diff --git "a/checkpoint-78030/trainer_state.json" "b/checkpoint-78030/trainer_state.json" new file mode 100644--- /dev/null +++ "b/checkpoint-78030/trainer_state.json" @@ -0,0 +1,468223 @@ +{ + "best_metric": 0.2753095030784607, + "best_model_checkpoint": "./medium-article-titles-engagement-all/checkpoint-78030", + "epoch": 3.0, + "global_step": 78030, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 1.0000000000000002e-06, + "loss": 0.6875, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 2.0000000000000003e-06, + "loss": 0.6914, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 3e-06, + "loss": 0.6886, + "step": 3 + }, + { + "epoch": 0.0, + "learning_rate": 4.000000000000001e-06, + "loss": 0.7051, + "step": 4 + }, + { + "epoch": 0.0, + "learning_rate": 5e-06, + "loss": 0.6932, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 6e-06, + "loss": 0.6967, + "step": 6 + }, + { + "epoch": 0.0, + "learning_rate": 7.000000000000001e-06, + "loss": 0.6975, + "step": 7 + }, + { + "epoch": 0.0, + "learning_rate": 8.000000000000001e-06, + "loss": 0.7076, + "step": 8 + }, + { + "epoch": 0.0, + "learning_rate": 9e-06, + "loss": 0.6945, + "step": 9 + }, + { + "epoch": 0.0, + "learning_rate": 1e-05, + "loss": 0.7013, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 1.1000000000000001e-05, + "loss": 0.6972, + "step": 11 + }, + { + "epoch": 0.0, + "learning_rate": 1.2e-05, + "loss": 0.6986, + "step": 12 + }, + { + "epoch": 0.0, + "learning_rate": 1.3000000000000001e-05, + "loss": 0.7057, + "step": 13 + }, + { + "epoch": 0.0, + "learning_rate": 1.4000000000000001e-05, + "loss": 0.6878, + "step": 14 + }, + { + "epoch": 0.0, + "learning_rate": 1.5e-05, + "loss": 0.6905, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.6911, + "step": 16 + }, + { + "epoch": 0.0, + "learning_rate": 1.7000000000000003e-05, + "loss": 0.689, + "step": 17 + }, + { + "epoch": 0.0, + "learning_rate": 1.8e-05, + "loss": 0.7046, + "step": 18 + }, + { + "epoch": 0.0, + "learning_rate": 1.9e-05, + "loss": 0.6864, + "step": 19 + }, + { + "epoch": 0.0, + "learning_rate": 2e-05, + "loss": 0.6986, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 2.1e-05, + "loss": 0.692, + "step": 21 + }, + { + "epoch": 0.0, + "learning_rate": 2.2000000000000003e-05, + "loss": 0.6996, + "step": 22 + }, + { + "epoch": 0.0, + "learning_rate": 2.3000000000000003e-05, + "loss": 0.7021, + "step": 23 + }, + { + "epoch": 0.0, + "learning_rate": 2.4e-05, + "loss": 0.6944, + "step": 24 + }, + { + "epoch": 0.0, + "learning_rate": 2.5e-05, + "loss": 0.6939, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 2.6000000000000002e-05, + "loss": 0.6931, + "step": 26 + }, + { + "epoch": 0.0, + "learning_rate": 2.7000000000000002e-05, + "loss": 0.6865, + "step": 27 + }, + { + "epoch": 0.0, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.6789, + "step": 28 + }, + { + "epoch": 0.0, + "learning_rate": 2.9e-05, + "loss": 0.6938, + "step": 29 + }, + { + "epoch": 0.0, + "learning_rate": 3e-05, + "loss": 0.707, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 3.1e-05, + "loss": 0.6978, + "step": 31 + }, + { + "epoch": 0.0, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.6878, + "step": 32 + }, + { + "epoch": 0.0, + "learning_rate": 3.3e-05, + "loss": 0.7003, + "step": 33 + }, + { + "epoch": 0.0, + "learning_rate": 3.4000000000000007e-05, + "loss": 0.6971, + "step": 34 + }, + { + "epoch": 0.0, + "learning_rate": 3.5e-05, + "loss": 0.6861, + "step": 35 + }, + { + "epoch": 0.0, + "learning_rate": 3.6e-05, + "loss": 0.695, + "step": 36 + }, + { + "epoch": 0.0, + "learning_rate": 3.7e-05, + "loss": 0.6992, + "step": 37 + }, + { + "epoch": 0.0, + "learning_rate": 3.8e-05, + "loss": 0.6953, + "step": 38 + }, + { + "epoch": 0.0, + "learning_rate": 3.9000000000000006e-05, + "loss": 0.703, + "step": 39 + }, + { + "epoch": 0.0, + "learning_rate": 4e-05, + "loss": 0.7045, + "step": 40 + }, + { + "epoch": 0.0, + "learning_rate": 4.1e-05, + "loss": 0.6891, + "step": 41 + }, + { + "epoch": 0.0, + "learning_rate": 4.2e-05, + "loss": 0.6778, + "step": 42 + }, + { + "epoch": 0.0, + "learning_rate": 4.3e-05, + "loss": 0.6909, + "step": 43 + }, + { + "epoch": 0.0, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.6907, + "step": 44 + }, + { + "epoch": 0.0, + "learning_rate": 4.5e-05, + "loss": 0.7057, + "step": 45 + }, + { + "epoch": 0.0, + "learning_rate": 4.600000000000001e-05, + "loss": 0.7291, + "step": 46 + }, + { + "epoch": 0.0, + "learning_rate": 4.7e-05, + "loss": 0.6999, + "step": 47 + }, + { + "epoch": 0.0, + "learning_rate": 4.8e-05, + "loss": 0.693, + "step": 48 + }, + { + "epoch": 0.0, + "learning_rate": 4.9e-05, + "loss": 0.6911, + "step": 49 + }, + { + "epoch": 0.0, + "learning_rate": 5e-05, + "loss": 0.6846, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 4.9999358809951274e-05, + "loss": 0.6862, + "step": 51 + }, + { + "epoch": 0.0, + "learning_rate": 4.999871761990254e-05, + "loss": 0.6899, + "step": 52 + }, + { + "epoch": 0.0, + "learning_rate": 4.999807642985381e-05, + "loss": 0.6806, + "step": 53 + }, + { + "epoch": 0.0, + "learning_rate": 4.999743523980508e-05, + "loss": 0.6772, + "step": 54 + }, + { + "epoch": 0.0, + "learning_rate": 4.9996794049756354e-05, + "loss": 0.6709, + "step": 55 + }, + { + "epoch": 0.0, + "learning_rate": 4.999615285970762e-05, + "loss": 0.7035, + "step": 56 + }, + { + "epoch": 0.0, + "learning_rate": 4.999551166965889e-05, + "loss": 0.7047, + "step": 57 + }, + { + "epoch": 0.0, + "learning_rate": 4.9994870479610156e-05, + "loss": 0.6776, + "step": 58 + }, + { + "epoch": 0.0, + "learning_rate": 4.999422928956143e-05, + "loss": 0.6891, + "step": 59 + }, + { + "epoch": 0.0, + "learning_rate": 4.99935880995127e-05, + "loss": 0.7329, + "step": 60 + }, + { + "epoch": 0.0, + "learning_rate": 4.9992946909463964e-05, + "loss": 0.6923, + "step": 61 + }, + { + "epoch": 0.0, + "learning_rate": 4.9992305719415236e-05, + "loss": 0.6773, + "step": 62 + }, + { + "epoch": 0.0, + "learning_rate": 4.999166452936651e-05, + "loss": 0.7002, + "step": 63 + }, + { + "epoch": 0.0, + "learning_rate": 4.999102333931778e-05, + "loss": 0.677, + "step": 64 + }, + { + "epoch": 0.0, + "learning_rate": 4.9990382149269044e-05, + "loss": 0.6776, + "step": 65 + }, + { + "epoch": 0.0, + "learning_rate": 4.9989740959220316e-05, + "loss": 0.6733, + "step": 66 + }, + { + "epoch": 0.0, + "learning_rate": 4.998909976917158e-05, + "loss": 0.6679, + "step": 67 + }, + { + "epoch": 0.0, + "learning_rate": 4.998845857912285e-05, + "loss": 0.6719, + "step": 68 + }, + { + "epoch": 0.0, + "learning_rate": 4.9987817389074124e-05, + "loss": 0.734, + "step": 69 + }, + { + "epoch": 0.0, + "learning_rate": 4.9987176199025396e-05, + "loss": 0.753, + "step": 70 + }, + { + "epoch": 0.0, + "learning_rate": 4.998653500897667e-05, + "loss": 0.6637, + "step": 71 + }, + { + "epoch": 0.0, + "learning_rate": 4.998589381892793e-05, + "loss": 0.6924, + "step": 72 + }, + { + "epoch": 0.0, + "learning_rate": 4.9985252628879204e-05, + "loss": 0.6542, + "step": 73 + }, + { + "epoch": 0.0, + "learning_rate": 4.998461143883047e-05, + "loss": 0.6428, + "step": 74 + }, + { + "epoch": 0.0, + "learning_rate": 4.998397024878174e-05, + "loss": 0.6298, + "step": 75 + }, + { + "epoch": 0.0, + "learning_rate": 4.9983329058733006e-05, + "loss": 0.7027, + "step": 76 + }, + { + "epoch": 0.0, + "learning_rate": 4.998268786868428e-05, + "loss": 0.7037, + "step": 77 + }, + { + "epoch": 0.0, + "learning_rate": 4.998204667863555e-05, + "loss": 0.6772, + "step": 78 + }, + { + "epoch": 0.0, + "learning_rate": 4.998140548858682e-05, + "loss": 0.6772, + "step": 79 + }, + { + "epoch": 0.0, + "learning_rate": 4.998076429853809e-05, + "loss": 0.7103, + "step": 80 + }, + { + "epoch": 0.0, + "learning_rate": 4.998012310848936e-05, + "loss": 0.7194, + "step": 81 + }, + { + "epoch": 0.0, + "learning_rate": 4.997948191844063e-05, + "loss": 0.6671, + "step": 82 + }, + { + "epoch": 0.0, + "learning_rate": 4.9978840728391895e-05, + "loss": 0.7094, + "step": 83 + }, + { + "epoch": 0.0, + "learning_rate": 4.9978199538343166e-05, + "loss": 0.6957, + "step": 84 + }, + { + "epoch": 0.0, + "learning_rate": 4.997755834829443e-05, + "loss": 0.701, + "step": 85 + }, + { + "epoch": 0.0, + "learning_rate": 4.997691715824571e-05, + "loss": 0.7063, + "step": 86 + }, + { + "epoch": 0.0, + "learning_rate": 4.997627596819698e-05, + "loss": 0.6921, + "step": 87 + }, + { + "epoch": 0.0, + "learning_rate": 4.9975634778148246e-05, + "loss": 0.6906, + "step": 88 + }, + { + "epoch": 0.0, + "learning_rate": 4.997499358809952e-05, + "loss": 0.6544, + "step": 89 + }, + { + "epoch": 0.0, + "learning_rate": 4.997435239805078e-05, + "loss": 0.6991, + "step": 90 + }, + { + "epoch": 0.0, + "learning_rate": 4.9973711208002055e-05, + "loss": 0.6843, + "step": 91 + }, + { + "epoch": 0.0, + "learning_rate": 4.997307001795332e-05, + "loss": 0.7162, + "step": 92 + }, + { + "epoch": 0.0, + "learning_rate": 4.997242882790459e-05, + "loss": 0.6903, + "step": 93 + }, + { + "epoch": 0.0, + "learning_rate": 4.997178763785586e-05, + "loss": 0.6632, + "step": 94 + }, + { + "epoch": 0.0, + "learning_rate": 4.9971146447807135e-05, + "loss": 0.6624, + "step": 95 + }, + { + "epoch": 0.0, + "learning_rate": 4.9970505257758406e-05, + "loss": 0.6999, + "step": 96 + }, + { + "epoch": 0.0, + "learning_rate": 4.996986406770967e-05, + "loss": 0.6853, + "step": 97 + }, + { + "epoch": 0.0, + "learning_rate": 4.996922287766094e-05, + "loss": 0.6804, + "step": 98 + }, + { + "epoch": 0.0, + "learning_rate": 4.996858168761221e-05, + "loss": 0.6837, + "step": 99 + }, + { + "epoch": 0.0, + "learning_rate": 4.996794049756348e-05, + "loss": 0.6734, + "step": 100 + }, + { + "epoch": 0.0, + "learning_rate": 4.9967299307514745e-05, + "loss": 0.6601, + "step": 101 + }, + { + "epoch": 0.0, + "learning_rate": 4.9966658117466016e-05, + "loss": 0.69, + "step": 102 + }, + { + "epoch": 0.0, + "learning_rate": 4.996601692741729e-05, + "loss": 0.6816, + "step": 103 + }, + { + "epoch": 0.0, + "learning_rate": 4.996537573736856e-05, + "loss": 0.6952, + "step": 104 + }, + { + "epoch": 0.0, + "learning_rate": 4.996473454731983e-05, + "loss": 0.6727, + "step": 105 + }, + { + "epoch": 0.0, + "learning_rate": 4.9964093357271097e-05, + "loss": 0.6923, + "step": 106 + }, + { + "epoch": 0.0, + "learning_rate": 4.996345216722237e-05, + "loss": 0.6931, + "step": 107 + }, + { + "epoch": 0.0, + "learning_rate": 4.996281097717363e-05, + "loss": 0.6269, + "step": 108 + }, + { + "epoch": 0.0, + "learning_rate": 4.9962169787124905e-05, + "loss": 0.6633, + "step": 109 + }, + { + "epoch": 0.0, + "learning_rate": 4.996152859707618e-05, + "loss": 0.6427, + "step": 110 + }, + { + "epoch": 0.0, + "learning_rate": 4.996088740702745e-05, + "loss": 0.7126, + "step": 111 + }, + { + "epoch": 0.0, + "learning_rate": 4.996024621697871e-05, + "loss": 0.6922, + "step": 112 + }, + { + "epoch": 0.0, + "learning_rate": 4.9959605026929985e-05, + "loss": 0.6797, + "step": 113 + }, + { + "epoch": 0.0, + "learning_rate": 4.995896383688126e-05, + "loss": 0.6848, + "step": 114 + }, + { + "epoch": 0.0, + "learning_rate": 4.995832264683252e-05, + "loss": 0.6809, + "step": 115 + }, + { + "epoch": 0.0, + "learning_rate": 4.9957681456783793e-05, + "loss": 0.6594, + "step": 116 + }, + { + "epoch": 0.0, + "learning_rate": 4.995704026673506e-05, + "loss": 0.669, + "step": 117 + }, + { + "epoch": 0.0, + "learning_rate": 4.995639907668633e-05, + "loss": 0.6672, + "step": 118 + }, + { + "epoch": 0.0, + "learning_rate": 4.99557578866376e-05, + "loss": 0.6758, + "step": 119 + }, + { + "epoch": 0.0, + "learning_rate": 4.9955116696588873e-05, + "loss": 0.6773, + "step": 120 + }, + { + "epoch": 0.0, + "learning_rate": 4.995447550654014e-05, + "loss": 0.6838, + "step": 121 + }, + { + "epoch": 0.0, + "learning_rate": 4.995383431649141e-05, + "loss": 0.7066, + "step": 122 + }, + { + "epoch": 0.0, + "learning_rate": 4.995319312644268e-05, + "loss": 0.6666, + "step": 123 + }, + { + "epoch": 0.0, + "learning_rate": 4.995255193639395e-05, + "loss": 0.6994, + "step": 124 + }, + { + "epoch": 0.0, + "learning_rate": 4.995191074634522e-05, + "loss": 0.6532, + "step": 125 + }, + { + "epoch": 0.0, + "learning_rate": 4.995126955629649e-05, + "loss": 0.63, + "step": 126 + }, + { + "epoch": 0.0, + "learning_rate": 4.995062836624776e-05, + "loss": 0.655, + "step": 127 + }, + { + "epoch": 0.0, + "learning_rate": 4.994998717619903e-05, + "loss": 0.7004, + "step": 128 + }, + { + "epoch": 0.0, + "learning_rate": 4.99493459861503e-05, + "loss": 0.7173, + "step": 129 + }, + { + "epoch": 0.0, + "learning_rate": 4.9948704796101564e-05, + "loss": 0.6564, + "step": 130 + }, + { + "epoch": 0.01, + "learning_rate": 4.9948063606052835e-05, + "loss": 0.6169, + "step": 131 + }, + { + "epoch": 0.01, + "learning_rate": 4.994742241600411e-05, + "loss": 0.6415, + "step": 132 + }, + { + "epoch": 0.01, + "learning_rate": 4.994678122595537e-05, + "loss": 0.669, + "step": 133 + }, + { + "epoch": 0.01, + "learning_rate": 4.9946140035906644e-05, + "loss": 0.696, + "step": 134 + }, + { + "epoch": 0.01, + "learning_rate": 4.9945498845857915e-05, + "loss": 0.7049, + "step": 135 + }, + { + "epoch": 0.01, + "learning_rate": 4.994485765580919e-05, + "loss": 0.7374, + "step": 136 + }, + { + "epoch": 0.01, + "learning_rate": 4.994421646576045e-05, + "loss": 0.7051, + "step": 137 + }, + { + "epoch": 0.01, + "learning_rate": 4.9943575275711724e-05, + "loss": 0.6312, + "step": 138 + }, + { + "epoch": 0.01, + "learning_rate": 4.994293408566299e-05, + "loss": 0.6345, + "step": 139 + }, + { + "epoch": 0.01, + "learning_rate": 4.994229289561426e-05, + "loss": 0.6694, + "step": 140 + }, + { + "epoch": 0.01, + "learning_rate": 4.994165170556553e-05, + "loss": 0.6714, + "step": 141 + }, + { + "epoch": 0.01, + "learning_rate": 4.9941010515516804e-05, + "loss": 0.6624, + "step": 142 + }, + { + "epoch": 0.01, + "learning_rate": 4.9940369325468076e-05, + "loss": 0.7267, + "step": 143 + }, + { + "epoch": 0.01, + "learning_rate": 4.993972813541934e-05, + "loss": 0.6634, + "step": 144 + }, + { + "epoch": 0.01, + "learning_rate": 4.993908694537061e-05, + "loss": 0.6911, + "step": 145 + }, + { + "epoch": 0.01, + "learning_rate": 4.993844575532188e-05, + "loss": 0.693, + "step": 146 + }, + { + "epoch": 0.01, + "learning_rate": 4.993780456527315e-05, + "loss": 0.6908, + "step": 147 + }, + { + "epoch": 0.01, + "learning_rate": 4.993716337522442e-05, + "loss": 0.6699, + "step": 148 + }, + { + "epoch": 0.01, + "learning_rate": 4.9936522185175686e-05, + "loss": 0.7021, + "step": 149 + }, + { + "epoch": 0.01, + "learning_rate": 4.993588099512696e-05, + "loss": 0.6818, + "step": 150 + }, + { + "epoch": 0.01, + "learning_rate": 4.993523980507823e-05, + "loss": 0.6696, + "step": 151 + }, + { + "epoch": 0.01, + "learning_rate": 4.99345986150295e-05, + "loss": 0.6535, + "step": 152 + }, + { + "epoch": 0.01, + "learning_rate": 4.9933957424980766e-05, + "loss": 0.6887, + "step": 153 + }, + { + "epoch": 0.01, + "learning_rate": 4.993331623493204e-05, + "loss": 0.7054, + "step": 154 + }, + { + "epoch": 0.01, + "learning_rate": 4.99326750448833e-05, + "loss": 0.6678, + "step": 155 + }, + { + "epoch": 0.01, + "learning_rate": 4.9932033854834574e-05, + "loss": 0.658, + "step": 156 + }, + { + "epoch": 0.01, + "learning_rate": 4.9931392664785846e-05, + "loss": 0.6759, + "step": 157 + }, + { + "epoch": 0.01, + "learning_rate": 4.993075147473712e-05, + "loss": 0.664, + "step": 158 + }, + { + "epoch": 0.01, + "learning_rate": 4.993011028468839e-05, + "loss": 0.6591, + "step": 159 + }, + { + "epoch": 0.01, + "learning_rate": 4.9929469094639654e-05, + "loss": 0.6613, + "step": 160 + }, + { + "epoch": 0.01, + "learning_rate": 4.9928827904590926e-05, + "loss": 0.6653, + "step": 161 + }, + { + "epoch": 0.01, + "learning_rate": 4.992818671454219e-05, + "loss": 0.7277, + "step": 162 + }, + { + "epoch": 0.01, + "learning_rate": 4.992754552449346e-05, + "loss": 0.6702, + "step": 163 + }, + { + "epoch": 0.01, + "learning_rate": 4.992690433444473e-05, + "loss": 0.7418, + "step": 164 + }, + { + "epoch": 0.01, + "learning_rate": 4.9926263144396e-05, + "loss": 0.6932, + "step": 165 + }, + { + "epoch": 0.01, + "learning_rate": 4.992562195434727e-05, + "loss": 0.6749, + "step": 166 + }, + { + "epoch": 0.01, + "learning_rate": 4.992498076429854e-05, + "loss": 0.7208, + "step": 167 + }, + { + "epoch": 0.01, + "learning_rate": 4.9924339574249814e-05, + "loss": 0.6776, + "step": 168 + }, + { + "epoch": 0.01, + "learning_rate": 4.992369838420108e-05, + "loss": 0.6869, + "step": 169 + }, + { + "epoch": 0.01, + "learning_rate": 4.992305719415235e-05, + "loss": 0.6796, + "step": 170 + }, + { + "epoch": 0.01, + "learning_rate": 4.9922416004103616e-05, + "loss": 0.6983, + "step": 171 + }, + { + "epoch": 0.01, + "learning_rate": 4.992177481405489e-05, + "loss": 0.6474, + "step": 172 + }, + { + "epoch": 0.01, + "learning_rate": 4.992113362400615e-05, + "loss": 0.6994, + "step": 173 + }, + { + "epoch": 0.01, + "learning_rate": 4.9920492433957424e-05, + "loss": 0.6765, + "step": 174 + }, + { + "epoch": 0.01, + "learning_rate": 4.99198512439087e-05, + "loss": 0.6874, + "step": 175 + }, + { + "epoch": 0.01, + "learning_rate": 4.991921005385997e-05, + "loss": 0.6622, + "step": 176 + }, + { + "epoch": 0.01, + "learning_rate": 4.991856886381124e-05, + "loss": 0.691, + "step": 177 + }, + { + "epoch": 0.01, + "learning_rate": 4.9917927673762504e-05, + "loss": 0.6644, + "step": 178 + }, + { + "epoch": 0.01, + "learning_rate": 4.9917286483713776e-05, + "loss": 0.698, + "step": 179 + }, + { + "epoch": 0.01, + "learning_rate": 4.991664529366504e-05, + "loss": 0.7033, + "step": 180 + }, + { + "epoch": 0.01, + "learning_rate": 4.991600410361631e-05, + "loss": 0.6996, + "step": 181 + }, + { + "epoch": 0.01, + "learning_rate": 4.9915362913567584e-05, + "loss": 0.6781, + "step": 182 + }, + { + "epoch": 0.01, + "learning_rate": 4.9914721723518856e-05, + "loss": 0.6521, + "step": 183 + }, + { + "epoch": 0.01, + "learning_rate": 4.991408053347013e-05, + "loss": 0.6644, + "step": 184 + }, + { + "epoch": 0.01, + "learning_rate": 4.991343934342139e-05, + "loss": 0.6924, + "step": 185 + }, + { + "epoch": 0.01, + "learning_rate": 4.9912798153372664e-05, + "loss": 0.6455, + "step": 186 + }, + { + "epoch": 0.01, + "learning_rate": 4.991215696332393e-05, + "loss": 0.6645, + "step": 187 + }, + { + "epoch": 0.01, + "learning_rate": 4.99115157732752e-05, + "loss": 0.7119, + "step": 188 + }, + { + "epoch": 0.01, + "learning_rate": 4.9910874583226466e-05, + "loss": 0.657, + "step": 189 + }, + { + "epoch": 0.01, + "learning_rate": 4.991023339317774e-05, + "loss": 0.6966, + "step": 190 + }, + { + "epoch": 0.01, + "learning_rate": 4.990959220312901e-05, + "loss": 0.6488, + "step": 191 + }, + { + "epoch": 0.01, + "learning_rate": 4.990895101308028e-05, + "loss": 0.6564, + "step": 192 + }, + { + "epoch": 0.01, + "learning_rate": 4.990830982303155e-05, + "loss": 0.6699, + "step": 193 + }, + { + "epoch": 0.01, + "learning_rate": 4.990766863298282e-05, + "loss": 0.6909, + "step": 194 + }, + { + "epoch": 0.01, + "learning_rate": 4.990702744293409e-05, + "loss": 0.6717, + "step": 195 + }, + { + "epoch": 0.01, + "learning_rate": 4.9906386252885355e-05, + "loss": 0.6881, + "step": 196 + }, + { + "epoch": 0.01, + "learning_rate": 4.9905745062836626e-05, + "loss": 0.6703, + "step": 197 + }, + { + "epoch": 0.01, + "learning_rate": 4.99051038727879e-05, + "loss": 0.6634, + "step": 198 + }, + { + "epoch": 0.01, + "learning_rate": 4.990446268273917e-05, + "loss": 0.6661, + "step": 199 + }, + { + "epoch": 0.01, + "learning_rate": 4.9903821492690435e-05, + "loss": 0.6691, + "step": 200 + }, + { + "epoch": 0.01, + "learning_rate": 4.9903180302641706e-05, + "loss": 0.6542, + "step": 201 + }, + { + "epoch": 0.01, + "learning_rate": 4.990253911259298e-05, + "loss": 0.6716, + "step": 202 + }, + { + "epoch": 0.01, + "learning_rate": 4.990189792254424e-05, + "loss": 0.6792, + "step": 203 + }, + { + "epoch": 0.01, + "learning_rate": 4.9901256732495515e-05, + "loss": 0.6455, + "step": 204 + }, + { + "epoch": 0.01, + "learning_rate": 4.990061554244678e-05, + "loss": 0.6509, + "step": 205 + }, + { + "epoch": 0.01, + "learning_rate": 4.989997435239805e-05, + "loss": 0.6808, + "step": 206 + }, + { + "epoch": 0.01, + "learning_rate": 4.989933316234932e-05, + "loss": 0.6797, + "step": 207 + }, + { + "epoch": 0.01, + "learning_rate": 4.9898691972300595e-05, + "loss": 0.6784, + "step": 208 + }, + { + "epoch": 0.01, + "learning_rate": 4.989805078225186e-05, + "loss": 0.6672, + "step": 209 + }, + { + "epoch": 0.01, + "learning_rate": 4.989740959220313e-05, + "loss": 0.7168, + "step": 210 + }, + { + "epoch": 0.01, + "learning_rate": 4.98967684021544e-05, + "loss": 0.6476, + "step": 211 + }, + { + "epoch": 0.01, + "learning_rate": 4.989612721210567e-05, + "loss": 0.6922, + "step": 212 + }, + { + "epoch": 0.01, + "learning_rate": 4.989548602205694e-05, + "loss": 0.6435, + "step": 213 + }, + { + "epoch": 0.01, + "learning_rate": 4.989484483200821e-05, + "loss": 0.6573, + "step": 214 + }, + { + "epoch": 0.01, + "learning_rate": 4.989420364195948e-05, + "loss": 0.6504, + "step": 215 + }, + { + "epoch": 0.01, + "learning_rate": 4.989356245191075e-05, + "loss": 0.6728, + "step": 216 + }, + { + "epoch": 0.01, + "learning_rate": 4.989292126186202e-05, + "loss": 0.6659, + "step": 217 + }, + { + "epoch": 0.01, + "learning_rate": 4.9892280071813285e-05, + "loss": 0.6871, + "step": 218 + }, + { + "epoch": 0.01, + "learning_rate": 4.9891638881764557e-05, + "loss": 0.6709, + "step": 219 + }, + { + "epoch": 0.01, + "learning_rate": 4.989099769171583e-05, + "loss": 0.6336, + "step": 220 + }, + { + "epoch": 0.01, + "learning_rate": 4.989035650166709e-05, + "loss": 0.68, + "step": 221 + }, + { + "epoch": 0.01, + "learning_rate": 4.9889715311618365e-05, + "loss": 0.6536, + "step": 222 + }, + { + "epoch": 0.01, + "learning_rate": 4.988907412156964e-05, + "loss": 0.6244, + "step": 223 + }, + { + "epoch": 0.01, + "learning_rate": 4.988843293152091e-05, + "loss": 0.6552, + "step": 224 + }, + { + "epoch": 0.01, + "learning_rate": 4.988779174147217e-05, + "loss": 0.6667, + "step": 225 + }, + { + "epoch": 0.01, + "learning_rate": 4.9887150551423445e-05, + "loss": 0.6739, + "step": 226 + }, + { + "epoch": 0.01, + "learning_rate": 4.988650936137471e-05, + "loss": 0.6252, + "step": 227 + }, + { + "epoch": 0.01, + "learning_rate": 4.988586817132598e-05, + "loss": 0.7028, + "step": 228 + }, + { + "epoch": 0.01, + "learning_rate": 4.9885226981277253e-05, + "loss": 0.6613, + "step": 229 + }, + { + "epoch": 0.01, + "learning_rate": 4.9884585791228525e-05, + "loss": 0.6663, + "step": 230 + }, + { + "epoch": 0.01, + "learning_rate": 4.98839446011798e-05, + "loss": 0.6131, + "step": 231 + }, + { + "epoch": 0.01, + "learning_rate": 4.988330341113106e-05, + "loss": 0.7338, + "step": 232 + }, + { + "epoch": 0.01, + "learning_rate": 4.9882662221082334e-05, + "loss": 0.6186, + "step": 233 + }, + { + "epoch": 0.01, + "learning_rate": 4.98820210310336e-05, + "loss": 0.6284, + "step": 234 + }, + { + "epoch": 0.01, + "learning_rate": 4.988137984098487e-05, + "loss": 0.6667, + "step": 235 + }, + { + "epoch": 0.01, + "learning_rate": 4.9880738650936135e-05, + "loss": 0.7164, + "step": 236 + }, + { + "epoch": 0.01, + "learning_rate": 4.988009746088741e-05, + "loss": 0.6672, + "step": 237 + }, + { + "epoch": 0.01, + "learning_rate": 4.987945627083868e-05, + "loss": 0.6465, + "step": 238 + }, + { + "epoch": 0.01, + "learning_rate": 4.987881508078995e-05, + "loss": 0.6449, + "step": 239 + }, + { + "epoch": 0.01, + "learning_rate": 4.987817389074122e-05, + "loss": 0.6182, + "step": 240 + }, + { + "epoch": 0.01, + "learning_rate": 4.987753270069249e-05, + "loss": 0.6723, + "step": 241 + }, + { + "epoch": 0.01, + "learning_rate": 4.987689151064376e-05, + "loss": 0.646, + "step": 242 + }, + { + "epoch": 0.01, + "learning_rate": 4.9876250320595024e-05, + "loss": 0.6653, + "step": 243 + }, + { + "epoch": 0.01, + "learning_rate": 4.9875609130546295e-05, + "loss": 0.6917, + "step": 244 + }, + { + "epoch": 0.01, + "learning_rate": 4.987496794049756e-05, + "loss": 0.6323, + "step": 245 + }, + { + "epoch": 0.01, + "learning_rate": 4.987432675044884e-05, + "loss": 0.7016, + "step": 246 + }, + { + "epoch": 0.01, + "learning_rate": 4.987368556040011e-05, + "loss": 0.6315, + "step": 247 + }, + { + "epoch": 0.01, + "learning_rate": 4.9873044370351375e-05, + "loss": 0.6481, + "step": 248 + }, + { + "epoch": 0.01, + "learning_rate": 4.987240318030265e-05, + "loss": 0.6611, + "step": 249 + }, + { + "epoch": 0.01, + "learning_rate": 4.987176199025391e-05, + "loss": 0.6377, + "step": 250 + }, + { + "epoch": 0.01, + "learning_rate": 4.9871120800205184e-05, + "loss": 0.6525, + "step": 251 + }, + { + "epoch": 0.01, + "learning_rate": 4.987047961015645e-05, + "loss": 0.6918, + "step": 252 + }, + { + "epoch": 0.01, + "learning_rate": 4.986983842010772e-05, + "loss": 0.6251, + "step": 253 + }, + { + "epoch": 0.01, + "learning_rate": 4.986919723005899e-05, + "loss": 0.6595, + "step": 254 + }, + { + "epoch": 0.01, + "learning_rate": 4.9868556040010264e-05, + "loss": 0.6906, + "step": 255 + }, + { + "epoch": 0.01, + "learning_rate": 4.9867914849961536e-05, + "loss": 0.6699, + "step": 256 + }, + { + "epoch": 0.01, + "learning_rate": 4.98672736599128e-05, + "loss": 0.6785, + "step": 257 + }, + { + "epoch": 0.01, + "learning_rate": 4.986663246986407e-05, + "loss": 0.6728, + "step": 258 + }, + { + "epoch": 0.01, + "learning_rate": 4.986599127981534e-05, + "loss": 0.6465, + "step": 259 + }, + { + "epoch": 0.01, + "learning_rate": 4.986535008976661e-05, + "loss": 0.6385, + "step": 260 + }, + { + "epoch": 0.01, + "learning_rate": 4.9864708899717874e-05, + "loss": 0.6783, + "step": 261 + }, + { + "epoch": 0.01, + "learning_rate": 4.9864067709669146e-05, + "loss": 0.6517, + "step": 262 + }, + { + "epoch": 0.01, + "learning_rate": 4.986342651962042e-05, + "loss": 0.6056, + "step": 263 + }, + { + "epoch": 0.01, + "learning_rate": 4.986278532957169e-05, + "loss": 0.6538, + "step": 264 + }, + { + "epoch": 0.01, + "learning_rate": 4.986214413952296e-05, + "loss": 0.6287, + "step": 265 + }, + { + "epoch": 0.01, + "learning_rate": 4.9861502949474226e-05, + "loss": 0.6738, + "step": 266 + }, + { + "epoch": 0.01, + "learning_rate": 4.98608617594255e-05, + "loss": 0.6437, + "step": 267 + }, + { + "epoch": 0.01, + "learning_rate": 4.986022056937676e-05, + "loss": 0.6096, + "step": 268 + }, + { + "epoch": 0.01, + "learning_rate": 4.9859579379328034e-05, + "loss": 0.6127, + "step": 269 + }, + { + "epoch": 0.01, + "learning_rate": 4.9858938189279306e-05, + "loss": 0.6439, + "step": 270 + }, + { + "epoch": 0.01, + "learning_rate": 4.985829699923058e-05, + "loss": 0.693, + "step": 271 + }, + { + "epoch": 0.01, + "learning_rate": 4.985765580918184e-05, + "loss": 0.6826, + "step": 272 + }, + { + "epoch": 0.01, + "learning_rate": 4.9857014619133114e-05, + "loss": 0.6542, + "step": 273 + }, + { + "epoch": 0.01, + "learning_rate": 4.9856373429084386e-05, + "loss": 0.6542, + "step": 274 + }, + { + "epoch": 0.01, + "learning_rate": 4.985573223903565e-05, + "loss": 0.7074, + "step": 275 + }, + { + "epoch": 0.01, + "learning_rate": 4.985509104898692e-05, + "loss": 0.6379, + "step": 276 + }, + { + "epoch": 0.01, + "learning_rate": 4.985444985893819e-05, + "loss": 0.6805, + "step": 277 + }, + { + "epoch": 0.01, + "learning_rate": 4.985380866888946e-05, + "loss": 0.6123, + "step": 278 + }, + { + "epoch": 0.01, + "learning_rate": 4.985316747884073e-05, + "loss": 0.6137, + "step": 279 + }, + { + "epoch": 0.01, + "learning_rate": 4.9852526288792e-05, + "loss": 0.6421, + "step": 280 + }, + { + "epoch": 0.01, + "learning_rate": 4.985188509874327e-05, + "loss": 0.635, + "step": 281 + }, + { + "epoch": 0.01, + "learning_rate": 4.985124390869454e-05, + "loss": 0.6754, + "step": 282 + }, + { + "epoch": 0.01, + "learning_rate": 4.985060271864581e-05, + "loss": 0.6953, + "step": 283 + }, + { + "epoch": 0.01, + "learning_rate": 4.9849961528597076e-05, + "loss": 0.6877, + "step": 284 + }, + { + "epoch": 0.01, + "learning_rate": 4.984932033854835e-05, + "loss": 0.6194, + "step": 285 + }, + { + "epoch": 0.01, + "learning_rate": 4.984867914849962e-05, + "loss": 0.67, + "step": 286 + }, + { + "epoch": 0.01, + "learning_rate": 4.984803795845089e-05, + "loss": 0.6293, + "step": 287 + }, + { + "epoch": 0.01, + "learning_rate": 4.9847396768402156e-05, + "loss": 0.6375, + "step": 288 + }, + { + "epoch": 0.01, + "learning_rate": 4.984675557835343e-05, + "loss": 0.694, + "step": 289 + }, + { + "epoch": 0.01, + "learning_rate": 4.984611438830469e-05, + "loss": 0.6346, + "step": 290 + }, + { + "epoch": 0.01, + "learning_rate": 4.9845473198255964e-05, + "loss": 0.7256, + "step": 291 + }, + { + "epoch": 0.01, + "learning_rate": 4.9844832008207236e-05, + "loss": 0.6914, + "step": 292 + }, + { + "epoch": 0.01, + "learning_rate": 4.98441908181585e-05, + "loss": 0.6517, + "step": 293 + }, + { + "epoch": 0.01, + "learning_rate": 4.984354962810977e-05, + "loss": 0.6647, + "step": 294 + }, + { + "epoch": 0.01, + "learning_rate": 4.9842908438061044e-05, + "loss": 0.7181, + "step": 295 + }, + { + "epoch": 0.01, + "learning_rate": 4.9842267248012316e-05, + "loss": 0.6915, + "step": 296 + }, + { + "epoch": 0.01, + "learning_rate": 4.984162605796358e-05, + "loss": 0.6824, + "step": 297 + }, + { + "epoch": 0.01, + "learning_rate": 4.984098486791485e-05, + "loss": 0.6154, + "step": 298 + }, + { + "epoch": 0.01, + "learning_rate": 4.984034367786612e-05, + "loss": 0.6559, + "step": 299 + }, + { + "epoch": 0.01, + "learning_rate": 4.983970248781739e-05, + "loss": 0.702, + "step": 300 + }, + { + "epoch": 0.01, + "learning_rate": 4.983906129776866e-05, + "loss": 0.6856, + "step": 301 + }, + { + "epoch": 0.01, + "learning_rate": 4.983842010771993e-05, + "loss": 0.683, + "step": 302 + }, + { + "epoch": 0.01, + "learning_rate": 4.9837778917671205e-05, + "loss": 0.6675, + "step": 303 + }, + { + "epoch": 0.01, + "learning_rate": 4.983713772762247e-05, + "loss": 0.6311, + "step": 304 + }, + { + "epoch": 0.01, + "learning_rate": 4.983649653757374e-05, + "loss": 0.636, + "step": 305 + }, + { + "epoch": 0.01, + "learning_rate": 4.9835855347525006e-05, + "loss": 0.6535, + "step": 306 + }, + { + "epoch": 0.01, + "learning_rate": 4.983521415747628e-05, + "loss": 0.6404, + "step": 307 + }, + { + "epoch": 0.01, + "learning_rate": 4.983457296742754e-05, + "loss": 0.6698, + "step": 308 + }, + { + "epoch": 0.01, + "learning_rate": 4.9833931777378815e-05, + "loss": 0.6385, + "step": 309 + }, + { + "epoch": 0.01, + "learning_rate": 4.9833290587330086e-05, + "loss": 0.6764, + "step": 310 + }, + { + "epoch": 0.01, + "learning_rate": 4.983264939728136e-05, + "loss": 0.66, + "step": 311 + }, + { + "epoch": 0.01, + "learning_rate": 4.983200820723263e-05, + "loss": 0.7051, + "step": 312 + }, + { + "epoch": 0.01, + "learning_rate": 4.9831367017183895e-05, + "loss": 0.6494, + "step": 313 + }, + { + "epoch": 0.01, + "learning_rate": 4.9830725827135166e-05, + "loss": 0.6507, + "step": 314 + }, + { + "epoch": 0.01, + "learning_rate": 4.983008463708643e-05, + "loss": 0.6481, + "step": 315 + }, + { + "epoch": 0.01, + "learning_rate": 4.98294434470377e-05, + "loss": 0.708, + "step": 316 + }, + { + "epoch": 0.01, + "learning_rate": 4.982880225698897e-05, + "loss": 0.6506, + "step": 317 + }, + { + "epoch": 0.01, + "learning_rate": 4.9828161066940246e-05, + "loss": 0.6046, + "step": 318 + }, + { + "epoch": 0.01, + "learning_rate": 4.982751987689152e-05, + "loss": 0.619, + "step": 319 + }, + { + "epoch": 0.01, + "learning_rate": 4.982687868684278e-05, + "loss": 0.6548, + "step": 320 + }, + { + "epoch": 0.01, + "learning_rate": 4.9826237496794055e-05, + "loss": 0.6492, + "step": 321 + }, + { + "epoch": 0.01, + "learning_rate": 4.982559630674532e-05, + "loss": 0.5974, + "step": 322 + }, + { + "epoch": 0.01, + "learning_rate": 4.982495511669659e-05, + "loss": 0.6586, + "step": 323 + }, + { + "epoch": 0.01, + "learning_rate": 4.9824313926647856e-05, + "loss": 0.6798, + "step": 324 + }, + { + "epoch": 0.01, + "learning_rate": 4.982367273659913e-05, + "loss": 0.6038, + "step": 325 + }, + { + "epoch": 0.01, + "learning_rate": 4.98230315465504e-05, + "loss": 0.6457, + "step": 326 + }, + { + "epoch": 0.01, + "learning_rate": 4.982239035650167e-05, + "loss": 0.5946, + "step": 327 + }, + { + "epoch": 0.01, + "learning_rate": 4.982174916645294e-05, + "loss": 0.7132, + "step": 328 + }, + { + "epoch": 0.01, + "learning_rate": 4.982110797640421e-05, + "loss": 0.6233, + "step": 329 + }, + { + "epoch": 0.01, + "learning_rate": 4.982046678635548e-05, + "loss": 0.5911, + "step": 330 + }, + { + "epoch": 0.01, + "learning_rate": 4.9819825596306745e-05, + "loss": 0.5719, + "step": 331 + }, + { + "epoch": 0.01, + "learning_rate": 4.981918440625802e-05, + "loss": 0.6379, + "step": 332 + }, + { + "epoch": 0.01, + "learning_rate": 4.981854321620928e-05, + "loss": 0.6674, + "step": 333 + }, + { + "epoch": 0.01, + "learning_rate": 4.981790202616056e-05, + "loss": 0.6875, + "step": 334 + }, + { + "epoch": 0.01, + "learning_rate": 4.981726083611183e-05, + "loss": 0.5306, + "step": 335 + }, + { + "epoch": 0.01, + "learning_rate": 4.98166196460631e-05, + "loss": 0.6616, + "step": 336 + }, + { + "epoch": 0.01, + "learning_rate": 4.981597845601437e-05, + "loss": 0.6349, + "step": 337 + }, + { + "epoch": 0.01, + "learning_rate": 4.981533726596563e-05, + "loss": 0.7384, + "step": 338 + }, + { + "epoch": 0.01, + "learning_rate": 4.9814696075916905e-05, + "loss": 0.7376, + "step": 339 + }, + { + "epoch": 0.01, + "learning_rate": 4.981405488586817e-05, + "loss": 0.6593, + "step": 340 + }, + { + "epoch": 0.01, + "learning_rate": 4.981341369581944e-05, + "loss": 0.6127, + "step": 341 + }, + { + "epoch": 0.01, + "learning_rate": 4.9812772505770713e-05, + "loss": 0.5999, + "step": 342 + }, + { + "epoch": 0.01, + "learning_rate": 4.9812131315721985e-05, + "loss": 0.6769, + "step": 343 + }, + { + "epoch": 0.01, + "learning_rate": 4.981149012567326e-05, + "loss": 0.6062, + "step": 344 + }, + { + "epoch": 0.01, + "learning_rate": 4.981084893562452e-05, + "loss": 0.6903, + "step": 345 + }, + { + "epoch": 0.01, + "learning_rate": 4.9810207745575794e-05, + "loss": 0.7044, + "step": 346 + }, + { + "epoch": 0.01, + "learning_rate": 4.980956655552706e-05, + "loss": 0.667, + "step": 347 + }, + { + "epoch": 0.01, + "learning_rate": 4.980892536547833e-05, + "loss": 0.6666, + "step": 348 + }, + { + "epoch": 0.01, + "learning_rate": 4.9808284175429595e-05, + "loss": 0.6397, + "step": 349 + }, + { + "epoch": 0.01, + "learning_rate": 4.980764298538087e-05, + "loss": 0.6529, + "step": 350 + }, + { + "epoch": 0.01, + "learning_rate": 4.980700179533214e-05, + "loss": 0.6008, + "step": 351 + }, + { + "epoch": 0.01, + "learning_rate": 4.980636060528341e-05, + "loss": 0.6296, + "step": 352 + }, + { + "epoch": 0.01, + "learning_rate": 4.980571941523468e-05, + "loss": 0.6476, + "step": 353 + }, + { + "epoch": 0.01, + "learning_rate": 4.980507822518595e-05, + "loss": 0.6591, + "step": 354 + }, + { + "epoch": 0.01, + "learning_rate": 4.980443703513722e-05, + "loss": 0.666, + "step": 355 + }, + { + "epoch": 0.01, + "learning_rate": 4.9803795845088484e-05, + "loss": 0.6782, + "step": 356 + }, + { + "epoch": 0.01, + "learning_rate": 4.9803154655039755e-05, + "loss": 0.5872, + "step": 357 + }, + { + "epoch": 0.01, + "learning_rate": 4.980251346499103e-05, + "loss": 0.71, + "step": 358 + }, + { + "epoch": 0.01, + "learning_rate": 4.98018722749423e-05, + "loss": 0.6091, + "step": 359 + }, + { + "epoch": 0.01, + "learning_rate": 4.9801231084893564e-05, + "loss": 0.7609, + "step": 360 + }, + { + "epoch": 0.01, + "learning_rate": 4.9800589894844835e-05, + "loss": 0.6844, + "step": 361 + }, + { + "epoch": 0.01, + "learning_rate": 4.979994870479611e-05, + "loss": 0.7514, + "step": 362 + }, + { + "epoch": 0.01, + "learning_rate": 4.979930751474737e-05, + "loss": 0.7359, + "step": 363 + }, + { + "epoch": 0.01, + "learning_rate": 4.9798666324698644e-05, + "loss": 0.6408, + "step": 364 + }, + { + "epoch": 0.01, + "learning_rate": 4.979802513464991e-05, + "loss": 0.6557, + "step": 365 + }, + { + "epoch": 0.01, + "learning_rate": 4.979738394460118e-05, + "loss": 0.6067, + "step": 366 + }, + { + "epoch": 0.01, + "learning_rate": 4.979674275455245e-05, + "loss": 0.6569, + "step": 367 + }, + { + "epoch": 0.01, + "learning_rate": 4.9796101564503724e-05, + "loss": 0.6196, + "step": 368 + }, + { + "epoch": 0.01, + "learning_rate": 4.979546037445499e-05, + "loss": 0.6015, + "step": 369 + }, + { + "epoch": 0.01, + "learning_rate": 4.979481918440626e-05, + "loss": 0.6689, + "step": 370 + }, + { + "epoch": 0.01, + "learning_rate": 4.979417799435753e-05, + "loss": 0.6738, + "step": 371 + }, + { + "epoch": 0.01, + "learning_rate": 4.97935368043088e-05, + "loss": 0.5636, + "step": 372 + }, + { + "epoch": 0.01, + "learning_rate": 4.979289561426007e-05, + "loss": 0.6747, + "step": 373 + }, + { + "epoch": 0.01, + "learning_rate": 4.979225442421134e-05, + "loss": 0.5829, + "step": 374 + }, + { + "epoch": 0.01, + "learning_rate": 4.979161323416261e-05, + "loss": 0.6046, + "step": 375 + }, + { + "epoch": 0.01, + "learning_rate": 4.979097204411388e-05, + "loss": 0.7024, + "step": 376 + }, + { + "epoch": 0.01, + "learning_rate": 4.979033085406515e-05, + "loss": 0.6996, + "step": 377 + }, + { + "epoch": 0.01, + "learning_rate": 4.9789689664016414e-05, + "loss": 0.6969, + "step": 378 + }, + { + "epoch": 0.01, + "learning_rate": 4.9789048473967686e-05, + "loss": 0.6852, + "step": 379 + }, + { + "epoch": 0.01, + "learning_rate": 4.978840728391896e-05, + "loss": 0.7081, + "step": 380 + }, + { + "epoch": 0.01, + "learning_rate": 4.978776609387022e-05, + "loss": 0.5689, + "step": 381 + }, + { + "epoch": 0.01, + "learning_rate": 4.9787124903821494e-05, + "loss": 0.634, + "step": 382 + }, + { + "epoch": 0.01, + "learning_rate": 4.9786483713772766e-05, + "loss": 0.6174, + "step": 383 + }, + { + "epoch": 0.01, + "learning_rate": 4.978584252372404e-05, + "loss": 0.7074, + "step": 384 + }, + { + "epoch": 0.01, + "learning_rate": 4.97852013336753e-05, + "loss": 0.6342, + "step": 385 + }, + { + "epoch": 0.01, + "learning_rate": 4.9784560143626574e-05, + "loss": 0.6801, + "step": 386 + }, + { + "epoch": 0.01, + "learning_rate": 4.978391895357784e-05, + "loss": 0.5336, + "step": 387 + }, + { + "epoch": 0.01, + "learning_rate": 4.978327776352911e-05, + "loss": 0.6806, + "step": 388 + }, + { + "epoch": 0.01, + "learning_rate": 4.978263657348038e-05, + "loss": 0.7306, + "step": 389 + }, + { + "epoch": 0.01, + "learning_rate": 4.9781995383431654e-05, + "loss": 0.6278, + "step": 390 + }, + { + "epoch": 0.02, + "learning_rate": 4.9781354193382926e-05, + "loss": 0.6593, + "step": 391 + }, + { + "epoch": 0.02, + "learning_rate": 4.978071300333419e-05, + "loss": 0.6317, + "step": 392 + }, + { + "epoch": 0.02, + "learning_rate": 4.978007181328546e-05, + "loss": 0.6302, + "step": 393 + }, + { + "epoch": 0.02, + "learning_rate": 4.977943062323673e-05, + "loss": 0.7447, + "step": 394 + }, + { + "epoch": 0.02, + "learning_rate": 4.9778789433188e-05, + "loss": 0.677, + "step": 395 + }, + { + "epoch": 0.02, + "learning_rate": 4.9778148243139264e-05, + "loss": 0.6667, + "step": 396 + }, + { + "epoch": 0.02, + "learning_rate": 4.9777507053090536e-05, + "loss": 0.6245, + "step": 397 + }, + { + "epoch": 0.02, + "learning_rate": 4.977686586304181e-05, + "loss": 0.6026, + "step": 398 + }, + { + "epoch": 0.02, + "learning_rate": 4.977622467299308e-05, + "loss": 0.676, + "step": 399 + }, + { + "epoch": 0.02, + "learning_rate": 4.977558348294435e-05, + "loss": 0.6143, + "step": 400 + }, + { + "epoch": 0.02, + "learning_rate": 4.9774942292895616e-05, + "loss": 0.657, + "step": 401 + }, + { + "epoch": 0.02, + "learning_rate": 4.977430110284689e-05, + "loss": 0.6636, + "step": 402 + }, + { + "epoch": 0.02, + "learning_rate": 4.977365991279815e-05, + "loss": 0.6205, + "step": 403 + }, + { + "epoch": 0.02, + "learning_rate": 4.9773018722749424e-05, + "loss": 0.599, + "step": 404 + }, + { + "epoch": 0.02, + "learning_rate": 4.977237753270069e-05, + "loss": 0.7161, + "step": 405 + }, + { + "epoch": 0.02, + "learning_rate": 4.977173634265197e-05, + "loss": 0.6414, + "step": 406 + }, + { + "epoch": 0.02, + "learning_rate": 4.977109515260324e-05, + "loss": 0.6789, + "step": 407 + }, + { + "epoch": 0.02, + "learning_rate": 4.9770453962554504e-05, + "loss": 0.6112, + "step": 408 + }, + { + "epoch": 0.02, + "learning_rate": 4.9769812772505776e-05, + "loss": 0.698, + "step": 409 + }, + { + "epoch": 0.02, + "learning_rate": 4.976917158245704e-05, + "loss": 0.6958, + "step": 410 + }, + { + "epoch": 0.02, + "learning_rate": 4.976853039240831e-05, + "loss": 0.6233, + "step": 411 + }, + { + "epoch": 0.02, + "learning_rate": 4.976788920235958e-05, + "loss": 0.6292, + "step": 412 + }, + { + "epoch": 0.02, + "learning_rate": 4.976724801231085e-05, + "loss": 0.6567, + "step": 413 + }, + { + "epoch": 0.02, + "learning_rate": 4.976660682226212e-05, + "loss": 0.8298, + "step": 414 + }, + { + "epoch": 0.02, + "learning_rate": 4.976596563221339e-05, + "loss": 0.7153, + "step": 415 + }, + { + "epoch": 0.02, + "learning_rate": 4.9765324442164665e-05, + "loss": 0.6625, + "step": 416 + }, + { + "epoch": 0.02, + "learning_rate": 4.976468325211593e-05, + "loss": 0.6642, + "step": 417 + }, + { + "epoch": 0.02, + "learning_rate": 4.97640420620672e-05, + "loss": 0.6737, + "step": 418 + }, + { + "epoch": 0.02, + "learning_rate": 4.9763400872018466e-05, + "loss": 0.6366, + "step": 419 + }, + { + "epoch": 0.02, + "learning_rate": 4.976275968196974e-05, + "loss": 0.6784, + "step": 420 + }, + { + "epoch": 0.02, + "learning_rate": 4.9762118491921e-05, + "loss": 0.6855, + "step": 421 + }, + { + "epoch": 0.02, + "learning_rate": 4.9761477301872275e-05, + "loss": 0.6458, + "step": 422 + }, + { + "epoch": 0.02, + "learning_rate": 4.9760836111823546e-05, + "loss": 0.6852, + "step": 423 + }, + { + "epoch": 0.02, + "learning_rate": 4.976019492177482e-05, + "loss": 0.6874, + "step": 424 + }, + { + "epoch": 0.02, + "learning_rate": 4.975955373172609e-05, + "loss": 0.6537, + "step": 425 + }, + { + "epoch": 0.02, + "learning_rate": 4.9758912541677355e-05, + "loss": 0.6629, + "step": 426 + }, + { + "epoch": 0.02, + "learning_rate": 4.9758271351628626e-05, + "loss": 0.641, + "step": 427 + }, + { + "epoch": 0.02, + "learning_rate": 4.975763016157989e-05, + "loss": 0.666, + "step": 428 + }, + { + "epoch": 0.02, + "learning_rate": 4.975698897153116e-05, + "loss": 0.6118, + "step": 429 + }, + { + "epoch": 0.02, + "learning_rate": 4.9756347781482435e-05, + "loss": 0.638, + "step": 430 + }, + { + "epoch": 0.02, + "learning_rate": 4.9755706591433707e-05, + "loss": 0.603, + "step": 431 + }, + { + "epoch": 0.02, + "learning_rate": 4.975506540138497e-05, + "loss": 0.6276, + "step": 432 + }, + { + "epoch": 0.02, + "learning_rate": 4.975442421133624e-05, + "loss": 0.6184, + "step": 433 + }, + { + "epoch": 0.02, + "learning_rate": 4.9753783021287515e-05, + "loss": 0.708, + "step": 434 + }, + { + "epoch": 0.02, + "learning_rate": 4.975314183123878e-05, + "loss": 0.7014, + "step": 435 + }, + { + "epoch": 0.02, + "learning_rate": 4.975250064119005e-05, + "loss": 0.7006, + "step": 436 + }, + { + "epoch": 0.02, + "learning_rate": 4.9751859451141316e-05, + "loss": 0.6508, + "step": 437 + }, + { + "epoch": 0.02, + "learning_rate": 4.975121826109259e-05, + "loss": 0.5984, + "step": 438 + }, + { + "epoch": 0.02, + "learning_rate": 4.975057707104386e-05, + "loss": 0.573, + "step": 439 + }, + { + "epoch": 0.02, + "learning_rate": 4.974993588099513e-05, + "loss": 0.6654, + "step": 440 + }, + { + "epoch": 0.02, + "learning_rate": 4.9749294690946397e-05, + "loss": 0.632, + "step": 441 + }, + { + "epoch": 0.02, + "learning_rate": 4.974865350089767e-05, + "loss": 0.6371, + "step": 442 + }, + { + "epoch": 0.02, + "learning_rate": 4.974801231084894e-05, + "loss": 0.7639, + "step": 443 + }, + { + "epoch": 0.02, + "learning_rate": 4.9747371120800205e-05, + "loss": 0.6173, + "step": 444 + }, + { + "epoch": 0.02, + "learning_rate": 4.974672993075148e-05, + "loss": 0.742, + "step": 445 + }, + { + "epoch": 0.02, + "learning_rate": 4.974608874070275e-05, + "loss": 0.6789, + "step": 446 + }, + { + "epoch": 0.02, + "learning_rate": 4.974544755065402e-05, + "loss": 0.6756, + "step": 447 + }, + { + "epoch": 0.02, + "learning_rate": 4.9744806360605285e-05, + "loss": 0.6618, + "step": 448 + }, + { + "epoch": 0.02, + "learning_rate": 4.974416517055656e-05, + "loss": 0.6753, + "step": 449 + }, + { + "epoch": 0.02, + "learning_rate": 4.974352398050782e-05, + "loss": 0.6843, + "step": 450 + }, + { + "epoch": 0.02, + "learning_rate": 4.9742882790459093e-05, + "loss": 0.6798, + "step": 451 + }, + { + "epoch": 0.02, + "learning_rate": 4.9742241600410365e-05, + "loss": 0.6273, + "step": 452 + }, + { + "epoch": 0.02, + "learning_rate": 4.974160041036163e-05, + "loss": 0.6631, + "step": 453 + }, + { + "epoch": 0.02, + "learning_rate": 4.97409592203129e-05, + "loss": 0.5897, + "step": 454 + }, + { + "epoch": 0.02, + "learning_rate": 4.9740318030264173e-05, + "loss": 0.629, + "step": 455 + }, + { + "epoch": 0.02, + "learning_rate": 4.9739676840215445e-05, + "loss": 0.6795, + "step": 456 + }, + { + "epoch": 0.02, + "learning_rate": 4.973903565016671e-05, + "loss": 0.6424, + "step": 457 + }, + { + "epoch": 0.02, + "learning_rate": 4.973839446011798e-05, + "loss": 0.6624, + "step": 458 + }, + { + "epoch": 0.02, + "learning_rate": 4.973775327006925e-05, + "loss": 0.6148, + "step": 459 + }, + { + "epoch": 0.02, + "learning_rate": 4.973711208002052e-05, + "loss": 0.6378, + "step": 460 + }, + { + "epoch": 0.02, + "learning_rate": 4.973647088997179e-05, + "loss": 0.6271, + "step": 461 + }, + { + "epoch": 0.02, + "learning_rate": 4.973582969992306e-05, + "loss": 0.6655, + "step": 462 + }, + { + "epoch": 0.02, + "learning_rate": 4.9735188509874334e-05, + "loss": 0.6797, + "step": 463 + }, + { + "epoch": 0.02, + "learning_rate": 4.97345473198256e-05, + "loss": 0.635, + "step": 464 + }, + { + "epoch": 0.02, + "learning_rate": 4.973390612977687e-05, + "loss": 0.6964, + "step": 465 + }, + { + "epoch": 0.02, + "learning_rate": 4.9733264939728135e-05, + "loss": 0.6723, + "step": 466 + }, + { + "epoch": 0.02, + "learning_rate": 4.973262374967941e-05, + "loss": 0.6123, + "step": 467 + }, + { + "epoch": 0.02, + "learning_rate": 4.973198255963067e-05, + "loss": 0.7147, + "step": 468 + }, + { + "epoch": 0.02, + "learning_rate": 4.9731341369581944e-05, + "loss": 0.6126, + "step": 469 + }, + { + "epoch": 0.02, + "learning_rate": 4.9730700179533215e-05, + "loss": 0.6339, + "step": 470 + }, + { + "epoch": 0.02, + "learning_rate": 4.973005898948449e-05, + "loss": 0.6323, + "step": 471 + }, + { + "epoch": 0.02, + "learning_rate": 4.972941779943576e-05, + "loss": 0.5681, + "step": 472 + }, + { + "epoch": 0.02, + "learning_rate": 4.9728776609387024e-05, + "loss": 0.6254, + "step": 473 + }, + { + "epoch": 0.02, + "learning_rate": 4.9728135419338295e-05, + "loss": 0.6086, + "step": 474 + }, + { + "epoch": 0.02, + "learning_rate": 4.972749422928956e-05, + "loss": 0.6252, + "step": 475 + }, + { + "epoch": 0.02, + "learning_rate": 4.972685303924083e-05, + "loss": 0.6609, + "step": 476 + }, + { + "epoch": 0.02, + "learning_rate": 4.97262118491921e-05, + "loss": 0.6388, + "step": 477 + }, + { + "epoch": 0.02, + "learning_rate": 4.9725570659143376e-05, + "loss": 0.6624, + "step": 478 + }, + { + "epoch": 0.02, + "learning_rate": 4.972492946909465e-05, + "loss": 0.6756, + "step": 479 + }, + { + "epoch": 0.02, + "learning_rate": 4.972428827904591e-05, + "loss": 0.7042, + "step": 480 + }, + { + "epoch": 0.02, + "learning_rate": 4.9723647088997184e-05, + "loss": 0.6019, + "step": 481 + }, + { + "epoch": 0.02, + "learning_rate": 4.972300589894845e-05, + "loss": 0.6086, + "step": 482 + }, + { + "epoch": 0.02, + "learning_rate": 4.972236470889972e-05, + "loss": 0.6523, + "step": 483 + }, + { + "epoch": 0.02, + "learning_rate": 4.9721723518850986e-05, + "loss": 0.6145, + "step": 484 + }, + { + "epoch": 0.02, + "learning_rate": 4.972108232880226e-05, + "loss": 0.7168, + "step": 485 + }, + { + "epoch": 0.02, + "learning_rate": 4.972044113875353e-05, + "loss": 0.5385, + "step": 486 + }, + { + "epoch": 0.02, + "learning_rate": 4.97197999487048e-05, + "loss": 0.6165, + "step": 487 + }, + { + "epoch": 0.02, + "learning_rate": 4.971915875865607e-05, + "loss": 0.6073, + "step": 488 + }, + { + "epoch": 0.02, + "learning_rate": 4.971851756860734e-05, + "loss": 0.6785, + "step": 489 + }, + { + "epoch": 0.02, + "learning_rate": 4.971787637855861e-05, + "loss": 0.6888, + "step": 490 + }, + { + "epoch": 0.02, + "learning_rate": 4.9717235188509874e-05, + "loss": 0.6306, + "step": 491 + }, + { + "epoch": 0.02, + "learning_rate": 4.9716593998461146e-05, + "loss": 0.5827, + "step": 492 + }, + { + "epoch": 0.02, + "learning_rate": 4.971595280841241e-05, + "loss": 0.6549, + "step": 493 + }, + { + "epoch": 0.02, + "learning_rate": 4.971531161836369e-05, + "loss": 0.7205, + "step": 494 + }, + { + "epoch": 0.02, + "learning_rate": 4.9714670428314954e-05, + "loss": 0.5896, + "step": 495 + }, + { + "epoch": 0.02, + "learning_rate": 4.9714029238266226e-05, + "loss": 0.6253, + "step": 496 + }, + { + "epoch": 0.02, + "learning_rate": 4.97133880482175e-05, + "loss": 0.6638, + "step": 497 + }, + { + "epoch": 0.02, + "learning_rate": 4.971274685816876e-05, + "loss": 0.6144, + "step": 498 + }, + { + "epoch": 0.02, + "learning_rate": 4.9712105668120034e-05, + "loss": 0.6595, + "step": 499 + }, + { + "epoch": 0.02, + "learning_rate": 4.97114644780713e-05, + "loss": 0.6494, + "step": 500 + }, + { + "epoch": 0.02, + "learning_rate": 4.971082328802257e-05, + "loss": 0.714, + "step": 501 + }, + { + "epoch": 0.02, + "learning_rate": 4.971018209797384e-05, + "loss": 0.6523, + "step": 502 + }, + { + "epoch": 0.02, + "learning_rate": 4.9709540907925114e-05, + "loss": 0.6692, + "step": 503 + }, + { + "epoch": 0.02, + "learning_rate": 4.970889971787638e-05, + "loss": 0.6682, + "step": 504 + }, + { + "epoch": 0.02, + "learning_rate": 4.970825852782765e-05, + "loss": 0.6382, + "step": 505 + }, + { + "epoch": 0.02, + "learning_rate": 4.970761733777892e-05, + "loss": 0.617, + "step": 506 + }, + { + "epoch": 0.02, + "learning_rate": 4.970697614773019e-05, + "loss": 0.6718, + "step": 507 + }, + { + "epoch": 0.02, + "learning_rate": 4.970633495768146e-05, + "loss": 0.6417, + "step": 508 + }, + { + "epoch": 0.02, + "learning_rate": 4.9705693767632724e-05, + "loss": 0.649, + "step": 509 + }, + { + "epoch": 0.02, + "learning_rate": 4.9705052577583996e-05, + "loss": 0.6783, + "step": 510 + }, + { + "epoch": 0.02, + "learning_rate": 4.970441138753527e-05, + "loss": 0.6132, + "step": 511 + }, + { + "epoch": 0.02, + "learning_rate": 4.970377019748654e-05, + "loss": 0.6431, + "step": 512 + }, + { + "epoch": 0.02, + "learning_rate": 4.9703129007437804e-05, + "loss": 0.6062, + "step": 513 + }, + { + "epoch": 0.02, + "learning_rate": 4.9702487817389076e-05, + "loss": 0.647, + "step": 514 + }, + { + "epoch": 0.02, + "learning_rate": 4.970184662734035e-05, + "loss": 0.597, + "step": 515 + }, + { + "epoch": 0.02, + "learning_rate": 4.970120543729161e-05, + "loss": 0.6179, + "step": 516 + }, + { + "epoch": 0.02, + "learning_rate": 4.9700564247242884e-05, + "loss": 0.6671, + "step": 517 + }, + { + "epoch": 0.02, + "learning_rate": 4.9699923057194156e-05, + "loss": 0.6752, + "step": 518 + }, + { + "epoch": 0.02, + "learning_rate": 4.969928186714543e-05, + "loss": 0.597, + "step": 519 + }, + { + "epoch": 0.02, + "learning_rate": 4.969864067709669e-05, + "loss": 0.5915, + "step": 520 + }, + { + "epoch": 0.02, + "learning_rate": 4.9697999487047964e-05, + "loss": 0.6389, + "step": 521 + }, + { + "epoch": 0.02, + "learning_rate": 4.969735829699923e-05, + "loss": 0.6883, + "step": 522 + }, + { + "epoch": 0.02, + "learning_rate": 4.96967171069505e-05, + "loss": 0.656, + "step": 523 + }, + { + "epoch": 0.02, + "learning_rate": 4.969607591690177e-05, + "loss": 0.5763, + "step": 524 + }, + { + "epoch": 0.02, + "learning_rate": 4.969543472685304e-05, + "loss": 0.5899, + "step": 525 + }, + { + "epoch": 0.02, + "learning_rate": 4.969479353680431e-05, + "loss": 0.6352, + "step": 526 + }, + { + "epoch": 0.02, + "learning_rate": 4.969415234675558e-05, + "loss": 0.6731, + "step": 527 + }, + { + "epoch": 0.02, + "learning_rate": 4.969351115670685e-05, + "loss": 0.6156, + "step": 528 + }, + { + "epoch": 0.02, + "learning_rate": 4.969286996665812e-05, + "loss": 0.6267, + "step": 529 + }, + { + "epoch": 0.02, + "learning_rate": 4.969222877660939e-05, + "loss": 0.6753, + "step": 530 + }, + { + "epoch": 0.02, + "learning_rate": 4.969158758656066e-05, + "loss": 0.7151, + "step": 531 + }, + { + "epoch": 0.02, + "learning_rate": 4.9690946396511926e-05, + "loss": 0.5535, + "step": 532 + }, + { + "epoch": 0.02, + "learning_rate": 4.96903052064632e-05, + "loss": 0.6686, + "step": 533 + }, + { + "epoch": 0.02, + "learning_rate": 4.968966401641447e-05, + "loss": 0.5681, + "step": 534 + }, + { + "epoch": 0.02, + "learning_rate": 4.968902282636574e-05, + "loss": 0.6417, + "step": 535 + }, + { + "epoch": 0.02, + "learning_rate": 4.9688381636317006e-05, + "loss": 0.6123, + "step": 536 + }, + { + "epoch": 0.02, + "learning_rate": 4.968774044626828e-05, + "loss": 0.5953, + "step": 537 + }, + { + "epoch": 0.02, + "learning_rate": 4.968709925621954e-05, + "loss": 0.6945, + "step": 538 + }, + { + "epoch": 0.02, + "learning_rate": 4.9686458066170815e-05, + "loss": 0.6493, + "step": 539 + }, + { + "epoch": 0.02, + "learning_rate": 4.9685816876122086e-05, + "loss": 0.5992, + "step": 540 + }, + { + "epoch": 0.02, + "learning_rate": 4.968517568607335e-05, + "loss": 0.5946, + "step": 541 + }, + { + "epoch": 0.02, + "learning_rate": 4.968453449602462e-05, + "loss": 0.7016, + "step": 542 + }, + { + "epoch": 0.02, + "learning_rate": 4.9683893305975895e-05, + "loss": 0.6745, + "step": 543 + }, + { + "epoch": 0.02, + "learning_rate": 4.9683252115927167e-05, + "loss": 0.6934, + "step": 544 + }, + { + "epoch": 0.02, + "learning_rate": 4.968261092587843e-05, + "loss": 0.5999, + "step": 545 + }, + { + "epoch": 0.02, + "learning_rate": 4.96819697358297e-05, + "loss": 0.6332, + "step": 546 + }, + { + "epoch": 0.02, + "learning_rate": 4.968132854578097e-05, + "loss": 0.6845, + "step": 547 + }, + { + "epoch": 0.02, + "learning_rate": 4.968068735573224e-05, + "loss": 0.6592, + "step": 548 + }, + { + "epoch": 0.02, + "learning_rate": 4.968004616568351e-05, + "loss": 0.7044, + "step": 549 + }, + { + "epoch": 0.02, + "learning_rate": 4.967940497563478e-05, + "loss": 0.6782, + "step": 550 + }, + { + "epoch": 0.02, + "learning_rate": 4.9678763785586055e-05, + "loss": 0.6153, + "step": 551 + }, + { + "epoch": 0.02, + "learning_rate": 4.967812259553732e-05, + "loss": 0.6497, + "step": 552 + }, + { + "epoch": 0.02, + "learning_rate": 4.967748140548859e-05, + "loss": 0.64, + "step": 553 + }, + { + "epoch": 0.02, + "learning_rate": 4.9676840215439857e-05, + "loss": 0.5976, + "step": 554 + }, + { + "epoch": 0.02, + "learning_rate": 4.967619902539113e-05, + "loss": 0.6447, + "step": 555 + }, + { + "epoch": 0.02, + "learning_rate": 4.967555783534239e-05, + "loss": 0.679, + "step": 556 + }, + { + "epoch": 0.02, + "learning_rate": 4.9674916645293665e-05, + "loss": 0.7191, + "step": 557 + }, + { + "epoch": 0.02, + "learning_rate": 4.967427545524494e-05, + "loss": 0.5963, + "step": 558 + }, + { + "epoch": 0.02, + "learning_rate": 4.967363426519621e-05, + "loss": 0.6973, + "step": 559 + }, + { + "epoch": 0.02, + "learning_rate": 4.967299307514748e-05, + "loss": 0.5447, + "step": 560 + }, + { + "epoch": 0.02, + "learning_rate": 4.9672351885098745e-05, + "loss": 0.5777, + "step": 561 + }, + { + "epoch": 0.02, + "learning_rate": 4.967171069505002e-05, + "loss": 0.6521, + "step": 562 + }, + { + "epoch": 0.02, + "learning_rate": 4.967106950500128e-05, + "loss": 0.6685, + "step": 563 + }, + { + "epoch": 0.02, + "learning_rate": 4.9670428314952553e-05, + "loss": 0.6211, + "step": 564 + }, + { + "epoch": 0.02, + "learning_rate": 4.966978712490382e-05, + "loss": 0.6365, + "step": 565 + }, + { + "epoch": 0.02, + "learning_rate": 4.96691459348551e-05, + "loss": 0.6297, + "step": 566 + }, + { + "epoch": 0.02, + "learning_rate": 4.966850474480637e-05, + "loss": 0.6609, + "step": 567 + }, + { + "epoch": 0.02, + "learning_rate": 4.9667863554757634e-05, + "loss": 0.6742, + "step": 568 + }, + { + "epoch": 0.02, + "learning_rate": 4.9667222364708905e-05, + "loss": 0.6379, + "step": 569 + }, + { + "epoch": 0.02, + "learning_rate": 4.966658117466017e-05, + "loss": 0.6345, + "step": 570 + }, + { + "epoch": 0.02, + "learning_rate": 4.966593998461144e-05, + "loss": 0.6507, + "step": 571 + }, + { + "epoch": 0.02, + "learning_rate": 4.966529879456271e-05, + "loss": 0.6817, + "step": 572 + }, + { + "epoch": 0.02, + "learning_rate": 4.966465760451398e-05, + "loss": 0.6335, + "step": 573 + }, + { + "epoch": 0.02, + "learning_rate": 4.966401641446525e-05, + "loss": 0.6243, + "step": 574 + }, + { + "epoch": 0.02, + "learning_rate": 4.966337522441652e-05, + "loss": 0.669, + "step": 575 + }, + { + "epoch": 0.02, + "learning_rate": 4.9662734034367794e-05, + "loss": 0.7154, + "step": 576 + }, + { + "epoch": 0.02, + "learning_rate": 4.966209284431906e-05, + "loss": 0.5784, + "step": 577 + }, + { + "epoch": 0.02, + "learning_rate": 4.966145165427033e-05, + "loss": 0.5828, + "step": 578 + }, + { + "epoch": 0.02, + "learning_rate": 4.9660810464221595e-05, + "loss": 0.6705, + "step": 579 + }, + { + "epoch": 0.02, + "learning_rate": 4.966016927417287e-05, + "loss": 0.5542, + "step": 580 + }, + { + "epoch": 0.02, + "learning_rate": 4.965952808412413e-05, + "loss": 0.6549, + "step": 581 + }, + { + "epoch": 0.02, + "learning_rate": 4.9658886894075404e-05, + "loss": 0.6165, + "step": 582 + }, + { + "epoch": 0.02, + "learning_rate": 4.9658245704026675e-05, + "loss": 0.6875, + "step": 583 + }, + { + "epoch": 0.02, + "learning_rate": 4.965760451397795e-05, + "loss": 0.6797, + "step": 584 + }, + { + "epoch": 0.02, + "learning_rate": 4.965696332392922e-05, + "loss": 0.6613, + "step": 585 + }, + { + "epoch": 0.02, + "learning_rate": 4.9656322133880484e-05, + "loss": 0.6784, + "step": 586 + }, + { + "epoch": 0.02, + "learning_rate": 4.9655680943831755e-05, + "loss": 0.6743, + "step": 587 + }, + { + "epoch": 0.02, + "learning_rate": 4.965503975378302e-05, + "loss": 0.5665, + "step": 588 + }, + { + "epoch": 0.02, + "learning_rate": 4.965439856373429e-05, + "loss": 0.6927, + "step": 589 + }, + { + "epoch": 0.02, + "learning_rate": 4.9653757373685564e-05, + "loss": 0.6604, + "step": 590 + }, + { + "epoch": 0.02, + "learning_rate": 4.9653116183636836e-05, + "loss": 0.5986, + "step": 591 + }, + { + "epoch": 0.02, + "learning_rate": 4.96524749935881e-05, + "loss": 0.5905, + "step": 592 + }, + { + "epoch": 0.02, + "learning_rate": 4.965183380353937e-05, + "loss": 0.6559, + "step": 593 + }, + { + "epoch": 0.02, + "learning_rate": 4.9651192613490644e-05, + "loss": 0.6169, + "step": 594 + }, + { + "epoch": 0.02, + "learning_rate": 4.965055142344191e-05, + "loss": 0.6748, + "step": 595 + }, + { + "epoch": 0.02, + "learning_rate": 4.964991023339318e-05, + "loss": 0.6785, + "step": 596 + }, + { + "epoch": 0.02, + "learning_rate": 4.9649269043344446e-05, + "loss": 0.7179, + "step": 597 + }, + { + "epoch": 0.02, + "learning_rate": 4.964862785329572e-05, + "loss": 0.6151, + "step": 598 + }, + { + "epoch": 0.02, + "learning_rate": 4.964798666324699e-05, + "loss": 0.626, + "step": 599 + }, + { + "epoch": 0.02, + "learning_rate": 4.964734547319826e-05, + "loss": 0.5889, + "step": 600 + }, + { + "epoch": 0.02, + "learning_rate": 4.9646704283149526e-05, + "loss": 0.6531, + "step": 601 + }, + { + "epoch": 0.02, + "learning_rate": 4.96460630931008e-05, + "loss": 0.6651, + "step": 602 + }, + { + "epoch": 0.02, + "learning_rate": 4.964542190305207e-05, + "loss": 0.643, + "step": 603 + }, + { + "epoch": 0.02, + "learning_rate": 4.9644780713003334e-05, + "loss": 0.6693, + "step": 604 + }, + { + "epoch": 0.02, + "learning_rate": 4.9644139522954606e-05, + "loss": 0.636, + "step": 605 + }, + { + "epoch": 0.02, + "learning_rate": 4.964349833290588e-05, + "loss": 0.5811, + "step": 606 + }, + { + "epoch": 0.02, + "learning_rate": 4.964285714285715e-05, + "loss": 0.7276, + "step": 607 + }, + { + "epoch": 0.02, + "learning_rate": 4.9642215952808414e-05, + "loss": 0.6619, + "step": 608 + }, + { + "epoch": 0.02, + "learning_rate": 4.9641574762759686e-05, + "loss": 0.5992, + "step": 609 + }, + { + "epoch": 0.02, + "learning_rate": 4.964093357271095e-05, + "loss": 0.572, + "step": 610 + }, + { + "epoch": 0.02, + "learning_rate": 4.964029238266222e-05, + "loss": 0.4896, + "step": 611 + }, + { + "epoch": 0.02, + "learning_rate": 4.9639651192613494e-05, + "loss": 0.5452, + "step": 612 + }, + { + "epoch": 0.02, + "learning_rate": 4.963901000256476e-05, + "loss": 0.6812, + "step": 613 + }, + { + "epoch": 0.02, + "learning_rate": 4.963836881251603e-05, + "loss": 0.6382, + "step": 614 + }, + { + "epoch": 0.02, + "learning_rate": 4.96377276224673e-05, + "loss": 0.6762, + "step": 615 + }, + { + "epoch": 0.02, + "learning_rate": 4.9637086432418574e-05, + "loss": 0.6938, + "step": 616 + }, + { + "epoch": 0.02, + "learning_rate": 4.963644524236984e-05, + "loss": 0.6374, + "step": 617 + }, + { + "epoch": 0.02, + "learning_rate": 4.963580405232111e-05, + "loss": 0.6314, + "step": 618 + }, + { + "epoch": 0.02, + "learning_rate": 4.9635162862272376e-05, + "loss": 0.643, + "step": 619 + }, + { + "epoch": 0.02, + "learning_rate": 4.963452167222365e-05, + "loss": 0.6124, + "step": 620 + }, + { + "epoch": 0.02, + "learning_rate": 4.963388048217492e-05, + "loss": 0.7339, + "step": 621 + }, + { + "epoch": 0.02, + "learning_rate": 4.963323929212619e-05, + "loss": 0.5743, + "step": 622 + }, + { + "epoch": 0.02, + "learning_rate": 4.963259810207746e-05, + "loss": 0.6549, + "step": 623 + }, + { + "epoch": 0.02, + "learning_rate": 4.963195691202873e-05, + "loss": 0.5656, + "step": 624 + }, + { + "epoch": 0.02, + "learning_rate": 4.963131572198e-05, + "loss": 0.637, + "step": 625 + }, + { + "epoch": 0.02, + "learning_rate": 4.9630674531931264e-05, + "loss": 0.6799, + "step": 626 + }, + { + "epoch": 0.02, + "learning_rate": 4.9630033341882536e-05, + "loss": 0.6278, + "step": 627 + }, + { + "epoch": 0.02, + "learning_rate": 4.96293921518338e-05, + "loss": 0.6471, + "step": 628 + }, + { + "epoch": 0.02, + "learning_rate": 4.962875096178507e-05, + "loss": 0.6452, + "step": 629 + }, + { + "epoch": 0.02, + "learning_rate": 4.9628109771736344e-05, + "loss": 0.5867, + "step": 630 + }, + { + "epoch": 0.02, + "learning_rate": 4.9627468581687616e-05, + "loss": 0.6439, + "step": 631 + }, + { + "epoch": 0.02, + "learning_rate": 4.962682739163889e-05, + "loss": 0.6095, + "step": 632 + }, + { + "epoch": 0.02, + "learning_rate": 4.962618620159015e-05, + "loss": 0.6533, + "step": 633 + }, + { + "epoch": 0.02, + "learning_rate": 4.9625545011541425e-05, + "loss": 0.6396, + "step": 634 + }, + { + "epoch": 0.02, + "learning_rate": 4.962490382149269e-05, + "loss": 0.6318, + "step": 635 + }, + { + "epoch": 0.02, + "learning_rate": 4.962426263144396e-05, + "loss": 0.63, + "step": 636 + }, + { + "epoch": 0.02, + "learning_rate": 4.9623621441395226e-05, + "loss": 0.6757, + "step": 637 + }, + { + "epoch": 0.02, + "learning_rate": 4.9622980251346505e-05, + "loss": 0.6398, + "step": 638 + }, + { + "epoch": 0.02, + "learning_rate": 4.9622339061297776e-05, + "loss": 0.6507, + "step": 639 + }, + { + "epoch": 0.02, + "learning_rate": 4.962169787124904e-05, + "loss": 0.5963, + "step": 640 + }, + { + "epoch": 0.02, + "learning_rate": 4.962105668120031e-05, + "loss": 0.6532, + "step": 641 + }, + { + "epoch": 0.02, + "learning_rate": 4.962041549115158e-05, + "loss": 0.7162, + "step": 642 + }, + { + "epoch": 0.02, + "learning_rate": 4.961977430110285e-05, + "loss": 0.7003, + "step": 643 + }, + { + "epoch": 0.02, + "learning_rate": 4.9619133111054115e-05, + "loss": 0.6422, + "step": 644 + }, + { + "epoch": 0.02, + "learning_rate": 4.9618491921005386e-05, + "loss": 0.6904, + "step": 645 + }, + { + "epoch": 0.02, + "learning_rate": 4.961785073095666e-05, + "loss": 0.6163, + "step": 646 + }, + { + "epoch": 0.02, + "learning_rate": 4.961720954090793e-05, + "loss": 0.667, + "step": 647 + }, + { + "epoch": 0.02, + "learning_rate": 4.96165683508592e-05, + "loss": 0.6777, + "step": 648 + }, + { + "epoch": 0.02, + "learning_rate": 4.9615927160810466e-05, + "loss": 0.5493, + "step": 649 + }, + { + "epoch": 0.02, + "learning_rate": 4.961528597076174e-05, + "loss": 0.6507, + "step": 650 + }, + { + "epoch": 0.03, + "learning_rate": 4.9614644780713e-05, + "loss": 0.7253, + "step": 651 + }, + { + "epoch": 0.03, + "learning_rate": 4.9614003590664275e-05, + "loss": 0.6924, + "step": 652 + }, + { + "epoch": 0.03, + "learning_rate": 4.961336240061554e-05, + "loss": 0.6597, + "step": 653 + }, + { + "epoch": 0.03, + "learning_rate": 4.961272121056682e-05, + "loss": 0.5915, + "step": 654 + }, + { + "epoch": 0.03, + "learning_rate": 4.961208002051808e-05, + "loss": 0.6417, + "step": 655 + }, + { + "epoch": 0.03, + "learning_rate": 4.9611438830469355e-05, + "loss": 0.6283, + "step": 656 + }, + { + "epoch": 0.03, + "learning_rate": 4.9610797640420627e-05, + "loss": 0.6488, + "step": 657 + }, + { + "epoch": 0.03, + "learning_rate": 4.961015645037189e-05, + "loss": 0.6755, + "step": 658 + }, + { + "epoch": 0.03, + "learning_rate": 4.960951526032316e-05, + "loss": 0.6043, + "step": 659 + }, + { + "epoch": 0.03, + "learning_rate": 4.960887407027443e-05, + "loss": 0.6193, + "step": 660 + }, + { + "epoch": 0.03, + "learning_rate": 4.96082328802257e-05, + "loss": 0.6466, + "step": 661 + }, + { + "epoch": 0.03, + "learning_rate": 4.960759169017697e-05, + "loss": 0.6463, + "step": 662 + }, + { + "epoch": 0.03, + "learning_rate": 4.960695050012824e-05, + "loss": 0.6494, + "step": 663 + }, + { + "epoch": 0.03, + "learning_rate": 4.960630931007951e-05, + "loss": 0.6123, + "step": 664 + }, + { + "epoch": 0.03, + "learning_rate": 4.960566812003078e-05, + "loss": 0.6113, + "step": 665 + }, + { + "epoch": 0.03, + "learning_rate": 4.960502692998205e-05, + "loss": 0.6986, + "step": 666 + }, + { + "epoch": 0.03, + "learning_rate": 4.960438573993332e-05, + "loss": 0.6115, + "step": 667 + }, + { + "epoch": 0.03, + "learning_rate": 4.960374454988459e-05, + "loss": 0.6127, + "step": 668 + }, + { + "epoch": 0.03, + "learning_rate": 4.960310335983585e-05, + "loss": 0.6837, + "step": 669 + }, + { + "epoch": 0.03, + "learning_rate": 4.9602462169787125e-05, + "loss": 0.6183, + "step": 670 + }, + { + "epoch": 0.03, + "learning_rate": 4.96018209797384e-05, + "loss": 0.7084, + "step": 671 + }, + { + "epoch": 0.03, + "learning_rate": 4.960117978968967e-05, + "loss": 0.6384, + "step": 672 + }, + { + "epoch": 0.03, + "learning_rate": 4.960053859964093e-05, + "loss": 0.6168, + "step": 673 + }, + { + "epoch": 0.03, + "learning_rate": 4.9599897409592205e-05, + "loss": 0.6042, + "step": 674 + }, + { + "epoch": 0.03, + "learning_rate": 4.959925621954348e-05, + "loss": 0.6329, + "step": 675 + }, + { + "epoch": 0.03, + "learning_rate": 4.959861502949474e-05, + "loss": 0.5845, + "step": 676 + }, + { + "epoch": 0.03, + "learning_rate": 4.9597973839446013e-05, + "loss": 0.6863, + "step": 677 + }, + { + "epoch": 0.03, + "learning_rate": 4.9597332649397285e-05, + "loss": 0.6907, + "step": 678 + }, + { + "epoch": 0.03, + "learning_rate": 4.959669145934856e-05, + "loss": 0.6315, + "step": 679 + }, + { + "epoch": 0.03, + "learning_rate": 4.959605026929982e-05, + "loss": 0.6329, + "step": 680 + }, + { + "epoch": 0.03, + "learning_rate": 4.9595409079251094e-05, + "loss": 0.6551, + "step": 681 + }, + { + "epoch": 0.03, + "learning_rate": 4.959476788920236e-05, + "loss": 0.5753, + "step": 682 + }, + { + "epoch": 0.03, + "learning_rate": 4.959412669915363e-05, + "loss": 0.6599, + "step": 683 + }, + { + "epoch": 0.03, + "learning_rate": 4.95934855091049e-05, + "loss": 0.6489, + "step": 684 + }, + { + "epoch": 0.03, + "learning_rate": 4.959284431905617e-05, + "loss": 0.6618, + "step": 685 + }, + { + "epoch": 0.03, + "learning_rate": 4.959220312900744e-05, + "loss": 0.6326, + "step": 686 + }, + { + "epoch": 0.03, + "learning_rate": 4.959156193895871e-05, + "loss": 0.6609, + "step": 687 + }, + { + "epoch": 0.03, + "learning_rate": 4.959092074890998e-05, + "loss": 0.6211, + "step": 688 + }, + { + "epoch": 0.03, + "learning_rate": 4.959027955886125e-05, + "loss": 0.6377, + "step": 689 + }, + { + "epoch": 0.03, + "learning_rate": 4.958963836881252e-05, + "loss": 0.6092, + "step": 690 + }, + { + "epoch": 0.03, + "learning_rate": 4.9588997178763784e-05, + "loss": 0.6439, + "step": 691 + }, + { + "epoch": 0.03, + "learning_rate": 4.9588355988715055e-05, + "loss": 0.6341, + "step": 692 + }, + { + "epoch": 0.03, + "learning_rate": 4.958771479866633e-05, + "loss": 0.6453, + "step": 693 + }, + { + "epoch": 0.03, + "learning_rate": 4.95870736086176e-05, + "loss": 0.6493, + "step": 694 + }, + { + "epoch": 0.03, + "learning_rate": 4.958643241856887e-05, + "loss": 0.5392, + "step": 695 + }, + { + "epoch": 0.03, + "learning_rate": 4.9585791228520135e-05, + "loss": 0.6367, + "step": 696 + }, + { + "epoch": 0.03, + "learning_rate": 4.958515003847141e-05, + "loss": 0.7526, + "step": 697 + }, + { + "epoch": 0.03, + "learning_rate": 4.958450884842267e-05, + "loss": 0.7398, + "step": 698 + }, + { + "epoch": 0.03, + "learning_rate": 4.9583867658373944e-05, + "loss": 0.6471, + "step": 699 + }, + { + "epoch": 0.03, + "learning_rate": 4.958322646832521e-05, + "loss": 0.6762, + "step": 700 + }, + { + "epoch": 0.03, + "learning_rate": 4.958258527827648e-05, + "loss": 0.6052, + "step": 701 + }, + { + "epoch": 0.03, + "learning_rate": 4.958194408822775e-05, + "loss": 0.5991, + "step": 702 + }, + { + "epoch": 0.03, + "learning_rate": 4.9581302898179024e-05, + "loss": 0.6992, + "step": 703 + }, + { + "epoch": 0.03, + "learning_rate": 4.9580661708130296e-05, + "loss": 0.5962, + "step": 704 + }, + { + "epoch": 0.03, + "learning_rate": 4.958002051808156e-05, + "loss": 0.6302, + "step": 705 + }, + { + "epoch": 0.03, + "learning_rate": 4.957937932803283e-05, + "loss": 0.6557, + "step": 706 + }, + { + "epoch": 0.03, + "learning_rate": 4.95787381379841e-05, + "loss": 0.7107, + "step": 707 + }, + { + "epoch": 0.03, + "learning_rate": 4.957809694793537e-05, + "loss": 0.7375, + "step": 708 + }, + { + "epoch": 0.03, + "learning_rate": 4.9577455757886634e-05, + "loss": 0.5859, + "step": 709 + }, + { + "epoch": 0.03, + "learning_rate": 4.957681456783791e-05, + "loss": 0.7068, + "step": 710 + }, + { + "epoch": 0.03, + "learning_rate": 4.9576173377789184e-05, + "loss": 0.6699, + "step": 711 + }, + { + "epoch": 0.03, + "learning_rate": 4.957553218774045e-05, + "loss": 0.6456, + "step": 712 + }, + { + "epoch": 0.03, + "learning_rate": 4.957489099769172e-05, + "loss": 0.6466, + "step": 713 + }, + { + "epoch": 0.03, + "learning_rate": 4.9574249807642986e-05, + "loss": 0.7084, + "step": 714 + }, + { + "epoch": 0.03, + "learning_rate": 4.957360861759426e-05, + "loss": 0.6147, + "step": 715 + }, + { + "epoch": 0.03, + "learning_rate": 4.957296742754552e-05, + "loss": 0.7033, + "step": 716 + }, + { + "epoch": 0.03, + "learning_rate": 4.9572326237496794e-05, + "loss": 0.6553, + "step": 717 + }, + { + "epoch": 0.03, + "learning_rate": 4.9571685047448066e-05, + "loss": 0.6992, + "step": 718 + }, + { + "epoch": 0.03, + "learning_rate": 4.957104385739934e-05, + "loss": 0.6454, + "step": 719 + }, + { + "epoch": 0.03, + "learning_rate": 4.957040266735061e-05, + "loss": 0.5606, + "step": 720 + }, + { + "epoch": 0.03, + "learning_rate": 4.9569761477301874e-05, + "loss": 0.6489, + "step": 721 + }, + { + "epoch": 0.03, + "learning_rate": 4.9569120287253146e-05, + "loss": 0.621, + "step": 722 + }, + { + "epoch": 0.03, + "learning_rate": 4.956847909720441e-05, + "loss": 0.6198, + "step": 723 + }, + { + "epoch": 0.03, + "learning_rate": 4.956783790715568e-05, + "loss": 0.5835, + "step": 724 + }, + { + "epoch": 0.03, + "learning_rate": 4.956719671710695e-05, + "loss": 0.6816, + "step": 725 + }, + { + "epoch": 0.03, + "learning_rate": 4.9566555527058226e-05, + "loss": 0.6091, + "step": 726 + }, + { + "epoch": 0.03, + "learning_rate": 4.95659143370095e-05, + "loss": 0.611, + "step": 727 + }, + { + "epoch": 0.03, + "learning_rate": 4.956527314696076e-05, + "loss": 0.6053, + "step": 728 + }, + { + "epoch": 0.03, + "learning_rate": 4.9564631956912034e-05, + "loss": 0.689, + "step": 729 + }, + { + "epoch": 0.03, + "learning_rate": 4.95639907668633e-05, + "loss": 0.6261, + "step": 730 + }, + { + "epoch": 0.03, + "learning_rate": 4.956334957681457e-05, + "loss": 0.7889, + "step": 731 + }, + { + "epoch": 0.03, + "learning_rate": 4.9562708386765836e-05, + "loss": 0.6633, + "step": 732 + }, + { + "epoch": 0.03, + "learning_rate": 4.956206719671711e-05, + "loss": 0.5869, + "step": 733 + }, + { + "epoch": 0.03, + "learning_rate": 4.956142600666838e-05, + "loss": 0.6152, + "step": 734 + }, + { + "epoch": 0.03, + "learning_rate": 4.956078481661965e-05, + "loss": 0.6178, + "step": 735 + }, + { + "epoch": 0.03, + "learning_rate": 4.956014362657092e-05, + "loss": 0.6232, + "step": 736 + }, + { + "epoch": 0.03, + "learning_rate": 4.955950243652219e-05, + "loss": 0.6795, + "step": 737 + }, + { + "epoch": 0.03, + "learning_rate": 4.955886124647346e-05, + "loss": 0.5793, + "step": 738 + }, + { + "epoch": 0.03, + "learning_rate": 4.9558220056424724e-05, + "loss": 0.6005, + "step": 739 + }, + { + "epoch": 0.03, + "learning_rate": 4.9557578866375996e-05, + "loss": 0.6802, + "step": 740 + }, + { + "epoch": 0.03, + "learning_rate": 4.955693767632726e-05, + "loss": 0.6088, + "step": 741 + }, + { + "epoch": 0.03, + "learning_rate": 4.955629648627854e-05, + "loss": 0.6884, + "step": 742 + }, + { + "epoch": 0.03, + "learning_rate": 4.9555655296229804e-05, + "loss": 0.6526, + "step": 743 + }, + { + "epoch": 0.03, + "learning_rate": 4.9555014106181076e-05, + "loss": 0.5922, + "step": 744 + }, + { + "epoch": 0.03, + "learning_rate": 4.955437291613235e-05, + "loss": 0.6442, + "step": 745 + }, + { + "epoch": 0.03, + "learning_rate": 4.955373172608361e-05, + "loss": 0.6115, + "step": 746 + }, + { + "epoch": 0.03, + "learning_rate": 4.9553090536034885e-05, + "loss": 0.635, + "step": 747 + }, + { + "epoch": 0.03, + "learning_rate": 4.955244934598615e-05, + "loss": 0.6548, + "step": 748 + }, + { + "epoch": 0.03, + "learning_rate": 4.955180815593742e-05, + "loss": 0.6193, + "step": 749 + }, + { + "epoch": 0.03, + "learning_rate": 4.955116696588869e-05, + "loss": 0.6401, + "step": 750 + }, + { + "epoch": 0.03, + "learning_rate": 4.9550525775839965e-05, + "loss": 0.6261, + "step": 751 + }, + { + "epoch": 0.03, + "learning_rate": 4.954988458579123e-05, + "loss": 0.6344, + "step": 752 + }, + { + "epoch": 0.03, + "learning_rate": 4.95492433957425e-05, + "loss": 0.6219, + "step": 753 + }, + { + "epoch": 0.03, + "learning_rate": 4.954860220569377e-05, + "loss": 0.6517, + "step": 754 + }, + { + "epoch": 0.03, + "learning_rate": 4.954796101564504e-05, + "loss": 0.5861, + "step": 755 + }, + { + "epoch": 0.03, + "learning_rate": 4.954731982559631e-05, + "loss": 0.6639, + "step": 756 + }, + { + "epoch": 0.03, + "learning_rate": 4.9546678635547575e-05, + "loss": 0.684, + "step": 757 + }, + { + "epoch": 0.03, + "learning_rate": 4.9546037445498846e-05, + "loss": 0.6374, + "step": 758 + }, + { + "epoch": 0.03, + "learning_rate": 4.954539625545012e-05, + "loss": 0.6116, + "step": 759 + }, + { + "epoch": 0.03, + "learning_rate": 4.954475506540139e-05, + "loss": 0.7166, + "step": 760 + }, + { + "epoch": 0.03, + "learning_rate": 4.9544113875352655e-05, + "loss": 0.6642, + "step": 761 + }, + { + "epoch": 0.03, + "learning_rate": 4.9543472685303926e-05, + "loss": 0.605, + "step": 762 + }, + { + "epoch": 0.03, + "learning_rate": 4.95428314952552e-05, + "loss": 0.6058, + "step": 763 + }, + { + "epoch": 0.03, + "learning_rate": 4.954219030520646e-05, + "loss": 0.6922, + "step": 764 + }, + { + "epoch": 0.03, + "learning_rate": 4.9541549115157735e-05, + "loss": 0.6777, + "step": 765 + }, + { + "epoch": 0.03, + "learning_rate": 4.9540907925109007e-05, + "loss": 0.6443, + "step": 766 + }, + { + "epoch": 0.03, + "learning_rate": 4.954026673506028e-05, + "loss": 0.6693, + "step": 767 + }, + { + "epoch": 0.03, + "learning_rate": 4.953962554501154e-05, + "loss": 0.6427, + "step": 768 + }, + { + "epoch": 0.03, + "learning_rate": 4.9538984354962815e-05, + "loss": 0.6772, + "step": 769 + }, + { + "epoch": 0.03, + "learning_rate": 4.953834316491408e-05, + "loss": 0.6608, + "step": 770 + }, + { + "epoch": 0.03, + "learning_rate": 4.953770197486535e-05, + "loss": 0.6393, + "step": 771 + }, + { + "epoch": 0.03, + "learning_rate": 4.953706078481662e-05, + "loss": 0.6128, + "step": 772 + }, + { + "epoch": 0.03, + "learning_rate": 4.953641959476789e-05, + "loss": 0.6396, + "step": 773 + }, + { + "epoch": 0.03, + "learning_rate": 4.953577840471916e-05, + "loss": 0.6322, + "step": 774 + }, + { + "epoch": 0.03, + "learning_rate": 4.953513721467043e-05, + "loss": 0.6788, + "step": 775 + }, + { + "epoch": 0.03, + "learning_rate": 4.95344960246217e-05, + "loss": 0.6351, + "step": 776 + }, + { + "epoch": 0.03, + "learning_rate": 4.953385483457297e-05, + "loss": 0.5719, + "step": 777 + }, + { + "epoch": 0.03, + "learning_rate": 4.953321364452424e-05, + "loss": 0.7127, + "step": 778 + }, + { + "epoch": 0.03, + "learning_rate": 4.9532572454475505e-05, + "loss": 0.664, + "step": 779 + }, + { + "epoch": 0.03, + "learning_rate": 4.953193126442678e-05, + "loss": 0.5958, + "step": 780 + }, + { + "epoch": 0.03, + "learning_rate": 4.953129007437805e-05, + "loss": 0.709, + "step": 781 + }, + { + "epoch": 0.03, + "learning_rate": 4.953064888432932e-05, + "loss": 0.6559, + "step": 782 + }, + { + "epoch": 0.03, + "learning_rate": 4.953000769428059e-05, + "loss": 0.6339, + "step": 783 + }, + { + "epoch": 0.03, + "learning_rate": 4.952936650423186e-05, + "loss": 0.6798, + "step": 784 + }, + { + "epoch": 0.03, + "learning_rate": 4.952872531418313e-05, + "loss": 0.6199, + "step": 785 + }, + { + "epoch": 0.03, + "learning_rate": 4.9528084124134393e-05, + "loss": 0.6618, + "step": 786 + }, + { + "epoch": 0.03, + "learning_rate": 4.9527442934085665e-05, + "loss": 0.6302, + "step": 787 + }, + { + "epoch": 0.03, + "learning_rate": 4.952680174403693e-05, + "loss": 0.6198, + "step": 788 + }, + { + "epoch": 0.03, + "learning_rate": 4.95261605539882e-05, + "loss": 0.6493, + "step": 789 + }, + { + "epoch": 0.03, + "learning_rate": 4.9525519363939473e-05, + "loss": 0.7005, + "step": 790 + }, + { + "epoch": 0.03, + "learning_rate": 4.9524878173890745e-05, + "loss": 0.6302, + "step": 791 + }, + { + "epoch": 0.03, + "learning_rate": 4.952423698384202e-05, + "loss": 0.6607, + "step": 792 + }, + { + "epoch": 0.03, + "learning_rate": 4.952359579379328e-05, + "loss": 0.5616, + "step": 793 + }, + { + "epoch": 0.03, + "learning_rate": 4.9522954603744554e-05, + "loss": 0.6194, + "step": 794 + }, + { + "epoch": 0.03, + "learning_rate": 4.952231341369582e-05, + "loss": 0.6627, + "step": 795 + }, + { + "epoch": 0.03, + "learning_rate": 4.952167222364709e-05, + "loss": 0.6716, + "step": 796 + }, + { + "epoch": 0.03, + "learning_rate": 4.9521031033598355e-05, + "loss": 0.6097, + "step": 797 + }, + { + "epoch": 0.03, + "learning_rate": 4.9520389843549634e-05, + "loss": 0.6457, + "step": 798 + }, + { + "epoch": 0.03, + "learning_rate": 4.9519748653500905e-05, + "loss": 0.603, + "step": 799 + }, + { + "epoch": 0.03, + "learning_rate": 4.951910746345217e-05, + "loss": 0.6673, + "step": 800 + }, + { + "epoch": 0.03, + "learning_rate": 4.951846627340344e-05, + "loss": 0.5897, + "step": 801 + }, + { + "epoch": 0.03, + "learning_rate": 4.951782508335471e-05, + "loss": 0.7311, + "step": 802 + }, + { + "epoch": 0.03, + "learning_rate": 4.951718389330598e-05, + "loss": 0.6246, + "step": 803 + }, + { + "epoch": 0.03, + "learning_rate": 4.9516542703257244e-05, + "loss": 0.6788, + "step": 804 + }, + { + "epoch": 0.03, + "learning_rate": 4.9515901513208515e-05, + "loss": 0.6693, + "step": 805 + }, + { + "epoch": 0.03, + "learning_rate": 4.951526032315979e-05, + "loss": 0.6324, + "step": 806 + }, + { + "epoch": 0.03, + "learning_rate": 4.951461913311106e-05, + "loss": 0.5892, + "step": 807 + }, + { + "epoch": 0.03, + "learning_rate": 4.951397794306233e-05, + "loss": 0.6213, + "step": 808 + }, + { + "epoch": 0.03, + "learning_rate": 4.9513336753013595e-05, + "loss": 0.6206, + "step": 809 + }, + { + "epoch": 0.03, + "learning_rate": 4.951269556296487e-05, + "loss": 0.6968, + "step": 810 + }, + { + "epoch": 0.03, + "learning_rate": 4.951205437291613e-05, + "loss": 0.6917, + "step": 811 + }, + { + "epoch": 0.03, + "learning_rate": 4.9511413182867404e-05, + "loss": 0.5594, + "step": 812 + }, + { + "epoch": 0.03, + "learning_rate": 4.951077199281867e-05, + "loss": 0.5473, + "step": 813 + }, + { + "epoch": 0.03, + "learning_rate": 4.951013080276995e-05, + "loss": 0.6289, + "step": 814 + }, + { + "epoch": 0.03, + "learning_rate": 4.950948961272121e-05, + "loss": 0.6278, + "step": 815 + }, + { + "epoch": 0.03, + "learning_rate": 4.9508848422672484e-05, + "loss": 0.6463, + "step": 816 + }, + { + "epoch": 0.03, + "learning_rate": 4.9508207232623756e-05, + "loss": 0.5428, + "step": 817 + }, + { + "epoch": 0.03, + "learning_rate": 4.950756604257502e-05, + "loss": 0.6682, + "step": 818 + }, + { + "epoch": 0.03, + "learning_rate": 4.950692485252629e-05, + "loss": 0.6326, + "step": 819 + }, + { + "epoch": 0.03, + "learning_rate": 4.950628366247756e-05, + "loss": 0.6056, + "step": 820 + }, + { + "epoch": 0.03, + "learning_rate": 4.950564247242883e-05, + "loss": 0.6131, + "step": 821 + }, + { + "epoch": 0.03, + "learning_rate": 4.95050012823801e-05, + "loss": 0.6319, + "step": 822 + }, + { + "epoch": 0.03, + "learning_rate": 4.950436009233137e-05, + "loss": 0.6517, + "step": 823 + }, + { + "epoch": 0.03, + "learning_rate": 4.950371890228264e-05, + "loss": 0.6375, + "step": 824 + }, + { + "epoch": 0.03, + "learning_rate": 4.950307771223391e-05, + "loss": 0.5787, + "step": 825 + }, + { + "epoch": 0.03, + "learning_rate": 4.950243652218518e-05, + "loss": 0.7115, + "step": 826 + }, + { + "epoch": 0.03, + "learning_rate": 4.9501795332136446e-05, + "loss": 0.6594, + "step": 827 + }, + { + "epoch": 0.03, + "learning_rate": 4.950115414208772e-05, + "loss": 0.657, + "step": 828 + }, + { + "epoch": 0.03, + "learning_rate": 4.950051295203898e-05, + "loss": 0.6241, + "step": 829 + }, + { + "epoch": 0.03, + "learning_rate": 4.9499871761990254e-05, + "loss": 0.6049, + "step": 830 + }, + { + "epoch": 0.03, + "learning_rate": 4.9499230571941526e-05, + "loss": 0.6005, + "step": 831 + }, + { + "epoch": 0.03, + "learning_rate": 4.94985893818928e-05, + "loss": 0.5644, + "step": 832 + }, + { + "epoch": 0.03, + "learning_rate": 4.949794819184406e-05, + "loss": 0.6173, + "step": 833 + }, + { + "epoch": 0.03, + "learning_rate": 4.9497307001795334e-05, + "loss": 0.664, + "step": 834 + }, + { + "epoch": 0.03, + "learning_rate": 4.9496665811746606e-05, + "loss": 0.6406, + "step": 835 + }, + { + "epoch": 0.03, + "learning_rate": 4.949602462169787e-05, + "loss": 0.6089, + "step": 836 + }, + { + "epoch": 0.03, + "learning_rate": 4.949538343164914e-05, + "loss": 0.6305, + "step": 837 + }, + { + "epoch": 0.03, + "learning_rate": 4.9494742241600414e-05, + "loss": 0.5968, + "step": 838 + }, + { + "epoch": 0.03, + "learning_rate": 4.9494101051551686e-05, + "loss": 0.6085, + "step": 839 + }, + { + "epoch": 0.03, + "learning_rate": 4.949345986150295e-05, + "loss": 0.6612, + "step": 840 + }, + { + "epoch": 0.03, + "learning_rate": 4.949281867145422e-05, + "loss": 0.6401, + "step": 841 + }, + { + "epoch": 0.03, + "learning_rate": 4.949217748140549e-05, + "loss": 0.6339, + "step": 842 + }, + { + "epoch": 0.03, + "learning_rate": 4.949153629135676e-05, + "loss": 0.6294, + "step": 843 + }, + { + "epoch": 0.03, + "learning_rate": 4.949089510130803e-05, + "loss": 0.6112, + "step": 844 + }, + { + "epoch": 0.03, + "learning_rate": 4.9490253911259296e-05, + "loss": 0.5932, + "step": 845 + }, + { + "epoch": 0.03, + "learning_rate": 4.948961272121057e-05, + "loss": 0.5756, + "step": 846 + }, + { + "epoch": 0.03, + "learning_rate": 4.948897153116184e-05, + "loss": 0.5911, + "step": 847 + }, + { + "epoch": 0.03, + "learning_rate": 4.948833034111311e-05, + "loss": 0.638, + "step": 848 + }, + { + "epoch": 0.03, + "learning_rate": 4.9487689151064376e-05, + "loss": 0.6916, + "step": 849 + }, + { + "epoch": 0.03, + "learning_rate": 4.948704796101565e-05, + "loss": 0.7131, + "step": 850 + }, + { + "epoch": 0.03, + "learning_rate": 4.948640677096691e-05, + "loss": 0.6246, + "step": 851 + }, + { + "epoch": 0.03, + "learning_rate": 4.9485765580918184e-05, + "loss": 0.6692, + "step": 852 + }, + { + "epoch": 0.03, + "learning_rate": 4.9485124390869456e-05, + "loss": 0.7067, + "step": 853 + }, + { + "epoch": 0.03, + "learning_rate": 4.948448320082073e-05, + "loss": 0.633, + "step": 854 + }, + { + "epoch": 0.03, + "learning_rate": 4.9483842010772e-05, + "loss": 0.7198, + "step": 855 + }, + { + "epoch": 0.03, + "learning_rate": 4.9483200820723264e-05, + "loss": 0.645, + "step": 856 + }, + { + "epoch": 0.03, + "learning_rate": 4.9482559630674536e-05, + "loss": 0.6665, + "step": 857 + }, + { + "epoch": 0.03, + "learning_rate": 4.94819184406258e-05, + "loss": 0.6515, + "step": 858 + }, + { + "epoch": 0.03, + "learning_rate": 4.948127725057707e-05, + "loss": 0.6729, + "step": 859 + }, + { + "epoch": 0.03, + "learning_rate": 4.948063606052834e-05, + "loss": 0.64, + "step": 860 + }, + { + "epoch": 0.03, + "learning_rate": 4.947999487047961e-05, + "loss": 0.6764, + "step": 861 + }, + { + "epoch": 0.03, + "learning_rate": 4.947935368043088e-05, + "loss": 0.6701, + "step": 862 + }, + { + "epoch": 0.03, + "learning_rate": 4.947871249038215e-05, + "loss": 0.6919, + "step": 863 + }, + { + "epoch": 0.03, + "learning_rate": 4.9478071300333425e-05, + "loss": 0.6646, + "step": 864 + }, + { + "epoch": 0.03, + "learning_rate": 4.947743011028469e-05, + "loss": 0.6714, + "step": 865 + }, + { + "epoch": 0.03, + "learning_rate": 4.947678892023596e-05, + "loss": 0.6658, + "step": 866 + }, + { + "epoch": 0.03, + "learning_rate": 4.9476147730187226e-05, + "loss": 0.6834, + "step": 867 + }, + { + "epoch": 0.03, + "learning_rate": 4.94755065401385e-05, + "loss": 0.6594, + "step": 868 + }, + { + "epoch": 0.03, + "learning_rate": 4.947486535008976e-05, + "loss": 0.6667, + "step": 869 + }, + { + "epoch": 0.03, + "learning_rate": 4.947422416004104e-05, + "loss": 0.6677, + "step": 870 + }, + { + "epoch": 0.03, + "learning_rate": 4.947358296999231e-05, + "loss": 0.629, + "step": 871 + }, + { + "epoch": 0.03, + "learning_rate": 4.947294177994358e-05, + "loss": 0.6192, + "step": 872 + }, + { + "epoch": 0.03, + "learning_rate": 4.947230058989485e-05, + "loss": 0.5925, + "step": 873 + }, + { + "epoch": 0.03, + "learning_rate": 4.9471659399846115e-05, + "loss": 0.6705, + "step": 874 + }, + { + "epoch": 0.03, + "learning_rate": 4.9471018209797386e-05, + "loss": 0.5997, + "step": 875 + }, + { + "epoch": 0.03, + "learning_rate": 4.947037701974865e-05, + "loss": 0.6135, + "step": 876 + }, + { + "epoch": 0.03, + "learning_rate": 4.946973582969992e-05, + "loss": 0.6076, + "step": 877 + }, + { + "epoch": 0.03, + "learning_rate": 4.9469094639651195e-05, + "loss": 0.5946, + "step": 878 + }, + { + "epoch": 0.03, + "learning_rate": 4.9468453449602467e-05, + "loss": 0.6681, + "step": 879 + }, + { + "epoch": 0.03, + "learning_rate": 4.946781225955374e-05, + "loss": 0.6808, + "step": 880 + }, + { + "epoch": 0.03, + "learning_rate": 4.9467171069505e-05, + "loss": 0.6558, + "step": 881 + }, + { + "epoch": 0.03, + "learning_rate": 4.9466529879456275e-05, + "loss": 0.6518, + "step": 882 + }, + { + "epoch": 0.03, + "learning_rate": 4.946588868940754e-05, + "loss": 0.6226, + "step": 883 + }, + { + "epoch": 0.03, + "learning_rate": 4.946524749935881e-05, + "loss": 0.6351, + "step": 884 + }, + { + "epoch": 0.03, + "learning_rate": 4.9464606309310077e-05, + "loss": 0.6952, + "step": 885 + }, + { + "epoch": 0.03, + "learning_rate": 4.9463965119261355e-05, + "loss": 0.583, + "step": 886 + }, + { + "epoch": 0.03, + "learning_rate": 4.946332392921262e-05, + "loss": 0.5459, + "step": 887 + }, + { + "epoch": 0.03, + "learning_rate": 4.946268273916389e-05, + "loss": 0.6808, + "step": 888 + }, + { + "epoch": 0.03, + "learning_rate": 4.946204154911516e-05, + "loss": 0.6128, + "step": 889 + }, + { + "epoch": 0.03, + "learning_rate": 4.946140035906643e-05, + "loss": 0.6721, + "step": 890 + }, + { + "epoch": 0.03, + "learning_rate": 4.94607591690177e-05, + "loss": 0.6236, + "step": 891 + }, + { + "epoch": 0.03, + "learning_rate": 4.9460117978968965e-05, + "loss": 0.6064, + "step": 892 + }, + { + "epoch": 0.03, + "learning_rate": 4.945947678892024e-05, + "loss": 0.6573, + "step": 893 + }, + { + "epoch": 0.03, + "learning_rate": 4.945883559887151e-05, + "loss": 0.7008, + "step": 894 + }, + { + "epoch": 0.03, + "learning_rate": 4.945819440882278e-05, + "loss": 0.6129, + "step": 895 + }, + { + "epoch": 0.03, + "learning_rate": 4.9457553218774045e-05, + "loss": 0.5817, + "step": 896 + }, + { + "epoch": 0.03, + "learning_rate": 4.945691202872532e-05, + "loss": 0.632, + "step": 897 + }, + { + "epoch": 0.03, + "learning_rate": 4.945627083867659e-05, + "loss": 0.6305, + "step": 898 + }, + { + "epoch": 0.03, + "learning_rate": 4.9455629648627853e-05, + "loss": 0.6501, + "step": 899 + }, + { + "epoch": 0.03, + "learning_rate": 4.9454988458579125e-05, + "loss": 0.6227, + "step": 900 + }, + { + "epoch": 0.03, + "learning_rate": 4.945434726853039e-05, + "loss": 0.5778, + "step": 901 + }, + { + "epoch": 0.03, + "learning_rate": 4.945370607848167e-05, + "loss": 0.6382, + "step": 902 + }, + { + "epoch": 0.03, + "learning_rate": 4.9453064888432934e-05, + "loss": 0.6131, + "step": 903 + }, + { + "epoch": 0.03, + "learning_rate": 4.9452423698384205e-05, + "loss": 0.7437, + "step": 904 + }, + { + "epoch": 0.03, + "learning_rate": 4.945178250833548e-05, + "loss": 0.7296, + "step": 905 + }, + { + "epoch": 0.03, + "learning_rate": 4.945114131828674e-05, + "loss": 0.6221, + "step": 906 + }, + { + "epoch": 0.03, + "learning_rate": 4.9450500128238014e-05, + "loss": 0.6076, + "step": 907 + }, + { + "epoch": 0.03, + "learning_rate": 4.944985893818928e-05, + "loss": 0.628, + "step": 908 + }, + { + "epoch": 0.03, + "learning_rate": 4.944921774814055e-05, + "loss": 0.6738, + "step": 909 + }, + { + "epoch": 0.03, + "learning_rate": 4.944857655809182e-05, + "loss": 0.6696, + "step": 910 + }, + { + "epoch": 0.04, + "learning_rate": 4.9447935368043094e-05, + "loss": 0.5837, + "step": 911 + }, + { + "epoch": 0.04, + "learning_rate": 4.944729417799436e-05, + "loss": 0.6761, + "step": 912 + }, + { + "epoch": 0.04, + "learning_rate": 4.944665298794563e-05, + "loss": 0.6168, + "step": 913 + }, + { + "epoch": 0.04, + "learning_rate": 4.94460117978969e-05, + "loss": 0.683, + "step": 914 + }, + { + "epoch": 0.04, + "learning_rate": 4.944537060784817e-05, + "loss": 0.5655, + "step": 915 + }, + { + "epoch": 0.04, + "learning_rate": 4.944472941779944e-05, + "loss": 0.5927, + "step": 916 + }, + { + "epoch": 0.04, + "learning_rate": 4.9444088227750704e-05, + "loss": 0.649, + "step": 917 + }, + { + "epoch": 0.04, + "learning_rate": 4.9443447037701975e-05, + "loss": 0.648, + "step": 918 + }, + { + "epoch": 0.04, + "learning_rate": 4.944280584765325e-05, + "loss": 0.5429, + "step": 919 + }, + { + "epoch": 0.04, + "learning_rate": 4.944216465760452e-05, + "loss": 0.7035, + "step": 920 + }, + { + "epoch": 0.04, + "learning_rate": 4.9441523467555784e-05, + "loss": 0.6246, + "step": 921 + }, + { + "epoch": 0.04, + "learning_rate": 4.9440882277507055e-05, + "loss": 0.6124, + "step": 922 + }, + { + "epoch": 0.04, + "learning_rate": 4.944024108745833e-05, + "loss": 0.6932, + "step": 923 + }, + { + "epoch": 0.04, + "learning_rate": 4.943959989740959e-05, + "loss": 0.6222, + "step": 924 + }, + { + "epoch": 0.04, + "learning_rate": 4.9438958707360864e-05, + "loss": 0.6933, + "step": 925 + }, + { + "epoch": 0.04, + "learning_rate": 4.9438317517312136e-05, + "loss": 0.597, + "step": 926 + }, + { + "epoch": 0.04, + "learning_rate": 4.943767632726341e-05, + "loss": 0.6565, + "step": 927 + }, + { + "epoch": 0.04, + "learning_rate": 4.943703513721467e-05, + "loss": 0.6734, + "step": 928 + }, + { + "epoch": 0.04, + "learning_rate": 4.9436393947165944e-05, + "loss": 0.628, + "step": 929 + }, + { + "epoch": 0.04, + "learning_rate": 4.943575275711721e-05, + "loss": 0.686, + "step": 930 + }, + { + "epoch": 0.04, + "learning_rate": 4.943511156706848e-05, + "loss": 0.6623, + "step": 931 + }, + { + "epoch": 0.04, + "learning_rate": 4.943447037701975e-05, + "loss": 0.6719, + "step": 932 + }, + { + "epoch": 0.04, + "learning_rate": 4.943382918697102e-05, + "loss": 0.6355, + "step": 933 + }, + { + "epoch": 0.04, + "learning_rate": 4.943318799692229e-05, + "loss": 0.612, + "step": 934 + }, + { + "epoch": 0.04, + "learning_rate": 4.943254680687356e-05, + "loss": 0.6643, + "step": 935 + }, + { + "epoch": 0.04, + "learning_rate": 4.943190561682483e-05, + "loss": 0.5757, + "step": 936 + }, + { + "epoch": 0.04, + "learning_rate": 4.94312644267761e-05, + "loss": 0.6696, + "step": 937 + }, + { + "epoch": 0.04, + "learning_rate": 4.943062323672737e-05, + "loss": 0.7139, + "step": 938 + }, + { + "epoch": 0.04, + "learning_rate": 4.9429982046678634e-05, + "loss": 0.6178, + "step": 939 + }, + { + "epoch": 0.04, + "learning_rate": 4.9429340856629906e-05, + "loss": 0.6183, + "step": 940 + }, + { + "epoch": 0.04, + "learning_rate": 4.942869966658118e-05, + "loss": 0.6509, + "step": 941 + }, + { + "epoch": 0.04, + "learning_rate": 4.942805847653245e-05, + "loss": 0.5784, + "step": 942 + }, + { + "epoch": 0.04, + "learning_rate": 4.942741728648372e-05, + "loss": 0.5949, + "step": 943 + }, + { + "epoch": 0.04, + "learning_rate": 4.9426776096434986e-05, + "loss": 0.685, + "step": 944 + }, + { + "epoch": 0.04, + "learning_rate": 4.942613490638626e-05, + "loss": 0.6262, + "step": 945 + }, + { + "epoch": 0.04, + "learning_rate": 4.942549371633752e-05, + "loss": 0.6199, + "step": 946 + }, + { + "epoch": 0.04, + "learning_rate": 4.9424852526288794e-05, + "loss": 0.6449, + "step": 947 + }, + { + "epoch": 0.04, + "learning_rate": 4.942421133624006e-05, + "loss": 0.6261, + "step": 948 + }, + { + "epoch": 0.04, + "learning_rate": 4.942357014619133e-05, + "loss": 0.5987, + "step": 949 + }, + { + "epoch": 0.04, + "learning_rate": 4.94229289561426e-05, + "loss": 0.6173, + "step": 950 + }, + { + "epoch": 0.04, + "learning_rate": 4.9422287766093874e-05, + "loss": 0.6068, + "step": 951 + }, + { + "epoch": 0.04, + "learning_rate": 4.9421646576045146e-05, + "loss": 0.6703, + "step": 952 + }, + { + "epoch": 0.04, + "learning_rate": 4.942100538599641e-05, + "loss": 0.6945, + "step": 953 + }, + { + "epoch": 0.04, + "learning_rate": 4.942036419594768e-05, + "loss": 0.5756, + "step": 954 + }, + { + "epoch": 0.04, + "learning_rate": 4.941972300589895e-05, + "loss": 0.6707, + "step": 955 + }, + { + "epoch": 0.04, + "learning_rate": 4.941908181585022e-05, + "loss": 0.6794, + "step": 956 + }, + { + "epoch": 0.04, + "learning_rate": 4.9418440625801484e-05, + "loss": 0.6802, + "step": 957 + }, + { + "epoch": 0.04, + "learning_rate": 4.941779943575276e-05, + "loss": 0.6641, + "step": 958 + }, + { + "epoch": 0.04, + "learning_rate": 4.9417158245704034e-05, + "loss": 0.5789, + "step": 959 + }, + { + "epoch": 0.04, + "learning_rate": 4.94165170556553e-05, + "loss": 0.6074, + "step": 960 + }, + { + "epoch": 0.04, + "learning_rate": 4.941587586560657e-05, + "loss": 0.6838, + "step": 961 + }, + { + "epoch": 0.04, + "learning_rate": 4.9415234675557836e-05, + "loss": 0.6742, + "step": 962 + }, + { + "epoch": 0.04, + "learning_rate": 4.941459348550911e-05, + "loss": 0.6201, + "step": 963 + }, + { + "epoch": 0.04, + "learning_rate": 4.941395229546037e-05, + "loss": 0.6545, + "step": 964 + }, + { + "epoch": 0.04, + "learning_rate": 4.9413311105411644e-05, + "loss": 0.6457, + "step": 965 + }, + { + "epoch": 0.04, + "learning_rate": 4.9412669915362916e-05, + "loss": 0.6683, + "step": 966 + }, + { + "epoch": 0.04, + "learning_rate": 4.941202872531419e-05, + "loss": 0.5826, + "step": 967 + }, + { + "epoch": 0.04, + "learning_rate": 4.941138753526546e-05, + "loss": 0.6712, + "step": 968 + }, + { + "epoch": 0.04, + "learning_rate": 4.9410746345216725e-05, + "loss": 0.6487, + "step": 969 + }, + { + "epoch": 0.04, + "learning_rate": 4.9410105155167996e-05, + "loss": 0.6526, + "step": 970 + }, + { + "epoch": 0.04, + "learning_rate": 4.940946396511926e-05, + "loss": 0.6315, + "step": 971 + }, + { + "epoch": 0.04, + "learning_rate": 4.940882277507053e-05, + "loss": 0.6124, + "step": 972 + }, + { + "epoch": 0.04, + "learning_rate": 4.94081815850218e-05, + "loss": 0.5969, + "step": 973 + }, + { + "epoch": 0.04, + "learning_rate": 4.9407540394973076e-05, + "loss": 0.7066, + "step": 974 + }, + { + "epoch": 0.04, + "learning_rate": 4.940689920492434e-05, + "loss": 0.5995, + "step": 975 + }, + { + "epoch": 0.04, + "learning_rate": 4.940625801487561e-05, + "loss": 0.6133, + "step": 976 + }, + { + "epoch": 0.04, + "learning_rate": 4.9405616824826885e-05, + "loss": 0.5847, + "step": 977 + }, + { + "epoch": 0.04, + "learning_rate": 4.940497563477815e-05, + "loss": 0.6204, + "step": 978 + }, + { + "epoch": 0.04, + "learning_rate": 4.940433444472942e-05, + "loss": 0.6569, + "step": 979 + }, + { + "epoch": 0.04, + "learning_rate": 4.9403693254680686e-05, + "loss": 0.5952, + "step": 980 + }, + { + "epoch": 0.04, + "learning_rate": 4.940305206463196e-05, + "loss": 0.6449, + "step": 981 + }, + { + "epoch": 0.04, + "learning_rate": 4.940241087458323e-05, + "loss": 0.6288, + "step": 982 + }, + { + "epoch": 0.04, + "learning_rate": 4.94017696845345e-05, + "loss": 0.7162, + "step": 983 + }, + { + "epoch": 0.04, + "learning_rate": 4.9401128494485766e-05, + "loss": 0.6493, + "step": 984 + }, + { + "epoch": 0.04, + "learning_rate": 4.940048730443704e-05, + "loss": 0.7039, + "step": 985 + }, + { + "epoch": 0.04, + "learning_rate": 4.939984611438831e-05, + "loss": 0.6812, + "step": 986 + }, + { + "epoch": 0.04, + "learning_rate": 4.9399204924339575e-05, + "loss": 0.6324, + "step": 987 + }, + { + "epoch": 0.04, + "learning_rate": 4.9398563734290846e-05, + "loss": 0.6342, + "step": 988 + }, + { + "epoch": 0.04, + "learning_rate": 4.939792254424211e-05, + "loss": 0.5952, + "step": 989 + }, + { + "epoch": 0.04, + "learning_rate": 4.939728135419339e-05, + "loss": 0.6421, + "step": 990 + }, + { + "epoch": 0.04, + "learning_rate": 4.9396640164144655e-05, + "loss": 0.6702, + "step": 991 + }, + { + "epoch": 0.04, + "learning_rate": 4.9395998974095927e-05, + "loss": 0.584, + "step": 992 + }, + { + "epoch": 0.04, + "learning_rate": 4.939535778404719e-05, + "loss": 0.5872, + "step": 993 + }, + { + "epoch": 0.04, + "learning_rate": 4.939471659399846e-05, + "loss": 0.7133, + "step": 994 + }, + { + "epoch": 0.04, + "learning_rate": 4.9394075403949735e-05, + "loss": 0.625, + "step": 995 + }, + { + "epoch": 0.04, + "learning_rate": 4.9393434213901e-05, + "loss": 0.6517, + "step": 996 + }, + { + "epoch": 0.04, + "learning_rate": 4.939279302385227e-05, + "loss": 0.631, + "step": 997 + }, + { + "epoch": 0.04, + "learning_rate": 4.939215183380354e-05, + "loss": 0.6681, + "step": 998 + }, + { + "epoch": 0.04, + "learning_rate": 4.9391510643754815e-05, + "loss": 0.5949, + "step": 999 + }, + { + "epoch": 0.04, + "learning_rate": 4.939086945370608e-05, + "loss": 0.6614, + "step": 1000 + }, + { + "epoch": 0.04, + "learning_rate": 4.939022826365735e-05, + "loss": 0.6788, + "step": 1001 + }, + { + "epoch": 0.04, + "learning_rate": 4.938958707360862e-05, + "loss": 0.6285, + "step": 1002 + }, + { + "epoch": 0.04, + "learning_rate": 4.938894588355989e-05, + "loss": 0.5653, + "step": 1003 + }, + { + "epoch": 0.04, + "learning_rate": 4.938830469351116e-05, + "loss": 0.5902, + "step": 1004 + }, + { + "epoch": 0.04, + "learning_rate": 4.9387663503462425e-05, + "loss": 0.6269, + "step": 1005 + }, + { + "epoch": 0.04, + "learning_rate": 4.93870223134137e-05, + "loss": 0.6283, + "step": 1006 + }, + { + "epoch": 0.04, + "learning_rate": 4.938638112336497e-05, + "loss": 0.678, + "step": 1007 + }, + { + "epoch": 0.04, + "learning_rate": 4.938573993331624e-05, + "loss": 0.6407, + "step": 1008 + }, + { + "epoch": 0.04, + "learning_rate": 4.9385098743267505e-05, + "loss": 0.6411, + "step": 1009 + }, + { + "epoch": 0.04, + "learning_rate": 4.938445755321878e-05, + "loss": 0.6333, + "step": 1010 + }, + { + "epoch": 0.04, + "learning_rate": 4.938381636317004e-05, + "loss": 0.6142, + "step": 1011 + }, + { + "epoch": 0.04, + "learning_rate": 4.9383175173121313e-05, + "loss": 0.7537, + "step": 1012 + }, + { + "epoch": 0.04, + "learning_rate": 4.9382533983072585e-05, + "loss": 0.6601, + "step": 1013 + }, + { + "epoch": 0.04, + "learning_rate": 4.938189279302386e-05, + "loss": 0.7013, + "step": 1014 + }, + { + "epoch": 0.04, + "learning_rate": 4.938125160297513e-05, + "loss": 0.5944, + "step": 1015 + }, + { + "epoch": 0.04, + "learning_rate": 4.9380610412926394e-05, + "loss": 0.628, + "step": 1016 + }, + { + "epoch": 0.04, + "learning_rate": 4.9379969222877665e-05, + "loss": 0.6064, + "step": 1017 + }, + { + "epoch": 0.04, + "learning_rate": 4.937932803282893e-05, + "loss": 0.6185, + "step": 1018 + }, + { + "epoch": 0.04, + "learning_rate": 4.93786868427802e-05, + "loss": 0.5976, + "step": 1019 + }, + { + "epoch": 0.04, + "learning_rate": 4.937804565273147e-05, + "loss": 0.6966, + "step": 1020 + }, + { + "epoch": 0.04, + "learning_rate": 4.937740446268274e-05, + "loss": 0.5636, + "step": 1021 + }, + { + "epoch": 0.04, + "learning_rate": 4.937676327263401e-05, + "loss": 0.5755, + "step": 1022 + }, + { + "epoch": 0.04, + "learning_rate": 4.937612208258528e-05, + "loss": 0.6425, + "step": 1023 + }, + { + "epoch": 0.04, + "learning_rate": 4.9375480892536554e-05, + "loss": 0.645, + "step": 1024 + }, + { + "epoch": 0.04, + "learning_rate": 4.937483970248782e-05, + "loss": 0.5782, + "step": 1025 + }, + { + "epoch": 0.04, + "learning_rate": 4.937419851243909e-05, + "loss": 0.6828, + "step": 1026 + }, + { + "epoch": 0.04, + "learning_rate": 4.9373557322390355e-05, + "loss": 0.6423, + "step": 1027 + }, + { + "epoch": 0.04, + "learning_rate": 4.937291613234163e-05, + "loss": 0.6207, + "step": 1028 + }, + { + "epoch": 0.04, + "learning_rate": 4.93722749422929e-05, + "loss": 0.599, + "step": 1029 + }, + { + "epoch": 0.04, + "learning_rate": 4.937163375224417e-05, + "loss": 0.6785, + "step": 1030 + }, + { + "epoch": 0.04, + "learning_rate": 4.937099256219544e-05, + "loss": 0.5936, + "step": 1031 + }, + { + "epoch": 0.04, + "learning_rate": 4.937035137214671e-05, + "loss": 0.6905, + "step": 1032 + }, + { + "epoch": 0.04, + "learning_rate": 4.936971018209798e-05, + "loss": 0.6007, + "step": 1033 + }, + { + "epoch": 0.04, + "learning_rate": 4.9369068992049244e-05, + "loss": 0.6559, + "step": 1034 + }, + { + "epoch": 0.04, + "learning_rate": 4.9368427802000516e-05, + "loss": 0.6406, + "step": 1035 + }, + { + "epoch": 0.04, + "learning_rate": 4.936778661195178e-05, + "loss": 0.6758, + "step": 1036 + }, + { + "epoch": 0.04, + "learning_rate": 4.936714542190305e-05, + "loss": 0.6325, + "step": 1037 + }, + { + "epoch": 0.04, + "learning_rate": 4.9366504231854324e-05, + "loss": 0.6318, + "step": 1038 + }, + { + "epoch": 0.04, + "learning_rate": 4.9365863041805596e-05, + "loss": 0.6581, + "step": 1039 + }, + { + "epoch": 0.04, + "learning_rate": 4.936522185175687e-05, + "loss": 0.6115, + "step": 1040 + }, + { + "epoch": 0.04, + "learning_rate": 4.936458066170813e-05, + "loss": 0.629, + "step": 1041 + }, + { + "epoch": 0.04, + "learning_rate": 4.9363939471659404e-05, + "loss": 0.6885, + "step": 1042 + }, + { + "epoch": 0.04, + "learning_rate": 4.936329828161067e-05, + "loss": 0.6346, + "step": 1043 + }, + { + "epoch": 0.04, + "learning_rate": 4.936265709156194e-05, + "loss": 0.6513, + "step": 1044 + }, + { + "epoch": 0.04, + "learning_rate": 4.9362015901513206e-05, + "loss": 0.5622, + "step": 1045 + }, + { + "epoch": 0.04, + "learning_rate": 4.9361374711464484e-05, + "loss": 0.5957, + "step": 1046 + }, + { + "epoch": 0.04, + "learning_rate": 4.936073352141575e-05, + "loss": 0.6285, + "step": 1047 + }, + { + "epoch": 0.04, + "learning_rate": 4.936009233136702e-05, + "loss": 0.5611, + "step": 1048 + }, + { + "epoch": 0.04, + "learning_rate": 4.935945114131829e-05, + "loss": 0.5589, + "step": 1049 + }, + { + "epoch": 0.04, + "learning_rate": 4.935880995126956e-05, + "loss": 0.5902, + "step": 1050 + }, + { + "epoch": 0.04, + "learning_rate": 4.935816876122083e-05, + "loss": 0.6467, + "step": 1051 + }, + { + "epoch": 0.04, + "learning_rate": 4.9357527571172094e-05, + "loss": 0.5789, + "step": 1052 + }, + { + "epoch": 0.04, + "learning_rate": 4.9356886381123366e-05, + "loss": 0.6692, + "step": 1053 + }, + { + "epoch": 0.04, + "learning_rate": 4.935624519107464e-05, + "loss": 0.5411, + "step": 1054 + }, + { + "epoch": 0.04, + "learning_rate": 4.935560400102591e-05, + "loss": 0.5624, + "step": 1055 + }, + { + "epoch": 0.04, + "learning_rate": 4.9354962810977174e-05, + "loss": 0.7045, + "step": 1056 + }, + { + "epoch": 0.04, + "learning_rate": 4.9354321620928446e-05, + "loss": 0.6405, + "step": 1057 + }, + { + "epoch": 0.04, + "learning_rate": 4.935368043087972e-05, + "loss": 0.6338, + "step": 1058 + }, + { + "epoch": 0.04, + "learning_rate": 4.935303924083098e-05, + "loss": 0.6587, + "step": 1059 + }, + { + "epoch": 0.04, + "learning_rate": 4.9352398050782254e-05, + "loss": 0.6025, + "step": 1060 + }, + { + "epoch": 0.04, + "learning_rate": 4.935175686073352e-05, + "loss": 0.6361, + "step": 1061 + }, + { + "epoch": 0.04, + "learning_rate": 4.93511156706848e-05, + "loss": 0.6261, + "step": 1062 + }, + { + "epoch": 0.04, + "learning_rate": 4.935047448063606e-05, + "loss": 0.6134, + "step": 1063 + }, + { + "epoch": 0.04, + "learning_rate": 4.9349833290587334e-05, + "loss": 0.6238, + "step": 1064 + }, + { + "epoch": 0.04, + "learning_rate": 4.93491921005386e-05, + "loss": 0.6982, + "step": 1065 + }, + { + "epoch": 0.04, + "learning_rate": 4.934855091048987e-05, + "loss": 0.6934, + "step": 1066 + }, + { + "epoch": 0.04, + "learning_rate": 4.934790972044114e-05, + "loss": 0.6582, + "step": 1067 + }, + { + "epoch": 0.04, + "learning_rate": 4.934726853039241e-05, + "loss": 0.6006, + "step": 1068 + }, + { + "epoch": 0.04, + "learning_rate": 4.934662734034368e-05, + "loss": 0.6132, + "step": 1069 + }, + { + "epoch": 0.04, + "learning_rate": 4.934598615029495e-05, + "loss": 0.6555, + "step": 1070 + }, + { + "epoch": 0.04, + "learning_rate": 4.934534496024622e-05, + "loss": 0.647, + "step": 1071 + }, + { + "epoch": 0.04, + "learning_rate": 4.934470377019749e-05, + "loss": 0.6107, + "step": 1072 + }, + { + "epoch": 0.04, + "learning_rate": 4.934406258014876e-05, + "loss": 0.6296, + "step": 1073 + }, + { + "epoch": 0.04, + "learning_rate": 4.9343421390100024e-05, + "loss": 0.6079, + "step": 1074 + }, + { + "epoch": 0.04, + "learning_rate": 4.9342780200051296e-05, + "loss": 0.6084, + "step": 1075 + }, + { + "epoch": 0.04, + "learning_rate": 4.934213901000257e-05, + "loss": 0.5968, + "step": 1076 + }, + { + "epoch": 0.04, + "learning_rate": 4.934149781995383e-05, + "loss": 0.6041, + "step": 1077 + }, + { + "epoch": 0.04, + "learning_rate": 4.9340856629905104e-05, + "loss": 0.6348, + "step": 1078 + }, + { + "epoch": 0.04, + "learning_rate": 4.9340215439856376e-05, + "loss": 0.6837, + "step": 1079 + }, + { + "epoch": 0.04, + "learning_rate": 4.933957424980765e-05, + "loss": 0.528, + "step": 1080 + }, + { + "epoch": 0.04, + "learning_rate": 4.933893305975891e-05, + "loss": 0.5639, + "step": 1081 + }, + { + "epoch": 0.04, + "learning_rate": 4.9338291869710185e-05, + "loss": 0.6176, + "step": 1082 + }, + { + "epoch": 0.04, + "learning_rate": 4.933765067966145e-05, + "loss": 0.6413, + "step": 1083 + }, + { + "epoch": 0.04, + "learning_rate": 4.933700948961272e-05, + "loss": 0.6985, + "step": 1084 + }, + { + "epoch": 0.04, + "learning_rate": 4.933636829956399e-05, + "loss": 0.5917, + "step": 1085 + }, + { + "epoch": 0.04, + "learning_rate": 4.9335727109515265e-05, + "loss": 0.5843, + "step": 1086 + }, + { + "epoch": 0.04, + "learning_rate": 4.9335085919466536e-05, + "loss": 0.6769, + "step": 1087 + }, + { + "epoch": 0.04, + "learning_rate": 4.93344447294178e-05, + "loss": 0.6577, + "step": 1088 + }, + { + "epoch": 0.04, + "learning_rate": 4.933380353936907e-05, + "loss": 0.5663, + "step": 1089 + }, + { + "epoch": 0.04, + "learning_rate": 4.933316234932034e-05, + "loss": 0.6394, + "step": 1090 + }, + { + "epoch": 0.04, + "learning_rate": 4.933252115927161e-05, + "loss": 0.6373, + "step": 1091 + }, + { + "epoch": 0.04, + "learning_rate": 4.933187996922288e-05, + "loss": 0.5672, + "step": 1092 + }, + { + "epoch": 0.04, + "learning_rate": 4.9331238779174146e-05, + "loss": 0.6452, + "step": 1093 + }, + { + "epoch": 0.04, + "learning_rate": 4.933059758912542e-05, + "loss": 0.6157, + "step": 1094 + }, + { + "epoch": 0.04, + "learning_rate": 4.932995639907669e-05, + "loss": 0.6502, + "step": 1095 + }, + { + "epoch": 0.04, + "learning_rate": 4.932931520902796e-05, + "loss": 0.6534, + "step": 1096 + }, + { + "epoch": 0.04, + "learning_rate": 4.9328674018979226e-05, + "loss": 0.6426, + "step": 1097 + }, + { + "epoch": 0.04, + "learning_rate": 4.93280328289305e-05, + "loss": 0.7128, + "step": 1098 + }, + { + "epoch": 0.04, + "learning_rate": 4.932739163888176e-05, + "loss": 0.6342, + "step": 1099 + }, + { + "epoch": 0.04, + "learning_rate": 4.9326750448833035e-05, + "loss": 0.6487, + "step": 1100 + }, + { + "epoch": 0.04, + "learning_rate": 4.9326109258784307e-05, + "loss": 0.6581, + "step": 1101 + }, + { + "epoch": 0.04, + "learning_rate": 4.932546806873558e-05, + "loss": 0.6064, + "step": 1102 + }, + { + "epoch": 0.04, + "learning_rate": 4.932482687868685e-05, + "loss": 0.653, + "step": 1103 + }, + { + "epoch": 0.04, + "learning_rate": 4.9324185688638115e-05, + "loss": 0.6097, + "step": 1104 + }, + { + "epoch": 0.04, + "learning_rate": 4.9323544498589387e-05, + "loss": 0.6136, + "step": 1105 + }, + { + "epoch": 0.04, + "learning_rate": 4.932290330854065e-05, + "loss": 0.6213, + "step": 1106 + }, + { + "epoch": 0.04, + "learning_rate": 4.932226211849192e-05, + "loss": 0.671, + "step": 1107 + }, + { + "epoch": 0.04, + "learning_rate": 4.932162092844319e-05, + "loss": 0.6416, + "step": 1108 + }, + { + "epoch": 0.04, + "learning_rate": 4.932097973839446e-05, + "loss": 0.6239, + "step": 1109 + }, + { + "epoch": 0.04, + "learning_rate": 4.932033854834573e-05, + "loss": 0.6333, + "step": 1110 + }, + { + "epoch": 0.04, + "learning_rate": 4.9319697358297e-05, + "loss": 0.5954, + "step": 1111 + }, + { + "epoch": 0.04, + "learning_rate": 4.9319056168248275e-05, + "loss": 0.6552, + "step": 1112 + }, + { + "epoch": 0.04, + "learning_rate": 4.931841497819954e-05, + "loss": 0.6534, + "step": 1113 + }, + { + "epoch": 0.04, + "learning_rate": 4.931777378815081e-05, + "loss": 0.651, + "step": 1114 + }, + { + "epoch": 0.04, + "learning_rate": 4.931713259810208e-05, + "loss": 0.6218, + "step": 1115 + }, + { + "epoch": 0.04, + "learning_rate": 4.931649140805335e-05, + "loss": 0.6085, + "step": 1116 + }, + { + "epoch": 0.04, + "learning_rate": 4.931585021800461e-05, + "loss": 0.6014, + "step": 1117 + }, + { + "epoch": 0.04, + "learning_rate": 4.931520902795589e-05, + "loss": 0.5863, + "step": 1118 + }, + { + "epoch": 0.04, + "learning_rate": 4.9314567837907164e-05, + "loss": 0.635, + "step": 1119 + }, + { + "epoch": 0.04, + "learning_rate": 4.931392664785843e-05, + "loss": 0.6832, + "step": 1120 + }, + { + "epoch": 0.04, + "learning_rate": 4.93132854578097e-05, + "loss": 0.656, + "step": 1121 + }, + { + "epoch": 0.04, + "learning_rate": 4.9312644267760965e-05, + "loss": 0.7034, + "step": 1122 + }, + { + "epoch": 0.04, + "learning_rate": 4.931200307771224e-05, + "loss": 0.6, + "step": 1123 + }, + { + "epoch": 0.04, + "learning_rate": 4.93113618876635e-05, + "loss": 0.6522, + "step": 1124 + }, + { + "epoch": 0.04, + "learning_rate": 4.9310720697614774e-05, + "loss": 0.5999, + "step": 1125 + }, + { + "epoch": 0.04, + "learning_rate": 4.9310079507566045e-05, + "loss": 0.7038, + "step": 1126 + }, + { + "epoch": 0.04, + "learning_rate": 4.930943831751732e-05, + "loss": 0.616, + "step": 1127 + }, + { + "epoch": 0.04, + "learning_rate": 4.930879712746859e-05, + "loss": 0.6541, + "step": 1128 + }, + { + "epoch": 0.04, + "learning_rate": 4.9308155937419854e-05, + "loss": 0.6738, + "step": 1129 + }, + { + "epoch": 0.04, + "learning_rate": 4.9307514747371125e-05, + "loss": 0.6189, + "step": 1130 + }, + { + "epoch": 0.04, + "learning_rate": 4.930687355732239e-05, + "loss": 0.6934, + "step": 1131 + }, + { + "epoch": 0.04, + "learning_rate": 4.930623236727366e-05, + "loss": 0.623, + "step": 1132 + }, + { + "epoch": 0.04, + "learning_rate": 4.930559117722493e-05, + "loss": 0.6856, + "step": 1133 + }, + { + "epoch": 0.04, + "learning_rate": 4.9304949987176205e-05, + "loss": 0.6286, + "step": 1134 + }, + { + "epoch": 0.04, + "learning_rate": 4.930430879712747e-05, + "loss": 0.6758, + "step": 1135 + }, + { + "epoch": 0.04, + "learning_rate": 4.930366760707874e-05, + "loss": 0.6318, + "step": 1136 + }, + { + "epoch": 0.04, + "learning_rate": 4.9303026417030014e-05, + "loss": 0.649, + "step": 1137 + }, + { + "epoch": 0.04, + "learning_rate": 4.930238522698128e-05, + "loss": 0.6923, + "step": 1138 + }, + { + "epoch": 0.04, + "learning_rate": 4.930174403693255e-05, + "loss": 0.6717, + "step": 1139 + }, + { + "epoch": 0.04, + "learning_rate": 4.9301102846883815e-05, + "loss": 0.6491, + "step": 1140 + }, + { + "epoch": 0.04, + "learning_rate": 4.930046165683509e-05, + "loss": 0.6663, + "step": 1141 + }, + { + "epoch": 0.04, + "learning_rate": 4.929982046678636e-05, + "loss": 0.6723, + "step": 1142 + }, + { + "epoch": 0.04, + "learning_rate": 4.929917927673763e-05, + "loss": 0.5857, + "step": 1143 + }, + { + "epoch": 0.04, + "learning_rate": 4.9298538086688895e-05, + "loss": 0.635, + "step": 1144 + }, + { + "epoch": 0.04, + "learning_rate": 4.929789689664017e-05, + "loss": 0.5941, + "step": 1145 + }, + { + "epoch": 0.04, + "learning_rate": 4.929725570659144e-05, + "loss": 0.6931, + "step": 1146 + }, + { + "epoch": 0.04, + "learning_rate": 4.9296614516542704e-05, + "loss": 0.6619, + "step": 1147 + }, + { + "epoch": 0.04, + "learning_rate": 4.9295973326493976e-05, + "loss": 0.6134, + "step": 1148 + }, + { + "epoch": 0.04, + "learning_rate": 4.929533213644524e-05, + "loss": 0.6593, + "step": 1149 + }, + { + "epoch": 0.04, + "learning_rate": 4.929469094639652e-05, + "loss": 0.6358, + "step": 1150 + }, + { + "epoch": 0.04, + "learning_rate": 4.9294049756347784e-05, + "loss": 0.6118, + "step": 1151 + }, + { + "epoch": 0.04, + "learning_rate": 4.9293408566299056e-05, + "loss": 0.6388, + "step": 1152 + }, + { + "epoch": 0.04, + "learning_rate": 4.929276737625032e-05, + "loss": 0.6084, + "step": 1153 + }, + { + "epoch": 0.04, + "learning_rate": 4.929212618620159e-05, + "loss": 0.6638, + "step": 1154 + }, + { + "epoch": 0.04, + "learning_rate": 4.9291484996152864e-05, + "loss": 0.6678, + "step": 1155 + }, + { + "epoch": 0.04, + "learning_rate": 4.929084380610413e-05, + "loss": 0.6467, + "step": 1156 + }, + { + "epoch": 0.04, + "learning_rate": 4.92902026160554e-05, + "loss": 0.6449, + "step": 1157 + }, + { + "epoch": 0.04, + "learning_rate": 4.928956142600667e-05, + "loss": 0.5494, + "step": 1158 + }, + { + "epoch": 0.04, + "learning_rate": 4.9288920235957944e-05, + "loss": 0.6831, + "step": 1159 + }, + { + "epoch": 0.04, + "learning_rate": 4.928827904590921e-05, + "loss": 0.672, + "step": 1160 + }, + { + "epoch": 0.04, + "learning_rate": 4.928763785586048e-05, + "loss": 0.6409, + "step": 1161 + }, + { + "epoch": 0.04, + "learning_rate": 4.9286996665811746e-05, + "loss": 0.5729, + "step": 1162 + }, + { + "epoch": 0.04, + "learning_rate": 4.928635547576302e-05, + "loss": 0.6389, + "step": 1163 + }, + { + "epoch": 0.04, + "learning_rate": 4.928571428571429e-05, + "loss": 0.6022, + "step": 1164 + }, + { + "epoch": 0.04, + "learning_rate": 4.9285073095665554e-05, + "loss": 0.6287, + "step": 1165 + }, + { + "epoch": 0.04, + "learning_rate": 4.9284431905616826e-05, + "loss": 0.693, + "step": 1166 + }, + { + "epoch": 0.04, + "learning_rate": 4.92837907155681e-05, + "loss": 0.745, + "step": 1167 + }, + { + "epoch": 0.04, + "learning_rate": 4.928314952551937e-05, + "loss": 0.635, + "step": 1168 + }, + { + "epoch": 0.04, + "learning_rate": 4.9282508335470634e-05, + "loss": 0.5731, + "step": 1169 + }, + { + "epoch": 0.04, + "learning_rate": 4.9281867145421906e-05, + "loss": 0.6408, + "step": 1170 + }, + { + "epoch": 0.05, + "learning_rate": 4.928122595537317e-05, + "loss": 0.6551, + "step": 1171 + }, + { + "epoch": 0.05, + "learning_rate": 4.928058476532444e-05, + "loss": 0.639, + "step": 1172 + }, + { + "epoch": 0.05, + "learning_rate": 4.9279943575275714e-05, + "loss": 0.6585, + "step": 1173 + }, + { + "epoch": 0.05, + "learning_rate": 4.9279302385226986e-05, + "loss": 0.6089, + "step": 1174 + }, + { + "epoch": 0.05, + "learning_rate": 4.927866119517826e-05, + "loss": 0.6626, + "step": 1175 + }, + { + "epoch": 0.05, + "learning_rate": 4.927802000512952e-05, + "loss": 0.6871, + "step": 1176 + }, + { + "epoch": 0.05, + "learning_rate": 4.9277378815080794e-05, + "loss": 0.6376, + "step": 1177 + }, + { + "epoch": 0.05, + "learning_rate": 4.927673762503206e-05, + "loss": 0.6115, + "step": 1178 + }, + { + "epoch": 0.05, + "learning_rate": 4.927609643498333e-05, + "loss": 0.6276, + "step": 1179 + }, + { + "epoch": 0.05, + "learning_rate": 4.9275455244934596e-05, + "loss": 0.6519, + "step": 1180 + }, + { + "epoch": 0.05, + "learning_rate": 4.927481405488587e-05, + "loss": 0.6937, + "step": 1181 + }, + { + "epoch": 0.05, + "learning_rate": 4.927417286483714e-05, + "loss": 0.5894, + "step": 1182 + }, + { + "epoch": 0.05, + "learning_rate": 4.927353167478841e-05, + "loss": 0.6057, + "step": 1183 + }, + { + "epoch": 0.05, + "learning_rate": 4.927289048473968e-05, + "loss": 0.5605, + "step": 1184 + }, + { + "epoch": 0.05, + "learning_rate": 4.927224929469095e-05, + "loss": 0.6183, + "step": 1185 + }, + { + "epoch": 0.05, + "learning_rate": 4.927160810464222e-05, + "loss": 0.6345, + "step": 1186 + }, + { + "epoch": 0.05, + "learning_rate": 4.9270966914593484e-05, + "loss": 0.6178, + "step": 1187 + }, + { + "epoch": 0.05, + "learning_rate": 4.9270325724544756e-05, + "loss": 0.6586, + "step": 1188 + }, + { + "epoch": 0.05, + "learning_rate": 4.926968453449603e-05, + "loss": 0.705, + "step": 1189 + }, + { + "epoch": 0.05, + "learning_rate": 4.92690433444473e-05, + "loss": 0.6513, + "step": 1190 + }, + { + "epoch": 0.05, + "learning_rate": 4.926840215439857e-05, + "loss": 0.5945, + "step": 1191 + }, + { + "epoch": 0.05, + "learning_rate": 4.9267760964349836e-05, + "loss": 0.6358, + "step": 1192 + }, + { + "epoch": 0.05, + "learning_rate": 4.926711977430111e-05, + "loss": 0.6028, + "step": 1193 + }, + { + "epoch": 0.05, + "learning_rate": 4.926647858425237e-05, + "loss": 0.5721, + "step": 1194 + }, + { + "epoch": 0.05, + "learning_rate": 4.9265837394203645e-05, + "loss": 0.5724, + "step": 1195 + }, + { + "epoch": 0.05, + "learning_rate": 4.926519620415491e-05, + "loss": 0.6459, + "step": 1196 + }, + { + "epoch": 0.05, + "learning_rate": 4.926455501410618e-05, + "loss": 0.7056, + "step": 1197 + }, + { + "epoch": 0.05, + "learning_rate": 4.926391382405745e-05, + "loss": 0.6596, + "step": 1198 + }, + { + "epoch": 0.05, + "learning_rate": 4.9263272634008725e-05, + "loss": 0.6196, + "step": 1199 + }, + { + "epoch": 0.05, + "learning_rate": 4.9262631443959996e-05, + "loss": 0.7047, + "step": 1200 + }, + { + "epoch": 0.05, + "learning_rate": 4.926199025391126e-05, + "loss": 0.69, + "step": 1201 + }, + { + "epoch": 0.05, + "learning_rate": 4.926134906386253e-05, + "loss": 0.6427, + "step": 1202 + }, + { + "epoch": 0.05, + "learning_rate": 4.92607078738138e-05, + "loss": 0.734, + "step": 1203 + }, + { + "epoch": 0.05, + "learning_rate": 4.926006668376507e-05, + "loss": 0.5632, + "step": 1204 + }, + { + "epoch": 0.05, + "learning_rate": 4.9259425493716335e-05, + "loss": 0.6676, + "step": 1205 + }, + { + "epoch": 0.05, + "learning_rate": 4.925878430366761e-05, + "loss": 0.5978, + "step": 1206 + }, + { + "epoch": 0.05, + "learning_rate": 4.925814311361888e-05, + "loss": 0.6823, + "step": 1207 + }, + { + "epoch": 0.05, + "learning_rate": 4.925750192357015e-05, + "loss": 0.6469, + "step": 1208 + }, + { + "epoch": 0.05, + "learning_rate": 4.925686073352142e-05, + "loss": 0.6398, + "step": 1209 + }, + { + "epoch": 0.05, + "learning_rate": 4.9256219543472686e-05, + "loss": 0.5871, + "step": 1210 + }, + { + "epoch": 0.05, + "learning_rate": 4.925557835342396e-05, + "loss": 0.6135, + "step": 1211 + }, + { + "epoch": 0.05, + "learning_rate": 4.925493716337522e-05, + "loss": 0.5608, + "step": 1212 + }, + { + "epoch": 0.05, + "learning_rate": 4.9254295973326495e-05, + "loss": 0.6551, + "step": 1213 + }, + { + "epoch": 0.05, + "learning_rate": 4.9253654783277767e-05, + "loss": 0.6439, + "step": 1214 + }, + { + "epoch": 0.05, + "learning_rate": 4.925301359322904e-05, + "loss": 0.5674, + "step": 1215 + }, + { + "epoch": 0.05, + "learning_rate": 4.92523724031803e-05, + "loss": 0.588, + "step": 1216 + }, + { + "epoch": 0.05, + "learning_rate": 4.9251731213131575e-05, + "loss": 0.5982, + "step": 1217 + }, + { + "epoch": 0.05, + "learning_rate": 4.925109002308285e-05, + "loss": 0.6641, + "step": 1218 + }, + { + "epoch": 0.05, + "learning_rate": 4.925044883303411e-05, + "loss": 0.5723, + "step": 1219 + }, + { + "epoch": 0.05, + "learning_rate": 4.924980764298538e-05, + "loss": 0.658, + "step": 1220 + }, + { + "epoch": 0.05, + "learning_rate": 4.924916645293665e-05, + "loss": 0.6445, + "step": 1221 + }, + { + "epoch": 0.05, + "learning_rate": 4.924852526288793e-05, + "loss": 0.5902, + "step": 1222 + }, + { + "epoch": 0.05, + "learning_rate": 4.924788407283919e-05, + "loss": 0.6475, + "step": 1223 + }, + { + "epoch": 0.05, + "learning_rate": 4.9247242882790463e-05, + "loss": 0.6893, + "step": 1224 + }, + { + "epoch": 0.05, + "learning_rate": 4.924660169274173e-05, + "loss": 0.5852, + "step": 1225 + }, + { + "epoch": 0.05, + "learning_rate": 4.9245960502693e-05, + "loss": 0.5909, + "step": 1226 + }, + { + "epoch": 0.05, + "learning_rate": 4.924531931264427e-05, + "loss": 0.6027, + "step": 1227 + }, + { + "epoch": 0.05, + "learning_rate": 4.924467812259554e-05, + "loss": 0.6328, + "step": 1228 + }, + { + "epoch": 0.05, + "learning_rate": 4.924403693254681e-05, + "loss": 0.6724, + "step": 1229 + }, + { + "epoch": 0.05, + "learning_rate": 4.924339574249808e-05, + "loss": 0.6299, + "step": 1230 + }, + { + "epoch": 0.05, + "learning_rate": 4.924275455244935e-05, + "loss": 0.6725, + "step": 1231 + }, + { + "epoch": 0.05, + "learning_rate": 4.924211336240062e-05, + "loss": 0.5775, + "step": 1232 + }, + { + "epoch": 0.05, + "learning_rate": 4.924147217235189e-05, + "loss": 0.6055, + "step": 1233 + }, + { + "epoch": 0.05, + "learning_rate": 4.9240830982303153e-05, + "loss": 0.5909, + "step": 1234 + }, + { + "epoch": 0.05, + "learning_rate": 4.9240189792254425e-05, + "loss": 0.591, + "step": 1235 + }, + { + "epoch": 0.05, + "learning_rate": 4.92395486022057e-05, + "loss": 0.6522, + "step": 1236 + }, + { + "epoch": 0.05, + "learning_rate": 4.923890741215696e-05, + "loss": 0.5668, + "step": 1237 + }, + { + "epoch": 0.05, + "learning_rate": 4.9238266222108234e-05, + "loss": 0.5922, + "step": 1238 + }, + { + "epoch": 0.05, + "learning_rate": 4.9237625032059505e-05, + "loss": 0.6538, + "step": 1239 + }, + { + "epoch": 0.05, + "learning_rate": 4.923698384201078e-05, + "loss": 0.647, + "step": 1240 + }, + { + "epoch": 0.05, + "learning_rate": 4.923634265196204e-05, + "loss": 0.6113, + "step": 1241 + }, + { + "epoch": 0.05, + "learning_rate": 4.9235701461913314e-05, + "loss": 0.6678, + "step": 1242 + }, + { + "epoch": 0.05, + "learning_rate": 4.923506027186458e-05, + "loss": 0.6545, + "step": 1243 + }, + { + "epoch": 0.05, + "learning_rate": 4.923441908181585e-05, + "loss": 0.5631, + "step": 1244 + }, + { + "epoch": 0.05, + "learning_rate": 4.923377789176712e-05, + "loss": 0.6112, + "step": 1245 + }, + { + "epoch": 0.05, + "learning_rate": 4.9233136701718394e-05, + "loss": 0.6783, + "step": 1246 + }, + { + "epoch": 0.05, + "learning_rate": 4.9232495511669665e-05, + "loss": 0.5764, + "step": 1247 + }, + { + "epoch": 0.05, + "learning_rate": 4.923185432162093e-05, + "loss": 0.592, + "step": 1248 + }, + { + "epoch": 0.05, + "learning_rate": 4.92312131315722e-05, + "loss": 0.5283, + "step": 1249 + }, + { + "epoch": 0.05, + "learning_rate": 4.923057194152347e-05, + "loss": 0.5451, + "step": 1250 + }, + { + "epoch": 0.05, + "learning_rate": 4.922993075147474e-05, + "loss": 0.6145, + "step": 1251 + }, + { + "epoch": 0.05, + "learning_rate": 4.9229289561426004e-05, + "loss": 0.6176, + "step": 1252 + }, + { + "epoch": 0.05, + "learning_rate": 4.9228648371377275e-05, + "loss": 0.5813, + "step": 1253 + }, + { + "epoch": 0.05, + "learning_rate": 4.922800718132855e-05, + "loss": 0.6836, + "step": 1254 + }, + { + "epoch": 0.05, + "learning_rate": 4.922736599127982e-05, + "loss": 0.6281, + "step": 1255 + }, + { + "epoch": 0.05, + "learning_rate": 4.922672480123109e-05, + "loss": 0.623, + "step": 1256 + }, + { + "epoch": 0.05, + "learning_rate": 4.9226083611182355e-05, + "loss": 0.6133, + "step": 1257 + }, + { + "epoch": 0.05, + "learning_rate": 4.922544242113363e-05, + "loss": 0.7053, + "step": 1258 + }, + { + "epoch": 0.05, + "learning_rate": 4.922480123108489e-05, + "loss": 0.6016, + "step": 1259 + }, + { + "epoch": 0.05, + "learning_rate": 4.9224160041036164e-05, + "loss": 0.6518, + "step": 1260 + }, + { + "epoch": 0.05, + "learning_rate": 4.9223518850987436e-05, + "loss": 0.6177, + "step": 1261 + }, + { + "epoch": 0.05, + "learning_rate": 4.922287766093871e-05, + "loss": 0.587, + "step": 1262 + }, + { + "epoch": 0.05, + "learning_rate": 4.922223647088998e-05, + "loss": 0.5702, + "step": 1263 + }, + { + "epoch": 0.05, + "learning_rate": 4.9221595280841244e-05, + "loss": 0.6374, + "step": 1264 + }, + { + "epoch": 0.05, + "learning_rate": 4.9220954090792516e-05, + "loss": 0.6785, + "step": 1265 + }, + { + "epoch": 0.05, + "learning_rate": 4.922031290074378e-05, + "loss": 0.6157, + "step": 1266 + }, + { + "epoch": 0.05, + "learning_rate": 4.921967171069505e-05, + "loss": 0.6355, + "step": 1267 + }, + { + "epoch": 0.05, + "learning_rate": 4.921903052064632e-05, + "loss": 0.6323, + "step": 1268 + }, + { + "epoch": 0.05, + "learning_rate": 4.921838933059759e-05, + "loss": 0.5838, + "step": 1269 + }, + { + "epoch": 0.05, + "learning_rate": 4.921774814054886e-05, + "loss": 0.6125, + "step": 1270 + }, + { + "epoch": 0.05, + "learning_rate": 4.921710695050013e-05, + "loss": 0.6357, + "step": 1271 + }, + { + "epoch": 0.05, + "learning_rate": 4.9216465760451404e-05, + "loss": 0.6265, + "step": 1272 + }, + { + "epoch": 0.05, + "learning_rate": 4.921582457040267e-05, + "loss": 0.7485, + "step": 1273 + }, + { + "epoch": 0.05, + "learning_rate": 4.921518338035394e-05, + "loss": 0.6537, + "step": 1274 + }, + { + "epoch": 0.05, + "learning_rate": 4.9214542190305206e-05, + "loss": 0.6388, + "step": 1275 + }, + { + "epoch": 0.05, + "learning_rate": 4.921390100025648e-05, + "loss": 0.607, + "step": 1276 + }, + { + "epoch": 0.05, + "learning_rate": 4.921325981020775e-05, + "loss": 0.5697, + "step": 1277 + }, + { + "epoch": 0.05, + "learning_rate": 4.921261862015902e-05, + "loss": 0.6787, + "step": 1278 + }, + { + "epoch": 0.05, + "learning_rate": 4.9211977430110286e-05, + "loss": 0.641, + "step": 1279 + }, + { + "epoch": 0.05, + "learning_rate": 4.921133624006156e-05, + "loss": 0.5854, + "step": 1280 + }, + { + "epoch": 0.05, + "learning_rate": 4.921069505001283e-05, + "loss": 0.6134, + "step": 1281 + }, + { + "epoch": 0.05, + "learning_rate": 4.9210053859964094e-05, + "loss": 0.6512, + "step": 1282 + }, + { + "epoch": 0.05, + "learning_rate": 4.9209412669915366e-05, + "loss": 0.6616, + "step": 1283 + }, + { + "epoch": 0.05, + "learning_rate": 4.920877147986663e-05, + "loss": 0.6472, + "step": 1284 + }, + { + "epoch": 0.05, + "learning_rate": 4.92081302898179e-05, + "loss": 0.6214, + "step": 1285 + }, + { + "epoch": 0.05, + "learning_rate": 4.9207489099769174e-05, + "loss": 0.6634, + "step": 1286 + }, + { + "epoch": 0.05, + "learning_rate": 4.9206847909720446e-05, + "loss": 0.6288, + "step": 1287 + }, + { + "epoch": 0.05, + "learning_rate": 4.920620671967172e-05, + "loss": 0.6036, + "step": 1288 + }, + { + "epoch": 0.05, + "learning_rate": 4.920556552962298e-05, + "loss": 0.6674, + "step": 1289 + }, + { + "epoch": 0.05, + "learning_rate": 4.9204924339574254e-05, + "loss": 0.5812, + "step": 1290 + }, + { + "epoch": 0.05, + "learning_rate": 4.920428314952552e-05, + "loss": 0.6144, + "step": 1291 + }, + { + "epoch": 0.05, + "learning_rate": 4.920364195947679e-05, + "loss": 0.6196, + "step": 1292 + }, + { + "epoch": 0.05, + "learning_rate": 4.9203000769428056e-05, + "loss": 0.6584, + "step": 1293 + }, + { + "epoch": 0.05, + "learning_rate": 4.9202359579379334e-05, + "loss": 0.7248, + "step": 1294 + }, + { + "epoch": 0.05, + "learning_rate": 4.92017183893306e-05, + "loss": 0.6769, + "step": 1295 + }, + { + "epoch": 0.05, + "learning_rate": 4.920107719928187e-05, + "loss": 0.6086, + "step": 1296 + }, + { + "epoch": 0.05, + "learning_rate": 4.920043600923314e-05, + "loss": 0.5609, + "step": 1297 + }, + { + "epoch": 0.05, + "learning_rate": 4.919979481918441e-05, + "loss": 0.7002, + "step": 1298 + }, + { + "epoch": 0.05, + "learning_rate": 4.919915362913568e-05, + "loss": 0.6245, + "step": 1299 + }, + { + "epoch": 0.05, + "learning_rate": 4.9198512439086944e-05, + "loss": 0.6626, + "step": 1300 + }, + { + "epoch": 0.05, + "learning_rate": 4.9197871249038216e-05, + "loss": 0.6618, + "step": 1301 + }, + { + "epoch": 0.05, + "learning_rate": 4.919723005898949e-05, + "loss": 0.6489, + "step": 1302 + }, + { + "epoch": 0.05, + "learning_rate": 4.919658886894076e-05, + "loss": 0.6577, + "step": 1303 + }, + { + "epoch": 0.05, + "learning_rate": 4.9195947678892025e-05, + "loss": 0.659, + "step": 1304 + }, + { + "epoch": 0.05, + "learning_rate": 4.9195306488843296e-05, + "loss": 0.6283, + "step": 1305 + }, + { + "epoch": 0.05, + "learning_rate": 4.919466529879457e-05, + "loss": 0.7055, + "step": 1306 + }, + { + "epoch": 0.05, + "learning_rate": 4.919402410874583e-05, + "loss": 0.647, + "step": 1307 + }, + { + "epoch": 0.05, + "learning_rate": 4.9193382918697105e-05, + "loss": 0.622, + "step": 1308 + }, + { + "epoch": 0.05, + "learning_rate": 4.919274172864837e-05, + "loss": 0.6822, + "step": 1309 + }, + { + "epoch": 0.05, + "learning_rate": 4.919210053859965e-05, + "loss": 0.622, + "step": 1310 + }, + { + "epoch": 0.05, + "learning_rate": 4.919145934855091e-05, + "loss": 0.6557, + "step": 1311 + }, + { + "epoch": 0.05, + "learning_rate": 4.9190818158502185e-05, + "loss": 0.6302, + "step": 1312 + }, + { + "epoch": 0.05, + "learning_rate": 4.919017696845345e-05, + "loss": 0.6532, + "step": 1313 + }, + { + "epoch": 0.05, + "learning_rate": 4.918953577840472e-05, + "loss": 0.5987, + "step": 1314 + }, + { + "epoch": 0.05, + "learning_rate": 4.918889458835599e-05, + "loss": 0.6085, + "step": 1315 + }, + { + "epoch": 0.05, + "learning_rate": 4.918825339830726e-05, + "loss": 0.62, + "step": 1316 + }, + { + "epoch": 0.05, + "learning_rate": 4.918761220825853e-05, + "loss": 0.6285, + "step": 1317 + }, + { + "epoch": 0.05, + "learning_rate": 4.91869710182098e-05, + "loss": 0.6036, + "step": 1318 + }, + { + "epoch": 0.05, + "learning_rate": 4.918632982816107e-05, + "loss": 0.5829, + "step": 1319 + }, + { + "epoch": 0.05, + "learning_rate": 4.918568863811234e-05, + "loss": 0.5784, + "step": 1320 + }, + { + "epoch": 0.05, + "learning_rate": 4.918504744806361e-05, + "loss": 0.5855, + "step": 1321 + }, + { + "epoch": 0.05, + "learning_rate": 4.9184406258014875e-05, + "loss": 0.636, + "step": 1322 + }, + { + "epoch": 0.05, + "learning_rate": 4.9183765067966146e-05, + "loss": 0.6152, + "step": 1323 + }, + { + "epoch": 0.05, + "learning_rate": 4.918312387791742e-05, + "loss": 0.6826, + "step": 1324 + }, + { + "epoch": 0.05, + "learning_rate": 4.918248268786868e-05, + "loss": 0.6726, + "step": 1325 + }, + { + "epoch": 0.05, + "learning_rate": 4.9181841497819955e-05, + "loss": 0.5277, + "step": 1326 + }, + { + "epoch": 0.05, + "learning_rate": 4.9181200307771227e-05, + "loss": 0.5672, + "step": 1327 + }, + { + "epoch": 0.05, + "learning_rate": 4.91805591177225e-05, + "loss": 0.5618, + "step": 1328 + }, + { + "epoch": 0.05, + "learning_rate": 4.917991792767376e-05, + "loss": 0.6732, + "step": 1329 + }, + { + "epoch": 0.05, + "learning_rate": 4.9179276737625035e-05, + "loss": 0.6796, + "step": 1330 + }, + { + "epoch": 0.05, + "learning_rate": 4.91786355475763e-05, + "loss": 0.7051, + "step": 1331 + }, + { + "epoch": 0.05, + "learning_rate": 4.917799435752757e-05, + "loss": 0.7058, + "step": 1332 + }, + { + "epoch": 0.05, + "learning_rate": 4.917735316747884e-05, + "loss": 0.6149, + "step": 1333 + }, + { + "epoch": 0.05, + "learning_rate": 4.9176711977430115e-05, + "loss": 0.6348, + "step": 1334 + }, + { + "epoch": 0.05, + "learning_rate": 4.917607078738139e-05, + "loss": 0.6328, + "step": 1335 + }, + { + "epoch": 0.05, + "learning_rate": 4.917542959733265e-05, + "loss": 0.6617, + "step": 1336 + }, + { + "epoch": 0.05, + "learning_rate": 4.9174788407283923e-05, + "loss": 0.7131, + "step": 1337 + }, + { + "epoch": 0.05, + "learning_rate": 4.917414721723519e-05, + "loss": 0.6842, + "step": 1338 + }, + { + "epoch": 0.05, + "learning_rate": 4.917350602718646e-05, + "loss": 0.6455, + "step": 1339 + }, + { + "epoch": 0.05, + "learning_rate": 4.9172864837137725e-05, + "loss": 0.6419, + "step": 1340 + }, + { + "epoch": 0.05, + "learning_rate": 4.9172223647089e-05, + "loss": 0.6435, + "step": 1341 + }, + { + "epoch": 0.05, + "learning_rate": 4.917158245704027e-05, + "loss": 0.5967, + "step": 1342 + }, + { + "epoch": 0.05, + "learning_rate": 4.917094126699154e-05, + "loss": 0.6013, + "step": 1343 + }, + { + "epoch": 0.05, + "learning_rate": 4.917030007694281e-05, + "loss": 0.6068, + "step": 1344 + }, + { + "epoch": 0.05, + "learning_rate": 4.916965888689408e-05, + "loss": 0.6448, + "step": 1345 + }, + { + "epoch": 0.05, + "learning_rate": 4.916901769684535e-05, + "loss": 0.601, + "step": 1346 + }, + { + "epoch": 0.05, + "learning_rate": 4.9168376506796613e-05, + "loss": 0.617, + "step": 1347 + }, + { + "epoch": 0.05, + "learning_rate": 4.9167735316747885e-05, + "loss": 0.6143, + "step": 1348 + }, + { + "epoch": 0.05, + "learning_rate": 4.916709412669916e-05, + "loss": 0.6447, + "step": 1349 + }, + { + "epoch": 0.05, + "learning_rate": 4.916645293665043e-05, + "loss": 0.603, + "step": 1350 + }, + { + "epoch": 0.05, + "learning_rate": 4.91658117466017e-05, + "loss": 0.5887, + "step": 1351 + }, + { + "epoch": 0.05, + "learning_rate": 4.9165170556552965e-05, + "loss": 0.6288, + "step": 1352 + }, + { + "epoch": 0.05, + "learning_rate": 4.916452936650424e-05, + "loss": 0.6114, + "step": 1353 + }, + { + "epoch": 0.05, + "learning_rate": 4.91638881764555e-05, + "loss": 0.5982, + "step": 1354 + }, + { + "epoch": 0.05, + "learning_rate": 4.9163246986406774e-05, + "loss": 0.608, + "step": 1355 + }, + { + "epoch": 0.05, + "learning_rate": 4.916260579635804e-05, + "loss": 0.631, + "step": 1356 + }, + { + "epoch": 0.05, + "learning_rate": 4.916196460630931e-05, + "loss": 0.6505, + "step": 1357 + }, + { + "epoch": 0.05, + "learning_rate": 4.916132341626058e-05, + "loss": 0.6624, + "step": 1358 + }, + { + "epoch": 0.05, + "learning_rate": 4.9160682226211854e-05, + "loss": 0.6554, + "step": 1359 + }, + { + "epoch": 0.05, + "learning_rate": 4.9160041036163125e-05, + "loss": 0.6429, + "step": 1360 + }, + { + "epoch": 0.05, + "learning_rate": 4.915939984611439e-05, + "loss": 0.6293, + "step": 1361 + }, + { + "epoch": 0.05, + "learning_rate": 4.915875865606566e-05, + "loss": 0.57, + "step": 1362 + }, + { + "epoch": 0.05, + "learning_rate": 4.915811746601693e-05, + "loss": 0.6667, + "step": 1363 + }, + { + "epoch": 0.05, + "learning_rate": 4.91574762759682e-05, + "loss": 0.5903, + "step": 1364 + }, + { + "epoch": 0.05, + "learning_rate": 4.9156835085919464e-05, + "loss": 0.6208, + "step": 1365 + }, + { + "epoch": 0.05, + "learning_rate": 4.915619389587074e-05, + "loss": 0.6872, + "step": 1366 + }, + { + "epoch": 0.05, + "learning_rate": 4.915555270582201e-05, + "loss": 0.6062, + "step": 1367 + }, + { + "epoch": 0.05, + "learning_rate": 4.915491151577328e-05, + "loss": 0.6548, + "step": 1368 + }, + { + "epoch": 0.05, + "learning_rate": 4.915427032572455e-05, + "loss": 0.57, + "step": 1369 + }, + { + "epoch": 0.05, + "learning_rate": 4.9153629135675816e-05, + "loss": 0.595, + "step": 1370 + }, + { + "epoch": 0.05, + "learning_rate": 4.915298794562709e-05, + "loss": 0.586, + "step": 1371 + }, + { + "epoch": 0.05, + "learning_rate": 4.915234675557835e-05, + "loss": 0.5957, + "step": 1372 + }, + { + "epoch": 0.05, + "learning_rate": 4.9151705565529624e-05, + "loss": 0.6159, + "step": 1373 + }, + { + "epoch": 0.05, + "learning_rate": 4.9151064375480896e-05, + "loss": 0.5759, + "step": 1374 + }, + { + "epoch": 0.05, + "learning_rate": 4.915042318543217e-05, + "loss": 0.6943, + "step": 1375 + }, + { + "epoch": 0.05, + "learning_rate": 4.914978199538343e-05, + "loss": 0.7015, + "step": 1376 + }, + { + "epoch": 0.05, + "learning_rate": 4.9149140805334704e-05, + "loss": 0.5499, + "step": 1377 + }, + { + "epoch": 0.05, + "learning_rate": 4.9148499615285976e-05, + "loss": 0.5437, + "step": 1378 + }, + { + "epoch": 0.05, + "learning_rate": 4.914785842523724e-05, + "loss": 0.5856, + "step": 1379 + }, + { + "epoch": 0.05, + "learning_rate": 4.914721723518851e-05, + "loss": 0.5968, + "step": 1380 + }, + { + "epoch": 0.05, + "learning_rate": 4.914657604513978e-05, + "loss": 0.7115, + "step": 1381 + }, + { + "epoch": 0.05, + "learning_rate": 4.9145934855091056e-05, + "loss": 0.6926, + "step": 1382 + }, + { + "epoch": 0.05, + "learning_rate": 4.914529366504232e-05, + "loss": 0.5951, + "step": 1383 + }, + { + "epoch": 0.05, + "learning_rate": 4.914465247499359e-05, + "loss": 0.6251, + "step": 1384 + }, + { + "epoch": 0.05, + "learning_rate": 4.914401128494486e-05, + "loss": 0.6082, + "step": 1385 + }, + { + "epoch": 0.05, + "learning_rate": 4.914337009489613e-05, + "loss": 0.6378, + "step": 1386 + }, + { + "epoch": 0.05, + "learning_rate": 4.91427289048474e-05, + "loss": 0.6871, + "step": 1387 + }, + { + "epoch": 0.05, + "learning_rate": 4.9142087714798666e-05, + "loss": 0.6214, + "step": 1388 + }, + { + "epoch": 0.05, + "learning_rate": 4.914144652474994e-05, + "loss": 0.6322, + "step": 1389 + }, + { + "epoch": 0.05, + "learning_rate": 4.914080533470121e-05, + "loss": 0.6372, + "step": 1390 + }, + { + "epoch": 0.05, + "learning_rate": 4.914016414465248e-05, + "loss": 0.565, + "step": 1391 + }, + { + "epoch": 0.05, + "learning_rate": 4.9139522954603746e-05, + "loss": 0.6815, + "step": 1392 + }, + { + "epoch": 0.05, + "learning_rate": 4.913888176455502e-05, + "loss": 0.6708, + "step": 1393 + }, + { + "epoch": 0.05, + "learning_rate": 4.913824057450628e-05, + "loss": 0.5909, + "step": 1394 + }, + { + "epoch": 0.05, + "learning_rate": 4.9137599384457554e-05, + "loss": 0.6585, + "step": 1395 + }, + { + "epoch": 0.05, + "learning_rate": 4.9136958194408826e-05, + "loss": 0.622, + "step": 1396 + }, + { + "epoch": 0.05, + "learning_rate": 4.913631700436009e-05, + "loss": 0.6626, + "step": 1397 + }, + { + "epoch": 0.05, + "learning_rate": 4.913567581431137e-05, + "loss": 0.628, + "step": 1398 + }, + { + "epoch": 0.05, + "learning_rate": 4.9135034624262634e-05, + "loss": 0.6432, + "step": 1399 + }, + { + "epoch": 0.05, + "learning_rate": 4.9134393434213906e-05, + "loss": 0.6189, + "step": 1400 + }, + { + "epoch": 0.05, + "learning_rate": 4.913375224416517e-05, + "loss": 0.6339, + "step": 1401 + }, + { + "epoch": 0.05, + "learning_rate": 4.913311105411644e-05, + "loss": 0.6565, + "step": 1402 + }, + { + "epoch": 0.05, + "learning_rate": 4.913246986406771e-05, + "loss": 0.6029, + "step": 1403 + }, + { + "epoch": 0.05, + "learning_rate": 4.913182867401898e-05, + "loss": 0.6654, + "step": 1404 + }, + { + "epoch": 0.05, + "learning_rate": 4.913118748397025e-05, + "loss": 0.597, + "step": 1405 + }, + { + "epoch": 0.05, + "learning_rate": 4.913054629392152e-05, + "loss": 0.6455, + "step": 1406 + }, + { + "epoch": 0.05, + "learning_rate": 4.9129905103872795e-05, + "loss": 0.6752, + "step": 1407 + }, + { + "epoch": 0.05, + "learning_rate": 4.912926391382406e-05, + "loss": 0.5796, + "step": 1408 + }, + { + "epoch": 0.05, + "learning_rate": 4.912862272377533e-05, + "loss": 0.6874, + "step": 1409 + }, + { + "epoch": 0.05, + "learning_rate": 4.9127981533726596e-05, + "loss": 0.6278, + "step": 1410 + }, + { + "epoch": 0.05, + "learning_rate": 4.912734034367787e-05, + "loss": 0.5879, + "step": 1411 + }, + { + "epoch": 0.05, + "learning_rate": 4.912669915362913e-05, + "loss": 0.6165, + "step": 1412 + }, + { + "epoch": 0.05, + "learning_rate": 4.9126057963580404e-05, + "loss": 0.6514, + "step": 1413 + }, + { + "epoch": 0.05, + "learning_rate": 4.9125416773531676e-05, + "loss": 0.6353, + "step": 1414 + }, + { + "epoch": 0.05, + "learning_rate": 4.912477558348295e-05, + "loss": 0.6035, + "step": 1415 + }, + { + "epoch": 0.05, + "learning_rate": 4.912413439343422e-05, + "loss": 0.6358, + "step": 1416 + }, + { + "epoch": 0.05, + "learning_rate": 4.9123493203385485e-05, + "loss": 0.6145, + "step": 1417 + }, + { + "epoch": 0.05, + "learning_rate": 4.9122852013336756e-05, + "loss": 0.6515, + "step": 1418 + }, + { + "epoch": 0.05, + "learning_rate": 4.912221082328802e-05, + "loss": 0.6028, + "step": 1419 + }, + { + "epoch": 0.05, + "learning_rate": 4.912156963323929e-05, + "loss": 0.6653, + "step": 1420 + }, + { + "epoch": 0.05, + "learning_rate": 4.9120928443190565e-05, + "loss": 0.621, + "step": 1421 + }, + { + "epoch": 0.05, + "learning_rate": 4.9120287253141836e-05, + "loss": 0.6455, + "step": 1422 + }, + { + "epoch": 0.05, + "learning_rate": 4.911964606309311e-05, + "loss": 0.6475, + "step": 1423 + }, + { + "epoch": 0.05, + "learning_rate": 4.911900487304437e-05, + "loss": 0.6359, + "step": 1424 + }, + { + "epoch": 0.05, + "learning_rate": 4.9118363682995645e-05, + "loss": 0.6438, + "step": 1425 + }, + { + "epoch": 0.05, + "learning_rate": 4.911772249294691e-05, + "loss": 0.6579, + "step": 1426 + }, + { + "epoch": 0.05, + "learning_rate": 4.911708130289818e-05, + "loss": 0.5775, + "step": 1427 + }, + { + "epoch": 0.05, + "learning_rate": 4.9116440112849446e-05, + "loss": 0.5902, + "step": 1428 + }, + { + "epoch": 0.05, + "learning_rate": 4.911579892280072e-05, + "loss": 0.6136, + "step": 1429 + }, + { + "epoch": 0.05, + "learning_rate": 4.911515773275199e-05, + "loss": 0.5889, + "step": 1430 + }, + { + "epoch": 0.06, + "learning_rate": 4.911451654270326e-05, + "loss": 0.6635, + "step": 1431 + }, + { + "epoch": 0.06, + "learning_rate": 4.911387535265453e-05, + "loss": 0.5915, + "step": 1432 + }, + { + "epoch": 0.06, + "learning_rate": 4.91132341626058e-05, + "loss": 0.6214, + "step": 1433 + }, + { + "epoch": 0.06, + "learning_rate": 4.911259297255707e-05, + "loss": 0.6529, + "step": 1434 + }, + { + "epoch": 0.06, + "learning_rate": 4.9111951782508335e-05, + "loss": 0.5913, + "step": 1435 + }, + { + "epoch": 0.06, + "learning_rate": 4.9111310592459607e-05, + "loss": 0.6892, + "step": 1436 + }, + { + "epoch": 0.06, + "learning_rate": 4.911066940241088e-05, + "loss": 0.6186, + "step": 1437 + }, + { + "epoch": 0.06, + "learning_rate": 4.911002821236215e-05, + "loss": 0.6913, + "step": 1438 + }, + { + "epoch": 0.06, + "learning_rate": 4.9109387022313415e-05, + "loss": 0.609, + "step": 1439 + }, + { + "epoch": 0.06, + "learning_rate": 4.910874583226469e-05, + "loss": 0.6419, + "step": 1440 + }, + { + "epoch": 0.06, + "learning_rate": 4.910810464221596e-05, + "loss": 0.6323, + "step": 1441 + }, + { + "epoch": 0.06, + "learning_rate": 4.910746345216722e-05, + "loss": 0.6244, + "step": 1442 + }, + { + "epoch": 0.06, + "learning_rate": 4.9106822262118495e-05, + "loss": 0.637, + "step": 1443 + }, + { + "epoch": 0.06, + "learning_rate": 4.910618107206976e-05, + "loss": 0.6946, + "step": 1444 + }, + { + "epoch": 0.06, + "learning_rate": 4.910553988202103e-05, + "loss": 0.5923, + "step": 1445 + }, + { + "epoch": 0.06, + "learning_rate": 4.91048986919723e-05, + "loss": 0.6135, + "step": 1446 + }, + { + "epoch": 0.06, + "learning_rate": 4.9104257501923575e-05, + "loss": 0.6177, + "step": 1447 + }, + { + "epoch": 0.06, + "learning_rate": 4.910361631187484e-05, + "loss": 0.7206, + "step": 1448 + }, + { + "epoch": 0.06, + "learning_rate": 4.910297512182611e-05, + "loss": 0.6419, + "step": 1449 + }, + { + "epoch": 0.06, + "learning_rate": 4.9102333931777383e-05, + "loss": 0.6737, + "step": 1450 + }, + { + "epoch": 0.06, + "learning_rate": 4.910169274172865e-05, + "loss": 0.6063, + "step": 1451 + }, + { + "epoch": 0.06, + "learning_rate": 4.910105155167992e-05, + "loss": 0.6647, + "step": 1452 + }, + { + "epoch": 0.06, + "learning_rate": 4.9100410361631185e-05, + "loss": 0.6421, + "step": 1453 + }, + { + "epoch": 0.06, + "learning_rate": 4.9099769171582464e-05, + "loss": 0.567, + "step": 1454 + }, + { + "epoch": 0.06, + "learning_rate": 4.909912798153373e-05, + "loss": 0.5998, + "step": 1455 + }, + { + "epoch": 0.06, + "learning_rate": 4.9098486791485e-05, + "loss": 0.5743, + "step": 1456 + }, + { + "epoch": 0.06, + "learning_rate": 4.9097845601436265e-05, + "loss": 0.573, + "step": 1457 + }, + { + "epoch": 0.06, + "learning_rate": 4.909720441138754e-05, + "loss": 0.5804, + "step": 1458 + }, + { + "epoch": 0.06, + "learning_rate": 4.909656322133881e-05, + "loss": 0.6175, + "step": 1459 + }, + { + "epoch": 0.06, + "learning_rate": 4.9095922031290074e-05, + "loss": 0.6146, + "step": 1460 + }, + { + "epoch": 0.06, + "learning_rate": 4.9095280841241345e-05, + "loss": 0.6359, + "step": 1461 + }, + { + "epoch": 0.06, + "learning_rate": 4.909463965119262e-05, + "loss": 0.5769, + "step": 1462 + }, + { + "epoch": 0.06, + "learning_rate": 4.909399846114389e-05, + "loss": 0.705, + "step": 1463 + }, + { + "epoch": 0.06, + "learning_rate": 4.9093357271095154e-05, + "loss": 0.6578, + "step": 1464 + }, + { + "epoch": 0.06, + "learning_rate": 4.9092716081046425e-05, + "loss": 0.5924, + "step": 1465 + }, + { + "epoch": 0.06, + "learning_rate": 4.909207489099769e-05, + "loss": 0.5688, + "step": 1466 + }, + { + "epoch": 0.06, + "learning_rate": 4.909143370094896e-05, + "loss": 0.585, + "step": 1467 + }, + { + "epoch": 0.06, + "learning_rate": 4.9090792510900234e-05, + "loss": 0.5521, + "step": 1468 + }, + { + "epoch": 0.06, + "learning_rate": 4.90901513208515e-05, + "loss": 0.62, + "step": 1469 + }, + { + "epoch": 0.06, + "learning_rate": 4.908951013080278e-05, + "loss": 0.5435, + "step": 1470 + }, + { + "epoch": 0.06, + "learning_rate": 4.908886894075404e-05, + "loss": 0.6618, + "step": 1471 + }, + { + "epoch": 0.06, + "learning_rate": 4.9088227750705314e-05, + "loss": 0.644, + "step": 1472 + }, + { + "epoch": 0.06, + "learning_rate": 4.908758656065658e-05, + "loss": 0.6348, + "step": 1473 + }, + { + "epoch": 0.06, + "learning_rate": 4.908694537060785e-05, + "loss": 0.5485, + "step": 1474 + }, + { + "epoch": 0.06, + "learning_rate": 4.908630418055912e-05, + "loss": 0.6092, + "step": 1475 + }, + { + "epoch": 0.06, + "learning_rate": 4.908566299051039e-05, + "loss": 0.5486, + "step": 1476 + }, + { + "epoch": 0.06, + "learning_rate": 4.908502180046166e-05, + "loss": 0.5601, + "step": 1477 + }, + { + "epoch": 0.06, + "learning_rate": 4.908438061041293e-05, + "loss": 0.7027, + "step": 1478 + }, + { + "epoch": 0.06, + "learning_rate": 4.90837394203642e-05, + "loss": 0.7096, + "step": 1479 + }, + { + "epoch": 0.06, + "learning_rate": 4.908309823031547e-05, + "loss": 0.7466, + "step": 1480 + }, + { + "epoch": 0.06, + "learning_rate": 4.908245704026674e-05, + "loss": 0.5599, + "step": 1481 + }, + { + "epoch": 0.06, + "learning_rate": 4.9081815850218004e-05, + "loss": 0.6452, + "step": 1482 + }, + { + "epoch": 0.06, + "learning_rate": 4.9081174660169276e-05, + "loss": 0.6201, + "step": 1483 + }, + { + "epoch": 0.06, + "learning_rate": 4.908053347012055e-05, + "loss": 0.6781, + "step": 1484 + }, + { + "epoch": 0.06, + "learning_rate": 4.907989228007181e-05, + "loss": 0.6665, + "step": 1485 + }, + { + "epoch": 0.06, + "learning_rate": 4.9079251090023084e-05, + "loss": 0.6057, + "step": 1486 + }, + { + "epoch": 0.06, + "learning_rate": 4.9078609899974356e-05, + "loss": 0.6086, + "step": 1487 + }, + { + "epoch": 0.06, + "learning_rate": 4.907796870992563e-05, + "loss": 0.5958, + "step": 1488 + }, + { + "epoch": 0.06, + "learning_rate": 4.907732751987689e-05, + "loss": 0.5324, + "step": 1489 + }, + { + "epoch": 0.06, + "learning_rate": 4.9076686329828164e-05, + "loss": 0.7078, + "step": 1490 + }, + { + "epoch": 0.06, + "learning_rate": 4.907604513977943e-05, + "loss": 0.6593, + "step": 1491 + }, + { + "epoch": 0.06, + "learning_rate": 4.90754039497307e-05, + "loss": 0.5809, + "step": 1492 + }, + { + "epoch": 0.06, + "learning_rate": 4.907476275968197e-05, + "loss": 0.6722, + "step": 1493 + }, + { + "epoch": 0.06, + "learning_rate": 4.9074121569633244e-05, + "loss": 0.6252, + "step": 1494 + }, + { + "epoch": 0.06, + "learning_rate": 4.9073480379584516e-05, + "loss": 0.6895, + "step": 1495 + }, + { + "epoch": 0.06, + "learning_rate": 4.907283918953578e-05, + "loss": 0.6776, + "step": 1496 + }, + { + "epoch": 0.06, + "learning_rate": 4.907219799948705e-05, + "loss": 0.6233, + "step": 1497 + }, + { + "epoch": 0.06, + "learning_rate": 4.907155680943832e-05, + "loss": 0.6746, + "step": 1498 + }, + { + "epoch": 0.06, + "learning_rate": 4.907091561938959e-05, + "loss": 0.6212, + "step": 1499 + }, + { + "epoch": 0.06, + "learning_rate": 4.9070274429340854e-05, + "loss": 0.6253, + "step": 1500 + }, + { + "epoch": 0.06, + "learning_rate": 4.9069633239292126e-05, + "loss": 0.6754, + "step": 1501 + }, + { + "epoch": 0.06, + "learning_rate": 4.90689920492434e-05, + "loss": 0.6604, + "step": 1502 + }, + { + "epoch": 0.06, + "learning_rate": 4.906835085919467e-05, + "loss": 0.6508, + "step": 1503 + }, + { + "epoch": 0.06, + "learning_rate": 4.906770966914594e-05, + "loss": 0.627, + "step": 1504 + }, + { + "epoch": 0.06, + "learning_rate": 4.9067068479097206e-05, + "loss": 0.6426, + "step": 1505 + }, + { + "epoch": 0.06, + "learning_rate": 4.906642728904848e-05, + "loss": 0.6868, + "step": 1506 + }, + { + "epoch": 0.06, + "learning_rate": 4.906578609899974e-05, + "loss": 0.6397, + "step": 1507 + }, + { + "epoch": 0.06, + "learning_rate": 4.9065144908951014e-05, + "loss": 0.6338, + "step": 1508 + }, + { + "epoch": 0.06, + "learning_rate": 4.9064503718902286e-05, + "loss": 0.6773, + "step": 1509 + }, + { + "epoch": 0.06, + "learning_rate": 4.906386252885356e-05, + "loss": 0.6633, + "step": 1510 + }, + { + "epoch": 0.06, + "learning_rate": 4.906322133880483e-05, + "loss": 0.6243, + "step": 1511 + }, + { + "epoch": 0.06, + "learning_rate": 4.9062580148756094e-05, + "loss": 0.5949, + "step": 1512 + }, + { + "epoch": 0.06, + "learning_rate": 4.9061938958707366e-05, + "loss": 0.5645, + "step": 1513 + }, + { + "epoch": 0.06, + "learning_rate": 4.906129776865863e-05, + "loss": 0.6767, + "step": 1514 + }, + { + "epoch": 0.06, + "learning_rate": 4.90606565786099e-05, + "loss": 0.6083, + "step": 1515 + }, + { + "epoch": 0.06, + "learning_rate": 4.906001538856117e-05, + "loss": 0.5816, + "step": 1516 + }, + { + "epoch": 0.06, + "learning_rate": 4.905937419851244e-05, + "loss": 0.5912, + "step": 1517 + }, + { + "epoch": 0.06, + "learning_rate": 4.905873300846371e-05, + "loss": 0.6104, + "step": 1518 + }, + { + "epoch": 0.06, + "learning_rate": 4.905809181841498e-05, + "loss": 0.6706, + "step": 1519 + }, + { + "epoch": 0.06, + "learning_rate": 4.9057450628366255e-05, + "loss": 0.6456, + "step": 1520 + }, + { + "epoch": 0.06, + "learning_rate": 4.905680943831752e-05, + "loss": 0.6168, + "step": 1521 + }, + { + "epoch": 0.06, + "learning_rate": 4.905616824826879e-05, + "loss": 0.6606, + "step": 1522 + }, + { + "epoch": 0.06, + "learning_rate": 4.9055527058220056e-05, + "loss": 0.6383, + "step": 1523 + }, + { + "epoch": 0.06, + "learning_rate": 4.905488586817133e-05, + "loss": 0.6938, + "step": 1524 + }, + { + "epoch": 0.06, + "learning_rate": 4.905424467812259e-05, + "loss": 0.6256, + "step": 1525 + }, + { + "epoch": 0.06, + "learning_rate": 4.905360348807387e-05, + "loss": 0.6126, + "step": 1526 + }, + { + "epoch": 0.06, + "learning_rate": 4.9052962298025136e-05, + "loss": 0.5779, + "step": 1527 + }, + { + "epoch": 0.06, + "learning_rate": 4.905232110797641e-05, + "loss": 0.6518, + "step": 1528 + }, + { + "epoch": 0.06, + "learning_rate": 4.905167991792768e-05, + "loss": 0.6155, + "step": 1529 + }, + { + "epoch": 0.06, + "learning_rate": 4.9051038727878945e-05, + "loss": 0.5489, + "step": 1530 + }, + { + "epoch": 0.06, + "learning_rate": 4.9050397537830216e-05, + "loss": 0.6167, + "step": 1531 + }, + { + "epoch": 0.06, + "learning_rate": 4.904975634778148e-05, + "loss": 0.6366, + "step": 1532 + }, + { + "epoch": 0.06, + "learning_rate": 4.904911515773275e-05, + "loss": 0.5476, + "step": 1533 + }, + { + "epoch": 0.06, + "learning_rate": 4.9048473967684025e-05, + "loss": 0.6639, + "step": 1534 + }, + { + "epoch": 0.06, + "learning_rate": 4.9047832777635296e-05, + "loss": 0.6305, + "step": 1535 + }, + { + "epoch": 0.06, + "learning_rate": 4.904719158758656e-05, + "loss": 0.6105, + "step": 1536 + }, + { + "epoch": 0.06, + "learning_rate": 4.904655039753783e-05, + "loss": 0.6404, + "step": 1537 + }, + { + "epoch": 0.06, + "learning_rate": 4.9045909207489105e-05, + "loss": 0.5228, + "step": 1538 + }, + { + "epoch": 0.06, + "learning_rate": 4.904526801744037e-05, + "loss": 0.615, + "step": 1539 + }, + { + "epoch": 0.06, + "learning_rate": 4.904462682739164e-05, + "loss": 0.6747, + "step": 1540 + }, + { + "epoch": 0.06, + "learning_rate": 4.9043985637342906e-05, + "loss": 0.6394, + "step": 1541 + }, + { + "epoch": 0.06, + "learning_rate": 4.9043344447294185e-05, + "loss": 0.7077, + "step": 1542 + }, + { + "epoch": 0.06, + "learning_rate": 4.904270325724545e-05, + "loss": 0.5882, + "step": 1543 + }, + { + "epoch": 0.06, + "learning_rate": 4.904206206719672e-05, + "loss": 0.6305, + "step": 1544 + }, + { + "epoch": 0.06, + "learning_rate": 4.9041420877147986e-05, + "loss": 0.5848, + "step": 1545 + }, + { + "epoch": 0.06, + "learning_rate": 4.904077968709926e-05, + "loss": 0.7045, + "step": 1546 + }, + { + "epoch": 0.06, + "learning_rate": 4.904013849705053e-05, + "loss": 0.6226, + "step": 1547 + }, + { + "epoch": 0.06, + "learning_rate": 4.9039497307001795e-05, + "loss": 0.6427, + "step": 1548 + }, + { + "epoch": 0.06, + "learning_rate": 4.9038856116953067e-05, + "loss": 0.5956, + "step": 1549 + }, + { + "epoch": 0.06, + "learning_rate": 4.903821492690434e-05, + "loss": 0.6875, + "step": 1550 + }, + { + "epoch": 0.06, + "learning_rate": 4.903757373685561e-05, + "loss": 0.6297, + "step": 1551 + }, + { + "epoch": 0.06, + "learning_rate": 4.9036932546806875e-05, + "loss": 0.6821, + "step": 1552 + }, + { + "epoch": 0.06, + "learning_rate": 4.903629135675815e-05, + "loss": 0.6136, + "step": 1553 + }, + { + "epoch": 0.06, + "learning_rate": 4.903565016670941e-05, + "loss": 0.5458, + "step": 1554 + }, + { + "epoch": 0.06, + "learning_rate": 4.903500897666068e-05, + "loss": 0.6467, + "step": 1555 + }, + { + "epoch": 0.06, + "learning_rate": 4.9034367786611955e-05, + "loss": 0.6171, + "step": 1556 + }, + { + "epoch": 0.06, + "learning_rate": 4.903372659656322e-05, + "loss": 0.6489, + "step": 1557 + }, + { + "epoch": 0.06, + "learning_rate": 4.90330854065145e-05, + "loss": 0.6294, + "step": 1558 + }, + { + "epoch": 0.06, + "learning_rate": 4.9032444216465763e-05, + "loss": 0.6763, + "step": 1559 + }, + { + "epoch": 0.06, + "learning_rate": 4.9031803026417035e-05, + "loss": 0.6508, + "step": 1560 + }, + { + "epoch": 0.06, + "learning_rate": 4.90311618363683e-05, + "loss": 0.6122, + "step": 1561 + }, + { + "epoch": 0.06, + "learning_rate": 4.903052064631957e-05, + "loss": 0.5615, + "step": 1562 + }, + { + "epoch": 0.06, + "learning_rate": 4.902987945627084e-05, + "loss": 0.6471, + "step": 1563 + }, + { + "epoch": 0.06, + "learning_rate": 4.902923826622211e-05, + "loss": 0.6647, + "step": 1564 + }, + { + "epoch": 0.06, + "learning_rate": 4.902859707617338e-05, + "loss": 0.6345, + "step": 1565 + }, + { + "epoch": 0.06, + "learning_rate": 4.902795588612465e-05, + "loss": 0.5945, + "step": 1566 + }, + { + "epoch": 0.06, + "learning_rate": 4.9027314696075924e-05, + "loss": 0.6018, + "step": 1567 + }, + { + "epoch": 0.06, + "learning_rate": 4.902667350602719e-05, + "loss": 0.6818, + "step": 1568 + }, + { + "epoch": 0.06, + "learning_rate": 4.902603231597846e-05, + "loss": 0.5939, + "step": 1569 + }, + { + "epoch": 0.06, + "learning_rate": 4.9025391125929725e-05, + "loss": 0.6133, + "step": 1570 + }, + { + "epoch": 0.06, + "learning_rate": 4.9024749935881e-05, + "loss": 0.7027, + "step": 1571 + }, + { + "epoch": 0.06, + "learning_rate": 4.902410874583226e-05, + "loss": 0.5554, + "step": 1572 + }, + { + "epoch": 0.06, + "learning_rate": 4.9023467555783534e-05, + "loss": 0.7075, + "step": 1573 + }, + { + "epoch": 0.06, + "learning_rate": 4.9022826365734805e-05, + "loss": 0.6406, + "step": 1574 + }, + { + "epoch": 0.06, + "learning_rate": 4.902218517568608e-05, + "loss": 0.6325, + "step": 1575 + }, + { + "epoch": 0.06, + "learning_rate": 4.902154398563735e-05, + "loss": 0.6549, + "step": 1576 + }, + { + "epoch": 0.06, + "learning_rate": 4.9020902795588614e-05, + "loss": 0.6225, + "step": 1577 + }, + { + "epoch": 0.06, + "learning_rate": 4.9020261605539885e-05, + "loss": 0.6921, + "step": 1578 + }, + { + "epoch": 0.06, + "learning_rate": 4.901962041549115e-05, + "loss": 0.6311, + "step": 1579 + }, + { + "epoch": 0.06, + "learning_rate": 4.901897922544242e-05, + "loss": 0.6632, + "step": 1580 + }, + { + "epoch": 0.06, + "learning_rate": 4.9018338035393694e-05, + "loss": 0.6971, + "step": 1581 + }, + { + "epoch": 0.06, + "learning_rate": 4.9017696845344965e-05, + "loss": 0.7333, + "step": 1582 + }, + { + "epoch": 0.06, + "learning_rate": 4.901705565529624e-05, + "loss": 0.6685, + "step": 1583 + }, + { + "epoch": 0.06, + "learning_rate": 4.90164144652475e-05, + "loss": 0.6219, + "step": 1584 + }, + { + "epoch": 0.06, + "learning_rate": 4.9015773275198774e-05, + "loss": 0.6196, + "step": 1585 + }, + { + "epoch": 0.06, + "learning_rate": 4.901513208515004e-05, + "loss": 0.5533, + "step": 1586 + }, + { + "epoch": 0.06, + "learning_rate": 4.901449089510131e-05, + "loss": 0.6547, + "step": 1587 + }, + { + "epoch": 0.06, + "learning_rate": 4.9013849705052575e-05, + "loss": 0.5972, + "step": 1588 + }, + { + "epoch": 0.06, + "learning_rate": 4.901320851500385e-05, + "loss": 0.5945, + "step": 1589 + }, + { + "epoch": 0.06, + "learning_rate": 4.901256732495512e-05, + "loss": 0.589, + "step": 1590 + }, + { + "epoch": 0.06, + "learning_rate": 4.901192613490639e-05, + "loss": 0.5729, + "step": 1591 + }, + { + "epoch": 0.06, + "learning_rate": 4.901128494485766e-05, + "loss": 0.6187, + "step": 1592 + }, + { + "epoch": 0.06, + "learning_rate": 4.901064375480893e-05, + "loss": 0.6351, + "step": 1593 + }, + { + "epoch": 0.06, + "learning_rate": 4.90100025647602e-05, + "loss": 0.5812, + "step": 1594 + }, + { + "epoch": 0.06, + "learning_rate": 4.9009361374711464e-05, + "loss": 0.677, + "step": 1595 + }, + { + "epoch": 0.06, + "learning_rate": 4.9008720184662736e-05, + "loss": 0.5333, + "step": 1596 + }, + { + "epoch": 0.06, + "learning_rate": 4.900807899461401e-05, + "loss": 0.6034, + "step": 1597 + }, + { + "epoch": 0.06, + "learning_rate": 4.900743780456528e-05, + "loss": 0.6371, + "step": 1598 + }, + { + "epoch": 0.06, + "learning_rate": 4.9006796614516544e-05, + "loss": 0.6656, + "step": 1599 + }, + { + "epoch": 0.06, + "learning_rate": 4.9006155424467816e-05, + "loss": 0.5699, + "step": 1600 + }, + { + "epoch": 0.06, + "learning_rate": 4.900551423441909e-05, + "loss": 0.5831, + "step": 1601 + }, + { + "epoch": 0.06, + "learning_rate": 4.900487304437035e-05, + "loss": 0.7082, + "step": 1602 + }, + { + "epoch": 0.06, + "learning_rate": 4.9004231854321624e-05, + "loss": 0.7421, + "step": 1603 + }, + { + "epoch": 0.06, + "learning_rate": 4.900359066427289e-05, + "loss": 0.5927, + "step": 1604 + }, + { + "epoch": 0.06, + "learning_rate": 4.900294947422416e-05, + "loss": 0.702, + "step": 1605 + }, + { + "epoch": 0.06, + "learning_rate": 4.900230828417543e-05, + "loss": 0.5389, + "step": 1606 + }, + { + "epoch": 0.06, + "learning_rate": 4.9001667094126704e-05, + "loss": 0.6463, + "step": 1607 + }, + { + "epoch": 0.06, + "learning_rate": 4.900102590407797e-05, + "loss": 0.6512, + "step": 1608 + }, + { + "epoch": 0.06, + "learning_rate": 4.900038471402924e-05, + "loss": 0.5983, + "step": 1609 + }, + { + "epoch": 0.06, + "learning_rate": 4.899974352398051e-05, + "loss": 0.6095, + "step": 1610 + }, + { + "epoch": 0.06, + "learning_rate": 4.899910233393178e-05, + "loss": 0.6714, + "step": 1611 + }, + { + "epoch": 0.06, + "learning_rate": 4.899846114388305e-05, + "loss": 0.6435, + "step": 1612 + }, + { + "epoch": 0.06, + "learning_rate": 4.8997819953834314e-05, + "loss": 0.6294, + "step": 1613 + }, + { + "epoch": 0.06, + "learning_rate": 4.899717876378559e-05, + "loss": 0.6043, + "step": 1614 + }, + { + "epoch": 0.06, + "learning_rate": 4.899653757373686e-05, + "loss": 0.6378, + "step": 1615 + }, + { + "epoch": 0.06, + "learning_rate": 4.899589638368813e-05, + "loss": 0.5828, + "step": 1616 + }, + { + "epoch": 0.06, + "learning_rate": 4.8995255193639394e-05, + "loss": 0.6478, + "step": 1617 + }, + { + "epoch": 0.06, + "learning_rate": 4.8994614003590666e-05, + "loss": 0.6474, + "step": 1618 + }, + { + "epoch": 0.06, + "learning_rate": 4.899397281354194e-05, + "loss": 0.6378, + "step": 1619 + }, + { + "epoch": 0.06, + "learning_rate": 4.89933316234932e-05, + "loss": 0.5968, + "step": 1620 + }, + { + "epoch": 0.06, + "learning_rate": 4.8992690433444474e-05, + "loss": 0.6027, + "step": 1621 + }, + { + "epoch": 0.06, + "learning_rate": 4.8992049243395746e-05, + "loss": 0.5646, + "step": 1622 + }, + { + "epoch": 0.06, + "learning_rate": 4.899140805334702e-05, + "loss": 0.6123, + "step": 1623 + }, + { + "epoch": 0.06, + "learning_rate": 4.899076686329828e-05, + "loss": 0.6467, + "step": 1624 + }, + { + "epoch": 0.06, + "learning_rate": 4.8990125673249554e-05, + "loss": 0.5895, + "step": 1625 + }, + { + "epoch": 0.06, + "learning_rate": 4.898948448320082e-05, + "loss": 0.61, + "step": 1626 + }, + { + "epoch": 0.06, + "learning_rate": 4.898884329315209e-05, + "loss": 0.5703, + "step": 1627 + }, + { + "epoch": 0.06, + "learning_rate": 4.898820210310336e-05, + "loss": 0.6541, + "step": 1628 + }, + { + "epoch": 0.06, + "learning_rate": 4.898756091305463e-05, + "loss": 0.5745, + "step": 1629 + }, + { + "epoch": 0.06, + "learning_rate": 4.8986919723005906e-05, + "loss": 0.6244, + "step": 1630 + }, + { + "epoch": 0.06, + "learning_rate": 4.898627853295717e-05, + "loss": 0.5967, + "step": 1631 + }, + { + "epoch": 0.06, + "learning_rate": 4.898563734290844e-05, + "loss": 0.5921, + "step": 1632 + }, + { + "epoch": 0.06, + "learning_rate": 4.898499615285971e-05, + "loss": 0.5899, + "step": 1633 + }, + { + "epoch": 0.06, + "learning_rate": 4.898435496281098e-05, + "loss": 0.7235, + "step": 1634 + }, + { + "epoch": 0.06, + "learning_rate": 4.8983713772762244e-05, + "loss": 0.5555, + "step": 1635 + }, + { + "epoch": 0.06, + "learning_rate": 4.8983072582713516e-05, + "loss": 0.6444, + "step": 1636 + }, + { + "epoch": 0.06, + "learning_rate": 4.898243139266479e-05, + "loss": 0.6408, + "step": 1637 + }, + { + "epoch": 0.06, + "learning_rate": 4.898179020261606e-05, + "loss": 0.6061, + "step": 1638 + }, + { + "epoch": 0.06, + "learning_rate": 4.898114901256733e-05, + "loss": 0.5591, + "step": 1639 + }, + { + "epoch": 0.06, + "learning_rate": 4.8980507822518596e-05, + "loss": 0.6799, + "step": 1640 + }, + { + "epoch": 0.06, + "learning_rate": 4.897986663246987e-05, + "loss": 0.615, + "step": 1641 + }, + { + "epoch": 0.06, + "learning_rate": 4.897922544242113e-05, + "loss": 0.5706, + "step": 1642 + }, + { + "epoch": 0.06, + "learning_rate": 4.8978584252372405e-05, + "loss": 0.6768, + "step": 1643 + }, + { + "epoch": 0.06, + "learning_rate": 4.897794306232367e-05, + "loss": 0.7311, + "step": 1644 + }, + { + "epoch": 0.06, + "learning_rate": 4.897730187227494e-05, + "loss": 0.7007, + "step": 1645 + }, + { + "epoch": 0.06, + "learning_rate": 4.897666068222621e-05, + "loss": 0.7004, + "step": 1646 + }, + { + "epoch": 0.06, + "learning_rate": 4.8976019492177485e-05, + "loss": 0.588, + "step": 1647 + }, + { + "epoch": 0.06, + "learning_rate": 4.8975378302128756e-05, + "loss": 0.6163, + "step": 1648 + }, + { + "epoch": 0.06, + "learning_rate": 4.897473711208002e-05, + "loss": 0.6017, + "step": 1649 + }, + { + "epoch": 0.06, + "learning_rate": 4.897409592203129e-05, + "loss": 0.6494, + "step": 1650 + }, + { + "epoch": 0.06, + "learning_rate": 4.897345473198256e-05, + "loss": 0.6948, + "step": 1651 + }, + { + "epoch": 0.06, + "learning_rate": 4.897281354193383e-05, + "loss": 0.6091, + "step": 1652 + }, + { + "epoch": 0.06, + "learning_rate": 4.89721723518851e-05, + "loss": 0.553, + "step": 1653 + }, + { + "epoch": 0.06, + "learning_rate": 4.897153116183637e-05, + "loss": 0.609, + "step": 1654 + }, + { + "epoch": 0.06, + "learning_rate": 4.8970889971787645e-05, + "loss": 0.5845, + "step": 1655 + }, + { + "epoch": 0.06, + "learning_rate": 4.897024878173891e-05, + "loss": 0.6119, + "step": 1656 + }, + { + "epoch": 0.06, + "learning_rate": 4.896960759169018e-05, + "loss": 0.6245, + "step": 1657 + }, + { + "epoch": 0.06, + "learning_rate": 4.8968966401641447e-05, + "loss": 0.6395, + "step": 1658 + }, + { + "epoch": 0.06, + "learning_rate": 4.896832521159272e-05, + "loss": 0.6125, + "step": 1659 + }, + { + "epoch": 0.06, + "learning_rate": 4.896768402154398e-05, + "loss": 0.5929, + "step": 1660 + }, + { + "epoch": 0.06, + "learning_rate": 4.8967042831495255e-05, + "loss": 0.6047, + "step": 1661 + }, + { + "epoch": 0.06, + "learning_rate": 4.8966401641446527e-05, + "loss": 0.6173, + "step": 1662 + }, + { + "epoch": 0.06, + "learning_rate": 4.89657604513978e-05, + "loss": 0.5336, + "step": 1663 + }, + { + "epoch": 0.06, + "learning_rate": 4.896511926134907e-05, + "loss": 0.5639, + "step": 1664 + }, + { + "epoch": 0.06, + "learning_rate": 4.8964478071300335e-05, + "loss": 0.6123, + "step": 1665 + }, + { + "epoch": 0.06, + "learning_rate": 4.896383688125161e-05, + "loss": 0.5762, + "step": 1666 + }, + { + "epoch": 0.06, + "learning_rate": 4.896319569120287e-05, + "loss": 0.7567, + "step": 1667 + }, + { + "epoch": 0.06, + "learning_rate": 4.896255450115414e-05, + "loss": 0.6202, + "step": 1668 + }, + { + "epoch": 0.06, + "learning_rate": 4.8961913311105415e-05, + "loss": 0.6349, + "step": 1669 + }, + { + "epoch": 0.06, + "learning_rate": 4.896127212105669e-05, + "loss": 0.6862, + "step": 1670 + }, + { + "epoch": 0.06, + "learning_rate": 4.896063093100796e-05, + "loss": 0.6039, + "step": 1671 + }, + { + "epoch": 0.06, + "learning_rate": 4.8959989740959223e-05, + "loss": 0.575, + "step": 1672 + }, + { + "epoch": 0.06, + "learning_rate": 4.8959348550910495e-05, + "loss": 0.6856, + "step": 1673 + }, + { + "epoch": 0.06, + "learning_rate": 4.895870736086176e-05, + "loss": 0.5616, + "step": 1674 + }, + { + "epoch": 0.06, + "learning_rate": 4.895806617081303e-05, + "loss": 0.709, + "step": 1675 + }, + { + "epoch": 0.06, + "learning_rate": 4.89574249807643e-05, + "loss": 0.6329, + "step": 1676 + }, + { + "epoch": 0.06, + "learning_rate": 4.895678379071557e-05, + "loss": 0.6855, + "step": 1677 + }, + { + "epoch": 0.06, + "learning_rate": 4.895614260066684e-05, + "loss": 0.729, + "step": 1678 + }, + { + "epoch": 0.06, + "learning_rate": 4.895550141061811e-05, + "loss": 0.6756, + "step": 1679 + }, + { + "epoch": 0.06, + "learning_rate": 4.8954860220569384e-05, + "loss": 0.5932, + "step": 1680 + }, + { + "epoch": 0.06, + "learning_rate": 4.895421903052065e-05, + "loss": 0.6026, + "step": 1681 + }, + { + "epoch": 0.06, + "learning_rate": 4.895357784047192e-05, + "loss": 0.636, + "step": 1682 + }, + { + "epoch": 0.06, + "learning_rate": 4.8952936650423185e-05, + "loss": 0.5471, + "step": 1683 + }, + { + "epoch": 0.06, + "learning_rate": 4.895229546037446e-05, + "loss": 0.6388, + "step": 1684 + }, + { + "epoch": 0.06, + "learning_rate": 4.895165427032573e-05, + "loss": 0.609, + "step": 1685 + }, + { + "epoch": 0.06, + "learning_rate": 4.8951013080277e-05, + "loss": 0.6503, + "step": 1686 + }, + { + "epoch": 0.06, + "learning_rate": 4.8950371890228265e-05, + "loss": 0.6376, + "step": 1687 + }, + { + "epoch": 0.06, + "learning_rate": 4.894973070017954e-05, + "loss": 0.5297, + "step": 1688 + }, + { + "epoch": 0.06, + "learning_rate": 4.894908951013081e-05, + "loss": 0.5657, + "step": 1689 + }, + { + "epoch": 0.06, + "learning_rate": 4.8948448320082074e-05, + "loss": 0.6095, + "step": 1690 + }, + { + "epoch": 0.07, + "learning_rate": 4.8947807130033345e-05, + "loss": 0.633, + "step": 1691 + }, + { + "epoch": 0.07, + "learning_rate": 4.894716593998461e-05, + "loss": 0.6075, + "step": 1692 + }, + { + "epoch": 0.07, + "learning_rate": 4.894652474993588e-05, + "loss": 0.6343, + "step": 1693 + }, + { + "epoch": 0.07, + "learning_rate": 4.8945883559887154e-05, + "loss": 0.5291, + "step": 1694 + }, + { + "epoch": 0.07, + "learning_rate": 4.8945242369838425e-05, + "loss": 0.6001, + "step": 1695 + }, + { + "epoch": 0.07, + "learning_rate": 4.894460117978969e-05, + "loss": 0.5824, + "step": 1696 + }, + { + "epoch": 0.07, + "learning_rate": 4.894395998974096e-05, + "loss": 0.6392, + "step": 1697 + }, + { + "epoch": 0.07, + "learning_rate": 4.8943318799692234e-05, + "loss": 0.5621, + "step": 1698 + }, + { + "epoch": 0.07, + "learning_rate": 4.89426776096435e-05, + "loss": 0.6695, + "step": 1699 + }, + { + "epoch": 0.07, + "learning_rate": 4.894203641959477e-05, + "loss": 0.6112, + "step": 1700 + }, + { + "epoch": 0.07, + "learning_rate": 4.8941395229546035e-05, + "loss": 0.6534, + "step": 1701 + }, + { + "epoch": 0.07, + "learning_rate": 4.8940754039497314e-05, + "loss": 0.6175, + "step": 1702 + }, + { + "epoch": 0.07, + "learning_rate": 4.894011284944858e-05, + "loss": 0.7202, + "step": 1703 + }, + { + "epoch": 0.07, + "learning_rate": 4.893947165939985e-05, + "loss": 0.5994, + "step": 1704 + }, + { + "epoch": 0.07, + "learning_rate": 4.8938830469351116e-05, + "loss": 0.607, + "step": 1705 + }, + { + "epoch": 0.07, + "learning_rate": 4.893818927930239e-05, + "loss": 0.6859, + "step": 1706 + }, + { + "epoch": 0.07, + "learning_rate": 4.893754808925366e-05, + "loss": 0.6079, + "step": 1707 + }, + { + "epoch": 0.07, + "learning_rate": 4.8936906899204924e-05, + "loss": 0.6224, + "step": 1708 + }, + { + "epoch": 0.07, + "learning_rate": 4.8936265709156196e-05, + "loss": 0.6835, + "step": 1709 + }, + { + "epoch": 0.07, + "learning_rate": 4.893562451910747e-05, + "loss": 0.637, + "step": 1710 + }, + { + "epoch": 0.07, + "learning_rate": 4.893498332905874e-05, + "loss": 0.5689, + "step": 1711 + }, + { + "epoch": 0.07, + "learning_rate": 4.8934342139010004e-05, + "loss": 0.7275, + "step": 1712 + }, + { + "epoch": 0.07, + "learning_rate": 4.8933700948961276e-05, + "loss": 0.6113, + "step": 1713 + }, + { + "epoch": 0.07, + "learning_rate": 4.893305975891254e-05, + "loss": 0.5406, + "step": 1714 + }, + { + "epoch": 0.07, + "learning_rate": 4.893241856886381e-05, + "loss": 0.5479, + "step": 1715 + }, + { + "epoch": 0.07, + "learning_rate": 4.8931777378815084e-05, + "loss": 0.6797, + "step": 1716 + }, + { + "epoch": 0.07, + "learning_rate": 4.893113618876635e-05, + "loss": 0.6219, + "step": 1717 + }, + { + "epoch": 0.07, + "learning_rate": 4.893049499871763e-05, + "loss": 0.6358, + "step": 1718 + }, + { + "epoch": 0.07, + "learning_rate": 4.892985380866889e-05, + "loss": 0.65, + "step": 1719 + }, + { + "epoch": 0.07, + "learning_rate": 4.8929212618620164e-05, + "loss": 0.5935, + "step": 1720 + }, + { + "epoch": 0.07, + "learning_rate": 4.892857142857143e-05, + "loss": 0.6652, + "step": 1721 + }, + { + "epoch": 0.07, + "learning_rate": 4.89279302385227e-05, + "loss": 0.6178, + "step": 1722 + }, + { + "epoch": 0.07, + "learning_rate": 4.8927289048473966e-05, + "loss": 0.6581, + "step": 1723 + }, + { + "epoch": 0.07, + "learning_rate": 4.892664785842524e-05, + "loss": 0.6344, + "step": 1724 + }, + { + "epoch": 0.07, + "learning_rate": 4.892600666837651e-05, + "loss": 0.6472, + "step": 1725 + }, + { + "epoch": 0.07, + "learning_rate": 4.892536547832778e-05, + "loss": 0.6094, + "step": 1726 + }, + { + "epoch": 0.07, + "learning_rate": 4.892472428827905e-05, + "loss": 0.6105, + "step": 1727 + }, + { + "epoch": 0.07, + "learning_rate": 4.892408309823032e-05, + "loss": 0.7314, + "step": 1728 + }, + { + "epoch": 0.07, + "learning_rate": 4.892344190818159e-05, + "loss": 0.6524, + "step": 1729 + }, + { + "epoch": 0.07, + "learning_rate": 4.8922800718132854e-05, + "loss": 0.6256, + "step": 1730 + }, + { + "epoch": 0.07, + "learning_rate": 4.8922159528084126e-05, + "loss": 0.6266, + "step": 1731 + }, + { + "epoch": 0.07, + "learning_rate": 4.892151833803539e-05, + "loss": 0.6736, + "step": 1732 + }, + { + "epoch": 0.07, + "learning_rate": 4.892087714798666e-05, + "loss": 0.6636, + "step": 1733 + }, + { + "epoch": 0.07, + "learning_rate": 4.8920235957937934e-05, + "loss": 0.6431, + "step": 1734 + }, + { + "epoch": 0.07, + "learning_rate": 4.8919594767889206e-05, + "loss": 0.6361, + "step": 1735 + }, + { + "epoch": 0.07, + "learning_rate": 4.891895357784048e-05, + "loss": 0.6558, + "step": 1736 + }, + { + "epoch": 0.07, + "learning_rate": 4.891831238779174e-05, + "loss": 0.6108, + "step": 1737 + }, + { + "epoch": 0.07, + "learning_rate": 4.8917671197743014e-05, + "loss": 0.5897, + "step": 1738 + }, + { + "epoch": 0.07, + "learning_rate": 4.891703000769428e-05, + "loss": 0.6459, + "step": 1739 + }, + { + "epoch": 0.07, + "learning_rate": 4.891638881764555e-05, + "loss": 0.6658, + "step": 1740 + }, + { + "epoch": 0.07, + "learning_rate": 4.891574762759682e-05, + "loss": 0.6137, + "step": 1741 + }, + { + "epoch": 0.07, + "learning_rate": 4.8915106437548095e-05, + "loss": 0.6194, + "step": 1742 + }, + { + "epoch": 0.07, + "learning_rate": 4.8914465247499366e-05, + "loss": 0.6039, + "step": 1743 + }, + { + "epoch": 0.07, + "learning_rate": 4.891382405745063e-05, + "loss": 0.5716, + "step": 1744 + }, + { + "epoch": 0.07, + "learning_rate": 4.89131828674019e-05, + "loss": 0.6369, + "step": 1745 + }, + { + "epoch": 0.07, + "learning_rate": 4.891254167735317e-05, + "loss": 0.5854, + "step": 1746 + }, + { + "epoch": 0.07, + "learning_rate": 4.891190048730444e-05, + "loss": 0.6977, + "step": 1747 + }, + { + "epoch": 0.07, + "learning_rate": 4.8911259297255704e-05, + "loss": 0.5641, + "step": 1748 + }, + { + "epoch": 0.07, + "learning_rate": 4.8910618107206976e-05, + "loss": 0.5781, + "step": 1749 + }, + { + "epoch": 0.07, + "learning_rate": 4.890997691715825e-05, + "loss": 0.571, + "step": 1750 + }, + { + "epoch": 0.07, + "learning_rate": 4.890933572710952e-05, + "loss": 0.7375, + "step": 1751 + }, + { + "epoch": 0.07, + "learning_rate": 4.890869453706079e-05, + "loss": 0.573, + "step": 1752 + }, + { + "epoch": 0.07, + "learning_rate": 4.8908053347012056e-05, + "loss": 0.6639, + "step": 1753 + }, + { + "epoch": 0.07, + "learning_rate": 4.890741215696333e-05, + "loss": 0.6445, + "step": 1754 + }, + { + "epoch": 0.07, + "learning_rate": 4.890677096691459e-05, + "loss": 0.6196, + "step": 1755 + }, + { + "epoch": 0.07, + "learning_rate": 4.8906129776865865e-05, + "loss": 0.5837, + "step": 1756 + }, + { + "epoch": 0.07, + "learning_rate": 4.8905488586817136e-05, + "loss": 0.6093, + "step": 1757 + }, + { + "epoch": 0.07, + "learning_rate": 4.890484739676841e-05, + "loss": 0.6167, + "step": 1758 + }, + { + "epoch": 0.07, + "learning_rate": 4.890420620671967e-05, + "loss": 0.5782, + "step": 1759 + }, + { + "epoch": 0.07, + "learning_rate": 4.8903565016670945e-05, + "loss": 0.6434, + "step": 1760 + }, + { + "epoch": 0.07, + "learning_rate": 4.8902923826622216e-05, + "loss": 0.6465, + "step": 1761 + }, + { + "epoch": 0.07, + "learning_rate": 4.890228263657348e-05, + "loss": 0.634, + "step": 1762 + }, + { + "epoch": 0.07, + "learning_rate": 4.890164144652475e-05, + "loss": 0.6448, + "step": 1763 + }, + { + "epoch": 0.07, + "learning_rate": 4.890100025647602e-05, + "loss": 0.5755, + "step": 1764 + }, + { + "epoch": 0.07, + "learning_rate": 4.890035906642729e-05, + "loss": 0.5746, + "step": 1765 + }, + { + "epoch": 0.07, + "learning_rate": 4.889971787637856e-05, + "loss": 0.5936, + "step": 1766 + }, + { + "epoch": 0.07, + "learning_rate": 4.889907668632983e-05, + "loss": 0.6127, + "step": 1767 + }, + { + "epoch": 0.07, + "learning_rate": 4.88984354962811e-05, + "loss": 0.6915, + "step": 1768 + }, + { + "epoch": 0.07, + "learning_rate": 4.889779430623237e-05, + "loss": 0.5974, + "step": 1769 + }, + { + "epoch": 0.07, + "learning_rate": 4.889715311618364e-05, + "loss": 0.5524, + "step": 1770 + }, + { + "epoch": 0.07, + "learning_rate": 4.8896511926134907e-05, + "loss": 0.5161, + "step": 1771 + }, + { + "epoch": 0.07, + "learning_rate": 4.889587073608618e-05, + "loss": 0.5786, + "step": 1772 + }, + { + "epoch": 0.07, + "learning_rate": 4.889522954603744e-05, + "loss": 0.6008, + "step": 1773 + }, + { + "epoch": 0.07, + "learning_rate": 4.889458835598872e-05, + "loss": 0.5791, + "step": 1774 + }, + { + "epoch": 0.07, + "learning_rate": 4.889394716593999e-05, + "loss": 0.5647, + "step": 1775 + }, + { + "epoch": 0.07, + "learning_rate": 4.889330597589126e-05, + "loss": 0.5631, + "step": 1776 + }, + { + "epoch": 0.07, + "learning_rate": 4.889266478584252e-05, + "loss": 0.6482, + "step": 1777 + }, + { + "epoch": 0.07, + "learning_rate": 4.8892023595793795e-05, + "loss": 0.5982, + "step": 1778 + }, + { + "epoch": 0.07, + "learning_rate": 4.889138240574507e-05, + "loss": 0.5999, + "step": 1779 + }, + { + "epoch": 0.07, + "learning_rate": 4.889074121569633e-05, + "loss": 0.7286, + "step": 1780 + }, + { + "epoch": 0.07, + "learning_rate": 4.88901000256476e-05, + "loss": 0.678, + "step": 1781 + }, + { + "epoch": 0.07, + "learning_rate": 4.8889458835598875e-05, + "loss": 0.6949, + "step": 1782 + }, + { + "epoch": 0.07, + "learning_rate": 4.888881764555015e-05, + "loss": 0.5565, + "step": 1783 + }, + { + "epoch": 0.07, + "learning_rate": 4.888817645550141e-05, + "loss": 0.5937, + "step": 1784 + }, + { + "epoch": 0.07, + "learning_rate": 4.8887535265452683e-05, + "loss": 0.5403, + "step": 1785 + }, + { + "epoch": 0.07, + "learning_rate": 4.888689407540395e-05, + "loss": 0.6042, + "step": 1786 + }, + { + "epoch": 0.07, + "learning_rate": 4.888625288535522e-05, + "loss": 0.6813, + "step": 1787 + }, + { + "epoch": 0.07, + "learning_rate": 4.888561169530649e-05, + "loss": 0.6041, + "step": 1788 + }, + { + "epoch": 0.07, + "learning_rate": 4.888497050525776e-05, + "loss": 0.6038, + "step": 1789 + }, + { + "epoch": 0.07, + "learning_rate": 4.8884329315209035e-05, + "loss": 0.5639, + "step": 1790 + }, + { + "epoch": 0.07, + "learning_rate": 4.88836881251603e-05, + "loss": 0.5949, + "step": 1791 + }, + { + "epoch": 0.07, + "learning_rate": 4.888304693511157e-05, + "loss": 0.6252, + "step": 1792 + }, + { + "epoch": 0.07, + "learning_rate": 4.888240574506284e-05, + "loss": 0.5679, + "step": 1793 + }, + { + "epoch": 0.07, + "learning_rate": 4.888176455501411e-05, + "loss": 0.6862, + "step": 1794 + }, + { + "epoch": 0.07, + "learning_rate": 4.8881123364965374e-05, + "loss": 0.5273, + "step": 1795 + }, + { + "epoch": 0.07, + "learning_rate": 4.8880482174916645e-05, + "loss": 0.5925, + "step": 1796 + }, + { + "epoch": 0.07, + "learning_rate": 4.887984098486792e-05, + "loss": 0.672, + "step": 1797 + }, + { + "epoch": 0.07, + "learning_rate": 4.887919979481919e-05, + "loss": 0.6115, + "step": 1798 + }, + { + "epoch": 0.07, + "learning_rate": 4.887855860477046e-05, + "loss": 0.61, + "step": 1799 + }, + { + "epoch": 0.07, + "learning_rate": 4.8877917414721725e-05, + "loss": 0.6693, + "step": 1800 + }, + { + "epoch": 0.07, + "learning_rate": 4.8877276224673e-05, + "loss": 0.5672, + "step": 1801 + }, + { + "epoch": 0.07, + "learning_rate": 4.887663503462426e-05, + "loss": 0.6129, + "step": 1802 + }, + { + "epoch": 0.07, + "learning_rate": 4.8875993844575534e-05, + "loss": 0.6769, + "step": 1803 + }, + { + "epoch": 0.07, + "learning_rate": 4.88753526545268e-05, + "loss": 0.6643, + "step": 1804 + }, + { + "epoch": 0.07, + "learning_rate": 4.887471146447807e-05, + "loss": 0.5615, + "step": 1805 + }, + { + "epoch": 0.07, + "learning_rate": 4.887407027442935e-05, + "loss": 0.6067, + "step": 1806 + }, + { + "epoch": 0.07, + "learning_rate": 4.8873429084380614e-05, + "loss": 0.6046, + "step": 1807 + }, + { + "epoch": 0.07, + "learning_rate": 4.8872787894331886e-05, + "loss": 0.6, + "step": 1808 + }, + { + "epoch": 0.07, + "learning_rate": 4.887214670428315e-05, + "loss": 0.5904, + "step": 1809 + }, + { + "epoch": 0.07, + "learning_rate": 4.887150551423442e-05, + "loss": 0.5915, + "step": 1810 + }, + { + "epoch": 0.07, + "learning_rate": 4.887086432418569e-05, + "loss": 0.5364, + "step": 1811 + }, + { + "epoch": 0.07, + "learning_rate": 4.887022313413696e-05, + "loss": 0.5867, + "step": 1812 + }, + { + "epoch": 0.07, + "learning_rate": 4.886958194408823e-05, + "loss": 0.6545, + "step": 1813 + }, + { + "epoch": 0.07, + "learning_rate": 4.88689407540395e-05, + "loss": 0.5791, + "step": 1814 + }, + { + "epoch": 0.07, + "learning_rate": 4.8868299563990774e-05, + "loss": 0.5714, + "step": 1815 + }, + { + "epoch": 0.07, + "learning_rate": 4.886765837394204e-05, + "loss": 0.635, + "step": 1816 + }, + { + "epoch": 0.07, + "learning_rate": 4.886701718389331e-05, + "loss": 0.6492, + "step": 1817 + }, + { + "epoch": 0.07, + "learning_rate": 4.8866375993844576e-05, + "loss": 0.5566, + "step": 1818 + }, + { + "epoch": 0.07, + "learning_rate": 4.886573480379585e-05, + "loss": 0.6237, + "step": 1819 + }, + { + "epoch": 0.07, + "learning_rate": 4.886509361374711e-05, + "loss": 0.6745, + "step": 1820 + }, + { + "epoch": 0.07, + "learning_rate": 4.8864452423698384e-05, + "loss": 0.7586, + "step": 1821 + }, + { + "epoch": 0.07, + "learning_rate": 4.8863811233649656e-05, + "loss": 0.6689, + "step": 1822 + }, + { + "epoch": 0.07, + "learning_rate": 4.886317004360093e-05, + "loss": 0.5851, + "step": 1823 + }, + { + "epoch": 0.07, + "learning_rate": 4.88625288535522e-05, + "loss": 0.6302, + "step": 1824 + }, + { + "epoch": 0.07, + "learning_rate": 4.8861887663503464e-05, + "loss": 0.5105, + "step": 1825 + }, + { + "epoch": 0.07, + "learning_rate": 4.8861246473454736e-05, + "loss": 0.6356, + "step": 1826 + }, + { + "epoch": 0.07, + "learning_rate": 4.8860605283406e-05, + "loss": 0.6728, + "step": 1827 + }, + { + "epoch": 0.07, + "learning_rate": 4.885996409335727e-05, + "loss": 0.585, + "step": 1828 + }, + { + "epoch": 0.07, + "learning_rate": 4.8859322903308544e-05, + "loss": 0.6197, + "step": 1829 + }, + { + "epoch": 0.07, + "learning_rate": 4.8858681713259816e-05, + "loss": 0.6467, + "step": 1830 + }, + { + "epoch": 0.07, + "learning_rate": 4.885804052321108e-05, + "loss": 0.5473, + "step": 1831 + }, + { + "epoch": 0.07, + "learning_rate": 4.885739933316235e-05, + "loss": 0.5542, + "step": 1832 + }, + { + "epoch": 0.07, + "learning_rate": 4.8856758143113624e-05, + "loss": 0.6517, + "step": 1833 + }, + { + "epoch": 0.07, + "learning_rate": 4.885611695306489e-05, + "loss": 0.6377, + "step": 1834 + }, + { + "epoch": 0.07, + "learning_rate": 4.885547576301616e-05, + "loss": 0.6616, + "step": 1835 + }, + { + "epoch": 0.07, + "learning_rate": 4.8854834572967426e-05, + "loss": 0.6461, + "step": 1836 + }, + { + "epoch": 0.07, + "learning_rate": 4.88541933829187e-05, + "loss": 0.5778, + "step": 1837 + }, + { + "epoch": 0.07, + "learning_rate": 4.885355219286997e-05, + "loss": 0.5698, + "step": 1838 + }, + { + "epoch": 0.07, + "learning_rate": 4.885291100282124e-05, + "loss": 0.6905, + "step": 1839 + }, + { + "epoch": 0.07, + "learning_rate": 4.8852269812772506e-05, + "loss": 0.6626, + "step": 1840 + }, + { + "epoch": 0.07, + "learning_rate": 4.885162862272378e-05, + "loss": 0.7315, + "step": 1841 + }, + { + "epoch": 0.07, + "learning_rate": 4.885098743267505e-05, + "loss": 0.6443, + "step": 1842 + }, + { + "epoch": 0.07, + "learning_rate": 4.8850346242626314e-05, + "loss": 0.5984, + "step": 1843 + }, + { + "epoch": 0.07, + "learning_rate": 4.8849705052577586e-05, + "loss": 0.6006, + "step": 1844 + }, + { + "epoch": 0.07, + "learning_rate": 4.884906386252886e-05, + "loss": 0.5807, + "step": 1845 + }, + { + "epoch": 0.07, + "learning_rate": 4.884842267248013e-05, + "loss": 0.6739, + "step": 1846 + }, + { + "epoch": 0.07, + "learning_rate": 4.8847781482431394e-05, + "loss": 0.5694, + "step": 1847 + }, + { + "epoch": 0.07, + "learning_rate": 4.8847140292382666e-05, + "loss": 0.6093, + "step": 1848 + }, + { + "epoch": 0.07, + "learning_rate": 4.884649910233394e-05, + "loss": 0.5458, + "step": 1849 + }, + { + "epoch": 0.07, + "learning_rate": 4.88458579122852e-05, + "loss": 0.5525, + "step": 1850 + }, + { + "epoch": 0.07, + "learning_rate": 4.8845216722236474e-05, + "loss": 0.6082, + "step": 1851 + }, + { + "epoch": 0.07, + "learning_rate": 4.884457553218774e-05, + "loss": 0.5432, + "step": 1852 + }, + { + "epoch": 0.07, + "learning_rate": 4.884393434213901e-05, + "loss": 0.6128, + "step": 1853 + }, + { + "epoch": 0.07, + "learning_rate": 4.884329315209028e-05, + "loss": 0.5693, + "step": 1854 + }, + { + "epoch": 0.07, + "learning_rate": 4.8842651962041555e-05, + "loss": 0.5341, + "step": 1855 + }, + { + "epoch": 0.07, + "learning_rate": 4.884201077199282e-05, + "loss": 0.5849, + "step": 1856 + }, + { + "epoch": 0.07, + "learning_rate": 4.884136958194409e-05, + "loss": 0.6836, + "step": 1857 + }, + { + "epoch": 0.07, + "learning_rate": 4.884072839189536e-05, + "loss": 0.6892, + "step": 1858 + }, + { + "epoch": 0.07, + "learning_rate": 4.884008720184663e-05, + "loss": 0.6175, + "step": 1859 + }, + { + "epoch": 0.07, + "learning_rate": 4.88394460117979e-05, + "loss": 0.5449, + "step": 1860 + }, + { + "epoch": 0.07, + "learning_rate": 4.8838804821749165e-05, + "loss": 0.653, + "step": 1861 + }, + { + "epoch": 0.07, + "learning_rate": 4.883816363170044e-05, + "loss": 0.5429, + "step": 1862 + }, + { + "epoch": 0.07, + "learning_rate": 4.883752244165171e-05, + "loss": 0.6077, + "step": 1863 + }, + { + "epoch": 0.07, + "learning_rate": 4.883688125160298e-05, + "loss": 0.6587, + "step": 1864 + }, + { + "epoch": 0.07, + "learning_rate": 4.8836240061554245e-05, + "loss": 0.6552, + "step": 1865 + }, + { + "epoch": 0.07, + "learning_rate": 4.8835598871505516e-05, + "loss": 0.7016, + "step": 1866 + }, + { + "epoch": 0.07, + "learning_rate": 4.883495768145679e-05, + "loss": 0.8168, + "step": 1867 + }, + { + "epoch": 0.07, + "learning_rate": 4.883431649140805e-05, + "loss": 0.6538, + "step": 1868 + }, + { + "epoch": 0.07, + "learning_rate": 4.8833675301359325e-05, + "loss": 0.5996, + "step": 1869 + }, + { + "epoch": 0.07, + "learning_rate": 4.8833034111310596e-05, + "loss": 0.7295, + "step": 1870 + }, + { + "epoch": 0.07, + "learning_rate": 4.883239292126187e-05, + "loss": 0.6114, + "step": 1871 + }, + { + "epoch": 0.07, + "learning_rate": 4.883175173121313e-05, + "loss": 0.5798, + "step": 1872 + }, + { + "epoch": 0.07, + "learning_rate": 4.8831110541164405e-05, + "loss": 0.5492, + "step": 1873 + }, + { + "epoch": 0.07, + "learning_rate": 4.883046935111567e-05, + "loss": 0.6663, + "step": 1874 + }, + { + "epoch": 0.07, + "learning_rate": 4.882982816106694e-05, + "loss": 0.6703, + "step": 1875 + }, + { + "epoch": 0.07, + "learning_rate": 4.882918697101821e-05, + "loss": 0.6454, + "step": 1876 + }, + { + "epoch": 0.07, + "learning_rate": 4.882854578096948e-05, + "loss": 0.6421, + "step": 1877 + }, + { + "epoch": 0.07, + "learning_rate": 4.8827904590920757e-05, + "loss": 0.568, + "step": 1878 + }, + { + "epoch": 0.07, + "learning_rate": 4.882726340087202e-05, + "loss": 0.6382, + "step": 1879 + }, + { + "epoch": 0.07, + "learning_rate": 4.882662221082329e-05, + "loss": 0.6171, + "step": 1880 + }, + { + "epoch": 0.07, + "learning_rate": 4.882598102077456e-05, + "loss": 0.5733, + "step": 1881 + }, + { + "epoch": 0.07, + "learning_rate": 4.882533983072583e-05, + "loss": 0.637, + "step": 1882 + }, + { + "epoch": 0.07, + "learning_rate": 4.8824698640677095e-05, + "loss": 0.5999, + "step": 1883 + }, + { + "epoch": 0.07, + "learning_rate": 4.8824057450628367e-05, + "loss": 0.7039, + "step": 1884 + }, + { + "epoch": 0.07, + "learning_rate": 4.882341626057964e-05, + "loss": 0.6252, + "step": 1885 + }, + { + "epoch": 0.07, + "learning_rate": 4.882277507053091e-05, + "loss": 0.6408, + "step": 1886 + }, + { + "epoch": 0.07, + "learning_rate": 4.882213388048218e-05, + "loss": 0.6055, + "step": 1887 + }, + { + "epoch": 0.07, + "learning_rate": 4.882149269043345e-05, + "loss": 0.6588, + "step": 1888 + }, + { + "epoch": 0.07, + "learning_rate": 4.882085150038472e-05, + "loss": 0.5946, + "step": 1889 + }, + { + "epoch": 0.07, + "learning_rate": 4.882021031033598e-05, + "loss": 0.6608, + "step": 1890 + }, + { + "epoch": 0.07, + "learning_rate": 4.8819569120287255e-05, + "loss": 0.6791, + "step": 1891 + }, + { + "epoch": 0.07, + "learning_rate": 4.881892793023852e-05, + "loss": 0.6113, + "step": 1892 + }, + { + "epoch": 0.07, + "learning_rate": 4.881828674018979e-05, + "loss": 0.5347, + "step": 1893 + }, + { + "epoch": 0.07, + "learning_rate": 4.8817645550141063e-05, + "loss": 0.5989, + "step": 1894 + }, + { + "epoch": 0.07, + "learning_rate": 4.8817004360092335e-05, + "loss": 0.5798, + "step": 1895 + }, + { + "epoch": 0.07, + "learning_rate": 4.881636317004361e-05, + "loss": 0.6573, + "step": 1896 + }, + { + "epoch": 0.07, + "learning_rate": 4.881572197999487e-05, + "loss": 0.599, + "step": 1897 + }, + { + "epoch": 0.07, + "learning_rate": 4.8815080789946143e-05, + "loss": 0.6671, + "step": 1898 + }, + { + "epoch": 0.07, + "learning_rate": 4.881443959989741e-05, + "loss": 0.6578, + "step": 1899 + }, + { + "epoch": 0.07, + "learning_rate": 4.881379840984868e-05, + "loss": 0.6153, + "step": 1900 + }, + { + "epoch": 0.07, + "learning_rate": 4.881315721979995e-05, + "loss": 0.6497, + "step": 1901 + }, + { + "epoch": 0.07, + "learning_rate": 4.8812516029751224e-05, + "loss": 0.6371, + "step": 1902 + }, + { + "epoch": 0.07, + "learning_rate": 4.8811874839702495e-05, + "loss": 0.5744, + "step": 1903 + }, + { + "epoch": 0.07, + "learning_rate": 4.881123364965376e-05, + "loss": 0.5692, + "step": 1904 + }, + { + "epoch": 0.07, + "learning_rate": 4.881059245960503e-05, + "loss": 0.5892, + "step": 1905 + }, + { + "epoch": 0.07, + "learning_rate": 4.88099512695563e-05, + "loss": 0.5309, + "step": 1906 + }, + { + "epoch": 0.07, + "learning_rate": 4.880931007950757e-05, + "loss": 0.7456, + "step": 1907 + }, + { + "epoch": 0.07, + "learning_rate": 4.8808668889458834e-05, + "loss": 0.6542, + "step": 1908 + }, + { + "epoch": 0.07, + "learning_rate": 4.8808027699410105e-05, + "loss": 0.5666, + "step": 1909 + }, + { + "epoch": 0.07, + "learning_rate": 4.880738650936138e-05, + "loss": 0.6197, + "step": 1910 + }, + { + "epoch": 0.07, + "learning_rate": 4.880674531931265e-05, + "loss": 0.5662, + "step": 1911 + }, + { + "epoch": 0.07, + "learning_rate": 4.880610412926392e-05, + "loss": 0.656, + "step": 1912 + }, + { + "epoch": 0.07, + "learning_rate": 4.8805462939215185e-05, + "loss": 0.6348, + "step": 1913 + }, + { + "epoch": 0.07, + "learning_rate": 4.880482174916646e-05, + "loss": 0.5762, + "step": 1914 + }, + { + "epoch": 0.07, + "learning_rate": 4.880418055911772e-05, + "loss": 0.5963, + "step": 1915 + }, + { + "epoch": 0.07, + "learning_rate": 4.8803539369068994e-05, + "loss": 0.7337, + "step": 1916 + }, + { + "epoch": 0.07, + "learning_rate": 4.8802898179020265e-05, + "loss": 0.6742, + "step": 1917 + }, + { + "epoch": 0.07, + "learning_rate": 4.880225698897154e-05, + "loss": 0.6299, + "step": 1918 + }, + { + "epoch": 0.07, + "learning_rate": 4.88016157989228e-05, + "loss": 0.573, + "step": 1919 + }, + { + "epoch": 0.07, + "learning_rate": 4.8800974608874074e-05, + "loss": 0.652, + "step": 1920 + }, + { + "epoch": 0.07, + "learning_rate": 4.8800333418825346e-05, + "loss": 0.5417, + "step": 1921 + }, + { + "epoch": 0.07, + "learning_rate": 4.879969222877661e-05, + "loss": 0.6001, + "step": 1922 + }, + { + "epoch": 0.07, + "learning_rate": 4.879905103872788e-05, + "loss": 0.6307, + "step": 1923 + }, + { + "epoch": 0.07, + "learning_rate": 4.879840984867915e-05, + "loss": 0.5833, + "step": 1924 + }, + { + "epoch": 0.07, + "learning_rate": 4.879776865863042e-05, + "loss": 0.583, + "step": 1925 + }, + { + "epoch": 0.07, + "learning_rate": 4.879712746858169e-05, + "loss": 0.565, + "step": 1926 + }, + { + "epoch": 0.07, + "learning_rate": 4.879648627853296e-05, + "loss": 0.5528, + "step": 1927 + }, + { + "epoch": 0.07, + "learning_rate": 4.879584508848423e-05, + "loss": 0.595, + "step": 1928 + }, + { + "epoch": 0.07, + "learning_rate": 4.87952038984355e-05, + "loss": 0.6222, + "step": 1929 + }, + { + "epoch": 0.07, + "learning_rate": 4.879456270838677e-05, + "loss": 0.7301, + "step": 1930 + }, + { + "epoch": 0.07, + "learning_rate": 4.8793921518338036e-05, + "loss": 0.6074, + "step": 1931 + }, + { + "epoch": 0.07, + "learning_rate": 4.879328032828931e-05, + "loss": 0.6113, + "step": 1932 + }, + { + "epoch": 0.07, + "learning_rate": 4.879263913824057e-05, + "loss": 0.5621, + "step": 1933 + }, + { + "epoch": 0.07, + "learning_rate": 4.879199794819185e-05, + "loss": 0.5604, + "step": 1934 + }, + { + "epoch": 0.07, + "learning_rate": 4.8791356758143116e-05, + "loss": 0.6102, + "step": 1935 + }, + { + "epoch": 0.07, + "learning_rate": 4.879071556809439e-05, + "loss": 0.5538, + "step": 1936 + }, + { + "epoch": 0.07, + "learning_rate": 4.879007437804565e-05, + "loss": 0.5715, + "step": 1937 + }, + { + "epoch": 0.07, + "learning_rate": 4.8789433187996924e-05, + "loss": 0.5472, + "step": 1938 + }, + { + "epoch": 0.07, + "learning_rate": 4.8788791997948196e-05, + "loss": 0.5779, + "step": 1939 + }, + { + "epoch": 0.07, + "learning_rate": 4.878815080789946e-05, + "loss": 0.6519, + "step": 1940 + }, + { + "epoch": 0.07, + "learning_rate": 4.878750961785073e-05, + "loss": 0.6301, + "step": 1941 + }, + { + "epoch": 0.07, + "learning_rate": 4.8786868427802004e-05, + "loss": 0.5443, + "step": 1942 + }, + { + "epoch": 0.07, + "learning_rate": 4.8786227237753276e-05, + "loss": 0.6331, + "step": 1943 + }, + { + "epoch": 0.07, + "learning_rate": 4.878558604770454e-05, + "loss": 0.5716, + "step": 1944 + }, + { + "epoch": 0.07, + "learning_rate": 4.878494485765581e-05, + "loss": 0.664, + "step": 1945 + }, + { + "epoch": 0.07, + "learning_rate": 4.878430366760708e-05, + "loss": 0.5228, + "step": 1946 + }, + { + "epoch": 0.07, + "learning_rate": 4.878366247755835e-05, + "loss": 0.5835, + "step": 1947 + }, + { + "epoch": 0.07, + "learning_rate": 4.878302128750962e-05, + "loss": 0.5758, + "step": 1948 + }, + { + "epoch": 0.07, + "learning_rate": 4.8782380097460886e-05, + "loss": 0.6526, + "step": 1949 + }, + { + "epoch": 0.07, + "learning_rate": 4.8781738907412164e-05, + "loss": 0.584, + "step": 1950 + }, + { + "epoch": 0.08, + "learning_rate": 4.878109771736343e-05, + "loss": 0.5634, + "step": 1951 + }, + { + "epoch": 0.08, + "learning_rate": 4.87804565273147e-05, + "loss": 0.5962, + "step": 1952 + }, + { + "epoch": 0.08, + "learning_rate": 4.8779815337265966e-05, + "loss": 0.6422, + "step": 1953 + }, + { + "epoch": 0.08, + "learning_rate": 4.877917414721724e-05, + "loss": 0.6658, + "step": 1954 + }, + { + "epoch": 0.08, + "learning_rate": 4.87785329571685e-05, + "loss": 0.5725, + "step": 1955 + }, + { + "epoch": 0.08, + "learning_rate": 4.8777891767119774e-05, + "loss": 0.6181, + "step": 1956 + }, + { + "epoch": 0.08, + "learning_rate": 4.8777250577071046e-05, + "loss": 0.6111, + "step": 1957 + }, + { + "epoch": 0.08, + "learning_rate": 4.877660938702232e-05, + "loss": 0.5611, + "step": 1958 + }, + { + "epoch": 0.08, + "learning_rate": 4.877596819697359e-05, + "loss": 0.6343, + "step": 1959 + }, + { + "epoch": 0.08, + "learning_rate": 4.8775327006924854e-05, + "loss": 0.6908, + "step": 1960 + }, + { + "epoch": 0.08, + "learning_rate": 4.8774685816876126e-05, + "loss": 0.595, + "step": 1961 + }, + { + "epoch": 0.08, + "learning_rate": 4.877404462682739e-05, + "loss": 0.6135, + "step": 1962 + }, + { + "epoch": 0.08, + "learning_rate": 4.877340343677866e-05, + "loss": 0.6177, + "step": 1963 + }, + { + "epoch": 0.08, + "learning_rate": 4.877276224672993e-05, + "loss": 0.6507, + "step": 1964 + }, + { + "epoch": 0.08, + "learning_rate": 4.87721210566812e-05, + "loss": 0.5859, + "step": 1965 + }, + { + "epoch": 0.08, + "learning_rate": 4.877147986663248e-05, + "loss": 0.6024, + "step": 1966 + }, + { + "epoch": 0.08, + "learning_rate": 4.877083867658374e-05, + "loss": 0.5839, + "step": 1967 + }, + { + "epoch": 0.08, + "learning_rate": 4.8770197486535015e-05, + "loss": 0.6807, + "step": 1968 + }, + { + "epoch": 0.08, + "learning_rate": 4.876955629648628e-05, + "loss": 0.6424, + "step": 1969 + }, + { + "epoch": 0.08, + "learning_rate": 4.876891510643755e-05, + "loss": 0.5972, + "step": 1970 + }, + { + "epoch": 0.08, + "learning_rate": 4.8768273916388816e-05, + "loss": 0.6742, + "step": 1971 + }, + { + "epoch": 0.08, + "learning_rate": 4.876763272634009e-05, + "loss": 0.5999, + "step": 1972 + }, + { + "epoch": 0.08, + "learning_rate": 4.876699153629136e-05, + "loss": 0.542, + "step": 1973 + }, + { + "epoch": 0.08, + "learning_rate": 4.876635034624263e-05, + "loss": 0.6847, + "step": 1974 + }, + { + "epoch": 0.08, + "learning_rate": 4.87657091561939e-05, + "loss": 0.6067, + "step": 1975 + }, + { + "epoch": 0.08, + "learning_rate": 4.876506796614517e-05, + "loss": 0.6505, + "step": 1976 + }, + { + "epoch": 0.08, + "learning_rate": 4.876442677609644e-05, + "loss": 0.5813, + "step": 1977 + }, + { + "epoch": 0.08, + "learning_rate": 4.8763785586047705e-05, + "loss": 0.5854, + "step": 1978 + }, + { + "epoch": 0.08, + "learning_rate": 4.8763144395998976e-05, + "loss": 0.6133, + "step": 1979 + }, + { + "epoch": 0.08, + "learning_rate": 4.876250320595024e-05, + "loss": 0.5854, + "step": 1980 + }, + { + "epoch": 0.08, + "learning_rate": 4.876186201590151e-05, + "loss": 0.6058, + "step": 1981 + }, + { + "epoch": 0.08, + "learning_rate": 4.8761220825852785e-05, + "loss": 0.7788, + "step": 1982 + }, + { + "epoch": 0.08, + "learning_rate": 4.8760579635804056e-05, + "loss": 0.6281, + "step": 1983 + }, + { + "epoch": 0.08, + "learning_rate": 4.875993844575533e-05, + "loss": 0.6201, + "step": 1984 + }, + { + "epoch": 0.08, + "learning_rate": 4.875929725570659e-05, + "loss": 0.605, + "step": 1985 + }, + { + "epoch": 0.08, + "learning_rate": 4.8758656065657865e-05, + "loss": 0.702, + "step": 1986 + }, + { + "epoch": 0.08, + "learning_rate": 4.875801487560913e-05, + "loss": 0.6272, + "step": 1987 + }, + { + "epoch": 0.08, + "learning_rate": 4.87573736855604e-05, + "loss": 0.5936, + "step": 1988 + }, + { + "epoch": 0.08, + "learning_rate": 4.875673249551167e-05, + "loss": 0.5849, + "step": 1989 + }, + { + "epoch": 0.08, + "learning_rate": 4.8756091305462945e-05, + "loss": 0.585, + "step": 1990 + }, + { + "epoch": 0.08, + "learning_rate": 4.875545011541421e-05, + "loss": 0.5868, + "step": 1991 + }, + { + "epoch": 0.08, + "learning_rate": 4.875480892536548e-05, + "loss": 0.5701, + "step": 1992 + }, + { + "epoch": 0.08, + "learning_rate": 4.875416773531675e-05, + "loss": 0.6718, + "step": 1993 + }, + { + "epoch": 0.08, + "learning_rate": 4.875352654526802e-05, + "loss": 0.6909, + "step": 1994 + }, + { + "epoch": 0.08, + "learning_rate": 4.875288535521929e-05, + "loss": 0.5308, + "step": 1995 + }, + { + "epoch": 0.08, + "learning_rate": 4.8752244165170555e-05, + "loss": 0.5639, + "step": 1996 + }, + { + "epoch": 0.08, + "learning_rate": 4.8751602975121827e-05, + "loss": 0.6452, + "step": 1997 + }, + { + "epoch": 0.08, + "learning_rate": 4.87509617850731e-05, + "loss": 0.6824, + "step": 1998 + }, + { + "epoch": 0.08, + "learning_rate": 4.875032059502437e-05, + "loss": 0.6205, + "step": 1999 + }, + { + "epoch": 0.08, + "learning_rate": 4.8749679404975635e-05, + "loss": 0.6132, + "step": 2000 + }, + { + "epoch": 0.08, + "learning_rate": 4.874903821492691e-05, + "loss": 0.5818, + "step": 2001 + }, + { + "epoch": 0.08, + "learning_rate": 4.874839702487818e-05, + "loss": 0.6644, + "step": 2002 + }, + { + "epoch": 0.08, + "learning_rate": 4.874775583482944e-05, + "loss": 0.6289, + "step": 2003 + }, + { + "epoch": 0.08, + "learning_rate": 4.8747114644780715e-05, + "loss": 0.6919, + "step": 2004 + }, + { + "epoch": 0.08, + "learning_rate": 4.874647345473199e-05, + "loss": 0.6568, + "step": 2005 + }, + { + "epoch": 0.08, + "learning_rate": 4.874583226468326e-05, + "loss": 0.5291, + "step": 2006 + }, + { + "epoch": 0.08, + "learning_rate": 4.8745191074634523e-05, + "loss": 0.5969, + "step": 2007 + }, + { + "epoch": 0.08, + "learning_rate": 4.8744549884585795e-05, + "loss": 0.6067, + "step": 2008 + }, + { + "epoch": 0.08, + "learning_rate": 4.874390869453706e-05, + "loss": 0.7035, + "step": 2009 + }, + { + "epoch": 0.08, + "learning_rate": 4.874326750448833e-05, + "loss": 0.5688, + "step": 2010 + }, + { + "epoch": 0.08, + "learning_rate": 4.8742626314439604e-05, + "loss": 0.63, + "step": 2011 + }, + { + "epoch": 0.08, + "learning_rate": 4.874198512439087e-05, + "loss": 0.6587, + "step": 2012 + }, + { + "epoch": 0.08, + "learning_rate": 4.874134393434214e-05, + "loss": 0.657, + "step": 2013 + }, + { + "epoch": 0.08, + "learning_rate": 4.874070274429341e-05, + "loss": 0.6463, + "step": 2014 + }, + { + "epoch": 0.08, + "learning_rate": 4.8740061554244684e-05, + "loss": 0.5687, + "step": 2015 + }, + { + "epoch": 0.08, + "learning_rate": 4.873942036419595e-05, + "loss": 0.7023, + "step": 2016 + }, + { + "epoch": 0.08, + "learning_rate": 4.873877917414722e-05, + "loss": 0.6479, + "step": 2017 + }, + { + "epoch": 0.08, + "learning_rate": 4.8738137984098485e-05, + "loss": 0.6219, + "step": 2018 + }, + { + "epoch": 0.08, + "learning_rate": 4.873749679404976e-05, + "loss": 0.6684, + "step": 2019 + }, + { + "epoch": 0.08, + "learning_rate": 4.873685560400103e-05, + "loss": 0.5434, + "step": 2020 + }, + { + "epoch": 0.08, + "learning_rate": 4.8736214413952294e-05, + "loss": 0.5572, + "step": 2021 + }, + { + "epoch": 0.08, + "learning_rate": 4.873557322390357e-05, + "loss": 0.6421, + "step": 2022 + }, + { + "epoch": 0.08, + "learning_rate": 4.873493203385484e-05, + "loss": 0.7051, + "step": 2023 + }, + { + "epoch": 0.08, + "learning_rate": 4.873429084380611e-05, + "loss": 0.5695, + "step": 2024 + }, + { + "epoch": 0.08, + "learning_rate": 4.8733649653757374e-05, + "loss": 0.616, + "step": 2025 + }, + { + "epoch": 0.08, + "learning_rate": 4.8733008463708645e-05, + "loss": 0.6535, + "step": 2026 + }, + { + "epoch": 0.08, + "learning_rate": 4.873236727365991e-05, + "loss": 0.6629, + "step": 2027 + }, + { + "epoch": 0.08, + "learning_rate": 4.873172608361118e-05, + "loss": 0.6259, + "step": 2028 + }, + { + "epoch": 0.08, + "learning_rate": 4.8731084893562454e-05, + "loss": 0.6701, + "step": 2029 + }, + { + "epoch": 0.08, + "learning_rate": 4.8730443703513725e-05, + "loss": 0.5325, + "step": 2030 + }, + { + "epoch": 0.08, + "learning_rate": 4.8729802513465e-05, + "loss": 0.6557, + "step": 2031 + }, + { + "epoch": 0.08, + "learning_rate": 4.872916132341626e-05, + "loss": 0.5836, + "step": 2032 + }, + { + "epoch": 0.08, + "learning_rate": 4.8728520133367534e-05, + "loss": 0.6297, + "step": 2033 + }, + { + "epoch": 0.08, + "learning_rate": 4.87278789433188e-05, + "loss": 0.6296, + "step": 2034 + }, + { + "epoch": 0.08, + "learning_rate": 4.872723775327007e-05, + "loss": 0.6093, + "step": 2035 + }, + { + "epoch": 0.08, + "learning_rate": 4.8726596563221335e-05, + "loss": 0.5989, + "step": 2036 + }, + { + "epoch": 0.08, + "learning_rate": 4.872595537317261e-05, + "loss": 0.5982, + "step": 2037 + }, + { + "epoch": 0.08, + "learning_rate": 4.8725314183123886e-05, + "loss": 0.5456, + "step": 2038 + }, + { + "epoch": 0.08, + "learning_rate": 4.872467299307515e-05, + "loss": 0.5665, + "step": 2039 + }, + { + "epoch": 0.08, + "learning_rate": 4.872403180302642e-05, + "loss": 0.5833, + "step": 2040 + }, + { + "epoch": 0.08, + "learning_rate": 4.872339061297769e-05, + "loss": 0.5769, + "step": 2041 + }, + { + "epoch": 0.08, + "learning_rate": 4.872274942292896e-05, + "loss": 0.6448, + "step": 2042 + }, + { + "epoch": 0.08, + "learning_rate": 4.8722108232880224e-05, + "loss": 0.6299, + "step": 2043 + }, + { + "epoch": 0.08, + "learning_rate": 4.8721467042831496e-05, + "loss": 0.5972, + "step": 2044 + }, + { + "epoch": 0.08, + "learning_rate": 4.872082585278277e-05, + "loss": 0.549, + "step": 2045 + }, + { + "epoch": 0.08, + "learning_rate": 4.872018466273404e-05, + "loss": 0.6087, + "step": 2046 + }, + { + "epoch": 0.08, + "learning_rate": 4.871954347268531e-05, + "loss": 0.5925, + "step": 2047 + }, + { + "epoch": 0.08, + "learning_rate": 4.8718902282636576e-05, + "loss": 0.5374, + "step": 2048 + }, + { + "epoch": 0.08, + "learning_rate": 4.871826109258785e-05, + "loss": 0.579, + "step": 2049 + }, + { + "epoch": 0.08, + "learning_rate": 4.871761990253911e-05, + "loss": 0.5307, + "step": 2050 + }, + { + "epoch": 0.08, + "learning_rate": 4.8716978712490384e-05, + "loss": 0.6172, + "step": 2051 + }, + { + "epoch": 0.08, + "learning_rate": 4.871633752244165e-05, + "loss": 0.6372, + "step": 2052 + }, + { + "epoch": 0.08, + "learning_rate": 4.871569633239292e-05, + "loss": 0.64, + "step": 2053 + }, + { + "epoch": 0.08, + "learning_rate": 4.87150551423442e-05, + "loss": 0.7051, + "step": 2054 + }, + { + "epoch": 0.08, + "learning_rate": 4.8714413952295464e-05, + "loss": 0.6517, + "step": 2055 + }, + { + "epoch": 0.08, + "learning_rate": 4.8713772762246736e-05, + "loss": 0.6921, + "step": 2056 + }, + { + "epoch": 0.08, + "learning_rate": 4.8713131572198e-05, + "loss": 0.5696, + "step": 2057 + }, + { + "epoch": 0.08, + "learning_rate": 4.871249038214927e-05, + "loss": 0.6658, + "step": 2058 + }, + { + "epoch": 0.08, + "learning_rate": 4.871184919210054e-05, + "loss": 0.5546, + "step": 2059 + }, + { + "epoch": 0.08, + "learning_rate": 4.871120800205181e-05, + "loss": 0.5976, + "step": 2060 + }, + { + "epoch": 0.08, + "learning_rate": 4.871056681200308e-05, + "loss": 0.6195, + "step": 2061 + }, + { + "epoch": 0.08, + "learning_rate": 4.870992562195435e-05, + "loss": 0.6153, + "step": 2062 + }, + { + "epoch": 0.08, + "learning_rate": 4.8709284431905624e-05, + "loss": 0.6809, + "step": 2063 + }, + { + "epoch": 0.08, + "learning_rate": 4.870864324185689e-05, + "loss": 0.6701, + "step": 2064 + }, + { + "epoch": 0.08, + "learning_rate": 4.870800205180816e-05, + "loss": 0.6335, + "step": 2065 + }, + { + "epoch": 0.08, + "learning_rate": 4.8707360861759426e-05, + "loss": 0.6249, + "step": 2066 + }, + { + "epoch": 0.08, + "learning_rate": 4.87067196717107e-05, + "loss": 0.6994, + "step": 2067 + }, + { + "epoch": 0.08, + "learning_rate": 4.870607848166196e-05, + "loss": 0.6526, + "step": 2068 + }, + { + "epoch": 0.08, + "learning_rate": 4.8705437291613234e-05, + "loss": 0.6362, + "step": 2069 + }, + { + "epoch": 0.08, + "learning_rate": 4.8704796101564506e-05, + "loss": 0.5909, + "step": 2070 + }, + { + "epoch": 0.08, + "learning_rate": 4.870415491151578e-05, + "loss": 0.5845, + "step": 2071 + }, + { + "epoch": 0.08, + "learning_rate": 4.870351372146705e-05, + "loss": 0.651, + "step": 2072 + }, + { + "epoch": 0.08, + "learning_rate": 4.8702872531418314e-05, + "loss": 0.6038, + "step": 2073 + }, + { + "epoch": 0.08, + "learning_rate": 4.8702231341369586e-05, + "loss": 0.589, + "step": 2074 + }, + { + "epoch": 0.08, + "learning_rate": 4.870159015132085e-05, + "loss": 0.5965, + "step": 2075 + }, + { + "epoch": 0.08, + "learning_rate": 4.870094896127212e-05, + "loss": 0.5941, + "step": 2076 + }, + { + "epoch": 0.08, + "learning_rate": 4.8700307771223395e-05, + "loss": 0.6019, + "step": 2077 + }, + { + "epoch": 0.08, + "learning_rate": 4.8699666581174666e-05, + "loss": 0.6082, + "step": 2078 + }, + { + "epoch": 0.08, + "learning_rate": 4.869902539112593e-05, + "loss": 0.6446, + "step": 2079 + }, + { + "epoch": 0.08, + "learning_rate": 4.86983842010772e-05, + "loss": 0.6271, + "step": 2080 + }, + { + "epoch": 0.08, + "learning_rate": 4.8697743011028475e-05, + "loss": 0.5734, + "step": 2081 + }, + { + "epoch": 0.08, + "learning_rate": 4.869710182097974e-05, + "loss": 0.5891, + "step": 2082 + }, + { + "epoch": 0.08, + "learning_rate": 4.869646063093101e-05, + "loss": 0.6945, + "step": 2083 + }, + { + "epoch": 0.08, + "learning_rate": 4.8695819440882276e-05, + "loss": 0.5222, + "step": 2084 + }, + { + "epoch": 0.08, + "learning_rate": 4.869517825083355e-05, + "loss": 0.6309, + "step": 2085 + }, + { + "epoch": 0.08, + "learning_rate": 4.869453706078482e-05, + "loss": 0.6462, + "step": 2086 + }, + { + "epoch": 0.08, + "learning_rate": 4.869389587073609e-05, + "loss": 0.6294, + "step": 2087 + }, + { + "epoch": 0.08, + "learning_rate": 4.8693254680687356e-05, + "loss": 0.6094, + "step": 2088 + }, + { + "epoch": 0.08, + "learning_rate": 4.869261349063863e-05, + "loss": 0.6052, + "step": 2089 + }, + { + "epoch": 0.08, + "learning_rate": 4.86919723005899e-05, + "loss": 0.5657, + "step": 2090 + }, + { + "epoch": 0.08, + "learning_rate": 4.8691331110541165e-05, + "loss": 0.5609, + "step": 2091 + }, + { + "epoch": 0.08, + "learning_rate": 4.8690689920492436e-05, + "loss": 0.6405, + "step": 2092 + }, + { + "epoch": 0.08, + "learning_rate": 4.869004873044371e-05, + "loss": 0.706, + "step": 2093 + }, + { + "epoch": 0.08, + "learning_rate": 4.868940754039498e-05, + "loss": 0.6146, + "step": 2094 + }, + { + "epoch": 0.08, + "learning_rate": 4.8688766350346245e-05, + "loss": 0.6173, + "step": 2095 + }, + { + "epoch": 0.08, + "learning_rate": 4.8688125160297516e-05, + "loss": 0.5499, + "step": 2096 + }, + { + "epoch": 0.08, + "learning_rate": 4.868748397024878e-05, + "loss": 0.6117, + "step": 2097 + }, + { + "epoch": 0.08, + "learning_rate": 4.868684278020005e-05, + "loss": 0.6316, + "step": 2098 + }, + { + "epoch": 0.08, + "learning_rate": 4.8686201590151325e-05, + "loss": 0.6414, + "step": 2099 + }, + { + "epoch": 0.08, + "learning_rate": 4.868556040010259e-05, + "loss": 0.5267, + "step": 2100 + }, + { + "epoch": 0.08, + "learning_rate": 4.868491921005386e-05, + "loss": 0.6117, + "step": 2101 + }, + { + "epoch": 0.08, + "learning_rate": 4.868427802000513e-05, + "loss": 0.6, + "step": 2102 + }, + { + "epoch": 0.08, + "learning_rate": 4.8683636829956405e-05, + "loss": 0.6201, + "step": 2103 + }, + { + "epoch": 0.08, + "learning_rate": 4.868299563990767e-05, + "loss": 0.6936, + "step": 2104 + }, + { + "epoch": 0.08, + "learning_rate": 4.868235444985894e-05, + "loss": 0.6606, + "step": 2105 + }, + { + "epoch": 0.08, + "learning_rate": 4.8681713259810207e-05, + "loss": 0.6144, + "step": 2106 + }, + { + "epoch": 0.08, + "learning_rate": 4.868107206976148e-05, + "loss": 0.618, + "step": 2107 + }, + { + "epoch": 0.08, + "learning_rate": 4.868043087971275e-05, + "loss": 0.5864, + "step": 2108 + }, + { + "epoch": 0.08, + "learning_rate": 4.8679789689664015e-05, + "loss": 0.5251, + "step": 2109 + }, + { + "epoch": 0.08, + "learning_rate": 4.8679148499615293e-05, + "loss": 0.5311, + "step": 2110 + }, + { + "epoch": 0.08, + "learning_rate": 4.867850730956656e-05, + "loss": 0.6838, + "step": 2111 + }, + { + "epoch": 0.08, + "learning_rate": 4.867786611951783e-05, + "loss": 0.6209, + "step": 2112 + }, + { + "epoch": 0.08, + "learning_rate": 4.8677224929469095e-05, + "loss": 0.664, + "step": 2113 + }, + { + "epoch": 0.08, + "learning_rate": 4.867658373942037e-05, + "loss": 0.6296, + "step": 2114 + }, + { + "epoch": 0.08, + "learning_rate": 4.867594254937163e-05, + "loss": 0.6189, + "step": 2115 + }, + { + "epoch": 0.08, + "learning_rate": 4.86753013593229e-05, + "loss": 0.5662, + "step": 2116 + }, + { + "epoch": 0.08, + "learning_rate": 4.8674660169274175e-05, + "loss": 0.6322, + "step": 2117 + }, + { + "epoch": 0.08, + "learning_rate": 4.867401897922545e-05, + "loss": 0.6927, + "step": 2118 + }, + { + "epoch": 0.08, + "learning_rate": 4.867337778917672e-05, + "loss": 0.6293, + "step": 2119 + }, + { + "epoch": 0.08, + "learning_rate": 4.8672736599127983e-05, + "loss": 0.6963, + "step": 2120 + }, + { + "epoch": 0.08, + "learning_rate": 4.8672095409079255e-05, + "loss": 0.6762, + "step": 2121 + }, + { + "epoch": 0.08, + "learning_rate": 4.867145421903052e-05, + "loss": 0.6835, + "step": 2122 + }, + { + "epoch": 0.08, + "learning_rate": 4.867081302898179e-05, + "loss": 0.6095, + "step": 2123 + }, + { + "epoch": 0.08, + "learning_rate": 4.867017183893306e-05, + "loss": 0.5673, + "step": 2124 + }, + { + "epoch": 0.08, + "learning_rate": 4.866953064888433e-05, + "loss": 0.5191, + "step": 2125 + }, + { + "epoch": 0.08, + "learning_rate": 4.866888945883561e-05, + "loss": 0.6553, + "step": 2126 + }, + { + "epoch": 0.08, + "learning_rate": 4.866824826878687e-05, + "loss": 0.6594, + "step": 2127 + }, + { + "epoch": 0.08, + "learning_rate": 4.8667607078738144e-05, + "loss": 0.5812, + "step": 2128 + }, + { + "epoch": 0.08, + "learning_rate": 4.866696588868941e-05, + "loss": 0.5523, + "step": 2129 + }, + { + "epoch": 0.08, + "learning_rate": 4.866632469864068e-05, + "loss": 0.5867, + "step": 2130 + }, + { + "epoch": 0.08, + "learning_rate": 4.8665683508591945e-05, + "loss": 0.6198, + "step": 2131 + }, + { + "epoch": 0.08, + "learning_rate": 4.866504231854322e-05, + "loss": 0.654, + "step": 2132 + }, + { + "epoch": 0.08, + "learning_rate": 4.866440112849449e-05, + "loss": 0.6445, + "step": 2133 + }, + { + "epoch": 0.08, + "learning_rate": 4.866375993844576e-05, + "loss": 0.6036, + "step": 2134 + }, + { + "epoch": 0.08, + "learning_rate": 4.866311874839703e-05, + "loss": 0.5906, + "step": 2135 + }, + { + "epoch": 0.08, + "learning_rate": 4.86624775583483e-05, + "loss": 0.5831, + "step": 2136 + }, + { + "epoch": 0.08, + "learning_rate": 4.866183636829957e-05, + "loss": 0.6036, + "step": 2137 + }, + { + "epoch": 0.08, + "learning_rate": 4.8661195178250834e-05, + "loss": 0.6155, + "step": 2138 + }, + { + "epoch": 0.08, + "learning_rate": 4.8660553988202105e-05, + "loss": 0.5803, + "step": 2139 + }, + { + "epoch": 0.08, + "learning_rate": 4.865991279815337e-05, + "loss": 0.6205, + "step": 2140 + }, + { + "epoch": 0.08, + "learning_rate": 4.865927160810464e-05, + "loss": 0.6661, + "step": 2141 + }, + { + "epoch": 0.08, + "learning_rate": 4.8658630418055914e-05, + "loss": 0.5773, + "step": 2142 + }, + { + "epoch": 0.08, + "learning_rate": 4.8657989228007186e-05, + "loss": 0.5538, + "step": 2143 + }, + { + "epoch": 0.08, + "learning_rate": 4.865734803795846e-05, + "loss": 0.5649, + "step": 2144 + }, + { + "epoch": 0.08, + "learning_rate": 4.865670684790972e-05, + "loss": 0.5696, + "step": 2145 + }, + { + "epoch": 0.08, + "learning_rate": 4.8656065657860994e-05, + "loss": 0.5144, + "step": 2146 + }, + { + "epoch": 0.08, + "learning_rate": 4.865542446781226e-05, + "loss": 0.5694, + "step": 2147 + }, + { + "epoch": 0.08, + "learning_rate": 4.865478327776353e-05, + "loss": 0.5939, + "step": 2148 + }, + { + "epoch": 0.08, + "learning_rate": 4.86541420877148e-05, + "loss": 0.5872, + "step": 2149 + }, + { + "epoch": 0.08, + "learning_rate": 4.8653500897666074e-05, + "loss": 0.6207, + "step": 2150 + }, + { + "epoch": 0.08, + "learning_rate": 4.865285970761734e-05, + "loss": 0.5524, + "step": 2151 + }, + { + "epoch": 0.08, + "learning_rate": 4.865221851756861e-05, + "loss": 0.5911, + "step": 2152 + }, + { + "epoch": 0.08, + "learning_rate": 4.865157732751988e-05, + "loss": 0.7259, + "step": 2153 + }, + { + "epoch": 0.08, + "learning_rate": 4.865093613747115e-05, + "loss": 0.6452, + "step": 2154 + }, + { + "epoch": 0.08, + "learning_rate": 4.865029494742242e-05, + "loss": 0.4494, + "step": 2155 + }, + { + "epoch": 0.08, + "learning_rate": 4.8649653757373684e-05, + "loss": 0.6007, + "step": 2156 + }, + { + "epoch": 0.08, + "learning_rate": 4.8649012567324956e-05, + "loss": 0.5753, + "step": 2157 + }, + { + "epoch": 0.08, + "learning_rate": 4.864837137727623e-05, + "loss": 0.6679, + "step": 2158 + }, + { + "epoch": 0.08, + "learning_rate": 4.86477301872275e-05, + "loss": 0.6274, + "step": 2159 + }, + { + "epoch": 0.08, + "learning_rate": 4.8647088997178764e-05, + "loss": 0.6741, + "step": 2160 + }, + { + "epoch": 0.08, + "learning_rate": 4.8646447807130036e-05, + "loss": 0.6151, + "step": 2161 + }, + { + "epoch": 0.08, + "learning_rate": 4.864580661708131e-05, + "loss": 0.6264, + "step": 2162 + }, + { + "epoch": 0.08, + "learning_rate": 4.864516542703257e-05, + "loss": 0.6311, + "step": 2163 + }, + { + "epoch": 0.08, + "learning_rate": 4.8644524236983844e-05, + "loss": 0.6161, + "step": 2164 + }, + { + "epoch": 0.08, + "learning_rate": 4.8643883046935116e-05, + "loss": 0.5773, + "step": 2165 + }, + { + "epoch": 0.08, + "learning_rate": 4.864324185688639e-05, + "loss": 0.6097, + "step": 2166 + }, + { + "epoch": 0.08, + "learning_rate": 4.864260066683765e-05, + "loss": 0.5846, + "step": 2167 + }, + { + "epoch": 0.08, + "learning_rate": 4.8641959476788924e-05, + "loss": 0.6785, + "step": 2168 + }, + { + "epoch": 0.08, + "learning_rate": 4.864131828674019e-05, + "loss": 0.5607, + "step": 2169 + }, + { + "epoch": 0.08, + "learning_rate": 4.864067709669146e-05, + "loss": 0.6124, + "step": 2170 + }, + { + "epoch": 0.08, + "learning_rate": 4.864003590664273e-05, + "loss": 0.6033, + "step": 2171 + }, + { + "epoch": 0.08, + "learning_rate": 4.8639394716594e-05, + "loss": 0.5021, + "step": 2172 + }, + { + "epoch": 0.08, + "learning_rate": 4.863875352654527e-05, + "loss": 0.5789, + "step": 2173 + }, + { + "epoch": 0.08, + "learning_rate": 4.863811233649654e-05, + "loss": 0.6078, + "step": 2174 + }, + { + "epoch": 0.08, + "learning_rate": 4.863747114644781e-05, + "loss": 0.5639, + "step": 2175 + }, + { + "epoch": 0.08, + "learning_rate": 4.863682995639908e-05, + "loss": 0.6212, + "step": 2176 + }, + { + "epoch": 0.08, + "learning_rate": 4.863618876635035e-05, + "loss": 0.5788, + "step": 2177 + }, + { + "epoch": 0.08, + "learning_rate": 4.8635547576301614e-05, + "loss": 0.5021, + "step": 2178 + }, + { + "epoch": 0.08, + "learning_rate": 4.8634906386252886e-05, + "loss": 0.5758, + "step": 2179 + }, + { + "epoch": 0.08, + "learning_rate": 4.863426519620416e-05, + "loss": 0.6485, + "step": 2180 + }, + { + "epoch": 0.08, + "learning_rate": 4.863362400615542e-05, + "loss": 0.6013, + "step": 2181 + }, + { + "epoch": 0.08, + "learning_rate": 4.86329828161067e-05, + "loss": 0.656, + "step": 2182 + }, + { + "epoch": 0.08, + "learning_rate": 4.8632341626057966e-05, + "loss": 0.6913, + "step": 2183 + }, + { + "epoch": 0.08, + "learning_rate": 4.863170043600924e-05, + "loss": 0.6421, + "step": 2184 + }, + { + "epoch": 0.08, + "learning_rate": 4.86310592459605e-05, + "loss": 0.6415, + "step": 2185 + }, + { + "epoch": 0.08, + "learning_rate": 4.8630418055911774e-05, + "loss": 0.5955, + "step": 2186 + }, + { + "epoch": 0.08, + "learning_rate": 4.862977686586304e-05, + "loss": 0.5389, + "step": 2187 + }, + { + "epoch": 0.08, + "learning_rate": 4.862913567581431e-05, + "loss": 0.7388, + "step": 2188 + }, + { + "epoch": 0.08, + "learning_rate": 4.862849448576558e-05, + "loss": 0.614, + "step": 2189 + }, + { + "epoch": 0.08, + "learning_rate": 4.8627853295716855e-05, + "loss": 0.6931, + "step": 2190 + }, + { + "epoch": 0.08, + "learning_rate": 4.8627212105668126e-05, + "loss": 0.5913, + "step": 2191 + }, + { + "epoch": 0.08, + "learning_rate": 4.862657091561939e-05, + "loss": 0.6201, + "step": 2192 + }, + { + "epoch": 0.08, + "learning_rate": 4.862592972557066e-05, + "loss": 0.5762, + "step": 2193 + }, + { + "epoch": 0.08, + "learning_rate": 4.862528853552193e-05, + "loss": 0.5514, + "step": 2194 + }, + { + "epoch": 0.08, + "learning_rate": 4.86246473454732e-05, + "loss": 0.6147, + "step": 2195 + }, + { + "epoch": 0.08, + "learning_rate": 4.8624006155424465e-05, + "loss": 0.5656, + "step": 2196 + }, + { + "epoch": 0.08, + "learning_rate": 4.8623364965375736e-05, + "loss": 0.6131, + "step": 2197 + }, + { + "epoch": 0.08, + "learning_rate": 4.8622723775327015e-05, + "loss": 0.649, + "step": 2198 + }, + { + "epoch": 0.08, + "learning_rate": 4.862208258527828e-05, + "loss": 0.6293, + "step": 2199 + }, + { + "epoch": 0.08, + "learning_rate": 4.862144139522955e-05, + "loss": 0.5807, + "step": 2200 + }, + { + "epoch": 0.08, + "learning_rate": 4.8620800205180816e-05, + "loss": 0.675, + "step": 2201 + }, + { + "epoch": 0.08, + "learning_rate": 4.862015901513209e-05, + "loss": 0.5949, + "step": 2202 + }, + { + "epoch": 0.08, + "learning_rate": 4.861951782508335e-05, + "loss": 0.6271, + "step": 2203 + }, + { + "epoch": 0.08, + "learning_rate": 4.8618876635034625e-05, + "loss": 0.6587, + "step": 2204 + }, + { + "epoch": 0.08, + "learning_rate": 4.8618235444985896e-05, + "loss": 0.635, + "step": 2205 + }, + { + "epoch": 0.08, + "learning_rate": 4.861759425493717e-05, + "loss": 0.6052, + "step": 2206 + }, + { + "epoch": 0.08, + "learning_rate": 4.861695306488844e-05, + "loss": 0.6112, + "step": 2207 + }, + { + "epoch": 0.08, + "learning_rate": 4.8616311874839705e-05, + "loss": 0.6892, + "step": 2208 + }, + { + "epoch": 0.08, + "learning_rate": 4.8615670684790977e-05, + "loss": 0.5682, + "step": 2209 + }, + { + "epoch": 0.08, + "learning_rate": 4.861502949474224e-05, + "loss": 0.6415, + "step": 2210 + }, + { + "epoch": 0.09, + "learning_rate": 4.861438830469351e-05, + "loss": 0.5642, + "step": 2211 + }, + { + "epoch": 0.09, + "learning_rate": 4.861374711464478e-05, + "loss": 0.6398, + "step": 2212 + }, + { + "epoch": 0.09, + "learning_rate": 4.861310592459605e-05, + "loss": 0.6433, + "step": 2213 + }, + { + "epoch": 0.09, + "learning_rate": 4.861246473454732e-05, + "loss": 0.6022, + "step": 2214 + }, + { + "epoch": 0.09, + "learning_rate": 4.861182354449859e-05, + "loss": 0.7042, + "step": 2215 + }, + { + "epoch": 0.09, + "learning_rate": 4.8611182354449865e-05, + "loss": 0.6465, + "step": 2216 + }, + { + "epoch": 0.09, + "learning_rate": 4.861054116440113e-05, + "loss": 0.6248, + "step": 2217 + }, + { + "epoch": 0.09, + "learning_rate": 4.86098999743524e-05, + "loss": 0.687, + "step": 2218 + }, + { + "epoch": 0.09, + "learning_rate": 4.8609258784303667e-05, + "loss": 0.6054, + "step": 2219 + }, + { + "epoch": 0.09, + "learning_rate": 4.860861759425494e-05, + "loss": 0.6891, + "step": 2220 + }, + { + "epoch": 0.09, + "learning_rate": 4.860797640420621e-05, + "loss": 0.647, + "step": 2221 + }, + { + "epoch": 0.09, + "learning_rate": 4.860733521415748e-05, + "loss": 0.6693, + "step": 2222 + }, + { + "epoch": 0.09, + "learning_rate": 4.860669402410875e-05, + "loss": 0.6419, + "step": 2223 + }, + { + "epoch": 0.09, + "learning_rate": 4.860605283406002e-05, + "loss": 0.6725, + "step": 2224 + }, + { + "epoch": 0.09, + "learning_rate": 4.860541164401129e-05, + "loss": 0.5911, + "step": 2225 + }, + { + "epoch": 0.09, + "learning_rate": 4.8604770453962555e-05, + "loss": 0.6126, + "step": 2226 + }, + { + "epoch": 0.09, + "learning_rate": 4.860412926391383e-05, + "loss": 0.5843, + "step": 2227 + }, + { + "epoch": 0.09, + "learning_rate": 4.860348807386509e-05, + "loss": 0.6526, + "step": 2228 + }, + { + "epoch": 0.09, + "learning_rate": 4.8602846883816363e-05, + "loss": 0.6444, + "step": 2229 + }, + { + "epoch": 0.09, + "learning_rate": 4.8602205693767635e-05, + "loss": 0.6158, + "step": 2230 + }, + { + "epoch": 0.09, + "learning_rate": 4.860156450371891e-05, + "loss": 0.6544, + "step": 2231 + }, + { + "epoch": 0.09, + "learning_rate": 4.860092331367018e-05, + "loss": 0.6796, + "step": 2232 + }, + { + "epoch": 0.09, + "learning_rate": 4.8600282123621444e-05, + "loss": 0.7162, + "step": 2233 + }, + { + "epoch": 0.09, + "learning_rate": 4.8599640933572715e-05, + "loss": 0.5953, + "step": 2234 + }, + { + "epoch": 0.09, + "learning_rate": 4.859899974352398e-05, + "loss": 0.5041, + "step": 2235 + }, + { + "epoch": 0.09, + "learning_rate": 4.859835855347525e-05, + "loss": 0.6489, + "step": 2236 + }, + { + "epoch": 0.09, + "learning_rate": 4.8597717363426524e-05, + "loss": 0.6956, + "step": 2237 + }, + { + "epoch": 0.09, + "learning_rate": 4.8597076173377795e-05, + "loss": 0.6327, + "step": 2238 + }, + { + "epoch": 0.09, + "learning_rate": 4.859643498332906e-05, + "loss": 0.6763, + "step": 2239 + }, + { + "epoch": 0.09, + "learning_rate": 4.859579379328033e-05, + "loss": 0.6409, + "step": 2240 + }, + { + "epoch": 0.09, + "learning_rate": 4.8595152603231604e-05, + "loss": 0.5568, + "step": 2241 + }, + { + "epoch": 0.09, + "learning_rate": 4.859451141318287e-05, + "loss": 0.5881, + "step": 2242 + }, + { + "epoch": 0.09, + "learning_rate": 4.859387022313414e-05, + "loss": 0.63, + "step": 2243 + }, + { + "epoch": 0.09, + "learning_rate": 4.8593229033085405e-05, + "loss": 0.6065, + "step": 2244 + }, + { + "epoch": 0.09, + "learning_rate": 4.859258784303668e-05, + "loss": 0.6482, + "step": 2245 + }, + { + "epoch": 0.09, + "learning_rate": 4.859194665298795e-05, + "loss": 0.599, + "step": 2246 + }, + { + "epoch": 0.09, + "learning_rate": 4.859130546293922e-05, + "loss": 0.5569, + "step": 2247 + }, + { + "epoch": 0.09, + "learning_rate": 4.8590664272890485e-05, + "loss": 0.6408, + "step": 2248 + }, + { + "epoch": 0.09, + "learning_rate": 4.859002308284176e-05, + "loss": 0.6984, + "step": 2249 + }, + { + "epoch": 0.09, + "learning_rate": 4.858938189279303e-05, + "loss": 0.6598, + "step": 2250 + }, + { + "epoch": 0.09, + "learning_rate": 4.8588740702744294e-05, + "loss": 0.5853, + "step": 2251 + }, + { + "epoch": 0.09, + "learning_rate": 4.8588099512695565e-05, + "loss": 0.5966, + "step": 2252 + }, + { + "epoch": 0.09, + "learning_rate": 4.858745832264684e-05, + "loss": 0.564, + "step": 2253 + }, + { + "epoch": 0.09, + "learning_rate": 4.858681713259811e-05, + "loss": 0.6112, + "step": 2254 + }, + { + "epoch": 0.09, + "learning_rate": 4.8586175942549374e-05, + "loss": 0.5761, + "step": 2255 + }, + { + "epoch": 0.09, + "learning_rate": 4.8585534752500646e-05, + "loss": 0.5189, + "step": 2256 + }, + { + "epoch": 0.09, + "learning_rate": 4.858489356245191e-05, + "loss": 0.5592, + "step": 2257 + }, + { + "epoch": 0.09, + "learning_rate": 4.858425237240318e-05, + "loss": 0.6267, + "step": 2258 + }, + { + "epoch": 0.09, + "learning_rate": 4.8583611182354454e-05, + "loss": 0.5668, + "step": 2259 + }, + { + "epoch": 0.09, + "learning_rate": 4.858296999230572e-05, + "loss": 0.5899, + "step": 2260 + }, + { + "epoch": 0.09, + "learning_rate": 4.858232880225699e-05, + "loss": 0.676, + "step": 2261 + }, + { + "epoch": 0.09, + "learning_rate": 4.858168761220826e-05, + "loss": 0.7028, + "step": 2262 + }, + { + "epoch": 0.09, + "learning_rate": 4.8581046422159534e-05, + "loss": 0.7725, + "step": 2263 + }, + { + "epoch": 0.09, + "learning_rate": 4.85804052321108e-05, + "loss": 0.6076, + "step": 2264 + }, + { + "epoch": 0.09, + "learning_rate": 4.857976404206207e-05, + "loss": 0.5816, + "step": 2265 + }, + { + "epoch": 0.09, + "learning_rate": 4.8579122852013336e-05, + "loss": 0.4928, + "step": 2266 + }, + { + "epoch": 0.09, + "learning_rate": 4.857848166196461e-05, + "loss": 0.6202, + "step": 2267 + }, + { + "epoch": 0.09, + "learning_rate": 4.857784047191588e-05, + "loss": 0.5996, + "step": 2268 + }, + { + "epoch": 0.09, + "learning_rate": 4.8577199281867144e-05, + "loss": 0.6668, + "step": 2269 + }, + { + "epoch": 0.09, + "learning_rate": 4.857655809181842e-05, + "loss": 0.629, + "step": 2270 + }, + { + "epoch": 0.09, + "learning_rate": 4.857591690176969e-05, + "loss": 0.6755, + "step": 2271 + }, + { + "epoch": 0.09, + "learning_rate": 4.857527571172096e-05, + "loss": 0.7112, + "step": 2272 + }, + { + "epoch": 0.09, + "learning_rate": 4.8574634521672224e-05, + "loss": 0.7101, + "step": 2273 + }, + { + "epoch": 0.09, + "learning_rate": 4.8573993331623496e-05, + "loss": 0.6397, + "step": 2274 + }, + { + "epoch": 0.09, + "learning_rate": 4.857335214157476e-05, + "loss": 0.5886, + "step": 2275 + }, + { + "epoch": 0.09, + "learning_rate": 4.857271095152603e-05, + "loss": 0.621, + "step": 2276 + }, + { + "epoch": 0.09, + "learning_rate": 4.8572069761477304e-05, + "loss": 0.6247, + "step": 2277 + }, + { + "epoch": 0.09, + "learning_rate": 4.8571428571428576e-05, + "loss": 0.6524, + "step": 2278 + }, + { + "epoch": 0.09, + "learning_rate": 4.857078738137985e-05, + "loss": 0.6474, + "step": 2279 + }, + { + "epoch": 0.09, + "learning_rate": 4.857014619133111e-05, + "loss": 0.6488, + "step": 2280 + }, + { + "epoch": 0.09, + "learning_rate": 4.8569505001282384e-05, + "loss": 0.6678, + "step": 2281 + }, + { + "epoch": 0.09, + "learning_rate": 4.856886381123365e-05, + "loss": 0.6529, + "step": 2282 + }, + { + "epoch": 0.09, + "learning_rate": 4.856822262118492e-05, + "loss": 0.6093, + "step": 2283 + }, + { + "epoch": 0.09, + "learning_rate": 4.8567581431136186e-05, + "loss": 0.6129, + "step": 2284 + }, + { + "epoch": 0.09, + "learning_rate": 4.856694024108746e-05, + "loss": 0.6316, + "step": 2285 + }, + { + "epoch": 0.09, + "learning_rate": 4.8566299051038736e-05, + "loss": 0.6037, + "step": 2286 + }, + { + "epoch": 0.09, + "learning_rate": 4.856565786099e-05, + "loss": 0.5908, + "step": 2287 + }, + { + "epoch": 0.09, + "learning_rate": 4.856501667094127e-05, + "loss": 0.6138, + "step": 2288 + }, + { + "epoch": 0.09, + "learning_rate": 4.856437548089254e-05, + "loss": 0.5909, + "step": 2289 + }, + { + "epoch": 0.09, + "learning_rate": 4.856373429084381e-05, + "loss": 0.6259, + "step": 2290 + }, + { + "epoch": 0.09, + "learning_rate": 4.8563093100795074e-05, + "loss": 0.6401, + "step": 2291 + }, + { + "epoch": 0.09, + "learning_rate": 4.8562451910746346e-05, + "loss": 0.6609, + "step": 2292 + }, + { + "epoch": 0.09, + "learning_rate": 4.856181072069762e-05, + "loss": 0.6468, + "step": 2293 + }, + { + "epoch": 0.09, + "learning_rate": 4.856116953064889e-05, + "loss": 0.6027, + "step": 2294 + }, + { + "epoch": 0.09, + "learning_rate": 4.856052834060016e-05, + "loss": 0.6181, + "step": 2295 + }, + { + "epoch": 0.09, + "learning_rate": 4.8559887150551426e-05, + "loss": 0.6634, + "step": 2296 + }, + { + "epoch": 0.09, + "learning_rate": 4.85592459605027e-05, + "loss": 0.6874, + "step": 2297 + }, + { + "epoch": 0.09, + "learning_rate": 4.855860477045396e-05, + "loss": 0.5787, + "step": 2298 + }, + { + "epoch": 0.09, + "learning_rate": 4.8557963580405235e-05, + "loss": 0.5517, + "step": 2299 + }, + { + "epoch": 0.09, + "learning_rate": 4.85573223903565e-05, + "loss": 0.6391, + "step": 2300 + }, + { + "epoch": 0.09, + "learning_rate": 4.855668120030777e-05, + "loss": 0.6284, + "step": 2301 + }, + { + "epoch": 0.09, + "learning_rate": 4.855604001025904e-05, + "loss": 0.5329, + "step": 2302 + }, + { + "epoch": 0.09, + "learning_rate": 4.8555398820210315e-05, + "loss": 0.563, + "step": 2303 + }, + { + "epoch": 0.09, + "learning_rate": 4.8554757630161586e-05, + "loss": 0.6569, + "step": 2304 + }, + { + "epoch": 0.09, + "learning_rate": 4.855411644011285e-05, + "loss": 0.6351, + "step": 2305 + }, + { + "epoch": 0.09, + "learning_rate": 4.855347525006412e-05, + "loss": 0.6396, + "step": 2306 + }, + { + "epoch": 0.09, + "learning_rate": 4.855283406001539e-05, + "loss": 0.627, + "step": 2307 + }, + { + "epoch": 0.09, + "learning_rate": 4.855219286996666e-05, + "loss": 0.5643, + "step": 2308 + }, + { + "epoch": 0.09, + "learning_rate": 4.855155167991793e-05, + "loss": 0.6363, + "step": 2309 + }, + { + "epoch": 0.09, + "learning_rate": 4.85509104898692e-05, + "loss": 0.6331, + "step": 2310 + }, + { + "epoch": 0.09, + "learning_rate": 4.855026929982047e-05, + "loss": 0.659, + "step": 2311 + }, + { + "epoch": 0.09, + "learning_rate": 4.854962810977174e-05, + "loss": 0.6275, + "step": 2312 + }, + { + "epoch": 0.09, + "learning_rate": 4.854898691972301e-05, + "loss": 0.6737, + "step": 2313 + }, + { + "epoch": 0.09, + "learning_rate": 4.8548345729674276e-05, + "loss": 0.6587, + "step": 2314 + }, + { + "epoch": 0.09, + "learning_rate": 4.854770453962555e-05, + "loss": 0.6252, + "step": 2315 + }, + { + "epoch": 0.09, + "learning_rate": 4.854706334957681e-05, + "loss": 0.5079, + "step": 2316 + }, + { + "epoch": 0.09, + "learning_rate": 4.8546422159528085e-05, + "loss": 0.561, + "step": 2317 + }, + { + "epoch": 0.09, + "learning_rate": 4.8545780969479356e-05, + "loss": 0.5993, + "step": 2318 + }, + { + "epoch": 0.09, + "learning_rate": 4.854513977943063e-05, + "loss": 0.6644, + "step": 2319 + }, + { + "epoch": 0.09, + "learning_rate": 4.854449858938189e-05, + "loss": 0.6057, + "step": 2320 + }, + { + "epoch": 0.09, + "learning_rate": 4.8543857399333165e-05, + "loss": 0.5927, + "step": 2321 + }, + { + "epoch": 0.09, + "learning_rate": 4.8543216209284437e-05, + "loss": 0.6378, + "step": 2322 + }, + { + "epoch": 0.09, + "learning_rate": 4.85425750192357e-05, + "loss": 0.5755, + "step": 2323 + }, + { + "epoch": 0.09, + "learning_rate": 4.854193382918697e-05, + "loss": 0.5502, + "step": 2324 + }, + { + "epoch": 0.09, + "learning_rate": 4.8541292639138245e-05, + "loss": 0.6007, + "step": 2325 + }, + { + "epoch": 0.09, + "learning_rate": 4.854065144908952e-05, + "loss": 0.6069, + "step": 2326 + }, + { + "epoch": 0.09, + "learning_rate": 4.854001025904078e-05, + "loss": 0.5689, + "step": 2327 + }, + { + "epoch": 0.09, + "learning_rate": 4.853936906899205e-05, + "loss": 0.6137, + "step": 2328 + }, + { + "epoch": 0.09, + "learning_rate": 4.853872787894332e-05, + "loss": 0.5437, + "step": 2329 + }, + { + "epoch": 0.09, + "learning_rate": 4.853808668889459e-05, + "loss": 0.5873, + "step": 2330 + }, + { + "epoch": 0.09, + "learning_rate": 4.853744549884586e-05, + "loss": 0.6183, + "step": 2331 + }, + { + "epoch": 0.09, + "learning_rate": 4.8536804308797127e-05, + "loss": 0.6845, + "step": 2332 + }, + { + "epoch": 0.09, + "learning_rate": 4.85361631187484e-05, + "loss": 0.693, + "step": 2333 + }, + { + "epoch": 0.09, + "learning_rate": 4.853552192869967e-05, + "loss": 0.7116, + "step": 2334 + }, + { + "epoch": 0.09, + "learning_rate": 4.853488073865094e-05, + "loss": 0.5882, + "step": 2335 + }, + { + "epoch": 0.09, + "learning_rate": 4.853423954860221e-05, + "loss": 0.5694, + "step": 2336 + }, + { + "epoch": 0.09, + "learning_rate": 4.853359835855348e-05, + "loss": 0.5918, + "step": 2337 + }, + { + "epoch": 0.09, + "learning_rate": 4.853295716850474e-05, + "loss": 0.6417, + "step": 2338 + }, + { + "epoch": 0.09, + "learning_rate": 4.8532315978456015e-05, + "loss": 0.5841, + "step": 2339 + }, + { + "epoch": 0.09, + "learning_rate": 4.853167478840729e-05, + "loss": 0.5638, + "step": 2340 + }, + { + "epoch": 0.09, + "learning_rate": 4.853103359835856e-05, + "loss": 0.5541, + "step": 2341 + }, + { + "epoch": 0.09, + "learning_rate": 4.853039240830983e-05, + "loss": 0.6391, + "step": 2342 + }, + { + "epoch": 0.09, + "learning_rate": 4.8529751218261095e-05, + "loss": 0.6685, + "step": 2343 + }, + { + "epoch": 0.09, + "learning_rate": 4.852911002821237e-05, + "loss": 0.5924, + "step": 2344 + }, + { + "epoch": 0.09, + "learning_rate": 4.852846883816363e-05, + "loss": 0.6065, + "step": 2345 + }, + { + "epoch": 0.09, + "learning_rate": 4.8527827648114904e-05, + "loss": 0.6033, + "step": 2346 + }, + { + "epoch": 0.09, + "learning_rate": 4.852718645806617e-05, + "loss": 0.6347, + "step": 2347 + }, + { + "epoch": 0.09, + "learning_rate": 4.852654526801744e-05, + "loss": 0.5403, + "step": 2348 + }, + { + "epoch": 0.09, + "learning_rate": 4.852590407796871e-05, + "loss": 0.5924, + "step": 2349 + }, + { + "epoch": 0.09, + "learning_rate": 4.8525262887919984e-05, + "loss": 0.6247, + "step": 2350 + }, + { + "epoch": 0.09, + "learning_rate": 4.8524621697871255e-05, + "loss": 0.5713, + "step": 2351 + }, + { + "epoch": 0.09, + "learning_rate": 4.852398050782252e-05, + "loss": 0.5204, + "step": 2352 + }, + { + "epoch": 0.09, + "learning_rate": 4.852333931777379e-05, + "loss": 0.4871, + "step": 2353 + }, + { + "epoch": 0.09, + "learning_rate": 4.852269812772506e-05, + "loss": 0.7046, + "step": 2354 + }, + { + "epoch": 0.09, + "learning_rate": 4.852205693767633e-05, + "loss": 0.6716, + "step": 2355 + }, + { + "epoch": 0.09, + "learning_rate": 4.8521415747627594e-05, + "loss": 0.6924, + "step": 2356 + }, + { + "epoch": 0.09, + "learning_rate": 4.8520774557578865e-05, + "loss": 0.5835, + "step": 2357 + }, + { + "epoch": 0.09, + "learning_rate": 4.8520133367530144e-05, + "loss": 0.6427, + "step": 2358 + }, + { + "epoch": 0.09, + "learning_rate": 4.851949217748141e-05, + "loss": 0.6313, + "step": 2359 + }, + { + "epoch": 0.09, + "learning_rate": 4.851885098743268e-05, + "loss": 0.5891, + "step": 2360 + }, + { + "epoch": 0.09, + "learning_rate": 4.8518209797383945e-05, + "loss": 0.6414, + "step": 2361 + }, + { + "epoch": 0.09, + "learning_rate": 4.851756860733522e-05, + "loss": 0.5705, + "step": 2362 + }, + { + "epoch": 0.09, + "learning_rate": 4.851692741728648e-05, + "loss": 0.6142, + "step": 2363 + }, + { + "epoch": 0.09, + "learning_rate": 4.8516286227237754e-05, + "loss": 0.6857, + "step": 2364 + }, + { + "epoch": 0.09, + "learning_rate": 4.8515645037189025e-05, + "loss": 0.6452, + "step": 2365 + }, + { + "epoch": 0.09, + "learning_rate": 4.85150038471403e-05, + "loss": 0.5818, + "step": 2366 + }, + { + "epoch": 0.09, + "learning_rate": 4.851436265709157e-05, + "loss": 0.5243, + "step": 2367 + }, + { + "epoch": 0.09, + "learning_rate": 4.8513721467042834e-05, + "loss": 0.6333, + "step": 2368 + }, + { + "epoch": 0.09, + "learning_rate": 4.8513080276994106e-05, + "loss": 0.5597, + "step": 2369 + }, + { + "epoch": 0.09, + "learning_rate": 4.851243908694537e-05, + "loss": 0.7061, + "step": 2370 + }, + { + "epoch": 0.09, + "learning_rate": 4.851179789689664e-05, + "loss": 0.6934, + "step": 2371 + }, + { + "epoch": 0.09, + "learning_rate": 4.851115670684791e-05, + "loss": 0.6182, + "step": 2372 + }, + { + "epoch": 0.09, + "learning_rate": 4.851051551679918e-05, + "loss": 0.6586, + "step": 2373 + }, + { + "epoch": 0.09, + "learning_rate": 4.850987432675045e-05, + "loss": 0.5994, + "step": 2374 + }, + { + "epoch": 0.09, + "learning_rate": 4.850923313670172e-05, + "loss": 0.6198, + "step": 2375 + }, + { + "epoch": 0.09, + "learning_rate": 4.8508591946652994e-05, + "loss": 0.5751, + "step": 2376 + }, + { + "epoch": 0.09, + "learning_rate": 4.850795075660426e-05, + "loss": 0.6235, + "step": 2377 + }, + { + "epoch": 0.09, + "learning_rate": 4.850730956655553e-05, + "loss": 0.6116, + "step": 2378 + }, + { + "epoch": 0.09, + "learning_rate": 4.8506668376506796e-05, + "loss": 0.6413, + "step": 2379 + }, + { + "epoch": 0.09, + "learning_rate": 4.850602718645807e-05, + "loss": 0.5869, + "step": 2380 + }, + { + "epoch": 0.09, + "learning_rate": 4.850538599640934e-05, + "loss": 0.6344, + "step": 2381 + }, + { + "epoch": 0.09, + "learning_rate": 4.850474480636061e-05, + "loss": 0.6852, + "step": 2382 + }, + { + "epoch": 0.09, + "learning_rate": 4.8504103616311876e-05, + "loss": 0.5876, + "step": 2383 + }, + { + "epoch": 0.09, + "learning_rate": 4.850346242626315e-05, + "loss": 0.6559, + "step": 2384 + }, + { + "epoch": 0.09, + "learning_rate": 4.850282123621442e-05, + "loss": 0.5929, + "step": 2385 + }, + { + "epoch": 0.09, + "learning_rate": 4.8502180046165684e-05, + "loss": 0.6077, + "step": 2386 + }, + { + "epoch": 0.09, + "learning_rate": 4.8501538856116956e-05, + "loss": 0.6322, + "step": 2387 + }, + { + "epoch": 0.09, + "learning_rate": 4.850089766606822e-05, + "loss": 0.592, + "step": 2388 + }, + { + "epoch": 0.09, + "learning_rate": 4.850025647601949e-05, + "loss": 0.5373, + "step": 2389 + }, + { + "epoch": 0.09, + "learning_rate": 4.8499615285970764e-05, + "loss": 0.7603, + "step": 2390 + }, + { + "epoch": 0.09, + "learning_rate": 4.8498974095922036e-05, + "loss": 0.6687, + "step": 2391 + }, + { + "epoch": 0.09, + "learning_rate": 4.84983329058733e-05, + "loss": 0.5658, + "step": 2392 + }, + { + "epoch": 0.09, + "learning_rate": 4.849769171582457e-05, + "loss": 0.6713, + "step": 2393 + }, + { + "epoch": 0.09, + "learning_rate": 4.8497050525775844e-05, + "loss": 0.4786, + "step": 2394 + }, + { + "epoch": 0.09, + "learning_rate": 4.849640933572711e-05, + "loss": 0.6855, + "step": 2395 + }, + { + "epoch": 0.09, + "learning_rate": 4.849576814567838e-05, + "loss": 0.612, + "step": 2396 + }, + { + "epoch": 0.09, + "learning_rate": 4.849512695562965e-05, + "loss": 0.5938, + "step": 2397 + }, + { + "epoch": 0.09, + "learning_rate": 4.8494485765580924e-05, + "loss": 0.63, + "step": 2398 + }, + { + "epoch": 0.09, + "learning_rate": 4.849384457553219e-05, + "loss": 0.6266, + "step": 2399 + }, + { + "epoch": 0.09, + "learning_rate": 4.849320338548346e-05, + "loss": 0.63, + "step": 2400 + }, + { + "epoch": 0.09, + "learning_rate": 4.8492562195434726e-05, + "loss": 0.5501, + "step": 2401 + }, + { + "epoch": 0.09, + "learning_rate": 4.8491921005386e-05, + "loss": 0.581, + "step": 2402 + }, + { + "epoch": 0.09, + "learning_rate": 4.849127981533727e-05, + "loss": 0.6147, + "step": 2403 + }, + { + "epoch": 0.09, + "learning_rate": 4.8490638625288534e-05, + "loss": 0.6283, + "step": 2404 + }, + { + "epoch": 0.09, + "learning_rate": 4.8489997435239806e-05, + "loss": 0.6994, + "step": 2405 + }, + { + "epoch": 0.09, + "learning_rate": 4.848935624519108e-05, + "loss": 0.6216, + "step": 2406 + }, + { + "epoch": 0.09, + "learning_rate": 4.848871505514235e-05, + "loss": 0.6704, + "step": 2407 + }, + { + "epoch": 0.09, + "learning_rate": 4.8488073865093614e-05, + "loss": 0.6755, + "step": 2408 + }, + { + "epoch": 0.09, + "learning_rate": 4.8487432675044886e-05, + "loss": 0.5179, + "step": 2409 + }, + { + "epoch": 0.09, + "learning_rate": 4.848679148499615e-05, + "loss": 0.5712, + "step": 2410 + }, + { + "epoch": 0.09, + "learning_rate": 4.848615029494742e-05, + "loss": 0.6466, + "step": 2411 + }, + { + "epoch": 0.09, + "learning_rate": 4.8485509104898695e-05, + "loss": 0.5277, + "step": 2412 + }, + { + "epoch": 0.09, + "learning_rate": 4.8484867914849966e-05, + "loss": 0.6169, + "step": 2413 + }, + { + "epoch": 0.09, + "learning_rate": 4.848422672480124e-05, + "loss": 0.5766, + "step": 2414 + }, + { + "epoch": 0.09, + "learning_rate": 4.84835855347525e-05, + "loss": 0.6517, + "step": 2415 + }, + { + "epoch": 0.09, + "learning_rate": 4.8482944344703775e-05, + "loss": 0.7031, + "step": 2416 + }, + { + "epoch": 0.09, + "learning_rate": 4.848230315465504e-05, + "loss": 0.6008, + "step": 2417 + }, + { + "epoch": 0.09, + "learning_rate": 4.848166196460631e-05, + "loss": 0.5552, + "step": 2418 + }, + { + "epoch": 0.09, + "learning_rate": 4.848102077455758e-05, + "loss": 0.6552, + "step": 2419 + }, + { + "epoch": 0.09, + "learning_rate": 4.848037958450885e-05, + "loss": 0.6762, + "step": 2420 + }, + { + "epoch": 0.09, + "learning_rate": 4.847973839446012e-05, + "loss": 0.6046, + "step": 2421 + }, + { + "epoch": 0.09, + "learning_rate": 4.847909720441139e-05, + "loss": 0.6071, + "step": 2422 + }, + { + "epoch": 0.09, + "learning_rate": 4.847845601436266e-05, + "loss": 0.6087, + "step": 2423 + }, + { + "epoch": 0.09, + "learning_rate": 4.847781482431393e-05, + "loss": 0.6446, + "step": 2424 + }, + { + "epoch": 0.09, + "learning_rate": 4.84771736342652e-05, + "loss": 0.5699, + "step": 2425 + }, + { + "epoch": 0.09, + "learning_rate": 4.8476532444216465e-05, + "loss": 0.5204, + "step": 2426 + }, + { + "epoch": 0.09, + "learning_rate": 4.8475891254167736e-05, + "loss": 0.6482, + "step": 2427 + }, + { + "epoch": 0.09, + "learning_rate": 4.847525006411901e-05, + "loss": 0.6246, + "step": 2428 + }, + { + "epoch": 0.09, + "learning_rate": 4.847460887407027e-05, + "loss": 0.6576, + "step": 2429 + }, + { + "epoch": 0.09, + "learning_rate": 4.847396768402155e-05, + "loss": 0.6998, + "step": 2430 + }, + { + "epoch": 0.09, + "learning_rate": 4.8473326493972816e-05, + "loss": 0.5757, + "step": 2431 + }, + { + "epoch": 0.09, + "learning_rate": 4.847268530392409e-05, + "loss": 0.5635, + "step": 2432 + }, + { + "epoch": 0.09, + "learning_rate": 4.847204411387535e-05, + "loss": 0.5692, + "step": 2433 + }, + { + "epoch": 0.09, + "learning_rate": 4.8471402923826625e-05, + "loss": 0.6281, + "step": 2434 + }, + { + "epoch": 0.09, + "learning_rate": 4.847076173377789e-05, + "loss": 0.6075, + "step": 2435 + }, + { + "epoch": 0.09, + "learning_rate": 4.847012054372916e-05, + "loss": 0.5741, + "step": 2436 + }, + { + "epoch": 0.09, + "learning_rate": 4.846947935368043e-05, + "loss": 0.6414, + "step": 2437 + }, + { + "epoch": 0.09, + "learning_rate": 4.8468838163631705e-05, + "loss": 0.6316, + "step": 2438 + }, + { + "epoch": 0.09, + "learning_rate": 4.846819697358298e-05, + "loss": 0.5855, + "step": 2439 + }, + { + "epoch": 0.09, + "learning_rate": 4.846755578353424e-05, + "loss": 0.65, + "step": 2440 + }, + { + "epoch": 0.09, + "learning_rate": 4.846691459348551e-05, + "loss": 0.6501, + "step": 2441 + }, + { + "epoch": 0.09, + "learning_rate": 4.846627340343678e-05, + "loss": 0.6412, + "step": 2442 + }, + { + "epoch": 0.09, + "learning_rate": 4.846563221338805e-05, + "loss": 0.6516, + "step": 2443 + }, + { + "epoch": 0.09, + "learning_rate": 4.8464991023339315e-05, + "loss": 0.574, + "step": 2444 + }, + { + "epoch": 0.09, + "learning_rate": 4.846434983329059e-05, + "loss": 0.5815, + "step": 2445 + }, + { + "epoch": 0.09, + "learning_rate": 4.8463708643241865e-05, + "loss": 0.6076, + "step": 2446 + }, + { + "epoch": 0.09, + "learning_rate": 4.846306745319313e-05, + "loss": 0.6383, + "step": 2447 + }, + { + "epoch": 0.09, + "learning_rate": 4.84624262631444e-05, + "loss": 0.7021, + "step": 2448 + }, + { + "epoch": 0.09, + "learning_rate": 4.846178507309567e-05, + "loss": 0.5549, + "step": 2449 + }, + { + "epoch": 0.09, + "learning_rate": 4.846114388304694e-05, + "loss": 0.6817, + "step": 2450 + }, + { + "epoch": 0.09, + "learning_rate": 4.8460502692998203e-05, + "loss": 0.6481, + "step": 2451 + }, + { + "epoch": 0.09, + "learning_rate": 4.8459861502949475e-05, + "loss": 0.5839, + "step": 2452 + }, + { + "epoch": 0.09, + "learning_rate": 4.845922031290075e-05, + "loss": 0.617, + "step": 2453 + }, + { + "epoch": 0.09, + "learning_rate": 4.845857912285202e-05, + "loss": 0.6308, + "step": 2454 + }, + { + "epoch": 0.09, + "learning_rate": 4.845793793280329e-05, + "loss": 0.5999, + "step": 2455 + }, + { + "epoch": 0.09, + "learning_rate": 4.8457296742754555e-05, + "loss": 0.5733, + "step": 2456 + }, + { + "epoch": 0.09, + "learning_rate": 4.845665555270583e-05, + "loss": 0.5887, + "step": 2457 + }, + { + "epoch": 0.09, + "learning_rate": 4.845601436265709e-05, + "loss": 0.55, + "step": 2458 + }, + { + "epoch": 0.09, + "learning_rate": 4.8455373172608364e-05, + "loss": 0.5702, + "step": 2459 + }, + { + "epoch": 0.09, + "learning_rate": 4.845473198255963e-05, + "loss": 0.5764, + "step": 2460 + }, + { + "epoch": 0.09, + "learning_rate": 4.84540907925109e-05, + "loss": 0.5558, + "step": 2461 + }, + { + "epoch": 0.09, + "learning_rate": 4.845344960246217e-05, + "loss": 0.5971, + "step": 2462 + }, + { + "epoch": 0.09, + "learning_rate": 4.8452808412413444e-05, + "loss": 0.6227, + "step": 2463 + }, + { + "epoch": 0.09, + "learning_rate": 4.8452167222364715e-05, + "loss": 0.68, + "step": 2464 + }, + { + "epoch": 0.09, + "learning_rate": 4.845152603231598e-05, + "loss": 0.5733, + "step": 2465 + }, + { + "epoch": 0.09, + "learning_rate": 4.845088484226725e-05, + "loss": 0.6785, + "step": 2466 + }, + { + "epoch": 0.09, + "learning_rate": 4.845024365221852e-05, + "loss": 0.652, + "step": 2467 + }, + { + "epoch": 0.09, + "learning_rate": 4.844960246216979e-05, + "loss": 0.637, + "step": 2468 + }, + { + "epoch": 0.09, + "learning_rate": 4.844896127212106e-05, + "loss": 0.603, + "step": 2469 + }, + { + "epoch": 0.09, + "learning_rate": 4.844832008207233e-05, + "loss": 0.5757, + "step": 2470 + }, + { + "epoch": 0.1, + "learning_rate": 4.84476788920236e-05, + "loss": 0.6481, + "step": 2471 + }, + { + "epoch": 0.1, + "learning_rate": 4.844703770197487e-05, + "loss": 0.6586, + "step": 2472 + }, + { + "epoch": 0.1, + "learning_rate": 4.844639651192614e-05, + "loss": 0.5794, + "step": 2473 + }, + { + "epoch": 0.1, + "learning_rate": 4.8445755321877405e-05, + "loss": 0.6427, + "step": 2474 + }, + { + "epoch": 0.1, + "learning_rate": 4.844511413182868e-05, + "loss": 0.5743, + "step": 2475 + }, + { + "epoch": 0.1, + "learning_rate": 4.844447294177994e-05, + "loss": 0.6421, + "step": 2476 + }, + { + "epoch": 0.1, + "learning_rate": 4.8443831751731214e-05, + "loss": 0.5911, + "step": 2477 + }, + { + "epoch": 0.1, + "learning_rate": 4.8443190561682486e-05, + "loss": 0.5927, + "step": 2478 + }, + { + "epoch": 0.1, + "learning_rate": 4.844254937163376e-05, + "loss": 0.6799, + "step": 2479 + }, + { + "epoch": 0.1, + "learning_rate": 4.844190818158502e-05, + "loss": 0.6733, + "step": 2480 + }, + { + "epoch": 0.1, + "learning_rate": 4.8441266991536294e-05, + "loss": 0.7262, + "step": 2481 + }, + { + "epoch": 0.1, + "learning_rate": 4.8440625801487566e-05, + "loss": 0.576, + "step": 2482 + }, + { + "epoch": 0.1, + "learning_rate": 4.843998461143883e-05, + "loss": 0.5426, + "step": 2483 + }, + { + "epoch": 0.1, + "learning_rate": 4.84393434213901e-05, + "loss": 0.6422, + "step": 2484 + }, + { + "epoch": 0.1, + "learning_rate": 4.8438702231341374e-05, + "loss": 0.6475, + "step": 2485 + }, + { + "epoch": 0.1, + "learning_rate": 4.8438061041292646e-05, + "loss": 0.6503, + "step": 2486 + }, + { + "epoch": 0.1, + "learning_rate": 4.843741985124391e-05, + "loss": 0.5621, + "step": 2487 + }, + { + "epoch": 0.1, + "learning_rate": 4.843677866119518e-05, + "loss": 0.6803, + "step": 2488 + }, + { + "epoch": 0.1, + "learning_rate": 4.843613747114645e-05, + "loss": 0.6457, + "step": 2489 + }, + { + "epoch": 0.1, + "learning_rate": 4.843549628109772e-05, + "loss": 0.5811, + "step": 2490 + }, + { + "epoch": 0.1, + "learning_rate": 4.843485509104899e-05, + "loss": 0.5991, + "step": 2491 + }, + { + "epoch": 0.1, + "learning_rate": 4.8434213901000256e-05, + "loss": 0.636, + "step": 2492 + }, + { + "epoch": 0.1, + "learning_rate": 4.843357271095153e-05, + "loss": 0.5851, + "step": 2493 + }, + { + "epoch": 0.1, + "learning_rate": 4.84329315209028e-05, + "loss": 0.6803, + "step": 2494 + }, + { + "epoch": 0.1, + "learning_rate": 4.843229033085407e-05, + "loss": 0.6255, + "step": 2495 + }, + { + "epoch": 0.1, + "learning_rate": 4.8431649140805336e-05, + "loss": 0.6352, + "step": 2496 + }, + { + "epoch": 0.1, + "learning_rate": 4.843100795075661e-05, + "loss": 0.6284, + "step": 2497 + }, + { + "epoch": 0.1, + "learning_rate": 4.843036676070787e-05, + "loss": 0.6519, + "step": 2498 + }, + { + "epoch": 0.1, + "learning_rate": 4.8429725570659144e-05, + "loss": 0.627, + "step": 2499 + }, + { + "epoch": 0.1, + "learning_rate": 4.8429084380610416e-05, + "loss": 0.635, + "step": 2500 + }, + { + "epoch": 0.1, + "learning_rate": 4.842844319056169e-05, + "loss": 0.6574, + "step": 2501 + }, + { + "epoch": 0.1, + "learning_rate": 4.842780200051296e-05, + "loss": 0.5769, + "step": 2502 + }, + { + "epoch": 0.1, + "learning_rate": 4.8427160810464224e-05, + "loss": 0.601, + "step": 2503 + }, + { + "epoch": 0.1, + "learning_rate": 4.8426519620415496e-05, + "loss": 0.6209, + "step": 2504 + }, + { + "epoch": 0.1, + "learning_rate": 4.842587843036676e-05, + "loss": 0.6434, + "step": 2505 + }, + { + "epoch": 0.1, + "learning_rate": 4.842523724031803e-05, + "loss": 0.6291, + "step": 2506 + }, + { + "epoch": 0.1, + "learning_rate": 4.84245960502693e-05, + "loss": 0.5923, + "step": 2507 + }, + { + "epoch": 0.1, + "learning_rate": 4.842395486022057e-05, + "loss": 0.6192, + "step": 2508 + }, + { + "epoch": 0.1, + "learning_rate": 4.842331367017184e-05, + "loss": 0.6277, + "step": 2509 + }, + { + "epoch": 0.1, + "learning_rate": 4.842267248012311e-05, + "loss": 0.5482, + "step": 2510 + }, + { + "epoch": 0.1, + "learning_rate": 4.8422031290074384e-05, + "loss": 0.6588, + "step": 2511 + }, + { + "epoch": 0.1, + "learning_rate": 4.842139010002565e-05, + "loss": 0.5471, + "step": 2512 + }, + { + "epoch": 0.1, + "learning_rate": 4.842074890997692e-05, + "loss": 0.5699, + "step": 2513 + }, + { + "epoch": 0.1, + "learning_rate": 4.8420107719928186e-05, + "loss": 0.546, + "step": 2514 + }, + { + "epoch": 0.1, + "learning_rate": 4.841946652987946e-05, + "loss": 0.5595, + "step": 2515 + }, + { + "epoch": 0.1, + "learning_rate": 4.841882533983072e-05, + "loss": 0.6171, + "step": 2516 + }, + { + "epoch": 0.1, + "learning_rate": 4.8418184149781994e-05, + "loss": 0.6397, + "step": 2517 + }, + { + "epoch": 0.1, + "learning_rate": 4.841754295973327e-05, + "loss": 0.5407, + "step": 2518 + }, + { + "epoch": 0.1, + "learning_rate": 4.841690176968454e-05, + "loss": 0.668, + "step": 2519 + }, + { + "epoch": 0.1, + "learning_rate": 4.841626057963581e-05, + "loss": 0.5989, + "step": 2520 + }, + { + "epoch": 0.1, + "learning_rate": 4.8415619389587074e-05, + "loss": 0.6019, + "step": 2521 + }, + { + "epoch": 0.1, + "learning_rate": 4.8414978199538346e-05, + "loss": 0.6194, + "step": 2522 + }, + { + "epoch": 0.1, + "learning_rate": 4.841433700948961e-05, + "loss": 0.6456, + "step": 2523 + }, + { + "epoch": 0.1, + "learning_rate": 4.841369581944088e-05, + "loss": 0.6635, + "step": 2524 + }, + { + "epoch": 0.1, + "learning_rate": 4.8413054629392155e-05, + "loss": 0.6142, + "step": 2525 + }, + { + "epoch": 0.1, + "learning_rate": 4.8412413439343426e-05, + "loss": 0.6152, + "step": 2526 + }, + { + "epoch": 0.1, + "learning_rate": 4.84117722492947e-05, + "loss": 0.6466, + "step": 2527 + }, + { + "epoch": 0.1, + "learning_rate": 4.841113105924596e-05, + "loss": 0.5871, + "step": 2528 + }, + { + "epoch": 0.1, + "learning_rate": 4.8410489869197235e-05, + "loss": 0.6296, + "step": 2529 + }, + { + "epoch": 0.1, + "learning_rate": 4.84098486791485e-05, + "loss": 0.5383, + "step": 2530 + }, + { + "epoch": 0.1, + "learning_rate": 4.840920748909977e-05, + "loss": 0.6021, + "step": 2531 + }, + { + "epoch": 0.1, + "learning_rate": 4.8408566299051036e-05, + "loss": 0.5892, + "step": 2532 + }, + { + "epoch": 0.1, + "learning_rate": 4.840792510900231e-05, + "loss": 0.6262, + "step": 2533 + }, + { + "epoch": 0.1, + "learning_rate": 4.840728391895358e-05, + "loss": 0.5905, + "step": 2534 + }, + { + "epoch": 0.1, + "learning_rate": 4.840664272890485e-05, + "loss": 0.6469, + "step": 2535 + }, + { + "epoch": 0.1, + "learning_rate": 4.840600153885612e-05, + "loss": 0.6304, + "step": 2536 + }, + { + "epoch": 0.1, + "learning_rate": 4.840536034880739e-05, + "loss": 0.6575, + "step": 2537 + }, + { + "epoch": 0.1, + "learning_rate": 4.840471915875866e-05, + "loss": 0.6326, + "step": 2538 + }, + { + "epoch": 0.1, + "learning_rate": 4.8404077968709925e-05, + "loss": 0.6822, + "step": 2539 + }, + { + "epoch": 0.1, + "learning_rate": 4.8403436778661196e-05, + "loss": 0.6274, + "step": 2540 + }, + { + "epoch": 0.1, + "learning_rate": 4.840279558861247e-05, + "loss": 0.4992, + "step": 2541 + }, + { + "epoch": 0.1, + "learning_rate": 4.840215439856374e-05, + "loss": 0.5355, + "step": 2542 + }, + { + "epoch": 0.1, + "learning_rate": 4.8401513208515005e-05, + "loss": 0.576, + "step": 2543 + }, + { + "epoch": 0.1, + "learning_rate": 4.8400872018466277e-05, + "loss": 0.5992, + "step": 2544 + }, + { + "epoch": 0.1, + "learning_rate": 4.840023082841755e-05, + "loss": 0.72, + "step": 2545 + }, + { + "epoch": 0.1, + "learning_rate": 4.839958963836881e-05, + "loss": 0.6533, + "step": 2546 + }, + { + "epoch": 0.1, + "learning_rate": 4.8398948448320085e-05, + "loss": 0.5774, + "step": 2547 + }, + { + "epoch": 0.1, + "learning_rate": 4.839830725827135e-05, + "loss": 0.6521, + "step": 2548 + }, + { + "epoch": 0.1, + "learning_rate": 4.839766606822262e-05, + "loss": 0.5891, + "step": 2549 + }, + { + "epoch": 0.1, + "learning_rate": 4.839702487817389e-05, + "loss": 0.5717, + "step": 2550 + }, + { + "epoch": 0.1, + "learning_rate": 4.8396383688125165e-05, + "loss": 0.7153, + "step": 2551 + }, + { + "epoch": 0.1, + "learning_rate": 4.839574249807643e-05, + "loss": 0.6342, + "step": 2552 + }, + { + "epoch": 0.1, + "learning_rate": 4.83951013080277e-05, + "loss": 0.6422, + "step": 2553 + }, + { + "epoch": 0.1, + "learning_rate": 4.839446011797897e-05, + "loss": 0.6327, + "step": 2554 + }, + { + "epoch": 0.1, + "learning_rate": 4.839381892793024e-05, + "loss": 0.6183, + "step": 2555 + }, + { + "epoch": 0.1, + "learning_rate": 4.839317773788151e-05, + "loss": 0.5797, + "step": 2556 + }, + { + "epoch": 0.1, + "learning_rate": 4.839253654783278e-05, + "loss": 0.5807, + "step": 2557 + }, + { + "epoch": 0.1, + "learning_rate": 4.8391895357784053e-05, + "loss": 0.7585, + "step": 2558 + }, + { + "epoch": 0.1, + "learning_rate": 4.839125416773532e-05, + "loss": 0.5524, + "step": 2559 + }, + { + "epoch": 0.1, + "learning_rate": 4.839061297768659e-05, + "loss": 0.5888, + "step": 2560 + }, + { + "epoch": 0.1, + "learning_rate": 4.8389971787637855e-05, + "loss": 0.6792, + "step": 2561 + }, + { + "epoch": 0.1, + "learning_rate": 4.838933059758913e-05, + "loss": 0.5442, + "step": 2562 + }, + { + "epoch": 0.1, + "learning_rate": 4.83886894075404e-05, + "loss": 0.6722, + "step": 2563 + }, + { + "epoch": 0.1, + "learning_rate": 4.8388048217491663e-05, + "loss": 0.5331, + "step": 2564 + }, + { + "epoch": 0.1, + "learning_rate": 4.8387407027442935e-05, + "loss": 0.6282, + "step": 2565 + }, + { + "epoch": 0.1, + "learning_rate": 4.838676583739421e-05, + "loss": 0.5224, + "step": 2566 + }, + { + "epoch": 0.1, + "learning_rate": 4.838612464734548e-05, + "loss": 0.6259, + "step": 2567 + }, + { + "epoch": 0.1, + "learning_rate": 4.8385483457296744e-05, + "loss": 0.6882, + "step": 2568 + }, + { + "epoch": 0.1, + "learning_rate": 4.8384842267248015e-05, + "loss": 0.5863, + "step": 2569 + }, + { + "epoch": 0.1, + "learning_rate": 4.838420107719928e-05, + "loss": 0.6598, + "step": 2570 + }, + { + "epoch": 0.1, + "learning_rate": 4.838355988715055e-05, + "loss": 0.621, + "step": 2571 + }, + { + "epoch": 0.1, + "learning_rate": 4.8382918697101824e-05, + "loss": 0.5723, + "step": 2572 + }, + { + "epoch": 0.1, + "learning_rate": 4.8382277507053095e-05, + "loss": 0.5418, + "step": 2573 + }, + { + "epoch": 0.1, + "learning_rate": 4.838163631700437e-05, + "loss": 0.5863, + "step": 2574 + }, + { + "epoch": 0.1, + "learning_rate": 4.838099512695563e-05, + "loss": 0.6449, + "step": 2575 + }, + { + "epoch": 0.1, + "learning_rate": 4.8380353936906904e-05, + "loss": 0.5531, + "step": 2576 + }, + { + "epoch": 0.1, + "learning_rate": 4.837971274685817e-05, + "loss": 0.6895, + "step": 2577 + }, + { + "epoch": 0.1, + "learning_rate": 4.837907155680944e-05, + "loss": 0.5555, + "step": 2578 + }, + { + "epoch": 0.1, + "learning_rate": 4.8378430366760705e-05, + "loss": 0.5685, + "step": 2579 + }, + { + "epoch": 0.1, + "learning_rate": 4.837778917671198e-05, + "loss": 0.6086, + "step": 2580 + }, + { + "epoch": 0.1, + "learning_rate": 4.837714798666325e-05, + "loss": 0.5661, + "step": 2581 + }, + { + "epoch": 0.1, + "learning_rate": 4.837650679661452e-05, + "loss": 0.702, + "step": 2582 + }, + { + "epoch": 0.1, + "learning_rate": 4.837586560656579e-05, + "loss": 0.5788, + "step": 2583 + }, + { + "epoch": 0.1, + "learning_rate": 4.837522441651706e-05, + "loss": 0.5632, + "step": 2584 + }, + { + "epoch": 0.1, + "learning_rate": 4.837458322646833e-05, + "loss": 0.5503, + "step": 2585 + }, + { + "epoch": 0.1, + "learning_rate": 4.8373942036419594e-05, + "loss": 0.6682, + "step": 2586 + }, + { + "epoch": 0.1, + "learning_rate": 4.8373300846370865e-05, + "loss": 0.5739, + "step": 2587 + }, + { + "epoch": 0.1, + "learning_rate": 4.837265965632213e-05, + "loss": 0.6104, + "step": 2588 + }, + { + "epoch": 0.1, + "learning_rate": 4.83720184662734e-05, + "loss": 0.6155, + "step": 2589 + }, + { + "epoch": 0.1, + "learning_rate": 4.837137727622468e-05, + "loss": 0.5106, + "step": 2590 + }, + { + "epoch": 0.1, + "learning_rate": 4.8370736086175946e-05, + "loss": 0.6891, + "step": 2591 + }, + { + "epoch": 0.1, + "learning_rate": 4.837009489612722e-05, + "loss": 0.5708, + "step": 2592 + }, + { + "epoch": 0.1, + "learning_rate": 4.836945370607848e-05, + "loss": 0.5647, + "step": 2593 + }, + { + "epoch": 0.1, + "learning_rate": 4.8368812516029754e-05, + "loss": 0.6405, + "step": 2594 + }, + { + "epoch": 0.1, + "learning_rate": 4.836817132598102e-05, + "loss": 0.612, + "step": 2595 + }, + { + "epoch": 0.1, + "learning_rate": 4.836753013593229e-05, + "loss": 0.5935, + "step": 2596 + }, + { + "epoch": 0.1, + "learning_rate": 4.836688894588356e-05, + "loss": 0.5648, + "step": 2597 + }, + { + "epoch": 0.1, + "learning_rate": 4.8366247755834834e-05, + "loss": 0.7012, + "step": 2598 + }, + { + "epoch": 0.1, + "learning_rate": 4.8365606565786106e-05, + "loss": 0.5952, + "step": 2599 + }, + { + "epoch": 0.1, + "learning_rate": 4.836496537573737e-05, + "loss": 0.6949, + "step": 2600 + }, + { + "epoch": 0.1, + "learning_rate": 4.836432418568864e-05, + "loss": 0.5561, + "step": 2601 + }, + { + "epoch": 0.1, + "learning_rate": 4.836368299563991e-05, + "loss": 0.5267, + "step": 2602 + }, + { + "epoch": 0.1, + "learning_rate": 4.836304180559118e-05, + "loss": 0.7005, + "step": 2603 + }, + { + "epoch": 0.1, + "learning_rate": 4.8362400615542444e-05, + "loss": 0.6048, + "step": 2604 + }, + { + "epoch": 0.1, + "learning_rate": 4.8361759425493716e-05, + "loss": 0.6365, + "step": 2605 + }, + { + "epoch": 0.1, + "learning_rate": 4.8361118235444994e-05, + "loss": 0.5982, + "step": 2606 + }, + { + "epoch": 0.1, + "learning_rate": 4.836047704539626e-05, + "loss": 0.6054, + "step": 2607 + }, + { + "epoch": 0.1, + "learning_rate": 4.835983585534753e-05, + "loss": 0.6156, + "step": 2608 + }, + { + "epoch": 0.1, + "learning_rate": 4.8359194665298796e-05, + "loss": 0.5841, + "step": 2609 + }, + { + "epoch": 0.1, + "learning_rate": 4.835855347525007e-05, + "loss": 0.564, + "step": 2610 + }, + { + "epoch": 0.1, + "learning_rate": 4.835791228520133e-05, + "loss": 0.5996, + "step": 2611 + }, + { + "epoch": 0.1, + "learning_rate": 4.8357271095152604e-05, + "loss": 0.5457, + "step": 2612 + }, + { + "epoch": 0.1, + "learning_rate": 4.8356629905103876e-05, + "loss": 0.6527, + "step": 2613 + }, + { + "epoch": 0.1, + "learning_rate": 4.835598871505515e-05, + "loss": 0.547, + "step": 2614 + }, + { + "epoch": 0.1, + "learning_rate": 4.835534752500642e-05, + "loss": 0.5768, + "step": 2615 + }, + { + "epoch": 0.1, + "learning_rate": 4.8354706334957684e-05, + "loss": 0.6149, + "step": 2616 + }, + { + "epoch": 0.1, + "learning_rate": 4.8354065144908956e-05, + "loss": 0.574, + "step": 2617 + }, + { + "epoch": 0.1, + "learning_rate": 4.835342395486022e-05, + "loss": 0.6757, + "step": 2618 + }, + { + "epoch": 0.1, + "learning_rate": 4.835278276481149e-05, + "loss": 0.5539, + "step": 2619 + }, + { + "epoch": 0.1, + "learning_rate": 4.835214157476276e-05, + "loss": 0.5923, + "step": 2620 + }, + { + "epoch": 0.1, + "learning_rate": 4.835150038471403e-05, + "loss": 0.6408, + "step": 2621 + }, + { + "epoch": 0.1, + "learning_rate": 4.83508591946653e-05, + "loss": 0.6969, + "step": 2622 + }, + { + "epoch": 0.1, + "learning_rate": 4.835021800461657e-05, + "loss": 0.606, + "step": 2623 + }, + { + "epoch": 0.1, + "learning_rate": 4.8349576814567844e-05, + "loss": 0.5734, + "step": 2624 + }, + { + "epoch": 0.1, + "learning_rate": 4.834893562451911e-05, + "loss": 0.4745, + "step": 2625 + }, + { + "epoch": 0.1, + "learning_rate": 4.834829443447038e-05, + "loss": 0.6543, + "step": 2626 + }, + { + "epoch": 0.1, + "learning_rate": 4.8347653244421646e-05, + "loss": 0.5765, + "step": 2627 + }, + { + "epoch": 0.1, + "learning_rate": 4.834701205437292e-05, + "loss": 0.5725, + "step": 2628 + }, + { + "epoch": 0.1, + "learning_rate": 4.834637086432419e-05, + "loss": 0.7563, + "step": 2629 + }, + { + "epoch": 0.1, + "learning_rate": 4.834572967427546e-05, + "loss": 0.5741, + "step": 2630 + }, + { + "epoch": 0.1, + "learning_rate": 4.8345088484226726e-05, + "loss": 0.7573, + "step": 2631 + }, + { + "epoch": 0.1, + "learning_rate": 4.8344447294178e-05, + "loss": 0.5905, + "step": 2632 + }, + { + "epoch": 0.1, + "learning_rate": 4.834380610412927e-05, + "loss": 0.5704, + "step": 2633 + }, + { + "epoch": 0.1, + "learning_rate": 4.8343164914080535e-05, + "loss": 0.6546, + "step": 2634 + }, + { + "epoch": 0.1, + "learning_rate": 4.8342523724031806e-05, + "loss": 0.6003, + "step": 2635 + }, + { + "epoch": 0.1, + "learning_rate": 4.834188253398307e-05, + "loss": 0.6059, + "step": 2636 + }, + { + "epoch": 0.1, + "learning_rate": 4.834124134393434e-05, + "loss": 0.566, + "step": 2637 + }, + { + "epoch": 0.1, + "learning_rate": 4.8340600153885615e-05, + "loss": 0.6115, + "step": 2638 + }, + { + "epoch": 0.1, + "learning_rate": 4.8339958963836886e-05, + "loss": 0.5746, + "step": 2639 + }, + { + "epoch": 0.1, + "learning_rate": 4.833931777378815e-05, + "loss": 0.6346, + "step": 2640 + }, + { + "epoch": 0.1, + "learning_rate": 4.833867658373942e-05, + "loss": 0.5324, + "step": 2641 + }, + { + "epoch": 0.1, + "learning_rate": 4.8338035393690695e-05, + "loss": 0.5775, + "step": 2642 + }, + { + "epoch": 0.1, + "learning_rate": 4.833739420364196e-05, + "loss": 0.6161, + "step": 2643 + }, + { + "epoch": 0.1, + "learning_rate": 4.833675301359323e-05, + "loss": 0.5861, + "step": 2644 + }, + { + "epoch": 0.1, + "learning_rate": 4.83361118235445e-05, + "loss": 0.628, + "step": 2645 + }, + { + "epoch": 0.1, + "learning_rate": 4.8335470633495775e-05, + "loss": 0.605, + "step": 2646 + }, + { + "epoch": 0.1, + "learning_rate": 4.833482944344704e-05, + "loss": 0.5298, + "step": 2647 + }, + { + "epoch": 0.1, + "learning_rate": 4.833418825339831e-05, + "loss": 0.5698, + "step": 2648 + }, + { + "epoch": 0.1, + "learning_rate": 4.8333547063349576e-05, + "loss": 0.6307, + "step": 2649 + }, + { + "epoch": 0.1, + "learning_rate": 4.833290587330085e-05, + "loss": 0.5793, + "step": 2650 + }, + { + "epoch": 0.1, + "learning_rate": 4.833226468325212e-05, + "loss": 0.6046, + "step": 2651 + }, + { + "epoch": 0.1, + "learning_rate": 4.8331623493203385e-05, + "loss": 0.6921, + "step": 2652 + }, + { + "epoch": 0.1, + "learning_rate": 4.8330982303154656e-05, + "loss": 0.5393, + "step": 2653 + }, + { + "epoch": 0.1, + "learning_rate": 4.833034111310593e-05, + "loss": 0.6346, + "step": 2654 + }, + { + "epoch": 0.1, + "learning_rate": 4.83296999230572e-05, + "loss": 0.5497, + "step": 2655 + }, + { + "epoch": 0.1, + "learning_rate": 4.8329058733008465e-05, + "loss": 0.6673, + "step": 2656 + }, + { + "epoch": 0.1, + "learning_rate": 4.8328417542959737e-05, + "loss": 0.561, + "step": 2657 + }, + { + "epoch": 0.1, + "learning_rate": 4.8327776352911e-05, + "loss": 0.5441, + "step": 2658 + }, + { + "epoch": 0.1, + "learning_rate": 4.832713516286227e-05, + "loss": 0.5566, + "step": 2659 + }, + { + "epoch": 0.1, + "learning_rate": 4.8326493972813545e-05, + "loss": 0.5574, + "step": 2660 + }, + { + "epoch": 0.1, + "learning_rate": 4.832585278276482e-05, + "loss": 0.5166, + "step": 2661 + }, + { + "epoch": 0.1, + "learning_rate": 4.832521159271609e-05, + "loss": 0.6454, + "step": 2662 + }, + { + "epoch": 0.1, + "learning_rate": 4.832457040266735e-05, + "loss": 0.6047, + "step": 2663 + }, + { + "epoch": 0.1, + "learning_rate": 4.8323929212618625e-05, + "loss": 0.7754, + "step": 2664 + }, + { + "epoch": 0.1, + "learning_rate": 4.832328802256989e-05, + "loss": 0.6438, + "step": 2665 + }, + { + "epoch": 0.1, + "learning_rate": 4.832264683252116e-05, + "loss": 0.676, + "step": 2666 + }, + { + "epoch": 0.1, + "learning_rate": 4.832200564247243e-05, + "loss": 0.6369, + "step": 2667 + }, + { + "epoch": 0.1, + "learning_rate": 4.83213644524237e-05, + "loss": 0.5368, + "step": 2668 + }, + { + "epoch": 0.1, + "learning_rate": 4.832072326237497e-05, + "loss": 0.649, + "step": 2669 + }, + { + "epoch": 0.1, + "learning_rate": 4.832008207232624e-05, + "loss": 0.5439, + "step": 2670 + }, + { + "epoch": 0.1, + "learning_rate": 4.8319440882277513e-05, + "loss": 0.5839, + "step": 2671 + }, + { + "epoch": 0.1, + "learning_rate": 4.831879969222878e-05, + "loss": 0.5691, + "step": 2672 + }, + { + "epoch": 0.1, + "learning_rate": 4.831815850218005e-05, + "loss": 0.6706, + "step": 2673 + }, + { + "epoch": 0.1, + "learning_rate": 4.8317517312131315e-05, + "loss": 0.6193, + "step": 2674 + }, + { + "epoch": 0.1, + "learning_rate": 4.831687612208259e-05, + "loss": 0.6145, + "step": 2675 + }, + { + "epoch": 0.1, + "learning_rate": 4.831623493203385e-05, + "loss": 0.5759, + "step": 2676 + }, + { + "epoch": 0.1, + "learning_rate": 4.8315593741985123e-05, + "loss": 0.6655, + "step": 2677 + }, + { + "epoch": 0.1, + "learning_rate": 4.83149525519364e-05, + "loss": 0.7086, + "step": 2678 + }, + { + "epoch": 0.1, + "learning_rate": 4.831431136188767e-05, + "loss": 0.6029, + "step": 2679 + }, + { + "epoch": 0.1, + "learning_rate": 4.831367017183894e-05, + "loss": 0.6175, + "step": 2680 + }, + { + "epoch": 0.1, + "learning_rate": 4.8313028981790204e-05, + "loss": 0.6298, + "step": 2681 + }, + { + "epoch": 0.1, + "learning_rate": 4.8312387791741475e-05, + "loss": 0.6068, + "step": 2682 + }, + { + "epoch": 0.1, + "learning_rate": 4.831174660169274e-05, + "loss": 0.5562, + "step": 2683 + }, + { + "epoch": 0.1, + "learning_rate": 4.831110541164401e-05, + "loss": 0.6325, + "step": 2684 + }, + { + "epoch": 0.1, + "learning_rate": 4.8310464221595284e-05, + "loss": 0.5959, + "step": 2685 + }, + { + "epoch": 0.1, + "learning_rate": 4.8309823031546555e-05, + "loss": 0.5889, + "step": 2686 + }, + { + "epoch": 0.1, + "learning_rate": 4.830918184149783e-05, + "loss": 0.5245, + "step": 2687 + }, + { + "epoch": 0.1, + "learning_rate": 4.830854065144909e-05, + "loss": 0.7133, + "step": 2688 + }, + { + "epoch": 0.1, + "learning_rate": 4.8307899461400364e-05, + "loss": 0.6122, + "step": 2689 + }, + { + "epoch": 0.1, + "learning_rate": 4.830725827135163e-05, + "loss": 0.6502, + "step": 2690 + }, + { + "epoch": 0.1, + "learning_rate": 4.83066170813029e-05, + "loss": 0.7062, + "step": 2691 + }, + { + "epoch": 0.1, + "learning_rate": 4.8305975891254165e-05, + "loss": 0.6643, + "step": 2692 + }, + { + "epoch": 0.1, + "learning_rate": 4.830533470120544e-05, + "loss": 0.5784, + "step": 2693 + }, + { + "epoch": 0.1, + "learning_rate": 4.830469351115671e-05, + "loss": 0.5818, + "step": 2694 + }, + { + "epoch": 0.1, + "learning_rate": 4.830405232110798e-05, + "loss": 0.6207, + "step": 2695 + }, + { + "epoch": 0.1, + "learning_rate": 4.830341113105925e-05, + "loss": 0.5741, + "step": 2696 + }, + { + "epoch": 0.1, + "learning_rate": 4.830276994101052e-05, + "loss": 0.5582, + "step": 2697 + }, + { + "epoch": 0.1, + "learning_rate": 4.830212875096179e-05, + "loss": 0.5828, + "step": 2698 + }, + { + "epoch": 0.1, + "learning_rate": 4.8301487560913054e-05, + "loss": 0.5538, + "step": 2699 + }, + { + "epoch": 0.1, + "learning_rate": 4.8300846370864326e-05, + "loss": 0.5617, + "step": 2700 + }, + { + "epoch": 0.1, + "learning_rate": 4.83002051808156e-05, + "loss": 0.5872, + "step": 2701 + }, + { + "epoch": 0.1, + "learning_rate": 4.829956399076687e-05, + "loss": 0.5561, + "step": 2702 + }, + { + "epoch": 0.1, + "learning_rate": 4.8298922800718134e-05, + "loss": 0.6239, + "step": 2703 + }, + { + "epoch": 0.1, + "learning_rate": 4.8298281610669406e-05, + "loss": 0.6421, + "step": 2704 + }, + { + "epoch": 0.1, + "learning_rate": 4.829764042062068e-05, + "loss": 0.6947, + "step": 2705 + }, + { + "epoch": 0.1, + "learning_rate": 4.829699923057194e-05, + "loss": 0.5283, + "step": 2706 + }, + { + "epoch": 0.1, + "learning_rate": 4.8296358040523214e-05, + "loss": 0.5945, + "step": 2707 + }, + { + "epoch": 0.1, + "learning_rate": 4.829571685047448e-05, + "loss": 0.5909, + "step": 2708 + }, + { + "epoch": 0.1, + "learning_rate": 4.829507566042575e-05, + "loss": 0.6507, + "step": 2709 + }, + { + "epoch": 0.1, + "learning_rate": 4.829443447037702e-05, + "loss": 0.6503, + "step": 2710 + }, + { + "epoch": 0.1, + "learning_rate": 4.8293793280328294e-05, + "loss": 0.5312, + "step": 2711 + }, + { + "epoch": 0.1, + "learning_rate": 4.829315209027956e-05, + "loss": 0.5675, + "step": 2712 + }, + { + "epoch": 0.1, + "learning_rate": 4.829251090023083e-05, + "loss": 0.5432, + "step": 2713 + }, + { + "epoch": 0.1, + "learning_rate": 4.82918697101821e-05, + "loss": 0.5833, + "step": 2714 + }, + { + "epoch": 0.1, + "learning_rate": 4.829122852013337e-05, + "loss": 0.5931, + "step": 2715 + }, + { + "epoch": 0.1, + "learning_rate": 4.829058733008464e-05, + "loss": 0.585, + "step": 2716 + }, + { + "epoch": 0.1, + "learning_rate": 4.828994614003591e-05, + "loss": 0.5368, + "step": 2717 + }, + { + "epoch": 0.1, + "learning_rate": 4.828930494998718e-05, + "loss": 0.5897, + "step": 2718 + }, + { + "epoch": 0.1, + "learning_rate": 4.828866375993845e-05, + "loss": 0.553, + "step": 2719 + }, + { + "epoch": 0.1, + "learning_rate": 4.828802256988972e-05, + "loss": 0.6818, + "step": 2720 + }, + { + "epoch": 0.1, + "learning_rate": 4.8287381379840984e-05, + "loss": 0.543, + "step": 2721 + }, + { + "epoch": 0.1, + "learning_rate": 4.8286740189792256e-05, + "loss": 0.5774, + "step": 2722 + }, + { + "epoch": 0.1, + "learning_rate": 4.828609899974353e-05, + "loss": 0.6244, + "step": 2723 + }, + { + "epoch": 0.1, + "learning_rate": 4.828545780969479e-05, + "loss": 0.6153, + "step": 2724 + }, + { + "epoch": 0.1, + "learning_rate": 4.8284816619646064e-05, + "loss": 0.6049, + "step": 2725 + }, + { + "epoch": 0.1, + "learning_rate": 4.8284175429597336e-05, + "loss": 0.6603, + "step": 2726 + }, + { + "epoch": 0.1, + "learning_rate": 4.828353423954861e-05, + "loss": 0.6293, + "step": 2727 + }, + { + "epoch": 0.1, + "learning_rate": 4.828289304949987e-05, + "loss": 0.5752, + "step": 2728 + }, + { + "epoch": 0.1, + "learning_rate": 4.8282251859451144e-05, + "loss": 0.5907, + "step": 2729 + }, + { + "epoch": 0.1, + "learning_rate": 4.828161066940241e-05, + "loss": 0.5979, + "step": 2730 + }, + { + "epoch": 0.1, + "learning_rate": 4.828096947935368e-05, + "loss": 0.6068, + "step": 2731 + }, + { + "epoch": 0.11, + "learning_rate": 4.828032828930495e-05, + "loss": 0.5727, + "step": 2732 + }, + { + "epoch": 0.11, + "learning_rate": 4.8279687099256224e-05, + "loss": 0.6181, + "step": 2733 + }, + { + "epoch": 0.11, + "learning_rate": 4.8279045909207496e-05, + "loss": 0.642, + "step": 2734 + }, + { + "epoch": 0.11, + "learning_rate": 4.827840471915876e-05, + "loss": 0.5774, + "step": 2735 + }, + { + "epoch": 0.11, + "learning_rate": 4.827776352911003e-05, + "loss": 0.6597, + "step": 2736 + }, + { + "epoch": 0.11, + "learning_rate": 4.82771223390613e-05, + "loss": 0.6662, + "step": 2737 + }, + { + "epoch": 0.11, + "learning_rate": 4.827648114901257e-05, + "loss": 0.6261, + "step": 2738 + }, + { + "epoch": 0.11, + "learning_rate": 4.8275839958963834e-05, + "loss": 0.533, + "step": 2739 + }, + { + "epoch": 0.11, + "learning_rate": 4.8275198768915106e-05, + "loss": 0.6659, + "step": 2740 + }, + { + "epoch": 0.11, + "learning_rate": 4.827455757886638e-05, + "loss": 0.6633, + "step": 2741 + }, + { + "epoch": 0.11, + "learning_rate": 4.827391638881765e-05, + "loss": 0.5725, + "step": 2742 + }, + { + "epoch": 0.11, + "learning_rate": 4.827327519876892e-05, + "loss": 0.5961, + "step": 2743 + }, + { + "epoch": 0.11, + "learning_rate": 4.8272634008720186e-05, + "loss": 0.6891, + "step": 2744 + }, + { + "epoch": 0.11, + "learning_rate": 4.827199281867146e-05, + "loss": 0.5911, + "step": 2745 + }, + { + "epoch": 0.11, + "learning_rate": 4.827135162862272e-05, + "loss": 0.5732, + "step": 2746 + }, + { + "epoch": 0.11, + "learning_rate": 4.8270710438573995e-05, + "loss": 0.7139, + "step": 2747 + }, + { + "epoch": 0.11, + "learning_rate": 4.827006924852526e-05, + "loss": 0.5869, + "step": 2748 + }, + { + "epoch": 0.11, + "learning_rate": 4.826942805847654e-05, + "loss": 0.6558, + "step": 2749 + }, + { + "epoch": 0.11, + "learning_rate": 4.826878686842781e-05, + "loss": 0.5731, + "step": 2750 + }, + { + "epoch": 0.11, + "learning_rate": 4.8268145678379075e-05, + "loss": 0.6155, + "step": 2751 + }, + { + "epoch": 0.11, + "learning_rate": 4.8267504488330346e-05, + "loss": 0.6226, + "step": 2752 + }, + { + "epoch": 0.11, + "learning_rate": 4.826686329828161e-05, + "loss": 0.5601, + "step": 2753 + }, + { + "epoch": 0.11, + "learning_rate": 4.826622210823288e-05, + "loss": 0.5867, + "step": 2754 + }, + { + "epoch": 0.11, + "learning_rate": 4.826558091818415e-05, + "loss": 0.5604, + "step": 2755 + }, + { + "epoch": 0.11, + "learning_rate": 4.826493972813542e-05, + "loss": 0.5707, + "step": 2756 + }, + { + "epoch": 0.11, + "learning_rate": 4.826429853808669e-05, + "loss": 0.5231, + "step": 2757 + }, + { + "epoch": 0.11, + "learning_rate": 4.826365734803796e-05, + "loss": 0.7331, + "step": 2758 + }, + { + "epoch": 0.11, + "learning_rate": 4.8263016157989235e-05, + "loss": 0.537, + "step": 2759 + }, + { + "epoch": 0.11, + "learning_rate": 4.82623749679405e-05, + "loss": 0.5532, + "step": 2760 + }, + { + "epoch": 0.11, + "learning_rate": 4.826173377789177e-05, + "loss": 0.6193, + "step": 2761 + }, + { + "epoch": 0.11, + "learning_rate": 4.8261092587843036e-05, + "loss": 0.5266, + "step": 2762 + }, + { + "epoch": 0.11, + "learning_rate": 4.826045139779431e-05, + "loss": 0.5672, + "step": 2763 + }, + { + "epoch": 0.11, + "learning_rate": 4.825981020774557e-05, + "loss": 0.6226, + "step": 2764 + }, + { + "epoch": 0.11, + "learning_rate": 4.8259169017696845e-05, + "loss": 0.6217, + "step": 2765 + }, + { + "epoch": 0.11, + "learning_rate": 4.8258527827648117e-05, + "loss": 0.6108, + "step": 2766 + }, + { + "epoch": 0.11, + "learning_rate": 4.825788663759939e-05, + "loss": 0.5335, + "step": 2767 + }, + { + "epoch": 0.11, + "learning_rate": 4.825724544755066e-05, + "loss": 0.5643, + "step": 2768 + }, + { + "epoch": 0.11, + "learning_rate": 4.8256604257501925e-05, + "loss": 0.6882, + "step": 2769 + }, + { + "epoch": 0.11, + "learning_rate": 4.8255963067453197e-05, + "loss": 0.7252, + "step": 2770 + }, + { + "epoch": 0.11, + "learning_rate": 4.825532187740446e-05, + "loss": 0.6466, + "step": 2771 + }, + { + "epoch": 0.11, + "learning_rate": 4.825468068735573e-05, + "loss": 0.6332, + "step": 2772 + }, + { + "epoch": 0.11, + "learning_rate": 4.8254039497307005e-05, + "loss": 0.5505, + "step": 2773 + }, + { + "epoch": 0.11, + "learning_rate": 4.825339830725828e-05, + "loss": 0.5707, + "step": 2774 + }, + { + "epoch": 0.11, + "learning_rate": 4.825275711720954e-05, + "loss": 0.4838, + "step": 2775 + }, + { + "epoch": 0.11, + "learning_rate": 4.825211592716081e-05, + "loss": 0.6431, + "step": 2776 + }, + { + "epoch": 0.11, + "learning_rate": 4.8251474737112085e-05, + "loss": 0.5509, + "step": 2777 + }, + { + "epoch": 0.11, + "learning_rate": 4.825083354706335e-05, + "loss": 0.6864, + "step": 2778 + }, + { + "epoch": 0.11, + "learning_rate": 4.825019235701462e-05, + "loss": 0.5604, + "step": 2779 + }, + { + "epoch": 0.11, + "learning_rate": 4.824955116696589e-05, + "loss": 0.6631, + "step": 2780 + }, + { + "epoch": 0.11, + "learning_rate": 4.824890997691716e-05, + "loss": 0.5618, + "step": 2781 + }, + { + "epoch": 0.11, + "learning_rate": 4.824826878686843e-05, + "loss": 0.6284, + "step": 2782 + }, + { + "epoch": 0.11, + "learning_rate": 4.82476275968197e-05, + "loss": 0.5846, + "step": 2783 + }, + { + "epoch": 0.11, + "learning_rate": 4.824698640677097e-05, + "loss": 0.6087, + "step": 2784 + }, + { + "epoch": 0.11, + "learning_rate": 4.824634521672224e-05, + "loss": 0.5855, + "step": 2785 + }, + { + "epoch": 0.11, + "learning_rate": 4.824570402667351e-05, + "loss": 0.6115, + "step": 2786 + }, + { + "epoch": 0.11, + "learning_rate": 4.8245062836624775e-05, + "loss": 0.696, + "step": 2787 + }, + { + "epoch": 0.11, + "learning_rate": 4.824442164657605e-05, + "loss": 0.5793, + "step": 2788 + }, + { + "epoch": 0.11, + "learning_rate": 4.824378045652732e-05, + "loss": 0.5863, + "step": 2789 + }, + { + "epoch": 0.11, + "learning_rate": 4.824313926647859e-05, + "loss": 0.6209, + "step": 2790 + }, + { + "epoch": 0.11, + "learning_rate": 4.8242498076429855e-05, + "loss": 0.6366, + "step": 2791 + }, + { + "epoch": 0.11, + "learning_rate": 4.824185688638113e-05, + "loss": 0.588, + "step": 2792 + }, + { + "epoch": 0.11, + "learning_rate": 4.824121569633239e-05, + "loss": 0.6022, + "step": 2793 + }, + { + "epoch": 0.11, + "learning_rate": 4.8240574506283664e-05, + "loss": 0.6549, + "step": 2794 + }, + { + "epoch": 0.11, + "learning_rate": 4.8239933316234935e-05, + "loss": 0.644, + "step": 2795 + }, + { + "epoch": 0.11, + "learning_rate": 4.82392921261862e-05, + "loss": 0.6327, + "step": 2796 + }, + { + "epoch": 0.11, + "learning_rate": 4.823865093613747e-05, + "loss": 0.6594, + "step": 2797 + }, + { + "epoch": 0.11, + "learning_rate": 4.8238009746088744e-05, + "loss": 0.6104, + "step": 2798 + }, + { + "epoch": 0.11, + "learning_rate": 4.8237368556040015e-05, + "loss": 0.5564, + "step": 2799 + }, + { + "epoch": 0.11, + "learning_rate": 4.823672736599128e-05, + "loss": 0.6167, + "step": 2800 + }, + { + "epoch": 0.11, + "learning_rate": 4.823608617594255e-05, + "loss": 0.5898, + "step": 2801 + }, + { + "epoch": 0.11, + "learning_rate": 4.8235444985893824e-05, + "loss": 0.5908, + "step": 2802 + }, + { + "epoch": 0.11, + "learning_rate": 4.823480379584509e-05, + "loss": 0.6133, + "step": 2803 + }, + { + "epoch": 0.11, + "learning_rate": 4.823416260579636e-05, + "loss": 0.5954, + "step": 2804 + }, + { + "epoch": 0.11, + "learning_rate": 4.823352141574763e-05, + "loss": 0.6622, + "step": 2805 + }, + { + "epoch": 0.11, + "learning_rate": 4.8232880225698904e-05, + "loss": 0.6413, + "step": 2806 + }, + { + "epoch": 0.11, + "learning_rate": 4.823223903565017e-05, + "loss": 0.5431, + "step": 2807 + }, + { + "epoch": 0.11, + "learning_rate": 4.823159784560144e-05, + "loss": 0.6559, + "step": 2808 + }, + { + "epoch": 0.11, + "learning_rate": 4.8230956655552705e-05, + "loss": 0.5468, + "step": 2809 + }, + { + "epoch": 0.11, + "learning_rate": 4.823031546550398e-05, + "loss": 0.5726, + "step": 2810 + }, + { + "epoch": 0.11, + "learning_rate": 4.822967427545525e-05, + "loss": 0.5833, + "step": 2811 + }, + { + "epoch": 0.11, + "learning_rate": 4.8229033085406514e-05, + "loss": 0.5681, + "step": 2812 + }, + { + "epoch": 0.11, + "learning_rate": 4.8228391895357786e-05, + "loss": 0.6491, + "step": 2813 + }, + { + "epoch": 0.11, + "learning_rate": 4.822775070530906e-05, + "loss": 0.6424, + "step": 2814 + }, + { + "epoch": 0.11, + "learning_rate": 4.822710951526033e-05, + "loss": 0.647, + "step": 2815 + }, + { + "epoch": 0.11, + "learning_rate": 4.8226468325211594e-05, + "loss": 0.5628, + "step": 2816 + }, + { + "epoch": 0.11, + "learning_rate": 4.8225827135162866e-05, + "loss": 0.6356, + "step": 2817 + }, + { + "epoch": 0.11, + "learning_rate": 4.822518594511413e-05, + "loss": 0.5638, + "step": 2818 + }, + { + "epoch": 0.11, + "learning_rate": 4.82245447550654e-05, + "loss": 0.5554, + "step": 2819 + }, + { + "epoch": 0.11, + "learning_rate": 4.8223903565016674e-05, + "loss": 0.6588, + "step": 2820 + }, + { + "epoch": 0.11, + "learning_rate": 4.8223262374967946e-05, + "loss": 0.5748, + "step": 2821 + }, + { + "epoch": 0.11, + "learning_rate": 4.822262118491922e-05, + "loss": 0.7032, + "step": 2822 + }, + { + "epoch": 0.11, + "learning_rate": 4.822197999487048e-05, + "loss": 0.7085, + "step": 2823 + }, + { + "epoch": 0.11, + "learning_rate": 4.8221338804821754e-05, + "loss": 0.5667, + "step": 2824 + }, + { + "epoch": 0.11, + "learning_rate": 4.822069761477302e-05, + "loss": 0.5511, + "step": 2825 + }, + { + "epoch": 0.11, + "learning_rate": 4.822005642472429e-05, + "loss": 0.6383, + "step": 2826 + }, + { + "epoch": 0.11, + "learning_rate": 4.8219415234675556e-05, + "loss": 0.6517, + "step": 2827 + }, + { + "epoch": 0.11, + "learning_rate": 4.821877404462683e-05, + "loss": 0.6, + "step": 2828 + }, + { + "epoch": 0.11, + "learning_rate": 4.82181328545781e-05, + "loss": 0.5052, + "step": 2829 + }, + { + "epoch": 0.11, + "learning_rate": 4.821749166452937e-05, + "loss": 0.6272, + "step": 2830 + }, + { + "epoch": 0.11, + "learning_rate": 4.821685047448064e-05, + "loss": 0.5066, + "step": 2831 + }, + { + "epoch": 0.11, + "learning_rate": 4.821620928443191e-05, + "loss": 0.601, + "step": 2832 + }, + { + "epoch": 0.11, + "learning_rate": 4.821556809438318e-05, + "loss": 0.5526, + "step": 2833 + }, + { + "epoch": 0.11, + "learning_rate": 4.8214926904334444e-05, + "loss": 0.5614, + "step": 2834 + }, + { + "epoch": 0.11, + "learning_rate": 4.8214285714285716e-05, + "loss": 0.5487, + "step": 2835 + }, + { + "epoch": 0.11, + "learning_rate": 4.821364452423698e-05, + "loss": 0.5203, + "step": 2836 + }, + { + "epoch": 0.11, + "learning_rate": 4.821300333418825e-05, + "loss": 0.6025, + "step": 2837 + }, + { + "epoch": 0.11, + "learning_rate": 4.821236214413953e-05, + "loss": 0.6287, + "step": 2838 + }, + { + "epoch": 0.11, + "learning_rate": 4.8211720954090796e-05, + "loss": 0.6595, + "step": 2839 + }, + { + "epoch": 0.11, + "learning_rate": 4.821107976404207e-05, + "loss": 0.5079, + "step": 2840 + }, + { + "epoch": 0.11, + "learning_rate": 4.821043857399333e-05, + "loss": 0.7033, + "step": 2841 + }, + { + "epoch": 0.11, + "learning_rate": 4.8209797383944604e-05, + "loss": 0.4991, + "step": 2842 + }, + { + "epoch": 0.11, + "learning_rate": 4.820915619389587e-05, + "loss": 0.5895, + "step": 2843 + }, + { + "epoch": 0.11, + "learning_rate": 4.820851500384714e-05, + "loss": 0.6176, + "step": 2844 + }, + { + "epoch": 0.11, + "learning_rate": 4.820787381379841e-05, + "loss": 0.6382, + "step": 2845 + }, + { + "epoch": 0.11, + "learning_rate": 4.8207232623749684e-05, + "loss": 0.601, + "step": 2846 + }, + { + "epoch": 0.11, + "learning_rate": 4.8206591433700956e-05, + "loss": 0.6089, + "step": 2847 + }, + { + "epoch": 0.11, + "learning_rate": 4.820595024365222e-05, + "loss": 0.6423, + "step": 2848 + }, + { + "epoch": 0.11, + "learning_rate": 4.820530905360349e-05, + "loss": 0.5894, + "step": 2849 + }, + { + "epoch": 0.11, + "learning_rate": 4.820466786355476e-05, + "loss": 0.6899, + "step": 2850 + }, + { + "epoch": 0.11, + "learning_rate": 4.820402667350603e-05, + "loss": 0.6189, + "step": 2851 + }, + { + "epoch": 0.11, + "learning_rate": 4.8203385483457294e-05, + "loss": 0.6685, + "step": 2852 + }, + { + "epoch": 0.11, + "learning_rate": 4.8202744293408566e-05, + "loss": 0.5562, + "step": 2853 + }, + { + "epoch": 0.11, + "learning_rate": 4.820210310335984e-05, + "loss": 0.5904, + "step": 2854 + }, + { + "epoch": 0.11, + "learning_rate": 4.820146191331111e-05, + "loss": 0.5035, + "step": 2855 + }, + { + "epoch": 0.11, + "learning_rate": 4.820082072326238e-05, + "loss": 0.62, + "step": 2856 + }, + { + "epoch": 0.11, + "learning_rate": 4.8200179533213646e-05, + "loss": 0.6277, + "step": 2857 + }, + { + "epoch": 0.11, + "learning_rate": 4.819953834316492e-05, + "loss": 0.5465, + "step": 2858 + }, + { + "epoch": 0.11, + "learning_rate": 4.819889715311618e-05, + "loss": 0.6101, + "step": 2859 + }, + { + "epoch": 0.11, + "learning_rate": 4.8198255963067455e-05, + "loss": 0.6991, + "step": 2860 + }, + { + "epoch": 0.11, + "learning_rate": 4.8197614773018726e-05, + "loss": 0.6433, + "step": 2861 + }, + { + "epoch": 0.11, + "learning_rate": 4.819697358297e-05, + "loss": 0.635, + "step": 2862 + }, + { + "epoch": 0.11, + "learning_rate": 4.819633239292126e-05, + "loss": 0.5229, + "step": 2863 + }, + { + "epoch": 0.11, + "learning_rate": 4.8195691202872535e-05, + "loss": 0.5714, + "step": 2864 + }, + { + "epoch": 0.11, + "learning_rate": 4.8195050012823806e-05, + "loss": 0.5807, + "step": 2865 + }, + { + "epoch": 0.11, + "learning_rate": 4.819440882277507e-05, + "loss": 0.7208, + "step": 2866 + }, + { + "epoch": 0.11, + "learning_rate": 4.819376763272634e-05, + "loss": 0.6496, + "step": 2867 + }, + { + "epoch": 0.11, + "learning_rate": 4.819312644267761e-05, + "loss": 0.6608, + "step": 2868 + }, + { + "epoch": 0.11, + "learning_rate": 4.819248525262888e-05, + "loss": 0.545, + "step": 2869 + }, + { + "epoch": 0.11, + "learning_rate": 4.819184406258015e-05, + "loss": 0.4993, + "step": 2870 + }, + { + "epoch": 0.11, + "learning_rate": 4.819120287253142e-05, + "loss": 0.6209, + "step": 2871 + }, + { + "epoch": 0.11, + "learning_rate": 4.819056168248269e-05, + "loss": 0.6118, + "step": 2872 + }, + { + "epoch": 0.11, + "learning_rate": 4.818992049243396e-05, + "loss": 0.5993, + "step": 2873 + }, + { + "epoch": 0.11, + "learning_rate": 4.818927930238523e-05, + "loss": 0.644, + "step": 2874 + }, + { + "epoch": 0.11, + "learning_rate": 4.8188638112336496e-05, + "loss": 0.6399, + "step": 2875 + }, + { + "epoch": 0.11, + "learning_rate": 4.818799692228777e-05, + "loss": 0.6929, + "step": 2876 + }, + { + "epoch": 0.11, + "learning_rate": 4.818735573223904e-05, + "loss": 0.6498, + "step": 2877 + }, + { + "epoch": 0.11, + "learning_rate": 4.818671454219031e-05, + "loss": 0.6478, + "step": 2878 + }, + { + "epoch": 0.11, + "learning_rate": 4.8186073352141577e-05, + "loss": 0.5597, + "step": 2879 + }, + { + "epoch": 0.11, + "learning_rate": 4.818543216209285e-05, + "loss": 0.5828, + "step": 2880 + }, + { + "epoch": 0.11, + "learning_rate": 4.818479097204411e-05, + "loss": 0.6842, + "step": 2881 + }, + { + "epoch": 0.11, + "learning_rate": 4.8184149781995385e-05, + "loss": 0.5975, + "step": 2882 + }, + { + "epoch": 0.11, + "learning_rate": 4.818350859194666e-05, + "loss": 0.6956, + "step": 2883 + }, + { + "epoch": 0.11, + "learning_rate": 4.818286740189792e-05, + "loss": 0.6114, + "step": 2884 + }, + { + "epoch": 0.11, + "learning_rate": 4.818222621184919e-05, + "loss": 0.6296, + "step": 2885 + }, + { + "epoch": 0.11, + "learning_rate": 4.8181585021800465e-05, + "loss": 0.5806, + "step": 2886 + }, + { + "epoch": 0.11, + "learning_rate": 4.818094383175174e-05, + "loss": 0.6447, + "step": 2887 + }, + { + "epoch": 0.11, + "learning_rate": 4.8180302641703e-05, + "loss": 0.6931, + "step": 2888 + }, + { + "epoch": 0.11, + "learning_rate": 4.817966145165427e-05, + "loss": 0.5979, + "step": 2889 + }, + { + "epoch": 0.11, + "learning_rate": 4.817902026160554e-05, + "loss": 0.6468, + "step": 2890 + }, + { + "epoch": 0.11, + "learning_rate": 4.817837907155681e-05, + "loss": 0.6368, + "step": 2891 + }, + { + "epoch": 0.11, + "learning_rate": 4.817773788150808e-05, + "loss": 0.6699, + "step": 2892 + }, + { + "epoch": 0.11, + "learning_rate": 4.8177096691459353e-05, + "loss": 0.5966, + "step": 2893 + }, + { + "epoch": 0.11, + "learning_rate": 4.8176455501410625e-05, + "loss": 0.5513, + "step": 2894 + }, + { + "epoch": 0.11, + "learning_rate": 4.817581431136189e-05, + "loss": 0.6738, + "step": 2895 + }, + { + "epoch": 0.11, + "learning_rate": 4.817517312131316e-05, + "loss": 0.5904, + "step": 2896 + }, + { + "epoch": 0.11, + "learning_rate": 4.817453193126443e-05, + "loss": 0.6049, + "step": 2897 + }, + { + "epoch": 0.11, + "learning_rate": 4.81738907412157e-05, + "loss": 0.5826, + "step": 2898 + }, + { + "epoch": 0.11, + "learning_rate": 4.8173249551166963e-05, + "loss": 0.652, + "step": 2899 + }, + { + "epoch": 0.11, + "learning_rate": 4.8172608361118235e-05, + "loss": 0.667, + "step": 2900 + }, + { + "epoch": 0.11, + "learning_rate": 4.817196717106951e-05, + "loss": 0.6141, + "step": 2901 + }, + { + "epoch": 0.11, + "learning_rate": 4.817132598102078e-05, + "loss": 0.6781, + "step": 2902 + }, + { + "epoch": 0.11, + "learning_rate": 4.817068479097205e-05, + "loss": 0.5849, + "step": 2903 + }, + { + "epoch": 0.11, + "learning_rate": 4.8170043600923315e-05, + "loss": 0.6586, + "step": 2904 + }, + { + "epoch": 0.11, + "learning_rate": 4.816940241087459e-05, + "loss": 0.602, + "step": 2905 + }, + { + "epoch": 0.11, + "learning_rate": 4.816876122082585e-05, + "loss": 0.6121, + "step": 2906 + }, + { + "epoch": 0.11, + "learning_rate": 4.8168120030777124e-05, + "loss": 0.651, + "step": 2907 + }, + { + "epoch": 0.11, + "learning_rate": 4.816747884072839e-05, + "loss": 0.6755, + "step": 2908 + }, + { + "epoch": 0.11, + "learning_rate": 4.816683765067967e-05, + "loss": 0.628, + "step": 2909 + }, + { + "epoch": 0.11, + "learning_rate": 4.816619646063094e-05, + "loss": 0.5866, + "step": 2910 + }, + { + "epoch": 0.11, + "learning_rate": 4.8165555270582204e-05, + "loss": 0.6611, + "step": 2911 + }, + { + "epoch": 0.11, + "learning_rate": 4.8164914080533475e-05, + "loss": 0.5656, + "step": 2912 + }, + { + "epoch": 0.11, + "learning_rate": 4.816427289048474e-05, + "loss": 0.5851, + "step": 2913 + }, + { + "epoch": 0.11, + "learning_rate": 4.816363170043601e-05, + "loss": 0.5861, + "step": 2914 + }, + { + "epoch": 0.11, + "learning_rate": 4.816299051038728e-05, + "loss": 0.6883, + "step": 2915 + }, + { + "epoch": 0.11, + "learning_rate": 4.816234932033855e-05, + "loss": 0.6101, + "step": 2916 + }, + { + "epoch": 0.11, + "learning_rate": 4.816170813028982e-05, + "loss": 0.6191, + "step": 2917 + }, + { + "epoch": 0.11, + "learning_rate": 4.816106694024109e-05, + "loss": 0.6505, + "step": 2918 + }, + { + "epoch": 0.11, + "learning_rate": 4.8160425750192364e-05, + "loss": 0.7212, + "step": 2919 + }, + { + "epoch": 0.11, + "learning_rate": 4.815978456014363e-05, + "loss": 0.6198, + "step": 2920 + }, + { + "epoch": 0.11, + "learning_rate": 4.81591433700949e-05, + "loss": 0.6226, + "step": 2921 + }, + { + "epoch": 0.11, + "learning_rate": 4.8158502180046165e-05, + "loss": 0.6156, + "step": 2922 + }, + { + "epoch": 0.11, + "learning_rate": 4.815786098999744e-05, + "loss": 0.6381, + "step": 2923 + }, + { + "epoch": 0.11, + "learning_rate": 4.81572197999487e-05, + "loss": 0.606, + "step": 2924 + }, + { + "epoch": 0.11, + "learning_rate": 4.8156578609899974e-05, + "loss": 0.6124, + "step": 2925 + }, + { + "epoch": 0.11, + "learning_rate": 4.8155937419851246e-05, + "loss": 0.5828, + "step": 2926 + }, + { + "epoch": 0.11, + "learning_rate": 4.815529622980252e-05, + "loss": 0.617, + "step": 2927 + }, + { + "epoch": 0.11, + "learning_rate": 4.815465503975379e-05, + "loss": 0.6559, + "step": 2928 + }, + { + "epoch": 0.11, + "learning_rate": 4.8154013849705054e-05, + "loss": 0.563, + "step": 2929 + }, + { + "epoch": 0.11, + "learning_rate": 4.8153372659656326e-05, + "loss": 0.6352, + "step": 2930 + }, + { + "epoch": 0.11, + "learning_rate": 4.815273146960759e-05, + "loss": 0.6729, + "step": 2931 + }, + { + "epoch": 0.11, + "learning_rate": 4.815209027955886e-05, + "loss": 0.6244, + "step": 2932 + }, + { + "epoch": 0.11, + "learning_rate": 4.8151449089510134e-05, + "loss": 0.6027, + "step": 2933 + }, + { + "epoch": 0.11, + "learning_rate": 4.8150807899461406e-05, + "loss": 0.631, + "step": 2934 + }, + { + "epoch": 0.11, + "learning_rate": 4.815016670941267e-05, + "loss": 0.548, + "step": 2935 + }, + { + "epoch": 0.11, + "learning_rate": 4.814952551936394e-05, + "loss": 0.5959, + "step": 2936 + }, + { + "epoch": 0.11, + "learning_rate": 4.8148884329315214e-05, + "loss": 0.5802, + "step": 2937 + }, + { + "epoch": 0.11, + "learning_rate": 4.814824313926648e-05, + "loss": 0.6362, + "step": 2938 + }, + { + "epoch": 0.11, + "learning_rate": 4.814760194921775e-05, + "loss": 0.6076, + "step": 2939 + }, + { + "epoch": 0.11, + "learning_rate": 4.8146960759169016e-05, + "loss": 0.5652, + "step": 2940 + }, + { + "epoch": 0.11, + "learning_rate": 4.814631956912029e-05, + "loss": 0.5973, + "step": 2941 + }, + { + "epoch": 0.11, + "learning_rate": 4.814567837907156e-05, + "loss": 0.5709, + "step": 2942 + }, + { + "epoch": 0.11, + "learning_rate": 4.814503718902283e-05, + "loss": 0.6402, + "step": 2943 + }, + { + "epoch": 0.11, + "learning_rate": 4.8144395998974096e-05, + "loss": 0.6445, + "step": 2944 + }, + { + "epoch": 0.11, + "learning_rate": 4.814375480892537e-05, + "loss": 0.5905, + "step": 2945 + }, + { + "epoch": 0.11, + "learning_rate": 4.814311361887664e-05, + "loss": 0.5192, + "step": 2946 + }, + { + "epoch": 0.11, + "learning_rate": 4.8142472428827904e-05, + "loss": 0.6331, + "step": 2947 + }, + { + "epoch": 0.11, + "learning_rate": 4.8141831238779176e-05, + "loss": 0.5151, + "step": 2948 + }, + { + "epoch": 0.11, + "learning_rate": 4.814119004873045e-05, + "loss": 0.619, + "step": 2949 + }, + { + "epoch": 0.11, + "learning_rate": 4.814054885868172e-05, + "loss": 0.6083, + "step": 2950 + }, + { + "epoch": 0.11, + "learning_rate": 4.8139907668632984e-05, + "loss": 0.5158, + "step": 2951 + }, + { + "epoch": 0.11, + "learning_rate": 4.8139266478584256e-05, + "loss": 0.6135, + "step": 2952 + }, + { + "epoch": 0.11, + "learning_rate": 4.813862528853552e-05, + "loss": 0.5415, + "step": 2953 + }, + { + "epoch": 0.11, + "learning_rate": 4.813798409848679e-05, + "loss": 0.6406, + "step": 2954 + }, + { + "epoch": 0.11, + "learning_rate": 4.8137342908438064e-05, + "loss": 0.5842, + "step": 2955 + }, + { + "epoch": 0.11, + "learning_rate": 4.813670171838933e-05, + "loss": 0.6806, + "step": 2956 + }, + { + "epoch": 0.11, + "learning_rate": 4.81360605283406e-05, + "loss": 0.6058, + "step": 2957 + }, + { + "epoch": 0.11, + "learning_rate": 4.813541933829187e-05, + "loss": 0.6466, + "step": 2958 + }, + { + "epoch": 0.11, + "learning_rate": 4.8134778148243144e-05, + "loss": 0.5783, + "step": 2959 + }, + { + "epoch": 0.11, + "learning_rate": 4.813413695819441e-05, + "loss": 0.6344, + "step": 2960 + }, + { + "epoch": 0.11, + "learning_rate": 4.813349576814568e-05, + "loss": 0.6575, + "step": 2961 + }, + { + "epoch": 0.11, + "learning_rate": 4.8132854578096946e-05, + "loss": 0.6069, + "step": 2962 + }, + { + "epoch": 0.11, + "learning_rate": 4.813221338804822e-05, + "loss": 0.5447, + "step": 2963 + }, + { + "epoch": 0.11, + "learning_rate": 4.813157219799949e-05, + "loss": 0.7508, + "step": 2964 + }, + { + "epoch": 0.11, + "learning_rate": 4.813093100795076e-05, + "loss": 0.6558, + "step": 2965 + }, + { + "epoch": 0.11, + "learning_rate": 4.813028981790203e-05, + "loss": 0.5367, + "step": 2966 + }, + { + "epoch": 0.11, + "learning_rate": 4.81296486278533e-05, + "loss": 0.6356, + "step": 2967 + }, + { + "epoch": 0.11, + "learning_rate": 4.812900743780457e-05, + "loss": 0.5604, + "step": 2968 + }, + { + "epoch": 0.11, + "learning_rate": 4.8128366247755835e-05, + "loss": 0.6462, + "step": 2969 + }, + { + "epoch": 0.11, + "learning_rate": 4.8127725057707106e-05, + "loss": 0.6353, + "step": 2970 + }, + { + "epoch": 0.11, + "learning_rate": 4.812708386765837e-05, + "loss": 0.5342, + "step": 2971 + }, + { + "epoch": 0.11, + "learning_rate": 4.812644267760964e-05, + "loss": 0.5803, + "step": 2972 + }, + { + "epoch": 0.11, + "learning_rate": 4.8125801487560915e-05, + "loss": 0.6289, + "step": 2973 + }, + { + "epoch": 0.11, + "learning_rate": 4.8125160297512186e-05, + "loss": 0.6035, + "step": 2974 + }, + { + "epoch": 0.11, + "learning_rate": 4.812451910746346e-05, + "loss": 0.578, + "step": 2975 + }, + { + "epoch": 0.11, + "learning_rate": 4.812387791741472e-05, + "loss": 0.6948, + "step": 2976 + }, + { + "epoch": 0.11, + "learning_rate": 4.8123236727365995e-05, + "loss": 0.5908, + "step": 2977 + }, + { + "epoch": 0.11, + "learning_rate": 4.812259553731726e-05, + "loss": 0.626, + "step": 2978 + }, + { + "epoch": 0.11, + "learning_rate": 4.812195434726853e-05, + "loss": 0.611, + "step": 2979 + }, + { + "epoch": 0.11, + "learning_rate": 4.8121313157219796e-05, + "loss": 0.6115, + "step": 2980 + }, + { + "epoch": 0.11, + "learning_rate": 4.8120671967171075e-05, + "loss": 0.5476, + "step": 2981 + }, + { + "epoch": 0.11, + "learning_rate": 4.8120030777122347e-05, + "loss": 0.6723, + "step": 2982 + }, + { + "epoch": 0.11, + "learning_rate": 4.811938958707361e-05, + "loss": 0.5919, + "step": 2983 + }, + { + "epoch": 0.11, + "learning_rate": 4.811874839702488e-05, + "loss": 0.5312, + "step": 2984 + }, + { + "epoch": 0.11, + "learning_rate": 4.811810720697615e-05, + "loss": 0.5351, + "step": 2985 + }, + { + "epoch": 0.11, + "learning_rate": 4.811746601692742e-05, + "loss": 0.5858, + "step": 2986 + }, + { + "epoch": 0.11, + "learning_rate": 4.8116824826878685e-05, + "loss": 0.6524, + "step": 2987 + }, + { + "epoch": 0.11, + "learning_rate": 4.8116183636829956e-05, + "loss": 0.5887, + "step": 2988 + }, + { + "epoch": 0.11, + "learning_rate": 4.811554244678123e-05, + "loss": 0.6442, + "step": 2989 + }, + { + "epoch": 0.11, + "learning_rate": 4.81149012567325e-05, + "loss": 0.5482, + "step": 2990 + }, + { + "epoch": 0.11, + "learning_rate": 4.811426006668377e-05, + "loss": 0.571, + "step": 2991 + }, + { + "epoch": 0.12, + "learning_rate": 4.8113618876635037e-05, + "loss": 0.6136, + "step": 2992 + }, + { + "epoch": 0.12, + "learning_rate": 4.811297768658631e-05, + "loss": 0.551, + "step": 2993 + }, + { + "epoch": 0.12, + "learning_rate": 4.811233649653757e-05, + "loss": 0.6743, + "step": 2994 + }, + { + "epoch": 0.12, + "learning_rate": 4.8111695306488845e-05, + "loss": 0.6765, + "step": 2995 + }, + { + "epoch": 0.12, + "learning_rate": 4.811105411644011e-05, + "loss": 0.6995, + "step": 2996 + }, + { + "epoch": 0.12, + "learning_rate": 4.811041292639138e-05, + "loss": 0.5916, + "step": 2997 + }, + { + "epoch": 0.12, + "learning_rate": 4.810977173634266e-05, + "loss": 0.4935, + "step": 2998 + }, + { + "epoch": 0.12, + "learning_rate": 4.8109130546293925e-05, + "loss": 0.736, + "step": 2999 + }, + { + "epoch": 0.12, + "learning_rate": 4.81084893562452e-05, + "loss": 0.615, + "step": 3000 + }, + { + "epoch": 0.12, + "learning_rate": 4.810784816619646e-05, + "loss": 0.5856, + "step": 3001 + }, + { + "epoch": 0.12, + "learning_rate": 4.8107206976147733e-05, + "loss": 0.5849, + "step": 3002 + }, + { + "epoch": 0.12, + "learning_rate": 4.8106565786099e-05, + "loss": 0.544, + "step": 3003 + }, + { + "epoch": 0.12, + "learning_rate": 4.810592459605027e-05, + "loss": 0.5699, + "step": 3004 + }, + { + "epoch": 0.12, + "learning_rate": 4.810528340600154e-05, + "loss": 0.6792, + "step": 3005 + }, + { + "epoch": 0.12, + "learning_rate": 4.8104642215952813e-05, + "loss": 0.5704, + "step": 3006 + }, + { + "epoch": 0.12, + "learning_rate": 4.8104001025904085e-05, + "loss": 0.5725, + "step": 3007 + }, + { + "epoch": 0.12, + "learning_rate": 4.810335983585535e-05, + "loss": 0.5894, + "step": 3008 + }, + { + "epoch": 0.12, + "learning_rate": 4.810271864580662e-05, + "loss": 0.6052, + "step": 3009 + }, + { + "epoch": 0.12, + "learning_rate": 4.810207745575789e-05, + "loss": 0.6745, + "step": 3010 + }, + { + "epoch": 0.12, + "learning_rate": 4.810143626570916e-05, + "loss": 0.6395, + "step": 3011 + }, + { + "epoch": 0.12, + "learning_rate": 4.8100795075660423e-05, + "loss": 0.5997, + "step": 3012 + }, + { + "epoch": 0.12, + "learning_rate": 4.8100153885611695e-05, + "loss": 0.6049, + "step": 3013 + }, + { + "epoch": 0.12, + "learning_rate": 4.809951269556297e-05, + "loss": 0.5928, + "step": 3014 + }, + { + "epoch": 0.12, + "learning_rate": 4.809887150551424e-05, + "loss": 0.5229, + "step": 3015 + }, + { + "epoch": 0.12, + "learning_rate": 4.809823031546551e-05, + "loss": 0.5184, + "step": 3016 + }, + { + "epoch": 0.12, + "learning_rate": 4.8097589125416775e-05, + "loss": 0.5862, + "step": 3017 + }, + { + "epoch": 0.12, + "learning_rate": 4.809694793536805e-05, + "loss": 0.5489, + "step": 3018 + }, + { + "epoch": 0.12, + "learning_rate": 4.809630674531931e-05, + "loss": 0.6231, + "step": 3019 + }, + { + "epoch": 0.12, + "learning_rate": 4.8095665555270584e-05, + "loss": 0.5552, + "step": 3020 + }, + { + "epoch": 0.12, + "learning_rate": 4.8095024365221855e-05, + "loss": 0.5526, + "step": 3021 + }, + { + "epoch": 0.12, + "learning_rate": 4.809438317517313e-05, + "loss": 0.6294, + "step": 3022 + }, + { + "epoch": 0.12, + "learning_rate": 4.809374198512439e-05, + "loss": 0.5669, + "step": 3023 + }, + { + "epoch": 0.12, + "learning_rate": 4.8093100795075664e-05, + "loss": 0.6091, + "step": 3024 + }, + { + "epoch": 0.12, + "learning_rate": 4.8092459605026935e-05, + "loss": 0.6419, + "step": 3025 + }, + { + "epoch": 0.12, + "learning_rate": 4.80918184149782e-05, + "loss": 0.5443, + "step": 3026 + }, + { + "epoch": 0.12, + "learning_rate": 4.809117722492947e-05, + "loss": 0.628, + "step": 3027 + }, + { + "epoch": 0.12, + "learning_rate": 4.809053603488074e-05, + "loss": 0.511, + "step": 3028 + }, + { + "epoch": 0.12, + "learning_rate": 4.808989484483201e-05, + "loss": 0.6491, + "step": 3029 + }, + { + "epoch": 0.12, + "learning_rate": 4.808925365478328e-05, + "loss": 0.5309, + "step": 3030 + }, + { + "epoch": 0.12, + "learning_rate": 4.808861246473455e-05, + "loss": 0.5819, + "step": 3031 + }, + { + "epoch": 0.12, + "learning_rate": 4.808797127468582e-05, + "loss": 0.6367, + "step": 3032 + }, + { + "epoch": 0.12, + "learning_rate": 4.808733008463709e-05, + "loss": 0.5768, + "step": 3033 + }, + { + "epoch": 0.12, + "learning_rate": 4.808668889458836e-05, + "loss": 0.6466, + "step": 3034 + }, + { + "epoch": 0.12, + "learning_rate": 4.8086047704539626e-05, + "loss": 0.6834, + "step": 3035 + }, + { + "epoch": 0.12, + "learning_rate": 4.80854065144909e-05, + "loss": 0.5699, + "step": 3036 + }, + { + "epoch": 0.12, + "learning_rate": 4.808476532444217e-05, + "loss": 0.5967, + "step": 3037 + }, + { + "epoch": 0.12, + "learning_rate": 4.808412413439344e-05, + "loss": 0.6904, + "step": 3038 + }, + { + "epoch": 0.12, + "learning_rate": 4.8083482944344706e-05, + "loss": 0.654, + "step": 3039 + }, + { + "epoch": 0.12, + "learning_rate": 4.808284175429598e-05, + "loss": 0.5711, + "step": 3040 + }, + { + "epoch": 0.12, + "learning_rate": 4.808220056424724e-05, + "loss": 0.6394, + "step": 3041 + }, + { + "epoch": 0.12, + "learning_rate": 4.8081559374198514e-05, + "loss": 0.6171, + "step": 3042 + }, + { + "epoch": 0.12, + "learning_rate": 4.8080918184149786e-05, + "loss": 0.7088, + "step": 3043 + }, + { + "epoch": 0.12, + "learning_rate": 4.808027699410105e-05, + "loss": 0.5951, + "step": 3044 + }, + { + "epoch": 0.12, + "learning_rate": 4.807963580405232e-05, + "loss": 0.6325, + "step": 3045 + }, + { + "epoch": 0.12, + "learning_rate": 4.8078994614003594e-05, + "loss": 0.6295, + "step": 3046 + }, + { + "epoch": 0.12, + "learning_rate": 4.8078353423954866e-05, + "loss": 0.6659, + "step": 3047 + }, + { + "epoch": 0.12, + "learning_rate": 4.807771223390613e-05, + "loss": 0.682, + "step": 3048 + }, + { + "epoch": 0.12, + "learning_rate": 4.80770710438574e-05, + "loss": 0.5896, + "step": 3049 + }, + { + "epoch": 0.12, + "learning_rate": 4.807642985380867e-05, + "loss": 0.6238, + "step": 3050 + }, + { + "epoch": 0.12, + "learning_rate": 4.807578866375994e-05, + "loss": 0.6323, + "step": 3051 + }, + { + "epoch": 0.12, + "learning_rate": 4.807514747371121e-05, + "loss": 0.6326, + "step": 3052 + }, + { + "epoch": 0.12, + "learning_rate": 4.807450628366248e-05, + "loss": 0.5884, + "step": 3053 + }, + { + "epoch": 0.12, + "learning_rate": 4.8073865093613754e-05, + "loss": 0.548, + "step": 3054 + }, + { + "epoch": 0.12, + "learning_rate": 4.807322390356502e-05, + "loss": 0.6529, + "step": 3055 + }, + { + "epoch": 0.12, + "learning_rate": 4.807258271351629e-05, + "loss": 0.585, + "step": 3056 + }, + { + "epoch": 0.12, + "learning_rate": 4.8071941523467556e-05, + "loss": 0.65, + "step": 3057 + }, + { + "epoch": 0.12, + "learning_rate": 4.807130033341883e-05, + "loss": 0.5924, + "step": 3058 + }, + { + "epoch": 0.12, + "learning_rate": 4.807065914337009e-05, + "loss": 0.5309, + "step": 3059 + }, + { + "epoch": 0.12, + "learning_rate": 4.8070017953321364e-05, + "loss": 0.5491, + "step": 3060 + }, + { + "epoch": 0.12, + "learning_rate": 4.8069376763272636e-05, + "loss": 0.6194, + "step": 3061 + }, + { + "epoch": 0.12, + "learning_rate": 4.806873557322391e-05, + "loss": 0.5653, + "step": 3062 + }, + { + "epoch": 0.12, + "learning_rate": 4.806809438317518e-05, + "loss": 0.5774, + "step": 3063 + }, + { + "epoch": 0.12, + "learning_rate": 4.8067453193126444e-05, + "loss": 0.5753, + "step": 3064 + }, + { + "epoch": 0.12, + "learning_rate": 4.8066812003077716e-05, + "loss": 0.6248, + "step": 3065 + }, + { + "epoch": 0.12, + "learning_rate": 4.806617081302898e-05, + "loss": 0.5918, + "step": 3066 + }, + { + "epoch": 0.12, + "learning_rate": 4.806552962298025e-05, + "loss": 0.5243, + "step": 3067 + }, + { + "epoch": 0.12, + "learning_rate": 4.806488843293152e-05, + "loss": 0.5649, + "step": 3068 + }, + { + "epoch": 0.12, + "learning_rate": 4.8064247242882796e-05, + "loss": 0.5345, + "step": 3069 + }, + { + "epoch": 0.12, + "learning_rate": 4.806360605283407e-05, + "loss": 0.6475, + "step": 3070 + }, + { + "epoch": 0.12, + "learning_rate": 4.806296486278533e-05, + "loss": 0.6431, + "step": 3071 + }, + { + "epoch": 0.12, + "learning_rate": 4.8062323672736604e-05, + "loss": 0.6128, + "step": 3072 + }, + { + "epoch": 0.12, + "learning_rate": 4.806168248268787e-05, + "loss": 0.6429, + "step": 3073 + }, + { + "epoch": 0.12, + "learning_rate": 4.806104129263914e-05, + "loss": 0.5803, + "step": 3074 + }, + { + "epoch": 0.12, + "learning_rate": 4.8060400102590406e-05, + "loss": 0.6141, + "step": 3075 + }, + { + "epoch": 0.12, + "learning_rate": 4.805975891254168e-05, + "loss": 0.6094, + "step": 3076 + }, + { + "epoch": 0.12, + "learning_rate": 4.805911772249295e-05, + "loss": 0.5262, + "step": 3077 + }, + { + "epoch": 0.12, + "learning_rate": 4.805847653244422e-05, + "loss": 0.562, + "step": 3078 + }, + { + "epoch": 0.12, + "learning_rate": 4.805783534239549e-05, + "loss": 0.6812, + "step": 3079 + }, + { + "epoch": 0.12, + "learning_rate": 4.805719415234676e-05, + "loss": 0.5678, + "step": 3080 + }, + { + "epoch": 0.12, + "learning_rate": 4.805655296229803e-05, + "loss": 0.5988, + "step": 3081 + }, + { + "epoch": 0.12, + "learning_rate": 4.8055911772249295e-05, + "loss": 0.5972, + "step": 3082 + }, + { + "epoch": 0.12, + "learning_rate": 4.8055270582200566e-05, + "loss": 0.6466, + "step": 3083 + }, + { + "epoch": 0.12, + "learning_rate": 4.805462939215183e-05, + "loss": 0.6161, + "step": 3084 + }, + { + "epoch": 0.12, + "learning_rate": 4.80539882021031e-05, + "loss": 0.5933, + "step": 3085 + }, + { + "epoch": 0.12, + "learning_rate": 4.8053347012054375e-05, + "loss": 0.6014, + "step": 3086 + }, + { + "epoch": 0.12, + "learning_rate": 4.8052705822005646e-05, + "loss": 0.54, + "step": 3087 + }, + { + "epoch": 0.12, + "learning_rate": 4.805206463195692e-05, + "loss": 0.6164, + "step": 3088 + }, + { + "epoch": 0.12, + "learning_rate": 4.805142344190818e-05, + "loss": 0.5847, + "step": 3089 + }, + { + "epoch": 0.12, + "learning_rate": 4.8050782251859455e-05, + "loss": 0.5734, + "step": 3090 + }, + { + "epoch": 0.12, + "learning_rate": 4.805014106181072e-05, + "loss": 0.5868, + "step": 3091 + }, + { + "epoch": 0.12, + "learning_rate": 4.804949987176199e-05, + "loss": 0.5161, + "step": 3092 + }, + { + "epoch": 0.12, + "learning_rate": 4.804885868171326e-05, + "loss": 0.6001, + "step": 3093 + }, + { + "epoch": 0.12, + "learning_rate": 4.8048217491664535e-05, + "loss": 0.6159, + "step": 3094 + }, + { + "epoch": 0.12, + "learning_rate": 4.80475763016158e-05, + "loss": 0.6444, + "step": 3095 + }, + { + "epoch": 0.12, + "learning_rate": 4.804693511156707e-05, + "loss": 0.6333, + "step": 3096 + }, + { + "epoch": 0.12, + "learning_rate": 4.804629392151834e-05, + "loss": 0.5417, + "step": 3097 + }, + { + "epoch": 0.12, + "learning_rate": 4.804565273146961e-05, + "loss": 0.733, + "step": 3098 + }, + { + "epoch": 0.12, + "learning_rate": 4.804501154142088e-05, + "loss": 0.5393, + "step": 3099 + }, + { + "epoch": 0.12, + "learning_rate": 4.8044370351372145e-05, + "loss": 0.6819, + "step": 3100 + }, + { + "epoch": 0.12, + "learning_rate": 4.8043729161323417e-05, + "loss": 0.5558, + "step": 3101 + }, + { + "epoch": 0.12, + "learning_rate": 4.804308797127469e-05, + "loss": 0.6018, + "step": 3102 + }, + { + "epoch": 0.12, + "learning_rate": 4.804244678122596e-05, + "loss": 0.5652, + "step": 3103 + }, + { + "epoch": 0.12, + "learning_rate": 4.8041805591177225e-05, + "loss": 0.647, + "step": 3104 + }, + { + "epoch": 0.12, + "learning_rate": 4.8041164401128497e-05, + "loss": 0.6136, + "step": 3105 + }, + { + "epoch": 0.12, + "learning_rate": 4.804052321107977e-05, + "loss": 0.62, + "step": 3106 + }, + { + "epoch": 0.12, + "learning_rate": 4.803988202103103e-05, + "loss": 0.6523, + "step": 3107 + }, + { + "epoch": 0.12, + "learning_rate": 4.8039240830982305e-05, + "loss": 0.6305, + "step": 3108 + }, + { + "epoch": 0.12, + "learning_rate": 4.803859964093358e-05, + "loss": 0.6652, + "step": 3109 + }, + { + "epoch": 0.12, + "learning_rate": 4.803795845088485e-05, + "loss": 0.5564, + "step": 3110 + }, + { + "epoch": 0.12, + "learning_rate": 4.803731726083611e-05, + "loss": 0.5607, + "step": 3111 + }, + { + "epoch": 0.12, + "learning_rate": 4.8036676070787385e-05, + "loss": 0.633, + "step": 3112 + }, + { + "epoch": 0.12, + "learning_rate": 4.803603488073865e-05, + "loss": 0.6865, + "step": 3113 + }, + { + "epoch": 0.12, + "learning_rate": 4.803539369068992e-05, + "loss": 0.5189, + "step": 3114 + }, + { + "epoch": 0.12, + "learning_rate": 4.8034752500641193e-05, + "loss": 0.6033, + "step": 3115 + }, + { + "epoch": 0.12, + "learning_rate": 4.803411131059246e-05, + "loss": 0.565, + "step": 3116 + }, + { + "epoch": 0.12, + "learning_rate": 4.803347012054373e-05, + "loss": 0.6253, + "step": 3117 + }, + { + "epoch": 0.12, + "learning_rate": 4.8032828930495e-05, + "loss": 0.5689, + "step": 3118 + }, + { + "epoch": 0.12, + "learning_rate": 4.8032187740446274e-05, + "loss": 0.6514, + "step": 3119 + }, + { + "epoch": 0.12, + "learning_rate": 4.803154655039754e-05, + "loss": 0.6443, + "step": 3120 + }, + { + "epoch": 0.12, + "learning_rate": 4.803090536034881e-05, + "loss": 0.6624, + "step": 3121 + }, + { + "epoch": 0.12, + "learning_rate": 4.8030264170300075e-05, + "loss": 0.7009, + "step": 3122 + }, + { + "epoch": 0.12, + "learning_rate": 4.802962298025135e-05, + "loss": 0.5989, + "step": 3123 + }, + { + "epoch": 0.12, + "learning_rate": 4.802898179020262e-05, + "loss": 0.6454, + "step": 3124 + }, + { + "epoch": 0.12, + "learning_rate": 4.802834060015389e-05, + "loss": 0.6046, + "step": 3125 + }, + { + "epoch": 0.12, + "learning_rate": 4.802769941010516e-05, + "loss": 0.5573, + "step": 3126 + }, + { + "epoch": 0.12, + "learning_rate": 4.802705822005643e-05, + "loss": 0.5935, + "step": 3127 + }, + { + "epoch": 0.12, + "learning_rate": 4.80264170300077e-05, + "loss": 0.5208, + "step": 3128 + }, + { + "epoch": 0.12, + "learning_rate": 4.8025775839958964e-05, + "loss": 0.6065, + "step": 3129 + }, + { + "epoch": 0.12, + "learning_rate": 4.8025134649910235e-05, + "loss": 0.5683, + "step": 3130 + }, + { + "epoch": 0.12, + "learning_rate": 4.80244934598615e-05, + "loss": 0.6327, + "step": 3131 + }, + { + "epoch": 0.12, + "learning_rate": 4.802385226981277e-05, + "loss": 0.5938, + "step": 3132 + }, + { + "epoch": 0.12, + "learning_rate": 4.8023211079764044e-05, + "loss": 0.5962, + "step": 3133 + }, + { + "epoch": 0.12, + "learning_rate": 4.8022569889715315e-05, + "loss": 0.5505, + "step": 3134 + }, + { + "epoch": 0.12, + "learning_rate": 4.802192869966659e-05, + "loss": 0.6212, + "step": 3135 + }, + { + "epoch": 0.12, + "learning_rate": 4.802128750961785e-05, + "loss": 0.6158, + "step": 3136 + }, + { + "epoch": 0.12, + "learning_rate": 4.8020646319569124e-05, + "loss": 0.5215, + "step": 3137 + }, + { + "epoch": 0.12, + "learning_rate": 4.802000512952039e-05, + "loss": 0.6526, + "step": 3138 + }, + { + "epoch": 0.12, + "learning_rate": 4.801936393947166e-05, + "loss": 0.5649, + "step": 3139 + }, + { + "epoch": 0.12, + "learning_rate": 4.8018722749422925e-05, + "loss": 0.5454, + "step": 3140 + }, + { + "epoch": 0.12, + "learning_rate": 4.8018081559374204e-05, + "loss": 0.6875, + "step": 3141 + }, + { + "epoch": 0.12, + "learning_rate": 4.8017440369325476e-05, + "loss": 0.5836, + "step": 3142 + }, + { + "epoch": 0.12, + "learning_rate": 4.801679917927674e-05, + "loss": 0.6105, + "step": 3143 + }, + { + "epoch": 0.12, + "learning_rate": 4.801615798922801e-05, + "loss": 0.5632, + "step": 3144 + }, + { + "epoch": 0.12, + "learning_rate": 4.801551679917928e-05, + "loss": 0.5537, + "step": 3145 + }, + { + "epoch": 0.12, + "learning_rate": 4.801487560913055e-05, + "loss": 0.5009, + "step": 3146 + }, + { + "epoch": 0.12, + "learning_rate": 4.8014234419081814e-05, + "loss": 0.5758, + "step": 3147 + }, + { + "epoch": 0.12, + "learning_rate": 4.8013593229033086e-05, + "loss": 0.5773, + "step": 3148 + }, + { + "epoch": 0.12, + "learning_rate": 4.801295203898436e-05, + "loss": 0.633, + "step": 3149 + }, + { + "epoch": 0.12, + "learning_rate": 4.801231084893563e-05, + "loss": 0.7365, + "step": 3150 + }, + { + "epoch": 0.12, + "learning_rate": 4.80116696588869e-05, + "loss": 0.6112, + "step": 3151 + }, + { + "epoch": 0.12, + "learning_rate": 4.8011028468838166e-05, + "loss": 0.6926, + "step": 3152 + }, + { + "epoch": 0.12, + "learning_rate": 4.801038727878944e-05, + "loss": 0.5993, + "step": 3153 + }, + { + "epoch": 0.12, + "learning_rate": 4.80097460887407e-05, + "loss": 0.5915, + "step": 3154 + }, + { + "epoch": 0.12, + "learning_rate": 4.8009104898691974e-05, + "loss": 0.535, + "step": 3155 + }, + { + "epoch": 0.12, + "learning_rate": 4.800846370864324e-05, + "loss": 0.6594, + "step": 3156 + }, + { + "epoch": 0.12, + "learning_rate": 4.800782251859452e-05, + "loss": 0.6401, + "step": 3157 + }, + { + "epoch": 0.12, + "learning_rate": 4.800718132854578e-05, + "loss": 0.6396, + "step": 3158 + }, + { + "epoch": 0.12, + "learning_rate": 4.8006540138497054e-05, + "loss": 0.5599, + "step": 3159 + }, + { + "epoch": 0.12, + "learning_rate": 4.8005898948448326e-05, + "loss": 0.6055, + "step": 3160 + }, + { + "epoch": 0.12, + "learning_rate": 4.800525775839959e-05, + "loss": 0.5738, + "step": 3161 + }, + { + "epoch": 0.12, + "learning_rate": 4.800461656835086e-05, + "loss": 0.6558, + "step": 3162 + }, + { + "epoch": 0.12, + "learning_rate": 4.800397537830213e-05, + "loss": 0.5668, + "step": 3163 + }, + { + "epoch": 0.12, + "learning_rate": 4.80033341882534e-05, + "loss": 0.6372, + "step": 3164 + }, + { + "epoch": 0.12, + "learning_rate": 4.800269299820467e-05, + "loss": 0.5727, + "step": 3165 + }, + { + "epoch": 0.12, + "learning_rate": 4.800205180815594e-05, + "loss": 0.7476, + "step": 3166 + }, + { + "epoch": 0.12, + "learning_rate": 4.800141061810721e-05, + "loss": 0.6534, + "step": 3167 + }, + { + "epoch": 0.12, + "learning_rate": 4.800076942805848e-05, + "loss": 0.5795, + "step": 3168 + }, + { + "epoch": 0.12, + "learning_rate": 4.800012823800975e-05, + "loss": 0.5947, + "step": 3169 + }, + { + "epoch": 0.12, + "learning_rate": 4.7999487047961016e-05, + "loss": 0.6103, + "step": 3170 + }, + { + "epoch": 0.12, + "learning_rate": 4.799884585791229e-05, + "loss": 0.595, + "step": 3171 + }, + { + "epoch": 0.12, + "learning_rate": 4.799820466786355e-05, + "loss": 0.5449, + "step": 3172 + }, + { + "epoch": 0.12, + "learning_rate": 4.7997563477814824e-05, + "loss": 0.6104, + "step": 3173 + }, + { + "epoch": 0.12, + "learning_rate": 4.7996922287766096e-05, + "loss": 0.5528, + "step": 3174 + }, + { + "epoch": 0.12, + "learning_rate": 4.799628109771737e-05, + "loss": 0.5726, + "step": 3175 + }, + { + "epoch": 0.12, + "learning_rate": 4.799563990766864e-05, + "loss": 0.6386, + "step": 3176 + }, + { + "epoch": 0.12, + "learning_rate": 4.7994998717619904e-05, + "loss": 0.5394, + "step": 3177 + }, + { + "epoch": 0.12, + "learning_rate": 4.7994357527571176e-05, + "loss": 0.5692, + "step": 3178 + }, + { + "epoch": 0.12, + "learning_rate": 4.799371633752244e-05, + "loss": 0.5951, + "step": 3179 + }, + { + "epoch": 0.12, + "learning_rate": 4.799307514747371e-05, + "loss": 0.5475, + "step": 3180 + }, + { + "epoch": 0.12, + "learning_rate": 4.7992433957424984e-05, + "loss": 0.5864, + "step": 3181 + }, + { + "epoch": 0.12, + "learning_rate": 4.7991792767376256e-05, + "loss": 0.5977, + "step": 3182 + }, + { + "epoch": 0.12, + "learning_rate": 4.799115157732752e-05, + "loss": 0.6622, + "step": 3183 + }, + { + "epoch": 0.12, + "learning_rate": 4.799051038727879e-05, + "loss": 0.5659, + "step": 3184 + }, + { + "epoch": 0.12, + "learning_rate": 4.7989869197230065e-05, + "loss": 0.6036, + "step": 3185 + }, + { + "epoch": 0.12, + "learning_rate": 4.798922800718133e-05, + "loss": 0.5337, + "step": 3186 + }, + { + "epoch": 0.12, + "learning_rate": 4.79885868171326e-05, + "loss": 0.6624, + "step": 3187 + }, + { + "epoch": 0.12, + "learning_rate": 4.7987945627083866e-05, + "loss": 0.6045, + "step": 3188 + }, + { + "epoch": 0.12, + "learning_rate": 4.798730443703514e-05, + "loss": 0.571, + "step": 3189 + }, + { + "epoch": 0.12, + "learning_rate": 4.798666324698641e-05, + "loss": 0.6289, + "step": 3190 + }, + { + "epoch": 0.12, + "learning_rate": 4.798602205693768e-05, + "loss": 0.5022, + "step": 3191 + }, + { + "epoch": 0.12, + "learning_rate": 4.7985380866888946e-05, + "loss": 0.5648, + "step": 3192 + }, + { + "epoch": 0.12, + "learning_rate": 4.798473967684022e-05, + "loss": 0.6449, + "step": 3193 + }, + { + "epoch": 0.12, + "learning_rate": 4.798409848679149e-05, + "loss": 0.6586, + "step": 3194 + }, + { + "epoch": 0.12, + "learning_rate": 4.7983457296742755e-05, + "loss": 0.6116, + "step": 3195 + }, + { + "epoch": 0.12, + "learning_rate": 4.7982816106694026e-05, + "loss": 0.6219, + "step": 3196 + }, + { + "epoch": 0.12, + "learning_rate": 4.79821749166453e-05, + "loss": 0.6038, + "step": 3197 + }, + { + "epoch": 0.12, + "learning_rate": 4.798153372659657e-05, + "loss": 0.6274, + "step": 3198 + }, + { + "epoch": 0.12, + "learning_rate": 4.7980892536547835e-05, + "loss": 0.5436, + "step": 3199 + }, + { + "epoch": 0.12, + "learning_rate": 4.7980251346499106e-05, + "loss": 0.58, + "step": 3200 + }, + { + "epoch": 0.12, + "learning_rate": 4.797961015645037e-05, + "loss": 0.6542, + "step": 3201 + }, + { + "epoch": 0.12, + "learning_rate": 4.797896896640164e-05, + "loss": 0.5981, + "step": 3202 + }, + { + "epoch": 0.12, + "learning_rate": 4.7978327776352915e-05, + "loss": 0.6354, + "step": 3203 + }, + { + "epoch": 0.12, + "learning_rate": 4.797768658630418e-05, + "loss": 0.5521, + "step": 3204 + }, + { + "epoch": 0.12, + "learning_rate": 4.797704539625545e-05, + "loss": 0.6605, + "step": 3205 + }, + { + "epoch": 0.12, + "learning_rate": 4.797640420620672e-05, + "loss": 0.5499, + "step": 3206 + }, + { + "epoch": 0.12, + "learning_rate": 4.7975763016157995e-05, + "loss": 0.5991, + "step": 3207 + }, + { + "epoch": 0.12, + "learning_rate": 4.797512182610926e-05, + "loss": 0.6151, + "step": 3208 + }, + { + "epoch": 0.12, + "learning_rate": 4.797448063606053e-05, + "loss": 0.6849, + "step": 3209 + }, + { + "epoch": 0.12, + "learning_rate": 4.7973839446011796e-05, + "loss": 0.6853, + "step": 3210 + }, + { + "epoch": 0.12, + "learning_rate": 4.797319825596307e-05, + "loss": 0.6103, + "step": 3211 + }, + { + "epoch": 0.12, + "learning_rate": 4.797255706591434e-05, + "loss": 0.5883, + "step": 3212 + }, + { + "epoch": 0.12, + "learning_rate": 4.797191587586561e-05, + "loss": 0.6259, + "step": 3213 + }, + { + "epoch": 0.12, + "learning_rate": 4.797127468581688e-05, + "loss": 0.6578, + "step": 3214 + }, + { + "epoch": 0.12, + "learning_rate": 4.797063349576815e-05, + "loss": 0.5454, + "step": 3215 + }, + { + "epoch": 0.12, + "learning_rate": 4.796999230571942e-05, + "loss": 0.5899, + "step": 3216 + }, + { + "epoch": 0.12, + "learning_rate": 4.7969351115670685e-05, + "loss": 0.6161, + "step": 3217 + }, + { + "epoch": 0.12, + "learning_rate": 4.796870992562196e-05, + "loss": 0.5952, + "step": 3218 + }, + { + "epoch": 0.12, + "learning_rate": 4.796806873557322e-05, + "loss": 0.647, + "step": 3219 + }, + { + "epoch": 0.12, + "learning_rate": 4.796742754552449e-05, + "loss": 0.5909, + "step": 3220 + }, + { + "epoch": 0.12, + "learning_rate": 4.7966786355475765e-05, + "loss": 0.6136, + "step": 3221 + }, + { + "epoch": 0.12, + "learning_rate": 4.796614516542704e-05, + "loss": 0.612, + "step": 3222 + }, + { + "epoch": 0.12, + "learning_rate": 4.796550397537831e-05, + "loss": 0.6103, + "step": 3223 + }, + { + "epoch": 0.12, + "learning_rate": 4.796486278532957e-05, + "loss": 0.5713, + "step": 3224 + }, + { + "epoch": 0.12, + "learning_rate": 4.7964221595280845e-05, + "loss": 0.5418, + "step": 3225 + }, + { + "epoch": 0.12, + "learning_rate": 4.796358040523211e-05, + "loss": 0.6185, + "step": 3226 + }, + { + "epoch": 0.12, + "learning_rate": 4.796293921518338e-05, + "loss": 0.6479, + "step": 3227 + }, + { + "epoch": 0.12, + "learning_rate": 4.796229802513465e-05, + "loss": 0.6522, + "step": 3228 + }, + { + "epoch": 0.12, + "learning_rate": 4.7961656835085925e-05, + "loss": 0.5781, + "step": 3229 + }, + { + "epoch": 0.12, + "learning_rate": 4.79610156450372e-05, + "loss": 0.6331, + "step": 3230 + }, + { + "epoch": 0.12, + "learning_rate": 4.796037445498846e-05, + "loss": 0.6201, + "step": 3231 + }, + { + "epoch": 0.12, + "learning_rate": 4.7959733264939734e-05, + "loss": 0.6247, + "step": 3232 + }, + { + "epoch": 0.12, + "learning_rate": 4.7959092074891e-05, + "loss": 0.5439, + "step": 3233 + }, + { + "epoch": 0.12, + "learning_rate": 4.795845088484227e-05, + "loss": 0.6085, + "step": 3234 + }, + { + "epoch": 0.12, + "learning_rate": 4.7957809694793535e-05, + "loss": 0.5434, + "step": 3235 + }, + { + "epoch": 0.12, + "learning_rate": 4.795716850474481e-05, + "loss": 0.5181, + "step": 3236 + }, + { + "epoch": 0.12, + "learning_rate": 4.795652731469608e-05, + "loss": 0.6093, + "step": 3237 + }, + { + "epoch": 0.12, + "learning_rate": 4.795588612464735e-05, + "loss": 0.5799, + "step": 3238 + }, + { + "epoch": 0.12, + "learning_rate": 4.795524493459862e-05, + "loss": 0.558, + "step": 3239 + }, + { + "epoch": 0.12, + "learning_rate": 4.795460374454989e-05, + "loss": 0.6896, + "step": 3240 + }, + { + "epoch": 0.12, + "learning_rate": 4.795396255450116e-05, + "loss": 0.6266, + "step": 3241 + }, + { + "epoch": 0.12, + "learning_rate": 4.7953321364452424e-05, + "loss": 0.6813, + "step": 3242 + }, + { + "epoch": 0.12, + "learning_rate": 4.7952680174403695e-05, + "loss": 0.5868, + "step": 3243 + }, + { + "epoch": 0.12, + "learning_rate": 4.795203898435496e-05, + "loss": 0.6677, + "step": 3244 + }, + { + "epoch": 0.12, + "learning_rate": 4.795139779430623e-05, + "loss": 0.5816, + "step": 3245 + }, + { + "epoch": 0.12, + "learning_rate": 4.7950756604257504e-05, + "loss": 0.6025, + "step": 3246 + }, + { + "epoch": 0.12, + "learning_rate": 4.7950115414208775e-05, + "loss": 0.6285, + "step": 3247 + }, + { + "epoch": 0.12, + "learning_rate": 4.794947422416005e-05, + "loss": 0.6645, + "step": 3248 + }, + { + "epoch": 0.12, + "learning_rate": 4.794883303411131e-05, + "loss": 0.623, + "step": 3249 + }, + { + "epoch": 0.12, + "learning_rate": 4.7948191844062584e-05, + "loss": 0.5449, + "step": 3250 + }, + { + "epoch": 0.12, + "learning_rate": 4.794755065401385e-05, + "loss": 0.624, + "step": 3251 + }, + { + "epoch": 0.13, + "learning_rate": 4.794690946396512e-05, + "loss": 0.5597, + "step": 3252 + }, + { + "epoch": 0.13, + "learning_rate": 4.794626827391639e-05, + "loss": 0.582, + "step": 3253 + }, + { + "epoch": 0.13, + "learning_rate": 4.7945627083867664e-05, + "loss": 0.5266, + "step": 3254 + }, + { + "epoch": 0.13, + "learning_rate": 4.794498589381893e-05, + "loss": 0.6048, + "step": 3255 + }, + { + "epoch": 0.13, + "learning_rate": 4.79443447037702e-05, + "loss": 0.7219, + "step": 3256 + }, + { + "epoch": 0.13, + "learning_rate": 4.794370351372147e-05, + "loss": 0.6589, + "step": 3257 + }, + { + "epoch": 0.13, + "learning_rate": 4.794306232367274e-05, + "loss": 0.5728, + "step": 3258 + }, + { + "epoch": 0.13, + "learning_rate": 4.794242113362401e-05, + "loss": 0.6379, + "step": 3259 + }, + { + "epoch": 0.13, + "learning_rate": 4.7941779943575274e-05, + "loss": 0.6256, + "step": 3260 + }, + { + "epoch": 0.13, + "learning_rate": 4.7941138753526546e-05, + "loss": 0.6981, + "step": 3261 + }, + { + "epoch": 0.13, + "learning_rate": 4.794049756347782e-05, + "loss": 0.6538, + "step": 3262 + }, + { + "epoch": 0.13, + "learning_rate": 4.793985637342909e-05, + "loss": 0.6105, + "step": 3263 + }, + { + "epoch": 0.13, + "learning_rate": 4.7939215183380354e-05, + "loss": 0.5636, + "step": 3264 + }, + { + "epoch": 0.13, + "learning_rate": 4.7938573993331626e-05, + "loss": 0.5981, + "step": 3265 + }, + { + "epoch": 0.13, + "learning_rate": 4.79379328032829e-05, + "loss": 0.5728, + "step": 3266 + }, + { + "epoch": 0.13, + "learning_rate": 4.793729161323416e-05, + "loss": 0.6081, + "step": 3267 + }, + { + "epoch": 0.13, + "learning_rate": 4.7936650423185434e-05, + "loss": 0.5961, + "step": 3268 + }, + { + "epoch": 0.13, + "learning_rate": 4.7936009233136706e-05, + "loss": 0.5984, + "step": 3269 + }, + { + "epoch": 0.13, + "learning_rate": 4.793536804308798e-05, + "loss": 0.635, + "step": 3270 + }, + { + "epoch": 0.13, + "learning_rate": 4.793472685303924e-05, + "loss": 0.608, + "step": 3271 + }, + { + "epoch": 0.13, + "learning_rate": 4.7934085662990514e-05, + "loss": 0.5942, + "step": 3272 + }, + { + "epoch": 0.13, + "learning_rate": 4.793344447294178e-05, + "loss": 0.6118, + "step": 3273 + }, + { + "epoch": 0.13, + "learning_rate": 4.793280328289305e-05, + "loss": 0.5594, + "step": 3274 + }, + { + "epoch": 0.13, + "learning_rate": 4.793216209284432e-05, + "loss": 0.6025, + "step": 3275 + }, + { + "epoch": 0.13, + "learning_rate": 4.793152090279559e-05, + "loss": 0.6169, + "step": 3276 + }, + { + "epoch": 0.13, + "learning_rate": 4.793087971274686e-05, + "loss": 0.6089, + "step": 3277 + }, + { + "epoch": 0.13, + "learning_rate": 4.793023852269813e-05, + "loss": 0.6567, + "step": 3278 + }, + { + "epoch": 0.13, + "learning_rate": 4.79295973326494e-05, + "loss": 0.5361, + "step": 3279 + }, + { + "epoch": 0.13, + "learning_rate": 4.792895614260067e-05, + "loss": 0.6247, + "step": 3280 + }, + { + "epoch": 0.13, + "learning_rate": 4.792831495255194e-05, + "loss": 0.5877, + "step": 3281 + }, + { + "epoch": 0.13, + "learning_rate": 4.7927673762503204e-05, + "loss": 0.6482, + "step": 3282 + }, + { + "epoch": 0.13, + "learning_rate": 4.7927032572454476e-05, + "loss": 0.5394, + "step": 3283 + }, + { + "epoch": 0.13, + "learning_rate": 4.792639138240575e-05, + "loss": 0.6693, + "step": 3284 + }, + { + "epoch": 0.13, + "learning_rate": 4.792575019235702e-05, + "loss": 0.6527, + "step": 3285 + }, + { + "epoch": 0.13, + "learning_rate": 4.792510900230829e-05, + "loss": 0.5254, + "step": 3286 + }, + { + "epoch": 0.13, + "learning_rate": 4.7924467812259556e-05, + "loss": 0.6128, + "step": 3287 + }, + { + "epoch": 0.13, + "learning_rate": 4.792382662221083e-05, + "loss": 0.5416, + "step": 3288 + }, + { + "epoch": 0.13, + "learning_rate": 4.792318543216209e-05, + "loss": 0.6424, + "step": 3289 + }, + { + "epoch": 0.13, + "learning_rate": 4.7922544242113364e-05, + "loss": 0.5477, + "step": 3290 + }, + { + "epoch": 0.13, + "learning_rate": 4.792190305206463e-05, + "loss": 0.5556, + "step": 3291 + }, + { + "epoch": 0.13, + "learning_rate": 4.79212618620159e-05, + "loss": 0.58, + "step": 3292 + }, + { + "epoch": 0.13, + "learning_rate": 4.792062067196717e-05, + "loss": 0.5974, + "step": 3293 + }, + { + "epoch": 0.13, + "learning_rate": 4.7919979481918444e-05, + "loss": 0.5674, + "step": 3294 + }, + { + "epoch": 0.13, + "learning_rate": 4.7919338291869716e-05, + "loss": 0.6526, + "step": 3295 + }, + { + "epoch": 0.13, + "learning_rate": 4.791869710182098e-05, + "loss": 0.7134, + "step": 3296 + }, + { + "epoch": 0.13, + "learning_rate": 4.791805591177225e-05, + "loss": 0.6921, + "step": 3297 + }, + { + "epoch": 0.13, + "learning_rate": 4.791741472172352e-05, + "loss": 0.6833, + "step": 3298 + }, + { + "epoch": 0.13, + "learning_rate": 4.791677353167479e-05, + "loss": 0.5606, + "step": 3299 + }, + { + "epoch": 0.13, + "learning_rate": 4.7916132341626054e-05, + "loss": 0.6493, + "step": 3300 + }, + { + "epoch": 0.13, + "learning_rate": 4.791549115157733e-05, + "loss": 0.6064, + "step": 3301 + }, + { + "epoch": 0.13, + "learning_rate": 4.7914849961528605e-05, + "loss": 0.5726, + "step": 3302 + }, + { + "epoch": 0.13, + "learning_rate": 4.791420877147987e-05, + "loss": 0.583, + "step": 3303 + }, + { + "epoch": 0.13, + "learning_rate": 4.791356758143114e-05, + "loss": 0.6372, + "step": 3304 + }, + { + "epoch": 0.13, + "learning_rate": 4.7912926391382406e-05, + "loss": 0.6086, + "step": 3305 + }, + { + "epoch": 0.13, + "learning_rate": 4.791228520133368e-05, + "loss": 0.6623, + "step": 3306 + }, + { + "epoch": 0.13, + "learning_rate": 4.791164401128494e-05, + "loss": 0.5057, + "step": 3307 + }, + { + "epoch": 0.13, + "learning_rate": 4.7911002821236215e-05, + "loss": 0.5421, + "step": 3308 + }, + { + "epoch": 0.13, + "learning_rate": 4.7910361631187486e-05, + "loss": 0.502, + "step": 3309 + }, + { + "epoch": 0.13, + "learning_rate": 4.790972044113876e-05, + "loss": 0.5597, + "step": 3310 + }, + { + "epoch": 0.13, + "learning_rate": 4.790907925109003e-05, + "loss": 0.6641, + "step": 3311 + }, + { + "epoch": 0.13, + "learning_rate": 4.7908438061041295e-05, + "loss": 0.6874, + "step": 3312 + }, + { + "epoch": 0.13, + "learning_rate": 4.7907796870992566e-05, + "loss": 0.6303, + "step": 3313 + }, + { + "epoch": 0.13, + "learning_rate": 4.790715568094383e-05, + "loss": 0.6674, + "step": 3314 + }, + { + "epoch": 0.13, + "learning_rate": 4.79065144908951e-05, + "loss": 0.5769, + "step": 3315 + }, + { + "epoch": 0.13, + "learning_rate": 4.790587330084637e-05, + "loss": 0.6952, + "step": 3316 + }, + { + "epoch": 0.13, + "learning_rate": 4.7905232110797647e-05, + "loss": 0.6051, + "step": 3317 + }, + { + "epoch": 0.13, + "learning_rate": 4.790459092074891e-05, + "loss": 0.5959, + "step": 3318 + }, + { + "epoch": 0.13, + "learning_rate": 4.790394973070018e-05, + "loss": 0.6388, + "step": 3319 + }, + { + "epoch": 0.13, + "learning_rate": 4.7903308540651455e-05, + "loss": 0.5984, + "step": 3320 + }, + { + "epoch": 0.13, + "learning_rate": 4.790266735060272e-05, + "loss": 0.5767, + "step": 3321 + }, + { + "epoch": 0.13, + "learning_rate": 4.790202616055399e-05, + "loss": 0.6219, + "step": 3322 + }, + { + "epoch": 0.13, + "learning_rate": 4.7901384970505256e-05, + "loss": 0.6597, + "step": 3323 + }, + { + "epoch": 0.13, + "learning_rate": 4.790074378045653e-05, + "loss": 0.5944, + "step": 3324 + }, + { + "epoch": 0.13, + "learning_rate": 4.79001025904078e-05, + "loss": 0.6376, + "step": 3325 + }, + { + "epoch": 0.13, + "learning_rate": 4.789946140035907e-05, + "loss": 0.5667, + "step": 3326 + }, + { + "epoch": 0.13, + "learning_rate": 4.7898820210310337e-05, + "loss": 0.7188, + "step": 3327 + }, + { + "epoch": 0.13, + "learning_rate": 4.789817902026161e-05, + "loss": 0.6505, + "step": 3328 + }, + { + "epoch": 0.13, + "learning_rate": 4.789753783021288e-05, + "loss": 0.6144, + "step": 3329 + }, + { + "epoch": 0.13, + "learning_rate": 4.7896896640164145e-05, + "loss": 0.5842, + "step": 3330 + }, + { + "epoch": 0.13, + "learning_rate": 4.789625545011542e-05, + "loss": 0.6254, + "step": 3331 + }, + { + "epoch": 0.13, + "learning_rate": 4.789561426006668e-05, + "loss": 0.655, + "step": 3332 + }, + { + "epoch": 0.13, + "learning_rate": 4.789497307001795e-05, + "loss": 0.6262, + "step": 3333 + }, + { + "epoch": 0.13, + "learning_rate": 4.7894331879969225e-05, + "loss": 0.6092, + "step": 3334 + }, + { + "epoch": 0.13, + "learning_rate": 4.78936906899205e-05, + "loss": 0.6073, + "step": 3335 + }, + { + "epoch": 0.13, + "learning_rate": 4.789304949987176e-05, + "loss": 0.6226, + "step": 3336 + }, + { + "epoch": 0.13, + "learning_rate": 4.7892408309823033e-05, + "loss": 0.6267, + "step": 3337 + }, + { + "epoch": 0.13, + "learning_rate": 4.7891767119774305e-05, + "loss": 0.5936, + "step": 3338 + }, + { + "epoch": 0.13, + "learning_rate": 4.789112592972557e-05, + "loss": 0.6444, + "step": 3339 + }, + { + "epoch": 0.13, + "learning_rate": 4.789048473967684e-05, + "loss": 0.5452, + "step": 3340 + }, + { + "epoch": 0.13, + "learning_rate": 4.7889843549628114e-05, + "loss": 0.5656, + "step": 3341 + }, + { + "epoch": 0.13, + "learning_rate": 4.7889202359579385e-05, + "loss": 0.5852, + "step": 3342 + }, + { + "epoch": 0.13, + "learning_rate": 4.788856116953065e-05, + "loss": 0.5939, + "step": 3343 + }, + { + "epoch": 0.13, + "learning_rate": 4.788791997948192e-05, + "loss": 0.6323, + "step": 3344 + }, + { + "epoch": 0.13, + "learning_rate": 4.788727878943319e-05, + "loss": 0.5372, + "step": 3345 + }, + { + "epoch": 0.13, + "learning_rate": 4.788663759938446e-05, + "loss": 0.6205, + "step": 3346 + }, + { + "epoch": 0.13, + "learning_rate": 4.788599640933573e-05, + "loss": 0.5579, + "step": 3347 + }, + { + "epoch": 0.13, + "learning_rate": 4.7885355219286995e-05, + "loss": 0.6428, + "step": 3348 + }, + { + "epoch": 0.13, + "learning_rate": 4.788471402923827e-05, + "loss": 0.5769, + "step": 3349 + }, + { + "epoch": 0.13, + "learning_rate": 4.788407283918954e-05, + "loss": 0.6171, + "step": 3350 + }, + { + "epoch": 0.13, + "learning_rate": 4.788343164914081e-05, + "loss": 0.6857, + "step": 3351 + }, + { + "epoch": 0.13, + "learning_rate": 4.7882790459092075e-05, + "loss": 0.6197, + "step": 3352 + }, + { + "epoch": 0.13, + "learning_rate": 4.788214926904335e-05, + "loss": 0.5914, + "step": 3353 + }, + { + "epoch": 0.13, + "learning_rate": 4.788150807899461e-05, + "loss": 0.6237, + "step": 3354 + }, + { + "epoch": 0.13, + "learning_rate": 4.7880866888945884e-05, + "loss": 0.6727, + "step": 3355 + }, + { + "epoch": 0.13, + "learning_rate": 4.7880225698897155e-05, + "loss": 0.6071, + "step": 3356 + }, + { + "epoch": 0.13, + "learning_rate": 4.787958450884843e-05, + "loss": 0.5476, + "step": 3357 + }, + { + "epoch": 0.13, + "learning_rate": 4.78789433187997e-05, + "loss": 0.6158, + "step": 3358 + }, + { + "epoch": 0.13, + "learning_rate": 4.7878302128750964e-05, + "loss": 0.558, + "step": 3359 + }, + { + "epoch": 0.13, + "learning_rate": 4.7877660938702235e-05, + "loss": 0.5556, + "step": 3360 + }, + { + "epoch": 0.13, + "learning_rate": 4.78770197486535e-05, + "loss": 0.662, + "step": 3361 + }, + { + "epoch": 0.13, + "learning_rate": 4.787637855860477e-05, + "loss": 0.6168, + "step": 3362 + }, + { + "epoch": 0.13, + "learning_rate": 4.7875737368556044e-05, + "loss": 0.5374, + "step": 3363 + }, + { + "epoch": 0.13, + "learning_rate": 4.787509617850731e-05, + "loss": 0.506, + "step": 3364 + }, + { + "epoch": 0.13, + "learning_rate": 4.787445498845858e-05, + "loss": 0.6628, + "step": 3365 + }, + { + "epoch": 0.13, + "learning_rate": 4.787381379840985e-05, + "loss": 0.4524, + "step": 3366 + }, + { + "epoch": 0.13, + "learning_rate": 4.7873172608361124e-05, + "loss": 0.6065, + "step": 3367 + }, + { + "epoch": 0.13, + "learning_rate": 4.787253141831239e-05, + "loss": 0.6723, + "step": 3368 + }, + { + "epoch": 0.13, + "learning_rate": 4.787189022826366e-05, + "loss": 0.6131, + "step": 3369 + }, + { + "epoch": 0.13, + "learning_rate": 4.7871249038214926e-05, + "loss": 0.6114, + "step": 3370 + }, + { + "epoch": 0.13, + "learning_rate": 4.78706078481662e-05, + "loss": 0.588, + "step": 3371 + }, + { + "epoch": 0.13, + "learning_rate": 4.786996665811747e-05, + "loss": 0.6444, + "step": 3372 + }, + { + "epoch": 0.13, + "learning_rate": 4.786932546806874e-05, + "loss": 0.5971, + "step": 3373 + }, + { + "epoch": 0.13, + "learning_rate": 4.786868427802001e-05, + "loss": 0.5748, + "step": 3374 + }, + { + "epoch": 0.13, + "learning_rate": 4.786804308797128e-05, + "loss": 0.6699, + "step": 3375 + }, + { + "epoch": 0.13, + "learning_rate": 4.786740189792255e-05, + "loss": 0.6888, + "step": 3376 + }, + { + "epoch": 0.13, + "learning_rate": 4.7866760707873814e-05, + "loss": 0.6156, + "step": 3377 + }, + { + "epoch": 0.13, + "learning_rate": 4.7866119517825086e-05, + "loss": 0.6402, + "step": 3378 + }, + { + "epoch": 0.13, + "learning_rate": 4.786547832777635e-05, + "loss": 0.6284, + "step": 3379 + }, + { + "epoch": 0.13, + "learning_rate": 4.786483713772762e-05, + "loss": 0.5217, + "step": 3380 + }, + { + "epoch": 0.13, + "learning_rate": 4.7864195947678894e-05, + "loss": 0.6463, + "step": 3381 + }, + { + "epoch": 0.13, + "learning_rate": 4.7863554757630166e-05, + "loss": 0.5703, + "step": 3382 + }, + { + "epoch": 0.13, + "learning_rate": 4.786291356758144e-05, + "loss": 0.6059, + "step": 3383 + }, + { + "epoch": 0.13, + "learning_rate": 4.78622723775327e-05, + "loss": 0.5818, + "step": 3384 + }, + { + "epoch": 0.13, + "learning_rate": 4.7861631187483974e-05, + "loss": 0.6879, + "step": 3385 + }, + { + "epoch": 0.13, + "learning_rate": 4.786098999743524e-05, + "loss": 0.5622, + "step": 3386 + }, + { + "epoch": 0.13, + "learning_rate": 4.786034880738651e-05, + "loss": 0.6216, + "step": 3387 + }, + { + "epoch": 0.13, + "learning_rate": 4.7859707617337776e-05, + "loss": 0.6751, + "step": 3388 + }, + { + "epoch": 0.13, + "learning_rate": 4.7859066427289054e-05, + "loss": 0.5506, + "step": 3389 + }, + { + "epoch": 0.13, + "learning_rate": 4.7858425237240326e-05, + "loss": 0.5492, + "step": 3390 + }, + { + "epoch": 0.13, + "learning_rate": 4.785778404719159e-05, + "loss": 0.6195, + "step": 3391 + }, + { + "epoch": 0.13, + "learning_rate": 4.785714285714286e-05, + "loss": 0.6158, + "step": 3392 + }, + { + "epoch": 0.13, + "learning_rate": 4.785650166709413e-05, + "loss": 0.6035, + "step": 3393 + }, + { + "epoch": 0.13, + "learning_rate": 4.78558604770454e-05, + "loss": 0.6335, + "step": 3394 + }, + { + "epoch": 0.13, + "learning_rate": 4.7855219286996664e-05, + "loss": 0.534, + "step": 3395 + }, + { + "epoch": 0.13, + "learning_rate": 4.7854578096947936e-05, + "loss": 0.6331, + "step": 3396 + }, + { + "epoch": 0.13, + "learning_rate": 4.785393690689921e-05, + "loss": 0.6922, + "step": 3397 + }, + { + "epoch": 0.13, + "learning_rate": 4.785329571685048e-05, + "loss": 0.5735, + "step": 3398 + }, + { + "epoch": 0.13, + "learning_rate": 4.785265452680175e-05, + "loss": 0.5906, + "step": 3399 + }, + { + "epoch": 0.13, + "learning_rate": 4.7852013336753016e-05, + "loss": 0.5334, + "step": 3400 + }, + { + "epoch": 0.13, + "learning_rate": 4.785137214670429e-05, + "loss": 0.6192, + "step": 3401 + }, + { + "epoch": 0.13, + "learning_rate": 4.785073095665555e-05, + "loss": 0.5569, + "step": 3402 + }, + { + "epoch": 0.13, + "learning_rate": 4.7850089766606824e-05, + "loss": 0.5837, + "step": 3403 + }, + { + "epoch": 0.13, + "learning_rate": 4.784944857655809e-05, + "loss": 0.6226, + "step": 3404 + }, + { + "epoch": 0.13, + "learning_rate": 4.784880738650936e-05, + "loss": 0.5841, + "step": 3405 + }, + { + "epoch": 0.13, + "learning_rate": 4.784816619646063e-05, + "loss": 0.5473, + "step": 3406 + }, + { + "epoch": 0.13, + "learning_rate": 4.7847525006411904e-05, + "loss": 0.5731, + "step": 3407 + }, + { + "epoch": 0.13, + "learning_rate": 4.7846883816363176e-05, + "loss": 0.6182, + "step": 3408 + }, + { + "epoch": 0.13, + "learning_rate": 4.784624262631444e-05, + "loss": 0.5833, + "step": 3409 + }, + { + "epoch": 0.13, + "learning_rate": 4.784560143626571e-05, + "loss": 0.6883, + "step": 3410 + }, + { + "epoch": 0.13, + "learning_rate": 4.784496024621698e-05, + "loss": 0.6945, + "step": 3411 + }, + { + "epoch": 0.13, + "learning_rate": 4.784431905616825e-05, + "loss": 0.6704, + "step": 3412 + }, + { + "epoch": 0.13, + "learning_rate": 4.784367786611952e-05, + "loss": 0.5589, + "step": 3413 + }, + { + "epoch": 0.13, + "learning_rate": 4.784303667607079e-05, + "loss": 0.5927, + "step": 3414 + }, + { + "epoch": 0.13, + "learning_rate": 4.784239548602206e-05, + "loss": 0.6042, + "step": 3415 + }, + { + "epoch": 0.13, + "learning_rate": 4.784175429597333e-05, + "loss": 0.6035, + "step": 3416 + }, + { + "epoch": 0.13, + "learning_rate": 4.78411131059246e-05, + "loss": 0.6054, + "step": 3417 + }, + { + "epoch": 0.13, + "learning_rate": 4.7840471915875866e-05, + "loss": 0.6665, + "step": 3418 + }, + { + "epoch": 0.13, + "learning_rate": 4.783983072582714e-05, + "loss": 0.651, + "step": 3419 + }, + { + "epoch": 0.13, + "learning_rate": 4.78391895357784e-05, + "loss": 0.6212, + "step": 3420 + }, + { + "epoch": 0.13, + "learning_rate": 4.7838548345729675e-05, + "loss": 0.5811, + "step": 3421 + }, + { + "epoch": 0.13, + "learning_rate": 4.7837907155680946e-05, + "loss": 0.5158, + "step": 3422 + }, + { + "epoch": 0.13, + "learning_rate": 4.783726596563222e-05, + "loss": 0.5397, + "step": 3423 + }, + { + "epoch": 0.13, + "learning_rate": 4.783662477558348e-05, + "loss": 0.5386, + "step": 3424 + }, + { + "epoch": 0.13, + "learning_rate": 4.7835983585534755e-05, + "loss": 0.5972, + "step": 3425 + }, + { + "epoch": 0.13, + "learning_rate": 4.7835342395486026e-05, + "loss": 0.6178, + "step": 3426 + }, + { + "epoch": 0.13, + "learning_rate": 4.783470120543729e-05, + "loss": 0.5773, + "step": 3427 + }, + { + "epoch": 0.13, + "learning_rate": 4.783406001538856e-05, + "loss": 0.6169, + "step": 3428 + }, + { + "epoch": 0.13, + "learning_rate": 4.7833418825339835e-05, + "loss": 0.6272, + "step": 3429 + }, + { + "epoch": 0.13, + "learning_rate": 4.7832777635291107e-05, + "loss": 0.5746, + "step": 3430 + }, + { + "epoch": 0.13, + "learning_rate": 4.783213644524237e-05, + "loss": 0.6819, + "step": 3431 + }, + { + "epoch": 0.13, + "learning_rate": 4.783149525519364e-05, + "loss": 0.6359, + "step": 3432 + }, + { + "epoch": 0.13, + "learning_rate": 4.783085406514491e-05, + "loss": 0.6052, + "step": 3433 + }, + { + "epoch": 0.13, + "learning_rate": 4.783021287509618e-05, + "loss": 0.5729, + "step": 3434 + }, + { + "epoch": 0.13, + "learning_rate": 4.782957168504745e-05, + "loss": 0.5986, + "step": 3435 + }, + { + "epoch": 0.13, + "learning_rate": 4.7828930494998717e-05, + "loss": 0.5815, + "step": 3436 + }, + { + "epoch": 0.13, + "learning_rate": 4.782828930494999e-05, + "loss": 0.5886, + "step": 3437 + }, + { + "epoch": 0.13, + "learning_rate": 4.782764811490126e-05, + "loss": 0.6268, + "step": 3438 + }, + { + "epoch": 0.13, + "learning_rate": 4.782700692485253e-05, + "loss": 0.6803, + "step": 3439 + }, + { + "epoch": 0.13, + "learning_rate": 4.7826365734803797e-05, + "loss": 0.5337, + "step": 3440 + }, + { + "epoch": 0.13, + "learning_rate": 4.782572454475507e-05, + "loss": 0.6191, + "step": 3441 + }, + { + "epoch": 0.13, + "learning_rate": 4.782508335470633e-05, + "loss": 0.5815, + "step": 3442 + }, + { + "epoch": 0.13, + "learning_rate": 4.7824442164657605e-05, + "loss": 0.5757, + "step": 3443 + }, + { + "epoch": 0.13, + "learning_rate": 4.782380097460888e-05, + "loss": 0.6882, + "step": 3444 + }, + { + "epoch": 0.13, + "learning_rate": 4.782315978456015e-05, + "loss": 0.6204, + "step": 3445 + }, + { + "epoch": 0.13, + "learning_rate": 4.782251859451142e-05, + "loss": 0.5776, + "step": 3446 + }, + { + "epoch": 0.13, + "learning_rate": 4.7821877404462685e-05, + "loss": 0.5166, + "step": 3447 + }, + { + "epoch": 0.13, + "learning_rate": 4.782123621441396e-05, + "loss": 0.6135, + "step": 3448 + }, + { + "epoch": 0.13, + "learning_rate": 4.782059502436522e-05, + "loss": 0.6128, + "step": 3449 + }, + { + "epoch": 0.13, + "learning_rate": 4.7819953834316493e-05, + "loss": 0.7172, + "step": 3450 + }, + { + "epoch": 0.13, + "learning_rate": 4.781931264426776e-05, + "loss": 0.576, + "step": 3451 + }, + { + "epoch": 0.13, + "learning_rate": 4.781867145421903e-05, + "loss": 0.6315, + "step": 3452 + }, + { + "epoch": 0.13, + "learning_rate": 4.78180302641703e-05, + "loss": 0.5707, + "step": 3453 + }, + { + "epoch": 0.13, + "learning_rate": 4.7817389074121574e-05, + "loss": 0.6101, + "step": 3454 + }, + { + "epoch": 0.13, + "learning_rate": 4.7816747884072845e-05, + "loss": 0.5604, + "step": 3455 + }, + { + "epoch": 0.13, + "learning_rate": 4.781610669402411e-05, + "loss": 0.5777, + "step": 3456 + }, + { + "epoch": 0.13, + "learning_rate": 4.781546550397538e-05, + "loss": 0.5303, + "step": 3457 + }, + { + "epoch": 0.13, + "learning_rate": 4.781482431392665e-05, + "loss": 0.5472, + "step": 3458 + }, + { + "epoch": 0.13, + "learning_rate": 4.781418312387792e-05, + "loss": 0.5932, + "step": 3459 + }, + { + "epoch": 0.13, + "learning_rate": 4.7813541933829184e-05, + "loss": 0.5833, + "step": 3460 + }, + { + "epoch": 0.13, + "learning_rate": 4.781290074378046e-05, + "loss": 0.6528, + "step": 3461 + }, + { + "epoch": 0.13, + "learning_rate": 4.7812259553731734e-05, + "loss": 0.578, + "step": 3462 + }, + { + "epoch": 0.13, + "learning_rate": 4.7811618363683e-05, + "loss": 0.536, + "step": 3463 + }, + { + "epoch": 0.13, + "learning_rate": 4.781097717363427e-05, + "loss": 0.5432, + "step": 3464 + }, + { + "epoch": 0.13, + "learning_rate": 4.7810335983585535e-05, + "loss": 0.5785, + "step": 3465 + }, + { + "epoch": 0.13, + "learning_rate": 4.780969479353681e-05, + "loss": 0.6791, + "step": 3466 + }, + { + "epoch": 0.13, + "learning_rate": 4.780905360348807e-05, + "loss": 0.5796, + "step": 3467 + }, + { + "epoch": 0.13, + "learning_rate": 4.7808412413439344e-05, + "loss": 0.5814, + "step": 3468 + }, + { + "epoch": 0.13, + "learning_rate": 4.7807771223390615e-05, + "loss": 0.5134, + "step": 3469 + }, + { + "epoch": 0.13, + "learning_rate": 4.780713003334189e-05, + "loss": 0.5947, + "step": 3470 + }, + { + "epoch": 0.13, + "learning_rate": 4.780648884329316e-05, + "loss": 0.4875, + "step": 3471 + }, + { + "epoch": 0.13, + "learning_rate": 4.7805847653244424e-05, + "loss": 0.6572, + "step": 3472 + }, + { + "epoch": 0.13, + "learning_rate": 4.7805206463195695e-05, + "loss": 0.5536, + "step": 3473 + }, + { + "epoch": 0.13, + "learning_rate": 4.780456527314696e-05, + "loss": 0.5897, + "step": 3474 + }, + { + "epoch": 0.13, + "learning_rate": 4.780392408309823e-05, + "loss": 0.5771, + "step": 3475 + }, + { + "epoch": 0.13, + "learning_rate": 4.78032828930495e-05, + "loss": 0.6344, + "step": 3476 + }, + { + "epoch": 0.13, + "learning_rate": 4.7802641703000776e-05, + "loss": 0.6524, + "step": 3477 + }, + { + "epoch": 0.13, + "learning_rate": 4.780200051295204e-05, + "loss": 0.6201, + "step": 3478 + }, + { + "epoch": 0.13, + "learning_rate": 4.780135932290331e-05, + "loss": 0.5894, + "step": 3479 + }, + { + "epoch": 0.13, + "learning_rate": 4.7800718132854584e-05, + "loss": 0.6245, + "step": 3480 + }, + { + "epoch": 0.13, + "learning_rate": 4.780007694280585e-05, + "loss": 0.5745, + "step": 3481 + }, + { + "epoch": 0.13, + "learning_rate": 4.779943575275712e-05, + "loss": 0.7138, + "step": 3482 + }, + { + "epoch": 0.13, + "learning_rate": 4.7798794562708386e-05, + "loss": 0.5445, + "step": 3483 + }, + { + "epoch": 0.13, + "learning_rate": 4.779815337265966e-05, + "loss": 0.5681, + "step": 3484 + }, + { + "epoch": 0.13, + "learning_rate": 4.779751218261093e-05, + "loss": 0.6588, + "step": 3485 + }, + { + "epoch": 0.13, + "learning_rate": 4.77968709925622e-05, + "loss": 0.6091, + "step": 3486 + }, + { + "epoch": 0.13, + "learning_rate": 4.7796229802513466e-05, + "loss": 0.591, + "step": 3487 + }, + { + "epoch": 0.13, + "learning_rate": 4.779558861246474e-05, + "loss": 0.6504, + "step": 3488 + }, + { + "epoch": 0.13, + "learning_rate": 4.779494742241601e-05, + "loss": 0.5786, + "step": 3489 + }, + { + "epoch": 0.13, + "learning_rate": 4.7794306232367274e-05, + "loss": 0.5509, + "step": 3490 + }, + { + "epoch": 0.13, + "learning_rate": 4.7793665042318546e-05, + "loss": 0.556, + "step": 3491 + }, + { + "epoch": 0.13, + "learning_rate": 4.779302385226981e-05, + "loss": 0.5571, + "step": 3492 + }, + { + "epoch": 0.13, + "learning_rate": 4.779238266222108e-05, + "loss": 0.6531, + "step": 3493 + }, + { + "epoch": 0.13, + "learning_rate": 4.7791741472172354e-05, + "loss": 0.6009, + "step": 3494 + }, + { + "epoch": 0.13, + "learning_rate": 4.7791100282123626e-05, + "loss": 0.6239, + "step": 3495 + }, + { + "epoch": 0.13, + "learning_rate": 4.779045909207489e-05, + "loss": 0.5676, + "step": 3496 + }, + { + "epoch": 0.13, + "learning_rate": 4.778981790202616e-05, + "loss": 0.6384, + "step": 3497 + }, + { + "epoch": 0.13, + "learning_rate": 4.7789176711977434e-05, + "loss": 0.5087, + "step": 3498 + }, + { + "epoch": 0.13, + "learning_rate": 4.77885355219287e-05, + "loss": 0.5241, + "step": 3499 + }, + { + "epoch": 0.13, + "learning_rate": 4.778789433187997e-05, + "loss": 0.6632, + "step": 3500 + }, + { + "epoch": 0.13, + "learning_rate": 4.778725314183124e-05, + "loss": 0.605, + "step": 3501 + }, + { + "epoch": 0.13, + "learning_rate": 4.7786611951782514e-05, + "loss": 0.5453, + "step": 3502 + }, + { + "epoch": 0.13, + "learning_rate": 4.778597076173378e-05, + "loss": 0.5842, + "step": 3503 + }, + { + "epoch": 0.13, + "learning_rate": 4.778532957168505e-05, + "loss": 0.6038, + "step": 3504 + }, + { + "epoch": 0.13, + "learning_rate": 4.7784688381636316e-05, + "loss": 0.636, + "step": 3505 + }, + { + "epoch": 0.13, + "learning_rate": 4.778404719158759e-05, + "loss": 0.6276, + "step": 3506 + }, + { + "epoch": 0.13, + "learning_rate": 4.778340600153886e-05, + "loss": 0.5201, + "step": 3507 + }, + { + "epoch": 0.13, + "learning_rate": 4.7782764811490124e-05, + "loss": 0.6036, + "step": 3508 + }, + { + "epoch": 0.13, + "learning_rate": 4.7782123621441396e-05, + "loss": 0.6309, + "step": 3509 + }, + { + "epoch": 0.13, + "learning_rate": 4.778148243139267e-05, + "loss": 0.5521, + "step": 3510 + }, + { + "epoch": 0.13, + "learning_rate": 4.778084124134394e-05, + "loss": 0.5682, + "step": 3511 + }, + { + "epoch": 0.14, + "learning_rate": 4.7780200051295204e-05, + "loss": 0.5951, + "step": 3512 + }, + { + "epoch": 0.14, + "learning_rate": 4.7779558861246476e-05, + "loss": 0.5931, + "step": 3513 + }, + { + "epoch": 0.14, + "learning_rate": 4.777891767119774e-05, + "loss": 0.6143, + "step": 3514 + }, + { + "epoch": 0.14, + "learning_rate": 4.777827648114901e-05, + "loss": 0.6134, + "step": 3515 + }, + { + "epoch": 0.14, + "learning_rate": 4.7777635291100284e-05, + "loss": 0.5735, + "step": 3516 + }, + { + "epoch": 0.14, + "learning_rate": 4.7776994101051556e-05, + "loss": 0.5178, + "step": 3517 + }, + { + "epoch": 0.14, + "learning_rate": 4.777635291100283e-05, + "loss": 0.5869, + "step": 3518 + }, + { + "epoch": 0.14, + "learning_rate": 4.777571172095409e-05, + "loss": 0.5633, + "step": 3519 + }, + { + "epoch": 0.14, + "learning_rate": 4.7775070530905365e-05, + "loss": 0.6178, + "step": 3520 + }, + { + "epoch": 0.14, + "learning_rate": 4.777442934085663e-05, + "loss": 0.652, + "step": 3521 + }, + { + "epoch": 0.14, + "learning_rate": 4.77737881508079e-05, + "loss": 0.539, + "step": 3522 + }, + { + "epoch": 0.14, + "learning_rate": 4.7773146960759166e-05, + "loss": 0.5502, + "step": 3523 + }, + { + "epoch": 0.14, + "learning_rate": 4.777250577071044e-05, + "loss": 0.6021, + "step": 3524 + }, + { + "epoch": 0.14, + "learning_rate": 4.777186458066171e-05, + "loss": 0.7079, + "step": 3525 + }, + { + "epoch": 0.14, + "learning_rate": 4.777122339061298e-05, + "loss": 0.6137, + "step": 3526 + }, + { + "epoch": 0.14, + "learning_rate": 4.777058220056425e-05, + "loss": 0.6351, + "step": 3527 + }, + { + "epoch": 0.14, + "learning_rate": 4.776994101051552e-05, + "loss": 0.5177, + "step": 3528 + }, + { + "epoch": 0.14, + "learning_rate": 4.776929982046679e-05, + "loss": 0.6541, + "step": 3529 + }, + { + "epoch": 0.14, + "learning_rate": 4.7768658630418055e-05, + "loss": 0.6157, + "step": 3530 + }, + { + "epoch": 0.14, + "learning_rate": 4.7768017440369326e-05, + "loss": 0.7362, + "step": 3531 + }, + { + "epoch": 0.14, + "learning_rate": 4.776737625032059e-05, + "loss": 0.6922, + "step": 3532 + }, + { + "epoch": 0.14, + "learning_rate": 4.776673506027187e-05, + "loss": 0.6382, + "step": 3533 + }, + { + "epoch": 0.14, + "learning_rate": 4.776609387022314e-05, + "loss": 0.58, + "step": 3534 + }, + { + "epoch": 0.14, + "learning_rate": 4.7765452680174406e-05, + "loss": 0.7287, + "step": 3535 + }, + { + "epoch": 0.14, + "learning_rate": 4.776481149012568e-05, + "loss": 0.598, + "step": 3536 + }, + { + "epoch": 0.14, + "learning_rate": 4.776417030007694e-05, + "loss": 0.6132, + "step": 3537 + }, + { + "epoch": 0.14, + "learning_rate": 4.7763529110028215e-05, + "loss": 0.6001, + "step": 3538 + }, + { + "epoch": 0.14, + "learning_rate": 4.776288791997948e-05, + "loss": 0.5689, + "step": 3539 + }, + { + "epoch": 0.14, + "learning_rate": 4.776224672993075e-05, + "loss": 0.5662, + "step": 3540 + }, + { + "epoch": 0.14, + "learning_rate": 4.776160553988202e-05, + "loss": 0.5593, + "step": 3541 + }, + { + "epoch": 0.14, + "learning_rate": 4.7760964349833295e-05, + "loss": 0.6265, + "step": 3542 + }, + { + "epoch": 0.14, + "learning_rate": 4.7760323159784567e-05, + "loss": 0.6231, + "step": 3543 + }, + { + "epoch": 0.14, + "learning_rate": 4.775968196973583e-05, + "loss": 0.6123, + "step": 3544 + }, + { + "epoch": 0.14, + "learning_rate": 4.77590407796871e-05, + "loss": 0.7177, + "step": 3545 + }, + { + "epoch": 0.14, + "learning_rate": 4.775839958963837e-05, + "loss": 0.5221, + "step": 3546 + }, + { + "epoch": 0.14, + "learning_rate": 4.775775839958964e-05, + "loss": 0.5669, + "step": 3547 + }, + { + "epoch": 0.14, + "learning_rate": 4.7757117209540905e-05, + "loss": 0.6125, + "step": 3548 + }, + { + "epoch": 0.14, + "learning_rate": 4.775647601949218e-05, + "loss": 0.6091, + "step": 3549 + }, + { + "epoch": 0.14, + "learning_rate": 4.775583482944345e-05, + "loss": 0.6057, + "step": 3550 + }, + { + "epoch": 0.14, + "learning_rate": 4.775519363939472e-05, + "loss": 0.5907, + "step": 3551 + }, + { + "epoch": 0.14, + "learning_rate": 4.775455244934599e-05, + "loss": 0.5864, + "step": 3552 + }, + { + "epoch": 0.14, + "learning_rate": 4.775391125929726e-05, + "loss": 0.6337, + "step": 3553 + }, + { + "epoch": 0.14, + "learning_rate": 4.775327006924853e-05, + "loss": 0.5737, + "step": 3554 + }, + { + "epoch": 0.14, + "learning_rate": 4.775262887919979e-05, + "loss": 0.5324, + "step": 3555 + }, + { + "epoch": 0.14, + "learning_rate": 4.7751987689151065e-05, + "loss": 0.4973, + "step": 3556 + }, + { + "epoch": 0.14, + "learning_rate": 4.775134649910234e-05, + "loss": 0.6675, + "step": 3557 + }, + { + "epoch": 0.14, + "learning_rate": 4.775070530905361e-05, + "loss": 0.6196, + "step": 3558 + }, + { + "epoch": 0.14, + "learning_rate": 4.775006411900488e-05, + "loss": 0.5964, + "step": 3559 + }, + { + "epoch": 0.14, + "learning_rate": 4.7749422928956145e-05, + "loss": 0.6583, + "step": 3560 + }, + { + "epoch": 0.14, + "learning_rate": 4.774878173890742e-05, + "loss": 0.594, + "step": 3561 + }, + { + "epoch": 0.14, + "learning_rate": 4.774814054885868e-05, + "loss": 0.5299, + "step": 3562 + }, + { + "epoch": 0.14, + "learning_rate": 4.7747499358809953e-05, + "loss": 0.8001, + "step": 3563 + }, + { + "epoch": 0.14, + "learning_rate": 4.774685816876122e-05, + "loss": 0.5688, + "step": 3564 + }, + { + "epoch": 0.14, + "learning_rate": 4.77462169787125e-05, + "loss": 0.6908, + "step": 3565 + }, + { + "epoch": 0.14, + "learning_rate": 4.774557578866376e-05, + "loss": 0.5371, + "step": 3566 + }, + { + "epoch": 0.14, + "learning_rate": 4.7744934598615034e-05, + "loss": 0.6379, + "step": 3567 + }, + { + "epoch": 0.14, + "learning_rate": 4.7744293408566305e-05, + "loss": 0.6161, + "step": 3568 + }, + { + "epoch": 0.14, + "learning_rate": 4.774365221851757e-05, + "loss": 0.5356, + "step": 3569 + }, + { + "epoch": 0.14, + "learning_rate": 4.774301102846884e-05, + "loss": 0.6512, + "step": 3570 + }, + { + "epoch": 0.14, + "learning_rate": 4.774236983842011e-05, + "loss": 0.6013, + "step": 3571 + }, + { + "epoch": 0.14, + "learning_rate": 4.774172864837138e-05, + "loss": 0.6948, + "step": 3572 + }, + { + "epoch": 0.14, + "learning_rate": 4.774108745832265e-05, + "loss": 0.5854, + "step": 3573 + }, + { + "epoch": 0.14, + "learning_rate": 4.774044626827392e-05, + "loss": 0.6865, + "step": 3574 + }, + { + "epoch": 0.14, + "learning_rate": 4.773980507822519e-05, + "loss": 0.6222, + "step": 3575 + }, + { + "epoch": 0.14, + "learning_rate": 4.773916388817646e-05, + "loss": 0.5871, + "step": 3576 + }, + { + "epoch": 0.14, + "learning_rate": 4.773852269812773e-05, + "loss": 0.6247, + "step": 3577 + }, + { + "epoch": 0.14, + "learning_rate": 4.7737881508078995e-05, + "loss": 0.621, + "step": 3578 + }, + { + "epoch": 0.14, + "learning_rate": 4.773724031803027e-05, + "loss": 0.5371, + "step": 3579 + }, + { + "epoch": 0.14, + "learning_rate": 4.773659912798153e-05, + "loss": 0.5661, + "step": 3580 + }, + { + "epoch": 0.14, + "learning_rate": 4.7735957937932804e-05, + "loss": 0.6779, + "step": 3581 + }, + { + "epoch": 0.14, + "learning_rate": 4.7735316747884075e-05, + "loss": 0.5849, + "step": 3582 + }, + { + "epoch": 0.14, + "learning_rate": 4.773467555783535e-05, + "loss": 0.6428, + "step": 3583 + }, + { + "epoch": 0.14, + "learning_rate": 4.773403436778661e-05, + "loss": 0.6095, + "step": 3584 + }, + { + "epoch": 0.14, + "learning_rate": 4.7733393177737884e-05, + "loss": 0.591, + "step": 3585 + }, + { + "epoch": 0.14, + "learning_rate": 4.7732751987689156e-05, + "loss": 0.5325, + "step": 3586 + }, + { + "epoch": 0.14, + "learning_rate": 4.773211079764042e-05, + "loss": 0.5921, + "step": 3587 + }, + { + "epoch": 0.14, + "learning_rate": 4.773146960759169e-05, + "loss": 0.6079, + "step": 3588 + }, + { + "epoch": 0.14, + "learning_rate": 4.7730828417542964e-05, + "loss": 0.5376, + "step": 3589 + }, + { + "epoch": 0.14, + "learning_rate": 4.7730187227494236e-05, + "loss": 0.6749, + "step": 3590 + }, + { + "epoch": 0.14, + "learning_rate": 4.77295460374455e-05, + "loss": 0.5283, + "step": 3591 + }, + { + "epoch": 0.14, + "learning_rate": 4.772890484739677e-05, + "loss": 0.5894, + "step": 3592 + }, + { + "epoch": 0.14, + "learning_rate": 4.772826365734804e-05, + "loss": 0.6036, + "step": 3593 + }, + { + "epoch": 0.14, + "learning_rate": 4.772762246729931e-05, + "loss": 0.6062, + "step": 3594 + }, + { + "epoch": 0.14, + "learning_rate": 4.772698127725058e-05, + "loss": 0.6397, + "step": 3595 + }, + { + "epoch": 0.14, + "learning_rate": 4.7726340087201846e-05, + "loss": 0.6739, + "step": 3596 + }, + { + "epoch": 0.14, + "learning_rate": 4.772569889715312e-05, + "loss": 0.6572, + "step": 3597 + }, + { + "epoch": 0.14, + "learning_rate": 4.772505770710439e-05, + "loss": 0.5768, + "step": 3598 + }, + { + "epoch": 0.14, + "learning_rate": 4.772441651705566e-05, + "loss": 0.6443, + "step": 3599 + }, + { + "epoch": 0.14, + "learning_rate": 4.7723775327006926e-05, + "loss": 0.6456, + "step": 3600 + }, + { + "epoch": 0.14, + "learning_rate": 4.77231341369582e-05, + "loss": 0.5903, + "step": 3601 + }, + { + "epoch": 0.14, + "learning_rate": 4.772249294690946e-05, + "loss": 0.6364, + "step": 3602 + }, + { + "epoch": 0.14, + "learning_rate": 4.7721851756860734e-05, + "loss": 0.6835, + "step": 3603 + }, + { + "epoch": 0.14, + "learning_rate": 4.7721210566812006e-05, + "loss": 0.5237, + "step": 3604 + }, + { + "epoch": 0.14, + "learning_rate": 4.772056937676328e-05, + "loss": 0.5233, + "step": 3605 + }, + { + "epoch": 0.14, + "learning_rate": 4.771992818671455e-05, + "loss": 0.6058, + "step": 3606 + }, + { + "epoch": 0.14, + "learning_rate": 4.7719286996665814e-05, + "loss": 0.735, + "step": 3607 + }, + { + "epoch": 0.14, + "learning_rate": 4.7718645806617086e-05, + "loss": 0.6085, + "step": 3608 + }, + { + "epoch": 0.14, + "learning_rate": 4.771800461656835e-05, + "loss": 0.5757, + "step": 3609 + }, + { + "epoch": 0.14, + "learning_rate": 4.771736342651962e-05, + "loss": 0.5785, + "step": 3610 + }, + { + "epoch": 0.14, + "learning_rate": 4.771672223647089e-05, + "loss": 0.5943, + "step": 3611 + }, + { + "epoch": 0.14, + "learning_rate": 4.771608104642216e-05, + "loss": 0.5703, + "step": 3612 + }, + { + "epoch": 0.14, + "learning_rate": 4.771543985637343e-05, + "loss": 0.638, + "step": 3613 + }, + { + "epoch": 0.14, + "learning_rate": 4.77147986663247e-05, + "loss": 0.5821, + "step": 3614 + }, + { + "epoch": 0.14, + "learning_rate": 4.7714157476275974e-05, + "loss": 0.6589, + "step": 3615 + }, + { + "epoch": 0.14, + "learning_rate": 4.771351628622724e-05, + "loss": 0.7318, + "step": 3616 + }, + { + "epoch": 0.14, + "learning_rate": 4.771287509617851e-05, + "loss": 0.6087, + "step": 3617 + }, + { + "epoch": 0.14, + "learning_rate": 4.7712233906129776e-05, + "loss": 0.6587, + "step": 3618 + }, + { + "epoch": 0.14, + "learning_rate": 4.771159271608105e-05, + "loss": 0.6079, + "step": 3619 + }, + { + "epoch": 0.14, + "learning_rate": 4.771095152603231e-05, + "loss": 0.5609, + "step": 3620 + }, + { + "epoch": 0.14, + "learning_rate": 4.771031033598359e-05, + "loss": 0.5956, + "step": 3621 + }, + { + "epoch": 0.14, + "learning_rate": 4.770966914593486e-05, + "loss": 0.5719, + "step": 3622 + }, + { + "epoch": 0.14, + "learning_rate": 4.770902795588613e-05, + "loss": 0.6246, + "step": 3623 + }, + { + "epoch": 0.14, + "learning_rate": 4.77083867658374e-05, + "loss": 0.6679, + "step": 3624 + }, + { + "epoch": 0.14, + "learning_rate": 4.7707745575788664e-05, + "loss": 0.5433, + "step": 3625 + }, + { + "epoch": 0.14, + "learning_rate": 4.7707104385739936e-05, + "loss": 0.6139, + "step": 3626 + }, + { + "epoch": 0.14, + "learning_rate": 4.77064631956912e-05, + "loss": 0.6552, + "step": 3627 + }, + { + "epoch": 0.14, + "learning_rate": 4.770582200564247e-05, + "loss": 0.5621, + "step": 3628 + }, + { + "epoch": 0.14, + "learning_rate": 4.7705180815593744e-05, + "loss": 0.5299, + "step": 3629 + }, + { + "epoch": 0.14, + "learning_rate": 4.7704539625545016e-05, + "loss": 0.6787, + "step": 3630 + }, + { + "epoch": 0.14, + "learning_rate": 4.770389843549629e-05, + "loss": 0.5948, + "step": 3631 + }, + { + "epoch": 0.14, + "learning_rate": 4.770325724544755e-05, + "loss": 0.6081, + "step": 3632 + }, + { + "epoch": 0.14, + "learning_rate": 4.7702616055398825e-05, + "loss": 0.6024, + "step": 3633 + }, + { + "epoch": 0.14, + "learning_rate": 4.770197486535009e-05, + "loss": 0.6333, + "step": 3634 + }, + { + "epoch": 0.14, + "learning_rate": 4.770133367530136e-05, + "loss": 0.681, + "step": 3635 + }, + { + "epoch": 0.14, + "learning_rate": 4.7700692485252626e-05, + "loss": 0.6487, + "step": 3636 + }, + { + "epoch": 0.14, + "learning_rate": 4.7700051295203905e-05, + "loss": 0.5903, + "step": 3637 + }, + { + "epoch": 0.14, + "learning_rate": 4.769941010515517e-05, + "loss": 0.5308, + "step": 3638 + }, + { + "epoch": 0.14, + "learning_rate": 4.769876891510644e-05, + "loss": 0.5722, + "step": 3639 + }, + { + "epoch": 0.14, + "learning_rate": 4.769812772505771e-05, + "loss": 0.5762, + "step": 3640 + }, + { + "epoch": 0.14, + "learning_rate": 4.769748653500898e-05, + "loss": 0.6293, + "step": 3641 + }, + { + "epoch": 0.14, + "learning_rate": 4.769684534496025e-05, + "loss": 0.6073, + "step": 3642 + }, + { + "epoch": 0.14, + "learning_rate": 4.7696204154911515e-05, + "loss": 0.5185, + "step": 3643 + }, + { + "epoch": 0.14, + "learning_rate": 4.7695562964862786e-05, + "loss": 0.5903, + "step": 3644 + }, + { + "epoch": 0.14, + "learning_rate": 4.769492177481406e-05, + "loss": 0.5793, + "step": 3645 + }, + { + "epoch": 0.14, + "learning_rate": 4.769428058476533e-05, + "loss": 0.6735, + "step": 3646 + }, + { + "epoch": 0.14, + "learning_rate": 4.7693639394716595e-05, + "loss": 0.6043, + "step": 3647 + }, + { + "epoch": 0.14, + "learning_rate": 4.7692998204667866e-05, + "loss": 0.5582, + "step": 3648 + }, + { + "epoch": 0.14, + "learning_rate": 4.769235701461914e-05, + "loss": 0.5907, + "step": 3649 + }, + { + "epoch": 0.14, + "learning_rate": 4.76917158245704e-05, + "loss": 0.5973, + "step": 3650 + }, + { + "epoch": 0.14, + "learning_rate": 4.7691074634521675e-05, + "loss": 0.6262, + "step": 3651 + }, + { + "epoch": 0.14, + "learning_rate": 4.769043344447294e-05, + "loss": 0.6133, + "step": 3652 + }, + { + "epoch": 0.14, + "learning_rate": 4.768979225442421e-05, + "loss": 0.5925, + "step": 3653 + }, + { + "epoch": 0.14, + "learning_rate": 4.768915106437548e-05, + "loss": 0.589, + "step": 3654 + }, + { + "epoch": 0.14, + "learning_rate": 4.7688509874326755e-05, + "loss": 0.6423, + "step": 3655 + }, + { + "epoch": 0.14, + "learning_rate": 4.768786868427802e-05, + "loss": 0.5545, + "step": 3656 + }, + { + "epoch": 0.14, + "learning_rate": 4.768722749422929e-05, + "loss": 0.5424, + "step": 3657 + }, + { + "epoch": 0.14, + "learning_rate": 4.768658630418056e-05, + "loss": 0.7075, + "step": 3658 + }, + { + "epoch": 0.14, + "learning_rate": 4.768594511413183e-05, + "loss": 0.5181, + "step": 3659 + }, + { + "epoch": 0.14, + "learning_rate": 4.76853039240831e-05, + "loss": 0.5054, + "step": 3660 + }, + { + "epoch": 0.14, + "learning_rate": 4.768466273403437e-05, + "loss": 0.5796, + "step": 3661 + }, + { + "epoch": 0.14, + "learning_rate": 4.768402154398564e-05, + "loss": 0.5919, + "step": 3662 + }, + { + "epoch": 0.14, + "learning_rate": 4.768338035393691e-05, + "loss": 0.604, + "step": 3663 + }, + { + "epoch": 0.14, + "learning_rate": 4.768273916388818e-05, + "loss": 0.6968, + "step": 3664 + }, + { + "epoch": 0.14, + "learning_rate": 4.7682097973839445e-05, + "loss": 0.5666, + "step": 3665 + }, + { + "epoch": 0.14, + "learning_rate": 4.768145678379072e-05, + "loss": 0.6291, + "step": 3666 + }, + { + "epoch": 0.14, + "learning_rate": 4.768081559374199e-05, + "loss": 0.6046, + "step": 3667 + }, + { + "epoch": 0.14, + "learning_rate": 4.768017440369325e-05, + "loss": 0.5917, + "step": 3668 + }, + { + "epoch": 0.14, + "learning_rate": 4.7679533213644525e-05, + "loss": 0.5944, + "step": 3669 + }, + { + "epoch": 0.14, + "learning_rate": 4.76788920235958e-05, + "loss": 0.576, + "step": 3670 + }, + { + "epoch": 0.14, + "learning_rate": 4.767825083354707e-05, + "loss": 0.5638, + "step": 3671 + }, + { + "epoch": 0.14, + "learning_rate": 4.7677609643498333e-05, + "loss": 0.651, + "step": 3672 + }, + { + "epoch": 0.14, + "learning_rate": 4.7676968453449605e-05, + "loss": 0.6104, + "step": 3673 + }, + { + "epoch": 0.14, + "learning_rate": 4.767632726340087e-05, + "loss": 0.624, + "step": 3674 + }, + { + "epoch": 0.14, + "learning_rate": 4.767568607335214e-05, + "loss": 0.5468, + "step": 3675 + }, + { + "epoch": 0.14, + "learning_rate": 4.7675044883303414e-05, + "loss": 0.6536, + "step": 3676 + }, + { + "epoch": 0.14, + "learning_rate": 4.7674403693254685e-05, + "loss": 0.5978, + "step": 3677 + }, + { + "epoch": 0.14, + "learning_rate": 4.767376250320596e-05, + "loss": 0.6453, + "step": 3678 + }, + { + "epoch": 0.14, + "learning_rate": 4.767312131315722e-05, + "loss": 0.5991, + "step": 3679 + }, + { + "epoch": 0.14, + "learning_rate": 4.7672480123108494e-05, + "loss": 0.5594, + "step": 3680 + }, + { + "epoch": 0.14, + "learning_rate": 4.767183893305976e-05, + "loss": 0.6364, + "step": 3681 + }, + { + "epoch": 0.14, + "learning_rate": 4.767119774301103e-05, + "loss": 0.6152, + "step": 3682 + }, + { + "epoch": 0.14, + "learning_rate": 4.7670556552962295e-05, + "loss": 0.6597, + "step": 3683 + }, + { + "epoch": 0.14, + "learning_rate": 4.766991536291357e-05, + "loss": 0.5756, + "step": 3684 + }, + { + "epoch": 0.14, + "learning_rate": 4.766927417286484e-05, + "loss": 0.561, + "step": 3685 + }, + { + "epoch": 0.14, + "learning_rate": 4.766863298281611e-05, + "loss": 0.5846, + "step": 3686 + }, + { + "epoch": 0.14, + "learning_rate": 4.766799179276738e-05, + "loss": 0.5573, + "step": 3687 + }, + { + "epoch": 0.14, + "learning_rate": 4.766735060271865e-05, + "loss": 0.5545, + "step": 3688 + }, + { + "epoch": 0.14, + "learning_rate": 4.766670941266992e-05, + "loss": 0.6391, + "step": 3689 + }, + { + "epoch": 0.14, + "learning_rate": 4.7666068222621184e-05, + "loss": 0.5885, + "step": 3690 + }, + { + "epoch": 0.14, + "learning_rate": 4.7665427032572455e-05, + "loss": 0.5657, + "step": 3691 + }, + { + "epoch": 0.14, + "learning_rate": 4.766478584252372e-05, + "loss": 0.5892, + "step": 3692 + }, + { + "epoch": 0.14, + "learning_rate": 4.7664144652475e-05, + "loss": 0.5855, + "step": 3693 + }, + { + "epoch": 0.14, + "learning_rate": 4.766350346242627e-05, + "loss": 0.5808, + "step": 3694 + }, + { + "epoch": 0.14, + "learning_rate": 4.7662862272377535e-05, + "loss": 0.5733, + "step": 3695 + }, + { + "epoch": 0.14, + "learning_rate": 4.766222108232881e-05, + "loss": 0.6601, + "step": 3696 + }, + { + "epoch": 0.14, + "learning_rate": 4.766157989228007e-05, + "loss": 0.6468, + "step": 3697 + }, + { + "epoch": 0.14, + "learning_rate": 4.7660938702231344e-05, + "loss": 0.598, + "step": 3698 + }, + { + "epoch": 0.14, + "learning_rate": 4.766029751218261e-05, + "loss": 0.5414, + "step": 3699 + }, + { + "epoch": 0.14, + "learning_rate": 4.765965632213388e-05, + "loss": 0.5842, + "step": 3700 + }, + { + "epoch": 0.14, + "learning_rate": 4.765901513208515e-05, + "loss": 0.5421, + "step": 3701 + }, + { + "epoch": 0.14, + "learning_rate": 4.7658373942036424e-05, + "loss": 0.5685, + "step": 3702 + }, + { + "epoch": 0.14, + "learning_rate": 4.7657732751987696e-05, + "loss": 0.5815, + "step": 3703 + }, + { + "epoch": 0.14, + "learning_rate": 4.765709156193896e-05, + "loss": 0.5701, + "step": 3704 + }, + { + "epoch": 0.14, + "learning_rate": 4.765645037189023e-05, + "loss": 0.6203, + "step": 3705 + }, + { + "epoch": 0.14, + "learning_rate": 4.76558091818415e-05, + "loss": 0.5688, + "step": 3706 + }, + { + "epoch": 0.14, + "learning_rate": 4.765516799179277e-05, + "loss": 0.6002, + "step": 3707 + }, + { + "epoch": 0.14, + "learning_rate": 4.7654526801744034e-05, + "loss": 0.6083, + "step": 3708 + }, + { + "epoch": 0.14, + "learning_rate": 4.765388561169531e-05, + "loss": 0.5702, + "step": 3709 + }, + { + "epoch": 0.14, + "learning_rate": 4.765324442164658e-05, + "loss": 0.6105, + "step": 3710 + }, + { + "epoch": 0.14, + "learning_rate": 4.765260323159785e-05, + "loss": 0.7105, + "step": 3711 + }, + { + "epoch": 0.14, + "learning_rate": 4.765196204154912e-05, + "loss": 0.648, + "step": 3712 + }, + { + "epoch": 0.14, + "learning_rate": 4.7651320851500386e-05, + "loss": 0.5784, + "step": 3713 + }, + { + "epoch": 0.14, + "learning_rate": 4.765067966145166e-05, + "loss": 0.6326, + "step": 3714 + }, + { + "epoch": 0.14, + "learning_rate": 4.765003847140292e-05, + "loss": 0.6485, + "step": 3715 + }, + { + "epoch": 0.14, + "learning_rate": 4.7649397281354194e-05, + "loss": 0.5842, + "step": 3716 + }, + { + "epoch": 0.14, + "learning_rate": 4.7648756091305466e-05, + "loss": 0.5558, + "step": 3717 + }, + { + "epoch": 0.14, + "learning_rate": 4.764811490125674e-05, + "loss": 0.634, + "step": 3718 + }, + { + "epoch": 0.14, + "learning_rate": 4.7647473711208e-05, + "loss": 0.568, + "step": 3719 + }, + { + "epoch": 0.14, + "learning_rate": 4.7646832521159274e-05, + "loss": 0.5813, + "step": 3720 + }, + { + "epoch": 0.14, + "learning_rate": 4.7646191331110546e-05, + "loss": 0.6158, + "step": 3721 + }, + { + "epoch": 0.14, + "learning_rate": 4.764555014106181e-05, + "loss": 0.6346, + "step": 3722 + }, + { + "epoch": 0.14, + "learning_rate": 4.764490895101308e-05, + "loss": 0.6262, + "step": 3723 + }, + { + "epoch": 0.14, + "learning_rate": 4.764426776096435e-05, + "loss": 0.5826, + "step": 3724 + }, + { + "epoch": 0.14, + "learning_rate": 4.7643626570915626e-05, + "loss": 0.6156, + "step": 3725 + }, + { + "epoch": 0.14, + "learning_rate": 4.764298538086689e-05, + "loss": 0.5721, + "step": 3726 + }, + { + "epoch": 0.14, + "learning_rate": 4.764234419081816e-05, + "loss": 0.5835, + "step": 3727 + }, + { + "epoch": 0.14, + "learning_rate": 4.764170300076943e-05, + "loss": 0.6688, + "step": 3728 + }, + { + "epoch": 0.14, + "learning_rate": 4.76410618107207e-05, + "loss": 0.6833, + "step": 3729 + }, + { + "epoch": 0.14, + "learning_rate": 4.764042062067197e-05, + "loss": 0.53, + "step": 3730 + }, + { + "epoch": 0.14, + "learning_rate": 4.7639779430623236e-05, + "loss": 0.6307, + "step": 3731 + }, + { + "epoch": 0.14, + "learning_rate": 4.763913824057451e-05, + "loss": 0.6052, + "step": 3732 + }, + { + "epoch": 0.14, + "learning_rate": 4.763849705052578e-05, + "loss": 0.5855, + "step": 3733 + }, + { + "epoch": 0.14, + "learning_rate": 4.763785586047705e-05, + "loss": 0.5722, + "step": 3734 + }, + { + "epoch": 0.14, + "learning_rate": 4.7637214670428316e-05, + "loss": 0.5885, + "step": 3735 + }, + { + "epoch": 0.14, + "learning_rate": 4.763657348037959e-05, + "loss": 0.7122, + "step": 3736 + }, + { + "epoch": 0.14, + "learning_rate": 4.763593229033085e-05, + "loss": 0.6388, + "step": 3737 + }, + { + "epoch": 0.14, + "learning_rate": 4.7635291100282124e-05, + "loss": 0.5595, + "step": 3738 + }, + { + "epoch": 0.14, + "learning_rate": 4.7634649910233396e-05, + "loss": 0.6562, + "step": 3739 + }, + { + "epoch": 0.14, + "learning_rate": 4.763400872018466e-05, + "loss": 0.5922, + "step": 3740 + }, + { + "epoch": 0.14, + "learning_rate": 4.763336753013593e-05, + "loss": 0.6749, + "step": 3741 + }, + { + "epoch": 0.14, + "learning_rate": 4.7632726340087205e-05, + "loss": 0.5722, + "step": 3742 + }, + { + "epoch": 0.14, + "learning_rate": 4.7632085150038476e-05, + "loss": 0.6957, + "step": 3743 + }, + { + "epoch": 0.14, + "learning_rate": 4.763144395998974e-05, + "loss": 0.613, + "step": 3744 + }, + { + "epoch": 0.14, + "learning_rate": 4.763080276994101e-05, + "loss": 0.6239, + "step": 3745 + }, + { + "epoch": 0.14, + "learning_rate": 4.7630161579892285e-05, + "loss": 0.5591, + "step": 3746 + }, + { + "epoch": 0.14, + "learning_rate": 4.762952038984355e-05, + "loss": 0.624, + "step": 3747 + }, + { + "epoch": 0.14, + "learning_rate": 4.762887919979482e-05, + "loss": 0.6706, + "step": 3748 + }, + { + "epoch": 0.14, + "learning_rate": 4.762823800974609e-05, + "loss": 0.6047, + "step": 3749 + }, + { + "epoch": 0.14, + "learning_rate": 4.7627596819697365e-05, + "loss": 0.6901, + "step": 3750 + }, + { + "epoch": 0.14, + "learning_rate": 4.762695562964863e-05, + "loss": 0.6174, + "step": 3751 + }, + { + "epoch": 0.14, + "learning_rate": 4.76263144395999e-05, + "loss": 0.5557, + "step": 3752 + }, + { + "epoch": 0.14, + "learning_rate": 4.7625673249551166e-05, + "loss": 0.5845, + "step": 3753 + }, + { + "epoch": 0.14, + "learning_rate": 4.762503205950244e-05, + "loss": 0.5389, + "step": 3754 + }, + { + "epoch": 0.14, + "learning_rate": 4.762439086945371e-05, + "loss": 0.6193, + "step": 3755 + }, + { + "epoch": 0.14, + "learning_rate": 4.7623749679404975e-05, + "loss": 0.6478, + "step": 3756 + }, + { + "epoch": 0.14, + "learning_rate": 4.7623108489356246e-05, + "loss": 0.6177, + "step": 3757 + }, + { + "epoch": 0.14, + "learning_rate": 4.762246729930752e-05, + "loss": 0.6073, + "step": 3758 + }, + { + "epoch": 0.14, + "learning_rate": 4.762182610925879e-05, + "loss": 0.5432, + "step": 3759 + }, + { + "epoch": 0.14, + "learning_rate": 4.7621184919210055e-05, + "loss": 0.6274, + "step": 3760 + }, + { + "epoch": 0.14, + "learning_rate": 4.7620543729161326e-05, + "loss": 0.5794, + "step": 3761 + }, + { + "epoch": 0.14, + "learning_rate": 4.761990253911259e-05, + "loss": 0.5988, + "step": 3762 + }, + { + "epoch": 0.14, + "learning_rate": 4.761926134906386e-05, + "loss": 0.5003, + "step": 3763 + }, + { + "epoch": 0.14, + "learning_rate": 4.7618620159015135e-05, + "loss": 0.5524, + "step": 3764 + }, + { + "epoch": 0.14, + "learning_rate": 4.7617978968966407e-05, + "loss": 0.5089, + "step": 3765 + }, + { + "epoch": 0.14, + "learning_rate": 4.761733777891768e-05, + "loss": 0.4822, + "step": 3766 + }, + { + "epoch": 0.14, + "learning_rate": 4.761669658886894e-05, + "loss": 0.5658, + "step": 3767 + }, + { + "epoch": 0.14, + "learning_rate": 4.7616055398820215e-05, + "loss": 0.5958, + "step": 3768 + }, + { + "epoch": 0.14, + "learning_rate": 4.761541420877148e-05, + "loss": 0.5887, + "step": 3769 + }, + { + "epoch": 0.14, + "learning_rate": 4.761477301872275e-05, + "loss": 0.6046, + "step": 3770 + }, + { + "epoch": 0.14, + "learning_rate": 4.7614131828674017e-05, + "loss": 0.5866, + "step": 3771 + }, + { + "epoch": 0.15, + "learning_rate": 4.761349063862529e-05, + "loss": 0.6789, + "step": 3772 + }, + { + "epoch": 0.15, + "learning_rate": 4.761284944857656e-05, + "loss": 0.5824, + "step": 3773 + }, + { + "epoch": 0.15, + "learning_rate": 4.761220825852783e-05, + "loss": 0.627, + "step": 3774 + }, + { + "epoch": 0.15, + "learning_rate": 4.7611567068479103e-05, + "loss": 0.6081, + "step": 3775 + }, + { + "epoch": 0.15, + "learning_rate": 4.761092587843037e-05, + "loss": 0.5903, + "step": 3776 + }, + { + "epoch": 0.15, + "learning_rate": 4.761028468838164e-05, + "loss": 0.4826, + "step": 3777 + }, + { + "epoch": 0.15, + "learning_rate": 4.7609643498332905e-05, + "loss": 0.5433, + "step": 3778 + }, + { + "epoch": 0.15, + "learning_rate": 4.760900230828418e-05, + "loss": 0.6252, + "step": 3779 + }, + { + "epoch": 0.15, + "learning_rate": 4.760836111823544e-05, + "loss": 0.818, + "step": 3780 + }, + { + "epoch": 0.15, + "learning_rate": 4.760771992818672e-05, + "loss": 0.5378, + "step": 3781 + }, + { + "epoch": 0.15, + "learning_rate": 4.760707873813799e-05, + "loss": 0.555, + "step": 3782 + }, + { + "epoch": 0.15, + "learning_rate": 4.760643754808926e-05, + "loss": 0.6524, + "step": 3783 + }, + { + "epoch": 0.15, + "learning_rate": 4.760579635804053e-05, + "loss": 0.5999, + "step": 3784 + }, + { + "epoch": 0.15, + "learning_rate": 4.7605155167991793e-05, + "loss": 0.5832, + "step": 3785 + }, + { + "epoch": 0.15, + "learning_rate": 4.7604513977943065e-05, + "loss": 0.5493, + "step": 3786 + }, + { + "epoch": 0.15, + "learning_rate": 4.760387278789433e-05, + "loss": 0.6089, + "step": 3787 + }, + { + "epoch": 0.15, + "learning_rate": 4.76032315978456e-05, + "loss": 0.6015, + "step": 3788 + }, + { + "epoch": 0.15, + "learning_rate": 4.7602590407796874e-05, + "loss": 0.6493, + "step": 3789 + }, + { + "epoch": 0.15, + "learning_rate": 4.7601949217748145e-05, + "loss": 0.612, + "step": 3790 + }, + { + "epoch": 0.15, + "learning_rate": 4.760130802769942e-05, + "loss": 0.7069, + "step": 3791 + }, + { + "epoch": 0.15, + "learning_rate": 4.760066683765068e-05, + "loss": 0.5458, + "step": 3792 + }, + { + "epoch": 0.15, + "learning_rate": 4.7600025647601954e-05, + "loss": 0.6337, + "step": 3793 + }, + { + "epoch": 0.15, + "learning_rate": 4.759938445755322e-05, + "loss": 0.5398, + "step": 3794 + }, + { + "epoch": 0.15, + "learning_rate": 4.759874326750449e-05, + "loss": 0.6164, + "step": 3795 + }, + { + "epoch": 0.15, + "learning_rate": 4.7598102077455755e-05, + "loss": 0.4931, + "step": 3796 + }, + { + "epoch": 0.15, + "learning_rate": 4.7597460887407034e-05, + "loss": 0.6449, + "step": 3797 + }, + { + "epoch": 0.15, + "learning_rate": 4.75968196973583e-05, + "loss": 0.5984, + "step": 3798 + }, + { + "epoch": 0.15, + "learning_rate": 4.759617850730957e-05, + "loss": 0.6085, + "step": 3799 + }, + { + "epoch": 0.15, + "learning_rate": 4.759553731726084e-05, + "loss": 0.5992, + "step": 3800 + }, + { + "epoch": 0.15, + "learning_rate": 4.759489612721211e-05, + "loss": 0.6327, + "step": 3801 + }, + { + "epoch": 0.15, + "learning_rate": 4.759425493716338e-05, + "loss": 0.6234, + "step": 3802 + }, + { + "epoch": 0.15, + "learning_rate": 4.7593613747114644e-05, + "loss": 0.5556, + "step": 3803 + }, + { + "epoch": 0.15, + "learning_rate": 4.7592972557065915e-05, + "loss": 0.5112, + "step": 3804 + }, + { + "epoch": 0.15, + "learning_rate": 4.759233136701719e-05, + "loss": 0.5835, + "step": 3805 + }, + { + "epoch": 0.15, + "learning_rate": 4.759169017696846e-05, + "loss": 0.6742, + "step": 3806 + }, + { + "epoch": 0.15, + "learning_rate": 4.7591048986919724e-05, + "loss": 0.5493, + "step": 3807 + }, + { + "epoch": 0.15, + "learning_rate": 4.7590407796870996e-05, + "loss": 0.6312, + "step": 3808 + }, + { + "epoch": 0.15, + "learning_rate": 4.758976660682227e-05, + "loss": 0.6738, + "step": 3809 + }, + { + "epoch": 0.15, + "learning_rate": 4.758912541677353e-05, + "loss": 0.6913, + "step": 3810 + }, + { + "epoch": 0.15, + "learning_rate": 4.7588484226724804e-05, + "loss": 0.639, + "step": 3811 + }, + { + "epoch": 0.15, + "learning_rate": 4.758784303667607e-05, + "loss": 0.6251, + "step": 3812 + }, + { + "epoch": 0.15, + "learning_rate": 4.758720184662735e-05, + "loss": 0.7062, + "step": 3813 + }, + { + "epoch": 0.15, + "learning_rate": 4.758656065657861e-05, + "loss": 0.65, + "step": 3814 + }, + { + "epoch": 0.15, + "learning_rate": 4.7585919466529884e-05, + "loss": 0.5746, + "step": 3815 + }, + { + "epoch": 0.15, + "learning_rate": 4.758527827648115e-05, + "loss": 0.6088, + "step": 3816 + }, + { + "epoch": 0.15, + "learning_rate": 4.758463708643242e-05, + "loss": 0.6471, + "step": 3817 + }, + { + "epoch": 0.15, + "learning_rate": 4.758399589638369e-05, + "loss": 0.5595, + "step": 3818 + }, + { + "epoch": 0.15, + "learning_rate": 4.758335470633496e-05, + "loss": 0.579, + "step": 3819 + }, + { + "epoch": 0.15, + "learning_rate": 4.758271351628623e-05, + "loss": 0.6898, + "step": 3820 + }, + { + "epoch": 0.15, + "learning_rate": 4.75820723262375e-05, + "loss": 0.5021, + "step": 3821 + }, + { + "epoch": 0.15, + "learning_rate": 4.758143113618877e-05, + "loss": 0.6629, + "step": 3822 + }, + { + "epoch": 0.15, + "learning_rate": 4.758078994614004e-05, + "loss": 0.6832, + "step": 3823 + }, + { + "epoch": 0.15, + "learning_rate": 4.758014875609131e-05, + "loss": 0.5242, + "step": 3824 + }, + { + "epoch": 0.15, + "learning_rate": 4.7579507566042574e-05, + "loss": 0.7327, + "step": 3825 + }, + { + "epoch": 0.15, + "learning_rate": 4.7578866375993846e-05, + "loss": 0.6846, + "step": 3826 + }, + { + "epoch": 0.15, + "learning_rate": 4.757822518594512e-05, + "loss": 0.6417, + "step": 3827 + }, + { + "epoch": 0.15, + "learning_rate": 4.757758399589638e-05, + "loss": 0.6458, + "step": 3828 + }, + { + "epoch": 0.15, + "learning_rate": 4.7576942805847654e-05, + "loss": 0.5937, + "step": 3829 + }, + { + "epoch": 0.15, + "learning_rate": 4.7576301615798926e-05, + "loss": 0.6296, + "step": 3830 + }, + { + "epoch": 0.15, + "learning_rate": 4.75756604257502e-05, + "loss": 0.5764, + "step": 3831 + }, + { + "epoch": 0.15, + "learning_rate": 4.757501923570146e-05, + "loss": 0.5974, + "step": 3832 + }, + { + "epoch": 0.15, + "learning_rate": 4.7574378045652734e-05, + "loss": 0.6011, + "step": 3833 + }, + { + "epoch": 0.15, + "learning_rate": 4.7573736855604e-05, + "loss": 0.535, + "step": 3834 + }, + { + "epoch": 0.15, + "learning_rate": 4.757309566555527e-05, + "loss": 0.6232, + "step": 3835 + }, + { + "epoch": 0.15, + "learning_rate": 4.757245447550654e-05, + "loss": 0.6218, + "step": 3836 + }, + { + "epoch": 0.15, + "learning_rate": 4.7571813285457814e-05, + "loss": 0.6287, + "step": 3837 + }, + { + "epoch": 0.15, + "learning_rate": 4.7571172095409086e-05, + "loss": 0.6297, + "step": 3838 + }, + { + "epoch": 0.15, + "learning_rate": 4.757053090536035e-05, + "loss": 0.6171, + "step": 3839 + }, + { + "epoch": 0.15, + "learning_rate": 4.756988971531162e-05, + "loss": 0.6114, + "step": 3840 + }, + { + "epoch": 0.15, + "learning_rate": 4.756924852526289e-05, + "loss": 0.6296, + "step": 3841 + }, + { + "epoch": 0.15, + "learning_rate": 4.756860733521416e-05, + "loss": 0.5737, + "step": 3842 + }, + { + "epoch": 0.15, + "learning_rate": 4.7567966145165424e-05, + "loss": 0.5536, + "step": 3843 + }, + { + "epoch": 0.15, + "learning_rate": 4.7567324955116696e-05, + "loss": 0.6645, + "step": 3844 + }, + { + "epoch": 0.15, + "learning_rate": 4.756668376506797e-05, + "loss": 0.7116, + "step": 3845 + }, + { + "epoch": 0.15, + "learning_rate": 4.756604257501924e-05, + "loss": 0.5691, + "step": 3846 + }, + { + "epoch": 0.15, + "learning_rate": 4.756540138497051e-05, + "loss": 0.6229, + "step": 3847 + }, + { + "epoch": 0.15, + "learning_rate": 4.7564760194921776e-05, + "loss": 0.645, + "step": 3848 + }, + { + "epoch": 0.15, + "learning_rate": 4.756411900487305e-05, + "loss": 0.5911, + "step": 3849 + }, + { + "epoch": 0.15, + "learning_rate": 4.756347781482431e-05, + "loss": 0.5432, + "step": 3850 + }, + { + "epoch": 0.15, + "learning_rate": 4.7562836624775584e-05, + "loss": 0.6063, + "step": 3851 + }, + { + "epoch": 0.15, + "learning_rate": 4.7562195434726856e-05, + "loss": 0.6019, + "step": 3852 + }, + { + "epoch": 0.15, + "learning_rate": 4.756155424467813e-05, + "loss": 0.5553, + "step": 3853 + }, + { + "epoch": 0.15, + "learning_rate": 4.75609130546294e-05, + "loss": 0.5842, + "step": 3854 + }, + { + "epoch": 0.15, + "learning_rate": 4.7560271864580665e-05, + "loss": 0.5967, + "step": 3855 + }, + { + "epoch": 0.15, + "learning_rate": 4.7559630674531936e-05, + "loss": 0.5497, + "step": 3856 + }, + { + "epoch": 0.15, + "learning_rate": 4.75589894844832e-05, + "loss": 0.618, + "step": 3857 + }, + { + "epoch": 0.15, + "learning_rate": 4.755834829443447e-05, + "loss": 0.554, + "step": 3858 + }, + { + "epoch": 0.15, + "learning_rate": 4.755770710438574e-05, + "loss": 0.6885, + "step": 3859 + }, + { + "epoch": 0.15, + "learning_rate": 4.755706591433701e-05, + "loss": 0.5567, + "step": 3860 + }, + { + "epoch": 0.15, + "learning_rate": 4.755642472428828e-05, + "loss": 0.5831, + "step": 3861 + }, + { + "epoch": 0.15, + "learning_rate": 4.755578353423955e-05, + "loss": 0.6449, + "step": 3862 + }, + { + "epoch": 0.15, + "learning_rate": 4.7555142344190825e-05, + "loss": 0.591, + "step": 3863 + }, + { + "epoch": 0.15, + "learning_rate": 4.755450115414209e-05, + "loss": 0.5751, + "step": 3864 + }, + { + "epoch": 0.15, + "learning_rate": 4.755385996409336e-05, + "loss": 0.6159, + "step": 3865 + }, + { + "epoch": 0.15, + "learning_rate": 4.7553218774044626e-05, + "loss": 0.704, + "step": 3866 + }, + { + "epoch": 0.15, + "learning_rate": 4.75525775839959e-05, + "loss": 0.5487, + "step": 3867 + }, + { + "epoch": 0.15, + "learning_rate": 4.755193639394716e-05, + "loss": 0.6087, + "step": 3868 + }, + { + "epoch": 0.15, + "learning_rate": 4.755129520389844e-05, + "loss": 0.6618, + "step": 3869 + }, + { + "epoch": 0.15, + "learning_rate": 4.7550654013849706e-05, + "loss": 0.6749, + "step": 3870 + }, + { + "epoch": 0.15, + "learning_rate": 4.755001282380098e-05, + "loss": 0.6147, + "step": 3871 + }, + { + "epoch": 0.15, + "learning_rate": 4.754937163375225e-05, + "loss": 0.5972, + "step": 3872 + }, + { + "epoch": 0.15, + "learning_rate": 4.7548730443703515e-05, + "loss": 0.6098, + "step": 3873 + }, + { + "epoch": 0.15, + "learning_rate": 4.7548089253654786e-05, + "loss": 0.6426, + "step": 3874 + }, + { + "epoch": 0.15, + "learning_rate": 4.754744806360605e-05, + "loss": 0.719, + "step": 3875 + }, + { + "epoch": 0.15, + "learning_rate": 4.754680687355732e-05, + "loss": 0.6011, + "step": 3876 + }, + { + "epoch": 0.15, + "learning_rate": 4.7546165683508595e-05, + "loss": 0.5558, + "step": 3877 + }, + { + "epoch": 0.15, + "learning_rate": 4.7545524493459867e-05, + "loss": 0.6238, + "step": 3878 + }, + { + "epoch": 0.15, + "learning_rate": 4.754488330341113e-05, + "loss": 0.5331, + "step": 3879 + }, + { + "epoch": 0.15, + "learning_rate": 4.75442421133624e-05, + "loss": 0.665, + "step": 3880 + }, + { + "epoch": 0.15, + "learning_rate": 4.7543600923313675e-05, + "loss": 0.588, + "step": 3881 + }, + { + "epoch": 0.15, + "learning_rate": 4.754295973326494e-05, + "loss": 0.5468, + "step": 3882 + }, + { + "epoch": 0.15, + "learning_rate": 4.754231854321621e-05, + "loss": 0.602, + "step": 3883 + }, + { + "epoch": 0.15, + "learning_rate": 4.7541677353167477e-05, + "loss": 0.6038, + "step": 3884 + }, + { + "epoch": 0.15, + "learning_rate": 4.7541036163118755e-05, + "loss": 0.5117, + "step": 3885 + }, + { + "epoch": 0.15, + "learning_rate": 4.754039497307002e-05, + "loss": 0.6303, + "step": 3886 + }, + { + "epoch": 0.15, + "learning_rate": 4.753975378302129e-05, + "loss": 0.6252, + "step": 3887 + }, + { + "epoch": 0.15, + "learning_rate": 4.753911259297256e-05, + "loss": 0.5344, + "step": 3888 + }, + { + "epoch": 0.15, + "learning_rate": 4.753847140292383e-05, + "loss": 0.6011, + "step": 3889 + }, + { + "epoch": 0.15, + "learning_rate": 4.75378302128751e-05, + "loss": 0.5378, + "step": 3890 + }, + { + "epoch": 0.15, + "learning_rate": 4.7537189022826365e-05, + "loss": 0.6966, + "step": 3891 + }, + { + "epoch": 0.15, + "learning_rate": 4.753654783277764e-05, + "loss": 0.6463, + "step": 3892 + }, + { + "epoch": 0.15, + "learning_rate": 4.753590664272891e-05, + "loss": 0.6041, + "step": 3893 + }, + { + "epoch": 0.15, + "learning_rate": 4.753526545268018e-05, + "loss": 0.5829, + "step": 3894 + }, + { + "epoch": 0.15, + "learning_rate": 4.7534624262631445e-05, + "loss": 0.59, + "step": 3895 + }, + { + "epoch": 0.15, + "learning_rate": 4.753398307258272e-05, + "loss": 0.5822, + "step": 3896 + }, + { + "epoch": 0.15, + "learning_rate": 4.753334188253398e-05, + "loss": 0.5913, + "step": 3897 + }, + { + "epoch": 0.15, + "learning_rate": 4.7532700692485253e-05, + "loss": 0.6078, + "step": 3898 + }, + { + "epoch": 0.15, + "learning_rate": 4.7532059502436525e-05, + "loss": 0.6246, + "step": 3899 + }, + { + "epoch": 0.15, + "learning_rate": 4.753141831238779e-05, + "loss": 0.6293, + "step": 3900 + }, + { + "epoch": 0.15, + "learning_rate": 4.753077712233906e-05, + "loss": 0.6673, + "step": 3901 + }, + { + "epoch": 0.15, + "learning_rate": 4.7530135932290334e-05, + "loss": 0.711, + "step": 3902 + }, + { + "epoch": 0.15, + "learning_rate": 4.7529494742241605e-05, + "loss": 0.5969, + "step": 3903 + }, + { + "epoch": 0.15, + "learning_rate": 4.752885355219287e-05, + "loss": 0.5786, + "step": 3904 + }, + { + "epoch": 0.15, + "learning_rate": 4.752821236214414e-05, + "loss": 0.5535, + "step": 3905 + }, + { + "epoch": 0.15, + "learning_rate": 4.752757117209541e-05, + "loss": 0.5901, + "step": 3906 + }, + { + "epoch": 0.15, + "learning_rate": 4.752692998204668e-05, + "loss": 0.6202, + "step": 3907 + }, + { + "epoch": 0.15, + "learning_rate": 4.752628879199795e-05, + "loss": 0.6203, + "step": 3908 + }, + { + "epoch": 0.15, + "learning_rate": 4.752564760194922e-05, + "loss": 0.5778, + "step": 3909 + }, + { + "epoch": 0.15, + "learning_rate": 4.7525006411900494e-05, + "loss": 0.6614, + "step": 3910 + }, + { + "epoch": 0.15, + "learning_rate": 4.752436522185176e-05, + "loss": 0.6459, + "step": 3911 + }, + { + "epoch": 0.15, + "learning_rate": 4.752372403180303e-05, + "loss": 0.5804, + "step": 3912 + }, + { + "epoch": 0.15, + "learning_rate": 4.7523082841754295e-05, + "loss": 0.5353, + "step": 3913 + }, + { + "epoch": 0.15, + "learning_rate": 4.752244165170557e-05, + "loss": 0.465, + "step": 3914 + }, + { + "epoch": 0.15, + "learning_rate": 4.752180046165683e-05, + "loss": 0.6009, + "step": 3915 + }, + { + "epoch": 0.15, + "learning_rate": 4.7521159271608104e-05, + "loss": 0.6134, + "step": 3916 + }, + { + "epoch": 0.15, + "learning_rate": 4.7520518081559375e-05, + "loss": 0.543, + "step": 3917 + }, + { + "epoch": 0.15, + "learning_rate": 4.751987689151065e-05, + "loss": 0.7149, + "step": 3918 + }, + { + "epoch": 0.15, + "learning_rate": 4.751923570146192e-05, + "loss": 0.5725, + "step": 3919 + }, + { + "epoch": 0.15, + "learning_rate": 4.7518594511413184e-05, + "loss": 0.545, + "step": 3920 + }, + { + "epoch": 0.15, + "learning_rate": 4.7517953321364456e-05, + "loss": 0.5982, + "step": 3921 + }, + { + "epoch": 0.15, + "learning_rate": 4.751731213131572e-05, + "loss": 0.638, + "step": 3922 + }, + { + "epoch": 0.15, + "learning_rate": 4.751667094126699e-05, + "loss": 0.5092, + "step": 3923 + }, + { + "epoch": 0.15, + "learning_rate": 4.7516029751218264e-05, + "loss": 0.6218, + "step": 3924 + }, + { + "epoch": 0.15, + "learning_rate": 4.7515388561169536e-05, + "loss": 0.5464, + "step": 3925 + }, + { + "epoch": 0.15, + "learning_rate": 4.751474737112081e-05, + "loss": 0.7073, + "step": 3926 + }, + { + "epoch": 0.15, + "learning_rate": 4.751410618107207e-05, + "loss": 0.5739, + "step": 3927 + }, + { + "epoch": 0.15, + "learning_rate": 4.7513464991023344e-05, + "loss": 0.6355, + "step": 3928 + }, + { + "epoch": 0.15, + "learning_rate": 4.751282380097461e-05, + "loss": 0.5994, + "step": 3929 + }, + { + "epoch": 0.15, + "learning_rate": 4.751218261092588e-05, + "loss": 0.6739, + "step": 3930 + }, + { + "epoch": 0.15, + "learning_rate": 4.7511541420877146e-05, + "loss": 0.6128, + "step": 3931 + }, + { + "epoch": 0.15, + "learning_rate": 4.751090023082842e-05, + "loss": 0.6411, + "step": 3932 + }, + { + "epoch": 0.15, + "learning_rate": 4.751025904077969e-05, + "loss": 0.5385, + "step": 3933 + }, + { + "epoch": 0.15, + "learning_rate": 4.750961785073096e-05, + "loss": 0.5722, + "step": 3934 + }, + { + "epoch": 0.15, + "learning_rate": 4.750897666068223e-05, + "loss": 0.6108, + "step": 3935 + }, + { + "epoch": 0.15, + "learning_rate": 4.75083354706335e-05, + "loss": 0.6275, + "step": 3936 + }, + { + "epoch": 0.15, + "learning_rate": 4.750769428058477e-05, + "loss": 0.4888, + "step": 3937 + }, + { + "epoch": 0.15, + "learning_rate": 4.7507053090536034e-05, + "loss": 0.6383, + "step": 3938 + }, + { + "epoch": 0.15, + "learning_rate": 4.7506411900487306e-05, + "loss": 0.5981, + "step": 3939 + }, + { + "epoch": 0.15, + "learning_rate": 4.750577071043857e-05, + "loss": 0.6075, + "step": 3940 + }, + { + "epoch": 0.15, + "learning_rate": 4.750512952038985e-05, + "loss": 0.5484, + "step": 3941 + }, + { + "epoch": 0.15, + "learning_rate": 4.750448833034112e-05, + "loss": 0.5396, + "step": 3942 + }, + { + "epoch": 0.15, + "learning_rate": 4.7503847140292386e-05, + "loss": 0.5477, + "step": 3943 + }, + { + "epoch": 0.15, + "learning_rate": 4.750320595024366e-05, + "loss": 0.5923, + "step": 3944 + }, + { + "epoch": 0.15, + "learning_rate": 4.750256476019492e-05, + "loss": 0.5817, + "step": 3945 + }, + { + "epoch": 0.15, + "learning_rate": 4.7501923570146194e-05, + "loss": 0.5456, + "step": 3946 + }, + { + "epoch": 0.15, + "learning_rate": 4.750128238009746e-05, + "loss": 0.5436, + "step": 3947 + }, + { + "epoch": 0.15, + "learning_rate": 4.750064119004873e-05, + "loss": 0.5864, + "step": 3948 + }, + { + "epoch": 0.15, + "learning_rate": 4.75e-05, + "loss": 0.5927, + "step": 3949 + }, + { + "epoch": 0.15, + "learning_rate": 4.7499358809951274e-05, + "loss": 0.6182, + "step": 3950 + }, + { + "epoch": 0.15, + "learning_rate": 4.7498717619902546e-05, + "loss": 0.616, + "step": 3951 + }, + { + "epoch": 0.15, + "learning_rate": 4.749807642985381e-05, + "loss": 0.6664, + "step": 3952 + }, + { + "epoch": 0.15, + "learning_rate": 4.749743523980508e-05, + "loss": 0.6514, + "step": 3953 + }, + { + "epoch": 0.15, + "learning_rate": 4.749679404975635e-05, + "loss": 0.4279, + "step": 3954 + }, + { + "epoch": 0.15, + "learning_rate": 4.749615285970762e-05, + "loss": 0.6325, + "step": 3955 + }, + { + "epoch": 0.15, + "learning_rate": 4.7495511669658884e-05, + "loss": 0.657, + "step": 3956 + }, + { + "epoch": 0.15, + "learning_rate": 4.749487047961016e-05, + "loss": 0.5784, + "step": 3957 + }, + { + "epoch": 0.15, + "learning_rate": 4.749422928956143e-05, + "loss": 0.5475, + "step": 3958 + }, + { + "epoch": 0.15, + "learning_rate": 4.74935880995127e-05, + "loss": 0.5481, + "step": 3959 + }, + { + "epoch": 0.15, + "learning_rate": 4.749294690946397e-05, + "loss": 0.6793, + "step": 3960 + }, + { + "epoch": 0.15, + "learning_rate": 4.7492305719415236e-05, + "loss": 0.6593, + "step": 3961 + }, + { + "epoch": 0.15, + "learning_rate": 4.749166452936651e-05, + "loss": 0.5818, + "step": 3962 + }, + { + "epoch": 0.15, + "learning_rate": 4.749102333931777e-05, + "loss": 0.5867, + "step": 3963 + }, + { + "epoch": 0.15, + "learning_rate": 4.7490382149269044e-05, + "loss": 0.5461, + "step": 3964 + }, + { + "epoch": 0.15, + "learning_rate": 4.7489740959220316e-05, + "loss": 0.6123, + "step": 3965 + }, + { + "epoch": 0.15, + "learning_rate": 4.748909976917159e-05, + "loss": 0.6489, + "step": 3966 + }, + { + "epoch": 0.15, + "learning_rate": 4.748845857912285e-05, + "loss": 0.6897, + "step": 3967 + }, + { + "epoch": 0.15, + "learning_rate": 4.7487817389074125e-05, + "loss": 0.6131, + "step": 3968 + }, + { + "epoch": 0.15, + "learning_rate": 4.7487176199025396e-05, + "loss": 0.5955, + "step": 3969 + }, + { + "epoch": 0.15, + "learning_rate": 4.748653500897666e-05, + "loss": 0.6151, + "step": 3970 + }, + { + "epoch": 0.15, + "learning_rate": 4.748589381892793e-05, + "loss": 0.5871, + "step": 3971 + }, + { + "epoch": 0.15, + "learning_rate": 4.74852526288792e-05, + "loss": 0.5798, + "step": 3972 + }, + { + "epoch": 0.15, + "learning_rate": 4.7484611438830476e-05, + "loss": 0.5538, + "step": 3973 + }, + { + "epoch": 0.15, + "learning_rate": 4.748397024878174e-05, + "loss": 0.6438, + "step": 3974 + }, + { + "epoch": 0.15, + "learning_rate": 4.748332905873301e-05, + "loss": 0.6546, + "step": 3975 + }, + { + "epoch": 0.15, + "learning_rate": 4.748268786868428e-05, + "loss": 0.6099, + "step": 3976 + }, + { + "epoch": 0.15, + "learning_rate": 4.748204667863555e-05, + "loss": 0.5803, + "step": 3977 + }, + { + "epoch": 0.15, + "learning_rate": 4.748140548858682e-05, + "loss": 0.4538, + "step": 3978 + }, + { + "epoch": 0.15, + "learning_rate": 4.7480764298538086e-05, + "loss": 0.6345, + "step": 3979 + }, + { + "epoch": 0.15, + "learning_rate": 4.748012310848936e-05, + "loss": 0.6238, + "step": 3980 + }, + { + "epoch": 0.15, + "learning_rate": 4.747948191844063e-05, + "loss": 0.5935, + "step": 3981 + }, + { + "epoch": 0.15, + "learning_rate": 4.74788407283919e-05, + "loss": 0.6254, + "step": 3982 + }, + { + "epoch": 0.15, + "learning_rate": 4.7478199538343166e-05, + "loss": 0.6507, + "step": 3983 + }, + { + "epoch": 0.15, + "learning_rate": 4.747755834829444e-05, + "loss": 0.5997, + "step": 3984 + }, + { + "epoch": 0.15, + "learning_rate": 4.74769171582457e-05, + "loss": 0.606, + "step": 3985 + }, + { + "epoch": 0.15, + "learning_rate": 4.7476275968196975e-05, + "loss": 0.5106, + "step": 3986 + }, + { + "epoch": 0.15, + "learning_rate": 4.7475634778148247e-05, + "loss": 0.6172, + "step": 3987 + }, + { + "epoch": 0.15, + "learning_rate": 4.747499358809951e-05, + "loss": 0.6329, + "step": 3988 + }, + { + "epoch": 0.15, + "learning_rate": 4.747435239805078e-05, + "loss": 0.5764, + "step": 3989 + }, + { + "epoch": 0.15, + "learning_rate": 4.7473711208002055e-05, + "loss": 0.6322, + "step": 3990 + }, + { + "epoch": 0.15, + "learning_rate": 4.747307001795333e-05, + "loss": 0.5565, + "step": 3991 + }, + { + "epoch": 0.15, + "learning_rate": 4.747242882790459e-05, + "loss": 0.5419, + "step": 3992 + }, + { + "epoch": 0.15, + "learning_rate": 4.747178763785586e-05, + "loss": 0.6003, + "step": 3993 + }, + { + "epoch": 0.15, + "learning_rate": 4.747114644780713e-05, + "loss": 0.6074, + "step": 3994 + }, + { + "epoch": 0.15, + "learning_rate": 4.74705052577584e-05, + "loss": 0.5389, + "step": 3995 + }, + { + "epoch": 0.15, + "learning_rate": 4.746986406770967e-05, + "loss": 0.6593, + "step": 3996 + }, + { + "epoch": 0.15, + "learning_rate": 4.746922287766094e-05, + "loss": 0.5913, + "step": 3997 + }, + { + "epoch": 0.15, + "learning_rate": 4.7468581687612215e-05, + "loss": 0.5442, + "step": 3998 + }, + { + "epoch": 0.15, + "learning_rate": 4.746794049756348e-05, + "loss": 0.6488, + "step": 3999 + }, + { + "epoch": 0.15, + "learning_rate": 4.746729930751475e-05, + "loss": 0.5802, + "step": 4000 + }, + { + "epoch": 0.15, + "learning_rate": 4.746665811746602e-05, + "loss": 0.5482, + "step": 4001 + }, + { + "epoch": 0.15, + "learning_rate": 4.746601692741729e-05, + "loss": 0.6009, + "step": 4002 + }, + { + "epoch": 0.15, + "learning_rate": 4.746537573736855e-05, + "loss": 0.5869, + "step": 4003 + }, + { + "epoch": 0.15, + "learning_rate": 4.7464734547319825e-05, + "loss": 0.5968, + "step": 4004 + }, + { + "epoch": 0.15, + "learning_rate": 4.74640933572711e-05, + "loss": 0.627, + "step": 4005 + }, + { + "epoch": 0.15, + "learning_rate": 4.746345216722237e-05, + "loss": 0.5893, + "step": 4006 + }, + { + "epoch": 0.15, + "learning_rate": 4.746281097717364e-05, + "loss": 0.6585, + "step": 4007 + }, + { + "epoch": 0.15, + "learning_rate": 4.7462169787124905e-05, + "loss": 0.7039, + "step": 4008 + }, + { + "epoch": 0.15, + "learning_rate": 4.746152859707618e-05, + "loss": 0.6968, + "step": 4009 + }, + { + "epoch": 0.15, + "learning_rate": 4.746088740702744e-05, + "loss": 0.6535, + "step": 4010 + }, + { + "epoch": 0.15, + "learning_rate": 4.7460246216978714e-05, + "loss": 0.515, + "step": 4011 + }, + { + "epoch": 0.15, + "learning_rate": 4.7459605026929985e-05, + "loss": 0.5498, + "step": 4012 + }, + { + "epoch": 0.15, + "learning_rate": 4.745896383688126e-05, + "loss": 0.6346, + "step": 4013 + }, + { + "epoch": 0.15, + "learning_rate": 4.745832264683253e-05, + "loss": 0.5708, + "step": 4014 + }, + { + "epoch": 0.15, + "learning_rate": 4.7457681456783794e-05, + "loss": 0.5836, + "step": 4015 + }, + { + "epoch": 0.15, + "learning_rate": 4.7457040266735065e-05, + "loss": 0.5924, + "step": 4016 + }, + { + "epoch": 0.15, + "learning_rate": 4.745639907668633e-05, + "loss": 0.613, + "step": 4017 + }, + { + "epoch": 0.15, + "learning_rate": 4.74557578866376e-05, + "loss": 0.6645, + "step": 4018 + }, + { + "epoch": 0.15, + "learning_rate": 4.745511669658887e-05, + "loss": 0.6082, + "step": 4019 + }, + { + "epoch": 0.15, + "learning_rate": 4.745447550654014e-05, + "loss": 0.6651, + "step": 4020 + }, + { + "epoch": 0.15, + "learning_rate": 4.745383431649141e-05, + "loss": 0.6497, + "step": 4021 + }, + { + "epoch": 0.15, + "learning_rate": 4.745319312644268e-05, + "loss": 0.6069, + "step": 4022 + }, + { + "epoch": 0.15, + "learning_rate": 4.7452551936393954e-05, + "loss": 0.6437, + "step": 4023 + }, + { + "epoch": 0.15, + "learning_rate": 4.745191074634522e-05, + "loss": 0.6566, + "step": 4024 + }, + { + "epoch": 0.15, + "learning_rate": 4.745126955629649e-05, + "loss": 0.6697, + "step": 4025 + }, + { + "epoch": 0.15, + "learning_rate": 4.7450628366247755e-05, + "loss": 0.594, + "step": 4026 + }, + { + "epoch": 0.15, + "learning_rate": 4.744998717619903e-05, + "loss": 0.6124, + "step": 4027 + }, + { + "epoch": 0.15, + "learning_rate": 4.744934598615029e-05, + "loss": 0.6516, + "step": 4028 + }, + { + "epoch": 0.15, + "learning_rate": 4.744870479610157e-05, + "loss": 0.5962, + "step": 4029 + }, + { + "epoch": 0.15, + "learning_rate": 4.7448063606052835e-05, + "loss": 0.6069, + "step": 4030 + }, + { + "epoch": 0.15, + "learning_rate": 4.744742241600411e-05, + "loss": 0.5814, + "step": 4031 + }, + { + "epoch": 0.16, + "learning_rate": 4.744678122595538e-05, + "loss": 0.5511, + "step": 4032 + }, + { + "epoch": 0.16, + "learning_rate": 4.7446140035906644e-05, + "loss": 0.6392, + "step": 4033 + }, + { + "epoch": 0.16, + "learning_rate": 4.7445498845857916e-05, + "loss": 0.6821, + "step": 4034 + }, + { + "epoch": 0.16, + "learning_rate": 4.744485765580918e-05, + "loss": 0.6107, + "step": 4035 + }, + { + "epoch": 0.16, + "learning_rate": 4.744421646576045e-05, + "loss": 0.5188, + "step": 4036 + }, + { + "epoch": 0.16, + "learning_rate": 4.7443575275711724e-05, + "loss": 0.5203, + "step": 4037 + }, + { + "epoch": 0.16, + "learning_rate": 4.7442934085662996e-05, + "loss": 0.5857, + "step": 4038 + }, + { + "epoch": 0.16, + "learning_rate": 4.744229289561426e-05, + "loss": 0.586, + "step": 4039 + }, + { + "epoch": 0.16, + "learning_rate": 4.744165170556553e-05, + "loss": 0.5953, + "step": 4040 + }, + { + "epoch": 0.16, + "learning_rate": 4.7441010515516804e-05, + "loss": 0.4162, + "step": 4041 + }, + { + "epoch": 0.16, + "learning_rate": 4.744036932546807e-05, + "loss": 0.6111, + "step": 4042 + }, + { + "epoch": 0.16, + "learning_rate": 4.743972813541934e-05, + "loss": 0.5897, + "step": 4043 + }, + { + "epoch": 0.16, + "learning_rate": 4.7439086945370606e-05, + "loss": 0.5772, + "step": 4044 + }, + { + "epoch": 0.16, + "learning_rate": 4.7438445755321884e-05, + "loss": 0.6951, + "step": 4045 + }, + { + "epoch": 0.16, + "learning_rate": 4.743780456527315e-05, + "loss": 0.5878, + "step": 4046 + }, + { + "epoch": 0.16, + "learning_rate": 4.743716337522442e-05, + "loss": 0.6627, + "step": 4047 + }, + { + "epoch": 0.16, + "learning_rate": 4.7436522185175686e-05, + "loss": 0.6808, + "step": 4048 + }, + { + "epoch": 0.16, + "learning_rate": 4.743588099512696e-05, + "loss": 0.5384, + "step": 4049 + }, + { + "epoch": 0.16, + "learning_rate": 4.743523980507823e-05, + "loss": 0.7647, + "step": 4050 + }, + { + "epoch": 0.16, + "learning_rate": 4.7434598615029494e-05, + "loss": 0.6147, + "step": 4051 + }, + { + "epoch": 0.16, + "learning_rate": 4.7433957424980766e-05, + "loss": 0.6266, + "step": 4052 + }, + { + "epoch": 0.16, + "learning_rate": 4.743331623493204e-05, + "loss": 0.6134, + "step": 4053 + }, + { + "epoch": 0.16, + "learning_rate": 4.743267504488331e-05, + "loss": 0.5298, + "step": 4054 + }, + { + "epoch": 0.16, + "learning_rate": 4.7432033854834574e-05, + "loss": 0.6126, + "step": 4055 + }, + { + "epoch": 0.16, + "learning_rate": 4.7431392664785846e-05, + "loss": 0.6648, + "step": 4056 + }, + { + "epoch": 0.16, + "learning_rate": 4.743075147473711e-05, + "loss": 0.6442, + "step": 4057 + }, + { + "epoch": 0.16, + "learning_rate": 4.743011028468838e-05, + "loss": 0.6127, + "step": 4058 + }, + { + "epoch": 0.16, + "learning_rate": 4.7429469094639654e-05, + "loss": 0.5449, + "step": 4059 + }, + { + "epoch": 0.16, + "learning_rate": 4.742882790459092e-05, + "loss": 0.5629, + "step": 4060 + }, + { + "epoch": 0.16, + "learning_rate": 4.742818671454219e-05, + "loss": 0.6286, + "step": 4061 + }, + { + "epoch": 0.16, + "learning_rate": 4.742754552449346e-05, + "loss": 0.5815, + "step": 4062 + }, + { + "epoch": 0.16, + "learning_rate": 4.7426904334444734e-05, + "loss": 0.5803, + "step": 4063 + }, + { + "epoch": 0.16, + "learning_rate": 4.7426263144396e-05, + "loss": 0.5774, + "step": 4064 + }, + { + "epoch": 0.16, + "learning_rate": 4.742562195434727e-05, + "loss": 0.5923, + "step": 4065 + }, + { + "epoch": 0.16, + "learning_rate": 4.7424980764298536e-05, + "loss": 0.5347, + "step": 4066 + }, + { + "epoch": 0.16, + "learning_rate": 4.742433957424981e-05, + "loss": 0.6638, + "step": 4067 + }, + { + "epoch": 0.16, + "learning_rate": 4.742369838420108e-05, + "loss": 0.5596, + "step": 4068 + }, + { + "epoch": 0.16, + "learning_rate": 4.742305719415235e-05, + "loss": 0.6044, + "step": 4069 + }, + { + "epoch": 0.16, + "learning_rate": 4.742241600410362e-05, + "loss": 0.5785, + "step": 4070 + }, + { + "epoch": 0.16, + "learning_rate": 4.742177481405489e-05, + "loss": 0.57, + "step": 4071 + }, + { + "epoch": 0.16, + "learning_rate": 4.742113362400616e-05, + "loss": 0.5985, + "step": 4072 + }, + { + "epoch": 0.16, + "learning_rate": 4.7420492433957424e-05, + "loss": 0.5622, + "step": 4073 + }, + { + "epoch": 0.16, + "learning_rate": 4.7419851243908696e-05, + "loss": 0.6076, + "step": 4074 + }, + { + "epoch": 0.16, + "learning_rate": 4.741921005385996e-05, + "loss": 0.6279, + "step": 4075 + }, + { + "epoch": 0.16, + "learning_rate": 4.741856886381123e-05, + "loss": 0.7098, + "step": 4076 + }, + { + "epoch": 0.16, + "learning_rate": 4.7417927673762505e-05, + "loss": 0.6854, + "step": 4077 + }, + { + "epoch": 0.16, + "learning_rate": 4.7417286483713776e-05, + "loss": 0.5057, + "step": 4078 + }, + { + "epoch": 0.16, + "learning_rate": 4.741664529366505e-05, + "loss": 0.6143, + "step": 4079 + }, + { + "epoch": 0.16, + "learning_rate": 4.741600410361631e-05, + "loss": 0.5703, + "step": 4080 + }, + { + "epoch": 0.16, + "learning_rate": 4.7415362913567585e-05, + "loss": 0.6024, + "step": 4081 + }, + { + "epoch": 0.16, + "learning_rate": 4.741472172351885e-05, + "loss": 0.6894, + "step": 4082 + }, + { + "epoch": 0.16, + "learning_rate": 4.741408053347012e-05, + "loss": 0.5623, + "step": 4083 + }, + { + "epoch": 0.16, + "learning_rate": 4.741343934342139e-05, + "loss": 0.6266, + "step": 4084 + }, + { + "epoch": 0.16, + "learning_rate": 4.7412798153372665e-05, + "loss": 0.7227, + "step": 4085 + }, + { + "epoch": 0.16, + "learning_rate": 4.7412156963323936e-05, + "loss": 0.6827, + "step": 4086 + }, + { + "epoch": 0.16, + "learning_rate": 4.74115157732752e-05, + "loss": 0.6023, + "step": 4087 + }, + { + "epoch": 0.16, + "learning_rate": 4.741087458322647e-05, + "loss": 0.5969, + "step": 4088 + }, + { + "epoch": 0.16, + "learning_rate": 4.741023339317774e-05, + "loss": 0.6502, + "step": 4089 + }, + { + "epoch": 0.16, + "learning_rate": 4.740959220312901e-05, + "loss": 0.6992, + "step": 4090 + }, + { + "epoch": 0.16, + "learning_rate": 4.7408951013080275e-05, + "loss": 0.5491, + "step": 4091 + }, + { + "epoch": 0.16, + "learning_rate": 4.7408309823031546e-05, + "loss": 0.5786, + "step": 4092 + }, + { + "epoch": 0.16, + "learning_rate": 4.740766863298282e-05, + "loss": 0.5578, + "step": 4093 + }, + { + "epoch": 0.16, + "learning_rate": 4.740702744293409e-05, + "loss": 0.6036, + "step": 4094 + }, + { + "epoch": 0.16, + "learning_rate": 4.740638625288536e-05, + "loss": 0.56, + "step": 4095 + }, + { + "epoch": 0.16, + "learning_rate": 4.7405745062836626e-05, + "loss": 0.5823, + "step": 4096 + }, + { + "epoch": 0.16, + "learning_rate": 4.74051038727879e-05, + "loss": 0.5731, + "step": 4097 + }, + { + "epoch": 0.16, + "learning_rate": 4.740446268273916e-05, + "loss": 0.55, + "step": 4098 + }, + { + "epoch": 0.16, + "learning_rate": 4.7403821492690435e-05, + "loss": 0.6844, + "step": 4099 + }, + { + "epoch": 0.16, + "learning_rate": 4.7403180302641707e-05, + "loss": 0.7929, + "step": 4100 + }, + { + "epoch": 0.16, + "learning_rate": 4.740253911259298e-05, + "loss": 0.5819, + "step": 4101 + }, + { + "epoch": 0.16, + "learning_rate": 4.740189792254424e-05, + "loss": 0.6348, + "step": 4102 + }, + { + "epoch": 0.16, + "learning_rate": 4.7401256732495515e-05, + "loss": 0.5941, + "step": 4103 + }, + { + "epoch": 0.16, + "learning_rate": 4.740061554244679e-05, + "loss": 0.6127, + "step": 4104 + }, + { + "epoch": 0.16, + "learning_rate": 4.739997435239805e-05, + "loss": 0.6879, + "step": 4105 + }, + { + "epoch": 0.16, + "learning_rate": 4.739933316234932e-05, + "loss": 0.6338, + "step": 4106 + }, + { + "epoch": 0.16, + "learning_rate": 4.739869197230059e-05, + "loss": 0.5703, + "step": 4107 + }, + { + "epoch": 0.16, + "learning_rate": 4.739805078225186e-05, + "loss": 0.6163, + "step": 4108 + }, + { + "epoch": 0.16, + "learning_rate": 4.739740959220313e-05, + "loss": 0.7145, + "step": 4109 + }, + { + "epoch": 0.16, + "learning_rate": 4.7396768402154403e-05, + "loss": 0.5573, + "step": 4110 + }, + { + "epoch": 0.16, + "learning_rate": 4.739612721210567e-05, + "loss": 0.5751, + "step": 4111 + }, + { + "epoch": 0.16, + "learning_rate": 4.739548602205694e-05, + "loss": 0.5254, + "step": 4112 + }, + { + "epoch": 0.16, + "learning_rate": 4.739484483200821e-05, + "loss": 0.5169, + "step": 4113 + }, + { + "epoch": 0.16, + "learning_rate": 4.739420364195948e-05, + "loss": 0.5678, + "step": 4114 + }, + { + "epoch": 0.16, + "learning_rate": 4.739356245191075e-05, + "loss": 0.5401, + "step": 4115 + }, + { + "epoch": 0.16, + "learning_rate": 4.739292126186201e-05, + "loss": 0.5987, + "step": 4116 + }, + { + "epoch": 0.16, + "learning_rate": 4.739228007181329e-05, + "loss": 0.636, + "step": 4117 + }, + { + "epoch": 0.16, + "learning_rate": 4.739163888176456e-05, + "loss": 0.7071, + "step": 4118 + }, + { + "epoch": 0.16, + "learning_rate": 4.739099769171583e-05, + "loss": 0.5031, + "step": 4119 + }, + { + "epoch": 0.16, + "learning_rate": 4.73903565016671e-05, + "loss": 0.6484, + "step": 4120 + }, + { + "epoch": 0.16, + "learning_rate": 4.7389715311618365e-05, + "loss": 0.659, + "step": 4121 + }, + { + "epoch": 0.16, + "learning_rate": 4.738907412156964e-05, + "loss": 0.5576, + "step": 4122 + }, + { + "epoch": 0.16, + "learning_rate": 4.73884329315209e-05, + "loss": 0.6205, + "step": 4123 + }, + { + "epoch": 0.16, + "learning_rate": 4.7387791741472174e-05, + "loss": 0.5314, + "step": 4124 + }, + { + "epoch": 0.16, + "learning_rate": 4.7387150551423445e-05, + "loss": 0.5895, + "step": 4125 + }, + { + "epoch": 0.16, + "learning_rate": 4.738650936137472e-05, + "loss": 0.5128, + "step": 4126 + }, + { + "epoch": 0.16, + "learning_rate": 4.738586817132598e-05, + "loss": 0.6646, + "step": 4127 + }, + { + "epoch": 0.16, + "learning_rate": 4.7385226981277254e-05, + "loss": 0.5162, + "step": 4128 + }, + { + "epoch": 0.16, + "learning_rate": 4.7384585791228525e-05, + "loss": 0.593, + "step": 4129 + }, + { + "epoch": 0.16, + "learning_rate": 4.738394460117979e-05, + "loss": 0.6501, + "step": 4130 + }, + { + "epoch": 0.16, + "learning_rate": 4.738330341113106e-05, + "loss": 0.544, + "step": 4131 + }, + { + "epoch": 0.16, + "learning_rate": 4.738266222108233e-05, + "loss": 0.6178, + "step": 4132 + }, + { + "epoch": 0.16, + "learning_rate": 4.7382021031033605e-05, + "loss": 0.685, + "step": 4133 + }, + { + "epoch": 0.16, + "learning_rate": 4.738137984098487e-05, + "loss": 0.6316, + "step": 4134 + }, + { + "epoch": 0.16, + "learning_rate": 4.738073865093614e-05, + "loss": 0.5262, + "step": 4135 + }, + { + "epoch": 0.16, + "learning_rate": 4.738009746088741e-05, + "loss": 0.5688, + "step": 4136 + }, + { + "epoch": 0.16, + "learning_rate": 4.737945627083868e-05, + "loss": 0.6067, + "step": 4137 + }, + { + "epoch": 0.16, + "learning_rate": 4.737881508078995e-05, + "loss": 0.5979, + "step": 4138 + }, + { + "epoch": 0.16, + "learning_rate": 4.7378173890741215e-05, + "loss": 0.6443, + "step": 4139 + }, + { + "epoch": 0.16, + "learning_rate": 4.737753270069249e-05, + "loss": 0.6695, + "step": 4140 + }, + { + "epoch": 0.16, + "learning_rate": 4.737689151064376e-05, + "loss": 0.6689, + "step": 4141 + }, + { + "epoch": 0.16, + "learning_rate": 4.737625032059503e-05, + "loss": 0.6188, + "step": 4142 + }, + { + "epoch": 0.16, + "learning_rate": 4.7375609130546296e-05, + "loss": 0.6006, + "step": 4143 + }, + { + "epoch": 0.16, + "learning_rate": 4.737496794049757e-05, + "loss": 0.6139, + "step": 4144 + }, + { + "epoch": 0.16, + "learning_rate": 4.737432675044883e-05, + "loss": 0.5224, + "step": 4145 + }, + { + "epoch": 0.16, + "learning_rate": 4.7373685560400104e-05, + "loss": 0.5723, + "step": 4146 + }, + { + "epoch": 0.16, + "learning_rate": 4.7373044370351376e-05, + "loss": 0.6039, + "step": 4147 + }, + { + "epoch": 0.16, + "learning_rate": 4.737240318030264e-05, + "loss": 0.6052, + "step": 4148 + }, + { + "epoch": 0.16, + "learning_rate": 4.737176199025391e-05, + "loss": 0.5522, + "step": 4149 + }, + { + "epoch": 0.16, + "learning_rate": 4.7371120800205184e-05, + "loss": 0.5577, + "step": 4150 + }, + { + "epoch": 0.16, + "learning_rate": 4.7370479610156456e-05, + "loss": 0.5577, + "step": 4151 + }, + { + "epoch": 0.16, + "learning_rate": 4.736983842010772e-05, + "loss": 0.6519, + "step": 4152 + }, + { + "epoch": 0.16, + "learning_rate": 4.736919723005899e-05, + "loss": 0.5373, + "step": 4153 + }, + { + "epoch": 0.16, + "learning_rate": 4.736855604001026e-05, + "loss": 0.5967, + "step": 4154 + }, + { + "epoch": 0.16, + "learning_rate": 4.736791484996153e-05, + "loss": 0.6394, + "step": 4155 + }, + { + "epoch": 0.16, + "learning_rate": 4.73672736599128e-05, + "loss": 0.5378, + "step": 4156 + }, + { + "epoch": 0.16, + "learning_rate": 4.736663246986407e-05, + "loss": 0.641, + "step": 4157 + }, + { + "epoch": 0.16, + "learning_rate": 4.7365991279815344e-05, + "loss": 0.601, + "step": 4158 + }, + { + "epoch": 0.16, + "learning_rate": 4.736535008976661e-05, + "loss": 0.5341, + "step": 4159 + }, + { + "epoch": 0.16, + "learning_rate": 4.736470889971788e-05, + "loss": 0.5343, + "step": 4160 + }, + { + "epoch": 0.16, + "learning_rate": 4.7364067709669146e-05, + "loss": 0.5517, + "step": 4161 + }, + { + "epoch": 0.16, + "learning_rate": 4.736342651962042e-05, + "loss": 0.6189, + "step": 4162 + }, + { + "epoch": 0.16, + "learning_rate": 4.736278532957168e-05, + "loss": 0.587, + "step": 4163 + }, + { + "epoch": 0.16, + "learning_rate": 4.7362144139522954e-05, + "loss": 0.519, + "step": 4164 + }, + { + "epoch": 0.16, + "learning_rate": 4.7361502949474226e-05, + "loss": 0.5811, + "step": 4165 + }, + { + "epoch": 0.16, + "learning_rate": 4.73608617594255e-05, + "loss": 0.7612, + "step": 4166 + }, + { + "epoch": 0.16, + "learning_rate": 4.736022056937677e-05, + "loss": 0.5558, + "step": 4167 + }, + { + "epoch": 0.16, + "learning_rate": 4.7359579379328034e-05, + "loss": 0.6152, + "step": 4168 + }, + { + "epoch": 0.16, + "learning_rate": 4.7358938189279306e-05, + "loss": 0.5495, + "step": 4169 + }, + { + "epoch": 0.16, + "learning_rate": 4.735829699923057e-05, + "loss": 0.5819, + "step": 4170 + }, + { + "epoch": 0.16, + "learning_rate": 4.735765580918184e-05, + "loss": 0.6288, + "step": 4171 + }, + { + "epoch": 0.16, + "learning_rate": 4.7357014619133114e-05, + "loss": 0.64, + "step": 4172 + }, + { + "epoch": 0.16, + "learning_rate": 4.7356373429084386e-05, + "loss": 0.5339, + "step": 4173 + }, + { + "epoch": 0.16, + "learning_rate": 4.735573223903566e-05, + "loss": 0.5011, + "step": 4174 + }, + { + "epoch": 0.16, + "learning_rate": 4.735509104898692e-05, + "loss": 0.5719, + "step": 4175 + }, + { + "epoch": 0.16, + "learning_rate": 4.7354449858938194e-05, + "loss": 0.5658, + "step": 4176 + }, + { + "epoch": 0.16, + "learning_rate": 4.735380866888946e-05, + "loss": 0.6811, + "step": 4177 + }, + { + "epoch": 0.16, + "learning_rate": 4.735316747884073e-05, + "loss": 0.5507, + "step": 4178 + }, + { + "epoch": 0.16, + "learning_rate": 4.7352526288791996e-05, + "loss": 0.5182, + "step": 4179 + }, + { + "epoch": 0.16, + "learning_rate": 4.735188509874327e-05, + "loss": 0.5754, + "step": 4180 + }, + { + "epoch": 0.16, + "learning_rate": 4.735124390869454e-05, + "loss": 0.593, + "step": 4181 + }, + { + "epoch": 0.16, + "learning_rate": 4.735060271864581e-05, + "loss": 0.6133, + "step": 4182 + }, + { + "epoch": 0.16, + "learning_rate": 4.734996152859708e-05, + "loss": 0.6719, + "step": 4183 + }, + { + "epoch": 0.16, + "learning_rate": 4.734932033854835e-05, + "loss": 0.5422, + "step": 4184 + }, + { + "epoch": 0.16, + "learning_rate": 4.734867914849962e-05, + "loss": 0.586, + "step": 4185 + }, + { + "epoch": 0.16, + "learning_rate": 4.7348037958450884e-05, + "loss": 0.6016, + "step": 4186 + }, + { + "epoch": 0.16, + "learning_rate": 4.7347396768402156e-05, + "loss": 0.5867, + "step": 4187 + }, + { + "epoch": 0.16, + "learning_rate": 4.734675557835342e-05, + "loss": 0.5893, + "step": 4188 + }, + { + "epoch": 0.16, + "learning_rate": 4.73461143883047e-05, + "loss": 0.7295, + "step": 4189 + }, + { + "epoch": 0.16, + "learning_rate": 4.7345473198255965e-05, + "loss": 0.6951, + "step": 4190 + }, + { + "epoch": 0.16, + "learning_rate": 4.7344832008207236e-05, + "loss": 0.4994, + "step": 4191 + }, + { + "epoch": 0.16, + "learning_rate": 4.734419081815851e-05, + "loss": 0.5626, + "step": 4192 + }, + { + "epoch": 0.16, + "learning_rate": 4.734354962810977e-05, + "loss": 0.5926, + "step": 4193 + }, + { + "epoch": 0.16, + "learning_rate": 4.7342908438061045e-05, + "loss": 0.6074, + "step": 4194 + }, + { + "epoch": 0.16, + "learning_rate": 4.734226724801231e-05, + "loss": 0.5161, + "step": 4195 + }, + { + "epoch": 0.16, + "learning_rate": 4.734162605796358e-05, + "loss": 0.6213, + "step": 4196 + }, + { + "epoch": 0.16, + "learning_rate": 4.734098486791485e-05, + "loss": 0.5689, + "step": 4197 + }, + { + "epoch": 0.16, + "learning_rate": 4.7340343677866125e-05, + "loss": 0.6532, + "step": 4198 + }, + { + "epoch": 0.16, + "learning_rate": 4.733970248781739e-05, + "loss": 0.6541, + "step": 4199 + }, + { + "epoch": 0.16, + "learning_rate": 4.733906129776866e-05, + "loss": 0.6666, + "step": 4200 + }, + { + "epoch": 0.16, + "learning_rate": 4.733842010771993e-05, + "loss": 0.5653, + "step": 4201 + }, + { + "epoch": 0.16, + "learning_rate": 4.73377789176712e-05, + "loss": 0.6187, + "step": 4202 + }, + { + "epoch": 0.16, + "learning_rate": 4.733713772762247e-05, + "loss": 0.5413, + "step": 4203 + }, + { + "epoch": 0.16, + "learning_rate": 4.7336496537573735e-05, + "loss": 0.5973, + "step": 4204 + }, + { + "epoch": 0.16, + "learning_rate": 4.733585534752501e-05, + "loss": 0.5687, + "step": 4205 + }, + { + "epoch": 0.16, + "learning_rate": 4.733521415747628e-05, + "loss": 0.5754, + "step": 4206 + }, + { + "epoch": 0.16, + "learning_rate": 4.733457296742755e-05, + "loss": 0.5346, + "step": 4207 + }, + { + "epoch": 0.16, + "learning_rate": 4.7333931777378815e-05, + "loss": 0.6761, + "step": 4208 + }, + { + "epoch": 0.16, + "learning_rate": 4.7333290587330087e-05, + "loss": 0.5468, + "step": 4209 + }, + { + "epoch": 0.16, + "learning_rate": 4.733264939728136e-05, + "loss": 0.5869, + "step": 4210 + }, + { + "epoch": 0.16, + "learning_rate": 4.733200820723262e-05, + "loss": 0.5588, + "step": 4211 + }, + { + "epoch": 0.16, + "learning_rate": 4.7331367017183895e-05, + "loss": 0.7361, + "step": 4212 + }, + { + "epoch": 0.16, + "learning_rate": 4.7330725827135167e-05, + "loss": 0.582, + "step": 4213 + }, + { + "epoch": 0.16, + "learning_rate": 4.733008463708644e-05, + "loss": 0.6724, + "step": 4214 + }, + { + "epoch": 0.16, + "learning_rate": 4.73294434470377e-05, + "loss": 0.5832, + "step": 4215 + }, + { + "epoch": 0.16, + "learning_rate": 4.7328802256988975e-05, + "loss": 0.6042, + "step": 4216 + }, + { + "epoch": 0.16, + "learning_rate": 4.732816106694024e-05, + "loss": 0.5739, + "step": 4217 + }, + { + "epoch": 0.16, + "learning_rate": 4.732751987689151e-05, + "loss": 0.581, + "step": 4218 + }, + { + "epoch": 0.16, + "learning_rate": 4.732687868684278e-05, + "loss": 0.6092, + "step": 4219 + }, + { + "epoch": 0.16, + "learning_rate": 4.732623749679405e-05, + "loss": 0.6667, + "step": 4220 + }, + { + "epoch": 0.16, + "learning_rate": 4.732559630674533e-05, + "loss": 0.6068, + "step": 4221 + }, + { + "epoch": 0.16, + "learning_rate": 4.732495511669659e-05, + "loss": 0.5781, + "step": 4222 + }, + { + "epoch": 0.16, + "learning_rate": 4.7324313926647863e-05, + "loss": 0.5475, + "step": 4223 + }, + { + "epoch": 0.16, + "learning_rate": 4.732367273659913e-05, + "loss": 0.5369, + "step": 4224 + }, + { + "epoch": 0.16, + "learning_rate": 4.73230315465504e-05, + "loss": 0.5848, + "step": 4225 + }, + { + "epoch": 0.16, + "learning_rate": 4.7322390356501665e-05, + "loss": 0.5299, + "step": 4226 + }, + { + "epoch": 0.16, + "learning_rate": 4.732174916645294e-05, + "loss": 0.5572, + "step": 4227 + }, + { + "epoch": 0.16, + "learning_rate": 4.732110797640421e-05, + "loss": 0.5812, + "step": 4228 + }, + { + "epoch": 0.16, + "learning_rate": 4.732046678635548e-05, + "loss": 0.5972, + "step": 4229 + }, + { + "epoch": 0.16, + "learning_rate": 4.731982559630675e-05, + "loss": 0.5851, + "step": 4230 + }, + { + "epoch": 0.16, + "learning_rate": 4.731918440625802e-05, + "loss": 0.6487, + "step": 4231 + }, + { + "epoch": 0.16, + "learning_rate": 4.731854321620929e-05, + "loss": 0.5647, + "step": 4232 + }, + { + "epoch": 0.16, + "learning_rate": 4.7317902026160553e-05, + "loss": 0.6963, + "step": 4233 + }, + { + "epoch": 0.16, + "learning_rate": 4.7317260836111825e-05, + "loss": 0.5146, + "step": 4234 + }, + { + "epoch": 0.16, + "learning_rate": 4.731661964606309e-05, + "loss": 0.5376, + "step": 4235 + }, + { + "epoch": 0.16, + "learning_rate": 4.731597845601436e-05, + "loss": 0.599, + "step": 4236 + }, + { + "epoch": 0.16, + "learning_rate": 4.7315337265965634e-05, + "loss": 0.5922, + "step": 4237 + }, + { + "epoch": 0.16, + "learning_rate": 4.7314696075916905e-05, + "loss": 0.5554, + "step": 4238 + }, + { + "epoch": 0.16, + "learning_rate": 4.731405488586818e-05, + "loss": 0.5823, + "step": 4239 + }, + { + "epoch": 0.16, + "learning_rate": 4.731341369581944e-05, + "loss": 0.6816, + "step": 4240 + }, + { + "epoch": 0.16, + "learning_rate": 4.7312772505770714e-05, + "loss": 0.6758, + "step": 4241 + }, + { + "epoch": 0.16, + "learning_rate": 4.731213131572198e-05, + "loss": 0.621, + "step": 4242 + }, + { + "epoch": 0.16, + "learning_rate": 4.731149012567325e-05, + "loss": 0.5426, + "step": 4243 + }, + { + "epoch": 0.16, + "learning_rate": 4.731084893562452e-05, + "loss": 0.6371, + "step": 4244 + }, + { + "epoch": 0.16, + "learning_rate": 4.7310207745575794e-05, + "loss": 0.6473, + "step": 4245 + }, + { + "epoch": 0.16, + "learning_rate": 4.7309566555527065e-05, + "loss": 0.5433, + "step": 4246 + }, + { + "epoch": 0.16, + "learning_rate": 4.730892536547833e-05, + "loss": 0.4775, + "step": 4247 + }, + { + "epoch": 0.16, + "learning_rate": 4.73082841754296e-05, + "loss": 0.6058, + "step": 4248 + }, + { + "epoch": 0.16, + "learning_rate": 4.730764298538087e-05, + "loss": 0.6015, + "step": 4249 + }, + { + "epoch": 0.16, + "learning_rate": 4.730700179533214e-05, + "loss": 0.5933, + "step": 4250 + }, + { + "epoch": 0.16, + "learning_rate": 4.7306360605283404e-05, + "loss": 0.6277, + "step": 4251 + }, + { + "epoch": 0.16, + "learning_rate": 4.7305719415234675e-05, + "loss": 0.682, + "step": 4252 + }, + { + "epoch": 0.16, + "learning_rate": 4.730507822518595e-05, + "loss": 0.5526, + "step": 4253 + }, + { + "epoch": 0.16, + "learning_rate": 4.730443703513722e-05, + "loss": 0.5945, + "step": 4254 + }, + { + "epoch": 0.16, + "learning_rate": 4.730379584508849e-05, + "loss": 0.5704, + "step": 4255 + }, + { + "epoch": 0.16, + "learning_rate": 4.7303154655039756e-05, + "loss": 0.6004, + "step": 4256 + }, + { + "epoch": 0.16, + "learning_rate": 4.730251346499103e-05, + "loss": 0.5936, + "step": 4257 + }, + { + "epoch": 0.16, + "learning_rate": 4.730187227494229e-05, + "loss": 0.4962, + "step": 4258 + }, + { + "epoch": 0.16, + "learning_rate": 4.7301231084893564e-05, + "loss": 0.5997, + "step": 4259 + }, + { + "epoch": 0.16, + "learning_rate": 4.7300589894844836e-05, + "loss": 0.511, + "step": 4260 + }, + { + "epoch": 0.16, + "learning_rate": 4.729994870479611e-05, + "loss": 0.6747, + "step": 4261 + }, + { + "epoch": 0.16, + "learning_rate": 4.729930751474737e-05, + "loss": 0.5489, + "step": 4262 + }, + { + "epoch": 0.16, + "learning_rate": 4.7298666324698644e-05, + "loss": 0.5643, + "step": 4263 + }, + { + "epoch": 0.16, + "learning_rate": 4.7298025134649916e-05, + "loss": 0.5665, + "step": 4264 + }, + { + "epoch": 0.16, + "learning_rate": 4.729738394460118e-05, + "loss": 0.5816, + "step": 4265 + }, + { + "epoch": 0.16, + "learning_rate": 4.729674275455245e-05, + "loss": 0.6546, + "step": 4266 + }, + { + "epoch": 0.16, + "learning_rate": 4.729610156450372e-05, + "loss": 0.7327, + "step": 4267 + }, + { + "epoch": 0.16, + "learning_rate": 4.729546037445499e-05, + "loss": 0.6058, + "step": 4268 + }, + { + "epoch": 0.16, + "learning_rate": 4.729481918440626e-05, + "loss": 0.6188, + "step": 4269 + }, + { + "epoch": 0.16, + "learning_rate": 4.729417799435753e-05, + "loss": 0.5511, + "step": 4270 + }, + { + "epoch": 0.16, + "learning_rate": 4.72935368043088e-05, + "loss": 0.6726, + "step": 4271 + }, + { + "epoch": 0.16, + "learning_rate": 4.729289561426007e-05, + "loss": 0.6898, + "step": 4272 + }, + { + "epoch": 0.16, + "learning_rate": 4.729225442421134e-05, + "loss": 0.609, + "step": 4273 + }, + { + "epoch": 0.16, + "learning_rate": 4.7291613234162606e-05, + "loss": 0.6202, + "step": 4274 + }, + { + "epoch": 0.16, + "learning_rate": 4.729097204411388e-05, + "loss": 0.5142, + "step": 4275 + }, + { + "epoch": 0.16, + "learning_rate": 4.729033085406514e-05, + "loss": 0.5956, + "step": 4276 + }, + { + "epoch": 0.16, + "learning_rate": 4.728968966401642e-05, + "loss": 0.6762, + "step": 4277 + }, + { + "epoch": 0.16, + "learning_rate": 4.7289048473967686e-05, + "loss": 0.5116, + "step": 4278 + }, + { + "epoch": 0.16, + "learning_rate": 4.728840728391896e-05, + "loss": 0.55, + "step": 4279 + }, + { + "epoch": 0.16, + "learning_rate": 4.728776609387022e-05, + "loss": 0.6122, + "step": 4280 + }, + { + "epoch": 0.16, + "learning_rate": 4.7287124903821494e-05, + "loss": 0.6114, + "step": 4281 + }, + { + "epoch": 0.16, + "learning_rate": 4.7286483713772766e-05, + "loss": 0.5619, + "step": 4282 + }, + { + "epoch": 0.16, + "learning_rate": 4.728584252372403e-05, + "loss": 0.6289, + "step": 4283 + }, + { + "epoch": 0.16, + "learning_rate": 4.72852013336753e-05, + "loss": 0.6274, + "step": 4284 + }, + { + "epoch": 0.16, + "learning_rate": 4.7284560143626574e-05, + "loss": 0.5807, + "step": 4285 + }, + { + "epoch": 0.16, + "learning_rate": 4.7283918953577846e-05, + "loss": 0.6433, + "step": 4286 + }, + { + "epoch": 0.16, + "learning_rate": 4.728327776352911e-05, + "loss": 0.5195, + "step": 4287 + }, + { + "epoch": 0.16, + "learning_rate": 4.728263657348038e-05, + "loss": 0.5658, + "step": 4288 + }, + { + "epoch": 0.16, + "learning_rate": 4.728199538343165e-05, + "loss": 0.5749, + "step": 4289 + }, + { + "epoch": 0.16, + "learning_rate": 4.728135419338292e-05, + "loss": 0.5844, + "step": 4290 + }, + { + "epoch": 0.16, + "learning_rate": 4.728071300333419e-05, + "loss": 0.5808, + "step": 4291 + }, + { + "epoch": 0.17, + "learning_rate": 4.7280071813285456e-05, + "loss": 0.6023, + "step": 4292 + }, + { + "epoch": 0.17, + "learning_rate": 4.7279430623236735e-05, + "loss": 0.6188, + "step": 4293 + }, + { + "epoch": 0.17, + "learning_rate": 4.7278789433188e-05, + "loss": 0.6388, + "step": 4294 + }, + { + "epoch": 0.17, + "learning_rate": 4.727814824313927e-05, + "loss": 0.6272, + "step": 4295 + }, + { + "epoch": 0.17, + "learning_rate": 4.7277507053090536e-05, + "loss": 0.6737, + "step": 4296 + }, + { + "epoch": 0.17, + "learning_rate": 4.727686586304181e-05, + "loss": 0.6318, + "step": 4297 + }, + { + "epoch": 0.17, + "learning_rate": 4.727622467299307e-05, + "loss": 0.559, + "step": 4298 + }, + { + "epoch": 0.17, + "learning_rate": 4.7275583482944344e-05, + "loss": 0.6313, + "step": 4299 + }, + { + "epoch": 0.17, + "learning_rate": 4.7274942292895616e-05, + "loss": 0.6255, + "step": 4300 + }, + { + "epoch": 0.17, + "learning_rate": 4.727430110284689e-05, + "loss": 0.6021, + "step": 4301 + }, + { + "epoch": 0.17, + "learning_rate": 4.727365991279816e-05, + "loss": 0.5523, + "step": 4302 + }, + { + "epoch": 0.17, + "learning_rate": 4.7273018722749425e-05, + "loss": 0.5353, + "step": 4303 + }, + { + "epoch": 0.17, + "learning_rate": 4.7272377532700696e-05, + "loss": 0.5184, + "step": 4304 + }, + { + "epoch": 0.17, + "learning_rate": 4.727173634265196e-05, + "loss": 0.6335, + "step": 4305 + }, + { + "epoch": 0.17, + "learning_rate": 4.727109515260323e-05, + "loss": 0.5952, + "step": 4306 + }, + { + "epoch": 0.17, + "learning_rate": 4.72704539625545e-05, + "loss": 0.5842, + "step": 4307 + }, + { + "epoch": 0.17, + "learning_rate": 4.726981277250577e-05, + "loss": 0.5088, + "step": 4308 + }, + { + "epoch": 0.17, + "learning_rate": 4.726917158245704e-05, + "loss": 0.6022, + "step": 4309 + }, + { + "epoch": 0.17, + "learning_rate": 4.726853039240831e-05, + "loss": 0.5433, + "step": 4310 + }, + { + "epoch": 0.17, + "learning_rate": 4.7267889202359585e-05, + "loss": 0.6969, + "step": 4311 + }, + { + "epoch": 0.17, + "learning_rate": 4.726724801231085e-05, + "loss": 0.5898, + "step": 4312 + }, + { + "epoch": 0.17, + "learning_rate": 4.726660682226212e-05, + "loss": 0.5819, + "step": 4313 + }, + { + "epoch": 0.17, + "learning_rate": 4.7265965632213386e-05, + "loss": 0.555, + "step": 4314 + }, + { + "epoch": 0.17, + "learning_rate": 4.726532444216466e-05, + "loss": 0.5732, + "step": 4315 + }, + { + "epoch": 0.17, + "learning_rate": 4.726468325211593e-05, + "loss": 0.5966, + "step": 4316 + }, + { + "epoch": 0.17, + "learning_rate": 4.72640420620672e-05, + "loss": 0.512, + "step": 4317 + }, + { + "epoch": 0.17, + "learning_rate": 4.726340087201847e-05, + "loss": 0.6257, + "step": 4318 + }, + { + "epoch": 0.17, + "learning_rate": 4.726275968196974e-05, + "loss": 0.6199, + "step": 4319 + }, + { + "epoch": 0.17, + "learning_rate": 4.726211849192101e-05, + "loss": 0.6007, + "step": 4320 + }, + { + "epoch": 0.17, + "learning_rate": 4.7261477301872275e-05, + "loss": 0.5328, + "step": 4321 + }, + { + "epoch": 0.17, + "learning_rate": 4.7260836111823547e-05, + "loss": 0.5671, + "step": 4322 + }, + { + "epoch": 0.17, + "learning_rate": 4.726019492177481e-05, + "loss": 0.5665, + "step": 4323 + }, + { + "epoch": 0.17, + "learning_rate": 4.725955373172608e-05, + "loss": 0.6894, + "step": 4324 + }, + { + "epoch": 0.17, + "learning_rate": 4.7258912541677355e-05, + "loss": 0.6036, + "step": 4325 + }, + { + "epoch": 0.17, + "learning_rate": 4.725827135162863e-05, + "loss": 0.6526, + "step": 4326 + }, + { + "epoch": 0.17, + "learning_rate": 4.72576301615799e-05, + "loss": 0.6847, + "step": 4327 + }, + { + "epoch": 0.17, + "learning_rate": 4.725698897153116e-05, + "loss": 0.5221, + "step": 4328 + }, + { + "epoch": 0.17, + "learning_rate": 4.7256347781482435e-05, + "loss": 0.5431, + "step": 4329 + }, + { + "epoch": 0.17, + "learning_rate": 4.72557065914337e-05, + "loss": 0.5796, + "step": 4330 + }, + { + "epoch": 0.17, + "learning_rate": 4.725506540138497e-05, + "loss": 0.6393, + "step": 4331 + }, + { + "epoch": 0.17, + "learning_rate": 4.725442421133624e-05, + "loss": 0.6054, + "step": 4332 + }, + { + "epoch": 0.17, + "learning_rate": 4.7253783021287515e-05, + "loss": 0.6018, + "step": 4333 + }, + { + "epoch": 0.17, + "learning_rate": 4.725314183123879e-05, + "loss": 0.6504, + "step": 4334 + }, + { + "epoch": 0.17, + "learning_rate": 4.725250064119005e-05, + "loss": 0.5705, + "step": 4335 + }, + { + "epoch": 0.17, + "learning_rate": 4.7251859451141323e-05, + "loss": 0.6197, + "step": 4336 + }, + { + "epoch": 0.17, + "learning_rate": 4.725121826109259e-05, + "loss": 0.5678, + "step": 4337 + }, + { + "epoch": 0.17, + "learning_rate": 4.725057707104386e-05, + "loss": 0.6352, + "step": 4338 + }, + { + "epoch": 0.17, + "learning_rate": 4.7249935880995125e-05, + "loss": 0.6888, + "step": 4339 + }, + { + "epoch": 0.17, + "learning_rate": 4.72492946909464e-05, + "loss": 0.6692, + "step": 4340 + }, + { + "epoch": 0.17, + "learning_rate": 4.724865350089767e-05, + "loss": 0.667, + "step": 4341 + }, + { + "epoch": 0.17, + "learning_rate": 4.724801231084894e-05, + "loss": 0.5836, + "step": 4342 + }, + { + "epoch": 0.17, + "learning_rate": 4.724737112080021e-05, + "loss": 0.6215, + "step": 4343 + }, + { + "epoch": 0.17, + "learning_rate": 4.724672993075148e-05, + "loss": 0.5184, + "step": 4344 + }, + { + "epoch": 0.17, + "learning_rate": 4.724608874070275e-05, + "loss": 0.5639, + "step": 4345 + }, + { + "epoch": 0.17, + "learning_rate": 4.7245447550654014e-05, + "loss": 0.6161, + "step": 4346 + }, + { + "epoch": 0.17, + "learning_rate": 4.7244806360605285e-05, + "loss": 0.6319, + "step": 4347 + }, + { + "epoch": 0.17, + "learning_rate": 4.724416517055655e-05, + "loss": 0.584, + "step": 4348 + }, + { + "epoch": 0.17, + "learning_rate": 4.724352398050783e-05, + "loss": 0.631, + "step": 4349 + }, + { + "epoch": 0.17, + "learning_rate": 4.7242882790459094e-05, + "loss": 0.5528, + "step": 4350 + }, + { + "epoch": 0.17, + "learning_rate": 4.7242241600410365e-05, + "loss": 0.5391, + "step": 4351 + }, + { + "epoch": 0.17, + "learning_rate": 4.724160041036164e-05, + "loss": 0.553, + "step": 4352 + }, + { + "epoch": 0.17, + "learning_rate": 4.72409592203129e-05, + "loss": 0.6193, + "step": 4353 + }, + { + "epoch": 0.17, + "learning_rate": 4.7240318030264174e-05, + "loss": 0.6741, + "step": 4354 + }, + { + "epoch": 0.17, + "learning_rate": 4.723967684021544e-05, + "loss": 0.5969, + "step": 4355 + }, + { + "epoch": 0.17, + "learning_rate": 4.723903565016671e-05, + "loss": 0.5991, + "step": 4356 + }, + { + "epoch": 0.17, + "learning_rate": 4.723839446011798e-05, + "loss": 0.5992, + "step": 4357 + }, + { + "epoch": 0.17, + "learning_rate": 4.7237753270069254e-05, + "loss": 0.5096, + "step": 4358 + }, + { + "epoch": 0.17, + "learning_rate": 4.723711208002052e-05, + "loss": 0.5208, + "step": 4359 + }, + { + "epoch": 0.17, + "learning_rate": 4.723647088997179e-05, + "loss": 0.5423, + "step": 4360 + }, + { + "epoch": 0.17, + "learning_rate": 4.723582969992306e-05, + "loss": 0.6968, + "step": 4361 + }, + { + "epoch": 0.17, + "learning_rate": 4.723518850987433e-05, + "loss": 0.5837, + "step": 4362 + }, + { + "epoch": 0.17, + "learning_rate": 4.72345473198256e-05, + "loss": 0.5838, + "step": 4363 + }, + { + "epoch": 0.17, + "learning_rate": 4.7233906129776864e-05, + "loss": 0.6247, + "step": 4364 + }, + { + "epoch": 0.17, + "learning_rate": 4.723326493972814e-05, + "loss": 0.603, + "step": 4365 + }, + { + "epoch": 0.17, + "learning_rate": 4.723262374967941e-05, + "loss": 0.6424, + "step": 4366 + }, + { + "epoch": 0.17, + "learning_rate": 4.723198255963068e-05, + "loss": 0.4916, + "step": 4367 + }, + { + "epoch": 0.17, + "learning_rate": 4.7231341369581944e-05, + "loss": 0.6064, + "step": 4368 + }, + { + "epoch": 0.17, + "learning_rate": 4.7230700179533216e-05, + "loss": 0.5808, + "step": 4369 + }, + { + "epoch": 0.17, + "learning_rate": 4.723005898948449e-05, + "loss": 0.5807, + "step": 4370 + }, + { + "epoch": 0.17, + "learning_rate": 4.722941779943575e-05, + "loss": 0.5025, + "step": 4371 + }, + { + "epoch": 0.17, + "learning_rate": 4.7228776609387024e-05, + "loss": 0.6418, + "step": 4372 + }, + { + "epoch": 0.17, + "learning_rate": 4.7228135419338296e-05, + "loss": 0.6518, + "step": 4373 + }, + { + "epoch": 0.17, + "learning_rate": 4.722749422928957e-05, + "loss": 0.655, + "step": 4374 + }, + { + "epoch": 0.17, + "learning_rate": 4.722685303924083e-05, + "loss": 0.5849, + "step": 4375 + }, + { + "epoch": 0.17, + "learning_rate": 4.7226211849192104e-05, + "loss": 0.6447, + "step": 4376 + }, + { + "epoch": 0.17, + "learning_rate": 4.722557065914337e-05, + "loss": 0.6257, + "step": 4377 + }, + { + "epoch": 0.17, + "learning_rate": 4.722492946909464e-05, + "loss": 0.6365, + "step": 4378 + }, + { + "epoch": 0.17, + "learning_rate": 4.722428827904591e-05, + "loss": 0.6047, + "step": 4379 + }, + { + "epoch": 0.17, + "learning_rate": 4.722364708899718e-05, + "loss": 0.56, + "step": 4380 + }, + { + "epoch": 0.17, + "learning_rate": 4.7223005898948456e-05, + "loss": 0.5992, + "step": 4381 + }, + { + "epoch": 0.17, + "learning_rate": 4.722236470889972e-05, + "loss": 0.5559, + "step": 4382 + }, + { + "epoch": 0.17, + "learning_rate": 4.722172351885099e-05, + "loss": 0.5763, + "step": 4383 + }, + { + "epoch": 0.17, + "learning_rate": 4.722108232880226e-05, + "loss": 0.6112, + "step": 4384 + }, + { + "epoch": 0.17, + "learning_rate": 4.722044113875353e-05, + "loss": 0.586, + "step": 4385 + }, + { + "epoch": 0.17, + "learning_rate": 4.7219799948704794e-05, + "loss": 0.55, + "step": 4386 + }, + { + "epoch": 0.17, + "learning_rate": 4.7219158758656066e-05, + "loss": 0.5546, + "step": 4387 + }, + { + "epoch": 0.17, + "learning_rate": 4.721851756860734e-05, + "loss": 0.6137, + "step": 4388 + }, + { + "epoch": 0.17, + "learning_rate": 4.721787637855861e-05, + "loss": 0.6472, + "step": 4389 + }, + { + "epoch": 0.17, + "learning_rate": 4.721723518850988e-05, + "loss": 0.5558, + "step": 4390 + }, + { + "epoch": 0.17, + "learning_rate": 4.7216593998461146e-05, + "loss": 0.574, + "step": 4391 + }, + { + "epoch": 0.17, + "learning_rate": 4.721595280841242e-05, + "loss": 0.6059, + "step": 4392 + }, + { + "epoch": 0.17, + "learning_rate": 4.721531161836368e-05, + "loss": 0.6378, + "step": 4393 + }, + { + "epoch": 0.17, + "learning_rate": 4.7214670428314954e-05, + "loss": 0.5148, + "step": 4394 + }, + { + "epoch": 0.17, + "learning_rate": 4.721402923826622e-05, + "loss": 0.6649, + "step": 4395 + }, + { + "epoch": 0.17, + "learning_rate": 4.721338804821749e-05, + "loss": 0.5216, + "step": 4396 + }, + { + "epoch": 0.17, + "learning_rate": 4.721274685816876e-05, + "loss": 0.676, + "step": 4397 + }, + { + "epoch": 0.17, + "learning_rate": 4.7212105668120034e-05, + "loss": 0.5587, + "step": 4398 + }, + { + "epoch": 0.17, + "learning_rate": 4.7211464478071306e-05, + "loss": 0.5982, + "step": 4399 + }, + { + "epoch": 0.17, + "learning_rate": 4.721082328802257e-05, + "loss": 0.5004, + "step": 4400 + }, + { + "epoch": 0.17, + "learning_rate": 4.721018209797384e-05, + "loss": 0.6243, + "step": 4401 + }, + { + "epoch": 0.17, + "learning_rate": 4.720954090792511e-05, + "loss": 0.5149, + "step": 4402 + }, + { + "epoch": 0.17, + "learning_rate": 4.720889971787638e-05, + "loss": 0.5144, + "step": 4403 + }, + { + "epoch": 0.17, + "learning_rate": 4.720825852782765e-05, + "loss": 0.5816, + "step": 4404 + }, + { + "epoch": 0.17, + "learning_rate": 4.720761733777892e-05, + "loss": 0.546, + "step": 4405 + }, + { + "epoch": 0.17, + "learning_rate": 4.7206976147730195e-05, + "loss": 0.6192, + "step": 4406 + }, + { + "epoch": 0.17, + "learning_rate": 4.720633495768146e-05, + "loss": 0.6243, + "step": 4407 + }, + { + "epoch": 0.17, + "learning_rate": 4.720569376763273e-05, + "loss": 0.54, + "step": 4408 + }, + { + "epoch": 0.17, + "learning_rate": 4.7205052577583996e-05, + "loss": 0.6031, + "step": 4409 + }, + { + "epoch": 0.17, + "learning_rate": 4.720441138753527e-05, + "loss": 0.6214, + "step": 4410 + }, + { + "epoch": 0.17, + "learning_rate": 4.720377019748653e-05, + "loss": 0.6286, + "step": 4411 + }, + { + "epoch": 0.17, + "learning_rate": 4.7203129007437805e-05, + "loss": 0.516, + "step": 4412 + }, + { + "epoch": 0.17, + "learning_rate": 4.7202487817389076e-05, + "loss": 0.6451, + "step": 4413 + }, + { + "epoch": 0.17, + "learning_rate": 4.720184662734035e-05, + "loss": 0.486, + "step": 4414 + }, + { + "epoch": 0.17, + "learning_rate": 4.720120543729162e-05, + "loss": 0.6477, + "step": 4415 + }, + { + "epoch": 0.17, + "learning_rate": 4.7200564247242885e-05, + "loss": 0.6287, + "step": 4416 + }, + { + "epoch": 0.17, + "learning_rate": 4.7199923057194156e-05, + "loss": 0.5775, + "step": 4417 + }, + { + "epoch": 0.17, + "learning_rate": 4.719928186714542e-05, + "loss": 0.6089, + "step": 4418 + }, + { + "epoch": 0.17, + "learning_rate": 4.719864067709669e-05, + "loss": 0.5793, + "step": 4419 + }, + { + "epoch": 0.17, + "learning_rate": 4.7197999487047965e-05, + "loss": 0.5389, + "step": 4420 + }, + { + "epoch": 0.17, + "learning_rate": 4.7197358296999236e-05, + "loss": 0.6258, + "step": 4421 + }, + { + "epoch": 0.17, + "learning_rate": 4.71967171069505e-05, + "loss": 0.6146, + "step": 4422 + }, + { + "epoch": 0.17, + "learning_rate": 4.719607591690177e-05, + "loss": 0.632, + "step": 4423 + }, + { + "epoch": 0.17, + "learning_rate": 4.7195434726853045e-05, + "loss": 0.7046, + "step": 4424 + }, + { + "epoch": 0.17, + "learning_rate": 4.719479353680431e-05, + "loss": 0.5813, + "step": 4425 + }, + { + "epoch": 0.17, + "learning_rate": 4.719415234675558e-05, + "loss": 0.6155, + "step": 4426 + }, + { + "epoch": 0.17, + "learning_rate": 4.7193511156706846e-05, + "loss": 0.675, + "step": 4427 + }, + { + "epoch": 0.17, + "learning_rate": 4.719286996665812e-05, + "loss": 0.6442, + "step": 4428 + }, + { + "epoch": 0.17, + "learning_rate": 4.719222877660939e-05, + "loss": 0.5733, + "step": 4429 + }, + { + "epoch": 0.17, + "learning_rate": 4.719158758656066e-05, + "loss": 0.5476, + "step": 4430 + }, + { + "epoch": 0.17, + "learning_rate": 4.7190946396511926e-05, + "loss": 0.6014, + "step": 4431 + }, + { + "epoch": 0.17, + "learning_rate": 4.71903052064632e-05, + "loss": 0.6407, + "step": 4432 + }, + { + "epoch": 0.17, + "learning_rate": 4.718966401641447e-05, + "loss": 0.636, + "step": 4433 + }, + { + "epoch": 0.17, + "learning_rate": 4.7189022826365735e-05, + "loss": 0.5967, + "step": 4434 + }, + { + "epoch": 0.17, + "learning_rate": 4.7188381636317007e-05, + "loss": 0.6432, + "step": 4435 + }, + { + "epoch": 0.17, + "learning_rate": 4.718774044626827e-05, + "loss": 0.6058, + "step": 4436 + }, + { + "epoch": 0.17, + "learning_rate": 4.718709925621955e-05, + "loss": 0.6089, + "step": 4437 + }, + { + "epoch": 0.17, + "learning_rate": 4.7186458066170815e-05, + "loss": 0.5694, + "step": 4438 + }, + { + "epoch": 0.17, + "learning_rate": 4.718581687612209e-05, + "loss": 0.652, + "step": 4439 + }, + { + "epoch": 0.17, + "learning_rate": 4.718517568607335e-05, + "loss": 0.5875, + "step": 4440 + }, + { + "epoch": 0.17, + "learning_rate": 4.718453449602462e-05, + "loss": 0.5649, + "step": 4441 + }, + { + "epoch": 0.17, + "learning_rate": 4.7183893305975895e-05, + "loss": 0.5983, + "step": 4442 + }, + { + "epoch": 0.17, + "learning_rate": 4.718325211592716e-05, + "loss": 0.5683, + "step": 4443 + }, + { + "epoch": 0.17, + "learning_rate": 4.718261092587843e-05, + "loss": 0.626, + "step": 4444 + }, + { + "epoch": 0.17, + "learning_rate": 4.7181969735829703e-05, + "loss": 0.6125, + "step": 4445 + }, + { + "epoch": 0.17, + "learning_rate": 4.7181328545780975e-05, + "loss": 0.6418, + "step": 4446 + }, + { + "epoch": 0.17, + "learning_rate": 4.718068735573224e-05, + "loss": 0.5969, + "step": 4447 + }, + { + "epoch": 0.17, + "learning_rate": 4.718004616568351e-05, + "loss": 0.5605, + "step": 4448 + }, + { + "epoch": 0.17, + "learning_rate": 4.717940497563478e-05, + "loss": 0.5404, + "step": 4449 + }, + { + "epoch": 0.17, + "learning_rate": 4.717876378558605e-05, + "loss": 0.6189, + "step": 4450 + }, + { + "epoch": 0.17, + "learning_rate": 4.717812259553732e-05, + "loss": 0.6192, + "step": 4451 + }, + { + "epoch": 0.17, + "learning_rate": 4.7177481405488585e-05, + "loss": 0.6922, + "step": 4452 + }, + { + "epoch": 0.17, + "learning_rate": 4.7176840215439864e-05, + "loss": 0.6333, + "step": 4453 + }, + { + "epoch": 0.17, + "learning_rate": 4.717619902539113e-05, + "loss": 0.5805, + "step": 4454 + }, + { + "epoch": 0.17, + "learning_rate": 4.71755578353424e-05, + "loss": 0.6553, + "step": 4455 + }, + { + "epoch": 0.17, + "learning_rate": 4.7174916645293665e-05, + "loss": 0.5568, + "step": 4456 + }, + { + "epoch": 0.17, + "learning_rate": 4.717427545524494e-05, + "loss": 0.5854, + "step": 4457 + }, + { + "epoch": 0.17, + "learning_rate": 4.71736342651962e-05, + "loss": 0.6353, + "step": 4458 + }, + { + "epoch": 0.17, + "learning_rate": 4.7172993075147474e-05, + "loss": 0.5605, + "step": 4459 + }, + { + "epoch": 0.17, + "learning_rate": 4.7172351885098745e-05, + "loss": 0.5943, + "step": 4460 + }, + { + "epoch": 0.17, + "learning_rate": 4.717171069505002e-05, + "loss": 0.5739, + "step": 4461 + }, + { + "epoch": 0.17, + "learning_rate": 4.717106950500129e-05, + "loss": 0.6443, + "step": 4462 + }, + { + "epoch": 0.17, + "learning_rate": 4.7170428314952554e-05, + "loss": 0.6452, + "step": 4463 + }, + { + "epoch": 0.17, + "learning_rate": 4.7169787124903825e-05, + "loss": 0.6716, + "step": 4464 + }, + { + "epoch": 0.17, + "learning_rate": 4.716914593485509e-05, + "loss": 0.5987, + "step": 4465 + }, + { + "epoch": 0.17, + "learning_rate": 4.716850474480636e-05, + "loss": 0.5885, + "step": 4466 + }, + { + "epoch": 0.17, + "learning_rate": 4.716786355475763e-05, + "loss": 0.6257, + "step": 4467 + }, + { + "epoch": 0.17, + "learning_rate": 4.71672223647089e-05, + "loss": 0.7254, + "step": 4468 + }, + { + "epoch": 0.17, + "learning_rate": 4.716658117466017e-05, + "loss": 0.6247, + "step": 4469 + }, + { + "epoch": 0.17, + "learning_rate": 4.716593998461144e-05, + "loss": 0.6141, + "step": 4470 + }, + { + "epoch": 0.17, + "learning_rate": 4.7165298794562714e-05, + "loss": 0.6759, + "step": 4471 + }, + { + "epoch": 0.17, + "learning_rate": 4.716465760451398e-05, + "loss": 0.6168, + "step": 4472 + }, + { + "epoch": 0.17, + "learning_rate": 4.716401641446525e-05, + "loss": 0.6348, + "step": 4473 + }, + { + "epoch": 0.17, + "learning_rate": 4.7163375224416515e-05, + "loss": 0.6005, + "step": 4474 + }, + { + "epoch": 0.17, + "learning_rate": 4.716273403436779e-05, + "loss": 0.5263, + "step": 4475 + }, + { + "epoch": 0.17, + "learning_rate": 4.716209284431906e-05, + "loss": 0.6194, + "step": 4476 + }, + { + "epoch": 0.17, + "learning_rate": 4.716145165427033e-05, + "loss": 0.6562, + "step": 4477 + }, + { + "epoch": 0.17, + "learning_rate": 4.71608104642216e-05, + "loss": 0.6216, + "step": 4478 + }, + { + "epoch": 0.17, + "learning_rate": 4.716016927417287e-05, + "loss": 0.4861, + "step": 4479 + }, + { + "epoch": 0.17, + "learning_rate": 4.715952808412414e-05, + "loss": 0.6238, + "step": 4480 + }, + { + "epoch": 0.17, + "learning_rate": 4.7158886894075404e-05, + "loss": 0.5524, + "step": 4481 + }, + { + "epoch": 0.17, + "learning_rate": 4.7158245704026676e-05, + "loss": 0.4684, + "step": 4482 + }, + { + "epoch": 0.17, + "learning_rate": 4.715760451397794e-05, + "loss": 0.597, + "step": 4483 + }, + { + "epoch": 0.17, + "learning_rate": 4.715696332392921e-05, + "loss": 0.5535, + "step": 4484 + }, + { + "epoch": 0.17, + "learning_rate": 4.7156322133880484e-05, + "loss": 0.5654, + "step": 4485 + }, + { + "epoch": 0.17, + "learning_rate": 4.7155680943831756e-05, + "loss": 0.6848, + "step": 4486 + }, + { + "epoch": 0.17, + "learning_rate": 4.715503975378303e-05, + "loss": 0.5146, + "step": 4487 + }, + { + "epoch": 0.17, + "learning_rate": 4.715439856373429e-05, + "loss": 0.5172, + "step": 4488 + }, + { + "epoch": 0.17, + "learning_rate": 4.7153757373685564e-05, + "loss": 0.6836, + "step": 4489 + }, + { + "epoch": 0.17, + "learning_rate": 4.715311618363683e-05, + "loss": 0.5864, + "step": 4490 + }, + { + "epoch": 0.17, + "learning_rate": 4.71524749935881e-05, + "loss": 0.6238, + "step": 4491 + }, + { + "epoch": 0.17, + "learning_rate": 4.715183380353937e-05, + "loss": 0.6605, + "step": 4492 + }, + { + "epoch": 0.17, + "learning_rate": 4.7151192613490644e-05, + "loss": 0.525, + "step": 4493 + }, + { + "epoch": 0.17, + "learning_rate": 4.715055142344191e-05, + "loss": 0.6957, + "step": 4494 + }, + { + "epoch": 0.17, + "learning_rate": 4.714991023339318e-05, + "loss": 0.5468, + "step": 4495 + }, + { + "epoch": 0.17, + "learning_rate": 4.714926904334445e-05, + "loss": 0.5329, + "step": 4496 + }, + { + "epoch": 0.17, + "learning_rate": 4.714862785329572e-05, + "loss": 0.6059, + "step": 4497 + }, + { + "epoch": 0.17, + "learning_rate": 4.714798666324699e-05, + "loss": 0.5616, + "step": 4498 + }, + { + "epoch": 0.17, + "learning_rate": 4.7147345473198254e-05, + "loss": 0.6077, + "step": 4499 + }, + { + "epoch": 0.17, + "learning_rate": 4.7146704283149526e-05, + "loss": 0.6465, + "step": 4500 + }, + { + "epoch": 0.17, + "learning_rate": 4.71460630931008e-05, + "loss": 0.6803, + "step": 4501 + }, + { + "epoch": 0.17, + "learning_rate": 4.714542190305207e-05, + "loss": 0.6392, + "step": 4502 + }, + { + "epoch": 0.17, + "learning_rate": 4.714478071300334e-05, + "loss": 0.5412, + "step": 4503 + }, + { + "epoch": 0.17, + "learning_rate": 4.7144139522954606e-05, + "loss": 0.5646, + "step": 4504 + }, + { + "epoch": 0.17, + "learning_rate": 4.714349833290588e-05, + "loss": 0.7062, + "step": 4505 + }, + { + "epoch": 0.17, + "learning_rate": 4.714285714285714e-05, + "loss": 0.6141, + "step": 4506 + }, + { + "epoch": 0.17, + "learning_rate": 4.7142215952808414e-05, + "loss": 0.5637, + "step": 4507 + }, + { + "epoch": 0.17, + "learning_rate": 4.7141574762759686e-05, + "loss": 0.5713, + "step": 4508 + }, + { + "epoch": 0.17, + "learning_rate": 4.714093357271096e-05, + "loss": 0.6184, + "step": 4509 + }, + { + "epoch": 0.17, + "learning_rate": 4.714029238266222e-05, + "loss": 0.5896, + "step": 4510 + }, + { + "epoch": 0.17, + "learning_rate": 4.7139651192613494e-05, + "loss": 0.5266, + "step": 4511 + }, + { + "epoch": 0.17, + "learning_rate": 4.7139010002564766e-05, + "loss": 0.579, + "step": 4512 + }, + { + "epoch": 0.17, + "learning_rate": 4.713836881251603e-05, + "loss": 0.5981, + "step": 4513 + }, + { + "epoch": 0.17, + "learning_rate": 4.71377276224673e-05, + "loss": 0.621, + "step": 4514 + }, + { + "epoch": 0.17, + "learning_rate": 4.713708643241857e-05, + "loss": 0.5521, + "step": 4515 + }, + { + "epoch": 0.17, + "learning_rate": 4.713644524236984e-05, + "loss": 0.5932, + "step": 4516 + }, + { + "epoch": 0.17, + "learning_rate": 4.713580405232111e-05, + "loss": 0.5831, + "step": 4517 + }, + { + "epoch": 0.17, + "learning_rate": 4.713516286227238e-05, + "loss": 0.5828, + "step": 4518 + }, + { + "epoch": 0.17, + "learning_rate": 4.713452167222365e-05, + "loss": 0.5041, + "step": 4519 + }, + { + "epoch": 0.17, + "learning_rate": 4.713388048217492e-05, + "loss": 0.5171, + "step": 4520 + }, + { + "epoch": 0.17, + "learning_rate": 4.713323929212619e-05, + "loss": 0.6281, + "step": 4521 + }, + { + "epoch": 0.17, + "learning_rate": 4.7132598102077456e-05, + "loss": 0.6741, + "step": 4522 + }, + { + "epoch": 0.17, + "learning_rate": 4.713195691202873e-05, + "loss": 0.5554, + "step": 4523 + }, + { + "epoch": 0.17, + "learning_rate": 4.713131572197999e-05, + "loss": 0.6356, + "step": 4524 + }, + { + "epoch": 0.17, + "learning_rate": 4.713067453193127e-05, + "loss": 0.6057, + "step": 4525 + }, + { + "epoch": 0.17, + "learning_rate": 4.7130033341882536e-05, + "loss": 0.649, + "step": 4526 + }, + { + "epoch": 0.17, + "learning_rate": 4.712939215183381e-05, + "loss": 0.5724, + "step": 4527 + }, + { + "epoch": 0.17, + "learning_rate": 4.712875096178507e-05, + "loss": 0.5407, + "step": 4528 + }, + { + "epoch": 0.17, + "learning_rate": 4.7128109771736345e-05, + "loss": 0.5135, + "step": 4529 + }, + { + "epoch": 0.17, + "learning_rate": 4.7127468581687616e-05, + "loss": 0.5733, + "step": 4530 + }, + { + "epoch": 0.17, + "learning_rate": 4.712682739163888e-05, + "loss": 0.5076, + "step": 4531 + }, + { + "epoch": 0.17, + "learning_rate": 4.712618620159015e-05, + "loss": 0.6183, + "step": 4532 + }, + { + "epoch": 0.17, + "learning_rate": 4.7125545011541425e-05, + "loss": 0.6041, + "step": 4533 + }, + { + "epoch": 0.17, + "learning_rate": 4.7124903821492696e-05, + "loss": 0.5179, + "step": 4534 + }, + { + "epoch": 0.17, + "learning_rate": 4.712426263144396e-05, + "loss": 0.6142, + "step": 4535 + }, + { + "epoch": 0.17, + "learning_rate": 4.712362144139523e-05, + "loss": 0.5965, + "step": 4536 + }, + { + "epoch": 0.17, + "learning_rate": 4.71229802513465e-05, + "loss": 0.6869, + "step": 4537 + }, + { + "epoch": 0.17, + "learning_rate": 4.712233906129777e-05, + "loss": 0.5551, + "step": 4538 + }, + { + "epoch": 0.17, + "learning_rate": 4.712169787124904e-05, + "loss": 0.6074, + "step": 4539 + }, + { + "epoch": 0.17, + "learning_rate": 4.7121056681200306e-05, + "loss": 0.5822, + "step": 4540 + }, + { + "epoch": 0.17, + "learning_rate": 4.7120415491151585e-05, + "loss": 0.5924, + "step": 4541 + }, + { + "epoch": 0.17, + "learning_rate": 4.711977430110285e-05, + "loss": 0.6158, + "step": 4542 + }, + { + "epoch": 0.17, + "learning_rate": 4.711913311105412e-05, + "loss": 0.5584, + "step": 4543 + }, + { + "epoch": 0.17, + "learning_rate": 4.7118491921005387e-05, + "loss": 0.5599, + "step": 4544 + }, + { + "epoch": 0.17, + "learning_rate": 4.711785073095666e-05, + "loss": 0.5744, + "step": 4545 + }, + { + "epoch": 0.17, + "learning_rate": 4.711720954090792e-05, + "loss": 0.596, + "step": 4546 + }, + { + "epoch": 0.17, + "learning_rate": 4.7116568350859195e-05, + "loss": 0.63, + "step": 4547 + }, + { + "epoch": 0.17, + "learning_rate": 4.7115927160810467e-05, + "loss": 0.5674, + "step": 4548 + }, + { + "epoch": 0.17, + "learning_rate": 4.711528597076174e-05, + "loss": 0.6929, + "step": 4549 + }, + { + "epoch": 0.17, + "learning_rate": 4.711464478071301e-05, + "loss": 0.6787, + "step": 4550 + }, + { + "epoch": 0.17, + "learning_rate": 4.7114003590664275e-05, + "loss": 0.6127, + "step": 4551 + }, + { + "epoch": 0.18, + "learning_rate": 4.711336240061555e-05, + "loss": 0.5965, + "step": 4552 + }, + { + "epoch": 0.18, + "learning_rate": 4.711272121056681e-05, + "loss": 0.5857, + "step": 4553 + }, + { + "epoch": 0.18, + "learning_rate": 4.711208002051808e-05, + "loss": 0.5624, + "step": 4554 + }, + { + "epoch": 0.18, + "learning_rate": 4.711143883046935e-05, + "loss": 0.5467, + "step": 4555 + }, + { + "epoch": 0.18, + "learning_rate": 4.711079764042062e-05, + "loss": 0.5811, + "step": 4556 + }, + { + "epoch": 0.18, + "learning_rate": 4.711015645037189e-05, + "loss": 0.6517, + "step": 4557 + }, + { + "epoch": 0.18, + "learning_rate": 4.7109515260323163e-05, + "loss": 0.5305, + "step": 4558 + }, + { + "epoch": 0.18, + "learning_rate": 4.7108874070274435e-05, + "loss": 0.6101, + "step": 4559 + }, + { + "epoch": 0.18, + "learning_rate": 4.71082328802257e-05, + "loss": 0.5271, + "step": 4560 + }, + { + "epoch": 0.18, + "learning_rate": 4.710759169017697e-05, + "loss": 0.5738, + "step": 4561 + }, + { + "epoch": 0.18, + "learning_rate": 4.710695050012824e-05, + "loss": 0.5694, + "step": 4562 + }, + { + "epoch": 0.18, + "learning_rate": 4.710630931007951e-05, + "loss": 0.6748, + "step": 4563 + }, + { + "epoch": 0.18, + "learning_rate": 4.710566812003078e-05, + "loss": 0.5259, + "step": 4564 + }, + { + "epoch": 0.18, + "learning_rate": 4.710502692998205e-05, + "loss": 0.487, + "step": 4565 + }, + { + "epoch": 0.18, + "learning_rate": 4.7104385739933324e-05, + "loss": 0.7207, + "step": 4566 + }, + { + "epoch": 0.18, + "learning_rate": 4.710374454988459e-05, + "loss": 0.6225, + "step": 4567 + }, + { + "epoch": 0.18, + "learning_rate": 4.710310335983586e-05, + "loss": 0.5753, + "step": 4568 + }, + { + "epoch": 0.18, + "learning_rate": 4.7102462169787125e-05, + "loss": 0.6096, + "step": 4569 + }, + { + "epoch": 0.18, + "learning_rate": 4.71018209797384e-05, + "loss": 0.575, + "step": 4570 + }, + { + "epoch": 0.18, + "learning_rate": 4.710117978968966e-05, + "loss": 0.612, + "step": 4571 + }, + { + "epoch": 0.18, + "learning_rate": 4.7100538599640934e-05, + "loss": 0.5788, + "step": 4572 + }, + { + "epoch": 0.18, + "learning_rate": 4.7099897409592205e-05, + "loss": 0.5608, + "step": 4573 + }, + { + "epoch": 0.18, + "learning_rate": 4.709925621954348e-05, + "loss": 0.648, + "step": 4574 + }, + { + "epoch": 0.18, + "learning_rate": 4.709861502949475e-05, + "loss": 0.5482, + "step": 4575 + }, + { + "epoch": 0.18, + "learning_rate": 4.7097973839446014e-05, + "loss": 0.6526, + "step": 4576 + }, + { + "epoch": 0.18, + "learning_rate": 4.7097332649397285e-05, + "loss": 0.5263, + "step": 4577 + }, + { + "epoch": 0.18, + "learning_rate": 4.709669145934855e-05, + "loss": 0.6628, + "step": 4578 + }, + { + "epoch": 0.18, + "learning_rate": 4.709605026929982e-05, + "loss": 0.6053, + "step": 4579 + }, + { + "epoch": 0.18, + "learning_rate": 4.7095409079251094e-05, + "loss": 0.5606, + "step": 4580 + }, + { + "epoch": 0.18, + "learning_rate": 4.7094767889202365e-05, + "loss": 0.6684, + "step": 4581 + }, + { + "epoch": 0.18, + "learning_rate": 4.709412669915363e-05, + "loss": 0.6127, + "step": 4582 + }, + { + "epoch": 0.18, + "learning_rate": 4.70934855091049e-05, + "loss": 0.5833, + "step": 4583 + }, + { + "epoch": 0.18, + "learning_rate": 4.7092844319056174e-05, + "loss": 0.5706, + "step": 4584 + }, + { + "epoch": 0.18, + "learning_rate": 4.709220312900744e-05, + "loss": 0.5664, + "step": 4585 + }, + { + "epoch": 0.18, + "learning_rate": 4.709156193895871e-05, + "loss": 0.6148, + "step": 4586 + }, + { + "epoch": 0.18, + "learning_rate": 4.7090920748909975e-05, + "loss": 0.5983, + "step": 4587 + }, + { + "epoch": 0.18, + "learning_rate": 4.709027955886125e-05, + "loss": 0.4693, + "step": 4588 + }, + { + "epoch": 0.18, + "learning_rate": 4.708963836881252e-05, + "loss": 0.5842, + "step": 4589 + }, + { + "epoch": 0.18, + "learning_rate": 4.708899717876379e-05, + "loss": 0.6348, + "step": 4590 + }, + { + "epoch": 0.18, + "learning_rate": 4.7088355988715056e-05, + "loss": 0.4883, + "step": 4591 + }, + { + "epoch": 0.18, + "learning_rate": 4.708771479866633e-05, + "loss": 0.5917, + "step": 4592 + }, + { + "epoch": 0.18, + "learning_rate": 4.70870736086176e-05, + "loss": 0.5566, + "step": 4593 + }, + { + "epoch": 0.18, + "learning_rate": 4.7086432418568864e-05, + "loss": 0.6369, + "step": 4594 + }, + { + "epoch": 0.18, + "learning_rate": 4.7085791228520136e-05, + "loss": 0.5627, + "step": 4595 + }, + { + "epoch": 0.18, + "learning_rate": 4.70851500384714e-05, + "loss": 0.6069, + "step": 4596 + }, + { + "epoch": 0.18, + "learning_rate": 4.708450884842268e-05, + "loss": 0.5539, + "step": 4597 + }, + { + "epoch": 0.18, + "learning_rate": 4.7083867658373944e-05, + "loss": 0.5298, + "step": 4598 + }, + { + "epoch": 0.18, + "learning_rate": 4.7083226468325216e-05, + "loss": 0.6629, + "step": 4599 + }, + { + "epoch": 0.18, + "learning_rate": 4.708258527827648e-05, + "loss": 0.6146, + "step": 4600 + }, + { + "epoch": 0.18, + "learning_rate": 4.708194408822775e-05, + "loss": 0.5739, + "step": 4601 + }, + { + "epoch": 0.18, + "learning_rate": 4.7081302898179024e-05, + "loss": 0.6416, + "step": 4602 + }, + { + "epoch": 0.18, + "learning_rate": 4.708066170813029e-05, + "loss": 0.5726, + "step": 4603 + }, + { + "epoch": 0.18, + "learning_rate": 4.708002051808156e-05, + "loss": 0.5224, + "step": 4604 + }, + { + "epoch": 0.18, + "learning_rate": 4.707937932803283e-05, + "loss": 0.5656, + "step": 4605 + }, + { + "epoch": 0.18, + "learning_rate": 4.7078738137984104e-05, + "loss": 0.5647, + "step": 4606 + }, + { + "epoch": 0.18, + "learning_rate": 4.707809694793537e-05, + "loss": 0.6905, + "step": 4607 + }, + { + "epoch": 0.18, + "learning_rate": 4.707745575788664e-05, + "loss": 0.6759, + "step": 4608 + }, + { + "epoch": 0.18, + "learning_rate": 4.7076814567837906e-05, + "loss": 0.5656, + "step": 4609 + }, + { + "epoch": 0.18, + "learning_rate": 4.707617337778918e-05, + "loss": 0.5792, + "step": 4610 + }, + { + "epoch": 0.18, + "learning_rate": 4.707553218774045e-05, + "loss": 0.6676, + "step": 4611 + }, + { + "epoch": 0.18, + "learning_rate": 4.7074890997691714e-05, + "loss": 0.6103, + "step": 4612 + }, + { + "epoch": 0.18, + "learning_rate": 4.707424980764299e-05, + "loss": 0.5438, + "step": 4613 + }, + { + "epoch": 0.18, + "learning_rate": 4.707360861759426e-05, + "loss": 0.6458, + "step": 4614 + }, + { + "epoch": 0.18, + "learning_rate": 4.707296742754553e-05, + "loss": 0.563, + "step": 4615 + }, + { + "epoch": 0.18, + "learning_rate": 4.7072326237496794e-05, + "loss": 0.6546, + "step": 4616 + }, + { + "epoch": 0.18, + "learning_rate": 4.7071685047448066e-05, + "loss": 0.5847, + "step": 4617 + }, + { + "epoch": 0.18, + "learning_rate": 4.707104385739933e-05, + "loss": 0.6131, + "step": 4618 + }, + { + "epoch": 0.18, + "learning_rate": 4.70704026673506e-05, + "loss": 0.6495, + "step": 4619 + }, + { + "epoch": 0.18, + "learning_rate": 4.7069761477301874e-05, + "loss": 0.559, + "step": 4620 + }, + { + "epoch": 0.18, + "learning_rate": 4.7069120287253146e-05, + "loss": 0.5554, + "step": 4621 + }, + { + "epoch": 0.18, + "learning_rate": 4.706847909720442e-05, + "loss": 0.7412, + "step": 4622 + }, + { + "epoch": 0.18, + "learning_rate": 4.706783790715568e-05, + "loss": 0.6236, + "step": 4623 + }, + { + "epoch": 0.18, + "learning_rate": 4.7067196717106954e-05, + "loss": 0.5672, + "step": 4624 + }, + { + "epoch": 0.18, + "learning_rate": 4.706655552705822e-05, + "loss": 0.569, + "step": 4625 + }, + { + "epoch": 0.18, + "learning_rate": 4.706591433700949e-05, + "loss": 0.5707, + "step": 4626 + }, + { + "epoch": 0.18, + "learning_rate": 4.7065273146960756e-05, + "loss": 0.5467, + "step": 4627 + }, + { + "epoch": 0.18, + "learning_rate": 4.706463195691203e-05, + "loss": 0.6011, + "step": 4628 + }, + { + "epoch": 0.18, + "learning_rate": 4.7063990766863306e-05, + "loss": 0.6088, + "step": 4629 + }, + { + "epoch": 0.18, + "learning_rate": 4.706334957681457e-05, + "loss": 0.5818, + "step": 4630 + }, + { + "epoch": 0.18, + "learning_rate": 4.706270838676584e-05, + "loss": 0.5538, + "step": 4631 + }, + { + "epoch": 0.18, + "learning_rate": 4.706206719671711e-05, + "loss": 0.7274, + "step": 4632 + }, + { + "epoch": 0.18, + "learning_rate": 4.706142600666838e-05, + "loss": 0.6135, + "step": 4633 + }, + { + "epoch": 0.18, + "learning_rate": 4.7060784816619645e-05, + "loss": 0.6194, + "step": 4634 + }, + { + "epoch": 0.18, + "learning_rate": 4.7060143626570916e-05, + "loss": 0.5702, + "step": 4635 + }, + { + "epoch": 0.18, + "learning_rate": 4.705950243652219e-05, + "loss": 0.5176, + "step": 4636 + }, + { + "epoch": 0.18, + "learning_rate": 4.705886124647346e-05, + "loss": 0.6437, + "step": 4637 + }, + { + "epoch": 0.18, + "learning_rate": 4.705822005642473e-05, + "loss": 0.5749, + "step": 4638 + }, + { + "epoch": 0.18, + "learning_rate": 4.7057578866375996e-05, + "loss": 0.6498, + "step": 4639 + }, + { + "epoch": 0.18, + "learning_rate": 4.705693767632727e-05, + "loss": 0.5808, + "step": 4640 + }, + { + "epoch": 0.18, + "learning_rate": 4.705629648627853e-05, + "loss": 0.5708, + "step": 4641 + }, + { + "epoch": 0.18, + "learning_rate": 4.7055655296229805e-05, + "loss": 0.6483, + "step": 4642 + }, + { + "epoch": 0.18, + "learning_rate": 4.705501410618107e-05, + "loss": 0.6187, + "step": 4643 + }, + { + "epoch": 0.18, + "learning_rate": 4.705437291613234e-05, + "loss": 0.5926, + "step": 4644 + }, + { + "epoch": 0.18, + "learning_rate": 4.705373172608361e-05, + "loss": 0.539, + "step": 4645 + }, + { + "epoch": 0.18, + "learning_rate": 4.7053090536034885e-05, + "loss": 0.6166, + "step": 4646 + }, + { + "epoch": 0.18, + "learning_rate": 4.7052449345986156e-05, + "loss": 0.5641, + "step": 4647 + }, + { + "epoch": 0.18, + "learning_rate": 4.705180815593742e-05, + "loss": 0.5428, + "step": 4648 + }, + { + "epoch": 0.18, + "learning_rate": 4.705116696588869e-05, + "loss": 0.4762, + "step": 4649 + }, + { + "epoch": 0.18, + "learning_rate": 4.705052577583996e-05, + "loss": 0.5417, + "step": 4650 + }, + { + "epoch": 0.18, + "learning_rate": 4.704988458579123e-05, + "loss": 0.5945, + "step": 4651 + }, + { + "epoch": 0.18, + "learning_rate": 4.70492433957425e-05, + "loss": 0.5832, + "step": 4652 + }, + { + "epoch": 0.18, + "learning_rate": 4.704860220569377e-05, + "loss": 0.6842, + "step": 4653 + }, + { + "epoch": 0.18, + "learning_rate": 4.704796101564504e-05, + "loss": 0.5595, + "step": 4654 + }, + { + "epoch": 0.18, + "learning_rate": 4.704731982559631e-05, + "loss": 0.6015, + "step": 4655 + }, + { + "epoch": 0.18, + "learning_rate": 4.704667863554758e-05, + "loss": 0.6772, + "step": 4656 + }, + { + "epoch": 0.18, + "learning_rate": 4.7046037445498847e-05, + "loss": 0.5907, + "step": 4657 + }, + { + "epoch": 0.18, + "learning_rate": 4.704539625545012e-05, + "loss": 0.6288, + "step": 4658 + }, + { + "epoch": 0.18, + "learning_rate": 4.704475506540138e-05, + "loss": 0.5985, + "step": 4659 + }, + { + "epoch": 0.18, + "learning_rate": 4.7044113875352655e-05, + "loss": 0.4647, + "step": 4660 + }, + { + "epoch": 0.18, + "learning_rate": 4.704347268530393e-05, + "loss": 0.4758, + "step": 4661 + }, + { + "epoch": 0.18, + "learning_rate": 4.70428314952552e-05, + "loss": 0.5328, + "step": 4662 + }, + { + "epoch": 0.18, + "learning_rate": 4.704219030520646e-05, + "loss": 0.5929, + "step": 4663 + }, + { + "epoch": 0.18, + "learning_rate": 4.7041549115157735e-05, + "loss": 0.5704, + "step": 4664 + }, + { + "epoch": 0.18, + "learning_rate": 4.704090792510901e-05, + "loss": 0.5964, + "step": 4665 + }, + { + "epoch": 0.18, + "learning_rate": 4.704026673506027e-05, + "loss": 0.5821, + "step": 4666 + }, + { + "epoch": 0.18, + "learning_rate": 4.703962554501154e-05, + "loss": 0.5052, + "step": 4667 + }, + { + "epoch": 0.18, + "learning_rate": 4.7038984354962815e-05, + "loss": 0.4712, + "step": 4668 + }, + { + "epoch": 0.18, + "learning_rate": 4.703834316491409e-05, + "loss": 0.6604, + "step": 4669 + }, + { + "epoch": 0.18, + "learning_rate": 4.703770197486535e-05, + "loss": 0.6457, + "step": 4670 + }, + { + "epoch": 0.18, + "learning_rate": 4.7037060784816623e-05, + "loss": 0.6597, + "step": 4671 + }, + { + "epoch": 0.18, + "learning_rate": 4.703641959476789e-05, + "loss": 0.547, + "step": 4672 + }, + { + "epoch": 0.18, + "learning_rate": 4.703577840471916e-05, + "loss": 0.5944, + "step": 4673 + }, + { + "epoch": 0.18, + "learning_rate": 4.703513721467043e-05, + "loss": 0.7664, + "step": 4674 + }, + { + "epoch": 0.18, + "learning_rate": 4.70344960246217e-05, + "loss": 0.6026, + "step": 4675 + }, + { + "epoch": 0.18, + "learning_rate": 4.703385483457297e-05, + "loss": 0.7211, + "step": 4676 + }, + { + "epoch": 0.18, + "learning_rate": 4.703321364452424e-05, + "loss": 0.693, + "step": 4677 + }, + { + "epoch": 0.18, + "learning_rate": 4.703257245447551e-05, + "loss": 0.5438, + "step": 4678 + }, + { + "epoch": 0.18, + "learning_rate": 4.703193126442678e-05, + "loss": 0.6277, + "step": 4679 + }, + { + "epoch": 0.18, + "learning_rate": 4.703129007437805e-05, + "loss": 0.6268, + "step": 4680 + }, + { + "epoch": 0.18, + "learning_rate": 4.7030648884329314e-05, + "loss": 0.5416, + "step": 4681 + }, + { + "epoch": 0.18, + "learning_rate": 4.7030007694280585e-05, + "loss": 0.6283, + "step": 4682 + }, + { + "epoch": 0.18, + "learning_rate": 4.702936650423186e-05, + "loss": 0.6028, + "step": 4683 + }, + { + "epoch": 0.18, + "learning_rate": 4.702872531418312e-05, + "loss": 0.6546, + "step": 4684 + }, + { + "epoch": 0.18, + "learning_rate": 4.70280841241344e-05, + "loss": 0.5768, + "step": 4685 + }, + { + "epoch": 0.18, + "learning_rate": 4.7027442934085665e-05, + "loss": 0.5909, + "step": 4686 + }, + { + "epoch": 0.18, + "learning_rate": 4.702680174403694e-05, + "loss": 0.5863, + "step": 4687 + }, + { + "epoch": 0.18, + "learning_rate": 4.70261605539882e-05, + "loss": 0.5708, + "step": 4688 + }, + { + "epoch": 0.18, + "learning_rate": 4.7025519363939474e-05, + "loss": 0.6255, + "step": 4689 + }, + { + "epoch": 0.18, + "learning_rate": 4.7024878173890745e-05, + "loss": 0.6281, + "step": 4690 + }, + { + "epoch": 0.18, + "learning_rate": 4.702423698384201e-05, + "loss": 0.6446, + "step": 4691 + }, + { + "epoch": 0.18, + "learning_rate": 4.702359579379328e-05, + "loss": 0.6442, + "step": 4692 + }, + { + "epoch": 0.18, + "learning_rate": 4.7022954603744554e-05, + "loss": 0.5469, + "step": 4693 + }, + { + "epoch": 0.18, + "learning_rate": 4.7022313413695826e-05, + "loss": 0.686, + "step": 4694 + }, + { + "epoch": 0.18, + "learning_rate": 4.702167222364709e-05, + "loss": 0.5075, + "step": 4695 + }, + { + "epoch": 0.18, + "learning_rate": 4.702103103359836e-05, + "loss": 0.5293, + "step": 4696 + }, + { + "epoch": 0.18, + "learning_rate": 4.702038984354963e-05, + "loss": 0.566, + "step": 4697 + }, + { + "epoch": 0.18, + "learning_rate": 4.70197486535009e-05, + "loss": 0.5746, + "step": 4698 + }, + { + "epoch": 0.18, + "learning_rate": 4.701910746345217e-05, + "loss": 0.6299, + "step": 4699 + }, + { + "epoch": 0.18, + "learning_rate": 4.7018466273403435e-05, + "loss": 0.6351, + "step": 4700 + }, + { + "epoch": 0.18, + "learning_rate": 4.7017825083354714e-05, + "loss": 0.5723, + "step": 4701 + }, + { + "epoch": 0.18, + "learning_rate": 4.701718389330598e-05, + "loss": 0.5745, + "step": 4702 + }, + { + "epoch": 0.18, + "learning_rate": 4.701654270325725e-05, + "loss": 0.6462, + "step": 4703 + }, + { + "epoch": 0.18, + "learning_rate": 4.7015901513208516e-05, + "loss": 0.4602, + "step": 4704 + }, + { + "epoch": 0.18, + "learning_rate": 4.701526032315979e-05, + "loss": 0.6417, + "step": 4705 + }, + { + "epoch": 0.18, + "learning_rate": 4.701461913311105e-05, + "loss": 0.6252, + "step": 4706 + }, + { + "epoch": 0.18, + "learning_rate": 4.7013977943062324e-05, + "loss": 0.5382, + "step": 4707 + }, + { + "epoch": 0.18, + "learning_rate": 4.7013336753013596e-05, + "loss": 0.5269, + "step": 4708 + }, + { + "epoch": 0.18, + "learning_rate": 4.701269556296487e-05, + "loss": 0.6131, + "step": 4709 + }, + { + "epoch": 0.18, + "learning_rate": 4.701205437291614e-05, + "loss": 0.63, + "step": 4710 + }, + { + "epoch": 0.18, + "learning_rate": 4.7011413182867404e-05, + "loss": 0.6798, + "step": 4711 + }, + { + "epoch": 0.18, + "learning_rate": 4.7010771992818676e-05, + "loss": 0.5975, + "step": 4712 + }, + { + "epoch": 0.18, + "learning_rate": 4.701013080276994e-05, + "loss": 0.5996, + "step": 4713 + }, + { + "epoch": 0.18, + "learning_rate": 4.700948961272121e-05, + "loss": 0.5979, + "step": 4714 + }, + { + "epoch": 0.18, + "learning_rate": 4.700884842267248e-05, + "loss": 0.6607, + "step": 4715 + }, + { + "epoch": 0.18, + "learning_rate": 4.700820723262375e-05, + "loss": 0.56, + "step": 4716 + }, + { + "epoch": 0.18, + "learning_rate": 4.700756604257502e-05, + "loss": 0.6063, + "step": 4717 + }, + { + "epoch": 0.18, + "learning_rate": 4.700692485252629e-05, + "loss": 0.6724, + "step": 4718 + }, + { + "epoch": 0.18, + "learning_rate": 4.7006283662477564e-05, + "loss": 0.5757, + "step": 4719 + }, + { + "epoch": 0.18, + "learning_rate": 4.700564247242883e-05, + "loss": 0.6443, + "step": 4720 + }, + { + "epoch": 0.18, + "learning_rate": 4.70050012823801e-05, + "loss": 0.6196, + "step": 4721 + }, + { + "epoch": 0.18, + "learning_rate": 4.7004360092331366e-05, + "loss": 0.5893, + "step": 4722 + }, + { + "epoch": 0.18, + "learning_rate": 4.700371890228264e-05, + "loss": 0.601, + "step": 4723 + }, + { + "epoch": 0.18, + "learning_rate": 4.700307771223391e-05, + "loss": 0.6022, + "step": 4724 + }, + { + "epoch": 0.18, + "learning_rate": 4.700243652218518e-05, + "loss": 0.6509, + "step": 4725 + }, + { + "epoch": 0.18, + "learning_rate": 4.700179533213645e-05, + "loss": 0.6407, + "step": 4726 + }, + { + "epoch": 0.18, + "learning_rate": 4.700115414208772e-05, + "loss": 0.5816, + "step": 4727 + }, + { + "epoch": 0.18, + "learning_rate": 4.700051295203899e-05, + "loss": 0.6475, + "step": 4728 + }, + { + "epoch": 0.18, + "learning_rate": 4.6999871761990254e-05, + "loss": 0.6023, + "step": 4729 + }, + { + "epoch": 0.18, + "learning_rate": 4.6999230571941526e-05, + "loss": 0.6682, + "step": 4730 + }, + { + "epoch": 0.18, + "learning_rate": 4.699858938189279e-05, + "loss": 0.6494, + "step": 4731 + }, + { + "epoch": 0.18, + "learning_rate": 4.699794819184406e-05, + "loss": 0.5818, + "step": 4732 + }, + { + "epoch": 0.18, + "learning_rate": 4.6997307001795334e-05, + "loss": 0.6182, + "step": 4733 + }, + { + "epoch": 0.18, + "learning_rate": 4.6996665811746606e-05, + "loss": 0.5798, + "step": 4734 + }, + { + "epoch": 0.18, + "learning_rate": 4.699602462169788e-05, + "loss": 0.6428, + "step": 4735 + }, + { + "epoch": 0.18, + "learning_rate": 4.699538343164914e-05, + "loss": 0.6252, + "step": 4736 + }, + { + "epoch": 0.18, + "learning_rate": 4.6994742241600414e-05, + "loss": 0.6074, + "step": 4737 + }, + { + "epoch": 0.18, + "learning_rate": 4.699410105155168e-05, + "loss": 0.5529, + "step": 4738 + }, + { + "epoch": 0.18, + "learning_rate": 4.699345986150295e-05, + "loss": 0.6913, + "step": 4739 + }, + { + "epoch": 0.18, + "learning_rate": 4.699281867145422e-05, + "loss": 0.6416, + "step": 4740 + }, + { + "epoch": 0.18, + "learning_rate": 4.6992177481405495e-05, + "loss": 0.5797, + "step": 4741 + }, + { + "epoch": 0.18, + "learning_rate": 4.699153629135676e-05, + "loss": 0.6534, + "step": 4742 + }, + { + "epoch": 0.18, + "learning_rate": 4.699089510130803e-05, + "loss": 0.6146, + "step": 4743 + }, + { + "epoch": 0.18, + "learning_rate": 4.69902539112593e-05, + "loss": 0.5675, + "step": 4744 + }, + { + "epoch": 0.18, + "learning_rate": 4.698961272121057e-05, + "loss": 0.6391, + "step": 4745 + }, + { + "epoch": 0.18, + "learning_rate": 4.698897153116184e-05, + "loss": 0.5494, + "step": 4746 + }, + { + "epoch": 0.18, + "learning_rate": 4.6988330341113105e-05, + "loss": 0.5758, + "step": 4747 + }, + { + "epoch": 0.18, + "learning_rate": 4.6987689151064376e-05, + "loss": 0.6139, + "step": 4748 + }, + { + "epoch": 0.18, + "learning_rate": 4.698704796101565e-05, + "loss": 0.6301, + "step": 4749 + }, + { + "epoch": 0.18, + "learning_rate": 4.698640677096692e-05, + "loss": 0.5869, + "step": 4750 + }, + { + "epoch": 0.18, + "learning_rate": 4.6985765580918185e-05, + "loss": 0.6063, + "step": 4751 + }, + { + "epoch": 0.18, + "learning_rate": 4.6985124390869456e-05, + "loss": 0.6585, + "step": 4752 + }, + { + "epoch": 0.18, + "learning_rate": 4.698448320082073e-05, + "loss": 0.5885, + "step": 4753 + }, + { + "epoch": 0.18, + "learning_rate": 4.698384201077199e-05, + "loss": 0.6407, + "step": 4754 + }, + { + "epoch": 0.18, + "learning_rate": 4.6983200820723265e-05, + "loss": 0.5817, + "step": 4755 + }, + { + "epoch": 0.18, + "learning_rate": 4.698255963067453e-05, + "loss": 0.5083, + "step": 4756 + }, + { + "epoch": 0.18, + "learning_rate": 4.698191844062581e-05, + "loss": 0.609, + "step": 4757 + }, + { + "epoch": 0.18, + "learning_rate": 4.698127725057707e-05, + "loss": 0.6464, + "step": 4758 + }, + { + "epoch": 0.18, + "learning_rate": 4.6980636060528345e-05, + "loss": 0.5647, + "step": 4759 + }, + { + "epoch": 0.18, + "learning_rate": 4.697999487047961e-05, + "loss": 0.6387, + "step": 4760 + }, + { + "epoch": 0.18, + "learning_rate": 4.697935368043088e-05, + "loss": 0.654, + "step": 4761 + }, + { + "epoch": 0.18, + "learning_rate": 4.697871249038215e-05, + "loss": 0.6366, + "step": 4762 + }, + { + "epoch": 0.18, + "learning_rate": 4.697807130033342e-05, + "loss": 0.5495, + "step": 4763 + }, + { + "epoch": 0.18, + "learning_rate": 4.697743011028469e-05, + "loss": 0.587, + "step": 4764 + }, + { + "epoch": 0.18, + "learning_rate": 4.697678892023596e-05, + "loss": 0.6368, + "step": 4765 + }, + { + "epoch": 0.18, + "learning_rate": 4.697614773018723e-05, + "loss": 0.6009, + "step": 4766 + }, + { + "epoch": 0.18, + "learning_rate": 4.69755065401385e-05, + "loss": 0.5486, + "step": 4767 + }, + { + "epoch": 0.18, + "learning_rate": 4.697486535008977e-05, + "loss": 0.5708, + "step": 4768 + }, + { + "epoch": 0.18, + "learning_rate": 4.6974224160041035e-05, + "loss": 0.6205, + "step": 4769 + }, + { + "epoch": 0.18, + "learning_rate": 4.6973582969992307e-05, + "loss": 0.5895, + "step": 4770 + }, + { + "epoch": 0.18, + "learning_rate": 4.697294177994358e-05, + "loss": 0.6426, + "step": 4771 + }, + { + "epoch": 0.18, + "learning_rate": 4.697230058989484e-05, + "loss": 0.6063, + "step": 4772 + }, + { + "epoch": 0.18, + "learning_rate": 4.697165939984612e-05, + "loss": 0.5654, + "step": 4773 + }, + { + "epoch": 0.18, + "learning_rate": 4.697101820979739e-05, + "loss": 0.5334, + "step": 4774 + }, + { + "epoch": 0.18, + "learning_rate": 4.697037701974866e-05, + "loss": 0.559, + "step": 4775 + }, + { + "epoch": 0.18, + "learning_rate": 4.696973582969992e-05, + "loss": 0.5443, + "step": 4776 + }, + { + "epoch": 0.18, + "learning_rate": 4.6969094639651195e-05, + "loss": 0.5247, + "step": 4777 + }, + { + "epoch": 0.18, + "learning_rate": 4.696845344960246e-05, + "loss": 0.6458, + "step": 4778 + }, + { + "epoch": 0.18, + "learning_rate": 4.696781225955373e-05, + "loss": 0.6179, + "step": 4779 + }, + { + "epoch": 0.18, + "learning_rate": 4.6967171069505003e-05, + "loss": 0.5594, + "step": 4780 + }, + { + "epoch": 0.18, + "learning_rate": 4.6966529879456275e-05, + "loss": 0.6443, + "step": 4781 + }, + { + "epoch": 0.18, + "learning_rate": 4.696588868940755e-05, + "loss": 0.6364, + "step": 4782 + }, + { + "epoch": 0.18, + "learning_rate": 4.696524749935881e-05, + "loss": 0.5997, + "step": 4783 + }, + { + "epoch": 0.18, + "learning_rate": 4.6964606309310084e-05, + "loss": 0.6246, + "step": 4784 + }, + { + "epoch": 0.18, + "learning_rate": 4.696396511926135e-05, + "loss": 0.562, + "step": 4785 + }, + { + "epoch": 0.18, + "learning_rate": 4.696332392921262e-05, + "loss": 0.6093, + "step": 4786 + }, + { + "epoch": 0.18, + "learning_rate": 4.6962682739163885e-05, + "loss": 0.6156, + "step": 4787 + }, + { + "epoch": 0.18, + "learning_rate": 4.696204154911516e-05, + "loss": 0.5873, + "step": 4788 + }, + { + "epoch": 0.18, + "learning_rate": 4.6961400359066435e-05, + "loss": 0.7025, + "step": 4789 + }, + { + "epoch": 0.18, + "learning_rate": 4.69607591690177e-05, + "loss": 0.5956, + "step": 4790 + }, + { + "epoch": 0.18, + "learning_rate": 4.696011797896897e-05, + "loss": 0.5442, + "step": 4791 + }, + { + "epoch": 0.18, + "learning_rate": 4.695947678892024e-05, + "loss": 0.571, + "step": 4792 + }, + { + "epoch": 0.18, + "learning_rate": 4.695883559887151e-05, + "loss": 0.5825, + "step": 4793 + }, + { + "epoch": 0.18, + "learning_rate": 4.6958194408822774e-05, + "loss": 0.6007, + "step": 4794 + }, + { + "epoch": 0.18, + "learning_rate": 4.6957553218774045e-05, + "loss": 0.6088, + "step": 4795 + }, + { + "epoch": 0.18, + "learning_rate": 4.695691202872532e-05, + "loss": 0.6915, + "step": 4796 + }, + { + "epoch": 0.18, + "learning_rate": 4.695627083867659e-05, + "loss": 0.5742, + "step": 4797 + }, + { + "epoch": 0.18, + "learning_rate": 4.695562964862786e-05, + "loss": 0.5361, + "step": 4798 + }, + { + "epoch": 0.18, + "learning_rate": 4.6954988458579125e-05, + "loss": 0.5461, + "step": 4799 + }, + { + "epoch": 0.18, + "learning_rate": 4.69543472685304e-05, + "loss": 0.6443, + "step": 4800 + }, + { + "epoch": 0.18, + "learning_rate": 4.695370607848166e-05, + "loss": 0.5922, + "step": 4801 + }, + { + "epoch": 0.18, + "learning_rate": 4.6953064888432934e-05, + "loss": 0.5705, + "step": 4802 + }, + { + "epoch": 0.18, + "learning_rate": 4.69524236983842e-05, + "loss": 0.526, + "step": 4803 + }, + { + "epoch": 0.18, + "learning_rate": 4.695178250833547e-05, + "loss": 0.5989, + "step": 4804 + }, + { + "epoch": 0.18, + "learning_rate": 4.695114131828674e-05, + "loss": 0.6113, + "step": 4805 + }, + { + "epoch": 0.18, + "learning_rate": 4.6950500128238014e-05, + "loss": 0.6128, + "step": 4806 + }, + { + "epoch": 0.18, + "learning_rate": 4.6949858938189286e-05, + "loss": 0.661, + "step": 4807 + }, + { + "epoch": 0.18, + "learning_rate": 4.694921774814055e-05, + "loss": 0.5885, + "step": 4808 + }, + { + "epoch": 0.18, + "learning_rate": 4.694857655809182e-05, + "loss": 0.6166, + "step": 4809 + }, + { + "epoch": 0.18, + "learning_rate": 4.694793536804309e-05, + "loss": 0.6241, + "step": 4810 + }, + { + "epoch": 0.18, + "learning_rate": 4.694729417799436e-05, + "loss": 0.5964, + "step": 4811 + }, + { + "epoch": 0.19, + "learning_rate": 4.694665298794563e-05, + "loss": 0.5769, + "step": 4812 + }, + { + "epoch": 0.19, + "learning_rate": 4.69460117978969e-05, + "loss": 0.6301, + "step": 4813 + }, + { + "epoch": 0.19, + "learning_rate": 4.694537060784817e-05, + "loss": 0.6181, + "step": 4814 + }, + { + "epoch": 0.19, + "learning_rate": 4.694472941779944e-05, + "loss": 0.6564, + "step": 4815 + }, + { + "epoch": 0.19, + "learning_rate": 4.694408822775071e-05, + "loss": 0.5523, + "step": 4816 + }, + { + "epoch": 0.19, + "learning_rate": 4.6943447037701976e-05, + "loss": 0.5761, + "step": 4817 + }, + { + "epoch": 0.19, + "learning_rate": 4.694280584765325e-05, + "loss": 0.5884, + "step": 4818 + }, + { + "epoch": 0.19, + "learning_rate": 4.694216465760451e-05, + "loss": 0.6512, + "step": 4819 + }, + { + "epoch": 0.19, + "learning_rate": 4.6941523467555784e-05, + "loss": 0.5287, + "step": 4820 + }, + { + "epoch": 0.19, + "learning_rate": 4.6940882277507056e-05, + "loss": 0.5834, + "step": 4821 + }, + { + "epoch": 0.19, + "learning_rate": 4.694024108745833e-05, + "loss": 0.5967, + "step": 4822 + }, + { + "epoch": 0.19, + "learning_rate": 4.693959989740959e-05, + "loss": 0.5779, + "step": 4823 + }, + { + "epoch": 0.19, + "learning_rate": 4.6938958707360864e-05, + "loss": 0.5309, + "step": 4824 + }, + { + "epoch": 0.19, + "learning_rate": 4.6938317517312136e-05, + "loss": 0.5955, + "step": 4825 + }, + { + "epoch": 0.19, + "learning_rate": 4.69376763272634e-05, + "loss": 0.4966, + "step": 4826 + }, + { + "epoch": 0.19, + "learning_rate": 4.693703513721467e-05, + "loss": 0.5817, + "step": 4827 + }, + { + "epoch": 0.19, + "learning_rate": 4.6936393947165944e-05, + "loss": 0.6193, + "step": 4828 + }, + { + "epoch": 0.19, + "learning_rate": 4.6935752757117216e-05, + "loss": 0.5906, + "step": 4829 + }, + { + "epoch": 0.19, + "learning_rate": 4.693511156706848e-05, + "loss": 0.6704, + "step": 4830 + }, + { + "epoch": 0.19, + "learning_rate": 4.693447037701975e-05, + "loss": 0.5177, + "step": 4831 + }, + { + "epoch": 0.19, + "learning_rate": 4.693382918697102e-05, + "loss": 0.549, + "step": 4832 + }, + { + "epoch": 0.19, + "learning_rate": 4.693318799692229e-05, + "loss": 0.6151, + "step": 4833 + }, + { + "epoch": 0.19, + "learning_rate": 4.693254680687356e-05, + "loss": 0.5557, + "step": 4834 + }, + { + "epoch": 0.19, + "learning_rate": 4.6931905616824826e-05, + "loss": 0.5102, + "step": 4835 + }, + { + "epoch": 0.19, + "learning_rate": 4.69312644267761e-05, + "loss": 0.4801, + "step": 4836 + }, + { + "epoch": 0.19, + "learning_rate": 4.693062323672737e-05, + "loss": 0.5567, + "step": 4837 + }, + { + "epoch": 0.19, + "learning_rate": 4.692998204667864e-05, + "loss": 0.5381, + "step": 4838 + }, + { + "epoch": 0.19, + "learning_rate": 4.6929340856629906e-05, + "loss": 0.571, + "step": 4839 + }, + { + "epoch": 0.19, + "learning_rate": 4.692869966658118e-05, + "loss": 0.5207, + "step": 4840 + }, + { + "epoch": 0.19, + "learning_rate": 4.692805847653244e-05, + "loss": 0.5807, + "step": 4841 + }, + { + "epoch": 0.19, + "learning_rate": 4.6927417286483714e-05, + "loss": 0.6093, + "step": 4842 + }, + { + "epoch": 0.19, + "learning_rate": 4.6926776096434986e-05, + "loss": 0.6211, + "step": 4843 + }, + { + "epoch": 0.19, + "learning_rate": 4.692613490638625e-05, + "loss": 0.6314, + "step": 4844 + }, + { + "epoch": 0.19, + "learning_rate": 4.692549371633753e-05, + "loss": 0.6207, + "step": 4845 + }, + { + "epoch": 0.19, + "learning_rate": 4.6924852526288794e-05, + "loss": 0.6268, + "step": 4846 + }, + { + "epoch": 0.19, + "learning_rate": 4.6924211336240066e-05, + "loss": 0.6785, + "step": 4847 + }, + { + "epoch": 0.19, + "learning_rate": 4.692357014619133e-05, + "loss": 0.5979, + "step": 4848 + }, + { + "epoch": 0.19, + "learning_rate": 4.69229289561426e-05, + "loss": 0.6942, + "step": 4849 + }, + { + "epoch": 0.19, + "learning_rate": 4.692228776609387e-05, + "loss": 0.5929, + "step": 4850 + }, + { + "epoch": 0.19, + "learning_rate": 4.692164657604514e-05, + "loss": 0.5063, + "step": 4851 + }, + { + "epoch": 0.19, + "learning_rate": 4.692100538599641e-05, + "loss": 0.5387, + "step": 4852 + }, + { + "epoch": 0.19, + "learning_rate": 4.692036419594768e-05, + "loss": 0.6208, + "step": 4853 + }, + { + "epoch": 0.19, + "learning_rate": 4.6919723005898955e-05, + "loss": 0.6011, + "step": 4854 + }, + { + "epoch": 0.19, + "learning_rate": 4.691908181585022e-05, + "loss": 0.5025, + "step": 4855 + }, + { + "epoch": 0.19, + "learning_rate": 4.691844062580149e-05, + "loss": 0.638, + "step": 4856 + }, + { + "epoch": 0.19, + "learning_rate": 4.6917799435752756e-05, + "loss": 0.5437, + "step": 4857 + }, + { + "epoch": 0.19, + "learning_rate": 4.691715824570403e-05, + "loss": 0.5296, + "step": 4858 + }, + { + "epoch": 0.19, + "learning_rate": 4.691651705565529e-05, + "loss": 0.6477, + "step": 4859 + }, + { + "epoch": 0.19, + "learning_rate": 4.6915875865606565e-05, + "loss": 0.5679, + "step": 4860 + }, + { + "epoch": 0.19, + "learning_rate": 4.691523467555784e-05, + "loss": 0.6199, + "step": 4861 + }, + { + "epoch": 0.19, + "learning_rate": 4.691459348550911e-05, + "loss": 0.566, + "step": 4862 + }, + { + "epoch": 0.19, + "learning_rate": 4.691395229546038e-05, + "loss": 0.6112, + "step": 4863 + }, + { + "epoch": 0.19, + "learning_rate": 4.6913311105411645e-05, + "loss": 0.5627, + "step": 4864 + }, + { + "epoch": 0.19, + "learning_rate": 4.6912669915362916e-05, + "loss": 0.561, + "step": 4865 + }, + { + "epoch": 0.19, + "learning_rate": 4.691202872531418e-05, + "loss": 0.6431, + "step": 4866 + }, + { + "epoch": 0.19, + "learning_rate": 4.691138753526545e-05, + "loss": 0.5931, + "step": 4867 + }, + { + "epoch": 0.19, + "learning_rate": 4.6910746345216725e-05, + "loss": 0.5877, + "step": 4868 + }, + { + "epoch": 0.19, + "learning_rate": 4.6910105155167996e-05, + "loss": 0.5925, + "step": 4869 + }, + { + "epoch": 0.19, + "learning_rate": 4.690946396511927e-05, + "loss": 0.5967, + "step": 4870 + }, + { + "epoch": 0.19, + "learning_rate": 4.690882277507053e-05, + "loss": 0.503, + "step": 4871 + }, + { + "epoch": 0.19, + "learning_rate": 4.6908181585021805e-05, + "loss": 0.5351, + "step": 4872 + }, + { + "epoch": 0.19, + "learning_rate": 4.690754039497307e-05, + "loss": 0.53, + "step": 4873 + }, + { + "epoch": 0.19, + "learning_rate": 4.690689920492434e-05, + "loss": 0.5574, + "step": 4874 + }, + { + "epoch": 0.19, + "learning_rate": 4.6906258014875606e-05, + "loss": 0.5128, + "step": 4875 + }, + { + "epoch": 0.19, + "learning_rate": 4.690561682482688e-05, + "loss": 0.6416, + "step": 4876 + }, + { + "epoch": 0.19, + "learning_rate": 4.690497563477816e-05, + "loss": 0.5695, + "step": 4877 + }, + { + "epoch": 0.19, + "learning_rate": 4.690433444472942e-05, + "loss": 0.5595, + "step": 4878 + }, + { + "epoch": 0.19, + "learning_rate": 4.690369325468069e-05, + "loss": 0.5277, + "step": 4879 + }, + { + "epoch": 0.19, + "learning_rate": 4.690305206463196e-05, + "loss": 0.5877, + "step": 4880 + }, + { + "epoch": 0.19, + "learning_rate": 4.690241087458323e-05, + "loss": 0.5838, + "step": 4881 + }, + { + "epoch": 0.19, + "learning_rate": 4.6901769684534495e-05, + "loss": 0.6062, + "step": 4882 + }, + { + "epoch": 0.19, + "learning_rate": 4.690112849448577e-05, + "loss": 0.6655, + "step": 4883 + }, + { + "epoch": 0.19, + "learning_rate": 4.690048730443704e-05, + "loss": 0.5386, + "step": 4884 + }, + { + "epoch": 0.19, + "learning_rate": 4.689984611438831e-05, + "loss": 0.6526, + "step": 4885 + }, + { + "epoch": 0.19, + "learning_rate": 4.689920492433958e-05, + "loss": 0.5867, + "step": 4886 + }, + { + "epoch": 0.19, + "learning_rate": 4.689856373429085e-05, + "loss": 0.6654, + "step": 4887 + }, + { + "epoch": 0.19, + "learning_rate": 4.689792254424212e-05, + "loss": 0.6673, + "step": 4888 + }, + { + "epoch": 0.19, + "learning_rate": 4.689728135419338e-05, + "loss": 0.4851, + "step": 4889 + }, + { + "epoch": 0.19, + "learning_rate": 4.6896640164144655e-05, + "loss": 0.6594, + "step": 4890 + }, + { + "epoch": 0.19, + "learning_rate": 4.689599897409592e-05, + "loss": 0.5233, + "step": 4891 + }, + { + "epoch": 0.19, + "learning_rate": 4.689535778404719e-05, + "loss": 0.6422, + "step": 4892 + }, + { + "epoch": 0.19, + "learning_rate": 4.6894716593998463e-05, + "loss": 0.6083, + "step": 4893 + }, + { + "epoch": 0.19, + "learning_rate": 4.6894075403949735e-05, + "loss": 0.4848, + "step": 4894 + }, + { + "epoch": 0.19, + "learning_rate": 4.689343421390101e-05, + "loss": 0.5765, + "step": 4895 + }, + { + "epoch": 0.19, + "learning_rate": 4.689279302385227e-05, + "loss": 0.5261, + "step": 4896 + }, + { + "epoch": 0.19, + "learning_rate": 4.6892151833803544e-05, + "loss": 0.679, + "step": 4897 + }, + { + "epoch": 0.19, + "learning_rate": 4.689151064375481e-05, + "loss": 0.5814, + "step": 4898 + }, + { + "epoch": 0.19, + "learning_rate": 4.689086945370608e-05, + "loss": 0.5189, + "step": 4899 + }, + { + "epoch": 0.19, + "learning_rate": 4.689022826365735e-05, + "loss": 0.5318, + "step": 4900 + }, + { + "epoch": 0.19, + "learning_rate": 4.6889587073608624e-05, + "loss": 0.606, + "step": 4901 + }, + { + "epoch": 0.19, + "learning_rate": 4.688894588355989e-05, + "loss": 0.6135, + "step": 4902 + }, + { + "epoch": 0.19, + "learning_rate": 4.688830469351116e-05, + "loss": 0.6067, + "step": 4903 + }, + { + "epoch": 0.19, + "learning_rate": 4.688766350346243e-05, + "loss": 0.5142, + "step": 4904 + }, + { + "epoch": 0.19, + "learning_rate": 4.68870223134137e-05, + "loss": 0.6499, + "step": 4905 + }, + { + "epoch": 0.19, + "learning_rate": 4.688638112336497e-05, + "loss": 0.5905, + "step": 4906 + }, + { + "epoch": 0.19, + "learning_rate": 4.6885739933316234e-05, + "loss": 0.6693, + "step": 4907 + }, + { + "epoch": 0.19, + "learning_rate": 4.6885098743267505e-05, + "loss": 0.5773, + "step": 4908 + }, + { + "epoch": 0.19, + "learning_rate": 4.688445755321878e-05, + "loss": 0.6536, + "step": 4909 + }, + { + "epoch": 0.19, + "learning_rate": 4.688381636317005e-05, + "loss": 0.6502, + "step": 4910 + }, + { + "epoch": 0.19, + "learning_rate": 4.6883175173121314e-05, + "loss": 0.6023, + "step": 4911 + }, + { + "epoch": 0.19, + "learning_rate": 4.6882533983072585e-05, + "loss": 0.5624, + "step": 4912 + }, + { + "epoch": 0.19, + "learning_rate": 4.688189279302386e-05, + "loss": 0.587, + "step": 4913 + }, + { + "epoch": 0.19, + "learning_rate": 4.688125160297512e-05, + "loss": 0.5743, + "step": 4914 + }, + { + "epoch": 0.19, + "learning_rate": 4.6880610412926394e-05, + "loss": 0.5296, + "step": 4915 + }, + { + "epoch": 0.19, + "learning_rate": 4.6879969222877666e-05, + "loss": 0.5974, + "step": 4916 + }, + { + "epoch": 0.19, + "learning_rate": 4.687932803282894e-05, + "loss": 0.6777, + "step": 4917 + }, + { + "epoch": 0.19, + "learning_rate": 4.68786868427802e-05, + "loss": 0.566, + "step": 4918 + }, + { + "epoch": 0.19, + "learning_rate": 4.6878045652731474e-05, + "loss": 0.6246, + "step": 4919 + }, + { + "epoch": 0.19, + "learning_rate": 4.687740446268274e-05, + "loss": 0.5538, + "step": 4920 + }, + { + "epoch": 0.19, + "learning_rate": 4.687676327263401e-05, + "loss": 0.5939, + "step": 4921 + }, + { + "epoch": 0.19, + "learning_rate": 4.687612208258528e-05, + "loss": 0.6284, + "step": 4922 + }, + { + "epoch": 0.19, + "learning_rate": 4.687548089253655e-05, + "loss": 0.6213, + "step": 4923 + }, + { + "epoch": 0.19, + "learning_rate": 4.687483970248782e-05, + "loss": 0.6166, + "step": 4924 + }, + { + "epoch": 0.19, + "learning_rate": 4.687419851243909e-05, + "loss": 0.5709, + "step": 4925 + }, + { + "epoch": 0.19, + "learning_rate": 4.687355732239036e-05, + "loss": 0.6103, + "step": 4926 + }, + { + "epoch": 0.19, + "learning_rate": 4.687291613234163e-05, + "loss": 0.5543, + "step": 4927 + }, + { + "epoch": 0.19, + "learning_rate": 4.68722749422929e-05, + "loss": 0.6566, + "step": 4928 + }, + { + "epoch": 0.19, + "learning_rate": 4.6871633752244164e-05, + "loss": 0.5986, + "step": 4929 + }, + { + "epoch": 0.19, + "learning_rate": 4.6870992562195436e-05, + "loss": 0.6072, + "step": 4930 + }, + { + "epoch": 0.19, + "learning_rate": 4.687035137214671e-05, + "loss": 0.5912, + "step": 4931 + }, + { + "epoch": 0.19, + "learning_rate": 4.686971018209797e-05, + "loss": 0.5631, + "step": 4932 + }, + { + "epoch": 0.19, + "learning_rate": 4.686906899204925e-05, + "loss": 0.6023, + "step": 4933 + }, + { + "epoch": 0.19, + "learning_rate": 4.6868427802000516e-05, + "loss": 0.5959, + "step": 4934 + }, + { + "epoch": 0.19, + "learning_rate": 4.686778661195179e-05, + "loss": 0.5347, + "step": 4935 + }, + { + "epoch": 0.19, + "learning_rate": 4.686714542190305e-05, + "loss": 0.6858, + "step": 4936 + }, + { + "epoch": 0.19, + "learning_rate": 4.6866504231854324e-05, + "loss": 0.5683, + "step": 4937 + }, + { + "epoch": 0.19, + "learning_rate": 4.686586304180559e-05, + "loss": 0.5273, + "step": 4938 + }, + { + "epoch": 0.19, + "learning_rate": 4.686522185175686e-05, + "loss": 0.5746, + "step": 4939 + }, + { + "epoch": 0.19, + "learning_rate": 4.686458066170813e-05, + "loss": 0.7113, + "step": 4940 + }, + { + "epoch": 0.19, + "learning_rate": 4.6863939471659404e-05, + "loss": 0.6121, + "step": 4941 + }, + { + "epoch": 0.19, + "learning_rate": 4.6863298281610676e-05, + "loss": 0.572, + "step": 4942 + }, + { + "epoch": 0.19, + "learning_rate": 4.686265709156194e-05, + "loss": 0.5806, + "step": 4943 + }, + { + "epoch": 0.19, + "learning_rate": 4.686201590151321e-05, + "loss": 0.5361, + "step": 4944 + }, + { + "epoch": 0.19, + "learning_rate": 4.686137471146448e-05, + "loss": 0.5706, + "step": 4945 + }, + { + "epoch": 0.19, + "learning_rate": 4.686073352141575e-05, + "loss": 0.5275, + "step": 4946 + }, + { + "epoch": 0.19, + "learning_rate": 4.6860092331367014e-05, + "loss": 0.5798, + "step": 4947 + }, + { + "epoch": 0.19, + "learning_rate": 4.6859451141318286e-05, + "loss": 0.5655, + "step": 4948 + }, + { + "epoch": 0.19, + "learning_rate": 4.6858809951269564e-05, + "loss": 0.576, + "step": 4949 + }, + { + "epoch": 0.19, + "learning_rate": 4.685816876122083e-05, + "loss": 0.6035, + "step": 4950 + }, + { + "epoch": 0.19, + "learning_rate": 4.68575275711721e-05, + "loss": 0.616, + "step": 4951 + }, + { + "epoch": 0.19, + "learning_rate": 4.6856886381123366e-05, + "loss": 0.5336, + "step": 4952 + }, + { + "epoch": 0.19, + "learning_rate": 4.685624519107464e-05, + "loss": 0.5261, + "step": 4953 + }, + { + "epoch": 0.19, + "learning_rate": 4.68556040010259e-05, + "loss": 0.623, + "step": 4954 + }, + { + "epoch": 0.19, + "learning_rate": 4.6854962810977174e-05, + "loss": 0.5751, + "step": 4955 + }, + { + "epoch": 0.19, + "learning_rate": 4.6854321620928446e-05, + "loss": 0.5982, + "step": 4956 + }, + { + "epoch": 0.19, + "learning_rate": 4.685368043087972e-05, + "loss": 0.5458, + "step": 4957 + }, + { + "epoch": 0.19, + "learning_rate": 4.685303924083099e-05, + "loss": 0.7074, + "step": 4958 + }, + { + "epoch": 0.19, + "learning_rate": 4.6852398050782254e-05, + "loss": 0.7939, + "step": 4959 + }, + { + "epoch": 0.19, + "learning_rate": 4.6851756860733526e-05, + "loss": 0.6492, + "step": 4960 + }, + { + "epoch": 0.19, + "learning_rate": 4.685111567068479e-05, + "loss": 0.6039, + "step": 4961 + }, + { + "epoch": 0.19, + "learning_rate": 4.685047448063606e-05, + "loss": 0.5248, + "step": 4962 + }, + { + "epoch": 0.19, + "learning_rate": 4.684983329058733e-05, + "loss": 0.5626, + "step": 4963 + }, + { + "epoch": 0.19, + "learning_rate": 4.68491921005386e-05, + "loss": 0.5988, + "step": 4964 + }, + { + "epoch": 0.19, + "learning_rate": 4.684855091048987e-05, + "loss": 0.6549, + "step": 4965 + }, + { + "epoch": 0.19, + "learning_rate": 4.684790972044114e-05, + "loss": 0.634, + "step": 4966 + }, + { + "epoch": 0.19, + "learning_rate": 4.6847268530392415e-05, + "loss": 0.6306, + "step": 4967 + }, + { + "epoch": 0.19, + "learning_rate": 4.684662734034368e-05, + "loss": 0.6658, + "step": 4968 + }, + { + "epoch": 0.19, + "learning_rate": 4.684598615029495e-05, + "loss": 0.5903, + "step": 4969 + }, + { + "epoch": 0.19, + "learning_rate": 4.6845344960246216e-05, + "loss": 0.6242, + "step": 4970 + }, + { + "epoch": 0.19, + "learning_rate": 4.684470377019749e-05, + "loss": 0.5466, + "step": 4971 + }, + { + "epoch": 0.19, + "learning_rate": 4.684406258014876e-05, + "loss": 0.6745, + "step": 4972 + }, + { + "epoch": 0.19, + "learning_rate": 4.684342139010003e-05, + "loss": 0.5747, + "step": 4973 + }, + { + "epoch": 0.19, + "learning_rate": 4.6842780200051296e-05, + "loss": 0.6543, + "step": 4974 + }, + { + "epoch": 0.19, + "learning_rate": 4.684213901000257e-05, + "loss": 0.6193, + "step": 4975 + }, + { + "epoch": 0.19, + "learning_rate": 4.684149781995384e-05, + "loss": 0.5615, + "step": 4976 + }, + { + "epoch": 0.19, + "learning_rate": 4.6840856629905105e-05, + "loss": 0.6949, + "step": 4977 + }, + { + "epoch": 0.19, + "learning_rate": 4.6840215439856376e-05, + "loss": 0.5501, + "step": 4978 + }, + { + "epoch": 0.19, + "learning_rate": 4.683957424980764e-05, + "loss": 0.6365, + "step": 4979 + }, + { + "epoch": 0.19, + "learning_rate": 4.683893305975891e-05, + "loss": 0.6164, + "step": 4980 + }, + { + "epoch": 0.19, + "learning_rate": 4.6838291869710185e-05, + "loss": 0.6008, + "step": 4981 + }, + { + "epoch": 0.19, + "learning_rate": 4.6837650679661456e-05, + "loss": 0.5691, + "step": 4982 + }, + { + "epoch": 0.19, + "learning_rate": 4.683700948961272e-05, + "loss": 0.5886, + "step": 4983 + }, + { + "epoch": 0.19, + "learning_rate": 4.683636829956399e-05, + "loss": 0.6141, + "step": 4984 + }, + { + "epoch": 0.19, + "learning_rate": 4.6835727109515265e-05, + "loss": 0.6997, + "step": 4985 + }, + { + "epoch": 0.19, + "learning_rate": 4.683508591946653e-05, + "loss": 0.5545, + "step": 4986 + }, + { + "epoch": 0.19, + "learning_rate": 4.68344447294178e-05, + "loss": 0.6489, + "step": 4987 + }, + { + "epoch": 0.19, + "learning_rate": 4.683380353936907e-05, + "loss": 0.5465, + "step": 4988 + }, + { + "epoch": 0.19, + "learning_rate": 4.6833162349320345e-05, + "loss": 0.6494, + "step": 4989 + }, + { + "epoch": 0.19, + "learning_rate": 4.683252115927161e-05, + "loss": 0.629, + "step": 4990 + }, + { + "epoch": 0.19, + "learning_rate": 4.683187996922288e-05, + "loss": 0.5672, + "step": 4991 + }, + { + "epoch": 0.19, + "learning_rate": 4.6831238779174147e-05, + "loss": 0.6275, + "step": 4992 + }, + { + "epoch": 0.19, + "learning_rate": 4.683059758912542e-05, + "loss": 0.5966, + "step": 4993 + }, + { + "epoch": 0.19, + "learning_rate": 4.682995639907669e-05, + "loss": 0.5555, + "step": 4994 + }, + { + "epoch": 0.19, + "learning_rate": 4.6829315209027955e-05, + "loss": 0.6053, + "step": 4995 + }, + { + "epoch": 0.19, + "learning_rate": 4.682867401897923e-05, + "loss": 0.6393, + "step": 4996 + }, + { + "epoch": 0.19, + "learning_rate": 4.68280328289305e-05, + "loss": 0.6314, + "step": 4997 + }, + { + "epoch": 0.19, + "learning_rate": 4.682739163888177e-05, + "loss": 0.5826, + "step": 4998 + }, + { + "epoch": 0.19, + "learning_rate": 4.6826750448833035e-05, + "loss": 0.5955, + "step": 4999 + }, + { + "epoch": 0.19, + "learning_rate": 4.682610925878431e-05, + "loss": 0.5731, + "step": 5000 + }, + { + "epoch": 0.19, + "learning_rate": 4.682546806873557e-05, + "loss": 0.6276, + "step": 5001 + }, + { + "epoch": 0.19, + "learning_rate": 4.6824826878686843e-05, + "loss": 0.5248, + "step": 5002 + }, + { + "epoch": 0.19, + "learning_rate": 4.6824185688638115e-05, + "loss": 0.6036, + "step": 5003 + }, + { + "epoch": 0.19, + "learning_rate": 4.682354449858938e-05, + "loss": 0.5736, + "step": 5004 + }, + { + "epoch": 0.19, + "learning_rate": 4.682290330854066e-05, + "loss": 0.6041, + "step": 5005 + }, + { + "epoch": 0.19, + "learning_rate": 4.6822262118491923e-05, + "loss": 0.6786, + "step": 5006 + }, + { + "epoch": 0.19, + "learning_rate": 4.6821620928443195e-05, + "loss": 0.5703, + "step": 5007 + }, + { + "epoch": 0.19, + "learning_rate": 4.682097973839446e-05, + "loss": 0.6433, + "step": 5008 + }, + { + "epoch": 0.19, + "learning_rate": 4.682033854834573e-05, + "loss": 0.5022, + "step": 5009 + }, + { + "epoch": 0.19, + "learning_rate": 4.6819697358297e-05, + "loss": 0.637, + "step": 5010 + }, + { + "epoch": 0.19, + "learning_rate": 4.681905616824827e-05, + "loss": 0.6253, + "step": 5011 + }, + { + "epoch": 0.19, + "learning_rate": 4.681841497819954e-05, + "loss": 0.6766, + "step": 5012 + }, + { + "epoch": 0.19, + "learning_rate": 4.681777378815081e-05, + "loss": 0.5623, + "step": 5013 + }, + { + "epoch": 0.19, + "learning_rate": 4.6817132598102084e-05, + "loss": 0.5592, + "step": 5014 + }, + { + "epoch": 0.19, + "learning_rate": 4.681649140805335e-05, + "loss": 0.5491, + "step": 5015 + }, + { + "epoch": 0.19, + "learning_rate": 4.681585021800462e-05, + "loss": 0.5824, + "step": 5016 + }, + { + "epoch": 0.19, + "learning_rate": 4.6815209027955885e-05, + "loss": 0.5621, + "step": 5017 + }, + { + "epoch": 0.19, + "learning_rate": 4.681456783790716e-05, + "loss": 0.5155, + "step": 5018 + }, + { + "epoch": 0.19, + "learning_rate": 4.681392664785842e-05, + "loss": 0.5783, + "step": 5019 + }, + { + "epoch": 0.19, + "learning_rate": 4.6813285457809694e-05, + "loss": 0.5397, + "step": 5020 + }, + { + "epoch": 0.19, + "learning_rate": 4.681264426776097e-05, + "loss": 0.5839, + "step": 5021 + }, + { + "epoch": 0.19, + "learning_rate": 4.681200307771224e-05, + "loss": 0.6245, + "step": 5022 + }, + { + "epoch": 0.19, + "learning_rate": 4.681136188766351e-05, + "loss": 0.6327, + "step": 5023 + }, + { + "epoch": 0.19, + "learning_rate": 4.6810720697614774e-05, + "loss": 0.5642, + "step": 5024 + }, + { + "epoch": 0.19, + "learning_rate": 4.6810079507566045e-05, + "loss": 0.5198, + "step": 5025 + }, + { + "epoch": 0.19, + "learning_rate": 4.680943831751731e-05, + "loss": 0.58, + "step": 5026 + }, + { + "epoch": 0.19, + "learning_rate": 4.680879712746858e-05, + "loss": 0.6232, + "step": 5027 + }, + { + "epoch": 0.19, + "learning_rate": 4.6808155937419854e-05, + "loss": 0.6513, + "step": 5028 + }, + { + "epoch": 0.19, + "learning_rate": 4.6807514747371126e-05, + "loss": 0.6342, + "step": 5029 + }, + { + "epoch": 0.19, + "learning_rate": 4.68068735573224e-05, + "loss": 0.6206, + "step": 5030 + }, + { + "epoch": 0.19, + "learning_rate": 4.680623236727366e-05, + "loss": 0.585, + "step": 5031 + }, + { + "epoch": 0.19, + "learning_rate": 4.6805591177224934e-05, + "loss": 0.6504, + "step": 5032 + }, + { + "epoch": 0.19, + "learning_rate": 4.68049499871762e-05, + "loss": 0.6458, + "step": 5033 + }, + { + "epoch": 0.19, + "learning_rate": 4.680430879712747e-05, + "loss": 0.5137, + "step": 5034 + }, + { + "epoch": 0.19, + "learning_rate": 4.6803667607078736e-05, + "loss": 0.5859, + "step": 5035 + }, + { + "epoch": 0.19, + "learning_rate": 4.680302641703001e-05, + "loss": 0.554, + "step": 5036 + }, + { + "epoch": 0.19, + "learning_rate": 4.680238522698128e-05, + "loss": 0.5356, + "step": 5037 + }, + { + "epoch": 0.19, + "learning_rate": 4.680174403693255e-05, + "loss": 0.6332, + "step": 5038 + }, + { + "epoch": 0.19, + "learning_rate": 4.680110284688382e-05, + "loss": 0.5437, + "step": 5039 + }, + { + "epoch": 0.19, + "learning_rate": 4.680046165683509e-05, + "loss": 0.6007, + "step": 5040 + }, + { + "epoch": 0.19, + "learning_rate": 4.679982046678636e-05, + "loss": 0.5872, + "step": 5041 + }, + { + "epoch": 0.19, + "learning_rate": 4.6799179276737624e-05, + "loss": 0.6196, + "step": 5042 + }, + { + "epoch": 0.19, + "learning_rate": 4.6798538086688896e-05, + "loss": 0.6961, + "step": 5043 + }, + { + "epoch": 0.19, + "learning_rate": 4.679789689664017e-05, + "loss": 0.5139, + "step": 5044 + }, + { + "epoch": 0.19, + "learning_rate": 4.679725570659144e-05, + "loss": 0.4955, + "step": 5045 + }, + { + "epoch": 0.19, + "learning_rate": 4.6796614516542704e-05, + "loss": 0.5502, + "step": 5046 + }, + { + "epoch": 0.19, + "learning_rate": 4.6795973326493976e-05, + "loss": 0.577, + "step": 5047 + }, + { + "epoch": 0.19, + "learning_rate": 4.679533213644525e-05, + "loss": 0.5314, + "step": 5048 + }, + { + "epoch": 0.19, + "learning_rate": 4.679469094639651e-05, + "loss": 0.5513, + "step": 5049 + }, + { + "epoch": 0.19, + "learning_rate": 4.6794049756347784e-05, + "loss": 0.5778, + "step": 5050 + }, + { + "epoch": 0.19, + "learning_rate": 4.679340856629905e-05, + "loss": 0.6601, + "step": 5051 + }, + { + "epoch": 0.19, + "learning_rate": 4.679276737625032e-05, + "loss": 0.5792, + "step": 5052 + }, + { + "epoch": 0.19, + "learning_rate": 4.679212618620159e-05, + "loss": 0.5804, + "step": 5053 + }, + { + "epoch": 0.19, + "learning_rate": 4.6791484996152864e-05, + "loss": 0.6209, + "step": 5054 + }, + { + "epoch": 0.19, + "learning_rate": 4.679084380610413e-05, + "loss": 0.6258, + "step": 5055 + }, + { + "epoch": 0.19, + "learning_rate": 4.67902026160554e-05, + "loss": 0.6085, + "step": 5056 + }, + { + "epoch": 0.19, + "learning_rate": 4.678956142600667e-05, + "loss": 0.6191, + "step": 5057 + }, + { + "epoch": 0.19, + "learning_rate": 4.678892023595794e-05, + "loss": 0.5227, + "step": 5058 + }, + { + "epoch": 0.19, + "learning_rate": 4.678827904590921e-05, + "loss": 0.5953, + "step": 5059 + }, + { + "epoch": 0.19, + "learning_rate": 4.678763785586048e-05, + "loss": 0.5491, + "step": 5060 + }, + { + "epoch": 0.19, + "learning_rate": 4.678699666581175e-05, + "loss": 0.5294, + "step": 5061 + }, + { + "epoch": 0.19, + "learning_rate": 4.678635547576302e-05, + "loss": 0.6189, + "step": 5062 + }, + { + "epoch": 0.19, + "learning_rate": 4.678571428571429e-05, + "loss": 0.5825, + "step": 5063 + }, + { + "epoch": 0.19, + "learning_rate": 4.6785073095665554e-05, + "loss": 0.677, + "step": 5064 + }, + { + "epoch": 0.19, + "learning_rate": 4.6784431905616826e-05, + "loss": 0.5888, + "step": 5065 + }, + { + "epoch": 0.19, + "learning_rate": 4.67837907155681e-05, + "loss": 0.6305, + "step": 5066 + }, + { + "epoch": 0.19, + "learning_rate": 4.678314952551936e-05, + "loss": 0.588, + "step": 5067 + }, + { + "epoch": 0.19, + "learning_rate": 4.6782508335470634e-05, + "loss": 0.6132, + "step": 5068 + }, + { + "epoch": 0.19, + "learning_rate": 4.6781867145421906e-05, + "loss": 0.6411, + "step": 5069 + }, + { + "epoch": 0.19, + "learning_rate": 4.678122595537318e-05, + "loss": 0.5999, + "step": 5070 + }, + { + "epoch": 0.19, + "learning_rate": 4.678058476532444e-05, + "loss": 0.7271, + "step": 5071 + }, + { + "epoch": 0.2, + "learning_rate": 4.6779943575275714e-05, + "loss": 0.573, + "step": 5072 + }, + { + "epoch": 0.2, + "learning_rate": 4.6779302385226986e-05, + "loss": 0.5538, + "step": 5073 + }, + { + "epoch": 0.2, + "learning_rate": 4.677866119517825e-05, + "loss": 0.654, + "step": 5074 + }, + { + "epoch": 0.2, + "learning_rate": 4.677802000512952e-05, + "loss": 0.5867, + "step": 5075 + }, + { + "epoch": 0.2, + "learning_rate": 4.6777378815080795e-05, + "loss": 0.5143, + "step": 5076 + }, + { + "epoch": 0.2, + "learning_rate": 4.6776737625032066e-05, + "loss": 0.5868, + "step": 5077 + }, + { + "epoch": 0.2, + "learning_rate": 4.677609643498333e-05, + "loss": 0.5615, + "step": 5078 + }, + { + "epoch": 0.2, + "learning_rate": 4.67754552449346e-05, + "loss": 0.5795, + "step": 5079 + }, + { + "epoch": 0.2, + "learning_rate": 4.677481405488587e-05, + "loss": 0.6157, + "step": 5080 + }, + { + "epoch": 0.2, + "learning_rate": 4.677417286483714e-05, + "loss": 0.6026, + "step": 5081 + }, + { + "epoch": 0.2, + "learning_rate": 4.677353167478841e-05, + "loss": 0.5639, + "step": 5082 + }, + { + "epoch": 0.2, + "learning_rate": 4.6772890484739676e-05, + "loss": 0.5639, + "step": 5083 + }, + { + "epoch": 0.2, + "learning_rate": 4.677224929469095e-05, + "loss": 0.567, + "step": 5084 + }, + { + "epoch": 0.2, + "learning_rate": 4.677160810464222e-05, + "loss": 0.6273, + "step": 5085 + }, + { + "epoch": 0.2, + "learning_rate": 4.677096691459349e-05, + "loss": 0.5433, + "step": 5086 + }, + { + "epoch": 0.2, + "learning_rate": 4.6770325724544756e-05, + "loss": 0.5786, + "step": 5087 + }, + { + "epoch": 0.2, + "learning_rate": 4.676968453449603e-05, + "loss": 0.6174, + "step": 5088 + }, + { + "epoch": 0.2, + "learning_rate": 4.676904334444729e-05, + "loss": 0.6047, + "step": 5089 + }, + { + "epoch": 0.2, + "learning_rate": 4.6768402154398565e-05, + "loss": 0.5787, + "step": 5090 + }, + { + "epoch": 0.2, + "learning_rate": 4.6767760964349836e-05, + "loss": 0.6221, + "step": 5091 + }, + { + "epoch": 0.2, + "learning_rate": 4.67671197743011e-05, + "loss": 0.6555, + "step": 5092 + }, + { + "epoch": 0.2, + "learning_rate": 4.676647858425238e-05, + "loss": 0.6033, + "step": 5093 + }, + { + "epoch": 0.2, + "learning_rate": 4.6765837394203645e-05, + "loss": 0.5402, + "step": 5094 + }, + { + "epoch": 0.2, + "learning_rate": 4.6765196204154917e-05, + "loss": 0.5503, + "step": 5095 + }, + { + "epoch": 0.2, + "learning_rate": 4.676455501410618e-05, + "loss": 0.5423, + "step": 5096 + }, + { + "epoch": 0.2, + "learning_rate": 4.676391382405745e-05, + "loss": 0.6089, + "step": 5097 + }, + { + "epoch": 0.2, + "learning_rate": 4.676327263400872e-05, + "loss": 0.596, + "step": 5098 + }, + { + "epoch": 0.2, + "learning_rate": 4.676263144395999e-05, + "loss": 0.6027, + "step": 5099 + }, + { + "epoch": 0.2, + "learning_rate": 4.676199025391126e-05, + "loss": 0.6439, + "step": 5100 + }, + { + "epoch": 0.2, + "learning_rate": 4.676134906386253e-05, + "loss": 0.5456, + "step": 5101 + }, + { + "epoch": 0.2, + "learning_rate": 4.6760707873813805e-05, + "loss": 0.5536, + "step": 5102 + }, + { + "epoch": 0.2, + "learning_rate": 4.676006668376507e-05, + "loss": 0.5809, + "step": 5103 + }, + { + "epoch": 0.2, + "learning_rate": 4.675942549371634e-05, + "loss": 0.5418, + "step": 5104 + }, + { + "epoch": 0.2, + "learning_rate": 4.6758784303667607e-05, + "loss": 0.5793, + "step": 5105 + }, + { + "epoch": 0.2, + "learning_rate": 4.675814311361888e-05, + "loss": 0.6781, + "step": 5106 + }, + { + "epoch": 0.2, + "learning_rate": 4.675750192357014e-05, + "loss": 0.5759, + "step": 5107 + }, + { + "epoch": 0.2, + "learning_rate": 4.6756860733521415e-05, + "loss": 0.5901, + "step": 5108 + }, + { + "epoch": 0.2, + "learning_rate": 4.6756219543472693e-05, + "loss": 0.6193, + "step": 5109 + }, + { + "epoch": 0.2, + "learning_rate": 4.675557835342396e-05, + "loss": 0.488, + "step": 5110 + }, + { + "epoch": 0.2, + "learning_rate": 4.675493716337523e-05, + "loss": 0.6391, + "step": 5111 + }, + { + "epoch": 0.2, + "learning_rate": 4.6754295973326495e-05, + "loss": 0.5785, + "step": 5112 + }, + { + "epoch": 0.2, + "learning_rate": 4.675365478327777e-05, + "loss": 0.5507, + "step": 5113 + }, + { + "epoch": 0.2, + "learning_rate": 4.675301359322903e-05, + "loss": 0.6194, + "step": 5114 + }, + { + "epoch": 0.2, + "learning_rate": 4.6752372403180303e-05, + "loss": 0.6225, + "step": 5115 + }, + { + "epoch": 0.2, + "learning_rate": 4.6751731213131575e-05, + "loss": 0.7444, + "step": 5116 + }, + { + "epoch": 0.2, + "learning_rate": 4.675109002308285e-05, + "loss": 0.557, + "step": 5117 + }, + { + "epoch": 0.2, + "learning_rate": 4.675044883303412e-05, + "loss": 0.6107, + "step": 5118 + }, + { + "epoch": 0.2, + "learning_rate": 4.6749807642985384e-05, + "loss": 0.5304, + "step": 5119 + }, + { + "epoch": 0.2, + "learning_rate": 4.6749166452936655e-05, + "loss": 0.6371, + "step": 5120 + }, + { + "epoch": 0.2, + "learning_rate": 4.674852526288792e-05, + "loss": 0.6386, + "step": 5121 + }, + { + "epoch": 0.2, + "learning_rate": 4.674788407283919e-05, + "loss": 0.6243, + "step": 5122 + }, + { + "epoch": 0.2, + "learning_rate": 4.674724288279046e-05, + "loss": 0.5942, + "step": 5123 + }, + { + "epoch": 0.2, + "learning_rate": 4.674660169274173e-05, + "loss": 0.6242, + "step": 5124 + }, + { + "epoch": 0.2, + "learning_rate": 4.6745960502693e-05, + "loss": 0.5982, + "step": 5125 + }, + { + "epoch": 0.2, + "learning_rate": 4.674531931264427e-05, + "loss": 0.5626, + "step": 5126 + }, + { + "epoch": 0.2, + "learning_rate": 4.6744678122595544e-05, + "loss": 0.5969, + "step": 5127 + }, + { + "epoch": 0.2, + "learning_rate": 4.674403693254681e-05, + "loss": 0.5378, + "step": 5128 + }, + { + "epoch": 0.2, + "learning_rate": 4.674339574249808e-05, + "loss": 0.6174, + "step": 5129 + }, + { + "epoch": 0.2, + "learning_rate": 4.6742754552449345e-05, + "loss": 0.5376, + "step": 5130 + }, + { + "epoch": 0.2, + "learning_rate": 4.674211336240062e-05, + "loss": 0.5881, + "step": 5131 + }, + { + "epoch": 0.2, + "learning_rate": 4.674147217235189e-05, + "loss": 0.5697, + "step": 5132 + }, + { + "epoch": 0.2, + "learning_rate": 4.674083098230316e-05, + "loss": 0.6, + "step": 5133 + }, + { + "epoch": 0.2, + "learning_rate": 4.6740189792254425e-05, + "loss": 0.5468, + "step": 5134 + }, + { + "epoch": 0.2, + "learning_rate": 4.67395486022057e-05, + "loss": 0.546, + "step": 5135 + }, + { + "epoch": 0.2, + "learning_rate": 4.673890741215697e-05, + "loss": 0.4769, + "step": 5136 + }, + { + "epoch": 0.2, + "learning_rate": 4.6738266222108234e-05, + "loss": 0.5646, + "step": 5137 + }, + { + "epoch": 0.2, + "learning_rate": 4.6737625032059505e-05, + "loss": 0.5868, + "step": 5138 + }, + { + "epoch": 0.2, + "learning_rate": 4.673698384201077e-05, + "loss": 0.5825, + "step": 5139 + }, + { + "epoch": 0.2, + "learning_rate": 4.673634265196204e-05, + "loss": 0.4626, + "step": 5140 + }, + { + "epoch": 0.2, + "learning_rate": 4.6735701461913314e-05, + "loss": 0.628, + "step": 5141 + }, + { + "epoch": 0.2, + "learning_rate": 4.6735060271864586e-05, + "loss": 0.5928, + "step": 5142 + }, + { + "epoch": 0.2, + "learning_rate": 4.673441908181585e-05, + "loss": 0.5738, + "step": 5143 + }, + { + "epoch": 0.2, + "learning_rate": 4.673377789176712e-05, + "loss": 0.568, + "step": 5144 + }, + { + "epoch": 0.2, + "learning_rate": 4.6733136701718394e-05, + "loss": 0.5169, + "step": 5145 + }, + { + "epoch": 0.2, + "learning_rate": 4.673249551166966e-05, + "loss": 0.512, + "step": 5146 + }, + { + "epoch": 0.2, + "learning_rate": 4.673185432162093e-05, + "loss": 0.634, + "step": 5147 + }, + { + "epoch": 0.2, + "learning_rate": 4.67312131315722e-05, + "loss": 0.6587, + "step": 5148 + }, + { + "epoch": 0.2, + "learning_rate": 4.6730571941523474e-05, + "loss": 0.5889, + "step": 5149 + }, + { + "epoch": 0.2, + "learning_rate": 4.672993075147474e-05, + "loss": 0.6714, + "step": 5150 + }, + { + "epoch": 0.2, + "learning_rate": 4.672928956142601e-05, + "loss": 0.5582, + "step": 5151 + }, + { + "epoch": 0.2, + "learning_rate": 4.6728648371377276e-05, + "loss": 0.5783, + "step": 5152 + }, + { + "epoch": 0.2, + "learning_rate": 4.672800718132855e-05, + "loss": 0.6975, + "step": 5153 + }, + { + "epoch": 0.2, + "learning_rate": 4.672736599127982e-05, + "loss": 0.5943, + "step": 5154 + }, + { + "epoch": 0.2, + "learning_rate": 4.6726724801231084e-05, + "loss": 0.5627, + "step": 5155 + }, + { + "epoch": 0.2, + "learning_rate": 4.6726083611182356e-05, + "loss": 0.5018, + "step": 5156 + }, + { + "epoch": 0.2, + "learning_rate": 4.672544242113363e-05, + "loss": 0.6527, + "step": 5157 + }, + { + "epoch": 0.2, + "learning_rate": 4.67248012310849e-05, + "loss": 0.6742, + "step": 5158 + }, + { + "epoch": 0.2, + "learning_rate": 4.6724160041036164e-05, + "loss": 0.5349, + "step": 5159 + }, + { + "epoch": 0.2, + "learning_rate": 4.6723518850987436e-05, + "loss": 0.6326, + "step": 5160 + }, + { + "epoch": 0.2, + "learning_rate": 4.67228776609387e-05, + "loss": 0.6218, + "step": 5161 + }, + { + "epoch": 0.2, + "learning_rate": 4.672223647088997e-05, + "loss": 0.6196, + "step": 5162 + }, + { + "epoch": 0.2, + "learning_rate": 4.6721595280841244e-05, + "loss": 0.6017, + "step": 5163 + }, + { + "epoch": 0.2, + "learning_rate": 4.6720954090792516e-05, + "loss": 0.5623, + "step": 5164 + }, + { + "epoch": 0.2, + "learning_rate": 4.672031290074379e-05, + "loss": 0.5695, + "step": 5165 + }, + { + "epoch": 0.2, + "learning_rate": 4.671967171069505e-05, + "loss": 0.6177, + "step": 5166 + }, + { + "epoch": 0.2, + "learning_rate": 4.6719030520646324e-05, + "loss": 0.5574, + "step": 5167 + }, + { + "epoch": 0.2, + "learning_rate": 4.671838933059759e-05, + "loss": 0.5576, + "step": 5168 + }, + { + "epoch": 0.2, + "learning_rate": 4.671774814054886e-05, + "loss": 0.58, + "step": 5169 + }, + { + "epoch": 0.2, + "learning_rate": 4.6717106950500126e-05, + "loss": 0.5317, + "step": 5170 + }, + { + "epoch": 0.2, + "learning_rate": 4.67164657604514e-05, + "loss": 0.5148, + "step": 5171 + }, + { + "epoch": 0.2, + "learning_rate": 4.671582457040267e-05, + "loss": 0.6196, + "step": 5172 + }, + { + "epoch": 0.2, + "learning_rate": 4.671518338035394e-05, + "loss": 0.6438, + "step": 5173 + }, + { + "epoch": 0.2, + "learning_rate": 4.671454219030521e-05, + "loss": 0.6089, + "step": 5174 + }, + { + "epoch": 0.2, + "learning_rate": 4.671390100025648e-05, + "loss": 0.6583, + "step": 5175 + }, + { + "epoch": 0.2, + "learning_rate": 4.671325981020775e-05, + "loss": 0.5577, + "step": 5176 + }, + { + "epoch": 0.2, + "learning_rate": 4.6712618620159014e-05, + "loss": 0.6854, + "step": 5177 + }, + { + "epoch": 0.2, + "learning_rate": 4.6711977430110286e-05, + "loss": 0.5523, + "step": 5178 + }, + { + "epoch": 0.2, + "learning_rate": 4.671133624006155e-05, + "loss": 0.698, + "step": 5179 + }, + { + "epoch": 0.2, + "learning_rate": 4.671069505001282e-05, + "loss": 0.6223, + "step": 5180 + }, + { + "epoch": 0.2, + "learning_rate": 4.67100538599641e-05, + "loss": 0.5072, + "step": 5181 + }, + { + "epoch": 0.2, + "learning_rate": 4.6709412669915366e-05, + "loss": 0.546, + "step": 5182 + }, + { + "epoch": 0.2, + "learning_rate": 4.670877147986664e-05, + "loss": 0.5211, + "step": 5183 + }, + { + "epoch": 0.2, + "learning_rate": 4.67081302898179e-05, + "loss": 0.5804, + "step": 5184 + }, + { + "epoch": 0.2, + "learning_rate": 4.6707489099769175e-05, + "loss": 0.5962, + "step": 5185 + }, + { + "epoch": 0.2, + "learning_rate": 4.670684790972044e-05, + "loss": 0.6097, + "step": 5186 + }, + { + "epoch": 0.2, + "learning_rate": 4.670620671967171e-05, + "loss": 0.5772, + "step": 5187 + }, + { + "epoch": 0.2, + "learning_rate": 4.670556552962298e-05, + "loss": 0.5714, + "step": 5188 + }, + { + "epoch": 0.2, + "learning_rate": 4.6704924339574255e-05, + "loss": 0.6043, + "step": 5189 + }, + { + "epoch": 0.2, + "learning_rate": 4.6704283149525526e-05, + "loss": 0.639, + "step": 5190 + }, + { + "epoch": 0.2, + "learning_rate": 4.670364195947679e-05, + "loss": 0.6737, + "step": 5191 + }, + { + "epoch": 0.2, + "learning_rate": 4.670300076942806e-05, + "loss": 0.6327, + "step": 5192 + }, + { + "epoch": 0.2, + "learning_rate": 4.670235957937933e-05, + "loss": 0.6581, + "step": 5193 + }, + { + "epoch": 0.2, + "learning_rate": 4.67017183893306e-05, + "loss": 0.5306, + "step": 5194 + }, + { + "epoch": 0.2, + "learning_rate": 4.6701077199281865e-05, + "loss": 0.6201, + "step": 5195 + }, + { + "epoch": 0.2, + "learning_rate": 4.6700436009233136e-05, + "loss": 0.5382, + "step": 5196 + }, + { + "epoch": 0.2, + "learning_rate": 4.669979481918441e-05, + "loss": 0.6625, + "step": 5197 + }, + { + "epoch": 0.2, + "learning_rate": 4.669915362913568e-05, + "loss": 0.6208, + "step": 5198 + }, + { + "epoch": 0.2, + "learning_rate": 4.669851243908695e-05, + "loss": 0.6254, + "step": 5199 + }, + { + "epoch": 0.2, + "learning_rate": 4.6697871249038216e-05, + "loss": 0.624, + "step": 5200 + }, + { + "epoch": 0.2, + "learning_rate": 4.669723005898949e-05, + "loss": 0.5401, + "step": 5201 + }, + { + "epoch": 0.2, + "learning_rate": 4.669658886894075e-05, + "loss": 0.5562, + "step": 5202 + }, + { + "epoch": 0.2, + "learning_rate": 4.6695947678892025e-05, + "loss": 0.5306, + "step": 5203 + }, + { + "epoch": 0.2, + "learning_rate": 4.6695306488843296e-05, + "loss": 0.528, + "step": 5204 + }, + { + "epoch": 0.2, + "learning_rate": 4.669466529879457e-05, + "loss": 0.5939, + "step": 5205 + }, + { + "epoch": 0.2, + "learning_rate": 4.669402410874583e-05, + "loss": 0.6019, + "step": 5206 + }, + { + "epoch": 0.2, + "learning_rate": 4.6693382918697105e-05, + "loss": 0.5868, + "step": 5207 + }, + { + "epoch": 0.2, + "learning_rate": 4.6692741728648377e-05, + "loss": 0.6379, + "step": 5208 + }, + { + "epoch": 0.2, + "learning_rate": 4.669210053859964e-05, + "loss": 0.5879, + "step": 5209 + }, + { + "epoch": 0.2, + "learning_rate": 4.669145934855091e-05, + "loss": 0.6178, + "step": 5210 + }, + { + "epoch": 0.2, + "learning_rate": 4.669081815850218e-05, + "loss": 0.5871, + "step": 5211 + }, + { + "epoch": 0.2, + "learning_rate": 4.669017696845345e-05, + "loss": 0.5746, + "step": 5212 + }, + { + "epoch": 0.2, + "learning_rate": 4.668953577840472e-05, + "loss": 0.5611, + "step": 5213 + }, + { + "epoch": 0.2, + "learning_rate": 4.668889458835599e-05, + "loss": 0.5741, + "step": 5214 + }, + { + "epoch": 0.2, + "learning_rate": 4.668825339830726e-05, + "loss": 0.5771, + "step": 5215 + }, + { + "epoch": 0.2, + "learning_rate": 4.668761220825853e-05, + "loss": 0.5523, + "step": 5216 + }, + { + "epoch": 0.2, + "learning_rate": 4.66869710182098e-05, + "loss": 0.5295, + "step": 5217 + }, + { + "epoch": 0.2, + "learning_rate": 4.668632982816107e-05, + "loss": 0.5763, + "step": 5218 + }, + { + "epoch": 0.2, + "learning_rate": 4.668568863811234e-05, + "loss": 0.6377, + "step": 5219 + }, + { + "epoch": 0.2, + "learning_rate": 4.668504744806361e-05, + "loss": 0.5776, + "step": 5220 + }, + { + "epoch": 0.2, + "learning_rate": 4.668440625801488e-05, + "loss": 0.5157, + "step": 5221 + }, + { + "epoch": 0.2, + "learning_rate": 4.668376506796615e-05, + "loss": 0.5782, + "step": 5222 + }, + { + "epoch": 0.2, + "learning_rate": 4.668312387791742e-05, + "loss": 0.5513, + "step": 5223 + }, + { + "epoch": 0.2, + "learning_rate": 4.668248268786868e-05, + "loss": 0.5229, + "step": 5224 + }, + { + "epoch": 0.2, + "learning_rate": 4.6681841497819955e-05, + "loss": 0.6252, + "step": 5225 + }, + { + "epoch": 0.2, + "learning_rate": 4.668120030777123e-05, + "loss": 0.655, + "step": 5226 + }, + { + "epoch": 0.2, + "learning_rate": 4.668055911772249e-05, + "loss": 0.6757, + "step": 5227 + }, + { + "epoch": 0.2, + "learning_rate": 4.6679917927673763e-05, + "loss": 0.6246, + "step": 5228 + }, + { + "epoch": 0.2, + "learning_rate": 4.6679276737625035e-05, + "loss": 0.5907, + "step": 5229 + }, + { + "epoch": 0.2, + "learning_rate": 4.667863554757631e-05, + "loss": 0.6046, + "step": 5230 + }, + { + "epoch": 0.2, + "learning_rate": 4.667799435752757e-05, + "loss": 0.5686, + "step": 5231 + }, + { + "epoch": 0.2, + "learning_rate": 4.6677353167478844e-05, + "loss": 0.5821, + "step": 5232 + }, + { + "epoch": 0.2, + "learning_rate": 4.667671197743011e-05, + "loss": 0.6616, + "step": 5233 + }, + { + "epoch": 0.2, + "learning_rate": 4.667607078738138e-05, + "loss": 0.5814, + "step": 5234 + }, + { + "epoch": 0.2, + "learning_rate": 4.667542959733265e-05, + "loss": 0.5914, + "step": 5235 + }, + { + "epoch": 0.2, + "learning_rate": 4.6674788407283924e-05, + "loss": 0.581, + "step": 5236 + }, + { + "epoch": 0.2, + "learning_rate": 4.6674147217235195e-05, + "loss": 0.5701, + "step": 5237 + }, + { + "epoch": 0.2, + "learning_rate": 4.667350602718646e-05, + "loss": 0.5772, + "step": 5238 + }, + { + "epoch": 0.2, + "learning_rate": 4.667286483713773e-05, + "loss": 0.6093, + "step": 5239 + }, + { + "epoch": 0.2, + "learning_rate": 4.6672223647089e-05, + "loss": 0.5635, + "step": 5240 + }, + { + "epoch": 0.2, + "learning_rate": 4.667158245704027e-05, + "loss": 0.5625, + "step": 5241 + }, + { + "epoch": 0.2, + "learning_rate": 4.6670941266991534e-05, + "loss": 0.5888, + "step": 5242 + }, + { + "epoch": 0.2, + "learning_rate": 4.6670300076942805e-05, + "loss": 0.6004, + "step": 5243 + }, + { + "epoch": 0.2, + "learning_rate": 4.666965888689408e-05, + "loss": 0.6015, + "step": 5244 + }, + { + "epoch": 0.2, + "learning_rate": 4.666901769684535e-05, + "loss": 0.5768, + "step": 5245 + }, + { + "epoch": 0.2, + "learning_rate": 4.666837650679662e-05, + "loss": 0.6725, + "step": 5246 + }, + { + "epoch": 0.2, + "learning_rate": 4.6667735316747885e-05, + "loss": 0.4963, + "step": 5247 + }, + { + "epoch": 0.2, + "learning_rate": 4.666709412669916e-05, + "loss": 0.5127, + "step": 5248 + }, + { + "epoch": 0.2, + "learning_rate": 4.666645293665042e-05, + "loss": 0.6378, + "step": 5249 + }, + { + "epoch": 0.2, + "learning_rate": 4.6665811746601694e-05, + "loss": 0.7011, + "step": 5250 + }, + { + "epoch": 0.2, + "learning_rate": 4.666517055655296e-05, + "loss": 0.6631, + "step": 5251 + }, + { + "epoch": 0.2, + "learning_rate": 4.666452936650423e-05, + "loss": 0.6314, + "step": 5252 + }, + { + "epoch": 0.2, + "learning_rate": 4.666388817645551e-05, + "loss": 0.7052, + "step": 5253 + }, + { + "epoch": 0.2, + "learning_rate": 4.6663246986406774e-05, + "loss": 0.5302, + "step": 5254 + }, + { + "epoch": 0.2, + "learning_rate": 4.6662605796358046e-05, + "loss": 0.5898, + "step": 5255 + }, + { + "epoch": 0.2, + "learning_rate": 4.666196460630931e-05, + "loss": 0.5882, + "step": 5256 + }, + { + "epoch": 0.2, + "learning_rate": 4.666132341626058e-05, + "loss": 0.5771, + "step": 5257 + }, + { + "epoch": 0.2, + "learning_rate": 4.666068222621185e-05, + "loss": 0.5704, + "step": 5258 + }, + { + "epoch": 0.2, + "learning_rate": 4.666004103616312e-05, + "loss": 0.6902, + "step": 5259 + }, + { + "epoch": 0.2, + "learning_rate": 4.665939984611439e-05, + "loss": 0.6392, + "step": 5260 + }, + { + "epoch": 0.2, + "learning_rate": 4.665875865606566e-05, + "loss": 0.6004, + "step": 5261 + }, + { + "epoch": 0.2, + "learning_rate": 4.6658117466016934e-05, + "loss": 0.5555, + "step": 5262 + }, + { + "epoch": 0.2, + "learning_rate": 4.66574762759682e-05, + "loss": 0.4959, + "step": 5263 + }, + { + "epoch": 0.2, + "learning_rate": 4.665683508591947e-05, + "loss": 0.6183, + "step": 5264 + }, + { + "epoch": 0.2, + "learning_rate": 4.6656193895870736e-05, + "loss": 0.5951, + "step": 5265 + }, + { + "epoch": 0.2, + "learning_rate": 4.665555270582201e-05, + "loss": 0.5081, + "step": 5266 + }, + { + "epoch": 0.2, + "learning_rate": 4.665491151577327e-05, + "loss": 0.5022, + "step": 5267 + }, + { + "epoch": 0.2, + "learning_rate": 4.6654270325724544e-05, + "loss": 0.5994, + "step": 5268 + }, + { + "epoch": 0.2, + "learning_rate": 4.665362913567582e-05, + "loss": 0.5952, + "step": 5269 + }, + { + "epoch": 0.2, + "learning_rate": 4.665298794562709e-05, + "loss": 0.5023, + "step": 5270 + }, + { + "epoch": 0.2, + "learning_rate": 4.665234675557836e-05, + "loss": 0.6063, + "step": 5271 + }, + { + "epoch": 0.2, + "learning_rate": 4.6651705565529624e-05, + "loss": 0.5252, + "step": 5272 + }, + { + "epoch": 0.2, + "learning_rate": 4.6651064375480896e-05, + "loss": 0.6505, + "step": 5273 + }, + { + "epoch": 0.2, + "learning_rate": 4.665042318543216e-05, + "loss": 0.5819, + "step": 5274 + }, + { + "epoch": 0.2, + "learning_rate": 4.664978199538343e-05, + "loss": 0.6907, + "step": 5275 + }, + { + "epoch": 0.2, + "learning_rate": 4.6649140805334704e-05, + "loss": 0.7126, + "step": 5276 + }, + { + "epoch": 0.2, + "learning_rate": 4.6648499615285976e-05, + "loss": 0.574, + "step": 5277 + }, + { + "epoch": 0.2, + "learning_rate": 4.664785842523725e-05, + "loss": 0.59, + "step": 5278 + }, + { + "epoch": 0.2, + "learning_rate": 4.664721723518851e-05, + "loss": 0.5663, + "step": 5279 + }, + { + "epoch": 0.2, + "learning_rate": 4.6646576045139784e-05, + "loss": 0.6479, + "step": 5280 + }, + { + "epoch": 0.2, + "learning_rate": 4.664593485509105e-05, + "loss": 0.5896, + "step": 5281 + }, + { + "epoch": 0.2, + "learning_rate": 4.664529366504232e-05, + "loss": 0.6409, + "step": 5282 + }, + { + "epoch": 0.2, + "learning_rate": 4.6644652474993586e-05, + "loss": 0.647, + "step": 5283 + }, + { + "epoch": 0.2, + "learning_rate": 4.664401128494486e-05, + "loss": 0.6376, + "step": 5284 + }, + { + "epoch": 0.2, + "learning_rate": 4.664337009489613e-05, + "loss": 0.5604, + "step": 5285 + }, + { + "epoch": 0.2, + "learning_rate": 4.66427289048474e-05, + "loss": 0.6132, + "step": 5286 + }, + { + "epoch": 0.2, + "learning_rate": 4.664208771479867e-05, + "loss": 0.5876, + "step": 5287 + }, + { + "epoch": 0.2, + "learning_rate": 4.664144652474994e-05, + "loss": 0.5175, + "step": 5288 + }, + { + "epoch": 0.2, + "learning_rate": 4.664080533470121e-05, + "loss": 0.6641, + "step": 5289 + }, + { + "epoch": 0.2, + "learning_rate": 4.6640164144652474e-05, + "loss": 0.6356, + "step": 5290 + }, + { + "epoch": 0.2, + "learning_rate": 4.6639522954603746e-05, + "loss": 0.6144, + "step": 5291 + }, + { + "epoch": 0.2, + "learning_rate": 4.663888176455502e-05, + "loss": 0.5879, + "step": 5292 + }, + { + "epoch": 0.2, + "learning_rate": 4.663824057450629e-05, + "loss": 0.6119, + "step": 5293 + }, + { + "epoch": 0.2, + "learning_rate": 4.6637599384457554e-05, + "loss": 0.5954, + "step": 5294 + }, + { + "epoch": 0.2, + "learning_rate": 4.6636958194408826e-05, + "loss": 0.5908, + "step": 5295 + }, + { + "epoch": 0.2, + "learning_rate": 4.66363170043601e-05, + "loss": 0.6174, + "step": 5296 + }, + { + "epoch": 0.2, + "learning_rate": 4.663567581431136e-05, + "loss": 0.6002, + "step": 5297 + }, + { + "epoch": 0.2, + "learning_rate": 4.6635034624262635e-05, + "loss": 0.6199, + "step": 5298 + }, + { + "epoch": 0.2, + "learning_rate": 4.66343934342139e-05, + "loss": 0.6316, + "step": 5299 + }, + { + "epoch": 0.2, + "learning_rate": 4.663375224416517e-05, + "loss": 0.5353, + "step": 5300 + }, + { + "epoch": 0.2, + "learning_rate": 4.663311105411644e-05, + "loss": 0.5862, + "step": 5301 + }, + { + "epoch": 0.2, + "learning_rate": 4.6632469864067715e-05, + "loss": 0.543, + "step": 5302 + }, + { + "epoch": 0.2, + "learning_rate": 4.663182867401898e-05, + "loss": 0.5369, + "step": 5303 + }, + { + "epoch": 0.2, + "learning_rate": 4.663118748397025e-05, + "loss": 0.6287, + "step": 5304 + }, + { + "epoch": 0.2, + "learning_rate": 4.663054629392152e-05, + "loss": 0.6805, + "step": 5305 + }, + { + "epoch": 0.2, + "learning_rate": 4.662990510387279e-05, + "loss": 0.6093, + "step": 5306 + }, + { + "epoch": 0.2, + "learning_rate": 4.662926391382406e-05, + "loss": 0.5704, + "step": 5307 + }, + { + "epoch": 0.2, + "learning_rate": 4.662862272377533e-05, + "loss": 0.5802, + "step": 5308 + }, + { + "epoch": 0.2, + "learning_rate": 4.66279815337266e-05, + "loss": 0.5539, + "step": 5309 + }, + { + "epoch": 0.2, + "learning_rate": 4.662734034367787e-05, + "loss": 0.5823, + "step": 5310 + }, + { + "epoch": 0.2, + "learning_rate": 4.662669915362914e-05, + "loss": 0.5331, + "step": 5311 + }, + { + "epoch": 0.2, + "learning_rate": 4.6626057963580405e-05, + "loss": 0.5336, + "step": 5312 + }, + { + "epoch": 0.2, + "learning_rate": 4.6625416773531676e-05, + "loss": 0.617, + "step": 5313 + }, + { + "epoch": 0.2, + "learning_rate": 4.662477558348295e-05, + "loss": 0.5859, + "step": 5314 + }, + { + "epoch": 0.2, + "learning_rate": 4.662413439343421e-05, + "loss": 0.6401, + "step": 5315 + }, + { + "epoch": 0.2, + "learning_rate": 4.6623493203385485e-05, + "loss": 0.7276, + "step": 5316 + }, + { + "epoch": 0.2, + "learning_rate": 4.6622852013336757e-05, + "loss": 0.568, + "step": 5317 + }, + { + "epoch": 0.2, + "learning_rate": 4.662221082328803e-05, + "loss": 0.5588, + "step": 5318 + }, + { + "epoch": 0.2, + "learning_rate": 4.662156963323929e-05, + "loss": 0.5681, + "step": 5319 + }, + { + "epoch": 0.2, + "learning_rate": 4.6620928443190565e-05, + "loss": 0.53, + "step": 5320 + }, + { + "epoch": 0.2, + "learning_rate": 4.662028725314183e-05, + "loss": 0.5828, + "step": 5321 + }, + { + "epoch": 0.2, + "learning_rate": 4.66196460630931e-05, + "loss": 0.73, + "step": 5322 + }, + { + "epoch": 0.2, + "learning_rate": 4.661900487304437e-05, + "loss": 0.6132, + "step": 5323 + }, + { + "epoch": 0.2, + "learning_rate": 4.6618363682995645e-05, + "loss": 0.6479, + "step": 5324 + }, + { + "epoch": 0.2, + "learning_rate": 4.661772249294692e-05, + "loss": 0.6019, + "step": 5325 + }, + { + "epoch": 0.2, + "learning_rate": 4.661708130289818e-05, + "loss": 0.7477, + "step": 5326 + }, + { + "epoch": 0.2, + "learning_rate": 4.661644011284945e-05, + "loss": 0.6018, + "step": 5327 + }, + { + "epoch": 0.2, + "learning_rate": 4.661579892280072e-05, + "loss": 0.6638, + "step": 5328 + }, + { + "epoch": 0.2, + "learning_rate": 4.661515773275199e-05, + "loss": 0.5634, + "step": 5329 + }, + { + "epoch": 0.2, + "learning_rate": 4.6614516542703255e-05, + "loss": 0.5584, + "step": 5330 + }, + { + "epoch": 0.2, + "learning_rate": 4.661387535265453e-05, + "loss": 0.6544, + "step": 5331 + }, + { + "epoch": 0.2, + "learning_rate": 4.66132341626058e-05, + "loss": 0.5249, + "step": 5332 + }, + { + "epoch": 0.21, + "learning_rate": 4.661259297255707e-05, + "loss": 0.6973, + "step": 5333 + }, + { + "epoch": 0.21, + "learning_rate": 4.661195178250834e-05, + "loss": 0.5914, + "step": 5334 + }, + { + "epoch": 0.21, + "learning_rate": 4.661131059245961e-05, + "loss": 0.5386, + "step": 5335 + }, + { + "epoch": 0.21, + "learning_rate": 4.661066940241088e-05, + "loss": 0.5955, + "step": 5336 + }, + { + "epoch": 0.21, + "learning_rate": 4.6610028212362143e-05, + "loss": 0.5445, + "step": 5337 + }, + { + "epoch": 0.21, + "learning_rate": 4.6609387022313415e-05, + "loss": 0.5693, + "step": 5338 + }, + { + "epoch": 0.21, + "learning_rate": 4.660874583226468e-05, + "loss": 0.5038, + "step": 5339 + }, + { + "epoch": 0.21, + "learning_rate": 4.660810464221595e-05, + "loss": 0.6442, + "step": 5340 + }, + { + "epoch": 0.21, + "learning_rate": 4.660746345216723e-05, + "loss": 0.7669, + "step": 5341 + }, + { + "epoch": 0.21, + "learning_rate": 4.6606822262118495e-05, + "loss": 0.519, + "step": 5342 + }, + { + "epoch": 0.21, + "learning_rate": 4.660618107206977e-05, + "loss": 0.6603, + "step": 5343 + }, + { + "epoch": 0.21, + "learning_rate": 4.660553988202103e-05, + "loss": 0.5469, + "step": 5344 + }, + { + "epoch": 0.21, + "learning_rate": 4.6604898691972304e-05, + "loss": 0.6075, + "step": 5345 + }, + { + "epoch": 0.21, + "learning_rate": 4.660425750192357e-05, + "loss": 0.566, + "step": 5346 + }, + { + "epoch": 0.21, + "learning_rate": 4.660361631187484e-05, + "loss": 0.5811, + "step": 5347 + }, + { + "epoch": 0.21, + "learning_rate": 4.660297512182611e-05, + "loss": 0.5052, + "step": 5348 + }, + { + "epoch": 0.21, + "learning_rate": 4.6602333931777384e-05, + "loss": 0.6377, + "step": 5349 + }, + { + "epoch": 0.21, + "learning_rate": 4.6601692741728655e-05, + "loss": 0.6139, + "step": 5350 + }, + { + "epoch": 0.21, + "learning_rate": 4.660105155167992e-05, + "loss": 0.615, + "step": 5351 + }, + { + "epoch": 0.21, + "learning_rate": 4.660041036163119e-05, + "loss": 0.6027, + "step": 5352 + }, + { + "epoch": 0.21, + "learning_rate": 4.659976917158246e-05, + "loss": 0.574, + "step": 5353 + }, + { + "epoch": 0.21, + "learning_rate": 4.659912798153373e-05, + "loss": 0.5316, + "step": 5354 + }, + { + "epoch": 0.21, + "learning_rate": 4.6598486791484994e-05, + "loss": 0.591, + "step": 5355 + }, + { + "epoch": 0.21, + "learning_rate": 4.6597845601436265e-05, + "loss": 0.543, + "step": 5356 + }, + { + "epoch": 0.21, + "learning_rate": 4.659720441138754e-05, + "loss": 0.6273, + "step": 5357 + }, + { + "epoch": 0.21, + "learning_rate": 4.659656322133881e-05, + "loss": 0.5608, + "step": 5358 + }, + { + "epoch": 0.21, + "learning_rate": 4.659592203129008e-05, + "loss": 0.5836, + "step": 5359 + }, + { + "epoch": 0.21, + "learning_rate": 4.6595280841241345e-05, + "loss": 0.6234, + "step": 5360 + }, + { + "epoch": 0.21, + "learning_rate": 4.659463965119262e-05, + "loss": 0.5188, + "step": 5361 + }, + { + "epoch": 0.21, + "learning_rate": 4.659399846114388e-05, + "loss": 0.4952, + "step": 5362 + }, + { + "epoch": 0.21, + "learning_rate": 4.6593357271095154e-05, + "loss": 0.5285, + "step": 5363 + }, + { + "epoch": 0.21, + "learning_rate": 4.6592716081046426e-05, + "loss": 0.5491, + "step": 5364 + }, + { + "epoch": 0.21, + "learning_rate": 4.65920748909977e-05, + "loss": 0.5964, + "step": 5365 + }, + { + "epoch": 0.21, + "learning_rate": 4.659143370094896e-05, + "loss": 0.6806, + "step": 5366 + }, + { + "epoch": 0.21, + "learning_rate": 4.6590792510900234e-05, + "loss": 0.6277, + "step": 5367 + }, + { + "epoch": 0.21, + "learning_rate": 4.6590151320851506e-05, + "loss": 0.5767, + "step": 5368 + }, + { + "epoch": 0.21, + "learning_rate": 4.658951013080277e-05, + "loss": 0.454, + "step": 5369 + }, + { + "epoch": 0.21, + "learning_rate": 4.658886894075404e-05, + "loss": 0.6661, + "step": 5370 + }, + { + "epoch": 0.21, + "learning_rate": 4.658822775070531e-05, + "loss": 0.6588, + "step": 5371 + }, + { + "epoch": 0.21, + "learning_rate": 4.658758656065658e-05, + "loss": 0.5507, + "step": 5372 + }, + { + "epoch": 0.21, + "learning_rate": 4.658694537060785e-05, + "loss": 0.5832, + "step": 5373 + }, + { + "epoch": 0.21, + "learning_rate": 4.658630418055912e-05, + "loss": 0.5431, + "step": 5374 + }, + { + "epoch": 0.21, + "learning_rate": 4.658566299051039e-05, + "loss": 0.5797, + "step": 5375 + }, + { + "epoch": 0.21, + "learning_rate": 4.658502180046166e-05, + "loss": 0.5506, + "step": 5376 + }, + { + "epoch": 0.21, + "learning_rate": 4.658438061041293e-05, + "loss": 0.6467, + "step": 5377 + }, + { + "epoch": 0.21, + "learning_rate": 4.6583739420364196e-05, + "loss": 0.8012, + "step": 5378 + }, + { + "epoch": 0.21, + "learning_rate": 4.658309823031547e-05, + "loss": 0.5685, + "step": 5379 + }, + { + "epoch": 0.21, + "learning_rate": 4.658245704026674e-05, + "loss": 0.5922, + "step": 5380 + }, + { + "epoch": 0.21, + "learning_rate": 4.658181585021801e-05, + "loss": 0.5853, + "step": 5381 + }, + { + "epoch": 0.21, + "learning_rate": 4.6581174660169276e-05, + "loss": 0.5345, + "step": 5382 + }, + { + "epoch": 0.21, + "learning_rate": 4.658053347012055e-05, + "loss": 0.5917, + "step": 5383 + }, + { + "epoch": 0.21, + "learning_rate": 4.657989228007181e-05, + "loss": 0.6386, + "step": 5384 + }, + { + "epoch": 0.21, + "learning_rate": 4.6579251090023084e-05, + "loss": 0.6291, + "step": 5385 + }, + { + "epoch": 0.21, + "learning_rate": 4.6578609899974356e-05, + "loss": 0.5439, + "step": 5386 + }, + { + "epoch": 0.21, + "learning_rate": 4.657796870992562e-05, + "loss": 0.572, + "step": 5387 + }, + { + "epoch": 0.21, + "learning_rate": 4.657732751987689e-05, + "loss": 0.6382, + "step": 5388 + }, + { + "epoch": 0.21, + "learning_rate": 4.6576686329828164e-05, + "loss": 0.5574, + "step": 5389 + }, + { + "epoch": 0.21, + "learning_rate": 4.6576045139779436e-05, + "loss": 0.5943, + "step": 5390 + }, + { + "epoch": 0.21, + "learning_rate": 4.65754039497307e-05, + "loss": 0.4964, + "step": 5391 + }, + { + "epoch": 0.21, + "learning_rate": 4.657476275968197e-05, + "loss": 0.6869, + "step": 5392 + }, + { + "epoch": 0.21, + "learning_rate": 4.657412156963324e-05, + "loss": 0.5958, + "step": 5393 + }, + { + "epoch": 0.21, + "learning_rate": 4.657348037958451e-05, + "loss": 0.5491, + "step": 5394 + }, + { + "epoch": 0.21, + "learning_rate": 4.657283918953578e-05, + "loss": 0.5732, + "step": 5395 + }, + { + "epoch": 0.21, + "learning_rate": 4.657219799948705e-05, + "loss": 0.5806, + "step": 5396 + }, + { + "epoch": 0.21, + "learning_rate": 4.6571556809438324e-05, + "loss": 0.6242, + "step": 5397 + }, + { + "epoch": 0.21, + "learning_rate": 4.657091561938959e-05, + "loss": 0.5502, + "step": 5398 + }, + { + "epoch": 0.21, + "learning_rate": 4.657027442934086e-05, + "loss": 0.5814, + "step": 5399 + }, + { + "epoch": 0.21, + "learning_rate": 4.6569633239292126e-05, + "loss": 0.6112, + "step": 5400 + }, + { + "epoch": 0.21, + "learning_rate": 4.65689920492434e-05, + "loss": 0.6705, + "step": 5401 + }, + { + "epoch": 0.21, + "learning_rate": 4.656835085919466e-05, + "loss": 0.5535, + "step": 5402 + }, + { + "epoch": 0.21, + "learning_rate": 4.6567709669145934e-05, + "loss": 0.625, + "step": 5403 + }, + { + "epoch": 0.21, + "learning_rate": 4.6567068479097206e-05, + "loss": 0.6209, + "step": 5404 + }, + { + "epoch": 0.21, + "learning_rate": 4.656642728904848e-05, + "loss": 0.5386, + "step": 5405 + }, + { + "epoch": 0.21, + "learning_rate": 4.656578609899975e-05, + "loss": 0.5421, + "step": 5406 + }, + { + "epoch": 0.21, + "learning_rate": 4.6565144908951014e-05, + "loss": 0.5467, + "step": 5407 + }, + { + "epoch": 0.21, + "learning_rate": 4.6564503718902286e-05, + "loss": 0.5501, + "step": 5408 + }, + { + "epoch": 0.21, + "learning_rate": 4.656386252885355e-05, + "loss": 0.5158, + "step": 5409 + }, + { + "epoch": 0.21, + "learning_rate": 4.656322133880482e-05, + "loss": 0.5581, + "step": 5410 + }, + { + "epoch": 0.21, + "learning_rate": 4.656258014875609e-05, + "loss": 0.5994, + "step": 5411 + }, + { + "epoch": 0.21, + "learning_rate": 4.656193895870736e-05, + "loss": 0.577, + "step": 5412 + }, + { + "epoch": 0.21, + "learning_rate": 4.656129776865864e-05, + "loss": 0.5983, + "step": 5413 + }, + { + "epoch": 0.21, + "learning_rate": 4.65606565786099e-05, + "loss": 0.5565, + "step": 5414 + }, + { + "epoch": 0.21, + "learning_rate": 4.6560015388561175e-05, + "loss": 0.5526, + "step": 5415 + }, + { + "epoch": 0.21, + "learning_rate": 4.655937419851244e-05, + "loss": 0.5496, + "step": 5416 + }, + { + "epoch": 0.21, + "learning_rate": 4.655873300846371e-05, + "loss": 0.6581, + "step": 5417 + }, + { + "epoch": 0.21, + "learning_rate": 4.6558091818414976e-05, + "loss": 0.5592, + "step": 5418 + }, + { + "epoch": 0.21, + "learning_rate": 4.655745062836625e-05, + "loss": 0.5274, + "step": 5419 + }, + { + "epoch": 0.21, + "learning_rate": 4.655680943831752e-05, + "loss": 0.7018, + "step": 5420 + }, + { + "epoch": 0.21, + "learning_rate": 4.655616824826879e-05, + "loss": 0.5182, + "step": 5421 + }, + { + "epoch": 0.21, + "learning_rate": 4.655552705822006e-05, + "loss": 0.5293, + "step": 5422 + }, + { + "epoch": 0.21, + "learning_rate": 4.655488586817133e-05, + "loss": 0.5827, + "step": 5423 + }, + { + "epoch": 0.21, + "learning_rate": 4.65542446781226e-05, + "loss": 0.7101, + "step": 5424 + }, + { + "epoch": 0.21, + "learning_rate": 4.6553603488073865e-05, + "loss": 0.6239, + "step": 5425 + }, + { + "epoch": 0.21, + "learning_rate": 4.6552962298025136e-05, + "loss": 0.6252, + "step": 5426 + }, + { + "epoch": 0.21, + "learning_rate": 4.65523211079764e-05, + "loss": 0.6311, + "step": 5427 + }, + { + "epoch": 0.21, + "learning_rate": 4.655167991792767e-05, + "loss": 0.6025, + "step": 5428 + }, + { + "epoch": 0.21, + "learning_rate": 4.6551038727878945e-05, + "loss": 0.5547, + "step": 5429 + }, + { + "epoch": 0.21, + "learning_rate": 4.6550397537830217e-05, + "loss": 0.6314, + "step": 5430 + }, + { + "epoch": 0.21, + "learning_rate": 4.654975634778149e-05, + "loss": 0.632, + "step": 5431 + }, + { + "epoch": 0.21, + "learning_rate": 4.654911515773275e-05, + "loss": 0.5568, + "step": 5432 + }, + { + "epoch": 0.21, + "learning_rate": 4.6548473967684025e-05, + "loss": 0.5001, + "step": 5433 + }, + { + "epoch": 0.21, + "learning_rate": 4.654783277763529e-05, + "loss": 0.6588, + "step": 5434 + }, + { + "epoch": 0.21, + "learning_rate": 4.654719158758656e-05, + "loss": 0.5753, + "step": 5435 + }, + { + "epoch": 0.21, + "learning_rate": 4.654655039753783e-05, + "loss": 0.6104, + "step": 5436 + }, + { + "epoch": 0.21, + "learning_rate": 4.6545909207489105e-05, + "loss": 0.6258, + "step": 5437 + }, + { + "epoch": 0.21, + "learning_rate": 4.654526801744037e-05, + "loss": 0.5521, + "step": 5438 + }, + { + "epoch": 0.21, + "learning_rate": 4.654462682739164e-05, + "loss": 0.59, + "step": 5439 + }, + { + "epoch": 0.21, + "learning_rate": 4.654398563734291e-05, + "loss": 0.5617, + "step": 5440 + }, + { + "epoch": 0.21, + "learning_rate": 4.654334444729418e-05, + "loss": 0.5689, + "step": 5441 + }, + { + "epoch": 0.21, + "learning_rate": 4.654270325724545e-05, + "loss": 0.6689, + "step": 5442 + }, + { + "epoch": 0.21, + "learning_rate": 4.6542062067196715e-05, + "loss": 0.5564, + "step": 5443 + }, + { + "epoch": 0.21, + "learning_rate": 4.654142087714799e-05, + "loss": 0.5268, + "step": 5444 + }, + { + "epoch": 0.21, + "learning_rate": 4.654077968709926e-05, + "loss": 0.5685, + "step": 5445 + }, + { + "epoch": 0.21, + "learning_rate": 4.654013849705053e-05, + "loss": 0.6208, + "step": 5446 + }, + { + "epoch": 0.21, + "learning_rate": 4.65394973070018e-05, + "loss": 0.7016, + "step": 5447 + }, + { + "epoch": 0.21, + "learning_rate": 4.653885611695307e-05, + "loss": 0.6366, + "step": 5448 + }, + { + "epoch": 0.21, + "learning_rate": 4.653821492690434e-05, + "loss": 0.5983, + "step": 5449 + }, + { + "epoch": 0.21, + "learning_rate": 4.6537573736855603e-05, + "loss": 0.5775, + "step": 5450 + }, + { + "epoch": 0.21, + "learning_rate": 4.6536932546806875e-05, + "loss": 0.515, + "step": 5451 + }, + { + "epoch": 0.21, + "learning_rate": 4.653629135675815e-05, + "loss": 0.5865, + "step": 5452 + }, + { + "epoch": 0.21, + "learning_rate": 4.653565016670942e-05, + "loss": 0.6692, + "step": 5453 + }, + { + "epoch": 0.21, + "learning_rate": 4.6535008976660684e-05, + "loss": 0.6176, + "step": 5454 + }, + { + "epoch": 0.21, + "learning_rate": 4.6534367786611955e-05, + "loss": 0.6461, + "step": 5455 + }, + { + "epoch": 0.21, + "learning_rate": 4.653372659656323e-05, + "loss": 0.5905, + "step": 5456 + }, + { + "epoch": 0.21, + "learning_rate": 4.653308540651449e-05, + "loss": 0.5898, + "step": 5457 + }, + { + "epoch": 0.21, + "learning_rate": 4.6532444216465764e-05, + "loss": 0.5898, + "step": 5458 + }, + { + "epoch": 0.21, + "learning_rate": 4.653180302641703e-05, + "loss": 0.6145, + "step": 5459 + }, + { + "epoch": 0.21, + "learning_rate": 4.65311618363683e-05, + "loss": 0.554, + "step": 5460 + }, + { + "epoch": 0.21, + "learning_rate": 4.653052064631957e-05, + "loss": 0.6214, + "step": 5461 + }, + { + "epoch": 0.21, + "learning_rate": 4.6529879456270844e-05, + "loss": 0.6342, + "step": 5462 + }, + { + "epoch": 0.21, + "learning_rate": 4.652923826622211e-05, + "loss": 0.647, + "step": 5463 + }, + { + "epoch": 0.21, + "learning_rate": 4.652859707617338e-05, + "loss": 0.5719, + "step": 5464 + }, + { + "epoch": 0.21, + "learning_rate": 4.652795588612465e-05, + "loss": 0.59, + "step": 5465 + }, + { + "epoch": 0.21, + "learning_rate": 4.652731469607592e-05, + "loss": 0.5942, + "step": 5466 + }, + { + "epoch": 0.21, + "learning_rate": 4.652667350602719e-05, + "loss": 0.7153, + "step": 5467 + }, + { + "epoch": 0.21, + "learning_rate": 4.652603231597846e-05, + "loss": 0.5711, + "step": 5468 + }, + { + "epoch": 0.21, + "learning_rate": 4.652539112592973e-05, + "loss": 0.6442, + "step": 5469 + }, + { + "epoch": 0.21, + "learning_rate": 4.6524749935881e-05, + "loss": 0.5762, + "step": 5470 + }, + { + "epoch": 0.21, + "learning_rate": 4.652410874583227e-05, + "loss": 0.5912, + "step": 5471 + }, + { + "epoch": 0.21, + "learning_rate": 4.6523467555783534e-05, + "loss": 0.5215, + "step": 5472 + }, + { + "epoch": 0.21, + "learning_rate": 4.6522826365734805e-05, + "loss": 0.666, + "step": 5473 + }, + { + "epoch": 0.21, + "learning_rate": 4.652218517568608e-05, + "loss": 0.5662, + "step": 5474 + }, + { + "epoch": 0.21, + "learning_rate": 4.652154398563734e-05, + "loss": 0.551, + "step": 5475 + }, + { + "epoch": 0.21, + "learning_rate": 4.6520902795588614e-05, + "loss": 0.5074, + "step": 5476 + }, + { + "epoch": 0.21, + "learning_rate": 4.6520261605539886e-05, + "loss": 0.6114, + "step": 5477 + }, + { + "epoch": 0.21, + "learning_rate": 4.651962041549116e-05, + "loss": 0.625, + "step": 5478 + }, + { + "epoch": 0.21, + "learning_rate": 4.651897922544242e-05, + "loss": 0.6639, + "step": 5479 + }, + { + "epoch": 0.21, + "learning_rate": 4.6518338035393694e-05, + "loss": 0.6172, + "step": 5480 + }, + { + "epoch": 0.21, + "learning_rate": 4.651769684534496e-05, + "loss": 0.4844, + "step": 5481 + }, + { + "epoch": 0.21, + "learning_rate": 4.651705565529623e-05, + "loss": 0.5828, + "step": 5482 + }, + { + "epoch": 0.21, + "learning_rate": 4.65164144652475e-05, + "loss": 0.5234, + "step": 5483 + }, + { + "epoch": 0.21, + "learning_rate": 4.6515773275198774e-05, + "loss": 0.5537, + "step": 5484 + }, + { + "epoch": 0.21, + "learning_rate": 4.6515132085150046e-05, + "loss": 0.6986, + "step": 5485 + }, + { + "epoch": 0.21, + "learning_rate": 4.651449089510131e-05, + "loss": 0.5683, + "step": 5486 + }, + { + "epoch": 0.21, + "learning_rate": 4.651384970505258e-05, + "loss": 0.5353, + "step": 5487 + }, + { + "epoch": 0.21, + "learning_rate": 4.651320851500385e-05, + "loss": 0.516, + "step": 5488 + }, + { + "epoch": 0.21, + "learning_rate": 4.651256732495512e-05, + "loss": 0.5547, + "step": 5489 + }, + { + "epoch": 0.21, + "learning_rate": 4.6511926134906384e-05, + "loss": 0.5918, + "step": 5490 + }, + { + "epoch": 0.21, + "learning_rate": 4.6511284944857656e-05, + "loss": 0.6052, + "step": 5491 + }, + { + "epoch": 0.21, + "learning_rate": 4.651064375480893e-05, + "loss": 0.5407, + "step": 5492 + }, + { + "epoch": 0.21, + "learning_rate": 4.65100025647602e-05, + "loss": 0.5987, + "step": 5493 + }, + { + "epoch": 0.21, + "learning_rate": 4.650936137471147e-05, + "loss": 0.5224, + "step": 5494 + }, + { + "epoch": 0.21, + "learning_rate": 4.6508720184662736e-05, + "loss": 0.5782, + "step": 5495 + }, + { + "epoch": 0.21, + "learning_rate": 4.650807899461401e-05, + "loss": 0.5489, + "step": 5496 + }, + { + "epoch": 0.21, + "learning_rate": 4.650743780456527e-05, + "loss": 0.6225, + "step": 5497 + }, + { + "epoch": 0.21, + "learning_rate": 4.6506796614516544e-05, + "loss": 0.614, + "step": 5498 + }, + { + "epoch": 0.21, + "learning_rate": 4.650615542446781e-05, + "loss": 0.5313, + "step": 5499 + }, + { + "epoch": 0.21, + "learning_rate": 4.650551423441908e-05, + "loss": 0.681, + "step": 5500 + }, + { + "epoch": 0.21, + "learning_rate": 4.650487304437036e-05, + "loss": 0.6095, + "step": 5501 + }, + { + "epoch": 0.21, + "learning_rate": 4.6504231854321624e-05, + "loss": 0.5847, + "step": 5502 + }, + { + "epoch": 0.21, + "learning_rate": 4.6503590664272896e-05, + "loss": 0.6103, + "step": 5503 + }, + { + "epoch": 0.21, + "learning_rate": 4.650294947422416e-05, + "loss": 0.5991, + "step": 5504 + }, + { + "epoch": 0.21, + "learning_rate": 4.650230828417543e-05, + "loss": 0.5298, + "step": 5505 + }, + { + "epoch": 0.21, + "learning_rate": 4.65016670941267e-05, + "loss": 0.5801, + "step": 5506 + }, + { + "epoch": 0.21, + "learning_rate": 4.650102590407797e-05, + "loss": 0.62, + "step": 5507 + }, + { + "epoch": 0.21, + "learning_rate": 4.650038471402924e-05, + "loss": 0.6038, + "step": 5508 + }, + { + "epoch": 0.21, + "learning_rate": 4.649974352398051e-05, + "loss": 0.5885, + "step": 5509 + }, + { + "epoch": 0.21, + "learning_rate": 4.6499102333931784e-05, + "loss": 0.5524, + "step": 5510 + }, + { + "epoch": 0.21, + "learning_rate": 4.649846114388305e-05, + "loss": 0.5932, + "step": 5511 + }, + { + "epoch": 0.21, + "learning_rate": 4.649781995383432e-05, + "loss": 0.6138, + "step": 5512 + }, + { + "epoch": 0.21, + "learning_rate": 4.6497178763785586e-05, + "loss": 0.7358, + "step": 5513 + }, + { + "epoch": 0.21, + "learning_rate": 4.649653757373686e-05, + "loss": 0.5857, + "step": 5514 + }, + { + "epoch": 0.21, + "learning_rate": 4.649589638368812e-05, + "loss": 0.567, + "step": 5515 + }, + { + "epoch": 0.21, + "learning_rate": 4.6495255193639394e-05, + "loss": 0.551, + "step": 5516 + }, + { + "epoch": 0.21, + "learning_rate": 4.6494614003590666e-05, + "loss": 0.6159, + "step": 5517 + }, + { + "epoch": 0.21, + "learning_rate": 4.649397281354194e-05, + "loss": 0.5931, + "step": 5518 + }, + { + "epoch": 0.21, + "learning_rate": 4.649333162349321e-05, + "loss": 0.5431, + "step": 5519 + }, + { + "epoch": 0.21, + "learning_rate": 4.6492690433444475e-05, + "loss": 0.5148, + "step": 5520 + }, + { + "epoch": 0.21, + "learning_rate": 4.6492049243395746e-05, + "loss": 0.5871, + "step": 5521 + }, + { + "epoch": 0.21, + "learning_rate": 4.649140805334701e-05, + "loss": 0.5439, + "step": 5522 + }, + { + "epoch": 0.21, + "learning_rate": 4.649076686329828e-05, + "loss": 0.6247, + "step": 5523 + }, + { + "epoch": 0.21, + "learning_rate": 4.6490125673249555e-05, + "loss": 0.6191, + "step": 5524 + }, + { + "epoch": 0.21, + "learning_rate": 4.6489484483200826e-05, + "loss": 0.649, + "step": 5525 + }, + { + "epoch": 0.21, + "learning_rate": 4.648884329315209e-05, + "loss": 0.5219, + "step": 5526 + }, + { + "epoch": 0.21, + "learning_rate": 4.648820210310336e-05, + "loss": 0.648, + "step": 5527 + }, + { + "epoch": 0.21, + "learning_rate": 4.6487560913054635e-05, + "loss": 0.6092, + "step": 5528 + }, + { + "epoch": 0.21, + "learning_rate": 4.64869197230059e-05, + "loss": 0.6472, + "step": 5529 + }, + { + "epoch": 0.21, + "learning_rate": 4.648627853295717e-05, + "loss": 0.6038, + "step": 5530 + }, + { + "epoch": 0.21, + "learning_rate": 4.6485637342908436e-05, + "loss": 0.5968, + "step": 5531 + }, + { + "epoch": 0.21, + "learning_rate": 4.648499615285971e-05, + "loss": 0.6111, + "step": 5532 + }, + { + "epoch": 0.21, + "learning_rate": 4.648435496281098e-05, + "loss": 0.6421, + "step": 5533 + }, + { + "epoch": 0.21, + "learning_rate": 4.648371377276225e-05, + "loss": 0.5402, + "step": 5534 + }, + { + "epoch": 0.21, + "learning_rate": 4.6483072582713516e-05, + "loss": 0.6076, + "step": 5535 + }, + { + "epoch": 0.21, + "learning_rate": 4.648243139266479e-05, + "loss": 0.5935, + "step": 5536 + }, + { + "epoch": 0.21, + "learning_rate": 4.648179020261606e-05, + "loss": 0.6121, + "step": 5537 + }, + { + "epoch": 0.21, + "learning_rate": 4.6481149012567325e-05, + "loss": 0.565, + "step": 5538 + }, + { + "epoch": 0.21, + "learning_rate": 4.6480507822518596e-05, + "loss": 0.6663, + "step": 5539 + }, + { + "epoch": 0.21, + "learning_rate": 4.647986663246987e-05, + "loss": 0.5856, + "step": 5540 + }, + { + "epoch": 0.21, + "learning_rate": 4.647922544242114e-05, + "loss": 0.5321, + "step": 5541 + }, + { + "epoch": 0.21, + "learning_rate": 4.6478584252372405e-05, + "loss": 0.6067, + "step": 5542 + }, + { + "epoch": 0.21, + "learning_rate": 4.6477943062323677e-05, + "loss": 0.6012, + "step": 5543 + }, + { + "epoch": 0.21, + "learning_rate": 4.647730187227494e-05, + "loss": 0.5555, + "step": 5544 + }, + { + "epoch": 0.21, + "learning_rate": 4.647666068222621e-05, + "loss": 0.5968, + "step": 5545 + }, + { + "epoch": 0.21, + "learning_rate": 4.6476019492177485e-05, + "loss": 0.7153, + "step": 5546 + }, + { + "epoch": 0.21, + "learning_rate": 4.647537830212875e-05, + "loss": 0.6138, + "step": 5547 + }, + { + "epoch": 0.21, + "learning_rate": 4.647473711208002e-05, + "loss": 0.5491, + "step": 5548 + }, + { + "epoch": 0.21, + "learning_rate": 4.647409592203129e-05, + "loss": 0.5321, + "step": 5549 + }, + { + "epoch": 0.21, + "learning_rate": 4.6473454731982565e-05, + "loss": 0.5436, + "step": 5550 + }, + { + "epoch": 0.21, + "learning_rate": 4.647281354193383e-05, + "loss": 0.5639, + "step": 5551 + }, + { + "epoch": 0.21, + "learning_rate": 4.64721723518851e-05, + "loss": 0.7412, + "step": 5552 + }, + { + "epoch": 0.21, + "learning_rate": 4.647153116183637e-05, + "loss": 0.5087, + "step": 5553 + }, + { + "epoch": 0.21, + "learning_rate": 4.647088997178764e-05, + "loss": 0.5832, + "step": 5554 + }, + { + "epoch": 0.21, + "learning_rate": 4.647024878173891e-05, + "loss": 0.6689, + "step": 5555 + }, + { + "epoch": 0.21, + "learning_rate": 4.646960759169018e-05, + "loss": 0.596, + "step": 5556 + }, + { + "epoch": 0.21, + "learning_rate": 4.6468966401641453e-05, + "loss": 0.5114, + "step": 5557 + }, + { + "epoch": 0.21, + "learning_rate": 4.646832521159272e-05, + "loss": 0.5777, + "step": 5558 + }, + { + "epoch": 0.21, + "learning_rate": 4.646768402154399e-05, + "loss": 0.5803, + "step": 5559 + }, + { + "epoch": 0.21, + "learning_rate": 4.6467042831495255e-05, + "loss": 0.644, + "step": 5560 + }, + { + "epoch": 0.21, + "learning_rate": 4.646640164144653e-05, + "loss": 0.6441, + "step": 5561 + }, + { + "epoch": 0.21, + "learning_rate": 4.646576045139779e-05, + "loss": 0.5715, + "step": 5562 + }, + { + "epoch": 0.21, + "learning_rate": 4.6465119261349063e-05, + "loss": 0.5279, + "step": 5563 + }, + { + "epoch": 0.21, + "learning_rate": 4.6464478071300335e-05, + "loss": 0.5682, + "step": 5564 + }, + { + "epoch": 0.21, + "learning_rate": 4.646383688125161e-05, + "loss": 0.6644, + "step": 5565 + }, + { + "epoch": 0.21, + "learning_rate": 4.646319569120288e-05, + "loss": 0.6803, + "step": 5566 + }, + { + "epoch": 0.21, + "learning_rate": 4.6462554501154144e-05, + "loss": 0.5186, + "step": 5567 + }, + { + "epoch": 0.21, + "learning_rate": 4.6461913311105415e-05, + "loss": 0.6546, + "step": 5568 + }, + { + "epoch": 0.21, + "learning_rate": 4.646127212105668e-05, + "loss": 0.5576, + "step": 5569 + }, + { + "epoch": 0.21, + "learning_rate": 4.646063093100795e-05, + "loss": 0.582, + "step": 5570 + }, + { + "epoch": 0.21, + "learning_rate": 4.645998974095922e-05, + "loss": 0.6048, + "step": 5571 + }, + { + "epoch": 0.21, + "learning_rate": 4.6459348550910495e-05, + "loss": 0.5945, + "step": 5572 + }, + { + "epoch": 0.21, + "learning_rate": 4.645870736086177e-05, + "loss": 0.6932, + "step": 5573 + }, + { + "epoch": 0.21, + "learning_rate": 4.645806617081303e-05, + "loss": 0.5361, + "step": 5574 + }, + { + "epoch": 0.21, + "learning_rate": 4.6457424980764304e-05, + "loss": 0.5175, + "step": 5575 + }, + { + "epoch": 0.21, + "learning_rate": 4.645678379071557e-05, + "loss": 0.5148, + "step": 5576 + }, + { + "epoch": 0.21, + "learning_rate": 4.645614260066684e-05, + "loss": 0.6955, + "step": 5577 + }, + { + "epoch": 0.21, + "learning_rate": 4.6455501410618105e-05, + "loss": 0.5867, + "step": 5578 + }, + { + "epoch": 0.21, + "learning_rate": 4.645486022056938e-05, + "loss": 0.6531, + "step": 5579 + }, + { + "epoch": 0.21, + "learning_rate": 4.645421903052065e-05, + "loss": 0.5469, + "step": 5580 + }, + { + "epoch": 0.21, + "learning_rate": 4.645357784047192e-05, + "loss": 0.5953, + "step": 5581 + }, + { + "epoch": 0.21, + "learning_rate": 4.645293665042319e-05, + "loss": 0.6051, + "step": 5582 + }, + { + "epoch": 0.21, + "learning_rate": 4.645229546037446e-05, + "loss": 0.551, + "step": 5583 + }, + { + "epoch": 0.21, + "learning_rate": 4.645165427032573e-05, + "loss": 0.735, + "step": 5584 + }, + { + "epoch": 0.21, + "learning_rate": 4.6451013080276994e-05, + "loss": 0.6505, + "step": 5585 + }, + { + "epoch": 0.21, + "learning_rate": 4.6450371890228266e-05, + "loss": 0.5006, + "step": 5586 + }, + { + "epoch": 0.21, + "learning_rate": 4.644973070017953e-05, + "loss": 0.5265, + "step": 5587 + }, + { + "epoch": 0.21, + "learning_rate": 4.64490895101308e-05, + "loss": 0.5776, + "step": 5588 + }, + { + "epoch": 0.21, + "learning_rate": 4.6448448320082074e-05, + "loss": 0.5305, + "step": 5589 + }, + { + "epoch": 0.21, + "learning_rate": 4.6447807130033346e-05, + "loss": 0.5099, + "step": 5590 + }, + { + "epoch": 0.21, + "learning_rate": 4.644716593998462e-05, + "loss": 0.6556, + "step": 5591 + }, + { + "epoch": 0.21, + "learning_rate": 4.644652474993588e-05, + "loss": 0.5326, + "step": 5592 + }, + { + "epoch": 0.22, + "learning_rate": 4.6445883559887154e-05, + "loss": 0.5802, + "step": 5593 + }, + { + "epoch": 0.22, + "learning_rate": 4.644524236983842e-05, + "loss": 0.6122, + "step": 5594 + }, + { + "epoch": 0.22, + "learning_rate": 4.644460117978969e-05, + "loss": 0.5594, + "step": 5595 + }, + { + "epoch": 0.22, + "learning_rate": 4.644395998974096e-05, + "loss": 0.5501, + "step": 5596 + }, + { + "epoch": 0.22, + "learning_rate": 4.6443318799692234e-05, + "loss": 0.6115, + "step": 5597 + }, + { + "epoch": 0.22, + "learning_rate": 4.64426776096435e-05, + "loss": 0.6331, + "step": 5598 + }, + { + "epoch": 0.22, + "learning_rate": 4.644203641959477e-05, + "loss": 0.5185, + "step": 5599 + }, + { + "epoch": 0.22, + "learning_rate": 4.644139522954604e-05, + "loss": 0.5254, + "step": 5600 + }, + { + "epoch": 0.22, + "learning_rate": 4.644075403949731e-05, + "loss": 0.6304, + "step": 5601 + }, + { + "epoch": 0.22, + "learning_rate": 4.644011284944858e-05, + "loss": 0.6405, + "step": 5602 + }, + { + "epoch": 0.22, + "learning_rate": 4.6439471659399844e-05, + "loss": 0.5492, + "step": 5603 + }, + { + "epoch": 0.22, + "learning_rate": 4.6438830469351116e-05, + "loss": 0.5974, + "step": 5604 + }, + { + "epoch": 0.22, + "learning_rate": 4.643818927930239e-05, + "loss": 0.5956, + "step": 5605 + }, + { + "epoch": 0.22, + "learning_rate": 4.643754808925366e-05, + "loss": 0.5482, + "step": 5606 + }, + { + "epoch": 0.22, + "learning_rate": 4.6436906899204924e-05, + "loss": 0.5237, + "step": 5607 + }, + { + "epoch": 0.22, + "learning_rate": 4.6436265709156196e-05, + "loss": 0.7228, + "step": 5608 + }, + { + "epoch": 0.22, + "learning_rate": 4.643562451910747e-05, + "loss": 0.7752, + "step": 5609 + }, + { + "epoch": 0.22, + "learning_rate": 4.643498332905873e-05, + "loss": 0.5768, + "step": 5610 + }, + { + "epoch": 0.22, + "learning_rate": 4.6434342139010004e-05, + "loss": 0.6226, + "step": 5611 + }, + { + "epoch": 0.22, + "learning_rate": 4.6433700948961276e-05, + "loss": 0.5911, + "step": 5612 + }, + { + "epoch": 0.22, + "learning_rate": 4.643305975891255e-05, + "loss": 0.5261, + "step": 5613 + }, + { + "epoch": 0.22, + "learning_rate": 4.643241856886381e-05, + "loss": 0.5392, + "step": 5614 + }, + { + "epoch": 0.22, + "learning_rate": 4.6431777378815084e-05, + "loss": 0.5181, + "step": 5615 + }, + { + "epoch": 0.22, + "learning_rate": 4.643113618876635e-05, + "loss": 0.5861, + "step": 5616 + }, + { + "epoch": 0.22, + "learning_rate": 4.643049499871762e-05, + "loss": 0.5705, + "step": 5617 + }, + { + "epoch": 0.22, + "learning_rate": 4.642985380866889e-05, + "loss": 0.5978, + "step": 5618 + }, + { + "epoch": 0.22, + "learning_rate": 4.642921261862016e-05, + "loss": 0.5561, + "step": 5619 + }, + { + "epoch": 0.22, + "learning_rate": 4.642857142857143e-05, + "loss": 0.5816, + "step": 5620 + }, + { + "epoch": 0.22, + "learning_rate": 4.64279302385227e-05, + "loss": 0.5757, + "step": 5621 + }, + { + "epoch": 0.22, + "learning_rate": 4.642728904847397e-05, + "loss": 0.6273, + "step": 5622 + }, + { + "epoch": 0.22, + "learning_rate": 4.642664785842524e-05, + "loss": 0.5953, + "step": 5623 + }, + { + "epoch": 0.22, + "learning_rate": 4.642600666837651e-05, + "loss": 0.5772, + "step": 5624 + }, + { + "epoch": 0.22, + "learning_rate": 4.6425365478327774e-05, + "loss": 0.5418, + "step": 5625 + }, + { + "epoch": 0.22, + "learning_rate": 4.6424724288279046e-05, + "loss": 0.5462, + "step": 5626 + }, + { + "epoch": 0.22, + "learning_rate": 4.642408309823032e-05, + "loss": 0.5593, + "step": 5627 + }, + { + "epoch": 0.22, + "learning_rate": 4.642344190818159e-05, + "loss": 0.5794, + "step": 5628 + }, + { + "epoch": 0.22, + "learning_rate": 4.642280071813286e-05, + "loss": 0.5286, + "step": 5629 + }, + { + "epoch": 0.22, + "learning_rate": 4.6422159528084126e-05, + "loss": 0.6122, + "step": 5630 + }, + { + "epoch": 0.22, + "learning_rate": 4.64215183380354e-05, + "loss": 0.5273, + "step": 5631 + }, + { + "epoch": 0.22, + "learning_rate": 4.642087714798666e-05, + "loss": 0.6219, + "step": 5632 + }, + { + "epoch": 0.22, + "learning_rate": 4.6420235957937935e-05, + "loss": 0.5335, + "step": 5633 + }, + { + "epoch": 0.22, + "learning_rate": 4.6419594767889206e-05, + "loss": 0.5406, + "step": 5634 + }, + { + "epoch": 0.22, + "learning_rate": 4.641895357784047e-05, + "loss": 0.6527, + "step": 5635 + }, + { + "epoch": 0.22, + "learning_rate": 4.641831238779174e-05, + "loss": 0.5785, + "step": 5636 + }, + { + "epoch": 0.22, + "learning_rate": 4.6417671197743015e-05, + "loss": 0.6429, + "step": 5637 + }, + { + "epoch": 0.22, + "learning_rate": 4.6417030007694286e-05, + "loss": 0.5776, + "step": 5638 + }, + { + "epoch": 0.22, + "learning_rate": 4.641638881764555e-05, + "loss": 0.5597, + "step": 5639 + }, + { + "epoch": 0.22, + "learning_rate": 4.641574762759682e-05, + "loss": 0.6294, + "step": 5640 + }, + { + "epoch": 0.22, + "learning_rate": 4.641510643754809e-05, + "loss": 0.6494, + "step": 5641 + }, + { + "epoch": 0.22, + "learning_rate": 4.641446524749936e-05, + "loss": 0.673, + "step": 5642 + }, + { + "epoch": 0.22, + "learning_rate": 4.641382405745063e-05, + "loss": 0.5786, + "step": 5643 + }, + { + "epoch": 0.22, + "learning_rate": 4.64131828674019e-05, + "loss": 0.6436, + "step": 5644 + }, + { + "epoch": 0.22, + "learning_rate": 4.6412541677353175e-05, + "loss": 0.6521, + "step": 5645 + }, + { + "epoch": 0.22, + "learning_rate": 4.641190048730444e-05, + "loss": 0.5838, + "step": 5646 + }, + { + "epoch": 0.22, + "learning_rate": 4.641125929725571e-05, + "loss": 0.6263, + "step": 5647 + }, + { + "epoch": 0.22, + "learning_rate": 4.6410618107206976e-05, + "loss": 0.5954, + "step": 5648 + }, + { + "epoch": 0.22, + "learning_rate": 4.640997691715825e-05, + "loss": 0.6075, + "step": 5649 + }, + { + "epoch": 0.22, + "learning_rate": 4.640933572710951e-05, + "loss": 0.5736, + "step": 5650 + }, + { + "epoch": 0.22, + "learning_rate": 4.6408694537060785e-05, + "loss": 0.5989, + "step": 5651 + }, + { + "epoch": 0.22, + "learning_rate": 4.6408053347012057e-05, + "loss": 0.5866, + "step": 5652 + }, + { + "epoch": 0.22, + "learning_rate": 4.640741215696333e-05, + "loss": 0.6521, + "step": 5653 + }, + { + "epoch": 0.22, + "learning_rate": 4.64067709669146e-05, + "loss": 0.6085, + "step": 5654 + }, + { + "epoch": 0.22, + "learning_rate": 4.6406129776865865e-05, + "loss": 0.5742, + "step": 5655 + }, + { + "epoch": 0.22, + "learning_rate": 4.6405488586817137e-05, + "loss": 0.5399, + "step": 5656 + }, + { + "epoch": 0.22, + "learning_rate": 4.64048473967684e-05, + "loss": 0.5232, + "step": 5657 + }, + { + "epoch": 0.22, + "learning_rate": 4.640420620671967e-05, + "loss": 0.5699, + "step": 5658 + }, + { + "epoch": 0.22, + "learning_rate": 4.640356501667094e-05, + "loss": 0.5641, + "step": 5659 + }, + { + "epoch": 0.22, + "learning_rate": 4.640292382662221e-05, + "loss": 0.6098, + "step": 5660 + }, + { + "epoch": 0.22, + "learning_rate": 4.640228263657349e-05, + "loss": 0.6014, + "step": 5661 + }, + { + "epoch": 0.22, + "learning_rate": 4.640164144652475e-05, + "loss": 0.5692, + "step": 5662 + }, + { + "epoch": 0.22, + "learning_rate": 4.6401000256476025e-05, + "loss": 0.53, + "step": 5663 + }, + { + "epoch": 0.22, + "learning_rate": 4.640035906642729e-05, + "loss": 0.5979, + "step": 5664 + }, + { + "epoch": 0.22, + "learning_rate": 4.639971787637856e-05, + "loss": 0.5783, + "step": 5665 + }, + { + "epoch": 0.22, + "learning_rate": 4.639907668632983e-05, + "loss": 0.6796, + "step": 5666 + }, + { + "epoch": 0.22, + "learning_rate": 4.63984354962811e-05, + "loss": 0.5388, + "step": 5667 + }, + { + "epoch": 0.22, + "learning_rate": 4.639779430623237e-05, + "loss": 0.5698, + "step": 5668 + }, + { + "epoch": 0.22, + "learning_rate": 4.639715311618364e-05, + "loss": 0.6168, + "step": 5669 + }, + { + "epoch": 0.22, + "learning_rate": 4.6396511926134914e-05, + "loss": 0.6915, + "step": 5670 + }, + { + "epoch": 0.22, + "learning_rate": 4.639587073608618e-05, + "loss": 0.6028, + "step": 5671 + }, + { + "epoch": 0.22, + "learning_rate": 4.639522954603745e-05, + "loss": 0.5695, + "step": 5672 + }, + { + "epoch": 0.22, + "learning_rate": 4.6394588355988715e-05, + "loss": 0.5594, + "step": 5673 + }, + { + "epoch": 0.22, + "learning_rate": 4.639394716593999e-05, + "loss": 0.5375, + "step": 5674 + }, + { + "epoch": 0.22, + "learning_rate": 4.639330597589125e-05, + "loss": 0.5695, + "step": 5675 + }, + { + "epoch": 0.22, + "learning_rate": 4.6392664785842524e-05, + "loss": 0.5203, + "step": 5676 + }, + { + "epoch": 0.22, + "learning_rate": 4.6392023595793795e-05, + "loss": 0.7212, + "step": 5677 + }, + { + "epoch": 0.22, + "learning_rate": 4.639138240574507e-05, + "loss": 0.5858, + "step": 5678 + }, + { + "epoch": 0.22, + "learning_rate": 4.639074121569634e-05, + "loss": 0.5615, + "step": 5679 + }, + { + "epoch": 0.22, + "learning_rate": 4.6390100025647604e-05, + "loss": 0.5463, + "step": 5680 + }, + { + "epoch": 0.22, + "learning_rate": 4.6389458835598875e-05, + "loss": 0.5192, + "step": 5681 + }, + { + "epoch": 0.22, + "learning_rate": 4.638881764555014e-05, + "loss": 0.584, + "step": 5682 + }, + { + "epoch": 0.22, + "learning_rate": 4.638817645550141e-05, + "loss": 0.5931, + "step": 5683 + }, + { + "epoch": 0.22, + "learning_rate": 4.6387535265452684e-05, + "loss": 0.7229, + "step": 5684 + }, + { + "epoch": 0.22, + "learning_rate": 4.6386894075403955e-05, + "loss": 0.508, + "step": 5685 + }, + { + "epoch": 0.22, + "learning_rate": 4.638625288535522e-05, + "loss": 0.664, + "step": 5686 + }, + { + "epoch": 0.22, + "learning_rate": 4.638561169530649e-05, + "loss": 0.53, + "step": 5687 + }, + { + "epoch": 0.22, + "learning_rate": 4.6384970505257764e-05, + "loss": 0.6235, + "step": 5688 + }, + { + "epoch": 0.22, + "learning_rate": 4.638432931520903e-05, + "loss": 0.5581, + "step": 5689 + }, + { + "epoch": 0.22, + "learning_rate": 4.63836881251603e-05, + "loss": 0.5821, + "step": 5690 + }, + { + "epoch": 0.22, + "learning_rate": 4.6383046935111565e-05, + "loss": 0.6749, + "step": 5691 + }, + { + "epoch": 0.22, + "learning_rate": 4.638240574506284e-05, + "loss": 0.6073, + "step": 5692 + }, + { + "epoch": 0.22, + "learning_rate": 4.638176455501411e-05, + "loss": 0.5436, + "step": 5693 + }, + { + "epoch": 0.22, + "learning_rate": 4.638112336496538e-05, + "loss": 0.614, + "step": 5694 + }, + { + "epoch": 0.22, + "learning_rate": 4.6380482174916645e-05, + "loss": 0.5721, + "step": 5695 + }, + { + "epoch": 0.22, + "learning_rate": 4.637984098486792e-05, + "loss": 0.5342, + "step": 5696 + }, + { + "epoch": 0.22, + "learning_rate": 4.637919979481919e-05, + "loss": 0.5067, + "step": 5697 + }, + { + "epoch": 0.22, + "learning_rate": 4.6378558604770454e-05, + "loss": 0.6698, + "step": 5698 + }, + { + "epoch": 0.22, + "learning_rate": 4.6377917414721726e-05, + "loss": 0.5963, + "step": 5699 + }, + { + "epoch": 0.22, + "learning_rate": 4.6377276224673e-05, + "loss": 0.5558, + "step": 5700 + }, + { + "epoch": 0.22, + "learning_rate": 4.637663503462427e-05, + "loss": 0.5459, + "step": 5701 + }, + { + "epoch": 0.22, + "learning_rate": 4.6375993844575534e-05, + "loss": 0.6113, + "step": 5702 + }, + { + "epoch": 0.22, + "learning_rate": 4.6375352654526806e-05, + "loss": 0.617, + "step": 5703 + }, + { + "epoch": 0.22, + "learning_rate": 4.637471146447807e-05, + "loss": 0.6173, + "step": 5704 + }, + { + "epoch": 0.22, + "learning_rate": 4.637407027442934e-05, + "loss": 0.5912, + "step": 5705 + }, + { + "epoch": 0.22, + "learning_rate": 4.6373429084380614e-05, + "loss": 0.4969, + "step": 5706 + }, + { + "epoch": 0.22, + "learning_rate": 4.637278789433188e-05, + "loss": 0.5408, + "step": 5707 + }, + { + "epoch": 0.22, + "learning_rate": 4.637214670428315e-05, + "loss": 0.5964, + "step": 5708 + }, + { + "epoch": 0.22, + "learning_rate": 4.637150551423442e-05, + "loss": 0.553, + "step": 5709 + }, + { + "epoch": 0.22, + "learning_rate": 4.6370864324185694e-05, + "loss": 0.5779, + "step": 5710 + }, + { + "epoch": 0.22, + "learning_rate": 4.637022313413696e-05, + "loss": 0.5708, + "step": 5711 + }, + { + "epoch": 0.22, + "learning_rate": 4.636958194408823e-05, + "loss": 0.5281, + "step": 5712 + }, + { + "epoch": 0.22, + "learning_rate": 4.6368940754039496e-05, + "loss": 0.5291, + "step": 5713 + }, + { + "epoch": 0.22, + "learning_rate": 4.636829956399077e-05, + "loss": 0.5461, + "step": 5714 + }, + { + "epoch": 0.22, + "learning_rate": 4.636765837394204e-05, + "loss": 0.5757, + "step": 5715 + }, + { + "epoch": 0.22, + "learning_rate": 4.636701718389331e-05, + "loss": 0.5553, + "step": 5716 + }, + { + "epoch": 0.22, + "learning_rate": 4.636637599384458e-05, + "loss": 0.5481, + "step": 5717 + }, + { + "epoch": 0.22, + "learning_rate": 4.636573480379585e-05, + "loss": 0.5963, + "step": 5718 + }, + { + "epoch": 0.22, + "learning_rate": 4.636509361374712e-05, + "loss": 0.4933, + "step": 5719 + }, + { + "epoch": 0.22, + "learning_rate": 4.6364452423698384e-05, + "loss": 0.6487, + "step": 5720 + }, + { + "epoch": 0.22, + "learning_rate": 4.6363811233649656e-05, + "loss": 0.6875, + "step": 5721 + }, + { + "epoch": 0.22, + "learning_rate": 4.636317004360092e-05, + "loss": 0.5185, + "step": 5722 + }, + { + "epoch": 0.22, + "learning_rate": 4.636252885355219e-05, + "loss": 0.6193, + "step": 5723 + }, + { + "epoch": 0.22, + "learning_rate": 4.6361887663503464e-05, + "loss": 0.5585, + "step": 5724 + }, + { + "epoch": 0.22, + "learning_rate": 4.6361246473454736e-05, + "loss": 0.6047, + "step": 5725 + }, + { + "epoch": 0.22, + "learning_rate": 4.636060528340601e-05, + "loss": 0.6999, + "step": 5726 + }, + { + "epoch": 0.22, + "learning_rate": 4.635996409335727e-05, + "loss": 0.5637, + "step": 5727 + }, + { + "epoch": 0.22, + "learning_rate": 4.6359322903308544e-05, + "loss": 0.5981, + "step": 5728 + }, + { + "epoch": 0.22, + "learning_rate": 4.635868171325981e-05, + "loss": 0.6999, + "step": 5729 + }, + { + "epoch": 0.22, + "learning_rate": 4.635804052321108e-05, + "loss": 0.5431, + "step": 5730 + }, + { + "epoch": 0.22, + "learning_rate": 4.6357399333162346e-05, + "loss": 0.6336, + "step": 5731 + }, + { + "epoch": 0.22, + "learning_rate": 4.6356758143113624e-05, + "loss": 0.5632, + "step": 5732 + }, + { + "epoch": 0.22, + "learning_rate": 4.6356116953064896e-05, + "loss": 0.5572, + "step": 5733 + }, + { + "epoch": 0.22, + "learning_rate": 4.635547576301616e-05, + "loss": 0.5725, + "step": 5734 + }, + { + "epoch": 0.22, + "learning_rate": 4.635483457296743e-05, + "loss": 0.559, + "step": 5735 + }, + { + "epoch": 0.22, + "learning_rate": 4.63541933829187e-05, + "loss": 0.6181, + "step": 5736 + }, + { + "epoch": 0.22, + "learning_rate": 4.635355219286997e-05, + "loss": 0.6611, + "step": 5737 + }, + { + "epoch": 0.22, + "learning_rate": 4.6352911002821234e-05, + "loss": 0.5649, + "step": 5738 + }, + { + "epoch": 0.22, + "learning_rate": 4.6352269812772506e-05, + "loss": 0.5417, + "step": 5739 + }, + { + "epoch": 0.22, + "learning_rate": 4.635162862272378e-05, + "loss": 0.5809, + "step": 5740 + }, + { + "epoch": 0.22, + "learning_rate": 4.635098743267505e-05, + "loss": 0.5768, + "step": 5741 + }, + { + "epoch": 0.22, + "learning_rate": 4.635034624262632e-05, + "loss": 0.4475, + "step": 5742 + }, + { + "epoch": 0.22, + "learning_rate": 4.6349705052577586e-05, + "loss": 0.5636, + "step": 5743 + }, + { + "epoch": 0.22, + "learning_rate": 4.634906386252886e-05, + "loss": 0.615, + "step": 5744 + }, + { + "epoch": 0.22, + "learning_rate": 4.634842267248012e-05, + "loss": 0.4388, + "step": 5745 + }, + { + "epoch": 0.22, + "learning_rate": 4.6347781482431395e-05, + "loss": 0.5449, + "step": 5746 + }, + { + "epoch": 0.22, + "learning_rate": 4.634714029238266e-05, + "loss": 0.5511, + "step": 5747 + }, + { + "epoch": 0.22, + "learning_rate": 4.634649910233393e-05, + "loss": 0.5524, + "step": 5748 + }, + { + "epoch": 0.22, + "learning_rate": 4.63458579122852e-05, + "loss": 0.5933, + "step": 5749 + }, + { + "epoch": 0.22, + "learning_rate": 4.6345216722236475e-05, + "loss": 0.5959, + "step": 5750 + }, + { + "epoch": 0.22, + "learning_rate": 4.6344575532187746e-05, + "loss": 0.5411, + "step": 5751 + }, + { + "epoch": 0.22, + "learning_rate": 4.634393434213901e-05, + "loss": 0.6129, + "step": 5752 + }, + { + "epoch": 0.22, + "learning_rate": 4.634329315209028e-05, + "loss": 0.5237, + "step": 5753 + }, + { + "epoch": 0.22, + "learning_rate": 4.634265196204155e-05, + "loss": 0.5533, + "step": 5754 + }, + { + "epoch": 0.22, + "learning_rate": 4.634201077199282e-05, + "loss": 0.5495, + "step": 5755 + }, + { + "epoch": 0.22, + "learning_rate": 4.634136958194409e-05, + "loss": 0.5745, + "step": 5756 + }, + { + "epoch": 0.22, + "learning_rate": 4.634072839189536e-05, + "loss": 0.5485, + "step": 5757 + }, + { + "epoch": 0.22, + "learning_rate": 4.634008720184663e-05, + "loss": 0.5561, + "step": 5758 + }, + { + "epoch": 0.22, + "learning_rate": 4.63394460117979e-05, + "loss": 0.5496, + "step": 5759 + }, + { + "epoch": 0.22, + "learning_rate": 4.633880482174917e-05, + "loss": 0.5064, + "step": 5760 + }, + { + "epoch": 0.22, + "learning_rate": 4.6338163631700436e-05, + "loss": 0.6714, + "step": 5761 + }, + { + "epoch": 0.22, + "learning_rate": 4.633752244165171e-05, + "loss": 0.5971, + "step": 5762 + }, + { + "epoch": 0.22, + "learning_rate": 4.633688125160297e-05, + "loss": 0.5829, + "step": 5763 + }, + { + "epoch": 0.22, + "learning_rate": 4.6336240061554245e-05, + "loss": 0.5244, + "step": 5764 + }, + { + "epoch": 0.22, + "learning_rate": 4.6335598871505517e-05, + "loss": 0.6446, + "step": 5765 + }, + { + "epoch": 0.22, + "learning_rate": 4.633495768145679e-05, + "loss": 0.6204, + "step": 5766 + }, + { + "epoch": 0.22, + "learning_rate": 4.633431649140805e-05, + "loss": 0.5605, + "step": 5767 + }, + { + "epoch": 0.22, + "learning_rate": 4.6333675301359325e-05, + "loss": 0.5209, + "step": 5768 + }, + { + "epoch": 0.22, + "learning_rate": 4.63330341113106e-05, + "loss": 0.6049, + "step": 5769 + }, + { + "epoch": 0.22, + "learning_rate": 4.633239292126186e-05, + "loss": 0.5841, + "step": 5770 + }, + { + "epoch": 0.22, + "learning_rate": 4.633175173121313e-05, + "loss": 0.6319, + "step": 5771 + }, + { + "epoch": 0.22, + "learning_rate": 4.6331110541164405e-05, + "loss": 0.5427, + "step": 5772 + }, + { + "epoch": 0.22, + "learning_rate": 4.633046935111568e-05, + "loss": 0.6518, + "step": 5773 + }, + { + "epoch": 0.22, + "learning_rate": 4.632982816106694e-05, + "loss": 0.6227, + "step": 5774 + }, + { + "epoch": 0.22, + "learning_rate": 4.632918697101821e-05, + "loss": 0.5731, + "step": 5775 + }, + { + "epoch": 0.22, + "learning_rate": 4.632854578096948e-05, + "loss": 0.5664, + "step": 5776 + }, + { + "epoch": 0.22, + "learning_rate": 4.632790459092075e-05, + "loss": 0.6197, + "step": 5777 + }, + { + "epoch": 0.22, + "learning_rate": 4.632726340087202e-05, + "loss": 0.5783, + "step": 5778 + }, + { + "epoch": 0.22, + "learning_rate": 4.632662221082329e-05, + "loss": 0.5997, + "step": 5779 + }, + { + "epoch": 0.22, + "learning_rate": 4.632598102077456e-05, + "loss": 0.4902, + "step": 5780 + }, + { + "epoch": 0.22, + "learning_rate": 4.632533983072583e-05, + "loss": 0.6295, + "step": 5781 + }, + { + "epoch": 0.22, + "learning_rate": 4.63246986406771e-05, + "loss": 0.5251, + "step": 5782 + }, + { + "epoch": 0.22, + "learning_rate": 4.632405745062837e-05, + "loss": 0.5685, + "step": 5783 + }, + { + "epoch": 0.22, + "learning_rate": 4.632341626057964e-05, + "loss": 0.5813, + "step": 5784 + }, + { + "epoch": 0.22, + "learning_rate": 4.6322775070530903e-05, + "loss": 0.5529, + "step": 5785 + }, + { + "epoch": 0.22, + "learning_rate": 4.6322133880482175e-05, + "loss": 0.6171, + "step": 5786 + }, + { + "epoch": 0.22, + "learning_rate": 4.632149269043345e-05, + "loss": 0.6308, + "step": 5787 + }, + { + "epoch": 0.22, + "learning_rate": 4.632085150038472e-05, + "loss": 0.5517, + "step": 5788 + }, + { + "epoch": 0.22, + "learning_rate": 4.632021031033599e-05, + "loss": 0.6332, + "step": 5789 + }, + { + "epoch": 0.22, + "learning_rate": 4.6319569120287255e-05, + "loss": 0.6257, + "step": 5790 + }, + { + "epoch": 0.22, + "learning_rate": 4.631892793023853e-05, + "loss": 0.547, + "step": 5791 + }, + { + "epoch": 0.22, + "learning_rate": 4.631828674018979e-05, + "loss": 0.6428, + "step": 5792 + }, + { + "epoch": 0.22, + "learning_rate": 4.6317645550141064e-05, + "loss": 0.5865, + "step": 5793 + }, + { + "epoch": 0.22, + "learning_rate": 4.631700436009233e-05, + "loss": 0.5675, + "step": 5794 + }, + { + "epoch": 0.22, + "learning_rate": 4.63163631700436e-05, + "loss": 0.5927, + "step": 5795 + }, + { + "epoch": 0.22, + "learning_rate": 4.631572197999487e-05, + "loss": 0.5974, + "step": 5796 + }, + { + "epoch": 0.22, + "learning_rate": 4.6315080789946144e-05, + "loss": 0.6075, + "step": 5797 + }, + { + "epoch": 0.22, + "learning_rate": 4.6314439599897415e-05, + "loss": 0.5486, + "step": 5798 + }, + { + "epoch": 0.22, + "learning_rate": 4.631379840984868e-05, + "loss": 0.6037, + "step": 5799 + }, + { + "epoch": 0.22, + "learning_rate": 4.631315721979995e-05, + "loss": 0.597, + "step": 5800 + }, + { + "epoch": 0.22, + "learning_rate": 4.631251602975122e-05, + "loss": 0.6399, + "step": 5801 + }, + { + "epoch": 0.22, + "learning_rate": 4.631187483970249e-05, + "loss": 0.6056, + "step": 5802 + }, + { + "epoch": 0.22, + "learning_rate": 4.6311233649653754e-05, + "loss": 0.6122, + "step": 5803 + }, + { + "epoch": 0.22, + "learning_rate": 4.631059245960503e-05, + "loss": 0.6072, + "step": 5804 + }, + { + "epoch": 0.22, + "learning_rate": 4.6309951269556304e-05, + "loss": 0.5792, + "step": 5805 + }, + { + "epoch": 0.22, + "learning_rate": 4.630931007950757e-05, + "loss": 0.6876, + "step": 5806 + }, + { + "epoch": 0.22, + "learning_rate": 4.630866888945884e-05, + "loss": 0.53, + "step": 5807 + }, + { + "epoch": 0.22, + "learning_rate": 4.6308027699410105e-05, + "loss": 0.5565, + "step": 5808 + }, + { + "epoch": 0.22, + "learning_rate": 4.630738650936138e-05, + "loss": 0.56, + "step": 5809 + }, + { + "epoch": 0.22, + "learning_rate": 4.630674531931264e-05, + "loss": 0.6542, + "step": 5810 + }, + { + "epoch": 0.22, + "learning_rate": 4.6306104129263914e-05, + "loss": 0.5303, + "step": 5811 + }, + { + "epoch": 0.22, + "learning_rate": 4.6305462939215186e-05, + "loss": 0.6102, + "step": 5812 + }, + { + "epoch": 0.22, + "learning_rate": 4.630482174916646e-05, + "loss": 0.4873, + "step": 5813 + }, + { + "epoch": 0.22, + "learning_rate": 4.630418055911773e-05, + "loss": 0.6638, + "step": 5814 + }, + { + "epoch": 0.22, + "learning_rate": 4.6303539369068994e-05, + "loss": 0.5966, + "step": 5815 + }, + { + "epoch": 0.22, + "learning_rate": 4.6302898179020266e-05, + "loss": 0.585, + "step": 5816 + }, + { + "epoch": 0.22, + "learning_rate": 4.630225698897153e-05, + "loss": 0.5656, + "step": 5817 + }, + { + "epoch": 0.22, + "learning_rate": 4.63016157989228e-05, + "loss": 0.6232, + "step": 5818 + }, + { + "epoch": 0.22, + "learning_rate": 4.630097460887407e-05, + "loss": 0.5896, + "step": 5819 + }, + { + "epoch": 0.22, + "learning_rate": 4.630033341882534e-05, + "loss": 0.5808, + "step": 5820 + }, + { + "epoch": 0.22, + "learning_rate": 4.629969222877661e-05, + "loss": 0.5159, + "step": 5821 + }, + { + "epoch": 0.22, + "learning_rate": 4.629905103872788e-05, + "loss": 0.6708, + "step": 5822 + }, + { + "epoch": 0.22, + "learning_rate": 4.6298409848679154e-05, + "loss": 0.4733, + "step": 5823 + }, + { + "epoch": 0.22, + "learning_rate": 4.629776865863042e-05, + "loss": 0.5394, + "step": 5824 + }, + { + "epoch": 0.22, + "learning_rate": 4.629712746858169e-05, + "loss": 0.6082, + "step": 5825 + }, + { + "epoch": 0.22, + "learning_rate": 4.6296486278532956e-05, + "loss": 0.4752, + "step": 5826 + }, + { + "epoch": 0.22, + "learning_rate": 4.629584508848423e-05, + "loss": 0.5838, + "step": 5827 + }, + { + "epoch": 0.22, + "learning_rate": 4.62952038984355e-05, + "loss": 0.5687, + "step": 5828 + }, + { + "epoch": 0.22, + "learning_rate": 4.629456270838677e-05, + "loss": 0.5728, + "step": 5829 + }, + { + "epoch": 0.22, + "learning_rate": 4.629392151833804e-05, + "loss": 0.4755, + "step": 5830 + }, + { + "epoch": 0.22, + "learning_rate": 4.629328032828931e-05, + "loss": 0.4989, + "step": 5831 + }, + { + "epoch": 0.22, + "learning_rate": 4.629263913824058e-05, + "loss": 0.6057, + "step": 5832 + }, + { + "epoch": 0.22, + "learning_rate": 4.6291997948191844e-05, + "loss": 0.5516, + "step": 5833 + }, + { + "epoch": 0.22, + "learning_rate": 4.6291356758143116e-05, + "loss": 0.6755, + "step": 5834 + }, + { + "epoch": 0.22, + "learning_rate": 4.629071556809438e-05, + "loss": 0.6486, + "step": 5835 + }, + { + "epoch": 0.22, + "learning_rate": 4.629007437804565e-05, + "loss": 0.5444, + "step": 5836 + }, + { + "epoch": 0.22, + "learning_rate": 4.6289433187996924e-05, + "loss": 0.6927, + "step": 5837 + }, + { + "epoch": 0.22, + "learning_rate": 4.6288791997948196e-05, + "loss": 0.5554, + "step": 5838 + }, + { + "epoch": 0.22, + "learning_rate": 4.628815080789947e-05, + "loss": 0.5573, + "step": 5839 + }, + { + "epoch": 0.22, + "learning_rate": 4.628750961785073e-05, + "loss": 0.7242, + "step": 5840 + }, + { + "epoch": 0.22, + "learning_rate": 4.6286868427802004e-05, + "loss": 0.6407, + "step": 5841 + }, + { + "epoch": 0.22, + "learning_rate": 4.628622723775327e-05, + "loss": 0.4915, + "step": 5842 + }, + { + "epoch": 0.22, + "learning_rate": 4.628558604770454e-05, + "loss": 0.6218, + "step": 5843 + }, + { + "epoch": 0.22, + "learning_rate": 4.628494485765581e-05, + "loss": 0.5854, + "step": 5844 + }, + { + "epoch": 0.22, + "learning_rate": 4.6284303667607084e-05, + "loss": 0.7167, + "step": 5845 + }, + { + "epoch": 0.22, + "learning_rate": 4.628366247755835e-05, + "loss": 0.5403, + "step": 5846 + }, + { + "epoch": 0.22, + "learning_rate": 4.628302128750962e-05, + "loss": 0.6191, + "step": 5847 + }, + { + "epoch": 0.22, + "learning_rate": 4.628238009746089e-05, + "loss": 0.6486, + "step": 5848 + }, + { + "epoch": 0.22, + "learning_rate": 4.628173890741216e-05, + "loss": 0.6275, + "step": 5849 + }, + { + "epoch": 0.22, + "learning_rate": 4.628109771736343e-05, + "loss": 0.5982, + "step": 5850 + }, + { + "epoch": 0.22, + "learning_rate": 4.6280456527314694e-05, + "loss": 0.6111, + "step": 5851 + }, + { + "epoch": 0.22, + "learning_rate": 4.6279815337265966e-05, + "loss": 0.5598, + "step": 5852 + }, + { + "epoch": 0.23, + "learning_rate": 4.627917414721724e-05, + "loss": 0.5768, + "step": 5853 + }, + { + "epoch": 0.23, + "learning_rate": 4.627853295716851e-05, + "loss": 0.524, + "step": 5854 + }, + { + "epoch": 0.23, + "learning_rate": 4.6277891767119775e-05, + "loss": 0.5483, + "step": 5855 + }, + { + "epoch": 0.23, + "learning_rate": 4.6277250577071046e-05, + "loss": 0.5656, + "step": 5856 + }, + { + "epoch": 0.23, + "learning_rate": 4.627660938702232e-05, + "loss": 0.5493, + "step": 5857 + }, + { + "epoch": 0.23, + "learning_rate": 4.627596819697358e-05, + "loss": 0.5621, + "step": 5858 + }, + { + "epoch": 0.23, + "learning_rate": 4.6275327006924855e-05, + "loss": 0.4702, + "step": 5859 + }, + { + "epoch": 0.23, + "learning_rate": 4.6274685816876126e-05, + "loss": 0.5802, + "step": 5860 + }, + { + "epoch": 0.23, + "learning_rate": 4.62740446268274e-05, + "loss": 0.5735, + "step": 5861 + }, + { + "epoch": 0.23, + "learning_rate": 4.627340343677866e-05, + "loss": 0.6663, + "step": 5862 + }, + { + "epoch": 0.23, + "learning_rate": 4.6272762246729935e-05, + "loss": 0.5396, + "step": 5863 + }, + { + "epoch": 0.23, + "learning_rate": 4.62721210566812e-05, + "loss": 0.5909, + "step": 5864 + }, + { + "epoch": 0.23, + "learning_rate": 4.627147986663247e-05, + "loss": 0.6109, + "step": 5865 + }, + { + "epoch": 0.23, + "learning_rate": 4.627083867658374e-05, + "loss": 0.4909, + "step": 5866 + }, + { + "epoch": 0.23, + "learning_rate": 4.627019748653501e-05, + "loss": 0.6462, + "step": 5867 + }, + { + "epoch": 0.23, + "learning_rate": 4.626955629648628e-05, + "loss": 0.6522, + "step": 5868 + }, + { + "epoch": 0.23, + "learning_rate": 4.626891510643755e-05, + "loss": 0.6736, + "step": 5869 + }, + { + "epoch": 0.23, + "learning_rate": 4.626827391638882e-05, + "loss": 0.5949, + "step": 5870 + }, + { + "epoch": 0.23, + "learning_rate": 4.626763272634009e-05, + "loss": 0.6469, + "step": 5871 + }, + { + "epoch": 0.23, + "learning_rate": 4.626699153629136e-05, + "loss": 0.5804, + "step": 5872 + }, + { + "epoch": 0.23, + "learning_rate": 4.6266350346242625e-05, + "loss": 0.6585, + "step": 5873 + }, + { + "epoch": 0.23, + "learning_rate": 4.6265709156193896e-05, + "loss": 0.6039, + "step": 5874 + }, + { + "epoch": 0.23, + "learning_rate": 4.626506796614517e-05, + "loss": 0.586, + "step": 5875 + }, + { + "epoch": 0.23, + "learning_rate": 4.626442677609644e-05, + "loss": 0.5606, + "step": 5876 + }, + { + "epoch": 0.23, + "learning_rate": 4.626378558604771e-05, + "loss": 0.5511, + "step": 5877 + }, + { + "epoch": 0.23, + "learning_rate": 4.6263144395998977e-05, + "loss": 0.6224, + "step": 5878 + }, + { + "epoch": 0.23, + "learning_rate": 4.626250320595025e-05, + "loss": 0.5954, + "step": 5879 + }, + { + "epoch": 0.23, + "learning_rate": 4.626186201590151e-05, + "loss": 0.6125, + "step": 5880 + }, + { + "epoch": 0.23, + "learning_rate": 4.6261220825852785e-05, + "loss": 0.5921, + "step": 5881 + }, + { + "epoch": 0.23, + "learning_rate": 4.626057963580405e-05, + "loss": 0.5984, + "step": 5882 + }, + { + "epoch": 0.23, + "learning_rate": 4.625993844575532e-05, + "loss": 0.5462, + "step": 5883 + }, + { + "epoch": 0.23, + "learning_rate": 4.625929725570659e-05, + "loss": 0.5477, + "step": 5884 + }, + { + "epoch": 0.23, + "learning_rate": 4.6258656065657865e-05, + "loss": 0.6276, + "step": 5885 + }, + { + "epoch": 0.23, + "learning_rate": 4.625801487560914e-05, + "loss": 0.5431, + "step": 5886 + }, + { + "epoch": 0.23, + "learning_rate": 4.62573736855604e-05, + "loss": 0.6324, + "step": 5887 + }, + { + "epoch": 0.23, + "learning_rate": 4.6256732495511673e-05, + "loss": 0.4963, + "step": 5888 + }, + { + "epoch": 0.23, + "learning_rate": 4.625609130546294e-05, + "loss": 0.5515, + "step": 5889 + }, + { + "epoch": 0.23, + "learning_rate": 4.625545011541421e-05, + "loss": 0.5586, + "step": 5890 + }, + { + "epoch": 0.23, + "learning_rate": 4.6254808925365475e-05, + "loss": 0.5662, + "step": 5891 + }, + { + "epoch": 0.23, + "learning_rate": 4.6254167735316754e-05, + "loss": 0.6435, + "step": 5892 + }, + { + "epoch": 0.23, + "learning_rate": 4.6253526545268025e-05, + "loss": 0.5003, + "step": 5893 + }, + { + "epoch": 0.23, + "learning_rate": 4.625288535521929e-05, + "loss": 0.5494, + "step": 5894 + }, + { + "epoch": 0.23, + "learning_rate": 4.625224416517056e-05, + "loss": 0.5056, + "step": 5895 + }, + { + "epoch": 0.23, + "learning_rate": 4.625160297512183e-05, + "loss": 0.6057, + "step": 5896 + }, + { + "epoch": 0.23, + "learning_rate": 4.62509617850731e-05, + "loss": 0.5023, + "step": 5897 + }, + { + "epoch": 0.23, + "learning_rate": 4.6250320595024363e-05, + "loss": 0.6655, + "step": 5898 + }, + { + "epoch": 0.23, + "learning_rate": 4.6249679404975635e-05, + "loss": 0.6478, + "step": 5899 + }, + { + "epoch": 0.23, + "learning_rate": 4.624903821492691e-05, + "loss": 0.5099, + "step": 5900 + }, + { + "epoch": 0.23, + "learning_rate": 4.624839702487818e-05, + "loss": 0.6, + "step": 5901 + }, + { + "epoch": 0.23, + "learning_rate": 4.624775583482945e-05, + "loss": 0.522, + "step": 5902 + }, + { + "epoch": 0.23, + "learning_rate": 4.6247114644780715e-05, + "loss": 0.5412, + "step": 5903 + }, + { + "epoch": 0.23, + "learning_rate": 4.624647345473199e-05, + "loss": 0.4974, + "step": 5904 + }, + { + "epoch": 0.23, + "learning_rate": 4.624583226468325e-05, + "loss": 0.572, + "step": 5905 + }, + { + "epoch": 0.23, + "learning_rate": 4.6245191074634524e-05, + "loss": 0.6212, + "step": 5906 + }, + { + "epoch": 0.23, + "learning_rate": 4.624454988458579e-05, + "loss": 0.6175, + "step": 5907 + }, + { + "epoch": 0.23, + "learning_rate": 4.624390869453706e-05, + "loss": 0.6581, + "step": 5908 + }, + { + "epoch": 0.23, + "learning_rate": 4.624326750448833e-05, + "loss": 0.5977, + "step": 5909 + }, + { + "epoch": 0.23, + "learning_rate": 4.6242626314439604e-05, + "loss": 0.4987, + "step": 5910 + }, + { + "epoch": 0.23, + "learning_rate": 4.6241985124390875e-05, + "loss": 0.5882, + "step": 5911 + }, + { + "epoch": 0.23, + "learning_rate": 4.624134393434214e-05, + "loss": 0.6158, + "step": 5912 + }, + { + "epoch": 0.23, + "learning_rate": 4.624070274429341e-05, + "loss": 0.6415, + "step": 5913 + }, + { + "epoch": 0.23, + "learning_rate": 4.624006155424468e-05, + "loss": 0.4937, + "step": 5914 + }, + { + "epoch": 0.23, + "learning_rate": 4.623942036419595e-05, + "loss": 0.495, + "step": 5915 + }, + { + "epoch": 0.23, + "learning_rate": 4.623877917414722e-05, + "loss": 0.5685, + "step": 5916 + }, + { + "epoch": 0.23, + "learning_rate": 4.623813798409849e-05, + "loss": 0.4408, + "step": 5917 + }, + { + "epoch": 0.23, + "learning_rate": 4.623749679404976e-05, + "loss": 0.5566, + "step": 5918 + }, + { + "epoch": 0.23, + "learning_rate": 4.623685560400103e-05, + "loss": 0.6394, + "step": 5919 + }, + { + "epoch": 0.23, + "learning_rate": 4.62362144139523e-05, + "loss": 0.6567, + "step": 5920 + }, + { + "epoch": 0.23, + "learning_rate": 4.6235573223903566e-05, + "loss": 0.5668, + "step": 5921 + }, + { + "epoch": 0.23, + "learning_rate": 4.623493203385484e-05, + "loss": 0.4959, + "step": 5922 + }, + { + "epoch": 0.23, + "learning_rate": 4.62342908438061e-05, + "loss": 0.6902, + "step": 5923 + }, + { + "epoch": 0.23, + "learning_rate": 4.6233649653757374e-05, + "loss": 0.5148, + "step": 5924 + }, + { + "epoch": 0.23, + "learning_rate": 4.6233008463708646e-05, + "loss": 0.578, + "step": 5925 + }, + { + "epoch": 0.23, + "learning_rate": 4.623236727365992e-05, + "loss": 0.6326, + "step": 5926 + }, + { + "epoch": 0.23, + "learning_rate": 4.623172608361118e-05, + "loss": 0.5758, + "step": 5927 + }, + { + "epoch": 0.23, + "learning_rate": 4.6231084893562454e-05, + "loss": 0.6133, + "step": 5928 + }, + { + "epoch": 0.23, + "learning_rate": 4.6230443703513726e-05, + "loss": 0.6324, + "step": 5929 + }, + { + "epoch": 0.23, + "learning_rate": 4.622980251346499e-05, + "loss": 0.5162, + "step": 5930 + }, + { + "epoch": 0.23, + "learning_rate": 4.622916132341626e-05, + "loss": 0.6086, + "step": 5931 + }, + { + "epoch": 0.23, + "learning_rate": 4.6228520133367534e-05, + "loss": 0.7215, + "step": 5932 + }, + { + "epoch": 0.23, + "learning_rate": 4.6227878943318806e-05, + "loss": 0.6307, + "step": 5933 + }, + { + "epoch": 0.23, + "learning_rate": 4.622723775327007e-05, + "loss": 0.6129, + "step": 5934 + }, + { + "epoch": 0.23, + "learning_rate": 4.622659656322134e-05, + "loss": 0.6414, + "step": 5935 + }, + { + "epoch": 0.23, + "learning_rate": 4.622595537317261e-05, + "loss": 0.5673, + "step": 5936 + }, + { + "epoch": 0.23, + "learning_rate": 4.622531418312388e-05, + "loss": 0.5046, + "step": 5937 + }, + { + "epoch": 0.23, + "learning_rate": 4.622467299307515e-05, + "loss": 0.5373, + "step": 5938 + }, + { + "epoch": 0.23, + "learning_rate": 4.6224031803026416e-05, + "loss": 0.5754, + "step": 5939 + }, + { + "epoch": 0.23, + "learning_rate": 4.622339061297769e-05, + "loss": 0.5567, + "step": 5940 + }, + { + "epoch": 0.23, + "learning_rate": 4.622274942292896e-05, + "loss": 0.6566, + "step": 5941 + }, + { + "epoch": 0.23, + "learning_rate": 4.622210823288023e-05, + "loss": 0.5699, + "step": 5942 + }, + { + "epoch": 0.23, + "learning_rate": 4.6221467042831496e-05, + "loss": 0.5069, + "step": 5943 + }, + { + "epoch": 0.23, + "learning_rate": 4.622082585278277e-05, + "loss": 0.6328, + "step": 5944 + }, + { + "epoch": 0.23, + "learning_rate": 4.622018466273403e-05, + "loss": 0.5732, + "step": 5945 + }, + { + "epoch": 0.23, + "learning_rate": 4.6219543472685304e-05, + "loss": 0.5305, + "step": 5946 + }, + { + "epoch": 0.23, + "learning_rate": 4.6218902282636576e-05, + "loss": 0.607, + "step": 5947 + }, + { + "epoch": 0.23, + "learning_rate": 4.621826109258785e-05, + "loss": 0.6067, + "step": 5948 + }, + { + "epoch": 0.23, + "learning_rate": 4.621761990253912e-05, + "loss": 0.6104, + "step": 5949 + }, + { + "epoch": 0.23, + "learning_rate": 4.6216978712490384e-05, + "loss": 0.6207, + "step": 5950 + }, + { + "epoch": 0.23, + "learning_rate": 4.6216337522441656e-05, + "loss": 0.6952, + "step": 5951 + }, + { + "epoch": 0.23, + "learning_rate": 4.621569633239292e-05, + "loss": 0.6317, + "step": 5952 + }, + { + "epoch": 0.23, + "learning_rate": 4.621505514234419e-05, + "loss": 0.6649, + "step": 5953 + }, + { + "epoch": 0.23, + "learning_rate": 4.621441395229546e-05, + "loss": 0.6678, + "step": 5954 + }, + { + "epoch": 0.23, + "learning_rate": 4.621377276224673e-05, + "loss": 0.5548, + "step": 5955 + }, + { + "epoch": 0.23, + "learning_rate": 4.6213131572198e-05, + "loss": 0.5464, + "step": 5956 + }, + { + "epoch": 0.23, + "learning_rate": 4.621249038214927e-05, + "loss": 0.5816, + "step": 5957 + }, + { + "epoch": 0.23, + "learning_rate": 4.6211849192100545e-05, + "loss": 0.54, + "step": 5958 + }, + { + "epoch": 0.23, + "learning_rate": 4.621120800205181e-05, + "loss": 0.6094, + "step": 5959 + }, + { + "epoch": 0.23, + "learning_rate": 4.621056681200308e-05, + "loss": 0.633, + "step": 5960 + }, + { + "epoch": 0.23, + "learning_rate": 4.6209925621954346e-05, + "loss": 0.6383, + "step": 5961 + }, + { + "epoch": 0.23, + "learning_rate": 4.620928443190562e-05, + "loss": 0.6573, + "step": 5962 + }, + { + "epoch": 0.23, + "learning_rate": 4.620864324185688e-05, + "loss": 0.597, + "step": 5963 + }, + { + "epoch": 0.23, + "learning_rate": 4.620800205180816e-05, + "loss": 0.6484, + "step": 5964 + }, + { + "epoch": 0.23, + "learning_rate": 4.620736086175943e-05, + "loss": 0.6276, + "step": 5965 + }, + { + "epoch": 0.23, + "learning_rate": 4.62067196717107e-05, + "loss": 0.6105, + "step": 5966 + }, + { + "epoch": 0.23, + "learning_rate": 4.620607848166197e-05, + "loss": 0.5778, + "step": 5967 + }, + { + "epoch": 0.23, + "learning_rate": 4.6205437291613235e-05, + "loss": 0.6225, + "step": 5968 + }, + { + "epoch": 0.23, + "learning_rate": 4.6204796101564506e-05, + "loss": 0.5402, + "step": 5969 + }, + { + "epoch": 0.23, + "learning_rate": 4.620415491151577e-05, + "loss": 0.6649, + "step": 5970 + }, + { + "epoch": 0.23, + "learning_rate": 4.620351372146704e-05, + "loss": 0.5124, + "step": 5971 + }, + { + "epoch": 0.23, + "learning_rate": 4.6202872531418315e-05, + "loss": 0.5504, + "step": 5972 + }, + { + "epoch": 0.23, + "learning_rate": 4.6202231341369586e-05, + "loss": 0.6115, + "step": 5973 + }, + { + "epoch": 0.23, + "learning_rate": 4.620159015132086e-05, + "loss": 0.5358, + "step": 5974 + }, + { + "epoch": 0.23, + "learning_rate": 4.620094896127212e-05, + "loss": 0.579, + "step": 5975 + }, + { + "epoch": 0.23, + "learning_rate": 4.6200307771223395e-05, + "loss": 0.6116, + "step": 5976 + }, + { + "epoch": 0.23, + "learning_rate": 4.619966658117466e-05, + "loss": 0.5972, + "step": 5977 + }, + { + "epoch": 0.23, + "learning_rate": 4.619902539112593e-05, + "loss": 0.5908, + "step": 5978 + }, + { + "epoch": 0.23, + "learning_rate": 4.6198384201077196e-05, + "loss": 0.6189, + "step": 5979 + }, + { + "epoch": 0.23, + "learning_rate": 4.6197743011028475e-05, + "loss": 0.6632, + "step": 5980 + }, + { + "epoch": 0.23, + "learning_rate": 4.619710182097974e-05, + "loss": 0.6326, + "step": 5981 + }, + { + "epoch": 0.23, + "learning_rate": 4.619646063093101e-05, + "loss": 0.5087, + "step": 5982 + }, + { + "epoch": 0.23, + "learning_rate": 4.619581944088228e-05, + "loss": 0.519, + "step": 5983 + }, + { + "epoch": 0.23, + "learning_rate": 4.619517825083355e-05, + "loss": 0.4974, + "step": 5984 + }, + { + "epoch": 0.23, + "learning_rate": 4.619453706078482e-05, + "loss": 0.5338, + "step": 5985 + }, + { + "epoch": 0.23, + "learning_rate": 4.6193895870736085e-05, + "loss": 0.5305, + "step": 5986 + }, + { + "epoch": 0.23, + "learning_rate": 4.6193254680687357e-05, + "loss": 0.5598, + "step": 5987 + }, + { + "epoch": 0.23, + "learning_rate": 4.619261349063863e-05, + "loss": 0.5538, + "step": 5988 + }, + { + "epoch": 0.23, + "learning_rate": 4.61919723005899e-05, + "loss": 0.5909, + "step": 5989 + }, + { + "epoch": 0.23, + "learning_rate": 4.6191331110541165e-05, + "loss": 0.7141, + "step": 5990 + }, + { + "epoch": 0.23, + "learning_rate": 4.619068992049244e-05, + "loss": 0.5987, + "step": 5991 + }, + { + "epoch": 0.23, + "learning_rate": 4.619004873044371e-05, + "loss": 0.5889, + "step": 5992 + }, + { + "epoch": 0.23, + "learning_rate": 4.618940754039497e-05, + "loss": 0.6761, + "step": 5993 + }, + { + "epoch": 0.23, + "learning_rate": 4.6188766350346245e-05, + "loss": 0.5909, + "step": 5994 + }, + { + "epoch": 0.23, + "learning_rate": 4.618812516029751e-05, + "loss": 0.6906, + "step": 5995 + }, + { + "epoch": 0.23, + "learning_rate": 4.618748397024878e-05, + "loss": 0.5841, + "step": 5996 + }, + { + "epoch": 0.23, + "learning_rate": 4.618684278020005e-05, + "loss": 0.5556, + "step": 5997 + }, + { + "epoch": 0.23, + "learning_rate": 4.6186201590151325e-05, + "loss": 0.6706, + "step": 5998 + }, + { + "epoch": 0.23, + "learning_rate": 4.618556040010259e-05, + "loss": 0.5151, + "step": 5999 + }, + { + "epoch": 0.23, + "learning_rate": 4.618491921005386e-05, + "loss": 0.6174, + "step": 6000 + }, + { + "epoch": 0.23, + "learning_rate": 4.6184278020005133e-05, + "loss": 0.5677, + "step": 6001 + }, + { + "epoch": 0.23, + "learning_rate": 4.61836368299564e-05, + "loss": 0.6347, + "step": 6002 + }, + { + "epoch": 0.23, + "learning_rate": 4.618299563990767e-05, + "loss": 0.5541, + "step": 6003 + }, + { + "epoch": 0.23, + "learning_rate": 4.618235444985894e-05, + "loss": 0.5345, + "step": 6004 + }, + { + "epoch": 0.23, + "learning_rate": 4.6181713259810214e-05, + "loss": 0.5893, + "step": 6005 + }, + { + "epoch": 0.23, + "learning_rate": 4.618107206976148e-05, + "loss": 0.6355, + "step": 6006 + }, + { + "epoch": 0.23, + "learning_rate": 4.618043087971275e-05, + "loss": 0.5532, + "step": 6007 + }, + { + "epoch": 0.23, + "learning_rate": 4.6179789689664015e-05, + "loss": 0.6266, + "step": 6008 + }, + { + "epoch": 0.23, + "learning_rate": 4.617914849961529e-05, + "loss": 0.6379, + "step": 6009 + }, + { + "epoch": 0.23, + "learning_rate": 4.617850730956656e-05, + "loss": 0.621, + "step": 6010 + }, + { + "epoch": 0.23, + "learning_rate": 4.6177866119517824e-05, + "loss": 0.5495, + "step": 6011 + }, + { + "epoch": 0.23, + "learning_rate": 4.6177224929469095e-05, + "loss": 0.6342, + "step": 6012 + }, + { + "epoch": 0.23, + "learning_rate": 4.617658373942037e-05, + "loss": 0.6084, + "step": 6013 + }, + { + "epoch": 0.23, + "learning_rate": 4.617594254937164e-05, + "loss": 0.5716, + "step": 6014 + }, + { + "epoch": 0.23, + "learning_rate": 4.6175301359322904e-05, + "loss": 0.6521, + "step": 6015 + }, + { + "epoch": 0.23, + "learning_rate": 4.6174660169274175e-05, + "loss": 0.6254, + "step": 6016 + }, + { + "epoch": 0.23, + "learning_rate": 4.617401897922545e-05, + "loss": 0.5272, + "step": 6017 + }, + { + "epoch": 0.23, + "learning_rate": 4.617337778917671e-05, + "loss": 0.5703, + "step": 6018 + }, + { + "epoch": 0.23, + "learning_rate": 4.6172736599127984e-05, + "loss": 0.5562, + "step": 6019 + }, + { + "epoch": 0.23, + "learning_rate": 4.6172095409079255e-05, + "loss": 0.5542, + "step": 6020 + }, + { + "epoch": 0.23, + "learning_rate": 4.617145421903053e-05, + "loss": 0.5391, + "step": 6021 + }, + { + "epoch": 0.23, + "learning_rate": 4.617081302898179e-05, + "loss": 0.5786, + "step": 6022 + }, + { + "epoch": 0.23, + "learning_rate": 4.6170171838933064e-05, + "loss": 0.5329, + "step": 6023 + }, + { + "epoch": 0.23, + "learning_rate": 4.616953064888433e-05, + "loss": 0.5426, + "step": 6024 + }, + { + "epoch": 0.23, + "learning_rate": 4.61688894588356e-05, + "loss": 0.5155, + "step": 6025 + }, + { + "epoch": 0.23, + "learning_rate": 4.616824826878687e-05, + "loss": 0.5329, + "step": 6026 + }, + { + "epoch": 0.23, + "learning_rate": 4.616760707873814e-05, + "loss": 0.5004, + "step": 6027 + }, + { + "epoch": 0.23, + "learning_rate": 4.616696588868941e-05, + "loss": 0.5695, + "step": 6028 + }, + { + "epoch": 0.23, + "learning_rate": 4.616632469864068e-05, + "loss": 0.6446, + "step": 6029 + }, + { + "epoch": 0.23, + "learning_rate": 4.616568350859195e-05, + "loss": 0.5484, + "step": 6030 + }, + { + "epoch": 0.23, + "learning_rate": 4.616504231854322e-05, + "loss": 0.5256, + "step": 6031 + }, + { + "epoch": 0.23, + "learning_rate": 4.616440112849449e-05, + "loss": 0.557, + "step": 6032 + }, + { + "epoch": 0.23, + "learning_rate": 4.6163759938445754e-05, + "loss": 0.6909, + "step": 6033 + }, + { + "epoch": 0.23, + "learning_rate": 4.6163118748397026e-05, + "loss": 0.6541, + "step": 6034 + }, + { + "epoch": 0.23, + "learning_rate": 4.61624775583483e-05, + "loss": 0.6766, + "step": 6035 + }, + { + "epoch": 0.23, + "learning_rate": 4.616183636829957e-05, + "loss": 0.5117, + "step": 6036 + }, + { + "epoch": 0.23, + "learning_rate": 4.616119517825084e-05, + "loss": 0.5983, + "step": 6037 + }, + { + "epoch": 0.23, + "learning_rate": 4.6160553988202106e-05, + "loss": 0.5281, + "step": 6038 + }, + { + "epoch": 0.23, + "learning_rate": 4.615991279815338e-05, + "loss": 0.5756, + "step": 6039 + }, + { + "epoch": 0.23, + "learning_rate": 4.615927160810464e-05, + "loss": 0.5518, + "step": 6040 + }, + { + "epoch": 0.23, + "learning_rate": 4.6158630418055914e-05, + "loss": 0.5817, + "step": 6041 + }, + { + "epoch": 0.23, + "learning_rate": 4.615798922800718e-05, + "loss": 0.591, + "step": 6042 + }, + { + "epoch": 0.23, + "learning_rate": 4.615734803795845e-05, + "loss": 0.7059, + "step": 6043 + }, + { + "epoch": 0.23, + "learning_rate": 4.615670684790972e-05, + "loss": 0.6569, + "step": 6044 + }, + { + "epoch": 0.23, + "learning_rate": 4.6156065657860994e-05, + "loss": 0.64, + "step": 6045 + }, + { + "epoch": 0.23, + "learning_rate": 4.6155424467812266e-05, + "loss": 0.5862, + "step": 6046 + }, + { + "epoch": 0.23, + "learning_rate": 4.615478327776353e-05, + "loss": 0.5562, + "step": 6047 + }, + { + "epoch": 0.23, + "learning_rate": 4.61541420877148e-05, + "loss": 0.5319, + "step": 6048 + }, + { + "epoch": 0.23, + "learning_rate": 4.615350089766607e-05, + "loss": 0.7351, + "step": 6049 + }, + { + "epoch": 0.23, + "learning_rate": 4.615285970761734e-05, + "loss": 0.5951, + "step": 6050 + }, + { + "epoch": 0.23, + "learning_rate": 4.6152218517568604e-05, + "loss": 0.6389, + "step": 6051 + }, + { + "epoch": 0.23, + "learning_rate": 4.615157732751988e-05, + "loss": 0.5636, + "step": 6052 + }, + { + "epoch": 0.23, + "learning_rate": 4.6150936137471154e-05, + "loss": 0.6227, + "step": 6053 + }, + { + "epoch": 0.23, + "learning_rate": 4.615029494742242e-05, + "loss": 0.5062, + "step": 6054 + }, + { + "epoch": 0.23, + "learning_rate": 4.614965375737369e-05, + "loss": 0.6332, + "step": 6055 + }, + { + "epoch": 0.23, + "learning_rate": 4.6149012567324956e-05, + "loss": 0.6113, + "step": 6056 + }, + { + "epoch": 0.23, + "learning_rate": 4.614837137727623e-05, + "loss": 0.6059, + "step": 6057 + }, + { + "epoch": 0.23, + "learning_rate": 4.614773018722749e-05, + "loss": 0.5658, + "step": 6058 + }, + { + "epoch": 0.23, + "learning_rate": 4.6147088997178764e-05, + "loss": 0.5739, + "step": 6059 + }, + { + "epoch": 0.23, + "learning_rate": 4.6146447807130036e-05, + "loss": 0.5191, + "step": 6060 + }, + { + "epoch": 0.23, + "learning_rate": 4.614580661708131e-05, + "loss": 0.6188, + "step": 6061 + }, + { + "epoch": 0.23, + "learning_rate": 4.614516542703258e-05, + "loss": 0.5981, + "step": 6062 + }, + { + "epoch": 0.23, + "learning_rate": 4.6144524236983844e-05, + "loss": 0.6429, + "step": 6063 + }, + { + "epoch": 0.23, + "learning_rate": 4.6143883046935116e-05, + "loss": 0.5869, + "step": 6064 + }, + { + "epoch": 0.23, + "learning_rate": 4.614324185688638e-05, + "loss": 0.5157, + "step": 6065 + }, + { + "epoch": 0.23, + "learning_rate": 4.614260066683765e-05, + "loss": 0.6358, + "step": 6066 + }, + { + "epoch": 0.23, + "learning_rate": 4.614195947678892e-05, + "loss": 0.6167, + "step": 6067 + }, + { + "epoch": 0.23, + "learning_rate": 4.614131828674019e-05, + "loss": 0.6407, + "step": 6068 + }, + { + "epoch": 0.23, + "learning_rate": 4.614067709669146e-05, + "loss": 0.585, + "step": 6069 + }, + { + "epoch": 0.23, + "learning_rate": 4.614003590664273e-05, + "loss": 0.5906, + "step": 6070 + }, + { + "epoch": 0.23, + "learning_rate": 4.6139394716594005e-05, + "loss": 0.4804, + "step": 6071 + }, + { + "epoch": 0.23, + "learning_rate": 4.613875352654527e-05, + "loss": 0.522, + "step": 6072 + }, + { + "epoch": 0.23, + "learning_rate": 4.613811233649654e-05, + "loss": 0.6403, + "step": 6073 + }, + { + "epoch": 0.23, + "learning_rate": 4.6137471146447806e-05, + "loss": 0.5789, + "step": 6074 + }, + { + "epoch": 0.23, + "learning_rate": 4.613682995639908e-05, + "loss": 0.477, + "step": 6075 + }, + { + "epoch": 0.23, + "learning_rate": 4.613618876635035e-05, + "loss": 0.6475, + "step": 6076 + }, + { + "epoch": 0.23, + "learning_rate": 4.613554757630162e-05, + "loss": 0.5668, + "step": 6077 + }, + { + "epoch": 0.23, + "learning_rate": 4.6134906386252886e-05, + "loss": 0.57, + "step": 6078 + }, + { + "epoch": 0.23, + "learning_rate": 4.613426519620416e-05, + "loss": 0.6119, + "step": 6079 + }, + { + "epoch": 0.23, + "learning_rate": 4.613362400615543e-05, + "loss": 0.4616, + "step": 6080 + }, + { + "epoch": 0.23, + "learning_rate": 4.6132982816106695e-05, + "loss": 0.6168, + "step": 6081 + }, + { + "epoch": 0.23, + "learning_rate": 4.6132341626057966e-05, + "loss": 0.5913, + "step": 6082 + }, + { + "epoch": 0.23, + "learning_rate": 4.613170043600923e-05, + "loss": 0.5433, + "step": 6083 + }, + { + "epoch": 0.23, + "learning_rate": 4.61310592459605e-05, + "loss": 0.6161, + "step": 6084 + }, + { + "epoch": 0.23, + "learning_rate": 4.6130418055911775e-05, + "loss": 0.5519, + "step": 6085 + }, + { + "epoch": 0.23, + "learning_rate": 4.6129776865863046e-05, + "loss": 0.549, + "step": 6086 + }, + { + "epoch": 0.23, + "learning_rate": 4.612913567581431e-05, + "loss": 0.6524, + "step": 6087 + }, + { + "epoch": 0.23, + "learning_rate": 4.612849448576558e-05, + "loss": 0.5937, + "step": 6088 + }, + { + "epoch": 0.23, + "learning_rate": 4.6127853295716855e-05, + "loss": 0.4922, + "step": 6089 + }, + { + "epoch": 0.23, + "learning_rate": 4.612721210566812e-05, + "loss": 0.6147, + "step": 6090 + }, + { + "epoch": 0.23, + "learning_rate": 4.612657091561939e-05, + "loss": 0.6816, + "step": 6091 + }, + { + "epoch": 0.23, + "learning_rate": 4.612592972557066e-05, + "loss": 0.5596, + "step": 6092 + }, + { + "epoch": 0.23, + "learning_rate": 4.6125288535521935e-05, + "loss": 0.5087, + "step": 6093 + }, + { + "epoch": 0.23, + "learning_rate": 4.61246473454732e-05, + "loss": 0.6579, + "step": 6094 + }, + { + "epoch": 0.23, + "learning_rate": 4.612400615542447e-05, + "loss": 0.5068, + "step": 6095 + }, + { + "epoch": 0.23, + "learning_rate": 4.6123364965375736e-05, + "loss": 0.547, + "step": 6096 + }, + { + "epoch": 0.23, + "learning_rate": 4.612272377532701e-05, + "loss": 0.5608, + "step": 6097 + }, + { + "epoch": 0.23, + "learning_rate": 4.612208258527828e-05, + "loss": 0.5428, + "step": 6098 + }, + { + "epoch": 0.23, + "learning_rate": 4.6121441395229545e-05, + "loss": 0.6018, + "step": 6099 + }, + { + "epoch": 0.23, + "learning_rate": 4.6120800205180817e-05, + "loss": 0.5365, + "step": 6100 + }, + { + "epoch": 0.23, + "learning_rate": 4.612015901513209e-05, + "loss": 0.5825, + "step": 6101 + }, + { + "epoch": 0.23, + "learning_rate": 4.611951782508336e-05, + "loss": 0.5596, + "step": 6102 + }, + { + "epoch": 0.23, + "learning_rate": 4.6118876635034625e-05, + "loss": 0.6014, + "step": 6103 + }, + { + "epoch": 0.23, + "learning_rate": 4.61182354449859e-05, + "loss": 0.6752, + "step": 6104 + }, + { + "epoch": 0.23, + "learning_rate": 4.611759425493716e-05, + "loss": 0.6246, + "step": 6105 + }, + { + "epoch": 0.23, + "learning_rate": 4.611695306488843e-05, + "loss": 0.58, + "step": 6106 + }, + { + "epoch": 0.23, + "learning_rate": 4.6116311874839705e-05, + "loss": 0.6046, + "step": 6107 + }, + { + "epoch": 0.23, + "learning_rate": 4.611567068479098e-05, + "loss": 0.5292, + "step": 6108 + }, + { + "epoch": 0.23, + "learning_rate": 4.611502949474225e-05, + "loss": 0.5402, + "step": 6109 + }, + { + "epoch": 0.23, + "learning_rate": 4.6114388304693513e-05, + "loss": 0.6048, + "step": 6110 + }, + { + "epoch": 0.23, + "learning_rate": 4.6113747114644785e-05, + "loss": 0.6445, + "step": 6111 + }, + { + "epoch": 0.23, + "learning_rate": 4.611310592459605e-05, + "loss": 0.4978, + "step": 6112 + }, + { + "epoch": 0.24, + "learning_rate": 4.611246473454732e-05, + "loss": 0.6019, + "step": 6113 + }, + { + "epoch": 0.24, + "learning_rate": 4.611182354449859e-05, + "loss": 0.6249, + "step": 6114 + }, + { + "epoch": 0.24, + "learning_rate": 4.611118235444986e-05, + "loss": 0.5328, + "step": 6115 + }, + { + "epoch": 0.24, + "learning_rate": 4.611054116440113e-05, + "loss": 0.5431, + "step": 6116 + }, + { + "epoch": 0.24, + "learning_rate": 4.61098999743524e-05, + "loss": 0.6271, + "step": 6117 + }, + { + "epoch": 0.24, + "learning_rate": 4.6109258784303674e-05, + "loss": 0.6261, + "step": 6118 + }, + { + "epoch": 0.24, + "learning_rate": 4.610861759425494e-05, + "loss": 0.6652, + "step": 6119 + }, + { + "epoch": 0.24, + "learning_rate": 4.610797640420621e-05, + "loss": 0.6685, + "step": 6120 + }, + { + "epoch": 0.24, + "learning_rate": 4.6107335214157475e-05, + "loss": 0.5423, + "step": 6121 + }, + { + "epoch": 0.24, + "learning_rate": 4.610669402410875e-05, + "loss": 0.6413, + "step": 6122 + }, + { + "epoch": 0.24, + "learning_rate": 4.610605283406001e-05, + "loss": 0.5567, + "step": 6123 + }, + { + "epoch": 0.24, + "learning_rate": 4.610541164401129e-05, + "loss": 0.6245, + "step": 6124 + }, + { + "epoch": 0.24, + "learning_rate": 4.610477045396256e-05, + "loss": 0.4717, + "step": 6125 + }, + { + "epoch": 0.24, + "learning_rate": 4.610412926391383e-05, + "loss": 0.5756, + "step": 6126 + }, + { + "epoch": 0.24, + "learning_rate": 4.61034880738651e-05, + "loss": 0.5149, + "step": 6127 + }, + { + "epoch": 0.24, + "learning_rate": 4.6102846883816364e-05, + "loss": 0.5734, + "step": 6128 + }, + { + "epoch": 0.24, + "learning_rate": 4.6102205693767635e-05, + "loss": 0.5934, + "step": 6129 + }, + { + "epoch": 0.24, + "learning_rate": 4.61015645037189e-05, + "loss": 0.6365, + "step": 6130 + }, + { + "epoch": 0.24, + "learning_rate": 4.610092331367017e-05, + "loss": 0.6593, + "step": 6131 + }, + { + "epoch": 0.24, + "learning_rate": 4.6100282123621444e-05, + "loss": 0.582, + "step": 6132 + }, + { + "epoch": 0.24, + "learning_rate": 4.6099640933572715e-05, + "loss": 0.6276, + "step": 6133 + }, + { + "epoch": 0.24, + "learning_rate": 4.609899974352399e-05, + "loss": 0.603, + "step": 6134 + }, + { + "epoch": 0.24, + "learning_rate": 4.609835855347525e-05, + "loss": 0.5733, + "step": 6135 + }, + { + "epoch": 0.24, + "learning_rate": 4.6097717363426524e-05, + "loss": 0.566, + "step": 6136 + }, + { + "epoch": 0.24, + "learning_rate": 4.609707617337779e-05, + "loss": 0.5222, + "step": 6137 + }, + { + "epoch": 0.24, + "learning_rate": 4.609643498332906e-05, + "loss": 0.5549, + "step": 6138 + }, + { + "epoch": 0.24, + "learning_rate": 4.6095793793280325e-05, + "loss": 0.6373, + "step": 6139 + }, + { + "epoch": 0.24, + "learning_rate": 4.6095152603231604e-05, + "loss": 0.5414, + "step": 6140 + }, + { + "epoch": 0.24, + "learning_rate": 4.609451141318287e-05, + "loss": 0.6174, + "step": 6141 + }, + { + "epoch": 0.24, + "learning_rate": 4.609387022313414e-05, + "loss": 0.578, + "step": 6142 + }, + { + "epoch": 0.24, + "learning_rate": 4.609322903308541e-05, + "loss": 0.6216, + "step": 6143 + }, + { + "epoch": 0.24, + "learning_rate": 4.609258784303668e-05, + "loss": 0.5332, + "step": 6144 + }, + { + "epoch": 0.24, + "learning_rate": 4.609194665298795e-05, + "loss": 0.581, + "step": 6145 + }, + { + "epoch": 0.24, + "learning_rate": 4.6091305462939214e-05, + "loss": 0.591, + "step": 6146 + }, + { + "epoch": 0.24, + "learning_rate": 4.6090664272890486e-05, + "loss": 0.5505, + "step": 6147 + }, + { + "epoch": 0.24, + "learning_rate": 4.609002308284176e-05, + "loss": 0.5069, + "step": 6148 + }, + { + "epoch": 0.24, + "learning_rate": 4.608938189279303e-05, + "loss": 0.6147, + "step": 6149 + }, + { + "epoch": 0.24, + "learning_rate": 4.6088740702744294e-05, + "loss": 0.606, + "step": 6150 + }, + { + "epoch": 0.24, + "learning_rate": 4.6088099512695566e-05, + "loss": 0.4822, + "step": 6151 + }, + { + "epoch": 0.24, + "learning_rate": 4.608745832264684e-05, + "loss": 0.575, + "step": 6152 + }, + { + "epoch": 0.24, + "learning_rate": 4.60868171325981e-05, + "loss": 0.5584, + "step": 6153 + }, + { + "epoch": 0.24, + "learning_rate": 4.6086175942549374e-05, + "loss": 0.6322, + "step": 6154 + }, + { + "epoch": 0.24, + "learning_rate": 4.608553475250064e-05, + "loss": 0.4814, + "step": 6155 + }, + { + "epoch": 0.24, + "learning_rate": 4.608489356245191e-05, + "loss": 0.6738, + "step": 6156 + }, + { + "epoch": 0.24, + "learning_rate": 4.608425237240318e-05, + "loss": 0.6168, + "step": 6157 + }, + { + "epoch": 0.24, + "learning_rate": 4.6083611182354454e-05, + "loss": 0.6255, + "step": 6158 + }, + { + "epoch": 0.24, + "learning_rate": 4.608296999230572e-05, + "loss": 0.5121, + "step": 6159 + }, + { + "epoch": 0.24, + "learning_rate": 4.608232880225699e-05, + "loss": 0.5655, + "step": 6160 + }, + { + "epoch": 0.24, + "learning_rate": 4.608168761220826e-05, + "loss": 0.6404, + "step": 6161 + }, + { + "epoch": 0.24, + "learning_rate": 4.608104642215953e-05, + "loss": 0.5396, + "step": 6162 + }, + { + "epoch": 0.24, + "learning_rate": 4.60804052321108e-05, + "loss": 0.626, + "step": 6163 + }, + { + "epoch": 0.24, + "learning_rate": 4.607976404206207e-05, + "loss": 0.6246, + "step": 6164 + }, + { + "epoch": 0.24, + "learning_rate": 4.607912285201334e-05, + "loss": 0.6368, + "step": 6165 + }, + { + "epoch": 0.24, + "learning_rate": 4.607848166196461e-05, + "loss": 0.5394, + "step": 6166 + }, + { + "epoch": 0.24, + "learning_rate": 4.607784047191588e-05, + "loss": 0.5953, + "step": 6167 + }, + { + "epoch": 0.24, + "learning_rate": 4.6077199281867144e-05, + "loss": 0.6012, + "step": 6168 + }, + { + "epoch": 0.24, + "learning_rate": 4.6076558091818416e-05, + "loss": 0.5205, + "step": 6169 + }, + { + "epoch": 0.24, + "learning_rate": 4.607591690176969e-05, + "loss": 0.4994, + "step": 6170 + }, + { + "epoch": 0.24, + "learning_rate": 4.607527571172095e-05, + "loss": 0.6054, + "step": 6171 + }, + { + "epoch": 0.24, + "learning_rate": 4.6074634521672224e-05, + "loss": 0.6038, + "step": 6172 + }, + { + "epoch": 0.24, + "learning_rate": 4.6073993331623496e-05, + "loss": 0.7019, + "step": 6173 + }, + { + "epoch": 0.24, + "learning_rate": 4.607335214157477e-05, + "loss": 0.6601, + "step": 6174 + }, + { + "epoch": 0.24, + "learning_rate": 4.607271095152603e-05, + "loss": 0.5021, + "step": 6175 + }, + { + "epoch": 0.24, + "learning_rate": 4.6072069761477304e-05, + "loss": 0.5935, + "step": 6176 + }, + { + "epoch": 0.24, + "learning_rate": 4.607142857142857e-05, + "loss": 0.4836, + "step": 6177 + }, + { + "epoch": 0.24, + "learning_rate": 4.607078738137984e-05, + "loss": 0.6146, + "step": 6178 + }, + { + "epoch": 0.24, + "learning_rate": 4.607014619133111e-05, + "loss": 0.6164, + "step": 6179 + }, + { + "epoch": 0.24, + "learning_rate": 4.6069505001282384e-05, + "loss": 0.5735, + "step": 6180 + }, + { + "epoch": 0.24, + "learning_rate": 4.6068863811233656e-05, + "loss": 0.5589, + "step": 6181 + }, + { + "epoch": 0.24, + "learning_rate": 4.606822262118492e-05, + "loss": 0.7249, + "step": 6182 + }, + { + "epoch": 0.24, + "learning_rate": 4.606758143113619e-05, + "loss": 0.6363, + "step": 6183 + }, + { + "epoch": 0.24, + "learning_rate": 4.606694024108746e-05, + "loss": 0.5387, + "step": 6184 + }, + { + "epoch": 0.24, + "learning_rate": 4.606629905103873e-05, + "loss": 0.5789, + "step": 6185 + }, + { + "epoch": 0.24, + "learning_rate": 4.6065657860989994e-05, + "loss": 0.5432, + "step": 6186 + }, + { + "epoch": 0.24, + "learning_rate": 4.6065016670941266e-05, + "loss": 0.5301, + "step": 6187 + }, + { + "epoch": 0.24, + "learning_rate": 4.606437548089254e-05, + "loss": 0.5345, + "step": 6188 + }, + { + "epoch": 0.24, + "learning_rate": 4.606373429084381e-05, + "loss": 0.6192, + "step": 6189 + }, + { + "epoch": 0.24, + "learning_rate": 4.606309310079508e-05, + "loss": 0.5586, + "step": 6190 + }, + { + "epoch": 0.24, + "learning_rate": 4.6062451910746346e-05, + "loss": 0.5051, + "step": 6191 + }, + { + "epoch": 0.24, + "learning_rate": 4.606181072069762e-05, + "loss": 0.5642, + "step": 6192 + }, + { + "epoch": 0.24, + "learning_rate": 4.606116953064888e-05, + "loss": 0.5772, + "step": 6193 + }, + { + "epoch": 0.24, + "learning_rate": 4.6060528340600155e-05, + "loss": 0.5757, + "step": 6194 + }, + { + "epoch": 0.24, + "learning_rate": 4.605988715055142e-05, + "loss": 0.5657, + "step": 6195 + }, + { + "epoch": 0.24, + "learning_rate": 4.60592459605027e-05, + "loss": 0.7057, + "step": 6196 + }, + { + "epoch": 0.24, + "learning_rate": 4.605860477045397e-05, + "loss": 0.5289, + "step": 6197 + }, + { + "epoch": 0.24, + "learning_rate": 4.6057963580405235e-05, + "loss": 0.4925, + "step": 6198 + }, + { + "epoch": 0.24, + "learning_rate": 4.6057322390356506e-05, + "loss": 0.4999, + "step": 6199 + }, + { + "epoch": 0.24, + "learning_rate": 4.605668120030777e-05, + "loss": 0.6793, + "step": 6200 + }, + { + "epoch": 0.24, + "learning_rate": 4.605604001025904e-05, + "loss": 0.5546, + "step": 6201 + }, + { + "epoch": 0.24, + "learning_rate": 4.605539882021031e-05, + "loss": 0.5522, + "step": 6202 + }, + { + "epoch": 0.24, + "learning_rate": 4.605475763016158e-05, + "loss": 0.5371, + "step": 6203 + }, + { + "epoch": 0.24, + "learning_rate": 4.605411644011285e-05, + "loss": 0.5577, + "step": 6204 + }, + { + "epoch": 0.24, + "learning_rate": 4.605347525006412e-05, + "loss": 0.591, + "step": 6205 + }, + { + "epoch": 0.24, + "learning_rate": 4.6052834060015395e-05, + "loss": 0.5749, + "step": 6206 + }, + { + "epoch": 0.24, + "learning_rate": 4.605219286996666e-05, + "loss": 0.6605, + "step": 6207 + }, + { + "epoch": 0.24, + "learning_rate": 4.605155167991793e-05, + "loss": 0.6135, + "step": 6208 + }, + { + "epoch": 0.24, + "learning_rate": 4.6050910489869196e-05, + "loss": 0.5394, + "step": 6209 + }, + { + "epoch": 0.24, + "learning_rate": 4.605026929982047e-05, + "loss": 0.5066, + "step": 6210 + }, + { + "epoch": 0.24, + "learning_rate": 4.604962810977173e-05, + "loss": 0.6791, + "step": 6211 + }, + { + "epoch": 0.24, + "learning_rate": 4.604898691972301e-05, + "loss": 0.6146, + "step": 6212 + }, + { + "epoch": 0.24, + "learning_rate": 4.604834572967428e-05, + "loss": 0.5194, + "step": 6213 + }, + { + "epoch": 0.24, + "learning_rate": 4.604770453962555e-05, + "loss": 0.5181, + "step": 6214 + }, + { + "epoch": 0.24, + "learning_rate": 4.604706334957682e-05, + "loss": 0.5287, + "step": 6215 + }, + { + "epoch": 0.24, + "learning_rate": 4.6046422159528085e-05, + "loss": 0.5754, + "step": 6216 + }, + { + "epoch": 0.24, + "learning_rate": 4.604578096947936e-05, + "loss": 0.6055, + "step": 6217 + }, + { + "epoch": 0.24, + "learning_rate": 4.604513977943062e-05, + "loss": 0.5094, + "step": 6218 + }, + { + "epoch": 0.24, + "learning_rate": 4.604449858938189e-05, + "loss": 0.6622, + "step": 6219 + }, + { + "epoch": 0.24, + "learning_rate": 4.6043857399333165e-05, + "loss": 0.5579, + "step": 6220 + }, + { + "epoch": 0.24, + "learning_rate": 4.604321620928444e-05, + "loss": 0.6123, + "step": 6221 + }, + { + "epoch": 0.24, + "learning_rate": 4.604257501923571e-05, + "loss": 0.534, + "step": 6222 + }, + { + "epoch": 0.24, + "learning_rate": 4.6041933829186973e-05, + "loss": 0.6074, + "step": 6223 + }, + { + "epoch": 0.24, + "learning_rate": 4.6041292639138245e-05, + "loss": 0.67, + "step": 6224 + }, + { + "epoch": 0.24, + "learning_rate": 4.604065144908951e-05, + "loss": 0.5935, + "step": 6225 + }, + { + "epoch": 0.24, + "learning_rate": 4.604001025904078e-05, + "loss": 0.5572, + "step": 6226 + }, + { + "epoch": 0.24, + "learning_rate": 4.603936906899205e-05, + "loss": 0.653, + "step": 6227 + }, + { + "epoch": 0.24, + "learning_rate": 4.603872787894332e-05, + "loss": 0.5891, + "step": 6228 + }, + { + "epoch": 0.24, + "learning_rate": 4.603808668889459e-05, + "loss": 0.5647, + "step": 6229 + }, + { + "epoch": 0.24, + "learning_rate": 4.603744549884586e-05, + "loss": 0.5638, + "step": 6230 + }, + { + "epoch": 0.24, + "learning_rate": 4.6036804308797134e-05, + "loss": 0.5488, + "step": 6231 + }, + { + "epoch": 0.24, + "learning_rate": 4.60361631187484e-05, + "loss": 0.6288, + "step": 6232 + }, + { + "epoch": 0.24, + "learning_rate": 4.603552192869967e-05, + "loss": 0.5457, + "step": 6233 + }, + { + "epoch": 0.24, + "learning_rate": 4.6034880738650935e-05, + "loss": 0.5186, + "step": 6234 + }, + { + "epoch": 0.24, + "learning_rate": 4.603423954860221e-05, + "loss": 0.5557, + "step": 6235 + }, + { + "epoch": 0.24, + "learning_rate": 4.603359835855348e-05, + "loss": 0.6138, + "step": 6236 + }, + { + "epoch": 0.24, + "learning_rate": 4.603295716850475e-05, + "loss": 0.638, + "step": 6237 + }, + { + "epoch": 0.24, + "learning_rate": 4.6032315978456015e-05, + "loss": 0.6018, + "step": 6238 + }, + { + "epoch": 0.24, + "learning_rate": 4.603167478840729e-05, + "loss": 0.704, + "step": 6239 + }, + { + "epoch": 0.24, + "learning_rate": 4.603103359835856e-05, + "loss": 0.6363, + "step": 6240 + }, + { + "epoch": 0.24, + "learning_rate": 4.6030392408309824e-05, + "loss": 0.6208, + "step": 6241 + }, + { + "epoch": 0.24, + "learning_rate": 4.6029751218261095e-05, + "loss": 0.54, + "step": 6242 + }, + { + "epoch": 0.24, + "learning_rate": 4.602911002821236e-05, + "loss": 0.792, + "step": 6243 + }, + { + "epoch": 0.24, + "learning_rate": 4.602846883816363e-05, + "loss": 0.5381, + "step": 6244 + }, + { + "epoch": 0.24, + "learning_rate": 4.6027827648114904e-05, + "loss": 0.5377, + "step": 6245 + }, + { + "epoch": 0.24, + "learning_rate": 4.6027186458066175e-05, + "loss": 0.6523, + "step": 6246 + }, + { + "epoch": 0.24, + "learning_rate": 4.602654526801744e-05, + "loss": 0.6986, + "step": 6247 + }, + { + "epoch": 0.24, + "learning_rate": 4.602590407796871e-05, + "loss": 0.5532, + "step": 6248 + }, + { + "epoch": 0.24, + "learning_rate": 4.6025262887919984e-05, + "loss": 0.5568, + "step": 6249 + }, + { + "epoch": 0.24, + "learning_rate": 4.602462169787125e-05, + "loss": 0.5819, + "step": 6250 + }, + { + "epoch": 0.24, + "learning_rate": 4.602398050782252e-05, + "loss": 0.6041, + "step": 6251 + }, + { + "epoch": 0.24, + "learning_rate": 4.602333931777379e-05, + "loss": 0.6746, + "step": 6252 + }, + { + "epoch": 0.24, + "learning_rate": 4.6022698127725064e-05, + "loss": 0.6277, + "step": 6253 + }, + { + "epoch": 0.24, + "learning_rate": 4.602205693767633e-05, + "loss": 0.5177, + "step": 6254 + }, + { + "epoch": 0.24, + "learning_rate": 4.60214157476276e-05, + "loss": 0.5763, + "step": 6255 + }, + { + "epoch": 0.24, + "learning_rate": 4.6020774557578866e-05, + "loss": 0.5992, + "step": 6256 + }, + { + "epoch": 0.24, + "learning_rate": 4.602013336753014e-05, + "loss": 0.6385, + "step": 6257 + }, + { + "epoch": 0.24, + "learning_rate": 4.601949217748141e-05, + "loss": 0.551, + "step": 6258 + }, + { + "epoch": 0.24, + "learning_rate": 4.6018850987432674e-05, + "loss": 0.6754, + "step": 6259 + }, + { + "epoch": 0.24, + "learning_rate": 4.6018209797383946e-05, + "loss": 0.6148, + "step": 6260 + }, + { + "epoch": 0.24, + "learning_rate": 4.601756860733522e-05, + "loss": 0.5925, + "step": 6261 + }, + { + "epoch": 0.24, + "learning_rate": 4.601692741728649e-05, + "loss": 0.5378, + "step": 6262 + }, + { + "epoch": 0.24, + "learning_rate": 4.6016286227237754e-05, + "loss": 0.5641, + "step": 6263 + }, + { + "epoch": 0.24, + "learning_rate": 4.6015645037189026e-05, + "loss": 0.4311, + "step": 6264 + }, + { + "epoch": 0.24, + "learning_rate": 4.601500384714029e-05, + "loss": 0.5801, + "step": 6265 + }, + { + "epoch": 0.24, + "learning_rate": 4.601436265709156e-05, + "loss": 0.6468, + "step": 6266 + }, + { + "epoch": 0.24, + "learning_rate": 4.6013721467042834e-05, + "loss": 0.585, + "step": 6267 + }, + { + "epoch": 0.24, + "learning_rate": 4.6013080276994106e-05, + "loss": 0.6846, + "step": 6268 + }, + { + "epoch": 0.24, + "learning_rate": 4.601243908694538e-05, + "loss": 0.6196, + "step": 6269 + }, + { + "epoch": 0.24, + "learning_rate": 4.601179789689664e-05, + "loss": 0.5874, + "step": 6270 + }, + { + "epoch": 0.24, + "learning_rate": 4.6011156706847914e-05, + "loss": 0.5378, + "step": 6271 + }, + { + "epoch": 0.24, + "learning_rate": 4.601051551679918e-05, + "loss": 0.5975, + "step": 6272 + }, + { + "epoch": 0.24, + "learning_rate": 4.600987432675045e-05, + "loss": 0.5796, + "step": 6273 + }, + { + "epoch": 0.24, + "learning_rate": 4.6009233136701716e-05, + "loss": 0.5805, + "step": 6274 + }, + { + "epoch": 0.24, + "learning_rate": 4.600859194665299e-05, + "loss": 0.6439, + "step": 6275 + }, + { + "epoch": 0.24, + "learning_rate": 4.600795075660426e-05, + "loss": 0.619, + "step": 6276 + }, + { + "epoch": 0.24, + "learning_rate": 4.600730956655553e-05, + "loss": 0.4708, + "step": 6277 + }, + { + "epoch": 0.24, + "learning_rate": 4.60066683765068e-05, + "loss": 0.5633, + "step": 6278 + }, + { + "epoch": 0.24, + "learning_rate": 4.600602718645807e-05, + "loss": 0.4858, + "step": 6279 + }, + { + "epoch": 0.24, + "learning_rate": 4.600538599640934e-05, + "loss": 0.5028, + "step": 6280 + }, + { + "epoch": 0.24, + "learning_rate": 4.6004744806360604e-05, + "loss": 0.559, + "step": 6281 + }, + { + "epoch": 0.24, + "learning_rate": 4.6004103616311876e-05, + "loss": 0.5585, + "step": 6282 + }, + { + "epoch": 0.24, + "learning_rate": 4.600346242626314e-05, + "loss": 0.6232, + "step": 6283 + }, + { + "epoch": 0.24, + "learning_rate": 4.600282123621442e-05, + "loss": 0.6454, + "step": 6284 + }, + { + "epoch": 0.24, + "learning_rate": 4.600218004616569e-05, + "loss": 0.5029, + "step": 6285 + }, + { + "epoch": 0.24, + "learning_rate": 4.6001538856116956e-05, + "loss": 0.5314, + "step": 6286 + }, + { + "epoch": 0.24, + "learning_rate": 4.600089766606823e-05, + "loss": 0.6847, + "step": 6287 + }, + { + "epoch": 0.24, + "learning_rate": 4.600025647601949e-05, + "loss": 0.5464, + "step": 6288 + }, + { + "epoch": 0.24, + "learning_rate": 4.5999615285970764e-05, + "loss": 0.5775, + "step": 6289 + }, + { + "epoch": 0.24, + "learning_rate": 4.599897409592203e-05, + "loss": 0.6973, + "step": 6290 + }, + { + "epoch": 0.24, + "learning_rate": 4.59983329058733e-05, + "loss": 0.4945, + "step": 6291 + }, + { + "epoch": 0.24, + "learning_rate": 4.599769171582457e-05, + "loss": 0.6407, + "step": 6292 + }, + { + "epoch": 0.24, + "learning_rate": 4.5997050525775845e-05, + "loss": 0.6703, + "step": 6293 + }, + { + "epoch": 0.24, + "learning_rate": 4.5996409335727116e-05, + "loss": 0.5926, + "step": 6294 + }, + { + "epoch": 0.24, + "learning_rate": 4.599576814567838e-05, + "loss": 0.5232, + "step": 6295 + }, + { + "epoch": 0.24, + "learning_rate": 4.599512695562965e-05, + "loss": 0.5571, + "step": 6296 + }, + { + "epoch": 0.24, + "learning_rate": 4.599448576558092e-05, + "loss": 0.4914, + "step": 6297 + }, + { + "epoch": 0.24, + "learning_rate": 4.599384457553219e-05, + "loss": 0.595, + "step": 6298 + }, + { + "epoch": 0.24, + "learning_rate": 4.5993203385483454e-05, + "loss": 0.5817, + "step": 6299 + }, + { + "epoch": 0.24, + "learning_rate": 4.599256219543473e-05, + "loss": 0.5784, + "step": 6300 + }, + { + "epoch": 0.24, + "learning_rate": 4.5991921005386e-05, + "loss": 0.5444, + "step": 6301 + }, + { + "epoch": 0.24, + "learning_rate": 4.599127981533727e-05, + "loss": 0.629, + "step": 6302 + }, + { + "epoch": 0.24, + "learning_rate": 4.599063862528854e-05, + "loss": 0.6729, + "step": 6303 + }, + { + "epoch": 0.24, + "learning_rate": 4.5989997435239806e-05, + "loss": 0.6282, + "step": 6304 + }, + { + "epoch": 0.24, + "learning_rate": 4.598935624519108e-05, + "loss": 0.5706, + "step": 6305 + }, + { + "epoch": 0.24, + "learning_rate": 4.598871505514234e-05, + "loss": 0.5921, + "step": 6306 + }, + { + "epoch": 0.24, + "learning_rate": 4.5988073865093615e-05, + "loss": 0.5986, + "step": 6307 + }, + { + "epoch": 0.24, + "learning_rate": 4.5987432675044886e-05, + "loss": 0.5852, + "step": 6308 + }, + { + "epoch": 0.24, + "learning_rate": 4.598679148499616e-05, + "loss": 0.5939, + "step": 6309 + }, + { + "epoch": 0.24, + "learning_rate": 4.598615029494742e-05, + "loss": 0.6385, + "step": 6310 + }, + { + "epoch": 0.24, + "learning_rate": 4.5985509104898695e-05, + "loss": 0.6158, + "step": 6311 + }, + { + "epoch": 0.24, + "learning_rate": 4.5984867914849966e-05, + "loss": 0.7266, + "step": 6312 + }, + { + "epoch": 0.24, + "learning_rate": 4.598422672480123e-05, + "loss": 0.6473, + "step": 6313 + }, + { + "epoch": 0.24, + "learning_rate": 4.59835855347525e-05, + "loss": 0.5546, + "step": 6314 + }, + { + "epoch": 0.24, + "learning_rate": 4.598294434470377e-05, + "loss": 0.6311, + "step": 6315 + }, + { + "epoch": 0.24, + "learning_rate": 4.598230315465504e-05, + "loss": 0.5478, + "step": 6316 + }, + { + "epoch": 0.24, + "learning_rate": 4.598166196460631e-05, + "loss": 0.5656, + "step": 6317 + }, + { + "epoch": 0.24, + "learning_rate": 4.598102077455758e-05, + "loss": 0.5409, + "step": 6318 + }, + { + "epoch": 0.24, + "learning_rate": 4.598037958450885e-05, + "loss": 0.5995, + "step": 6319 + }, + { + "epoch": 0.24, + "learning_rate": 4.597973839446012e-05, + "loss": 0.5936, + "step": 6320 + }, + { + "epoch": 0.24, + "learning_rate": 4.597909720441139e-05, + "loss": 0.5803, + "step": 6321 + }, + { + "epoch": 0.24, + "learning_rate": 4.5978456014362657e-05, + "loss": 0.6117, + "step": 6322 + }, + { + "epoch": 0.24, + "learning_rate": 4.597781482431393e-05, + "loss": 0.6598, + "step": 6323 + }, + { + "epoch": 0.24, + "learning_rate": 4.59771736342652e-05, + "loss": 0.5779, + "step": 6324 + }, + { + "epoch": 0.24, + "learning_rate": 4.597653244421647e-05, + "loss": 0.576, + "step": 6325 + }, + { + "epoch": 0.24, + "learning_rate": 4.597589125416774e-05, + "loss": 0.5318, + "step": 6326 + }, + { + "epoch": 0.24, + "learning_rate": 4.597525006411901e-05, + "loss": 0.5791, + "step": 6327 + }, + { + "epoch": 0.24, + "learning_rate": 4.597460887407027e-05, + "loss": 0.5312, + "step": 6328 + }, + { + "epoch": 0.24, + "learning_rate": 4.5973967684021545e-05, + "loss": 0.584, + "step": 6329 + }, + { + "epoch": 0.24, + "learning_rate": 4.597332649397282e-05, + "loss": 0.5315, + "step": 6330 + }, + { + "epoch": 0.24, + "learning_rate": 4.597268530392408e-05, + "loss": 0.6176, + "step": 6331 + }, + { + "epoch": 0.24, + "learning_rate": 4.597204411387535e-05, + "loss": 0.5575, + "step": 6332 + }, + { + "epoch": 0.24, + "learning_rate": 4.5971402923826625e-05, + "loss": 0.554, + "step": 6333 + }, + { + "epoch": 0.24, + "learning_rate": 4.59707617337779e-05, + "loss": 0.6603, + "step": 6334 + }, + { + "epoch": 0.24, + "learning_rate": 4.597012054372916e-05, + "loss": 0.5841, + "step": 6335 + }, + { + "epoch": 0.24, + "learning_rate": 4.5969479353680433e-05, + "loss": 0.5684, + "step": 6336 + }, + { + "epoch": 0.24, + "learning_rate": 4.59688381636317e-05, + "loss": 0.5651, + "step": 6337 + }, + { + "epoch": 0.24, + "learning_rate": 4.596819697358297e-05, + "loss": 0.542, + "step": 6338 + }, + { + "epoch": 0.24, + "learning_rate": 4.596755578353424e-05, + "loss": 0.6341, + "step": 6339 + }, + { + "epoch": 0.24, + "learning_rate": 4.5966914593485514e-05, + "loss": 0.5364, + "step": 6340 + }, + { + "epoch": 0.24, + "learning_rate": 4.5966273403436785e-05, + "loss": 0.5162, + "step": 6341 + }, + { + "epoch": 0.24, + "learning_rate": 4.596563221338805e-05, + "loss": 0.5122, + "step": 6342 + }, + { + "epoch": 0.24, + "learning_rate": 4.596499102333932e-05, + "loss": 0.6079, + "step": 6343 + }, + { + "epoch": 0.24, + "learning_rate": 4.596434983329059e-05, + "loss": 0.6349, + "step": 6344 + }, + { + "epoch": 0.24, + "learning_rate": 4.596370864324186e-05, + "loss": 0.5116, + "step": 6345 + }, + { + "epoch": 0.24, + "learning_rate": 4.5963067453193124e-05, + "loss": 0.6316, + "step": 6346 + }, + { + "epoch": 0.24, + "learning_rate": 4.5962426263144395e-05, + "loss": 0.6367, + "step": 6347 + }, + { + "epoch": 0.24, + "learning_rate": 4.596178507309567e-05, + "loss": 0.6036, + "step": 6348 + }, + { + "epoch": 0.24, + "learning_rate": 4.596114388304694e-05, + "loss": 0.5808, + "step": 6349 + }, + { + "epoch": 0.24, + "learning_rate": 4.596050269299821e-05, + "loss": 0.6588, + "step": 6350 + }, + { + "epoch": 0.24, + "learning_rate": 4.5959861502949475e-05, + "loss": 0.6268, + "step": 6351 + }, + { + "epoch": 0.24, + "learning_rate": 4.595922031290075e-05, + "loss": 0.5382, + "step": 6352 + }, + { + "epoch": 0.24, + "learning_rate": 4.595857912285201e-05, + "loss": 0.6129, + "step": 6353 + }, + { + "epoch": 0.24, + "learning_rate": 4.5957937932803284e-05, + "loss": 0.5343, + "step": 6354 + }, + { + "epoch": 0.24, + "learning_rate": 4.595729674275455e-05, + "loss": 0.6501, + "step": 6355 + }, + { + "epoch": 0.24, + "learning_rate": 4.595665555270583e-05, + "loss": 0.5372, + "step": 6356 + }, + { + "epoch": 0.24, + "learning_rate": 4.59560143626571e-05, + "loss": 0.5134, + "step": 6357 + }, + { + "epoch": 0.24, + "learning_rate": 4.5955373172608364e-05, + "loss": 0.5991, + "step": 6358 + }, + { + "epoch": 0.24, + "learning_rate": 4.5954731982559636e-05, + "loss": 0.5824, + "step": 6359 + }, + { + "epoch": 0.24, + "learning_rate": 4.59540907925109e-05, + "loss": 0.6238, + "step": 6360 + }, + { + "epoch": 0.24, + "learning_rate": 4.595344960246217e-05, + "loss": 0.5708, + "step": 6361 + }, + { + "epoch": 0.24, + "learning_rate": 4.595280841241344e-05, + "loss": 0.5599, + "step": 6362 + }, + { + "epoch": 0.24, + "learning_rate": 4.595216722236471e-05, + "loss": 0.525, + "step": 6363 + }, + { + "epoch": 0.24, + "learning_rate": 4.595152603231598e-05, + "loss": 0.5772, + "step": 6364 + }, + { + "epoch": 0.24, + "learning_rate": 4.595088484226725e-05, + "loss": 0.5347, + "step": 6365 + }, + { + "epoch": 0.24, + "learning_rate": 4.5950243652218524e-05, + "loss": 0.7044, + "step": 6366 + }, + { + "epoch": 0.24, + "learning_rate": 4.594960246216979e-05, + "loss": 0.4502, + "step": 6367 + }, + { + "epoch": 0.24, + "learning_rate": 4.594896127212106e-05, + "loss": 0.6131, + "step": 6368 + }, + { + "epoch": 0.24, + "learning_rate": 4.5948320082072326e-05, + "loss": 0.5537, + "step": 6369 + }, + { + "epoch": 0.24, + "learning_rate": 4.59476788920236e-05, + "loss": 0.6036, + "step": 6370 + }, + { + "epoch": 0.24, + "learning_rate": 4.594703770197486e-05, + "loss": 0.6596, + "step": 6371 + }, + { + "epoch": 0.24, + "learning_rate": 4.594639651192614e-05, + "loss": 0.5488, + "step": 6372 + }, + { + "epoch": 0.25, + "learning_rate": 4.5945755321877406e-05, + "loss": 0.6098, + "step": 6373 + }, + { + "epoch": 0.25, + "learning_rate": 4.594511413182868e-05, + "loss": 0.6044, + "step": 6374 + }, + { + "epoch": 0.25, + "learning_rate": 4.594447294177995e-05, + "loss": 0.6423, + "step": 6375 + }, + { + "epoch": 0.25, + "learning_rate": 4.5943831751731214e-05, + "loss": 0.5949, + "step": 6376 + }, + { + "epoch": 0.25, + "learning_rate": 4.5943190561682486e-05, + "loss": 0.5951, + "step": 6377 + }, + { + "epoch": 0.25, + "learning_rate": 4.594254937163375e-05, + "loss": 0.6238, + "step": 6378 + }, + { + "epoch": 0.25, + "learning_rate": 4.594190818158502e-05, + "loss": 0.5847, + "step": 6379 + }, + { + "epoch": 0.25, + "learning_rate": 4.5941266991536294e-05, + "loss": 0.6243, + "step": 6380 + }, + { + "epoch": 0.25, + "learning_rate": 4.5940625801487566e-05, + "loss": 0.6115, + "step": 6381 + }, + { + "epoch": 0.25, + "learning_rate": 4.593998461143883e-05, + "loss": 0.6234, + "step": 6382 + }, + { + "epoch": 0.25, + "learning_rate": 4.59393434213901e-05, + "loss": 0.6111, + "step": 6383 + }, + { + "epoch": 0.25, + "learning_rate": 4.5938702231341374e-05, + "loss": 0.6136, + "step": 6384 + }, + { + "epoch": 0.25, + "learning_rate": 4.593806104129264e-05, + "loss": 0.5509, + "step": 6385 + }, + { + "epoch": 0.25, + "learning_rate": 4.593741985124391e-05, + "loss": 0.5956, + "step": 6386 + }, + { + "epoch": 0.25, + "learning_rate": 4.5936778661195176e-05, + "loss": 0.6449, + "step": 6387 + }, + { + "epoch": 0.25, + "learning_rate": 4.5936137471146454e-05, + "loss": 0.6022, + "step": 6388 + }, + { + "epoch": 0.25, + "learning_rate": 4.593549628109772e-05, + "loss": 0.5207, + "step": 6389 + }, + { + "epoch": 0.25, + "learning_rate": 4.593485509104899e-05, + "loss": 0.5693, + "step": 6390 + }, + { + "epoch": 0.25, + "learning_rate": 4.593421390100026e-05, + "loss": 0.5047, + "step": 6391 + }, + { + "epoch": 0.25, + "learning_rate": 4.593357271095153e-05, + "loss": 0.625, + "step": 6392 + }, + { + "epoch": 0.25, + "learning_rate": 4.59329315209028e-05, + "loss": 0.5473, + "step": 6393 + }, + { + "epoch": 0.25, + "learning_rate": 4.5932290330854064e-05, + "loss": 0.608, + "step": 6394 + }, + { + "epoch": 0.25, + "learning_rate": 4.5931649140805336e-05, + "loss": 0.637, + "step": 6395 + }, + { + "epoch": 0.25, + "learning_rate": 4.593100795075661e-05, + "loss": 0.7163, + "step": 6396 + }, + { + "epoch": 0.25, + "learning_rate": 4.593036676070788e-05, + "loss": 0.5941, + "step": 6397 + }, + { + "epoch": 0.25, + "learning_rate": 4.5929725570659144e-05, + "loss": 0.5261, + "step": 6398 + }, + { + "epoch": 0.25, + "learning_rate": 4.5929084380610416e-05, + "loss": 0.5011, + "step": 6399 + }, + { + "epoch": 0.25, + "learning_rate": 4.592844319056169e-05, + "loss": 0.577, + "step": 6400 + }, + { + "epoch": 0.25, + "learning_rate": 4.592780200051295e-05, + "loss": 0.6783, + "step": 6401 + }, + { + "epoch": 0.25, + "learning_rate": 4.5927160810464224e-05, + "loss": 0.5689, + "step": 6402 + }, + { + "epoch": 0.25, + "learning_rate": 4.592651962041549e-05, + "loss": 0.6306, + "step": 6403 + }, + { + "epoch": 0.25, + "learning_rate": 4.592587843036676e-05, + "loss": 0.5784, + "step": 6404 + }, + { + "epoch": 0.25, + "learning_rate": 4.592523724031803e-05, + "loss": 0.6583, + "step": 6405 + }, + { + "epoch": 0.25, + "learning_rate": 4.5924596050269305e-05, + "loss": 0.568, + "step": 6406 + }, + { + "epoch": 0.25, + "learning_rate": 4.592395486022057e-05, + "loss": 0.5712, + "step": 6407 + }, + { + "epoch": 0.25, + "learning_rate": 4.592331367017184e-05, + "loss": 0.5974, + "step": 6408 + }, + { + "epoch": 0.25, + "learning_rate": 4.592267248012311e-05, + "loss": 0.5873, + "step": 6409 + }, + { + "epoch": 0.25, + "learning_rate": 4.592203129007438e-05, + "loss": 0.629, + "step": 6410 + }, + { + "epoch": 0.25, + "learning_rate": 4.592139010002565e-05, + "loss": 0.5996, + "step": 6411 + }, + { + "epoch": 0.25, + "learning_rate": 4.592074890997692e-05, + "loss": 0.6852, + "step": 6412 + }, + { + "epoch": 0.25, + "learning_rate": 4.592010771992819e-05, + "loss": 0.6338, + "step": 6413 + }, + { + "epoch": 0.25, + "learning_rate": 4.591946652987946e-05, + "loss": 0.5276, + "step": 6414 + }, + { + "epoch": 0.25, + "learning_rate": 4.591882533983073e-05, + "loss": 0.5289, + "step": 6415 + }, + { + "epoch": 0.25, + "learning_rate": 4.5918184149781995e-05, + "loss": 0.548, + "step": 6416 + }, + { + "epoch": 0.25, + "learning_rate": 4.5917542959733266e-05, + "loss": 0.5436, + "step": 6417 + }, + { + "epoch": 0.25, + "learning_rate": 4.591690176968454e-05, + "loss": 0.5285, + "step": 6418 + }, + { + "epoch": 0.25, + "learning_rate": 4.59162605796358e-05, + "loss": 0.5589, + "step": 6419 + }, + { + "epoch": 0.25, + "learning_rate": 4.5915619389587075e-05, + "loss": 0.5051, + "step": 6420 + }, + { + "epoch": 0.25, + "learning_rate": 4.5914978199538346e-05, + "loss": 0.5653, + "step": 6421 + }, + { + "epoch": 0.25, + "learning_rate": 4.591433700948962e-05, + "loss": 0.5663, + "step": 6422 + }, + { + "epoch": 0.25, + "learning_rate": 4.591369581944088e-05, + "loss": 0.5911, + "step": 6423 + }, + { + "epoch": 0.25, + "learning_rate": 4.5913054629392155e-05, + "loss": 0.5395, + "step": 6424 + }, + { + "epoch": 0.25, + "learning_rate": 4.591241343934342e-05, + "loss": 0.5375, + "step": 6425 + }, + { + "epoch": 0.25, + "learning_rate": 4.591177224929469e-05, + "loss": 0.5487, + "step": 6426 + }, + { + "epoch": 0.25, + "learning_rate": 4.591113105924596e-05, + "loss": 0.5391, + "step": 6427 + }, + { + "epoch": 0.25, + "learning_rate": 4.5910489869197235e-05, + "loss": 0.518, + "step": 6428 + }, + { + "epoch": 0.25, + "learning_rate": 4.5909848679148507e-05, + "loss": 0.5974, + "step": 6429 + }, + { + "epoch": 0.25, + "learning_rate": 4.590920748909977e-05, + "loss": 0.6032, + "step": 6430 + }, + { + "epoch": 0.25, + "learning_rate": 4.590856629905104e-05, + "loss": 0.5045, + "step": 6431 + }, + { + "epoch": 0.25, + "learning_rate": 4.590792510900231e-05, + "loss": 0.6319, + "step": 6432 + }, + { + "epoch": 0.25, + "learning_rate": 4.590728391895358e-05, + "loss": 0.609, + "step": 6433 + }, + { + "epoch": 0.25, + "learning_rate": 4.5906642728904845e-05, + "loss": 0.5623, + "step": 6434 + }, + { + "epoch": 0.25, + "learning_rate": 4.5906001538856117e-05, + "loss": 0.6023, + "step": 6435 + }, + { + "epoch": 0.25, + "learning_rate": 4.590536034880739e-05, + "loss": 0.676, + "step": 6436 + }, + { + "epoch": 0.25, + "learning_rate": 4.590471915875866e-05, + "loss": 0.5371, + "step": 6437 + }, + { + "epoch": 0.25, + "learning_rate": 4.590407796870993e-05, + "loss": 0.6278, + "step": 6438 + }, + { + "epoch": 0.25, + "learning_rate": 4.59034367786612e-05, + "loss": 0.532, + "step": 6439 + }, + { + "epoch": 0.25, + "learning_rate": 4.590279558861247e-05, + "loss": 0.6171, + "step": 6440 + }, + { + "epoch": 0.25, + "learning_rate": 4.590215439856373e-05, + "loss": 0.5539, + "step": 6441 + }, + { + "epoch": 0.25, + "learning_rate": 4.5901513208515005e-05, + "loss": 0.5397, + "step": 6442 + }, + { + "epoch": 0.25, + "learning_rate": 4.590087201846627e-05, + "loss": 0.569, + "step": 6443 + }, + { + "epoch": 0.25, + "learning_rate": 4.590023082841755e-05, + "loss": 0.5313, + "step": 6444 + }, + { + "epoch": 0.25, + "learning_rate": 4.589958963836882e-05, + "loss": 0.6468, + "step": 6445 + }, + { + "epoch": 0.25, + "learning_rate": 4.5898948448320085e-05, + "loss": 0.5581, + "step": 6446 + }, + { + "epoch": 0.25, + "learning_rate": 4.589830725827136e-05, + "loss": 0.5349, + "step": 6447 + }, + { + "epoch": 0.25, + "learning_rate": 4.589766606822262e-05, + "loss": 0.6538, + "step": 6448 + }, + { + "epoch": 0.25, + "learning_rate": 4.5897024878173893e-05, + "loss": 0.5774, + "step": 6449 + }, + { + "epoch": 0.25, + "learning_rate": 4.589638368812516e-05, + "loss": 0.5196, + "step": 6450 + }, + { + "epoch": 0.25, + "learning_rate": 4.589574249807643e-05, + "loss": 0.5, + "step": 6451 + }, + { + "epoch": 0.25, + "learning_rate": 4.58951013080277e-05, + "loss": 0.626, + "step": 6452 + }, + { + "epoch": 0.25, + "learning_rate": 4.5894460117978974e-05, + "loss": 0.5552, + "step": 6453 + }, + { + "epoch": 0.25, + "learning_rate": 4.5893818927930245e-05, + "loss": 0.6161, + "step": 6454 + }, + { + "epoch": 0.25, + "learning_rate": 4.589317773788151e-05, + "loss": 0.6253, + "step": 6455 + }, + { + "epoch": 0.25, + "learning_rate": 4.589253654783278e-05, + "loss": 0.6359, + "step": 6456 + }, + { + "epoch": 0.25, + "learning_rate": 4.589189535778405e-05, + "loss": 0.5712, + "step": 6457 + }, + { + "epoch": 0.25, + "learning_rate": 4.589125416773532e-05, + "loss": 0.6155, + "step": 6458 + }, + { + "epoch": 0.25, + "learning_rate": 4.5890612977686584e-05, + "loss": 0.5986, + "step": 6459 + }, + { + "epoch": 0.25, + "learning_rate": 4.588997178763786e-05, + "loss": 0.6397, + "step": 6460 + }, + { + "epoch": 0.25, + "learning_rate": 4.588933059758913e-05, + "loss": 0.5965, + "step": 6461 + }, + { + "epoch": 0.25, + "learning_rate": 4.58886894075404e-05, + "loss": 0.6234, + "step": 6462 + }, + { + "epoch": 0.25, + "learning_rate": 4.588804821749167e-05, + "loss": 0.606, + "step": 6463 + }, + { + "epoch": 0.25, + "learning_rate": 4.5887407027442935e-05, + "loss": 0.6014, + "step": 6464 + }, + { + "epoch": 0.25, + "learning_rate": 4.588676583739421e-05, + "loss": 0.5272, + "step": 6465 + }, + { + "epoch": 0.25, + "learning_rate": 4.588612464734547e-05, + "loss": 0.5911, + "step": 6466 + }, + { + "epoch": 0.25, + "learning_rate": 4.5885483457296744e-05, + "loss": 0.5509, + "step": 6467 + }, + { + "epoch": 0.25, + "learning_rate": 4.5884842267248015e-05, + "loss": 0.6349, + "step": 6468 + }, + { + "epoch": 0.25, + "learning_rate": 4.588420107719929e-05, + "loss": 0.5839, + "step": 6469 + }, + { + "epoch": 0.25, + "learning_rate": 4.588355988715055e-05, + "loss": 0.5176, + "step": 6470 + }, + { + "epoch": 0.25, + "learning_rate": 4.5882918697101824e-05, + "loss": 0.5659, + "step": 6471 + }, + { + "epoch": 0.25, + "learning_rate": 4.5882277507053096e-05, + "loss": 0.6735, + "step": 6472 + }, + { + "epoch": 0.25, + "learning_rate": 4.588163631700436e-05, + "loss": 0.4758, + "step": 6473 + }, + { + "epoch": 0.25, + "learning_rate": 4.588099512695563e-05, + "loss": 0.5928, + "step": 6474 + }, + { + "epoch": 0.25, + "learning_rate": 4.58803539369069e-05, + "loss": 0.5516, + "step": 6475 + }, + { + "epoch": 0.25, + "learning_rate": 4.587971274685817e-05, + "loss": 0.5851, + "step": 6476 + }, + { + "epoch": 0.25, + "learning_rate": 4.587907155680944e-05, + "loss": 0.5381, + "step": 6477 + }, + { + "epoch": 0.25, + "learning_rate": 4.587843036676071e-05, + "loss": 0.5725, + "step": 6478 + }, + { + "epoch": 0.25, + "learning_rate": 4.587778917671198e-05, + "loss": 0.5635, + "step": 6479 + }, + { + "epoch": 0.25, + "learning_rate": 4.587714798666325e-05, + "loss": 0.5695, + "step": 6480 + }, + { + "epoch": 0.25, + "learning_rate": 4.587650679661452e-05, + "loss": 0.5876, + "step": 6481 + }, + { + "epoch": 0.25, + "learning_rate": 4.5875865606565786e-05, + "loss": 0.5823, + "step": 6482 + }, + { + "epoch": 0.25, + "learning_rate": 4.587522441651706e-05, + "loss": 0.5202, + "step": 6483 + }, + { + "epoch": 0.25, + "learning_rate": 4.587458322646833e-05, + "loss": 0.604, + "step": 6484 + }, + { + "epoch": 0.25, + "learning_rate": 4.58739420364196e-05, + "loss": 0.5383, + "step": 6485 + }, + { + "epoch": 0.25, + "learning_rate": 4.5873300846370866e-05, + "loss": 0.6686, + "step": 6486 + }, + { + "epoch": 0.25, + "learning_rate": 4.587265965632214e-05, + "loss": 0.5384, + "step": 6487 + }, + { + "epoch": 0.25, + "learning_rate": 4.58720184662734e-05, + "loss": 0.4824, + "step": 6488 + }, + { + "epoch": 0.25, + "learning_rate": 4.5871377276224674e-05, + "loss": 0.6148, + "step": 6489 + }, + { + "epoch": 0.25, + "learning_rate": 4.5870736086175946e-05, + "loss": 0.5645, + "step": 6490 + }, + { + "epoch": 0.25, + "learning_rate": 4.587009489612721e-05, + "loss": 0.5493, + "step": 6491 + }, + { + "epoch": 0.25, + "learning_rate": 4.586945370607848e-05, + "loss": 0.6073, + "step": 6492 + }, + { + "epoch": 0.25, + "learning_rate": 4.5868812516029754e-05, + "loss": 0.516, + "step": 6493 + }, + { + "epoch": 0.25, + "learning_rate": 4.5868171325981026e-05, + "loss": 0.5885, + "step": 6494 + }, + { + "epoch": 0.25, + "learning_rate": 4.586753013593229e-05, + "loss": 0.5885, + "step": 6495 + }, + { + "epoch": 0.25, + "learning_rate": 4.586688894588356e-05, + "loss": 0.5209, + "step": 6496 + }, + { + "epoch": 0.25, + "learning_rate": 4.586624775583483e-05, + "loss": 0.6111, + "step": 6497 + }, + { + "epoch": 0.25, + "learning_rate": 4.58656065657861e-05, + "loss": 0.6133, + "step": 6498 + }, + { + "epoch": 0.25, + "learning_rate": 4.586496537573737e-05, + "loss": 0.5724, + "step": 6499 + }, + { + "epoch": 0.25, + "learning_rate": 4.586432418568864e-05, + "loss": 0.6271, + "step": 6500 + }, + { + "epoch": 0.25, + "learning_rate": 4.5863682995639914e-05, + "loss": 0.6271, + "step": 6501 + }, + { + "epoch": 0.25, + "learning_rate": 4.586304180559118e-05, + "loss": 0.6045, + "step": 6502 + }, + { + "epoch": 0.25, + "learning_rate": 4.586240061554245e-05, + "loss": 0.604, + "step": 6503 + }, + { + "epoch": 0.25, + "learning_rate": 4.5861759425493716e-05, + "loss": 0.498, + "step": 6504 + }, + { + "epoch": 0.25, + "learning_rate": 4.586111823544499e-05, + "loss": 0.5711, + "step": 6505 + }, + { + "epoch": 0.25, + "learning_rate": 4.586047704539625e-05, + "loss": 0.5217, + "step": 6506 + }, + { + "epoch": 0.25, + "learning_rate": 4.5859835855347524e-05, + "loss": 0.5287, + "step": 6507 + }, + { + "epoch": 0.25, + "learning_rate": 4.5859194665298796e-05, + "loss": 0.5921, + "step": 6508 + }, + { + "epoch": 0.25, + "learning_rate": 4.585855347525007e-05, + "loss": 0.5874, + "step": 6509 + }, + { + "epoch": 0.25, + "learning_rate": 4.585791228520134e-05, + "loss": 0.5523, + "step": 6510 + }, + { + "epoch": 0.25, + "learning_rate": 4.5857271095152604e-05, + "loss": 0.5917, + "step": 6511 + }, + { + "epoch": 0.25, + "learning_rate": 4.5856629905103876e-05, + "loss": 0.5999, + "step": 6512 + }, + { + "epoch": 0.25, + "learning_rate": 4.585598871505514e-05, + "loss": 0.5679, + "step": 6513 + }, + { + "epoch": 0.25, + "learning_rate": 4.585534752500641e-05, + "loss": 0.6271, + "step": 6514 + }, + { + "epoch": 0.25, + "learning_rate": 4.585470633495768e-05, + "loss": 0.636, + "step": 6515 + }, + { + "epoch": 0.25, + "learning_rate": 4.5854065144908956e-05, + "loss": 0.5215, + "step": 6516 + }, + { + "epoch": 0.25, + "learning_rate": 4.585342395486023e-05, + "loss": 0.5519, + "step": 6517 + }, + { + "epoch": 0.25, + "learning_rate": 4.585278276481149e-05, + "loss": 0.5928, + "step": 6518 + }, + { + "epoch": 0.25, + "learning_rate": 4.5852141574762765e-05, + "loss": 0.5576, + "step": 6519 + }, + { + "epoch": 0.25, + "learning_rate": 4.585150038471403e-05, + "loss": 0.5804, + "step": 6520 + }, + { + "epoch": 0.25, + "learning_rate": 4.58508591946653e-05, + "loss": 0.5912, + "step": 6521 + }, + { + "epoch": 0.25, + "learning_rate": 4.5850218004616566e-05, + "loss": 0.5188, + "step": 6522 + }, + { + "epoch": 0.25, + "learning_rate": 4.584957681456784e-05, + "loss": 0.6246, + "step": 6523 + }, + { + "epoch": 0.25, + "learning_rate": 4.584893562451911e-05, + "loss": 0.5795, + "step": 6524 + }, + { + "epoch": 0.25, + "learning_rate": 4.584829443447038e-05, + "loss": 0.5415, + "step": 6525 + }, + { + "epoch": 0.25, + "learning_rate": 4.584765324442165e-05, + "loss": 0.5335, + "step": 6526 + }, + { + "epoch": 0.25, + "learning_rate": 4.584701205437292e-05, + "loss": 0.605, + "step": 6527 + }, + { + "epoch": 0.25, + "learning_rate": 4.584637086432419e-05, + "loss": 0.5116, + "step": 6528 + }, + { + "epoch": 0.25, + "learning_rate": 4.5845729674275455e-05, + "loss": 0.5525, + "step": 6529 + }, + { + "epoch": 0.25, + "learning_rate": 4.5845088484226726e-05, + "loss": 0.5354, + "step": 6530 + }, + { + "epoch": 0.25, + "learning_rate": 4.584444729417799e-05, + "loss": 0.6093, + "step": 6531 + }, + { + "epoch": 0.25, + "learning_rate": 4.584380610412927e-05, + "loss": 0.5288, + "step": 6532 + }, + { + "epoch": 0.25, + "learning_rate": 4.5843164914080535e-05, + "loss": 0.6368, + "step": 6533 + }, + { + "epoch": 0.25, + "learning_rate": 4.5842523724031806e-05, + "loss": 0.6075, + "step": 6534 + }, + { + "epoch": 0.25, + "learning_rate": 4.584188253398308e-05, + "loss": 0.5257, + "step": 6535 + }, + { + "epoch": 0.25, + "learning_rate": 4.584124134393434e-05, + "loss": 0.6076, + "step": 6536 + }, + { + "epoch": 0.25, + "learning_rate": 4.5840600153885615e-05, + "loss": 0.5743, + "step": 6537 + }, + { + "epoch": 0.25, + "learning_rate": 4.583995896383688e-05, + "loss": 0.5817, + "step": 6538 + }, + { + "epoch": 0.25, + "learning_rate": 4.583931777378815e-05, + "loss": 0.5981, + "step": 6539 + }, + { + "epoch": 0.25, + "learning_rate": 4.583867658373942e-05, + "loss": 0.5181, + "step": 6540 + }, + { + "epoch": 0.25, + "learning_rate": 4.5838035393690695e-05, + "loss": 0.5737, + "step": 6541 + }, + { + "epoch": 0.25, + "learning_rate": 4.583739420364196e-05, + "loss": 0.6536, + "step": 6542 + }, + { + "epoch": 0.25, + "learning_rate": 4.583675301359323e-05, + "loss": 0.4801, + "step": 6543 + }, + { + "epoch": 0.25, + "learning_rate": 4.58361118235445e-05, + "loss": 0.5046, + "step": 6544 + }, + { + "epoch": 0.25, + "learning_rate": 4.583547063349577e-05, + "loss": 0.5334, + "step": 6545 + }, + { + "epoch": 0.25, + "learning_rate": 4.583482944344704e-05, + "loss": 0.5108, + "step": 6546 + }, + { + "epoch": 0.25, + "learning_rate": 4.5834188253398305e-05, + "loss": 0.5691, + "step": 6547 + }, + { + "epoch": 0.25, + "learning_rate": 4.583354706334958e-05, + "loss": 0.5977, + "step": 6548 + }, + { + "epoch": 0.25, + "learning_rate": 4.583290587330085e-05, + "loss": 0.6014, + "step": 6549 + }, + { + "epoch": 0.25, + "learning_rate": 4.583226468325212e-05, + "loss": 0.5727, + "step": 6550 + }, + { + "epoch": 0.25, + "learning_rate": 4.5831623493203385e-05, + "loss": 0.5282, + "step": 6551 + }, + { + "epoch": 0.25, + "learning_rate": 4.583098230315466e-05, + "loss": 0.5121, + "step": 6552 + }, + { + "epoch": 0.25, + "learning_rate": 4.583034111310593e-05, + "loss": 0.5503, + "step": 6553 + }, + { + "epoch": 0.25, + "learning_rate": 4.582969992305719e-05, + "loss": 0.5239, + "step": 6554 + }, + { + "epoch": 0.25, + "learning_rate": 4.5829058733008465e-05, + "loss": 0.6199, + "step": 6555 + }, + { + "epoch": 0.25, + "learning_rate": 4.582841754295974e-05, + "loss": 0.5306, + "step": 6556 + }, + { + "epoch": 0.25, + "learning_rate": 4.582777635291101e-05, + "loss": 0.5985, + "step": 6557 + }, + { + "epoch": 0.25, + "learning_rate": 4.5827135162862273e-05, + "loss": 0.6763, + "step": 6558 + }, + { + "epoch": 0.25, + "learning_rate": 4.5826493972813545e-05, + "loss": 0.666, + "step": 6559 + }, + { + "epoch": 0.25, + "learning_rate": 4.582585278276481e-05, + "loss": 0.7133, + "step": 6560 + }, + { + "epoch": 0.25, + "learning_rate": 4.582521159271608e-05, + "loss": 0.599, + "step": 6561 + }, + { + "epoch": 0.25, + "learning_rate": 4.5824570402667354e-05, + "loss": 0.6222, + "step": 6562 + }, + { + "epoch": 0.25, + "learning_rate": 4.582392921261862e-05, + "loss": 0.6234, + "step": 6563 + }, + { + "epoch": 0.25, + "learning_rate": 4.582328802256989e-05, + "loss": 0.5927, + "step": 6564 + }, + { + "epoch": 0.25, + "learning_rate": 4.582264683252116e-05, + "loss": 0.6262, + "step": 6565 + }, + { + "epoch": 0.25, + "learning_rate": 4.5822005642472434e-05, + "loss": 0.648, + "step": 6566 + }, + { + "epoch": 0.25, + "learning_rate": 4.58213644524237e-05, + "loss": 0.543, + "step": 6567 + }, + { + "epoch": 0.25, + "learning_rate": 4.582072326237497e-05, + "loss": 0.6333, + "step": 6568 + }, + { + "epoch": 0.25, + "learning_rate": 4.5820082072326235e-05, + "loss": 0.5758, + "step": 6569 + }, + { + "epoch": 0.25, + "learning_rate": 4.581944088227751e-05, + "loss": 0.587, + "step": 6570 + }, + { + "epoch": 0.25, + "learning_rate": 4.581879969222878e-05, + "loss": 0.5627, + "step": 6571 + }, + { + "epoch": 0.25, + "learning_rate": 4.581815850218005e-05, + "loss": 0.5882, + "step": 6572 + }, + { + "epoch": 0.25, + "learning_rate": 4.581751731213132e-05, + "loss": 0.6138, + "step": 6573 + }, + { + "epoch": 0.25, + "learning_rate": 4.581687612208259e-05, + "loss": 0.6367, + "step": 6574 + }, + { + "epoch": 0.25, + "learning_rate": 4.581623493203386e-05, + "loss": 0.5952, + "step": 6575 + }, + { + "epoch": 0.25, + "learning_rate": 4.5815593741985124e-05, + "loss": 0.5009, + "step": 6576 + }, + { + "epoch": 0.25, + "learning_rate": 4.5814952551936395e-05, + "loss": 0.6016, + "step": 6577 + }, + { + "epoch": 0.25, + "learning_rate": 4.581431136188766e-05, + "loss": 0.6068, + "step": 6578 + }, + { + "epoch": 0.25, + "learning_rate": 4.581367017183893e-05, + "loss": 0.5109, + "step": 6579 + }, + { + "epoch": 0.25, + "learning_rate": 4.5813028981790204e-05, + "loss": 0.5945, + "step": 6580 + }, + { + "epoch": 0.25, + "learning_rate": 4.5812387791741475e-05, + "loss": 0.5288, + "step": 6581 + }, + { + "epoch": 0.25, + "learning_rate": 4.581174660169275e-05, + "loss": 0.4935, + "step": 6582 + }, + { + "epoch": 0.25, + "learning_rate": 4.581110541164401e-05, + "loss": 0.5928, + "step": 6583 + }, + { + "epoch": 0.25, + "learning_rate": 4.5810464221595284e-05, + "loss": 0.6098, + "step": 6584 + }, + { + "epoch": 0.25, + "learning_rate": 4.580982303154655e-05, + "loss": 0.5122, + "step": 6585 + }, + { + "epoch": 0.25, + "learning_rate": 4.580918184149782e-05, + "loss": 0.5579, + "step": 6586 + }, + { + "epoch": 0.25, + "learning_rate": 4.580854065144909e-05, + "loss": 0.6238, + "step": 6587 + }, + { + "epoch": 0.25, + "learning_rate": 4.5807899461400364e-05, + "loss": 0.6052, + "step": 6588 + }, + { + "epoch": 0.25, + "learning_rate": 4.5807258271351636e-05, + "loss": 0.615, + "step": 6589 + }, + { + "epoch": 0.25, + "learning_rate": 4.58066170813029e-05, + "loss": 0.6555, + "step": 6590 + }, + { + "epoch": 0.25, + "learning_rate": 4.580597589125417e-05, + "loss": 0.6282, + "step": 6591 + }, + { + "epoch": 0.25, + "learning_rate": 4.580533470120544e-05, + "loss": 0.7015, + "step": 6592 + }, + { + "epoch": 0.25, + "learning_rate": 4.580469351115671e-05, + "loss": 0.6178, + "step": 6593 + }, + { + "epoch": 0.25, + "learning_rate": 4.5804052321107974e-05, + "loss": 0.5236, + "step": 6594 + }, + { + "epoch": 0.25, + "learning_rate": 4.5803411131059246e-05, + "loss": 0.5617, + "step": 6595 + }, + { + "epoch": 0.25, + "learning_rate": 4.580276994101052e-05, + "loss": 0.5464, + "step": 6596 + }, + { + "epoch": 0.25, + "learning_rate": 4.580212875096179e-05, + "loss": 0.5842, + "step": 6597 + }, + { + "epoch": 0.25, + "learning_rate": 4.580148756091306e-05, + "loss": 0.5508, + "step": 6598 + }, + { + "epoch": 0.25, + "learning_rate": 4.5800846370864326e-05, + "loss": 0.5848, + "step": 6599 + }, + { + "epoch": 0.25, + "learning_rate": 4.58002051808156e-05, + "loss": 0.6581, + "step": 6600 + }, + { + "epoch": 0.25, + "learning_rate": 4.579956399076686e-05, + "loss": 0.4694, + "step": 6601 + }, + { + "epoch": 0.25, + "learning_rate": 4.5798922800718134e-05, + "loss": 0.534, + "step": 6602 + }, + { + "epoch": 0.25, + "learning_rate": 4.57982816106694e-05, + "loss": 0.5952, + "step": 6603 + }, + { + "epoch": 0.25, + "learning_rate": 4.579764042062068e-05, + "loss": 0.6302, + "step": 6604 + }, + { + "epoch": 0.25, + "learning_rate": 4.579699923057195e-05, + "loss": 0.4754, + "step": 6605 + }, + { + "epoch": 0.25, + "learning_rate": 4.5796358040523214e-05, + "loss": 0.5433, + "step": 6606 + }, + { + "epoch": 0.25, + "learning_rate": 4.5795716850474486e-05, + "loss": 0.5184, + "step": 6607 + }, + { + "epoch": 0.25, + "learning_rate": 4.579507566042575e-05, + "loss": 0.5124, + "step": 6608 + }, + { + "epoch": 0.25, + "learning_rate": 4.579443447037702e-05, + "loss": 0.5921, + "step": 6609 + }, + { + "epoch": 0.25, + "learning_rate": 4.579379328032829e-05, + "loss": 0.7201, + "step": 6610 + }, + { + "epoch": 0.25, + "learning_rate": 4.579315209027956e-05, + "loss": 0.5862, + "step": 6611 + }, + { + "epoch": 0.25, + "learning_rate": 4.579251090023083e-05, + "loss": 0.613, + "step": 6612 + }, + { + "epoch": 0.25, + "learning_rate": 4.57918697101821e-05, + "loss": 0.4656, + "step": 6613 + }, + { + "epoch": 0.25, + "learning_rate": 4.5791228520133374e-05, + "loss": 0.5923, + "step": 6614 + }, + { + "epoch": 0.25, + "learning_rate": 4.579058733008464e-05, + "loss": 0.6342, + "step": 6615 + }, + { + "epoch": 0.25, + "learning_rate": 4.578994614003591e-05, + "loss": 0.4485, + "step": 6616 + }, + { + "epoch": 0.25, + "learning_rate": 4.5789304949987176e-05, + "loss": 0.5142, + "step": 6617 + }, + { + "epoch": 0.25, + "learning_rate": 4.578866375993845e-05, + "loss": 0.5755, + "step": 6618 + }, + { + "epoch": 0.25, + "learning_rate": 4.578802256988971e-05, + "loss": 0.6033, + "step": 6619 + }, + { + "epoch": 0.25, + "learning_rate": 4.578738137984099e-05, + "loss": 0.6084, + "step": 6620 + }, + { + "epoch": 0.25, + "learning_rate": 4.5786740189792256e-05, + "loss": 0.5439, + "step": 6621 + }, + { + "epoch": 0.25, + "learning_rate": 4.578609899974353e-05, + "loss": 0.6245, + "step": 6622 + }, + { + "epoch": 0.25, + "learning_rate": 4.57854578096948e-05, + "loss": 0.6131, + "step": 6623 + }, + { + "epoch": 0.25, + "learning_rate": 4.5784816619646064e-05, + "loss": 0.5613, + "step": 6624 + }, + { + "epoch": 0.25, + "learning_rate": 4.5784175429597336e-05, + "loss": 0.5416, + "step": 6625 + }, + { + "epoch": 0.25, + "learning_rate": 4.57835342395486e-05, + "loss": 0.6287, + "step": 6626 + }, + { + "epoch": 0.25, + "learning_rate": 4.578289304949987e-05, + "loss": 0.6506, + "step": 6627 + }, + { + "epoch": 0.25, + "learning_rate": 4.5782251859451145e-05, + "loss": 0.5348, + "step": 6628 + }, + { + "epoch": 0.25, + "learning_rate": 4.5781610669402416e-05, + "loss": 0.6056, + "step": 6629 + }, + { + "epoch": 0.25, + "learning_rate": 4.578096947935368e-05, + "loss": 0.6527, + "step": 6630 + }, + { + "epoch": 0.25, + "learning_rate": 4.578032828930495e-05, + "loss": 0.6311, + "step": 6631 + }, + { + "epoch": 0.25, + "learning_rate": 4.5779687099256225e-05, + "loss": 0.603, + "step": 6632 + }, + { + "epoch": 0.26, + "learning_rate": 4.577904590920749e-05, + "loss": 0.5476, + "step": 6633 + }, + { + "epoch": 0.26, + "learning_rate": 4.577840471915876e-05, + "loss": 0.5824, + "step": 6634 + }, + { + "epoch": 0.26, + "learning_rate": 4.5777763529110026e-05, + "loss": 0.5917, + "step": 6635 + }, + { + "epoch": 0.26, + "learning_rate": 4.5777122339061305e-05, + "loss": 0.5043, + "step": 6636 + }, + { + "epoch": 0.26, + "learning_rate": 4.577648114901257e-05, + "loss": 0.6001, + "step": 6637 + }, + { + "epoch": 0.26, + "learning_rate": 4.577583995896384e-05, + "loss": 0.6075, + "step": 6638 + }, + { + "epoch": 0.26, + "learning_rate": 4.5775198768915106e-05, + "loss": 0.538, + "step": 6639 + }, + { + "epoch": 0.26, + "learning_rate": 4.577455757886638e-05, + "loss": 0.6252, + "step": 6640 + }, + { + "epoch": 0.26, + "learning_rate": 4.577391638881765e-05, + "loss": 0.6145, + "step": 6641 + }, + { + "epoch": 0.26, + "learning_rate": 4.5773275198768915e-05, + "loss": 0.5066, + "step": 6642 + }, + { + "epoch": 0.26, + "learning_rate": 4.5772634008720186e-05, + "loss": 0.5869, + "step": 6643 + }, + { + "epoch": 0.26, + "learning_rate": 4.577199281867146e-05, + "loss": 0.4995, + "step": 6644 + }, + { + "epoch": 0.26, + "learning_rate": 4.577135162862273e-05, + "loss": 0.5054, + "step": 6645 + }, + { + "epoch": 0.26, + "learning_rate": 4.5770710438573995e-05, + "loss": 0.5393, + "step": 6646 + }, + { + "epoch": 0.26, + "learning_rate": 4.5770069248525266e-05, + "loss": 0.6076, + "step": 6647 + }, + { + "epoch": 0.26, + "learning_rate": 4.576942805847653e-05, + "loss": 0.5516, + "step": 6648 + }, + { + "epoch": 0.26, + "learning_rate": 4.57687868684278e-05, + "loss": 0.4264, + "step": 6649 + }, + { + "epoch": 0.26, + "learning_rate": 4.5768145678379075e-05, + "loss": 0.5523, + "step": 6650 + }, + { + "epoch": 0.26, + "learning_rate": 4.576750448833034e-05, + "loss": 0.5339, + "step": 6651 + }, + { + "epoch": 0.26, + "learning_rate": 4.576686329828161e-05, + "loss": 0.5786, + "step": 6652 + }, + { + "epoch": 0.26, + "learning_rate": 4.576622210823288e-05, + "loss": 0.6487, + "step": 6653 + }, + { + "epoch": 0.26, + "learning_rate": 4.5765580918184155e-05, + "loss": 0.4725, + "step": 6654 + }, + { + "epoch": 0.26, + "learning_rate": 4.576493972813542e-05, + "loss": 0.5844, + "step": 6655 + }, + { + "epoch": 0.26, + "learning_rate": 4.576429853808669e-05, + "loss": 0.6294, + "step": 6656 + }, + { + "epoch": 0.26, + "learning_rate": 4.5763657348037957e-05, + "loss": 0.573, + "step": 6657 + }, + { + "epoch": 0.26, + "learning_rate": 4.576301615798923e-05, + "loss": 0.628, + "step": 6658 + }, + { + "epoch": 0.26, + "learning_rate": 4.57623749679405e-05, + "loss": 0.6416, + "step": 6659 + }, + { + "epoch": 0.26, + "learning_rate": 4.576173377789177e-05, + "loss": 0.5653, + "step": 6660 + }, + { + "epoch": 0.26, + "learning_rate": 4.5761092587843043e-05, + "loss": 0.5054, + "step": 6661 + }, + { + "epoch": 0.26, + "learning_rate": 4.576045139779431e-05, + "loss": 0.6188, + "step": 6662 + }, + { + "epoch": 0.26, + "learning_rate": 4.575981020774558e-05, + "loss": 0.5523, + "step": 6663 + }, + { + "epoch": 0.26, + "learning_rate": 4.5759169017696845e-05, + "loss": 0.5688, + "step": 6664 + }, + { + "epoch": 0.26, + "learning_rate": 4.575852782764812e-05, + "loss": 0.6205, + "step": 6665 + }, + { + "epoch": 0.26, + "learning_rate": 4.575788663759938e-05, + "loss": 0.5579, + "step": 6666 + }, + { + "epoch": 0.26, + "learning_rate": 4.575724544755065e-05, + "loss": 0.5963, + "step": 6667 + }, + { + "epoch": 0.26, + "learning_rate": 4.5756604257501925e-05, + "loss": 0.5528, + "step": 6668 + }, + { + "epoch": 0.26, + "learning_rate": 4.57559630674532e-05, + "loss": 0.552, + "step": 6669 + }, + { + "epoch": 0.26, + "learning_rate": 4.575532187740447e-05, + "loss": 0.5543, + "step": 6670 + }, + { + "epoch": 0.26, + "learning_rate": 4.5754680687355733e-05, + "loss": 0.6105, + "step": 6671 + }, + { + "epoch": 0.26, + "learning_rate": 4.5754039497307005e-05, + "loss": 0.5138, + "step": 6672 + }, + { + "epoch": 0.26, + "learning_rate": 4.575339830725827e-05, + "loss": 0.5497, + "step": 6673 + }, + { + "epoch": 0.26, + "learning_rate": 4.575275711720954e-05, + "loss": 0.5337, + "step": 6674 + }, + { + "epoch": 0.26, + "learning_rate": 4.5752115927160814e-05, + "loss": 0.5367, + "step": 6675 + }, + { + "epoch": 0.26, + "learning_rate": 4.5751474737112085e-05, + "loss": 0.6414, + "step": 6676 + }, + { + "epoch": 0.26, + "learning_rate": 4.575083354706336e-05, + "loss": 0.5404, + "step": 6677 + }, + { + "epoch": 0.26, + "learning_rate": 4.575019235701462e-05, + "loss": 0.5879, + "step": 6678 + }, + { + "epoch": 0.26, + "learning_rate": 4.5749551166965894e-05, + "loss": 0.7301, + "step": 6679 + }, + { + "epoch": 0.26, + "learning_rate": 4.574890997691716e-05, + "loss": 0.5658, + "step": 6680 + }, + { + "epoch": 0.26, + "learning_rate": 4.574826878686843e-05, + "loss": 0.5239, + "step": 6681 + }, + { + "epoch": 0.26, + "learning_rate": 4.5747627596819695e-05, + "loss": 0.551, + "step": 6682 + }, + { + "epoch": 0.26, + "learning_rate": 4.574698640677097e-05, + "loss": 0.5821, + "step": 6683 + }, + { + "epoch": 0.26, + "learning_rate": 4.574634521672224e-05, + "loss": 0.5777, + "step": 6684 + }, + { + "epoch": 0.26, + "learning_rate": 4.574570402667351e-05, + "loss": 0.4886, + "step": 6685 + }, + { + "epoch": 0.26, + "learning_rate": 4.574506283662478e-05, + "loss": 0.6235, + "step": 6686 + }, + { + "epoch": 0.26, + "learning_rate": 4.574442164657605e-05, + "loss": 0.6043, + "step": 6687 + }, + { + "epoch": 0.26, + "learning_rate": 4.574378045652732e-05, + "loss": 0.5181, + "step": 6688 + }, + { + "epoch": 0.26, + "learning_rate": 4.5743139266478584e-05, + "loss": 0.496, + "step": 6689 + }, + { + "epoch": 0.26, + "learning_rate": 4.5742498076429855e-05, + "loss": 0.6316, + "step": 6690 + }, + { + "epoch": 0.26, + "learning_rate": 4.574185688638112e-05, + "loss": 0.599, + "step": 6691 + }, + { + "epoch": 0.26, + "learning_rate": 4.57412156963324e-05, + "loss": 0.5443, + "step": 6692 + }, + { + "epoch": 0.26, + "learning_rate": 4.5740574506283664e-05, + "loss": 0.5643, + "step": 6693 + }, + { + "epoch": 0.26, + "learning_rate": 4.5739933316234936e-05, + "loss": 0.6395, + "step": 6694 + }, + { + "epoch": 0.26, + "learning_rate": 4.573929212618621e-05, + "loss": 0.5862, + "step": 6695 + }, + { + "epoch": 0.26, + "learning_rate": 4.573865093613747e-05, + "loss": 0.5705, + "step": 6696 + }, + { + "epoch": 0.26, + "learning_rate": 4.5738009746088744e-05, + "loss": 0.5098, + "step": 6697 + }, + { + "epoch": 0.26, + "learning_rate": 4.573736855604001e-05, + "loss": 0.5931, + "step": 6698 + }, + { + "epoch": 0.26, + "learning_rate": 4.573672736599128e-05, + "loss": 0.5811, + "step": 6699 + }, + { + "epoch": 0.26, + "learning_rate": 4.573608617594255e-05, + "loss": 0.5904, + "step": 6700 + }, + { + "epoch": 0.26, + "learning_rate": 4.5735444985893824e-05, + "loss": 0.5391, + "step": 6701 + }, + { + "epoch": 0.26, + "learning_rate": 4.573480379584509e-05, + "loss": 0.6247, + "step": 6702 + }, + { + "epoch": 0.26, + "learning_rate": 4.573416260579636e-05, + "loss": 0.6344, + "step": 6703 + }, + { + "epoch": 0.26, + "learning_rate": 4.573352141574763e-05, + "loss": 0.6459, + "step": 6704 + }, + { + "epoch": 0.26, + "learning_rate": 4.57328802256989e-05, + "loss": 0.6105, + "step": 6705 + }, + { + "epoch": 0.26, + "learning_rate": 4.573223903565017e-05, + "loss": 0.5496, + "step": 6706 + }, + { + "epoch": 0.26, + "learning_rate": 4.5731597845601434e-05, + "loss": 0.5903, + "step": 6707 + }, + { + "epoch": 0.26, + "learning_rate": 4.573095665555271e-05, + "loss": 0.5842, + "step": 6708 + }, + { + "epoch": 0.26, + "learning_rate": 4.573031546550398e-05, + "loss": 0.5971, + "step": 6709 + }, + { + "epoch": 0.26, + "learning_rate": 4.572967427545525e-05, + "loss": 0.533, + "step": 6710 + }, + { + "epoch": 0.26, + "learning_rate": 4.5729033085406514e-05, + "loss": 0.5423, + "step": 6711 + }, + { + "epoch": 0.26, + "learning_rate": 4.5728391895357786e-05, + "loss": 0.5622, + "step": 6712 + }, + { + "epoch": 0.26, + "learning_rate": 4.572775070530906e-05, + "loss": 0.5394, + "step": 6713 + }, + { + "epoch": 0.26, + "learning_rate": 4.572710951526032e-05, + "loss": 0.5383, + "step": 6714 + }, + { + "epoch": 0.26, + "learning_rate": 4.5726468325211594e-05, + "loss": 0.6094, + "step": 6715 + }, + { + "epoch": 0.26, + "learning_rate": 4.5725827135162866e-05, + "loss": 0.6462, + "step": 6716 + }, + { + "epoch": 0.26, + "learning_rate": 4.572518594511414e-05, + "loss": 0.5362, + "step": 6717 + }, + { + "epoch": 0.26, + "learning_rate": 4.57245447550654e-05, + "loss": 0.5498, + "step": 6718 + }, + { + "epoch": 0.26, + "learning_rate": 4.5723903565016674e-05, + "loss": 0.5046, + "step": 6719 + }, + { + "epoch": 0.26, + "learning_rate": 4.572326237496794e-05, + "loss": 0.5551, + "step": 6720 + }, + { + "epoch": 0.26, + "learning_rate": 4.572262118491921e-05, + "loss": 0.6073, + "step": 6721 + }, + { + "epoch": 0.26, + "learning_rate": 4.572197999487048e-05, + "loss": 0.618, + "step": 6722 + }, + { + "epoch": 0.26, + "learning_rate": 4.572133880482175e-05, + "loss": 0.5544, + "step": 6723 + }, + { + "epoch": 0.26, + "learning_rate": 4.572069761477302e-05, + "loss": 0.6053, + "step": 6724 + }, + { + "epoch": 0.26, + "learning_rate": 4.572005642472429e-05, + "loss": 0.642, + "step": 6725 + }, + { + "epoch": 0.26, + "learning_rate": 4.571941523467556e-05, + "loss": 0.5491, + "step": 6726 + }, + { + "epoch": 0.26, + "learning_rate": 4.571877404462683e-05, + "loss": 0.6339, + "step": 6727 + }, + { + "epoch": 0.26, + "learning_rate": 4.57181328545781e-05, + "loss": 0.61, + "step": 6728 + }, + { + "epoch": 0.26, + "learning_rate": 4.5717491664529364e-05, + "loss": 0.5942, + "step": 6729 + }, + { + "epoch": 0.26, + "learning_rate": 4.5716850474480636e-05, + "loss": 0.5907, + "step": 6730 + }, + { + "epoch": 0.26, + "learning_rate": 4.571620928443191e-05, + "loss": 0.5075, + "step": 6731 + }, + { + "epoch": 0.26, + "learning_rate": 4.571556809438318e-05, + "loss": 0.6427, + "step": 6732 + }, + { + "epoch": 0.26, + "learning_rate": 4.571492690433445e-05, + "loss": 0.6082, + "step": 6733 + }, + { + "epoch": 0.26, + "learning_rate": 4.5714285714285716e-05, + "loss": 0.514, + "step": 6734 + }, + { + "epoch": 0.26, + "learning_rate": 4.571364452423699e-05, + "loss": 0.6093, + "step": 6735 + }, + { + "epoch": 0.26, + "learning_rate": 4.571300333418825e-05, + "loss": 0.5464, + "step": 6736 + }, + { + "epoch": 0.26, + "learning_rate": 4.5712362144139524e-05, + "loss": 0.5774, + "step": 6737 + }, + { + "epoch": 0.26, + "learning_rate": 4.571172095409079e-05, + "loss": 0.5729, + "step": 6738 + }, + { + "epoch": 0.26, + "learning_rate": 4.571107976404206e-05, + "loss": 0.5805, + "step": 6739 + }, + { + "epoch": 0.26, + "learning_rate": 4.571043857399333e-05, + "loss": 0.5905, + "step": 6740 + }, + { + "epoch": 0.26, + "learning_rate": 4.5709797383944605e-05, + "loss": 0.5141, + "step": 6741 + }, + { + "epoch": 0.26, + "learning_rate": 4.5709156193895876e-05, + "loss": 0.5279, + "step": 6742 + }, + { + "epoch": 0.26, + "learning_rate": 4.570851500384714e-05, + "loss": 0.5052, + "step": 6743 + }, + { + "epoch": 0.26, + "learning_rate": 4.570787381379841e-05, + "loss": 0.6382, + "step": 6744 + }, + { + "epoch": 0.26, + "learning_rate": 4.570723262374968e-05, + "loss": 0.5823, + "step": 6745 + }, + { + "epoch": 0.26, + "learning_rate": 4.570659143370095e-05, + "loss": 0.5377, + "step": 6746 + }, + { + "epoch": 0.26, + "learning_rate": 4.570595024365222e-05, + "loss": 0.5965, + "step": 6747 + }, + { + "epoch": 0.26, + "learning_rate": 4.570530905360349e-05, + "loss": 0.6889, + "step": 6748 + }, + { + "epoch": 0.26, + "learning_rate": 4.5704667863554765e-05, + "loss": 0.566, + "step": 6749 + }, + { + "epoch": 0.26, + "learning_rate": 4.570402667350603e-05, + "loss": 0.5339, + "step": 6750 + }, + { + "epoch": 0.26, + "learning_rate": 4.57033854834573e-05, + "loss": 0.5808, + "step": 6751 + }, + { + "epoch": 0.26, + "learning_rate": 4.5702744293408566e-05, + "loss": 0.5289, + "step": 6752 + }, + { + "epoch": 0.26, + "learning_rate": 4.570210310335984e-05, + "loss": 0.6074, + "step": 6753 + }, + { + "epoch": 0.26, + "learning_rate": 4.57014619133111e-05, + "loss": 0.4924, + "step": 6754 + }, + { + "epoch": 0.26, + "learning_rate": 4.5700820723262375e-05, + "loss": 0.5906, + "step": 6755 + }, + { + "epoch": 0.26, + "learning_rate": 4.5700179533213646e-05, + "loss": 0.5996, + "step": 6756 + }, + { + "epoch": 0.26, + "learning_rate": 4.569953834316492e-05, + "loss": 0.5984, + "step": 6757 + }, + { + "epoch": 0.26, + "learning_rate": 4.569889715311619e-05, + "loss": 0.6152, + "step": 6758 + }, + { + "epoch": 0.26, + "learning_rate": 4.5698255963067455e-05, + "loss": 0.6147, + "step": 6759 + }, + { + "epoch": 0.26, + "learning_rate": 4.5697614773018727e-05, + "loss": 0.5853, + "step": 6760 + }, + { + "epoch": 0.26, + "learning_rate": 4.569697358296999e-05, + "loss": 0.5384, + "step": 6761 + }, + { + "epoch": 0.26, + "learning_rate": 4.569633239292126e-05, + "loss": 0.6162, + "step": 6762 + }, + { + "epoch": 0.26, + "learning_rate": 4.569569120287253e-05, + "loss": 0.6251, + "step": 6763 + }, + { + "epoch": 0.26, + "learning_rate": 4.5695050012823807e-05, + "loss": 0.6484, + "step": 6764 + }, + { + "epoch": 0.26, + "learning_rate": 4.569440882277507e-05, + "loss": 0.6177, + "step": 6765 + }, + { + "epoch": 0.26, + "learning_rate": 4.569376763272634e-05, + "loss": 0.6295, + "step": 6766 + }, + { + "epoch": 0.26, + "learning_rate": 4.5693126442677615e-05, + "loss": 0.5266, + "step": 6767 + }, + { + "epoch": 0.26, + "learning_rate": 4.569248525262888e-05, + "loss": 0.5782, + "step": 6768 + }, + { + "epoch": 0.26, + "learning_rate": 4.569184406258015e-05, + "loss": 0.5451, + "step": 6769 + }, + { + "epoch": 0.26, + "learning_rate": 4.5691202872531417e-05, + "loss": 0.564, + "step": 6770 + }, + { + "epoch": 0.26, + "learning_rate": 4.569056168248269e-05, + "loss": 0.6213, + "step": 6771 + }, + { + "epoch": 0.26, + "learning_rate": 4.568992049243396e-05, + "loss": 0.5826, + "step": 6772 + }, + { + "epoch": 0.26, + "learning_rate": 4.568927930238523e-05, + "loss": 0.5633, + "step": 6773 + }, + { + "epoch": 0.26, + "learning_rate": 4.5688638112336503e-05, + "loss": 0.6256, + "step": 6774 + }, + { + "epoch": 0.26, + "learning_rate": 4.568799692228777e-05, + "loss": 0.6109, + "step": 6775 + }, + { + "epoch": 0.26, + "learning_rate": 4.568735573223904e-05, + "loss": 0.6121, + "step": 6776 + }, + { + "epoch": 0.26, + "learning_rate": 4.5686714542190305e-05, + "loss": 0.5725, + "step": 6777 + }, + { + "epoch": 0.26, + "learning_rate": 4.568607335214158e-05, + "loss": 0.5941, + "step": 6778 + }, + { + "epoch": 0.26, + "learning_rate": 4.568543216209284e-05, + "loss": 0.5831, + "step": 6779 + }, + { + "epoch": 0.26, + "learning_rate": 4.568479097204412e-05, + "loss": 0.6597, + "step": 6780 + }, + { + "epoch": 0.26, + "learning_rate": 4.5684149781995385e-05, + "loss": 0.5839, + "step": 6781 + }, + { + "epoch": 0.26, + "learning_rate": 4.568350859194666e-05, + "loss": 0.6222, + "step": 6782 + }, + { + "epoch": 0.26, + "learning_rate": 4.568286740189793e-05, + "loss": 0.6304, + "step": 6783 + }, + { + "epoch": 0.26, + "learning_rate": 4.5682226211849194e-05, + "loss": 0.6545, + "step": 6784 + }, + { + "epoch": 0.26, + "learning_rate": 4.5681585021800465e-05, + "loss": 0.5053, + "step": 6785 + }, + { + "epoch": 0.26, + "learning_rate": 4.568094383175173e-05, + "loss": 0.5334, + "step": 6786 + }, + { + "epoch": 0.26, + "learning_rate": 4.5680302641703e-05, + "loss": 0.5469, + "step": 6787 + }, + { + "epoch": 0.26, + "learning_rate": 4.5679661451654274e-05, + "loss": 0.4691, + "step": 6788 + }, + { + "epoch": 0.26, + "learning_rate": 4.5679020261605545e-05, + "loss": 0.6102, + "step": 6789 + }, + { + "epoch": 0.26, + "learning_rate": 4.567837907155681e-05, + "loss": 0.6669, + "step": 6790 + }, + { + "epoch": 0.26, + "learning_rate": 4.567773788150808e-05, + "loss": 0.6349, + "step": 6791 + }, + { + "epoch": 0.26, + "learning_rate": 4.5677096691459354e-05, + "loss": 0.6148, + "step": 6792 + }, + { + "epoch": 0.26, + "learning_rate": 4.567645550141062e-05, + "loss": 0.4918, + "step": 6793 + }, + { + "epoch": 0.26, + "learning_rate": 4.567581431136189e-05, + "loss": 0.572, + "step": 6794 + }, + { + "epoch": 0.26, + "learning_rate": 4.5675173121313155e-05, + "loss": 0.5411, + "step": 6795 + }, + { + "epoch": 0.26, + "learning_rate": 4.5674531931264434e-05, + "loss": 0.618, + "step": 6796 + }, + { + "epoch": 0.26, + "learning_rate": 4.56738907412157e-05, + "loss": 0.5611, + "step": 6797 + }, + { + "epoch": 0.26, + "learning_rate": 4.567324955116697e-05, + "loss": 0.5854, + "step": 6798 + }, + { + "epoch": 0.26, + "learning_rate": 4.5672608361118235e-05, + "loss": 0.5589, + "step": 6799 + }, + { + "epoch": 0.26, + "learning_rate": 4.567196717106951e-05, + "loss": 0.5643, + "step": 6800 + }, + { + "epoch": 0.26, + "learning_rate": 4.567132598102078e-05, + "loss": 0.6132, + "step": 6801 + }, + { + "epoch": 0.26, + "learning_rate": 4.5670684790972044e-05, + "loss": 0.5558, + "step": 6802 + }, + { + "epoch": 0.26, + "learning_rate": 4.5670043600923315e-05, + "loss": 0.5847, + "step": 6803 + }, + { + "epoch": 0.26, + "learning_rate": 4.566940241087459e-05, + "loss": 0.6526, + "step": 6804 + }, + { + "epoch": 0.26, + "learning_rate": 4.566876122082586e-05, + "loss": 0.5358, + "step": 6805 + }, + { + "epoch": 0.26, + "learning_rate": 4.5668120030777124e-05, + "loss": 0.5741, + "step": 6806 + }, + { + "epoch": 0.26, + "learning_rate": 4.5667478840728396e-05, + "loss": 0.6329, + "step": 6807 + }, + { + "epoch": 0.26, + "learning_rate": 4.566683765067966e-05, + "loss": 0.5465, + "step": 6808 + }, + { + "epoch": 0.26, + "learning_rate": 4.566619646063093e-05, + "loss": 0.5671, + "step": 6809 + }, + { + "epoch": 0.26, + "learning_rate": 4.5665555270582204e-05, + "loss": 0.5319, + "step": 6810 + }, + { + "epoch": 0.26, + "learning_rate": 4.566491408053347e-05, + "loss": 0.5339, + "step": 6811 + }, + { + "epoch": 0.26, + "learning_rate": 4.566427289048474e-05, + "loss": 0.5986, + "step": 6812 + }, + { + "epoch": 0.26, + "learning_rate": 4.566363170043601e-05, + "loss": 0.714, + "step": 6813 + }, + { + "epoch": 0.26, + "learning_rate": 4.5662990510387284e-05, + "loss": 0.5541, + "step": 6814 + }, + { + "epoch": 0.26, + "learning_rate": 4.566234932033855e-05, + "loss": 0.5633, + "step": 6815 + }, + { + "epoch": 0.26, + "learning_rate": 4.566170813028982e-05, + "loss": 0.5141, + "step": 6816 + }, + { + "epoch": 0.26, + "learning_rate": 4.5661066940241086e-05, + "loss": 0.646, + "step": 6817 + }, + { + "epoch": 0.26, + "learning_rate": 4.566042575019236e-05, + "loss": 0.6037, + "step": 6818 + }, + { + "epoch": 0.26, + "learning_rate": 4.565978456014363e-05, + "loss": 0.5549, + "step": 6819 + }, + { + "epoch": 0.26, + "learning_rate": 4.56591433700949e-05, + "loss": 0.6267, + "step": 6820 + }, + { + "epoch": 0.26, + "learning_rate": 4.565850218004617e-05, + "loss": 0.5958, + "step": 6821 + }, + { + "epoch": 0.26, + "learning_rate": 4.565786098999744e-05, + "loss": 0.5062, + "step": 6822 + }, + { + "epoch": 0.26, + "learning_rate": 4.565721979994871e-05, + "loss": 0.6329, + "step": 6823 + }, + { + "epoch": 0.26, + "learning_rate": 4.5656578609899974e-05, + "loss": 0.5686, + "step": 6824 + }, + { + "epoch": 0.26, + "learning_rate": 4.5655937419851246e-05, + "loss": 0.5609, + "step": 6825 + }, + { + "epoch": 0.26, + "learning_rate": 4.565529622980251e-05, + "loss": 0.5842, + "step": 6826 + }, + { + "epoch": 0.26, + "learning_rate": 4.565465503975378e-05, + "loss": 0.5275, + "step": 6827 + }, + { + "epoch": 0.26, + "learning_rate": 4.5654013849705054e-05, + "loss": 0.566, + "step": 6828 + }, + { + "epoch": 0.26, + "learning_rate": 4.5653372659656326e-05, + "loss": 0.5593, + "step": 6829 + }, + { + "epoch": 0.26, + "learning_rate": 4.56527314696076e-05, + "loss": 0.5801, + "step": 6830 + }, + { + "epoch": 0.26, + "learning_rate": 4.565209027955886e-05, + "loss": 0.6023, + "step": 6831 + }, + { + "epoch": 0.26, + "learning_rate": 4.5651449089510134e-05, + "loss": 0.5501, + "step": 6832 + }, + { + "epoch": 0.26, + "learning_rate": 4.56508078994614e-05, + "loss": 0.5966, + "step": 6833 + }, + { + "epoch": 0.26, + "learning_rate": 4.565016670941267e-05, + "loss": 0.5815, + "step": 6834 + }, + { + "epoch": 0.26, + "learning_rate": 4.564952551936394e-05, + "loss": 0.6026, + "step": 6835 + }, + { + "epoch": 0.26, + "learning_rate": 4.5648884329315214e-05, + "loss": 0.662, + "step": 6836 + }, + { + "epoch": 0.26, + "learning_rate": 4.5648243139266486e-05, + "loss": 0.6095, + "step": 6837 + }, + { + "epoch": 0.26, + "learning_rate": 4.564760194921775e-05, + "loss": 0.653, + "step": 6838 + }, + { + "epoch": 0.26, + "learning_rate": 4.564696075916902e-05, + "loss": 0.5636, + "step": 6839 + }, + { + "epoch": 0.26, + "learning_rate": 4.564631956912029e-05, + "loss": 0.5577, + "step": 6840 + }, + { + "epoch": 0.26, + "learning_rate": 4.564567837907156e-05, + "loss": 0.7025, + "step": 6841 + }, + { + "epoch": 0.26, + "learning_rate": 4.5645037189022824e-05, + "loss": 0.5811, + "step": 6842 + }, + { + "epoch": 0.26, + "learning_rate": 4.5644395998974096e-05, + "loss": 0.5874, + "step": 6843 + }, + { + "epoch": 0.26, + "learning_rate": 4.564375480892537e-05, + "loss": 0.5873, + "step": 6844 + }, + { + "epoch": 0.26, + "learning_rate": 4.564311361887664e-05, + "loss": 0.6004, + "step": 6845 + }, + { + "epoch": 0.26, + "learning_rate": 4.564247242882791e-05, + "loss": 0.5716, + "step": 6846 + }, + { + "epoch": 0.26, + "learning_rate": 4.5641831238779176e-05, + "loss": 0.4999, + "step": 6847 + }, + { + "epoch": 0.26, + "learning_rate": 4.564119004873045e-05, + "loss": 0.5173, + "step": 6848 + }, + { + "epoch": 0.26, + "learning_rate": 4.564054885868171e-05, + "loss": 0.4893, + "step": 6849 + }, + { + "epoch": 0.26, + "learning_rate": 4.5639907668632984e-05, + "loss": 0.5675, + "step": 6850 + }, + { + "epoch": 0.26, + "learning_rate": 4.563926647858425e-05, + "loss": 0.6027, + "step": 6851 + }, + { + "epoch": 0.26, + "learning_rate": 4.563862528853553e-05, + "loss": 0.5779, + "step": 6852 + }, + { + "epoch": 0.26, + "learning_rate": 4.563798409848679e-05, + "loss": 0.6083, + "step": 6853 + }, + { + "epoch": 0.26, + "learning_rate": 4.5637342908438065e-05, + "loss": 0.6425, + "step": 6854 + }, + { + "epoch": 0.26, + "learning_rate": 4.5636701718389336e-05, + "loss": 0.4939, + "step": 6855 + }, + { + "epoch": 0.26, + "learning_rate": 4.56360605283406e-05, + "loss": 0.644, + "step": 6856 + }, + { + "epoch": 0.26, + "learning_rate": 4.563541933829187e-05, + "loss": 0.5346, + "step": 6857 + }, + { + "epoch": 0.26, + "learning_rate": 4.563477814824314e-05, + "loss": 0.5509, + "step": 6858 + }, + { + "epoch": 0.26, + "learning_rate": 4.563413695819441e-05, + "loss": 0.6687, + "step": 6859 + }, + { + "epoch": 0.26, + "learning_rate": 4.563349576814568e-05, + "loss": 0.4958, + "step": 6860 + }, + { + "epoch": 0.26, + "learning_rate": 4.563285457809695e-05, + "loss": 0.6196, + "step": 6861 + }, + { + "epoch": 0.26, + "learning_rate": 4.563221338804822e-05, + "loss": 0.586, + "step": 6862 + }, + { + "epoch": 0.26, + "learning_rate": 4.563157219799949e-05, + "loss": 0.6328, + "step": 6863 + }, + { + "epoch": 0.26, + "learning_rate": 4.563093100795076e-05, + "loss": 0.5574, + "step": 6864 + }, + { + "epoch": 0.26, + "learning_rate": 4.5630289817902026e-05, + "loss": 0.5621, + "step": 6865 + }, + { + "epoch": 0.26, + "learning_rate": 4.56296486278533e-05, + "loss": 0.5452, + "step": 6866 + }, + { + "epoch": 0.26, + "learning_rate": 4.562900743780456e-05, + "loss": 0.5599, + "step": 6867 + }, + { + "epoch": 0.26, + "learning_rate": 4.562836624775584e-05, + "loss": 0.5509, + "step": 6868 + }, + { + "epoch": 0.26, + "learning_rate": 4.5627725057707106e-05, + "loss": 0.5439, + "step": 6869 + }, + { + "epoch": 0.26, + "learning_rate": 4.562708386765838e-05, + "loss": 0.6168, + "step": 6870 + }, + { + "epoch": 0.26, + "learning_rate": 4.562644267760964e-05, + "loss": 0.6177, + "step": 6871 + }, + { + "epoch": 0.26, + "learning_rate": 4.5625801487560915e-05, + "loss": 0.4974, + "step": 6872 + }, + { + "epoch": 0.26, + "learning_rate": 4.5625160297512187e-05, + "loss": 0.6118, + "step": 6873 + }, + { + "epoch": 0.26, + "learning_rate": 4.562451910746345e-05, + "loss": 0.7857, + "step": 6874 + }, + { + "epoch": 0.26, + "learning_rate": 4.562387791741472e-05, + "loss": 0.5902, + "step": 6875 + }, + { + "epoch": 0.26, + "learning_rate": 4.5623236727365995e-05, + "loss": 0.6203, + "step": 6876 + }, + { + "epoch": 0.26, + "learning_rate": 4.562259553731727e-05, + "loss": 0.5493, + "step": 6877 + }, + { + "epoch": 0.26, + "learning_rate": 4.562195434726853e-05, + "loss": 0.6108, + "step": 6878 + }, + { + "epoch": 0.26, + "learning_rate": 4.56213131572198e-05, + "loss": 0.5462, + "step": 6879 + }, + { + "epoch": 0.26, + "learning_rate": 4.562067196717107e-05, + "loss": 0.5877, + "step": 6880 + }, + { + "epoch": 0.26, + "learning_rate": 4.562003077712234e-05, + "loss": 0.6352, + "step": 6881 + }, + { + "epoch": 0.26, + "learning_rate": 4.561938958707361e-05, + "loss": 0.689, + "step": 6882 + }, + { + "epoch": 0.26, + "learning_rate": 4.5618748397024877e-05, + "loss": 0.5671, + "step": 6883 + }, + { + "epoch": 0.26, + "learning_rate": 4.561810720697615e-05, + "loss": 0.5493, + "step": 6884 + }, + { + "epoch": 0.26, + "learning_rate": 4.561746601692742e-05, + "loss": 0.5921, + "step": 6885 + }, + { + "epoch": 0.26, + "learning_rate": 4.561682482687869e-05, + "loss": 0.4715, + "step": 6886 + }, + { + "epoch": 0.26, + "learning_rate": 4.561618363682996e-05, + "loss": 0.6091, + "step": 6887 + }, + { + "epoch": 0.26, + "learning_rate": 4.561554244678123e-05, + "loss": 0.5923, + "step": 6888 + }, + { + "epoch": 0.26, + "learning_rate": 4.561490125673249e-05, + "loss": 0.6306, + "step": 6889 + }, + { + "epoch": 0.26, + "learning_rate": 4.5614260066683765e-05, + "loss": 0.5856, + "step": 6890 + }, + { + "epoch": 0.26, + "learning_rate": 4.561361887663504e-05, + "loss": 0.5525, + "step": 6891 + }, + { + "epoch": 0.26, + "learning_rate": 4.561297768658631e-05, + "loss": 0.6269, + "step": 6892 + }, + { + "epoch": 0.27, + "learning_rate": 4.561233649653758e-05, + "loss": 0.6072, + "step": 6893 + }, + { + "epoch": 0.27, + "learning_rate": 4.5611695306488845e-05, + "loss": 0.5077, + "step": 6894 + }, + { + "epoch": 0.27, + "learning_rate": 4.561105411644012e-05, + "loss": 0.5631, + "step": 6895 + }, + { + "epoch": 0.27, + "learning_rate": 4.561041292639138e-05, + "loss": 0.5704, + "step": 6896 + }, + { + "epoch": 0.27, + "learning_rate": 4.5609771736342654e-05, + "loss": 0.5832, + "step": 6897 + }, + { + "epoch": 0.27, + "learning_rate": 4.560913054629392e-05, + "loss": 0.5525, + "step": 6898 + }, + { + "epoch": 0.27, + "learning_rate": 4.560848935624519e-05, + "loss": 0.5667, + "step": 6899 + }, + { + "epoch": 0.27, + "learning_rate": 4.560784816619646e-05, + "loss": 0.5324, + "step": 6900 + }, + { + "epoch": 0.27, + "learning_rate": 4.5607206976147734e-05, + "loss": 0.5661, + "step": 6901 + }, + { + "epoch": 0.27, + "learning_rate": 4.5606565786099005e-05, + "loss": 0.6146, + "step": 6902 + }, + { + "epoch": 0.27, + "learning_rate": 4.560592459605027e-05, + "loss": 0.5538, + "step": 6903 + }, + { + "epoch": 0.27, + "learning_rate": 4.560528340600154e-05, + "loss": 0.599, + "step": 6904 + }, + { + "epoch": 0.27, + "learning_rate": 4.560464221595281e-05, + "loss": 0.5447, + "step": 6905 + }, + { + "epoch": 0.27, + "learning_rate": 4.560400102590408e-05, + "loss": 0.6151, + "step": 6906 + }, + { + "epoch": 0.27, + "learning_rate": 4.560335983585535e-05, + "loss": 0.6214, + "step": 6907 + }, + { + "epoch": 0.27, + "learning_rate": 4.560271864580662e-05, + "loss": 0.4803, + "step": 6908 + }, + { + "epoch": 0.27, + "learning_rate": 4.5602077455757894e-05, + "loss": 0.5611, + "step": 6909 + }, + { + "epoch": 0.27, + "learning_rate": 4.560143626570916e-05, + "loss": 0.54, + "step": 6910 + }, + { + "epoch": 0.27, + "learning_rate": 4.560079507566043e-05, + "loss": 0.614, + "step": 6911 + }, + { + "epoch": 0.27, + "learning_rate": 4.5600153885611695e-05, + "loss": 0.5373, + "step": 6912 + }, + { + "epoch": 0.27, + "learning_rate": 4.559951269556297e-05, + "loss": 0.6067, + "step": 6913 + }, + { + "epoch": 0.27, + "learning_rate": 4.559887150551423e-05, + "loss": 0.549, + "step": 6914 + }, + { + "epoch": 0.27, + "learning_rate": 4.5598230315465504e-05, + "loss": 0.5307, + "step": 6915 + }, + { + "epoch": 0.27, + "learning_rate": 4.5597589125416775e-05, + "loss": 0.6186, + "step": 6916 + }, + { + "epoch": 0.27, + "learning_rate": 4.559694793536805e-05, + "loss": 0.5549, + "step": 6917 + }, + { + "epoch": 0.27, + "learning_rate": 4.559630674531932e-05, + "loss": 0.6294, + "step": 6918 + }, + { + "epoch": 0.27, + "learning_rate": 4.5595665555270584e-05, + "loss": 0.4727, + "step": 6919 + }, + { + "epoch": 0.27, + "learning_rate": 4.5595024365221856e-05, + "loss": 0.4728, + "step": 6920 + }, + { + "epoch": 0.27, + "learning_rate": 4.559438317517312e-05, + "loss": 0.6172, + "step": 6921 + }, + { + "epoch": 0.27, + "learning_rate": 4.559374198512439e-05, + "loss": 0.5592, + "step": 6922 + }, + { + "epoch": 0.27, + "learning_rate": 4.5593100795075664e-05, + "loss": 0.6671, + "step": 6923 + }, + { + "epoch": 0.27, + "learning_rate": 4.5592459605026936e-05, + "loss": 0.592, + "step": 6924 + }, + { + "epoch": 0.27, + "learning_rate": 4.55918184149782e-05, + "loss": 0.4372, + "step": 6925 + }, + { + "epoch": 0.27, + "learning_rate": 4.559117722492947e-05, + "loss": 0.4929, + "step": 6926 + }, + { + "epoch": 0.27, + "learning_rate": 4.5590536034880744e-05, + "loss": 0.631, + "step": 6927 + }, + { + "epoch": 0.27, + "learning_rate": 4.558989484483201e-05, + "loss": 0.6556, + "step": 6928 + }, + { + "epoch": 0.27, + "learning_rate": 4.558925365478328e-05, + "loss": 0.6305, + "step": 6929 + }, + { + "epoch": 0.27, + "learning_rate": 4.5588612464734546e-05, + "loss": 0.6041, + "step": 6930 + }, + { + "epoch": 0.27, + "learning_rate": 4.558797127468582e-05, + "loss": 0.6017, + "step": 6931 + }, + { + "epoch": 0.27, + "learning_rate": 4.558733008463709e-05, + "loss": 0.59, + "step": 6932 + }, + { + "epoch": 0.27, + "learning_rate": 4.558668889458836e-05, + "loss": 0.5323, + "step": 6933 + }, + { + "epoch": 0.27, + "learning_rate": 4.5586047704539626e-05, + "loss": 0.6266, + "step": 6934 + }, + { + "epoch": 0.27, + "learning_rate": 4.55854065144909e-05, + "loss": 0.5003, + "step": 6935 + }, + { + "epoch": 0.27, + "learning_rate": 4.558476532444217e-05, + "loss": 0.5572, + "step": 6936 + }, + { + "epoch": 0.27, + "learning_rate": 4.5584124134393434e-05, + "loss": 0.5801, + "step": 6937 + }, + { + "epoch": 0.27, + "learning_rate": 4.5583482944344706e-05, + "loss": 0.5261, + "step": 6938 + }, + { + "epoch": 0.27, + "learning_rate": 4.558284175429597e-05, + "loss": 0.5827, + "step": 6939 + }, + { + "epoch": 0.27, + "learning_rate": 4.558220056424725e-05, + "loss": 0.6387, + "step": 6940 + }, + { + "epoch": 0.27, + "learning_rate": 4.5581559374198514e-05, + "loss": 0.5711, + "step": 6941 + }, + { + "epoch": 0.27, + "learning_rate": 4.5580918184149786e-05, + "loss": 0.6041, + "step": 6942 + }, + { + "epoch": 0.27, + "learning_rate": 4.558027699410105e-05, + "loss": 0.5832, + "step": 6943 + }, + { + "epoch": 0.27, + "learning_rate": 4.557963580405232e-05, + "loss": 0.5197, + "step": 6944 + }, + { + "epoch": 0.27, + "learning_rate": 4.5578994614003594e-05, + "loss": 0.6342, + "step": 6945 + }, + { + "epoch": 0.27, + "learning_rate": 4.557835342395486e-05, + "loss": 0.5437, + "step": 6946 + }, + { + "epoch": 0.27, + "learning_rate": 4.557771223390613e-05, + "loss": 0.5702, + "step": 6947 + }, + { + "epoch": 0.27, + "learning_rate": 4.55770710438574e-05, + "loss": 0.6241, + "step": 6948 + }, + { + "epoch": 0.27, + "learning_rate": 4.5576429853808674e-05, + "loss": 0.4985, + "step": 6949 + }, + { + "epoch": 0.27, + "learning_rate": 4.557578866375994e-05, + "loss": 0.5562, + "step": 6950 + }, + { + "epoch": 0.27, + "learning_rate": 4.557514747371121e-05, + "loss": 0.556, + "step": 6951 + }, + { + "epoch": 0.27, + "learning_rate": 4.5574506283662476e-05, + "loss": 0.698, + "step": 6952 + }, + { + "epoch": 0.27, + "learning_rate": 4.557386509361375e-05, + "loss": 0.6031, + "step": 6953 + }, + { + "epoch": 0.27, + "learning_rate": 4.557322390356502e-05, + "loss": 0.6465, + "step": 6954 + }, + { + "epoch": 0.27, + "learning_rate": 4.5572582713516284e-05, + "loss": 0.6265, + "step": 6955 + }, + { + "epoch": 0.27, + "learning_rate": 4.557194152346756e-05, + "loss": 0.5332, + "step": 6956 + }, + { + "epoch": 0.27, + "learning_rate": 4.557130033341883e-05, + "loss": 0.5942, + "step": 6957 + }, + { + "epoch": 0.27, + "learning_rate": 4.55706591433701e-05, + "loss": 0.5542, + "step": 6958 + }, + { + "epoch": 0.27, + "learning_rate": 4.5570017953321364e-05, + "loss": 0.6581, + "step": 6959 + }, + { + "epoch": 0.27, + "learning_rate": 4.5569376763272636e-05, + "loss": 0.5727, + "step": 6960 + }, + { + "epoch": 0.27, + "learning_rate": 4.556873557322391e-05, + "loss": 0.64, + "step": 6961 + }, + { + "epoch": 0.27, + "learning_rate": 4.556809438317517e-05, + "loss": 0.5921, + "step": 6962 + }, + { + "epoch": 0.27, + "learning_rate": 4.5567453193126445e-05, + "loss": 0.5666, + "step": 6963 + }, + { + "epoch": 0.27, + "learning_rate": 4.5566812003077716e-05, + "loss": 0.5033, + "step": 6964 + }, + { + "epoch": 0.27, + "learning_rate": 4.556617081302899e-05, + "loss": 0.648, + "step": 6965 + }, + { + "epoch": 0.27, + "learning_rate": 4.556552962298025e-05, + "loss": 0.4895, + "step": 6966 + }, + { + "epoch": 0.27, + "learning_rate": 4.5564888432931525e-05, + "loss": 0.6547, + "step": 6967 + }, + { + "epoch": 0.27, + "learning_rate": 4.556424724288279e-05, + "loss": 0.5148, + "step": 6968 + }, + { + "epoch": 0.27, + "learning_rate": 4.556360605283406e-05, + "loss": 0.6223, + "step": 6969 + }, + { + "epoch": 0.27, + "learning_rate": 4.556296486278533e-05, + "loss": 0.6428, + "step": 6970 + }, + { + "epoch": 0.27, + "learning_rate": 4.55623236727366e-05, + "loss": 0.4996, + "step": 6971 + }, + { + "epoch": 0.27, + "learning_rate": 4.556168248268787e-05, + "loss": 0.5875, + "step": 6972 + }, + { + "epoch": 0.27, + "learning_rate": 4.556104129263914e-05, + "loss": 0.5632, + "step": 6973 + }, + { + "epoch": 0.27, + "learning_rate": 4.556040010259041e-05, + "loss": 0.581, + "step": 6974 + }, + { + "epoch": 0.27, + "learning_rate": 4.555975891254168e-05, + "loss": 0.5551, + "step": 6975 + }, + { + "epoch": 0.27, + "learning_rate": 4.555911772249295e-05, + "loss": 0.6666, + "step": 6976 + }, + { + "epoch": 0.27, + "learning_rate": 4.5558476532444215e-05, + "loss": 0.5591, + "step": 6977 + }, + { + "epoch": 0.27, + "learning_rate": 4.5557835342395486e-05, + "loss": 0.537, + "step": 6978 + }, + { + "epoch": 0.27, + "learning_rate": 4.555719415234676e-05, + "loss": 0.4858, + "step": 6979 + }, + { + "epoch": 0.27, + "learning_rate": 4.555655296229803e-05, + "loss": 0.7543, + "step": 6980 + }, + { + "epoch": 0.27, + "learning_rate": 4.55559117722493e-05, + "loss": 0.6151, + "step": 6981 + }, + { + "epoch": 0.27, + "learning_rate": 4.5555270582200566e-05, + "loss": 0.6214, + "step": 6982 + }, + { + "epoch": 0.27, + "learning_rate": 4.555462939215184e-05, + "loss": 0.5573, + "step": 6983 + }, + { + "epoch": 0.27, + "learning_rate": 4.55539882021031e-05, + "loss": 0.4926, + "step": 6984 + }, + { + "epoch": 0.27, + "learning_rate": 4.5553347012054375e-05, + "loss": 0.5411, + "step": 6985 + }, + { + "epoch": 0.27, + "learning_rate": 4.555270582200564e-05, + "loss": 0.5845, + "step": 6986 + }, + { + "epoch": 0.27, + "learning_rate": 4.555206463195691e-05, + "loss": 0.7337, + "step": 6987 + }, + { + "epoch": 0.27, + "learning_rate": 4.555142344190818e-05, + "loss": 0.5984, + "step": 6988 + }, + { + "epoch": 0.27, + "learning_rate": 4.5550782251859455e-05, + "loss": 0.5382, + "step": 6989 + }, + { + "epoch": 0.27, + "learning_rate": 4.555014106181073e-05, + "loss": 0.4964, + "step": 6990 + }, + { + "epoch": 0.27, + "learning_rate": 4.554949987176199e-05, + "loss": 0.6744, + "step": 6991 + }, + { + "epoch": 0.27, + "learning_rate": 4.554885868171326e-05, + "loss": 0.685, + "step": 6992 + }, + { + "epoch": 0.27, + "learning_rate": 4.554821749166453e-05, + "loss": 0.5188, + "step": 6993 + }, + { + "epoch": 0.27, + "learning_rate": 4.55475763016158e-05, + "loss": 0.5403, + "step": 6994 + }, + { + "epoch": 0.27, + "learning_rate": 4.554693511156707e-05, + "loss": 0.5501, + "step": 6995 + }, + { + "epoch": 0.27, + "learning_rate": 4.5546293921518343e-05, + "loss": 0.5717, + "step": 6996 + }, + { + "epoch": 0.27, + "learning_rate": 4.5545652731469615e-05, + "loss": 0.5265, + "step": 6997 + }, + { + "epoch": 0.27, + "learning_rate": 4.554501154142088e-05, + "loss": 0.5683, + "step": 6998 + }, + { + "epoch": 0.27, + "learning_rate": 4.554437035137215e-05, + "loss": 0.5895, + "step": 6999 + }, + { + "epoch": 0.27, + "learning_rate": 4.554372916132342e-05, + "loss": 0.5543, + "step": 7000 + }, + { + "epoch": 0.27, + "learning_rate": 4.554308797127469e-05, + "loss": 0.5596, + "step": 7001 + }, + { + "epoch": 0.27, + "learning_rate": 4.554244678122595e-05, + "loss": 0.5354, + "step": 7002 + }, + { + "epoch": 0.27, + "learning_rate": 4.5541805591177225e-05, + "loss": 0.5468, + "step": 7003 + }, + { + "epoch": 0.27, + "learning_rate": 4.55411644011285e-05, + "loss": 0.5754, + "step": 7004 + }, + { + "epoch": 0.27, + "learning_rate": 4.554052321107977e-05, + "loss": 0.4953, + "step": 7005 + }, + { + "epoch": 0.27, + "learning_rate": 4.553988202103104e-05, + "loss": 0.4824, + "step": 7006 + }, + { + "epoch": 0.27, + "learning_rate": 4.5539240830982305e-05, + "loss": 0.666, + "step": 7007 + }, + { + "epoch": 0.27, + "learning_rate": 4.553859964093358e-05, + "loss": 0.6375, + "step": 7008 + }, + { + "epoch": 0.27, + "learning_rate": 4.553795845088484e-05, + "loss": 0.6024, + "step": 7009 + }, + { + "epoch": 0.27, + "learning_rate": 4.5537317260836114e-05, + "loss": 0.5548, + "step": 7010 + }, + { + "epoch": 0.27, + "learning_rate": 4.553667607078738e-05, + "loss": 0.6234, + "step": 7011 + }, + { + "epoch": 0.27, + "learning_rate": 4.553603488073866e-05, + "loss": 0.5383, + "step": 7012 + }, + { + "epoch": 0.27, + "learning_rate": 4.553539369068992e-05, + "loss": 0.6828, + "step": 7013 + }, + { + "epoch": 0.27, + "learning_rate": 4.5534752500641194e-05, + "loss": 0.6409, + "step": 7014 + }, + { + "epoch": 0.27, + "learning_rate": 4.5534111310592465e-05, + "loss": 0.6699, + "step": 7015 + }, + { + "epoch": 0.27, + "learning_rate": 4.553347012054373e-05, + "loss": 0.5641, + "step": 7016 + }, + { + "epoch": 0.27, + "learning_rate": 4.5532828930495e-05, + "loss": 0.5641, + "step": 7017 + }, + { + "epoch": 0.27, + "learning_rate": 4.553218774044627e-05, + "loss": 0.6824, + "step": 7018 + }, + { + "epoch": 0.27, + "learning_rate": 4.553154655039754e-05, + "loss": 0.5664, + "step": 7019 + }, + { + "epoch": 0.27, + "learning_rate": 4.553090536034881e-05, + "loss": 0.5482, + "step": 7020 + }, + { + "epoch": 0.27, + "learning_rate": 4.553026417030008e-05, + "loss": 0.5812, + "step": 7021 + }, + { + "epoch": 0.27, + "learning_rate": 4.552962298025135e-05, + "loss": 0.5591, + "step": 7022 + }, + { + "epoch": 0.27, + "learning_rate": 4.552898179020262e-05, + "loss": 0.5764, + "step": 7023 + }, + { + "epoch": 0.27, + "learning_rate": 4.552834060015389e-05, + "loss": 0.572, + "step": 7024 + }, + { + "epoch": 0.27, + "learning_rate": 4.5527699410105155e-05, + "loss": 0.5523, + "step": 7025 + }, + { + "epoch": 0.27, + "learning_rate": 4.552705822005643e-05, + "loss": 0.6028, + "step": 7026 + }, + { + "epoch": 0.27, + "learning_rate": 4.552641703000769e-05, + "loss": 0.5751, + "step": 7027 + }, + { + "epoch": 0.27, + "learning_rate": 4.552577583995897e-05, + "loss": 0.6174, + "step": 7028 + }, + { + "epoch": 0.27, + "learning_rate": 4.5525134649910236e-05, + "loss": 0.5846, + "step": 7029 + }, + { + "epoch": 0.27, + "learning_rate": 4.552449345986151e-05, + "loss": 0.5816, + "step": 7030 + }, + { + "epoch": 0.27, + "learning_rate": 4.552385226981277e-05, + "loss": 0.6053, + "step": 7031 + }, + { + "epoch": 0.27, + "learning_rate": 4.5523211079764044e-05, + "loss": 0.5805, + "step": 7032 + }, + { + "epoch": 0.27, + "learning_rate": 4.5522569889715316e-05, + "loss": 0.556, + "step": 7033 + }, + { + "epoch": 0.27, + "learning_rate": 4.552192869966658e-05, + "loss": 0.6198, + "step": 7034 + }, + { + "epoch": 0.27, + "learning_rate": 4.552128750961785e-05, + "loss": 0.6297, + "step": 7035 + }, + { + "epoch": 0.27, + "learning_rate": 4.5520646319569124e-05, + "loss": 0.5, + "step": 7036 + }, + { + "epoch": 0.27, + "learning_rate": 4.5520005129520396e-05, + "loss": 0.5285, + "step": 7037 + }, + { + "epoch": 0.27, + "learning_rate": 4.551936393947166e-05, + "loss": 0.5683, + "step": 7038 + }, + { + "epoch": 0.27, + "learning_rate": 4.551872274942293e-05, + "loss": 0.6333, + "step": 7039 + }, + { + "epoch": 0.27, + "learning_rate": 4.55180815593742e-05, + "loss": 0.5335, + "step": 7040 + }, + { + "epoch": 0.27, + "learning_rate": 4.551744036932547e-05, + "loss": 0.5545, + "step": 7041 + }, + { + "epoch": 0.27, + "learning_rate": 4.551679917927674e-05, + "loss": 0.5367, + "step": 7042 + }, + { + "epoch": 0.27, + "learning_rate": 4.5516157989228006e-05, + "loss": 0.5936, + "step": 7043 + }, + { + "epoch": 0.27, + "learning_rate": 4.5515516799179284e-05, + "loss": 0.5226, + "step": 7044 + }, + { + "epoch": 0.27, + "learning_rate": 4.551487560913055e-05, + "loss": 0.4598, + "step": 7045 + }, + { + "epoch": 0.27, + "learning_rate": 4.551423441908182e-05, + "loss": 0.5769, + "step": 7046 + }, + { + "epoch": 0.27, + "learning_rate": 4.5513593229033086e-05, + "loss": 0.5963, + "step": 7047 + }, + { + "epoch": 0.27, + "learning_rate": 4.551295203898436e-05, + "loss": 0.4878, + "step": 7048 + }, + { + "epoch": 0.27, + "learning_rate": 4.551231084893562e-05, + "loss": 0.6382, + "step": 7049 + }, + { + "epoch": 0.27, + "learning_rate": 4.5511669658886894e-05, + "loss": 0.5279, + "step": 7050 + }, + { + "epoch": 0.27, + "learning_rate": 4.5511028468838166e-05, + "loss": 0.5453, + "step": 7051 + }, + { + "epoch": 0.27, + "learning_rate": 4.551038727878944e-05, + "loss": 0.5591, + "step": 7052 + }, + { + "epoch": 0.27, + "learning_rate": 4.550974608874071e-05, + "loss": 0.5784, + "step": 7053 + }, + { + "epoch": 0.27, + "learning_rate": 4.5509104898691974e-05, + "loss": 0.3719, + "step": 7054 + }, + { + "epoch": 0.27, + "learning_rate": 4.5508463708643246e-05, + "loss": 0.574, + "step": 7055 + }, + { + "epoch": 0.27, + "learning_rate": 4.550782251859451e-05, + "loss": 0.6026, + "step": 7056 + }, + { + "epoch": 0.27, + "learning_rate": 4.550718132854578e-05, + "loss": 0.5303, + "step": 7057 + }, + { + "epoch": 0.27, + "learning_rate": 4.550654013849705e-05, + "loss": 0.5146, + "step": 7058 + }, + { + "epoch": 0.27, + "learning_rate": 4.550589894844832e-05, + "loss": 0.549, + "step": 7059 + }, + { + "epoch": 0.27, + "learning_rate": 4.550525775839959e-05, + "loss": 0.4975, + "step": 7060 + }, + { + "epoch": 0.27, + "learning_rate": 4.550461656835086e-05, + "loss": 0.4747, + "step": 7061 + }, + { + "epoch": 0.27, + "learning_rate": 4.5503975378302134e-05, + "loss": 0.6723, + "step": 7062 + }, + { + "epoch": 0.27, + "learning_rate": 4.55033341882534e-05, + "loss": 0.5859, + "step": 7063 + }, + { + "epoch": 0.27, + "learning_rate": 4.550269299820467e-05, + "loss": 0.5512, + "step": 7064 + }, + { + "epoch": 0.27, + "learning_rate": 4.5502051808155936e-05, + "loss": 0.4922, + "step": 7065 + }, + { + "epoch": 0.27, + "learning_rate": 4.550141061810721e-05, + "loss": 0.5731, + "step": 7066 + }, + { + "epoch": 0.27, + "learning_rate": 4.550076942805848e-05, + "loss": 0.5629, + "step": 7067 + }, + { + "epoch": 0.27, + "learning_rate": 4.550012823800975e-05, + "loss": 0.6036, + "step": 7068 + }, + { + "epoch": 0.27, + "learning_rate": 4.549948704796102e-05, + "loss": 0.6346, + "step": 7069 + }, + { + "epoch": 0.27, + "learning_rate": 4.549884585791229e-05, + "loss": 0.6036, + "step": 7070 + }, + { + "epoch": 0.27, + "learning_rate": 4.549820466786356e-05, + "loss": 0.7241, + "step": 7071 + }, + { + "epoch": 0.27, + "learning_rate": 4.5497563477814824e-05, + "loss": 0.6259, + "step": 7072 + }, + { + "epoch": 0.27, + "learning_rate": 4.5496922287766096e-05, + "loss": 0.5513, + "step": 7073 + }, + { + "epoch": 0.27, + "learning_rate": 4.549628109771736e-05, + "loss": 0.6174, + "step": 7074 + }, + { + "epoch": 0.27, + "learning_rate": 4.549563990766863e-05, + "loss": 0.5735, + "step": 7075 + }, + { + "epoch": 0.27, + "learning_rate": 4.5494998717619905e-05, + "loss": 0.4934, + "step": 7076 + }, + { + "epoch": 0.27, + "learning_rate": 4.5494357527571176e-05, + "loss": 0.5915, + "step": 7077 + }, + { + "epoch": 0.27, + "learning_rate": 4.549371633752245e-05, + "loss": 0.6383, + "step": 7078 + }, + { + "epoch": 0.27, + "learning_rate": 4.549307514747371e-05, + "loss": 0.5655, + "step": 7079 + }, + { + "epoch": 0.27, + "learning_rate": 4.5492433957424985e-05, + "loss": 0.5298, + "step": 7080 + }, + { + "epoch": 0.27, + "learning_rate": 4.549179276737625e-05, + "loss": 0.5513, + "step": 7081 + }, + { + "epoch": 0.27, + "learning_rate": 4.549115157732752e-05, + "loss": 0.6554, + "step": 7082 + }, + { + "epoch": 0.27, + "learning_rate": 4.549051038727879e-05, + "loss": 0.5686, + "step": 7083 + }, + { + "epoch": 0.27, + "learning_rate": 4.5489869197230065e-05, + "loss": 0.6024, + "step": 7084 + }, + { + "epoch": 0.27, + "learning_rate": 4.548922800718133e-05, + "loss": 0.4935, + "step": 7085 + }, + { + "epoch": 0.27, + "learning_rate": 4.54885868171326e-05, + "loss": 0.6251, + "step": 7086 + }, + { + "epoch": 0.27, + "learning_rate": 4.548794562708387e-05, + "loss": 0.5964, + "step": 7087 + }, + { + "epoch": 0.27, + "learning_rate": 4.548730443703514e-05, + "loss": 0.5243, + "step": 7088 + }, + { + "epoch": 0.27, + "learning_rate": 4.548666324698641e-05, + "loss": 0.5087, + "step": 7089 + }, + { + "epoch": 0.27, + "learning_rate": 4.5486022056937675e-05, + "loss": 0.6071, + "step": 7090 + }, + { + "epoch": 0.27, + "learning_rate": 4.5485380866888946e-05, + "loss": 0.5545, + "step": 7091 + }, + { + "epoch": 0.27, + "learning_rate": 4.548473967684022e-05, + "loss": 0.5669, + "step": 7092 + }, + { + "epoch": 0.27, + "learning_rate": 4.548409848679149e-05, + "loss": 0.5792, + "step": 7093 + }, + { + "epoch": 0.27, + "learning_rate": 4.5483457296742755e-05, + "loss": 0.5434, + "step": 7094 + }, + { + "epoch": 0.27, + "learning_rate": 4.5482816106694027e-05, + "loss": 0.4819, + "step": 7095 + }, + { + "epoch": 0.27, + "learning_rate": 4.54821749166453e-05, + "loss": 0.6296, + "step": 7096 + }, + { + "epoch": 0.27, + "learning_rate": 4.548153372659656e-05, + "loss": 0.7193, + "step": 7097 + }, + { + "epoch": 0.27, + "learning_rate": 4.5480892536547835e-05, + "loss": 0.5997, + "step": 7098 + }, + { + "epoch": 0.27, + "learning_rate": 4.54802513464991e-05, + "loss": 0.4898, + "step": 7099 + }, + { + "epoch": 0.27, + "learning_rate": 4.547961015645038e-05, + "loss": 0.4908, + "step": 7100 + }, + { + "epoch": 0.27, + "learning_rate": 4.547896896640164e-05, + "loss": 0.5866, + "step": 7101 + }, + { + "epoch": 0.27, + "learning_rate": 4.5478327776352915e-05, + "loss": 0.5966, + "step": 7102 + }, + { + "epoch": 0.27, + "learning_rate": 4.547768658630418e-05, + "loss": 0.5355, + "step": 7103 + }, + { + "epoch": 0.27, + "learning_rate": 4.547704539625545e-05, + "loss": 0.5712, + "step": 7104 + }, + { + "epoch": 0.27, + "learning_rate": 4.547640420620672e-05, + "loss": 0.5587, + "step": 7105 + }, + { + "epoch": 0.27, + "learning_rate": 4.547576301615799e-05, + "loss": 0.643, + "step": 7106 + }, + { + "epoch": 0.27, + "learning_rate": 4.547512182610926e-05, + "loss": 0.6829, + "step": 7107 + }, + { + "epoch": 0.27, + "learning_rate": 4.547448063606053e-05, + "loss": 0.5112, + "step": 7108 + }, + { + "epoch": 0.27, + "learning_rate": 4.5473839446011803e-05, + "loss": 0.5157, + "step": 7109 + }, + { + "epoch": 0.27, + "learning_rate": 4.547319825596307e-05, + "loss": 0.532, + "step": 7110 + }, + { + "epoch": 0.27, + "learning_rate": 4.547255706591434e-05, + "loss": 0.5377, + "step": 7111 + }, + { + "epoch": 0.27, + "learning_rate": 4.5471915875865605e-05, + "loss": 0.5906, + "step": 7112 + }, + { + "epoch": 0.27, + "learning_rate": 4.547127468581688e-05, + "loss": 0.6311, + "step": 7113 + }, + { + "epoch": 0.27, + "learning_rate": 4.547063349576815e-05, + "loss": 0.6408, + "step": 7114 + }, + { + "epoch": 0.27, + "learning_rate": 4.5469992305719413e-05, + "loss": 0.5885, + "step": 7115 + }, + { + "epoch": 0.27, + "learning_rate": 4.546935111567069e-05, + "loss": 0.4992, + "step": 7116 + }, + { + "epoch": 0.27, + "learning_rate": 4.546870992562196e-05, + "loss": 0.5379, + "step": 7117 + }, + { + "epoch": 0.27, + "learning_rate": 4.546806873557323e-05, + "loss": 0.5671, + "step": 7118 + }, + { + "epoch": 0.27, + "learning_rate": 4.5467427545524494e-05, + "loss": 0.6092, + "step": 7119 + }, + { + "epoch": 0.27, + "learning_rate": 4.5466786355475765e-05, + "loss": 0.5542, + "step": 7120 + }, + { + "epoch": 0.27, + "learning_rate": 4.546614516542703e-05, + "loss": 0.5227, + "step": 7121 + }, + { + "epoch": 0.27, + "learning_rate": 4.54655039753783e-05, + "loss": 0.481, + "step": 7122 + }, + { + "epoch": 0.27, + "learning_rate": 4.5464862785329574e-05, + "loss": 0.5879, + "step": 7123 + }, + { + "epoch": 0.27, + "learning_rate": 4.5464221595280845e-05, + "loss": 0.5372, + "step": 7124 + }, + { + "epoch": 0.27, + "learning_rate": 4.546358040523212e-05, + "loss": 0.6259, + "step": 7125 + }, + { + "epoch": 0.27, + "learning_rate": 4.546293921518338e-05, + "loss": 0.5625, + "step": 7126 + }, + { + "epoch": 0.27, + "learning_rate": 4.5462298025134654e-05, + "loss": 0.6364, + "step": 7127 + }, + { + "epoch": 0.27, + "learning_rate": 4.546165683508592e-05, + "loss": 0.6546, + "step": 7128 + }, + { + "epoch": 0.27, + "learning_rate": 4.546101564503719e-05, + "loss": 0.5892, + "step": 7129 + }, + { + "epoch": 0.27, + "learning_rate": 4.5460374454988455e-05, + "loss": 0.5171, + "step": 7130 + }, + { + "epoch": 0.27, + "learning_rate": 4.545973326493973e-05, + "loss": 0.5684, + "step": 7131 + }, + { + "epoch": 0.27, + "learning_rate": 4.5459092074891e-05, + "loss": 0.578, + "step": 7132 + }, + { + "epoch": 0.27, + "learning_rate": 4.545845088484227e-05, + "loss": 0.5475, + "step": 7133 + }, + { + "epoch": 0.27, + "learning_rate": 4.545780969479354e-05, + "loss": 0.525, + "step": 7134 + }, + { + "epoch": 0.27, + "learning_rate": 4.545716850474481e-05, + "loss": 0.5252, + "step": 7135 + }, + { + "epoch": 0.27, + "learning_rate": 4.545652731469608e-05, + "loss": 0.5376, + "step": 7136 + }, + { + "epoch": 0.27, + "learning_rate": 4.5455886124647344e-05, + "loss": 0.6053, + "step": 7137 + }, + { + "epoch": 0.27, + "learning_rate": 4.5455244934598615e-05, + "loss": 0.5119, + "step": 7138 + }, + { + "epoch": 0.27, + "learning_rate": 4.545460374454989e-05, + "loss": 0.5845, + "step": 7139 + }, + { + "epoch": 0.27, + "learning_rate": 4.545396255450116e-05, + "loss": 0.5464, + "step": 7140 + }, + { + "epoch": 0.27, + "learning_rate": 4.545332136445243e-05, + "loss": 0.601, + "step": 7141 + }, + { + "epoch": 0.27, + "learning_rate": 4.5452680174403696e-05, + "loss": 0.4966, + "step": 7142 + }, + { + "epoch": 0.27, + "learning_rate": 4.545203898435497e-05, + "loss": 0.5985, + "step": 7143 + }, + { + "epoch": 0.27, + "learning_rate": 4.545139779430623e-05, + "loss": 0.5781, + "step": 7144 + }, + { + "epoch": 0.27, + "learning_rate": 4.5450756604257504e-05, + "loss": 0.4998, + "step": 7145 + }, + { + "epoch": 0.27, + "learning_rate": 4.545011541420877e-05, + "loss": 0.5469, + "step": 7146 + }, + { + "epoch": 0.27, + "learning_rate": 4.544947422416004e-05, + "loss": 0.5185, + "step": 7147 + }, + { + "epoch": 0.27, + "learning_rate": 4.544883303411131e-05, + "loss": 0.4856, + "step": 7148 + }, + { + "epoch": 0.27, + "learning_rate": 4.5448191844062584e-05, + "loss": 0.598, + "step": 7149 + }, + { + "epoch": 0.27, + "learning_rate": 4.5447550654013856e-05, + "loss": 0.618, + "step": 7150 + }, + { + "epoch": 0.27, + "learning_rate": 4.544690946396512e-05, + "loss": 0.5955, + "step": 7151 + }, + { + "epoch": 0.27, + "learning_rate": 4.544626827391639e-05, + "loss": 0.6584, + "step": 7152 + }, + { + "epoch": 0.28, + "learning_rate": 4.544562708386766e-05, + "loss": 0.5028, + "step": 7153 + }, + { + "epoch": 0.28, + "learning_rate": 4.544498589381893e-05, + "loss": 0.6866, + "step": 7154 + }, + { + "epoch": 0.28, + "learning_rate": 4.54443447037702e-05, + "loss": 0.4972, + "step": 7155 + }, + { + "epoch": 0.28, + "learning_rate": 4.544370351372147e-05, + "loss": 0.5695, + "step": 7156 + }, + { + "epoch": 0.28, + "learning_rate": 4.5443062323672744e-05, + "loss": 0.5974, + "step": 7157 + }, + { + "epoch": 0.28, + "learning_rate": 4.544242113362401e-05, + "loss": 0.5913, + "step": 7158 + }, + { + "epoch": 0.28, + "learning_rate": 4.544177994357528e-05, + "loss": 0.6014, + "step": 7159 + }, + { + "epoch": 0.28, + "learning_rate": 4.5441138753526546e-05, + "loss": 0.6133, + "step": 7160 + }, + { + "epoch": 0.28, + "learning_rate": 4.544049756347782e-05, + "loss": 0.5168, + "step": 7161 + }, + { + "epoch": 0.28, + "learning_rate": 4.543985637342908e-05, + "loss": 0.5338, + "step": 7162 + }, + { + "epoch": 0.28, + "learning_rate": 4.5439215183380354e-05, + "loss": 0.5688, + "step": 7163 + }, + { + "epoch": 0.28, + "learning_rate": 4.5438573993331626e-05, + "loss": 0.5939, + "step": 7164 + }, + { + "epoch": 0.28, + "learning_rate": 4.54379328032829e-05, + "loss": 0.5789, + "step": 7165 + }, + { + "epoch": 0.28, + "learning_rate": 4.543729161323417e-05, + "loss": 0.6248, + "step": 7166 + }, + { + "epoch": 0.28, + "learning_rate": 4.5436650423185434e-05, + "loss": 0.5643, + "step": 7167 + }, + { + "epoch": 0.28, + "learning_rate": 4.5436009233136706e-05, + "loss": 0.6014, + "step": 7168 + }, + { + "epoch": 0.28, + "learning_rate": 4.543536804308797e-05, + "loss": 0.5442, + "step": 7169 + }, + { + "epoch": 0.28, + "learning_rate": 4.543472685303924e-05, + "loss": 0.5649, + "step": 7170 + }, + { + "epoch": 0.28, + "learning_rate": 4.543408566299051e-05, + "loss": 0.581, + "step": 7171 + }, + { + "epoch": 0.28, + "learning_rate": 4.5433444472941786e-05, + "loss": 0.6564, + "step": 7172 + }, + { + "epoch": 0.28, + "learning_rate": 4.543280328289305e-05, + "loss": 0.5541, + "step": 7173 + }, + { + "epoch": 0.28, + "learning_rate": 4.543216209284432e-05, + "loss": 0.6141, + "step": 7174 + }, + { + "epoch": 0.28, + "learning_rate": 4.5431520902795594e-05, + "loss": 0.5035, + "step": 7175 + }, + { + "epoch": 0.28, + "learning_rate": 4.543087971274686e-05, + "loss": 0.4537, + "step": 7176 + }, + { + "epoch": 0.28, + "learning_rate": 4.543023852269813e-05, + "loss": 0.4102, + "step": 7177 + }, + { + "epoch": 0.28, + "learning_rate": 4.5429597332649396e-05, + "loss": 0.5637, + "step": 7178 + }, + { + "epoch": 0.28, + "learning_rate": 4.542895614260067e-05, + "loss": 0.5416, + "step": 7179 + }, + { + "epoch": 0.28, + "learning_rate": 4.542831495255194e-05, + "loss": 0.6245, + "step": 7180 + }, + { + "epoch": 0.28, + "learning_rate": 4.542767376250321e-05, + "loss": 0.5866, + "step": 7181 + }, + { + "epoch": 0.28, + "learning_rate": 4.5427032572454476e-05, + "loss": 0.623, + "step": 7182 + }, + { + "epoch": 0.28, + "learning_rate": 4.542639138240575e-05, + "loss": 0.594, + "step": 7183 + }, + { + "epoch": 0.28, + "learning_rate": 4.542575019235702e-05, + "loss": 0.5881, + "step": 7184 + }, + { + "epoch": 0.28, + "learning_rate": 4.5425109002308285e-05, + "loss": 0.543, + "step": 7185 + }, + { + "epoch": 0.28, + "learning_rate": 4.5424467812259556e-05, + "loss": 0.4257, + "step": 7186 + }, + { + "epoch": 0.28, + "learning_rate": 4.542382662221082e-05, + "loss": 0.5923, + "step": 7187 + }, + { + "epoch": 0.28, + "learning_rate": 4.54231854321621e-05, + "loss": 0.5653, + "step": 7188 + }, + { + "epoch": 0.28, + "learning_rate": 4.5422544242113365e-05, + "loss": 0.4866, + "step": 7189 + }, + { + "epoch": 0.28, + "learning_rate": 4.5421903052064636e-05, + "loss": 0.5343, + "step": 7190 + }, + { + "epoch": 0.28, + "learning_rate": 4.54212618620159e-05, + "loss": 0.5307, + "step": 7191 + }, + { + "epoch": 0.28, + "learning_rate": 4.542062067196717e-05, + "loss": 0.6182, + "step": 7192 + }, + { + "epoch": 0.28, + "learning_rate": 4.5419979481918445e-05, + "loss": 0.5837, + "step": 7193 + }, + { + "epoch": 0.28, + "learning_rate": 4.541933829186971e-05, + "loss": 0.6543, + "step": 7194 + }, + { + "epoch": 0.28, + "learning_rate": 4.541869710182098e-05, + "loss": 0.5315, + "step": 7195 + }, + { + "epoch": 0.28, + "learning_rate": 4.541805591177225e-05, + "loss": 0.6447, + "step": 7196 + }, + { + "epoch": 0.28, + "learning_rate": 4.5417414721723525e-05, + "loss": 0.598, + "step": 7197 + }, + { + "epoch": 0.28, + "learning_rate": 4.541677353167479e-05, + "loss": 0.5991, + "step": 7198 + }, + { + "epoch": 0.28, + "learning_rate": 4.541613234162606e-05, + "loss": 0.678, + "step": 7199 + }, + { + "epoch": 0.28, + "learning_rate": 4.5415491151577326e-05, + "loss": 0.7154, + "step": 7200 + }, + { + "epoch": 0.28, + "learning_rate": 4.54148499615286e-05, + "loss": 0.6278, + "step": 7201 + }, + { + "epoch": 0.28, + "learning_rate": 4.541420877147987e-05, + "loss": 0.6056, + "step": 7202 + }, + { + "epoch": 0.28, + "learning_rate": 4.5413567581431135e-05, + "loss": 0.5628, + "step": 7203 + }, + { + "epoch": 0.28, + "learning_rate": 4.541292639138241e-05, + "loss": 0.5719, + "step": 7204 + }, + { + "epoch": 0.28, + "learning_rate": 4.541228520133368e-05, + "loss": 0.4631, + "step": 7205 + }, + { + "epoch": 0.28, + "learning_rate": 4.541164401128495e-05, + "loss": 0.6487, + "step": 7206 + }, + { + "epoch": 0.28, + "learning_rate": 4.5411002821236215e-05, + "loss": 0.5557, + "step": 7207 + }, + { + "epoch": 0.28, + "learning_rate": 4.5410361631187487e-05, + "loss": 0.5144, + "step": 7208 + }, + { + "epoch": 0.28, + "learning_rate": 4.540972044113875e-05, + "loss": 0.5348, + "step": 7209 + }, + { + "epoch": 0.28, + "learning_rate": 4.540907925109002e-05, + "loss": 0.5354, + "step": 7210 + }, + { + "epoch": 0.28, + "learning_rate": 4.5408438061041295e-05, + "loss": 0.5238, + "step": 7211 + }, + { + "epoch": 0.28, + "learning_rate": 4.540779687099257e-05, + "loss": 0.5949, + "step": 7212 + }, + { + "epoch": 0.28, + "learning_rate": 4.540715568094384e-05, + "loss": 0.6535, + "step": 7213 + }, + { + "epoch": 0.28, + "learning_rate": 4.54065144908951e-05, + "loss": 0.6067, + "step": 7214 + }, + { + "epoch": 0.28, + "learning_rate": 4.5405873300846375e-05, + "loss": 0.5901, + "step": 7215 + }, + { + "epoch": 0.28, + "learning_rate": 4.540523211079764e-05, + "loss": 0.5868, + "step": 7216 + }, + { + "epoch": 0.28, + "learning_rate": 4.540459092074891e-05, + "loss": 0.7009, + "step": 7217 + }, + { + "epoch": 0.28, + "learning_rate": 4.540394973070018e-05, + "loss": 0.5512, + "step": 7218 + }, + { + "epoch": 0.28, + "learning_rate": 4.540330854065145e-05, + "loss": 0.6185, + "step": 7219 + }, + { + "epoch": 0.28, + "learning_rate": 4.540266735060272e-05, + "loss": 0.5238, + "step": 7220 + }, + { + "epoch": 0.28, + "learning_rate": 4.540202616055399e-05, + "loss": 0.5706, + "step": 7221 + }, + { + "epoch": 0.28, + "learning_rate": 4.5401384970505263e-05, + "loss": 0.71, + "step": 7222 + }, + { + "epoch": 0.28, + "learning_rate": 4.540074378045653e-05, + "loss": 0.5516, + "step": 7223 + }, + { + "epoch": 0.28, + "learning_rate": 4.54001025904078e-05, + "loss": 0.6057, + "step": 7224 + }, + { + "epoch": 0.28, + "learning_rate": 4.5399461400359065e-05, + "loss": 0.5123, + "step": 7225 + }, + { + "epoch": 0.28, + "learning_rate": 4.539882021031034e-05, + "loss": 0.6006, + "step": 7226 + }, + { + "epoch": 0.28, + "learning_rate": 4.539817902026161e-05, + "loss": 0.656, + "step": 7227 + }, + { + "epoch": 0.28, + "learning_rate": 4.539753783021288e-05, + "loss": 0.4964, + "step": 7228 + }, + { + "epoch": 0.28, + "learning_rate": 4.539689664016415e-05, + "loss": 0.6367, + "step": 7229 + }, + { + "epoch": 0.28, + "learning_rate": 4.539625545011542e-05, + "loss": 0.5457, + "step": 7230 + }, + { + "epoch": 0.28, + "learning_rate": 4.539561426006669e-05, + "loss": 0.5942, + "step": 7231 + }, + { + "epoch": 0.28, + "learning_rate": 4.5394973070017954e-05, + "loss": 0.5808, + "step": 7232 + }, + { + "epoch": 0.28, + "learning_rate": 4.5394331879969225e-05, + "loss": 0.559, + "step": 7233 + }, + { + "epoch": 0.28, + "learning_rate": 4.539369068992049e-05, + "loss": 0.53, + "step": 7234 + }, + { + "epoch": 0.28, + "learning_rate": 4.539304949987176e-05, + "loss": 0.6106, + "step": 7235 + }, + { + "epoch": 0.28, + "learning_rate": 4.5392408309823034e-05, + "loss": 0.6069, + "step": 7236 + }, + { + "epoch": 0.28, + "learning_rate": 4.5391767119774305e-05, + "loss": 0.5302, + "step": 7237 + }, + { + "epoch": 0.28, + "learning_rate": 4.539112592972558e-05, + "loss": 0.6562, + "step": 7238 + }, + { + "epoch": 0.28, + "learning_rate": 4.539048473967684e-05, + "loss": 0.5659, + "step": 7239 + }, + { + "epoch": 0.28, + "learning_rate": 4.5389843549628114e-05, + "loss": 0.4893, + "step": 7240 + }, + { + "epoch": 0.28, + "learning_rate": 4.538920235957938e-05, + "loss": 0.5064, + "step": 7241 + }, + { + "epoch": 0.28, + "learning_rate": 4.538856116953065e-05, + "loss": 0.5024, + "step": 7242 + }, + { + "epoch": 0.28, + "learning_rate": 4.538791997948192e-05, + "loss": 0.5183, + "step": 7243 + }, + { + "epoch": 0.28, + "learning_rate": 4.5387278789433194e-05, + "loss": 0.6433, + "step": 7244 + }, + { + "epoch": 0.28, + "learning_rate": 4.538663759938446e-05, + "loss": 0.5972, + "step": 7245 + }, + { + "epoch": 0.28, + "learning_rate": 4.538599640933573e-05, + "loss": 0.5406, + "step": 7246 + }, + { + "epoch": 0.28, + "learning_rate": 4.5385355219287e-05, + "loss": 0.5528, + "step": 7247 + }, + { + "epoch": 0.28, + "learning_rate": 4.538471402923827e-05, + "loss": 0.6045, + "step": 7248 + }, + { + "epoch": 0.28, + "learning_rate": 4.538407283918954e-05, + "loss": 0.608, + "step": 7249 + }, + { + "epoch": 0.28, + "learning_rate": 4.5383431649140804e-05, + "loss": 0.5358, + "step": 7250 + }, + { + "epoch": 0.28, + "learning_rate": 4.5382790459092076e-05, + "loss": 0.5819, + "step": 7251 + }, + { + "epoch": 0.28, + "learning_rate": 4.538214926904335e-05, + "loss": 0.4508, + "step": 7252 + }, + { + "epoch": 0.28, + "learning_rate": 4.538150807899462e-05, + "loss": 0.6494, + "step": 7253 + }, + { + "epoch": 0.28, + "learning_rate": 4.5380866888945884e-05, + "loss": 0.7168, + "step": 7254 + }, + { + "epoch": 0.28, + "learning_rate": 4.5380225698897156e-05, + "loss": 0.5483, + "step": 7255 + }, + { + "epoch": 0.28, + "learning_rate": 4.537958450884843e-05, + "loss": 0.6005, + "step": 7256 + }, + { + "epoch": 0.28, + "learning_rate": 4.537894331879969e-05, + "loss": 0.5309, + "step": 7257 + }, + { + "epoch": 0.28, + "learning_rate": 4.5378302128750964e-05, + "loss": 0.7358, + "step": 7258 + }, + { + "epoch": 0.28, + "learning_rate": 4.537766093870223e-05, + "loss": 0.5189, + "step": 7259 + }, + { + "epoch": 0.28, + "learning_rate": 4.537701974865351e-05, + "loss": 0.5226, + "step": 7260 + }, + { + "epoch": 0.28, + "learning_rate": 4.537637855860477e-05, + "loss": 0.5522, + "step": 7261 + }, + { + "epoch": 0.28, + "learning_rate": 4.5375737368556044e-05, + "loss": 0.6329, + "step": 7262 + }, + { + "epoch": 0.28, + "learning_rate": 4.537509617850731e-05, + "loss": 0.6328, + "step": 7263 + }, + { + "epoch": 0.28, + "learning_rate": 4.537445498845858e-05, + "loss": 0.6598, + "step": 7264 + }, + { + "epoch": 0.28, + "learning_rate": 4.537381379840985e-05, + "loss": 0.5376, + "step": 7265 + }, + { + "epoch": 0.28, + "learning_rate": 4.537317260836112e-05, + "loss": 0.6139, + "step": 7266 + }, + { + "epoch": 0.28, + "learning_rate": 4.537253141831239e-05, + "loss": 0.5427, + "step": 7267 + }, + { + "epoch": 0.28, + "learning_rate": 4.537189022826366e-05, + "loss": 0.5962, + "step": 7268 + }, + { + "epoch": 0.28, + "learning_rate": 4.537124903821493e-05, + "loss": 0.6982, + "step": 7269 + }, + { + "epoch": 0.28, + "learning_rate": 4.53706078481662e-05, + "loss": 0.4955, + "step": 7270 + }, + { + "epoch": 0.28, + "learning_rate": 4.536996665811747e-05, + "loss": 0.5683, + "step": 7271 + }, + { + "epoch": 0.28, + "learning_rate": 4.5369325468068734e-05, + "loss": 0.5666, + "step": 7272 + }, + { + "epoch": 0.28, + "learning_rate": 4.5368684278020006e-05, + "loss": 0.6161, + "step": 7273 + }, + { + "epoch": 0.28, + "learning_rate": 4.536804308797128e-05, + "loss": 0.5775, + "step": 7274 + }, + { + "epoch": 0.28, + "learning_rate": 4.536740189792254e-05, + "loss": 0.5483, + "step": 7275 + }, + { + "epoch": 0.28, + "learning_rate": 4.536676070787382e-05, + "loss": 0.543, + "step": 7276 + }, + { + "epoch": 0.28, + "learning_rate": 4.5366119517825086e-05, + "loss": 0.5874, + "step": 7277 + }, + { + "epoch": 0.28, + "learning_rate": 4.536547832777636e-05, + "loss": 0.6751, + "step": 7278 + }, + { + "epoch": 0.28, + "learning_rate": 4.536483713772762e-05, + "loss": 0.5656, + "step": 7279 + }, + { + "epoch": 0.28, + "learning_rate": 4.5364195947678894e-05, + "loss": 0.5923, + "step": 7280 + }, + { + "epoch": 0.28, + "learning_rate": 4.536355475763016e-05, + "loss": 0.5253, + "step": 7281 + }, + { + "epoch": 0.28, + "learning_rate": 4.536291356758143e-05, + "loss": 0.53, + "step": 7282 + }, + { + "epoch": 0.28, + "learning_rate": 4.53622723775327e-05, + "loss": 0.5556, + "step": 7283 + }, + { + "epoch": 0.28, + "learning_rate": 4.5361631187483974e-05, + "loss": 0.5476, + "step": 7284 + }, + { + "epoch": 0.28, + "learning_rate": 4.5360989997435246e-05, + "loss": 0.6413, + "step": 7285 + }, + { + "epoch": 0.28, + "learning_rate": 4.536034880738651e-05, + "loss": 0.5202, + "step": 7286 + }, + { + "epoch": 0.28, + "learning_rate": 4.535970761733778e-05, + "loss": 0.5718, + "step": 7287 + }, + { + "epoch": 0.28, + "learning_rate": 4.535906642728905e-05, + "loss": 0.5313, + "step": 7288 + }, + { + "epoch": 0.28, + "learning_rate": 4.535842523724032e-05, + "loss": 0.5985, + "step": 7289 + }, + { + "epoch": 0.28, + "learning_rate": 4.5357784047191584e-05, + "loss": 0.5327, + "step": 7290 + }, + { + "epoch": 0.28, + "learning_rate": 4.5357142857142856e-05, + "loss": 0.6899, + "step": 7291 + }, + { + "epoch": 0.28, + "learning_rate": 4.535650166709413e-05, + "loss": 0.5384, + "step": 7292 + }, + { + "epoch": 0.28, + "learning_rate": 4.53558604770454e-05, + "loss": 0.5043, + "step": 7293 + }, + { + "epoch": 0.28, + "learning_rate": 4.535521928699667e-05, + "loss": 0.619, + "step": 7294 + }, + { + "epoch": 0.28, + "learning_rate": 4.5354578096947936e-05, + "loss": 0.6472, + "step": 7295 + }, + { + "epoch": 0.28, + "learning_rate": 4.535393690689921e-05, + "loss": 0.5948, + "step": 7296 + }, + { + "epoch": 0.28, + "learning_rate": 4.535329571685047e-05, + "loss": 0.5814, + "step": 7297 + }, + { + "epoch": 0.28, + "learning_rate": 4.5352654526801745e-05, + "loss": 0.5535, + "step": 7298 + }, + { + "epoch": 0.28, + "learning_rate": 4.5352013336753016e-05, + "loss": 0.5865, + "step": 7299 + }, + { + "epoch": 0.28, + "learning_rate": 4.535137214670429e-05, + "loss": 0.5805, + "step": 7300 + }, + { + "epoch": 0.28, + "learning_rate": 4.535073095665556e-05, + "loss": 0.5825, + "step": 7301 + }, + { + "epoch": 0.28, + "learning_rate": 4.5350089766606825e-05, + "loss": 0.4864, + "step": 7302 + }, + { + "epoch": 0.28, + "learning_rate": 4.5349448576558096e-05, + "loss": 0.5393, + "step": 7303 + }, + { + "epoch": 0.28, + "learning_rate": 4.534880738650936e-05, + "loss": 0.5416, + "step": 7304 + }, + { + "epoch": 0.28, + "learning_rate": 4.534816619646063e-05, + "loss": 0.5713, + "step": 7305 + }, + { + "epoch": 0.28, + "learning_rate": 4.53475250064119e-05, + "loss": 0.479, + "step": 7306 + }, + { + "epoch": 0.28, + "learning_rate": 4.534688381636317e-05, + "loss": 0.5752, + "step": 7307 + }, + { + "epoch": 0.28, + "learning_rate": 4.534624262631444e-05, + "loss": 0.5703, + "step": 7308 + }, + { + "epoch": 0.28, + "learning_rate": 4.534560143626571e-05, + "loss": 0.6387, + "step": 7309 + }, + { + "epoch": 0.28, + "learning_rate": 4.5344960246216985e-05, + "loss": 0.5797, + "step": 7310 + }, + { + "epoch": 0.28, + "learning_rate": 4.534431905616825e-05, + "loss": 0.563, + "step": 7311 + }, + { + "epoch": 0.28, + "learning_rate": 4.534367786611952e-05, + "loss": 0.6327, + "step": 7312 + }, + { + "epoch": 0.28, + "learning_rate": 4.5343036676070786e-05, + "loss": 0.6546, + "step": 7313 + }, + { + "epoch": 0.28, + "learning_rate": 4.534239548602206e-05, + "loss": 0.5625, + "step": 7314 + }, + { + "epoch": 0.28, + "learning_rate": 4.534175429597333e-05, + "loss": 0.5712, + "step": 7315 + }, + { + "epoch": 0.28, + "learning_rate": 4.53411131059246e-05, + "loss": 0.4828, + "step": 7316 + }, + { + "epoch": 0.28, + "learning_rate": 4.5340471915875866e-05, + "loss": 0.6361, + "step": 7317 + }, + { + "epoch": 0.28, + "learning_rate": 4.533983072582714e-05, + "loss": 0.5785, + "step": 7318 + }, + { + "epoch": 0.28, + "learning_rate": 4.533918953577841e-05, + "loss": 0.6235, + "step": 7319 + }, + { + "epoch": 0.28, + "learning_rate": 4.5338548345729675e-05, + "loss": 0.4809, + "step": 7320 + }, + { + "epoch": 0.28, + "learning_rate": 4.5337907155680947e-05, + "loss": 0.5284, + "step": 7321 + }, + { + "epoch": 0.28, + "learning_rate": 4.533726596563221e-05, + "loss": 0.5713, + "step": 7322 + }, + { + "epoch": 0.28, + "learning_rate": 4.533662477558348e-05, + "loss": 0.584, + "step": 7323 + }, + { + "epoch": 0.28, + "learning_rate": 4.5335983585534755e-05, + "loss": 0.5526, + "step": 7324 + }, + { + "epoch": 0.28, + "learning_rate": 4.533534239548603e-05, + "loss": 0.5924, + "step": 7325 + }, + { + "epoch": 0.28, + "learning_rate": 4.533470120543729e-05, + "loss": 0.4309, + "step": 7326 + }, + { + "epoch": 0.28, + "learning_rate": 4.533406001538856e-05, + "loss": 0.5369, + "step": 7327 + }, + { + "epoch": 0.28, + "learning_rate": 4.5333418825339835e-05, + "loss": 0.5124, + "step": 7328 + }, + { + "epoch": 0.28, + "learning_rate": 4.53327776352911e-05, + "loss": 0.4848, + "step": 7329 + }, + { + "epoch": 0.28, + "learning_rate": 4.533213644524237e-05, + "loss": 0.5628, + "step": 7330 + }, + { + "epoch": 0.28, + "learning_rate": 4.5331495255193643e-05, + "loss": 0.6139, + "step": 7331 + }, + { + "epoch": 0.28, + "learning_rate": 4.5330854065144915e-05, + "loss": 0.6209, + "step": 7332 + }, + { + "epoch": 0.28, + "learning_rate": 4.533021287509618e-05, + "loss": 0.5658, + "step": 7333 + }, + { + "epoch": 0.28, + "learning_rate": 4.532957168504745e-05, + "loss": 0.6258, + "step": 7334 + }, + { + "epoch": 0.28, + "learning_rate": 4.532893049499872e-05, + "loss": 0.4958, + "step": 7335 + }, + { + "epoch": 0.28, + "learning_rate": 4.532828930494999e-05, + "loss": 0.5573, + "step": 7336 + }, + { + "epoch": 0.28, + "learning_rate": 4.532764811490126e-05, + "loss": 0.5281, + "step": 7337 + }, + { + "epoch": 0.28, + "learning_rate": 4.5327006924852525e-05, + "loss": 0.568, + "step": 7338 + }, + { + "epoch": 0.28, + "learning_rate": 4.53263657348038e-05, + "loss": 0.5271, + "step": 7339 + }, + { + "epoch": 0.28, + "learning_rate": 4.532572454475507e-05, + "loss": 0.5829, + "step": 7340 + }, + { + "epoch": 0.28, + "learning_rate": 4.532508335470634e-05, + "loss": 0.5978, + "step": 7341 + }, + { + "epoch": 0.28, + "learning_rate": 4.5324442164657605e-05, + "loss": 0.5254, + "step": 7342 + }, + { + "epoch": 0.28, + "learning_rate": 4.532380097460888e-05, + "loss": 0.5207, + "step": 7343 + }, + { + "epoch": 0.28, + "learning_rate": 4.532315978456015e-05, + "loss": 0.5954, + "step": 7344 + }, + { + "epoch": 0.28, + "learning_rate": 4.5322518594511414e-05, + "loss": 0.5934, + "step": 7345 + }, + { + "epoch": 0.28, + "learning_rate": 4.5321877404462685e-05, + "loss": 0.6337, + "step": 7346 + }, + { + "epoch": 0.28, + "learning_rate": 4.532123621441395e-05, + "loss": 0.5366, + "step": 7347 + }, + { + "epoch": 0.28, + "learning_rate": 4.532059502436523e-05, + "loss": 0.6181, + "step": 7348 + }, + { + "epoch": 0.28, + "learning_rate": 4.5319953834316494e-05, + "loss": 0.5372, + "step": 7349 + }, + { + "epoch": 0.28, + "learning_rate": 4.5319312644267765e-05, + "loss": 0.5067, + "step": 7350 + }, + { + "epoch": 0.28, + "learning_rate": 4.531867145421903e-05, + "loss": 0.566, + "step": 7351 + }, + { + "epoch": 0.28, + "learning_rate": 4.53180302641703e-05, + "loss": 0.5917, + "step": 7352 + }, + { + "epoch": 0.28, + "learning_rate": 4.5317389074121574e-05, + "loss": 0.5975, + "step": 7353 + }, + { + "epoch": 0.28, + "learning_rate": 4.531674788407284e-05, + "loss": 0.6303, + "step": 7354 + }, + { + "epoch": 0.28, + "learning_rate": 4.531610669402411e-05, + "loss": 0.6555, + "step": 7355 + }, + { + "epoch": 0.28, + "learning_rate": 4.531546550397538e-05, + "loss": 0.5823, + "step": 7356 + }, + { + "epoch": 0.28, + "learning_rate": 4.5314824313926654e-05, + "loss": 0.6033, + "step": 7357 + }, + { + "epoch": 0.28, + "learning_rate": 4.531418312387792e-05, + "loss": 0.613, + "step": 7358 + }, + { + "epoch": 0.28, + "learning_rate": 4.531354193382919e-05, + "loss": 0.5161, + "step": 7359 + }, + { + "epoch": 0.28, + "learning_rate": 4.5312900743780455e-05, + "loss": 0.6071, + "step": 7360 + }, + { + "epoch": 0.28, + "learning_rate": 4.531225955373173e-05, + "loss": 0.6292, + "step": 7361 + }, + { + "epoch": 0.28, + "learning_rate": 4.5311618363683e-05, + "loss": 0.5817, + "step": 7362 + }, + { + "epoch": 0.28, + "learning_rate": 4.5310977173634264e-05, + "loss": 0.5767, + "step": 7363 + }, + { + "epoch": 0.28, + "learning_rate": 4.531033598358554e-05, + "loss": 0.5294, + "step": 7364 + }, + { + "epoch": 0.28, + "learning_rate": 4.530969479353681e-05, + "loss": 0.5453, + "step": 7365 + }, + { + "epoch": 0.28, + "learning_rate": 4.530905360348808e-05, + "loss": 0.5922, + "step": 7366 + }, + { + "epoch": 0.28, + "learning_rate": 4.5308412413439344e-05, + "loss": 0.6194, + "step": 7367 + }, + { + "epoch": 0.28, + "learning_rate": 4.5307771223390616e-05, + "loss": 0.5634, + "step": 7368 + }, + { + "epoch": 0.28, + "learning_rate": 4.530713003334188e-05, + "loss": 0.4897, + "step": 7369 + }, + { + "epoch": 0.28, + "learning_rate": 4.530648884329315e-05, + "loss": 0.588, + "step": 7370 + }, + { + "epoch": 0.28, + "learning_rate": 4.5305847653244424e-05, + "loss": 0.5713, + "step": 7371 + }, + { + "epoch": 0.28, + "learning_rate": 4.5305206463195696e-05, + "loss": 0.535, + "step": 7372 + }, + { + "epoch": 0.28, + "learning_rate": 4.530456527314697e-05, + "loss": 0.5542, + "step": 7373 + }, + { + "epoch": 0.28, + "learning_rate": 4.530392408309823e-05, + "loss": 0.4981, + "step": 7374 + }, + { + "epoch": 0.28, + "learning_rate": 4.5303282893049504e-05, + "loss": 0.5515, + "step": 7375 + }, + { + "epoch": 0.28, + "learning_rate": 4.530264170300077e-05, + "loss": 0.557, + "step": 7376 + }, + { + "epoch": 0.28, + "learning_rate": 4.530200051295204e-05, + "loss": 0.575, + "step": 7377 + }, + { + "epoch": 0.28, + "learning_rate": 4.5301359322903306e-05, + "loss": 0.6222, + "step": 7378 + }, + { + "epoch": 0.28, + "learning_rate": 4.530071813285458e-05, + "loss": 0.4371, + "step": 7379 + }, + { + "epoch": 0.28, + "learning_rate": 4.530007694280585e-05, + "loss": 0.6317, + "step": 7380 + }, + { + "epoch": 0.28, + "learning_rate": 4.529943575275712e-05, + "loss": 0.6773, + "step": 7381 + }, + { + "epoch": 0.28, + "learning_rate": 4.529879456270839e-05, + "loss": 0.5117, + "step": 7382 + }, + { + "epoch": 0.28, + "learning_rate": 4.529815337265966e-05, + "loss": 0.5504, + "step": 7383 + }, + { + "epoch": 0.28, + "learning_rate": 4.529751218261093e-05, + "loss": 0.526, + "step": 7384 + }, + { + "epoch": 0.28, + "learning_rate": 4.5296870992562194e-05, + "loss": 0.5792, + "step": 7385 + }, + { + "epoch": 0.28, + "learning_rate": 4.5296229802513466e-05, + "loss": 0.6137, + "step": 7386 + }, + { + "epoch": 0.28, + "learning_rate": 4.529558861246474e-05, + "loss": 0.6762, + "step": 7387 + }, + { + "epoch": 0.28, + "learning_rate": 4.529494742241601e-05, + "loss": 0.5406, + "step": 7388 + }, + { + "epoch": 0.28, + "learning_rate": 4.529430623236728e-05, + "loss": 0.6943, + "step": 7389 + }, + { + "epoch": 0.28, + "learning_rate": 4.5293665042318546e-05, + "loss": 0.6579, + "step": 7390 + }, + { + "epoch": 0.28, + "learning_rate": 4.529302385226982e-05, + "loss": 0.5847, + "step": 7391 + }, + { + "epoch": 0.28, + "learning_rate": 4.529238266222108e-05, + "loss": 0.6231, + "step": 7392 + }, + { + "epoch": 0.28, + "learning_rate": 4.5291741472172354e-05, + "loss": 0.6056, + "step": 7393 + }, + { + "epoch": 0.28, + "learning_rate": 4.529110028212362e-05, + "loss": 0.7549, + "step": 7394 + }, + { + "epoch": 0.28, + "learning_rate": 4.529045909207489e-05, + "loss": 0.5563, + "step": 7395 + }, + { + "epoch": 0.28, + "learning_rate": 4.528981790202616e-05, + "loss": 0.603, + "step": 7396 + }, + { + "epoch": 0.28, + "learning_rate": 4.5289176711977434e-05, + "loss": 0.634, + "step": 7397 + }, + { + "epoch": 0.28, + "learning_rate": 4.5288535521928706e-05, + "loss": 0.5438, + "step": 7398 + }, + { + "epoch": 0.28, + "learning_rate": 4.528789433187997e-05, + "loss": 0.5278, + "step": 7399 + }, + { + "epoch": 0.28, + "learning_rate": 4.528725314183124e-05, + "loss": 0.5222, + "step": 7400 + }, + { + "epoch": 0.28, + "learning_rate": 4.528661195178251e-05, + "loss": 0.5344, + "step": 7401 + }, + { + "epoch": 0.28, + "learning_rate": 4.528597076173378e-05, + "loss": 0.6177, + "step": 7402 + }, + { + "epoch": 0.28, + "learning_rate": 4.528532957168505e-05, + "loss": 0.6158, + "step": 7403 + }, + { + "epoch": 0.28, + "learning_rate": 4.528468838163632e-05, + "loss": 0.5984, + "step": 7404 + }, + { + "epoch": 0.28, + "learning_rate": 4.528404719158759e-05, + "loss": 0.5476, + "step": 7405 + }, + { + "epoch": 0.28, + "learning_rate": 4.528340600153886e-05, + "loss": 0.5719, + "step": 7406 + }, + { + "epoch": 0.28, + "learning_rate": 4.528276481149013e-05, + "loss": 0.6045, + "step": 7407 + }, + { + "epoch": 0.28, + "learning_rate": 4.5282123621441396e-05, + "loss": 0.6157, + "step": 7408 + }, + { + "epoch": 0.28, + "learning_rate": 4.528148243139267e-05, + "loss": 0.5987, + "step": 7409 + }, + { + "epoch": 0.28, + "learning_rate": 4.528084124134393e-05, + "loss": 0.5609, + "step": 7410 + }, + { + "epoch": 0.28, + "learning_rate": 4.5280200051295205e-05, + "loss": 0.5997, + "step": 7411 + }, + { + "epoch": 0.28, + "learning_rate": 4.5279558861246476e-05, + "loss": 0.5539, + "step": 7412 + }, + { + "epoch": 0.29, + "learning_rate": 4.527891767119775e-05, + "loss": 0.5688, + "step": 7413 + }, + { + "epoch": 0.29, + "learning_rate": 4.527827648114901e-05, + "loss": 0.6197, + "step": 7414 + }, + { + "epoch": 0.29, + "learning_rate": 4.5277635291100285e-05, + "loss": 0.6477, + "step": 7415 + }, + { + "epoch": 0.29, + "learning_rate": 4.5276994101051556e-05, + "loss": 0.6178, + "step": 7416 + }, + { + "epoch": 0.29, + "learning_rate": 4.527635291100282e-05, + "loss": 0.5289, + "step": 7417 + }, + { + "epoch": 0.29, + "learning_rate": 4.527571172095409e-05, + "loss": 0.549, + "step": 7418 + }, + { + "epoch": 0.29, + "learning_rate": 4.527507053090536e-05, + "loss": 0.5786, + "step": 7419 + }, + { + "epoch": 0.29, + "learning_rate": 4.5274429340856636e-05, + "loss": 0.5774, + "step": 7420 + }, + { + "epoch": 0.29, + "learning_rate": 4.52737881508079e-05, + "loss": 0.6161, + "step": 7421 + }, + { + "epoch": 0.29, + "learning_rate": 4.527314696075917e-05, + "loss": 0.5844, + "step": 7422 + }, + { + "epoch": 0.29, + "learning_rate": 4.527250577071044e-05, + "loss": 0.6773, + "step": 7423 + }, + { + "epoch": 0.29, + "learning_rate": 4.527186458066171e-05, + "loss": 0.6065, + "step": 7424 + }, + { + "epoch": 0.29, + "learning_rate": 4.527122339061298e-05, + "loss": 0.5592, + "step": 7425 + }, + { + "epoch": 0.29, + "learning_rate": 4.5270582200564246e-05, + "loss": 0.6306, + "step": 7426 + }, + { + "epoch": 0.29, + "learning_rate": 4.526994101051552e-05, + "loss": 0.5115, + "step": 7427 + }, + { + "epoch": 0.29, + "learning_rate": 4.526929982046679e-05, + "loss": 0.5315, + "step": 7428 + }, + { + "epoch": 0.29, + "learning_rate": 4.526865863041806e-05, + "loss": 0.6457, + "step": 7429 + }, + { + "epoch": 0.29, + "learning_rate": 4.5268017440369327e-05, + "loss": 0.576, + "step": 7430 + }, + { + "epoch": 0.29, + "learning_rate": 4.52673762503206e-05, + "loss": 0.5489, + "step": 7431 + }, + { + "epoch": 0.29, + "learning_rate": 4.526673506027186e-05, + "loss": 0.6051, + "step": 7432 + }, + { + "epoch": 0.29, + "learning_rate": 4.5266093870223135e-05, + "loss": 0.6207, + "step": 7433 + }, + { + "epoch": 0.29, + "learning_rate": 4.526545268017441e-05, + "loss": 0.5748, + "step": 7434 + }, + { + "epoch": 0.29, + "learning_rate": 4.526481149012567e-05, + "loss": 0.6501, + "step": 7435 + }, + { + "epoch": 0.29, + "learning_rate": 4.526417030007695e-05, + "loss": 0.5544, + "step": 7436 + }, + { + "epoch": 0.29, + "learning_rate": 4.5263529110028215e-05, + "loss": 0.5741, + "step": 7437 + }, + { + "epoch": 0.29, + "learning_rate": 4.526288791997949e-05, + "loss": 0.5289, + "step": 7438 + }, + { + "epoch": 0.29, + "learning_rate": 4.526224672993075e-05, + "loss": 0.6188, + "step": 7439 + }, + { + "epoch": 0.29, + "learning_rate": 4.526160553988202e-05, + "loss": 0.636, + "step": 7440 + }, + { + "epoch": 0.29, + "learning_rate": 4.526096434983329e-05, + "loss": 0.607, + "step": 7441 + }, + { + "epoch": 0.29, + "learning_rate": 4.526032315978456e-05, + "loss": 0.5807, + "step": 7442 + }, + { + "epoch": 0.29, + "learning_rate": 4.525968196973583e-05, + "loss": 0.5293, + "step": 7443 + }, + { + "epoch": 0.29, + "learning_rate": 4.5259040779687103e-05, + "loss": 0.6146, + "step": 7444 + }, + { + "epoch": 0.29, + "learning_rate": 4.5258399589638375e-05, + "loss": 0.4781, + "step": 7445 + }, + { + "epoch": 0.29, + "learning_rate": 4.525775839958964e-05, + "loss": 0.6047, + "step": 7446 + }, + { + "epoch": 0.29, + "learning_rate": 4.525711720954091e-05, + "loss": 0.5748, + "step": 7447 + }, + { + "epoch": 0.29, + "learning_rate": 4.525647601949218e-05, + "loss": 0.5843, + "step": 7448 + }, + { + "epoch": 0.29, + "learning_rate": 4.525583482944345e-05, + "loss": 0.5922, + "step": 7449 + }, + { + "epoch": 0.29, + "learning_rate": 4.5255193639394713e-05, + "loss": 0.5603, + "step": 7450 + }, + { + "epoch": 0.29, + "learning_rate": 4.5254552449345985e-05, + "loss": 0.5082, + "step": 7451 + }, + { + "epoch": 0.29, + "learning_rate": 4.5253911259297264e-05, + "loss": 0.5638, + "step": 7452 + }, + { + "epoch": 0.29, + "learning_rate": 4.525327006924853e-05, + "loss": 0.6153, + "step": 7453 + }, + { + "epoch": 0.29, + "learning_rate": 4.52526288791998e-05, + "loss": 0.5776, + "step": 7454 + }, + { + "epoch": 0.29, + "learning_rate": 4.5251987689151065e-05, + "loss": 0.5122, + "step": 7455 + }, + { + "epoch": 0.29, + "learning_rate": 4.525134649910234e-05, + "loss": 0.6915, + "step": 7456 + }, + { + "epoch": 0.29, + "learning_rate": 4.52507053090536e-05, + "loss": 0.5891, + "step": 7457 + }, + { + "epoch": 0.29, + "learning_rate": 4.5250064119004874e-05, + "loss": 0.5045, + "step": 7458 + }, + { + "epoch": 0.29, + "learning_rate": 4.5249422928956145e-05, + "loss": 0.6013, + "step": 7459 + }, + { + "epoch": 0.29, + "learning_rate": 4.524878173890742e-05, + "loss": 0.6201, + "step": 7460 + }, + { + "epoch": 0.29, + "learning_rate": 4.524814054885869e-05, + "loss": 0.6604, + "step": 7461 + }, + { + "epoch": 0.29, + "learning_rate": 4.5247499358809954e-05, + "loss": 0.5317, + "step": 7462 + }, + { + "epoch": 0.29, + "learning_rate": 4.5246858168761225e-05, + "loss": 0.6233, + "step": 7463 + }, + { + "epoch": 0.29, + "learning_rate": 4.524621697871249e-05, + "loss": 0.6313, + "step": 7464 + }, + { + "epoch": 0.29, + "learning_rate": 4.524557578866376e-05, + "loss": 0.6023, + "step": 7465 + }, + { + "epoch": 0.29, + "learning_rate": 4.524493459861503e-05, + "loss": 0.5765, + "step": 7466 + }, + { + "epoch": 0.29, + "learning_rate": 4.52442934085663e-05, + "loss": 0.5698, + "step": 7467 + }, + { + "epoch": 0.29, + "learning_rate": 4.524365221851757e-05, + "loss": 0.5995, + "step": 7468 + }, + { + "epoch": 0.29, + "learning_rate": 4.524301102846884e-05, + "loss": 0.5608, + "step": 7469 + }, + { + "epoch": 0.29, + "learning_rate": 4.5242369838420114e-05, + "loss": 0.5373, + "step": 7470 + }, + { + "epoch": 0.29, + "learning_rate": 4.524172864837138e-05, + "loss": 0.5766, + "step": 7471 + }, + { + "epoch": 0.29, + "learning_rate": 4.524108745832265e-05, + "loss": 0.5292, + "step": 7472 + }, + { + "epoch": 0.29, + "learning_rate": 4.5240446268273915e-05, + "loss": 0.5876, + "step": 7473 + }, + { + "epoch": 0.29, + "learning_rate": 4.523980507822519e-05, + "loss": 0.5386, + "step": 7474 + }, + { + "epoch": 0.29, + "learning_rate": 4.523916388817646e-05, + "loss": 0.4582, + "step": 7475 + }, + { + "epoch": 0.29, + "learning_rate": 4.523852269812773e-05, + "loss": 0.5699, + "step": 7476 + }, + { + "epoch": 0.29, + "learning_rate": 4.5237881508078996e-05, + "loss": 0.6113, + "step": 7477 + }, + { + "epoch": 0.29, + "learning_rate": 4.523724031803027e-05, + "loss": 0.6432, + "step": 7478 + }, + { + "epoch": 0.29, + "learning_rate": 4.523659912798154e-05, + "loss": 0.5711, + "step": 7479 + }, + { + "epoch": 0.29, + "learning_rate": 4.5235957937932804e-05, + "loss": 0.5476, + "step": 7480 + }, + { + "epoch": 0.29, + "learning_rate": 4.5235316747884076e-05, + "loss": 0.5977, + "step": 7481 + }, + { + "epoch": 0.29, + "learning_rate": 4.523467555783534e-05, + "loss": 0.4998, + "step": 7482 + }, + { + "epoch": 0.29, + "learning_rate": 4.523403436778661e-05, + "loss": 0.5362, + "step": 7483 + }, + { + "epoch": 0.29, + "learning_rate": 4.5233393177737884e-05, + "loss": 0.5864, + "step": 7484 + }, + { + "epoch": 0.29, + "learning_rate": 4.5232751987689156e-05, + "loss": 0.532, + "step": 7485 + }, + { + "epoch": 0.29, + "learning_rate": 4.523211079764042e-05, + "loss": 0.4993, + "step": 7486 + }, + { + "epoch": 0.29, + "learning_rate": 4.523146960759169e-05, + "loss": 0.5607, + "step": 7487 + }, + { + "epoch": 0.29, + "learning_rate": 4.5230828417542964e-05, + "loss": 0.5531, + "step": 7488 + }, + { + "epoch": 0.29, + "learning_rate": 4.523018722749423e-05, + "loss": 0.5453, + "step": 7489 + }, + { + "epoch": 0.29, + "learning_rate": 4.52295460374455e-05, + "loss": 0.6364, + "step": 7490 + }, + { + "epoch": 0.29, + "learning_rate": 4.522890484739677e-05, + "loss": 0.5772, + "step": 7491 + }, + { + "epoch": 0.29, + "learning_rate": 4.5228263657348044e-05, + "loss": 0.5467, + "step": 7492 + }, + { + "epoch": 0.29, + "learning_rate": 4.522762246729931e-05, + "loss": 0.6432, + "step": 7493 + }, + { + "epoch": 0.29, + "learning_rate": 4.522698127725058e-05, + "loss": 0.4884, + "step": 7494 + }, + { + "epoch": 0.29, + "learning_rate": 4.5226340087201846e-05, + "loss": 0.5072, + "step": 7495 + }, + { + "epoch": 0.29, + "learning_rate": 4.522569889715312e-05, + "loss": 0.6557, + "step": 7496 + }, + { + "epoch": 0.29, + "learning_rate": 4.522505770710439e-05, + "loss": 0.6432, + "step": 7497 + }, + { + "epoch": 0.29, + "learning_rate": 4.5224416517055654e-05, + "loss": 0.5682, + "step": 7498 + }, + { + "epoch": 0.29, + "learning_rate": 4.5223775327006926e-05, + "loss": 0.7032, + "step": 7499 + }, + { + "epoch": 0.29, + "learning_rate": 4.52231341369582e-05, + "loss": 0.6176, + "step": 7500 + }, + { + "epoch": 0.29, + "learning_rate": 4.522249294690947e-05, + "loss": 0.4809, + "step": 7501 + }, + { + "epoch": 0.29, + "learning_rate": 4.5221851756860734e-05, + "loss": 0.5598, + "step": 7502 + }, + { + "epoch": 0.29, + "learning_rate": 4.5221210566812006e-05, + "loss": 0.5179, + "step": 7503 + }, + { + "epoch": 0.29, + "learning_rate": 4.522056937676327e-05, + "loss": 0.5907, + "step": 7504 + }, + { + "epoch": 0.29, + "learning_rate": 4.521992818671454e-05, + "loss": 0.5699, + "step": 7505 + }, + { + "epoch": 0.29, + "learning_rate": 4.5219286996665814e-05, + "loss": 0.6174, + "step": 7506 + }, + { + "epoch": 0.29, + "learning_rate": 4.521864580661708e-05, + "loss": 0.5661, + "step": 7507 + }, + { + "epoch": 0.29, + "learning_rate": 4.521800461656836e-05, + "loss": 0.5422, + "step": 7508 + }, + { + "epoch": 0.29, + "learning_rate": 4.521736342651962e-05, + "loss": 0.5729, + "step": 7509 + }, + { + "epoch": 0.29, + "learning_rate": 4.5216722236470894e-05, + "loss": 0.5652, + "step": 7510 + }, + { + "epoch": 0.29, + "learning_rate": 4.521608104642216e-05, + "loss": 0.5764, + "step": 7511 + }, + { + "epoch": 0.29, + "learning_rate": 4.521543985637343e-05, + "loss": 0.6216, + "step": 7512 + }, + { + "epoch": 0.29, + "learning_rate": 4.5214798666324696e-05, + "loss": 0.4521, + "step": 7513 + }, + { + "epoch": 0.29, + "learning_rate": 4.521415747627597e-05, + "loss": 0.5265, + "step": 7514 + }, + { + "epoch": 0.29, + "learning_rate": 4.521351628622724e-05, + "loss": 0.5639, + "step": 7515 + }, + { + "epoch": 0.29, + "learning_rate": 4.521287509617851e-05, + "loss": 0.5778, + "step": 7516 + }, + { + "epoch": 0.29, + "learning_rate": 4.521223390612978e-05, + "loss": 0.5738, + "step": 7517 + }, + { + "epoch": 0.29, + "learning_rate": 4.521159271608105e-05, + "loss": 0.5381, + "step": 7518 + }, + { + "epoch": 0.29, + "learning_rate": 4.521095152603232e-05, + "loss": 0.6019, + "step": 7519 + }, + { + "epoch": 0.29, + "learning_rate": 4.5210310335983585e-05, + "loss": 0.5419, + "step": 7520 + }, + { + "epoch": 0.29, + "learning_rate": 4.5209669145934856e-05, + "loss": 0.6869, + "step": 7521 + }, + { + "epoch": 0.29, + "learning_rate": 4.520902795588612e-05, + "loss": 0.5857, + "step": 7522 + }, + { + "epoch": 0.29, + "learning_rate": 4.520838676583739e-05, + "loss": 0.556, + "step": 7523 + }, + { + "epoch": 0.29, + "learning_rate": 4.520774557578867e-05, + "loss": 0.5629, + "step": 7524 + }, + { + "epoch": 0.29, + "learning_rate": 4.5207104385739936e-05, + "loss": 0.5838, + "step": 7525 + }, + { + "epoch": 0.29, + "learning_rate": 4.520646319569121e-05, + "loss": 0.5421, + "step": 7526 + }, + { + "epoch": 0.29, + "learning_rate": 4.520582200564247e-05, + "loss": 0.5193, + "step": 7527 + }, + { + "epoch": 0.29, + "learning_rate": 4.5205180815593745e-05, + "loss": 0.5429, + "step": 7528 + }, + { + "epoch": 0.29, + "learning_rate": 4.520453962554501e-05, + "loss": 0.6593, + "step": 7529 + }, + { + "epoch": 0.29, + "learning_rate": 4.520389843549628e-05, + "loss": 0.5061, + "step": 7530 + }, + { + "epoch": 0.29, + "learning_rate": 4.520325724544755e-05, + "loss": 0.5199, + "step": 7531 + }, + { + "epoch": 0.29, + "learning_rate": 4.5202616055398825e-05, + "loss": 0.6128, + "step": 7532 + }, + { + "epoch": 0.29, + "learning_rate": 4.5201974865350097e-05, + "loss": 0.6016, + "step": 7533 + }, + { + "epoch": 0.29, + "learning_rate": 4.520133367530136e-05, + "loss": 0.669, + "step": 7534 + }, + { + "epoch": 0.29, + "learning_rate": 4.520069248525263e-05, + "loss": 0.4733, + "step": 7535 + }, + { + "epoch": 0.29, + "learning_rate": 4.52000512952039e-05, + "loss": 0.5435, + "step": 7536 + }, + { + "epoch": 0.29, + "learning_rate": 4.519941010515517e-05, + "loss": 0.6166, + "step": 7537 + }, + { + "epoch": 0.29, + "learning_rate": 4.5198768915106435e-05, + "loss": 0.6092, + "step": 7538 + }, + { + "epoch": 0.29, + "learning_rate": 4.5198127725057706e-05, + "loss": 0.5631, + "step": 7539 + }, + { + "epoch": 0.29, + "learning_rate": 4.519748653500898e-05, + "loss": 0.5731, + "step": 7540 + }, + { + "epoch": 0.29, + "learning_rate": 4.519684534496025e-05, + "loss": 0.6107, + "step": 7541 + }, + { + "epoch": 0.29, + "learning_rate": 4.519620415491152e-05, + "loss": 0.6498, + "step": 7542 + }, + { + "epoch": 0.29, + "learning_rate": 4.5195562964862787e-05, + "loss": 0.6083, + "step": 7543 + }, + { + "epoch": 0.29, + "learning_rate": 4.519492177481406e-05, + "loss": 0.5226, + "step": 7544 + }, + { + "epoch": 0.29, + "learning_rate": 4.519428058476532e-05, + "loss": 0.5712, + "step": 7545 + }, + { + "epoch": 0.29, + "learning_rate": 4.5193639394716595e-05, + "loss": 0.604, + "step": 7546 + }, + { + "epoch": 0.29, + "learning_rate": 4.519299820466787e-05, + "loss": 0.541, + "step": 7547 + }, + { + "epoch": 0.29, + "learning_rate": 4.519235701461914e-05, + "loss": 0.4961, + "step": 7548 + }, + { + "epoch": 0.29, + "learning_rate": 4.519171582457041e-05, + "loss": 0.5905, + "step": 7549 + }, + { + "epoch": 0.29, + "learning_rate": 4.5191074634521675e-05, + "loss": 0.5726, + "step": 7550 + }, + { + "epoch": 0.29, + "learning_rate": 4.519043344447295e-05, + "loss": 0.4696, + "step": 7551 + }, + { + "epoch": 0.29, + "learning_rate": 4.518979225442421e-05, + "loss": 0.553, + "step": 7552 + }, + { + "epoch": 0.29, + "learning_rate": 4.5189151064375483e-05, + "loss": 0.4972, + "step": 7553 + }, + { + "epoch": 0.29, + "learning_rate": 4.518850987432675e-05, + "loss": 0.681, + "step": 7554 + }, + { + "epoch": 0.29, + "learning_rate": 4.518786868427802e-05, + "loss": 0.5728, + "step": 7555 + }, + { + "epoch": 0.29, + "learning_rate": 4.518722749422929e-05, + "loss": 0.5606, + "step": 7556 + }, + { + "epoch": 0.29, + "learning_rate": 4.5186586304180563e-05, + "loss": 0.5725, + "step": 7557 + }, + { + "epoch": 0.29, + "learning_rate": 4.5185945114131835e-05, + "loss": 0.5286, + "step": 7558 + }, + { + "epoch": 0.29, + "learning_rate": 4.51853039240831e-05, + "loss": 0.4907, + "step": 7559 + }, + { + "epoch": 0.29, + "learning_rate": 4.518466273403437e-05, + "loss": 0.517, + "step": 7560 + }, + { + "epoch": 0.29, + "learning_rate": 4.518402154398564e-05, + "loss": 0.5663, + "step": 7561 + }, + { + "epoch": 0.29, + "learning_rate": 4.518338035393691e-05, + "loss": 0.5588, + "step": 7562 + }, + { + "epoch": 0.29, + "learning_rate": 4.518273916388818e-05, + "loss": 0.6801, + "step": 7563 + }, + { + "epoch": 0.29, + "learning_rate": 4.518209797383945e-05, + "loss": 0.5481, + "step": 7564 + }, + { + "epoch": 0.29, + "learning_rate": 4.518145678379072e-05, + "loss": 0.5764, + "step": 7565 + }, + { + "epoch": 0.29, + "learning_rate": 4.518081559374199e-05, + "loss": 0.5634, + "step": 7566 + }, + { + "epoch": 0.29, + "learning_rate": 4.518017440369326e-05, + "loss": 0.5464, + "step": 7567 + }, + { + "epoch": 0.29, + "learning_rate": 4.5179533213644525e-05, + "loss": 0.6377, + "step": 7568 + }, + { + "epoch": 0.29, + "learning_rate": 4.51788920235958e-05, + "loss": 0.5315, + "step": 7569 + }, + { + "epoch": 0.29, + "learning_rate": 4.517825083354706e-05, + "loss": 0.5312, + "step": 7570 + }, + { + "epoch": 0.29, + "learning_rate": 4.5177609643498334e-05, + "loss": 0.5794, + "step": 7571 + }, + { + "epoch": 0.29, + "learning_rate": 4.5176968453449605e-05, + "loss": 0.5341, + "step": 7572 + }, + { + "epoch": 0.29, + "learning_rate": 4.517632726340088e-05, + "loss": 0.5838, + "step": 7573 + }, + { + "epoch": 0.29, + "learning_rate": 4.517568607335214e-05, + "loss": 0.6528, + "step": 7574 + }, + { + "epoch": 0.29, + "learning_rate": 4.5175044883303414e-05, + "loss": 0.5832, + "step": 7575 + }, + { + "epoch": 0.29, + "learning_rate": 4.5174403693254685e-05, + "loss": 0.5341, + "step": 7576 + }, + { + "epoch": 0.29, + "learning_rate": 4.517376250320595e-05, + "loss": 0.5143, + "step": 7577 + }, + { + "epoch": 0.29, + "learning_rate": 4.517312131315722e-05, + "loss": 0.5375, + "step": 7578 + }, + { + "epoch": 0.29, + "learning_rate": 4.517248012310849e-05, + "loss": 0.5811, + "step": 7579 + }, + { + "epoch": 0.29, + "learning_rate": 4.5171838933059766e-05, + "loss": 0.5185, + "step": 7580 + }, + { + "epoch": 0.29, + "learning_rate": 4.517119774301103e-05, + "loss": 0.5773, + "step": 7581 + }, + { + "epoch": 0.29, + "learning_rate": 4.51705565529623e-05, + "loss": 0.5602, + "step": 7582 + }, + { + "epoch": 0.29, + "learning_rate": 4.516991536291357e-05, + "loss": 0.6487, + "step": 7583 + }, + { + "epoch": 0.29, + "learning_rate": 4.516927417286484e-05, + "loss": 0.5917, + "step": 7584 + }, + { + "epoch": 0.29, + "learning_rate": 4.516863298281611e-05, + "loss": 0.5924, + "step": 7585 + }, + { + "epoch": 0.29, + "learning_rate": 4.5167991792767376e-05, + "loss": 0.6413, + "step": 7586 + }, + { + "epoch": 0.29, + "learning_rate": 4.516735060271865e-05, + "loss": 0.5901, + "step": 7587 + }, + { + "epoch": 0.29, + "learning_rate": 4.516670941266992e-05, + "loss": 0.61, + "step": 7588 + }, + { + "epoch": 0.29, + "learning_rate": 4.516606822262119e-05, + "loss": 0.5343, + "step": 7589 + }, + { + "epoch": 0.29, + "learning_rate": 4.5165427032572456e-05, + "loss": 0.5875, + "step": 7590 + }, + { + "epoch": 0.29, + "learning_rate": 4.516478584252373e-05, + "loss": 0.6984, + "step": 7591 + }, + { + "epoch": 0.29, + "learning_rate": 4.516414465247499e-05, + "loss": 0.548, + "step": 7592 + }, + { + "epoch": 0.29, + "learning_rate": 4.5163503462426264e-05, + "loss": 0.6072, + "step": 7593 + }, + { + "epoch": 0.29, + "learning_rate": 4.5162862272377536e-05, + "loss": 0.567, + "step": 7594 + }, + { + "epoch": 0.29, + "learning_rate": 4.51622210823288e-05, + "loss": 0.6068, + "step": 7595 + }, + { + "epoch": 0.29, + "learning_rate": 4.516157989228008e-05, + "loss": 0.6293, + "step": 7596 + }, + { + "epoch": 0.29, + "learning_rate": 4.5160938702231344e-05, + "loss": 0.5836, + "step": 7597 + }, + { + "epoch": 0.29, + "learning_rate": 4.5160297512182616e-05, + "loss": 0.5774, + "step": 7598 + }, + { + "epoch": 0.29, + "learning_rate": 4.515965632213388e-05, + "loss": 0.5421, + "step": 7599 + }, + { + "epoch": 0.29, + "learning_rate": 4.515901513208515e-05, + "loss": 0.6596, + "step": 7600 + }, + { + "epoch": 0.29, + "learning_rate": 4.515837394203642e-05, + "loss": 0.5338, + "step": 7601 + }, + { + "epoch": 0.29, + "learning_rate": 4.515773275198769e-05, + "loss": 0.5438, + "step": 7602 + }, + { + "epoch": 0.29, + "learning_rate": 4.515709156193896e-05, + "loss": 0.6307, + "step": 7603 + }, + { + "epoch": 0.29, + "learning_rate": 4.515645037189023e-05, + "loss": 0.6251, + "step": 7604 + }, + { + "epoch": 0.29, + "learning_rate": 4.5155809181841504e-05, + "loss": 0.6246, + "step": 7605 + }, + { + "epoch": 0.29, + "learning_rate": 4.515516799179277e-05, + "loss": 0.6227, + "step": 7606 + }, + { + "epoch": 0.29, + "learning_rate": 4.515452680174404e-05, + "loss": 0.5766, + "step": 7607 + }, + { + "epoch": 0.29, + "learning_rate": 4.5153885611695306e-05, + "loss": 0.6224, + "step": 7608 + }, + { + "epoch": 0.29, + "learning_rate": 4.515324442164658e-05, + "loss": 0.6034, + "step": 7609 + }, + { + "epoch": 0.29, + "learning_rate": 4.515260323159784e-05, + "loss": 0.6151, + "step": 7610 + }, + { + "epoch": 0.29, + "learning_rate": 4.5151962041549114e-05, + "loss": 0.6615, + "step": 7611 + }, + { + "epoch": 0.29, + "learning_rate": 4.515132085150039e-05, + "loss": 0.5753, + "step": 7612 + }, + { + "epoch": 0.29, + "learning_rate": 4.515067966145166e-05, + "loss": 0.5646, + "step": 7613 + }, + { + "epoch": 0.29, + "learning_rate": 4.515003847140293e-05, + "loss": 0.6294, + "step": 7614 + }, + { + "epoch": 0.29, + "learning_rate": 4.5149397281354194e-05, + "loss": 0.5658, + "step": 7615 + }, + { + "epoch": 0.29, + "learning_rate": 4.5148756091305466e-05, + "loss": 0.5717, + "step": 7616 + }, + { + "epoch": 0.29, + "learning_rate": 4.514811490125673e-05, + "loss": 0.5, + "step": 7617 + }, + { + "epoch": 0.29, + "learning_rate": 4.5147473711208e-05, + "loss": 0.5202, + "step": 7618 + }, + { + "epoch": 0.29, + "learning_rate": 4.5146832521159274e-05, + "loss": 0.5707, + "step": 7619 + }, + { + "epoch": 0.29, + "learning_rate": 4.5146191331110546e-05, + "loss": 0.5886, + "step": 7620 + }, + { + "epoch": 0.29, + "learning_rate": 4.514555014106182e-05, + "loss": 0.6102, + "step": 7621 + }, + { + "epoch": 0.29, + "learning_rate": 4.514490895101308e-05, + "loss": 0.6254, + "step": 7622 + }, + { + "epoch": 0.29, + "learning_rate": 4.5144267760964354e-05, + "loss": 0.5817, + "step": 7623 + }, + { + "epoch": 0.29, + "learning_rate": 4.514362657091562e-05, + "loss": 0.4662, + "step": 7624 + }, + { + "epoch": 0.29, + "learning_rate": 4.514298538086689e-05, + "loss": 0.6006, + "step": 7625 + }, + { + "epoch": 0.29, + "learning_rate": 4.5142344190818156e-05, + "loss": 0.5578, + "step": 7626 + }, + { + "epoch": 0.29, + "learning_rate": 4.514170300076943e-05, + "loss": 0.4879, + "step": 7627 + }, + { + "epoch": 0.29, + "learning_rate": 4.51410618107207e-05, + "loss": 0.5553, + "step": 7628 + }, + { + "epoch": 0.29, + "learning_rate": 4.514042062067197e-05, + "loss": 0.5595, + "step": 7629 + }, + { + "epoch": 0.29, + "learning_rate": 4.513977943062324e-05, + "loss": 0.4845, + "step": 7630 + }, + { + "epoch": 0.29, + "learning_rate": 4.513913824057451e-05, + "loss": 0.5403, + "step": 7631 + }, + { + "epoch": 0.29, + "learning_rate": 4.513849705052578e-05, + "loss": 0.5465, + "step": 7632 + }, + { + "epoch": 0.29, + "learning_rate": 4.5137855860477045e-05, + "loss": 0.5193, + "step": 7633 + }, + { + "epoch": 0.29, + "learning_rate": 4.5137214670428316e-05, + "loss": 0.6094, + "step": 7634 + }, + { + "epoch": 0.29, + "learning_rate": 4.513657348037959e-05, + "loss": 0.6528, + "step": 7635 + }, + { + "epoch": 0.29, + "learning_rate": 4.513593229033086e-05, + "loss": 0.6356, + "step": 7636 + }, + { + "epoch": 0.29, + "learning_rate": 4.5135291100282125e-05, + "loss": 0.6965, + "step": 7637 + }, + { + "epoch": 0.29, + "learning_rate": 4.5134649910233396e-05, + "loss": 0.6417, + "step": 7638 + }, + { + "epoch": 0.29, + "learning_rate": 4.513400872018467e-05, + "loss": 0.5453, + "step": 7639 + }, + { + "epoch": 0.29, + "learning_rate": 4.513336753013593e-05, + "loss": 0.5627, + "step": 7640 + }, + { + "epoch": 0.29, + "learning_rate": 4.5132726340087205e-05, + "loss": 0.5754, + "step": 7641 + }, + { + "epoch": 0.29, + "learning_rate": 4.513208515003847e-05, + "loss": 0.5669, + "step": 7642 + }, + { + "epoch": 0.29, + "learning_rate": 4.513144395998974e-05, + "loss": 0.5776, + "step": 7643 + }, + { + "epoch": 0.29, + "learning_rate": 4.513080276994101e-05, + "loss": 0.5178, + "step": 7644 + }, + { + "epoch": 0.29, + "learning_rate": 4.5130161579892285e-05, + "loss": 0.5825, + "step": 7645 + }, + { + "epoch": 0.29, + "learning_rate": 4.512952038984355e-05, + "loss": 0.5, + "step": 7646 + }, + { + "epoch": 0.29, + "learning_rate": 4.512887919979482e-05, + "loss": 0.617, + "step": 7647 + }, + { + "epoch": 0.29, + "learning_rate": 4.512823800974609e-05, + "loss": 0.6205, + "step": 7648 + }, + { + "epoch": 0.29, + "learning_rate": 4.512759681969736e-05, + "loss": 0.5868, + "step": 7649 + }, + { + "epoch": 0.29, + "learning_rate": 4.512695562964863e-05, + "loss": 0.5759, + "step": 7650 + }, + { + "epoch": 0.29, + "learning_rate": 4.51263144395999e-05, + "loss": 0.5681, + "step": 7651 + }, + { + "epoch": 0.29, + "learning_rate": 4.512567324955117e-05, + "loss": 0.6393, + "step": 7652 + }, + { + "epoch": 0.29, + "learning_rate": 4.512503205950244e-05, + "loss": 0.5596, + "step": 7653 + }, + { + "epoch": 0.29, + "learning_rate": 4.512439086945371e-05, + "loss": 0.5712, + "step": 7654 + }, + { + "epoch": 0.29, + "learning_rate": 4.5123749679404975e-05, + "loss": 0.594, + "step": 7655 + }, + { + "epoch": 0.29, + "learning_rate": 4.5123108489356247e-05, + "loss": 0.6228, + "step": 7656 + }, + { + "epoch": 0.29, + "learning_rate": 4.512246729930752e-05, + "loss": 0.5251, + "step": 7657 + }, + { + "epoch": 0.29, + "learning_rate": 4.512182610925878e-05, + "loss": 0.5235, + "step": 7658 + }, + { + "epoch": 0.29, + "learning_rate": 4.5121184919210055e-05, + "loss": 0.4976, + "step": 7659 + }, + { + "epoch": 0.29, + "learning_rate": 4.512054372916133e-05, + "loss": 0.6316, + "step": 7660 + }, + { + "epoch": 0.29, + "learning_rate": 4.51199025391126e-05, + "loss": 0.5936, + "step": 7661 + }, + { + "epoch": 0.29, + "learning_rate": 4.511926134906386e-05, + "loss": 0.5319, + "step": 7662 + }, + { + "epoch": 0.29, + "learning_rate": 4.5118620159015135e-05, + "loss": 0.5905, + "step": 7663 + }, + { + "epoch": 0.29, + "learning_rate": 4.51179789689664e-05, + "loss": 0.5543, + "step": 7664 + }, + { + "epoch": 0.29, + "learning_rate": 4.511733777891767e-05, + "loss": 0.5797, + "step": 7665 + }, + { + "epoch": 0.29, + "learning_rate": 4.5116696588868943e-05, + "loss": 0.5476, + "step": 7666 + }, + { + "epoch": 0.29, + "learning_rate": 4.511605539882021e-05, + "loss": 0.5491, + "step": 7667 + }, + { + "epoch": 0.29, + "learning_rate": 4.511541420877149e-05, + "loss": 0.5482, + "step": 7668 + }, + { + "epoch": 0.29, + "learning_rate": 4.511477301872275e-05, + "loss": 0.5023, + "step": 7669 + }, + { + "epoch": 0.29, + "learning_rate": 4.5114131828674024e-05, + "loss": 0.632, + "step": 7670 + }, + { + "epoch": 0.29, + "learning_rate": 4.511349063862529e-05, + "loss": 0.5881, + "step": 7671 + }, + { + "epoch": 0.29, + "learning_rate": 4.511284944857656e-05, + "loss": 0.5012, + "step": 7672 + }, + { + "epoch": 0.3, + "learning_rate": 4.5112208258527825e-05, + "loss": 0.6656, + "step": 7673 + }, + { + "epoch": 0.3, + "learning_rate": 4.51115670684791e-05, + "loss": 0.6535, + "step": 7674 + }, + { + "epoch": 0.3, + "learning_rate": 4.511092587843037e-05, + "loss": 0.5665, + "step": 7675 + }, + { + "epoch": 0.3, + "learning_rate": 4.511028468838164e-05, + "loss": 0.4989, + "step": 7676 + }, + { + "epoch": 0.3, + "learning_rate": 4.510964349833291e-05, + "loss": 0.6781, + "step": 7677 + }, + { + "epoch": 0.3, + "learning_rate": 4.510900230828418e-05, + "loss": 0.6836, + "step": 7678 + }, + { + "epoch": 0.3, + "learning_rate": 4.510836111823545e-05, + "loss": 0.5343, + "step": 7679 + }, + { + "epoch": 0.3, + "learning_rate": 4.5107719928186714e-05, + "loss": 0.5467, + "step": 7680 + }, + { + "epoch": 0.3, + "learning_rate": 4.5107078738137985e-05, + "loss": 0.5971, + "step": 7681 + }, + { + "epoch": 0.3, + "learning_rate": 4.510643754808925e-05, + "loss": 0.5982, + "step": 7682 + }, + { + "epoch": 0.3, + "learning_rate": 4.510579635804052e-05, + "loss": 0.6139, + "step": 7683 + }, + { + "epoch": 0.3, + "learning_rate": 4.51051551679918e-05, + "loss": 0.4958, + "step": 7684 + }, + { + "epoch": 0.3, + "learning_rate": 4.5104513977943065e-05, + "loss": 0.6407, + "step": 7685 + }, + { + "epoch": 0.3, + "learning_rate": 4.510387278789434e-05, + "loss": 0.4724, + "step": 7686 + }, + { + "epoch": 0.3, + "learning_rate": 4.51032315978456e-05, + "loss": 0.5998, + "step": 7687 + }, + { + "epoch": 0.3, + "learning_rate": 4.5102590407796874e-05, + "loss": 0.5989, + "step": 7688 + }, + { + "epoch": 0.3, + "learning_rate": 4.510194921774814e-05, + "loss": 0.6534, + "step": 7689 + }, + { + "epoch": 0.3, + "learning_rate": 4.510130802769941e-05, + "loss": 0.5418, + "step": 7690 + }, + { + "epoch": 0.3, + "learning_rate": 4.510066683765068e-05, + "loss": 0.6211, + "step": 7691 + }, + { + "epoch": 0.3, + "learning_rate": 4.5100025647601954e-05, + "loss": 0.4659, + "step": 7692 + }, + { + "epoch": 0.3, + "learning_rate": 4.5099384457553226e-05, + "loss": 0.5351, + "step": 7693 + }, + { + "epoch": 0.3, + "learning_rate": 4.509874326750449e-05, + "loss": 0.7076, + "step": 7694 + }, + { + "epoch": 0.3, + "learning_rate": 4.509810207745576e-05, + "loss": 0.5136, + "step": 7695 + }, + { + "epoch": 0.3, + "learning_rate": 4.509746088740703e-05, + "loss": 0.6126, + "step": 7696 + }, + { + "epoch": 0.3, + "learning_rate": 4.50968196973583e-05, + "loss": 0.5033, + "step": 7697 + }, + { + "epoch": 0.3, + "learning_rate": 4.5096178507309564e-05, + "loss": 0.4999, + "step": 7698 + }, + { + "epoch": 0.3, + "learning_rate": 4.5095537317260836e-05, + "loss": 0.6423, + "step": 7699 + }, + { + "epoch": 0.3, + "learning_rate": 4.509489612721211e-05, + "loss": 0.5322, + "step": 7700 + }, + { + "epoch": 0.3, + "learning_rate": 4.509425493716338e-05, + "loss": 0.5587, + "step": 7701 + }, + { + "epoch": 0.3, + "learning_rate": 4.509361374711465e-05, + "loss": 0.6017, + "step": 7702 + }, + { + "epoch": 0.3, + "learning_rate": 4.5092972557065916e-05, + "loss": 0.6241, + "step": 7703 + }, + { + "epoch": 0.3, + "learning_rate": 4.509233136701719e-05, + "loss": 0.6332, + "step": 7704 + }, + { + "epoch": 0.3, + "learning_rate": 4.509169017696845e-05, + "loss": 0.587, + "step": 7705 + }, + { + "epoch": 0.3, + "learning_rate": 4.5091048986919724e-05, + "loss": 0.5733, + "step": 7706 + }, + { + "epoch": 0.3, + "learning_rate": 4.5090407796870996e-05, + "loss": 0.5358, + "step": 7707 + }, + { + "epoch": 0.3, + "learning_rate": 4.508976660682227e-05, + "loss": 0.5866, + "step": 7708 + }, + { + "epoch": 0.3, + "learning_rate": 4.508912541677353e-05, + "loss": 0.6481, + "step": 7709 + }, + { + "epoch": 0.3, + "learning_rate": 4.5088484226724804e-05, + "loss": 0.5565, + "step": 7710 + }, + { + "epoch": 0.3, + "learning_rate": 4.5087843036676076e-05, + "loss": 0.6222, + "step": 7711 + }, + { + "epoch": 0.3, + "learning_rate": 4.508720184662734e-05, + "loss": 0.5636, + "step": 7712 + }, + { + "epoch": 0.3, + "learning_rate": 4.508656065657861e-05, + "loss": 0.4953, + "step": 7713 + }, + { + "epoch": 0.3, + "learning_rate": 4.508591946652988e-05, + "loss": 0.5767, + "step": 7714 + }, + { + "epoch": 0.3, + "learning_rate": 4.508527827648115e-05, + "loss": 0.6457, + "step": 7715 + }, + { + "epoch": 0.3, + "learning_rate": 4.508463708643242e-05, + "loss": 0.5432, + "step": 7716 + }, + { + "epoch": 0.3, + "learning_rate": 4.508399589638369e-05, + "loss": 0.5706, + "step": 7717 + }, + { + "epoch": 0.3, + "learning_rate": 4.5083354706334964e-05, + "loss": 0.5754, + "step": 7718 + }, + { + "epoch": 0.3, + "learning_rate": 4.508271351628623e-05, + "loss": 0.6299, + "step": 7719 + }, + { + "epoch": 0.3, + "learning_rate": 4.50820723262375e-05, + "loss": 0.5773, + "step": 7720 + }, + { + "epoch": 0.3, + "learning_rate": 4.5081431136188766e-05, + "loss": 0.5622, + "step": 7721 + }, + { + "epoch": 0.3, + "learning_rate": 4.508078994614004e-05, + "loss": 0.571, + "step": 7722 + }, + { + "epoch": 0.3, + "learning_rate": 4.508014875609131e-05, + "loss": 0.5398, + "step": 7723 + }, + { + "epoch": 0.3, + "learning_rate": 4.507950756604258e-05, + "loss": 0.547, + "step": 7724 + }, + { + "epoch": 0.3, + "learning_rate": 4.5078866375993846e-05, + "loss": 0.6336, + "step": 7725 + }, + { + "epoch": 0.3, + "learning_rate": 4.507822518594512e-05, + "loss": 0.7175, + "step": 7726 + }, + { + "epoch": 0.3, + "learning_rate": 4.507758399589639e-05, + "loss": 0.6011, + "step": 7727 + }, + { + "epoch": 0.3, + "learning_rate": 4.5076942805847654e-05, + "loss": 0.5673, + "step": 7728 + }, + { + "epoch": 0.3, + "learning_rate": 4.5076301615798926e-05, + "loss": 0.5275, + "step": 7729 + }, + { + "epoch": 0.3, + "learning_rate": 4.507566042575019e-05, + "loss": 0.5746, + "step": 7730 + }, + { + "epoch": 0.3, + "learning_rate": 4.507501923570146e-05, + "loss": 0.5661, + "step": 7731 + }, + { + "epoch": 0.3, + "learning_rate": 4.5074378045652734e-05, + "loss": 0.5965, + "step": 7732 + }, + { + "epoch": 0.3, + "learning_rate": 4.5073736855604006e-05, + "loss": 0.5797, + "step": 7733 + }, + { + "epoch": 0.3, + "learning_rate": 4.507309566555527e-05, + "loss": 0.5534, + "step": 7734 + }, + { + "epoch": 0.3, + "learning_rate": 4.507245447550654e-05, + "loss": 0.5518, + "step": 7735 + }, + { + "epoch": 0.3, + "learning_rate": 4.5071813285457815e-05, + "loss": 0.5815, + "step": 7736 + }, + { + "epoch": 0.3, + "learning_rate": 4.507117209540908e-05, + "loss": 0.574, + "step": 7737 + }, + { + "epoch": 0.3, + "learning_rate": 4.507053090536035e-05, + "loss": 0.608, + "step": 7738 + }, + { + "epoch": 0.3, + "learning_rate": 4.506988971531162e-05, + "loss": 0.6093, + "step": 7739 + }, + { + "epoch": 0.3, + "learning_rate": 4.5069248525262895e-05, + "loss": 0.6456, + "step": 7740 + }, + { + "epoch": 0.3, + "learning_rate": 4.506860733521416e-05, + "loss": 0.6949, + "step": 7741 + }, + { + "epoch": 0.3, + "learning_rate": 4.506796614516543e-05, + "loss": 0.5891, + "step": 7742 + }, + { + "epoch": 0.3, + "learning_rate": 4.5067324955116696e-05, + "loss": 0.6156, + "step": 7743 + }, + { + "epoch": 0.3, + "learning_rate": 4.506668376506797e-05, + "loss": 0.5391, + "step": 7744 + }, + { + "epoch": 0.3, + "learning_rate": 4.506604257501924e-05, + "loss": 0.6012, + "step": 7745 + }, + { + "epoch": 0.3, + "learning_rate": 4.5065401384970505e-05, + "loss": 0.5361, + "step": 7746 + }, + { + "epoch": 0.3, + "learning_rate": 4.5064760194921776e-05, + "loss": 0.6679, + "step": 7747 + }, + { + "epoch": 0.3, + "learning_rate": 4.506411900487305e-05, + "loss": 0.5385, + "step": 7748 + }, + { + "epoch": 0.3, + "learning_rate": 4.506347781482432e-05, + "loss": 0.5495, + "step": 7749 + }, + { + "epoch": 0.3, + "learning_rate": 4.5062836624775585e-05, + "loss": 0.6316, + "step": 7750 + }, + { + "epoch": 0.3, + "learning_rate": 4.5062195434726856e-05, + "loss": 0.5137, + "step": 7751 + }, + { + "epoch": 0.3, + "learning_rate": 4.506155424467812e-05, + "loss": 0.6002, + "step": 7752 + }, + { + "epoch": 0.3, + "learning_rate": 4.506091305462939e-05, + "loss": 0.564, + "step": 7753 + }, + { + "epoch": 0.3, + "learning_rate": 4.5060271864580665e-05, + "loss": 0.6313, + "step": 7754 + }, + { + "epoch": 0.3, + "learning_rate": 4.505963067453193e-05, + "loss": 0.5934, + "step": 7755 + }, + { + "epoch": 0.3, + "learning_rate": 4.505898948448321e-05, + "loss": 0.5516, + "step": 7756 + }, + { + "epoch": 0.3, + "learning_rate": 4.505834829443447e-05, + "loss": 0.5273, + "step": 7757 + }, + { + "epoch": 0.3, + "learning_rate": 4.5057707104385745e-05, + "loss": 0.5311, + "step": 7758 + }, + { + "epoch": 0.3, + "learning_rate": 4.505706591433701e-05, + "loss": 0.642, + "step": 7759 + }, + { + "epoch": 0.3, + "learning_rate": 4.505642472428828e-05, + "loss": 0.5537, + "step": 7760 + }, + { + "epoch": 0.3, + "learning_rate": 4.5055783534239546e-05, + "loss": 0.5501, + "step": 7761 + }, + { + "epoch": 0.3, + "learning_rate": 4.505514234419082e-05, + "loss": 0.5207, + "step": 7762 + }, + { + "epoch": 0.3, + "learning_rate": 4.505450115414209e-05, + "loss": 0.5477, + "step": 7763 + }, + { + "epoch": 0.3, + "learning_rate": 4.505385996409336e-05, + "loss": 0.5729, + "step": 7764 + }, + { + "epoch": 0.3, + "learning_rate": 4.505321877404463e-05, + "loss": 0.556, + "step": 7765 + }, + { + "epoch": 0.3, + "learning_rate": 4.50525775839959e-05, + "loss": 0.5966, + "step": 7766 + }, + { + "epoch": 0.3, + "learning_rate": 4.505193639394717e-05, + "loss": 0.4895, + "step": 7767 + }, + { + "epoch": 0.3, + "learning_rate": 4.5051295203898435e-05, + "loss": 0.5299, + "step": 7768 + }, + { + "epoch": 0.3, + "learning_rate": 4.505065401384971e-05, + "loss": 0.5438, + "step": 7769 + }, + { + "epoch": 0.3, + "learning_rate": 4.505001282380097e-05, + "loss": 0.6141, + "step": 7770 + }, + { + "epoch": 0.3, + "learning_rate": 4.504937163375224e-05, + "loss": 0.4537, + "step": 7771 + }, + { + "epoch": 0.3, + "learning_rate": 4.504873044370352e-05, + "loss": 0.6299, + "step": 7772 + }, + { + "epoch": 0.3, + "learning_rate": 4.504808925365479e-05, + "loss": 0.5838, + "step": 7773 + }, + { + "epoch": 0.3, + "learning_rate": 4.504744806360606e-05, + "loss": 0.592, + "step": 7774 + }, + { + "epoch": 0.3, + "learning_rate": 4.504680687355732e-05, + "loss": 0.6139, + "step": 7775 + }, + { + "epoch": 0.3, + "learning_rate": 4.5046165683508595e-05, + "loss": 0.553, + "step": 7776 + }, + { + "epoch": 0.3, + "learning_rate": 4.504552449345986e-05, + "loss": 0.6087, + "step": 7777 + }, + { + "epoch": 0.3, + "learning_rate": 4.504488330341113e-05, + "loss": 0.638, + "step": 7778 + }, + { + "epoch": 0.3, + "learning_rate": 4.5044242113362403e-05, + "loss": 0.5516, + "step": 7779 + }, + { + "epoch": 0.3, + "learning_rate": 4.5043600923313675e-05, + "loss": 0.5997, + "step": 7780 + }, + { + "epoch": 0.3, + "learning_rate": 4.504295973326495e-05, + "loss": 0.5116, + "step": 7781 + }, + { + "epoch": 0.3, + "learning_rate": 4.504231854321621e-05, + "loss": 0.5588, + "step": 7782 + }, + { + "epoch": 0.3, + "learning_rate": 4.5041677353167484e-05, + "loss": 0.5877, + "step": 7783 + }, + { + "epoch": 0.3, + "learning_rate": 4.504103616311875e-05, + "loss": 0.5459, + "step": 7784 + }, + { + "epoch": 0.3, + "learning_rate": 4.504039497307002e-05, + "loss": 0.5099, + "step": 7785 + }, + { + "epoch": 0.3, + "learning_rate": 4.5039753783021285e-05, + "loss": 0.6083, + "step": 7786 + }, + { + "epoch": 0.3, + "learning_rate": 4.503911259297256e-05, + "loss": 0.5647, + "step": 7787 + }, + { + "epoch": 0.3, + "learning_rate": 4.503847140292383e-05, + "loss": 0.4622, + "step": 7788 + }, + { + "epoch": 0.3, + "learning_rate": 4.50378302128751e-05, + "loss": 0.5593, + "step": 7789 + }, + { + "epoch": 0.3, + "learning_rate": 4.503718902282637e-05, + "loss": 0.6578, + "step": 7790 + }, + { + "epoch": 0.3, + "learning_rate": 4.503654783277764e-05, + "loss": 0.5401, + "step": 7791 + }, + { + "epoch": 0.3, + "learning_rate": 4.503590664272891e-05, + "loss": 0.4737, + "step": 7792 + }, + { + "epoch": 0.3, + "learning_rate": 4.5035265452680174e-05, + "loss": 0.6197, + "step": 7793 + }, + { + "epoch": 0.3, + "learning_rate": 4.5034624262631445e-05, + "loss": 0.6404, + "step": 7794 + }, + { + "epoch": 0.3, + "learning_rate": 4.503398307258272e-05, + "loss": 0.5915, + "step": 7795 + }, + { + "epoch": 0.3, + "learning_rate": 4.503334188253399e-05, + "loss": 0.5214, + "step": 7796 + }, + { + "epoch": 0.3, + "learning_rate": 4.5032700692485254e-05, + "loss": 0.5813, + "step": 7797 + }, + { + "epoch": 0.3, + "learning_rate": 4.5032059502436525e-05, + "loss": 0.5565, + "step": 7798 + }, + { + "epoch": 0.3, + "learning_rate": 4.50314183123878e-05, + "loss": 0.5271, + "step": 7799 + }, + { + "epoch": 0.3, + "learning_rate": 4.503077712233906e-05, + "loss": 0.5244, + "step": 7800 + }, + { + "epoch": 0.3, + "learning_rate": 4.5030135932290334e-05, + "loss": 0.5099, + "step": 7801 + }, + { + "epoch": 0.3, + "learning_rate": 4.50294947422416e-05, + "loss": 0.6401, + "step": 7802 + }, + { + "epoch": 0.3, + "learning_rate": 4.502885355219287e-05, + "loss": 0.5556, + "step": 7803 + }, + { + "epoch": 0.3, + "learning_rate": 4.502821236214414e-05, + "loss": 0.5349, + "step": 7804 + }, + { + "epoch": 0.3, + "learning_rate": 4.5027571172095414e-05, + "loss": 0.5534, + "step": 7805 + }, + { + "epoch": 0.3, + "learning_rate": 4.502692998204668e-05, + "loss": 0.6423, + "step": 7806 + }, + { + "epoch": 0.3, + "learning_rate": 4.502628879199795e-05, + "loss": 0.5945, + "step": 7807 + }, + { + "epoch": 0.3, + "learning_rate": 4.502564760194922e-05, + "loss": 0.5892, + "step": 7808 + }, + { + "epoch": 0.3, + "learning_rate": 4.502500641190049e-05, + "loss": 0.6572, + "step": 7809 + }, + { + "epoch": 0.3, + "learning_rate": 4.502436522185176e-05, + "loss": 0.5986, + "step": 7810 + }, + { + "epoch": 0.3, + "learning_rate": 4.502372403180303e-05, + "loss": 0.5458, + "step": 7811 + }, + { + "epoch": 0.3, + "learning_rate": 4.50230828417543e-05, + "loss": 0.4454, + "step": 7812 + }, + { + "epoch": 0.3, + "learning_rate": 4.502244165170557e-05, + "loss": 0.6451, + "step": 7813 + }, + { + "epoch": 0.3, + "learning_rate": 4.502180046165684e-05, + "loss": 0.5116, + "step": 7814 + }, + { + "epoch": 0.3, + "learning_rate": 4.5021159271608104e-05, + "loss": 0.583, + "step": 7815 + }, + { + "epoch": 0.3, + "learning_rate": 4.5020518081559376e-05, + "loss": 0.6309, + "step": 7816 + }, + { + "epoch": 0.3, + "learning_rate": 4.501987689151065e-05, + "loss": 0.5686, + "step": 7817 + }, + { + "epoch": 0.3, + "learning_rate": 4.501923570146191e-05, + "loss": 0.6243, + "step": 7818 + }, + { + "epoch": 0.3, + "learning_rate": 4.5018594511413184e-05, + "loss": 0.6289, + "step": 7819 + }, + { + "epoch": 0.3, + "learning_rate": 4.5017953321364456e-05, + "loss": 0.6032, + "step": 7820 + }, + { + "epoch": 0.3, + "learning_rate": 4.501731213131573e-05, + "loss": 0.5682, + "step": 7821 + }, + { + "epoch": 0.3, + "learning_rate": 4.501667094126699e-05, + "loss": 0.5908, + "step": 7822 + }, + { + "epoch": 0.3, + "learning_rate": 4.5016029751218264e-05, + "loss": 0.573, + "step": 7823 + }, + { + "epoch": 0.3, + "learning_rate": 4.501538856116953e-05, + "loss": 0.5341, + "step": 7824 + }, + { + "epoch": 0.3, + "learning_rate": 4.50147473711208e-05, + "loss": 0.6375, + "step": 7825 + }, + { + "epoch": 0.3, + "learning_rate": 4.501410618107207e-05, + "loss": 0.6236, + "step": 7826 + }, + { + "epoch": 0.3, + "learning_rate": 4.501346499102334e-05, + "loss": 0.641, + "step": 7827 + }, + { + "epoch": 0.3, + "learning_rate": 4.5012823800974616e-05, + "loss": 0.5287, + "step": 7828 + }, + { + "epoch": 0.3, + "learning_rate": 4.501218261092588e-05, + "loss": 0.5714, + "step": 7829 + }, + { + "epoch": 0.3, + "learning_rate": 4.501154142087715e-05, + "loss": 0.5708, + "step": 7830 + }, + { + "epoch": 0.3, + "learning_rate": 4.501090023082842e-05, + "loss": 0.6054, + "step": 7831 + }, + { + "epoch": 0.3, + "learning_rate": 4.501025904077969e-05, + "loss": 0.5814, + "step": 7832 + }, + { + "epoch": 0.3, + "learning_rate": 4.5009617850730954e-05, + "loss": 0.5378, + "step": 7833 + }, + { + "epoch": 0.3, + "learning_rate": 4.5008976660682226e-05, + "loss": 0.556, + "step": 7834 + }, + { + "epoch": 0.3, + "learning_rate": 4.50083354706335e-05, + "loss": 0.5395, + "step": 7835 + }, + { + "epoch": 0.3, + "learning_rate": 4.500769428058477e-05, + "loss": 0.5419, + "step": 7836 + }, + { + "epoch": 0.3, + "learning_rate": 4.500705309053604e-05, + "loss": 0.5674, + "step": 7837 + }, + { + "epoch": 0.3, + "learning_rate": 4.5006411900487306e-05, + "loss": 0.5612, + "step": 7838 + }, + { + "epoch": 0.3, + "learning_rate": 4.500577071043858e-05, + "loss": 0.5707, + "step": 7839 + }, + { + "epoch": 0.3, + "learning_rate": 4.500512952038984e-05, + "loss": 0.5864, + "step": 7840 + }, + { + "epoch": 0.3, + "learning_rate": 4.5004488330341114e-05, + "loss": 0.6316, + "step": 7841 + }, + { + "epoch": 0.3, + "learning_rate": 4.500384714029238e-05, + "loss": 0.5821, + "step": 7842 + }, + { + "epoch": 0.3, + "learning_rate": 4.500320595024365e-05, + "loss": 0.601, + "step": 7843 + }, + { + "epoch": 0.3, + "learning_rate": 4.500256476019493e-05, + "loss": 0.6437, + "step": 7844 + }, + { + "epoch": 0.3, + "learning_rate": 4.5001923570146194e-05, + "loss": 0.4942, + "step": 7845 + }, + { + "epoch": 0.3, + "learning_rate": 4.5001282380097466e-05, + "loss": 0.5789, + "step": 7846 + }, + { + "epoch": 0.3, + "learning_rate": 4.500064119004873e-05, + "loss": 0.6079, + "step": 7847 + }, + { + "epoch": 0.3, + "learning_rate": 4.5e-05, + "loss": 0.506, + "step": 7848 + }, + { + "epoch": 0.3, + "learning_rate": 4.499935880995127e-05, + "loss": 0.4994, + "step": 7849 + }, + { + "epoch": 0.3, + "learning_rate": 4.499871761990254e-05, + "loss": 0.6539, + "step": 7850 + }, + { + "epoch": 0.3, + "learning_rate": 4.499807642985381e-05, + "loss": 0.5337, + "step": 7851 + }, + { + "epoch": 0.3, + "learning_rate": 4.499743523980508e-05, + "loss": 0.5801, + "step": 7852 + }, + { + "epoch": 0.3, + "learning_rate": 4.4996794049756355e-05, + "loss": 0.5868, + "step": 7853 + }, + { + "epoch": 0.3, + "learning_rate": 4.499615285970762e-05, + "loss": 0.4864, + "step": 7854 + }, + { + "epoch": 0.3, + "learning_rate": 4.499551166965889e-05, + "loss": 0.6435, + "step": 7855 + }, + { + "epoch": 0.3, + "learning_rate": 4.4994870479610156e-05, + "loss": 0.612, + "step": 7856 + }, + { + "epoch": 0.3, + "learning_rate": 4.499422928956143e-05, + "loss": 0.6401, + "step": 7857 + }, + { + "epoch": 0.3, + "learning_rate": 4.499358809951269e-05, + "loss": 0.5703, + "step": 7858 + }, + { + "epoch": 0.3, + "learning_rate": 4.4992946909463965e-05, + "loss": 0.5394, + "step": 7859 + }, + { + "epoch": 0.3, + "learning_rate": 4.4992305719415236e-05, + "loss": 0.5609, + "step": 7860 + }, + { + "epoch": 0.3, + "learning_rate": 4.499166452936651e-05, + "loss": 0.5253, + "step": 7861 + }, + { + "epoch": 0.3, + "learning_rate": 4.499102333931778e-05, + "loss": 0.5343, + "step": 7862 + }, + { + "epoch": 0.3, + "learning_rate": 4.4990382149269045e-05, + "loss": 0.6038, + "step": 7863 + }, + { + "epoch": 0.3, + "learning_rate": 4.4989740959220316e-05, + "loss": 0.6487, + "step": 7864 + }, + { + "epoch": 0.3, + "learning_rate": 4.498909976917158e-05, + "loss": 0.5842, + "step": 7865 + }, + { + "epoch": 0.3, + "learning_rate": 4.498845857912285e-05, + "loss": 0.601, + "step": 7866 + }, + { + "epoch": 0.3, + "learning_rate": 4.4987817389074125e-05, + "loss": 0.6425, + "step": 7867 + }, + { + "epoch": 0.3, + "learning_rate": 4.4987176199025397e-05, + "loss": 0.6695, + "step": 7868 + }, + { + "epoch": 0.3, + "learning_rate": 4.498653500897666e-05, + "loss": 0.5942, + "step": 7869 + }, + { + "epoch": 0.3, + "learning_rate": 4.498589381892793e-05, + "loss": 0.6026, + "step": 7870 + }, + { + "epoch": 0.3, + "learning_rate": 4.4985252628879205e-05, + "loss": 0.5981, + "step": 7871 + }, + { + "epoch": 0.3, + "learning_rate": 4.498461143883047e-05, + "loss": 0.5804, + "step": 7872 + }, + { + "epoch": 0.3, + "learning_rate": 4.498397024878174e-05, + "loss": 0.5739, + "step": 7873 + }, + { + "epoch": 0.3, + "learning_rate": 4.4983329058733006e-05, + "loss": 0.5229, + "step": 7874 + }, + { + "epoch": 0.3, + "learning_rate": 4.498268786868428e-05, + "loss": 0.5213, + "step": 7875 + }, + { + "epoch": 0.3, + "learning_rate": 4.498204667863555e-05, + "loss": 0.5447, + "step": 7876 + }, + { + "epoch": 0.3, + "learning_rate": 4.498140548858682e-05, + "loss": 0.555, + "step": 7877 + }, + { + "epoch": 0.3, + "learning_rate": 4.4980764298538087e-05, + "loss": 0.7162, + "step": 7878 + }, + { + "epoch": 0.3, + "learning_rate": 4.498012310848936e-05, + "loss": 0.6087, + "step": 7879 + }, + { + "epoch": 0.3, + "learning_rate": 4.497948191844063e-05, + "loss": 0.6668, + "step": 7880 + }, + { + "epoch": 0.3, + "learning_rate": 4.4978840728391895e-05, + "loss": 0.5863, + "step": 7881 + }, + { + "epoch": 0.3, + "learning_rate": 4.497819953834317e-05, + "loss": 0.5789, + "step": 7882 + }, + { + "epoch": 0.3, + "learning_rate": 4.497755834829444e-05, + "loss": 0.7341, + "step": 7883 + }, + { + "epoch": 0.3, + "learning_rate": 4.497691715824571e-05, + "loss": 0.6407, + "step": 7884 + }, + { + "epoch": 0.3, + "learning_rate": 4.4976275968196975e-05, + "loss": 0.6117, + "step": 7885 + }, + { + "epoch": 0.3, + "learning_rate": 4.497563477814825e-05, + "loss": 0.609, + "step": 7886 + }, + { + "epoch": 0.3, + "learning_rate": 4.497499358809951e-05, + "loss": 0.5364, + "step": 7887 + }, + { + "epoch": 0.3, + "learning_rate": 4.4974352398050783e-05, + "loss": 0.57, + "step": 7888 + }, + { + "epoch": 0.3, + "learning_rate": 4.4973711208002055e-05, + "loss": 0.5432, + "step": 7889 + }, + { + "epoch": 0.3, + "learning_rate": 4.497307001795332e-05, + "loss": 0.6147, + "step": 7890 + }, + { + "epoch": 0.3, + "learning_rate": 4.497242882790459e-05, + "loss": 0.5315, + "step": 7891 + }, + { + "epoch": 0.3, + "learning_rate": 4.4971787637855863e-05, + "loss": 0.5839, + "step": 7892 + }, + { + "epoch": 0.3, + "learning_rate": 4.4971146447807135e-05, + "loss": 0.6723, + "step": 7893 + }, + { + "epoch": 0.3, + "learning_rate": 4.49705052577584e-05, + "loss": 0.5123, + "step": 7894 + }, + { + "epoch": 0.3, + "learning_rate": 4.496986406770967e-05, + "loss": 0.6027, + "step": 7895 + }, + { + "epoch": 0.3, + "learning_rate": 4.496922287766094e-05, + "loss": 0.6591, + "step": 7896 + }, + { + "epoch": 0.3, + "learning_rate": 4.496858168761221e-05, + "loss": 0.5579, + "step": 7897 + }, + { + "epoch": 0.3, + "learning_rate": 4.496794049756348e-05, + "loss": 0.5352, + "step": 7898 + }, + { + "epoch": 0.3, + "learning_rate": 4.496729930751475e-05, + "loss": 0.5487, + "step": 7899 + }, + { + "epoch": 0.3, + "learning_rate": 4.4966658117466024e-05, + "loss": 0.5274, + "step": 7900 + }, + { + "epoch": 0.3, + "learning_rate": 4.496601692741729e-05, + "loss": 0.7005, + "step": 7901 + }, + { + "epoch": 0.3, + "learning_rate": 4.496537573736856e-05, + "loss": 0.6054, + "step": 7902 + }, + { + "epoch": 0.3, + "learning_rate": 4.4964734547319825e-05, + "loss": 0.5386, + "step": 7903 + }, + { + "epoch": 0.3, + "learning_rate": 4.49640933572711e-05, + "loss": 0.5626, + "step": 7904 + }, + { + "epoch": 0.3, + "learning_rate": 4.496345216722237e-05, + "loss": 0.5553, + "step": 7905 + }, + { + "epoch": 0.3, + "learning_rate": 4.4962810977173634e-05, + "loss": 0.6202, + "step": 7906 + }, + { + "epoch": 0.3, + "learning_rate": 4.4962169787124905e-05, + "loss": 0.5881, + "step": 7907 + }, + { + "epoch": 0.3, + "learning_rate": 4.496152859707618e-05, + "loss": 0.4908, + "step": 7908 + }, + { + "epoch": 0.3, + "learning_rate": 4.496088740702745e-05, + "loss": 0.5443, + "step": 7909 + }, + { + "epoch": 0.3, + "learning_rate": 4.4960246216978714e-05, + "loss": 0.5562, + "step": 7910 + }, + { + "epoch": 0.3, + "learning_rate": 4.4959605026929985e-05, + "loss": 0.5983, + "step": 7911 + }, + { + "epoch": 0.3, + "learning_rate": 4.495896383688125e-05, + "loss": 0.5698, + "step": 7912 + }, + { + "epoch": 0.3, + "learning_rate": 4.495832264683252e-05, + "loss": 0.6393, + "step": 7913 + }, + { + "epoch": 0.3, + "learning_rate": 4.4957681456783794e-05, + "loss": 0.5049, + "step": 7914 + }, + { + "epoch": 0.3, + "learning_rate": 4.495704026673506e-05, + "loss": 0.575, + "step": 7915 + }, + { + "epoch": 0.3, + "learning_rate": 4.495639907668634e-05, + "loss": 0.6949, + "step": 7916 + }, + { + "epoch": 0.3, + "learning_rate": 4.49557578866376e-05, + "loss": 0.5606, + "step": 7917 + }, + { + "epoch": 0.3, + "learning_rate": 4.4955116696588874e-05, + "loss": 0.5794, + "step": 7918 + }, + { + "epoch": 0.3, + "learning_rate": 4.495447550654014e-05, + "loss": 0.5683, + "step": 7919 + }, + { + "epoch": 0.3, + "learning_rate": 4.495383431649141e-05, + "loss": 0.5046, + "step": 7920 + }, + { + "epoch": 0.3, + "learning_rate": 4.4953193126442676e-05, + "loss": 0.4995, + "step": 7921 + }, + { + "epoch": 0.3, + "learning_rate": 4.495255193639395e-05, + "loss": 0.6432, + "step": 7922 + }, + { + "epoch": 0.3, + "learning_rate": 4.495191074634522e-05, + "loss": 0.5529, + "step": 7923 + }, + { + "epoch": 0.3, + "learning_rate": 4.495126955629649e-05, + "loss": 0.6133, + "step": 7924 + }, + { + "epoch": 0.3, + "learning_rate": 4.495062836624776e-05, + "loss": 0.5728, + "step": 7925 + }, + { + "epoch": 0.3, + "learning_rate": 4.494998717619903e-05, + "loss": 0.5106, + "step": 7926 + }, + { + "epoch": 0.3, + "learning_rate": 4.49493459861503e-05, + "loss": 0.5407, + "step": 7927 + }, + { + "epoch": 0.3, + "learning_rate": 4.4948704796101564e-05, + "loss": 0.4948, + "step": 7928 + }, + { + "epoch": 0.3, + "learning_rate": 4.4948063606052836e-05, + "loss": 0.5528, + "step": 7929 + }, + { + "epoch": 0.3, + "learning_rate": 4.49474224160041e-05, + "loss": 0.6755, + "step": 7930 + }, + { + "epoch": 0.3, + "learning_rate": 4.494678122595537e-05, + "loss": 0.6352, + "step": 7931 + }, + { + "epoch": 0.3, + "learning_rate": 4.494614003590665e-05, + "loss": 0.6138, + "step": 7932 + }, + { + "epoch": 0.3, + "learning_rate": 4.4945498845857916e-05, + "loss": 0.5008, + "step": 7933 + }, + { + "epoch": 0.31, + "learning_rate": 4.494485765580919e-05, + "loss": 0.6474, + "step": 7934 + }, + { + "epoch": 0.31, + "learning_rate": 4.494421646576045e-05, + "loss": 0.582, + "step": 7935 + }, + { + "epoch": 0.31, + "learning_rate": 4.4943575275711724e-05, + "loss": 0.4619, + "step": 7936 + }, + { + "epoch": 0.31, + "learning_rate": 4.494293408566299e-05, + "loss": 0.5617, + "step": 7937 + }, + { + "epoch": 0.31, + "learning_rate": 4.494229289561426e-05, + "loss": 0.5488, + "step": 7938 + }, + { + "epoch": 0.31, + "learning_rate": 4.494165170556553e-05, + "loss": 0.6317, + "step": 7939 + }, + { + "epoch": 0.31, + "learning_rate": 4.4941010515516804e-05, + "loss": 0.5418, + "step": 7940 + }, + { + "epoch": 0.31, + "learning_rate": 4.4940369325468076e-05, + "loss": 0.604, + "step": 7941 + }, + { + "epoch": 0.31, + "learning_rate": 4.493972813541934e-05, + "loss": 0.5178, + "step": 7942 + }, + { + "epoch": 0.31, + "learning_rate": 4.493908694537061e-05, + "loss": 0.6271, + "step": 7943 + }, + { + "epoch": 0.31, + "learning_rate": 4.493844575532188e-05, + "loss": 0.5517, + "step": 7944 + }, + { + "epoch": 0.31, + "learning_rate": 4.493780456527315e-05, + "loss": 0.5636, + "step": 7945 + }, + { + "epoch": 0.31, + "learning_rate": 4.4937163375224414e-05, + "loss": 0.6249, + "step": 7946 + }, + { + "epoch": 0.31, + "learning_rate": 4.4936522185175686e-05, + "loss": 0.5454, + "step": 7947 + }, + { + "epoch": 0.31, + "learning_rate": 4.493588099512696e-05, + "loss": 0.6215, + "step": 7948 + }, + { + "epoch": 0.31, + "learning_rate": 4.493523980507823e-05, + "loss": 0.577, + "step": 7949 + }, + { + "epoch": 0.31, + "learning_rate": 4.49345986150295e-05, + "loss": 0.5921, + "step": 7950 + }, + { + "epoch": 0.31, + "learning_rate": 4.4933957424980766e-05, + "loss": 0.5934, + "step": 7951 + }, + { + "epoch": 0.31, + "learning_rate": 4.493331623493204e-05, + "loss": 0.6501, + "step": 7952 + }, + { + "epoch": 0.31, + "learning_rate": 4.49326750448833e-05, + "loss": 0.5895, + "step": 7953 + }, + { + "epoch": 0.31, + "learning_rate": 4.4932033854834574e-05, + "loss": 0.644, + "step": 7954 + }, + { + "epoch": 0.31, + "learning_rate": 4.4931392664785846e-05, + "loss": 0.5604, + "step": 7955 + }, + { + "epoch": 0.31, + "learning_rate": 4.493075147473712e-05, + "loss": 0.5724, + "step": 7956 + }, + { + "epoch": 0.31, + "learning_rate": 4.493011028468838e-05, + "loss": 0.5848, + "step": 7957 + }, + { + "epoch": 0.31, + "learning_rate": 4.4929469094639654e-05, + "loss": 0.5728, + "step": 7958 + }, + { + "epoch": 0.31, + "learning_rate": 4.4928827904590926e-05, + "loss": 0.6116, + "step": 7959 + }, + { + "epoch": 0.31, + "learning_rate": 4.492818671454219e-05, + "loss": 0.5341, + "step": 7960 + }, + { + "epoch": 0.31, + "learning_rate": 4.492754552449346e-05, + "loss": 0.6541, + "step": 7961 + }, + { + "epoch": 0.31, + "learning_rate": 4.492690433444473e-05, + "loss": 0.6302, + "step": 7962 + }, + { + "epoch": 0.31, + "learning_rate": 4.4926263144396e-05, + "loss": 0.52, + "step": 7963 + }, + { + "epoch": 0.31, + "learning_rate": 4.492562195434727e-05, + "loss": 0.5998, + "step": 7964 + }, + { + "epoch": 0.31, + "learning_rate": 4.492498076429854e-05, + "loss": 0.6483, + "step": 7965 + }, + { + "epoch": 0.31, + "learning_rate": 4.492433957424981e-05, + "loss": 0.6551, + "step": 7966 + }, + { + "epoch": 0.31, + "learning_rate": 4.492369838420108e-05, + "loss": 0.5273, + "step": 7967 + }, + { + "epoch": 0.31, + "learning_rate": 4.492305719415235e-05, + "loss": 0.5429, + "step": 7968 + }, + { + "epoch": 0.31, + "learning_rate": 4.4922416004103616e-05, + "loss": 0.5496, + "step": 7969 + }, + { + "epoch": 0.31, + "learning_rate": 4.492177481405489e-05, + "loss": 0.6348, + "step": 7970 + }, + { + "epoch": 0.31, + "learning_rate": 4.492113362400616e-05, + "loss": 0.5014, + "step": 7971 + }, + { + "epoch": 0.31, + "learning_rate": 4.492049243395743e-05, + "loss": 0.5445, + "step": 7972 + }, + { + "epoch": 0.31, + "learning_rate": 4.4919851243908696e-05, + "loss": 0.6208, + "step": 7973 + }, + { + "epoch": 0.31, + "learning_rate": 4.491921005385997e-05, + "loss": 0.5268, + "step": 7974 + }, + { + "epoch": 0.31, + "learning_rate": 4.491856886381123e-05, + "loss": 0.5489, + "step": 7975 + }, + { + "epoch": 0.31, + "learning_rate": 4.4917927673762505e-05, + "loss": 0.6103, + "step": 7976 + }, + { + "epoch": 0.31, + "learning_rate": 4.4917286483713776e-05, + "loss": 0.5239, + "step": 7977 + }, + { + "epoch": 0.31, + "learning_rate": 4.491664529366504e-05, + "loss": 0.6258, + "step": 7978 + }, + { + "epoch": 0.31, + "learning_rate": 4.491600410361631e-05, + "loss": 0.4866, + "step": 7979 + }, + { + "epoch": 0.31, + "learning_rate": 4.4915362913567585e-05, + "loss": 0.5275, + "step": 7980 + }, + { + "epoch": 0.31, + "learning_rate": 4.4914721723518857e-05, + "loss": 0.5947, + "step": 7981 + }, + { + "epoch": 0.31, + "learning_rate": 4.491408053347012e-05, + "loss": 0.5141, + "step": 7982 + }, + { + "epoch": 0.31, + "learning_rate": 4.491343934342139e-05, + "loss": 0.6188, + "step": 7983 + }, + { + "epoch": 0.31, + "learning_rate": 4.491279815337266e-05, + "loss": 0.5082, + "step": 7984 + }, + { + "epoch": 0.31, + "learning_rate": 4.491215696332393e-05, + "loss": 0.5365, + "step": 7985 + }, + { + "epoch": 0.31, + "learning_rate": 4.49115157732752e-05, + "loss": 0.5742, + "step": 7986 + }, + { + "epoch": 0.31, + "learning_rate": 4.491087458322647e-05, + "loss": 0.4871, + "step": 7987 + }, + { + "epoch": 0.31, + "learning_rate": 4.4910233393177745e-05, + "loss": 0.6408, + "step": 7988 + }, + { + "epoch": 0.31, + "learning_rate": 4.490959220312901e-05, + "loss": 0.5577, + "step": 7989 + }, + { + "epoch": 0.31, + "learning_rate": 4.490895101308028e-05, + "loss": 0.5956, + "step": 7990 + }, + { + "epoch": 0.31, + "learning_rate": 4.4908309823031547e-05, + "loss": 0.633, + "step": 7991 + }, + { + "epoch": 0.31, + "learning_rate": 4.490766863298282e-05, + "loss": 0.6256, + "step": 7992 + }, + { + "epoch": 0.31, + "learning_rate": 4.490702744293408e-05, + "loss": 0.5414, + "step": 7993 + }, + { + "epoch": 0.31, + "learning_rate": 4.4906386252885355e-05, + "loss": 0.5722, + "step": 7994 + }, + { + "epoch": 0.31, + "learning_rate": 4.490574506283663e-05, + "loss": 0.5787, + "step": 7995 + }, + { + "epoch": 0.31, + "learning_rate": 4.49051038727879e-05, + "loss": 0.652, + "step": 7996 + }, + { + "epoch": 0.31, + "learning_rate": 4.490446268273917e-05, + "loss": 0.4981, + "step": 7997 + }, + { + "epoch": 0.31, + "learning_rate": 4.4903821492690435e-05, + "loss": 0.6042, + "step": 7998 + }, + { + "epoch": 0.31, + "learning_rate": 4.490318030264171e-05, + "loss": 0.5907, + "step": 7999 + }, + { + "epoch": 0.31, + "learning_rate": 4.490253911259297e-05, + "loss": 0.6757, + "step": 8000 + }, + { + "epoch": 0.31, + "learning_rate": 4.4901897922544243e-05, + "loss": 0.6341, + "step": 8001 + }, + { + "epoch": 0.31, + "learning_rate": 4.490125673249551e-05, + "loss": 0.5404, + "step": 8002 + }, + { + "epoch": 0.31, + "learning_rate": 4.490061554244678e-05, + "loss": 0.6753, + "step": 8003 + }, + { + "epoch": 0.31, + "learning_rate": 4.489997435239806e-05, + "loss": 0.6099, + "step": 8004 + }, + { + "epoch": 0.31, + "learning_rate": 4.4899333162349324e-05, + "loss": 0.6006, + "step": 8005 + }, + { + "epoch": 0.31, + "learning_rate": 4.4898691972300595e-05, + "loss": 0.4657, + "step": 8006 + }, + { + "epoch": 0.31, + "learning_rate": 4.489805078225186e-05, + "loss": 0.5104, + "step": 8007 + }, + { + "epoch": 0.31, + "learning_rate": 4.489740959220313e-05, + "loss": 0.5161, + "step": 8008 + }, + { + "epoch": 0.31, + "learning_rate": 4.48967684021544e-05, + "loss": 0.6295, + "step": 8009 + }, + { + "epoch": 0.31, + "learning_rate": 4.489612721210567e-05, + "loss": 0.5003, + "step": 8010 + }, + { + "epoch": 0.31, + "learning_rate": 4.489548602205694e-05, + "loss": 0.5675, + "step": 8011 + }, + { + "epoch": 0.31, + "learning_rate": 4.489484483200821e-05, + "loss": 0.6212, + "step": 8012 + }, + { + "epoch": 0.31, + "learning_rate": 4.4894203641959484e-05, + "loss": 0.54, + "step": 8013 + }, + { + "epoch": 0.31, + "learning_rate": 4.489356245191075e-05, + "loss": 0.6086, + "step": 8014 + }, + { + "epoch": 0.31, + "learning_rate": 4.489292126186202e-05, + "loss": 0.4813, + "step": 8015 + }, + { + "epoch": 0.31, + "learning_rate": 4.4892280071813285e-05, + "loss": 0.563, + "step": 8016 + }, + { + "epoch": 0.31, + "learning_rate": 4.489163888176456e-05, + "loss": 0.5371, + "step": 8017 + }, + { + "epoch": 0.31, + "learning_rate": 4.489099769171582e-05, + "loss": 0.5178, + "step": 8018 + }, + { + "epoch": 0.31, + "learning_rate": 4.4890356501667094e-05, + "loss": 0.5464, + "step": 8019 + }, + { + "epoch": 0.31, + "learning_rate": 4.4889715311618365e-05, + "loss": 0.6123, + "step": 8020 + }, + { + "epoch": 0.31, + "learning_rate": 4.488907412156964e-05, + "loss": 0.6079, + "step": 8021 + }, + { + "epoch": 0.31, + "learning_rate": 4.488843293152091e-05, + "loss": 0.7197, + "step": 8022 + }, + { + "epoch": 0.31, + "learning_rate": 4.4887791741472174e-05, + "loss": 0.5317, + "step": 8023 + }, + { + "epoch": 0.31, + "learning_rate": 4.4887150551423445e-05, + "loss": 0.6835, + "step": 8024 + }, + { + "epoch": 0.31, + "learning_rate": 4.488650936137471e-05, + "loss": 0.5951, + "step": 8025 + }, + { + "epoch": 0.31, + "learning_rate": 4.488586817132598e-05, + "loss": 0.571, + "step": 8026 + }, + { + "epoch": 0.31, + "learning_rate": 4.4885226981277254e-05, + "loss": 0.6278, + "step": 8027 + }, + { + "epoch": 0.31, + "learning_rate": 4.4884585791228526e-05, + "loss": 0.4994, + "step": 8028 + }, + { + "epoch": 0.31, + "learning_rate": 4.488394460117979e-05, + "loss": 0.6269, + "step": 8029 + }, + { + "epoch": 0.31, + "learning_rate": 4.488330341113106e-05, + "loss": 0.6056, + "step": 8030 + }, + { + "epoch": 0.31, + "learning_rate": 4.4882662221082334e-05, + "loss": 0.6192, + "step": 8031 + }, + { + "epoch": 0.31, + "learning_rate": 4.48820210310336e-05, + "loss": 0.5737, + "step": 8032 + }, + { + "epoch": 0.31, + "learning_rate": 4.488137984098487e-05, + "loss": 0.6227, + "step": 8033 + }, + { + "epoch": 0.31, + "learning_rate": 4.4880738650936136e-05, + "loss": 0.5843, + "step": 8034 + }, + { + "epoch": 0.31, + "learning_rate": 4.488009746088741e-05, + "loss": 0.6222, + "step": 8035 + }, + { + "epoch": 0.31, + "learning_rate": 4.487945627083868e-05, + "loss": 0.5366, + "step": 8036 + }, + { + "epoch": 0.31, + "learning_rate": 4.487881508078995e-05, + "loss": 0.593, + "step": 8037 + }, + { + "epoch": 0.31, + "learning_rate": 4.4878173890741216e-05, + "loss": 0.5774, + "step": 8038 + }, + { + "epoch": 0.31, + "learning_rate": 4.487753270069249e-05, + "loss": 0.5598, + "step": 8039 + }, + { + "epoch": 0.31, + "learning_rate": 4.487689151064376e-05, + "loss": 0.5484, + "step": 8040 + }, + { + "epoch": 0.31, + "learning_rate": 4.4876250320595024e-05, + "loss": 0.598, + "step": 8041 + }, + { + "epoch": 0.31, + "learning_rate": 4.4875609130546296e-05, + "loss": 0.5789, + "step": 8042 + }, + { + "epoch": 0.31, + "learning_rate": 4.487496794049757e-05, + "loss": 0.5757, + "step": 8043 + }, + { + "epoch": 0.31, + "learning_rate": 4.487432675044884e-05, + "loss": 0.5971, + "step": 8044 + }, + { + "epoch": 0.31, + "learning_rate": 4.4873685560400104e-05, + "loss": 0.4949, + "step": 8045 + }, + { + "epoch": 0.31, + "learning_rate": 4.4873044370351376e-05, + "loss": 0.5898, + "step": 8046 + }, + { + "epoch": 0.31, + "learning_rate": 4.487240318030264e-05, + "loss": 0.4917, + "step": 8047 + }, + { + "epoch": 0.31, + "learning_rate": 4.487176199025391e-05, + "loss": 0.5574, + "step": 8048 + }, + { + "epoch": 0.31, + "learning_rate": 4.4871120800205184e-05, + "loss": 0.584, + "step": 8049 + }, + { + "epoch": 0.31, + "learning_rate": 4.487047961015645e-05, + "loss": 0.5325, + "step": 8050 + }, + { + "epoch": 0.31, + "learning_rate": 4.486983842010772e-05, + "loss": 0.5479, + "step": 8051 + }, + { + "epoch": 0.31, + "learning_rate": 4.486919723005899e-05, + "loss": 0.6836, + "step": 8052 + }, + { + "epoch": 0.31, + "learning_rate": 4.4868556040010264e-05, + "loss": 0.6763, + "step": 8053 + }, + { + "epoch": 0.31, + "learning_rate": 4.486791484996153e-05, + "loss": 0.5854, + "step": 8054 + }, + { + "epoch": 0.31, + "learning_rate": 4.48672736599128e-05, + "loss": 0.5723, + "step": 8055 + }, + { + "epoch": 0.31, + "learning_rate": 4.4866632469864066e-05, + "loss": 0.4701, + "step": 8056 + }, + { + "epoch": 0.31, + "learning_rate": 4.486599127981534e-05, + "loss": 0.6007, + "step": 8057 + }, + { + "epoch": 0.31, + "learning_rate": 4.486535008976661e-05, + "loss": 0.516, + "step": 8058 + }, + { + "epoch": 0.31, + "learning_rate": 4.486470889971788e-05, + "loss": 0.5491, + "step": 8059 + }, + { + "epoch": 0.31, + "learning_rate": 4.486406770966915e-05, + "loss": 0.7217, + "step": 8060 + }, + { + "epoch": 0.31, + "learning_rate": 4.486342651962042e-05, + "loss": 0.6107, + "step": 8061 + }, + { + "epoch": 0.31, + "learning_rate": 4.486278532957169e-05, + "loss": 0.4746, + "step": 8062 + }, + { + "epoch": 0.31, + "learning_rate": 4.4862144139522954e-05, + "loss": 0.6077, + "step": 8063 + }, + { + "epoch": 0.31, + "learning_rate": 4.4861502949474226e-05, + "loss": 0.5666, + "step": 8064 + }, + { + "epoch": 0.31, + "learning_rate": 4.486086175942549e-05, + "loss": 0.5012, + "step": 8065 + }, + { + "epoch": 0.31, + "learning_rate": 4.486022056937676e-05, + "loss": 0.5418, + "step": 8066 + }, + { + "epoch": 0.31, + "learning_rate": 4.4859579379328034e-05, + "loss": 0.5232, + "step": 8067 + }, + { + "epoch": 0.31, + "learning_rate": 4.4858938189279306e-05, + "loss": 0.4972, + "step": 8068 + }, + { + "epoch": 0.31, + "learning_rate": 4.485829699923058e-05, + "loss": 0.5056, + "step": 8069 + }, + { + "epoch": 0.31, + "learning_rate": 4.485765580918184e-05, + "loss": 0.545, + "step": 8070 + }, + { + "epoch": 0.31, + "learning_rate": 4.4857014619133115e-05, + "loss": 0.5964, + "step": 8071 + }, + { + "epoch": 0.31, + "learning_rate": 4.485637342908438e-05, + "loss": 0.5258, + "step": 8072 + }, + { + "epoch": 0.31, + "learning_rate": 4.485573223903565e-05, + "loss": 0.5935, + "step": 8073 + }, + { + "epoch": 0.31, + "learning_rate": 4.4855091048986916e-05, + "loss": 0.5552, + "step": 8074 + }, + { + "epoch": 0.31, + "learning_rate": 4.485444985893819e-05, + "loss": 0.5373, + "step": 8075 + }, + { + "epoch": 0.31, + "learning_rate": 4.4853808668889466e-05, + "loss": 0.5589, + "step": 8076 + }, + { + "epoch": 0.31, + "learning_rate": 4.485316747884073e-05, + "loss": 0.5742, + "step": 8077 + }, + { + "epoch": 0.31, + "learning_rate": 4.4852526288792e-05, + "loss": 0.6514, + "step": 8078 + }, + { + "epoch": 0.31, + "learning_rate": 4.485188509874327e-05, + "loss": 0.6942, + "step": 8079 + }, + { + "epoch": 0.31, + "learning_rate": 4.485124390869454e-05, + "loss": 0.5813, + "step": 8080 + }, + { + "epoch": 0.31, + "learning_rate": 4.4850602718645805e-05, + "loss": 0.5412, + "step": 8081 + }, + { + "epoch": 0.31, + "learning_rate": 4.4849961528597076e-05, + "loss": 0.5254, + "step": 8082 + }, + { + "epoch": 0.31, + "learning_rate": 4.484932033854835e-05, + "loss": 0.5163, + "step": 8083 + }, + { + "epoch": 0.31, + "learning_rate": 4.484867914849962e-05, + "loss": 0.6308, + "step": 8084 + }, + { + "epoch": 0.31, + "learning_rate": 4.484803795845089e-05, + "loss": 0.604, + "step": 8085 + }, + { + "epoch": 0.31, + "learning_rate": 4.4847396768402156e-05, + "loss": 0.6547, + "step": 8086 + }, + { + "epoch": 0.31, + "learning_rate": 4.484675557835343e-05, + "loss": 0.5074, + "step": 8087 + }, + { + "epoch": 0.31, + "learning_rate": 4.484611438830469e-05, + "loss": 0.5178, + "step": 8088 + }, + { + "epoch": 0.31, + "learning_rate": 4.4845473198255965e-05, + "loss": 0.6542, + "step": 8089 + }, + { + "epoch": 0.31, + "learning_rate": 4.484483200820723e-05, + "loss": 0.5826, + "step": 8090 + }, + { + "epoch": 0.31, + "learning_rate": 4.48441908181585e-05, + "loss": 0.5467, + "step": 8091 + }, + { + "epoch": 0.31, + "learning_rate": 4.484354962810977e-05, + "loss": 0.6082, + "step": 8092 + }, + { + "epoch": 0.31, + "learning_rate": 4.4842908438061045e-05, + "loss": 0.6188, + "step": 8093 + }, + { + "epoch": 0.31, + "learning_rate": 4.4842267248012317e-05, + "loss": 0.6089, + "step": 8094 + }, + { + "epoch": 0.31, + "learning_rate": 4.484162605796358e-05, + "loss": 0.5371, + "step": 8095 + }, + { + "epoch": 0.31, + "learning_rate": 4.484098486791485e-05, + "loss": 0.5451, + "step": 8096 + }, + { + "epoch": 0.31, + "learning_rate": 4.484034367786612e-05, + "loss": 0.5527, + "step": 8097 + }, + { + "epoch": 0.31, + "learning_rate": 4.483970248781739e-05, + "loss": 0.5771, + "step": 8098 + }, + { + "epoch": 0.31, + "learning_rate": 4.483906129776866e-05, + "loss": 0.5769, + "step": 8099 + }, + { + "epoch": 0.31, + "learning_rate": 4.483842010771993e-05, + "loss": 0.6203, + "step": 8100 + }, + { + "epoch": 0.31, + "learning_rate": 4.4837778917671205e-05, + "loss": 0.5271, + "step": 8101 + }, + { + "epoch": 0.31, + "learning_rate": 4.483713772762247e-05, + "loss": 0.5474, + "step": 8102 + }, + { + "epoch": 0.31, + "learning_rate": 4.483649653757374e-05, + "loss": 0.6173, + "step": 8103 + }, + { + "epoch": 0.31, + "learning_rate": 4.483585534752501e-05, + "loss": 0.6387, + "step": 8104 + }, + { + "epoch": 0.31, + "learning_rate": 4.483521415747628e-05, + "loss": 0.6678, + "step": 8105 + }, + { + "epoch": 0.31, + "learning_rate": 4.483457296742754e-05, + "loss": 0.5991, + "step": 8106 + }, + { + "epoch": 0.31, + "learning_rate": 4.4833931777378815e-05, + "loss": 0.4804, + "step": 8107 + }, + { + "epoch": 0.31, + "learning_rate": 4.483329058733009e-05, + "loss": 0.5481, + "step": 8108 + }, + { + "epoch": 0.31, + "learning_rate": 4.483264939728136e-05, + "loss": 0.5444, + "step": 8109 + }, + { + "epoch": 0.31, + "learning_rate": 4.483200820723263e-05, + "loss": 0.5514, + "step": 8110 + }, + { + "epoch": 0.31, + "learning_rate": 4.4831367017183895e-05, + "loss": 0.5943, + "step": 8111 + }, + { + "epoch": 0.31, + "learning_rate": 4.483072582713517e-05, + "loss": 0.6629, + "step": 8112 + }, + { + "epoch": 0.31, + "learning_rate": 4.483008463708643e-05, + "loss": 0.5644, + "step": 8113 + }, + { + "epoch": 0.31, + "learning_rate": 4.4829443447037703e-05, + "loss": 0.5942, + "step": 8114 + }, + { + "epoch": 0.31, + "learning_rate": 4.4828802256988975e-05, + "loss": 0.6698, + "step": 8115 + }, + { + "epoch": 0.31, + "learning_rate": 4.482816106694025e-05, + "loss": 0.4247, + "step": 8116 + }, + { + "epoch": 0.31, + "learning_rate": 4.482751987689151e-05, + "loss": 0.5993, + "step": 8117 + }, + { + "epoch": 0.31, + "learning_rate": 4.4826878686842784e-05, + "loss": 0.5871, + "step": 8118 + }, + { + "epoch": 0.31, + "learning_rate": 4.4826237496794055e-05, + "loss": 0.5905, + "step": 8119 + }, + { + "epoch": 0.31, + "learning_rate": 4.482559630674532e-05, + "loss": 0.6865, + "step": 8120 + }, + { + "epoch": 0.31, + "learning_rate": 4.482495511669659e-05, + "loss": 0.5768, + "step": 8121 + }, + { + "epoch": 0.31, + "learning_rate": 4.482431392664786e-05, + "loss": 0.5253, + "step": 8122 + }, + { + "epoch": 0.31, + "learning_rate": 4.482367273659913e-05, + "loss": 0.5333, + "step": 8123 + }, + { + "epoch": 0.31, + "learning_rate": 4.48230315465504e-05, + "loss": 0.5559, + "step": 8124 + }, + { + "epoch": 0.31, + "learning_rate": 4.482239035650167e-05, + "loss": 0.5565, + "step": 8125 + }, + { + "epoch": 0.31, + "learning_rate": 4.482174916645294e-05, + "loss": 0.5857, + "step": 8126 + }, + { + "epoch": 0.31, + "learning_rate": 4.482110797640421e-05, + "loss": 0.5733, + "step": 8127 + }, + { + "epoch": 0.31, + "learning_rate": 4.482046678635548e-05, + "loss": 0.611, + "step": 8128 + }, + { + "epoch": 0.31, + "learning_rate": 4.4819825596306745e-05, + "loss": 0.6061, + "step": 8129 + }, + { + "epoch": 0.31, + "learning_rate": 4.481918440625802e-05, + "loss": 0.6762, + "step": 8130 + }, + { + "epoch": 0.31, + "learning_rate": 4.481854321620929e-05, + "loss": 0.5641, + "step": 8131 + }, + { + "epoch": 0.31, + "learning_rate": 4.481790202616056e-05, + "loss": 0.6341, + "step": 8132 + }, + { + "epoch": 0.31, + "learning_rate": 4.4817260836111825e-05, + "loss": 0.5858, + "step": 8133 + }, + { + "epoch": 0.31, + "learning_rate": 4.48166196460631e-05, + "loss": 0.4921, + "step": 8134 + }, + { + "epoch": 0.31, + "learning_rate": 4.481597845601436e-05, + "loss": 0.521, + "step": 8135 + }, + { + "epoch": 0.31, + "learning_rate": 4.4815337265965634e-05, + "loss": 0.4781, + "step": 8136 + }, + { + "epoch": 0.31, + "learning_rate": 4.4814696075916906e-05, + "loss": 0.5219, + "step": 8137 + }, + { + "epoch": 0.31, + "learning_rate": 4.481405488586817e-05, + "loss": 0.5842, + "step": 8138 + }, + { + "epoch": 0.31, + "learning_rate": 4.481341369581944e-05, + "loss": 0.5746, + "step": 8139 + }, + { + "epoch": 0.31, + "learning_rate": 4.4812772505770714e-05, + "loss": 0.6069, + "step": 8140 + }, + { + "epoch": 0.31, + "learning_rate": 4.4812131315721986e-05, + "loss": 0.4972, + "step": 8141 + }, + { + "epoch": 0.31, + "learning_rate": 4.481149012567325e-05, + "loss": 0.5348, + "step": 8142 + }, + { + "epoch": 0.31, + "learning_rate": 4.481084893562452e-05, + "loss": 0.5891, + "step": 8143 + }, + { + "epoch": 0.31, + "learning_rate": 4.481020774557579e-05, + "loss": 0.5876, + "step": 8144 + }, + { + "epoch": 0.31, + "learning_rate": 4.480956655552706e-05, + "loss": 0.5551, + "step": 8145 + }, + { + "epoch": 0.31, + "learning_rate": 4.480892536547833e-05, + "loss": 0.5624, + "step": 8146 + }, + { + "epoch": 0.31, + "learning_rate": 4.48082841754296e-05, + "loss": 0.6297, + "step": 8147 + }, + { + "epoch": 0.31, + "learning_rate": 4.4807642985380874e-05, + "loss": 0.5742, + "step": 8148 + }, + { + "epoch": 0.31, + "learning_rate": 4.480700179533214e-05, + "loss": 0.4821, + "step": 8149 + }, + { + "epoch": 0.31, + "learning_rate": 4.480636060528341e-05, + "loss": 0.6965, + "step": 8150 + }, + { + "epoch": 0.31, + "learning_rate": 4.4805719415234676e-05, + "loss": 0.5739, + "step": 8151 + }, + { + "epoch": 0.31, + "learning_rate": 4.480507822518595e-05, + "loss": 0.6721, + "step": 8152 + }, + { + "epoch": 0.31, + "learning_rate": 4.480443703513721e-05, + "loss": 0.5875, + "step": 8153 + }, + { + "epoch": 0.31, + "learning_rate": 4.4803795845088484e-05, + "loss": 0.569, + "step": 8154 + }, + { + "epoch": 0.31, + "learning_rate": 4.4803154655039756e-05, + "loss": 0.6399, + "step": 8155 + }, + { + "epoch": 0.31, + "learning_rate": 4.480251346499103e-05, + "loss": 0.5666, + "step": 8156 + }, + { + "epoch": 0.31, + "learning_rate": 4.48018722749423e-05, + "loss": 0.5987, + "step": 8157 + }, + { + "epoch": 0.31, + "learning_rate": 4.4801231084893564e-05, + "loss": 0.6114, + "step": 8158 + }, + { + "epoch": 0.31, + "learning_rate": 4.4800589894844836e-05, + "loss": 0.6721, + "step": 8159 + }, + { + "epoch": 0.31, + "learning_rate": 4.47999487047961e-05, + "loss": 0.5721, + "step": 8160 + }, + { + "epoch": 0.31, + "learning_rate": 4.479930751474737e-05, + "loss": 0.5556, + "step": 8161 + }, + { + "epoch": 0.31, + "learning_rate": 4.479866632469864e-05, + "loss": 0.5873, + "step": 8162 + }, + { + "epoch": 0.31, + "learning_rate": 4.479802513464991e-05, + "loss": 0.5691, + "step": 8163 + }, + { + "epoch": 0.31, + "learning_rate": 4.479738394460119e-05, + "loss": 0.6191, + "step": 8164 + }, + { + "epoch": 0.31, + "learning_rate": 4.479674275455245e-05, + "loss": 0.6163, + "step": 8165 + }, + { + "epoch": 0.31, + "learning_rate": 4.4796101564503724e-05, + "loss": 0.5712, + "step": 8166 + }, + { + "epoch": 0.31, + "learning_rate": 4.479546037445499e-05, + "loss": 0.5794, + "step": 8167 + }, + { + "epoch": 0.31, + "learning_rate": 4.479481918440626e-05, + "loss": 0.5425, + "step": 8168 + }, + { + "epoch": 0.31, + "learning_rate": 4.4794177994357526e-05, + "loss": 0.4934, + "step": 8169 + }, + { + "epoch": 0.31, + "learning_rate": 4.47935368043088e-05, + "loss": 0.5818, + "step": 8170 + }, + { + "epoch": 0.31, + "learning_rate": 4.479289561426007e-05, + "loss": 0.6712, + "step": 8171 + }, + { + "epoch": 0.31, + "learning_rate": 4.479225442421134e-05, + "loss": 0.5214, + "step": 8172 + }, + { + "epoch": 0.31, + "learning_rate": 4.479161323416261e-05, + "loss": 0.5627, + "step": 8173 + }, + { + "epoch": 0.31, + "learning_rate": 4.479097204411388e-05, + "loss": 0.5664, + "step": 8174 + }, + { + "epoch": 0.31, + "learning_rate": 4.479033085406515e-05, + "loss": 0.5956, + "step": 8175 + }, + { + "epoch": 0.31, + "learning_rate": 4.4789689664016414e-05, + "loss": 0.5244, + "step": 8176 + }, + { + "epoch": 0.31, + "learning_rate": 4.4789048473967686e-05, + "loss": 0.5792, + "step": 8177 + }, + { + "epoch": 0.31, + "learning_rate": 4.478840728391895e-05, + "loss": 0.5414, + "step": 8178 + }, + { + "epoch": 0.31, + "learning_rate": 4.478776609387022e-05, + "loss": 0.4889, + "step": 8179 + }, + { + "epoch": 0.31, + "learning_rate": 4.4787124903821494e-05, + "loss": 0.6444, + "step": 8180 + }, + { + "epoch": 0.31, + "learning_rate": 4.4786483713772766e-05, + "loss": 0.5717, + "step": 8181 + }, + { + "epoch": 0.31, + "learning_rate": 4.478584252372404e-05, + "loss": 0.5337, + "step": 8182 + }, + { + "epoch": 0.31, + "learning_rate": 4.47852013336753e-05, + "loss": 0.4915, + "step": 8183 + }, + { + "epoch": 0.31, + "learning_rate": 4.4784560143626575e-05, + "loss": 0.5958, + "step": 8184 + }, + { + "epoch": 0.31, + "learning_rate": 4.478391895357784e-05, + "loss": 0.5193, + "step": 8185 + }, + { + "epoch": 0.31, + "learning_rate": 4.478327776352911e-05, + "loss": 0.5253, + "step": 8186 + }, + { + "epoch": 0.31, + "learning_rate": 4.478263657348038e-05, + "loss": 0.6022, + "step": 8187 + }, + { + "epoch": 0.31, + "learning_rate": 4.4781995383431655e-05, + "loss": 0.5115, + "step": 8188 + }, + { + "epoch": 0.31, + "learning_rate": 4.478135419338292e-05, + "loss": 0.486, + "step": 8189 + }, + { + "epoch": 0.31, + "learning_rate": 4.478071300333419e-05, + "loss": 0.538, + "step": 8190 + }, + { + "epoch": 0.31, + "learning_rate": 4.478007181328546e-05, + "loss": 0.5051, + "step": 8191 + }, + { + "epoch": 0.31, + "learning_rate": 4.477943062323673e-05, + "loss": 0.5134, + "step": 8192 + }, + { + "epoch": 0.31, + "learning_rate": 4.4778789433188e-05, + "loss": 0.4847, + "step": 8193 + }, + { + "epoch": 0.32, + "learning_rate": 4.4778148243139265e-05, + "loss": 0.5398, + "step": 8194 + }, + { + "epoch": 0.32, + "learning_rate": 4.4777507053090536e-05, + "loss": 0.5702, + "step": 8195 + }, + { + "epoch": 0.32, + "learning_rate": 4.477686586304181e-05, + "loss": 0.5464, + "step": 8196 + }, + { + "epoch": 0.32, + "learning_rate": 4.477622467299308e-05, + "loss": 0.7215, + "step": 8197 + }, + { + "epoch": 0.32, + "learning_rate": 4.4775583482944345e-05, + "loss": 0.61, + "step": 8198 + }, + { + "epoch": 0.32, + "learning_rate": 4.4774942292895616e-05, + "loss": 0.5528, + "step": 8199 + }, + { + "epoch": 0.32, + "learning_rate": 4.477430110284689e-05, + "loss": 0.4328, + "step": 8200 + }, + { + "epoch": 0.32, + "learning_rate": 4.477365991279815e-05, + "loss": 0.6461, + "step": 8201 + }, + { + "epoch": 0.32, + "learning_rate": 4.4773018722749425e-05, + "loss": 0.6188, + "step": 8202 + }, + { + "epoch": 0.32, + "learning_rate": 4.4772377532700697e-05, + "loss": 0.5965, + "step": 8203 + }, + { + "epoch": 0.32, + "learning_rate": 4.477173634265197e-05, + "loss": 0.5592, + "step": 8204 + }, + { + "epoch": 0.32, + "learning_rate": 4.477109515260323e-05, + "loss": 0.6207, + "step": 8205 + }, + { + "epoch": 0.32, + "learning_rate": 4.4770453962554505e-05, + "loss": 0.5645, + "step": 8206 + }, + { + "epoch": 0.32, + "learning_rate": 4.476981277250577e-05, + "loss": 0.5914, + "step": 8207 + }, + { + "epoch": 0.32, + "learning_rate": 4.476917158245704e-05, + "loss": 0.6889, + "step": 8208 + }, + { + "epoch": 0.32, + "learning_rate": 4.476853039240831e-05, + "loss": 0.6069, + "step": 8209 + }, + { + "epoch": 0.32, + "learning_rate": 4.476788920235958e-05, + "loss": 0.5072, + "step": 8210 + }, + { + "epoch": 0.32, + "learning_rate": 4.476724801231085e-05, + "loss": 0.6217, + "step": 8211 + }, + { + "epoch": 0.32, + "learning_rate": 4.476660682226212e-05, + "loss": 0.507, + "step": 8212 + }, + { + "epoch": 0.32, + "learning_rate": 4.476596563221339e-05, + "loss": 0.6289, + "step": 8213 + }, + { + "epoch": 0.32, + "learning_rate": 4.476532444216466e-05, + "loss": 0.5962, + "step": 8214 + }, + { + "epoch": 0.32, + "learning_rate": 4.476468325211593e-05, + "loss": 0.5978, + "step": 8215 + }, + { + "epoch": 0.32, + "learning_rate": 4.4764042062067195e-05, + "loss": 0.5786, + "step": 8216 + }, + { + "epoch": 0.32, + "learning_rate": 4.476340087201847e-05, + "loss": 0.6396, + "step": 8217 + }, + { + "epoch": 0.32, + "learning_rate": 4.476275968196974e-05, + "loss": 0.5596, + "step": 8218 + }, + { + "epoch": 0.32, + "learning_rate": 4.476211849192101e-05, + "loss": 0.6311, + "step": 8219 + }, + { + "epoch": 0.32, + "learning_rate": 4.476147730187228e-05, + "loss": 0.5974, + "step": 8220 + }, + { + "epoch": 0.32, + "learning_rate": 4.476083611182355e-05, + "loss": 0.5765, + "step": 8221 + }, + { + "epoch": 0.32, + "learning_rate": 4.476019492177482e-05, + "loss": 0.5754, + "step": 8222 + }, + { + "epoch": 0.32, + "learning_rate": 4.4759553731726083e-05, + "loss": 0.5315, + "step": 8223 + }, + { + "epoch": 0.32, + "learning_rate": 4.4758912541677355e-05, + "loss": 0.5359, + "step": 8224 + }, + { + "epoch": 0.32, + "learning_rate": 4.475827135162862e-05, + "loss": 0.6352, + "step": 8225 + }, + { + "epoch": 0.32, + "learning_rate": 4.475763016157989e-05, + "loss": 0.5912, + "step": 8226 + }, + { + "epoch": 0.32, + "learning_rate": 4.4756988971531164e-05, + "loss": 0.5405, + "step": 8227 + }, + { + "epoch": 0.32, + "learning_rate": 4.4756347781482435e-05, + "loss": 0.6025, + "step": 8228 + }, + { + "epoch": 0.32, + "learning_rate": 4.475570659143371e-05, + "loss": 0.5438, + "step": 8229 + }, + { + "epoch": 0.32, + "learning_rate": 4.475506540138497e-05, + "loss": 0.6106, + "step": 8230 + }, + { + "epoch": 0.32, + "learning_rate": 4.4754424211336244e-05, + "loss": 0.563, + "step": 8231 + }, + { + "epoch": 0.32, + "learning_rate": 4.475378302128751e-05, + "loss": 0.5386, + "step": 8232 + }, + { + "epoch": 0.32, + "learning_rate": 4.475314183123878e-05, + "loss": 0.5133, + "step": 8233 + }, + { + "epoch": 0.32, + "learning_rate": 4.4752500641190045e-05, + "loss": 0.623, + "step": 8234 + }, + { + "epoch": 0.32, + "learning_rate": 4.475185945114132e-05, + "loss": 0.4978, + "step": 8235 + }, + { + "epoch": 0.32, + "learning_rate": 4.4751218261092595e-05, + "loss": 0.5696, + "step": 8236 + }, + { + "epoch": 0.32, + "learning_rate": 4.475057707104386e-05, + "loss": 0.5322, + "step": 8237 + }, + { + "epoch": 0.32, + "learning_rate": 4.474993588099513e-05, + "loss": 0.4814, + "step": 8238 + }, + { + "epoch": 0.32, + "learning_rate": 4.47492946909464e-05, + "loss": 0.5864, + "step": 8239 + }, + { + "epoch": 0.32, + "learning_rate": 4.474865350089767e-05, + "loss": 0.5157, + "step": 8240 + }, + { + "epoch": 0.32, + "learning_rate": 4.4748012310848934e-05, + "loss": 0.5481, + "step": 8241 + }, + { + "epoch": 0.32, + "learning_rate": 4.4747371120800205e-05, + "loss": 0.5176, + "step": 8242 + }, + { + "epoch": 0.32, + "learning_rate": 4.474672993075148e-05, + "loss": 0.5171, + "step": 8243 + }, + { + "epoch": 0.32, + "learning_rate": 4.474608874070275e-05, + "loss": 0.6355, + "step": 8244 + }, + { + "epoch": 0.32, + "learning_rate": 4.474544755065402e-05, + "loss": 0.6683, + "step": 8245 + }, + { + "epoch": 0.32, + "learning_rate": 4.4744806360605285e-05, + "loss": 0.6355, + "step": 8246 + }, + { + "epoch": 0.32, + "learning_rate": 4.474416517055656e-05, + "loss": 0.7829, + "step": 8247 + }, + { + "epoch": 0.32, + "learning_rate": 4.474352398050782e-05, + "loss": 0.5261, + "step": 8248 + }, + { + "epoch": 0.32, + "learning_rate": 4.4742882790459094e-05, + "loss": 0.49, + "step": 8249 + }, + { + "epoch": 0.32, + "learning_rate": 4.474224160041036e-05, + "loss": 0.5628, + "step": 8250 + }, + { + "epoch": 0.32, + "learning_rate": 4.474160041036163e-05, + "loss": 0.6591, + "step": 8251 + }, + { + "epoch": 0.32, + "learning_rate": 4.47409592203129e-05, + "loss": 0.5568, + "step": 8252 + }, + { + "epoch": 0.32, + "learning_rate": 4.4740318030264174e-05, + "loss": 0.562, + "step": 8253 + }, + { + "epoch": 0.32, + "learning_rate": 4.4739676840215446e-05, + "loss": 0.4864, + "step": 8254 + }, + { + "epoch": 0.32, + "learning_rate": 4.473903565016671e-05, + "loss": 0.5893, + "step": 8255 + }, + { + "epoch": 0.32, + "learning_rate": 4.473839446011798e-05, + "loss": 0.5052, + "step": 8256 + }, + { + "epoch": 0.32, + "learning_rate": 4.473775327006925e-05, + "loss": 0.4749, + "step": 8257 + }, + { + "epoch": 0.32, + "learning_rate": 4.473711208002052e-05, + "loss": 0.565, + "step": 8258 + }, + { + "epoch": 0.32, + "learning_rate": 4.473647088997179e-05, + "loss": 0.6189, + "step": 8259 + }, + { + "epoch": 0.32, + "learning_rate": 4.473582969992306e-05, + "loss": 0.6117, + "step": 8260 + }, + { + "epoch": 0.32, + "learning_rate": 4.473518850987433e-05, + "loss": 0.5794, + "step": 8261 + }, + { + "epoch": 0.32, + "learning_rate": 4.47345473198256e-05, + "loss": 0.4981, + "step": 8262 + }, + { + "epoch": 0.32, + "learning_rate": 4.473390612977687e-05, + "loss": 0.6414, + "step": 8263 + }, + { + "epoch": 0.32, + "learning_rate": 4.4733264939728136e-05, + "loss": 0.5316, + "step": 8264 + }, + { + "epoch": 0.32, + "learning_rate": 4.473262374967941e-05, + "loss": 0.4871, + "step": 8265 + }, + { + "epoch": 0.32, + "learning_rate": 4.473198255963067e-05, + "loss": 0.5807, + "step": 8266 + }, + { + "epoch": 0.32, + "learning_rate": 4.4731341369581944e-05, + "loss": 0.5608, + "step": 8267 + }, + { + "epoch": 0.32, + "learning_rate": 4.4730700179533216e-05, + "loss": 0.6149, + "step": 8268 + }, + { + "epoch": 0.32, + "learning_rate": 4.473005898948449e-05, + "loss": 0.5393, + "step": 8269 + }, + { + "epoch": 0.32, + "learning_rate": 4.472941779943575e-05, + "loss": 0.5513, + "step": 8270 + }, + { + "epoch": 0.32, + "learning_rate": 4.4728776609387024e-05, + "loss": 0.5518, + "step": 8271 + }, + { + "epoch": 0.32, + "learning_rate": 4.4728135419338296e-05, + "loss": 0.587, + "step": 8272 + }, + { + "epoch": 0.32, + "learning_rate": 4.472749422928956e-05, + "loss": 0.4899, + "step": 8273 + }, + { + "epoch": 0.32, + "learning_rate": 4.472685303924083e-05, + "loss": 0.5483, + "step": 8274 + }, + { + "epoch": 0.32, + "learning_rate": 4.4726211849192104e-05, + "loss": 0.544, + "step": 8275 + }, + { + "epoch": 0.32, + "learning_rate": 4.4725570659143376e-05, + "loss": 0.5785, + "step": 8276 + }, + { + "epoch": 0.32, + "learning_rate": 4.472492946909464e-05, + "loss": 0.6495, + "step": 8277 + }, + { + "epoch": 0.32, + "learning_rate": 4.472428827904591e-05, + "loss": 0.649, + "step": 8278 + }, + { + "epoch": 0.32, + "learning_rate": 4.472364708899718e-05, + "loss": 0.4872, + "step": 8279 + }, + { + "epoch": 0.32, + "learning_rate": 4.472300589894845e-05, + "loss": 0.6099, + "step": 8280 + }, + { + "epoch": 0.32, + "learning_rate": 4.472236470889972e-05, + "loss": 0.6097, + "step": 8281 + }, + { + "epoch": 0.32, + "learning_rate": 4.4721723518850986e-05, + "loss": 0.6401, + "step": 8282 + }, + { + "epoch": 0.32, + "learning_rate": 4.472108232880226e-05, + "loss": 0.5427, + "step": 8283 + }, + { + "epoch": 0.32, + "learning_rate": 4.472044113875353e-05, + "loss": 0.5718, + "step": 8284 + }, + { + "epoch": 0.32, + "learning_rate": 4.47197999487048e-05, + "loss": 0.5692, + "step": 8285 + }, + { + "epoch": 0.32, + "learning_rate": 4.4719158758656066e-05, + "loss": 0.5302, + "step": 8286 + }, + { + "epoch": 0.32, + "learning_rate": 4.471851756860734e-05, + "loss": 0.5331, + "step": 8287 + }, + { + "epoch": 0.32, + "learning_rate": 4.471787637855861e-05, + "loss": 0.5269, + "step": 8288 + }, + { + "epoch": 0.32, + "learning_rate": 4.4717235188509874e-05, + "loss": 0.5554, + "step": 8289 + }, + { + "epoch": 0.32, + "learning_rate": 4.4716593998461146e-05, + "loss": 0.5875, + "step": 8290 + }, + { + "epoch": 0.32, + "learning_rate": 4.471595280841242e-05, + "loss": 0.6319, + "step": 8291 + }, + { + "epoch": 0.32, + "learning_rate": 4.471531161836369e-05, + "loss": 0.5254, + "step": 8292 + }, + { + "epoch": 0.32, + "learning_rate": 4.4714670428314955e-05, + "loss": 0.61, + "step": 8293 + }, + { + "epoch": 0.32, + "learning_rate": 4.4714029238266226e-05, + "loss": 0.5337, + "step": 8294 + }, + { + "epoch": 0.32, + "learning_rate": 4.471338804821749e-05, + "loss": 0.5591, + "step": 8295 + }, + { + "epoch": 0.32, + "learning_rate": 4.471274685816876e-05, + "loss": 0.4948, + "step": 8296 + }, + { + "epoch": 0.32, + "learning_rate": 4.4712105668120035e-05, + "loss": 0.592, + "step": 8297 + }, + { + "epoch": 0.32, + "learning_rate": 4.47114644780713e-05, + "loss": 0.5184, + "step": 8298 + }, + { + "epoch": 0.32, + "learning_rate": 4.471082328802257e-05, + "loss": 0.7385, + "step": 8299 + }, + { + "epoch": 0.32, + "learning_rate": 4.471018209797384e-05, + "loss": 0.4295, + "step": 8300 + }, + { + "epoch": 0.32, + "learning_rate": 4.4709540907925115e-05, + "loss": 0.5951, + "step": 8301 + }, + { + "epoch": 0.32, + "learning_rate": 4.470889971787638e-05, + "loss": 0.5515, + "step": 8302 + }, + { + "epoch": 0.32, + "learning_rate": 4.470825852782765e-05, + "loss": 0.5732, + "step": 8303 + }, + { + "epoch": 0.32, + "learning_rate": 4.4707617337778916e-05, + "loss": 0.5841, + "step": 8304 + }, + { + "epoch": 0.32, + "learning_rate": 4.470697614773019e-05, + "loss": 0.5996, + "step": 8305 + }, + { + "epoch": 0.32, + "learning_rate": 4.470633495768146e-05, + "loss": 0.483, + "step": 8306 + }, + { + "epoch": 0.32, + "learning_rate": 4.470569376763273e-05, + "loss": 0.5282, + "step": 8307 + }, + { + "epoch": 0.32, + "learning_rate": 4.4705052577584e-05, + "loss": 0.6048, + "step": 8308 + }, + { + "epoch": 0.32, + "learning_rate": 4.470441138753527e-05, + "loss": 0.5042, + "step": 8309 + }, + { + "epoch": 0.32, + "learning_rate": 4.470377019748654e-05, + "loss": 0.6443, + "step": 8310 + }, + { + "epoch": 0.32, + "learning_rate": 4.4703129007437805e-05, + "loss": 0.5357, + "step": 8311 + }, + { + "epoch": 0.32, + "learning_rate": 4.4702487817389076e-05, + "loss": 0.4597, + "step": 8312 + }, + { + "epoch": 0.32, + "learning_rate": 4.470184662734034e-05, + "loss": 0.5435, + "step": 8313 + }, + { + "epoch": 0.32, + "learning_rate": 4.470120543729161e-05, + "loss": 0.6034, + "step": 8314 + }, + { + "epoch": 0.32, + "learning_rate": 4.4700564247242885e-05, + "loss": 0.5322, + "step": 8315 + }, + { + "epoch": 0.32, + "learning_rate": 4.4699923057194157e-05, + "loss": 0.5879, + "step": 8316 + }, + { + "epoch": 0.32, + "learning_rate": 4.469928186714543e-05, + "loss": 0.7021, + "step": 8317 + }, + { + "epoch": 0.32, + "learning_rate": 4.469864067709669e-05, + "loss": 0.6025, + "step": 8318 + }, + { + "epoch": 0.32, + "learning_rate": 4.4697999487047965e-05, + "loss": 0.6142, + "step": 8319 + }, + { + "epoch": 0.32, + "learning_rate": 4.469735829699923e-05, + "loss": 0.5601, + "step": 8320 + }, + { + "epoch": 0.32, + "learning_rate": 4.46967171069505e-05, + "loss": 0.4916, + "step": 8321 + }, + { + "epoch": 0.32, + "learning_rate": 4.4696075916901767e-05, + "loss": 0.5905, + "step": 8322 + }, + { + "epoch": 0.32, + "learning_rate": 4.469543472685304e-05, + "loss": 0.5166, + "step": 8323 + }, + { + "epoch": 0.32, + "learning_rate": 4.469479353680432e-05, + "loss": 0.5596, + "step": 8324 + }, + { + "epoch": 0.32, + "learning_rate": 4.469415234675558e-05, + "loss": 0.6768, + "step": 8325 + }, + { + "epoch": 0.32, + "learning_rate": 4.469351115670685e-05, + "loss": 0.4988, + "step": 8326 + }, + { + "epoch": 0.32, + "learning_rate": 4.469286996665812e-05, + "loss": 0.5136, + "step": 8327 + }, + { + "epoch": 0.32, + "learning_rate": 4.469222877660939e-05, + "loss": 0.506, + "step": 8328 + }, + { + "epoch": 0.32, + "learning_rate": 4.4691587586560655e-05, + "loss": 0.5979, + "step": 8329 + }, + { + "epoch": 0.32, + "learning_rate": 4.469094639651193e-05, + "loss": 0.5761, + "step": 8330 + }, + { + "epoch": 0.32, + "learning_rate": 4.46903052064632e-05, + "loss": 0.5813, + "step": 8331 + }, + { + "epoch": 0.32, + "learning_rate": 4.468966401641447e-05, + "loss": 0.5455, + "step": 8332 + }, + { + "epoch": 0.32, + "learning_rate": 4.468902282636574e-05, + "loss": 0.4698, + "step": 8333 + }, + { + "epoch": 0.32, + "learning_rate": 4.468838163631701e-05, + "loss": 0.6155, + "step": 8334 + }, + { + "epoch": 0.32, + "learning_rate": 4.468774044626828e-05, + "loss": 0.5747, + "step": 8335 + }, + { + "epoch": 0.32, + "learning_rate": 4.4687099256219543e-05, + "loss": 0.5128, + "step": 8336 + }, + { + "epoch": 0.32, + "learning_rate": 4.4686458066170815e-05, + "loss": 0.6983, + "step": 8337 + }, + { + "epoch": 0.32, + "learning_rate": 4.468581687612208e-05, + "loss": 0.7667, + "step": 8338 + }, + { + "epoch": 0.32, + "learning_rate": 4.468517568607335e-05, + "loss": 0.5372, + "step": 8339 + }, + { + "epoch": 0.32, + "learning_rate": 4.4684534496024624e-05, + "loss": 0.5216, + "step": 8340 + }, + { + "epoch": 0.32, + "learning_rate": 4.4683893305975895e-05, + "loss": 0.5262, + "step": 8341 + }, + { + "epoch": 0.32, + "learning_rate": 4.468325211592717e-05, + "loss": 0.5357, + "step": 8342 + }, + { + "epoch": 0.32, + "learning_rate": 4.468261092587843e-05, + "loss": 0.5657, + "step": 8343 + }, + { + "epoch": 0.32, + "learning_rate": 4.4681969735829704e-05, + "loss": 0.5757, + "step": 8344 + }, + { + "epoch": 0.32, + "learning_rate": 4.468132854578097e-05, + "loss": 0.5648, + "step": 8345 + }, + { + "epoch": 0.32, + "learning_rate": 4.468068735573224e-05, + "loss": 0.6005, + "step": 8346 + }, + { + "epoch": 0.32, + "learning_rate": 4.468004616568351e-05, + "loss": 0.5642, + "step": 8347 + }, + { + "epoch": 0.32, + "learning_rate": 4.4679404975634784e-05, + "loss": 0.4727, + "step": 8348 + }, + { + "epoch": 0.32, + "learning_rate": 4.467876378558605e-05, + "loss": 0.4809, + "step": 8349 + }, + { + "epoch": 0.32, + "learning_rate": 4.467812259553732e-05, + "loss": 0.4907, + "step": 8350 + }, + { + "epoch": 0.32, + "learning_rate": 4.467748140548859e-05, + "loss": 0.4778, + "step": 8351 + }, + { + "epoch": 0.32, + "learning_rate": 4.467684021543986e-05, + "loss": 0.4571, + "step": 8352 + }, + { + "epoch": 0.32, + "learning_rate": 4.467619902539113e-05, + "loss": 0.5397, + "step": 8353 + }, + { + "epoch": 0.32, + "learning_rate": 4.4675557835342394e-05, + "loss": 0.6227, + "step": 8354 + }, + { + "epoch": 0.32, + "learning_rate": 4.4674916645293665e-05, + "loss": 0.471, + "step": 8355 + }, + { + "epoch": 0.32, + "learning_rate": 4.467427545524494e-05, + "loss": 0.6201, + "step": 8356 + }, + { + "epoch": 0.32, + "learning_rate": 4.467363426519621e-05, + "loss": 0.6289, + "step": 8357 + }, + { + "epoch": 0.32, + "learning_rate": 4.4672993075147474e-05, + "loss": 0.539, + "step": 8358 + }, + { + "epoch": 0.32, + "learning_rate": 4.4672351885098745e-05, + "loss": 0.5883, + "step": 8359 + }, + { + "epoch": 0.32, + "learning_rate": 4.467171069505002e-05, + "loss": 0.676, + "step": 8360 + }, + { + "epoch": 0.32, + "learning_rate": 4.467106950500128e-05, + "loss": 0.5489, + "step": 8361 + }, + { + "epoch": 0.32, + "learning_rate": 4.4670428314952554e-05, + "loss": 0.5722, + "step": 8362 + }, + { + "epoch": 0.32, + "learning_rate": 4.4669787124903826e-05, + "loss": 0.6174, + "step": 8363 + }, + { + "epoch": 0.32, + "learning_rate": 4.46691459348551e-05, + "loss": 0.5459, + "step": 8364 + }, + { + "epoch": 0.32, + "learning_rate": 4.466850474480636e-05, + "loss": 0.521, + "step": 8365 + }, + { + "epoch": 0.32, + "learning_rate": 4.4667863554757634e-05, + "loss": 0.5368, + "step": 8366 + }, + { + "epoch": 0.32, + "learning_rate": 4.46672223647089e-05, + "loss": 0.6514, + "step": 8367 + }, + { + "epoch": 0.32, + "learning_rate": 4.466658117466017e-05, + "loss": 0.5218, + "step": 8368 + }, + { + "epoch": 0.32, + "learning_rate": 4.466593998461144e-05, + "loss": 0.6029, + "step": 8369 + }, + { + "epoch": 0.32, + "learning_rate": 4.466529879456271e-05, + "loss": 0.5652, + "step": 8370 + }, + { + "epoch": 0.32, + "learning_rate": 4.466465760451398e-05, + "loss": 0.707, + "step": 8371 + }, + { + "epoch": 0.32, + "learning_rate": 4.466401641446525e-05, + "loss": 0.5996, + "step": 8372 + }, + { + "epoch": 0.32, + "learning_rate": 4.466337522441652e-05, + "loss": 0.5575, + "step": 8373 + }, + { + "epoch": 0.32, + "learning_rate": 4.466273403436779e-05, + "loss": 0.4527, + "step": 8374 + }, + { + "epoch": 0.32, + "learning_rate": 4.466209284431906e-05, + "loss": 0.6506, + "step": 8375 + }, + { + "epoch": 0.32, + "learning_rate": 4.4661451654270324e-05, + "loss": 0.6583, + "step": 8376 + }, + { + "epoch": 0.32, + "learning_rate": 4.4660810464221596e-05, + "loss": 0.5663, + "step": 8377 + }, + { + "epoch": 0.32, + "learning_rate": 4.466016927417287e-05, + "loss": 0.6902, + "step": 8378 + }, + { + "epoch": 0.32, + "learning_rate": 4.465952808412414e-05, + "loss": 0.6875, + "step": 8379 + }, + { + "epoch": 0.32, + "learning_rate": 4.465888689407541e-05, + "loss": 0.4993, + "step": 8380 + }, + { + "epoch": 0.32, + "learning_rate": 4.4658245704026676e-05, + "loss": 0.5254, + "step": 8381 + }, + { + "epoch": 0.32, + "learning_rate": 4.465760451397795e-05, + "loss": 0.6597, + "step": 8382 + }, + { + "epoch": 0.32, + "learning_rate": 4.465696332392921e-05, + "loss": 0.6047, + "step": 8383 + }, + { + "epoch": 0.32, + "learning_rate": 4.4656322133880484e-05, + "loss": 0.6689, + "step": 8384 + }, + { + "epoch": 0.32, + "learning_rate": 4.465568094383175e-05, + "loss": 0.5464, + "step": 8385 + }, + { + "epoch": 0.32, + "learning_rate": 4.465503975378302e-05, + "loss": 0.5406, + "step": 8386 + }, + { + "epoch": 0.32, + "learning_rate": 4.465439856373429e-05, + "loss": 0.6198, + "step": 8387 + }, + { + "epoch": 0.32, + "learning_rate": 4.4653757373685564e-05, + "loss": 0.5518, + "step": 8388 + }, + { + "epoch": 0.32, + "learning_rate": 4.4653116183636836e-05, + "loss": 0.5864, + "step": 8389 + }, + { + "epoch": 0.32, + "learning_rate": 4.46524749935881e-05, + "loss": 0.5955, + "step": 8390 + }, + { + "epoch": 0.32, + "learning_rate": 4.465183380353937e-05, + "loss": 0.5913, + "step": 8391 + }, + { + "epoch": 0.32, + "learning_rate": 4.465119261349064e-05, + "loss": 0.575, + "step": 8392 + }, + { + "epoch": 0.32, + "learning_rate": 4.465055142344191e-05, + "loss": 0.6273, + "step": 8393 + }, + { + "epoch": 0.32, + "learning_rate": 4.4649910233393174e-05, + "loss": 0.4683, + "step": 8394 + }, + { + "epoch": 0.32, + "learning_rate": 4.464926904334445e-05, + "loss": 0.614, + "step": 8395 + }, + { + "epoch": 0.32, + "learning_rate": 4.4648627853295724e-05, + "loss": 0.5684, + "step": 8396 + }, + { + "epoch": 0.32, + "learning_rate": 4.464798666324699e-05, + "loss": 0.6482, + "step": 8397 + }, + { + "epoch": 0.32, + "learning_rate": 4.464734547319826e-05, + "loss": 0.6162, + "step": 8398 + }, + { + "epoch": 0.32, + "learning_rate": 4.4646704283149526e-05, + "loss": 0.646, + "step": 8399 + }, + { + "epoch": 0.32, + "learning_rate": 4.46460630931008e-05, + "loss": 0.5829, + "step": 8400 + }, + { + "epoch": 0.32, + "learning_rate": 4.464542190305206e-05, + "loss": 0.5941, + "step": 8401 + }, + { + "epoch": 0.32, + "learning_rate": 4.4644780713003334e-05, + "loss": 0.5996, + "step": 8402 + }, + { + "epoch": 0.32, + "learning_rate": 4.4644139522954606e-05, + "loss": 0.6146, + "step": 8403 + }, + { + "epoch": 0.32, + "learning_rate": 4.464349833290588e-05, + "loss": 0.5628, + "step": 8404 + }, + { + "epoch": 0.32, + "learning_rate": 4.464285714285715e-05, + "loss": 0.5065, + "step": 8405 + }, + { + "epoch": 0.32, + "learning_rate": 4.4642215952808415e-05, + "loss": 0.6123, + "step": 8406 + }, + { + "epoch": 0.32, + "learning_rate": 4.4641574762759686e-05, + "loss": 0.5467, + "step": 8407 + }, + { + "epoch": 0.32, + "learning_rate": 4.464093357271095e-05, + "loss": 0.554, + "step": 8408 + }, + { + "epoch": 0.32, + "learning_rate": 4.464029238266222e-05, + "loss": 0.4757, + "step": 8409 + }, + { + "epoch": 0.32, + "learning_rate": 4.463965119261349e-05, + "loss": 0.5519, + "step": 8410 + }, + { + "epoch": 0.32, + "learning_rate": 4.463901000256476e-05, + "loss": 0.5572, + "step": 8411 + }, + { + "epoch": 0.32, + "learning_rate": 4.463836881251603e-05, + "loss": 0.6472, + "step": 8412 + }, + { + "epoch": 0.32, + "learning_rate": 4.46377276224673e-05, + "loss": 0.5497, + "step": 8413 + }, + { + "epoch": 0.32, + "learning_rate": 4.4637086432418575e-05, + "loss": 0.5282, + "step": 8414 + }, + { + "epoch": 0.32, + "learning_rate": 4.463644524236984e-05, + "loss": 0.5764, + "step": 8415 + }, + { + "epoch": 0.32, + "learning_rate": 4.463580405232111e-05, + "loss": 0.67, + "step": 8416 + }, + { + "epoch": 0.32, + "learning_rate": 4.4635162862272376e-05, + "loss": 0.5955, + "step": 8417 + }, + { + "epoch": 0.32, + "learning_rate": 4.463452167222365e-05, + "loss": 0.5117, + "step": 8418 + }, + { + "epoch": 0.32, + "learning_rate": 4.463388048217492e-05, + "loss": 0.5054, + "step": 8419 + }, + { + "epoch": 0.32, + "learning_rate": 4.463323929212619e-05, + "loss": 0.589, + "step": 8420 + }, + { + "epoch": 0.32, + "learning_rate": 4.4632598102077456e-05, + "loss": 0.5851, + "step": 8421 + }, + { + "epoch": 0.32, + "learning_rate": 4.463195691202873e-05, + "loss": 0.5579, + "step": 8422 + }, + { + "epoch": 0.32, + "learning_rate": 4.463131572198e-05, + "loss": 0.5568, + "step": 8423 + }, + { + "epoch": 0.32, + "learning_rate": 4.4630674531931265e-05, + "loss": 0.5432, + "step": 8424 + }, + { + "epoch": 0.32, + "learning_rate": 4.4630033341882536e-05, + "loss": 0.5968, + "step": 8425 + }, + { + "epoch": 0.32, + "learning_rate": 4.46293921518338e-05, + "loss": 0.4908, + "step": 8426 + }, + { + "epoch": 0.32, + "learning_rate": 4.462875096178507e-05, + "loss": 0.6334, + "step": 8427 + }, + { + "epoch": 0.32, + "learning_rate": 4.4628109771736345e-05, + "loss": 0.6447, + "step": 8428 + }, + { + "epoch": 0.32, + "learning_rate": 4.4627468581687617e-05, + "loss": 0.5777, + "step": 8429 + }, + { + "epoch": 0.32, + "learning_rate": 4.462682739163888e-05, + "loss": 0.5463, + "step": 8430 + }, + { + "epoch": 0.32, + "learning_rate": 4.462618620159015e-05, + "loss": 0.5494, + "step": 8431 + }, + { + "epoch": 0.32, + "learning_rate": 4.4625545011541425e-05, + "loss": 0.5007, + "step": 8432 + }, + { + "epoch": 0.32, + "learning_rate": 4.462490382149269e-05, + "loss": 0.697, + "step": 8433 + }, + { + "epoch": 0.32, + "learning_rate": 4.462426263144396e-05, + "loss": 0.8105, + "step": 8434 + }, + { + "epoch": 0.32, + "learning_rate": 4.462362144139523e-05, + "loss": 0.6072, + "step": 8435 + }, + { + "epoch": 0.32, + "learning_rate": 4.4622980251346505e-05, + "loss": 0.5348, + "step": 8436 + }, + { + "epoch": 0.32, + "learning_rate": 4.462233906129777e-05, + "loss": 0.5558, + "step": 8437 + }, + { + "epoch": 0.32, + "learning_rate": 4.462169787124904e-05, + "loss": 0.5994, + "step": 8438 + }, + { + "epoch": 0.32, + "learning_rate": 4.462105668120031e-05, + "loss": 0.5724, + "step": 8439 + }, + { + "epoch": 0.32, + "learning_rate": 4.462041549115158e-05, + "loss": 0.5034, + "step": 8440 + }, + { + "epoch": 0.32, + "learning_rate": 4.461977430110285e-05, + "loss": 0.5386, + "step": 8441 + }, + { + "epoch": 0.32, + "learning_rate": 4.4619133111054115e-05, + "loss": 0.6507, + "step": 8442 + }, + { + "epoch": 0.32, + "learning_rate": 4.461849192100539e-05, + "loss": 0.5968, + "step": 8443 + }, + { + "epoch": 0.32, + "learning_rate": 4.461785073095666e-05, + "loss": 0.5734, + "step": 8444 + }, + { + "epoch": 0.32, + "learning_rate": 4.461720954090793e-05, + "loss": 0.6165, + "step": 8445 + }, + { + "epoch": 0.32, + "learning_rate": 4.4616568350859195e-05, + "loss": 0.5785, + "step": 8446 + }, + { + "epoch": 0.32, + "learning_rate": 4.461592716081047e-05, + "loss": 0.6404, + "step": 8447 + }, + { + "epoch": 0.32, + "learning_rate": 4.461528597076173e-05, + "loss": 0.5599, + "step": 8448 + }, + { + "epoch": 0.32, + "learning_rate": 4.4614644780713003e-05, + "loss": 0.5289, + "step": 8449 + }, + { + "epoch": 0.32, + "learning_rate": 4.4614003590664275e-05, + "loss": 0.6064, + "step": 8450 + }, + { + "epoch": 0.32, + "learning_rate": 4.461336240061555e-05, + "loss": 0.5112, + "step": 8451 + }, + { + "epoch": 0.32, + "learning_rate": 4.461272121056682e-05, + "loss": 0.5393, + "step": 8452 + }, + { + "epoch": 0.32, + "learning_rate": 4.4612080020518084e-05, + "loss": 0.4976, + "step": 8453 + }, + { + "epoch": 0.33, + "learning_rate": 4.4611438830469355e-05, + "loss": 0.5244, + "step": 8454 + }, + { + "epoch": 0.33, + "learning_rate": 4.461079764042062e-05, + "loss": 0.6119, + "step": 8455 + }, + { + "epoch": 0.33, + "learning_rate": 4.461015645037189e-05, + "loss": 0.586, + "step": 8456 + }, + { + "epoch": 0.33, + "learning_rate": 4.460951526032316e-05, + "loss": 0.5326, + "step": 8457 + }, + { + "epoch": 0.33, + "learning_rate": 4.460887407027443e-05, + "loss": 0.6281, + "step": 8458 + }, + { + "epoch": 0.33, + "learning_rate": 4.46082328802257e-05, + "loss": 0.4957, + "step": 8459 + }, + { + "epoch": 0.33, + "learning_rate": 4.460759169017697e-05, + "loss": 0.5856, + "step": 8460 + }, + { + "epoch": 0.33, + "learning_rate": 4.4606950500128244e-05, + "loss": 0.5929, + "step": 8461 + }, + { + "epoch": 0.33, + "learning_rate": 4.460630931007951e-05, + "loss": 0.6137, + "step": 8462 + }, + { + "epoch": 0.33, + "learning_rate": 4.460566812003078e-05, + "loss": 0.5301, + "step": 8463 + }, + { + "epoch": 0.33, + "learning_rate": 4.4605026929982045e-05, + "loss": 0.5196, + "step": 8464 + }, + { + "epoch": 0.33, + "learning_rate": 4.460438573993332e-05, + "loss": 0.5996, + "step": 8465 + }, + { + "epoch": 0.33, + "learning_rate": 4.460374454988458e-05, + "loss": 0.5088, + "step": 8466 + }, + { + "epoch": 0.33, + "learning_rate": 4.460310335983586e-05, + "loss": 0.542, + "step": 8467 + }, + { + "epoch": 0.33, + "learning_rate": 4.460246216978713e-05, + "loss": 0.7661, + "step": 8468 + }, + { + "epoch": 0.33, + "learning_rate": 4.46018209797384e-05, + "loss": 0.5526, + "step": 8469 + }, + { + "epoch": 0.33, + "learning_rate": 4.460117978968967e-05, + "loss": 0.7504, + "step": 8470 + }, + { + "epoch": 0.33, + "learning_rate": 4.4600538599640934e-05, + "loss": 0.8382, + "step": 8471 + }, + { + "epoch": 0.33, + "learning_rate": 4.4599897409592206e-05, + "loss": 0.6157, + "step": 8472 + }, + { + "epoch": 0.33, + "learning_rate": 4.459925621954347e-05, + "loss": 0.5429, + "step": 8473 + }, + { + "epoch": 0.33, + "learning_rate": 4.459861502949474e-05, + "loss": 0.5574, + "step": 8474 + }, + { + "epoch": 0.33, + "learning_rate": 4.4597973839446014e-05, + "loss": 0.547, + "step": 8475 + }, + { + "epoch": 0.33, + "learning_rate": 4.4597332649397286e-05, + "loss": 0.5222, + "step": 8476 + }, + { + "epoch": 0.33, + "learning_rate": 4.459669145934856e-05, + "loss": 0.6044, + "step": 8477 + }, + { + "epoch": 0.33, + "learning_rate": 4.459605026929982e-05, + "loss": 0.567, + "step": 8478 + }, + { + "epoch": 0.33, + "learning_rate": 4.4595409079251094e-05, + "loss": 0.6187, + "step": 8479 + }, + { + "epoch": 0.33, + "learning_rate": 4.459476788920236e-05, + "loss": 0.6082, + "step": 8480 + }, + { + "epoch": 0.33, + "learning_rate": 4.459412669915363e-05, + "loss": 0.6045, + "step": 8481 + }, + { + "epoch": 0.33, + "learning_rate": 4.4593485509104896e-05, + "loss": 0.5425, + "step": 8482 + }, + { + "epoch": 0.33, + "learning_rate": 4.459284431905617e-05, + "loss": 0.5788, + "step": 8483 + }, + { + "epoch": 0.33, + "learning_rate": 4.4592203129007446e-05, + "loss": 0.6039, + "step": 8484 + }, + { + "epoch": 0.33, + "learning_rate": 4.459156193895871e-05, + "loss": 0.5453, + "step": 8485 + }, + { + "epoch": 0.33, + "learning_rate": 4.459092074890998e-05, + "loss": 0.6296, + "step": 8486 + }, + { + "epoch": 0.33, + "learning_rate": 4.459027955886125e-05, + "loss": 0.5769, + "step": 8487 + }, + { + "epoch": 0.33, + "learning_rate": 4.458963836881252e-05, + "loss": 0.4666, + "step": 8488 + }, + { + "epoch": 0.33, + "learning_rate": 4.4588997178763784e-05, + "loss": 0.5986, + "step": 8489 + }, + { + "epoch": 0.33, + "learning_rate": 4.4588355988715056e-05, + "loss": 0.6026, + "step": 8490 + }, + { + "epoch": 0.33, + "learning_rate": 4.458771479866633e-05, + "loss": 0.569, + "step": 8491 + }, + { + "epoch": 0.33, + "learning_rate": 4.45870736086176e-05, + "loss": 0.6584, + "step": 8492 + }, + { + "epoch": 0.33, + "learning_rate": 4.458643241856887e-05, + "loss": 0.5207, + "step": 8493 + }, + { + "epoch": 0.33, + "learning_rate": 4.4585791228520136e-05, + "loss": 0.5926, + "step": 8494 + }, + { + "epoch": 0.33, + "learning_rate": 4.458515003847141e-05, + "loss": 0.5518, + "step": 8495 + }, + { + "epoch": 0.33, + "learning_rate": 4.458450884842267e-05, + "loss": 0.5685, + "step": 8496 + }, + { + "epoch": 0.33, + "learning_rate": 4.4583867658373944e-05, + "loss": 0.6054, + "step": 8497 + }, + { + "epoch": 0.33, + "learning_rate": 4.458322646832521e-05, + "loss": 0.592, + "step": 8498 + }, + { + "epoch": 0.33, + "learning_rate": 4.458258527827648e-05, + "loss": 0.5857, + "step": 8499 + }, + { + "epoch": 0.33, + "learning_rate": 4.458194408822775e-05, + "loss": 0.5644, + "step": 8500 + }, + { + "epoch": 0.33, + "learning_rate": 4.4581302898179024e-05, + "loss": 0.609, + "step": 8501 + }, + { + "epoch": 0.33, + "learning_rate": 4.4580661708130296e-05, + "loss": 0.526, + "step": 8502 + }, + { + "epoch": 0.33, + "learning_rate": 4.458002051808156e-05, + "loss": 0.5252, + "step": 8503 + }, + { + "epoch": 0.33, + "learning_rate": 4.457937932803283e-05, + "loss": 0.546, + "step": 8504 + }, + { + "epoch": 0.33, + "learning_rate": 4.45787381379841e-05, + "loss": 0.5879, + "step": 8505 + }, + { + "epoch": 0.33, + "learning_rate": 4.457809694793537e-05, + "loss": 0.5987, + "step": 8506 + }, + { + "epoch": 0.33, + "learning_rate": 4.457745575788664e-05, + "loss": 0.6078, + "step": 8507 + }, + { + "epoch": 0.33, + "learning_rate": 4.457681456783791e-05, + "loss": 0.5735, + "step": 8508 + }, + { + "epoch": 0.33, + "learning_rate": 4.457617337778918e-05, + "loss": 0.5509, + "step": 8509 + }, + { + "epoch": 0.33, + "learning_rate": 4.457553218774045e-05, + "loss": 0.4713, + "step": 8510 + }, + { + "epoch": 0.33, + "learning_rate": 4.457489099769172e-05, + "loss": 0.5488, + "step": 8511 + }, + { + "epoch": 0.33, + "learning_rate": 4.4574249807642986e-05, + "loss": 0.4776, + "step": 8512 + }, + { + "epoch": 0.33, + "learning_rate": 4.457360861759426e-05, + "loss": 0.6395, + "step": 8513 + }, + { + "epoch": 0.33, + "learning_rate": 4.457296742754552e-05, + "loss": 0.5624, + "step": 8514 + }, + { + "epoch": 0.33, + "learning_rate": 4.4572326237496794e-05, + "loss": 0.5347, + "step": 8515 + }, + { + "epoch": 0.33, + "learning_rate": 4.4571685047448066e-05, + "loss": 0.5646, + "step": 8516 + }, + { + "epoch": 0.33, + "learning_rate": 4.457104385739934e-05, + "loss": 0.5876, + "step": 8517 + }, + { + "epoch": 0.33, + "learning_rate": 4.45704026673506e-05, + "loss": 0.5156, + "step": 8518 + }, + { + "epoch": 0.33, + "learning_rate": 4.4569761477301875e-05, + "loss": 0.6344, + "step": 8519 + }, + { + "epoch": 0.33, + "learning_rate": 4.4569120287253146e-05, + "loss": 0.5436, + "step": 8520 + }, + { + "epoch": 0.33, + "learning_rate": 4.456847909720441e-05, + "loss": 0.513, + "step": 8521 + }, + { + "epoch": 0.33, + "learning_rate": 4.456783790715568e-05, + "loss": 0.6599, + "step": 8522 + }, + { + "epoch": 0.33, + "learning_rate": 4.4567196717106955e-05, + "loss": 0.6396, + "step": 8523 + }, + { + "epoch": 0.33, + "learning_rate": 4.4566555527058226e-05, + "loss": 0.512, + "step": 8524 + }, + { + "epoch": 0.33, + "learning_rate": 4.456591433700949e-05, + "loss": 0.4499, + "step": 8525 + }, + { + "epoch": 0.33, + "learning_rate": 4.456527314696076e-05, + "loss": 0.5084, + "step": 8526 + }, + { + "epoch": 0.33, + "learning_rate": 4.456463195691203e-05, + "loss": 0.5946, + "step": 8527 + }, + { + "epoch": 0.33, + "learning_rate": 4.45639907668633e-05, + "loss": 0.564, + "step": 8528 + }, + { + "epoch": 0.33, + "learning_rate": 4.456334957681457e-05, + "loss": 0.6641, + "step": 8529 + }, + { + "epoch": 0.33, + "learning_rate": 4.4562708386765836e-05, + "loss": 0.5371, + "step": 8530 + }, + { + "epoch": 0.33, + "learning_rate": 4.456206719671711e-05, + "loss": 0.5412, + "step": 8531 + }, + { + "epoch": 0.33, + "learning_rate": 4.456142600666838e-05, + "loss": 0.4866, + "step": 8532 + }, + { + "epoch": 0.33, + "learning_rate": 4.456078481661965e-05, + "loss": 0.5538, + "step": 8533 + }, + { + "epoch": 0.33, + "learning_rate": 4.4560143626570916e-05, + "loss": 0.6568, + "step": 8534 + }, + { + "epoch": 0.33, + "learning_rate": 4.455950243652219e-05, + "loss": 0.5408, + "step": 8535 + }, + { + "epoch": 0.33, + "learning_rate": 4.455886124647345e-05, + "loss": 0.5664, + "step": 8536 + }, + { + "epoch": 0.33, + "learning_rate": 4.4558220056424725e-05, + "loss": 0.6182, + "step": 8537 + }, + { + "epoch": 0.33, + "learning_rate": 4.4557578866375997e-05, + "loss": 0.543, + "step": 8538 + }, + { + "epoch": 0.33, + "learning_rate": 4.455693767632727e-05, + "loss": 0.6033, + "step": 8539 + }, + { + "epoch": 0.33, + "learning_rate": 4.455629648627854e-05, + "loss": 0.5936, + "step": 8540 + }, + { + "epoch": 0.33, + "learning_rate": 4.4555655296229805e-05, + "loss": 0.5191, + "step": 8541 + }, + { + "epoch": 0.33, + "learning_rate": 4.455501410618108e-05, + "loss": 0.6485, + "step": 8542 + }, + { + "epoch": 0.33, + "learning_rate": 4.455437291613234e-05, + "loss": 0.5846, + "step": 8543 + }, + { + "epoch": 0.33, + "learning_rate": 4.455373172608361e-05, + "loss": 0.56, + "step": 8544 + }, + { + "epoch": 0.33, + "learning_rate": 4.455309053603488e-05, + "loss": 0.6173, + "step": 8545 + }, + { + "epoch": 0.33, + "learning_rate": 4.455244934598615e-05, + "loss": 0.5275, + "step": 8546 + }, + { + "epoch": 0.33, + "learning_rate": 4.455180815593742e-05, + "loss": 0.568, + "step": 8547 + }, + { + "epoch": 0.33, + "learning_rate": 4.455116696588869e-05, + "loss": 0.6102, + "step": 8548 + }, + { + "epoch": 0.33, + "learning_rate": 4.4550525775839965e-05, + "loss": 0.4958, + "step": 8549 + }, + { + "epoch": 0.33, + "learning_rate": 4.454988458579123e-05, + "loss": 0.5811, + "step": 8550 + }, + { + "epoch": 0.33, + "learning_rate": 4.45492433957425e-05, + "loss": 0.5928, + "step": 8551 + }, + { + "epoch": 0.33, + "learning_rate": 4.454860220569377e-05, + "loss": 0.6326, + "step": 8552 + }, + { + "epoch": 0.33, + "learning_rate": 4.454796101564504e-05, + "loss": 0.5356, + "step": 8553 + }, + { + "epoch": 0.33, + "learning_rate": 4.45473198255963e-05, + "loss": 0.5512, + "step": 8554 + }, + { + "epoch": 0.33, + "learning_rate": 4.454667863554758e-05, + "loss": 0.5513, + "step": 8555 + }, + { + "epoch": 0.33, + "learning_rate": 4.4546037445498854e-05, + "loss": 0.5236, + "step": 8556 + }, + { + "epoch": 0.33, + "learning_rate": 4.454539625545012e-05, + "loss": 0.5549, + "step": 8557 + }, + { + "epoch": 0.33, + "learning_rate": 4.454475506540139e-05, + "loss": 0.601, + "step": 8558 + }, + { + "epoch": 0.33, + "learning_rate": 4.4544113875352655e-05, + "loss": 0.5881, + "step": 8559 + }, + { + "epoch": 0.33, + "learning_rate": 4.454347268530393e-05, + "loss": 0.5898, + "step": 8560 + }, + { + "epoch": 0.33, + "learning_rate": 4.454283149525519e-05, + "loss": 0.503, + "step": 8561 + }, + { + "epoch": 0.33, + "learning_rate": 4.4542190305206464e-05, + "loss": 0.4955, + "step": 8562 + }, + { + "epoch": 0.33, + "learning_rate": 4.4541549115157735e-05, + "loss": 0.6298, + "step": 8563 + }, + { + "epoch": 0.33, + "learning_rate": 4.454090792510901e-05, + "loss": 0.5059, + "step": 8564 + }, + { + "epoch": 0.33, + "learning_rate": 4.454026673506028e-05, + "loss": 0.5371, + "step": 8565 + }, + { + "epoch": 0.33, + "learning_rate": 4.4539625545011544e-05, + "loss": 0.6468, + "step": 8566 + }, + { + "epoch": 0.33, + "learning_rate": 4.4538984354962815e-05, + "loss": 0.6016, + "step": 8567 + }, + { + "epoch": 0.33, + "learning_rate": 4.453834316491408e-05, + "loss": 0.5248, + "step": 8568 + }, + { + "epoch": 0.33, + "learning_rate": 4.453770197486535e-05, + "loss": 0.5268, + "step": 8569 + }, + { + "epoch": 0.33, + "learning_rate": 4.453706078481662e-05, + "loss": 0.4643, + "step": 8570 + }, + { + "epoch": 0.33, + "learning_rate": 4.453641959476789e-05, + "loss": 0.4879, + "step": 8571 + }, + { + "epoch": 0.33, + "learning_rate": 4.453577840471916e-05, + "loss": 0.6953, + "step": 8572 + }, + { + "epoch": 0.33, + "learning_rate": 4.453513721467043e-05, + "loss": 0.7358, + "step": 8573 + }, + { + "epoch": 0.33, + "learning_rate": 4.4534496024621704e-05, + "loss": 0.56, + "step": 8574 + }, + { + "epoch": 0.33, + "learning_rate": 4.453385483457297e-05, + "loss": 0.5507, + "step": 8575 + }, + { + "epoch": 0.33, + "learning_rate": 4.453321364452424e-05, + "loss": 0.4916, + "step": 8576 + }, + { + "epoch": 0.33, + "learning_rate": 4.4532572454475505e-05, + "loss": 0.5772, + "step": 8577 + }, + { + "epoch": 0.33, + "learning_rate": 4.453193126442678e-05, + "loss": 0.5232, + "step": 8578 + }, + { + "epoch": 0.33, + "learning_rate": 4.453129007437805e-05, + "loss": 0.5868, + "step": 8579 + }, + { + "epoch": 0.33, + "learning_rate": 4.453064888432932e-05, + "loss": 0.499, + "step": 8580 + }, + { + "epoch": 0.33, + "learning_rate": 4.4530007694280585e-05, + "loss": 0.5896, + "step": 8581 + }, + { + "epoch": 0.33, + "learning_rate": 4.452936650423186e-05, + "loss": 0.5295, + "step": 8582 + }, + { + "epoch": 0.33, + "learning_rate": 4.452872531418313e-05, + "loss": 0.6621, + "step": 8583 + }, + { + "epoch": 0.33, + "learning_rate": 4.4528084124134394e-05, + "loss": 0.62, + "step": 8584 + }, + { + "epoch": 0.33, + "learning_rate": 4.4527442934085666e-05, + "loss": 0.5527, + "step": 8585 + }, + { + "epoch": 0.33, + "learning_rate": 4.452680174403693e-05, + "loss": 0.5928, + "step": 8586 + }, + { + "epoch": 0.33, + "learning_rate": 4.45261605539882e-05, + "loss": 0.5056, + "step": 8587 + }, + { + "epoch": 0.33, + "learning_rate": 4.4525519363939474e-05, + "loss": 0.6687, + "step": 8588 + }, + { + "epoch": 0.33, + "learning_rate": 4.4524878173890746e-05, + "loss": 0.6651, + "step": 8589 + }, + { + "epoch": 0.33, + "learning_rate": 4.452423698384201e-05, + "loss": 0.514, + "step": 8590 + }, + { + "epoch": 0.33, + "learning_rate": 4.452359579379328e-05, + "loss": 0.569, + "step": 8591 + }, + { + "epoch": 0.33, + "learning_rate": 4.4522954603744554e-05, + "loss": 0.5633, + "step": 8592 + }, + { + "epoch": 0.33, + "learning_rate": 4.452231341369582e-05, + "loss": 0.5758, + "step": 8593 + }, + { + "epoch": 0.33, + "learning_rate": 4.452167222364709e-05, + "loss": 0.481, + "step": 8594 + }, + { + "epoch": 0.33, + "learning_rate": 4.452103103359836e-05, + "loss": 0.6111, + "step": 8595 + }, + { + "epoch": 0.33, + "learning_rate": 4.4520389843549634e-05, + "loss": 0.52, + "step": 8596 + }, + { + "epoch": 0.33, + "learning_rate": 4.45197486535009e-05, + "loss": 0.5581, + "step": 8597 + }, + { + "epoch": 0.33, + "learning_rate": 4.451910746345217e-05, + "loss": 0.6316, + "step": 8598 + }, + { + "epoch": 0.33, + "learning_rate": 4.4518466273403436e-05, + "loss": 0.4735, + "step": 8599 + }, + { + "epoch": 0.33, + "learning_rate": 4.451782508335471e-05, + "loss": 0.5669, + "step": 8600 + }, + { + "epoch": 0.33, + "learning_rate": 4.451718389330598e-05, + "loss": 0.6258, + "step": 8601 + }, + { + "epoch": 0.33, + "learning_rate": 4.4516542703257244e-05, + "loss": 0.6813, + "step": 8602 + }, + { + "epoch": 0.33, + "learning_rate": 4.4515901513208516e-05, + "loss": 0.5675, + "step": 8603 + }, + { + "epoch": 0.33, + "learning_rate": 4.451526032315979e-05, + "loss": 0.6272, + "step": 8604 + }, + { + "epoch": 0.33, + "learning_rate": 4.451461913311106e-05, + "loss": 0.6091, + "step": 8605 + }, + { + "epoch": 0.33, + "learning_rate": 4.4513977943062324e-05, + "loss": 0.4786, + "step": 8606 + }, + { + "epoch": 0.33, + "learning_rate": 4.4513336753013596e-05, + "loss": 0.576, + "step": 8607 + }, + { + "epoch": 0.33, + "learning_rate": 4.451269556296486e-05, + "loss": 0.488, + "step": 8608 + }, + { + "epoch": 0.33, + "learning_rate": 4.451205437291613e-05, + "loss": 0.584, + "step": 8609 + }, + { + "epoch": 0.33, + "learning_rate": 4.4511413182867404e-05, + "loss": 0.4877, + "step": 8610 + }, + { + "epoch": 0.33, + "learning_rate": 4.4510771992818676e-05, + "loss": 0.5283, + "step": 8611 + }, + { + "epoch": 0.33, + "learning_rate": 4.451013080276995e-05, + "loss": 0.5545, + "step": 8612 + }, + { + "epoch": 0.33, + "learning_rate": 4.450948961272121e-05, + "loss": 0.5301, + "step": 8613 + }, + { + "epoch": 0.33, + "learning_rate": 4.4508848422672484e-05, + "loss": 0.5562, + "step": 8614 + }, + { + "epoch": 0.33, + "learning_rate": 4.450820723262375e-05, + "loss": 0.4665, + "step": 8615 + }, + { + "epoch": 0.33, + "learning_rate": 4.450756604257502e-05, + "loss": 0.5947, + "step": 8616 + }, + { + "epoch": 0.33, + "learning_rate": 4.4506924852526286e-05, + "loss": 0.5629, + "step": 8617 + }, + { + "epoch": 0.33, + "learning_rate": 4.450628366247756e-05, + "loss": 0.6121, + "step": 8618 + }, + { + "epoch": 0.33, + "learning_rate": 4.450564247242883e-05, + "loss": 0.505, + "step": 8619 + }, + { + "epoch": 0.33, + "learning_rate": 4.45050012823801e-05, + "loss": 0.6102, + "step": 8620 + }, + { + "epoch": 0.33, + "learning_rate": 4.450436009233137e-05, + "loss": 0.5233, + "step": 8621 + }, + { + "epoch": 0.33, + "learning_rate": 4.450371890228264e-05, + "loss": 0.6834, + "step": 8622 + }, + { + "epoch": 0.33, + "learning_rate": 4.450307771223391e-05, + "loss": 0.6189, + "step": 8623 + }, + { + "epoch": 0.33, + "learning_rate": 4.4502436522185174e-05, + "loss": 0.5953, + "step": 8624 + }, + { + "epoch": 0.33, + "learning_rate": 4.4501795332136446e-05, + "loss": 0.4646, + "step": 8625 + }, + { + "epoch": 0.33, + "learning_rate": 4.450115414208771e-05, + "loss": 0.4995, + "step": 8626 + }, + { + "epoch": 0.33, + "learning_rate": 4.450051295203899e-05, + "loss": 0.6056, + "step": 8627 + }, + { + "epoch": 0.33, + "learning_rate": 4.449987176199026e-05, + "loss": 0.5437, + "step": 8628 + }, + { + "epoch": 0.33, + "learning_rate": 4.4499230571941526e-05, + "loss": 0.5747, + "step": 8629 + }, + { + "epoch": 0.33, + "learning_rate": 4.44985893818928e-05, + "loss": 0.5623, + "step": 8630 + }, + { + "epoch": 0.33, + "learning_rate": 4.449794819184406e-05, + "loss": 0.629, + "step": 8631 + }, + { + "epoch": 0.33, + "learning_rate": 4.4497307001795335e-05, + "loss": 0.5062, + "step": 8632 + }, + { + "epoch": 0.33, + "learning_rate": 4.44966658117466e-05, + "loss": 0.527, + "step": 8633 + }, + { + "epoch": 0.33, + "learning_rate": 4.449602462169787e-05, + "loss": 0.5914, + "step": 8634 + }, + { + "epoch": 0.33, + "learning_rate": 4.449538343164914e-05, + "loss": 0.6227, + "step": 8635 + }, + { + "epoch": 0.33, + "learning_rate": 4.4494742241600415e-05, + "loss": 0.6255, + "step": 8636 + }, + { + "epoch": 0.33, + "learning_rate": 4.4494101051551686e-05, + "loss": 0.5885, + "step": 8637 + }, + { + "epoch": 0.33, + "learning_rate": 4.449345986150295e-05, + "loss": 0.4969, + "step": 8638 + }, + { + "epoch": 0.33, + "learning_rate": 4.449281867145422e-05, + "loss": 0.5665, + "step": 8639 + }, + { + "epoch": 0.33, + "learning_rate": 4.449217748140549e-05, + "loss": 0.5367, + "step": 8640 + }, + { + "epoch": 0.33, + "learning_rate": 4.449153629135676e-05, + "loss": 0.5566, + "step": 8641 + }, + { + "epoch": 0.33, + "learning_rate": 4.4490895101308025e-05, + "loss": 0.5312, + "step": 8642 + }, + { + "epoch": 0.33, + "learning_rate": 4.4490253911259296e-05, + "loss": 0.5489, + "step": 8643 + }, + { + "epoch": 0.33, + "learning_rate": 4.448961272121057e-05, + "loss": 0.5783, + "step": 8644 + }, + { + "epoch": 0.33, + "learning_rate": 4.448897153116184e-05, + "loss": 0.47, + "step": 8645 + }, + { + "epoch": 0.33, + "learning_rate": 4.448833034111311e-05, + "loss": 0.6031, + "step": 8646 + }, + { + "epoch": 0.33, + "learning_rate": 4.4487689151064376e-05, + "loss": 0.5451, + "step": 8647 + }, + { + "epoch": 0.33, + "learning_rate": 4.448704796101565e-05, + "loss": 0.6982, + "step": 8648 + }, + { + "epoch": 0.33, + "learning_rate": 4.448640677096691e-05, + "loss": 0.5791, + "step": 8649 + }, + { + "epoch": 0.33, + "learning_rate": 4.4485765580918185e-05, + "loss": 0.5708, + "step": 8650 + }, + { + "epoch": 0.33, + "learning_rate": 4.4485124390869457e-05, + "loss": 0.5494, + "step": 8651 + }, + { + "epoch": 0.33, + "learning_rate": 4.448448320082073e-05, + "loss": 0.5811, + "step": 8652 + }, + { + "epoch": 0.33, + "learning_rate": 4.448384201077199e-05, + "loss": 0.5815, + "step": 8653 + }, + { + "epoch": 0.33, + "learning_rate": 4.4483200820723265e-05, + "loss": 0.6, + "step": 8654 + }, + { + "epoch": 0.33, + "learning_rate": 4.448255963067454e-05, + "loss": 0.6354, + "step": 8655 + }, + { + "epoch": 0.33, + "learning_rate": 4.44819184406258e-05, + "loss": 0.4936, + "step": 8656 + }, + { + "epoch": 0.33, + "learning_rate": 4.448127725057707e-05, + "loss": 0.5552, + "step": 8657 + }, + { + "epoch": 0.33, + "learning_rate": 4.448063606052834e-05, + "loss": 0.6033, + "step": 8658 + }, + { + "epoch": 0.33, + "learning_rate": 4.447999487047961e-05, + "loss": 0.5372, + "step": 8659 + }, + { + "epoch": 0.33, + "learning_rate": 4.447935368043088e-05, + "loss": 0.5252, + "step": 8660 + }, + { + "epoch": 0.33, + "learning_rate": 4.4478712490382153e-05, + "loss": 0.5616, + "step": 8661 + }, + { + "epoch": 0.33, + "learning_rate": 4.4478071300333425e-05, + "loss": 0.6027, + "step": 8662 + }, + { + "epoch": 0.33, + "learning_rate": 4.447743011028469e-05, + "loss": 0.6886, + "step": 8663 + }, + { + "epoch": 0.33, + "learning_rate": 4.447678892023596e-05, + "loss": 0.5648, + "step": 8664 + }, + { + "epoch": 0.33, + "learning_rate": 4.447614773018723e-05, + "loss": 0.5501, + "step": 8665 + }, + { + "epoch": 0.33, + "learning_rate": 4.44755065401385e-05, + "loss": 0.5457, + "step": 8666 + }, + { + "epoch": 0.33, + "learning_rate": 4.447486535008977e-05, + "loss": 0.5983, + "step": 8667 + }, + { + "epoch": 0.33, + "learning_rate": 4.447422416004104e-05, + "loss": 0.6195, + "step": 8668 + }, + { + "epoch": 0.33, + "learning_rate": 4.447358296999231e-05, + "loss": 0.5424, + "step": 8669 + }, + { + "epoch": 0.33, + "learning_rate": 4.447294177994358e-05, + "loss": 0.6192, + "step": 8670 + }, + { + "epoch": 0.33, + "learning_rate": 4.447230058989485e-05, + "loss": 0.5799, + "step": 8671 + }, + { + "epoch": 0.33, + "learning_rate": 4.4471659399846115e-05, + "loss": 0.5444, + "step": 8672 + }, + { + "epoch": 0.33, + "learning_rate": 4.447101820979739e-05, + "loss": 0.6123, + "step": 8673 + }, + { + "epoch": 0.33, + "learning_rate": 4.447037701974865e-05, + "loss": 0.6593, + "step": 8674 + }, + { + "epoch": 0.33, + "learning_rate": 4.4469735829699924e-05, + "loss": 0.6279, + "step": 8675 + }, + { + "epoch": 0.33, + "learning_rate": 4.4469094639651195e-05, + "loss": 0.5951, + "step": 8676 + }, + { + "epoch": 0.33, + "learning_rate": 4.446845344960247e-05, + "loss": 0.5151, + "step": 8677 + }, + { + "epoch": 0.33, + "learning_rate": 4.446781225955373e-05, + "loss": 0.6075, + "step": 8678 + }, + { + "epoch": 0.33, + "learning_rate": 4.4467171069505004e-05, + "loss": 0.5753, + "step": 8679 + }, + { + "epoch": 0.33, + "learning_rate": 4.4466529879456275e-05, + "loss": 0.6379, + "step": 8680 + }, + { + "epoch": 0.33, + "learning_rate": 4.446588868940754e-05, + "loss": 0.6054, + "step": 8681 + }, + { + "epoch": 0.33, + "learning_rate": 4.446524749935881e-05, + "loss": 0.5593, + "step": 8682 + }, + { + "epoch": 0.33, + "learning_rate": 4.4464606309310084e-05, + "loss": 0.5612, + "step": 8683 + }, + { + "epoch": 0.33, + "learning_rate": 4.4463965119261355e-05, + "loss": 0.5745, + "step": 8684 + }, + { + "epoch": 0.33, + "learning_rate": 4.446332392921262e-05, + "loss": 0.5894, + "step": 8685 + }, + { + "epoch": 0.33, + "learning_rate": 4.446268273916389e-05, + "loss": 0.5891, + "step": 8686 + }, + { + "epoch": 0.33, + "learning_rate": 4.446204154911516e-05, + "loss": 0.6131, + "step": 8687 + }, + { + "epoch": 0.33, + "learning_rate": 4.446140035906643e-05, + "loss": 0.5895, + "step": 8688 + }, + { + "epoch": 0.33, + "learning_rate": 4.44607591690177e-05, + "loss": 0.5779, + "step": 8689 + }, + { + "epoch": 0.33, + "learning_rate": 4.4460117978968965e-05, + "loss": 0.5145, + "step": 8690 + }, + { + "epoch": 0.33, + "learning_rate": 4.445947678892024e-05, + "loss": 0.6314, + "step": 8691 + }, + { + "epoch": 0.33, + "learning_rate": 4.445883559887151e-05, + "loss": 0.5687, + "step": 8692 + }, + { + "epoch": 0.33, + "learning_rate": 4.445819440882278e-05, + "loss": 0.5881, + "step": 8693 + }, + { + "epoch": 0.33, + "learning_rate": 4.4457553218774046e-05, + "loss": 0.6712, + "step": 8694 + }, + { + "epoch": 0.33, + "learning_rate": 4.445691202872532e-05, + "loss": 0.5677, + "step": 8695 + }, + { + "epoch": 0.33, + "learning_rate": 4.445627083867658e-05, + "loss": 0.6094, + "step": 8696 + }, + { + "epoch": 0.33, + "learning_rate": 4.4455629648627854e-05, + "loss": 0.6274, + "step": 8697 + }, + { + "epoch": 0.33, + "learning_rate": 4.4454988458579126e-05, + "loss": 0.5708, + "step": 8698 + }, + { + "epoch": 0.33, + "learning_rate": 4.44543472685304e-05, + "loss": 0.571, + "step": 8699 + }, + { + "epoch": 0.33, + "learning_rate": 4.445370607848167e-05, + "loss": 0.565, + "step": 8700 + }, + { + "epoch": 0.33, + "learning_rate": 4.4453064888432934e-05, + "loss": 0.5172, + "step": 8701 + }, + { + "epoch": 0.33, + "learning_rate": 4.4452423698384206e-05, + "loss": 0.6369, + "step": 8702 + }, + { + "epoch": 0.33, + "learning_rate": 4.445178250833547e-05, + "loss": 0.6064, + "step": 8703 + }, + { + "epoch": 0.33, + "learning_rate": 4.445114131828674e-05, + "loss": 0.6832, + "step": 8704 + }, + { + "epoch": 0.33, + "learning_rate": 4.445050012823801e-05, + "loss": 0.545, + "step": 8705 + }, + { + "epoch": 0.33, + "learning_rate": 4.444985893818928e-05, + "loss": 0.5337, + "step": 8706 + }, + { + "epoch": 0.33, + "learning_rate": 4.444921774814055e-05, + "loss": 0.5004, + "step": 8707 + }, + { + "epoch": 0.33, + "learning_rate": 4.444857655809182e-05, + "loss": 0.5554, + "step": 8708 + }, + { + "epoch": 0.33, + "learning_rate": 4.4447935368043094e-05, + "loss": 0.5299, + "step": 8709 + }, + { + "epoch": 0.33, + "learning_rate": 4.444729417799436e-05, + "loss": 0.6594, + "step": 8710 + }, + { + "epoch": 0.33, + "learning_rate": 4.444665298794563e-05, + "loss": 0.5537, + "step": 8711 + }, + { + "epoch": 0.33, + "learning_rate": 4.4446011797896896e-05, + "loss": 0.5927, + "step": 8712 + }, + { + "epoch": 0.33, + "learning_rate": 4.444537060784817e-05, + "loss": 0.5401, + "step": 8713 + }, + { + "epoch": 0.34, + "learning_rate": 4.444472941779943e-05, + "loss": 0.6057, + "step": 8714 + }, + { + "epoch": 0.34, + "learning_rate": 4.444408822775071e-05, + "loss": 0.4679, + "step": 8715 + }, + { + "epoch": 0.34, + "learning_rate": 4.444344703770198e-05, + "loss": 0.6532, + "step": 8716 + }, + { + "epoch": 0.34, + "learning_rate": 4.444280584765325e-05, + "loss": 0.5308, + "step": 8717 + }, + { + "epoch": 0.34, + "learning_rate": 4.444216465760452e-05, + "loss": 0.5832, + "step": 8718 + }, + { + "epoch": 0.34, + "learning_rate": 4.4441523467555784e-05, + "loss": 0.4797, + "step": 8719 + }, + { + "epoch": 0.34, + "learning_rate": 4.4440882277507056e-05, + "loss": 0.5775, + "step": 8720 + }, + { + "epoch": 0.34, + "learning_rate": 4.444024108745832e-05, + "loss": 0.5155, + "step": 8721 + }, + { + "epoch": 0.34, + "learning_rate": 4.443959989740959e-05, + "loss": 0.5288, + "step": 8722 + }, + { + "epoch": 0.34, + "learning_rate": 4.4438958707360864e-05, + "loss": 0.5125, + "step": 8723 + }, + { + "epoch": 0.34, + "learning_rate": 4.4438317517312136e-05, + "loss": 0.5652, + "step": 8724 + }, + { + "epoch": 0.34, + "learning_rate": 4.443767632726341e-05, + "loss": 0.5388, + "step": 8725 + }, + { + "epoch": 0.34, + "learning_rate": 4.443703513721467e-05, + "loss": 0.5803, + "step": 8726 + }, + { + "epoch": 0.34, + "learning_rate": 4.4436393947165944e-05, + "loss": 0.4764, + "step": 8727 + }, + { + "epoch": 0.34, + "learning_rate": 4.443575275711721e-05, + "loss": 0.6986, + "step": 8728 + }, + { + "epoch": 0.34, + "learning_rate": 4.443511156706848e-05, + "loss": 0.5308, + "step": 8729 + }, + { + "epoch": 0.34, + "learning_rate": 4.4434470377019746e-05, + "loss": 0.4395, + "step": 8730 + }, + { + "epoch": 0.34, + "learning_rate": 4.443382918697102e-05, + "loss": 0.6091, + "step": 8731 + }, + { + "epoch": 0.34, + "learning_rate": 4.443318799692229e-05, + "loss": 0.5866, + "step": 8732 + }, + { + "epoch": 0.34, + "learning_rate": 4.443254680687356e-05, + "loss": 0.6178, + "step": 8733 + }, + { + "epoch": 0.34, + "learning_rate": 4.443190561682483e-05, + "loss": 0.6319, + "step": 8734 + }, + { + "epoch": 0.34, + "learning_rate": 4.44312644267761e-05, + "loss": 0.6602, + "step": 8735 + }, + { + "epoch": 0.34, + "learning_rate": 4.443062323672737e-05, + "loss": 0.6313, + "step": 8736 + }, + { + "epoch": 0.34, + "learning_rate": 4.4429982046678634e-05, + "loss": 0.6359, + "step": 8737 + }, + { + "epoch": 0.34, + "learning_rate": 4.4429340856629906e-05, + "loss": 0.5891, + "step": 8738 + }, + { + "epoch": 0.34, + "learning_rate": 4.442869966658118e-05, + "loss": 0.6262, + "step": 8739 + }, + { + "epoch": 0.34, + "learning_rate": 4.442805847653245e-05, + "loss": 0.6282, + "step": 8740 + }, + { + "epoch": 0.34, + "learning_rate": 4.4427417286483715e-05, + "loss": 0.4623, + "step": 8741 + }, + { + "epoch": 0.34, + "learning_rate": 4.4426776096434986e-05, + "loss": 0.602, + "step": 8742 + }, + { + "epoch": 0.34, + "learning_rate": 4.442613490638626e-05, + "loss": 0.6928, + "step": 8743 + }, + { + "epoch": 0.34, + "learning_rate": 4.442549371633752e-05, + "loss": 0.5637, + "step": 8744 + }, + { + "epoch": 0.34, + "learning_rate": 4.4424852526288795e-05, + "loss": 0.5023, + "step": 8745 + }, + { + "epoch": 0.34, + "learning_rate": 4.442421133624006e-05, + "loss": 0.6135, + "step": 8746 + }, + { + "epoch": 0.34, + "learning_rate": 4.442357014619133e-05, + "loss": 0.5071, + "step": 8747 + }, + { + "epoch": 0.34, + "learning_rate": 4.44229289561426e-05, + "loss": 0.5028, + "step": 8748 + }, + { + "epoch": 0.34, + "learning_rate": 4.4422287766093875e-05, + "loss": 0.6358, + "step": 8749 + }, + { + "epoch": 0.34, + "learning_rate": 4.442164657604514e-05, + "loss": 0.586, + "step": 8750 + }, + { + "epoch": 0.34, + "learning_rate": 4.442100538599641e-05, + "loss": 0.5267, + "step": 8751 + }, + { + "epoch": 0.34, + "learning_rate": 4.442036419594768e-05, + "loss": 0.5325, + "step": 8752 + }, + { + "epoch": 0.34, + "learning_rate": 4.441972300589895e-05, + "loss": 0.529, + "step": 8753 + }, + { + "epoch": 0.34, + "learning_rate": 4.441908181585022e-05, + "loss": 0.5653, + "step": 8754 + }, + { + "epoch": 0.34, + "learning_rate": 4.441844062580149e-05, + "loss": 0.6391, + "step": 8755 + }, + { + "epoch": 0.34, + "learning_rate": 4.441779943575276e-05, + "loss": 0.582, + "step": 8756 + }, + { + "epoch": 0.34, + "learning_rate": 4.441715824570403e-05, + "loss": 0.5306, + "step": 8757 + }, + { + "epoch": 0.34, + "learning_rate": 4.44165170556553e-05, + "loss": 0.6226, + "step": 8758 + }, + { + "epoch": 0.34, + "learning_rate": 4.4415875865606565e-05, + "loss": 0.6711, + "step": 8759 + }, + { + "epoch": 0.34, + "learning_rate": 4.4415234675557837e-05, + "loss": 0.5711, + "step": 8760 + }, + { + "epoch": 0.34, + "learning_rate": 4.441459348550911e-05, + "loss": 0.5608, + "step": 8761 + }, + { + "epoch": 0.34, + "learning_rate": 4.441395229546037e-05, + "loss": 0.6079, + "step": 8762 + }, + { + "epoch": 0.34, + "learning_rate": 4.4413311105411645e-05, + "loss": 0.6067, + "step": 8763 + }, + { + "epoch": 0.34, + "learning_rate": 4.4412669915362917e-05, + "loss": 0.5683, + "step": 8764 + }, + { + "epoch": 0.34, + "learning_rate": 4.441202872531419e-05, + "loss": 0.6122, + "step": 8765 + }, + { + "epoch": 0.34, + "learning_rate": 4.441138753526545e-05, + "loss": 0.5795, + "step": 8766 + }, + { + "epoch": 0.34, + "learning_rate": 4.4410746345216725e-05, + "loss": 0.5721, + "step": 8767 + }, + { + "epoch": 0.34, + "learning_rate": 4.441010515516799e-05, + "loss": 0.6031, + "step": 8768 + }, + { + "epoch": 0.34, + "learning_rate": 4.440946396511926e-05, + "loss": 0.5145, + "step": 8769 + }, + { + "epoch": 0.34, + "learning_rate": 4.440882277507053e-05, + "loss": 0.6036, + "step": 8770 + }, + { + "epoch": 0.34, + "learning_rate": 4.4408181585021805e-05, + "loss": 0.5342, + "step": 8771 + }, + { + "epoch": 0.34, + "learning_rate": 4.440754039497308e-05, + "loss": 0.5377, + "step": 8772 + }, + { + "epoch": 0.34, + "learning_rate": 4.440689920492434e-05, + "loss": 0.6469, + "step": 8773 + }, + { + "epoch": 0.34, + "learning_rate": 4.4406258014875613e-05, + "loss": 0.5312, + "step": 8774 + }, + { + "epoch": 0.34, + "learning_rate": 4.440561682482688e-05, + "loss": 0.5765, + "step": 8775 + }, + { + "epoch": 0.34, + "learning_rate": 4.440497563477815e-05, + "loss": 0.5734, + "step": 8776 + }, + { + "epoch": 0.34, + "learning_rate": 4.4404334444729415e-05, + "loss": 0.562, + "step": 8777 + }, + { + "epoch": 0.34, + "learning_rate": 4.440369325468069e-05, + "loss": 0.484, + "step": 8778 + }, + { + "epoch": 0.34, + "learning_rate": 4.440305206463196e-05, + "loss": 0.5952, + "step": 8779 + }, + { + "epoch": 0.34, + "learning_rate": 4.440241087458323e-05, + "loss": 0.5669, + "step": 8780 + }, + { + "epoch": 0.34, + "learning_rate": 4.44017696845345e-05, + "loss": 0.5917, + "step": 8781 + }, + { + "epoch": 0.34, + "learning_rate": 4.440112849448577e-05, + "loss": 0.5991, + "step": 8782 + }, + { + "epoch": 0.34, + "learning_rate": 4.440048730443704e-05, + "loss": 0.5956, + "step": 8783 + }, + { + "epoch": 0.34, + "learning_rate": 4.4399846114388303e-05, + "loss": 0.5935, + "step": 8784 + }, + { + "epoch": 0.34, + "learning_rate": 4.4399204924339575e-05, + "loss": 0.5771, + "step": 8785 + }, + { + "epoch": 0.34, + "learning_rate": 4.439856373429084e-05, + "loss": 0.4787, + "step": 8786 + }, + { + "epoch": 0.34, + "learning_rate": 4.439792254424212e-05, + "loss": 0.6257, + "step": 8787 + }, + { + "epoch": 0.34, + "learning_rate": 4.439728135419339e-05, + "loss": 0.4667, + "step": 8788 + }, + { + "epoch": 0.34, + "learning_rate": 4.4396640164144655e-05, + "loss": 0.5521, + "step": 8789 + }, + { + "epoch": 0.34, + "learning_rate": 4.439599897409593e-05, + "loss": 0.5951, + "step": 8790 + }, + { + "epoch": 0.34, + "learning_rate": 4.439535778404719e-05, + "loss": 0.6056, + "step": 8791 + }, + { + "epoch": 0.34, + "learning_rate": 4.4394716593998464e-05, + "loss": 0.5413, + "step": 8792 + }, + { + "epoch": 0.34, + "learning_rate": 4.439407540394973e-05, + "loss": 0.6237, + "step": 8793 + }, + { + "epoch": 0.34, + "learning_rate": 4.4393434213901e-05, + "loss": 0.4904, + "step": 8794 + }, + { + "epoch": 0.34, + "learning_rate": 4.439279302385227e-05, + "loss": 0.6256, + "step": 8795 + }, + { + "epoch": 0.34, + "learning_rate": 4.4392151833803544e-05, + "loss": 0.5831, + "step": 8796 + }, + { + "epoch": 0.34, + "learning_rate": 4.4391510643754815e-05, + "loss": 0.6471, + "step": 8797 + }, + { + "epoch": 0.34, + "learning_rate": 4.439086945370608e-05, + "loss": 0.6336, + "step": 8798 + }, + { + "epoch": 0.34, + "learning_rate": 4.439022826365735e-05, + "loss": 0.5065, + "step": 8799 + }, + { + "epoch": 0.34, + "learning_rate": 4.438958707360862e-05, + "loss": 0.5708, + "step": 8800 + }, + { + "epoch": 0.34, + "learning_rate": 4.438894588355989e-05, + "loss": 0.6035, + "step": 8801 + }, + { + "epoch": 0.34, + "learning_rate": 4.4388304693511154e-05, + "loss": 0.51, + "step": 8802 + }, + { + "epoch": 0.34, + "learning_rate": 4.438766350346243e-05, + "loss": 0.5028, + "step": 8803 + }, + { + "epoch": 0.34, + "learning_rate": 4.43870223134137e-05, + "loss": 0.5887, + "step": 8804 + }, + { + "epoch": 0.34, + "learning_rate": 4.438638112336497e-05, + "loss": 0.5578, + "step": 8805 + }, + { + "epoch": 0.34, + "learning_rate": 4.438573993331624e-05, + "loss": 0.5142, + "step": 8806 + }, + { + "epoch": 0.34, + "learning_rate": 4.4385098743267506e-05, + "loss": 0.5464, + "step": 8807 + }, + { + "epoch": 0.34, + "learning_rate": 4.438445755321878e-05, + "loss": 0.5294, + "step": 8808 + }, + { + "epoch": 0.34, + "learning_rate": 4.438381636317004e-05, + "loss": 0.4524, + "step": 8809 + }, + { + "epoch": 0.34, + "learning_rate": 4.4383175173121314e-05, + "loss": 0.5761, + "step": 8810 + }, + { + "epoch": 0.34, + "learning_rate": 4.4382533983072586e-05, + "loss": 0.6029, + "step": 8811 + }, + { + "epoch": 0.34, + "learning_rate": 4.438189279302386e-05, + "loss": 0.6398, + "step": 8812 + }, + { + "epoch": 0.34, + "learning_rate": 4.438125160297512e-05, + "loss": 0.6633, + "step": 8813 + }, + { + "epoch": 0.34, + "learning_rate": 4.4380610412926394e-05, + "loss": 0.6057, + "step": 8814 + }, + { + "epoch": 0.34, + "learning_rate": 4.4379969222877666e-05, + "loss": 0.625, + "step": 8815 + }, + { + "epoch": 0.34, + "learning_rate": 4.437932803282893e-05, + "loss": 0.6958, + "step": 8816 + }, + { + "epoch": 0.34, + "learning_rate": 4.43786868427802e-05, + "loss": 0.6101, + "step": 8817 + }, + { + "epoch": 0.34, + "learning_rate": 4.437804565273147e-05, + "loss": 0.5973, + "step": 8818 + }, + { + "epoch": 0.34, + "learning_rate": 4.437740446268274e-05, + "loss": 0.5454, + "step": 8819 + }, + { + "epoch": 0.34, + "learning_rate": 4.437676327263401e-05, + "loss": 0.5913, + "step": 8820 + }, + { + "epoch": 0.34, + "learning_rate": 4.437612208258528e-05, + "loss": 0.6017, + "step": 8821 + }, + { + "epoch": 0.34, + "learning_rate": 4.437548089253655e-05, + "loss": 0.494, + "step": 8822 + }, + { + "epoch": 0.34, + "learning_rate": 4.437483970248782e-05, + "loss": 0.6007, + "step": 8823 + }, + { + "epoch": 0.34, + "learning_rate": 4.437419851243909e-05, + "loss": 0.5673, + "step": 8824 + }, + { + "epoch": 0.34, + "learning_rate": 4.4373557322390356e-05, + "loss": 0.5238, + "step": 8825 + }, + { + "epoch": 0.34, + "learning_rate": 4.437291613234163e-05, + "loss": 0.6716, + "step": 8826 + }, + { + "epoch": 0.34, + "learning_rate": 4.43722749422929e-05, + "loss": 0.5319, + "step": 8827 + }, + { + "epoch": 0.34, + "learning_rate": 4.437163375224417e-05, + "loss": 0.5781, + "step": 8828 + }, + { + "epoch": 0.34, + "learning_rate": 4.4370992562195436e-05, + "loss": 0.55, + "step": 8829 + }, + { + "epoch": 0.34, + "learning_rate": 4.437035137214671e-05, + "loss": 0.5078, + "step": 8830 + }, + { + "epoch": 0.34, + "learning_rate": 4.436971018209797e-05, + "loss": 0.59, + "step": 8831 + }, + { + "epoch": 0.34, + "learning_rate": 4.4369068992049244e-05, + "loss": 0.6293, + "step": 8832 + }, + { + "epoch": 0.34, + "learning_rate": 4.4368427802000516e-05, + "loss": 0.6091, + "step": 8833 + }, + { + "epoch": 0.34, + "learning_rate": 4.436778661195178e-05, + "loss": 0.592, + "step": 8834 + }, + { + "epoch": 0.34, + "learning_rate": 4.436714542190305e-05, + "loss": 0.5411, + "step": 8835 + }, + { + "epoch": 0.34, + "learning_rate": 4.4366504231854324e-05, + "loss": 0.4376, + "step": 8836 + }, + { + "epoch": 0.34, + "learning_rate": 4.4365863041805596e-05, + "loss": 0.6065, + "step": 8837 + }, + { + "epoch": 0.34, + "learning_rate": 4.436522185175686e-05, + "loss": 0.6921, + "step": 8838 + }, + { + "epoch": 0.34, + "learning_rate": 4.436458066170813e-05, + "loss": 0.558, + "step": 8839 + }, + { + "epoch": 0.34, + "learning_rate": 4.43639394716594e-05, + "loss": 0.6565, + "step": 8840 + }, + { + "epoch": 0.34, + "learning_rate": 4.436329828161067e-05, + "loss": 0.6229, + "step": 8841 + }, + { + "epoch": 0.34, + "learning_rate": 4.436265709156194e-05, + "loss": 0.5987, + "step": 8842 + }, + { + "epoch": 0.34, + "learning_rate": 4.436201590151321e-05, + "loss": 0.6404, + "step": 8843 + }, + { + "epoch": 0.34, + "learning_rate": 4.4361374711464485e-05, + "loss": 0.6127, + "step": 8844 + }, + { + "epoch": 0.34, + "learning_rate": 4.436073352141575e-05, + "loss": 0.5308, + "step": 8845 + }, + { + "epoch": 0.34, + "learning_rate": 4.436009233136702e-05, + "loss": 0.5888, + "step": 8846 + }, + { + "epoch": 0.34, + "learning_rate": 4.4359451141318286e-05, + "loss": 0.5852, + "step": 8847 + }, + { + "epoch": 0.34, + "learning_rate": 4.435880995126956e-05, + "loss": 0.573, + "step": 8848 + }, + { + "epoch": 0.34, + "learning_rate": 4.435816876122082e-05, + "loss": 0.587, + "step": 8849 + }, + { + "epoch": 0.34, + "learning_rate": 4.4357527571172094e-05, + "loss": 0.5337, + "step": 8850 + }, + { + "epoch": 0.34, + "learning_rate": 4.4356886381123366e-05, + "loss": 0.5503, + "step": 8851 + }, + { + "epoch": 0.34, + "learning_rate": 4.435624519107464e-05, + "loss": 0.5981, + "step": 8852 + }, + { + "epoch": 0.34, + "learning_rate": 4.435560400102591e-05, + "loss": 0.6705, + "step": 8853 + }, + { + "epoch": 0.34, + "learning_rate": 4.4354962810977175e-05, + "loss": 0.589, + "step": 8854 + }, + { + "epoch": 0.34, + "learning_rate": 4.4354321620928446e-05, + "loss": 0.6211, + "step": 8855 + }, + { + "epoch": 0.34, + "learning_rate": 4.435368043087971e-05, + "loss": 0.6081, + "step": 8856 + }, + { + "epoch": 0.34, + "learning_rate": 4.435303924083098e-05, + "loss": 0.633, + "step": 8857 + }, + { + "epoch": 0.34, + "learning_rate": 4.4352398050782255e-05, + "loss": 0.5802, + "step": 8858 + }, + { + "epoch": 0.34, + "learning_rate": 4.4351756860733526e-05, + "loss": 0.7109, + "step": 8859 + }, + { + "epoch": 0.34, + "learning_rate": 4.43511156706848e-05, + "loss": 0.5092, + "step": 8860 + }, + { + "epoch": 0.34, + "learning_rate": 4.435047448063606e-05, + "loss": 0.5743, + "step": 8861 + }, + { + "epoch": 0.34, + "learning_rate": 4.4349833290587335e-05, + "loss": 0.5592, + "step": 8862 + }, + { + "epoch": 0.34, + "learning_rate": 4.43491921005386e-05, + "loss": 0.5535, + "step": 8863 + }, + { + "epoch": 0.34, + "learning_rate": 4.434855091048987e-05, + "loss": 0.5637, + "step": 8864 + }, + { + "epoch": 0.34, + "learning_rate": 4.4347909720441136e-05, + "loss": 0.5626, + "step": 8865 + }, + { + "epoch": 0.34, + "learning_rate": 4.434726853039241e-05, + "loss": 0.528, + "step": 8866 + }, + { + "epoch": 0.34, + "learning_rate": 4.434662734034368e-05, + "loss": 0.5875, + "step": 8867 + }, + { + "epoch": 0.34, + "learning_rate": 4.434598615029495e-05, + "loss": 0.6103, + "step": 8868 + }, + { + "epoch": 0.34, + "learning_rate": 4.434534496024622e-05, + "loss": 0.5939, + "step": 8869 + }, + { + "epoch": 0.34, + "learning_rate": 4.434470377019749e-05, + "loss": 0.6075, + "step": 8870 + }, + { + "epoch": 0.34, + "learning_rate": 4.434406258014876e-05, + "loss": 0.5509, + "step": 8871 + }, + { + "epoch": 0.34, + "learning_rate": 4.4343421390100025e-05, + "loss": 0.5603, + "step": 8872 + }, + { + "epoch": 0.34, + "learning_rate": 4.4342780200051297e-05, + "loss": 0.6138, + "step": 8873 + }, + { + "epoch": 0.34, + "learning_rate": 4.434213901000256e-05, + "loss": 0.573, + "step": 8874 + }, + { + "epoch": 0.34, + "learning_rate": 4.434149781995384e-05, + "loss": 0.5892, + "step": 8875 + }, + { + "epoch": 0.34, + "learning_rate": 4.434085662990511e-05, + "loss": 0.529, + "step": 8876 + }, + { + "epoch": 0.34, + "learning_rate": 4.434021543985638e-05, + "loss": 0.5836, + "step": 8877 + }, + { + "epoch": 0.34, + "learning_rate": 4.433957424980765e-05, + "loss": 0.5257, + "step": 8878 + }, + { + "epoch": 0.34, + "learning_rate": 4.433893305975891e-05, + "loss": 0.6047, + "step": 8879 + }, + { + "epoch": 0.34, + "learning_rate": 4.4338291869710185e-05, + "loss": 0.5706, + "step": 8880 + }, + { + "epoch": 0.34, + "learning_rate": 4.433765067966145e-05, + "loss": 0.5343, + "step": 8881 + }, + { + "epoch": 0.34, + "learning_rate": 4.433700948961272e-05, + "loss": 0.6146, + "step": 8882 + }, + { + "epoch": 0.34, + "learning_rate": 4.433636829956399e-05, + "loss": 0.5617, + "step": 8883 + }, + { + "epoch": 0.34, + "learning_rate": 4.4335727109515265e-05, + "loss": 0.6151, + "step": 8884 + }, + { + "epoch": 0.34, + "learning_rate": 4.433508591946654e-05, + "loss": 0.5351, + "step": 8885 + }, + { + "epoch": 0.34, + "learning_rate": 4.43344447294178e-05, + "loss": 0.5274, + "step": 8886 + }, + { + "epoch": 0.34, + "learning_rate": 4.4333803539369073e-05, + "loss": 0.6626, + "step": 8887 + }, + { + "epoch": 0.34, + "learning_rate": 4.433316234932034e-05, + "loss": 0.6112, + "step": 8888 + }, + { + "epoch": 0.34, + "learning_rate": 4.433252115927161e-05, + "loss": 0.5311, + "step": 8889 + }, + { + "epoch": 0.34, + "learning_rate": 4.4331879969222875e-05, + "loss": 0.5306, + "step": 8890 + }, + { + "epoch": 0.34, + "learning_rate": 4.433123877917415e-05, + "loss": 0.5374, + "step": 8891 + }, + { + "epoch": 0.34, + "learning_rate": 4.433059758912542e-05, + "loss": 0.5619, + "step": 8892 + }, + { + "epoch": 0.34, + "learning_rate": 4.432995639907669e-05, + "loss": 0.5548, + "step": 8893 + }, + { + "epoch": 0.34, + "learning_rate": 4.432931520902796e-05, + "loss": 0.5704, + "step": 8894 + }, + { + "epoch": 0.34, + "learning_rate": 4.432867401897923e-05, + "loss": 0.5817, + "step": 8895 + }, + { + "epoch": 0.34, + "learning_rate": 4.43280328289305e-05, + "loss": 0.5817, + "step": 8896 + }, + { + "epoch": 0.34, + "learning_rate": 4.4327391638881764e-05, + "loss": 0.5322, + "step": 8897 + }, + { + "epoch": 0.34, + "learning_rate": 4.4326750448833035e-05, + "loss": 0.5936, + "step": 8898 + }, + { + "epoch": 0.34, + "learning_rate": 4.432610925878431e-05, + "loss": 0.6623, + "step": 8899 + }, + { + "epoch": 0.34, + "learning_rate": 4.432546806873558e-05, + "loss": 0.5917, + "step": 8900 + }, + { + "epoch": 0.34, + "learning_rate": 4.4324826878686844e-05, + "loss": 0.5367, + "step": 8901 + }, + { + "epoch": 0.34, + "learning_rate": 4.4324185688638115e-05, + "loss": 0.5396, + "step": 8902 + }, + { + "epoch": 0.34, + "learning_rate": 4.432354449858939e-05, + "loss": 0.6593, + "step": 8903 + }, + { + "epoch": 0.34, + "learning_rate": 4.432290330854065e-05, + "loss": 0.5788, + "step": 8904 + }, + { + "epoch": 0.34, + "learning_rate": 4.4322262118491924e-05, + "loss": 0.5923, + "step": 8905 + }, + { + "epoch": 0.34, + "learning_rate": 4.432162092844319e-05, + "loss": 0.5958, + "step": 8906 + }, + { + "epoch": 0.34, + "learning_rate": 4.432097973839446e-05, + "loss": 0.6045, + "step": 8907 + }, + { + "epoch": 0.34, + "learning_rate": 4.432033854834573e-05, + "loss": 0.5737, + "step": 8908 + }, + { + "epoch": 0.34, + "learning_rate": 4.4319697358297004e-05, + "loss": 0.5282, + "step": 8909 + }, + { + "epoch": 0.34, + "learning_rate": 4.431905616824827e-05, + "loss": 0.5896, + "step": 8910 + }, + { + "epoch": 0.34, + "learning_rate": 4.431841497819954e-05, + "loss": 0.5236, + "step": 8911 + }, + { + "epoch": 0.34, + "learning_rate": 4.431777378815081e-05, + "loss": 0.5974, + "step": 8912 + }, + { + "epoch": 0.34, + "learning_rate": 4.431713259810208e-05, + "loss": 0.5656, + "step": 8913 + }, + { + "epoch": 0.34, + "learning_rate": 4.431649140805335e-05, + "loss": 0.6029, + "step": 8914 + }, + { + "epoch": 0.34, + "learning_rate": 4.431585021800462e-05, + "loss": 0.5781, + "step": 8915 + }, + { + "epoch": 0.34, + "learning_rate": 4.431520902795589e-05, + "loss": 0.5721, + "step": 8916 + }, + { + "epoch": 0.34, + "learning_rate": 4.431456783790716e-05, + "loss": 0.5506, + "step": 8917 + }, + { + "epoch": 0.34, + "learning_rate": 4.431392664785843e-05, + "loss": 0.5428, + "step": 8918 + }, + { + "epoch": 0.34, + "learning_rate": 4.4313285457809694e-05, + "loss": 0.5391, + "step": 8919 + }, + { + "epoch": 0.34, + "learning_rate": 4.4312644267760966e-05, + "loss": 0.5348, + "step": 8920 + }, + { + "epoch": 0.34, + "learning_rate": 4.431200307771224e-05, + "loss": 0.6221, + "step": 8921 + }, + { + "epoch": 0.34, + "learning_rate": 4.43113618876635e-05, + "loss": 0.6339, + "step": 8922 + }, + { + "epoch": 0.34, + "learning_rate": 4.4310720697614774e-05, + "loss": 0.5758, + "step": 8923 + }, + { + "epoch": 0.34, + "learning_rate": 4.4310079507566046e-05, + "loss": 0.6146, + "step": 8924 + }, + { + "epoch": 0.34, + "learning_rate": 4.430943831751732e-05, + "loss": 0.6444, + "step": 8925 + }, + { + "epoch": 0.34, + "learning_rate": 4.430879712746858e-05, + "loss": 0.5334, + "step": 8926 + }, + { + "epoch": 0.34, + "learning_rate": 4.4308155937419854e-05, + "loss": 0.595, + "step": 8927 + }, + { + "epoch": 0.34, + "learning_rate": 4.430751474737112e-05, + "loss": 0.6691, + "step": 8928 + }, + { + "epoch": 0.34, + "learning_rate": 4.430687355732239e-05, + "loss": 0.6233, + "step": 8929 + }, + { + "epoch": 0.34, + "learning_rate": 4.430623236727366e-05, + "loss": 0.5665, + "step": 8930 + }, + { + "epoch": 0.34, + "learning_rate": 4.4305591177224934e-05, + "loss": 0.5113, + "step": 8931 + }, + { + "epoch": 0.34, + "learning_rate": 4.4304949987176206e-05, + "loss": 0.6325, + "step": 8932 + }, + { + "epoch": 0.34, + "learning_rate": 4.430430879712747e-05, + "loss": 0.5382, + "step": 8933 + }, + { + "epoch": 0.34, + "learning_rate": 4.430366760707874e-05, + "loss": 0.6223, + "step": 8934 + }, + { + "epoch": 0.34, + "learning_rate": 4.430302641703001e-05, + "loss": 0.5443, + "step": 8935 + }, + { + "epoch": 0.34, + "learning_rate": 4.430238522698128e-05, + "loss": 0.5993, + "step": 8936 + }, + { + "epoch": 0.34, + "learning_rate": 4.4301744036932544e-05, + "loss": 0.5736, + "step": 8937 + }, + { + "epoch": 0.34, + "learning_rate": 4.4301102846883816e-05, + "loss": 0.5404, + "step": 8938 + }, + { + "epoch": 0.34, + "learning_rate": 4.430046165683509e-05, + "loss": 0.5824, + "step": 8939 + }, + { + "epoch": 0.34, + "learning_rate": 4.429982046678636e-05, + "loss": 0.5929, + "step": 8940 + }, + { + "epoch": 0.34, + "learning_rate": 4.429917927673763e-05, + "loss": 0.5532, + "step": 8941 + }, + { + "epoch": 0.34, + "learning_rate": 4.4298538086688896e-05, + "loss": 0.6189, + "step": 8942 + }, + { + "epoch": 0.34, + "learning_rate": 4.429789689664017e-05, + "loss": 0.6164, + "step": 8943 + }, + { + "epoch": 0.34, + "learning_rate": 4.429725570659143e-05, + "loss": 0.5661, + "step": 8944 + }, + { + "epoch": 0.34, + "learning_rate": 4.4296614516542704e-05, + "loss": 0.5422, + "step": 8945 + }, + { + "epoch": 0.34, + "learning_rate": 4.429597332649397e-05, + "loss": 0.5947, + "step": 8946 + }, + { + "epoch": 0.34, + "learning_rate": 4.429533213644525e-05, + "loss": 0.5458, + "step": 8947 + }, + { + "epoch": 0.34, + "learning_rate": 4.429469094639652e-05, + "loss": 0.6398, + "step": 8948 + }, + { + "epoch": 0.34, + "learning_rate": 4.4294049756347784e-05, + "loss": 0.6646, + "step": 8949 + }, + { + "epoch": 0.34, + "learning_rate": 4.4293408566299056e-05, + "loss": 0.5751, + "step": 8950 + }, + { + "epoch": 0.34, + "learning_rate": 4.429276737625032e-05, + "loss": 0.6475, + "step": 8951 + }, + { + "epoch": 0.34, + "learning_rate": 4.429212618620159e-05, + "loss": 0.5174, + "step": 8952 + }, + { + "epoch": 0.34, + "learning_rate": 4.429148499615286e-05, + "loss": 0.6546, + "step": 8953 + }, + { + "epoch": 0.34, + "learning_rate": 4.429084380610413e-05, + "loss": 0.5993, + "step": 8954 + }, + { + "epoch": 0.34, + "learning_rate": 4.42902026160554e-05, + "loss": 0.5987, + "step": 8955 + }, + { + "epoch": 0.34, + "learning_rate": 4.428956142600667e-05, + "loss": 0.5784, + "step": 8956 + }, + { + "epoch": 0.34, + "learning_rate": 4.4288920235957945e-05, + "loss": 0.5869, + "step": 8957 + }, + { + "epoch": 0.34, + "learning_rate": 4.428827904590921e-05, + "loss": 0.6017, + "step": 8958 + }, + { + "epoch": 0.34, + "learning_rate": 4.428763785586048e-05, + "loss": 0.4574, + "step": 8959 + }, + { + "epoch": 0.34, + "learning_rate": 4.4286996665811746e-05, + "loss": 0.5734, + "step": 8960 + }, + { + "epoch": 0.34, + "learning_rate": 4.428635547576302e-05, + "loss": 0.514, + "step": 8961 + }, + { + "epoch": 0.34, + "learning_rate": 4.428571428571428e-05, + "loss": 0.5665, + "step": 8962 + }, + { + "epoch": 0.34, + "learning_rate": 4.428507309566556e-05, + "loss": 0.5148, + "step": 8963 + }, + { + "epoch": 0.34, + "learning_rate": 4.4284431905616826e-05, + "loss": 0.6107, + "step": 8964 + }, + { + "epoch": 0.34, + "learning_rate": 4.42837907155681e-05, + "loss": 0.5437, + "step": 8965 + }, + { + "epoch": 0.34, + "learning_rate": 4.428314952551937e-05, + "loss": 0.6195, + "step": 8966 + }, + { + "epoch": 0.34, + "learning_rate": 4.4282508335470635e-05, + "loss": 0.6468, + "step": 8967 + }, + { + "epoch": 0.34, + "learning_rate": 4.4281867145421906e-05, + "loss": 0.4928, + "step": 8968 + }, + { + "epoch": 0.34, + "learning_rate": 4.428122595537317e-05, + "loss": 0.5522, + "step": 8969 + }, + { + "epoch": 0.34, + "learning_rate": 4.428058476532444e-05, + "loss": 0.6571, + "step": 8970 + }, + { + "epoch": 0.34, + "learning_rate": 4.4279943575275715e-05, + "loss": 0.4876, + "step": 8971 + }, + { + "epoch": 0.34, + "learning_rate": 4.4279302385226986e-05, + "loss": 0.5953, + "step": 8972 + }, + { + "epoch": 0.34, + "learning_rate": 4.427866119517825e-05, + "loss": 0.5776, + "step": 8973 + }, + { + "epoch": 0.35, + "learning_rate": 4.427802000512952e-05, + "loss": 0.5726, + "step": 8974 + }, + { + "epoch": 0.35, + "learning_rate": 4.4277378815080795e-05, + "loss": 0.6528, + "step": 8975 + }, + { + "epoch": 0.35, + "learning_rate": 4.427673762503206e-05, + "loss": 0.6299, + "step": 8976 + }, + { + "epoch": 0.35, + "learning_rate": 4.427609643498333e-05, + "loss": 0.6474, + "step": 8977 + }, + { + "epoch": 0.35, + "learning_rate": 4.4275455244934596e-05, + "loss": 0.6463, + "step": 8978 + }, + { + "epoch": 0.35, + "learning_rate": 4.427481405488587e-05, + "loss": 0.6528, + "step": 8979 + }, + { + "epoch": 0.35, + "learning_rate": 4.427417286483714e-05, + "loss": 0.5221, + "step": 8980 + }, + { + "epoch": 0.35, + "learning_rate": 4.427353167478841e-05, + "loss": 0.5521, + "step": 8981 + }, + { + "epoch": 0.35, + "learning_rate": 4.4272890484739676e-05, + "loss": 0.5462, + "step": 8982 + }, + { + "epoch": 0.35, + "learning_rate": 4.427224929469095e-05, + "loss": 0.5861, + "step": 8983 + }, + { + "epoch": 0.35, + "learning_rate": 4.427160810464222e-05, + "loss": 0.6044, + "step": 8984 + }, + { + "epoch": 0.35, + "learning_rate": 4.4270966914593485e-05, + "loss": 0.5208, + "step": 8985 + }, + { + "epoch": 0.35, + "learning_rate": 4.4270325724544757e-05, + "loss": 0.5967, + "step": 8986 + }, + { + "epoch": 0.35, + "learning_rate": 4.426968453449603e-05, + "loss": 0.6221, + "step": 8987 + }, + { + "epoch": 0.35, + "learning_rate": 4.42690433444473e-05, + "loss": 0.5583, + "step": 8988 + }, + { + "epoch": 0.35, + "learning_rate": 4.4268402154398565e-05, + "loss": 0.5523, + "step": 8989 + }, + { + "epoch": 0.35, + "learning_rate": 4.426776096434984e-05, + "loss": 0.5238, + "step": 8990 + }, + { + "epoch": 0.35, + "learning_rate": 4.42671197743011e-05, + "loss": 0.5288, + "step": 8991 + }, + { + "epoch": 0.35, + "learning_rate": 4.426647858425237e-05, + "loss": 0.611, + "step": 8992 + }, + { + "epoch": 0.35, + "learning_rate": 4.4265837394203645e-05, + "loss": 0.6405, + "step": 8993 + }, + { + "epoch": 0.35, + "learning_rate": 4.426519620415491e-05, + "loss": 0.4891, + "step": 8994 + }, + { + "epoch": 0.35, + "learning_rate": 4.426455501410618e-05, + "loss": 0.5962, + "step": 8995 + }, + { + "epoch": 0.35, + "learning_rate": 4.4263913824057453e-05, + "loss": 0.5259, + "step": 8996 + }, + { + "epoch": 0.35, + "learning_rate": 4.4263272634008725e-05, + "loss": 0.5666, + "step": 8997 + }, + { + "epoch": 0.35, + "learning_rate": 4.426263144395999e-05, + "loss": 0.5251, + "step": 8998 + }, + { + "epoch": 0.35, + "learning_rate": 4.426199025391126e-05, + "loss": 0.6714, + "step": 8999 + }, + { + "epoch": 0.35, + "learning_rate": 4.426134906386253e-05, + "loss": 0.5852, + "step": 9000 + }, + { + "epoch": 0.35, + "learning_rate": 4.42607078738138e-05, + "loss": 0.6663, + "step": 9001 + }, + { + "epoch": 0.35, + "learning_rate": 4.426006668376507e-05, + "loss": 0.5352, + "step": 9002 + }, + { + "epoch": 0.35, + "learning_rate": 4.425942549371634e-05, + "loss": 0.5897, + "step": 9003 + }, + { + "epoch": 0.35, + "learning_rate": 4.4258784303667614e-05, + "loss": 0.6067, + "step": 9004 + }, + { + "epoch": 0.35, + "learning_rate": 4.425814311361888e-05, + "loss": 0.6009, + "step": 9005 + }, + { + "epoch": 0.35, + "learning_rate": 4.425750192357015e-05, + "loss": 0.5639, + "step": 9006 + }, + { + "epoch": 0.35, + "learning_rate": 4.4256860733521415e-05, + "loss": 0.6072, + "step": 9007 + }, + { + "epoch": 0.35, + "learning_rate": 4.425621954347269e-05, + "loss": 0.5925, + "step": 9008 + }, + { + "epoch": 0.35, + "learning_rate": 4.425557835342395e-05, + "loss": 0.5288, + "step": 9009 + }, + { + "epoch": 0.35, + "learning_rate": 4.4254937163375224e-05, + "loss": 0.6382, + "step": 9010 + }, + { + "epoch": 0.35, + "learning_rate": 4.4254295973326495e-05, + "loss": 0.5468, + "step": 9011 + }, + { + "epoch": 0.35, + "learning_rate": 4.425365478327777e-05, + "loss": 0.5319, + "step": 9012 + }, + { + "epoch": 0.35, + "learning_rate": 4.425301359322904e-05, + "loss": 0.5536, + "step": 9013 + }, + { + "epoch": 0.35, + "learning_rate": 4.4252372403180304e-05, + "loss": 0.6033, + "step": 9014 + }, + { + "epoch": 0.35, + "learning_rate": 4.4251731213131575e-05, + "loss": 0.5423, + "step": 9015 + }, + { + "epoch": 0.35, + "learning_rate": 4.425109002308284e-05, + "loss": 0.4626, + "step": 9016 + }, + { + "epoch": 0.35, + "learning_rate": 4.425044883303411e-05, + "loss": 0.6654, + "step": 9017 + }, + { + "epoch": 0.35, + "learning_rate": 4.424980764298538e-05, + "loss": 0.5885, + "step": 9018 + }, + { + "epoch": 0.35, + "learning_rate": 4.4249166452936655e-05, + "loss": 0.653, + "step": 9019 + }, + { + "epoch": 0.35, + "learning_rate": 4.424852526288793e-05, + "loss": 0.5982, + "step": 9020 + }, + { + "epoch": 0.35, + "learning_rate": 4.424788407283919e-05, + "loss": 0.563, + "step": 9021 + }, + { + "epoch": 0.35, + "learning_rate": 4.4247242882790464e-05, + "loss": 0.5913, + "step": 9022 + }, + { + "epoch": 0.35, + "learning_rate": 4.424660169274173e-05, + "loss": 0.5659, + "step": 9023 + }, + { + "epoch": 0.35, + "learning_rate": 4.4245960502693e-05, + "loss": 0.6231, + "step": 9024 + }, + { + "epoch": 0.35, + "learning_rate": 4.4245319312644265e-05, + "loss": 0.5678, + "step": 9025 + }, + { + "epoch": 0.35, + "learning_rate": 4.424467812259554e-05, + "loss": 0.5146, + "step": 9026 + }, + { + "epoch": 0.35, + "learning_rate": 4.424403693254681e-05, + "loss": 0.5865, + "step": 9027 + }, + { + "epoch": 0.35, + "learning_rate": 4.424339574249808e-05, + "loss": 0.5339, + "step": 9028 + }, + { + "epoch": 0.35, + "learning_rate": 4.424275455244935e-05, + "loss": 0.5308, + "step": 9029 + }, + { + "epoch": 0.35, + "learning_rate": 4.424211336240062e-05, + "loss": 0.6516, + "step": 9030 + }, + { + "epoch": 0.35, + "learning_rate": 4.424147217235189e-05, + "loss": 0.5952, + "step": 9031 + }, + { + "epoch": 0.35, + "learning_rate": 4.4240830982303154e-05, + "loss": 0.6467, + "step": 9032 + }, + { + "epoch": 0.35, + "learning_rate": 4.4240189792254426e-05, + "loss": 0.582, + "step": 9033 + }, + { + "epoch": 0.35, + "learning_rate": 4.423954860220569e-05, + "loss": 0.4656, + "step": 9034 + }, + { + "epoch": 0.35, + "learning_rate": 4.423890741215697e-05, + "loss": 0.5585, + "step": 9035 + }, + { + "epoch": 0.35, + "learning_rate": 4.4238266222108234e-05, + "loss": 0.4733, + "step": 9036 + }, + { + "epoch": 0.35, + "learning_rate": 4.4237625032059506e-05, + "loss": 0.5201, + "step": 9037 + }, + { + "epoch": 0.35, + "learning_rate": 4.423698384201078e-05, + "loss": 0.6286, + "step": 9038 + }, + { + "epoch": 0.35, + "learning_rate": 4.423634265196204e-05, + "loss": 0.5455, + "step": 9039 + }, + { + "epoch": 0.35, + "learning_rate": 4.4235701461913314e-05, + "loss": 0.6422, + "step": 9040 + }, + { + "epoch": 0.35, + "learning_rate": 4.423506027186458e-05, + "loss": 0.5765, + "step": 9041 + }, + { + "epoch": 0.35, + "learning_rate": 4.423441908181585e-05, + "loss": 0.6116, + "step": 9042 + }, + { + "epoch": 0.35, + "learning_rate": 4.423377789176712e-05, + "loss": 0.4873, + "step": 9043 + }, + { + "epoch": 0.35, + "learning_rate": 4.4233136701718394e-05, + "loss": 0.6701, + "step": 9044 + }, + { + "epoch": 0.35, + "learning_rate": 4.4232495511669666e-05, + "loss": 0.6306, + "step": 9045 + }, + { + "epoch": 0.35, + "learning_rate": 4.423185432162093e-05, + "loss": 0.5308, + "step": 9046 + }, + { + "epoch": 0.35, + "learning_rate": 4.42312131315722e-05, + "loss": 0.5996, + "step": 9047 + }, + { + "epoch": 0.35, + "learning_rate": 4.423057194152347e-05, + "loss": 0.5734, + "step": 9048 + }, + { + "epoch": 0.35, + "learning_rate": 4.422993075147474e-05, + "loss": 0.5504, + "step": 9049 + }, + { + "epoch": 0.35, + "learning_rate": 4.4229289561426004e-05, + "loss": 0.5099, + "step": 9050 + }, + { + "epoch": 0.35, + "learning_rate": 4.4228648371377276e-05, + "loss": 0.5517, + "step": 9051 + }, + { + "epoch": 0.35, + "learning_rate": 4.422800718132855e-05, + "loss": 0.5632, + "step": 9052 + }, + { + "epoch": 0.35, + "learning_rate": 4.422736599127982e-05, + "loss": 0.4644, + "step": 9053 + }, + { + "epoch": 0.35, + "learning_rate": 4.422672480123109e-05, + "loss": 0.5442, + "step": 9054 + }, + { + "epoch": 0.35, + "learning_rate": 4.4226083611182356e-05, + "loss": 0.5222, + "step": 9055 + }, + { + "epoch": 0.35, + "learning_rate": 4.422544242113363e-05, + "loss": 0.5001, + "step": 9056 + }, + { + "epoch": 0.35, + "learning_rate": 4.422480123108489e-05, + "loss": 0.5483, + "step": 9057 + }, + { + "epoch": 0.35, + "learning_rate": 4.4224160041036164e-05, + "loss": 0.498, + "step": 9058 + }, + { + "epoch": 0.35, + "learning_rate": 4.4223518850987436e-05, + "loss": 0.5663, + "step": 9059 + }, + { + "epoch": 0.35, + "learning_rate": 4.422287766093871e-05, + "loss": 0.6125, + "step": 9060 + }, + { + "epoch": 0.35, + "learning_rate": 4.422223647088997e-05, + "loss": 0.5647, + "step": 9061 + }, + { + "epoch": 0.35, + "learning_rate": 4.4221595280841244e-05, + "loss": 0.6316, + "step": 9062 + }, + { + "epoch": 0.35, + "learning_rate": 4.4220954090792516e-05, + "loss": 0.5232, + "step": 9063 + }, + { + "epoch": 0.35, + "learning_rate": 4.422031290074378e-05, + "loss": 0.5933, + "step": 9064 + }, + { + "epoch": 0.35, + "learning_rate": 4.421967171069505e-05, + "loss": 0.6423, + "step": 9065 + }, + { + "epoch": 0.35, + "learning_rate": 4.421903052064632e-05, + "loss": 0.4689, + "step": 9066 + }, + { + "epoch": 0.35, + "learning_rate": 4.421838933059759e-05, + "loss": 0.6972, + "step": 9067 + }, + { + "epoch": 0.35, + "learning_rate": 4.421774814054886e-05, + "loss": 0.5566, + "step": 9068 + }, + { + "epoch": 0.35, + "learning_rate": 4.421710695050013e-05, + "loss": 0.5578, + "step": 9069 + }, + { + "epoch": 0.35, + "learning_rate": 4.42164657604514e-05, + "loss": 0.5646, + "step": 9070 + }, + { + "epoch": 0.35, + "learning_rate": 4.421582457040267e-05, + "loss": 0.5979, + "step": 9071 + }, + { + "epoch": 0.35, + "learning_rate": 4.421518338035394e-05, + "loss": 0.6261, + "step": 9072 + }, + { + "epoch": 0.35, + "learning_rate": 4.4214542190305206e-05, + "loss": 0.6917, + "step": 9073 + }, + { + "epoch": 0.35, + "learning_rate": 4.421390100025648e-05, + "loss": 0.5388, + "step": 9074 + }, + { + "epoch": 0.35, + "learning_rate": 4.421325981020775e-05, + "loss": 0.5251, + "step": 9075 + }, + { + "epoch": 0.35, + "learning_rate": 4.421261862015902e-05, + "loss": 0.5706, + "step": 9076 + }, + { + "epoch": 0.35, + "learning_rate": 4.4211977430110286e-05, + "loss": 0.5575, + "step": 9077 + }, + { + "epoch": 0.35, + "learning_rate": 4.421133624006156e-05, + "loss": 0.4916, + "step": 9078 + }, + { + "epoch": 0.35, + "learning_rate": 4.421069505001282e-05, + "loss": 0.4738, + "step": 9079 + }, + { + "epoch": 0.35, + "learning_rate": 4.4210053859964095e-05, + "loss": 0.5042, + "step": 9080 + }, + { + "epoch": 0.35, + "learning_rate": 4.4209412669915366e-05, + "loss": 0.6254, + "step": 9081 + }, + { + "epoch": 0.35, + "learning_rate": 4.420877147986663e-05, + "loss": 0.5618, + "step": 9082 + }, + { + "epoch": 0.35, + "learning_rate": 4.42081302898179e-05, + "loss": 0.5895, + "step": 9083 + }, + { + "epoch": 0.35, + "learning_rate": 4.4207489099769175e-05, + "loss": 0.578, + "step": 9084 + }, + { + "epoch": 0.35, + "learning_rate": 4.4206847909720446e-05, + "loss": 0.6778, + "step": 9085 + }, + { + "epoch": 0.35, + "learning_rate": 4.420620671967171e-05, + "loss": 0.5139, + "step": 9086 + }, + { + "epoch": 0.35, + "learning_rate": 4.420556552962298e-05, + "loss": 0.4865, + "step": 9087 + }, + { + "epoch": 0.35, + "learning_rate": 4.420492433957425e-05, + "loss": 0.6093, + "step": 9088 + }, + { + "epoch": 0.35, + "learning_rate": 4.420428314952552e-05, + "loss": 0.5306, + "step": 9089 + }, + { + "epoch": 0.35, + "learning_rate": 4.420364195947679e-05, + "loss": 0.6052, + "step": 9090 + }, + { + "epoch": 0.35, + "learning_rate": 4.420300076942806e-05, + "loss": 0.4233, + "step": 9091 + }, + { + "epoch": 0.35, + "learning_rate": 4.4202359579379335e-05, + "loss": 0.5492, + "step": 9092 + }, + { + "epoch": 0.35, + "learning_rate": 4.42017183893306e-05, + "loss": 0.6347, + "step": 9093 + }, + { + "epoch": 0.35, + "learning_rate": 4.420107719928187e-05, + "loss": 0.5181, + "step": 9094 + }, + { + "epoch": 0.35, + "learning_rate": 4.4200436009233137e-05, + "loss": 0.6536, + "step": 9095 + }, + { + "epoch": 0.35, + "learning_rate": 4.419979481918441e-05, + "loss": 0.6164, + "step": 9096 + }, + { + "epoch": 0.35, + "learning_rate": 4.419915362913567e-05, + "loss": 0.4769, + "step": 9097 + }, + { + "epoch": 0.35, + "learning_rate": 4.4198512439086945e-05, + "loss": 0.5121, + "step": 9098 + }, + { + "epoch": 0.35, + "learning_rate": 4.4197871249038217e-05, + "loss": 0.5362, + "step": 9099 + }, + { + "epoch": 0.35, + "learning_rate": 4.419723005898949e-05, + "loss": 0.5881, + "step": 9100 + }, + { + "epoch": 0.35, + "learning_rate": 4.419658886894076e-05, + "loss": 0.5296, + "step": 9101 + }, + { + "epoch": 0.35, + "learning_rate": 4.4195947678892025e-05, + "loss": 0.4987, + "step": 9102 + }, + { + "epoch": 0.35, + "learning_rate": 4.41953064888433e-05, + "loss": 0.6525, + "step": 9103 + }, + { + "epoch": 0.35, + "learning_rate": 4.419466529879456e-05, + "loss": 0.5988, + "step": 9104 + }, + { + "epoch": 0.35, + "learning_rate": 4.419402410874583e-05, + "loss": 0.5019, + "step": 9105 + }, + { + "epoch": 0.35, + "learning_rate": 4.41933829186971e-05, + "loss": 0.6468, + "step": 9106 + }, + { + "epoch": 0.35, + "learning_rate": 4.419274172864838e-05, + "loss": 0.5332, + "step": 9107 + }, + { + "epoch": 0.35, + "learning_rate": 4.419210053859965e-05, + "loss": 0.5515, + "step": 9108 + }, + { + "epoch": 0.35, + "learning_rate": 4.4191459348550913e-05, + "loss": 0.6113, + "step": 9109 + }, + { + "epoch": 0.35, + "learning_rate": 4.4190818158502185e-05, + "loss": 0.4531, + "step": 9110 + }, + { + "epoch": 0.35, + "learning_rate": 4.419017696845345e-05, + "loss": 0.6312, + "step": 9111 + }, + { + "epoch": 0.35, + "learning_rate": 4.418953577840472e-05, + "loss": 0.6958, + "step": 9112 + }, + { + "epoch": 0.35, + "learning_rate": 4.418889458835599e-05, + "loss": 0.6942, + "step": 9113 + }, + { + "epoch": 0.35, + "learning_rate": 4.418825339830726e-05, + "loss": 0.6334, + "step": 9114 + }, + { + "epoch": 0.35, + "learning_rate": 4.418761220825853e-05, + "loss": 0.7115, + "step": 9115 + }, + { + "epoch": 0.35, + "learning_rate": 4.41869710182098e-05, + "loss": 0.5723, + "step": 9116 + }, + { + "epoch": 0.35, + "learning_rate": 4.4186329828161074e-05, + "loss": 0.5612, + "step": 9117 + }, + { + "epoch": 0.35, + "learning_rate": 4.418568863811234e-05, + "loss": 0.5672, + "step": 9118 + }, + { + "epoch": 0.35, + "learning_rate": 4.418504744806361e-05, + "loss": 0.5801, + "step": 9119 + }, + { + "epoch": 0.35, + "learning_rate": 4.4184406258014875e-05, + "loss": 0.6344, + "step": 9120 + }, + { + "epoch": 0.35, + "learning_rate": 4.418376506796615e-05, + "loss": 0.592, + "step": 9121 + }, + { + "epoch": 0.35, + "learning_rate": 4.418312387791741e-05, + "loss": 0.5755, + "step": 9122 + }, + { + "epoch": 0.35, + "learning_rate": 4.418248268786869e-05, + "loss": 0.5869, + "step": 9123 + }, + { + "epoch": 0.35, + "learning_rate": 4.4181841497819955e-05, + "loss": 0.5607, + "step": 9124 + }, + { + "epoch": 0.35, + "learning_rate": 4.418120030777123e-05, + "loss": 0.557, + "step": 9125 + }, + { + "epoch": 0.35, + "learning_rate": 4.41805591177225e-05, + "loss": 0.6241, + "step": 9126 + }, + { + "epoch": 0.35, + "learning_rate": 4.4179917927673764e-05, + "loss": 0.499, + "step": 9127 + }, + { + "epoch": 0.35, + "learning_rate": 4.4179276737625035e-05, + "loss": 0.5718, + "step": 9128 + }, + { + "epoch": 0.35, + "learning_rate": 4.41786355475763e-05, + "loss": 0.5437, + "step": 9129 + }, + { + "epoch": 0.35, + "learning_rate": 4.417799435752757e-05, + "loss": 0.6543, + "step": 9130 + }, + { + "epoch": 0.35, + "learning_rate": 4.4177353167478844e-05, + "loss": 0.5918, + "step": 9131 + }, + { + "epoch": 0.35, + "learning_rate": 4.4176711977430115e-05, + "loss": 0.6151, + "step": 9132 + }, + { + "epoch": 0.35, + "learning_rate": 4.417607078738138e-05, + "loss": 0.6329, + "step": 9133 + }, + { + "epoch": 0.35, + "learning_rate": 4.417542959733265e-05, + "loss": 0.5926, + "step": 9134 + }, + { + "epoch": 0.35, + "learning_rate": 4.4174788407283924e-05, + "loss": 0.5559, + "step": 9135 + }, + { + "epoch": 0.35, + "learning_rate": 4.417414721723519e-05, + "loss": 0.5785, + "step": 9136 + }, + { + "epoch": 0.35, + "learning_rate": 4.417350602718646e-05, + "loss": 0.5926, + "step": 9137 + }, + { + "epoch": 0.35, + "learning_rate": 4.4172864837137725e-05, + "loss": 0.5805, + "step": 9138 + }, + { + "epoch": 0.35, + "learning_rate": 4.4172223647089e-05, + "loss": 0.5047, + "step": 9139 + }, + { + "epoch": 0.35, + "learning_rate": 4.417158245704027e-05, + "loss": 0.5756, + "step": 9140 + }, + { + "epoch": 0.35, + "learning_rate": 4.417094126699154e-05, + "loss": 0.5835, + "step": 9141 + }, + { + "epoch": 0.35, + "learning_rate": 4.4170300076942806e-05, + "loss": 0.5451, + "step": 9142 + }, + { + "epoch": 0.35, + "learning_rate": 4.416965888689408e-05, + "loss": 0.5246, + "step": 9143 + }, + { + "epoch": 0.35, + "learning_rate": 4.416901769684535e-05, + "loss": 0.4877, + "step": 9144 + }, + { + "epoch": 0.35, + "learning_rate": 4.4168376506796614e-05, + "loss": 0.4968, + "step": 9145 + }, + { + "epoch": 0.35, + "learning_rate": 4.4167735316747886e-05, + "loss": 0.5903, + "step": 9146 + }, + { + "epoch": 0.35, + "learning_rate": 4.416709412669916e-05, + "loss": 0.4729, + "step": 9147 + }, + { + "epoch": 0.35, + "learning_rate": 4.416645293665043e-05, + "loss": 0.5702, + "step": 9148 + }, + { + "epoch": 0.35, + "learning_rate": 4.4165811746601694e-05, + "loss": 0.6017, + "step": 9149 + }, + { + "epoch": 0.35, + "learning_rate": 4.4165170556552966e-05, + "loss": 0.5627, + "step": 9150 + }, + { + "epoch": 0.35, + "learning_rate": 4.416452936650423e-05, + "loss": 0.4163, + "step": 9151 + }, + { + "epoch": 0.35, + "learning_rate": 4.41638881764555e-05, + "loss": 0.6079, + "step": 9152 + }, + { + "epoch": 0.35, + "learning_rate": 4.4163246986406774e-05, + "loss": 0.5663, + "step": 9153 + }, + { + "epoch": 0.35, + "learning_rate": 4.416260579635804e-05, + "loss": 0.6231, + "step": 9154 + }, + { + "epoch": 0.35, + "learning_rate": 4.416196460630931e-05, + "loss": 0.6465, + "step": 9155 + }, + { + "epoch": 0.35, + "learning_rate": 4.416132341626058e-05, + "loss": 0.5468, + "step": 9156 + }, + { + "epoch": 0.35, + "learning_rate": 4.4160682226211854e-05, + "loss": 0.5239, + "step": 9157 + }, + { + "epoch": 0.35, + "learning_rate": 4.416004103616312e-05, + "loss": 0.5425, + "step": 9158 + }, + { + "epoch": 0.35, + "learning_rate": 4.415939984611439e-05, + "loss": 0.5695, + "step": 9159 + }, + { + "epoch": 0.35, + "learning_rate": 4.4158758656065656e-05, + "loss": 0.6037, + "step": 9160 + }, + { + "epoch": 0.35, + "learning_rate": 4.415811746601693e-05, + "loss": 0.5651, + "step": 9161 + }, + { + "epoch": 0.35, + "learning_rate": 4.41574762759682e-05, + "loss": 0.5638, + "step": 9162 + }, + { + "epoch": 0.35, + "learning_rate": 4.415683508591947e-05, + "loss": 0.5647, + "step": 9163 + }, + { + "epoch": 0.35, + "learning_rate": 4.415619389587074e-05, + "loss": 0.5462, + "step": 9164 + }, + { + "epoch": 0.35, + "learning_rate": 4.415555270582201e-05, + "loss": 0.6058, + "step": 9165 + }, + { + "epoch": 0.35, + "learning_rate": 4.415491151577328e-05, + "loss": 0.5374, + "step": 9166 + }, + { + "epoch": 0.35, + "learning_rate": 4.4154270325724544e-05, + "loss": 0.5093, + "step": 9167 + }, + { + "epoch": 0.35, + "learning_rate": 4.4153629135675816e-05, + "loss": 0.5379, + "step": 9168 + }, + { + "epoch": 0.35, + "learning_rate": 4.415298794562708e-05, + "loss": 0.6256, + "step": 9169 + }, + { + "epoch": 0.35, + "learning_rate": 4.415234675557835e-05, + "loss": 0.5655, + "step": 9170 + }, + { + "epoch": 0.35, + "learning_rate": 4.4151705565529624e-05, + "loss": 0.5761, + "step": 9171 + }, + { + "epoch": 0.35, + "learning_rate": 4.4151064375480896e-05, + "loss": 0.5975, + "step": 9172 + }, + { + "epoch": 0.35, + "learning_rate": 4.415042318543217e-05, + "loss": 0.607, + "step": 9173 + }, + { + "epoch": 0.35, + "learning_rate": 4.414978199538343e-05, + "loss": 0.5372, + "step": 9174 + }, + { + "epoch": 0.35, + "learning_rate": 4.4149140805334704e-05, + "loss": 0.5186, + "step": 9175 + }, + { + "epoch": 0.35, + "learning_rate": 4.414849961528597e-05, + "loss": 0.4994, + "step": 9176 + }, + { + "epoch": 0.35, + "learning_rate": 4.414785842523724e-05, + "loss": 0.6624, + "step": 9177 + }, + { + "epoch": 0.35, + "learning_rate": 4.4147217235188506e-05, + "loss": 0.5441, + "step": 9178 + }, + { + "epoch": 0.35, + "learning_rate": 4.4146576045139785e-05, + "loss": 0.6016, + "step": 9179 + }, + { + "epoch": 0.35, + "learning_rate": 4.4145934855091056e-05, + "loss": 0.5469, + "step": 9180 + }, + { + "epoch": 0.35, + "learning_rate": 4.414529366504232e-05, + "loss": 0.5187, + "step": 9181 + }, + { + "epoch": 0.35, + "learning_rate": 4.414465247499359e-05, + "loss": 0.5723, + "step": 9182 + }, + { + "epoch": 0.35, + "learning_rate": 4.414401128494486e-05, + "loss": 0.6361, + "step": 9183 + }, + { + "epoch": 0.35, + "learning_rate": 4.414337009489613e-05, + "loss": 0.474, + "step": 9184 + }, + { + "epoch": 0.35, + "learning_rate": 4.4142728904847394e-05, + "loss": 0.5982, + "step": 9185 + }, + { + "epoch": 0.35, + "learning_rate": 4.4142087714798666e-05, + "loss": 0.5752, + "step": 9186 + }, + { + "epoch": 0.35, + "learning_rate": 4.414144652474994e-05, + "loss": 0.5798, + "step": 9187 + }, + { + "epoch": 0.35, + "learning_rate": 4.414080533470121e-05, + "loss": 0.5651, + "step": 9188 + }, + { + "epoch": 0.35, + "learning_rate": 4.414016414465248e-05, + "loss": 0.4885, + "step": 9189 + }, + { + "epoch": 0.35, + "learning_rate": 4.4139522954603746e-05, + "loss": 0.6284, + "step": 9190 + }, + { + "epoch": 0.35, + "learning_rate": 4.413888176455502e-05, + "loss": 0.6554, + "step": 9191 + }, + { + "epoch": 0.35, + "learning_rate": 4.413824057450628e-05, + "loss": 0.504, + "step": 9192 + }, + { + "epoch": 0.35, + "learning_rate": 4.4137599384457555e-05, + "loss": 0.6471, + "step": 9193 + }, + { + "epoch": 0.35, + "learning_rate": 4.413695819440882e-05, + "loss": 0.5617, + "step": 9194 + }, + { + "epoch": 0.35, + "learning_rate": 4.41363170043601e-05, + "loss": 0.5948, + "step": 9195 + }, + { + "epoch": 0.35, + "learning_rate": 4.413567581431136e-05, + "loss": 0.562, + "step": 9196 + }, + { + "epoch": 0.35, + "learning_rate": 4.4135034624262635e-05, + "loss": 0.5022, + "step": 9197 + }, + { + "epoch": 0.35, + "learning_rate": 4.4134393434213906e-05, + "loss": 0.6018, + "step": 9198 + }, + { + "epoch": 0.35, + "learning_rate": 4.413375224416517e-05, + "loss": 0.492, + "step": 9199 + }, + { + "epoch": 0.35, + "learning_rate": 4.413311105411644e-05, + "loss": 0.6347, + "step": 9200 + }, + { + "epoch": 0.35, + "learning_rate": 4.413246986406771e-05, + "loss": 0.5562, + "step": 9201 + }, + { + "epoch": 0.35, + "learning_rate": 4.413182867401898e-05, + "loss": 0.548, + "step": 9202 + }, + { + "epoch": 0.35, + "learning_rate": 4.413118748397025e-05, + "loss": 0.6217, + "step": 9203 + }, + { + "epoch": 0.35, + "learning_rate": 4.413054629392152e-05, + "loss": 0.6226, + "step": 9204 + }, + { + "epoch": 0.35, + "learning_rate": 4.412990510387279e-05, + "loss": 0.5251, + "step": 9205 + }, + { + "epoch": 0.35, + "learning_rate": 4.412926391382406e-05, + "loss": 0.5225, + "step": 9206 + }, + { + "epoch": 0.35, + "learning_rate": 4.412862272377533e-05, + "loss": 0.5629, + "step": 9207 + }, + { + "epoch": 0.35, + "learning_rate": 4.4127981533726597e-05, + "loss": 0.5678, + "step": 9208 + }, + { + "epoch": 0.35, + "learning_rate": 4.412734034367787e-05, + "loss": 0.6065, + "step": 9209 + }, + { + "epoch": 0.35, + "learning_rate": 4.412669915362913e-05, + "loss": 0.6257, + "step": 9210 + }, + { + "epoch": 0.35, + "learning_rate": 4.412605796358041e-05, + "loss": 0.5655, + "step": 9211 + }, + { + "epoch": 0.35, + "learning_rate": 4.412541677353168e-05, + "loss": 0.6963, + "step": 9212 + }, + { + "epoch": 0.35, + "learning_rate": 4.412477558348295e-05, + "loss": 0.5042, + "step": 9213 + }, + { + "epoch": 0.35, + "learning_rate": 4.412413439343421e-05, + "loss": 0.5133, + "step": 9214 + }, + { + "epoch": 0.35, + "learning_rate": 4.4123493203385485e-05, + "loss": 0.6073, + "step": 9215 + }, + { + "epoch": 0.35, + "learning_rate": 4.412285201333676e-05, + "loss": 0.4683, + "step": 9216 + }, + { + "epoch": 0.35, + "learning_rate": 4.412221082328802e-05, + "loss": 0.5042, + "step": 9217 + }, + { + "epoch": 0.35, + "learning_rate": 4.412156963323929e-05, + "loss": 0.5841, + "step": 9218 + }, + { + "epoch": 0.35, + "learning_rate": 4.4120928443190565e-05, + "loss": 0.5625, + "step": 9219 + }, + { + "epoch": 0.35, + "learning_rate": 4.412028725314184e-05, + "loss": 0.5934, + "step": 9220 + }, + { + "epoch": 0.35, + "learning_rate": 4.41196460630931e-05, + "loss": 0.5673, + "step": 9221 + }, + { + "epoch": 0.35, + "learning_rate": 4.4119004873044373e-05, + "loss": 0.5785, + "step": 9222 + }, + { + "epoch": 0.35, + "learning_rate": 4.411836368299564e-05, + "loss": 0.5335, + "step": 9223 + }, + { + "epoch": 0.35, + "learning_rate": 4.411772249294691e-05, + "loss": 0.5535, + "step": 9224 + }, + { + "epoch": 0.35, + "learning_rate": 4.411708130289818e-05, + "loss": 0.6375, + "step": 9225 + }, + { + "epoch": 0.35, + "learning_rate": 4.411644011284945e-05, + "loss": 0.6777, + "step": 9226 + }, + { + "epoch": 0.35, + "learning_rate": 4.411579892280072e-05, + "loss": 0.5568, + "step": 9227 + }, + { + "epoch": 0.35, + "learning_rate": 4.411515773275199e-05, + "loss": 0.6349, + "step": 9228 + }, + { + "epoch": 0.35, + "learning_rate": 4.411451654270326e-05, + "loss": 0.6233, + "step": 9229 + }, + { + "epoch": 0.35, + "learning_rate": 4.411387535265453e-05, + "loss": 0.5441, + "step": 9230 + }, + { + "epoch": 0.35, + "learning_rate": 4.41132341626058e-05, + "loss": 0.568, + "step": 9231 + }, + { + "epoch": 0.35, + "learning_rate": 4.411259297255707e-05, + "loss": 0.5689, + "step": 9232 + }, + { + "epoch": 0.35, + "learning_rate": 4.4111951782508335e-05, + "loss": 0.6053, + "step": 9233 + }, + { + "epoch": 0.36, + "learning_rate": 4.411131059245961e-05, + "loss": 0.5779, + "step": 9234 + }, + { + "epoch": 0.36, + "learning_rate": 4.411066940241088e-05, + "loss": 0.5599, + "step": 9235 + }, + { + "epoch": 0.36, + "learning_rate": 4.411002821236215e-05, + "loss": 0.596, + "step": 9236 + }, + { + "epoch": 0.36, + "learning_rate": 4.4109387022313415e-05, + "loss": 0.5414, + "step": 9237 + }, + { + "epoch": 0.36, + "learning_rate": 4.410874583226469e-05, + "loss": 0.5403, + "step": 9238 + }, + { + "epoch": 0.36, + "learning_rate": 4.410810464221595e-05, + "loss": 0.5998, + "step": 9239 + }, + { + "epoch": 0.36, + "learning_rate": 4.4107463452167224e-05, + "loss": 0.5374, + "step": 9240 + }, + { + "epoch": 0.36, + "learning_rate": 4.4106822262118495e-05, + "loss": 0.5637, + "step": 9241 + }, + { + "epoch": 0.36, + "learning_rate": 4.410618107206976e-05, + "loss": 0.6064, + "step": 9242 + }, + { + "epoch": 0.36, + "learning_rate": 4.410553988202103e-05, + "loss": 0.6374, + "step": 9243 + }, + { + "epoch": 0.36, + "learning_rate": 4.4104898691972304e-05, + "loss": 0.5539, + "step": 9244 + }, + { + "epoch": 0.36, + "learning_rate": 4.4104257501923576e-05, + "loss": 0.5353, + "step": 9245 + }, + { + "epoch": 0.36, + "learning_rate": 4.410361631187484e-05, + "loss": 0.5692, + "step": 9246 + }, + { + "epoch": 0.36, + "learning_rate": 4.410297512182611e-05, + "loss": 0.5501, + "step": 9247 + }, + { + "epoch": 0.36, + "learning_rate": 4.410233393177738e-05, + "loss": 0.5826, + "step": 9248 + }, + { + "epoch": 0.36, + "learning_rate": 4.410169274172865e-05, + "loss": 0.5138, + "step": 9249 + }, + { + "epoch": 0.36, + "learning_rate": 4.410105155167992e-05, + "loss": 0.586, + "step": 9250 + }, + { + "epoch": 0.36, + "learning_rate": 4.410041036163119e-05, + "loss": 0.5921, + "step": 9251 + }, + { + "epoch": 0.36, + "learning_rate": 4.4099769171582464e-05, + "loss": 0.5842, + "step": 9252 + }, + { + "epoch": 0.36, + "learning_rate": 4.409912798153373e-05, + "loss": 0.6832, + "step": 9253 + }, + { + "epoch": 0.36, + "learning_rate": 4.4098486791485e-05, + "loss": 0.56, + "step": 9254 + }, + { + "epoch": 0.36, + "learning_rate": 4.4097845601436266e-05, + "loss": 0.5601, + "step": 9255 + }, + { + "epoch": 0.36, + "learning_rate": 4.409720441138754e-05, + "loss": 0.578, + "step": 9256 + }, + { + "epoch": 0.36, + "learning_rate": 4.40965632213388e-05, + "loss": 0.5728, + "step": 9257 + }, + { + "epoch": 0.36, + "learning_rate": 4.4095922031290074e-05, + "loss": 0.5988, + "step": 9258 + }, + { + "epoch": 0.36, + "learning_rate": 4.4095280841241346e-05, + "loss": 0.6971, + "step": 9259 + }, + { + "epoch": 0.36, + "learning_rate": 4.409463965119262e-05, + "loss": 0.5021, + "step": 9260 + }, + { + "epoch": 0.36, + "learning_rate": 4.409399846114389e-05, + "loss": 0.5845, + "step": 9261 + }, + { + "epoch": 0.36, + "learning_rate": 4.4093357271095154e-05, + "loss": 0.5933, + "step": 9262 + }, + { + "epoch": 0.36, + "learning_rate": 4.4092716081046426e-05, + "loss": 0.5751, + "step": 9263 + }, + { + "epoch": 0.36, + "learning_rate": 4.409207489099769e-05, + "loss": 0.608, + "step": 9264 + }, + { + "epoch": 0.36, + "learning_rate": 4.409143370094896e-05, + "loss": 0.4675, + "step": 9265 + }, + { + "epoch": 0.36, + "learning_rate": 4.409079251090023e-05, + "loss": 0.6036, + "step": 9266 + }, + { + "epoch": 0.36, + "learning_rate": 4.4090151320851506e-05, + "loss": 0.6011, + "step": 9267 + }, + { + "epoch": 0.36, + "learning_rate": 4.408951013080278e-05, + "loss": 0.531, + "step": 9268 + }, + { + "epoch": 0.36, + "learning_rate": 4.408886894075404e-05, + "loss": 0.652, + "step": 9269 + }, + { + "epoch": 0.36, + "learning_rate": 4.4088227750705314e-05, + "loss": 0.5358, + "step": 9270 + }, + { + "epoch": 0.36, + "learning_rate": 4.408758656065658e-05, + "loss": 0.5678, + "step": 9271 + }, + { + "epoch": 0.36, + "learning_rate": 4.408694537060785e-05, + "loss": 0.6678, + "step": 9272 + }, + { + "epoch": 0.36, + "learning_rate": 4.4086304180559116e-05, + "loss": 0.534, + "step": 9273 + }, + { + "epoch": 0.36, + "learning_rate": 4.408566299051039e-05, + "loss": 0.5792, + "step": 9274 + }, + { + "epoch": 0.36, + "learning_rate": 4.408502180046166e-05, + "loss": 0.6454, + "step": 9275 + }, + { + "epoch": 0.36, + "learning_rate": 4.408438061041293e-05, + "loss": 0.5707, + "step": 9276 + }, + { + "epoch": 0.36, + "learning_rate": 4.40837394203642e-05, + "loss": 0.5194, + "step": 9277 + }, + { + "epoch": 0.36, + "learning_rate": 4.408309823031547e-05, + "loss": 0.4847, + "step": 9278 + }, + { + "epoch": 0.36, + "learning_rate": 4.408245704026674e-05, + "loss": 0.5361, + "step": 9279 + }, + { + "epoch": 0.36, + "learning_rate": 4.4081815850218004e-05, + "loss": 0.5382, + "step": 9280 + }, + { + "epoch": 0.36, + "learning_rate": 4.4081174660169276e-05, + "loss": 0.635, + "step": 9281 + }, + { + "epoch": 0.36, + "learning_rate": 4.408053347012054e-05, + "loss": 0.5718, + "step": 9282 + }, + { + "epoch": 0.36, + "learning_rate": 4.407989228007182e-05, + "loss": 0.5809, + "step": 9283 + }, + { + "epoch": 0.36, + "learning_rate": 4.4079251090023084e-05, + "loss": 0.6252, + "step": 9284 + }, + { + "epoch": 0.36, + "learning_rate": 4.4078609899974356e-05, + "loss": 0.4559, + "step": 9285 + }, + { + "epoch": 0.36, + "learning_rate": 4.407796870992563e-05, + "loss": 0.7329, + "step": 9286 + }, + { + "epoch": 0.36, + "learning_rate": 4.407732751987689e-05, + "loss": 0.563, + "step": 9287 + }, + { + "epoch": 0.36, + "learning_rate": 4.4076686329828164e-05, + "loss": 0.5797, + "step": 9288 + }, + { + "epoch": 0.36, + "learning_rate": 4.407604513977943e-05, + "loss": 0.5879, + "step": 9289 + }, + { + "epoch": 0.36, + "learning_rate": 4.40754039497307e-05, + "loss": 0.5249, + "step": 9290 + }, + { + "epoch": 0.36, + "learning_rate": 4.407476275968197e-05, + "loss": 0.5868, + "step": 9291 + }, + { + "epoch": 0.36, + "learning_rate": 4.4074121569633245e-05, + "loss": 0.5162, + "step": 9292 + }, + { + "epoch": 0.36, + "learning_rate": 4.407348037958451e-05, + "loss": 0.5285, + "step": 9293 + }, + { + "epoch": 0.36, + "learning_rate": 4.407283918953578e-05, + "loss": 0.6131, + "step": 9294 + }, + { + "epoch": 0.36, + "learning_rate": 4.407219799948705e-05, + "loss": 0.6162, + "step": 9295 + }, + { + "epoch": 0.36, + "learning_rate": 4.407155680943832e-05, + "loss": 0.6831, + "step": 9296 + }, + { + "epoch": 0.36, + "learning_rate": 4.407091561938959e-05, + "loss": 0.6165, + "step": 9297 + }, + { + "epoch": 0.36, + "learning_rate": 4.4070274429340855e-05, + "loss": 0.5539, + "step": 9298 + }, + { + "epoch": 0.36, + "learning_rate": 4.4069633239292126e-05, + "loss": 0.6098, + "step": 9299 + }, + { + "epoch": 0.36, + "learning_rate": 4.40689920492434e-05, + "loss": 0.6522, + "step": 9300 + }, + { + "epoch": 0.36, + "learning_rate": 4.406835085919467e-05, + "loss": 0.5255, + "step": 9301 + }, + { + "epoch": 0.36, + "learning_rate": 4.4067709669145935e-05, + "loss": 0.6155, + "step": 9302 + }, + { + "epoch": 0.36, + "learning_rate": 4.4067068479097206e-05, + "loss": 0.5068, + "step": 9303 + }, + { + "epoch": 0.36, + "learning_rate": 4.406642728904848e-05, + "loss": 0.62, + "step": 9304 + }, + { + "epoch": 0.36, + "learning_rate": 4.406578609899974e-05, + "loss": 0.5457, + "step": 9305 + }, + { + "epoch": 0.36, + "learning_rate": 4.4065144908951015e-05, + "loss": 0.5366, + "step": 9306 + }, + { + "epoch": 0.36, + "learning_rate": 4.4064503718902286e-05, + "loss": 0.5811, + "step": 9307 + }, + { + "epoch": 0.36, + "learning_rate": 4.406386252885356e-05, + "loss": 0.5048, + "step": 9308 + }, + { + "epoch": 0.36, + "learning_rate": 4.406322133880482e-05, + "loss": 0.613, + "step": 9309 + }, + { + "epoch": 0.36, + "learning_rate": 4.4062580148756095e-05, + "loss": 0.5253, + "step": 9310 + }, + { + "epoch": 0.36, + "learning_rate": 4.406193895870736e-05, + "loss": 0.5669, + "step": 9311 + }, + { + "epoch": 0.36, + "learning_rate": 4.406129776865863e-05, + "loss": 0.5713, + "step": 9312 + }, + { + "epoch": 0.36, + "learning_rate": 4.40606565786099e-05, + "loss": 0.5417, + "step": 9313 + }, + { + "epoch": 0.36, + "learning_rate": 4.406001538856117e-05, + "loss": 0.5575, + "step": 9314 + }, + { + "epoch": 0.36, + "learning_rate": 4.405937419851244e-05, + "loss": 0.5065, + "step": 9315 + }, + { + "epoch": 0.36, + "learning_rate": 4.405873300846371e-05, + "loss": 0.5171, + "step": 9316 + }, + { + "epoch": 0.36, + "learning_rate": 4.405809181841498e-05, + "loss": 0.6333, + "step": 9317 + }, + { + "epoch": 0.36, + "learning_rate": 4.405745062836625e-05, + "loss": 0.6819, + "step": 9318 + }, + { + "epoch": 0.36, + "learning_rate": 4.405680943831752e-05, + "loss": 0.4751, + "step": 9319 + }, + { + "epoch": 0.36, + "learning_rate": 4.4056168248268785e-05, + "loss": 0.5264, + "step": 9320 + }, + { + "epoch": 0.36, + "learning_rate": 4.4055527058220057e-05, + "loss": 0.5423, + "step": 9321 + }, + { + "epoch": 0.36, + "learning_rate": 4.405488586817133e-05, + "loss": 0.5307, + "step": 9322 + }, + { + "epoch": 0.36, + "learning_rate": 4.40542446781226e-05, + "loss": 0.5527, + "step": 9323 + }, + { + "epoch": 0.36, + "learning_rate": 4.405360348807387e-05, + "loss": 0.5023, + "step": 9324 + }, + { + "epoch": 0.36, + "learning_rate": 4.405296229802514e-05, + "loss": 0.6149, + "step": 9325 + }, + { + "epoch": 0.36, + "learning_rate": 4.405232110797641e-05, + "loss": 0.4944, + "step": 9326 + }, + { + "epoch": 0.36, + "learning_rate": 4.405167991792767e-05, + "loss": 0.7054, + "step": 9327 + }, + { + "epoch": 0.36, + "learning_rate": 4.4051038727878945e-05, + "loss": 0.5857, + "step": 9328 + }, + { + "epoch": 0.36, + "learning_rate": 4.405039753783021e-05, + "loss": 0.6116, + "step": 9329 + }, + { + "epoch": 0.36, + "learning_rate": 4.404975634778148e-05, + "loss": 0.5057, + "step": 9330 + }, + { + "epoch": 0.36, + "learning_rate": 4.4049115157732753e-05, + "loss": 0.418, + "step": 9331 + }, + { + "epoch": 0.36, + "learning_rate": 4.4048473967684025e-05, + "loss": 0.4906, + "step": 9332 + }, + { + "epoch": 0.36, + "learning_rate": 4.40478327776353e-05, + "loss": 0.6364, + "step": 9333 + }, + { + "epoch": 0.36, + "learning_rate": 4.404719158758656e-05, + "loss": 0.517, + "step": 9334 + }, + { + "epoch": 0.36, + "learning_rate": 4.4046550397537834e-05, + "loss": 0.5238, + "step": 9335 + }, + { + "epoch": 0.36, + "learning_rate": 4.40459092074891e-05, + "loss": 0.6191, + "step": 9336 + }, + { + "epoch": 0.36, + "learning_rate": 4.404526801744037e-05, + "loss": 0.5255, + "step": 9337 + }, + { + "epoch": 0.36, + "learning_rate": 4.4044626827391635e-05, + "loss": 0.5625, + "step": 9338 + }, + { + "epoch": 0.36, + "learning_rate": 4.4043985637342914e-05, + "loss": 0.603, + "step": 9339 + }, + { + "epoch": 0.36, + "learning_rate": 4.4043344447294185e-05, + "loss": 0.6767, + "step": 9340 + }, + { + "epoch": 0.36, + "learning_rate": 4.404270325724545e-05, + "loss": 0.5752, + "step": 9341 + }, + { + "epoch": 0.36, + "learning_rate": 4.404206206719672e-05, + "loss": 0.5898, + "step": 9342 + }, + { + "epoch": 0.36, + "learning_rate": 4.404142087714799e-05, + "loss": 0.4674, + "step": 9343 + }, + { + "epoch": 0.36, + "learning_rate": 4.404077968709926e-05, + "loss": 0.5849, + "step": 9344 + }, + { + "epoch": 0.36, + "learning_rate": 4.4040138497050524e-05, + "loss": 0.5315, + "step": 9345 + }, + { + "epoch": 0.36, + "learning_rate": 4.4039497307001795e-05, + "loss": 0.5977, + "step": 9346 + }, + { + "epoch": 0.36, + "learning_rate": 4.403885611695307e-05, + "loss": 0.5368, + "step": 9347 + }, + { + "epoch": 0.36, + "learning_rate": 4.403821492690434e-05, + "loss": 0.5228, + "step": 9348 + }, + { + "epoch": 0.36, + "learning_rate": 4.403757373685561e-05, + "loss": 0.5726, + "step": 9349 + }, + { + "epoch": 0.36, + "learning_rate": 4.4036932546806875e-05, + "loss": 0.5876, + "step": 9350 + }, + { + "epoch": 0.36, + "learning_rate": 4.403629135675815e-05, + "loss": 0.4469, + "step": 9351 + }, + { + "epoch": 0.36, + "learning_rate": 4.403565016670941e-05, + "loss": 0.5683, + "step": 9352 + }, + { + "epoch": 0.36, + "learning_rate": 4.4035008976660684e-05, + "loss": 0.5117, + "step": 9353 + }, + { + "epoch": 0.36, + "learning_rate": 4.403436778661195e-05, + "loss": 0.6131, + "step": 9354 + }, + { + "epoch": 0.36, + "learning_rate": 4.403372659656323e-05, + "loss": 0.5453, + "step": 9355 + }, + { + "epoch": 0.36, + "learning_rate": 4.403308540651449e-05, + "loss": 0.5063, + "step": 9356 + }, + { + "epoch": 0.36, + "learning_rate": 4.4032444216465764e-05, + "loss": 0.5377, + "step": 9357 + }, + { + "epoch": 0.36, + "learning_rate": 4.4031803026417036e-05, + "loss": 0.7024, + "step": 9358 + }, + { + "epoch": 0.36, + "learning_rate": 4.40311618363683e-05, + "loss": 0.5592, + "step": 9359 + }, + { + "epoch": 0.36, + "learning_rate": 4.403052064631957e-05, + "loss": 0.5762, + "step": 9360 + }, + { + "epoch": 0.36, + "learning_rate": 4.402987945627084e-05, + "loss": 0.6076, + "step": 9361 + }, + { + "epoch": 0.36, + "learning_rate": 4.402923826622211e-05, + "loss": 0.5793, + "step": 9362 + }, + { + "epoch": 0.36, + "learning_rate": 4.402859707617338e-05, + "loss": 0.5247, + "step": 9363 + }, + { + "epoch": 0.36, + "learning_rate": 4.402795588612465e-05, + "loss": 0.5346, + "step": 9364 + }, + { + "epoch": 0.36, + "learning_rate": 4.402731469607592e-05, + "loss": 0.5375, + "step": 9365 + }, + { + "epoch": 0.36, + "learning_rate": 4.402667350602719e-05, + "loss": 0.5949, + "step": 9366 + }, + { + "epoch": 0.36, + "learning_rate": 4.402603231597846e-05, + "loss": 0.5478, + "step": 9367 + }, + { + "epoch": 0.36, + "learning_rate": 4.4025391125929726e-05, + "loss": 0.6779, + "step": 9368 + }, + { + "epoch": 0.36, + "learning_rate": 4.4024749935881e-05, + "loss": 0.5343, + "step": 9369 + }, + { + "epoch": 0.36, + "learning_rate": 4.402410874583226e-05, + "loss": 0.6185, + "step": 9370 + }, + { + "epoch": 0.36, + "learning_rate": 4.402346755578354e-05, + "loss": 0.5288, + "step": 9371 + }, + { + "epoch": 0.36, + "learning_rate": 4.4022826365734806e-05, + "loss": 0.4614, + "step": 9372 + }, + { + "epoch": 0.36, + "learning_rate": 4.402218517568608e-05, + "loss": 0.5696, + "step": 9373 + }, + { + "epoch": 0.36, + "learning_rate": 4.402154398563734e-05, + "loss": 0.6105, + "step": 9374 + }, + { + "epoch": 0.36, + "learning_rate": 4.4020902795588614e-05, + "loss": 0.6731, + "step": 9375 + }, + { + "epoch": 0.36, + "learning_rate": 4.4020261605539886e-05, + "loss": 0.4831, + "step": 9376 + }, + { + "epoch": 0.36, + "learning_rate": 4.401962041549115e-05, + "loss": 0.5793, + "step": 9377 + }, + { + "epoch": 0.36, + "learning_rate": 4.401897922544242e-05, + "loss": 0.5056, + "step": 9378 + }, + { + "epoch": 0.36, + "learning_rate": 4.4018338035393694e-05, + "loss": 0.5847, + "step": 9379 + }, + { + "epoch": 0.36, + "learning_rate": 4.4017696845344966e-05, + "loss": 0.4507, + "step": 9380 + }, + { + "epoch": 0.36, + "learning_rate": 4.401705565529623e-05, + "loss": 0.5189, + "step": 9381 + }, + { + "epoch": 0.36, + "learning_rate": 4.40164144652475e-05, + "loss": 0.5944, + "step": 9382 + }, + { + "epoch": 0.36, + "learning_rate": 4.401577327519877e-05, + "loss": 0.6385, + "step": 9383 + }, + { + "epoch": 0.36, + "learning_rate": 4.401513208515004e-05, + "loss": 0.4986, + "step": 9384 + }, + { + "epoch": 0.36, + "learning_rate": 4.401449089510131e-05, + "loss": 0.591, + "step": 9385 + }, + { + "epoch": 0.36, + "learning_rate": 4.4013849705052576e-05, + "loss": 0.5928, + "step": 9386 + }, + { + "epoch": 0.36, + "learning_rate": 4.401320851500385e-05, + "loss": 0.4893, + "step": 9387 + }, + { + "epoch": 0.36, + "learning_rate": 4.401256732495512e-05, + "loss": 0.6728, + "step": 9388 + }, + { + "epoch": 0.36, + "learning_rate": 4.401192613490639e-05, + "loss": 0.5537, + "step": 9389 + }, + { + "epoch": 0.36, + "learning_rate": 4.4011284944857656e-05, + "loss": 0.552, + "step": 9390 + }, + { + "epoch": 0.36, + "learning_rate": 4.401064375480893e-05, + "loss": 0.5846, + "step": 9391 + }, + { + "epoch": 0.36, + "learning_rate": 4.401000256476019e-05, + "loss": 0.4602, + "step": 9392 + }, + { + "epoch": 0.36, + "learning_rate": 4.4009361374711464e-05, + "loss": 0.6659, + "step": 9393 + }, + { + "epoch": 0.36, + "learning_rate": 4.4008720184662736e-05, + "loss": 0.6435, + "step": 9394 + }, + { + "epoch": 0.36, + "learning_rate": 4.400807899461401e-05, + "loss": 0.6234, + "step": 9395 + }, + { + "epoch": 0.36, + "learning_rate": 4.400743780456528e-05, + "loss": 0.6965, + "step": 9396 + }, + { + "epoch": 0.36, + "learning_rate": 4.4006796614516544e-05, + "loss": 0.508, + "step": 9397 + }, + { + "epoch": 0.36, + "learning_rate": 4.4006155424467816e-05, + "loss": 0.4922, + "step": 9398 + }, + { + "epoch": 0.36, + "learning_rate": 4.400551423441908e-05, + "loss": 0.5737, + "step": 9399 + }, + { + "epoch": 0.36, + "learning_rate": 4.400487304437035e-05, + "loss": 0.5543, + "step": 9400 + }, + { + "epoch": 0.36, + "learning_rate": 4.400423185432162e-05, + "loss": 0.5388, + "step": 9401 + }, + { + "epoch": 0.36, + "learning_rate": 4.400359066427289e-05, + "loss": 0.6197, + "step": 9402 + }, + { + "epoch": 0.36, + "learning_rate": 4.400294947422416e-05, + "loss": 0.6082, + "step": 9403 + }, + { + "epoch": 0.36, + "learning_rate": 4.400230828417543e-05, + "loss": 0.4548, + "step": 9404 + }, + { + "epoch": 0.36, + "learning_rate": 4.4001667094126705e-05, + "loss": 0.622, + "step": 9405 + }, + { + "epoch": 0.36, + "learning_rate": 4.400102590407797e-05, + "loss": 0.5456, + "step": 9406 + }, + { + "epoch": 0.36, + "learning_rate": 4.400038471402924e-05, + "loss": 0.5597, + "step": 9407 + }, + { + "epoch": 0.36, + "learning_rate": 4.3999743523980506e-05, + "loss": 0.5949, + "step": 9408 + }, + { + "epoch": 0.36, + "learning_rate": 4.399910233393178e-05, + "loss": 0.5443, + "step": 9409 + }, + { + "epoch": 0.36, + "learning_rate": 4.399846114388305e-05, + "loss": 0.568, + "step": 9410 + }, + { + "epoch": 0.36, + "learning_rate": 4.399781995383432e-05, + "loss": 0.5472, + "step": 9411 + }, + { + "epoch": 0.36, + "learning_rate": 4.399717876378559e-05, + "loss": 0.6077, + "step": 9412 + }, + { + "epoch": 0.36, + "learning_rate": 4.399653757373686e-05, + "loss": 0.6171, + "step": 9413 + }, + { + "epoch": 0.36, + "learning_rate": 4.399589638368813e-05, + "loss": 0.5865, + "step": 9414 + }, + { + "epoch": 0.36, + "learning_rate": 4.3995255193639395e-05, + "loss": 0.5566, + "step": 9415 + }, + { + "epoch": 0.36, + "learning_rate": 4.3994614003590666e-05, + "loss": 0.6651, + "step": 9416 + }, + { + "epoch": 0.36, + "learning_rate": 4.399397281354193e-05, + "loss": 0.5921, + "step": 9417 + }, + { + "epoch": 0.36, + "learning_rate": 4.39933316234932e-05, + "loss": 0.6212, + "step": 9418 + }, + { + "epoch": 0.36, + "learning_rate": 4.3992690433444475e-05, + "loss": 0.5679, + "step": 9419 + }, + { + "epoch": 0.36, + "learning_rate": 4.3992049243395746e-05, + "loss": 0.529, + "step": 9420 + }, + { + "epoch": 0.36, + "learning_rate": 4.399140805334702e-05, + "loss": 0.6165, + "step": 9421 + }, + { + "epoch": 0.36, + "learning_rate": 4.399076686329828e-05, + "loss": 0.5766, + "step": 9422 + }, + { + "epoch": 0.36, + "learning_rate": 4.3990125673249555e-05, + "loss": 0.6037, + "step": 9423 + }, + { + "epoch": 0.36, + "learning_rate": 4.398948448320082e-05, + "loss": 0.5947, + "step": 9424 + }, + { + "epoch": 0.36, + "learning_rate": 4.398884329315209e-05, + "loss": 0.5919, + "step": 9425 + }, + { + "epoch": 0.36, + "learning_rate": 4.3988202103103356e-05, + "loss": 0.6319, + "step": 9426 + }, + { + "epoch": 0.36, + "learning_rate": 4.3987560913054635e-05, + "loss": 0.5629, + "step": 9427 + }, + { + "epoch": 0.36, + "learning_rate": 4.398691972300591e-05, + "loss": 0.6236, + "step": 9428 + }, + { + "epoch": 0.36, + "learning_rate": 4.398627853295717e-05, + "loss": 0.5578, + "step": 9429 + }, + { + "epoch": 0.36, + "learning_rate": 4.398563734290844e-05, + "loss": 0.6084, + "step": 9430 + }, + { + "epoch": 0.36, + "learning_rate": 4.398499615285971e-05, + "loss": 0.4476, + "step": 9431 + }, + { + "epoch": 0.36, + "learning_rate": 4.398435496281098e-05, + "loss": 0.6575, + "step": 9432 + }, + { + "epoch": 0.36, + "learning_rate": 4.3983713772762245e-05, + "loss": 0.4811, + "step": 9433 + }, + { + "epoch": 0.36, + "learning_rate": 4.3983072582713517e-05, + "loss": 0.5643, + "step": 9434 + }, + { + "epoch": 0.36, + "learning_rate": 4.398243139266479e-05, + "loss": 0.5833, + "step": 9435 + }, + { + "epoch": 0.36, + "learning_rate": 4.398179020261606e-05, + "loss": 0.5701, + "step": 9436 + }, + { + "epoch": 0.36, + "learning_rate": 4.398114901256733e-05, + "loss": 0.5758, + "step": 9437 + }, + { + "epoch": 0.36, + "learning_rate": 4.39805078225186e-05, + "loss": 0.6513, + "step": 9438 + }, + { + "epoch": 0.36, + "learning_rate": 4.397986663246987e-05, + "loss": 0.5702, + "step": 9439 + }, + { + "epoch": 0.36, + "learning_rate": 4.397922544242113e-05, + "loss": 0.5781, + "step": 9440 + }, + { + "epoch": 0.36, + "learning_rate": 4.3978584252372405e-05, + "loss": 0.5287, + "step": 9441 + }, + { + "epoch": 0.36, + "learning_rate": 4.397794306232367e-05, + "loss": 0.4945, + "step": 9442 + }, + { + "epoch": 0.36, + "learning_rate": 4.397730187227495e-05, + "loss": 0.608, + "step": 9443 + }, + { + "epoch": 0.36, + "learning_rate": 4.3976660682226213e-05, + "loss": 0.6789, + "step": 9444 + }, + { + "epoch": 0.36, + "learning_rate": 4.3976019492177485e-05, + "loss": 0.4617, + "step": 9445 + }, + { + "epoch": 0.36, + "learning_rate": 4.397537830212876e-05, + "loss": 0.5639, + "step": 9446 + }, + { + "epoch": 0.36, + "learning_rate": 4.397473711208002e-05, + "loss": 0.52, + "step": 9447 + }, + { + "epoch": 0.36, + "learning_rate": 4.3974095922031294e-05, + "loss": 0.548, + "step": 9448 + }, + { + "epoch": 0.36, + "learning_rate": 4.397345473198256e-05, + "loss": 0.5793, + "step": 9449 + }, + { + "epoch": 0.36, + "learning_rate": 4.397281354193383e-05, + "loss": 0.6878, + "step": 9450 + }, + { + "epoch": 0.36, + "learning_rate": 4.39721723518851e-05, + "loss": 0.4981, + "step": 9451 + }, + { + "epoch": 0.36, + "learning_rate": 4.3971531161836374e-05, + "loss": 0.6311, + "step": 9452 + }, + { + "epoch": 0.36, + "learning_rate": 4.397088997178764e-05, + "loss": 0.5654, + "step": 9453 + }, + { + "epoch": 0.36, + "learning_rate": 4.397024878173891e-05, + "loss": 0.6053, + "step": 9454 + }, + { + "epoch": 0.36, + "learning_rate": 4.396960759169018e-05, + "loss": 0.5366, + "step": 9455 + }, + { + "epoch": 0.36, + "learning_rate": 4.396896640164145e-05, + "loss": 0.6149, + "step": 9456 + }, + { + "epoch": 0.36, + "learning_rate": 4.396832521159272e-05, + "loss": 0.5851, + "step": 9457 + }, + { + "epoch": 0.36, + "learning_rate": 4.3967684021543984e-05, + "loss": 0.5722, + "step": 9458 + }, + { + "epoch": 0.36, + "learning_rate": 4.396704283149526e-05, + "loss": 0.5935, + "step": 9459 + }, + { + "epoch": 0.36, + "learning_rate": 4.396640164144653e-05, + "loss": 0.5347, + "step": 9460 + }, + { + "epoch": 0.36, + "learning_rate": 4.39657604513978e-05, + "loss": 0.5762, + "step": 9461 + }, + { + "epoch": 0.36, + "learning_rate": 4.3965119261349064e-05, + "loss": 0.4949, + "step": 9462 + }, + { + "epoch": 0.36, + "learning_rate": 4.3964478071300335e-05, + "loss": 0.5862, + "step": 9463 + }, + { + "epoch": 0.36, + "learning_rate": 4.396383688125161e-05, + "loss": 0.558, + "step": 9464 + }, + { + "epoch": 0.36, + "learning_rate": 4.396319569120287e-05, + "loss": 0.5964, + "step": 9465 + }, + { + "epoch": 0.36, + "learning_rate": 4.3962554501154144e-05, + "loss": 0.5496, + "step": 9466 + }, + { + "epoch": 0.36, + "learning_rate": 4.3961913311105415e-05, + "loss": 0.5065, + "step": 9467 + }, + { + "epoch": 0.36, + "learning_rate": 4.396127212105669e-05, + "loss": 0.5299, + "step": 9468 + }, + { + "epoch": 0.36, + "learning_rate": 4.396063093100795e-05, + "loss": 0.6382, + "step": 9469 + }, + { + "epoch": 0.36, + "learning_rate": 4.3959989740959224e-05, + "loss": 0.5174, + "step": 9470 + }, + { + "epoch": 0.36, + "learning_rate": 4.395934855091049e-05, + "loss": 0.6052, + "step": 9471 + }, + { + "epoch": 0.36, + "learning_rate": 4.395870736086176e-05, + "loss": 0.5658, + "step": 9472 + }, + { + "epoch": 0.36, + "learning_rate": 4.395806617081303e-05, + "loss": 0.553, + "step": 9473 + }, + { + "epoch": 0.36, + "learning_rate": 4.39574249807643e-05, + "loss": 0.6136, + "step": 9474 + }, + { + "epoch": 0.36, + "learning_rate": 4.395678379071557e-05, + "loss": 0.6178, + "step": 9475 + }, + { + "epoch": 0.36, + "learning_rate": 4.395614260066684e-05, + "loss": 0.6735, + "step": 9476 + }, + { + "epoch": 0.36, + "learning_rate": 4.395550141061811e-05, + "loss": 0.502, + "step": 9477 + }, + { + "epoch": 0.36, + "learning_rate": 4.395486022056938e-05, + "loss": 0.5001, + "step": 9478 + }, + { + "epoch": 0.36, + "learning_rate": 4.395421903052065e-05, + "loss": 0.6693, + "step": 9479 + }, + { + "epoch": 0.36, + "learning_rate": 4.3953577840471914e-05, + "loss": 0.5865, + "step": 9480 + }, + { + "epoch": 0.36, + "learning_rate": 4.3952936650423186e-05, + "loss": 0.4968, + "step": 9481 + }, + { + "epoch": 0.36, + "learning_rate": 4.395229546037446e-05, + "loss": 0.5903, + "step": 9482 + }, + { + "epoch": 0.36, + "learning_rate": 4.395165427032573e-05, + "loss": 0.6499, + "step": 9483 + }, + { + "epoch": 0.36, + "learning_rate": 4.3951013080277e-05, + "loss": 0.6203, + "step": 9484 + }, + { + "epoch": 0.36, + "learning_rate": 4.3950371890228266e-05, + "loss": 0.5485, + "step": 9485 + }, + { + "epoch": 0.36, + "learning_rate": 4.394973070017954e-05, + "loss": 0.6827, + "step": 9486 + }, + { + "epoch": 0.36, + "learning_rate": 4.39490895101308e-05, + "loss": 0.5493, + "step": 9487 + }, + { + "epoch": 0.36, + "learning_rate": 4.3948448320082074e-05, + "loss": 0.5038, + "step": 9488 + }, + { + "epoch": 0.36, + "learning_rate": 4.394780713003334e-05, + "loss": 0.497, + "step": 9489 + }, + { + "epoch": 0.36, + "learning_rate": 4.394716593998461e-05, + "loss": 0.5672, + "step": 9490 + }, + { + "epoch": 0.36, + "learning_rate": 4.394652474993588e-05, + "loss": 0.5414, + "step": 9491 + }, + { + "epoch": 0.36, + "learning_rate": 4.3945883559887154e-05, + "loss": 0.6298, + "step": 9492 + }, + { + "epoch": 0.36, + "learning_rate": 4.3945242369838426e-05, + "loss": 0.5477, + "step": 9493 + }, + { + "epoch": 0.37, + "learning_rate": 4.394460117978969e-05, + "loss": 0.5739, + "step": 9494 + }, + { + "epoch": 0.37, + "learning_rate": 4.394395998974096e-05, + "loss": 0.5521, + "step": 9495 + }, + { + "epoch": 0.37, + "learning_rate": 4.394331879969223e-05, + "loss": 0.5899, + "step": 9496 + }, + { + "epoch": 0.37, + "learning_rate": 4.39426776096435e-05, + "loss": 0.7203, + "step": 9497 + }, + { + "epoch": 0.37, + "learning_rate": 4.394203641959477e-05, + "loss": 0.575, + "step": 9498 + }, + { + "epoch": 0.37, + "learning_rate": 4.394139522954604e-05, + "loss": 0.6602, + "step": 9499 + }, + { + "epoch": 0.37, + "learning_rate": 4.3940754039497314e-05, + "loss": 0.4822, + "step": 9500 + }, + { + "epoch": 0.37, + "learning_rate": 4.394011284944858e-05, + "loss": 0.5366, + "step": 9501 + }, + { + "epoch": 0.37, + "learning_rate": 4.393947165939985e-05, + "loss": 0.5116, + "step": 9502 + }, + { + "epoch": 0.37, + "learning_rate": 4.3938830469351116e-05, + "loss": 0.509, + "step": 9503 + }, + { + "epoch": 0.37, + "learning_rate": 4.393818927930239e-05, + "loss": 0.5768, + "step": 9504 + }, + { + "epoch": 0.37, + "learning_rate": 4.393754808925365e-05, + "loss": 0.5454, + "step": 9505 + }, + { + "epoch": 0.37, + "learning_rate": 4.3936906899204924e-05, + "loss": 0.6316, + "step": 9506 + }, + { + "epoch": 0.37, + "learning_rate": 4.3936265709156196e-05, + "loss": 0.5434, + "step": 9507 + }, + { + "epoch": 0.37, + "learning_rate": 4.393562451910747e-05, + "loss": 0.5579, + "step": 9508 + }, + { + "epoch": 0.37, + "learning_rate": 4.393498332905874e-05, + "loss": 0.5199, + "step": 9509 + }, + { + "epoch": 0.37, + "learning_rate": 4.3934342139010004e-05, + "loss": 0.6068, + "step": 9510 + }, + { + "epoch": 0.37, + "learning_rate": 4.3933700948961276e-05, + "loss": 0.502, + "step": 9511 + }, + { + "epoch": 0.37, + "learning_rate": 4.393305975891254e-05, + "loss": 0.6169, + "step": 9512 + }, + { + "epoch": 0.37, + "learning_rate": 4.393241856886381e-05, + "loss": 0.6402, + "step": 9513 + }, + { + "epoch": 0.37, + "learning_rate": 4.393177737881508e-05, + "loss": 0.5549, + "step": 9514 + }, + { + "epoch": 0.37, + "learning_rate": 4.3931136188766356e-05, + "loss": 0.5513, + "step": 9515 + }, + { + "epoch": 0.37, + "learning_rate": 4.393049499871762e-05, + "loss": 0.6416, + "step": 9516 + }, + { + "epoch": 0.37, + "learning_rate": 4.392985380866889e-05, + "loss": 0.5659, + "step": 9517 + }, + { + "epoch": 0.37, + "learning_rate": 4.3929212618620165e-05, + "loss": 0.6124, + "step": 9518 + }, + { + "epoch": 0.37, + "learning_rate": 4.392857142857143e-05, + "loss": 0.5883, + "step": 9519 + }, + { + "epoch": 0.37, + "learning_rate": 4.39279302385227e-05, + "loss": 0.5319, + "step": 9520 + }, + { + "epoch": 0.37, + "learning_rate": 4.3927289048473966e-05, + "loss": 0.562, + "step": 9521 + }, + { + "epoch": 0.37, + "learning_rate": 4.392664785842524e-05, + "loss": 0.5222, + "step": 9522 + }, + { + "epoch": 0.37, + "learning_rate": 4.392600666837651e-05, + "loss": 0.5152, + "step": 9523 + }, + { + "epoch": 0.37, + "learning_rate": 4.392536547832778e-05, + "loss": 0.525, + "step": 9524 + }, + { + "epoch": 0.37, + "learning_rate": 4.3924724288279046e-05, + "loss": 0.5466, + "step": 9525 + }, + { + "epoch": 0.37, + "learning_rate": 4.392408309823032e-05, + "loss": 0.6033, + "step": 9526 + }, + { + "epoch": 0.37, + "learning_rate": 4.392344190818159e-05, + "loss": 0.5665, + "step": 9527 + }, + { + "epoch": 0.37, + "learning_rate": 4.3922800718132855e-05, + "loss": 0.5787, + "step": 9528 + }, + { + "epoch": 0.37, + "learning_rate": 4.3922159528084126e-05, + "loss": 0.6386, + "step": 9529 + }, + { + "epoch": 0.37, + "learning_rate": 4.392151833803539e-05, + "loss": 0.4473, + "step": 9530 + }, + { + "epoch": 0.37, + "learning_rate": 4.392087714798667e-05, + "loss": 0.4643, + "step": 9531 + }, + { + "epoch": 0.37, + "learning_rate": 4.3920235957937935e-05, + "loss": 0.5677, + "step": 9532 + }, + { + "epoch": 0.37, + "learning_rate": 4.3919594767889206e-05, + "loss": 0.5942, + "step": 9533 + }, + { + "epoch": 0.37, + "learning_rate": 4.391895357784047e-05, + "loss": 0.5989, + "step": 9534 + }, + { + "epoch": 0.37, + "learning_rate": 4.391831238779174e-05, + "loss": 0.5575, + "step": 9535 + }, + { + "epoch": 0.37, + "learning_rate": 4.3917671197743015e-05, + "loss": 0.54, + "step": 9536 + }, + { + "epoch": 0.37, + "learning_rate": 4.391703000769428e-05, + "loss": 0.5883, + "step": 9537 + }, + { + "epoch": 0.37, + "learning_rate": 4.391638881764555e-05, + "loss": 0.5484, + "step": 9538 + }, + { + "epoch": 0.37, + "learning_rate": 4.391574762759682e-05, + "loss": 0.4656, + "step": 9539 + }, + { + "epoch": 0.37, + "learning_rate": 4.3915106437548095e-05, + "loss": 0.5898, + "step": 9540 + }, + { + "epoch": 0.37, + "learning_rate": 4.391446524749936e-05, + "loss": 0.5296, + "step": 9541 + }, + { + "epoch": 0.37, + "learning_rate": 4.391382405745063e-05, + "loss": 0.4358, + "step": 9542 + }, + { + "epoch": 0.37, + "learning_rate": 4.3913182867401897e-05, + "loss": 0.6538, + "step": 9543 + }, + { + "epoch": 0.37, + "learning_rate": 4.391254167735317e-05, + "loss": 0.4963, + "step": 9544 + }, + { + "epoch": 0.37, + "learning_rate": 4.391190048730444e-05, + "loss": 0.5574, + "step": 9545 + }, + { + "epoch": 0.37, + "learning_rate": 4.3911259297255705e-05, + "loss": 0.5227, + "step": 9546 + }, + { + "epoch": 0.37, + "learning_rate": 4.391061810720698e-05, + "loss": 0.5575, + "step": 9547 + }, + { + "epoch": 0.37, + "learning_rate": 4.390997691715825e-05, + "loss": 0.6103, + "step": 9548 + }, + { + "epoch": 0.37, + "learning_rate": 4.390933572710952e-05, + "loss": 0.5216, + "step": 9549 + }, + { + "epoch": 0.37, + "learning_rate": 4.3908694537060785e-05, + "loss": 0.5667, + "step": 9550 + }, + { + "epoch": 0.37, + "learning_rate": 4.390805334701206e-05, + "loss": 0.4168, + "step": 9551 + }, + { + "epoch": 0.37, + "learning_rate": 4.390741215696332e-05, + "loss": 0.5973, + "step": 9552 + }, + { + "epoch": 0.37, + "learning_rate": 4.3906770966914593e-05, + "loss": 0.5961, + "step": 9553 + }, + { + "epoch": 0.37, + "learning_rate": 4.3906129776865865e-05, + "loss": 0.5796, + "step": 9554 + }, + { + "epoch": 0.37, + "learning_rate": 4.390548858681714e-05, + "loss": 0.5255, + "step": 9555 + }, + { + "epoch": 0.37, + "learning_rate": 4.390484739676841e-05, + "loss": 0.6409, + "step": 9556 + }, + { + "epoch": 0.37, + "learning_rate": 4.3904206206719673e-05, + "loss": 0.6189, + "step": 9557 + }, + { + "epoch": 0.37, + "learning_rate": 4.3903565016670945e-05, + "loss": 0.5723, + "step": 9558 + }, + { + "epoch": 0.37, + "learning_rate": 4.390292382662221e-05, + "loss": 0.4899, + "step": 9559 + }, + { + "epoch": 0.37, + "learning_rate": 4.390228263657348e-05, + "loss": 0.5708, + "step": 9560 + }, + { + "epoch": 0.37, + "learning_rate": 4.390164144652475e-05, + "loss": 0.5407, + "step": 9561 + }, + { + "epoch": 0.37, + "learning_rate": 4.390100025647602e-05, + "loss": 0.5669, + "step": 9562 + }, + { + "epoch": 0.37, + "learning_rate": 4.390035906642729e-05, + "loss": 0.4627, + "step": 9563 + }, + { + "epoch": 0.37, + "learning_rate": 4.389971787637856e-05, + "loss": 0.5477, + "step": 9564 + }, + { + "epoch": 0.37, + "learning_rate": 4.3899076686329834e-05, + "loss": 0.6117, + "step": 9565 + }, + { + "epoch": 0.37, + "learning_rate": 4.38984354962811e-05, + "loss": 0.5721, + "step": 9566 + }, + { + "epoch": 0.37, + "learning_rate": 4.389779430623237e-05, + "loss": 0.4952, + "step": 9567 + }, + { + "epoch": 0.37, + "learning_rate": 4.3897153116183635e-05, + "loss": 0.5608, + "step": 9568 + }, + { + "epoch": 0.37, + "learning_rate": 4.389651192613491e-05, + "loss": 0.5282, + "step": 9569 + }, + { + "epoch": 0.37, + "learning_rate": 4.389587073608618e-05, + "loss": 0.6349, + "step": 9570 + }, + { + "epoch": 0.37, + "learning_rate": 4.389522954603745e-05, + "loss": 0.505, + "step": 9571 + }, + { + "epoch": 0.37, + "learning_rate": 4.389458835598872e-05, + "loss": 0.5481, + "step": 9572 + }, + { + "epoch": 0.37, + "learning_rate": 4.389394716593999e-05, + "loss": 0.5361, + "step": 9573 + }, + { + "epoch": 0.37, + "learning_rate": 4.389330597589126e-05, + "loss": 0.6887, + "step": 9574 + }, + { + "epoch": 0.37, + "learning_rate": 4.3892664785842524e-05, + "loss": 0.6089, + "step": 9575 + }, + { + "epoch": 0.37, + "learning_rate": 4.3892023595793795e-05, + "loss": 0.5113, + "step": 9576 + }, + { + "epoch": 0.37, + "learning_rate": 4.389138240574506e-05, + "loss": 0.5813, + "step": 9577 + }, + { + "epoch": 0.37, + "learning_rate": 4.389074121569633e-05, + "loss": 0.6779, + "step": 9578 + }, + { + "epoch": 0.37, + "learning_rate": 4.3890100025647604e-05, + "loss": 0.5443, + "step": 9579 + }, + { + "epoch": 0.37, + "learning_rate": 4.3889458835598876e-05, + "loss": 0.5105, + "step": 9580 + }, + { + "epoch": 0.37, + "learning_rate": 4.388881764555015e-05, + "loss": 0.5524, + "step": 9581 + }, + { + "epoch": 0.37, + "learning_rate": 4.388817645550141e-05, + "loss": 0.4871, + "step": 9582 + }, + { + "epoch": 0.37, + "learning_rate": 4.3887535265452684e-05, + "loss": 0.5757, + "step": 9583 + }, + { + "epoch": 0.37, + "learning_rate": 4.388689407540395e-05, + "loss": 0.6287, + "step": 9584 + }, + { + "epoch": 0.37, + "learning_rate": 4.388625288535522e-05, + "loss": 0.5703, + "step": 9585 + }, + { + "epoch": 0.37, + "learning_rate": 4.3885611695306485e-05, + "loss": 0.5822, + "step": 9586 + }, + { + "epoch": 0.37, + "learning_rate": 4.3884970505257764e-05, + "loss": 0.6639, + "step": 9587 + }, + { + "epoch": 0.37, + "learning_rate": 4.388432931520903e-05, + "loss": 0.5129, + "step": 9588 + }, + { + "epoch": 0.37, + "learning_rate": 4.38836881251603e-05, + "loss": 0.5573, + "step": 9589 + }, + { + "epoch": 0.37, + "learning_rate": 4.388304693511157e-05, + "loss": 0.5846, + "step": 9590 + }, + { + "epoch": 0.37, + "learning_rate": 4.388240574506284e-05, + "loss": 0.7084, + "step": 9591 + }, + { + "epoch": 0.37, + "learning_rate": 4.388176455501411e-05, + "loss": 0.5206, + "step": 9592 + }, + { + "epoch": 0.37, + "learning_rate": 4.3881123364965374e-05, + "loss": 0.5858, + "step": 9593 + }, + { + "epoch": 0.37, + "learning_rate": 4.3880482174916646e-05, + "loss": 0.5644, + "step": 9594 + }, + { + "epoch": 0.37, + "learning_rate": 4.387984098486792e-05, + "loss": 0.4757, + "step": 9595 + }, + { + "epoch": 0.37, + "learning_rate": 4.387919979481919e-05, + "loss": 0.596, + "step": 9596 + }, + { + "epoch": 0.37, + "learning_rate": 4.3878558604770454e-05, + "loss": 0.5237, + "step": 9597 + }, + { + "epoch": 0.37, + "learning_rate": 4.3877917414721726e-05, + "loss": 0.5381, + "step": 9598 + }, + { + "epoch": 0.37, + "learning_rate": 4.3877276224673e-05, + "loss": 0.6628, + "step": 9599 + }, + { + "epoch": 0.37, + "learning_rate": 4.387663503462426e-05, + "loss": 0.5813, + "step": 9600 + }, + { + "epoch": 0.37, + "learning_rate": 4.3875993844575534e-05, + "loss": 0.468, + "step": 9601 + }, + { + "epoch": 0.37, + "learning_rate": 4.38753526545268e-05, + "loss": 0.625, + "step": 9602 + }, + { + "epoch": 0.37, + "learning_rate": 4.387471146447808e-05, + "loss": 0.5073, + "step": 9603 + }, + { + "epoch": 0.37, + "learning_rate": 4.387407027442934e-05, + "loss": 0.5163, + "step": 9604 + }, + { + "epoch": 0.37, + "learning_rate": 4.3873429084380614e-05, + "loss": 0.6419, + "step": 9605 + }, + { + "epoch": 0.37, + "learning_rate": 4.387278789433188e-05, + "loss": 0.5155, + "step": 9606 + }, + { + "epoch": 0.37, + "learning_rate": 4.387214670428315e-05, + "loss": 0.66, + "step": 9607 + }, + { + "epoch": 0.37, + "learning_rate": 4.387150551423442e-05, + "loss": 0.5532, + "step": 9608 + }, + { + "epoch": 0.37, + "learning_rate": 4.387086432418569e-05, + "loss": 0.6075, + "step": 9609 + }, + { + "epoch": 0.37, + "learning_rate": 4.387022313413696e-05, + "loss": 0.5425, + "step": 9610 + }, + { + "epoch": 0.37, + "learning_rate": 4.386958194408823e-05, + "loss": 0.5396, + "step": 9611 + }, + { + "epoch": 0.37, + "learning_rate": 4.38689407540395e-05, + "loss": 0.6039, + "step": 9612 + }, + { + "epoch": 0.37, + "learning_rate": 4.386829956399077e-05, + "loss": 0.4884, + "step": 9613 + }, + { + "epoch": 0.37, + "learning_rate": 4.386765837394204e-05, + "loss": 0.5086, + "step": 9614 + }, + { + "epoch": 0.37, + "learning_rate": 4.386701718389331e-05, + "loss": 0.4358, + "step": 9615 + }, + { + "epoch": 0.37, + "learning_rate": 4.3866375993844576e-05, + "loss": 0.5569, + "step": 9616 + }, + { + "epoch": 0.37, + "learning_rate": 4.386573480379585e-05, + "loss": 0.6097, + "step": 9617 + }, + { + "epoch": 0.37, + "learning_rate": 4.386509361374711e-05, + "loss": 0.6122, + "step": 9618 + }, + { + "epoch": 0.37, + "learning_rate": 4.386445242369839e-05, + "loss": 0.4538, + "step": 9619 + }, + { + "epoch": 0.37, + "learning_rate": 4.3863811233649656e-05, + "loss": 0.5734, + "step": 9620 + }, + { + "epoch": 0.37, + "learning_rate": 4.386317004360093e-05, + "loss": 0.5637, + "step": 9621 + }, + { + "epoch": 0.37, + "learning_rate": 4.386252885355219e-05, + "loss": 0.5748, + "step": 9622 + }, + { + "epoch": 0.37, + "learning_rate": 4.3861887663503464e-05, + "loss": 0.5225, + "step": 9623 + }, + { + "epoch": 0.37, + "learning_rate": 4.3861246473454736e-05, + "loss": 0.5442, + "step": 9624 + }, + { + "epoch": 0.37, + "learning_rate": 4.3860605283406e-05, + "loss": 0.551, + "step": 9625 + }, + { + "epoch": 0.37, + "learning_rate": 4.385996409335727e-05, + "loss": 0.6664, + "step": 9626 + }, + { + "epoch": 0.37, + "learning_rate": 4.3859322903308545e-05, + "loss": 0.5812, + "step": 9627 + }, + { + "epoch": 0.37, + "learning_rate": 4.3858681713259816e-05, + "loss": 0.574, + "step": 9628 + }, + { + "epoch": 0.37, + "learning_rate": 4.385804052321108e-05, + "loss": 0.6197, + "step": 9629 + }, + { + "epoch": 0.37, + "learning_rate": 4.385739933316235e-05, + "loss": 0.576, + "step": 9630 + }, + { + "epoch": 0.37, + "learning_rate": 4.385675814311362e-05, + "loss": 0.6144, + "step": 9631 + }, + { + "epoch": 0.37, + "learning_rate": 4.385611695306489e-05, + "loss": 0.5403, + "step": 9632 + }, + { + "epoch": 0.37, + "learning_rate": 4.385547576301616e-05, + "loss": 0.6049, + "step": 9633 + }, + { + "epoch": 0.37, + "learning_rate": 4.3854834572967426e-05, + "loss": 0.6125, + "step": 9634 + }, + { + "epoch": 0.37, + "learning_rate": 4.38541933829187e-05, + "loss": 0.5737, + "step": 9635 + }, + { + "epoch": 0.37, + "learning_rate": 4.385355219286997e-05, + "loss": 0.5348, + "step": 9636 + }, + { + "epoch": 0.37, + "learning_rate": 4.385291100282124e-05, + "loss": 0.572, + "step": 9637 + }, + { + "epoch": 0.37, + "learning_rate": 4.3852269812772506e-05, + "loss": 0.5296, + "step": 9638 + }, + { + "epoch": 0.37, + "learning_rate": 4.385162862272378e-05, + "loss": 0.5863, + "step": 9639 + }, + { + "epoch": 0.37, + "learning_rate": 4.385098743267504e-05, + "loss": 0.5869, + "step": 9640 + }, + { + "epoch": 0.37, + "learning_rate": 4.3850346242626315e-05, + "loss": 0.5466, + "step": 9641 + }, + { + "epoch": 0.37, + "learning_rate": 4.3849705052577586e-05, + "loss": 0.4956, + "step": 9642 + }, + { + "epoch": 0.37, + "learning_rate": 4.384906386252886e-05, + "loss": 0.6111, + "step": 9643 + }, + { + "epoch": 0.37, + "learning_rate": 4.384842267248013e-05, + "loss": 0.5032, + "step": 9644 + }, + { + "epoch": 0.37, + "learning_rate": 4.3847781482431395e-05, + "loss": 0.6063, + "step": 9645 + }, + { + "epoch": 0.37, + "learning_rate": 4.3847140292382667e-05, + "loss": 0.4822, + "step": 9646 + }, + { + "epoch": 0.37, + "learning_rate": 4.384649910233393e-05, + "loss": 0.6523, + "step": 9647 + }, + { + "epoch": 0.37, + "learning_rate": 4.38458579122852e-05, + "loss": 0.5259, + "step": 9648 + }, + { + "epoch": 0.37, + "learning_rate": 4.384521672223647e-05, + "loss": 0.5407, + "step": 9649 + }, + { + "epoch": 0.37, + "learning_rate": 4.384457553218774e-05, + "loss": 0.5313, + "step": 9650 + }, + { + "epoch": 0.37, + "learning_rate": 4.384393434213901e-05, + "loss": 0.613, + "step": 9651 + }, + { + "epoch": 0.37, + "learning_rate": 4.384329315209028e-05, + "loss": 0.5252, + "step": 9652 + }, + { + "epoch": 0.37, + "learning_rate": 4.3842651962041555e-05, + "loss": 0.6507, + "step": 9653 + }, + { + "epoch": 0.37, + "learning_rate": 4.384201077199282e-05, + "loss": 0.6914, + "step": 9654 + }, + { + "epoch": 0.37, + "learning_rate": 4.384136958194409e-05, + "loss": 0.5759, + "step": 9655 + }, + { + "epoch": 0.37, + "learning_rate": 4.3840728391895357e-05, + "loss": 0.5838, + "step": 9656 + }, + { + "epoch": 0.37, + "learning_rate": 4.384008720184663e-05, + "loss": 0.5415, + "step": 9657 + }, + { + "epoch": 0.37, + "learning_rate": 4.38394460117979e-05, + "loss": 0.6187, + "step": 9658 + }, + { + "epoch": 0.37, + "learning_rate": 4.383880482174917e-05, + "loss": 0.4915, + "step": 9659 + }, + { + "epoch": 0.37, + "learning_rate": 4.3838163631700443e-05, + "loss": 0.623, + "step": 9660 + }, + { + "epoch": 0.37, + "learning_rate": 4.383752244165171e-05, + "loss": 0.6085, + "step": 9661 + }, + { + "epoch": 0.37, + "learning_rate": 4.383688125160298e-05, + "loss": 0.6174, + "step": 9662 + }, + { + "epoch": 0.37, + "learning_rate": 4.3836240061554245e-05, + "loss": 0.5771, + "step": 9663 + }, + { + "epoch": 0.37, + "learning_rate": 4.383559887150552e-05, + "loss": 0.552, + "step": 9664 + }, + { + "epoch": 0.37, + "learning_rate": 4.383495768145678e-05, + "loss": 0.5642, + "step": 9665 + }, + { + "epoch": 0.37, + "learning_rate": 4.3834316491408053e-05, + "loss": 0.5699, + "step": 9666 + }, + { + "epoch": 0.37, + "learning_rate": 4.3833675301359325e-05, + "loss": 0.646, + "step": 9667 + }, + { + "epoch": 0.37, + "learning_rate": 4.38330341113106e-05, + "loss": 0.6895, + "step": 9668 + }, + { + "epoch": 0.37, + "learning_rate": 4.383239292126187e-05, + "loss": 0.4776, + "step": 9669 + }, + { + "epoch": 0.37, + "learning_rate": 4.3831751731213134e-05, + "loss": 0.5204, + "step": 9670 + }, + { + "epoch": 0.37, + "learning_rate": 4.3831110541164405e-05, + "loss": 0.6052, + "step": 9671 + }, + { + "epoch": 0.37, + "learning_rate": 4.383046935111567e-05, + "loss": 0.5233, + "step": 9672 + }, + { + "epoch": 0.37, + "learning_rate": 4.382982816106694e-05, + "loss": 0.5938, + "step": 9673 + }, + { + "epoch": 0.37, + "learning_rate": 4.382918697101821e-05, + "loss": 0.5419, + "step": 9674 + }, + { + "epoch": 0.37, + "learning_rate": 4.3828545780969485e-05, + "loss": 0.5301, + "step": 9675 + }, + { + "epoch": 0.37, + "learning_rate": 4.382790459092075e-05, + "loss": 0.6571, + "step": 9676 + }, + { + "epoch": 0.37, + "learning_rate": 4.382726340087202e-05, + "loss": 0.5774, + "step": 9677 + }, + { + "epoch": 0.37, + "learning_rate": 4.3826622210823294e-05, + "loss": 0.5406, + "step": 9678 + }, + { + "epoch": 0.37, + "learning_rate": 4.382598102077456e-05, + "loss": 0.5437, + "step": 9679 + }, + { + "epoch": 0.37, + "learning_rate": 4.382533983072583e-05, + "loss": 0.6224, + "step": 9680 + }, + { + "epoch": 0.37, + "learning_rate": 4.3824698640677095e-05, + "loss": 0.5408, + "step": 9681 + }, + { + "epoch": 0.37, + "learning_rate": 4.382405745062837e-05, + "loss": 0.5249, + "step": 9682 + }, + { + "epoch": 0.37, + "learning_rate": 4.382341626057964e-05, + "loss": 0.5262, + "step": 9683 + }, + { + "epoch": 0.37, + "learning_rate": 4.382277507053091e-05, + "loss": 0.5413, + "step": 9684 + }, + { + "epoch": 0.37, + "learning_rate": 4.3822133880482175e-05, + "loss": 0.6076, + "step": 9685 + }, + { + "epoch": 0.37, + "learning_rate": 4.382149269043345e-05, + "loss": 0.6594, + "step": 9686 + }, + { + "epoch": 0.37, + "learning_rate": 4.382085150038472e-05, + "loss": 0.5409, + "step": 9687 + }, + { + "epoch": 0.37, + "learning_rate": 4.3820210310335984e-05, + "loss": 0.626, + "step": 9688 + }, + { + "epoch": 0.37, + "learning_rate": 4.3819569120287255e-05, + "loss": 0.5779, + "step": 9689 + }, + { + "epoch": 0.37, + "learning_rate": 4.381892793023852e-05, + "loss": 0.6202, + "step": 9690 + }, + { + "epoch": 0.37, + "learning_rate": 4.38182867401898e-05, + "loss": 0.5518, + "step": 9691 + }, + { + "epoch": 0.37, + "learning_rate": 4.3817645550141064e-05, + "loss": 0.5631, + "step": 9692 + }, + { + "epoch": 0.37, + "learning_rate": 4.3817004360092336e-05, + "loss": 0.5094, + "step": 9693 + }, + { + "epoch": 0.37, + "learning_rate": 4.38163631700436e-05, + "loss": 0.7, + "step": 9694 + }, + { + "epoch": 0.37, + "learning_rate": 4.381572197999487e-05, + "loss": 0.5744, + "step": 9695 + }, + { + "epoch": 0.37, + "learning_rate": 4.3815080789946144e-05, + "loss": 0.6047, + "step": 9696 + }, + { + "epoch": 0.37, + "learning_rate": 4.381443959989741e-05, + "loss": 0.4863, + "step": 9697 + }, + { + "epoch": 0.37, + "learning_rate": 4.381379840984868e-05, + "loss": 0.5898, + "step": 9698 + }, + { + "epoch": 0.37, + "learning_rate": 4.381315721979995e-05, + "loss": 0.5559, + "step": 9699 + }, + { + "epoch": 0.37, + "learning_rate": 4.3812516029751224e-05, + "loss": 0.533, + "step": 9700 + }, + { + "epoch": 0.37, + "learning_rate": 4.381187483970249e-05, + "loss": 0.6015, + "step": 9701 + }, + { + "epoch": 0.37, + "learning_rate": 4.381123364965376e-05, + "loss": 0.635, + "step": 9702 + }, + { + "epoch": 0.37, + "learning_rate": 4.3810592459605026e-05, + "loss": 0.5683, + "step": 9703 + }, + { + "epoch": 0.37, + "learning_rate": 4.38099512695563e-05, + "loss": 0.5396, + "step": 9704 + }, + { + "epoch": 0.37, + "learning_rate": 4.380931007950757e-05, + "loss": 0.5002, + "step": 9705 + }, + { + "epoch": 0.37, + "learning_rate": 4.3808668889458834e-05, + "loss": 0.4738, + "step": 9706 + }, + { + "epoch": 0.37, + "learning_rate": 4.3808027699410106e-05, + "loss": 0.5655, + "step": 9707 + }, + { + "epoch": 0.37, + "learning_rate": 4.380738650936138e-05, + "loss": 0.5361, + "step": 9708 + }, + { + "epoch": 0.37, + "learning_rate": 4.380674531931265e-05, + "loss": 0.6336, + "step": 9709 + }, + { + "epoch": 0.37, + "learning_rate": 4.3806104129263914e-05, + "loss": 0.53, + "step": 9710 + }, + { + "epoch": 0.37, + "learning_rate": 4.3805462939215186e-05, + "loss": 0.6147, + "step": 9711 + }, + { + "epoch": 0.37, + "learning_rate": 4.380482174916645e-05, + "loss": 0.5792, + "step": 9712 + }, + { + "epoch": 0.37, + "learning_rate": 4.380418055911772e-05, + "loss": 0.5322, + "step": 9713 + }, + { + "epoch": 0.37, + "learning_rate": 4.3803539369068994e-05, + "loss": 0.5806, + "step": 9714 + }, + { + "epoch": 0.37, + "learning_rate": 4.3802898179020266e-05, + "loss": 0.4737, + "step": 9715 + }, + { + "epoch": 0.37, + "learning_rate": 4.380225698897154e-05, + "loss": 0.6595, + "step": 9716 + }, + { + "epoch": 0.37, + "learning_rate": 4.38016157989228e-05, + "loss": 0.6054, + "step": 9717 + }, + { + "epoch": 0.37, + "learning_rate": 4.3800974608874074e-05, + "loss": 0.6057, + "step": 9718 + }, + { + "epoch": 0.37, + "learning_rate": 4.380033341882534e-05, + "loss": 0.5857, + "step": 9719 + }, + { + "epoch": 0.37, + "learning_rate": 4.379969222877661e-05, + "loss": 0.4941, + "step": 9720 + }, + { + "epoch": 0.37, + "learning_rate": 4.3799051038727876e-05, + "loss": 0.4962, + "step": 9721 + }, + { + "epoch": 0.37, + "learning_rate": 4.379840984867915e-05, + "loss": 0.4627, + "step": 9722 + }, + { + "epoch": 0.37, + "learning_rate": 4.379776865863042e-05, + "loss": 0.6207, + "step": 9723 + }, + { + "epoch": 0.37, + "learning_rate": 4.379712746858169e-05, + "loss": 0.5814, + "step": 9724 + }, + { + "epoch": 0.37, + "learning_rate": 4.379648627853296e-05, + "loss": 0.5417, + "step": 9725 + }, + { + "epoch": 0.37, + "learning_rate": 4.379584508848423e-05, + "loss": 0.5017, + "step": 9726 + }, + { + "epoch": 0.37, + "learning_rate": 4.37952038984355e-05, + "loss": 0.5128, + "step": 9727 + }, + { + "epoch": 0.37, + "learning_rate": 4.3794562708386764e-05, + "loss": 0.5236, + "step": 9728 + }, + { + "epoch": 0.37, + "learning_rate": 4.3793921518338036e-05, + "loss": 0.665, + "step": 9729 + }, + { + "epoch": 0.37, + "learning_rate": 4.379328032828931e-05, + "loss": 0.5147, + "step": 9730 + }, + { + "epoch": 0.37, + "learning_rate": 4.379263913824058e-05, + "loss": 0.5848, + "step": 9731 + }, + { + "epoch": 0.37, + "learning_rate": 4.379199794819185e-05, + "loss": 0.5964, + "step": 9732 + }, + { + "epoch": 0.37, + "learning_rate": 4.3791356758143116e-05, + "loss": 0.5501, + "step": 9733 + }, + { + "epoch": 0.37, + "learning_rate": 4.379071556809439e-05, + "loss": 0.5951, + "step": 9734 + }, + { + "epoch": 0.37, + "learning_rate": 4.379007437804565e-05, + "loss": 0.5618, + "step": 9735 + }, + { + "epoch": 0.37, + "learning_rate": 4.3789433187996925e-05, + "loss": 0.5945, + "step": 9736 + }, + { + "epoch": 0.37, + "learning_rate": 4.378879199794819e-05, + "loss": 0.5415, + "step": 9737 + }, + { + "epoch": 0.37, + "learning_rate": 4.378815080789946e-05, + "loss": 0.6327, + "step": 9738 + }, + { + "epoch": 0.37, + "learning_rate": 4.378750961785073e-05, + "loss": 0.5689, + "step": 9739 + }, + { + "epoch": 0.37, + "learning_rate": 4.3786868427802005e-05, + "loss": 0.6256, + "step": 9740 + }, + { + "epoch": 0.37, + "learning_rate": 4.3786227237753276e-05, + "loss": 0.5701, + "step": 9741 + }, + { + "epoch": 0.37, + "learning_rate": 4.378558604770454e-05, + "loss": 0.5568, + "step": 9742 + }, + { + "epoch": 0.37, + "learning_rate": 4.378494485765581e-05, + "loss": 0.5561, + "step": 9743 + }, + { + "epoch": 0.37, + "learning_rate": 4.378430366760708e-05, + "loss": 0.6252, + "step": 9744 + }, + { + "epoch": 0.37, + "learning_rate": 4.378366247755835e-05, + "loss": 0.5428, + "step": 9745 + }, + { + "epoch": 0.37, + "learning_rate": 4.378302128750962e-05, + "loss": 0.5335, + "step": 9746 + }, + { + "epoch": 0.37, + "learning_rate": 4.378238009746089e-05, + "loss": 0.615, + "step": 9747 + }, + { + "epoch": 0.37, + "learning_rate": 4.378173890741216e-05, + "loss": 0.6174, + "step": 9748 + }, + { + "epoch": 0.37, + "learning_rate": 4.378109771736343e-05, + "loss": 0.5468, + "step": 9749 + }, + { + "epoch": 0.37, + "learning_rate": 4.37804565273147e-05, + "loss": 0.6111, + "step": 9750 + }, + { + "epoch": 0.37, + "learning_rate": 4.3779815337265966e-05, + "loss": 0.5068, + "step": 9751 + }, + { + "epoch": 0.37, + "learning_rate": 4.377917414721724e-05, + "loss": 0.5104, + "step": 9752 + }, + { + "epoch": 0.37, + "learning_rate": 4.37785329571685e-05, + "loss": 0.5458, + "step": 9753 + }, + { + "epoch": 0.38, + "learning_rate": 4.3777891767119775e-05, + "loss": 0.504, + "step": 9754 + }, + { + "epoch": 0.38, + "learning_rate": 4.3777250577071046e-05, + "loss": 0.5602, + "step": 9755 + }, + { + "epoch": 0.38, + "learning_rate": 4.377660938702232e-05, + "loss": 0.5389, + "step": 9756 + }, + { + "epoch": 0.38, + "learning_rate": 4.377596819697358e-05, + "loss": 0.5406, + "step": 9757 + }, + { + "epoch": 0.38, + "learning_rate": 4.3775327006924855e-05, + "loss": 0.5531, + "step": 9758 + }, + { + "epoch": 0.38, + "learning_rate": 4.3774685816876127e-05, + "loss": 0.5894, + "step": 9759 + }, + { + "epoch": 0.38, + "learning_rate": 4.377404462682739e-05, + "loss": 0.5646, + "step": 9760 + }, + { + "epoch": 0.38, + "learning_rate": 4.377340343677866e-05, + "loss": 0.6152, + "step": 9761 + }, + { + "epoch": 0.38, + "learning_rate": 4.377276224672993e-05, + "loss": 0.5128, + "step": 9762 + }, + { + "epoch": 0.38, + "learning_rate": 4.377212105668121e-05, + "loss": 0.5346, + "step": 9763 + }, + { + "epoch": 0.38, + "learning_rate": 4.377147986663247e-05, + "loss": 0.5487, + "step": 9764 + }, + { + "epoch": 0.38, + "learning_rate": 4.377083867658374e-05, + "loss": 0.5218, + "step": 9765 + }, + { + "epoch": 0.38, + "learning_rate": 4.377019748653501e-05, + "loss": 0.5096, + "step": 9766 + }, + { + "epoch": 0.38, + "learning_rate": 4.376955629648628e-05, + "loss": 0.5304, + "step": 9767 + }, + { + "epoch": 0.38, + "learning_rate": 4.376891510643755e-05, + "loss": 0.5219, + "step": 9768 + }, + { + "epoch": 0.38, + "learning_rate": 4.376827391638882e-05, + "loss": 0.5267, + "step": 9769 + }, + { + "epoch": 0.38, + "learning_rate": 4.376763272634009e-05, + "loss": 0.5896, + "step": 9770 + }, + { + "epoch": 0.38, + "learning_rate": 4.376699153629136e-05, + "loss": 0.5656, + "step": 9771 + }, + { + "epoch": 0.38, + "learning_rate": 4.376635034624263e-05, + "loss": 0.5585, + "step": 9772 + }, + { + "epoch": 0.38, + "learning_rate": 4.37657091561939e-05, + "loss": 0.6023, + "step": 9773 + }, + { + "epoch": 0.38, + "learning_rate": 4.376506796614517e-05, + "loss": 0.5344, + "step": 9774 + }, + { + "epoch": 0.38, + "learning_rate": 4.376442677609643e-05, + "loss": 0.4962, + "step": 9775 + }, + { + "epoch": 0.38, + "learning_rate": 4.3763785586047705e-05, + "loss": 0.5631, + "step": 9776 + }, + { + "epoch": 0.38, + "learning_rate": 4.376314439599898e-05, + "loss": 0.5663, + "step": 9777 + }, + { + "epoch": 0.38, + "learning_rate": 4.376250320595024e-05, + "loss": 0.6351, + "step": 9778 + }, + { + "epoch": 0.38, + "learning_rate": 4.376186201590152e-05, + "loss": 0.4507, + "step": 9779 + }, + { + "epoch": 0.38, + "learning_rate": 4.3761220825852785e-05, + "loss": 0.5411, + "step": 9780 + }, + { + "epoch": 0.38, + "learning_rate": 4.376057963580406e-05, + "loss": 0.6491, + "step": 9781 + }, + { + "epoch": 0.38, + "learning_rate": 4.375993844575532e-05, + "loss": 0.4307, + "step": 9782 + }, + { + "epoch": 0.38, + "learning_rate": 4.3759297255706594e-05, + "loss": 0.4473, + "step": 9783 + }, + { + "epoch": 0.38, + "learning_rate": 4.375865606565786e-05, + "loss": 0.6395, + "step": 9784 + }, + { + "epoch": 0.38, + "learning_rate": 4.375801487560913e-05, + "loss": 0.5466, + "step": 9785 + }, + { + "epoch": 0.38, + "learning_rate": 4.37573736855604e-05, + "loss": 0.5728, + "step": 9786 + }, + { + "epoch": 0.38, + "learning_rate": 4.3756732495511674e-05, + "loss": 0.7599, + "step": 9787 + }, + { + "epoch": 0.38, + "learning_rate": 4.3756091305462945e-05, + "loss": 0.618, + "step": 9788 + }, + { + "epoch": 0.38, + "learning_rate": 4.375545011541421e-05, + "loss": 0.4574, + "step": 9789 + }, + { + "epoch": 0.38, + "learning_rate": 4.375480892536548e-05, + "loss": 0.4724, + "step": 9790 + }, + { + "epoch": 0.38, + "learning_rate": 4.375416773531675e-05, + "loss": 0.5329, + "step": 9791 + }, + { + "epoch": 0.38, + "learning_rate": 4.375352654526802e-05, + "loss": 0.5289, + "step": 9792 + }, + { + "epoch": 0.38, + "learning_rate": 4.3752885355219284e-05, + "loss": 0.4686, + "step": 9793 + }, + { + "epoch": 0.38, + "learning_rate": 4.3752244165170555e-05, + "loss": 0.5773, + "step": 9794 + }, + { + "epoch": 0.38, + "learning_rate": 4.375160297512183e-05, + "loss": 0.5445, + "step": 9795 + }, + { + "epoch": 0.38, + "learning_rate": 4.37509617850731e-05, + "loss": 0.7723, + "step": 9796 + }, + { + "epoch": 0.38, + "learning_rate": 4.375032059502437e-05, + "loss": 0.5716, + "step": 9797 + }, + { + "epoch": 0.38, + "learning_rate": 4.3749679404975635e-05, + "loss": 0.6059, + "step": 9798 + }, + { + "epoch": 0.38, + "learning_rate": 4.374903821492691e-05, + "loss": 0.6274, + "step": 9799 + }, + { + "epoch": 0.38, + "learning_rate": 4.374839702487817e-05, + "loss": 0.6422, + "step": 9800 + }, + { + "epoch": 0.38, + "learning_rate": 4.3747755834829444e-05, + "loss": 0.4537, + "step": 9801 + }, + { + "epoch": 0.38, + "learning_rate": 4.3747114644780716e-05, + "loss": 0.5552, + "step": 9802 + }, + { + "epoch": 0.38, + "learning_rate": 4.374647345473199e-05, + "loss": 0.6009, + "step": 9803 + }, + { + "epoch": 0.38, + "learning_rate": 4.374583226468326e-05, + "loss": 0.4983, + "step": 9804 + }, + { + "epoch": 0.38, + "learning_rate": 4.3745191074634524e-05, + "loss": 0.5531, + "step": 9805 + }, + { + "epoch": 0.38, + "learning_rate": 4.3744549884585796e-05, + "loss": 0.5108, + "step": 9806 + }, + { + "epoch": 0.38, + "learning_rate": 4.374390869453706e-05, + "loss": 0.5218, + "step": 9807 + }, + { + "epoch": 0.38, + "learning_rate": 4.374326750448833e-05, + "loss": 0.5845, + "step": 9808 + }, + { + "epoch": 0.38, + "learning_rate": 4.37426263144396e-05, + "loss": 0.5262, + "step": 9809 + }, + { + "epoch": 0.38, + "learning_rate": 4.374198512439087e-05, + "loss": 0.4786, + "step": 9810 + }, + { + "epoch": 0.38, + "learning_rate": 4.374134393434214e-05, + "loss": 0.4962, + "step": 9811 + }, + { + "epoch": 0.38, + "learning_rate": 4.374070274429341e-05, + "loss": 0.4725, + "step": 9812 + }, + { + "epoch": 0.38, + "learning_rate": 4.3740061554244684e-05, + "loss": 0.5341, + "step": 9813 + }, + { + "epoch": 0.38, + "learning_rate": 4.373942036419595e-05, + "loss": 0.4781, + "step": 9814 + }, + { + "epoch": 0.38, + "learning_rate": 4.373877917414722e-05, + "loss": 0.5631, + "step": 9815 + }, + { + "epoch": 0.38, + "learning_rate": 4.3738137984098486e-05, + "loss": 0.6004, + "step": 9816 + }, + { + "epoch": 0.38, + "learning_rate": 4.373749679404976e-05, + "loss": 0.5542, + "step": 9817 + }, + { + "epoch": 0.38, + "learning_rate": 4.373685560400103e-05, + "loss": 0.5775, + "step": 9818 + }, + { + "epoch": 0.38, + "learning_rate": 4.37362144139523e-05, + "loss": 0.6144, + "step": 9819 + }, + { + "epoch": 0.38, + "learning_rate": 4.373557322390357e-05, + "loss": 0.4468, + "step": 9820 + }, + { + "epoch": 0.38, + "learning_rate": 4.373493203385484e-05, + "loss": 0.4681, + "step": 9821 + }, + { + "epoch": 0.38, + "learning_rate": 4.373429084380611e-05, + "loss": 0.5133, + "step": 9822 + }, + { + "epoch": 0.38, + "learning_rate": 4.3733649653757374e-05, + "loss": 0.4899, + "step": 9823 + }, + { + "epoch": 0.38, + "learning_rate": 4.3733008463708646e-05, + "loss": 0.6388, + "step": 9824 + }, + { + "epoch": 0.38, + "learning_rate": 4.373236727365991e-05, + "loss": 0.4989, + "step": 9825 + }, + { + "epoch": 0.38, + "learning_rate": 4.373172608361118e-05, + "loss": 0.5597, + "step": 9826 + }, + { + "epoch": 0.38, + "learning_rate": 4.3731084893562454e-05, + "loss": 0.5365, + "step": 9827 + }, + { + "epoch": 0.38, + "learning_rate": 4.3730443703513726e-05, + "loss": 0.6292, + "step": 9828 + }, + { + "epoch": 0.38, + "learning_rate": 4.3729802513465e-05, + "loss": 0.5982, + "step": 9829 + }, + { + "epoch": 0.38, + "learning_rate": 4.372916132341626e-05, + "loss": 0.5566, + "step": 9830 + }, + { + "epoch": 0.38, + "learning_rate": 4.3728520133367534e-05, + "loss": 0.5331, + "step": 9831 + }, + { + "epoch": 0.38, + "learning_rate": 4.37278789433188e-05, + "loss": 0.5579, + "step": 9832 + }, + { + "epoch": 0.38, + "learning_rate": 4.372723775327007e-05, + "loss": 0.6135, + "step": 9833 + }, + { + "epoch": 0.38, + "learning_rate": 4.3726596563221336e-05, + "loss": 0.6179, + "step": 9834 + }, + { + "epoch": 0.38, + "learning_rate": 4.3725955373172614e-05, + "loss": 0.6876, + "step": 9835 + }, + { + "epoch": 0.38, + "learning_rate": 4.372531418312388e-05, + "loss": 0.5172, + "step": 9836 + }, + { + "epoch": 0.38, + "learning_rate": 4.372467299307515e-05, + "loss": 0.5265, + "step": 9837 + }, + { + "epoch": 0.38, + "learning_rate": 4.372403180302642e-05, + "loss": 0.5494, + "step": 9838 + }, + { + "epoch": 0.38, + "learning_rate": 4.372339061297769e-05, + "loss": 0.6344, + "step": 9839 + }, + { + "epoch": 0.38, + "learning_rate": 4.372274942292896e-05, + "loss": 0.5731, + "step": 9840 + }, + { + "epoch": 0.38, + "learning_rate": 4.3722108232880224e-05, + "loss": 0.5769, + "step": 9841 + }, + { + "epoch": 0.38, + "learning_rate": 4.3721467042831496e-05, + "loss": 0.6547, + "step": 9842 + }, + { + "epoch": 0.38, + "learning_rate": 4.372082585278277e-05, + "loss": 0.6115, + "step": 9843 + }, + { + "epoch": 0.38, + "learning_rate": 4.372018466273404e-05, + "loss": 0.5051, + "step": 9844 + }, + { + "epoch": 0.38, + "learning_rate": 4.3719543472685304e-05, + "loss": 0.4776, + "step": 9845 + }, + { + "epoch": 0.38, + "learning_rate": 4.3718902282636576e-05, + "loss": 0.6089, + "step": 9846 + }, + { + "epoch": 0.38, + "learning_rate": 4.371826109258785e-05, + "loss": 0.6087, + "step": 9847 + }, + { + "epoch": 0.38, + "learning_rate": 4.371761990253911e-05, + "loss": 0.6164, + "step": 9848 + }, + { + "epoch": 0.38, + "learning_rate": 4.3716978712490385e-05, + "loss": 0.5848, + "step": 9849 + }, + { + "epoch": 0.38, + "learning_rate": 4.371633752244165e-05, + "loss": 0.5605, + "step": 9850 + }, + { + "epoch": 0.38, + "learning_rate": 4.371569633239293e-05, + "loss": 0.5281, + "step": 9851 + }, + { + "epoch": 0.38, + "learning_rate": 4.371505514234419e-05, + "loss": 0.5051, + "step": 9852 + }, + { + "epoch": 0.38, + "learning_rate": 4.3714413952295465e-05, + "loss": 0.5378, + "step": 9853 + }, + { + "epoch": 0.38, + "learning_rate": 4.371377276224673e-05, + "loss": 0.5242, + "step": 9854 + }, + { + "epoch": 0.38, + "learning_rate": 4.3713131572198e-05, + "loss": 0.5376, + "step": 9855 + }, + { + "epoch": 0.38, + "learning_rate": 4.371249038214927e-05, + "loss": 0.4943, + "step": 9856 + }, + { + "epoch": 0.38, + "learning_rate": 4.371184919210054e-05, + "loss": 0.6165, + "step": 9857 + }, + { + "epoch": 0.38, + "learning_rate": 4.371120800205181e-05, + "loss": 0.3977, + "step": 9858 + }, + { + "epoch": 0.38, + "learning_rate": 4.371056681200308e-05, + "loss": 0.5666, + "step": 9859 + }, + { + "epoch": 0.38, + "learning_rate": 4.370992562195435e-05, + "loss": 0.6058, + "step": 9860 + }, + { + "epoch": 0.38, + "learning_rate": 4.370928443190562e-05, + "loss": 0.5345, + "step": 9861 + }, + { + "epoch": 0.38, + "learning_rate": 4.370864324185689e-05, + "loss": 0.6421, + "step": 9862 + }, + { + "epoch": 0.38, + "learning_rate": 4.3708002051808155e-05, + "loss": 0.7414, + "step": 9863 + }, + { + "epoch": 0.38, + "learning_rate": 4.3707360861759426e-05, + "loss": 0.4815, + "step": 9864 + }, + { + "epoch": 0.38, + "learning_rate": 4.37067196717107e-05, + "loss": 0.6294, + "step": 9865 + }, + { + "epoch": 0.38, + "learning_rate": 4.370607848166196e-05, + "loss": 0.465, + "step": 9866 + }, + { + "epoch": 0.38, + "learning_rate": 4.370543729161324e-05, + "loss": 0.5357, + "step": 9867 + }, + { + "epoch": 0.38, + "learning_rate": 4.3704796101564506e-05, + "loss": 0.5397, + "step": 9868 + }, + { + "epoch": 0.38, + "learning_rate": 4.370415491151578e-05, + "loss": 0.4629, + "step": 9869 + }, + { + "epoch": 0.38, + "learning_rate": 4.370351372146704e-05, + "loss": 0.4805, + "step": 9870 + }, + { + "epoch": 0.38, + "learning_rate": 4.3702872531418315e-05, + "loss": 0.5009, + "step": 9871 + }, + { + "epoch": 0.38, + "learning_rate": 4.370223134136958e-05, + "loss": 0.5066, + "step": 9872 + }, + { + "epoch": 0.38, + "learning_rate": 4.370159015132085e-05, + "loss": 0.58, + "step": 9873 + }, + { + "epoch": 0.38, + "learning_rate": 4.370094896127212e-05, + "loss": 0.518, + "step": 9874 + }, + { + "epoch": 0.38, + "learning_rate": 4.3700307771223395e-05, + "loss": 0.6339, + "step": 9875 + }, + { + "epoch": 0.38, + "learning_rate": 4.369966658117467e-05, + "loss": 0.5091, + "step": 9876 + }, + { + "epoch": 0.38, + "learning_rate": 4.369902539112593e-05, + "loss": 0.5032, + "step": 9877 + }, + { + "epoch": 0.38, + "learning_rate": 4.36983842010772e-05, + "loss": 0.4899, + "step": 9878 + }, + { + "epoch": 0.38, + "learning_rate": 4.369774301102847e-05, + "loss": 0.4643, + "step": 9879 + }, + { + "epoch": 0.38, + "learning_rate": 4.369710182097974e-05, + "loss": 0.5495, + "step": 9880 + }, + { + "epoch": 0.38, + "learning_rate": 4.3696460630931005e-05, + "loss": 0.6095, + "step": 9881 + }, + { + "epoch": 0.38, + "learning_rate": 4.369581944088228e-05, + "loss": 0.4782, + "step": 9882 + }, + { + "epoch": 0.38, + "learning_rate": 4.369517825083355e-05, + "loss": 0.5562, + "step": 9883 + }, + { + "epoch": 0.38, + "learning_rate": 4.369453706078482e-05, + "loss": 0.5976, + "step": 9884 + }, + { + "epoch": 0.38, + "learning_rate": 4.369389587073609e-05, + "loss": 0.661, + "step": 9885 + }, + { + "epoch": 0.38, + "learning_rate": 4.369325468068736e-05, + "loss": 0.6638, + "step": 9886 + }, + { + "epoch": 0.38, + "learning_rate": 4.369261349063863e-05, + "loss": 0.5315, + "step": 9887 + }, + { + "epoch": 0.38, + "learning_rate": 4.3691972300589893e-05, + "loss": 0.4968, + "step": 9888 + }, + { + "epoch": 0.38, + "learning_rate": 4.3691331110541165e-05, + "loss": 0.5719, + "step": 9889 + }, + { + "epoch": 0.38, + "learning_rate": 4.369068992049244e-05, + "loss": 0.5498, + "step": 9890 + }, + { + "epoch": 0.38, + "learning_rate": 4.369004873044371e-05, + "loss": 0.6032, + "step": 9891 + }, + { + "epoch": 0.38, + "learning_rate": 4.368940754039498e-05, + "loss": 0.4819, + "step": 9892 + }, + { + "epoch": 0.38, + "learning_rate": 4.3688766350346245e-05, + "loss": 0.5665, + "step": 9893 + }, + { + "epoch": 0.38, + "learning_rate": 4.368812516029752e-05, + "loss": 0.6636, + "step": 9894 + }, + { + "epoch": 0.38, + "learning_rate": 4.368748397024878e-05, + "loss": 0.5686, + "step": 9895 + }, + { + "epoch": 0.38, + "learning_rate": 4.3686842780200054e-05, + "loss": 0.6346, + "step": 9896 + }, + { + "epoch": 0.38, + "learning_rate": 4.368620159015132e-05, + "loss": 0.5396, + "step": 9897 + }, + { + "epoch": 0.38, + "learning_rate": 4.368556040010259e-05, + "loss": 0.5584, + "step": 9898 + }, + { + "epoch": 0.38, + "learning_rate": 4.368491921005386e-05, + "loss": 0.5553, + "step": 9899 + }, + { + "epoch": 0.38, + "learning_rate": 4.3684278020005134e-05, + "loss": 0.517, + "step": 9900 + }, + { + "epoch": 0.38, + "learning_rate": 4.3683636829956405e-05, + "loss": 0.6631, + "step": 9901 + }, + { + "epoch": 0.38, + "learning_rate": 4.368299563990767e-05, + "loss": 0.5673, + "step": 9902 + }, + { + "epoch": 0.38, + "learning_rate": 4.368235444985894e-05, + "loss": 0.62, + "step": 9903 + }, + { + "epoch": 0.38, + "learning_rate": 4.368171325981021e-05, + "loss": 0.489, + "step": 9904 + }, + { + "epoch": 0.38, + "learning_rate": 4.368107206976148e-05, + "loss": 0.5907, + "step": 9905 + }, + { + "epoch": 0.38, + "learning_rate": 4.368043087971275e-05, + "loss": 0.5309, + "step": 9906 + }, + { + "epoch": 0.38, + "learning_rate": 4.367978968966402e-05, + "loss": 0.5045, + "step": 9907 + }, + { + "epoch": 0.38, + "learning_rate": 4.367914849961529e-05, + "loss": 0.5585, + "step": 9908 + }, + { + "epoch": 0.38, + "learning_rate": 4.367850730956656e-05, + "loss": 0.5391, + "step": 9909 + }, + { + "epoch": 0.38, + "learning_rate": 4.367786611951783e-05, + "loss": 0.5642, + "step": 9910 + }, + { + "epoch": 0.38, + "learning_rate": 4.3677224929469095e-05, + "loss": 0.5177, + "step": 9911 + }, + { + "epoch": 0.38, + "learning_rate": 4.367658373942037e-05, + "loss": 0.5124, + "step": 9912 + }, + { + "epoch": 0.38, + "learning_rate": 4.367594254937163e-05, + "loss": 0.5777, + "step": 9913 + }, + { + "epoch": 0.38, + "learning_rate": 4.3675301359322904e-05, + "loss": 0.5605, + "step": 9914 + }, + { + "epoch": 0.38, + "learning_rate": 4.3674660169274176e-05, + "loss": 0.5809, + "step": 9915 + }, + { + "epoch": 0.38, + "learning_rate": 4.367401897922545e-05, + "loss": 0.6302, + "step": 9916 + }, + { + "epoch": 0.38, + "learning_rate": 4.367337778917671e-05, + "loss": 0.5946, + "step": 9917 + }, + { + "epoch": 0.38, + "learning_rate": 4.3672736599127984e-05, + "loss": 0.5752, + "step": 9918 + }, + { + "epoch": 0.38, + "learning_rate": 4.3672095409079256e-05, + "loss": 0.5892, + "step": 9919 + }, + { + "epoch": 0.38, + "learning_rate": 4.367145421903052e-05, + "loss": 0.7134, + "step": 9920 + }, + { + "epoch": 0.38, + "learning_rate": 4.367081302898179e-05, + "loss": 0.5403, + "step": 9921 + }, + { + "epoch": 0.38, + "learning_rate": 4.367017183893306e-05, + "loss": 0.5168, + "step": 9922 + }, + { + "epoch": 0.38, + "learning_rate": 4.3669530648884336e-05, + "loss": 0.551, + "step": 9923 + }, + { + "epoch": 0.38, + "learning_rate": 4.36688894588356e-05, + "loss": 0.6352, + "step": 9924 + }, + { + "epoch": 0.38, + "learning_rate": 4.366824826878687e-05, + "loss": 0.5932, + "step": 9925 + }, + { + "epoch": 0.38, + "learning_rate": 4.366760707873814e-05, + "loss": 0.5253, + "step": 9926 + }, + { + "epoch": 0.38, + "learning_rate": 4.366696588868941e-05, + "loss": 0.4984, + "step": 9927 + }, + { + "epoch": 0.38, + "learning_rate": 4.366632469864068e-05, + "loss": 0.6404, + "step": 9928 + }, + { + "epoch": 0.38, + "learning_rate": 4.3665683508591946e-05, + "loss": 0.5102, + "step": 9929 + }, + { + "epoch": 0.38, + "learning_rate": 4.366504231854322e-05, + "loss": 0.5611, + "step": 9930 + }, + { + "epoch": 0.38, + "learning_rate": 4.366440112849449e-05, + "loss": 0.5691, + "step": 9931 + }, + { + "epoch": 0.38, + "learning_rate": 4.366375993844576e-05, + "loss": 0.6302, + "step": 9932 + }, + { + "epoch": 0.38, + "learning_rate": 4.3663118748397026e-05, + "loss": 0.525, + "step": 9933 + }, + { + "epoch": 0.38, + "learning_rate": 4.36624775583483e-05, + "loss": 0.4826, + "step": 9934 + }, + { + "epoch": 0.38, + "learning_rate": 4.366183636829956e-05, + "loss": 0.5423, + "step": 9935 + }, + { + "epoch": 0.38, + "learning_rate": 4.3661195178250834e-05, + "loss": 0.6523, + "step": 9936 + }, + { + "epoch": 0.38, + "learning_rate": 4.3660553988202106e-05, + "loss": 0.5367, + "step": 9937 + }, + { + "epoch": 0.38, + "learning_rate": 4.365991279815337e-05, + "loss": 0.4719, + "step": 9938 + }, + { + "epoch": 0.38, + "learning_rate": 4.365927160810465e-05, + "loss": 0.4829, + "step": 9939 + }, + { + "epoch": 0.38, + "learning_rate": 4.3658630418055914e-05, + "loss": 0.6272, + "step": 9940 + }, + { + "epoch": 0.38, + "learning_rate": 4.3657989228007186e-05, + "loss": 0.4895, + "step": 9941 + }, + { + "epoch": 0.38, + "learning_rate": 4.365734803795845e-05, + "loss": 0.5715, + "step": 9942 + }, + { + "epoch": 0.38, + "learning_rate": 4.365670684790972e-05, + "loss": 0.5765, + "step": 9943 + }, + { + "epoch": 0.38, + "learning_rate": 4.365606565786099e-05, + "loss": 0.5512, + "step": 9944 + }, + { + "epoch": 0.38, + "learning_rate": 4.365542446781226e-05, + "loss": 0.4473, + "step": 9945 + }, + { + "epoch": 0.38, + "learning_rate": 4.365478327776353e-05, + "loss": 0.5996, + "step": 9946 + }, + { + "epoch": 0.38, + "learning_rate": 4.36541420877148e-05, + "loss": 0.5026, + "step": 9947 + }, + { + "epoch": 0.38, + "learning_rate": 4.3653500897666074e-05, + "loss": 0.6041, + "step": 9948 + }, + { + "epoch": 0.38, + "learning_rate": 4.365285970761734e-05, + "loss": 0.6055, + "step": 9949 + }, + { + "epoch": 0.38, + "learning_rate": 4.365221851756861e-05, + "loss": 0.7124, + "step": 9950 + }, + { + "epoch": 0.38, + "learning_rate": 4.3651577327519876e-05, + "loss": 0.5714, + "step": 9951 + }, + { + "epoch": 0.38, + "learning_rate": 4.365093613747115e-05, + "loss": 0.6254, + "step": 9952 + }, + { + "epoch": 0.38, + "learning_rate": 4.365029494742241e-05, + "loss": 0.6542, + "step": 9953 + }, + { + "epoch": 0.38, + "learning_rate": 4.3649653757373684e-05, + "loss": 0.6067, + "step": 9954 + }, + { + "epoch": 0.38, + "learning_rate": 4.3649012567324956e-05, + "loss": 0.615, + "step": 9955 + }, + { + "epoch": 0.38, + "learning_rate": 4.364837137727623e-05, + "loss": 0.574, + "step": 9956 + }, + { + "epoch": 0.38, + "learning_rate": 4.36477301872275e-05, + "loss": 0.6413, + "step": 9957 + }, + { + "epoch": 0.38, + "learning_rate": 4.3647088997178764e-05, + "loss": 0.6198, + "step": 9958 + }, + { + "epoch": 0.38, + "learning_rate": 4.3646447807130036e-05, + "loss": 0.5712, + "step": 9959 + }, + { + "epoch": 0.38, + "learning_rate": 4.36458066170813e-05, + "loss": 0.5934, + "step": 9960 + }, + { + "epoch": 0.38, + "learning_rate": 4.364516542703257e-05, + "loss": 0.5507, + "step": 9961 + }, + { + "epoch": 0.38, + "learning_rate": 4.3644524236983845e-05, + "loss": 0.4915, + "step": 9962 + }, + { + "epoch": 0.38, + "learning_rate": 4.3643883046935116e-05, + "loss": 0.5768, + "step": 9963 + }, + { + "epoch": 0.38, + "learning_rate": 4.364324185688639e-05, + "loss": 0.6045, + "step": 9964 + }, + { + "epoch": 0.38, + "learning_rate": 4.364260066683765e-05, + "loss": 0.5909, + "step": 9965 + }, + { + "epoch": 0.38, + "learning_rate": 4.3641959476788925e-05, + "loss": 0.5339, + "step": 9966 + }, + { + "epoch": 0.38, + "learning_rate": 4.364131828674019e-05, + "loss": 0.5079, + "step": 9967 + }, + { + "epoch": 0.38, + "learning_rate": 4.364067709669146e-05, + "loss": 0.5928, + "step": 9968 + }, + { + "epoch": 0.38, + "learning_rate": 4.3640035906642726e-05, + "loss": 0.6499, + "step": 9969 + }, + { + "epoch": 0.38, + "learning_rate": 4.3639394716594e-05, + "loss": 0.5639, + "step": 9970 + }, + { + "epoch": 0.38, + "learning_rate": 4.363875352654527e-05, + "loss": 0.5559, + "step": 9971 + }, + { + "epoch": 0.38, + "learning_rate": 4.363811233649654e-05, + "loss": 0.5669, + "step": 9972 + }, + { + "epoch": 0.38, + "learning_rate": 4.363747114644781e-05, + "loss": 0.5232, + "step": 9973 + }, + { + "epoch": 0.38, + "learning_rate": 4.363682995639908e-05, + "loss": 0.5918, + "step": 9974 + }, + { + "epoch": 0.38, + "learning_rate": 4.363618876635035e-05, + "loss": 0.4965, + "step": 9975 + }, + { + "epoch": 0.38, + "learning_rate": 4.3635547576301615e-05, + "loss": 0.5984, + "step": 9976 + }, + { + "epoch": 0.38, + "learning_rate": 4.3634906386252886e-05, + "loss": 0.5773, + "step": 9977 + }, + { + "epoch": 0.38, + "learning_rate": 4.363426519620416e-05, + "loss": 0.594, + "step": 9978 + }, + { + "epoch": 0.38, + "learning_rate": 4.363362400615543e-05, + "loss": 0.5794, + "step": 9979 + }, + { + "epoch": 0.38, + "learning_rate": 4.3632982816106695e-05, + "loss": 0.6208, + "step": 9980 + }, + { + "epoch": 0.38, + "learning_rate": 4.3632341626057967e-05, + "loss": 0.6266, + "step": 9981 + }, + { + "epoch": 0.38, + "learning_rate": 4.363170043600924e-05, + "loss": 0.5104, + "step": 9982 + }, + { + "epoch": 0.38, + "learning_rate": 4.36310592459605e-05, + "loss": 0.5467, + "step": 9983 + }, + { + "epoch": 0.38, + "learning_rate": 4.3630418055911775e-05, + "loss": 0.6182, + "step": 9984 + }, + { + "epoch": 0.38, + "learning_rate": 4.362977686586304e-05, + "loss": 0.5824, + "step": 9985 + }, + { + "epoch": 0.38, + "learning_rate": 4.362913567581431e-05, + "loss": 0.5886, + "step": 9986 + }, + { + "epoch": 0.38, + "learning_rate": 4.362849448576558e-05, + "loss": 0.5536, + "step": 9987 + }, + { + "epoch": 0.38, + "learning_rate": 4.3627853295716855e-05, + "loss": 0.5639, + "step": 9988 + }, + { + "epoch": 0.38, + "learning_rate": 4.362721210566813e-05, + "loss": 0.5733, + "step": 9989 + }, + { + "epoch": 0.38, + "learning_rate": 4.362657091561939e-05, + "loss": 0.5807, + "step": 9990 + }, + { + "epoch": 0.38, + "learning_rate": 4.362592972557066e-05, + "loss": 0.5875, + "step": 9991 + }, + { + "epoch": 0.38, + "learning_rate": 4.362528853552193e-05, + "loss": 0.491, + "step": 9992 + }, + { + "epoch": 0.38, + "learning_rate": 4.36246473454732e-05, + "loss": 0.4846, + "step": 9993 + }, + { + "epoch": 0.38, + "learning_rate": 4.3624006155424465e-05, + "loss": 0.5881, + "step": 9994 + }, + { + "epoch": 0.38, + "learning_rate": 4.3623364965375743e-05, + "loss": 0.598, + "step": 9995 + }, + { + "epoch": 0.38, + "learning_rate": 4.362272377532701e-05, + "loss": 0.5306, + "step": 9996 + }, + { + "epoch": 0.38, + "learning_rate": 4.362208258527828e-05, + "loss": 0.5221, + "step": 9997 + }, + { + "epoch": 0.38, + "learning_rate": 4.362144139522955e-05, + "loss": 0.5392, + "step": 9998 + }, + { + "epoch": 0.38, + "learning_rate": 4.362080020518082e-05, + "loss": 0.5599, + "step": 9999 + }, + { + "epoch": 0.38, + "learning_rate": 4.362015901513209e-05, + "loss": 0.595, + "step": 10000 + }, + { + "epoch": 0.38, + "learning_rate": 4.3619517825083353e-05, + "loss": 0.5503, + "step": 10001 + }, + { + "epoch": 0.38, + "learning_rate": 4.3618876635034625e-05, + "loss": 0.553, + "step": 10002 + }, + { + "epoch": 0.38, + "learning_rate": 4.36182354449859e-05, + "loss": 0.4927, + "step": 10003 + }, + { + "epoch": 0.38, + "learning_rate": 4.361759425493717e-05, + "loss": 0.5907, + "step": 10004 + }, + { + "epoch": 0.38, + "learning_rate": 4.3616953064888434e-05, + "loss": 0.5177, + "step": 10005 + }, + { + "epoch": 0.38, + "learning_rate": 4.3616311874839705e-05, + "loss": 0.4543, + "step": 10006 + }, + { + "epoch": 0.38, + "learning_rate": 4.361567068479098e-05, + "loss": 0.4947, + "step": 10007 + }, + { + "epoch": 0.38, + "learning_rate": 4.361502949474224e-05, + "loss": 0.5541, + "step": 10008 + }, + { + "epoch": 0.38, + "learning_rate": 4.3614388304693514e-05, + "loss": 0.575, + "step": 10009 + }, + { + "epoch": 0.38, + "learning_rate": 4.361374711464478e-05, + "loss": 0.5934, + "step": 10010 + }, + { + "epoch": 0.38, + "learning_rate": 4.361310592459606e-05, + "loss": 0.5468, + "step": 10011 + }, + { + "epoch": 0.38, + "learning_rate": 4.361246473454732e-05, + "loss": 0.5982, + "step": 10012 + }, + { + "epoch": 0.38, + "learning_rate": 4.3611823544498594e-05, + "loss": 0.4948, + "step": 10013 + }, + { + "epoch": 0.39, + "learning_rate": 4.361118235444986e-05, + "loss": 0.5034, + "step": 10014 + }, + { + "epoch": 0.39, + "learning_rate": 4.361054116440113e-05, + "loss": 0.4579, + "step": 10015 + }, + { + "epoch": 0.39, + "learning_rate": 4.36098999743524e-05, + "loss": 0.5136, + "step": 10016 + }, + { + "epoch": 0.39, + "learning_rate": 4.360925878430367e-05, + "loss": 0.5961, + "step": 10017 + }, + { + "epoch": 0.39, + "learning_rate": 4.360861759425494e-05, + "loss": 0.5473, + "step": 10018 + }, + { + "epoch": 0.39, + "learning_rate": 4.360797640420621e-05, + "loss": 0.6665, + "step": 10019 + }, + { + "epoch": 0.39, + "learning_rate": 4.360733521415748e-05, + "loss": 0.612, + "step": 10020 + }, + { + "epoch": 0.39, + "learning_rate": 4.360669402410875e-05, + "loss": 0.5972, + "step": 10021 + }, + { + "epoch": 0.39, + "learning_rate": 4.360605283406002e-05, + "loss": 0.5437, + "step": 10022 + }, + { + "epoch": 0.39, + "learning_rate": 4.3605411644011284e-05, + "loss": 0.6813, + "step": 10023 + }, + { + "epoch": 0.39, + "learning_rate": 4.3604770453962555e-05, + "loss": 0.6311, + "step": 10024 + }, + { + "epoch": 0.39, + "learning_rate": 4.360412926391383e-05, + "loss": 0.6191, + "step": 10025 + }, + { + "epoch": 0.39, + "learning_rate": 4.360348807386509e-05, + "loss": 0.4873, + "step": 10026 + }, + { + "epoch": 0.39, + "learning_rate": 4.360284688381637e-05, + "loss": 0.5583, + "step": 10027 + }, + { + "epoch": 0.39, + "learning_rate": 4.3602205693767636e-05, + "loss": 0.6197, + "step": 10028 + }, + { + "epoch": 0.39, + "learning_rate": 4.360156450371891e-05, + "loss": 0.551, + "step": 10029 + }, + { + "epoch": 0.39, + "learning_rate": 4.360092331367017e-05, + "loss": 0.6926, + "step": 10030 + }, + { + "epoch": 0.39, + "learning_rate": 4.3600282123621444e-05, + "loss": 0.6023, + "step": 10031 + }, + { + "epoch": 0.39, + "learning_rate": 4.359964093357271e-05, + "loss": 0.5925, + "step": 10032 + }, + { + "epoch": 0.39, + "learning_rate": 4.359899974352398e-05, + "loss": 0.6257, + "step": 10033 + }, + { + "epoch": 0.39, + "learning_rate": 4.359835855347525e-05, + "loss": 0.5617, + "step": 10034 + }, + { + "epoch": 0.39, + "learning_rate": 4.3597717363426524e-05, + "loss": 0.5823, + "step": 10035 + }, + { + "epoch": 0.39, + "learning_rate": 4.3597076173377796e-05, + "loss": 0.5883, + "step": 10036 + }, + { + "epoch": 0.39, + "learning_rate": 4.359643498332906e-05, + "loss": 0.6145, + "step": 10037 + }, + { + "epoch": 0.39, + "learning_rate": 4.359579379328033e-05, + "loss": 0.5804, + "step": 10038 + }, + { + "epoch": 0.39, + "learning_rate": 4.35951526032316e-05, + "loss": 0.6096, + "step": 10039 + }, + { + "epoch": 0.39, + "learning_rate": 4.359451141318287e-05, + "loss": 0.5522, + "step": 10040 + }, + { + "epoch": 0.39, + "learning_rate": 4.3593870223134134e-05, + "loss": 0.5628, + "step": 10041 + }, + { + "epoch": 0.39, + "learning_rate": 4.3593229033085406e-05, + "loss": 0.5559, + "step": 10042 + }, + { + "epoch": 0.39, + "learning_rate": 4.359258784303668e-05, + "loss": 0.5456, + "step": 10043 + }, + { + "epoch": 0.39, + "learning_rate": 4.359194665298795e-05, + "loss": 0.6936, + "step": 10044 + }, + { + "epoch": 0.39, + "learning_rate": 4.359130546293922e-05, + "loss": 0.5694, + "step": 10045 + }, + { + "epoch": 0.39, + "learning_rate": 4.3590664272890486e-05, + "loss": 0.5406, + "step": 10046 + }, + { + "epoch": 0.39, + "learning_rate": 4.359002308284176e-05, + "loss": 0.5335, + "step": 10047 + }, + { + "epoch": 0.39, + "learning_rate": 4.358938189279302e-05, + "loss": 0.6092, + "step": 10048 + }, + { + "epoch": 0.39, + "learning_rate": 4.3588740702744294e-05, + "loss": 0.5872, + "step": 10049 + }, + { + "epoch": 0.39, + "learning_rate": 4.3588099512695566e-05, + "loss": 0.5882, + "step": 10050 + }, + { + "epoch": 0.39, + "learning_rate": 4.358745832264684e-05, + "loss": 0.6187, + "step": 10051 + }, + { + "epoch": 0.39, + "learning_rate": 4.358681713259811e-05, + "loss": 0.4796, + "step": 10052 + }, + { + "epoch": 0.39, + "learning_rate": 4.3586175942549374e-05, + "loss": 0.5007, + "step": 10053 + }, + { + "epoch": 0.39, + "learning_rate": 4.3585534752500646e-05, + "loss": 0.6542, + "step": 10054 + }, + { + "epoch": 0.39, + "learning_rate": 4.358489356245191e-05, + "loss": 0.5249, + "step": 10055 + }, + { + "epoch": 0.39, + "learning_rate": 4.358425237240318e-05, + "loss": 0.5676, + "step": 10056 + }, + { + "epoch": 0.39, + "learning_rate": 4.358361118235445e-05, + "loss": 0.4875, + "step": 10057 + }, + { + "epoch": 0.39, + "learning_rate": 4.358296999230572e-05, + "loss": 0.5577, + "step": 10058 + }, + { + "epoch": 0.39, + "learning_rate": 4.358232880225699e-05, + "loss": 0.5969, + "step": 10059 + }, + { + "epoch": 0.39, + "learning_rate": 4.358168761220826e-05, + "loss": 0.5327, + "step": 10060 + }, + { + "epoch": 0.39, + "learning_rate": 4.3581046422159534e-05, + "loss": 0.5232, + "step": 10061 + }, + { + "epoch": 0.39, + "learning_rate": 4.35804052321108e-05, + "loss": 0.4772, + "step": 10062 + }, + { + "epoch": 0.39, + "learning_rate": 4.357976404206207e-05, + "loss": 0.629, + "step": 10063 + }, + { + "epoch": 0.39, + "learning_rate": 4.3579122852013336e-05, + "loss": 0.6112, + "step": 10064 + }, + { + "epoch": 0.39, + "learning_rate": 4.357848166196461e-05, + "loss": 0.5262, + "step": 10065 + }, + { + "epoch": 0.39, + "learning_rate": 4.357784047191588e-05, + "loss": 0.6984, + "step": 10066 + }, + { + "epoch": 0.39, + "learning_rate": 4.357719928186715e-05, + "loss": 0.6167, + "step": 10067 + }, + { + "epoch": 0.39, + "learning_rate": 4.3576558091818416e-05, + "loss": 0.4805, + "step": 10068 + }, + { + "epoch": 0.39, + "learning_rate": 4.357591690176969e-05, + "loss": 0.6047, + "step": 10069 + }, + { + "epoch": 0.39, + "learning_rate": 4.357527571172096e-05, + "loss": 0.5667, + "step": 10070 + }, + { + "epoch": 0.39, + "learning_rate": 4.3574634521672225e-05, + "loss": 0.4967, + "step": 10071 + }, + { + "epoch": 0.39, + "learning_rate": 4.3573993331623496e-05, + "loss": 0.5115, + "step": 10072 + }, + { + "epoch": 0.39, + "learning_rate": 4.357335214157476e-05, + "loss": 0.5243, + "step": 10073 + }, + { + "epoch": 0.39, + "learning_rate": 4.357271095152603e-05, + "loss": 0.5791, + "step": 10074 + }, + { + "epoch": 0.39, + "learning_rate": 4.3572069761477305e-05, + "loss": 0.5658, + "step": 10075 + }, + { + "epoch": 0.39, + "learning_rate": 4.3571428571428576e-05, + "loss": 0.5207, + "step": 10076 + }, + { + "epoch": 0.39, + "learning_rate": 4.357078738137984e-05, + "loss": 0.714, + "step": 10077 + }, + { + "epoch": 0.39, + "learning_rate": 4.357014619133111e-05, + "loss": 0.6105, + "step": 10078 + }, + { + "epoch": 0.39, + "learning_rate": 4.3569505001282385e-05, + "loss": 0.5822, + "step": 10079 + }, + { + "epoch": 0.39, + "learning_rate": 4.356886381123365e-05, + "loss": 0.5518, + "step": 10080 + }, + { + "epoch": 0.39, + "learning_rate": 4.356822262118492e-05, + "loss": 0.5486, + "step": 10081 + }, + { + "epoch": 0.39, + "learning_rate": 4.3567581431136186e-05, + "loss": 0.5987, + "step": 10082 + }, + { + "epoch": 0.39, + "learning_rate": 4.3566940241087465e-05, + "loss": 0.5249, + "step": 10083 + }, + { + "epoch": 0.39, + "learning_rate": 4.356629905103873e-05, + "loss": 0.5996, + "step": 10084 + }, + { + "epoch": 0.39, + "learning_rate": 4.356565786099e-05, + "loss": 0.6106, + "step": 10085 + }, + { + "epoch": 0.39, + "learning_rate": 4.3565016670941266e-05, + "loss": 0.5323, + "step": 10086 + }, + { + "epoch": 0.39, + "learning_rate": 4.356437548089254e-05, + "loss": 0.4846, + "step": 10087 + }, + { + "epoch": 0.39, + "learning_rate": 4.356373429084381e-05, + "loss": 0.6395, + "step": 10088 + }, + { + "epoch": 0.39, + "learning_rate": 4.3563093100795075e-05, + "loss": 0.547, + "step": 10089 + }, + { + "epoch": 0.39, + "learning_rate": 4.3562451910746346e-05, + "loss": 0.5616, + "step": 10090 + }, + { + "epoch": 0.39, + "learning_rate": 4.356181072069762e-05, + "loss": 0.4647, + "step": 10091 + }, + { + "epoch": 0.39, + "learning_rate": 4.356116953064889e-05, + "loss": 0.56, + "step": 10092 + }, + { + "epoch": 0.39, + "learning_rate": 4.3560528340600155e-05, + "loss": 0.6717, + "step": 10093 + }, + { + "epoch": 0.39, + "learning_rate": 4.3559887150551427e-05, + "loss": 0.5408, + "step": 10094 + }, + { + "epoch": 0.39, + "learning_rate": 4.355924596050269e-05, + "loss": 0.6168, + "step": 10095 + }, + { + "epoch": 0.39, + "learning_rate": 4.355860477045396e-05, + "loss": 0.5657, + "step": 10096 + }, + { + "epoch": 0.39, + "learning_rate": 4.3557963580405235e-05, + "loss": 0.7164, + "step": 10097 + }, + { + "epoch": 0.39, + "learning_rate": 4.35573223903565e-05, + "loss": 0.5314, + "step": 10098 + }, + { + "epoch": 0.39, + "learning_rate": 4.355668120030778e-05, + "loss": 0.474, + "step": 10099 + }, + { + "epoch": 0.39, + "learning_rate": 4.355604001025904e-05, + "loss": 0.4791, + "step": 10100 + }, + { + "epoch": 0.39, + "learning_rate": 4.3555398820210315e-05, + "loss": 0.5636, + "step": 10101 + }, + { + "epoch": 0.39, + "learning_rate": 4.355475763016158e-05, + "loss": 0.5727, + "step": 10102 + }, + { + "epoch": 0.39, + "learning_rate": 4.355411644011285e-05, + "loss": 0.642, + "step": 10103 + }, + { + "epoch": 0.39, + "learning_rate": 4.355347525006412e-05, + "loss": 0.545, + "step": 10104 + }, + { + "epoch": 0.39, + "learning_rate": 4.355283406001539e-05, + "loss": 0.488, + "step": 10105 + }, + { + "epoch": 0.39, + "learning_rate": 4.355219286996666e-05, + "loss": 0.5627, + "step": 10106 + }, + { + "epoch": 0.39, + "learning_rate": 4.355155167991793e-05, + "loss": 0.5668, + "step": 10107 + }, + { + "epoch": 0.39, + "learning_rate": 4.3550910489869203e-05, + "loss": 0.5278, + "step": 10108 + }, + { + "epoch": 0.39, + "learning_rate": 4.355026929982047e-05, + "loss": 0.6097, + "step": 10109 + }, + { + "epoch": 0.39, + "learning_rate": 4.354962810977174e-05, + "loss": 0.6293, + "step": 10110 + }, + { + "epoch": 0.39, + "learning_rate": 4.3548986919723005e-05, + "loss": 0.6425, + "step": 10111 + }, + { + "epoch": 0.39, + "learning_rate": 4.354834572967428e-05, + "loss": 0.4955, + "step": 10112 + }, + { + "epoch": 0.39, + "learning_rate": 4.354770453962554e-05, + "loss": 0.619, + "step": 10113 + }, + { + "epoch": 0.39, + "learning_rate": 4.3547063349576813e-05, + "loss": 0.5793, + "step": 10114 + }, + { + "epoch": 0.39, + "learning_rate": 4.3546422159528085e-05, + "loss": 0.5887, + "step": 10115 + }, + { + "epoch": 0.39, + "learning_rate": 4.354578096947936e-05, + "loss": 0.6903, + "step": 10116 + }, + { + "epoch": 0.39, + "learning_rate": 4.354513977943063e-05, + "loss": 0.5472, + "step": 10117 + }, + { + "epoch": 0.39, + "learning_rate": 4.3544498589381894e-05, + "loss": 0.5473, + "step": 10118 + }, + { + "epoch": 0.39, + "learning_rate": 4.3543857399333165e-05, + "loss": 0.4781, + "step": 10119 + }, + { + "epoch": 0.39, + "learning_rate": 4.354321620928443e-05, + "loss": 0.6349, + "step": 10120 + }, + { + "epoch": 0.39, + "learning_rate": 4.35425750192357e-05, + "loss": 0.5057, + "step": 10121 + }, + { + "epoch": 0.39, + "learning_rate": 4.3541933829186974e-05, + "loss": 0.5237, + "step": 10122 + }, + { + "epoch": 0.39, + "learning_rate": 4.3541292639138245e-05, + "loss": 0.5825, + "step": 10123 + }, + { + "epoch": 0.39, + "learning_rate": 4.354065144908952e-05, + "loss": 0.6909, + "step": 10124 + }, + { + "epoch": 0.39, + "learning_rate": 4.354001025904078e-05, + "loss": 0.4715, + "step": 10125 + }, + { + "epoch": 0.39, + "learning_rate": 4.3539369068992054e-05, + "loss": 0.5941, + "step": 10126 + }, + { + "epoch": 0.39, + "learning_rate": 4.353872787894332e-05, + "loss": 0.6281, + "step": 10127 + }, + { + "epoch": 0.39, + "learning_rate": 4.353808668889459e-05, + "loss": 0.5967, + "step": 10128 + }, + { + "epoch": 0.39, + "learning_rate": 4.3537445498845855e-05, + "loss": 0.4914, + "step": 10129 + }, + { + "epoch": 0.39, + "learning_rate": 4.353680430879713e-05, + "loss": 0.5548, + "step": 10130 + }, + { + "epoch": 0.39, + "learning_rate": 4.35361631187484e-05, + "loss": 0.5453, + "step": 10131 + }, + { + "epoch": 0.39, + "learning_rate": 4.353552192869967e-05, + "loss": 0.6145, + "step": 10132 + }, + { + "epoch": 0.39, + "learning_rate": 4.353488073865094e-05, + "loss": 0.6212, + "step": 10133 + }, + { + "epoch": 0.39, + "learning_rate": 4.353423954860221e-05, + "loss": 0.5083, + "step": 10134 + }, + { + "epoch": 0.39, + "learning_rate": 4.353359835855348e-05, + "loss": 0.5074, + "step": 10135 + }, + { + "epoch": 0.39, + "learning_rate": 4.3532957168504744e-05, + "loss": 0.5634, + "step": 10136 + }, + { + "epoch": 0.39, + "learning_rate": 4.3532315978456016e-05, + "loss": 0.5156, + "step": 10137 + }, + { + "epoch": 0.39, + "learning_rate": 4.353167478840729e-05, + "loss": 0.5557, + "step": 10138 + }, + { + "epoch": 0.39, + "learning_rate": 4.353103359835856e-05, + "loss": 0.5007, + "step": 10139 + }, + { + "epoch": 0.39, + "learning_rate": 4.3530392408309824e-05, + "loss": 0.4673, + "step": 10140 + }, + { + "epoch": 0.39, + "learning_rate": 4.3529751218261096e-05, + "loss": 0.4671, + "step": 10141 + }, + { + "epoch": 0.39, + "learning_rate": 4.352911002821237e-05, + "loss": 0.5526, + "step": 10142 + }, + { + "epoch": 0.39, + "learning_rate": 4.352846883816363e-05, + "loss": 0.5158, + "step": 10143 + }, + { + "epoch": 0.39, + "learning_rate": 4.3527827648114904e-05, + "loss": 0.5074, + "step": 10144 + }, + { + "epoch": 0.39, + "learning_rate": 4.352718645806617e-05, + "loss": 0.5107, + "step": 10145 + }, + { + "epoch": 0.39, + "learning_rate": 4.352654526801744e-05, + "loss": 0.5604, + "step": 10146 + }, + { + "epoch": 0.39, + "learning_rate": 4.352590407796871e-05, + "loss": 0.5327, + "step": 10147 + }, + { + "epoch": 0.39, + "learning_rate": 4.3525262887919984e-05, + "loss": 0.4741, + "step": 10148 + }, + { + "epoch": 0.39, + "learning_rate": 4.352462169787125e-05, + "loss": 0.6109, + "step": 10149 + }, + { + "epoch": 0.39, + "learning_rate": 4.352398050782252e-05, + "loss": 0.5522, + "step": 10150 + }, + { + "epoch": 0.39, + "learning_rate": 4.352333931777379e-05, + "loss": 0.5601, + "step": 10151 + }, + { + "epoch": 0.39, + "learning_rate": 4.352269812772506e-05, + "loss": 0.6388, + "step": 10152 + }, + { + "epoch": 0.39, + "learning_rate": 4.352205693767633e-05, + "loss": 0.571, + "step": 10153 + }, + { + "epoch": 0.39, + "learning_rate": 4.35214157476276e-05, + "loss": 0.5753, + "step": 10154 + }, + { + "epoch": 0.39, + "learning_rate": 4.352077455757887e-05, + "loss": 0.5289, + "step": 10155 + }, + { + "epoch": 0.39, + "learning_rate": 4.352013336753014e-05, + "loss": 0.5507, + "step": 10156 + }, + { + "epoch": 0.39, + "learning_rate": 4.351949217748141e-05, + "loss": 0.475, + "step": 10157 + }, + { + "epoch": 0.39, + "learning_rate": 4.3518850987432674e-05, + "loss": 0.5782, + "step": 10158 + }, + { + "epoch": 0.39, + "learning_rate": 4.3518209797383946e-05, + "loss": 0.5657, + "step": 10159 + }, + { + "epoch": 0.39, + "learning_rate": 4.351756860733522e-05, + "loss": 0.5212, + "step": 10160 + }, + { + "epoch": 0.39, + "learning_rate": 4.351692741728648e-05, + "loss": 0.5278, + "step": 10161 + }, + { + "epoch": 0.39, + "learning_rate": 4.3516286227237754e-05, + "loss": 0.5832, + "step": 10162 + }, + { + "epoch": 0.39, + "learning_rate": 4.3515645037189026e-05, + "loss": 0.5286, + "step": 10163 + }, + { + "epoch": 0.39, + "learning_rate": 4.35150038471403e-05, + "loss": 0.5349, + "step": 10164 + }, + { + "epoch": 0.39, + "learning_rate": 4.351436265709156e-05, + "loss": 0.5533, + "step": 10165 + }, + { + "epoch": 0.39, + "learning_rate": 4.3513721467042834e-05, + "loss": 0.6156, + "step": 10166 + }, + { + "epoch": 0.39, + "learning_rate": 4.35130802769941e-05, + "loss": 0.5825, + "step": 10167 + }, + { + "epoch": 0.39, + "learning_rate": 4.351243908694537e-05, + "loss": 0.5927, + "step": 10168 + }, + { + "epoch": 0.39, + "learning_rate": 4.351179789689664e-05, + "loss": 0.5857, + "step": 10169 + }, + { + "epoch": 0.39, + "learning_rate": 4.351115670684791e-05, + "loss": 0.5197, + "step": 10170 + }, + { + "epoch": 0.39, + "learning_rate": 4.3510515516799186e-05, + "loss": 0.5889, + "step": 10171 + }, + { + "epoch": 0.39, + "learning_rate": 4.350987432675045e-05, + "loss": 0.6776, + "step": 10172 + }, + { + "epoch": 0.39, + "learning_rate": 4.350923313670172e-05, + "loss": 0.5768, + "step": 10173 + }, + { + "epoch": 0.39, + "learning_rate": 4.350859194665299e-05, + "loss": 0.5819, + "step": 10174 + }, + { + "epoch": 0.39, + "learning_rate": 4.350795075660426e-05, + "loss": 0.6189, + "step": 10175 + }, + { + "epoch": 0.39, + "learning_rate": 4.3507309566555524e-05, + "loss": 0.5417, + "step": 10176 + }, + { + "epoch": 0.39, + "learning_rate": 4.3506668376506796e-05, + "loss": 0.5617, + "step": 10177 + }, + { + "epoch": 0.39, + "learning_rate": 4.350602718645807e-05, + "loss": 0.5596, + "step": 10178 + }, + { + "epoch": 0.39, + "learning_rate": 4.350538599640934e-05, + "loss": 0.5121, + "step": 10179 + }, + { + "epoch": 0.39, + "learning_rate": 4.350474480636061e-05, + "loss": 0.6033, + "step": 10180 + }, + { + "epoch": 0.39, + "learning_rate": 4.3504103616311876e-05, + "loss": 0.6129, + "step": 10181 + }, + { + "epoch": 0.39, + "learning_rate": 4.350346242626315e-05, + "loss": 0.5303, + "step": 10182 + }, + { + "epoch": 0.39, + "learning_rate": 4.350282123621441e-05, + "loss": 0.6165, + "step": 10183 + }, + { + "epoch": 0.39, + "learning_rate": 4.3502180046165685e-05, + "loss": 0.7298, + "step": 10184 + }, + { + "epoch": 0.39, + "learning_rate": 4.3501538856116956e-05, + "loss": 0.5379, + "step": 10185 + }, + { + "epoch": 0.39, + "learning_rate": 4.350089766606822e-05, + "loss": 0.5507, + "step": 10186 + }, + { + "epoch": 0.39, + "learning_rate": 4.35002564760195e-05, + "loss": 0.5293, + "step": 10187 + }, + { + "epoch": 0.39, + "learning_rate": 4.3499615285970765e-05, + "loss": 0.5822, + "step": 10188 + }, + { + "epoch": 0.39, + "learning_rate": 4.3498974095922036e-05, + "loss": 0.6623, + "step": 10189 + }, + { + "epoch": 0.39, + "learning_rate": 4.34983329058733e-05, + "loss": 0.5689, + "step": 10190 + }, + { + "epoch": 0.39, + "learning_rate": 4.349769171582457e-05, + "loss": 0.6399, + "step": 10191 + }, + { + "epoch": 0.39, + "learning_rate": 4.349705052577584e-05, + "loss": 0.6041, + "step": 10192 + }, + { + "epoch": 0.39, + "learning_rate": 4.349640933572711e-05, + "loss": 0.5366, + "step": 10193 + }, + { + "epoch": 0.39, + "learning_rate": 4.349576814567838e-05, + "loss": 0.5405, + "step": 10194 + }, + { + "epoch": 0.39, + "learning_rate": 4.349512695562965e-05, + "loss": 0.6163, + "step": 10195 + }, + { + "epoch": 0.39, + "learning_rate": 4.3494485765580925e-05, + "loss": 0.6626, + "step": 10196 + }, + { + "epoch": 0.39, + "learning_rate": 4.349384457553219e-05, + "loss": 0.5265, + "step": 10197 + }, + { + "epoch": 0.39, + "learning_rate": 4.349320338548346e-05, + "loss": 0.514, + "step": 10198 + }, + { + "epoch": 0.39, + "learning_rate": 4.3492562195434726e-05, + "loss": 0.5085, + "step": 10199 + }, + { + "epoch": 0.39, + "learning_rate": 4.3491921005386e-05, + "loss": 0.5413, + "step": 10200 + }, + { + "epoch": 0.39, + "learning_rate": 4.349127981533726e-05, + "loss": 0.5422, + "step": 10201 + }, + { + "epoch": 0.39, + "learning_rate": 4.3490638625288535e-05, + "loss": 0.5224, + "step": 10202 + }, + { + "epoch": 0.39, + "learning_rate": 4.3489997435239807e-05, + "loss": 0.6299, + "step": 10203 + }, + { + "epoch": 0.39, + "learning_rate": 4.348935624519108e-05, + "loss": 0.5933, + "step": 10204 + }, + { + "epoch": 0.39, + "learning_rate": 4.348871505514235e-05, + "loss": 0.4935, + "step": 10205 + }, + { + "epoch": 0.39, + "learning_rate": 4.3488073865093615e-05, + "loss": 0.5321, + "step": 10206 + }, + { + "epoch": 0.39, + "learning_rate": 4.3487432675044887e-05, + "loss": 0.5247, + "step": 10207 + }, + { + "epoch": 0.39, + "learning_rate": 4.348679148499615e-05, + "loss": 0.704, + "step": 10208 + }, + { + "epoch": 0.39, + "learning_rate": 4.348615029494742e-05, + "loss": 0.5528, + "step": 10209 + }, + { + "epoch": 0.39, + "learning_rate": 4.3485509104898695e-05, + "loss": 0.5048, + "step": 10210 + }, + { + "epoch": 0.39, + "learning_rate": 4.348486791484997e-05, + "loss": 0.6288, + "step": 10211 + }, + { + "epoch": 0.39, + "learning_rate": 4.348422672480124e-05, + "loss": 0.5054, + "step": 10212 + }, + { + "epoch": 0.39, + "learning_rate": 4.34835855347525e-05, + "loss": 0.5775, + "step": 10213 + }, + { + "epoch": 0.39, + "learning_rate": 4.3482944344703775e-05, + "loss": 0.5177, + "step": 10214 + }, + { + "epoch": 0.39, + "learning_rate": 4.348230315465504e-05, + "loss": 0.5254, + "step": 10215 + }, + { + "epoch": 0.39, + "learning_rate": 4.348166196460631e-05, + "loss": 0.4985, + "step": 10216 + }, + { + "epoch": 0.39, + "learning_rate": 4.348102077455758e-05, + "loss": 0.5218, + "step": 10217 + }, + { + "epoch": 0.39, + "learning_rate": 4.348037958450885e-05, + "loss": 0.5792, + "step": 10218 + }, + { + "epoch": 0.39, + "learning_rate": 4.347973839446012e-05, + "loss": 0.5437, + "step": 10219 + }, + { + "epoch": 0.39, + "learning_rate": 4.347909720441139e-05, + "loss": 0.4755, + "step": 10220 + }, + { + "epoch": 0.39, + "learning_rate": 4.3478456014362664e-05, + "loss": 0.6426, + "step": 10221 + }, + { + "epoch": 0.39, + "learning_rate": 4.347781482431393e-05, + "loss": 0.5283, + "step": 10222 + }, + { + "epoch": 0.39, + "learning_rate": 4.34771736342652e-05, + "loss": 0.6087, + "step": 10223 + }, + { + "epoch": 0.39, + "learning_rate": 4.3476532444216465e-05, + "loss": 0.5288, + "step": 10224 + }, + { + "epoch": 0.39, + "learning_rate": 4.347589125416774e-05, + "loss": 0.5563, + "step": 10225 + }, + { + "epoch": 0.39, + "learning_rate": 4.347525006411901e-05, + "loss": 0.4884, + "step": 10226 + }, + { + "epoch": 0.39, + "learning_rate": 4.347460887407028e-05, + "loss": 0.6693, + "step": 10227 + }, + { + "epoch": 0.39, + "learning_rate": 4.3473967684021545e-05, + "loss": 0.5464, + "step": 10228 + }, + { + "epoch": 0.39, + "learning_rate": 4.347332649397282e-05, + "loss": 0.5245, + "step": 10229 + }, + { + "epoch": 0.39, + "learning_rate": 4.347268530392409e-05, + "loss": 0.5692, + "step": 10230 + }, + { + "epoch": 0.39, + "learning_rate": 4.3472044113875354e-05, + "loss": 0.5342, + "step": 10231 + }, + { + "epoch": 0.39, + "learning_rate": 4.3471402923826625e-05, + "loss": 0.5913, + "step": 10232 + }, + { + "epoch": 0.39, + "learning_rate": 4.347076173377789e-05, + "loss": 0.4886, + "step": 10233 + }, + { + "epoch": 0.39, + "learning_rate": 4.347012054372916e-05, + "loss": 0.5406, + "step": 10234 + }, + { + "epoch": 0.39, + "learning_rate": 4.3469479353680434e-05, + "loss": 0.5325, + "step": 10235 + }, + { + "epoch": 0.39, + "learning_rate": 4.3468838163631705e-05, + "loss": 0.6788, + "step": 10236 + }, + { + "epoch": 0.39, + "learning_rate": 4.346819697358297e-05, + "loss": 0.5487, + "step": 10237 + }, + { + "epoch": 0.39, + "learning_rate": 4.346755578353424e-05, + "loss": 0.5292, + "step": 10238 + }, + { + "epoch": 0.39, + "learning_rate": 4.3466914593485514e-05, + "loss": 0.661, + "step": 10239 + }, + { + "epoch": 0.39, + "learning_rate": 4.346627340343678e-05, + "loss": 0.6276, + "step": 10240 + }, + { + "epoch": 0.39, + "learning_rate": 4.346563221338805e-05, + "loss": 0.6402, + "step": 10241 + }, + { + "epoch": 0.39, + "learning_rate": 4.3464991023339315e-05, + "loss": 0.6873, + "step": 10242 + }, + { + "epoch": 0.39, + "learning_rate": 4.3464349833290594e-05, + "loss": 0.5639, + "step": 10243 + }, + { + "epoch": 0.39, + "learning_rate": 4.346370864324186e-05, + "loss": 0.5433, + "step": 10244 + }, + { + "epoch": 0.39, + "learning_rate": 4.346306745319313e-05, + "loss": 0.5044, + "step": 10245 + }, + { + "epoch": 0.39, + "learning_rate": 4.3462426263144395e-05, + "loss": 0.5151, + "step": 10246 + }, + { + "epoch": 0.39, + "learning_rate": 4.346178507309567e-05, + "loss": 0.586, + "step": 10247 + }, + { + "epoch": 0.39, + "learning_rate": 4.346114388304694e-05, + "loss": 0.6336, + "step": 10248 + }, + { + "epoch": 0.39, + "learning_rate": 4.3460502692998204e-05, + "loss": 0.5436, + "step": 10249 + }, + { + "epoch": 0.39, + "learning_rate": 4.3459861502949476e-05, + "loss": 0.5236, + "step": 10250 + }, + { + "epoch": 0.39, + "learning_rate": 4.345922031290075e-05, + "loss": 0.5922, + "step": 10251 + }, + { + "epoch": 0.39, + "learning_rate": 4.345857912285202e-05, + "loss": 0.5752, + "step": 10252 + }, + { + "epoch": 0.39, + "learning_rate": 4.3457937932803284e-05, + "loss": 0.6306, + "step": 10253 + }, + { + "epoch": 0.39, + "learning_rate": 4.3457296742754556e-05, + "loss": 0.4414, + "step": 10254 + }, + { + "epoch": 0.39, + "learning_rate": 4.345665555270582e-05, + "loss": 0.5246, + "step": 10255 + }, + { + "epoch": 0.39, + "learning_rate": 4.345601436265709e-05, + "loss": 0.5915, + "step": 10256 + }, + { + "epoch": 0.39, + "learning_rate": 4.3455373172608364e-05, + "loss": 0.5436, + "step": 10257 + }, + { + "epoch": 0.39, + "learning_rate": 4.345473198255963e-05, + "loss": 0.6562, + "step": 10258 + }, + { + "epoch": 0.39, + "learning_rate": 4.345409079251091e-05, + "loss": 0.6972, + "step": 10259 + }, + { + "epoch": 0.39, + "learning_rate": 4.345344960246217e-05, + "loss": 0.558, + "step": 10260 + }, + { + "epoch": 0.39, + "learning_rate": 4.3452808412413444e-05, + "loss": 0.4778, + "step": 10261 + }, + { + "epoch": 0.39, + "learning_rate": 4.345216722236471e-05, + "loss": 0.6044, + "step": 10262 + }, + { + "epoch": 0.39, + "learning_rate": 4.345152603231598e-05, + "loss": 0.5998, + "step": 10263 + }, + { + "epoch": 0.39, + "learning_rate": 4.3450884842267246e-05, + "loss": 0.5027, + "step": 10264 + }, + { + "epoch": 0.39, + "learning_rate": 4.345024365221852e-05, + "loss": 0.6212, + "step": 10265 + }, + { + "epoch": 0.39, + "learning_rate": 4.344960246216979e-05, + "loss": 0.5845, + "step": 10266 + }, + { + "epoch": 0.39, + "learning_rate": 4.344896127212106e-05, + "loss": 0.5846, + "step": 10267 + }, + { + "epoch": 0.39, + "learning_rate": 4.344832008207233e-05, + "loss": 0.5162, + "step": 10268 + }, + { + "epoch": 0.39, + "learning_rate": 4.34476788920236e-05, + "loss": 0.5182, + "step": 10269 + }, + { + "epoch": 0.39, + "learning_rate": 4.344703770197487e-05, + "loss": 0.5729, + "step": 10270 + }, + { + "epoch": 0.39, + "learning_rate": 4.3446396511926134e-05, + "loss": 0.6318, + "step": 10271 + }, + { + "epoch": 0.39, + "learning_rate": 4.3445755321877406e-05, + "loss": 0.637, + "step": 10272 + }, + { + "epoch": 0.39, + "learning_rate": 4.344511413182867e-05, + "loss": 0.5848, + "step": 10273 + }, + { + "epoch": 0.4, + "learning_rate": 4.344447294177994e-05, + "loss": 0.5955, + "step": 10274 + }, + { + "epoch": 0.4, + "learning_rate": 4.344383175173122e-05, + "loss": 0.5268, + "step": 10275 + }, + { + "epoch": 0.4, + "learning_rate": 4.3443190561682486e-05, + "loss": 0.5509, + "step": 10276 + }, + { + "epoch": 0.4, + "learning_rate": 4.344254937163376e-05, + "loss": 0.5859, + "step": 10277 + }, + { + "epoch": 0.4, + "learning_rate": 4.344190818158502e-05, + "loss": 0.6051, + "step": 10278 + }, + { + "epoch": 0.4, + "learning_rate": 4.3441266991536294e-05, + "loss": 0.5962, + "step": 10279 + }, + { + "epoch": 0.4, + "learning_rate": 4.344062580148756e-05, + "loss": 0.5147, + "step": 10280 + }, + { + "epoch": 0.4, + "learning_rate": 4.343998461143883e-05, + "loss": 0.5518, + "step": 10281 + }, + { + "epoch": 0.4, + "learning_rate": 4.34393434213901e-05, + "loss": 0.6561, + "step": 10282 + }, + { + "epoch": 0.4, + "learning_rate": 4.3438702231341374e-05, + "loss": 0.573, + "step": 10283 + }, + { + "epoch": 0.4, + "learning_rate": 4.3438061041292646e-05, + "loss": 0.5577, + "step": 10284 + }, + { + "epoch": 0.4, + "learning_rate": 4.343741985124391e-05, + "loss": 0.5199, + "step": 10285 + }, + { + "epoch": 0.4, + "learning_rate": 4.343677866119518e-05, + "loss": 0.4917, + "step": 10286 + }, + { + "epoch": 0.4, + "learning_rate": 4.343613747114645e-05, + "loss": 0.6768, + "step": 10287 + }, + { + "epoch": 0.4, + "learning_rate": 4.343549628109772e-05, + "loss": 0.5096, + "step": 10288 + }, + { + "epoch": 0.4, + "learning_rate": 4.3434855091048984e-05, + "loss": 0.496, + "step": 10289 + }, + { + "epoch": 0.4, + "learning_rate": 4.3434213901000256e-05, + "loss": 0.5168, + "step": 10290 + }, + { + "epoch": 0.4, + "learning_rate": 4.343357271095153e-05, + "loss": 0.6787, + "step": 10291 + }, + { + "epoch": 0.4, + "learning_rate": 4.34329315209028e-05, + "loss": 0.6202, + "step": 10292 + }, + { + "epoch": 0.4, + "learning_rate": 4.343229033085407e-05, + "loss": 0.5736, + "step": 10293 + }, + { + "epoch": 0.4, + "learning_rate": 4.3431649140805336e-05, + "loss": 0.5821, + "step": 10294 + }, + { + "epoch": 0.4, + "learning_rate": 4.343100795075661e-05, + "loss": 0.4534, + "step": 10295 + }, + { + "epoch": 0.4, + "learning_rate": 4.343036676070787e-05, + "loss": 0.6192, + "step": 10296 + }, + { + "epoch": 0.4, + "learning_rate": 4.3429725570659145e-05, + "loss": 0.5561, + "step": 10297 + }, + { + "epoch": 0.4, + "learning_rate": 4.3429084380610416e-05, + "loss": 0.627, + "step": 10298 + }, + { + "epoch": 0.4, + "learning_rate": 4.342844319056169e-05, + "loss": 0.6432, + "step": 10299 + }, + { + "epoch": 0.4, + "learning_rate": 4.342780200051295e-05, + "loss": 0.4945, + "step": 10300 + }, + { + "epoch": 0.4, + "learning_rate": 4.3427160810464225e-05, + "loss": 0.5987, + "step": 10301 + }, + { + "epoch": 0.4, + "learning_rate": 4.3426519620415496e-05, + "loss": 0.6375, + "step": 10302 + }, + { + "epoch": 0.4, + "learning_rate": 4.342587843036676e-05, + "loss": 0.6277, + "step": 10303 + }, + { + "epoch": 0.4, + "learning_rate": 4.342523724031803e-05, + "loss": 0.5564, + "step": 10304 + }, + { + "epoch": 0.4, + "learning_rate": 4.34245960502693e-05, + "loss": 0.5371, + "step": 10305 + }, + { + "epoch": 0.4, + "learning_rate": 4.342395486022057e-05, + "loss": 0.6502, + "step": 10306 + }, + { + "epoch": 0.4, + "learning_rate": 4.342331367017184e-05, + "loss": 0.5181, + "step": 10307 + }, + { + "epoch": 0.4, + "learning_rate": 4.342267248012311e-05, + "loss": 0.6306, + "step": 10308 + }, + { + "epoch": 0.4, + "learning_rate": 4.342203129007438e-05, + "loss": 0.6441, + "step": 10309 + }, + { + "epoch": 0.4, + "learning_rate": 4.342139010002565e-05, + "loss": 0.5465, + "step": 10310 + }, + { + "epoch": 0.4, + "learning_rate": 4.342074890997692e-05, + "loss": 0.5618, + "step": 10311 + }, + { + "epoch": 0.4, + "learning_rate": 4.3420107719928186e-05, + "loss": 0.5921, + "step": 10312 + }, + { + "epoch": 0.4, + "learning_rate": 4.341946652987946e-05, + "loss": 0.5548, + "step": 10313 + }, + { + "epoch": 0.4, + "learning_rate": 4.341882533983073e-05, + "loss": 0.5602, + "step": 10314 + }, + { + "epoch": 0.4, + "learning_rate": 4.3418184149782e-05, + "loss": 0.5857, + "step": 10315 + }, + { + "epoch": 0.4, + "learning_rate": 4.3417542959733267e-05, + "loss": 0.538, + "step": 10316 + }, + { + "epoch": 0.4, + "learning_rate": 4.341690176968454e-05, + "loss": 0.5908, + "step": 10317 + }, + { + "epoch": 0.4, + "learning_rate": 4.34162605796358e-05, + "loss": 0.584, + "step": 10318 + }, + { + "epoch": 0.4, + "learning_rate": 4.3415619389587075e-05, + "loss": 0.6275, + "step": 10319 + }, + { + "epoch": 0.4, + "learning_rate": 4.341497819953835e-05, + "loss": 0.5527, + "step": 10320 + }, + { + "epoch": 0.4, + "learning_rate": 4.341433700948961e-05, + "loss": 0.5576, + "step": 10321 + }, + { + "epoch": 0.4, + "learning_rate": 4.341369581944088e-05, + "loss": 0.5492, + "step": 10322 + }, + { + "epoch": 0.4, + "learning_rate": 4.3413054629392155e-05, + "loss": 0.5506, + "step": 10323 + }, + { + "epoch": 0.4, + "learning_rate": 4.341241343934343e-05, + "loss": 0.6107, + "step": 10324 + }, + { + "epoch": 0.4, + "learning_rate": 4.341177224929469e-05, + "loss": 0.6101, + "step": 10325 + }, + { + "epoch": 0.4, + "learning_rate": 4.341113105924596e-05, + "loss": 0.6107, + "step": 10326 + }, + { + "epoch": 0.4, + "learning_rate": 4.341048986919723e-05, + "loss": 0.5902, + "step": 10327 + }, + { + "epoch": 0.4, + "learning_rate": 4.34098486791485e-05, + "loss": 0.7695, + "step": 10328 + }, + { + "epoch": 0.4, + "learning_rate": 4.340920748909977e-05, + "loss": 0.5286, + "step": 10329 + }, + { + "epoch": 0.4, + "learning_rate": 4.340856629905104e-05, + "loss": 0.514, + "step": 10330 + }, + { + "epoch": 0.4, + "learning_rate": 4.3407925109002315e-05, + "loss": 0.6081, + "step": 10331 + }, + { + "epoch": 0.4, + "learning_rate": 4.340728391895358e-05, + "loss": 0.5639, + "step": 10332 + }, + { + "epoch": 0.4, + "learning_rate": 4.340664272890485e-05, + "loss": 0.5353, + "step": 10333 + }, + { + "epoch": 0.4, + "learning_rate": 4.340600153885612e-05, + "loss": 0.5593, + "step": 10334 + }, + { + "epoch": 0.4, + "learning_rate": 4.340536034880739e-05, + "loss": 0.5689, + "step": 10335 + }, + { + "epoch": 0.4, + "learning_rate": 4.3404719158758653e-05, + "loss": 0.6367, + "step": 10336 + }, + { + "epoch": 0.4, + "learning_rate": 4.3404077968709925e-05, + "loss": 0.5552, + "step": 10337 + }, + { + "epoch": 0.4, + "learning_rate": 4.34034367786612e-05, + "loss": 0.5547, + "step": 10338 + }, + { + "epoch": 0.4, + "learning_rate": 4.340279558861247e-05, + "loss": 0.5864, + "step": 10339 + }, + { + "epoch": 0.4, + "learning_rate": 4.340215439856374e-05, + "loss": 0.6654, + "step": 10340 + }, + { + "epoch": 0.4, + "learning_rate": 4.3401513208515005e-05, + "loss": 0.598, + "step": 10341 + }, + { + "epoch": 0.4, + "learning_rate": 4.340087201846628e-05, + "loss": 0.5392, + "step": 10342 + }, + { + "epoch": 0.4, + "learning_rate": 4.340023082841754e-05, + "loss": 0.5358, + "step": 10343 + }, + { + "epoch": 0.4, + "learning_rate": 4.3399589638368814e-05, + "loss": 0.5311, + "step": 10344 + }, + { + "epoch": 0.4, + "learning_rate": 4.339894844832008e-05, + "loss": 0.6803, + "step": 10345 + }, + { + "epoch": 0.4, + "learning_rate": 4.339830725827135e-05, + "loss": 0.6673, + "step": 10346 + }, + { + "epoch": 0.4, + "learning_rate": 4.339766606822263e-05, + "loss": 0.5792, + "step": 10347 + }, + { + "epoch": 0.4, + "learning_rate": 4.3397024878173894e-05, + "loss": 0.6677, + "step": 10348 + }, + { + "epoch": 0.4, + "learning_rate": 4.3396383688125165e-05, + "loss": 0.5015, + "step": 10349 + }, + { + "epoch": 0.4, + "learning_rate": 4.339574249807643e-05, + "loss": 0.5425, + "step": 10350 + }, + { + "epoch": 0.4, + "learning_rate": 4.33951013080277e-05, + "loss": 0.6295, + "step": 10351 + }, + { + "epoch": 0.4, + "learning_rate": 4.339446011797897e-05, + "loss": 0.6489, + "step": 10352 + }, + { + "epoch": 0.4, + "learning_rate": 4.339381892793024e-05, + "loss": 0.6226, + "step": 10353 + }, + { + "epoch": 0.4, + "learning_rate": 4.339317773788151e-05, + "loss": 0.604, + "step": 10354 + }, + { + "epoch": 0.4, + "learning_rate": 4.339253654783278e-05, + "loss": 0.4802, + "step": 10355 + }, + { + "epoch": 0.4, + "learning_rate": 4.3391895357784054e-05, + "loss": 0.656, + "step": 10356 + }, + { + "epoch": 0.4, + "learning_rate": 4.339125416773532e-05, + "loss": 0.5636, + "step": 10357 + }, + { + "epoch": 0.4, + "learning_rate": 4.339061297768659e-05, + "loss": 0.5482, + "step": 10358 + }, + { + "epoch": 0.4, + "learning_rate": 4.3389971787637855e-05, + "loss": 0.5563, + "step": 10359 + }, + { + "epoch": 0.4, + "learning_rate": 4.338933059758913e-05, + "loss": 0.5445, + "step": 10360 + }, + { + "epoch": 0.4, + "learning_rate": 4.338868940754039e-05, + "loss": 0.579, + "step": 10361 + }, + { + "epoch": 0.4, + "learning_rate": 4.3388048217491664e-05, + "loss": 0.5559, + "step": 10362 + }, + { + "epoch": 0.4, + "learning_rate": 4.3387407027442936e-05, + "loss": 0.683, + "step": 10363 + }, + { + "epoch": 0.4, + "learning_rate": 4.338676583739421e-05, + "loss": 0.499, + "step": 10364 + }, + { + "epoch": 0.4, + "learning_rate": 4.338612464734548e-05, + "loss": 0.567, + "step": 10365 + }, + { + "epoch": 0.4, + "learning_rate": 4.3385483457296744e-05, + "loss": 0.5242, + "step": 10366 + }, + { + "epoch": 0.4, + "learning_rate": 4.3384842267248016e-05, + "loss": 0.576, + "step": 10367 + }, + { + "epoch": 0.4, + "learning_rate": 4.338420107719928e-05, + "loss": 0.5902, + "step": 10368 + }, + { + "epoch": 0.4, + "learning_rate": 4.338355988715055e-05, + "loss": 0.6493, + "step": 10369 + }, + { + "epoch": 0.4, + "learning_rate": 4.3382918697101824e-05, + "loss": 0.5179, + "step": 10370 + }, + { + "epoch": 0.4, + "learning_rate": 4.3382277507053096e-05, + "loss": 0.6596, + "step": 10371 + }, + { + "epoch": 0.4, + "learning_rate": 4.338163631700437e-05, + "loss": 0.5208, + "step": 10372 + }, + { + "epoch": 0.4, + "learning_rate": 4.338099512695563e-05, + "loss": 0.5538, + "step": 10373 + }, + { + "epoch": 0.4, + "learning_rate": 4.3380353936906904e-05, + "loss": 0.6228, + "step": 10374 + }, + { + "epoch": 0.4, + "learning_rate": 4.337971274685817e-05, + "loss": 0.5153, + "step": 10375 + }, + { + "epoch": 0.4, + "learning_rate": 4.337907155680944e-05, + "loss": 0.6286, + "step": 10376 + }, + { + "epoch": 0.4, + "learning_rate": 4.3378430366760706e-05, + "loss": 0.6348, + "step": 10377 + }, + { + "epoch": 0.4, + "learning_rate": 4.337778917671198e-05, + "loss": 0.5032, + "step": 10378 + }, + { + "epoch": 0.4, + "learning_rate": 4.337714798666325e-05, + "loss": 0.5548, + "step": 10379 + }, + { + "epoch": 0.4, + "learning_rate": 4.337650679661452e-05, + "loss": 0.4541, + "step": 10380 + }, + { + "epoch": 0.4, + "learning_rate": 4.337586560656579e-05, + "loss": 0.5632, + "step": 10381 + }, + { + "epoch": 0.4, + "learning_rate": 4.337522441651706e-05, + "loss": 0.5326, + "step": 10382 + }, + { + "epoch": 0.4, + "learning_rate": 4.337458322646833e-05, + "loss": 0.5637, + "step": 10383 + }, + { + "epoch": 0.4, + "learning_rate": 4.3373942036419594e-05, + "loss": 0.5621, + "step": 10384 + }, + { + "epoch": 0.4, + "learning_rate": 4.3373300846370866e-05, + "loss": 0.4965, + "step": 10385 + }, + { + "epoch": 0.4, + "learning_rate": 4.337265965632214e-05, + "loss": 0.5517, + "step": 10386 + }, + { + "epoch": 0.4, + "learning_rate": 4.337201846627341e-05, + "loss": 0.5298, + "step": 10387 + }, + { + "epoch": 0.4, + "learning_rate": 4.3371377276224674e-05, + "loss": 0.4818, + "step": 10388 + }, + { + "epoch": 0.4, + "learning_rate": 4.3370736086175946e-05, + "loss": 0.5238, + "step": 10389 + }, + { + "epoch": 0.4, + "learning_rate": 4.337009489612722e-05, + "loss": 0.5174, + "step": 10390 + }, + { + "epoch": 0.4, + "learning_rate": 4.336945370607848e-05, + "loss": 0.558, + "step": 10391 + }, + { + "epoch": 0.4, + "learning_rate": 4.3368812516029754e-05, + "loss": 0.722, + "step": 10392 + }, + { + "epoch": 0.4, + "learning_rate": 4.336817132598102e-05, + "loss": 0.5301, + "step": 10393 + }, + { + "epoch": 0.4, + "learning_rate": 4.336753013593229e-05, + "loss": 0.5421, + "step": 10394 + }, + { + "epoch": 0.4, + "learning_rate": 4.336688894588356e-05, + "loss": 0.4854, + "step": 10395 + }, + { + "epoch": 0.4, + "learning_rate": 4.3366247755834834e-05, + "loss": 0.6562, + "step": 10396 + }, + { + "epoch": 0.4, + "learning_rate": 4.33656065657861e-05, + "loss": 0.5835, + "step": 10397 + }, + { + "epoch": 0.4, + "learning_rate": 4.336496537573737e-05, + "loss": 0.6251, + "step": 10398 + }, + { + "epoch": 0.4, + "learning_rate": 4.336432418568864e-05, + "loss": 0.4841, + "step": 10399 + }, + { + "epoch": 0.4, + "learning_rate": 4.336368299563991e-05, + "loss": 0.5249, + "step": 10400 + }, + { + "epoch": 0.4, + "learning_rate": 4.336304180559118e-05, + "loss": 0.5812, + "step": 10401 + }, + { + "epoch": 0.4, + "learning_rate": 4.3362400615542444e-05, + "loss": 0.5432, + "step": 10402 + }, + { + "epoch": 0.4, + "learning_rate": 4.336175942549372e-05, + "loss": 0.4581, + "step": 10403 + }, + { + "epoch": 0.4, + "learning_rate": 4.336111823544499e-05, + "loss": 0.5919, + "step": 10404 + }, + { + "epoch": 0.4, + "learning_rate": 4.336047704539626e-05, + "loss": 0.6182, + "step": 10405 + }, + { + "epoch": 0.4, + "learning_rate": 4.3359835855347525e-05, + "loss": 0.5192, + "step": 10406 + }, + { + "epoch": 0.4, + "learning_rate": 4.3359194665298796e-05, + "loss": 0.5798, + "step": 10407 + }, + { + "epoch": 0.4, + "learning_rate": 4.335855347525007e-05, + "loss": 0.5309, + "step": 10408 + }, + { + "epoch": 0.4, + "learning_rate": 4.335791228520133e-05, + "loss": 0.6301, + "step": 10409 + }, + { + "epoch": 0.4, + "learning_rate": 4.3357271095152605e-05, + "loss": 0.5896, + "step": 10410 + }, + { + "epoch": 0.4, + "learning_rate": 4.3356629905103876e-05, + "loss": 0.6159, + "step": 10411 + }, + { + "epoch": 0.4, + "learning_rate": 4.335598871505515e-05, + "loss": 0.5349, + "step": 10412 + }, + { + "epoch": 0.4, + "learning_rate": 4.335534752500641e-05, + "loss": 0.4984, + "step": 10413 + }, + { + "epoch": 0.4, + "learning_rate": 4.3354706334957685e-05, + "loss": 0.4742, + "step": 10414 + }, + { + "epoch": 0.4, + "learning_rate": 4.335406514490895e-05, + "loss": 0.5663, + "step": 10415 + }, + { + "epoch": 0.4, + "learning_rate": 4.335342395486022e-05, + "loss": 0.5923, + "step": 10416 + }, + { + "epoch": 0.4, + "learning_rate": 4.335278276481149e-05, + "loss": 0.6354, + "step": 10417 + }, + { + "epoch": 0.4, + "learning_rate": 4.335214157476276e-05, + "loss": 0.5914, + "step": 10418 + }, + { + "epoch": 0.4, + "learning_rate": 4.3351500384714037e-05, + "loss": 0.574, + "step": 10419 + }, + { + "epoch": 0.4, + "learning_rate": 4.33508591946653e-05, + "loss": 0.5309, + "step": 10420 + }, + { + "epoch": 0.4, + "learning_rate": 4.335021800461657e-05, + "loss": 0.5565, + "step": 10421 + }, + { + "epoch": 0.4, + "learning_rate": 4.334957681456784e-05, + "loss": 0.6937, + "step": 10422 + }, + { + "epoch": 0.4, + "learning_rate": 4.334893562451911e-05, + "loss": 0.6174, + "step": 10423 + }, + { + "epoch": 0.4, + "learning_rate": 4.3348294434470375e-05, + "loss": 0.5713, + "step": 10424 + }, + { + "epoch": 0.4, + "learning_rate": 4.3347653244421646e-05, + "loss": 0.5886, + "step": 10425 + }, + { + "epoch": 0.4, + "learning_rate": 4.334701205437292e-05, + "loss": 0.6153, + "step": 10426 + }, + { + "epoch": 0.4, + "learning_rate": 4.334637086432419e-05, + "loss": 0.5072, + "step": 10427 + }, + { + "epoch": 0.4, + "learning_rate": 4.334572967427546e-05, + "loss": 0.4938, + "step": 10428 + }, + { + "epoch": 0.4, + "learning_rate": 4.3345088484226727e-05, + "loss": 0.5888, + "step": 10429 + }, + { + "epoch": 0.4, + "learning_rate": 4.3344447294178e-05, + "loss": 0.4895, + "step": 10430 + }, + { + "epoch": 0.4, + "learning_rate": 4.334380610412926e-05, + "loss": 0.5673, + "step": 10431 + }, + { + "epoch": 0.4, + "learning_rate": 4.3343164914080535e-05, + "loss": 0.516, + "step": 10432 + }, + { + "epoch": 0.4, + "learning_rate": 4.33425237240318e-05, + "loss": 0.4615, + "step": 10433 + }, + { + "epoch": 0.4, + "learning_rate": 4.334188253398307e-05, + "loss": 0.5841, + "step": 10434 + }, + { + "epoch": 0.4, + "learning_rate": 4.334124134393435e-05, + "loss": 0.5523, + "step": 10435 + }, + { + "epoch": 0.4, + "learning_rate": 4.3340600153885615e-05, + "loss": 0.5755, + "step": 10436 + }, + { + "epoch": 0.4, + "learning_rate": 4.333995896383689e-05, + "loss": 0.5269, + "step": 10437 + }, + { + "epoch": 0.4, + "learning_rate": 4.333931777378815e-05, + "loss": 0.5521, + "step": 10438 + }, + { + "epoch": 0.4, + "learning_rate": 4.3338676583739423e-05, + "loss": 0.5114, + "step": 10439 + }, + { + "epoch": 0.4, + "learning_rate": 4.333803539369069e-05, + "loss": 0.459, + "step": 10440 + }, + { + "epoch": 0.4, + "learning_rate": 4.333739420364196e-05, + "loss": 0.626, + "step": 10441 + }, + { + "epoch": 0.4, + "learning_rate": 4.333675301359323e-05, + "loss": 0.6119, + "step": 10442 + }, + { + "epoch": 0.4, + "learning_rate": 4.3336111823544504e-05, + "loss": 0.57, + "step": 10443 + }, + { + "epoch": 0.4, + "learning_rate": 4.3335470633495775e-05, + "loss": 0.6022, + "step": 10444 + }, + { + "epoch": 0.4, + "learning_rate": 4.333482944344704e-05, + "loss": 0.4927, + "step": 10445 + }, + { + "epoch": 0.4, + "learning_rate": 4.333418825339831e-05, + "loss": 0.634, + "step": 10446 + }, + { + "epoch": 0.4, + "learning_rate": 4.333354706334958e-05, + "loss": 0.5756, + "step": 10447 + }, + { + "epoch": 0.4, + "learning_rate": 4.333290587330085e-05, + "loss": 0.6134, + "step": 10448 + }, + { + "epoch": 0.4, + "learning_rate": 4.3332264683252113e-05, + "loss": 0.6082, + "step": 10449 + }, + { + "epoch": 0.4, + "learning_rate": 4.3331623493203385e-05, + "loss": 0.538, + "step": 10450 + }, + { + "epoch": 0.4, + "learning_rate": 4.333098230315466e-05, + "loss": 0.5896, + "step": 10451 + }, + { + "epoch": 0.4, + "learning_rate": 4.333034111310593e-05, + "loss": 0.4359, + "step": 10452 + }, + { + "epoch": 0.4, + "learning_rate": 4.33296999230572e-05, + "loss": 0.5013, + "step": 10453 + }, + { + "epoch": 0.4, + "learning_rate": 4.3329058733008465e-05, + "loss": 0.5999, + "step": 10454 + }, + { + "epoch": 0.4, + "learning_rate": 4.332841754295974e-05, + "loss": 0.5093, + "step": 10455 + }, + { + "epoch": 0.4, + "learning_rate": 4.3327776352911e-05, + "loss": 0.5128, + "step": 10456 + }, + { + "epoch": 0.4, + "learning_rate": 4.3327135162862274e-05, + "loss": 0.6372, + "step": 10457 + }, + { + "epoch": 0.4, + "learning_rate": 4.3326493972813545e-05, + "loss": 0.5556, + "step": 10458 + }, + { + "epoch": 0.4, + "learning_rate": 4.332585278276482e-05, + "loss": 0.5746, + "step": 10459 + }, + { + "epoch": 0.4, + "learning_rate": 4.332521159271608e-05, + "loss": 0.632, + "step": 10460 + }, + { + "epoch": 0.4, + "learning_rate": 4.3324570402667354e-05, + "loss": 0.4639, + "step": 10461 + }, + { + "epoch": 0.4, + "learning_rate": 4.3323929212618625e-05, + "loss": 0.5415, + "step": 10462 + }, + { + "epoch": 0.4, + "learning_rate": 4.332328802256989e-05, + "loss": 0.5699, + "step": 10463 + }, + { + "epoch": 0.4, + "learning_rate": 4.332264683252116e-05, + "loss": 0.494, + "step": 10464 + }, + { + "epoch": 0.4, + "learning_rate": 4.332200564247243e-05, + "loss": 0.4798, + "step": 10465 + }, + { + "epoch": 0.4, + "learning_rate": 4.33213644524237e-05, + "loss": 0.5071, + "step": 10466 + }, + { + "epoch": 0.4, + "learning_rate": 4.332072326237497e-05, + "loss": 0.6004, + "step": 10467 + }, + { + "epoch": 0.4, + "learning_rate": 4.332008207232624e-05, + "loss": 0.6362, + "step": 10468 + }, + { + "epoch": 0.4, + "learning_rate": 4.331944088227751e-05, + "loss": 0.5825, + "step": 10469 + }, + { + "epoch": 0.4, + "learning_rate": 4.331879969222878e-05, + "loss": 0.5839, + "step": 10470 + }, + { + "epoch": 0.4, + "learning_rate": 4.331815850218005e-05, + "loss": 0.5479, + "step": 10471 + }, + { + "epoch": 0.4, + "learning_rate": 4.3317517312131316e-05, + "loss": 0.5734, + "step": 10472 + }, + { + "epoch": 0.4, + "learning_rate": 4.331687612208259e-05, + "loss": 0.6236, + "step": 10473 + }, + { + "epoch": 0.4, + "learning_rate": 4.331623493203386e-05, + "loss": 0.5635, + "step": 10474 + }, + { + "epoch": 0.4, + "learning_rate": 4.331559374198513e-05, + "loss": 0.5622, + "step": 10475 + }, + { + "epoch": 0.4, + "learning_rate": 4.3314952551936396e-05, + "loss": 0.5443, + "step": 10476 + }, + { + "epoch": 0.4, + "learning_rate": 4.331431136188767e-05, + "loss": 0.5671, + "step": 10477 + }, + { + "epoch": 0.4, + "learning_rate": 4.331367017183893e-05, + "loss": 0.5522, + "step": 10478 + }, + { + "epoch": 0.4, + "learning_rate": 4.3313028981790204e-05, + "loss": 0.607, + "step": 10479 + }, + { + "epoch": 0.4, + "learning_rate": 4.3312387791741476e-05, + "loss": 0.545, + "step": 10480 + }, + { + "epoch": 0.4, + "learning_rate": 4.331174660169274e-05, + "loss": 0.6437, + "step": 10481 + }, + { + "epoch": 0.4, + "learning_rate": 4.331110541164401e-05, + "loss": 0.5374, + "step": 10482 + }, + { + "epoch": 0.4, + "learning_rate": 4.3310464221595284e-05, + "loss": 0.5748, + "step": 10483 + }, + { + "epoch": 0.4, + "learning_rate": 4.3309823031546556e-05, + "loss": 0.6666, + "step": 10484 + }, + { + "epoch": 0.4, + "learning_rate": 4.330918184149782e-05, + "loss": 0.6265, + "step": 10485 + }, + { + "epoch": 0.4, + "learning_rate": 4.330854065144909e-05, + "loss": 0.575, + "step": 10486 + }, + { + "epoch": 0.4, + "learning_rate": 4.330789946140036e-05, + "loss": 0.5783, + "step": 10487 + }, + { + "epoch": 0.4, + "learning_rate": 4.330725827135163e-05, + "loss": 0.6074, + "step": 10488 + }, + { + "epoch": 0.4, + "learning_rate": 4.33066170813029e-05, + "loss": 0.579, + "step": 10489 + }, + { + "epoch": 0.4, + "learning_rate": 4.3305975891254166e-05, + "loss": 0.5904, + "step": 10490 + }, + { + "epoch": 0.4, + "learning_rate": 4.3305334701205444e-05, + "loss": 0.4889, + "step": 10491 + }, + { + "epoch": 0.4, + "learning_rate": 4.330469351115671e-05, + "loss": 0.67, + "step": 10492 + }, + { + "epoch": 0.4, + "learning_rate": 4.330405232110798e-05, + "loss": 0.5282, + "step": 10493 + }, + { + "epoch": 0.4, + "learning_rate": 4.3303411131059246e-05, + "loss": 0.5395, + "step": 10494 + }, + { + "epoch": 0.4, + "learning_rate": 4.330276994101052e-05, + "loss": 0.6308, + "step": 10495 + }, + { + "epoch": 0.4, + "learning_rate": 4.330212875096178e-05, + "loss": 0.4893, + "step": 10496 + }, + { + "epoch": 0.4, + "learning_rate": 4.3301487560913054e-05, + "loss": 0.554, + "step": 10497 + }, + { + "epoch": 0.4, + "learning_rate": 4.3300846370864326e-05, + "loss": 0.5055, + "step": 10498 + }, + { + "epoch": 0.4, + "learning_rate": 4.33002051808156e-05, + "loss": 0.5326, + "step": 10499 + }, + { + "epoch": 0.4, + "learning_rate": 4.329956399076687e-05, + "loss": 0.5948, + "step": 10500 + }, + { + "epoch": 0.4, + "learning_rate": 4.3298922800718134e-05, + "loss": 0.5544, + "step": 10501 + }, + { + "epoch": 0.4, + "learning_rate": 4.3298281610669406e-05, + "loss": 0.5386, + "step": 10502 + }, + { + "epoch": 0.4, + "learning_rate": 4.329764042062067e-05, + "loss": 0.5081, + "step": 10503 + }, + { + "epoch": 0.4, + "learning_rate": 4.329699923057194e-05, + "loss": 0.569, + "step": 10504 + }, + { + "epoch": 0.4, + "learning_rate": 4.329635804052321e-05, + "loss": 0.5992, + "step": 10505 + }, + { + "epoch": 0.4, + "learning_rate": 4.329571685047448e-05, + "loss": 0.5428, + "step": 10506 + }, + { + "epoch": 0.4, + "learning_rate": 4.329507566042576e-05, + "loss": 0.5792, + "step": 10507 + }, + { + "epoch": 0.4, + "learning_rate": 4.329443447037702e-05, + "loss": 0.5572, + "step": 10508 + }, + { + "epoch": 0.4, + "learning_rate": 4.3293793280328294e-05, + "loss": 0.5675, + "step": 10509 + }, + { + "epoch": 0.4, + "learning_rate": 4.329315209027956e-05, + "loss": 0.5471, + "step": 10510 + }, + { + "epoch": 0.4, + "learning_rate": 4.329251090023083e-05, + "loss": 0.5919, + "step": 10511 + }, + { + "epoch": 0.4, + "learning_rate": 4.3291869710182096e-05, + "loss": 0.5497, + "step": 10512 + }, + { + "epoch": 0.4, + "learning_rate": 4.329122852013337e-05, + "loss": 0.5285, + "step": 10513 + }, + { + "epoch": 0.4, + "learning_rate": 4.329058733008464e-05, + "loss": 0.5629, + "step": 10514 + }, + { + "epoch": 0.4, + "learning_rate": 4.328994614003591e-05, + "loss": 0.5425, + "step": 10515 + }, + { + "epoch": 0.4, + "learning_rate": 4.328930494998718e-05, + "loss": 0.6192, + "step": 10516 + }, + { + "epoch": 0.4, + "learning_rate": 4.328866375993845e-05, + "loss": 0.6157, + "step": 10517 + }, + { + "epoch": 0.4, + "learning_rate": 4.328802256988972e-05, + "loss": 0.5117, + "step": 10518 + }, + { + "epoch": 0.4, + "learning_rate": 4.3287381379840985e-05, + "loss": 0.5179, + "step": 10519 + }, + { + "epoch": 0.4, + "learning_rate": 4.3286740189792256e-05, + "loss": 0.4572, + "step": 10520 + }, + { + "epoch": 0.4, + "learning_rate": 4.328609899974352e-05, + "loss": 0.4202, + "step": 10521 + }, + { + "epoch": 0.4, + "learning_rate": 4.328545780969479e-05, + "loss": 0.6019, + "step": 10522 + }, + { + "epoch": 0.4, + "learning_rate": 4.3284816619646065e-05, + "loss": 0.6175, + "step": 10523 + }, + { + "epoch": 0.4, + "learning_rate": 4.3284175429597336e-05, + "loss": 0.4934, + "step": 10524 + }, + { + "epoch": 0.4, + "learning_rate": 4.328353423954861e-05, + "loss": 0.5089, + "step": 10525 + }, + { + "epoch": 0.4, + "learning_rate": 4.328289304949987e-05, + "loss": 0.4699, + "step": 10526 + }, + { + "epoch": 0.4, + "learning_rate": 4.3282251859451145e-05, + "loss": 0.4837, + "step": 10527 + }, + { + "epoch": 0.4, + "learning_rate": 4.328161066940241e-05, + "loss": 0.5892, + "step": 10528 + }, + { + "epoch": 0.4, + "learning_rate": 4.328096947935368e-05, + "loss": 0.573, + "step": 10529 + }, + { + "epoch": 0.4, + "learning_rate": 4.328032828930495e-05, + "loss": 0.5592, + "step": 10530 + }, + { + "epoch": 0.4, + "learning_rate": 4.3279687099256225e-05, + "loss": 0.6189, + "step": 10531 + }, + { + "epoch": 0.4, + "learning_rate": 4.327904590920749e-05, + "loss": 0.5323, + "step": 10532 + }, + { + "epoch": 0.4, + "learning_rate": 4.327840471915876e-05, + "loss": 0.465, + "step": 10533 + }, + { + "epoch": 0.4, + "learning_rate": 4.327776352911003e-05, + "loss": 0.487, + "step": 10534 + }, + { + "epoch": 0.41, + "learning_rate": 4.32771223390613e-05, + "loss": 0.543, + "step": 10535 + }, + { + "epoch": 0.41, + "learning_rate": 4.327648114901257e-05, + "loss": 0.636, + "step": 10536 + }, + { + "epoch": 0.41, + "learning_rate": 4.3275839958963835e-05, + "loss": 0.4408, + "step": 10537 + }, + { + "epoch": 0.41, + "learning_rate": 4.3275198768915107e-05, + "loss": 0.5124, + "step": 10538 + }, + { + "epoch": 0.41, + "learning_rate": 4.327455757886638e-05, + "loss": 0.645, + "step": 10539 + }, + { + "epoch": 0.41, + "learning_rate": 4.327391638881765e-05, + "loss": 0.5104, + "step": 10540 + }, + { + "epoch": 0.41, + "learning_rate": 4.3273275198768915e-05, + "loss": 0.5175, + "step": 10541 + }, + { + "epoch": 0.41, + "learning_rate": 4.3272634008720187e-05, + "loss": 0.5298, + "step": 10542 + }, + { + "epoch": 0.41, + "learning_rate": 4.327199281867146e-05, + "loss": 0.5426, + "step": 10543 + }, + { + "epoch": 0.41, + "learning_rate": 4.327135162862272e-05, + "loss": 0.5951, + "step": 10544 + }, + { + "epoch": 0.41, + "learning_rate": 4.3270710438573995e-05, + "loss": 0.5582, + "step": 10545 + }, + { + "epoch": 0.41, + "learning_rate": 4.327006924852527e-05, + "loss": 0.5305, + "step": 10546 + }, + { + "epoch": 0.41, + "learning_rate": 4.326942805847654e-05, + "loss": 0.6086, + "step": 10547 + }, + { + "epoch": 0.41, + "learning_rate": 4.32687868684278e-05, + "loss": 0.6285, + "step": 10548 + }, + { + "epoch": 0.41, + "learning_rate": 4.3268145678379075e-05, + "loss": 0.5243, + "step": 10549 + }, + { + "epoch": 0.41, + "learning_rate": 4.326750448833034e-05, + "loss": 0.4714, + "step": 10550 + }, + { + "epoch": 0.41, + "learning_rate": 4.326686329828161e-05, + "loss": 0.5631, + "step": 10551 + }, + { + "epoch": 0.41, + "learning_rate": 4.3266222108232883e-05, + "loss": 0.5136, + "step": 10552 + }, + { + "epoch": 0.41, + "learning_rate": 4.326558091818415e-05, + "loss": 0.5383, + "step": 10553 + }, + { + "epoch": 0.41, + "learning_rate": 4.326493972813542e-05, + "loss": 0.5671, + "step": 10554 + }, + { + "epoch": 0.41, + "learning_rate": 4.326429853808669e-05, + "loss": 0.5071, + "step": 10555 + }, + { + "epoch": 0.41, + "learning_rate": 4.3263657348037964e-05, + "loss": 0.5311, + "step": 10556 + }, + { + "epoch": 0.41, + "learning_rate": 4.326301615798923e-05, + "loss": 0.5461, + "step": 10557 + }, + { + "epoch": 0.41, + "learning_rate": 4.32623749679405e-05, + "loss": 0.5715, + "step": 10558 + }, + { + "epoch": 0.41, + "learning_rate": 4.326173377789177e-05, + "loss": 0.5531, + "step": 10559 + }, + { + "epoch": 0.41, + "learning_rate": 4.326109258784304e-05, + "loss": 0.5349, + "step": 10560 + }, + { + "epoch": 0.41, + "learning_rate": 4.326045139779431e-05, + "loss": 0.6403, + "step": 10561 + }, + { + "epoch": 0.41, + "learning_rate": 4.325981020774558e-05, + "loss": 0.6287, + "step": 10562 + }, + { + "epoch": 0.41, + "learning_rate": 4.325916901769685e-05, + "loss": 0.6118, + "step": 10563 + }, + { + "epoch": 0.41, + "learning_rate": 4.325852782764812e-05, + "loss": 0.506, + "step": 10564 + }, + { + "epoch": 0.41, + "learning_rate": 4.325788663759939e-05, + "loss": 0.5973, + "step": 10565 + }, + { + "epoch": 0.41, + "learning_rate": 4.3257245447550654e-05, + "loss": 0.5592, + "step": 10566 + }, + { + "epoch": 0.41, + "learning_rate": 4.3256604257501925e-05, + "loss": 0.6387, + "step": 10567 + }, + { + "epoch": 0.41, + "learning_rate": 4.32559630674532e-05, + "loss": 0.5231, + "step": 10568 + }, + { + "epoch": 0.41, + "learning_rate": 4.325532187740446e-05, + "loss": 0.5729, + "step": 10569 + }, + { + "epoch": 0.41, + "learning_rate": 4.3254680687355734e-05, + "loss": 0.6019, + "step": 10570 + }, + { + "epoch": 0.41, + "learning_rate": 4.3254039497307005e-05, + "loss": 0.6298, + "step": 10571 + }, + { + "epoch": 0.41, + "learning_rate": 4.325339830725828e-05, + "loss": 0.5212, + "step": 10572 + }, + { + "epoch": 0.41, + "learning_rate": 4.325275711720954e-05, + "loss": 0.4963, + "step": 10573 + }, + { + "epoch": 0.41, + "learning_rate": 4.3252115927160814e-05, + "loss": 0.5448, + "step": 10574 + }, + { + "epoch": 0.41, + "learning_rate": 4.325147473711208e-05, + "loss": 0.5351, + "step": 10575 + }, + { + "epoch": 0.41, + "learning_rate": 4.325083354706335e-05, + "loss": 0.5616, + "step": 10576 + }, + { + "epoch": 0.41, + "learning_rate": 4.325019235701462e-05, + "loss": 0.6294, + "step": 10577 + }, + { + "epoch": 0.41, + "learning_rate": 4.324955116696589e-05, + "loss": 0.5972, + "step": 10578 + }, + { + "epoch": 0.41, + "learning_rate": 4.3248909976917166e-05, + "loss": 0.5468, + "step": 10579 + }, + { + "epoch": 0.41, + "learning_rate": 4.324826878686843e-05, + "loss": 0.5382, + "step": 10580 + }, + { + "epoch": 0.41, + "learning_rate": 4.32476275968197e-05, + "loss": 0.4965, + "step": 10581 + }, + { + "epoch": 0.41, + "learning_rate": 4.324698640677097e-05, + "loss": 0.6207, + "step": 10582 + }, + { + "epoch": 0.41, + "learning_rate": 4.324634521672224e-05, + "loss": 0.5824, + "step": 10583 + }, + { + "epoch": 0.41, + "learning_rate": 4.3245704026673504e-05, + "loss": 0.5456, + "step": 10584 + }, + { + "epoch": 0.41, + "learning_rate": 4.3245062836624776e-05, + "loss": 0.5344, + "step": 10585 + }, + { + "epoch": 0.41, + "learning_rate": 4.324442164657605e-05, + "loss": 0.5985, + "step": 10586 + }, + { + "epoch": 0.41, + "learning_rate": 4.324378045652732e-05, + "loss": 0.5676, + "step": 10587 + }, + { + "epoch": 0.41, + "learning_rate": 4.324313926647859e-05, + "loss": 0.557, + "step": 10588 + }, + { + "epoch": 0.41, + "learning_rate": 4.3242498076429856e-05, + "loss": 0.603, + "step": 10589 + }, + { + "epoch": 0.41, + "learning_rate": 4.324185688638113e-05, + "loss": 0.5438, + "step": 10590 + }, + { + "epoch": 0.41, + "learning_rate": 4.324121569633239e-05, + "loss": 0.5963, + "step": 10591 + }, + { + "epoch": 0.41, + "learning_rate": 4.3240574506283664e-05, + "loss": 0.5673, + "step": 10592 + }, + { + "epoch": 0.41, + "learning_rate": 4.323993331623493e-05, + "loss": 0.5945, + "step": 10593 + }, + { + "epoch": 0.41, + "learning_rate": 4.32392921261862e-05, + "loss": 0.5322, + "step": 10594 + }, + { + "epoch": 0.41, + "learning_rate": 4.323865093613748e-05, + "loss": 0.6162, + "step": 10595 + }, + { + "epoch": 0.41, + "learning_rate": 4.3238009746088744e-05, + "loss": 0.6469, + "step": 10596 + }, + { + "epoch": 0.41, + "learning_rate": 4.3237368556040016e-05, + "loss": 0.5726, + "step": 10597 + }, + { + "epoch": 0.41, + "learning_rate": 4.323672736599128e-05, + "loss": 0.4812, + "step": 10598 + }, + { + "epoch": 0.41, + "learning_rate": 4.323608617594255e-05, + "loss": 0.5429, + "step": 10599 + }, + { + "epoch": 0.41, + "learning_rate": 4.323544498589382e-05, + "loss": 0.5773, + "step": 10600 + }, + { + "epoch": 0.41, + "learning_rate": 4.323480379584509e-05, + "loss": 0.6242, + "step": 10601 + }, + { + "epoch": 0.41, + "learning_rate": 4.323416260579636e-05, + "loss": 0.5887, + "step": 10602 + }, + { + "epoch": 0.41, + "learning_rate": 4.323352141574763e-05, + "loss": 0.5651, + "step": 10603 + }, + { + "epoch": 0.41, + "learning_rate": 4.3232880225698904e-05, + "loss": 0.5369, + "step": 10604 + }, + { + "epoch": 0.41, + "learning_rate": 4.323223903565017e-05, + "loss": 0.5521, + "step": 10605 + }, + { + "epoch": 0.41, + "learning_rate": 4.323159784560144e-05, + "loss": 0.5404, + "step": 10606 + }, + { + "epoch": 0.41, + "learning_rate": 4.3230956655552706e-05, + "loss": 0.5614, + "step": 10607 + }, + { + "epoch": 0.41, + "learning_rate": 4.323031546550398e-05, + "loss": 0.5627, + "step": 10608 + }, + { + "epoch": 0.41, + "learning_rate": 4.322967427545524e-05, + "loss": 0.6952, + "step": 10609 + }, + { + "epoch": 0.41, + "learning_rate": 4.3229033085406514e-05, + "loss": 0.5354, + "step": 10610 + }, + { + "epoch": 0.41, + "learning_rate": 4.3228391895357786e-05, + "loss": 0.468, + "step": 10611 + }, + { + "epoch": 0.41, + "learning_rate": 4.322775070530906e-05, + "loss": 0.6299, + "step": 10612 + }, + { + "epoch": 0.41, + "learning_rate": 4.322710951526033e-05, + "loss": 0.5773, + "step": 10613 + }, + { + "epoch": 0.41, + "learning_rate": 4.3226468325211594e-05, + "loss": 0.6349, + "step": 10614 + }, + { + "epoch": 0.41, + "learning_rate": 4.3225827135162866e-05, + "loss": 0.5098, + "step": 10615 + }, + { + "epoch": 0.41, + "learning_rate": 4.322518594511413e-05, + "loss": 0.51, + "step": 10616 + }, + { + "epoch": 0.41, + "learning_rate": 4.32245447550654e-05, + "loss": 0.5029, + "step": 10617 + }, + { + "epoch": 0.41, + "learning_rate": 4.3223903565016674e-05, + "loss": 0.5604, + "step": 10618 + }, + { + "epoch": 0.41, + "learning_rate": 4.3223262374967946e-05, + "loss": 0.6018, + "step": 10619 + }, + { + "epoch": 0.41, + "learning_rate": 4.322262118491921e-05, + "loss": 0.5437, + "step": 10620 + }, + { + "epoch": 0.41, + "learning_rate": 4.322197999487048e-05, + "loss": 0.4634, + "step": 10621 + }, + { + "epoch": 0.41, + "learning_rate": 4.3221338804821755e-05, + "loss": 0.6021, + "step": 10622 + }, + { + "epoch": 0.41, + "learning_rate": 4.322069761477302e-05, + "loss": 0.5789, + "step": 10623 + }, + { + "epoch": 0.41, + "learning_rate": 4.322005642472429e-05, + "loss": 0.5348, + "step": 10624 + }, + { + "epoch": 0.41, + "learning_rate": 4.3219415234675556e-05, + "loss": 0.4879, + "step": 10625 + }, + { + "epoch": 0.41, + "learning_rate": 4.321877404462683e-05, + "loss": 0.5904, + "step": 10626 + }, + { + "epoch": 0.41, + "learning_rate": 4.32181328545781e-05, + "loss": 0.6495, + "step": 10627 + }, + { + "epoch": 0.41, + "learning_rate": 4.321749166452937e-05, + "loss": 0.6355, + "step": 10628 + }, + { + "epoch": 0.41, + "learning_rate": 4.3216850474480636e-05, + "loss": 0.555, + "step": 10629 + }, + { + "epoch": 0.41, + "learning_rate": 4.321620928443191e-05, + "loss": 0.5594, + "step": 10630 + }, + { + "epoch": 0.41, + "learning_rate": 4.321556809438318e-05, + "loss": 0.5524, + "step": 10631 + }, + { + "epoch": 0.41, + "learning_rate": 4.3214926904334445e-05, + "loss": 0.586, + "step": 10632 + }, + { + "epoch": 0.41, + "learning_rate": 4.3214285714285716e-05, + "loss": 0.5796, + "step": 10633 + }, + { + "epoch": 0.41, + "learning_rate": 4.321364452423699e-05, + "loss": 0.564, + "step": 10634 + }, + { + "epoch": 0.41, + "learning_rate": 4.321300333418826e-05, + "loss": 0.4877, + "step": 10635 + }, + { + "epoch": 0.41, + "learning_rate": 4.3212362144139525e-05, + "loss": 0.5309, + "step": 10636 + }, + { + "epoch": 0.41, + "learning_rate": 4.3211720954090796e-05, + "loss": 0.5692, + "step": 10637 + }, + { + "epoch": 0.41, + "learning_rate": 4.321107976404206e-05, + "loss": 0.5416, + "step": 10638 + }, + { + "epoch": 0.41, + "learning_rate": 4.321043857399333e-05, + "loss": 0.4386, + "step": 10639 + }, + { + "epoch": 0.41, + "learning_rate": 4.3209797383944605e-05, + "loss": 0.6329, + "step": 10640 + }, + { + "epoch": 0.41, + "learning_rate": 4.320915619389587e-05, + "loss": 0.5668, + "step": 10641 + }, + { + "epoch": 0.41, + "learning_rate": 4.320851500384714e-05, + "loss": 0.5828, + "step": 10642 + }, + { + "epoch": 0.41, + "learning_rate": 4.320787381379841e-05, + "loss": 0.5407, + "step": 10643 + }, + { + "epoch": 0.41, + "learning_rate": 4.3207232623749685e-05, + "loss": 0.6059, + "step": 10644 + }, + { + "epoch": 0.41, + "learning_rate": 4.320659143370095e-05, + "loss": 0.4572, + "step": 10645 + }, + { + "epoch": 0.41, + "learning_rate": 4.320595024365222e-05, + "loss": 0.4968, + "step": 10646 + }, + { + "epoch": 0.41, + "learning_rate": 4.3205309053603486e-05, + "loss": 0.5006, + "step": 10647 + }, + { + "epoch": 0.41, + "learning_rate": 4.320466786355476e-05, + "loss": 0.627, + "step": 10648 + }, + { + "epoch": 0.41, + "learning_rate": 4.320402667350603e-05, + "loss": 0.5166, + "step": 10649 + }, + { + "epoch": 0.41, + "learning_rate": 4.3203385483457295e-05, + "loss": 0.5497, + "step": 10650 + }, + { + "epoch": 0.41, + "learning_rate": 4.320274429340857e-05, + "loss": 0.6141, + "step": 10651 + }, + { + "epoch": 0.41, + "learning_rate": 4.320210310335984e-05, + "loss": 0.5275, + "step": 10652 + }, + { + "epoch": 0.41, + "learning_rate": 4.320146191331111e-05, + "loss": 0.5333, + "step": 10653 + }, + { + "epoch": 0.41, + "learning_rate": 4.3200820723262375e-05, + "loss": 0.5552, + "step": 10654 + }, + { + "epoch": 0.41, + "learning_rate": 4.320017953321365e-05, + "loss": 0.5327, + "step": 10655 + }, + { + "epoch": 0.41, + "learning_rate": 4.319953834316491e-05, + "loss": 0.6137, + "step": 10656 + }, + { + "epoch": 0.41, + "learning_rate": 4.319889715311618e-05, + "loss": 0.4906, + "step": 10657 + }, + { + "epoch": 0.41, + "learning_rate": 4.3198255963067455e-05, + "loss": 0.501, + "step": 10658 + }, + { + "epoch": 0.41, + "learning_rate": 4.319761477301873e-05, + "loss": 0.6508, + "step": 10659 + }, + { + "epoch": 0.41, + "learning_rate": 4.319697358297e-05, + "loss": 0.5431, + "step": 10660 + }, + { + "epoch": 0.41, + "learning_rate": 4.319633239292126e-05, + "loss": 0.6786, + "step": 10661 + }, + { + "epoch": 0.41, + "learning_rate": 4.3195691202872535e-05, + "loss": 0.5946, + "step": 10662 + }, + { + "epoch": 0.41, + "learning_rate": 4.31950500128238e-05, + "loss": 0.5442, + "step": 10663 + }, + { + "epoch": 0.41, + "learning_rate": 4.319440882277507e-05, + "loss": 0.5613, + "step": 10664 + }, + { + "epoch": 0.41, + "learning_rate": 4.319376763272634e-05, + "loss": 0.7859, + "step": 10665 + }, + { + "epoch": 0.41, + "learning_rate": 4.319312644267761e-05, + "loss": 0.5275, + "step": 10666 + }, + { + "epoch": 0.41, + "learning_rate": 4.319248525262889e-05, + "loss": 0.5705, + "step": 10667 + }, + { + "epoch": 0.41, + "learning_rate": 4.319184406258015e-05, + "loss": 0.5419, + "step": 10668 + }, + { + "epoch": 0.41, + "learning_rate": 4.3191202872531424e-05, + "loss": 0.4872, + "step": 10669 + }, + { + "epoch": 0.41, + "learning_rate": 4.319056168248269e-05, + "loss": 0.5203, + "step": 10670 + }, + { + "epoch": 0.41, + "learning_rate": 4.318992049243396e-05, + "loss": 0.4988, + "step": 10671 + }, + { + "epoch": 0.41, + "learning_rate": 4.3189279302385225e-05, + "loss": 0.5341, + "step": 10672 + }, + { + "epoch": 0.41, + "learning_rate": 4.31886381123365e-05, + "loss": 0.569, + "step": 10673 + }, + { + "epoch": 0.41, + "learning_rate": 4.318799692228777e-05, + "loss": 0.5567, + "step": 10674 + }, + { + "epoch": 0.41, + "learning_rate": 4.318735573223904e-05, + "loss": 0.4583, + "step": 10675 + }, + { + "epoch": 0.41, + "learning_rate": 4.318671454219031e-05, + "loss": 0.6393, + "step": 10676 + }, + { + "epoch": 0.41, + "learning_rate": 4.318607335214158e-05, + "loss": 0.5632, + "step": 10677 + }, + { + "epoch": 0.41, + "learning_rate": 4.318543216209285e-05, + "loss": 0.5851, + "step": 10678 + }, + { + "epoch": 0.41, + "learning_rate": 4.3184790972044114e-05, + "loss": 0.6287, + "step": 10679 + }, + { + "epoch": 0.41, + "learning_rate": 4.3184149781995385e-05, + "loss": 0.5043, + "step": 10680 + }, + { + "epoch": 0.41, + "learning_rate": 4.318350859194665e-05, + "loss": 0.5928, + "step": 10681 + }, + { + "epoch": 0.41, + "learning_rate": 4.318286740189792e-05, + "loss": 0.5698, + "step": 10682 + }, + { + "epoch": 0.41, + "learning_rate": 4.3182226211849194e-05, + "loss": 0.6731, + "step": 10683 + }, + { + "epoch": 0.41, + "learning_rate": 4.3181585021800465e-05, + "loss": 0.6382, + "step": 10684 + }, + { + "epoch": 0.41, + "learning_rate": 4.318094383175174e-05, + "loss": 0.6463, + "step": 10685 + }, + { + "epoch": 0.41, + "learning_rate": 4.3180302641703e-05, + "loss": 0.634, + "step": 10686 + }, + { + "epoch": 0.41, + "learning_rate": 4.3179661451654274e-05, + "loss": 0.641, + "step": 10687 + }, + { + "epoch": 0.41, + "learning_rate": 4.317902026160554e-05, + "loss": 0.6659, + "step": 10688 + }, + { + "epoch": 0.41, + "learning_rate": 4.317837907155681e-05, + "loss": 0.6661, + "step": 10689 + }, + { + "epoch": 0.41, + "learning_rate": 4.317773788150808e-05, + "loss": 0.5192, + "step": 10690 + }, + { + "epoch": 0.41, + "learning_rate": 4.3177096691459354e-05, + "loss": 0.5228, + "step": 10691 + }, + { + "epoch": 0.41, + "learning_rate": 4.317645550141062e-05, + "loss": 0.5814, + "step": 10692 + }, + { + "epoch": 0.41, + "learning_rate": 4.317581431136189e-05, + "loss": 0.6213, + "step": 10693 + }, + { + "epoch": 0.41, + "learning_rate": 4.317517312131316e-05, + "loss": 0.5874, + "step": 10694 + }, + { + "epoch": 0.41, + "learning_rate": 4.317453193126443e-05, + "loss": 0.6111, + "step": 10695 + }, + { + "epoch": 0.41, + "learning_rate": 4.31738907412157e-05, + "loss": 0.5389, + "step": 10696 + }, + { + "epoch": 0.41, + "learning_rate": 4.3173249551166964e-05, + "loss": 0.564, + "step": 10697 + }, + { + "epoch": 0.41, + "learning_rate": 4.3172608361118236e-05, + "loss": 0.5662, + "step": 10698 + }, + { + "epoch": 0.41, + "learning_rate": 4.317196717106951e-05, + "loss": 0.5676, + "step": 10699 + }, + { + "epoch": 0.41, + "learning_rate": 4.317132598102078e-05, + "loss": 0.5143, + "step": 10700 + }, + { + "epoch": 0.41, + "learning_rate": 4.3170684790972044e-05, + "loss": 0.5045, + "step": 10701 + }, + { + "epoch": 0.41, + "learning_rate": 4.3170043600923316e-05, + "loss": 0.5154, + "step": 10702 + }, + { + "epoch": 0.41, + "learning_rate": 4.316940241087459e-05, + "loss": 0.6139, + "step": 10703 + }, + { + "epoch": 0.41, + "learning_rate": 4.316876122082585e-05, + "loss": 0.6055, + "step": 10704 + }, + { + "epoch": 0.41, + "learning_rate": 4.3168120030777124e-05, + "loss": 0.6913, + "step": 10705 + }, + { + "epoch": 0.41, + "learning_rate": 4.3167478840728396e-05, + "loss": 0.5529, + "step": 10706 + }, + { + "epoch": 0.41, + "learning_rate": 4.316683765067967e-05, + "loss": 0.5308, + "step": 10707 + }, + { + "epoch": 0.41, + "learning_rate": 4.316619646063093e-05, + "loss": 0.5512, + "step": 10708 + }, + { + "epoch": 0.41, + "learning_rate": 4.3165555270582204e-05, + "loss": 0.4446, + "step": 10709 + }, + { + "epoch": 0.41, + "learning_rate": 4.316491408053347e-05, + "loss": 0.5572, + "step": 10710 + }, + { + "epoch": 0.41, + "learning_rate": 4.316427289048474e-05, + "loss": 0.5577, + "step": 10711 + }, + { + "epoch": 0.41, + "learning_rate": 4.316363170043601e-05, + "loss": 0.5864, + "step": 10712 + }, + { + "epoch": 0.41, + "learning_rate": 4.316299051038728e-05, + "loss": 0.5895, + "step": 10713 + }, + { + "epoch": 0.41, + "learning_rate": 4.316234932033855e-05, + "loss": 0.6783, + "step": 10714 + }, + { + "epoch": 0.41, + "learning_rate": 4.316170813028982e-05, + "loss": 0.4197, + "step": 10715 + }, + { + "epoch": 0.41, + "learning_rate": 4.316106694024109e-05, + "loss": 0.678, + "step": 10716 + }, + { + "epoch": 0.41, + "learning_rate": 4.316042575019236e-05, + "loss": 0.4791, + "step": 10717 + }, + { + "epoch": 0.41, + "learning_rate": 4.315978456014363e-05, + "loss": 0.5929, + "step": 10718 + }, + { + "epoch": 0.41, + "learning_rate": 4.3159143370094894e-05, + "loss": 0.4489, + "step": 10719 + }, + { + "epoch": 0.41, + "learning_rate": 4.3158502180046166e-05, + "loss": 0.5855, + "step": 10720 + }, + { + "epoch": 0.41, + "learning_rate": 4.315786098999744e-05, + "loss": 0.5083, + "step": 10721 + }, + { + "epoch": 0.41, + "learning_rate": 4.315721979994871e-05, + "loss": 0.5366, + "step": 10722 + }, + { + "epoch": 0.41, + "learning_rate": 4.315657860989998e-05, + "loss": 0.5837, + "step": 10723 + }, + { + "epoch": 0.41, + "learning_rate": 4.3155937419851246e-05, + "loss": 0.538, + "step": 10724 + }, + { + "epoch": 0.41, + "learning_rate": 4.315529622980252e-05, + "loss": 0.5899, + "step": 10725 + }, + { + "epoch": 0.41, + "learning_rate": 4.315465503975378e-05, + "loss": 0.6297, + "step": 10726 + }, + { + "epoch": 0.41, + "learning_rate": 4.3154013849705054e-05, + "loss": 0.5676, + "step": 10727 + }, + { + "epoch": 0.41, + "learning_rate": 4.315337265965632e-05, + "loss": 0.5502, + "step": 10728 + }, + { + "epoch": 0.41, + "learning_rate": 4.315273146960759e-05, + "loss": 0.5327, + "step": 10729 + }, + { + "epoch": 0.41, + "learning_rate": 4.315209027955886e-05, + "loss": 0.5431, + "step": 10730 + }, + { + "epoch": 0.41, + "learning_rate": 4.3151449089510134e-05, + "loss": 0.6054, + "step": 10731 + }, + { + "epoch": 0.41, + "learning_rate": 4.3150807899461406e-05, + "loss": 0.5551, + "step": 10732 + }, + { + "epoch": 0.41, + "learning_rate": 4.315016670941267e-05, + "loss": 0.5374, + "step": 10733 + }, + { + "epoch": 0.41, + "learning_rate": 4.314952551936394e-05, + "loss": 0.5438, + "step": 10734 + }, + { + "epoch": 0.41, + "learning_rate": 4.314888432931521e-05, + "loss": 0.5235, + "step": 10735 + }, + { + "epoch": 0.41, + "learning_rate": 4.314824313926648e-05, + "loss": 0.5163, + "step": 10736 + }, + { + "epoch": 0.41, + "learning_rate": 4.3147601949217744e-05, + "loss": 0.538, + "step": 10737 + }, + { + "epoch": 0.41, + "learning_rate": 4.3146960759169016e-05, + "loss": 0.6064, + "step": 10738 + }, + { + "epoch": 0.41, + "learning_rate": 4.3146319569120295e-05, + "loss": 0.5658, + "step": 10739 + }, + { + "epoch": 0.41, + "learning_rate": 4.314567837907156e-05, + "loss": 0.688, + "step": 10740 + }, + { + "epoch": 0.41, + "learning_rate": 4.314503718902283e-05, + "loss": 0.6604, + "step": 10741 + }, + { + "epoch": 0.41, + "learning_rate": 4.3144395998974096e-05, + "loss": 0.6046, + "step": 10742 + }, + { + "epoch": 0.41, + "learning_rate": 4.314375480892537e-05, + "loss": 0.5659, + "step": 10743 + }, + { + "epoch": 0.41, + "learning_rate": 4.314311361887663e-05, + "loss": 0.5359, + "step": 10744 + }, + { + "epoch": 0.41, + "learning_rate": 4.3142472428827905e-05, + "loss": 0.5319, + "step": 10745 + }, + { + "epoch": 0.41, + "learning_rate": 4.3141831238779176e-05, + "loss": 0.5426, + "step": 10746 + }, + { + "epoch": 0.41, + "learning_rate": 4.314119004873045e-05, + "loss": 0.5482, + "step": 10747 + }, + { + "epoch": 0.41, + "learning_rate": 4.314054885868172e-05, + "loss": 0.4899, + "step": 10748 + }, + { + "epoch": 0.41, + "learning_rate": 4.3139907668632985e-05, + "loss": 0.5766, + "step": 10749 + }, + { + "epoch": 0.41, + "learning_rate": 4.3139266478584256e-05, + "loss": 0.6562, + "step": 10750 + }, + { + "epoch": 0.41, + "learning_rate": 4.313862528853552e-05, + "loss": 0.5157, + "step": 10751 + }, + { + "epoch": 0.41, + "learning_rate": 4.313798409848679e-05, + "loss": 0.66, + "step": 10752 + }, + { + "epoch": 0.41, + "learning_rate": 4.313734290843806e-05, + "loss": 0.5796, + "step": 10753 + }, + { + "epoch": 0.41, + "learning_rate": 4.313670171838933e-05, + "loss": 0.5783, + "step": 10754 + }, + { + "epoch": 0.41, + "learning_rate": 4.313606052834061e-05, + "loss": 0.6771, + "step": 10755 + }, + { + "epoch": 0.41, + "learning_rate": 4.313541933829187e-05, + "loss": 0.5214, + "step": 10756 + }, + { + "epoch": 0.41, + "learning_rate": 4.3134778148243145e-05, + "loss": 0.4907, + "step": 10757 + }, + { + "epoch": 0.41, + "learning_rate": 4.313413695819441e-05, + "loss": 0.6055, + "step": 10758 + }, + { + "epoch": 0.41, + "learning_rate": 4.313349576814568e-05, + "loss": 0.5684, + "step": 10759 + }, + { + "epoch": 0.41, + "learning_rate": 4.3132854578096946e-05, + "loss": 0.4504, + "step": 10760 + }, + { + "epoch": 0.41, + "learning_rate": 4.313221338804822e-05, + "loss": 0.5958, + "step": 10761 + }, + { + "epoch": 0.41, + "learning_rate": 4.313157219799949e-05, + "loss": 0.5781, + "step": 10762 + }, + { + "epoch": 0.41, + "learning_rate": 4.313093100795076e-05, + "loss": 0.6041, + "step": 10763 + }, + { + "epoch": 0.41, + "learning_rate": 4.313028981790203e-05, + "loss": 0.6015, + "step": 10764 + }, + { + "epoch": 0.41, + "learning_rate": 4.31296486278533e-05, + "loss": 0.6041, + "step": 10765 + }, + { + "epoch": 0.41, + "learning_rate": 4.312900743780457e-05, + "loss": 0.5744, + "step": 10766 + }, + { + "epoch": 0.41, + "learning_rate": 4.3128366247755835e-05, + "loss": 0.6132, + "step": 10767 + }, + { + "epoch": 0.41, + "learning_rate": 4.312772505770711e-05, + "loss": 0.5374, + "step": 10768 + }, + { + "epoch": 0.41, + "learning_rate": 4.312708386765837e-05, + "loss": 0.5042, + "step": 10769 + }, + { + "epoch": 0.41, + "learning_rate": 4.312644267760964e-05, + "loss": 0.5222, + "step": 10770 + }, + { + "epoch": 0.41, + "learning_rate": 4.3125801487560915e-05, + "loss": 0.6937, + "step": 10771 + }, + { + "epoch": 0.41, + "learning_rate": 4.312516029751219e-05, + "loss": 0.6106, + "step": 10772 + }, + { + "epoch": 0.41, + "learning_rate": 4.312451910746346e-05, + "loss": 0.5995, + "step": 10773 + }, + { + "epoch": 0.41, + "learning_rate": 4.3123877917414723e-05, + "loss": 0.5728, + "step": 10774 + }, + { + "epoch": 0.41, + "learning_rate": 4.3123236727365995e-05, + "loss": 0.5534, + "step": 10775 + }, + { + "epoch": 0.41, + "learning_rate": 4.312259553731726e-05, + "loss": 0.5444, + "step": 10776 + }, + { + "epoch": 0.41, + "learning_rate": 4.312195434726853e-05, + "loss": 0.5323, + "step": 10777 + }, + { + "epoch": 0.41, + "learning_rate": 4.3121313157219804e-05, + "loss": 0.5264, + "step": 10778 + }, + { + "epoch": 0.41, + "learning_rate": 4.3120671967171075e-05, + "loss": 0.5253, + "step": 10779 + }, + { + "epoch": 0.41, + "learning_rate": 4.312003077712234e-05, + "loss": 0.6138, + "step": 10780 + }, + { + "epoch": 0.41, + "learning_rate": 4.311938958707361e-05, + "loss": 0.5514, + "step": 10781 + }, + { + "epoch": 0.41, + "learning_rate": 4.3118748397024884e-05, + "loss": 0.494, + "step": 10782 + }, + { + "epoch": 0.41, + "learning_rate": 4.311810720697615e-05, + "loss": 0.5244, + "step": 10783 + }, + { + "epoch": 0.41, + "learning_rate": 4.311746601692742e-05, + "loss": 0.582, + "step": 10784 + }, + { + "epoch": 0.41, + "learning_rate": 4.3116824826878685e-05, + "loss": 0.6602, + "step": 10785 + }, + { + "epoch": 0.41, + "learning_rate": 4.311618363682996e-05, + "loss": 0.5654, + "step": 10786 + }, + { + "epoch": 0.41, + "learning_rate": 4.311554244678123e-05, + "loss": 0.6319, + "step": 10787 + }, + { + "epoch": 0.41, + "learning_rate": 4.31149012567325e-05, + "loss": 0.5218, + "step": 10788 + }, + { + "epoch": 0.41, + "learning_rate": 4.3114260066683765e-05, + "loss": 0.6221, + "step": 10789 + }, + { + "epoch": 0.41, + "learning_rate": 4.311361887663504e-05, + "loss": 0.568, + "step": 10790 + }, + { + "epoch": 0.41, + "learning_rate": 4.311297768658631e-05, + "loss": 0.6765, + "step": 10791 + }, + { + "epoch": 0.41, + "learning_rate": 4.3112336496537574e-05, + "loss": 0.5534, + "step": 10792 + }, + { + "epoch": 0.41, + "learning_rate": 4.3111695306488845e-05, + "loss": 0.586, + "step": 10793 + }, + { + "epoch": 0.41, + "learning_rate": 4.311105411644012e-05, + "loss": 0.5613, + "step": 10794 + }, + { + "epoch": 0.42, + "learning_rate": 4.311041292639139e-05, + "loss": 0.5148, + "step": 10795 + }, + { + "epoch": 0.42, + "learning_rate": 4.3109771736342654e-05, + "loss": 0.5765, + "step": 10796 + }, + { + "epoch": 0.42, + "learning_rate": 4.3109130546293925e-05, + "loss": 0.5288, + "step": 10797 + }, + { + "epoch": 0.42, + "learning_rate": 4.310848935624519e-05, + "loss": 0.5879, + "step": 10798 + }, + { + "epoch": 0.42, + "learning_rate": 4.310784816619646e-05, + "loss": 0.4958, + "step": 10799 + }, + { + "epoch": 0.42, + "learning_rate": 4.3107206976147734e-05, + "loss": 0.4794, + "step": 10800 + }, + { + "epoch": 0.42, + "learning_rate": 4.3106565786099e-05, + "loss": 0.4931, + "step": 10801 + }, + { + "epoch": 0.42, + "learning_rate": 4.310592459605027e-05, + "loss": 0.5421, + "step": 10802 + }, + { + "epoch": 0.42, + "learning_rate": 4.310528340600154e-05, + "loss": 0.5578, + "step": 10803 + }, + { + "epoch": 0.42, + "learning_rate": 4.3104642215952814e-05, + "loss": 0.5261, + "step": 10804 + }, + { + "epoch": 0.42, + "learning_rate": 4.310400102590408e-05, + "loss": 0.5571, + "step": 10805 + }, + { + "epoch": 0.42, + "learning_rate": 4.310335983585535e-05, + "loss": 0.5894, + "step": 10806 + }, + { + "epoch": 0.42, + "learning_rate": 4.3102718645806616e-05, + "loss": 0.6482, + "step": 10807 + }, + { + "epoch": 0.42, + "learning_rate": 4.310207745575789e-05, + "loss": 0.5499, + "step": 10808 + }, + { + "epoch": 0.42, + "learning_rate": 4.310143626570916e-05, + "loss": 0.4607, + "step": 10809 + }, + { + "epoch": 0.42, + "learning_rate": 4.310079507566043e-05, + "loss": 0.5385, + "step": 10810 + }, + { + "epoch": 0.42, + "learning_rate": 4.31001538856117e-05, + "loss": 0.5874, + "step": 10811 + }, + { + "epoch": 0.42, + "learning_rate": 4.309951269556297e-05, + "loss": 0.5197, + "step": 10812 + }, + { + "epoch": 0.42, + "learning_rate": 4.309887150551424e-05, + "loss": 0.6077, + "step": 10813 + }, + { + "epoch": 0.42, + "learning_rate": 4.3098230315465504e-05, + "loss": 0.5712, + "step": 10814 + }, + { + "epoch": 0.42, + "learning_rate": 4.3097589125416776e-05, + "loss": 0.5524, + "step": 10815 + }, + { + "epoch": 0.42, + "learning_rate": 4.309694793536804e-05, + "loss": 0.5976, + "step": 10816 + }, + { + "epoch": 0.42, + "learning_rate": 4.309630674531931e-05, + "loss": 0.6141, + "step": 10817 + }, + { + "epoch": 0.42, + "learning_rate": 4.3095665555270584e-05, + "loss": 0.5275, + "step": 10818 + }, + { + "epoch": 0.42, + "learning_rate": 4.3095024365221856e-05, + "loss": 0.6217, + "step": 10819 + }, + { + "epoch": 0.42, + "learning_rate": 4.309438317517313e-05, + "loss": 0.5492, + "step": 10820 + }, + { + "epoch": 0.42, + "learning_rate": 4.309374198512439e-05, + "loss": 0.6037, + "step": 10821 + }, + { + "epoch": 0.42, + "learning_rate": 4.3093100795075664e-05, + "loss": 0.5357, + "step": 10822 + }, + { + "epoch": 0.42, + "learning_rate": 4.309245960502693e-05, + "loss": 0.4759, + "step": 10823 + }, + { + "epoch": 0.42, + "learning_rate": 4.30918184149782e-05, + "loss": 0.5691, + "step": 10824 + }, + { + "epoch": 0.42, + "learning_rate": 4.3091177224929466e-05, + "loss": 0.5554, + "step": 10825 + }, + { + "epoch": 0.42, + "learning_rate": 4.309053603488074e-05, + "loss": 0.5886, + "step": 10826 + }, + { + "epoch": 0.42, + "learning_rate": 4.3089894844832016e-05, + "loss": 0.5984, + "step": 10827 + }, + { + "epoch": 0.42, + "learning_rate": 4.308925365478328e-05, + "loss": 0.4907, + "step": 10828 + }, + { + "epoch": 0.42, + "learning_rate": 4.308861246473455e-05, + "loss": 0.5249, + "step": 10829 + }, + { + "epoch": 0.42, + "learning_rate": 4.308797127468582e-05, + "loss": 0.6663, + "step": 10830 + }, + { + "epoch": 0.42, + "learning_rate": 4.308733008463709e-05, + "loss": 0.5363, + "step": 10831 + }, + { + "epoch": 0.42, + "learning_rate": 4.3086688894588354e-05, + "loss": 0.4861, + "step": 10832 + }, + { + "epoch": 0.42, + "learning_rate": 4.3086047704539626e-05, + "loss": 0.6095, + "step": 10833 + }, + { + "epoch": 0.42, + "learning_rate": 4.30854065144909e-05, + "loss": 0.5304, + "step": 10834 + }, + { + "epoch": 0.42, + "learning_rate": 4.308476532444217e-05, + "loss": 0.576, + "step": 10835 + }, + { + "epoch": 0.42, + "learning_rate": 4.308412413439344e-05, + "loss": 0.5088, + "step": 10836 + }, + { + "epoch": 0.42, + "learning_rate": 4.3083482944344706e-05, + "loss": 0.621, + "step": 10837 + }, + { + "epoch": 0.42, + "learning_rate": 4.308284175429598e-05, + "loss": 0.4422, + "step": 10838 + }, + { + "epoch": 0.42, + "learning_rate": 4.308220056424724e-05, + "loss": 0.5612, + "step": 10839 + }, + { + "epoch": 0.42, + "learning_rate": 4.3081559374198514e-05, + "loss": 0.5582, + "step": 10840 + }, + { + "epoch": 0.42, + "learning_rate": 4.308091818414978e-05, + "loss": 0.6001, + "step": 10841 + }, + { + "epoch": 0.42, + "learning_rate": 4.308027699410105e-05, + "loss": 0.4859, + "step": 10842 + }, + { + "epoch": 0.42, + "learning_rate": 4.307963580405232e-05, + "loss": 0.5482, + "step": 10843 + }, + { + "epoch": 0.42, + "learning_rate": 4.3078994614003595e-05, + "loss": 0.5584, + "step": 10844 + }, + { + "epoch": 0.42, + "learning_rate": 4.3078353423954866e-05, + "loss": 0.5555, + "step": 10845 + }, + { + "epoch": 0.42, + "learning_rate": 4.307771223390613e-05, + "loss": 0.5539, + "step": 10846 + }, + { + "epoch": 0.42, + "learning_rate": 4.30770710438574e-05, + "loss": 0.5547, + "step": 10847 + }, + { + "epoch": 0.42, + "learning_rate": 4.307642985380867e-05, + "loss": 0.5814, + "step": 10848 + }, + { + "epoch": 0.42, + "learning_rate": 4.307578866375994e-05, + "loss": 0.5033, + "step": 10849 + }, + { + "epoch": 0.42, + "learning_rate": 4.307514747371121e-05, + "loss": 0.601, + "step": 10850 + }, + { + "epoch": 0.42, + "learning_rate": 4.307450628366248e-05, + "loss": 0.6035, + "step": 10851 + }, + { + "epoch": 0.42, + "learning_rate": 4.307386509361375e-05, + "loss": 0.5173, + "step": 10852 + }, + { + "epoch": 0.42, + "learning_rate": 4.307322390356502e-05, + "loss": 0.5747, + "step": 10853 + }, + { + "epoch": 0.42, + "learning_rate": 4.307258271351629e-05, + "loss": 0.6409, + "step": 10854 + }, + { + "epoch": 0.42, + "learning_rate": 4.3071941523467556e-05, + "loss": 0.6004, + "step": 10855 + }, + { + "epoch": 0.42, + "learning_rate": 4.307130033341883e-05, + "loss": 0.4381, + "step": 10856 + }, + { + "epoch": 0.42, + "learning_rate": 4.307065914337009e-05, + "loss": 0.6282, + "step": 10857 + }, + { + "epoch": 0.42, + "learning_rate": 4.3070017953321365e-05, + "loss": 0.537, + "step": 10858 + }, + { + "epoch": 0.42, + "learning_rate": 4.3069376763272636e-05, + "loss": 0.6275, + "step": 10859 + }, + { + "epoch": 0.42, + "learning_rate": 4.306873557322391e-05, + "loss": 0.6944, + "step": 10860 + }, + { + "epoch": 0.42, + "learning_rate": 4.306809438317517e-05, + "loss": 0.6612, + "step": 10861 + }, + { + "epoch": 0.42, + "learning_rate": 4.3067453193126445e-05, + "loss": 0.5943, + "step": 10862 + }, + { + "epoch": 0.42, + "learning_rate": 4.3066812003077716e-05, + "loss": 0.5628, + "step": 10863 + }, + { + "epoch": 0.42, + "learning_rate": 4.306617081302898e-05, + "loss": 0.5223, + "step": 10864 + }, + { + "epoch": 0.42, + "learning_rate": 4.306552962298025e-05, + "loss": 0.5444, + "step": 10865 + }, + { + "epoch": 0.42, + "learning_rate": 4.3064888432931525e-05, + "loss": 0.5843, + "step": 10866 + }, + { + "epoch": 0.42, + "learning_rate": 4.3064247242882797e-05, + "loss": 0.6346, + "step": 10867 + }, + { + "epoch": 0.42, + "learning_rate": 4.306360605283406e-05, + "loss": 0.5229, + "step": 10868 + }, + { + "epoch": 0.42, + "learning_rate": 4.306296486278533e-05, + "loss": 0.4958, + "step": 10869 + }, + { + "epoch": 0.42, + "learning_rate": 4.30623236727366e-05, + "loss": 0.4933, + "step": 10870 + }, + { + "epoch": 0.42, + "learning_rate": 4.306168248268787e-05, + "loss": 0.4624, + "step": 10871 + }, + { + "epoch": 0.42, + "learning_rate": 4.306104129263914e-05, + "loss": 0.5585, + "step": 10872 + }, + { + "epoch": 0.42, + "learning_rate": 4.3060400102590407e-05, + "loss": 0.6255, + "step": 10873 + }, + { + "epoch": 0.42, + "learning_rate": 4.305975891254168e-05, + "loss": 0.5577, + "step": 10874 + }, + { + "epoch": 0.42, + "learning_rate": 4.305911772249295e-05, + "loss": 0.4542, + "step": 10875 + }, + { + "epoch": 0.42, + "learning_rate": 4.305847653244422e-05, + "loss": 0.5693, + "step": 10876 + }, + { + "epoch": 0.42, + "learning_rate": 4.305783534239549e-05, + "loss": 0.5732, + "step": 10877 + }, + { + "epoch": 0.42, + "learning_rate": 4.305719415234676e-05, + "loss": 0.505, + "step": 10878 + }, + { + "epoch": 0.42, + "learning_rate": 4.305655296229802e-05, + "loss": 0.4996, + "step": 10879 + }, + { + "epoch": 0.42, + "learning_rate": 4.3055911772249295e-05, + "loss": 0.5364, + "step": 10880 + }, + { + "epoch": 0.42, + "learning_rate": 4.305527058220057e-05, + "loss": 0.6477, + "step": 10881 + }, + { + "epoch": 0.42, + "learning_rate": 4.305462939215184e-05, + "loss": 0.5106, + "step": 10882 + }, + { + "epoch": 0.42, + "learning_rate": 4.305398820210311e-05, + "loss": 0.6276, + "step": 10883 + }, + { + "epoch": 0.42, + "learning_rate": 4.3053347012054375e-05, + "loss": 0.5167, + "step": 10884 + }, + { + "epoch": 0.42, + "learning_rate": 4.305270582200565e-05, + "loss": 0.5444, + "step": 10885 + }, + { + "epoch": 0.42, + "learning_rate": 4.305206463195691e-05, + "loss": 0.5531, + "step": 10886 + }, + { + "epoch": 0.42, + "learning_rate": 4.3051423441908183e-05, + "loss": 0.6037, + "step": 10887 + }, + { + "epoch": 0.42, + "learning_rate": 4.305078225185945e-05, + "loss": 0.5392, + "step": 10888 + }, + { + "epoch": 0.42, + "learning_rate": 4.305014106181072e-05, + "loss": 0.5169, + "step": 10889 + }, + { + "epoch": 0.42, + "learning_rate": 4.304949987176199e-05, + "loss": 0.5129, + "step": 10890 + }, + { + "epoch": 0.42, + "learning_rate": 4.3048858681713264e-05, + "loss": 0.5354, + "step": 10891 + }, + { + "epoch": 0.42, + "learning_rate": 4.3048217491664535e-05, + "loss": 0.5744, + "step": 10892 + }, + { + "epoch": 0.42, + "learning_rate": 4.30475763016158e-05, + "loss": 0.631, + "step": 10893 + }, + { + "epoch": 0.42, + "learning_rate": 4.304693511156707e-05, + "loss": 0.6455, + "step": 10894 + }, + { + "epoch": 0.42, + "learning_rate": 4.304629392151834e-05, + "loss": 0.6297, + "step": 10895 + }, + { + "epoch": 0.42, + "learning_rate": 4.304565273146961e-05, + "loss": 0.4859, + "step": 10896 + }, + { + "epoch": 0.42, + "learning_rate": 4.3045011541420874e-05, + "loss": 0.5539, + "step": 10897 + }, + { + "epoch": 0.42, + "learning_rate": 4.3044370351372145e-05, + "loss": 0.5926, + "step": 10898 + }, + { + "epoch": 0.42, + "learning_rate": 4.3043729161323424e-05, + "loss": 0.5468, + "step": 10899 + }, + { + "epoch": 0.42, + "learning_rate": 4.304308797127469e-05, + "loss": 0.5328, + "step": 10900 + }, + { + "epoch": 0.42, + "learning_rate": 4.304244678122596e-05, + "loss": 0.6191, + "step": 10901 + }, + { + "epoch": 0.42, + "learning_rate": 4.3041805591177225e-05, + "loss": 0.6095, + "step": 10902 + }, + { + "epoch": 0.42, + "learning_rate": 4.30411644011285e-05, + "loss": 0.648, + "step": 10903 + }, + { + "epoch": 0.42, + "learning_rate": 4.304052321107976e-05, + "loss": 0.5054, + "step": 10904 + }, + { + "epoch": 0.42, + "learning_rate": 4.3039882021031034e-05, + "loss": 0.5029, + "step": 10905 + }, + { + "epoch": 0.42, + "learning_rate": 4.3039240830982305e-05, + "loss": 0.6157, + "step": 10906 + }, + { + "epoch": 0.42, + "learning_rate": 4.303859964093358e-05, + "loss": 0.5902, + "step": 10907 + }, + { + "epoch": 0.42, + "learning_rate": 4.303795845088485e-05, + "loss": 0.573, + "step": 10908 + }, + { + "epoch": 0.42, + "learning_rate": 4.3037317260836114e-05, + "loss": 0.5734, + "step": 10909 + }, + { + "epoch": 0.42, + "learning_rate": 4.3036676070787386e-05, + "loss": 0.5734, + "step": 10910 + }, + { + "epoch": 0.42, + "learning_rate": 4.303603488073865e-05, + "loss": 0.559, + "step": 10911 + }, + { + "epoch": 0.42, + "learning_rate": 4.303539369068992e-05, + "loss": 0.4641, + "step": 10912 + }, + { + "epoch": 0.42, + "learning_rate": 4.303475250064119e-05, + "loss": 0.469, + "step": 10913 + }, + { + "epoch": 0.42, + "learning_rate": 4.303411131059246e-05, + "loss": 0.6319, + "step": 10914 + }, + { + "epoch": 0.42, + "learning_rate": 4.303347012054373e-05, + "loss": 0.5635, + "step": 10915 + }, + { + "epoch": 0.42, + "learning_rate": 4.3032828930495e-05, + "loss": 0.6754, + "step": 10916 + }, + { + "epoch": 0.42, + "learning_rate": 4.3032187740446274e-05, + "loss": 0.5495, + "step": 10917 + }, + { + "epoch": 0.42, + "learning_rate": 4.303154655039754e-05, + "loss": 0.5653, + "step": 10918 + }, + { + "epoch": 0.42, + "learning_rate": 4.303090536034881e-05, + "loss": 0.4938, + "step": 10919 + }, + { + "epoch": 0.42, + "learning_rate": 4.3030264170300076e-05, + "loss": 0.5913, + "step": 10920 + }, + { + "epoch": 0.42, + "learning_rate": 4.302962298025135e-05, + "loss": 0.5275, + "step": 10921 + }, + { + "epoch": 0.42, + "learning_rate": 4.302898179020262e-05, + "loss": 0.5254, + "step": 10922 + }, + { + "epoch": 0.42, + "learning_rate": 4.302834060015389e-05, + "loss": 0.5314, + "step": 10923 + }, + { + "epoch": 0.42, + "learning_rate": 4.3027699410105156e-05, + "loss": 0.4998, + "step": 10924 + }, + { + "epoch": 0.42, + "learning_rate": 4.302705822005643e-05, + "loss": 0.5098, + "step": 10925 + }, + { + "epoch": 0.42, + "learning_rate": 4.30264170300077e-05, + "loss": 0.6392, + "step": 10926 + }, + { + "epoch": 0.42, + "learning_rate": 4.3025775839958964e-05, + "loss": 0.532, + "step": 10927 + }, + { + "epoch": 0.42, + "learning_rate": 4.3025134649910236e-05, + "loss": 0.5849, + "step": 10928 + }, + { + "epoch": 0.42, + "learning_rate": 4.30244934598615e-05, + "loss": 0.5505, + "step": 10929 + }, + { + "epoch": 0.42, + "learning_rate": 4.302385226981277e-05, + "loss": 0.6384, + "step": 10930 + }, + { + "epoch": 0.42, + "learning_rate": 4.3023211079764044e-05, + "loss": 0.5864, + "step": 10931 + }, + { + "epoch": 0.42, + "learning_rate": 4.3022569889715316e-05, + "loss": 0.5433, + "step": 10932 + }, + { + "epoch": 0.42, + "learning_rate": 4.302192869966658e-05, + "loss": 0.596, + "step": 10933 + }, + { + "epoch": 0.42, + "learning_rate": 4.302128750961785e-05, + "loss": 0.6687, + "step": 10934 + }, + { + "epoch": 0.42, + "learning_rate": 4.3020646319569124e-05, + "loss": 0.4802, + "step": 10935 + }, + { + "epoch": 0.42, + "learning_rate": 4.302000512952039e-05, + "loss": 0.5995, + "step": 10936 + }, + { + "epoch": 0.42, + "learning_rate": 4.301936393947166e-05, + "loss": 0.5152, + "step": 10937 + }, + { + "epoch": 0.42, + "learning_rate": 4.301872274942293e-05, + "loss": 0.5101, + "step": 10938 + }, + { + "epoch": 0.42, + "learning_rate": 4.3018081559374204e-05, + "loss": 0.5139, + "step": 10939 + }, + { + "epoch": 0.42, + "learning_rate": 4.301744036932547e-05, + "loss": 0.5452, + "step": 10940 + }, + { + "epoch": 0.42, + "learning_rate": 4.301679917927674e-05, + "loss": 0.5695, + "step": 10941 + }, + { + "epoch": 0.42, + "learning_rate": 4.301615798922801e-05, + "loss": 0.5496, + "step": 10942 + }, + { + "epoch": 0.42, + "learning_rate": 4.301551679917928e-05, + "loss": 0.5691, + "step": 10943 + }, + { + "epoch": 0.42, + "learning_rate": 4.301487560913055e-05, + "loss": 0.7062, + "step": 10944 + }, + { + "epoch": 0.42, + "learning_rate": 4.3014234419081814e-05, + "loss": 0.4894, + "step": 10945 + }, + { + "epoch": 0.42, + "learning_rate": 4.3013593229033086e-05, + "loss": 0.5588, + "step": 10946 + }, + { + "epoch": 0.42, + "learning_rate": 4.301295203898436e-05, + "loss": 0.6613, + "step": 10947 + }, + { + "epoch": 0.42, + "learning_rate": 4.301231084893563e-05, + "loss": 0.5491, + "step": 10948 + }, + { + "epoch": 0.42, + "learning_rate": 4.3011669658886894e-05, + "loss": 0.566, + "step": 10949 + }, + { + "epoch": 0.42, + "learning_rate": 4.3011028468838166e-05, + "loss": 0.5573, + "step": 10950 + }, + { + "epoch": 0.42, + "learning_rate": 4.301038727878944e-05, + "loss": 0.6059, + "step": 10951 + }, + { + "epoch": 0.42, + "learning_rate": 4.30097460887407e-05, + "loss": 0.5569, + "step": 10952 + }, + { + "epoch": 0.42, + "learning_rate": 4.3009104898691974e-05, + "loss": 0.5638, + "step": 10953 + }, + { + "epoch": 0.42, + "learning_rate": 4.3008463708643246e-05, + "loss": 0.5903, + "step": 10954 + }, + { + "epoch": 0.42, + "learning_rate": 4.300782251859452e-05, + "loss": 0.5404, + "step": 10955 + }, + { + "epoch": 0.42, + "learning_rate": 4.300718132854578e-05, + "loss": 0.5389, + "step": 10956 + }, + { + "epoch": 0.42, + "learning_rate": 4.3006540138497055e-05, + "loss": 0.5905, + "step": 10957 + }, + { + "epoch": 0.42, + "learning_rate": 4.300589894844832e-05, + "loss": 0.5474, + "step": 10958 + }, + { + "epoch": 0.42, + "learning_rate": 4.300525775839959e-05, + "loss": 0.4842, + "step": 10959 + }, + { + "epoch": 0.42, + "learning_rate": 4.300461656835086e-05, + "loss": 0.5831, + "step": 10960 + }, + { + "epoch": 0.42, + "learning_rate": 4.300397537830213e-05, + "loss": 0.5914, + "step": 10961 + }, + { + "epoch": 0.42, + "learning_rate": 4.30033341882534e-05, + "loss": 0.5244, + "step": 10962 + }, + { + "epoch": 0.42, + "learning_rate": 4.300269299820467e-05, + "loss": 0.4438, + "step": 10963 + }, + { + "epoch": 0.42, + "learning_rate": 4.300205180815594e-05, + "loss": 0.5266, + "step": 10964 + }, + { + "epoch": 0.42, + "learning_rate": 4.300141061810721e-05, + "loss": 0.5588, + "step": 10965 + }, + { + "epoch": 0.42, + "learning_rate": 4.300076942805848e-05, + "loss": 0.5849, + "step": 10966 + }, + { + "epoch": 0.42, + "learning_rate": 4.3000128238009745e-05, + "loss": 0.6447, + "step": 10967 + }, + { + "epoch": 0.42, + "learning_rate": 4.2999487047961016e-05, + "loss": 0.5947, + "step": 10968 + }, + { + "epoch": 0.42, + "learning_rate": 4.299884585791229e-05, + "loss": 0.5452, + "step": 10969 + }, + { + "epoch": 0.42, + "learning_rate": 4.299820466786356e-05, + "loss": 0.5548, + "step": 10970 + }, + { + "epoch": 0.42, + "learning_rate": 4.299756347781483e-05, + "loss": 0.5662, + "step": 10971 + }, + { + "epoch": 0.42, + "learning_rate": 4.2996922287766096e-05, + "loss": 0.3963, + "step": 10972 + }, + { + "epoch": 0.42, + "learning_rate": 4.299628109771737e-05, + "loss": 0.578, + "step": 10973 + }, + { + "epoch": 0.42, + "learning_rate": 4.299563990766863e-05, + "loss": 0.6119, + "step": 10974 + }, + { + "epoch": 0.42, + "learning_rate": 4.2994998717619905e-05, + "loss": 0.5141, + "step": 10975 + }, + { + "epoch": 0.42, + "learning_rate": 4.299435752757117e-05, + "loss": 0.6186, + "step": 10976 + }, + { + "epoch": 0.42, + "learning_rate": 4.299371633752244e-05, + "loss": 0.5467, + "step": 10977 + }, + { + "epoch": 0.42, + "learning_rate": 4.299307514747371e-05, + "loss": 0.5843, + "step": 10978 + }, + { + "epoch": 0.42, + "learning_rate": 4.2992433957424985e-05, + "loss": 0.4997, + "step": 10979 + }, + { + "epoch": 0.42, + "learning_rate": 4.2991792767376257e-05, + "loss": 0.5369, + "step": 10980 + }, + { + "epoch": 0.42, + "learning_rate": 4.299115157732752e-05, + "loss": 0.4702, + "step": 10981 + }, + { + "epoch": 0.42, + "learning_rate": 4.299051038727879e-05, + "loss": 0.547, + "step": 10982 + }, + { + "epoch": 0.42, + "learning_rate": 4.298986919723006e-05, + "loss": 0.4823, + "step": 10983 + }, + { + "epoch": 0.42, + "learning_rate": 4.298922800718133e-05, + "loss": 0.5211, + "step": 10984 + }, + { + "epoch": 0.42, + "learning_rate": 4.2988586817132595e-05, + "loss": 0.7369, + "step": 10985 + }, + { + "epoch": 0.42, + "learning_rate": 4.2987945627083867e-05, + "loss": 0.6876, + "step": 10986 + }, + { + "epoch": 0.42, + "learning_rate": 4.2987304437035145e-05, + "loss": 0.5051, + "step": 10987 + }, + { + "epoch": 0.42, + "learning_rate": 4.298666324698641e-05, + "loss": 0.4301, + "step": 10988 + }, + { + "epoch": 0.42, + "learning_rate": 4.298602205693768e-05, + "loss": 0.5199, + "step": 10989 + }, + { + "epoch": 0.42, + "learning_rate": 4.298538086688895e-05, + "loss": 0.6557, + "step": 10990 + }, + { + "epoch": 0.42, + "learning_rate": 4.298473967684022e-05, + "loss": 0.5624, + "step": 10991 + }, + { + "epoch": 0.42, + "learning_rate": 4.298409848679148e-05, + "loss": 0.6043, + "step": 10992 + }, + { + "epoch": 0.42, + "learning_rate": 4.2983457296742755e-05, + "loss": 0.5968, + "step": 10993 + }, + { + "epoch": 0.42, + "learning_rate": 4.298281610669403e-05, + "loss": 0.4944, + "step": 10994 + }, + { + "epoch": 0.42, + "learning_rate": 4.29821749166453e-05, + "loss": 0.5577, + "step": 10995 + }, + { + "epoch": 0.42, + "learning_rate": 4.298153372659657e-05, + "loss": 0.5131, + "step": 10996 + }, + { + "epoch": 0.42, + "learning_rate": 4.2980892536547835e-05, + "loss": 0.5157, + "step": 10997 + }, + { + "epoch": 0.42, + "learning_rate": 4.298025134649911e-05, + "loss": 0.6975, + "step": 10998 + }, + { + "epoch": 0.42, + "learning_rate": 4.297961015645037e-05, + "loss": 0.5118, + "step": 10999 + }, + { + "epoch": 0.42, + "learning_rate": 4.2978968966401643e-05, + "loss": 0.559, + "step": 11000 + }, + { + "epoch": 0.42, + "learning_rate": 4.297832777635291e-05, + "loss": 0.6917, + "step": 11001 + }, + { + "epoch": 0.42, + "learning_rate": 4.297768658630418e-05, + "loss": 0.5821, + "step": 11002 + }, + { + "epoch": 0.42, + "learning_rate": 4.297704539625545e-05, + "loss": 0.5963, + "step": 11003 + }, + { + "epoch": 0.42, + "learning_rate": 4.2976404206206724e-05, + "loss": 0.5341, + "step": 11004 + }, + { + "epoch": 0.42, + "learning_rate": 4.2975763016157995e-05, + "loss": 0.5514, + "step": 11005 + }, + { + "epoch": 0.42, + "learning_rate": 4.297512182610926e-05, + "loss": 0.57, + "step": 11006 + }, + { + "epoch": 0.42, + "learning_rate": 4.297448063606053e-05, + "loss": 0.5928, + "step": 11007 + }, + { + "epoch": 0.42, + "learning_rate": 4.29738394460118e-05, + "loss": 0.5383, + "step": 11008 + }, + { + "epoch": 0.42, + "learning_rate": 4.297319825596307e-05, + "loss": 0.5097, + "step": 11009 + }, + { + "epoch": 0.42, + "learning_rate": 4.297255706591434e-05, + "loss": 0.5665, + "step": 11010 + }, + { + "epoch": 0.42, + "learning_rate": 4.297191587586561e-05, + "loss": 0.6282, + "step": 11011 + }, + { + "epoch": 0.42, + "learning_rate": 4.297127468581688e-05, + "loss": 0.5791, + "step": 11012 + }, + { + "epoch": 0.42, + "learning_rate": 4.297063349576815e-05, + "loss": 0.5074, + "step": 11013 + }, + { + "epoch": 0.42, + "learning_rate": 4.296999230571942e-05, + "loss": 0.5802, + "step": 11014 + }, + { + "epoch": 0.42, + "learning_rate": 4.2969351115670685e-05, + "loss": 0.5946, + "step": 11015 + }, + { + "epoch": 0.42, + "learning_rate": 4.296870992562196e-05, + "loss": 0.5764, + "step": 11016 + }, + { + "epoch": 0.42, + "learning_rate": 4.296806873557322e-05, + "loss": 0.5723, + "step": 11017 + }, + { + "epoch": 0.42, + "learning_rate": 4.2967427545524494e-05, + "loss": 0.6295, + "step": 11018 + }, + { + "epoch": 0.42, + "learning_rate": 4.2966786355475765e-05, + "loss": 0.557, + "step": 11019 + }, + { + "epoch": 0.42, + "learning_rate": 4.296614516542704e-05, + "loss": 0.4723, + "step": 11020 + }, + { + "epoch": 0.42, + "learning_rate": 4.29655039753783e-05, + "loss": 0.5742, + "step": 11021 + }, + { + "epoch": 0.42, + "learning_rate": 4.2964862785329574e-05, + "loss": 0.6388, + "step": 11022 + }, + { + "epoch": 0.42, + "learning_rate": 4.2964221595280846e-05, + "loss": 0.566, + "step": 11023 + }, + { + "epoch": 0.42, + "learning_rate": 4.296358040523211e-05, + "loss": 0.5858, + "step": 11024 + }, + { + "epoch": 0.42, + "learning_rate": 4.296293921518338e-05, + "loss": 0.4875, + "step": 11025 + }, + { + "epoch": 0.42, + "learning_rate": 4.2962298025134654e-05, + "loss": 0.5322, + "step": 11026 + }, + { + "epoch": 0.42, + "learning_rate": 4.2961656835085926e-05, + "loss": 0.6143, + "step": 11027 + }, + { + "epoch": 0.42, + "learning_rate": 4.296101564503719e-05, + "loss": 0.506, + "step": 11028 + }, + { + "epoch": 0.42, + "learning_rate": 4.296037445498846e-05, + "loss": 0.6391, + "step": 11029 + }, + { + "epoch": 0.42, + "learning_rate": 4.295973326493973e-05, + "loss": 0.4583, + "step": 11030 + }, + { + "epoch": 0.42, + "learning_rate": 4.2959092074891e-05, + "loss": 0.6095, + "step": 11031 + }, + { + "epoch": 0.42, + "learning_rate": 4.295845088484227e-05, + "loss": 0.4393, + "step": 11032 + }, + { + "epoch": 0.42, + "learning_rate": 4.2957809694793536e-05, + "loss": 0.6214, + "step": 11033 + }, + { + "epoch": 0.42, + "learning_rate": 4.295716850474481e-05, + "loss": 0.5607, + "step": 11034 + }, + { + "epoch": 0.42, + "learning_rate": 4.295652731469608e-05, + "loss": 0.6088, + "step": 11035 + }, + { + "epoch": 0.42, + "learning_rate": 4.295588612464735e-05, + "loss": 0.548, + "step": 11036 + }, + { + "epoch": 0.42, + "learning_rate": 4.2955244934598616e-05, + "loss": 0.6845, + "step": 11037 + }, + { + "epoch": 0.42, + "learning_rate": 4.295460374454989e-05, + "loss": 0.5638, + "step": 11038 + }, + { + "epoch": 0.42, + "learning_rate": 4.295396255450115e-05, + "loss": 0.6512, + "step": 11039 + }, + { + "epoch": 0.42, + "learning_rate": 4.2953321364452424e-05, + "loss": 0.58, + "step": 11040 + }, + { + "epoch": 0.42, + "learning_rate": 4.2952680174403696e-05, + "loss": 0.5676, + "step": 11041 + }, + { + "epoch": 0.42, + "learning_rate": 4.295203898435497e-05, + "loss": 0.5793, + "step": 11042 + }, + { + "epoch": 0.42, + "learning_rate": 4.295139779430624e-05, + "loss": 0.5716, + "step": 11043 + }, + { + "epoch": 0.42, + "learning_rate": 4.2950756604257504e-05, + "loss": 0.5894, + "step": 11044 + }, + { + "epoch": 0.42, + "learning_rate": 4.2950115414208776e-05, + "loss": 0.6467, + "step": 11045 + }, + { + "epoch": 0.42, + "learning_rate": 4.294947422416004e-05, + "loss": 0.5495, + "step": 11046 + }, + { + "epoch": 0.42, + "learning_rate": 4.294883303411131e-05, + "loss": 0.5982, + "step": 11047 + }, + { + "epoch": 0.42, + "learning_rate": 4.294819184406258e-05, + "loss": 0.5022, + "step": 11048 + }, + { + "epoch": 0.42, + "learning_rate": 4.294755065401385e-05, + "loss": 0.5505, + "step": 11049 + }, + { + "epoch": 0.42, + "learning_rate": 4.294690946396512e-05, + "loss": 0.5139, + "step": 11050 + }, + { + "epoch": 0.42, + "learning_rate": 4.294626827391639e-05, + "loss": 0.514, + "step": 11051 + }, + { + "epoch": 0.42, + "learning_rate": 4.2945627083867664e-05, + "loss": 0.5644, + "step": 11052 + }, + { + "epoch": 0.42, + "learning_rate": 4.294498589381893e-05, + "loss": 0.5593, + "step": 11053 + }, + { + "epoch": 0.42, + "learning_rate": 4.29443447037702e-05, + "loss": 0.7057, + "step": 11054 + }, + { + "epoch": 0.43, + "learning_rate": 4.2943703513721466e-05, + "loss": 0.513, + "step": 11055 + }, + { + "epoch": 0.43, + "learning_rate": 4.294306232367274e-05, + "loss": 0.4898, + "step": 11056 + }, + { + "epoch": 0.43, + "learning_rate": 4.2942421133624e-05, + "loss": 0.4919, + "step": 11057 + }, + { + "epoch": 0.43, + "learning_rate": 4.2941779943575274e-05, + "loss": 0.5887, + "step": 11058 + }, + { + "epoch": 0.43, + "learning_rate": 4.294113875352655e-05, + "loss": 0.5578, + "step": 11059 + }, + { + "epoch": 0.43, + "learning_rate": 4.294049756347782e-05, + "loss": 0.4903, + "step": 11060 + }, + { + "epoch": 0.43, + "learning_rate": 4.293985637342909e-05, + "loss": 0.5798, + "step": 11061 + }, + { + "epoch": 0.43, + "learning_rate": 4.2939215183380354e-05, + "loss": 0.5505, + "step": 11062 + }, + { + "epoch": 0.43, + "learning_rate": 4.2938573993331626e-05, + "loss": 0.5847, + "step": 11063 + }, + { + "epoch": 0.43, + "learning_rate": 4.293793280328289e-05, + "loss": 0.5069, + "step": 11064 + }, + { + "epoch": 0.43, + "learning_rate": 4.293729161323416e-05, + "loss": 0.5128, + "step": 11065 + }, + { + "epoch": 0.43, + "learning_rate": 4.2936650423185434e-05, + "loss": 0.5794, + "step": 11066 + }, + { + "epoch": 0.43, + "learning_rate": 4.2936009233136706e-05, + "loss": 0.5605, + "step": 11067 + }, + { + "epoch": 0.43, + "learning_rate": 4.293536804308798e-05, + "loss": 0.493, + "step": 11068 + }, + { + "epoch": 0.43, + "learning_rate": 4.293472685303924e-05, + "loss": 0.5111, + "step": 11069 + }, + { + "epoch": 0.43, + "learning_rate": 4.2934085662990515e-05, + "loss": 0.6134, + "step": 11070 + }, + { + "epoch": 0.43, + "learning_rate": 4.293344447294178e-05, + "loss": 0.591, + "step": 11071 + }, + { + "epoch": 0.43, + "learning_rate": 4.293280328289305e-05, + "loss": 0.5064, + "step": 11072 + }, + { + "epoch": 0.43, + "learning_rate": 4.2932162092844316e-05, + "loss": 0.6081, + "step": 11073 + }, + { + "epoch": 0.43, + "learning_rate": 4.293152090279559e-05, + "loss": 0.5016, + "step": 11074 + }, + { + "epoch": 0.43, + "learning_rate": 4.293087971274686e-05, + "loss": 0.6032, + "step": 11075 + }, + { + "epoch": 0.43, + "learning_rate": 4.293023852269813e-05, + "loss": 0.7387, + "step": 11076 + }, + { + "epoch": 0.43, + "learning_rate": 4.29295973326494e-05, + "loss": 0.5303, + "step": 11077 + }, + { + "epoch": 0.43, + "learning_rate": 4.292895614260067e-05, + "loss": 0.472, + "step": 11078 + }, + { + "epoch": 0.43, + "learning_rate": 4.292831495255194e-05, + "loss": 0.6947, + "step": 11079 + }, + { + "epoch": 0.43, + "learning_rate": 4.2927673762503205e-05, + "loss": 0.6303, + "step": 11080 + }, + { + "epoch": 0.43, + "learning_rate": 4.2927032572454476e-05, + "loss": 0.6243, + "step": 11081 + }, + { + "epoch": 0.43, + "learning_rate": 4.292639138240575e-05, + "loss": 0.5504, + "step": 11082 + }, + { + "epoch": 0.43, + "learning_rate": 4.292575019235702e-05, + "loss": 0.5685, + "step": 11083 + }, + { + "epoch": 0.43, + "learning_rate": 4.2925109002308285e-05, + "loss": 0.5541, + "step": 11084 + }, + { + "epoch": 0.43, + "learning_rate": 4.2924467812259556e-05, + "loss": 0.5258, + "step": 11085 + }, + { + "epoch": 0.43, + "learning_rate": 4.292382662221083e-05, + "loss": 0.5249, + "step": 11086 + }, + { + "epoch": 0.43, + "learning_rate": 4.292318543216209e-05, + "loss": 0.5669, + "step": 11087 + }, + { + "epoch": 0.43, + "learning_rate": 4.2922544242113365e-05, + "loss": 0.5418, + "step": 11088 + }, + { + "epoch": 0.43, + "learning_rate": 4.292190305206463e-05, + "loss": 0.6463, + "step": 11089 + }, + { + "epoch": 0.43, + "learning_rate": 4.29212618620159e-05, + "loss": 0.647, + "step": 11090 + }, + { + "epoch": 0.43, + "learning_rate": 4.292062067196717e-05, + "loss": 0.4851, + "step": 11091 + }, + { + "epoch": 0.43, + "learning_rate": 4.2919979481918445e-05, + "loss": 0.5483, + "step": 11092 + }, + { + "epoch": 0.43, + "learning_rate": 4.291933829186971e-05, + "loss": 0.6409, + "step": 11093 + }, + { + "epoch": 0.43, + "learning_rate": 4.291869710182098e-05, + "loss": 0.6248, + "step": 11094 + }, + { + "epoch": 0.43, + "learning_rate": 4.291805591177225e-05, + "loss": 0.5995, + "step": 11095 + }, + { + "epoch": 0.43, + "learning_rate": 4.291741472172352e-05, + "loss": 0.6039, + "step": 11096 + }, + { + "epoch": 0.43, + "learning_rate": 4.291677353167479e-05, + "loss": 0.5248, + "step": 11097 + }, + { + "epoch": 0.43, + "learning_rate": 4.291613234162606e-05, + "loss": 0.5657, + "step": 11098 + }, + { + "epoch": 0.43, + "learning_rate": 4.291549115157733e-05, + "loss": 0.6847, + "step": 11099 + }, + { + "epoch": 0.43, + "learning_rate": 4.29148499615286e-05, + "loss": 0.4837, + "step": 11100 + }, + { + "epoch": 0.43, + "learning_rate": 4.291420877147987e-05, + "loss": 0.5889, + "step": 11101 + }, + { + "epoch": 0.43, + "learning_rate": 4.2913567581431135e-05, + "loss": 0.5673, + "step": 11102 + }, + { + "epoch": 0.43, + "learning_rate": 4.291292639138241e-05, + "loss": 0.5314, + "step": 11103 + }, + { + "epoch": 0.43, + "learning_rate": 4.291228520133368e-05, + "loss": 0.5969, + "step": 11104 + }, + { + "epoch": 0.43, + "learning_rate": 4.291164401128494e-05, + "loss": 0.5127, + "step": 11105 + }, + { + "epoch": 0.43, + "learning_rate": 4.2911002821236215e-05, + "loss": 0.5292, + "step": 11106 + }, + { + "epoch": 0.43, + "learning_rate": 4.291036163118749e-05, + "loss": 0.4638, + "step": 11107 + }, + { + "epoch": 0.43, + "learning_rate": 4.290972044113876e-05, + "loss": 0.5354, + "step": 11108 + }, + { + "epoch": 0.43, + "learning_rate": 4.2909079251090023e-05, + "loss": 0.4726, + "step": 11109 + }, + { + "epoch": 0.43, + "learning_rate": 4.2908438061041295e-05, + "loss": 0.5344, + "step": 11110 + }, + { + "epoch": 0.43, + "learning_rate": 4.290779687099256e-05, + "loss": 0.6269, + "step": 11111 + }, + { + "epoch": 0.43, + "learning_rate": 4.290715568094383e-05, + "loss": 0.5551, + "step": 11112 + }, + { + "epoch": 0.43, + "learning_rate": 4.2906514490895104e-05, + "loss": 0.4903, + "step": 11113 + }, + { + "epoch": 0.43, + "learning_rate": 4.2905873300846375e-05, + "loss": 0.5867, + "step": 11114 + }, + { + "epoch": 0.43, + "learning_rate": 4.290523211079765e-05, + "loss": 0.5019, + "step": 11115 + }, + { + "epoch": 0.43, + "learning_rate": 4.290459092074891e-05, + "loss": 0.6204, + "step": 11116 + }, + { + "epoch": 0.43, + "learning_rate": 4.2903949730700184e-05, + "loss": 0.5753, + "step": 11117 + }, + { + "epoch": 0.43, + "learning_rate": 4.290330854065145e-05, + "loss": 0.617, + "step": 11118 + }, + { + "epoch": 0.43, + "learning_rate": 4.290266735060272e-05, + "loss": 0.556, + "step": 11119 + }, + { + "epoch": 0.43, + "learning_rate": 4.2902026160553985e-05, + "loss": 0.6244, + "step": 11120 + }, + { + "epoch": 0.43, + "learning_rate": 4.290138497050526e-05, + "loss": 0.6312, + "step": 11121 + }, + { + "epoch": 0.43, + "learning_rate": 4.290074378045653e-05, + "loss": 0.5162, + "step": 11122 + }, + { + "epoch": 0.43, + "learning_rate": 4.29001025904078e-05, + "loss": 0.5251, + "step": 11123 + }, + { + "epoch": 0.43, + "learning_rate": 4.289946140035907e-05, + "loss": 0.6254, + "step": 11124 + }, + { + "epoch": 0.43, + "learning_rate": 4.289882021031034e-05, + "loss": 0.5756, + "step": 11125 + }, + { + "epoch": 0.43, + "learning_rate": 4.289817902026161e-05, + "loss": 0.5488, + "step": 11126 + }, + { + "epoch": 0.43, + "learning_rate": 4.2897537830212874e-05, + "loss": 0.4786, + "step": 11127 + }, + { + "epoch": 0.43, + "learning_rate": 4.2896896640164145e-05, + "loss": 0.4566, + "step": 11128 + }, + { + "epoch": 0.43, + "learning_rate": 4.289625545011542e-05, + "loss": 0.6067, + "step": 11129 + }, + { + "epoch": 0.43, + "learning_rate": 4.289561426006669e-05, + "loss": 0.5364, + "step": 11130 + }, + { + "epoch": 0.43, + "learning_rate": 4.289497307001796e-05, + "loss": 0.505, + "step": 11131 + }, + { + "epoch": 0.43, + "learning_rate": 4.2894331879969225e-05, + "loss": 0.5004, + "step": 11132 + }, + { + "epoch": 0.43, + "learning_rate": 4.28936906899205e-05, + "loss": 0.5936, + "step": 11133 + }, + { + "epoch": 0.43, + "learning_rate": 4.289304949987176e-05, + "loss": 0.5622, + "step": 11134 + }, + { + "epoch": 0.43, + "learning_rate": 4.2892408309823034e-05, + "loss": 0.4841, + "step": 11135 + }, + { + "epoch": 0.43, + "learning_rate": 4.28917671197743e-05, + "loss": 0.5355, + "step": 11136 + }, + { + "epoch": 0.43, + "learning_rate": 4.289112592972557e-05, + "loss": 0.6184, + "step": 11137 + }, + { + "epoch": 0.43, + "learning_rate": 4.289048473967684e-05, + "loss": 0.5304, + "step": 11138 + }, + { + "epoch": 0.43, + "learning_rate": 4.2889843549628114e-05, + "loss": 0.4825, + "step": 11139 + }, + { + "epoch": 0.43, + "learning_rate": 4.2889202359579386e-05, + "loss": 0.5995, + "step": 11140 + }, + { + "epoch": 0.43, + "learning_rate": 4.288856116953065e-05, + "loss": 0.6034, + "step": 11141 + }, + { + "epoch": 0.43, + "learning_rate": 4.288791997948192e-05, + "loss": 0.5195, + "step": 11142 + }, + { + "epoch": 0.43, + "learning_rate": 4.288727878943319e-05, + "loss": 0.6459, + "step": 11143 + }, + { + "epoch": 0.43, + "learning_rate": 4.288663759938446e-05, + "loss": 0.5798, + "step": 11144 + }, + { + "epoch": 0.43, + "learning_rate": 4.2885996409335724e-05, + "loss": 0.5879, + "step": 11145 + }, + { + "epoch": 0.43, + "learning_rate": 4.2885355219286996e-05, + "loss": 0.4671, + "step": 11146 + }, + { + "epoch": 0.43, + "learning_rate": 4.2884714029238274e-05, + "loss": 0.5167, + "step": 11147 + }, + { + "epoch": 0.43, + "learning_rate": 4.288407283918954e-05, + "loss": 0.5844, + "step": 11148 + }, + { + "epoch": 0.43, + "learning_rate": 4.288343164914081e-05, + "loss": 0.594, + "step": 11149 + }, + { + "epoch": 0.43, + "learning_rate": 4.2882790459092076e-05, + "loss": 0.5277, + "step": 11150 + }, + { + "epoch": 0.43, + "learning_rate": 4.288214926904335e-05, + "loss": 0.466, + "step": 11151 + }, + { + "epoch": 0.43, + "learning_rate": 4.288150807899461e-05, + "loss": 0.5534, + "step": 11152 + }, + { + "epoch": 0.43, + "learning_rate": 4.2880866888945884e-05, + "loss": 0.5855, + "step": 11153 + }, + { + "epoch": 0.43, + "learning_rate": 4.2880225698897156e-05, + "loss": 0.497, + "step": 11154 + }, + { + "epoch": 0.43, + "learning_rate": 4.287958450884843e-05, + "loss": 0.5034, + "step": 11155 + }, + { + "epoch": 0.43, + "learning_rate": 4.28789433187997e-05, + "loss": 0.6233, + "step": 11156 + }, + { + "epoch": 0.43, + "learning_rate": 4.2878302128750964e-05, + "loss": 0.5586, + "step": 11157 + }, + { + "epoch": 0.43, + "learning_rate": 4.2877660938702236e-05, + "loss": 0.6395, + "step": 11158 + }, + { + "epoch": 0.43, + "learning_rate": 4.28770197486535e-05, + "loss": 0.6516, + "step": 11159 + }, + { + "epoch": 0.43, + "learning_rate": 4.287637855860477e-05, + "loss": 0.53, + "step": 11160 + }, + { + "epoch": 0.43, + "learning_rate": 4.287573736855604e-05, + "loss": 0.5122, + "step": 11161 + }, + { + "epoch": 0.43, + "learning_rate": 4.287509617850731e-05, + "loss": 0.5685, + "step": 11162 + }, + { + "epoch": 0.43, + "learning_rate": 4.287445498845858e-05, + "loss": 0.4702, + "step": 11163 + }, + { + "epoch": 0.43, + "learning_rate": 4.287381379840985e-05, + "loss": 0.5564, + "step": 11164 + }, + { + "epoch": 0.43, + "learning_rate": 4.2873172608361124e-05, + "loss": 0.5224, + "step": 11165 + }, + { + "epoch": 0.43, + "learning_rate": 4.287253141831239e-05, + "loss": 0.4554, + "step": 11166 + }, + { + "epoch": 0.43, + "learning_rate": 4.287189022826366e-05, + "loss": 0.594, + "step": 11167 + }, + { + "epoch": 0.43, + "learning_rate": 4.2871249038214926e-05, + "loss": 0.5927, + "step": 11168 + }, + { + "epoch": 0.43, + "learning_rate": 4.28706078481662e-05, + "loss": 0.5357, + "step": 11169 + }, + { + "epoch": 0.43, + "learning_rate": 4.286996665811747e-05, + "loss": 0.4918, + "step": 11170 + }, + { + "epoch": 0.43, + "learning_rate": 4.286932546806874e-05, + "loss": 0.4914, + "step": 11171 + }, + { + "epoch": 0.43, + "learning_rate": 4.2868684278020006e-05, + "loss": 0.6033, + "step": 11172 + }, + { + "epoch": 0.43, + "learning_rate": 4.286804308797128e-05, + "loss": 0.5316, + "step": 11173 + }, + { + "epoch": 0.43, + "learning_rate": 4.286740189792255e-05, + "loss": 0.581, + "step": 11174 + }, + { + "epoch": 0.43, + "learning_rate": 4.2866760707873814e-05, + "loss": 0.4881, + "step": 11175 + }, + { + "epoch": 0.43, + "learning_rate": 4.2866119517825086e-05, + "loss": 0.6143, + "step": 11176 + }, + { + "epoch": 0.43, + "learning_rate": 4.286547832777635e-05, + "loss": 0.6842, + "step": 11177 + }, + { + "epoch": 0.43, + "learning_rate": 4.286483713772762e-05, + "loss": 0.6155, + "step": 11178 + }, + { + "epoch": 0.43, + "learning_rate": 4.2864195947678895e-05, + "loss": 0.4912, + "step": 11179 + }, + { + "epoch": 0.43, + "learning_rate": 4.2863554757630166e-05, + "loss": 0.7158, + "step": 11180 + }, + { + "epoch": 0.43, + "learning_rate": 4.286291356758143e-05, + "loss": 0.6348, + "step": 11181 + }, + { + "epoch": 0.43, + "learning_rate": 4.28622723775327e-05, + "loss": 0.4188, + "step": 11182 + }, + { + "epoch": 0.43, + "learning_rate": 4.2861631187483975e-05, + "loss": 0.5362, + "step": 11183 + }, + { + "epoch": 0.43, + "learning_rate": 4.286098999743524e-05, + "loss": 0.5731, + "step": 11184 + }, + { + "epoch": 0.43, + "learning_rate": 4.286034880738651e-05, + "loss": 0.5777, + "step": 11185 + }, + { + "epoch": 0.43, + "learning_rate": 4.285970761733778e-05, + "loss": 0.5484, + "step": 11186 + }, + { + "epoch": 0.43, + "learning_rate": 4.2859066427289055e-05, + "loss": 0.5769, + "step": 11187 + }, + { + "epoch": 0.43, + "learning_rate": 4.285842523724032e-05, + "loss": 0.6036, + "step": 11188 + }, + { + "epoch": 0.43, + "learning_rate": 4.285778404719159e-05, + "loss": 0.5224, + "step": 11189 + }, + { + "epoch": 0.43, + "learning_rate": 4.2857142857142856e-05, + "loss": 0.5294, + "step": 11190 + }, + { + "epoch": 0.43, + "learning_rate": 4.285650166709413e-05, + "loss": 0.5747, + "step": 11191 + }, + { + "epoch": 0.43, + "learning_rate": 4.28558604770454e-05, + "loss": 0.5925, + "step": 11192 + }, + { + "epoch": 0.43, + "learning_rate": 4.2855219286996665e-05, + "loss": 0.5203, + "step": 11193 + }, + { + "epoch": 0.43, + "learning_rate": 4.2854578096947936e-05, + "loss": 0.5146, + "step": 11194 + }, + { + "epoch": 0.43, + "learning_rate": 4.285393690689921e-05, + "loss": 0.5188, + "step": 11195 + }, + { + "epoch": 0.43, + "learning_rate": 4.285329571685048e-05, + "loss": 0.585, + "step": 11196 + }, + { + "epoch": 0.43, + "learning_rate": 4.2852654526801745e-05, + "loss": 0.5075, + "step": 11197 + }, + { + "epoch": 0.43, + "learning_rate": 4.2852013336753016e-05, + "loss": 0.548, + "step": 11198 + }, + { + "epoch": 0.43, + "learning_rate": 4.285137214670428e-05, + "loss": 0.5475, + "step": 11199 + }, + { + "epoch": 0.43, + "learning_rate": 4.285073095665555e-05, + "loss": 0.4887, + "step": 11200 + }, + { + "epoch": 0.43, + "learning_rate": 4.2850089766606825e-05, + "loss": 0.4676, + "step": 11201 + }, + { + "epoch": 0.43, + "learning_rate": 4.2849448576558097e-05, + "loss": 0.5351, + "step": 11202 + }, + { + "epoch": 0.43, + "learning_rate": 4.284880738650937e-05, + "loss": 0.5457, + "step": 11203 + }, + { + "epoch": 0.43, + "learning_rate": 4.284816619646063e-05, + "loss": 0.529, + "step": 11204 + }, + { + "epoch": 0.43, + "learning_rate": 4.2847525006411905e-05, + "loss": 0.5363, + "step": 11205 + }, + { + "epoch": 0.43, + "learning_rate": 4.284688381636317e-05, + "loss": 0.5455, + "step": 11206 + }, + { + "epoch": 0.43, + "learning_rate": 4.284624262631444e-05, + "loss": 0.5937, + "step": 11207 + }, + { + "epoch": 0.43, + "learning_rate": 4.2845601436265707e-05, + "loss": 0.7407, + "step": 11208 + }, + { + "epoch": 0.43, + "learning_rate": 4.284496024621698e-05, + "loss": 0.6444, + "step": 11209 + }, + { + "epoch": 0.43, + "learning_rate": 4.284431905616825e-05, + "loss": 0.5066, + "step": 11210 + }, + { + "epoch": 0.43, + "learning_rate": 4.284367786611952e-05, + "loss": 0.5948, + "step": 11211 + }, + { + "epoch": 0.43, + "learning_rate": 4.2843036676070793e-05, + "loss": 0.562, + "step": 11212 + }, + { + "epoch": 0.43, + "learning_rate": 4.284239548602206e-05, + "loss": 0.6305, + "step": 11213 + }, + { + "epoch": 0.43, + "learning_rate": 4.284175429597333e-05, + "loss": 0.5374, + "step": 11214 + }, + { + "epoch": 0.43, + "learning_rate": 4.2841113105924595e-05, + "loss": 0.6195, + "step": 11215 + }, + { + "epoch": 0.43, + "learning_rate": 4.284047191587587e-05, + "loss": 0.5279, + "step": 11216 + }, + { + "epoch": 0.43, + "learning_rate": 4.283983072582713e-05, + "loss": 0.5282, + "step": 11217 + }, + { + "epoch": 0.43, + "learning_rate": 4.283918953577841e-05, + "loss": 0.5627, + "step": 11218 + }, + { + "epoch": 0.43, + "learning_rate": 4.283854834572968e-05, + "loss": 0.6113, + "step": 11219 + }, + { + "epoch": 0.43, + "learning_rate": 4.283790715568095e-05, + "loss": 0.604, + "step": 11220 + }, + { + "epoch": 0.43, + "learning_rate": 4.283726596563222e-05, + "loss": 0.6501, + "step": 11221 + }, + { + "epoch": 0.43, + "learning_rate": 4.2836624775583483e-05, + "loss": 0.5595, + "step": 11222 + }, + { + "epoch": 0.43, + "learning_rate": 4.2835983585534755e-05, + "loss": 0.6428, + "step": 11223 + }, + { + "epoch": 0.43, + "learning_rate": 4.283534239548602e-05, + "loss": 0.5217, + "step": 11224 + }, + { + "epoch": 0.43, + "learning_rate": 4.283470120543729e-05, + "loss": 0.5595, + "step": 11225 + }, + { + "epoch": 0.43, + "learning_rate": 4.2834060015388564e-05, + "loss": 0.5832, + "step": 11226 + }, + { + "epoch": 0.43, + "learning_rate": 4.2833418825339835e-05, + "loss": 0.4727, + "step": 11227 + }, + { + "epoch": 0.43, + "learning_rate": 4.283277763529111e-05, + "loss": 0.569, + "step": 11228 + }, + { + "epoch": 0.43, + "learning_rate": 4.283213644524237e-05, + "loss": 0.6637, + "step": 11229 + }, + { + "epoch": 0.43, + "learning_rate": 4.2831495255193644e-05, + "loss": 0.5427, + "step": 11230 + }, + { + "epoch": 0.43, + "learning_rate": 4.283085406514491e-05, + "loss": 0.5467, + "step": 11231 + }, + { + "epoch": 0.43, + "learning_rate": 4.283021287509618e-05, + "loss": 0.575, + "step": 11232 + }, + { + "epoch": 0.43, + "learning_rate": 4.2829571685047445e-05, + "loss": 0.5855, + "step": 11233 + }, + { + "epoch": 0.43, + "learning_rate": 4.282893049499872e-05, + "loss": 0.4777, + "step": 11234 + }, + { + "epoch": 0.43, + "learning_rate": 4.282828930494999e-05, + "loss": 0.5991, + "step": 11235 + }, + { + "epoch": 0.43, + "learning_rate": 4.282764811490126e-05, + "loss": 0.574, + "step": 11236 + }, + { + "epoch": 0.43, + "learning_rate": 4.282700692485253e-05, + "loss": 0.6401, + "step": 11237 + }, + { + "epoch": 0.43, + "learning_rate": 4.28263657348038e-05, + "loss": 0.5819, + "step": 11238 + }, + { + "epoch": 0.43, + "learning_rate": 4.282572454475507e-05, + "loss": 0.5437, + "step": 11239 + }, + { + "epoch": 0.43, + "learning_rate": 4.2825083354706334e-05, + "loss": 0.4904, + "step": 11240 + }, + { + "epoch": 0.43, + "learning_rate": 4.2824442164657605e-05, + "loss": 0.6335, + "step": 11241 + }, + { + "epoch": 0.43, + "learning_rate": 4.282380097460888e-05, + "loss": 0.5966, + "step": 11242 + }, + { + "epoch": 0.43, + "learning_rate": 4.282315978456015e-05, + "loss": 0.6254, + "step": 11243 + }, + { + "epoch": 0.43, + "learning_rate": 4.2822518594511414e-05, + "loss": 0.6087, + "step": 11244 + }, + { + "epoch": 0.43, + "learning_rate": 4.2821877404462686e-05, + "loss": 0.5931, + "step": 11245 + }, + { + "epoch": 0.43, + "learning_rate": 4.282123621441396e-05, + "loss": 0.5585, + "step": 11246 + }, + { + "epoch": 0.43, + "learning_rate": 4.282059502436522e-05, + "loss": 0.6203, + "step": 11247 + }, + { + "epoch": 0.43, + "learning_rate": 4.2819953834316494e-05, + "loss": 0.4781, + "step": 11248 + }, + { + "epoch": 0.43, + "learning_rate": 4.281931264426776e-05, + "loss": 0.455, + "step": 11249 + }, + { + "epoch": 0.43, + "learning_rate": 4.281867145421903e-05, + "loss": 0.6225, + "step": 11250 + }, + { + "epoch": 0.43, + "learning_rate": 4.28180302641703e-05, + "loss": 0.5466, + "step": 11251 + }, + { + "epoch": 0.43, + "learning_rate": 4.2817389074121574e-05, + "loss": 0.5672, + "step": 11252 + }, + { + "epoch": 0.43, + "learning_rate": 4.281674788407284e-05, + "loss": 0.4969, + "step": 11253 + }, + { + "epoch": 0.43, + "learning_rate": 4.281610669402411e-05, + "loss": 0.6054, + "step": 11254 + }, + { + "epoch": 0.43, + "learning_rate": 4.281546550397538e-05, + "loss": 0.5643, + "step": 11255 + }, + { + "epoch": 0.43, + "learning_rate": 4.281482431392665e-05, + "loss": 0.6434, + "step": 11256 + }, + { + "epoch": 0.43, + "learning_rate": 4.281418312387792e-05, + "loss": 0.4948, + "step": 11257 + }, + { + "epoch": 0.43, + "learning_rate": 4.281354193382919e-05, + "loss": 0.5762, + "step": 11258 + }, + { + "epoch": 0.43, + "learning_rate": 4.281290074378046e-05, + "loss": 0.5849, + "step": 11259 + }, + { + "epoch": 0.43, + "learning_rate": 4.281225955373173e-05, + "loss": 0.5353, + "step": 11260 + }, + { + "epoch": 0.43, + "learning_rate": 4.2811618363683e-05, + "loss": 0.5824, + "step": 11261 + }, + { + "epoch": 0.43, + "learning_rate": 4.2810977173634264e-05, + "loss": 0.5569, + "step": 11262 + }, + { + "epoch": 0.43, + "learning_rate": 4.2810335983585536e-05, + "loss": 0.6416, + "step": 11263 + }, + { + "epoch": 0.43, + "learning_rate": 4.280969479353681e-05, + "loss": 0.6957, + "step": 11264 + }, + { + "epoch": 0.43, + "learning_rate": 4.280905360348807e-05, + "loss": 0.4966, + "step": 11265 + }, + { + "epoch": 0.43, + "learning_rate": 4.2808412413439344e-05, + "loss": 0.4774, + "step": 11266 + }, + { + "epoch": 0.43, + "learning_rate": 4.2807771223390616e-05, + "loss": 0.5691, + "step": 11267 + }, + { + "epoch": 0.43, + "learning_rate": 4.280713003334189e-05, + "loss": 0.6111, + "step": 11268 + }, + { + "epoch": 0.43, + "learning_rate": 4.280648884329315e-05, + "loss": 0.5371, + "step": 11269 + }, + { + "epoch": 0.43, + "learning_rate": 4.2805847653244424e-05, + "loss": 0.4788, + "step": 11270 + }, + { + "epoch": 0.43, + "learning_rate": 4.280520646319569e-05, + "loss": 0.5586, + "step": 11271 + }, + { + "epoch": 0.43, + "learning_rate": 4.280456527314696e-05, + "loss": 0.6079, + "step": 11272 + }, + { + "epoch": 0.43, + "learning_rate": 4.280392408309823e-05, + "loss": 0.5132, + "step": 11273 + }, + { + "epoch": 0.43, + "learning_rate": 4.2803282893049504e-05, + "loss": 0.5018, + "step": 11274 + }, + { + "epoch": 0.43, + "learning_rate": 4.2802641703000776e-05, + "loss": 0.5953, + "step": 11275 + }, + { + "epoch": 0.43, + "learning_rate": 4.280200051295204e-05, + "loss": 0.5301, + "step": 11276 + }, + { + "epoch": 0.43, + "learning_rate": 4.280135932290331e-05, + "loss": 0.5693, + "step": 11277 + }, + { + "epoch": 0.43, + "learning_rate": 4.280071813285458e-05, + "loss": 0.5405, + "step": 11278 + }, + { + "epoch": 0.43, + "learning_rate": 4.280007694280585e-05, + "loss": 0.5386, + "step": 11279 + }, + { + "epoch": 0.43, + "learning_rate": 4.2799435752757114e-05, + "loss": 0.569, + "step": 11280 + }, + { + "epoch": 0.43, + "learning_rate": 4.2798794562708386e-05, + "loss": 0.6519, + "step": 11281 + }, + { + "epoch": 0.43, + "learning_rate": 4.279815337265966e-05, + "loss": 0.5845, + "step": 11282 + }, + { + "epoch": 0.43, + "learning_rate": 4.279751218261093e-05, + "loss": 0.5121, + "step": 11283 + }, + { + "epoch": 0.43, + "learning_rate": 4.27968709925622e-05, + "loss": 0.5542, + "step": 11284 + }, + { + "epoch": 0.43, + "learning_rate": 4.2796229802513466e-05, + "loss": 0.5794, + "step": 11285 + }, + { + "epoch": 0.43, + "learning_rate": 4.279558861246474e-05, + "loss": 0.6672, + "step": 11286 + }, + { + "epoch": 0.43, + "learning_rate": 4.2794947422416e-05, + "loss": 0.5052, + "step": 11287 + }, + { + "epoch": 0.43, + "learning_rate": 4.2794306232367274e-05, + "loss": 0.5979, + "step": 11288 + }, + { + "epoch": 0.43, + "learning_rate": 4.279366504231854e-05, + "loss": 0.4896, + "step": 11289 + }, + { + "epoch": 0.43, + "learning_rate": 4.279302385226982e-05, + "loss": 0.5891, + "step": 11290 + }, + { + "epoch": 0.43, + "learning_rate": 4.279238266222109e-05, + "loss": 0.5888, + "step": 11291 + }, + { + "epoch": 0.43, + "learning_rate": 4.2791741472172355e-05, + "loss": 0.5956, + "step": 11292 + }, + { + "epoch": 0.43, + "learning_rate": 4.2791100282123626e-05, + "loss": 0.5527, + "step": 11293 + }, + { + "epoch": 0.43, + "learning_rate": 4.279045909207489e-05, + "loss": 0.5329, + "step": 11294 + }, + { + "epoch": 0.43, + "learning_rate": 4.278981790202616e-05, + "loss": 0.564, + "step": 11295 + }, + { + "epoch": 0.43, + "learning_rate": 4.278917671197743e-05, + "loss": 0.5376, + "step": 11296 + }, + { + "epoch": 0.43, + "learning_rate": 4.27885355219287e-05, + "loss": 0.5176, + "step": 11297 + }, + { + "epoch": 0.43, + "learning_rate": 4.278789433187997e-05, + "loss": 0.6279, + "step": 11298 + }, + { + "epoch": 0.43, + "learning_rate": 4.278725314183124e-05, + "loss": 0.6354, + "step": 11299 + }, + { + "epoch": 0.43, + "learning_rate": 4.2786611951782515e-05, + "loss": 0.5443, + "step": 11300 + }, + { + "epoch": 0.43, + "learning_rate": 4.278597076173378e-05, + "loss": 0.4738, + "step": 11301 + }, + { + "epoch": 0.43, + "learning_rate": 4.278532957168505e-05, + "loss": 0.5253, + "step": 11302 + }, + { + "epoch": 0.43, + "learning_rate": 4.2784688381636316e-05, + "loss": 0.5681, + "step": 11303 + }, + { + "epoch": 0.43, + "learning_rate": 4.278404719158759e-05, + "loss": 0.5697, + "step": 11304 + }, + { + "epoch": 0.43, + "learning_rate": 4.278340600153885e-05, + "loss": 0.5295, + "step": 11305 + }, + { + "epoch": 0.43, + "learning_rate": 4.2782764811490125e-05, + "loss": 0.582, + "step": 11306 + }, + { + "epoch": 0.43, + "learning_rate": 4.2782123621441396e-05, + "loss": 0.4803, + "step": 11307 + }, + { + "epoch": 0.43, + "learning_rate": 4.278148243139267e-05, + "loss": 0.5312, + "step": 11308 + }, + { + "epoch": 0.43, + "learning_rate": 4.278084124134394e-05, + "loss": 0.5483, + "step": 11309 + }, + { + "epoch": 0.43, + "learning_rate": 4.2780200051295205e-05, + "loss": 0.5792, + "step": 11310 + }, + { + "epoch": 0.43, + "learning_rate": 4.2779558861246477e-05, + "loss": 0.5223, + "step": 11311 + }, + { + "epoch": 0.43, + "learning_rate": 4.277891767119774e-05, + "loss": 0.5228, + "step": 11312 + }, + { + "epoch": 0.43, + "learning_rate": 4.277827648114901e-05, + "loss": 0.6864, + "step": 11313 + }, + { + "epoch": 0.43, + "learning_rate": 4.2777635291100285e-05, + "loss": 0.5994, + "step": 11314 + }, + { + "epoch": 0.44, + "learning_rate": 4.2776994101051557e-05, + "loss": 0.5511, + "step": 11315 + }, + { + "epoch": 0.44, + "learning_rate": 4.277635291100283e-05, + "loss": 0.6041, + "step": 11316 + }, + { + "epoch": 0.44, + "learning_rate": 4.277571172095409e-05, + "loss": 0.5759, + "step": 11317 + }, + { + "epoch": 0.44, + "learning_rate": 4.2775070530905365e-05, + "loss": 0.5936, + "step": 11318 + }, + { + "epoch": 0.44, + "learning_rate": 4.277442934085663e-05, + "loss": 0.5937, + "step": 11319 + }, + { + "epoch": 0.44, + "learning_rate": 4.27737881508079e-05, + "loss": 0.6039, + "step": 11320 + }, + { + "epoch": 0.44, + "learning_rate": 4.2773146960759167e-05, + "loss": 0.5569, + "step": 11321 + }, + { + "epoch": 0.44, + "learning_rate": 4.277250577071044e-05, + "loss": 0.6375, + "step": 11322 + }, + { + "epoch": 0.44, + "learning_rate": 4.277186458066171e-05, + "loss": 0.4749, + "step": 11323 + }, + { + "epoch": 0.44, + "learning_rate": 4.277122339061298e-05, + "loss": 0.5635, + "step": 11324 + }, + { + "epoch": 0.44, + "learning_rate": 4.2770582200564253e-05, + "loss": 0.5578, + "step": 11325 + }, + { + "epoch": 0.44, + "learning_rate": 4.276994101051552e-05, + "loss": 0.596, + "step": 11326 + }, + { + "epoch": 0.44, + "learning_rate": 4.276929982046679e-05, + "loss": 0.6014, + "step": 11327 + }, + { + "epoch": 0.44, + "learning_rate": 4.2768658630418055e-05, + "loss": 0.5758, + "step": 11328 + }, + { + "epoch": 0.44, + "learning_rate": 4.276801744036933e-05, + "loss": 0.6048, + "step": 11329 + }, + { + "epoch": 0.44, + "learning_rate": 4.27673762503206e-05, + "loss": 0.5664, + "step": 11330 + }, + { + "epoch": 0.44, + "learning_rate": 4.276673506027187e-05, + "loss": 0.4611, + "step": 11331 + }, + { + "epoch": 0.44, + "learning_rate": 4.2766093870223135e-05, + "loss": 0.5072, + "step": 11332 + }, + { + "epoch": 0.44, + "learning_rate": 4.276545268017441e-05, + "loss": 0.5889, + "step": 11333 + }, + { + "epoch": 0.44, + "learning_rate": 4.276481149012568e-05, + "loss": 0.5743, + "step": 11334 + }, + { + "epoch": 0.44, + "learning_rate": 4.2764170300076943e-05, + "loss": 0.6372, + "step": 11335 + }, + { + "epoch": 0.44, + "learning_rate": 4.2763529110028215e-05, + "loss": 0.5925, + "step": 11336 + }, + { + "epoch": 0.44, + "learning_rate": 4.276288791997948e-05, + "loss": 0.5643, + "step": 11337 + }, + { + "epoch": 0.44, + "learning_rate": 4.276224672993075e-05, + "loss": 0.4814, + "step": 11338 + }, + { + "epoch": 0.44, + "learning_rate": 4.2761605539882024e-05, + "loss": 0.5663, + "step": 11339 + }, + { + "epoch": 0.44, + "learning_rate": 4.2760964349833295e-05, + "loss": 0.5952, + "step": 11340 + }, + { + "epoch": 0.44, + "learning_rate": 4.276032315978456e-05, + "loss": 0.589, + "step": 11341 + }, + { + "epoch": 0.44, + "learning_rate": 4.275968196973583e-05, + "loss": 0.5065, + "step": 11342 + }, + { + "epoch": 0.44, + "learning_rate": 4.2759040779687104e-05, + "loss": 0.6021, + "step": 11343 + }, + { + "epoch": 0.44, + "learning_rate": 4.275839958963837e-05, + "loss": 0.5699, + "step": 11344 + }, + { + "epoch": 0.44, + "learning_rate": 4.275775839958964e-05, + "loss": 0.5276, + "step": 11345 + }, + { + "epoch": 0.44, + "learning_rate": 4.275711720954091e-05, + "loss": 0.5665, + "step": 11346 + }, + { + "epoch": 0.44, + "learning_rate": 4.2756476019492184e-05, + "loss": 0.5632, + "step": 11347 + }, + { + "epoch": 0.44, + "learning_rate": 4.275583482944345e-05, + "loss": 0.5265, + "step": 11348 + }, + { + "epoch": 0.44, + "learning_rate": 4.275519363939472e-05, + "loss": 0.5902, + "step": 11349 + }, + { + "epoch": 0.44, + "learning_rate": 4.2754552449345985e-05, + "loss": 0.5701, + "step": 11350 + }, + { + "epoch": 0.44, + "learning_rate": 4.275391125929726e-05, + "loss": 0.5928, + "step": 11351 + }, + { + "epoch": 0.44, + "learning_rate": 4.275327006924853e-05, + "loss": 0.539, + "step": 11352 + }, + { + "epoch": 0.44, + "learning_rate": 4.2752628879199794e-05, + "loss": 0.6312, + "step": 11353 + }, + { + "epoch": 0.44, + "learning_rate": 4.2751987689151065e-05, + "loss": 0.5332, + "step": 11354 + }, + { + "epoch": 0.44, + "learning_rate": 4.275134649910234e-05, + "loss": 0.5392, + "step": 11355 + }, + { + "epoch": 0.44, + "learning_rate": 4.275070530905361e-05, + "loss": 0.5779, + "step": 11356 + }, + { + "epoch": 0.44, + "learning_rate": 4.2750064119004874e-05, + "loss": 0.6987, + "step": 11357 + }, + { + "epoch": 0.44, + "learning_rate": 4.2749422928956146e-05, + "loss": 0.4968, + "step": 11358 + }, + { + "epoch": 0.44, + "learning_rate": 4.274878173890741e-05, + "loss": 0.5755, + "step": 11359 + }, + { + "epoch": 0.44, + "learning_rate": 4.274814054885868e-05, + "loss": 0.5835, + "step": 11360 + }, + { + "epoch": 0.44, + "learning_rate": 4.2747499358809954e-05, + "loss": 0.5987, + "step": 11361 + }, + { + "epoch": 0.44, + "learning_rate": 4.2746858168761226e-05, + "loss": 0.4533, + "step": 11362 + }, + { + "epoch": 0.44, + "learning_rate": 4.27462169787125e-05, + "loss": 0.6412, + "step": 11363 + }, + { + "epoch": 0.44, + "learning_rate": 4.274557578866376e-05, + "loss": 0.5339, + "step": 11364 + }, + { + "epoch": 0.44, + "learning_rate": 4.2744934598615034e-05, + "loss": 0.5425, + "step": 11365 + }, + { + "epoch": 0.44, + "learning_rate": 4.27442934085663e-05, + "loss": 0.5972, + "step": 11366 + }, + { + "epoch": 0.44, + "learning_rate": 4.274365221851757e-05, + "loss": 0.5525, + "step": 11367 + }, + { + "epoch": 0.44, + "learning_rate": 4.2743011028468836e-05, + "loss": 0.5526, + "step": 11368 + }, + { + "epoch": 0.44, + "learning_rate": 4.274236983842011e-05, + "loss": 0.5649, + "step": 11369 + }, + { + "epoch": 0.44, + "learning_rate": 4.274172864837138e-05, + "loss": 0.5552, + "step": 11370 + }, + { + "epoch": 0.44, + "learning_rate": 4.274108745832265e-05, + "loss": 0.5566, + "step": 11371 + }, + { + "epoch": 0.44, + "learning_rate": 4.274044626827392e-05, + "loss": 0.526, + "step": 11372 + }, + { + "epoch": 0.44, + "learning_rate": 4.273980507822519e-05, + "loss": 0.5253, + "step": 11373 + }, + { + "epoch": 0.44, + "learning_rate": 4.273916388817646e-05, + "loss": 0.6633, + "step": 11374 + }, + { + "epoch": 0.44, + "learning_rate": 4.2738522698127724e-05, + "loss": 0.6506, + "step": 11375 + }, + { + "epoch": 0.44, + "learning_rate": 4.2737881508078996e-05, + "loss": 0.5733, + "step": 11376 + }, + { + "epoch": 0.44, + "learning_rate": 4.273724031803026e-05, + "loss": 0.5197, + "step": 11377 + }, + { + "epoch": 0.44, + "learning_rate": 4.273659912798154e-05, + "loss": 0.6548, + "step": 11378 + }, + { + "epoch": 0.44, + "learning_rate": 4.273595793793281e-05, + "loss": 0.6122, + "step": 11379 + }, + { + "epoch": 0.44, + "learning_rate": 4.2735316747884076e-05, + "loss": 0.6094, + "step": 11380 + }, + { + "epoch": 0.44, + "learning_rate": 4.273467555783535e-05, + "loss": 0.6938, + "step": 11381 + }, + { + "epoch": 0.44, + "learning_rate": 4.273403436778661e-05, + "loss": 0.5638, + "step": 11382 + }, + { + "epoch": 0.44, + "learning_rate": 4.2733393177737884e-05, + "loss": 0.5944, + "step": 11383 + }, + { + "epoch": 0.44, + "learning_rate": 4.273275198768915e-05, + "loss": 0.501, + "step": 11384 + }, + { + "epoch": 0.44, + "learning_rate": 4.273211079764042e-05, + "loss": 0.5469, + "step": 11385 + }, + { + "epoch": 0.44, + "learning_rate": 4.273146960759169e-05, + "loss": 0.5736, + "step": 11386 + }, + { + "epoch": 0.44, + "learning_rate": 4.2730828417542964e-05, + "loss": 0.5883, + "step": 11387 + }, + { + "epoch": 0.44, + "learning_rate": 4.2730187227494236e-05, + "loss": 0.52, + "step": 11388 + }, + { + "epoch": 0.44, + "learning_rate": 4.27295460374455e-05, + "loss": 0.5344, + "step": 11389 + }, + { + "epoch": 0.44, + "learning_rate": 4.272890484739677e-05, + "loss": 0.5847, + "step": 11390 + }, + { + "epoch": 0.44, + "learning_rate": 4.272826365734804e-05, + "loss": 0.6458, + "step": 11391 + }, + { + "epoch": 0.44, + "learning_rate": 4.272762246729931e-05, + "loss": 0.475, + "step": 11392 + }, + { + "epoch": 0.44, + "learning_rate": 4.2726981277250574e-05, + "loss": 0.5855, + "step": 11393 + }, + { + "epoch": 0.44, + "learning_rate": 4.2726340087201846e-05, + "loss": 0.5574, + "step": 11394 + }, + { + "epoch": 0.44, + "learning_rate": 4.272569889715312e-05, + "loss": 0.5058, + "step": 11395 + }, + { + "epoch": 0.44, + "learning_rate": 4.272505770710439e-05, + "loss": 0.5836, + "step": 11396 + }, + { + "epoch": 0.44, + "learning_rate": 4.272441651705566e-05, + "loss": 0.5023, + "step": 11397 + }, + { + "epoch": 0.44, + "learning_rate": 4.2723775327006926e-05, + "loss": 0.4969, + "step": 11398 + }, + { + "epoch": 0.44, + "learning_rate": 4.27231341369582e-05, + "loss": 0.5636, + "step": 11399 + }, + { + "epoch": 0.44, + "learning_rate": 4.272249294690946e-05, + "loss": 0.5105, + "step": 11400 + }, + { + "epoch": 0.44, + "learning_rate": 4.2721851756860734e-05, + "loss": 0.5087, + "step": 11401 + }, + { + "epoch": 0.44, + "learning_rate": 4.2721210566812006e-05, + "loss": 0.5418, + "step": 11402 + }, + { + "epoch": 0.44, + "learning_rate": 4.272056937676328e-05, + "loss": 0.6475, + "step": 11403 + }, + { + "epoch": 0.44, + "learning_rate": 4.271992818671454e-05, + "loss": 0.5671, + "step": 11404 + }, + { + "epoch": 0.44, + "learning_rate": 4.2719286996665815e-05, + "loss": 0.5175, + "step": 11405 + }, + { + "epoch": 0.44, + "learning_rate": 4.2718645806617086e-05, + "loss": 0.5743, + "step": 11406 + }, + { + "epoch": 0.44, + "learning_rate": 4.271800461656835e-05, + "loss": 0.5975, + "step": 11407 + }, + { + "epoch": 0.44, + "learning_rate": 4.271736342651962e-05, + "loss": 0.5599, + "step": 11408 + }, + { + "epoch": 0.44, + "learning_rate": 4.271672223647089e-05, + "loss": 0.5838, + "step": 11409 + }, + { + "epoch": 0.44, + "learning_rate": 4.271608104642216e-05, + "loss": 0.5595, + "step": 11410 + }, + { + "epoch": 0.44, + "learning_rate": 4.271543985637343e-05, + "loss": 0.5926, + "step": 11411 + }, + { + "epoch": 0.44, + "learning_rate": 4.27147986663247e-05, + "loss": 0.5086, + "step": 11412 + }, + { + "epoch": 0.44, + "learning_rate": 4.271415747627597e-05, + "loss": 0.5113, + "step": 11413 + }, + { + "epoch": 0.44, + "learning_rate": 4.271351628622724e-05, + "loss": 0.5933, + "step": 11414 + }, + { + "epoch": 0.44, + "learning_rate": 4.271287509617851e-05, + "loss": 0.5607, + "step": 11415 + }, + { + "epoch": 0.44, + "learning_rate": 4.2712233906129776e-05, + "loss": 0.7216, + "step": 11416 + }, + { + "epoch": 0.44, + "learning_rate": 4.271159271608105e-05, + "loss": 0.565, + "step": 11417 + }, + { + "epoch": 0.44, + "learning_rate": 4.271095152603232e-05, + "loss": 0.4995, + "step": 11418 + }, + { + "epoch": 0.44, + "learning_rate": 4.271031033598359e-05, + "loss": 0.6665, + "step": 11419 + }, + { + "epoch": 0.44, + "learning_rate": 4.2709669145934856e-05, + "loss": 0.5423, + "step": 11420 + }, + { + "epoch": 0.44, + "learning_rate": 4.270902795588613e-05, + "loss": 0.6656, + "step": 11421 + }, + { + "epoch": 0.44, + "learning_rate": 4.270838676583739e-05, + "loss": 0.5293, + "step": 11422 + }, + { + "epoch": 0.44, + "learning_rate": 4.2707745575788665e-05, + "loss": 0.6178, + "step": 11423 + }, + { + "epoch": 0.44, + "learning_rate": 4.2707104385739937e-05, + "loss": 0.4715, + "step": 11424 + }, + { + "epoch": 0.44, + "learning_rate": 4.27064631956912e-05, + "loss": 0.5467, + "step": 11425 + }, + { + "epoch": 0.44, + "learning_rate": 4.270582200564247e-05, + "loss": 0.6648, + "step": 11426 + }, + { + "epoch": 0.44, + "learning_rate": 4.2705180815593745e-05, + "loss": 0.5012, + "step": 11427 + }, + { + "epoch": 0.44, + "learning_rate": 4.270453962554502e-05, + "loss": 0.5787, + "step": 11428 + }, + { + "epoch": 0.44, + "learning_rate": 4.270389843549628e-05, + "loss": 0.5983, + "step": 11429 + }, + { + "epoch": 0.44, + "learning_rate": 4.270325724544755e-05, + "loss": 0.598, + "step": 11430 + }, + { + "epoch": 0.44, + "learning_rate": 4.270261605539882e-05, + "loss": 0.6179, + "step": 11431 + }, + { + "epoch": 0.44, + "learning_rate": 4.270197486535009e-05, + "loss": 0.5475, + "step": 11432 + }, + { + "epoch": 0.44, + "learning_rate": 4.270133367530136e-05, + "loss": 0.6045, + "step": 11433 + }, + { + "epoch": 0.44, + "learning_rate": 4.270069248525263e-05, + "loss": 0.6356, + "step": 11434 + }, + { + "epoch": 0.44, + "learning_rate": 4.2700051295203905e-05, + "loss": 0.5504, + "step": 11435 + }, + { + "epoch": 0.44, + "learning_rate": 4.269941010515517e-05, + "loss": 0.6235, + "step": 11436 + }, + { + "epoch": 0.44, + "learning_rate": 4.269876891510644e-05, + "loss": 0.5221, + "step": 11437 + }, + { + "epoch": 0.44, + "learning_rate": 4.269812772505771e-05, + "loss": 0.6104, + "step": 11438 + }, + { + "epoch": 0.44, + "learning_rate": 4.269748653500898e-05, + "loss": 0.6464, + "step": 11439 + }, + { + "epoch": 0.44, + "learning_rate": 4.269684534496024e-05, + "loss": 0.5597, + "step": 11440 + }, + { + "epoch": 0.44, + "learning_rate": 4.2696204154911515e-05, + "loss": 0.6405, + "step": 11441 + }, + { + "epoch": 0.44, + "learning_rate": 4.269556296486279e-05, + "loss": 0.5348, + "step": 11442 + }, + { + "epoch": 0.44, + "learning_rate": 4.269492177481406e-05, + "loss": 0.5663, + "step": 11443 + }, + { + "epoch": 0.44, + "learning_rate": 4.269428058476533e-05, + "loss": 0.613, + "step": 11444 + }, + { + "epoch": 0.44, + "learning_rate": 4.2693639394716595e-05, + "loss": 0.546, + "step": 11445 + }, + { + "epoch": 0.44, + "learning_rate": 4.269299820466787e-05, + "loss": 0.6069, + "step": 11446 + }, + { + "epoch": 0.44, + "learning_rate": 4.269235701461913e-05, + "loss": 0.5524, + "step": 11447 + }, + { + "epoch": 0.44, + "learning_rate": 4.2691715824570404e-05, + "loss": 0.5194, + "step": 11448 + }, + { + "epoch": 0.44, + "learning_rate": 4.269107463452167e-05, + "loss": 0.5705, + "step": 11449 + }, + { + "epoch": 0.44, + "learning_rate": 4.269043344447295e-05, + "loss": 0.5838, + "step": 11450 + }, + { + "epoch": 0.44, + "learning_rate": 4.268979225442422e-05, + "loss": 0.4795, + "step": 11451 + }, + { + "epoch": 0.44, + "learning_rate": 4.2689151064375484e-05, + "loss": 0.5263, + "step": 11452 + }, + { + "epoch": 0.44, + "learning_rate": 4.2688509874326755e-05, + "loss": 0.5161, + "step": 11453 + }, + { + "epoch": 0.44, + "learning_rate": 4.268786868427802e-05, + "loss": 0.6349, + "step": 11454 + }, + { + "epoch": 0.44, + "learning_rate": 4.268722749422929e-05, + "loss": 0.5356, + "step": 11455 + }, + { + "epoch": 0.44, + "learning_rate": 4.268658630418056e-05, + "loss": 0.5782, + "step": 11456 + }, + { + "epoch": 0.44, + "learning_rate": 4.268594511413183e-05, + "loss": 0.4503, + "step": 11457 + }, + { + "epoch": 0.44, + "learning_rate": 4.26853039240831e-05, + "loss": 0.4366, + "step": 11458 + }, + { + "epoch": 0.44, + "learning_rate": 4.268466273403437e-05, + "loss": 0.6333, + "step": 11459 + }, + { + "epoch": 0.44, + "learning_rate": 4.2684021543985644e-05, + "loss": 0.5288, + "step": 11460 + }, + { + "epoch": 0.44, + "learning_rate": 4.268338035393691e-05, + "loss": 0.6327, + "step": 11461 + }, + { + "epoch": 0.44, + "learning_rate": 4.268273916388818e-05, + "loss": 0.557, + "step": 11462 + }, + { + "epoch": 0.44, + "learning_rate": 4.2682097973839445e-05, + "loss": 0.5903, + "step": 11463 + }, + { + "epoch": 0.44, + "learning_rate": 4.268145678379072e-05, + "loss": 0.5534, + "step": 11464 + }, + { + "epoch": 0.44, + "learning_rate": 4.268081559374198e-05, + "loss": 0.5439, + "step": 11465 + }, + { + "epoch": 0.44, + "learning_rate": 4.2680174403693254e-05, + "loss": 0.6316, + "step": 11466 + }, + { + "epoch": 0.44, + "learning_rate": 4.2679533213644525e-05, + "loss": 0.5283, + "step": 11467 + }, + { + "epoch": 0.44, + "learning_rate": 4.26788920235958e-05, + "loss": 0.5559, + "step": 11468 + }, + { + "epoch": 0.44, + "learning_rate": 4.267825083354707e-05, + "loss": 0.5368, + "step": 11469 + }, + { + "epoch": 0.44, + "learning_rate": 4.2677609643498334e-05, + "loss": 0.5298, + "step": 11470 + }, + { + "epoch": 0.44, + "learning_rate": 4.2676968453449606e-05, + "loss": 0.6417, + "step": 11471 + }, + { + "epoch": 0.44, + "learning_rate": 4.267632726340087e-05, + "loss": 0.5825, + "step": 11472 + }, + { + "epoch": 0.44, + "learning_rate": 4.267568607335214e-05, + "loss": 0.5353, + "step": 11473 + }, + { + "epoch": 0.44, + "learning_rate": 4.2675044883303414e-05, + "loss": 0.6284, + "step": 11474 + }, + { + "epoch": 0.44, + "learning_rate": 4.2674403693254686e-05, + "loss": 0.4793, + "step": 11475 + }, + { + "epoch": 0.44, + "learning_rate": 4.267376250320595e-05, + "loss": 0.5158, + "step": 11476 + }, + { + "epoch": 0.44, + "learning_rate": 4.267312131315722e-05, + "loss": 0.5608, + "step": 11477 + }, + { + "epoch": 0.44, + "learning_rate": 4.2672480123108494e-05, + "loss": 0.5836, + "step": 11478 + }, + { + "epoch": 0.44, + "learning_rate": 4.267183893305976e-05, + "loss": 0.4513, + "step": 11479 + }, + { + "epoch": 0.44, + "learning_rate": 4.267119774301103e-05, + "loss": 0.5008, + "step": 11480 + }, + { + "epoch": 0.44, + "learning_rate": 4.2670556552962296e-05, + "loss": 0.4825, + "step": 11481 + }, + { + "epoch": 0.44, + "learning_rate": 4.266991536291357e-05, + "loss": 0.5625, + "step": 11482 + }, + { + "epoch": 0.44, + "learning_rate": 4.266927417286484e-05, + "loss": 0.4977, + "step": 11483 + }, + { + "epoch": 0.44, + "learning_rate": 4.266863298281611e-05, + "loss": 0.5223, + "step": 11484 + }, + { + "epoch": 0.44, + "learning_rate": 4.2667991792767376e-05, + "loss": 0.4534, + "step": 11485 + }, + { + "epoch": 0.44, + "learning_rate": 4.266735060271865e-05, + "loss": 0.5603, + "step": 11486 + }, + { + "epoch": 0.44, + "learning_rate": 4.266670941266992e-05, + "loss": 0.5217, + "step": 11487 + }, + { + "epoch": 0.44, + "learning_rate": 4.2666068222621184e-05, + "loss": 0.578, + "step": 11488 + }, + { + "epoch": 0.44, + "learning_rate": 4.2665427032572456e-05, + "loss": 0.6068, + "step": 11489 + }, + { + "epoch": 0.44, + "learning_rate": 4.266478584252373e-05, + "loss": 0.6068, + "step": 11490 + }, + { + "epoch": 0.44, + "learning_rate": 4.2664144652475e-05, + "loss": 0.4961, + "step": 11491 + }, + { + "epoch": 0.44, + "learning_rate": 4.2663503462426264e-05, + "loss": 0.5802, + "step": 11492 + }, + { + "epoch": 0.44, + "learning_rate": 4.2662862272377536e-05, + "loss": 0.539, + "step": 11493 + }, + { + "epoch": 0.44, + "learning_rate": 4.26622210823288e-05, + "loss": 0.6264, + "step": 11494 + }, + { + "epoch": 0.44, + "learning_rate": 4.266157989228007e-05, + "loss": 0.5757, + "step": 11495 + }, + { + "epoch": 0.44, + "learning_rate": 4.2660938702231344e-05, + "loss": 0.5255, + "step": 11496 + }, + { + "epoch": 0.44, + "learning_rate": 4.266029751218261e-05, + "loss": 0.5484, + "step": 11497 + }, + { + "epoch": 0.44, + "learning_rate": 4.265965632213388e-05, + "loss": 0.4485, + "step": 11498 + }, + { + "epoch": 0.44, + "learning_rate": 4.265901513208515e-05, + "loss": 0.5017, + "step": 11499 + }, + { + "epoch": 0.44, + "learning_rate": 4.2658373942036424e-05, + "loss": 0.5103, + "step": 11500 + }, + { + "epoch": 0.44, + "learning_rate": 4.265773275198769e-05, + "loss": 0.5458, + "step": 11501 + }, + { + "epoch": 0.44, + "learning_rate": 4.265709156193896e-05, + "loss": 0.5194, + "step": 11502 + }, + { + "epoch": 0.44, + "learning_rate": 4.265645037189023e-05, + "loss": 0.6201, + "step": 11503 + }, + { + "epoch": 0.44, + "learning_rate": 4.26558091818415e-05, + "loss": 0.5958, + "step": 11504 + }, + { + "epoch": 0.44, + "learning_rate": 4.265516799179277e-05, + "loss": 0.4649, + "step": 11505 + }, + { + "epoch": 0.44, + "learning_rate": 4.265452680174404e-05, + "loss": 0.5735, + "step": 11506 + }, + { + "epoch": 0.44, + "learning_rate": 4.265388561169531e-05, + "loss": 0.5469, + "step": 11507 + }, + { + "epoch": 0.44, + "learning_rate": 4.265324442164658e-05, + "loss": 0.4864, + "step": 11508 + }, + { + "epoch": 0.44, + "learning_rate": 4.265260323159785e-05, + "loss": 0.6713, + "step": 11509 + }, + { + "epoch": 0.44, + "learning_rate": 4.2651962041549114e-05, + "loss": 0.556, + "step": 11510 + }, + { + "epoch": 0.44, + "learning_rate": 4.2651320851500386e-05, + "loss": 0.4865, + "step": 11511 + }, + { + "epoch": 0.44, + "learning_rate": 4.265067966145166e-05, + "loss": 0.5554, + "step": 11512 + }, + { + "epoch": 0.44, + "learning_rate": 4.265003847140292e-05, + "loss": 0.646, + "step": 11513 + }, + { + "epoch": 0.44, + "learning_rate": 4.2649397281354195e-05, + "loss": 0.5336, + "step": 11514 + }, + { + "epoch": 0.44, + "learning_rate": 4.2648756091305466e-05, + "loss": 0.5889, + "step": 11515 + }, + { + "epoch": 0.44, + "learning_rate": 4.264811490125674e-05, + "loss": 0.5486, + "step": 11516 + }, + { + "epoch": 0.44, + "learning_rate": 4.2647473711208e-05, + "loss": 0.5828, + "step": 11517 + }, + { + "epoch": 0.44, + "learning_rate": 4.2646832521159275e-05, + "loss": 0.5101, + "step": 11518 + }, + { + "epoch": 0.44, + "learning_rate": 4.264619133111054e-05, + "loss": 0.6087, + "step": 11519 + }, + { + "epoch": 0.44, + "learning_rate": 4.264555014106181e-05, + "loss": 0.4132, + "step": 11520 + }, + { + "epoch": 0.44, + "learning_rate": 4.264490895101308e-05, + "loss": 0.5472, + "step": 11521 + }, + { + "epoch": 0.44, + "learning_rate": 4.2644267760964355e-05, + "loss": 0.5719, + "step": 11522 + }, + { + "epoch": 0.44, + "learning_rate": 4.2643626570915626e-05, + "loss": 0.5141, + "step": 11523 + }, + { + "epoch": 0.44, + "learning_rate": 4.264298538086689e-05, + "loss": 0.5558, + "step": 11524 + }, + { + "epoch": 0.44, + "learning_rate": 4.264234419081816e-05, + "loss": 0.6272, + "step": 11525 + }, + { + "epoch": 0.44, + "learning_rate": 4.264170300076943e-05, + "loss": 0.4997, + "step": 11526 + }, + { + "epoch": 0.44, + "learning_rate": 4.26410618107207e-05, + "loss": 0.6146, + "step": 11527 + }, + { + "epoch": 0.44, + "learning_rate": 4.2640420620671965e-05, + "loss": 0.5363, + "step": 11528 + }, + { + "epoch": 0.44, + "learning_rate": 4.2639779430623236e-05, + "loss": 0.5523, + "step": 11529 + }, + { + "epoch": 0.44, + "learning_rate": 4.263913824057451e-05, + "loss": 0.5647, + "step": 11530 + }, + { + "epoch": 0.44, + "learning_rate": 4.263849705052578e-05, + "loss": 0.5604, + "step": 11531 + }, + { + "epoch": 0.44, + "learning_rate": 4.263785586047705e-05, + "loss": 0.5462, + "step": 11532 + }, + { + "epoch": 0.44, + "learning_rate": 4.2637214670428316e-05, + "loss": 0.6006, + "step": 11533 + }, + { + "epoch": 0.44, + "learning_rate": 4.263657348037959e-05, + "loss": 0.4606, + "step": 11534 + }, + { + "epoch": 0.44, + "learning_rate": 4.263593229033085e-05, + "loss": 0.5277, + "step": 11535 + }, + { + "epoch": 0.44, + "learning_rate": 4.2635291100282125e-05, + "loss": 0.6486, + "step": 11536 + }, + { + "epoch": 0.44, + "learning_rate": 4.263464991023339e-05, + "loss": 0.5984, + "step": 11537 + }, + { + "epoch": 0.44, + "learning_rate": 4.263400872018467e-05, + "loss": 0.5016, + "step": 11538 + }, + { + "epoch": 0.44, + "learning_rate": 4.263336753013594e-05, + "loss": 0.5235, + "step": 11539 + }, + { + "epoch": 0.44, + "learning_rate": 4.2632726340087205e-05, + "loss": 0.587, + "step": 11540 + }, + { + "epoch": 0.44, + "learning_rate": 4.263208515003848e-05, + "loss": 0.6774, + "step": 11541 + }, + { + "epoch": 0.44, + "learning_rate": 4.263144395998974e-05, + "loss": 0.5156, + "step": 11542 + }, + { + "epoch": 0.44, + "learning_rate": 4.263080276994101e-05, + "loss": 0.5559, + "step": 11543 + }, + { + "epoch": 0.44, + "learning_rate": 4.263016157989228e-05, + "loss": 0.5741, + "step": 11544 + }, + { + "epoch": 0.44, + "learning_rate": 4.262952038984355e-05, + "loss": 0.5959, + "step": 11545 + }, + { + "epoch": 0.44, + "learning_rate": 4.262887919979482e-05, + "loss": 0.4934, + "step": 11546 + }, + { + "epoch": 0.44, + "learning_rate": 4.2628238009746093e-05, + "loss": 0.6041, + "step": 11547 + }, + { + "epoch": 0.44, + "learning_rate": 4.2627596819697365e-05, + "loss": 0.5204, + "step": 11548 + }, + { + "epoch": 0.44, + "learning_rate": 4.262695562964863e-05, + "loss": 0.5162, + "step": 11549 + }, + { + "epoch": 0.44, + "learning_rate": 4.26263144395999e-05, + "loss": 0.5722, + "step": 11550 + }, + { + "epoch": 0.44, + "learning_rate": 4.262567324955117e-05, + "loss": 0.5395, + "step": 11551 + }, + { + "epoch": 0.44, + "learning_rate": 4.262503205950244e-05, + "loss": 0.6482, + "step": 11552 + }, + { + "epoch": 0.44, + "learning_rate": 4.26243908694537e-05, + "loss": 0.5668, + "step": 11553 + }, + { + "epoch": 0.44, + "learning_rate": 4.2623749679404975e-05, + "loss": 0.5611, + "step": 11554 + }, + { + "epoch": 0.44, + "learning_rate": 4.262310848935625e-05, + "loss": 0.4727, + "step": 11555 + }, + { + "epoch": 0.44, + "learning_rate": 4.262246729930752e-05, + "loss": 0.5924, + "step": 11556 + }, + { + "epoch": 0.44, + "learning_rate": 4.262182610925879e-05, + "loss": 0.619, + "step": 11557 + }, + { + "epoch": 0.44, + "learning_rate": 4.2621184919210055e-05, + "loss": 0.6, + "step": 11558 + }, + { + "epoch": 0.44, + "learning_rate": 4.262054372916133e-05, + "loss": 0.5667, + "step": 11559 + }, + { + "epoch": 0.44, + "learning_rate": 4.261990253911259e-05, + "loss": 0.5288, + "step": 11560 + }, + { + "epoch": 0.44, + "learning_rate": 4.2619261349063864e-05, + "loss": 0.553, + "step": 11561 + }, + { + "epoch": 0.44, + "learning_rate": 4.2618620159015135e-05, + "loss": 0.5238, + "step": 11562 + }, + { + "epoch": 0.44, + "learning_rate": 4.261797896896641e-05, + "loss": 0.5495, + "step": 11563 + }, + { + "epoch": 0.44, + "learning_rate": 4.261733777891767e-05, + "loss": 0.5076, + "step": 11564 + }, + { + "epoch": 0.44, + "learning_rate": 4.2616696588868944e-05, + "loss": 0.5239, + "step": 11565 + }, + { + "epoch": 0.44, + "learning_rate": 4.2616055398820215e-05, + "loss": 0.538, + "step": 11566 + }, + { + "epoch": 0.44, + "learning_rate": 4.261541420877148e-05, + "loss": 0.5485, + "step": 11567 + }, + { + "epoch": 0.44, + "learning_rate": 4.261477301872275e-05, + "loss": 0.6006, + "step": 11568 + }, + { + "epoch": 0.44, + "learning_rate": 4.261413182867402e-05, + "loss": 0.5853, + "step": 11569 + }, + { + "epoch": 0.44, + "learning_rate": 4.261349063862529e-05, + "loss": 0.5803, + "step": 11570 + }, + { + "epoch": 0.44, + "learning_rate": 4.261284944857656e-05, + "loss": 0.6374, + "step": 11571 + }, + { + "epoch": 0.44, + "learning_rate": 4.261220825852783e-05, + "loss": 0.5659, + "step": 11572 + }, + { + "epoch": 0.44, + "learning_rate": 4.26115670684791e-05, + "loss": 0.4802, + "step": 11573 + }, + { + "epoch": 0.44, + "learning_rate": 4.261092587843037e-05, + "loss": 0.5668, + "step": 11574 + }, + { + "epoch": 0.45, + "learning_rate": 4.261028468838164e-05, + "loss": 0.5816, + "step": 11575 + }, + { + "epoch": 0.45, + "learning_rate": 4.2609643498332905e-05, + "loss": 0.5038, + "step": 11576 + }, + { + "epoch": 0.45, + "learning_rate": 4.260900230828418e-05, + "loss": 0.5774, + "step": 11577 + }, + { + "epoch": 0.45, + "learning_rate": 4.260836111823545e-05, + "loss": 0.6074, + "step": 11578 + }, + { + "epoch": 0.45, + "learning_rate": 4.260771992818672e-05, + "loss": 0.5301, + "step": 11579 + }, + { + "epoch": 0.45, + "learning_rate": 4.2607078738137986e-05, + "loss": 0.5942, + "step": 11580 + }, + { + "epoch": 0.45, + "learning_rate": 4.260643754808926e-05, + "loss": 0.6232, + "step": 11581 + }, + { + "epoch": 0.45, + "learning_rate": 4.260579635804052e-05, + "loss": 0.558, + "step": 11582 + }, + { + "epoch": 0.45, + "learning_rate": 4.2605155167991794e-05, + "loss": 0.5925, + "step": 11583 + }, + { + "epoch": 0.45, + "learning_rate": 4.2604513977943066e-05, + "loss": 0.5599, + "step": 11584 + }, + { + "epoch": 0.45, + "learning_rate": 4.260387278789433e-05, + "loss": 0.577, + "step": 11585 + }, + { + "epoch": 0.45, + "learning_rate": 4.26032315978456e-05, + "loss": 0.681, + "step": 11586 + }, + { + "epoch": 0.45, + "learning_rate": 4.2602590407796874e-05, + "loss": 0.6119, + "step": 11587 + }, + { + "epoch": 0.45, + "learning_rate": 4.2601949217748146e-05, + "loss": 0.6478, + "step": 11588 + }, + { + "epoch": 0.45, + "learning_rate": 4.260130802769941e-05, + "loss": 0.5063, + "step": 11589 + }, + { + "epoch": 0.45, + "learning_rate": 4.260066683765068e-05, + "loss": 0.5399, + "step": 11590 + }, + { + "epoch": 0.45, + "learning_rate": 4.260002564760195e-05, + "loss": 0.5959, + "step": 11591 + }, + { + "epoch": 0.45, + "learning_rate": 4.259938445755322e-05, + "loss": 0.5795, + "step": 11592 + }, + { + "epoch": 0.45, + "learning_rate": 4.259874326750449e-05, + "loss": 0.5483, + "step": 11593 + }, + { + "epoch": 0.45, + "learning_rate": 4.259810207745576e-05, + "loss": 0.4664, + "step": 11594 + }, + { + "epoch": 0.45, + "learning_rate": 4.2597460887407034e-05, + "loss": 0.546, + "step": 11595 + }, + { + "epoch": 0.45, + "learning_rate": 4.25968196973583e-05, + "loss": 0.4798, + "step": 11596 + }, + { + "epoch": 0.45, + "learning_rate": 4.259617850730957e-05, + "loss": 0.4918, + "step": 11597 + }, + { + "epoch": 0.45, + "learning_rate": 4.2595537317260836e-05, + "loss": 0.653, + "step": 11598 + }, + { + "epoch": 0.45, + "learning_rate": 4.259489612721211e-05, + "loss": 0.5093, + "step": 11599 + }, + { + "epoch": 0.45, + "learning_rate": 4.259425493716337e-05, + "loss": 0.6154, + "step": 11600 + }, + { + "epoch": 0.45, + "learning_rate": 4.2593613747114644e-05, + "loss": 0.5482, + "step": 11601 + }, + { + "epoch": 0.45, + "learning_rate": 4.2592972557065916e-05, + "loss": 0.5536, + "step": 11602 + }, + { + "epoch": 0.45, + "learning_rate": 4.259233136701719e-05, + "loss": 0.6764, + "step": 11603 + }, + { + "epoch": 0.45, + "learning_rate": 4.259169017696846e-05, + "loss": 0.4999, + "step": 11604 + }, + { + "epoch": 0.45, + "learning_rate": 4.2591048986919724e-05, + "loss": 0.5123, + "step": 11605 + }, + { + "epoch": 0.45, + "learning_rate": 4.2590407796870996e-05, + "loss": 0.6017, + "step": 11606 + }, + { + "epoch": 0.45, + "learning_rate": 4.258976660682226e-05, + "loss": 0.5257, + "step": 11607 + }, + { + "epoch": 0.45, + "learning_rate": 4.258912541677353e-05, + "loss": 0.5793, + "step": 11608 + }, + { + "epoch": 0.45, + "learning_rate": 4.25884842267248e-05, + "loss": 0.4958, + "step": 11609 + }, + { + "epoch": 0.45, + "learning_rate": 4.2587843036676076e-05, + "loss": 0.6964, + "step": 11610 + }, + { + "epoch": 0.45, + "learning_rate": 4.258720184662735e-05, + "loss": 0.5854, + "step": 11611 + }, + { + "epoch": 0.45, + "learning_rate": 4.258656065657861e-05, + "loss": 0.5021, + "step": 11612 + }, + { + "epoch": 0.45, + "learning_rate": 4.2585919466529884e-05, + "loss": 0.5594, + "step": 11613 + }, + { + "epoch": 0.45, + "learning_rate": 4.258527827648115e-05, + "loss": 0.5729, + "step": 11614 + }, + { + "epoch": 0.45, + "learning_rate": 4.258463708643242e-05, + "loss": 0.6048, + "step": 11615 + }, + { + "epoch": 0.45, + "learning_rate": 4.2583995896383686e-05, + "loss": 0.478, + "step": 11616 + }, + { + "epoch": 0.45, + "learning_rate": 4.258335470633496e-05, + "loss": 0.5952, + "step": 11617 + }, + { + "epoch": 0.45, + "learning_rate": 4.258271351628623e-05, + "loss": 0.5209, + "step": 11618 + }, + { + "epoch": 0.45, + "learning_rate": 4.25820723262375e-05, + "loss": 0.5596, + "step": 11619 + }, + { + "epoch": 0.45, + "learning_rate": 4.258143113618877e-05, + "loss": 0.5146, + "step": 11620 + }, + { + "epoch": 0.45, + "learning_rate": 4.258078994614004e-05, + "loss": 0.4271, + "step": 11621 + }, + { + "epoch": 0.45, + "learning_rate": 4.258014875609131e-05, + "loss": 0.474, + "step": 11622 + }, + { + "epoch": 0.45, + "learning_rate": 4.2579507566042574e-05, + "loss": 0.5482, + "step": 11623 + }, + { + "epoch": 0.45, + "learning_rate": 4.2578866375993846e-05, + "loss": 0.5966, + "step": 11624 + }, + { + "epoch": 0.45, + "learning_rate": 4.257822518594511e-05, + "loss": 0.613, + "step": 11625 + }, + { + "epoch": 0.45, + "learning_rate": 4.257758399589639e-05, + "loss": 0.4818, + "step": 11626 + }, + { + "epoch": 0.45, + "learning_rate": 4.2576942805847655e-05, + "loss": 0.4976, + "step": 11627 + }, + { + "epoch": 0.45, + "learning_rate": 4.2576301615798926e-05, + "loss": 0.594, + "step": 11628 + }, + { + "epoch": 0.45, + "learning_rate": 4.25756604257502e-05, + "loss": 0.5579, + "step": 11629 + }, + { + "epoch": 0.45, + "learning_rate": 4.257501923570146e-05, + "loss": 0.6035, + "step": 11630 + }, + { + "epoch": 0.45, + "learning_rate": 4.2574378045652735e-05, + "loss": 0.5509, + "step": 11631 + }, + { + "epoch": 0.45, + "learning_rate": 4.2573736855604e-05, + "loss": 0.5089, + "step": 11632 + }, + { + "epoch": 0.45, + "learning_rate": 4.257309566555527e-05, + "loss": 0.5621, + "step": 11633 + }, + { + "epoch": 0.45, + "learning_rate": 4.257245447550654e-05, + "loss": 0.5019, + "step": 11634 + }, + { + "epoch": 0.45, + "learning_rate": 4.2571813285457815e-05, + "loss": 0.6193, + "step": 11635 + }, + { + "epoch": 0.45, + "learning_rate": 4.257117209540908e-05, + "loss": 0.58, + "step": 11636 + }, + { + "epoch": 0.45, + "learning_rate": 4.257053090536035e-05, + "loss": 0.593, + "step": 11637 + }, + { + "epoch": 0.45, + "learning_rate": 4.256988971531162e-05, + "loss": 0.4517, + "step": 11638 + }, + { + "epoch": 0.45, + "learning_rate": 4.256924852526289e-05, + "loss": 0.6475, + "step": 11639 + }, + { + "epoch": 0.45, + "learning_rate": 4.256860733521416e-05, + "loss": 0.5312, + "step": 11640 + }, + { + "epoch": 0.45, + "learning_rate": 4.2567966145165425e-05, + "loss": 0.4827, + "step": 11641 + }, + { + "epoch": 0.45, + "learning_rate": 4.2567324955116696e-05, + "loss": 0.4806, + "step": 11642 + }, + { + "epoch": 0.45, + "learning_rate": 4.256668376506797e-05, + "loss": 0.5624, + "step": 11643 + }, + { + "epoch": 0.45, + "learning_rate": 4.256604257501924e-05, + "loss": 0.5886, + "step": 11644 + }, + { + "epoch": 0.45, + "learning_rate": 4.2565401384970505e-05, + "loss": 0.5401, + "step": 11645 + }, + { + "epoch": 0.45, + "learning_rate": 4.2564760194921777e-05, + "loss": 0.5523, + "step": 11646 + }, + { + "epoch": 0.45, + "learning_rate": 4.256411900487305e-05, + "loss": 0.4712, + "step": 11647 + }, + { + "epoch": 0.45, + "learning_rate": 4.256347781482431e-05, + "loss": 0.4808, + "step": 11648 + }, + { + "epoch": 0.45, + "learning_rate": 4.2562836624775585e-05, + "loss": 0.477, + "step": 11649 + }, + { + "epoch": 0.45, + "learning_rate": 4.2562195434726857e-05, + "loss": 0.5547, + "step": 11650 + }, + { + "epoch": 0.45, + "learning_rate": 4.256155424467813e-05, + "loss": 0.6424, + "step": 11651 + }, + { + "epoch": 0.45, + "learning_rate": 4.256091305462939e-05, + "loss": 0.5377, + "step": 11652 + }, + { + "epoch": 0.45, + "learning_rate": 4.2560271864580665e-05, + "loss": 0.5521, + "step": 11653 + }, + { + "epoch": 0.45, + "learning_rate": 4.255963067453193e-05, + "loss": 0.5946, + "step": 11654 + }, + { + "epoch": 0.45, + "learning_rate": 4.25589894844832e-05, + "loss": 0.5657, + "step": 11655 + }, + { + "epoch": 0.45, + "learning_rate": 4.255834829443447e-05, + "loss": 0.4849, + "step": 11656 + }, + { + "epoch": 0.45, + "learning_rate": 4.255770710438574e-05, + "loss": 0.603, + "step": 11657 + }, + { + "epoch": 0.45, + "learning_rate": 4.255706591433701e-05, + "loss": 0.667, + "step": 11658 + }, + { + "epoch": 0.45, + "learning_rate": 4.255642472428828e-05, + "loss": 0.5223, + "step": 11659 + }, + { + "epoch": 0.45, + "learning_rate": 4.2555783534239553e-05, + "loss": 0.5169, + "step": 11660 + }, + { + "epoch": 0.45, + "learning_rate": 4.255514234419082e-05, + "loss": 0.6238, + "step": 11661 + }, + { + "epoch": 0.45, + "learning_rate": 4.255450115414209e-05, + "loss": 0.519, + "step": 11662 + }, + { + "epoch": 0.45, + "learning_rate": 4.2553859964093355e-05, + "loss": 0.5907, + "step": 11663 + }, + { + "epoch": 0.45, + "learning_rate": 4.255321877404463e-05, + "loss": 0.7038, + "step": 11664 + }, + { + "epoch": 0.45, + "learning_rate": 4.25525775839959e-05, + "loss": 0.5747, + "step": 11665 + }, + { + "epoch": 0.45, + "learning_rate": 4.255193639394717e-05, + "loss": 0.4685, + "step": 11666 + }, + { + "epoch": 0.45, + "learning_rate": 4.255129520389844e-05, + "loss": 0.5453, + "step": 11667 + }, + { + "epoch": 0.45, + "learning_rate": 4.255065401384971e-05, + "loss": 0.4432, + "step": 11668 + }, + { + "epoch": 0.45, + "learning_rate": 4.255001282380098e-05, + "loss": 0.4656, + "step": 11669 + }, + { + "epoch": 0.45, + "learning_rate": 4.2549371633752244e-05, + "loss": 0.5801, + "step": 11670 + }, + { + "epoch": 0.45, + "learning_rate": 4.2548730443703515e-05, + "loss": 0.5013, + "step": 11671 + }, + { + "epoch": 0.45, + "learning_rate": 4.254808925365478e-05, + "loss": 0.5214, + "step": 11672 + }, + { + "epoch": 0.45, + "learning_rate": 4.254744806360605e-05, + "loss": 0.563, + "step": 11673 + }, + { + "epoch": 0.45, + "learning_rate": 4.2546806873557324e-05, + "loss": 0.5561, + "step": 11674 + }, + { + "epoch": 0.45, + "learning_rate": 4.2546165683508595e-05, + "loss": 0.5252, + "step": 11675 + }, + { + "epoch": 0.45, + "learning_rate": 4.254552449345987e-05, + "loss": 0.5208, + "step": 11676 + }, + { + "epoch": 0.45, + "learning_rate": 4.254488330341113e-05, + "loss": 0.5233, + "step": 11677 + }, + { + "epoch": 0.45, + "learning_rate": 4.2544242113362404e-05, + "loss": 0.5205, + "step": 11678 + }, + { + "epoch": 0.45, + "learning_rate": 4.254360092331367e-05, + "loss": 0.5646, + "step": 11679 + }, + { + "epoch": 0.45, + "learning_rate": 4.254295973326494e-05, + "loss": 0.6785, + "step": 11680 + }, + { + "epoch": 0.45, + "learning_rate": 4.2542318543216205e-05, + "loss": 0.6117, + "step": 11681 + }, + { + "epoch": 0.45, + "learning_rate": 4.2541677353167484e-05, + "loss": 0.6267, + "step": 11682 + }, + { + "epoch": 0.45, + "learning_rate": 4.2541036163118755e-05, + "loss": 0.5719, + "step": 11683 + }, + { + "epoch": 0.45, + "learning_rate": 4.254039497307002e-05, + "loss": 0.5054, + "step": 11684 + }, + { + "epoch": 0.45, + "learning_rate": 4.253975378302129e-05, + "loss": 0.4985, + "step": 11685 + }, + { + "epoch": 0.45, + "learning_rate": 4.253911259297256e-05, + "loss": 0.5191, + "step": 11686 + }, + { + "epoch": 0.45, + "learning_rate": 4.253847140292383e-05, + "loss": 0.4805, + "step": 11687 + }, + { + "epoch": 0.45, + "learning_rate": 4.2537830212875094e-05, + "loss": 0.6382, + "step": 11688 + }, + { + "epoch": 0.45, + "learning_rate": 4.2537189022826365e-05, + "loss": 0.5338, + "step": 11689 + }, + { + "epoch": 0.45, + "learning_rate": 4.253654783277764e-05, + "loss": 0.6392, + "step": 11690 + }, + { + "epoch": 0.45, + "learning_rate": 4.253590664272891e-05, + "loss": 0.5475, + "step": 11691 + }, + { + "epoch": 0.45, + "learning_rate": 4.253526545268018e-05, + "loss": 0.5163, + "step": 11692 + }, + { + "epoch": 0.45, + "learning_rate": 4.2534624262631446e-05, + "loss": 0.5378, + "step": 11693 + }, + { + "epoch": 0.45, + "learning_rate": 4.253398307258272e-05, + "loss": 0.5251, + "step": 11694 + }, + { + "epoch": 0.45, + "learning_rate": 4.253334188253398e-05, + "loss": 0.5135, + "step": 11695 + }, + { + "epoch": 0.45, + "learning_rate": 4.2532700692485254e-05, + "loss": 0.6067, + "step": 11696 + }, + { + "epoch": 0.45, + "learning_rate": 4.253205950243652e-05, + "loss": 0.5092, + "step": 11697 + }, + { + "epoch": 0.45, + "learning_rate": 4.25314183123878e-05, + "loss": 0.5678, + "step": 11698 + }, + { + "epoch": 0.45, + "learning_rate": 4.253077712233907e-05, + "loss": 0.582, + "step": 11699 + }, + { + "epoch": 0.45, + "learning_rate": 4.2530135932290334e-05, + "loss": 0.6103, + "step": 11700 + }, + { + "epoch": 0.45, + "learning_rate": 4.2529494742241606e-05, + "loss": 0.5782, + "step": 11701 + }, + { + "epoch": 0.45, + "learning_rate": 4.252885355219287e-05, + "loss": 0.5426, + "step": 11702 + }, + { + "epoch": 0.45, + "learning_rate": 4.252821236214414e-05, + "loss": 0.5681, + "step": 11703 + }, + { + "epoch": 0.45, + "learning_rate": 4.252757117209541e-05, + "loss": 0.517, + "step": 11704 + }, + { + "epoch": 0.45, + "learning_rate": 4.252692998204668e-05, + "loss": 0.5601, + "step": 11705 + }, + { + "epoch": 0.45, + "learning_rate": 4.252628879199795e-05, + "loss": 0.654, + "step": 11706 + }, + { + "epoch": 0.45, + "learning_rate": 4.252564760194922e-05, + "loss": 0.5722, + "step": 11707 + }, + { + "epoch": 0.45, + "learning_rate": 4.2525006411900494e-05, + "loss": 0.454, + "step": 11708 + }, + { + "epoch": 0.45, + "learning_rate": 4.252436522185176e-05, + "loss": 0.5828, + "step": 11709 + }, + { + "epoch": 0.45, + "learning_rate": 4.252372403180303e-05, + "loss": 0.5459, + "step": 11710 + }, + { + "epoch": 0.45, + "learning_rate": 4.2523082841754296e-05, + "loss": 0.5077, + "step": 11711 + }, + { + "epoch": 0.45, + "learning_rate": 4.252244165170557e-05, + "loss": 0.5684, + "step": 11712 + }, + { + "epoch": 0.45, + "learning_rate": 4.252180046165683e-05, + "loss": 0.5644, + "step": 11713 + }, + { + "epoch": 0.45, + "learning_rate": 4.2521159271608104e-05, + "loss": 0.4768, + "step": 11714 + }, + { + "epoch": 0.45, + "learning_rate": 4.2520518081559376e-05, + "loss": 0.5259, + "step": 11715 + }, + { + "epoch": 0.45, + "learning_rate": 4.251987689151065e-05, + "loss": 0.4632, + "step": 11716 + }, + { + "epoch": 0.45, + "learning_rate": 4.251923570146192e-05, + "loss": 0.5708, + "step": 11717 + }, + { + "epoch": 0.45, + "learning_rate": 4.2518594511413184e-05, + "loss": 0.4982, + "step": 11718 + }, + { + "epoch": 0.45, + "learning_rate": 4.2517953321364456e-05, + "loss": 0.6057, + "step": 11719 + }, + { + "epoch": 0.45, + "learning_rate": 4.251731213131572e-05, + "loss": 0.5652, + "step": 11720 + }, + { + "epoch": 0.45, + "learning_rate": 4.251667094126699e-05, + "loss": 0.5125, + "step": 11721 + }, + { + "epoch": 0.45, + "learning_rate": 4.2516029751218264e-05, + "loss": 0.5702, + "step": 11722 + }, + { + "epoch": 0.45, + "learning_rate": 4.2515388561169536e-05, + "loss": 0.5242, + "step": 11723 + }, + { + "epoch": 0.45, + "learning_rate": 4.25147473711208e-05, + "loss": 0.5487, + "step": 11724 + }, + { + "epoch": 0.45, + "learning_rate": 4.251410618107207e-05, + "loss": 0.4265, + "step": 11725 + }, + { + "epoch": 0.45, + "learning_rate": 4.2513464991023344e-05, + "loss": 0.5626, + "step": 11726 + }, + { + "epoch": 0.45, + "learning_rate": 4.251282380097461e-05, + "loss": 0.6161, + "step": 11727 + }, + { + "epoch": 0.45, + "learning_rate": 4.251218261092588e-05, + "loss": 0.6168, + "step": 11728 + }, + { + "epoch": 0.45, + "learning_rate": 4.2511541420877146e-05, + "loss": 0.4906, + "step": 11729 + }, + { + "epoch": 0.45, + "learning_rate": 4.251090023082842e-05, + "loss": 0.5288, + "step": 11730 + }, + { + "epoch": 0.45, + "learning_rate": 4.251025904077969e-05, + "loss": 0.5791, + "step": 11731 + }, + { + "epoch": 0.45, + "learning_rate": 4.250961785073096e-05, + "loss": 0.4185, + "step": 11732 + }, + { + "epoch": 0.45, + "learning_rate": 4.2508976660682226e-05, + "loss": 0.5267, + "step": 11733 + }, + { + "epoch": 0.45, + "learning_rate": 4.25083354706335e-05, + "loss": 0.5289, + "step": 11734 + }, + { + "epoch": 0.45, + "learning_rate": 4.250769428058477e-05, + "loss": 0.5954, + "step": 11735 + }, + { + "epoch": 0.45, + "learning_rate": 4.2507053090536034e-05, + "loss": 0.6151, + "step": 11736 + }, + { + "epoch": 0.45, + "learning_rate": 4.2506411900487306e-05, + "loss": 0.6293, + "step": 11737 + }, + { + "epoch": 0.45, + "learning_rate": 4.250577071043858e-05, + "loss": 0.665, + "step": 11738 + }, + { + "epoch": 0.45, + "learning_rate": 4.250512952038985e-05, + "loss": 0.454, + "step": 11739 + }, + { + "epoch": 0.45, + "learning_rate": 4.2504488330341115e-05, + "loss": 0.524, + "step": 11740 + }, + { + "epoch": 0.45, + "learning_rate": 4.2503847140292386e-05, + "loss": 0.6948, + "step": 11741 + }, + { + "epoch": 0.45, + "learning_rate": 4.250320595024365e-05, + "loss": 0.6174, + "step": 11742 + }, + { + "epoch": 0.45, + "learning_rate": 4.250256476019492e-05, + "loss": 0.6199, + "step": 11743 + }, + { + "epoch": 0.45, + "learning_rate": 4.2501923570146195e-05, + "loss": 0.4088, + "step": 11744 + }, + { + "epoch": 0.45, + "learning_rate": 4.250128238009746e-05, + "loss": 0.5397, + "step": 11745 + }, + { + "epoch": 0.45, + "learning_rate": 4.250064119004873e-05, + "loss": 0.5611, + "step": 11746 + }, + { + "epoch": 0.45, + "learning_rate": 4.25e-05, + "loss": 0.5361, + "step": 11747 + }, + { + "epoch": 0.45, + "learning_rate": 4.2499358809951275e-05, + "loss": 0.4836, + "step": 11748 + }, + { + "epoch": 0.45, + "learning_rate": 4.249871761990254e-05, + "loss": 0.5391, + "step": 11749 + }, + { + "epoch": 0.45, + "learning_rate": 4.249807642985381e-05, + "loss": 0.6718, + "step": 11750 + }, + { + "epoch": 0.45, + "learning_rate": 4.2497435239805076e-05, + "loss": 0.6121, + "step": 11751 + }, + { + "epoch": 0.45, + "learning_rate": 4.249679404975635e-05, + "loss": 0.5303, + "step": 11752 + }, + { + "epoch": 0.45, + "learning_rate": 4.249615285970762e-05, + "loss": 0.5455, + "step": 11753 + }, + { + "epoch": 0.45, + "learning_rate": 4.249551166965889e-05, + "loss": 0.5736, + "step": 11754 + }, + { + "epoch": 0.45, + "learning_rate": 4.249487047961016e-05, + "loss": 0.5105, + "step": 11755 + }, + { + "epoch": 0.45, + "learning_rate": 4.249422928956143e-05, + "loss": 0.5668, + "step": 11756 + }, + { + "epoch": 0.45, + "learning_rate": 4.24935880995127e-05, + "loss": 0.5375, + "step": 11757 + }, + { + "epoch": 0.45, + "learning_rate": 4.2492946909463965e-05, + "loss": 0.4993, + "step": 11758 + }, + { + "epoch": 0.45, + "learning_rate": 4.2492305719415237e-05, + "loss": 0.5617, + "step": 11759 + }, + { + "epoch": 0.45, + "learning_rate": 4.24916645293665e-05, + "loss": 0.5086, + "step": 11760 + }, + { + "epoch": 0.45, + "learning_rate": 4.249102333931777e-05, + "loss": 0.5673, + "step": 11761 + }, + { + "epoch": 0.45, + "learning_rate": 4.2490382149269045e-05, + "loss": 0.5093, + "step": 11762 + }, + { + "epoch": 0.45, + "learning_rate": 4.248974095922032e-05, + "loss": 0.6268, + "step": 11763 + }, + { + "epoch": 0.45, + "learning_rate": 4.248909976917159e-05, + "loss": 0.4927, + "step": 11764 + }, + { + "epoch": 0.45, + "learning_rate": 4.248845857912285e-05, + "loss": 0.6672, + "step": 11765 + }, + { + "epoch": 0.45, + "learning_rate": 4.2487817389074125e-05, + "loss": 0.4673, + "step": 11766 + }, + { + "epoch": 0.45, + "learning_rate": 4.248717619902539e-05, + "loss": 0.5272, + "step": 11767 + }, + { + "epoch": 0.45, + "learning_rate": 4.248653500897666e-05, + "loss": 0.542, + "step": 11768 + }, + { + "epoch": 0.45, + "learning_rate": 4.2485893818927927e-05, + "loss": 0.5375, + "step": 11769 + }, + { + "epoch": 0.45, + "learning_rate": 4.2485252628879205e-05, + "loss": 0.5125, + "step": 11770 + }, + { + "epoch": 0.45, + "learning_rate": 4.248461143883048e-05, + "loss": 0.4654, + "step": 11771 + }, + { + "epoch": 0.45, + "learning_rate": 4.248397024878174e-05, + "loss": 0.5798, + "step": 11772 + }, + { + "epoch": 0.45, + "learning_rate": 4.2483329058733013e-05, + "loss": 0.4871, + "step": 11773 + }, + { + "epoch": 0.45, + "learning_rate": 4.248268786868428e-05, + "loss": 0.5029, + "step": 11774 + }, + { + "epoch": 0.45, + "learning_rate": 4.248204667863555e-05, + "loss": 0.5427, + "step": 11775 + }, + { + "epoch": 0.45, + "learning_rate": 4.2481405488586815e-05, + "loss": 0.6867, + "step": 11776 + }, + { + "epoch": 0.45, + "learning_rate": 4.248076429853809e-05, + "loss": 0.5394, + "step": 11777 + }, + { + "epoch": 0.45, + "learning_rate": 4.248012310848936e-05, + "loss": 0.5691, + "step": 11778 + }, + { + "epoch": 0.45, + "learning_rate": 4.247948191844063e-05, + "loss": 0.5186, + "step": 11779 + }, + { + "epoch": 0.45, + "learning_rate": 4.24788407283919e-05, + "loss": 0.6305, + "step": 11780 + }, + { + "epoch": 0.45, + "learning_rate": 4.247819953834317e-05, + "loss": 0.614, + "step": 11781 + }, + { + "epoch": 0.45, + "learning_rate": 4.247755834829444e-05, + "loss": 0.45, + "step": 11782 + }, + { + "epoch": 0.45, + "learning_rate": 4.2476917158245704e-05, + "loss": 0.6998, + "step": 11783 + }, + { + "epoch": 0.45, + "learning_rate": 4.2476275968196975e-05, + "loss": 0.4326, + "step": 11784 + }, + { + "epoch": 0.45, + "learning_rate": 4.247563477814824e-05, + "loss": 0.5113, + "step": 11785 + }, + { + "epoch": 0.45, + "learning_rate": 4.247499358809952e-05, + "loss": 0.4842, + "step": 11786 + }, + { + "epoch": 0.45, + "learning_rate": 4.2474352398050784e-05, + "loss": 0.4506, + "step": 11787 + }, + { + "epoch": 0.45, + "learning_rate": 4.2473711208002055e-05, + "loss": 0.5792, + "step": 11788 + }, + { + "epoch": 0.45, + "learning_rate": 4.247307001795333e-05, + "loss": 0.5677, + "step": 11789 + }, + { + "epoch": 0.45, + "learning_rate": 4.247242882790459e-05, + "loss": 0.553, + "step": 11790 + }, + { + "epoch": 0.45, + "learning_rate": 4.2471787637855864e-05, + "loss": 0.5646, + "step": 11791 + }, + { + "epoch": 0.45, + "learning_rate": 4.247114644780713e-05, + "loss": 0.4576, + "step": 11792 + }, + { + "epoch": 0.45, + "learning_rate": 4.24705052577584e-05, + "loss": 0.5206, + "step": 11793 + }, + { + "epoch": 0.45, + "learning_rate": 4.246986406770967e-05, + "loss": 0.4814, + "step": 11794 + }, + { + "epoch": 0.45, + "learning_rate": 4.2469222877660944e-05, + "loss": 0.5178, + "step": 11795 + }, + { + "epoch": 0.45, + "learning_rate": 4.246858168761221e-05, + "loss": 0.4351, + "step": 11796 + }, + { + "epoch": 0.45, + "learning_rate": 4.246794049756348e-05, + "loss": 0.605, + "step": 11797 + }, + { + "epoch": 0.45, + "learning_rate": 4.246729930751475e-05, + "loss": 0.4908, + "step": 11798 + }, + { + "epoch": 0.45, + "learning_rate": 4.246665811746602e-05, + "loss": 0.5189, + "step": 11799 + }, + { + "epoch": 0.45, + "learning_rate": 4.246601692741729e-05, + "loss": 0.4424, + "step": 11800 + }, + { + "epoch": 0.45, + "learning_rate": 4.2465375737368554e-05, + "loss": 0.5339, + "step": 11801 + }, + { + "epoch": 0.45, + "learning_rate": 4.2464734547319825e-05, + "loss": 0.5203, + "step": 11802 + }, + { + "epoch": 0.45, + "learning_rate": 4.24640933572711e-05, + "loss": 0.4814, + "step": 11803 + }, + { + "epoch": 0.45, + "learning_rate": 4.246345216722237e-05, + "loss": 0.6446, + "step": 11804 + }, + { + "epoch": 0.45, + "learning_rate": 4.2462810977173634e-05, + "loss": 0.6145, + "step": 11805 + }, + { + "epoch": 0.45, + "learning_rate": 4.2462169787124906e-05, + "loss": 0.757, + "step": 11806 + }, + { + "epoch": 0.45, + "learning_rate": 4.246152859707618e-05, + "loss": 0.5988, + "step": 11807 + }, + { + "epoch": 0.45, + "learning_rate": 4.246088740702744e-05, + "loss": 0.5149, + "step": 11808 + }, + { + "epoch": 0.45, + "learning_rate": 4.2460246216978714e-05, + "loss": 0.5474, + "step": 11809 + }, + { + "epoch": 0.45, + "learning_rate": 4.2459605026929986e-05, + "loss": 0.5142, + "step": 11810 + }, + { + "epoch": 0.45, + "learning_rate": 4.245896383688126e-05, + "loss": 0.5906, + "step": 11811 + }, + { + "epoch": 0.45, + "learning_rate": 4.245832264683252e-05, + "loss": 0.5102, + "step": 11812 + }, + { + "epoch": 0.45, + "learning_rate": 4.2457681456783794e-05, + "loss": 0.675, + "step": 11813 + }, + { + "epoch": 0.45, + "learning_rate": 4.245704026673506e-05, + "loss": 0.5222, + "step": 11814 + }, + { + "epoch": 0.45, + "learning_rate": 4.245639907668633e-05, + "loss": 0.479, + "step": 11815 + }, + { + "epoch": 0.45, + "learning_rate": 4.24557578866376e-05, + "loss": 0.5296, + "step": 11816 + }, + { + "epoch": 0.45, + "learning_rate": 4.245511669658887e-05, + "loss": 0.5193, + "step": 11817 + }, + { + "epoch": 0.45, + "learning_rate": 4.245447550654014e-05, + "loss": 0.5505, + "step": 11818 + }, + { + "epoch": 0.45, + "learning_rate": 4.245383431649141e-05, + "loss": 0.5574, + "step": 11819 + }, + { + "epoch": 0.45, + "learning_rate": 4.245319312644268e-05, + "loss": 0.4831, + "step": 11820 + }, + { + "epoch": 0.45, + "learning_rate": 4.245255193639395e-05, + "loss": 0.5455, + "step": 11821 + }, + { + "epoch": 0.45, + "learning_rate": 4.245191074634522e-05, + "loss": 0.5816, + "step": 11822 + }, + { + "epoch": 0.45, + "learning_rate": 4.2451269556296484e-05, + "loss": 0.643, + "step": 11823 + }, + { + "epoch": 0.45, + "learning_rate": 4.2450628366247756e-05, + "loss": 0.736, + "step": 11824 + }, + { + "epoch": 0.45, + "learning_rate": 4.244998717619903e-05, + "loss": 0.6683, + "step": 11825 + }, + { + "epoch": 0.45, + "learning_rate": 4.24493459861503e-05, + "loss": 0.551, + "step": 11826 + }, + { + "epoch": 0.45, + "learning_rate": 4.244870479610157e-05, + "loss": 0.6513, + "step": 11827 + }, + { + "epoch": 0.45, + "learning_rate": 4.2448063606052836e-05, + "loss": 0.581, + "step": 11828 + }, + { + "epoch": 0.45, + "learning_rate": 4.244742241600411e-05, + "loss": 0.6264, + "step": 11829 + }, + { + "epoch": 0.45, + "learning_rate": 4.244678122595537e-05, + "loss": 0.5528, + "step": 11830 + }, + { + "epoch": 0.45, + "learning_rate": 4.2446140035906644e-05, + "loss": 0.4682, + "step": 11831 + }, + { + "epoch": 0.45, + "learning_rate": 4.244549884585791e-05, + "loss": 0.5614, + "step": 11832 + }, + { + "epoch": 0.45, + "learning_rate": 4.244485765580918e-05, + "loss": 0.5428, + "step": 11833 + }, + { + "epoch": 0.45, + "learning_rate": 4.244421646576045e-05, + "loss": 0.5146, + "step": 11834 + }, + { + "epoch": 0.46, + "learning_rate": 4.2443575275711724e-05, + "loss": 0.5817, + "step": 11835 + }, + { + "epoch": 0.46, + "learning_rate": 4.2442934085662996e-05, + "loss": 0.58, + "step": 11836 + }, + { + "epoch": 0.46, + "learning_rate": 4.244229289561426e-05, + "loss": 0.5391, + "step": 11837 + }, + { + "epoch": 0.46, + "learning_rate": 4.244165170556553e-05, + "loss": 0.6082, + "step": 11838 + }, + { + "epoch": 0.46, + "learning_rate": 4.24410105155168e-05, + "loss": 0.6616, + "step": 11839 + }, + { + "epoch": 0.46, + "learning_rate": 4.244036932546807e-05, + "loss": 0.5402, + "step": 11840 + }, + { + "epoch": 0.46, + "learning_rate": 4.2439728135419334e-05, + "loss": 0.5038, + "step": 11841 + }, + { + "epoch": 0.46, + "learning_rate": 4.243908694537061e-05, + "loss": 0.5683, + "step": 11842 + }, + { + "epoch": 0.46, + "learning_rate": 4.2438445755321885e-05, + "loss": 0.4961, + "step": 11843 + }, + { + "epoch": 0.46, + "learning_rate": 4.243780456527315e-05, + "loss": 0.5172, + "step": 11844 + }, + { + "epoch": 0.46, + "learning_rate": 4.243716337522442e-05, + "loss": 0.6036, + "step": 11845 + }, + { + "epoch": 0.46, + "learning_rate": 4.2436522185175686e-05, + "loss": 0.5033, + "step": 11846 + }, + { + "epoch": 0.46, + "learning_rate": 4.243588099512696e-05, + "loss": 0.5091, + "step": 11847 + }, + { + "epoch": 0.46, + "learning_rate": 4.243523980507822e-05, + "loss": 0.5719, + "step": 11848 + }, + { + "epoch": 0.46, + "learning_rate": 4.2434598615029495e-05, + "loss": 0.4831, + "step": 11849 + }, + { + "epoch": 0.46, + "learning_rate": 4.2433957424980766e-05, + "loss": 0.5198, + "step": 11850 + }, + { + "epoch": 0.46, + "learning_rate": 4.243331623493204e-05, + "loss": 0.582, + "step": 11851 + }, + { + "epoch": 0.46, + "learning_rate": 4.243267504488331e-05, + "loss": 0.6047, + "step": 11852 + }, + { + "epoch": 0.46, + "learning_rate": 4.2432033854834575e-05, + "loss": 0.5672, + "step": 11853 + }, + { + "epoch": 0.46, + "learning_rate": 4.2431392664785846e-05, + "loss": 0.5456, + "step": 11854 + }, + { + "epoch": 0.46, + "learning_rate": 4.243075147473711e-05, + "loss": 0.5749, + "step": 11855 + }, + { + "epoch": 0.46, + "learning_rate": 4.243011028468838e-05, + "loss": 0.4949, + "step": 11856 + }, + { + "epoch": 0.46, + "learning_rate": 4.242946909463965e-05, + "loss": 0.4931, + "step": 11857 + }, + { + "epoch": 0.46, + "learning_rate": 4.2428827904590926e-05, + "loss": 0.5015, + "step": 11858 + }, + { + "epoch": 0.46, + "learning_rate": 4.242818671454219e-05, + "loss": 0.5095, + "step": 11859 + }, + { + "epoch": 0.46, + "learning_rate": 4.242754552449346e-05, + "loss": 0.4272, + "step": 11860 + }, + { + "epoch": 0.46, + "learning_rate": 4.2426904334444735e-05, + "loss": 0.6071, + "step": 11861 + }, + { + "epoch": 0.46, + "learning_rate": 4.2426263144396e-05, + "loss": 0.5283, + "step": 11862 + }, + { + "epoch": 0.46, + "learning_rate": 4.242562195434727e-05, + "loss": 0.5153, + "step": 11863 + }, + { + "epoch": 0.46, + "learning_rate": 4.2424980764298536e-05, + "loss": 0.4747, + "step": 11864 + }, + { + "epoch": 0.46, + "learning_rate": 4.242433957424981e-05, + "loss": 0.4986, + "step": 11865 + }, + { + "epoch": 0.46, + "learning_rate": 4.242369838420108e-05, + "loss": 0.5304, + "step": 11866 + }, + { + "epoch": 0.46, + "learning_rate": 4.242305719415235e-05, + "loss": 0.728, + "step": 11867 + }, + { + "epoch": 0.46, + "learning_rate": 4.2422416004103616e-05, + "loss": 0.5278, + "step": 11868 + }, + { + "epoch": 0.46, + "learning_rate": 4.242177481405489e-05, + "loss": 0.5127, + "step": 11869 + }, + { + "epoch": 0.46, + "learning_rate": 4.242113362400616e-05, + "loss": 0.5177, + "step": 11870 + }, + { + "epoch": 0.46, + "learning_rate": 4.2420492433957425e-05, + "loss": 0.4826, + "step": 11871 + }, + { + "epoch": 0.46, + "learning_rate": 4.2419851243908697e-05, + "loss": 0.4722, + "step": 11872 + }, + { + "epoch": 0.46, + "learning_rate": 4.241921005385996e-05, + "loss": 0.4906, + "step": 11873 + }, + { + "epoch": 0.46, + "learning_rate": 4.241856886381123e-05, + "loss": 0.5404, + "step": 11874 + }, + { + "epoch": 0.46, + "learning_rate": 4.2417927673762505e-05, + "loss": 0.5212, + "step": 11875 + }, + { + "epoch": 0.46, + "learning_rate": 4.241728648371378e-05, + "loss": 0.5156, + "step": 11876 + }, + { + "epoch": 0.46, + "learning_rate": 4.241664529366504e-05, + "loss": 0.5457, + "step": 11877 + }, + { + "epoch": 0.46, + "learning_rate": 4.241600410361631e-05, + "loss": 0.5604, + "step": 11878 + }, + { + "epoch": 0.46, + "learning_rate": 4.2415362913567585e-05, + "loss": 0.5803, + "step": 11879 + }, + { + "epoch": 0.46, + "learning_rate": 4.241472172351885e-05, + "loss": 0.483, + "step": 11880 + }, + { + "epoch": 0.46, + "learning_rate": 4.241408053347012e-05, + "loss": 0.5545, + "step": 11881 + }, + { + "epoch": 0.46, + "learning_rate": 4.2413439343421393e-05, + "loss": 0.5466, + "step": 11882 + }, + { + "epoch": 0.46, + "learning_rate": 4.2412798153372665e-05, + "loss": 0.5636, + "step": 11883 + }, + { + "epoch": 0.46, + "learning_rate": 4.241215696332393e-05, + "loss": 0.5878, + "step": 11884 + }, + { + "epoch": 0.46, + "learning_rate": 4.24115157732752e-05, + "loss": 0.6499, + "step": 11885 + }, + { + "epoch": 0.46, + "learning_rate": 4.2410874583226474e-05, + "loss": 0.5546, + "step": 11886 + }, + { + "epoch": 0.46, + "learning_rate": 4.241023339317774e-05, + "loss": 0.6116, + "step": 11887 + }, + { + "epoch": 0.46, + "learning_rate": 4.240959220312901e-05, + "loss": 0.6889, + "step": 11888 + }, + { + "epoch": 0.46, + "learning_rate": 4.2408951013080275e-05, + "loss": 0.4714, + "step": 11889 + }, + { + "epoch": 0.46, + "learning_rate": 4.240830982303155e-05, + "loss": 0.5754, + "step": 11890 + }, + { + "epoch": 0.46, + "learning_rate": 4.240766863298282e-05, + "loss": 0.5744, + "step": 11891 + }, + { + "epoch": 0.46, + "learning_rate": 4.240702744293409e-05, + "loss": 0.5003, + "step": 11892 + }, + { + "epoch": 0.46, + "learning_rate": 4.2406386252885355e-05, + "loss": 0.5888, + "step": 11893 + }, + { + "epoch": 0.46, + "learning_rate": 4.240574506283663e-05, + "loss": 0.5723, + "step": 11894 + }, + { + "epoch": 0.46, + "learning_rate": 4.24051038727879e-05, + "loss": 0.5869, + "step": 11895 + }, + { + "epoch": 0.46, + "learning_rate": 4.2404462682739164e-05, + "loss": 0.4705, + "step": 11896 + }, + { + "epoch": 0.46, + "learning_rate": 4.2403821492690435e-05, + "loss": 0.627, + "step": 11897 + }, + { + "epoch": 0.46, + "learning_rate": 4.240318030264171e-05, + "loss": 0.5318, + "step": 11898 + }, + { + "epoch": 0.46, + "learning_rate": 4.240253911259298e-05, + "loss": 0.5168, + "step": 11899 + }, + { + "epoch": 0.46, + "learning_rate": 4.2401897922544244e-05, + "loss": 0.4641, + "step": 11900 + }, + { + "epoch": 0.46, + "learning_rate": 4.2401256732495515e-05, + "loss": 0.592, + "step": 11901 + }, + { + "epoch": 0.46, + "learning_rate": 4.240061554244678e-05, + "loss": 0.5326, + "step": 11902 + }, + { + "epoch": 0.46, + "learning_rate": 4.239997435239805e-05, + "loss": 0.6424, + "step": 11903 + }, + { + "epoch": 0.46, + "learning_rate": 4.2399333162349324e-05, + "loss": 0.5492, + "step": 11904 + }, + { + "epoch": 0.46, + "learning_rate": 4.239869197230059e-05, + "loss": 0.4691, + "step": 11905 + }, + { + "epoch": 0.46, + "learning_rate": 4.239805078225186e-05, + "loss": 0.4884, + "step": 11906 + }, + { + "epoch": 0.46, + "learning_rate": 4.239740959220313e-05, + "loss": 0.5722, + "step": 11907 + }, + { + "epoch": 0.46, + "learning_rate": 4.2396768402154404e-05, + "loss": 0.6658, + "step": 11908 + }, + { + "epoch": 0.46, + "learning_rate": 4.239612721210567e-05, + "loss": 0.6786, + "step": 11909 + }, + { + "epoch": 0.46, + "learning_rate": 4.239548602205694e-05, + "loss": 0.4995, + "step": 11910 + }, + { + "epoch": 0.46, + "learning_rate": 4.2394844832008205e-05, + "loss": 0.5505, + "step": 11911 + }, + { + "epoch": 0.46, + "learning_rate": 4.239420364195948e-05, + "loss": 0.688, + "step": 11912 + }, + { + "epoch": 0.46, + "learning_rate": 4.239356245191075e-05, + "loss": 0.5349, + "step": 11913 + }, + { + "epoch": 0.46, + "learning_rate": 4.239292126186202e-05, + "loss": 0.5766, + "step": 11914 + }, + { + "epoch": 0.46, + "learning_rate": 4.239228007181329e-05, + "loss": 0.5207, + "step": 11915 + }, + { + "epoch": 0.46, + "learning_rate": 4.239163888176456e-05, + "loss": 0.5993, + "step": 11916 + }, + { + "epoch": 0.46, + "learning_rate": 4.239099769171583e-05, + "loss": 0.6287, + "step": 11917 + }, + { + "epoch": 0.46, + "learning_rate": 4.2390356501667094e-05, + "loss": 0.5549, + "step": 11918 + }, + { + "epoch": 0.46, + "learning_rate": 4.2389715311618366e-05, + "loss": 0.5197, + "step": 11919 + }, + { + "epoch": 0.46, + "learning_rate": 4.238907412156963e-05, + "loss": 0.6414, + "step": 11920 + }, + { + "epoch": 0.46, + "learning_rate": 4.23884329315209e-05, + "loss": 0.5521, + "step": 11921 + }, + { + "epoch": 0.46, + "learning_rate": 4.2387791741472174e-05, + "loss": 0.5456, + "step": 11922 + }, + { + "epoch": 0.46, + "learning_rate": 4.2387150551423446e-05, + "loss": 0.512, + "step": 11923 + }, + { + "epoch": 0.46, + "learning_rate": 4.238650936137472e-05, + "loss": 0.5845, + "step": 11924 + }, + { + "epoch": 0.46, + "learning_rate": 4.238586817132598e-05, + "loss": 0.578, + "step": 11925 + }, + { + "epoch": 0.46, + "learning_rate": 4.2385226981277254e-05, + "loss": 0.509, + "step": 11926 + }, + { + "epoch": 0.46, + "learning_rate": 4.238458579122852e-05, + "loss": 0.5127, + "step": 11927 + }, + { + "epoch": 0.46, + "learning_rate": 4.238394460117979e-05, + "loss": 0.503, + "step": 11928 + }, + { + "epoch": 0.46, + "learning_rate": 4.2383303411131056e-05, + "loss": 0.5263, + "step": 11929 + }, + { + "epoch": 0.46, + "learning_rate": 4.2382662221082334e-05, + "loss": 0.5716, + "step": 11930 + }, + { + "epoch": 0.46, + "learning_rate": 4.2382021031033606e-05, + "loss": 0.5068, + "step": 11931 + }, + { + "epoch": 0.46, + "learning_rate": 4.238137984098487e-05, + "loss": 0.5374, + "step": 11932 + }, + { + "epoch": 0.46, + "learning_rate": 4.238073865093614e-05, + "loss": 0.5951, + "step": 11933 + }, + { + "epoch": 0.46, + "learning_rate": 4.238009746088741e-05, + "loss": 0.5969, + "step": 11934 + }, + { + "epoch": 0.46, + "learning_rate": 4.237945627083868e-05, + "loss": 0.5144, + "step": 11935 + }, + { + "epoch": 0.46, + "learning_rate": 4.2378815080789944e-05, + "loss": 0.6123, + "step": 11936 + }, + { + "epoch": 0.46, + "learning_rate": 4.2378173890741216e-05, + "loss": 0.5349, + "step": 11937 + }, + { + "epoch": 0.46, + "learning_rate": 4.237753270069249e-05, + "loss": 0.562, + "step": 11938 + }, + { + "epoch": 0.46, + "learning_rate": 4.237689151064376e-05, + "loss": 0.5446, + "step": 11939 + }, + { + "epoch": 0.46, + "learning_rate": 4.237625032059503e-05, + "loss": 0.5813, + "step": 11940 + }, + { + "epoch": 0.46, + "learning_rate": 4.2375609130546296e-05, + "loss": 0.6389, + "step": 11941 + }, + { + "epoch": 0.46, + "learning_rate": 4.237496794049757e-05, + "loss": 0.5881, + "step": 11942 + }, + { + "epoch": 0.46, + "learning_rate": 4.237432675044883e-05, + "loss": 0.5508, + "step": 11943 + }, + { + "epoch": 0.46, + "learning_rate": 4.2373685560400104e-05, + "loss": 0.6425, + "step": 11944 + }, + { + "epoch": 0.46, + "learning_rate": 4.237304437035137e-05, + "loss": 0.504, + "step": 11945 + }, + { + "epoch": 0.46, + "learning_rate": 4.237240318030265e-05, + "loss": 0.5958, + "step": 11946 + }, + { + "epoch": 0.46, + "learning_rate": 4.237176199025391e-05, + "loss": 0.5394, + "step": 11947 + }, + { + "epoch": 0.46, + "learning_rate": 4.2371120800205184e-05, + "loss": 0.5438, + "step": 11948 + }, + { + "epoch": 0.46, + "learning_rate": 4.2370479610156456e-05, + "loss": 0.4463, + "step": 11949 + }, + { + "epoch": 0.46, + "learning_rate": 4.236983842010772e-05, + "loss": 0.5601, + "step": 11950 + }, + { + "epoch": 0.46, + "learning_rate": 4.236919723005899e-05, + "loss": 0.4925, + "step": 11951 + }, + { + "epoch": 0.46, + "learning_rate": 4.236855604001026e-05, + "loss": 0.5785, + "step": 11952 + }, + { + "epoch": 0.46, + "learning_rate": 4.236791484996153e-05, + "loss": 0.6515, + "step": 11953 + }, + { + "epoch": 0.46, + "learning_rate": 4.23672736599128e-05, + "loss": 0.6647, + "step": 11954 + }, + { + "epoch": 0.46, + "learning_rate": 4.236663246986407e-05, + "loss": 0.5445, + "step": 11955 + }, + { + "epoch": 0.46, + "learning_rate": 4.236599127981534e-05, + "loss": 0.5564, + "step": 11956 + }, + { + "epoch": 0.46, + "learning_rate": 4.236535008976661e-05, + "loss": 0.4385, + "step": 11957 + }, + { + "epoch": 0.46, + "learning_rate": 4.236470889971788e-05, + "loss": 0.524, + "step": 11958 + }, + { + "epoch": 0.46, + "learning_rate": 4.2364067709669146e-05, + "loss": 0.5648, + "step": 11959 + }, + { + "epoch": 0.46, + "learning_rate": 4.236342651962042e-05, + "loss": 0.5541, + "step": 11960 + }, + { + "epoch": 0.46, + "learning_rate": 4.236278532957168e-05, + "loss": 0.5671, + "step": 11961 + }, + { + "epoch": 0.46, + "learning_rate": 4.2362144139522955e-05, + "loss": 0.6022, + "step": 11962 + }, + { + "epoch": 0.46, + "learning_rate": 4.2361502949474226e-05, + "loss": 0.477, + "step": 11963 + }, + { + "epoch": 0.46, + "learning_rate": 4.23608617594255e-05, + "loss": 0.5538, + "step": 11964 + }, + { + "epoch": 0.46, + "learning_rate": 4.236022056937676e-05, + "loss": 0.5054, + "step": 11965 + }, + { + "epoch": 0.46, + "learning_rate": 4.2359579379328035e-05, + "loss": 0.4291, + "step": 11966 + }, + { + "epoch": 0.46, + "learning_rate": 4.2358938189279306e-05, + "loss": 0.5205, + "step": 11967 + }, + { + "epoch": 0.46, + "learning_rate": 4.235829699923057e-05, + "loss": 0.6077, + "step": 11968 + }, + { + "epoch": 0.46, + "learning_rate": 4.235765580918184e-05, + "loss": 0.5642, + "step": 11969 + }, + { + "epoch": 0.46, + "learning_rate": 4.2357014619133115e-05, + "loss": 0.5649, + "step": 11970 + }, + { + "epoch": 0.46, + "learning_rate": 4.2356373429084386e-05, + "loss": 0.4602, + "step": 11971 + }, + { + "epoch": 0.46, + "learning_rate": 4.235573223903565e-05, + "loss": 0.6458, + "step": 11972 + }, + { + "epoch": 0.46, + "learning_rate": 4.235509104898692e-05, + "loss": 0.5706, + "step": 11973 + }, + { + "epoch": 0.46, + "learning_rate": 4.235444985893819e-05, + "loss": 0.6437, + "step": 11974 + }, + { + "epoch": 0.46, + "learning_rate": 4.235380866888946e-05, + "loss": 0.638, + "step": 11975 + }, + { + "epoch": 0.46, + "learning_rate": 4.235316747884073e-05, + "loss": 0.421, + "step": 11976 + }, + { + "epoch": 0.46, + "learning_rate": 4.2352526288791996e-05, + "loss": 0.5538, + "step": 11977 + }, + { + "epoch": 0.46, + "learning_rate": 4.235188509874327e-05, + "loss": 0.5316, + "step": 11978 + }, + { + "epoch": 0.46, + "learning_rate": 4.235124390869454e-05, + "loss": 0.5934, + "step": 11979 + }, + { + "epoch": 0.46, + "learning_rate": 4.235060271864581e-05, + "loss": 0.522, + "step": 11980 + }, + { + "epoch": 0.46, + "learning_rate": 4.2349961528597077e-05, + "loss": 0.5055, + "step": 11981 + }, + { + "epoch": 0.46, + "learning_rate": 4.234932033854835e-05, + "loss": 0.5978, + "step": 11982 + }, + { + "epoch": 0.46, + "learning_rate": 4.234867914849961e-05, + "loss": 0.6822, + "step": 11983 + }, + { + "epoch": 0.46, + "learning_rate": 4.2348037958450885e-05, + "loss": 0.6214, + "step": 11984 + }, + { + "epoch": 0.46, + "learning_rate": 4.234739676840216e-05, + "loss": 0.4789, + "step": 11985 + }, + { + "epoch": 0.46, + "learning_rate": 4.234675557835343e-05, + "loss": 0.5373, + "step": 11986 + }, + { + "epoch": 0.46, + "learning_rate": 4.23461143883047e-05, + "loss": 0.5347, + "step": 11987 + }, + { + "epoch": 0.46, + "learning_rate": 4.2345473198255965e-05, + "loss": 0.6236, + "step": 11988 + }, + { + "epoch": 0.46, + "learning_rate": 4.234483200820724e-05, + "loss": 0.5161, + "step": 11989 + }, + { + "epoch": 0.46, + "learning_rate": 4.23441908181585e-05, + "loss": 0.5808, + "step": 11990 + }, + { + "epoch": 0.46, + "learning_rate": 4.234354962810977e-05, + "loss": 0.4663, + "step": 11991 + }, + { + "epoch": 0.46, + "learning_rate": 4.234290843806104e-05, + "loss": 0.4997, + "step": 11992 + }, + { + "epoch": 0.46, + "learning_rate": 4.234226724801231e-05, + "loss": 0.5041, + "step": 11993 + }, + { + "epoch": 0.46, + "learning_rate": 4.234162605796358e-05, + "loss": 0.6365, + "step": 11994 + }, + { + "epoch": 0.46, + "learning_rate": 4.2340984867914853e-05, + "loss": 0.5879, + "step": 11995 + }, + { + "epoch": 0.46, + "learning_rate": 4.2340343677866125e-05, + "loss": 0.5504, + "step": 11996 + }, + { + "epoch": 0.46, + "learning_rate": 4.233970248781739e-05, + "loss": 0.548, + "step": 11997 + }, + { + "epoch": 0.46, + "learning_rate": 4.233906129776866e-05, + "loss": 0.4776, + "step": 11998 + }, + { + "epoch": 0.46, + "learning_rate": 4.233842010771993e-05, + "loss": 0.5824, + "step": 11999 + }, + { + "epoch": 0.46, + "learning_rate": 4.23377789176712e-05, + "loss": 0.5661, + "step": 12000 + }, + { + "epoch": 0.46, + "learning_rate": 4.2337137727622463e-05, + "loss": 0.6114, + "step": 12001 + }, + { + "epoch": 0.46, + "learning_rate": 4.233649653757374e-05, + "loss": 0.6086, + "step": 12002 + }, + { + "epoch": 0.46, + "learning_rate": 4.2335855347525014e-05, + "loss": 0.4698, + "step": 12003 + }, + { + "epoch": 0.46, + "learning_rate": 4.233521415747628e-05, + "loss": 0.639, + "step": 12004 + }, + { + "epoch": 0.46, + "learning_rate": 4.233457296742755e-05, + "loss": 0.5525, + "step": 12005 + }, + { + "epoch": 0.46, + "learning_rate": 4.2333931777378815e-05, + "loss": 0.6398, + "step": 12006 + }, + { + "epoch": 0.46, + "learning_rate": 4.233329058733009e-05, + "loss": 0.5965, + "step": 12007 + }, + { + "epoch": 0.46, + "learning_rate": 4.233264939728135e-05, + "loss": 0.592, + "step": 12008 + }, + { + "epoch": 0.46, + "learning_rate": 4.2332008207232624e-05, + "loss": 0.5183, + "step": 12009 + }, + { + "epoch": 0.46, + "learning_rate": 4.2331367017183895e-05, + "loss": 0.5765, + "step": 12010 + }, + { + "epoch": 0.46, + "learning_rate": 4.233072582713517e-05, + "loss": 0.4918, + "step": 12011 + }, + { + "epoch": 0.46, + "learning_rate": 4.233008463708644e-05, + "loss": 0.5039, + "step": 12012 + }, + { + "epoch": 0.46, + "learning_rate": 4.2329443447037704e-05, + "loss": 0.5667, + "step": 12013 + }, + { + "epoch": 0.46, + "learning_rate": 4.2328802256988975e-05, + "loss": 0.5317, + "step": 12014 + }, + { + "epoch": 0.46, + "learning_rate": 4.232816106694024e-05, + "loss": 0.5013, + "step": 12015 + }, + { + "epoch": 0.46, + "learning_rate": 4.232751987689151e-05, + "loss": 0.5407, + "step": 12016 + }, + { + "epoch": 0.46, + "learning_rate": 4.232687868684278e-05, + "loss": 0.5449, + "step": 12017 + }, + { + "epoch": 0.46, + "learning_rate": 4.2326237496794055e-05, + "loss": 0.501, + "step": 12018 + }, + { + "epoch": 0.46, + "learning_rate": 4.232559630674532e-05, + "loss": 0.4998, + "step": 12019 + }, + { + "epoch": 0.46, + "learning_rate": 4.232495511669659e-05, + "loss": 0.5316, + "step": 12020 + }, + { + "epoch": 0.46, + "learning_rate": 4.2324313926647864e-05, + "loss": 0.5284, + "step": 12021 + }, + { + "epoch": 0.46, + "learning_rate": 4.232367273659913e-05, + "loss": 0.7048, + "step": 12022 + }, + { + "epoch": 0.46, + "learning_rate": 4.23230315465504e-05, + "loss": 0.6294, + "step": 12023 + }, + { + "epoch": 0.46, + "learning_rate": 4.2322390356501665e-05, + "loss": 0.5834, + "step": 12024 + }, + { + "epoch": 0.46, + "learning_rate": 4.232174916645294e-05, + "loss": 0.4969, + "step": 12025 + }, + { + "epoch": 0.46, + "learning_rate": 4.232110797640421e-05, + "loss": 0.5904, + "step": 12026 + }, + { + "epoch": 0.46, + "learning_rate": 4.232046678635548e-05, + "loss": 0.5258, + "step": 12027 + }, + { + "epoch": 0.46, + "learning_rate": 4.2319825596306746e-05, + "loss": 0.5232, + "step": 12028 + }, + { + "epoch": 0.46, + "learning_rate": 4.231918440625802e-05, + "loss": 0.5678, + "step": 12029 + }, + { + "epoch": 0.46, + "learning_rate": 4.231854321620929e-05, + "loss": 0.5608, + "step": 12030 + }, + { + "epoch": 0.46, + "learning_rate": 4.2317902026160554e-05, + "loss": 0.5481, + "step": 12031 + }, + { + "epoch": 0.46, + "learning_rate": 4.2317260836111826e-05, + "loss": 0.4797, + "step": 12032 + }, + { + "epoch": 0.46, + "learning_rate": 4.231661964606309e-05, + "loss": 0.5896, + "step": 12033 + }, + { + "epoch": 0.46, + "learning_rate": 4.231597845601437e-05, + "loss": 0.5362, + "step": 12034 + }, + { + "epoch": 0.46, + "learning_rate": 4.2315337265965634e-05, + "loss": 0.5766, + "step": 12035 + }, + { + "epoch": 0.46, + "learning_rate": 4.2314696075916906e-05, + "loss": 0.5898, + "step": 12036 + }, + { + "epoch": 0.46, + "learning_rate": 4.231405488586817e-05, + "loss": 0.4385, + "step": 12037 + }, + { + "epoch": 0.46, + "learning_rate": 4.231341369581944e-05, + "loss": 0.5906, + "step": 12038 + }, + { + "epoch": 0.46, + "learning_rate": 4.2312772505770714e-05, + "loss": 0.5887, + "step": 12039 + }, + { + "epoch": 0.46, + "learning_rate": 4.231213131572198e-05, + "loss": 0.5611, + "step": 12040 + }, + { + "epoch": 0.46, + "learning_rate": 4.231149012567325e-05, + "loss": 0.5651, + "step": 12041 + }, + { + "epoch": 0.46, + "learning_rate": 4.231084893562452e-05, + "loss": 0.5935, + "step": 12042 + }, + { + "epoch": 0.46, + "learning_rate": 4.2310207745575794e-05, + "loss": 0.5958, + "step": 12043 + }, + { + "epoch": 0.46, + "learning_rate": 4.230956655552706e-05, + "loss": 0.5094, + "step": 12044 + }, + { + "epoch": 0.46, + "learning_rate": 4.230892536547833e-05, + "loss": 0.5421, + "step": 12045 + }, + { + "epoch": 0.46, + "learning_rate": 4.2308284175429596e-05, + "loss": 0.5707, + "step": 12046 + }, + { + "epoch": 0.46, + "learning_rate": 4.230764298538087e-05, + "loss": 0.5936, + "step": 12047 + }, + { + "epoch": 0.46, + "learning_rate": 4.230700179533214e-05, + "loss": 0.5196, + "step": 12048 + }, + { + "epoch": 0.46, + "learning_rate": 4.2306360605283404e-05, + "loss": 0.6253, + "step": 12049 + }, + { + "epoch": 0.46, + "learning_rate": 4.2305719415234676e-05, + "loss": 0.5177, + "step": 12050 + }, + { + "epoch": 0.46, + "learning_rate": 4.230507822518595e-05, + "loss": 0.4913, + "step": 12051 + }, + { + "epoch": 0.46, + "learning_rate": 4.230443703513722e-05, + "loss": 0.559, + "step": 12052 + }, + { + "epoch": 0.46, + "learning_rate": 4.2303795845088484e-05, + "loss": 0.5699, + "step": 12053 + }, + { + "epoch": 0.46, + "learning_rate": 4.2303154655039756e-05, + "loss": 0.5652, + "step": 12054 + }, + { + "epoch": 0.46, + "learning_rate": 4.230251346499102e-05, + "loss": 0.5802, + "step": 12055 + }, + { + "epoch": 0.46, + "learning_rate": 4.230187227494229e-05, + "loss": 0.5239, + "step": 12056 + }, + { + "epoch": 0.46, + "learning_rate": 4.2301231084893564e-05, + "loss": 0.5143, + "step": 12057 + }, + { + "epoch": 0.46, + "learning_rate": 4.2300589894844836e-05, + "loss": 0.5644, + "step": 12058 + }, + { + "epoch": 0.46, + "learning_rate": 4.229994870479611e-05, + "loss": 0.5638, + "step": 12059 + }, + { + "epoch": 0.46, + "learning_rate": 4.229930751474737e-05, + "loss": 0.5364, + "step": 12060 + }, + { + "epoch": 0.46, + "learning_rate": 4.2298666324698644e-05, + "loss": 0.5569, + "step": 12061 + }, + { + "epoch": 0.46, + "learning_rate": 4.229802513464991e-05, + "loss": 0.4145, + "step": 12062 + }, + { + "epoch": 0.46, + "learning_rate": 4.229738394460118e-05, + "loss": 0.6914, + "step": 12063 + }, + { + "epoch": 0.46, + "learning_rate": 4.2296742754552446e-05, + "loss": 0.4436, + "step": 12064 + }, + { + "epoch": 0.46, + "learning_rate": 4.229610156450372e-05, + "loss": 0.6146, + "step": 12065 + }, + { + "epoch": 0.46, + "learning_rate": 4.229546037445499e-05, + "loss": 0.5788, + "step": 12066 + }, + { + "epoch": 0.46, + "learning_rate": 4.229481918440626e-05, + "loss": 0.4962, + "step": 12067 + }, + { + "epoch": 0.46, + "learning_rate": 4.229417799435753e-05, + "loss": 0.5441, + "step": 12068 + }, + { + "epoch": 0.46, + "learning_rate": 4.22935368043088e-05, + "loss": 0.4701, + "step": 12069 + }, + { + "epoch": 0.46, + "learning_rate": 4.229289561426007e-05, + "loss": 0.4776, + "step": 12070 + }, + { + "epoch": 0.46, + "learning_rate": 4.2292254424211335e-05, + "loss": 0.6993, + "step": 12071 + }, + { + "epoch": 0.46, + "learning_rate": 4.2291613234162606e-05, + "loss": 0.6217, + "step": 12072 + }, + { + "epoch": 0.46, + "learning_rate": 4.229097204411388e-05, + "loss": 0.5406, + "step": 12073 + }, + { + "epoch": 0.46, + "learning_rate": 4.229033085406515e-05, + "loss": 0.6227, + "step": 12074 + }, + { + "epoch": 0.46, + "learning_rate": 4.228968966401642e-05, + "loss": 0.5225, + "step": 12075 + }, + { + "epoch": 0.46, + "learning_rate": 4.2289048473967686e-05, + "loss": 0.6884, + "step": 12076 + }, + { + "epoch": 0.46, + "learning_rate": 4.228840728391896e-05, + "loss": 0.5499, + "step": 12077 + }, + { + "epoch": 0.46, + "learning_rate": 4.228776609387022e-05, + "loss": 0.6102, + "step": 12078 + }, + { + "epoch": 0.46, + "learning_rate": 4.2287124903821495e-05, + "loss": 0.5413, + "step": 12079 + }, + { + "epoch": 0.46, + "learning_rate": 4.228648371377276e-05, + "loss": 0.5778, + "step": 12080 + }, + { + "epoch": 0.46, + "learning_rate": 4.228584252372403e-05, + "loss": 0.4672, + "step": 12081 + }, + { + "epoch": 0.46, + "learning_rate": 4.22852013336753e-05, + "loss": 0.6324, + "step": 12082 + }, + { + "epoch": 0.46, + "learning_rate": 4.2284560143626575e-05, + "loss": 0.4738, + "step": 12083 + }, + { + "epoch": 0.46, + "learning_rate": 4.2283918953577846e-05, + "loss": 0.452, + "step": 12084 + }, + { + "epoch": 0.46, + "learning_rate": 4.228327776352911e-05, + "loss": 0.5389, + "step": 12085 + }, + { + "epoch": 0.46, + "learning_rate": 4.228263657348038e-05, + "loss": 0.5791, + "step": 12086 + }, + { + "epoch": 0.46, + "learning_rate": 4.228199538343165e-05, + "loss": 0.6544, + "step": 12087 + }, + { + "epoch": 0.46, + "learning_rate": 4.228135419338292e-05, + "loss": 0.4793, + "step": 12088 + }, + { + "epoch": 0.46, + "learning_rate": 4.2280713003334185e-05, + "loss": 0.5271, + "step": 12089 + }, + { + "epoch": 0.46, + "learning_rate": 4.228007181328546e-05, + "loss": 0.5539, + "step": 12090 + }, + { + "epoch": 0.46, + "learning_rate": 4.2279430623236735e-05, + "loss": 0.5564, + "step": 12091 + }, + { + "epoch": 0.46, + "learning_rate": 4.2278789433188e-05, + "loss": 0.5202, + "step": 12092 + }, + { + "epoch": 0.46, + "learning_rate": 4.227814824313927e-05, + "loss": 0.4964, + "step": 12093 + }, + { + "epoch": 0.46, + "learning_rate": 4.2277507053090537e-05, + "loss": 0.5342, + "step": 12094 + }, + { + "epoch": 0.47, + "learning_rate": 4.227686586304181e-05, + "loss": 0.5457, + "step": 12095 + }, + { + "epoch": 0.47, + "learning_rate": 4.227622467299307e-05, + "loss": 0.5309, + "step": 12096 + }, + { + "epoch": 0.47, + "learning_rate": 4.2275583482944345e-05, + "loss": 0.4863, + "step": 12097 + }, + { + "epoch": 0.47, + "learning_rate": 4.227494229289562e-05, + "loss": 0.464, + "step": 12098 + }, + { + "epoch": 0.47, + "learning_rate": 4.227430110284689e-05, + "loss": 0.6255, + "step": 12099 + }, + { + "epoch": 0.47, + "learning_rate": 4.227365991279816e-05, + "loss": 0.5119, + "step": 12100 + }, + { + "epoch": 0.47, + "learning_rate": 4.2273018722749425e-05, + "loss": 0.5538, + "step": 12101 + }, + { + "epoch": 0.47, + "learning_rate": 4.22723775327007e-05, + "loss": 0.5757, + "step": 12102 + }, + { + "epoch": 0.47, + "learning_rate": 4.227173634265196e-05, + "loss": 0.5902, + "step": 12103 + }, + { + "epoch": 0.47, + "learning_rate": 4.2271095152603233e-05, + "loss": 0.5905, + "step": 12104 + }, + { + "epoch": 0.47, + "learning_rate": 4.22704539625545e-05, + "loss": 0.6459, + "step": 12105 + }, + { + "epoch": 0.47, + "learning_rate": 4.226981277250578e-05, + "loss": 0.4203, + "step": 12106 + }, + { + "epoch": 0.47, + "learning_rate": 4.226917158245704e-05, + "loss": 0.4745, + "step": 12107 + }, + { + "epoch": 0.47, + "learning_rate": 4.2268530392408313e-05, + "loss": 0.5733, + "step": 12108 + }, + { + "epoch": 0.47, + "learning_rate": 4.2267889202359585e-05, + "loss": 0.6422, + "step": 12109 + }, + { + "epoch": 0.47, + "learning_rate": 4.226724801231085e-05, + "loss": 0.4755, + "step": 12110 + }, + { + "epoch": 0.47, + "learning_rate": 4.226660682226212e-05, + "loss": 0.5451, + "step": 12111 + }, + { + "epoch": 0.47, + "learning_rate": 4.226596563221339e-05, + "loss": 0.5173, + "step": 12112 + }, + { + "epoch": 0.47, + "learning_rate": 4.226532444216466e-05, + "loss": 0.4827, + "step": 12113 + }, + { + "epoch": 0.47, + "learning_rate": 4.226468325211593e-05, + "loss": 0.5802, + "step": 12114 + }, + { + "epoch": 0.47, + "learning_rate": 4.22640420620672e-05, + "loss": 0.6085, + "step": 12115 + }, + { + "epoch": 0.47, + "learning_rate": 4.226340087201847e-05, + "loss": 0.5992, + "step": 12116 + }, + { + "epoch": 0.47, + "learning_rate": 4.226275968196974e-05, + "loss": 0.5972, + "step": 12117 + }, + { + "epoch": 0.47, + "learning_rate": 4.226211849192101e-05, + "loss": 0.5816, + "step": 12118 + }, + { + "epoch": 0.47, + "learning_rate": 4.2261477301872275e-05, + "loss": 0.6695, + "step": 12119 + }, + { + "epoch": 0.47, + "learning_rate": 4.226083611182355e-05, + "loss": 0.3927, + "step": 12120 + }, + { + "epoch": 0.47, + "learning_rate": 4.226019492177481e-05, + "loss": 0.539, + "step": 12121 + }, + { + "epoch": 0.47, + "learning_rate": 4.2259553731726084e-05, + "loss": 0.4812, + "step": 12122 + }, + { + "epoch": 0.47, + "learning_rate": 4.2258912541677355e-05, + "loss": 0.5395, + "step": 12123 + }, + { + "epoch": 0.47, + "learning_rate": 4.225827135162863e-05, + "loss": 0.5003, + "step": 12124 + }, + { + "epoch": 0.47, + "learning_rate": 4.225763016157989e-05, + "loss": 0.465, + "step": 12125 + }, + { + "epoch": 0.47, + "learning_rate": 4.2256988971531164e-05, + "loss": 0.5143, + "step": 12126 + }, + { + "epoch": 0.47, + "learning_rate": 4.2256347781482435e-05, + "loss": 0.5276, + "step": 12127 + }, + { + "epoch": 0.47, + "learning_rate": 4.22557065914337e-05, + "loss": 0.5189, + "step": 12128 + }, + { + "epoch": 0.47, + "learning_rate": 4.225506540138497e-05, + "loss": 0.6124, + "step": 12129 + }, + { + "epoch": 0.47, + "learning_rate": 4.2254424211336244e-05, + "loss": 0.6429, + "step": 12130 + }, + { + "epoch": 0.47, + "learning_rate": 4.2253783021287516e-05, + "loss": 0.5427, + "step": 12131 + }, + { + "epoch": 0.47, + "learning_rate": 4.225314183123878e-05, + "loss": 0.6214, + "step": 12132 + }, + { + "epoch": 0.47, + "learning_rate": 4.225250064119005e-05, + "loss": 0.6442, + "step": 12133 + }, + { + "epoch": 0.47, + "learning_rate": 4.225185945114132e-05, + "loss": 0.4941, + "step": 12134 + }, + { + "epoch": 0.47, + "learning_rate": 4.225121826109259e-05, + "loss": 0.6157, + "step": 12135 + }, + { + "epoch": 0.47, + "learning_rate": 4.225057707104386e-05, + "loss": 0.5218, + "step": 12136 + }, + { + "epoch": 0.47, + "learning_rate": 4.2249935880995126e-05, + "loss": 0.5234, + "step": 12137 + }, + { + "epoch": 0.47, + "learning_rate": 4.22492946909464e-05, + "loss": 0.5095, + "step": 12138 + }, + { + "epoch": 0.47, + "learning_rate": 4.224865350089767e-05, + "loss": 0.6407, + "step": 12139 + }, + { + "epoch": 0.47, + "learning_rate": 4.224801231084894e-05, + "loss": 0.552, + "step": 12140 + }, + { + "epoch": 0.47, + "learning_rate": 4.2247371120800206e-05, + "loss": 0.5505, + "step": 12141 + }, + { + "epoch": 0.47, + "learning_rate": 4.224672993075148e-05, + "loss": 0.5304, + "step": 12142 + }, + { + "epoch": 0.47, + "learning_rate": 4.224608874070274e-05, + "loss": 0.5956, + "step": 12143 + }, + { + "epoch": 0.47, + "learning_rate": 4.2245447550654014e-05, + "loss": 0.6062, + "step": 12144 + }, + { + "epoch": 0.47, + "learning_rate": 4.2244806360605286e-05, + "loss": 0.4255, + "step": 12145 + }, + { + "epoch": 0.47, + "learning_rate": 4.224416517055656e-05, + "loss": 0.6674, + "step": 12146 + }, + { + "epoch": 0.47, + "learning_rate": 4.224352398050783e-05, + "loss": 0.5427, + "step": 12147 + }, + { + "epoch": 0.47, + "learning_rate": 4.2242882790459094e-05, + "loss": 0.5663, + "step": 12148 + }, + { + "epoch": 0.47, + "learning_rate": 4.2242241600410366e-05, + "loss": 0.5659, + "step": 12149 + }, + { + "epoch": 0.47, + "learning_rate": 4.224160041036163e-05, + "loss": 0.6006, + "step": 12150 + }, + { + "epoch": 0.47, + "learning_rate": 4.22409592203129e-05, + "loss": 0.4933, + "step": 12151 + }, + { + "epoch": 0.47, + "learning_rate": 4.224031803026417e-05, + "loss": 0.5108, + "step": 12152 + }, + { + "epoch": 0.47, + "learning_rate": 4.223967684021544e-05, + "loss": 0.4834, + "step": 12153 + }, + { + "epoch": 0.47, + "learning_rate": 4.223903565016671e-05, + "loss": 0.6072, + "step": 12154 + }, + { + "epoch": 0.47, + "learning_rate": 4.223839446011798e-05, + "loss": 0.6264, + "step": 12155 + }, + { + "epoch": 0.47, + "learning_rate": 4.2237753270069254e-05, + "loss": 0.5402, + "step": 12156 + }, + { + "epoch": 0.47, + "learning_rate": 4.223711208002052e-05, + "loss": 0.6813, + "step": 12157 + }, + { + "epoch": 0.47, + "learning_rate": 4.223647088997179e-05, + "loss": 0.4983, + "step": 12158 + }, + { + "epoch": 0.47, + "learning_rate": 4.2235829699923056e-05, + "loss": 0.6084, + "step": 12159 + }, + { + "epoch": 0.47, + "learning_rate": 4.223518850987433e-05, + "loss": 0.5975, + "step": 12160 + }, + { + "epoch": 0.47, + "learning_rate": 4.223454731982559e-05, + "loss": 0.5829, + "step": 12161 + }, + { + "epoch": 0.47, + "learning_rate": 4.223390612977687e-05, + "loss": 0.647, + "step": 12162 + }, + { + "epoch": 0.47, + "learning_rate": 4.223326493972814e-05, + "loss": 0.5935, + "step": 12163 + }, + { + "epoch": 0.47, + "learning_rate": 4.223262374967941e-05, + "loss": 0.5266, + "step": 12164 + }, + { + "epoch": 0.47, + "learning_rate": 4.223198255963068e-05, + "loss": 0.5338, + "step": 12165 + }, + { + "epoch": 0.47, + "learning_rate": 4.2231341369581944e-05, + "loss": 0.5583, + "step": 12166 + }, + { + "epoch": 0.47, + "learning_rate": 4.2230700179533216e-05, + "loss": 0.6014, + "step": 12167 + }, + { + "epoch": 0.47, + "learning_rate": 4.223005898948448e-05, + "loss": 0.4814, + "step": 12168 + }, + { + "epoch": 0.47, + "learning_rate": 4.222941779943575e-05, + "loss": 0.6071, + "step": 12169 + }, + { + "epoch": 0.47, + "learning_rate": 4.2228776609387024e-05, + "loss": 0.634, + "step": 12170 + }, + { + "epoch": 0.47, + "learning_rate": 4.2228135419338296e-05, + "loss": 0.6243, + "step": 12171 + }, + { + "epoch": 0.47, + "learning_rate": 4.222749422928957e-05, + "loss": 0.5913, + "step": 12172 + }, + { + "epoch": 0.47, + "learning_rate": 4.222685303924083e-05, + "loss": 0.5421, + "step": 12173 + }, + { + "epoch": 0.47, + "learning_rate": 4.2226211849192104e-05, + "loss": 0.6109, + "step": 12174 + }, + { + "epoch": 0.47, + "learning_rate": 4.222557065914337e-05, + "loss": 0.5893, + "step": 12175 + }, + { + "epoch": 0.47, + "learning_rate": 4.222492946909464e-05, + "loss": 0.5172, + "step": 12176 + }, + { + "epoch": 0.47, + "learning_rate": 4.2224288279045906e-05, + "loss": 0.572, + "step": 12177 + }, + { + "epoch": 0.47, + "learning_rate": 4.2223647088997185e-05, + "loss": 0.5527, + "step": 12178 + }, + { + "epoch": 0.47, + "learning_rate": 4.222300589894845e-05, + "loss": 0.5621, + "step": 12179 + }, + { + "epoch": 0.47, + "learning_rate": 4.222236470889972e-05, + "loss": 0.5875, + "step": 12180 + }, + { + "epoch": 0.47, + "learning_rate": 4.222172351885099e-05, + "loss": 0.5269, + "step": 12181 + }, + { + "epoch": 0.47, + "learning_rate": 4.222108232880226e-05, + "loss": 0.5465, + "step": 12182 + }, + { + "epoch": 0.47, + "learning_rate": 4.222044113875353e-05, + "loss": 0.5632, + "step": 12183 + }, + { + "epoch": 0.47, + "learning_rate": 4.2219799948704795e-05, + "loss": 0.5407, + "step": 12184 + }, + { + "epoch": 0.47, + "learning_rate": 4.2219158758656066e-05, + "loss": 0.6259, + "step": 12185 + }, + { + "epoch": 0.47, + "learning_rate": 4.221851756860734e-05, + "loss": 0.4882, + "step": 12186 + }, + { + "epoch": 0.47, + "learning_rate": 4.221787637855861e-05, + "loss": 0.4721, + "step": 12187 + }, + { + "epoch": 0.47, + "learning_rate": 4.2217235188509875e-05, + "loss": 0.5784, + "step": 12188 + }, + { + "epoch": 0.47, + "learning_rate": 4.2216593998461146e-05, + "loss": 0.5022, + "step": 12189 + }, + { + "epoch": 0.47, + "learning_rate": 4.221595280841242e-05, + "loss": 0.5291, + "step": 12190 + }, + { + "epoch": 0.47, + "learning_rate": 4.221531161836368e-05, + "loss": 0.5528, + "step": 12191 + }, + { + "epoch": 0.47, + "learning_rate": 4.2214670428314955e-05, + "loss": 0.5919, + "step": 12192 + }, + { + "epoch": 0.47, + "learning_rate": 4.221402923826622e-05, + "loss": 0.5577, + "step": 12193 + }, + { + "epoch": 0.47, + "learning_rate": 4.22133880482175e-05, + "loss": 0.5958, + "step": 12194 + }, + { + "epoch": 0.47, + "learning_rate": 4.221274685816876e-05, + "loss": 0.4638, + "step": 12195 + }, + { + "epoch": 0.47, + "learning_rate": 4.2212105668120035e-05, + "loss": 0.5504, + "step": 12196 + }, + { + "epoch": 0.47, + "learning_rate": 4.22114644780713e-05, + "loss": 0.5521, + "step": 12197 + }, + { + "epoch": 0.47, + "learning_rate": 4.221082328802257e-05, + "loss": 0.5545, + "step": 12198 + }, + { + "epoch": 0.47, + "learning_rate": 4.221018209797384e-05, + "loss": 0.4605, + "step": 12199 + }, + { + "epoch": 0.47, + "learning_rate": 4.220954090792511e-05, + "loss": 0.5816, + "step": 12200 + }, + { + "epoch": 0.47, + "learning_rate": 4.220889971787638e-05, + "loss": 0.581, + "step": 12201 + }, + { + "epoch": 0.47, + "learning_rate": 4.220825852782765e-05, + "loss": 0.5437, + "step": 12202 + }, + { + "epoch": 0.47, + "learning_rate": 4.220761733777892e-05, + "loss": 0.5475, + "step": 12203 + }, + { + "epoch": 0.47, + "learning_rate": 4.220697614773019e-05, + "loss": 0.4816, + "step": 12204 + }, + { + "epoch": 0.47, + "learning_rate": 4.220633495768146e-05, + "loss": 0.5396, + "step": 12205 + }, + { + "epoch": 0.47, + "learning_rate": 4.2205693767632725e-05, + "loss": 0.5735, + "step": 12206 + }, + { + "epoch": 0.47, + "learning_rate": 4.2205052577583997e-05, + "loss": 0.5308, + "step": 12207 + }, + { + "epoch": 0.47, + "learning_rate": 4.220441138753527e-05, + "loss": 0.4971, + "step": 12208 + }, + { + "epoch": 0.47, + "learning_rate": 4.220377019748653e-05, + "loss": 0.5764, + "step": 12209 + }, + { + "epoch": 0.47, + "learning_rate": 4.2203129007437805e-05, + "loss": 0.4276, + "step": 12210 + }, + { + "epoch": 0.47, + "learning_rate": 4.220248781738908e-05, + "loss": 0.5691, + "step": 12211 + }, + { + "epoch": 0.47, + "learning_rate": 4.220184662734035e-05, + "loss": 0.6666, + "step": 12212 + }, + { + "epoch": 0.47, + "learning_rate": 4.220120543729161e-05, + "loss": 0.4881, + "step": 12213 + }, + { + "epoch": 0.47, + "learning_rate": 4.2200564247242885e-05, + "loss": 0.5682, + "step": 12214 + }, + { + "epoch": 0.47, + "learning_rate": 4.219992305719415e-05, + "loss": 0.5581, + "step": 12215 + }, + { + "epoch": 0.47, + "learning_rate": 4.219928186714542e-05, + "loss": 0.5414, + "step": 12216 + }, + { + "epoch": 0.47, + "learning_rate": 4.2198640677096693e-05, + "loss": 0.5141, + "step": 12217 + }, + { + "epoch": 0.47, + "learning_rate": 4.2197999487047965e-05, + "loss": 0.4749, + "step": 12218 + }, + { + "epoch": 0.47, + "learning_rate": 4.219735829699924e-05, + "loss": 0.6847, + "step": 12219 + }, + { + "epoch": 0.47, + "learning_rate": 4.21967171069505e-05, + "loss": 0.5843, + "step": 12220 + }, + { + "epoch": 0.47, + "learning_rate": 4.2196075916901774e-05, + "loss": 0.4779, + "step": 12221 + }, + { + "epoch": 0.47, + "learning_rate": 4.219543472685304e-05, + "loss": 0.5331, + "step": 12222 + }, + { + "epoch": 0.47, + "learning_rate": 4.219479353680431e-05, + "loss": 0.5647, + "step": 12223 + }, + { + "epoch": 0.47, + "learning_rate": 4.2194152346755575e-05, + "loss": 0.5174, + "step": 12224 + }, + { + "epoch": 0.47, + "learning_rate": 4.219351115670685e-05, + "loss": 0.5578, + "step": 12225 + }, + { + "epoch": 0.47, + "learning_rate": 4.219286996665812e-05, + "loss": 0.5125, + "step": 12226 + }, + { + "epoch": 0.47, + "learning_rate": 4.219222877660939e-05, + "loss": 0.5468, + "step": 12227 + }, + { + "epoch": 0.47, + "learning_rate": 4.219158758656066e-05, + "loss": 0.5459, + "step": 12228 + }, + { + "epoch": 0.47, + "learning_rate": 4.219094639651193e-05, + "loss": 0.5439, + "step": 12229 + }, + { + "epoch": 0.47, + "learning_rate": 4.21903052064632e-05, + "loss": 0.5242, + "step": 12230 + }, + { + "epoch": 0.47, + "learning_rate": 4.2189664016414464e-05, + "loss": 0.5683, + "step": 12231 + }, + { + "epoch": 0.47, + "learning_rate": 4.2189022826365735e-05, + "loss": 0.5235, + "step": 12232 + }, + { + "epoch": 0.47, + "learning_rate": 4.218838163631701e-05, + "loss": 0.6344, + "step": 12233 + }, + { + "epoch": 0.47, + "learning_rate": 4.218774044626828e-05, + "loss": 0.5321, + "step": 12234 + }, + { + "epoch": 0.47, + "learning_rate": 4.218709925621955e-05, + "loss": 0.5508, + "step": 12235 + }, + { + "epoch": 0.47, + "learning_rate": 4.2186458066170815e-05, + "loss": 0.5834, + "step": 12236 + }, + { + "epoch": 0.47, + "learning_rate": 4.218581687612209e-05, + "loss": 0.5341, + "step": 12237 + }, + { + "epoch": 0.47, + "learning_rate": 4.218517568607335e-05, + "loss": 0.4933, + "step": 12238 + }, + { + "epoch": 0.47, + "learning_rate": 4.2184534496024624e-05, + "loss": 0.582, + "step": 12239 + }, + { + "epoch": 0.47, + "learning_rate": 4.218389330597589e-05, + "loss": 0.6801, + "step": 12240 + }, + { + "epoch": 0.47, + "learning_rate": 4.218325211592716e-05, + "loss": 0.5893, + "step": 12241 + }, + { + "epoch": 0.47, + "learning_rate": 4.218261092587843e-05, + "loss": 0.482, + "step": 12242 + }, + { + "epoch": 0.47, + "learning_rate": 4.2181969735829704e-05, + "loss": 0.5299, + "step": 12243 + }, + { + "epoch": 0.47, + "learning_rate": 4.2181328545780976e-05, + "loss": 0.6225, + "step": 12244 + }, + { + "epoch": 0.47, + "learning_rate": 4.218068735573224e-05, + "loss": 0.5973, + "step": 12245 + }, + { + "epoch": 0.47, + "learning_rate": 4.218004616568351e-05, + "loss": 0.5832, + "step": 12246 + }, + { + "epoch": 0.47, + "learning_rate": 4.217940497563478e-05, + "loss": 0.6499, + "step": 12247 + }, + { + "epoch": 0.47, + "learning_rate": 4.217876378558605e-05, + "loss": 0.5256, + "step": 12248 + }, + { + "epoch": 0.47, + "learning_rate": 4.2178122595537314e-05, + "loss": 0.5736, + "step": 12249 + }, + { + "epoch": 0.47, + "learning_rate": 4.217748140548859e-05, + "loss": 0.5879, + "step": 12250 + }, + { + "epoch": 0.47, + "learning_rate": 4.217684021543986e-05, + "loss": 0.5638, + "step": 12251 + }, + { + "epoch": 0.47, + "learning_rate": 4.217619902539113e-05, + "loss": 0.5488, + "step": 12252 + }, + { + "epoch": 0.47, + "learning_rate": 4.21755578353424e-05, + "loss": 0.6537, + "step": 12253 + }, + { + "epoch": 0.47, + "learning_rate": 4.2174916645293666e-05, + "loss": 0.6209, + "step": 12254 + }, + { + "epoch": 0.47, + "learning_rate": 4.217427545524494e-05, + "loss": 0.5392, + "step": 12255 + }, + { + "epoch": 0.47, + "learning_rate": 4.21736342651962e-05, + "loss": 0.585, + "step": 12256 + }, + { + "epoch": 0.47, + "learning_rate": 4.2172993075147474e-05, + "loss": 0.4906, + "step": 12257 + }, + { + "epoch": 0.47, + "learning_rate": 4.2172351885098746e-05, + "loss": 0.5955, + "step": 12258 + }, + { + "epoch": 0.47, + "learning_rate": 4.217171069505002e-05, + "loss": 0.575, + "step": 12259 + }, + { + "epoch": 0.47, + "learning_rate": 4.217106950500129e-05, + "loss": 0.5988, + "step": 12260 + }, + { + "epoch": 0.47, + "learning_rate": 4.2170428314952554e-05, + "loss": 0.4937, + "step": 12261 + }, + { + "epoch": 0.47, + "learning_rate": 4.2169787124903826e-05, + "loss": 0.4956, + "step": 12262 + }, + { + "epoch": 0.47, + "learning_rate": 4.216914593485509e-05, + "loss": 0.5, + "step": 12263 + }, + { + "epoch": 0.47, + "learning_rate": 4.216850474480636e-05, + "loss": 0.5117, + "step": 12264 + }, + { + "epoch": 0.47, + "learning_rate": 4.216786355475763e-05, + "loss": 0.4871, + "step": 12265 + }, + { + "epoch": 0.47, + "learning_rate": 4.2167222364708906e-05, + "loss": 0.5353, + "step": 12266 + }, + { + "epoch": 0.47, + "learning_rate": 4.216658117466017e-05, + "loss": 0.5091, + "step": 12267 + }, + { + "epoch": 0.47, + "learning_rate": 4.216593998461144e-05, + "loss": 0.6402, + "step": 12268 + }, + { + "epoch": 0.47, + "learning_rate": 4.2165298794562714e-05, + "loss": 0.6363, + "step": 12269 + }, + { + "epoch": 0.47, + "learning_rate": 4.216465760451398e-05, + "loss": 0.5462, + "step": 12270 + }, + { + "epoch": 0.47, + "learning_rate": 4.216401641446525e-05, + "loss": 0.6112, + "step": 12271 + }, + { + "epoch": 0.47, + "learning_rate": 4.2163375224416516e-05, + "loss": 0.5152, + "step": 12272 + }, + { + "epoch": 0.47, + "learning_rate": 4.216273403436779e-05, + "loss": 0.5278, + "step": 12273 + }, + { + "epoch": 0.47, + "learning_rate": 4.216209284431906e-05, + "loss": 0.5366, + "step": 12274 + }, + { + "epoch": 0.47, + "learning_rate": 4.216145165427033e-05, + "loss": 0.5947, + "step": 12275 + }, + { + "epoch": 0.47, + "learning_rate": 4.2160810464221596e-05, + "loss": 0.7104, + "step": 12276 + }, + { + "epoch": 0.47, + "learning_rate": 4.216016927417287e-05, + "loss": 0.6322, + "step": 12277 + }, + { + "epoch": 0.47, + "learning_rate": 4.215952808412414e-05, + "loss": 0.5937, + "step": 12278 + }, + { + "epoch": 0.47, + "learning_rate": 4.2158886894075404e-05, + "loss": 0.5521, + "step": 12279 + }, + { + "epoch": 0.47, + "learning_rate": 4.2158245704026676e-05, + "loss": 0.5783, + "step": 12280 + }, + { + "epoch": 0.47, + "learning_rate": 4.215760451397794e-05, + "loss": 0.564, + "step": 12281 + }, + { + "epoch": 0.47, + "learning_rate": 4.215696332392922e-05, + "loss": 0.485, + "step": 12282 + }, + { + "epoch": 0.47, + "learning_rate": 4.2156322133880484e-05, + "loss": 0.6022, + "step": 12283 + }, + { + "epoch": 0.47, + "learning_rate": 4.2155680943831756e-05, + "loss": 0.4773, + "step": 12284 + }, + { + "epoch": 0.47, + "learning_rate": 4.215503975378302e-05, + "loss": 0.5027, + "step": 12285 + }, + { + "epoch": 0.47, + "learning_rate": 4.215439856373429e-05, + "loss": 0.56, + "step": 12286 + }, + { + "epoch": 0.47, + "learning_rate": 4.2153757373685565e-05, + "loss": 0.5485, + "step": 12287 + }, + { + "epoch": 0.47, + "learning_rate": 4.215311618363683e-05, + "loss": 0.5243, + "step": 12288 + }, + { + "epoch": 0.47, + "learning_rate": 4.21524749935881e-05, + "loss": 0.5265, + "step": 12289 + }, + { + "epoch": 0.47, + "learning_rate": 4.215183380353937e-05, + "loss": 0.6111, + "step": 12290 + }, + { + "epoch": 0.47, + "learning_rate": 4.2151192613490645e-05, + "loss": 0.516, + "step": 12291 + }, + { + "epoch": 0.47, + "learning_rate": 4.215055142344191e-05, + "loss": 0.5087, + "step": 12292 + }, + { + "epoch": 0.47, + "learning_rate": 4.214991023339318e-05, + "loss": 0.5885, + "step": 12293 + }, + { + "epoch": 0.47, + "learning_rate": 4.2149269043344446e-05, + "loss": 0.569, + "step": 12294 + }, + { + "epoch": 0.47, + "learning_rate": 4.214862785329572e-05, + "loss": 0.5339, + "step": 12295 + }, + { + "epoch": 0.47, + "learning_rate": 4.214798666324699e-05, + "loss": 0.4566, + "step": 12296 + }, + { + "epoch": 0.47, + "learning_rate": 4.2147345473198255e-05, + "loss": 0.6076, + "step": 12297 + }, + { + "epoch": 0.47, + "learning_rate": 4.2146704283149526e-05, + "loss": 0.5199, + "step": 12298 + }, + { + "epoch": 0.47, + "learning_rate": 4.21460630931008e-05, + "loss": 0.609, + "step": 12299 + }, + { + "epoch": 0.47, + "learning_rate": 4.214542190305207e-05, + "loss": 0.5794, + "step": 12300 + }, + { + "epoch": 0.47, + "learning_rate": 4.2144780713003335e-05, + "loss": 0.4939, + "step": 12301 + }, + { + "epoch": 0.47, + "learning_rate": 4.2144139522954606e-05, + "loss": 0.5884, + "step": 12302 + }, + { + "epoch": 0.47, + "learning_rate": 4.214349833290587e-05, + "loss": 0.7111, + "step": 12303 + }, + { + "epoch": 0.47, + "learning_rate": 4.214285714285714e-05, + "loss": 0.6247, + "step": 12304 + }, + { + "epoch": 0.47, + "learning_rate": 4.2142215952808415e-05, + "loss": 0.5459, + "step": 12305 + }, + { + "epoch": 0.47, + "learning_rate": 4.2141574762759686e-05, + "loss": 0.4454, + "step": 12306 + }, + { + "epoch": 0.47, + "learning_rate": 4.214093357271096e-05, + "loss": 0.6927, + "step": 12307 + }, + { + "epoch": 0.47, + "learning_rate": 4.214029238266222e-05, + "loss": 0.5172, + "step": 12308 + }, + { + "epoch": 0.47, + "learning_rate": 4.2139651192613495e-05, + "loss": 0.5717, + "step": 12309 + }, + { + "epoch": 0.47, + "learning_rate": 4.213901000256476e-05, + "loss": 0.5442, + "step": 12310 + }, + { + "epoch": 0.47, + "learning_rate": 4.213836881251603e-05, + "loss": 0.5771, + "step": 12311 + }, + { + "epoch": 0.47, + "learning_rate": 4.2137727622467296e-05, + "loss": 0.5761, + "step": 12312 + }, + { + "epoch": 0.47, + "learning_rate": 4.213708643241857e-05, + "loss": 0.4679, + "step": 12313 + }, + { + "epoch": 0.47, + "learning_rate": 4.213644524236984e-05, + "loss": 0.5485, + "step": 12314 + }, + { + "epoch": 0.47, + "learning_rate": 4.213580405232111e-05, + "loss": 0.5058, + "step": 12315 + }, + { + "epoch": 0.47, + "learning_rate": 4.213516286227238e-05, + "loss": 0.4025, + "step": 12316 + }, + { + "epoch": 0.47, + "learning_rate": 4.213452167222365e-05, + "loss": 0.5833, + "step": 12317 + }, + { + "epoch": 0.47, + "learning_rate": 4.213388048217492e-05, + "loss": 0.6246, + "step": 12318 + }, + { + "epoch": 0.47, + "learning_rate": 4.2133239292126185e-05, + "loss": 0.5836, + "step": 12319 + }, + { + "epoch": 0.47, + "learning_rate": 4.213259810207746e-05, + "loss": 0.5812, + "step": 12320 + }, + { + "epoch": 0.47, + "learning_rate": 4.213195691202873e-05, + "loss": 0.491, + "step": 12321 + }, + { + "epoch": 0.47, + "learning_rate": 4.213131572198e-05, + "loss": 0.6128, + "step": 12322 + }, + { + "epoch": 0.47, + "learning_rate": 4.213067453193127e-05, + "loss": 0.5199, + "step": 12323 + }, + { + "epoch": 0.47, + "learning_rate": 4.213003334188254e-05, + "loss": 0.478, + "step": 12324 + }, + { + "epoch": 0.47, + "learning_rate": 4.212939215183381e-05, + "loss": 0.5038, + "step": 12325 + }, + { + "epoch": 0.47, + "learning_rate": 4.212875096178507e-05, + "loss": 0.4848, + "step": 12326 + }, + { + "epoch": 0.47, + "learning_rate": 4.2128109771736345e-05, + "loss": 0.425, + "step": 12327 + }, + { + "epoch": 0.47, + "learning_rate": 4.212746858168761e-05, + "loss": 0.619, + "step": 12328 + }, + { + "epoch": 0.47, + "learning_rate": 4.212682739163888e-05, + "loss": 0.5105, + "step": 12329 + }, + { + "epoch": 0.47, + "learning_rate": 4.2126186201590153e-05, + "loss": 0.6648, + "step": 12330 + }, + { + "epoch": 0.47, + "learning_rate": 4.2125545011541425e-05, + "loss": 0.5328, + "step": 12331 + }, + { + "epoch": 0.47, + "learning_rate": 4.21249038214927e-05, + "loss": 0.7082, + "step": 12332 + }, + { + "epoch": 0.47, + "learning_rate": 4.212426263144396e-05, + "loss": 0.5765, + "step": 12333 + }, + { + "epoch": 0.47, + "learning_rate": 4.2123621441395234e-05, + "loss": 0.6075, + "step": 12334 + }, + { + "epoch": 0.47, + "learning_rate": 4.21229802513465e-05, + "loss": 0.5132, + "step": 12335 + }, + { + "epoch": 0.47, + "learning_rate": 4.212233906129777e-05, + "loss": 0.5204, + "step": 12336 + }, + { + "epoch": 0.47, + "learning_rate": 4.2121697871249035e-05, + "loss": 0.4683, + "step": 12337 + }, + { + "epoch": 0.47, + "learning_rate": 4.2121056681200314e-05, + "loss": 0.662, + "step": 12338 + }, + { + "epoch": 0.47, + "learning_rate": 4.212041549115158e-05, + "loss": 0.5442, + "step": 12339 + }, + { + "epoch": 0.47, + "learning_rate": 4.211977430110285e-05, + "loss": 0.6129, + "step": 12340 + }, + { + "epoch": 0.47, + "learning_rate": 4.211913311105412e-05, + "loss": 0.5489, + "step": 12341 + }, + { + "epoch": 0.47, + "learning_rate": 4.211849192100539e-05, + "loss": 0.6382, + "step": 12342 + }, + { + "epoch": 0.47, + "learning_rate": 4.211785073095666e-05, + "loss": 0.6035, + "step": 12343 + }, + { + "epoch": 0.47, + "learning_rate": 4.2117209540907924e-05, + "loss": 0.5983, + "step": 12344 + }, + { + "epoch": 0.47, + "learning_rate": 4.2116568350859195e-05, + "loss": 0.5833, + "step": 12345 + }, + { + "epoch": 0.47, + "learning_rate": 4.211592716081047e-05, + "loss": 0.5742, + "step": 12346 + }, + { + "epoch": 0.47, + "learning_rate": 4.211528597076174e-05, + "loss": 0.5177, + "step": 12347 + }, + { + "epoch": 0.47, + "learning_rate": 4.2114644780713004e-05, + "loss": 0.5395, + "step": 12348 + }, + { + "epoch": 0.47, + "learning_rate": 4.2114003590664275e-05, + "loss": 0.563, + "step": 12349 + }, + { + "epoch": 0.47, + "learning_rate": 4.211336240061555e-05, + "loss": 0.6256, + "step": 12350 + }, + { + "epoch": 0.47, + "learning_rate": 4.211272121056681e-05, + "loss": 0.5626, + "step": 12351 + }, + { + "epoch": 0.47, + "learning_rate": 4.2112080020518084e-05, + "loss": 0.5268, + "step": 12352 + }, + { + "epoch": 0.47, + "learning_rate": 4.211143883046935e-05, + "loss": 0.5381, + "step": 12353 + }, + { + "epoch": 0.47, + "learning_rate": 4.211079764042063e-05, + "loss": 0.5455, + "step": 12354 + }, + { + "epoch": 0.48, + "learning_rate": 4.211015645037189e-05, + "loss": 0.5822, + "step": 12355 + }, + { + "epoch": 0.48, + "learning_rate": 4.2109515260323164e-05, + "loss": 0.5234, + "step": 12356 + }, + { + "epoch": 0.48, + "learning_rate": 4.210887407027443e-05, + "loss": 0.5251, + "step": 12357 + }, + { + "epoch": 0.48, + "learning_rate": 4.21082328802257e-05, + "loss": 0.4861, + "step": 12358 + }, + { + "epoch": 0.48, + "learning_rate": 4.210759169017697e-05, + "loss": 0.5322, + "step": 12359 + }, + { + "epoch": 0.48, + "learning_rate": 4.210695050012824e-05, + "loss": 0.5235, + "step": 12360 + }, + { + "epoch": 0.48, + "learning_rate": 4.210630931007951e-05, + "loss": 0.5409, + "step": 12361 + }, + { + "epoch": 0.48, + "learning_rate": 4.210566812003078e-05, + "loss": 0.5597, + "step": 12362 + }, + { + "epoch": 0.48, + "learning_rate": 4.210502692998205e-05, + "loss": 0.4881, + "step": 12363 + }, + { + "epoch": 0.48, + "learning_rate": 4.210438573993332e-05, + "loss": 0.5911, + "step": 12364 + }, + { + "epoch": 0.48, + "learning_rate": 4.210374454988459e-05, + "loss": 0.5624, + "step": 12365 + }, + { + "epoch": 0.48, + "learning_rate": 4.2103103359835854e-05, + "loss": 0.5823, + "step": 12366 + }, + { + "epoch": 0.48, + "learning_rate": 4.2102462169787126e-05, + "loss": 0.6175, + "step": 12367 + }, + { + "epoch": 0.48, + "learning_rate": 4.21018209797384e-05, + "loss": 0.4981, + "step": 12368 + }, + { + "epoch": 0.48, + "learning_rate": 4.210117978968966e-05, + "loss": 0.6046, + "step": 12369 + }, + { + "epoch": 0.48, + "learning_rate": 4.2100538599640934e-05, + "loss": 0.5718, + "step": 12370 + }, + { + "epoch": 0.48, + "learning_rate": 4.2099897409592206e-05, + "loss": 0.6384, + "step": 12371 + }, + { + "epoch": 0.48, + "learning_rate": 4.209925621954348e-05, + "loss": 0.5032, + "step": 12372 + }, + { + "epoch": 0.48, + "learning_rate": 4.209861502949474e-05, + "loss": 0.5351, + "step": 12373 + }, + { + "epoch": 0.48, + "learning_rate": 4.2097973839446014e-05, + "loss": 0.5679, + "step": 12374 + }, + { + "epoch": 0.48, + "learning_rate": 4.209733264939728e-05, + "loss": 0.5039, + "step": 12375 + }, + { + "epoch": 0.48, + "learning_rate": 4.209669145934855e-05, + "loss": 0.5633, + "step": 12376 + }, + { + "epoch": 0.48, + "learning_rate": 4.209605026929982e-05, + "loss": 0.5156, + "step": 12377 + }, + { + "epoch": 0.48, + "learning_rate": 4.2095409079251094e-05, + "loss": 0.4492, + "step": 12378 + }, + { + "epoch": 0.48, + "learning_rate": 4.2094767889202366e-05, + "loss": 0.598, + "step": 12379 + }, + { + "epoch": 0.48, + "learning_rate": 4.209412669915363e-05, + "loss": 0.5956, + "step": 12380 + }, + { + "epoch": 0.48, + "learning_rate": 4.20934855091049e-05, + "loss": 0.5236, + "step": 12381 + }, + { + "epoch": 0.48, + "learning_rate": 4.209284431905617e-05, + "loss": 0.5805, + "step": 12382 + }, + { + "epoch": 0.48, + "learning_rate": 4.209220312900744e-05, + "loss": 0.6414, + "step": 12383 + }, + { + "epoch": 0.48, + "learning_rate": 4.2091561938958704e-05, + "loss": 0.5263, + "step": 12384 + }, + { + "epoch": 0.48, + "learning_rate": 4.2090920748909976e-05, + "loss": 0.5527, + "step": 12385 + }, + { + "epoch": 0.48, + "learning_rate": 4.209027955886125e-05, + "loss": 0.53, + "step": 12386 + }, + { + "epoch": 0.48, + "learning_rate": 4.208963836881252e-05, + "loss": 0.5023, + "step": 12387 + }, + { + "epoch": 0.48, + "learning_rate": 4.208899717876379e-05, + "loss": 0.6101, + "step": 12388 + }, + { + "epoch": 0.48, + "learning_rate": 4.2088355988715056e-05, + "loss": 0.6111, + "step": 12389 + }, + { + "epoch": 0.48, + "learning_rate": 4.208771479866633e-05, + "loss": 0.6324, + "step": 12390 + }, + { + "epoch": 0.48, + "learning_rate": 4.208707360861759e-05, + "loss": 0.6266, + "step": 12391 + }, + { + "epoch": 0.48, + "learning_rate": 4.2086432418568864e-05, + "loss": 0.6428, + "step": 12392 + }, + { + "epoch": 0.48, + "learning_rate": 4.2085791228520136e-05, + "loss": 0.5513, + "step": 12393 + }, + { + "epoch": 0.48, + "learning_rate": 4.208515003847141e-05, + "loss": 0.5898, + "step": 12394 + }, + { + "epoch": 0.48, + "learning_rate": 4.208450884842268e-05, + "loss": 0.4978, + "step": 12395 + }, + { + "epoch": 0.48, + "learning_rate": 4.2083867658373944e-05, + "loss": 0.5454, + "step": 12396 + }, + { + "epoch": 0.48, + "learning_rate": 4.2083226468325216e-05, + "loss": 0.4975, + "step": 12397 + }, + { + "epoch": 0.48, + "learning_rate": 4.208258527827648e-05, + "loss": 0.627, + "step": 12398 + }, + { + "epoch": 0.48, + "learning_rate": 4.208194408822775e-05, + "loss": 0.4933, + "step": 12399 + }, + { + "epoch": 0.48, + "learning_rate": 4.208130289817902e-05, + "loss": 0.5291, + "step": 12400 + }, + { + "epoch": 0.48, + "learning_rate": 4.208066170813029e-05, + "loss": 0.5966, + "step": 12401 + }, + { + "epoch": 0.48, + "learning_rate": 4.208002051808156e-05, + "loss": 0.4404, + "step": 12402 + }, + { + "epoch": 0.48, + "learning_rate": 4.207937932803283e-05, + "loss": 0.5217, + "step": 12403 + }, + { + "epoch": 0.48, + "learning_rate": 4.2078738137984105e-05, + "loss": 0.5364, + "step": 12404 + }, + { + "epoch": 0.48, + "learning_rate": 4.207809694793537e-05, + "loss": 0.5084, + "step": 12405 + }, + { + "epoch": 0.48, + "learning_rate": 4.207745575788664e-05, + "loss": 0.5881, + "step": 12406 + }, + { + "epoch": 0.48, + "learning_rate": 4.2076814567837906e-05, + "loss": 0.5691, + "step": 12407 + }, + { + "epoch": 0.48, + "learning_rate": 4.207617337778918e-05, + "loss": 0.5293, + "step": 12408 + }, + { + "epoch": 0.48, + "learning_rate": 4.207553218774044e-05, + "loss": 0.5767, + "step": 12409 + }, + { + "epoch": 0.48, + "learning_rate": 4.207489099769172e-05, + "loss": 0.5535, + "step": 12410 + }, + { + "epoch": 0.48, + "learning_rate": 4.2074249807642986e-05, + "loss": 0.5768, + "step": 12411 + }, + { + "epoch": 0.48, + "learning_rate": 4.207360861759426e-05, + "loss": 0.5462, + "step": 12412 + }, + { + "epoch": 0.48, + "learning_rate": 4.207296742754553e-05, + "loss": 0.4989, + "step": 12413 + }, + { + "epoch": 0.48, + "learning_rate": 4.2072326237496795e-05, + "loss": 0.4899, + "step": 12414 + }, + { + "epoch": 0.48, + "learning_rate": 4.2071685047448066e-05, + "loss": 0.4617, + "step": 12415 + }, + { + "epoch": 0.48, + "learning_rate": 4.207104385739933e-05, + "loss": 0.4662, + "step": 12416 + }, + { + "epoch": 0.48, + "learning_rate": 4.20704026673506e-05, + "loss": 0.5623, + "step": 12417 + }, + { + "epoch": 0.48, + "learning_rate": 4.2069761477301875e-05, + "loss": 0.4798, + "step": 12418 + }, + { + "epoch": 0.48, + "learning_rate": 4.2069120287253147e-05, + "loss": 0.5607, + "step": 12419 + }, + { + "epoch": 0.48, + "learning_rate": 4.206847909720441e-05, + "loss": 0.5453, + "step": 12420 + }, + { + "epoch": 0.48, + "learning_rate": 4.206783790715568e-05, + "loss": 0.4799, + "step": 12421 + }, + { + "epoch": 0.48, + "learning_rate": 4.2067196717106955e-05, + "loss": 0.537, + "step": 12422 + }, + { + "epoch": 0.48, + "learning_rate": 4.206655552705822e-05, + "loss": 0.5861, + "step": 12423 + }, + { + "epoch": 0.48, + "learning_rate": 4.206591433700949e-05, + "loss": 0.5242, + "step": 12424 + }, + { + "epoch": 0.48, + "learning_rate": 4.2065273146960756e-05, + "loss": 0.479, + "step": 12425 + }, + { + "epoch": 0.48, + "learning_rate": 4.2064631956912035e-05, + "loss": 0.698, + "step": 12426 + }, + { + "epoch": 0.48, + "learning_rate": 4.20639907668633e-05, + "loss": 0.5487, + "step": 12427 + }, + { + "epoch": 0.48, + "learning_rate": 4.206334957681457e-05, + "loss": 0.5773, + "step": 12428 + }, + { + "epoch": 0.48, + "learning_rate": 4.2062708386765837e-05, + "loss": 0.5434, + "step": 12429 + }, + { + "epoch": 0.48, + "learning_rate": 4.206206719671711e-05, + "loss": 0.5888, + "step": 12430 + }, + { + "epoch": 0.48, + "learning_rate": 4.206142600666838e-05, + "loss": 0.4912, + "step": 12431 + }, + { + "epoch": 0.48, + "learning_rate": 4.2060784816619645e-05, + "loss": 0.604, + "step": 12432 + }, + { + "epoch": 0.48, + "learning_rate": 4.206014362657092e-05, + "loss": 0.5345, + "step": 12433 + }, + { + "epoch": 0.48, + "learning_rate": 4.205950243652219e-05, + "loss": 0.5597, + "step": 12434 + }, + { + "epoch": 0.48, + "learning_rate": 4.205886124647346e-05, + "loss": 0.5792, + "step": 12435 + }, + { + "epoch": 0.48, + "learning_rate": 4.2058220056424725e-05, + "loss": 0.5995, + "step": 12436 + }, + { + "epoch": 0.48, + "learning_rate": 4.2057578866376e-05, + "loss": 0.6262, + "step": 12437 + }, + { + "epoch": 0.48, + "learning_rate": 4.205693767632726e-05, + "loss": 0.5436, + "step": 12438 + }, + { + "epoch": 0.48, + "learning_rate": 4.2056296486278533e-05, + "loss": 0.6337, + "step": 12439 + }, + { + "epoch": 0.48, + "learning_rate": 4.2055655296229805e-05, + "loss": 0.5081, + "step": 12440 + }, + { + "epoch": 0.48, + "learning_rate": 4.205501410618107e-05, + "loss": 0.4923, + "step": 12441 + }, + { + "epoch": 0.48, + "learning_rate": 4.205437291613235e-05, + "loss": 0.5526, + "step": 12442 + }, + { + "epoch": 0.48, + "learning_rate": 4.2053731726083613e-05, + "loss": 0.5757, + "step": 12443 + }, + { + "epoch": 0.48, + "learning_rate": 4.2053090536034885e-05, + "loss": 0.5214, + "step": 12444 + }, + { + "epoch": 0.48, + "learning_rate": 4.205244934598615e-05, + "loss": 0.7004, + "step": 12445 + }, + { + "epoch": 0.48, + "learning_rate": 4.205180815593742e-05, + "loss": 0.648, + "step": 12446 + }, + { + "epoch": 0.48, + "learning_rate": 4.205116696588869e-05, + "loss": 0.5758, + "step": 12447 + }, + { + "epoch": 0.48, + "learning_rate": 4.205052577583996e-05, + "loss": 0.443, + "step": 12448 + }, + { + "epoch": 0.48, + "learning_rate": 4.204988458579123e-05, + "loss": 0.4642, + "step": 12449 + }, + { + "epoch": 0.48, + "learning_rate": 4.20492433957425e-05, + "loss": 0.6328, + "step": 12450 + }, + { + "epoch": 0.48, + "learning_rate": 4.2048602205693774e-05, + "loss": 0.5039, + "step": 12451 + }, + { + "epoch": 0.48, + "learning_rate": 4.204796101564504e-05, + "loss": 0.4763, + "step": 12452 + }, + { + "epoch": 0.48, + "learning_rate": 4.204731982559631e-05, + "loss": 0.5959, + "step": 12453 + }, + { + "epoch": 0.48, + "learning_rate": 4.2046678635547575e-05, + "loss": 0.571, + "step": 12454 + }, + { + "epoch": 0.48, + "learning_rate": 4.204603744549885e-05, + "loss": 0.5966, + "step": 12455 + }, + { + "epoch": 0.48, + "learning_rate": 4.204539625545012e-05, + "loss": 0.5418, + "step": 12456 + }, + { + "epoch": 0.48, + "learning_rate": 4.2044755065401384e-05, + "loss": 0.5661, + "step": 12457 + }, + { + "epoch": 0.48, + "learning_rate": 4.2044113875352655e-05, + "loss": 0.474, + "step": 12458 + }, + { + "epoch": 0.48, + "learning_rate": 4.204347268530393e-05, + "loss": 0.5644, + "step": 12459 + }, + { + "epoch": 0.48, + "learning_rate": 4.20428314952552e-05, + "loss": 0.5976, + "step": 12460 + }, + { + "epoch": 0.48, + "learning_rate": 4.2042190305206464e-05, + "loss": 0.5401, + "step": 12461 + }, + { + "epoch": 0.48, + "learning_rate": 4.2041549115157735e-05, + "loss": 0.5042, + "step": 12462 + }, + { + "epoch": 0.48, + "learning_rate": 4.2040907925109e-05, + "loss": 0.6832, + "step": 12463 + }, + { + "epoch": 0.48, + "learning_rate": 4.204026673506027e-05, + "loss": 0.5869, + "step": 12464 + }, + { + "epoch": 0.48, + "learning_rate": 4.2039625545011544e-05, + "loss": 0.5615, + "step": 12465 + }, + { + "epoch": 0.48, + "learning_rate": 4.2038984354962816e-05, + "loss": 0.6085, + "step": 12466 + }, + { + "epoch": 0.48, + "learning_rate": 4.203834316491409e-05, + "loss": 0.6825, + "step": 12467 + }, + { + "epoch": 0.48, + "learning_rate": 4.203770197486535e-05, + "loss": 0.5656, + "step": 12468 + }, + { + "epoch": 0.48, + "learning_rate": 4.2037060784816624e-05, + "loss": 0.4337, + "step": 12469 + }, + { + "epoch": 0.48, + "learning_rate": 4.203641959476789e-05, + "loss": 0.5492, + "step": 12470 + }, + { + "epoch": 0.48, + "learning_rate": 4.203577840471916e-05, + "loss": 0.5326, + "step": 12471 + }, + { + "epoch": 0.48, + "learning_rate": 4.2035137214670426e-05, + "loss": 0.5117, + "step": 12472 + }, + { + "epoch": 0.48, + "learning_rate": 4.20344960246217e-05, + "loss": 0.5699, + "step": 12473 + }, + { + "epoch": 0.48, + "learning_rate": 4.203385483457297e-05, + "loss": 0.573, + "step": 12474 + }, + { + "epoch": 0.48, + "learning_rate": 4.203321364452424e-05, + "loss": 0.633, + "step": 12475 + }, + { + "epoch": 0.48, + "learning_rate": 4.203257245447551e-05, + "loss": 0.56, + "step": 12476 + }, + { + "epoch": 0.48, + "learning_rate": 4.203193126442678e-05, + "loss": 0.5308, + "step": 12477 + }, + { + "epoch": 0.48, + "learning_rate": 4.203129007437805e-05, + "loss": 0.5395, + "step": 12478 + }, + { + "epoch": 0.48, + "learning_rate": 4.2030648884329314e-05, + "loss": 0.4798, + "step": 12479 + }, + { + "epoch": 0.48, + "learning_rate": 4.2030007694280586e-05, + "loss": 0.501, + "step": 12480 + }, + { + "epoch": 0.48, + "learning_rate": 4.202936650423186e-05, + "loss": 0.5759, + "step": 12481 + }, + { + "epoch": 0.48, + "learning_rate": 4.202872531418313e-05, + "loss": 0.613, + "step": 12482 + }, + { + "epoch": 0.48, + "learning_rate": 4.20280841241344e-05, + "loss": 0.4861, + "step": 12483 + }, + { + "epoch": 0.48, + "learning_rate": 4.2027442934085666e-05, + "loss": 0.5782, + "step": 12484 + }, + { + "epoch": 0.48, + "learning_rate": 4.202680174403694e-05, + "loss": 0.4559, + "step": 12485 + }, + { + "epoch": 0.48, + "learning_rate": 4.20261605539882e-05, + "loss": 0.5213, + "step": 12486 + }, + { + "epoch": 0.48, + "learning_rate": 4.2025519363939474e-05, + "loss": 0.4805, + "step": 12487 + }, + { + "epoch": 0.48, + "learning_rate": 4.202487817389074e-05, + "loss": 0.5608, + "step": 12488 + }, + { + "epoch": 0.48, + "learning_rate": 4.202423698384201e-05, + "loss": 0.5826, + "step": 12489 + }, + { + "epoch": 0.48, + "learning_rate": 4.202359579379328e-05, + "loss": 0.5008, + "step": 12490 + }, + { + "epoch": 0.48, + "learning_rate": 4.2022954603744554e-05, + "loss": 0.556, + "step": 12491 + }, + { + "epoch": 0.48, + "learning_rate": 4.2022313413695826e-05, + "loss": 0.57, + "step": 12492 + }, + { + "epoch": 0.48, + "learning_rate": 4.202167222364709e-05, + "loss": 0.5534, + "step": 12493 + }, + { + "epoch": 0.48, + "learning_rate": 4.202103103359836e-05, + "loss": 0.5438, + "step": 12494 + }, + { + "epoch": 0.48, + "learning_rate": 4.202038984354963e-05, + "loss": 0.556, + "step": 12495 + }, + { + "epoch": 0.48, + "learning_rate": 4.20197486535009e-05, + "loss": 0.6346, + "step": 12496 + }, + { + "epoch": 0.48, + "learning_rate": 4.2019107463452164e-05, + "loss": 0.4868, + "step": 12497 + }, + { + "epoch": 0.48, + "learning_rate": 4.201846627340344e-05, + "loss": 0.5643, + "step": 12498 + }, + { + "epoch": 0.48, + "learning_rate": 4.201782508335471e-05, + "loss": 0.4755, + "step": 12499 + }, + { + "epoch": 0.48, + "learning_rate": 4.201718389330598e-05, + "loss": 0.5911, + "step": 12500 + }, + { + "epoch": 0.48, + "learning_rate": 4.201654270325725e-05, + "loss": 0.6122, + "step": 12501 + }, + { + "epoch": 0.48, + "learning_rate": 4.2015901513208516e-05, + "loss": 0.5051, + "step": 12502 + }, + { + "epoch": 0.48, + "learning_rate": 4.201526032315979e-05, + "loss": 0.6022, + "step": 12503 + }, + { + "epoch": 0.48, + "learning_rate": 4.201461913311105e-05, + "loss": 0.5101, + "step": 12504 + }, + { + "epoch": 0.48, + "learning_rate": 4.2013977943062324e-05, + "loss": 0.51, + "step": 12505 + }, + { + "epoch": 0.48, + "learning_rate": 4.2013336753013596e-05, + "loss": 0.4906, + "step": 12506 + }, + { + "epoch": 0.48, + "learning_rate": 4.201269556296487e-05, + "loss": 0.423, + "step": 12507 + }, + { + "epoch": 0.48, + "learning_rate": 4.201205437291613e-05, + "loss": 0.5407, + "step": 12508 + }, + { + "epoch": 0.48, + "learning_rate": 4.2011413182867404e-05, + "loss": 0.6137, + "step": 12509 + }, + { + "epoch": 0.48, + "learning_rate": 4.2010771992818676e-05, + "loss": 0.5301, + "step": 12510 + }, + { + "epoch": 0.48, + "learning_rate": 4.201013080276994e-05, + "loss": 0.4944, + "step": 12511 + }, + { + "epoch": 0.48, + "learning_rate": 4.200948961272121e-05, + "loss": 0.5227, + "step": 12512 + }, + { + "epoch": 0.48, + "learning_rate": 4.200884842267248e-05, + "loss": 0.7119, + "step": 12513 + }, + { + "epoch": 0.48, + "learning_rate": 4.2008207232623756e-05, + "loss": 0.4835, + "step": 12514 + }, + { + "epoch": 0.48, + "learning_rate": 4.200756604257502e-05, + "loss": 0.6287, + "step": 12515 + }, + { + "epoch": 0.48, + "learning_rate": 4.200692485252629e-05, + "loss": 0.6933, + "step": 12516 + }, + { + "epoch": 0.48, + "learning_rate": 4.200628366247756e-05, + "loss": 0.5561, + "step": 12517 + }, + { + "epoch": 0.48, + "learning_rate": 4.200564247242883e-05, + "loss": 0.5729, + "step": 12518 + }, + { + "epoch": 0.48, + "learning_rate": 4.20050012823801e-05, + "loss": 0.5111, + "step": 12519 + }, + { + "epoch": 0.48, + "learning_rate": 4.2004360092331366e-05, + "loss": 0.4003, + "step": 12520 + }, + { + "epoch": 0.48, + "learning_rate": 4.200371890228264e-05, + "loss": 0.4293, + "step": 12521 + }, + { + "epoch": 0.48, + "learning_rate": 4.200307771223391e-05, + "loss": 0.6479, + "step": 12522 + }, + { + "epoch": 0.48, + "learning_rate": 4.200243652218518e-05, + "loss": 0.4819, + "step": 12523 + }, + { + "epoch": 0.48, + "learning_rate": 4.2001795332136446e-05, + "loss": 0.6011, + "step": 12524 + }, + { + "epoch": 0.48, + "learning_rate": 4.200115414208772e-05, + "loss": 0.4285, + "step": 12525 + }, + { + "epoch": 0.48, + "learning_rate": 4.200051295203898e-05, + "loss": 0.4767, + "step": 12526 + }, + { + "epoch": 0.48, + "learning_rate": 4.1999871761990255e-05, + "loss": 0.4648, + "step": 12527 + }, + { + "epoch": 0.48, + "learning_rate": 4.1999230571941526e-05, + "loss": 0.6184, + "step": 12528 + }, + { + "epoch": 0.48, + "learning_rate": 4.199858938189279e-05, + "loss": 0.5733, + "step": 12529 + }, + { + "epoch": 0.48, + "learning_rate": 4.199794819184406e-05, + "loss": 0.5873, + "step": 12530 + }, + { + "epoch": 0.48, + "learning_rate": 4.1997307001795335e-05, + "loss": 0.5154, + "step": 12531 + }, + { + "epoch": 0.48, + "learning_rate": 4.1996665811746607e-05, + "loss": 0.5726, + "step": 12532 + }, + { + "epoch": 0.48, + "learning_rate": 4.199602462169787e-05, + "loss": 0.5555, + "step": 12533 + }, + { + "epoch": 0.48, + "learning_rate": 4.199538343164914e-05, + "loss": 0.4518, + "step": 12534 + }, + { + "epoch": 0.48, + "learning_rate": 4.199474224160041e-05, + "loss": 0.5006, + "step": 12535 + }, + { + "epoch": 0.48, + "learning_rate": 4.199410105155168e-05, + "loss": 0.5557, + "step": 12536 + }, + { + "epoch": 0.48, + "learning_rate": 4.199345986150295e-05, + "loss": 0.5699, + "step": 12537 + }, + { + "epoch": 0.48, + "learning_rate": 4.199281867145422e-05, + "loss": 0.5042, + "step": 12538 + }, + { + "epoch": 0.48, + "learning_rate": 4.1992177481405495e-05, + "loss": 0.5144, + "step": 12539 + }, + { + "epoch": 0.48, + "learning_rate": 4.199153629135676e-05, + "loss": 0.5678, + "step": 12540 + }, + { + "epoch": 0.48, + "learning_rate": 4.199089510130803e-05, + "loss": 0.4623, + "step": 12541 + }, + { + "epoch": 0.48, + "learning_rate": 4.1990253911259297e-05, + "loss": 0.5198, + "step": 12542 + }, + { + "epoch": 0.48, + "learning_rate": 4.198961272121057e-05, + "loss": 0.578, + "step": 12543 + }, + { + "epoch": 0.48, + "learning_rate": 4.198897153116183e-05, + "loss": 0.5196, + "step": 12544 + }, + { + "epoch": 0.48, + "learning_rate": 4.1988330341113105e-05, + "loss": 0.5772, + "step": 12545 + }, + { + "epoch": 0.48, + "learning_rate": 4.198768915106438e-05, + "loss": 0.5527, + "step": 12546 + }, + { + "epoch": 0.48, + "learning_rate": 4.198704796101565e-05, + "loss": 0.4631, + "step": 12547 + }, + { + "epoch": 0.48, + "learning_rate": 4.198640677096692e-05, + "loss": 0.6901, + "step": 12548 + }, + { + "epoch": 0.48, + "learning_rate": 4.1985765580918185e-05, + "loss": 0.4901, + "step": 12549 + }, + { + "epoch": 0.48, + "learning_rate": 4.198512439086946e-05, + "loss": 0.6017, + "step": 12550 + }, + { + "epoch": 0.48, + "learning_rate": 4.198448320082072e-05, + "loss": 0.6677, + "step": 12551 + }, + { + "epoch": 0.48, + "learning_rate": 4.1983842010771993e-05, + "loss": 0.6808, + "step": 12552 + }, + { + "epoch": 0.48, + "learning_rate": 4.1983200820723265e-05, + "loss": 0.545, + "step": 12553 + }, + { + "epoch": 0.48, + "learning_rate": 4.198255963067454e-05, + "loss": 0.5778, + "step": 12554 + }, + { + "epoch": 0.48, + "learning_rate": 4.198191844062581e-05, + "loss": 0.5561, + "step": 12555 + }, + { + "epoch": 0.48, + "learning_rate": 4.1981277250577074e-05, + "loss": 0.6532, + "step": 12556 + }, + { + "epoch": 0.48, + "learning_rate": 4.1980636060528345e-05, + "loss": 0.5537, + "step": 12557 + }, + { + "epoch": 0.48, + "learning_rate": 4.197999487047961e-05, + "loss": 0.577, + "step": 12558 + }, + { + "epoch": 0.48, + "learning_rate": 4.197935368043088e-05, + "loss": 0.5563, + "step": 12559 + }, + { + "epoch": 0.48, + "learning_rate": 4.197871249038215e-05, + "loss": 0.5337, + "step": 12560 + }, + { + "epoch": 0.48, + "learning_rate": 4.197807130033342e-05, + "loss": 0.5664, + "step": 12561 + }, + { + "epoch": 0.48, + "learning_rate": 4.197743011028469e-05, + "loss": 0.617, + "step": 12562 + }, + { + "epoch": 0.48, + "learning_rate": 4.197678892023596e-05, + "loss": 0.5628, + "step": 12563 + }, + { + "epoch": 0.48, + "learning_rate": 4.1976147730187234e-05, + "loss": 0.6059, + "step": 12564 + }, + { + "epoch": 0.48, + "learning_rate": 4.19755065401385e-05, + "loss": 0.5945, + "step": 12565 + }, + { + "epoch": 0.48, + "learning_rate": 4.197486535008977e-05, + "loss": 0.678, + "step": 12566 + }, + { + "epoch": 0.48, + "learning_rate": 4.1974224160041035e-05, + "loss": 0.6038, + "step": 12567 + }, + { + "epoch": 0.48, + "learning_rate": 4.197358296999231e-05, + "loss": 0.5828, + "step": 12568 + }, + { + "epoch": 0.48, + "learning_rate": 4.197294177994358e-05, + "loss": 0.621, + "step": 12569 + }, + { + "epoch": 0.48, + "learning_rate": 4.197230058989485e-05, + "loss": 0.5974, + "step": 12570 + }, + { + "epoch": 0.48, + "learning_rate": 4.1971659399846115e-05, + "loss": 0.4678, + "step": 12571 + }, + { + "epoch": 0.48, + "learning_rate": 4.197101820979739e-05, + "loss": 0.5057, + "step": 12572 + }, + { + "epoch": 0.48, + "learning_rate": 4.197037701974866e-05, + "loss": 0.5845, + "step": 12573 + }, + { + "epoch": 0.48, + "learning_rate": 4.1969735829699924e-05, + "loss": 0.6281, + "step": 12574 + }, + { + "epoch": 0.48, + "learning_rate": 4.1969094639651195e-05, + "loss": 0.5538, + "step": 12575 + }, + { + "epoch": 0.48, + "learning_rate": 4.196845344960246e-05, + "loss": 0.5802, + "step": 12576 + }, + { + "epoch": 0.48, + "learning_rate": 4.196781225955373e-05, + "loss": 0.5324, + "step": 12577 + }, + { + "epoch": 0.48, + "learning_rate": 4.1967171069505004e-05, + "loss": 0.609, + "step": 12578 + }, + { + "epoch": 0.48, + "learning_rate": 4.1966529879456276e-05, + "loss": 0.598, + "step": 12579 + }, + { + "epoch": 0.48, + "learning_rate": 4.196588868940754e-05, + "loss": 0.6565, + "step": 12580 + }, + { + "epoch": 0.48, + "learning_rate": 4.196524749935881e-05, + "loss": 0.519, + "step": 12581 + }, + { + "epoch": 0.48, + "learning_rate": 4.1964606309310084e-05, + "loss": 0.5275, + "step": 12582 + }, + { + "epoch": 0.48, + "learning_rate": 4.196396511926135e-05, + "loss": 0.5645, + "step": 12583 + }, + { + "epoch": 0.48, + "learning_rate": 4.196332392921262e-05, + "loss": 0.4953, + "step": 12584 + }, + { + "epoch": 0.48, + "learning_rate": 4.1962682739163886e-05, + "loss": 0.5902, + "step": 12585 + }, + { + "epoch": 0.48, + "learning_rate": 4.1962041549115164e-05, + "loss": 0.5894, + "step": 12586 + }, + { + "epoch": 0.48, + "learning_rate": 4.196140035906643e-05, + "loss": 0.5109, + "step": 12587 + }, + { + "epoch": 0.48, + "learning_rate": 4.19607591690177e-05, + "loss": 0.6071, + "step": 12588 + }, + { + "epoch": 0.48, + "learning_rate": 4.1960117978968966e-05, + "loss": 0.5331, + "step": 12589 + }, + { + "epoch": 0.48, + "learning_rate": 4.195947678892024e-05, + "loss": 0.4763, + "step": 12590 + }, + { + "epoch": 0.48, + "learning_rate": 4.195883559887151e-05, + "loss": 0.6596, + "step": 12591 + }, + { + "epoch": 0.48, + "learning_rate": 4.1958194408822774e-05, + "loss": 0.5173, + "step": 12592 + }, + { + "epoch": 0.48, + "learning_rate": 4.1957553218774046e-05, + "loss": 0.5015, + "step": 12593 + }, + { + "epoch": 0.48, + "learning_rate": 4.195691202872532e-05, + "loss": 0.5453, + "step": 12594 + }, + { + "epoch": 0.48, + "learning_rate": 4.195627083867659e-05, + "loss": 0.539, + "step": 12595 + }, + { + "epoch": 0.48, + "learning_rate": 4.1955629648627854e-05, + "loss": 0.5369, + "step": 12596 + }, + { + "epoch": 0.48, + "learning_rate": 4.1954988458579126e-05, + "loss": 0.6236, + "step": 12597 + }, + { + "epoch": 0.48, + "learning_rate": 4.195434726853039e-05, + "loss": 0.5233, + "step": 12598 + }, + { + "epoch": 0.48, + "learning_rate": 4.195370607848166e-05, + "loss": 0.4461, + "step": 12599 + }, + { + "epoch": 0.48, + "learning_rate": 4.1953064888432934e-05, + "loss": 0.5283, + "step": 12600 + }, + { + "epoch": 0.48, + "learning_rate": 4.19524236983842e-05, + "loss": 0.5316, + "step": 12601 + }, + { + "epoch": 0.48, + "learning_rate": 4.195178250833548e-05, + "loss": 0.6468, + "step": 12602 + }, + { + "epoch": 0.48, + "learning_rate": 4.195114131828674e-05, + "loss": 0.5617, + "step": 12603 + }, + { + "epoch": 0.48, + "learning_rate": 4.1950500128238014e-05, + "loss": 0.5049, + "step": 12604 + }, + { + "epoch": 0.48, + "learning_rate": 4.194985893818928e-05, + "loss": 0.6082, + "step": 12605 + }, + { + "epoch": 0.48, + "learning_rate": 4.194921774814055e-05, + "loss": 0.4425, + "step": 12606 + }, + { + "epoch": 0.48, + "learning_rate": 4.1948576558091816e-05, + "loss": 0.4867, + "step": 12607 + }, + { + "epoch": 0.48, + "learning_rate": 4.194793536804309e-05, + "loss": 0.6075, + "step": 12608 + }, + { + "epoch": 0.48, + "learning_rate": 4.194729417799436e-05, + "loss": 0.5384, + "step": 12609 + }, + { + "epoch": 0.48, + "learning_rate": 4.194665298794563e-05, + "loss": 0.6104, + "step": 12610 + }, + { + "epoch": 0.48, + "learning_rate": 4.19460117978969e-05, + "loss": 0.4917, + "step": 12611 + }, + { + "epoch": 0.48, + "learning_rate": 4.194537060784817e-05, + "loss": 0.563, + "step": 12612 + }, + { + "epoch": 0.48, + "learning_rate": 4.194472941779944e-05, + "loss": 0.4954, + "step": 12613 + }, + { + "epoch": 0.48, + "learning_rate": 4.1944088227750704e-05, + "loss": 0.5153, + "step": 12614 + }, + { + "epoch": 0.49, + "learning_rate": 4.1943447037701976e-05, + "loss": 0.4423, + "step": 12615 + }, + { + "epoch": 0.49, + "learning_rate": 4.194280584765324e-05, + "loss": 0.5892, + "step": 12616 + }, + { + "epoch": 0.49, + "learning_rate": 4.194216465760451e-05, + "loss": 0.5486, + "step": 12617 + }, + { + "epoch": 0.49, + "learning_rate": 4.1941523467555784e-05, + "loss": 0.4446, + "step": 12618 + }, + { + "epoch": 0.49, + "learning_rate": 4.1940882277507056e-05, + "loss": 0.4511, + "step": 12619 + }, + { + "epoch": 0.49, + "learning_rate": 4.194024108745833e-05, + "loss": 0.5866, + "step": 12620 + }, + { + "epoch": 0.49, + "learning_rate": 4.193959989740959e-05, + "loss": 0.5393, + "step": 12621 + }, + { + "epoch": 0.49, + "learning_rate": 4.1938958707360865e-05, + "loss": 0.4939, + "step": 12622 + }, + { + "epoch": 0.49, + "learning_rate": 4.193831751731213e-05, + "loss": 0.5167, + "step": 12623 + }, + { + "epoch": 0.49, + "learning_rate": 4.19376763272634e-05, + "loss": 0.5715, + "step": 12624 + }, + { + "epoch": 0.49, + "learning_rate": 4.193703513721467e-05, + "loss": 0.4915, + "step": 12625 + }, + { + "epoch": 0.49, + "learning_rate": 4.1936393947165945e-05, + "loss": 0.6603, + "step": 12626 + }, + { + "epoch": 0.49, + "learning_rate": 4.1935752757117216e-05, + "loss": 0.5756, + "step": 12627 + }, + { + "epoch": 0.49, + "learning_rate": 4.193511156706848e-05, + "loss": 0.5537, + "step": 12628 + }, + { + "epoch": 0.49, + "learning_rate": 4.193447037701975e-05, + "loss": 0.4528, + "step": 12629 + }, + { + "epoch": 0.49, + "learning_rate": 4.193382918697102e-05, + "loss": 0.4852, + "step": 12630 + }, + { + "epoch": 0.49, + "learning_rate": 4.193318799692229e-05, + "loss": 0.5473, + "step": 12631 + }, + { + "epoch": 0.49, + "learning_rate": 4.1932546806873555e-05, + "loss": 0.5856, + "step": 12632 + }, + { + "epoch": 0.49, + "learning_rate": 4.1931905616824826e-05, + "loss": 0.5668, + "step": 12633 + }, + { + "epoch": 0.49, + "learning_rate": 4.19312644267761e-05, + "loss": 0.5007, + "step": 12634 + }, + { + "epoch": 0.49, + "learning_rate": 4.193062323672737e-05, + "loss": 0.5687, + "step": 12635 + }, + { + "epoch": 0.49, + "learning_rate": 4.192998204667864e-05, + "loss": 0.5294, + "step": 12636 + }, + { + "epoch": 0.49, + "learning_rate": 4.1929340856629906e-05, + "loss": 0.5132, + "step": 12637 + }, + { + "epoch": 0.49, + "learning_rate": 4.192869966658118e-05, + "loss": 0.566, + "step": 12638 + }, + { + "epoch": 0.49, + "learning_rate": 4.192805847653244e-05, + "loss": 0.5521, + "step": 12639 + }, + { + "epoch": 0.49, + "learning_rate": 4.1927417286483715e-05, + "loss": 0.5363, + "step": 12640 + }, + { + "epoch": 0.49, + "learning_rate": 4.1926776096434986e-05, + "loss": 0.5001, + "step": 12641 + }, + { + "epoch": 0.49, + "learning_rate": 4.192613490638626e-05, + "loss": 0.45, + "step": 12642 + }, + { + "epoch": 0.49, + "learning_rate": 4.192549371633753e-05, + "loss": 0.5504, + "step": 12643 + }, + { + "epoch": 0.49, + "learning_rate": 4.1924852526288795e-05, + "loss": 0.5751, + "step": 12644 + }, + { + "epoch": 0.49, + "learning_rate": 4.1924211336240067e-05, + "loss": 0.4559, + "step": 12645 + }, + { + "epoch": 0.49, + "learning_rate": 4.192357014619133e-05, + "loss": 0.6372, + "step": 12646 + }, + { + "epoch": 0.49, + "learning_rate": 4.19229289561426e-05, + "loss": 0.4999, + "step": 12647 + }, + { + "epoch": 0.49, + "learning_rate": 4.192228776609387e-05, + "loss": 0.534, + "step": 12648 + }, + { + "epoch": 0.49, + "learning_rate": 4.192164657604514e-05, + "loss": 0.5888, + "step": 12649 + }, + { + "epoch": 0.49, + "learning_rate": 4.192100538599641e-05, + "loss": 0.4246, + "step": 12650 + }, + { + "epoch": 0.49, + "learning_rate": 4.192036419594768e-05, + "loss": 0.6861, + "step": 12651 + }, + { + "epoch": 0.49, + "learning_rate": 4.1919723005898955e-05, + "loss": 0.6271, + "step": 12652 + }, + { + "epoch": 0.49, + "learning_rate": 4.191908181585022e-05, + "loss": 0.5554, + "step": 12653 + }, + { + "epoch": 0.49, + "learning_rate": 4.191844062580149e-05, + "loss": 0.5679, + "step": 12654 + }, + { + "epoch": 0.49, + "learning_rate": 4.191779943575276e-05, + "loss": 0.5957, + "step": 12655 + }, + { + "epoch": 0.49, + "learning_rate": 4.191715824570403e-05, + "loss": 0.462, + "step": 12656 + }, + { + "epoch": 0.49, + "learning_rate": 4.191651705565529e-05, + "loss": 0.4477, + "step": 12657 + }, + { + "epoch": 0.49, + "learning_rate": 4.191587586560657e-05, + "loss": 0.6565, + "step": 12658 + }, + { + "epoch": 0.49, + "learning_rate": 4.191523467555784e-05, + "loss": 0.5765, + "step": 12659 + }, + { + "epoch": 0.49, + "learning_rate": 4.191459348550911e-05, + "loss": 0.6415, + "step": 12660 + }, + { + "epoch": 0.49, + "learning_rate": 4.191395229546038e-05, + "loss": 0.5816, + "step": 12661 + }, + { + "epoch": 0.49, + "learning_rate": 4.1913311105411645e-05, + "loss": 0.6732, + "step": 12662 + }, + { + "epoch": 0.49, + "learning_rate": 4.191266991536292e-05, + "loss": 0.4366, + "step": 12663 + }, + { + "epoch": 0.49, + "learning_rate": 4.191202872531418e-05, + "loss": 0.6171, + "step": 12664 + }, + { + "epoch": 0.49, + "learning_rate": 4.1911387535265453e-05, + "loss": 0.6257, + "step": 12665 + }, + { + "epoch": 0.49, + "learning_rate": 4.1910746345216725e-05, + "loss": 0.5655, + "step": 12666 + }, + { + "epoch": 0.49, + "learning_rate": 4.1910105155168e-05, + "loss": 0.5822, + "step": 12667 + }, + { + "epoch": 0.49, + "learning_rate": 4.190946396511926e-05, + "loss": 0.5898, + "step": 12668 + }, + { + "epoch": 0.49, + "learning_rate": 4.1908822775070534e-05, + "loss": 0.5193, + "step": 12669 + }, + { + "epoch": 0.49, + "learning_rate": 4.1908181585021805e-05, + "loss": 0.5374, + "step": 12670 + }, + { + "epoch": 0.49, + "learning_rate": 4.190754039497307e-05, + "loss": 0.5841, + "step": 12671 + }, + { + "epoch": 0.49, + "learning_rate": 4.190689920492434e-05, + "loss": 0.5248, + "step": 12672 + }, + { + "epoch": 0.49, + "learning_rate": 4.190625801487561e-05, + "loss": 0.516, + "step": 12673 + }, + { + "epoch": 0.49, + "learning_rate": 4.1905616824826885e-05, + "loss": 0.5765, + "step": 12674 + }, + { + "epoch": 0.49, + "learning_rate": 4.190497563477815e-05, + "loss": 0.6361, + "step": 12675 + }, + { + "epoch": 0.49, + "learning_rate": 4.190433444472942e-05, + "loss": 0.517, + "step": 12676 + }, + { + "epoch": 0.49, + "learning_rate": 4.190369325468069e-05, + "loss": 0.5153, + "step": 12677 + }, + { + "epoch": 0.49, + "learning_rate": 4.190305206463196e-05, + "loss": 0.4602, + "step": 12678 + }, + { + "epoch": 0.49, + "learning_rate": 4.190241087458323e-05, + "loss": 0.5514, + "step": 12679 + }, + { + "epoch": 0.49, + "learning_rate": 4.1901769684534495e-05, + "loss": 0.5636, + "step": 12680 + }, + { + "epoch": 0.49, + "learning_rate": 4.190112849448577e-05, + "loss": 0.5801, + "step": 12681 + }, + { + "epoch": 0.49, + "learning_rate": 4.190048730443704e-05, + "loss": 0.5243, + "step": 12682 + }, + { + "epoch": 0.49, + "learning_rate": 4.189984611438831e-05, + "loss": 0.5027, + "step": 12683 + }, + { + "epoch": 0.49, + "learning_rate": 4.1899204924339575e-05, + "loss": 0.5609, + "step": 12684 + }, + { + "epoch": 0.49, + "learning_rate": 4.189856373429085e-05, + "loss": 0.586, + "step": 12685 + }, + { + "epoch": 0.49, + "learning_rate": 4.189792254424211e-05, + "loss": 0.4124, + "step": 12686 + }, + { + "epoch": 0.49, + "learning_rate": 4.1897281354193384e-05, + "loss": 0.4539, + "step": 12687 + }, + { + "epoch": 0.49, + "learning_rate": 4.1896640164144656e-05, + "loss": 0.4993, + "step": 12688 + }, + { + "epoch": 0.49, + "learning_rate": 4.189599897409592e-05, + "loss": 0.59, + "step": 12689 + }, + { + "epoch": 0.49, + "learning_rate": 4.18953577840472e-05, + "loss": 0.5296, + "step": 12690 + }, + { + "epoch": 0.49, + "learning_rate": 4.1894716593998464e-05, + "loss": 0.4846, + "step": 12691 + }, + { + "epoch": 0.49, + "learning_rate": 4.1894075403949736e-05, + "loss": 0.4759, + "step": 12692 + }, + { + "epoch": 0.49, + "learning_rate": 4.1893434213901e-05, + "loss": 0.612, + "step": 12693 + }, + { + "epoch": 0.49, + "learning_rate": 4.189279302385227e-05, + "loss": 0.6812, + "step": 12694 + }, + { + "epoch": 0.49, + "learning_rate": 4.189215183380354e-05, + "loss": 0.5749, + "step": 12695 + }, + { + "epoch": 0.49, + "learning_rate": 4.189151064375481e-05, + "loss": 0.4672, + "step": 12696 + }, + { + "epoch": 0.49, + "learning_rate": 4.189086945370608e-05, + "loss": 0.5021, + "step": 12697 + }, + { + "epoch": 0.49, + "learning_rate": 4.189022826365735e-05, + "loss": 0.4717, + "step": 12698 + }, + { + "epoch": 0.49, + "learning_rate": 4.1889587073608624e-05, + "loss": 0.516, + "step": 12699 + }, + { + "epoch": 0.49, + "learning_rate": 4.188894588355989e-05, + "loss": 0.4953, + "step": 12700 + }, + { + "epoch": 0.49, + "learning_rate": 4.188830469351116e-05, + "loss": 0.5508, + "step": 12701 + }, + { + "epoch": 0.49, + "learning_rate": 4.1887663503462426e-05, + "loss": 0.6251, + "step": 12702 + }, + { + "epoch": 0.49, + "learning_rate": 4.18870223134137e-05, + "loss": 0.5181, + "step": 12703 + }, + { + "epoch": 0.49, + "learning_rate": 4.188638112336496e-05, + "loss": 0.5387, + "step": 12704 + }, + { + "epoch": 0.49, + "learning_rate": 4.1885739933316234e-05, + "loss": 0.4928, + "step": 12705 + }, + { + "epoch": 0.49, + "learning_rate": 4.1885098743267506e-05, + "loss": 0.7365, + "step": 12706 + }, + { + "epoch": 0.49, + "learning_rate": 4.188445755321878e-05, + "loss": 0.5597, + "step": 12707 + }, + { + "epoch": 0.49, + "learning_rate": 4.188381636317005e-05, + "loss": 0.5983, + "step": 12708 + }, + { + "epoch": 0.49, + "learning_rate": 4.1883175173121314e-05, + "loss": 0.5265, + "step": 12709 + }, + { + "epoch": 0.49, + "learning_rate": 4.1882533983072586e-05, + "loss": 0.5406, + "step": 12710 + }, + { + "epoch": 0.49, + "learning_rate": 4.188189279302385e-05, + "loss": 0.5253, + "step": 12711 + }, + { + "epoch": 0.49, + "learning_rate": 4.188125160297512e-05, + "loss": 0.5856, + "step": 12712 + }, + { + "epoch": 0.49, + "learning_rate": 4.1880610412926394e-05, + "loss": 0.569, + "step": 12713 + }, + { + "epoch": 0.49, + "learning_rate": 4.1879969222877666e-05, + "loss": 0.487, + "step": 12714 + }, + { + "epoch": 0.49, + "learning_rate": 4.187932803282894e-05, + "loss": 0.5671, + "step": 12715 + }, + { + "epoch": 0.49, + "learning_rate": 4.18786868427802e-05, + "loss": 0.546, + "step": 12716 + }, + { + "epoch": 0.49, + "learning_rate": 4.1878045652731474e-05, + "loss": 0.5051, + "step": 12717 + }, + { + "epoch": 0.49, + "learning_rate": 4.187740446268274e-05, + "loss": 0.6141, + "step": 12718 + }, + { + "epoch": 0.49, + "learning_rate": 4.187676327263401e-05, + "loss": 0.5938, + "step": 12719 + }, + { + "epoch": 0.49, + "learning_rate": 4.1876122082585276e-05, + "loss": 0.5677, + "step": 12720 + }, + { + "epoch": 0.49, + "learning_rate": 4.187548089253655e-05, + "loss": 0.555, + "step": 12721 + }, + { + "epoch": 0.49, + "learning_rate": 4.187483970248782e-05, + "loss": 0.5793, + "step": 12722 + }, + { + "epoch": 0.49, + "learning_rate": 4.187419851243909e-05, + "loss": 0.5065, + "step": 12723 + }, + { + "epoch": 0.49, + "learning_rate": 4.187355732239036e-05, + "loss": 0.576, + "step": 12724 + }, + { + "epoch": 0.49, + "learning_rate": 4.187291613234163e-05, + "loss": 0.5304, + "step": 12725 + }, + { + "epoch": 0.49, + "learning_rate": 4.18722749422929e-05, + "loss": 0.5327, + "step": 12726 + }, + { + "epoch": 0.49, + "learning_rate": 4.1871633752244164e-05, + "loss": 0.552, + "step": 12727 + }, + { + "epoch": 0.49, + "learning_rate": 4.1870992562195436e-05, + "loss": 0.5855, + "step": 12728 + }, + { + "epoch": 0.49, + "learning_rate": 4.187035137214671e-05, + "loss": 0.4428, + "step": 12729 + }, + { + "epoch": 0.49, + "learning_rate": 4.186971018209798e-05, + "loss": 0.4779, + "step": 12730 + }, + { + "epoch": 0.49, + "learning_rate": 4.1869068992049244e-05, + "loss": 0.5029, + "step": 12731 + }, + { + "epoch": 0.49, + "learning_rate": 4.1868427802000516e-05, + "loss": 0.537, + "step": 12732 + }, + { + "epoch": 0.49, + "learning_rate": 4.186778661195179e-05, + "loss": 0.503, + "step": 12733 + }, + { + "epoch": 0.49, + "learning_rate": 4.186714542190305e-05, + "loss": 0.4692, + "step": 12734 + }, + { + "epoch": 0.49, + "learning_rate": 4.1866504231854325e-05, + "loss": 0.5221, + "step": 12735 + }, + { + "epoch": 0.49, + "learning_rate": 4.186586304180559e-05, + "loss": 0.7074, + "step": 12736 + }, + { + "epoch": 0.49, + "learning_rate": 4.186522185175686e-05, + "loss": 0.5617, + "step": 12737 + }, + { + "epoch": 0.49, + "learning_rate": 4.186458066170813e-05, + "loss": 0.5833, + "step": 12738 + }, + { + "epoch": 0.49, + "learning_rate": 4.1863939471659405e-05, + "loss": 0.5841, + "step": 12739 + }, + { + "epoch": 0.49, + "learning_rate": 4.186329828161067e-05, + "loss": 0.5668, + "step": 12740 + }, + { + "epoch": 0.49, + "learning_rate": 4.186265709156194e-05, + "loss": 0.5435, + "step": 12741 + }, + { + "epoch": 0.49, + "learning_rate": 4.186201590151321e-05, + "loss": 0.5235, + "step": 12742 + }, + { + "epoch": 0.49, + "learning_rate": 4.186137471146448e-05, + "loss": 0.4221, + "step": 12743 + }, + { + "epoch": 0.49, + "learning_rate": 4.186073352141575e-05, + "loss": 0.5832, + "step": 12744 + }, + { + "epoch": 0.49, + "learning_rate": 4.1860092331367015e-05, + "loss": 0.4941, + "step": 12745 + }, + { + "epoch": 0.49, + "learning_rate": 4.185945114131829e-05, + "loss": 0.5817, + "step": 12746 + }, + { + "epoch": 0.49, + "learning_rate": 4.185880995126956e-05, + "loss": 0.6827, + "step": 12747 + }, + { + "epoch": 0.49, + "learning_rate": 4.185816876122083e-05, + "loss": 0.5367, + "step": 12748 + }, + { + "epoch": 0.49, + "learning_rate": 4.1857527571172095e-05, + "loss": 0.4222, + "step": 12749 + }, + { + "epoch": 0.49, + "learning_rate": 4.1856886381123366e-05, + "loss": 0.589, + "step": 12750 + }, + { + "epoch": 0.49, + "learning_rate": 4.185624519107464e-05, + "loss": 0.4973, + "step": 12751 + }, + { + "epoch": 0.49, + "learning_rate": 4.18556040010259e-05, + "loss": 0.4923, + "step": 12752 + }, + { + "epoch": 0.49, + "learning_rate": 4.1854962810977175e-05, + "loss": 0.5777, + "step": 12753 + }, + { + "epoch": 0.49, + "learning_rate": 4.1854321620928447e-05, + "loss": 0.4868, + "step": 12754 + }, + { + "epoch": 0.49, + "learning_rate": 4.185368043087972e-05, + "loss": 0.4954, + "step": 12755 + }, + { + "epoch": 0.49, + "learning_rate": 4.185303924083098e-05, + "loss": 0.6656, + "step": 12756 + }, + { + "epoch": 0.49, + "learning_rate": 4.1852398050782255e-05, + "loss": 0.5853, + "step": 12757 + }, + { + "epoch": 0.49, + "learning_rate": 4.185175686073352e-05, + "loss": 0.5817, + "step": 12758 + }, + { + "epoch": 0.49, + "learning_rate": 4.185111567068479e-05, + "loss": 0.5248, + "step": 12759 + }, + { + "epoch": 0.49, + "learning_rate": 4.185047448063606e-05, + "loss": 0.4831, + "step": 12760 + }, + { + "epoch": 0.49, + "learning_rate": 4.184983329058733e-05, + "loss": 0.6435, + "step": 12761 + }, + { + "epoch": 0.49, + "learning_rate": 4.184919210053861e-05, + "loss": 0.5112, + "step": 12762 + }, + { + "epoch": 0.49, + "learning_rate": 4.184855091048987e-05, + "loss": 0.4387, + "step": 12763 + }, + { + "epoch": 0.49, + "learning_rate": 4.184790972044114e-05, + "loss": 0.4961, + "step": 12764 + }, + { + "epoch": 0.49, + "learning_rate": 4.184726853039241e-05, + "loss": 0.5167, + "step": 12765 + }, + { + "epoch": 0.49, + "learning_rate": 4.184662734034368e-05, + "loss": 0.5408, + "step": 12766 + }, + { + "epoch": 0.49, + "learning_rate": 4.1845986150294945e-05, + "loss": 0.5668, + "step": 12767 + }, + { + "epoch": 0.49, + "learning_rate": 4.184534496024622e-05, + "loss": 0.6723, + "step": 12768 + }, + { + "epoch": 0.49, + "learning_rate": 4.184470377019749e-05, + "loss": 0.5877, + "step": 12769 + }, + { + "epoch": 0.49, + "learning_rate": 4.184406258014876e-05, + "loss": 0.4503, + "step": 12770 + }, + { + "epoch": 0.49, + "learning_rate": 4.184342139010003e-05, + "loss": 0.5289, + "step": 12771 + }, + { + "epoch": 0.49, + "learning_rate": 4.18427802000513e-05, + "loss": 0.4868, + "step": 12772 + }, + { + "epoch": 0.49, + "learning_rate": 4.184213901000257e-05, + "loss": 0.4722, + "step": 12773 + }, + { + "epoch": 0.49, + "learning_rate": 4.1841497819953833e-05, + "loss": 0.4573, + "step": 12774 + }, + { + "epoch": 0.49, + "learning_rate": 4.1840856629905105e-05, + "loss": 0.6117, + "step": 12775 + }, + { + "epoch": 0.49, + "learning_rate": 4.184021543985637e-05, + "loss": 0.6844, + "step": 12776 + }, + { + "epoch": 0.49, + "learning_rate": 4.183957424980764e-05, + "loss": 0.5371, + "step": 12777 + }, + { + "epoch": 0.49, + "learning_rate": 4.1838933059758914e-05, + "loss": 0.5578, + "step": 12778 + }, + { + "epoch": 0.49, + "learning_rate": 4.1838291869710185e-05, + "loss": 0.4895, + "step": 12779 + }, + { + "epoch": 0.49, + "learning_rate": 4.183765067966146e-05, + "loss": 0.6065, + "step": 12780 + }, + { + "epoch": 0.49, + "learning_rate": 4.183700948961272e-05, + "loss": 0.5022, + "step": 12781 + }, + { + "epoch": 0.49, + "learning_rate": 4.1836368299563994e-05, + "loss": 0.5219, + "step": 12782 + }, + { + "epoch": 0.49, + "learning_rate": 4.183572710951526e-05, + "loss": 0.5188, + "step": 12783 + }, + { + "epoch": 0.49, + "learning_rate": 4.183508591946653e-05, + "loss": 0.4612, + "step": 12784 + }, + { + "epoch": 0.49, + "learning_rate": 4.18344447294178e-05, + "loss": 0.4863, + "step": 12785 + }, + { + "epoch": 0.49, + "learning_rate": 4.1833803539369074e-05, + "loss": 0.5703, + "step": 12786 + }, + { + "epoch": 0.49, + "learning_rate": 4.1833162349320345e-05, + "loss": 0.6066, + "step": 12787 + }, + { + "epoch": 0.49, + "learning_rate": 4.183252115927161e-05, + "loss": 0.4712, + "step": 12788 + }, + { + "epoch": 0.49, + "learning_rate": 4.183187996922288e-05, + "loss": 0.6362, + "step": 12789 + }, + { + "epoch": 0.49, + "learning_rate": 4.183123877917415e-05, + "loss": 0.5355, + "step": 12790 + }, + { + "epoch": 0.49, + "learning_rate": 4.183059758912542e-05, + "loss": 0.5119, + "step": 12791 + }, + { + "epoch": 0.49, + "learning_rate": 4.1829956399076684e-05, + "loss": 0.5742, + "step": 12792 + }, + { + "epoch": 0.49, + "learning_rate": 4.1829315209027955e-05, + "loss": 0.61, + "step": 12793 + }, + { + "epoch": 0.49, + "learning_rate": 4.182867401897923e-05, + "loss": 0.7572, + "step": 12794 + }, + { + "epoch": 0.49, + "learning_rate": 4.18280328289305e-05, + "loss": 0.5627, + "step": 12795 + }, + { + "epoch": 0.49, + "learning_rate": 4.182739163888177e-05, + "loss": 0.5247, + "step": 12796 + }, + { + "epoch": 0.49, + "learning_rate": 4.1826750448833035e-05, + "loss": 0.532, + "step": 12797 + }, + { + "epoch": 0.49, + "learning_rate": 4.182610925878431e-05, + "loss": 0.5735, + "step": 12798 + }, + { + "epoch": 0.49, + "learning_rate": 4.182546806873557e-05, + "loss": 0.4582, + "step": 12799 + }, + { + "epoch": 0.49, + "learning_rate": 4.1824826878686844e-05, + "loss": 0.5449, + "step": 12800 + }, + { + "epoch": 0.49, + "learning_rate": 4.1824185688638116e-05, + "loss": 0.6108, + "step": 12801 + }, + { + "epoch": 0.49, + "learning_rate": 4.182354449858939e-05, + "loss": 0.5189, + "step": 12802 + }, + { + "epoch": 0.49, + "learning_rate": 4.182290330854065e-05, + "loss": 0.4757, + "step": 12803 + }, + { + "epoch": 0.49, + "learning_rate": 4.1822262118491924e-05, + "loss": 0.5657, + "step": 12804 + }, + { + "epoch": 0.49, + "learning_rate": 4.1821620928443196e-05, + "loss": 0.5706, + "step": 12805 + }, + { + "epoch": 0.49, + "learning_rate": 4.182097973839446e-05, + "loss": 0.4918, + "step": 12806 + }, + { + "epoch": 0.49, + "learning_rate": 4.182033854834573e-05, + "loss": 0.5354, + "step": 12807 + }, + { + "epoch": 0.49, + "learning_rate": 4.1819697358297e-05, + "loss": 0.5175, + "step": 12808 + }, + { + "epoch": 0.49, + "learning_rate": 4.181905616824827e-05, + "loss": 0.5047, + "step": 12809 + }, + { + "epoch": 0.49, + "learning_rate": 4.181841497819954e-05, + "loss": 0.527, + "step": 12810 + }, + { + "epoch": 0.49, + "learning_rate": 4.181777378815081e-05, + "loss": 0.4653, + "step": 12811 + }, + { + "epoch": 0.49, + "learning_rate": 4.181713259810208e-05, + "loss": 0.496, + "step": 12812 + }, + { + "epoch": 0.49, + "learning_rate": 4.181649140805335e-05, + "loss": 0.4777, + "step": 12813 + }, + { + "epoch": 0.49, + "learning_rate": 4.181585021800462e-05, + "loss": 0.5567, + "step": 12814 + }, + { + "epoch": 0.49, + "learning_rate": 4.1815209027955886e-05, + "loss": 0.5187, + "step": 12815 + }, + { + "epoch": 0.49, + "learning_rate": 4.181456783790716e-05, + "loss": 0.5495, + "step": 12816 + }, + { + "epoch": 0.49, + "learning_rate": 4.181392664785842e-05, + "loss": 0.548, + "step": 12817 + }, + { + "epoch": 0.49, + "learning_rate": 4.18132854578097e-05, + "loss": 0.6947, + "step": 12818 + }, + { + "epoch": 0.49, + "learning_rate": 4.1812644267760966e-05, + "loss": 0.479, + "step": 12819 + }, + { + "epoch": 0.49, + "learning_rate": 4.181200307771224e-05, + "loss": 0.4702, + "step": 12820 + }, + { + "epoch": 0.49, + "learning_rate": 4.18113618876635e-05, + "loss": 0.5823, + "step": 12821 + }, + { + "epoch": 0.49, + "learning_rate": 4.1810720697614774e-05, + "loss": 0.6175, + "step": 12822 + }, + { + "epoch": 0.49, + "learning_rate": 4.1810079507566046e-05, + "loss": 0.4953, + "step": 12823 + }, + { + "epoch": 0.49, + "learning_rate": 4.180943831751731e-05, + "loss": 0.5339, + "step": 12824 + }, + { + "epoch": 0.49, + "learning_rate": 4.180879712746858e-05, + "loss": 0.4323, + "step": 12825 + }, + { + "epoch": 0.49, + "learning_rate": 4.1808155937419854e-05, + "loss": 0.6503, + "step": 12826 + }, + { + "epoch": 0.49, + "learning_rate": 4.1807514747371126e-05, + "loss": 0.5574, + "step": 12827 + }, + { + "epoch": 0.49, + "learning_rate": 4.180687355732239e-05, + "loss": 0.561, + "step": 12828 + }, + { + "epoch": 0.49, + "learning_rate": 4.180623236727366e-05, + "loss": 0.5008, + "step": 12829 + }, + { + "epoch": 0.49, + "learning_rate": 4.1805591177224934e-05, + "loss": 0.483, + "step": 12830 + }, + { + "epoch": 0.49, + "learning_rate": 4.18049499871762e-05, + "loss": 0.7044, + "step": 12831 + }, + { + "epoch": 0.49, + "learning_rate": 4.180430879712747e-05, + "loss": 0.5208, + "step": 12832 + }, + { + "epoch": 0.49, + "learning_rate": 4.1803667607078736e-05, + "loss": 0.5509, + "step": 12833 + }, + { + "epoch": 0.49, + "learning_rate": 4.1803026417030014e-05, + "loss": 0.3417, + "step": 12834 + }, + { + "epoch": 0.49, + "learning_rate": 4.180238522698128e-05, + "loss": 0.5653, + "step": 12835 + }, + { + "epoch": 0.49, + "learning_rate": 4.180174403693255e-05, + "loss": 0.6629, + "step": 12836 + }, + { + "epoch": 0.49, + "learning_rate": 4.1801102846883816e-05, + "loss": 0.6457, + "step": 12837 + }, + { + "epoch": 0.49, + "learning_rate": 4.180046165683509e-05, + "loss": 0.4722, + "step": 12838 + }, + { + "epoch": 0.49, + "learning_rate": 4.179982046678636e-05, + "loss": 0.5164, + "step": 12839 + }, + { + "epoch": 0.49, + "learning_rate": 4.1799179276737624e-05, + "loss": 0.5383, + "step": 12840 + }, + { + "epoch": 0.49, + "learning_rate": 4.1798538086688896e-05, + "loss": 0.6425, + "step": 12841 + }, + { + "epoch": 0.49, + "learning_rate": 4.179789689664017e-05, + "loss": 0.5812, + "step": 12842 + }, + { + "epoch": 0.49, + "learning_rate": 4.179725570659144e-05, + "loss": 0.5338, + "step": 12843 + }, + { + "epoch": 0.49, + "learning_rate": 4.1796614516542704e-05, + "loss": 0.5372, + "step": 12844 + }, + { + "epoch": 0.49, + "learning_rate": 4.1795973326493976e-05, + "loss": 0.5507, + "step": 12845 + }, + { + "epoch": 0.49, + "learning_rate": 4.179533213644524e-05, + "loss": 0.6367, + "step": 12846 + }, + { + "epoch": 0.49, + "learning_rate": 4.179469094639651e-05, + "loss": 0.6109, + "step": 12847 + }, + { + "epoch": 0.49, + "learning_rate": 4.1794049756347785e-05, + "loss": 0.5069, + "step": 12848 + }, + { + "epoch": 0.49, + "learning_rate": 4.179340856629905e-05, + "loss": 0.5483, + "step": 12849 + }, + { + "epoch": 0.49, + "learning_rate": 4.179276737625033e-05, + "loss": 0.4352, + "step": 12850 + }, + { + "epoch": 0.49, + "learning_rate": 4.179212618620159e-05, + "loss": 0.5058, + "step": 12851 + }, + { + "epoch": 0.49, + "learning_rate": 4.1791484996152865e-05, + "loss": 0.6761, + "step": 12852 + }, + { + "epoch": 0.49, + "learning_rate": 4.179084380610413e-05, + "loss": 0.5455, + "step": 12853 + }, + { + "epoch": 0.49, + "learning_rate": 4.17902026160554e-05, + "loss": 0.5144, + "step": 12854 + }, + { + "epoch": 0.49, + "learning_rate": 4.1789561426006666e-05, + "loss": 0.5477, + "step": 12855 + }, + { + "epoch": 0.49, + "learning_rate": 4.178892023595794e-05, + "loss": 0.523, + "step": 12856 + }, + { + "epoch": 0.49, + "learning_rate": 4.178827904590921e-05, + "loss": 0.5139, + "step": 12857 + }, + { + "epoch": 0.49, + "learning_rate": 4.178763785586048e-05, + "loss": 0.5174, + "step": 12858 + }, + { + "epoch": 0.49, + "learning_rate": 4.178699666581175e-05, + "loss": 0.4726, + "step": 12859 + }, + { + "epoch": 0.49, + "learning_rate": 4.178635547576302e-05, + "loss": 0.4793, + "step": 12860 + }, + { + "epoch": 0.49, + "learning_rate": 4.178571428571429e-05, + "loss": 0.5138, + "step": 12861 + }, + { + "epoch": 0.49, + "learning_rate": 4.1785073095665555e-05, + "loss": 0.487, + "step": 12862 + }, + { + "epoch": 0.49, + "learning_rate": 4.1784431905616826e-05, + "loss": 0.5834, + "step": 12863 + }, + { + "epoch": 0.49, + "learning_rate": 4.178379071556809e-05, + "loss": 0.5173, + "step": 12864 + }, + { + "epoch": 0.49, + "learning_rate": 4.178314952551936e-05, + "loss": 0.6262, + "step": 12865 + }, + { + "epoch": 0.49, + "learning_rate": 4.1782508335470635e-05, + "loss": 0.566, + "step": 12866 + }, + { + "epoch": 0.49, + "learning_rate": 4.1781867145421907e-05, + "loss": 0.4565, + "step": 12867 + }, + { + "epoch": 0.49, + "learning_rate": 4.178122595537318e-05, + "loss": 0.5536, + "step": 12868 + }, + { + "epoch": 0.49, + "learning_rate": 4.178058476532444e-05, + "loss": 0.6524, + "step": 12869 + }, + { + "epoch": 0.49, + "learning_rate": 4.1779943575275715e-05, + "loss": 0.4717, + "step": 12870 + }, + { + "epoch": 0.49, + "learning_rate": 4.177930238522698e-05, + "loss": 0.5238, + "step": 12871 + }, + { + "epoch": 0.49, + "learning_rate": 4.177866119517825e-05, + "loss": 0.4694, + "step": 12872 + }, + { + "epoch": 0.49, + "learning_rate": 4.177802000512952e-05, + "loss": 0.5186, + "step": 12873 + }, + { + "epoch": 0.49, + "learning_rate": 4.1777378815080795e-05, + "loss": 0.5501, + "step": 12874 + }, + { + "epoch": 0.5, + "learning_rate": 4.177673762503207e-05, + "loss": 0.5468, + "step": 12875 + }, + { + "epoch": 0.5, + "learning_rate": 4.177609643498333e-05, + "loss": 0.5965, + "step": 12876 + }, + { + "epoch": 0.5, + "learning_rate": 4.17754552449346e-05, + "loss": 0.4955, + "step": 12877 + }, + { + "epoch": 0.5, + "learning_rate": 4.177481405488587e-05, + "loss": 0.6256, + "step": 12878 + }, + { + "epoch": 0.5, + "learning_rate": 4.177417286483714e-05, + "loss": 0.5998, + "step": 12879 + }, + { + "epoch": 0.5, + "learning_rate": 4.1773531674788405e-05, + "loss": 0.4673, + "step": 12880 + }, + { + "epoch": 0.5, + "learning_rate": 4.177289048473968e-05, + "loss": 0.5082, + "step": 12881 + }, + { + "epoch": 0.5, + "learning_rate": 4.177224929469095e-05, + "loss": 0.493, + "step": 12882 + }, + { + "epoch": 0.5, + "learning_rate": 4.177160810464222e-05, + "loss": 0.4594, + "step": 12883 + }, + { + "epoch": 0.5, + "learning_rate": 4.177096691459349e-05, + "loss": 0.5238, + "step": 12884 + }, + { + "epoch": 0.5, + "learning_rate": 4.177032572454476e-05, + "loss": 0.5495, + "step": 12885 + }, + { + "epoch": 0.5, + "learning_rate": 4.176968453449603e-05, + "loss": 0.568, + "step": 12886 + }, + { + "epoch": 0.5, + "learning_rate": 4.1769043344447293e-05, + "loss": 0.5229, + "step": 12887 + }, + { + "epoch": 0.5, + "learning_rate": 4.1768402154398565e-05, + "loss": 0.5344, + "step": 12888 + }, + { + "epoch": 0.5, + "learning_rate": 4.176776096434984e-05, + "loss": 0.528, + "step": 12889 + }, + { + "epoch": 0.5, + "learning_rate": 4.176711977430111e-05, + "loss": 0.5353, + "step": 12890 + }, + { + "epoch": 0.5, + "learning_rate": 4.1766478584252374e-05, + "loss": 0.6542, + "step": 12891 + }, + { + "epoch": 0.5, + "learning_rate": 4.1765837394203645e-05, + "loss": 0.5307, + "step": 12892 + }, + { + "epoch": 0.5, + "learning_rate": 4.176519620415492e-05, + "loss": 0.544, + "step": 12893 + }, + { + "epoch": 0.5, + "learning_rate": 4.176455501410618e-05, + "loss": 0.5521, + "step": 12894 + }, + { + "epoch": 0.5, + "learning_rate": 4.1763913824057454e-05, + "loss": 0.576, + "step": 12895 + }, + { + "epoch": 0.5, + "learning_rate": 4.176327263400872e-05, + "loss": 0.5257, + "step": 12896 + }, + { + "epoch": 0.5, + "learning_rate": 4.176263144395999e-05, + "loss": 0.4757, + "step": 12897 + }, + { + "epoch": 0.5, + "learning_rate": 4.176199025391126e-05, + "loss": 0.5523, + "step": 12898 + }, + { + "epoch": 0.5, + "learning_rate": 4.1761349063862534e-05, + "loss": 0.5002, + "step": 12899 + }, + { + "epoch": 0.5, + "learning_rate": 4.17607078738138e-05, + "loss": 0.6141, + "step": 12900 + }, + { + "epoch": 0.5, + "learning_rate": 4.176006668376507e-05, + "loss": 0.5334, + "step": 12901 + }, + { + "epoch": 0.5, + "learning_rate": 4.175942549371634e-05, + "loss": 0.6195, + "step": 12902 + }, + { + "epoch": 0.5, + "learning_rate": 4.175878430366761e-05, + "loss": 0.5669, + "step": 12903 + }, + { + "epoch": 0.5, + "learning_rate": 4.175814311361888e-05, + "loss": 0.5236, + "step": 12904 + }, + { + "epoch": 0.5, + "learning_rate": 4.1757501923570144e-05, + "loss": 0.5576, + "step": 12905 + }, + { + "epoch": 0.5, + "learning_rate": 4.175686073352142e-05, + "loss": 0.5933, + "step": 12906 + }, + { + "epoch": 0.5, + "learning_rate": 4.175621954347269e-05, + "loss": 0.4587, + "step": 12907 + }, + { + "epoch": 0.5, + "learning_rate": 4.175557835342396e-05, + "loss": 0.466, + "step": 12908 + }, + { + "epoch": 0.5, + "learning_rate": 4.1754937163375224e-05, + "loss": 0.4703, + "step": 12909 + }, + { + "epoch": 0.5, + "learning_rate": 4.1754295973326495e-05, + "loss": 0.5098, + "step": 12910 + }, + { + "epoch": 0.5, + "learning_rate": 4.175365478327777e-05, + "loss": 0.5839, + "step": 12911 + }, + { + "epoch": 0.5, + "learning_rate": 4.175301359322903e-05, + "loss": 0.5388, + "step": 12912 + }, + { + "epoch": 0.5, + "learning_rate": 4.1752372403180304e-05, + "loss": 0.5129, + "step": 12913 + }, + { + "epoch": 0.5, + "learning_rate": 4.1751731213131576e-05, + "loss": 0.643, + "step": 12914 + }, + { + "epoch": 0.5, + "learning_rate": 4.175109002308285e-05, + "loss": 0.6929, + "step": 12915 + }, + { + "epoch": 0.5, + "learning_rate": 4.175044883303411e-05, + "loss": 0.5525, + "step": 12916 + }, + { + "epoch": 0.5, + "learning_rate": 4.1749807642985384e-05, + "loss": 0.489, + "step": 12917 + }, + { + "epoch": 0.5, + "learning_rate": 4.174916645293665e-05, + "loss": 0.4997, + "step": 12918 + }, + { + "epoch": 0.5, + "learning_rate": 4.174852526288792e-05, + "loss": 0.6618, + "step": 12919 + }, + { + "epoch": 0.5, + "learning_rate": 4.174788407283919e-05, + "loss": 0.5219, + "step": 12920 + }, + { + "epoch": 0.5, + "learning_rate": 4.174724288279046e-05, + "loss": 0.612, + "step": 12921 + }, + { + "epoch": 0.5, + "learning_rate": 4.1746601692741736e-05, + "loss": 0.6051, + "step": 12922 + }, + { + "epoch": 0.5, + "learning_rate": 4.1745960502693e-05, + "loss": 0.5795, + "step": 12923 + }, + { + "epoch": 0.5, + "learning_rate": 4.174531931264427e-05, + "loss": 0.5907, + "step": 12924 + }, + { + "epoch": 0.5, + "learning_rate": 4.174467812259554e-05, + "loss": 0.5139, + "step": 12925 + }, + { + "epoch": 0.5, + "learning_rate": 4.174403693254681e-05, + "loss": 0.5615, + "step": 12926 + }, + { + "epoch": 0.5, + "learning_rate": 4.1743395742498074e-05, + "loss": 0.516, + "step": 12927 + }, + { + "epoch": 0.5, + "learning_rate": 4.1742754552449346e-05, + "loss": 0.577, + "step": 12928 + }, + { + "epoch": 0.5, + "learning_rate": 4.174211336240062e-05, + "loss": 0.5963, + "step": 12929 + }, + { + "epoch": 0.5, + "learning_rate": 4.174147217235189e-05, + "loss": 0.5544, + "step": 12930 + }, + { + "epoch": 0.5, + "learning_rate": 4.174083098230316e-05, + "loss": 0.542, + "step": 12931 + }, + { + "epoch": 0.5, + "learning_rate": 4.1740189792254426e-05, + "loss": 0.5256, + "step": 12932 + }, + { + "epoch": 0.5, + "learning_rate": 4.17395486022057e-05, + "loss": 0.5966, + "step": 12933 + }, + { + "epoch": 0.5, + "learning_rate": 4.173890741215696e-05, + "loss": 0.5652, + "step": 12934 + }, + { + "epoch": 0.5, + "learning_rate": 4.1738266222108234e-05, + "loss": 0.4978, + "step": 12935 + }, + { + "epoch": 0.5, + "learning_rate": 4.17376250320595e-05, + "loss": 0.4791, + "step": 12936 + }, + { + "epoch": 0.5, + "learning_rate": 4.173698384201077e-05, + "loss": 0.5468, + "step": 12937 + }, + { + "epoch": 0.5, + "learning_rate": 4.173634265196204e-05, + "loss": 0.4869, + "step": 12938 + }, + { + "epoch": 0.5, + "learning_rate": 4.1735701461913314e-05, + "loss": 0.5161, + "step": 12939 + }, + { + "epoch": 0.5, + "learning_rate": 4.1735060271864586e-05, + "loss": 0.5816, + "step": 12940 + }, + { + "epoch": 0.5, + "learning_rate": 4.173441908181585e-05, + "loss": 0.5715, + "step": 12941 + }, + { + "epoch": 0.5, + "learning_rate": 4.173377789176712e-05, + "loss": 0.4635, + "step": 12942 + }, + { + "epoch": 0.5, + "learning_rate": 4.173313670171839e-05, + "loss": 0.5803, + "step": 12943 + }, + { + "epoch": 0.5, + "learning_rate": 4.173249551166966e-05, + "loss": 0.4533, + "step": 12944 + }, + { + "epoch": 0.5, + "learning_rate": 4.173185432162093e-05, + "loss": 0.6316, + "step": 12945 + }, + { + "epoch": 0.5, + "learning_rate": 4.17312131315722e-05, + "loss": 0.5982, + "step": 12946 + }, + { + "epoch": 0.5, + "learning_rate": 4.1730571941523474e-05, + "loss": 0.654, + "step": 12947 + }, + { + "epoch": 0.5, + "learning_rate": 4.172993075147474e-05, + "loss": 0.5925, + "step": 12948 + }, + { + "epoch": 0.5, + "learning_rate": 4.172928956142601e-05, + "loss": 0.6498, + "step": 12949 + }, + { + "epoch": 0.5, + "learning_rate": 4.1728648371377276e-05, + "loss": 0.5761, + "step": 12950 + }, + { + "epoch": 0.5, + "learning_rate": 4.172800718132855e-05, + "loss": 0.5533, + "step": 12951 + }, + { + "epoch": 0.5, + "learning_rate": 4.172736599127981e-05, + "loss": 0.5736, + "step": 12952 + }, + { + "epoch": 0.5, + "learning_rate": 4.1726724801231084e-05, + "loss": 0.4485, + "step": 12953 + }, + { + "epoch": 0.5, + "learning_rate": 4.1726083611182356e-05, + "loss": 0.4917, + "step": 12954 + }, + { + "epoch": 0.5, + "learning_rate": 4.172544242113363e-05, + "loss": 0.5596, + "step": 12955 + }, + { + "epoch": 0.5, + "learning_rate": 4.17248012310849e-05, + "loss": 0.4968, + "step": 12956 + }, + { + "epoch": 0.5, + "learning_rate": 4.1724160041036165e-05, + "loss": 0.4793, + "step": 12957 + }, + { + "epoch": 0.5, + "learning_rate": 4.1723518850987436e-05, + "loss": 0.5781, + "step": 12958 + }, + { + "epoch": 0.5, + "learning_rate": 4.17228776609387e-05, + "loss": 0.5408, + "step": 12959 + }, + { + "epoch": 0.5, + "learning_rate": 4.172223647088997e-05, + "loss": 0.4407, + "step": 12960 + }, + { + "epoch": 0.5, + "learning_rate": 4.1721595280841245e-05, + "loss": 0.5137, + "step": 12961 + }, + { + "epoch": 0.5, + "learning_rate": 4.1720954090792516e-05, + "loss": 0.5081, + "step": 12962 + }, + { + "epoch": 0.5, + "learning_rate": 4.172031290074378e-05, + "loss": 0.6407, + "step": 12963 + }, + { + "epoch": 0.5, + "learning_rate": 4.171967171069505e-05, + "loss": 0.53, + "step": 12964 + }, + { + "epoch": 0.5, + "learning_rate": 4.1719030520646325e-05, + "loss": 0.6911, + "step": 12965 + }, + { + "epoch": 0.5, + "learning_rate": 4.171838933059759e-05, + "loss": 0.503, + "step": 12966 + }, + { + "epoch": 0.5, + "learning_rate": 4.171774814054886e-05, + "loss": 0.5731, + "step": 12967 + }, + { + "epoch": 0.5, + "learning_rate": 4.1717106950500126e-05, + "loss": 0.4614, + "step": 12968 + }, + { + "epoch": 0.5, + "learning_rate": 4.17164657604514e-05, + "loss": 0.4651, + "step": 12969 + }, + { + "epoch": 0.5, + "learning_rate": 4.171582457040267e-05, + "loss": 0.5203, + "step": 12970 + }, + { + "epoch": 0.5, + "learning_rate": 4.171518338035394e-05, + "loss": 0.6492, + "step": 12971 + }, + { + "epoch": 0.5, + "learning_rate": 4.1714542190305206e-05, + "loss": 0.5071, + "step": 12972 + }, + { + "epoch": 0.5, + "learning_rate": 4.171390100025648e-05, + "loss": 0.5039, + "step": 12973 + }, + { + "epoch": 0.5, + "learning_rate": 4.171325981020775e-05, + "loss": 0.4577, + "step": 12974 + }, + { + "epoch": 0.5, + "learning_rate": 4.1712618620159015e-05, + "loss": 0.4899, + "step": 12975 + }, + { + "epoch": 0.5, + "learning_rate": 4.1711977430110286e-05, + "loss": 0.5474, + "step": 12976 + }, + { + "epoch": 0.5, + "learning_rate": 4.171133624006156e-05, + "loss": 0.531, + "step": 12977 + }, + { + "epoch": 0.5, + "learning_rate": 4.171069505001283e-05, + "loss": 0.5222, + "step": 12978 + }, + { + "epoch": 0.5, + "learning_rate": 4.1710053859964095e-05, + "loss": 0.6202, + "step": 12979 + }, + { + "epoch": 0.5, + "learning_rate": 4.1709412669915367e-05, + "loss": 0.4873, + "step": 12980 + }, + { + "epoch": 0.5, + "learning_rate": 4.170877147986663e-05, + "loss": 0.5297, + "step": 12981 + }, + { + "epoch": 0.5, + "learning_rate": 4.17081302898179e-05, + "loss": 0.5618, + "step": 12982 + }, + { + "epoch": 0.5, + "learning_rate": 4.1707489099769175e-05, + "loss": 0.6336, + "step": 12983 + }, + { + "epoch": 0.5, + "learning_rate": 4.170684790972044e-05, + "loss": 0.4993, + "step": 12984 + }, + { + "epoch": 0.5, + "learning_rate": 4.170620671967171e-05, + "loss": 0.611, + "step": 12985 + }, + { + "epoch": 0.5, + "learning_rate": 4.170556552962298e-05, + "loss": 0.5657, + "step": 12986 + }, + { + "epoch": 0.5, + "learning_rate": 4.1704924339574255e-05, + "loss": 0.6416, + "step": 12987 + }, + { + "epoch": 0.5, + "learning_rate": 4.170428314952552e-05, + "loss": 0.6594, + "step": 12988 + }, + { + "epoch": 0.5, + "learning_rate": 4.170364195947679e-05, + "loss": 0.5135, + "step": 12989 + }, + { + "epoch": 0.5, + "learning_rate": 4.170300076942806e-05, + "loss": 0.4339, + "step": 12990 + }, + { + "epoch": 0.5, + "learning_rate": 4.170235957937933e-05, + "loss": 0.5104, + "step": 12991 + }, + { + "epoch": 0.5, + "learning_rate": 4.17017183893306e-05, + "loss": 0.554, + "step": 12992 + }, + { + "epoch": 0.5, + "learning_rate": 4.1701077199281865e-05, + "loss": 0.6675, + "step": 12993 + }, + { + "epoch": 0.5, + "learning_rate": 4.1700436009233144e-05, + "loss": 0.5739, + "step": 12994 + }, + { + "epoch": 0.5, + "learning_rate": 4.169979481918441e-05, + "loss": 0.4475, + "step": 12995 + }, + { + "epoch": 0.5, + "learning_rate": 4.169915362913568e-05, + "loss": 0.5406, + "step": 12996 + }, + { + "epoch": 0.5, + "learning_rate": 4.1698512439086945e-05, + "loss": 0.5068, + "step": 12997 + }, + { + "epoch": 0.5, + "learning_rate": 4.169787124903822e-05, + "loss": 0.5942, + "step": 12998 + }, + { + "epoch": 0.5, + "learning_rate": 4.169723005898948e-05, + "loss": 0.5443, + "step": 12999 + }, + { + "epoch": 0.5, + "learning_rate": 4.1696588868940753e-05, + "loss": 0.7364, + "step": 13000 + }, + { + "epoch": 0.5, + "learning_rate": 4.1695947678892025e-05, + "loss": 0.5459, + "step": 13001 + }, + { + "epoch": 0.5, + "learning_rate": 4.16953064888433e-05, + "loss": 0.5616, + "step": 13002 + }, + { + "epoch": 0.5, + "learning_rate": 4.169466529879457e-05, + "loss": 0.5753, + "step": 13003 + }, + { + "epoch": 0.5, + "learning_rate": 4.1694024108745834e-05, + "loss": 0.5252, + "step": 13004 + }, + { + "epoch": 0.5, + "learning_rate": 4.1693382918697105e-05, + "loss": 0.6536, + "step": 13005 + }, + { + "epoch": 0.5, + "learning_rate": 4.169274172864837e-05, + "loss": 0.5211, + "step": 13006 + }, + { + "epoch": 0.5, + "learning_rate": 4.169210053859964e-05, + "loss": 0.4824, + "step": 13007 + }, + { + "epoch": 0.5, + "learning_rate": 4.169145934855091e-05, + "loss": 0.5679, + "step": 13008 + }, + { + "epoch": 0.5, + "learning_rate": 4.169081815850218e-05, + "loss": 0.5993, + "step": 13009 + }, + { + "epoch": 0.5, + "learning_rate": 4.169017696845346e-05, + "loss": 0.5617, + "step": 13010 + }, + { + "epoch": 0.5, + "learning_rate": 4.168953577840472e-05, + "loss": 0.6223, + "step": 13011 + }, + { + "epoch": 0.5, + "learning_rate": 4.1688894588355994e-05, + "loss": 0.5891, + "step": 13012 + }, + { + "epoch": 0.5, + "learning_rate": 4.168825339830726e-05, + "loss": 0.4513, + "step": 13013 + }, + { + "epoch": 0.5, + "learning_rate": 4.168761220825853e-05, + "loss": 0.4068, + "step": 13014 + }, + { + "epoch": 0.5, + "learning_rate": 4.1686971018209795e-05, + "loss": 0.5536, + "step": 13015 + }, + { + "epoch": 0.5, + "learning_rate": 4.168632982816107e-05, + "loss": 0.5315, + "step": 13016 + }, + { + "epoch": 0.5, + "learning_rate": 4.168568863811234e-05, + "loss": 0.5487, + "step": 13017 + }, + { + "epoch": 0.5, + "learning_rate": 4.168504744806361e-05, + "loss": 0.4771, + "step": 13018 + }, + { + "epoch": 0.5, + "learning_rate": 4.168440625801488e-05, + "loss": 0.6794, + "step": 13019 + }, + { + "epoch": 0.5, + "learning_rate": 4.168376506796615e-05, + "loss": 0.5625, + "step": 13020 + }, + { + "epoch": 0.5, + "learning_rate": 4.168312387791742e-05, + "loss": 0.6859, + "step": 13021 + }, + { + "epoch": 0.5, + "learning_rate": 4.1682482687868684e-05, + "loss": 0.4762, + "step": 13022 + }, + { + "epoch": 0.5, + "learning_rate": 4.1681841497819956e-05, + "loss": 0.5818, + "step": 13023 + }, + { + "epoch": 0.5, + "learning_rate": 4.168120030777122e-05, + "loss": 0.6515, + "step": 13024 + }, + { + "epoch": 0.5, + "learning_rate": 4.168055911772249e-05, + "loss": 0.5097, + "step": 13025 + }, + { + "epoch": 0.5, + "learning_rate": 4.1679917927673764e-05, + "loss": 0.605, + "step": 13026 + }, + { + "epoch": 0.5, + "learning_rate": 4.1679276737625036e-05, + "loss": 0.593, + "step": 13027 + }, + { + "epoch": 0.5, + "learning_rate": 4.167863554757631e-05, + "loss": 0.5103, + "step": 13028 + }, + { + "epoch": 0.5, + "learning_rate": 4.167799435752757e-05, + "loss": 0.5142, + "step": 13029 + }, + { + "epoch": 0.5, + "learning_rate": 4.1677353167478844e-05, + "loss": 0.6449, + "step": 13030 + }, + { + "epoch": 0.5, + "learning_rate": 4.167671197743011e-05, + "loss": 0.4398, + "step": 13031 + }, + { + "epoch": 0.5, + "learning_rate": 4.167607078738138e-05, + "loss": 0.6279, + "step": 13032 + }, + { + "epoch": 0.5, + "learning_rate": 4.167542959733265e-05, + "loss": 0.5675, + "step": 13033 + }, + { + "epoch": 0.5, + "learning_rate": 4.1674788407283924e-05, + "loss": 0.628, + "step": 13034 + }, + { + "epoch": 0.5, + "learning_rate": 4.1674147217235196e-05, + "loss": 0.5997, + "step": 13035 + }, + { + "epoch": 0.5, + "learning_rate": 4.167350602718646e-05, + "loss": 0.5586, + "step": 13036 + }, + { + "epoch": 0.5, + "learning_rate": 4.167286483713773e-05, + "loss": 0.5015, + "step": 13037 + }, + { + "epoch": 0.5, + "learning_rate": 4.1672223647089e-05, + "loss": 0.6281, + "step": 13038 + }, + { + "epoch": 0.5, + "learning_rate": 4.167158245704027e-05, + "loss": 0.5382, + "step": 13039 + }, + { + "epoch": 0.5, + "learning_rate": 4.1670941266991534e-05, + "loss": 0.6194, + "step": 13040 + }, + { + "epoch": 0.5, + "learning_rate": 4.1670300076942806e-05, + "loss": 0.4788, + "step": 13041 + }, + { + "epoch": 0.5, + "learning_rate": 4.166965888689408e-05, + "loss": 0.506, + "step": 13042 + }, + { + "epoch": 0.5, + "learning_rate": 4.166901769684535e-05, + "loss": 0.5788, + "step": 13043 + }, + { + "epoch": 0.5, + "learning_rate": 4.166837650679662e-05, + "loss": 0.5648, + "step": 13044 + }, + { + "epoch": 0.5, + "learning_rate": 4.1667735316747886e-05, + "loss": 0.5587, + "step": 13045 + }, + { + "epoch": 0.5, + "learning_rate": 4.166709412669916e-05, + "loss": 0.573, + "step": 13046 + }, + { + "epoch": 0.5, + "learning_rate": 4.166645293665042e-05, + "loss": 0.5958, + "step": 13047 + }, + { + "epoch": 0.5, + "learning_rate": 4.1665811746601694e-05, + "loss": 0.5842, + "step": 13048 + }, + { + "epoch": 0.5, + "learning_rate": 4.1665170556552966e-05, + "loss": 0.5646, + "step": 13049 + }, + { + "epoch": 0.5, + "learning_rate": 4.166452936650424e-05, + "loss": 0.5496, + "step": 13050 + }, + { + "epoch": 0.5, + "learning_rate": 4.16638881764555e-05, + "loss": 0.5847, + "step": 13051 + }, + { + "epoch": 0.5, + "learning_rate": 4.1663246986406774e-05, + "loss": 0.5111, + "step": 13052 + }, + { + "epoch": 0.5, + "learning_rate": 4.1662605796358046e-05, + "loss": 0.5017, + "step": 13053 + }, + { + "epoch": 0.5, + "learning_rate": 4.166196460630931e-05, + "loss": 0.5624, + "step": 13054 + }, + { + "epoch": 0.5, + "learning_rate": 4.166132341626058e-05, + "loss": 0.4698, + "step": 13055 + }, + { + "epoch": 0.5, + "learning_rate": 4.166068222621185e-05, + "loss": 0.4701, + "step": 13056 + }, + { + "epoch": 0.5, + "learning_rate": 4.166004103616312e-05, + "loss": 0.5216, + "step": 13057 + }, + { + "epoch": 0.5, + "learning_rate": 4.165939984611439e-05, + "loss": 0.5629, + "step": 13058 + }, + { + "epoch": 0.5, + "learning_rate": 4.165875865606566e-05, + "loss": 0.6451, + "step": 13059 + }, + { + "epoch": 0.5, + "learning_rate": 4.165811746601693e-05, + "loss": 0.4826, + "step": 13060 + }, + { + "epoch": 0.5, + "learning_rate": 4.16574762759682e-05, + "loss": 0.6196, + "step": 13061 + }, + { + "epoch": 0.5, + "learning_rate": 4.165683508591947e-05, + "loss": 0.6359, + "step": 13062 + }, + { + "epoch": 0.5, + "learning_rate": 4.1656193895870736e-05, + "loss": 0.5082, + "step": 13063 + }, + { + "epoch": 0.5, + "learning_rate": 4.165555270582201e-05, + "loss": 0.4867, + "step": 13064 + }, + { + "epoch": 0.5, + "learning_rate": 4.165491151577327e-05, + "loss": 0.5397, + "step": 13065 + }, + { + "epoch": 0.5, + "learning_rate": 4.165427032572455e-05, + "loss": 0.5618, + "step": 13066 + }, + { + "epoch": 0.5, + "learning_rate": 4.1653629135675816e-05, + "loss": 0.6416, + "step": 13067 + }, + { + "epoch": 0.5, + "learning_rate": 4.165298794562709e-05, + "loss": 0.4654, + "step": 13068 + }, + { + "epoch": 0.5, + "learning_rate": 4.165234675557835e-05, + "loss": 0.6199, + "step": 13069 + }, + { + "epoch": 0.5, + "learning_rate": 4.1651705565529625e-05, + "loss": 0.4616, + "step": 13070 + }, + { + "epoch": 0.5, + "learning_rate": 4.1651064375480896e-05, + "loss": 0.6444, + "step": 13071 + }, + { + "epoch": 0.5, + "learning_rate": 4.165042318543216e-05, + "loss": 0.5947, + "step": 13072 + }, + { + "epoch": 0.5, + "learning_rate": 4.164978199538343e-05, + "loss": 0.6258, + "step": 13073 + }, + { + "epoch": 0.5, + "learning_rate": 4.1649140805334705e-05, + "loss": 0.5212, + "step": 13074 + }, + { + "epoch": 0.5, + "learning_rate": 4.1648499615285976e-05, + "loss": 0.5615, + "step": 13075 + }, + { + "epoch": 0.5, + "learning_rate": 4.164785842523724e-05, + "loss": 0.4811, + "step": 13076 + }, + { + "epoch": 0.5, + "learning_rate": 4.164721723518851e-05, + "loss": 0.5466, + "step": 13077 + }, + { + "epoch": 0.5, + "learning_rate": 4.164657604513978e-05, + "loss": 0.5795, + "step": 13078 + }, + { + "epoch": 0.5, + "learning_rate": 4.164593485509105e-05, + "loss": 0.636, + "step": 13079 + }, + { + "epoch": 0.5, + "learning_rate": 4.164529366504232e-05, + "loss": 0.4425, + "step": 13080 + }, + { + "epoch": 0.5, + "learning_rate": 4.1644652474993586e-05, + "loss": 0.5212, + "step": 13081 + }, + { + "epoch": 0.5, + "learning_rate": 4.1644011284944865e-05, + "loss": 0.5164, + "step": 13082 + }, + { + "epoch": 0.5, + "learning_rate": 4.164337009489613e-05, + "loss": 0.5415, + "step": 13083 + }, + { + "epoch": 0.5, + "learning_rate": 4.16427289048474e-05, + "loss": 0.5345, + "step": 13084 + }, + { + "epoch": 0.5, + "learning_rate": 4.1642087714798666e-05, + "loss": 0.4879, + "step": 13085 + }, + { + "epoch": 0.5, + "learning_rate": 4.164144652474994e-05, + "loss": 0.4452, + "step": 13086 + }, + { + "epoch": 0.5, + "learning_rate": 4.16408053347012e-05, + "loss": 0.492, + "step": 13087 + }, + { + "epoch": 0.5, + "learning_rate": 4.1640164144652475e-05, + "loss": 0.6486, + "step": 13088 + }, + { + "epoch": 0.5, + "learning_rate": 4.1639522954603747e-05, + "loss": 0.5373, + "step": 13089 + }, + { + "epoch": 0.5, + "learning_rate": 4.163888176455502e-05, + "loss": 0.4852, + "step": 13090 + }, + { + "epoch": 0.5, + "learning_rate": 4.163824057450629e-05, + "loss": 0.4556, + "step": 13091 + }, + { + "epoch": 0.5, + "learning_rate": 4.1637599384457555e-05, + "loss": 0.493, + "step": 13092 + }, + { + "epoch": 0.5, + "learning_rate": 4.1636958194408827e-05, + "loss": 0.5973, + "step": 13093 + }, + { + "epoch": 0.5, + "learning_rate": 4.163631700436009e-05, + "loss": 0.5214, + "step": 13094 + }, + { + "epoch": 0.5, + "learning_rate": 4.163567581431136e-05, + "loss": 0.5008, + "step": 13095 + }, + { + "epoch": 0.5, + "learning_rate": 4.163503462426263e-05, + "loss": 0.5595, + "step": 13096 + }, + { + "epoch": 0.5, + "learning_rate": 4.16343934342139e-05, + "loss": 0.6055, + "step": 13097 + }, + { + "epoch": 0.5, + "learning_rate": 4.163375224416518e-05, + "loss": 0.516, + "step": 13098 + }, + { + "epoch": 0.5, + "learning_rate": 4.163311105411644e-05, + "loss": 0.5879, + "step": 13099 + }, + { + "epoch": 0.5, + "learning_rate": 4.1632469864067715e-05, + "loss": 0.6319, + "step": 13100 + }, + { + "epoch": 0.5, + "learning_rate": 4.163182867401898e-05, + "loss": 0.4987, + "step": 13101 + }, + { + "epoch": 0.5, + "learning_rate": 4.163118748397025e-05, + "loss": 0.5232, + "step": 13102 + }, + { + "epoch": 0.5, + "learning_rate": 4.163054629392152e-05, + "loss": 0.5983, + "step": 13103 + }, + { + "epoch": 0.5, + "learning_rate": 4.162990510387279e-05, + "loss": 0.6152, + "step": 13104 + }, + { + "epoch": 0.5, + "learning_rate": 4.162926391382406e-05, + "loss": 0.5112, + "step": 13105 + }, + { + "epoch": 0.5, + "learning_rate": 4.162862272377533e-05, + "loss": 0.5469, + "step": 13106 + }, + { + "epoch": 0.5, + "learning_rate": 4.1627981533726604e-05, + "loss": 0.4329, + "step": 13107 + }, + { + "epoch": 0.5, + "learning_rate": 4.162734034367787e-05, + "loss": 0.7064, + "step": 13108 + }, + { + "epoch": 0.5, + "learning_rate": 4.162669915362914e-05, + "loss": 0.5475, + "step": 13109 + }, + { + "epoch": 0.5, + "learning_rate": 4.1626057963580405e-05, + "loss": 0.5503, + "step": 13110 + }, + { + "epoch": 0.5, + "learning_rate": 4.162541677353168e-05, + "loss": 0.56, + "step": 13111 + }, + { + "epoch": 0.5, + "learning_rate": 4.162477558348294e-05, + "loss": 0.5271, + "step": 13112 + }, + { + "epoch": 0.5, + "learning_rate": 4.1624134393434214e-05, + "loss": 0.4435, + "step": 13113 + }, + { + "epoch": 0.5, + "learning_rate": 4.1623493203385485e-05, + "loss": 0.4521, + "step": 13114 + }, + { + "epoch": 0.5, + "learning_rate": 4.162285201333676e-05, + "loss": 0.5289, + "step": 13115 + }, + { + "epoch": 0.5, + "learning_rate": 4.162221082328803e-05, + "loss": 0.5181, + "step": 13116 + }, + { + "epoch": 0.5, + "learning_rate": 4.1621569633239294e-05, + "loss": 0.5076, + "step": 13117 + }, + { + "epoch": 0.5, + "learning_rate": 4.1620928443190565e-05, + "loss": 0.6174, + "step": 13118 + }, + { + "epoch": 0.5, + "learning_rate": 4.162028725314183e-05, + "loss": 0.7447, + "step": 13119 + }, + { + "epoch": 0.5, + "learning_rate": 4.16196460630931e-05, + "loss": 0.5253, + "step": 13120 + }, + { + "epoch": 0.5, + "learning_rate": 4.1619004873044374e-05, + "loss": 0.573, + "step": 13121 + }, + { + "epoch": 0.5, + "learning_rate": 4.1618363682995645e-05, + "loss": 0.6337, + "step": 13122 + }, + { + "epoch": 0.5, + "learning_rate": 4.161772249294691e-05, + "loss": 0.6291, + "step": 13123 + }, + { + "epoch": 0.5, + "learning_rate": 4.161708130289818e-05, + "loss": 0.5717, + "step": 13124 + }, + { + "epoch": 0.5, + "learning_rate": 4.1616440112849454e-05, + "loss": 0.5883, + "step": 13125 + }, + { + "epoch": 0.5, + "learning_rate": 4.161579892280072e-05, + "loss": 0.5968, + "step": 13126 + }, + { + "epoch": 0.5, + "learning_rate": 4.161515773275199e-05, + "loss": 0.5076, + "step": 13127 + }, + { + "epoch": 0.5, + "learning_rate": 4.1614516542703255e-05, + "loss": 0.5595, + "step": 13128 + }, + { + "epoch": 0.5, + "learning_rate": 4.161387535265453e-05, + "loss": 0.5011, + "step": 13129 + }, + { + "epoch": 0.5, + "learning_rate": 4.16132341626058e-05, + "loss": 0.5863, + "step": 13130 + }, + { + "epoch": 0.5, + "learning_rate": 4.161259297255707e-05, + "loss": 0.6104, + "step": 13131 + }, + { + "epoch": 0.5, + "learning_rate": 4.1611951782508335e-05, + "loss": 0.5007, + "step": 13132 + }, + { + "epoch": 0.5, + "learning_rate": 4.161131059245961e-05, + "loss": 0.5462, + "step": 13133 + }, + { + "epoch": 0.5, + "learning_rate": 4.161066940241088e-05, + "loss": 0.6064, + "step": 13134 + }, + { + "epoch": 0.5, + "learning_rate": 4.1610028212362144e-05, + "loss": 0.4944, + "step": 13135 + }, + { + "epoch": 0.51, + "learning_rate": 4.1609387022313416e-05, + "loss": 0.59, + "step": 13136 + }, + { + "epoch": 0.51, + "learning_rate": 4.160874583226469e-05, + "loss": 0.5402, + "step": 13137 + }, + { + "epoch": 0.51, + "learning_rate": 4.160810464221596e-05, + "loss": 0.5326, + "step": 13138 + }, + { + "epoch": 0.51, + "learning_rate": 4.1607463452167224e-05, + "loss": 0.4514, + "step": 13139 + }, + { + "epoch": 0.51, + "learning_rate": 4.1606822262118496e-05, + "loss": 0.5482, + "step": 13140 + }, + { + "epoch": 0.51, + "learning_rate": 4.160618107206976e-05, + "loss": 0.569, + "step": 13141 + }, + { + "epoch": 0.51, + "learning_rate": 4.160553988202103e-05, + "loss": 0.4744, + "step": 13142 + }, + { + "epoch": 0.51, + "learning_rate": 4.1604898691972304e-05, + "loss": 0.5793, + "step": 13143 + }, + { + "epoch": 0.51, + "learning_rate": 4.160425750192357e-05, + "loss": 0.7002, + "step": 13144 + }, + { + "epoch": 0.51, + "learning_rate": 4.160361631187484e-05, + "loss": 0.4753, + "step": 13145 + }, + { + "epoch": 0.51, + "learning_rate": 4.160297512182611e-05, + "loss": 0.5583, + "step": 13146 + }, + { + "epoch": 0.51, + "learning_rate": 4.1602333931777384e-05, + "loss": 0.5245, + "step": 13147 + }, + { + "epoch": 0.51, + "learning_rate": 4.160169274172865e-05, + "loss": 0.5371, + "step": 13148 + }, + { + "epoch": 0.51, + "learning_rate": 4.160105155167992e-05, + "loss": 0.5542, + "step": 13149 + }, + { + "epoch": 0.51, + "learning_rate": 4.1600410361631186e-05, + "loss": 0.5228, + "step": 13150 + }, + { + "epoch": 0.51, + "learning_rate": 4.159976917158246e-05, + "loss": 0.6298, + "step": 13151 + }, + { + "epoch": 0.51, + "learning_rate": 4.159912798153373e-05, + "loss": 0.5225, + "step": 13152 + }, + { + "epoch": 0.51, + "learning_rate": 4.1598486791484994e-05, + "loss": 0.6283, + "step": 13153 + }, + { + "epoch": 0.51, + "learning_rate": 4.159784560143627e-05, + "loss": 0.4784, + "step": 13154 + }, + { + "epoch": 0.51, + "learning_rate": 4.159720441138754e-05, + "loss": 0.5255, + "step": 13155 + }, + { + "epoch": 0.51, + "learning_rate": 4.159656322133881e-05, + "loss": 0.5424, + "step": 13156 + }, + { + "epoch": 0.51, + "learning_rate": 4.1595922031290074e-05, + "loss": 0.5824, + "step": 13157 + }, + { + "epoch": 0.51, + "learning_rate": 4.1595280841241346e-05, + "loss": 0.4859, + "step": 13158 + }, + { + "epoch": 0.51, + "learning_rate": 4.159463965119261e-05, + "loss": 0.5587, + "step": 13159 + }, + { + "epoch": 0.51, + "learning_rate": 4.159399846114388e-05, + "loss": 0.5216, + "step": 13160 + }, + { + "epoch": 0.51, + "learning_rate": 4.1593357271095154e-05, + "loss": 0.6098, + "step": 13161 + }, + { + "epoch": 0.51, + "learning_rate": 4.1592716081046426e-05, + "loss": 0.6317, + "step": 13162 + }, + { + "epoch": 0.51, + "learning_rate": 4.15920748909977e-05, + "loss": 0.447, + "step": 13163 + }, + { + "epoch": 0.51, + "learning_rate": 4.159143370094896e-05, + "loss": 0.5823, + "step": 13164 + }, + { + "epoch": 0.51, + "learning_rate": 4.1590792510900234e-05, + "loss": 0.5423, + "step": 13165 + }, + { + "epoch": 0.51, + "learning_rate": 4.15901513208515e-05, + "loss": 0.6286, + "step": 13166 + }, + { + "epoch": 0.51, + "learning_rate": 4.158951013080277e-05, + "loss": 0.4182, + "step": 13167 + }, + { + "epoch": 0.51, + "learning_rate": 4.1588868940754036e-05, + "loss": 0.5793, + "step": 13168 + }, + { + "epoch": 0.51, + "learning_rate": 4.158822775070531e-05, + "loss": 0.4317, + "step": 13169 + }, + { + "epoch": 0.51, + "learning_rate": 4.1587586560656586e-05, + "loss": 0.5042, + "step": 13170 + }, + { + "epoch": 0.51, + "learning_rate": 4.158694537060785e-05, + "loss": 0.5399, + "step": 13171 + }, + { + "epoch": 0.51, + "learning_rate": 4.158630418055912e-05, + "loss": 0.4732, + "step": 13172 + }, + { + "epoch": 0.51, + "learning_rate": 4.158566299051039e-05, + "loss": 0.5525, + "step": 13173 + }, + { + "epoch": 0.51, + "learning_rate": 4.158502180046166e-05, + "loss": 0.5815, + "step": 13174 + }, + { + "epoch": 0.51, + "learning_rate": 4.1584380610412924e-05, + "loss": 0.5488, + "step": 13175 + }, + { + "epoch": 0.51, + "learning_rate": 4.1583739420364196e-05, + "loss": 0.4538, + "step": 13176 + }, + { + "epoch": 0.51, + "learning_rate": 4.158309823031547e-05, + "loss": 0.5767, + "step": 13177 + }, + { + "epoch": 0.51, + "learning_rate": 4.158245704026674e-05, + "loss": 0.4654, + "step": 13178 + }, + { + "epoch": 0.51, + "learning_rate": 4.158181585021801e-05, + "loss": 0.5615, + "step": 13179 + }, + { + "epoch": 0.51, + "learning_rate": 4.1581174660169276e-05, + "loss": 0.6393, + "step": 13180 + }, + { + "epoch": 0.51, + "learning_rate": 4.158053347012055e-05, + "loss": 0.5185, + "step": 13181 + }, + { + "epoch": 0.51, + "learning_rate": 4.157989228007181e-05, + "loss": 0.5808, + "step": 13182 + }, + { + "epoch": 0.51, + "learning_rate": 4.1579251090023085e-05, + "loss": 0.5025, + "step": 13183 + }, + { + "epoch": 0.51, + "learning_rate": 4.157860989997435e-05, + "loss": 0.6408, + "step": 13184 + }, + { + "epoch": 0.51, + "learning_rate": 4.157796870992562e-05, + "loss": 0.582, + "step": 13185 + }, + { + "epoch": 0.51, + "learning_rate": 4.157732751987689e-05, + "loss": 0.6433, + "step": 13186 + }, + { + "epoch": 0.51, + "learning_rate": 4.1576686329828165e-05, + "loss": 0.5272, + "step": 13187 + }, + { + "epoch": 0.51, + "learning_rate": 4.1576045139779436e-05, + "loss": 0.5881, + "step": 13188 + }, + { + "epoch": 0.51, + "learning_rate": 4.15754039497307e-05, + "loss": 0.5043, + "step": 13189 + }, + { + "epoch": 0.51, + "learning_rate": 4.157476275968197e-05, + "loss": 0.6382, + "step": 13190 + }, + { + "epoch": 0.51, + "learning_rate": 4.157412156963324e-05, + "loss": 0.4768, + "step": 13191 + }, + { + "epoch": 0.51, + "learning_rate": 4.157348037958451e-05, + "loss": 0.5976, + "step": 13192 + }, + { + "epoch": 0.51, + "learning_rate": 4.157283918953578e-05, + "loss": 0.5001, + "step": 13193 + }, + { + "epoch": 0.51, + "learning_rate": 4.157219799948705e-05, + "loss": 0.5361, + "step": 13194 + }, + { + "epoch": 0.51, + "learning_rate": 4.157155680943832e-05, + "loss": 0.5319, + "step": 13195 + }, + { + "epoch": 0.51, + "learning_rate": 4.157091561938959e-05, + "loss": 0.4934, + "step": 13196 + }, + { + "epoch": 0.51, + "learning_rate": 4.157027442934086e-05, + "loss": 0.5428, + "step": 13197 + }, + { + "epoch": 0.51, + "learning_rate": 4.1569633239292126e-05, + "loss": 0.4729, + "step": 13198 + }, + { + "epoch": 0.51, + "learning_rate": 4.15689920492434e-05, + "loss": 0.4842, + "step": 13199 + }, + { + "epoch": 0.51, + "learning_rate": 4.156835085919466e-05, + "loss": 0.5954, + "step": 13200 + }, + { + "epoch": 0.51, + "learning_rate": 4.1567709669145935e-05, + "loss": 0.5759, + "step": 13201 + }, + { + "epoch": 0.51, + "learning_rate": 4.1567068479097207e-05, + "loss": 0.5118, + "step": 13202 + }, + { + "epoch": 0.51, + "learning_rate": 4.156642728904848e-05, + "loss": 0.4718, + "step": 13203 + }, + { + "epoch": 0.51, + "learning_rate": 4.156578609899974e-05, + "loss": 0.5186, + "step": 13204 + }, + { + "epoch": 0.51, + "learning_rate": 4.1565144908951015e-05, + "loss": 0.5312, + "step": 13205 + }, + { + "epoch": 0.51, + "learning_rate": 4.156450371890229e-05, + "loss": 0.5364, + "step": 13206 + }, + { + "epoch": 0.51, + "learning_rate": 4.156386252885355e-05, + "loss": 0.5479, + "step": 13207 + }, + { + "epoch": 0.51, + "learning_rate": 4.156322133880482e-05, + "loss": 0.6735, + "step": 13208 + }, + { + "epoch": 0.51, + "learning_rate": 4.1562580148756095e-05, + "loss": 0.4888, + "step": 13209 + }, + { + "epoch": 0.51, + "learning_rate": 4.156193895870737e-05, + "loss": 0.5341, + "step": 13210 + }, + { + "epoch": 0.51, + "learning_rate": 4.156129776865863e-05, + "loss": 0.5603, + "step": 13211 + }, + { + "epoch": 0.51, + "learning_rate": 4.1560656578609903e-05, + "loss": 0.4675, + "step": 13212 + }, + { + "epoch": 0.51, + "learning_rate": 4.1560015388561175e-05, + "loss": 0.5244, + "step": 13213 + }, + { + "epoch": 0.51, + "learning_rate": 4.155937419851244e-05, + "loss": 0.6002, + "step": 13214 + }, + { + "epoch": 0.51, + "learning_rate": 4.155873300846371e-05, + "loss": 0.5581, + "step": 13215 + }, + { + "epoch": 0.51, + "learning_rate": 4.155809181841498e-05, + "loss": 0.5165, + "step": 13216 + }, + { + "epoch": 0.51, + "learning_rate": 4.155745062836625e-05, + "loss": 0.6209, + "step": 13217 + }, + { + "epoch": 0.51, + "learning_rate": 4.155680943831752e-05, + "loss": 0.5122, + "step": 13218 + }, + { + "epoch": 0.51, + "learning_rate": 4.155616824826879e-05, + "loss": 0.5112, + "step": 13219 + }, + { + "epoch": 0.51, + "learning_rate": 4.155552705822006e-05, + "loss": 0.529, + "step": 13220 + }, + { + "epoch": 0.51, + "learning_rate": 4.155488586817133e-05, + "loss": 0.5325, + "step": 13221 + }, + { + "epoch": 0.51, + "learning_rate": 4.15542446781226e-05, + "loss": 0.5902, + "step": 13222 + }, + { + "epoch": 0.51, + "learning_rate": 4.1553603488073865e-05, + "loss": 0.6792, + "step": 13223 + }, + { + "epoch": 0.51, + "learning_rate": 4.155296229802514e-05, + "loss": 0.6788, + "step": 13224 + }, + { + "epoch": 0.51, + "learning_rate": 4.15523211079764e-05, + "loss": 0.5995, + "step": 13225 + }, + { + "epoch": 0.51, + "learning_rate": 4.155167991792768e-05, + "loss": 0.4769, + "step": 13226 + }, + { + "epoch": 0.51, + "learning_rate": 4.1551038727878945e-05, + "loss": 0.5272, + "step": 13227 + }, + { + "epoch": 0.51, + "learning_rate": 4.155039753783022e-05, + "loss": 0.631, + "step": 13228 + }, + { + "epoch": 0.51, + "learning_rate": 4.154975634778148e-05, + "loss": 0.5645, + "step": 13229 + }, + { + "epoch": 0.51, + "learning_rate": 4.1549115157732754e-05, + "loss": 0.4865, + "step": 13230 + }, + { + "epoch": 0.51, + "learning_rate": 4.1548473967684025e-05, + "loss": 0.6237, + "step": 13231 + }, + { + "epoch": 0.51, + "learning_rate": 4.154783277763529e-05, + "loss": 0.5596, + "step": 13232 + }, + { + "epoch": 0.51, + "learning_rate": 4.154719158758656e-05, + "loss": 0.5631, + "step": 13233 + }, + { + "epoch": 0.51, + "learning_rate": 4.1546550397537834e-05, + "loss": 0.5487, + "step": 13234 + }, + { + "epoch": 0.51, + "learning_rate": 4.1545909207489105e-05, + "loss": 0.524, + "step": 13235 + }, + { + "epoch": 0.51, + "learning_rate": 4.154526801744037e-05, + "loss": 0.6427, + "step": 13236 + }, + { + "epoch": 0.51, + "learning_rate": 4.154462682739164e-05, + "loss": 0.6846, + "step": 13237 + }, + { + "epoch": 0.51, + "learning_rate": 4.154398563734291e-05, + "loss": 0.6009, + "step": 13238 + }, + { + "epoch": 0.51, + "learning_rate": 4.154334444729418e-05, + "loss": 0.6549, + "step": 13239 + }, + { + "epoch": 0.51, + "learning_rate": 4.154270325724545e-05, + "loss": 0.5728, + "step": 13240 + }, + { + "epoch": 0.51, + "learning_rate": 4.1542062067196715e-05, + "loss": 0.568, + "step": 13241 + }, + { + "epoch": 0.51, + "learning_rate": 4.1541420877147994e-05, + "loss": 0.584, + "step": 13242 + }, + { + "epoch": 0.51, + "learning_rate": 4.154077968709926e-05, + "loss": 0.5776, + "step": 13243 + }, + { + "epoch": 0.51, + "learning_rate": 4.154013849705053e-05, + "loss": 0.6098, + "step": 13244 + }, + { + "epoch": 0.51, + "learning_rate": 4.1539497307001796e-05, + "loss": 0.598, + "step": 13245 + }, + { + "epoch": 0.51, + "learning_rate": 4.153885611695307e-05, + "loss": 0.582, + "step": 13246 + }, + { + "epoch": 0.51, + "learning_rate": 4.153821492690433e-05, + "loss": 0.5371, + "step": 13247 + }, + { + "epoch": 0.51, + "learning_rate": 4.1537573736855604e-05, + "loss": 0.5373, + "step": 13248 + }, + { + "epoch": 0.51, + "learning_rate": 4.1536932546806876e-05, + "loss": 0.5208, + "step": 13249 + }, + { + "epoch": 0.51, + "learning_rate": 4.153629135675815e-05, + "loss": 0.5131, + "step": 13250 + }, + { + "epoch": 0.51, + "learning_rate": 4.153565016670942e-05, + "loss": 0.5789, + "step": 13251 + }, + { + "epoch": 0.51, + "learning_rate": 4.1535008976660684e-05, + "loss": 0.5303, + "step": 13252 + }, + { + "epoch": 0.51, + "learning_rate": 4.1534367786611956e-05, + "loss": 0.5793, + "step": 13253 + }, + { + "epoch": 0.51, + "learning_rate": 4.153372659656322e-05, + "loss": 0.5332, + "step": 13254 + }, + { + "epoch": 0.51, + "learning_rate": 4.153308540651449e-05, + "loss": 0.5626, + "step": 13255 + }, + { + "epoch": 0.51, + "learning_rate": 4.153244421646576e-05, + "loss": 0.5704, + "step": 13256 + }, + { + "epoch": 0.51, + "learning_rate": 4.153180302641703e-05, + "loss": 0.5117, + "step": 13257 + }, + { + "epoch": 0.51, + "learning_rate": 4.153116183636831e-05, + "loss": 0.4769, + "step": 13258 + }, + { + "epoch": 0.51, + "learning_rate": 4.153052064631957e-05, + "loss": 0.5188, + "step": 13259 + }, + { + "epoch": 0.51, + "learning_rate": 4.1529879456270844e-05, + "loss": 0.509, + "step": 13260 + }, + { + "epoch": 0.51, + "learning_rate": 4.152923826622211e-05, + "loss": 0.595, + "step": 13261 + }, + { + "epoch": 0.51, + "learning_rate": 4.152859707617338e-05, + "loss": 0.4577, + "step": 13262 + }, + { + "epoch": 0.51, + "learning_rate": 4.1527955886124646e-05, + "loss": 0.6526, + "step": 13263 + }, + { + "epoch": 0.51, + "learning_rate": 4.152731469607592e-05, + "loss": 0.5901, + "step": 13264 + }, + { + "epoch": 0.51, + "learning_rate": 4.152667350602719e-05, + "loss": 0.6153, + "step": 13265 + }, + { + "epoch": 0.51, + "learning_rate": 4.152603231597846e-05, + "loss": 0.4506, + "step": 13266 + }, + { + "epoch": 0.51, + "learning_rate": 4.152539112592973e-05, + "loss": 0.5161, + "step": 13267 + }, + { + "epoch": 0.51, + "learning_rate": 4.1524749935881e-05, + "loss": 0.5763, + "step": 13268 + }, + { + "epoch": 0.51, + "learning_rate": 4.152410874583227e-05, + "loss": 0.5859, + "step": 13269 + }, + { + "epoch": 0.51, + "learning_rate": 4.1523467555783534e-05, + "loss": 0.5646, + "step": 13270 + }, + { + "epoch": 0.51, + "learning_rate": 4.1522826365734806e-05, + "loss": 0.5223, + "step": 13271 + }, + { + "epoch": 0.51, + "learning_rate": 4.152218517568607e-05, + "loss": 0.6599, + "step": 13272 + }, + { + "epoch": 0.51, + "learning_rate": 4.152154398563734e-05, + "loss": 0.6245, + "step": 13273 + }, + { + "epoch": 0.51, + "learning_rate": 4.1520902795588614e-05, + "loss": 0.623, + "step": 13274 + }, + { + "epoch": 0.51, + "learning_rate": 4.1520261605539886e-05, + "loss": 0.5075, + "step": 13275 + }, + { + "epoch": 0.51, + "learning_rate": 4.151962041549116e-05, + "loss": 0.5332, + "step": 13276 + }, + { + "epoch": 0.51, + "learning_rate": 4.151897922544242e-05, + "loss": 0.6047, + "step": 13277 + }, + { + "epoch": 0.51, + "learning_rate": 4.1518338035393694e-05, + "loss": 0.5597, + "step": 13278 + }, + { + "epoch": 0.51, + "learning_rate": 4.151769684534496e-05, + "loss": 0.5901, + "step": 13279 + }, + { + "epoch": 0.51, + "learning_rate": 4.151705565529623e-05, + "loss": 0.5666, + "step": 13280 + }, + { + "epoch": 0.51, + "learning_rate": 4.15164144652475e-05, + "loss": 0.5182, + "step": 13281 + }, + { + "epoch": 0.51, + "learning_rate": 4.1515773275198774e-05, + "loss": 0.53, + "step": 13282 + }, + { + "epoch": 0.51, + "learning_rate": 4.151513208515004e-05, + "loss": 0.5974, + "step": 13283 + }, + { + "epoch": 0.51, + "learning_rate": 4.151449089510131e-05, + "loss": 0.4799, + "step": 13284 + }, + { + "epoch": 0.51, + "learning_rate": 4.151384970505258e-05, + "loss": 0.5969, + "step": 13285 + }, + { + "epoch": 0.51, + "learning_rate": 4.151320851500385e-05, + "loss": 0.5395, + "step": 13286 + }, + { + "epoch": 0.51, + "learning_rate": 4.151256732495512e-05, + "loss": 0.5897, + "step": 13287 + }, + { + "epoch": 0.51, + "learning_rate": 4.1511926134906384e-05, + "loss": 0.5669, + "step": 13288 + }, + { + "epoch": 0.51, + "learning_rate": 4.1511284944857656e-05, + "loss": 0.5323, + "step": 13289 + }, + { + "epoch": 0.51, + "learning_rate": 4.151064375480893e-05, + "loss": 0.5031, + "step": 13290 + }, + { + "epoch": 0.51, + "learning_rate": 4.15100025647602e-05, + "loss": 0.5548, + "step": 13291 + }, + { + "epoch": 0.51, + "learning_rate": 4.1509361374711465e-05, + "loss": 0.5512, + "step": 13292 + }, + { + "epoch": 0.51, + "learning_rate": 4.1508720184662736e-05, + "loss": 0.5478, + "step": 13293 + }, + { + "epoch": 0.51, + "learning_rate": 4.150807899461401e-05, + "loss": 0.4891, + "step": 13294 + }, + { + "epoch": 0.51, + "learning_rate": 4.150743780456527e-05, + "loss": 0.6286, + "step": 13295 + }, + { + "epoch": 0.51, + "learning_rate": 4.1506796614516545e-05, + "loss": 0.5071, + "step": 13296 + }, + { + "epoch": 0.51, + "learning_rate": 4.1506155424467816e-05, + "loss": 0.4979, + "step": 13297 + }, + { + "epoch": 0.51, + "learning_rate": 4.150551423441909e-05, + "loss": 0.6326, + "step": 13298 + }, + { + "epoch": 0.51, + "learning_rate": 4.150487304437035e-05, + "loss": 0.5976, + "step": 13299 + }, + { + "epoch": 0.51, + "learning_rate": 4.1504231854321625e-05, + "loss": 0.5545, + "step": 13300 + }, + { + "epoch": 0.51, + "learning_rate": 4.150359066427289e-05, + "loss": 0.532, + "step": 13301 + }, + { + "epoch": 0.51, + "learning_rate": 4.150294947422416e-05, + "loss": 0.6217, + "step": 13302 + }, + { + "epoch": 0.51, + "learning_rate": 4.150230828417543e-05, + "loss": 0.4753, + "step": 13303 + }, + { + "epoch": 0.51, + "learning_rate": 4.15016670941267e-05, + "loss": 0.5227, + "step": 13304 + }, + { + "epoch": 0.51, + "learning_rate": 4.150102590407797e-05, + "loss": 0.5659, + "step": 13305 + }, + { + "epoch": 0.51, + "learning_rate": 4.150038471402924e-05, + "loss": 0.4978, + "step": 13306 + }, + { + "epoch": 0.51, + "learning_rate": 4.149974352398051e-05, + "loss": 0.5955, + "step": 13307 + }, + { + "epoch": 0.51, + "learning_rate": 4.149910233393178e-05, + "loss": 0.6348, + "step": 13308 + }, + { + "epoch": 0.51, + "learning_rate": 4.149846114388305e-05, + "loss": 0.6607, + "step": 13309 + }, + { + "epoch": 0.51, + "learning_rate": 4.1497819953834315e-05, + "loss": 0.5804, + "step": 13310 + }, + { + "epoch": 0.51, + "learning_rate": 4.1497178763785586e-05, + "loss": 0.5003, + "step": 13311 + }, + { + "epoch": 0.51, + "learning_rate": 4.149653757373686e-05, + "loss": 0.4883, + "step": 13312 + }, + { + "epoch": 0.51, + "learning_rate": 4.149589638368812e-05, + "loss": 0.5334, + "step": 13313 + }, + { + "epoch": 0.51, + "learning_rate": 4.14952551936394e-05, + "loss": 0.4947, + "step": 13314 + }, + { + "epoch": 0.51, + "learning_rate": 4.1494614003590667e-05, + "loss": 0.5305, + "step": 13315 + }, + { + "epoch": 0.51, + "learning_rate": 4.149397281354194e-05, + "loss": 0.6269, + "step": 13316 + }, + { + "epoch": 0.51, + "learning_rate": 4.14933316234932e-05, + "loss": 0.5145, + "step": 13317 + }, + { + "epoch": 0.51, + "learning_rate": 4.1492690433444475e-05, + "loss": 0.458, + "step": 13318 + }, + { + "epoch": 0.51, + "learning_rate": 4.149204924339574e-05, + "loss": 0.5877, + "step": 13319 + }, + { + "epoch": 0.51, + "learning_rate": 4.149140805334701e-05, + "loss": 0.5472, + "step": 13320 + }, + { + "epoch": 0.51, + "learning_rate": 4.149076686329828e-05, + "loss": 0.4859, + "step": 13321 + }, + { + "epoch": 0.51, + "learning_rate": 4.1490125673249555e-05, + "loss": 0.6502, + "step": 13322 + }, + { + "epoch": 0.51, + "learning_rate": 4.148948448320083e-05, + "loss": 0.5189, + "step": 13323 + }, + { + "epoch": 0.51, + "learning_rate": 4.148884329315209e-05, + "loss": 0.5124, + "step": 13324 + }, + { + "epoch": 0.51, + "learning_rate": 4.1488202103103363e-05, + "loss": 0.5993, + "step": 13325 + }, + { + "epoch": 0.51, + "learning_rate": 4.148756091305463e-05, + "loss": 0.52, + "step": 13326 + }, + { + "epoch": 0.51, + "learning_rate": 4.14869197230059e-05, + "loss": 0.4692, + "step": 13327 + }, + { + "epoch": 0.51, + "learning_rate": 4.1486278532957165e-05, + "loss": 0.505, + "step": 13328 + }, + { + "epoch": 0.51, + "learning_rate": 4.148563734290844e-05, + "loss": 0.6179, + "step": 13329 + }, + { + "epoch": 0.51, + "learning_rate": 4.1484996152859715e-05, + "loss": 0.6315, + "step": 13330 + }, + { + "epoch": 0.51, + "learning_rate": 4.148435496281098e-05, + "loss": 0.6218, + "step": 13331 + }, + { + "epoch": 0.51, + "learning_rate": 4.148371377276225e-05, + "loss": 0.6045, + "step": 13332 + }, + { + "epoch": 0.51, + "learning_rate": 4.148307258271352e-05, + "loss": 0.5691, + "step": 13333 + }, + { + "epoch": 0.51, + "learning_rate": 4.148243139266479e-05, + "loss": 0.6415, + "step": 13334 + }, + { + "epoch": 0.51, + "learning_rate": 4.1481790202616053e-05, + "loss": 0.5144, + "step": 13335 + }, + { + "epoch": 0.51, + "learning_rate": 4.1481149012567325e-05, + "loss": 0.551, + "step": 13336 + }, + { + "epoch": 0.51, + "learning_rate": 4.14805078225186e-05, + "loss": 0.5388, + "step": 13337 + }, + { + "epoch": 0.51, + "learning_rate": 4.147986663246987e-05, + "loss": 0.5369, + "step": 13338 + }, + { + "epoch": 0.51, + "learning_rate": 4.147922544242114e-05, + "loss": 0.5109, + "step": 13339 + }, + { + "epoch": 0.51, + "learning_rate": 4.1478584252372405e-05, + "loss": 0.556, + "step": 13340 + }, + { + "epoch": 0.51, + "learning_rate": 4.147794306232368e-05, + "loss": 0.5493, + "step": 13341 + }, + { + "epoch": 0.51, + "learning_rate": 4.147730187227494e-05, + "loss": 0.5118, + "step": 13342 + }, + { + "epoch": 0.51, + "learning_rate": 4.1476660682226214e-05, + "loss": 0.5557, + "step": 13343 + }, + { + "epoch": 0.51, + "learning_rate": 4.147601949217748e-05, + "loss": 0.541, + "step": 13344 + }, + { + "epoch": 0.51, + "learning_rate": 4.147537830212875e-05, + "loss": 0.5321, + "step": 13345 + }, + { + "epoch": 0.51, + "learning_rate": 4.147473711208002e-05, + "loss": 0.5573, + "step": 13346 + }, + { + "epoch": 0.51, + "learning_rate": 4.1474095922031294e-05, + "loss": 0.5225, + "step": 13347 + }, + { + "epoch": 0.51, + "learning_rate": 4.1473454731982565e-05, + "loss": 0.4571, + "step": 13348 + }, + { + "epoch": 0.51, + "learning_rate": 4.147281354193383e-05, + "loss": 0.5623, + "step": 13349 + }, + { + "epoch": 0.51, + "learning_rate": 4.14721723518851e-05, + "loss": 0.5345, + "step": 13350 + }, + { + "epoch": 0.51, + "learning_rate": 4.147153116183637e-05, + "loss": 0.4836, + "step": 13351 + }, + { + "epoch": 0.51, + "learning_rate": 4.147088997178764e-05, + "loss": 0.4548, + "step": 13352 + }, + { + "epoch": 0.51, + "learning_rate": 4.147024878173891e-05, + "loss": 0.509, + "step": 13353 + }, + { + "epoch": 0.51, + "learning_rate": 4.146960759169018e-05, + "loss": 0.5693, + "step": 13354 + }, + { + "epoch": 0.51, + "learning_rate": 4.146896640164145e-05, + "loss": 0.5753, + "step": 13355 + }, + { + "epoch": 0.51, + "learning_rate": 4.146832521159272e-05, + "loss": 0.4153, + "step": 13356 + }, + { + "epoch": 0.51, + "learning_rate": 4.146768402154399e-05, + "loss": 0.646, + "step": 13357 + }, + { + "epoch": 0.51, + "learning_rate": 4.1467042831495256e-05, + "loss": 0.5428, + "step": 13358 + }, + { + "epoch": 0.51, + "learning_rate": 4.146640164144653e-05, + "loss": 0.518, + "step": 13359 + }, + { + "epoch": 0.51, + "learning_rate": 4.146576045139779e-05, + "loss": 0.6475, + "step": 13360 + }, + { + "epoch": 0.51, + "learning_rate": 4.1465119261349064e-05, + "loss": 0.5372, + "step": 13361 + }, + { + "epoch": 0.51, + "learning_rate": 4.1464478071300336e-05, + "loss": 0.615, + "step": 13362 + }, + { + "epoch": 0.51, + "learning_rate": 4.146383688125161e-05, + "loss": 0.6274, + "step": 13363 + }, + { + "epoch": 0.51, + "learning_rate": 4.146319569120287e-05, + "loss": 0.645, + "step": 13364 + }, + { + "epoch": 0.51, + "learning_rate": 4.1462554501154144e-05, + "loss": 0.4871, + "step": 13365 + }, + { + "epoch": 0.51, + "learning_rate": 4.1461913311105416e-05, + "loss": 0.54, + "step": 13366 + }, + { + "epoch": 0.51, + "learning_rate": 4.146127212105668e-05, + "loss": 0.6233, + "step": 13367 + }, + { + "epoch": 0.51, + "learning_rate": 4.146063093100795e-05, + "loss": 0.4047, + "step": 13368 + }, + { + "epoch": 0.51, + "learning_rate": 4.1459989740959224e-05, + "loss": 0.5519, + "step": 13369 + }, + { + "epoch": 0.51, + "learning_rate": 4.1459348550910496e-05, + "loss": 0.583, + "step": 13370 + }, + { + "epoch": 0.51, + "learning_rate": 4.145870736086176e-05, + "loss": 0.5433, + "step": 13371 + }, + { + "epoch": 0.51, + "learning_rate": 4.145806617081303e-05, + "loss": 0.5623, + "step": 13372 + }, + { + "epoch": 0.51, + "learning_rate": 4.14574249807643e-05, + "loss": 0.5672, + "step": 13373 + }, + { + "epoch": 0.51, + "learning_rate": 4.145678379071557e-05, + "loss": 0.6667, + "step": 13374 + }, + { + "epoch": 0.51, + "learning_rate": 4.145614260066684e-05, + "loss": 0.5463, + "step": 13375 + }, + { + "epoch": 0.51, + "learning_rate": 4.1455501410618106e-05, + "loss": 0.5096, + "step": 13376 + }, + { + "epoch": 0.51, + "learning_rate": 4.145486022056938e-05, + "loss": 0.5221, + "step": 13377 + }, + { + "epoch": 0.51, + "learning_rate": 4.145421903052065e-05, + "loss": 0.5685, + "step": 13378 + }, + { + "epoch": 0.51, + "learning_rate": 4.145357784047192e-05, + "loss": 0.5747, + "step": 13379 + }, + { + "epoch": 0.51, + "learning_rate": 4.1452936650423186e-05, + "loss": 0.5864, + "step": 13380 + }, + { + "epoch": 0.51, + "learning_rate": 4.145229546037446e-05, + "loss": 0.6063, + "step": 13381 + }, + { + "epoch": 0.51, + "learning_rate": 4.145165427032572e-05, + "loss": 0.5574, + "step": 13382 + }, + { + "epoch": 0.51, + "learning_rate": 4.1451013080276994e-05, + "loss": 0.5485, + "step": 13383 + }, + { + "epoch": 0.51, + "learning_rate": 4.1450371890228266e-05, + "loss": 0.4669, + "step": 13384 + }, + { + "epoch": 0.51, + "learning_rate": 4.144973070017954e-05, + "loss": 0.5691, + "step": 13385 + }, + { + "epoch": 0.51, + "learning_rate": 4.144908951013081e-05, + "loss": 0.5998, + "step": 13386 + }, + { + "epoch": 0.51, + "learning_rate": 4.1448448320082074e-05, + "loss": 0.5978, + "step": 13387 + }, + { + "epoch": 0.51, + "learning_rate": 4.1447807130033346e-05, + "loss": 0.5867, + "step": 13388 + }, + { + "epoch": 0.51, + "learning_rate": 4.144716593998461e-05, + "loss": 0.5183, + "step": 13389 + }, + { + "epoch": 0.51, + "learning_rate": 4.144652474993588e-05, + "loss": 0.5071, + "step": 13390 + }, + { + "epoch": 0.51, + "learning_rate": 4.144588355988715e-05, + "loss": 0.613, + "step": 13391 + }, + { + "epoch": 0.51, + "learning_rate": 4.144524236983842e-05, + "loss": 0.5133, + "step": 13392 + }, + { + "epoch": 0.51, + "learning_rate": 4.144460117978969e-05, + "loss": 0.4935, + "step": 13393 + }, + { + "epoch": 0.51, + "learning_rate": 4.144395998974096e-05, + "loss": 0.5578, + "step": 13394 + }, + { + "epoch": 0.51, + "learning_rate": 4.1443318799692235e-05, + "loss": 0.532, + "step": 13395 + }, + { + "epoch": 0.52, + "learning_rate": 4.14426776096435e-05, + "loss": 0.5613, + "step": 13396 + }, + { + "epoch": 0.52, + "learning_rate": 4.144203641959477e-05, + "loss": 0.5212, + "step": 13397 + }, + { + "epoch": 0.52, + "learning_rate": 4.1441395229546036e-05, + "loss": 0.61, + "step": 13398 + }, + { + "epoch": 0.52, + "learning_rate": 4.144075403949731e-05, + "loss": 0.5407, + "step": 13399 + }, + { + "epoch": 0.52, + "learning_rate": 4.144011284944858e-05, + "loss": 0.4929, + "step": 13400 + }, + { + "epoch": 0.52, + "learning_rate": 4.1439471659399844e-05, + "loss": 0.553, + "step": 13401 + }, + { + "epoch": 0.52, + "learning_rate": 4.143883046935112e-05, + "loss": 0.5393, + "step": 13402 + }, + { + "epoch": 0.52, + "learning_rate": 4.143818927930239e-05, + "loss": 0.4884, + "step": 13403 + }, + { + "epoch": 0.52, + "learning_rate": 4.143754808925366e-05, + "loss": 0.5629, + "step": 13404 + }, + { + "epoch": 0.52, + "learning_rate": 4.1436906899204925e-05, + "loss": 0.5784, + "step": 13405 + }, + { + "epoch": 0.52, + "learning_rate": 4.1436265709156196e-05, + "loss": 0.5393, + "step": 13406 + }, + { + "epoch": 0.52, + "learning_rate": 4.143562451910746e-05, + "loss": 0.5259, + "step": 13407 + }, + { + "epoch": 0.52, + "learning_rate": 4.143498332905873e-05, + "loss": 0.5714, + "step": 13408 + }, + { + "epoch": 0.52, + "learning_rate": 4.1434342139010005e-05, + "loss": 0.589, + "step": 13409 + }, + { + "epoch": 0.52, + "learning_rate": 4.1433700948961276e-05, + "loss": 0.6164, + "step": 13410 + }, + { + "epoch": 0.52, + "learning_rate": 4.143305975891255e-05, + "loss": 0.5232, + "step": 13411 + }, + { + "epoch": 0.52, + "learning_rate": 4.143241856886381e-05, + "loss": 0.5699, + "step": 13412 + }, + { + "epoch": 0.52, + "learning_rate": 4.1431777378815085e-05, + "loss": 0.5153, + "step": 13413 + }, + { + "epoch": 0.52, + "learning_rate": 4.143113618876635e-05, + "loss": 0.5896, + "step": 13414 + }, + { + "epoch": 0.52, + "learning_rate": 4.143049499871762e-05, + "loss": 0.5595, + "step": 13415 + }, + { + "epoch": 0.52, + "learning_rate": 4.1429853808668886e-05, + "loss": 0.5167, + "step": 13416 + }, + { + "epoch": 0.52, + "learning_rate": 4.142921261862016e-05, + "loss": 0.5837, + "step": 13417 + }, + { + "epoch": 0.52, + "learning_rate": 4.1428571428571437e-05, + "loss": 0.5412, + "step": 13418 + }, + { + "epoch": 0.52, + "learning_rate": 4.14279302385227e-05, + "loss": 0.59, + "step": 13419 + }, + { + "epoch": 0.52, + "learning_rate": 4.142728904847397e-05, + "loss": 0.5103, + "step": 13420 + }, + { + "epoch": 0.52, + "learning_rate": 4.142664785842524e-05, + "loss": 0.5684, + "step": 13421 + }, + { + "epoch": 0.52, + "learning_rate": 4.142600666837651e-05, + "loss": 0.5054, + "step": 13422 + }, + { + "epoch": 0.52, + "learning_rate": 4.1425365478327775e-05, + "loss": 0.4907, + "step": 13423 + }, + { + "epoch": 0.52, + "learning_rate": 4.1424724288279047e-05, + "loss": 0.6558, + "step": 13424 + }, + { + "epoch": 0.52, + "learning_rate": 4.142408309823032e-05, + "loss": 0.5662, + "step": 13425 + }, + { + "epoch": 0.52, + "learning_rate": 4.142344190818159e-05, + "loss": 0.5505, + "step": 13426 + }, + { + "epoch": 0.52, + "learning_rate": 4.142280071813286e-05, + "loss": 0.5618, + "step": 13427 + }, + { + "epoch": 0.52, + "learning_rate": 4.142215952808413e-05, + "loss": 0.5265, + "step": 13428 + }, + { + "epoch": 0.52, + "learning_rate": 4.14215183380354e-05, + "loss": 0.4589, + "step": 13429 + }, + { + "epoch": 0.52, + "learning_rate": 4.142087714798666e-05, + "loss": 0.5108, + "step": 13430 + }, + { + "epoch": 0.52, + "learning_rate": 4.1420235957937935e-05, + "loss": 0.5164, + "step": 13431 + }, + { + "epoch": 0.52, + "learning_rate": 4.14195947678892e-05, + "loss": 0.4718, + "step": 13432 + }, + { + "epoch": 0.52, + "learning_rate": 4.141895357784047e-05, + "loss": 0.4601, + "step": 13433 + }, + { + "epoch": 0.52, + "learning_rate": 4.141831238779174e-05, + "loss": 0.4946, + "step": 13434 + }, + { + "epoch": 0.52, + "learning_rate": 4.1417671197743015e-05, + "loss": 0.6225, + "step": 13435 + }, + { + "epoch": 0.52, + "learning_rate": 4.141703000769429e-05, + "loss": 0.6665, + "step": 13436 + }, + { + "epoch": 0.52, + "learning_rate": 4.141638881764555e-05, + "loss": 0.5091, + "step": 13437 + }, + { + "epoch": 0.52, + "learning_rate": 4.1415747627596823e-05, + "loss": 0.6065, + "step": 13438 + }, + { + "epoch": 0.52, + "learning_rate": 4.141510643754809e-05, + "loss": 0.4809, + "step": 13439 + }, + { + "epoch": 0.52, + "learning_rate": 4.141446524749936e-05, + "loss": 0.549, + "step": 13440 + }, + { + "epoch": 0.52, + "learning_rate": 4.141382405745063e-05, + "loss": 0.6257, + "step": 13441 + }, + { + "epoch": 0.52, + "learning_rate": 4.1413182867401904e-05, + "loss": 0.4801, + "step": 13442 + }, + { + "epoch": 0.52, + "learning_rate": 4.141254167735317e-05, + "loss": 0.4583, + "step": 13443 + }, + { + "epoch": 0.52, + "learning_rate": 4.141190048730444e-05, + "loss": 0.5177, + "step": 13444 + }, + { + "epoch": 0.52, + "learning_rate": 4.141125929725571e-05, + "loss": 0.5876, + "step": 13445 + }, + { + "epoch": 0.52, + "learning_rate": 4.141061810720698e-05, + "loss": 0.5635, + "step": 13446 + }, + { + "epoch": 0.52, + "learning_rate": 4.140997691715825e-05, + "loss": 0.454, + "step": 13447 + }, + { + "epoch": 0.52, + "learning_rate": 4.1409335727109514e-05, + "loss": 0.5711, + "step": 13448 + }, + { + "epoch": 0.52, + "learning_rate": 4.1408694537060785e-05, + "loss": 0.4838, + "step": 13449 + }, + { + "epoch": 0.52, + "learning_rate": 4.140805334701206e-05, + "loss": 0.4614, + "step": 13450 + }, + { + "epoch": 0.52, + "learning_rate": 4.140741215696333e-05, + "loss": 0.564, + "step": 13451 + }, + { + "epoch": 0.52, + "learning_rate": 4.1406770966914594e-05, + "loss": 0.4145, + "step": 13452 + }, + { + "epoch": 0.52, + "learning_rate": 4.1406129776865865e-05, + "loss": 0.5326, + "step": 13453 + }, + { + "epoch": 0.52, + "learning_rate": 4.140548858681714e-05, + "loss": 0.5336, + "step": 13454 + }, + { + "epoch": 0.52, + "learning_rate": 4.14048473967684e-05, + "loss": 0.5267, + "step": 13455 + }, + { + "epoch": 0.52, + "learning_rate": 4.1404206206719674e-05, + "loss": 0.5463, + "step": 13456 + }, + { + "epoch": 0.52, + "learning_rate": 4.1403565016670945e-05, + "loss": 0.5689, + "step": 13457 + }, + { + "epoch": 0.52, + "learning_rate": 4.140292382662222e-05, + "loss": 0.5351, + "step": 13458 + }, + { + "epoch": 0.52, + "learning_rate": 4.140228263657348e-05, + "loss": 0.4585, + "step": 13459 + }, + { + "epoch": 0.52, + "learning_rate": 4.1401641446524754e-05, + "loss": 0.5282, + "step": 13460 + }, + { + "epoch": 0.52, + "learning_rate": 4.140100025647602e-05, + "loss": 0.563, + "step": 13461 + }, + { + "epoch": 0.52, + "learning_rate": 4.140035906642729e-05, + "loss": 0.5318, + "step": 13462 + }, + { + "epoch": 0.52, + "learning_rate": 4.139971787637856e-05, + "loss": 0.5782, + "step": 13463 + }, + { + "epoch": 0.52, + "learning_rate": 4.139907668632983e-05, + "loss": 0.4916, + "step": 13464 + }, + { + "epoch": 0.52, + "learning_rate": 4.13984354962811e-05, + "loss": 0.5091, + "step": 13465 + }, + { + "epoch": 0.52, + "learning_rate": 4.139779430623237e-05, + "loss": 0.6457, + "step": 13466 + }, + { + "epoch": 0.52, + "learning_rate": 4.139715311618364e-05, + "loss": 0.46, + "step": 13467 + }, + { + "epoch": 0.52, + "learning_rate": 4.139651192613491e-05, + "loss": 0.5759, + "step": 13468 + }, + { + "epoch": 0.52, + "learning_rate": 4.139587073608618e-05, + "loss": 0.5252, + "step": 13469 + }, + { + "epoch": 0.52, + "learning_rate": 4.1395229546037444e-05, + "loss": 0.5117, + "step": 13470 + }, + { + "epoch": 0.52, + "learning_rate": 4.1394588355988716e-05, + "loss": 0.6272, + "step": 13471 + }, + { + "epoch": 0.52, + "learning_rate": 4.139394716593999e-05, + "loss": 0.5185, + "step": 13472 + }, + { + "epoch": 0.52, + "learning_rate": 4.139330597589125e-05, + "loss": 0.6034, + "step": 13473 + }, + { + "epoch": 0.52, + "learning_rate": 4.139266478584253e-05, + "loss": 0.5615, + "step": 13474 + }, + { + "epoch": 0.52, + "learning_rate": 4.1392023595793796e-05, + "loss": 0.6651, + "step": 13475 + }, + { + "epoch": 0.52, + "learning_rate": 4.139138240574507e-05, + "loss": 0.5127, + "step": 13476 + }, + { + "epoch": 0.52, + "learning_rate": 4.139074121569633e-05, + "loss": 0.527, + "step": 13477 + }, + { + "epoch": 0.52, + "learning_rate": 4.1390100025647604e-05, + "loss": 0.5343, + "step": 13478 + }, + { + "epoch": 0.52, + "learning_rate": 4.138945883559887e-05, + "loss": 0.4908, + "step": 13479 + }, + { + "epoch": 0.52, + "learning_rate": 4.138881764555014e-05, + "loss": 0.4996, + "step": 13480 + }, + { + "epoch": 0.52, + "learning_rate": 4.138817645550141e-05, + "loss": 0.5008, + "step": 13481 + }, + { + "epoch": 0.52, + "learning_rate": 4.1387535265452684e-05, + "loss": 0.5272, + "step": 13482 + }, + { + "epoch": 0.52, + "learning_rate": 4.1386894075403956e-05, + "loss": 0.5412, + "step": 13483 + }, + { + "epoch": 0.52, + "learning_rate": 4.138625288535522e-05, + "loss": 0.4739, + "step": 13484 + }, + { + "epoch": 0.52, + "learning_rate": 4.138561169530649e-05, + "loss": 0.5306, + "step": 13485 + }, + { + "epoch": 0.52, + "learning_rate": 4.138497050525776e-05, + "loss": 0.5228, + "step": 13486 + }, + { + "epoch": 0.52, + "learning_rate": 4.138432931520903e-05, + "loss": 0.5741, + "step": 13487 + }, + { + "epoch": 0.52, + "learning_rate": 4.1383688125160294e-05, + "loss": 0.5146, + "step": 13488 + }, + { + "epoch": 0.52, + "learning_rate": 4.1383046935111566e-05, + "loss": 0.5207, + "step": 13489 + }, + { + "epoch": 0.52, + "learning_rate": 4.1382405745062844e-05, + "loss": 0.5887, + "step": 13490 + }, + { + "epoch": 0.52, + "learning_rate": 4.138176455501411e-05, + "loss": 0.4704, + "step": 13491 + }, + { + "epoch": 0.52, + "learning_rate": 4.138112336496538e-05, + "loss": 0.5639, + "step": 13492 + }, + { + "epoch": 0.52, + "learning_rate": 4.1380482174916646e-05, + "loss": 0.4192, + "step": 13493 + }, + { + "epoch": 0.52, + "learning_rate": 4.137984098486792e-05, + "loss": 0.5249, + "step": 13494 + }, + { + "epoch": 0.52, + "learning_rate": 4.137919979481918e-05, + "loss": 0.4658, + "step": 13495 + }, + { + "epoch": 0.52, + "learning_rate": 4.1378558604770454e-05, + "loss": 0.6041, + "step": 13496 + }, + { + "epoch": 0.52, + "learning_rate": 4.1377917414721726e-05, + "loss": 0.592, + "step": 13497 + }, + { + "epoch": 0.52, + "learning_rate": 4.1377276224673e-05, + "loss": 0.5823, + "step": 13498 + }, + { + "epoch": 0.52, + "learning_rate": 4.137663503462427e-05, + "loss": 0.4648, + "step": 13499 + }, + { + "epoch": 0.52, + "learning_rate": 4.1375993844575534e-05, + "loss": 0.6019, + "step": 13500 + }, + { + "epoch": 0.52, + "learning_rate": 4.1375352654526806e-05, + "loss": 0.5929, + "step": 13501 + }, + { + "epoch": 0.52, + "learning_rate": 4.137471146447807e-05, + "loss": 0.5231, + "step": 13502 + }, + { + "epoch": 0.52, + "learning_rate": 4.137407027442934e-05, + "loss": 0.5733, + "step": 13503 + }, + { + "epoch": 0.52, + "learning_rate": 4.137342908438061e-05, + "loss": 0.4934, + "step": 13504 + }, + { + "epoch": 0.52, + "learning_rate": 4.137278789433188e-05, + "loss": 0.5816, + "step": 13505 + }, + { + "epoch": 0.52, + "learning_rate": 4.137214670428315e-05, + "loss": 0.5997, + "step": 13506 + }, + { + "epoch": 0.52, + "learning_rate": 4.137150551423442e-05, + "loss": 0.5203, + "step": 13507 + }, + { + "epoch": 0.52, + "learning_rate": 4.1370864324185695e-05, + "loss": 0.4716, + "step": 13508 + }, + { + "epoch": 0.52, + "learning_rate": 4.137022313413696e-05, + "loss": 0.4933, + "step": 13509 + }, + { + "epoch": 0.52, + "learning_rate": 4.136958194408823e-05, + "loss": 0.4445, + "step": 13510 + }, + { + "epoch": 0.52, + "learning_rate": 4.1368940754039496e-05, + "loss": 0.4663, + "step": 13511 + }, + { + "epoch": 0.52, + "learning_rate": 4.136829956399077e-05, + "loss": 0.6356, + "step": 13512 + }, + { + "epoch": 0.52, + "learning_rate": 4.136765837394204e-05, + "loss": 0.584, + "step": 13513 + }, + { + "epoch": 0.52, + "learning_rate": 4.136701718389331e-05, + "loss": 0.5875, + "step": 13514 + }, + { + "epoch": 0.52, + "learning_rate": 4.1366375993844576e-05, + "loss": 0.6627, + "step": 13515 + }, + { + "epoch": 0.52, + "learning_rate": 4.136573480379585e-05, + "loss": 0.5689, + "step": 13516 + }, + { + "epoch": 0.52, + "learning_rate": 4.136509361374712e-05, + "loss": 0.6332, + "step": 13517 + }, + { + "epoch": 0.52, + "learning_rate": 4.1364452423698385e-05, + "loss": 0.5981, + "step": 13518 + }, + { + "epoch": 0.52, + "learning_rate": 4.1363811233649656e-05, + "loss": 0.597, + "step": 13519 + }, + { + "epoch": 0.52, + "learning_rate": 4.136317004360092e-05, + "loss": 0.5264, + "step": 13520 + }, + { + "epoch": 0.52, + "learning_rate": 4.136252885355219e-05, + "loss": 0.6163, + "step": 13521 + }, + { + "epoch": 0.52, + "learning_rate": 4.1361887663503465e-05, + "loss": 0.4957, + "step": 13522 + }, + { + "epoch": 0.52, + "learning_rate": 4.1361246473454736e-05, + "loss": 0.5647, + "step": 13523 + }, + { + "epoch": 0.52, + "learning_rate": 4.1360605283406e-05, + "loss": 0.5052, + "step": 13524 + }, + { + "epoch": 0.52, + "learning_rate": 4.135996409335727e-05, + "loss": 0.6524, + "step": 13525 + }, + { + "epoch": 0.52, + "learning_rate": 4.1359322903308545e-05, + "loss": 0.4734, + "step": 13526 + }, + { + "epoch": 0.52, + "learning_rate": 4.135868171325981e-05, + "loss": 0.547, + "step": 13527 + }, + { + "epoch": 0.52, + "learning_rate": 4.135804052321108e-05, + "loss": 0.4778, + "step": 13528 + }, + { + "epoch": 0.52, + "learning_rate": 4.135739933316235e-05, + "loss": 0.5618, + "step": 13529 + }, + { + "epoch": 0.52, + "learning_rate": 4.1356758143113625e-05, + "loss": 0.5447, + "step": 13530 + }, + { + "epoch": 0.52, + "learning_rate": 4.135611695306489e-05, + "loss": 0.4776, + "step": 13531 + }, + { + "epoch": 0.52, + "learning_rate": 4.135547576301616e-05, + "loss": 0.5739, + "step": 13532 + }, + { + "epoch": 0.52, + "learning_rate": 4.1354834572967426e-05, + "loss": 0.5622, + "step": 13533 + }, + { + "epoch": 0.52, + "learning_rate": 4.13541933829187e-05, + "loss": 0.5877, + "step": 13534 + }, + { + "epoch": 0.52, + "learning_rate": 4.135355219286997e-05, + "loss": 0.5163, + "step": 13535 + }, + { + "epoch": 0.52, + "learning_rate": 4.1352911002821235e-05, + "loss": 0.4955, + "step": 13536 + }, + { + "epoch": 0.52, + "learning_rate": 4.1352269812772507e-05, + "loss": 0.5248, + "step": 13537 + }, + { + "epoch": 0.52, + "learning_rate": 4.135162862272378e-05, + "loss": 0.5797, + "step": 13538 + }, + { + "epoch": 0.52, + "learning_rate": 4.135098743267505e-05, + "loss": 0.6175, + "step": 13539 + }, + { + "epoch": 0.52, + "learning_rate": 4.1350346242626315e-05, + "loss": 0.5574, + "step": 13540 + }, + { + "epoch": 0.52, + "learning_rate": 4.134970505257759e-05, + "loss": 0.5443, + "step": 13541 + }, + { + "epoch": 0.52, + "learning_rate": 4.134906386252885e-05, + "loss": 0.5915, + "step": 13542 + }, + { + "epoch": 0.52, + "learning_rate": 4.134842267248012e-05, + "loss": 0.6431, + "step": 13543 + }, + { + "epoch": 0.52, + "learning_rate": 4.1347781482431395e-05, + "loss": 0.5814, + "step": 13544 + }, + { + "epoch": 0.52, + "learning_rate": 4.134714029238267e-05, + "loss": 0.5677, + "step": 13545 + }, + { + "epoch": 0.52, + "learning_rate": 4.134649910233394e-05, + "loss": 0.569, + "step": 13546 + }, + { + "epoch": 0.52, + "learning_rate": 4.1345857912285203e-05, + "loss": 0.5497, + "step": 13547 + }, + { + "epoch": 0.52, + "learning_rate": 4.1345216722236475e-05, + "loss": 0.6019, + "step": 13548 + }, + { + "epoch": 0.52, + "learning_rate": 4.134457553218774e-05, + "loss": 0.5319, + "step": 13549 + }, + { + "epoch": 0.52, + "learning_rate": 4.134393434213901e-05, + "loss": 0.5081, + "step": 13550 + }, + { + "epoch": 0.52, + "learning_rate": 4.134329315209028e-05, + "loss": 0.5159, + "step": 13551 + }, + { + "epoch": 0.52, + "learning_rate": 4.134265196204155e-05, + "loss": 0.5895, + "step": 13552 + }, + { + "epoch": 0.52, + "learning_rate": 4.134201077199282e-05, + "loss": 0.5479, + "step": 13553 + }, + { + "epoch": 0.52, + "learning_rate": 4.134136958194409e-05, + "loss": 0.5427, + "step": 13554 + }, + { + "epoch": 0.52, + "learning_rate": 4.1340728391895364e-05, + "loss": 0.4339, + "step": 13555 + }, + { + "epoch": 0.52, + "learning_rate": 4.134008720184663e-05, + "loss": 0.5607, + "step": 13556 + }, + { + "epoch": 0.52, + "learning_rate": 4.13394460117979e-05, + "loss": 0.5181, + "step": 13557 + }, + { + "epoch": 0.52, + "learning_rate": 4.1338804821749165e-05, + "loss": 0.4744, + "step": 13558 + }, + { + "epoch": 0.52, + "learning_rate": 4.133816363170044e-05, + "loss": 0.5444, + "step": 13559 + }, + { + "epoch": 0.52, + "learning_rate": 4.13375224416517e-05, + "loss": 0.4738, + "step": 13560 + }, + { + "epoch": 0.52, + "learning_rate": 4.1336881251602974e-05, + "loss": 0.5445, + "step": 13561 + }, + { + "epoch": 0.52, + "learning_rate": 4.133624006155425e-05, + "loss": 0.4952, + "step": 13562 + }, + { + "epoch": 0.52, + "learning_rate": 4.133559887150552e-05, + "loss": 0.5123, + "step": 13563 + }, + { + "epoch": 0.52, + "learning_rate": 4.133495768145679e-05, + "loss": 0.5673, + "step": 13564 + }, + { + "epoch": 0.52, + "learning_rate": 4.1334316491408054e-05, + "loss": 0.5712, + "step": 13565 + }, + { + "epoch": 0.52, + "learning_rate": 4.1333675301359325e-05, + "loss": 0.5487, + "step": 13566 + }, + { + "epoch": 0.52, + "learning_rate": 4.133303411131059e-05, + "loss": 0.5369, + "step": 13567 + }, + { + "epoch": 0.52, + "learning_rate": 4.133239292126186e-05, + "loss": 0.5234, + "step": 13568 + }, + { + "epoch": 0.52, + "learning_rate": 4.1331751731213134e-05, + "loss": 0.606, + "step": 13569 + }, + { + "epoch": 0.52, + "learning_rate": 4.1331110541164405e-05, + "loss": 0.5445, + "step": 13570 + }, + { + "epoch": 0.52, + "learning_rate": 4.133046935111568e-05, + "loss": 0.4175, + "step": 13571 + }, + { + "epoch": 0.52, + "learning_rate": 4.132982816106694e-05, + "loss": 0.5535, + "step": 13572 + }, + { + "epoch": 0.52, + "learning_rate": 4.1329186971018214e-05, + "loss": 0.5124, + "step": 13573 + }, + { + "epoch": 0.52, + "learning_rate": 4.132854578096948e-05, + "loss": 0.5313, + "step": 13574 + }, + { + "epoch": 0.52, + "learning_rate": 4.132790459092075e-05, + "loss": 0.4994, + "step": 13575 + }, + { + "epoch": 0.52, + "learning_rate": 4.1327263400872015e-05, + "loss": 0.5431, + "step": 13576 + }, + { + "epoch": 0.52, + "learning_rate": 4.132662221082329e-05, + "loss": 0.4434, + "step": 13577 + }, + { + "epoch": 0.52, + "learning_rate": 4.132598102077456e-05, + "loss": 0.6533, + "step": 13578 + }, + { + "epoch": 0.52, + "learning_rate": 4.132533983072583e-05, + "loss": 0.5446, + "step": 13579 + }, + { + "epoch": 0.52, + "learning_rate": 4.13246986406771e-05, + "loss": 0.4901, + "step": 13580 + }, + { + "epoch": 0.52, + "learning_rate": 4.132405745062837e-05, + "loss": 0.5938, + "step": 13581 + }, + { + "epoch": 0.52, + "learning_rate": 4.132341626057964e-05, + "loss": 0.3878, + "step": 13582 + }, + { + "epoch": 0.52, + "learning_rate": 4.1322775070530904e-05, + "loss": 0.5911, + "step": 13583 + }, + { + "epoch": 0.52, + "learning_rate": 4.1322133880482176e-05, + "loss": 0.5277, + "step": 13584 + }, + { + "epoch": 0.52, + "learning_rate": 4.132149269043345e-05, + "loss": 0.6027, + "step": 13585 + }, + { + "epoch": 0.52, + "learning_rate": 4.132085150038472e-05, + "loss": 0.4819, + "step": 13586 + }, + { + "epoch": 0.52, + "learning_rate": 4.132021031033599e-05, + "loss": 0.6423, + "step": 13587 + }, + { + "epoch": 0.52, + "learning_rate": 4.1319569120287256e-05, + "loss": 0.5658, + "step": 13588 + }, + { + "epoch": 0.52, + "learning_rate": 4.131892793023853e-05, + "loss": 0.511, + "step": 13589 + }, + { + "epoch": 0.52, + "learning_rate": 4.131828674018979e-05, + "loss": 0.5093, + "step": 13590 + }, + { + "epoch": 0.52, + "learning_rate": 4.1317645550141064e-05, + "loss": 0.5116, + "step": 13591 + }, + { + "epoch": 0.52, + "learning_rate": 4.131700436009233e-05, + "loss": 0.5874, + "step": 13592 + }, + { + "epoch": 0.52, + "learning_rate": 4.13163631700436e-05, + "loss": 0.5659, + "step": 13593 + }, + { + "epoch": 0.52, + "learning_rate": 4.131572197999487e-05, + "loss": 0.5843, + "step": 13594 + }, + { + "epoch": 0.52, + "learning_rate": 4.1315080789946144e-05, + "loss": 0.535, + "step": 13595 + }, + { + "epoch": 0.52, + "learning_rate": 4.1314439599897416e-05, + "loss": 0.6391, + "step": 13596 + }, + { + "epoch": 0.52, + "learning_rate": 4.131379840984868e-05, + "loss": 0.4932, + "step": 13597 + }, + { + "epoch": 0.52, + "learning_rate": 4.131315721979995e-05, + "loss": 0.637, + "step": 13598 + }, + { + "epoch": 0.52, + "learning_rate": 4.131251602975122e-05, + "loss": 0.6492, + "step": 13599 + }, + { + "epoch": 0.52, + "learning_rate": 4.131187483970249e-05, + "loss": 0.5145, + "step": 13600 + }, + { + "epoch": 0.52, + "learning_rate": 4.131123364965376e-05, + "loss": 0.5551, + "step": 13601 + }, + { + "epoch": 0.52, + "learning_rate": 4.131059245960503e-05, + "loss": 0.4889, + "step": 13602 + }, + { + "epoch": 0.52, + "learning_rate": 4.13099512695563e-05, + "loss": 0.542, + "step": 13603 + }, + { + "epoch": 0.52, + "learning_rate": 4.130931007950757e-05, + "loss": 0.5713, + "step": 13604 + }, + { + "epoch": 0.52, + "learning_rate": 4.130866888945884e-05, + "loss": 0.5788, + "step": 13605 + }, + { + "epoch": 0.52, + "learning_rate": 4.1308027699410106e-05, + "loss": 0.4694, + "step": 13606 + }, + { + "epoch": 0.52, + "learning_rate": 4.130738650936138e-05, + "loss": 0.5028, + "step": 13607 + }, + { + "epoch": 0.52, + "learning_rate": 4.130674531931264e-05, + "loss": 0.5286, + "step": 13608 + }, + { + "epoch": 0.52, + "learning_rate": 4.1306104129263914e-05, + "loss": 0.4758, + "step": 13609 + }, + { + "epoch": 0.52, + "learning_rate": 4.1305462939215186e-05, + "loss": 0.5468, + "step": 13610 + }, + { + "epoch": 0.52, + "learning_rate": 4.130482174916646e-05, + "loss": 0.4956, + "step": 13611 + }, + { + "epoch": 0.52, + "learning_rate": 4.130418055911772e-05, + "loss": 0.4847, + "step": 13612 + }, + { + "epoch": 0.52, + "learning_rate": 4.1303539369068994e-05, + "loss": 0.477, + "step": 13613 + }, + { + "epoch": 0.52, + "learning_rate": 4.1302898179020266e-05, + "loss": 0.5169, + "step": 13614 + }, + { + "epoch": 0.52, + "learning_rate": 4.130225698897153e-05, + "loss": 0.4847, + "step": 13615 + }, + { + "epoch": 0.52, + "learning_rate": 4.13016157989228e-05, + "loss": 0.4406, + "step": 13616 + }, + { + "epoch": 0.52, + "learning_rate": 4.1300974608874074e-05, + "loss": 0.5504, + "step": 13617 + }, + { + "epoch": 0.52, + "learning_rate": 4.1300333418825346e-05, + "loss": 0.5806, + "step": 13618 + }, + { + "epoch": 0.52, + "learning_rate": 4.129969222877661e-05, + "loss": 0.6304, + "step": 13619 + }, + { + "epoch": 0.52, + "learning_rate": 4.129905103872788e-05, + "loss": 0.5445, + "step": 13620 + }, + { + "epoch": 0.52, + "learning_rate": 4.129840984867915e-05, + "loss": 0.5708, + "step": 13621 + }, + { + "epoch": 0.52, + "learning_rate": 4.129776865863042e-05, + "loss": 0.5488, + "step": 13622 + }, + { + "epoch": 0.52, + "learning_rate": 4.129712746858169e-05, + "loss": 0.4858, + "step": 13623 + }, + { + "epoch": 0.52, + "learning_rate": 4.1296486278532956e-05, + "loss": 0.6456, + "step": 13624 + }, + { + "epoch": 0.52, + "learning_rate": 4.129584508848423e-05, + "loss": 0.5917, + "step": 13625 + }, + { + "epoch": 0.52, + "learning_rate": 4.12952038984355e-05, + "loss": 0.5653, + "step": 13626 + }, + { + "epoch": 0.52, + "learning_rate": 4.129456270838677e-05, + "loss": 0.5417, + "step": 13627 + }, + { + "epoch": 0.52, + "learning_rate": 4.1293921518338036e-05, + "loss": 0.6646, + "step": 13628 + }, + { + "epoch": 0.52, + "learning_rate": 4.129328032828931e-05, + "loss": 0.587, + "step": 13629 + }, + { + "epoch": 0.52, + "learning_rate": 4.129263913824057e-05, + "loss": 0.4727, + "step": 13630 + }, + { + "epoch": 0.52, + "learning_rate": 4.1291997948191845e-05, + "loss": 0.5445, + "step": 13631 + }, + { + "epoch": 0.52, + "learning_rate": 4.1291356758143116e-05, + "loss": 0.4667, + "step": 13632 + }, + { + "epoch": 0.52, + "learning_rate": 4.129071556809439e-05, + "loss": 0.598, + "step": 13633 + }, + { + "epoch": 0.52, + "learning_rate": 4.129007437804566e-05, + "loss": 0.5171, + "step": 13634 + }, + { + "epoch": 0.52, + "learning_rate": 4.1289433187996925e-05, + "loss": 0.4829, + "step": 13635 + }, + { + "epoch": 0.52, + "learning_rate": 4.1288791997948196e-05, + "loss": 0.5179, + "step": 13636 + }, + { + "epoch": 0.52, + "learning_rate": 4.128815080789946e-05, + "loss": 0.5408, + "step": 13637 + }, + { + "epoch": 0.52, + "learning_rate": 4.128750961785073e-05, + "loss": 0.6036, + "step": 13638 + }, + { + "epoch": 0.52, + "learning_rate": 4.1286868427802e-05, + "loss": 0.5905, + "step": 13639 + }, + { + "epoch": 0.52, + "learning_rate": 4.128622723775327e-05, + "loss": 0.516, + "step": 13640 + }, + { + "epoch": 0.52, + "learning_rate": 4.128558604770454e-05, + "loss": 0.5956, + "step": 13641 + }, + { + "epoch": 0.52, + "learning_rate": 4.128494485765581e-05, + "loss": 0.5856, + "step": 13642 + }, + { + "epoch": 0.52, + "learning_rate": 4.1284303667607085e-05, + "loss": 0.5192, + "step": 13643 + }, + { + "epoch": 0.52, + "learning_rate": 4.128366247755835e-05, + "loss": 0.479, + "step": 13644 + }, + { + "epoch": 0.52, + "learning_rate": 4.128302128750962e-05, + "loss": 0.4986, + "step": 13645 + }, + { + "epoch": 0.52, + "learning_rate": 4.1282380097460887e-05, + "loss": 0.4943, + "step": 13646 + }, + { + "epoch": 0.52, + "learning_rate": 4.128173890741216e-05, + "loss": 0.4298, + "step": 13647 + }, + { + "epoch": 0.52, + "learning_rate": 4.128109771736342e-05, + "loss": 0.644, + "step": 13648 + }, + { + "epoch": 0.52, + "learning_rate": 4.1280456527314695e-05, + "loss": 0.5959, + "step": 13649 + }, + { + "epoch": 0.52, + "learning_rate": 4.127981533726597e-05, + "loss": 0.6782, + "step": 13650 + }, + { + "epoch": 0.52, + "learning_rate": 4.127917414721724e-05, + "loss": 0.5744, + "step": 13651 + }, + { + "epoch": 0.52, + "learning_rate": 4.127853295716851e-05, + "loss": 0.5001, + "step": 13652 + }, + { + "epoch": 0.52, + "learning_rate": 4.1277891767119775e-05, + "loss": 0.4527, + "step": 13653 + }, + { + "epoch": 0.52, + "learning_rate": 4.127725057707105e-05, + "loss": 0.5331, + "step": 13654 + }, + { + "epoch": 0.52, + "learning_rate": 4.127660938702231e-05, + "loss": 0.5717, + "step": 13655 + }, + { + "epoch": 0.53, + "learning_rate": 4.127596819697358e-05, + "loss": 0.579, + "step": 13656 + }, + { + "epoch": 0.53, + "learning_rate": 4.1275327006924855e-05, + "loss": 0.4002, + "step": 13657 + }, + { + "epoch": 0.53, + "learning_rate": 4.127468581687613e-05, + "loss": 0.5329, + "step": 13658 + }, + { + "epoch": 0.53, + "learning_rate": 4.12740446268274e-05, + "loss": 0.5615, + "step": 13659 + }, + { + "epoch": 0.53, + "learning_rate": 4.1273403436778663e-05, + "loss": 0.6599, + "step": 13660 + }, + { + "epoch": 0.53, + "learning_rate": 4.1272762246729935e-05, + "loss": 0.4674, + "step": 13661 + }, + { + "epoch": 0.53, + "learning_rate": 4.12721210566812e-05, + "loss": 0.4976, + "step": 13662 + }, + { + "epoch": 0.53, + "learning_rate": 4.127147986663247e-05, + "loss": 0.63, + "step": 13663 + }, + { + "epoch": 0.53, + "learning_rate": 4.127083867658374e-05, + "loss": 0.6626, + "step": 13664 + }, + { + "epoch": 0.53, + "learning_rate": 4.127019748653501e-05, + "loss": 0.5605, + "step": 13665 + }, + { + "epoch": 0.53, + "learning_rate": 4.126955629648628e-05, + "loss": 0.4675, + "step": 13666 + }, + { + "epoch": 0.53, + "learning_rate": 4.126891510643755e-05, + "loss": 0.545, + "step": 13667 + }, + { + "epoch": 0.53, + "learning_rate": 4.1268273916388824e-05, + "loss": 0.4907, + "step": 13668 + }, + { + "epoch": 0.53, + "learning_rate": 4.126763272634009e-05, + "loss": 0.5086, + "step": 13669 + }, + { + "epoch": 0.53, + "learning_rate": 4.126699153629136e-05, + "loss": 0.5639, + "step": 13670 + }, + { + "epoch": 0.53, + "learning_rate": 4.1266350346242625e-05, + "loss": 0.5191, + "step": 13671 + }, + { + "epoch": 0.53, + "learning_rate": 4.12657091561939e-05, + "loss": 0.4772, + "step": 13672 + }, + { + "epoch": 0.53, + "learning_rate": 4.126506796614517e-05, + "loss": 0.5236, + "step": 13673 + }, + { + "epoch": 0.53, + "learning_rate": 4.126442677609644e-05, + "loss": 0.5704, + "step": 13674 + }, + { + "epoch": 0.53, + "learning_rate": 4.1263785586047705e-05, + "loss": 0.4479, + "step": 13675 + }, + { + "epoch": 0.53, + "learning_rate": 4.126314439599898e-05, + "loss": 0.5214, + "step": 13676 + }, + { + "epoch": 0.53, + "learning_rate": 4.126250320595025e-05, + "loss": 0.5675, + "step": 13677 + }, + { + "epoch": 0.53, + "learning_rate": 4.1261862015901514e-05, + "loss": 0.5949, + "step": 13678 + }, + { + "epoch": 0.53, + "learning_rate": 4.1261220825852785e-05, + "loss": 0.6083, + "step": 13679 + }, + { + "epoch": 0.53, + "learning_rate": 4.126057963580405e-05, + "loss": 0.5812, + "step": 13680 + }, + { + "epoch": 0.53, + "learning_rate": 4.125993844575532e-05, + "loss": 0.6152, + "step": 13681 + }, + { + "epoch": 0.53, + "learning_rate": 4.1259297255706594e-05, + "loss": 0.5428, + "step": 13682 + }, + { + "epoch": 0.53, + "learning_rate": 4.1258656065657865e-05, + "loss": 0.5723, + "step": 13683 + }, + { + "epoch": 0.53, + "learning_rate": 4.125801487560913e-05, + "loss": 0.5723, + "step": 13684 + }, + { + "epoch": 0.53, + "learning_rate": 4.12573736855604e-05, + "loss": 0.5002, + "step": 13685 + }, + { + "epoch": 0.53, + "learning_rate": 4.1256732495511674e-05, + "loss": 0.5052, + "step": 13686 + }, + { + "epoch": 0.53, + "learning_rate": 4.125609130546294e-05, + "loss": 0.5542, + "step": 13687 + }, + { + "epoch": 0.53, + "learning_rate": 4.125545011541421e-05, + "loss": 0.6027, + "step": 13688 + }, + { + "epoch": 0.53, + "learning_rate": 4.125480892536548e-05, + "loss": 0.5428, + "step": 13689 + }, + { + "epoch": 0.53, + "learning_rate": 4.1254167735316754e-05, + "loss": 0.5689, + "step": 13690 + }, + { + "epoch": 0.53, + "learning_rate": 4.125352654526802e-05, + "loss": 0.5703, + "step": 13691 + }, + { + "epoch": 0.53, + "learning_rate": 4.125288535521929e-05, + "loss": 0.4656, + "step": 13692 + }, + { + "epoch": 0.53, + "learning_rate": 4.1252244165170556e-05, + "loss": 0.6045, + "step": 13693 + }, + { + "epoch": 0.53, + "learning_rate": 4.125160297512183e-05, + "loss": 0.672, + "step": 13694 + }, + { + "epoch": 0.53, + "learning_rate": 4.12509617850731e-05, + "loss": 0.5575, + "step": 13695 + }, + { + "epoch": 0.53, + "learning_rate": 4.1250320595024364e-05, + "loss": 0.4471, + "step": 13696 + }, + { + "epoch": 0.53, + "learning_rate": 4.1249679404975636e-05, + "loss": 0.5607, + "step": 13697 + }, + { + "epoch": 0.53, + "learning_rate": 4.124903821492691e-05, + "loss": 0.5052, + "step": 13698 + }, + { + "epoch": 0.53, + "learning_rate": 4.124839702487818e-05, + "loss": 0.549, + "step": 13699 + }, + { + "epoch": 0.53, + "learning_rate": 4.1247755834829444e-05, + "loss": 0.5724, + "step": 13700 + }, + { + "epoch": 0.53, + "learning_rate": 4.1247114644780716e-05, + "loss": 0.5059, + "step": 13701 + }, + { + "epoch": 0.53, + "learning_rate": 4.124647345473198e-05, + "loss": 0.5725, + "step": 13702 + }, + { + "epoch": 0.53, + "learning_rate": 4.124583226468325e-05, + "loss": 0.5097, + "step": 13703 + }, + { + "epoch": 0.53, + "learning_rate": 4.1245191074634524e-05, + "loss": 0.4755, + "step": 13704 + }, + { + "epoch": 0.53, + "learning_rate": 4.1244549884585796e-05, + "loss": 0.6354, + "step": 13705 + }, + { + "epoch": 0.53, + "learning_rate": 4.124390869453707e-05, + "loss": 0.5803, + "step": 13706 + }, + { + "epoch": 0.53, + "learning_rate": 4.124326750448833e-05, + "loss": 0.4899, + "step": 13707 + }, + { + "epoch": 0.53, + "learning_rate": 4.1242626314439604e-05, + "loss": 0.4354, + "step": 13708 + }, + { + "epoch": 0.53, + "learning_rate": 4.124198512439087e-05, + "loss": 0.5182, + "step": 13709 + }, + { + "epoch": 0.53, + "learning_rate": 4.124134393434214e-05, + "loss": 0.5155, + "step": 13710 + }, + { + "epoch": 0.53, + "learning_rate": 4.1240702744293406e-05, + "loss": 0.5481, + "step": 13711 + }, + { + "epoch": 0.53, + "learning_rate": 4.124006155424468e-05, + "loss": 0.5153, + "step": 13712 + }, + { + "epoch": 0.53, + "learning_rate": 4.123942036419595e-05, + "loss": 0.4977, + "step": 13713 + }, + { + "epoch": 0.53, + "learning_rate": 4.123877917414722e-05, + "loss": 0.561, + "step": 13714 + }, + { + "epoch": 0.53, + "learning_rate": 4.123813798409849e-05, + "loss": 0.6113, + "step": 13715 + }, + { + "epoch": 0.53, + "learning_rate": 4.123749679404976e-05, + "loss": 0.3978, + "step": 13716 + }, + { + "epoch": 0.53, + "learning_rate": 4.123685560400103e-05, + "loss": 0.6106, + "step": 13717 + }, + { + "epoch": 0.53, + "learning_rate": 4.1236214413952294e-05, + "loss": 0.4766, + "step": 13718 + }, + { + "epoch": 0.53, + "learning_rate": 4.1235573223903566e-05, + "loss": 0.5047, + "step": 13719 + }, + { + "epoch": 0.53, + "learning_rate": 4.123493203385483e-05, + "loss": 0.4795, + "step": 13720 + }, + { + "epoch": 0.53, + "learning_rate": 4.12342908438061e-05, + "loss": 0.4782, + "step": 13721 + }, + { + "epoch": 0.53, + "learning_rate": 4.123364965375738e-05, + "loss": 0.5757, + "step": 13722 + }, + { + "epoch": 0.53, + "learning_rate": 4.1233008463708646e-05, + "loss": 0.6264, + "step": 13723 + }, + { + "epoch": 0.53, + "learning_rate": 4.123236727365992e-05, + "loss": 0.5136, + "step": 13724 + }, + { + "epoch": 0.53, + "learning_rate": 4.123172608361118e-05, + "loss": 0.5096, + "step": 13725 + }, + { + "epoch": 0.53, + "learning_rate": 4.1231084893562454e-05, + "loss": 0.6092, + "step": 13726 + }, + { + "epoch": 0.53, + "learning_rate": 4.123044370351372e-05, + "loss": 0.5453, + "step": 13727 + }, + { + "epoch": 0.53, + "learning_rate": 4.122980251346499e-05, + "loss": 0.5136, + "step": 13728 + }, + { + "epoch": 0.53, + "learning_rate": 4.122916132341626e-05, + "loss": 0.5107, + "step": 13729 + }, + { + "epoch": 0.53, + "learning_rate": 4.1228520133367535e-05, + "loss": 0.4979, + "step": 13730 + }, + { + "epoch": 0.53, + "learning_rate": 4.1227878943318806e-05, + "loss": 0.5722, + "step": 13731 + }, + { + "epoch": 0.53, + "learning_rate": 4.122723775327007e-05, + "loss": 0.5701, + "step": 13732 + }, + { + "epoch": 0.53, + "learning_rate": 4.122659656322134e-05, + "loss": 0.6122, + "step": 13733 + }, + { + "epoch": 0.53, + "learning_rate": 4.122595537317261e-05, + "loss": 0.6538, + "step": 13734 + }, + { + "epoch": 0.53, + "learning_rate": 4.122531418312388e-05, + "loss": 0.6525, + "step": 13735 + }, + { + "epoch": 0.53, + "learning_rate": 4.1224672993075144e-05, + "loss": 0.5508, + "step": 13736 + }, + { + "epoch": 0.53, + "learning_rate": 4.1224031803026416e-05, + "loss": 0.5304, + "step": 13737 + }, + { + "epoch": 0.53, + "learning_rate": 4.122339061297769e-05, + "loss": 0.553, + "step": 13738 + }, + { + "epoch": 0.53, + "learning_rate": 4.122274942292896e-05, + "loss": 0.6109, + "step": 13739 + }, + { + "epoch": 0.53, + "learning_rate": 4.122210823288023e-05, + "loss": 0.5415, + "step": 13740 + }, + { + "epoch": 0.53, + "learning_rate": 4.1221467042831496e-05, + "loss": 0.549, + "step": 13741 + }, + { + "epoch": 0.53, + "learning_rate": 4.122082585278277e-05, + "loss": 0.5835, + "step": 13742 + }, + { + "epoch": 0.53, + "learning_rate": 4.122018466273403e-05, + "loss": 0.6688, + "step": 13743 + }, + { + "epoch": 0.53, + "learning_rate": 4.1219543472685305e-05, + "loss": 0.5243, + "step": 13744 + }, + { + "epoch": 0.53, + "learning_rate": 4.1218902282636576e-05, + "loss": 0.5462, + "step": 13745 + }, + { + "epoch": 0.53, + "learning_rate": 4.121826109258785e-05, + "loss": 0.5075, + "step": 13746 + }, + { + "epoch": 0.53, + "learning_rate": 4.121761990253911e-05, + "loss": 0.5021, + "step": 13747 + }, + { + "epoch": 0.53, + "learning_rate": 4.1216978712490385e-05, + "loss": 0.6322, + "step": 13748 + }, + { + "epoch": 0.53, + "learning_rate": 4.1216337522441656e-05, + "loss": 0.525, + "step": 13749 + }, + { + "epoch": 0.53, + "learning_rate": 4.121569633239292e-05, + "loss": 0.5324, + "step": 13750 + }, + { + "epoch": 0.53, + "learning_rate": 4.121505514234419e-05, + "loss": 0.5739, + "step": 13751 + }, + { + "epoch": 0.53, + "learning_rate": 4.121441395229546e-05, + "loss": 0.5258, + "step": 13752 + }, + { + "epoch": 0.53, + "learning_rate": 4.121377276224673e-05, + "loss": 0.5556, + "step": 13753 + }, + { + "epoch": 0.53, + "learning_rate": 4.1213131572198e-05, + "loss": 0.4938, + "step": 13754 + }, + { + "epoch": 0.53, + "learning_rate": 4.121249038214927e-05, + "loss": 0.5137, + "step": 13755 + }, + { + "epoch": 0.53, + "learning_rate": 4.121184919210054e-05, + "loss": 0.4712, + "step": 13756 + }, + { + "epoch": 0.53, + "learning_rate": 4.121120800205181e-05, + "loss": 0.5768, + "step": 13757 + }, + { + "epoch": 0.53, + "learning_rate": 4.121056681200308e-05, + "loss": 0.5458, + "step": 13758 + }, + { + "epoch": 0.53, + "learning_rate": 4.1209925621954347e-05, + "loss": 0.5071, + "step": 13759 + }, + { + "epoch": 0.53, + "learning_rate": 4.120928443190562e-05, + "loss": 0.5385, + "step": 13760 + }, + { + "epoch": 0.53, + "learning_rate": 4.120864324185689e-05, + "loss": 0.568, + "step": 13761 + }, + { + "epoch": 0.53, + "learning_rate": 4.120800205180816e-05, + "loss": 0.4747, + "step": 13762 + }, + { + "epoch": 0.53, + "learning_rate": 4.120736086175943e-05, + "loss": 0.5077, + "step": 13763 + }, + { + "epoch": 0.53, + "learning_rate": 4.12067196717107e-05, + "loss": 0.5885, + "step": 13764 + }, + { + "epoch": 0.53, + "learning_rate": 4.120607848166196e-05, + "loss": 0.6294, + "step": 13765 + }, + { + "epoch": 0.53, + "learning_rate": 4.1205437291613235e-05, + "loss": 0.4821, + "step": 13766 + }, + { + "epoch": 0.53, + "learning_rate": 4.120479610156451e-05, + "loss": 0.6099, + "step": 13767 + }, + { + "epoch": 0.53, + "learning_rate": 4.120415491151577e-05, + "loss": 0.5393, + "step": 13768 + }, + { + "epoch": 0.53, + "learning_rate": 4.120351372146704e-05, + "loss": 0.6633, + "step": 13769 + }, + { + "epoch": 0.53, + "learning_rate": 4.1202872531418315e-05, + "loss": 0.5803, + "step": 13770 + }, + { + "epoch": 0.53, + "learning_rate": 4.120223134136959e-05, + "loss": 0.5277, + "step": 13771 + }, + { + "epoch": 0.53, + "learning_rate": 4.120159015132085e-05, + "loss": 0.5344, + "step": 13772 + }, + { + "epoch": 0.53, + "learning_rate": 4.1200948961272123e-05, + "loss": 0.6554, + "step": 13773 + }, + { + "epoch": 0.53, + "learning_rate": 4.1200307771223395e-05, + "loss": 0.4029, + "step": 13774 + }, + { + "epoch": 0.53, + "learning_rate": 4.119966658117466e-05, + "loss": 0.5434, + "step": 13775 + }, + { + "epoch": 0.53, + "learning_rate": 4.119902539112593e-05, + "loss": 0.6626, + "step": 13776 + }, + { + "epoch": 0.53, + "learning_rate": 4.1198384201077204e-05, + "loss": 0.4903, + "step": 13777 + }, + { + "epoch": 0.53, + "learning_rate": 4.1197743011028475e-05, + "loss": 0.6364, + "step": 13778 + }, + { + "epoch": 0.53, + "learning_rate": 4.119710182097974e-05, + "loss": 0.5071, + "step": 13779 + }, + { + "epoch": 0.53, + "learning_rate": 4.119646063093101e-05, + "loss": 0.5829, + "step": 13780 + }, + { + "epoch": 0.53, + "learning_rate": 4.119581944088228e-05, + "loss": 0.5726, + "step": 13781 + }, + { + "epoch": 0.53, + "learning_rate": 4.119517825083355e-05, + "loss": 0.5018, + "step": 13782 + }, + { + "epoch": 0.53, + "learning_rate": 4.119453706078482e-05, + "loss": 0.5407, + "step": 13783 + }, + { + "epoch": 0.53, + "learning_rate": 4.1193895870736085e-05, + "loss": 0.5964, + "step": 13784 + }, + { + "epoch": 0.53, + "learning_rate": 4.119325468068736e-05, + "loss": 0.5254, + "step": 13785 + }, + { + "epoch": 0.53, + "learning_rate": 4.119261349063863e-05, + "loss": 0.6187, + "step": 13786 + }, + { + "epoch": 0.53, + "learning_rate": 4.11919723005899e-05, + "loss": 0.5228, + "step": 13787 + }, + { + "epoch": 0.53, + "learning_rate": 4.1191331110541165e-05, + "loss": 0.5822, + "step": 13788 + }, + { + "epoch": 0.53, + "learning_rate": 4.119068992049244e-05, + "loss": 0.4653, + "step": 13789 + }, + { + "epoch": 0.53, + "learning_rate": 4.11900487304437e-05, + "loss": 0.5572, + "step": 13790 + }, + { + "epoch": 0.53, + "learning_rate": 4.1189407540394974e-05, + "loss": 0.6305, + "step": 13791 + }, + { + "epoch": 0.53, + "learning_rate": 4.1188766350346245e-05, + "loss": 0.5288, + "step": 13792 + }, + { + "epoch": 0.53, + "learning_rate": 4.118812516029752e-05, + "loss": 0.5836, + "step": 13793 + }, + { + "epoch": 0.53, + "learning_rate": 4.118748397024879e-05, + "loss": 0.4721, + "step": 13794 + }, + { + "epoch": 0.53, + "learning_rate": 4.1186842780200054e-05, + "loss": 0.6638, + "step": 13795 + }, + { + "epoch": 0.53, + "learning_rate": 4.1186201590151326e-05, + "loss": 0.4857, + "step": 13796 + }, + { + "epoch": 0.53, + "learning_rate": 4.118556040010259e-05, + "loss": 0.5066, + "step": 13797 + }, + { + "epoch": 0.53, + "learning_rate": 4.118491921005386e-05, + "loss": 0.5895, + "step": 13798 + }, + { + "epoch": 0.53, + "learning_rate": 4.118427802000513e-05, + "loss": 0.5158, + "step": 13799 + }, + { + "epoch": 0.53, + "learning_rate": 4.11836368299564e-05, + "loss": 0.5551, + "step": 13800 + }, + { + "epoch": 0.53, + "learning_rate": 4.118299563990767e-05, + "loss": 0.5221, + "step": 13801 + }, + { + "epoch": 0.53, + "learning_rate": 4.118235444985894e-05, + "loss": 0.5268, + "step": 13802 + }, + { + "epoch": 0.53, + "learning_rate": 4.1181713259810214e-05, + "loss": 0.5917, + "step": 13803 + }, + { + "epoch": 0.53, + "learning_rate": 4.118107206976148e-05, + "loss": 0.4441, + "step": 13804 + }, + { + "epoch": 0.53, + "learning_rate": 4.118043087971275e-05, + "loss": 0.5147, + "step": 13805 + }, + { + "epoch": 0.53, + "learning_rate": 4.1179789689664016e-05, + "loss": 0.6467, + "step": 13806 + }, + { + "epoch": 0.53, + "learning_rate": 4.117914849961529e-05, + "loss": 0.5824, + "step": 13807 + }, + { + "epoch": 0.53, + "learning_rate": 4.117850730956655e-05, + "loss": 0.4098, + "step": 13808 + }, + { + "epoch": 0.53, + "learning_rate": 4.1177866119517824e-05, + "loss": 0.5886, + "step": 13809 + }, + { + "epoch": 0.53, + "learning_rate": 4.11772249294691e-05, + "loss": 0.4908, + "step": 13810 + }, + { + "epoch": 0.53, + "learning_rate": 4.117658373942037e-05, + "loss": 0.4282, + "step": 13811 + }, + { + "epoch": 0.53, + "learning_rate": 4.117594254937164e-05, + "loss": 0.6381, + "step": 13812 + }, + { + "epoch": 0.53, + "learning_rate": 4.1175301359322904e-05, + "loss": 0.5049, + "step": 13813 + }, + { + "epoch": 0.53, + "learning_rate": 4.1174660169274176e-05, + "loss": 0.5535, + "step": 13814 + }, + { + "epoch": 0.53, + "learning_rate": 4.117401897922544e-05, + "loss": 0.5922, + "step": 13815 + }, + { + "epoch": 0.53, + "learning_rate": 4.117337778917671e-05, + "loss": 0.4447, + "step": 13816 + }, + { + "epoch": 0.53, + "learning_rate": 4.1172736599127984e-05, + "loss": 0.4161, + "step": 13817 + }, + { + "epoch": 0.53, + "learning_rate": 4.1172095409079256e-05, + "loss": 0.5701, + "step": 13818 + }, + { + "epoch": 0.53, + "learning_rate": 4.117145421903053e-05, + "loss": 0.5988, + "step": 13819 + }, + { + "epoch": 0.53, + "learning_rate": 4.117081302898179e-05, + "loss": 0.4797, + "step": 13820 + }, + { + "epoch": 0.53, + "learning_rate": 4.1170171838933064e-05, + "loss": 0.5562, + "step": 13821 + }, + { + "epoch": 0.53, + "learning_rate": 4.116953064888433e-05, + "loss": 0.5042, + "step": 13822 + }, + { + "epoch": 0.53, + "learning_rate": 4.11688894588356e-05, + "loss": 0.419, + "step": 13823 + }, + { + "epoch": 0.53, + "learning_rate": 4.1168248268786866e-05, + "loss": 0.5923, + "step": 13824 + }, + { + "epoch": 0.53, + "learning_rate": 4.116760707873814e-05, + "loss": 0.6396, + "step": 13825 + }, + { + "epoch": 0.53, + "learning_rate": 4.116696588868941e-05, + "loss": 0.5031, + "step": 13826 + }, + { + "epoch": 0.53, + "learning_rate": 4.116632469864068e-05, + "loss": 0.6747, + "step": 13827 + }, + { + "epoch": 0.53, + "learning_rate": 4.116568350859195e-05, + "loss": 0.5481, + "step": 13828 + }, + { + "epoch": 0.53, + "learning_rate": 4.116504231854322e-05, + "loss": 0.6573, + "step": 13829 + }, + { + "epoch": 0.53, + "learning_rate": 4.116440112849449e-05, + "loss": 0.6479, + "step": 13830 + }, + { + "epoch": 0.53, + "learning_rate": 4.1163759938445754e-05, + "loss": 0.5425, + "step": 13831 + }, + { + "epoch": 0.53, + "learning_rate": 4.1163118748397026e-05, + "loss": 0.6054, + "step": 13832 + }, + { + "epoch": 0.53, + "learning_rate": 4.11624775583483e-05, + "loss": 0.5908, + "step": 13833 + }, + { + "epoch": 0.53, + "learning_rate": 4.116183636829957e-05, + "loss": 0.5734, + "step": 13834 + }, + { + "epoch": 0.53, + "learning_rate": 4.1161195178250834e-05, + "loss": 0.4696, + "step": 13835 + }, + { + "epoch": 0.53, + "learning_rate": 4.1160553988202106e-05, + "loss": 0.5212, + "step": 13836 + }, + { + "epoch": 0.53, + "learning_rate": 4.115991279815338e-05, + "loss": 0.4801, + "step": 13837 + }, + { + "epoch": 0.53, + "learning_rate": 4.115927160810464e-05, + "loss": 0.583, + "step": 13838 + }, + { + "epoch": 0.53, + "learning_rate": 4.1158630418055914e-05, + "loss": 0.5427, + "step": 13839 + }, + { + "epoch": 0.53, + "learning_rate": 4.115798922800718e-05, + "loss": 0.532, + "step": 13840 + }, + { + "epoch": 0.53, + "learning_rate": 4.115734803795845e-05, + "loss": 0.4005, + "step": 13841 + }, + { + "epoch": 0.53, + "learning_rate": 4.115670684790972e-05, + "loss": 0.4798, + "step": 13842 + }, + { + "epoch": 0.53, + "learning_rate": 4.1156065657860995e-05, + "loss": 0.4659, + "step": 13843 + }, + { + "epoch": 0.53, + "learning_rate": 4.115542446781226e-05, + "loss": 0.5249, + "step": 13844 + }, + { + "epoch": 0.53, + "learning_rate": 4.115478327776353e-05, + "loss": 0.5819, + "step": 13845 + }, + { + "epoch": 0.53, + "learning_rate": 4.11541420877148e-05, + "loss": 0.6137, + "step": 13846 + }, + { + "epoch": 0.53, + "learning_rate": 4.115350089766607e-05, + "loss": 0.4725, + "step": 13847 + }, + { + "epoch": 0.53, + "learning_rate": 4.115285970761734e-05, + "loss": 0.5179, + "step": 13848 + }, + { + "epoch": 0.53, + "learning_rate": 4.115221851756861e-05, + "loss": 0.5518, + "step": 13849 + }, + { + "epoch": 0.53, + "learning_rate": 4.115157732751988e-05, + "loss": 0.5507, + "step": 13850 + }, + { + "epoch": 0.53, + "learning_rate": 4.115093613747115e-05, + "loss": 0.5695, + "step": 13851 + }, + { + "epoch": 0.53, + "learning_rate": 4.115029494742242e-05, + "loss": 0.5116, + "step": 13852 + }, + { + "epoch": 0.53, + "learning_rate": 4.1149653757373685e-05, + "loss": 0.5004, + "step": 13853 + }, + { + "epoch": 0.53, + "learning_rate": 4.1149012567324956e-05, + "loss": 0.5755, + "step": 13854 + }, + { + "epoch": 0.53, + "learning_rate": 4.114837137727623e-05, + "loss": 0.5896, + "step": 13855 + }, + { + "epoch": 0.53, + "learning_rate": 4.114773018722749e-05, + "loss": 0.4692, + "step": 13856 + }, + { + "epoch": 0.53, + "learning_rate": 4.1147088997178765e-05, + "loss": 0.5388, + "step": 13857 + }, + { + "epoch": 0.53, + "learning_rate": 4.1146447807130036e-05, + "loss": 0.4517, + "step": 13858 + }, + { + "epoch": 0.53, + "learning_rate": 4.114580661708131e-05, + "loss": 0.5095, + "step": 13859 + }, + { + "epoch": 0.53, + "learning_rate": 4.114516542703257e-05, + "loss": 0.5502, + "step": 13860 + }, + { + "epoch": 0.53, + "learning_rate": 4.1144524236983845e-05, + "loss": 0.545, + "step": 13861 + }, + { + "epoch": 0.53, + "learning_rate": 4.114388304693511e-05, + "loss": 0.5444, + "step": 13862 + }, + { + "epoch": 0.53, + "learning_rate": 4.114324185688638e-05, + "loss": 0.4857, + "step": 13863 + }, + { + "epoch": 0.53, + "learning_rate": 4.114260066683765e-05, + "loss": 0.6171, + "step": 13864 + }, + { + "epoch": 0.53, + "learning_rate": 4.1141959476788925e-05, + "loss": 0.5852, + "step": 13865 + }, + { + "epoch": 0.53, + "learning_rate": 4.1141318286740197e-05, + "loss": 0.5337, + "step": 13866 + }, + { + "epoch": 0.53, + "learning_rate": 4.114067709669146e-05, + "loss": 0.4832, + "step": 13867 + }, + { + "epoch": 0.53, + "learning_rate": 4.114003590664273e-05, + "loss": 0.5621, + "step": 13868 + }, + { + "epoch": 0.53, + "learning_rate": 4.1139394716594e-05, + "loss": 0.4715, + "step": 13869 + }, + { + "epoch": 0.53, + "learning_rate": 4.113875352654527e-05, + "loss": 0.3625, + "step": 13870 + }, + { + "epoch": 0.53, + "learning_rate": 4.1138112336496535e-05, + "loss": 0.5619, + "step": 13871 + }, + { + "epoch": 0.53, + "learning_rate": 4.1137471146447807e-05, + "loss": 0.5245, + "step": 13872 + }, + { + "epoch": 0.53, + "learning_rate": 4.113682995639908e-05, + "loss": 0.4928, + "step": 13873 + }, + { + "epoch": 0.53, + "learning_rate": 4.113618876635035e-05, + "loss": 0.4971, + "step": 13874 + }, + { + "epoch": 0.53, + "learning_rate": 4.113554757630162e-05, + "loss": 0.7615, + "step": 13875 + }, + { + "epoch": 0.53, + "learning_rate": 4.113490638625289e-05, + "loss": 0.55, + "step": 13876 + }, + { + "epoch": 0.53, + "learning_rate": 4.113426519620416e-05, + "loss": 0.4905, + "step": 13877 + }, + { + "epoch": 0.53, + "learning_rate": 4.113362400615542e-05, + "loss": 0.6104, + "step": 13878 + }, + { + "epoch": 0.53, + "learning_rate": 4.1132982816106695e-05, + "loss": 0.47, + "step": 13879 + }, + { + "epoch": 0.53, + "learning_rate": 4.113234162605796e-05, + "loss": 0.5016, + "step": 13880 + }, + { + "epoch": 0.53, + "learning_rate": 4.113170043600923e-05, + "loss": 0.5439, + "step": 13881 + }, + { + "epoch": 0.53, + "learning_rate": 4.113105924596051e-05, + "loss": 0.5685, + "step": 13882 + }, + { + "epoch": 0.53, + "learning_rate": 4.1130418055911775e-05, + "loss": 0.6398, + "step": 13883 + }, + { + "epoch": 0.53, + "learning_rate": 4.112977686586305e-05, + "loss": 0.5907, + "step": 13884 + }, + { + "epoch": 0.53, + "learning_rate": 4.112913567581431e-05, + "loss": 0.4173, + "step": 13885 + }, + { + "epoch": 0.53, + "learning_rate": 4.1128494485765583e-05, + "loss": 0.5118, + "step": 13886 + }, + { + "epoch": 0.53, + "learning_rate": 4.112785329571685e-05, + "loss": 0.487, + "step": 13887 + }, + { + "epoch": 0.53, + "learning_rate": 4.112721210566812e-05, + "loss": 0.4324, + "step": 13888 + }, + { + "epoch": 0.53, + "learning_rate": 4.112657091561939e-05, + "loss": 0.5837, + "step": 13889 + }, + { + "epoch": 0.53, + "learning_rate": 4.1125929725570664e-05, + "loss": 0.6375, + "step": 13890 + }, + { + "epoch": 0.53, + "learning_rate": 4.1125288535521935e-05, + "loss": 0.5551, + "step": 13891 + }, + { + "epoch": 0.53, + "learning_rate": 4.11246473454732e-05, + "loss": 0.4525, + "step": 13892 + }, + { + "epoch": 0.53, + "learning_rate": 4.112400615542447e-05, + "loss": 0.5924, + "step": 13893 + }, + { + "epoch": 0.53, + "learning_rate": 4.112336496537574e-05, + "loss": 0.557, + "step": 13894 + }, + { + "epoch": 0.53, + "learning_rate": 4.112272377532701e-05, + "loss": 0.4867, + "step": 13895 + }, + { + "epoch": 0.53, + "learning_rate": 4.1122082585278274e-05, + "loss": 0.3705, + "step": 13896 + }, + { + "epoch": 0.53, + "learning_rate": 4.1121441395229545e-05, + "loss": 0.5629, + "step": 13897 + }, + { + "epoch": 0.53, + "learning_rate": 4.112080020518082e-05, + "loss": 0.4749, + "step": 13898 + }, + { + "epoch": 0.53, + "learning_rate": 4.112015901513209e-05, + "loss": 0.6564, + "step": 13899 + }, + { + "epoch": 0.53, + "learning_rate": 4.111951782508336e-05, + "loss": 0.382, + "step": 13900 + }, + { + "epoch": 0.53, + "learning_rate": 4.1118876635034625e-05, + "loss": 0.5029, + "step": 13901 + }, + { + "epoch": 0.53, + "learning_rate": 4.11182354449859e-05, + "loss": 0.5333, + "step": 13902 + }, + { + "epoch": 0.53, + "learning_rate": 4.111759425493716e-05, + "loss": 0.5087, + "step": 13903 + }, + { + "epoch": 0.53, + "learning_rate": 4.1116953064888434e-05, + "loss": 0.5302, + "step": 13904 + }, + { + "epoch": 0.53, + "learning_rate": 4.1116311874839705e-05, + "loss": 0.5507, + "step": 13905 + }, + { + "epoch": 0.53, + "learning_rate": 4.111567068479098e-05, + "loss": 0.5792, + "step": 13906 + }, + { + "epoch": 0.53, + "learning_rate": 4.111502949474224e-05, + "loss": 0.5698, + "step": 13907 + }, + { + "epoch": 0.53, + "learning_rate": 4.1114388304693514e-05, + "loss": 0.4552, + "step": 13908 + }, + { + "epoch": 0.53, + "learning_rate": 4.1113747114644786e-05, + "loss": 0.5352, + "step": 13909 + }, + { + "epoch": 0.53, + "learning_rate": 4.111310592459605e-05, + "loss": 0.5981, + "step": 13910 + }, + { + "epoch": 0.53, + "learning_rate": 4.111246473454732e-05, + "loss": 0.5049, + "step": 13911 + }, + { + "epoch": 0.53, + "learning_rate": 4.111182354449859e-05, + "loss": 0.4848, + "step": 13912 + }, + { + "epoch": 0.53, + "learning_rate": 4.111118235444986e-05, + "loss": 0.5615, + "step": 13913 + }, + { + "epoch": 0.53, + "learning_rate": 4.111054116440113e-05, + "loss": 0.5501, + "step": 13914 + }, + { + "epoch": 0.53, + "learning_rate": 4.11098999743524e-05, + "loss": 0.4761, + "step": 13915 + }, + { + "epoch": 0.54, + "learning_rate": 4.110925878430367e-05, + "loss": 0.505, + "step": 13916 + }, + { + "epoch": 0.54, + "learning_rate": 4.110861759425494e-05, + "loss": 0.5888, + "step": 13917 + }, + { + "epoch": 0.54, + "learning_rate": 4.110797640420621e-05, + "loss": 0.515, + "step": 13918 + }, + { + "epoch": 0.54, + "learning_rate": 4.1107335214157476e-05, + "loss": 0.6044, + "step": 13919 + }, + { + "epoch": 0.54, + "learning_rate": 4.110669402410875e-05, + "loss": 0.5261, + "step": 13920 + }, + { + "epoch": 0.54, + "learning_rate": 4.110605283406002e-05, + "loss": 0.4797, + "step": 13921 + }, + { + "epoch": 0.54, + "learning_rate": 4.110541164401129e-05, + "loss": 0.4723, + "step": 13922 + }, + { + "epoch": 0.54, + "learning_rate": 4.1104770453962556e-05, + "loss": 0.5608, + "step": 13923 + }, + { + "epoch": 0.54, + "learning_rate": 4.110412926391383e-05, + "loss": 0.5088, + "step": 13924 + }, + { + "epoch": 0.54, + "learning_rate": 4.110348807386509e-05, + "loss": 0.5096, + "step": 13925 + }, + { + "epoch": 0.54, + "learning_rate": 4.1102846883816364e-05, + "loss": 0.4688, + "step": 13926 + }, + { + "epoch": 0.54, + "learning_rate": 4.1102205693767636e-05, + "loss": 0.618, + "step": 13927 + }, + { + "epoch": 0.54, + "learning_rate": 4.11015645037189e-05, + "loss": 0.4292, + "step": 13928 + }, + { + "epoch": 0.54, + "learning_rate": 4.110092331367017e-05, + "loss": 0.5592, + "step": 13929 + }, + { + "epoch": 0.54, + "learning_rate": 4.1100282123621444e-05, + "loss": 0.531, + "step": 13930 + }, + { + "epoch": 0.54, + "learning_rate": 4.1099640933572716e-05, + "loss": 0.6271, + "step": 13931 + }, + { + "epoch": 0.54, + "learning_rate": 4.109899974352398e-05, + "loss": 0.5762, + "step": 13932 + }, + { + "epoch": 0.54, + "learning_rate": 4.109835855347525e-05, + "loss": 0.5217, + "step": 13933 + }, + { + "epoch": 0.54, + "learning_rate": 4.109771736342652e-05, + "loss": 0.608, + "step": 13934 + }, + { + "epoch": 0.54, + "learning_rate": 4.109707617337779e-05, + "loss": 0.6721, + "step": 13935 + }, + { + "epoch": 0.54, + "learning_rate": 4.109643498332906e-05, + "loss": 0.5643, + "step": 13936 + }, + { + "epoch": 0.54, + "learning_rate": 4.109579379328033e-05, + "loss": 0.6004, + "step": 13937 + }, + { + "epoch": 0.54, + "learning_rate": 4.1095152603231604e-05, + "loss": 0.5459, + "step": 13938 + }, + { + "epoch": 0.54, + "learning_rate": 4.109451141318287e-05, + "loss": 0.5953, + "step": 13939 + }, + { + "epoch": 0.54, + "learning_rate": 4.109387022313414e-05, + "loss": 0.5416, + "step": 13940 + }, + { + "epoch": 0.54, + "learning_rate": 4.1093229033085406e-05, + "loss": 0.5441, + "step": 13941 + }, + { + "epoch": 0.54, + "learning_rate": 4.109258784303668e-05, + "loss": 0.6103, + "step": 13942 + }, + { + "epoch": 0.54, + "learning_rate": 4.109194665298794e-05, + "loss": 0.4647, + "step": 13943 + }, + { + "epoch": 0.54, + "learning_rate": 4.1091305462939214e-05, + "loss": 0.5731, + "step": 13944 + }, + { + "epoch": 0.54, + "learning_rate": 4.1090664272890486e-05, + "loss": 0.5155, + "step": 13945 + }, + { + "epoch": 0.54, + "learning_rate": 4.109002308284176e-05, + "loss": 0.5551, + "step": 13946 + }, + { + "epoch": 0.54, + "learning_rate": 4.108938189279303e-05, + "loss": 0.4977, + "step": 13947 + }, + { + "epoch": 0.54, + "learning_rate": 4.1088740702744294e-05, + "loss": 0.5159, + "step": 13948 + }, + { + "epoch": 0.54, + "learning_rate": 4.1088099512695566e-05, + "loss": 0.55, + "step": 13949 + }, + { + "epoch": 0.54, + "learning_rate": 4.108745832264683e-05, + "loss": 0.5862, + "step": 13950 + }, + { + "epoch": 0.54, + "learning_rate": 4.10868171325981e-05, + "loss": 0.5592, + "step": 13951 + }, + { + "epoch": 0.54, + "learning_rate": 4.108617594254937e-05, + "loss": 0.5773, + "step": 13952 + }, + { + "epoch": 0.54, + "learning_rate": 4.1085534752500646e-05, + "loss": 0.4228, + "step": 13953 + }, + { + "epoch": 0.54, + "learning_rate": 4.108489356245192e-05, + "loss": 0.5274, + "step": 13954 + }, + { + "epoch": 0.54, + "learning_rate": 4.108425237240318e-05, + "loss": 0.454, + "step": 13955 + }, + { + "epoch": 0.54, + "learning_rate": 4.1083611182354455e-05, + "loss": 0.5568, + "step": 13956 + }, + { + "epoch": 0.54, + "learning_rate": 4.108296999230572e-05, + "loss": 0.6183, + "step": 13957 + }, + { + "epoch": 0.54, + "learning_rate": 4.108232880225699e-05, + "loss": 0.602, + "step": 13958 + }, + { + "epoch": 0.54, + "learning_rate": 4.1081687612208256e-05, + "loss": 0.5016, + "step": 13959 + }, + { + "epoch": 0.54, + "learning_rate": 4.108104642215953e-05, + "loss": 0.603, + "step": 13960 + }, + { + "epoch": 0.54, + "learning_rate": 4.10804052321108e-05, + "loss": 0.5553, + "step": 13961 + }, + { + "epoch": 0.54, + "learning_rate": 4.107976404206207e-05, + "loss": 0.5405, + "step": 13962 + }, + { + "epoch": 0.54, + "learning_rate": 4.107912285201334e-05, + "loss": 0.5146, + "step": 13963 + }, + { + "epoch": 0.54, + "learning_rate": 4.107848166196461e-05, + "loss": 0.4993, + "step": 13964 + }, + { + "epoch": 0.54, + "learning_rate": 4.107784047191588e-05, + "loss": 0.4142, + "step": 13965 + }, + { + "epoch": 0.54, + "learning_rate": 4.1077199281867145e-05, + "loss": 0.4899, + "step": 13966 + }, + { + "epoch": 0.54, + "learning_rate": 4.1076558091818416e-05, + "loss": 0.5542, + "step": 13967 + }, + { + "epoch": 0.54, + "learning_rate": 4.107591690176968e-05, + "loss": 0.5466, + "step": 13968 + }, + { + "epoch": 0.54, + "learning_rate": 4.107527571172095e-05, + "loss": 0.6022, + "step": 13969 + }, + { + "epoch": 0.54, + "learning_rate": 4.107463452167223e-05, + "loss": 0.6144, + "step": 13970 + }, + { + "epoch": 0.54, + "learning_rate": 4.1073993331623496e-05, + "loss": 0.3967, + "step": 13971 + }, + { + "epoch": 0.54, + "learning_rate": 4.107335214157477e-05, + "loss": 0.6373, + "step": 13972 + }, + { + "epoch": 0.54, + "learning_rate": 4.107271095152603e-05, + "loss": 0.5343, + "step": 13973 + }, + { + "epoch": 0.54, + "learning_rate": 4.1072069761477305e-05, + "loss": 0.5051, + "step": 13974 + }, + { + "epoch": 0.54, + "learning_rate": 4.107142857142857e-05, + "loss": 0.534, + "step": 13975 + }, + { + "epoch": 0.54, + "learning_rate": 4.107078738137984e-05, + "loss": 0.5004, + "step": 13976 + }, + { + "epoch": 0.54, + "learning_rate": 4.107014619133111e-05, + "loss": 0.4749, + "step": 13977 + }, + { + "epoch": 0.54, + "learning_rate": 4.1069505001282385e-05, + "loss": 0.5535, + "step": 13978 + }, + { + "epoch": 0.54, + "learning_rate": 4.106886381123366e-05, + "loss": 0.6237, + "step": 13979 + }, + { + "epoch": 0.54, + "learning_rate": 4.106822262118492e-05, + "loss": 0.6648, + "step": 13980 + }, + { + "epoch": 0.54, + "learning_rate": 4.106758143113619e-05, + "loss": 0.6196, + "step": 13981 + }, + { + "epoch": 0.54, + "learning_rate": 4.106694024108746e-05, + "loss": 0.4818, + "step": 13982 + }, + { + "epoch": 0.54, + "learning_rate": 4.106629905103873e-05, + "loss": 0.5809, + "step": 13983 + }, + { + "epoch": 0.54, + "learning_rate": 4.1065657860989995e-05, + "loss": 0.47, + "step": 13984 + }, + { + "epoch": 0.54, + "learning_rate": 4.1065016670941267e-05, + "loss": 0.4843, + "step": 13985 + }, + { + "epoch": 0.54, + "learning_rate": 4.106437548089254e-05, + "loss": 0.4517, + "step": 13986 + }, + { + "epoch": 0.54, + "learning_rate": 4.106373429084381e-05, + "loss": 0.6959, + "step": 13987 + }, + { + "epoch": 0.54, + "learning_rate": 4.106309310079508e-05, + "loss": 0.5468, + "step": 13988 + }, + { + "epoch": 0.54, + "learning_rate": 4.106245191074635e-05, + "loss": 0.5664, + "step": 13989 + }, + { + "epoch": 0.54, + "learning_rate": 4.106181072069762e-05, + "loss": 0.5804, + "step": 13990 + }, + { + "epoch": 0.54, + "learning_rate": 4.106116953064888e-05, + "loss": 0.5771, + "step": 13991 + }, + { + "epoch": 0.54, + "learning_rate": 4.1060528340600155e-05, + "loss": 0.5042, + "step": 13992 + }, + { + "epoch": 0.54, + "learning_rate": 4.105988715055143e-05, + "loss": 0.5656, + "step": 13993 + }, + { + "epoch": 0.54, + "learning_rate": 4.10592459605027e-05, + "loss": 0.7265, + "step": 13994 + }, + { + "epoch": 0.54, + "learning_rate": 4.1058604770453963e-05, + "loss": 0.5166, + "step": 13995 + }, + { + "epoch": 0.54, + "learning_rate": 4.1057963580405235e-05, + "loss": 0.614, + "step": 13996 + }, + { + "epoch": 0.54, + "learning_rate": 4.105732239035651e-05, + "loss": 0.6223, + "step": 13997 + }, + { + "epoch": 0.54, + "learning_rate": 4.105668120030777e-05, + "loss": 0.5713, + "step": 13998 + }, + { + "epoch": 0.54, + "learning_rate": 4.1056040010259044e-05, + "loss": 0.5331, + "step": 13999 + }, + { + "epoch": 0.54, + "learning_rate": 4.105539882021031e-05, + "loss": 0.465, + "step": 14000 + }, + { + "epoch": 0.54, + "learning_rate": 4.105475763016158e-05, + "loss": 0.4936, + "step": 14001 + }, + { + "epoch": 0.54, + "learning_rate": 4.105411644011285e-05, + "loss": 0.5762, + "step": 14002 + }, + { + "epoch": 0.54, + "learning_rate": 4.1053475250064124e-05, + "loss": 0.4683, + "step": 14003 + }, + { + "epoch": 0.54, + "learning_rate": 4.105283406001539e-05, + "loss": 0.53, + "step": 14004 + }, + { + "epoch": 0.54, + "learning_rate": 4.105219286996666e-05, + "loss": 0.6495, + "step": 14005 + }, + { + "epoch": 0.54, + "learning_rate": 4.105155167991793e-05, + "loss": 0.5149, + "step": 14006 + }, + { + "epoch": 0.54, + "learning_rate": 4.10509104898692e-05, + "loss": 0.5549, + "step": 14007 + }, + { + "epoch": 0.54, + "learning_rate": 4.105026929982047e-05, + "loss": 0.5856, + "step": 14008 + }, + { + "epoch": 0.54, + "learning_rate": 4.104962810977174e-05, + "loss": 0.6919, + "step": 14009 + }, + { + "epoch": 0.54, + "learning_rate": 4.104898691972301e-05, + "loss": 0.5424, + "step": 14010 + }, + { + "epoch": 0.54, + "learning_rate": 4.104834572967428e-05, + "loss": 0.6047, + "step": 14011 + }, + { + "epoch": 0.54, + "learning_rate": 4.104770453962555e-05, + "loss": 0.5258, + "step": 14012 + }, + { + "epoch": 0.54, + "learning_rate": 4.1047063349576814e-05, + "loss": 0.467, + "step": 14013 + }, + { + "epoch": 0.54, + "learning_rate": 4.1046422159528085e-05, + "loss": 0.5157, + "step": 14014 + }, + { + "epoch": 0.54, + "learning_rate": 4.104578096947936e-05, + "loss": 0.6223, + "step": 14015 + }, + { + "epoch": 0.54, + "learning_rate": 4.104513977943062e-05, + "loss": 0.5475, + "step": 14016 + }, + { + "epoch": 0.54, + "learning_rate": 4.1044498589381894e-05, + "loss": 0.4586, + "step": 14017 + }, + { + "epoch": 0.54, + "learning_rate": 4.1043857399333165e-05, + "loss": 0.5621, + "step": 14018 + }, + { + "epoch": 0.54, + "learning_rate": 4.104321620928444e-05, + "loss": 0.4851, + "step": 14019 + }, + { + "epoch": 0.54, + "learning_rate": 4.10425750192357e-05, + "loss": 0.5383, + "step": 14020 + }, + { + "epoch": 0.54, + "learning_rate": 4.1041933829186974e-05, + "loss": 0.482, + "step": 14021 + }, + { + "epoch": 0.54, + "learning_rate": 4.104129263913824e-05, + "loss": 0.6434, + "step": 14022 + }, + { + "epoch": 0.54, + "learning_rate": 4.104065144908951e-05, + "loss": 0.5546, + "step": 14023 + }, + { + "epoch": 0.54, + "learning_rate": 4.104001025904078e-05, + "loss": 0.5299, + "step": 14024 + }, + { + "epoch": 0.54, + "learning_rate": 4.1039369068992054e-05, + "loss": 0.5208, + "step": 14025 + }, + { + "epoch": 0.54, + "learning_rate": 4.1038727878943326e-05, + "loss": 0.5104, + "step": 14026 + }, + { + "epoch": 0.54, + "learning_rate": 4.103808668889459e-05, + "loss": 0.5635, + "step": 14027 + }, + { + "epoch": 0.54, + "learning_rate": 4.103744549884586e-05, + "loss": 0.5324, + "step": 14028 + }, + { + "epoch": 0.54, + "learning_rate": 4.103680430879713e-05, + "loss": 0.5206, + "step": 14029 + }, + { + "epoch": 0.54, + "learning_rate": 4.10361631187484e-05, + "loss": 0.5008, + "step": 14030 + }, + { + "epoch": 0.54, + "learning_rate": 4.1035521928699664e-05, + "loss": 0.5309, + "step": 14031 + }, + { + "epoch": 0.54, + "learning_rate": 4.1034880738650936e-05, + "loss": 0.5257, + "step": 14032 + }, + { + "epoch": 0.54, + "learning_rate": 4.103423954860221e-05, + "loss": 0.5596, + "step": 14033 + }, + { + "epoch": 0.54, + "learning_rate": 4.103359835855348e-05, + "loss": 0.5191, + "step": 14034 + }, + { + "epoch": 0.54, + "learning_rate": 4.103295716850475e-05, + "loss": 0.5533, + "step": 14035 + }, + { + "epoch": 0.54, + "learning_rate": 4.1032315978456016e-05, + "loss": 0.5754, + "step": 14036 + }, + { + "epoch": 0.54, + "learning_rate": 4.103167478840729e-05, + "loss": 0.5331, + "step": 14037 + }, + { + "epoch": 0.54, + "learning_rate": 4.103103359835855e-05, + "loss": 0.5135, + "step": 14038 + }, + { + "epoch": 0.54, + "learning_rate": 4.1030392408309824e-05, + "loss": 0.5453, + "step": 14039 + }, + { + "epoch": 0.54, + "learning_rate": 4.102975121826109e-05, + "loss": 0.4828, + "step": 14040 + }, + { + "epoch": 0.54, + "learning_rate": 4.102911002821237e-05, + "loss": 0.4584, + "step": 14041 + }, + { + "epoch": 0.54, + "learning_rate": 4.102846883816364e-05, + "loss": 0.5823, + "step": 14042 + }, + { + "epoch": 0.54, + "learning_rate": 4.1027827648114904e-05, + "loss": 0.4882, + "step": 14043 + }, + { + "epoch": 0.54, + "learning_rate": 4.1027186458066176e-05, + "loss": 0.5075, + "step": 14044 + }, + { + "epoch": 0.54, + "learning_rate": 4.102654526801744e-05, + "loss": 0.5483, + "step": 14045 + }, + { + "epoch": 0.54, + "learning_rate": 4.102590407796871e-05, + "loss": 0.507, + "step": 14046 + }, + { + "epoch": 0.54, + "learning_rate": 4.102526288791998e-05, + "loss": 0.4965, + "step": 14047 + }, + { + "epoch": 0.54, + "learning_rate": 4.102462169787125e-05, + "loss": 0.6613, + "step": 14048 + }, + { + "epoch": 0.54, + "learning_rate": 4.102398050782252e-05, + "loss": 0.5669, + "step": 14049 + }, + { + "epoch": 0.54, + "learning_rate": 4.102333931777379e-05, + "loss": 0.5908, + "step": 14050 + }, + { + "epoch": 0.54, + "learning_rate": 4.1022698127725064e-05, + "loss": 0.5007, + "step": 14051 + }, + { + "epoch": 0.54, + "learning_rate": 4.102205693767633e-05, + "loss": 0.5386, + "step": 14052 + }, + { + "epoch": 0.54, + "learning_rate": 4.10214157476276e-05, + "loss": 0.4926, + "step": 14053 + }, + { + "epoch": 0.54, + "learning_rate": 4.1020774557578866e-05, + "loss": 0.5347, + "step": 14054 + }, + { + "epoch": 0.54, + "learning_rate": 4.102013336753014e-05, + "loss": 0.5935, + "step": 14055 + }, + { + "epoch": 0.54, + "learning_rate": 4.10194921774814e-05, + "loss": 0.4859, + "step": 14056 + }, + { + "epoch": 0.54, + "learning_rate": 4.1018850987432674e-05, + "loss": 0.4757, + "step": 14057 + }, + { + "epoch": 0.54, + "learning_rate": 4.1018209797383946e-05, + "loss": 0.5124, + "step": 14058 + }, + { + "epoch": 0.54, + "learning_rate": 4.101756860733522e-05, + "loss": 0.4623, + "step": 14059 + }, + { + "epoch": 0.54, + "learning_rate": 4.101692741728649e-05, + "loss": 0.5298, + "step": 14060 + }, + { + "epoch": 0.54, + "learning_rate": 4.1016286227237754e-05, + "loss": 0.5324, + "step": 14061 + }, + { + "epoch": 0.54, + "learning_rate": 4.1015645037189026e-05, + "loss": 0.4994, + "step": 14062 + }, + { + "epoch": 0.54, + "learning_rate": 4.101500384714029e-05, + "loss": 0.4957, + "step": 14063 + }, + { + "epoch": 0.54, + "learning_rate": 4.101436265709156e-05, + "loss": 0.5638, + "step": 14064 + }, + { + "epoch": 0.54, + "learning_rate": 4.1013721467042835e-05, + "loss": 0.5643, + "step": 14065 + }, + { + "epoch": 0.54, + "learning_rate": 4.1013080276994106e-05, + "loss": 0.5696, + "step": 14066 + }, + { + "epoch": 0.54, + "learning_rate": 4.101243908694537e-05, + "loss": 0.5432, + "step": 14067 + }, + { + "epoch": 0.54, + "learning_rate": 4.101179789689664e-05, + "loss": 0.4498, + "step": 14068 + }, + { + "epoch": 0.54, + "learning_rate": 4.1011156706847915e-05, + "loss": 0.4781, + "step": 14069 + }, + { + "epoch": 0.54, + "learning_rate": 4.101051551679918e-05, + "loss": 0.6087, + "step": 14070 + }, + { + "epoch": 0.54, + "learning_rate": 4.100987432675045e-05, + "loss": 0.6423, + "step": 14071 + }, + { + "epoch": 0.54, + "learning_rate": 4.1009233136701716e-05, + "loss": 0.578, + "step": 14072 + }, + { + "epoch": 0.54, + "learning_rate": 4.100859194665299e-05, + "loss": 0.4654, + "step": 14073 + }, + { + "epoch": 0.54, + "learning_rate": 4.100795075660426e-05, + "loss": 0.474, + "step": 14074 + }, + { + "epoch": 0.54, + "learning_rate": 4.100730956655553e-05, + "loss": 0.4542, + "step": 14075 + }, + { + "epoch": 0.54, + "learning_rate": 4.1006668376506796e-05, + "loss": 0.6003, + "step": 14076 + }, + { + "epoch": 0.54, + "learning_rate": 4.100602718645807e-05, + "loss": 0.4692, + "step": 14077 + }, + { + "epoch": 0.54, + "learning_rate": 4.100538599640934e-05, + "loss": 0.5752, + "step": 14078 + }, + { + "epoch": 0.54, + "learning_rate": 4.1004744806360605e-05, + "loss": 0.4578, + "step": 14079 + }, + { + "epoch": 0.54, + "learning_rate": 4.1004103616311876e-05, + "loss": 0.5026, + "step": 14080 + }, + { + "epoch": 0.54, + "learning_rate": 4.100346242626315e-05, + "loss": 0.5639, + "step": 14081 + }, + { + "epoch": 0.54, + "learning_rate": 4.100282123621442e-05, + "loss": 0.4288, + "step": 14082 + }, + { + "epoch": 0.54, + "learning_rate": 4.1002180046165685e-05, + "loss": 0.5267, + "step": 14083 + }, + { + "epoch": 0.54, + "learning_rate": 4.1001538856116956e-05, + "loss": 0.4717, + "step": 14084 + }, + { + "epoch": 0.54, + "learning_rate": 4.100089766606822e-05, + "loss": 0.6007, + "step": 14085 + }, + { + "epoch": 0.54, + "learning_rate": 4.100025647601949e-05, + "loss": 0.4214, + "step": 14086 + }, + { + "epoch": 0.54, + "learning_rate": 4.0999615285970765e-05, + "loss": 0.5682, + "step": 14087 + }, + { + "epoch": 0.54, + "learning_rate": 4.099897409592203e-05, + "loss": 0.5762, + "step": 14088 + }, + { + "epoch": 0.54, + "learning_rate": 4.09983329058733e-05, + "loss": 0.5822, + "step": 14089 + }, + { + "epoch": 0.54, + "learning_rate": 4.099769171582457e-05, + "loss": 0.4803, + "step": 14090 + }, + { + "epoch": 0.54, + "learning_rate": 4.0997050525775845e-05, + "loss": 0.428, + "step": 14091 + }, + { + "epoch": 0.54, + "learning_rate": 4.099640933572711e-05, + "loss": 0.5344, + "step": 14092 + }, + { + "epoch": 0.54, + "learning_rate": 4.099576814567838e-05, + "loss": 0.5008, + "step": 14093 + }, + { + "epoch": 0.54, + "learning_rate": 4.0995126955629647e-05, + "loss": 0.5279, + "step": 14094 + }, + { + "epoch": 0.54, + "learning_rate": 4.099448576558092e-05, + "loss": 0.5468, + "step": 14095 + }, + { + "epoch": 0.54, + "learning_rate": 4.099384457553219e-05, + "loss": 0.4723, + "step": 14096 + }, + { + "epoch": 0.54, + "learning_rate": 4.099320338548346e-05, + "loss": 0.4875, + "step": 14097 + }, + { + "epoch": 0.54, + "learning_rate": 4.0992562195434733e-05, + "loss": 0.5259, + "step": 14098 + }, + { + "epoch": 0.54, + "learning_rate": 4.0991921005386e-05, + "loss": 0.5736, + "step": 14099 + }, + { + "epoch": 0.54, + "learning_rate": 4.099127981533727e-05, + "loss": 0.6499, + "step": 14100 + }, + { + "epoch": 0.54, + "learning_rate": 4.0990638625288535e-05, + "loss": 0.5321, + "step": 14101 + }, + { + "epoch": 0.54, + "learning_rate": 4.098999743523981e-05, + "loss": 0.6262, + "step": 14102 + }, + { + "epoch": 0.54, + "learning_rate": 4.098935624519107e-05, + "loss": 0.4818, + "step": 14103 + }, + { + "epoch": 0.54, + "learning_rate": 4.098871505514234e-05, + "loss": 0.4597, + "step": 14104 + }, + { + "epoch": 0.54, + "learning_rate": 4.0988073865093615e-05, + "loss": 0.5556, + "step": 14105 + }, + { + "epoch": 0.54, + "learning_rate": 4.098743267504489e-05, + "loss": 0.5561, + "step": 14106 + }, + { + "epoch": 0.54, + "learning_rate": 4.098679148499616e-05, + "loss": 0.5221, + "step": 14107 + }, + { + "epoch": 0.54, + "learning_rate": 4.0986150294947423e-05, + "loss": 0.5297, + "step": 14108 + }, + { + "epoch": 0.54, + "learning_rate": 4.0985509104898695e-05, + "loss": 0.4998, + "step": 14109 + }, + { + "epoch": 0.54, + "learning_rate": 4.098486791484996e-05, + "loss": 0.5332, + "step": 14110 + }, + { + "epoch": 0.54, + "learning_rate": 4.098422672480123e-05, + "loss": 0.6082, + "step": 14111 + }, + { + "epoch": 0.54, + "learning_rate": 4.09835855347525e-05, + "loss": 0.5087, + "step": 14112 + }, + { + "epoch": 0.54, + "learning_rate": 4.0982944344703775e-05, + "loss": 0.4222, + "step": 14113 + }, + { + "epoch": 0.54, + "learning_rate": 4.098230315465505e-05, + "loss": 0.5246, + "step": 14114 + }, + { + "epoch": 0.54, + "learning_rate": 4.098166196460631e-05, + "loss": 0.4982, + "step": 14115 + }, + { + "epoch": 0.54, + "learning_rate": 4.0981020774557584e-05, + "loss": 0.4938, + "step": 14116 + }, + { + "epoch": 0.54, + "learning_rate": 4.098037958450885e-05, + "loss": 0.6251, + "step": 14117 + }, + { + "epoch": 0.54, + "learning_rate": 4.097973839446012e-05, + "loss": 0.4222, + "step": 14118 + }, + { + "epoch": 0.54, + "learning_rate": 4.0979097204411385e-05, + "loss": 0.5066, + "step": 14119 + }, + { + "epoch": 0.54, + "learning_rate": 4.097845601436266e-05, + "loss": 0.5038, + "step": 14120 + }, + { + "epoch": 0.54, + "learning_rate": 4.097781482431393e-05, + "loss": 0.7409, + "step": 14121 + }, + { + "epoch": 0.54, + "learning_rate": 4.09771736342652e-05, + "loss": 0.4409, + "step": 14122 + }, + { + "epoch": 0.54, + "learning_rate": 4.097653244421647e-05, + "loss": 0.632, + "step": 14123 + }, + { + "epoch": 0.54, + "learning_rate": 4.097589125416774e-05, + "loss": 0.5573, + "step": 14124 + }, + { + "epoch": 0.54, + "learning_rate": 4.097525006411901e-05, + "loss": 0.544, + "step": 14125 + }, + { + "epoch": 0.54, + "learning_rate": 4.0974608874070274e-05, + "loss": 0.4945, + "step": 14126 + }, + { + "epoch": 0.54, + "learning_rate": 4.0973967684021545e-05, + "loss": 0.5735, + "step": 14127 + }, + { + "epoch": 0.54, + "learning_rate": 4.097332649397281e-05, + "loss": 0.4519, + "step": 14128 + }, + { + "epoch": 0.54, + "learning_rate": 4.097268530392408e-05, + "loss": 0.5377, + "step": 14129 + }, + { + "epoch": 0.54, + "learning_rate": 4.0972044113875354e-05, + "loss": 0.5123, + "step": 14130 + }, + { + "epoch": 0.54, + "learning_rate": 4.0971402923826626e-05, + "loss": 0.4551, + "step": 14131 + }, + { + "epoch": 0.54, + "learning_rate": 4.09707617337779e-05, + "loss": 0.5546, + "step": 14132 + }, + { + "epoch": 0.54, + "learning_rate": 4.097012054372916e-05, + "loss": 0.5251, + "step": 14133 + }, + { + "epoch": 0.54, + "learning_rate": 4.0969479353680434e-05, + "loss": 0.5434, + "step": 14134 + }, + { + "epoch": 0.54, + "learning_rate": 4.09688381636317e-05, + "loss": 0.582, + "step": 14135 + }, + { + "epoch": 0.54, + "learning_rate": 4.096819697358297e-05, + "loss": 0.6954, + "step": 14136 + }, + { + "epoch": 0.54, + "learning_rate": 4.096755578353424e-05, + "loss": 0.4262, + "step": 14137 + }, + { + "epoch": 0.54, + "learning_rate": 4.0966914593485514e-05, + "loss": 0.5299, + "step": 14138 + }, + { + "epoch": 0.54, + "learning_rate": 4.096627340343678e-05, + "loss": 0.4695, + "step": 14139 + }, + { + "epoch": 0.54, + "learning_rate": 4.096563221338805e-05, + "loss": 0.4818, + "step": 14140 + }, + { + "epoch": 0.54, + "learning_rate": 4.096499102333932e-05, + "loss": 0.5513, + "step": 14141 + }, + { + "epoch": 0.54, + "learning_rate": 4.096434983329059e-05, + "loss": 0.6223, + "step": 14142 + }, + { + "epoch": 0.54, + "learning_rate": 4.096370864324186e-05, + "loss": 0.5691, + "step": 14143 + }, + { + "epoch": 0.54, + "learning_rate": 4.0963067453193124e-05, + "loss": 0.6137, + "step": 14144 + }, + { + "epoch": 0.54, + "learning_rate": 4.0962426263144396e-05, + "loss": 0.4789, + "step": 14145 + }, + { + "epoch": 0.54, + "learning_rate": 4.096178507309567e-05, + "loss": 0.484, + "step": 14146 + }, + { + "epoch": 0.54, + "learning_rate": 4.096114388304694e-05, + "loss": 0.6214, + "step": 14147 + }, + { + "epoch": 0.54, + "learning_rate": 4.0960502692998204e-05, + "loss": 0.5864, + "step": 14148 + }, + { + "epoch": 0.54, + "learning_rate": 4.0959861502949476e-05, + "loss": 0.5445, + "step": 14149 + }, + { + "epoch": 0.54, + "learning_rate": 4.095922031290075e-05, + "loss": 0.4708, + "step": 14150 + }, + { + "epoch": 0.54, + "learning_rate": 4.095857912285201e-05, + "loss": 0.5788, + "step": 14151 + }, + { + "epoch": 0.54, + "learning_rate": 4.0957937932803284e-05, + "loss": 0.5147, + "step": 14152 + }, + { + "epoch": 0.54, + "learning_rate": 4.0957296742754556e-05, + "loss": 0.58, + "step": 14153 + }, + { + "epoch": 0.54, + "learning_rate": 4.095665555270583e-05, + "loss": 0.5639, + "step": 14154 + }, + { + "epoch": 0.54, + "learning_rate": 4.095601436265709e-05, + "loss": 0.6084, + "step": 14155 + }, + { + "epoch": 0.54, + "learning_rate": 4.0955373172608364e-05, + "loss": 0.505, + "step": 14156 + }, + { + "epoch": 0.54, + "learning_rate": 4.0954731982559636e-05, + "loss": 0.5178, + "step": 14157 + }, + { + "epoch": 0.54, + "learning_rate": 4.09540907925109e-05, + "loss": 0.4192, + "step": 14158 + }, + { + "epoch": 0.54, + "learning_rate": 4.095344960246217e-05, + "loss": 0.5764, + "step": 14159 + }, + { + "epoch": 0.54, + "learning_rate": 4.095280841241344e-05, + "loss": 0.4743, + "step": 14160 + }, + { + "epoch": 0.54, + "learning_rate": 4.095216722236471e-05, + "loss": 0.5278, + "step": 14161 + }, + { + "epoch": 0.54, + "learning_rate": 4.095152603231598e-05, + "loss": 0.5189, + "step": 14162 + }, + { + "epoch": 0.54, + "learning_rate": 4.095088484226725e-05, + "loss": 0.5626, + "step": 14163 + }, + { + "epoch": 0.54, + "learning_rate": 4.095024365221852e-05, + "loss": 0.4603, + "step": 14164 + }, + { + "epoch": 0.54, + "learning_rate": 4.094960246216979e-05, + "loss": 0.5929, + "step": 14165 + }, + { + "epoch": 0.54, + "learning_rate": 4.094896127212106e-05, + "loss": 0.5344, + "step": 14166 + }, + { + "epoch": 0.54, + "learning_rate": 4.0948320082072326e-05, + "loss": 0.626, + "step": 14167 + }, + { + "epoch": 0.54, + "learning_rate": 4.09476788920236e-05, + "loss": 0.5213, + "step": 14168 + }, + { + "epoch": 0.54, + "learning_rate": 4.094703770197487e-05, + "loss": 0.4979, + "step": 14169 + }, + { + "epoch": 0.54, + "learning_rate": 4.094639651192614e-05, + "loss": 0.5081, + "step": 14170 + }, + { + "epoch": 0.54, + "learning_rate": 4.0945755321877406e-05, + "loss": 0.4937, + "step": 14171 + }, + { + "epoch": 0.54, + "learning_rate": 4.094511413182868e-05, + "loss": 0.6152, + "step": 14172 + }, + { + "epoch": 0.54, + "learning_rate": 4.094447294177994e-05, + "loss": 0.6186, + "step": 14173 + }, + { + "epoch": 0.54, + "learning_rate": 4.0943831751731214e-05, + "loss": 0.4472, + "step": 14174 + }, + { + "epoch": 0.54, + "learning_rate": 4.0943190561682486e-05, + "loss": 0.5652, + "step": 14175 + }, + { + "epoch": 0.55, + "learning_rate": 4.094254937163375e-05, + "loss": 0.5233, + "step": 14176 + }, + { + "epoch": 0.55, + "learning_rate": 4.094190818158502e-05, + "loss": 0.4323, + "step": 14177 + }, + { + "epoch": 0.55, + "learning_rate": 4.0941266991536295e-05, + "loss": 0.5027, + "step": 14178 + }, + { + "epoch": 0.55, + "learning_rate": 4.0940625801487566e-05, + "loss": 0.6144, + "step": 14179 + }, + { + "epoch": 0.55, + "learning_rate": 4.093998461143883e-05, + "loss": 0.467, + "step": 14180 + }, + { + "epoch": 0.55, + "learning_rate": 4.09393434213901e-05, + "loss": 0.605, + "step": 14181 + }, + { + "epoch": 0.55, + "learning_rate": 4.093870223134137e-05, + "loss": 0.6161, + "step": 14182 + }, + { + "epoch": 0.55, + "learning_rate": 4.093806104129264e-05, + "loss": 0.5625, + "step": 14183 + }, + { + "epoch": 0.55, + "learning_rate": 4.093741985124391e-05, + "loss": 0.4345, + "step": 14184 + }, + { + "epoch": 0.55, + "learning_rate": 4.093677866119518e-05, + "loss": 0.5291, + "step": 14185 + }, + { + "epoch": 0.55, + "learning_rate": 4.0936137471146455e-05, + "loss": 0.6276, + "step": 14186 + }, + { + "epoch": 0.55, + "learning_rate": 4.093549628109772e-05, + "loss": 0.5342, + "step": 14187 + }, + { + "epoch": 0.55, + "learning_rate": 4.093485509104899e-05, + "loss": 0.4546, + "step": 14188 + }, + { + "epoch": 0.55, + "learning_rate": 4.0934213901000256e-05, + "loss": 0.5999, + "step": 14189 + }, + { + "epoch": 0.55, + "learning_rate": 4.093357271095153e-05, + "loss": 0.5444, + "step": 14190 + }, + { + "epoch": 0.55, + "learning_rate": 4.093293152090279e-05, + "loss": 0.5089, + "step": 14191 + }, + { + "epoch": 0.55, + "learning_rate": 4.0932290330854065e-05, + "loss": 0.5517, + "step": 14192 + }, + { + "epoch": 0.55, + "learning_rate": 4.0931649140805336e-05, + "loss": 0.5345, + "step": 14193 + }, + { + "epoch": 0.55, + "learning_rate": 4.093100795075661e-05, + "loss": 0.4314, + "step": 14194 + }, + { + "epoch": 0.55, + "learning_rate": 4.093036676070788e-05, + "loss": 0.5836, + "step": 14195 + }, + { + "epoch": 0.55, + "learning_rate": 4.0929725570659145e-05, + "loss": 0.4837, + "step": 14196 + }, + { + "epoch": 0.55, + "learning_rate": 4.0929084380610417e-05, + "loss": 0.4445, + "step": 14197 + }, + { + "epoch": 0.55, + "learning_rate": 4.092844319056168e-05, + "loss": 0.529, + "step": 14198 + }, + { + "epoch": 0.55, + "learning_rate": 4.092780200051295e-05, + "loss": 0.5075, + "step": 14199 + }, + { + "epoch": 0.55, + "learning_rate": 4.092716081046422e-05, + "loss": 0.6732, + "step": 14200 + }, + { + "epoch": 0.55, + "learning_rate": 4.0926519620415497e-05, + "loss": 0.4532, + "step": 14201 + }, + { + "epoch": 0.55, + "learning_rate": 4.092587843036677e-05, + "loss": 0.537, + "step": 14202 + }, + { + "epoch": 0.55, + "learning_rate": 4.092523724031803e-05, + "loss": 0.6632, + "step": 14203 + }, + { + "epoch": 0.55, + "learning_rate": 4.0924596050269305e-05, + "loss": 0.5266, + "step": 14204 + }, + { + "epoch": 0.55, + "learning_rate": 4.092395486022057e-05, + "loss": 0.5994, + "step": 14205 + }, + { + "epoch": 0.55, + "learning_rate": 4.092331367017184e-05, + "loss": 0.5289, + "step": 14206 + }, + { + "epoch": 0.55, + "learning_rate": 4.0922672480123107e-05, + "loss": 0.578, + "step": 14207 + }, + { + "epoch": 0.55, + "learning_rate": 4.092203129007438e-05, + "loss": 0.5346, + "step": 14208 + }, + { + "epoch": 0.55, + "learning_rate": 4.092139010002565e-05, + "loss": 0.4908, + "step": 14209 + }, + { + "epoch": 0.55, + "learning_rate": 4.092074890997692e-05, + "loss": 0.5836, + "step": 14210 + }, + { + "epoch": 0.55, + "learning_rate": 4.0920107719928193e-05, + "loss": 0.5341, + "step": 14211 + }, + { + "epoch": 0.55, + "learning_rate": 4.091946652987946e-05, + "loss": 0.5853, + "step": 14212 + }, + { + "epoch": 0.55, + "learning_rate": 4.091882533983073e-05, + "loss": 0.6157, + "step": 14213 + }, + { + "epoch": 0.55, + "learning_rate": 4.0918184149781995e-05, + "loss": 0.5317, + "step": 14214 + }, + { + "epoch": 0.55, + "learning_rate": 4.091754295973327e-05, + "loss": 0.5172, + "step": 14215 + }, + { + "epoch": 0.55, + "learning_rate": 4.091690176968453e-05, + "loss": 0.6029, + "step": 14216 + }, + { + "epoch": 0.55, + "learning_rate": 4.0916260579635803e-05, + "loss": 0.5912, + "step": 14217 + }, + { + "epoch": 0.55, + "learning_rate": 4.0915619389587075e-05, + "loss": 0.5993, + "step": 14218 + }, + { + "epoch": 0.55, + "learning_rate": 4.091497819953835e-05, + "loss": 0.4845, + "step": 14219 + }, + { + "epoch": 0.55, + "learning_rate": 4.091433700948962e-05, + "loss": 0.5302, + "step": 14220 + }, + { + "epoch": 0.55, + "learning_rate": 4.0913695819440884e-05, + "loss": 0.6127, + "step": 14221 + }, + { + "epoch": 0.55, + "learning_rate": 4.0913054629392155e-05, + "loss": 0.5214, + "step": 14222 + }, + { + "epoch": 0.55, + "learning_rate": 4.091241343934342e-05, + "loss": 0.501, + "step": 14223 + }, + { + "epoch": 0.55, + "learning_rate": 4.091177224929469e-05, + "loss": 0.5511, + "step": 14224 + }, + { + "epoch": 0.55, + "learning_rate": 4.0911131059245964e-05, + "loss": 0.6608, + "step": 14225 + }, + { + "epoch": 0.55, + "learning_rate": 4.0910489869197235e-05, + "loss": 0.5422, + "step": 14226 + }, + { + "epoch": 0.55, + "learning_rate": 4.09098486791485e-05, + "loss": 0.5563, + "step": 14227 + }, + { + "epoch": 0.55, + "learning_rate": 4.090920748909977e-05, + "loss": 0.5314, + "step": 14228 + }, + { + "epoch": 0.55, + "learning_rate": 4.0908566299051044e-05, + "loss": 0.4605, + "step": 14229 + }, + { + "epoch": 0.55, + "learning_rate": 4.090792510900231e-05, + "loss": 0.62, + "step": 14230 + }, + { + "epoch": 0.55, + "learning_rate": 4.090728391895358e-05, + "loss": 0.5636, + "step": 14231 + }, + { + "epoch": 0.55, + "learning_rate": 4.0906642728904845e-05, + "loss": 0.518, + "step": 14232 + }, + { + "epoch": 0.55, + "learning_rate": 4.090600153885612e-05, + "loss": 0.4912, + "step": 14233 + }, + { + "epoch": 0.55, + "learning_rate": 4.090536034880739e-05, + "loss": 0.6567, + "step": 14234 + }, + { + "epoch": 0.55, + "learning_rate": 4.090471915875866e-05, + "loss": 0.4572, + "step": 14235 + }, + { + "epoch": 0.55, + "learning_rate": 4.0904077968709925e-05, + "loss": 0.4181, + "step": 14236 + }, + { + "epoch": 0.55, + "learning_rate": 4.09034367786612e-05, + "loss": 0.4595, + "step": 14237 + }, + { + "epoch": 0.55, + "learning_rate": 4.090279558861247e-05, + "loss": 0.5569, + "step": 14238 + }, + { + "epoch": 0.55, + "learning_rate": 4.0902154398563734e-05, + "loss": 0.5089, + "step": 14239 + }, + { + "epoch": 0.55, + "learning_rate": 4.0901513208515005e-05, + "loss": 0.4972, + "step": 14240 + }, + { + "epoch": 0.55, + "learning_rate": 4.090087201846628e-05, + "loss": 0.5762, + "step": 14241 + }, + { + "epoch": 0.55, + "learning_rate": 4.090023082841755e-05, + "loss": 0.4415, + "step": 14242 + }, + { + "epoch": 0.55, + "learning_rate": 4.0899589638368814e-05, + "loss": 0.6411, + "step": 14243 + }, + { + "epoch": 0.55, + "learning_rate": 4.0898948448320086e-05, + "loss": 0.5297, + "step": 14244 + }, + { + "epoch": 0.55, + "learning_rate": 4.089830725827135e-05, + "loss": 0.6593, + "step": 14245 + }, + { + "epoch": 0.55, + "learning_rate": 4.089766606822262e-05, + "loss": 0.5077, + "step": 14246 + }, + { + "epoch": 0.55, + "learning_rate": 4.0897024878173894e-05, + "loss": 0.5272, + "step": 14247 + }, + { + "epoch": 0.55, + "learning_rate": 4.089638368812516e-05, + "loss": 0.6239, + "step": 14248 + }, + { + "epoch": 0.55, + "learning_rate": 4.089574249807643e-05, + "loss": 0.4577, + "step": 14249 + }, + { + "epoch": 0.55, + "learning_rate": 4.08951013080277e-05, + "loss": 0.5301, + "step": 14250 + }, + { + "epoch": 0.55, + "learning_rate": 4.0894460117978974e-05, + "loss": 0.5242, + "step": 14251 + }, + { + "epoch": 0.55, + "learning_rate": 4.089381892793024e-05, + "loss": 0.4371, + "step": 14252 + }, + { + "epoch": 0.55, + "learning_rate": 4.089317773788151e-05, + "loss": 0.522, + "step": 14253 + }, + { + "epoch": 0.55, + "learning_rate": 4.0892536547832776e-05, + "loss": 0.5243, + "step": 14254 + }, + { + "epoch": 0.55, + "learning_rate": 4.089189535778405e-05, + "loss": 0.5798, + "step": 14255 + }, + { + "epoch": 0.55, + "learning_rate": 4.089125416773532e-05, + "loss": 0.5475, + "step": 14256 + }, + { + "epoch": 0.55, + "learning_rate": 4.089061297768659e-05, + "loss": 0.5446, + "step": 14257 + }, + { + "epoch": 0.55, + "learning_rate": 4.088997178763786e-05, + "loss": 0.522, + "step": 14258 + }, + { + "epoch": 0.55, + "learning_rate": 4.088933059758913e-05, + "loss": 0.5658, + "step": 14259 + }, + { + "epoch": 0.55, + "learning_rate": 4.08886894075404e-05, + "loss": 0.5396, + "step": 14260 + }, + { + "epoch": 0.55, + "learning_rate": 4.0888048217491664e-05, + "loss": 0.5241, + "step": 14261 + }, + { + "epoch": 0.55, + "learning_rate": 4.0887407027442936e-05, + "loss": 0.5436, + "step": 14262 + }, + { + "epoch": 0.55, + "learning_rate": 4.08867658373942e-05, + "loss": 0.5698, + "step": 14263 + }, + { + "epoch": 0.55, + "learning_rate": 4.088612464734547e-05, + "loss": 0.5917, + "step": 14264 + }, + { + "epoch": 0.55, + "learning_rate": 4.0885483457296744e-05, + "loss": 0.4872, + "step": 14265 + }, + { + "epoch": 0.55, + "learning_rate": 4.0884842267248016e-05, + "loss": 0.4923, + "step": 14266 + }, + { + "epoch": 0.55, + "learning_rate": 4.088420107719929e-05, + "loss": 0.5409, + "step": 14267 + }, + { + "epoch": 0.55, + "learning_rate": 4.088355988715055e-05, + "loss": 0.55, + "step": 14268 + }, + { + "epoch": 0.55, + "learning_rate": 4.0882918697101824e-05, + "loss": 0.6478, + "step": 14269 + }, + { + "epoch": 0.55, + "learning_rate": 4.088227750705309e-05, + "loss": 0.6159, + "step": 14270 + }, + { + "epoch": 0.55, + "learning_rate": 4.088163631700436e-05, + "loss": 0.6188, + "step": 14271 + }, + { + "epoch": 0.55, + "learning_rate": 4.0880995126955626e-05, + "loss": 0.556, + "step": 14272 + }, + { + "epoch": 0.55, + "learning_rate": 4.0880353936906904e-05, + "loss": 0.5574, + "step": 14273 + }, + { + "epoch": 0.55, + "learning_rate": 4.0879712746858176e-05, + "loss": 0.4942, + "step": 14274 + }, + { + "epoch": 0.55, + "learning_rate": 4.087907155680944e-05, + "loss": 0.4944, + "step": 14275 + }, + { + "epoch": 0.55, + "learning_rate": 4.087843036676071e-05, + "loss": 0.5107, + "step": 14276 + }, + { + "epoch": 0.55, + "learning_rate": 4.087778917671198e-05, + "loss": 0.5195, + "step": 14277 + }, + { + "epoch": 0.55, + "learning_rate": 4.087714798666325e-05, + "loss": 0.5608, + "step": 14278 + }, + { + "epoch": 0.55, + "learning_rate": 4.0876506796614514e-05, + "loss": 0.5195, + "step": 14279 + }, + { + "epoch": 0.55, + "learning_rate": 4.0875865606565786e-05, + "loss": 0.5292, + "step": 14280 + }, + { + "epoch": 0.55, + "learning_rate": 4.087522441651706e-05, + "loss": 0.6467, + "step": 14281 + }, + { + "epoch": 0.55, + "learning_rate": 4.087458322646833e-05, + "loss": 0.5306, + "step": 14282 + }, + { + "epoch": 0.55, + "learning_rate": 4.08739420364196e-05, + "loss": 0.5809, + "step": 14283 + }, + { + "epoch": 0.55, + "learning_rate": 4.0873300846370866e-05, + "loss": 0.5476, + "step": 14284 + }, + { + "epoch": 0.55, + "learning_rate": 4.087265965632214e-05, + "loss": 0.5287, + "step": 14285 + }, + { + "epoch": 0.55, + "learning_rate": 4.08720184662734e-05, + "loss": 0.5684, + "step": 14286 + }, + { + "epoch": 0.55, + "learning_rate": 4.0871377276224675e-05, + "loss": 0.5967, + "step": 14287 + }, + { + "epoch": 0.55, + "learning_rate": 4.087073608617594e-05, + "loss": 0.5497, + "step": 14288 + }, + { + "epoch": 0.55, + "learning_rate": 4.087009489612721e-05, + "loss": 0.4805, + "step": 14289 + }, + { + "epoch": 0.55, + "learning_rate": 4.086945370607848e-05, + "loss": 0.4638, + "step": 14290 + }, + { + "epoch": 0.55, + "learning_rate": 4.0868812516029755e-05, + "loss": 0.6153, + "step": 14291 + }, + { + "epoch": 0.55, + "learning_rate": 4.0868171325981026e-05, + "loss": 0.5545, + "step": 14292 + }, + { + "epoch": 0.55, + "learning_rate": 4.086753013593229e-05, + "loss": 0.4707, + "step": 14293 + }, + { + "epoch": 0.55, + "learning_rate": 4.086688894588356e-05, + "loss": 0.5711, + "step": 14294 + }, + { + "epoch": 0.55, + "learning_rate": 4.086624775583483e-05, + "loss": 0.5104, + "step": 14295 + }, + { + "epoch": 0.55, + "learning_rate": 4.08656065657861e-05, + "loss": 0.6746, + "step": 14296 + }, + { + "epoch": 0.55, + "learning_rate": 4.086496537573737e-05, + "loss": 0.5156, + "step": 14297 + }, + { + "epoch": 0.55, + "learning_rate": 4.086432418568864e-05, + "loss": 0.5729, + "step": 14298 + }, + { + "epoch": 0.55, + "learning_rate": 4.086368299563991e-05, + "loss": 0.6436, + "step": 14299 + }, + { + "epoch": 0.55, + "learning_rate": 4.086304180559118e-05, + "loss": 0.6031, + "step": 14300 + }, + { + "epoch": 0.55, + "learning_rate": 4.086240061554245e-05, + "loss": 0.4949, + "step": 14301 + }, + { + "epoch": 0.55, + "learning_rate": 4.0861759425493716e-05, + "loss": 0.4351, + "step": 14302 + }, + { + "epoch": 0.55, + "learning_rate": 4.086111823544499e-05, + "loss": 0.4751, + "step": 14303 + }, + { + "epoch": 0.55, + "learning_rate": 4.086047704539625e-05, + "loss": 0.5025, + "step": 14304 + }, + { + "epoch": 0.55, + "learning_rate": 4.0859835855347525e-05, + "loss": 0.6102, + "step": 14305 + }, + { + "epoch": 0.55, + "learning_rate": 4.0859194665298796e-05, + "loss": 0.5642, + "step": 14306 + }, + { + "epoch": 0.55, + "learning_rate": 4.085855347525007e-05, + "loss": 0.594, + "step": 14307 + }, + { + "epoch": 0.55, + "learning_rate": 4.085791228520133e-05, + "loss": 0.4807, + "step": 14308 + }, + { + "epoch": 0.55, + "learning_rate": 4.0857271095152605e-05, + "loss": 0.468, + "step": 14309 + }, + { + "epoch": 0.55, + "learning_rate": 4.0856629905103877e-05, + "loss": 0.4695, + "step": 14310 + }, + { + "epoch": 0.55, + "learning_rate": 4.085598871505514e-05, + "loss": 0.6433, + "step": 14311 + }, + { + "epoch": 0.55, + "learning_rate": 4.085534752500641e-05, + "loss": 0.5529, + "step": 14312 + }, + { + "epoch": 0.55, + "learning_rate": 4.0854706334957685e-05, + "loss": 0.6177, + "step": 14313 + }, + { + "epoch": 0.55, + "learning_rate": 4.085406514490896e-05, + "loss": 0.6003, + "step": 14314 + }, + { + "epoch": 0.55, + "learning_rate": 4.085342395486022e-05, + "loss": 0.6018, + "step": 14315 + }, + { + "epoch": 0.55, + "learning_rate": 4.085278276481149e-05, + "loss": 0.5567, + "step": 14316 + }, + { + "epoch": 0.55, + "learning_rate": 4.085214157476276e-05, + "loss": 0.4337, + "step": 14317 + }, + { + "epoch": 0.55, + "learning_rate": 4.085150038471403e-05, + "loss": 0.5752, + "step": 14318 + }, + { + "epoch": 0.55, + "learning_rate": 4.08508591946653e-05, + "loss": 0.5413, + "step": 14319 + }, + { + "epoch": 0.55, + "learning_rate": 4.085021800461657e-05, + "loss": 0.5151, + "step": 14320 + }, + { + "epoch": 0.55, + "learning_rate": 4.084957681456784e-05, + "loss": 0.5879, + "step": 14321 + }, + { + "epoch": 0.55, + "learning_rate": 4.084893562451911e-05, + "loss": 0.5274, + "step": 14322 + }, + { + "epoch": 0.55, + "learning_rate": 4.084829443447038e-05, + "loss": 0.5031, + "step": 14323 + }, + { + "epoch": 0.55, + "learning_rate": 4.084765324442165e-05, + "loss": 0.6269, + "step": 14324 + }, + { + "epoch": 0.55, + "learning_rate": 4.084701205437292e-05, + "loss": 0.617, + "step": 14325 + }, + { + "epoch": 0.55, + "learning_rate": 4.084637086432418e-05, + "loss": 0.5779, + "step": 14326 + }, + { + "epoch": 0.55, + "learning_rate": 4.0845729674275455e-05, + "loss": 0.5454, + "step": 14327 + }, + { + "epoch": 0.55, + "learning_rate": 4.084508848422673e-05, + "loss": 0.5194, + "step": 14328 + }, + { + "epoch": 0.55, + "learning_rate": 4.0844447294178e-05, + "loss": 0.5454, + "step": 14329 + }, + { + "epoch": 0.55, + "learning_rate": 4.084380610412927e-05, + "loss": 0.5453, + "step": 14330 + }, + { + "epoch": 0.55, + "learning_rate": 4.0843164914080535e-05, + "loss": 0.5185, + "step": 14331 + }, + { + "epoch": 0.55, + "learning_rate": 4.084252372403181e-05, + "loss": 0.5371, + "step": 14332 + }, + { + "epoch": 0.55, + "learning_rate": 4.084188253398307e-05, + "loss": 0.6287, + "step": 14333 + }, + { + "epoch": 0.55, + "learning_rate": 4.0841241343934344e-05, + "loss": 0.5595, + "step": 14334 + }, + { + "epoch": 0.55, + "learning_rate": 4.084060015388561e-05, + "loss": 0.5608, + "step": 14335 + }, + { + "epoch": 0.55, + "learning_rate": 4.083995896383688e-05, + "loss": 0.5189, + "step": 14336 + }, + { + "epoch": 0.55, + "learning_rate": 4.083931777378815e-05, + "loss": 0.4781, + "step": 14337 + }, + { + "epoch": 0.55, + "learning_rate": 4.0838676583739424e-05, + "loss": 0.558, + "step": 14338 + }, + { + "epoch": 0.55, + "learning_rate": 4.0838035393690695e-05, + "loss": 0.4753, + "step": 14339 + }, + { + "epoch": 0.55, + "learning_rate": 4.083739420364196e-05, + "loss": 0.6213, + "step": 14340 + }, + { + "epoch": 0.55, + "learning_rate": 4.083675301359323e-05, + "loss": 0.5062, + "step": 14341 + }, + { + "epoch": 0.55, + "learning_rate": 4.08361118235445e-05, + "loss": 0.6571, + "step": 14342 + }, + { + "epoch": 0.55, + "learning_rate": 4.083547063349577e-05, + "loss": 0.4743, + "step": 14343 + }, + { + "epoch": 0.55, + "learning_rate": 4.083482944344704e-05, + "loss": 0.445, + "step": 14344 + }, + { + "epoch": 0.55, + "learning_rate": 4.083418825339831e-05, + "loss": 0.488, + "step": 14345 + }, + { + "epoch": 0.55, + "learning_rate": 4.0833547063349584e-05, + "loss": 0.5566, + "step": 14346 + }, + { + "epoch": 0.55, + "learning_rate": 4.083290587330085e-05, + "loss": 0.5127, + "step": 14347 + }, + { + "epoch": 0.55, + "learning_rate": 4.083226468325212e-05, + "loss": 0.6329, + "step": 14348 + }, + { + "epoch": 0.55, + "learning_rate": 4.0831623493203385e-05, + "loss": 0.4916, + "step": 14349 + }, + { + "epoch": 0.55, + "learning_rate": 4.083098230315466e-05, + "loss": 0.5674, + "step": 14350 + }, + { + "epoch": 0.55, + "learning_rate": 4.083034111310592e-05, + "loss": 0.6095, + "step": 14351 + }, + { + "epoch": 0.55, + "learning_rate": 4.0829699923057194e-05, + "loss": 0.5547, + "step": 14352 + }, + { + "epoch": 0.55, + "learning_rate": 4.0829058733008465e-05, + "loss": 0.5397, + "step": 14353 + }, + { + "epoch": 0.55, + "learning_rate": 4.082841754295974e-05, + "loss": 0.6095, + "step": 14354 + }, + { + "epoch": 0.55, + "learning_rate": 4.082777635291101e-05, + "loss": 0.6202, + "step": 14355 + }, + { + "epoch": 0.55, + "learning_rate": 4.0827135162862274e-05, + "loss": 0.545, + "step": 14356 + }, + { + "epoch": 0.55, + "learning_rate": 4.0826493972813546e-05, + "loss": 0.5369, + "step": 14357 + }, + { + "epoch": 0.55, + "learning_rate": 4.082585278276481e-05, + "loss": 0.4205, + "step": 14358 + }, + { + "epoch": 0.55, + "learning_rate": 4.082521159271608e-05, + "loss": 0.5137, + "step": 14359 + }, + { + "epoch": 0.55, + "learning_rate": 4.082457040266735e-05, + "loss": 0.5519, + "step": 14360 + }, + { + "epoch": 0.55, + "learning_rate": 4.0823929212618626e-05, + "loss": 0.5996, + "step": 14361 + }, + { + "epoch": 0.55, + "learning_rate": 4.08232880225699e-05, + "loss": 0.5084, + "step": 14362 + }, + { + "epoch": 0.55, + "learning_rate": 4.082264683252116e-05, + "loss": 0.503, + "step": 14363 + }, + { + "epoch": 0.55, + "learning_rate": 4.0822005642472434e-05, + "loss": 0.5964, + "step": 14364 + }, + { + "epoch": 0.55, + "learning_rate": 4.08213644524237e-05, + "loss": 0.6975, + "step": 14365 + }, + { + "epoch": 0.55, + "learning_rate": 4.082072326237497e-05, + "loss": 0.6685, + "step": 14366 + }, + { + "epoch": 0.55, + "learning_rate": 4.0820082072326236e-05, + "loss": 0.5221, + "step": 14367 + }, + { + "epoch": 0.55, + "learning_rate": 4.081944088227751e-05, + "loss": 0.5567, + "step": 14368 + }, + { + "epoch": 0.55, + "learning_rate": 4.081879969222878e-05, + "loss": 0.6816, + "step": 14369 + }, + { + "epoch": 0.55, + "learning_rate": 4.081815850218005e-05, + "loss": 0.5196, + "step": 14370 + }, + { + "epoch": 0.55, + "learning_rate": 4.081751731213132e-05, + "loss": 0.6195, + "step": 14371 + }, + { + "epoch": 0.55, + "learning_rate": 4.081687612208259e-05, + "loss": 0.5651, + "step": 14372 + }, + { + "epoch": 0.55, + "learning_rate": 4.081623493203386e-05, + "loss": 0.5858, + "step": 14373 + }, + { + "epoch": 0.55, + "learning_rate": 4.0815593741985124e-05, + "loss": 0.5253, + "step": 14374 + }, + { + "epoch": 0.55, + "learning_rate": 4.0814952551936396e-05, + "loss": 0.6152, + "step": 14375 + }, + { + "epoch": 0.55, + "learning_rate": 4.081431136188766e-05, + "loss": 0.621, + "step": 14376 + }, + { + "epoch": 0.55, + "learning_rate": 4.081367017183893e-05, + "loss": 0.4604, + "step": 14377 + }, + { + "epoch": 0.55, + "learning_rate": 4.0813028981790204e-05, + "loss": 0.5405, + "step": 14378 + }, + { + "epoch": 0.55, + "learning_rate": 4.0812387791741476e-05, + "loss": 0.5574, + "step": 14379 + }, + { + "epoch": 0.55, + "learning_rate": 4.081174660169275e-05, + "loss": 0.7167, + "step": 14380 + }, + { + "epoch": 0.55, + "learning_rate": 4.081110541164401e-05, + "loss": 0.5668, + "step": 14381 + }, + { + "epoch": 0.55, + "learning_rate": 4.0810464221595284e-05, + "loss": 0.5325, + "step": 14382 + }, + { + "epoch": 0.55, + "learning_rate": 4.080982303154655e-05, + "loss": 0.5467, + "step": 14383 + }, + { + "epoch": 0.55, + "learning_rate": 4.080918184149782e-05, + "loss": 0.494, + "step": 14384 + }, + { + "epoch": 0.55, + "learning_rate": 4.080854065144909e-05, + "loss": 0.6176, + "step": 14385 + }, + { + "epoch": 0.55, + "learning_rate": 4.0807899461400364e-05, + "loss": 0.474, + "step": 14386 + }, + { + "epoch": 0.55, + "learning_rate": 4.080725827135163e-05, + "loss": 0.6368, + "step": 14387 + }, + { + "epoch": 0.55, + "learning_rate": 4.08066170813029e-05, + "loss": 0.4229, + "step": 14388 + }, + { + "epoch": 0.55, + "learning_rate": 4.080597589125417e-05, + "loss": 0.518, + "step": 14389 + }, + { + "epoch": 0.55, + "learning_rate": 4.080533470120544e-05, + "loss": 0.5807, + "step": 14390 + }, + { + "epoch": 0.55, + "learning_rate": 4.080469351115671e-05, + "loss": 0.5228, + "step": 14391 + }, + { + "epoch": 0.55, + "learning_rate": 4.0804052321107974e-05, + "loss": 0.4584, + "step": 14392 + }, + { + "epoch": 0.55, + "learning_rate": 4.0803411131059246e-05, + "loss": 0.5571, + "step": 14393 + }, + { + "epoch": 0.55, + "learning_rate": 4.080276994101052e-05, + "loss": 0.5857, + "step": 14394 + }, + { + "epoch": 0.55, + "learning_rate": 4.080212875096179e-05, + "loss": 0.5485, + "step": 14395 + }, + { + "epoch": 0.55, + "learning_rate": 4.0801487560913054e-05, + "loss": 0.4969, + "step": 14396 + }, + { + "epoch": 0.55, + "learning_rate": 4.0800846370864326e-05, + "loss": 0.4782, + "step": 14397 + }, + { + "epoch": 0.55, + "learning_rate": 4.08002051808156e-05, + "loss": 0.5293, + "step": 14398 + }, + { + "epoch": 0.55, + "learning_rate": 4.079956399076686e-05, + "loss": 0.4851, + "step": 14399 + }, + { + "epoch": 0.55, + "learning_rate": 4.0798922800718135e-05, + "loss": 0.5896, + "step": 14400 + }, + { + "epoch": 0.55, + "learning_rate": 4.0798281610669406e-05, + "loss": 0.5702, + "step": 14401 + }, + { + "epoch": 0.55, + "learning_rate": 4.079764042062068e-05, + "loss": 0.5843, + "step": 14402 + }, + { + "epoch": 0.55, + "learning_rate": 4.079699923057194e-05, + "loss": 0.5411, + "step": 14403 + }, + { + "epoch": 0.55, + "learning_rate": 4.0796358040523215e-05, + "loss": 0.6047, + "step": 14404 + }, + { + "epoch": 0.55, + "learning_rate": 4.079571685047448e-05, + "loss": 0.5083, + "step": 14405 + }, + { + "epoch": 0.55, + "learning_rate": 4.079507566042575e-05, + "loss": 0.4831, + "step": 14406 + }, + { + "epoch": 0.55, + "learning_rate": 4.079443447037702e-05, + "loss": 0.4656, + "step": 14407 + }, + { + "epoch": 0.55, + "learning_rate": 4.079379328032829e-05, + "loss": 0.5098, + "step": 14408 + }, + { + "epoch": 0.55, + "learning_rate": 4.079315209027956e-05, + "loss": 0.4834, + "step": 14409 + }, + { + "epoch": 0.55, + "learning_rate": 4.079251090023083e-05, + "loss": 0.5121, + "step": 14410 + }, + { + "epoch": 0.55, + "learning_rate": 4.07918697101821e-05, + "loss": 0.5747, + "step": 14411 + }, + { + "epoch": 0.55, + "learning_rate": 4.079122852013337e-05, + "loss": 0.6002, + "step": 14412 + }, + { + "epoch": 0.55, + "learning_rate": 4.079058733008464e-05, + "loss": 0.5601, + "step": 14413 + }, + { + "epoch": 0.55, + "learning_rate": 4.0789946140035905e-05, + "loss": 0.4875, + "step": 14414 + }, + { + "epoch": 0.55, + "learning_rate": 4.0789304949987176e-05, + "loss": 0.6266, + "step": 14415 + }, + { + "epoch": 0.55, + "learning_rate": 4.078866375993845e-05, + "loss": 0.5695, + "step": 14416 + }, + { + "epoch": 0.55, + "learning_rate": 4.078802256988972e-05, + "loss": 0.5022, + "step": 14417 + }, + { + "epoch": 0.55, + "learning_rate": 4.078738137984099e-05, + "loss": 0.5004, + "step": 14418 + }, + { + "epoch": 0.55, + "learning_rate": 4.0786740189792256e-05, + "loss": 0.5025, + "step": 14419 + }, + { + "epoch": 0.55, + "learning_rate": 4.078609899974353e-05, + "loss": 0.6067, + "step": 14420 + }, + { + "epoch": 0.55, + "learning_rate": 4.078545780969479e-05, + "loss": 0.51, + "step": 14421 + }, + { + "epoch": 0.55, + "learning_rate": 4.0784816619646065e-05, + "loss": 0.5346, + "step": 14422 + }, + { + "epoch": 0.55, + "learning_rate": 4.078417542959733e-05, + "loss": 0.6039, + "step": 14423 + }, + { + "epoch": 0.55, + "learning_rate": 4.07835342395486e-05, + "loss": 0.6061, + "step": 14424 + }, + { + "epoch": 0.55, + "learning_rate": 4.078289304949987e-05, + "loss": 0.5539, + "step": 14425 + }, + { + "epoch": 0.55, + "learning_rate": 4.0782251859451145e-05, + "loss": 0.6363, + "step": 14426 + }, + { + "epoch": 0.55, + "learning_rate": 4.078161066940242e-05, + "loss": 0.509, + "step": 14427 + }, + { + "epoch": 0.55, + "learning_rate": 4.078096947935368e-05, + "loss": 0.5882, + "step": 14428 + }, + { + "epoch": 0.55, + "learning_rate": 4.078032828930495e-05, + "loss": 0.5551, + "step": 14429 + }, + { + "epoch": 0.55, + "learning_rate": 4.077968709925622e-05, + "loss": 0.5127, + "step": 14430 + }, + { + "epoch": 0.55, + "learning_rate": 4.077904590920749e-05, + "loss": 0.5397, + "step": 14431 + }, + { + "epoch": 0.55, + "learning_rate": 4.0778404719158755e-05, + "loss": 0.6084, + "step": 14432 + }, + { + "epoch": 0.55, + "learning_rate": 4.0777763529110033e-05, + "loss": 0.4062, + "step": 14433 + }, + { + "epoch": 0.55, + "learning_rate": 4.0777122339061305e-05, + "loss": 0.5383, + "step": 14434 + }, + { + "epoch": 0.55, + "learning_rate": 4.077648114901257e-05, + "loss": 0.5117, + "step": 14435 + }, + { + "epoch": 0.56, + "learning_rate": 4.077583995896384e-05, + "loss": 0.5043, + "step": 14436 + }, + { + "epoch": 0.56, + "learning_rate": 4.077519876891511e-05, + "loss": 0.5216, + "step": 14437 + }, + { + "epoch": 0.56, + "learning_rate": 4.077455757886638e-05, + "loss": 0.455, + "step": 14438 + }, + { + "epoch": 0.56, + "learning_rate": 4.0773916388817643e-05, + "loss": 0.6572, + "step": 14439 + }, + { + "epoch": 0.56, + "learning_rate": 4.0773275198768915e-05, + "loss": 0.4757, + "step": 14440 + }, + { + "epoch": 0.56, + "learning_rate": 4.077263400872019e-05, + "loss": 0.4636, + "step": 14441 + }, + { + "epoch": 0.56, + "learning_rate": 4.077199281867146e-05, + "loss": 0.4513, + "step": 14442 + }, + { + "epoch": 0.56, + "learning_rate": 4.077135162862273e-05, + "loss": 0.5815, + "step": 14443 + }, + { + "epoch": 0.56, + "learning_rate": 4.0770710438573995e-05, + "loss": 0.4999, + "step": 14444 + }, + { + "epoch": 0.56, + "learning_rate": 4.077006924852527e-05, + "loss": 0.5605, + "step": 14445 + }, + { + "epoch": 0.56, + "learning_rate": 4.076942805847653e-05, + "loss": 0.5516, + "step": 14446 + }, + { + "epoch": 0.56, + "learning_rate": 4.0768786868427804e-05, + "loss": 0.5677, + "step": 14447 + }, + { + "epoch": 0.56, + "learning_rate": 4.076814567837907e-05, + "loss": 0.5899, + "step": 14448 + }, + { + "epoch": 0.56, + "learning_rate": 4.076750448833035e-05, + "loss": 0.5813, + "step": 14449 + }, + { + "epoch": 0.56, + "learning_rate": 4.076686329828161e-05, + "loss": 0.6369, + "step": 14450 + }, + { + "epoch": 0.56, + "learning_rate": 4.0766222108232884e-05, + "loss": 0.5631, + "step": 14451 + }, + { + "epoch": 0.56, + "learning_rate": 4.0765580918184155e-05, + "loss": 0.4261, + "step": 14452 + }, + { + "epoch": 0.56, + "learning_rate": 4.076493972813542e-05, + "loss": 0.4991, + "step": 14453 + }, + { + "epoch": 0.56, + "learning_rate": 4.076429853808669e-05, + "loss": 0.6553, + "step": 14454 + }, + { + "epoch": 0.56, + "learning_rate": 4.076365734803796e-05, + "loss": 0.5791, + "step": 14455 + }, + { + "epoch": 0.56, + "learning_rate": 4.076301615798923e-05, + "loss": 0.5405, + "step": 14456 + }, + { + "epoch": 0.56, + "learning_rate": 4.07623749679405e-05, + "loss": 0.5013, + "step": 14457 + }, + { + "epoch": 0.56, + "learning_rate": 4.076173377789177e-05, + "loss": 0.5622, + "step": 14458 + }, + { + "epoch": 0.56, + "learning_rate": 4.076109258784304e-05, + "loss": 0.6419, + "step": 14459 + }, + { + "epoch": 0.56, + "learning_rate": 4.076045139779431e-05, + "loss": 0.4766, + "step": 14460 + }, + { + "epoch": 0.56, + "learning_rate": 4.075981020774558e-05, + "loss": 0.5476, + "step": 14461 + }, + { + "epoch": 0.56, + "learning_rate": 4.0759169017696845e-05, + "loss": 0.4714, + "step": 14462 + }, + { + "epoch": 0.56, + "learning_rate": 4.075852782764812e-05, + "loss": 0.6295, + "step": 14463 + }, + { + "epoch": 0.56, + "learning_rate": 4.075788663759938e-05, + "loss": 0.4742, + "step": 14464 + }, + { + "epoch": 0.56, + "learning_rate": 4.0757245447550654e-05, + "loss": 0.5698, + "step": 14465 + }, + { + "epoch": 0.56, + "learning_rate": 4.0756604257501926e-05, + "loss": 0.5806, + "step": 14466 + }, + { + "epoch": 0.56, + "learning_rate": 4.07559630674532e-05, + "loss": 0.5358, + "step": 14467 + }, + { + "epoch": 0.56, + "learning_rate": 4.075532187740446e-05, + "loss": 0.5254, + "step": 14468 + }, + { + "epoch": 0.56, + "learning_rate": 4.0754680687355734e-05, + "loss": 0.5118, + "step": 14469 + }, + { + "epoch": 0.56, + "learning_rate": 4.0754039497307006e-05, + "loss": 0.6049, + "step": 14470 + }, + { + "epoch": 0.56, + "learning_rate": 4.075339830725827e-05, + "loss": 0.4956, + "step": 14471 + }, + { + "epoch": 0.56, + "learning_rate": 4.075275711720954e-05, + "loss": 0.6545, + "step": 14472 + }, + { + "epoch": 0.56, + "learning_rate": 4.0752115927160814e-05, + "loss": 0.5347, + "step": 14473 + }, + { + "epoch": 0.56, + "learning_rate": 4.0751474737112086e-05, + "loss": 0.4639, + "step": 14474 + }, + { + "epoch": 0.56, + "learning_rate": 4.075083354706335e-05, + "loss": 0.6256, + "step": 14475 + }, + { + "epoch": 0.56, + "learning_rate": 4.075019235701462e-05, + "loss": 0.438, + "step": 14476 + }, + { + "epoch": 0.56, + "learning_rate": 4.074955116696589e-05, + "loss": 0.5374, + "step": 14477 + }, + { + "epoch": 0.56, + "learning_rate": 4.074890997691716e-05, + "loss": 0.4977, + "step": 14478 + }, + { + "epoch": 0.56, + "learning_rate": 4.074826878686843e-05, + "loss": 0.347, + "step": 14479 + }, + { + "epoch": 0.56, + "learning_rate": 4.0747627596819696e-05, + "loss": 0.5677, + "step": 14480 + }, + { + "epoch": 0.56, + "learning_rate": 4.074698640677097e-05, + "loss": 0.5461, + "step": 14481 + }, + { + "epoch": 0.56, + "learning_rate": 4.074634521672224e-05, + "loss": 0.4717, + "step": 14482 + }, + { + "epoch": 0.56, + "learning_rate": 4.074570402667351e-05, + "loss": 0.6237, + "step": 14483 + }, + { + "epoch": 0.56, + "learning_rate": 4.0745062836624776e-05, + "loss": 0.5611, + "step": 14484 + }, + { + "epoch": 0.56, + "learning_rate": 4.074442164657605e-05, + "loss": 0.4797, + "step": 14485 + }, + { + "epoch": 0.56, + "learning_rate": 4.074378045652731e-05, + "loss": 0.4858, + "step": 14486 + }, + { + "epoch": 0.56, + "learning_rate": 4.0743139266478584e-05, + "loss": 0.5038, + "step": 14487 + }, + { + "epoch": 0.56, + "learning_rate": 4.0742498076429856e-05, + "loss": 0.5512, + "step": 14488 + }, + { + "epoch": 0.56, + "learning_rate": 4.074185688638113e-05, + "loss": 0.3927, + "step": 14489 + }, + { + "epoch": 0.56, + "learning_rate": 4.07412156963324e-05, + "loss": 0.701, + "step": 14490 + }, + { + "epoch": 0.56, + "learning_rate": 4.0740574506283664e-05, + "loss": 0.4788, + "step": 14491 + }, + { + "epoch": 0.56, + "learning_rate": 4.0739933316234936e-05, + "loss": 0.5605, + "step": 14492 + }, + { + "epoch": 0.56, + "learning_rate": 4.07392921261862e-05, + "loss": 0.5666, + "step": 14493 + }, + { + "epoch": 0.56, + "learning_rate": 4.073865093613747e-05, + "loss": 0.5506, + "step": 14494 + }, + { + "epoch": 0.56, + "learning_rate": 4.073800974608874e-05, + "loss": 0.5033, + "step": 14495 + }, + { + "epoch": 0.56, + "learning_rate": 4.073736855604001e-05, + "loss": 0.5447, + "step": 14496 + }, + { + "epoch": 0.56, + "learning_rate": 4.073672736599128e-05, + "loss": 0.7116, + "step": 14497 + }, + { + "epoch": 0.56, + "learning_rate": 4.073608617594255e-05, + "loss": 0.5943, + "step": 14498 + }, + { + "epoch": 0.56, + "learning_rate": 4.0735444985893824e-05, + "loss": 0.4911, + "step": 14499 + }, + { + "epoch": 0.56, + "learning_rate": 4.073480379584509e-05, + "loss": 0.5651, + "step": 14500 + }, + { + "epoch": 0.56, + "learning_rate": 4.073416260579636e-05, + "loss": 0.5934, + "step": 14501 + }, + { + "epoch": 0.56, + "learning_rate": 4.0733521415747626e-05, + "loss": 0.7392, + "step": 14502 + }, + { + "epoch": 0.56, + "learning_rate": 4.07328802256989e-05, + "loss": 0.5336, + "step": 14503 + }, + { + "epoch": 0.56, + "learning_rate": 4.073223903565016e-05, + "loss": 0.5289, + "step": 14504 + }, + { + "epoch": 0.56, + "learning_rate": 4.073159784560144e-05, + "loss": 0.572, + "step": 14505 + }, + { + "epoch": 0.56, + "learning_rate": 4.073095665555271e-05, + "loss": 0.4989, + "step": 14506 + }, + { + "epoch": 0.56, + "learning_rate": 4.073031546550398e-05, + "loss": 0.5647, + "step": 14507 + }, + { + "epoch": 0.56, + "learning_rate": 4.072967427545525e-05, + "loss": 0.5379, + "step": 14508 + }, + { + "epoch": 0.56, + "learning_rate": 4.0729033085406514e-05, + "loss": 0.6117, + "step": 14509 + }, + { + "epoch": 0.56, + "learning_rate": 4.0728391895357786e-05, + "loss": 0.5658, + "step": 14510 + }, + { + "epoch": 0.56, + "learning_rate": 4.072775070530905e-05, + "loss": 0.4704, + "step": 14511 + }, + { + "epoch": 0.56, + "learning_rate": 4.072710951526032e-05, + "loss": 0.5218, + "step": 14512 + }, + { + "epoch": 0.56, + "learning_rate": 4.0726468325211595e-05, + "loss": 0.4864, + "step": 14513 + }, + { + "epoch": 0.56, + "learning_rate": 4.0725827135162866e-05, + "loss": 0.4698, + "step": 14514 + }, + { + "epoch": 0.56, + "learning_rate": 4.072518594511414e-05, + "loss": 0.4843, + "step": 14515 + }, + { + "epoch": 0.56, + "learning_rate": 4.07245447550654e-05, + "loss": 0.5615, + "step": 14516 + }, + { + "epoch": 0.56, + "learning_rate": 4.0723903565016675e-05, + "loss": 0.4837, + "step": 14517 + }, + { + "epoch": 0.56, + "learning_rate": 4.072326237496794e-05, + "loss": 0.5851, + "step": 14518 + }, + { + "epoch": 0.56, + "learning_rate": 4.072262118491921e-05, + "loss": 0.5631, + "step": 14519 + }, + { + "epoch": 0.56, + "learning_rate": 4.0721979994870476e-05, + "loss": 0.5542, + "step": 14520 + }, + { + "epoch": 0.56, + "learning_rate": 4.0721338804821755e-05, + "loss": 0.4085, + "step": 14521 + }, + { + "epoch": 0.56, + "learning_rate": 4.072069761477302e-05, + "loss": 0.5203, + "step": 14522 + }, + { + "epoch": 0.56, + "learning_rate": 4.072005642472429e-05, + "loss": 0.5437, + "step": 14523 + }, + { + "epoch": 0.56, + "learning_rate": 4.071941523467556e-05, + "loss": 0.4763, + "step": 14524 + }, + { + "epoch": 0.56, + "learning_rate": 4.071877404462683e-05, + "loss": 0.494, + "step": 14525 + }, + { + "epoch": 0.56, + "learning_rate": 4.07181328545781e-05, + "loss": 0.5031, + "step": 14526 + }, + { + "epoch": 0.56, + "learning_rate": 4.0717491664529365e-05, + "loss": 0.5479, + "step": 14527 + }, + { + "epoch": 0.56, + "learning_rate": 4.0716850474480636e-05, + "loss": 0.4336, + "step": 14528 + }, + { + "epoch": 0.56, + "learning_rate": 4.071620928443191e-05, + "loss": 0.4905, + "step": 14529 + }, + { + "epoch": 0.56, + "learning_rate": 4.071556809438318e-05, + "loss": 0.4915, + "step": 14530 + }, + { + "epoch": 0.56, + "learning_rate": 4.071492690433445e-05, + "loss": 0.5426, + "step": 14531 + }, + { + "epoch": 0.56, + "learning_rate": 4.0714285714285717e-05, + "loss": 0.6177, + "step": 14532 + }, + { + "epoch": 0.56, + "learning_rate": 4.071364452423699e-05, + "loss": 0.5909, + "step": 14533 + }, + { + "epoch": 0.56, + "learning_rate": 4.071300333418825e-05, + "loss": 0.5013, + "step": 14534 + }, + { + "epoch": 0.56, + "learning_rate": 4.0712362144139525e-05, + "loss": 0.5791, + "step": 14535 + }, + { + "epoch": 0.56, + "learning_rate": 4.071172095409079e-05, + "loss": 0.5072, + "step": 14536 + }, + { + "epoch": 0.56, + "learning_rate": 4.071107976404206e-05, + "loss": 0.4902, + "step": 14537 + }, + { + "epoch": 0.56, + "learning_rate": 4.071043857399333e-05, + "loss": 0.5476, + "step": 14538 + }, + { + "epoch": 0.56, + "learning_rate": 4.0709797383944605e-05, + "loss": 0.4983, + "step": 14539 + }, + { + "epoch": 0.56, + "learning_rate": 4.070915619389588e-05, + "loss": 0.4155, + "step": 14540 + }, + { + "epoch": 0.56, + "learning_rate": 4.070851500384714e-05, + "loss": 0.5144, + "step": 14541 + }, + { + "epoch": 0.56, + "learning_rate": 4.070787381379841e-05, + "loss": 0.5895, + "step": 14542 + }, + { + "epoch": 0.56, + "learning_rate": 4.070723262374968e-05, + "loss": 0.5402, + "step": 14543 + }, + { + "epoch": 0.56, + "learning_rate": 4.070659143370095e-05, + "loss": 0.4175, + "step": 14544 + }, + { + "epoch": 0.56, + "learning_rate": 4.070595024365222e-05, + "loss": 0.6113, + "step": 14545 + }, + { + "epoch": 0.56, + "learning_rate": 4.0705309053603493e-05, + "loss": 0.4417, + "step": 14546 + }, + { + "epoch": 0.56, + "learning_rate": 4.070466786355476e-05, + "loss": 0.6375, + "step": 14547 + }, + { + "epoch": 0.56, + "learning_rate": 4.070402667350603e-05, + "loss": 0.5133, + "step": 14548 + }, + { + "epoch": 0.56, + "learning_rate": 4.07033854834573e-05, + "loss": 0.5677, + "step": 14549 + }, + { + "epoch": 0.56, + "learning_rate": 4.070274429340857e-05, + "loss": 0.5355, + "step": 14550 + }, + { + "epoch": 0.56, + "learning_rate": 4.070210310335984e-05, + "loss": 0.6027, + "step": 14551 + }, + { + "epoch": 0.56, + "learning_rate": 4.0701461913311103e-05, + "loss": 0.6435, + "step": 14552 + }, + { + "epoch": 0.56, + "learning_rate": 4.0700820723262375e-05, + "loss": 0.7756, + "step": 14553 + }, + { + "epoch": 0.56, + "learning_rate": 4.070017953321365e-05, + "loss": 0.5034, + "step": 14554 + }, + { + "epoch": 0.56, + "learning_rate": 4.069953834316492e-05, + "loss": 0.5319, + "step": 14555 + }, + { + "epoch": 0.56, + "learning_rate": 4.0698897153116184e-05, + "loss": 0.5093, + "step": 14556 + }, + { + "epoch": 0.56, + "learning_rate": 4.0698255963067455e-05, + "loss": 0.5255, + "step": 14557 + }, + { + "epoch": 0.56, + "learning_rate": 4.069761477301873e-05, + "loss": 0.4688, + "step": 14558 + }, + { + "epoch": 0.56, + "learning_rate": 4.069697358296999e-05, + "loss": 0.5527, + "step": 14559 + }, + { + "epoch": 0.56, + "learning_rate": 4.0696332392921264e-05, + "loss": 0.4921, + "step": 14560 + }, + { + "epoch": 0.56, + "learning_rate": 4.0695691202872535e-05, + "loss": 0.5144, + "step": 14561 + }, + { + "epoch": 0.56, + "learning_rate": 4.069505001282381e-05, + "loss": 0.5032, + "step": 14562 + }, + { + "epoch": 0.56, + "learning_rate": 4.069440882277507e-05, + "loss": 0.6189, + "step": 14563 + }, + { + "epoch": 0.56, + "learning_rate": 4.0693767632726344e-05, + "loss": 0.4907, + "step": 14564 + }, + { + "epoch": 0.56, + "learning_rate": 4.069312644267761e-05, + "loss": 0.5472, + "step": 14565 + }, + { + "epoch": 0.56, + "learning_rate": 4.069248525262888e-05, + "loss": 0.5099, + "step": 14566 + }, + { + "epoch": 0.56, + "learning_rate": 4.069184406258015e-05, + "loss": 0.5141, + "step": 14567 + }, + { + "epoch": 0.56, + "learning_rate": 4.069120287253142e-05, + "loss": 0.466, + "step": 14568 + }, + { + "epoch": 0.56, + "learning_rate": 4.069056168248269e-05, + "loss": 0.5046, + "step": 14569 + }, + { + "epoch": 0.56, + "learning_rate": 4.068992049243396e-05, + "loss": 0.5409, + "step": 14570 + }, + { + "epoch": 0.56, + "learning_rate": 4.068927930238523e-05, + "loss": 0.457, + "step": 14571 + }, + { + "epoch": 0.56, + "learning_rate": 4.06886381123365e-05, + "loss": 0.5365, + "step": 14572 + }, + { + "epoch": 0.56, + "learning_rate": 4.068799692228777e-05, + "loss": 0.5646, + "step": 14573 + }, + { + "epoch": 0.56, + "learning_rate": 4.0687355732239034e-05, + "loss": 0.4474, + "step": 14574 + }, + { + "epoch": 0.56, + "learning_rate": 4.0686714542190305e-05, + "loss": 0.4835, + "step": 14575 + }, + { + "epoch": 0.56, + "learning_rate": 4.068607335214158e-05, + "loss": 0.5193, + "step": 14576 + }, + { + "epoch": 0.56, + "learning_rate": 4.068543216209285e-05, + "loss": 0.6077, + "step": 14577 + }, + { + "epoch": 0.56, + "learning_rate": 4.068479097204412e-05, + "loss": 0.5116, + "step": 14578 + }, + { + "epoch": 0.56, + "learning_rate": 4.0684149781995386e-05, + "loss": 0.555, + "step": 14579 + }, + { + "epoch": 0.56, + "learning_rate": 4.068350859194666e-05, + "loss": 0.4665, + "step": 14580 + }, + { + "epoch": 0.56, + "learning_rate": 4.068286740189792e-05, + "loss": 0.4356, + "step": 14581 + }, + { + "epoch": 0.56, + "learning_rate": 4.0682226211849194e-05, + "loss": 0.69, + "step": 14582 + }, + { + "epoch": 0.56, + "learning_rate": 4.068158502180046e-05, + "loss": 0.5718, + "step": 14583 + }, + { + "epoch": 0.56, + "learning_rate": 4.068094383175173e-05, + "loss": 0.5575, + "step": 14584 + }, + { + "epoch": 0.56, + "learning_rate": 4.0680302641703e-05, + "loss": 0.5773, + "step": 14585 + }, + { + "epoch": 0.56, + "learning_rate": 4.0679661451654274e-05, + "loss": 0.4445, + "step": 14586 + }, + { + "epoch": 0.56, + "learning_rate": 4.0679020261605546e-05, + "loss": 0.5177, + "step": 14587 + }, + { + "epoch": 0.56, + "learning_rate": 4.067837907155681e-05, + "loss": 0.5304, + "step": 14588 + }, + { + "epoch": 0.56, + "learning_rate": 4.067773788150808e-05, + "loss": 0.408, + "step": 14589 + }, + { + "epoch": 0.56, + "learning_rate": 4.067709669145935e-05, + "loss": 0.6015, + "step": 14590 + }, + { + "epoch": 0.56, + "learning_rate": 4.067645550141062e-05, + "loss": 0.4551, + "step": 14591 + }, + { + "epoch": 0.56, + "learning_rate": 4.0675814311361884e-05, + "loss": 0.5016, + "step": 14592 + }, + { + "epoch": 0.56, + "learning_rate": 4.067517312131316e-05, + "loss": 0.5263, + "step": 14593 + }, + { + "epoch": 0.56, + "learning_rate": 4.0674531931264434e-05, + "loss": 0.4819, + "step": 14594 + }, + { + "epoch": 0.56, + "learning_rate": 4.06738907412157e-05, + "loss": 0.606, + "step": 14595 + }, + { + "epoch": 0.56, + "learning_rate": 4.067324955116697e-05, + "loss": 0.6641, + "step": 14596 + }, + { + "epoch": 0.56, + "learning_rate": 4.0672608361118236e-05, + "loss": 0.4715, + "step": 14597 + }, + { + "epoch": 0.56, + "learning_rate": 4.067196717106951e-05, + "loss": 0.5096, + "step": 14598 + }, + { + "epoch": 0.56, + "learning_rate": 4.067132598102077e-05, + "loss": 0.5534, + "step": 14599 + }, + { + "epoch": 0.56, + "learning_rate": 4.0670684790972044e-05, + "loss": 0.4836, + "step": 14600 + }, + { + "epoch": 0.56, + "learning_rate": 4.0670043600923316e-05, + "loss": 0.5738, + "step": 14601 + }, + { + "epoch": 0.56, + "learning_rate": 4.066940241087459e-05, + "loss": 0.5, + "step": 14602 + }, + { + "epoch": 0.56, + "learning_rate": 4.066876122082586e-05, + "loss": 0.4379, + "step": 14603 + }, + { + "epoch": 0.56, + "learning_rate": 4.0668120030777124e-05, + "loss": 0.4781, + "step": 14604 + }, + { + "epoch": 0.56, + "learning_rate": 4.0667478840728396e-05, + "loss": 0.5116, + "step": 14605 + }, + { + "epoch": 0.56, + "learning_rate": 4.066683765067966e-05, + "loss": 0.5725, + "step": 14606 + }, + { + "epoch": 0.56, + "learning_rate": 4.066619646063093e-05, + "loss": 0.5066, + "step": 14607 + }, + { + "epoch": 0.56, + "learning_rate": 4.06655552705822e-05, + "loss": 0.4825, + "step": 14608 + }, + { + "epoch": 0.56, + "learning_rate": 4.0664914080533476e-05, + "loss": 0.464, + "step": 14609 + }, + { + "epoch": 0.56, + "learning_rate": 4.066427289048474e-05, + "loss": 0.6205, + "step": 14610 + }, + { + "epoch": 0.56, + "learning_rate": 4.066363170043601e-05, + "loss": 0.5732, + "step": 14611 + }, + { + "epoch": 0.56, + "learning_rate": 4.0662990510387284e-05, + "loss": 0.4832, + "step": 14612 + }, + { + "epoch": 0.56, + "learning_rate": 4.066234932033855e-05, + "loss": 0.5584, + "step": 14613 + }, + { + "epoch": 0.56, + "learning_rate": 4.066170813028982e-05, + "loss": 0.5284, + "step": 14614 + }, + { + "epoch": 0.56, + "learning_rate": 4.0661066940241086e-05, + "loss": 0.5898, + "step": 14615 + }, + { + "epoch": 0.56, + "learning_rate": 4.066042575019236e-05, + "loss": 0.4908, + "step": 14616 + }, + { + "epoch": 0.56, + "learning_rate": 4.065978456014363e-05, + "loss": 0.496, + "step": 14617 + }, + { + "epoch": 0.56, + "learning_rate": 4.06591433700949e-05, + "loss": 0.4733, + "step": 14618 + }, + { + "epoch": 0.56, + "learning_rate": 4.0658502180046166e-05, + "loss": 0.5899, + "step": 14619 + }, + { + "epoch": 0.56, + "learning_rate": 4.065786098999744e-05, + "loss": 0.5847, + "step": 14620 + }, + { + "epoch": 0.56, + "learning_rate": 4.065721979994871e-05, + "loss": 0.5247, + "step": 14621 + }, + { + "epoch": 0.56, + "learning_rate": 4.0656578609899975e-05, + "loss": 0.5771, + "step": 14622 + }, + { + "epoch": 0.56, + "learning_rate": 4.0655937419851246e-05, + "loss": 0.5038, + "step": 14623 + }, + { + "epoch": 0.56, + "learning_rate": 4.065529622980251e-05, + "loss": 0.5681, + "step": 14624 + }, + { + "epoch": 0.56, + "learning_rate": 4.065465503975378e-05, + "loss": 0.5986, + "step": 14625 + }, + { + "epoch": 0.56, + "learning_rate": 4.0654013849705055e-05, + "loss": 0.4701, + "step": 14626 + }, + { + "epoch": 0.56, + "learning_rate": 4.0653372659656326e-05, + "loss": 0.3904, + "step": 14627 + }, + { + "epoch": 0.56, + "learning_rate": 4.065273146960759e-05, + "loss": 0.5965, + "step": 14628 + }, + { + "epoch": 0.56, + "learning_rate": 4.065209027955886e-05, + "loss": 0.7199, + "step": 14629 + }, + { + "epoch": 0.56, + "learning_rate": 4.0651449089510135e-05, + "loss": 0.5869, + "step": 14630 + }, + { + "epoch": 0.56, + "learning_rate": 4.06508078994614e-05, + "loss": 0.5275, + "step": 14631 + }, + { + "epoch": 0.56, + "learning_rate": 4.065016670941267e-05, + "loss": 0.462, + "step": 14632 + }, + { + "epoch": 0.56, + "learning_rate": 4.064952551936394e-05, + "loss": 0.4875, + "step": 14633 + }, + { + "epoch": 0.56, + "learning_rate": 4.0648884329315215e-05, + "loss": 0.4158, + "step": 14634 + }, + { + "epoch": 0.56, + "learning_rate": 4.064824313926648e-05, + "loss": 0.6053, + "step": 14635 + }, + { + "epoch": 0.56, + "learning_rate": 4.064760194921775e-05, + "loss": 0.565, + "step": 14636 + }, + { + "epoch": 0.56, + "learning_rate": 4.0646960759169016e-05, + "loss": 0.5015, + "step": 14637 + }, + { + "epoch": 0.56, + "learning_rate": 4.064631956912029e-05, + "loss": 0.57, + "step": 14638 + }, + { + "epoch": 0.56, + "learning_rate": 4.064567837907156e-05, + "loss": 0.5038, + "step": 14639 + }, + { + "epoch": 0.56, + "learning_rate": 4.0645037189022825e-05, + "loss": 0.442, + "step": 14640 + }, + { + "epoch": 0.56, + "learning_rate": 4.0644395998974096e-05, + "loss": 0.5875, + "step": 14641 + }, + { + "epoch": 0.56, + "learning_rate": 4.064375480892537e-05, + "loss": 0.5076, + "step": 14642 + }, + { + "epoch": 0.56, + "learning_rate": 4.064311361887664e-05, + "loss": 0.5352, + "step": 14643 + }, + { + "epoch": 0.56, + "learning_rate": 4.0642472428827905e-05, + "loss": 0.5008, + "step": 14644 + }, + { + "epoch": 0.56, + "learning_rate": 4.0641831238779177e-05, + "loss": 0.4247, + "step": 14645 + }, + { + "epoch": 0.56, + "learning_rate": 4.064119004873044e-05, + "loss": 0.4926, + "step": 14646 + }, + { + "epoch": 0.56, + "learning_rate": 4.064054885868171e-05, + "loss": 0.6171, + "step": 14647 + }, + { + "epoch": 0.56, + "learning_rate": 4.0639907668632985e-05, + "loss": 0.6929, + "step": 14648 + }, + { + "epoch": 0.56, + "learning_rate": 4.063926647858426e-05, + "loss": 0.6001, + "step": 14649 + }, + { + "epoch": 0.56, + "learning_rate": 4.063862528853553e-05, + "loss": 0.4807, + "step": 14650 + }, + { + "epoch": 0.56, + "learning_rate": 4.063798409848679e-05, + "loss": 0.4698, + "step": 14651 + }, + { + "epoch": 0.56, + "learning_rate": 4.0637342908438065e-05, + "loss": 0.5524, + "step": 14652 + }, + { + "epoch": 0.56, + "learning_rate": 4.063670171838933e-05, + "loss": 0.5183, + "step": 14653 + }, + { + "epoch": 0.56, + "learning_rate": 4.06360605283406e-05, + "loss": 0.5955, + "step": 14654 + }, + { + "epoch": 0.56, + "learning_rate": 4.063541933829187e-05, + "loss": 0.446, + "step": 14655 + }, + { + "epoch": 0.56, + "learning_rate": 4.063477814824314e-05, + "loss": 0.5835, + "step": 14656 + }, + { + "epoch": 0.56, + "learning_rate": 4.063413695819441e-05, + "loss": 0.4392, + "step": 14657 + }, + { + "epoch": 0.56, + "learning_rate": 4.063349576814568e-05, + "loss": 0.5626, + "step": 14658 + }, + { + "epoch": 0.56, + "learning_rate": 4.0632854578096953e-05, + "loss": 0.5783, + "step": 14659 + }, + { + "epoch": 0.56, + "learning_rate": 4.063221338804822e-05, + "loss": 0.5189, + "step": 14660 + }, + { + "epoch": 0.56, + "learning_rate": 4.063157219799949e-05, + "loss": 0.5486, + "step": 14661 + }, + { + "epoch": 0.56, + "learning_rate": 4.0630931007950755e-05, + "loss": 0.489, + "step": 14662 + }, + { + "epoch": 0.56, + "learning_rate": 4.063028981790203e-05, + "loss": 0.4993, + "step": 14663 + }, + { + "epoch": 0.56, + "learning_rate": 4.062964862785329e-05, + "loss": 0.428, + "step": 14664 + }, + { + "epoch": 0.56, + "learning_rate": 4.062900743780457e-05, + "loss": 0.4376, + "step": 14665 + }, + { + "epoch": 0.56, + "learning_rate": 4.062836624775584e-05, + "loss": 0.673, + "step": 14666 + }, + { + "epoch": 0.56, + "learning_rate": 4.062772505770711e-05, + "loss": 0.4366, + "step": 14667 + }, + { + "epoch": 0.56, + "learning_rate": 4.062708386765838e-05, + "loss": 0.5594, + "step": 14668 + }, + { + "epoch": 0.56, + "learning_rate": 4.0626442677609644e-05, + "loss": 0.6635, + "step": 14669 + }, + { + "epoch": 0.56, + "learning_rate": 4.0625801487560915e-05, + "loss": 0.635, + "step": 14670 + }, + { + "epoch": 0.56, + "learning_rate": 4.062516029751218e-05, + "loss": 0.3875, + "step": 14671 + }, + { + "epoch": 0.56, + "learning_rate": 4.062451910746345e-05, + "loss": 0.6273, + "step": 14672 + }, + { + "epoch": 0.56, + "learning_rate": 4.0623877917414724e-05, + "loss": 0.5673, + "step": 14673 + }, + { + "epoch": 0.56, + "learning_rate": 4.0623236727365995e-05, + "loss": 0.6271, + "step": 14674 + }, + { + "epoch": 0.56, + "learning_rate": 4.062259553731727e-05, + "loss": 0.4352, + "step": 14675 + }, + { + "epoch": 0.56, + "learning_rate": 4.062195434726853e-05, + "loss": 0.4935, + "step": 14676 + }, + { + "epoch": 0.56, + "learning_rate": 4.0621313157219804e-05, + "loss": 0.5974, + "step": 14677 + }, + { + "epoch": 0.56, + "learning_rate": 4.062067196717107e-05, + "loss": 0.5302, + "step": 14678 + }, + { + "epoch": 0.56, + "learning_rate": 4.062003077712234e-05, + "loss": 0.4958, + "step": 14679 + }, + { + "epoch": 0.56, + "learning_rate": 4.0619389587073605e-05, + "loss": 0.4447, + "step": 14680 + }, + { + "epoch": 0.56, + "learning_rate": 4.0618748397024884e-05, + "loss": 0.4852, + "step": 14681 + }, + { + "epoch": 0.56, + "learning_rate": 4.061810720697615e-05, + "loss": 0.4864, + "step": 14682 + }, + { + "epoch": 0.56, + "learning_rate": 4.061746601692742e-05, + "loss": 0.5178, + "step": 14683 + }, + { + "epoch": 0.56, + "learning_rate": 4.061682482687869e-05, + "loss": 0.4525, + "step": 14684 + }, + { + "epoch": 0.56, + "learning_rate": 4.061618363682996e-05, + "loss": 0.5899, + "step": 14685 + }, + { + "epoch": 0.56, + "learning_rate": 4.061554244678123e-05, + "loss": 0.6074, + "step": 14686 + }, + { + "epoch": 0.56, + "learning_rate": 4.0614901256732494e-05, + "loss": 0.5746, + "step": 14687 + }, + { + "epoch": 0.56, + "learning_rate": 4.0614260066683766e-05, + "loss": 0.6107, + "step": 14688 + }, + { + "epoch": 0.56, + "learning_rate": 4.061361887663504e-05, + "loss": 0.5928, + "step": 14689 + }, + { + "epoch": 0.56, + "learning_rate": 4.061297768658631e-05, + "loss": 0.4859, + "step": 14690 + }, + { + "epoch": 0.56, + "learning_rate": 4.0612336496537574e-05, + "loss": 0.5333, + "step": 14691 + }, + { + "epoch": 0.56, + "learning_rate": 4.0611695306488846e-05, + "loss": 0.4674, + "step": 14692 + }, + { + "epoch": 0.56, + "learning_rate": 4.061105411644012e-05, + "loss": 0.5518, + "step": 14693 + }, + { + "epoch": 0.56, + "learning_rate": 4.061041292639138e-05, + "loss": 0.5052, + "step": 14694 + }, + { + "epoch": 0.56, + "learning_rate": 4.0609771736342654e-05, + "loss": 0.5084, + "step": 14695 + }, + { + "epoch": 0.57, + "learning_rate": 4.060913054629392e-05, + "loss": 0.6132, + "step": 14696 + }, + { + "epoch": 0.57, + "learning_rate": 4.060848935624519e-05, + "loss": 0.511, + "step": 14697 + }, + { + "epoch": 0.57, + "learning_rate": 4.060784816619646e-05, + "loss": 0.4383, + "step": 14698 + }, + { + "epoch": 0.57, + "learning_rate": 4.0607206976147734e-05, + "loss": 0.4774, + "step": 14699 + }, + { + "epoch": 0.57, + "learning_rate": 4.0606565786099e-05, + "loss": 0.5314, + "step": 14700 + }, + { + "epoch": 0.57, + "learning_rate": 4.060592459605027e-05, + "loss": 0.5966, + "step": 14701 + }, + { + "epoch": 0.57, + "learning_rate": 4.060528340600154e-05, + "loss": 0.5434, + "step": 14702 + }, + { + "epoch": 0.57, + "learning_rate": 4.060464221595281e-05, + "loss": 0.692, + "step": 14703 + }, + { + "epoch": 0.57, + "learning_rate": 4.060400102590408e-05, + "loss": 0.5304, + "step": 14704 + }, + { + "epoch": 0.57, + "learning_rate": 4.060335983585535e-05, + "loss": 0.4189, + "step": 14705 + }, + { + "epoch": 0.57, + "learning_rate": 4.060271864580662e-05, + "loss": 0.4908, + "step": 14706 + }, + { + "epoch": 0.57, + "learning_rate": 4.060207745575789e-05, + "loss": 0.4834, + "step": 14707 + }, + { + "epoch": 0.57, + "learning_rate": 4.060143626570916e-05, + "loss": 0.4767, + "step": 14708 + }, + { + "epoch": 0.57, + "learning_rate": 4.0600795075660424e-05, + "loss": 0.5159, + "step": 14709 + }, + { + "epoch": 0.57, + "learning_rate": 4.0600153885611696e-05, + "loss": 0.3904, + "step": 14710 + }, + { + "epoch": 0.57, + "learning_rate": 4.059951269556297e-05, + "loss": 0.4474, + "step": 14711 + }, + { + "epoch": 0.57, + "learning_rate": 4.059887150551423e-05, + "loss": 0.6198, + "step": 14712 + }, + { + "epoch": 0.57, + "learning_rate": 4.0598230315465504e-05, + "loss": 0.4335, + "step": 14713 + }, + { + "epoch": 0.57, + "learning_rate": 4.0597589125416776e-05, + "loss": 0.4905, + "step": 14714 + }, + { + "epoch": 0.57, + "learning_rate": 4.059694793536805e-05, + "loss": 0.5585, + "step": 14715 + }, + { + "epoch": 0.57, + "learning_rate": 4.059630674531931e-05, + "loss": 0.5595, + "step": 14716 + }, + { + "epoch": 0.57, + "learning_rate": 4.0595665555270584e-05, + "loss": 0.388, + "step": 14717 + }, + { + "epoch": 0.57, + "learning_rate": 4.059502436522185e-05, + "loss": 0.6083, + "step": 14718 + }, + { + "epoch": 0.57, + "learning_rate": 4.059438317517312e-05, + "loss": 0.5177, + "step": 14719 + }, + { + "epoch": 0.57, + "learning_rate": 4.059374198512439e-05, + "loss": 0.5566, + "step": 14720 + }, + { + "epoch": 0.57, + "learning_rate": 4.0593100795075664e-05, + "loss": 0.5623, + "step": 14721 + }, + { + "epoch": 0.57, + "learning_rate": 4.0592459605026936e-05, + "loss": 0.5487, + "step": 14722 + }, + { + "epoch": 0.57, + "learning_rate": 4.05918184149782e-05, + "loss": 0.4545, + "step": 14723 + }, + { + "epoch": 0.57, + "learning_rate": 4.059117722492947e-05, + "loss": 0.6777, + "step": 14724 + }, + { + "epoch": 0.57, + "learning_rate": 4.059053603488074e-05, + "loss": 0.489, + "step": 14725 + }, + { + "epoch": 0.57, + "learning_rate": 4.058989484483201e-05, + "loss": 0.4691, + "step": 14726 + }, + { + "epoch": 0.57, + "learning_rate": 4.058925365478328e-05, + "loss": 0.6134, + "step": 14727 + }, + { + "epoch": 0.57, + "learning_rate": 4.0588612464734546e-05, + "loss": 0.4739, + "step": 14728 + }, + { + "epoch": 0.57, + "learning_rate": 4.058797127468582e-05, + "loss": 0.4799, + "step": 14729 + }, + { + "epoch": 0.57, + "learning_rate": 4.058733008463709e-05, + "loss": 0.5663, + "step": 14730 + }, + { + "epoch": 0.57, + "learning_rate": 4.058668889458836e-05, + "loss": 0.4367, + "step": 14731 + }, + { + "epoch": 0.57, + "learning_rate": 4.0586047704539626e-05, + "loss": 0.5346, + "step": 14732 + }, + { + "epoch": 0.57, + "learning_rate": 4.05854065144909e-05, + "loss": 0.5645, + "step": 14733 + }, + { + "epoch": 0.57, + "learning_rate": 4.058476532444216e-05, + "loss": 0.5393, + "step": 14734 + }, + { + "epoch": 0.57, + "learning_rate": 4.0584124134393435e-05, + "loss": 0.5397, + "step": 14735 + }, + { + "epoch": 0.57, + "learning_rate": 4.0583482944344706e-05, + "loss": 0.4105, + "step": 14736 + }, + { + "epoch": 0.57, + "learning_rate": 4.058284175429598e-05, + "loss": 0.5781, + "step": 14737 + }, + { + "epoch": 0.57, + "learning_rate": 4.058220056424725e-05, + "loss": 0.4057, + "step": 14738 + }, + { + "epoch": 0.57, + "learning_rate": 4.0581559374198515e-05, + "loss": 0.4119, + "step": 14739 + }, + { + "epoch": 0.57, + "learning_rate": 4.0580918184149786e-05, + "loss": 0.5762, + "step": 14740 + }, + { + "epoch": 0.57, + "learning_rate": 4.058027699410105e-05, + "loss": 0.4834, + "step": 14741 + }, + { + "epoch": 0.57, + "learning_rate": 4.057963580405232e-05, + "loss": 0.4354, + "step": 14742 + }, + { + "epoch": 0.57, + "learning_rate": 4.057899461400359e-05, + "loss": 0.433, + "step": 14743 + }, + { + "epoch": 0.57, + "learning_rate": 4.057835342395486e-05, + "loss": 0.5745, + "step": 14744 + }, + { + "epoch": 0.57, + "learning_rate": 4.057771223390613e-05, + "loss": 0.5303, + "step": 14745 + }, + { + "epoch": 0.57, + "learning_rate": 4.05770710438574e-05, + "loss": 0.5415, + "step": 14746 + }, + { + "epoch": 0.57, + "learning_rate": 4.0576429853808675e-05, + "loss": 0.5984, + "step": 14747 + }, + { + "epoch": 0.57, + "learning_rate": 4.057578866375994e-05, + "loss": 0.5516, + "step": 14748 + }, + { + "epoch": 0.57, + "learning_rate": 4.057514747371121e-05, + "loss": 0.6037, + "step": 14749 + }, + { + "epoch": 0.57, + "learning_rate": 4.0574506283662476e-05, + "loss": 0.4659, + "step": 14750 + }, + { + "epoch": 0.57, + "learning_rate": 4.057386509361375e-05, + "loss": 0.5408, + "step": 14751 + }, + { + "epoch": 0.57, + "learning_rate": 4.057322390356501e-05, + "loss": 0.5744, + "step": 14752 + }, + { + "epoch": 0.57, + "learning_rate": 4.057258271351629e-05, + "loss": 0.469, + "step": 14753 + }, + { + "epoch": 0.57, + "learning_rate": 4.057194152346756e-05, + "loss": 0.6256, + "step": 14754 + }, + { + "epoch": 0.57, + "learning_rate": 4.057130033341883e-05, + "loss": 0.4978, + "step": 14755 + }, + { + "epoch": 0.57, + "learning_rate": 4.05706591433701e-05, + "loss": 0.5007, + "step": 14756 + }, + { + "epoch": 0.57, + "learning_rate": 4.0570017953321365e-05, + "loss": 0.432, + "step": 14757 + }, + { + "epoch": 0.57, + "learning_rate": 4.0569376763272637e-05, + "loss": 0.5198, + "step": 14758 + }, + { + "epoch": 0.57, + "learning_rate": 4.05687355732239e-05, + "loss": 0.5101, + "step": 14759 + }, + { + "epoch": 0.57, + "learning_rate": 4.056809438317517e-05, + "loss": 0.5989, + "step": 14760 + }, + { + "epoch": 0.57, + "learning_rate": 4.0567453193126445e-05, + "loss": 0.4709, + "step": 14761 + }, + { + "epoch": 0.57, + "learning_rate": 4.056681200307772e-05, + "loss": 0.5365, + "step": 14762 + }, + { + "epoch": 0.57, + "learning_rate": 4.056617081302899e-05, + "loss": 0.5638, + "step": 14763 + }, + { + "epoch": 0.57, + "learning_rate": 4.056552962298025e-05, + "loss": 0.5681, + "step": 14764 + }, + { + "epoch": 0.57, + "learning_rate": 4.0564888432931525e-05, + "loss": 0.5109, + "step": 14765 + }, + { + "epoch": 0.57, + "learning_rate": 4.056424724288279e-05, + "loss": 0.4822, + "step": 14766 + }, + { + "epoch": 0.57, + "learning_rate": 4.056360605283406e-05, + "loss": 0.5907, + "step": 14767 + }, + { + "epoch": 0.57, + "learning_rate": 4.056296486278533e-05, + "loss": 0.5234, + "step": 14768 + }, + { + "epoch": 0.57, + "learning_rate": 4.0562323672736605e-05, + "loss": 0.5383, + "step": 14769 + }, + { + "epoch": 0.57, + "learning_rate": 4.056168248268787e-05, + "loss": 0.491, + "step": 14770 + }, + { + "epoch": 0.57, + "learning_rate": 4.056104129263914e-05, + "loss": 0.5904, + "step": 14771 + }, + { + "epoch": 0.57, + "learning_rate": 4.0560400102590414e-05, + "loss": 0.563, + "step": 14772 + }, + { + "epoch": 0.57, + "learning_rate": 4.055975891254168e-05, + "loss": 0.5404, + "step": 14773 + }, + { + "epoch": 0.57, + "learning_rate": 4.055911772249295e-05, + "loss": 0.5034, + "step": 14774 + }, + { + "epoch": 0.57, + "learning_rate": 4.0558476532444215e-05, + "loss": 0.4859, + "step": 14775 + }, + { + "epoch": 0.57, + "learning_rate": 4.055783534239549e-05, + "loss": 0.6769, + "step": 14776 + }, + { + "epoch": 0.57, + "learning_rate": 4.055719415234676e-05, + "loss": 0.5275, + "step": 14777 + }, + { + "epoch": 0.57, + "learning_rate": 4.055655296229803e-05, + "loss": 0.4635, + "step": 14778 + }, + { + "epoch": 0.57, + "learning_rate": 4.0555911772249295e-05, + "loss": 0.555, + "step": 14779 + }, + { + "epoch": 0.57, + "learning_rate": 4.055527058220057e-05, + "loss": 0.6537, + "step": 14780 + }, + { + "epoch": 0.57, + "learning_rate": 4.055462939215184e-05, + "loss": 0.4616, + "step": 14781 + }, + { + "epoch": 0.57, + "learning_rate": 4.0553988202103104e-05, + "loss": 0.5025, + "step": 14782 + }, + { + "epoch": 0.57, + "learning_rate": 4.0553347012054375e-05, + "loss": 0.5569, + "step": 14783 + }, + { + "epoch": 0.57, + "learning_rate": 4.055270582200564e-05, + "loss": 0.5937, + "step": 14784 + }, + { + "epoch": 0.57, + "learning_rate": 4.055206463195691e-05, + "loss": 0.5195, + "step": 14785 + }, + { + "epoch": 0.57, + "learning_rate": 4.0551423441908184e-05, + "loss": 0.5144, + "step": 14786 + }, + { + "epoch": 0.57, + "learning_rate": 4.0550782251859455e-05, + "loss": 0.6228, + "step": 14787 + }, + { + "epoch": 0.57, + "learning_rate": 4.055014106181072e-05, + "loss": 0.467, + "step": 14788 + }, + { + "epoch": 0.57, + "learning_rate": 4.054949987176199e-05, + "loss": 0.544, + "step": 14789 + }, + { + "epoch": 0.57, + "learning_rate": 4.0548858681713264e-05, + "loss": 0.4795, + "step": 14790 + }, + { + "epoch": 0.57, + "learning_rate": 4.054821749166453e-05, + "loss": 0.5158, + "step": 14791 + }, + { + "epoch": 0.57, + "learning_rate": 4.05475763016158e-05, + "loss": 0.5023, + "step": 14792 + }, + { + "epoch": 0.57, + "learning_rate": 4.054693511156707e-05, + "loss": 0.4217, + "step": 14793 + }, + { + "epoch": 0.57, + "learning_rate": 4.0546293921518344e-05, + "loss": 0.5723, + "step": 14794 + }, + { + "epoch": 0.57, + "learning_rate": 4.054565273146961e-05, + "loss": 0.5331, + "step": 14795 + }, + { + "epoch": 0.57, + "learning_rate": 4.054501154142088e-05, + "loss": 0.4895, + "step": 14796 + }, + { + "epoch": 0.57, + "learning_rate": 4.0544370351372145e-05, + "loss": 0.5283, + "step": 14797 + }, + { + "epoch": 0.57, + "learning_rate": 4.054372916132342e-05, + "loss": 0.5408, + "step": 14798 + }, + { + "epoch": 0.57, + "learning_rate": 4.054308797127469e-05, + "loss": 0.5817, + "step": 14799 + }, + { + "epoch": 0.57, + "learning_rate": 4.0542446781225954e-05, + "loss": 0.5226, + "step": 14800 + }, + { + "epoch": 0.57, + "learning_rate": 4.0541805591177226e-05, + "loss": 0.5793, + "step": 14801 + }, + { + "epoch": 0.57, + "learning_rate": 4.05411644011285e-05, + "loss": 0.6567, + "step": 14802 + }, + { + "epoch": 0.57, + "learning_rate": 4.054052321107977e-05, + "loss": 0.6667, + "step": 14803 + }, + { + "epoch": 0.57, + "learning_rate": 4.0539882021031034e-05, + "loss": 0.5882, + "step": 14804 + }, + { + "epoch": 0.57, + "learning_rate": 4.0539240830982306e-05, + "loss": 0.5225, + "step": 14805 + }, + { + "epoch": 0.57, + "learning_rate": 4.053859964093357e-05, + "loss": 0.5732, + "step": 14806 + }, + { + "epoch": 0.57, + "learning_rate": 4.053795845088484e-05, + "loss": 0.5031, + "step": 14807 + }, + { + "epoch": 0.57, + "learning_rate": 4.0537317260836114e-05, + "loss": 0.5368, + "step": 14808 + }, + { + "epoch": 0.57, + "learning_rate": 4.0536676070787386e-05, + "loss": 0.5288, + "step": 14809 + }, + { + "epoch": 0.57, + "learning_rate": 4.053603488073866e-05, + "loss": 0.4312, + "step": 14810 + }, + { + "epoch": 0.57, + "learning_rate": 4.053539369068992e-05, + "loss": 0.5146, + "step": 14811 + }, + { + "epoch": 0.57, + "learning_rate": 4.0534752500641194e-05, + "loss": 0.5032, + "step": 14812 + }, + { + "epoch": 0.57, + "learning_rate": 4.053411131059246e-05, + "loss": 0.5203, + "step": 14813 + }, + { + "epoch": 0.57, + "learning_rate": 4.053347012054373e-05, + "loss": 0.5901, + "step": 14814 + }, + { + "epoch": 0.57, + "learning_rate": 4.0532828930494996e-05, + "loss": 0.5682, + "step": 14815 + }, + { + "epoch": 0.57, + "learning_rate": 4.053218774044627e-05, + "loss": 0.4755, + "step": 14816 + }, + { + "epoch": 0.57, + "learning_rate": 4.053154655039754e-05, + "loss": 0.5596, + "step": 14817 + }, + { + "epoch": 0.57, + "learning_rate": 4.053090536034881e-05, + "loss": 0.5102, + "step": 14818 + }, + { + "epoch": 0.57, + "learning_rate": 4.053026417030008e-05, + "loss": 0.5188, + "step": 14819 + }, + { + "epoch": 0.57, + "learning_rate": 4.052962298025135e-05, + "loss": 0.6224, + "step": 14820 + }, + { + "epoch": 0.57, + "learning_rate": 4.052898179020262e-05, + "loss": 0.524, + "step": 14821 + }, + { + "epoch": 0.57, + "learning_rate": 4.0528340600153884e-05, + "loss": 0.6225, + "step": 14822 + }, + { + "epoch": 0.57, + "learning_rate": 4.0527699410105156e-05, + "loss": 0.5727, + "step": 14823 + }, + { + "epoch": 0.57, + "learning_rate": 4.052705822005642e-05, + "loss": 0.5273, + "step": 14824 + }, + { + "epoch": 0.57, + "learning_rate": 4.05264170300077e-05, + "loss": 0.6563, + "step": 14825 + }, + { + "epoch": 0.57, + "learning_rate": 4.052577583995897e-05, + "loss": 0.4824, + "step": 14826 + }, + { + "epoch": 0.57, + "learning_rate": 4.0525134649910236e-05, + "loss": 0.5197, + "step": 14827 + }, + { + "epoch": 0.57, + "learning_rate": 4.052449345986151e-05, + "loss": 0.5169, + "step": 14828 + }, + { + "epoch": 0.57, + "learning_rate": 4.052385226981277e-05, + "loss": 0.4969, + "step": 14829 + }, + { + "epoch": 0.57, + "learning_rate": 4.0523211079764044e-05, + "loss": 0.5233, + "step": 14830 + }, + { + "epoch": 0.57, + "learning_rate": 4.052256988971531e-05, + "loss": 0.4509, + "step": 14831 + }, + { + "epoch": 0.57, + "learning_rate": 4.052192869966658e-05, + "loss": 0.5939, + "step": 14832 + }, + { + "epoch": 0.57, + "learning_rate": 4.052128750961785e-05, + "loss": 0.4774, + "step": 14833 + }, + { + "epoch": 0.57, + "learning_rate": 4.0520646319569124e-05, + "loss": 0.6082, + "step": 14834 + }, + { + "epoch": 0.57, + "learning_rate": 4.0520005129520396e-05, + "loss": 0.4735, + "step": 14835 + }, + { + "epoch": 0.57, + "learning_rate": 4.051936393947166e-05, + "loss": 0.4869, + "step": 14836 + }, + { + "epoch": 0.57, + "learning_rate": 4.051872274942293e-05, + "loss": 0.604, + "step": 14837 + }, + { + "epoch": 0.57, + "learning_rate": 4.05180815593742e-05, + "loss": 0.5774, + "step": 14838 + }, + { + "epoch": 0.57, + "learning_rate": 4.051744036932547e-05, + "loss": 0.5874, + "step": 14839 + }, + { + "epoch": 0.57, + "learning_rate": 4.0516799179276734e-05, + "loss": 0.5056, + "step": 14840 + }, + { + "epoch": 0.57, + "learning_rate": 4.051615798922801e-05, + "loss": 0.3865, + "step": 14841 + }, + { + "epoch": 0.57, + "learning_rate": 4.051551679917928e-05, + "loss": 0.5192, + "step": 14842 + }, + { + "epoch": 0.57, + "learning_rate": 4.051487560913055e-05, + "loss": 0.504, + "step": 14843 + }, + { + "epoch": 0.57, + "learning_rate": 4.051423441908182e-05, + "loss": 0.456, + "step": 14844 + }, + { + "epoch": 0.57, + "learning_rate": 4.0513593229033086e-05, + "loss": 0.6041, + "step": 14845 + }, + { + "epoch": 0.57, + "learning_rate": 4.051295203898436e-05, + "loss": 0.4224, + "step": 14846 + }, + { + "epoch": 0.57, + "learning_rate": 4.051231084893562e-05, + "loss": 0.4988, + "step": 14847 + }, + { + "epoch": 0.57, + "learning_rate": 4.0511669658886895e-05, + "loss": 0.5219, + "step": 14848 + }, + { + "epoch": 0.57, + "learning_rate": 4.0511028468838166e-05, + "loss": 0.4235, + "step": 14849 + }, + { + "epoch": 0.57, + "learning_rate": 4.051038727878944e-05, + "loss": 0.4629, + "step": 14850 + }, + { + "epoch": 0.57, + "learning_rate": 4.05097460887407e-05, + "loss": 0.6152, + "step": 14851 + }, + { + "epoch": 0.57, + "learning_rate": 4.0509104898691975e-05, + "loss": 0.4912, + "step": 14852 + }, + { + "epoch": 0.57, + "learning_rate": 4.0508463708643246e-05, + "loss": 0.6376, + "step": 14853 + }, + { + "epoch": 0.57, + "learning_rate": 4.050782251859451e-05, + "loss": 0.4105, + "step": 14854 + }, + { + "epoch": 0.57, + "learning_rate": 4.050718132854578e-05, + "loss": 0.5854, + "step": 14855 + }, + { + "epoch": 0.57, + "learning_rate": 4.050654013849705e-05, + "loss": 0.5578, + "step": 14856 + }, + { + "epoch": 0.57, + "learning_rate": 4.0505898948448326e-05, + "loss": 0.5442, + "step": 14857 + }, + { + "epoch": 0.57, + "learning_rate": 4.050525775839959e-05, + "loss": 0.6673, + "step": 14858 + }, + { + "epoch": 0.57, + "learning_rate": 4.050461656835086e-05, + "loss": 0.5369, + "step": 14859 + }, + { + "epoch": 0.57, + "learning_rate": 4.050397537830213e-05, + "loss": 0.4821, + "step": 14860 + }, + { + "epoch": 0.57, + "learning_rate": 4.05033341882534e-05, + "loss": 0.4559, + "step": 14861 + }, + { + "epoch": 0.57, + "learning_rate": 4.050269299820467e-05, + "loss": 0.6437, + "step": 14862 + }, + { + "epoch": 0.57, + "learning_rate": 4.0502051808155936e-05, + "loss": 0.6153, + "step": 14863 + }, + { + "epoch": 0.57, + "learning_rate": 4.050141061810721e-05, + "loss": 0.5897, + "step": 14864 + }, + { + "epoch": 0.57, + "learning_rate": 4.050076942805848e-05, + "loss": 0.4603, + "step": 14865 + }, + { + "epoch": 0.57, + "learning_rate": 4.050012823800975e-05, + "loss": 0.5825, + "step": 14866 + }, + { + "epoch": 0.57, + "learning_rate": 4.0499487047961017e-05, + "loss": 0.5936, + "step": 14867 + }, + { + "epoch": 0.57, + "learning_rate": 4.049884585791229e-05, + "loss": 0.5242, + "step": 14868 + }, + { + "epoch": 0.57, + "learning_rate": 4.049820466786355e-05, + "loss": 0.582, + "step": 14869 + }, + { + "epoch": 0.57, + "learning_rate": 4.0497563477814825e-05, + "loss": 0.5189, + "step": 14870 + }, + { + "epoch": 0.57, + "learning_rate": 4.04969222877661e-05, + "loss": 0.5916, + "step": 14871 + }, + { + "epoch": 0.57, + "learning_rate": 4.049628109771736e-05, + "loss": 0.5043, + "step": 14872 + }, + { + "epoch": 0.57, + "learning_rate": 4.049563990766863e-05, + "loss": 0.5037, + "step": 14873 + }, + { + "epoch": 0.57, + "learning_rate": 4.0494998717619905e-05, + "loss": 0.6189, + "step": 14874 + }, + { + "epoch": 0.57, + "learning_rate": 4.049435752757118e-05, + "loss": 0.588, + "step": 14875 + }, + { + "epoch": 0.57, + "learning_rate": 4.049371633752244e-05, + "loss": 0.5111, + "step": 14876 + }, + { + "epoch": 0.57, + "learning_rate": 4.049307514747371e-05, + "loss": 0.5661, + "step": 14877 + }, + { + "epoch": 0.57, + "learning_rate": 4.049243395742498e-05, + "loss": 0.511, + "step": 14878 + }, + { + "epoch": 0.57, + "learning_rate": 4.049179276737625e-05, + "loss": 0.5431, + "step": 14879 + }, + { + "epoch": 0.57, + "learning_rate": 4.049115157732752e-05, + "loss": 0.4778, + "step": 14880 + }, + { + "epoch": 0.57, + "learning_rate": 4.0490510387278793e-05, + "loss": 0.5178, + "step": 14881 + }, + { + "epoch": 0.57, + "learning_rate": 4.0489869197230065e-05, + "loss": 0.645, + "step": 14882 + }, + { + "epoch": 0.57, + "learning_rate": 4.048922800718133e-05, + "loss": 0.5356, + "step": 14883 + }, + { + "epoch": 0.57, + "learning_rate": 4.04885868171326e-05, + "loss": 0.5322, + "step": 14884 + }, + { + "epoch": 0.57, + "learning_rate": 4.048794562708387e-05, + "loss": 0.5832, + "step": 14885 + }, + { + "epoch": 0.57, + "learning_rate": 4.048730443703514e-05, + "loss": 0.5819, + "step": 14886 + }, + { + "epoch": 0.57, + "learning_rate": 4.0486663246986403e-05, + "loss": 0.5743, + "step": 14887 + }, + { + "epoch": 0.57, + "learning_rate": 4.0486022056937675e-05, + "loss": 0.5545, + "step": 14888 + }, + { + "epoch": 0.57, + "learning_rate": 4.048538086688895e-05, + "loss": 0.5423, + "step": 14889 + }, + { + "epoch": 0.57, + "learning_rate": 4.048473967684022e-05, + "loss": 0.5418, + "step": 14890 + }, + { + "epoch": 0.57, + "learning_rate": 4.048409848679149e-05, + "loss": 0.4664, + "step": 14891 + }, + { + "epoch": 0.57, + "learning_rate": 4.0483457296742755e-05, + "loss": 0.5358, + "step": 14892 + }, + { + "epoch": 0.57, + "learning_rate": 4.048281610669403e-05, + "loss": 0.4539, + "step": 14893 + }, + { + "epoch": 0.57, + "learning_rate": 4.048217491664529e-05, + "loss": 0.4742, + "step": 14894 + }, + { + "epoch": 0.57, + "learning_rate": 4.0481533726596564e-05, + "loss": 0.4546, + "step": 14895 + }, + { + "epoch": 0.57, + "learning_rate": 4.0480892536547835e-05, + "loss": 0.6352, + "step": 14896 + }, + { + "epoch": 0.57, + "learning_rate": 4.048025134649911e-05, + "loss": 0.444, + "step": 14897 + }, + { + "epoch": 0.57, + "learning_rate": 4.047961015645038e-05, + "loss": 0.4971, + "step": 14898 + }, + { + "epoch": 0.57, + "learning_rate": 4.0478968966401644e-05, + "loss": 0.5284, + "step": 14899 + }, + { + "epoch": 0.57, + "learning_rate": 4.0478327776352915e-05, + "loss": 0.5505, + "step": 14900 + }, + { + "epoch": 0.57, + "learning_rate": 4.047768658630418e-05, + "loss": 0.5126, + "step": 14901 + }, + { + "epoch": 0.57, + "learning_rate": 4.047704539625545e-05, + "loss": 0.4721, + "step": 14902 + }, + { + "epoch": 0.57, + "learning_rate": 4.047640420620672e-05, + "loss": 0.6113, + "step": 14903 + }, + { + "epoch": 0.57, + "learning_rate": 4.047576301615799e-05, + "loss": 0.5906, + "step": 14904 + }, + { + "epoch": 0.57, + "learning_rate": 4.047512182610926e-05, + "loss": 0.5385, + "step": 14905 + }, + { + "epoch": 0.57, + "learning_rate": 4.047448063606053e-05, + "loss": 0.619, + "step": 14906 + }, + { + "epoch": 0.57, + "learning_rate": 4.0473839446011804e-05, + "loss": 0.5557, + "step": 14907 + }, + { + "epoch": 0.57, + "learning_rate": 4.047319825596307e-05, + "loss": 0.4663, + "step": 14908 + }, + { + "epoch": 0.57, + "learning_rate": 4.047255706591434e-05, + "loss": 0.4605, + "step": 14909 + }, + { + "epoch": 0.57, + "learning_rate": 4.0471915875865605e-05, + "loss": 0.4545, + "step": 14910 + }, + { + "epoch": 0.57, + "learning_rate": 4.047127468581688e-05, + "loss": 0.5268, + "step": 14911 + }, + { + "epoch": 0.57, + "learning_rate": 4.047063349576814e-05, + "loss": 0.5279, + "step": 14912 + }, + { + "epoch": 0.57, + "learning_rate": 4.046999230571942e-05, + "loss": 0.4244, + "step": 14913 + }, + { + "epoch": 0.57, + "learning_rate": 4.046935111567069e-05, + "loss": 0.503, + "step": 14914 + }, + { + "epoch": 0.57, + "learning_rate": 4.046870992562196e-05, + "loss": 0.5039, + "step": 14915 + }, + { + "epoch": 0.57, + "learning_rate": 4.046806873557323e-05, + "loss": 0.5302, + "step": 14916 + }, + { + "epoch": 0.57, + "learning_rate": 4.0467427545524494e-05, + "loss": 0.5224, + "step": 14917 + }, + { + "epoch": 0.57, + "learning_rate": 4.0466786355475766e-05, + "loss": 0.5091, + "step": 14918 + }, + { + "epoch": 0.57, + "learning_rate": 4.046614516542703e-05, + "loss": 0.5228, + "step": 14919 + }, + { + "epoch": 0.57, + "learning_rate": 4.04655039753783e-05, + "loss": 0.4241, + "step": 14920 + }, + { + "epoch": 0.57, + "learning_rate": 4.0464862785329574e-05, + "loss": 0.5867, + "step": 14921 + }, + { + "epoch": 0.57, + "learning_rate": 4.0464221595280846e-05, + "loss": 0.6403, + "step": 14922 + }, + { + "epoch": 0.57, + "learning_rate": 4.046358040523212e-05, + "loss": 0.6416, + "step": 14923 + }, + { + "epoch": 0.57, + "learning_rate": 4.046293921518338e-05, + "loss": 0.4014, + "step": 14924 + }, + { + "epoch": 0.57, + "learning_rate": 4.0462298025134654e-05, + "loss": 0.6851, + "step": 14925 + }, + { + "epoch": 0.57, + "learning_rate": 4.046165683508592e-05, + "loss": 0.4269, + "step": 14926 + }, + { + "epoch": 0.57, + "learning_rate": 4.046101564503719e-05, + "loss": 0.446, + "step": 14927 + }, + { + "epoch": 0.57, + "learning_rate": 4.0460374454988456e-05, + "loss": 0.4816, + "step": 14928 + }, + { + "epoch": 0.57, + "learning_rate": 4.0459733264939734e-05, + "loss": 0.6044, + "step": 14929 + }, + { + "epoch": 0.57, + "learning_rate": 4.0459092074891e-05, + "loss": 0.6373, + "step": 14930 + }, + { + "epoch": 0.57, + "learning_rate": 4.045845088484227e-05, + "loss": 0.5654, + "step": 14931 + }, + { + "epoch": 0.57, + "learning_rate": 4.045780969479354e-05, + "loss": 0.5287, + "step": 14932 + }, + { + "epoch": 0.57, + "learning_rate": 4.045716850474481e-05, + "loss": 0.5878, + "step": 14933 + }, + { + "epoch": 0.57, + "learning_rate": 4.045652731469608e-05, + "loss": 0.483, + "step": 14934 + }, + { + "epoch": 0.57, + "learning_rate": 4.0455886124647344e-05, + "loss": 0.5526, + "step": 14935 + }, + { + "epoch": 0.57, + "learning_rate": 4.0455244934598616e-05, + "loss": 0.462, + "step": 14936 + }, + { + "epoch": 0.57, + "learning_rate": 4.045460374454989e-05, + "loss": 0.5045, + "step": 14937 + }, + { + "epoch": 0.57, + "learning_rate": 4.045396255450116e-05, + "loss": 0.4549, + "step": 14938 + }, + { + "epoch": 0.57, + "learning_rate": 4.0453321364452424e-05, + "loss": 0.5373, + "step": 14939 + }, + { + "epoch": 0.57, + "learning_rate": 4.0452680174403696e-05, + "loss": 0.5385, + "step": 14940 + }, + { + "epoch": 0.57, + "learning_rate": 4.045203898435497e-05, + "loss": 0.5145, + "step": 14941 + }, + { + "epoch": 0.57, + "learning_rate": 4.045139779430623e-05, + "loss": 0.5556, + "step": 14942 + }, + { + "epoch": 0.57, + "learning_rate": 4.0450756604257504e-05, + "loss": 0.4603, + "step": 14943 + }, + { + "epoch": 0.57, + "learning_rate": 4.045011541420877e-05, + "loss": 0.6672, + "step": 14944 + }, + { + "epoch": 0.57, + "learning_rate": 4.044947422416004e-05, + "loss": 0.5125, + "step": 14945 + }, + { + "epoch": 0.57, + "learning_rate": 4.044883303411131e-05, + "loss": 0.6032, + "step": 14946 + }, + { + "epoch": 0.57, + "learning_rate": 4.0448191844062584e-05, + "loss": 0.6427, + "step": 14947 + }, + { + "epoch": 0.57, + "learning_rate": 4.044755065401385e-05, + "loss": 0.5063, + "step": 14948 + }, + { + "epoch": 0.57, + "learning_rate": 4.044690946396512e-05, + "loss": 0.4801, + "step": 14949 + }, + { + "epoch": 0.57, + "learning_rate": 4.044626827391639e-05, + "loss": 0.5243, + "step": 14950 + }, + { + "epoch": 0.57, + "learning_rate": 4.044562708386766e-05, + "loss": 0.5011, + "step": 14951 + }, + { + "epoch": 0.57, + "learning_rate": 4.044498589381893e-05, + "loss": 0.4673, + "step": 14952 + }, + { + "epoch": 0.57, + "learning_rate": 4.04443447037702e-05, + "loss": 0.6014, + "step": 14953 + }, + { + "epoch": 0.57, + "learning_rate": 4.044370351372147e-05, + "loss": 0.5287, + "step": 14954 + }, + { + "epoch": 0.57, + "learning_rate": 4.044306232367274e-05, + "loss": 0.5732, + "step": 14955 + }, + { + "epoch": 0.58, + "learning_rate": 4.044242113362401e-05, + "loss": 0.5166, + "step": 14956 + }, + { + "epoch": 0.58, + "learning_rate": 4.0441779943575275e-05, + "loss": 0.6071, + "step": 14957 + }, + { + "epoch": 0.58, + "learning_rate": 4.0441138753526546e-05, + "loss": 0.5964, + "step": 14958 + }, + { + "epoch": 0.58, + "learning_rate": 4.044049756347782e-05, + "loss": 0.5099, + "step": 14959 + }, + { + "epoch": 0.58, + "learning_rate": 4.043985637342908e-05, + "loss": 0.5304, + "step": 14960 + }, + { + "epoch": 0.58, + "learning_rate": 4.0439215183380355e-05, + "loss": 0.4615, + "step": 14961 + }, + { + "epoch": 0.58, + "learning_rate": 4.0438573993331626e-05, + "loss": 0.5725, + "step": 14962 + }, + { + "epoch": 0.58, + "learning_rate": 4.04379328032829e-05, + "loss": 0.6344, + "step": 14963 + }, + { + "epoch": 0.58, + "learning_rate": 4.043729161323416e-05, + "loss": 0.6011, + "step": 14964 + }, + { + "epoch": 0.58, + "learning_rate": 4.0436650423185435e-05, + "loss": 0.4335, + "step": 14965 + }, + { + "epoch": 0.58, + "learning_rate": 4.04360092331367e-05, + "loss": 0.5228, + "step": 14966 + }, + { + "epoch": 0.58, + "learning_rate": 4.043536804308797e-05, + "loss": 0.5483, + "step": 14967 + }, + { + "epoch": 0.58, + "learning_rate": 4.043472685303924e-05, + "loss": 0.4389, + "step": 14968 + }, + { + "epoch": 0.58, + "learning_rate": 4.0434085662990515e-05, + "loss": 0.5645, + "step": 14969 + }, + { + "epoch": 0.58, + "learning_rate": 4.0433444472941787e-05, + "loss": 0.4414, + "step": 14970 + }, + { + "epoch": 0.58, + "learning_rate": 4.043280328289305e-05, + "loss": 0.5303, + "step": 14971 + }, + { + "epoch": 0.58, + "learning_rate": 4.043216209284432e-05, + "loss": 0.4918, + "step": 14972 + }, + { + "epoch": 0.58, + "learning_rate": 4.043152090279559e-05, + "loss": 0.6301, + "step": 14973 + }, + { + "epoch": 0.58, + "learning_rate": 4.043087971274686e-05, + "loss": 0.5172, + "step": 14974 + }, + { + "epoch": 0.58, + "learning_rate": 4.0430238522698125e-05, + "loss": 0.5737, + "step": 14975 + }, + { + "epoch": 0.58, + "learning_rate": 4.0429597332649396e-05, + "loss": 0.4533, + "step": 14976 + }, + { + "epoch": 0.58, + "learning_rate": 4.042895614260067e-05, + "loss": 0.5379, + "step": 14977 + }, + { + "epoch": 0.58, + "learning_rate": 4.042831495255194e-05, + "loss": 0.6217, + "step": 14978 + }, + { + "epoch": 0.58, + "learning_rate": 4.042767376250321e-05, + "loss": 0.5998, + "step": 14979 + }, + { + "epoch": 0.58, + "learning_rate": 4.0427032572454477e-05, + "loss": 0.4695, + "step": 14980 + }, + { + "epoch": 0.58, + "learning_rate": 4.042639138240575e-05, + "loss": 0.5757, + "step": 14981 + }, + { + "epoch": 0.58, + "learning_rate": 4.042575019235701e-05, + "loss": 0.544, + "step": 14982 + }, + { + "epoch": 0.58, + "learning_rate": 4.0425109002308285e-05, + "loss": 0.5496, + "step": 14983 + }, + { + "epoch": 0.58, + "learning_rate": 4.042446781225955e-05, + "loss": 0.4888, + "step": 14984 + }, + { + "epoch": 0.58, + "learning_rate": 4.042382662221083e-05, + "loss": 0.5857, + "step": 14985 + }, + { + "epoch": 0.58, + "learning_rate": 4.04231854321621e-05, + "loss": 0.4849, + "step": 14986 + }, + { + "epoch": 0.58, + "learning_rate": 4.0422544242113365e-05, + "loss": 0.7051, + "step": 14987 + }, + { + "epoch": 0.58, + "learning_rate": 4.042190305206464e-05, + "loss": 0.5775, + "step": 14988 + }, + { + "epoch": 0.58, + "learning_rate": 4.04212618620159e-05, + "loss": 0.6401, + "step": 14989 + }, + { + "epoch": 0.58, + "learning_rate": 4.0420620671967173e-05, + "loss": 0.6221, + "step": 14990 + }, + { + "epoch": 0.58, + "learning_rate": 4.041997948191844e-05, + "loss": 0.5254, + "step": 14991 + }, + { + "epoch": 0.58, + "learning_rate": 4.041933829186971e-05, + "loss": 0.3985, + "step": 14992 + }, + { + "epoch": 0.58, + "learning_rate": 4.041869710182098e-05, + "loss": 0.5938, + "step": 14993 + }, + { + "epoch": 0.58, + "learning_rate": 4.0418055911772253e-05, + "loss": 0.5437, + "step": 14994 + }, + { + "epoch": 0.58, + "learning_rate": 4.0417414721723525e-05, + "loss": 0.5063, + "step": 14995 + }, + { + "epoch": 0.58, + "learning_rate": 4.041677353167479e-05, + "loss": 0.568, + "step": 14996 + }, + { + "epoch": 0.58, + "learning_rate": 4.041613234162606e-05, + "loss": 0.5752, + "step": 14997 + }, + { + "epoch": 0.58, + "learning_rate": 4.041549115157733e-05, + "loss": 0.4758, + "step": 14998 + }, + { + "epoch": 0.58, + "learning_rate": 4.04148499615286e-05, + "loss": 0.5717, + "step": 14999 + }, + { + "epoch": 0.58, + "learning_rate": 4.0414208771479863e-05, + "loss": 0.528, + "step": 15000 + }, + { + "epoch": 0.58, + "learning_rate": 4.041356758143114e-05, + "loss": 0.5857, + "step": 15001 + }, + { + "epoch": 0.58, + "learning_rate": 4.041292639138241e-05, + "loss": 0.5069, + "step": 15002 + }, + { + "epoch": 0.58, + "learning_rate": 4.041228520133368e-05, + "loss": 0.5461, + "step": 15003 + }, + { + "epoch": 0.58, + "learning_rate": 4.041164401128495e-05, + "loss": 0.6366, + "step": 15004 + }, + { + "epoch": 0.58, + "learning_rate": 4.0411002821236215e-05, + "loss": 0.4952, + "step": 15005 + }, + { + "epoch": 0.58, + "learning_rate": 4.041036163118749e-05, + "loss": 0.4906, + "step": 15006 + }, + { + "epoch": 0.58, + "learning_rate": 4.040972044113875e-05, + "loss": 0.5209, + "step": 15007 + }, + { + "epoch": 0.58, + "learning_rate": 4.0409079251090024e-05, + "loss": 0.4873, + "step": 15008 + }, + { + "epoch": 0.58, + "learning_rate": 4.0408438061041295e-05, + "loss": 0.5341, + "step": 15009 + }, + { + "epoch": 0.58, + "learning_rate": 4.040779687099257e-05, + "loss": 0.524, + "step": 15010 + }, + { + "epoch": 0.58, + "learning_rate": 4.040715568094383e-05, + "loss": 0.5268, + "step": 15011 + }, + { + "epoch": 0.58, + "learning_rate": 4.0406514490895104e-05, + "loss": 0.4905, + "step": 15012 + }, + { + "epoch": 0.58, + "learning_rate": 4.0405873300846375e-05, + "loss": 0.471, + "step": 15013 + }, + { + "epoch": 0.58, + "learning_rate": 4.040523211079764e-05, + "loss": 0.5432, + "step": 15014 + }, + { + "epoch": 0.58, + "learning_rate": 4.040459092074891e-05, + "loss": 0.505, + "step": 15015 + }, + { + "epoch": 0.58, + "learning_rate": 4.040394973070018e-05, + "loss": 0.467, + "step": 15016 + }, + { + "epoch": 0.58, + "learning_rate": 4.0403308540651456e-05, + "loss": 0.5419, + "step": 15017 + }, + { + "epoch": 0.58, + "learning_rate": 4.040266735060272e-05, + "loss": 0.5864, + "step": 15018 + }, + { + "epoch": 0.58, + "learning_rate": 4.040202616055399e-05, + "loss": 0.5855, + "step": 15019 + }, + { + "epoch": 0.58, + "learning_rate": 4.040138497050526e-05, + "loss": 0.67, + "step": 15020 + }, + { + "epoch": 0.58, + "learning_rate": 4.040074378045653e-05, + "loss": 0.5526, + "step": 15021 + }, + { + "epoch": 0.58, + "learning_rate": 4.04001025904078e-05, + "loss": 0.4781, + "step": 15022 + }, + { + "epoch": 0.58, + "learning_rate": 4.0399461400359066e-05, + "loss": 0.5174, + "step": 15023 + }, + { + "epoch": 0.58, + "learning_rate": 4.039882021031034e-05, + "loss": 0.5133, + "step": 15024 + }, + { + "epoch": 0.58, + "learning_rate": 4.039817902026161e-05, + "loss": 0.5853, + "step": 15025 + }, + { + "epoch": 0.58, + "learning_rate": 4.039753783021288e-05, + "loss": 0.4469, + "step": 15026 + }, + { + "epoch": 0.58, + "learning_rate": 4.0396896640164146e-05, + "loss": 0.3994, + "step": 15027 + }, + { + "epoch": 0.58, + "learning_rate": 4.039625545011542e-05, + "loss": 0.6058, + "step": 15028 + }, + { + "epoch": 0.58, + "learning_rate": 4.039561426006668e-05, + "loss": 0.5839, + "step": 15029 + }, + { + "epoch": 0.58, + "learning_rate": 4.0394973070017954e-05, + "loss": 0.6385, + "step": 15030 + }, + { + "epoch": 0.58, + "learning_rate": 4.0394331879969226e-05, + "loss": 0.5155, + "step": 15031 + }, + { + "epoch": 0.58, + "learning_rate": 4.039369068992049e-05, + "loss": 0.5881, + "step": 15032 + }, + { + "epoch": 0.58, + "learning_rate": 4.039304949987176e-05, + "loss": 0.423, + "step": 15033 + }, + { + "epoch": 0.58, + "learning_rate": 4.0392408309823034e-05, + "loss": 0.5827, + "step": 15034 + }, + { + "epoch": 0.58, + "learning_rate": 4.0391767119774306e-05, + "loss": 0.4713, + "step": 15035 + }, + { + "epoch": 0.58, + "learning_rate": 4.039112592972557e-05, + "loss": 0.5202, + "step": 15036 + }, + { + "epoch": 0.58, + "learning_rate": 4.039048473967684e-05, + "loss": 0.52, + "step": 15037 + }, + { + "epoch": 0.58, + "learning_rate": 4.038984354962811e-05, + "loss": 0.5844, + "step": 15038 + }, + { + "epoch": 0.58, + "learning_rate": 4.038920235957938e-05, + "loss": 0.5017, + "step": 15039 + }, + { + "epoch": 0.58, + "learning_rate": 4.038856116953065e-05, + "loss": 0.545, + "step": 15040 + }, + { + "epoch": 0.58, + "learning_rate": 4.038791997948192e-05, + "loss": 0.3681, + "step": 15041 + }, + { + "epoch": 0.58, + "learning_rate": 4.0387278789433194e-05, + "loss": 0.5867, + "step": 15042 + }, + { + "epoch": 0.58, + "learning_rate": 4.038663759938446e-05, + "loss": 0.5421, + "step": 15043 + }, + { + "epoch": 0.58, + "learning_rate": 4.038599640933573e-05, + "loss": 0.5455, + "step": 15044 + }, + { + "epoch": 0.58, + "learning_rate": 4.0385355219286996e-05, + "loss": 0.5487, + "step": 15045 + }, + { + "epoch": 0.58, + "learning_rate": 4.038471402923827e-05, + "loss": 0.4978, + "step": 15046 + }, + { + "epoch": 0.58, + "learning_rate": 4.038407283918953e-05, + "loss": 0.5453, + "step": 15047 + }, + { + "epoch": 0.58, + "learning_rate": 4.0383431649140804e-05, + "loss": 0.6163, + "step": 15048 + }, + { + "epoch": 0.58, + "learning_rate": 4.0382790459092076e-05, + "loss": 0.5631, + "step": 15049 + }, + { + "epoch": 0.58, + "learning_rate": 4.038214926904335e-05, + "loss": 0.5171, + "step": 15050 + }, + { + "epoch": 0.58, + "learning_rate": 4.038150807899462e-05, + "loss": 0.5955, + "step": 15051 + }, + { + "epoch": 0.58, + "learning_rate": 4.0380866888945884e-05, + "loss": 0.6181, + "step": 15052 + }, + { + "epoch": 0.58, + "learning_rate": 4.0380225698897156e-05, + "loss": 0.6281, + "step": 15053 + }, + { + "epoch": 0.58, + "learning_rate": 4.037958450884842e-05, + "loss": 0.5669, + "step": 15054 + }, + { + "epoch": 0.58, + "learning_rate": 4.037894331879969e-05, + "loss": 0.524, + "step": 15055 + }, + { + "epoch": 0.58, + "learning_rate": 4.0378302128750964e-05, + "loss": 0.6353, + "step": 15056 + }, + { + "epoch": 0.58, + "learning_rate": 4.0377660938702236e-05, + "loss": 0.4906, + "step": 15057 + }, + { + "epoch": 0.58, + "learning_rate": 4.037701974865351e-05, + "loss": 0.4875, + "step": 15058 + }, + { + "epoch": 0.58, + "learning_rate": 4.037637855860477e-05, + "loss": 0.558, + "step": 15059 + }, + { + "epoch": 0.58, + "learning_rate": 4.0375737368556044e-05, + "loss": 0.473, + "step": 15060 + }, + { + "epoch": 0.58, + "learning_rate": 4.037509617850731e-05, + "loss": 0.547, + "step": 15061 + }, + { + "epoch": 0.58, + "learning_rate": 4.037445498845858e-05, + "loss": 0.4652, + "step": 15062 + }, + { + "epoch": 0.58, + "learning_rate": 4.0373813798409846e-05, + "loss": 0.5505, + "step": 15063 + }, + { + "epoch": 0.58, + "learning_rate": 4.037317260836112e-05, + "loss": 0.4824, + "step": 15064 + }, + { + "epoch": 0.58, + "learning_rate": 4.037253141831239e-05, + "loss": 0.4884, + "step": 15065 + }, + { + "epoch": 0.58, + "learning_rate": 4.037189022826366e-05, + "loss": 0.4724, + "step": 15066 + }, + { + "epoch": 0.58, + "learning_rate": 4.037124903821493e-05, + "loss": 0.5969, + "step": 15067 + }, + { + "epoch": 0.58, + "learning_rate": 4.03706078481662e-05, + "loss": 0.5169, + "step": 15068 + }, + { + "epoch": 0.58, + "learning_rate": 4.036996665811747e-05, + "loss": 0.4678, + "step": 15069 + }, + { + "epoch": 0.58, + "learning_rate": 4.0369325468068735e-05, + "loss": 0.4685, + "step": 15070 + }, + { + "epoch": 0.58, + "learning_rate": 4.0368684278020006e-05, + "loss": 0.4477, + "step": 15071 + }, + { + "epoch": 0.58, + "learning_rate": 4.036804308797127e-05, + "loss": 0.5061, + "step": 15072 + }, + { + "epoch": 0.58, + "learning_rate": 4.036740189792255e-05, + "loss": 0.5545, + "step": 15073 + }, + { + "epoch": 0.58, + "learning_rate": 4.0366760707873815e-05, + "loss": 0.5998, + "step": 15074 + }, + { + "epoch": 0.58, + "learning_rate": 4.0366119517825086e-05, + "loss": 0.5851, + "step": 15075 + }, + { + "epoch": 0.58, + "learning_rate": 4.036547832777636e-05, + "loss": 0.582, + "step": 15076 + }, + { + "epoch": 0.58, + "learning_rate": 4.036483713772762e-05, + "loss": 0.4656, + "step": 15077 + }, + { + "epoch": 0.58, + "learning_rate": 4.0364195947678895e-05, + "loss": 0.4433, + "step": 15078 + }, + { + "epoch": 0.58, + "learning_rate": 4.036355475763016e-05, + "loss": 0.5813, + "step": 15079 + }, + { + "epoch": 0.58, + "learning_rate": 4.036291356758143e-05, + "loss": 0.4537, + "step": 15080 + }, + { + "epoch": 0.58, + "learning_rate": 4.03622723775327e-05, + "loss": 0.605, + "step": 15081 + }, + { + "epoch": 0.58, + "learning_rate": 4.0361631187483975e-05, + "loss": 0.4299, + "step": 15082 + }, + { + "epoch": 0.58, + "learning_rate": 4.036098999743524e-05, + "loss": 0.4743, + "step": 15083 + }, + { + "epoch": 0.58, + "learning_rate": 4.036034880738651e-05, + "loss": 0.543, + "step": 15084 + }, + { + "epoch": 0.58, + "learning_rate": 4.035970761733778e-05, + "loss": 0.5992, + "step": 15085 + }, + { + "epoch": 0.58, + "learning_rate": 4.035906642728905e-05, + "loss": 0.6081, + "step": 15086 + }, + { + "epoch": 0.58, + "learning_rate": 4.035842523724032e-05, + "loss": 0.4255, + "step": 15087 + }, + { + "epoch": 0.58, + "learning_rate": 4.0357784047191585e-05, + "loss": 0.4331, + "step": 15088 + }, + { + "epoch": 0.58, + "learning_rate": 4.035714285714286e-05, + "loss": 0.5346, + "step": 15089 + }, + { + "epoch": 0.58, + "learning_rate": 4.035650166709413e-05, + "loss": 0.5123, + "step": 15090 + }, + { + "epoch": 0.58, + "learning_rate": 4.03558604770454e-05, + "loss": 0.47, + "step": 15091 + }, + { + "epoch": 0.58, + "learning_rate": 4.0355219286996665e-05, + "loss": 0.5116, + "step": 15092 + }, + { + "epoch": 0.58, + "learning_rate": 4.0354578096947937e-05, + "loss": 0.5914, + "step": 15093 + }, + { + "epoch": 0.58, + "learning_rate": 4.035393690689921e-05, + "loss": 0.6227, + "step": 15094 + }, + { + "epoch": 0.58, + "learning_rate": 4.035329571685047e-05, + "loss": 0.5836, + "step": 15095 + }, + { + "epoch": 0.58, + "learning_rate": 4.0352654526801745e-05, + "loss": 0.5425, + "step": 15096 + }, + { + "epoch": 0.58, + "learning_rate": 4.035201333675302e-05, + "loss": 0.7024, + "step": 15097 + }, + { + "epoch": 0.58, + "learning_rate": 4.035137214670429e-05, + "loss": 0.4472, + "step": 15098 + }, + { + "epoch": 0.58, + "learning_rate": 4.035073095665555e-05, + "loss": 0.5328, + "step": 15099 + }, + { + "epoch": 0.58, + "learning_rate": 4.0350089766606825e-05, + "loss": 0.5396, + "step": 15100 + }, + { + "epoch": 0.58, + "learning_rate": 4.03494485765581e-05, + "loss": 0.6033, + "step": 15101 + }, + { + "epoch": 0.58, + "learning_rate": 4.034880738650936e-05, + "loss": 0.5929, + "step": 15102 + }, + { + "epoch": 0.58, + "learning_rate": 4.0348166196460633e-05, + "loss": 0.5337, + "step": 15103 + }, + { + "epoch": 0.58, + "learning_rate": 4.03475250064119e-05, + "loss": 0.5408, + "step": 15104 + }, + { + "epoch": 0.58, + "learning_rate": 4.034688381636318e-05, + "loss": 0.4386, + "step": 15105 + }, + { + "epoch": 0.58, + "learning_rate": 4.034624262631444e-05, + "loss": 0.5685, + "step": 15106 + }, + { + "epoch": 0.58, + "learning_rate": 4.0345601436265714e-05, + "loss": 0.5525, + "step": 15107 + }, + { + "epoch": 0.58, + "learning_rate": 4.034496024621698e-05, + "loss": 0.6618, + "step": 15108 + }, + { + "epoch": 0.58, + "learning_rate": 4.034431905616825e-05, + "loss": 0.56, + "step": 15109 + }, + { + "epoch": 0.58, + "learning_rate": 4.034367786611952e-05, + "loss": 0.5028, + "step": 15110 + }, + { + "epoch": 0.58, + "learning_rate": 4.034303667607079e-05, + "loss": 0.5099, + "step": 15111 + }, + { + "epoch": 0.58, + "learning_rate": 4.034239548602206e-05, + "loss": 0.7088, + "step": 15112 + }, + { + "epoch": 0.58, + "learning_rate": 4.034175429597333e-05, + "loss": 0.4516, + "step": 15113 + }, + { + "epoch": 0.58, + "learning_rate": 4.03411131059246e-05, + "loss": 0.5383, + "step": 15114 + }, + { + "epoch": 0.58, + "learning_rate": 4.034047191587587e-05, + "loss": 0.5312, + "step": 15115 + }, + { + "epoch": 0.58, + "learning_rate": 4.033983072582714e-05, + "loss": 0.4267, + "step": 15116 + }, + { + "epoch": 0.58, + "learning_rate": 4.0339189535778404e-05, + "loss": 0.4855, + "step": 15117 + }, + { + "epoch": 0.58, + "learning_rate": 4.0338548345729675e-05, + "loss": 0.6019, + "step": 15118 + }, + { + "epoch": 0.58, + "learning_rate": 4.033790715568095e-05, + "loss": 0.5859, + "step": 15119 + }, + { + "epoch": 0.58, + "learning_rate": 4.033726596563221e-05, + "loss": 0.553, + "step": 15120 + }, + { + "epoch": 0.58, + "learning_rate": 4.0336624775583484e-05, + "loss": 0.5371, + "step": 15121 + }, + { + "epoch": 0.58, + "learning_rate": 4.0335983585534755e-05, + "loss": 0.5603, + "step": 15122 + }, + { + "epoch": 0.58, + "learning_rate": 4.033534239548603e-05, + "loss": 0.6013, + "step": 15123 + }, + { + "epoch": 0.58, + "learning_rate": 4.033470120543729e-05, + "loss": 0.4336, + "step": 15124 + }, + { + "epoch": 0.58, + "learning_rate": 4.0334060015388564e-05, + "loss": 0.5348, + "step": 15125 + }, + { + "epoch": 0.58, + "learning_rate": 4.033341882533983e-05, + "loss": 0.5983, + "step": 15126 + }, + { + "epoch": 0.58, + "learning_rate": 4.03327776352911e-05, + "loss": 0.5592, + "step": 15127 + }, + { + "epoch": 0.58, + "learning_rate": 4.033213644524237e-05, + "loss": 0.584, + "step": 15128 + }, + { + "epoch": 0.58, + "learning_rate": 4.0331495255193644e-05, + "loss": 0.4974, + "step": 15129 + }, + { + "epoch": 0.58, + "learning_rate": 4.0330854065144916e-05, + "loss": 0.5709, + "step": 15130 + }, + { + "epoch": 0.58, + "learning_rate": 4.033021287509618e-05, + "loss": 0.5683, + "step": 15131 + }, + { + "epoch": 0.58, + "learning_rate": 4.032957168504745e-05, + "loss": 0.4814, + "step": 15132 + }, + { + "epoch": 0.58, + "learning_rate": 4.032893049499872e-05, + "loss": 0.543, + "step": 15133 + }, + { + "epoch": 0.58, + "learning_rate": 4.032828930494999e-05, + "loss": 0.4835, + "step": 15134 + }, + { + "epoch": 0.58, + "learning_rate": 4.0327648114901254e-05, + "loss": 0.6577, + "step": 15135 + }, + { + "epoch": 0.58, + "learning_rate": 4.0327006924852526e-05, + "loss": 0.5646, + "step": 15136 + }, + { + "epoch": 0.58, + "learning_rate": 4.03263657348038e-05, + "loss": 0.5286, + "step": 15137 + }, + { + "epoch": 0.58, + "learning_rate": 4.032572454475507e-05, + "loss": 0.577, + "step": 15138 + }, + { + "epoch": 0.58, + "learning_rate": 4.032508335470634e-05, + "loss": 0.5047, + "step": 15139 + }, + { + "epoch": 0.58, + "learning_rate": 4.0324442164657606e-05, + "loss": 0.495, + "step": 15140 + }, + { + "epoch": 0.58, + "learning_rate": 4.032380097460888e-05, + "loss": 0.5734, + "step": 15141 + }, + { + "epoch": 0.58, + "learning_rate": 4.032315978456014e-05, + "loss": 0.5073, + "step": 15142 + }, + { + "epoch": 0.58, + "learning_rate": 4.0322518594511414e-05, + "loss": 0.571, + "step": 15143 + }, + { + "epoch": 0.58, + "learning_rate": 4.0321877404462686e-05, + "loss": 0.5892, + "step": 15144 + }, + { + "epoch": 0.58, + "learning_rate": 4.032123621441396e-05, + "loss": 0.5293, + "step": 15145 + }, + { + "epoch": 0.58, + "learning_rate": 4.032059502436523e-05, + "loss": 0.5489, + "step": 15146 + }, + { + "epoch": 0.58, + "learning_rate": 4.0319953834316494e-05, + "loss": 0.5406, + "step": 15147 + }, + { + "epoch": 0.58, + "learning_rate": 4.0319312644267766e-05, + "loss": 0.5684, + "step": 15148 + }, + { + "epoch": 0.58, + "learning_rate": 4.031867145421903e-05, + "loss": 0.4756, + "step": 15149 + }, + { + "epoch": 0.58, + "learning_rate": 4.03180302641703e-05, + "loss": 0.6251, + "step": 15150 + }, + { + "epoch": 0.58, + "learning_rate": 4.031738907412157e-05, + "loss": 0.3709, + "step": 15151 + }, + { + "epoch": 0.58, + "learning_rate": 4.031674788407284e-05, + "loss": 0.6123, + "step": 15152 + }, + { + "epoch": 0.58, + "learning_rate": 4.031610669402411e-05, + "loss": 0.4754, + "step": 15153 + }, + { + "epoch": 0.58, + "learning_rate": 4.031546550397538e-05, + "loss": 0.4371, + "step": 15154 + }, + { + "epoch": 0.58, + "learning_rate": 4.0314824313926654e-05, + "loss": 0.4788, + "step": 15155 + }, + { + "epoch": 0.58, + "learning_rate": 4.031418312387792e-05, + "loss": 0.6595, + "step": 15156 + }, + { + "epoch": 0.58, + "learning_rate": 4.031354193382919e-05, + "loss": 0.5468, + "step": 15157 + }, + { + "epoch": 0.58, + "learning_rate": 4.0312900743780456e-05, + "loss": 0.4821, + "step": 15158 + }, + { + "epoch": 0.58, + "learning_rate": 4.031225955373173e-05, + "loss": 0.4963, + "step": 15159 + }, + { + "epoch": 0.58, + "learning_rate": 4.031161836368299e-05, + "loss": 0.522, + "step": 15160 + }, + { + "epoch": 0.58, + "learning_rate": 4.031097717363427e-05, + "loss": 0.5188, + "step": 15161 + }, + { + "epoch": 0.58, + "learning_rate": 4.0310335983585536e-05, + "loss": 0.4625, + "step": 15162 + }, + { + "epoch": 0.58, + "learning_rate": 4.030969479353681e-05, + "loss": 0.5486, + "step": 15163 + }, + { + "epoch": 0.58, + "learning_rate": 4.030905360348808e-05, + "loss": 0.4308, + "step": 15164 + }, + { + "epoch": 0.58, + "learning_rate": 4.0308412413439344e-05, + "loss": 0.5038, + "step": 15165 + }, + { + "epoch": 0.58, + "learning_rate": 4.0307771223390616e-05, + "loss": 0.5922, + "step": 15166 + }, + { + "epoch": 0.58, + "learning_rate": 4.030713003334188e-05, + "loss": 0.5286, + "step": 15167 + }, + { + "epoch": 0.58, + "learning_rate": 4.030648884329315e-05, + "loss": 0.5477, + "step": 15168 + }, + { + "epoch": 0.58, + "learning_rate": 4.0305847653244424e-05, + "loss": 0.5268, + "step": 15169 + }, + { + "epoch": 0.58, + "learning_rate": 4.0305206463195696e-05, + "loss": 0.4986, + "step": 15170 + }, + { + "epoch": 0.58, + "learning_rate": 4.030456527314696e-05, + "loss": 0.6092, + "step": 15171 + }, + { + "epoch": 0.58, + "learning_rate": 4.030392408309823e-05, + "loss": 0.5333, + "step": 15172 + }, + { + "epoch": 0.58, + "learning_rate": 4.0303282893049505e-05, + "loss": 0.4687, + "step": 15173 + }, + { + "epoch": 0.58, + "learning_rate": 4.030264170300077e-05, + "loss": 0.3616, + "step": 15174 + }, + { + "epoch": 0.58, + "learning_rate": 4.030200051295204e-05, + "loss": 0.6105, + "step": 15175 + }, + { + "epoch": 0.58, + "learning_rate": 4.0301359322903306e-05, + "loss": 0.5471, + "step": 15176 + }, + { + "epoch": 0.58, + "learning_rate": 4.0300718132854585e-05, + "loss": 0.6116, + "step": 15177 + }, + { + "epoch": 0.58, + "learning_rate": 4.030007694280585e-05, + "loss": 0.6426, + "step": 15178 + }, + { + "epoch": 0.58, + "learning_rate": 4.029943575275712e-05, + "loss": 0.5856, + "step": 15179 + }, + { + "epoch": 0.58, + "learning_rate": 4.0298794562708386e-05, + "loss": 0.4698, + "step": 15180 + }, + { + "epoch": 0.58, + "learning_rate": 4.029815337265966e-05, + "loss": 0.5468, + "step": 15181 + }, + { + "epoch": 0.58, + "learning_rate": 4.029751218261093e-05, + "loss": 0.5762, + "step": 15182 + }, + { + "epoch": 0.58, + "learning_rate": 4.0296870992562195e-05, + "loss": 0.6226, + "step": 15183 + }, + { + "epoch": 0.58, + "learning_rate": 4.0296229802513466e-05, + "loss": 0.5107, + "step": 15184 + }, + { + "epoch": 0.58, + "learning_rate": 4.029558861246474e-05, + "loss": 0.5828, + "step": 15185 + }, + { + "epoch": 0.58, + "learning_rate": 4.029494742241601e-05, + "loss": 0.6346, + "step": 15186 + }, + { + "epoch": 0.58, + "learning_rate": 4.0294306232367275e-05, + "loss": 0.5282, + "step": 15187 + }, + { + "epoch": 0.58, + "learning_rate": 4.0293665042318546e-05, + "loss": 0.4615, + "step": 15188 + }, + { + "epoch": 0.58, + "learning_rate": 4.029302385226981e-05, + "loss": 0.5035, + "step": 15189 + }, + { + "epoch": 0.58, + "learning_rate": 4.029238266222108e-05, + "loss": 0.5613, + "step": 15190 + }, + { + "epoch": 0.58, + "learning_rate": 4.0291741472172355e-05, + "loss": 0.5416, + "step": 15191 + }, + { + "epoch": 0.58, + "learning_rate": 4.029110028212362e-05, + "loss": 0.5464, + "step": 15192 + }, + { + "epoch": 0.58, + "learning_rate": 4.029045909207489e-05, + "loss": 0.5538, + "step": 15193 + }, + { + "epoch": 0.58, + "learning_rate": 4.028981790202616e-05, + "loss": 0.6291, + "step": 15194 + }, + { + "epoch": 0.58, + "learning_rate": 4.0289176711977435e-05, + "loss": 0.5745, + "step": 15195 + }, + { + "epoch": 0.58, + "learning_rate": 4.02885355219287e-05, + "loss": 0.5809, + "step": 15196 + }, + { + "epoch": 0.58, + "learning_rate": 4.028789433187997e-05, + "loss": 0.5511, + "step": 15197 + }, + { + "epoch": 0.58, + "learning_rate": 4.0287253141831236e-05, + "loss": 0.5834, + "step": 15198 + }, + { + "epoch": 0.58, + "learning_rate": 4.028661195178251e-05, + "loss": 0.5994, + "step": 15199 + }, + { + "epoch": 0.58, + "learning_rate": 4.028597076173378e-05, + "loss": 0.4588, + "step": 15200 + }, + { + "epoch": 0.58, + "learning_rate": 4.028532957168505e-05, + "loss": 0.6336, + "step": 15201 + }, + { + "epoch": 0.58, + "learning_rate": 4.028468838163632e-05, + "loss": 0.7222, + "step": 15202 + }, + { + "epoch": 0.58, + "learning_rate": 4.028404719158759e-05, + "loss": 0.6282, + "step": 15203 + }, + { + "epoch": 0.58, + "learning_rate": 4.028340600153886e-05, + "loss": 0.5407, + "step": 15204 + }, + { + "epoch": 0.58, + "learning_rate": 4.0282764811490125e-05, + "loss": 0.5423, + "step": 15205 + }, + { + "epoch": 0.58, + "learning_rate": 4.02821236214414e-05, + "loss": 0.5523, + "step": 15206 + }, + { + "epoch": 0.58, + "learning_rate": 4.028148243139266e-05, + "loss": 0.5523, + "step": 15207 + }, + { + "epoch": 0.58, + "learning_rate": 4.028084124134393e-05, + "loss": 0.5412, + "step": 15208 + }, + { + "epoch": 0.58, + "learning_rate": 4.0280200051295205e-05, + "loss": 0.5603, + "step": 15209 + }, + { + "epoch": 0.58, + "learning_rate": 4.027955886124648e-05, + "loss": 0.514, + "step": 15210 + }, + { + "epoch": 0.58, + "learning_rate": 4.027891767119775e-05, + "loss": 0.5943, + "step": 15211 + }, + { + "epoch": 0.58, + "learning_rate": 4.027827648114901e-05, + "loss": 0.4901, + "step": 15212 + }, + { + "epoch": 0.58, + "learning_rate": 4.0277635291100285e-05, + "loss": 0.5545, + "step": 15213 + }, + { + "epoch": 0.58, + "learning_rate": 4.027699410105155e-05, + "loss": 0.6017, + "step": 15214 + }, + { + "epoch": 0.58, + "learning_rate": 4.027635291100282e-05, + "loss": 0.4986, + "step": 15215 + }, + { + "epoch": 0.59, + "learning_rate": 4.0275711720954093e-05, + "loss": 0.5431, + "step": 15216 + }, + { + "epoch": 0.59, + "learning_rate": 4.0275070530905365e-05, + "loss": 0.5546, + "step": 15217 + }, + { + "epoch": 0.59, + "learning_rate": 4.027442934085664e-05, + "loss": 0.5826, + "step": 15218 + }, + { + "epoch": 0.59, + "learning_rate": 4.02737881508079e-05, + "loss": 0.5443, + "step": 15219 + }, + { + "epoch": 0.59, + "learning_rate": 4.0273146960759174e-05, + "loss": 0.519, + "step": 15220 + }, + { + "epoch": 0.59, + "learning_rate": 4.027250577071044e-05, + "loss": 0.5199, + "step": 15221 + }, + { + "epoch": 0.59, + "learning_rate": 4.027186458066171e-05, + "loss": 0.5685, + "step": 15222 + }, + { + "epoch": 0.59, + "learning_rate": 4.0271223390612975e-05, + "loss": 0.4276, + "step": 15223 + }, + { + "epoch": 0.59, + "learning_rate": 4.027058220056425e-05, + "loss": 0.4981, + "step": 15224 + }, + { + "epoch": 0.59, + "learning_rate": 4.026994101051552e-05, + "loss": 0.4538, + "step": 15225 + }, + { + "epoch": 0.59, + "learning_rate": 4.026929982046679e-05, + "loss": 0.5054, + "step": 15226 + }, + { + "epoch": 0.59, + "learning_rate": 4.026865863041806e-05, + "loss": 0.4958, + "step": 15227 + }, + { + "epoch": 0.59, + "learning_rate": 4.026801744036933e-05, + "loss": 0.5298, + "step": 15228 + }, + { + "epoch": 0.59, + "learning_rate": 4.02673762503206e-05, + "loss": 0.6394, + "step": 15229 + }, + { + "epoch": 0.59, + "learning_rate": 4.0266735060271864e-05, + "loss": 0.5835, + "step": 15230 + }, + { + "epoch": 0.59, + "learning_rate": 4.0266093870223135e-05, + "loss": 0.5149, + "step": 15231 + }, + { + "epoch": 0.59, + "learning_rate": 4.02654526801744e-05, + "loss": 0.4706, + "step": 15232 + }, + { + "epoch": 0.59, + "learning_rate": 4.026481149012568e-05, + "loss": 0.5511, + "step": 15233 + }, + { + "epoch": 0.59, + "learning_rate": 4.0264170300076944e-05, + "loss": 0.4604, + "step": 15234 + }, + { + "epoch": 0.59, + "learning_rate": 4.0263529110028215e-05, + "loss": 0.5207, + "step": 15235 + }, + { + "epoch": 0.59, + "learning_rate": 4.026288791997949e-05, + "loss": 0.5248, + "step": 15236 + }, + { + "epoch": 0.59, + "learning_rate": 4.026224672993075e-05, + "loss": 0.4722, + "step": 15237 + }, + { + "epoch": 0.59, + "learning_rate": 4.0261605539882024e-05, + "loss": 0.4883, + "step": 15238 + }, + { + "epoch": 0.59, + "learning_rate": 4.026096434983329e-05, + "loss": 0.5937, + "step": 15239 + }, + { + "epoch": 0.59, + "learning_rate": 4.026032315978456e-05, + "loss": 0.583, + "step": 15240 + }, + { + "epoch": 0.59, + "learning_rate": 4.025968196973583e-05, + "loss": 0.4938, + "step": 15241 + }, + { + "epoch": 0.59, + "learning_rate": 4.0259040779687104e-05, + "loss": 0.5881, + "step": 15242 + }, + { + "epoch": 0.59, + "learning_rate": 4.025839958963837e-05, + "loss": 0.4984, + "step": 15243 + }, + { + "epoch": 0.59, + "learning_rate": 4.025775839958964e-05, + "loss": 0.5798, + "step": 15244 + }, + { + "epoch": 0.59, + "learning_rate": 4.025711720954091e-05, + "loss": 0.5776, + "step": 15245 + }, + { + "epoch": 0.59, + "learning_rate": 4.025647601949218e-05, + "loss": 0.5983, + "step": 15246 + }, + { + "epoch": 0.59, + "learning_rate": 4.025583482944345e-05, + "loss": 0.4855, + "step": 15247 + }, + { + "epoch": 0.59, + "learning_rate": 4.0255193639394714e-05, + "loss": 0.4644, + "step": 15248 + }, + { + "epoch": 0.59, + "learning_rate": 4.025455244934599e-05, + "loss": 0.5539, + "step": 15249 + }, + { + "epoch": 0.59, + "learning_rate": 4.025391125929726e-05, + "loss": 0.5749, + "step": 15250 + }, + { + "epoch": 0.59, + "learning_rate": 4.025327006924853e-05, + "loss": 0.4572, + "step": 15251 + }, + { + "epoch": 0.59, + "learning_rate": 4.0252628879199794e-05, + "loss": 0.4652, + "step": 15252 + }, + { + "epoch": 0.59, + "learning_rate": 4.0251987689151066e-05, + "loss": 0.5618, + "step": 15253 + }, + { + "epoch": 0.59, + "learning_rate": 4.025134649910234e-05, + "loss": 0.6241, + "step": 15254 + }, + { + "epoch": 0.59, + "learning_rate": 4.02507053090536e-05, + "loss": 0.5296, + "step": 15255 + }, + { + "epoch": 0.59, + "learning_rate": 4.0250064119004874e-05, + "loss": 0.4542, + "step": 15256 + }, + { + "epoch": 0.59, + "learning_rate": 4.0249422928956146e-05, + "loss": 0.5051, + "step": 15257 + }, + { + "epoch": 0.59, + "learning_rate": 4.024878173890742e-05, + "loss": 0.4379, + "step": 15258 + }, + { + "epoch": 0.59, + "learning_rate": 4.024814054885868e-05, + "loss": 0.8004, + "step": 15259 + }, + { + "epoch": 0.59, + "learning_rate": 4.0247499358809954e-05, + "loss": 0.4657, + "step": 15260 + }, + { + "epoch": 0.59, + "learning_rate": 4.024685816876122e-05, + "loss": 0.4685, + "step": 15261 + }, + { + "epoch": 0.59, + "learning_rate": 4.024621697871249e-05, + "loss": 0.4982, + "step": 15262 + }, + { + "epoch": 0.59, + "learning_rate": 4.024557578866376e-05, + "loss": 0.5792, + "step": 15263 + }, + { + "epoch": 0.59, + "learning_rate": 4.024493459861503e-05, + "loss": 0.4834, + "step": 15264 + }, + { + "epoch": 0.59, + "learning_rate": 4.0244293408566306e-05, + "loss": 0.4591, + "step": 15265 + }, + { + "epoch": 0.59, + "learning_rate": 4.024365221851757e-05, + "loss": 0.4891, + "step": 15266 + }, + { + "epoch": 0.59, + "learning_rate": 4.024301102846884e-05, + "loss": 0.498, + "step": 15267 + }, + { + "epoch": 0.59, + "learning_rate": 4.024236983842011e-05, + "loss": 0.569, + "step": 15268 + }, + { + "epoch": 0.59, + "learning_rate": 4.024172864837138e-05, + "loss": 0.4201, + "step": 15269 + }, + { + "epoch": 0.59, + "learning_rate": 4.0241087458322644e-05, + "loss": 0.5411, + "step": 15270 + }, + { + "epoch": 0.59, + "learning_rate": 4.0240446268273916e-05, + "loss": 0.5329, + "step": 15271 + }, + { + "epoch": 0.59, + "learning_rate": 4.023980507822519e-05, + "loss": 0.624, + "step": 15272 + }, + { + "epoch": 0.59, + "learning_rate": 4.023916388817646e-05, + "loss": 0.4337, + "step": 15273 + }, + { + "epoch": 0.59, + "learning_rate": 4.023852269812773e-05, + "loss": 0.605, + "step": 15274 + }, + { + "epoch": 0.59, + "learning_rate": 4.0237881508078996e-05, + "loss": 0.5058, + "step": 15275 + }, + { + "epoch": 0.59, + "learning_rate": 4.023724031803027e-05, + "loss": 0.4915, + "step": 15276 + }, + { + "epoch": 0.59, + "learning_rate": 4.023659912798153e-05, + "loss": 0.5657, + "step": 15277 + }, + { + "epoch": 0.59, + "learning_rate": 4.0235957937932804e-05, + "loss": 0.5291, + "step": 15278 + }, + { + "epoch": 0.59, + "learning_rate": 4.023531674788407e-05, + "loss": 0.4649, + "step": 15279 + }, + { + "epoch": 0.59, + "learning_rate": 4.023467555783534e-05, + "loss": 0.4588, + "step": 15280 + }, + { + "epoch": 0.59, + "learning_rate": 4.023403436778661e-05, + "loss": 0.4456, + "step": 15281 + }, + { + "epoch": 0.59, + "learning_rate": 4.0233393177737884e-05, + "loss": 0.5986, + "step": 15282 + }, + { + "epoch": 0.59, + "learning_rate": 4.0232751987689156e-05, + "loss": 0.5872, + "step": 15283 + }, + { + "epoch": 0.59, + "learning_rate": 4.023211079764042e-05, + "loss": 0.5364, + "step": 15284 + }, + { + "epoch": 0.59, + "learning_rate": 4.023146960759169e-05, + "loss": 0.4485, + "step": 15285 + }, + { + "epoch": 0.59, + "learning_rate": 4.023082841754296e-05, + "loss": 0.5374, + "step": 15286 + }, + { + "epoch": 0.59, + "learning_rate": 4.023018722749423e-05, + "loss": 0.5081, + "step": 15287 + }, + { + "epoch": 0.59, + "learning_rate": 4.02295460374455e-05, + "loss": 0.5131, + "step": 15288 + }, + { + "epoch": 0.59, + "learning_rate": 4.022890484739677e-05, + "loss": 0.5677, + "step": 15289 + }, + { + "epoch": 0.59, + "learning_rate": 4.0228263657348045e-05, + "loss": 0.5394, + "step": 15290 + }, + { + "epoch": 0.59, + "learning_rate": 4.022762246729931e-05, + "loss": 0.5596, + "step": 15291 + }, + { + "epoch": 0.59, + "learning_rate": 4.022698127725058e-05, + "loss": 0.5249, + "step": 15292 + }, + { + "epoch": 0.59, + "learning_rate": 4.0226340087201846e-05, + "loss": 0.5386, + "step": 15293 + }, + { + "epoch": 0.59, + "learning_rate": 4.022569889715312e-05, + "loss": 0.4917, + "step": 15294 + }, + { + "epoch": 0.59, + "learning_rate": 4.022505770710438e-05, + "loss": 0.5834, + "step": 15295 + }, + { + "epoch": 0.59, + "learning_rate": 4.0224416517055655e-05, + "loss": 0.4297, + "step": 15296 + }, + { + "epoch": 0.59, + "learning_rate": 4.0223775327006926e-05, + "loss": 0.5418, + "step": 15297 + }, + { + "epoch": 0.59, + "learning_rate": 4.02231341369582e-05, + "loss": 0.6367, + "step": 15298 + }, + { + "epoch": 0.59, + "learning_rate": 4.022249294690947e-05, + "loss": 0.4757, + "step": 15299 + }, + { + "epoch": 0.59, + "learning_rate": 4.0221851756860735e-05, + "loss": 0.5697, + "step": 15300 + }, + { + "epoch": 0.59, + "learning_rate": 4.0221210566812006e-05, + "loss": 0.4769, + "step": 15301 + }, + { + "epoch": 0.59, + "learning_rate": 4.022056937676327e-05, + "loss": 0.4765, + "step": 15302 + }, + { + "epoch": 0.59, + "learning_rate": 4.021992818671454e-05, + "loss": 0.4722, + "step": 15303 + }, + { + "epoch": 0.59, + "learning_rate": 4.0219286996665815e-05, + "loss": 0.5792, + "step": 15304 + }, + { + "epoch": 0.59, + "learning_rate": 4.0218645806617087e-05, + "loss": 0.5482, + "step": 15305 + }, + { + "epoch": 0.59, + "learning_rate": 4.021800461656836e-05, + "loss": 0.6111, + "step": 15306 + }, + { + "epoch": 0.59, + "learning_rate": 4.021736342651962e-05, + "loss": 0.6031, + "step": 15307 + }, + { + "epoch": 0.59, + "learning_rate": 4.0216722236470895e-05, + "loss": 0.4317, + "step": 15308 + }, + { + "epoch": 0.59, + "learning_rate": 4.021608104642216e-05, + "loss": 0.5843, + "step": 15309 + }, + { + "epoch": 0.59, + "learning_rate": 4.021543985637343e-05, + "loss": 0.5601, + "step": 15310 + }, + { + "epoch": 0.59, + "learning_rate": 4.0214798666324696e-05, + "loss": 0.5991, + "step": 15311 + }, + { + "epoch": 0.59, + "learning_rate": 4.021415747627597e-05, + "loss": 0.6117, + "step": 15312 + }, + { + "epoch": 0.59, + "learning_rate": 4.021351628622724e-05, + "loss": 0.5187, + "step": 15313 + }, + { + "epoch": 0.59, + "learning_rate": 4.021287509617851e-05, + "loss": 0.646, + "step": 15314 + }, + { + "epoch": 0.59, + "learning_rate": 4.021223390612978e-05, + "loss": 0.5559, + "step": 15315 + }, + { + "epoch": 0.59, + "learning_rate": 4.021159271608105e-05, + "loss": 0.4742, + "step": 15316 + }, + { + "epoch": 0.59, + "learning_rate": 4.021095152603232e-05, + "loss": 0.6448, + "step": 15317 + }, + { + "epoch": 0.59, + "learning_rate": 4.0210310335983585e-05, + "loss": 0.5342, + "step": 15318 + }, + { + "epoch": 0.59, + "learning_rate": 4.020966914593486e-05, + "loss": 0.6407, + "step": 15319 + }, + { + "epoch": 0.59, + "learning_rate": 4.020902795588612e-05, + "loss": 0.534, + "step": 15320 + }, + { + "epoch": 0.59, + "learning_rate": 4.02083867658374e-05, + "loss": 0.5663, + "step": 15321 + }, + { + "epoch": 0.59, + "learning_rate": 4.0207745575788665e-05, + "loss": 0.61, + "step": 15322 + }, + { + "epoch": 0.59, + "learning_rate": 4.020710438573994e-05, + "loss": 0.5929, + "step": 15323 + }, + { + "epoch": 0.59, + "learning_rate": 4.020646319569121e-05, + "loss": 0.4973, + "step": 15324 + }, + { + "epoch": 0.59, + "learning_rate": 4.0205822005642473e-05, + "loss": 0.5616, + "step": 15325 + }, + { + "epoch": 0.59, + "learning_rate": 4.0205180815593745e-05, + "loss": 0.4811, + "step": 15326 + }, + { + "epoch": 0.59, + "learning_rate": 4.020453962554501e-05, + "loss": 0.6051, + "step": 15327 + }, + { + "epoch": 0.59, + "learning_rate": 4.020389843549628e-05, + "loss": 0.474, + "step": 15328 + }, + { + "epoch": 0.59, + "learning_rate": 4.0203257245447554e-05, + "loss": 0.5755, + "step": 15329 + }, + { + "epoch": 0.59, + "learning_rate": 4.0202616055398825e-05, + "loss": 0.5239, + "step": 15330 + }, + { + "epoch": 0.59, + "learning_rate": 4.020197486535009e-05, + "loss": 0.5653, + "step": 15331 + }, + { + "epoch": 0.59, + "learning_rate": 4.020133367530136e-05, + "loss": 0.5577, + "step": 15332 + }, + { + "epoch": 0.59, + "learning_rate": 4.0200692485252634e-05, + "loss": 0.535, + "step": 15333 + }, + { + "epoch": 0.59, + "learning_rate": 4.02000512952039e-05, + "loss": 0.5345, + "step": 15334 + }, + { + "epoch": 0.59, + "learning_rate": 4.019941010515517e-05, + "loss": 0.4653, + "step": 15335 + }, + { + "epoch": 0.59, + "learning_rate": 4.0198768915106435e-05, + "loss": 0.6142, + "step": 15336 + }, + { + "epoch": 0.59, + "learning_rate": 4.0198127725057714e-05, + "loss": 0.5013, + "step": 15337 + }, + { + "epoch": 0.59, + "learning_rate": 4.019748653500898e-05, + "loss": 0.4772, + "step": 15338 + }, + { + "epoch": 0.59, + "learning_rate": 4.019684534496025e-05, + "loss": 0.4017, + "step": 15339 + }, + { + "epoch": 0.59, + "learning_rate": 4.0196204154911515e-05, + "loss": 0.4529, + "step": 15340 + }, + { + "epoch": 0.59, + "learning_rate": 4.019556296486279e-05, + "loss": 0.5794, + "step": 15341 + }, + { + "epoch": 0.59, + "learning_rate": 4.019492177481406e-05, + "loss": 0.5586, + "step": 15342 + }, + { + "epoch": 0.59, + "learning_rate": 4.0194280584765324e-05, + "loss": 0.4093, + "step": 15343 + }, + { + "epoch": 0.59, + "learning_rate": 4.0193639394716595e-05, + "loss": 0.5253, + "step": 15344 + }, + { + "epoch": 0.59, + "learning_rate": 4.019299820466787e-05, + "loss": 0.5117, + "step": 15345 + }, + { + "epoch": 0.59, + "learning_rate": 4.019235701461914e-05, + "loss": 0.5296, + "step": 15346 + }, + { + "epoch": 0.59, + "learning_rate": 4.0191715824570404e-05, + "loss": 0.47, + "step": 15347 + }, + { + "epoch": 0.59, + "learning_rate": 4.0191074634521675e-05, + "loss": 0.5642, + "step": 15348 + }, + { + "epoch": 0.59, + "learning_rate": 4.019043344447294e-05, + "loss": 0.4681, + "step": 15349 + }, + { + "epoch": 0.59, + "learning_rate": 4.018979225442421e-05, + "loss": 0.5233, + "step": 15350 + }, + { + "epoch": 0.59, + "learning_rate": 4.0189151064375484e-05, + "loss": 0.5789, + "step": 15351 + }, + { + "epoch": 0.59, + "learning_rate": 4.018850987432675e-05, + "loss": 0.6446, + "step": 15352 + }, + { + "epoch": 0.59, + "learning_rate": 4.018786868427802e-05, + "loss": 0.5145, + "step": 15353 + }, + { + "epoch": 0.59, + "learning_rate": 4.018722749422929e-05, + "loss": 0.5597, + "step": 15354 + }, + { + "epoch": 0.59, + "learning_rate": 4.0186586304180564e-05, + "loss": 0.4449, + "step": 15355 + }, + { + "epoch": 0.59, + "learning_rate": 4.018594511413183e-05, + "loss": 0.5951, + "step": 15356 + }, + { + "epoch": 0.59, + "learning_rate": 4.01853039240831e-05, + "loss": 0.558, + "step": 15357 + }, + { + "epoch": 0.59, + "learning_rate": 4.0184662734034366e-05, + "loss": 0.4333, + "step": 15358 + }, + { + "epoch": 0.59, + "learning_rate": 4.018402154398564e-05, + "loss": 0.47, + "step": 15359 + }, + { + "epoch": 0.59, + "learning_rate": 4.018338035393691e-05, + "loss": 0.5366, + "step": 15360 + }, + { + "epoch": 0.59, + "learning_rate": 4.018273916388818e-05, + "loss": 0.505, + "step": 15361 + }, + { + "epoch": 0.59, + "learning_rate": 4.018209797383945e-05, + "loss": 0.5751, + "step": 15362 + }, + { + "epoch": 0.59, + "learning_rate": 4.018145678379072e-05, + "loss": 0.7288, + "step": 15363 + }, + { + "epoch": 0.59, + "learning_rate": 4.018081559374199e-05, + "loss": 0.6227, + "step": 15364 + }, + { + "epoch": 0.59, + "learning_rate": 4.0180174403693254e-05, + "loss": 0.5561, + "step": 15365 + }, + { + "epoch": 0.59, + "learning_rate": 4.0179533213644526e-05, + "loss": 0.5295, + "step": 15366 + }, + { + "epoch": 0.59, + "learning_rate": 4.017889202359579e-05, + "loss": 0.4767, + "step": 15367 + }, + { + "epoch": 0.59, + "learning_rate": 4.017825083354706e-05, + "loss": 0.7303, + "step": 15368 + }, + { + "epoch": 0.59, + "learning_rate": 4.0177609643498334e-05, + "loss": 0.5007, + "step": 15369 + }, + { + "epoch": 0.59, + "learning_rate": 4.0176968453449606e-05, + "loss": 0.4586, + "step": 15370 + }, + { + "epoch": 0.59, + "learning_rate": 4.017632726340088e-05, + "loss": 0.5955, + "step": 15371 + }, + { + "epoch": 0.59, + "learning_rate": 4.017568607335214e-05, + "loss": 0.609, + "step": 15372 + }, + { + "epoch": 0.59, + "learning_rate": 4.0175044883303414e-05, + "loss": 0.5084, + "step": 15373 + }, + { + "epoch": 0.59, + "learning_rate": 4.017440369325468e-05, + "loss": 0.462, + "step": 15374 + }, + { + "epoch": 0.59, + "learning_rate": 4.017376250320595e-05, + "loss": 0.4979, + "step": 15375 + }, + { + "epoch": 0.59, + "learning_rate": 4.017312131315722e-05, + "loss": 0.4594, + "step": 15376 + }, + { + "epoch": 0.59, + "learning_rate": 4.0172480123108494e-05, + "loss": 0.6078, + "step": 15377 + }, + { + "epoch": 0.59, + "learning_rate": 4.0171838933059766e-05, + "loss": 0.4602, + "step": 15378 + }, + { + "epoch": 0.59, + "learning_rate": 4.017119774301103e-05, + "loss": 0.5134, + "step": 15379 + }, + { + "epoch": 0.59, + "learning_rate": 4.01705565529623e-05, + "loss": 0.516, + "step": 15380 + }, + { + "epoch": 0.59, + "learning_rate": 4.016991536291357e-05, + "loss": 0.606, + "step": 15381 + }, + { + "epoch": 0.59, + "learning_rate": 4.016927417286484e-05, + "loss": 0.5767, + "step": 15382 + }, + { + "epoch": 0.59, + "learning_rate": 4.0168632982816104e-05, + "loss": 0.5977, + "step": 15383 + }, + { + "epoch": 0.59, + "learning_rate": 4.0167991792767376e-05, + "loss": 0.5044, + "step": 15384 + }, + { + "epoch": 0.59, + "learning_rate": 4.016735060271865e-05, + "loss": 0.5235, + "step": 15385 + }, + { + "epoch": 0.59, + "learning_rate": 4.016670941266992e-05, + "loss": 0.4507, + "step": 15386 + }, + { + "epoch": 0.59, + "learning_rate": 4.016606822262119e-05, + "loss": 0.5405, + "step": 15387 + }, + { + "epoch": 0.59, + "learning_rate": 4.0165427032572456e-05, + "loss": 0.5951, + "step": 15388 + }, + { + "epoch": 0.59, + "learning_rate": 4.016478584252373e-05, + "loss": 0.5484, + "step": 15389 + }, + { + "epoch": 0.59, + "learning_rate": 4.016414465247499e-05, + "loss": 0.5925, + "step": 15390 + }, + { + "epoch": 0.59, + "learning_rate": 4.0163503462426264e-05, + "loss": 0.5283, + "step": 15391 + }, + { + "epoch": 0.59, + "learning_rate": 4.0162862272377536e-05, + "loss": 0.5849, + "step": 15392 + }, + { + "epoch": 0.59, + "learning_rate": 4.016222108232881e-05, + "loss": 0.4959, + "step": 15393 + }, + { + "epoch": 0.59, + "learning_rate": 4.016157989228007e-05, + "loss": 0.5279, + "step": 15394 + }, + { + "epoch": 0.59, + "learning_rate": 4.0160938702231345e-05, + "loss": 0.5009, + "step": 15395 + }, + { + "epoch": 0.59, + "learning_rate": 4.0160297512182616e-05, + "loss": 0.5562, + "step": 15396 + }, + { + "epoch": 0.59, + "learning_rate": 4.015965632213388e-05, + "loss": 0.602, + "step": 15397 + }, + { + "epoch": 0.59, + "learning_rate": 4.015901513208515e-05, + "loss": 0.4103, + "step": 15398 + }, + { + "epoch": 0.59, + "learning_rate": 4.015837394203642e-05, + "loss": 0.6087, + "step": 15399 + }, + { + "epoch": 0.59, + "learning_rate": 4.015773275198769e-05, + "loss": 0.5564, + "step": 15400 + }, + { + "epoch": 0.59, + "learning_rate": 4.015709156193896e-05, + "loss": 0.5444, + "step": 15401 + }, + { + "epoch": 0.59, + "learning_rate": 4.015645037189023e-05, + "loss": 0.4839, + "step": 15402 + }, + { + "epoch": 0.59, + "learning_rate": 4.01558091818415e-05, + "loss": 0.5403, + "step": 15403 + }, + { + "epoch": 0.59, + "learning_rate": 4.015516799179277e-05, + "loss": 0.4745, + "step": 15404 + }, + { + "epoch": 0.59, + "learning_rate": 4.015452680174404e-05, + "loss": 0.5585, + "step": 15405 + }, + { + "epoch": 0.59, + "learning_rate": 4.0153885611695306e-05, + "loss": 0.5043, + "step": 15406 + }, + { + "epoch": 0.59, + "learning_rate": 4.015324442164658e-05, + "loss": 0.5581, + "step": 15407 + }, + { + "epoch": 0.59, + "learning_rate": 4.015260323159784e-05, + "loss": 0.5685, + "step": 15408 + }, + { + "epoch": 0.59, + "learning_rate": 4.015196204154912e-05, + "loss": 0.5731, + "step": 15409 + }, + { + "epoch": 0.59, + "learning_rate": 4.0151320851500386e-05, + "loss": 0.4854, + "step": 15410 + }, + { + "epoch": 0.59, + "learning_rate": 4.015067966145166e-05, + "loss": 0.5049, + "step": 15411 + }, + { + "epoch": 0.59, + "learning_rate": 4.015003847140292e-05, + "loss": 0.502, + "step": 15412 + }, + { + "epoch": 0.59, + "learning_rate": 4.0149397281354195e-05, + "loss": 0.564, + "step": 15413 + }, + { + "epoch": 0.59, + "learning_rate": 4.0148756091305466e-05, + "loss": 0.4996, + "step": 15414 + }, + { + "epoch": 0.59, + "learning_rate": 4.014811490125673e-05, + "loss": 0.6287, + "step": 15415 + }, + { + "epoch": 0.59, + "learning_rate": 4.0147473711208e-05, + "loss": 0.4341, + "step": 15416 + }, + { + "epoch": 0.59, + "learning_rate": 4.0146832521159275e-05, + "loss": 0.5757, + "step": 15417 + }, + { + "epoch": 0.59, + "learning_rate": 4.0146191331110547e-05, + "loss": 0.5765, + "step": 15418 + }, + { + "epoch": 0.59, + "learning_rate": 4.014555014106181e-05, + "loss": 0.6131, + "step": 15419 + }, + { + "epoch": 0.59, + "learning_rate": 4.014490895101308e-05, + "loss": 0.447, + "step": 15420 + }, + { + "epoch": 0.59, + "learning_rate": 4.014426776096435e-05, + "loss": 0.5296, + "step": 15421 + }, + { + "epoch": 0.59, + "learning_rate": 4.014362657091562e-05, + "loss": 0.5121, + "step": 15422 + }, + { + "epoch": 0.59, + "learning_rate": 4.014298538086689e-05, + "loss": 0.5005, + "step": 15423 + }, + { + "epoch": 0.59, + "learning_rate": 4.0142344190818157e-05, + "loss": 0.5483, + "step": 15424 + }, + { + "epoch": 0.59, + "learning_rate": 4.0141703000769435e-05, + "loss": 0.5841, + "step": 15425 + }, + { + "epoch": 0.59, + "learning_rate": 4.01410618107207e-05, + "loss": 0.5907, + "step": 15426 + }, + { + "epoch": 0.59, + "learning_rate": 4.014042062067197e-05, + "loss": 0.6176, + "step": 15427 + }, + { + "epoch": 0.59, + "learning_rate": 4.0139779430623237e-05, + "loss": 0.5223, + "step": 15428 + }, + { + "epoch": 0.59, + "learning_rate": 4.013913824057451e-05, + "loss": 0.5133, + "step": 15429 + }, + { + "epoch": 0.59, + "learning_rate": 4.013849705052577e-05, + "loss": 0.4723, + "step": 15430 + }, + { + "epoch": 0.59, + "learning_rate": 4.0137855860477045e-05, + "loss": 0.4904, + "step": 15431 + }, + { + "epoch": 0.59, + "learning_rate": 4.013721467042832e-05, + "loss": 0.4719, + "step": 15432 + }, + { + "epoch": 0.59, + "learning_rate": 4.013657348037959e-05, + "loss": 0.49, + "step": 15433 + }, + { + "epoch": 0.59, + "learning_rate": 4.013593229033086e-05, + "loss": 0.5043, + "step": 15434 + }, + { + "epoch": 0.59, + "learning_rate": 4.0135291100282125e-05, + "loss": 0.4372, + "step": 15435 + }, + { + "epoch": 0.59, + "learning_rate": 4.01346499102334e-05, + "loss": 0.4734, + "step": 15436 + }, + { + "epoch": 0.59, + "learning_rate": 4.013400872018466e-05, + "loss": 0.5386, + "step": 15437 + }, + { + "epoch": 0.59, + "learning_rate": 4.0133367530135933e-05, + "loss": 0.5631, + "step": 15438 + }, + { + "epoch": 0.59, + "learning_rate": 4.01327263400872e-05, + "loss": 0.5111, + "step": 15439 + }, + { + "epoch": 0.59, + "learning_rate": 4.013208515003847e-05, + "loss": 0.5358, + "step": 15440 + }, + { + "epoch": 0.59, + "learning_rate": 4.013144395998974e-05, + "loss": 0.4679, + "step": 15441 + }, + { + "epoch": 0.59, + "learning_rate": 4.0130802769941014e-05, + "loss": 0.5646, + "step": 15442 + }, + { + "epoch": 0.59, + "learning_rate": 4.0130161579892285e-05, + "loss": 0.5142, + "step": 15443 + }, + { + "epoch": 0.59, + "learning_rate": 4.012952038984355e-05, + "loss": 0.5075, + "step": 15444 + }, + { + "epoch": 0.59, + "learning_rate": 4.012887919979482e-05, + "loss": 0.5432, + "step": 15445 + }, + { + "epoch": 0.59, + "learning_rate": 4.012823800974609e-05, + "loss": 0.4232, + "step": 15446 + }, + { + "epoch": 0.59, + "learning_rate": 4.012759681969736e-05, + "loss": 0.5453, + "step": 15447 + }, + { + "epoch": 0.59, + "learning_rate": 4.012695562964863e-05, + "loss": 0.5338, + "step": 15448 + }, + { + "epoch": 0.59, + "learning_rate": 4.01263144395999e-05, + "loss": 0.4857, + "step": 15449 + }, + { + "epoch": 0.59, + "learning_rate": 4.0125673249551174e-05, + "loss": 0.5503, + "step": 15450 + }, + { + "epoch": 0.59, + "learning_rate": 4.012503205950244e-05, + "loss": 0.4761, + "step": 15451 + }, + { + "epoch": 0.59, + "learning_rate": 4.012439086945371e-05, + "loss": 0.5228, + "step": 15452 + }, + { + "epoch": 0.59, + "learning_rate": 4.0123749679404975e-05, + "loss": 0.5265, + "step": 15453 + }, + { + "epoch": 0.59, + "learning_rate": 4.012310848935625e-05, + "loss": 0.5829, + "step": 15454 + }, + { + "epoch": 0.59, + "learning_rate": 4.012246729930751e-05, + "loss": 0.6186, + "step": 15455 + }, + { + "epoch": 0.59, + "learning_rate": 4.0121826109258784e-05, + "loss": 0.45, + "step": 15456 + }, + { + "epoch": 0.59, + "learning_rate": 4.0121184919210055e-05, + "loss": 0.5552, + "step": 15457 + }, + { + "epoch": 0.59, + "learning_rate": 4.012054372916133e-05, + "loss": 0.6128, + "step": 15458 + }, + { + "epoch": 0.59, + "learning_rate": 4.01199025391126e-05, + "loss": 0.4965, + "step": 15459 + }, + { + "epoch": 0.59, + "learning_rate": 4.0119261349063864e-05, + "loss": 0.564, + "step": 15460 + }, + { + "epoch": 0.59, + "learning_rate": 4.0118620159015135e-05, + "loss": 0.5282, + "step": 15461 + }, + { + "epoch": 0.59, + "learning_rate": 4.01179789689664e-05, + "loss": 0.5669, + "step": 15462 + }, + { + "epoch": 0.59, + "learning_rate": 4.011733777891767e-05, + "loss": 0.4468, + "step": 15463 + }, + { + "epoch": 0.59, + "learning_rate": 4.0116696588868944e-05, + "loss": 0.5868, + "step": 15464 + }, + { + "epoch": 0.59, + "learning_rate": 4.0116055398820216e-05, + "loss": 0.5264, + "step": 15465 + }, + { + "epoch": 0.59, + "learning_rate": 4.011541420877148e-05, + "loss": 0.5813, + "step": 15466 + }, + { + "epoch": 0.59, + "learning_rate": 4.011477301872275e-05, + "loss": 0.5616, + "step": 15467 + }, + { + "epoch": 0.59, + "learning_rate": 4.0114131828674024e-05, + "loss": 0.5303, + "step": 15468 + }, + { + "epoch": 0.59, + "learning_rate": 4.011349063862529e-05, + "loss": 0.5172, + "step": 15469 + }, + { + "epoch": 0.59, + "learning_rate": 4.011284944857656e-05, + "loss": 0.5255, + "step": 15470 + }, + { + "epoch": 0.59, + "learning_rate": 4.0112208258527826e-05, + "loss": 0.5503, + "step": 15471 + }, + { + "epoch": 0.59, + "learning_rate": 4.01115670684791e-05, + "loss": 0.4278, + "step": 15472 + }, + { + "epoch": 0.59, + "learning_rate": 4.011092587843037e-05, + "loss": 0.5645, + "step": 15473 + }, + { + "epoch": 0.59, + "learning_rate": 4.011028468838164e-05, + "loss": 0.5021, + "step": 15474 + }, + { + "epoch": 0.59, + "learning_rate": 4.0109643498332906e-05, + "loss": 0.4571, + "step": 15475 + }, + { + "epoch": 0.6, + "learning_rate": 4.010900230828418e-05, + "loss": 0.6133, + "step": 15476 + }, + { + "epoch": 0.6, + "learning_rate": 4.010836111823545e-05, + "loss": 0.6002, + "step": 15477 + }, + { + "epoch": 0.6, + "learning_rate": 4.0107719928186714e-05, + "loss": 0.4991, + "step": 15478 + }, + { + "epoch": 0.6, + "learning_rate": 4.0107078738137986e-05, + "loss": 0.5732, + "step": 15479 + }, + { + "epoch": 0.6, + "learning_rate": 4.010643754808925e-05, + "loss": 0.4849, + "step": 15480 + }, + { + "epoch": 0.6, + "learning_rate": 4.010579635804053e-05, + "loss": 0.5926, + "step": 15481 + }, + { + "epoch": 0.6, + "learning_rate": 4.0105155167991794e-05, + "loss": 0.4876, + "step": 15482 + }, + { + "epoch": 0.6, + "learning_rate": 4.0104513977943066e-05, + "loss": 0.5227, + "step": 15483 + }, + { + "epoch": 0.6, + "learning_rate": 4.010387278789434e-05, + "loss": 0.5608, + "step": 15484 + }, + { + "epoch": 0.6, + "learning_rate": 4.01032315978456e-05, + "loss": 0.5406, + "step": 15485 + }, + { + "epoch": 0.6, + "learning_rate": 4.0102590407796874e-05, + "loss": 0.4378, + "step": 15486 + }, + { + "epoch": 0.6, + "learning_rate": 4.010194921774814e-05, + "loss": 0.519, + "step": 15487 + }, + { + "epoch": 0.6, + "learning_rate": 4.010130802769941e-05, + "loss": 0.5761, + "step": 15488 + }, + { + "epoch": 0.6, + "learning_rate": 4.010066683765068e-05, + "loss": 0.4522, + "step": 15489 + }, + { + "epoch": 0.6, + "learning_rate": 4.0100025647601954e-05, + "loss": 0.4643, + "step": 15490 + }, + { + "epoch": 0.6, + "learning_rate": 4.009938445755322e-05, + "loss": 0.5142, + "step": 15491 + }, + { + "epoch": 0.6, + "learning_rate": 4.009874326750449e-05, + "loss": 0.4306, + "step": 15492 + }, + { + "epoch": 0.6, + "learning_rate": 4.009810207745576e-05, + "loss": 0.5583, + "step": 15493 + }, + { + "epoch": 0.6, + "learning_rate": 4.009746088740703e-05, + "loss": 0.4925, + "step": 15494 + }, + { + "epoch": 0.6, + "learning_rate": 4.00968196973583e-05, + "loss": 0.5556, + "step": 15495 + }, + { + "epoch": 0.6, + "learning_rate": 4.0096178507309564e-05, + "loss": 0.5262, + "step": 15496 + }, + { + "epoch": 0.6, + "learning_rate": 4.009553731726084e-05, + "loss": 0.5538, + "step": 15497 + }, + { + "epoch": 0.6, + "learning_rate": 4.009489612721211e-05, + "loss": 0.5271, + "step": 15498 + }, + { + "epoch": 0.6, + "learning_rate": 4.009425493716338e-05, + "loss": 0.5235, + "step": 15499 + }, + { + "epoch": 0.6, + "learning_rate": 4.0093613747114644e-05, + "loss": 0.5528, + "step": 15500 + }, + { + "epoch": 0.6, + "learning_rate": 4.0092972557065916e-05, + "loss": 0.5566, + "step": 15501 + }, + { + "epoch": 0.6, + "learning_rate": 4.009233136701719e-05, + "loss": 0.6002, + "step": 15502 + }, + { + "epoch": 0.6, + "learning_rate": 4.009169017696845e-05, + "loss": 0.6167, + "step": 15503 + }, + { + "epoch": 0.6, + "learning_rate": 4.0091048986919724e-05, + "loss": 0.5579, + "step": 15504 + }, + { + "epoch": 0.6, + "learning_rate": 4.0090407796870996e-05, + "loss": 0.6685, + "step": 15505 + }, + { + "epoch": 0.6, + "learning_rate": 4.008976660682227e-05, + "loss": 0.5359, + "step": 15506 + }, + { + "epoch": 0.6, + "learning_rate": 4.008912541677353e-05, + "loss": 0.6249, + "step": 15507 + }, + { + "epoch": 0.6, + "learning_rate": 4.0088484226724805e-05, + "loss": 0.4813, + "step": 15508 + }, + { + "epoch": 0.6, + "learning_rate": 4.008784303667607e-05, + "loss": 0.5733, + "step": 15509 + }, + { + "epoch": 0.6, + "learning_rate": 4.008720184662734e-05, + "loss": 0.5093, + "step": 15510 + }, + { + "epoch": 0.6, + "learning_rate": 4.008656065657861e-05, + "loss": 0.4211, + "step": 15511 + }, + { + "epoch": 0.6, + "learning_rate": 4.008591946652988e-05, + "loss": 0.5154, + "step": 15512 + }, + { + "epoch": 0.6, + "learning_rate": 4.0085278276481156e-05, + "loss": 0.4915, + "step": 15513 + }, + { + "epoch": 0.6, + "learning_rate": 4.008463708643242e-05, + "loss": 0.5233, + "step": 15514 + }, + { + "epoch": 0.6, + "learning_rate": 4.008399589638369e-05, + "loss": 0.4467, + "step": 15515 + }, + { + "epoch": 0.6, + "learning_rate": 4.008335470633496e-05, + "loss": 0.5281, + "step": 15516 + }, + { + "epoch": 0.6, + "learning_rate": 4.008271351628623e-05, + "loss": 0.5528, + "step": 15517 + }, + { + "epoch": 0.6, + "learning_rate": 4.0082072326237495e-05, + "loss": 0.6026, + "step": 15518 + }, + { + "epoch": 0.6, + "learning_rate": 4.0081431136188766e-05, + "loss": 0.6031, + "step": 15519 + }, + { + "epoch": 0.6, + "learning_rate": 4.008078994614004e-05, + "loss": 0.5929, + "step": 15520 + }, + { + "epoch": 0.6, + "learning_rate": 4.008014875609131e-05, + "loss": 0.5312, + "step": 15521 + }, + { + "epoch": 0.6, + "learning_rate": 4.007950756604258e-05, + "loss": 0.5033, + "step": 15522 + }, + { + "epoch": 0.6, + "learning_rate": 4.0078866375993846e-05, + "loss": 0.4603, + "step": 15523 + }, + { + "epoch": 0.6, + "learning_rate": 4.007822518594512e-05, + "loss": 0.539, + "step": 15524 + }, + { + "epoch": 0.6, + "learning_rate": 4.007758399589638e-05, + "loss": 0.5877, + "step": 15525 + }, + { + "epoch": 0.6, + "learning_rate": 4.0076942805847655e-05, + "loss": 0.5179, + "step": 15526 + }, + { + "epoch": 0.6, + "learning_rate": 4.007630161579892e-05, + "loss": 0.5624, + "step": 15527 + }, + { + "epoch": 0.6, + "learning_rate": 4.007566042575019e-05, + "loss": 0.637, + "step": 15528 + }, + { + "epoch": 0.6, + "learning_rate": 4.007501923570146e-05, + "loss": 0.6183, + "step": 15529 + }, + { + "epoch": 0.6, + "learning_rate": 4.0074378045652735e-05, + "loss": 0.6338, + "step": 15530 + }, + { + "epoch": 0.6, + "learning_rate": 4.0073736855604007e-05, + "loss": 0.6401, + "step": 15531 + }, + { + "epoch": 0.6, + "learning_rate": 4.007309566555527e-05, + "loss": 0.512, + "step": 15532 + }, + { + "epoch": 0.6, + "learning_rate": 4.007245447550654e-05, + "loss": 0.5342, + "step": 15533 + }, + { + "epoch": 0.6, + "learning_rate": 4.007181328545781e-05, + "loss": 0.5381, + "step": 15534 + }, + { + "epoch": 0.6, + "learning_rate": 4.007117209540908e-05, + "loss": 0.505, + "step": 15535 + }, + { + "epoch": 0.6, + "learning_rate": 4.007053090536035e-05, + "loss": 0.514, + "step": 15536 + }, + { + "epoch": 0.6, + "learning_rate": 4.006988971531162e-05, + "loss": 0.4916, + "step": 15537 + }, + { + "epoch": 0.6, + "learning_rate": 4.0069248525262895e-05, + "loss": 0.5458, + "step": 15538 + }, + { + "epoch": 0.6, + "learning_rate": 4.006860733521416e-05, + "loss": 0.664, + "step": 15539 + }, + { + "epoch": 0.6, + "learning_rate": 4.006796614516543e-05, + "loss": 0.517, + "step": 15540 + }, + { + "epoch": 0.6, + "learning_rate": 4.00673249551167e-05, + "loss": 0.4252, + "step": 15541 + }, + { + "epoch": 0.6, + "learning_rate": 4.006668376506797e-05, + "loss": 0.5605, + "step": 15542 + }, + { + "epoch": 0.6, + "learning_rate": 4.006604257501923e-05, + "loss": 0.5126, + "step": 15543 + }, + { + "epoch": 0.6, + "learning_rate": 4.0065401384970505e-05, + "loss": 0.5047, + "step": 15544 + }, + { + "epoch": 0.6, + "learning_rate": 4.006476019492178e-05, + "loss": 0.5391, + "step": 15545 + }, + { + "epoch": 0.6, + "learning_rate": 4.006411900487305e-05, + "loss": 0.5995, + "step": 15546 + }, + { + "epoch": 0.6, + "learning_rate": 4.006347781482432e-05, + "loss": 0.4841, + "step": 15547 + }, + { + "epoch": 0.6, + "learning_rate": 4.0062836624775585e-05, + "loss": 0.5485, + "step": 15548 + }, + { + "epoch": 0.6, + "learning_rate": 4.006219543472686e-05, + "loss": 0.5614, + "step": 15549 + }, + { + "epoch": 0.6, + "learning_rate": 4.006155424467812e-05, + "loss": 0.5414, + "step": 15550 + }, + { + "epoch": 0.6, + "learning_rate": 4.0060913054629393e-05, + "loss": 0.4519, + "step": 15551 + }, + { + "epoch": 0.6, + "learning_rate": 4.0060271864580665e-05, + "loss": 0.5274, + "step": 15552 + }, + { + "epoch": 0.6, + "learning_rate": 4.005963067453194e-05, + "loss": 0.5196, + "step": 15553 + }, + { + "epoch": 0.6, + "learning_rate": 4.00589894844832e-05, + "loss": 0.4344, + "step": 15554 + }, + { + "epoch": 0.6, + "learning_rate": 4.0058348294434474e-05, + "loss": 0.6002, + "step": 15555 + }, + { + "epoch": 0.6, + "learning_rate": 4.0057707104385745e-05, + "loss": 0.5168, + "step": 15556 + }, + { + "epoch": 0.6, + "learning_rate": 4.005706591433701e-05, + "loss": 0.6346, + "step": 15557 + }, + { + "epoch": 0.6, + "learning_rate": 4.005642472428828e-05, + "loss": 0.5513, + "step": 15558 + }, + { + "epoch": 0.6, + "learning_rate": 4.005578353423955e-05, + "loss": 0.461, + "step": 15559 + }, + { + "epoch": 0.6, + "learning_rate": 4.005514234419082e-05, + "loss": 0.547, + "step": 15560 + }, + { + "epoch": 0.6, + "learning_rate": 4.005450115414209e-05, + "loss": 0.6419, + "step": 15561 + }, + { + "epoch": 0.6, + "learning_rate": 4.005385996409336e-05, + "loss": 0.4219, + "step": 15562 + }, + { + "epoch": 0.6, + "learning_rate": 4.005321877404463e-05, + "loss": 0.5671, + "step": 15563 + }, + { + "epoch": 0.6, + "learning_rate": 4.00525775839959e-05, + "loss": 0.5608, + "step": 15564 + }, + { + "epoch": 0.6, + "learning_rate": 4.005193639394717e-05, + "loss": 0.5883, + "step": 15565 + }, + { + "epoch": 0.6, + "learning_rate": 4.0051295203898435e-05, + "loss": 0.4817, + "step": 15566 + }, + { + "epoch": 0.6, + "learning_rate": 4.005065401384971e-05, + "loss": 0.6181, + "step": 15567 + }, + { + "epoch": 0.6, + "learning_rate": 4.005001282380097e-05, + "loss": 0.5053, + "step": 15568 + }, + { + "epoch": 0.6, + "learning_rate": 4.004937163375225e-05, + "loss": 0.3918, + "step": 15569 + }, + { + "epoch": 0.6, + "learning_rate": 4.0048730443703515e-05, + "loss": 0.6974, + "step": 15570 + }, + { + "epoch": 0.6, + "learning_rate": 4.004808925365479e-05, + "loss": 0.5616, + "step": 15571 + }, + { + "epoch": 0.6, + "learning_rate": 4.004744806360605e-05, + "loss": 0.548, + "step": 15572 + }, + { + "epoch": 0.6, + "learning_rate": 4.0046806873557324e-05, + "loss": 0.5225, + "step": 15573 + }, + { + "epoch": 0.6, + "learning_rate": 4.0046165683508596e-05, + "loss": 0.5101, + "step": 15574 + }, + { + "epoch": 0.6, + "learning_rate": 4.004552449345986e-05, + "loss": 0.479, + "step": 15575 + }, + { + "epoch": 0.6, + "learning_rate": 4.004488330341113e-05, + "loss": 0.61, + "step": 15576 + }, + { + "epoch": 0.6, + "learning_rate": 4.0044242113362404e-05, + "loss": 0.5797, + "step": 15577 + }, + { + "epoch": 0.6, + "learning_rate": 4.0043600923313676e-05, + "loss": 0.5553, + "step": 15578 + }, + { + "epoch": 0.6, + "learning_rate": 4.004295973326494e-05, + "loss": 0.5374, + "step": 15579 + }, + { + "epoch": 0.6, + "learning_rate": 4.004231854321621e-05, + "loss": 0.4984, + "step": 15580 + }, + { + "epoch": 0.6, + "learning_rate": 4.004167735316748e-05, + "loss": 0.5997, + "step": 15581 + }, + { + "epoch": 0.6, + "learning_rate": 4.004103616311875e-05, + "loss": 0.6198, + "step": 15582 + }, + { + "epoch": 0.6, + "learning_rate": 4.004039497307002e-05, + "loss": 0.5014, + "step": 15583 + }, + { + "epoch": 0.6, + "learning_rate": 4.0039753783021286e-05, + "loss": 0.4854, + "step": 15584 + }, + { + "epoch": 0.6, + "learning_rate": 4.0039112592972564e-05, + "loss": 0.5129, + "step": 15585 + }, + { + "epoch": 0.6, + "learning_rate": 4.003847140292383e-05, + "loss": 0.5749, + "step": 15586 + }, + { + "epoch": 0.6, + "learning_rate": 4.00378302128751e-05, + "loss": 0.5111, + "step": 15587 + }, + { + "epoch": 0.6, + "learning_rate": 4.0037189022826366e-05, + "loss": 0.562, + "step": 15588 + }, + { + "epoch": 0.6, + "learning_rate": 4.003654783277764e-05, + "loss": 0.5441, + "step": 15589 + }, + { + "epoch": 0.6, + "learning_rate": 4.00359066427289e-05, + "loss": 0.6088, + "step": 15590 + }, + { + "epoch": 0.6, + "learning_rate": 4.0035265452680174e-05, + "loss": 0.5826, + "step": 15591 + }, + { + "epoch": 0.6, + "learning_rate": 4.0034624262631446e-05, + "loss": 0.485, + "step": 15592 + }, + { + "epoch": 0.6, + "learning_rate": 4.003398307258272e-05, + "loss": 0.5255, + "step": 15593 + }, + { + "epoch": 0.6, + "learning_rate": 4.003334188253399e-05, + "loss": 0.5222, + "step": 15594 + }, + { + "epoch": 0.6, + "learning_rate": 4.0032700692485254e-05, + "loss": 0.6138, + "step": 15595 + }, + { + "epoch": 0.6, + "learning_rate": 4.0032059502436526e-05, + "loss": 0.5544, + "step": 15596 + }, + { + "epoch": 0.6, + "learning_rate": 4.003141831238779e-05, + "loss": 0.5675, + "step": 15597 + }, + { + "epoch": 0.6, + "learning_rate": 4.003077712233906e-05, + "loss": 0.5252, + "step": 15598 + }, + { + "epoch": 0.6, + "learning_rate": 4.003013593229033e-05, + "loss": 0.5178, + "step": 15599 + }, + { + "epoch": 0.6, + "learning_rate": 4.00294947422416e-05, + "loss": 0.4728, + "step": 15600 + }, + { + "epoch": 0.6, + "learning_rate": 4.002885355219287e-05, + "loss": 0.5602, + "step": 15601 + }, + { + "epoch": 0.6, + "learning_rate": 4.002821236214414e-05, + "loss": 0.5545, + "step": 15602 + }, + { + "epoch": 0.6, + "learning_rate": 4.0027571172095414e-05, + "loss": 0.4982, + "step": 15603 + }, + { + "epoch": 0.6, + "learning_rate": 4.002692998204668e-05, + "loss": 0.6366, + "step": 15604 + }, + { + "epoch": 0.6, + "learning_rate": 4.002628879199795e-05, + "loss": 0.4863, + "step": 15605 + }, + { + "epoch": 0.6, + "learning_rate": 4.0025647601949216e-05, + "loss": 0.4425, + "step": 15606 + }, + { + "epoch": 0.6, + "learning_rate": 4.002500641190049e-05, + "loss": 0.4366, + "step": 15607 + }, + { + "epoch": 0.6, + "learning_rate": 4.002436522185176e-05, + "loss": 0.5722, + "step": 15608 + }, + { + "epoch": 0.6, + "learning_rate": 4.002372403180303e-05, + "loss": 0.7448, + "step": 15609 + }, + { + "epoch": 0.6, + "learning_rate": 4.00230828417543e-05, + "loss": 0.5743, + "step": 15610 + }, + { + "epoch": 0.6, + "learning_rate": 4.002244165170557e-05, + "loss": 0.6367, + "step": 15611 + }, + { + "epoch": 0.6, + "learning_rate": 4.002180046165684e-05, + "loss": 0.5411, + "step": 15612 + }, + { + "epoch": 0.6, + "learning_rate": 4.0021159271608104e-05, + "loss": 0.5364, + "step": 15613 + }, + { + "epoch": 0.6, + "learning_rate": 4.0020518081559376e-05, + "loss": 0.5877, + "step": 15614 + }, + { + "epoch": 0.6, + "learning_rate": 4.001987689151064e-05, + "loss": 0.4904, + "step": 15615 + }, + { + "epoch": 0.6, + "learning_rate": 4.001923570146191e-05, + "loss": 0.525, + "step": 15616 + }, + { + "epoch": 0.6, + "learning_rate": 4.0018594511413184e-05, + "loss": 0.5561, + "step": 15617 + }, + { + "epoch": 0.6, + "learning_rate": 4.0017953321364456e-05, + "loss": 0.6503, + "step": 15618 + }, + { + "epoch": 0.6, + "learning_rate": 4.001731213131573e-05, + "loss": 0.5426, + "step": 15619 + }, + { + "epoch": 0.6, + "learning_rate": 4.001667094126699e-05, + "loss": 0.5777, + "step": 15620 + }, + { + "epoch": 0.6, + "learning_rate": 4.0016029751218265e-05, + "loss": 0.5527, + "step": 15621 + }, + { + "epoch": 0.6, + "learning_rate": 4.001538856116953e-05, + "loss": 0.6713, + "step": 15622 + }, + { + "epoch": 0.6, + "learning_rate": 4.00147473711208e-05, + "loss": 0.5215, + "step": 15623 + }, + { + "epoch": 0.6, + "learning_rate": 4.001410618107207e-05, + "loss": 0.4548, + "step": 15624 + }, + { + "epoch": 0.6, + "learning_rate": 4.0013464991023345e-05, + "loss": 0.4662, + "step": 15625 + }, + { + "epoch": 0.6, + "learning_rate": 4.001282380097461e-05, + "loss": 0.6361, + "step": 15626 + }, + { + "epoch": 0.6, + "learning_rate": 4.001218261092588e-05, + "loss": 0.5589, + "step": 15627 + }, + { + "epoch": 0.6, + "learning_rate": 4.001154142087715e-05, + "loss": 0.4314, + "step": 15628 + }, + { + "epoch": 0.6, + "learning_rate": 4.001090023082842e-05, + "loss": 0.5475, + "step": 15629 + }, + { + "epoch": 0.6, + "learning_rate": 4.001025904077969e-05, + "loss": 0.595, + "step": 15630 + }, + { + "epoch": 0.6, + "learning_rate": 4.0009617850730955e-05, + "loss": 0.5569, + "step": 15631 + }, + { + "epoch": 0.6, + "learning_rate": 4.0008976660682226e-05, + "loss": 0.4966, + "step": 15632 + }, + { + "epoch": 0.6, + "learning_rate": 4.00083354706335e-05, + "loss": 0.5188, + "step": 15633 + }, + { + "epoch": 0.6, + "learning_rate": 4.000769428058477e-05, + "loss": 0.5159, + "step": 15634 + }, + { + "epoch": 0.6, + "learning_rate": 4.0007053090536035e-05, + "loss": 0.5211, + "step": 15635 + }, + { + "epoch": 0.6, + "learning_rate": 4.0006411900487306e-05, + "loss": 0.5248, + "step": 15636 + }, + { + "epoch": 0.6, + "learning_rate": 4.000577071043858e-05, + "loss": 0.6201, + "step": 15637 + }, + { + "epoch": 0.6, + "learning_rate": 4.000512952038984e-05, + "loss": 0.4359, + "step": 15638 + }, + { + "epoch": 0.6, + "learning_rate": 4.0004488330341115e-05, + "loss": 0.5707, + "step": 15639 + }, + { + "epoch": 0.6, + "learning_rate": 4.000384714029238e-05, + "loss": 0.5188, + "step": 15640 + }, + { + "epoch": 0.6, + "learning_rate": 4.000320595024366e-05, + "loss": 0.5049, + "step": 15641 + }, + { + "epoch": 0.6, + "learning_rate": 4.000256476019492e-05, + "loss": 0.557, + "step": 15642 + }, + { + "epoch": 0.6, + "learning_rate": 4.0001923570146195e-05, + "loss": 0.5356, + "step": 15643 + }, + { + "epoch": 0.6, + "learning_rate": 4.000128238009746e-05, + "loss": 0.4994, + "step": 15644 + }, + { + "epoch": 0.6, + "learning_rate": 4.000064119004873e-05, + "loss": 0.5105, + "step": 15645 + }, + { + "epoch": 0.6, + "learning_rate": 4e-05, + "loss": 0.5103, + "step": 15646 + }, + { + "epoch": 0.6, + "learning_rate": 3.999935880995127e-05, + "loss": 0.482, + "step": 15647 + }, + { + "epoch": 0.6, + "learning_rate": 3.999871761990254e-05, + "loss": 0.4868, + "step": 15648 + }, + { + "epoch": 0.6, + "learning_rate": 3.999807642985381e-05, + "loss": 0.5805, + "step": 15649 + }, + { + "epoch": 0.6, + "learning_rate": 3.999743523980508e-05, + "loss": 0.5592, + "step": 15650 + }, + { + "epoch": 0.6, + "learning_rate": 3.999679404975635e-05, + "loss": 0.5865, + "step": 15651 + }, + { + "epoch": 0.6, + "learning_rate": 3.999615285970762e-05, + "loss": 0.552, + "step": 15652 + }, + { + "epoch": 0.6, + "learning_rate": 3.9995511669658885e-05, + "loss": 0.4742, + "step": 15653 + }, + { + "epoch": 0.6, + "learning_rate": 3.999487047961016e-05, + "loss": 0.4759, + "step": 15654 + }, + { + "epoch": 0.6, + "learning_rate": 3.999422928956143e-05, + "loss": 0.5934, + "step": 15655 + }, + { + "epoch": 0.6, + "learning_rate": 3.999358809951269e-05, + "loss": 0.5444, + "step": 15656 + }, + { + "epoch": 0.6, + "learning_rate": 3.999294690946397e-05, + "loss": 0.5772, + "step": 15657 + }, + { + "epoch": 0.6, + "learning_rate": 3.999230571941524e-05, + "loss": 0.5811, + "step": 15658 + }, + { + "epoch": 0.6, + "learning_rate": 3.999166452936651e-05, + "loss": 0.5992, + "step": 15659 + }, + { + "epoch": 0.6, + "learning_rate": 3.9991023339317773e-05, + "loss": 0.5646, + "step": 15660 + }, + { + "epoch": 0.6, + "learning_rate": 3.9990382149269045e-05, + "loss": 0.5729, + "step": 15661 + }, + { + "epoch": 0.6, + "learning_rate": 3.998974095922031e-05, + "loss": 0.6078, + "step": 15662 + }, + { + "epoch": 0.6, + "learning_rate": 3.998909976917158e-05, + "loss": 0.5875, + "step": 15663 + }, + { + "epoch": 0.6, + "learning_rate": 3.9988458579122854e-05, + "loss": 0.5394, + "step": 15664 + }, + { + "epoch": 0.6, + "learning_rate": 3.9987817389074125e-05, + "loss": 0.5237, + "step": 15665 + }, + { + "epoch": 0.6, + "learning_rate": 3.99871761990254e-05, + "loss": 0.4206, + "step": 15666 + }, + { + "epoch": 0.6, + "learning_rate": 3.998653500897666e-05, + "loss": 0.6087, + "step": 15667 + }, + { + "epoch": 0.6, + "learning_rate": 3.9985893818927934e-05, + "loss": 0.5526, + "step": 15668 + }, + { + "epoch": 0.6, + "learning_rate": 3.99852526288792e-05, + "loss": 0.5051, + "step": 15669 + }, + { + "epoch": 0.6, + "learning_rate": 3.998461143883047e-05, + "loss": 0.5628, + "step": 15670 + }, + { + "epoch": 0.6, + "learning_rate": 3.998397024878174e-05, + "loss": 0.5479, + "step": 15671 + }, + { + "epoch": 0.6, + "learning_rate": 3.998332905873301e-05, + "loss": 0.5571, + "step": 15672 + }, + { + "epoch": 0.6, + "learning_rate": 3.9982687868684285e-05, + "loss": 0.5293, + "step": 15673 + }, + { + "epoch": 0.6, + "learning_rate": 3.998204667863555e-05, + "loss": 0.6564, + "step": 15674 + }, + { + "epoch": 0.6, + "learning_rate": 3.998140548858682e-05, + "loss": 0.4948, + "step": 15675 + }, + { + "epoch": 0.6, + "learning_rate": 3.998076429853809e-05, + "loss": 0.603, + "step": 15676 + }, + { + "epoch": 0.6, + "learning_rate": 3.998012310848936e-05, + "loss": 0.5278, + "step": 15677 + }, + { + "epoch": 0.6, + "learning_rate": 3.9979481918440624e-05, + "loss": 0.5593, + "step": 15678 + }, + { + "epoch": 0.6, + "learning_rate": 3.9978840728391895e-05, + "loss": 0.4909, + "step": 15679 + }, + { + "epoch": 0.6, + "learning_rate": 3.997819953834317e-05, + "loss": 0.4827, + "step": 15680 + }, + { + "epoch": 0.6, + "learning_rate": 3.997755834829444e-05, + "loss": 0.552, + "step": 15681 + }, + { + "epoch": 0.6, + "learning_rate": 3.997691715824571e-05, + "loss": 0.5583, + "step": 15682 + }, + { + "epoch": 0.6, + "learning_rate": 3.9976275968196975e-05, + "loss": 0.4819, + "step": 15683 + }, + { + "epoch": 0.6, + "learning_rate": 3.997563477814825e-05, + "loss": 0.505, + "step": 15684 + }, + { + "epoch": 0.6, + "learning_rate": 3.997499358809951e-05, + "loss": 0.5778, + "step": 15685 + }, + { + "epoch": 0.6, + "learning_rate": 3.9974352398050784e-05, + "loss": 0.4921, + "step": 15686 + }, + { + "epoch": 0.6, + "learning_rate": 3.997371120800205e-05, + "loss": 0.4543, + "step": 15687 + }, + { + "epoch": 0.6, + "learning_rate": 3.997307001795332e-05, + "loss": 0.4985, + "step": 15688 + }, + { + "epoch": 0.6, + "learning_rate": 3.997242882790459e-05, + "loss": 0.5316, + "step": 15689 + }, + { + "epoch": 0.6, + "learning_rate": 3.9971787637855864e-05, + "loss": 0.5837, + "step": 15690 + }, + { + "epoch": 0.6, + "learning_rate": 3.9971146447807136e-05, + "loss": 0.5986, + "step": 15691 + }, + { + "epoch": 0.6, + "learning_rate": 3.99705052577584e-05, + "loss": 0.6676, + "step": 15692 + }, + { + "epoch": 0.6, + "learning_rate": 3.996986406770967e-05, + "loss": 0.5808, + "step": 15693 + }, + { + "epoch": 0.6, + "learning_rate": 3.996922287766094e-05, + "loss": 0.6726, + "step": 15694 + }, + { + "epoch": 0.6, + "learning_rate": 3.996858168761221e-05, + "loss": 0.4712, + "step": 15695 + }, + { + "epoch": 0.6, + "learning_rate": 3.996794049756348e-05, + "loss": 0.5104, + "step": 15696 + }, + { + "epoch": 0.6, + "learning_rate": 3.996729930751475e-05, + "loss": 0.4976, + "step": 15697 + }, + { + "epoch": 0.6, + "learning_rate": 3.9966658117466024e-05, + "loss": 0.4346, + "step": 15698 + }, + { + "epoch": 0.6, + "learning_rate": 3.996601692741729e-05, + "loss": 0.4833, + "step": 15699 + }, + { + "epoch": 0.6, + "learning_rate": 3.996537573736856e-05, + "loss": 0.6434, + "step": 15700 + }, + { + "epoch": 0.6, + "learning_rate": 3.9964734547319826e-05, + "loss": 0.5027, + "step": 15701 + }, + { + "epoch": 0.6, + "learning_rate": 3.99640933572711e-05, + "loss": 0.5719, + "step": 15702 + }, + { + "epoch": 0.6, + "learning_rate": 3.996345216722236e-05, + "loss": 0.5875, + "step": 15703 + }, + { + "epoch": 0.6, + "learning_rate": 3.9962810977173634e-05, + "loss": 0.5487, + "step": 15704 + }, + { + "epoch": 0.6, + "learning_rate": 3.9962169787124906e-05, + "loss": 0.5008, + "step": 15705 + }, + { + "epoch": 0.6, + "learning_rate": 3.996152859707618e-05, + "loss": 0.5956, + "step": 15706 + }, + { + "epoch": 0.6, + "learning_rate": 3.996088740702745e-05, + "loss": 0.5387, + "step": 15707 + }, + { + "epoch": 0.6, + "learning_rate": 3.9960246216978714e-05, + "loss": 0.5453, + "step": 15708 + }, + { + "epoch": 0.6, + "learning_rate": 3.9959605026929986e-05, + "loss": 0.5758, + "step": 15709 + }, + { + "epoch": 0.6, + "learning_rate": 3.995896383688125e-05, + "loss": 0.5301, + "step": 15710 + }, + { + "epoch": 0.6, + "learning_rate": 3.995832264683252e-05, + "loss": 0.5613, + "step": 15711 + }, + { + "epoch": 0.6, + "learning_rate": 3.9957681456783794e-05, + "loss": 0.5017, + "step": 15712 + }, + { + "epoch": 0.6, + "learning_rate": 3.9957040266735066e-05, + "loss": 0.5683, + "step": 15713 + }, + { + "epoch": 0.6, + "learning_rate": 3.995639907668633e-05, + "loss": 0.5616, + "step": 15714 + }, + { + "epoch": 0.6, + "learning_rate": 3.99557578866376e-05, + "loss": 0.57, + "step": 15715 + }, + { + "epoch": 0.6, + "learning_rate": 3.9955116696588874e-05, + "loss": 0.4999, + "step": 15716 + }, + { + "epoch": 0.6, + "learning_rate": 3.995447550654014e-05, + "loss": 0.4768, + "step": 15717 + }, + { + "epoch": 0.6, + "learning_rate": 3.995383431649141e-05, + "loss": 0.6215, + "step": 15718 + }, + { + "epoch": 0.6, + "learning_rate": 3.9953193126442676e-05, + "loss": 0.586, + "step": 15719 + }, + { + "epoch": 0.6, + "learning_rate": 3.995255193639395e-05, + "loss": 0.5892, + "step": 15720 + }, + { + "epoch": 0.6, + "learning_rate": 3.995191074634522e-05, + "loss": 0.5566, + "step": 15721 + }, + { + "epoch": 0.6, + "learning_rate": 3.995126955629649e-05, + "loss": 0.5405, + "step": 15722 + }, + { + "epoch": 0.6, + "learning_rate": 3.9950628366247756e-05, + "loss": 0.5961, + "step": 15723 + }, + { + "epoch": 0.6, + "learning_rate": 3.994998717619903e-05, + "loss": 0.5172, + "step": 15724 + }, + { + "epoch": 0.6, + "learning_rate": 3.99493459861503e-05, + "loss": 0.5809, + "step": 15725 + }, + { + "epoch": 0.6, + "learning_rate": 3.9948704796101564e-05, + "loss": 0.5189, + "step": 15726 + }, + { + "epoch": 0.6, + "learning_rate": 3.9948063606052836e-05, + "loss": 0.4386, + "step": 15727 + }, + { + "epoch": 0.6, + "learning_rate": 3.99474224160041e-05, + "loss": 0.5005, + "step": 15728 + }, + { + "epoch": 0.6, + "learning_rate": 3.994678122595538e-05, + "loss": 0.4679, + "step": 15729 + }, + { + "epoch": 0.6, + "learning_rate": 3.9946140035906645e-05, + "loss": 0.4746, + "step": 15730 + }, + { + "epoch": 0.6, + "learning_rate": 3.9945498845857916e-05, + "loss": 0.5865, + "step": 15731 + }, + { + "epoch": 0.6, + "learning_rate": 3.994485765580918e-05, + "loss": 0.5319, + "step": 15732 + }, + { + "epoch": 0.6, + "learning_rate": 3.994421646576045e-05, + "loss": 0.4829, + "step": 15733 + }, + { + "epoch": 0.6, + "learning_rate": 3.9943575275711725e-05, + "loss": 0.581, + "step": 15734 + }, + { + "epoch": 0.6, + "learning_rate": 3.994293408566299e-05, + "loss": 0.4764, + "step": 15735 + }, + { + "epoch": 0.6, + "learning_rate": 3.994229289561426e-05, + "loss": 0.4838, + "step": 15736 + }, + { + "epoch": 0.61, + "learning_rate": 3.994165170556553e-05, + "loss": 0.4944, + "step": 15737 + }, + { + "epoch": 0.61, + "learning_rate": 3.9941010515516805e-05, + "loss": 0.4268, + "step": 15738 + }, + { + "epoch": 0.61, + "learning_rate": 3.994036932546807e-05, + "loss": 0.4839, + "step": 15739 + }, + { + "epoch": 0.61, + "learning_rate": 3.993972813541934e-05, + "loss": 0.5994, + "step": 15740 + }, + { + "epoch": 0.61, + "learning_rate": 3.9939086945370606e-05, + "loss": 0.6366, + "step": 15741 + }, + { + "epoch": 0.61, + "learning_rate": 3.993844575532188e-05, + "loss": 0.5584, + "step": 15742 + }, + { + "epoch": 0.61, + "learning_rate": 3.993780456527315e-05, + "loss": 0.5378, + "step": 15743 + }, + { + "epoch": 0.61, + "learning_rate": 3.9937163375224415e-05, + "loss": 0.5845, + "step": 15744 + }, + { + "epoch": 0.61, + "learning_rate": 3.993652218517569e-05, + "loss": 0.5427, + "step": 15745 + }, + { + "epoch": 0.61, + "learning_rate": 3.993588099512696e-05, + "loss": 0.494, + "step": 15746 + }, + { + "epoch": 0.61, + "learning_rate": 3.993523980507823e-05, + "loss": 0.5372, + "step": 15747 + }, + { + "epoch": 0.61, + "learning_rate": 3.9934598615029495e-05, + "loss": 0.562, + "step": 15748 + }, + { + "epoch": 0.61, + "learning_rate": 3.9933957424980766e-05, + "loss": 0.537, + "step": 15749 + }, + { + "epoch": 0.61, + "learning_rate": 3.993331623493203e-05, + "loss": 0.6075, + "step": 15750 + }, + { + "epoch": 0.61, + "learning_rate": 3.99326750448833e-05, + "loss": 0.6007, + "step": 15751 + }, + { + "epoch": 0.61, + "learning_rate": 3.9932033854834575e-05, + "loss": 0.492, + "step": 15752 + }, + { + "epoch": 0.61, + "learning_rate": 3.9931392664785847e-05, + "loss": 0.5396, + "step": 15753 + }, + { + "epoch": 0.61, + "learning_rate": 3.993075147473712e-05, + "loss": 0.5647, + "step": 15754 + }, + { + "epoch": 0.61, + "learning_rate": 3.993011028468838e-05, + "loss": 0.5203, + "step": 15755 + }, + { + "epoch": 0.61, + "learning_rate": 3.9929469094639655e-05, + "loss": 0.5191, + "step": 15756 + }, + { + "epoch": 0.61, + "learning_rate": 3.992882790459092e-05, + "loss": 0.5883, + "step": 15757 + }, + { + "epoch": 0.61, + "learning_rate": 3.992818671454219e-05, + "loss": 0.481, + "step": 15758 + }, + { + "epoch": 0.61, + "learning_rate": 3.9927545524493457e-05, + "loss": 0.5379, + "step": 15759 + }, + { + "epoch": 0.61, + "learning_rate": 3.992690433444473e-05, + "loss": 0.4999, + "step": 15760 + }, + { + "epoch": 0.61, + "learning_rate": 3.9926263144396e-05, + "loss": 0.4607, + "step": 15761 + }, + { + "epoch": 0.61, + "learning_rate": 3.992562195434727e-05, + "loss": 0.4995, + "step": 15762 + }, + { + "epoch": 0.61, + "learning_rate": 3.9924980764298543e-05, + "loss": 0.5553, + "step": 15763 + }, + { + "epoch": 0.61, + "learning_rate": 3.992433957424981e-05, + "loss": 0.4747, + "step": 15764 + }, + { + "epoch": 0.61, + "learning_rate": 3.992369838420108e-05, + "loss": 0.5859, + "step": 15765 + }, + { + "epoch": 0.61, + "learning_rate": 3.9923057194152345e-05, + "loss": 0.4973, + "step": 15766 + }, + { + "epoch": 0.61, + "learning_rate": 3.992241600410362e-05, + "loss": 0.4853, + "step": 15767 + }, + { + "epoch": 0.61, + "learning_rate": 3.992177481405489e-05, + "loss": 0.4861, + "step": 15768 + }, + { + "epoch": 0.61, + "learning_rate": 3.992113362400616e-05, + "loss": 0.5779, + "step": 15769 + }, + { + "epoch": 0.61, + "learning_rate": 3.992049243395743e-05, + "loss": 0.4434, + "step": 15770 + }, + { + "epoch": 0.61, + "learning_rate": 3.99198512439087e-05, + "loss": 0.5726, + "step": 15771 + }, + { + "epoch": 0.61, + "learning_rate": 3.991921005385997e-05, + "loss": 0.5809, + "step": 15772 + }, + { + "epoch": 0.61, + "learning_rate": 3.9918568863811233e-05, + "loss": 0.6307, + "step": 15773 + }, + { + "epoch": 0.61, + "learning_rate": 3.9917927673762505e-05, + "loss": 0.4892, + "step": 15774 + }, + { + "epoch": 0.61, + "learning_rate": 3.991728648371377e-05, + "loss": 0.5573, + "step": 15775 + }, + { + "epoch": 0.61, + "learning_rate": 3.991664529366504e-05, + "loss": 0.5609, + "step": 15776 + }, + { + "epoch": 0.61, + "learning_rate": 3.9916004103616314e-05, + "loss": 0.538, + "step": 15777 + }, + { + "epoch": 0.61, + "learning_rate": 3.9915362913567585e-05, + "loss": 0.4906, + "step": 15778 + }, + { + "epoch": 0.61, + "learning_rate": 3.991472172351886e-05, + "loss": 0.5829, + "step": 15779 + }, + { + "epoch": 0.61, + "learning_rate": 3.991408053347012e-05, + "loss": 0.5967, + "step": 15780 + }, + { + "epoch": 0.61, + "learning_rate": 3.9913439343421394e-05, + "loss": 0.533, + "step": 15781 + }, + { + "epoch": 0.61, + "learning_rate": 3.991279815337266e-05, + "loss": 0.5251, + "step": 15782 + }, + { + "epoch": 0.61, + "learning_rate": 3.991215696332393e-05, + "loss": 0.5219, + "step": 15783 + }, + { + "epoch": 0.61, + "learning_rate": 3.99115157732752e-05, + "loss": 0.582, + "step": 15784 + }, + { + "epoch": 0.61, + "learning_rate": 3.9910874583226474e-05, + "loss": 0.4553, + "step": 15785 + }, + { + "epoch": 0.61, + "learning_rate": 3.991023339317774e-05, + "loss": 0.4588, + "step": 15786 + }, + { + "epoch": 0.61, + "learning_rate": 3.990959220312901e-05, + "loss": 0.4342, + "step": 15787 + }, + { + "epoch": 0.61, + "learning_rate": 3.990895101308028e-05, + "loss": 0.4874, + "step": 15788 + }, + { + "epoch": 0.61, + "learning_rate": 3.990830982303155e-05, + "loss": 0.4693, + "step": 15789 + }, + { + "epoch": 0.61, + "learning_rate": 3.990766863298282e-05, + "loss": 0.4482, + "step": 15790 + }, + { + "epoch": 0.61, + "learning_rate": 3.9907027442934084e-05, + "loss": 0.5264, + "step": 15791 + }, + { + "epoch": 0.61, + "learning_rate": 3.9906386252885355e-05, + "loss": 0.5413, + "step": 15792 + }, + { + "epoch": 0.61, + "learning_rate": 3.990574506283663e-05, + "loss": 0.4613, + "step": 15793 + }, + { + "epoch": 0.61, + "learning_rate": 3.99051038727879e-05, + "loss": 0.4602, + "step": 15794 + }, + { + "epoch": 0.61, + "learning_rate": 3.9904462682739164e-05, + "loss": 0.5523, + "step": 15795 + }, + { + "epoch": 0.61, + "learning_rate": 3.9903821492690436e-05, + "loss": 0.6029, + "step": 15796 + }, + { + "epoch": 0.61, + "learning_rate": 3.990318030264171e-05, + "loss": 0.5122, + "step": 15797 + }, + { + "epoch": 0.61, + "learning_rate": 3.990253911259297e-05, + "loss": 0.5968, + "step": 15798 + }, + { + "epoch": 0.61, + "learning_rate": 3.9901897922544244e-05, + "loss": 0.4285, + "step": 15799 + }, + { + "epoch": 0.61, + "learning_rate": 3.9901256732495516e-05, + "loss": 0.4804, + "step": 15800 + }, + { + "epoch": 0.61, + "learning_rate": 3.990061554244679e-05, + "loss": 0.5248, + "step": 15801 + }, + { + "epoch": 0.61, + "learning_rate": 3.989997435239805e-05, + "loss": 0.4707, + "step": 15802 + }, + { + "epoch": 0.61, + "learning_rate": 3.9899333162349324e-05, + "loss": 0.6384, + "step": 15803 + }, + { + "epoch": 0.61, + "learning_rate": 3.989869197230059e-05, + "loss": 0.4831, + "step": 15804 + }, + { + "epoch": 0.61, + "learning_rate": 3.989805078225186e-05, + "loss": 0.6374, + "step": 15805 + }, + { + "epoch": 0.61, + "learning_rate": 3.989740959220313e-05, + "loss": 0.472, + "step": 15806 + }, + { + "epoch": 0.61, + "learning_rate": 3.98967684021544e-05, + "loss": 0.5168, + "step": 15807 + }, + { + "epoch": 0.61, + "learning_rate": 3.989612721210567e-05, + "loss": 0.4017, + "step": 15808 + }, + { + "epoch": 0.61, + "learning_rate": 3.989548602205694e-05, + "loss": 0.6168, + "step": 15809 + }, + { + "epoch": 0.61, + "learning_rate": 3.989484483200821e-05, + "loss": 0.6222, + "step": 15810 + }, + { + "epoch": 0.61, + "learning_rate": 3.989420364195948e-05, + "loss": 0.4975, + "step": 15811 + }, + { + "epoch": 0.61, + "learning_rate": 3.989356245191075e-05, + "loss": 0.5057, + "step": 15812 + }, + { + "epoch": 0.61, + "learning_rate": 3.9892921261862014e-05, + "loss": 0.4397, + "step": 15813 + }, + { + "epoch": 0.61, + "learning_rate": 3.9892280071813286e-05, + "loss": 0.607, + "step": 15814 + }, + { + "epoch": 0.61, + "learning_rate": 3.989163888176456e-05, + "loss": 0.4272, + "step": 15815 + }, + { + "epoch": 0.61, + "learning_rate": 3.989099769171582e-05, + "loss": 0.6249, + "step": 15816 + }, + { + "epoch": 0.61, + "learning_rate": 3.98903565016671e-05, + "loss": 0.5126, + "step": 15817 + }, + { + "epoch": 0.61, + "learning_rate": 3.9889715311618366e-05, + "loss": 0.6032, + "step": 15818 + }, + { + "epoch": 0.61, + "learning_rate": 3.988907412156964e-05, + "loss": 0.5324, + "step": 15819 + }, + { + "epoch": 0.61, + "learning_rate": 3.98884329315209e-05, + "loss": 0.6293, + "step": 15820 + }, + { + "epoch": 0.61, + "learning_rate": 3.9887791741472174e-05, + "loss": 0.5117, + "step": 15821 + }, + { + "epoch": 0.61, + "learning_rate": 3.988715055142344e-05, + "loss": 0.4846, + "step": 15822 + }, + { + "epoch": 0.61, + "learning_rate": 3.988650936137471e-05, + "loss": 0.5044, + "step": 15823 + }, + { + "epoch": 0.61, + "learning_rate": 3.988586817132598e-05, + "loss": 0.5156, + "step": 15824 + }, + { + "epoch": 0.61, + "learning_rate": 3.9885226981277254e-05, + "loss": 0.5213, + "step": 15825 + }, + { + "epoch": 0.61, + "learning_rate": 3.9884585791228526e-05, + "loss": 0.5515, + "step": 15826 + }, + { + "epoch": 0.61, + "learning_rate": 3.988394460117979e-05, + "loss": 0.5234, + "step": 15827 + }, + { + "epoch": 0.61, + "learning_rate": 3.988330341113106e-05, + "loss": 0.5406, + "step": 15828 + }, + { + "epoch": 0.61, + "learning_rate": 3.988266222108233e-05, + "loss": 0.4051, + "step": 15829 + }, + { + "epoch": 0.61, + "learning_rate": 3.98820210310336e-05, + "loss": 0.5769, + "step": 15830 + }, + { + "epoch": 0.61, + "learning_rate": 3.9881379840984864e-05, + "loss": 0.6034, + "step": 15831 + }, + { + "epoch": 0.61, + "learning_rate": 3.9880738650936136e-05, + "loss": 0.4435, + "step": 15832 + }, + { + "epoch": 0.61, + "learning_rate": 3.9880097460887414e-05, + "loss": 0.4504, + "step": 15833 + }, + { + "epoch": 0.61, + "learning_rate": 3.987945627083868e-05, + "loss": 0.4555, + "step": 15834 + }, + { + "epoch": 0.61, + "learning_rate": 3.987881508078995e-05, + "loss": 0.4728, + "step": 15835 + }, + { + "epoch": 0.61, + "learning_rate": 3.9878173890741216e-05, + "loss": 0.5402, + "step": 15836 + }, + { + "epoch": 0.61, + "learning_rate": 3.987753270069249e-05, + "loss": 0.6027, + "step": 15837 + }, + { + "epoch": 0.61, + "learning_rate": 3.987689151064375e-05, + "loss": 0.5156, + "step": 15838 + }, + { + "epoch": 0.61, + "learning_rate": 3.9876250320595024e-05, + "loss": 0.4694, + "step": 15839 + }, + { + "epoch": 0.61, + "learning_rate": 3.9875609130546296e-05, + "loss": 0.5262, + "step": 15840 + }, + { + "epoch": 0.61, + "learning_rate": 3.987496794049757e-05, + "loss": 0.5187, + "step": 15841 + }, + { + "epoch": 0.61, + "learning_rate": 3.987432675044884e-05, + "loss": 0.4706, + "step": 15842 + }, + { + "epoch": 0.61, + "learning_rate": 3.9873685560400105e-05, + "loss": 0.5403, + "step": 15843 + }, + { + "epoch": 0.61, + "learning_rate": 3.9873044370351376e-05, + "loss": 0.6269, + "step": 15844 + }, + { + "epoch": 0.61, + "learning_rate": 3.987240318030264e-05, + "loss": 0.4685, + "step": 15845 + }, + { + "epoch": 0.61, + "learning_rate": 3.987176199025391e-05, + "loss": 0.5107, + "step": 15846 + }, + { + "epoch": 0.61, + "learning_rate": 3.987112080020518e-05, + "loss": 0.6017, + "step": 15847 + }, + { + "epoch": 0.61, + "learning_rate": 3.987047961015645e-05, + "loss": 0.4327, + "step": 15848 + }, + { + "epoch": 0.61, + "learning_rate": 3.986983842010772e-05, + "loss": 0.4824, + "step": 15849 + }, + { + "epoch": 0.61, + "learning_rate": 3.986919723005899e-05, + "loss": 0.5529, + "step": 15850 + }, + { + "epoch": 0.61, + "learning_rate": 3.9868556040010265e-05, + "loss": 0.53, + "step": 15851 + }, + { + "epoch": 0.61, + "learning_rate": 3.986791484996153e-05, + "loss": 0.4524, + "step": 15852 + }, + { + "epoch": 0.61, + "learning_rate": 3.98672736599128e-05, + "loss": 0.5877, + "step": 15853 + }, + { + "epoch": 0.61, + "learning_rate": 3.9866632469864066e-05, + "loss": 0.3788, + "step": 15854 + }, + { + "epoch": 0.61, + "learning_rate": 3.986599127981534e-05, + "loss": 0.6086, + "step": 15855 + }, + { + "epoch": 0.61, + "learning_rate": 3.986535008976661e-05, + "loss": 0.5322, + "step": 15856 + }, + { + "epoch": 0.61, + "learning_rate": 3.986470889971788e-05, + "loss": 0.5424, + "step": 15857 + }, + { + "epoch": 0.61, + "learning_rate": 3.986406770966915e-05, + "loss": 0.5234, + "step": 15858 + }, + { + "epoch": 0.61, + "learning_rate": 3.986342651962042e-05, + "loss": 0.5429, + "step": 15859 + }, + { + "epoch": 0.61, + "learning_rate": 3.986278532957169e-05, + "loss": 0.6678, + "step": 15860 + }, + { + "epoch": 0.61, + "learning_rate": 3.9862144139522955e-05, + "loss": 0.5123, + "step": 15861 + }, + { + "epoch": 0.61, + "learning_rate": 3.9861502949474227e-05, + "loss": 0.469, + "step": 15862 + }, + { + "epoch": 0.61, + "learning_rate": 3.986086175942549e-05, + "loss": 0.5385, + "step": 15863 + }, + { + "epoch": 0.61, + "learning_rate": 3.986022056937676e-05, + "loss": 0.5356, + "step": 15864 + }, + { + "epoch": 0.61, + "learning_rate": 3.9859579379328035e-05, + "loss": 0.5674, + "step": 15865 + }, + { + "epoch": 0.61, + "learning_rate": 3.9858938189279307e-05, + "loss": 0.4608, + "step": 15866 + }, + { + "epoch": 0.61, + "learning_rate": 3.985829699923058e-05, + "loss": 0.5875, + "step": 15867 + }, + { + "epoch": 0.61, + "learning_rate": 3.985765580918184e-05, + "loss": 0.5312, + "step": 15868 + }, + { + "epoch": 0.61, + "learning_rate": 3.9857014619133115e-05, + "loss": 0.3816, + "step": 15869 + }, + { + "epoch": 0.61, + "learning_rate": 3.985637342908438e-05, + "loss": 0.693, + "step": 15870 + }, + { + "epoch": 0.61, + "learning_rate": 3.985573223903565e-05, + "loss": 0.5436, + "step": 15871 + }, + { + "epoch": 0.61, + "learning_rate": 3.985509104898692e-05, + "loss": 0.4506, + "step": 15872 + }, + { + "epoch": 0.61, + "learning_rate": 3.9854449858938195e-05, + "loss": 0.6469, + "step": 15873 + }, + { + "epoch": 0.61, + "learning_rate": 3.985380866888946e-05, + "loss": 0.5158, + "step": 15874 + }, + { + "epoch": 0.61, + "learning_rate": 3.985316747884073e-05, + "loss": 0.462, + "step": 15875 + }, + { + "epoch": 0.61, + "learning_rate": 3.9852526288792003e-05, + "loss": 0.5771, + "step": 15876 + }, + { + "epoch": 0.61, + "learning_rate": 3.985188509874327e-05, + "loss": 0.5101, + "step": 15877 + }, + { + "epoch": 0.61, + "learning_rate": 3.985124390869454e-05, + "loss": 0.6255, + "step": 15878 + }, + { + "epoch": 0.61, + "learning_rate": 3.9850602718645805e-05, + "loss": 0.7225, + "step": 15879 + }, + { + "epoch": 0.61, + "learning_rate": 3.984996152859708e-05, + "loss": 0.5573, + "step": 15880 + }, + { + "epoch": 0.61, + "learning_rate": 3.984932033854835e-05, + "loss": 0.41, + "step": 15881 + }, + { + "epoch": 0.61, + "learning_rate": 3.984867914849962e-05, + "loss": 0.4673, + "step": 15882 + }, + { + "epoch": 0.61, + "learning_rate": 3.9848037958450885e-05, + "loss": 0.5427, + "step": 15883 + }, + { + "epoch": 0.61, + "learning_rate": 3.984739676840216e-05, + "loss": 0.5201, + "step": 15884 + }, + { + "epoch": 0.61, + "learning_rate": 3.984675557835343e-05, + "loss": 0.6204, + "step": 15885 + }, + { + "epoch": 0.61, + "learning_rate": 3.9846114388304693e-05, + "loss": 0.5514, + "step": 15886 + }, + { + "epoch": 0.61, + "learning_rate": 3.9845473198255965e-05, + "loss": 0.5801, + "step": 15887 + }, + { + "epoch": 0.61, + "learning_rate": 3.984483200820723e-05, + "loss": 0.4679, + "step": 15888 + }, + { + "epoch": 0.61, + "learning_rate": 3.984419081815851e-05, + "loss": 0.557, + "step": 15889 + }, + { + "epoch": 0.61, + "learning_rate": 3.9843549628109774e-05, + "loss": 0.507, + "step": 15890 + }, + { + "epoch": 0.61, + "learning_rate": 3.9842908438061045e-05, + "loss": 0.5249, + "step": 15891 + }, + { + "epoch": 0.61, + "learning_rate": 3.984226724801231e-05, + "loss": 0.6775, + "step": 15892 + }, + { + "epoch": 0.61, + "learning_rate": 3.984162605796358e-05, + "loss": 0.5791, + "step": 15893 + }, + { + "epoch": 0.61, + "learning_rate": 3.9840984867914854e-05, + "loss": 0.557, + "step": 15894 + }, + { + "epoch": 0.61, + "learning_rate": 3.984034367786612e-05, + "loss": 0.4402, + "step": 15895 + }, + { + "epoch": 0.61, + "learning_rate": 3.983970248781739e-05, + "loss": 0.6432, + "step": 15896 + }, + { + "epoch": 0.61, + "learning_rate": 3.983906129776866e-05, + "loss": 0.6165, + "step": 15897 + }, + { + "epoch": 0.61, + "learning_rate": 3.9838420107719934e-05, + "loss": 0.5449, + "step": 15898 + }, + { + "epoch": 0.61, + "learning_rate": 3.98377789176712e-05, + "loss": 0.5013, + "step": 15899 + }, + { + "epoch": 0.61, + "learning_rate": 3.983713772762247e-05, + "loss": 0.681, + "step": 15900 + }, + { + "epoch": 0.61, + "learning_rate": 3.9836496537573735e-05, + "loss": 0.4646, + "step": 15901 + }, + { + "epoch": 0.61, + "learning_rate": 3.983585534752501e-05, + "loss": 0.5276, + "step": 15902 + }, + { + "epoch": 0.61, + "learning_rate": 3.983521415747628e-05, + "loss": 0.5787, + "step": 15903 + }, + { + "epoch": 0.61, + "learning_rate": 3.9834572967427544e-05, + "loss": 0.5204, + "step": 15904 + }, + { + "epoch": 0.61, + "learning_rate": 3.983393177737882e-05, + "loss": 0.5165, + "step": 15905 + }, + { + "epoch": 0.61, + "learning_rate": 3.983329058733009e-05, + "loss": 0.6845, + "step": 15906 + }, + { + "epoch": 0.61, + "learning_rate": 3.983264939728136e-05, + "loss": 0.6337, + "step": 15907 + }, + { + "epoch": 0.61, + "learning_rate": 3.9832008207232624e-05, + "loss": 0.5379, + "step": 15908 + }, + { + "epoch": 0.61, + "learning_rate": 3.9831367017183896e-05, + "loss": 0.4503, + "step": 15909 + }, + { + "epoch": 0.61, + "learning_rate": 3.983072582713516e-05, + "loss": 0.5491, + "step": 15910 + }, + { + "epoch": 0.61, + "learning_rate": 3.983008463708643e-05, + "loss": 0.5291, + "step": 15911 + }, + { + "epoch": 0.61, + "learning_rate": 3.9829443447037704e-05, + "loss": 0.4798, + "step": 15912 + }, + { + "epoch": 0.61, + "learning_rate": 3.9828802256988976e-05, + "loss": 0.4994, + "step": 15913 + }, + { + "epoch": 0.61, + "learning_rate": 3.982816106694025e-05, + "loss": 0.6105, + "step": 15914 + }, + { + "epoch": 0.61, + "learning_rate": 3.982751987689151e-05, + "loss": 0.5165, + "step": 15915 + }, + { + "epoch": 0.61, + "learning_rate": 3.9826878686842784e-05, + "loss": 0.639, + "step": 15916 + }, + { + "epoch": 0.61, + "learning_rate": 3.982623749679405e-05, + "loss": 0.5743, + "step": 15917 + }, + { + "epoch": 0.61, + "learning_rate": 3.982559630674532e-05, + "loss": 0.4928, + "step": 15918 + }, + { + "epoch": 0.61, + "learning_rate": 3.9824955116696586e-05, + "loss": 0.52, + "step": 15919 + }, + { + "epoch": 0.61, + "learning_rate": 3.982431392664786e-05, + "loss": 0.5951, + "step": 15920 + }, + { + "epoch": 0.61, + "learning_rate": 3.9823672736599136e-05, + "loss": 0.5252, + "step": 15921 + }, + { + "epoch": 0.61, + "learning_rate": 3.98230315465504e-05, + "loss": 0.4806, + "step": 15922 + }, + { + "epoch": 0.61, + "learning_rate": 3.982239035650167e-05, + "loss": 0.4854, + "step": 15923 + }, + { + "epoch": 0.61, + "learning_rate": 3.982174916645294e-05, + "loss": 0.5439, + "step": 15924 + }, + { + "epoch": 0.61, + "learning_rate": 3.982110797640421e-05, + "loss": 0.5408, + "step": 15925 + }, + { + "epoch": 0.61, + "learning_rate": 3.9820466786355474e-05, + "loss": 0.6327, + "step": 15926 + }, + { + "epoch": 0.61, + "learning_rate": 3.9819825596306746e-05, + "loss": 0.5511, + "step": 15927 + }, + { + "epoch": 0.61, + "learning_rate": 3.981918440625802e-05, + "loss": 0.4828, + "step": 15928 + }, + { + "epoch": 0.61, + "learning_rate": 3.981854321620929e-05, + "loss": 0.4881, + "step": 15929 + }, + { + "epoch": 0.61, + "learning_rate": 3.981790202616056e-05, + "loss": 0.5646, + "step": 15930 + }, + { + "epoch": 0.61, + "learning_rate": 3.9817260836111826e-05, + "loss": 0.5229, + "step": 15931 + }, + { + "epoch": 0.61, + "learning_rate": 3.98166196460631e-05, + "loss": 0.5705, + "step": 15932 + }, + { + "epoch": 0.61, + "learning_rate": 3.981597845601436e-05, + "loss": 0.5217, + "step": 15933 + }, + { + "epoch": 0.61, + "learning_rate": 3.9815337265965634e-05, + "loss": 0.5197, + "step": 15934 + }, + { + "epoch": 0.61, + "learning_rate": 3.98146960759169e-05, + "loss": 0.5097, + "step": 15935 + }, + { + "epoch": 0.61, + "learning_rate": 3.981405488586817e-05, + "loss": 0.5491, + "step": 15936 + }, + { + "epoch": 0.61, + "learning_rate": 3.981341369581944e-05, + "loss": 0.6324, + "step": 15937 + }, + { + "epoch": 0.61, + "learning_rate": 3.9812772505770714e-05, + "loss": 0.4559, + "step": 15938 + }, + { + "epoch": 0.61, + "learning_rate": 3.9812131315721986e-05, + "loss": 0.4833, + "step": 15939 + }, + { + "epoch": 0.61, + "learning_rate": 3.981149012567325e-05, + "loss": 0.4315, + "step": 15940 + }, + { + "epoch": 0.61, + "learning_rate": 3.981084893562452e-05, + "loss": 0.4886, + "step": 15941 + }, + { + "epoch": 0.61, + "learning_rate": 3.981020774557579e-05, + "loss": 0.6114, + "step": 15942 + }, + { + "epoch": 0.61, + "learning_rate": 3.980956655552706e-05, + "loss": 0.4777, + "step": 15943 + }, + { + "epoch": 0.61, + "learning_rate": 3.980892536547833e-05, + "loss": 0.5006, + "step": 15944 + }, + { + "epoch": 0.61, + "learning_rate": 3.98082841754296e-05, + "loss": 0.429, + "step": 15945 + }, + { + "epoch": 0.61, + "learning_rate": 3.980764298538087e-05, + "loss": 0.5118, + "step": 15946 + }, + { + "epoch": 0.61, + "learning_rate": 3.980700179533214e-05, + "loss": 0.4851, + "step": 15947 + }, + { + "epoch": 0.61, + "learning_rate": 3.980636060528341e-05, + "loss": 0.4554, + "step": 15948 + }, + { + "epoch": 0.61, + "learning_rate": 3.9805719415234676e-05, + "loss": 0.6129, + "step": 15949 + }, + { + "epoch": 0.61, + "learning_rate": 3.980507822518595e-05, + "loss": 0.5387, + "step": 15950 + }, + { + "epoch": 0.61, + "learning_rate": 3.980443703513721e-05, + "loss": 0.7085, + "step": 15951 + }, + { + "epoch": 0.61, + "learning_rate": 3.9803795845088484e-05, + "loss": 0.6067, + "step": 15952 + }, + { + "epoch": 0.61, + "learning_rate": 3.9803154655039756e-05, + "loss": 0.6034, + "step": 15953 + }, + { + "epoch": 0.61, + "learning_rate": 3.980251346499103e-05, + "loss": 0.6172, + "step": 15954 + }, + { + "epoch": 0.61, + "learning_rate": 3.980187227494229e-05, + "loss": 0.5906, + "step": 15955 + }, + { + "epoch": 0.61, + "learning_rate": 3.9801231084893565e-05, + "loss": 0.6358, + "step": 15956 + }, + { + "epoch": 0.61, + "learning_rate": 3.9800589894844836e-05, + "loss": 0.4822, + "step": 15957 + }, + { + "epoch": 0.61, + "learning_rate": 3.97999487047961e-05, + "loss": 0.4608, + "step": 15958 + }, + { + "epoch": 0.61, + "learning_rate": 3.979930751474737e-05, + "loss": 0.506, + "step": 15959 + }, + { + "epoch": 0.61, + "learning_rate": 3.9798666324698645e-05, + "loss": 0.6087, + "step": 15960 + }, + { + "epoch": 0.61, + "learning_rate": 3.9798025134649916e-05, + "loss": 0.4506, + "step": 15961 + }, + { + "epoch": 0.61, + "learning_rate": 3.979738394460118e-05, + "loss": 0.4902, + "step": 15962 + }, + { + "epoch": 0.61, + "learning_rate": 3.979674275455245e-05, + "loss": 0.476, + "step": 15963 + }, + { + "epoch": 0.61, + "learning_rate": 3.979610156450372e-05, + "loss": 0.5031, + "step": 15964 + }, + { + "epoch": 0.61, + "learning_rate": 3.979546037445499e-05, + "loss": 0.5322, + "step": 15965 + }, + { + "epoch": 0.61, + "learning_rate": 3.979481918440626e-05, + "loss": 0.7429, + "step": 15966 + }, + { + "epoch": 0.61, + "learning_rate": 3.9794177994357526e-05, + "loss": 0.4505, + "step": 15967 + }, + { + "epoch": 0.61, + "learning_rate": 3.97935368043088e-05, + "loss": 0.4884, + "step": 15968 + }, + { + "epoch": 0.61, + "learning_rate": 3.979289561426007e-05, + "loss": 0.513, + "step": 15969 + }, + { + "epoch": 0.61, + "learning_rate": 3.979225442421134e-05, + "loss": 0.5652, + "step": 15970 + }, + { + "epoch": 0.61, + "learning_rate": 3.9791613234162606e-05, + "loss": 0.5284, + "step": 15971 + }, + { + "epoch": 0.61, + "learning_rate": 3.979097204411388e-05, + "loss": 0.5692, + "step": 15972 + }, + { + "epoch": 0.61, + "learning_rate": 3.979033085406514e-05, + "loss": 0.6365, + "step": 15973 + }, + { + "epoch": 0.61, + "learning_rate": 3.9789689664016415e-05, + "loss": 0.5379, + "step": 15974 + }, + { + "epoch": 0.61, + "learning_rate": 3.9789048473967687e-05, + "loss": 0.5597, + "step": 15975 + }, + { + "epoch": 0.61, + "learning_rate": 3.978840728391895e-05, + "loss": 0.6069, + "step": 15976 + }, + { + "epoch": 0.61, + "learning_rate": 3.978776609387023e-05, + "loss": 0.408, + "step": 15977 + }, + { + "epoch": 0.61, + "learning_rate": 3.9787124903821495e-05, + "loss": 0.5063, + "step": 15978 + }, + { + "epoch": 0.61, + "learning_rate": 3.978648371377277e-05, + "loss": 0.5806, + "step": 15979 + }, + { + "epoch": 0.61, + "learning_rate": 3.978584252372403e-05, + "loss": 0.6489, + "step": 15980 + }, + { + "epoch": 0.61, + "learning_rate": 3.97852013336753e-05, + "loss": 0.5275, + "step": 15981 + }, + { + "epoch": 0.61, + "learning_rate": 3.978456014362657e-05, + "loss": 0.545, + "step": 15982 + }, + { + "epoch": 0.61, + "learning_rate": 3.978391895357784e-05, + "loss": 0.4088, + "step": 15983 + }, + { + "epoch": 0.61, + "learning_rate": 3.978327776352911e-05, + "loss": 0.4152, + "step": 15984 + }, + { + "epoch": 0.61, + "learning_rate": 3.978263657348038e-05, + "loss": 0.5748, + "step": 15985 + }, + { + "epoch": 0.61, + "learning_rate": 3.9781995383431655e-05, + "loss": 0.4515, + "step": 15986 + }, + { + "epoch": 0.61, + "learning_rate": 3.978135419338292e-05, + "loss": 0.5173, + "step": 15987 + }, + { + "epoch": 0.61, + "learning_rate": 3.978071300333419e-05, + "loss": 0.4956, + "step": 15988 + }, + { + "epoch": 0.61, + "learning_rate": 3.978007181328546e-05, + "loss": 0.5911, + "step": 15989 + }, + { + "epoch": 0.61, + "learning_rate": 3.977943062323673e-05, + "loss": 0.5413, + "step": 15990 + }, + { + "epoch": 0.61, + "learning_rate": 3.977878943318799e-05, + "loss": 0.5522, + "step": 15991 + }, + { + "epoch": 0.61, + "learning_rate": 3.9778148243139265e-05, + "loss": 0.5471, + "step": 15992 + }, + { + "epoch": 0.61, + "learning_rate": 3.9777507053090544e-05, + "loss": 0.4346, + "step": 15993 + }, + { + "epoch": 0.61, + "learning_rate": 3.977686586304181e-05, + "loss": 0.6327, + "step": 15994 + }, + { + "epoch": 0.61, + "learning_rate": 3.977622467299308e-05, + "loss": 0.5385, + "step": 15995 + }, + { + "epoch": 0.61, + "learning_rate": 3.9775583482944345e-05, + "loss": 0.5025, + "step": 15996 + }, + { + "epoch": 0.62, + "learning_rate": 3.977494229289562e-05, + "loss": 0.4732, + "step": 15997 + }, + { + "epoch": 0.62, + "learning_rate": 3.977430110284688e-05, + "loss": 0.6402, + "step": 15998 + }, + { + "epoch": 0.62, + "learning_rate": 3.9773659912798154e-05, + "loss": 0.487, + "step": 15999 + }, + { + "epoch": 0.62, + "learning_rate": 3.9773018722749425e-05, + "loss": 0.4848, + "step": 16000 + }, + { + "epoch": 0.62, + "learning_rate": 3.97723775327007e-05, + "loss": 0.53, + "step": 16001 + }, + { + "epoch": 0.62, + "learning_rate": 3.977173634265197e-05, + "loss": 0.4512, + "step": 16002 + }, + { + "epoch": 0.62, + "learning_rate": 3.9771095152603234e-05, + "loss": 0.6061, + "step": 16003 + }, + { + "epoch": 0.62, + "learning_rate": 3.9770453962554505e-05, + "loss": 0.5148, + "step": 16004 + }, + { + "epoch": 0.62, + "learning_rate": 3.976981277250577e-05, + "loss": 0.5629, + "step": 16005 + }, + { + "epoch": 0.62, + "learning_rate": 3.976917158245704e-05, + "loss": 0.4506, + "step": 16006 + }, + { + "epoch": 0.62, + "learning_rate": 3.976853039240831e-05, + "loss": 0.565, + "step": 16007 + }, + { + "epoch": 0.62, + "learning_rate": 3.976788920235958e-05, + "loss": 0.439, + "step": 16008 + }, + { + "epoch": 0.62, + "learning_rate": 3.976724801231085e-05, + "loss": 0.5912, + "step": 16009 + }, + { + "epoch": 0.62, + "learning_rate": 3.976660682226212e-05, + "loss": 0.5843, + "step": 16010 + }, + { + "epoch": 0.62, + "learning_rate": 3.9765965632213394e-05, + "loss": 0.5841, + "step": 16011 + }, + { + "epoch": 0.62, + "learning_rate": 3.976532444216466e-05, + "loss": 0.6543, + "step": 16012 + }, + { + "epoch": 0.62, + "learning_rate": 3.976468325211593e-05, + "loss": 0.5063, + "step": 16013 + }, + { + "epoch": 0.62, + "learning_rate": 3.9764042062067195e-05, + "loss": 0.5878, + "step": 16014 + }, + { + "epoch": 0.62, + "learning_rate": 3.976340087201847e-05, + "loss": 0.5982, + "step": 16015 + }, + { + "epoch": 0.62, + "learning_rate": 3.976275968196974e-05, + "loss": 0.6039, + "step": 16016 + }, + { + "epoch": 0.62, + "learning_rate": 3.976211849192101e-05, + "loss": 0.52, + "step": 16017 + }, + { + "epoch": 0.62, + "learning_rate": 3.9761477301872275e-05, + "loss": 0.4933, + "step": 16018 + }, + { + "epoch": 0.62, + "learning_rate": 3.976083611182355e-05, + "loss": 0.5256, + "step": 16019 + }, + { + "epoch": 0.62, + "learning_rate": 3.976019492177482e-05, + "loss": 0.6582, + "step": 16020 + }, + { + "epoch": 0.62, + "learning_rate": 3.9759553731726084e-05, + "loss": 0.5953, + "step": 16021 + }, + { + "epoch": 0.62, + "learning_rate": 3.9758912541677356e-05, + "loss": 0.509, + "step": 16022 + }, + { + "epoch": 0.62, + "learning_rate": 3.975827135162862e-05, + "loss": 0.4883, + "step": 16023 + }, + { + "epoch": 0.62, + "learning_rate": 3.975763016157989e-05, + "loss": 0.5099, + "step": 16024 + }, + { + "epoch": 0.62, + "learning_rate": 3.9756988971531164e-05, + "loss": 0.5443, + "step": 16025 + }, + { + "epoch": 0.62, + "learning_rate": 3.9756347781482436e-05, + "loss": 0.5745, + "step": 16026 + }, + { + "epoch": 0.62, + "learning_rate": 3.97557065914337e-05, + "loss": 0.4864, + "step": 16027 + }, + { + "epoch": 0.62, + "learning_rate": 3.975506540138497e-05, + "loss": 0.6391, + "step": 16028 + }, + { + "epoch": 0.62, + "learning_rate": 3.9754424211336244e-05, + "loss": 0.5605, + "step": 16029 + }, + { + "epoch": 0.62, + "learning_rate": 3.975378302128751e-05, + "loss": 0.5423, + "step": 16030 + }, + { + "epoch": 0.62, + "learning_rate": 3.975314183123878e-05, + "loss": 0.5788, + "step": 16031 + }, + { + "epoch": 0.62, + "learning_rate": 3.975250064119005e-05, + "loss": 0.503, + "step": 16032 + }, + { + "epoch": 0.62, + "learning_rate": 3.9751859451141324e-05, + "loss": 0.495, + "step": 16033 + }, + { + "epoch": 0.62, + "learning_rate": 3.975121826109259e-05, + "loss": 0.536, + "step": 16034 + }, + { + "epoch": 0.62, + "learning_rate": 3.975057707104386e-05, + "loss": 0.5478, + "step": 16035 + }, + { + "epoch": 0.62, + "learning_rate": 3.9749935880995126e-05, + "loss": 0.4544, + "step": 16036 + }, + { + "epoch": 0.62, + "learning_rate": 3.97492946909464e-05, + "loss": 0.4658, + "step": 16037 + }, + { + "epoch": 0.62, + "learning_rate": 3.974865350089767e-05, + "loss": 0.4753, + "step": 16038 + }, + { + "epoch": 0.62, + "learning_rate": 3.9748012310848934e-05, + "loss": 0.5405, + "step": 16039 + }, + { + "epoch": 0.62, + "learning_rate": 3.9747371120800206e-05, + "loss": 0.573, + "step": 16040 + }, + { + "epoch": 0.62, + "learning_rate": 3.974672993075148e-05, + "loss": 0.6207, + "step": 16041 + }, + { + "epoch": 0.62, + "learning_rate": 3.974608874070275e-05, + "loss": 0.4811, + "step": 16042 + }, + { + "epoch": 0.62, + "learning_rate": 3.9745447550654014e-05, + "loss": 0.5161, + "step": 16043 + }, + { + "epoch": 0.62, + "learning_rate": 3.9744806360605286e-05, + "loss": 0.5079, + "step": 16044 + }, + { + "epoch": 0.62, + "learning_rate": 3.974416517055656e-05, + "loss": 0.5223, + "step": 16045 + }, + { + "epoch": 0.62, + "learning_rate": 3.974352398050782e-05, + "loss": 0.5717, + "step": 16046 + }, + { + "epoch": 0.62, + "learning_rate": 3.9742882790459094e-05, + "loss": 0.4761, + "step": 16047 + }, + { + "epoch": 0.62, + "learning_rate": 3.974224160041036e-05, + "loss": 0.5682, + "step": 16048 + }, + { + "epoch": 0.62, + "learning_rate": 3.974160041036164e-05, + "loss": 0.5505, + "step": 16049 + }, + { + "epoch": 0.62, + "learning_rate": 3.97409592203129e-05, + "loss": 0.5695, + "step": 16050 + }, + { + "epoch": 0.62, + "learning_rate": 3.9740318030264174e-05, + "loss": 0.4142, + "step": 16051 + }, + { + "epoch": 0.62, + "learning_rate": 3.973967684021544e-05, + "loss": 0.498, + "step": 16052 + }, + { + "epoch": 0.62, + "learning_rate": 3.973903565016671e-05, + "loss": 0.4284, + "step": 16053 + }, + { + "epoch": 0.62, + "learning_rate": 3.973839446011798e-05, + "loss": 0.5764, + "step": 16054 + }, + { + "epoch": 0.62, + "learning_rate": 3.973775327006925e-05, + "loss": 0.5198, + "step": 16055 + }, + { + "epoch": 0.62, + "learning_rate": 3.973711208002052e-05, + "loss": 0.5456, + "step": 16056 + }, + { + "epoch": 0.62, + "learning_rate": 3.973647088997179e-05, + "loss": 0.6311, + "step": 16057 + }, + { + "epoch": 0.62, + "learning_rate": 3.973582969992306e-05, + "loss": 0.518, + "step": 16058 + }, + { + "epoch": 0.62, + "learning_rate": 3.973518850987433e-05, + "loss": 0.5029, + "step": 16059 + }, + { + "epoch": 0.62, + "learning_rate": 3.97345473198256e-05, + "loss": 0.5327, + "step": 16060 + }, + { + "epoch": 0.62, + "learning_rate": 3.9733906129776864e-05, + "loss": 0.4838, + "step": 16061 + }, + { + "epoch": 0.62, + "learning_rate": 3.9733264939728136e-05, + "loss": 0.51, + "step": 16062 + }, + { + "epoch": 0.62, + "learning_rate": 3.973262374967941e-05, + "loss": 0.5193, + "step": 16063 + }, + { + "epoch": 0.62, + "learning_rate": 3.973198255963067e-05, + "loss": 0.6994, + "step": 16064 + }, + { + "epoch": 0.62, + "learning_rate": 3.973134136958195e-05, + "loss": 0.4702, + "step": 16065 + }, + { + "epoch": 0.62, + "learning_rate": 3.9730700179533216e-05, + "loss": 0.5765, + "step": 16066 + }, + { + "epoch": 0.62, + "learning_rate": 3.973005898948449e-05, + "loss": 0.5451, + "step": 16067 + }, + { + "epoch": 0.62, + "learning_rate": 3.972941779943575e-05, + "loss": 0.5168, + "step": 16068 + }, + { + "epoch": 0.62, + "learning_rate": 3.9728776609387025e-05, + "loss": 0.595, + "step": 16069 + }, + { + "epoch": 0.62, + "learning_rate": 3.972813541933829e-05, + "loss": 0.615, + "step": 16070 + }, + { + "epoch": 0.62, + "learning_rate": 3.972749422928956e-05, + "loss": 0.5149, + "step": 16071 + }, + { + "epoch": 0.62, + "learning_rate": 3.972685303924083e-05, + "loss": 0.4116, + "step": 16072 + }, + { + "epoch": 0.62, + "learning_rate": 3.9726211849192105e-05, + "loss": 0.5609, + "step": 16073 + }, + { + "epoch": 0.62, + "learning_rate": 3.9725570659143376e-05, + "loss": 0.5381, + "step": 16074 + }, + { + "epoch": 0.62, + "learning_rate": 3.972492946909464e-05, + "loss": 0.4869, + "step": 16075 + }, + { + "epoch": 0.62, + "learning_rate": 3.972428827904591e-05, + "loss": 0.4975, + "step": 16076 + }, + { + "epoch": 0.62, + "learning_rate": 3.972364708899718e-05, + "loss": 0.5121, + "step": 16077 + }, + { + "epoch": 0.62, + "learning_rate": 3.972300589894845e-05, + "loss": 0.561, + "step": 16078 + }, + { + "epoch": 0.62, + "learning_rate": 3.9722364708899715e-05, + "loss": 0.5238, + "step": 16079 + }, + { + "epoch": 0.62, + "learning_rate": 3.9721723518850986e-05, + "loss": 0.444, + "step": 16080 + }, + { + "epoch": 0.62, + "learning_rate": 3.9721082328802265e-05, + "loss": 0.6023, + "step": 16081 + }, + { + "epoch": 0.62, + "learning_rate": 3.972044113875353e-05, + "loss": 0.5993, + "step": 16082 + }, + { + "epoch": 0.62, + "learning_rate": 3.97197999487048e-05, + "loss": 0.5386, + "step": 16083 + }, + { + "epoch": 0.62, + "learning_rate": 3.9719158758656066e-05, + "loss": 0.6026, + "step": 16084 + }, + { + "epoch": 0.62, + "learning_rate": 3.971851756860734e-05, + "loss": 0.4823, + "step": 16085 + }, + { + "epoch": 0.62, + "learning_rate": 3.97178763785586e-05, + "loss": 0.4956, + "step": 16086 + }, + { + "epoch": 0.62, + "learning_rate": 3.9717235188509875e-05, + "loss": 0.5257, + "step": 16087 + }, + { + "epoch": 0.62, + "learning_rate": 3.9716593998461147e-05, + "loss": 0.4581, + "step": 16088 + }, + { + "epoch": 0.62, + "learning_rate": 3.971595280841242e-05, + "loss": 0.5832, + "step": 16089 + }, + { + "epoch": 0.62, + "learning_rate": 3.971531161836369e-05, + "loss": 0.5396, + "step": 16090 + }, + { + "epoch": 0.62, + "learning_rate": 3.9714670428314955e-05, + "loss": 0.5147, + "step": 16091 + }, + { + "epoch": 0.62, + "learning_rate": 3.971402923826623e-05, + "loss": 0.4994, + "step": 16092 + }, + { + "epoch": 0.62, + "learning_rate": 3.971338804821749e-05, + "loss": 0.6058, + "step": 16093 + }, + { + "epoch": 0.62, + "learning_rate": 3.971274685816876e-05, + "loss": 0.5656, + "step": 16094 + }, + { + "epoch": 0.62, + "learning_rate": 3.971210566812003e-05, + "loss": 0.5331, + "step": 16095 + }, + { + "epoch": 0.62, + "learning_rate": 3.97114644780713e-05, + "loss": 0.5019, + "step": 16096 + }, + { + "epoch": 0.62, + "learning_rate": 3.971082328802257e-05, + "loss": 0.4978, + "step": 16097 + }, + { + "epoch": 0.62, + "learning_rate": 3.9710182097973843e-05, + "loss": 0.5823, + "step": 16098 + }, + { + "epoch": 0.62, + "learning_rate": 3.9709540907925115e-05, + "loss": 0.5049, + "step": 16099 + }, + { + "epoch": 0.62, + "learning_rate": 3.970889971787638e-05, + "loss": 0.5344, + "step": 16100 + }, + { + "epoch": 0.62, + "learning_rate": 3.970825852782765e-05, + "loss": 0.5089, + "step": 16101 + }, + { + "epoch": 0.62, + "learning_rate": 3.970761733777892e-05, + "loss": 0.4722, + "step": 16102 + }, + { + "epoch": 0.62, + "learning_rate": 3.970697614773019e-05, + "loss": 0.5211, + "step": 16103 + }, + { + "epoch": 0.62, + "learning_rate": 3.970633495768146e-05, + "loss": 0.474, + "step": 16104 + }, + { + "epoch": 0.62, + "learning_rate": 3.970569376763273e-05, + "loss": 0.5499, + "step": 16105 + }, + { + "epoch": 0.62, + "learning_rate": 3.9705052577584e-05, + "loss": 0.6128, + "step": 16106 + }, + { + "epoch": 0.62, + "learning_rate": 3.970441138753527e-05, + "loss": 0.568, + "step": 16107 + }, + { + "epoch": 0.62, + "learning_rate": 3.970377019748654e-05, + "loss": 0.5059, + "step": 16108 + }, + { + "epoch": 0.62, + "learning_rate": 3.9703129007437805e-05, + "loss": 0.5211, + "step": 16109 + }, + { + "epoch": 0.62, + "learning_rate": 3.970248781738908e-05, + "loss": 0.519, + "step": 16110 + }, + { + "epoch": 0.62, + "learning_rate": 3.970184662734034e-05, + "loss": 0.4889, + "step": 16111 + }, + { + "epoch": 0.62, + "learning_rate": 3.9701205437291614e-05, + "loss": 0.5347, + "step": 16112 + }, + { + "epoch": 0.62, + "learning_rate": 3.9700564247242885e-05, + "loss": 0.5618, + "step": 16113 + }, + { + "epoch": 0.62, + "learning_rate": 3.969992305719416e-05, + "loss": 0.4683, + "step": 16114 + }, + { + "epoch": 0.62, + "learning_rate": 3.969928186714542e-05, + "loss": 0.4814, + "step": 16115 + }, + { + "epoch": 0.62, + "learning_rate": 3.9698640677096694e-05, + "loss": 0.4914, + "step": 16116 + }, + { + "epoch": 0.62, + "learning_rate": 3.9697999487047965e-05, + "loss": 0.5074, + "step": 16117 + }, + { + "epoch": 0.62, + "learning_rate": 3.969735829699923e-05, + "loss": 0.5601, + "step": 16118 + }, + { + "epoch": 0.62, + "learning_rate": 3.96967171069505e-05, + "loss": 0.49, + "step": 16119 + }, + { + "epoch": 0.62, + "learning_rate": 3.9696075916901774e-05, + "loss": 0.5955, + "step": 16120 + }, + { + "epoch": 0.62, + "learning_rate": 3.9695434726853045e-05, + "loss": 0.5166, + "step": 16121 + }, + { + "epoch": 0.62, + "learning_rate": 3.969479353680431e-05, + "loss": 0.5815, + "step": 16122 + }, + { + "epoch": 0.62, + "learning_rate": 3.969415234675558e-05, + "loss": 0.5079, + "step": 16123 + }, + { + "epoch": 0.62, + "learning_rate": 3.969351115670685e-05, + "loss": 0.475, + "step": 16124 + }, + { + "epoch": 0.62, + "learning_rate": 3.969286996665812e-05, + "loss": 0.5082, + "step": 16125 + }, + { + "epoch": 0.62, + "learning_rate": 3.969222877660939e-05, + "loss": 0.4389, + "step": 16126 + }, + { + "epoch": 0.62, + "learning_rate": 3.9691587586560655e-05, + "loss": 0.5433, + "step": 16127 + }, + { + "epoch": 0.62, + "learning_rate": 3.969094639651193e-05, + "loss": 0.466, + "step": 16128 + }, + { + "epoch": 0.62, + "learning_rate": 3.96903052064632e-05, + "loss": 0.49, + "step": 16129 + }, + { + "epoch": 0.62, + "learning_rate": 3.968966401641447e-05, + "loss": 0.5512, + "step": 16130 + }, + { + "epoch": 0.62, + "learning_rate": 3.9689022826365736e-05, + "loss": 0.607, + "step": 16131 + }, + { + "epoch": 0.62, + "learning_rate": 3.968838163631701e-05, + "loss": 0.5585, + "step": 16132 + }, + { + "epoch": 0.62, + "learning_rate": 3.968774044626827e-05, + "loss": 0.3926, + "step": 16133 + }, + { + "epoch": 0.62, + "learning_rate": 3.9687099256219544e-05, + "loss": 0.5663, + "step": 16134 + }, + { + "epoch": 0.62, + "learning_rate": 3.9686458066170816e-05, + "loss": 0.6051, + "step": 16135 + }, + { + "epoch": 0.62, + "learning_rate": 3.968581687612208e-05, + "loss": 0.5743, + "step": 16136 + }, + { + "epoch": 0.62, + "learning_rate": 3.968517568607336e-05, + "loss": 0.7205, + "step": 16137 + }, + { + "epoch": 0.62, + "learning_rate": 3.9684534496024624e-05, + "loss": 0.5354, + "step": 16138 + }, + { + "epoch": 0.62, + "learning_rate": 3.9683893305975896e-05, + "loss": 0.5125, + "step": 16139 + }, + { + "epoch": 0.62, + "learning_rate": 3.968325211592716e-05, + "loss": 0.4761, + "step": 16140 + }, + { + "epoch": 0.62, + "learning_rate": 3.968261092587843e-05, + "loss": 0.5724, + "step": 16141 + }, + { + "epoch": 0.62, + "learning_rate": 3.96819697358297e-05, + "loss": 0.52, + "step": 16142 + }, + { + "epoch": 0.62, + "learning_rate": 3.968132854578097e-05, + "loss": 0.5157, + "step": 16143 + }, + { + "epoch": 0.62, + "learning_rate": 3.968068735573224e-05, + "loss": 0.4734, + "step": 16144 + }, + { + "epoch": 0.62, + "learning_rate": 3.968004616568351e-05, + "loss": 0.5699, + "step": 16145 + }, + { + "epoch": 0.62, + "learning_rate": 3.9679404975634784e-05, + "loss": 0.4867, + "step": 16146 + }, + { + "epoch": 0.62, + "learning_rate": 3.967876378558605e-05, + "loss": 0.5716, + "step": 16147 + }, + { + "epoch": 0.62, + "learning_rate": 3.967812259553732e-05, + "loss": 0.5634, + "step": 16148 + }, + { + "epoch": 0.62, + "learning_rate": 3.9677481405488586e-05, + "loss": 0.4994, + "step": 16149 + }, + { + "epoch": 0.62, + "learning_rate": 3.967684021543986e-05, + "loss": 0.5211, + "step": 16150 + }, + { + "epoch": 0.62, + "learning_rate": 3.967619902539112e-05, + "loss": 0.5184, + "step": 16151 + }, + { + "epoch": 0.62, + "learning_rate": 3.9675557835342394e-05, + "loss": 0.5071, + "step": 16152 + }, + { + "epoch": 0.62, + "learning_rate": 3.967491664529367e-05, + "loss": 0.5697, + "step": 16153 + }, + { + "epoch": 0.62, + "learning_rate": 3.967427545524494e-05, + "loss": 0.5096, + "step": 16154 + }, + { + "epoch": 0.62, + "learning_rate": 3.967363426519621e-05, + "loss": 0.4581, + "step": 16155 + }, + { + "epoch": 0.62, + "learning_rate": 3.9672993075147474e-05, + "loss": 0.5007, + "step": 16156 + }, + { + "epoch": 0.62, + "learning_rate": 3.9672351885098746e-05, + "loss": 0.4967, + "step": 16157 + }, + { + "epoch": 0.62, + "learning_rate": 3.967171069505001e-05, + "loss": 0.6541, + "step": 16158 + }, + { + "epoch": 0.62, + "learning_rate": 3.967106950500128e-05, + "loss": 0.5132, + "step": 16159 + }, + { + "epoch": 0.62, + "learning_rate": 3.9670428314952554e-05, + "loss": 0.5792, + "step": 16160 + }, + { + "epoch": 0.62, + "learning_rate": 3.9669787124903826e-05, + "loss": 0.4709, + "step": 16161 + }, + { + "epoch": 0.62, + "learning_rate": 3.96691459348551e-05, + "loss": 0.4767, + "step": 16162 + }, + { + "epoch": 0.62, + "learning_rate": 3.966850474480636e-05, + "loss": 0.4803, + "step": 16163 + }, + { + "epoch": 0.62, + "learning_rate": 3.9667863554757634e-05, + "loss": 0.5797, + "step": 16164 + }, + { + "epoch": 0.62, + "learning_rate": 3.96672223647089e-05, + "loss": 0.4932, + "step": 16165 + }, + { + "epoch": 0.62, + "learning_rate": 3.966658117466017e-05, + "loss": 0.4856, + "step": 16166 + }, + { + "epoch": 0.62, + "learning_rate": 3.9665939984611436e-05, + "loss": 0.5306, + "step": 16167 + }, + { + "epoch": 0.62, + "learning_rate": 3.966529879456271e-05, + "loss": 0.6584, + "step": 16168 + }, + { + "epoch": 0.62, + "learning_rate": 3.966465760451398e-05, + "loss": 0.5892, + "step": 16169 + }, + { + "epoch": 0.62, + "learning_rate": 3.966401641446525e-05, + "loss": 0.5485, + "step": 16170 + }, + { + "epoch": 0.62, + "learning_rate": 3.966337522441652e-05, + "loss": 0.4884, + "step": 16171 + }, + { + "epoch": 0.62, + "learning_rate": 3.966273403436779e-05, + "loss": 0.5378, + "step": 16172 + }, + { + "epoch": 0.62, + "learning_rate": 3.966209284431906e-05, + "loss": 0.6156, + "step": 16173 + }, + { + "epoch": 0.62, + "learning_rate": 3.9661451654270324e-05, + "loss": 0.4414, + "step": 16174 + }, + { + "epoch": 0.62, + "learning_rate": 3.9660810464221596e-05, + "loss": 0.5067, + "step": 16175 + }, + { + "epoch": 0.62, + "learning_rate": 3.966016927417287e-05, + "loss": 0.5214, + "step": 16176 + }, + { + "epoch": 0.62, + "learning_rate": 3.965952808412414e-05, + "loss": 0.4874, + "step": 16177 + }, + { + "epoch": 0.62, + "learning_rate": 3.9658886894075405e-05, + "loss": 0.4873, + "step": 16178 + }, + { + "epoch": 0.62, + "learning_rate": 3.9658245704026676e-05, + "loss": 0.5247, + "step": 16179 + }, + { + "epoch": 0.62, + "learning_rate": 3.965760451397795e-05, + "loss": 0.4103, + "step": 16180 + }, + { + "epoch": 0.62, + "learning_rate": 3.965696332392921e-05, + "loss": 0.5596, + "step": 16181 + }, + { + "epoch": 0.62, + "learning_rate": 3.9656322133880485e-05, + "loss": 0.5893, + "step": 16182 + }, + { + "epoch": 0.62, + "learning_rate": 3.965568094383175e-05, + "loss": 0.53, + "step": 16183 + }, + { + "epoch": 0.62, + "learning_rate": 3.965503975378302e-05, + "loss": 0.5566, + "step": 16184 + }, + { + "epoch": 0.62, + "learning_rate": 3.965439856373429e-05, + "loss": 0.5218, + "step": 16185 + }, + { + "epoch": 0.62, + "learning_rate": 3.9653757373685565e-05, + "loss": 0.3681, + "step": 16186 + }, + { + "epoch": 0.62, + "learning_rate": 3.965311618363683e-05, + "loss": 0.514, + "step": 16187 + }, + { + "epoch": 0.62, + "learning_rate": 3.96524749935881e-05, + "loss": 0.4082, + "step": 16188 + }, + { + "epoch": 0.62, + "learning_rate": 3.965183380353937e-05, + "loss": 0.5452, + "step": 16189 + }, + { + "epoch": 0.62, + "learning_rate": 3.965119261349064e-05, + "loss": 0.5848, + "step": 16190 + }, + { + "epoch": 0.62, + "learning_rate": 3.965055142344191e-05, + "loss": 0.6026, + "step": 16191 + }, + { + "epoch": 0.62, + "learning_rate": 3.964991023339318e-05, + "loss": 0.53, + "step": 16192 + }, + { + "epoch": 0.62, + "learning_rate": 3.964926904334445e-05, + "loss": 0.4833, + "step": 16193 + }, + { + "epoch": 0.62, + "learning_rate": 3.964862785329572e-05, + "loss": 0.5639, + "step": 16194 + }, + { + "epoch": 0.62, + "learning_rate": 3.964798666324699e-05, + "loss": 0.4481, + "step": 16195 + }, + { + "epoch": 0.62, + "learning_rate": 3.9647345473198255e-05, + "loss": 0.4972, + "step": 16196 + }, + { + "epoch": 0.62, + "learning_rate": 3.9646704283149527e-05, + "loss": 0.5466, + "step": 16197 + }, + { + "epoch": 0.62, + "learning_rate": 3.96460630931008e-05, + "loss": 0.4928, + "step": 16198 + }, + { + "epoch": 0.62, + "learning_rate": 3.964542190305206e-05, + "loss": 0.5461, + "step": 16199 + }, + { + "epoch": 0.62, + "learning_rate": 3.9644780713003335e-05, + "loss": 0.4576, + "step": 16200 + }, + { + "epoch": 0.62, + "learning_rate": 3.9644139522954607e-05, + "loss": 0.4867, + "step": 16201 + }, + { + "epoch": 0.62, + "learning_rate": 3.964349833290588e-05, + "loss": 0.5448, + "step": 16202 + }, + { + "epoch": 0.62, + "learning_rate": 3.964285714285714e-05, + "loss": 0.6266, + "step": 16203 + }, + { + "epoch": 0.62, + "learning_rate": 3.9642215952808415e-05, + "loss": 0.5152, + "step": 16204 + }, + { + "epoch": 0.62, + "learning_rate": 3.964157476275968e-05, + "loss": 0.4875, + "step": 16205 + }, + { + "epoch": 0.62, + "learning_rate": 3.964093357271095e-05, + "loss": 0.5553, + "step": 16206 + }, + { + "epoch": 0.62, + "learning_rate": 3.964029238266222e-05, + "loss": 0.5801, + "step": 16207 + }, + { + "epoch": 0.62, + "learning_rate": 3.9639651192613495e-05, + "loss": 0.5586, + "step": 16208 + }, + { + "epoch": 0.62, + "learning_rate": 3.963901000256477e-05, + "loss": 0.5001, + "step": 16209 + }, + { + "epoch": 0.62, + "learning_rate": 3.963836881251603e-05, + "loss": 0.6202, + "step": 16210 + }, + { + "epoch": 0.62, + "learning_rate": 3.9637727622467303e-05, + "loss": 0.5356, + "step": 16211 + }, + { + "epoch": 0.62, + "learning_rate": 3.963708643241857e-05, + "loss": 0.4533, + "step": 16212 + }, + { + "epoch": 0.62, + "learning_rate": 3.963644524236984e-05, + "loss": 0.5428, + "step": 16213 + }, + { + "epoch": 0.62, + "learning_rate": 3.9635804052321105e-05, + "loss": 0.5026, + "step": 16214 + }, + { + "epoch": 0.62, + "learning_rate": 3.963516286227238e-05, + "loss": 0.5123, + "step": 16215 + }, + { + "epoch": 0.62, + "learning_rate": 3.963452167222365e-05, + "loss": 0.5902, + "step": 16216 + }, + { + "epoch": 0.62, + "learning_rate": 3.963388048217492e-05, + "loss": 0.6169, + "step": 16217 + }, + { + "epoch": 0.62, + "learning_rate": 3.963323929212619e-05, + "loss": 0.5587, + "step": 16218 + }, + { + "epoch": 0.62, + "learning_rate": 3.963259810207746e-05, + "loss": 0.4914, + "step": 16219 + }, + { + "epoch": 0.62, + "learning_rate": 3.963195691202873e-05, + "loss": 0.4369, + "step": 16220 + }, + { + "epoch": 0.62, + "learning_rate": 3.9631315721979993e-05, + "loss": 0.5832, + "step": 16221 + }, + { + "epoch": 0.62, + "learning_rate": 3.9630674531931265e-05, + "loss": 0.5509, + "step": 16222 + }, + { + "epoch": 0.62, + "learning_rate": 3.963003334188253e-05, + "loss": 0.4563, + "step": 16223 + }, + { + "epoch": 0.62, + "learning_rate": 3.96293921518338e-05, + "loss": 0.5729, + "step": 16224 + }, + { + "epoch": 0.62, + "learning_rate": 3.962875096178508e-05, + "loss": 0.522, + "step": 16225 + }, + { + "epoch": 0.62, + "learning_rate": 3.9628109771736345e-05, + "loss": 0.5372, + "step": 16226 + }, + { + "epoch": 0.62, + "learning_rate": 3.962746858168762e-05, + "loss": 0.4693, + "step": 16227 + }, + { + "epoch": 0.62, + "learning_rate": 3.962682739163888e-05, + "loss": 0.4892, + "step": 16228 + }, + { + "epoch": 0.62, + "learning_rate": 3.9626186201590154e-05, + "loss": 0.4887, + "step": 16229 + }, + { + "epoch": 0.62, + "learning_rate": 3.962554501154142e-05, + "loss": 0.5463, + "step": 16230 + }, + { + "epoch": 0.62, + "learning_rate": 3.962490382149269e-05, + "loss": 0.688, + "step": 16231 + }, + { + "epoch": 0.62, + "learning_rate": 3.962426263144396e-05, + "loss": 0.4756, + "step": 16232 + }, + { + "epoch": 0.62, + "learning_rate": 3.9623621441395234e-05, + "loss": 0.4304, + "step": 16233 + }, + { + "epoch": 0.62, + "learning_rate": 3.9622980251346505e-05, + "loss": 0.558, + "step": 16234 + }, + { + "epoch": 0.62, + "learning_rate": 3.962233906129777e-05, + "loss": 0.594, + "step": 16235 + }, + { + "epoch": 0.62, + "learning_rate": 3.962169787124904e-05, + "loss": 0.4809, + "step": 16236 + }, + { + "epoch": 0.62, + "learning_rate": 3.962105668120031e-05, + "loss": 0.5221, + "step": 16237 + }, + { + "epoch": 0.62, + "learning_rate": 3.962041549115158e-05, + "loss": 0.533, + "step": 16238 + }, + { + "epoch": 0.62, + "learning_rate": 3.9619774301102844e-05, + "loss": 0.4283, + "step": 16239 + }, + { + "epoch": 0.62, + "learning_rate": 3.9619133111054115e-05, + "loss": 0.4121, + "step": 16240 + }, + { + "epoch": 0.62, + "learning_rate": 3.9618491921005394e-05, + "loss": 0.5424, + "step": 16241 + }, + { + "epoch": 0.62, + "learning_rate": 3.961785073095666e-05, + "loss": 0.7312, + "step": 16242 + }, + { + "epoch": 0.62, + "learning_rate": 3.961720954090793e-05, + "loss": 0.5755, + "step": 16243 + }, + { + "epoch": 0.62, + "learning_rate": 3.9616568350859196e-05, + "loss": 0.5814, + "step": 16244 + }, + { + "epoch": 0.62, + "learning_rate": 3.961592716081047e-05, + "loss": 0.5373, + "step": 16245 + }, + { + "epoch": 0.62, + "learning_rate": 3.961528597076173e-05, + "loss": 0.6484, + "step": 16246 + }, + { + "epoch": 0.62, + "learning_rate": 3.9614644780713004e-05, + "loss": 0.6385, + "step": 16247 + }, + { + "epoch": 0.62, + "learning_rate": 3.9614003590664276e-05, + "loss": 0.5182, + "step": 16248 + }, + { + "epoch": 0.62, + "learning_rate": 3.961336240061555e-05, + "loss": 0.5169, + "step": 16249 + }, + { + "epoch": 0.62, + "learning_rate": 3.961272121056682e-05, + "loss": 0.6185, + "step": 16250 + }, + { + "epoch": 0.62, + "learning_rate": 3.9612080020518084e-05, + "loss": 0.4233, + "step": 16251 + }, + { + "epoch": 0.62, + "learning_rate": 3.9611438830469356e-05, + "loss": 0.5882, + "step": 16252 + }, + { + "epoch": 0.62, + "learning_rate": 3.961079764042062e-05, + "loss": 0.5115, + "step": 16253 + }, + { + "epoch": 0.62, + "learning_rate": 3.961015645037189e-05, + "loss": 0.4957, + "step": 16254 + }, + { + "epoch": 0.62, + "learning_rate": 3.960951526032316e-05, + "loss": 0.5681, + "step": 16255 + }, + { + "epoch": 0.62, + "learning_rate": 3.960887407027443e-05, + "loss": 0.5373, + "step": 16256 + }, + { + "epoch": 0.63, + "learning_rate": 3.96082328802257e-05, + "loss": 0.566, + "step": 16257 + }, + { + "epoch": 0.63, + "learning_rate": 3.960759169017697e-05, + "loss": 0.5744, + "step": 16258 + }, + { + "epoch": 0.63, + "learning_rate": 3.9606950500128244e-05, + "loss": 0.5961, + "step": 16259 + }, + { + "epoch": 0.63, + "learning_rate": 3.960630931007951e-05, + "loss": 0.6061, + "step": 16260 + }, + { + "epoch": 0.63, + "learning_rate": 3.960566812003078e-05, + "loss": 0.5601, + "step": 16261 + }, + { + "epoch": 0.63, + "learning_rate": 3.9605026929982046e-05, + "loss": 0.5149, + "step": 16262 + }, + { + "epoch": 0.63, + "learning_rate": 3.960438573993332e-05, + "loss": 0.4868, + "step": 16263 + }, + { + "epoch": 0.63, + "learning_rate": 3.960374454988459e-05, + "loss": 0.4861, + "step": 16264 + }, + { + "epoch": 0.63, + "learning_rate": 3.960310335983586e-05, + "loss": 0.492, + "step": 16265 + }, + { + "epoch": 0.63, + "learning_rate": 3.9602462169787126e-05, + "loss": 0.4696, + "step": 16266 + }, + { + "epoch": 0.63, + "learning_rate": 3.96018209797384e-05, + "loss": 0.5021, + "step": 16267 + }, + { + "epoch": 0.63, + "learning_rate": 3.960117978968967e-05, + "loss": 0.5557, + "step": 16268 + }, + { + "epoch": 0.63, + "learning_rate": 3.9600538599640934e-05, + "loss": 0.4858, + "step": 16269 + }, + { + "epoch": 0.63, + "learning_rate": 3.9599897409592206e-05, + "loss": 0.5231, + "step": 16270 + }, + { + "epoch": 0.63, + "learning_rate": 3.959925621954347e-05, + "loss": 0.4836, + "step": 16271 + }, + { + "epoch": 0.63, + "learning_rate": 3.959861502949474e-05, + "loss": 0.6224, + "step": 16272 + }, + { + "epoch": 0.63, + "learning_rate": 3.9597973839446014e-05, + "loss": 0.4699, + "step": 16273 + }, + { + "epoch": 0.63, + "learning_rate": 3.9597332649397286e-05, + "loss": 0.5433, + "step": 16274 + }, + { + "epoch": 0.63, + "learning_rate": 3.959669145934855e-05, + "loss": 0.538, + "step": 16275 + }, + { + "epoch": 0.63, + "learning_rate": 3.959605026929982e-05, + "loss": 0.5946, + "step": 16276 + }, + { + "epoch": 0.63, + "learning_rate": 3.9595409079251094e-05, + "loss": 0.6279, + "step": 16277 + }, + { + "epoch": 0.63, + "learning_rate": 3.959476788920236e-05, + "loss": 0.6023, + "step": 16278 + }, + { + "epoch": 0.63, + "learning_rate": 3.959412669915363e-05, + "loss": 0.4322, + "step": 16279 + }, + { + "epoch": 0.63, + "learning_rate": 3.95934855091049e-05, + "loss": 0.5053, + "step": 16280 + }, + { + "epoch": 0.63, + "learning_rate": 3.9592844319056175e-05, + "loss": 0.5239, + "step": 16281 + }, + { + "epoch": 0.63, + "learning_rate": 3.959220312900744e-05, + "loss": 0.4383, + "step": 16282 + }, + { + "epoch": 0.63, + "learning_rate": 3.959156193895871e-05, + "loss": 0.5039, + "step": 16283 + }, + { + "epoch": 0.63, + "learning_rate": 3.9590920748909976e-05, + "loss": 0.4821, + "step": 16284 + }, + { + "epoch": 0.63, + "learning_rate": 3.959027955886125e-05, + "loss": 0.4504, + "step": 16285 + }, + { + "epoch": 0.63, + "learning_rate": 3.958963836881252e-05, + "loss": 0.4754, + "step": 16286 + }, + { + "epoch": 0.63, + "learning_rate": 3.9588997178763784e-05, + "loss": 0.5738, + "step": 16287 + }, + { + "epoch": 0.63, + "learning_rate": 3.9588355988715056e-05, + "loss": 0.5617, + "step": 16288 + }, + { + "epoch": 0.63, + "learning_rate": 3.958771479866633e-05, + "loss": 0.4045, + "step": 16289 + }, + { + "epoch": 0.63, + "learning_rate": 3.95870736086176e-05, + "loss": 0.4566, + "step": 16290 + }, + { + "epoch": 0.63, + "learning_rate": 3.9586432418568865e-05, + "loss": 0.5595, + "step": 16291 + }, + { + "epoch": 0.63, + "learning_rate": 3.9585791228520136e-05, + "loss": 0.5356, + "step": 16292 + }, + { + "epoch": 0.63, + "learning_rate": 3.95851500384714e-05, + "loss": 0.5721, + "step": 16293 + }, + { + "epoch": 0.63, + "learning_rate": 3.958450884842267e-05, + "loss": 0.6553, + "step": 16294 + }, + { + "epoch": 0.63, + "learning_rate": 3.9583867658373945e-05, + "loss": 0.4808, + "step": 16295 + }, + { + "epoch": 0.63, + "learning_rate": 3.958322646832521e-05, + "loss": 0.5826, + "step": 16296 + }, + { + "epoch": 0.63, + "learning_rate": 3.958258527827649e-05, + "loss": 0.6578, + "step": 16297 + }, + { + "epoch": 0.63, + "learning_rate": 3.958194408822775e-05, + "loss": 0.6017, + "step": 16298 + }, + { + "epoch": 0.63, + "learning_rate": 3.9581302898179025e-05, + "loss": 0.5482, + "step": 16299 + }, + { + "epoch": 0.63, + "learning_rate": 3.958066170813029e-05, + "loss": 0.5084, + "step": 16300 + }, + { + "epoch": 0.63, + "learning_rate": 3.958002051808156e-05, + "loss": 0.5206, + "step": 16301 + }, + { + "epoch": 0.63, + "learning_rate": 3.9579379328032826e-05, + "loss": 0.5667, + "step": 16302 + }, + { + "epoch": 0.63, + "learning_rate": 3.95787381379841e-05, + "loss": 0.5148, + "step": 16303 + }, + { + "epoch": 0.63, + "learning_rate": 3.957809694793537e-05, + "loss": 0.5919, + "step": 16304 + }, + { + "epoch": 0.63, + "learning_rate": 3.957745575788664e-05, + "loss": 0.677, + "step": 16305 + }, + { + "epoch": 0.63, + "learning_rate": 3.957681456783791e-05, + "loss": 0.5799, + "step": 16306 + }, + { + "epoch": 0.63, + "learning_rate": 3.957617337778918e-05, + "loss": 0.4668, + "step": 16307 + }, + { + "epoch": 0.63, + "learning_rate": 3.957553218774045e-05, + "loss": 0.5865, + "step": 16308 + }, + { + "epoch": 0.63, + "learning_rate": 3.9574890997691715e-05, + "loss": 0.5769, + "step": 16309 + }, + { + "epoch": 0.63, + "learning_rate": 3.9574249807642987e-05, + "loss": 0.4675, + "step": 16310 + }, + { + "epoch": 0.63, + "learning_rate": 3.957360861759425e-05, + "loss": 0.5002, + "step": 16311 + }, + { + "epoch": 0.63, + "learning_rate": 3.957296742754552e-05, + "loss": 0.4514, + "step": 16312 + }, + { + "epoch": 0.63, + "learning_rate": 3.95723262374968e-05, + "loss": 0.571, + "step": 16313 + }, + { + "epoch": 0.63, + "learning_rate": 3.957168504744807e-05, + "loss": 0.6237, + "step": 16314 + }, + { + "epoch": 0.63, + "learning_rate": 3.957104385739934e-05, + "loss": 0.5089, + "step": 16315 + }, + { + "epoch": 0.63, + "learning_rate": 3.95704026673506e-05, + "loss": 0.5906, + "step": 16316 + }, + { + "epoch": 0.63, + "learning_rate": 3.9569761477301875e-05, + "loss": 0.4464, + "step": 16317 + }, + { + "epoch": 0.63, + "learning_rate": 3.956912028725314e-05, + "loss": 0.5435, + "step": 16318 + }, + { + "epoch": 0.63, + "learning_rate": 3.956847909720441e-05, + "loss": 0.5024, + "step": 16319 + }, + { + "epoch": 0.63, + "learning_rate": 3.956783790715568e-05, + "loss": 0.5407, + "step": 16320 + }, + { + "epoch": 0.63, + "learning_rate": 3.9567196717106955e-05, + "loss": 0.5638, + "step": 16321 + }, + { + "epoch": 0.63, + "learning_rate": 3.956655552705823e-05, + "loss": 0.4664, + "step": 16322 + }, + { + "epoch": 0.63, + "learning_rate": 3.956591433700949e-05, + "loss": 0.5642, + "step": 16323 + }, + { + "epoch": 0.63, + "learning_rate": 3.9565273146960763e-05, + "loss": 0.4338, + "step": 16324 + }, + { + "epoch": 0.63, + "learning_rate": 3.956463195691203e-05, + "loss": 0.5845, + "step": 16325 + }, + { + "epoch": 0.63, + "learning_rate": 3.95639907668633e-05, + "loss": 0.5359, + "step": 16326 + }, + { + "epoch": 0.63, + "learning_rate": 3.9563349576814565e-05, + "loss": 0.4901, + "step": 16327 + }, + { + "epoch": 0.63, + "learning_rate": 3.956270838676584e-05, + "loss": 0.5052, + "step": 16328 + }, + { + "epoch": 0.63, + "learning_rate": 3.956206719671711e-05, + "loss": 0.5991, + "step": 16329 + }, + { + "epoch": 0.63, + "learning_rate": 3.956142600666838e-05, + "loss": 0.4431, + "step": 16330 + }, + { + "epoch": 0.63, + "learning_rate": 3.956078481661965e-05, + "loss": 0.5466, + "step": 16331 + }, + { + "epoch": 0.63, + "learning_rate": 3.956014362657092e-05, + "loss": 0.5668, + "step": 16332 + }, + { + "epoch": 0.63, + "learning_rate": 3.955950243652219e-05, + "loss": 0.5709, + "step": 16333 + }, + { + "epoch": 0.63, + "learning_rate": 3.9558861246473454e-05, + "loss": 0.564, + "step": 16334 + }, + { + "epoch": 0.63, + "learning_rate": 3.9558220056424725e-05, + "loss": 0.5982, + "step": 16335 + }, + { + "epoch": 0.63, + "learning_rate": 3.9557578866376e-05, + "loss": 0.5979, + "step": 16336 + }, + { + "epoch": 0.63, + "learning_rate": 3.955693767632727e-05, + "loss": 0.5433, + "step": 16337 + }, + { + "epoch": 0.63, + "learning_rate": 3.9556296486278534e-05, + "loss": 0.5154, + "step": 16338 + }, + { + "epoch": 0.63, + "learning_rate": 3.9555655296229805e-05, + "loss": 0.5192, + "step": 16339 + }, + { + "epoch": 0.63, + "learning_rate": 3.955501410618108e-05, + "loss": 0.386, + "step": 16340 + }, + { + "epoch": 0.63, + "learning_rate": 3.955437291613234e-05, + "loss": 0.5486, + "step": 16341 + }, + { + "epoch": 0.63, + "learning_rate": 3.9553731726083614e-05, + "loss": 0.4647, + "step": 16342 + }, + { + "epoch": 0.63, + "learning_rate": 3.955309053603488e-05, + "loss": 0.513, + "step": 16343 + }, + { + "epoch": 0.63, + "learning_rate": 3.955244934598615e-05, + "loss": 0.5756, + "step": 16344 + }, + { + "epoch": 0.63, + "learning_rate": 3.955180815593742e-05, + "loss": 0.5343, + "step": 16345 + }, + { + "epoch": 0.63, + "learning_rate": 3.9551166965888694e-05, + "loss": 0.4309, + "step": 16346 + }, + { + "epoch": 0.63, + "learning_rate": 3.955052577583996e-05, + "loss": 0.5576, + "step": 16347 + }, + { + "epoch": 0.63, + "learning_rate": 3.954988458579123e-05, + "loss": 0.5685, + "step": 16348 + }, + { + "epoch": 0.63, + "learning_rate": 3.95492433957425e-05, + "loss": 0.5484, + "step": 16349 + }, + { + "epoch": 0.63, + "learning_rate": 3.954860220569377e-05, + "loss": 0.5671, + "step": 16350 + }, + { + "epoch": 0.63, + "learning_rate": 3.954796101564504e-05, + "loss": 0.564, + "step": 16351 + }, + { + "epoch": 0.63, + "learning_rate": 3.954731982559631e-05, + "loss": 0.5289, + "step": 16352 + }, + { + "epoch": 0.63, + "learning_rate": 3.954667863554758e-05, + "loss": 0.4485, + "step": 16353 + }, + { + "epoch": 0.63, + "learning_rate": 3.954603744549885e-05, + "loss": 0.5039, + "step": 16354 + }, + { + "epoch": 0.63, + "learning_rate": 3.954539625545012e-05, + "loss": 0.5748, + "step": 16355 + }, + { + "epoch": 0.63, + "learning_rate": 3.9544755065401384e-05, + "loss": 0.5348, + "step": 16356 + }, + { + "epoch": 0.63, + "learning_rate": 3.9544113875352656e-05, + "loss": 0.5244, + "step": 16357 + }, + { + "epoch": 0.63, + "learning_rate": 3.954347268530393e-05, + "loss": 0.6146, + "step": 16358 + }, + { + "epoch": 0.63, + "learning_rate": 3.954283149525519e-05, + "loss": 0.5067, + "step": 16359 + }, + { + "epoch": 0.63, + "learning_rate": 3.9542190305206464e-05, + "loss": 0.4949, + "step": 16360 + }, + { + "epoch": 0.63, + "learning_rate": 3.9541549115157736e-05, + "loss": 0.5738, + "step": 16361 + }, + { + "epoch": 0.63, + "learning_rate": 3.954090792510901e-05, + "loss": 0.5018, + "step": 16362 + }, + { + "epoch": 0.63, + "learning_rate": 3.954026673506027e-05, + "loss": 0.5653, + "step": 16363 + }, + { + "epoch": 0.63, + "learning_rate": 3.9539625545011544e-05, + "loss": 0.4609, + "step": 16364 + }, + { + "epoch": 0.63, + "learning_rate": 3.953898435496281e-05, + "loss": 0.6116, + "step": 16365 + }, + { + "epoch": 0.63, + "learning_rate": 3.953834316491408e-05, + "loss": 0.449, + "step": 16366 + }, + { + "epoch": 0.63, + "learning_rate": 3.953770197486535e-05, + "loss": 0.6163, + "step": 16367 + }, + { + "epoch": 0.63, + "learning_rate": 3.9537060784816624e-05, + "loss": 0.6128, + "step": 16368 + }, + { + "epoch": 0.63, + "learning_rate": 3.9536419594767896e-05, + "loss": 0.4954, + "step": 16369 + }, + { + "epoch": 0.63, + "learning_rate": 3.953577840471916e-05, + "loss": 0.4933, + "step": 16370 + }, + { + "epoch": 0.63, + "learning_rate": 3.953513721467043e-05, + "loss": 0.5457, + "step": 16371 + }, + { + "epoch": 0.63, + "learning_rate": 3.95344960246217e-05, + "loss": 0.4836, + "step": 16372 + }, + { + "epoch": 0.63, + "learning_rate": 3.953385483457297e-05, + "loss": 0.5931, + "step": 16373 + }, + { + "epoch": 0.63, + "learning_rate": 3.9533213644524234e-05, + "loss": 0.4345, + "step": 16374 + }, + { + "epoch": 0.63, + "learning_rate": 3.9532572454475506e-05, + "loss": 0.3874, + "step": 16375 + }, + { + "epoch": 0.63, + "learning_rate": 3.953193126442678e-05, + "loss": 0.5218, + "step": 16376 + }, + { + "epoch": 0.63, + "learning_rate": 3.953129007437805e-05, + "loss": 0.5289, + "step": 16377 + }, + { + "epoch": 0.63, + "learning_rate": 3.953064888432932e-05, + "loss": 0.5419, + "step": 16378 + }, + { + "epoch": 0.63, + "learning_rate": 3.9530007694280586e-05, + "loss": 0.4262, + "step": 16379 + }, + { + "epoch": 0.63, + "learning_rate": 3.952936650423186e-05, + "loss": 0.5133, + "step": 16380 + }, + { + "epoch": 0.63, + "learning_rate": 3.952872531418312e-05, + "loss": 0.5087, + "step": 16381 + }, + { + "epoch": 0.63, + "learning_rate": 3.9528084124134394e-05, + "loss": 0.5044, + "step": 16382 + }, + { + "epoch": 0.63, + "learning_rate": 3.952744293408566e-05, + "loss": 0.4642, + "step": 16383 + }, + { + "epoch": 0.63, + "learning_rate": 3.952680174403693e-05, + "loss": 0.4577, + "step": 16384 + }, + { + "epoch": 0.63, + "learning_rate": 3.952616055398821e-05, + "loss": 0.6164, + "step": 16385 + }, + { + "epoch": 0.63, + "learning_rate": 3.9525519363939474e-05, + "loss": 0.5518, + "step": 16386 + }, + { + "epoch": 0.63, + "learning_rate": 3.9524878173890746e-05, + "loss": 0.5769, + "step": 16387 + }, + { + "epoch": 0.63, + "learning_rate": 3.952423698384201e-05, + "loss": 0.5526, + "step": 16388 + }, + { + "epoch": 0.63, + "learning_rate": 3.952359579379328e-05, + "loss": 0.5022, + "step": 16389 + }, + { + "epoch": 0.63, + "learning_rate": 3.952295460374455e-05, + "loss": 0.5557, + "step": 16390 + }, + { + "epoch": 0.63, + "learning_rate": 3.952231341369582e-05, + "loss": 0.5828, + "step": 16391 + }, + { + "epoch": 0.63, + "learning_rate": 3.952167222364709e-05, + "loss": 0.569, + "step": 16392 + }, + { + "epoch": 0.63, + "learning_rate": 3.952103103359836e-05, + "loss": 0.6437, + "step": 16393 + }, + { + "epoch": 0.63, + "learning_rate": 3.9520389843549635e-05, + "loss": 0.534, + "step": 16394 + }, + { + "epoch": 0.63, + "learning_rate": 3.95197486535009e-05, + "loss": 0.6118, + "step": 16395 + }, + { + "epoch": 0.63, + "learning_rate": 3.951910746345217e-05, + "loss": 0.4944, + "step": 16396 + }, + { + "epoch": 0.63, + "learning_rate": 3.9518466273403436e-05, + "loss": 0.3909, + "step": 16397 + }, + { + "epoch": 0.63, + "learning_rate": 3.951782508335471e-05, + "loss": 0.6561, + "step": 16398 + }, + { + "epoch": 0.63, + "learning_rate": 3.951718389330597e-05, + "loss": 0.5157, + "step": 16399 + }, + { + "epoch": 0.63, + "learning_rate": 3.9516542703257245e-05, + "loss": 0.6377, + "step": 16400 + }, + { + "epoch": 0.63, + "learning_rate": 3.9515901513208516e-05, + "loss": 0.4989, + "step": 16401 + }, + { + "epoch": 0.63, + "learning_rate": 3.951526032315979e-05, + "loss": 0.4494, + "step": 16402 + }, + { + "epoch": 0.63, + "learning_rate": 3.951461913311106e-05, + "loss": 0.5813, + "step": 16403 + }, + { + "epoch": 0.63, + "learning_rate": 3.9513977943062325e-05, + "loss": 0.5151, + "step": 16404 + }, + { + "epoch": 0.63, + "learning_rate": 3.9513336753013596e-05, + "loss": 0.5994, + "step": 16405 + }, + { + "epoch": 0.63, + "learning_rate": 3.951269556296486e-05, + "loss": 0.4958, + "step": 16406 + }, + { + "epoch": 0.63, + "learning_rate": 3.951205437291613e-05, + "loss": 0.5425, + "step": 16407 + }, + { + "epoch": 0.63, + "learning_rate": 3.9511413182867405e-05, + "loss": 0.4877, + "step": 16408 + }, + { + "epoch": 0.63, + "learning_rate": 3.9510771992818676e-05, + "loss": 0.6328, + "step": 16409 + }, + { + "epoch": 0.63, + "learning_rate": 3.951013080276994e-05, + "loss": 0.5677, + "step": 16410 + }, + { + "epoch": 0.63, + "learning_rate": 3.950948961272121e-05, + "loss": 0.432, + "step": 16411 + }, + { + "epoch": 0.63, + "learning_rate": 3.9508848422672485e-05, + "loss": 0.4893, + "step": 16412 + }, + { + "epoch": 0.63, + "learning_rate": 3.950820723262375e-05, + "loss": 0.6201, + "step": 16413 + }, + { + "epoch": 0.63, + "learning_rate": 3.950756604257502e-05, + "loss": 0.5101, + "step": 16414 + }, + { + "epoch": 0.63, + "learning_rate": 3.9506924852526286e-05, + "loss": 0.5067, + "step": 16415 + }, + { + "epoch": 0.63, + "learning_rate": 3.950628366247756e-05, + "loss": 0.4717, + "step": 16416 + }, + { + "epoch": 0.63, + "learning_rate": 3.950564247242883e-05, + "loss": 0.5195, + "step": 16417 + }, + { + "epoch": 0.63, + "learning_rate": 3.95050012823801e-05, + "loss": 0.5065, + "step": 16418 + }, + { + "epoch": 0.63, + "learning_rate": 3.9504360092331366e-05, + "loss": 0.458, + "step": 16419 + }, + { + "epoch": 0.63, + "learning_rate": 3.950371890228264e-05, + "loss": 0.4763, + "step": 16420 + }, + { + "epoch": 0.63, + "learning_rate": 3.950307771223391e-05, + "loss": 0.5909, + "step": 16421 + }, + { + "epoch": 0.63, + "learning_rate": 3.9502436522185175e-05, + "loss": 0.5224, + "step": 16422 + }, + { + "epoch": 0.63, + "learning_rate": 3.9501795332136447e-05, + "loss": 0.4391, + "step": 16423 + }, + { + "epoch": 0.63, + "learning_rate": 3.950115414208772e-05, + "loss": 0.4962, + "step": 16424 + }, + { + "epoch": 0.63, + "learning_rate": 3.950051295203899e-05, + "loss": 0.4501, + "step": 16425 + }, + { + "epoch": 0.63, + "learning_rate": 3.9499871761990255e-05, + "loss": 0.5861, + "step": 16426 + }, + { + "epoch": 0.63, + "learning_rate": 3.949923057194153e-05, + "loss": 0.6139, + "step": 16427 + }, + { + "epoch": 0.63, + "learning_rate": 3.94985893818928e-05, + "loss": 0.7407, + "step": 16428 + }, + { + "epoch": 0.63, + "learning_rate": 3.949794819184406e-05, + "loss": 0.5524, + "step": 16429 + }, + { + "epoch": 0.63, + "learning_rate": 3.9497307001795335e-05, + "loss": 0.6109, + "step": 16430 + }, + { + "epoch": 0.63, + "learning_rate": 3.94966658117466e-05, + "loss": 0.5625, + "step": 16431 + }, + { + "epoch": 0.63, + "learning_rate": 3.949602462169787e-05, + "loss": 0.5101, + "step": 16432 + }, + { + "epoch": 0.63, + "learning_rate": 3.9495383431649143e-05, + "loss": 0.451, + "step": 16433 + }, + { + "epoch": 0.63, + "learning_rate": 3.9494742241600415e-05, + "loss": 0.5168, + "step": 16434 + }, + { + "epoch": 0.63, + "learning_rate": 3.949410105155168e-05, + "loss": 0.4635, + "step": 16435 + }, + { + "epoch": 0.63, + "learning_rate": 3.949345986150295e-05, + "loss": 0.5224, + "step": 16436 + }, + { + "epoch": 0.63, + "learning_rate": 3.9492818671454224e-05, + "loss": 0.4593, + "step": 16437 + }, + { + "epoch": 0.63, + "learning_rate": 3.949217748140549e-05, + "loss": 0.6036, + "step": 16438 + }, + { + "epoch": 0.63, + "learning_rate": 3.949153629135676e-05, + "loss": 0.6251, + "step": 16439 + }, + { + "epoch": 0.63, + "learning_rate": 3.949089510130803e-05, + "loss": 0.5497, + "step": 16440 + }, + { + "epoch": 0.63, + "learning_rate": 3.9490253911259304e-05, + "loss": 0.7035, + "step": 16441 + }, + { + "epoch": 0.63, + "learning_rate": 3.948961272121057e-05, + "loss": 0.5321, + "step": 16442 + }, + { + "epoch": 0.63, + "learning_rate": 3.948897153116184e-05, + "loss": 0.5447, + "step": 16443 + }, + { + "epoch": 0.63, + "learning_rate": 3.9488330341113105e-05, + "loss": 0.5165, + "step": 16444 + }, + { + "epoch": 0.63, + "learning_rate": 3.948768915106438e-05, + "loss": 0.4981, + "step": 16445 + }, + { + "epoch": 0.63, + "learning_rate": 3.948704796101565e-05, + "loss": 0.5737, + "step": 16446 + }, + { + "epoch": 0.63, + "learning_rate": 3.9486406770966914e-05, + "loss": 0.4785, + "step": 16447 + }, + { + "epoch": 0.63, + "learning_rate": 3.9485765580918185e-05, + "loss": 0.6231, + "step": 16448 + }, + { + "epoch": 0.63, + "learning_rate": 3.948512439086946e-05, + "loss": 0.5577, + "step": 16449 + }, + { + "epoch": 0.63, + "learning_rate": 3.948448320082073e-05, + "loss": 0.53, + "step": 16450 + }, + { + "epoch": 0.63, + "learning_rate": 3.9483842010771994e-05, + "loss": 0.4176, + "step": 16451 + }, + { + "epoch": 0.63, + "learning_rate": 3.9483200820723265e-05, + "loss": 0.5201, + "step": 16452 + }, + { + "epoch": 0.63, + "learning_rate": 3.948255963067453e-05, + "loss": 0.5429, + "step": 16453 + }, + { + "epoch": 0.63, + "learning_rate": 3.94819184406258e-05, + "loss": 0.5851, + "step": 16454 + }, + { + "epoch": 0.63, + "learning_rate": 3.9481277250577074e-05, + "loss": 0.525, + "step": 16455 + }, + { + "epoch": 0.63, + "learning_rate": 3.9480636060528345e-05, + "loss": 0.5475, + "step": 16456 + }, + { + "epoch": 0.63, + "learning_rate": 3.947999487047962e-05, + "loss": 0.433, + "step": 16457 + }, + { + "epoch": 0.63, + "learning_rate": 3.947935368043088e-05, + "loss": 0.4966, + "step": 16458 + }, + { + "epoch": 0.63, + "learning_rate": 3.9478712490382154e-05, + "loss": 0.4966, + "step": 16459 + }, + { + "epoch": 0.63, + "learning_rate": 3.947807130033342e-05, + "loss": 0.6238, + "step": 16460 + }, + { + "epoch": 0.63, + "learning_rate": 3.947743011028469e-05, + "loss": 0.3166, + "step": 16461 + }, + { + "epoch": 0.63, + "learning_rate": 3.9476788920235955e-05, + "loss": 0.4247, + "step": 16462 + }, + { + "epoch": 0.63, + "learning_rate": 3.947614773018723e-05, + "loss": 0.5269, + "step": 16463 + }, + { + "epoch": 0.63, + "learning_rate": 3.94755065401385e-05, + "loss": 0.4485, + "step": 16464 + }, + { + "epoch": 0.63, + "learning_rate": 3.947486535008977e-05, + "loss": 0.4631, + "step": 16465 + }, + { + "epoch": 0.63, + "learning_rate": 3.947422416004104e-05, + "loss": 0.5694, + "step": 16466 + }, + { + "epoch": 0.63, + "learning_rate": 3.947358296999231e-05, + "loss": 0.6444, + "step": 16467 + }, + { + "epoch": 0.63, + "learning_rate": 3.947294177994358e-05, + "loss": 0.5444, + "step": 16468 + }, + { + "epoch": 0.63, + "learning_rate": 3.9472300589894844e-05, + "loss": 0.5908, + "step": 16469 + }, + { + "epoch": 0.63, + "learning_rate": 3.9471659399846116e-05, + "loss": 0.4155, + "step": 16470 + }, + { + "epoch": 0.63, + "learning_rate": 3.947101820979738e-05, + "loss": 0.5691, + "step": 16471 + }, + { + "epoch": 0.63, + "learning_rate": 3.947037701974865e-05, + "loss": 0.4197, + "step": 16472 + }, + { + "epoch": 0.63, + "learning_rate": 3.946973582969993e-05, + "loss": 0.5229, + "step": 16473 + }, + { + "epoch": 0.63, + "learning_rate": 3.9469094639651196e-05, + "loss": 0.4782, + "step": 16474 + }, + { + "epoch": 0.63, + "learning_rate": 3.946845344960247e-05, + "loss": 0.6443, + "step": 16475 + }, + { + "epoch": 0.63, + "learning_rate": 3.946781225955373e-05, + "loss": 0.492, + "step": 16476 + }, + { + "epoch": 0.63, + "learning_rate": 3.9467171069505004e-05, + "loss": 0.4828, + "step": 16477 + }, + { + "epoch": 0.63, + "learning_rate": 3.946652987945627e-05, + "loss": 0.6264, + "step": 16478 + }, + { + "epoch": 0.63, + "learning_rate": 3.946588868940754e-05, + "loss": 0.5339, + "step": 16479 + }, + { + "epoch": 0.63, + "learning_rate": 3.946524749935881e-05, + "loss": 0.3521, + "step": 16480 + }, + { + "epoch": 0.63, + "learning_rate": 3.9464606309310084e-05, + "loss": 0.4731, + "step": 16481 + }, + { + "epoch": 0.63, + "learning_rate": 3.9463965119261356e-05, + "loss": 0.6522, + "step": 16482 + }, + { + "epoch": 0.63, + "learning_rate": 3.946332392921262e-05, + "loss": 0.4844, + "step": 16483 + }, + { + "epoch": 0.63, + "learning_rate": 3.946268273916389e-05, + "loss": 0.555, + "step": 16484 + }, + { + "epoch": 0.63, + "learning_rate": 3.946204154911516e-05, + "loss": 0.5063, + "step": 16485 + }, + { + "epoch": 0.63, + "learning_rate": 3.946140035906643e-05, + "loss": 0.6291, + "step": 16486 + }, + { + "epoch": 0.63, + "learning_rate": 3.9460759169017694e-05, + "loss": 0.4871, + "step": 16487 + }, + { + "epoch": 0.63, + "learning_rate": 3.9460117978968966e-05, + "loss": 0.5745, + "step": 16488 + }, + { + "epoch": 0.63, + "learning_rate": 3.945947678892024e-05, + "loss": 0.5199, + "step": 16489 + }, + { + "epoch": 0.63, + "learning_rate": 3.945883559887151e-05, + "loss": 0.5862, + "step": 16490 + }, + { + "epoch": 0.63, + "learning_rate": 3.945819440882278e-05, + "loss": 0.5183, + "step": 16491 + }, + { + "epoch": 0.63, + "learning_rate": 3.9457553218774046e-05, + "loss": 0.5484, + "step": 16492 + }, + { + "epoch": 0.63, + "learning_rate": 3.945691202872532e-05, + "loss": 0.5882, + "step": 16493 + }, + { + "epoch": 0.63, + "learning_rate": 3.945627083867658e-05, + "loss": 0.515, + "step": 16494 + }, + { + "epoch": 0.63, + "learning_rate": 3.9455629648627854e-05, + "loss": 0.4819, + "step": 16495 + }, + { + "epoch": 0.63, + "learning_rate": 3.9454988458579126e-05, + "loss": 0.5943, + "step": 16496 + }, + { + "epoch": 0.63, + "learning_rate": 3.94543472685304e-05, + "loss": 0.5596, + "step": 16497 + }, + { + "epoch": 0.63, + "learning_rate": 3.945370607848166e-05, + "loss": 0.5989, + "step": 16498 + }, + { + "epoch": 0.63, + "learning_rate": 3.9453064888432934e-05, + "loss": 0.6612, + "step": 16499 + }, + { + "epoch": 0.63, + "learning_rate": 3.9452423698384206e-05, + "loss": 0.4442, + "step": 16500 + }, + { + "epoch": 0.63, + "learning_rate": 3.945178250833547e-05, + "loss": 0.4706, + "step": 16501 + }, + { + "epoch": 0.63, + "learning_rate": 3.945114131828674e-05, + "loss": 0.5202, + "step": 16502 + }, + { + "epoch": 0.63, + "learning_rate": 3.945050012823801e-05, + "loss": 0.5141, + "step": 16503 + }, + { + "epoch": 0.63, + "learning_rate": 3.944985893818928e-05, + "loss": 0.5239, + "step": 16504 + }, + { + "epoch": 0.63, + "learning_rate": 3.944921774814055e-05, + "loss": 0.6163, + "step": 16505 + }, + { + "epoch": 0.63, + "learning_rate": 3.944857655809182e-05, + "loss": 0.5332, + "step": 16506 + }, + { + "epoch": 0.63, + "learning_rate": 3.944793536804309e-05, + "loss": 0.6222, + "step": 16507 + }, + { + "epoch": 0.63, + "learning_rate": 3.944729417799436e-05, + "loss": 0.5424, + "step": 16508 + }, + { + "epoch": 0.63, + "learning_rate": 3.944665298794563e-05, + "loss": 0.5291, + "step": 16509 + }, + { + "epoch": 0.63, + "learning_rate": 3.9446011797896896e-05, + "loss": 0.5171, + "step": 16510 + }, + { + "epoch": 0.63, + "learning_rate": 3.944537060784817e-05, + "loss": 0.6157, + "step": 16511 + }, + { + "epoch": 0.63, + "learning_rate": 3.944472941779944e-05, + "loss": 0.5981, + "step": 16512 + }, + { + "epoch": 0.63, + "learning_rate": 3.944408822775071e-05, + "loss": 0.49, + "step": 16513 + }, + { + "epoch": 0.63, + "learning_rate": 3.9443447037701976e-05, + "loss": 0.4901, + "step": 16514 + }, + { + "epoch": 0.63, + "learning_rate": 3.944280584765325e-05, + "loss": 0.6013, + "step": 16515 + }, + { + "epoch": 0.63, + "learning_rate": 3.944216465760451e-05, + "loss": 0.5557, + "step": 16516 + }, + { + "epoch": 0.64, + "learning_rate": 3.9441523467555785e-05, + "loss": 0.5645, + "step": 16517 + }, + { + "epoch": 0.64, + "learning_rate": 3.9440882277507056e-05, + "loss": 0.5855, + "step": 16518 + }, + { + "epoch": 0.64, + "learning_rate": 3.944024108745832e-05, + "loss": 0.5773, + "step": 16519 + }, + { + "epoch": 0.64, + "learning_rate": 3.943959989740959e-05, + "loss": 0.5126, + "step": 16520 + }, + { + "epoch": 0.64, + "learning_rate": 3.9438958707360865e-05, + "loss": 0.4402, + "step": 16521 + }, + { + "epoch": 0.64, + "learning_rate": 3.9438317517312136e-05, + "loss": 0.5316, + "step": 16522 + }, + { + "epoch": 0.64, + "learning_rate": 3.94376763272634e-05, + "loss": 0.471, + "step": 16523 + }, + { + "epoch": 0.64, + "learning_rate": 3.943703513721467e-05, + "loss": 0.5523, + "step": 16524 + }, + { + "epoch": 0.64, + "learning_rate": 3.943639394716594e-05, + "loss": 0.4237, + "step": 16525 + }, + { + "epoch": 0.64, + "learning_rate": 3.943575275711721e-05, + "loss": 0.5676, + "step": 16526 + }, + { + "epoch": 0.64, + "learning_rate": 3.943511156706848e-05, + "loss": 0.6445, + "step": 16527 + }, + { + "epoch": 0.64, + "learning_rate": 3.943447037701975e-05, + "loss": 0.4736, + "step": 16528 + }, + { + "epoch": 0.64, + "learning_rate": 3.9433829186971025e-05, + "loss": 0.4448, + "step": 16529 + }, + { + "epoch": 0.64, + "learning_rate": 3.943318799692229e-05, + "loss": 0.5654, + "step": 16530 + }, + { + "epoch": 0.64, + "learning_rate": 3.943254680687356e-05, + "loss": 0.6123, + "step": 16531 + }, + { + "epoch": 0.64, + "learning_rate": 3.9431905616824827e-05, + "loss": 0.6288, + "step": 16532 + }, + { + "epoch": 0.64, + "learning_rate": 3.94312644267761e-05, + "loss": 0.5234, + "step": 16533 + }, + { + "epoch": 0.64, + "learning_rate": 3.943062323672736e-05, + "loss": 0.5734, + "step": 16534 + }, + { + "epoch": 0.64, + "learning_rate": 3.9429982046678635e-05, + "loss": 0.5796, + "step": 16535 + }, + { + "epoch": 0.64, + "learning_rate": 3.9429340856629907e-05, + "loss": 0.5582, + "step": 16536 + }, + { + "epoch": 0.64, + "learning_rate": 3.942869966658118e-05, + "loss": 0.5447, + "step": 16537 + }, + { + "epoch": 0.64, + "learning_rate": 3.942805847653245e-05, + "loss": 0.5367, + "step": 16538 + }, + { + "epoch": 0.64, + "learning_rate": 3.9427417286483715e-05, + "loss": 0.5144, + "step": 16539 + }, + { + "epoch": 0.64, + "learning_rate": 3.942677609643499e-05, + "loss": 0.6543, + "step": 16540 + }, + { + "epoch": 0.64, + "learning_rate": 3.942613490638625e-05, + "loss": 0.437, + "step": 16541 + }, + { + "epoch": 0.64, + "learning_rate": 3.942549371633752e-05, + "loss": 0.6325, + "step": 16542 + }, + { + "epoch": 0.64, + "learning_rate": 3.942485252628879e-05, + "loss": 0.5004, + "step": 16543 + }, + { + "epoch": 0.64, + "learning_rate": 3.942421133624006e-05, + "loss": 0.5227, + "step": 16544 + }, + { + "epoch": 0.64, + "learning_rate": 3.942357014619134e-05, + "loss": 0.4431, + "step": 16545 + }, + { + "epoch": 0.64, + "learning_rate": 3.9422928956142603e-05, + "loss": 0.4095, + "step": 16546 + }, + { + "epoch": 0.64, + "learning_rate": 3.9422287766093875e-05, + "loss": 0.4851, + "step": 16547 + }, + { + "epoch": 0.64, + "learning_rate": 3.942164657604514e-05, + "loss": 0.6195, + "step": 16548 + }, + { + "epoch": 0.64, + "learning_rate": 3.942100538599641e-05, + "loss": 0.5993, + "step": 16549 + }, + { + "epoch": 0.64, + "learning_rate": 3.942036419594768e-05, + "loss": 0.4589, + "step": 16550 + }, + { + "epoch": 0.64, + "learning_rate": 3.941972300589895e-05, + "loss": 0.592, + "step": 16551 + }, + { + "epoch": 0.64, + "learning_rate": 3.941908181585022e-05, + "loss": 0.6095, + "step": 16552 + }, + { + "epoch": 0.64, + "learning_rate": 3.941844062580149e-05, + "loss": 0.5099, + "step": 16553 + }, + { + "epoch": 0.64, + "learning_rate": 3.9417799435752764e-05, + "loss": 0.6451, + "step": 16554 + }, + { + "epoch": 0.64, + "learning_rate": 3.941715824570403e-05, + "loss": 0.5402, + "step": 16555 + }, + { + "epoch": 0.64, + "learning_rate": 3.94165170556553e-05, + "loss": 0.5049, + "step": 16556 + }, + { + "epoch": 0.64, + "learning_rate": 3.9415875865606565e-05, + "loss": 0.5547, + "step": 16557 + }, + { + "epoch": 0.64, + "learning_rate": 3.941523467555784e-05, + "loss": 0.4602, + "step": 16558 + }, + { + "epoch": 0.64, + "learning_rate": 3.94145934855091e-05, + "loss": 0.5528, + "step": 16559 + }, + { + "epoch": 0.64, + "learning_rate": 3.9413952295460374e-05, + "loss": 0.4882, + "step": 16560 + }, + { + "epoch": 0.64, + "learning_rate": 3.9413311105411645e-05, + "loss": 0.5626, + "step": 16561 + }, + { + "epoch": 0.64, + "learning_rate": 3.941266991536292e-05, + "loss": 0.5917, + "step": 16562 + }, + { + "epoch": 0.64, + "learning_rate": 3.941202872531419e-05, + "loss": 0.4952, + "step": 16563 + }, + { + "epoch": 0.64, + "learning_rate": 3.9411387535265454e-05, + "loss": 0.4943, + "step": 16564 + }, + { + "epoch": 0.64, + "learning_rate": 3.9410746345216725e-05, + "loss": 0.4498, + "step": 16565 + }, + { + "epoch": 0.64, + "learning_rate": 3.941010515516799e-05, + "loss": 0.5167, + "step": 16566 + }, + { + "epoch": 0.64, + "learning_rate": 3.940946396511926e-05, + "loss": 0.4649, + "step": 16567 + }, + { + "epoch": 0.64, + "learning_rate": 3.9408822775070534e-05, + "loss": 0.6277, + "step": 16568 + }, + { + "epoch": 0.64, + "learning_rate": 3.9408181585021805e-05, + "loss": 0.5787, + "step": 16569 + }, + { + "epoch": 0.64, + "learning_rate": 3.940754039497307e-05, + "loss": 0.4566, + "step": 16570 + }, + { + "epoch": 0.64, + "learning_rate": 3.940689920492434e-05, + "loss": 0.4664, + "step": 16571 + }, + { + "epoch": 0.64, + "learning_rate": 3.9406258014875614e-05, + "loss": 0.4307, + "step": 16572 + }, + { + "epoch": 0.64, + "learning_rate": 3.940561682482688e-05, + "loss": 0.6261, + "step": 16573 + }, + { + "epoch": 0.64, + "learning_rate": 3.940497563477815e-05, + "loss": 0.5564, + "step": 16574 + }, + { + "epoch": 0.64, + "learning_rate": 3.9404334444729415e-05, + "loss": 0.4771, + "step": 16575 + }, + { + "epoch": 0.64, + "learning_rate": 3.940369325468069e-05, + "loss": 0.416, + "step": 16576 + }, + { + "epoch": 0.64, + "learning_rate": 3.940305206463196e-05, + "loss": 0.4927, + "step": 16577 + }, + { + "epoch": 0.64, + "learning_rate": 3.940241087458323e-05, + "loss": 0.5925, + "step": 16578 + }, + { + "epoch": 0.64, + "learning_rate": 3.9401769684534496e-05, + "loss": 0.4712, + "step": 16579 + }, + { + "epoch": 0.64, + "learning_rate": 3.940112849448577e-05, + "loss": 0.5082, + "step": 16580 + }, + { + "epoch": 0.64, + "learning_rate": 3.940048730443704e-05, + "loss": 0.493, + "step": 16581 + }, + { + "epoch": 0.64, + "learning_rate": 3.9399846114388304e-05, + "loss": 0.5534, + "step": 16582 + }, + { + "epoch": 0.64, + "learning_rate": 3.9399204924339576e-05, + "loss": 0.5088, + "step": 16583 + }, + { + "epoch": 0.64, + "learning_rate": 3.939856373429085e-05, + "loss": 0.5574, + "step": 16584 + }, + { + "epoch": 0.64, + "learning_rate": 3.939792254424212e-05, + "loss": 0.4111, + "step": 16585 + }, + { + "epoch": 0.64, + "learning_rate": 3.9397281354193384e-05, + "loss": 0.5003, + "step": 16586 + }, + { + "epoch": 0.64, + "learning_rate": 3.9396640164144656e-05, + "loss": 0.4524, + "step": 16587 + }, + { + "epoch": 0.64, + "learning_rate": 3.939599897409592e-05, + "loss": 0.515, + "step": 16588 + }, + { + "epoch": 0.64, + "learning_rate": 3.939535778404719e-05, + "loss": 0.5155, + "step": 16589 + }, + { + "epoch": 0.64, + "learning_rate": 3.9394716593998464e-05, + "loss": 0.5867, + "step": 16590 + }, + { + "epoch": 0.64, + "learning_rate": 3.939407540394973e-05, + "loss": 0.4961, + "step": 16591 + }, + { + "epoch": 0.64, + "learning_rate": 3.9393434213901e-05, + "loss": 0.5865, + "step": 16592 + }, + { + "epoch": 0.64, + "learning_rate": 3.939279302385227e-05, + "loss": 0.5116, + "step": 16593 + }, + { + "epoch": 0.64, + "learning_rate": 3.9392151833803544e-05, + "loss": 0.5187, + "step": 16594 + }, + { + "epoch": 0.64, + "learning_rate": 3.939151064375481e-05, + "loss": 0.569, + "step": 16595 + }, + { + "epoch": 0.64, + "learning_rate": 3.939086945370608e-05, + "loss": 0.6125, + "step": 16596 + }, + { + "epoch": 0.64, + "learning_rate": 3.9390228263657346e-05, + "loss": 0.5157, + "step": 16597 + }, + { + "epoch": 0.64, + "learning_rate": 3.938958707360862e-05, + "loss": 0.4514, + "step": 16598 + }, + { + "epoch": 0.64, + "learning_rate": 3.938894588355989e-05, + "loss": 0.5609, + "step": 16599 + }, + { + "epoch": 0.64, + "learning_rate": 3.938830469351116e-05, + "loss": 0.517, + "step": 16600 + }, + { + "epoch": 0.64, + "learning_rate": 3.938766350346243e-05, + "loss": 0.4667, + "step": 16601 + }, + { + "epoch": 0.64, + "learning_rate": 3.93870223134137e-05, + "loss": 0.5336, + "step": 16602 + }, + { + "epoch": 0.64, + "learning_rate": 3.938638112336497e-05, + "loss": 0.529, + "step": 16603 + }, + { + "epoch": 0.64, + "learning_rate": 3.9385739933316234e-05, + "loss": 0.4986, + "step": 16604 + }, + { + "epoch": 0.64, + "learning_rate": 3.9385098743267506e-05, + "loss": 0.438, + "step": 16605 + }, + { + "epoch": 0.64, + "learning_rate": 3.938445755321877e-05, + "loss": 0.5506, + "step": 16606 + }, + { + "epoch": 0.64, + "learning_rate": 3.938381636317004e-05, + "loss": 0.4483, + "step": 16607 + }, + { + "epoch": 0.64, + "learning_rate": 3.9383175173121314e-05, + "loss": 0.5787, + "step": 16608 + }, + { + "epoch": 0.64, + "learning_rate": 3.9382533983072586e-05, + "loss": 0.4792, + "step": 16609 + }, + { + "epoch": 0.64, + "learning_rate": 3.938189279302386e-05, + "loss": 0.5116, + "step": 16610 + }, + { + "epoch": 0.64, + "learning_rate": 3.938125160297512e-05, + "loss": 0.5856, + "step": 16611 + }, + { + "epoch": 0.64, + "learning_rate": 3.9380610412926394e-05, + "loss": 0.689, + "step": 16612 + }, + { + "epoch": 0.64, + "learning_rate": 3.937996922287766e-05, + "loss": 0.4858, + "step": 16613 + }, + { + "epoch": 0.64, + "learning_rate": 3.937932803282893e-05, + "loss": 0.505, + "step": 16614 + }, + { + "epoch": 0.64, + "learning_rate": 3.93786868427802e-05, + "loss": 0.6311, + "step": 16615 + }, + { + "epoch": 0.64, + "learning_rate": 3.9378045652731475e-05, + "loss": 0.5274, + "step": 16616 + }, + { + "epoch": 0.64, + "learning_rate": 3.9377404462682746e-05, + "loss": 0.5994, + "step": 16617 + }, + { + "epoch": 0.64, + "learning_rate": 3.937676327263401e-05, + "loss": 0.5364, + "step": 16618 + }, + { + "epoch": 0.64, + "learning_rate": 3.937612208258528e-05, + "loss": 0.5762, + "step": 16619 + }, + { + "epoch": 0.64, + "learning_rate": 3.937548089253655e-05, + "loss": 0.5408, + "step": 16620 + }, + { + "epoch": 0.64, + "learning_rate": 3.937483970248782e-05, + "loss": 0.5444, + "step": 16621 + }, + { + "epoch": 0.64, + "learning_rate": 3.9374198512439085e-05, + "loss": 0.4689, + "step": 16622 + }, + { + "epoch": 0.64, + "learning_rate": 3.9373557322390356e-05, + "loss": 0.4639, + "step": 16623 + }, + { + "epoch": 0.64, + "learning_rate": 3.937291613234163e-05, + "loss": 0.5444, + "step": 16624 + }, + { + "epoch": 0.64, + "learning_rate": 3.93722749422929e-05, + "loss": 0.5582, + "step": 16625 + }, + { + "epoch": 0.64, + "learning_rate": 3.937163375224417e-05, + "loss": 0.5582, + "step": 16626 + }, + { + "epoch": 0.64, + "learning_rate": 3.9370992562195436e-05, + "loss": 0.4368, + "step": 16627 + }, + { + "epoch": 0.64, + "learning_rate": 3.937035137214671e-05, + "loss": 0.5408, + "step": 16628 + }, + { + "epoch": 0.64, + "learning_rate": 3.936971018209797e-05, + "loss": 0.5154, + "step": 16629 + }, + { + "epoch": 0.64, + "learning_rate": 3.9369068992049245e-05, + "loss": 0.4541, + "step": 16630 + }, + { + "epoch": 0.64, + "learning_rate": 3.936842780200051e-05, + "loss": 0.6084, + "step": 16631 + }, + { + "epoch": 0.64, + "learning_rate": 3.936778661195178e-05, + "loss": 0.4026, + "step": 16632 + }, + { + "epoch": 0.64, + "learning_rate": 3.936714542190306e-05, + "loss": 0.4245, + "step": 16633 + }, + { + "epoch": 0.64, + "learning_rate": 3.9366504231854325e-05, + "loss": 0.5518, + "step": 16634 + }, + { + "epoch": 0.64, + "learning_rate": 3.9365863041805596e-05, + "loss": 0.424, + "step": 16635 + }, + { + "epoch": 0.64, + "learning_rate": 3.936522185175686e-05, + "loss": 0.5981, + "step": 16636 + }, + { + "epoch": 0.64, + "learning_rate": 3.936458066170813e-05, + "loss": 0.6631, + "step": 16637 + }, + { + "epoch": 0.64, + "learning_rate": 3.93639394716594e-05, + "loss": 0.5858, + "step": 16638 + }, + { + "epoch": 0.64, + "learning_rate": 3.936329828161067e-05, + "loss": 0.5734, + "step": 16639 + }, + { + "epoch": 0.64, + "learning_rate": 3.936265709156194e-05, + "loss": 0.5155, + "step": 16640 + }, + { + "epoch": 0.64, + "learning_rate": 3.936201590151321e-05, + "loss": 0.52, + "step": 16641 + }, + { + "epoch": 0.64, + "learning_rate": 3.9361374711464485e-05, + "loss": 0.4926, + "step": 16642 + }, + { + "epoch": 0.64, + "learning_rate": 3.936073352141575e-05, + "loss": 0.5235, + "step": 16643 + }, + { + "epoch": 0.64, + "learning_rate": 3.936009233136702e-05, + "loss": 0.4974, + "step": 16644 + }, + { + "epoch": 0.64, + "learning_rate": 3.9359451141318287e-05, + "loss": 0.4881, + "step": 16645 + }, + { + "epoch": 0.64, + "learning_rate": 3.935880995126956e-05, + "loss": 0.6063, + "step": 16646 + }, + { + "epoch": 0.64, + "learning_rate": 3.935816876122082e-05, + "loss": 0.4382, + "step": 16647 + }, + { + "epoch": 0.64, + "learning_rate": 3.9357527571172095e-05, + "loss": 0.5107, + "step": 16648 + }, + { + "epoch": 0.64, + "learning_rate": 3.935688638112337e-05, + "loss": 0.5384, + "step": 16649 + }, + { + "epoch": 0.64, + "learning_rate": 3.935624519107464e-05, + "loss": 0.5668, + "step": 16650 + }, + { + "epoch": 0.64, + "learning_rate": 3.935560400102591e-05, + "loss": 0.4973, + "step": 16651 + }, + { + "epoch": 0.64, + "learning_rate": 3.9354962810977175e-05, + "loss": 0.4784, + "step": 16652 + }, + { + "epoch": 0.64, + "learning_rate": 3.935432162092845e-05, + "loss": 0.4774, + "step": 16653 + }, + { + "epoch": 0.64, + "learning_rate": 3.935368043087971e-05, + "loss": 0.5528, + "step": 16654 + }, + { + "epoch": 0.64, + "learning_rate": 3.9353039240830983e-05, + "loss": 0.4847, + "step": 16655 + }, + { + "epoch": 0.64, + "learning_rate": 3.9352398050782255e-05, + "loss": 0.5845, + "step": 16656 + }, + { + "epoch": 0.64, + "learning_rate": 3.935175686073353e-05, + "loss": 0.5521, + "step": 16657 + }, + { + "epoch": 0.64, + "learning_rate": 3.935111567068479e-05, + "loss": 0.5119, + "step": 16658 + }, + { + "epoch": 0.64, + "learning_rate": 3.9350474480636063e-05, + "loss": 0.4855, + "step": 16659 + }, + { + "epoch": 0.64, + "learning_rate": 3.9349833290587335e-05, + "loss": 0.4962, + "step": 16660 + }, + { + "epoch": 0.64, + "learning_rate": 3.93491921005386e-05, + "loss": 0.4944, + "step": 16661 + }, + { + "epoch": 0.64, + "learning_rate": 3.934855091048987e-05, + "loss": 0.5119, + "step": 16662 + }, + { + "epoch": 0.64, + "learning_rate": 3.934790972044114e-05, + "loss": 0.5448, + "step": 16663 + }, + { + "epoch": 0.64, + "learning_rate": 3.934726853039241e-05, + "loss": 0.5398, + "step": 16664 + }, + { + "epoch": 0.64, + "learning_rate": 3.934662734034368e-05, + "loss": 0.5341, + "step": 16665 + }, + { + "epoch": 0.64, + "learning_rate": 3.934598615029495e-05, + "loss": 0.5241, + "step": 16666 + }, + { + "epoch": 0.64, + "learning_rate": 3.934534496024622e-05, + "loss": 0.4612, + "step": 16667 + }, + { + "epoch": 0.64, + "learning_rate": 3.934470377019749e-05, + "loss": 0.5125, + "step": 16668 + }, + { + "epoch": 0.64, + "learning_rate": 3.934406258014876e-05, + "loss": 0.4056, + "step": 16669 + }, + { + "epoch": 0.64, + "learning_rate": 3.9343421390100025e-05, + "loss": 0.5999, + "step": 16670 + }, + { + "epoch": 0.64, + "learning_rate": 3.93427802000513e-05, + "loss": 0.484, + "step": 16671 + }, + { + "epoch": 0.64, + "learning_rate": 3.934213901000257e-05, + "loss": 0.5045, + "step": 16672 + }, + { + "epoch": 0.64, + "learning_rate": 3.934149781995384e-05, + "loss": 0.6612, + "step": 16673 + }, + { + "epoch": 0.64, + "learning_rate": 3.9340856629905105e-05, + "loss": 0.4757, + "step": 16674 + }, + { + "epoch": 0.64, + "learning_rate": 3.934021543985638e-05, + "loss": 0.5283, + "step": 16675 + }, + { + "epoch": 0.64, + "learning_rate": 3.933957424980764e-05, + "loss": 0.5068, + "step": 16676 + }, + { + "epoch": 0.64, + "learning_rate": 3.9338933059758914e-05, + "loss": 0.5029, + "step": 16677 + }, + { + "epoch": 0.64, + "learning_rate": 3.9338291869710185e-05, + "loss": 0.6073, + "step": 16678 + }, + { + "epoch": 0.64, + "learning_rate": 3.933765067966145e-05, + "loss": 0.5757, + "step": 16679 + }, + { + "epoch": 0.64, + "learning_rate": 3.933700948961272e-05, + "loss": 0.5624, + "step": 16680 + }, + { + "epoch": 0.64, + "learning_rate": 3.9336368299563994e-05, + "loss": 0.685, + "step": 16681 + }, + { + "epoch": 0.64, + "learning_rate": 3.9335727109515266e-05, + "loss": 0.5548, + "step": 16682 + }, + { + "epoch": 0.64, + "learning_rate": 3.933508591946653e-05, + "loss": 0.5274, + "step": 16683 + }, + { + "epoch": 0.64, + "learning_rate": 3.93344447294178e-05, + "loss": 0.5466, + "step": 16684 + }, + { + "epoch": 0.64, + "learning_rate": 3.933380353936907e-05, + "loss": 0.421, + "step": 16685 + }, + { + "epoch": 0.64, + "learning_rate": 3.933316234932034e-05, + "loss": 0.645, + "step": 16686 + }, + { + "epoch": 0.64, + "learning_rate": 3.933252115927161e-05, + "loss": 0.4927, + "step": 16687 + }, + { + "epoch": 0.64, + "learning_rate": 3.933187996922288e-05, + "loss": 0.4892, + "step": 16688 + }, + { + "epoch": 0.64, + "learning_rate": 3.9331238779174154e-05, + "loss": 0.5418, + "step": 16689 + }, + { + "epoch": 0.64, + "learning_rate": 3.933059758912542e-05, + "loss": 0.6399, + "step": 16690 + }, + { + "epoch": 0.64, + "learning_rate": 3.932995639907669e-05, + "loss": 0.5261, + "step": 16691 + }, + { + "epoch": 0.64, + "learning_rate": 3.9329315209027956e-05, + "loss": 0.6259, + "step": 16692 + }, + { + "epoch": 0.64, + "learning_rate": 3.932867401897923e-05, + "loss": 0.5264, + "step": 16693 + }, + { + "epoch": 0.64, + "learning_rate": 3.932803282893049e-05, + "loss": 0.5281, + "step": 16694 + }, + { + "epoch": 0.64, + "learning_rate": 3.9327391638881764e-05, + "loss": 0.5507, + "step": 16695 + }, + { + "epoch": 0.64, + "learning_rate": 3.9326750448833036e-05, + "loss": 0.5035, + "step": 16696 + }, + { + "epoch": 0.64, + "learning_rate": 3.932610925878431e-05, + "loss": 0.5181, + "step": 16697 + }, + { + "epoch": 0.64, + "learning_rate": 3.932546806873558e-05, + "loss": 0.4544, + "step": 16698 + }, + { + "epoch": 0.64, + "learning_rate": 3.9324826878686844e-05, + "loss": 0.4701, + "step": 16699 + }, + { + "epoch": 0.64, + "learning_rate": 3.9324185688638116e-05, + "loss": 0.5006, + "step": 16700 + }, + { + "epoch": 0.64, + "learning_rate": 3.932354449858938e-05, + "loss": 0.4679, + "step": 16701 + }, + { + "epoch": 0.64, + "learning_rate": 3.932290330854065e-05, + "loss": 0.5187, + "step": 16702 + }, + { + "epoch": 0.64, + "learning_rate": 3.932226211849192e-05, + "loss": 0.5946, + "step": 16703 + }, + { + "epoch": 0.64, + "learning_rate": 3.932162092844319e-05, + "loss": 0.6074, + "step": 16704 + }, + { + "epoch": 0.64, + "learning_rate": 3.932097973839447e-05, + "loss": 0.5016, + "step": 16705 + }, + { + "epoch": 0.64, + "learning_rate": 3.932033854834573e-05, + "loss": 0.4696, + "step": 16706 + }, + { + "epoch": 0.64, + "learning_rate": 3.9319697358297004e-05, + "loss": 0.5769, + "step": 16707 + }, + { + "epoch": 0.64, + "learning_rate": 3.931905616824827e-05, + "loss": 0.4161, + "step": 16708 + }, + { + "epoch": 0.64, + "learning_rate": 3.931841497819954e-05, + "loss": 0.4854, + "step": 16709 + }, + { + "epoch": 0.64, + "learning_rate": 3.9317773788150806e-05, + "loss": 0.4792, + "step": 16710 + }, + { + "epoch": 0.64, + "learning_rate": 3.931713259810208e-05, + "loss": 0.5376, + "step": 16711 + }, + { + "epoch": 0.64, + "learning_rate": 3.931649140805335e-05, + "loss": 0.6037, + "step": 16712 + }, + { + "epoch": 0.64, + "learning_rate": 3.931585021800462e-05, + "loss": 0.5094, + "step": 16713 + }, + { + "epoch": 0.64, + "learning_rate": 3.931520902795589e-05, + "loss": 0.559, + "step": 16714 + }, + { + "epoch": 0.64, + "learning_rate": 3.931456783790716e-05, + "loss": 0.47, + "step": 16715 + }, + { + "epoch": 0.64, + "learning_rate": 3.931392664785843e-05, + "loss": 0.5092, + "step": 16716 + }, + { + "epoch": 0.64, + "learning_rate": 3.9313285457809694e-05, + "loss": 0.5367, + "step": 16717 + }, + { + "epoch": 0.64, + "learning_rate": 3.9312644267760966e-05, + "loss": 0.5795, + "step": 16718 + }, + { + "epoch": 0.64, + "learning_rate": 3.931200307771223e-05, + "loss": 0.5909, + "step": 16719 + }, + { + "epoch": 0.64, + "learning_rate": 3.93113618876635e-05, + "loss": 0.5546, + "step": 16720 + }, + { + "epoch": 0.64, + "learning_rate": 3.9310720697614774e-05, + "loss": 0.6774, + "step": 16721 + }, + { + "epoch": 0.64, + "learning_rate": 3.9310079507566046e-05, + "loss": 0.5758, + "step": 16722 + }, + { + "epoch": 0.64, + "learning_rate": 3.930943831751732e-05, + "loss": 0.32, + "step": 16723 + }, + { + "epoch": 0.64, + "learning_rate": 3.930879712746858e-05, + "loss": 0.4976, + "step": 16724 + }, + { + "epoch": 0.64, + "learning_rate": 3.9308155937419854e-05, + "loss": 0.4816, + "step": 16725 + }, + { + "epoch": 0.64, + "learning_rate": 3.930751474737112e-05, + "loss": 0.5098, + "step": 16726 + }, + { + "epoch": 0.64, + "learning_rate": 3.930687355732239e-05, + "loss": 0.4741, + "step": 16727 + }, + { + "epoch": 0.64, + "learning_rate": 3.930623236727366e-05, + "loss": 0.535, + "step": 16728 + }, + { + "epoch": 0.64, + "learning_rate": 3.9305591177224935e-05, + "loss": 0.5895, + "step": 16729 + }, + { + "epoch": 0.64, + "learning_rate": 3.93049499871762e-05, + "loss": 0.4564, + "step": 16730 + }, + { + "epoch": 0.64, + "learning_rate": 3.930430879712747e-05, + "loss": 0.5468, + "step": 16731 + }, + { + "epoch": 0.64, + "learning_rate": 3.930366760707874e-05, + "loss": 0.5092, + "step": 16732 + }, + { + "epoch": 0.64, + "learning_rate": 3.930302641703001e-05, + "loss": 0.4819, + "step": 16733 + }, + { + "epoch": 0.64, + "learning_rate": 3.930238522698128e-05, + "loss": 0.5898, + "step": 16734 + }, + { + "epoch": 0.64, + "learning_rate": 3.9301744036932545e-05, + "loss": 0.7092, + "step": 16735 + }, + { + "epoch": 0.64, + "learning_rate": 3.9301102846883816e-05, + "loss": 0.5236, + "step": 16736 + }, + { + "epoch": 0.64, + "learning_rate": 3.930046165683509e-05, + "loss": 0.4927, + "step": 16737 + }, + { + "epoch": 0.64, + "learning_rate": 3.929982046678636e-05, + "loss": 0.5247, + "step": 16738 + }, + { + "epoch": 0.64, + "learning_rate": 3.9299179276737625e-05, + "loss": 0.6105, + "step": 16739 + }, + { + "epoch": 0.64, + "learning_rate": 3.9298538086688896e-05, + "loss": 0.5805, + "step": 16740 + }, + { + "epoch": 0.64, + "learning_rate": 3.929789689664017e-05, + "loss": 0.4892, + "step": 16741 + }, + { + "epoch": 0.64, + "learning_rate": 3.929725570659143e-05, + "loss": 0.5344, + "step": 16742 + }, + { + "epoch": 0.64, + "learning_rate": 3.9296614516542705e-05, + "loss": 0.578, + "step": 16743 + }, + { + "epoch": 0.64, + "learning_rate": 3.9295973326493976e-05, + "loss": 0.6221, + "step": 16744 + }, + { + "epoch": 0.64, + "learning_rate": 3.929533213644525e-05, + "loss": 0.6342, + "step": 16745 + }, + { + "epoch": 0.64, + "learning_rate": 3.929469094639651e-05, + "loss": 0.49, + "step": 16746 + }, + { + "epoch": 0.64, + "learning_rate": 3.9294049756347785e-05, + "loss": 0.5761, + "step": 16747 + }, + { + "epoch": 0.64, + "learning_rate": 3.929340856629905e-05, + "loss": 0.6066, + "step": 16748 + }, + { + "epoch": 0.64, + "learning_rate": 3.929276737625032e-05, + "loss": 0.4375, + "step": 16749 + }, + { + "epoch": 0.64, + "learning_rate": 3.929212618620159e-05, + "loss": 0.6229, + "step": 16750 + }, + { + "epoch": 0.64, + "learning_rate": 3.929148499615286e-05, + "loss": 0.482, + "step": 16751 + }, + { + "epoch": 0.64, + "learning_rate": 3.929084380610413e-05, + "loss": 0.5925, + "step": 16752 + }, + { + "epoch": 0.64, + "learning_rate": 3.92902026160554e-05, + "loss": 0.4467, + "step": 16753 + }, + { + "epoch": 0.64, + "learning_rate": 3.928956142600667e-05, + "loss": 0.5649, + "step": 16754 + }, + { + "epoch": 0.64, + "learning_rate": 3.928892023595794e-05, + "loss": 0.4101, + "step": 16755 + }, + { + "epoch": 0.64, + "learning_rate": 3.928827904590921e-05, + "loss": 0.5833, + "step": 16756 + }, + { + "epoch": 0.64, + "learning_rate": 3.9287637855860475e-05, + "loss": 0.6071, + "step": 16757 + }, + { + "epoch": 0.64, + "learning_rate": 3.9286996665811747e-05, + "loss": 0.5784, + "step": 16758 + }, + { + "epoch": 0.64, + "learning_rate": 3.928635547576302e-05, + "loss": 0.5506, + "step": 16759 + }, + { + "epoch": 0.64, + "learning_rate": 3.928571428571429e-05, + "loss": 0.5466, + "step": 16760 + }, + { + "epoch": 0.64, + "learning_rate": 3.928507309566556e-05, + "loss": 0.4904, + "step": 16761 + }, + { + "epoch": 0.64, + "learning_rate": 3.928443190561683e-05, + "loss": 0.5459, + "step": 16762 + }, + { + "epoch": 0.64, + "learning_rate": 3.92837907155681e-05, + "loss": 0.4972, + "step": 16763 + }, + { + "epoch": 0.64, + "learning_rate": 3.928314952551936e-05, + "loss": 0.5341, + "step": 16764 + }, + { + "epoch": 0.64, + "learning_rate": 3.9282508335470635e-05, + "loss": 0.5071, + "step": 16765 + }, + { + "epoch": 0.64, + "learning_rate": 3.92818671454219e-05, + "loss": 0.5509, + "step": 16766 + }, + { + "epoch": 0.64, + "learning_rate": 3.928122595537317e-05, + "loss": 0.5095, + "step": 16767 + }, + { + "epoch": 0.64, + "learning_rate": 3.9280584765324443e-05, + "loss": 0.4965, + "step": 16768 + }, + { + "epoch": 0.64, + "learning_rate": 3.9279943575275715e-05, + "loss": 0.539, + "step": 16769 + }, + { + "epoch": 0.64, + "learning_rate": 3.927930238522699e-05, + "loss": 0.5729, + "step": 16770 + }, + { + "epoch": 0.64, + "learning_rate": 3.927866119517825e-05, + "loss": 0.5493, + "step": 16771 + }, + { + "epoch": 0.64, + "learning_rate": 3.9278020005129524e-05, + "loss": 0.4975, + "step": 16772 + }, + { + "epoch": 0.64, + "learning_rate": 3.927737881508079e-05, + "loss": 0.6071, + "step": 16773 + }, + { + "epoch": 0.64, + "learning_rate": 3.927673762503206e-05, + "loss": 0.5151, + "step": 16774 + }, + { + "epoch": 0.64, + "learning_rate": 3.9276096434983325e-05, + "loss": 0.636, + "step": 16775 + }, + { + "epoch": 0.64, + "learning_rate": 3.9275455244934604e-05, + "loss": 0.5363, + "step": 16776 + }, + { + "epoch": 0.65, + "learning_rate": 3.9274814054885875e-05, + "loss": 0.5455, + "step": 16777 + }, + { + "epoch": 0.65, + "learning_rate": 3.927417286483714e-05, + "loss": 0.5732, + "step": 16778 + }, + { + "epoch": 0.65, + "learning_rate": 3.927353167478841e-05, + "loss": 0.4977, + "step": 16779 + }, + { + "epoch": 0.65, + "learning_rate": 3.927289048473968e-05, + "loss": 0.5365, + "step": 16780 + }, + { + "epoch": 0.65, + "learning_rate": 3.927224929469095e-05, + "loss": 0.4704, + "step": 16781 + }, + { + "epoch": 0.65, + "learning_rate": 3.9271608104642214e-05, + "loss": 0.5189, + "step": 16782 + }, + { + "epoch": 0.65, + "learning_rate": 3.9270966914593485e-05, + "loss": 0.5602, + "step": 16783 + }, + { + "epoch": 0.65, + "learning_rate": 3.927032572454476e-05, + "loss": 0.4928, + "step": 16784 + }, + { + "epoch": 0.65, + "learning_rate": 3.926968453449603e-05, + "loss": 0.5534, + "step": 16785 + }, + { + "epoch": 0.65, + "learning_rate": 3.92690433444473e-05, + "loss": 0.4099, + "step": 16786 + }, + { + "epoch": 0.65, + "learning_rate": 3.9268402154398565e-05, + "loss": 0.5849, + "step": 16787 + }, + { + "epoch": 0.65, + "learning_rate": 3.926776096434984e-05, + "loss": 0.4891, + "step": 16788 + }, + { + "epoch": 0.65, + "learning_rate": 3.92671197743011e-05, + "loss": 0.5514, + "step": 16789 + }, + { + "epoch": 0.65, + "learning_rate": 3.9266478584252374e-05, + "loss": 0.5636, + "step": 16790 + }, + { + "epoch": 0.65, + "learning_rate": 3.926583739420364e-05, + "loss": 0.5583, + "step": 16791 + }, + { + "epoch": 0.65, + "learning_rate": 3.926519620415491e-05, + "loss": 0.52, + "step": 16792 + }, + { + "epoch": 0.65, + "learning_rate": 3.926455501410618e-05, + "loss": 0.5333, + "step": 16793 + }, + { + "epoch": 0.65, + "learning_rate": 3.9263913824057454e-05, + "loss": 0.4931, + "step": 16794 + }, + { + "epoch": 0.65, + "learning_rate": 3.9263272634008726e-05, + "loss": 0.5332, + "step": 16795 + }, + { + "epoch": 0.65, + "learning_rate": 3.926263144395999e-05, + "loss": 0.5433, + "step": 16796 + }, + { + "epoch": 0.65, + "learning_rate": 3.926199025391126e-05, + "loss": 0.5379, + "step": 16797 + }, + { + "epoch": 0.65, + "learning_rate": 3.926134906386253e-05, + "loss": 0.56, + "step": 16798 + }, + { + "epoch": 0.65, + "learning_rate": 3.92607078738138e-05, + "loss": 0.5419, + "step": 16799 + }, + { + "epoch": 0.65, + "learning_rate": 3.926006668376507e-05, + "loss": 0.4915, + "step": 16800 + }, + { + "epoch": 0.65, + "learning_rate": 3.925942549371634e-05, + "loss": 0.7326, + "step": 16801 + }, + { + "epoch": 0.65, + "learning_rate": 3.9258784303667614e-05, + "loss": 0.5467, + "step": 16802 + }, + { + "epoch": 0.65, + "learning_rate": 3.925814311361888e-05, + "loss": 0.4319, + "step": 16803 + }, + { + "epoch": 0.65, + "learning_rate": 3.925750192357015e-05, + "loss": 0.6519, + "step": 16804 + }, + { + "epoch": 0.65, + "learning_rate": 3.9256860733521416e-05, + "loss": 0.5518, + "step": 16805 + }, + { + "epoch": 0.65, + "learning_rate": 3.925621954347269e-05, + "loss": 0.4997, + "step": 16806 + }, + { + "epoch": 0.65, + "learning_rate": 3.925557835342395e-05, + "loss": 0.573, + "step": 16807 + }, + { + "epoch": 0.65, + "learning_rate": 3.9254937163375224e-05, + "loss": 0.5731, + "step": 16808 + }, + { + "epoch": 0.65, + "learning_rate": 3.9254295973326496e-05, + "loss": 0.5051, + "step": 16809 + }, + { + "epoch": 0.65, + "learning_rate": 3.925365478327777e-05, + "loss": 0.4981, + "step": 16810 + }, + { + "epoch": 0.65, + "learning_rate": 3.925301359322904e-05, + "loss": 0.4798, + "step": 16811 + }, + { + "epoch": 0.65, + "learning_rate": 3.9252372403180304e-05, + "loss": 0.4704, + "step": 16812 + }, + { + "epoch": 0.65, + "learning_rate": 3.9251731213131576e-05, + "loss": 0.5668, + "step": 16813 + }, + { + "epoch": 0.65, + "learning_rate": 3.925109002308284e-05, + "loss": 0.439, + "step": 16814 + }, + { + "epoch": 0.65, + "learning_rate": 3.925044883303411e-05, + "loss": 0.5833, + "step": 16815 + }, + { + "epoch": 0.65, + "learning_rate": 3.9249807642985384e-05, + "loss": 0.6226, + "step": 16816 + }, + { + "epoch": 0.65, + "learning_rate": 3.9249166452936656e-05, + "loss": 0.479, + "step": 16817 + }, + { + "epoch": 0.65, + "learning_rate": 3.924852526288792e-05, + "loss": 0.5474, + "step": 16818 + }, + { + "epoch": 0.65, + "learning_rate": 3.924788407283919e-05, + "loss": 0.5035, + "step": 16819 + }, + { + "epoch": 0.65, + "learning_rate": 3.9247242882790464e-05, + "loss": 0.445, + "step": 16820 + }, + { + "epoch": 0.65, + "learning_rate": 3.924660169274173e-05, + "loss": 0.4646, + "step": 16821 + }, + { + "epoch": 0.65, + "learning_rate": 3.9245960502693e-05, + "loss": 0.4, + "step": 16822 + }, + { + "epoch": 0.65, + "learning_rate": 3.9245319312644266e-05, + "loss": 0.5033, + "step": 16823 + }, + { + "epoch": 0.65, + "learning_rate": 3.924467812259554e-05, + "loss": 0.6146, + "step": 16824 + }, + { + "epoch": 0.65, + "learning_rate": 3.924403693254681e-05, + "loss": 0.5452, + "step": 16825 + }, + { + "epoch": 0.65, + "learning_rate": 3.924339574249808e-05, + "loss": 0.5356, + "step": 16826 + }, + { + "epoch": 0.65, + "learning_rate": 3.9242754552449346e-05, + "loss": 0.5449, + "step": 16827 + }, + { + "epoch": 0.65, + "learning_rate": 3.924211336240062e-05, + "loss": 0.4982, + "step": 16828 + }, + { + "epoch": 0.65, + "learning_rate": 3.924147217235189e-05, + "loss": 0.5464, + "step": 16829 + }, + { + "epoch": 0.65, + "learning_rate": 3.9240830982303154e-05, + "loss": 0.6589, + "step": 16830 + }, + { + "epoch": 0.65, + "learning_rate": 3.9240189792254426e-05, + "loss": 0.5227, + "step": 16831 + }, + { + "epoch": 0.65, + "learning_rate": 3.92395486022057e-05, + "loss": 0.5875, + "step": 16832 + }, + { + "epoch": 0.65, + "learning_rate": 3.923890741215697e-05, + "loss": 0.4353, + "step": 16833 + }, + { + "epoch": 0.65, + "learning_rate": 3.9238266222108234e-05, + "loss": 0.4933, + "step": 16834 + }, + { + "epoch": 0.65, + "learning_rate": 3.9237625032059506e-05, + "loss": 0.6206, + "step": 16835 + }, + { + "epoch": 0.65, + "learning_rate": 3.923698384201077e-05, + "loss": 0.5259, + "step": 16836 + }, + { + "epoch": 0.65, + "learning_rate": 3.923634265196204e-05, + "loss": 0.5051, + "step": 16837 + }, + { + "epoch": 0.65, + "learning_rate": 3.9235701461913315e-05, + "loss": 0.5028, + "step": 16838 + }, + { + "epoch": 0.65, + "learning_rate": 3.923506027186458e-05, + "loss": 0.5638, + "step": 16839 + }, + { + "epoch": 0.65, + "learning_rate": 3.923441908181585e-05, + "loss": 0.6427, + "step": 16840 + }, + { + "epoch": 0.65, + "learning_rate": 3.923377789176712e-05, + "loss": 0.5281, + "step": 16841 + }, + { + "epoch": 0.65, + "learning_rate": 3.9233136701718395e-05, + "loss": 0.4617, + "step": 16842 + }, + { + "epoch": 0.65, + "learning_rate": 3.923249551166966e-05, + "loss": 0.521, + "step": 16843 + }, + { + "epoch": 0.65, + "learning_rate": 3.923185432162093e-05, + "loss": 0.5333, + "step": 16844 + }, + { + "epoch": 0.65, + "learning_rate": 3.9231213131572196e-05, + "loss": 0.5665, + "step": 16845 + }, + { + "epoch": 0.65, + "learning_rate": 3.923057194152347e-05, + "loss": 0.4678, + "step": 16846 + }, + { + "epoch": 0.65, + "learning_rate": 3.922993075147474e-05, + "loss": 0.5438, + "step": 16847 + }, + { + "epoch": 0.65, + "learning_rate": 3.922928956142601e-05, + "loss": 0.453, + "step": 16848 + }, + { + "epoch": 0.65, + "learning_rate": 3.922864837137728e-05, + "loss": 0.4656, + "step": 16849 + }, + { + "epoch": 0.65, + "learning_rate": 3.922800718132855e-05, + "loss": 0.5085, + "step": 16850 + }, + { + "epoch": 0.65, + "learning_rate": 3.922736599127982e-05, + "loss": 0.4638, + "step": 16851 + }, + { + "epoch": 0.65, + "learning_rate": 3.9226724801231085e-05, + "loss": 0.4859, + "step": 16852 + }, + { + "epoch": 0.65, + "learning_rate": 3.9226083611182356e-05, + "loss": 0.5636, + "step": 16853 + }, + { + "epoch": 0.65, + "learning_rate": 3.922544242113362e-05, + "loss": 0.5654, + "step": 16854 + }, + { + "epoch": 0.65, + "learning_rate": 3.922480123108489e-05, + "loss": 0.4735, + "step": 16855 + }, + { + "epoch": 0.65, + "learning_rate": 3.9224160041036165e-05, + "loss": 0.6318, + "step": 16856 + }, + { + "epoch": 0.65, + "learning_rate": 3.9223518850987436e-05, + "loss": 0.4551, + "step": 16857 + }, + { + "epoch": 0.65, + "learning_rate": 3.922287766093871e-05, + "loss": 0.5102, + "step": 16858 + }, + { + "epoch": 0.65, + "learning_rate": 3.922223647088997e-05, + "loss": 0.3986, + "step": 16859 + }, + { + "epoch": 0.65, + "learning_rate": 3.9221595280841245e-05, + "loss": 0.5568, + "step": 16860 + }, + { + "epoch": 0.65, + "learning_rate": 3.922095409079251e-05, + "loss": 0.6608, + "step": 16861 + }, + { + "epoch": 0.65, + "learning_rate": 3.922031290074378e-05, + "loss": 0.5331, + "step": 16862 + }, + { + "epoch": 0.65, + "learning_rate": 3.9219671710695046e-05, + "loss": 0.422, + "step": 16863 + }, + { + "epoch": 0.65, + "learning_rate": 3.9219030520646325e-05, + "loss": 0.5414, + "step": 16864 + }, + { + "epoch": 0.65, + "learning_rate": 3.92183893305976e-05, + "loss": 0.5022, + "step": 16865 + }, + { + "epoch": 0.65, + "learning_rate": 3.921774814054886e-05, + "loss": 0.5029, + "step": 16866 + }, + { + "epoch": 0.65, + "learning_rate": 3.921710695050013e-05, + "loss": 0.5507, + "step": 16867 + }, + { + "epoch": 0.65, + "learning_rate": 3.92164657604514e-05, + "loss": 0.4207, + "step": 16868 + }, + { + "epoch": 0.65, + "learning_rate": 3.921582457040267e-05, + "loss": 0.545, + "step": 16869 + }, + { + "epoch": 0.65, + "learning_rate": 3.9215183380353935e-05, + "loss": 0.4994, + "step": 16870 + }, + { + "epoch": 0.65, + "learning_rate": 3.921454219030521e-05, + "loss": 0.5144, + "step": 16871 + }, + { + "epoch": 0.65, + "learning_rate": 3.921390100025648e-05, + "loss": 0.5372, + "step": 16872 + }, + { + "epoch": 0.65, + "learning_rate": 3.921325981020775e-05, + "loss": 0.543, + "step": 16873 + }, + { + "epoch": 0.65, + "learning_rate": 3.921261862015902e-05, + "loss": 0.5043, + "step": 16874 + }, + { + "epoch": 0.65, + "learning_rate": 3.921197743011029e-05, + "loss": 0.5958, + "step": 16875 + }, + { + "epoch": 0.65, + "learning_rate": 3.921133624006156e-05, + "loss": 0.4842, + "step": 16876 + }, + { + "epoch": 0.65, + "learning_rate": 3.921069505001282e-05, + "loss": 0.5748, + "step": 16877 + }, + { + "epoch": 0.65, + "learning_rate": 3.9210053859964095e-05, + "loss": 0.5483, + "step": 16878 + }, + { + "epoch": 0.65, + "learning_rate": 3.920941266991536e-05, + "loss": 0.4843, + "step": 16879 + }, + { + "epoch": 0.65, + "learning_rate": 3.920877147986663e-05, + "loss": 0.5201, + "step": 16880 + }, + { + "epoch": 0.65, + "learning_rate": 3.9208130289817903e-05, + "loss": 0.5107, + "step": 16881 + }, + { + "epoch": 0.65, + "learning_rate": 3.9207489099769175e-05, + "loss": 0.5284, + "step": 16882 + }, + { + "epoch": 0.65, + "learning_rate": 3.920684790972045e-05, + "loss": 0.4761, + "step": 16883 + }, + { + "epoch": 0.65, + "learning_rate": 3.920620671967171e-05, + "loss": 0.5258, + "step": 16884 + }, + { + "epoch": 0.65, + "learning_rate": 3.9205565529622984e-05, + "loss": 0.6007, + "step": 16885 + }, + { + "epoch": 0.65, + "learning_rate": 3.920492433957425e-05, + "loss": 0.497, + "step": 16886 + }, + { + "epoch": 0.65, + "learning_rate": 3.920428314952552e-05, + "loss": 0.5847, + "step": 16887 + }, + { + "epoch": 0.65, + "learning_rate": 3.920364195947679e-05, + "loss": 0.6517, + "step": 16888 + }, + { + "epoch": 0.65, + "learning_rate": 3.9203000769428064e-05, + "loss": 0.5004, + "step": 16889 + }, + { + "epoch": 0.65, + "learning_rate": 3.920235957937933e-05, + "loss": 0.5344, + "step": 16890 + }, + { + "epoch": 0.65, + "learning_rate": 3.92017183893306e-05, + "loss": 0.4717, + "step": 16891 + }, + { + "epoch": 0.65, + "learning_rate": 3.920107719928187e-05, + "loss": 0.5614, + "step": 16892 + }, + { + "epoch": 0.65, + "learning_rate": 3.920043600923314e-05, + "loss": 0.6614, + "step": 16893 + }, + { + "epoch": 0.65, + "learning_rate": 3.919979481918441e-05, + "loss": 0.4299, + "step": 16894 + }, + { + "epoch": 0.65, + "learning_rate": 3.9199153629135674e-05, + "loss": 0.6534, + "step": 16895 + }, + { + "epoch": 0.65, + "learning_rate": 3.9198512439086945e-05, + "loss": 0.5977, + "step": 16896 + }, + { + "epoch": 0.65, + "learning_rate": 3.919787124903822e-05, + "loss": 0.6489, + "step": 16897 + }, + { + "epoch": 0.65, + "learning_rate": 3.919723005898949e-05, + "loss": 0.541, + "step": 16898 + }, + { + "epoch": 0.65, + "learning_rate": 3.9196588868940754e-05, + "loss": 0.4427, + "step": 16899 + }, + { + "epoch": 0.65, + "learning_rate": 3.9195947678892025e-05, + "loss": 0.568, + "step": 16900 + }, + { + "epoch": 0.65, + "learning_rate": 3.91953064888433e-05, + "loss": 0.6264, + "step": 16901 + }, + { + "epoch": 0.65, + "learning_rate": 3.919466529879456e-05, + "loss": 0.4856, + "step": 16902 + }, + { + "epoch": 0.65, + "learning_rate": 3.9194024108745834e-05, + "loss": 0.6205, + "step": 16903 + }, + { + "epoch": 0.65, + "learning_rate": 3.9193382918697106e-05, + "loss": 0.5313, + "step": 16904 + }, + { + "epoch": 0.65, + "learning_rate": 3.919274172864838e-05, + "loss": 0.4556, + "step": 16905 + }, + { + "epoch": 0.65, + "learning_rate": 3.919210053859964e-05, + "loss": 0.5378, + "step": 16906 + }, + { + "epoch": 0.65, + "learning_rate": 3.9191459348550914e-05, + "loss": 0.4838, + "step": 16907 + }, + { + "epoch": 0.65, + "learning_rate": 3.919081815850218e-05, + "loss": 0.5854, + "step": 16908 + }, + { + "epoch": 0.65, + "learning_rate": 3.919017696845345e-05, + "loss": 0.5268, + "step": 16909 + }, + { + "epoch": 0.65, + "learning_rate": 3.918953577840472e-05, + "loss": 0.5619, + "step": 16910 + }, + { + "epoch": 0.65, + "learning_rate": 3.918889458835599e-05, + "loss": 0.5389, + "step": 16911 + }, + { + "epoch": 0.65, + "learning_rate": 3.918825339830726e-05, + "loss": 0.5103, + "step": 16912 + }, + { + "epoch": 0.65, + "learning_rate": 3.918761220825853e-05, + "loss": 0.4771, + "step": 16913 + }, + { + "epoch": 0.65, + "learning_rate": 3.91869710182098e-05, + "loss": 0.5092, + "step": 16914 + }, + { + "epoch": 0.65, + "learning_rate": 3.918632982816107e-05, + "loss": 0.4887, + "step": 16915 + }, + { + "epoch": 0.65, + "learning_rate": 3.918568863811234e-05, + "loss": 0.4515, + "step": 16916 + }, + { + "epoch": 0.65, + "learning_rate": 3.9185047448063604e-05, + "loss": 0.5365, + "step": 16917 + }, + { + "epoch": 0.65, + "learning_rate": 3.9184406258014876e-05, + "loss": 0.5823, + "step": 16918 + }, + { + "epoch": 0.65, + "learning_rate": 3.918376506796615e-05, + "loss": 0.4836, + "step": 16919 + }, + { + "epoch": 0.65, + "learning_rate": 3.918312387791742e-05, + "loss": 0.5979, + "step": 16920 + }, + { + "epoch": 0.65, + "learning_rate": 3.918248268786869e-05, + "loss": 0.5564, + "step": 16921 + }, + { + "epoch": 0.65, + "learning_rate": 3.9181841497819956e-05, + "loss": 0.5359, + "step": 16922 + }, + { + "epoch": 0.65, + "learning_rate": 3.918120030777123e-05, + "loss": 0.5239, + "step": 16923 + }, + { + "epoch": 0.65, + "learning_rate": 3.918055911772249e-05, + "loss": 0.6255, + "step": 16924 + }, + { + "epoch": 0.65, + "learning_rate": 3.9179917927673764e-05, + "loss": 0.4445, + "step": 16925 + }, + { + "epoch": 0.65, + "learning_rate": 3.917927673762503e-05, + "loss": 0.5622, + "step": 16926 + }, + { + "epoch": 0.65, + "learning_rate": 3.91786355475763e-05, + "loss": 0.5222, + "step": 16927 + }, + { + "epoch": 0.65, + "learning_rate": 3.917799435752757e-05, + "loss": 0.6394, + "step": 16928 + }, + { + "epoch": 0.65, + "learning_rate": 3.9177353167478844e-05, + "loss": 0.6195, + "step": 16929 + }, + { + "epoch": 0.65, + "learning_rate": 3.9176711977430116e-05, + "loss": 0.5756, + "step": 16930 + }, + { + "epoch": 0.65, + "learning_rate": 3.917607078738138e-05, + "loss": 0.6447, + "step": 16931 + }, + { + "epoch": 0.65, + "learning_rate": 3.917542959733265e-05, + "loss": 0.4711, + "step": 16932 + }, + { + "epoch": 0.65, + "learning_rate": 3.917478840728392e-05, + "loss": 0.5966, + "step": 16933 + }, + { + "epoch": 0.65, + "learning_rate": 3.917414721723519e-05, + "loss": 0.4911, + "step": 16934 + }, + { + "epoch": 0.65, + "learning_rate": 3.9173506027186454e-05, + "loss": 0.4276, + "step": 16935 + }, + { + "epoch": 0.65, + "learning_rate": 3.917286483713773e-05, + "loss": 0.6001, + "step": 16936 + }, + { + "epoch": 0.65, + "learning_rate": 3.9172223647089004e-05, + "loss": 0.5403, + "step": 16937 + }, + { + "epoch": 0.65, + "learning_rate": 3.917158245704027e-05, + "loss": 0.5065, + "step": 16938 + }, + { + "epoch": 0.65, + "learning_rate": 3.917094126699154e-05, + "loss": 0.4965, + "step": 16939 + }, + { + "epoch": 0.65, + "learning_rate": 3.9170300076942806e-05, + "loss": 0.4641, + "step": 16940 + }, + { + "epoch": 0.65, + "learning_rate": 3.916965888689408e-05, + "loss": 0.5196, + "step": 16941 + }, + { + "epoch": 0.65, + "learning_rate": 3.916901769684534e-05, + "loss": 0.4812, + "step": 16942 + }, + { + "epoch": 0.65, + "learning_rate": 3.9168376506796614e-05, + "loss": 0.611, + "step": 16943 + }, + { + "epoch": 0.65, + "learning_rate": 3.9167735316747886e-05, + "loss": 0.586, + "step": 16944 + }, + { + "epoch": 0.65, + "learning_rate": 3.916709412669916e-05, + "loss": 0.5404, + "step": 16945 + }, + { + "epoch": 0.65, + "learning_rate": 3.916645293665043e-05, + "loss": 0.4786, + "step": 16946 + }, + { + "epoch": 0.65, + "learning_rate": 3.9165811746601694e-05, + "loss": 0.4959, + "step": 16947 + }, + { + "epoch": 0.65, + "learning_rate": 3.9165170556552966e-05, + "loss": 0.5183, + "step": 16948 + }, + { + "epoch": 0.65, + "learning_rate": 3.916452936650423e-05, + "loss": 0.601, + "step": 16949 + }, + { + "epoch": 0.65, + "learning_rate": 3.91638881764555e-05, + "loss": 0.6084, + "step": 16950 + }, + { + "epoch": 0.65, + "learning_rate": 3.916324698640677e-05, + "loss": 0.61, + "step": 16951 + }, + { + "epoch": 0.65, + "learning_rate": 3.916260579635804e-05, + "loss": 0.4364, + "step": 16952 + }, + { + "epoch": 0.65, + "learning_rate": 3.916196460630931e-05, + "loss": 0.4748, + "step": 16953 + }, + { + "epoch": 0.65, + "learning_rate": 3.916132341626058e-05, + "loss": 0.53, + "step": 16954 + }, + { + "epoch": 0.65, + "learning_rate": 3.9160682226211855e-05, + "loss": 0.5281, + "step": 16955 + }, + { + "epoch": 0.65, + "learning_rate": 3.916004103616312e-05, + "loss": 0.4826, + "step": 16956 + }, + { + "epoch": 0.65, + "learning_rate": 3.915939984611439e-05, + "loss": 0.5396, + "step": 16957 + }, + { + "epoch": 0.65, + "learning_rate": 3.9158758656065656e-05, + "loss": 0.4276, + "step": 16958 + }, + { + "epoch": 0.65, + "learning_rate": 3.915811746601693e-05, + "loss": 0.4837, + "step": 16959 + }, + { + "epoch": 0.65, + "learning_rate": 3.91574762759682e-05, + "loss": 0.4891, + "step": 16960 + }, + { + "epoch": 0.65, + "learning_rate": 3.915683508591947e-05, + "loss": 0.52, + "step": 16961 + }, + { + "epoch": 0.65, + "learning_rate": 3.9156193895870736e-05, + "loss": 0.5218, + "step": 16962 + }, + { + "epoch": 0.65, + "learning_rate": 3.915555270582201e-05, + "loss": 0.5146, + "step": 16963 + }, + { + "epoch": 0.65, + "learning_rate": 3.915491151577328e-05, + "loss": 0.6021, + "step": 16964 + }, + { + "epoch": 0.65, + "learning_rate": 3.9154270325724545e-05, + "loss": 0.6102, + "step": 16965 + }, + { + "epoch": 0.65, + "learning_rate": 3.9153629135675816e-05, + "loss": 0.4402, + "step": 16966 + }, + { + "epoch": 0.65, + "learning_rate": 3.915298794562708e-05, + "loss": 0.4004, + "step": 16967 + }, + { + "epoch": 0.65, + "learning_rate": 3.915234675557835e-05, + "loss": 0.4843, + "step": 16968 + }, + { + "epoch": 0.65, + "learning_rate": 3.9151705565529625e-05, + "loss": 0.565, + "step": 16969 + }, + { + "epoch": 0.65, + "learning_rate": 3.9151064375480896e-05, + "loss": 0.3628, + "step": 16970 + }, + { + "epoch": 0.65, + "learning_rate": 3.915042318543216e-05, + "loss": 0.4751, + "step": 16971 + }, + { + "epoch": 0.65, + "learning_rate": 3.914978199538343e-05, + "loss": 0.4937, + "step": 16972 + }, + { + "epoch": 0.65, + "learning_rate": 3.9149140805334705e-05, + "loss": 0.5352, + "step": 16973 + }, + { + "epoch": 0.65, + "learning_rate": 3.914849961528597e-05, + "loss": 0.5213, + "step": 16974 + }, + { + "epoch": 0.65, + "learning_rate": 3.914785842523724e-05, + "loss": 0.5333, + "step": 16975 + }, + { + "epoch": 0.65, + "learning_rate": 3.914721723518851e-05, + "loss": 0.4426, + "step": 16976 + }, + { + "epoch": 0.65, + "learning_rate": 3.9146576045139785e-05, + "loss": 0.5067, + "step": 16977 + }, + { + "epoch": 0.65, + "learning_rate": 3.914593485509105e-05, + "loss": 0.5031, + "step": 16978 + }, + { + "epoch": 0.65, + "learning_rate": 3.914529366504232e-05, + "loss": 0.5405, + "step": 16979 + }, + { + "epoch": 0.65, + "learning_rate": 3.9144652474993587e-05, + "loss": 0.4668, + "step": 16980 + }, + { + "epoch": 0.65, + "learning_rate": 3.914401128494486e-05, + "loss": 0.4182, + "step": 16981 + }, + { + "epoch": 0.65, + "learning_rate": 3.914337009489613e-05, + "loss": 0.63, + "step": 16982 + }, + { + "epoch": 0.65, + "learning_rate": 3.9142728904847395e-05, + "loss": 0.4209, + "step": 16983 + }, + { + "epoch": 0.65, + "learning_rate": 3.914208771479867e-05, + "loss": 0.5007, + "step": 16984 + }, + { + "epoch": 0.65, + "learning_rate": 3.914144652474994e-05, + "loss": 0.5384, + "step": 16985 + }, + { + "epoch": 0.65, + "learning_rate": 3.914080533470121e-05, + "loss": 0.7031, + "step": 16986 + }, + { + "epoch": 0.65, + "learning_rate": 3.9140164144652475e-05, + "loss": 0.4983, + "step": 16987 + }, + { + "epoch": 0.65, + "learning_rate": 3.913952295460375e-05, + "loss": 0.3968, + "step": 16988 + }, + { + "epoch": 0.65, + "learning_rate": 3.913888176455501e-05, + "loss": 0.6151, + "step": 16989 + }, + { + "epoch": 0.65, + "learning_rate": 3.9138240574506283e-05, + "loss": 0.5113, + "step": 16990 + }, + { + "epoch": 0.65, + "learning_rate": 3.9137599384457555e-05, + "loss": 0.5381, + "step": 16991 + }, + { + "epoch": 0.65, + "learning_rate": 3.913695819440883e-05, + "loss": 0.5264, + "step": 16992 + }, + { + "epoch": 0.65, + "learning_rate": 3.91363170043601e-05, + "loss": 0.4888, + "step": 16993 + }, + { + "epoch": 0.65, + "learning_rate": 3.9135675814311363e-05, + "loss": 0.5253, + "step": 16994 + }, + { + "epoch": 0.65, + "learning_rate": 3.9135034624262635e-05, + "loss": 0.4513, + "step": 16995 + }, + { + "epoch": 0.65, + "learning_rate": 3.91343934342139e-05, + "loss": 0.5356, + "step": 16996 + }, + { + "epoch": 0.65, + "learning_rate": 3.913375224416517e-05, + "loss": 0.5917, + "step": 16997 + }, + { + "epoch": 0.65, + "learning_rate": 3.9133111054116444e-05, + "loss": 0.6284, + "step": 16998 + }, + { + "epoch": 0.65, + "learning_rate": 3.913246986406771e-05, + "loss": 0.5118, + "step": 16999 + }, + { + "epoch": 0.65, + "learning_rate": 3.913182867401898e-05, + "loss": 0.5373, + "step": 17000 + }, + { + "epoch": 0.65, + "learning_rate": 3.913118748397025e-05, + "loss": 0.5616, + "step": 17001 + }, + { + "epoch": 0.65, + "learning_rate": 3.9130546293921524e-05, + "loss": 0.6254, + "step": 17002 + }, + { + "epoch": 0.65, + "learning_rate": 3.912990510387279e-05, + "loss": 0.4471, + "step": 17003 + }, + { + "epoch": 0.65, + "learning_rate": 3.912926391382406e-05, + "loss": 0.5542, + "step": 17004 + }, + { + "epoch": 0.65, + "learning_rate": 3.9128622723775325e-05, + "loss": 0.5869, + "step": 17005 + }, + { + "epoch": 0.65, + "learning_rate": 3.91279815337266e-05, + "loss": 0.5018, + "step": 17006 + }, + { + "epoch": 0.65, + "learning_rate": 3.912734034367787e-05, + "loss": 0.4992, + "step": 17007 + }, + { + "epoch": 0.65, + "learning_rate": 3.912669915362914e-05, + "loss": 0.5971, + "step": 17008 + }, + { + "epoch": 0.65, + "learning_rate": 3.912605796358041e-05, + "loss": 0.5948, + "step": 17009 + }, + { + "epoch": 0.65, + "learning_rate": 3.912541677353168e-05, + "loss": 0.53, + "step": 17010 + }, + { + "epoch": 0.65, + "learning_rate": 3.912477558348295e-05, + "loss": 0.5254, + "step": 17011 + }, + { + "epoch": 0.65, + "learning_rate": 3.9124134393434214e-05, + "loss": 0.6823, + "step": 17012 + }, + { + "epoch": 0.65, + "learning_rate": 3.9123493203385485e-05, + "loss": 0.5171, + "step": 17013 + }, + { + "epoch": 0.65, + "learning_rate": 3.912285201333675e-05, + "loss": 0.5835, + "step": 17014 + }, + { + "epoch": 0.65, + "learning_rate": 3.912221082328802e-05, + "loss": 0.5654, + "step": 17015 + }, + { + "epoch": 0.65, + "learning_rate": 3.9121569633239294e-05, + "loss": 0.595, + "step": 17016 + }, + { + "epoch": 0.65, + "learning_rate": 3.9120928443190566e-05, + "loss": 0.5581, + "step": 17017 + }, + { + "epoch": 0.65, + "learning_rate": 3.912028725314184e-05, + "loss": 0.4323, + "step": 17018 + }, + { + "epoch": 0.65, + "learning_rate": 3.91196460630931e-05, + "loss": 0.4992, + "step": 17019 + }, + { + "epoch": 0.65, + "learning_rate": 3.9119004873044374e-05, + "loss": 0.5449, + "step": 17020 + }, + { + "epoch": 0.65, + "learning_rate": 3.911836368299564e-05, + "loss": 0.5364, + "step": 17021 + }, + { + "epoch": 0.65, + "learning_rate": 3.911772249294691e-05, + "loss": 0.4796, + "step": 17022 + }, + { + "epoch": 0.65, + "learning_rate": 3.9117081302898176e-05, + "loss": 0.5846, + "step": 17023 + }, + { + "epoch": 0.65, + "learning_rate": 3.9116440112849454e-05, + "loss": 0.5787, + "step": 17024 + }, + { + "epoch": 0.65, + "learning_rate": 3.9115798922800726e-05, + "loss": 0.4341, + "step": 17025 + }, + { + "epoch": 0.65, + "learning_rate": 3.911515773275199e-05, + "loss": 0.5112, + "step": 17026 + }, + { + "epoch": 0.65, + "learning_rate": 3.911451654270326e-05, + "loss": 0.5184, + "step": 17027 + }, + { + "epoch": 0.65, + "learning_rate": 3.911387535265453e-05, + "loss": 0.5246, + "step": 17028 + }, + { + "epoch": 0.65, + "learning_rate": 3.91132341626058e-05, + "loss": 0.4601, + "step": 17029 + }, + { + "epoch": 0.65, + "learning_rate": 3.9112592972557064e-05, + "loss": 0.5064, + "step": 17030 + }, + { + "epoch": 0.65, + "learning_rate": 3.9111951782508336e-05, + "loss": 0.5036, + "step": 17031 + }, + { + "epoch": 0.65, + "learning_rate": 3.911131059245961e-05, + "loss": 0.4559, + "step": 17032 + }, + { + "epoch": 0.65, + "learning_rate": 3.911066940241088e-05, + "loss": 0.4684, + "step": 17033 + }, + { + "epoch": 0.65, + "learning_rate": 3.911002821236215e-05, + "loss": 0.5086, + "step": 17034 + }, + { + "epoch": 0.65, + "learning_rate": 3.9109387022313416e-05, + "loss": 0.5666, + "step": 17035 + }, + { + "epoch": 0.65, + "learning_rate": 3.910874583226469e-05, + "loss": 0.585, + "step": 17036 + }, + { + "epoch": 0.66, + "learning_rate": 3.910810464221595e-05, + "loss": 0.4488, + "step": 17037 + }, + { + "epoch": 0.66, + "learning_rate": 3.9107463452167224e-05, + "loss": 0.481, + "step": 17038 + }, + { + "epoch": 0.66, + "learning_rate": 3.910682226211849e-05, + "loss": 0.4873, + "step": 17039 + }, + { + "epoch": 0.66, + "learning_rate": 3.910618107206976e-05, + "loss": 0.5454, + "step": 17040 + }, + { + "epoch": 0.66, + "learning_rate": 3.910553988202103e-05, + "loss": 0.5487, + "step": 17041 + }, + { + "epoch": 0.66, + "learning_rate": 3.9104898691972304e-05, + "loss": 0.5646, + "step": 17042 + }, + { + "epoch": 0.66, + "learning_rate": 3.9104257501923576e-05, + "loss": 0.4079, + "step": 17043 + }, + { + "epoch": 0.66, + "learning_rate": 3.910361631187484e-05, + "loss": 0.5737, + "step": 17044 + }, + { + "epoch": 0.66, + "learning_rate": 3.910297512182611e-05, + "loss": 0.5229, + "step": 17045 + }, + { + "epoch": 0.66, + "learning_rate": 3.910233393177738e-05, + "loss": 0.6064, + "step": 17046 + }, + { + "epoch": 0.66, + "learning_rate": 3.910169274172865e-05, + "loss": 0.5459, + "step": 17047 + }, + { + "epoch": 0.66, + "learning_rate": 3.910105155167992e-05, + "loss": 0.5369, + "step": 17048 + }, + { + "epoch": 0.66, + "learning_rate": 3.910041036163119e-05, + "loss": 0.4498, + "step": 17049 + }, + { + "epoch": 0.66, + "learning_rate": 3.909976917158246e-05, + "loss": 0.4505, + "step": 17050 + }, + { + "epoch": 0.66, + "learning_rate": 3.909912798153373e-05, + "loss": 0.4298, + "step": 17051 + }, + { + "epoch": 0.66, + "learning_rate": 3.9098486791485e-05, + "loss": 0.6111, + "step": 17052 + }, + { + "epoch": 0.66, + "learning_rate": 3.9097845601436266e-05, + "loss": 0.4848, + "step": 17053 + }, + { + "epoch": 0.66, + "learning_rate": 3.909720441138754e-05, + "loss": 0.5899, + "step": 17054 + }, + { + "epoch": 0.66, + "learning_rate": 3.90965632213388e-05, + "loss": 0.6507, + "step": 17055 + }, + { + "epoch": 0.66, + "learning_rate": 3.9095922031290074e-05, + "loss": 0.5203, + "step": 17056 + }, + { + "epoch": 0.66, + "learning_rate": 3.9095280841241346e-05, + "loss": 0.5943, + "step": 17057 + }, + { + "epoch": 0.66, + "learning_rate": 3.909463965119262e-05, + "loss": 0.5622, + "step": 17058 + }, + { + "epoch": 0.66, + "learning_rate": 3.909399846114388e-05, + "loss": 0.5229, + "step": 17059 + }, + { + "epoch": 0.66, + "learning_rate": 3.9093357271095154e-05, + "loss": 0.5449, + "step": 17060 + }, + { + "epoch": 0.66, + "learning_rate": 3.9092716081046426e-05, + "loss": 0.4306, + "step": 17061 + }, + { + "epoch": 0.66, + "learning_rate": 3.909207489099769e-05, + "loss": 0.5291, + "step": 17062 + }, + { + "epoch": 0.66, + "learning_rate": 3.909143370094896e-05, + "loss": 0.6127, + "step": 17063 + }, + { + "epoch": 0.66, + "learning_rate": 3.9090792510900235e-05, + "loss": 0.4664, + "step": 17064 + }, + { + "epoch": 0.66, + "learning_rate": 3.9090151320851506e-05, + "loss": 0.6191, + "step": 17065 + }, + { + "epoch": 0.66, + "learning_rate": 3.908951013080277e-05, + "loss": 0.579, + "step": 17066 + }, + { + "epoch": 0.66, + "learning_rate": 3.908886894075404e-05, + "loss": 0.5213, + "step": 17067 + }, + { + "epoch": 0.66, + "learning_rate": 3.908822775070531e-05, + "loss": 0.6828, + "step": 17068 + }, + { + "epoch": 0.66, + "learning_rate": 3.908758656065658e-05, + "loss": 0.5416, + "step": 17069 + }, + { + "epoch": 0.66, + "learning_rate": 3.908694537060785e-05, + "loss": 0.4385, + "step": 17070 + }, + { + "epoch": 0.66, + "learning_rate": 3.9086304180559116e-05, + "loss": 0.6299, + "step": 17071 + }, + { + "epoch": 0.66, + "learning_rate": 3.908566299051039e-05, + "loss": 0.5454, + "step": 17072 + }, + { + "epoch": 0.66, + "learning_rate": 3.908502180046166e-05, + "loss": 0.4917, + "step": 17073 + }, + { + "epoch": 0.66, + "learning_rate": 3.908438061041293e-05, + "loss": 0.5451, + "step": 17074 + }, + { + "epoch": 0.66, + "learning_rate": 3.9083739420364196e-05, + "loss": 0.5494, + "step": 17075 + }, + { + "epoch": 0.66, + "learning_rate": 3.908309823031547e-05, + "loss": 0.5952, + "step": 17076 + }, + { + "epoch": 0.66, + "learning_rate": 3.908245704026673e-05, + "loss": 0.5967, + "step": 17077 + }, + { + "epoch": 0.66, + "learning_rate": 3.9081815850218005e-05, + "loss": 0.629, + "step": 17078 + }, + { + "epoch": 0.66, + "learning_rate": 3.9081174660169276e-05, + "loss": 0.5145, + "step": 17079 + }, + { + "epoch": 0.66, + "learning_rate": 3.908053347012055e-05, + "loss": 0.5695, + "step": 17080 + }, + { + "epoch": 0.66, + "learning_rate": 3.907989228007182e-05, + "loss": 0.5307, + "step": 17081 + }, + { + "epoch": 0.66, + "learning_rate": 3.9079251090023085e-05, + "loss": 0.4859, + "step": 17082 + }, + { + "epoch": 0.66, + "learning_rate": 3.9078609899974357e-05, + "loss": 0.5575, + "step": 17083 + }, + { + "epoch": 0.66, + "learning_rate": 3.907796870992562e-05, + "loss": 0.5111, + "step": 17084 + }, + { + "epoch": 0.66, + "learning_rate": 3.907732751987689e-05, + "loss": 0.4702, + "step": 17085 + }, + { + "epoch": 0.66, + "learning_rate": 3.907668632982816e-05, + "loss": 0.6378, + "step": 17086 + }, + { + "epoch": 0.66, + "learning_rate": 3.907604513977943e-05, + "loss": 0.538, + "step": 17087 + }, + { + "epoch": 0.66, + "learning_rate": 3.90754039497307e-05, + "loss": 0.5058, + "step": 17088 + }, + { + "epoch": 0.66, + "learning_rate": 3.907476275968197e-05, + "loss": 0.5263, + "step": 17089 + }, + { + "epoch": 0.66, + "learning_rate": 3.9074121569633245e-05, + "loss": 0.5875, + "step": 17090 + }, + { + "epoch": 0.66, + "learning_rate": 3.907348037958451e-05, + "loss": 0.4698, + "step": 17091 + }, + { + "epoch": 0.66, + "learning_rate": 3.907283918953578e-05, + "loss": 0.5418, + "step": 17092 + }, + { + "epoch": 0.66, + "learning_rate": 3.9072197999487047e-05, + "loss": 0.4621, + "step": 17093 + }, + { + "epoch": 0.66, + "learning_rate": 3.907155680943832e-05, + "loss": 0.6803, + "step": 17094 + }, + { + "epoch": 0.66, + "learning_rate": 3.907091561938958e-05, + "loss": 0.4782, + "step": 17095 + }, + { + "epoch": 0.66, + "learning_rate": 3.907027442934086e-05, + "loss": 0.5956, + "step": 17096 + }, + { + "epoch": 0.66, + "learning_rate": 3.9069633239292133e-05, + "loss": 0.4559, + "step": 17097 + }, + { + "epoch": 0.66, + "learning_rate": 3.90689920492434e-05, + "loss": 0.5731, + "step": 17098 + }, + { + "epoch": 0.66, + "learning_rate": 3.906835085919467e-05, + "loss": 0.5093, + "step": 17099 + }, + { + "epoch": 0.66, + "learning_rate": 3.9067709669145935e-05, + "loss": 0.484, + "step": 17100 + }, + { + "epoch": 0.66, + "learning_rate": 3.906706847909721e-05, + "loss": 0.6517, + "step": 17101 + }, + { + "epoch": 0.66, + "learning_rate": 3.906642728904847e-05, + "loss": 0.6641, + "step": 17102 + }, + { + "epoch": 0.66, + "learning_rate": 3.9065786098999743e-05, + "loss": 0.5393, + "step": 17103 + }, + { + "epoch": 0.66, + "learning_rate": 3.9065144908951015e-05, + "loss": 0.4489, + "step": 17104 + }, + { + "epoch": 0.66, + "learning_rate": 3.906450371890229e-05, + "loss": 0.5416, + "step": 17105 + }, + { + "epoch": 0.66, + "learning_rate": 3.906386252885356e-05, + "loss": 0.4811, + "step": 17106 + }, + { + "epoch": 0.66, + "learning_rate": 3.9063221338804824e-05, + "loss": 0.4629, + "step": 17107 + }, + { + "epoch": 0.66, + "learning_rate": 3.9062580148756095e-05, + "loss": 0.5468, + "step": 17108 + }, + { + "epoch": 0.66, + "learning_rate": 3.906193895870736e-05, + "loss": 0.4882, + "step": 17109 + }, + { + "epoch": 0.66, + "learning_rate": 3.906129776865863e-05, + "loss": 0.5736, + "step": 17110 + }, + { + "epoch": 0.66, + "learning_rate": 3.90606565786099e-05, + "loss": 0.5902, + "step": 17111 + }, + { + "epoch": 0.66, + "learning_rate": 3.906001538856117e-05, + "loss": 0.4433, + "step": 17112 + }, + { + "epoch": 0.66, + "learning_rate": 3.905937419851244e-05, + "loss": 0.6855, + "step": 17113 + }, + { + "epoch": 0.66, + "learning_rate": 3.905873300846371e-05, + "loss": 0.5164, + "step": 17114 + }, + { + "epoch": 0.66, + "learning_rate": 3.9058091818414984e-05, + "loss": 0.4798, + "step": 17115 + }, + { + "epoch": 0.66, + "learning_rate": 3.905745062836625e-05, + "loss": 0.4909, + "step": 17116 + }, + { + "epoch": 0.66, + "learning_rate": 3.905680943831752e-05, + "loss": 0.4961, + "step": 17117 + }, + { + "epoch": 0.66, + "learning_rate": 3.9056168248268785e-05, + "loss": 0.5232, + "step": 17118 + }, + { + "epoch": 0.66, + "learning_rate": 3.905552705822006e-05, + "loss": 0.4344, + "step": 17119 + }, + { + "epoch": 0.66, + "learning_rate": 3.905488586817133e-05, + "loss": 0.5856, + "step": 17120 + }, + { + "epoch": 0.66, + "learning_rate": 3.90542446781226e-05, + "loss": 0.3767, + "step": 17121 + }, + { + "epoch": 0.66, + "learning_rate": 3.9053603488073865e-05, + "loss": 0.4879, + "step": 17122 + }, + { + "epoch": 0.66, + "learning_rate": 3.905296229802514e-05, + "loss": 0.4634, + "step": 17123 + }, + { + "epoch": 0.66, + "learning_rate": 3.905232110797641e-05, + "loss": 0.5634, + "step": 17124 + }, + { + "epoch": 0.66, + "learning_rate": 3.9051679917927674e-05, + "loss": 0.612, + "step": 17125 + }, + { + "epoch": 0.66, + "learning_rate": 3.9051038727878945e-05, + "loss": 0.5523, + "step": 17126 + }, + { + "epoch": 0.66, + "learning_rate": 3.905039753783021e-05, + "loss": 0.5922, + "step": 17127 + }, + { + "epoch": 0.66, + "learning_rate": 3.904975634778148e-05, + "loss": 0.5303, + "step": 17128 + }, + { + "epoch": 0.66, + "learning_rate": 3.9049115157732754e-05, + "loss": 0.5674, + "step": 17129 + }, + { + "epoch": 0.66, + "learning_rate": 3.9048473967684026e-05, + "loss": 0.533, + "step": 17130 + }, + { + "epoch": 0.66, + "learning_rate": 3.904783277763529e-05, + "loss": 0.5754, + "step": 17131 + }, + { + "epoch": 0.66, + "learning_rate": 3.904719158758656e-05, + "loss": 0.5876, + "step": 17132 + }, + { + "epoch": 0.66, + "learning_rate": 3.9046550397537834e-05, + "loss": 0.473, + "step": 17133 + }, + { + "epoch": 0.66, + "learning_rate": 3.90459092074891e-05, + "loss": 0.5229, + "step": 17134 + }, + { + "epoch": 0.66, + "learning_rate": 3.904526801744037e-05, + "loss": 0.4266, + "step": 17135 + }, + { + "epoch": 0.66, + "learning_rate": 3.904462682739164e-05, + "loss": 0.553, + "step": 17136 + }, + { + "epoch": 0.66, + "learning_rate": 3.9043985637342914e-05, + "loss": 0.53, + "step": 17137 + }, + { + "epoch": 0.66, + "learning_rate": 3.904334444729418e-05, + "loss": 0.5479, + "step": 17138 + }, + { + "epoch": 0.66, + "learning_rate": 3.904270325724545e-05, + "loss": 0.4305, + "step": 17139 + }, + { + "epoch": 0.66, + "learning_rate": 3.9042062067196716e-05, + "loss": 0.6807, + "step": 17140 + }, + { + "epoch": 0.66, + "learning_rate": 3.904142087714799e-05, + "loss": 0.5831, + "step": 17141 + }, + { + "epoch": 0.66, + "learning_rate": 3.904077968709926e-05, + "loss": 0.5633, + "step": 17142 + }, + { + "epoch": 0.66, + "learning_rate": 3.9040138497050524e-05, + "loss": 0.4933, + "step": 17143 + }, + { + "epoch": 0.66, + "learning_rate": 3.9039497307001796e-05, + "loss": 0.639, + "step": 17144 + }, + { + "epoch": 0.66, + "learning_rate": 3.903885611695307e-05, + "loss": 0.522, + "step": 17145 + }, + { + "epoch": 0.66, + "learning_rate": 3.903821492690434e-05, + "loss": 0.5415, + "step": 17146 + }, + { + "epoch": 0.66, + "learning_rate": 3.9037573736855604e-05, + "loss": 0.4923, + "step": 17147 + }, + { + "epoch": 0.66, + "learning_rate": 3.9036932546806876e-05, + "loss": 0.5441, + "step": 17148 + }, + { + "epoch": 0.66, + "learning_rate": 3.903629135675814e-05, + "loss": 0.5546, + "step": 17149 + }, + { + "epoch": 0.66, + "learning_rate": 3.903565016670941e-05, + "loss": 0.5118, + "step": 17150 + }, + { + "epoch": 0.66, + "learning_rate": 3.9035008976660684e-05, + "loss": 0.4969, + "step": 17151 + }, + { + "epoch": 0.66, + "learning_rate": 3.9034367786611956e-05, + "loss": 0.5237, + "step": 17152 + }, + { + "epoch": 0.66, + "learning_rate": 3.903372659656323e-05, + "loss": 0.503, + "step": 17153 + }, + { + "epoch": 0.66, + "learning_rate": 3.903308540651449e-05, + "loss": 0.52, + "step": 17154 + }, + { + "epoch": 0.66, + "learning_rate": 3.9032444216465764e-05, + "loss": 0.5003, + "step": 17155 + }, + { + "epoch": 0.66, + "learning_rate": 3.903180302641703e-05, + "loss": 0.5684, + "step": 17156 + }, + { + "epoch": 0.66, + "learning_rate": 3.90311618363683e-05, + "loss": 0.4535, + "step": 17157 + }, + { + "epoch": 0.66, + "learning_rate": 3.9030520646319566e-05, + "loss": 0.4871, + "step": 17158 + }, + { + "epoch": 0.66, + "learning_rate": 3.902987945627084e-05, + "loss": 0.4754, + "step": 17159 + }, + { + "epoch": 0.66, + "learning_rate": 3.902923826622211e-05, + "loss": 0.5213, + "step": 17160 + }, + { + "epoch": 0.66, + "learning_rate": 3.902859707617338e-05, + "loss": 0.4965, + "step": 17161 + }, + { + "epoch": 0.66, + "learning_rate": 3.902795588612465e-05, + "loss": 0.5092, + "step": 17162 + }, + { + "epoch": 0.66, + "learning_rate": 3.902731469607592e-05, + "loss": 0.4412, + "step": 17163 + }, + { + "epoch": 0.66, + "learning_rate": 3.902667350602719e-05, + "loss": 0.5046, + "step": 17164 + }, + { + "epoch": 0.66, + "learning_rate": 3.9026032315978454e-05, + "loss": 0.4918, + "step": 17165 + }, + { + "epoch": 0.66, + "learning_rate": 3.9025391125929726e-05, + "loss": 0.5161, + "step": 17166 + }, + { + "epoch": 0.66, + "learning_rate": 3.902474993588099e-05, + "loss": 0.5195, + "step": 17167 + }, + { + "epoch": 0.66, + "learning_rate": 3.902410874583227e-05, + "loss": 0.5892, + "step": 17168 + }, + { + "epoch": 0.66, + "learning_rate": 3.902346755578354e-05, + "loss": 0.5553, + "step": 17169 + }, + { + "epoch": 0.66, + "learning_rate": 3.9022826365734806e-05, + "loss": 0.443, + "step": 17170 + }, + { + "epoch": 0.66, + "learning_rate": 3.902218517568608e-05, + "loss": 0.5915, + "step": 17171 + }, + { + "epoch": 0.66, + "learning_rate": 3.902154398563734e-05, + "loss": 0.5834, + "step": 17172 + }, + { + "epoch": 0.66, + "learning_rate": 3.9020902795588615e-05, + "loss": 0.4884, + "step": 17173 + }, + { + "epoch": 0.66, + "learning_rate": 3.902026160553988e-05, + "loss": 0.5664, + "step": 17174 + }, + { + "epoch": 0.66, + "learning_rate": 3.901962041549115e-05, + "loss": 0.6784, + "step": 17175 + }, + { + "epoch": 0.66, + "learning_rate": 3.901897922544242e-05, + "loss": 0.6061, + "step": 17176 + }, + { + "epoch": 0.66, + "learning_rate": 3.9018338035393695e-05, + "loss": 0.5889, + "step": 17177 + }, + { + "epoch": 0.66, + "learning_rate": 3.9017696845344966e-05, + "loss": 0.4907, + "step": 17178 + }, + { + "epoch": 0.66, + "learning_rate": 3.901705565529623e-05, + "loss": 0.4922, + "step": 17179 + }, + { + "epoch": 0.66, + "learning_rate": 3.90164144652475e-05, + "loss": 0.4501, + "step": 17180 + }, + { + "epoch": 0.66, + "learning_rate": 3.901577327519877e-05, + "loss": 0.5656, + "step": 17181 + }, + { + "epoch": 0.66, + "learning_rate": 3.901513208515004e-05, + "loss": 0.5387, + "step": 17182 + }, + { + "epoch": 0.66, + "learning_rate": 3.9014490895101305e-05, + "loss": 0.5251, + "step": 17183 + }, + { + "epoch": 0.66, + "learning_rate": 3.901384970505258e-05, + "loss": 0.5163, + "step": 17184 + }, + { + "epoch": 0.66, + "learning_rate": 3.9013208515003855e-05, + "loss": 0.5438, + "step": 17185 + }, + { + "epoch": 0.66, + "learning_rate": 3.901256732495512e-05, + "loss": 0.5553, + "step": 17186 + }, + { + "epoch": 0.66, + "learning_rate": 3.901192613490639e-05, + "loss": 0.4399, + "step": 17187 + }, + { + "epoch": 0.66, + "learning_rate": 3.9011284944857656e-05, + "loss": 0.4926, + "step": 17188 + }, + { + "epoch": 0.66, + "learning_rate": 3.901064375480893e-05, + "loss": 0.5093, + "step": 17189 + }, + { + "epoch": 0.66, + "learning_rate": 3.901000256476019e-05, + "loss": 0.5477, + "step": 17190 + }, + { + "epoch": 0.66, + "learning_rate": 3.9009361374711465e-05, + "loss": 0.5201, + "step": 17191 + }, + { + "epoch": 0.66, + "learning_rate": 3.9008720184662736e-05, + "loss": 0.4921, + "step": 17192 + }, + { + "epoch": 0.66, + "learning_rate": 3.900807899461401e-05, + "loss": 0.5376, + "step": 17193 + }, + { + "epoch": 0.66, + "learning_rate": 3.900743780456528e-05, + "loss": 0.5749, + "step": 17194 + }, + { + "epoch": 0.66, + "learning_rate": 3.9006796614516545e-05, + "loss": 0.3744, + "step": 17195 + }, + { + "epoch": 0.66, + "learning_rate": 3.9006155424467817e-05, + "loss": 0.544, + "step": 17196 + }, + { + "epoch": 0.66, + "learning_rate": 3.900551423441908e-05, + "loss": 0.4356, + "step": 17197 + }, + { + "epoch": 0.66, + "learning_rate": 3.900487304437035e-05, + "loss": 0.3903, + "step": 17198 + }, + { + "epoch": 0.66, + "learning_rate": 3.900423185432162e-05, + "loss": 0.4466, + "step": 17199 + }, + { + "epoch": 0.66, + "learning_rate": 3.900359066427289e-05, + "loss": 0.4825, + "step": 17200 + }, + { + "epoch": 0.66, + "learning_rate": 3.900294947422416e-05, + "loss": 0.4269, + "step": 17201 + }, + { + "epoch": 0.66, + "learning_rate": 3.900230828417543e-05, + "loss": 0.5175, + "step": 17202 + }, + { + "epoch": 0.66, + "learning_rate": 3.9001667094126705e-05, + "loss": 0.6214, + "step": 17203 + }, + { + "epoch": 0.66, + "learning_rate": 3.900102590407797e-05, + "loss": 0.4581, + "step": 17204 + }, + { + "epoch": 0.66, + "learning_rate": 3.900038471402924e-05, + "loss": 0.4701, + "step": 17205 + }, + { + "epoch": 0.66, + "learning_rate": 3.899974352398051e-05, + "loss": 0.5621, + "step": 17206 + }, + { + "epoch": 0.66, + "learning_rate": 3.899910233393178e-05, + "loss": 0.5722, + "step": 17207 + }, + { + "epoch": 0.66, + "learning_rate": 3.899846114388305e-05, + "loss": 0.5957, + "step": 17208 + }, + { + "epoch": 0.66, + "learning_rate": 3.899781995383432e-05, + "loss": 0.5853, + "step": 17209 + }, + { + "epoch": 0.66, + "learning_rate": 3.899717876378559e-05, + "loss": 0.3955, + "step": 17210 + }, + { + "epoch": 0.66, + "learning_rate": 3.899653757373686e-05, + "loss": 0.5461, + "step": 17211 + }, + { + "epoch": 0.66, + "learning_rate": 3.899589638368813e-05, + "loss": 0.5205, + "step": 17212 + }, + { + "epoch": 0.66, + "learning_rate": 3.8995255193639395e-05, + "loss": 0.4543, + "step": 17213 + }, + { + "epoch": 0.66, + "learning_rate": 3.899461400359067e-05, + "loss": 0.5194, + "step": 17214 + }, + { + "epoch": 0.66, + "learning_rate": 3.899397281354193e-05, + "loss": 0.5307, + "step": 17215 + }, + { + "epoch": 0.66, + "learning_rate": 3.8993331623493203e-05, + "loss": 0.5093, + "step": 17216 + }, + { + "epoch": 0.66, + "learning_rate": 3.8992690433444475e-05, + "loss": 0.5481, + "step": 17217 + }, + { + "epoch": 0.66, + "learning_rate": 3.899204924339575e-05, + "loss": 0.5563, + "step": 17218 + }, + { + "epoch": 0.66, + "learning_rate": 3.899140805334701e-05, + "loss": 0.4247, + "step": 17219 + }, + { + "epoch": 0.66, + "learning_rate": 3.8990766863298284e-05, + "loss": 0.5183, + "step": 17220 + }, + { + "epoch": 0.66, + "learning_rate": 3.8990125673249555e-05, + "loss": 0.6224, + "step": 17221 + }, + { + "epoch": 0.66, + "learning_rate": 3.898948448320082e-05, + "loss": 0.4936, + "step": 17222 + }, + { + "epoch": 0.66, + "learning_rate": 3.898884329315209e-05, + "loss": 0.5212, + "step": 17223 + }, + { + "epoch": 0.66, + "learning_rate": 3.8988202103103364e-05, + "loss": 0.5148, + "step": 17224 + }, + { + "epoch": 0.66, + "learning_rate": 3.8987560913054635e-05, + "loss": 0.5391, + "step": 17225 + }, + { + "epoch": 0.66, + "learning_rate": 3.89869197230059e-05, + "loss": 0.5058, + "step": 17226 + }, + { + "epoch": 0.66, + "learning_rate": 3.898627853295717e-05, + "loss": 0.4876, + "step": 17227 + }, + { + "epoch": 0.66, + "learning_rate": 3.898563734290844e-05, + "loss": 0.5164, + "step": 17228 + }, + { + "epoch": 0.66, + "learning_rate": 3.898499615285971e-05, + "loss": 0.5519, + "step": 17229 + }, + { + "epoch": 0.66, + "learning_rate": 3.898435496281098e-05, + "loss": 0.4112, + "step": 17230 + }, + { + "epoch": 0.66, + "learning_rate": 3.8983713772762245e-05, + "loss": 0.4356, + "step": 17231 + }, + { + "epoch": 0.66, + "learning_rate": 3.898307258271352e-05, + "loss": 0.4912, + "step": 17232 + }, + { + "epoch": 0.66, + "learning_rate": 3.898243139266479e-05, + "loss": 0.5055, + "step": 17233 + }, + { + "epoch": 0.66, + "learning_rate": 3.898179020261606e-05, + "loss": 0.5308, + "step": 17234 + }, + { + "epoch": 0.66, + "learning_rate": 3.8981149012567325e-05, + "loss": 0.5637, + "step": 17235 + }, + { + "epoch": 0.66, + "learning_rate": 3.89805078225186e-05, + "loss": 0.5697, + "step": 17236 + }, + { + "epoch": 0.66, + "learning_rate": 3.897986663246986e-05, + "loss": 0.5803, + "step": 17237 + }, + { + "epoch": 0.66, + "learning_rate": 3.8979225442421134e-05, + "loss": 0.5311, + "step": 17238 + }, + { + "epoch": 0.66, + "learning_rate": 3.8978584252372406e-05, + "loss": 0.6067, + "step": 17239 + }, + { + "epoch": 0.66, + "learning_rate": 3.897794306232368e-05, + "loss": 0.5164, + "step": 17240 + }, + { + "epoch": 0.66, + "learning_rate": 3.897730187227495e-05, + "loss": 0.4965, + "step": 17241 + }, + { + "epoch": 0.66, + "learning_rate": 3.8976660682226214e-05, + "loss": 0.4547, + "step": 17242 + }, + { + "epoch": 0.66, + "learning_rate": 3.8976019492177486e-05, + "loss": 0.4845, + "step": 17243 + }, + { + "epoch": 0.66, + "learning_rate": 3.897537830212875e-05, + "loss": 0.4929, + "step": 17244 + }, + { + "epoch": 0.66, + "learning_rate": 3.897473711208002e-05, + "loss": 0.5916, + "step": 17245 + }, + { + "epoch": 0.66, + "learning_rate": 3.897409592203129e-05, + "loss": 0.5459, + "step": 17246 + }, + { + "epoch": 0.66, + "learning_rate": 3.897345473198256e-05, + "loss": 0.5229, + "step": 17247 + }, + { + "epoch": 0.66, + "learning_rate": 3.897281354193383e-05, + "loss": 0.5499, + "step": 17248 + }, + { + "epoch": 0.66, + "learning_rate": 3.89721723518851e-05, + "loss": 0.6194, + "step": 17249 + }, + { + "epoch": 0.66, + "learning_rate": 3.8971531161836374e-05, + "loss": 0.5527, + "step": 17250 + }, + { + "epoch": 0.66, + "learning_rate": 3.897088997178764e-05, + "loss": 0.5045, + "step": 17251 + }, + { + "epoch": 0.66, + "learning_rate": 3.897024878173891e-05, + "loss": 0.518, + "step": 17252 + }, + { + "epoch": 0.66, + "learning_rate": 3.8969607591690176e-05, + "loss": 0.4945, + "step": 17253 + }, + { + "epoch": 0.66, + "learning_rate": 3.896896640164145e-05, + "loss": 0.6265, + "step": 17254 + }, + { + "epoch": 0.66, + "learning_rate": 3.896832521159271e-05, + "loss": 0.5952, + "step": 17255 + }, + { + "epoch": 0.66, + "learning_rate": 3.896768402154399e-05, + "loss": 0.64, + "step": 17256 + }, + { + "epoch": 0.66, + "learning_rate": 3.896704283149526e-05, + "loss": 0.5563, + "step": 17257 + }, + { + "epoch": 0.66, + "learning_rate": 3.896640164144653e-05, + "loss": 0.5035, + "step": 17258 + }, + { + "epoch": 0.66, + "learning_rate": 3.89657604513978e-05, + "loss": 0.5829, + "step": 17259 + }, + { + "epoch": 0.66, + "learning_rate": 3.8965119261349064e-05, + "loss": 0.4479, + "step": 17260 + }, + { + "epoch": 0.66, + "learning_rate": 3.8964478071300336e-05, + "loss": 0.5024, + "step": 17261 + }, + { + "epoch": 0.66, + "learning_rate": 3.89638368812516e-05, + "loss": 0.4987, + "step": 17262 + }, + { + "epoch": 0.66, + "learning_rate": 3.896319569120287e-05, + "loss": 0.5888, + "step": 17263 + }, + { + "epoch": 0.66, + "learning_rate": 3.8962554501154144e-05, + "loss": 0.6661, + "step": 17264 + }, + { + "epoch": 0.66, + "learning_rate": 3.8961913311105416e-05, + "loss": 0.6267, + "step": 17265 + }, + { + "epoch": 0.66, + "learning_rate": 3.896127212105669e-05, + "loss": 0.4373, + "step": 17266 + }, + { + "epoch": 0.66, + "learning_rate": 3.896063093100795e-05, + "loss": 0.5598, + "step": 17267 + }, + { + "epoch": 0.66, + "learning_rate": 3.8959989740959224e-05, + "loss": 0.4398, + "step": 17268 + }, + { + "epoch": 0.66, + "learning_rate": 3.895934855091049e-05, + "loss": 0.6598, + "step": 17269 + }, + { + "epoch": 0.66, + "learning_rate": 3.895870736086176e-05, + "loss": 0.4336, + "step": 17270 + }, + { + "epoch": 0.66, + "learning_rate": 3.8958066170813026e-05, + "loss": 0.5355, + "step": 17271 + }, + { + "epoch": 0.66, + "learning_rate": 3.8957424980764304e-05, + "loss": 0.4827, + "step": 17272 + }, + { + "epoch": 0.66, + "learning_rate": 3.895678379071557e-05, + "loss": 0.4543, + "step": 17273 + }, + { + "epoch": 0.66, + "learning_rate": 3.895614260066684e-05, + "loss": 0.5972, + "step": 17274 + }, + { + "epoch": 0.66, + "learning_rate": 3.895550141061811e-05, + "loss": 0.5495, + "step": 17275 + }, + { + "epoch": 0.66, + "learning_rate": 3.895486022056938e-05, + "loss": 0.5486, + "step": 17276 + }, + { + "epoch": 0.66, + "learning_rate": 3.895421903052065e-05, + "loss": 0.582, + "step": 17277 + }, + { + "epoch": 0.66, + "learning_rate": 3.8953577840471914e-05, + "loss": 0.6376, + "step": 17278 + }, + { + "epoch": 0.66, + "learning_rate": 3.8952936650423186e-05, + "loss": 0.4534, + "step": 17279 + }, + { + "epoch": 0.66, + "learning_rate": 3.895229546037446e-05, + "loss": 0.5017, + "step": 17280 + }, + { + "epoch": 0.66, + "learning_rate": 3.895165427032573e-05, + "loss": 0.5571, + "step": 17281 + }, + { + "epoch": 0.66, + "learning_rate": 3.8951013080276994e-05, + "loss": 0.5336, + "step": 17282 + }, + { + "epoch": 0.66, + "learning_rate": 3.8950371890228266e-05, + "loss": 0.5235, + "step": 17283 + }, + { + "epoch": 0.66, + "learning_rate": 3.894973070017954e-05, + "loss": 0.4323, + "step": 17284 + }, + { + "epoch": 0.66, + "learning_rate": 3.89490895101308e-05, + "loss": 0.5054, + "step": 17285 + }, + { + "epoch": 0.66, + "learning_rate": 3.8948448320082075e-05, + "loss": 0.5807, + "step": 17286 + }, + { + "epoch": 0.66, + "learning_rate": 3.894780713003334e-05, + "loss": 0.5373, + "step": 17287 + }, + { + "epoch": 0.66, + "learning_rate": 3.894716593998461e-05, + "loss": 0.4932, + "step": 17288 + }, + { + "epoch": 0.66, + "learning_rate": 3.894652474993588e-05, + "loss": 0.5088, + "step": 17289 + }, + { + "epoch": 0.66, + "learning_rate": 3.8945883559887155e-05, + "loss": 0.5242, + "step": 17290 + }, + { + "epoch": 0.66, + "learning_rate": 3.894524236983842e-05, + "loss": 0.5968, + "step": 17291 + }, + { + "epoch": 0.66, + "learning_rate": 3.894460117978969e-05, + "loss": 0.6465, + "step": 17292 + }, + { + "epoch": 0.66, + "learning_rate": 3.894395998974096e-05, + "loss": 0.4988, + "step": 17293 + }, + { + "epoch": 0.66, + "learning_rate": 3.894331879969223e-05, + "loss": 0.5998, + "step": 17294 + }, + { + "epoch": 0.66, + "learning_rate": 3.89426776096435e-05, + "loss": 0.5374, + "step": 17295 + }, + { + "epoch": 0.66, + "learning_rate": 3.894203641959477e-05, + "loss": 0.4351, + "step": 17296 + }, + { + "epoch": 0.67, + "learning_rate": 3.894139522954604e-05, + "loss": 0.532, + "step": 17297 + }, + { + "epoch": 0.67, + "learning_rate": 3.894075403949731e-05, + "loss": 0.6338, + "step": 17298 + }, + { + "epoch": 0.67, + "learning_rate": 3.894011284944858e-05, + "loss": 0.4529, + "step": 17299 + }, + { + "epoch": 0.67, + "learning_rate": 3.8939471659399845e-05, + "loss": 0.5602, + "step": 17300 + }, + { + "epoch": 0.67, + "learning_rate": 3.8938830469351116e-05, + "loss": 0.5725, + "step": 17301 + }, + { + "epoch": 0.67, + "learning_rate": 3.893818927930239e-05, + "loss": 0.5576, + "step": 17302 + }, + { + "epoch": 0.67, + "learning_rate": 3.893754808925365e-05, + "loss": 0.5759, + "step": 17303 + }, + { + "epoch": 0.67, + "learning_rate": 3.8936906899204925e-05, + "loss": 0.4889, + "step": 17304 + }, + { + "epoch": 0.67, + "learning_rate": 3.8936265709156197e-05, + "loss": 0.6052, + "step": 17305 + }, + { + "epoch": 0.67, + "learning_rate": 3.893562451910747e-05, + "loss": 0.5996, + "step": 17306 + }, + { + "epoch": 0.67, + "learning_rate": 3.893498332905873e-05, + "loss": 0.6268, + "step": 17307 + }, + { + "epoch": 0.67, + "learning_rate": 3.8934342139010005e-05, + "loss": 0.5218, + "step": 17308 + }, + { + "epoch": 0.67, + "learning_rate": 3.893370094896127e-05, + "loss": 0.5244, + "step": 17309 + }, + { + "epoch": 0.67, + "learning_rate": 3.893305975891254e-05, + "loss": 0.4725, + "step": 17310 + }, + { + "epoch": 0.67, + "learning_rate": 3.893241856886381e-05, + "loss": 0.4496, + "step": 17311 + }, + { + "epoch": 0.67, + "learning_rate": 3.8931777378815085e-05, + "loss": 0.596, + "step": 17312 + }, + { + "epoch": 0.67, + "learning_rate": 3.893113618876636e-05, + "loss": 0.4461, + "step": 17313 + }, + { + "epoch": 0.67, + "learning_rate": 3.893049499871762e-05, + "loss": 0.5595, + "step": 17314 + }, + { + "epoch": 0.67, + "learning_rate": 3.892985380866889e-05, + "loss": 0.5392, + "step": 17315 + }, + { + "epoch": 0.67, + "learning_rate": 3.892921261862016e-05, + "loss": 0.4643, + "step": 17316 + }, + { + "epoch": 0.67, + "learning_rate": 3.892857142857143e-05, + "loss": 0.5986, + "step": 17317 + }, + { + "epoch": 0.67, + "learning_rate": 3.8927930238522695e-05, + "loss": 0.4795, + "step": 17318 + }, + { + "epoch": 0.67, + "learning_rate": 3.892728904847397e-05, + "loss": 0.5105, + "step": 17319 + }, + { + "epoch": 0.67, + "learning_rate": 3.892664785842524e-05, + "loss": 0.4905, + "step": 17320 + }, + { + "epoch": 0.67, + "learning_rate": 3.892600666837651e-05, + "loss": 0.5628, + "step": 17321 + }, + { + "epoch": 0.67, + "learning_rate": 3.892536547832778e-05, + "loss": 0.4473, + "step": 17322 + }, + { + "epoch": 0.67, + "learning_rate": 3.892472428827905e-05, + "loss": 0.4978, + "step": 17323 + }, + { + "epoch": 0.67, + "learning_rate": 3.892408309823032e-05, + "loss": 0.5783, + "step": 17324 + }, + { + "epoch": 0.67, + "learning_rate": 3.8923441908181583e-05, + "loss": 0.5056, + "step": 17325 + }, + { + "epoch": 0.67, + "learning_rate": 3.8922800718132855e-05, + "loss": 0.4889, + "step": 17326 + }, + { + "epoch": 0.67, + "learning_rate": 3.892215952808412e-05, + "loss": 0.4627, + "step": 17327 + }, + { + "epoch": 0.67, + "learning_rate": 3.89215183380354e-05, + "loss": 0.4821, + "step": 17328 + }, + { + "epoch": 0.67, + "learning_rate": 3.892087714798667e-05, + "loss": 0.5096, + "step": 17329 + }, + { + "epoch": 0.67, + "learning_rate": 3.8920235957937935e-05, + "loss": 0.5524, + "step": 17330 + }, + { + "epoch": 0.67, + "learning_rate": 3.891959476788921e-05, + "loss": 0.5029, + "step": 17331 + }, + { + "epoch": 0.67, + "learning_rate": 3.891895357784047e-05, + "loss": 0.5511, + "step": 17332 + }, + { + "epoch": 0.67, + "learning_rate": 3.8918312387791744e-05, + "loss": 0.4476, + "step": 17333 + }, + { + "epoch": 0.67, + "learning_rate": 3.891767119774301e-05, + "loss": 0.5733, + "step": 17334 + }, + { + "epoch": 0.67, + "learning_rate": 3.891703000769428e-05, + "loss": 0.5775, + "step": 17335 + }, + { + "epoch": 0.67, + "learning_rate": 3.891638881764555e-05, + "loss": 0.4891, + "step": 17336 + }, + { + "epoch": 0.67, + "learning_rate": 3.8915747627596824e-05, + "loss": 0.4757, + "step": 17337 + }, + { + "epoch": 0.67, + "learning_rate": 3.8915106437548095e-05, + "loss": 0.4565, + "step": 17338 + }, + { + "epoch": 0.67, + "learning_rate": 3.891446524749936e-05, + "loss": 0.5359, + "step": 17339 + }, + { + "epoch": 0.67, + "learning_rate": 3.891382405745063e-05, + "loss": 0.5449, + "step": 17340 + }, + { + "epoch": 0.67, + "learning_rate": 3.89131828674019e-05, + "loss": 0.6006, + "step": 17341 + }, + { + "epoch": 0.67, + "learning_rate": 3.891254167735317e-05, + "loss": 0.4249, + "step": 17342 + }, + { + "epoch": 0.67, + "learning_rate": 3.8911900487304434e-05, + "loss": 0.6339, + "step": 17343 + }, + { + "epoch": 0.67, + "learning_rate": 3.891125929725571e-05, + "loss": 0.5066, + "step": 17344 + }, + { + "epoch": 0.67, + "learning_rate": 3.891061810720698e-05, + "loss": 0.5179, + "step": 17345 + }, + { + "epoch": 0.67, + "learning_rate": 3.890997691715825e-05, + "loss": 0.5266, + "step": 17346 + }, + { + "epoch": 0.67, + "learning_rate": 3.890933572710952e-05, + "loss": 0.6307, + "step": 17347 + }, + { + "epoch": 0.67, + "learning_rate": 3.8908694537060785e-05, + "loss": 0.6583, + "step": 17348 + }, + { + "epoch": 0.67, + "learning_rate": 3.890805334701206e-05, + "loss": 0.5502, + "step": 17349 + }, + { + "epoch": 0.67, + "learning_rate": 3.890741215696332e-05, + "loss": 0.5745, + "step": 17350 + }, + { + "epoch": 0.67, + "learning_rate": 3.8906770966914594e-05, + "loss": 0.5381, + "step": 17351 + }, + { + "epoch": 0.67, + "learning_rate": 3.8906129776865866e-05, + "loss": 0.4056, + "step": 17352 + }, + { + "epoch": 0.67, + "learning_rate": 3.890548858681714e-05, + "loss": 0.4357, + "step": 17353 + }, + { + "epoch": 0.67, + "learning_rate": 3.89048473967684e-05, + "loss": 0.51, + "step": 17354 + }, + { + "epoch": 0.67, + "learning_rate": 3.8904206206719674e-05, + "loss": 0.6044, + "step": 17355 + }, + { + "epoch": 0.67, + "learning_rate": 3.8903565016670946e-05, + "loss": 0.5561, + "step": 17356 + }, + { + "epoch": 0.67, + "learning_rate": 3.890292382662221e-05, + "loss": 0.3362, + "step": 17357 + }, + { + "epoch": 0.67, + "learning_rate": 3.890228263657348e-05, + "loss": 0.456, + "step": 17358 + }, + { + "epoch": 0.67, + "learning_rate": 3.890164144652475e-05, + "loss": 0.4977, + "step": 17359 + }, + { + "epoch": 0.67, + "learning_rate": 3.890100025647602e-05, + "loss": 0.5167, + "step": 17360 + }, + { + "epoch": 0.67, + "learning_rate": 3.890035906642729e-05, + "loss": 0.5744, + "step": 17361 + }, + { + "epoch": 0.67, + "learning_rate": 3.889971787637856e-05, + "loss": 0.4778, + "step": 17362 + }, + { + "epoch": 0.67, + "learning_rate": 3.889907668632983e-05, + "loss": 0.4934, + "step": 17363 + }, + { + "epoch": 0.67, + "learning_rate": 3.88984354962811e-05, + "loss": 0.518, + "step": 17364 + }, + { + "epoch": 0.67, + "learning_rate": 3.889779430623237e-05, + "loss": 0.5048, + "step": 17365 + }, + { + "epoch": 0.67, + "learning_rate": 3.8897153116183636e-05, + "loss": 0.4883, + "step": 17366 + }, + { + "epoch": 0.67, + "learning_rate": 3.889651192613491e-05, + "loss": 0.4482, + "step": 17367 + }, + { + "epoch": 0.67, + "learning_rate": 3.889587073608618e-05, + "loss": 0.5561, + "step": 17368 + }, + { + "epoch": 0.67, + "learning_rate": 3.889522954603745e-05, + "loss": 0.6091, + "step": 17369 + }, + { + "epoch": 0.67, + "learning_rate": 3.8894588355988716e-05, + "loss": 0.5076, + "step": 17370 + }, + { + "epoch": 0.67, + "learning_rate": 3.889394716593999e-05, + "loss": 0.5205, + "step": 17371 + }, + { + "epoch": 0.67, + "learning_rate": 3.889330597589126e-05, + "loss": 0.5108, + "step": 17372 + }, + { + "epoch": 0.67, + "learning_rate": 3.8892664785842524e-05, + "loss": 0.4846, + "step": 17373 + }, + { + "epoch": 0.67, + "learning_rate": 3.8892023595793796e-05, + "loss": 0.4661, + "step": 17374 + }, + { + "epoch": 0.67, + "learning_rate": 3.889138240574506e-05, + "loss": 0.5194, + "step": 17375 + }, + { + "epoch": 0.67, + "learning_rate": 3.889074121569633e-05, + "loss": 0.3926, + "step": 17376 + }, + { + "epoch": 0.67, + "learning_rate": 3.8890100025647604e-05, + "loss": 0.4978, + "step": 17377 + }, + { + "epoch": 0.67, + "learning_rate": 3.8889458835598876e-05, + "loss": 0.5795, + "step": 17378 + }, + { + "epoch": 0.67, + "learning_rate": 3.888881764555014e-05, + "loss": 0.5617, + "step": 17379 + }, + { + "epoch": 0.67, + "learning_rate": 3.888817645550141e-05, + "loss": 0.6181, + "step": 17380 + }, + { + "epoch": 0.67, + "learning_rate": 3.8887535265452684e-05, + "loss": 0.4495, + "step": 17381 + }, + { + "epoch": 0.67, + "learning_rate": 3.888689407540395e-05, + "loss": 0.4625, + "step": 17382 + }, + { + "epoch": 0.67, + "learning_rate": 3.888625288535522e-05, + "loss": 0.5252, + "step": 17383 + }, + { + "epoch": 0.67, + "learning_rate": 3.888561169530649e-05, + "loss": 0.538, + "step": 17384 + }, + { + "epoch": 0.67, + "learning_rate": 3.8884970505257764e-05, + "loss": 0.5724, + "step": 17385 + }, + { + "epoch": 0.67, + "learning_rate": 3.888432931520903e-05, + "loss": 0.4672, + "step": 17386 + }, + { + "epoch": 0.67, + "learning_rate": 3.88836881251603e-05, + "loss": 0.5618, + "step": 17387 + }, + { + "epoch": 0.67, + "learning_rate": 3.8883046935111566e-05, + "loss": 0.5129, + "step": 17388 + }, + { + "epoch": 0.67, + "learning_rate": 3.888240574506284e-05, + "loss": 0.5442, + "step": 17389 + }, + { + "epoch": 0.67, + "learning_rate": 3.888176455501411e-05, + "loss": 0.5417, + "step": 17390 + }, + { + "epoch": 0.67, + "learning_rate": 3.8881123364965374e-05, + "loss": 0.5397, + "step": 17391 + }, + { + "epoch": 0.67, + "learning_rate": 3.8880482174916646e-05, + "loss": 0.5092, + "step": 17392 + }, + { + "epoch": 0.67, + "learning_rate": 3.887984098486792e-05, + "loss": 0.4748, + "step": 17393 + }, + { + "epoch": 0.67, + "learning_rate": 3.887919979481919e-05, + "loss": 0.4943, + "step": 17394 + }, + { + "epoch": 0.67, + "learning_rate": 3.8878558604770454e-05, + "loss": 0.3912, + "step": 17395 + }, + { + "epoch": 0.67, + "learning_rate": 3.8877917414721726e-05, + "loss": 0.5435, + "step": 17396 + }, + { + "epoch": 0.67, + "learning_rate": 3.887727622467299e-05, + "loss": 0.5125, + "step": 17397 + }, + { + "epoch": 0.67, + "learning_rate": 3.887663503462426e-05, + "loss": 0.5795, + "step": 17398 + }, + { + "epoch": 0.67, + "learning_rate": 3.8875993844575535e-05, + "loss": 0.5246, + "step": 17399 + }, + { + "epoch": 0.67, + "learning_rate": 3.8875352654526806e-05, + "loss": 0.5276, + "step": 17400 + }, + { + "epoch": 0.67, + "learning_rate": 3.887471146447808e-05, + "loss": 0.4073, + "step": 17401 + }, + { + "epoch": 0.67, + "learning_rate": 3.887407027442934e-05, + "loss": 0.5642, + "step": 17402 + }, + { + "epoch": 0.67, + "learning_rate": 3.8873429084380615e-05, + "loss": 0.5281, + "step": 17403 + }, + { + "epoch": 0.67, + "learning_rate": 3.887278789433188e-05, + "loss": 0.5193, + "step": 17404 + }, + { + "epoch": 0.67, + "learning_rate": 3.887214670428315e-05, + "loss": 0.5707, + "step": 17405 + }, + { + "epoch": 0.67, + "learning_rate": 3.8871505514234416e-05, + "loss": 0.6435, + "step": 17406 + }, + { + "epoch": 0.67, + "learning_rate": 3.887086432418569e-05, + "loss": 0.5496, + "step": 17407 + }, + { + "epoch": 0.67, + "learning_rate": 3.887022313413696e-05, + "loss": 0.4666, + "step": 17408 + }, + { + "epoch": 0.67, + "learning_rate": 3.886958194408823e-05, + "loss": 0.4901, + "step": 17409 + }, + { + "epoch": 0.67, + "learning_rate": 3.88689407540395e-05, + "loss": 0.4237, + "step": 17410 + }, + { + "epoch": 0.67, + "learning_rate": 3.886829956399077e-05, + "loss": 0.6249, + "step": 17411 + }, + { + "epoch": 0.67, + "learning_rate": 3.886765837394204e-05, + "loss": 0.5226, + "step": 17412 + }, + { + "epoch": 0.67, + "learning_rate": 3.8867017183893305e-05, + "loss": 0.5109, + "step": 17413 + }, + { + "epoch": 0.67, + "learning_rate": 3.8866375993844576e-05, + "loss": 0.5213, + "step": 17414 + }, + { + "epoch": 0.67, + "learning_rate": 3.886573480379584e-05, + "loss": 0.4638, + "step": 17415 + }, + { + "epoch": 0.67, + "learning_rate": 3.886509361374712e-05, + "loss": 0.5927, + "step": 17416 + }, + { + "epoch": 0.67, + "learning_rate": 3.886445242369839e-05, + "loss": 0.4967, + "step": 17417 + }, + { + "epoch": 0.67, + "learning_rate": 3.8863811233649657e-05, + "loss": 0.5128, + "step": 17418 + }, + { + "epoch": 0.67, + "learning_rate": 3.886317004360093e-05, + "loss": 0.5964, + "step": 17419 + }, + { + "epoch": 0.67, + "learning_rate": 3.886252885355219e-05, + "loss": 0.5066, + "step": 17420 + }, + { + "epoch": 0.67, + "learning_rate": 3.8861887663503465e-05, + "loss": 0.4993, + "step": 17421 + }, + { + "epoch": 0.67, + "learning_rate": 3.886124647345473e-05, + "loss": 0.5141, + "step": 17422 + }, + { + "epoch": 0.67, + "learning_rate": 3.8860605283406e-05, + "loss": 0.5095, + "step": 17423 + }, + { + "epoch": 0.67, + "learning_rate": 3.885996409335727e-05, + "loss": 0.4787, + "step": 17424 + }, + { + "epoch": 0.67, + "learning_rate": 3.8859322903308545e-05, + "loss": 0.4767, + "step": 17425 + }, + { + "epoch": 0.67, + "learning_rate": 3.885868171325982e-05, + "loss": 0.6252, + "step": 17426 + }, + { + "epoch": 0.67, + "learning_rate": 3.885804052321108e-05, + "loss": 0.455, + "step": 17427 + }, + { + "epoch": 0.67, + "learning_rate": 3.885739933316235e-05, + "loss": 0.4777, + "step": 17428 + }, + { + "epoch": 0.67, + "learning_rate": 3.885675814311362e-05, + "loss": 0.4842, + "step": 17429 + }, + { + "epoch": 0.67, + "learning_rate": 3.885611695306489e-05, + "loss": 0.4196, + "step": 17430 + }, + { + "epoch": 0.67, + "learning_rate": 3.8855475763016155e-05, + "loss": 0.4797, + "step": 17431 + }, + { + "epoch": 0.67, + "learning_rate": 3.8854834572967433e-05, + "loss": 0.5769, + "step": 17432 + }, + { + "epoch": 0.67, + "learning_rate": 3.88541933829187e-05, + "loss": 0.5777, + "step": 17433 + }, + { + "epoch": 0.67, + "learning_rate": 3.885355219286997e-05, + "loss": 0.4329, + "step": 17434 + }, + { + "epoch": 0.67, + "learning_rate": 3.885291100282124e-05, + "loss": 0.4725, + "step": 17435 + }, + { + "epoch": 0.67, + "learning_rate": 3.885226981277251e-05, + "loss": 0.4648, + "step": 17436 + }, + { + "epoch": 0.67, + "learning_rate": 3.885162862272378e-05, + "loss": 0.4931, + "step": 17437 + }, + { + "epoch": 0.67, + "learning_rate": 3.8850987432675043e-05, + "loss": 0.5187, + "step": 17438 + }, + { + "epoch": 0.67, + "learning_rate": 3.8850346242626315e-05, + "loss": 0.5626, + "step": 17439 + }, + { + "epoch": 0.67, + "learning_rate": 3.884970505257759e-05, + "loss": 0.5124, + "step": 17440 + }, + { + "epoch": 0.67, + "learning_rate": 3.884906386252886e-05, + "loss": 0.5521, + "step": 17441 + }, + { + "epoch": 0.67, + "learning_rate": 3.8848422672480124e-05, + "loss": 0.7139, + "step": 17442 + }, + { + "epoch": 0.67, + "learning_rate": 3.8847781482431395e-05, + "loss": 0.4839, + "step": 17443 + }, + { + "epoch": 0.67, + "learning_rate": 3.884714029238267e-05, + "loss": 0.5563, + "step": 17444 + }, + { + "epoch": 0.67, + "learning_rate": 3.884649910233393e-05, + "loss": 0.4584, + "step": 17445 + }, + { + "epoch": 0.67, + "learning_rate": 3.8845857912285204e-05, + "loss": 0.4941, + "step": 17446 + }, + { + "epoch": 0.67, + "learning_rate": 3.884521672223647e-05, + "loss": 0.4492, + "step": 17447 + }, + { + "epoch": 0.67, + "learning_rate": 3.884457553218774e-05, + "loss": 0.4644, + "step": 17448 + }, + { + "epoch": 0.67, + "learning_rate": 3.884393434213901e-05, + "loss": 0.5449, + "step": 17449 + }, + { + "epoch": 0.67, + "learning_rate": 3.8843293152090284e-05, + "loss": 0.4575, + "step": 17450 + }, + { + "epoch": 0.67, + "learning_rate": 3.884265196204155e-05, + "loss": 0.5072, + "step": 17451 + }, + { + "epoch": 0.67, + "learning_rate": 3.884201077199282e-05, + "loss": 0.5049, + "step": 17452 + }, + { + "epoch": 0.67, + "learning_rate": 3.884136958194409e-05, + "loss": 0.6279, + "step": 17453 + }, + { + "epoch": 0.67, + "learning_rate": 3.884072839189536e-05, + "loss": 0.4286, + "step": 17454 + }, + { + "epoch": 0.67, + "learning_rate": 3.884008720184663e-05, + "loss": 0.6698, + "step": 17455 + }, + { + "epoch": 0.67, + "learning_rate": 3.88394460117979e-05, + "loss": 0.4574, + "step": 17456 + }, + { + "epoch": 0.67, + "learning_rate": 3.883880482174917e-05, + "loss": 0.5107, + "step": 17457 + }, + { + "epoch": 0.67, + "learning_rate": 3.883816363170044e-05, + "loss": 0.5415, + "step": 17458 + }, + { + "epoch": 0.67, + "learning_rate": 3.883752244165171e-05, + "loss": 0.5198, + "step": 17459 + }, + { + "epoch": 0.67, + "learning_rate": 3.8836881251602974e-05, + "loss": 0.529, + "step": 17460 + }, + { + "epoch": 0.67, + "learning_rate": 3.8836240061554245e-05, + "loss": 0.4389, + "step": 17461 + }, + { + "epoch": 0.67, + "learning_rate": 3.883559887150552e-05, + "loss": 0.5093, + "step": 17462 + }, + { + "epoch": 0.67, + "learning_rate": 3.883495768145678e-05, + "loss": 0.6256, + "step": 17463 + }, + { + "epoch": 0.67, + "learning_rate": 3.8834316491408054e-05, + "loss": 0.541, + "step": 17464 + }, + { + "epoch": 0.67, + "learning_rate": 3.8833675301359326e-05, + "loss": 0.5333, + "step": 17465 + }, + { + "epoch": 0.67, + "learning_rate": 3.88330341113106e-05, + "loss": 0.4642, + "step": 17466 + }, + { + "epoch": 0.67, + "learning_rate": 3.883239292126186e-05, + "loss": 0.5802, + "step": 17467 + }, + { + "epoch": 0.67, + "learning_rate": 3.8831751731213134e-05, + "loss": 0.6078, + "step": 17468 + }, + { + "epoch": 0.67, + "learning_rate": 3.88311105411644e-05, + "loss": 0.602, + "step": 17469 + }, + { + "epoch": 0.67, + "learning_rate": 3.883046935111567e-05, + "loss": 0.5547, + "step": 17470 + }, + { + "epoch": 0.67, + "learning_rate": 3.882982816106694e-05, + "loss": 0.4929, + "step": 17471 + }, + { + "epoch": 0.67, + "learning_rate": 3.8829186971018214e-05, + "loss": 0.56, + "step": 17472 + }, + { + "epoch": 0.67, + "learning_rate": 3.8828545780969486e-05, + "loss": 0.5524, + "step": 17473 + }, + { + "epoch": 0.67, + "learning_rate": 3.882790459092075e-05, + "loss": 0.4215, + "step": 17474 + }, + { + "epoch": 0.67, + "learning_rate": 3.882726340087202e-05, + "loss": 0.5048, + "step": 17475 + }, + { + "epoch": 0.67, + "learning_rate": 3.882662221082329e-05, + "loss": 0.5057, + "step": 17476 + }, + { + "epoch": 0.67, + "learning_rate": 3.882598102077456e-05, + "loss": 0.5547, + "step": 17477 + }, + { + "epoch": 0.67, + "learning_rate": 3.8825339830725824e-05, + "loss": 0.7576, + "step": 17478 + }, + { + "epoch": 0.67, + "learning_rate": 3.8824698640677096e-05, + "loss": 0.6253, + "step": 17479 + }, + { + "epoch": 0.67, + "learning_rate": 3.882405745062837e-05, + "loss": 0.5474, + "step": 17480 + }, + { + "epoch": 0.67, + "learning_rate": 3.882341626057964e-05, + "loss": 0.5152, + "step": 17481 + }, + { + "epoch": 0.67, + "learning_rate": 3.882277507053091e-05, + "loss": 0.5308, + "step": 17482 + }, + { + "epoch": 0.67, + "learning_rate": 3.8822133880482176e-05, + "loss": 0.5947, + "step": 17483 + }, + { + "epoch": 0.67, + "learning_rate": 3.882149269043345e-05, + "loss": 0.5842, + "step": 17484 + }, + { + "epoch": 0.67, + "learning_rate": 3.882085150038471e-05, + "loss": 0.4007, + "step": 17485 + }, + { + "epoch": 0.67, + "learning_rate": 3.8820210310335984e-05, + "loss": 0.5963, + "step": 17486 + }, + { + "epoch": 0.67, + "learning_rate": 3.881956912028725e-05, + "loss": 0.4655, + "step": 17487 + }, + { + "epoch": 0.67, + "learning_rate": 3.881892793023853e-05, + "loss": 0.5785, + "step": 17488 + }, + { + "epoch": 0.67, + "learning_rate": 3.88182867401898e-05, + "loss": 0.5003, + "step": 17489 + }, + { + "epoch": 0.67, + "learning_rate": 3.8817645550141064e-05, + "loss": 0.6, + "step": 17490 + }, + { + "epoch": 0.67, + "learning_rate": 3.8817004360092336e-05, + "loss": 0.5933, + "step": 17491 + }, + { + "epoch": 0.67, + "learning_rate": 3.88163631700436e-05, + "loss": 0.4853, + "step": 17492 + }, + { + "epoch": 0.67, + "learning_rate": 3.881572197999487e-05, + "loss": 0.59, + "step": 17493 + }, + { + "epoch": 0.67, + "learning_rate": 3.881508078994614e-05, + "loss": 0.4743, + "step": 17494 + }, + { + "epoch": 0.67, + "learning_rate": 3.881443959989741e-05, + "loss": 0.4195, + "step": 17495 + }, + { + "epoch": 0.67, + "learning_rate": 3.881379840984868e-05, + "loss": 0.6154, + "step": 17496 + }, + { + "epoch": 0.67, + "learning_rate": 3.881315721979995e-05, + "loss": 0.5362, + "step": 17497 + }, + { + "epoch": 0.67, + "learning_rate": 3.8812516029751224e-05, + "loss": 0.3976, + "step": 17498 + }, + { + "epoch": 0.67, + "learning_rate": 3.881187483970249e-05, + "loss": 0.4007, + "step": 17499 + }, + { + "epoch": 0.67, + "learning_rate": 3.881123364965376e-05, + "loss": 0.5911, + "step": 17500 + }, + { + "epoch": 0.67, + "learning_rate": 3.8810592459605026e-05, + "loss": 0.4579, + "step": 17501 + }, + { + "epoch": 0.67, + "learning_rate": 3.88099512695563e-05, + "loss": 0.6188, + "step": 17502 + }, + { + "epoch": 0.67, + "learning_rate": 3.880931007950756e-05, + "loss": 0.4969, + "step": 17503 + }, + { + "epoch": 0.67, + "learning_rate": 3.880866888945884e-05, + "loss": 0.616, + "step": 17504 + }, + { + "epoch": 0.67, + "learning_rate": 3.8808027699410106e-05, + "loss": 0.4425, + "step": 17505 + }, + { + "epoch": 0.67, + "learning_rate": 3.880738650936138e-05, + "loss": 0.4699, + "step": 17506 + }, + { + "epoch": 0.67, + "learning_rate": 3.880674531931265e-05, + "loss": 0.552, + "step": 17507 + }, + { + "epoch": 0.67, + "learning_rate": 3.8806104129263915e-05, + "loss": 0.4745, + "step": 17508 + }, + { + "epoch": 0.67, + "learning_rate": 3.8805462939215186e-05, + "loss": 0.5721, + "step": 17509 + }, + { + "epoch": 0.67, + "learning_rate": 3.880482174916645e-05, + "loss": 0.5772, + "step": 17510 + }, + { + "epoch": 0.67, + "learning_rate": 3.880418055911772e-05, + "loss": 0.5276, + "step": 17511 + }, + { + "epoch": 0.67, + "learning_rate": 3.8803539369068995e-05, + "loss": 0.6028, + "step": 17512 + }, + { + "epoch": 0.67, + "learning_rate": 3.8802898179020266e-05, + "loss": 0.4853, + "step": 17513 + }, + { + "epoch": 0.67, + "learning_rate": 3.880225698897153e-05, + "loss": 0.4928, + "step": 17514 + }, + { + "epoch": 0.67, + "learning_rate": 3.88016157989228e-05, + "loss": 0.4882, + "step": 17515 + }, + { + "epoch": 0.67, + "learning_rate": 3.8800974608874075e-05, + "loss": 0.5607, + "step": 17516 + }, + { + "epoch": 0.67, + "learning_rate": 3.880033341882534e-05, + "loss": 0.4936, + "step": 17517 + }, + { + "epoch": 0.67, + "learning_rate": 3.879969222877661e-05, + "loss": 0.4847, + "step": 17518 + }, + { + "epoch": 0.67, + "learning_rate": 3.8799051038727876e-05, + "loss": 0.5469, + "step": 17519 + }, + { + "epoch": 0.67, + "learning_rate": 3.879840984867915e-05, + "loss": 0.5348, + "step": 17520 + }, + { + "epoch": 0.67, + "learning_rate": 3.879776865863042e-05, + "loss": 0.4615, + "step": 17521 + }, + { + "epoch": 0.67, + "learning_rate": 3.879712746858169e-05, + "loss": 0.5361, + "step": 17522 + }, + { + "epoch": 0.67, + "learning_rate": 3.8796486278532956e-05, + "loss": 0.6262, + "step": 17523 + }, + { + "epoch": 0.67, + "learning_rate": 3.879584508848423e-05, + "loss": 0.5354, + "step": 17524 + }, + { + "epoch": 0.67, + "learning_rate": 3.87952038984355e-05, + "loss": 0.5558, + "step": 17525 + }, + { + "epoch": 0.67, + "learning_rate": 3.8794562708386765e-05, + "loss": 0.4793, + "step": 17526 + }, + { + "epoch": 0.67, + "learning_rate": 3.8793921518338036e-05, + "loss": 0.4744, + "step": 17527 + }, + { + "epoch": 0.67, + "learning_rate": 3.879328032828931e-05, + "loss": 0.5499, + "step": 17528 + }, + { + "epoch": 0.67, + "learning_rate": 3.879263913824058e-05, + "loss": 0.5137, + "step": 17529 + }, + { + "epoch": 0.67, + "learning_rate": 3.8791997948191845e-05, + "loss": 0.5454, + "step": 17530 + }, + { + "epoch": 0.67, + "learning_rate": 3.8791356758143117e-05, + "loss": 0.4753, + "step": 17531 + }, + { + "epoch": 0.67, + "learning_rate": 3.879071556809438e-05, + "loss": 0.4552, + "step": 17532 + }, + { + "epoch": 0.67, + "learning_rate": 3.879007437804565e-05, + "loss": 0.5314, + "step": 17533 + }, + { + "epoch": 0.67, + "learning_rate": 3.8789433187996925e-05, + "loss": 0.5479, + "step": 17534 + }, + { + "epoch": 0.67, + "learning_rate": 3.878879199794819e-05, + "loss": 0.5622, + "step": 17535 + }, + { + "epoch": 0.67, + "learning_rate": 3.878815080789946e-05, + "loss": 0.5187, + "step": 17536 + }, + { + "epoch": 0.67, + "learning_rate": 3.878750961785073e-05, + "loss": 0.503, + "step": 17537 + }, + { + "epoch": 0.67, + "learning_rate": 3.8786868427802005e-05, + "loss": 0.5512, + "step": 17538 + }, + { + "epoch": 0.67, + "learning_rate": 3.878622723775327e-05, + "loss": 0.5858, + "step": 17539 + }, + { + "epoch": 0.67, + "learning_rate": 3.878558604770454e-05, + "loss": 0.5491, + "step": 17540 + }, + { + "epoch": 0.67, + "learning_rate": 3.878494485765581e-05, + "loss": 0.4918, + "step": 17541 + }, + { + "epoch": 0.67, + "learning_rate": 3.878430366760708e-05, + "loss": 0.5649, + "step": 17542 + }, + { + "epoch": 0.67, + "learning_rate": 3.878366247755835e-05, + "loss": 0.5166, + "step": 17543 + }, + { + "epoch": 0.67, + "learning_rate": 3.878302128750962e-05, + "loss": 0.5053, + "step": 17544 + }, + { + "epoch": 0.67, + "learning_rate": 3.8782380097460894e-05, + "loss": 0.519, + "step": 17545 + }, + { + "epoch": 0.67, + "learning_rate": 3.878173890741216e-05, + "loss": 0.5165, + "step": 17546 + }, + { + "epoch": 0.67, + "learning_rate": 3.878109771736343e-05, + "loss": 0.608, + "step": 17547 + }, + { + "epoch": 0.67, + "learning_rate": 3.8780456527314695e-05, + "loss": 0.4931, + "step": 17548 + }, + { + "epoch": 0.67, + "learning_rate": 3.877981533726597e-05, + "loss": 0.5574, + "step": 17549 + }, + { + "epoch": 0.67, + "learning_rate": 3.877917414721723e-05, + "loss": 0.4914, + "step": 17550 + }, + { + "epoch": 0.67, + "learning_rate": 3.8778532957168503e-05, + "loss": 0.4648, + "step": 17551 + }, + { + "epoch": 0.67, + "learning_rate": 3.8777891767119775e-05, + "loss": 0.5238, + "step": 17552 + }, + { + "epoch": 0.67, + "learning_rate": 3.877725057707105e-05, + "loss": 0.5732, + "step": 17553 + }, + { + "epoch": 0.67, + "learning_rate": 3.877660938702232e-05, + "loss": 0.5226, + "step": 17554 + }, + { + "epoch": 0.67, + "learning_rate": 3.8775968196973584e-05, + "loss": 0.5251, + "step": 17555 + }, + { + "epoch": 0.67, + "learning_rate": 3.8775327006924855e-05, + "loss": 0.5083, + "step": 17556 + }, + { + "epoch": 0.68, + "learning_rate": 3.877468581687612e-05, + "loss": 0.6725, + "step": 17557 + }, + { + "epoch": 0.68, + "learning_rate": 3.877404462682739e-05, + "loss": 0.4824, + "step": 17558 + }, + { + "epoch": 0.68, + "learning_rate": 3.8773403436778664e-05, + "loss": 0.3867, + "step": 17559 + }, + { + "epoch": 0.68, + "learning_rate": 3.8772762246729935e-05, + "loss": 0.5171, + "step": 17560 + }, + { + "epoch": 0.68, + "learning_rate": 3.877212105668121e-05, + "loss": 0.4367, + "step": 17561 + }, + { + "epoch": 0.68, + "learning_rate": 3.877147986663247e-05, + "loss": 0.5174, + "step": 17562 + }, + { + "epoch": 0.68, + "learning_rate": 3.8770838676583744e-05, + "loss": 0.4298, + "step": 17563 + }, + { + "epoch": 0.68, + "learning_rate": 3.877019748653501e-05, + "loss": 0.5322, + "step": 17564 + }, + { + "epoch": 0.68, + "learning_rate": 3.876955629648628e-05, + "loss": 0.5923, + "step": 17565 + }, + { + "epoch": 0.68, + "learning_rate": 3.8768915106437545e-05, + "loss": 0.5718, + "step": 17566 + }, + { + "epoch": 0.68, + "learning_rate": 3.876827391638882e-05, + "loss": 0.454, + "step": 17567 + }, + { + "epoch": 0.68, + "learning_rate": 3.876763272634009e-05, + "loss": 0.5682, + "step": 17568 + }, + { + "epoch": 0.68, + "learning_rate": 3.876699153629136e-05, + "loss": 0.5, + "step": 17569 + }, + { + "epoch": 0.68, + "learning_rate": 3.876635034624263e-05, + "loss": 0.4802, + "step": 17570 + }, + { + "epoch": 0.68, + "learning_rate": 3.87657091561939e-05, + "loss": 0.4019, + "step": 17571 + }, + { + "epoch": 0.68, + "learning_rate": 3.876506796614517e-05, + "loss": 0.5224, + "step": 17572 + }, + { + "epoch": 0.68, + "learning_rate": 3.8764426776096434e-05, + "loss": 0.492, + "step": 17573 + }, + { + "epoch": 0.68, + "learning_rate": 3.8763785586047706e-05, + "loss": 0.513, + "step": 17574 + }, + { + "epoch": 0.68, + "learning_rate": 3.876314439599897e-05, + "loss": 0.6744, + "step": 17575 + }, + { + "epoch": 0.68, + "learning_rate": 3.876250320595025e-05, + "loss": 0.4662, + "step": 17576 + }, + { + "epoch": 0.68, + "learning_rate": 3.876186201590152e-05, + "loss": 0.4397, + "step": 17577 + }, + { + "epoch": 0.68, + "learning_rate": 3.8761220825852786e-05, + "loss": 0.4956, + "step": 17578 + }, + { + "epoch": 0.68, + "learning_rate": 3.876057963580406e-05, + "loss": 0.5064, + "step": 17579 + }, + { + "epoch": 0.68, + "learning_rate": 3.875993844575532e-05, + "loss": 0.5543, + "step": 17580 + }, + { + "epoch": 0.68, + "learning_rate": 3.8759297255706594e-05, + "loss": 0.6216, + "step": 17581 + }, + { + "epoch": 0.68, + "learning_rate": 3.875865606565786e-05, + "loss": 0.3919, + "step": 17582 + }, + { + "epoch": 0.68, + "learning_rate": 3.875801487560913e-05, + "loss": 0.5615, + "step": 17583 + }, + { + "epoch": 0.68, + "learning_rate": 3.87573736855604e-05, + "loss": 0.5069, + "step": 17584 + }, + { + "epoch": 0.68, + "learning_rate": 3.8756732495511674e-05, + "loss": 0.5305, + "step": 17585 + }, + { + "epoch": 0.68, + "learning_rate": 3.8756091305462946e-05, + "loss": 0.5305, + "step": 17586 + }, + { + "epoch": 0.68, + "learning_rate": 3.875545011541421e-05, + "loss": 0.5137, + "step": 17587 + }, + { + "epoch": 0.68, + "learning_rate": 3.875480892536548e-05, + "loss": 0.5545, + "step": 17588 + }, + { + "epoch": 0.68, + "learning_rate": 3.875416773531675e-05, + "loss": 0.6449, + "step": 17589 + }, + { + "epoch": 0.68, + "learning_rate": 3.875352654526802e-05, + "loss": 0.4969, + "step": 17590 + }, + { + "epoch": 0.68, + "learning_rate": 3.8752885355219284e-05, + "loss": 0.4913, + "step": 17591 + }, + { + "epoch": 0.68, + "learning_rate": 3.875224416517056e-05, + "loss": 0.52, + "step": 17592 + }, + { + "epoch": 0.68, + "learning_rate": 3.875160297512183e-05, + "loss": 0.5358, + "step": 17593 + }, + { + "epoch": 0.68, + "learning_rate": 3.87509617850731e-05, + "loss": 0.5175, + "step": 17594 + }, + { + "epoch": 0.68, + "learning_rate": 3.875032059502437e-05, + "loss": 0.5562, + "step": 17595 + }, + { + "epoch": 0.68, + "learning_rate": 3.8749679404975636e-05, + "loss": 0.5135, + "step": 17596 + }, + { + "epoch": 0.68, + "learning_rate": 3.874903821492691e-05, + "loss": 0.5559, + "step": 17597 + }, + { + "epoch": 0.68, + "learning_rate": 3.874839702487817e-05, + "loss": 0.4286, + "step": 17598 + }, + { + "epoch": 0.68, + "learning_rate": 3.8747755834829444e-05, + "loss": 0.4581, + "step": 17599 + }, + { + "epoch": 0.68, + "learning_rate": 3.8747114644780716e-05, + "loss": 0.6123, + "step": 17600 + }, + { + "epoch": 0.68, + "learning_rate": 3.874647345473199e-05, + "loss": 0.578, + "step": 17601 + }, + { + "epoch": 0.68, + "learning_rate": 3.874583226468325e-05, + "loss": 0.4813, + "step": 17602 + }, + { + "epoch": 0.68, + "learning_rate": 3.8745191074634524e-05, + "loss": 0.5405, + "step": 17603 + }, + { + "epoch": 0.68, + "learning_rate": 3.8744549884585796e-05, + "loss": 0.576, + "step": 17604 + }, + { + "epoch": 0.68, + "learning_rate": 3.874390869453706e-05, + "loss": 0.4921, + "step": 17605 + }, + { + "epoch": 0.68, + "learning_rate": 3.874326750448833e-05, + "loss": 0.4349, + "step": 17606 + }, + { + "epoch": 0.68, + "learning_rate": 3.87426263144396e-05, + "loss": 0.4534, + "step": 17607 + }, + { + "epoch": 0.68, + "learning_rate": 3.874198512439087e-05, + "loss": 0.5553, + "step": 17608 + }, + { + "epoch": 0.68, + "learning_rate": 3.874134393434214e-05, + "loss": 0.5307, + "step": 17609 + }, + { + "epoch": 0.68, + "learning_rate": 3.874070274429341e-05, + "loss": 0.4846, + "step": 17610 + }, + { + "epoch": 0.68, + "learning_rate": 3.874006155424468e-05, + "loss": 0.4509, + "step": 17611 + }, + { + "epoch": 0.68, + "learning_rate": 3.873942036419595e-05, + "loss": 0.6752, + "step": 17612 + }, + { + "epoch": 0.68, + "learning_rate": 3.873877917414722e-05, + "loss": 0.5303, + "step": 17613 + }, + { + "epoch": 0.68, + "learning_rate": 3.8738137984098486e-05, + "loss": 0.574, + "step": 17614 + }, + { + "epoch": 0.68, + "learning_rate": 3.873749679404976e-05, + "loss": 0.561, + "step": 17615 + }, + { + "epoch": 0.68, + "learning_rate": 3.873685560400103e-05, + "loss": 0.5257, + "step": 17616 + }, + { + "epoch": 0.68, + "learning_rate": 3.87362144139523e-05, + "loss": 0.6176, + "step": 17617 + }, + { + "epoch": 0.68, + "learning_rate": 3.8735573223903566e-05, + "loss": 0.5496, + "step": 17618 + }, + { + "epoch": 0.68, + "learning_rate": 3.873493203385484e-05, + "loss": 0.5531, + "step": 17619 + }, + { + "epoch": 0.68, + "learning_rate": 3.87342908438061e-05, + "loss": 0.5814, + "step": 17620 + }, + { + "epoch": 0.68, + "learning_rate": 3.8733649653757375e-05, + "loss": 0.4985, + "step": 17621 + }, + { + "epoch": 0.68, + "learning_rate": 3.8733008463708646e-05, + "loss": 0.4617, + "step": 17622 + }, + { + "epoch": 0.68, + "learning_rate": 3.873236727365991e-05, + "loss": 0.4725, + "step": 17623 + }, + { + "epoch": 0.68, + "learning_rate": 3.873172608361118e-05, + "loss": 0.5386, + "step": 17624 + }, + { + "epoch": 0.68, + "learning_rate": 3.8731084893562455e-05, + "loss": 0.505, + "step": 17625 + }, + { + "epoch": 0.68, + "learning_rate": 3.8730443703513726e-05, + "loss": 0.4853, + "step": 17626 + }, + { + "epoch": 0.68, + "learning_rate": 3.872980251346499e-05, + "loss": 0.5388, + "step": 17627 + }, + { + "epoch": 0.68, + "learning_rate": 3.872916132341626e-05, + "loss": 0.4534, + "step": 17628 + }, + { + "epoch": 0.68, + "learning_rate": 3.872852013336753e-05, + "loss": 0.5097, + "step": 17629 + }, + { + "epoch": 0.68, + "learning_rate": 3.87278789433188e-05, + "loss": 0.4407, + "step": 17630 + }, + { + "epoch": 0.68, + "learning_rate": 3.872723775327007e-05, + "loss": 0.5213, + "step": 17631 + }, + { + "epoch": 0.68, + "learning_rate": 3.872659656322134e-05, + "loss": 0.6288, + "step": 17632 + }, + { + "epoch": 0.68, + "learning_rate": 3.8725955373172615e-05, + "loss": 0.5175, + "step": 17633 + }, + { + "epoch": 0.68, + "learning_rate": 3.872531418312388e-05, + "loss": 0.5054, + "step": 17634 + }, + { + "epoch": 0.68, + "learning_rate": 3.872467299307515e-05, + "loss": 0.5794, + "step": 17635 + }, + { + "epoch": 0.68, + "learning_rate": 3.8724031803026416e-05, + "loss": 0.5467, + "step": 17636 + }, + { + "epoch": 0.68, + "learning_rate": 3.872339061297769e-05, + "loss": 0.5115, + "step": 17637 + }, + { + "epoch": 0.68, + "learning_rate": 3.872274942292895e-05, + "loss": 0.5412, + "step": 17638 + }, + { + "epoch": 0.68, + "learning_rate": 3.8722108232880225e-05, + "loss": 0.5372, + "step": 17639 + }, + { + "epoch": 0.68, + "learning_rate": 3.8721467042831497e-05, + "loss": 0.5877, + "step": 17640 + }, + { + "epoch": 0.68, + "learning_rate": 3.872082585278277e-05, + "loss": 0.4971, + "step": 17641 + }, + { + "epoch": 0.68, + "learning_rate": 3.872018466273404e-05, + "loss": 0.4964, + "step": 17642 + }, + { + "epoch": 0.68, + "learning_rate": 3.8719543472685305e-05, + "loss": 0.4655, + "step": 17643 + }, + { + "epoch": 0.68, + "learning_rate": 3.8718902282636577e-05, + "loss": 0.525, + "step": 17644 + }, + { + "epoch": 0.68, + "learning_rate": 3.871826109258784e-05, + "loss": 0.4914, + "step": 17645 + }, + { + "epoch": 0.68, + "learning_rate": 3.871761990253911e-05, + "loss": 0.5183, + "step": 17646 + }, + { + "epoch": 0.68, + "learning_rate": 3.871697871249038e-05, + "loss": 0.4262, + "step": 17647 + }, + { + "epoch": 0.68, + "learning_rate": 3.871633752244166e-05, + "loss": 0.5258, + "step": 17648 + }, + { + "epoch": 0.68, + "learning_rate": 3.871569633239293e-05, + "loss": 0.4482, + "step": 17649 + }, + { + "epoch": 0.68, + "learning_rate": 3.871505514234419e-05, + "loss": 0.6693, + "step": 17650 + }, + { + "epoch": 0.68, + "learning_rate": 3.8714413952295465e-05, + "loss": 0.4618, + "step": 17651 + }, + { + "epoch": 0.68, + "learning_rate": 3.871377276224673e-05, + "loss": 0.657, + "step": 17652 + }, + { + "epoch": 0.68, + "learning_rate": 3.8713131572198e-05, + "loss": 0.611, + "step": 17653 + }, + { + "epoch": 0.68, + "learning_rate": 3.871249038214927e-05, + "loss": 0.5465, + "step": 17654 + }, + { + "epoch": 0.68, + "learning_rate": 3.871184919210054e-05, + "loss": 0.5278, + "step": 17655 + }, + { + "epoch": 0.68, + "learning_rate": 3.871120800205181e-05, + "loss": 0.5409, + "step": 17656 + }, + { + "epoch": 0.68, + "learning_rate": 3.871056681200308e-05, + "loss": 0.4974, + "step": 17657 + }, + { + "epoch": 0.68, + "learning_rate": 3.8709925621954354e-05, + "loss": 0.4549, + "step": 17658 + }, + { + "epoch": 0.68, + "learning_rate": 3.870928443190562e-05, + "loss": 0.547, + "step": 17659 + }, + { + "epoch": 0.68, + "learning_rate": 3.870864324185689e-05, + "loss": 0.6778, + "step": 17660 + }, + { + "epoch": 0.68, + "learning_rate": 3.8708002051808155e-05, + "loss": 0.5656, + "step": 17661 + }, + { + "epoch": 0.68, + "learning_rate": 3.870736086175943e-05, + "loss": 0.5713, + "step": 17662 + }, + { + "epoch": 0.68, + "learning_rate": 3.870671967171069e-05, + "loss": 0.4946, + "step": 17663 + }, + { + "epoch": 0.68, + "learning_rate": 3.870607848166197e-05, + "loss": 0.4124, + "step": 17664 + }, + { + "epoch": 0.68, + "learning_rate": 3.8705437291613235e-05, + "loss": 0.5469, + "step": 17665 + }, + { + "epoch": 0.68, + "learning_rate": 3.870479610156451e-05, + "loss": 0.4558, + "step": 17666 + }, + { + "epoch": 0.68, + "learning_rate": 3.870415491151578e-05, + "loss": 0.5039, + "step": 17667 + }, + { + "epoch": 0.68, + "learning_rate": 3.8703513721467044e-05, + "loss": 0.5179, + "step": 17668 + }, + { + "epoch": 0.68, + "learning_rate": 3.8702872531418315e-05, + "loss": 0.5119, + "step": 17669 + }, + { + "epoch": 0.68, + "learning_rate": 3.870223134136958e-05, + "loss": 0.5106, + "step": 17670 + }, + { + "epoch": 0.68, + "learning_rate": 3.870159015132085e-05, + "loss": 0.4308, + "step": 17671 + }, + { + "epoch": 0.68, + "learning_rate": 3.8700948961272124e-05, + "loss": 0.5487, + "step": 17672 + }, + { + "epoch": 0.68, + "learning_rate": 3.8700307771223395e-05, + "loss": 0.4718, + "step": 17673 + }, + { + "epoch": 0.68, + "learning_rate": 3.869966658117466e-05, + "loss": 0.5389, + "step": 17674 + }, + { + "epoch": 0.68, + "learning_rate": 3.869902539112593e-05, + "loss": 0.5221, + "step": 17675 + }, + { + "epoch": 0.68, + "learning_rate": 3.8698384201077204e-05, + "loss": 0.5794, + "step": 17676 + }, + { + "epoch": 0.68, + "learning_rate": 3.869774301102847e-05, + "loss": 0.4672, + "step": 17677 + }, + { + "epoch": 0.68, + "learning_rate": 3.869710182097974e-05, + "loss": 0.5005, + "step": 17678 + }, + { + "epoch": 0.68, + "learning_rate": 3.8696460630931005e-05, + "loss": 0.4773, + "step": 17679 + }, + { + "epoch": 0.68, + "learning_rate": 3.8695819440882284e-05, + "loss": 0.5112, + "step": 17680 + }, + { + "epoch": 0.68, + "learning_rate": 3.869517825083355e-05, + "loss": 0.5545, + "step": 17681 + }, + { + "epoch": 0.68, + "learning_rate": 3.869453706078482e-05, + "loss": 0.4032, + "step": 17682 + }, + { + "epoch": 0.68, + "learning_rate": 3.8693895870736085e-05, + "loss": 0.6359, + "step": 17683 + }, + { + "epoch": 0.68, + "learning_rate": 3.869325468068736e-05, + "loss": 0.4453, + "step": 17684 + }, + { + "epoch": 0.68, + "learning_rate": 3.869261349063863e-05, + "loss": 0.4722, + "step": 17685 + }, + { + "epoch": 0.68, + "learning_rate": 3.8691972300589894e-05, + "loss": 0.5345, + "step": 17686 + }, + { + "epoch": 0.68, + "learning_rate": 3.8691331110541166e-05, + "loss": 0.3978, + "step": 17687 + }, + { + "epoch": 0.68, + "learning_rate": 3.869068992049244e-05, + "loss": 0.5376, + "step": 17688 + }, + { + "epoch": 0.68, + "learning_rate": 3.869004873044371e-05, + "loss": 0.4837, + "step": 17689 + }, + { + "epoch": 0.68, + "learning_rate": 3.8689407540394974e-05, + "loss": 0.646, + "step": 17690 + }, + { + "epoch": 0.68, + "learning_rate": 3.8688766350346246e-05, + "loss": 0.4994, + "step": 17691 + }, + { + "epoch": 0.68, + "learning_rate": 3.868812516029751e-05, + "loss": 0.3896, + "step": 17692 + }, + { + "epoch": 0.68, + "learning_rate": 3.868748397024878e-05, + "loss": 0.4809, + "step": 17693 + }, + { + "epoch": 0.68, + "learning_rate": 3.8686842780200054e-05, + "loss": 0.5338, + "step": 17694 + }, + { + "epoch": 0.68, + "learning_rate": 3.868620159015132e-05, + "loss": 0.5017, + "step": 17695 + }, + { + "epoch": 0.68, + "learning_rate": 3.868556040010259e-05, + "loss": 0.551, + "step": 17696 + }, + { + "epoch": 0.68, + "learning_rate": 3.868491921005386e-05, + "loss": 0.5178, + "step": 17697 + }, + { + "epoch": 0.68, + "learning_rate": 3.8684278020005134e-05, + "loss": 0.5038, + "step": 17698 + }, + { + "epoch": 0.68, + "learning_rate": 3.86836368299564e-05, + "loss": 0.5028, + "step": 17699 + }, + { + "epoch": 0.68, + "learning_rate": 3.868299563990767e-05, + "loss": 0.4999, + "step": 17700 + }, + { + "epoch": 0.68, + "learning_rate": 3.8682354449858936e-05, + "loss": 0.4161, + "step": 17701 + }, + { + "epoch": 0.68, + "learning_rate": 3.868171325981021e-05, + "loss": 0.6871, + "step": 17702 + }, + { + "epoch": 0.68, + "learning_rate": 3.868107206976148e-05, + "loss": 0.5045, + "step": 17703 + }, + { + "epoch": 0.68, + "learning_rate": 3.868043087971275e-05, + "loss": 0.5625, + "step": 17704 + }, + { + "epoch": 0.68, + "learning_rate": 3.867978968966402e-05, + "loss": 0.6448, + "step": 17705 + }, + { + "epoch": 0.68, + "learning_rate": 3.867914849961529e-05, + "loss": 0.4598, + "step": 17706 + }, + { + "epoch": 0.68, + "learning_rate": 3.867850730956656e-05, + "loss": 0.4935, + "step": 17707 + }, + { + "epoch": 0.68, + "learning_rate": 3.8677866119517824e-05, + "loss": 0.5082, + "step": 17708 + }, + { + "epoch": 0.68, + "learning_rate": 3.8677224929469096e-05, + "loss": 0.5247, + "step": 17709 + }, + { + "epoch": 0.68, + "learning_rate": 3.867658373942036e-05, + "loss": 0.5008, + "step": 17710 + }, + { + "epoch": 0.68, + "learning_rate": 3.867594254937163e-05, + "loss": 0.4362, + "step": 17711 + }, + { + "epoch": 0.68, + "learning_rate": 3.8675301359322904e-05, + "loss": 0.5287, + "step": 17712 + }, + { + "epoch": 0.68, + "learning_rate": 3.8674660169274176e-05, + "loss": 0.4785, + "step": 17713 + }, + { + "epoch": 0.68, + "learning_rate": 3.867401897922545e-05, + "loss": 0.3952, + "step": 17714 + }, + { + "epoch": 0.68, + "learning_rate": 3.867337778917671e-05, + "loss": 0.5395, + "step": 17715 + }, + { + "epoch": 0.68, + "learning_rate": 3.8672736599127984e-05, + "loss": 0.5186, + "step": 17716 + }, + { + "epoch": 0.68, + "learning_rate": 3.867209540907925e-05, + "loss": 0.4948, + "step": 17717 + }, + { + "epoch": 0.68, + "learning_rate": 3.867145421903052e-05, + "loss": 0.5151, + "step": 17718 + }, + { + "epoch": 0.68, + "learning_rate": 3.867081302898179e-05, + "loss": 0.5778, + "step": 17719 + }, + { + "epoch": 0.68, + "learning_rate": 3.8670171838933064e-05, + "loss": 0.4573, + "step": 17720 + }, + { + "epoch": 0.68, + "learning_rate": 3.8669530648884336e-05, + "loss": 0.5531, + "step": 17721 + }, + { + "epoch": 0.68, + "learning_rate": 3.86688894588356e-05, + "loss": 0.5332, + "step": 17722 + }, + { + "epoch": 0.68, + "learning_rate": 3.866824826878687e-05, + "loss": 0.5704, + "step": 17723 + }, + { + "epoch": 0.68, + "learning_rate": 3.866760707873814e-05, + "loss": 0.4928, + "step": 17724 + }, + { + "epoch": 0.68, + "learning_rate": 3.866696588868941e-05, + "loss": 0.5469, + "step": 17725 + }, + { + "epoch": 0.68, + "learning_rate": 3.8666324698640674e-05, + "loss": 0.4733, + "step": 17726 + }, + { + "epoch": 0.68, + "learning_rate": 3.8665683508591946e-05, + "loss": 0.6298, + "step": 17727 + }, + { + "epoch": 0.68, + "learning_rate": 3.866504231854322e-05, + "loss": 0.434, + "step": 17728 + }, + { + "epoch": 0.68, + "learning_rate": 3.866440112849449e-05, + "loss": 0.5057, + "step": 17729 + }, + { + "epoch": 0.68, + "learning_rate": 3.866375993844576e-05, + "loss": 0.5805, + "step": 17730 + }, + { + "epoch": 0.68, + "learning_rate": 3.8663118748397026e-05, + "loss": 0.4919, + "step": 17731 + }, + { + "epoch": 0.68, + "learning_rate": 3.86624775583483e-05, + "loss": 0.4382, + "step": 17732 + }, + { + "epoch": 0.68, + "learning_rate": 3.866183636829956e-05, + "loss": 0.501, + "step": 17733 + }, + { + "epoch": 0.68, + "learning_rate": 3.8661195178250835e-05, + "loss": 0.4842, + "step": 17734 + }, + { + "epoch": 0.68, + "learning_rate": 3.86605539882021e-05, + "loss": 0.5514, + "step": 17735 + }, + { + "epoch": 0.68, + "learning_rate": 3.865991279815338e-05, + "loss": 0.5851, + "step": 17736 + }, + { + "epoch": 0.68, + "learning_rate": 3.865927160810464e-05, + "loss": 0.4263, + "step": 17737 + }, + { + "epoch": 0.68, + "learning_rate": 3.8658630418055915e-05, + "loss": 0.4557, + "step": 17738 + }, + { + "epoch": 0.68, + "learning_rate": 3.8657989228007186e-05, + "loss": 0.4711, + "step": 17739 + }, + { + "epoch": 0.68, + "learning_rate": 3.865734803795845e-05, + "loss": 0.4972, + "step": 17740 + }, + { + "epoch": 0.68, + "learning_rate": 3.865670684790972e-05, + "loss": 0.5297, + "step": 17741 + }, + { + "epoch": 0.68, + "learning_rate": 3.865606565786099e-05, + "loss": 0.4383, + "step": 17742 + }, + { + "epoch": 0.68, + "learning_rate": 3.865542446781226e-05, + "loss": 0.5051, + "step": 17743 + }, + { + "epoch": 0.68, + "learning_rate": 3.865478327776353e-05, + "loss": 0.5553, + "step": 17744 + }, + { + "epoch": 0.68, + "learning_rate": 3.86541420877148e-05, + "loss": 0.4929, + "step": 17745 + }, + { + "epoch": 0.68, + "learning_rate": 3.865350089766607e-05, + "loss": 0.4429, + "step": 17746 + }, + { + "epoch": 0.68, + "learning_rate": 3.865285970761734e-05, + "loss": 0.6154, + "step": 17747 + }, + { + "epoch": 0.68, + "learning_rate": 3.865221851756861e-05, + "loss": 0.6278, + "step": 17748 + }, + { + "epoch": 0.68, + "learning_rate": 3.8651577327519876e-05, + "loss": 0.4961, + "step": 17749 + }, + { + "epoch": 0.68, + "learning_rate": 3.865093613747115e-05, + "loss": 0.428, + "step": 17750 + }, + { + "epoch": 0.68, + "learning_rate": 3.865029494742241e-05, + "loss": 0.4883, + "step": 17751 + }, + { + "epoch": 0.68, + "learning_rate": 3.864965375737369e-05, + "loss": 0.402, + "step": 17752 + }, + { + "epoch": 0.68, + "learning_rate": 3.8649012567324957e-05, + "loss": 0.4966, + "step": 17753 + }, + { + "epoch": 0.68, + "learning_rate": 3.864837137727623e-05, + "loss": 0.5992, + "step": 17754 + }, + { + "epoch": 0.68, + "learning_rate": 3.86477301872275e-05, + "loss": 0.4707, + "step": 17755 + }, + { + "epoch": 0.68, + "learning_rate": 3.8647088997178765e-05, + "loss": 0.4748, + "step": 17756 + }, + { + "epoch": 0.68, + "learning_rate": 3.864644780713004e-05, + "loss": 0.5826, + "step": 17757 + }, + { + "epoch": 0.68, + "learning_rate": 3.86458066170813e-05, + "loss": 0.4931, + "step": 17758 + }, + { + "epoch": 0.68, + "learning_rate": 3.864516542703257e-05, + "loss": 0.4605, + "step": 17759 + }, + { + "epoch": 0.68, + "learning_rate": 3.8644524236983845e-05, + "loss": 0.5938, + "step": 17760 + }, + { + "epoch": 0.68, + "learning_rate": 3.864388304693512e-05, + "loss": 0.4779, + "step": 17761 + }, + { + "epoch": 0.68, + "learning_rate": 3.864324185688638e-05, + "loss": 0.4361, + "step": 17762 + }, + { + "epoch": 0.68, + "learning_rate": 3.864260066683765e-05, + "loss": 0.5421, + "step": 17763 + }, + { + "epoch": 0.68, + "learning_rate": 3.8641959476788925e-05, + "loss": 0.6218, + "step": 17764 + }, + { + "epoch": 0.68, + "learning_rate": 3.864131828674019e-05, + "loss": 0.5539, + "step": 17765 + }, + { + "epoch": 0.68, + "learning_rate": 3.864067709669146e-05, + "loss": 0.5299, + "step": 17766 + }, + { + "epoch": 0.68, + "learning_rate": 3.864003590664273e-05, + "loss": 0.524, + "step": 17767 + }, + { + "epoch": 0.68, + "learning_rate": 3.8639394716594e-05, + "loss": 0.5426, + "step": 17768 + }, + { + "epoch": 0.68, + "learning_rate": 3.863875352654527e-05, + "loss": 0.4712, + "step": 17769 + }, + { + "epoch": 0.68, + "learning_rate": 3.863811233649654e-05, + "loss": 0.5693, + "step": 17770 + }, + { + "epoch": 0.68, + "learning_rate": 3.863747114644781e-05, + "loss": 0.4454, + "step": 17771 + }, + { + "epoch": 0.68, + "learning_rate": 3.863682995639908e-05, + "loss": 0.5225, + "step": 17772 + }, + { + "epoch": 0.68, + "learning_rate": 3.863618876635035e-05, + "loss": 0.643, + "step": 17773 + }, + { + "epoch": 0.68, + "learning_rate": 3.8635547576301615e-05, + "loss": 0.4344, + "step": 17774 + }, + { + "epoch": 0.68, + "learning_rate": 3.863490638625289e-05, + "loss": 0.3754, + "step": 17775 + }, + { + "epoch": 0.68, + "learning_rate": 3.863426519620416e-05, + "loss": 0.4686, + "step": 17776 + }, + { + "epoch": 0.68, + "learning_rate": 3.863362400615543e-05, + "loss": 0.6292, + "step": 17777 + }, + { + "epoch": 0.68, + "learning_rate": 3.8632982816106695e-05, + "loss": 0.5431, + "step": 17778 + }, + { + "epoch": 0.68, + "learning_rate": 3.863234162605797e-05, + "loss": 0.5107, + "step": 17779 + }, + { + "epoch": 0.68, + "learning_rate": 3.863170043600923e-05, + "loss": 0.5317, + "step": 17780 + }, + { + "epoch": 0.68, + "learning_rate": 3.8631059245960504e-05, + "loss": 0.4578, + "step": 17781 + }, + { + "epoch": 0.68, + "learning_rate": 3.8630418055911775e-05, + "loss": 0.5744, + "step": 17782 + }, + { + "epoch": 0.68, + "learning_rate": 3.862977686586304e-05, + "loss": 0.5657, + "step": 17783 + }, + { + "epoch": 0.68, + "learning_rate": 3.862913567581431e-05, + "loss": 0.5778, + "step": 17784 + }, + { + "epoch": 0.68, + "learning_rate": 3.8628494485765584e-05, + "loss": 0.4955, + "step": 17785 + }, + { + "epoch": 0.68, + "learning_rate": 3.8627853295716855e-05, + "loss": 0.5887, + "step": 17786 + }, + { + "epoch": 0.68, + "learning_rate": 3.862721210566812e-05, + "loss": 0.5261, + "step": 17787 + }, + { + "epoch": 0.68, + "learning_rate": 3.862657091561939e-05, + "loss": 0.5422, + "step": 17788 + }, + { + "epoch": 0.68, + "learning_rate": 3.862592972557066e-05, + "loss": 0.5062, + "step": 17789 + }, + { + "epoch": 0.68, + "learning_rate": 3.862528853552193e-05, + "loss": 0.4869, + "step": 17790 + }, + { + "epoch": 0.68, + "learning_rate": 3.86246473454732e-05, + "loss": 0.4708, + "step": 17791 + }, + { + "epoch": 0.68, + "learning_rate": 3.862400615542447e-05, + "loss": 0.5042, + "step": 17792 + }, + { + "epoch": 0.68, + "learning_rate": 3.8623364965375744e-05, + "loss": 0.5323, + "step": 17793 + }, + { + "epoch": 0.68, + "learning_rate": 3.862272377532701e-05, + "loss": 0.483, + "step": 17794 + }, + { + "epoch": 0.68, + "learning_rate": 3.862208258527828e-05, + "loss": 0.526, + "step": 17795 + }, + { + "epoch": 0.68, + "learning_rate": 3.8621441395229545e-05, + "loss": 0.6196, + "step": 17796 + }, + { + "epoch": 0.68, + "learning_rate": 3.862080020518082e-05, + "loss": 0.5498, + "step": 17797 + }, + { + "epoch": 0.68, + "learning_rate": 3.862015901513208e-05, + "loss": 0.4817, + "step": 17798 + }, + { + "epoch": 0.68, + "learning_rate": 3.8619517825083354e-05, + "loss": 0.5155, + "step": 17799 + }, + { + "epoch": 0.68, + "learning_rate": 3.8618876635034626e-05, + "loss": 0.5407, + "step": 17800 + }, + { + "epoch": 0.68, + "learning_rate": 3.86182354449859e-05, + "loss": 0.6439, + "step": 17801 + }, + { + "epoch": 0.68, + "learning_rate": 3.861759425493717e-05, + "loss": 0.5804, + "step": 17802 + }, + { + "epoch": 0.68, + "learning_rate": 3.8616953064888434e-05, + "loss": 0.5178, + "step": 17803 + }, + { + "epoch": 0.68, + "learning_rate": 3.8616311874839706e-05, + "loss": 0.4879, + "step": 17804 + }, + { + "epoch": 0.68, + "learning_rate": 3.861567068479097e-05, + "loss": 0.5845, + "step": 17805 + }, + { + "epoch": 0.68, + "learning_rate": 3.861502949474224e-05, + "loss": 0.709, + "step": 17806 + }, + { + "epoch": 0.68, + "learning_rate": 3.861438830469351e-05, + "loss": 0.5312, + "step": 17807 + }, + { + "epoch": 0.68, + "learning_rate": 3.8613747114644786e-05, + "loss": 0.4834, + "step": 17808 + }, + { + "epoch": 0.68, + "learning_rate": 3.861310592459606e-05, + "loss": 0.4824, + "step": 17809 + }, + { + "epoch": 0.68, + "learning_rate": 3.861246473454732e-05, + "loss": 0.4769, + "step": 17810 + }, + { + "epoch": 0.68, + "learning_rate": 3.8611823544498594e-05, + "loss": 0.4869, + "step": 17811 + }, + { + "epoch": 0.68, + "learning_rate": 3.861118235444986e-05, + "loss": 0.5015, + "step": 17812 + }, + { + "epoch": 0.68, + "learning_rate": 3.861054116440113e-05, + "loss": 0.5448, + "step": 17813 + }, + { + "epoch": 0.68, + "learning_rate": 3.8609899974352396e-05, + "loss": 0.5232, + "step": 17814 + }, + { + "epoch": 0.68, + "learning_rate": 3.860925878430367e-05, + "loss": 0.4864, + "step": 17815 + }, + { + "epoch": 0.68, + "learning_rate": 3.860861759425494e-05, + "loss": 0.5626, + "step": 17816 + }, + { + "epoch": 0.69, + "learning_rate": 3.860797640420621e-05, + "loss": 0.6056, + "step": 17817 + }, + { + "epoch": 0.69, + "learning_rate": 3.860733521415748e-05, + "loss": 0.5453, + "step": 17818 + }, + { + "epoch": 0.69, + "learning_rate": 3.860669402410875e-05, + "loss": 0.5076, + "step": 17819 + }, + { + "epoch": 0.69, + "learning_rate": 3.860605283406002e-05, + "loss": 0.4733, + "step": 17820 + }, + { + "epoch": 0.69, + "learning_rate": 3.8605411644011284e-05, + "loss": 0.4441, + "step": 17821 + }, + { + "epoch": 0.69, + "learning_rate": 3.8604770453962556e-05, + "loss": 0.5818, + "step": 17822 + }, + { + "epoch": 0.69, + "learning_rate": 3.860412926391382e-05, + "loss": 0.6361, + "step": 17823 + }, + { + "epoch": 0.69, + "learning_rate": 3.86034880738651e-05, + "loss": 0.4869, + "step": 17824 + }, + { + "epoch": 0.69, + "learning_rate": 3.8602846883816364e-05, + "loss": 0.4368, + "step": 17825 + }, + { + "epoch": 0.69, + "learning_rate": 3.8602205693767636e-05, + "loss": 0.4816, + "step": 17826 + }, + { + "epoch": 0.69, + "learning_rate": 3.860156450371891e-05, + "loss": 0.5564, + "step": 17827 + }, + { + "epoch": 0.69, + "learning_rate": 3.860092331367017e-05, + "loss": 0.4347, + "step": 17828 + }, + { + "epoch": 0.69, + "learning_rate": 3.8600282123621444e-05, + "loss": 0.5093, + "step": 17829 + }, + { + "epoch": 0.69, + "learning_rate": 3.859964093357271e-05, + "loss": 0.4884, + "step": 17830 + }, + { + "epoch": 0.69, + "learning_rate": 3.859899974352398e-05, + "loss": 0.4675, + "step": 17831 + }, + { + "epoch": 0.69, + "learning_rate": 3.859835855347525e-05, + "loss": 0.4794, + "step": 17832 + }, + { + "epoch": 0.69, + "learning_rate": 3.8597717363426524e-05, + "loss": 0.59, + "step": 17833 + }, + { + "epoch": 0.69, + "learning_rate": 3.859707617337779e-05, + "loss": 0.434, + "step": 17834 + }, + { + "epoch": 0.69, + "learning_rate": 3.859643498332906e-05, + "loss": 0.4998, + "step": 17835 + }, + { + "epoch": 0.69, + "learning_rate": 3.859579379328033e-05, + "loss": 0.6201, + "step": 17836 + }, + { + "epoch": 0.69, + "learning_rate": 3.85951526032316e-05, + "loss": 0.5433, + "step": 17837 + }, + { + "epoch": 0.69, + "learning_rate": 3.859451141318287e-05, + "loss": 0.5677, + "step": 17838 + }, + { + "epoch": 0.69, + "learning_rate": 3.8593870223134134e-05, + "loss": 0.5549, + "step": 17839 + }, + { + "epoch": 0.69, + "learning_rate": 3.859322903308541e-05, + "loss": 0.5243, + "step": 17840 + }, + { + "epoch": 0.69, + "learning_rate": 3.859258784303668e-05, + "loss": 0.5023, + "step": 17841 + }, + { + "epoch": 0.69, + "learning_rate": 3.859194665298795e-05, + "loss": 0.5646, + "step": 17842 + }, + { + "epoch": 0.69, + "learning_rate": 3.8591305462939215e-05, + "loss": 0.5877, + "step": 17843 + }, + { + "epoch": 0.69, + "learning_rate": 3.8590664272890486e-05, + "loss": 0.5499, + "step": 17844 + }, + { + "epoch": 0.69, + "learning_rate": 3.859002308284176e-05, + "loss": 0.5016, + "step": 17845 + }, + { + "epoch": 0.69, + "learning_rate": 3.858938189279302e-05, + "loss": 0.4575, + "step": 17846 + }, + { + "epoch": 0.69, + "learning_rate": 3.8588740702744295e-05, + "loss": 0.4777, + "step": 17847 + }, + { + "epoch": 0.69, + "learning_rate": 3.8588099512695566e-05, + "loss": 0.5477, + "step": 17848 + }, + { + "epoch": 0.69, + "learning_rate": 3.858745832264684e-05, + "loss": 0.6133, + "step": 17849 + }, + { + "epoch": 0.69, + "learning_rate": 3.85868171325981e-05, + "loss": 0.4741, + "step": 17850 + }, + { + "epoch": 0.69, + "learning_rate": 3.8586175942549375e-05, + "loss": 0.6232, + "step": 17851 + }, + { + "epoch": 0.69, + "learning_rate": 3.858553475250064e-05, + "loss": 0.467, + "step": 17852 + }, + { + "epoch": 0.69, + "learning_rate": 3.858489356245191e-05, + "loss": 0.5169, + "step": 17853 + }, + { + "epoch": 0.69, + "learning_rate": 3.858425237240318e-05, + "loss": 0.4875, + "step": 17854 + }, + { + "epoch": 0.69, + "learning_rate": 3.858361118235445e-05, + "loss": 0.5234, + "step": 17855 + }, + { + "epoch": 0.69, + "learning_rate": 3.858296999230572e-05, + "loss": 0.5438, + "step": 17856 + }, + { + "epoch": 0.69, + "learning_rate": 3.858232880225699e-05, + "loss": 0.5478, + "step": 17857 + }, + { + "epoch": 0.69, + "learning_rate": 3.858168761220826e-05, + "loss": 0.4731, + "step": 17858 + }, + { + "epoch": 0.69, + "learning_rate": 3.858104642215953e-05, + "loss": 0.6193, + "step": 17859 + }, + { + "epoch": 0.69, + "learning_rate": 3.85804052321108e-05, + "loss": 0.5821, + "step": 17860 + }, + { + "epoch": 0.69, + "learning_rate": 3.8579764042062065e-05, + "loss": 0.5059, + "step": 17861 + }, + { + "epoch": 0.69, + "learning_rate": 3.8579122852013336e-05, + "loss": 0.4773, + "step": 17862 + }, + { + "epoch": 0.69, + "learning_rate": 3.857848166196461e-05, + "loss": 0.6205, + "step": 17863 + }, + { + "epoch": 0.69, + "learning_rate": 3.857784047191588e-05, + "loss": 0.5618, + "step": 17864 + }, + { + "epoch": 0.69, + "learning_rate": 3.857719928186715e-05, + "loss": 0.5095, + "step": 17865 + }, + { + "epoch": 0.69, + "learning_rate": 3.8576558091818417e-05, + "loss": 0.4812, + "step": 17866 + }, + { + "epoch": 0.69, + "learning_rate": 3.857591690176969e-05, + "loss": 0.4486, + "step": 17867 + }, + { + "epoch": 0.69, + "learning_rate": 3.857527571172095e-05, + "loss": 0.5483, + "step": 17868 + }, + { + "epoch": 0.69, + "learning_rate": 3.8574634521672225e-05, + "loss": 0.4506, + "step": 17869 + }, + { + "epoch": 0.69, + "learning_rate": 3.857399333162349e-05, + "loss": 0.5584, + "step": 17870 + }, + { + "epoch": 0.69, + "learning_rate": 3.857335214157476e-05, + "loss": 0.4722, + "step": 17871 + }, + { + "epoch": 0.69, + "learning_rate": 3.857271095152603e-05, + "loss": 0.4956, + "step": 17872 + }, + { + "epoch": 0.69, + "learning_rate": 3.8572069761477305e-05, + "loss": 0.5101, + "step": 17873 + }, + { + "epoch": 0.69, + "learning_rate": 3.857142857142858e-05, + "loss": 0.7272, + "step": 17874 + }, + { + "epoch": 0.69, + "learning_rate": 3.857078738137984e-05, + "loss": 0.4403, + "step": 17875 + }, + { + "epoch": 0.69, + "learning_rate": 3.8570146191331113e-05, + "loss": 0.4004, + "step": 17876 + }, + { + "epoch": 0.69, + "learning_rate": 3.856950500128238e-05, + "loss": 0.5971, + "step": 17877 + }, + { + "epoch": 0.69, + "learning_rate": 3.856886381123365e-05, + "loss": 0.588, + "step": 17878 + }, + { + "epoch": 0.69, + "learning_rate": 3.856822262118492e-05, + "loss": 0.554, + "step": 17879 + }, + { + "epoch": 0.69, + "learning_rate": 3.8567581431136194e-05, + "loss": 0.6195, + "step": 17880 + }, + { + "epoch": 0.69, + "learning_rate": 3.8566940241087465e-05, + "loss": 0.5997, + "step": 17881 + }, + { + "epoch": 0.69, + "learning_rate": 3.856629905103873e-05, + "loss": 0.4613, + "step": 17882 + }, + { + "epoch": 0.69, + "learning_rate": 3.856565786099e-05, + "loss": 0.5721, + "step": 17883 + }, + { + "epoch": 0.69, + "learning_rate": 3.856501667094127e-05, + "loss": 0.5352, + "step": 17884 + }, + { + "epoch": 0.69, + "learning_rate": 3.856437548089254e-05, + "loss": 0.548, + "step": 17885 + }, + { + "epoch": 0.69, + "learning_rate": 3.8563734290843803e-05, + "loss": 0.4952, + "step": 17886 + }, + { + "epoch": 0.69, + "learning_rate": 3.8563093100795075e-05, + "loss": 0.4391, + "step": 17887 + }, + { + "epoch": 0.69, + "learning_rate": 3.856245191074635e-05, + "loss": 0.5611, + "step": 17888 + }, + { + "epoch": 0.69, + "learning_rate": 3.856181072069762e-05, + "loss": 0.4867, + "step": 17889 + }, + { + "epoch": 0.69, + "learning_rate": 3.856116953064889e-05, + "loss": 0.5014, + "step": 17890 + }, + { + "epoch": 0.69, + "learning_rate": 3.8560528340600155e-05, + "loss": 0.5635, + "step": 17891 + }, + { + "epoch": 0.69, + "learning_rate": 3.855988715055143e-05, + "loss": 0.4987, + "step": 17892 + }, + { + "epoch": 0.69, + "learning_rate": 3.855924596050269e-05, + "loss": 0.5274, + "step": 17893 + }, + { + "epoch": 0.69, + "learning_rate": 3.8558604770453964e-05, + "loss": 0.4773, + "step": 17894 + }, + { + "epoch": 0.69, + "learning_rate": 3.855796358040523e-05, + "loss": 0.4365, + "step": 17895 + }, + { + "epoch": 0.69, + "learning_rate": 3.855732239035651e-05, + "loss": 0.5628, + "step": 17896 + }, + { + "epoch": 0.69, + "learning_rate": 3.855668120030777e-05, + "loss": 0.5128, + "step": 17897 + }, + { + "epoch": 0.69, + "learning_rate": 3.8556040010259044e-05, + "loss": 0.5154, + "step": 17898 + }, + { + "epoch": 0.69, + "learning_rate": 3.8555398820210315e-05, + "loss": 0.565, + "step": 17899 + }, + { + "epoch": 0.69, + "learning_rate": 3.855475763016158e-05, + "loss": 0.5223, + "step": 17900 + }, + { + "epoch": 0.69, + "learning_rate": 3.855411644011285e-05, + "loss": 0.5484, + "step": 17901 + }, + { + "epoch": 0.69, + "learning_rate": 3.855347525006412e-05, + "loss": 0.4027, + "step": 17902 + }, + { + "epoch": 0.69, + "learning_rate": 3.855283406001539e-05, + "loss": 0.5727, + "step": 17903 + }, + { + "epoch": 0.69, + "learning_rate": 3.855219286996666e-05, + "loss": 0.529, + "step": 17904 + }, + { + "epoch": 0.69, + "learning_rate": 3.855155167991793e-05, + "loss": 0.4691, + "step": 17905 + }, + { + "epoch": 0.69, + "learning_rate": 3.85509104898692e-05, + "loss": 0.5968, + "step": 17906 + }, + { + "epoch": 0.69, + "learning_rate": 3.855026929982047e-05, + "loss": 0.5255, + "step": 17907 + }, + { + "epoch": 0.69, + "learning_rate": 3.854962810977174e-05, + "loss": 0.5262, + "step": 17908 + }, + { + "epoch": 0.69, + "learning_rate": 3.8548986919723006e-05, + "loss": 0.5765, + "step": 17909 + }, + { + "epoch": 0.69, + "learning_rate": 3.854834572967428e-05, + "loss": 0.4376, + "step": 17910 + }, + { + "epoch": 0.69, + "learning_rate": 3.854770453962554e-05, + "loss": 0.4826, + "step": 17911 + }, + { + "epoch": 0.69, + "learning_rate": 3.854706334957682e-05, + "loss": 0.4857, + "step": 17912 + }, + { + "epoch": 0.69, + "learning_rate": 3.8546422159528086e-05, + "loss": 0.5158, + "step": 17913 + }, + { + "epoch": 0.69, + "learning_rate": 3.854578096947936e-05, + "loss": 0.5129, + "step": 17914 + }, + { + "epoch": 0.69, + "learning_rate": 3.854513977943062e-05, + "loss": 0.4864, + "step": 17915 + }, + { + "epoch": 0.69, + "learning_rate": 3.8544498589381894e-05, + "loss": 0.4939, + "step": 17916 + }, + { + "epoch": 0.69, + "learning_rate": 3.8543857399333166e-05, + "loss": 0.5653, + "step": 17917 + }, + { + "epoch": 0.69, + "learning_rate": 3.854321620928443e-05, + "loss": 0.5535, + "step": 17918 + }, + { + "epoch": 0.69, + "learning_rate": 3.85425750192357e-05, + "loss": 0.5844, + "step": 17919 + }, + { + "epoch": 0.69, + "learning_rate": 3.8541933829186974e-05, + "loss": 0.4153, + "step": 17920 + }, + { + "epoch": 0.69, + "learning_rate": 3.8541292639138246e-05, + "loss": 0.4489, + "step": 17921 + }, + { + "epoch": 0.69, + "learning_rate": 3.854065144908951e-05, + "loss": 0.491, + "step": 17922 + }, + { + "epoch": 0.69, + "learning_rate": 3.854001025904078e-05, + "loss": 0.617, + "step": 17923 + }, + { + "epoch": 0.69, + "learning_rate": 3.853936906899205e-05, + "loss": 0.5303, + "step": 17924 + }, + { + "epoch": 0.69, + "learning_rate": 3.853872787894332e-05, + "loss": 0.4803, + "step": 17925 + }, + { + "epoch": 0.69, + "learning_rate": 3.853808668889459e-05, + "loss": 0.5474, + "step": 17926 + }, + { + "epoch": 0.69, + "learning_rate": 3.8537445498845856e-05, + "loss": 0.5289, + "step": 17927 + }, + { + "epoch": 0.69, + "learning_rate": 3.8536804308797134e-05, + "loss": 0.5382, + "step": 17928 + }, + { + "epoch": 0.69, + "learning_rate": 3.85361631187484e-05, + "loss": 0.5339, + "step": 17929 + }, + { + "epoch": 0.69, + "learning_rate": 3.853552192869967e-05, + "loss": 0.5817, + "step": 17930 + }, + { + "epoch": 0.69, + "learning_rate": 3.8534880738650936e-05, + "loss": 0.4542, + "step": 17931 + }, + { + "epoch": 0.69, + "learning_rate": 3.853423954860221e-05, + "loss": 0.4837, + "step": 17932 + }, + { + "epoch": 0.69, + "learning_rate": 3.853359835855347e-05, + "loss": 0.5612, + "step": 17933 + }, + { + "epoch": 0.69, + "learning_rate": 3.8532957168504744e-05, + "loss": 0.4894, + "step": 17934 + }, + { + "epoch": 0.69, + "learning_rate": 3.8532315978456016e-05, + "loss": 0.5057, + "step": 17935 + }, + { + "epoch": 0.69, + "learning_rate": 3.853167478840729e-05, + "loss": 0.4229, + "step": 17936 + }, + { + "epoch": 0.69, + "learning_rate": 3.853103359835856e-05, + "loss": 0.4886, + "step": 17937 + }, + { + "epoch": 0.69, + "learning_rate": 3.8530392408309824e-05, + "loss": 0.5622, + "step": 17938 + }, + { + "epoch": 0.69, + "learning_rate": 3.8529751218261096e-05, + "loss": 0.5267, + "step": 17939 + }, + { + "epoch": 0.69, + "learning_rate": 3.852911002821236e-05, + "loss": 0.4501, + "step": 17940 + }, + { + "epoch": 0.69, + "learning_rate": 3.852846883816363e-05, + "loss": 0.5278, + "step": 17941 + }, + { + "epoch": 0.69, + "learning_rate": 3.8527827648114904e-05, + "loss": 0.4776, + "step": 17942 + }, + { + "epoch": 0.69, + "learning_rate": 3.852718645806617e-05, + "loss": 0.4742, + "step": 17943 + }, + { + "epoch": 0.69, + "learning_rate": 3.852654526801744e-05, + "loss": 0.6058, + "step": 17944 + }, + { + "epoch": 0.69, + "learning_rate": 3.852590407796871e-05, + "loss": 0.6028, + "step": 17945 + }, + { + "epoch": 0.69, + "learning_rate": 3.8525262887919985e-05, + "loss": 0.5425, + "step": 17946 + }, + { + "epoch": 0.69, + "learning_rate": 3.852462169787125e-05, + "loss": 0.5782, + "step": 17947 + }, + { + "epoch": 0.69, + "learning_rate": 3.852398050782252e-05, + "loss": 0.5446, + "step": 17948 + }, + { + "epoch": 0.69, + "learning_rate": 3.8523339317773786e-05, + "loss": 0.6825, + "step": 17949 + }, + { + "epoch": 0.69, + "learning_rate": 3.852269812772506e-05, + "loss": 0.5257, + "step": 17950 + }, + { + "epoch": 0.69, + "learning_rate": 3.852205693767633e-05, + "loss": 0.5785, + "step": 17951 + }, + { + "epoch": 0.69, + "learning_rate": 3.85214157476276e-05, + "loss": 0.6923, + "step": 17952 + }, + { + "epoch": 0.69, + "learning_rate": 3.852077455757887e-05, + "loss": 0.4765, + "step": 17953 + }, + { + "epoch": 0.69, + "learning_rate": 3.852013336753014e-05, + "loss": 0.4802, + "step": 17954 + }, + { + "epoch": 0.69, + "learning_rate": 3.851949217748141e-05, + "loss": 0.4504, + "step": 17955 + }, + { + "epoch": 0.69, + "learning_rate": 3.8518850987432675e-05, + "loss": 0.5816, + "step": 17956 + }, + { + "epoch": 0.69, + "learning_rate": 3.8518209797383946e-05, + "loss": 0.5837, + "step": 17957 + }, + { + "epoch": 0.69, + "learning_rate": 3.851756860733521e-05, + "loss": 0.5183, + "step": 17958 + }, + { + "epoch": 0.69, + "learning_rate": 3.851692741728648e-05, + "loss": 0.5083, + "step": 17959 + }, + { + "epoch": 0.69, + "learning_rate": 3.8516286227237755e-05, + "loss": 0.4666, + "step": 17960 + }, + { + "epoch": 0.69, + "learning_rate": 3.8515645037189026e-05, + "loss": 0.4768, + "step": 17961 + }, + { + "epoch": 0.69, + "learning_rate": 3.85150038471403e-05, + "loss": 0.4972, + "step": 17962 + }, + { + "epoch": 0.69, + "learning_rate": 3.851436265709156e-05, + "loss": 0.5123, + "step": 17963 + }, + { + "epoch": 0.69, + "learning_rate": 3.8513721467042835e-05, + "loss": 0.6266, + "step": 17964 + }, + { + "epoch": 0.69, + "learning_rate": 3.85130802769941e-05, + "loss": 0.5558, + "step": 17965 + }, + { + "epoch": 0.69, + "learning_rate": 3.851243908694537e-05, + "loss": 0.5098, + "step": 17966 + }, + { + "epoch": 0.69, + "learning_rate": 3.851179789689664e-05, + "loss": 0.6863, + "step": 17967 + }, + { + "epoch": 0.69, + "learning_rate": 3.8511156706847915e-05, + "loss": 0.5367, + "step": 17968 + }, + { + "epoch": 0.69, + "learning_rate": 3.8510515516799187e-05, + "loss": 0.4949, + "step": 17969 + }, + { + "epoch": 0.69, + "learning_rate": 3.850987432675045e-05, + "loss": 0.6615, + "step": 17970 + }, + { + "epoch": 0.69, + "learning_rate": 3.850923313670172e-05, + "loss": 0.503, + "step": 17971 + }, + { + "epoch": 0.69, + "learning_rate": 3.850859194665299e-05, + "loss": 0.4763, + "step": 17972 + }, + { + "epoch": 0.69, + "learning_rate": 3.850795075660426e-05, + "loss": 0.5753, + "step": 17973 + }, + { + "epoch": 0.69, + "learning_rate": 3.8507309566555525e-05, + "loss": 0.4041, + "step": 17974 + }, + { + "epoch": 0.69, + "learning_rate": 3.8506668376506797e-05, + "loss": 0.7223, + "step": 17975 + }, + { + "epoch": 0.69, + "learning_rate": 3.850602718645807e-05, + "loss": 0.4732, + "step": 17976 + }, + { + "epoch": 0.69, + "learning_rate": 3.850538599640934e-05, + "loss": 0.4693, + "step": 17977 + }, + { + "epoch": 0.69, + "learning_rate": 3.850474480636061e-05, + "loss": 0.4504, + "step": 17978 + }, + { + "epoch": 0.69, + "learning_rate": 3.850410361631188e-05, + "loss": 0.5025, + "step": 17979 + }, + { + "epoch": 0.69, + "learning_rate": 3.850346242626315e-05, + "loss": 0.4572, + "step": 17980 + }, + { + "epoch": 0.69, + "learning_rate": 3.850282123621441e-05, + "loss": 0.5208, + "step": 17981 + }, + { + "epoch": 0.69, + "learning_rate": 3.8502180046165685e-05, + "loss": 0.5557, + "step": 17982 + }, + { + "epoch": 0.69, + "learning_rate": 3.850153885611695e-05, + "loss": 0.4105, + "step": 17983 + }, + { + "epoch": 0.69, + "learning_rate": 3.850089766606823e-05, + "loss": 0.6045, + "step": 17984 + }, + { + "epoch": 0.69, + "learning_rate": 3.850025647601949e-05, + "loss": 0.4881, + "step": 17985 + }, + { + "epoch": 0.69, + "learning_rate": 3.8499615285970765e-05, + "loss": 0.4523, + "step": 17986 + }, + { + "epoch": 0.69, + "learning_rate": 3.849897409592204e-05, + "loss": 0.5265, + "step": 17987 + }, + { + "epoch": 0.69, + "learning_rate": 3.84983329058733e-05, + "loss": 0.4339, + "step": 17988 + }, + { + "epoch": 0.69, + "learning_rate": 3.8497691715824573e-05, + "loss": 0.504, + "step": 17989 + }, + { + "epoch": 0.69, + "learning_rate": 3.849705052577584e-05, + "loss": 0.658, + "step": 17990 + }, + { + "epoch": 0.69, + "learning_rate": 3.849640933572711e-05, + "loss": 0.6479, + "step": 17991 + }, + { + "epoch": 0.69, + "learning_rate": 3.849576814567838e-05, + "loss": 0.6413, + "step": 17992 + }, + { + "epoch": 0.69, + "learning_rate": 3.8495126955629654e-05, + "loss": 0.4762, + "step": 17993 + }, + { + "epoch": 0.69, + "learning_rate": 3.849448576558092e-05, + "loss": 0.5855, + "step": 17994 + }, + { + "epoch": 0.69, + "learning_rate": 3.849384457553219e-05, + "loss": 0.5709, + "step": 17995 + }, + { + "epoch": 0.69, + "learning_rate": 3.849320338548346e-05, + "loss": 0.5612, + "step": 17996 + }, + { + "epoch": 0.69, + "learning_rate": 3.849256219543473e-05, + "loss": 0.4877, + "step": 17997 + }, + { + "epoch": 0.69, + "learning_rate": 3.8491921005386e-05, + "loss": 0.5991, + "step": 17998 + }, + { + "epoch": 0.69, + "learning_rate": 3.8491279815337264e-05, + "loss": 0.6326, + "step": 17999 + }, + { + "epoch": 0.69, + "learning_rate": 3.849063862528854e-05, + "loss": 0.5035, + "step": 18000 + }, + { + "epoch": 0.69, + "learning_rate": 3.848999743523981e-05, + "loss": 0.49, + "step": 18001 + }, + { + "epoch": 0.69, + "learning_rate": 3.848935624519108e-05, + "loss": 0.5199, + "step": 18002 + }, + { + "epoch": 0.69, + "learning_rate": 3.8488715055142344e-05, + "loss": 0.5529, + "step": 18003 + }, + { + "epoch": 0.69, + "learning_rate": 3.8488073865093615e-05, + "loss": 0.595, + "step": 18004 + }, + { + "epoch": 0.69, + "learning_rate": 3.848743267504489e-05, + "loss": 0.5346, + "step": 18005 + }, + { + "epoch": 0.69, + "learning_rate": 3.848679148499615e-05, + "loss": 0.5486, + "step": 18006 + }, + { + "epoch": 0.69, + "learning_rate": 3.8486150294947424e-05, + "loss": 0.5111, + "step": 18007 + }, + { + "epoch": 0.69, + "learning_rate": 3.8485509104898695e-05, + "loss": 0.5681, + "step": 18008 + }, + { + "epoch": 0.69, + "learning_rate": 3.848486791484997e-05, + "loss": 0.4939, + "step": 18009 + }, + { + "epoch": 0.69, + "learning_rate": 3.848422672480123e-05, + "loss": 0.5408, + "step": 18010 + }, + { + "epoch": 0.69, + "learning_rate": 3.8483585534752504e-05, + "loss": 0.5706, + "step": 18011 + }, + { + "epoch": 0.69, + "learning_rate": 3.848294434470377e-05, + "loss": 0.5934, + "step": 18012 + }, + { + "epoch": 0.69, + "learning_rate": 3.848230315465504e-05, + "loss": 0.5044, + "step": 18013 + }, + { + "epoch": 0.69, + "learning_rate": 3.848166196460631e-05, + "loss": 0.5638, + "step": 18014 + }, + { + "epoch": 0.69, + "learning_rate": 3.848102077455758e-05, + "loss": 0.6312, + "step": 18015 + }, + { + "epoch": 0.69, + "learning_rate": 3.848037958450885e-05, + "loss": 0.5376, + "step": 18016 + }, + { + "epoch": 0.69, + "learning_rate": 3.847973839446012e-05, + "loss": 0.504, + "step": 18017 + }, + { + "epoch": 0.69, + "learning_rate": 3.847909720441139e-05, + "loss": 0.465, + "step": 18018 + }, + { + "epoch": 0.69, + "learning_rate": 3.847845601436266e-05, + "loss": 0.5737, + "step": 18019 + }, + { + "epoch": 0.69, + "learning_rate": 3.847781482431393e-05, + "loss": 0.5183, + "step": 18020 + }, + { + "epoch": 0.69, + "learning_rate": 3.8477173634265194e-05, + "loss": 0.5659, + "step": 18021 + }, + { + "epoch": 0.69, + "learning_rate": 3.8476532444216466e-05, + "loss": 0.5812, + "step": 18022 + }, + { + "epoch": 0.69, + "learning_rate": 3.847589125416774e-05, + "loss": 0.5, + "step": 18023 + }, + { + "epoch": 0.69, + "learning_rate": 3.847525006411901e-05, + "loss": 0.5135, + "step": 18024 + }, + { + "epoch": 0.69, + "learning_rate": 3.847460887407028e-05, + "loss": 0.5222, + "step": 18025 + }, + { + "epoch": 0.69, + "learning_rate": 3.8473967684021546e-05, + "loss": 0.6212, + "step": 18026 + }, + { + "epoch": 0.69, + "learning_rate": 3.847332649397282e-05, + "loss": 0.4904, + "step": 18027 + }, + { + "epoch": 0.69, + "learning_rate": 3.847268530392408e-05, + "loss": 0.5362, + "step": 18028 + }, + { + "epoch": 0.69, + "learning_rate": 3.8472044113875354e-05, + "loss": 0.5122, + "step": 18029 + }, + { + "epoch": 0.69, + "learning_rate": 3.847140292382662e-05, + "loss": 0.5472, + "step": 18030 + }, + { + "epoch": 0.69, + "learning_rate": 3.847076173377789e-05, + "loss": 0.4993, + "step": 18031 + }, + { + "epoch": 0.69, + "learning_rate": 3.847012054372916e-05, + "loss": 0.5523, + "step": 18032 + }, + { + "epoch": 0.69, + "learning_rate": 3.8469479353680434e-05, + "loss": 0.4627, + "step": 18033 + }, + { + "epoch": 0.69, + "learning_rate": 3.8468838163631706e-05, + "loss": 0.4534, + "step": 18034 + }, + { + "epoch": 0.69, + "learning_rate": 3.846819697358297e-05, + "loss": 0.5625, + "step": 18035 + }, + { + "epoch": 0.69, + "learning_rate": 3.846755578353424e-05, + "loss": 0.4692, + "step": 18036 + }, + { + "epoch": 0.69, + "learning_rate": 3.846691459348551e-05, + "loss": 0.4618, + "step": 18037 + }, + { + "epoch": 0.69, + "learning_rate": 3.846627340343678e-05, + "loss": 0.5821, + "step": 18038 + }, + { + "epoch": 0.69, + "learning_rate": 3.846563221338805e-05, + "loss": 0.5486, + "step": 18039 + }, + { + "epoch": 0.69, + "learning_rate": 3.846499102333932e-05, + "loss": 0.5563, + "step": 18040 + }, + { + "epoch": 0.69, + "learning_rate": 3.8464349833290594e-05, + "loss": 0.6188, + "step": 18041 + }, + { + "epoch": 0.69, + "learning_rate": 3.846370864324186e-05, + "loss": 0.5871, + "step": 18042 + }, + { + "epoch": 0.69, + "learning_rate": 3.846306745319313e-05, + "loss": 0.4628, + "step": 18043 + }, + { + "epoch": 0.69, + "learning_rate": 3.8462426263144396e-05, + "loss": 0.4827, + "step": 18044 + }, + { + "epoch": 0.69, + "learning_rate": 3.846178507309567e-05, + "loss": 0.5313, + "step": 18045 + }, + { + "epoch": 0.69, + "learning_rate": 3.846114388304693e-05, + "loss": 0.6048, + "step": 18046 + }, + { + "epoch": 0.69, + "learning_rate": 3.8460502692998204e-05, + "loss": 0.4778, + "step": 18047 + }, + { + "epoch": 0.69, + "learning_rate": 3.8459861502949476e-05, + "loss": 0.673, + "step": 18048 + }, + { + "epoch": 0.69, + "learning_rate": 3.845922031290075e-05, + "loss": 0.6361, + "step": 18049 + }, + { + "epoch": 0.69, + "learning_rate": 3.845857912285202e-05, + "loss": 0.5268, + "step": 18050 + }, + { + "epoch": 0.69, + "learning_rate": 3.8457937932803284e-05, + "loss": 0.5674, + "step": 18051 + }, + { + "epoch": 0.69, + "learning_rate": 3.8457296742754556e-05, + "loss": 0.6396, + "step": 18052 + }, + { + "epoch": 0.69, + "learning_rate": 3.845665555270582e-05, + "loss": 0.548, + "step": 18053 + }, + { + "epoch": 0.69, + "learning_rate": 3.845601436265709e-05, + "loss": 0.4666, + "step": 18054 + }, + { + "epoch": 0.69, + "learning_rate": 3.845537317260836e-05, + "loss": 0.539, + "step": 18055 + }, + { + "epoch": 0.69, + "learning_rate": 3.8454731982559636e-05, + "loss": 0.5308, + "step": 18056 + }, + { + "epoch": 0.69, + "learning_rate": 3.84540907925109e-05, + "loss": 0.4303, + "step": 18057 + }, + { + "epoch": 0.69, + "learning_rate": 3.845344960246217e-05, + "loss": 0.5561, + "step": 18058 + }, + { + "epoch": 0.69, + "learning_rate": 3.8452808412413445e-05, + "loss": 0.5342, + "step": 18059 + }, + { + "epoch": 0.69, + "learning_rate": 3.845216722236471e-05, + "loss": 0.4978, + "step": 18060 + }, + { + "epoch": 0.69, + "learning_rate": 3.845152603231598e-05, + "loss": 0.4923, + "step": 18061 + }, + { + "epoch": 0.69, + "learning_rate": 3.8450884842267246e-05, + "loss": 0.529, + "step": 18062 + }, + { + "epoch": 0.69, + "learning_rate": 3.845024365221852e-05, + "loss": 0.5361, + "step": 18063 + }, + { + "epoch": 0.69, + "learning_rate": 3.844960246216979e-05, + "loss": 0.6384, + "step": 18064 + }, + { + "epoch": 0.69, + "learning_rate": 3.844896127212106e-05, + "loss": 0.4705, + "step": 18065 + }, + { + "epoch": 0.69, + "learning_rate": 3.8448320082072326e-05, + "loss": 0.5372, + "step": 18066 + }, + { + "epoch": 0.69, + "learning_rate": 3.84476788920236e-05, + "loss": 0.4996, + "step": 18067 + }, + { + "epoch": 0.69, + "learning_rate": 3.844703770197487e-05, + "loss": 0.447, + "step": 18068 + }, + { + "epoch": 0.69, + "learning_rate": 3.8446396511926135e-05, + "loss": 0.527, + "step": 18069 + }, + { + "epoch": 0.69, + "learning_rate": 3.8445755321877406e-05, + "loss": 0.5648, + "step": 18070 + }, + { + "epoch": 0.69, + "learning_rate": 3.844511413182867e-05, + "loss": 0.552, + "step": 18071 + }, + { + "epoch": 0.69, + "learning_rate": 3.844447294177995e-05, + "loss": 0.6014, + "step": 18072 + }, + { + "epoch": 0.69, + "learning_rate": 3.8443831751731215e-05, + "loss": 0.4386, + "step": 18073 + }, + { + "epoch": 0.69, + "learning_rate": 3.8443190561682486e-05, + "loss": 0.4692, + "step": 18074 + }, + { + "epoch": 0.69, + "learning_rate": 3.844254937163375e-05, + "loss": 0.4614, + "step": 18075 + }, + { + "epoch": 0.69, + "learning_rate": 3.844190818158502e-05, + "loss": 0.5178, + "step": 18076 + }, + { + "epoch": 0.7, + "learning_rate": 3.8441266991536295e-05, + "loss": 0.4953, + "step": 18077 + }, + { + "epoch": 0.7, + "learning_rate": 3.844062580148756e-05, + "loss": 0.5354, + "step": 18078 + }, + { + "epoch": 0.7, + "learning_rate": 3.843998461143883e-05, + "loss": 0.5241, + "step": 18079 + }, + { + "epoch": 0.7, + "learning_rate": 3.84393434213901e-05, + "loss": 0.5081, + "step": 18080 + }, + { + "epoch": 0.7, + "learning_rate": 3.8438702231341375e-05, + "loss": 0.4343, + "step": 18081 + }, + { + "epoch": 0.7, + "learning_rate": 3.843806104129264e-05, + "loss": 0.4352, + "step": 18082 + }, + { + "epoch": 0.7, + "learning_rate": 3.843741985124391e-05, + "loss": 0.5791, + "step": 18083 + }, + { + "epoch": 0.7, + "learning_rate": 3.8436778661195176e-05, + "loss": 0.5243, + "step": 18084 + }, + { + "epoch": 0.7, + "learning_rate": 3.843613747114645e-05, + "loss": 0.5563, + "step": 18085 + }, + { + "epoch": 0.7, + "learning_rate": 3.843549628109772e-05, + "loss": 0.4651, + "step": 18086 + }, + { + "epoch": 0.7, + "learning_rate": 3.8434855091048985e-05, + "loss": 0.4707, + "step": 18087 + }, + { + "epoch": 0.7, + "learning_rate": 3.843421390100026e-05, + "loss": 0.5829, + "step": 18088 + }, + { + "epoch": 0.7, + "learning_rate": 3.843357271095153e-05, + "loss": 0.5453, + "step": 18089 + }, + { + "epoch": 0.7, + "learning_rate": 3.84329315209028e-05, + "loss": 0.4641, + "step": 18090 + }, + { + "epoch": 0.7, + "learning_rate": 3.8432290330854065e-05, + "loss": 0.5222, + "step": 18091 + }, + { + "epoch": 0.7, + "learning_rate": 3.843164914080534e-05, + "loss": 0.4181, + "step": 18092 + }, + { + "epoch": 0.7, + "learning_rate": 3.84310079507566e-05, + "loss": 0.7471, + "step": 18093 + }, + { + "epoch": 0.7, + "learning_rate": 3.843036676070787e-05, + "loss": 0.6384, + "step": 18094 + }, + { + "epoch": 0.7, + "learning_rate": 3.8429725570659145e-05, + "loss": 0.5329, + "step": 18095 + }, + { + "epoch": 0.7, + "learning_rate": 3.842908438061042e-05, + "loss": 0.5134, + "step": 18096 + }, + { + "epoch": 0.7, + "learning_rate": 3.842844319056169e-05, + "loss": 0.4941, + "step": 18097 + }, + { + "epoch": 0.7, + "learning_rate": 3.8427802000512953e-05, + "loss": 0.537, + "step": 18098 + }, + { + "epoch": 0.7, + "learning_rate": 3.8427160810464225e-05, + "loss": 0.5505, + "step": 18099 + }, + { + "epoch": 0.7, + "learning_rate": 3.842651962041549e-05, + "loss": 0.5613, + "step": 18100 + }, + { + "epoch": 0.7, + "learning_rate": 3.842587843036676e-05, + "loss": 0.4361, + "step": 18101 + }, + { + "epoch": 0.7, + "learning_rate": 3.842523724031803e-05, + "loss": 0.5237, + "step": 18102 + }, + { + "epoch": 0.7, + "learning_rate": 3.84245960502693e-05, + "loss": 0.5466, + "step": 18103 + }, + { + "epoch": 0.7, + "learning_rate": 3.842395486022057e-05, + "loss": 0.4533, + "step": 18104 + }, + { + "epoch": 0.7, + "learning_rate": 3.842331367017184e-05, + "loss": 0.4374, + "step": 18105 + }, + { + "epoch": 0.7, + "learning_rate": 3.8422672480123114e-05, + "loss": 0.6049, + "step": 18106 + }, + { + "epoch": 0.7, + "learning_rate": 3.842203129007438e-05, + "loss": 0.632, + "step": 18107 + }, + { + "epoch": 0.7, + "learning_rate": 3.842139010002565e-05, + "loss": 0.6205, + "step": 18108 + }, + { + "epoch": 0.7, + "learning_rate": 3.8420748909976915e-05, + "loss": 0.4489, + "step": 18109 + }, + { + "epoch": 0.7, + "learning_rate": 3.842010771992819e-05, + "loss": 0.5566, + "step": 18110 + }, + { + "epoch": 0.7, + "learning_rate": 3.841946652987946e-05, + "loss": 0.6731, + "step": 18111 + }, + { + "epoch": 0.7, + "learning_rate": 3.841882533983073e-05, + "loss": 0.4367, + "step": 18112 + }, + { + "epoch": 0.7, + "learning_rate": 3.8418184149782e-05, + "loss": 0.5053, + "step": 18113 + }, + { + "epoch": 0.7, + "learning_rate": 3.841754295973327e-05, + "loss": 0.434, + "step": 18114 + }, + { + "epoch": 0.7, + "learning_rate": 3.841690176968454e-05, + "loss": 0.5725, + "step": 18115 + }, + { + "epoch": 0.7, + "learning_rate": 3.8416260579635804e-05, + "loss": 0.5385, + "step": 18116 + }, + { + "epoch": 0.7, + "learning_rate": 3.8415619389587075e-05, + "loss": 0.4967, + "step": 18117 + }, + { + "epoch": 0.7, + "learning_rate": 3.841497819953834e-05, + "loss": 0.5979, + "step": 18118 + }, + { + "epoch": 0.7, + "learning_rate": 3.841433700948961e-05, + "loss": 0.5081, + "step": 18119 + }, + { + "epoch": 0.7, + "learning_rate": 3.8413695819440884e-05, + "loss": 0.5006, + "step": 18120 + }, + { + "epoch": 0.7, + "learning_rate": 3.8413054629392155e-05, + "loss": 0.4967, + "step": 18121 + }, + { + "epoch": 0.7, + "learning_rate": 3.841241343934343e-05, + "loss": 0.4993, + "step": 18122 + }, + { + "epoch": 0.7, + "learning_rate": 3.841177224929469e-05, + "loss": 0.5051, + "step": 18123 + }, + { + "epoch": 0.7, + "learning_rate": 3.8411131059245964e-05, + "loss": 0.5067, + "step": 18124 + }, + { + "epoch": 0.7, + "learning_rate": 3.841048986919723e-05, + "loss": 0.4561, + "step": 18125 + }, + { + "epoch": 0.7, + "learning_rate": 3.84098486791485e-05, + "loss": 0.5318, + "step": 18126 + }, + { + "epoch": 0.7, + "learning_rate": 3.840920748909977e-05, + "loss": 0.505, + "step": 18127 + }, + { + "epoch": 0.7, + "learning_rate": 3.8408566299051044e-05, + "loss": 0.4773, + "step": 18128 + }, + { + "epoch": 0.7, + "learning_rate": 3.8407925109002316e-05, + "loss": 0.4851, + "step": 18129 + }, + { + "epoch": 0.7, + "learning_rate": 3.840728391895358e-05, + "loss": 0.4939, + "step": 18130 + }, + { + "epoch": 0.7, + "learning_rate": 3.840664272890485e-05, + "loss": 0.4653, + "step": 18131 + }, + { + "epoch": 0.7, + "learning_rate": 3.840600153885612e-05, + "loss": 0.4795, + "step": 18132 + }, + { + "epoch": 0.7, + "learning_rate": 3.840536034880739e-05, + "loss": 0.6099, + "step": 18133 + }, + { + "epoch": 0.7, + "learning_rate": 3.8404719158758654e-05, + "loss": 0.5782, + "step": 18134 + }, + { + "epoch": 0.7, + "learning_rate": 3.8404077968709926e-05, + "loss": 0.5763, + "step": 18135 + }, + { + "epoch": 0.7, + "learning_rate": 3.84034367786612e-05, + "loss": 0.5049, + "step": 18136 + }, + { + "epoch": 0.7, + "learning_rate": 3.840279558861247e-05, + "loss": 0.4548, + "step": 18137 + }, + { + "epoch": 0.7, + "learning_rate": 3.840215439856374e-05, + "loss": 0.4794, + "step": 18138 + }, + { + "epoch": 0.7, + "learning_rate": 3.8401513208515006e-05, + "loss": 0.6308, + "step": 18139 + }, + { + "epoch": 0.7, + "learning_rate": 3.840087201846628e-05, + "loss": 0.5622, + "step": 18140 + }, + { + "epoch": 0.7, + "learning_rate": 3.840023082841754e-05, + "loss": 0.5378, + "step": 18141 + }, + { + "epoch": 0.7, + "learning_rate": 3.8399589638368814e-05, + "loss": 0.4275, + "step": 18142 + }, + { + "epoch": 0.7, + "learning_rate": 3.839894844832008e-05, + "loss": 0.5504, + "step": 18143 + }, + { + "epoch": 0.7, + "learning_rate": 3.839830725827136e-05, + "loss": 0.4801, + "step": 18144 + }, + { + "epoch": 0.7, + "learning_rate": 3.839766606822262e-05, + "loss": 0.5512, + "step": 18145 + }, + { + "epoch": 0.7, + "learning_rate": 3.8397024878173894e-05, + "loss": 0.5026, + "step": 18146 + }, + { + "epoch": 0.7, + "learning_rate": 3.8396383688125166e-05, + "loss": 0.5249, + "step": 18147 + }, + { + "epoch": 0.7, + "learning_rate": 3.839574249807643e-05, + "loss": 0.543, + "step": 18148 + }, + { + "epoch": 0.7, + "learning_rate": 3.83951013080277e-05, + "loss": 0.5782, + "step": 18149 + }, + { + "epoch": 0.7, + "learning_rate": 3.839446011797897e-05, + "loss": 0.5247, + "step": 18150 + }, + { + "epoch": 0.7, + "learning_rate": 3.839381892793024e-05, + "loss": 0.4427, + "step": 18151 + }, + { + "epoch": 0.7, + "learning_rate": 3.839317773788151e-05, + "loss": 0.4796, + "step": 18152 + }, + { + "epoch": 0.7, + "learning_rate": 3.839253654783278e-05, + "loss": 0.4044, + "step": 18153 + }, + { + "epoch": 0.7, + "learning_rate": 3.839189535778405e-05, + "loss": 0.4918, + "step": 18154 + }, + { + "epoch": 0.7, + "learning_rate": 3.839125416773532e-05, + "loss": 0.4436, + "step": 18155 + }, + { + "epoch": 0.7, + "learning_rate": 3.839061297768659e-05, + "loss": 0.508, + "step": 18156 + }, + { + "epoch": 0.7, + "learning_rate": 3.8389971787637856e-05, + "loss": 0.4848, + "step": 18157 + }, + { + "epoch": 0.7, + "learning_rate": 3.838933059758913e-05, + "loss": 0.5639, + "step": 18158 + }, + { + "epoch": 0.7, + "learning_rate": 3.838868940754039e-05, + "loss": 0.6512, + "step": 18159 + }, + { + "epoch": 0.7, + "learning_rate": 3.838804821749167e-05, + "loss": 0.5254, + "step": 18160 + }, + { + "epoch": 0.7, + "learning_rate": 3.8387407027442936e-05, + "loss": 0.5886, + "step": 18161 + }, + { + "epoch": 0.7, + "learning_rate": 3.838676583739421e-05, + "loss": 0.4849, + "step": 18162 + }, + { + "epoch": 0.7, + "learning_rate": 3.838612464734547e-05, + "loss": 0.508, + "step": 18163 + }, + { + "epoch": 0.7, + "learning_rate": 3.8385483457296744e-05, + "loss": 0.6258, + "step": 18164 + }, + { + "epoch": 0.7, + "learning_rate": 3.8384842267248016e-05, + "loss": 0.6222, + "step": 18165 + }, + { + "epoch": 0.7, + "learning_rate": 3.838420107719928e-05, + "loss": 0.5565, + "step": 18166 + }, + { + "epoch": 0.7, + "learning_rate": 3.838355988715055e-05, + "loss": 0.4939, + "step": 18167 + }, + { + "epoch": 0.7, + "learning_rate": 3.8382918697101824e-05, + "loss": 0.4538, + "step": 18168 + }, + { + "epoch": 0.7, + "learning_rate": 3.8382277507053096e-05, + "loss": 0.5503, + "step": 18169 + }, + { + "epoch": 0.7, + "learning_rate": 3.838163631700436e-05, + "loss": 0.5971, + "step": 18170 + }, + { + "epoch": 0.7, + "learning_rate": 3.838099512695563e-05, + "loss": 0.4997, + "step": 18171 + }, + { + "epoch": 0.7, + "learning_rate": 3.83803539369069e-05, + "loss": 0.5018, + "step": 18172 + }, + { + "epoch": 0.7, + "learning_rate": 3.837971274685817e-05, + "loss": 0.4028, + "step": 18173 + }, + { + "epoch": 0.7, + "learning_rate": 3.837907155680944e-05, + "loss": 0.4439, + "step": 18174 + }, + { + "epoch": 0.7, + "learning_rate": 3.8378430366760706e-05, + "loss": 0.4788, + "step": 18175 + }, + { + "epoch": 0.7, + "learning_rate": 3.837778917671198e-05, + "loss": 0.5471, + "step": 18176 + }, + { + "epoch": 0.7, + "learning_rate": 3.837714798666325e-05, + "loss": 0.5436, + "step": 18177 + }, + { + "epoch": 0.7, + "learning_rate": 3.837650679661452e-05, + "loss": 0.5424, + "step": 18178 + }, + { + "epoch": 0.7, + "learning_rate": 3.8375865606565786e-05, + "loss": 0.4922, + "step": 18179 + }, + { + "epoch": 0.7, + "learning_rate": 3.837522441651706e-05, + "loss": 0.5308, + "step": 18180 + }, + { + "epoch": 0.7, + "learning_rate": 3.837458322646832e-05, + "loss": 0.5223, + "step": 18181 + }, + { + "epoch": 0.7, + "learning_rate": 3.8373942036419595e-05, + "loss": 0.579, + "step": 18182 + }, + { + "epoch": 0.7, + "learning_rate": 3.8373300846370866e-05, + "loss": 0.5366, + "step": 18183 + }, + { + "epoch": 0.7, + "learning_rate": 3.837265965632214e-05, + "loss": 0.6025, + "step": 18184 + }, + { + "epoch": 0.7, + "learning_rate": 3.837201846627341e-05, + "loss": 0.4248, + "step": 18185 + }, + { + "epoch": 0.7, + "learning_rate": 3.8371377276224675e-05, + "loss": 0.5476, + "step": 18186 + }, + { + "epoch": 0.7, + "learning_rate": 3.8370736086175946e-05, + "loss": 0.5922, + "step": 18187 + }, + { + "epoch": 0.7, + "learning_rate": 3.837009489612721e-05, + "loss": 0.4829, + "step": 18188 + }, + { + "epoch": 0.7, + "learning_rate": 3.836945370607848e-05, + "loss": 0.4553, + "step": 18189 + }, + { + "epoch": 0.7, + "learning_rate": 3.836881251602975e-05, + "loss": 0.5683, + "step": 18190 + }, + { + "epoch": 0.7, + "learning_rate": 3.836817132598102e-05, + "loss": 0.4849, + "step": 18191 + }, + { + "epoch": 0.7, + "learning_rate": 3.836753013593229e-05, + "loss": 0.5109, + "step": 18192 + }, + { + "epoch": 0.7, + "learning_rate": 3.836688894588356e-05, + "loss": 0.5453, + "step": 18193 + }, + { + "epoch": 0.7, + "learning_rate": 3.8366247755834835e-05, + "loss": 0.4958, + "step": 18194 + }, + { + "epoch": 0.7, + "learning_rate": 3.83656065657861e-05, + "loss": 0.446, + "step": 18195 + }, + { + "epoch": 0.7, + "learning_rate": 3.836496537573737e-05, + "loss": 0.4145, + "step": 18196 + }, + { + "epoch": 0.7, + "learning_rate": 3.8364324185688636e-05, + "loss": 0.466, + "step": 18197 + }, + { + "epoch": 0.7, + "learning_rate": 3.836368299563991e-05, + "loss": 0.5315, + "step": 18198 + }, + { + "epoch": 0.7, + "learning_rate": 3.836304180559118e-05, + "loss": 0.482, + "step": 18199 + }, + { + "epoch": 0.7, + "learning_rate": 3.836240061554245e-05, + "loss": 0.4989, + "step": 18200 + }, + { + "epoch": 0.7, + "learning_rate": 3.836175942549372e-05, + "loss": 0.5344, + "step": 18201 + }, + { + "epoch": 0.7, + "learning_rate": 3.836111823544499e-05, + "loss": 0.5884, + "step": 18202 + }, + { + "epoch": 0.7, + "learning_rate": 3.836047704539626e-05, + "loss": 0.597, + "step": 18203 + }, + { + "epoch": 0.7, + "learning_rate": 3.8359835855347525e-05, + "loss": 0.4161, + "step": 18204 + }, + { + "epoch": 0.7, + "learning_rate": 3.83591946652988e-05, + "loss": 0.5728, + "step": 18205 + }, + { + "epoch": 0.7, + "learning_rate": 3.835855347525006e-05, + "loss": 0.5862, + "step": 18206 + }, + { + "epoch": 0.7, + "learning_rate": 3.835791228520133e-05, + "loss": 0.444, + "step": 18207 + }, + { + "epoch": 0.7, + "learning_rate": 3.8357271095152605e-05, + "loss": 0.4037, + "step": 18208 + }, + { + "epoch": 0.7, + "learning_rate": 3.835662990510388e-05, + "loss": 0.4539, + "step": 18209 + }, + { + "epoch": 0.7, + "learning_rate": 3.835598871505515e-05, + "loss": 0.5017, + "step": 18210 + }, + { + "epoch": 0.7, + "learning_rate": 3.8355347525006413e-05, + "loss": 0.4876, + "step": 18211 + }, + { + "epoch": 0.7, + "learning_rate": 3.8354706334957685e-05, + "loss": 0.5795, + "step": 18212 + }, + { + "epoch": 0.7, + "learning_rate": 3.835406514490895e-05, + "loss": 0.6621, + "step": 18213 + }, + { + "epoch": 0.7, + "learning_rate": 3.835342395486022e-05, + "loss": 0.552, + "step": 18214 + }, + { + "epoch": 0.7, + "learning_rate": 3.8352782764811494e-05, + "loss": 0.5539, + "step": 18215 + }, + { + "epoch": 0.7, + "learning_rate": 3.8352141574762765e-05, + "loss": 0.4935, + "step": 18216 + }, + { + "epoch": 0.7, + "learning_rate": 3.835150038471403e-05, + "loss": 0.4313, + "step": 18217 + }, + { + "epoch": 0.7, + "learning_rate": 3.83508591946653e-05, + "loss": 0.6063, + "step": 18218 + }, + { + "epoch": 0.7, + "learning_rate": 3.8350218004616574e-05, + "loss": 0.4382, + "step": 18219 + }, + { + "epoch": 0.7, + "learning_rate": 3.834957681456784e-05, + "loss": 0.5583, + "step": 18220 + }, + { + "epoch": 0.7, + "learning_rate": 3.834893562451911e-05, + "loss": 0.6546, + "step": 18221 + }, + { + "epoch": 0.7, + "learning_rate": 3.8348294434470375e-05, + "loss": 0.6326, + "step": 18222 + }, + { + "epoch": 0.7, + "learning_rate": 3.834765324442165e-05, + "loss": 0.5235, + "step": 18223 + }, + { + "epoch": 0.7, + "learning_rate": 3.834701205437292e-05, + "loss": 0.5011, + "step": 18224 + }, + { + "epoch": 0.7, + "learning_rate": 3.834637086432419e-05, + "loss": 0.5246, + "step": 18225 + }, + { + "epoch": 0.7, + "learning_rate": 3.8345729674275455e-05, + "loss": 0.4745, + "step": 18226 + }, + { + "epoch": 0.7, + "learning_rate": 3.834508848422673e-05, + "loss": 0.5431, + "step": 18227 + }, + { + "epoch": 0.7, + "learning_rate": 3.8344447294178e-05, + "loss": 0.5486, + "step": 18228 + }, + { + "epoch": 0.7, + "learning_rate": 3.8343806104129264e-05, + "loss": 0.5659, + "step": 18229 + }, + { + "epoch": 0.7, + "learning_rate": 3.8343164914080535e-05, + "loss": 0.5358, + "step": 18230 + }, + { + "epoch": 0.7, + "learning_rate": 3.83425237240318e-05, + "loss": 0.5642, + "step": 18231 + }, + { + "epoch": 0.7, + "learning_rate": 3.834188253398308e-05, + "loss": 0.447, + "step": 18232 + }, + { + "epoch": 0.7, + "learning_rate": 3.8341241343934344e-05, + "loss": 0.4466, + "step": 18233 + }, + { + "epoch": 0.7, + "learning_rate": 3.8340600153885615e-05, + "loss": 0.5284, + "step": 18234 + }, + { + "epoch": 0.7, + "learning_rate": 3.833995896383688e-05, + "loss": 0.4991, + "step": 18235 + }, + { + "epoch": 0.7, + "learning_rate": 3.833931777378815e-05, + "loss": 0.5148, + "step": 18236 + }, + { + "epoch": 0.7, + "learning_rate": 3.8338676583739424e-05, + "loss": 0.4628, + "step": 18237 + }, + { + "epoch": 0.7, + "learning_rate": 3.833803539369069e-05, + "loss": 0.5856, + "step": 18238 + }, + { + "epoch": 0.7, + "learning_rate": 3.833739420364196e-05, + "loss": 0.6658, + "step": 18239 + }, + { + "epoch": 0.7, + "learning_rate": 3.833675301359323e-05, + "loss": 0.5233, + "step": 18240 + }, + { + "epoch": 0.7, + "learning_rate": 3.8336111823544504e-05, + "loss": 0.579, + "step": 18241 + }, + { + "epoch": 0.7, + "learning_rate": 3.833547063349577e-05, + "loss": 0.5983, + "step": 18242 + }, + { + "epoch": 0.7, + "learning_rate": 3.833482944344704e-05, + "loss": 0.4875, + "step": 18243 + }, + { + "epoch": 0.7, + "learning_rate": 3.8334188253398306e-05, + "loss": 0.4571, + "step": 18244 + }, + { + "epoch": 0.7, + "learning_rate": 3.833354706334958e-05, + "loss": 0.5837, + "step": 18245 + }, + { + "epoch": 0.7, + "learning_rate": 3.833290587330085e-05, + "loss": 0.5709, + "step": 18246 + }, + { + "epoch": 0.7, + "learning_rate": 3.8332264683252114e-05, + "loss": 0.5185, + "step": 18247 + }, + { + "epoch": 0.7, + "learning_rate": 3.833162349320339e-05, + "loss": 0.5565, + "step": 18248 + }, + { + "epoch": 0.7, + "learning_rate": 3.833098230315466e-05, + "loss": 0.4616, + "step": 18249 + }, + { + "epoch": 0.7, + "learning_rate": 3.833034111310593e-05, + "loss": 0.5451, + "step": 18250 + }, + { + "epoch": 0.7, + "learning_rate": 3.8329699923057194e-05, + "loss": 0.422, + "step": 18251 + }, + { + "epoch": 0.7, + "learning_rate": 3.8329058733008466e-05, + "loss": 0.5963, + "step": 18252 + }, + { + "epoch": 0.7, + "learning_rate": 3.832841754295973e-05, + "loss": 0.5449, + "step": 18253 + }, + { + "epoch": 0.7, + "learning_rate": 3.8327776352911e-05, + "loss": 0.5039, + "step": 18254 + }, + { + "epoch": 0.7, + "learning_rate": 3.8327135162862274e-05, + "loss": 0.5349, + "step": 18255 + }, + { + "epoch": 0.7, + "learning_rate": 3.8326493972813546e-05, + "loss": 0.5425, + "step": 18256 + }, + { + "epoch": 0.7, + "learning_rate": 3.832585278276482e-05, + "loss": 0.4125, + "step": 18257 + }, + { + "epoch": 0.7, + "learning_rate": 3.832521159271608e-05, + "loss": 0.5866, + "step": 18258 + }, + { + "epoch": 0.7, + "learning_rate": 3.8324570402667354e-05, + "loss": 0.5704, + "step": 18259 + }, + { + "epoch": 0.7, + "learning_rate": 3.832392921261862e-05, + "loss": 0.5844, + "step": 18260 + }, + { + "epoch": 0.7, + "learning_rate": 3.832328802256989e-05, + "loss": 0.5202, + "step": 18261 + }, + { + "epoch": 0.7, + "learning_rate": 3.8322646832521156e-05, + "loss": 0.4733, + "step": 18262 + }, + { + "epoch": 0.7, + "learning_rate": 3.832200564247243e-05, + "loss": 0.5001, + "step": 18263 + }, + { + "epoch": 0.7, + "learning_rate": 3.83213644524237e-05, + "loss": 0.5413, + "step": 18264 + }, + { + "epoch": 0.7, + "learning_rate": 3.832072326237497e-05, + "loss": 0.5339, + "step": 18265 + }, + { + "epoch": 0.7, + "learning_rate": 3.832008207232624e-05, + "loss": 0.5154, + "step": 18266 + }, + { + "epoch": 0.7, + "learning_rate": 3.831944088227751e-05, + "loss": 0.5788, + "step": 18267 + }, + { + "epoch": 0.7, + "learning_rate": 3.831879969222878e-05, + "loss": 0.443, + "step": 18268 + }, + { + "epoch": 0.7, + "learning_rate": 3.8318158502180044e-05, + "loss": 0.5226, + "step": 18269 + }, + { + "epoch": 0.7, + "learning_rate": 3.8317517312131316e-05, + "loss": 0.442, + "step": 18270 + }, + { + "epoch": 0.7, + "learning_rate": 3.831687612208259e-05, + "loss": 0.5391, + "step": 18271 + }, + { + "epoch": 0.7, + "learning_rate": 3.831623493203386e-05, + "loss": 0.6129, + "step": 18272 + }, + { + "epoch": 0.7, + "learning_rate": 3.831559374198513e-05, + "loss": 0.4493, + "step": 18273 + }, + { + "epoch": 0.7, + "learning_rate": 3.8314952551936396e-05, + "loss": 0.503, + "step": 18274 + }, + { + "epoch": 0.7, + "learning_rate": 3.831431136188767e-05, + "loss": 0.4377, + "step": 18275 + }, + { + "epoch": 0.7, + "learning_rate": 3.831367017183893e-05, + "loss": 0.4611, + "step": 18276 + }, + { + "epoch": 0.7, + "learning_rate": 3.8313028981790204e-05, + "loss": 0.4699, + "step": 18277 + }, + { + "epoch": 0.7, + "learning_rate": 3.831238779174147e-05, + "loss": 0.6517, + "step": 18278 + }, + { + "epoch": 0.7, + "learning_rate": 3.831174660169274e-05, + "loss": 0.4823, + "step": 18279 + }, + { + "epoch": 0.7, + "learning_rate": 3.831110541164401e-05, + "loss": 0.4582, + "step": 18280 + }, + { + "epoch": 0.7, + "learning_rate": 3.8310464221595285e-05, + "loss": 0.5478, + "step": 18281 + }, + { + "epoch": 0.7, + "learning_rate": 3.8309823031546556e-05, + "loss": 0.4664, + "step": 18282 + }, + { + "epoch": 0.7, + "learning_rate": 3.830918184149782e-05, + "loss": 0.55, + "step": 18283 + }, + { + "epoch": 0.7, + "learning_rate": 3.830854065144909e-05, + "loss": 0.491, + "step": 18284 + }, + { + "epoch": 0.7, + "learning_rate": 3.830789946140036e-05, + "loss": 0.5184, + "step": 18285 + }, + { + "epoch": 0.7, + "learning_rate": 3.830725827135163e-05, + "loss": 0.5092, + "step": 18286 + }, + { + "epoch": 0.7, + "learning_rate": 3.83066170813029e-05, + "loss": 0.5167, + "step": 18287 + }, + { + "epoch": 0.7, + "learning_rate": 3.830597589125417e-05, + "loss": 0.5088, + "step": 18288 + }, + { + "epoch": 0.7, + "learning_rate": 3.830533470120544e-05, + "loss": 0.4842, + "step": 18289 + }, + { + "epoch": 0.7, + "learning_rate": 3.830469351115671e-05, + "loss": 0.4904, + "step": 18290 + }, + { + "epoch": 0.7, + "learning_rate": 3.830405232110798e-05, + "loss": 0.5668, + "step": 18291 + }, + { + "epoch": 0.7, + "learning_rate": 3.8303411131059246e-05, + "loss": 0.4906, + "step": 18292 + }, + { + "epoch": 0.7, + "learning_rate": 3.830276994101052e-05, + "loss": 0.4191, + "step": 18293 + }, + { + "epoch": 0.7, + "learning_rate": 3.830212875096178e-05, + "loss": 0.4904, + "step": 18294 + }, + { + "epoch": 0.7, + "learning_rate": 3.8301487560913055e-05, + "loss": 0.5467, + "step": 18295 + }, + { + "epoch": 0.7, + "learning_rate": 3.8300846370864326e-05, + "loss": 0.4821, + "step": 18296 + }, + { + "epoch": 0.7, + "learning_rate": 3.83002051808156e-05, + "loss": 0.5117, + "step": 18297 + }, + { + "epoch": 0.7, + "learning_rate": 3.829956399076686e-05, + "loss": 0.6581, + "step": 18298 + }, + { + "epoch": 0.7, + "learning_rate": 3.8298922800718135e-05, + "loss": 0.5338, + "step": 18299 + }, + { + "epoch": 0.7, + "learning_rate": 3.8298281610669406e-05, + "loss": 0.6321, + "step": 18300 + }, + { + "epoch": 0.7, + "learning_rate": 3.829764042062067e-05, + "loss": 0.6853, + "step": 18301 + }, + { + "epoch": 0.7, + "learning_rate": 3.829699923057194e-05, + "loss": 0.4721, + "step": 18302 + }, + { + "epoch": 0.7, + "learning_rate": 3.829635804052321e-05, + "loss": 0.4644, + "step": 18303 + }, + { + "epoch": 0.7, + "learning_rate": 3.8295716850474487e-05, + "loss": 0.5295, + "step": 18304 + }, + { + "epoch": 0.7, + "learning_rate": 3.829507566042575e-05, + "loss": 0.6147, + "step": 18305 + }, + { + "epoch": 0.7, + "learning_rate": 3.829443447037702e-05, + "loss": 0.5772, + "step": 18306 + }, + { + "epoch": 0.7, + "learning_rate": 3.829379328032829e-05, + "loss": 0.4714, + "step": 18307 + }, + { + "epoch": 0.7, + "learning_rate": 3.829315209027956e-05, + "loss": 0.51, + "step": 18308 + }, + { + "epoch": 0.7, + "learning_rate": 3.829251090023083e-05, + "loss": 0.4827, + "step": 18309 + }, + { + "epoch": 0.7, + "learning_rate": 3.8291869710182097e-05, + "loss": 0.4466, + "step": 18310 + }, + { + "epoch": 0.7, + "learning_rate": 3.829122852013337e-05, + "loss": 0.515, + "step": 18311 + }, + { + "epoch": 0.7, + "learning_rate": 3.829058733008464e-05, + "loss": 0.536, + "step": 18312 + }, + { + "epoch": 0.7, + "learning_rate": 3.828994614003591e-05, + "loss": 0.5027, + "step": 18313 + }, + { + "epoch": 0.7, + "learning_rate": 3.828930494998718e-05, + "loss": 0.5672, + "step": 18314 + }, + { + "epoch": 0.7, + "learning_rate": 3.828866375993845e-05, + "loss": 0.5785, + "step": 18315 + }, + { + "epoch": 0.7, + "learning_rate": 3.828802256988972e-05, + "loss": 0.4244, + "step": 18316 + }, + { + "epoch": 0.7, + "learning_rate": 3.8287381379840985e-05, + "loss": 0.5652, + "step": 18317 + }, + { + "epoch": 0.7, + "learning_rate": 3.828674018979226e-05, + "loss": 0.4585, + "step": 18318 + }, + { + "epoch": 0.7, + "learning_rate": 3.828609899974352e-05, + "loss": 0.5334, + "step": 18319 + }, + { + "epoch": 0.7, + "learning_rate": 3.82854578096948e-05, + "loss": 0.5798, + "step": 18320 + }, + { + "epoch": 0.7, + "learning_rate": 3.8284816619646065e-05, + "loss": 0.5572, + "step": 18321 + }, + { + "epoch": 0.7, + "learning_rate": 3.828417542959734e-05, + "loss": 0.4982, + "step": 18322 + }, + { + "epoch": 0.7, + "learning_rate": 3.82835342395486e-05, + "loss": 0.4914, + "step": 18323 + }, + { + "epoch": 0.7, + "learning_rate": 3.8282893049499873e-05, + "loss": 0.5518, + "step": 18324 + }, + { + "epoch": 0.7, + "learning_rate": 3.8282251859451145e-05, + "loss": 0.442, + "step": 18325 + }, + { + "epoch": 0.7, + "learning_rate": 3.828161066940241e-05, + "loss": 0.452, + "step": 18326 + }, + { + "epoch": 0.7, + "learning_rate": 3.828096947935368e-05, + "loss": 0.5853, + "step": 18327 + }, + { + "epoch": 0.7, + "learning_rate": 3.8280328289304954e-05, + "loss": 0.4779, + "step": 18328 + }, + { + "epoch": 0.7, + "learning_rate": 3.8279687099256225e-05, + "loss": 0.5289, + "step": 18329 + }, + { + "epoch": 0.7, + "learning_rate": 3.827904590920749e-05, + "loss": 0.6204, + "step": 18330 + }, + { + "epoch": 0.7, + "learning_rate": 3.827840471915876e-05, + "loss": 0.566, + "step": 18331 + }, + { + "epoch": 0.7, + "learning_rate": 3.827776352911003e-05, + "loss": 0.545, + "step": 18332 + }, + { + "epoch": 0.7, + "learning_rate": 3.82771223390613e-05, + "loss": 0.5813, + "step": 18333 + }, + { + "epoch": 0.7, + "learning_rate": 3.827648114901257e-05, + "loss": 0.5049, + "step": 18334 + }, + { + "epoch": 0.7, + "learning_rate": 3.8275839958963835e-05, + "loss": 0.4601, + "step": 18335 + }, + { + "epoch": 0.7, + "learning_rate": 3.8275198768915114e-05, + "loss": 0.4946, + "step": 18336 + }, + { + "epoch": 0.7, + "learning_rate": 3.827455757886638e-05, + "loss": 0.4489, + "step": 18337 + }, + { + "epoch": 0.71, + "learning_rate": 3.827391638881765e-05, + "loss": 0.5133, + "step": 18338 + }, + { + "epoch": 0.71, + "learning_rate": 3.8273275198768915e-05, + "loss": 0.5725, + "step": 18339 + }, + { + "epoch": 0.71, + "learning_rate": 3.827263400872019e-05, + "loss": 0.5143, + "step": 18340 + }, + { + "epoch": 0.71, + "learning_rate": 3.827199281867145e-05, + "loss": 0.5342, + "step": 18341 + }, + { + "epoch": 0.71, + "learning_rate": 3.8271351628622724e-05, + "loss": 0.5489, + "step": 18342 + }, + { + "epoch": 0.71, + "learning_rate": 3.8270710438573995e-05, + "loss": 0.6103, + "step": 18343 + }, + { + "epoch": 0.71, + "learning_rate": 3.827006924852527e-05, + "loss": 0.5461, + "step": 18344 + }, + { + "epoch": 0.71, + "learning_rate": 3.826942805847654e-05, + "loss": 0.4614, + "step": 18345 + }, + { + "epoch": 0.71, + "learning_rate": 3.8268786868427804e-05, + "loss": 0.4908, + "step": 18346 + }, + { + "epoch": 0.71, + "learning_rate": 3.8268145678379076e-05, + "loss": 0.5046, + "step": 18347 + }, + { + "epoch": 0.71, + "learning_rate": 3.826750448833034e-05, + "loss": 0.5914, + "step": 18348 + }, + { + "epoch": 0.71, + "learning_rate": 3.826686329828161e-05, + "loss": 0.5301, + "step": 18349 + }, + { + "epoch": 0.71, + "learning_rate": 3.826622210823288e-05, + "loss": 0.5753, + "step": 18350 + }, + { + "epoch": 0.71, + "learning_rate": 3.826558091818415e-05, + "loss": 0.4935, + "step": 18351 + }, + { + "epoch": 0.71, + "learning_rate": 3.826493972813542e-05, + "loss": 0.5411, + "step": 18352 + }, + { + "epoch": 0.71, + "learning_rate": 3.826429853808669e-05, + "loss": 0.4356, + "step": 18353 + }, + { + "epoch": 0.71, + "learning_rate": 3.8263657348037964e-05, + "loss": 0.5096, + "step": 18354 + }, + { + "epoch": 0.71, + "learning_rate": 3.826301615798923e-05, + "loss": 0.5114, + "step": 18355 + }, + { + "epoch": 0.71, + "learning_rate": 3.82623749679405e-05, + "loss": 0.5426, + "step": 18356 + }, + { + "epoch": 0.71, + "learning_rate": 3.8261733777891766e-05, + "loss": 0.4656, + "step": 18357 + }, + { + "epoch": 0.71, + "learning_rate": 3.826109258784304e-05, + "loss": 0.467, + "step": 18358 + }, + { + "epoch": 0.71, + "learning_rate": 3.826045139779431e-05, + "loss": 0.4653, + "step": 18359 + }, + { + "epoch": 0.71, + "learning_rate": 3.825981020774558e-05, + "loss": 0.581, + "step": 18360 + }, + { + "epoch": 0.71, + "learning_rate": 3.825916901769685e-05, + "loss": 0.487, + "step": 18361 + }, + { + "epoch": 0.71, + "learning_rate": 3.825852782764812e-05, + "loss": 0.4622, + "step": 18362 + }, + { + "epoch": 0.71, + "learning_rate": 3.825788663759939e-05, + "loss": 0.4846, + "step": 18363 + }, + { + "epoch": 0.71, + "learning_rate": 3.8257245447550654e-05, + "loss": 0.5476, + "step": 18364 + }, + { + "epoch": 0.71, + "learning_rate": 3.8256604257501926e-05, + "loss": 0.4987, + "step": 18365 + }, + { + "epoch": 0.71, + "learning_rate": 3.825596306745319e-05, + "loss": 0.4622, + "step": 18366 + }, + { + "epoch": 0.71, + "learning_rate": 3.825532187740446e-05, + "loss": 0.5296, + "step": 18367 + }, + { + "epoch": 0.71, + "learning_rate": 3.8254680687355734e-05, + "loss": 0.4934, + "step": 18368 + }, + { + "epoch": 0.71, + "learning_rate": 3.8254039497307006e-05, + "loss": 0.5493, + "step": 18369 + }, + { + "epoch": 0.71, + "learning_rate": 3.825339830725828e-05, + "loss": 0.5029, + "step": 18370 + }, + { + "epoch": 0.71, + "learning_rate": 3.825275711720954e-05, + "loss": 0.4988, + "step": 18371 + }, + { + "epoch": 0.71, + "learning_rate": 3.8252115927160814e-05, + "loss": 0.6135, + "step": 18372 + }, + { + "epoch": 0.71, + "learning_rate": 3.825147473711208e-05, + "loss": 0.5126, + "step": 18373 + }, + { + "epoch": 0.71, + "learning_rate": 3.825083354706335e-05, + "loss": 0.5433, + "step": 18374 + }, + { + "epoch": 0.71, + "learning_rate": 3.825019235701462e-05, + "loss": 0.6851, + "step": 18375 + }, + { + "epoch": 0.71, + "learning_rate": 3.8249551166965894e-05, + "loss": 0.6216, + "step": 18376 + }, + { + "epoch": 0.71, + "learning_rate": 3.824890997691716e-05, + "loss": 0.5548, + "step": 18377 + }, + { + "epoch": 0.71, + "learning_rate": 3.824826878686843e-05, + "loss": 0.5043, + "step": 18378 + }, + { + "epoch": 0.71, + "learning_rate": 3.82476275968197e-05, + "loss": 0.506, + "step": 18379 + }, + { + "epoch": 0.71, + "learning_rate": 3.824698640677097e-05, + "loss": 0.5991, + "step": 18380 + }, + { + "epoch": 0.71, + "learning_rate": 3.824634521672224e-05, + "loss": 0.5482, + "step": 18381 + }, + { + "epoch": 0.71, + "learning_rate": 3.8245704026673504e-05, + "loss": 0.5719, + "step": 18382 + }, + { + "epoch": 0.71, + "learning_rate": 3.8245062836624776e-05, + "loss": 0.6514, + "step": 18383 + }, + { + "epoch": 0.71, + "learning_rate": 3.824442164657605e-05, + "loss": 0.5531, + "step": 18384 + }, + { + "epoch": 0.71, + "learning_rate": 3.824378045652732e-05, + "loss": 0.5158, + "step": 18385 + }, + { + "epoch": 0.71, + "learning_rate": 3.8243139266478584e-05, + "loss": 0.4828, + "step": 18386 + }, + { + "epoch": 0.71, + "learning_rate": 3.8242498076429856e-05, + "loss": 0.4854, + "step": 18387 + }, + { + "epoch": 0.71, + "learning_rate": 3.824185688638113e-05, + "loss": 0.5224, + "step": 18388 + }, + { + "epoch": 0.71, + "learning_rate": 3.824121569633239e-05, + "loss": 0.5642, + "step": 18389 + }, + { + "epoch": 0.71, + "learning_rate": 3.8240574506283664e-05, + "loss": 0.5518, + "step": 18390 + }, + { + "epoch": 0.71, + "learning_rate": 3.823993331623493e-05, + "loss": 0.5676, + "step": 18391 + }, + { + "epoch": 0.71, + "learning_rate": 3.823929212618621e-05, + "loss": 0.5526, + "step": 18392 + }, + { + "epoch": 0.71, + "learning_rate": 3.823865093613747e-05, + "loss": 0.6126, + "step": 18393 + }, + { + "epoch": 0.71, + "learning_rate": 3.8238009746088745e-05, + "loss": 0.5339, + "step": 18394 + }, + { + "epoch": 0.71, + "learning_rate": 3.823736855604001e-05, + "loss": 0.5854, + "step": 18395 + }, + { + "epoch": 0.71, + "learning_rate": 3.823672736599128e-05, + "loss": 0.498, + "step": 18396 + }, + { + "epoch": 0.71, + "learning_rate": 3.823608617594255e-05, + "loss": 0.4524, + "step": 18397 + }, + { + "epoch": 0.71, + "learning_rate": 3.823544498589382e-05, + "loss": 0.5321, + "step": 18398 + }, + { + "epoch": 0.71, + "learning_rate": 3.823480379584509e-05, + "loss": 0.5198, + "step": 18399 + }, + { + "epoch": 0.71, + "learning_rate": 3.823416260579636e-05, + "loss": 0.4652, + "step": 18400 + }, + { + "epoch": 0.71, + "learning_rate": 3.823352141574763e-05, + "loss": 0.4706, + "step": 18401 + }, + { + "epoch": 0.71, + "learning_rate": 3.82328802256989e-05, + "loss": 0.6196, + "step": 18402 + }, + { + "epoch": 0.71, + "learning_rate": 3.823223903565017e-05, + "loss": 0.4168, + "step": 18403 + }, + { + "epoch": 0.71, + "learning_rate": 3.8231597845601435e-05, + "loss": 0.4973, + "step": 18404 + }, + { + "epoch": 0.71, + "learning_rate": 3.8230956655552706e-05, + "loss": 0.5881, + "step": 18405 + }, + { + "epoch": 0.71, + "learning_rate": 3.823031546550398e-05, + "loss": 0.5372, + "step": 18406 + }, + { + "epoch": 0.71, + "learning_rate": 3.822967427545524e-05, + "loss": 0.4547, + "step": 18407 + }, + { + "epoch": 0.71, + "learning_rate": 3.822903308540652e-05, + "loss": 0.4999, + "step": 18408 + }, + { + "epoch": 0.71, + "learning_rate": 3.8228391895357786e-05, + "loss": 0.513, + "step": 18409 + }, + { + "epoch": 0.71, + "learning_rate": 3.822775070530906e-05, + "loss": 0.4948, + "step": 18410 + }, + { + "epoch": 0.71, + "learning_rate": 3.822710951526032e-05, + "loss": 0.4715, + "step": 18411 + }, + { + "epoch": 0.71, + "learning_rate": 3.8226468325211595e-05, + "loss": 0.5796, + "step": 18412 + }, + { + "epoch": 0.71, + "learning_rate": 3.822582713516286e-05, + "loss": 0.4733, + "step": 18413 + }, + { + "epoch": 0.71, + "learning_rate": 3.822518594511413e-05, + "loss": 0.46, + "step": 18414 + }, + { + "epoch": 0.71, + "learning_rate": 3.82245447550654e-05, + "loss": 0.4553, + "step": 18415 + }, + { + "epoch": 0.71, + "learning_rate": 3.8223903565016675e-05, + "loss": 0.6193, + "step": 18416 + }, + { + "epoch": 0.71, + "learning_rate": 3.8223262374967947e-05, + "loss": 0.5744, + "step": 18417 + }, + { + "epoch": 0.71, + "learning_rate": 3.822262118491921e-05, + "loss": 0.5475, + "step": 18418 + }, + { + "epoch": 0.71, + "learning_rate": 3.822197999487048e-05, + "loss": 0.3776, + "step": 18419 + }, + { + "epoch": 0.71, + "learning_rate": 3.822133880482175e-05, + "loss": 0.4782, + "step": 18420 + }, + { + "epoch": 0.71, + "learning_rate": 3.822069761477302e-05, + "loss": 0.5079, + "step": 18421 + }, + { + "epoch": 0.71, + "learning_rate": 3.8220056424724285e-05, + "loss": 0.4291, + "step": 18422 + }, + { + "epoch": 0.71, + "learning_rate": 3.8219415234675557e-05, + "loss": 0.4904, + "step": 18423 + }, + { + "epoch": 0.71, + "learning_rate": 3.821877404462683e-05, + "loss": 0.6036, + "step": 18424 + }, + { + "epoch": 0.71, + "learning_rate": 3.82181328545781e-05, + "loss": 0.6088, + "step": 18425 + }, + { + "epoch": 0.71, + "learning_rate": 3.821749166452937e-05, + "loss": 0.4562, + "step": 18426 + }, + { + "epoch": 0.71, + "learning_rate": 3.821685047448064e-05, + "loss": 0.5344, + "step": 18427 + }, + { + "epoch": 0.71, + "learning_rate": 3.821620928443191e-05, + "loss": 0.4397, + "step": 18428 + }, + { + "epoch": 0.71, + "learning_rate": 3.821556809438317e-05, + "loss": 0.5597, + "step": 18429 + }, + { + "epoch": 0.71, + "learning_rate": 3.8214926904334445e-05, + "loss": 0.5225, + "step": 18430 + }, + { + "epoch": 0.71, + "learning_rate": 3.821428571428572e-05, + "loss": 0.5257, + "step": 18431 + }, + { + "epoch": 0.71, + "learning_rate": 3.821364452423699e-05, + "loss": 0.4146, + "step": 18432 + }, + { + "epoch": 0.71, + "learning_rate": 3.821300333418826e-05, + "loss": 0.4971, + "step": 18433 + }, + { + "epoch": 0.71, + "learning_rate": 3.8212362144139525e-05, + "loss": 0.4513, + "step": 18434 + }, + { + "epoch": 0.71, + "learning_rate": 3.82117209540908e-05, + "loss": 0.5277, + "step": 18435 + }, + { + "epoch": 0.71, + "learning_rate": 3.821107976404206e-05, + "loss": 0.5493, + "step": 18436 + }, + { + "epoch": 0.71, + "learning_rate": 3.8210438573993333e-05, + "loss": 0.6237, + "step": 18437 + }, + { + "epoch": 0.71, + "learning_rate": 3.82097973839446e-05, + "loss": 0.4208, + "step": 18438 + }, + { + "epoch": 0.71, + "learning_rate": 3.820915619389587e-05, + "loss": 0.5338, + "step": 18439 + }, + { + "epoch": 0.71, + "learning_rate": 3.820851500384714e-05, + "loss": 0.4695, + "step": 18440 + }, + { + "epoch": 0.71, + "learning_rate": 3.8207873813798414e-05, + "loss": 0.5342, + "step": 18441 + }, + { + "epoch": 0.71, + "learning_rate": 3.8207232623749685e-05, + "loss": 0.5411, + "step": 18442 + }, + { + "epoch": 0.71, + "learning_rate": 3.820659143370095e-05, + "loss": 0.4836, + "step": 18443 + }, + { + "epoch": 0.71, + "learning_rate": 3.820595024365222e-05, + "loss": 0.423, + "step": 18444 + }, + { + "epoch": 0.71, + "learning_rate": 3.820530905360349e-05, + "loss": 0.5857, + "step": 18445 + }, + { + "epoch": 0.71, + "learning_rate": 3.820466786355476e-05, + "loss": 0.5771, + "step": 18446 + }, + { + "epoch": 0.71, + "learning_rate": 3.820402667350603e-05, + "loss": 0.4462, + "step": 18447 + }, + { + "epoch": 0.71, + "learning_rate": 3.82033854834573e-05, + "loss": 0.5349, + "step": 18448 + }, + { + "epoch": 0.71, + "learning_rate": 3.820274429340857e-05, + "loss": 0.3963, + "step": 18449 + }, + { + "epoch": 0.71, + "learning_rate": 3.820210310335984e-05, + "loss": 0.4368, + "step": 18450 + }, + { + "epoch": 0.71, + "learning_rate": 3.820146191331111e-05, + "loss": 0.4876, + "step": 18451 + }, + { + "epoch": 0.71, + "learning_rate": 3.8200820723262375e-05, + "loss": 0.5201, + "step": 18452 + }, + { + "epoch": 0.71, + "learning_rate": 3.820017953321365e-05, + "loss": 0.5278, + "step": 18453 + }, + { + "epoch": 0.71, + "learning_rate": 3.819953834316491e-05, + "loss": 0.4748, + "step": 18454 + }, + { + "epoch": 0.71, + "learning_rate": 3.8198897153116184e-05, + "loss": 0.3847, + "step": 18455 + }, + { + "epoch": 0.71, + "learning_rate": 3.8198255963067455e-05, + "loss": 0.475, + "step": 18456 + }, + { + "epoch": 0.71, + "learning_rate": 3.819761477301873e-05, + "loss": 0.5249, + "step": 18457 + }, + { + "epoch": 0.71, + "learning_rate": 3.819697358296999e-05, + "loss": 0.6197, + "step": 18458 + }, + { + "epoch": 0.71, + "learning_rate": 3.8196332392921264e-05, + "loss": 0.5618, + "step": 18459 + }, + { + "epoch": 0.71, + "learning_rate": 3.8195691202872536e-05, + "loss": 0.5225, + "step": 18460 + }, + { + "epoch": 0.71, + "learning_rate": 3.81950500128238e-05, + "loss": 0.5842, + "step": 18461 + }, + { + "epoch": 0.71, + "learning_rate": 3.819440882277507e-05, + "loss": 0.5257, + "step": 18462 + }, + { + "epoch": 0.71, + "learning_rate": 3.819376763272634e-05, + "loss": 0.5529, + "step": 18463 + }, + { + "epoch": 0.71, + "learning_rate": 3.8193126442677616e-05, + "loss": 0.4379, + "step": 18464 + }, + { + "epoch": 0.71, + "learning_rate": 3.819248525262888e-05, + "loss": 0.4895, + "step": 18465 + }, + { + "epoch": 0.71, + "learning_rate": 3.819184406258015e-05, + "loss": 0.4885, + "step": 18466 + }, + { + "epoch": 0.71, + "learning_rate": 3.819120287253142e-05, + "loss": 0.5705, + "step": 18467 + }, + { + "epoch": 0.71, + "learning_rate": 3.819056168248269e-05, + "loss": 0.6663, + "step": 18468 + }, + { + "epoch": 0.71, + "learning_rate": 3.818992049243396e-05, + "loss": 0.528, + "step": 18469 + }, + { + "epoch": 0.71, + "learning_rate": 3.8189279302385226e-05, + "loss": 0.606, + "step": 18470 + }, + { + "epoch": 0.71, + "learning_rate": 3.81886381123365e-05, + "loss": 0.4174, + "step": 18471 + }, + { + "epoch": 0.71, + "learning_rate": 3.818799692228777e-05, + "loss": 0.492, + "step": 18472 + }, + { + "epoch": 0.71, + "learning_rate": 3.818735573223904e-05, + "loss": 0.5547, + "step": 18473 + }, + { + "epoch": 0.71, + "learning_rate": 3.8186714542190306e-05, + "loss": 0.5049, + "step": 18474 + }, + { + "epoch": 0.71, + "learning_rate": 3.818607335214158e-05, + "loss": 0.5052, + "step": 18475 + }, + { + "epoch": 0.71, + "learning_rate": 3.818543216209284e-05, + "loss": 0.4459, + "step": 18476 + }, + { + "epoch": 0.71, + "learning_rate": 3.8184790972044114e-05, + "loss": 0.4489, + "step": 18477 + }, + { + "epoch": 0.71, + "learning_rate": 3.8184149781995386e-05, + "loss": 0.5569, + "step": 18478 + }, + { + "epoch": 0.71, + "learning_rate": 3.818350859194665e-05, + "loss": 0.4963, + "step": 18479 + }, + { + "epoch": 0.71, + "learning_rate": 3.818286740189793e-05, + "loss": 0.4328, + "step": 18480 + }, + { + "epoch": 0.71, + "learning_rate": 3.8182226211849194e-05, + "loss": 0.4454, + "step": 18481 + }, + { + "epoch": 0.71, + "learning_rate": 3.8181585021800466e-05, + "loss": 0.4665, + "step": 18482 + }, + { + "epoch": 0.71, + "learning_rate": 3.818094383175173e-05, + "loss": 0.478, + "step": 18483 + }, + { + "epoch": 0.71, + "learning_rate": 3.8180302641703e-05, + "loss": 0.6564, + "step": 18484 + }, + { + "epoch": 0.71, + "learning_rate": 3.817966145165427e-05, + "loss": 0.4245, + "step": 18485 + }, + { + "epoch": 0.71, + "learning_rate": 3.817902026160554e-05, + "loss": 0.536, + "step": 18486 + }, + { + "epoch": 0.71, + "learning_rate": 3.817837907155681e-05, + "loss": 0.6911, + "step": 18487 + }, + { + "epoch": 0.71, + "learning_rate": 3.817773788150808e-05, + "loss": 0.421, + "step": 18488 + }, + { + "epoch": 0.71, + "learning_rate": 3.8177096691459354e-05, + "loss": 0.5456, + "step": 18489 + }, + { + "epoch": 0.71, + "learning_rate": 3.817645550141062e-05, + "loss": 0.4497, + "step": 18490 + }, + { + "epoch": 0.71, + "learning_rate": 3.817581431136189e-05, + "loss": 0.4814, + "step": 18491 + }, + { + "epoch": 0.71, + "learning_rate": 3.8175173121313156e-05, + "loss": 0.48, + "step": 18492 + }, + { + "epoch": 0.71, + "learning_rate": 3.817453193126443e-05, + "loss": 0.5535, + "step": 18493 + }, + { + "epoch": 0.71, + "learning_rate": 3.817389074121569e-05, + "loss": 0.4992, + "step": 18494 + }, + { + "epoch": 0.71, + "learning_rate": 3.8173249551166964e-05, + "loss": 0.5093, + "step": 18495 + }, + { + "epoch": 0.71, + "learning_rate": 3.817260836111824e-05, + "loss": 0.4223, + "step": 18496 + }, + { + "epoch": 0.71, + "learning_rate": 3.817196717106951e-05, + "loss": 0.5782, + "step": 18497 + }, + { + "epoch": 0.71, + "learning_rate": 3.817132598102078e-05, + "loss": 0.5328, + "step": 18498 + }, + { + "epoch": 0.71, + "learning_rate": 3.8170684790972044e-05, + "loss": 0.4403, + "step": 18499 + }, + { + "epoch": 0.71, + "learning_rate": 3.8170043600923316e-05, + "loss": 0.5635, + "step": 18500 + }, + { + "epoch": 0.71, + "learning_rate": 3.816940241087458e-05, + "loss": 0.4694, + "step": 18501 + }, + { + "epoch": 0.71, + "learning_rate": 3.816876122082585e-05, + "loss": 0.5189, + "step": 18502 + }, + { + "epoch": 0.71, + "learning_rate": 3.8168120030777124e-05, + "loss": 0.5665, + "step": 18503 + }, + { + "epoch": 0.71, + "learning_rate": 3.8167478840728396e-05, + "loss": 0.5145, + "step": 18504 + }, + { + "epoch": 0.71, + "learning_rate": 3.816683765067967e-05, + "loss": 0.7517, + "step": 18505 + }, + { + "epoch": 0.71, + "learning_rate": 3.816619646063093e-05, + "loss": 0.5539, + "step": 18506 + }, + { + "epoch": 0.71, + "learning_rate": 3.8165555270582205e-05, + "loss": 0.5952, + "step": 18507 + }, + { + "epoch": 0.71, + "learning_rate": 3.816491408053347e-05, + "loss": 0.5498, + "step": 18508 + }, + { + "epoch": 0.71, + "learning_rate": 3.816427289048474e-05, + "loss": 0.4986, + "step": 18509 + }, + { + "epoch": 0.71, + "learning_rate": 3.8163631700436006e-05, + "loss": 0.5227, + "step": 18510 + }, + { + "epoch": 0.71, + "learning_rate": 3.816299051038728e-05, + "loss": 0.6048, + "step": 18511 + }, + { + "epoch": 0.71, + "learning_rate": 3.816234932033855e-05, + "loss": 0.3732, + "step": 18512 + }, + { + "epoch": 0.71, + "learning_rate": 3.816170813028982e-05, + "loss": 0.5406, + "step": 18513 + }, + { + "epoch": 0.71, + "learning_rate": 3.816106694024109e-05, + "loss": 0.4905, + "step": 18514 + }, + { + "epoch": 0.71, + "learning_rate": 3.816042575019236e-05, + "loss": 0.4109, + "step": 18515 + }, + { + "epoch": 0.71, + "learning_rate": 3.815978456014363e-05, + "loss": 0.4016, + "step": 18516 + }, + { + "epoch": 0.71, + "learning_rate": 3.8159143370094895e-05, + "loss": 0.4763, + "step": 18517 + }, + { + "epoch": 0.71, + "learning_rate": 3.8158502180046166e-05, + "loss": 0.4403, + "step": 18518 + }, + { + "epoch": 0.71, + "learning_rate": 3.815786098999744e-05, + "loss": 0.423, + "step": 18519 + }, + { + "epoch": 0.71, + "learning_rate": 3.815721979994871e-05, + "loss": 0.6258, + "step": 18520 + }, + { + "epoch": 0.71, + "learning_rate": 3.815657860989998e-05, + "loss": 0.4017, + "step": 18521 + }, + { + "epoch": 0.71, + "learning_rate": 3.8155937419851246e-05, + "loss": 0.6296, + "step": 18522 + }, + { + "epoch": 0.71, + "learning_rate": 3.815529622980252e-05, + "loss": 0.4644, + "step": 18523 + }, + { + "epoch": 0.71, + "learning_rate": 3.815465503975378e-05, + "loss": 0.4982, + "step": 18524 + }, + { + "epoch": 0.71, + "learning_rate": 3.8154013849705055e-05, + "loss": 0.4793, + "step": 18525 + }, + { + "epoch": 0.71, + "learning_rate": 3.815337265965632e-05, + "loss": 0.4859, + "step": 18526 + }, + { + "epoch": 0.71, + "learning_rate": 3.815273146960759e-05, + "loss": 0.5648, + "step": 18527 + }, + { + "epoch": 0.71, + "learning_rate": 3.815209027955886e-05, + "loss": 0.6346, + "step": 18528 + }, + { + "epoch": 0.71, + "learning_rate": 3.8151449089510135e-05, + "loss": 0.6464, + "step": 18529 + }, + { + "epoch": 0.71, + "learning_rate": 3.815080789946141e-05, + "loss": 0.4464, + "step": 18530 + }, + { + "epoch": 0.71, + "learning_rate": 3.815016670941267e-05, + "loss": 0.4779, + "step": 18531 + }, + { + "epoch": 0.71, + "learning_rate": 3.814952551936394e-05, + "loss": 0.5347, + "step": 18532 + }, + { + "epoch": 0.71, + "learning_rate": 3.814888432931521e-05, + "loss": 0.5321, + "step": 18533 + }, + { + "epoch": 0.71, + "learning_rate": 3.814824313926648e-05, + "loss": 0.4752, + "step": 18534 + }, + { + "epoch": 0.71, + "learning_rate": 3.814760194921775e-05, + "loss": 0.3789, + "step": 18535 + }, + { + "epoch": 0.71, + "learning_rate": 3.814696075916902e-05, + "loss": 0.5785, + "step": 18536 + }, + { + "epoch": 0.71, + "learning_rate": 3.814631956912029e-05, + "loss": 0.5087, + "step": 18537 + }, + { + "epoch": 0.71, + "learning_rate": 3.814567837907156e-05, + "loss": 0.5356, + "step": 18538 + }, + { + "epoch": 0.71, + "learning_rate": 3.814503718902283e-05, + "loss": 0.5831, + "step": 18539 + }, + { + "epoch": 0.71, + "learning_rate": 3.81443959989741e-05, + "loss": 0.5709, + "step": 18540 + }, + { + "epoch": 0.71, + "learning_rate": 3.814375480892537e-05, + "loss": 0.4786, + "step": 18541 + }, + { + "epoch": 0.71, + "learning_rate": 3.814311361887663e-05, + "loss": 0.5151, + "step": 18542 + }, + { + "epoch": 0.71, + "learning_rate": 3.8142472428827905e-05, + "loss": 0.5386, + "step": 18543 + }, + { + "epoch": 0.71, + "learning_rate": 3.814183123877918e-05, + "loss": 0.4869, + "step": 18544 + }, + { + "epoch": 0.71, + "learning_rate": 3.814119004873045e-05, + "loss": 0.538, + "step": 18545 + }, + { + "epoch": 0.71, + "learning_rate": 3.8140548858681713e-05, + "loss": 0.4237, + "step": 18546 + }, + { + "epoch": 0.71, + "learning_rate": 3.8139907668632985e-05, + "loss": 0.5163, + "step": 18547 + }, + { + "epoch": 0.71, + "learning_rate": 3.813926647858426e-05, + "loss": 0.5633, + "step": 18548 + }, + { + "epoch": 0.71, + "learning_rate": 3.813862528853552e-05, + "loss": 0.5531, + "step": 18549 + }, + { + "epoch": 0.71, + "learning_rate": 3.8137984098486794e-05, + "loss": 0.4756, + "step": 18550 + }, + { + "epoch": 0.71, + "learning_rate": 3.813734290843806e-05, + "loss": 0.4975, + "step": 18551 + }, + { + "epoch": 0.71, + "learning_rate": 3.813670171838934e-05, + "loss": 0.5986, + "step": 18552 + }, + { + "epoch": 0.71, + "learning_rate": 3.81360605283406e-05, + "loss": 0.4586, + "step": 18553 + }, + { + "epoch": 0.71, + "learning_rate": 3.8135419338291874e-05, + "loss": 0.5081, + "step": 18554 + }, + { + "epoch": 0.71, + "learning_rate": 3.813477814824314e-05, + "loss": 0.6209, + "step": 18555 + }, + { + "epoch": 0.71, + "learning_rate": 3.813413695819441e-05, + "loss": 0.6451, + "step": 18556 + }, + { + "epoch": 0.71, + "learning_rate": 3.813349576814568e-05, + "loss": 0.5955, + "step": 18557 + }, + { + "epoch": 0.71, + "learning_rate": 3.813285457809695e-05, + "loss": 0.4849, + "step": 18558 + }, + { + "epoch": 0.71, + "learning_rate": 3.813221338804822e-05, + "loss": 0.542, + "step": 18559 + }, + { + "epoch": 0.71, + "learning_rate": 3.813157219799949e-05, + "loss": 0.4995, + "step": 18560 + }, + { + "epoch": 0.71, + "learning_rate": 3.813093100795076e-05, + "loss": 0.4138, + "step": 18561 + }, + { + "epoch": 0.71, + "learning_rate": 3.813028981790203e-05, + "loss": 0.489, + "step": 18562 + }, + { + "epoch": 0.71, + "learning_rate": 3.81296486278533e-05, + "loss": 0.4496, + "step": 18563 + }, + { + "epoch": 0.71, + "learning_rate": 3.8129007437804564e-05, + "loss": 0.5358, + "step": 18564 + }, + { + "epoch": 0.71, + "learning_rate": 3.8128366247755835e-05, + "loss": 0.5421, + "step": 18565 + }, + { + "epoch": 0.71, + "learning_rate": 3.812772505770711e-05, + "loss": 0.4728, + "step": 18566 + }, + { + "epoch": 0.71, + "learning_rate": 3.812708386765837e-05, + "loss": 0.4197, + "step": 18567 + }, + { + "epoch": 0.71, + "learning_rate": 3.812644267760965e-05, + "loss": 0.5378, + "step": 18568 + }, + { + "epoch": 0.71, + "learning_rate": 3.8125801487560915e-05, + "loss": 0.4381, + "step": 18569 + }, + { + "epoch": 0.71, + "learning_rate": 3.812516029751219e-05, + "loss": 0.5561, + "step": 18570 + }, + { + "epoch": 0.71, + "learning_rate": 3.812451910746345e-05, + "loss": 0.4066, + "step": 18571 + }, + { + "epoch": 0.71, + "learning_rate": 3.8123877917414724e-05, + "loss": 0.5013, + "step": 18572 + }, + { + "epoch": 0.71, + "learning_rate": 3.812323672736599e-05, + "loss": 0.508, + "step": 18573 + }, + { + "epoch": 0.71, + "learning_rate": 3.812259553731726e-05, + "loss": 0.4523, + "step": 18574 + }, + { + "epoch": 0.71, + "learning_rate": 3.812195434726853e-05, + "loss": 0.4917, + "step": 18575 + }, + { + "epoch": 0.71, + "learning_rate": 3.8121313157219804e-05, + "loss": 0.5476, + "step": 18576 + }, + { + "epoch": 0.71, + "learning_rate": 3.8120671967171076e-05, + "loss": 0.5267, + "step": 18577 + }, + { + "epoch": 0.71, + "learning_rate": 3.812003077712234e-05, + "loss": 0.4613, + "step": 18578 + }, + { + "epoch": 0.71, + "learning_rate": 3.811938958707361e-05, + "loss": 0.5199, + "step": 18579 + }, + { + "epoch": 0.71, + "learning_rate": 3.811874839702488e-05, + "loss": 0.5759, + "step": 18580 + }, + { + "epoch": 0.71, + "learning_rate": 3.811810720697615e-05, + "loss": 0.5357, + "step": 18581 + }, + { + "epoch": 0.71, + "learning_rate": 3.8117466016927414e-05, + "loss": 0.5185, + "step": 18582 + }, + { + "epoch": 0.71, + "learning_rate": 3.8116824826878686e-05, + "loss": 0.6213, + "step": 18583 + }, + { + "epoch": 0.71, + "learning_rate": 3.811618363682996e-05, + "loss": 0.5344, + "step": 18584 + }, + { + "epoch": 0.71, + "learning_rate": 3.811554244678123e-05, + "loss": 0.6488, + "step": 18585 + }, + { + "epoch": 0.71, + "learning_rate": 3.81149012567325e-05, + "loss": 0.507, + "step": 18586 + }, + { + "epoch": 0.71, + "learning_rate": 3.8114260066683766e-05, + "loss": 0.4488, + "step": 18587 + }, + { + "epoch": 0.71, + "learning_rate": 3.811361887663504e-05, + "loss": 0.6832, + "step": 18588 + }, + { + "epoch": 0.71, + "learning_rate": 3.81129776865863e-05, + "loss": 0.6617, + "step": 18589 + }, + { + "epoch": 0.71, + "learning_rate": 3.8112336496537574e-05, + "loss": 0.6309, + "step": 18590 + }, + { + "epoch": 0.71, + "learning_rate": 3.8111695306488846e-05, + "loss": 0.567, + "step": 18591 + }, + { + "epoch": 0.71, + "learning_rate": 3.811105411644012e-05, + "loss": 0.4245, + "step": 18592 + }, + { + "epoch": 0.71, + "learning_rate": 3.811041292639139e-05, + "loss": 0.4486, + "step": 18593 + }, + { + "epoch": 0.71, + "learning_rate": 3.8109771736342654e-05, + "loss": 0.4353, + "step": 18594 + }, + { + "epoch": 0.71, + "learning_rate": 3.8109130546293926e-05, + "loss": 0.4774, + "step": 18595 + }, + { + "epoch": 0.71, + "learning_rate": 3.810848935624519e-05, + "loss": 0.5516, + "step": 18596 + }, + { + "epoch": 0.71, + "learning_rate": 3.810784816619646e-05, + "loss": 0.434, + "step": 18597 + }, + { + "epoch": 0.72, + "learning_rate": 3.810720697614773e-05, + "loss": 0.5027, + "step": 18598 + }, + { + "epoch": 0.72, + "learning_rate": 3.8106565786099e-05, + "loss": 0.5815, + "step": 18599 + }, + { + "epoch": 0.72, + "learning_rate": 3.810592459605027e-05, + "loss": 0.5542, + "step": 18600 + }, + { + "epoch": 0.72, + "learning_rate": 3.810528340600154e-05, + "loss": 0.5255, + "step": 18601 + }, + { + "epoch": 0.72, + "learning_rate": 3.8104642215952814e-05, + "loss": 0.5614, + "step": 18602 + }, + { + "epoch": 0.72, + "learning_rate": 3.810400102590408e-05, + "loss": 0.5178, + "step": 18603 + }, + { + "epoch": 0.72, + "learning_rate": 3.810335983585535e-05, + "loss": 0.6012, + "step": 18604 + }, + { + "epoch": 0.72, + "learning_rate": 3.8102718645806616e-05, + "loss": 0.5447, + "step": 18605 + }, + { + "epoch": 0.72, + "learning_rate": 3.810207745575789e-05, + "loss": 0.6133, + "step": 18606 + }, + { + "epoch": 0.72, + "learning_rate": 3.810143626570916e-05, + "loss": 0.4001, + "step": 18607 + }, + { + "epoch": 0.72, + "learning_rate": 3.810079507566043e-05, + "loss": 0.4444, + "step": 18608 + }, + { + "epoch": 0.72, + "learning_rate": 3.8100153885611696e-05, + "loss": 0.4829, + "step": 18609 + }, + { + "epoch": 0.72, + "learning_rate": 3.809951269556297e-05, + "loss": 0.5527, + "step": 18610 + }, + { + "epoch": 0.72, + "learning_rate": 3.809887150551424e-05, + "loss": 0.5139, + "step": 18611 + }, + { + "epoch": 0.72, + "learning_rate": 3.8098230315465504e-05, + "loss": 0.4888, + "step": 18612 + }, + { + "epoch": 0.72, + "learning_rate": 3.8097589125416776e-05, + "loss": 0.4649, + "step": 18613 + }, + { + "epoch": 0.72, + "learning_rate": 3.809694793536804e-05, + "loss": 0.5813, + "step": 18614 + }, + { + "epoch": 0.72, + "learning_rate": 3.809630674531931e-05, + "loss": 0.5556, + "step": 18615 + }, + { + "epoch": 0.72, + "learning_rate": 3.8095665555270585e-05, + "loss": 0.4999, + "step": 18616 + }, + { + "epoch": 0.72, + "learning_rate": 3.8095024365221856e-05, + "loss": 0.4299, + "step": 18617 + }, + { + "epoch": 0.72, + "learning_rate": 3.809438317517312e-05, + "loss": 0.505, + "step": 18618 + }, + { + "epoch": 0.72, + "learning_rate": 3.809374198512439e-05, + "loss": 0.4724, + "step": 18619 + }, + { + "epoch": 0.72, + "learning_rate": 3.8093100795075665e-05, + "loss": 0.5472, + "step": 18620 + }, + { + "epoch": 0.72, + "learning_rate": 3.809245960502693e-05, + "loss": 0.4578, + "step": 18621 + }, + { + "epoch": 0.72, + "learning_rate": 3.80918184149782e-05, + "loss": 0.5937, + "step": 18622 + }, + { + "epoch": 0.72, + "learning_rate": 3.809117722492947e-05, + "loss": 0.4646, + "step": 18623 + }, + { + "epoch": 0.72, + "learning_rate": 3.8090536034880745e-05, + "loss": 0.5024, + "step": 18624 + }, + { + "epoch": 0.72, + "learning_rate": 3.808989484483201e-05, + "loss": 0.5417, + "step": 18625 + }, + { + "epoch": 0.72, + "learning_rate": 3.808925365478328e-05, + "loss": 0.5357, + "step": 18626 + }, + { + "epoch": 0.72, + "learning_rate": 3.8088612464734546e-05, + "loss": 0.64, + "step": 18627 + }, + { + "epoch": 0.72, + "learning_rate": 3.808797127468582e-05, + "loss": 0.4687, + "step": 18628 + }, + { + "epoch": 0.72, + "learning_rate": 3.808733008463709e-05, + "loss": 0.5048, + "step": 18629 + }, + { + "epoch": 0.72, + "learning_rate": 3.8086688894588355e-05, + "loss": 0.5024, + "step": 18630 + }, + { + "epoch": 0.72, + "learning_rate": 3.8086047704539626e-05, + "loss": 0.5456, + "step": 18631 + }, + { + "epoch": 0.72, + "learning_rate": 3.80854065144909e-05, + "loss": 0.7135, + "step": 18632 + }, + { + "epoch": 0.72, + "learning_rate": 3.808476532444217e-05, + "loss": 0.3701, + "step": 18633 + }, + { + "epoch": 0.72, + "learning_rate": 3.8084124134393435e-05, + "loss": 0.5728, + "step": 18634 + }, + { + "epoch": 0.72, + "learning_rate": 3.8083482944344706e-05, + "loss": 0.5165, + "step": 18635 + }, + { + "epoch": 0.72, + "learning_rate": 3.808284175429597e-05, + "loss": 0.5586, + "step": 18636 + }, + { + "epoch": 0.72, + "learning_rate": 3.808220056424724e-05, + "loss": 0.6193, + "step": 18637 + }, + { + "epoch": 0.72, + "learning_rate": 3.8081559374198515e-05, + "loss": 0.5717, + "step": 18638 + }, + { + "epoch": 0.72, + "learning_rate": 3.808091818414978e-05, + "loss": 0.533, + "step": 18639 + }, + { + "epoch": 0.72, + "learning_rate": 3.808027699410106e-05, + "loss": 0.638, + "step": 18640 + }, + { + "epoch": 0.72, + "learning_rate": 3.807963580405232e-05, + "loss": 0.5414, + "step": 18641 + }, + { + "epoch": 0.72, + "learning_rate": 3.8078994614003595e-05, + "loss": 0.5471, + "step": 18642 + }, + { + "epoch": 0.72, + "learning_rate": 3.807835342395486e-05, + "loss": 0.4358, + "step": 18643 + }, + { + "epoch": 0.72, + "learning_rate": 3.807771223390613e-05, + "loss": 0.4986, + "step": 18644 + }, + { + "epoch": 0.72, + "learning_rate": 3.8077071043857397e-05, + "loss": 0.5509, + "step": 18645 + }, + { + "epoch": 0.72, + "learning_rate": 3.807642985380867e-05, + "loss": 0.577, + "step": 18646 + }, + { + "epoch": 0.72, + "learning_rate": 3.807578866375994e-05, + "loss": 0.5277, + "step": 18647 + }, + { + "epoch": 0.72, + "learning_rate": 3.807514747371121e-05, + "loss": 0.568, + "step": 18648 + }, + { + "epoch": 0.72, + "learning_rate": 3.8074506283662483e-05, + "loss": 0.6164, + "step": 18649 + }, + { + "epoch": 0.72, + "learning_rate": 3.807386509361375e-05, + "loss": 0.5023, + "step": 18650 + }, + { + "epoch": 0.72, + "learning_rate": 3.807322390356502e-05, + "loss": 0.5015, + "step": 18651 + }, + { + "epoch": 0.72, + "learning_rate": 3.8072582713516285e-05, + "loss": 0.608, + "step": 18652 + }, + { + "epoch": 0.72, + "learning_rate": 3.807194152346756e-05, + "loss": 0.4917, + "step": 18653 + }, + { + "epoch": 0.72, + "learning_rate": 3.807130033341882e-05, + "loss": 0.5266, + "step": 18654 + }, + { + "epoch": 0.72, + "learning_rate": 3.807065914337009e-05, + "loss": 0.5232, + "step": 18655 + }, + { + "epoch": 0.72, + "learning_rate": 3.807001795332137e-05, + "loss": 0.4543, + "step": 18656 + }, + { + "epoch": 0.72, + "learning_rate": 3.806937676327264e-05, + "loss": 0.4779, + "step": 18657 + }, + { + "epoch": 0.72, + "learning_rate": 3.806873557322391e-05, + "loss": 0.4632, + "step": 18658 + }, + { + "epoch": 0.72, + "learning_rate": 3.8068094383175173e-05, + "loss": 0.5032, + "step": 18659 + }, + { + "epoch": 0.72, + "learning_rate": 3.8067453193126445e-05, + "loss": 0.5075, + "step": 18660 + }, + { + "epoch": 0.72, + "learning_rate": 3.806681200307771e-05, + "loss": 0.4823, + "step": 18661 + }, + { + "epoch": 0.72, + "learning_rate": 3.806617081302898e-05, + "loss": 0.5617, + "step": 18662 + }, + { + "epoch": 0.72, + "learning_rate": 3.8065529622980254e-05, + "loss": 0.4889, + "step": 18663 + }, + { + "epoch": 0.72, + "learning_rate": 3.8064888432931525e-05, + "loss": 0.5637, + "step": 18664 + }, + { + "epoch": 0.72, + "learning_rate": 3.80642472428828e-05, + "loss": 0.5662, + "step": 18665 + }, + { + "epoch": 0.72, + "learning_rate": 3.806360605283406e-05, + "loss": 0.5442, + "step": 18666 + }, + { + "epoch": 0.72, + "learning_rate": 3.8062964862785334e-05, + "loss": 0.4902, + "step": 18667 + }, + { + "epoch": 0.72, + "learning_rate": 3.80623236727366e-05, + "loss": 0.5939, + "step": 18668 + }, + { + "epoch": 0.72, + "learning_rate": 3.806168248268787e-05, + "loss": 0.5579, + "step": 18669 + }, + { + "epoch": 0.72, + "learning_rate": 3.8061041292639135e-05, + "loss": 0.5303, + "step": 18670 + }, + { + "epoch": 0.72, + "learning_rate": 3.806040010259041e-05, + "loss": 0.3896, + "step": 18671 + }, + { + "epoch": 0.72, + "learning_rate": 3.805975891254168e-05, + "loss": 0.4987, + "step": 18672 + }, + { + "epoch": 0.72, + "learning_rate": 3.805911772249295e-05, + "loss": 0.4735, + "step": 18673 + }, + { + "epoch": 0.72, + "learning_rate": 3.805847653244422e-05, + "loss": 0.5595, + "step": 18674 + }, + { + "epoch": 0.72, + "learning_rate": 3.805783534239549e-05, + "loss": 0.5724, + "step": 18675 + }, + { + "epoch": 0.72, + "learning_rate": 3.805719415234676e-05, + "loss": 0.46, + "step": 18676 + }, + { + "epoch": 0.72, + "learning_rate": 3.8056552962298024e-05, + "loss": 0.583, + "step": 18677 + }, + { + "epoch": 0.72, + "learning_rate": 3.8055911772249295e-05, + "loss": 0.5075, + "step": 18678 + }, + { + "epoch": 0.72, + "learning_rate": 3.805527058220057e-05, + "loss": 0.5132, + "step": 18679 + }, + { + "epoch": 0.72, + "learning_rate": 3.805462939215184e-05, + "loss": 0.5672, + "step": 18680 + }, + { + "epoch": 0.72, + "learning_rate": 3.8053988202103104e-05, + "loss": 0.4294, + "step": 18681 + }, + { + "epoch": 0.72, + "learning_rate": 3.8053347012054376e-05, + "loss": 0.4749, + "step": 18682 + }, + { + "epoch": 0.72, + "learning_rate": 3.805270582200565e-05, + "loss": 0.5266, + "step": 18683 + }, + { + "epoch": 0.72, + "learning_rate": 3.805206463195691e-05, + "loss": 0.5827, + "step": 18684 + }, + { + "epoch": 0.72, + "learning_rate": 3.8051423441908184e-05, + "loss": 0.6399, + "step": 18685 + }, + { + "epoch": 0.72, + "learning_rate": 3.805078225185945e-05, + "loss": 0.4754, + "step": 18686 + }, + { + "epoch": 0.72, + "learning_rate": 3.805014106181072e-05, + "loss": 0.4731, + "step": 18687 + }, + { + "epoch": 0.72, + "learning_rate": 3.804949987176199e-05, + "loss": 0.5325, + "step": 18688 + }, + { + "epoch": 0.72, + "learning_rate": 3.8048858681713264e-05, + "loss": 0.4102, + "step": 18689 + }, + { + "epoch": 0.72, + "learning_rate": 3.804821749166453e-05, + "loss": 0.5578, + "step": 18690 + }, + { + "epoch": 0.72, + "learning_rate": 3.80475763016158e-05, + "loss": 0.6347, + "step": 18691 + }, + { + "epoch": 0.72, + "learning_rate": 3.804693511156707e-05, + "loss": 0.6201, + "step": 18692 + }, + { + "epoch": 0.72, + "learning_rate": 3.804629392151834e-05, + "loss": 0.4558, + "step": 18693 + }, + { + "epoch": 0.72, + "learning_rate": 3.804565273146961e-05, + "loss": 0.4426, + "step": 18694 + }, + { + "epoch": 0.72, + "learning_rate": 3.804501154142088e-05, + "loss": 0.4881, + "step": 18695 + }, + { + "epoch": 0.72, + "learning_rate": 3.804437035137215e-05, + "loss": 0.4882, + "step": 18696 + }, + { + "epoch": 0.72, + "learning_rate": 3.804372916132342e-05, + "loss": 0.5257, + "step": 18697 + }, + { + "epoch": 0.72, + "learning_rate": 3.804308797127469e-05, + "loss": 0.4614, + "step": 18698 + }, + { + "epoch": 0.72, + "learning_rate": 3.804244678122596e-05, + "loss": 0.4652, + "step": 18699 + }, + { + "epoch": 0.72, + "learning_rate": 3.8041805591177226e-05, + "loss": 0.6429, + "step": 18700 + }, + { + "epoch": 0.72, + "learning_rate": 3.80411644011285e-05, + "loss": 0.5411, + "step": 18701 + }, + { + "epoch": 0.72, + "learning_rate": 3.804052321107976e-05, + "loss": 0.4885, + "step": 18702 + }, + { + "epoch": 0.72, + "learning_rate": 3.8039882021031034e-05, + "loss": 0.6211, + "step": 18703 + }, + { + "epoch": 0.72, + "learning_rate": 3.8039240830982306e-05, + "loss": 0.5091, + "step": 18704 + }, + { + "epoch": 0.72, + "learning_rate": 3.803859964093358e-05, + "loss": 0.481, + "step": 18705 + }, + { + "epoch": 0.72, + "learning_rate": 3.803795845088484e-05, + "loss": 0.6212, + "step": 18706 + }, + { + "epoch": 0.72, + "learning_rate": 3.8037317260836114e-05, + "loss": 0.4871, + "step": 18707 + }, + { + "epoch": 0.72, + "learning_rate": 3.8036676070787386e-05, + "loss": 0.5769, + "step": 18708 + }, + { + "epoch": 0.72, + "learning_rate": 3.803603488073865e-05, + "loss": 0.5421, + "step": 18709 + }, + { + "epoch": 0.72, + "learning_rate": 3.803539369068992e-05, + "loss": 0.4864, + "step": 18710 + }, + { + "epoch": 0.72, + "learning_rate": 3.803475250064119e-05, + "loss": 0.4566, + "step": 18711 + }, + { + "epoch": 0.72, + "learning_rate": 3.8034111310592466e-05, + "loss": 0.4953, + "step": 18712 + }, + { + "epoch": 0.72, + "learning_rate": 3.803347012054373e-05, + "loss": 0.5006, + "step": 18713 + }, + { + "epoch": 0.72, + "learning_rate": 3.8032828930495e-05, + "loss": 0.533, + "step": 18714 + }, + { + "epoch": 0.72, + "learning_rate": 3.803218774044627e-05, + "loss": 0.5118, + "step": 18715 + }, + { + "epoch": 0.72, + "learning_rate": 3.803154655039754e-05, + "loss": 0.4711, + "step": 18716 + }, + { + "epoch": 0.72, + "learning_rate": 3.803090536034881e-05, + "loss": 0.4893, + "step": 18717 + }, + { + "epoch": 0.72, + "learning_rate": 3.8030264170300076e-05, + "loss": 0.5099, + "step": 18718 + }, + { + "epoch": 0.72, + "learning_rate": 3.802962298025135e-05, + "loss": 0.5223, + "step": 18719 + }, + { + "epoch": 0.72, + "learning_rate": 3.802898179020262e-05, + "loss": 0.5, + "step": 18720 + }, + { + "epoch": 0.72, + "learning_rate": 3.802834060015389e-05, + "loss": 0.4576, + "step": 18721 + }, + { + "epoch": 0.72, + "learning_rate": 3.8027699410105156e-05, + "loss": 0.5356, + "step": 18722 + }, + { + "epoch": 0.72, + "learning_rate": 3.802705822005643e-05, + "loss": 0.4884, + "step": 18723 + }, + { + "epoch": 0.72, + "learning_rate": 3.802641703000769e-05, + "loss": 0.6521, + "step": 18724 + }, + { + "epoch": 0.72, + "learning_rate": 3.8025775839958964e-05, + "loss": 0.4715, + "step": 18725 + }, + { + "epoch": 0.72, + "learning_rate": 3.8025134649910236e-05, + "loss": 0.5213, + "step": 18726 + }, + { + "epoch": 0.72, + "learning_rate": 3.80244934598615e-05, + "loss": 0.4761, + "step": 18727 + }, + { + "epoch": 0.72, + "learning_rate": 3.802385226981278e-05, + "loss": 0.511, + "step": 18728 + }, + { + "epoch": 0.72, + "learning_rate": 3.8023211079764045e-05, + "loss": 0.5508, + "step": 18729 + }, + { + "epoch": 0.72, + "learning_rate": 3.8022569889715316e-05, + "loss": 0.5315, + "step": 18730 + }, + { + "epoch": 0.72, + "learning_rate": 3.802192869966658e-05, + "loss": 0.4468, + "step": 18731 + }, + { + "epoch": 0.72, + "learning_rate": 3.802128750961785e-05, + "loss": 0.5499, + "step": 18732 + }, + { + "epoch": 0.72, + "learning_rate": 3.802064631956912e-05, + "loss": 0.4594, + "step": 18733 + }, + { + "epoch": 0.72, + "learning_rate": 3.802000512952039e-05, + "loss": 0.5036, + "step": 18734 + }, + { + "epoch": 0.72, + "learning_rate": 3.801936393947166e-05, + "loss": 0.4544, + "step": 18735 + }, + { + "epoch": 0.72, + "learning_rate": 3.801872274942293e-05, + "loss": 0.4559, + "step": 18736 + }, + { + "epoch": 0.72, + "learning_rate": 3.8018081559374205e-05, + "loss": 0.4659, + "step": 18737 + }, + { + "epoch": 0.72, + "learning_rate": 3.801744036932547e-05, + "loss": 0.6515, + "step": 18738 + }, + { + "epoch": 0.72, + "learning_rate": 3.801679917927674e-05, + "loss": 0.5221, + "step": 18739 + }, + { + "epoch": 0.72, + "learning_rate": 3.8016157989228006e-05, + "loss": 0.5534, + "step": 18740 + }, + { + "epoch": 0.72, + "learning_rate": 3.801551679917928e-05, + "loss": 0.4745, + "step": 18741 + }, + { + "epoch": 0.72, + "learning_rate": 3.801487560913054e-05, + "loss": 0.5221, + "step": 18742 + }, + { + "epoch": 0.72, + "learning_rate": 3.8014234419081815e-05, + "loss": 0.5767, + "step": 18743 + }, + { + "epoch": 0.72, + "learning_rate": 3.801359322903309e-05, + "loss": 0.5107, + "step": 18744 + }, + { + "epoch": 0.72, + "learning_rate": 3.801295203898436e-05, + "loss": 0.5612, + "step": 18745 + }, + { + "epoch": 0.72, + "learning_rate": 3.801231084893563e-05, + "loss": 0.5915, + "step": 18746 + }, + { + "epoch": 0.72, + "learning_rate": 3.8011669658886895e-05, + "loss": 0.4802, + "step": 18747 + }, + { + "epoch": 0.72, + "learning_rate": 3.8011028468838167e-05, + "loss": 0.4986, + "step": 18748 + }, + { + "epoch": 0.72, + "learning_rate": 3.801038727878943e-05, + "loss": 0.5723, + "step": 18749 + }, + { + "epoch": 0.72, + "learning_rate": 3.80097460887407e-05, + "loss": 0.5535, + "step": 18750 + }, + { + "epoch": 0.72, + "learning_rate": 3.8009104898691975e-05, + "loss": 0.4441, + "step": 18751 + }, + { + "epoch": 0.72, + "learning_rate": 3.8008463708643247e-05, + "loss": 0.5394, + "step": 18752 + }, + { + "epoch": 0.72, + "learning_rate": 3.800782251859452e-05, + "loss": 0.4923, + "step": 18753 + }, + { + "epoch": 0.72, + "learning_rate": 3.800718132854578e-05, + "loss": 0.4879, + "step": 18754 + }, + { + "epoch": 0.72, + "learning_rate": 3.8006540138497055e-05, + "loss": 0.5374, + "step": 18755 + }, + { + "epoch": 0.72, + "learning_rate": 3.800589894844832e-05, + "loss": 0.5785, + "step": 18756 + }, + { + "epoch": 0.72, + "learning_rate": 3.800525775839959e-05, + "loss": 0.4924, + "step": 18757 + }, + { + "epoch": 0.72, + "learning_rate": 3.8004616568350857e-05, + "loss": 0.5135, + "step": 18758 + }, + { + "epoch": 0.72, + "learning_rate": 3.800397537830213e-05, + "loss": 0.5322, + "step": 18759 + }, + { + "epoch": 0.72, + "learning_rate": 3.80033341882534e-05, + "loss": 0.5274, + "step": 18760 + }, + { + "epoch": 0.72, + "learning_rate": 3.800269299820467e-05, + "loss": 0.4157, + "step": 18761 + }, + { + "epoch": 0.72, + "learning_rate": 3.8002051808155943e-05, + "loss": 0.4286, + "step": 18762 + }, + { + "epoch": 0.72, + "learning_rate": 3.800141061810721e-05, + "loss": 0.4649, + "step": 18763 + }, + { + "epoch": 0.72, + "learning_rate": 3.800076942805848e-05, + "loss": 0.4779, + "step": 18764 + }, + { + "epoch": 0.72, + "learning_rate": 3.8000128238009745e-05, + "loss": 0.5102, + "step": 18765 + }, + { + "epoch": 0.72, + "learning_rate": 3.799948704796102e-05, + "loss": 0.4668, + "step": 18766 + }, + { + "epoch": 0.72, + "learning_rate": 3.799884585791229e-05, + "loss": 0.4935, + "step": 18767 + }, + { + "epoch": 0.72, + "learning_rate": 3.799820466786356e-05, + "loss": 0.5106, + "step": 18768 + }, + { + "epoch": 0.72, + "learning_rate": 3.7997563477814825e-05, + "loss": 0.5296, + "step": 18769 + }, + { + "epoch": 0.72, + "learning_rate": 3.79969222877661e-05, + "loss": 0.4471, + "step": 18770 + }, + { + "epoch": 0.72, + "learning_rate": 3.799628109771737e-05, + "loss": 0.4883, + "step": 18771 + }, + { + "epoch": 0.72, + "learning_rate": 3.7995639907668634e-05, + "loss": 0.5379, + "step": 18772 + }, + { + "epoch": 0.72, + "learning_rate": 3.7994998717619905e-05, + "loss": 0.5695, + "step": 18773 + }, + { + "epoch": 0.72, + "learning_rate": 3.799435752757117e-05, + "loss": 0.4985, + "step": 18774 + }, + { + "epoch": 0.72, + "learning_rate": 3.799371633752244e-05, + "loss": 0.4922, + "step": 18775 + }, + { + "epoch": 0.72, + "learning_rate": 3.7993075147473714e-05, + "loss": 0.5734, + "step": 18776 + }, + { + "epoch": 0.72, + "learning_rate": 3.7992433957424985e-05, + "loss": 0.4646, + "step": 18777 + }, + { + "epoch": 0.72, + "learning_rate": 3.799179276737625e-05, + "loss": 0.5995, + "step": 18778 + }, + { + "epoch": 0.72, + "learning_rate": 3.799115157732752e-05, + "loss": 0.5083, + "step": 18779 + }, + { + "epoch": 0.72, + "learning_rate": 3.7990510387278794e-05, + "loss": 0.5451, + "step": 18780 + }, + { + "epoch": 0.72, + "learning_rate": 3.798986919723006e-05, + "loss": 0.4626, + "step": 18781 + }, + { + "epoch": 0.72, + "learning_rate": 3.798922800718133e-05, + "loss": 0.4532, + "step": 18782 + }, + { + "epoch": 0.72, + "learning_rate": 3.79885868171326e-05, + "loss": 0.5815, + "step": 18783 + }, + { + "epoch": 0.72, + "learning_rate": 3.7987945627083874e-05, + "loss": 0.4728, + "step": 18784 + }, + { + "epoch": 0.72, + "learning_rate": 3.798730443703514e-05, + "loss": 0.575, + "step": 18785 + }, + { + "epoch": 0.72, + "learning_rate": 3.798666324698641e-05, + "loss": 0.6582, + "step": 18786 + }, + { + "epoch": 0.72, + "learning_rate": 3.7986022056937675e-05, + "loss": 0.5263, + "step": 18787 + }, + { + "epoch": 0.72, + "learning_rate": 3.798538086688895e-05, + "loss": 0.5161, + "step": 18788 + }, + { + "epoch": 0.72, + "learning_rate": 3.798473967684022e-05, + "loss": 0.3985, + "step": 18789 + }, + { + "epoch": 0.72, + "learning_rate": 3.7984098486791484e-05, + "loss": 0.5109, + "step": 18790 + }, + { + "epoch": 0.72, + "learning_rate": 3.7983457296742755e-05, + "loss": 0.5136, + "step": 18791 + }, + { + "epoch": 0.72, + "learning_rate": 3.798281610669403e-05, + "loss": 0.5472, + "step": 18792 + }, + { + "epoch": 0.72, + "learning_rate": 3.79821749166453e-05, + "loss": 0.5423, + "step": 18793 + }, + { + "epoch": 0.72, + "learning_rate": 3.7981533726596564e-05, + "loss": 0.4845, + "step": 18794 + }, + { + "epoch": 0.72, + "learning_rate": 3.7980892536547836e-05, + "loss": 0.4862, + "step": 18795 + }, + { + "epoch": 0.72, + "learning_rate": 3.79802513464991e-05, + "loss": 0.6183, + "step": 18796 + }, + { + "epoch": 0.72, + "learning_rate": 3.797961015645037e-05, + "loss": 0.5073, + "step": 18797 + }, + { + "epoch": 0.72, + "learning_rate": 3.7978968966401644e-05, + "loss": 0.4321, + "step": 18798 + }, + { + "epoch": 0.72, + "learning_rate": 3.797832777635291e-05, + "loss": 0.5503, + "step": 18799 + }, + { + "epoch": 0.72, + "learning_rate": 3.797768658630419e-05, + "loss": 0.4473, + "step": 18800 + }, + { + "epoch": 0.72, + "learning_rate": 3.797704539625545e-05, + "loss": 0.5529, + "step": 18801 + }, + { + "epoch": 0.72, + "learning_rate": 3.7976404206206724e-05, + "loss": 0.4634, + "step": 18802 + }, + { + "epoch": 0.72, + "learning_rate": 3.797576301615799e-05, + "loss": 0.5316, + "step": 18803 + }, + { + "epoch": 0.72, + "learning_rate": 3.797512182610926e-05, + "loss": 0.5391, + "step": 18804 + }, + { + "epoch": 0.72, + "learning_rate": 3.7974480636060526e-05, + "loss": 0.4205, + "step": 18805 + }, + { + "epoch": 0.72, + "learning_rate": 3.79738394460118e-05, + "loss": 0.5666, + "step": 18806 + }, + { + "epoch": 0.72, + "learning_rate": 3.797319825596307e-05, + "loss": 0.5645, + "step": 18807 + }, + { + "epoch": 0.72, + "learning_rate": 3.797255706591434e-05, + "loss": 0.4926, + "step": 18808 + }, + { + "epoch": 0.72, + "learning_rate": 3.797191587586561e-05, + "loss": 0.5067, + "step": 18809 + }, + { + "epoch": 0.72, + "learning_rate": 3.797127468581688e-05, + "loss": 0.5384, + "step": 18810 + }, + { + "epoch": 0.72, + "learning_rate": 3.797063349576815e-05, + "loss": 0.4348, + "step": 18811 + }, + { + "epoch": 0.72, + "learning_rate": 3.7969992305719414e-05, + "loss": 0.5757, + "step": 18812 + }, + { + "epoch": 0.72, + "learning_rate": 3.7969351115670686e-05, + "loss": 0.5505, + "step": 18813 + }, + { + "epoch": 0.72, + "learning_rate": 3.796870992562195e-05, + "loss": 0.4564, + "step": 18814 + }, + { + "epoch": 0.72, + "learning_rate": 3.796806873557322e-05, + "loss": 0.44, + "step": 18815 + }, + { + "epoch": 0.72, + "learning_rate": 3.79674275455245e-05, + "loss": 0.5725, + "step": 18816 + }, + { + "epoch": 0.72, + "learning_rate": 3.7966786355475766e-05, + "loss": 0.3732, + "step": 18817 + }, + { + "epoch": 0.72, + "learning_rate": 3.796614516542704e-05, + "loss": 0.5141, + "step": 18818 + }, + { + "epoch": 0.72, + "learning_rate": 3.79655039753783e-05, + "loss": 0.4921, + "step": 18819 + }, + { + "epoch": 0.72, + "learning_rate": 3.7964862785329574e-05, + "loss": 0.5117, + "step": 18820 + }, + { + "epoch": 0.72, + "learning_rate": 3.796422159528084e-05, + "loss": 0.534, + "step": 18821 + }, + { + "epoch": 0.72, + "learning_rate": 3.796358040523211e-05, + "loss": 0.5555, + "step": 18822 + }, + { + "epoch": 0.72, + "learning_rate": 3.796293921518338e-05, + "loss": 0.6055, + "step": 18823 + }, + { + "epoch": 0.72, + "learning_rate": 3.7962298025134654e-05, + "loss": 0.511, + "step": 18824 + }, + { + "epoch": 0.72, + "learning_rate": 3.7961656835085926e-05, + "loss": 0.5032, + "step": 18825 + }, + { + "epoch": 0.72, + "learning_rate": 3.796101564503719e-05, + "loss": 0.5542, + "step": 18826 + }, + { + "epoch": 0.72, + "learning_rate": 3.796037445498846e-05, + "loss": 0.5705, + "step": 18827 + }, + { + "epoch": 0.72, + "learning_rate": 3.795973326493973e-05, + "loss": 0.4815, + "step": 18828 + }, + { + "epoch": 0.72, + "learning_rate": 3.7959092074891e-05, + "loss": 0.4318, + "step": 18829 + }, + { + "epoch": 0.72, + "learning_rate": 3.7958450884842264e-05, + "loss": 0.5742, + "step": 18830 + }, + { + "epoch": 0.72, + "learning_rate": 3.7957809694793536e-05, + "loss": 0.5974, + "step": 18831 + }, + { + "epoch": 0.72, + "learning_rate": 3.795716850474481e-05, + "loss": 0.5427, + "step": 18832 + }, + { + "epoch": 0.72, + "learning_rate": 3.795652731469608e-05, + "loss": 0.4985, + "step": 18833 + }, + { + "epoch": 0.72, + "learning_rate": 3.795588612464735e-05, + "loss": 0.6254, + "step": 18834 + }, + { + "epoch": 0.72, + "learning_rate": 3.7955244934598616e-05, + "loss": 0.5893, + "step": 18835 + }, + { + "epoch": 0.72, + "learning_rate": 3.795460374454989e-05, + "loss": 0.4866, + "step": 18836 + }, + { + "epoch": 0.72, + "learning_rate": 3.795396255450115e-05, + "loss": 0.4681, + "step": 18837 + }, + { + "epoch": 0.72, + "learning_rate": 3.7953321364452424e-05, + "loss": 0.5388, + "step": 18838 + }, + { + "epoch": 0.72, + "learning_rate": 3.7952680174403696e-05, + "loss": 0.5113, + "step": 18839 + }, + { + "epoch": 0.72, + "learning_rate": 3.795203898435497e-05, + "loss": 0.5182, + "step": 18840 + }, + { + "epoch": 0.72, + "learning_rate": 3.795139779430623e-05, + "loss": 0.5201, + "step": 18841 + }, + { + "epoch": 0.72, + "learning_rate": 3.7950756604257505e-05, + "loss": 0.4952, + "step": 18842 + }, + { + "epoch": 0.72, + "learning_rate": 3.7950115414208776e-05, + "loss": 0.45, + "step": 18843 + }, + { + "epoch": 0.72, + "learning_rate": 3.794947422416004e-05, + "loss": 0.4818, + "step": 18844 + }, + { + "epoch": 0.72, + "learning_rate": 3.794883303411131e-05, + "loss": 0.5334, + "step": 18845 + }, + { + "epoch": 0.72, + "learning_rate": 3.794819184406258e-05, + "loss": 0.4867, + "step": 18846 + }, + { + "epoch": 0.72, + "learning_rate": 3.794755065401385e-05, + "loss": 0.4722, + "step": 18847 + }, + { + "epoch": 0.72, + "learning_rate": 3.794690946396512e-05, + "loss": 0.5007, + "step": 18848 + }, + { + "epoch": 0.72, + "learning_rate": 3.794626827391639e-05, + "loss": 0.5476, + "step": 18849 + }, + { + "epoch": 0.72, + "learning_rate": 3.794562708386766e-05, + "loss": 0.4763, + "step": 18850 + }, + { + "epoch": 0.72, + "learning_rate": 3.794498589381893e-05, + "loss": 0.4163, + "step": 18851 + }, + { + "epoch": 0.72, + "learning_rate": 3.79443447037702e-05, + "loss": 0.4816, + "step": 18852 + }, + { + "epoch": 0.72, + "learning_rate": 3.7943703513721466e-05, + "loss": 0.6884, + "step": 18853 + }, + { + "epoch": 0.72, + "learning_rate": 3.794306232367274e-05, + "loss": 0.4976, + "step": 18854 + }, + { + "epoch": 0.72, + "learning_rate": 3.794242113362401e-05, + "loss": 0.511, + "step": 18855 + }, + { + "epoch": 0.72, + "learning_rate": 3.794177994357528e-05, + "loss": 0.5119, + "step": 18856 + }, + { + "epoch": 0.72, + "learning_rate": 3.7941138753526546e-05, + "loss": 0.5052, + "step": 18857 + }, + { + "epoch": 0.73, + "learning_rate": 3.794049756347782e-05, + "loss": 0.4623, + "step": 18858 + }, + { + "epoch": 0.73, + "learning_rate": 3.793985637342908e-05, + "loss": 0.5981, + "step": 18859 + }, + { + "epoch": 0.73, + "learning_rate": 3.7939215183380355e-05, + "loss": 0.3626, + "step": 18860 + }, + { + "epoch": 0.73, + "learning_rate": 3.7938573993331627e-05, + "loss": 0.4364, + "step": 18861 + }, + { + "epoch": 0.73, + "learning_rate": 3.793793280328289e-05, + "loss": 0.5494, + "step": 18862 + }, + { + "epoch": 0.73, + "learning_rate": 3.793729161323416e-05, + "loss": 0.5138, + "step": 18863 + }, + { + "epoch": 0.73, + "learning_rate": 3.7936650423185435e-05, + "loss": 0.3633, + "step": 18864 + }, + { + "epoch": 0.73, + "learning_rate": 3.793600923313671e-05, + "loss": 0.5844, + "step": 18865 + }, + { + "epoch": 0.73, + "learning_rate": 3.793536804308797e-05, + "loss": 0.4523, + "step": 18866 + }, + { + "epoch": 0.73, + "learning_rate": 3.793472685303924e-05, + "loss": 0.5713, + "step": 18867 + }, + { + "epoch": 0.73, + "learning_rate": 3.793408566299051e-05, + "loss": 0.5063, + "step": 18868 + }, + { + "epoch": 0.73, + "learning_rate": 3.793344447294178e-05, + "loss": 0.5578, + "step": 18869 + }, + { + "epoch": 0.73, + "learning_rate": 3.793280328289305e-05, + "loss": 0.5523, + "step": 18870 + }, + { + "epoch": 0.73, + "learning_rate": 3.7932162092844317e-05, + "loss": 0.4426, + "step": 18871 + }, + { + "epoch": 0.73, + "learning_rate": 3.7931520902795595e-05, + "loss": 0.5623, + "step": 18872 + }, + { + "epoch": 0.73, + "learning_rate": 3.793087971274686e-05, + "loss": 0.4675, + "step": 18873 + }, + { + "epoch": 0.73, + "learning_rate": 3.793023852269813e-05, + "loss": 0.4831, + "step": 18874 + }, + { + "epoch": 0.73, + "learning_rate": 3.79295973326494e-05, + "loss": 0.4522, + "step": 18875 + }, + { + "epoch": 0.73, + "learning_rate": 3.792895614260067e-05, + "loss": 0.6649, + "step": 18876 + }, + { + "epoch": 0.73, + "learning_rate": 3.792831495255193e-05, + "loss": 0.4427, + "step": 18877 + }, + { + "epoch": 0.73, + "learning_rate": 3.7927673762503205e-05, + "loss": 0.5427, + "step": 18878 + }, + { + "epoch": 0.73, + "learning_rate": 3.792703257245448e-05, + "loss": 0.5232, + "step": 18879 + }, + { + "epoch": 0.73, + "learning_rate": 3.792639138240575e-05, + "loss": 0.4546, + "step": 18880 + }, + { + "epoch": 0.73, + "learning_rate": 3.792575019235702e-05, + "loss": 0.5055, + "step": 18881 + }, + { + "epoch": 0.73, + "learning_rate": 3.7925109002308285e-05, + "loss": 0.5338, + "step": 18882 + }, + { + "epoch": 0.73, + "learning_rate": 3.792446781225956e-05, + "loss": 0.511, + "step": 18883 + }, + { + "epoch": 0.73, + "learning_rate": 3.792382662221082e-05, + "loss": 0.4889, + "step": 18884 + }, + { + "epoch": 0.73, + "learning_rate": 3.7923185432162094e-05, + "loss": 0.6418, + "step": 18885 + }, + { + "epoch": 0.73, + "learning_rate": 3.7922544242113365e-05, + "loss": 0.5413, + "step": 18886 + }, + { + "epoch": 0.73, + "learning_rate": 3.792190305206463e-05, + "loss": 0.5055, + "step": 18887 + }, + { + "epoch": 0.73, + "learning_rate": 3.792126186201591e-05, + "loss": 0.5346, + "step": 18888 + }, + { + "epoch": 0.73, + "learning_rate": 3.7920620671967174e-05, + "loss": 0.5298, + "step": 18889 + }, + { + "epoch": 0.73, + "learning_rate": 3.7919979481918445e-05, + "loss": 0.5474, + "step": 18890 + }, + { + "epoch": 0.73, + "learning_rate": 3.791933829186971e-05, + "loss": 0.5612, + "step": 18891 + }, + { + "epoch": 0.73, + "learning_rate": 3.791869710182098e-05, + "loss": 0.479, + "step": 18892 + }, + { + "epoch": 0.73, + "learning_rate": 3.791805591177225e-05, + "loss": 0.5371, + "step": 18893 + }, + { + "epoch": 0.73, + "learning_rate": 3.791741472172352e-05, + "loss": 0.523, + "step": 18894 + }, + { + "epoch": 0.73, + "learning_rate": 3.791677353167479e-05, + "loss": 0.386, + "step": 18895 + }, + { + "epoch": 0.73, + "learning_rate": 3.791613234162606e-05, + "loss": 0.4257, + "step": 18896 + }, + { + "epoch": 0.73, + "learning_rate": 3.7915491151577334e-05, + "loss": 0.4805, + "step": 18897 + }, + { + "epoch": 0.73, + "learning_rate": 3.79148499615286e-05, + "loss": 0.5296, + "step": 18898 + }, + { + "epoch": 0.73, + "learning_rate": 3.791420877147987e-05, + "loss": 0.5913, + "step": 18899 + }, + { + "epoch": 0.73, + "learning_rate": 3.7913567581431135e-05, + "loss": 0.5594, + "step": 18900 + }, + { + "epoch": 0.73, + "learning_rate": 3.791292639138241e-05, + "loss": 0.4898, + "step": 18901 + }, + { + "epoch": 0.73, + "learning_rate": 3.791228520133367e-05, + "loss": 0.5368, + "step": 18902 + }, + { + "epoch": 0.73, + "learning_rate": 3.7911644011284944e-05, + "loss": 0.5612, + "step": 18903 + }, + { + "epoch": 0.73, + "learning_rate": 3.791100282123622e-05, + "loss": 0.5654, + "step": 18904 + }, + { + "epoch": 0.73, + "learning_rate": 3.791036163118749e-05, + "loss": 0.5551, + "step": 18905 + }, + { + "epoch": 0.73, + "learning_rate": 3.790972044113876e-05, + "loss": 0.5588, + "step": 18906 + }, + { + "epoch": 0.73, + "learning_rate": 3.7909079251090024e-05, + "loss": 0.4896, + "step": 18907 + }, + { + "epoch": 0.73, + "learning_rate": 3.7908438061041296e-05, + "loss": 0.5315, + "step": 18908 + }, + { + "epoch": 0.73, + "learning_rate": 3.790779687099256e-05, + "loss": 0.4848, + "step": 18909 + }, + { + "epoch": 0.73, + "learning_rate": 3.790715568094383e-05, + "loss": 0.4601, + "step": 18910 + }, + { + "epoch": 0.73, + "learning_rate": 3.7906514490895104e-05, + "loss": 0.4792, + "step": 18911 + }, + { + "epoch": 0.73, + "learning_rate": 3.7905873300846376e-05, + "loss": 0.4732, + "step": 18912 + }, + { + "epoch": 0.73, + "learning_rate": 3.790523211079765e-05, + "loss": 0.5228, + "step": 18913 + }, + { + "epoch": 0.73, + "learning_rate": 3.790459092074891e-05, + "loss": 0.4943, + "step": 18914 + }, + { + "epoch": 0.73, + "learning_rate": 3.7903949730700184e-05, + "loss": 0.5211, + "step": 18915 + }, + { + "epoch": 0.73, + "learning_rate": 3.790330854065145e-05, + "loss": 0.6238, + "step": 18916 + }, + { + "epoch": 0.73, + "learning_rate": 3.790266735060272e-05, + "loss": 0.4248, + "step": 18917 + }, + { + "epoch": 0.73, + "learning_rate": 3.7902026160553986e-05, + "loss": 0.5793, + "step": 18918 + }, + { + "epoch": 0.73, + "learning_rate": 3.790138497050526e-05, + "loss": 0.5645, + "step": 18919 + }, + { + "epoch": 0.73, + "learning_rate": 3.790074378045653e-05, + "loss": 0.521, + "step": 18920 + }, + { + "epoch": 0.73, + "learning_rate": 3.79001025904078e-05, + "loss": 0.5476, + "step": 18921 + }, + { + "epoch": 0.73, + "learning_rate": 3.789946140035907e-05, + "loss": 0.5448, + "step": 18922 + }, + { + "epoch": 0.73, + "learning_rate": 3.789882021031034e-05, + "loss": 0.5445, + "step": 18923 + }, + { + "epoch": 0.73, + "learning_rate": 3.789817902026161e-05, + "loss": 0.4899, + "step": 18924 + }, + { + "epoch": 0.73, + "learning_rate": 3.7897537830212874e-05, + "loss": 0.5192, + "step": 18925 + }, + { + "epoch": 0.73, + "learning_rate": 3.7896896640164146e-05, + "loss": 0.4852, + "step": 18926 + }, + { + "epoch": 0.73, + "learning_rate": 3.789625545011542e-05, + "loss": 0.4953, + "step": 18927 + }, + { + "epoch": 0.73, + "learning_rate": 3.789561426006669e-05, + "loss": 0.5603, + "step": 18928 + }, + { + "epoch": 0.73, + "learning_rate": 3.7894973070017954e-05, + "loss": 0.5483, + "step": 18929 + }, + { + "epoch": 0.73, + "learning_rate": 3.7894331879969226e-05, + "loss": 0.4915, + "step": 18930 + }, + { + "epoch": 0.73, + "learning_rate": 3.78936906899205e-05, + "loss": 0.4828, + "step": 18931 + }, + { + "epoch": 0.73, + "learning_rate": 3.789304949987176e-05, + "loss": 0.5223, + "step": 18932 + }, + { + "epoch": 0.73, + "learning_rate": 3.7892408309823034e-05, + "loss": 0.5506, + "step": 18933 + }, + { + "epoch": 0.73, + "learning_rate": 3.78917671197743e-05, + "loss": 0.4534, + "step": 18934 + }, + { + "epoch": 0.73, + "learning_rate": 3.789112592972557e-05, + "loss": 0.5478, + "step": 18935 + }, + { + "epoch": 0.73, + "learning_rate": 3.789048473967684e-05, + "loss": 0.4696, + "step": 18936 + }, + { + "epoch": 0.73, + "learning_rate": 3.7889843549628114e-05, + "loss": 0.518, + "step": 18937 + }, + { + "epoch": 0.73, + "learning_rate": 3.788920235957938e-05, + "loss": 0.44, + "step": 18938 + }, + { + "epoch": 0.73, + "learning_rate": 3.788856116953065e-05, + "loss": 0.5461, + "step": 18939 + }, + { + "epoch": 0.73, + "learning_rate": 3.788791997948192e-05, + "loss": 0.5342, + "step": 18940 + }, + { + "epoch": 0.73, + "learning_rate": 3.788727878943319e-05, + "loss": 0.4091, + "step": 18941 + }, + { + "epoch": 0.73, + "learning_rate": 3.788663759938446e-05, + "loss": 0.5164, + "step": 18942 + }, + { + "epoch": 0.73, + "learning_rate": 3.788599640933573e-05, + "loss": 0.483, + "step": 18943 + }, + { + "epoch": 0.73, + "learning_rate": 3.7885355219287e-05, + "loss": 0.4955, + "step": 18944 + }, + { + "epoch": 0.73, + "learning_rate": 3.788471402923827e-05, + "loss": 0.5437, + "step": 18945 + }, + { + "epoch": 0.73, + "learning_rate": 3.788407283918954e-05, + "loss": 0.6054, + "step": 18946 + }, + { + "epoch": 0.73, + "learning_rate": 3.7883431649140804e-05, + "loss": 0.5791, + "step": 18947 + }, + { + "epoch": 0.73, + "learning_rate": 3.7882790459092076e-05, + "loss": 0.4735, + "step": 18948 + }, + { + "epoch": 0.73, + "learning_rate": 3.788214926904335e-05, + "loss": 0.4692, + "step": 18949 + }, + { + "epoch": 0.73, + "learning_rate": 3.788150807899461e-05, + "loss": 0.4752, + "step": 18950 + }, + { + "epoch": 0.73, + "learning_rate": 3.7880866888945885e-05, + "loss": 0.5058, + "step": 18951 + }, + { + "epoch": 0.73, + "learning_rate": 3.7880225698897156e-05, + "loss": 0.5363, + "step": 18952 + }, + { + "epoch": 0.73, + "learning_rate": 3.787958450884843e-05, + "loss": 0.5598, + "step": 18953 + }, + { + "epoch": 0.73, + "learning_rate": 3.787894331879969e-05, + "loss": 0.5274, + "step": 18954 + }, + { + "epoch": 0.73, + "learning_rate": 3.7878302128750965e-05, + "loss": 0.5268, + "step": 18955 + }, + { + "epoch": 0.73, + "learning_rate": 3.787766093870223e-05, + "loss": 0.5828, + "step": 18956 + }, + { + "epoch": 0.73, + "learning_rate": 3.78770197486535e-05, + "loss": 0.5099, + "step": 18957 + }, + { + "epoch": 0.73, + "learning_rate": 3.787637855860477e-05, + "loss": 0.5465, + "step": 18958 + }, + { + "epoch": 0.73, + "learning_rate": 3.787573736855604e-05, + "loss": 0.5163, + "step": 18959 + }, + { + "epoch": 0.73, + "learning_rate": 3.7875096178507316e-05, + "loss": 0.5526, + "step": 18960 + }, + { + "epoch": 0.73, + "learning_rate": 3.787445498845858e-05, + "loss": 0.5017, + "step": 18961 + }, + { + "epoch": 0.73, + "learning_rate": 3.787381379840985e-05, + "loss": 0.4183, + "step": 18962 + }, + { + "epoch": 0.73, + "learning_rate": 3.787317260836112e-05, + "loss": 0.4401, + "step": 18963 + }, + { + "epoch": 0.73, + "learning_rate": 3.787253141831239e-05, + "loss": 0.5363, + "step": 18964 + }, + { + "epoch": 0.73, + "learning_rate": 3.7871890228263655e-05, + "loss": 0.5537, + "step": 18965 + }, + { + "epoch": 0.73, + "learning_rate": 3.7871249038214926e-05, + "loss": 0.5719, + "step": 18966 + }, + { + "epoch": 0.73, + "learning_rate": 3.78706078481662e-05, + "loss": 0.5067, + "step": 18967 + }, + { + "epoch": 0.73, + "learning_rate": 3.786996665811747e-05, + "loss": 0.5569, + "step": 18968 + }, + { + "epoch": 0.73, + "learning_rate": 3.786932546806874e-05, + "loss": 0.5521, + "step": 18969 + }, + { + "epoch": 0.73, + "learning_rate": 3.7868684278020006e-05, + "loss": 0.5607, + "step": 18970 + }, + { + "epoch": 0.73, + "learning_rate": 3.786804308797128e-05, + "loss": 0.5298, + "step": 18971 + }, + { + "epoch": 0.73, + "learning_rate": 3.786740189792254e-05, + "loss": 0.3796, + "step": 18972 + }, + { + "epoch": 0.73, + "learning_rate": 3.7866760707873815e-05, + "loss": 0.5353, + "step": 18973 + }, + { + "epoch": 0.73, + "learning_rate": 3.786611951782508e-05, + "loss": 0.4797, + "step": 18974 + }, + { + "epoch": 0.73, + "learning_rate": 3.786547832777635e-05, + "loss": 0.5634, + "step": 18975 + }, + { + "epoch": 0.73, + "learning_rate": 3.786483713772763e-05, + "loss": 0.5501, + "step": 18976 + }, + { + "epoch": 0.73, + "learning_rate": 3.7864195947678895e-05, + "loss": 0.6709, + "step": 18977 + }, + { + "epoch": 0.73, + "learning_rate": 3.786355475763017e-05, + "loss": 0.5083, + "step": 18978 + }, + { + "epoch": 0.73, + "learning_rate": 3.786291356758143e-05, + "loss": 0.5251, + "step": 18979 + }, + { + "epoch": 0.73, + "learning_rate": 3.78622723775327e-05, + "loss": 0.4917, + "step": 18980 + }, + { + "epoch": 0.73, + "learning_rate": 3.786163118748397e-05, + "loss": 0.5374, + "step": 18981 + }, + { + "epoch": 0.73, + "learning_rate": 3.786098999743524e-05, + "loss": 0.589, + "step": 18982 + }, + { + "epoch": 0.73, + "learning_rate": 3.786034880738651e-05, + "loss": 0.538, + "step": 18983 + }, + { + "epoch": 0.73, + "learning_rate": 3.7859707617337783e-05, + "loss": 0.54, + "step": 18984 + }, + { + "epoch": 0.73, + "learning_rate": 3.7859066427289055e-05, + "loss": 0.4958, + "step": 18985 + }, + { + "epoch": 0.73, + "learning_rate": 3.785842523724032e-05, + "loss": 0.5042, + "step": 18986 + }, + { + "epoch": 0.73, + "learning_rate": 3.785778404719159e-05, + "loss": 0.5605, + "step": 18987 + }, + { + "epoch": 0.73, + "learning_rate": 3.785714285714286e-05, + "loss": 0.4879, + "step": 18988 + }, + { + "epoch": 0.73, + "learning_rate": 3.785650166709413e-05, + "loss": 0.5405, + "step": 18989 + }, + { + "epoch": 0.73, + "learning_rate": 3.785586047704539e-05, + "loss": 0.4088, + "step": 18990 + }, + { + "epoch": 0.73, + "learning_rate": 3.7855219286996665e-05, + "loss": 0.6999, + "step": 18991 + }, + { + "epoch": 0.73, + "learning_rate": 3.785457809694794e-05, + "loss": 0.4604, + "step": 18992 + }, + { + "epoch": 0.73, + "learning_rate": 3.785393690689921e-05, + "loss": 0.5364, + "step": 18993 + }, + { + "epoch": 0.73, + "learning_rate": 3.785329571685048e-05, + "loss": 0.5104, + "step": 18994 + }, + { + "epoch": 0.73, + "learning_rate": 3.7852654526801745e-05, + "loss": 0.5161, + "step": 18995 + }, + { + "epoch": 0.73, + "learning_rate": 3.785201333675302e-05, + "loss": 0.4823, + "step": 18996 + }, + { + "epoch": 0.73, + "learning_rate": 3.785137214670428e-05, + "loss": 0.4075, + "step": 18997 + }, + { + "epoch": 0.73, + "learning_rate": 3.7850730956655554e-05, + "loss": 0.3757, + "step": 18998 + }, + { + "epoch": 0.73, + "learning_rate": 3.7850089766606825e-05, + "loss": 0.4425, + "step": 18999 + }, + { + "epoch": 0.73, + "learning_rate": 3.78494485765581e-05, + "loss": 0.4306, + "step": 19000 + }, + { + "epoch": 0.73, + "learning_rate": 3.784880738650936e-05, + "loss": 0.5299, + "step": 19001 + }, + { + "epoch": 0.73, + "learning_rate": 3.7848166196460634e-05, + "loss": 0.5652, + "step": 19002 + }, + { + "epoch": 0.73, + "learning_rate": 3.7847525006411905e-05, + "loss": 0.4684, + "step": 19003 + }, + { + "epoch": 0.73, + "learning_rate": 3.784688381636317e-05, + "loss": 0.5466, + "step": 19004 + }, + { + "epoch": 0.73, + "learning_rate": 3.784624262631444e-05, + "loss": 0.4555, + "step": 19005 + }, + { + "epoch": 0.73, + "learning_rate": 3.784560143626571e-05, + "loss": 0.5726, + "step": 19006 + }, + { + "epoch": 0.73, + "learning_rate": 3.784496024621698e-05, + "loss": 0.4693, + "step": 19007 + }, + { + "epoch": 0.73, + "learning_rate": 3.784431905616825e-05, + "loss": 0.4665, + "step": 19008 + }, + { + "epoch": 0.73, + "learning_rate": 3.784367786611952e-05, + "loss": 0.5344, + "step": 19009 + }, + { + "epoch": 0.73, + "learning_rate": 3.784303667607079e-05, + "loss": 0.5862, + "step": 19010 + }, + { + "epoch": 0.73, + "learning_rate": 3.784239548602206e-05, + "loss": 0.5605, + "step": 19011 + }, + { + "epoch": 0.73, + "learning_rate": 3.784175429597333e-05, + "loss": 0.452, + "step": 19012 + }, + { + "epoch": 0.73, + "learning_rate": 3.7841113105924595e-05, + "loss": 0.4626, + "step": 19013 + }, + { + "epoch": 0.73, + "learning_rate": 3.784047191587587e-05, + "loss": 0.442, + "step": 19014 + }, + { + "epoch": 0.73, + "learning_rate": 3.783983072582714e-05, + "loss": 0.4736, + "step": 19015 + }, + { + "epoch": 0.73, + "learning_rate": 3.783918953577841e-05, + "loss": 0.3904, + "step": 19016 + }, + { + "epoch": 0.73, + "learning_rate": 3.7838548345729676e-05, + "loss": 0.4568, + "step": 19017 + }, + { + "epoch": 0.73, + "learning_rate": 3.783790715568095e-05, + "loss": 0.5115, + "step": 19018 + }, + { + "epoch": 0.73, + "learning_rate": 3.783726596563221e-05, + "loss": 0.5411, + "step": 19019 + }, + { + "epoch": 0.73, + "learning_rate": 3.7836624775583484e-05, + "loss": 0.5109, + "step": 19020 + }, + { + "epoch": 0.73, + "learning_rate": 3.7835983585534756e-05, + "loss": 0.5033, + "step": 19021 + }, + { + "epoch": 0.73, + "learning_rate": 3.783534239548602e-05, + "loss": 0.4104, + "step": 19022 + }, + { + "epoch": 0.73, + "learning_rate": 3.783470120543729e-05, + "loss": 0.4782, + "step": 19023 + }, + { + "epoch": 0.73, + "learning_rate": 3.7834060015388564e-05, + "loss": 0.6092, + "step": 19024 + }, + { + "epoch": 0.73, + "learning_rate": 3.7833418825339836e-05, + "loss": 0.4742, + "step": 19025 + }, + { + "epoch": 0.73, + "learning_rate": 3.78327776352911e-05, + "loss": 0.5322, + "step": 19026 + }, + { + "epoch": 0.73, + "learning_rate": 3.783213644524237e-05, + "loss": 0.4496, + "step": 19027 + }, + { + "epoch": 0.73, + "learning_rate": 3.783149525519364e-05, + "loss": 0.544, + "step": 19028 + }, + { + "epoch": 0.73, + "learning_rate": 3.783085406514491e-05, + "loss": 0.5096, + "step": 19029 + }, + { + "epoch": 0.73, + "learning_rate": 3.783021287509618e-05, + "loss": 0.549, + "step": 19030 + }, + { + "epoch": 0.73, + "learning_rate": 3.782957168504745e-05, + "loss": 0.5317, + "step": 19031 + }, + { + "epoch": 0.73, + "learning_rate": 3.7828930494998724e-05, + "loss": 0.5338, + "step": 19032 + }, + { + "epoch": 0.73, + "learning_rate": 3.782828930494999e-05, + "loss": 0.6211, + "step": 19033 + }, + { + "epoch": 0.73, + "learning_rate": 3.782764811490126e-05, + "loss": 0.4816, + "step": 19034 + }, + { + "epoch": 0.73, + "learning_rate": 3.7827006924852526e-05, + "loss": 0.5772, + "step": 19035 + }, + { + "epoch": 0.73, + "learning_rate": 3.78263657348038e-05, + "loss": 0.3915, + "step": 19036 + }, + { + "epoch": 0.73, + "learning_rate": 3.782572454475506e-05, + "loss": 0.4916, + "step": 19037 + }, + { + "epoch": 0.73, + "learning_rate": 3.7825083354706334e-05, + "loss": 0.5828, + "step": 19038 + }, + { + "epoch": 0.73, + "learning_rate": 3.7824442164657606e-05, + "loss": 0.475, + "step": 19039 + }, + { + "epoch": 0.73, + "learning_rate": 3.782380097460888e-05, + "loss": 0.4754, + "step": 19040 + }, + { + "epoch": 0.73, + "learning_rate": 3.782315978456015e-05, + "loss": 0.4516, + "step": 19041 + }, + { + "epoch": 0.73, + "learning_rate": 3.7822518594511414e-05, + "loss": 0.4192, + "step": 19042 + }, + { + "epoch": 0.73, + "learning_rate": 3.7821877404462686e-05, + "loss": 0.425, + "step": 19043 + }, + { + "epoch": 0.73, + "learning_rate": 3.782123621441395e-05, + "loss": 0.4636, + "step": 19044 + }, + { + "epoch": 0.73, + "learning_rate": 3.782059502436522e-05, + "loss": 0.5389, + "step": 19045 + }, + { + "epoch": 0.73, + "learning_rate": 3.781995383431649e-05, + "loss": 0.4675, + "step": 19046 + }, + { + "epoch": 0.73, + "learning_rate": 3.781931264426776e-05, + "loss": 0.4812, + "step": 19047 + }, + { + "epoch": 0.73, + "learning_rate": 3.781867145421904e-05, + "loss": 0.3879, + "step": 19048 + }, + { + "epoch": 0.73, + "learning_rate": 3.78180302641703e-05, + "loss": 0.4624, + "step": 19049 + }, + { + "epoch": 0.73, + "learning_rate": 3.7817389074121574e-05, + "loss": 0.5528, + "step": 19050 + }, + { + "epoch": 0.73, + "learning_rate": 3.781674788407284e-05, + "loss": 0.5999, + "step": 19051 + }, + { + "epoch": 0.73, + "learning_rate": 3.781610669402411e-05, + "loss": 0.543, + "step": 19052 + }, + { + "epoch": 0.73, + "learning_rate": 3.7815465503975376e-05, + "loss": 0.5968, + "step": 19053 + }, + { + "epoch": 0.73, + "learning_rate": 3.781482431392665e-05, + "loss": 0.4578, + "step": 19054 + }, + { + "epoch": 0.73, + "learning_rate": 3.781418312387792e-05, + "loss": 0.5858, + "step": 19055 + }, + { + "epoch": 0.73, + "learning_rate": 3.781354193382919e-05, + "loss": 0.5166, + "step": 19056 + }, + { + "epoch": 0.73, + "learning_rate": 3.781290074378046e-05, + "loss": 0.543, + "step": 19057 + }, + { + "epoch": 0.73, + "learning_rate": 3.781225955373173e-05, + "loss": 0.4656, + "step": 19058 + }, + { + "epoch": 0.73, + "learning_rate": 3.7811618363683e-05, + "loss": 0.4616, + "step": 19059 + }, + { + "epoch": 0.73, + "learning_rate": 3.7810977173634264e-05, + "loss": 0.4965, + "step": 19060 + }, + { + "epoch": 0.73, + "learning_rate": 3.7810335983585536e-05, + "loss": 0.482, + "step": 19061 + }, + { + "epoch": 0.73, + "learning_rate": 3.78096947935368e-05, + "loss": 0.4806, + "step": 19062 + }, + { + "epoch": 0.73, + "learning_rate": 3.780905360348807e-05, + "loss": 0.5131, + "step": 19063 + }, + { + "epoch": 0.73, + "learning_rate": 3.7808412413439345e-05, + "loss": 0.3819, + "step": 19064 + }, + { + "epoch": 0.73, + "learning_rate": 3.7807771223390616e-05, + "loss": 0.5698, + "step": 19065 + }, + { + "epoch": 0.73, + "learning_rate": 3.780713003334189e-05, + "loss": 0.4502, + "step": 19066 + }, + { + "epoch": 0.73, + "learning_rate": 3.780648884329315e-05, + "loss": 0.4632, + "step": 19067 + }, + { + "epoch": 0.73, + "learning_rate": 3.7805847653244425e-05, + "loss": 0.5369, + "step": 19068 + }, + { + "epoch": 0.73, + "learning_rate": 3.780520646319569e-05, + "loss": 0.5152, + "step": 19069 + }, + { + "epoch": 0.73, + "learning_rate": 3.780456527314696e-05, + "loss": 0.5371, + "step": 19070 + }, + { + "epoch": 0.73, + "learning_rate": 3.780392408309823e-05, + "loss": 0.5839, + "step": 19071 + }, + { + "epoch": 0.73, + "learning_rate": 3.7803282893049505e-05, + "loss": 0.4844, + "step": 19072 + }, + { + "epoch": 0.73, + "learning_rate": 3.7802641703000776e-05, + "loss": 0.511, + "step": 19073 + }, + { + "epoch": 0.73, + "learning_rate": 3.780200051295204e-05, + "loss": 0.4978, + "step": 19074 + }, + { + "epoch": 0.73, + "learning_rate": 3.780135932290331e-05, + "loss": 0.4626, + "step": 19075 + }, + { + "epoch": 0.73, + "learning_rate": 3.780071813285458e-05, + "loss": 0.4806, + "step": 19076 + }, + { + "epoch": 0.73, + "learning_rate": 3.780007694280585e-05, + "loss": 0.4718, + "step": 19077 + }, + { + "epoch": 0.73, + "learning_rate": 3.7799435752757115e-05, + "loss": 0.507, + "step": 19078 + }, + { + "epoch": 0.73, + "learning_rate": 3.7798794562708386e-05, + "loss": 0.5286, + "step": 19079 + }, + { + "epoch": 0.73, + "learning_rate": 3.779815337265966e-05, + "loss": 0.4437, + "step": 19080 + }, + { + "epoch": 0.73, + "learning_rate": 3.779751218261093e-05, + "loss": 0.4272, + "step": 19081 + }, + { + "epoch": 0.73, + "learning_rate": 3.77968709925622e-05, + "loss": 0.5408, + "step": 19082 + }, + { + "epoch": 0.73, + "learning_rate": 3.7796229802513467e-05, + "loss": 0.5906, + "step": 19083 + }, + { + "epoch": 0.73, + "learning_rate": 3.779558861246474e-05, + "loss": 0.5262, + "step": 19084 + }, + { + "epoch": 0.73, + "learning_rate": 3.7794947422416e-05, + "loss": 0.5671, + "step": 19085 + }, + { + "epoch": 0.73, + "learning_rate": 3.7794306232367275e-05, + "loss": 0.5729, + "step": 19086 + }, + { + "epoch": 0.73, + "learning_rate": 3.779366504231855e-05, + "loss": 0.4803, + "step": 19087 + }, + { + "epoch": 0.73, + "learning_rate": 3.779302385226982e-05, + "loss": 0.5067, + "step": 19088 + }, + { + "epoch": 0.73, + "learning_rate": 3.779238266222108e-05, + "loss": 0.5124, + "step": 19089 + }, + { + "epoch": 0.73, + "learning_rate": 3.7791741472172355e-05, + "loss": 0.452, + "step": 19090 + }, + { + "epoch": 0.73, + "learning_rate": 3.779110028212363e-05, + "loss": 0.6059, + "step": 19091 + }, + { + "epoch": 0.73, + "learning_rate": 3.779045909207489e-05, + "loss": 0.5937, + "step": 19092 + }, + { + "epoch": 0.73, + "learning_rate": 3.778981790202616e-05, + "loss": 0.533, + "step": 19093 + }, + { + "epoch": 0.73, + "learning_rate": 3.778917671197743e-05, + "loss": 0.4653, + "step": 19094 + }, + { + "epoch": 0.73, + "learning_rate": 3.77885355219287e-05, + "loss": 0.5499, + "step": 19095 + }, + { + "epoch": 0.73, + "learning_rate": 3.778789433187997e-05, + "loss": 0.5141, + "step": 19096 + }, + { + "epoch": 0.73, + "learning_rate": 3.7787253141831243e-05, + "loss": 0.5539, + "step": 19097 + }, + { + "epoch": 0.73, + "learning_rate": 3.778661195178251e-05, + "loss": 0.4499, + "step": 19098 + }, + { + "epoch": 0.73, + "learning_rate": 3.778597076173378e-05, + "loss": 0.5907, + "step": 19099 + }, + { + "epoch": 0.73, + "learning_rate": 3.778532957168505e-05, + "loss": 0.422, + "step": 19100 + }, + { + "epoch": 0.73, + "learning_rate": 3.778468838163632e-05, + "loss": 0.4796, + "step": 19101 + }, + { + "epoch": 0.73, + "learning_rate": 3.778404719158759e-05, + "loss": 0.4983, + "step": 19102 + }, + { + "epoch": 0.73, + "learning_rate": 3.778340600153886e-05, + "loss": 0.4633, + "step": 19103 + }, + { + "epoch": 0.73, + "learning_rate": 3.778276481149013e-05, + "loss": 0.5135, + "step": 19104 + }, + { + "epoch": 0.73, + "learning_rate": 3.77821236214414e-05, + "loss": 0.4759, + "step": 19105 + }, + { + "epoch": 0.73, + "learning_rate": 3.778148243139267e-05, + "loss": 0.5059, + "step": 19106 + }, + { + "epoch": 0.73, + "learning_rate": 3.7780841241343934e-05, + "loss": 0.3842, + "step": 19107 + }, + { + "epoch": 0.73, + "learning_rate": 3.7780200051295205e-05, + "loss": 0.5326, + "step": 19108 + }, + { + "epoch": 0.73, + "learning_rate": 3.777955886124648e-05, + "loss": 0.4769, + "step": 19109 + }, + { + "epoch": 0.73, + "learning_rate": 3.777891767119774e-05, + "loss": 0.4869, + "step": 19110 + }, + { + "epoch": 0.73, + "learning_rate": 3.7778276481149014e-05, + "loss": 0.6052, + "step": 19111 + }, + { + "epoch": 0.73, + "learning_rate": 3.7777635291100285e-05, + "loss": 0.5863, + "step": 19112 + }, + { + "epoch": 0.73, + "learning_rate": 3.777699410105156e-05, + "loss": 0.467, + "step": 19113 + }, + { + "epoch": 0.73, + "learning_rate": 3.777635291100282e-05, + "loss": 0.5229, + "step": 19114 + }, + { + "epoch": 0.73, + "learning_rate": 3.7775711720954094e-05, + "loss": 0.6594, + "step": 19115 + }, + { + "epoch": 0.73, + "learning_rate": 3.777507053090536e-05, + "loss": 0.4569, + "step": 19116 + }, + { + "epoch": 0.73, + "learning_rate": 3.777442934085663e-05, + "loss": 0.5396, + "step": 19117 + }, + { + "epoch": 0.74, + "learning_rate": 3.77737881508079e-05, + "loss": 0.3499, + "step": 19118 + }, + { + "epoch": 0.74, + "learning_rate": 3.777314696075917e-05, + "loss": 0.6182, + "step": 19119 + }, + { + "epoch": 0.74, + "learning_rate": 3.7772505770710445e-05, + "loss": 0.5539, + "step": 19120 + }, + { + "epoch": 0.74, + "learning_rate": 3.777186458066171e-05, + "loss": 0.5432, + "step": 19121 + }, + { + "epoch": 0.74, + "learning_rate": 3.777122339061298e-05, + "loss": 0.4894, + "step": 19122 + }, + { + "epoch": 0.74, + "learning_rate": 3.777058220056425e-05, + "loss": 0.5941, + "step": 19123 + }, + { + "epoch": 0.74, + "learning_rate": 3.776994101051552e-05, + "loss": 0.5403, + "step": 19124 + }, + { + "epoch": 0.74, + "learning_rate": 3.7769299820466784e-05, + "loss": 0.5165, + "step": 19125 + }, + { + "epoch": 0.74, + "learning_rate": 3.7768658630418055e-05, + "loss": 0.4586, + "step": 19126 + }, + { + "epoch": 0.74, + "learning_rate": 3.776801744036933e-05, + "loss": 0.5831, + "step": 19127 + }, + { + "epoch": 0.74, + "learning_rate": 3.77673762503206e-05, + "loss": 0.6027, + "step": 19128 + }, + { + "epoch": 0.74, + "learning_rate": 3.776673506027187e-05, + "loss": 0.5135, + "step": 19129 + }, + { + "epoch": 0.74, + "learning_rate": 3.7766093870223136e-05, + "loss": 0.5435, + "step": 19130 + }, + { + "epoch": 0.74, + "learning_rate": 3.776545268017441e-05, + "loss": 0.488, + "step": 19131 + }, + { + "epoch": 0.74, + "learning_rate": 3.776481149012567e-05, + "loss": 0.4006, + "step": 19132 + }, + { + "epoch": 0.74, + "learning_rate": 3.7764170300076944e-05, + "loss": 0.6412, + "step": 19133 + }, + { + "epoch": 0.74, + "learning_rate": 3.776352911002821e-05, + "loss": 0.6119, + "step": 19134 + }, + { + "epoch": 0.74, + "learning_rate": 3.776288791997948e-05, + "loss": 0.4643, + "step": 19135 + }, + { + "epoch": 0.74, + "learning_rate": 3.776224672993076e-05, + "loss": 0.4865, + "step": 19136 + }, + { + "epoch": 0.74, + "learning_rate": 3.7761605539882024e-05, + "loss": 0.6329, + "step": 19137 + }, + { + "epoch": 0.74, + "learning_rate": 3.7760964349833296e-05, + "loss": 0.5596, + "step": 19138 + }, + { + "epoch": 0.74, + "learning_rate": 3.776032315978456e-05, + "loss": 0.5293, + "step": 19139 + }, + { + "epoch": 0.74, + "learning_rate": 3.775968196973583e-05, + "loss": 0.5462, + "step": 19140 + }, + { + "epoch": 0.74, + "learning_rate": 3.77590407796871e-05, + "loss": 0.5175, + "step": 19141 + }, + { + "epoch": 0.74, + "learning_rate": 3.775839958963837e-05, + "loss": 0.4787, + "step": 19142 + }, + { + "epoch": 0.74, + "learning_rate": 3.775775839958964e-05, + "loss": 0.4852, + "step": 19143 + }, + { + "epoch": 0.74, + "learning_rate": 3.775711720954091e-05, + "loss": 0.4359, + "step": 19144 + }, + { + "epoch": 0.74, + "learning_rate": 3.7756476019492184e-05, + "loss": 0.518, + "step": 19145 + }, + { + "epoch": 0.74, + "learning_rate": 3.775583482944345e-05, + "loss": 0.5565, + "step": 19146 + }, + { + "epoch": 0.74, + "learning_rate": 3.775519363939472e-05, + "loss": 0.5713, + "step": 19147 + }, + { + "epoch": 0.74, + "learning_rate": 3.7754552449345986e-05, + "loss": 0.5266, + "step": 19148 + }, + { + "epoch": 0.74, + "learning_rate": 3.775391125929726e-05, + "loss": 0.4625, + "step": 19149 + }, + { + "epoch": 0.74, + "learning_rate": 3.775327006924852e-05, + "loss": 0.6007, + "step": 19150 + }, + { + "epoch": 0.74, + "learning_rate": 3.7752628879199794e-05, + "loss": 0.5291, + "step": 19151 + }, + { + "epoch": 0.74, + "learning_rate": 3.7751987689151066e-05, + "loss": 0.511, + "step": 19152 + }, + { + "epoch": 0.74, + "learning_rate": 3.775134649910234e-05, + "loss": 0.4908, + "step": 19153 + }, + { + "epoch": 0.74, + "learning_rate": 3.775070530905361e-05, + "loss": 0.4826, + "step": 19154 + }, + { + "epoch": 0.74, + "learning_rate": 3.7750064119004874e-05, + "loss": 0.4462, + "step": 19155 + }, + { + "epoch": 0.74, + "learning_rate": 3.7749422928956146e-05, + "loss": 0.5907, + "step": 19156 + }, + { + "epoch": 0.74, + "learning_rate": 3.774878173890741e-05, + "loss": 0.5434, + "step": 19157 + }, + { + "epoch": 0.74, + "learning_rate": 3.774814054885868e-05, + "loss": 0.5643, + "step": 19158 + }, + { + "epoch": 0.74, + "learning_rate": 3.7747499358809954e-05, + "loss": 0.4671, + "step": 19159 + }, + { + "epoch": 0.74, + "learning_rate": 3.7746858168761226e-05, + "loss": 0.5035, + "step": 19160 + }, + { + "epoch": 0.74, + "learning_rate": 3.774621697871249e-05, + "loss": 0.4694, + "step": 19161 + }, + { + "epoch": 0.74, + "learning_rate": 3.774557578866376e-05, + "loss": 0.436, + "step": 19162 + }, + { + "epoch": 0.74, + "learning_rate": 3.7744934598615034e-05, + "loss": 0.5534, + "step": 19163 + }, + { + "epoch": 0.74, + "learning_rate": 3.77442934085663e-05, + "loss": 0.5067, + "step": 19164 + }, + { + "epoch": 0.74, + "learning_rate": 3.774365221851757e-05, + "loss": 0.6765, + "step": 19165 + }, + { + "epoch": 0.74, + "learning_rate": 3.7743011028468836e-05, + "loss": 0.4933, + "step": 19166 + }, + { + "epoch": 0.74, + "learning_rate": 3.774236983842011e-05, + "loss": 0.4582, + "step": 19167 + }, + { + "epoch": 0.74, + "learning_rate": 3.774172864837138e-05, + "loss": 0.4888, + "step": 19168 + }, + { + "epoch": 0.74, + "learning_rate": 3.774108745832265e-05, + "loss": 0.517, + "step": 19169 + }, + { + "epoch": 0.74, + "learning_rate": 3.7740446268273916e-05, + "loss": 0.5691, + "step": 19170 + }, + { + "epoch": 0.74, + "learning_rate": 3.773980507822519e-05, + "loss": 0.7017, + "step": 19171 + }, + { + "epoch": 0.74, + "learning_rate": 3.773916388817646e-05, + "loss": 0.5781, + "step": 19172 + }, + { + "epoch": 0.74, + "learning_rate": 3.7738522698127725e-05, + "loss": 0.5666, + "step": 19173 + }, + { + "epoch": 0.74, + "learning_rate": 3.7737881508078996e-05, + "loss": 0.4241, + "step": 19174 + }, + { + "epoch": 0.74, + "learning_rate": 3.773724031803027e-05, + "loss": 0.4889, + "step": 19175 + }, + { + "epoch": 0.74, + "learning_rate": 3.773659912798154e-05, + "loss": 0.5609, + "step": 19176 + }, + { + "epoch": 0.74, + "learning_rate": 3.7735957937932805e-05, + "loss": 0.4995, + "step": 19177 + }, + { + "epoch": 0.74, + "learning_rate": 3.7735316747884076e-05, + "loss": 0.599, + "step": 19178 + }, + { + "epoch": 0.74, + "learning_rate": 3.773467555783534e-05, + "loss": 0.4834, + "step": 19179 + }, + { + "epoch": 0.74, + "learning_rate": 3.773403436778661e-05, + "loss": 0.5392, + "step": 19180 + }, + { + "epoch": 0.74, + "learning_rate": 3.7733393177737885e-05, + "loss": 0.4959, + "step": 19181 + }, + { + "epoch": 0.74, + "learning_rate": 3.773275198768915e-05, + "loss": 0.5263, + "step": 19182 + }, + { + "epoch": 0.74, + "learning_rate": 3.773211079764042e-05, + "loss": 0.4795, + "step": 19183 + }, + { + "epoch": 0.74, + "learning_rate": 3.773146960759169e-05, + "loss": 0.3704, + "step": 19184 + }, + { + "epoch": 0.74, + "learning_rate": 3.7730828417542965e-05, + "loss": 0.5816, + "step": 19185 + }, + { + "epoch": 0.74, + "learning_rate": 3.773018722749423e-05, + "loss": 0.4579, + "step": 19186 + }, + { + "epoch": 0.74, + "learning_rate": 3.77295460374455e-05, + "loss": 0.5176, + "step": 19187 + }, + { + "epoch": 0.74, + "learning_rate": 3.7728904847396766e-05, + "loss": 0.4775, + "step": 19188 + }, + { + "epoch": 0.74, + "learning_rate": 3.772826365734804e-05, + "loss": 0.477, + "step": 19189 + }, + { + "epoch": 0.74, + "learning_rate": 3.772762246729931e-05, + "loss": 0.5384, + "step": 19190 + }, + { + "epoch": 0.74, + "learning_rate": 3.772698127725058e-05, + "loss": 0.5223, + "step": 19191 + }, + { + "epoch": 0.74, + "learning_rate": 3.772634008720185e-05, + "loss": 0.6739, + "step": 19192 + }, + { + "epoch": 0.74, + "learning_rate": 3.772569889715312e-05, + "loss": 0.4662, + "step": 19193 + }, + { + "epoch": 0.74, + "learning_rate": 3.772505770710439e-05, + "loss": 0.5069, + "step": 19194 + }, + { + "epoch": 0.74, + "learning_rate": 3.7724416517055655e-05, + "loss": 0.5188, + "step": 19195 + }, + { + "epoch": 0.74, + "learning_rate": 3.7723775327006927e-05, + "loss": 0.4567, + "step": 19196 + }, + { + "epoch": 0.74, + "learning_rate": 3.772313413695819e-05, + "loss": 0.5333, + "step": 19197 + }, + { + "epoch": 0.74, + "learning_rate": 3.772249294690946e-05, + "loss": 0.3837, + "step": 19198 + }, + { + "epoch": 0.74, + "learning_rate": 3.7721851756860735e-05, + "loss": 0.5404, + "step": 19199 + }, + { + "epoch": 0.74, + "learning_rate": 3.772121056681201e-05, + "loss": 0.4799, + "step": 19200 + }, + { + "epoch": 0.74, + "learning_rate": 3.772056937676328e-05, + "loss": 0.6078, + "step": 19201 + }, + { + "epoch": 0.74, + "learning_rate": 3.771992818671454e-05, + "loss": 0.6024, + "step": 19202 + }, + { + "epoch": 0.74, + "learning_rate": 3.7719286996665815e-05, + "loss": 0.6358, + "step": 19203 + }, + { + "epoch": 0.74, + "learning_rate": 3.771864580661708e-05, + "loss": 0.6626, + "step": 19204 + }, + { + "epoch": 0.74, + "learning_rate": 3.771800461656835e-05, + "loss": 0.4061, + "step": 19205 + }, + { + "epoch": 0.74, + "learning_rate": 3.771736342651962e-05, + "loss": 0.4676, + "step": 19206 + }, + { + "epoch": 0.74, + "learning_rate": 3.771672223647089e-05, + "loss": 0.5361, + "step": 19207 + }, + { + "epoch": 0.74, + "learning_rate": 3.771608104642217e-05, + "loss": 0.4125, + "step": 19208 + }, + { + "epoch": 0.74, + "learning_rate": 3.771543985637343e-05, + "loss": 0.5258, + "step": 19209 + }, + { + "epoch": 0.74, + "learning_rate": 3.7714798666324703e-05, + "loss": 0.45, + "step": 19210 + }, + { + "epoch": 0.74, + "learning_rate": 3.771415747627597e-05, + "loss": 0.5985, + "step": 19211 + }, + { + "epoch": 0.74, + "learning_rate": 3.771351628622724e-05, + "loss": 0.4933, + "step": 19212 + }, + { + "epoch": 0.74, + "learning_rate": 3.7712875096178505e-05, + "loss": 0.464, + "step": 19213 + }, + { + "epoch": 0.74, + "learning_rate": 3.771223390612978e-05, + "loss": 0.5025, + "step": 19214 + }, + { + "epoch": 0.74, + "learning_rate": 3.771159271608105e-05, + "loss": 0.5688, + "step": 19215 + }, + { + "epoch": 0.74, + "learning_rate": 3.771095152603232e-05, + "loss": 0.5059, + "step": 19216 + }, + { + "epoch": 0.74, + "learning_rate": 3.771031033598359e-05, + "loss": 0.6218, + "step": 19217 + }, + { + "epoch": 0.74, + "learning_rate": 3.770966914593486e-05, + "loss": 0.492, + "step": 19218 + }, + { + "epoch": 0.74, + "learning_rate": 3.770902795588613e-05, + "loss": 0.47, + "step": 19219 + }, + { + "epoch": 0.74, + "learning_rate": 3.7708386765837394e-05, + "loss": 0.5265, + "step": 19220 + }, + { + "epoch": 0.74, + "learning_rate": 3.7707745575788665e-05, + "loss": 0.5527, + "step": 19221 + }, + { + "epoch": 0.74, + "learning_rate": 3.770710438573993e-05, + "loss": 0.5042, + "step": 19222 + }, + { + "epoch": 0.74, + "learning_rate": 3.77064631956912e-05, + "loss": 0.5369, + "step": 19223 + }, + { + "epoch": 0.74, + "learning_rate": 3.7705822005642474e-05, + "loss": 0.544, + "step": 19224 + }, + { + "epoch": 0.74, + "learning_rate": 3.7705180815593745e-05, + "loss": 0.5559, + "step": 19225 + }, + { + "epoch": 0.74, + "learning_rate": 3.770453962554502e-05, + "loss": 0.5804, + "step": 19226 + }, + { + "epoch": 0.74, + "learning_rate": 3.770389843549628e-05, + "loss": 0.5314, + "step": 19227 + }, + { + "epoch": 0.74, + "learning_rate": 3.7703257245447554e-05, + "loss": 0.5974, + "step": 19228 + }, + { + "epoch": 0.74, + "learning_rate": 3.770261605539882e-05, + "loss": 0.5161, + "step": 19229 + }, + { + "epoch": 0.74, + "learning_rate": 3.770197486535009e-05, + "loss": 0.5623, + "step": 19230 + }, + { + "epoch": 0.74, + "learning_rate": 3.770133367530136e-05, + "loss": 0.5065, + "step": 19231 + }, + { + "epoch": 0.74, + "learning_rate": 3.7700692485252634e-05, + "loss": 0.6101, + "step": 19232 + }, + { + "epoch": 0.74, + "learning_rate": 3.77000512952039e-05, + "loss": 0.6745, + "step": 19233 + }, + { + "epoch": 0.74, + "learning_rate": 3.769941010515517e-05, + "loss": 0.5648, + "step": 19234 + }, + { + "epoch": 0.74, + "learning_rate": 3.769876891510644e-05, + "loss": 0.4478, + "step": 19235 + }, + { + "epoch": 0.74, + "learning_rate": 3.769812772505771e-05, + "loss": 0.564, + "step": 19236 + }, + { + "epoch": 0.74, + "learning_rate": 3.769748653500898e-05, + "loss": 0.4685, + "step": 19237 + }, + { + "epoch": 0.74, + "learning_rate": 3.7696845344960244e-05, + "loss": 0.4764, + "step": 19238 + }, + { + "epoch": 0.74, + "learning_rate": 3.7696204154911516e-05, + "loss": 0.5416, + "step": 19239 + }, + { + "epoch": 0.74, + "learning_rate": 3.769556296486279e-05, + "loss": 0.476, + "step": 19240 + }, + { + "epoch": 0.74, + "learning_rate": 3.769492177481406e-05, + "loss": 0.4252, + "step": 19241 + }, + { + "epoch": 0.74, + "learning_rate": 3.7694280584765324e-05, + "loss": 0.5373, + "step": 19242 + }, + { + "epoch": 0.74, + "learning_rate": 3.7693639394716596e-05, + "loss": 0.4463, + "step": 19243 + }, + { + "epoch": 0.74, + "learning_rate": 3.769299820466787e-05, + "loss": 0.5738, + "step": 19244 + }, + { + "epoch": 0.74, + "learning_rate": 3.769235701461913e-05, + "loss": 0.5723, + "step": 19245 + }, + { + "epoch": 0.74, + "learning_rate": 3.7691715824570404e-05, + "loss": 0.5296, + "step": 19246 + }, + { + "epoch": 0.74, + "learning_rate": 3.7691074634521676e-05, + "loss": 0.4327, + "step": 19247 + }, + { + "epoch": 0.74, + "learning_rate": 3.769043344447295e-05, + "loss": 0.5223, + "step": 19248 + }, + { + "epoch": 0.74, + "learning_rate": 3.768979225442421e-05, + "loss": 0.4879, + "step": 19249 + }, + { + "epoch": 0.74, + "learning_rate": 3.7689151064375484e-05, + "loss": 0.5432, + "step": 19250 + }, + { + "epoch": 0.74, + "learning_rate": 3.768850987432675e-05, + "loss": 0.4894, + "step": 19251 + }, + { + "epoch": 0.74, + "learning_rate": 3.768786868427802e-05, + "loss": 0.4939, + "step": 19252 + }, + { + "epoch": 0.74, + "learning_rate": 3.768722749422929e-05, + "loss": 0.3784, + "step": 19253 + }, + { + "epoch": 0.74, + "learning_rate": 3.768658630418056e-05, + "loss": 0.5384, + "step": 19254 + }, + { + "epoch": 0.74, + "learning_rate": 3.768594511413183e-05, + "loss": 0.4784, + "step": 19255 + }, + { + "epoch": 0.74, + "learning_rate": 3.76853039240831e-05, + "loss": 0.5041, + "step": 19256 + }, + { + "epoch": 0.74, + "learning_rate": 3.768466273403437e-05, + "loss": 0.5017, + "step": 19257 + }, + { + "epoch": 0.74, + "learning_rate": 3.768402154398564e-05, + "loss": 0.4271, + "step": 19258 + }, + { + "epoch": 0.74, + "learning_rate": 3.768338035393691e-05, + "loss": 0.4868, + "step": 19259 + }, + { + "epoch": 0.74, + "learning_rate": 3.7682739163888174e-05, + "loss": 0.4995, + "step": 19260 + }, + { + "epoch": 0.74, + "learning_rate": 3.7682097973839446e-05, + "loss": 0.4523, + "step": 19261 + }, + { + "epoch": 0.74, + "learning_rate": 3.768145678379072e-05, + "loss": 0.546, + "step": 19262 + }, + { + "epoch": 0.74, + "learning_rate": 3.768081559374199e-05, + "loss": 0.6548, + "step": 19263 + }, + { + "epoch": 0.74, + "learning_rate": 3.768017440369326e-05, + "loss": 0.5019, + "step": 19264 + }, + { + "epoch": 0.74, + "learning_rate": 3.7679533213644526e-05, + "loss": 0.5038, + "step": 19265 + }, + { + "epoch": 0.74, + "learning_rate": 3.76788920235958e-05, + "loss": 0.5923, + "step": 19266 + }, + { + "epoch": 0.74, + "learning_rate": 3.767825083354706e-05, + "loss": 0.5268, + "step": 19267 + }, + { + "epoch": 0.74, + "learning_rate": 3.7677609643498334e-05, + "loss": 0.5913, + "step": 19268 + }, + { + "epoch": 0.74, + "learning_rate": 3.7676968453449606e-05, + "loss": 0.5849, + "step": 19269 + }, + { + "epoch": 0.74, + "learning_rate": 3.767632726340087e-05, + "loss": 0.4914, + "step": 19270 + }, + { + "epoch": 0.74, + "learning_rate": 3.767568607335214e-05, + "loss": 0.5581, + "step": 19271 + }, + { + "epoch": 0.74, + "learning_rate": 3.7675044883303414e-05, + "loss": 0.4063, + "step": 19272 + }, + { + "epoch": 0.74, + "learning_rate": 3.7674403693254686e-05, + "loss": 0.3955, + "step": 19273 + }, + { + "epoch": 0.74, + "learning_rate": 3.767376250320595e-05, + "loss": 0.4776, + "step": 19274 + }, + { + "epoch": 0.74, + "learning_rate": 3.767312131315722e-05, + "loss": 0.5222, + "step": 19275 + }, + { + "epoch": 0.74, + "learning_rate": 3.767248012310849e-05, + "loss": 0.5115, + "step": 19276 + }, + { + "epoch": 0.74, + "learning_rate": 3.767183893305976e-05, + "loss": 0.4767, + "step": 19277 + }, + { + "epoch": 0.74, + "learning_rate": 3.767119774301103e-05, + "loss": 0.5087, + "step": 19278 + }, + { + "epoch": 0.74, + "learning_rate": 3.76705565529623e-05, + "loss": 0.5348, + "step": 19279 + }, + { + "epoch": 0.74, + "learning_rate": 3.7669915362913575e-05, + "loss": 0.5076, + "step": 19280 + }, + { + "epoch": 0.74, + "learning_rate": 3.766927417286484e-05, + "loss": 0.5973, + "step": 19281 + }, + { + "epoch": 0.74, + "learning_rate": 3.766863298281611e-05, + "loss": 0.3936, + "step": 19282 + }, + { + "epoch": 0.74, + "learning_rate": 3.7667991792767376e-05, + "loss": 0.5584, + "step": 19283 + }, + { + "epoch": 0.74, + "learning_rate": 3.766735060271865e-05, + "loss": 0.3799, + "step": 19284 + }, + { + "epoch": 0.74, + "learning_rate": 3.766670941266991e-05, + "loss": 0.6108, + "step": 19285 + }, + { + "epoch": 0.74, + "learning_rate": 3.7666068222621185e-05, + "loss": 0.4763, + "step": 19286 + }, + { + "epoch": 0.74, + "learning_rate": 3.7665427032572456e-05, + "loss": 0.5223, + "step": 19287 + }, + { + "epoch": 0.74, + "learning_rate": 3.766478584252373e-05, + "loss": 0.4574, + "step": 19288 + }, + { + "epoch": 0.74, + "learning_rate": 3.7664144652475e-05, + "loss": 0.4832, + "step": 19289 + }, + { + "epoch": 0.74, + "learning_rate": 3.7663503462426265e-05, + "loss": 0.6252, + "step": 19290 + }, + { + "epoch": 0.74, + "learning_rate": 3.7662862272377536e-05, + "loss": 0.5224, + "step": 19291 + }, + { + "epoch": 0.74, + "learning_rate": 3.76622210823288e-05, + "loss": 0.5391, + "step": 19292 + }, + { + "epoch": 0.74, + "learning_rate": 3.766157989228007e-05, + "loss": 0.5674, + "step": 19293 + }, + { + "epoch": 0.74, + "learning_rate": 3.766093870223134e-05, + "loss": 0.4699, + "step": 19294 + }, + { + "epoch": 0.74, + "learning_rate": 3.766029751218261e-05, + "loss": 0.6206, + "step": 19295 + }, + { + "epoch": 0.74, + "learning_rate": 3.765965632213389e-05, + "loss": 0.5023, + "step": 19296 + }, + { + "epoch": 0.74, + "learning_rate": 3.765901513208515e-05, + "loss": 0.5383, + "step": 19297 + }, + { + "epoch": 0.74, + "learning_rate": 3.7658373942036425e-05, + "loss": 0.5233, + "step": 19298 + }, + { + "epoch": 0.74, + "learning_rate": 3.765773275198769e-05, + "loss": 0.5165, + "step": 19299 + }, + { + "epoch": 0.74, + "learning_rate": 3.765709156193896e-05, + "loss": 0.5411, + "step": 19300 + }, + { + "epoch": 0.74, + "learning_rate": 3.7656450371890226e-05, + "loss": 0.4835, + "step": 19301 + }, + { + "epoch": 0.74, + "learning_rate": 3.76558091818415e-05, + "loss": 0.6979, + "step": 19302 + }, + { + "epoch": 0.74, + "learning_rate": 3.765516799179277e-05, + "loss": 0.521, + "step": 19303 + }, + { + "epoch": 0.74, + "learning_rate": 3.765452680174404e-05, + "loss": 0.5112, + "step": 19304 + }, + { + "epoch": 0.74, + "learning_rate": 3.765388561169531e-05, + "loss": 0.5015, + "step": 19305 + }, + { + "epoch": 0.74, + "learning_rate": 3.765324442164658e-05, + "loss": 0.4739, + "step": 19306 + }, + { + "epoch": 0.74, + "learning_rate": 3.765260323159785e-05, + "loss": 0.4097, + "step": 19307 + }, + { + "epoch": 0.74, + "learning_rate": 3.7651962041549115e-05, + "loss": 0.3941, + "step": 19308 + }, + { + "epoch": 0.74, + "learning_rate": 3.7651320851500387e-05, + "loss": 0.4053, + "step": 19309 + }, + { + "epoch": 0.74, + "learning_rate": 3.765067966145165e-05, + "loss": 0.4857, + "step": 19310 + }, + { + "epoch": 0.74, + "learning_rate": 3.765003847140292e-05, + "loss": 0.443, + "step": 19311 + }, + { + "epoch": 0.74, + "learning_rate": 3.7649397281354195e-05, + "loss": 0.5567, + "step": 19312 + }, + { + "epoch": 0.74, + "learning_rate": 3.764875609130547e-05, + "loss": 0.6004, + "step": 19313 + }, + { + "epoch": 0.74, + "learning_rate": 3.764811490125674e-05, + "loss": 0.548, + "step": 19314 + }, + { + "epoch": 0.74, + "learning_rate": 3.7647473711208e-05, + "loss": 0.4892, + "step": 19315 + }, + { + "epoch": 0.74, + "learning_rate": 3.7646832521159275e-05, + "loss": 0.5737, + "step": 19316 + }, + { + "epoch": 0.74, + "learning_rate": 3.764619133111054e-05, + "loss": 0.5425, + "step": 19317 + }, + { + "epoch": 0.74, + "learning_rate": 3.764555014106181e-05, + "loss": 0.5141, + "step": 19318 + }, + { + "epoch": 0.74, + "learning_rate": 3.7644908951013083e-05, + "loss": 0.5288, + "step": 19319 + }, + { + "epoch": 0.74, + "learning_rate": 3.7644267760964355e-05, + "loss": 0.4437, + "step": 19320 + }, + { + "epoch": 0.74, + "learning_rate": 3.764362657091562e-05, + "loss": 0.412, + "step": 19321 + }, + { + "epoch": 0.74, + "learning_rate": 3.764298538086689e-05, + "loss": 0.4522, + "step": 19322 + }, + { + "epoch": 0.74, + "learning_rate": 3.7642344190818164e-05, + "loss": 0.6571, + "step": 19323 + }, + { + "epoch": 0.74, + "learning_rate": 3.764170300076943e-05, + "loss": 0.4319, + "step": 19324 + }, + { + "epoch": 0.74, + "learning_rate": 3.76410618107207e-05, + "loss": 0.5942, + "step": 19325 + }, + { + "epoch": 0.74, + "learning_rate": 3.7640420620671965e-05, + "loss": 0.506, + "step": 19326 + }, + { + "epoch": 0.74, + "learning_rate": 3.763977943062324e-05, + "loss": 0.5052, + "step": 19327 + }, + { + "epoch": 0.74, + "learning_rate": 3.763913824057451e-05, + "loss": 0.4645, + "step": 19328 + }, + { + "epoch": 0.74, + "learning_rate": 3.763849705052578e-05, + "loss": 0.508, + "step": 19329 + }, + { + "epoch": 0.74, + "learning_rate": 3.7637855860477045e-05, + "loss": 0.579, + "step": 19330 + }, + { + "epoch": 0.74, + "learning_rate": 3.763721467042832e-05, + "loss": 0.5826, + "step": 19331 + }, + { + "epoch": 0.74, + "learning_rate": 3.763657348037959e-05, + "loss": 0.5413, + "step": 19332 + }, + { + "epoch": 0.74, + "learning_rate": 3.7635932290330854e-05, + "loss": 0.5523, + "step": 19333 + }, + { + "epoch": 0.74, + "learning_rate": 3.7635291100282125e-05, + "loss": 0.5035, + "step": 19334 + }, + { + "epoch": 0.74, + "learning_rate": 3.76346499102334e-05, + "loss": 0.5812, + "step": 19335 + }, + { + "epoch": 0.74, + "learning_rate": 3.763400872018467e-05, + "loss": 0.4179, + "step": 19336 + }, + { + "epoch": 0.74, + "learning_rate": 3.7633367530135934e-05, + "loss": 0.4671, + "step": 19337 + }, + { + "epoch": 0.74, + "learning_rate": 3.7632726340087205e-05, + "loss": 0.5562, + "step": 19338 + }, + { + "epoch": 0.74, + "learning_rate": 3.763208515003847e-05, + "loss": 0.3817, + "step": 19339 + }, + { + "epoch": 0.74, + "learning_rate": 3.763144395998974e-05, + "loss": 0.51, + "step": 19340 + }, + { + "epoch": 0.74, + "learning_rate": 3.7630802769941014e-05, + "loss": 0.5562, + "step": 19341 + }, + { + "epoch": 0.74, + "learning_rate": 3.763016157989228e-05, + "loss": 0.6611, + "step": 19342 + }, + { + "epoch": 0.74, + "learning_rate": 3.762952038984355e-05, + "loss": 0.5043, + "step": 19343 + }, + { + "epoch": 0.74, + "learning_rate": 3.762887919979482e-05, + "loss": 0.4008, + "step": 19344 + }, + { + "epoch": 0.74, + "learning_rate": 3.7628238009746094e-05, + "loss": 0.5917, + "step": 19345 + }, + { + "epoch": 0.74, + "learning_rate": 3.762759681969736e-05, + "loss": 0.5047, + "step": 19346 + }, + { + "epoch": 0.74, + "learning_rate": 3.762695562964863e-05, + "loss": 0.5519, + "step": 19347 + }, + { + "epoch": 0.74, + "learning_rate": 3.7626314439599895e-05, + "loss": 0.5548, + "step": 19348 + }, + { + "epoch": 0.74, + "learning_rate": 3.762567324955117e-05, + "loss": 0.5015, + "step": 19349 + }, + { + "epoch": 0.74, + "learning_rate": 3.762503205950244e-05, + "loss": 0.4958, + "step": 19350 + }, + { + "epoch": 0.74, + "learning_rate": 3.762439086945371e-05, + "loss": 0.4978, + "step": 19351 + }, + { + "epoch": 0.74, + "learning_rate": 3.762374967940498e-05, + "loss": 0.5845, + "step": 19352 + }, + { + "epoch": 0.74, + "learning_rate": 3.762310848935625e-05, + "loss": 0.471, + "step": 19353 + }, + { + "epoch": 0.74, + "learning_rate": 3.762246729930752e-05, + "loss": 0.592, + "step": 19354 + }, + { + "epoch": 0.74, + "learning_rate": 3.7621826109258784e-05, + "loss": 0.4892, + "step": 19355 + }, + { + "epoch": 0.74, + "learning_rate": 3.7621184919210056e-05, + "loss": 0.487, + "step": 19356 + }, + { + "epoch": 0.74, + "learning_rate": 3.762054372916132e-05, + "loss": 0.5321, + "step": 19357 + }, + { + "epoch": 0.74, + "learning_rate": 3.761990253911259e-05, + "loss": 0.615, + "step": 19358 + }, + { + "epoch": 0.74, + "learning_rate": 3.7619261349063864e-05, + "loss": 0.5081, + "step": 19359 + }, + { + "epoch": 0.74, + "learning_rate": 3.7618620159015136e-05, + "loss": 0.5184, + "step": 19360 + }, + { + "epoch": 0.74, + "learning_rate": 3.761797896896641e-05, + "loss": 0.5208, + "step": 19361 + }, + { + "epoch": 0.74, + "learning_rate": 3.761733777891767e-05, + "loss": 0.4596, + "step": 19362 + }, + { + "epoch": 0.74, + "learning_rate": 3.7616696588868944e-05, + "loss": 0.5623, + "step": 19363 + }, + { + "epoch": 0.74, + "learning_rate": 3.761605539882021e-05, + "loss": 0.481, + "step": 19364 + }, + { + "epoch": 0.74, + "learning_rate": 3.761541420877148e-05, + "loss": 0.6175, + "step": 19365 + }, + { + "epoch": 0.74, + "learning_rate": 3.7614773018722746e-05, + "loss": 0.4431, + "step": 19366 + }, + { + "epoch": 0.74, + "learning_rate": 3.761413182867402e-05, + "loss": 0.5083, + "step": 19367 + }, + { + "epoch": 0.74, + "learning_rate": 3.7613490638625296e-05, + "loss": 0.4284, + "step": 19368 + }, + { + "epoch": 0.74, + "learning_rate": 3.761284944857656e-05, + "loss": 0.4727, + "step": 19369 + }, + { + "epoch": 0.74, + "learning_rate": 3.761220825852783e-05, + "loss": 0.5921, + "step": 19370 + }, + { + "epoch": 0.74, + "learning_rate": 3.76115670684791e-05, + "loss": 0.5091, + "step": 19371 + }, + { + "epoch": 0.74, + "learning_rate": 3.761092587843037e-05, + "loss": 0.4179, + "step": 19372 + }, + { + "epoch": 0.74, + "learning_rate": 3.7610284688381634e-05, + "loss": 0.4983, + "step": 19373 + }, + { + "epoch": 0.74, + "learning_rate": 3.7609643498332906e-05, + "loss": 0.4445, + "step": 19374 + }, + { + "epoch": 0.74, + "learning_rate": 3.760900230828418e-05, + "loss": 0.4718, + "step": 19375 + }, + { + "epoch": 0.74, + "learning_rate": 3.760836111823545e-05, + "loss": 0.4751, + "step": 19376 + }, + { + "epoch": 0.74, + "learning_rate": 3.760771992818672e-05, + "loss": 0.3614, + "step": 19377 + }, + { + "epoch": 0.75, + "learning_rate": 3.7607078738137986e-05, + "loss": 0.5128, + "step": 19378 + }, + { + "epoch": 0.75, + "learning_rate": 3.760643754808926e-05, + "loss": 0.4614, + "step": 19379 + }, + { + "epoch": 0.75, + "learning_rate": 3.760579635804052e-05, + "loss": 0.3975, + "step": 19380 + }, + { + "epoch": 0.75, + "learning_rate": 3.7605155167991794e-05, + "loss": 0.4803, + "step": 19381 + }, + { + "epoch": 0.75, + "learning_rate": 3.760451397794306e-05, + "loss": 0.6398, + "step": 19382 + }, + { + "epoch": 0.75, + "learning_rate": 3.760387278789433e-05, + "loss": 0.4697, + "step": 19383 + }, + { + "epoch": 0.75, + "learning_rate": 3.76032315978456e-05, + "loss": 0.5509, + "step": 19384 + }, + { + "epoch": 0.75, + "learning_rate": 3.7602590407796874e-05, + "loss": 0.4563, + "step": 19385 + }, + { + "epoch": 0.75, + "learning_rate": 3.7601949217748146e-05, + "loss": 0.5602, + "step": 19386 + }, + { + "epoch": 0.75, + "learning_rate": 3.760130802769941e-05, + "loss": 0.4228, + "step": 19387 + }, + { + "epoch": 0.75, + "learning_rate": 3.760066683765068e-05, + "loss": 0.4647, + "step": 19388 + }, + { + "epoch": 0.75, + "learning_rate": 3.760002564760195e-05, + "loss": 0.6664, + "step": 19389 + }, + { + "epoch": 0.75, + "learning_rate": 3.759938445755322e-05, + "loss": 0.5587, + "step": 19390 + }, + { + "epoch": 0.75, + "learning_rate": 3.759874326750449e-05, + "loss": 0.556, + "step": 19391 + }, + { + "epoch": 0.75, + "learning_rate": 3.759810207745576e-05, + "loss": 0.4837, + "step": 19392 + }, + { + "epoch": 0.75, + "learning_rate": 3.759746088740703e-05, + "loss": 0.5679, + "step": 19393 + }, + { + "epoch": 0.75, + "learning_rate": 3.75968196973583e-05, + "loss": 0.4788, + "step": 19394 + }, + { + "epoch": 0.75, + "learning_rate": 3.759617850730957e-05, + "loss": 0.5323, + "step": 19395 + }, + { + "epoch": 0.75, + "learning_rate": 3.7595537317260836e-05, + "loss": 0.5635, + "step": 19396 + }, + { + "epoch": 0.75, + "learning_rate": 3.759489612721211e-05, + "loss": 0.5326, + "step": 19397 + }, + { + "epoch": 0.75, + "learning_rate": 3.759425493716337e-05, + "loss": 0.4347, + "step": 19398 + }, + { + "epoch": 0.75, + "learning_rate": 3.7593613747114645e-05, + "loss": 0.4792, + "step": 19399 + }, + { + "epoch": 0.75, + "learning_rate": 3.7592972557065916e-05, + "loss": 0.5861, + "step": 19400 + }, + { + "epoch": 0.75, + "learning_rate": 3.759233136701719e-05, + "loss": 0.5137, + "step": 19401 + }, + { + "epoch": 0.75, + "learning_rate": 3.759169017696845e-05, + "loss": 0.5011, + "step": 19402 + }, + { + "epoch": 0.75, + "learning_rate": 3.7591048986919725e-05, + "loss": 0.5354, + "step": 19403 + }, + { + "epoch": 0.75, + "learning_rate": 3.7590407796870996e-05, + "loss": 0.4732, + "step": 19404 + }, + { + "epoch": 0.75, + "learning_rate": 3.758976660682226e-05, + "loss": 0.4378, + "step": 19405 + }, + { + "epoch": 0.75, + "learning_rate": 3.758912541677353e-05, + "loss": 0.4819, + "step": 19406 + }, + { + "epoch": 0.75, + "learning_rate": 3.7588484226724805e-05, + "loss": 0.4981, + "step": 19407 + }, + { + "epoch": 0.75, + "learning_rate": 3.7587843036676076e-05, + "loss": 0.5894, + "step": 19408 + }, + { + "epoch": 0.75, + "learning_rate": 3.758720184662734e-05, + "loss": 0.4437, + "step": 19409 + }, + { + "epoch": 0.75, + "learning_rate": 3.758656065657861e-05, + "loss": 0.6012, + "step": 19410 + }, + { + "epoch": 0.75, + "learning_rate": 3.758591946652988e-05, + "loss": 0.5601, + "step": 19411 + }, + { + "epoch": 0.75, + "learning_rate": 3.758527827648115e-05, + "loss": 0.4778, + "step": 19412 + }, + { + "epoch": 0.75, + "learning_rate": 3.758463708643242e-05, + "loss": 0.5642, + "step": 19413 + }, + { + "epoch": 0.75, + "learning_rate": 3.7583995896383686e-05, + "loss": 0.4828, + "step": 19414 + }, + { + "epoch": 0.75, + "learning_rate": 3.758335470633496e-05, + "loss": 0.4693, + "step": 19415 + }, + { + "epoch": 0.75, + "learning_rate": 3.758271351628623e-05, + "loss": 0.453, + "step": 19416 + }, + { + "epoch": 0.75, + "learning_rate": 3.75820723262375e-05, + "loss": 0.6208, + "step": 19417 + }, + { + "epoch": 0.75, + "learning_rate": 3.7581431136188767e-05, + "loss": 0.5681, + "step": 19418 + }, + { + "epoch": 0.75, + "learning_rate": 3.758078994614004e-05, + "loss": 0.5904, + "step": 19419 + }, + { + "epoch": 0.75, + "learning_rate": 3.75801487560913e-05, + "loss": 0.5301, + "step": 19420 + }, + { + "epoch": 0.75, + "learning_rate": 3.7579507566042575e-05, + "loss": 0.5534, + "step": 19421 + }, + { + "epoch": 0.75, + "learning_rate": 3.757886637599385e-05, + "loss": 0.4267, + "step": 19422 + }, + { + "epoch": 0.75, + "learning_rate": 3.757822518594512e-05, + "loss": 0.5444, + "step": 19423 + }, + { + "epoch": 0.75, + "learning_rate": 3.757758399589639e-05, + "loss": 0.4439, + "step": 19424 + }, + { + "epoch": 0.75, + "learning_rate": 3.7576942805847655e-05, + "loss": 0.507, + "step": 19425 + }, + { + "epoch": 0.75, + "learning_rate": 3.757630161579893e-05, + "loss": 0.5465, + "step": 19426 + }, + { + "epoch": 0.75, + "learning_rate": 3.757566042575019e-05, + "loss": 0.4959, + "step": 19427 + }, + { + "epoch": 0.75, + "learning_rate": 3.757501923570146e-05, + "loss": 0.4969, + "step": 19428 + }, + { + "epoch": 0.75, + "learning_rate": 3.757437804565273e-05, + "loss": 0.5736, + "step": 19429 + }, + { + "epoch": 0.75, + "learning_rate": 3.7573736855604e-05, + "loss": 0.5732, + "step": 19430 + }, + { + "epoch": 0.75, + "learning_rate": 3.757309566555527e-05, + "loss": 0.533, + "step": 19431 + }, + { + "epoch": 0.75, + "learning_rate": 3.7572454475506543e-05, + "loss": 0.5732, + "step": 19432 + }, + { + "epoch": 0.75, + "learning_rate": 3.7571813285457815e-05, + "loss": 0.4137, + "step": 19433 + }, + { + "epoch": 0.75, + "learning_rate": 3.757117209540908e-05, + "loss": 0.4572, + "step": 19434 + }, + { + "epoch": 0.75, + "learning_rate": 3.757053090536035e-05, + "loss": 0.5398, + "step": 19435 + }, + { + "epoch": 0.75, + "learning_rate": 3.756988971531162e-05, + "loss": 0.5186, + "step": 19436 + }, + { + "epoch": 0.75, + "learning_rate": 3.756924852526289e-05, + "loss": 0.5884, + "step": 19437 + }, + { + "epoch": 0.75, + "learning_rate": 3.7568607335214153e-05, + "loss": 0.4737, + "step": 19438 + }, + { + "epoch": 0.75, + "learning_rate": 3.756796614516543e-05, + "loss": 0.5293, + "step": 19439 + }, + { + "epoch": 0.75, + "learning_rate": 3.7567324955116704e-05, + "loss": 0.4696, + "step": 19440 + }, + { + "epoch": 0.75, + "learning_rate": 3.756668376506797e-05, + "loss": 0.5432, + "step": 19441 + }, + { + "epoch": 0.75, + "learning_rate": 3.756604257501924e-05, + "loss": 0.5013, + "step": 19442 + }, + { + "epoch": 0.75, + "learning_rate": 3.7565401384970505e-05, + "loss": 0.5637, + "step": 19443 + }, + { + "epoch": 0.75, + "learning_rate": 3.756476019492178e-05, + "loss": 0.6656, + "step": 19444 + }, + { + "epoch": 0.75, + "learning_rate": 3.756411900487304e-05, + "loss": 0.4017, + "step": 19445 + }, + { + "epoch": 0.75, + "learning_rate": 3.7563477814824314e-05, + "loss": 0.4694, + "step": 19446 + }, + { + "epoch": 0.75, + "learning_rate": 3.7562836624775585e-05, + "loss": 0.6461, + "step": 19447 + }, + { + "epoch": 0.75, + "learning_rate": 3.756219543472686e-05, + "loss": 0.5159, + "step": 19448 + }, + { + "epoch": 0.75, + "learning_rate": 3.756155424467813e-05, + "loss": 0.502, + "step": 19449 + }, + { + "epoch": 0.75, + "learning_rate": 3.7560913054629394e-05, + "loss": 0.5283, + "step": 19450 + }, + { + "epoch": 0.75, + "learning_rate": 3.7560271864580665e-05, + "loss": 0.5237, + "step": 19451 + }, + { + "epoch": 0.75, + "learning_rate": 3.755963067453193e-05, + "loss": 0.5616, + "step": 19452 + }, + { + "epoch": 0.75, + "learning_rate": 3.75589894844832e-05, + "loss": 0.5957, + "step": 19453 + }, + { + "epoch": 0.75, + "learning_rate": 3.755834829443447e-05, + "loss": 0.4518, + "step": 19454 + }, + { + "epoch": 0.75, + "learning_rate": 3.755770710438574e-05, + "loss": 0.4107, + "step": 19455 + }, + { + "epoch": 0.75, + "learning_rate": 3.755706591433702e-05, + "loss": 0.5225, + "step": 19456 + }, + { + "epoch": 0.75, + "learning_rate": 3.755642472428828e-05, + "loss": 0.5282, + "step": 19457 + }, + { + "epoch": 0.75, + "learning_rate": 3.7555783534239554e-05, + "loss": 0.5081, + "step": 19458 + }, + { + "epoch": 0.75, + "learning_rate": 3.755514234419082e-05, + "loss": 0.4523, + "step": 19459 + }, + { + "epoch": 0.75, + "learning_rate": 3.755450115414209e-05, + "loss": 0.5195, + "step": 19460 + }, + { + "epoch": 0.75, + "learning_rate": 3.7553859964093355e-05, + "loss": 0.5387, + "step": 19461 + }, + { + "epoch": 0.75, + "learning_rate": 3.755321877404463e-05, + "loss": 0.5319, + "step": 19462 + }, + { + "epoch": 0.75, + "learning_rate": 3.75525775839959e-05, + "loss": 0.6581, + "step": 19463 + }, + { + "epoch": 0.75, + "learning_rate": 3.755193639394717e-05, + "loss": 0.5363, + "step": 19464 + }, + { + "epoch": 0.75, + "learning_rate": 3.755129520389844e-05, + "loss": 0.6356, + "step": 19465 + }, + { + "epoch": 0.75, + "learning_rate": 3.755065401384971e-05, + "loss": 0.6205, + "step": 19466 + }, + { + "epoch": 0.75, + "learning_rate": 3.755001282380098e-05, + "loss": 0.5632, + "step": 19467 + }, + { + "epoch": 0.75, + "learning_rate": 3.7549371633752244e-05, + "loss": 0.4769, + "step": 19468 + }, + { + "epoch": 0.75, + "learning_rate": 3.7548730443703516e-05, + "loss": 0.5159, + "step": 19469 + }, + { + "epoch": 0.75, + "learning_rate": 3.754808925365478e-05, + "loss": 0.5867, + "step": 19470 + }, + { + "epoch": 0.75, + "learning_rate": 3.754744806360605e-05, + "loss": 0.5241, + "step": 19471 + }, + { + "epoch": 0.75, + "learning_rate": 3.7546806873557324e-05, + "loss": 0.5914, + "step": 19472 + }, + { + "epoch": 0.75, + "learning_rate": 3.7546165683508596e-05, + "loss": 0.472, + "step": 19473 + }, + { + "epoch": 0.75, + "learning_rate": 3.754552449345987e-05, + "loss": 0.4947, + "step": 19474 + }, + { + "epoch": 0.75, + "learning_rate": 3.754488330341113e-05, + "loss": 0.54, + "step": 19475 + }, + { + "epoch": 0.75, + "learning_rate": 3.7544242113362404e-05, + "loss": 0.4701, + "step": 19476 + }, + { + "epoch": 0.75, + "learning_rate": 3.754360092331367e-05, + "loss": 0.4873, + "step": 19477 + }, + { + "epoch": 0.75, + "learning_rate": 3.754295973326494e-05, + "loss": 0.4646, + "step": 19478 + }, + { + "epoch": 0.75, + "learning_rate": 3.754231854321621e-05, + "loss": 0.4704, + "step": 19479 + }, + { + "epoch": 0.75, + "learning_rate": 3.7541677353167484e-05, + "loss": 0.4369, + "step": 19480 + }, + { + "epoch": 0.75, + "learning_rate": 3.754103616311875e-05, + "loss": 0.5874, + "step": 19481 + }, + { + "epoch": 0.75, + "learning_rate": 3.754039497307002e-05, + "loss": 0.557, + "step": 19482 + }, + { + "epoch": 0.75, + "learning_rate": 3.753975378302129e-05, + "loss": 0.6754, + "step": 19483 + }, + { + "epoch": 0.75, + "learning_rate": 3.753911259297256e-05, + "loss": 0.543, + "step": 19484 + }, + { + "epoch": 0.75, + "learning_rate": 3.753847140292383e-05, + "loss": 0.4881, + "step": 19485 + }, + { + "epoch": 0.75, + "learning_rate": 3.7537830212875094e-05, + "loss": 0.5251, + "step": 19486 + }, + { + "epoch": 0.75, + "learning_rate": 3.7537189022826366e-05, + "loss": 0.509, + "step": 19487 + }, + { + "epoch": 0.75, + "learning_rate": 3.753654783277764e-05, + "loss": 0.4032, + "step": 19488 + }, + { + "epoch": 0.75, + "learning_rate": 3.753590664272891e-05, + "loss": 0.5277, + "step": 19489 + }, + { + "epoch": 0.75, + "learning_rate": 3.7535265452680174e-05, + "loss": 0.5371, + "step": 19490 + }, + { + "epoch": 0.75, + "learning_rate": 3.7534624262631446e-05, + "loss": 0.4595, + "step": 19491 + }, + { + "epoch": 0.75, + "learning_rate": 3.753398307258272e-05, + "loss": 0.479, + "step": 19492 + }, + { + "epoch": 0.75, + "learning_rate": 3.753334188253398e-05, + "loss": 0.5136, + "step": 19493 + }, + { + "epoch": 0.75, + "learning_rate": 3.7532700692485254e-05, + "loss": 0.5234, + "step": 19494 + }, + { + "epoch": 0.75, + "learning_rate": 3.7532059502436526e-05, + "loss": 0.4835, + "step": 19495 + }, + { + "epoch": 0.75, + "learning_rate": 3.75314183123878e-05, + "loss": 0.4884, + "step": 19496 + }, + { + "epoch": 0.75, + "learning_rate": 3.753077712233906e-05, + "loss": 0.4793, + "step": 19497 + }, + { + "epoch": 0.75, + "learning_rate": 3.7530135932290334e-05, + "loss": 0.4601, + "step": 19498 + }, + { + "epoch": 0.75, + "learning_rate": 3.75294947422416e-05, + "loss": 0.4126, + "step": 19499 + }, + { + "epoch": 0.75, + "learning_rate": 3.752885355219287e-05, + "loss": 0.5577, + "step": 19500 + }, + { + "epoch": 0.75, + "learning_rate": 3.752821236214414e-05, + "loss": 0.5647, + "step": 19501 + }, + { + "epoch": 0.75, + "learning_rate": 3.752757117209541e-05, + "loss": 0.5845, + "step": 19502 + }, + { + "epoch": 0.75, + "learning_rate": 3.752692998204668e-05, + "loss": 0.4226, + "step": 19503 + }, + { + "epoch": 0.75, + "learning_rate": 3.752628879199795e-05, + "loss": 0.629, + "step": 19504 + }, + { + "epoch": 0.75, + "learning_rate": 3.752564760194922e-05, + "loss": 0.6077, + "step": 19505 + }, + { + "epoch": 0.75, + "learning_rate": 3.752500641190049e-05, + "loss": 0.5266, + "step": 19506 + }, + { + "epoch": 0.75, + "learning_rate": 3.752436522185176e-05, + "loss": 0.5539, + "step": 19507 + }, + { + "epoch": 0.75, + "learning_rate": 3.7523724031803025e-05, + "loss": 0.5561, + "step": 19508 + }, + { + "epoch": 0.75, + "learning_rate": 3.7523082841754296e-05, + "loss": 0.5152, + "step": 19509 + }, + { + "epoch": 0.75, + "learning_rate": 3.752244165170557e-05, + "loss": 0.5365, + "step": 19510 + }, + { + "epoch": 0.75, + "learning_rate": 3.752180046165684e-05, + "loss": 0.5734, + "step": 19511 + }, + { + "epoch": 0.75, + "learning_rate": 3.752115927160811e-05, + "loss": 0.4506, + "step": 19512 + }, + { + "epoch": 0.75, + "learning_rate": 3.7520518081559376e-05, + "loss": 0.5658, + "step": 19513 + }, + { + "epoch": 0.75, + "learning_rate": 3.751987689151065e-05, + "loss": 0.569, + "step": 19514 + }, + { + "epoch": 0.75, + "learning_rate": 3.751923570146191e-05, + "loss": 0.5103, + "step": 19515 + }, + { + "epoch": 0.75, + "learning_rate": 3.7518594511413185e-05, + "loss": 0.5685, + "step": 19516 + }, + { + "epoch": 0.75, + "learning_rate": 3.751795332136445e-05, + "loss": 0.4776, + "step": 19517 + }, + { + "epoch": 0.75, + "learning_rate": 3.751731213131572e-05, + "loss": 0.5406, + "step": 19518 + }, + { + "epoch": 0.75, + "learning_rate": 3.751667094126699e-05, + "loss": 0.4722, + "step": 19519 + }, + { + "epoch": 0.75, + "learning_rate": 3.7516029751218265e-05, + "loss": 0.5364, + "step": 19520 + }, + { + "epoch": 0.75, + "learning_rate": 3.7515388561169537e-05, + "loss": 0.4906, + "step": 19521 + }, + { + "epoch": 0.75, + "learning_rate": 3.75147473711208e-05, + "loss": 0.4181, + "step": 19522 + }, + { + "epoch": 0.75, + "learning_rate": 3.751410618107207e-05, + "loss": 0.5734, + "step": 19523 + }, + { + "epoch": 0.75, + "learning_rate": 3.751346499102334e-05, + "loss": 0.5179, + "step": 19524 + }, + { + "epoch": 0.75, + "learning_rate": 3.751282380097461e-05, + "loss": 0.5331, + "step": 19525 + }, + { + "epoch": 0.75, + "learning_rate": 3.7512182610925875e-05, + "loss": 0.5635, + "step": 19526 + }, + { + "epoch": 0.75, + "learning_rate": 3.7511541420877146e-05, + "loss": 0.521, + "step": 19527 + }, + { + "epoch": 0.75, + "learning_rate": 3.7510900230828425e-05, + "loss": 0.5534, + "step": 19528 + }, + { + "epoch": 0.75, + "learning_rate": 3.751025904077969e-05, + "loss": 0.4278, + "step": 19529 + }, + { + "epoch": 0.75, + "learning_rate": 3.750961785073096e-05, + "loss": 0.5294, + "step": 19530 + }, + { + "epoch": 0.75, + "learning_rate": 3.7508976660682227e-05, + "loss": 0.4944, + "step": 19531 + }, + { + "epoch": 0.75, + "learning_rate": 3.75083354706335e-05, + "loss": 0.4963, + "step": 19532 + }, + { + "epoch": 0.75, + "learning_rate": 3.750769428058476e-05, + "loss": 0.5936, + "step": 19533 + }, + { + "epoch": 0.75, + "learning_rate": 3.7507053090536035e-05, + "loss": 0.5432, + "step": 19534 + }, + { + "epoch": 0.75, + "learning_rate": 3.750641190048731e-05, + "loss": 0.5088, + "step": 19535 + }, + { + "epoch": 0.75, + "learning_rate": 3.750577071043858e-05, + "loss": 0.4608, + "step": 19536 + }, + { + "epoch": 0.75, + "learning_rate": 3.750512952038985e-05, + "loss": 0.4938, + "step": 19537 + }, + { + "epoch": 0.75, + "learning_rate": 3.7504488330341115e-05, + "loss": 0.4391, + "step": 19538 + }, + { + "epoch": 0.75, + "learning_rate": 3.750384714029239e-05, + "loss": 0.4591, + "step": 19539 + }, + { + "epoch": 0.75, + "learning_rate": 3.750320595024365e-05, + "loss": 0.434, + "step": 19540 + }, + { + "epoch": 0.75, + "learning_rate": 3.7502564760194923e-05, + "loss": 0.5299, + "step": 19541 + }, + { + "epoch": 0.75, + "learning_rate": 3.750192357014619e-05, + "loss": 0.526, + "step": 19542 + }, + { + "epoch": 0.75, + "learning_rate": 3.750128238009746e-05, + "loss": 0.5013, + "step": 19543 + }, + { + "epoch": 0.75, + "learning_rate": 3.750064119004873e-05, + "loss": 0.4705, + "step": 19544 + }, + { + "epoch": 0.75, + "learning_rate": 3.7500000000000003e-05, + "loss": 0.5892, + "step": 19545 + }, + { + "epoch": 0.75, + "learning_rate": 3.7499358809951275e-05, + "loss": 0.4808, + "step": 19546 + }, + { + "epoch": 0.75, + "learning_rate": 3.749871761990254e-05, + "loss": 0.6248, + "step": 19547 + }, + { + "epoch": 0.75, + "learning_rate": 3.749807642985381e-05, + "loss": 0.6032, + "step": 19548 + }, + { + "epoch": 0.75, + "learning_rate": 3.749743523980508e-05, + "loss": 0.438, + "step": 19549 + }, + { + "epoch": 0.75, + "learning_rate": 3.749679404975635e-05, + "loss": 0.4366, + "step": 19550 + }, + { + "epoch": 0.75, + "learning_rate": 3.749615285970762e-05, + "loss": 0.5428, + "step": 19551 + }, + { + "epoch": 0.75, + "learning_rate": 3.749551166965889e-05, + "loss": 0.6726, + "step": 19552 + }, + { + "epoch": 0.75, + "learning_rate": 3.749487047961016e-05, + "loss": 0.5066, + "step": 19553 + }, + { + "epoch": 0.75, + "learning_rate": 3.749422928956143e-05, + "loss": 0.5568, + "step": 19554 + }, + { + "epoch": 0.75, + "learning_rate": 3.74935880995127e-05, + "loss": 0.4748, + "step": 19555 + }, + { + "epoch": 0.75, + "learning_rate": 3.7492946909463965e-05, + "loss": 0.6059, + "step": 19556 + }, + { + "epoch": 0.75, + "learning_rate": 3.749230571941524e-05, + "loss": 0.5162, + "step": 19557 + }, + { + "epoch": 0.75, + "learning_rate": 3.74916645293665e-05, + "loss": 0.5557, + "step": 19558 + }, + { + "epoch": 0.75, + "learning_rate": 3.7491023339317774e-05, + "loss": 0.443, + "step": 19559 + }, + { + "epoch": 0.75, + "learning_rate": 3.7490382149269045e-05, + "loss": 0.4163, + "step": 19560 + }, + { + "epoch": 0.75, + "learning_rate": 3.748974095922032e-05, + "loss": 0.4154, + "step": 19561 + }, + { + "epoch": 0.75, + "learning_rate": 3.748909976917158e-05, + "loss": 0.5099, + "step": 19562 + }, + { + "epoch": 0.75, + "learning_rate": 3.7488458579122854e-05, + "loss": 0.5482, + "step": 19563 + }, + { + "epoch": 0.75, + "learning_rate": 3.7487817389074125e-05, + "loss": 0.4865, + "step": 19564 + }, + { + "epoch": 0.75, + "learning_rate": 3.748717619902539e-05, + "loss": 0.4581, + "step": 19565 + }, + { + "epoch": 0.75, + "learning_rate": 3.748653500897666e-05, + "loss": 0.5711, + "step": 19566 + }, + { + "epoch": 0.75, + "learning_rate": 3.7485893818927934e-05, + "loss": 0.4975, + "step": 19567 + }, + { + "epoch": 0.75, + "learning_rate": 3.7485252628879206e-05, + "loss": 0.4589, + "step": 19568 + }, + { + "epoch": 0.75, + "learning_rate": 3.748461143883047e-05, + "loss": 0.4225, + "step": 19569 + }, + { + "epoch": 0.75, + "learning_rate": 3.748397024878174e-05, + "loss": 0.4602, + "step": 19570 + }, + { + "epoch": 0.75, + "learning_rate": 3.748332905873301e-05, + "loss": 0.5015, + "step": 19571 + }, + { + "epoch": 0.75, + "learning_rate": 3.748268786868428e-05, + "loss": 0.5911, + "step": 19572 + }, + { + "epoch": 0.75, + "learning_rate": 3.748204667863555e-05, + "loss": 0.5568, + "step": 19573 + }, + { + "epoch": 0.75, + "learning_rate": 3.7481405488586816e-05, + "loss": 0.5378, + "step": 19574 + }, + { + "epoch": 0.75, + "learning_rate": 3.748076429853809e-05, + "loss": 0.4502, + "step": 19575 + }, + { + "epoch": 0.75, + "learning_rate": 3.748012310848936e-05, + "loss": 0.54, + "step": 19576 + }, + { + "epoch": 0.75, + "learning_rate": 3.747948191844063e-05, + "loss": 0.544, + "step": 19577 + }, + { + "epoch": 0.75, + "learning_rate": 3.7478840728391896e-05, + "loss": 0.4503, + "step": 19578 + }, + { + "epoch": 0.75, + "learning_rate": 3.747819953834317e-05, + "loss": 0.5919, + "step": 19579 + }, + { + "epoch": 0.75, + "learning_rate": 3.747755834829443e-05, + "loss": 0.49, + "step": 19580 + }, + { + "epoch": 0.75, + "learning_rate": 3.7476917158245704e-05, + "loss": 0.5311, + "step": 19581 + }, + { + "epoch": 0.75, + "learning_rate": 3.7476275968196976e-05, + "loss": 0.4682, + "step": 19582 + }, + { + "epoch": 0.75, + "learning_rate": 3.747563477814825e-05, + "loss": 0.612, + "step": 19583 + }, + { + "epoch": 0.75, + "learning_rate": 3.747499358809952e-05, + "loss": 0.4381, + "step": 19584 + }, + { + "epoch": 0.75, + "learning_rate": 3.7474352398050784e-05, + "loss": 0.5068, + "step": 19585 + }, + { + "epoch": 0.75, + "learning_rate": 3.7473711208002056e-05, + "loss": 0.5241, + "step": 19586 + }, + { + "epoch": 0.75, + "learning_rate": 3.747307001795332e-05, + "loss": 0.5863, + "step": 19587 + }, + { + "epoch": 0.75, + "learning_rate": 3.747242882790459e-05, + "loss": 0.4157, + "step": 19588 + }, + { + "epoch": 0.75, + "learning_rate": 3.747178763785586e-05, + "loss": 0.4245, + "step": 19589 + }, + { + "epoch": 0.75, + "learning_rate": 3.747114644780713e-05, + "loss": 0.4511, + "step": 19590 + }, + { + "epoch": 0.75, + "learning_rate": 3.74705052577584e-05, + "loss": 0.566, + "step": 19591 + }, + { + "epoch": 0.75, + "learning_rate": 3.746986406770967e-05, + "loss": 0.4542, + "step": 19592 + }, + { + "epoch": 0.75, + "learning_rate": 3.7469222877660944e-05, + "loss": 0.5298, + "step": 19593 + }, + { + "epoch": 0.75, + "learning_rate": 3.746858168761221e-05, + "loss": 0.468, + "step": 19594 + }, + { + "epoch": 0.75, + "learning_rate": 3.746794049756348e-05, + "loss": 0.5437, + "step": 19595 + }, + { + "epoch": 0.75, + "learning_rate": 3.7467299307514746e-05, + "loss": 0.495, + "step": 19596 + }, + { + "epoch": 0.75, + "learning_rate": 3.746665811746602e-05, + "loss": 0.5798, + "step": 19597 + }, + { + "epoch": 0.75, + "learning_rate": 3.746601692741728e-05, + "loss": 0.5509, + "step": 19598 + }, + { + "epoch": 0.75, + "learning_rate": 3.746537573736856e-05, + "loss": 0.5149, + "step": 19599 + }, + { + "epoch": 0.75, + "learning_rate": 3.746473454731983e-05, + "loss": 0.3982, + "step": 19600 + }, + { + "epoch": 0.75, + "learning_rate": 3.74640933572711e-05, + "loss": 0.4404, + "step": 19601 + }, + { + "epoch": 0.75, + "learning_rate": 3.746345216722237e-05, + "loss": 0.4684, + "step": 19602 + }, + { + "epoch": 0.75, + "learning_rate": 3.7462810977173634e-05, + "loss": 0.4922, + "step": 19603 + }, + { + "epoch": 0.75, + "learning_rate": 3.7462169787124906e-05, + "loss": 0.4262, + "step": 19604 + }, + { + "epoch": 0.75, + "learning_rate": 3.746152859707617e-05, + "loss": 0.6152, + "step": 19605 + }, + { + "epoch": 0.75, + "learning_rate": 3.746088740702744e-05, + "loss": 0.5033, + "step": 19606 + }, + { + "epoch": 0.75, + "learning_rate": 3.7460246216978714e-05, + "loss": 0.47, + "step": 19607 + }, + { + "epoch": 0.75, + "learning_rate": 3.7459605026929986e-05, + "loss": 0.4746, + "step": 19608 + }, + { + "epoch": 0.75, + "learning_rate": 3.745896383688126e-05, + "loss": 0.4193, + "step": 19609 + }, + { + "epoch": 0.75, + "learning_rate": 3.745832264683252e-05, + "loss": 0.5787, + "step": 19610 + }, + { + "epoch": 0.75, + "learning_rate": 3.7457681456783794e-05, + "loss": 0.558, + "step": 19611 + }, + { + "epoch": 0.75, + "learning_rate": 3.745704026673506e-05, + "loss": 0.4703, + "step": 19612 + }, + { + "epoch": 0.75, + "learning_rate": 3.745639907668633e-05, + "loss": 0.4492, + "step": 19613 + }, + { + "epoch": 0.75, + "learning_rate": 3.7455757886637596e-05, + "loss": 0.4493, + "step": 19614 + }, + { + "epoch": 0.75, + "learning_rate": 3.745511669658887e-05, + "loss": 0.5464, + "step": 19615 + }, + { + "epoch": 0.75, + "learning_rate": 3.745447550654014e-05, + "loss": 0.51, + "step": 19616 + }, + { + "epoch": 0.75, + "learning_rate": 3.745383431649141e-05, + "loss": 0.5881, + "step": 19617 + }, + { + "epoch": 0.75, + "learning_rate": 3.745319312644268e-05, + "loss": 0.5752, + "step": 19618 + }, + { + "epoch": 0.75, + "learning_rate": 3.745255193639395e-05, + "loss": 0.4536, + "step": 19619 + }, + { + "epoch": 0.75, + "learning_rate": 3.745191074634522e-05, + "loss": 0.5312, + "step": 19620 + }, + { + "epoch": 0.75, + "learning_rate": 3.7451269556296485e-05, + "loss": 0.4849, + "step": 19621 + }, + { + "epoch": 0.75, + "learning_rate": 3.7450628366247756e-05, + "loss": 0.5927, + "step": 19622 + }, + { + "epoch": 0.75, + "learning_rate": 3.744998717619903e-05, + "loss": 0.4108, + "step": 19623 + }, + { + "epoch": 0.75, + "learning_rate": 3.74493459861503e-05, + "loss": 0.6615, + "step": 19624 + }, + { + "epoch": 0.75, + "learning_rate": 3.7448704796101565e-05, + "loss": 0.4911, + "step": 19625 + }, + { + "epoch": 0.75, + "learning_rate": 3.7448063606052836e-05, + "loss": 0.395, + "step": 19626 + }, + { + "epoch": 0.75, + "learning_rate": 3.744742241600411e-05, + "loss": 0.491, + "step": 19627 + }, + { + "epoch": 0.75, + "learning_rate": 3.744678122595537e-05, + "loss": 0.6032, + "step": 19628 + }, + { + "epoch": 0.75, + "learning_rate": 3.7446140035906645e-05, + "loss": 0.5618, + "step": 19629 + }, + { + "epoch": 0.75, + "learning_rate": 3.744549884585791e-05, + "loss": 0.5056, + "step": 19630 + }, + { + "epoch": 0.75, + "learning_rate": 3.744485765580918e-05, + "loss": 0.5801, + "step": 19631 + }, + { + "epoch": 0.75, + "learning_rate": 3.744421646576045e-05, + "loss": 0.4237, + "step": 19632 + }, + { + "epoch": 0.75, + "learning_rate": 3.7443575275711725e-05, + "loss": 0.4612, + "step": 19633 + }, + { + "epoch": 0.75, + "learning_rate": 3.744293408566299e-05, + "loss": 0.504, + "step": 19634 + }, + { + "epoch": 0.75, + "learning_rate": 3.744229289561426e-05, + "loss": 0.5889, + "step": 19635 + }, + { + "epoch": 0.75, + "learning_rate": 3.744165170556553e-05, + "loss": 0.5193, + "step": 19636 + }, + { + "epoch": 0.75, + "learning_rate": 3.74410105155168e-05, + "loss": 0.4299, + "step": 19637 + }, + { + "epoch": 0.76, + "learning_rate": 3.744036932546807e-05, + "loss": 0.5095, + "step": 19638 + }, + { + "epoch": 0.76, + "learning_rate": 3.743972813541934e-05, + "loss": 0.5316, + "step": 19639 + }, + { + "epoch": 0.76, + "learning_rate": 3.743908694537061e-05, + "loss": 0.6207, + "step": 19640 + }, + { + "epoch": 0.76, + "learning_rate": 3.743844575532188e-05, + "loss": 0.4458, + "step": 19641 + }, + { + "epoch": 0.76, + "learning_rate": 3.743780456527315e-05, + "loss": 0.5189, + "step": 19642 + }, + { + "epoch": 0.76, + "learning_rate": 3.743716337522442e-05, + "loss": 0.575, + "step": 19643 + }, + { + "epoch": 0.76, + "learning_rate": 3.7436522185175687e-05, + "loss": 0.5432, + "step": 19644 + }, + { + "epoch": 0.76, + "learning_rate": 3.743588099512696e-05, + "loss": 0.4771, + "step": 19645 + }, + { + "epoch": 0.76, + "learning_rate": 3.743523980507822e-05, + "loss": 0.5634, + "step": 19646 + }, + { + "epoch": 0.76, + "learning_rate": 3.7434598615029495e-05, + "loss": 0.4664, + "step": 19647 + }, + { + "epoch": 0.76, + "learning_rate": 3.743395742498077e-05, + "loss": 0.4048, + "step": 19648 + }, + { + "epoch": 0.76, + "learning_rate": 3.743331623493204e-05, + "loss": 0.4932, + "step": 19649 + }, + { + "epoch": 0.76, + "learning_rate": 3.74326750448833e-05, + "loss": 0.5574, + "step": 19650 + }, + { + "epoch": 0.76, + "learning_rate": 3.7432033854834575e-05, + "loss": 0.5701, + "step": 19651 + }, + { + "epoch": 0.76, + "learning_rate": 3.743139266478585e-05, + "loss": 0.4362, + "step": 19652 + }, + { + "epoch": 0.76, + "learning_rate": 3.743075147473711e-05, + "loss": 0.4587, + "step": 19653 + }, + { + "epoch": 0.76, + "learning_rate": 3.7430110284688383e-05, + "loss": 0.6597, + "step": 19654 + }, + { + "epoch": 0.76, + "learning_rate": 3.7429469094639655e-05, + "loss": 0.4645, + "step": 19655 + }, + { + "epoch": 0.76, + "learning_rate": 3.742882790459093e-05, + "loss": 0.5143, + "step": 19656 + }, + { + "epoch": 0.76, + "learning_rate": 3.742818671454219e-05, + "loss": 0.5305, + "step": 19657 + }, + { + "epoch": 0.76, + "learning_rate": 3.7427545524493464e-05, + "loss": 0.4723, + "step": 19658 + }, + { + "epoch": 0.76, + "learning_rate": 3.742690433444473e-05, + "loss": 0.5453, + "step": 19659 + }, + { + "epoch": 0.76, + "learning_rate": 3.7426263144396e-05, + "loss": 0.6165, + "step": 19660 + }, + { + "epoch": 0.76, + "learning_rate": 3.742562195434727e-05, + "loss": 0.5081, + "step": 19661 + }, + { + "epoch": 0.76, + "learning_rate": 3.742498076429854e-05, + "loss": 0.516, + "step": 19662 + }, + { + "epoch": 0.76, + "learning_rate": 3.742433957424981e-05, + "loss": 0.6186, + "step": 19663 + }, + { + "epoch": 0.76, + "learning_rate": 3.742369838420108e-05, + "loss": 0.5242, + "step": 19664 + }, + { + "epoch": 0.76, + "learning_rate": 3.742305719415235e-05, + "loss": 0.5611, + "step": 19665 + }, + { + "epoch": 0.76, + "learning_rate": 3.742241600410362e-05, + "loss": 0.4102, + "step": 19666 + }, + { + "epoch": 0.76, + "learning_rate": 3.742177481405489e-05, + "loss": 0.6507, + "step": 19667 + }, + { + "epoch": 0.76, + "learning_rate": 3.7421133624006154e-05, + "loss": 0.5047, + "step": 19668 + }, + { + "epoch": 0.76, + "learning_rate": 3.7420492433957425e-05, + "loss": 0.4658, + "step": 19669 + }, + { + "epoch": 0.76, + "learning_rate": 3.74198512439087e-05, + "loss": 0.5989, + "step": 19670 + }, + { + "epoch": 0.76, + "learning_rate": 3.741921005385997e-05, + "loss": 0.5245, + "step": 19671 + }, + { + "epoch": 0.76, + "learning_rate": 3.741856886381124e-05, + "loss": 0.532, + "step": 19672 + }, + { + "epoch": 0.76, + "learning_rate": 3.7417927673762505e-05, + "loss": 0.5521, + "step": 19673 + }, + { + "epoch": 0.76, + "learning_rate": 3.741728648371378e-05, + "loss": 0.5279, + "step": 19674 + }, + { + "epoch": 0.76, + "learning_rate": 3.741664529366504e-05, + "loss": 0.568, + "step": 19675 + }, + { + "epoch": 0.76, + "learning_rate": 3.7416004103616314e-05, + "loss": 0.5192, + "step": 19676 + }, + { + "epoch": 0.76, + "learning_rate": 3.741536291356758e-05, + "loss": 0.5082, + "step": 19677 + }, + { + "epoch": 0.76, + "learning_rate": 3.741472172351885e-05, + "loss": 0.4844, + "step": 19678 + }, + { + "epoch": 0.76, + "learning_rate": 3.741408053347012e-05, + "loss": 0.5143, + "step": 19679 + }, + { + "epoch": 0.76, + "learning_rate": 3.7413439343421394e-05, + "loss": 0.4519, + "step": 19680 + }, + { + "epoch": 0.76, + "learning_rate": 3.7412798153372666e-05, + "loss": 0.4397, + "step": 19681 + }, + { + "epoch": 0.76, + "learning_rate": 3.741215696332393e-05, + "loss": 0.5847, + "step": 19682 + }, + { + "epoch": 0.76, + "learning_rate": 3.74115157732752e-05, + "loss": 0.5387, + "step": 19683 + }, + { + "epoch": 0.76, + "learning_rate": 3.741087458322647e-05, + "loss": 0.5976, + "step": 19684 + }, + { + "epoch": 0.76, + "learning_rate": 3.741023339317774e-05, + "loss": 0.4192, + "step": 19685 + }, + { + "epoch": 0.76, + "learning_rate": 3.7409592203129004e-05, + "loss": 0.4918, + "step": 19686 + }, + { + "epoch": 0.76, + "learning_rate": 3.740895101308028e-05, + "loss": 0.5389, + "step": 19687 + }, + { + "epoch": 0.76, + "learning_rate": 3.7408309823031554e-05, + "loss": 0.5591, + "step": 19688 + }, + { + "epoch": 0.76, + "learning_rate": 3.740766863298282e-05, + "loss": 0.4564, + "step": 19689 + }, + { + "epoch": 0.76, + "learning_rate": 3.740702744293409e-05, + "loss": 0.4407, + "step": 19690 + }, + { + "epoch": 0.76, + "learning_rate": 3.7406386252885356e-05, + "loss": 0.4477, + "step": 19691 + }, + { + "epoch": 0.76, + "learning_rate": 3.740574506283663e-05, + "loss": 0.5441, + "step": 19692 + }, + { + "epoch": 0.76, + "learning_rate": 3.740510387278789e-05, + "loss": 0.4914, + "step": 19693 + }, + { + "epoch": 0.76, + "learning_rate": 3.7404462682739164e-05, + "loss": 0.4048, + "step": 19694 + }, + { + "epoch": 0.76, + "learning_rate": 3.7403821492690436e-05, + "loss": 0.4777, + "step": 19695 + }, + { + "epoch": 0.76, + "learning_rate": 3.740318030264171e-05, + "loss": 0.5253, + "step": 19696 + }, + { + "epoch": 0.76, + "learning_rate": 3.740253911259298e-05, + "loss": 0.4838, + "step": 19697 + }, + { + "epoch": 0.76, + "learning_rate": 3.7401897922544244e-05, + "loss": 0.4442, + "step": 19698 + }, + { + "epoch": 0.76, + "learning_rate": 3.7401256732495516e-05, + "loss": 0.5363, + "step": 19699 + }, + { + "epoch": 0.76, + "learning_rate": 3.740061554244678e-05, + "loss": 0.5075, + "step": 19700 + }, + { + "epoch": 0.76, + "learning_rate": 3.739997435239805e-05, + "loss": 0.43, + "step": 19701 + }, + { + "epoch": 0.76, + "learning_rate": 3.739933316234932e-05, + "loss": 0.4709, + "step": 19702 + }, + { + "epoch": 0.76, + "learning_rate": 3.739869197230059e-05, + "loss": 0.4098, + "step": 19703 + }, + { + "epoch": 0.76, + "learning_rate": 3.739805078225186e-05, + "loss": 0.571, + "step": 19704 + }, + { + "epoch": 0.76, + "learning_rate": 3.739740959220313e-05, + "loss": 0.5873, + "step": 19705 + }, + { + "epoch": 0.76, + "learning_rate": 3.7396768402154404e-05, + "loss": 0.4522, + "step": 19706 + }, + { + "epoch": 0.76, + "learning_rate": 3.739612721210567e-05, + "loss": 0.6624, + "step": 19707 + }, + { + "epoch": 0.76, + "learning_rate": 3.739548602205694e-05, + "loss": 0.4467, + "step": 19708 + }, + { + "epoch": 0.76, + "learning_rate": 3.7394844832008206e-05, + "loss": 0.445, + "step": 19709 + }, + { + "epoch": 0.76, + "learning_rate": 3.739420364195948e-05, + "loss": 0.5665, + "step": 19710 + }, + { + "epoch": 0.76, + "learning_rate": 3.739356245191075e-05, + "loss": 0.4652, + "step": 19711 + }, + { + "epoch": 0.76, + "learning_rate": 3.739292126186202e-05, + "loss": 0.5818, + "step": 19712 + }, + { + "epoch": 0.76, + "learning_rate": 3.7392280071813286e-05, + "loss": 0.492, + "step": 19713 + }, + { + "epoch": 0.76, + "learning_rate": 3.739163888176456e-05, + "loss": 0.4284, + "step": 19714 + }, + { + "epoch": 0.76, + "learning_rate": 3.739099769171583e-05, + "loss": 0.5325, + "step": 19715 + }, + { + "epoch": 0.76, + "learning_rate": 3.7390356501667094e-05, + "loss": 0.3885, + "step": 19716 + }, + { + "epoch": 0.76, + "learning_rate": 3.7389715311618366e-05, + "loss": 0.544, + "step": 19717 + }, + { + "epoch": 0.76, + "learning_rate": 3.738907412156963e-05, + "loss": 0.4031, + "step": 19718 + }, + { + "epoch": 0.76, + "learning_rate": 3.73884329315209e-05, + "loss": 0.4801, + "step": 19719 + }, + { + "epoch": 0.76, + "learning_rate": 3.7387791741472174e-05, + "loss": 0.427, + "step": 19720 + }, + { + "epoch": 0.76, + "learning_rate": 3.7387150551423446e-05, + "loss": 0.641, + "step": 19721 + }, + { + "epoch": 0.76, + "learning_rate": 3.738650936137471e-05, + "loss": 0.4441, + "step": 19722 + }, + { + "epoch": 0.76, + "learning_rate": 3.738586817132598e-05, + "loss": 0.4787, + "step": 19723 + }, + { + "epoch": 0.76, + "learning_rate": 3.7385226981277255e-05, + "loss": 0.6555, + "step": 19724 + }, + { + "epoch": 0.76, + "learning_rate": 3.738458579122852e-05, + "loss": 0.4904, + "step": 19725 + }, + { + "epoch": 0.76, + "learning_rate": 3.738394460117979e-05, + "loss": 0.5052, + "step": 19726 + }, + { + "epoch": 0.76, + "learning_rate": 3.738330341113106e-05, + "loss": 0.5374, + "step": 19727 + }, + { + "epoch": 0.76, + "learning_rate": 3.7382662221082335e-05, + "loss": 0.4933, + "step": 19728 + }, + { + "epoch": 0.76, + "learning_rate": 3.73820210310336e-05, + "loss": 0.5587, + "step": 19729 + }, + { + "epoch": 0.76, + "learning_rate": 3.738137984098487e-05, + "loss": 0.4702, + "step": 19730 + }, + { + "epoch": 0.76, + "learning_rate": 3.7380738650936136e-05, + "loss": 0.5145, + "step": 19731 + }, + { + "epoch": 0.76, + "learning_rate": 3.738009746088741e-05, + "loss": 0.481, + "step": 19732 + }, + { + "epoch": 0.76, + "learning_rate": 3.737945627083868e-05, + "loss": 0.511, + "step": 19733 + }, + { + "epoch": 0.76, + "learning_rate": 3.7378815080789945e-05, + "loss": 0.5609, + "step": 19734 + }, + { + "epoch": 0.76, + "learning_rate": 3.7378173890741216e-05, + "loss": 0.497, + "step": 19735 + }, + { + "epoch": 0.76, + "learning_rate": 3.737753270069249e-05, + "loss": 0.4145, + "step": 19736 + }, + { + "epoch": 0.76, + "learning_rate": 3.737689151064376e-05, + "loss": 0.3814, + "step": 19737 + }, + { + "epoch": 0.76, + "learning_rate": 3.7376250320595025e-05, + "loss": 0.5568, + "step": 19738 + }, + { + "epoch": 0.76, + "learning_rate": 3.7375609130546296e-05, + "loss": 0.4815, + "step": 19739 + }, + { + "epoch": 0.76, + "learning_rate": 3.737496794049756e-05, + "loss": 0.5695, + "step": 19740 + }, + { + "epoch": 0.76, + "learning_rate": 3.737432675044883e-05, + "loss": 0.5137, + "step": 19741 + }, + { + "epoch": 0.76, + "learning_rate": 3.7373685560400105e-05, + "loss": 0.496, + "step": 19742 + }, + { + "epoch": 0.76, + "learning_rate": 3.7373044370351376e-05, + "loss": 0.4244, + "step": 19743 + }, + { + "epoch": 0.76, + "learning_rate": 3.737240318030265e-05, + "loss": 0.5355, + "step": 19744 + }, + { + "epoch": 0.76, + "learning_rate": 3.737176199025391e-05, + "loss": 0.4232, + "step": 19745 + }, + { + "epoch": 0.76, + "learning_rate": 3.7371120800205185e-05, + "loss": 0.4914, + "step": 19746 + }, + { + "epoch": 0.76, + "learning_rate": 3.737047961015645e-05, + "loss": 0.5079, + "step": 19747 + }, + { + "epoch": 0.76, + "learning_rate": 3.736983842010772e-05, + "loss": 0.4998, + "step": 19748 + }, + { + "epoch": 0.76, + "learning_rate": 3.7369197230058986e-05, + "loss": 0.4776, + "step": 19749 + }, + { + "epoch": 0.76, + "learning_rate": 3.736855604001026e-05, + "loss": 0.439, + "step": 19750 + }, + { + "epoch": 0.76, + "learning_rate": 3.736791484996153e-05, + "loss": 0.5784, + "step": 19751 + }, + { + "epoch": 0.76, + "learning_rate": 3.73672736599128e-05, + "loss": 0.4119, + "step": 19752 + }, + { + "epoch": 0.76, + "learning_rate": 3.736663246986407e-05, + "loss": 0.3826, + "step": 19753 + }, + { + "epoch": 0.76, + "learning_rate": 3.736599127981534e-05, + "loss": 0.4433, + "step": 19754 + }, + { + "epoch": 0.76, + "learning_rate": 3.736535008976661e-05, + "loss": 0.6012, + "step": 19755 + }, + { + "epoch": 0.76, + "learning_rate": 3.7364708899717875e-05, + "loss": 0.4036, + "step": 19756 + }, + { + "epoch": 0.76, + "learning_rate": 3.736406770966915e-05, + "loss": 0.4244, + "step": 19757 + }, + { + "epoch": 0.76, + "learning_rate": 3.736342651962041e-05, + "loss": 0.47, + "step": 19758 + }, + { + "epoch": 0.76, + "learning_rate": 3.736278532957169e-05, + "loss": 0.4912, + "step": 19759 + }, + { + "epoch": 0.76, + "learning_rate": 3.736214413952296e-05, + "loss": 0.4456, + "step": 19760 + }, + { + "epoch": 0.76, + "learning_rate": 3.736150294947423e-05, + "loss": 0.4872, + "step": 19761 + }, + { + "epoch": 0.76, + "learning_rate": 3.73608617594255e-05, + "loss": 0.5274, + "step": 19762 + }, + { + "epoch": 0.76, + "learning_rate": 3.736022056937676e-05, + "loss": 0.3844, + "step": 19763 + }, + { + "epoch": 0.76, + "learning_rate": 3.7359579379328035e-05, + "loss": 0.6617, + "step": 19764 + }, + { + "epoch": 0.76, + "learning_rate": 3.73589381892793e-05, + "loss": 0.6832, + "step": 19765 + }, + { + "epoch": 0.76, + "learning_rate": 3.735829699923057e-05, + "loss": 0.5646, + "step": 19766 + }, + { + "epoch": 0.76, + "learning_rate": 3.7357655809181843e-05, + "loss": 0.4467, + "step": 19767 + }, + { + "epoch": 0.76, + "learning_rate": 3.7357014619133115e-05, + "loss": 0.5008, + "step": 19768 + }, + { + "epoch": 0.76, + "learning_rate": 3.735637342908439e-05, + "loss": 0.5734, + "step": 19769 + }, + { + "epoch": 0.76, + "learning_rate": 3.735573223903565e-05, + "loss": 0.4635, + "step": 19770 + }, + { + "epoch": 0.76, + "learning_rate": 3.7355091048986924e-05, + "loss": 0.6942, + "step": 19771 + }, + { + "epoch": 0.76, + "learning_rate": 3.735444985893819e-05, + "loss": 0.5551, + "step": 19772 + }, + { + "epoch": 0.76, + "learning_rate": 3.735380866888946e-05, + "loss": 0.49, + "step": 19773 + }, + { + "epoch": 0.76, + "learning_rate": 3.7353167478840725e-05, + "loss": 0.5156, + "step": 19774 + }, + { + "epoch": 0.76, + "learning_rate": 3.7352526288792e-05, + "loss": 0.5283, + "step": 19775 + }, + { + "epoch": 0.76, + "learning_rate": 3.735188509874327e-05, + "loss": 0.4052, + "step": 19776 + }, + { + "epoch": 0.76, + "learning_rate": 3.735124390869454e-05, + "loss": 0.5264, + "step": 19777 + }, + { + "epoch": 0.76, + "learning_rate": 3.735060271864581e-05, + "loss": 0.5046, + "step": 19778 + }, + { + "epoch": 0.76, + "learning_rate": 3.734996152859708e-05, + "loss": 0.501, + "step": 19779 + }, + { + "epoch": 0.76, + "learning_rate": 3.734932033854835e-05, + "loss": 0.4878, + "step": 19780 + }, + { + "epoch": 0.76, + "learning_rate": 3.7348679148499614e-05, + "loss": 0.6089, + "step": 19781 + }, + { + "epoch": 0.76, + "learning_rate": 3.7348037958450885e-05, + "loss": 0.5883, + "step": 19782 + }, + { + "epoch": 0.76, + "learning_rate": 3.734739676840216e-05, + "loss": 0.5254, + "step": 19783 + }, + { + "epoch": 0.76, + "learning_rate": 3.734675557835343e-05, + "loss": 0.5556, + "step": 19784 + }, + { + "epoch": 0.76, + "learning_rate": 3.7346114388304694e-05, + "loss": 0.5637, + "step": 19785 + }, + { + "epoch": 0.76, + "learning_rate": 3.7345473198255965e-05, + "loss": 0.4875, + "step": 19786 + }, + { + "epoch": 0.76, + "learning_rate": 3.734483200820724e-05, + "loss": 0.4675, + "step": 19787 + }, + { + "epoch": 0.76, + "learning_rate": 3.73441908181585e-05, + "loss": 0.5692, + "step": 19788 + }, + { + "epoch": 0.76, + "learning_rate": 3.7343549628109774e-05, + "loss": 0.5221, + "step": 19789 + }, + { + "epoch": 0.76, + "learning_rate": 3.734290843806104e-05, + "loss": 0.5346, + "step": 19790 + }, + { + "epoch": 0.76, + "learning_rate": 3.734226724801231e-05, + "loss": 0.505, + "step": 19791 + }, + { + "epoch": 0.76, + "learning_rate": 3.734162605796358e-05, + "loss": 0.5564, + "step": 19792 + }, + { + "epoch": 0.76, + "learning_rate": 3.7340984867914854e-05, + "loss": 0.5676, + "step": 19793 + }, + { + "epoch": 0.76, + "learning_rate": 3.734034367786612e-05, + "loss": 0.4982, + "step": 19794 + }, + { + "epoch": 0.76, + "learning_rate": 3.733970248781739e-05, + "loss": 0.4465, + "step": 19795 + }, + { + "epoch": 0.76, + "learning_rate": 3.733906129776866e-05, + "loss": 0.522, + "step": 19796 + }, + { + "epoch": 0.76, + "learning_rate": 3.733842010771993e-05, + "loss": 0.4064, + "step": 19797 + }, + { + "epoch": 0.76, + "learning_rate": 3.73377789176712e-05, + "loss": 0.4554, + "step": 19798 + }, + { + "epoch": 0.76, + "learning_rate": 3.733713772762247e-05, + "loss": 0.4515, + "step": 19799 + }, + { + "epoch": 0.76, + "learning_rate": 3.733649653757374e-05, + "loss": 0.5619, + "step": 19800 + }, + { + "epoch": 0.76, + "learning_rate": 3.733585534752501e-05, + "loss": 0.4552, + "step": 19801 + }, + { + "epoch": 0.76, + "learning_rate": 3.733521415747628e-05, + "loss": 0.5517, + "step": 19802 + }, + { + "epoch": 0.76, + "learning_rate": 3.7334572967427544e-05, + "loss": 0.4076, + "step": 19803 + }, + { + "epoch": 0.76, + "learning_rate": 3.7333931777378816e-05, + "loss": 0.4823, + "step": 19804 + }, + { + "epoch": 0.76, + "learning_rate": 3.733329058733009e-05, + "loss": 0.6618, + "step": 19805 + }, + { + "epoch": 0.76, + "learning_rate": 3.733264939728135e-05, + "loss": 0.5465, + "step": 19806 + }, + { + "epoch": 0.76, + "learning_rate": 3.7332008207232624e-05, + "loss": 0.5078, + "step": 19807 + }, + { + "epoch": 0.76, + "learning_rate": 3.7331367017183896e-05, + "loss": 0.4651, + "step": 19808 + }, + { + "epoch": 0.76, + "learning_rate": 3.733072582713517e-05, + "loss": 0.4374, + "step": 19809 + }, + { + "epoch": 0.76, + "learning_rate": 3.733008463708643e-05, + "loss": 0.4718, + "step": 19810 + }, + { + "epoch": 0.76, + "learning_rate": 3.7329443447037704e-05, + "loss": 0.4922, + "step": 19811 + }, + { + "epoch": 0.76, + "learning_rate": 3.732880225698897e-05, + "loss": 0.5275, + "step": 19812 + }, + { + "epoch": 0.76, + "learning_rate": 3.732816106694024e-05, + "loss": 0.5339, + "step": 19813 + }, + { + "epoch": 0.76, + "learning_rate": 3.732751987689151e-05, + "loss": 0.4949, + "step": 19814 + }, + { + "epoch": 0.76, + "learning_rate": 3.7326878686842784e-05, + "loss": 0.4618, + "step": 19815 + }, + { + "epoch": 0.76, + "learning_rate": 3.7326237496794056e-05, + "loss": 0.4834, + "step": 19816 + }, + { + "epoch": 0.76, + "learning_rate": 3.732559630674532e-05, + "loss": 0.4582, + "step": 19817 + }, + { + "epoch": 0.76, + "learning_rate": 3.732495511669659e-05, + "loss": 0.4368, + "step": 19818 + }, + { + "epoch": 0.76, + "learning_rate": 3.732431392664786e-05, + "loss": 0.5448, + "step": 19819 + }, + { + "epoch": 0.76, + "learning_rate": 3.732367273659913e-05, + "loss": 0.4125, + "step": 19820 + }, + { + "epoch": 0.76, + "learning_rate": 3.7323031546550394e-05, + "loss": 0.5046, + "step": 19821 + }, + { + "epoch": 0.76, + "learning_rate": 3.7322390356501666e-05, + "loss": 0.5361, + "step": 19822 + }, + { + "epoch": 0.76, + "learning_rate": 3.732174916645294e-05, + "loss": 0.4488, + "step": 19823 + }, + { + "epoch": 0.76, + "learning_rate": 3.732110797640421e-05, + "loss": 0.4967, + "step": 19824 + }, + { + "epoch": 0.76, + "learning_rate": 3.732046678635548e-05, + "loss": 0.5332, + "step": 19825 + }, + { + "epoch": 0.76, + "learning_rate": 3.7319825596306746e-05, + "loss": 0.4465, + "step": 19826 + }, + { + "epoch": 0.76, + "learning_rate": 3.731918440625802e-05, + "loss": 0.3846, + "step": 19827 + }, + { + "epoch": 0.76, + "learning_rate": 3.731854321620928e-05, + "loss": 0.4724, + "step": 19828 + }, + { + "epoch": 0.76, + "learning_rate": 3.7317902026160554e-05, + "loss": 0.4581, + "step": 19829 + }, + { + "epoch": 0.76, + "learning_rate": 3.7317260836111826e-05, + "loss": 0.5, + "step": 19830 + }, + { + "epoch": 0.76, + "learning_rate": 3.73166196460631e-05, + "loss": 0.4556, + "step": 19831 + }, + { + "epoch": 0.76, + "learning_rate": 3.731597845601437e-05, + "loss": 0.4512, + "step": 19832 + }, + { + "epoch": 0.76, + "learning_rate": 3.7315337265965634e-05, + "loss": 0.4452, + "step": 19833 + }, + { + "epoch": 0.76, + "learning_rate": 3.7314696075916906e-05, + "loss": 0.4974, + "step": 19834 + }, + { + "epoch": 0.76, + "learning_rate": 3.731405488586817e-05, + "loss": 0.5558, + "step": 19835 + }, + { + "epoch": 0.76, + "learning_rate": 3.731341369581944e-05, + "loss": 0.4338, + "step": 19836 + }, + { + "epoch": 0.76, + "learning_rate": 3.731277250577071e-05, + "loss": 0.468, + "step": 19837 + }, + { + "epoch": 0.76, + "learning_rate": 3.731213131572198e-05, + "loss": 0.489, + "step": 19838 + }, + { + "epoch": 0.76, + "learning_rate": 3.731149012567325e-05, + "loss": 0.5111, + "step": 19839 + }, + { + "epoch": 0.76, + "learning_rate": 3.731084893562452e-05, + "loss": 0.5364, + "step": 19840 + }, + { + "epoch": 0.76, + "learning_rate": 3.7310207745575795e-05, + "loss": 0.5547, + "step": 19841 + }, + { + "epoch": 0.76, + "learning_rate": 3.730956655552706e-05, + "loss": 0.4433, + "step": 19842 + }, + { + "epoch": 0.76, + "learning_rate": 3.730892536547833e-05, + "loss": 0.5384, + "step": 19843 + }, + { + "epoch": 0.76, + "learning_rate": 3.7308284175429596e-05, + "loss": 0.4531, + "step": 19844 + }, + { + "epoch": 0.76, + "learning_rate": 3.730764298538087e-05, + "loss": 0.5682, + "step": 19845 + }, + { + "epoch": 0.76, + "learning_rate": 3.730700179533213e-05, + "loss": 0.621, + "step": 19846 + }, + { + "epoch": 0.76, + "learning_rate": 3.730636060528341e-05, + "loss": 0.4833, + "step": 19847 + }, + { + "epoch": 0.76, + "learning_rate": 3.730571941523468e-05, + "loss": 0.52, + "step": 19848 + }, + { + "epoch": 0.76, + "learning_rate": 3.730507822518595e-05, + "loss": 0.6008, + "step": 19849 + }, + { + "epoch": 0.76, + "learning_rate": 3.730443703513722e-05, + "loss": 0.5094, + "step": 19850 + }, + { + "epoch": 0.76, + "learning_rate": 3.7303795845088485e-05, + "loss": 0.5911, + "step": 19851 + }, + { + "epoch": 0.76, + "learning_rate": 3.7303154655039756e-05, + "loss": 0.4915, + "step": 19852 + }, + { + "epoch": 0.76, + "learning_rate": 3.730251346499102e-05, + "loss": 0.4299, + "step": 19853 + }, + { + "epoch": 0.76, + "learning_rate": 3.730187227494229e-05, + "loss": 0.396, + "step": 19854 + }, + { + "epoch": 0.76, + "learning_rate": 3.7301231084893565e-05, + "loss": 0.6205, + "step": 19855 + }, + { + "epoch": 0.76, + "learning_rate": 3.7300589894844837e-05, + "loss": 0.4358, + "step": 19856 + }, + { + "epoch": 0.76, + "learning_rate": 3.729994870479611e-05, + "loss": 0.4117, + "step": 19857 + }, + { + "epoch": 0.76, + "learning_rate": 3.729930751474737e-05, + "loss": 0.4538, + "step": 19858 + }, + { + "epoch": 0.76, + "learning_rate": 3.7298666324698645e-05, + "loss": 0.4962, + "step": 19859 + }, + { + "epoch": 0.76, + "learning_rate": 3.729802513464991e-05, + "loss": 0.445, + "step": 19860 + }, + { + "epoch": 0.76, + "learning_rate": 3.729738394460118e-05, + "loss": 0.4788, + "step": 19861 + }, + { + "epoch": 0.76, + "learning_rate": 3.7296742754552446e-05, + "loss": 0.4873, + "step": 19862 + }, + { + "epoch": 0.76, + "learning_rate": 3.729610156450372e-05, + "loss": 0.425, + "step": 19863 + }, + { + "epoch": 0.76, + "learning_rate": 3.729546037445499e-05, + "loss": 0.4475, + "step": 19864 + }, + { + "epoch": 0.76, + "learning_rate": 3.729481918440626e-05, + "loss": 0.5916, + "step": 19865 + }, + { + "epoch": 0.76, + "learning_rate": 3.729417799435753e-05, + "loss": 0.3873, + "step": 19866 + }, + { + "epoch": 0.76, + "learning_rate": 3.72935368043088e-05, + "loss": 0.5662, + "step": 19867 + }, + { + "epoch": 0.76, + "learning_rate": 3.729289561426007e-05, + "loss": 0.5504, + "step": 19868 + }, + { + "epoch": 0.76, + "learning_rate": 3.7292254424211335e-05, + "loss": 0.566, + "step": 19869 + }, + { + "epoch": 0.76, + "learning_rate": 3.729161323416261e-05, + "loss": 0.581, + "step": 19870 + }, + { + "epoch": 0.76, + "learning_rate": 3.729097204411388e-05, + "loss": 0.5868, + "step": 19871 + }, + { + "epoch": 0.76, + "learning_rate": 3.729033085406515e-05, + "loss": 0.4349, + "step": 19872 + }, + { + "epoch": 0.76, + "learning_rate": 3.7289689664016415e-05, + "loss": 0.3933, + "step": 19873 + }, + { + "epoch": 0.76, + "learning_rate": 3.728904847396769e-05, + "loss": 0.4434, + "step": 19874 + }, + { + "epoch": 0.76, + "learning_rate": 3.728840728391896e-05, + "loss": 0.5533, + "step": 19875 + }, + { + "epoch": 0.76, + "learning_rate": 3.7287766093870223e-05, + "loss": 0.4229, + "step": 19876 + }, + { + "epoch": 0.76, + "learning_rate": 3.7287124903821495e-05, + "loss": 0.5056, + "step": 19877 + }, + { + "epoch": 0.76, + "learning_rate": 3.728648371377276e-05, + "loss": 0.5789, + "step": 19878 + }, + { + "epoch": 0.76, + "learning_rate": 3.728584252372403e-05, + "loss": 0.5694, + "step": 19879 + }, + { + "epoch": 0.76, + "learning_rate": 3.7285201333675303e-05, + "loss": 0.5586, + "step": 19880 + }, + { + "epoch": 0.76, + "learning_rate": 3.7284560143626575e-05, + "loss": 0.555, + "step": 19881 + }, + { + "epoch": 0.76, + "learning_rate": 3.728391895357784e-05, + "loss": 0.484, + "step": 19882 + }, + { + "epoch": 0.76, + "learning_rate": 3.728327776352911e-05, + "loss": 0.4545, + "step": 19883 + }, + { + "epoch": 0.76, + "learning_rate": 3.7282636573480384e-05, + "loss": 0.5024, + "step": 19884 + }, + { + "epoch": 0.76, + "learning_rate": 3.728199538343165e-05, + "loss": 0.5486, + "step": 19885 + }, + { + "epoch": 0.76, + "learning_rate": 3.728135419338292e-05, + "loss": 0.5239, + "step": 19886 + }, + { + "epoch": 0.76, + "learning_rate": 3.728071300333419e-05, + "loss": 0.4512, + "step": 19887 + }, + { + "epoch": 0.76, + "learning_rate": 3.7280071813285464e-05, + "loss": 0.3787, + "step": 19888 + }, + { + "epoch": 0.76, + "learning_rate": 3.727943062323673e-05, + "loss": 0.5471, + "step": 19889 + }, + { + "epoch": 0.76, + "learning_rate": 3.7278789433188e-05, + "loss": 0.5514, + "step": 19890 + }, + { + "epoch": 0.76, + "learning_rate": 3.7278148243139265e-05, + "loss": 0.4267, + "step": 19891 + }, + { + "epoch": 0.76, + "learning_rate": 3.727750705309054e-05, + "loss": 0.5441, + "step": 19892 + }, + { + "epoch": 0.76, + "learning_rate": 3.727686586304181e-05, + "loss": 0.4776, + "step": 19893 + }, + { + "epoch": 0.76, + "learning_rate": 3.7276224672993074e-05, + "loss": 0.5344, + "step": 19894 + }, + { + "epoch": 0.76, + "learning_rate": 3.7275583482944345e-05, + "loss": 0.5306, + "step": 19895 + }, + { + "epoch": 0.76, + "learning_rate": 3.727494229289562e-05, + "loss": 0.4536, + "step": 19896 + }, + { + "epoch": 0.76, + "learning_rate": 3.727430110284689e-05, + "loss": 0.4787, + "step": 19897 + }, + { + "epoch": 0.77, + "learning_rate": 3.7273659912798154e-05, + "loss": 0.451, + "step": 19898 + }, + { + "epoch": 0.77, + "learning_rate": 3.7273018722749425e-05, + "loss": 0.575, + "step": 19899 + }, + { + "epoch": 0.77, + "learning_rate": 3.727237753270069e-05, + "loss": 0.5525, + "step": 19900 + }, + { + "epoch": 0.77, + "learning_rate": 3.727173634265196e-05, + "loss": 0.4533, + "step": 19901 + }, + { + "epoch": 0.77, + "learning_rate": 3.7271095152603234e-05, + "loss": 0.4863, + "step": 19902 + }, + { + "epoch": 0.77, + "learning_rate": 3.7270453962554506e-05, + "loss": 0.4289, + "step": 19903 + }, + { + "epoch": 0.77, + "learning_rate": 3.726981277250578e-05, + "loss": 0.5596, + "step": 19904 + }, + { + "epoch": 0.77, + "learning_rate": 3.726917158245704e-05, + "loss": 0.5665, + "step": 19905 + }, + { + "epoch": 0.77, + "learning_rate": 3.7268530392408314e-05, + "loss": 0.5164, + "step": 19906 + }, + { + "epoch": 0.77, + "learning_rate": 3.726788920235958e-05, + "loss": 0.5026, + "step": 19907 + }, + { + "epoch": 0.77, + "learning_rate": 3.726724801231085e-05, + "loss": 0.5725, + "step": 19908 + }, + { + "epoch": 0.77, + "learning_rate": 3.7266606822262116e-05, + "loss": 0.5348, + "step": 19909 + }, + { + "epoch": 0.77, + "learning_rate": 3.726596563221339e-05, + "loss": 0.4885, + "step": 19910 + }, + { + "epoch": 0.77, + "learning_rate": 3.726532444216466e-05, + "loss": 0.525, + "step": 19911 + }, + { + "epoch": 0.77, + "learning_rate": 3.726468325211593e-05, + "loss": 0.5719, + "step": 19912 + }, + { + "epoch": 0.77, + "learning_rate": 3.72640420620672e-05, + "loss": 0.4539, + "step": 19913 + }, + { + "epoch": 0.77, + "learning_rate": 3.726340087201847e-05, + "loss": 0.4927, + "step": 19914 + }, + { + "epoch": 0.77, + "learning_rate": 3.726275968196974e-05, + "loss": 0.5326, + "step": 19915 + }, + { + "epoch": 0.77, + "learning_rate": 3.7262118491921004e-05, + "loss": 0.532, + "step": 19916 + }, + { + "epoch": 0.77, + "learning_rate": 3.7261477301872276e-05, + "loss": 0.5071, + "step": 19917 + }, + { + "epoch": 0.77, + "learning_rate": 3.726083611182354e-05, + "loss": 0.53, + "step": 19918 + }, + { + "epoch": 0.77, + "learning_rate": 3.726019492177482e-05, + "loss": 0.5208, + "step": 19919 + }, + { + "epoch": 0.77, + "learning_rate": 3.725955373172609e-05, + "loss": 0.4805, + "step": 19920 + }, + { + "epoch": 0.77, + "learning_rate": 3.7258912541677356e-05, + "loss": 0.4775, + "step": 19921 + }, + { + "epoch": 0.77, + "learning_rate": 3.725827135162863e-05, + "loss": 0.5998, + "step": 19922 + }, + { + "epoch": 0.77, + "learning_rate": 3.725763016157989e-05, + "loss": 0.4767, + "step": 19923 + }, + { + "epoch": 0.77, + "learning_rate": 3.7256988971531164e-05, + "loss": 0.507, + "step": 19924 + }, + { + "epoch": 0.77, + "learning_rate": 3.725634778148243e-05, + "loss": 0.5176, + "step": 19925 + }, + { + "epoch": 0.77, + "learning_rate": 3.72557065914337e-05, + "loss": 0.5889, + "step": 19926 + }, + { + "epoch": 0.77, + "learning_rate": 3.725506540138497e-05, + "loss": 0.5137, + "step": 19927 + }, + { + "epoch": 0.77, + "learning_rate": 3.7254424211336244e-05, + "loss": 0.4986, + "step": 19928 + }, + { + "epoch": 0.77, + "learning_rate": 3.7253783021287516e-05, + "loss": 0.4496, + "step": 19929 + }, + { + "epoch": 0.77, + "learning_rate": 3.725314183123878e-05, + "loss": 0.5669, + "step": 19930 + }, + { + "epoch": 0.77, + "learning_rate": 3.725250064119005e-05, + "loss": 0.5296, + "step": 19931 + }, + { + "epoch": 0.77, + "learning_rate": 3.725185945114132e-05, + "loss": 0.3692, + "step": 19932 + }, + { + "epoch": 0.77, + "learning_rate": 3.725121826109259e-05, + "loss": 0.4278, + "step": 19933 + }, + { + "epoch": 0.77, + "learning_rate": 3.7250577071043854e-05, + "loss": 0.4403, + "step": 19934 + }, + { + "epoch": 0.77, + "learning_rate": 3.7249935880995126e-05, + "loss": 0.4052, + "step": 19935 + }, + { + "epoch": 0.77, + "learning_rate": 3.72492946909464e-05, + "loss": 0.512, + "step": 19936 + }, + { + "epoch": 0.77, + "learning_rate": 3.724865350089767e-05, + "loss": 0.4674, + "step": 19937 + }, + { + "epoch": 0.77, + "learning_rate": 3.724801231084894e-05, + "loss": 0.4454, + "step": 19938 + }, + { + "epoch": 0.77, + "learning_rate": 3.7247371120800206e-05, + "loss": 0.4164, + "step": 19939 + }, + { + "epoch": 0.77, + "learning_rate": 3.724672993075148e-05, + "loss": 0.5336, + "step": 19940 + }, + { + "epoch": 0.77, + "learning_rate": 3.724608874070274e-05, + "loss": 0.5744, + "step": 19941 + }, + { + "epoch": 0.77, + "learning_rate": 3.7245447550654014e-05, + "loss": 0.5346, + "step": 19942 + }, + { + "epoch": 0.77, + "learning_rate": 3.7244806360605286e-05, + "loss": 0.4888, + "step": 19943 + }, + { + "epoch": 0.77, + "learning_rate": 3.724416517055656e-05, + "loss": 0.4365, + "step": 19944 + }, + { + "epoch": 0.77, + "learning_rate": 3.724352398050782e-05, + "loss": 0.5186, + "step": 19945 + }, + { + "epoch": 0.77, + "learning_rate": 3.7242882790459094e-05, + "loss": 0.3841, + "step": 19946 + }, + { + "epoch": 0.77, + "learning_rate": 3.7242241600410366e-05, + "loss": 0.547, + "step": 19947 + }, + { + "epoch": 0.77, + "learning_rate": 3.724160041036163e-05, + "loss": 0.4208, + "step": 19948 + }, + { + "epoch": 0.77, + "learning_rate": 3.72409592203129e-05, + "loss": 0.6177, + "step": 19949 + }, + { + "epoch": 0.77, + "learning_rate": 3.724031803026417e-05, + "loss": 0.5155, + "step": 19950 + }, + { + "epoch": 0.77, + "learning_rate": 3.723967684021544e-05, + "loss": 0.4248, + "step": 19951 + }, + { + "epoch": 0.77, + "learning_rate": 3.723903565016671e-05, + "loss": 0.624, + "step": 19952 + }, + { + "epoch": 0.77, + "learning_rate": 3.723839446011798e-05, + "loss": 0.4127, + "step": 19953 + }, + { + "epoch": 0.77, + "learning_rate": 3.723775327006925e-05, + "loss": 0.4909, + "step": 19954 + }, + { + "epoch": 0.77, + "learning_rate": 3.723711208002052e-05, + "loss": 0.4906, + "step": 19955 + }, + { + "epoch": 0.77, + "learning_rate": 3.723647088997179e-05, + "loss": 0.4928, + "step": 19956 + }, + { + "epoch": 0.77, + "learning_rate": 3.7235829699923056e-05, + "loss": 0.3846, + "step": 19957 + }, + { + "epoch": 0.77, + "learning_rate": 3.723518850987433e-05, + "loss": 0.4423, + "step": 19958 + }, + { + "epoch": 0.77, + "learning_rate": 3.72345473198256e-05, + "loss": 0.5094, + "step": 19959 + }, + { + "epoch": 0.77, + "learning_rate": 3.723390612977687e-05, + "loss": 0.3993, + "step": 19960 + }, + { + "epoch": 0.77, + "learning_rate": 3.7233264939728136e-05, + "loss": 0.6069, + "step": 19961 + }, + { + "epoch": 0.77, + "learning_rate": 3.723262374967941e-05, + "loss": 0.5289, + "step": 19962 + }, + { + "epoch": 0.77, + "learning_rate": 3.723198255963067e-05, + "loss": 0.5269, + "step": 19963 + }, + { + "epoch": 0.77, + "learning_rate": 3.7231341369581945e-05, + "loss": 0.5788, + "step": 19964 + }, + { + "epoch": 0.77, + "learning_rate": 3.7230700179533216e-05, + "loss": 0.5572, + "step": 19965 + }, + { + "epoch": 0.77, + "learning_rate": 3.723005898948448e-05, + "loss": 0.6157, + "step": 19966 + }, + { + "epoch": 0.77, + "learning_rate": 3.722941779943575e-05, + "loss": 0.5183, + "step": 19967 + }, + { + "epoch": 0.77, + "learning_rate": 3.7228776609387025e-05, + "loss": 0.4886, + "step": 19968 + }, + { + "epoch": 0.77, + "learning_rate": 3.7228135419338297e-05, + "loss": 0.6488, + "step": 19969 + }, + { + "epoch": 0.77, + "learning_rate": 3.722749422928956e-05, + "loss": 0.418, + "step": 19970 + }, + { + "epoch": 0.77, + "learning_rate": 3.722685303924083e-05, + "loss": 0.4962, + "step": 19971 + }, + { + "epoch": 0.77, + "learning_rate": 3.72262118491921e-05, + "loss": 0.4568, + "step": 19972 + }, + { + "epoch": 0.77, + "learning_rate": 3.722557065914337e-05, + "loss": 0.4945, + "step": 19973 + }, + { + "epoch": 0.77, + "learning_rate": 3.722492946909464e-05, + "loss": 0.515, + "step": 19974 + }, + { + "epoch": 0.77, + "learning_rate": 3.722428827904591e-05, + "loss": 0.5032, + "step": 19975 + }, + { + "epoch": 0.77, + "learning_rate": 3.7223647088997185e-05, + "loss": 0.4068, + "step": 19976 + }, + { + "epoch": 0.77, + "learning_rate": 3.722300589894845e-05, + "loss": 0.565, + "step": 19977 + }, + { + "epoch": 0.77, + "learning_rate": 3.722236470889972e-05, + "loss": 0.6558, + "step": 19978 + }, + { + "epoch": 0.77, + "learning_rate": 3.7221723518850987e-05, + "loss": 0.5732, + "step": 19979 + }, + { + "epoch": 0.77, + "learning_rate": 3.722108232880226e-05, + "loss": 0.5087, + "step": 19980 + }, + { + "epoch": 0.77, + "learning_rate": 3.722044113875352e-05, + "loss": 0.4566, + "step": 19981 + }, + { + "epoch": 0.77, + "learning_rate": 3.7219799948704795e-05, + "loss": 0.4701, + "step": 19982 + }, + { + "epoch": 0.77, + "learning_rate": 3.721915875865607e-05, + "loss": 0.4825, + "step": 19983 + }, + { + "epoch": 0.77, + "learning_rate": 3.721851756860734e-05, + "loss": 0.5302, + "step": 19984 + }, + { + "epoch": 0.77, + "learning_rate": 3.721787637855861e-05, + "loss": 0.5416, + "step": 19985 + }, + { + "epoch": 0.77, + "learning_rate": 3.7217235188509875e-05, + "loss": 0.5723, + "step": 19986 + }, + { + "epoch": 0.77, + "learning_rate": 3.721659399846115e-05, + "loss": 0.4677, + "step": 19987 + }, + { + "epoch": 0.77, + "learning_rate": 3.721595280841241e-05, + "loss": 0.5722, + "step": 19988 + }, + { + "epoch": 0.77, + "learning_rate": 3.7215311618363683e-05, + "loss": 0.6627, + "step": 19989 + }, + { + "epoch": 0.77, + "learning_rate": 3.721467042831495e-05, + "loss": 0.601, + "step": 19990 + }, + { + "epoch": 0.77, + "learning_rate": 3.721402923826623e-05, + "loss": 0.4721, + "step": 19991 + }, + { + "epoch": 0.77, + "learning_rate": 3.72133880482175e-05, + "loss": 0.5487, + "step": 19992 + }, + { + "epoch": 0.77, + "learning_rate": 3.7212746858168764e-05, + "loss": 0.4497, + "step": 19993 + }, + { + "epoch": 0.77, + "learning_rate": 3.7212105668120035e-05, + "loss": 0.4212, + "step": 19994 + }, + { + "epoch": 0.77, + "learning_rate": 3.72114644780713e-05, + "loss": 0.5466, + "step": 19995 + }, + { + "epoch": 0.77, + "learning_rate": 3.721082328802257e-05, + "loss": 0.5169, + "step": 19996 + }, + { + "epoch": 0.77, + "learning_rate": 3.721018209797384e-05, + "loss": 0.5285, + "step": 19997 + }, + { + "epoch": 0.77, + "learning_rate": 3.720954090792511e-05, + "loss": 0.5086, + "step": 19998 + }, + { + "epoch": 0.77, + "learning_rate": 3.720889971787638e-05, + "loss": 0.541, + "step": 19999 + }, + { + "epoch": 0.77, + "learning_rate": 3.720825852782765e-05, + "loss": 0.5193, + "step": 20000 + }, + { + "epoch": 0.77, + "learning_rate": 3.7207617337778924e-05, + "loss": 0.5951, + "step": 20001 + }, + { + "epoch": 0.77, + "learning_rate": 3.720697614773019e-05, + "loss": 0.5549, + "step": 20002 + }, + { + "epoch": 0.77, + "learning_rate": 3.720633495768146e-05, + "loss": 0.564, + "step": 20003 + }, + { + "epoch": 0.77, + "learning_rate": 3.7205693767632725e-05, + "loss": 0.557, + "step": 20004 + }, + { + "epoch": 0.77, + "learning_rate": 3.7205052577584e-05, + "loss": 0.4955, + "step": 20005 + }, + { + "epoch": 0.77, + "learning_rate": 3.720441138753526e-05, + "loss": 0.4952, + "step": 20006 + }, + { + "epoch": 0.77, + "learning_rate": 3.720377019748654e-05, + "loss": 0.5043, + "step": 20007 + }, + { + "epoch": 0.77, + "learning_rate": 3.7203129007437805e-05, + "loss": 0.6426, + "step": 20008 + }, + { + "epoch": 0.77, + "learning_rate": 3.720248781738908e-05, + "loss": 0.459, + "step": 20009 + }, + { + "epoch": 0.77, + "learning_rate": 3.720184662734035e-05, + "loss": 0.4796, + "step": 20010 + }, + { + "epoch": 0.77, + "learning_rate": 3.7201205437291614e-05, + "loss": 0.4894, + "step": 20011 + }, + { + "epoch": 0.77, + "learning_rate": 3.7200564247242885e-05, + "loss": 0.5129, + "step": 20012 + }, + { + "epoch": 0.77, + "learning_rate": 3.719992305719415e-05, + "loss": 0.4606, + "step": 20013 + }, + { + "epoch": 0.77, + "learning_rate": 3.719928186714542e-05, + "loss": 0.5633, + "step": 20014 + }, + { + "epoch": 0.77, + "learning_rate": 3.7198640677096694e-05, + "loss": 0.5825, + "step": 20015 + }, + { + "epoch": 0.77, + "learning_rate": 3.7197999487047966e-05, + "loss": 0.5078, + "step": 20016 + }, + { + "epoch": 0.77, + "learning_rate": 3.719735829699923e-05, + "loss": 0.5226, + "step": 20017 + }, + { + "epoch": 0.77, + "learning_rate": 3.71967171069505e-05, + "loss": 0.5327, + "step": 20018 + }, + { + "epoch": 0.77, + "learning_rate": 3.7196075916901774e-05, + "loss": 0.386, + "step": 20019 + }, + { + "epoch": 0.77, + "learning_rate": 3.719543472685304e-05, + "loss": 0.5438, + "step": 20020 + }, + { + "epoch": 0.77, + "learning_rate": 3.719479353680431e-05, + "loss": 0.5493, + "step": 20021 + }, + { + "epoch": 0.77, + "learning_rate": 3.7194152346755576e-05, + "loss": 0.4648, + "step": 20022 + }, + { + "epoch": 0.77, + "learning_rate": 3.719351115670685e-05, + "loss": 0.5078, + "step": 20023 + }, + { + "epoch": 0.77, + "learning_rate": 3.719286996665812e-05, + "loss": 0.4553, + "step": 20024 + }, + { + "epoch": 0.77, + "learning_rate": 3.719222877660939e-05, + "loss": 0.4768, + "step": 20025 + }, + { + "epoch": 0.77, + "learning_rate": 3.719158758656066e-05, + "loss": 0.5444, + "step": 20026 + }, + { + "epoch": 0.77, + "learning_rate": 3.719094639651193e-05, + "loss": 0.4913, + "step": 20027 + }, + { + "epoch": 0.77, + "learning_rate": 3.71903052064632e-05, + "loss": 0.4435, + "step": 20028 + }, + { + "epoch": 0.77, + "learning_rate": 3.7189664016414464e-05, + "loss": 0.508, + "step": 20029 + }, + { + "epoch": 0.77, + "learning_rate": 3.7189022826365736e-05, + "loss": 0.5573, + "step": 20030 + }, + { + "epoch": 0.77, + "learning_rate": 3.718838163631701e-05, + "loss": 0.4706, + "step": 20031 + }, + { + "epoch": 0.77, + "learning_rate": 3.718774044626828e-05, + "loss": 0.5957, + "step": 20032 + }, + { + "epoch": 0.77, + "learning_rate": 3.7187099256219544e-05, + "loss": 0.4888, + "step": 20033 + }, + { + "epoch": 0.77, + "learning_rate": 3.7186458066170816e-05, + "loss": 0.5932, + "step": 20034 + }, + { + "epoch": 0.77, + "learning_rate": 3.718581687612209e-05, + "loss": 0.4279, + "step": 20035 + }, + { + "epoch": 0.77, + "learning_rate": 3.718517568607335e-05, + "loss": 0.5442, + "step": 20036 + }, + { + "epoch": 0.77, + "learning_rate": 3.7184534496024624e-05, + "loss": 0.3398, + "step": 20037 + }, + { + "epoch": 0.77, + "learning_rate": 3.718389330597589e-05, + "loss": 0.5084, + "step": 20038 + }, + { + "epoch": 0.77, + "learning_rate": 3.718325211592716e-05, + "loss": 0.4044, + "step": 20039 + }, + { + "epoch": 0.77, + "learning_rate": 3.718261092587843e-05, + "loss": 0.5552, + "step": 20040 + }, + { + "epoch": 0.77, + "learning_rate": 3.7181969735829704e-05, + "loss": 0.4619, + "step": 20041 + }, + { + "epoch": 0.77, + "learning_rate": 3.718132854578097e-05, + "loss": 0.4113, + "step": 20042 + }, + { + "epoch": 0.77, + "learning_rate": 3.718068735573224e-05, + "loss": 0.484, + "step": 20043 + }, + { + "epoch": 0.77, + "learning_rate": 3.718004616568351e-05, + "loss": 0.3618, + "step": 20044 + }, + { + "epoch": 0.77, + "learning_rate": 3.717940497563478e-05, + "loss": 0.324, + "step": 20045 + }, + { + "epoch": 0.77, + "learning_rate": 3.717876378558605e-05, + "loss": 0.5341, + "step": 20046 + }, + { + "epoch": 0.77, + "learning_rate": 3.717812259553732e-05, + "loss": 0.5326, + "step": 20047 + }, + { + "epoch": 0.77, + "learning_rate": 3.717748140548859e-05, + "loss": 0.4871, + "step": 20048 + }, + { + "epoch": 0.77, + "learning_rate": 3.717684021543986e-05, + "loss": 0.4301, + "step": 20049 + }, + { + "epoch": 0.77, + "learning_rate": 3.717619902539113e-05, + "loss": 0.645, + "step": 20050 + }, + { + "epoch": 0.77, + "learning_rate": 3.7175557835342394e-05, + "loss": 0.588, + "step": 20051 + }, + { + "epoch": 0.77, + "learning_rate": 3.7174916645293666e-05, + "loss": 0.5251, + "step": 20052 + }, + { + "epoch": 0.77, + "learning_rate": 3.717427545524494e-05, + "loss": 0.479, + "step": 20053 + }, + { + "epoch": 0.77, + "learning_rate": 3.71736342651962e-05, + "loss": 0.3765, + "step": 20054 + }, + { + "epoch": 0.77, + "learning_rate": 3.7172993075147474e-05, + "loss": 0.5242, + "step": 20055 + }, + { + "epoch": 0.77, + "learning_rate": 3.7172351885098746e-05, + "loss": 0.5987, + "step": 20056 + }, + { + "epoch": 0.77, + "learning_rate": 3.717171069505002e-05, + "loss": 0.5915, + "step": 20057 + }, + { + "epoch": 0.77, + "learning_rate": 3.717106950500128e-05, + "loss": 0.5296, + "step": 20058 + }, + { + "epoch": 0.77, + "learning_rate": 3.7170428314952555e-05, + "loss": 0.5208, + "step": 20059 + }, + { + "epoch": 0.77, + "learning_rate": 3.716978712490382e-05, + "loss": 0.5162, + "step": 20060 + }, + { + "epoch": 0.77, + "learning_rate": 3.716914593485509e-05, + "loss": 0.5079, + "step": 20061 + }, + { + "epoch": 0.77, + "learning_rate": 3.716850474480636e-05, + "loss": 0.5643, + "step": 20062 + }, + { + "epoch": 0.77, + "learning_rate": 3.7167863554757635e-05, + "loss": 0.5741, + "step": 20063 + }, + { + "epoch": 0.77, + "learning_rate": 3.7167222364708906e-05, + "loss": 0.5754, + "step": 20064 + }, + { + "epoch": 0.77, + "learning_rate": 3.716658117466017e-05, + "loss": 0.6107, + "step": 20065 + }, + { + "epoch": 0.77, + "learning_rate": 3.716593998461144e-05, + "loss": 0.5219, + "step": 20066 + }, + { + "epoch": 0.77, + "learning_rate": 3.716529879456271e-05, + "loss": 0.5009, + "step": 20067 + }, + { + "epoch": 0.77, + "learning_rate": 3.716465760451398e-05, + "loss": 0.5387, + "step": 20068 + }, + { + "epoch": 0.77, + "learning_rate": 3.7164016414465245e-05, + "loss": 0.5045, + "step": 20069 + }, + { + "epoch": 0.77, + "learning_rate": 3.7163375224416516e-05, + "loss": 0.5016, + "step": 20070 + }, + { + "epoch": 0.77, + "learning_rate": 3.716273403436779e-05, + "loss": 0.6426, + "step": 20071 + }, + { + "epoch": 0.77, + "learning_rate": 3.716209284431906e-05, + "loss": 0.4761, + "step": 20072 + }, + { + "epoch": 0.77, + "learning_rate": 3.716145165427033e-05, + "loss": 0.5416, + "step": 20073 + }, + { + "epoch": 0.77, + "learning_rate": 3.7160810464221596e-05, + "loss": 0.5294, + "step": 20074 + }, + { + "epoch": 0.77, + "learning_rate": 3.716016927417287e-05, + "loss": 0.4387, + "step": 20075 + }, + { + "epoch": 0.77, + "learning_rate": 3.715952808412413e-05, + "loss": 0.5698, + "step": 20076 + }, + { + "epoch": 0.77, + "learning_rate": 3.7158886894075405e-05, + "loss": 0.512, + "step": 20077 + }, + { + "epoch": 0.77, + "learning_rate": 3.715824570402667e-05, + "loss": 0.5518, + "step": 20078 + }, + { + "epoch": 0.77, + "learning_rate": 3.715760451397795e-05, + "loss": 0.4542, + "step": 20079 + }, + { + "epoch": 0.77, + "learning_rate": 3.715696332392922e-05, + "loss": 0.4229, + "step": 20080 + }, + { + "epoch": 0.77, + "learning_rate": 3.7156322133880485e-05, + "loss": 0.502, + "step": 20081 + }, + { + "epoch": 0.77, + "learning_rate": 3.7155680943831757e-05, + "loss": 0.4218, + "step": 20082 + }, + { + "epoch": 0.77, + "learning_rate": 3.715503975378302e-05, + "loss": 0.5636, + "step": 20083 + }, + { + "epoch": 0.77, + "learning_rate": 3.715439856373429e-05, + "loss": 0.4507, + "step": 20084 + }, + { + "epoch": 0.77, + "learning_rate": 3.715375737368556e-05, + "loss": 0.4931, + "step": 20085 + }, + { + "epoch": 0.77, + "learning_rate": 3.715311618363683e-05, + "loss": 0.5863, + "step": 20086 + }, + { + "epoch": 0.77, + "learning_rate": 3.71524749935881e-05, + "loss": 0.5919, + "step": 20087 + }, + { + "epoch": 0.77, + "learning_rate": 3.715183380353937e-05, + "loss": 0.5502, + "step": 20088 + }, + { + "epoch": 0.77, + "learning_rate": 3.7151192613490645e-05, + "loss": 0.4708, + "step": 20089 + }, + { + "epoch": 0.77, + "learning_rate": 3.715055142344191e-05, + "loss": 0.4963, + "step": 20090 + }, + { + "epoch": 0.77, + "learning_rate": 3.714991023339318e-05, + "loss": 0.5443, + "step": 20091 + }, + { + "epoch": 0.77, + "learning_rate": 3.714926904334445e-05, + "loss": 0.491, + "step": 20092 + }, + { + "epoch": 0.77, + "learning_rate": 3.714862785329572e-05, + "loss": 0.4633, + "step": 20093 + }, + { + "epoch": 0.77, + "learning_rate": 3.714798666324698e-05, + "loss": 0.3824, + "step": 20094 + }, + { + "epoch": 0.77, + "learning_rate": 3.714734547319826e-05, + "loss": 0.4856, + "step": 20095 + }, + { + "epoch": 0.77, + "learning_rate": 3.714670428314953e-05, + "loss": 0.5295, + "step": 20096 + }, + { + "epoch": 0.77, + "learning_rate": 3.71460630931008e-05, + "loss": 0.5131, + "step": 20097 + }, + { + "epoch": 0.77, + "learning_rate": 3.714542190305207e-05, + "loss": 0.5777, + "step": 20098 + }, + { + "epoch": 0.77, + "learning_rate": 3.7144780713003335e-05, + "loss": 0.4558, + "step": 20099 + }, + { + "epoch": 0.77, + "learning_rate": 3.714413952295461e-05, + "loss": 0.3795, + "step": 20100 + }, + { + "epoch": 0.77, + "learning_rate": 3.714349833290587e-05, + "loss": 0.4497, + "step": 20101 + }, + { + "epoch": 0.77, + "learning_rate": 3.7142857142857143e-05, + "loss": 0.4713, + "step": 20102 + }, + { + "epoch": 0.77, + "learning_rate": 3.7142215952808415e-05, + "loss": 0.5463, + "step": 20103 + }, + { + "epoch": 0.77, + "learning_rate": 3.714157476275969e-05, + "loss": 0.5548, + "step": 20104 + }, + { + "epoch": 0.77, + "learning_rate": 3.714093357271095e-05, + "loss": 0.428, + "step": 20105 + }, + { + "epoch": 0.77, + "learning_rate": 3.7140292382662224e-05, + "loss": 0.4982, + "step": 20106 + }, + { + "epoch": 0.77, + "learning_rate": 3.7139651192613495e-05, + "loss": 0.5586, + "step": 20107 + }, + { + "epoch": 0.77, + "learning_rate": 3.713901000256476e-05, + "loss": 0.5256, + "step": 20108 + }, + { + "epoch": 0.77, + "learning_rate": 3.713836881251603e-05, + "loss": 0.4383, + "step": 20109 + }, + { + "epoch": 0.77, + "learning_rate": 3.71377276224673e-05, + "loss": 0.532, + "step": 20110 + }, + { + "epoch": 0.77, + "learning_rate": 3.713708643241857e-05, + "loss": 0.4655, + "step": 20111 + }, + { + "epoch": 0.77, + "learning_rate": 3.713644524236984e-05, + "loss": 0.4134, + "step": 20112 + }, + { + "epoch": 0.77, + "learning_rate": 3.713580405232111e-05, + "loss": 0.4961, + "step": 20113 + }, + { + "epoch": 0.77, + "learning_rate": 3.713516286227238e-05, + "loss": 0.5117, + "step": 20114 + }, + { + "epoch": 0.77, + "learning_rate": 3.713452167222365e-05, + "loss": 0.5153, + "step": 20115 + }, + { + "epoch": 0.77, + "learning_rate": 3.713388048217492e-05, + "loss": 0.578, + "step": 20116 + }, + { + "epoch": 0.77, + "learning_rate": 3.7133239292126185e-05, + "loss": 0.5208, + "step": 20117 + }, + { + "epoch": 0.77, + "learning_rate": 3.713259810207746e-05, + "loss": 0.5443, + "step": 20118 + }, + { + "epoch": 0.77, + "learning_rate": 3.713195691202873e-05, + "loss": 0.5259, + "step": 20119 + }, + { + "epoch": 0.77, + "learning_rate": 3.713131572198e-05, + "loss": 0.4069, + "step": 20120 + }, + { + "epoch": 0.77, + "learning_rate": 3.7130674531931265e-05, + "loss": 0.4717, + "step": 20121 + }, + { + "epoch": 0.77, + "learning_rate": 3.713003334188254e-05, + "loss": 0.466, + "step": 20122 + }, + { + "epoch": 0.77, + "learning_rate": 3.71293921518338e-05, + "loss": 0.4455, + "step": 20123 + }, + { + "epoch": 0.77, + "learning_rate": 3.7128750961785074e-05, + "loss": 0.5922, + "step": 20124 + }, + { + "epoch": 0.77, + "learning_rate": 3.7128109771736346e-05, + "loss": 0.6085, + "step": 20125 + }, + { + "epoch": 0.77, + "learning_rate": 3.712746858168761e-05, + "loss": 0.51, + "step": 20126 + }, + { + "epoch": 0.77, + "learning_rate": 3.712682739163888e-05, + "loss": 0.4633, + "step": 20127 + }, + { + "epoch": 0.77, + "learning_rate": 3.7126186201590154e-05, + "loss": 0.5321, + "step": 20128 + }, + { + "epoch": 0.77, + "learning_rate": 3.7125545011541426e-05, + "loss": 0.4338, + "step": 20129 + }, + { + "epoch": 0.77, + "learning_rate": 3.712490382149269e-05, + "loss": 0.4209, + "step": 20130 + }, + { + "epoch": 0.77, + "learning_rate": 3.712426263144396e-05, + "loss": 0.6276, + "step": 20131 + }, + { + "epoch": 0.77, + "learning_rate": 3.712362144139523e-05, + "loss": 0.4405, + "step": 20132 + }, + { + "epoch": 0.77, + "learning_rate": 3.71229802513465e-05, + "loss": 0.4317, + "step": 20133 + }, + { + "epoch": 0.77, + "learning_rate": 3.712233906129777e-05, + "loss": 0.4837, + "step": 20134 + }, + { + "epoch": 0.77, + "learning_rate": 3.712169787124904e-05, + "loss": 0.4043, + "step": 20135 + }, + { + "epoch": 0.77, + "learning_rate": 3.7121056681200314e-05, + "loss": 0.4773, + "step": 20136 + }, + { + "epoch": 0.77, + "learning_rate": 3.712041549115158e-05, + "loss": 0.4605, + "step": 20137 + }, + { + "epoch": 0.77, + "learning_rate": 3.711977430110285e-05, + "loss": 0.4119, + "step": 20138 + }, + { + "epoch": 0.77, + "learning_rate": 3.7119133111054116e-05, + "loss": 0.5491, + "step": 20139 + }, + { + "epoch": 0.77, + "learning_rate": 3.711849192100539e-05, + "loss": 0.4581, + "step": 20140 + }, + { + "epoch": 0.77, + "learning_rate": 3.711785073095665e-05, + "loss": 0.4964, + "step": 20141 + }, + { + "epoch": 0.77, + "learning_rate": 3.7117209540907924e-05, + "loss": 0.5649, + "step": 20142 + }, + { + "epoch": 0.77, + "learning_rate": 3.7116568350859196e-05, + "loss": 0.4212, + "step": 20143 + }, + { + "epoch": 0.77, + "learning_rate": 3.711592716081047e-05, + "loss": 0.4517, + "step": 20144 + }, + { + "epoch": 0.77, + "learning_rate": 3.711528597076174e-05, + "loss": 0.4815, + "step": 20145 + }, + { + "epoch": 0.77, + "learning_rate": 3.7114644780713004e-05, + "loss": 0.5501, + "step": 20146 + }, + { + "epoch": 0.77, + "learning_rate": 3.7114003590664276e-05, + "loss": 0.5461, + "step": 20147 + }, + { + "epoch": 0.77, + "learning_rate": 3.711336240061554e-05, + "loss": 0.361, + "step": 20148 + }, + { + "epoch": 0.77, + "learning_rate": 3.711272121056681e-05, + "loss": 0.4065, + "step": 20149 + }, + { + "epoch": 0.77, + "learning_rate": 3.711208002051808e-05, + "loss": 0.4941, + "step": 20150 + }, + { + "epoch": 0.77, + "learning_rate": 3.7111438830469356e-05, + "loss": 0.5833, + "step": 20151 + }, + { + "epoch": 0.77, + "learning_rate": 3.711079764042063e-05, + "loss": 0.5519, + "step": 20152 + }, + { + "epoch": 0.77, + "learning_rate": 3.711015645037189e-05, + "loss": 0.4235, + "step": 20153 + }, + { + "epoch": 0.77, + "learning_rate": 3.7109515260323164e-05, + "loss": 0.5427, + "step": 20154 + }, + { + "epoch": 0.77, + "learning_rate": 3.710887407027443e-05, + "loss": 0.4727, + "step": 20155 + }, + { + "epoch": 0.77, + "learning_rate": 3.71082328802257e-05, + "loss": 0.5461, + "step": 20156 + }, + { + "epoch": 0.77, + "learning_rate": 3.7107591690176966e-05, + "loss": 0.4636, + "step": 20157 + }, + { + "epoch": 0.78, + "learning_rate": 3.710695050012824e-05, + "loss": 0.5774, + "step": 20158 + }, + { + "epoch": 0.78, + "learning_rate": 3.710630931007951e-05, + "loss": 0.5949, + "step": 20159 + }, + { + "epoch": 0.78, + "learning_rate": 3.710566812003078e-05, + "loss": 0.5731, + "step": 20160 + }, + { + "epoch": 0.78, + "learning_rate": 3.710502692998205e-05, + "loss": 0.5925, + "step": 20161 + }, + { + "epoch": 0.78, + "learning_rate": 3.710438573993332e-05, + "loss": 0.4778, + "step": 20162 + }, + { + "epoch": 0.78, + "learning_rate": 3.710374454988459e-05, + "loss": 0.5071, + "step": 20163 + }, + { + "epoch": 0.78, + "learning_rate": 3.7103103359835854e-05, + "loss": 0.404, + "step": 20164 + }, + { + "epoch": 0.78, + "learning_rate": 3.7102462169787126e-05, + "loss": 0.4068, + "step": 20165 + }, + { + "epoch": 0.78, + "learning_rate": 3.710182097973839e-05, + "loss": 0.4282, + "step": 20166 + }, + { + "epoch": 0.78, + "learning_rate": 3.710117978968967e-05, + "loss": 0.6162, + "step": 20167 + }, + { + "epoch": 0.78, + "learning_rate": 3.7100538599640934e-05, + "loss": 0.4927, + "step": 20168 + }, + { + "epoch": 0.78, + "learning_rate": 3.7099897409592206e-05, + "loss": 0.4412, + "step": 20169 + }, + { + "epoch": 0.78, + "learning_rate": 3.709925621954348e-05, + "loss": 0.561, + "step": 20170 + }, + { + "epoch": 0.78, + "learning_rate": 3.709861502949474e-05, + "loss": 0.6604, + "step": 20171 + }, + { + "epoch": 0.78, + "learning_rate": 3.7097973839446015e-05, + "loss": 0.5728, + "step": 20172 + }, + { + "epoch": 0.78, + "learning_rate": 3.709733264939728e-05, + "loss": 0.4575, + "step": 20173 + }, + { + "epoch": 0.78, + "learning_rate": 3.709669145934855e-05, + "loss": 0.5445, + "step": 20174 + }, + { + "epoch": 0.78, + "learning_rate": 3.709605026929982e-05, + "loss": 0.4618, + "step": 20175 + }, + { + "epoch": 0.78, + "learning_rate": 3.7095409079251095e-05, + "loss": 0.4286, + "step": 20176 + }, + { + "epoch": 0.78, + "learning_rate": 3.709476788920236e-05, + "loss": 0.6221, + "step": 20177 + }, + { + "epoch": 0.78, + "learning_rate": 3.709412669915363e-05, + "loss": 0.5518, + "step": 20178 + }, + { + "epoch": 0.78, + "learning_rate": 3.70934855091049e-05, + "loss": 0.6153, + "step": 20179 + }, + { + "epoch": 0.78, + "learning_rate": 3.709284431905617e-05, + "loss": 0.4571, + "step": 20180 + }, + { + "epoch": 0.78, + "learning_rate": 3.709220312900744e-05, + "loss": 0.4562, + "step": 20181 + }, + { + "epoch": 0.78, + "learning_rate": 3.7091561938958705e-05, + "loss": 0.5368, + "step": 20182 + }, + { + "epoch": 0.78, + "learning_rate": 3.7090920748909976e-05, + "loss": 0.5442, + "step": 20183 + }, + { + "epoch": 0.78, + "learning_rate": 3.709027955886125e-05, + "loss": 0.5134, + "step": 20184 + }, + { + "epoch": 0.78, + "learning_rate": 3.708963836881252e-05, + "loss": 0.4663, + "step": 20185 + }, + { + "epoch": 0.78, + "learning_rate": 3.7088997178763785e-05, + "loss": 0.4895, + "step": 20186 + }, + { + "epoch": 0.78, + "learning_rate": 3.7088355988715056e-05, + "loss": 0.5552, + "step": 20187 + }, + { + "epoch": 0.78, + "learning_rate": 3.708771479866633e-05, + "loss": 0.4696, + "step": 20188 + }, + { + "epoch": 0.78, + "learning_rate": 3.708707360861759e-05, + "loss": 0.424, + "step": 20189 + }, + { + "epoch": 0.78, + "learning_rate": 3.7086432418568865e-05, + "loss": 0.4856, + "step": 20190 + }, + { + "epoch": 0.78, + "learning_rate": 3.7085791228520137e-05, + "loss": 0.5504, + "step": 20191 + }, + { + "epoch": 0.78, + "learning_rate": 3.708515003847141e-05, + "loss": 0.4546, + "step": 20192 + }, + { + "epoch": 0.78, + "learning_rate": 3.708450884842267e-05, + "loss": 0.4784, + "step": 20193 + }, + { + "epoch": 0.78, + "learning_rate": 3.7083867658373945e-05, + "loss": 0.4797, + "step": 20194 + }, + { + "epoch": 0.78, + "learning_rate": 3.708322646832521e-05, + "loss": 0.4661, + "step": 20195 + }, + { + "epoch": 0.78, + "learning_rate": 3.708258527827648e-05, + "loss": 0.5332, + "step": 20196 + }, + { + "epoch": 0.78, + "learning_rate": 3.708194408822775e-05, + "loss": 0.6679, + "step": 20197 + }, + { + "epoch": 0.78, + "learning_rate": 3.708130289817902e-05, + "loss": 0.4403, + "step": 20198 + }, + { + "epoch": 0.78, + "learning_rate": 3.708066170813029e-05, + "loss": 0.3859, + "step": 20199 + }, + { + "epoch": 0.78, + "learning_rate": 3.708002051808156e-05, + "loss": 0.525, + "step": 20200 + }, + { + "epoch": 0.78, + "learning_rate": 3.707937932803283e-05, + "loss": 0.5041, + "step": 20201 + }, + { + "epoch": 0.78, + "learning_rate": 3.70787381379841e-05, + "loss": 0.6059, + "step": 20202 + }, + { + "epoch": 0.78, + "learning_rate": 3.707809694793537e-05, + "loss": 0.4545, + "step": 20203 + }, + { + "epoch": 0.78, + "learning_rate": 3.7077455757886635e-05, + "loss": 0.5247, + "step": 20204 + }, + { + "epoch": 0.78, + "learning_rate": 3.707681456783791e-05, + "loss": 0.4688, + "step": 20205 + }, + { + "epoch": 0.78, + "learning_rate": 3.707617337778918e-05, + "loss": 0.5259, + "step": 20206 + }, + { + "epoch": 0.78, + "learning_rate": 3.707553218774045e-05, + "loss": 0.701, + "step": 20207 + }, + { + "epoch": 0.78, + "learning_rate": 3.707489099769172e-05, + "loss": 0.545, + "step": 20208 + }, + { + "epoch": 0.78, + "learning_rate": 3.707424980764299e-05, + "loss": 0.4978, + "step": 20209 + }, + { + "epoch": 0.78, + "learning_rate": 3.707360861759426e-05, + "loss": 0.45, + "step": 20210 + }, + { + "epoch": 0.78, + "learning_rate": 3.7072967427545523e-05, + "loss": 0.4494, + "step": 20211 + }, + { + "epoch": 0.78, + "learning_rate": 3.7072326237496795e-05, + "loss": 0.5479, + "step": 20212 + }, + { + "epoch": 0.78, + "learning_rate": 3.707168504744807e-05, + "loss": 0.4264, + "step": 20213 + }, + { + "epoch": 0.78, + "learning_rate": 3.707104385739933e-05, + "loss": 0.3715, + "step": 20214 + }, + { + "epoch": 0.78, + "learning_rate": 3.7070402667350604e-05, + "loss": 0.4554, + "step": 20215 + }, + { + "epoch": 0.78, + "learning_rate": 3.7069761477301875e-05, + "loss": 0.4928, + "step": 20216 + }, + { + "epoch": 0.78, + "learning_rate": 3.706912028725315e-05, + "loss": 0.5198, + "step": 20217 + }, + { + "epoch": 0.78, + "learning_rate": 3.706847909720441e-05, + "loss": 0.5615, + "step": 20218 + }, + { + "epoch": 0.78, + "learning_rate": 3.7067837907155684e-05, + "loss": 0.4872, + "step": 20219 + }, + { + "epoch": 0.78, + "learning_rate": 3.706719671710695e-05, + "loss": 0.3993, + "step": 20220 + }, + { + "epoch": 0.78, + "learning_rate": 3.706655552705822e-05, + "loss": 0.6266, + "step": 20221 + }, + { + "epoch": 0.78, + "learning_rate": 3.706591433700949e-05, + "loss": 0.5464, + "step": 20222 + }, + { + "epoch": 0.78, + "learning_rate": 3.7065273146960764e-05, + "loss": 0.3829, + "step": 20223 + }, + { + "epoch": 0.78, + "learning_rate": 3.7064631956912035e-05, + "loss": 0.6988, + "step": 20224 + }, + { + "epoch": 0.78, + "learning_rate": 3.70639907668633e-05, + "loss": 0.3936, + "step": 20225 + }, + { + "epoch": 0.78, + "learning_rate": 3.706334957681457e-05, + "loss": 0.5308, + "step": 20226 + }, + { + "epoch": 0.78, + "learning_rate": 3.706270838676584e-05, + "loss": 0.3717, + "step": 20227 + }, + { + "epoch": 0.78, + "learning_rate": 3.706206719671711e-05, + "loss": 0.4797, + "step": 20228 + }, + { + "epoch": 0.78, + "learning_rate": 3.7061426006668374e-05, + "loss": 0.4816, + "step": 20229 + }, + { + "epoch": 0.78, + "learning_rate": 3.7060784816619645e-05, + "loss": 0.4982, + "step": 20230 + }, + { + "epoch": 0.78, + "learning_rate": 3.706014362657092e-05, + "loss": 0.4922, + "step": 20231 + }, + { + "epoch": 0.78, + "learning_rate": 3.705950243652219e-05, + "loss": 0.5609, + "step": 20232 + }, + { + "epoch": 0.78, + "learning_rate": 3.705886124647346e-05, + "loss": 0.4946, + "step": 20233 + }, + { + "epoch": 0.78, + "learning_rate": 3.7058220056424725e-05, + "loss": 0.5884, + "step": 20234 + }, + { + "epoch": 0.78, + "learning_rate": 3.7057578866376e-05, + "loss": 0.5551, + "step": 20235 + }, + { + "epoch": 0.78, + "learning_rate": 3.705693767632726e-05, + "loss": 0.5696, + "step": 20236 + }, + { + "epoch": 0.78, + "learning_rate": 3.7056296486278534e-05, + "loss": 0.4654, + "step": 20237 + }, + { + "epoch": 0.78, + "learning_rate": 3.70556552962298e-05, + "loss": 0.4439, + "step": 20238 + }, + { + "epoch": 0.78, + "learning_rate": 3.705501410618108e-05, + "loss": 0.4637, + "step": 20239 + }, + { + "epoch": 0.78, + "learning_rate": 3.705437291613235e-05, + "loss": 0.4365, + "step": 20240 + }, + { + "epoch": 0.78, + "learning_rate": 3.7053731726083614e-05, + "loss": 0.5327, + "step": 20241 + }, + { + "epoch": 0.78, + "learning_rate": 3.7053090536034886e-05, + "loss": 0.4885, + "step": 20242 + }, + { + "epoch": 0.78, + "learning_rate": 3.705244934598615e-05, + "loss": 0.5037, + "step": 20243 + }, + { + "epoch": 0.78, + "learning_rate": 3.705180815593742e-05, + "loss": 0.5554, + "step": 20244 + }, + { + "epoch": 0.78, + "learning_rate": 3.705116696588869e-05, + "loss": 0.4471, + "step": 20245 + }, + { + "epoch": 0.78, + "learning_rate": 3.705052577583996e-05, + "loss": 0.4648, + "step": 20246 + }, + { + "epoch": 0.78, + "learning_rate": 3.704988458579123e-05, + "loss": 0.5184, + "step": 20247 + }, + { + "epoch": 0.78, + "learning_rate": 3.70492433957425e-05, + "loss": 0.5741, + "step": 20248 + }, + { + "epoch": 0.78, + "learning_rate": 3.7048602205693774e-05, + "loss": 0.5719, + "step": 20249 + }, + { + "epoch": 0.78, + "learning_rate": 3.704796101564504e-05, + "loss": 0.4584, + "step": 20250 + }, + { + "epoch": 0.78, + "learning_rate": 3.704731982559631e-05, + "loss": 0.4241, + "step": 20251 + }, + { + "epoch": 0.78, + "learning_rate": 3.7046678635547576e-05, + "loss": 0.4442, + "step": 20252 + }, + { + "epoch": 0.78, + "learning_rate": 3.704603744549885e-05, + "loss": 0.523, + "step": 20253 + }, + { + "epoch": 0.78, + "learning_rate": 3.704539625545011e-05, + "loss": 0.5348, + "step": 20254 + }, + { + "epoch": 0.78, + "learning_rate": 3.704475506540139e-05, + "loss": 0.502, + "step": 20255 + }, + { + "epoch": 0.78, + "learning_rate": 3.7044113875352656e-05, + "loss": 0.6095, + "step": 20256 + }, + { + "epoch": 0.78, + "learning_rate": 3.704347268530393e-05, + "loss": 0.5478, + "step": 20257 + }, + { + "epoch": 0.78, + "learning_rate": 3.70428314952552e-05, + "loss": 0.5348, + "step": 20258 + }, + { + "epoch": 0.78, + "learning_rate": 3.7042190305206464e-05, + "loss": 0.5274, + "step": 20259 + }, + { + "epoch": 0.78, + "learning_rate": 3.7041549115157736e-05, + "loss": 0.5142, + "step": 20260 + }, + { + "epoch": 0.78, + "learning_rate": 3.7040907925109e-05, + "loss": 0.3954, + "step": 20261 + }, + { + "epoch": 0.78, + "learning_rate": 3.704026673506027e-05, + "loss": 0.5645, + "step": 20262 + }, + { + "epoch": 0.78, + "learning_rate": 3.7039625545011544e-05, + "loss": 0.5467, + "step": 20263 + }, + { + "epoch": 0.78, + "learning_rate": 3.7038984354962816e-05, + "loss": 0.4577, + "step": 20264 + }, + { + "epoch": 0.78, + "learning_rate": 3.703834316491408e-05, + "loss": 0.6004, + "step": 20265 + }, + { + "epoch": 0.78, + "learning_rate": 3.703770197486535e-05, + "loss": 0.4911, + "step": 20266 + }, + { + "epoch": 0.78, + "learning_rate": 3.7037060784816624e-05, + "loss": 0.5261, + "step": 20267 + }, + { + "epoch": 0.78, + "learning_rate": 3.703641959476789e-05, + "loss": 0.4577, + "step": 20268 + }, + { + "epoch": 0.78, + "learning_rate": 3.703577840471916e-05, + "loss": 0.5023, + "step": 20269 + }, + { + "epoch": 0.78, + "learning_rate": 3.7035137214670426e-05, + "loss": 0.5325, + "step": 20270 + }, + { + "epoch": 0.78, + "learning_rate": 3.70344960246217e-05, + "loss": 0.6051, + "step": 20271 + }, + { + "epoch": 0.78, + "learning_rate": 3.703385483457297e-05, + "loss": 0.4656, + "step": 20272 + }, + { + "epoch": 0.78, + "learning_rate": 3.703321364452424e-05, + "loss": 0.4858, + "step": 20273 + }, + { + "epoch": 0.78, + "learning_rate": 3.7032572454475506e-05, + "loss": 0.5489, + "step": 20274 + }, + { + "epoch": 0.78, + "learning_rate": 3.703193126442678e-05, + "loss": 0.4568, + "step": 20275 + }, + { + "epoch": 0.78, + "learning_rate": 3.703129007437805e-05, + "loss": 0.4162, + "step": 20276 + }, + { + "epoch": 0.78, + "learning_rate": 3.7030648884329314e-05, + "loss": 0.4431, + "step": 20277 + }, + { + "epoch": 0.78, + "learning_rate": 3.7030007694280586e-05, + "loss": 0.4539, + "step": 20278 + }, + { + "epoch": 0.78, + "learning_rate": 3.702936650423186e-05, + "loss": 0.5441, + "step": 20279 + }, + { + "epoch": 0.78, + "learning_rate": 3.702872531418313e-05, + "loss": 0.4741, + "step": 20280 + }, + { + "epoch": 0.78, + "learning_rate": 3.7028084124134395e-05, + "loss": 0.5486, + "step": 20281 + }, + { + "epoch": 0.78, + "learning_rate": 3.7027442934085666e-05, + "loss": 0.644, + "step": 20282 + }, + { + "epoch": 0.78, + "learning_rate": 3.702680174403693e-05, + "loss": 0.475, + "step": 20283 + }, + { + "epoch": 0.78, + "learning_rate": 3.70261605539882e-05, + "loss": 0.5835, + "step": 20284 + }, + { + "epoch": 0.78, + "learning_rate": 3.7025519363939475e-05, + "loss": 0.4723, + "step": 20285 + }, + { + "epoch": 0.78, + "learning_rate": 3.702487817389074e-05, + "loss": 0.6499, + "step": 20286 + }, + { + "epoch": 0.78, + "learning_rate": 3.702423698384201e-05, + "loss": 0.4918, + "step": 20287 + }, + { + "epoch": 0.78, + "learning_rate": 3.702359579379328e-05, + "loss": 0.4724, + "step": 20288 + }, + { + "epoch": 0.78, + "learning_rate": 3.7022954603744555e-05, + "loss": 0.4044, + "step": 20289 + }, + { + "epoch": 0.78, + "learning_rate": 3.702231341369582e-05, + "loss": 0.5197, + "step": 20290 + }, + { + "epoch": 0.78, + "learning_rate": 3.702167222364709e-05, + "loss": 0.5416, + "step": 20291 + }, + { + "epoch": 0.78, + "learning_rate": 3.7021031033598356e-05, + "loss": 0.5329, + "step": 20292 + }, + { + "epoch": 0.78, + "learning_rate": 3.702038984354963e-05, + "loss": 0.5753, + "step": 20293 + }, + { + "epoch": 0.78, + "learning_rate": 3.70197486535009e-05, + "loss": 0.5714, + "step": 20294 + }, + { + "epoch": 0.78, + "learning_rate": 3.701910746345217e-05, + "loss": 0.597, + "step": 20295 + }, + { + "epoch": 0.78, + "learning_rate": 3.701846627340344e-05, + "loss": 0.5067, + "step": 20296 + }, + { + "epoch": 0.78, + "learning_rate": 3.701782508335471e-05, + "loss": 0.5362, + "step": 20297 + }, + { + "epoch": 0.78, + "learning_rate": 3.701718389330598e-05, + "loss": 0.5675, + "step": 20298 + }, + { + "epoch": 0.78, + "learning_rate": 3.7016542703257245e-05, + "loss": 0.543, + "step": 20299 + }, + { + "epoch": 0.78, + "learning_rate": 3.7015901513208516e-05, + "loss": 0.4209, + "step": 20300 + }, + { + "epoch": 0.78, + "learning_rate": 3.701526032315978e-05, + "loss": 0.5399, + "step": 20301 + }, + { + "epoch": 0.78, + "learning_rate": 3.701461913311105e-05, + "loss": 0.5588, + "step": 20302 + }, + { + "epoch": 0.78, + "learning_rate": 3.7013977943062325e-05, + "loss": 0.452, + "step": 20303 + }, + { + "epoch": 0.78, + "learning_rate": 3.7013336753013597e-05, + "loss": 0.5088, + "step": 20304 + }, + { + "epoch": 0.78, + "learning_rate": 3.701269556296487e-05, + "loss": 0.4975, + "step": 20305 + }, + { + "epoch": 0.78, + "learning_rate": 3.701205437291613e-05, + "loss": 0.4097, + "step": 20306 + }, + { + "epoch": 0.78, + "learning_rate": 3.7011413182867405e-05, + "loss": 0.5348, + "step": 20307 + }, + { + "epoch": 0.78, + "learning_rate": 3.701077199281867e-05, + "loss": 0.4716, + "step": 20308 + }, + { + "epoch": 0.78, + "learning_rate": 3.701013080276994e-05, + "loss": 0.4764, + "step": 20309 + }, + { + "epoch": 0.78, + "learning_rate": 3.7009489612721207e-05, + "loss": 0.5454, + "step": 20310 + }, + { + "epoch": 0.78, + "learning_rate": 3.7008848422672485e-05, + "loss": 0.739, + "step": 20311 + }, + { + "epoch": 0.78, + "learning_rate": 3.700820723262376e-05, + "loss": 0.4795, + "step": 20312 + }, + { + "epoch": 0.78, + "learning_rate": 3.700756604257502e-05, + "loss": 0.5044, + "step": 20313 + }, + { + "epoch": 0.78, + "learning_rate": 3.7006924852526293e-05, + "loss": 0.4911, + "step": 20314 + }, + { + "epoch": 0.78, + "learning_rate": 3.700628366247756e-05, + "loss": 0.5199, + "step": 20315 + }, + { + "epoch": 0.78, + "learning_rate": 3.700564247242883e-05, + "loss": 0.7126, + "step": 20316 + }, + { + "epoch": 0.78, + "learning_rate": 3.7005001282380095e-05, + "loss": 0.372, + "step": 20317 + }, + { + "epoch": 0.78, + "learning_rate": 3.700436009233137e-05, + "loss": 0.4098, + "step": 20318 + }, + { + "epoch": 0.78, + "learning_rate": 3.700371890228264e-05, + "loss": 0.4157, + "step": 20319 + }, + { + "epoch": 0.78, + "learning_rate": 3.700307771223391e-05, + "loss": 0.4858, + "step": 20320 + }, + { + "epoch": 0.78, + "learning_rate": 3.700243652218518e-05, + "loss": 0.3747, + "step": 20321 + }, + { + "epoch": 0.78, + "learning_rate": 3.700179533213645e-05, + "loss": 0.5451, + "step": 20322 + }, + { + "epoch": 0.78, + "learning_rate": 3.700115414208772e-05, + "loss": 0.4943, + "step": 20323 + }, + { + "epoch": 0.78, + "learning_rate": 3.7000512952038983e-05, + "loss": 0.5474, + "step": 20324 + }, + { + "epoch": 0.78, + "learning_rate": 3.6999871761990255e-05, + "loss": 0.6053, + "step": 20325 + }, + { + "epoch": 0.78, + "learning_rate": 3.699923057194152e-05, + "loss": 0.6076, + "step": 20326 + }, + { + "epoch": 0.78, + "learning_rate": 3.69985893818928e-05, + "loss": 0.5303, + "step": 20327 + }, + { + "epoch": 0.78, + "learning_rate": 3.6997948191844064e-05, + "loss": 0.4962, + "step": 20328 + }, + { + "epoch": 0.78, + "learning_rate": 3.6997307001795335e-05, + "loss": 0.4239, + "step": 20329 + }, + { + "epoch": 0.78, + "learning_rate": 3.699666581174661e-05, + "loss": 0.5834, + "step": 20330 + }, + { + "epoch": 0.78, + "learning_rate": 3.699602462169787e-05, + "loss": 0.4674, + "step": 20331 + }, + { + "epoch": 0.78, + "learning_rate": 3.6995383431649144e-05, + "loss": 0.5815, + "step": 20332 + }, + { + "epoch": 0.78, + "learning_rate": 3.699474224160041e-05, + "loss": 0.5648, + "step": 20333 + }, + { + "epoch": 0.78, + "learning_rate": 3.699410105155168e-05, + "loss": 0.4462, + "step": 20334 + }, + { + "epoch": 0.78, + "learning_rate": 3.699345986150295e-05, + "loss": 0.5258, + "step": 20335 + }, + { + "epoch": 0.78, + "learning_rate": 3.6992818671454224e-05, + "loss": 0.5658, + "step": 20336 + }, + { + "epoch": 0.78, + "learning_rate": 3.699217748140549e-05, + "loss": 0.3995, + "step": 20337 + }, + { + "epoch": 0.78, + "learning_rate": 3.699153629135676e-05, + "loss": 0.5455, + "step": 20338 + }, + { + "epoch": 0.78, + "learning_rate": 3.699089510130803e-05, + "loss": 0.585, + "step": 20339 + }, + { + "epoch": 0.78, + "learning_rate": 3.69902539112593e-05, + "loss": 0.4367, + "step": 20340 + }, + { + "epoch": 0.78, + "learning_rate": 3.698961272121057e-05, + "loss": 0.578, + "step": 20341 + }, + { + "epoch": 0.78, + "learning_rate": 3.6988971531161834e-05, + "loss": 0.5389, + "step": 20342 + }, + { + "epoch": 0.78, + "learning_rate": 3.6988330341113105e-05, + "loss": 0.5173, + "step": 20343 + }, + { + "epoch": 0.78, + "learning_rate": 3.698768915106438e-05, + "loss": 0.5538, + "step": 20344 + }, + { + "epoch": 0.78, + "learning_rate": 3.698704796101565e-05, + "loss": 0.3962, + "step": 20345 + }, + { + "epoch": 0.78, + "learning_rate": 3.6986406770966914e-05, + "loss": 0.5844, + "step": 20346 + }, + { + "epoch": 0.78, + "learning_rate": 3.6985765580918185e-05, + "loss": 0.4038, + "step": 20347 + }, + { + "epoch": 0.78, + "learning_rate": 3.698512439086946e-05, + "loss": 0.4923, + "step": 20348 + }, + { + "epoch": 0.78, + "learning_rate": 3.698448320082072e-05, + "loss": 0.4611, + "step": 20349 + }, + { + "epoch": 0.78, + "learning_rate": 3.6983842010771994e-05, + "loss": 0.5875, + "step": 20350 + }, + { + "epoch": 0.78, + "learning_rate": 3.6983200820723266e-05, + "loss": 0.5056, + "step": 20351 + }, + { + "epoch": 0.78, + "learning_rate": 3.698255963067454e-05, + "loss": 0.5078, + "step": 20352 + }, + { + "epoch": 0.78, + "learning_rate": 3.69819184406258e-05, + "loss": 0.4208, + "step": 20353 + }, + { + "epoch": 0.78, + "learning_rate": 3.6981277250577074e-05, + "loss": 0.5338, + "step": 20354 + }, + { + "epoch": 0.78, + "learning_rate": 3.698063606052834e-05, + "loss": 0.4015, + "step": 20355 + }, + { + "epoch": 0.78, + "learning_rate": 3.697999487047961e-05, + "loss": 0.5843, + "step": 20356 + }, + { + "epoch": 0.78, + "learning_rate": 3.697935368043088e-05, + "loss": 0.4752, + "step": 20357 + }, + { + "epoch": 0.78, + "learning_rate": 3.697871249038215e-05, + "loss": 0.4617, + "step": 20358 + }, + { + "epoch": 0.78, + "learning_rate": 3.697807130033342e-05, + "loss": 0.4942, + "step": 20359 + }, + { + "epoch": 0.78, + "learning_rate": 3.697743011028469e-05, + "loss": 0.6092, + "step": 20360 + }, + { + "epoch": 0.78, + "learning_rate": 3.697678892023596e-05, + "loss": 0.4207, + "step": 20361 + }, + { + "epoch": 0.78, + "learning_rate": 3.697614773018723e-05, + "loss": 0.5197, + "step": 20362 + }, + { + "epoch": 0.78, + "learning_rate": 3.69755065401385e-05, + "loss": 0.5674, + "step": 20363 + }, + { + "epoch": 0.78, + "learning_rate": 3.6974865350089764e-05, + "loss": 0.4314, + "step": 20364 + }, + { + "epoch": 0.78, + "learning_rate": 3.6974224160041036e-05, + "loss": 0.4866, + "step": 20365 + }, + { + "epoch": 0.78, + "learning_rate": 3.697358296999231e-05, + "loss": 0.5786, + "step": 20366 + }, + { + "epoch": 0.78, + "learning_rate": 3.697294177994358e-05, + "loss": 0.4984, + "step": 20367 + }, + { + "epoch": 0.78, + "learning_rate": 3.697230058989485e-05, + "loss": 0.4966, + "step": 20368 + }, + { + "epoch": 0.78, + "learning_rate": 3.6971659399846116e-05, + "loss": 0.4994, + "step": 20369 + }, + { + "epoch": 0.78, + "learning_rate": 3.697101820979739e-05, + "loss": 0.4695, + "step": 20370 + }, + { + "epoch": 0.78, + "learning_rate": 3.697037701974865e-05, + "loss": 0.5256, + "step": 20371 + }, + { + "epoch": 0.78, + "learning_rate": 3.6969735829699924e-05, + "loss": 0.4322, + "step": 20372 + }, + { + "epoch": 0.78, + "learning_rate": 3.696909463965119e-05, + "loss": 0.6301, + "step": 20373 + }, + { + "epoch": 0.78, + "learning_rate": 3.696845344960246e-05, + "loss": 0.4033, + "step": 20374 + }, + { + "epoch": 0.78, + "learning_rate": 3.696781225955373e-05, + "loss": 0.4675, + "step": 20375 + }, + { + "epoch": 0.78, + "learning_rate": 3.6967171069505004e-05, + "loss": 0.5919, + "step": 20376 + }, + { + "epoch": 0.78, + "learning_rate": 3.6966529879456276e-05, + "loss": 0.5368, + "step": 20377 + }, + { + "epoch": 0.78, + "learning_rate": 3.696588868940754e-05, + "loss": 0.4726, + "step": 20378 + }, + { + "epoch": 0.78, + "learning_rate": 3.696524749935881e-05, + "loss": 0.4368, + "step": 20379 + }, + { + "epoch": 0.78, + "learning_rate": 3.696460630931008e-05, + "loss": 0.4287, + "step": 20380 + }, + { + "epoch": 0.78, + "learning_rate": 3.696396511926135e-05, + "loss": 0.5287, + "step": 20381 + }, + { + "epoch": 0.78, + "learning_rate": 3.696332392921262e-05, + "loss": 0.4614, + "step": 20382 + }, + { + "epoch": 0.78, + "learning_rate": 3.696268273916389e-05, + "loss": 0.6159, + "step": 20383 + }, + { + "epoch": 0.78, + "learning_rate": 3.6962041549115164e-05, + "loss": 0.5477, + "step": 20384 + }, + { + "epoch": 0.78, + "learning_rate": 3.696140035906643e-05, + "loss": 0.5677, + "step": 20385 + }, + { + "epoch": 0.78, + "learning_rate": 3.69607591690177e-05, + "loss": 0.569, + "step": 20386 + }, + { + "epoch": 0.78, + "learning_rate": 3.6960117978968966e-05, + "loss": 0.6159, + "step": 20387 + }, + { + "epoch": 0.78, + "learning_rate": 3.695947678892024e-05, + "loss": 0.5495, + "step": 20388 + }, + { + "epoch": 0.78, + "learning_rate": 3.69588355988715e-05, + "loss": 0.4441, + "step": 20389 + }, + { + "epoch": 0.78, + "learning_rate": 3.6958194408822774e-05, + "loss": 0.546, + "step": 20390 + }, + { + "epoch": 0.78, + "learning_rate": 3.6957553218774046e-05, + "loss": 0.4797, + "step": 20391 + }, + { + "epoch": 0.78, + "learning_rate": 3.695691202872532e-05, + "loss": 0.6183, + "step": 20392 + }, + { + "epoch": 0.78, + "learning_rate": 3.695627083867659e-05, + "loss": 0.5519, + "step": 20393 + }, + { + "epoch": 0.78, + "learning_rate": 3.6955629648627855e-05, + "loss": 0.6194, + "step": 20394 + }, + { + "epoch": 0.78, + "learning_rate": 3.6954988458579126e-05, + "loss": 0.4793, + "step": 20395 + }, + { + "epoch": 0.78, + "learning_rate": 3.695434726853039e-05, + "loss": 0.4419, + "step": 20396 + }, + { + "epoch": 0.78, + "learning_rate": 3.695370607848166e-05, + "loss": 0.5322, + "step": 20397 + }, + { + "epoch": 0.78, + "learning_rate": 3.695306488843293e-05, + "loss": 0.7176, + "step": 20398 + }, + { + "epoch": 0.78, + "learning_rate": 3.6952423698384206e-05, + "loss": 0.5937, + "step": 20399 + }, + { + "epoch": 0.78, + "learning_rate": 3.695178250833548e-05, + "loss": 0.6023, + "step": 20400 + }, + { + "epoch": 0.78, + "learning_rate": 3.695114131828674e-05, + "loss": 0.4867, + "step": 20401 + }, + { + "epoch": 0.78, + "learning_rate": 3.6950500128238015e-05, + "loss": 0.4979, + "step": 20402 + }, + { + "epoch": 0.78, + "learning_rate": 3.694985893818928e-05, + "loss": 0.4901, + "step": 20403 + }, + { + "epoch": 0.78, + "learning_rate": 3.694921774814055e-05, + "loss": 0.4216, + "step": 20404 + }, + { + "epoch": 0.78, + "learning_rate": 3.6948576558091816e-05, + "loss": 0.506, + "step": 20405 + }, + { + "epoch": 0.78, + "learning_rate": 3.694793536804309e-05, + "loss": 0.6106, + "step": 20406 + }, + { + "epoch": 0.78, + "learning_rate": 3.694729417799436e-05, + "loss": 0.5753, + "step": 20407 + }, + { + "epoch": 0.78, + "learning_rate": 3.694665298794563e-05, + "loss": 0.5304, + "step": 20408 + }, + { + "epoch": 0.78, + "learning_rate": 3.69460117978969e-05, + "loss": 0.5385, + "step": 20409 + }, + { + "epoch": 0.78, + "learning_rate": 3.694537060784817e-05, + "loss": 0.5007, + "step": 20410 + }, + { + "epoch": 0.78, + "learning_rate": 3.694472941779944e-05, + "loss": 0.5356, + "step": 20411 + }, + { + "epoch": 0.78, + "learning_rate": 3.6944088227750705e-05, + "loss": 0.5128, + "step": 20412 + }, + { + "epoch": 0.78, + "learning_rate": 3.6943447037701976e-05, + "loss": 0.4461, + "step": 20413 + }, + { + "epoch": 0.78, + "learning_rate": 3.694280584765324e-05, + "loss": 0.5136, + "step": 20414 + }, + { + "epoch": 0.78, + "learning_rate": 3.694216465760452e-05, + "loss": 0.4685, + "step": 20415 + }, + { + "epoch": 0.78, + "learning_rate": 3.6941523467555785e-05, + "loss": 0.5019, + "step": 20416 + }, + { + "epoch": 0.78, + "learning_rate": 3.6940882277507057e-05, + "loss": 0.4913, + "step": 20417 + }, + { + "epoch": 0.79, + "learning_rate": 3.694024108745833e-05, + "loss": 0.5008, + "step": 20418 + }, + { + "epoch": 0.79, + "learning_rate": 3.693959989740959e-05, + "loss": 0.545, + "step": 20419 + }, + { + "epoch": 0.79, + "learning_rate": 3.6938958707360865e-05, + "loss": 0.4543, + "step": 20420 + }, + { + "epoch": 0.79, + "learning_rate": 3.693831751731213e-05, + "loss": 0.5096, + "step": 20421 + }, + { + "epoch": 0.79, + "learning_rate": 3.69376763272634e-05, + "loss": 0.4341, + "step": 20422 + }, + { + "epoch": 0.79, + "learning_rate": 3.693703513721467e-05, + "loss": 0.5031, + "step": 20423 + }, + { + "epoch": 0.79, + "learning_rate": 3.6936393947165945e-05, + "loss": 0.575, + "step": 20424 + }, + { + "epoch": 0.79, + "learning_rate": 3.693575275711721e-05, + "loss": 0.5052, + "step": 20425 + }, + { + "epoch": 0.79, + "learning_rate": 3.693511156706848e-05, + "loss": 0.5584, + "step": 20426 + }, + { + "epoch": 0.79, + "learning_rate": 3.6934470377019753e-05, + "loss": 0.5299, + "step": 20427 + }, + { + "epoch": 0.79, + "learning_rate": 3.693382918697102e-05, + "loss": 0.4848, + "step": 20428 + }, + { + "epoch": 0.79, + "learning_rate": 3.693318799692229e-05, + "loss": 0.6429, + "step": 20429 + }, + { + "epoch": 0.79, + "learning_rate": 3.6932546806873555e-05, + "loss": 0.5582, + "step": 20430 + }, + { + "epoch": 0.79, + "learning_rate": 3.693190561682483e-05, + "loss": 0.5142, + "step": 20431 + }, + { + "epoch": 0.79, + "learning_rate": 3.69312644267761e-05, + "loss": 0.5401, + "step": 20432 + }, + { + "epoch": 0.79, + "learning_rate": 3.693062323672737e-05, + "loss": 0.469, + "step": 20433 + }, + { + "epoch": 0.79, + "learning_rate": 3.6929982046678635e-05, + "loss": 0.4272, + "step": 20434 + }, + { + "epoch": 0.79, + "learning_rate": 3.692934085662991e-05, + "loss": 0.5163, + "step": 20435 + }, + { + "epoch": 0.79, + "learning_rate": 3.692869966658118e-05, + "loss": 0.5252, + "step": 20436 + }, + { + "epoch": 0.79, + "learning_rate": 3.6928058476532443e-05, + "loss": 0.4297, + "step": 20437 + }, + { + "epoch": 0.79, + "learning_rate": 3.6927417286483715e-05, + "loss": 0.4481, + "step": 20438 + }, + { + "epoch": 0.79, + "learning_rate": 3.692677609643499e-05, + "loss": 0.5205, + "step": 20439 + }, + { + "epoch": 0.79, + "learning_rate": 3.692613490638626e-05, + "loss": 0.5263, + "step": 20440 + }, + { + "epoch": 0.79, + "learning_rate": 3.6925493716337524e-05, + "loss": 0.4479, + "step": 20441 + }, + { + "epoch": 0.79, + "learning_rate": 3.6924852526288795e-05, + "loss": 0.459, + "step": 20442 + }, + { + "epoch": 0.79, + "learning_rate": 3.692421133624006e-05, + "loss": 0.4405, + "step": 20443 + }, + { + "epoch": 0.79, + "learning_rate": 3.692357014619133e-05, + "loss": 0.5365, + "step": 20444 + }, + { + "epoch": 0.79, + "learning_rate": 3.6922928956142604e-05, + "loss": 0.4412, + "step": 20445 + }, + { + "epoch": 0.79, + "learning_rate": 3.692228776609387e-05, + "loss": 0.53, + "step": 20446 + }, + { + "epoch": 0.79, + "learning_rate": 3.692164657604514e-05, + "loss": 0.5014, + "step": 20447 + }, + { + "epoch": 0.79, + "learning_rate": 3.692100538599641e-05, + "loss": 0.5078, + "step": 20448 + }, + { + "epoch": 0.79, + "learning_rate": 3.6920364195947684e-05, + "loss": 0.4961, + "step": 20449 + }, + { + "epoch": 0.79, + "learning_rate": 3.691972300589895e-05, + "loss": 0.5834, + "step": 20450 + }, + { + "epoch": 0.79, + "learning_rate": 3.691908181585022e-05, + "loss": 0.5464, + "step": 20451 + }, + { + "epoch": 0.79, + "learning_rate": 3.6918440625801485e-05, + "loss": 0.5538, + "step": 20452 + }, + { + "epoch": 0.79, + "learning_rate": 3.691779943575276e-05, + "loss": 0.5348, + "step": 20453 + }, + { + "epoch": 0.79, + "learning_rate": 3.691715824570403e-05, + "loss": 0.4784, + "step": 20454 + }, + { + "epoch": 0.79, + "learning_rate": 3.69165170556553e-05, + "loss": 0.3686, + "step": 20455 + }, + { + "epoch": 0.79, + "learning_rate": 3.691587586560657e-05, + "loss": 0.5945, + "step": 20456 + }, + { + "epoch": 0.79, + "learning_rate": 3.691523467555784e-05, + "loss": 0.6523, + "step": 20457 + }, + { + "epoch": 0.79, + "learning_rate": 3.691459348550911e-05, + "loss": 0.5985, + "step": 20458 + }, + { + "epoch": 0.79, + "learning_rate": 3.6913952295460374e-05, + "loss": 0.5448, + "step": 20459 + }, + { + "epoch": 0.79, + "learning_rate": 3.6913311105411646e-05, + "loss": 0.4724, + "step": 20460 + }, + { + "epoch": 0.79, + "learning_rate": 3.691266991536291e-05, + "loss": 0.4715, + "step": 20461 + }, + { + "epoch": 0.79, + "learning_rate": 3.691202872531418e-05, + "loss": 0.3482, + "step": 20462 + }, + { + "epoch": 0.79, + "learning_rate": 3.6911387535265454e-05, + "loss": 0.5273, + "step": 20463 + }, + { + "epoch": 0.79, + "learning_rate": 3.6910746345216726e-05, + "loss": 0.4776, + "step": 20464 + }, + { + "epoch": 0.79, + "learning_rate": 3.6910105155168e-05, + "loss": 0.5099, + "step": 20465 + }, + { + "epoch": 0.79, + "learning_rate": 3.690946396511926e-05, + "loss": 0.5342, + "step": 20466 + }, + { + "epoch": 0.79, + "learning_rate": 3.6908822775070534e-05, + "loss": 0.5442, + "step": 20467 + }, + { + "epoch": 0.79, + "learning_rate": 3.69081815850218e-05, + "loss": 0.3782, + "step": 20468 + }, + { + "epoch": 0.79, + "learning_rate": 3.690754039497307e-05, + "loss": 0.4854, + "step": 20469 + }, + { + "epoch": 0.79, + "learning_rate": 3.6906899204924336e-05, + "loss": 0.5679, + "step": 20470 + }, + { + "epoch": 0.79, + "learning_rate": 3.6906258014875614e-05, + "loss": 0.4693, + "step": 20471 + }, + { + "epoch": 0.79, + "learning_rate": 3.6905616824826886e-05, + "loss": 0.7024, + "step": 20472 + }, + { + "epoch": 0.79, + "learning_rate": 3.690497563477815e-05, + "loss": 0.4765, + "step": 20473 + }, + { + "epoch": 0.79, + "learning_rate": 3.690433444472942e-05, + "loss": 0.5539, + "step": 20474 + }, + { + "epoch": 0.79, + "learning_rate": 3.690369325468069e-05, + "loss": 0.4747, + "step": 20475 + }, + { + "epoch": 0.79, + "learning_rate": 3.690305206463196e-05, + "loss": 0.5149, + "step": 20476 + }, + { + "epoch": 0.79, + "learning_rate": 3.6902410874583224e-05, + "loss": 0.6203, + "step": 20477 + }, + { + "epoch": 0.79, + "learning_rate": 3.6901769684534496e-05, + "loss": 0.4685, + "step": 20478 + }, + { + "epoch": 0.79, + "learning_rate": 3.690112849448577e-05, + "loss": 0.5004, + "step": 20479 + }, + { + "epoch": 0.79, + "learning_rate": 3.690048730443704e-05, + "loss": 0.4223, + "step": 20480 + }, + { + "epoch": 0.79, + "learning_rate": 3.689984611438831e-05, + "loss": 0.5298, + "step": 20481 + }, + { + "epoch": 0.79, + "learning_rate": 3.6899204924339576e-05, + "loss": 0.5191, + "step": 20482 + }, + { + "epoch": 0.79, + "learning_rate": 3.689856373429085e-05, + "loss": 0.4701, + "step": 20483 + }, + { + "epoch": 0.79, + "learning_rate": 3.689792254424211e-05, + "loss": 0.5325, + "step": 20484 + }, + { + "epoch": 0.79, + "learning_rate": 3.6897281354193384e-05, + "loss": 0.4951, + "step": 20485 + }, + { + "epoch": 0.79, + "learning_rate": 3.689664016414465e-05, + "loss": 0.4064, + "step": 20486 + }, + { + "epoch": 0.79, + "learning_rate": 3.689599897409593e-05, + "loss": 0.5014, + "step": 20487 + }, + { + "epoch": 0.79, + "learning_rate": 3.689535778404719e-05, + "loss": 0.4447, + "step": 20488 + }, + { + "epoch": 0.79, + "learning_rate": 3.6894716593998464e-05, + "loss": 0.4889, + "step": 20489 + }, + { + "epoch": 0.79, + "learning_rate": 3.6894075403949736e-05, + "loss": 0.457, + "step": 20490 + }, + { + "epoch": 0.79, + "learning_rate": 3.6893434213901e-05, + "loss": 0.494, + "step": 20491 + }, + { + "epoch": 0.79, + "learning_rate": 3.689279302385227e-05, + "loss": 0.4089, + "step": 20492 + }, + { + "epoch": 0.79, + "learning_rate": 3.689215183380354e-05, + "loss": 0.4936, + "step": 20493 + }, + { + "epoch": 0.79, + "learning_rate": 3.689151064375481e-05, + "loss": 0.4785, + "step": 20494 + }, + { + "epoch": 0.79, + "learning_rate": 3.689086945370608e-05, + "loss": 0.6149, + "step": 20495 + }, + { + "epoch": 0.79, + "learning_rate": 3.689022826365735e-05, + "loss": 0.529, + "step": 20496 + }, + { + "epoch": 0.79, + "learning_rate": 3.688958707360862e-05, + "loss": 0.4819, + "step": 20497 + }, + { + "epoch": 0.79, + "learning_rate": 3.688894588355989e-05, + "loss": 0.5027, + "step": 20498 + }, + { + "epoch": 0.79, + "learning_rate": 3.688830469351116e-05, + "loss": 0.4669, + "step": 20499 + }, + { + "epoch": 0.79, + "learning_rate": 3.6887663503462426e-05, + "loss": 0.5368, + "step": 20500 + }, + { + "epoch": 0.79, + "learning_rate": 3.68870223134137e-05, + "loss": 0.6153, + "step": 20501 + }, + { + "epoch": 0.79, + "learning_rate": 3.688638112336496e-05, + "loss": 0.5073, + "step": 20502 + }, + { + "epoch": 0.79, + "learning_rate": 3.688573993331624e-05, + "loss": 0.6027, + "step": 20503 + }, + { + "epoch": 0.79, + "learning_rate": 3.6885098743267506e-05, + "loss": 0.5926, + "step": 20504 + }, + { + "epoch": 0.79, + "learning_rate": 3.688445755321878e-05, + "loss": 0.5453, + "step": 20505 + }, + { + "epoch": 0.79, + "learning_rate": 3.688381636317004e-05, + "loss": 0.496, + "step": 20506 + }, + { + "epoch": 0.79, + "learning_rate": 3.6883175173121315e-05, + "loss": 0.5017, + "step": 20507 + }, + { + "epoch": 0.79, + "learning_rate": 3.6882533983072586e-05, + "loss": 0.5455, + "step": 20508 + }, + { + "epoch": 0.79, + "learning_rate": 3.688189279302385e-05, + "loss": 0.5322, + "step": 20509 + }, + { + "epoch": 0.79, + "learning_rate": 3.688125160297512e-05, + "loss": 0.643, + "step": 20510 + }, + { + "epoch": 0.79, + "learning_rate": 3.6880610412926395e-05, + "loss": 0.3427, + "step": 20511 + }, + { + "epoch": 0.79, + "learning_rate": 3.6879969222877666e-05, + "loss": 0.5366, + "step": 20512 + }, + { + "epoch": 0.79, + "learning_rate": 3.687932803282893e-05, + "loss": 0.5536, + "step": 20513 + }, + { + "epoch": 0.79, + "learning_rate": 3.68786868427802e-05, + "loss": 0.4954, + "step": 20514 + }, + { + "epoch": 0.79, + "learning_rate": 3.687804565273147e-05, + "loss": 0.5197, + "step": 20515 + }, + { + "epoch": 0.79, + "learning_rate": 3.687740446268274e-05, + "loss": 0.5349, + "step": 20516 + }, + { + "epoch": 0.79, + "learning_rate": 3.687676327263401e-05, + "loss": 0.526, + "step": 20517 + }, + { + "epoch": 0.79, + "learning_rate": 3.6876122082585276e-05, + "loss": 0.5176, + "step": 20518 + }, + { + "epoch": 0.79, + "learning_rate": 3.687548089253655e-05, + "loss": 0.5679, + "step": 20519 + }, + { + "epoch": 0.79, + "learning_rate": 3.687483970248782e-05, + "loss": 0.5943, + "step": 20520 + }, + { + "epoch": 0.79, + "learning_rate": 3.687419851243909e-05, + "loss": 0.5011, + "step": 20521 + }, + { + "epoch": 0.79, + "learning_rate": 3.6873557322390356e-05, + "loss": 0.4169, + "step": 20522 + }, + { + "epoch": 0.79, + "learning_rate": 3.687291613234163e-05, + "loss": 0.4722, + "step": 20523 + }, + { + "epoch": 0.79, + "learning_rate": 3.687227494229289e-05, + "loss": 0.6015, + "step": 20524 + }, + { + "epoch": 0.79, + "learning_rate": 3.6871633752244165e-05, + "loss": 0.4589, + "step": 20525 + }, + { + "epoch": 0.79, + "learning_rate": 3.6870992562195437e-05, + "loss": 0.5692, + "step": 20526 + }, + { + "epoch": 0.79, + "learning_rate": 3.687035137214671e-05, + "loss": 0.5462, + "step": 20527 + }, + { + "epoch": 0.79, + "learning_rate": 3.686971018209798e-05, + "loss": 0.5503, + "step": 20528 + }, + { + "epoch": 0.79, + "learning_rate": 3.6869068992049245e-05, + "loss": 0.5125, + "step": 20529 + }, + { + "epoch": 0.79, + "learning_rate": 3.686842780200052e-05, + "loss": 0.5449, + "step": 20530 + }, + { + "epoch": 0.79, + "learning_rate": 3.686778661195178e-05, + "loss": 0.4321, + "step": 20531 + }, + { + "epoch": 0.79, + "learning_rate": 3.686714542190305e-05, + "loss": 0.4583, + "step": 20532 + }, + { + "epoch": 0.79, + "learning_rate": 3.686650423185432e-05, + "loss": 0.4372, + "step": 20533 + }, + { + "epoch": 0.79, + "learning_rate": 3.686586304180559e-05, + "loss": 0.4879, + "step": 20534 + }, + { + "epoch": 0.79, + "learning_rate": 3.686522185175686e-05, + "loss": 0.4053, + "step": 20535 + }, + { + "epoch": 0.79, + "learning_rate": 3.686458066170813e-05, + "loss": 0.5645, + "step": 20536 + }, + { + "epoch": 0.79, + "learning_rate": 3.6863939471659405e-05, + "loss": 0.5071, + "step": 20537 + }, + { + "epoch": 0.79, + "learning_rate": 3.686329828161067e-05, + "loss": 0.4321, + "step": 20538 + }, + { + "epoch": 0.79, + "learning_rate": 3.686265709156194e-05, + "loss": 0.4573, + "step": 20539 + }, + { + "epoch": 0.79, + "learning_rate": 3.686201590151321e-05, + "loss": 0.4597, + "step": 20540 + }, + { + "epoch": 0.79, + "learning_rate": 3.686137471146448e-05, + "loss": 0.4651, + "step": 20541 + }, + { + "epoch": 0.79, + "learning_rate": 3.686073352141575e-05, + "loss": 0.5199, + "step": 20542 + }, + { + "epoch": 0.79, + "learning_rate": 3.686009233136702e-05, + "loss": 0.4939, + "step": 20543 + }, + { + "epoch": 0.79, + "learning_rate": 3.6859451141318294e-05, + "loss": 0.4809, + "step": 20544 + }, + { + "epoch": 0.79, + "learning_rate": 3.685880995126956e-05, + "loss": 0.5312, + "step": 20545 + }, + { + "epoch": 0.79, + "learning_rate": 3.685816876122083e-05, + "loss": 0.5693, + "step": 20546 + }, + { + "epoch": 0.79, + "learning_rate": 3.6857527571172095e-05, + "loss": 0.4041, + "step": 20547 + }, + { + "epoch": 0.79, + "learning_rate": 3.685688638112337e-05, + "loss": 0.4757, + "step": 20548 + }, + { + "epoch": 0.79, + "learning_rate": 3.685624519107463e-05, + "loss": 0.508, + "step": 20549 + }, + { + "epoch": 0.79, + "learning_rate": 3.6855604001025904e-05, + "loss": 0.5237, + "step": 20550 + }, + { + "epoch": 0.79, + "learning_rate": 3.6854962810977175e-05, + "loss": 0.5039, + "step": 20551 + }, + { + "epoch": 0.79, + "learning_rate": 3.685432162092845e-05, + "loss": 0.4912, + "step": 20552 + }, + { + "epoch": 0.79, + "learning_rate": 3.685368043087972e-05, + "loss": 0.423, + "step": 20553 + }, + { + "epoch": 0.79, + "learning_rate": 3.6853039240830984e-05, + "loss": 0.5145, + "step": 20554 + }, + { + "epoch": 0.79, + "learning_rate": 3.6852398050782255e-05, + "loss": 0.557, + "step": 20555 + }, + { + "epoch": 0.79, + "learning_rate": 3.685175686073352e-05, + "loss": 0.5406, + "step": 20556 + }, + { + "epoch": 0.79, + "learning_rate": 3.685111567068479e-05, + "loss": 0.5082, + "step": 20557 + }, + { + "epoch": 0.79, + "learning_rate": 3.685047448063606e-05, + "loss": 0.4858, + "step": 20558 + }, + { + "epoch": 0.79, + "learning_rate": 3.6849833290587335e-05, + "loss": 0.5038, + "step": 20559 + }, + { + "epoch": 0.79, + "learning_rate": 3.68491921005386e-05, + "loss": 0.5375, + "step": 20560 + }, + { + "epoch": 0.79, + "learning_rate": 3.684855091048987e-05, + "loss": 0.568, + "step": 20561 + }, + { + "epoch": 0.79, + "learning_rate": 3.6847909720441144e-05, + "loss": 0.4557, + "step": 20562 + }, + { + "epoch": 0.79, + "learning_rate": 3.684726853039241e-05, + "loss": 0.5441, + "step": 20563 + }, + { + "epoch": 0.79, + "learning_rate": 3.684662734034368e-05, + "loss": 0.6126, + "step": 20564 + }, + { + "epoch": 0.79, + "learning_rate": 3.6845986150294945e-05, + "loss": 0.4789, + "step": 20565 + }, + { + "epoch": 0.79, + "learning_rate": 3.684534496024622e-05, + "loss": 0.5016, + "step": 20566 + }, + { + "epoch": 0.79, + "learning_rate": 3.684470377019749e-05, + "loss": 0.4384, + "step": 20567 + }, + { + "epoch": 0.79, + "learning_rate": 3.684406258014876e-05, + "loss": 0.3581, + "step": 20568 + }, + { + "epoch": 0.79, + "learning_rate": 3.6843421390100025e-05, + "loss": 0.5035, + "step": 20569 + }, + { + "epoch": 0.79, + "learning_rate": 3.68427802000513e-05, + "loss": 0.4867, + "step": 20570 + }, + { + "epoch": 0.79, + "learning_rate": 3.684213901000257e-05, + "loss": 0.5664, + "step": 20571 + }, + { + "epoch": 0.79, + "learning_rate": 3.6841497819953834e-05, + "loss": 0.4675, + "step": 20572 + }, + { + "epoch": 0.79, + "learning_rate": 3.6840856629905106e-05, + "loss": 0.4548, + "step": 20573 + }, + { + "epoch": 0.79, + "learning_rate": 3.684021543985637e-05, + "loss": 0.5507, + "step": 20574 + }, + { + "epoch": 0.79, + "learning_rate": 3.683957424980765e-05, + "loss": 0.4903, + "step": 20575 + }, + { + "epoch": 0.79, + "learning_rate": 3.6838933059758914e-05, + "loss": 0.6058, + "step": 20576 + }, + { + "epoch": 0.79, + "learning_rate": 3.6838291869710186e-05, + "loss": 0.4725, + "step": 20577 + }, + { + "epoch": 0.79, + "learning_rate": 3.683765067966145e-05, + "loss": 0.5923, + "step": 20578 + }, + { + "epoch": 0.79, + "learning_rate": 3.683700948961272e-05, + "loss": 0.5751, + "step": 20579 + }, + { + "epoch": 0.79, + "learning_rate": 3.6836368299563994e-05, + "loss": 0.4434, + "step": 20580 + }, + { + "epoch": 0.79, + "learning_rate": 3.683572710951526e-05, + "loss": 0.5458, + "step": 20581 + }, + { + "epoch": 0.79, + "learning_rate": 3.683508591946653e-05, + "loss": 0.4874, + "step": 20582 + }, + { + "epoch": 0.79, + "learning_rate": 3.68344447294178e-05, + "loss": 0.4638, + "step": 20583 + }, + { + "epoch": 0.79, + "learning_rate": 3.6833803539369074e-05, + "loss": 0.5061, + "step": 20584 + }, + { + "epoch": 0.79, + "learning_rate": 3.683316234932034e-05, + "loss": 0.5385, + "step": 20585 + }, + { + "epoch": 0.79, + "learning_rate": 3.683252115927161e-05, + "loss": 0.6627, + "step": 20586 + }, + { + "epoch": 0.79, + "learning_rate": 3.683187996922288e-05, + "loss": 0.5454, + "step": 20587 + }, + { + "epoch": 0.79, + "learning_rate": 3.683123877917415e-05, + "loss": 0.4817, + "step": 20588 + }, + { + "epoch": 0.79, + "learning_rate": 3.683059758912542e-05, + "loss": 0.5606, + "step": 20589 + }, + { + "epoch": 0.79, + "learning_rate": 3.6829956399076684e-05, + "loss": 0.4919, + "step": 20590 + }, + { + "epoch": 0.79, + "learning_rate": 3.6829315209027956e-05, + "loss": 0.4411, + "step": 20591 + }, + { + "epoch": 0.79, + "learning_rate": 3.682867401897923e-05, + "loss": 0.6231, + "step": 20592 + }, + { + "epoch": 0.79, + "learning_rate": 3.68280328289305e-05, + "loss": 0.3937, + "step": 20593 + }, + { + "epoch": 0.79, + "learning_rate": 3.6827391638881764e-05, + "loss": 0.4405, + "step": 20594 + }, + { + "epoch": 0.79, + "learning_rate": 3.6826750448833036e-05, + "loss": 0.4403, + "step": 20595 + }, + { + "epoch": 0.79, + "learning_rate": 3.682610925878431e-05, + "loss": 0.6379, + "step": 20596 + }, + { + "epoch": 0.79, + "learning_rate": 3.682546806873557e-05, + "loss": 0.5205, + "step": 20597 + }, + { + "epoch": 0.79, + "learning_rate": 3.6824826878686844e-05, + "loss": 0.3659, + "step": 20598 + }, + { + "epoch": 0.79, + "learning_rate": 3.6824185688638116e-05, + "loss": 0.5581, + "step": 20599 + }, + { + "epoch": 0.79, + "learning_rate": 3.682354449858939e-05, + "loss": 0.6051, + "step": 20600 + }, + { + "epoch": 0.79, + "learning_rate": 3.682290330854065e-05, + "loss": 0.467, + "step": 20601 + }, + { + "epoch": 0.79, + "learning_rate": 3.6822262118491924e-05, + "loss": 0.5514, + "step": 20602 + }, + { + "epoch": 0.79, + "learning_rate": 3.682162092844319e-05, + "loss": 0.4453, + "step": 20603 + }, + { + "epoch": 0.79, + "learning_rate": 3.682097973839446e-05, + "loss": 0.5005, + "step": 20604 + }, + { + "epoch": 0.79, + "learning_rate": 3.682033854834573e-05, + "loss": 0.6112, + "step": 20605 + }, + { + "epoch": 0.79, + "learning_rate": 3.6819697358297e-05, + "loss": 0.5692, + "step": 20606 + }, + { + "epoch": 0.79, + "learning_rate": 3.681905616824827e-05, + "loss": 0.537, + "step": 20607 + }, + { + "epoch": 0.79, + "learning_rate": 3.681841497819954e-05, + "loss": 0.5051, + "step": 20608 + }, + { + "epoch": 0.79, + "learning_rate": 3.681777378815081e-05, + "loss": 0.5695, + "step": 20609 + }, + { + "epoch": 0.79, + "learning_rate": 3.681713259810208e-05, + "loss": 0.5238, + "step": 20610 + }, + { + "epoch": 0.79, + "learning_rate": 3.681649140805335e-05, + "loss": 0.5108, + "step": 20611 + }, + { + "epoch": 0.79, + "learning_rate": 3.6815850218004614e-05, + "loss": 0.4562, + "step": 20612 + }, + { + "epoch": 0.79, + "learning_rate": 3.6815209027955886e-05, + "loss": 0.5561, + "step": 20613 + }, + { + "epoch": 0.79, + "learning_rate": 3.681456783790716e-05, + "loss": 0.5316, + "step": 20614 + }, + { + "epoch": 0.79, + "learning_rate": 3.681392664785843e-05, + "loss": 0.4647, + "step": 20615 + }, + { + "epoch": 0.79, + "learning_rate": 3.68132854578097e-05, + "loss": 0.5679, + "step": 20616 + }, + { + "epoch": 0.79, + "learning_rate": 3.6812644267760966e-05, + "loss": 0.4188, + "step": 20617 + }, + { + "epoch": 0.79, + "learning_rate": 3.681200307771224e-05, + "loss": 0.4925, + "step": 20618 + }, + { + "epoch": 0.79, + "learning_rate": 3.68113618876635e-05, + "loss": 0.5502, + "step": 20619 + }, + { + "epoch": 0.79, + "learning_rate": 3.6810720697614775e-05, + "loss": 0.4749, + "step": 20620 + }, + { + "epoch": 0.79, + "learning_rate": 3.681007950756604e-05, + "loss": 0.5232, + "step": 20621 + }, + { + "epoch": 0.79, + "learning_rate": 3.680943831751731e-05, + "loss": 0.5612, + "step": 20622 + }, + { + "epoch": 0.79, + "learning_rate": 3.680879712746858e-05, + "loss": 0.4915, + "step": 20623 + }, + { + "epoch": 0.79, + "learning_rate": 3.6808155937419855e-05, + "loss": 0.4892, + "step": 20624 + }, + { + "epoch": 0.79, + "learning_rate": 3.6807514747371126e-05, + "loss": 0.5238, + "step": 20625 + }, + { + "epoch": 0.79, + "learning_rate": 3.680687355732239e-05, + "loss": 0.4539, + "step": 20626 + }, + { + "epoch": 0.79, + "learning_rate": 3.680623236727366e-05, + "loss": 0.4976, + "step": 20627 + }, + { + "epoch": 0.79, + "learning_rate": 3.680559117722493e-05, + "loss": 0.4939, + "step": 20628 + }, + { + "epoch": 0.79, + "learning_rate": 3.68049499871762e-05, + "loss": 0.541, + "step": 20629 + }, + { + "epoch": 0.79, + "learning_rate": 3.6804308797127465e-05, + "loss": 0.5512, + "step": 20630 + }, + { + "epoch": 0.79, + "learning_rate": 3.680366760707874e-05, + "loss": 0.4898, + "step": 20631 + }, + { + "epoch": 0.79, + "learning_rate": 3.6803026417030015e-05, + "loss": 0.4869, + "step": 20632 + }, + { + "epoch": 0.79, + "learning_rate": 3.680238522698128e-05, + "loss": 0.4071, + "step": 20633 + }, + { + "epoch": 0.79, + "learning_rate": 3.680174403693255e-05, + "loss": 0.449, + "step": 20634 + }, + { + "epoch": 0.79, + "learning_rate": 3.6801102846883816e-05, + "loss": 0.5361, + "step": 20635 + }, + { + "epoch": 0.79, + "learning_rate": 3.680046165683509e-05, + "loss": 0.5486, + "step": 20636 + }, + { + "epoch": 0.79, + "learning_rate": 3.679982046678635e-05, + "loss": 0.5862, + "step": 20637 + }, + { + "epoch": 0.79, + "learning_rate": 3.6799179276737625e-05, + "loss": 0.5803, + "step": 20638 + }, + { + "epoch": 0.79, + "learning_rate": 3.6798538086688897e-05, + "loss": 0.5868, + "step": 20639 + }, + { + "epoch": 0.79, + "learning_rate": 3.679789689664017e-05, + "loss": 0.5262, + "step": 20640 + }, + { + "epoch": 0.79, + "learning_rate": 3.679725570659144e-05, + "loss": 0.4402, + "step": 20641 + }, + { + "epoch": 0.79, + "learning_rate": 3.6796614516542705e-05, + "loss": 0.5854, + "step": 20642 + }, + { + "epoch": 0.79, + "learning_rate": 3.679597332649398e-05, + "loss": 0.494, + "step": 20643 + }, + { + "epoch": 0.79, + "learning_rate": 3.679533213644524e-05, + "loss": 0.4852, + "step": 20644 + }, + { + "epoch": 0.79, + "learning_rate": 3.679469094639651e-05, + "loss": 0.4953, + "step": 20645 + }, + { + "epoch": 0.79, + "learning_rate": 3.679404975634778e-05, + "loss": 0.4677, + "step": 20646 + }, + { + "epoch": 0.79, + "learning_rate": 3.679340856629906e-05, + "loss": 0.5287, + "step": 20647 + }, + { + "epoch": 0.79, + "learning_rate": 3.679276737625032e-05, + "loss": 0.4795, + "step": 20648 + }, + { + "epoch": 0.79, + "learning_rate": 3.6792126186201593e-05, + "loss": 0.4791, + "step": 20649 + }, + { + "epoch": 0.79, + "learning_rate": 3.6791484996152865e-05, + "loss": 0.5299, + "step": 20650 + }, + { + "epoch": 0.79, + "learning_rate": 3.679084380610413e-05, + "loss": 0.3982, + "step": 20651 + }, + { + "epoch": 0.79, + "learning_rate": 3.67902026160554e-05, + "loss": 0.486, + "step": 20652 + }, + { + "epoch": 0.79, + "learning_rate": 3.678956142600667e-05, + "loss": 0.4979, + "step": 20653 + }, + { + "epoch": 0.79, + "learning_rate": 3.678892023595794e-05, + "loss": 0.4407, + "step": 20654 + }, + { + "epoch": 0.79, + "learning_rate": 3.678827904590921e-05, + "loss": 0.452, + "step": 20655 + }, + { + "epoch": 0.79, + "learning_rate": 3.678763785586048e-05, + "loss": 0.4887, + "step": 20656 + }, + { + "epoch": 0.79, + "learning_rate": 3.678699666581175e-05, + "loss": 0.4896, + "step": 20657 + }, + { + "epoch": 0.79, + "learning_rate": 3.678635547576302e-05, + "loss": 0.4549, + "step": 20658 + }, + { + "epoch": 0.79, + "learning_rate": 3.678571428571429e-05, + "loss": 0.5571, + "step": 20659 + }, + { + "epoch": 0.79, + "learning_rate": 3.6785073095665555e-05, + "loss": 0.4722, + "step": 20660 + }, + { + "epoch": 0.79, + "learning_rate": 3.678443190561683e-05, + "loss": 0.4854, + "step": 20661 + }, + { + "epoch": 0.79, + "learning_rate": 3.678379071556809e-05, + "loss": 0.6293, + "step": 20662 + }, + { + "epoch": 0.79, + "learning_rate": 3.678314952551937e-05, + "loss": 0.5512, + "step": 20663 + }, + { + "epoch": 0.79, + "learning_rate": 3.6782508335470635e-05, + "loss": 0.558, + "step": 20664 + }, + { + "epoch": 0.79, + "learning_rate": 3.678186714542191e-05, + "loss": 0.4705, + "step": 20665 + }, + { + "epoch": 0.79, + "learning_rate": 3.678122595537317e-05, + "loss": 0.5616, + "step": 20666 + }, + { + "epoch": 0.79, + "learning_rate": 3.6780584765324444e-05, + "loss": 0.5337, + "step": 20667 + }, + { + "epoch": 0.79, + "learning_rate": 3.6779943575275715e-05, + "loss": 0.4918, + "step": 20668 + }, + { + "epoch": 0.79, + "learning_rate": 3.677930238522698e-05, + "loss": 0.4926, + "step": 20669 + }, + { + "epoch": 0.79, + "learning_rate": 3.677866119517825e-05, + "loss": 0.4587, + "step": 20670 + }, + { + "epoch": 0.79, + "learning_rate": 3.6778020005129524e-05, + "loss": 0.5537, + "step": 20671 + }, + { + "epoch": 0.79, + "learning_rate": 3.6777378815080795e-05, + "loss": 0.4803, + "step": 20672 + }, + { + "epoch": 0.79, + "learning_rate": 3.677673762503206e-05, + "loss": 0.5497, + "step": 20673 + }, + { + "epoch": 0.79, + "learning_rate": 3.677609643498333e-05, + "loss": 0.4895, + "step": 20674 + }, + { + "epoch": 0.79, + "learning_rate": 3.67754552449346e-05, + "loss": 0.5164, + "step": 20675 + }, + { + "epoch": 0.79, + "learning_rate": 3.677481405488587e-05, + "loss": 0.5245, + "step": 20676 + }, + { + "epoch": 0.79, + "learning_rate": 3.677417286483714e-05, + "loss": 0.5137, + "step": 20677 + }, + { + "epoch": 0.8, + "learning_rate": 3.6773531674788405e-05, + "loss": 0.448, + "step": 20678 + }, + { + "epoch": 0.8, + "learning_rate": 3.677289048473968e-05, + "loss": 0.4658, + "step": 20679 + }, + { + "epoch": 0.8, + "learning_rate": 3.677224929469095e-05, + "loss": 0.543, + "step": 20680 + }, + { + "epoch": 0.8, + "learning_rate": 3.677160810464222e-05, + "loss": 0.518, + "step": 20681 + }, + { + "epoch": 0.8, + "learning_rate": 3.6770966914593486e-05, + "loss": 0.5564, + "step": 20682 + }, + { + "epoch": 0.8, + "learning_rate": 3.677032572454476e-05, + "loss": 0.4432, + "step": 20683 + }, + { + "epoch": 0.8, + "learning_rate": 3.676968453449602e-05, + "loss": 0.5999, + "step": 20684 + }, + { + "epoch": 0.8, + "learning_rate": 3.6769043344447294e-05, + "loss": 0.4781, + "step": 20685 + }, + { + "epoch": 0.8, + "learning_rate": 3.6768402154398566e-05, + "loss": 0.6413, + "step": 20686 + }, + { + "epoch": 0.8, + "learning_rate": 3.676776096434984e-05, + "loss": 0.5175, + "step": 20687 + }, + { + "epoch": 0.8, + "learning_rate": 3.676711977430111e-05, + "loss": 0.4456, + "step": 20688 + }, + { + "epoch": 0.8, + "learning_rate": 3.6766478584252374e-05, + "loss": 0.5, + "step": 20689 + }, + { + "epoch": 0.8, + "learning_rate": 3.6765837394203646e-05, + "loss": 0.6563, + "step": 20690 + }, + { + "epoch": 0.8, + "learning_rate": 3.676519620415491e-05, + "loss": 0.6049, + "step": 20691 + }, + { + "epoch": 0.8, + "learning_rate": 3.676455501410618e-05, + "loss": 0.5471, + "step": 20692 + }, + { + "epoch": 0.8, + "learning_rate": 3.676391382405745e-05, + "loss": 0.554, + "step": 20693 + }, + { + "epoch": 0.8, + "learning_rate": 3.676327263400872e-05, + "loss": 0.5858, + "step": 20694 + }, + { + "epoch": 0.8, + "learning_rate": 3.676263144395999e-05, + "loss": 0.5347, + "step": 20695 + }, + { + "epoch": 0.8, + "learning_rate": 3.676199025391126e-05, + "loss": 0.5249, + "step": 20696 + }, + { + "epoch": 0.8, + "learning_rate": 3.6761349063862534e-05, + "loss": 0.492, + "step": 20697 + }, + { + "epoch": 0.8, + "learning_rate": 3.67607078738138e-05, + "loss": 0.4748, + "step": 20698 + }, + { + "epoch": 0.8, + "learning_rate": 3.676006668376507e-05, + "loss": 0.5837, + "step": 20699 + }, + { + "epoch": 0.8, + "learning_rate": 3.6759425493716336e-05, + "loss": 0.4791, + "step": 20700 + }, + { + "epoch": 0.8, + "learning_rate": 3.675878430366761e-05, + "loss": 0.4888, + "step": 20701 + }, + { + "epoch": 0.8, + "learning_rate": 3.675814311361888e-05, + "loss": 0.5887, + "step": 20702 + }, + { + "epoch": 0.8, + "learning_rate": 3.675750192357015e-05, + "loss": 0.6174, + "step": 20703 + }, + { + "epoch": 0.8, + "learning_rate": 3.675686073352142e-05, + "loss": 0.4281, + "step": 20704 + }, + { + "epoch": 0.8, + "learning_rate": 3.675621954347269e-05, + "loss": 0.473, + "step": 20705 + }, + { + "epoch": 0.8, + "learning_rate": 3.675557835342396e-05, + "loss": 0.4931, + "step": 20706 + }, + { + "epoch": 0.8, + "learning_rate": 3.6754937163375224e-05, + "loss": 0.4765, + "step": 20707 + }, + { + "epoch": 0.8, + "learning_rate": 3.6754295973326496e-05, + "loss": 0.4329, + "step": 20708 + }, + { + "epoch": 0.8, + "learning_rate": 3.675365478327776e-05, + "loss": 0.5596, + "step": 20709 + }, + { + "epoch": 0.8, + "learning_rate": 3.675301359322903e-05, + "loss": 0.4583, + "step": 20710 + }, + { + "epoch": 0.8, + "learning_rate": 3.6752372403180304e-05, + "loss": 0.52, + "step": 20711 + }, + { + "epoch": 0.8, + "learning_rate": 3.6751731213131576e-05, + "loss": 0.5702, + "step": 20712 + }, + { + "epoch": 0.8, + "learning_rate": 3.675109002308285e-05, + "loss": 0.5363, + "step": 20713 + }, + { + "epoch": 0.8, + "learning_rate": 3.675044883303411e-05, + "loss": 0.5496, + "step": 20714 + }, + { + "epoch": 0.8, + "learning_rate": 3.6749807642985384e-05, + "loss": 0.4065, + "step": 20715 + }, + { + "epoch": 0.8, + "learning_rate": 3.674916645293665e-05, + "loss": 0.4175, + "step": 20716 + }, + { + "epoch": 0.8, + "learning_rate": 3.674852526288792e-05, + "loss": 0.5531, + "step": 20717 + }, + { + "epoch": 0.8, + "learning_rate": 3.6747884072839186e-05, + "loss": 0.5, + "step": 20718 + }, + { + "epoch": 0.8, + "learning_rate": 3.6747242882790464e-05, + "loss": 0.5177, + "step": 20719 + }, + { + "epoch": 0.8, + "learning_rate": 3.674660169274173e-05, + "loss": 0.5758, + "step": 20720 + }, + { + "epoch": 0.8, + "learning_rate": 3.6745960502693e-05, + "loss": 0.6627, + "step": 20721 + }, + { + "epoch": 0.8, + "learning_rate": 3.674531931264427e-05, + "loss": 0.4997, + "step": 20722 + }, + { + "epoch": 0.8, + "learning_rate": 3.674467812259554e-05, + "loss": 0.531, + "step": 20723 + }, + { + "epoch": 0.8, + "learning_rate": 3.674403693254681e-05, + "loss": 0.5244, + "step": 20724 + }, + { + "epoch": 0.8, + "learning_rate": 3.6743395742498074e-05, + "loss": 0.4957, + "step": 20725 + }, + { + "epoch": 0.8, + "learning_rate": 3.6742754552449346e-05, + "loss": 0.4536, + "step": 20726 + }, + { + "epoch": 0.8, + "learning_rate": 3.674211336240062e-05, + "loss": 0.4314, + "step": 20727 + }, + { + "epoch": 0.8, + "learning_rate": 3.674147217235189e-05, + "loss": 0.464, + "step": 20728 + }, + { + "epoch": 0.8, + "learning_rate": 3.6740830982303155e-05, + "loss": 0.6002, + "step": 20729 + }, + { + "epoch": 0.8, + "learning_rate": 3.6740189792254426e-05, + "loss": 0.4681, + "step": 20730 + }, + { + "epoch": 0.8, + "learning_rate": 3.67395486022057e-05, + "loss": 0.4583, + "step": 20731 + }, + { + "epoch": 0.8, + "learning_rate": 3.673890741215696e-05, + "loss": 0.5159, + "step": 20732 + }, + { + "epoch": 0.8, + "learning_rate": 3.6738266222108235e-05, + "loss": 0.4686, + "step": 20733 + }, + { + "epoch": 0.8, + "learning_rate": 3.67376250320595e-05, + "loss": 0.4815, + "step": 20734 + }, + { + "epoch": 0.8, + "learning_rate": 3.673698384201078e-05, + "loss": 0.4636, + "step": 20735 + }, + { + "epoch": 0.8, + "learning_rate": 3.673634265196204e-05, + "loss": 0.4654, + "step": 20736 + }, + { + "epoch": 0.8, + "learning_rate": 3.6735701461913315e-05, + "loss": 0.4204, + "step": 20737 + }, + { + "epoch": 0.8, + "learning_rate": 3.673506027186458e-05, + "loss": 0.5262, + "step": 20738 + }, + { + "epoch": 0.8, + "learning_rate": 3.673441908181585e-05, + "loss": 0.4383, + "step": 20739 + }, + { + "epoch": 0.8, + "learning_rate": 3.673377789176712e-05, + "loss": 0.4991, + "step": 20740 + }, + { + "epoch": 0.8, + "learning_rate": 3.673313670171839e-05, + "loss": 0.5036, + "step": 20741 + }, + { + "epoch": 0.8, + "learning_rate": 3.673249551166966e-05, + "loss": 0.4489, + "step": 20742 + }, + { + "epoch": 0.8, + "learning_rate": 3.673185432162093e-05, + "loss": 0.4964, + "step": 20743 + }, + { + "epoch": 0.8, + "learning_rate": 3.67312131315722e-05, + "loss": 0.5346, + "step": 20744 + }, + { + "epoch": 0.8, + "learning_rate": 3.673057194152347e-05, + "loss": 0.4205, + "step": 20745 + }, + { + "epoch": 0.8, + "learning_rate": 3.672993075147474e-05, + "loss": 0.4973, + "step": 20746 + }, + { + "epoch": 0.8, + "learning_rate": 3.6729289561426005e-05, + "loss": 0.4338, + "step": 20747 + }, + { + "epoch": 0.8, + "learning_rate": 3.6728648371377277e-05, + "loss": 0.435, + "step": 20748 + }, + { + "epoch": 0.8, + "learning_rate": 3.672800718132855e-05, + "loss": 0.4892, + "step": 20749 + }, + { + "epoch": 0.8, + "learning_rate": 3.672736599127981e-05, + "loss": 0.5062, + "step": 20750 + }, + { + "epoch": 0.8, + "learning_rate": 3.672672480123109e-05, + "loss": 0.5928, + "step": 20751 + }, + { + "epoch": 0.8, + "learning_rate": 3.6726083611182357e-05, + "loss": 0.4595, + "step": 20752 + }, + { + "epoch": 0.8, + "learning_rate": 3.672544242113363e-05, + "loss": 0.5915, + "step": 20753 + }, + { + "epoch": 0.8, + "learning_rate": 3.672480123108489e-05, + "loss": 0.416, + "step": 20754 + }, + { + "epoch": 0.8, + "learning_rate": 3.6724160041036165e-05, + "loss": 0.4814, + "step": 20755 + }, + { + "epoch": 0.8, + "learning_rate": 3.672351885098743e-05, + "loss": 0.4805, + "step": 20756 + }, + { + "epoch": 0.8, + "learning_rate": 3.67228776609387e-05, + "loss": 0.6825, + "step": 20757 + }, + { + "epoch": 0.8, + "learning_rate": 3.672223647088997e-05, + "loss": 0.4357, + "step": 20758 + }, + { + "epoch": 0.8, + "learning_rate": 3.6721595280841245e-05, + "loss": 0.4267, + "step": 20759 + }, + { + "epoch": 0.8, + "learning_rate": 3.672095409079252e-05, + "loss": 0.443, + "step": 20760 + }, + { + "epoch": 0.8, + "learning_rate": 3.672031290074378e-05, + "loss": 0.457, + "step": 20761 + }, + { + "epoch": 0.8, + "learning_rate": 3.6719671710695053e-05, + "loss": 0.6225, + "step": 20762 + }, + { + "epoch": 0.8, + "learning_rate": 3.671903052064632e-05, + "loss": 0.5934, + "step": 20763 + }, + { + "epoch": 0.8, + "learning_rate": 3.671838933059759e-05, + "loss": 0.4713, + "step": 20764 + }, + { + "epoch": 0.8, + "learning_rate": 3.6717748140548855e-05, + "loss": 0.6159, + "step": 20765 + }, + { + "epoch": 0.8, + "learning_rate": 3.671710695050013e-05, + "loss": 0.5531, + "step": 20766 + }, + { + "epoch": 0.8, + "learning_rate": 3.67164657604514e-05, + "loss": 0.4831, + "step": 20767 + }, + { + "epoch": 0.8, + "learning_rate": 3.671582457040267e-05, + "loss": 0.5393, + "step": 20768 + }, + { + "epoch": 0.8, + "learning_rate": 3.671518338035394e-05, + "loss": 0.453, + "step": 20769 + }, + { + "epoch": 0.8, + "learning_rate": 3.671454219030521e-05, + "loss": 0.4482, + "step": 20770 + }, + { + "epoch": 0.8, + "learning_rate": 3.671390100025648e-05, + "loss": 0.6134, + "step": 20771 + }, + { + "epoch": 0.8, + "learning_rate": 3.6713259810207743e-05, + "loss": 0.5178, + "step": 20772 + }, + { + "epoch": 0.8, + "learning_rate": 3.6712618620159015e-05, + "loss": 0.4769, + "step": 20773 + }, + { + "epoch": 0.8, + "learning_rate": 3.671197743011029e-05, + "loss": 0.47, + "step": 20774 + }, + { + "epoch": 0.8, + "learning_rate": 3.671133624006156e-05, + "loss": 0.5311, + "step": 20775 + }, + { + "epoch": 0.8, + "learning_rate": 3.671069505001283e-05, + "loss": 0.575, + "step": 20776 + }, + { + "epoch": 0.8, + "learning_rate": 3.6710053859964095e-05, + "loss": 0.5262, + "step": 20777 + }, + { + "epoch": 0.8, + "learning_rate": 3.670941266991537e-05, + "loss": 0.4563, + "step": 20778 + }, + { + "epoch": 0.8, + "learning_rate": 3.670877147986663e-05, + "loss": 0.4179, + "step": 20779 + }, + { + "epoch": 0.8, + "learning_rate": 3.6708130289817904e-05, + "loss": 0.5243, + "step": 20780 + }, + { + "epoch": 0.8, + "learning_rate": 3.670748909976917e-05, + "loss": 0.5322, + "step": 20781 + }, + { + "epoch": 0.8, + "learning_rate": 3.670684790972044e-05, + "loss": 0.5204, + "step": 20782 + }, + { + "epoch": 0.8, + "learning_rate": 3.670620671967171e-05, + "loss": 0.5411, + "step": 20783 + }, + { + "epoch": 0.8, + "learning_rate": 3.6705565529622984e-05, + "loss": 0.4423, + "step": 20784 + }, + { + "epoch": 0.8, + "learning_rate": 3.6704924339574255e-05, + "loss": 0.5142, + "step": 20785 + }, + { + "epoch": 0.8, + "learning_rate": 3.670428314952552e-05, + "loss": 0.4327, + "step": 20786 + }, + { + "epoch": 0.8, + "learning_rate": 3.670364195947679e-05, + "loss": 0.4788, + "step": 20787 + }, + { + "epoch": 0.8, + "learning_rate": 3.670300076942806e-05, + "loss": 0.4128, + "step": 20788 + }, + { + "epoch": 0.8, + "learning_rate": 3.670235957937933e-05, + "loss": 0.511, + "step": 20789 + }, + { + "epoch": 0.8, + "learning_rate": 3.67017183893306e-05, + "loss": 0.4313, + "step": 20790 + }, + { + "epoch": 0.8, + "learning_rate": 3.670107719928187e-05, + "loss": 0.4837, + "step": 20791 + }, + { + "epoch": 0.8, + "learning_rate": 3.6700436009233144e-05, + "loss": 0.5795, + "step": 20792 + }, + { + "epoch": 0.8, + "learning_rate": 3.669979481918441e-05, + "loss": 0.4756, + "step": 20793 + }, + { + "epoch": 0.8, + "learning_rate": 3.669915362913568e-05, + "loss": 0.4405, + "step": 20794 + }, + { + "epoch": 0.8, + "learning_rate": 3.6698512439086946e-05, + "loss": 0.6275, + "step": 20795 + }, + { + "epoch": 0.8, + "learning_rate": 3.669787124903822e-05, + "loss": 0.544, + "step": 20796 + }, + { + "epoch": 0.8, + "learning_rate": 3.669723005898948e-05, + "loss": 0.3817, + "step": 20797 + }, + { + "epoch": 0.8, + "learning_rate": 3.6696588868940754e-05, + "loss": 0.5993, + "step": 20798 + }, + { + "epoch": 0.8, + "learning_rate": 3.6695947678892026e-05, + "loss": 0.5478, + "step": 20799 + }, + { + "epoch": 0.8, + "learning_rate": 3.66953064888433e-05, + "loss": 0.4581, + "step": 20800 + }, + { + "epoch": 0.8, + "learning_rate": 3.669466529879457e-05, + "loss": 0.439, + "step": 20801 + }, + { + "epoch": 0.8, + "learning_rate": 3.6694024108745834e-05, + "loss": 0.4358, + "step": 20802 + }, + { + "epoch": 0.8, + "learning_rate": 3.6693382918697106e-05, + "loss": 0.5624, + "step": 20803 + }, + { + "epoch": 0.8, + "learning_rate": 3.669274172864837e-05, + "loss": 0.4098, + "step": 20804 + }, + { + "epoch": 0.8, + "learning_rate": 3.669210053859964e-05, + "loss": 0.6263, + "step": 20805 + }, + { + "epoch": 0.8, + "learning_rate": 3.669145934855091e-05, + "loss": 0.6269, + "step": 20806 + }, + { + "epoch": 0.8, + "learning_rate": 3.6690818158502186e-05, + "loss": 0.4076, + "step": 20807 + }, + { + "epoch": 0.8, + "learning_rate": 3.669017696845345e-05, + "loss": 0.589, + "step": 20808 + }, + { + "epoch": 0.8, + "learning_rate": 3.668953577840472e-05, + "loss": 0.4277, + "step": 20809 + }, + { + "epoch": 0.8, + "learning_rate": 3.6688894588355994e-05, + "loss": 0.5778, + "step": 20810 + }, + { + "epoch": 0.8, + "learning_rate": 3.668825339830726e-05, + "loss": 0.5077, + "step": 20811 + }, + { + "epoch": 0.8, + "learning_rate": 3.668761220825853e-05, + "loss": 0.4792, + "step": 20812 + }, + { + "epoch": 0.8, + "learning_rate": 3.6686971018209796e-05, + "loss": 0.62, + "step": 20813 + }, + { + "epoch": 0.8, + "learning_rate": 3.668632982816107e-05, + "loss": 0.5242, + "step": 20814 + }, + { + "epoch": 0.8, + "learning_rate": 3.668568863811234e-05, + "loss": 0.5214, + "step": 20815 + }, + { + "epoch": 0.8, + "learning_rate": 3.668504744806361e-05, + "loss": 0.4773, + "step": 20816 + }, + { + "epoch": 0.8, + "learning_rate": 3.6684406258014876e-05, + "loss": 0.5008, + "step": 20817 + }, + { + "epoch": 0.8, + "learning_rate": 3.668376506796615e-05, + "loss": 0.5412, + "step": 20818 + }, + { + "epoch": 0.8, + "learning_rate": 3.668312387791742e-05, + "loss": 0.4099, + "step": 20819 + }, + { + "epoch": 0.8, + "learning_rate": 3.6682482687868684e-05, + "loss": 0.5601, + "step": 20820 + }, + { + "epoch": 0.8, + "learning_rate": 3.6681841497819956e-05, + "loss": 0.4862, + "step": 20821 + }, + { + "epoch": 0.8, + "learning_rate": 3.668120030777122e-05, + "loss": 0.5012, + "step": 20822 + }, + { + "epoch": 0.8, + "learning_rate": 3.66805591177225e-05, + "loss": 0.5241, + "step": 20823 + }, + { + "epoch": 0.8, + "learning_rate": 3.6679917927673764e-05, + "loss": 0.4955, + "step": 20824 + }, + { + "epoch": 0.8, + "learning_rate": 3.6679276737625036e-05, + "loss": 0.5528, + "step": 20825 + }, + { + "epoch": 0.8, + "learning_rate": 3.66786355475763e-05, + "loss": 0.4706, + "step": 20826 + }, + { + "epoch": 0.8, + "learning_rate": 3.667799435752757e-05, + "loss": 0.4707, + "step": 20827 + }, + { + "epoch": 0.8, + "learning_rate": 3.6677353167478844e-05, + "loss": 0.4649, + "step": 20828 + }, + { + "epoch": 0.8, + "learning_rate": 3.667671197743011e-05, + "loss": 0.4944, + "step": 20829 + }, + { + "epoch": 0.8, + "learning_rate": 3.667607078738138e-05, + "loss": 0.4299, + "step": 20830 + }, + { + "epoch": 0.8, + "learning_rate": 3.667542959733265e-05, + "loss": 0.5888, + "step": 20831 + }, + { + "epoch": 0.8, + "learning_rate": 3.6674788407283925e-05, + "loss": 0.4762, + "step": 20832 + }, + { + "epoch": 0.8, + "learning_rate": 3.667414721723519e-05, + "loss": 0.4459, + "step": 20833 + }, + { + "epoch": 0.8, + "learning_rate": 3.667350602718646e-05, + "loss": 0.5241, + "step": 20834 + }, + { + "epoch": 0.8, + "learning_rate": 3.6672864837137726e-05, + "loss": 0.5305, + "step": 20835 + }, + { + "epoch": 0.8, + "learning_rate": 3.6672223647089e-05, + "loss": 0.436, + "step": 20836 + }, + { + "epoch": 0.8, + "learning_rate": 3.667158245704027e-05, + "loss": 0.4753, + "step": 20837 + }, + { + "epoch": 0.8, + "learning_rate": 3.6670941266991534e-05, + "loss": 0.4385, + "step": 20838 + }, + { + "epoch": 0.8, + "learning_rate": 3.6670300076942806e-05, + "loss": 0.5471, + "step": 20839 + }, + { + "epoch": 0.8, + "learning_rate": 3.666965888689408e-05, + "loss": 0.4905, + "step": 20840 + }, + { + "epoch": 0.8, + "learning_rate": 3.666901769684535e-05, + "loss": 0.5197, + "step": 20841 + }, + { + "epoch": 0.8, + "learning_rate": 3.6668376506796615e-05, + "loss": 0.6123, + "step": 20842 + }, + { + "epoch": 0.8, + "learning_rate": 3.6667735316747886e-05, + "loss": 0.4216, + "step": 20843 + }, + { + "epoch": 0.8, + "learning_rate": 3.666709412669915e-05, + "loss": 0.5928, + "step": 20844 + }, + { + "epoch": 0.8, + "learning_rate": 3.666645293665042e-05, + "loss": 0.608, + "step": 20845 + }, + { + "epoch": 0.8, + "learning_rate": 3.6665811746601695e-05, + "loss": 0.4658, + "step": 20846 + }, + { + "epoch": 0.8, + "learning_rate": 3.6665170556552966e-05, + "loss": 0.4753, + "step": 20847 + }, + { + "epoch": 0.8, + "learning_rate": 3.666452936650424e-05, + "loss": 0.5105, + "step": 20848 + }, + { + "epoch": 0.8, + "learning_rate": 3.66638881764555e-05, + "loss": 0.5289, + "step": 20849 + }, + { + "epoch": 0.8, + "learning_rate": 3.6663246986406775e-05, + "loss": 0.4938, + "step": 20850 + }, + { + "epoch": 0.8, + "learning_rate": 3.666260579635804e-05, + "loss": 0.4708, + "step": 20851 + }, + { + "epoch": 0.8, + "learning_rate": 3.666196460630931e-05, + "loss": 0.5323, + "step": 20852 + }, + { + "epoch": 0.8, + "learning_rate": 3.6661323416260576e-05, + "loss": 0.3787, + "step": 20853 + }, + { + "epoch": 0.8, + "learning_rate": 3.666068222621185e-05, + "loss": 0.4819, + "step": 20854 + }, + { + "epoch": 0.8, + "learning_rate": 3.666004103616312e-05, + "loss": 0.5453, + "step": 20855 + }, + { + "epoch": 0.8, + "learning_rate": 3.665939984611439e-05, + "loss": 0.5156, + "step": 20856 + }, + { + "epoch": 0.8, + "learning_rate": 3.665875865606566e-05, + "loss": 0.5134, + "step": 20857 + }, + { + "epoch": 0.8, + "learning_rate": 3.665811746601693e-05, + "loss": 0.5569, + "step": 20858 + }, + { + "epoch": 0.8, + "learning_rate": 3.66574762759682e-05, + "loss": 0.372, + "step": 20859 + }, + { + "epoch": 0.8, + "learning_rate": 3.6656835085919465e-05, + "loss": 0.5287, + "step": 20860 + }, + { + "epoch": 0.8, + "learning_rate": 3.6656193895870737e-05, + "loss": 0.5276, + "step": 20861 + }, + { + "epoch": 0.8, + "learning_rate": 3.665555270582201e-05, + "loss": 0.4597, + "step": 20862 + }, + { + "epoch": 0.8, + "learning_rate": 3.665491151577328e-05, + "loss": 0.5824, + "step": 20863 + }, + { + "epoch": 0.8, + "learning_rate": 3.665427032572455e-05, + "loss": 0.5201, + "step": 20864 + }, + { + "epoch": 0.8, + "learning_rate": 3.665362913567582e-05, + "loss": 0.5379, + "step": 20865 + }, + { + "epoch": 0.8, + "learning_rate": 3.665298794562709e-05, + "loss": 0.5621, + "step": 20866 + }, + { + "epoch": 0.8, + "learning_rate": 3.665234675557835e-05, + "loss": 0.4799, + "step": 20867 + }, + { + "epoch": 0.8, + "learning_rate": 3.6651705565529625e-05, + "loss": 0.5669, + "step": 20868 + }, + { + "epoch": 0.8, + "learning_rate": 3.665106437548089e-05, + "loss": 0.4754, + "step": 20869 + }, + { + "epoch": 0.8, + "learning_rate": 3.665042318543216e-05, + "loss": 0.5477, + "step": 20870 + }, + { + "epoch": 0.8, + "learning_rate": 3.664978199538343e-05, + "loss": 0.5229, + "step": 20871 + }, + { + "epoch": 0.8, + "learning_rate": 3.6649140805334705e-05, + "loss": 0.474, + "step": 20872 + }, + { + "epoch": 0.8, + "learning_rate": 3.664849961528598e-05, + "loss": 0.4685, + "step": 20873 + }, + { + "epoch": 0.8, + "learning_rate": 3.664785842523724e-05, + "loss": 0.5197, + "step": 20874 + }, + { + "epoch": 0.8, + "learning_rate": 3.6647217235188513e-05, + "loss": 0.493, + "step": 20875 + }, + { + "epoch": 0.8, + "learning_rate": 3.664657604513978e-05, + "loss": 0.425, + "step": 20876 + }, + { + "epoch": 0.8, + "learning_rate": 3.664593485509105e-05, + "loss": 0.4782, + "step": 20877 + }, + { + "epoch": 0.8, + "learning_rate": 3.6645293665042315e-05, + "loss": 0.5574, + "step": 20878 + }, + { + "epoch": 0.8, + "learning_rate": 3.6644652474993594e-05, + "loss": 0.5562, + "step": 20879 + }, + { + "epoch": 0.8, + "learning_rate": 3.664401128494486e-05, + "loss": 0.4441, + "step": 20880 + }, + { + "epoch": 0.8, + "learning_rate": 3.664337009489613e-05, + "loss": 0.456, + "step": 20881 + }, + { + "epoch": 0.8, + "learning_rate": 3.66427289048474e-05, + "loss": 0.4802, + "step": 20882 + }, + { + "epoch": 0.8, + "learning_rate": 3.664208771479867e-05, + "loss": 0.4556, + "step": 20883 + }, + { + "epoch": 0.8, + "learning_rate": 3.664144652474994e-05, + "loss": 0.517, + "step": 20884 + }, + { + "epoch": 0.8, + "learning_rate": 3.6640805334701204e-05, + "loss": 0.4931, + "step": 20885 + }, + { + "epoch": 0.8, + "learning_rate": 3.6640164144652475e-05, + "loss": 0.5181, + "step": 20886 + }, + { + "epoch": 0.8, + "learning_rate": 3.663952295460375e-05, + "loss": 0.5854, + "step": 20887 + }, + { + "epoch": 0.8, + "learning_rate": 3.663888176455502e-05, + "loss": 0.5977, + "step": 20888 + }, + { + "epoch": 0.8, + "learning_rate": 3.6638240574506284e-05, + "loss": 0.4875, + "step": 20889 + }, + { + "epoch": 0.8, + "learning_rate": 3.6637599384457555e-05, + "loss": 0.4706, + "step": 20890 + }, + { + "epoch": 0.8, + "learning_rate": 3.663695819440883e-05, + "loss": 0.4896, + "step": 20891 + }, + { + "epoch": 0.8, + "learning_rate": 3.663631700436009e-05, + "loss": 0.5255, + "step": 20892 + }, + { + "epoch": 0.8, + "learning_rate": 3.6635675814311364e-05, + "loss": 0.5114, + "step": 20893 + }, + { + "epoch": 0.8, + "learning_rate": 3.663503462426263e-05, + "loss": 0.4288, + "step": 20894 + }, + { + "epoch": 0.8, + "learning_rate": 3.663439343421391e-05, + "loss": 0.5667, + "step": 20895 + }, + { + "epoch": 0.8, + "learning_rate": 3.663375224416517e-05, + "loss": 0.572, + "step": 20896 + }, + { + "epoch": 0.8, + "learning_rate": 3.6633111054116444e-05, + "loss": 0.5375, + "step": 20897 + }, + { + "epoch": 0.8, + "learning_rate": 3.663246986406771e-05, + "loss": 0.4612, + "step": 20898 + }, + { + "epoch": 0.8, + "learning_rate": 3.663182867401898e-05, + "loss": 0.4678, + "step": 20899 + }, + { + "epoch": 0.8, + "learning_rate": 3.663118748397025e-05, + "loss": 0.4836, + "step": 20900 + }, + { + "epoch": 0.8, + "learning_rate": 3.663054629392152e-05, + "loss": 0.5218, + "step": 20901 + }, + { + "epoch": 0.8, + "learning_rate": 3.662990510387279e-05, + "loss": 0.4405, + "step": 20902 + }, + { + "epoch": 0.8, + "learning_rate": 3.662926391382406e-05, + "loss": 0.4919, + "step": 20903 + }, + { + "epoch": 0.8, + "learning_rate": 3.662862272377533e-05, + "loss": 0.3759, + "step": 20904 + }, + { + "epoch": 0.8, + "learning_rate": 3.66279815337266e-05, + "loss": 0.5262, + "step": 20905 + }, + { + "epoch": 0.8, + "learning_rate": 3.662734034367787e-05, + "loss": 0.5769, + "step": 20906 + }, + { + "epoch": 0.8, + "learning_rate": 3.6626699153629134e-05, + "loss": 0.4468, + "step": 20907 + }, + { + "epoch": 0.8, + "learning_rate": 3.6626057963580406e-05, + "loss": 0.5024, + "step": 20908 + }, + { + "epoch": 0.8, + "learning_rate": 3.662541677353168e-05, + "loss": 0.4656, + "step": 20909 + }, + { + "epoch": 0.8, + "learning_rate": 3.662477558348294e-05, + "loss": 0.5094, + "step": 20910 + }, + { + "epoch": 0.8, + "learning_rate": 3.662413439343422e-05, + "loss": 0.4117, + "step": 20911 + }, + { + "epoch": 0.8, + "learning_rate": 3.6623493203385486e-05, + "loss": 0.5086, + "step": 20912 + }, + { + "epoch": 0.8, + "learning_rate": 3.662285201333676e-05, + "loss": 0.4647, + "step": 20913 + }, + { + "epoch": 0.8, + "learning_rate": 3.662221082328802e-05, + "loss": 0.4894, + "step": 20914 + }, + { + "epoch": 0.8, + "learning_rate": 3.6621569633239294e-05, + "loss": 0.3399, + "step": 20915 + }, + { + "epoch": 0.8, + "learning_rate": 3.662092844319056e-05, + "loss": 0.5258, + "step": 20916 + }, + { + "epoch": 0.8, + "learning_rate": 3.662028725314183e-05, + "loss": 0.5232, + "step": 20917 + }, + { + "epoch": 0.8, + "learning_rate": 3.66196460630931e-05, + "loss": 0.4362, + "step": 20918 + }, + { + "epoch": 0.8, + "learning_rate": 3.6619004873044374e-05, + "loss": 0.5765, + "step": 20919 + }, + { + "epoch": 0.8, + "learning_rate": 3.6618363682995646e-05, + "loss": 0.6774, + "step": 20920 + }, + { + "epoch": 0.8, + "learning_rate": 3.661772249294691e-05, + "loss": 0.4702, + "step": 20921 + }, + { + "epoch": 0.8, + "learning_rate": 3.661708130289818e-05, + "loss": 0.5662, + "step": 20922 + }, + { + "epoch": 0.8, + "learning_rate": 3.661644011284945e-05, + "loss": 0.4948, + "step": 20923 + }, + { + "epoch": 0.8, + "learning_rate": 3.661579892280072e-05, + "loss": 0.4782, + "step": 20924 + }, + { + "epoch": 0.8, + "learning_rate": 3.6615157732751984e-05, + "loss": 0.696, + "step": 20925 + }, + { + "epoch": 0.8, + "learning_rate": 3.6614516542703256e-05, + "loss": 0.5293, + "step": 20926 + }, + { + "epoch": 0.8, + "learning_rate": 3.661387535265453e-05, + "loss": 0.4799, + "step": 20927 + }, + { + "epoch": 0.8, + "learning_rate": 3.66132341626058e-05, + "loss": 0.4354, + "step": 20928 + }, + { + "epoch": 0.8, + "learning_rate": 3.661259297255707e-05, + "loss": 0.4284, + "step": 20929 + }, + { + "epoch": 0.8, + "learning_rate": 3.6611951782508336e-05, + "loss": 0.6269, + "step": 20930 + }, + { + "epoch": 0.8, + "learning_rate": 3.661131059245961e-05, + "loss": 0.5636, + "step": 20931 + }, + { + "epoch": 0.8, + "learning_rate": 3.661066940241087e-05, + "loss": 0.5135, + "step": 20932 + }, + { + "epoch": 0.8, + "learning_rate": 3.6610028212362144e-05, + "loss": 0.5014, + "step": 20933 + }, + { + "epoch": 0.8, + "learning_rate": 3.6609387022313416e-05, + "loss": 0.4957, + "step": 20934 + }, + { + "epoch": 0.8, + "learning_rate": 3.660874583226469e-05, + "loss": 0.456, + "step": 20935 + }, + { + "epoch": 0.8, + "learning_rate": 3.660810464221596e-05, + "loss": 0.4696, + "step": 20936 + }, + { + "epoch": 0.8, + "learning_rate": 3.6607463452167224e-05, + "loss": 0.498, + "step": 20937 + }, + { + "epoch": 0.8, + "learning_rate": 3.6606822262118496e-05, + "loss": 0.3849, + "step": 20938 + }, + { + "epoch": 0.81, + "learning_rate": 3.660618107206976e-05, + "loss": 0.4205, + "step": 20939 + }, + { + "epoch": 0.81, + "learning_rate": 3.660553988202103e-05, + "loss": 0.4683, + "step": 20940 + }, + { + "epoch": 0.81, + "learning_rate": 3.66048986919723e-05, + "loss": 0.4583, + "step": 20941 + }, + { + "epoch": 0.81, + "learning_rate": 3.660425750192357e-05, + "loss": 0.4556, + "step": 20942 + }, + { + "epoch": 0.81, + "learning_rate": 3.660361631187484e-05, + "loss": 0.5478, + "step": 20943 + }, + { + "epoch": 0.81, + "learning_rate": 3.660297512182611e-05, + "loss": 0.4735, + "step": 20944 + }, + { + "epoch": 0.81, + "learning_rate": 3.6602333931777385e-05, + "loss": 0.6629, + "step": 20945 + }, + { + "epoch": 0.81, + "learning_rate": 3.660169274172865e-05, + "loss": 0.4996, + "step": 20946 + }, + { + "epoch": 0.81, + "learning_rate": 3.660105155167992e-05, + "loss": 0.509, + "step": 20947 + }, + { + "epoch": 0.81, + "learning_rate": 3.6600410361631186e-05, + "loss": 0.43, + "step": 20948 + }, + { + "epoch": 0.81, + "learning_rate": 3.659976917158246e-05, + "loss": 0.5075, + "step": 20949 + }, + { + "epoch": 0.81, + "learning_rate": 3.659912798153373e-05, + "loss": 0.4697, + "step": 20950 + }, + { + "epoch": 0.81, + "learning_rate": 3.6598486791485e-05, + "loss": 0.5164, + "step": 20951 + }, + { + "epoch": 0.81, + "learning_rate": 3.6597845601436266e-05, + "loss": 0.4835, + "step": 20952 + }, + { + "epoch": 0.81, + "learning_rate": 3.659720441138754e-05, + "loss": 0.6664, + "step": 20953 + }, + { + "epoch": 0.81, + "learning_rate": 3.659656322133881e-05, + "loss": 0.5388, + "step": 20954 + }, + { + "epoch": 0.81, + "learning_rate": 3.6595922031290075e-05, + "loss": 0.4946, + "step": 20955 + }, + { + "epoch": 0.81, + "learning_rate": 3.6595280841241346e-05, + "loss": 0.6053, + "step": 20956 + }, + { + "epoch": 0.81, + "learning_rate": 3.659463965119261e-05, + "loss": 0.5272, + "step": 20957 + }, + { + "epoch": 0.81, + "learning_rate": 3.659399846114388e-05, + "loss": 0.6516, + "step": 20958 + }, + { + "epoch": 0.81, + "learning_rate": 3.6593357271095155e-05, + "loss": 0.4735, + "step": 20959 + }, + { + "epoch": 0.81, + "learning_rate": 3.6592716081046426e-05, + "loss": 0.4024, + "step": 20960 + }, + { + "epoch": 0.81, + "learning_rate": 3.659207489099769e-05, + "loss": 0.5011, + "step": 20961 + }, + { + "epoch": 0.81, + "learning_rate": 3.659143370094896e-05, + "loss": 0.5511, + "step": 20962 + }, + { + "epoch": 0.81, + "learning_rate": 3.6590792510900235e-05, + "loss": 0.5017, + "step": 20963 + }, + { + "epoch": 0.81, + "learning_rate": 3.65901513208515e-05, + "loss": 0.4641, + "step": 20964 + }, + { + "epoch": 0.81, + "learning_rate": 3.658951013080277e-05, + "loss": 0.5155, + "step": 20965 + }, + { + "epoch": 0.81, + "learning_rate": 3.6588868940754036e-05, + "loss": 0.5023, + "step": 20966 + }, + { + "epoch": 0.81, + "learning_rate": 3.6588227750705315e-05, + "loss": 0.5558, + "step": 20967 + }, + { + "epoch": 0.81, + "learning_rate": 3.658758656065658e-05, + "loss": 0.6723, + "step": 20968 + }, + { + "epoch": 0.81, + "learning_rate": 3.658694537060785e-05, + "loss": 0.4098, + "step": 20969 + }, + { + "epoch": 0.81, + "learning_rate": 3.658630418055912e-05, + "loss": 0.5759, + "step": 20970 + }, + { + "epoch": 0.81, + "learning_rate": 3.658566299051039e-05, + "loss": 0.488, + "step": 20971 + }, + { + "epoch": 0.81, + "learning_rate": 3.658502180046166e-05, + "loss": 0.5584, + "step": 20972 + }, + { + "epoch": 0.81, + "learning_rate": 3.6584380610412925e-05, + "loss": 0.5037, + "step": 20973 + }, + { + "epoch": 0.81, + "learning_rate": 3.6583739420364197e-05, + "loss": 0.3427, + "step": 20974 + }, + { + "epoch": 0.81, + "learning_rate": 3.658309823031547e-05, + "loss": 0.563, + "step": 20975 + }, + { + "epoch": 0.81, + "learning_rate": 3.658245704026674e-05, + "loss": 0.5393, + "step": 20976 + }, + { + "epoch": 0.81, + "learning_rate": 3.6581815850218005e-05, + "loss": 0.5728, + "step": 20977 + }, + { + "epoch": 0.81, + "learning_rate": 3.658117466016928e-05, + "loss": 0.4489, + "step": 20978 + }, + { + "epoch": 0.81, + "learning_rate": 3.658053347012055e-05, + "loss": 0.5272, + "step": 20979 + }, + { + "epoch": 0.81, + "learning_rate": 3.657989228007181e-05, + "loss": 0.5655, + "step": 20980 + }, + { + "epoch": 0.81, + "learning_rate": 3.6579251090023085e-05, + "loss": 0.5497, + "step": 20981 + }, + { + "epoch": 0.81, + "learning_rate": 3.657860989997435e-05, + "loss": 0.6573, + "step": 20982 + }, + { + "epoch": 0.81, + "learning_rate": 3.657796870992563e-05, + "loss": 0.4722, + "step": 20983 + }, + { + "epoch": 0.81, + "learning_rate": 3.6577327519876893e-05, + "loss": 0.5174, + "step": 20984 + }, + { + "epoch": 0.81, + "learning_rate": 3.6576686329828165e-05, + "loss": 0.4227, + "step": 20985 + }, + { + "epoch": 0.81, + "learning_rate": 3.657604513977943e-05, + "loss": 0.5725, + "step": 20986 + }, + { + "epoch": 0.81, + "learning_rate": 3.65754039497307e-05, + "loss": 0.4905, + "step": 20987 + }, + { + "epoch": 0.81, + "learning_rate": 3.6574762759681973e-05, + "loss": 0.5262, + "step": 20988 + }, + { + "epoch": 0.81, + "learning_rate": 3.657412156963324e-05, + "loss": 0.5078, + "step": 20989 + }, + { + "epoch": 0.81, + "learning_rate": 3.657348037958451e-05, + "loss": 0.6219, + "step": 20990 + }, + { + "epoch": 0.81, + "learning_rate": 3.657283918953578e-05, + "loss": 0.4544, + "step": 20991 + }, + { + "epoch": 0.81, + "learning_rate": 3.6572197999487054e-05, + "loss": 0.5032, + "step": 20992 + }, + { + "epoch": 0.81, + "learning_rate": 3.657155680943832e-05, + "loss": 0.6141, + "step": 20993 + }, + { + "epoch": 0.81, + "learning_rate": 3.657091561938959e-05, + "loss": 0.5019, + "step": 20994 + }, + { + "epoch": 0.81, + "learning_rate": 3.6570274429340855e-05, + "loss": 0.6399, + "step": 20995 + }, + { + "epoch": 0.81, + "learning_rate": 3.656963323929213e-05, + "loss": 0.5159, + "step": 20996 + }, + { + "epoch": 0.81, + "learning_rate": 3.65689920492434e-05, + "loss": 0.4658, + "step": 20997 + }, + { + "epoch": 0.81, + "learning_rate": 3.6568350859194664e-05, + "loss": 0.5432, + "step": 20998 + }, + { + "epoch": 0.81, + "learning_rate": 3.6567709669145935e-05, + "loss": 0.4871, + "step": 20999 + }, + { + "epoch": 0.81, + "learning_rate": 3.656706847909721e-05, + "loss": 0.4721, + "step": 21000 + }, + { + "epoch": 0.81, + "learning_rate": 3.656642728904848e-05, + "loss": 0.4711, + "step": 21001 + }, + { + "epoch": 0.81, + "learning_rate": 3.6565786098999744e-05, + "loss": 0.566, + "step": 21002 + }, + { + "epoch": 0.81, + "learning_rate": 3.6565144908951015e-05, + "loss": 0.4451, + "step": 21003 + }, + { + "epoch": 0.81, + "learning_rate": 3.656450371890228e-05, + "loss": 0.5013, + "step": 21004 + }, + { + "epoch": 0.81, + "learning_rate": 3.656386252885355e-05, + "loss": 0.5524, + "step": 21005 + }, + { + "epoch": 0.81, + "learning_rate": 3.6563221338804824e-05, + "loss": 0.5569, + "step": 21006 + }, + { + "epoch": 0.81, + "learning_rate": 3.6562580148756095e-05, + "loss": 0.5157, + "step": 21007 + }, + { + "epoch": 0.81, + "learning_rate": 3.656193895870737e-05, + "loss": 0.4863, + "step": 21008 + }, + { + "epoch": 0.81, + "learning_rate": 3.656129776865863e-05, + "loss": 0.3977, + "step": 21009 + }, + { + "epoch": 0.81, + "learning_rate": 3.6560656578609904e-05, + "loss": 0.468, + "step": 21010 + }, + { + "epoch": 0.81, + "learning_rate": 3.656001538856117e-05, + "loss": 0.4308, + "step": 21011 + }, + { + "epoch": 0.81, + "learning_rate": 3.655937419851244e-05, + "loss": 0.5331, + "step": 21012 + }, + { + "epoch": 0.81, + "learning_rate": 3.6558733008463705e-05, + "loss": 0.3891, + "step": 21013 + }, + { + "epoch": 0.81, + "learning_rate": 3.655809181841498e-05, + "loss": 0.4181, + "step": 21014 + }, + { + "epoch": 0.81, + "learning_rate": 3.655745062836625e-05, + "loss": 0.4609, + "step": 21015 + }, + { + "epoch": 0.81, + "learning_rate": 3.655680943831752e-05, + "loss": 0.7033, + "step": 21016 + }, + { + "epoch": 0.81, + "learning_rate": 3.655616824826879e-05, + "loss": 0.413, + "step": 21017 + }, + { + "epoch": 0.81, + "learning_rate": 3.655552705822006e-05, + "loss": 0.4344, + "step": 21018 + }, + { + "epoch": 0.81, + "learning_rate": 3.655488586817133e-05, + "loss": 0.5355, + "step": 21019 + }, + { + "epoch": 0.81, + "learning_rate": 3.6554244678122594e-05, + "loss": 0.4724, + "step": 21020 + }, + { + "epoch": 0.81, + "learning_rate": 3.6553603488073866e-05, + "loss": 0.6142, + "step": 21021 + }, + { + "epoch": 0.81, + "learning_rate": 3.655296229802514e-05, + "loss": 0.5381, + "step": 21022 + }, + { + "epoch": 0.81, + "learning_rate": 3.655232110797641e-05, + "loss": 0.4515, + "step": 21023 + }, + { + "epoch": 0.81, + "learning_rate": 3.655167991792768e-05, + "loss": 0.5217, + "step": 21024 + }, + { + "epoch": 0.81, + "learning_rate": 3.6551038727878946e-05, + "loss": 0.4475, + "step": 21025 + }, + { + "epoch": 0.81, + "learning_rate": 3.655039753783022e-05, + "loss": 0.6222, + "step": 21026 + }, + { + "epoch": 0.81, + "learning_rate": 3.654975634778148e-05, + "loss": 0.6144, + "step": 21027 + }, + { + "epoch": 0.81, + "learning_rate": 3.6549115157732754e-05, + "loss": 0.4782, + "step": 21028 + }, + { + "epoch": 0.81, + "learning_rate": 3.654847396768402e-05, + "loss": 0.4492, + "step": 21029 + }, + { + "epoch": 0.81, + "learning_rate": 3.654783277763529e-05, + "loss": 0.4603, + "step": 21030 + }, + { + "epoch": 0.81, + "learning_rate": 3.654719158758656e-05, + "loss": 0.6045, + "step": 21031 + }, + { + "epoch": 0.81, + "learning_rate": 3.6546550397537834e-05, + "loss": 0.5112, + "step": 21032 + }, + { + "epoch": 0.81, + "learning_rate": 3.6545909207489106e-05, + "loss": 0.3952, + "step": 21033 + }, + { + "epoch": 0.81, + "learning_rate": 3.654526801744037e-05, + "loss": 0.5273, + "step": 21034 + }, + { + "epoch": 0.81, + "learning_rate": 3.654462682739164e-05, + "loss": 0.3758, + "step": 21035 + }, + { + "epoch": 0.81, + "learning_rate": 3.654398563734291e-05, + "loss": 0.5223, + "step": 21036 + }, + { + "epoch": 0.81, + "learning_rate": 3.654334444729418e-05, + "loss": 0.4877, + "step": 21037 + }, + { + "epoch": 0.81, + "learning_rate": 3.654270325724545e-05, + "loss": 0.4562, + "step": 21038 + }, + { + "epoch": 0.81, + "learning_rate": 3.654206206719672e-05, + "loss": 0.4767, + "step": 21039 + }, + { + "epoch": 0.81, + "learning_rate": 3.654142087714799e-05, + "loss": 0.5185, + "step": 21040 + }, + { + "epoch": 0.81, + "learning_rate": 3.654077968709926e-05, + "loss": 0.4409, + "step": 21041 + }, + { + "epoch": 0.81, + "learning_rate": 3.654013849705053e-05, + "loss": 0.5255, + "step": 21042 + }, + { + "epoch": 0.81, + "learning_rate": 3.6539497307001796e-05, + "loss": 0.5706, + "step": 21043 + }, + { + "epoch": 0.81, + "learning_rate": 3.653885611695307e-05, + "loss": 0.4845, + "step": 21044 + }, + { + "epoch": 0.81, + "learning_rate": 3.653821492690433e-05, + "loss": 0.4706, + "step": 21045 + }, + { + "epoch": 0.81, + "learning_rate": 3.6537573736855604e-05, + "loss": 0.5734, + "step": 21046 + }, + { + "epoch": 0.81, + "learning_rate": 3.6536932546806876e-05, + "loss": 0.4451, + "step": 21047 + }, + { + "epoch": 0.81, + "learning_rate": 3.653629135675815e-05, + "loss": 0.5485, + "step": 21048 + }, + { + "epoch": 0.81, + "learning_rate": 3.653565016670941e-05, + "loss": 0.4845, + "step": 21049 + }, + { + "epoch": 0.81, + "learning_rate": 3.6535008976660684e-05, + "loss": 0.3635, + "step": 21050 + }, + { + "epoch": 0.81, + "learning_rate": 3.6534367786611956e-05, + "loss": 0.47, + "step": 21051 + }, + { + "epoch": 0.81, + "learning_rate": 3.653372659656322e-05, + "loss": 0.5354, + "step": 21052 + }, + { + "epoch": 0.81, + "learning_rate": 3.653308540651449e-05, + "loss": 0.4554, + "step": 21053 + }, + { + "epoch": 0.81, + "learning_rate": 3.653244421646576e-05, + "loss": 0.4829, + "step": 21054 + }, + { + "epoch": 0.81, + "learning_rate": 3.6531803026417036e-05, + "loss": 0.4679, + "step": 21055 + }, + { + "epoch": 0.81, + "learning_rate": 3.65311618363683e-05, + "loss": 0.5078, + "step": 21056 + }, + { + "epoch": 0.81, + "learning_rate": 3.653052064631957e-05, + "loss": 0.528, + "step": 21057 + }, + { + "epoch": 0.81, + "learning_rate": 3.652987945627084e-05, + "loss": 0.5009, + "step": 21058 + }, + { + "epoch": 0.81, + "learning_rate": 3.652923826622211e-05, + "loss": 0.5237, + "step": 21059 + }, + { + "epoch": 0.81, + "learning_rate": 3.652859707617338e-05, + "loss": 0.4181, + "step": 21060 + }, + { + "epoch": 0.81, + "learning_rate": 3.6527955886124646e-05, + "loss": 0.4376, + "step": 21061 + }, + { + "epoch": 0.81, + "learning_rate": 3.652731469607592e-05, + "loss": 0.5488, + "step": 21062 + }, + { + "epoch": 0.81, + "learning_rate": 3.652667350602719e-05, + "loss": 0.5076, + "step": 21063 + }, + { + "epoch": 0.81, + "learning_rate": 3.652603231597846e-05, + "loss": 0.6773, + "step": 21064 + }, + { + "epoch": 0.81, + "learning_rate": 3.6525391125929726e-05, + "loss": 0.4767, + "step": 21065 + }, + { + "epoch": 0.81, + "learning_rate": 3.6524749935881e-05, + "loss": 0.4761, + "step": 21066 + }, + { + "epoch": 0.81, + "learning_rate": 3.652410874583226e-05, + "loss": 0.4099, + "step": 21067 + }, + { + "epoch": 0.81, + "learning_rate": 3.6523467555783535e-05, + "loss": 0.4918, + "step": 21068 + }, + { + "epoch": 0.81, + "learning_rate": 3.6522826365734806e-05, + "loss": 0.4868, + "step": 21069 + }, + { + "epoch": 0.81, + "learning_rate": 3.652218517568607e-05, + "loss": 0.5656, + "step": 21070 + }, + { + "epoch": 0.81, + "learning_rate": 3.652154398563735e-05, + "loss": 0.5064, + "step": 21071 + }, + { + "epoch": 0.81, + "learning_rate": 3.6520902795588615e-05, + "loss": 0.4237, + "step": 21072 + }, + { + "epoch": 0.81, + "learning_rate": 3.6520261605539886e-05, + "loss": 0.4284, + "step": 21073 + }, + { + "epoch": 0.81, + "learning_rate": 3.651962041549115e-05, + "loss": 0.4191, + "step": 21074 + }, + { + "epoch": 0.81, + "learning_rate": 3.651897922544242e-05, + "loss": 0.5406, + "step": 21075 + }, + { + "epoch": 0.81, + "learning_rate": 3.651833803539369e-05, + "loss": 0.4981, + "step": 21076 + }, + { + "epoch": 0.81, + "learning_rate": 3.651769684534496e-05, + "loss": 0.5233, + "step": 21077 + }, + { + "epoch": 0.81, + "learning_rate": 3.651705565529623e-05, + "loss": 0.4693, + "step": 21078 + }, + { + "epoch": 0.81, + "learning_rate": 3.65164144652475e-05, + "loss": 0.3778, + "step": 21079 + }, + { + "epoch": 0.81, + "learning_rate": 3.6515773275198775e-05, + "loss": 0.3055, + "step": 21080 + }, + { + "epoch": 0.81, + "learning_rate": 3.651513208515004e-05, + "loss": 0.535, + "step": 21081 + }, + { + "epoch": 0.81, + "learning_rate": 3.651449089510131e-05, + "loss": 0.6207, + "step": 21082 + }, + { + "epoch": 0.81, + "learning_rate": 3.6513849705052577e-05, + "loss": 0.6576, + "step": 21083 + }, + { + "epoch": 0.81, + "learning_rate": 3.651320851500385e-05, + "loss": 0.5923, + "step": 21084 + }, + { + "epoch": 0.81, + "learning_rate": 3.651256732495511e-05, + "loss": 0.4279, + "step": 21085 + }, + { + "epoch": 0.81, + "learning_rate": 3.6511926134906385e-05, + "loss": 0.4467, + "step": 21086 + }, + { + "epoch": 0.81, + "learning_rate": 3.6511284944857657e-05, + "loss": 0.5267, + "step": 21087 + }, + { + "epoch": 0.81, + "learning_rate": 3.651064375480893e-05, + "loss": 0.5536, + "step": 21088 + }, + { + "epoch": 0.81, + "learning_rate": 3.65100025647602e-05, + "loss": 0.5004, + "step": 21089 + }, + { + "epoch": 0.81, + "learning_rate": 3.6509361374711465e-05, + "loss": 0.4588, + "step": 21090 + }, + { + "epoch": 0.81, + "learning_rate": 3.650872018466274e-05, + "loss": 0.6669, + "step": 21091 + }, + { + "epoch": 0.81, + "learning_rate": 3.6508078994614e-05, + "loss": 0.4604, + "step": 21092 + }, + { + "epoch": 0.81, + "learning_rate": 3.650743780456527e-05, + "loss": 0.5107, + "step": 21093 + }, + { + "epoch": 0.81, + "learning_rate": 3.6506796614516545e-05, + "loss": 0.4602, + "step": 21094 + }, + { + "epoch": 0.81, + "learning_rate": 3.650615542446782e-05, + "loss": 0.3994, + "step": 21095 + }, + { + "epoch": 0.81, + "learning_rate": 3.650551423441909e-05, + "loss": 0.519, + "step": 21096 + }, + { + "epoch": 0.81, + "learning_rate": 3.6504873044370353e-05, + "loss": 0.4432, + "step": 21097 + }, + { + "epoch": 0.81, + "learning_rate": 3.6504231854321625e-05, + "loss": 0.5045, + "step": 21098 + }, + { + "epoch": 0.81, + "learning_rate": 3.650359066427289e-05, + "loss": 0.4954, + "step": 21099 + }, + { + "epoch": 0.81, + "learning_rate": 3.650294947422416e-05, + "loss": 0.5464, + "step": 21100 + }, + { + "epoch": 0.81, + "learning_rate": 3.650230828417543e-05, + "loss": 0.5078, + "step": 21101 + }, + { + "epoch": 0.81, + "learning_rate": 3.65016670941267e-05, + "loss": 0.467, + "step": 21102 + }, + { + "epoch": 0.81, + "learning_rate": 3.650102590407797e-05, + "loss": 0.5862, + "step": 21103 + }, + { + "epoch": 0.81, + "learning_rate": 3.650038471402924e-05, + "loss": 0.4366, + "step": 21104 + }, + { + "epoch": 0.81, + "learning_rate": 3.6499743523980514e-05, + "loss": 0.4978, + "step": 21105 + }, + { + "epoch": 0.81, + "learning_rate": 3.649910233393178e-05, + "loss": 0.5823, + "step": 21106 + }, + { + "epoch": 0.81, + "learning_rate": 3.649846114388305e-05, + "loss": 0.4957, + "step": 21107 + }, + { + "epoch": 0.81, + "learning_rate": 3.6497819953834315e-05, + "loss": 0.546, + "step": 21108 + }, + { + "epoch": 0.81, + "learning_rate": 3.649717876378559e-05, + "loss": 0.4148, + "step": 21109 + }, + { + "epoch": 0.81, + "learning_rate": 3.649653757373686e-05, + "loss": 0.4491, + "step": 21110 + }, + { + "epoch": 0.81, + "learning_rate": 3.649589638368813e-05, + "loss": 0.4857, + "step": 21111 + }, + { + "epoch": 0.81, + "learning_rate": 3.6495255193639395e-05, + "loss": 0.5755, + "step": 21112 + }, + { + "epoch": 0.81, + "learning_rate": 3.649461400359067e-05, + "loss": 0.4463, + "step": 21113 + }, + { + "epoch": 0.81, + "learning_rate": 3.649397281354194e-05, + "loss": 0.4693, + "step": 21114 + }, + { + "epoch": 0.81, + "learning_rate": 3.6493331623493204e-05, + "loss": 0.5164, + "step": 21115 + }, + { + "epoch": 0.81, + "learning_rate": 3.6492690433444475e-05, + "loss": 0.53, + "step": 21116 + }, + { + "epoch": 0.81, + "learning_rate": 3.649204924339574e-05, + "loss": 0.5679, + "step": 21117 + }, + { + "epoch": 0.81, + "learning_rate": 3.649140805334701e-05, + "loss": 0.4734, + "step": 21118 + }, + { + "epoch": 0.81, + "learning_rate": 3.6490766863298284e-05, + "loss": 0.5064, + "step": 21119 + }, + { + "epoch": 0.81, + "learning_rate": 3.6490125673249555e-05, + "loss": 0.5405, + "step": 21120 + }, + { + "epoch": 0.81, + "learning_rate": 3.648948448320082e-05, + "loss": 0.465, + "step": 21121 + }, + { + "epoch": 0.81, + "learning_rate": 3.648884329315209e-05, + "loss": 0.4327, + "step": 21122 + }, + { + "epoch": 0.81, + "learning_rate": 3.6488202103103364e-05, + "loss": 0.5769, + "step": 21123 + }, + { + "epoch": 0.81, + "learning_rate": 3.648756091305463e-05, + "loss": 0.5865, + "step": 21124 + }, + { + "epoch": 0.81, + "learning_rate": 3.64869197230059e-05, + "loss": 0.4804, + "step": 21125 + }, + { + "epoch": 0.81, + "learning_rate": 3.6486278532957165e-05, + "loss": 0.4617, + "step": 21126 + }, + { + "epoch": 0.81, + "learning_rate": 3.6485637342908444e-05, + "loss": 0.6427, + "step": 21127 + }, + { + "epoch": 0.81, + "learning_rate": 3.648499615285971e-05, + "loss": 0.4148, + "step": 21128 + }, + { + "epoch": 0.81, + "learning_rate": 3.648435496281098e-05, + "loss": 0.4505, + "step": 21129 + }, + { + "epoch": 0.81, + "learning_rate": 3.6483713772762246e-05, + "loss": 0.4559, + "step": 21130 + }, + { + "epoch": 0.81, + "learning_rate": 3.648307258271352e-05, + "loss": 0.403, + "step": 21131 + }, + { + "epoch": 0.81, + "learning_rate": 3.648243139266479e-05, + "loss": 0.4758, + "step": 21132 + }, + { + "epoch": 0.81, + "learning_rate": 3.6481790202616054e-05, + "loss": 0.4856, + "step": 21133 + }, + { + "epoch": 0.81, + "learning_rate": 3.6481149012567326e-05, + "loss": 0.4694, + "step": 21134 + }, + { + "epoch": 0.81, + "learning_rate": 3.64805078225186e-05, + "loss": 0.5233, + "step": 21135 + }, + { + "epoch": 0.81, + "learning_rate": 3.647986663246987e-05, + "loss": 0.5232, + "step": 21136 + }, + { + "epoch": 0.81, + "learning_rate": 3.6479225442421134e-05, + "loss": 0.7223, + "step": 21137 + }, + { + "epoch": 0.81, + "learning_rate": 3.6478584252372406e-05, + "loss": 0.3198, + "step": 21138 + }, + { + "epoch": 0.81, + "learning_rate": 3.647794306232367e-05, + "loss": 0.4952, + "step": 21139 + }, + { + "epoch": 0.81, + "learning_rate": 3.647730187227494e-05, + "loss": 0.501, + "step": 21140 + }, + { + "epoch": 0.81, + "learning_rate": 3.6476660682226214e-05, + "loss": 0.5204, + "step": 21141 + }, + { + "epoch": 0.81, + "learning_rate": 3.647601949217748e-05, + "loss": 0.5256, + "step": 21142 + }, + { + "epoch": 0.81, + "learning_rate": 3.647537830212876e-05, + "loss": 0.4484, + "step": 21143 + }, + { + "epoch": 0.81, + "learning_rate": 3.647473711208002e-05, + "loss": 0.5026, + "step": 21144 + }, + { + "epoch": 0.81, + "learning_rate": 3.6474095922031294e-05, + "loss": 0.4265, + "step": 21145 + }, + { + "epoch": 0.81, + "learning_rate": 3.647345473198256e-05, + "loss": 0.3949, + "step": 21146 + }, + { + "epoch": 0.81, + "learning_rate": 3.647281354193383e-05, + "loss": 0.4285, + "step": 21147 + }, + { + "epoch": 0.81, + "learning_rate": 3.6472172351885096e-05, + "loss": 0.4726, + "step": 21148 + }, + { + "epoch": 0.81, + "learning_rate": 3.647153116183637e-05, + "loss": 0.5445, + "step": 21149 + }, + { + "epoch": 0.81, + "learning_rate": 3.647088997178764e-05, + "loss": 0.5919, + "step": 21150 + }, + { + "epoch": 0.81, + "learning_rate": 3.647024878173891e-05, + "loss": 0.619, + "step": 21151 + }, + { + "epoch": 0.81, + "learning_rate": 3.646960759169018e-05, + "loss": 0.4186, + "step": 21152 + }, + { + "epoch": 0.81, + "learning_rate": 3.646896640164145e-05, + "loss": 0.5818, + "step": 21153 + }, + { + "epoch": 0.81, + "learning_rate": 3.646832521159272e-05, + "loss": 0.4795, + "step": 21154 + }, + { + "epoch": 0.81, + "learning_rate": 3.6467684021543984e-05, + "loss": 0.5009, + "step": 21155 + }, + { + "epoch": 0.81, + "learning_rate": 3.6467042831495256e-05, + "loss": 0.474, + "step": 21156 + }, + { + "epoch": 0.81, + "learning_rate": 3.646640164144653e-05, + "loss": 0.4528, + "step": 21157 + }, + { + "epoch": 0.81, + "learning_rate": 3.646576045139779e-05, + "loss": 0.4144, + "step": 21158 + }, + { + "epoch": 0.81, + "learning_rate": 3.646511926134907e-05, + "loss": 0.5571, + "step": 21159 + }, + { + "epoch": 0.81, + "learning_rate": 3.6464478071300336e-05, + "loss": 0.5234, + "step": 21160 + }, + { + "epoch": 0.81, + "learning_rate": 3.646383688125161e-05, + "loss": 0.6214, + "step": 21161 + }, + { + "epoch": 0.81, + "learning_rate": 3.646319569120287e-05, + "loss": 0.4679, + "step": 21162 + }, + { + "epoch": 0.81, + "learning_rate": 3.6462554501154144e-05, + "loss": 0.4768, + "step": 21163 + }, + { + "epoch": 0.81, + "learning_rate": 3.646191331110541e-05, + "loss": 0.468, + "step": 21164 + }, + { + "epoch": 0.81, + "learning_rate": 3.646127212105668e-05, + "loss": 0.5692, + "step": 21165 + }, + { + "epoch": 0.81, + "learning_rate": 3.646063093100795e-05, + "loss": 0.6387, + "step": 21166 + }, + { + "epoch": 0.81, + "learning_rate": 3.6459989740959225e-05, + "loss": 0.5026, + "step": 21167 + }, + { + "epoch": 0.81, + "learning_rate": 3.6459348550910496e-05, + "loss": 0.6306, + "step": 21168 + }, + { + "epoch": 0.81, + "learning_rate": 3.645870736086176e-05, + "loss": 0.5095, + "step": 21169 + }, + { + "epoch": 0.81, + "learning_rate": 3.645806617081303e-05, + "loss": 0.4994, + "step": 21170 + }, + { + "epoch": 0.81, + "learning_rate": 3.64574249807643e-05, + "loss": 0.5904, + "step": 21171 + }, + { + "epoch": 0.81, + "learning_rate": 3.645678379071557e-05, + "loss": 0.5922, + "step": 21172 + }, + { + "epoch": 0.81, + "learning_rate": 3.6456142600666834e-05, + "loss": 0.4125, + "step": 21173 + }, + { + "epoch": 0.81, + "learning_rate": 3.6455501410618106e-05, + "loss": 0.5994, + "step": 21174 + }, + { + "epoch": 0.81, + "learning_rate": 3.645486022056938e-05, + "loss": 0.5657, + "step": 21175 + }, + { + "epoch": 0.81, + "learning_rate": 3.645421903052065e-05, + "loss": 0.5825, + "step": 21176 + }, + { + "epoch": 0.81, + "learning_rate": 3.645357784047192e-05, + "loss": 0.5213, + "step": 21177 + }, + { + "epoch": 0.81, + "learning_rate": 3.6452936650423186e-05, + "loss": 0.4199, + "step": 21178 + }, + { + "epoch": 0.81, + "learning_rate": 3.645229546037446e-05, + "loss": 0.576, + "step": 21179 + }, + { + "epoch": 0.81, + "learning_rate": 3.645165427032572e-05, + "loss": 0.5239, + "step": 21180 + }, + { + "epoch": 0.81, + "learning_rate": 3.6451013080276995e-05, + "loss": 0.6281, + "step": 21181 + }, + { + "epoch": 0.81, + "learning_rate": 3.6450371890228266e-05, + "loss": 0.5312, + "step": 21182 + }, + { + "epoch": 0.81, + "learning_rate": 3.644973070017954e-05, + "loss": 0.4697, + "step": 21183 + }, + { + "epoch": 0.81, + "learning_rate": 3.644908951013081e-05, + "loss": 0.3999, + "step": 21184 + }, + { + "epoch": 0.81, + "learning_rate": 3.6448448320082075e-05, + "loss": 0.4753, + "step": 21185 + }, + { + "epoch": 0.81, + "learning_rate": 3.6447807130033346e-05, + "loss": 0.5295, + "step": 21186 + }, + { + "epoch": 0.81, + "learning_rate": 3.644716593998461e-05, + "loss": 0.5282, + "step": 21187 + }, + { + "epoch": 0.81, + "learning_rate": 3.644652474993588e-05, + "loss": 0.4985, + "step": 21188 + }, + { + "epoch": 0.81, + "learning_rate": 3.644588355988715e-05, + "loss": 0.4874, + "step": 21189 + }, + { + "epoch": 0.81, + "learning_rate": 3.644524236983842e-05, + "loss": 0.5568, + "step": 21190 + }, + { + "epoch": 0.81, + "learning_rate": 3.644460117978969e-05, + "loss": 0.5765, + "step": 21191 + }, + { + "epoch": 0.81, + "learning_rate": 3.644395998974096e-05, + "loss": 0.4523, + "step": 21192 + }, + { + "epoch": 0.81, + "learning_rate": 3.6443318799692235e-05, + "loss": 0.4967, + "step": 21193 + }, + { + "epoch": 0.81, + "learning_rate": 3.64426776096435e-05, + "loss": 0.4925, + "step": 21194 + }, + { + "epoch": 0.81, + "learning_rate": 3.644203641959477e-05, + "loss": 0.4554, + "step": 21195 + }, + { + "epoch": 0.81, + "learning_rate": 3.6441395229546037e-05, + "loss": 0.606, + "step": 21196 + }, + { + "epoch": 0.81, + "learning_rate": 3.644075403949731e-05, + "loss": 0.5247, + "step": 21197 + }, + { + "epoch": 0.81, + "learning_rate": 3.644011284944858e-05, + "loss": 0.5385, + "step": 21198 + }, + { + "epoch": 0.82, + "learning_rate": 3.643947165939985e-05, + "loss": 0.4652, + "step": 21199 + }, + { + "epoch": 0.82, + "learning_rate": 3.643883046935112e-05, + "loss": 0.6022, + "step": 21200 + }, + { + "epoch": 0.82, + "learning_rate": 3.643818927930239e-05, + "loss": 0.4768, + "step": 21201 + }, + { + "epoch": 0.82, + "learning_rate": 3.643754808925366e-05, + "loss": 0.4578, + "step": 21202 + }, + { + "epoch": 0.82, + "learning_rate": 3.6436906899204925e-05, + "loss": 0.4731, + "step": 21203 + }, + { + "epoch": 0.82, + "learning_rate": 3.64362657091562e-05, + "loss": 0.4126, + "step": 21204 + }, + { + "epoch": 0.82, + "learning_rate": 3.643562451910746e-05, + "loss": 0.5366, + "step": 21205 + }, + { + "epoch": 0.82, + "learning_rate": 3.643498332905873e-05, + "loss": 0.5233, + "step": 21206 + }, + { + "epoch": 0.82, + "learning_rate": 3.6434342139010005e-05, + "loss": 0.5896, + "step": 21207 + }, + { + "epoch": 0.82, + "learning_rate": 3.643370094896128e-05, + "loss": 0.3822, + "step": 21208 + }, + { + "epoch": 0.82, + "learning_rate": 3.643305975891254e-05, + "loss": 0.4421, + "step": 21209 + }, + { + "epoch": 0.82, + "learning_rate": 3.6432418568863813e-05, + "loss": 0.5013, + "step": 21210 + }, + { + "epoch": 0.82, + "learning_rate": 3.6431777378815085e-05, + "loss": 0.5032, + "step": 21211 + }, + { + "epoch": 0.82, + "learning_rate": 3.643113618876635e-05, + "loss": 0.4031, + "step": 21212 + }, + { + "epoch": 0.82, + "learning_rate": 3.643049499871762e-05, + "loss": 0.5013, + "step": 21213 + }, + { + "epoch": 0.82, + "learning_rate": 3.642985380866889e-05, + "loss": 0.4674, + "step": 21214 + }, + { + "epoch": 0.82, + "learning_rate": 3.6429212618620165e-05, + "loss": 0.4777, + "step": 21215 + }, + { + "epoch": 0.82, + "learning_rate": 3.642857142857143e-05, + "loss": 0.5388, + "step": 21216 + }, + { + "epoch": 0.82, + "learning_rate": 3.64279302385227e-05, + "loss": 0.5159, + "step": 21217 + }, + { + "epoch": 0.82, + "learning_rate": 3.642728904847397e-05, + "loss": 0.5049, + "step": 21218 + }, + { + "epoch": 0.82, + "learning_rate": 3.642664785842524e-05, + "loss": 0.4691, + "step": 21219 + }, + { + "epoch": 0.82, + "learning_rate": 3.642600666837651e-05, + "loss": 0.4375, + "step": 21220 + }, + { + "epoch": 0.82, + "learning_rate": 3.6425365478327775e-05, + "loss": 0.4781, + "step": 21221 + }, + { + "epoch": 0.82, + "learning_rate": 3.642472428827905e-05, + "loss": 0.5945, + "step": 21222 + }, + { + "epoch": 0.82, + "learning_rate": 3.642408309823032e-05, + "loss": 0.4421, + "step": 21223 + }, + { + "epoch": 0.82, + "learning_rate": 3.642344190818159e-05, + "loss": 0.4851, + "step": 21224 + }, + { + "epoch": 0.82, + "learning_rate": 3.6422800718132855e-05, + "loss": 0.5876, + "step": 21225 + }, + { + "epoch": 0.82, + "learning_rate": 3.642215952808413e-05, + "loss": 0.6373, + "step": 21226 + }, + { + "epoch": 0.82, + "learning_rate": 3.642151833803539e-05, + "loss": 0.5682, + "step": 21227 + }, + { + "epoch": 0.82, + "learning_rate": 3.6420877147986664e-05, + "loss": 0.4753, + "step": 21228 + }, + { + "epoch": 0.82, + "learning_rate": 3.6420235957937935e-05, + "loss": 0.4714, + "step": 21229 + }, + { + "epoch": 0.82, + "learning_rate": 3.64195947678892e-05, + "loss": 0.4918, + "step": 21230 + }, + { + "epoch": 0.82, + "learning_rate": 3.641895357784048e-05, + "loss": 0.5962, + "step": 21231 + }, + { + "epoch": 0.82, + "learning_rate": 3.6418312387791744e-05, + "loss": 0.5606, + "step": 21232 + }, + { + "epoch": 0.82, + "learning_rate": 3.6417671197743016e-05, + "loss": 0.4884, + "step": 21233 + }, + { + "epoch": 0.82, + "learning_rate": 3.641703000769428e-05, + "loss": 0.5505, + "step": 21234 + }, + { + "epoch": 0.82, + "learning_rate": 3.641638881764555e-05, + "loss": 0.5275, + "step": 21235 + }, + { + "epoch": 0.82, + "learning_rate": 3.641574762759682e-05, + "loss": 0.4796, + "step": 21236 + }, + { + "epoch": 0.82, + "learning_rate": 3.641510643754809e-05, + "loss": 0.4066, + "step": 21237 + }, + { + "epoch": 0.82, + "learning_rate": 3.641446524749936e-05, + "loss": 0.4401, + "step": 21238 + }, + { + "epoch": 0.82, + "learning_rate": 3.641382405745063e-05, + "loss": 0.4845, + "step": 21239 + }, + { + "epoch": 0.82, + "learning_rate": 3.6413182867401904e-05, + "loss": 0.5201, + "step": 21240 + }, + { + "epoch": 0.82, + "learning_rate": 3.641254167735317e-05, + "loss": 0.4754, + "step": 21241 + }, + { + "epoch": 0.82, + "learning_rate": 3.641190048730444e-05, + "loss": 0.4764, + "step": 21242 + }, + { + "epoch": 0.82, + "learning_rate": 3.6411259297255706e-05, + "loss": 0.5524, + "step": 21243 + }, + { + "epoch": 0.82, + "learning_rate": 3.641061810720698e-05, + "loss": 0.4531, + "step": 21244 + }, + { + "epoch": 0.82, + "learning_rate": 3.640997691715824e-05, + "loss": 0.4682, + "step": 21245 + }, + { + "epoch": 0.82, + "learning_rate": 3.6409335727109514e-05, + "loss": 0.5502, + "step": 21246 + }, + { + "epoch": 0.82, + "learning_rate": 3.6408694537060786e-05, + "loss": 0.4941, + "step": 21247 + }, + { + "epoch": 0.82, + "learning_rate": 3.640805334701206e-05, + "loss": 0.494, + "step": 21248 + }, + { + "epoch": 0.82, + "learning_rate": 3.640741215696333e-05, + "loss": 0.5037, + "step": 21249 + }, + { + "epoch": 0.82, + "learning_rate": 3.6406770966914594e-05, + "loss": 0.627, + "step": 21250 + }, + { + "epoch": 0.82, + "learning_rate": 3.6406129776865866e-05, + "loss": 0.4336, + "step": 21251 + }, + { + "epoch": 0.82, + "learning_rate": 3.640548858681713e-05, + "loss": 0.5805, + "step": 21252 + }, + { + "epoch": 0.82, + "learning_rate": 3.64048473967684e-05, + "loss": 0.5405, + "step": 21253 + }, + { + "epoch": 0.82, + "learning_rate": 3.6404206206719674e-05, + "loss": 0.6005, + "step": 21254 + }, + { + "epoch": 0.82, + "learning_rate": 3.6403565016670946e-05, + "loss": 0.4348, + "step": 21255 + }, + { + "epoch": 0.82, + "learning_rate": 3.640292382662222e-05, + "loss": 0.4751, + "step": 21256 + }, + { + "epoch": 0.82, + "learning_rate": 3.640228263657348e-05, + "loss": 0.5367, + "step": 21257 + }, + { + "epoch": 0.82, + "learning_rate": 3.6401641446524754e-05, + "loss": 0.4562, + "step": 21258 + }, + { + "epoch": 0.82, + "learning_rate": 3.640100025647602e-05, + "loss": 0.5924, + "step": 21259 + }, + { + "epoch": 0.82, + "learning_rate": 3.640035906642729e-05, + "loss": 0.3893, + "step": 21260 + }, + { + "epoch": 0.82, + "learning_rate": 3.6399717876378556e-05, + "loss": 0.5289, + "step": 21261 + }, + { + "epoch": 0.82, + "learning_rate": 3.639907668632983e-05, + "loss": 0.4124, + "step": 21262 + }, + { + "epoch": 0.82, + "learning_rate": 3.63984354962811e-05, + "loss": 0.4334, + "step": 21263 + }, + { + "epoch": 0.82, + "learning_rate": 3.639779430623237e-05, + "loss": 0.4517, + "step": 21264 + }, + { + "epoch": 0.82, + "learning_rate": 3.639715311618364e-05, + "loss": 0.4868, + "step": 21265 + }, + { + "epoch": 0.82, + "learning_rate": 3.639651192613491e-05, + "loss": 0.5154, + "step": 21266 + }, + { + "epoch": 0.82, + "learning_rate": 3.639587073608618e-05, + "loss": 0.5721, + "step": 21267 + }, + { + "epoch": 0.82, + "learning_rate": 3.6395229546037444e-05, + "loss": 0.4224, + "step": 21268 + }, + { + "epoch": 0.82, + "learning_rate": 3.6394588355988716e-05, + "loss": 0.4789, + "step": 21269 + }, + { + "epoch": 0.82, + "learning_rate": 3.639394716593999e-05, + "loss": 0.5674, + "step": 21270 + }, + { + "epoch": 0.82, + "learning_rate": 3.639330597589126e-05, + "loss": 0.607, + "step": 21271 + }, + { + "epoch": 0.82, + "learning_rate": 3.6392664785842524e-05, + "loss": 0.4678, + "step": 21272 + }, + { + "epoch": 0.82, + "learning_rate": 3.6392023595793796e-05, + "loss": 0.6064, + "step": 21273 + }, + { + "epoch": 0.82, + "learning_rate": 3.639138240574507e-05, + "loss": 0.4693, + "step": 21274 + }, + { + "epoch": 0.82, + "learning_rate": 3.639074121569633e-05, + "loss": 0.4816, + "step": 21275 + }, + { + "epoch": 0.82, + "learning_rate": 3.6390100025647604e-05, + "loss": 0.4292, + "step": 21276 + }, + { + "epoch": 0.82, + "learning_rate": 3.638945883559887e-05, + "loss": 0.4231, + "step": 21277 + }, + { + "epoch": 0.82, + "learning_rate": 3.638881764555014e-05, + "loss": 0.5333, + "step": 21278 + }, + { + "epoch": 0.82, + "learning_rate": 3.638817645550141e-05, + "loss": 0.3957, + "step": 21279 + }, + { + "epoch": 0.82, + "learning_rate": 3.6387535265452685e-05, + "loss": 0.374, + "step": 21280 + }, + { + "epoch": 0.82, + "learning_rate": 3.638689407540395e-05, + "loss": 0.577, + "step": 21281 + }, + { + "epoch": 0.82, + "learning_rate": 3.638625288535522e-05, + "loss": 0.5367, + "step": 21282 + }, + { + "epoch": 0.82, + "learning_rate": 3.638561169530649e-05, + "loss": 0.4627, + "step": 21283 + }, + { + "epoch": 0.82, + "learning_rate": 3.638497050525776e-05, + "loss": 0.5354, + "step": 21284 + }, + { + "epoch": 0.82, + "learning_rate": 3.638432931520903e-05, + "loss": 0.5803, + "step": 21285 + }, + { + "epoch": 0.82, + "learning_rate": 3.6383688125160295e-05, + "loss": 0.681, + "step": 21286 + }, + { + "epoch": 0.82, + "learning_rate": 3.638304693511157e-05, + "loss": 0.5358, + "step": 21287 + }, + { + "epoch": 0.82, + "learning_rate": 3.638240574506284e-05, + "loss": 0.4977, + "step": 21288 + }, + { + "epoch": 0.82, + "learning_rate": 3.638176455501411e-05, + "loss": 0.5679, + "step": 21289 + }, + { + "epoch": 0.82, + "learning_rate": 3.6381123364965375e-05, + "loss": 0.5845, + "step": 21290 + }, + { + "epoch": 0.82, + "learning_rate": 3.6380482174916646e-05, + "loss": 0.6409, + "step": 21291 + }, + { + "epoch": 0.82, + "learning_rate": 3.637984098486792e-05, + "loss": 0.6395, + "step": 21292 + }, + { + "epoch": 0.82, + "learning_rate": 3.637919979481918e-05, + "loss": 0.5203, + "step": 21293 + }, + { + "epoch": 0.82, + "learning_rate": 3.6378558604770455e-05, + "loss": 0.5118, + "step": 21294 + }, + { + "epoch": 0.82, + "learning_rate": 3.6377917414721726e-05, + "loss": 0.5482, + "step": 21295 + }, + { + "epoch": 0.82, + "learning_rate": 3.6377276224673e-05, + "loss": 0.4821, + "step": 21296 + }, + { + "epoch": 0.82, + "learning_rate": 3.637663503462426e-05, + "loss": 0.6018, + "step": 21297 + }, + { + "epoch": 0.82, + "learning_rate": 3.6375993844575535e-05, + "loss": 0.4268, + "step": 21298 + }, + { + "epoch": 0.82, + "learning_rate": 3.63753526545268e-05, + "loss": 0.6697, + "step": 21299 + }, + { + "epoch": 0.82, + "learning_rate": 3.637471146447807e-05, + "loss": 0.5105, + "step": 21300 + }, + { + "epoch": 0.82, + "learning_rate": 3.637407027442934e-05, + "loss": 0.463, + "step": 21301 + }, + { + "epoch": 0.82, + "learning_rate": 3.637342908438061e-05, + "loss": 0.5288, + "step": 21302 + }, + { + "epoch": 0.82, + "learning_rate": 3.6372787894331887e-05, + "loss": 0.5785, + "step": 21303 + }, + { + "epoch": 0.82, + "learning_rate": 3.637214670428315e-05, + "loss": 0.447, + "step": 21304 + }, + { + "epoch": 0.82, + "learning_rate": 3.637150551423442e-05, + "loss": 0.4968, + "step": 21305 + }, + { + "epoch": 0.82, + "learning_rate": 3.637086432418569e-05, + "loss": 0.4217, + "step": 21306 + }, + { + "epoch": 0.82, + "learning_rate": 3.637022313413696e-05, + "loss": 0.5919, + "step": 21307 + }, + { + "epoch": 0.82, + "learning_rate": 3.6369581944088225e-05, + "loss": 0.4667, + "step": 21308 + }, + { + "epoch": 0.82, + "learning_rate": 3.6368940754039497e-05, + "loss": 0.4071, + "step": 21309 + }, + { + "epoch": 0.82, + "learning_rate": 3.636829956399077e-05, + "loss": 0.4471, + "step": 21310 + }, + { + "epoch": 0.82, + "learning_rate": 3.636765837394204e-05, + "loss": 0.5029, + "step": 21311 + }, + { + "epoch": 0.82, + "learning_rate": 3.636701718389331e-05, + "loss": 0.4676, + "step": 21312 + }, + { + "epoch": 0.82, + "learning_rate": 3.636637599384458e-05, + "loss": 0.6039, + "step": 21313 + }, + { + "epoch": 0.82, + "learning_rate": 3.636573480379585e-05, + "loss": 0.598, + "step": 21314 + }, + { + "epoch": 0.82, + "learning_rate": 3.636509361374711e-05, + "loss": 0.4986, + "step": 21315 + }, + { + "epoch": 0.82, + "learning_rate": 3.6364452423698385e-05, + "loss": 0.5845, + "step": 21316 + }, + { + "epoch": 0.82, + "learning_rate": 3.636381123364965e-05, + "loss": 0.5045, + "step": 21317 + }, + { + "epoch": 0.82, + "learning_rate": 3.636317004360092e-05, + "loss": 0.5342, + "step": 21318 + }, + { + "epoch": 0.82, + "learning_rate": 3.63625288535522e-05, + "loss": 0.5635, + "step": 21319 + }, + { + "epoch": 0.82, + "learning_rate": 3.6361887663503465e-05, + "loss": 0.4905, + "step": 21320 + }, + { + "epoch": 0.82, + "learning_rate": 3.636124647345474e-05, + "loss": 0.5265, + "step": 21321 + }, + { + "epoch": 0.82, + "learning_rate": 3.6360605283406e-05, + "loss": 0.5362, + "step": 21322 + }, + { + "epoch": 0.82, + "learning_rate": 3.6359964093357274e-05, + "loss": 0.6086, + "step": 21323 + }, + { + "epoch": 0.82, + "learning_rate": 3.635932290330854e-05, + "loss": 0.495, + "step": 21324 + }, + { + "epoch": 0.82, + "learning_rate": 3.635868171325981e-05, + "loss": 0.399, + "step": 21325 + }, + { + "epoch": 0.82, + "learning_rate": 3.635804052321108e-05, + "loss": 0.5252, + "step": 21326 + }, + { + "epoch": 0.82, + "learning_rate": 3.6357399333162354e-05, + "loss": 0.5451, + "step": 21327 + }, + { + "epoch": 0.82, + "learning_rate": 3.6356758143113625e-05, + "loss": 0.5975, + "step": 21328 + }, + { + "epoch": 0.82, + "learning_rate": 3.635611695306489e-05, + "loss": 0.5656, + "step": 21329 + }, + { + "epoch": 0.82, + "learning_rate": 3.635547576301616e-05, + "loss": 0.4758, + "step": 21330 + }, + { + "epoch": 0.82, + "learning_rate": 3.635483457296743e-05, + "loss": 0.4176, + "step": 21331 + }, + { + "epoch": 0.82, + "learning_rate": 3.63541933829187e-05, + "loss": 0.4831, + "step": 21332 + }, + { + "epoch": 0.82, + "learning_rate": 3.6353552192869964e-05, + "loss": 0.4673, + "step": 21333 + }, + { + "epoch": 0.82, + "learning_rate": 3.6352911002821235e-05, + "loss": 0.5907, + "step": 21334 + }, + { + "epoch": 0.82, + "learning_rate": 3.635226981277251e-05, + "loss": 0.4632, + "step": 21335 + }, + { + "epoch": 0.82, + "learning_rate": 3.635162862272378e-05, + "loss": 0.4933, + "step": 21336 + }, + { + "epoch": 0.82, + "learning_rate": 3.635098743267505e-05, + "loss": 0.5551, + "step": 21337 + }, + { + "epoch": 0.82, + "learning_rate": 3.6350346242626315e-05, + "loss": 0.4958, + "step": 21338 + }, + { + "epoch": 0.82, + "learning_rate": 3.634970505257759e-05, + "loss": 0.4641, + "step": 21339 + }, + { + "epoch": 0.82, + "learning_rate": 3.634906386252885e-05, + "loss": 0.4145, + "step": 21340 + }, + { + "epoch": 0.82, + "learning_rate": 3.6348422672480124e-05, + "loss": 0.4871, + "step": 21341 + }, + { + "epoch": 0.82, + "learning_rate": 3.6347781482431395e-05, + "loss": 0.5924, + "step": 21342 + }, + { + "epoch": 0.82, + "learning_rate": 3.634714029238267e-05, + "loss": 0.4414, + "step": 21343 + }, + { + "epoch": 0.82, + "learning_rate": 3.634649910233394e-05, + "loss": 0.3702, + "step": 21344 + }, + { + "epoch": 0.82, + "learning_rate": 3.6345857912285204e-05, + "loss": 0.4449, + "step": 21345 + }, + { + "epoch": 0.82, + "learning_rate": 3.6345216722236476e-05, + "loss": 0.4281, + "step": 21346 + }, + { + "epoch": 0.82, + "learning_rate": 3.634457553218774e-05, + "loss": 0.5529, + "step": 21347 + }, + { + "epoch": 0.82, + "learning_rate": 3.634393434213901e-05, + "loss": 0.5074, + "step": 21348 + }, + { + "epoch": 0.82, + "learning_rate": 3.634329315209028e-05, + "loss": 0.5577, + "step": 21349 + }, + { + "epoch": 0.82, + "learning_rate": 3.634265196204155e-05, + "loss": 0.5139, + "step": 21350 + }, + { + "epoch": 0.82, + "learning_rate": 3.634201077199282e-05, + "loss": 0.5787, + "step": 21351 + }, + { + "epoch": 0.82, + "learning_rate": 3.634136958194409e-05, + "loss": 0.6032, + "step": 21352 + }, + { + "epoch": 0.82, + "learning_rate": 3.6340728391895364e-05, + "loss": 0.5806, + "step": 21353 + }, + { + "epoch": 0.82, + "learning_rate": 3.634008720184663e-05, + "loss": 0.513, + "step": 21354 + }, + { + "epoch": 0.82, + "learning_rate": 3.63394460117979e-05, + "loss": 0.4677, + "step": 21355 + }, + { + "epoch": 0.82, + "learning_rate": 3.6338804821749166e-05, + "loss": 0.6009, + "step": 21356 + }, + { + "epoch": 0.82, + "learning_rate": 3.633816363170044e-05, + "loss": 0.4944, + "step": 21357 + }, + { + "epoch": 0.82, + "learning_rate": 3.633752244165171e-05, + "loss": 0.4789, + "step": 21358 + }, + { + "epoch": 0.82, + "learning_rate": 3.633688125160298e-05, + "loss": 0.4488, + "step": 21359 + }, + { + "epoch": 0.82, + "learning_rate": 3.6336240061554246e-05, + "loss": 0.508, + "step": 21360 + }, + { + "epoch": 0.82, + "learning_rate": 3.633559887150552e-05, + "loss": 0.4924, + "step": 21361 + }, + { + "epoch": 0.82, + "learning_rate": 3.633495768145679e-05, + "loss": 0.5083, + "step": 21362 + }, + { + "epoch": 0.82, + "learning_rate": 3.6334316491408054e-05, + "loss": 0.4475, + "step": 21363 + }, + { + "epoch": 0.82, + "learning_rate": 3.6333675301359326e-05, + "loss": 0.5102, + "step": 21364 + }, + { + "epoch": 0.82, + "learning_rate": 3.633303411131059e-05, + "loss": 0.4011, + "step": 21365 + }, + { + "epoch": 0.82, + "learning_rate": 3.633239292126186e-05, + "loss": 0.4537, + "step": 21366 + }, + { + "epoch": 0.82, + "learning_rate": 3.6331751731213134e-05, + "loss": 0.4647, + "step": 21367 + }, + { + "epoch": 0.82, + "learning_rate": 3.6331110541164406e-05, + "loss": 0.5478, + "step": 21368 + }, + { + "epoch": 0.82, + "learning_rate": 3.633046935111567e-05, + "loss": 0.4281, + "step": 21369 + }, + { + "epoch": 0.82, + "learning_rate": 3.632982816106694e-05, + "loss": 0.4661, + "step": 21370 + }, + { + "epoch": 0.82, + "learning_rate": 3.6329186971018214e-05, + "loss": 0.4817, + "step": 21371 + }, + { + "epoch": 0.82, + "learning_rate": 3.632854578096948e-05, + "loss": 0.433, + "step": 21372 + }, + { + "epoch": 0.82, + "learning_rate": 3.632790459092075e-05, + "loss": 0.5247, + "step": 21373 + }, + { + "epoch": 0.82, + "learning_rate": 3.6327263400872016e-05, + "loss": 0.4914, + "step": 21374 + }, + { + "epoch": 0.82, + "learning_rate": 3.6326622210823294e-05, + "loss": 0.4769, + "step": 21375 + }, + { + "epoch": 0.82, + "learning_rate": 3.632598102077456e-05, + "loss": 0.457, + "step": 21376 + }, + { + "epoch": 0.82, + "learning_rate": 3.632533983072583e-05, + "loss": 0.4021, + "step": 21377 + }, + { + "epoch": 0.82, + "learning_rate": 3.6324698640677096e-05, + "loss": 0.443, + "step": 21378 + }, + { + "epoch": 0.82, + "learning_rate": 3.632405745062837e-05, + "loss": 0.4965, + "step": 21379 + }, + { + "epoch": 0.82, + "learning_rate": 3.632341626057964e-05, + "loss": 0.5367, + "step": 21380 + }, + { + "epoch": 0.82, + "learning_rate": 3.6322775070530904e-05, + "loss": 0.4386, + "step": 21381 + }, + { + "epoch": 0.82, + "learning_rate": 3.6322133880482176e-05, + "loss": 0.461, + "step": 21382 + }, + { + "epoch": 0.82, + "learning_rate": 3.632149269043345e-05, + "loss": 0.4876, + "step": 21383 + }, + { + "epoch": 0.82, + "learning_rate": 3.632085150038472e-05, + "loss": 0.3662, + "step": 21384 + }, + { + "epoch": 0.82, + "learning_rate": 3.6320210310335984e-05, + "loss": 0.5042, + "step": 21385 + }, + { + "epoch": 0.82, + "learning_rate": 3.6319569120287256e-05, + "loss": 0.4486, + "step": 21386 + }, + { + "epoch": 0.82, + "learning_rate": 3.631892793023852e-05, + "loss": 0.5418, + "step": 21387 + }, + { + "epoch": 0.82, + "learning_rate": 3.631828674018979e-05, + "loss": 0.5932, + "step": 21388 + }, + { + "epoch": 0.82, + "learning_rate": 3.6317645550141065e-05, + "loss": 0.5545, + "step": 21389 + }, + { + "epoch": 0.82, + "learning_rate": 3.631700436009233e-05, + "loss": 0.4457, + "step": 21390 + }, + { + "epoch": 0.82, + "learning_rate": 3.631636317004361e-05, + "loss": 0.7139, + "step": 21391 + }, + { + "epoch": 0.82, + "learning_rate": 3.631572197999487e-05, + "loss": 0.4742, + "step": 21392 + }, + { + "epoch": 0.82, + "learning_rate": 3.6315080789946145e-05, + "loss": 0.5928, + "step": 21393 + }, + { + "epoch": 0.82, + "learning_rate": 3.631443959989741e-05, + "loss": 0.5793, + "step": 21394 + }, + { + "epoch": 0.82, + "learning_rate": 3.631379840984868e-05, + "loss": 0.6123, + "step": 21395 + }, + { + "epoch": 0.82, + "learning_rate": 3.6313157219799946e-05, + "loss": 0.5104, + "step": 21396 + }, + { + "epoch": 0.82, + "learning_rate": 3.631251602975122e-05, + "loss": 0.5381, + "step": 21397 + }, + { + "epoch": 0.82, + "learning_rate": 3.631187483970249e-05, + "loss": 0.5061, + "step": 21398 + }, + { + "epoch": 0.82, + "learning_rate": 3.631123364965376e-05, + "loss": 0.589, + "step": 21399 + }, + { + "epoch": 0.82, + "learning_rate": 3.631059245960503e-05, + "loss": 0.5237, + "step": 21400 + }, + { + "epoch": 0.82, + "learning_rate": 3.63099512695563e-05, + "loss": 0.5872, + "step": 21401 + }, + { + "epoch": 0.82, + "learning_rate": 3.630931007950757e-05, + "loss": 0.5408, + "step": 21402 + }, + { + "epoch": 0.82, + "learning_rate": 3.6308668889458835e-05, + "loss": 0.5092, + "step": 21403 + }, + { + "epoch": 0.82, + "learning_rate": 3.6308027699410106e-05, + "loss": 0.491, + "step": 21404 + }, + { + "epoch": 0.82, + "learning_rate": 3.630738650936137e-05, + "loss": 0.5849, + "step": 21405 + }, + { + "epoch": 0.82, + "learning_rate": 3.630674531931264e-05, + "loss": 0.5147, + "step": 21406 + }, + { + "epoch": 0.82, + "learning_rate": 3.6306104129263915e-05, + "loss": 0.471, + "step": 21407 + }, + { + "epoch": 0.82, + "learning_rate": 3.6305462939215186e-05, + "loss": 0.514, + "step": 21408 + }, + { + "epoch": 0.82, + "learning_rate": 3.630482174916646e-05, + "loss": 0.4831, + "step": 21409 + }, + { + "epoch": 0.82, + "learning_rate": 3.630418055911772e-05, + "loss": 0.5592, + "step": 21410 + }, + { + "epoch": 0.82, + "learning_rate": 3.6303539369068995e-05, + "loss": 0.4583, + "step": 21411 + }, + { + "epoch": 0.82, + "learning_rate": 3.630289817902026e-05, + "loss": 0.5195, + "step": 21412 + }, + { + "epoch": 0.82, + "learning_rate": 3.630225698897153e-05, + "loss": 0.5227, + "step": 21413 + }, + { + "epoch": 0.82, + "learning_rate": 3.63016157989228e-05, + "loss": 0.5123, + "step": 21414 + }, + { + "epoch": 0.82, + "learning_rate": 3.6300974608874075e-05, + "loss": 0.5467, + "step": 21415 + }, + { + "epoch": 0.82, + "learning_rate": 3.630033341882535e-05, + "loss": 0.4083, + "step": 21416 + }, + { + "epoch": 0.82, + "learning_rate": 3.629969222877661e-05, + "loss": 0.4378, + "step": 21417 + }, + { + "epoch": 0.82, + "learning_rate": 3.629905103872788e-05, + "loss": 0.4546, + "step": 21418 + }, + { + "epoch": 0.82, + "learning_rate": 3.629840984867915e-05, + "loss": 0.5236, + "step": 21419 + }, + { + "epoch": 0.82, + "learning_rate": 3.629776865863042e-05, + "loss": 0.3941, + "step": 21420 + }, + { + "epoch": 0.82, + "learning_rate": 3.6297127468581685e-05, + "loss": 0.4329, + "step": 21421 + }, + { + "epoch": 0.82, + "learning_rate": 3.6296486278532957e-05, + "loss": 0.5659, + "step": 21422 + }, + { + "epoch": 0.82, + "learning_rate": 3.629584508848423e-05, + "loss": 0.5395, + "step": 21423 + }, + { + "epoch": 0.82, + "learning_rate": 3.62952038984355e-05, + "loss": 0.47, + "step": 21424 + }, + { + "epoch": 0.82, + "learning_rate": 3.629456270838677e-05, + "loss": 0.5414, + "step": 21425 + }, + { + "epoch": 0.82, + "learning_rate": 3.629392151833804e-05, + "loss": 0.5949, + "step": 21426 + }, + { + "epoch": 0.82, + "learning_rate": 3.629328032828931e-05, + "loss": 0.5205, + "step": 21427 + }, + { + "epoch": 0.82, + "learning_rate": 3.629263913824057e-05, + "loss": 0.4355, + "step": 21428 + }, + { + "epoch": 0.82, + "learning_rate": 3.6291997948191845e-05, + "loss": 0.5024, + "step": 21429 + }, + { + "epoch": 0.82, + "learning_rate": 3.629135675814312e-05, + "loss": 0.5067, + "step": 21430 + }, + { + "epoch": 0.82, + "learning_rate": 3.629071556809439e-05, + "loss": 0.5353, + "step": 21431 + }, + { + "epoch": 0.82, + "learning_rate": 3.6290074378045653e-05, + "loss": 0.5324, + "step": 21432 + }, + { + "epoch": 0.82, + "learning_rate": 3.6289433187996925e-05, + "loss": 0.515, + "step": 21433 + }, + { + "epoch": 0.82, + "learning_rate": 3.62887919979482e-05, + "loss": 0.526, + "step": 21434 + }, + { + "epoch": 0.82, + "learning_rate": 3.628815080789946e-05, + "loss": 0.4719, + "step": 21435 + }, + { + "epoch": 0.82, + "learning_rate": 3.6287509617850734e-05, + "loss": 0.4199, + "step": 21436 + }, + { + "epoch": 0.82, + "learning_rate": 3.6286868427802e-05, + "loss": 0.5191, + "step": 21437 + }, + { + "epoch": 0.82, + "learning_rate": 3.628622723775327e-05, + "loss": 0.4197, + "step": 21438 + }, + { + "epoch": 0.82, + "learning_rate": 3.628558604770454e-05, + "loss": 0.5624, + "step": 21439 + }, + { + "epoch": 0.82, + "learning_rate": 3.6284944857655814e-05, + "loss": 0.4061, + "step": 21440 + }, + { + "epoch": 0.82, + "learning_rate": 3.628430366760708e-05, + "loss": 0.5054, + "step": 21441 + }, + { + "epoch": 0.82, + "learning_rate": 3.628366247755835e-05, + "loss": 0.5757, + "step": 21442 + }, + { + "epoch": 0.82, + "learning_rate": 3.628302128750962e-05, + "loss": 0.5925, + "step": 21443 + }, + { + "epoch": 0.82, + "learning_rate": 3.628238009746089e-05, + "loss": 0.4584, + "step": 21444 + }, + { + "epoch": 0.82, + "learning_rate": 3.628173890741216e-05, + "loss": 0.6803, + "step": 21445 + }, + { + "epoch": 0.82, + "learning_rate": 3.628109771736343e-05, + "loss": 0.4636, + "step": 21446 + }, + { + "epoch": 0.82, + "learning_rate": 3.62804565273147e-05, + "loss": 0.562, + "step": 21447 + }, + { + "epoch": 0.82, + "learning_rate": 3.627981533726597e-05, + "loss": 0.59, + "step": 21448 + }, + { + "epoch": 0.82, + "learning_rate": 3.627917414721724e-05, + "loss": 0.5914, + "step": 21449 + }, + { + "epoch": 0.82, + "learning_rate": 3.6278532957168504e-05, + "loss": 0.4236, + "step": 21450 + }, + { + "epoch": 0.82, + "learning_rate": 3.6277891767119775e-05, + "loss": 0.489, + "step": 21451 + }, + { + "epoch": 0.82, + "learning_rate": 3.627725057707105e-05, + "loss": 0.4849, + "step": 21452 + }, + { + "epoch": 0.82, + "learning_rate": 3.627660938702231e-05, + "loss": 0.5433, + "step": 21453 + }, + { + "epoch": 0.82, + "learning_rate": 3.6275968196973584e-05, + "loss": 0.5628, + "step": 21454 + }, + { + "epoch": 0.82, + "learning_rate": 3.6275327006924855e-05, + "loss": 0.5396, + "step": 21455 + }, + { + "epoch": 0.82, + "learning_rate": 3.627468581687613e-05, + "loss": 0.4594, + "step": 21456 + }, + { + "epoch": 0.82, + "learning_rate": 3.627404462682739e-05, + "loss": 0.5928, + "step": 21457 + }, + { + "epoch": 0.82, + "learning_rate": 3.6273403436778664e-05, + "loss": 0.4764, + "step": 21458 + }, + { + "epoch": 0.83, + "learning_rate": 3.627276224672993e-05, + "loss": 0.5267, + "step": 21459 + }, + { + "epoch": 0.83, + "learning_rate": 3.62721210566812e-05, + "loss": 0.6371, + "step": 21460 + }, + { + "epoch": 0.83, + "learning_rate": 3.627147986663247e-05, + "loss": 0.5489, + "step": 21461 + }, + { + "epoch": 0.83, + "learning_rate": 3.627083867658374e-05, + "loss": 0.4193, + "step": 21462 + }, + { + "epoch": 0.83, + "learning_rate": 3.6270197486535016e-05, + "loss": 0.5362, + "step": 21463 + }, + { + "epoch": 0.83, + "learning_rate": 3.626955629648628e-05, + "loss": 0.4529, + "step": 21464 + }, + { + "epoch": 0.83, + "learning_rate": 3.626891510643755e-05, + "loss": 0.4425, + "step": 21465 + }, + { + "epoch": 0.83, + "learning_rate": 3.626827391638882e-05, + "loss": 0.474, + "step": 21466 + }, + { + "epoch": 0.83, + "learning_rate": 3.626763272634009e-05, + "loss": 0.4375, + "step": 21467 + }, + { + "epoch": 0.83, + "learning_rate": 3.6266991536291354e-05, + "loss": 0.5047, + "step": 21468 + }, + { + "epoch": 0.83, + "learning_rate": 3.6266350346242626e-05, + "loss": 0.5742, + "step": 21469 + }, + { + "epoch": 0.83, + "learning_rate": 3.62657091561939e-05, + "loss": 0.5168, + "step": 21470 + }, + { + "epoch": 0.83, + "learning_rate": 3.626506796614517e-05, + "loss": 0.5781, + "step": 21471 + }, + { + "epoch": 0.83, + "learning_rate": 3.626442677609644e-05, + "loss": 0.4619, + "step": 21472 + }, + { + "epoch": 0.83, + "learning_rate": 3.6263785586047706e-05, + "loss": 0.4883, + "step": 21473 + }, + { + "epoch": 0.83, + "learning_rate": 3.626314439599898e-05, + "loss": 0.5607, + "step": 21474 + }, + { + "epoch": 0.83, + "learning_rate": 3.626250320595024e-05, + "loss": 0.5228, + "step": 21475 + }, + { + "epoch": 0.83, + "learning_rate": 3.6261862015901514e-05, + "loss": 0.5868, + "step": 21476 + }, + { + "epoch": 0.83, + "learning_rate": 3.626122082585278e-05, + "loss": 0.5002, + "step": 21477 + }, + { + "epoch": 0.83, + "learning_rate": 3.626057963580405e-05, + "loss": 0.4843, + "step": 21478 + }, + { + "epoch": 0.83, + "learning_rate": 3.625993844575533e-05, + "loss": 0.4826, + "step": 21479 + }, + { + "epoch": 0.83, + "learning_rate": 3.6259297255706594e-05, + "loss": 0.4834, + "step": 21480 + }, + { + "epoch": 0.83, + "learning_rate": 3.6258656065657866e-05, + "loss": 0.3979, + "step": 21481 + }, + { + "epoch": 0.83, + "learning_rate": 3.625801487560913e-05, + "loss": 0.4056, + "step": 21482 + }, + { + "epoch": 0.83, + "learning_rate": 3.62573736855604e-05, + "loss": 0.5353, + "step": 21483 + }, + { + "epoch": 0.83, + "learning_rate": 3.625673249551167e-05, + "loss": 0.5623, + "step": 21484 + }, + { + "epoch": 0.83, + "learning_rate": 3.625609130546294e-05, + "loss": 0.4835, + "step": 21485 + }, + { + "epoch": 0.83, + "learning_rate": 3.625545011541421e-05, + "loss": 0.4896, + "step": 21486 + }, + { + "epoch": 0.83, + "learning_rate": 3.625480892536548e-05, + "loss": 0.6278, + "step": 21487 + }, + { + "epoch": 0.83, + "learning_rate": 3.6254167735316754e-05, + "loss": 0.5536, + "step": 21488 + }, + { + "epoch": 0.83, + "learning_rate": 3.625352654526802e-05, + "loss": 0.4682, + "step": 21489 + }, + { + "epoch": 0.83, + "learning_rate": 3.625288535521929e-05, + "loss": 0.4862, + "step": 21490 + }, + { + "epoch": 0.83, + "learning_rate": 3.6252244165170556e-05, + "loss": 0.5422, + "step": 21491 + }, + { + "epoch": 0.83, + "learning_rate": 3.625160297512183e-05, + "loss": 0.3402, + "step": 21492 + }, + { + "epoch": 0.83, + "learning_rate": 3.625096178507309e-05, + "loss": 0.4271, + "step": 21493 + }, + { + "epoch": 0.83, + "learning_rate": 3.6250320595024364e-05, + "loss": 0.6256, + "step": 21494 + }, + { + "epoch": 0.83, + "learning_rate": 3.6249679404975636e-05, + "loss": 0.4155, + "step": 21495 + }, + { + "epoch": 0.83, + "learning_rate": 3.624903821492691e-05, + "loss": 0.4427, + "step": 21496 + }, + { + "epoch": 0.83, + "learning_rate": 3.624839702487818e-05, + "loss": 0.2692, + "step": 21497 + }, + { + "epoch": 0.83, + "learning_rate": 3.6247755834829444e-05, + "loss": 0.5003, + "step": 21498 + }, + { + "epoch": 0.83, + "learning_rate": 3.6247114644780716e-05, + "loss": 0.5066, + "step": 21499 + }, + { + "epoch": 0.83, + "learning_rate": 3.624647345473198e-05, + "loss": 0.5729, + "step": 21500 + }, + { + "epoch": 0.83, + "learning_rate": 3.624583226468325e-05, + "loss": 0.5088, + "step": 21501 + }, + { + "epoch": 0.83, + "learning_rate": 3.6245191074634525e-05, + "loss": 0.5086, + "step": 21502 + }, + { + "epoch": 0.83, + "learning_rate": 3.6244549884585796e-05, + "loss": 0.6588, + "step": 21503 + }, + { + "epoch": 0.83, + "learning_rate": 3.624390869453706e-05, + "loss": 0.3945, + "step": 21504 + }, + { + "epoch": 0.83, + "learning_rate": 3.624326750448833e-05, + "loss": 0.3742, + "step": 21505 + }, + { + "epoch": 0.83, + "learning_rate": 3.6242626314439605e-05, + "loss": 0.4754, + "step": 21506 + }, + { + "epoch": 0.83, + "learning_rate": 3.624198512439087e-05, + "loss": 0.5571, + "step": 21507 + }, + { + "epoch": 0.83, + "learning_rate": 3.624134393434214e-05, + "loss": 0.5991, + "step": 21508 + }, + { + "epoch": 0.83, + "learning_rate": 3.6240702744293406e-05, + "loss": 0.5281, + "step": 21509 + }, + { + "epoch": 0.83, + "learning_rate": 3.624006155424468e-05, + "loss": 0.521, + "step": 21510 + }, + { + "epoch": 0.83, + "learning_rate": 3.623942036419595e-05, + "loss": 0.3713, + "step": 21511 + }, + { + "epoch": 0.83, + "learning_rate": 3.623877917414722e-05, + "loss": 0.4834, + "step": 21512 + }, + { + "epoch": 0.83, + "learning_rate": 3.6238137984098486e-05, + "loss": 0.3851, + "step": 21513 + }, + { + "epoch": 0.83, + "learning_rate": 3.623749679404976e-05, + "loss": 0.4486, + "step": 21514 + }, + { + "epoch": 0.83, + "learning_rate": 3.623685560400103e-05, + "loss": 0.5805, + "step": 21515 + }, + { + "epoch": 0.83, + "learning_rate": 3.6236214413952295e-05, + "loss": 0.536, + "step": 21516 + }, + { + "epoch": 0.83, + "learning_rate": 3.6235573223903566e-05, + "loss": 0.4691, + "step": 21517 + }, + { + "epoch": 0.83, + "learning_rate": 3.623493203385484e-05, + "loss": 0.4455, + "step": 21518 + }, + { + "epoch": 0.83, + "learning_rate": 3.623429084380611e-05, + "loss": 0.5156, + "step": 21519 + }, + { + "epoch": 0.83, + "learning_rate": 3.6233649653757375e-05, + "loss": 0.4573, + "step": 21520 + }, + { + "epoch": 0.83, + "learning_rate": 3.6233008463708646e-05, + "loss": 0.5058, + "step": 21521 + }, + { + "epoch": 0.83, + "learning_rate": 3.623236727365991e-05, + "loss": 0.483, + "step": 21522 + }, + { + "epoch": 0.83, + "learning_rate": 3.623172608361118e-05, + "loss": 0.5835, + "step": 21523 + }, + { + "epoch": 0.83, + "learning_rate": 3.6231084893562455e-05, + "loss": 0.6338, + "step": 21524 + }, + { + "epoch": 0.83, + "learning_rate": 3.623044370351372e-05, + "loss": 0.4113, + "step": 21525 + }, + { + "epoch": 0.83, + "learning_rate": 3.622980251346499e-05, + "loss": 0.5649, + "step": 21526 + }, + { + "epoch": 0.83, + "learning_rate": 3.622916132341626e-05, + "loss": 0.4438, + "step": 21527 + }, + { + "epoch": 0.83, + "learning_rate": 3.6228520133367535e-05, + "loss": 0.3891, + "step": 21528 + }, + { + "epoch": 0.83, + "learning_rate": 3.62278789433188e-05, + "loss": 0.5431, + "step": 21529 + }, + { + "epoch": 0.83, + "learning_rate": 3.622723775327007e-05, + "loss": 0.4514, + "step": 21530 + }, + { + "epoch": 0.83, + "learning_rate": 3.6226596563221337e-05, + "loss": 0.5032, + "step": 21531 + }, + { + "epoch": 0.83, + "learning_rate": 3.622595537317261e-05, + "loss": 0.4404, + "step": 21532 + }, + { + "epoch": 0.83, + "learning_rate": 3.622531418312388e-05, + "loss": 0.502, + "step": 21533 + }, + { + "epoch": 0.83, + "learning_rate": 3.6224672993075145e-05, + "loss": 0.4513, + "step": 21534 + }, + { + "epoch": 0.83, + "learning_rate": 3.6224031803026423e-05, + "loss": 0.5104, + "step": 21535 + }, + { + "epoch": 0.83, + "learning_rate": 3.622339061297769e-05, + "loss": 0.5524, + "step": 21536 + }, + { + "epoch": 0.83, + "learning_rate": 3.622274942292896e-05, + "loss": 0.4936, + "step": 21537 + }, + { + "epoch": 0.83, + "learning_rate": 3.6222108232880225e-05, + "loss": 0.4996, + "step": 21538 + }, + { + "epoch": 0.83, + "learning_rate": 3.62214670428315e-05, + "loss": 0.4919, + "step": 21539 + }, + { + "epoch": 0.83, + "learning_rate": 3.622082585278277e-05, + "loss": 0.4158, + "step": 21540 + }, + { + "epoch": 0.83, + "learning_rate": 3.6220184662734033e-05, + "loss": 0.5038, + "step": 21541 + }, + { + "epoch": 0.83, + "learning_rate": 3.6219543472685305e-05, + "loss": 0.6079, + "step": 21542 + }, + { + "epoch": 0.83, + "learning_rate": 3.621890228263658e-05, + "loss": 0.5415, + "step": 21543 + }, + { + "epoch": 0.83, + "learning_rate": 3.621826109258785e-05, + "loss": 0.5475, + "step": 21544 + }, + { + "epoch": 0.83, + "learning_rate": 3.6217619902539113e-05, + "loss": 0.4772, + "step": 21545 + }, + { + "epoch": 0.83, + "learning_rate": 3.6216978712490385e-05, + "loss": 0.5336, + "step": 21546 + }, + { + "epoch": 0.83, + "learning_rate": 3.621633752244165e-05, + "loss": 0.5808, + "step": 21547 + }, + { + "epoch": 0.83, + "learning_rate": 3.621569633239292e-05, + "loss": 0.431, + "step": 21548 + }, + { + "epoch": 0.83, + "learning_rate": 3.6215055142344194e-05, + "loss": 0.4518, + "step": 21549 + }, + { + "epoch": 0.83, + "learning_rate": 3.621441395229546e-05, + "loss": 0.5571, + "step": 21550 + }, + { + "epoch": 0.83, + "learning_rate": 3.621377276224674e-05, + "loss": 0.635, + "step": 21551 + }, + { + "epoch": 0.83, + "learning_rate": 3.6213131572198e-05, + "loss": 0.4979, + "step": 21552 + }, + { + "epoch": 0.83, + "learning_rate": 3.6212490382149274e-05, + "loss": 0.4467, + "step": 21553 + }, + { + "epoch": 0.83, + "learning_rate": 3.621184919210054e-05, + "loss": 0.314, + "step": 21554 + }, + { + "epoch": 0.83, + "learning_rate": 3.621120800205181e-05, + "loss": 0.5231, + "step": 21555 + }, + { + "epoch": 0.83, + "learning_rate": 3.6210566812003075e-05, + "loss": 0.4193, + "step": 21556 + }, + { + "epoch": 0.83, + "learning_rate": 3.620992562195435e-05, + "loss": 0.5216, + "step": 21557 + }, + { + "epoch": 0.83, + "learning_rate": 3.620928443190562e-05, + "loss": 0.4106, + "step": 21558 + }, + { + "epoch": 0.83, + "learning_rate": 3.620864324185689e-05, + "loss": 0.5828, + "step": 21559 + }, + { + "epoch": 0.83, + "learning_rate": 3.620800205180816e-05, + "loss": 0.4304, + "step": 21560 + }, + { + "epoch": 0.83, + "learning_rate": 3.620736086175943e-05, + "loss": 0.4651, + "step": 21561 + }, + { + "epoch": 0.83, + "learning_rate": 3.62067196717107e-05, + "loss": 0.5359, + "step": 21562 + }, + { + "epoch": 0.83, + "learning_rate": 3.6206078481661964e-05, + "loss": 0.4889, + "step": 21563 + }, + { + "epoch": 0.83, + "learning_rate": 3.6205437291613235e-05, + "loss": 0.4749, + "step": 21564 + }, + { + "epoch": 0.83, + "learning_rate": 3.62047961015645e-05, + "loss": 0.4186, + "step": 21565 + }, + { + "epoch": 0.83, + "learning_rate": 3.620415491151577e-05, + "loss": 0.5683, + "step": 21566 + }, + { + "epoch": 0.83, + "learning_rate": 3.620351372146705e-05, + "loss": 0.5469, + "step": 21567 + }, + { + "epoch": 0.83, + "learning_rate": 3.6202872531418316e-05, + "loss": 0.4376, + "step": 21568 + }, + { + "epoch": 0.83, + "learning_rate": 3.620223134136959e-05, + "loss": 0.4976, + "step": 21569 + }, + { + "epoch": 0.83, + "learning_rate": 3.620159015132085e-05, + "loss": 0.5801, + "step": 21570 + }, + { + "epoch": 0.83, + "learning_rate": 3.6200948961272124e-05, + "loss": 0.5862, + "step": 21571 + }, + { + "epoch": 0.83, + "learning_rate": 3.620030777122339e-05, + "loss": 0.4628, + "step": 21572 + }, + { + "epoch": 0.83, + "learning_rate": 3.619966658117466e-05, + "loss": 0.6101, + "step": 21573 + }, + { + "epoch": 0.83, + "learning_rate": 3.619902539112593e-05, + "loss": 0.4737, + "step": 21574 + }, + { + "epoch": 0.83, + "learning_rate": 3.6198384201077204e-05, + "loss": 0.4841, + "step": 21575 + }, + { + "epoch": 0.83, + "learning_rate": 3.6197743011028476e-05, + "loss": 0.5013, + "step": 21576 + }, + { + "epoch": 0.83, + "learning_rate": 3.619710182097974e-05, + "loss": 0.4384, + "step": 21577 + }, + { + "epoch": 0.83, + "learning_rate": 3.619646063093101e-05, + "loss": 0.6, + "step": 21578 + }, + { + "epoch": 0.83, + "learning_rate": 3.619581944088228e-05, + "loss": 0.4563, + "step": 21579 + }, + { + "epoch": 0.83, + "learning_rate": 3.619517825083355e-05, + "loss": 0.4332, + "step": 21580 + }, + { + "epoch": 0.83, + "learning_rate": 3.6194537060784814e-05, + "loss": 0.5147, + "step": 21581 + }, + { + "epoch": 0.83, + "learning_rate": 3.6193895870736086e-05, + "loss": 0.4322, + "step": 21582 + }, + { + "epoch": 0.83, + "learning_rate": 3.619325468068736e-05, + "loss": 0.4051, + "step": 21583 + }, + { + "epoch": 0.83, + "learning_rate": 3.619261349063863e-05, + "loss": 0.5585, + "step": 21584 + }, + { + "epoch": 0.83, + "learning_rate": 3.61919723005899e-05, + "loss": 0.4289, + "step": 21585 + }, + { + "epoch": 0.83, + "learning_rate": 3.6191331110541166e-05, + "loss": 0.4897, + "step": 21586 + }, + { + "epoch": 0.83, + "learning_rate": 3.619068992049244e-05, + "loss": 0.4522, + "step": 21587 + }, + { + "epoch": 0.83, + "learning_rate": 3.61900487304437e-05, + "loss": 0.57, + "step": 21588 + }, + { + "epoch": 0.83, + "learning_rate": 3.6189407540394974e-05, + "loss": 0.4246, + "step": 21589 + }, + { + "epoch": 0.83, + "learning_rate": 3.6188766350346246e-05, + "loss": 0.5123, + "step": 21590 + }, + { + "epoch": 0.83, + "learning_rate": 3.618812516029752e-05, + "loss": 0.6214, + "step": 21591 + }, + { + "epoch": 0.83, + "learning_rate": 3.618748397024878e-05, + "loss": 0.4924, + "step": 21592 + }, + { + "epoch": 0.83, + "learning_rate": 3.6186842780200054e-05, + "loss": 0.4624, + "step": 21593 + }, + { + "epoch": 0.83, + "learning_rate": 3.6186201590151326e-05, + "loss": 0.5555, + "step": 21594 + }, + { + "epoch": 0.83, + "learning_rate": 3.618556040010259e-05, + "loss": 0.3996, + "step": 21595 + }, + { + "epoch": 0.83, + "learning_rate": 3.618491921005386e-05, + "loss": 0.4087, + "step": 21596 + }, + { + "epoch": 0.83, + "learning_rate": 3.618427802000513e-05, + "loss": 0.423, + "step": 21597 + }, + { + "epoch": 0.83, + "learning_rate": 3.61836368299564e-05, + "loss": 0.5383, + "step": 21598 + }, + { + "epoch": 0.83, + "learning_rate": 3.618299563990767e-05, + "loss": 0.4198, + "step": 21599 + }, + { + "epoch": 0.83, + "learning_rate": 3.618235444985894e-05, + "loss": 0.6484, + "step": 21600 + }, + { + "epoch": 0.83, + "learning_rate": 3.618171325981021e-05, + "loss": 0.5023, + "step": 21601 + }, + { + "epoch": 0.83, + "learning_rate": 3.618107206976148e-05, + "loss": 0.5092, + "step": 21602 + }, + { + "epoch": 0.83, + "learning_rate": 3.618043087971275e-05, + "loss": 0.4581, + "step": 21603 + }, + { + "epoch": 0.83, + "learning_rate": 3.6179789689664016e-05, + "loss": 0.5341, + "step": 21604 + }, + { + "epoch": 0.83, + "learning_rate": 3.617914849961529e-05, + "loss": 0.5687, + "step": 21605 + }, + { + "epoch": 0.83, + "learning_rate": 3.617850730956656e-05, + "loss": 0.5077, + "step": 21606 + }, + { + "epoch": 0.83, + "learning_rate": 3.617786611951783e-05, + "loss": 0.5232, + "step": 21607 + }, + { + "epoch": 0.83, + "learning_rate": 3.6177224929469096e-05, + "loss": 0.4447, + "step": 21608 + }, + { + "epoch": 0.83, + "learning_rate": 3.617658373942037e-05, + "loss": 0.491, + "step": 21609 + }, + { + "epoch": 0.83, + "learning_rate": 3.617594254937163e-05, + "loss": 0.4971, + "step": 21610 + }, + { + "epoch": 0.83, + "learning_rate": 3.6175301359322904e-05, + "loss": 0.4311, + "step": 21611 + }, + { + "epoch": 0.83, + "learning_rate": 3.6174660169274176e-05, + "loss": 0.4186, + "step": 21612 + }, + { + "epoch": 0.83, + "learning_rate": 3.617401897922544e-05, + "loss": 0.649, + "step": 21613 + }, + { + "epoch": 0.83, + "learning_rate": 3.617337778917671e-05, + "loss": 0.4692, + "step": 21614 + }, + { + "epoch": 0.83, + "learning_rate": 3.6172736599127985e-05, + "loss": 0.4382, + "step": 21615 + }, + { + "epoch": 0.83, + "learning_rate": 3.6172095409079256e-05, + "loss": 0.5316, + "step": 21616 + }, + { + "epoch": 0.83, + "learning_rate": 3.617145421903052e-05, + "loss": 0.4865, + "step": 21617 + }, + { + "epoch": 0.83, + "learning_rate": 3.617081302898179e-05, + "loss": 0.4413, + "step": 21618 + }, + { + "epoch": 0.83, + "learning_rate": 3.617017183893306e-05, + "loss": 0.6015, + "step": 21619 + }, + { + "epoch": 0.83, + "learning_rate": 3.616953064888433e-05, + "loss": 0.6094, + "step": 21620 + }, + { + "epoch": 0.83, + "learning_rate": 3.61688894588356e-05, + "loss": 0.4217, + "step": 21621 + }, + { + "epoch": 0.83, + "learning_rate": 3.6168248268786866e-05, + "loss": 0.5773, + "step": 21622 + }, + { + "epoch": 0.83, + "learning_rate": 3.6167607078738145e-05, + "loss": 0.5083, + "step": 21623 + }, + { + "epoch": 0.83, + "learning_rate": 3.616696588868941e-05, + "loss": 0.4248, + "step": 21624 + }, + { + "epoch": 0.83, + "learning_rate": 3.616632469864068e-05, + "loss": 0.5074, + "step": 21625 + }, + { + "epoch": 0.83, + "learning_rate": 3.6165683508591946e-05, + "loss": 0.4529, + "step": 21626 + }, + { + "epoch": 0.83, + "learning_rate": 3.616504231854322e-05, + "loss": 0.5379, + "step": 21627 + }, + { + "epoch": 0.83, + "learning_rate": 3.616440112849448e-05, + "loss": 0.4136, + "step": 21628 + }, + { + "epoch": 0.83, + "learning_rate": 3.6163759938445755e-05, + "loss": 0.456, + "step": 21629 + }, + { + "epoch": 0.83, + "learning_rate": 3.6163118748397026e-05, + "loss": 0.4956, + "step": 21630 + }, + { + "epoch": 0.83, + "learning_rate": 3.61624775583483e-05, + "loss": 0.497, + "step": 21631 + }, + { + "epoch": 0.83, + "learning_rate": 3.616183636829957e-05, + "loss": 0.5668, + "step": 21632 + }, + { + "epoch": 0.83, + "learning_rate": 3.6161195178250835e-05, + "loss": 0.5272, + "step": 21633 + }, + { + "epoch": 0.83, + "learning_rate": 3.6160553988202107e-05, + "loss": 0.4435, + "step": 21634 + }, + { + "epoch": 0.83, + "learning_rate": 3.615991279815337e-05, + "loss": 0.4793, + "step": 21635 + }, + { + "epoch": 0.83, + "learning_rate": 3.615927160810464e-05, + "loss": 0.5472, + "step": 21636 + }, + { + "epoch": 0.83, + "learning_rate": 3.615863041805591e-05, + "loss": 0.4811, + "step": 21637 + }, + { + "epoch": 0.83, + "learning_rate": 3.615798922800718e-05, + "loss": 0.5615, + "step": 21638 + }, + { + "epoch": 0.83, + "learning_rate": 3.615734803795846e-05, + "loss": 0.4481, + "step": 21639 + }, + { + "epoch": 0.83, + "learning_rate": 3.615670684790972e-05, + "loss": 0.4767, + "step": 21640 + }, + { + "epoch": 0.83, + "learning_rate": 3.6156065657860995e-05, + "loss": 0.5364, + "step": 21641 + }, + { + "epoch": 0.83, + "learning_rate": 3.615542446781226e-05, + "loss": 0.433, + "step": 21642 + }, + { + "epoch": 0.83, + "learning_rate": 3.615478327776353e-05, + "loss": 0.5759, + "step": 21643 + }, + { + "epoch": 0.83, + "learning_rate": 3.6154142087714797e-05, + "loss": 0.5024, + "step": 21644 + }, + { + "epoch": 0.83, + "learning_rate": 3.615350089766607e-05, + "loss": 0.5265, + "step": 21645 + }, + { + "epoch": 0.83, + "learning_rate": 3.615285970761734e-05, + "loss": 0.5423, + "step": 21646 + }, + { + "epoch": 0.83, + "learning_rate": 3.615221851756861e-05, + "loss": 0.4615, + "step": 21647 + }, + { + "epoch": 0.83, + "learning_rate": 3.6151577327519883e-05, + "loss": 0.4772, + "step": 21648 + }, + { + "epoch": 0.83, + "learning_rate": 3.615093613747115e-05, + "loss": 0.371, + "step": 21649 + }, + { + "epoch": 0.83, + "learning_rate": 3.615029494742242e-05, + "loss": 0.389, + "step": 21650 + }, + { + "epoch": 0.83, + "learning_rate": 3.6149653757373685e-05, + "loss": 0.5916, + "step": 21651 + }, + { + "epoch": 0.83, + "learning_rate": 3.614901256732496e-05, + "loss": 0.5237, + "step": 21652 + }, + { + "epoch": 0.83, + "learning_rate": 3.614837137727622e-05, + "loss": 0.5467, + "step": 21653 + }, + { + "epoch": 0.83, + "learning_rate": 3.6147730187227493e-05, + "loss": 0.3556, + "step": 21654 + }, + { + "epoch": 0.83, + "learning_rate": 3.6147088997178765e-05, + "loss": 0.4915, + "step": 21655 + }, + { + "epoch": 0.83, + "learning_rate": 3.614644780713004e-05, + "loss": 0.4236, + "step": 21656 + }, + { + "epoch": 0.83, + "learning_rate": 3.614580661708131e-05, + "loss": 0.4516, + "step": 21657 + }, + { + "epoch": 0.83, + "learning_rate": 3.6145165427032574e-05, + "loss": 0.3992, + "step": 21658 + }, + { + "epoch": 0.83, + "learning_rate": 3.6144524236983845e-05, + "loss": 0.6475, + "step": 21659 + }, + { + "epoch": 0.83, + "learning_rate": 3.614388304693511e-05, + "loss": 0.5197, + "step": 21660 + }, + { + "epoch": 0.83, + "learning_rate": 3.614324185688638e-05, + "loss": 0.4378, + "step": 21661 + }, + { + "epoch": 0.83, + "learning_rate": 3.6142600666837654e-05, + "loss": 0.5142, + "step": 21662 + }, + { + "epoch": 0.83, + "learning_rate": 3.6141959476788925e-05, + "loss": 0.4941, + "step": 21663 + }, + { + "epoch": 0.83, + "learning_rate": 3.614131828674019e-05, + "loss": 0.4828, + "step": 21664 + }, + { + "epoch": 0.83, + "learning_rate": 3.614067709669146e-05, + "loss": 0.4537, + "step": 21665 + }, + { + "epoch": 0.83, + "learning_rate": 3.6140035906642734e-05, + "loss": 0.4608, + "step": 21666 + }, + { + "epoch": 0.83, + "learning_rate": 3.6139394716594e-05, + "loss": 0.4366, + "step": 21667 + }, + { + "epoch": 0.83, + "learning_rate": 3.613875352654527e-05, + "loss": 0.553, + "step": 21668 + }, + { + "epoch": 0.83, + "learning_rate": 3.6138112336496535e-05, + "loss": 0.5977, + "step": 21669 + }, + { + "epoch": 0.83, + "learning_rate": 3.613747114644781e-05, + "loss": 0.5867, + "step": 21670 + }, + { + "epoch": 0.83, + "learning_rate": 3.613682995639908e-05, + "loss": 0.5419, + "step": 21671 + }, + { + "epoch": 0.83, + "learning_rate": 3.613618876635035e-05, + "loss": 0.4071, + "step": 21672 + }, + { + "epoch": 0.83, + "learning_rate": 3.6135547576301615e-05, + "loss": 0.4826, + "step": 21673 + }, + { + "epoch": 0.83, + "learning_rate": 3.613490638625289e-05, + "loss": 0.5626, + "step": 21674 + }, + { + "epoch": 0.83, + "learning_rate": 3.613426519620416e-05, + "loss": 0.4385, + "step": 21675 + }, + { + "epoch": 0.83, + "learning_rate": 3.6133624006155424e-05, + "loss": 0.5209, + "step": 21676 + }, + { + "epoch": 0.83, + "learning_rate": 3.6132982816106695e-05, + "loss": 0.5045, + "step": 21677 + }, + { + "epoch": 0.83, + "learning_rate": 3.613234162605797e-05, + "loss": 0.4856, + "step": 21678 + }, + { + "epoch": 0.83, + "learning_rate": 3.613170043600924e-05, + "loss": 0.5065, + "step": 21679 + }, + { + "epoch": 0.83, + "learning_rate": 3.6131059245960504e-05, + "loss": 0.4894, + "step": 21680 + }, + { + "epoch": 0.83, + "learning_rate": 3.6130418055911776e-05, + "loss": 0.5528, + "step": 21681 + }, + { + "epoch": 0.83, + "learning_rate": 3.612977686586304e-05, + "loss": 0.5086, + "step": 21682 + }, + { + "epoch": 0.83, + "learning_rate": 3.612913567581431e-05, + "loss": 0.38, + "step": 21683 + }, + { + "epoch": 0.83, + "learning_rate": 3.6128494485765584e-05, + "loss": 0.4533, + "step": 21684 + }, + { + "epoch": 0.83, + "learning_rate": 3.612785329571685e-05, + "loss": 0.4015, + "step": 21685 + }, + { + "epoch": 0.83, + "learning_rate": 3.612721210566812e-05, + "loss": 0.4272, + "step": 21686 + }, + { + "epoch": 0.83, + "learning_rate": 3.612657091561939e-05, + "loss": 0.3811, + "step": 21687 + }, + { + "epoch": 0.83, + "learning_rate": 3.6125929725570664e-05, + "loss": 0.5643, + "step": 21688 + }, + { + "epoch": 0.83, + "learning_rate": 3.612528853552193e-05, + "loss": 0.5244, + "step": 21689 + }, + { + "epoch": 0.83, + "learning_rate": 3.61246473454732e-05, + "loss": 0.5685, + "step": 21690 + }, + { + "epoch": 0.83, + "learning_rate": 3.6124006155424466e-05, + "loss": 0.5547, + "step": 21691 + }, + { + "epoch": 0.83, + "learning_rate": 3.612336496537574e-05, + "loss": 0.533, + "step": 21692 + }, + { + "epoch": 0.83, + "learning_rate": 3.612272377532701e-05, + "loss": 0.6401, + "step": 21693 + }, + { + "epoch": 0.83, + "learning_rate": 3.6122082585278274e-05, + "loss": 0.4959, + "step": 21694 + }, + { + "epoch": 0.83, + "learning_rate": 3.612144139522955e-05, + "loss": 0.4176, + "step": 21695 + }, + { + "epoch": 0.83, + "learning_rate": 3.612080020518082e-05, + "loss": 0.4171, + "step": 21696 + }, + { + "epoch": 0.83, + "learning_rate": 3.612015901513209e-05, + "loss": 0.6938, + "step": 21697 + }, + { + "epoch": 0.83, + "learning_rate": 3.6119517825083354e-05, + "loss": 0.4926, + "step": 21698 + }, + { + "epoch": 0.83, + "learning_rate": 3.6118876635034626e-05, + "loss": 0.5268, + "step": 21699 + }, + { + "epoch": 0.83, + "learning_rate": 3.611823544498589e-05, + "loss": 0.5176, + "step": 21700 + }, + { + "epoch": 0.83, + "learning_rate": 3.611759425493716e-05, + "loss": 0.5797, + "step": 21701 + }, + { + "epoch": 0.83, + "learning_rate": 3.6116953064888434e-05, + "loss": 0.4625, + "step": 21702 + }, + { + "epoch": 0.83, + "learning_rate": 3.6116311874839706e-05, + "loss": 0.5335, + "step": 21703 + }, + { + "epoch": 0.83, + "learning_rate": 3.611567068479098e-05, + "loss": 0.4801, + "step": 21704 + }, + { + "epoch": 0.83, + "learning_rate": 3.611502949474224e-05, + "loss": 0.4772, + "step": 21705 + }, + { + "epoch": 0.83, + "learning_rate": 3.6114388304693514e-05, + "loss": 0.4491, + "step": 21706 + }, + { + "epoch": 0.83, + "learning_rate": 3.611374711464478e-05, + "loss": 0.464, + "step": 21707 + }, + { + "epoch": 0.83, + "learning_rate": 3.611310592459605e-05, + "loss": 0.5962, + "step": 21708 + }, + { + "epoch": 0.83, + "learning_rate": 3.6112464734547316e-05, + "loss": 0.4462, + "step": 21709 + }, + { + "epoch": 0.83, + "learning_rate": 3.611182354449859e-05, + "loss": 0.5786, + "step": 21710 + }, + { + "epoch": 0.83, + "learning_rate": 3.6111182354449866e-05, + "loss": 0.4255, + "step": 21711 + }, + { + "epoch": 0.83, + "learning_rate": 3.611054116440113e-05, + "loss": 0.5036, + "step": 21712 + }, + { + "epoch": 0.83, + "learning_rate": 3.61098999743524e-05, + "loss": 0.6152, + "step": 21713 + }, + { + "epoch": 0.83, + "learning_rate": 3.610925878430367e-05, + "loss": 0.4049, + "step": 21714 + }, + { + "epoch": 0.83, + "learning_rate": 3.610861759425494e-05, + "loss": 0.4987, + "step": 21715 + }, + { + "epoch": 0.83, + "learning_rate": 3.6107976404206204e-05, + "loss": 0.5359, + "step": 21716 + }, + { + "epoch": 0.83, + "learning_rate": 3.6107335214157476e-05, + "loss": 0.6399, + "step": 21717 + }, + { + "epoch": 0.83, + "learning_rate": 3.610669402410875e-05, + "loss": 0.4826, + "step": 21718 + }, + { + "epoch": 0.84, + "learning_rate": 3.610605283406002e-05, + "loss": 0.5607, + "step": 21719 + }, + { + "epoch": 0.84, + "learning_rate": 3.610541164401129e-05, + "loss": 0.4902, + "step": 21720 + }, + { + "epoch": 0.84, + "learning_rate": 3.6104770453962556e-05, + "loss": 0.5687, + "step": 21721 + }, + { + "epoch": 0.84, + "learning_rate": 3.610412926391383e-05, + "loss": 0.3524, + "step": 21722 + }, + { + "epoch": 0.84, + "learning_rate": 3.610348807386509e-05, + "loss": 0.4642, + "step": 21723 + }, + { + "epoch": 0.84, + "learning_rate": 3.6102846883816365e-05, + "loss": 0.5211, + "step": 21724 + }, + { + "epoch": 0.84, + "learning_rate": 3.610220569376763e-05, + "loss": 0.4102, + "step": 21725 + }, + { + "epoch": 0.84, + "learning_rate": 3.61015645037189e-05, + "loss": 0.5514, + "step": 21726 + }, + { + "epoch": 0.84, + "learning_rate": 3.610092331367018e-05, + "loss": 0.4703, + "step": 21727 + }, + { + "epoch": 0.84, + "learning_rate": 3.6100282123621445e-05, + "loss": 0.5325, + "step": 21728 + }, + { + "epoch": 0.84, + "learning_rate": 3.6099640933572716e-05, + "loss": 0.4673, + "step": 21729 + }, + { + "epoch": 0.84, + "learning_rate": 3.609899974352398e-05, + "loss": 0.5674, + "step": 21730 + }, + { + "epoch": 0.84, + "learning_rate": 3.609835855347525e-05, + "loss": 0.4695, + "step": 21731 + }, + { + "epoch": 0.84, + "learning_rate": 3.609771736342652e-05, + "loss": 0.4655, + "step": 21732 + }, + { + "epoch": 0.84, + "learning_rate": 3.609707617337779e-05, + "loss": 0.5549, + "step": 21733 + }, + { + "epoch": 0.84, + "learning_rate": 3.609643498332906e-05, + "loss": 0.4674, + "step": 21734 + }, + { + "epoch": 0.84, + "learning_rate": 3.609579379328033e-05, + "loss": 0.4112, + "step": 21735 + }, + { + "epoch": 0.84, + "learning_rate": 3.6095152603231605e-05, + "loss": 0.5669, + "step": 21736 + }, + { + "epoch": 0.84, + "learning_rate": 3.609451141318287e-05, + "loss": 0.4512, + "step": 21737 + }, + { + "epoch": 0.84, + "learning_rate": 3.609387022313414e-05, + "loss": 0.523, + "step": 21738 + }, + { + "epoch": 0.84, + "learning_rate": 3.6093229033085406e-05, + "loss": 0.6165, + "step": 21739 + }, + { + "epoch": 0.84, + "learning_rate": 3.609258784303668e-05, + "loss": 0.4849, + "step": 21740 + }, + { + "epoch": 0.84, + "learning_rate": 3.609194665298794e-05, + "loss": 0.4723, + "step": 21741 + }, + { + "epoch": 0.84, + "learning_rate": 3.6091305462939215e-05, + "loss": 0.47, + "step": 21742 + }, + { + "epoch": 0.84, + "learning_rate": 3.6090664272890486e-05, + "loss": 0.4683, + "step": 21743 + }, + { + "epoch": 0.84, + "learning_rate": 3.609002308284176e-05, + "loss": 0.4351, + "step": 21744 + }, + { + "epoch": 0.84, + "learning_rate": 3.608938189279303e-05, + "loss": 0.5439, + "step": 21745 + }, + { + "epoch": 0.84, + "learning_rate": 3.6088740702744295e-05, + "loss": 0.6232, + "step": 21746 + }, + { + "epoch": 0.84, + "learning_rate": 3.6088099512695567e-05, + "loss": 0.5284, + "step": 21747 + }, + { + "epoch": 0.84, + "learning_rate": 3.608745832264683e-05, + "loss": 0.4917, + "step": 21748 + }, + { + "epoch": 0.84, + "learning_rate": 3.60868171325981e-05, + "loss": 0.4844, + "step": 21749 + }, + { + "epoch": 0.84, + "learning_rate": 3.6086175942549375e-05, + "loss": 0.6006, + "step": 21750 + }, + { + "epoch": 0.84, + "learning_rate": 3.608553475250065e-05, + "loss": 0.4468, + "step": 21751 + }, + { + "epoch": 0.84, + "learning_rate": 3.608489356245191e-05, + "loss": 0.3941, + "step": 21752 + }, + { + "epoch": 0.84, + "learning_rate": 3.608425237240318e-05, + "loss": 0.3896, + "step": 21753 + }, + { + "epoch": 0.84, + "learning_rate": 3.6083611182354455e-05, + "loss": 0.425, + "step": 21754 + }, + { + "epoch": 0.84, + "learning_rate": 3.608296999230572e-05, + "loss": 0.4503, + "step": 21755 + }, + { + "epoch": 0.84, + "learning_rate": 3.608232880225699e-05, + "loss": 0.5581, + "step": 21756 + }, + { + "epoch": 0.84, + "learning_rate": 3.608168761220826e-05, + "loss": 0.4755, + "step": 21757 + }, + { + "epoch": 0.84, + "learning_rate": 3.608104642215953e-05, + "loss": 0.581, + "step": 21758 + }, + { + "epoch": 0.84, + "learning_rate": 3.60804052321108e-05, + "loss": 0.4868, + "step": 21759 + }, + { + "epoch": 0.84, + "learning_rate": 3.607976404206207e-05, + "loss": 0.5247, + "step": 21760 + }, + { + "epoch": 0.84, + "learning_rate": 3.607912285201334e-05, + "loss": 0.4435, + "step": 21761 + }, + { + "epoch": 0.84, + "learning_rate": 3.607848166196461e-05, + "loss": 0.4749, + "step": 21762 + }, + { + "epoch": 0.84, + "learning_rate": 3.607784047191588e-05, + "loss": 0.4768, + "step": 21763 + }, + { + "epoch": 0.84, + "learning_rate": 3.6077199281867145e-05, + "loss": 0.4267, + "step": 21764 + }, + { + "epoch": 0.84, + "learning_rate": 3.607655809181842e-05, + "loss": 0.5214, + "step": 21765 + }, + { + "epoch": 0.84, + "learning_rate": 3.607591690176969e-05, + "loss": 0.3295, + "step": 21766 + }, + { + "epoch": 0.84, + "learning_rate": 3.607527571172096e-05, + "loss": 0.7009, + "step": 21767 + }, + { + "epoch": 0.84, + "learning_rate": 3.6074634521672225e-05, + "loss": 0.4745, + "step": 21768 + }, + { + "epoch": 0.84, + "learning_rate": 3.60739933316235e-05, + "loss": 0.4064, + "step": 21769 + }, + { + "epoch": 0.84, + "learning_rate": 3.607335214157476e-05, + "loss": 0.4708, + "step": 21770 + }, + { + "epoch": 0.84, + "learning_rate": 3.6072710951526034e-05, + "loss": 0.6675, + "step": 21771 + }, + { + "epoch": 0.84, + "learning_rate": 3.6072069761477305e-05, + "loss": 0.5559, + "step": 21772 + }, + { + "epoch": 0.84, + "learning_rate": 3.607142857142857e-05, + "loss": 0.6697, + "step": 21773 + }, + { + "epoch": 0.84, + "learning_rate": 3.607078738137984e-05, + "loss": 0.4855, + "step": 21774 + }, + { + "epoch": 0.84, + "learning_rate": 3.6070146191331114e-05, + "loss": 0.4584, + "step": 21775 + }, + { + "epoch": 0.84, + "learning_rate": 3.6069505001282385e-05, + "loss": 0.501, + "step": 21776 + }, + { + "epoch": 0.84, + "learning_rate": 3.606886381123365e-05, + "loss": 0.5028, + "step": 21777 + }, + { + "epoch": 0.84, + "learning_rate": 3.606822262118492e-05, + "loss": 0.4217, + "step": 21778 + }, + { + "epoch": 0.84, + "learning_rate": 3.606758143113619e-05, + "loss": 0.5324, + "step": 21779 + }, + { + "epoch": 0.84, + "learning_rate": 3.606694024108746e-05, + "loss": 0.4421, + "step": 21780 + }, + { + "epoch": 0.84, + "learning_rate": 3.606629905103873e-05, + "loss": 0.4927, + "step": 21781 + }, + { + "epoch": 0.84, + "learning_rate": 3.6065657860989995e-05, + "loss": 0.4691, + "step": 21782 + }, + { + "epoch": 0.84, + "learning_rate": 3.6065016670941274e-05, + "loss": 0.5694, + "step": 21783 + }, + { + "epoch": 0.84, + "learning_rate": 3.606437548089254e-05, + "loss": 0.5737, + "step": 21784 + }, + { + "epoch": 0.84, + "learning_rate": 3.606373429084381e-05, + "loss": 0.5552, + "step": 21785 + }, + { + "epoch": 0.84, + "learning_rate": 3.6063093100795075e-05, + "loss": 0.5154, + "step": 21786 + }, + { + "epoch": 0.84, + "learning_rate": 3.606245191074635e-05, + "loss": 0.4671, + "step": 21787 + }, + { + "epoch": 0.84, + "learning_rate": 3.606181072069761e-05, + "loss": 0.4334, + "step": 21788 + }, + { + "epoch": 0.84, + "learning_rate": 3.6061169530648884e-05, + "loss": 0.5747, + "step": 21789 + }, + { + "epoch": 0.84, + "learning_rate": 3.6060528340600156e-05, + "loss": 0.5255, + "step": 21790 + }, + { + "epoch": 0.84, + "learning_rate": 3.605988715055143e-05, + "loss": 0.4887, + "step": 21791 + }, + { + "epoch": 0.84, + "learning_rate": 3.60592459605027e-05, + "loss": 0.5137, + "step": 21792 + }, + { + "epoch": 0.84, + "learning_rate": 3.6058604770453964e-05, + "loss": 0.453, + "step": 21793 + }, + { + "epoch": 0.84, + "learning_rate": 3.6057963580405236e-05, + "loss": 0.4517, + "step": 21794 + }, + { + "epoch": 0.84, + "learning_rate": 3.60573223903565e-05, + "loss": 0.5837, + "step": 21795 + }, + { + "epoch": 0.84, + "learning_rate": 3.605668120030777e-05, + "loss": 0.5739, + "step": 21796 + }, + { + "epoch": 0.84, + "learning_rate": 3.605604001025904e-05, + "loss": 0.4353, + "step": 21797 + }, + { + "epoch": 0.84, + "learning_rate": 3.605539882021031e-05, + "loss": 0.4037, + "step": 21798 + }, + { + "epoch": 0.84, + "learning_rate": 3.605475763016159e-05, + "loss": 0.5122, + "step": 21799 + }, + { + "epoch": 0.84, + "learning_rate": 3.605411644011285e-05, + "loss": 0.5409, + "step": 21800 + }, + { + "epoch": 0.84, + "learning_rate": 3.6053475250064124e-05, + "loss": 0.6248, + "step": 21801 + }, + { + "epoch": 0.84, + "learning_rate": 3.605283406001539e-05, + "loss": 0.5005, + "step": 21802 + }, + { + "epoch": 0.84, + "learning_rate": 3.605219286996666e-05, + "loss": 0.4678, + "step": 21803 + }, + { + "epoch": 0.84, + "learning_rate": 3.6051551679917926e-05, + "loss": 0.4334, + "step": 21804 + }, + { + "epoch": 0.84, + "learning_rate": 3.60509104898692e-05, + "loss": 0.5033, + "step": 21805 + }, + { + "epoch": 0.84, + "learning_rate": 3.605026929982047e-05, + "loss": 0.4976, + "step": 21806 + }, + { + "epoch": 0.84, + "learning_rate": 3.604962810977174e-05, + "loss": 0.4514, + "step": 21807 + }, + { + "epoch": 0.84, + "learning_rate": 3.604898691972301e-05, + "loss": 0.4987, + "step": 21808 + }, + { + "epoch": 0.84, + "learning_rate": 3.604834572967428e-05, + "loss": 0.4724, + "step": 21809 + }, + { + "epoch": 0.84, + "learning_rate": 3.604770453962555e-05, + "loss": 0.4745, + "step": 21810 + }, + { + "epoch": 0.84, + "learning_rate": 3.6047063349576814e-05, + "loss": 0.439, + "step": 21811 + }, + { + "epoch": 0.84, + "learning_rate": 3.6046422159528086e-05, + "loss": 0.4514, + "step": 21812 + }, + { + "epoch": 0.84, + "learning_rate": 3.604578096947935e-05, + "loss": 0.4963, + "step": 21813 + }, + { + "epoch": 0.84, + "learning_rate": 3.604513977943062e-05, + "loss": 0.5076, + "step": 21814 + }, + { + "epoch": 0.84, + "learning_rate": 3.6044498589381894e-05, + "loss": 0.7132, + "step": 21815 + }, + { + "epoch": 0.84, + "learning_rate": 3.6043857399333166e-05, + "loss": 0.4418, + "step": 21816 + }, + { + "epoch": 0.84, + "learning_rate": 3.604321620928444e-05, + "loss": 0.4756, + "step": 21817 + }, + { + "epoch": 0.84, + "learning_rate": 3.60425750192357e-05, + "loss": 0.3741, + "step": 21818 + }, + { + "epoch": 0.84, + "learning_rate": 3.6041933829186974e-05, + "loss": 0.4807, + "step": 21819 + }, + { + "epoch": 0.84, + "learning_rate": 3.604129263913824e-05, + "loss": 0.6205, + "step": 21820 + }, + { + "epoch": 0.84, + "learning_rate": 3.604065144908951e-05, + "loss": 0.3524, + "step": 21821 + }, + { + "epoch": 0.84, + "learning_rate": 3.604001025904078e-05, + "loss": 0.5979, + "step": 21822 + }, + { + "epoch": 0.84, + "learning_rate": 3.6039369068992054e-05, + "loss": 0.6092, + "step": 21823 + }, + { + "epoch": 0.84, + "learning_rate": 3.603872787894332e-05, + "loss": 0.55, + "step": 21824 + }, + { + "epoch": 0.84, + "learning_rate": 3.603808668889459e-05, + "loss": 0.5845, + "step": 21825 + }, + { + "epoch": 0.84, + "learning_rate": 3.603744549884586e-05, + "loss": 0.6112, + "step": 21826 + }, + { + "epoch": 0.84, + "learning_rate": 3.603680430879713e-05, + "loss": 0.5738, + "step": 21827 + }, + { + "epoch": 0.84, + "learning_rate": 3.60361631187484e-05, + "loss": 0.5193, + "step": 21828 + }, + { + "epoch": 0.84, + "learning_rate": 3.6035521928699664e-05, + "loss": 0.4559, + "step": 21829 + }, + { + "epoch": 0.84, + "learning_rate": 3.6034880738650936e-05, + "loss": 0.5963, + "step": 21830 + }, + { + "epoch": 0.84, + "learning_rate": 3.603423954860221e-05, + "loss": 0.5788, + "step": 21831 + }, + { + "epoch": 0.84, + "learning_rate": 3.603359835855348e-05, + "loss": 0.5974, + "step": 21832 + }, + { + "epoch": 0.84, + "learning_rate": 3.6032957168504744e-05, + "loss": 0.4508, + "step": 21833 + }, + { + "epoch": 0.84, + "learning_rate": 3.6032315978456016e-05, + "loss": 0.5691, + "step": 21834 + }, + { + "epoch": 0.84, + "learning_rate": 3.603167478840729e-05, + "loss": 0.5182, + "step": 21835 + }, + { + "epoch": 0.84, + "learning_rate": 3.603103359835855e-05, + "loss": 0.5113, + "step": 21836 + }, + { + "epoch": 0.84, + "learning_rate": 3.6030392408309825e-05, + "loss": 0.5066, + "step": 21837 + }, + { + "epoch": 0.84, + "learning_rate": 3.6029751218261096e-05, + "loss": 0.5532, + "step": 21838 + }, + { + "epoch": 0.84, + "learning_rate": 3.602911002821237e-05, + "loss": 0.5749, + "step": 21839 + }, + { + "epoch": 0.84, + "learning_rate": 3.602846883816363e-05, + "loss": 0.4856, + "step": 21840 + }, + { + "epoch": 0.84, + "learning_rate": 3.6027827648114905e-05, + "loss": 0.6437, + "step": 21841 + }, + { + "epoch": 0.84, + "learning_rate": 3.602718645806617e-05, + "loss": 0.5107, + "step": 21842 + }, + { + "epoch": 0.84, + "learning_rate": 3.602654526801744e-05, + "loss": 0.5159, + "step": 21843 + }, + { + "epoch": 0.84, + "learning_rate": 3.602590407796871e-05, + "loss": 0.461, + "step": 21844 + }, + { + "epoch": 0.84, + "learning_rate": 3.602526288791998e-05, + "loss": 0.45, + "step": 21845 + }, + { + "epoch": 0.84, + "learning_rate": 3.602462169787125e-05, + "loss": 0.4572, + "step": 21846 + }, + { + "epoch": 0.84, + "learning_rate": 3.602398050782252e-05, + "loss": 0.5289, + "step": 21847 + }, + { + "epoch": 0.84, + "learning_rate": 3.602333931777379e-05, + "loss": 0.4804, + "step": 21848 + }, + { + "epoch": 0.84, + "learning_rate": 3.602269812772506e-05, + "loss": 0.5122, + "step": 21849 + }, + { + "epoch": 0.84, + "learning_rate": 3.602205693767633e-05, + "loss": 0.4251, + "step": 21850 + }, + { + "epoch": 0.84, + "learning_rate": 3.6021415747627595e-05, + "loss": 0.5088, + "step": 21851 + }, + { + "epoch": 0.84, + "learning_rate": 3.6020774557578866e-05, + "loss": 0.5442, + "step": 21852 + }, + { + "epoch": 0.84, + "learning_rate": 3.602013336753014e-05, + "loss": 0.4734, + "step": 21853 + }, + { + "epoch": 0.84, + "learning_rate": 3.601949217748141e-05, + "loss": 0.5376, + "step": 21854 + }, + { + "epoch": 0.84, + "learning_rate": 3.601885098743268e-05, + "loss": 0.4578, + "step": 21855 + }, + { + "epoch": 0.84, + "learning_rate": 3.6018209797383947e-05, + "loss": 0.5035, + "step": 21856 + }, + { + "epoch": 0.84, + "learning_rate": 3.601756860733522e-05, + "loss": 0.6708, + "step": 21857 + }, + { + "epoch": 0.84, + "learning_rate": 3.601692741728648e-05, + "loss": 0.4253, + "step": 21858 + }, + { + "epoch": 0.84, + "learning_rate": 3.6016286227237755e-05, + "loss": 0.3781, + "step": 21859 + }, + { + "epoch": 0.84, + "learning_rate": 3.601564503718902e-05, + "loss": 0.3275, + "step": 21860 + }, + { + "epoch": 0.84, + "learning_rate": 3.601500384714029e-05, + "loss": 0.4619, + "step": 21861 + }, + { + "epoch": 0.84, + "learning_rate": 3.601436265709156e-05, + "loss": 0.3188, + "step": 21862 + }, + { + "epoch": 0.84, + "learning_rate": 3.6013721467042835e-05, + "loss": 0.485, + "step": 21863 + }, + { + "epoch": 0.84, + "learning_rate": 3.601308027699411e-05, + "loss": 0.4356, + "step": 21864 + }, + { + "epoch": 0.84, + "learning_rate": 3.601243908694537e-05, + "loss": 0.44, + "step": 21865 + }, + { + "epoch": 0.84, + "learning_rate": 3.601179789689664e-05, + "loss": 0.5409, + "step": 21866 + }, + { + "epoch": 0.84, + "learning_rate": 3.601115670684791e-05, + "loss": 0.4568, + "step": 21867 + }, + { + "epoch": 0.84, + "learning_rate": 3.601051551679918e-05, + "loss": 0.5212, + "step": 21868 + }, + { + "epoch": 0.84, + "learning_rate": 3.6009874326750445e-05, + "loss": 0.5109, + "step": 21869 + }, + { + "epoch": 0.84, + "learning_rate": 3.600923313670172e-05, + "loss": 0.5984, + "step": 21870 + }, + { + "epoch": 0.84, + "learning_rate": 3.6008591946652995e-05, + "loss": 0.4104, + "step": 21871 + }, + { + "epoch": 0.84, + "learning_rate": 3.600795075660426e-05, + "loss": 0.4843, + "step": 21872 + }, + { + "epoch": 0.84, + "learning_rate": 3.600730956655553e-05, + "loss": 0.4641, + "step": 21873 + }, + { + "epoch": 0.84, + "learning_rate": 3.60066683765068e-05, + "loss": 0.5455, + "step": 21874 + }, + { + "epoch": 0.84, + "learning_rate": 3.600602718645807e-05, + "loss": 0.4621, + "step": 21875 + }, + { + "epoch": 0.84, + "learning_rate": 3.6005385996409333e-05, + "loss": 0.44, + "step": 21876 + }, + { + "epoch": 0.84, + "learning_rate": 3.6004744806360605e-05, + "loss": 0.3812, + "step": 21877 + }, + { + "epoch": 0.84, + "learning_rate": 3.600410361631188e-05, + "loss": 0.6563, + "step": 21878 + }, + { + "epoch": 0.84, + "learning_rate": 3.600346242626315e-05, + "loss": 0.487, + "step": 21879 + }, + { + "epoch": 0.84, + "learning_rate": 3.600282123621442e-05, + "loss": 0.4355, + "step": 21880 + }, + { + "epoch": 0.84, + "learning_rate": 3.6002180046165685e-05, + "loss": 0.4428, + "step": 21881 + }, + { + "epoch": 0.84, + "learning_rate": 3.600153885611696e-05, + "loss": 0.4626, + "step": 21882 + }, + { + "epoch": 0.84, + "learning_rate": 3.600089766606822e-05, + "loss": 0.5317, + "step": 21883 + }, + { + "epoch": 0.84, + "learning_rate": 3.6000256476019494e-05, + "loss": 0.4112, + "step": 21884 + }, + { + "epoch": 0.84, + "learning_rate": 3.599961528597076e-05, + "loss": 0.4292, + "step": 21885 + }, + { + "epoch": 0.84, + "learning_rate": 3.599897409592203e-05, + "loss": 0.47, + "step": 21886 + }, + { + "epoch": 0.84, + "learning_rate": 3.59983329058733e-05, + "loss": 0.4553, + "step": 21887 + }, + { + "epoch": 0.84, + "learning_rate": 3.5997691715824574e-05, + "loss": 0.4912, + "step": 21888 + }, + { + "epoch": 0.84, + "learning_rate": 3.5997050525775845e-05, + "loss": 0.4665, + "step": 21889 + }, + { + "epoch": 0.84, + "learning_rate": 3.599640933572711e-05, + "loss": 0.3954, + "step": 21890 + }, + { + "epoch": 0.84, + "learning_rate": 3.599576814567838e-05, + "loss": 0.4068, + "step": 21891 + }, + { + "epoch": 0.84, + "learning_rate": 3.599512695562965e-05, + "loss": 0.537, + "step": 21892 + }, + { + "epoch": 0.84, + "learning_rate": 3.599448576558092e-05, + "loss": 0.5836, + "step": 21893 + }, + { + "epoch": 0.84, + "learning_rate": 3.599384457553219e-05, + "loss": 0.5865, + "step": 21894 + }, + { + "epoch": 0.84, + "learning_rate": 3.599320338548346e-05, + "loss": 0.6558, + "step": 21895 + }, + { + "epoch": 0.84, + "learning_rate": 3.599256219543473e-05, + "loss": 0.4939, + "step": 21896 + }, + { + "epoch": 0.84, + "learning_rate": 3.5991921005386e-05, + "loss": 0.5093, + "step": 21897 + }, + { + "epoch": 0.84, + "learning_rate": 3.599127981533727e-05, + "loss": 0.4709, + "step": 21898 + }, + { + "epoch": 0.84, + "learning_rate": 3.5990638625288535e-05, + "loss": 0.4309, + "step": 21899 + }, + { + "epoch": 0.84, + "learning_rate": 3.598999743523981e-05, + "loss": 0.5815, + "step": 21900 + }, + { + "epoch": 0.84, + "learning_rate": 3.598935624519107e-05, + "loss": 0.5006, + "step": 21901 + }, + { + "epoch": 0.84, + "learning_rate": 3.5988715055142344e-05, + "loss": 0.5402, + "step": 21902 + }, + { + "epoch": 0.84, + "learning_rate": 3.5988073865093616e-05, + "loss": 0.5602, + "step": 21903 + }, + { + "epoch": 0.84, + "learning_rate": 3.598743267504489e-05, + "loss": 0.5826, + "step": 21904 + }, + { + "epoch": 0.84, + "learning_rate": 3.598679148499615e-05, + "loss": 0.5101, + "step": 21905 + }, + { + "epoch": 0.84, + "learning_rate": 3.5986150294947424e-05, + "loss": 0.531, + "step": 21906 + }, + { + "epoch": 0.84, + "learning_rate": 3.5985509104898696e-05, + "loss": 0.6236, + "step": 21907 + }, + { + "epoch": 0.84, + "learning_rate": 3.598486791484996e-05, + "loss": 0.6068, + "step": 21908 + }, + { + "epoch": 0.84, + "learning_rate": 3.598422672480123e-05, + "loss": 0.533, + "step": 21909 + }, + { + "epoch": 0.84, + "learning_rate": 3.5983585534752504e-05, + "loss": 0.5386, + "step": 21910 + }, + { + "epoch": 0.84, + "learning_rate": 3.5982944344703776e-05, + "loss": 0.5528, + "step": 21911 + }, + { + "epoch": 0.84, + "learning_rate": 3.598230315465504e-05, + "loss": 0.538, + "step": 21912 + }, + { + "epoch": 0.84, + "learning_rate": 3.598166196460631e-05, + "loss": 0.4862, + "step": 21913 + }, + { + "epoch": 0.84, + "learning_rate": 3.5981020774557584e-05, + "loss": 0.5257, + "step": 21914 + }, + { + "epoch": 0.84, + "learning_rate": 3.598037958450885e-05, + "loss": 0.4266, + "step": 21915 + }, + { + "epoch": 0.84, + "learning_rate": 3.597973839446012e-05, + "loss": 0.5262, + "step": 21916 + }, + { + "epoch": 0.84, + "learning_rate": 3.5979097204411386e-05, + "loss": 0.5884, + "step": 21917 + }, + { + "epoch": 0.84, + "learning_rate": 3.597845601436266e-05, + "loss": 0.4423, + "step": 21918 + }, + { + "epoch": 0.84, + "learning_rate": 3.597781482431393e-05, + "loss": 0.4629, + "step": 21919 + }, + { + "epoch": 0.84, + "learning_rate": 3.59771736342652e-05, + "loss": 0.5004, + "step": 21920 + }, + { + "epoch": 0.84, + "learning_rate": 3.5976532444216466e-05, + "loss": 0.4522, + "step": 21921 + }, + { + "epoch": 0.84, + "learning_rate": 3.597589125416774e-05, + "loss": 0.5246, + "step": 21922 + }, + { + "epoch": 0.84, + "learning_rate": 3.597525006411901e-05, + "loss": 0.5519, + "step": 21923 + }, + { + "epoch": 0.84, + "learning_rate": 3.5974608874070274e-05, + "loss": 0.4454, + "step": 21924 + }, + { + "epoch": 0.84, + "learning_rate": 3.5973967684021546e-05, + "loss": 0.5279, + "step": 21925 + }, + { + "epoch": 0.84, + "learning_rate": 3.597332649397282e-05, + "loss": 0.6292, + "step": 21926 + }, + { + "epoch": 0.84, + "learning_rate": 3.597268530392409e-05, + "loss": 0.5338, + "step": 21927 + }, + { + "epoch": 0.84, + "learning_rate": 3.5972044113875354e-05, + "loss": 0.4216, + "step": 21928 + }, + { + "epoch": 0.84, + "learning_rate": 3.5971402923826626e-05, + "loss": 0.4643, + "step": 21929 + }, + { + "epoch": 0.84, + "learning_rate": 3.597076173377789e-05, + "loss": 0.4425, + "step": 21930 + }, + { + "epoch": 0.84, + "learning_rate": 3.597012054372916e-05, + "loss": 0.6044, + "step": 21931 + }, + { + "epoch": 0.84, + "learning_rate": 3.5969479353680434e-05, + "loss": 0.4716, + "step": 21932 + }, + { + "epoch": 0.84, + "learning_rate": 3.59688381636317e-05, + "loss": 0.6377, + "step": 21933 + }, + { + "epoch": 0.84, + "learning_rate": 3.596819697358297e-05, + "loss": 0.4844, + "step": 21934 + }, + { + "epoch": 0.84, + "learning_rate": 3.596755578353424e-05, + "loss": 0.5832, + "step": 21935 + }, + { + "epoch": 0.84, + "learning_rate": 3.5966914593485514e-05, + "loss": 0.5755, + "step": 21936 + }, + { + "epoch": 0.84, + "learning_rate": 3.596627340343678e-05, + "loss": 0.4638, + "step": 21937 + }, + { + "epoch": 0.84, + "learning_rate": 3.596563221338805e-05, + "loss": 0.4807, + "step": 21938 + }, + { + "epoch": 0.84, + "learning_rate": 3.5964991023339316e-05, + "loss": 0.4192, + "step": 21939 + }, + { + "epoch": 0.84, + "learning_rate": 3.596434983329059e-05, + "loss": 0.4651, + "step": 21940 + }, + { + "epoch": 0.84, + "learning_rate": 3.596370864324186e-05, + "loss": 0.5549, + "step": 21941 + }, + { + "epoch": 0.84, + "learning_rate": 3.5963067453193124e-05, + "loss": 0.5468, + "step": 21942 + }, + { + "epoch": 0.84, + "learning_rate": 3.59624262631444e-05, + "loss": 0.5592, + "step": 21943 + }, + { + "epoch": 0.84, + "learning_rate": 3.596178507309567e-05, + "loss": 0.4791, + "step": 21944 + }, + { + "epoch": 0.84, + "learning_rate": 3.596114388304694e-05, + "loss": 0.5205, + "step": 21945 + }, + { + "epoch": 0.84, + "learning_rate": 3.5960502692998204e-05, + "loss": 0.4264, + "step": 21946 + }, + { + "epoch": 0.84, + "learning_rate": 3.5959861502949476e-05, + "loss": 0.5283, + "step": 21947 + }, + { + "epoch": 0.84, + "learning_rate": 3.595922031290074e-05, + "loss": 0.4323, + "step": 21948 + }, + { + "epoch": 0.84, + "learning_rate": 3.595857912285201e-05, + "loss": 0.5278, + "step": 21949 + }, + { + "epoch": 0.84, + "learning_rate": 3.5957937932803285e-05, + "loss": 0.492, + "step": 21950 + }, + { + "epoch": 0.84, + "learning_rate": 3.5957296742754556e-05, + "loss": 0.484, + "step": 21951 + }, + { + "epoch": 0.84, + "learning_rate": 3.595665555270583e-05, + "loss": 0.4532, + "step": 21952 + }, + { + "epoch": 0.84, + "learning_rate": 3.595601436265709e-05, + "loss": 0.4053, + "step": 21953 + }, + { + "epoch": 0.84, + "learning_rate": 3.5955373172608365e-05, + "loss": 0.5007, + "step": 21954 + }, + { + "epoch": 0.84, + "learning_rate": 3.595473198255963e-05, + "loss": 0.4271, + "step": 21955 + }, + { + "epoch": 0.84, + "learning_rate": 3.59540907925109e-05, + "loss": 0.5578, + "step": 21956 + }, + { + "epoch": 0.84, + "learning_rate": 3.5953449602462166e-05, + "loss": 0.6154, + "step": 21957 + }, + { + "epoch": 0.84, + "learning_rate": 3.595280841241344e-05, + "loss": 0.618, + "step": 21958 + }, + { + "epoch": 0.84, + "learning_rate": 3.5952167222364716e-05, + "loss": 0.4635, + "step": 21959 + }, + { + "epoch": 0.84, + "learning_rate": 3.595152603231598e-05, + "loss": 0.5624, + "step": 21960 + }, + { + "epoch": 0.84, + "learning_rate": 3.595088484226725e-05, + "loss": 0.4566, + "step": 21961 + }, + { + "epoch": 0.84, + "learning_rate": 3.595024365221852e-05, + "loss": 0.5093, + "step": 21962 + }, + { + "epoch": 0.84, + "learning_rate": 3.594960246216979e-05, + "loss": 0.4981, + "step": 21963 + }, + { + "epoch": 0.84, + "learning_rate": 3.5948961272121055e-05, + "loss": 0.5591, + "step": 21964 + }, + { + "epoch": 0.84, + "learning_rate": 3.5948320082072326e-05, + "loss": 0.4932, + "step": 21965 + }, + { + "epoch": 0.84, + "learning_rate": 3.59476788920236e-05, + "loss": 0.4837, + "step": 21966 + }, + { + "epoch": 0.84, + "learning_rate": 3.594703770197487e-05, + "loss": 0.3956, + "step": 21967 + }, + { + "epoch": 0.84, + "learning_rate": 3.594639651192614e-05, + "loss": 0.4612, + "step": 21968 + }, + { + "epoch": 0.84, + "learning_rate": 3.5945755321877407e-05, + "loss": 0.4889, + "step": 21969 + }, + { + "epoch": 0.84, + "learning_rate": 3.594511413182868e-05, + "loss": 0.4579, + "step": 21970 + }, + { + "epoch": 0.84, + "learning_rate": 3.594447294177994e-05, + "loss": 0.5552, + "step": 21971 + }, + { + "epoch": 0.84, + "learning_rate": 3.5943831751731215e-05, + "loss": 0.4813, + "step": 21972 + }, + { + "epoch": 0.84, + "learning_rate": 3.594319056168248e-05, + "loss": 0.5594, + "step": 21973 + }, + { + "epoch": 0.84, + "learning_rate": 3.594254937163375e-05, + "loss": 0.5393, + "step": 21974 + }, + { + "epoch": 0.84, + "learning_rate": 3.594190818158502e-05, + "loss": 0.531, + "step": 21975 + }, + { + "epoch": 0.84, + "learning_rate": 3.5941266991536295e-05, + "loss": 0.6136, + "step": 21976 + }, + { + "epoch": 0.84, + "learning_rate": 3.594062580148757e-05, + "loss": 0.424, + "step": 21977 + }, + { + "epoch": 0.84, + "learning_rate": 3.593998461143883e-05, + "loss": 0.4618, + "step": 21978 + }, + { + "epoch": 0.85, + "learning_rate": 3.59393434213901e-05, + "loss": 0.5338, + "step": 21979 + }, + { + "epoch": 0.85, + "learning_rate": 3.593870223134137e-05, + "loss": 0.542, + "step": 21980 + }, + { + "epoch": 0.85, + "learning_rate": 3.593806104129264e-05, + "loss": 0.4889, + "step": 21981 + }, + { + "epoch": 0.85, + "learning_rate": 3.593741985124391e-05, + "loss": 0.444, + "step": 21982 + }, + { + "epoch": 0.85, + "learning_rate": 3.5936778661195183e-05, + "loss": 0.5267, + "step": 21983 + }, + { + "epoch": 0.85, + "learning_rate": 3.593613747114645e-05, + "loss": 0.4571, + "step": 21984 + }, + { + "epoch": 0.85, + "learning_rate": 3.593549628109772e-05, + "loss": 0.5361, + "step": 21985 + }, + { + "epoch": 0.85, + "learning_rate": 3.593485509104899e-05, + "loss": 0.4618, + "step": 21986 + }, + { + "epoch": 0.85, + "learning_rate": 3.593421390100026e-05, + "loss": 0.4507, + "step": 21987 + }, + { + "epoch": 0.85, + "learning_rate": 3.593357271095153e-05, + "loss": 0.4599, + "step": 21988 + }, + { + "epoch": 0.85, + "learning_rate": 3.5932931520902793e-05, + "loss": 0.5798, + "step": 21989 + }, + { + "epoch": 0.85, + "learning_rate": 3.5932290330854065e-05, + "loss": 0.5008, + "step": 21990 + }, + { + "epoch": 0.85, + "learning_rate": 3.593164914080534e-05, + "loss": 0.4204, + "step": 21991 + }, + { + "epoch": 0.85, + "learning_rate": 3.593100795075661e-05, + "loss": 0.4866, + "step": 21992 + }, + { + "epoch": 0.85, + "learning_rate": 3.5930366760707874e-05, + "loss": 0.5339, + "step": 21993 + }, + { + "epoch": 0.85, + "learning_rate": 3.5929725570659145e-05, + "loss": 0.4635, + "step": 21994 + }, + { + "epoch": 0.85, + "learning_rate": 3.592908438061042e-05, + "loss": 0.5634, + "step": 21995 + }, + { + "epoch": 0.85, + "learning_rate": 3.592844319056168e-05, + "loss": 0.4608, + "step": 21996 + }, + { + "epoch": 0.85, + "learning_rate": 3.5927802000512954e-05, + "loss": 0.4431, + "step": 21997 + }, + { + "epoch": 0.85, + "learning_rate": 3.5927160810464225e-05, + "loss": 0.5457, + "step": 21998 + }, + { + "epoch": 0.85, + "learning_rate": 3.59265196204155e-05, + "loss": 0.4626, + "step": 21999 + }, + { + "epoch": 0.85, + "learning_rate": 3.592587843036676e-05, + "loss": 0.6074, + "step": 22000 + }, + { + "epoch": 0.85, + "learning_rate": 3.5925237240318034e-05, + "loss": 0.4393, + "step": 22001 + }, + { + "epoch": 0.85, + "learning_rate": 3.59245960502693e-05, + "loss": 0.4819, + "step": 22002 + }, + { + "epoch": 0.85, + "learning_rate": 3.592395486022057e-05, + "loss": 0.4222, + "step": 22003 + }, + { + "epoch": 0.85, + "learning_rate": 3.592331367017184e-05, + "loss": 0.4017, + "step": 22004 + }, + { + "epoch": 0.85, + "learning_rate": 3.592267248012311e-05, + "loss": 0.4493, + "step": 22005 + }, + { + "epoch": 0.85, + "learning_rate": 3.592203129007438e-05, + "loss": 0.5453, + "step": 22006 + }, + { + "epoch": 0.85, + "learning_rate": 3.592139010002565e-05, + "loss": 0.5765, + "step": 22007 + }, + { + "epoch": 0.85, + "learning_rate": 3.592074890997692e-05, + "loss": 0.5165, + "step": 22008 + }, + { + "epoch": 0.85, + "learning_rate": 3.592010771992819e-05, + "loss": 0.4813, + "step": 22009 + }, + { + "epoch": 0.85, + "learning_rate": 3.591946652987946e-05, + "loss": 0.5679, + "step": 22010 + }, + { + "epoch": 0.85, + "learning_rate": 3.5918825339830724e-05, + "loss": 0.4941, + "step": 22011 + }, + { + "epoch": 0.85, + "learning_rate": 3.5918184149781995e-05, + "loss": 0.6454, + "step": 22012 + }, + { + "epoch": 0.85, + "learning_rate": 3.591754295973327e-05, + "loss": 0.4918, + "step": 22013 + }, + { + "epoch": 0.85, + "learning_rate": 3.591690176968454e-05, + "loss": 0.4418, + "step": 22014 + }, + { + "epoch": 0.85, + "learning_rate": 3.591626057963581e-05, + "loss": 0.4293, + "step": 22015 + }, + { + "epoch": 0.85, + "learning_rate": 3.5915619389587076e-05, + "loss": 0.473, + "step": 22016 + }, + { + "epoch": 0.85, + "learning_rate": 3.591497819953835e-05, + "loss": 0.3982, + "step": 22017 + }, + { + "epoch": 0.85, + "learning_rate": 3.591433700948961e-05, + "loss": 0.5299, + "step": 22018 + }, + { + "epoch": 0.85, + "learning_rate": 3.5913695819440884e-05, + "loss": 0.4631, + "step": 22019 + }, + { + "epoch": 0.85, + "learning_rate": 3.591305462939215e-05, + "loss": 0.5171, + "step": 22020 + }, + { + "epoch": 0.85, + "learning_rate": 3.591241343934342e-05, + "loss": 0.4732, + "step": 22021 + }, + { + "epoch": 0.85, + "learning_rate": 3.591177224929469e-05, + "loss": 0.4643, + "step": 22022 + }, + { + "epoch": 0.85, + "learning_rate": 3.5911131059245964e-05, + "loss": 0.6147, + "step": 22023 + }, + { + "epoch": 0.85, + "learning_rate": 3.5910489869197236e-05, + "loss": 0.4331, + "step": 22024 + }, + { + "epoch": 0.85, + "learning_rate": 3.59098486791485e-05, + "loss": 0.5691, + "step": 22025 + }, + { + "epoch": 0.85, + "learning_rate": 3.590920748909977e-05, + "loss": 0.4766, + "step": 22026 + }, + { + "epoch": 0.85, + "learning_rate": 3.590856629905104e-05, + "loss": 0.5558, + "step": 22027 + }, + { + "epoch": 0.85, + "learning_rate": 3.590792510900231e-05, + "loss": 0.4436, + "step": 22028 + }, + { + "epoch": 0.85, + "learning_rate": 3.5907283918953574e-05, + "loss": 0.5091, + "step": 22029 + }, + { + "epoch": 0.85, + "learning_rate": 3.5906642728904846e-05, + "loss": 0.4407, + "step": 22030 + }, + { + "epoch": 0.85, + "learning_rate": 3.5906001538856124e-05, + "loss": 0.4828, + "step": 22031 + }, + { + "epoch": 0.85, + "learning_rate": 3.590536034880739e-05, + "loss": 0.589, + "step": 22032 + }, + { + "epoch": 0.85, + "learning_rate": 3.590471915875866e-05, + "loss": 0.481, + "step": 22033 + }, + { + "epoch": 0.85, + "learning_rate": 3.5904077968709926e-05, + "loss": 0.4884, + "step": 22034 + }, + { + "epoch": 0.85, + "learning_rate": 3.59034367786612e-05, + "loss": 0.5641, + "step": 22035 + }, + { + "epoch": 0.85, + "learning_rate": 3.590279558861246e-05, + "loss": 0.3718, + "step": 22036 + }, + { + "epoch": 0.85, + "learning_rate": 3.5902154398563734e-05, + "loss": 0.4534, + "step": 22037 + }, + { + "epoch": 0.85, + "learning_rate": 3.5901513208515006e-05, + "loss": 0.4096, + "step": 22038 + }, + { + "epoch": 0.85, + "learning_rate": 3.590087201846628e-05, + "loss": 0.4034, + "step": 22039 + }, + { + "epoch": 0.85, + "learning_rate": 3.590023082841755e-05, + "loss": 0.3912, + "step": 22040 + }, + { + "epoch": 0.85, + "learning_rate": 3.5899589638368814e-05, + "loss": 0.4794, + "step": 22041 + }, + { + "epoch": 0.85, + "learning_rate": 3.5898948448320086e-05, + "loss": 0.4855, + "step": 22042 + }, + { + "epoch": 0.85, + "learning_rate": 3.589830725827135e-05, + "loss": 0.5439, + "step": 22043 + }, + { + "epoch": 0.85, + "learning_rate": 3.589766606822262e-05, + "loss": 0.4482, + "step": 22044 + }, + { + "epoch": 0.85, + "learning_rate": 3.589702487817389e-05, + "loss": 0.4696, + "step": 22045 + }, + { + "epoch": 0.85, + "learning_rate": 3.589638368812516e-05, + "loss": 0.463, + "step": 22046 + }, + { + "epoch": 0.85, + "learning_rate": 3.589574249807643e-05, + "loss": 0.5784, + "step": 22047 + }, + { + "epoch": 0.85, + "learning_rate": 3.58951013080277e-05, + "loss": 0.4077, + "step": 22048 + }, + { + "epoch": 0.85, + "learning_rate": 3.5894460117978974e-05, + "loss": 0.482, + "step": 22049 + }, + { + "epoch": 0.85, + "learning_rate": 3.589381892793024e-05, + "loss": 0.4053, + "step": 22050 + }, + { + "epoch": 0.85, + "learning_rate": 3.589317773788151e-05, + "loss": 0.6512, + "step": 22051 + }, + { + "epoch": 0.85, + "learning_rate": 3.5892536547832776e-05, + "loss": 0.5518, + "step": 22052 + }, + { + "epoch": 0.85, + "learning_rate": 3.589189535778405e-05, + "loss": 0.4566, + "step": 22053 + }, + { + "epoch": 0.85, + "learning_rate": 3.589125416773532e-05, + "loss": 0.4719, + "step": 22054 + }, + { + "epoch": 0.85, + "learning_rate": 3.589061297768659e-05, + "loss": 0.512, + "step": 22055 + }, + { + "epoch": 0.85, + "learning_rate": 3.5889971787637856e-05, + "loss": 0.5341, + "step": 22056 + }, + { + "epoch": 0.85, + "learning_rate": 3.588933059758913e-05, + "loss": 0.5054, + "step": 22057 + }, + { + "epoch": 0.85, + "learning_rate": 3.58886894075404e-05, + "loss": 0.5422, + "step": 22058 + }, + { + "epoch": 0.85, + "learning_rate": 3.5888048217491665e-05, + "loss": 0.4657, + "step": 22059 + }, + { + "epoch": 0.85, + "learning_rate": 3.5887407027442936e-05, + "loss": 0.4109, + "step": 22060 + }, + { + "epoch": 0.85, + "learning_rate": 3.58867658373942e-05, + "loss": 0.5155, + "step": 22061 + }, + { + "epoch": 0.85, + "learning_rate": 3.588612464734547e-05, + "loss": 0.4426, + "step": 22062 + }, + { + "epoch": 0.85, + "learning_rate": 3.5885483457296745e-05, + "loss": 0.4982, + "step": 22063 + }, + { + "epoch": 0.85, + "learning_rate": 3.5884842267248016e-05, + "loss": 0.5423, + "step": 22064 + }, + { + "epoch": 0.85, + "learning_rate": 3.588420107719928e-05, + "loss": 0.5851, + "step": 22065 + }, + { + "epoch": 0.85, + "learning_rate": 3.588355988715055e-05, + "loss": 0.495, + "step": 22066 + }, + { + "epoch": 0.85, + "learning_rate": 3.5882918697101825e-05, + "loss": 0.462, + "step": 22067 + }, + { + "epoch": 0.85, + "learning_rate": 3.588227750705309e-05, + "loss": 0.5283, + "step": 22068 + }, + { + "epoch": 0.85, + "learning_rate": 3.588163631700436e-05, + "loss": 0.5034, + "step": 22069 + }, + { + "epoch": 0.85, + "learning_rate": 3.588099512695563e-05, + "loss": 0.4886, + "step": 22070 + }, + { + "epoch": 0.85, + "learning_rate": 3.5880353936906905e-05, + "loss": 0.4565, + "step": 22071 + }, + { + "epoch": 0.85, + "learning_rate": 3.587971274685817e-05, + "loss": 0.5412, + "step": 22072 + }, + { + "epoch": 0.85, + "learning_rate": 3.587907155680944e-05, + "loss": 0.4744, + "step": 22073 + }, + { + "epoch": 0.85, + "learning_rate": 3.5878430366760706e-05, + "loss": 0.5535, + "step": 22074 + }, + { + "epoch": 0.85, + "learning_rate": 3.587778917671198e-05, + "loss": 0.4227, + "step": 22075 + }, + { + "epoch": 0.85, + "learning_rate": 3.587714798666325e-05, + "loss": 0.5821, + "step": 22076 + }, + { + "epoch": 0.85, + "learning_rate": 3.5876506796614515e-05, + "loss": 0.4939, + "step": 22077 + }, + { + "epoch": 0.85, + "learning_rate": 3.5875865606565786e-05, + "loss": 0.5301, + "step": 22078 + }, + { + "epoch": 0.85, + "learning_rate": 3.587522441651706e-05, + "loss": 0.5847, + "step": 22079 + }, + { + "epoch": 0.85, + "learning_rate": 3.587458322646833e-05, + "loss": 0.4845, + "step": 22080 + }, + { + "epoch": 0.85, + "learning_rate": 3.5873942036419595e-05, + "loss": 0.5688, + "step": 22081 + }, + { + "epoch": 0.85, + "learning_rate": 3.5873300846370867e-05, + "loss": 0.4523, + "step": 22082 + }, + { + "epoch": 0.85, + "learning_rate": 3.587265965632213e-05, + "loss": 0.4976, + "step": 22083 + }, + { + "epoch": 0.85, + "learning_rate": 3.58720184662734e-05, + "loss": 0.5255, + "step": 22084 + }, + { + "epoch": 0.85, + "learning_rate": 3.5871377276224675e-05, + "loss": 0.5202, + "step": 22085 + }, + { + "epoch": 0.85, + "learning_rate": 3.587073608617595e-05, + "loss": 0.5, + "step": 22086 + }, + { + "epoch": 0.85, + "learning_rate": 3.587009489612722e-05, + "loss": 0.4952, + "step": 22087 + }, + { + "epoch": 0.85, + "learning_rate": 3.586945370607848e-05, + "loss": 0.5348, + "step": 22088 + }, + { + "epoch": 0.85, + "learning_rate": 3.5868812516029755e-05, + "loss": 0.4022, + "step": 22089 + }, + { + "epoch": 0.85, + "learning_rate": 3.586817132598102e-05, + "loss": 0.5221, + "step": 22090 + }, + { + "epoch": 0.85, + "learning_rate": 3.586753013593229e-05, + "loss": 0.4224, + "step": 22091 + }, + { + "epoch": 0.85, + "learning_rate": 3.586688894588356e-05, + "loss": 0.4711, + "step": 22092 + }, + { + "epoch": 0.85, + "learning_rate": 3.586624775583483e-05, + "loss": 0.4157, + "step": 22093 + }, + { + "epoch": 0.85, + "learning_rate": 3.58656065657861e-05, + "loss": 0.4437, + "step": 22094 + }, + { + "epoch": 0.85, + "learning_rate": 3.586496537573737e-05, + "loss": 0.4325, + "step": 22095 + }, + { + "epoch": 0.85, + "learning_rate": 3.5864324185688643e-05, + "loss": 0.6081, + "step": 22096 + }, + { + "epoch": 0.85, + "learning_rate": 3.586368299563991e-05, + "loss": 0.5659, + "step": 22097 + }, + { + "epoch": 0.85, + "learning_rate": 3.586304180559118e-05, + "loss": 0.4105, + "step": 22098 + }, + { + "epoch": 0.85, + "learning_rate": 3.5862400615542445e-05, + "loss": 0.5258, + "step": 22099 + }, + { + "epoch": 0.85, + "learning_rate": 3.586175942549372e-05, + "loss": 0.5205, + "step": 22100 + }, + { + "epoch": 0.85, + "learning_rate": 3.586111823544499e-05, + "loss": 0.5069, + "step": 22101 + }, + { + "epoch": 0.85, + "learning_rate": 3.586047704539626e-05, + "loss": 0.5783, + "step": 22102 + }, + { + "epoch": 0.85, + "learning_rate": 3.585983585534753e-05, + "loss": 0.6467, + "step": 22103 + }, + { + "epoch": 0.85, + "learning_rate": 3.58591946652988e-05, + "loss": 0.4976, + "step": 22104 + }, + { + "epoch": 0.85, + "learning_rate": 3.585855347525007e-05, + "loss": 0.4976, + "step": 22105 + }, + { + "epoch": 0.85, + "learning_rate": 3.5857912285201334e-05, + "loss": 0.4483, + "step": 22106 + }, + { + "epoch": 0.85, + "learning_rate": 3.5857271095152605e-05, + "loss": 0.5737, + "step": 22107 + }, + { + "epoch": 0.85, + "learning_rate": 3.585662990510387e-05, + "loss": 0.3979, + "step": 22108 + }, + { + "epoch": 0.85, + "learning_rate": 3.585598871505514e-05, + "loss": 0.5161, + "step": 22109 + }, + { + "epoch": 0.85, + "learning_rate": 3.5855347525006414e-05, + "loss": 0.5005, + "step": 22110 + }, + { + "epoch": 0.85, + "learning_rate": 3.5854706334957685e-05, + "loss": 0.4214, + "step": 22111 + }, + { + "epoch": 0.85, + "learning_rate": 3.585406514490896e-05, + "loss": 0.5318, + "step": 22112 + }, + { + "epoch": 0.85, + "learning_rate": 3.585342395486022e-05, + "loss": 0.4805, + "step": 22113 + }, + { + "epoch": 0.85, + "learning_rate": 3.5852782764811494e-05, + "loss": 0.4753, + "step": 22114 + }, + { + "epoch": 0.85, + "learning_rate": 3.585214157476276e-05, + "loss": 0.5043, + "step": 22115 + }, + { + "epoch": 0.85, + "learning_rate": 3.585150038471403e-05, + "loss": 0.5459, + "step": 22116 + }, + { + "epoch": 0.85, + "learning_rate": 3.5850859194665295e-05, + "loss": 0.4423, + "step": 22117 + }, + { + "epoch": 0.85, + "learning_rate": 3.585021800461657e-05, + "loss": 0.5893, + "step": 22118 + }, + { + "epoch": 0.85, + "learning_rate": 3.5849576814567846e-05, + "loss": 0.459, + "step": 22119 + }, + { + "epoch": 0.85, + "learning_rate": 3.584893562451911e-05, + "loss": 0.456, + "step": 22120 + }, + { + "epoch": 0.85, + "learning_rate": 3.584829443447038e-05, + "loss": 0.5251, + "step": 22121 + }, + { + "epoch": 0.85, + "learning_rate": 3.584765324442165e-05, + "loss": 0.5206, + "step": 22122 + }, + { + "epoch": 0.85, + "learning_rate": 3.584701205437292e-05, + "loss": 0.5164, + "step": 22123 + }, + { + "epoch": 0.85, + "learning_rate": 3.5846370864324184e-05, + "loss": 0.4575, + "step": 22124 + }, + { + "epoch": 0.85, + "learning_rate": 3.5845729674275456e-05, + "loss": 0.4542, + "step": 22125 + }, + { + "epoch": 0.85, + "learning_rate": 3.584508848422673e-05, + "loss": 0.5744, + "step": 22126 + }, + { + "epoch": 0.85, + "learning_rate": 3.5844447294178e-05, + "loss": 0.5583, + "step": 22127 + }, + { + "epoch": 0.85, + "learning_rate": 3.584380610412927e-05, + "loss": 0.4627, + "step": 22128 + }, + { + "epoch": 0.85, + "learning_rate": 3.5843164914080536e-05, + "loss": 0.447, + "step": 22129 + }, + { + "epoch": 0.85, + "learning_rate": 3.584252372403181e-05, + "loss": 0.4369, + "step": 22130 + }, + { + "epoch": 0.85, + "learning_rate": 3.584188253398307e-05, + "loss": 0.6338, + "step": 22131 + }, + { + "epoch": 0.85, + "learning_rate": 3.5841241343934344e-05, + "loss": 0.5413, + "step": 22132 + }, + { + "epoch": 0.85, + "learning_rate": 3.584060015388561e-05, + "loss": 0.5127, + "step": 22133 + }, + { + "epoch": 0.85, + "learning_rate": 3.583995896383688e-05, + "loss": 0.4515, + "step": 22134 + }, + { + "epoch": 0.85, + "learning_rate": 3.583931777378815e-05, + "loss": 0.438, + "step": 22135 + }, + { + "epoch": 0.85, + "learning_rate": 3.5838676583739424e-05, + "loss": 0.4882, + "step": 22136 + }, + { + "epoch": 0.85, + "learning_rate": 3.5838035393690696e-05, + "loss": 0.4801, + "step": 22137 + }, + { + "epoch": 0.85, + "learning_rate": 3.583739420364196e-05, + "loss": 0.4945, + "step": 22138 + }, + { + "epoch": 0.85, + "learning_rate": 3.583675301359323e-05, + "loss": 0.5761, + "step": 22139 + }, + { + "epoch": 0.85, + "learning_rate": 3.58361118235445e-05, + "loss": 0.515, + "step": 22140 + }, + { + "epoch": 0.85, + "learning_rate": 3.583547063349577e-05, + "loss": 0.5082, + "step": 22141 + }, + { + "epoch": 0.85, + "learning_rate": 3.583482944344704e-05, + "loss": 0.4191, + "step": 22142 + }, + { + "epoch": 0.85, + "learning_rate": 3.583418825339831e-05, + "loss": 0.4898, + "step": 22143 + }, + { + "epoch": 0.85, + "learning_rate": 3.583354706334958e-05, + "loss": 0.5491, + "step": 22144 + }, + { + "epoch": 0.85, + "learning_rate": 3.583290587330085e-05, + "loss": 0.5116, + "step": 22145 + }, + { + "epoch": 0.85, + "learning_rate": 3.583226468325212e-05, + "loss": 0.4239, + "step": 22146 + }, + { + "epoch": 0.85, + "learning_rate": 3.5831623493203386e-05, + "loss": 0.3829, + "step": 22147 + }, + { + "epoch": 0.85, + "learning_rate": 3.583098230315466e-05, + "loss": 0.4527, + "step": 22148 + }, + { + "epoch": 0.85, + "learning_rate": 3.583034111310592e-05, + "loss": 0.4127, + "step": 22149 + }, + { + "epoch": 0.85, + "learning_rate": 3.5829699923057194e-05, + "loss": 0.4505, + "step": 22150 + }, + { + "epoch": 0.85, + "learning_rate": 3.5829058733008466e-05, + "loss": 0.5654, + "step": 22151 + }, + { + "epoch": 0.85, + "learning_rate": 3.582841754295974e-05, + "loss": 0.4836, + "step": 22152 + }, + { + "epoch": 0.85, + "learning_rate": 3.5827776352911e-05, + "loss": 0.4409, + "step": 22153 + }, + { + "epoch": 0.85, + "learning_rate": 3.5827135162862274e-05, + "loss": 0.5401, + "step": 22154 + }, + { + "epoch": 0.85, + "learning_rate": 3.5826493972813546e-05, + "loss": 0.4943, + "step": 22155 + }, + { + "epoch": 0.85, + "learning_rate": 3.582585278276481e-05, + "loss": 0.5623, + "step": 22156 + }, + { + "epoch": 0.85, + "learning_rate": 3.582521159271608e-05, + "loss": 0.4752, + "step": 22157 + }, + { + "epoch": 0.85, + "learning_rate": 3.5824570402667354e-05, + "loss": 0.5076, + "step": 22158 + }, + { + "epoch": 0.85, + "learning_rate": 3.5823929212618626e-05, + "loss": 0.6048, + "step": 22159 + }, + { + "epoch": 0.85, + "learning_rate": 3.582328802256989e-05, + "loss": 0.4036, + "step": 22160 + }, + { + "epoch": 0.85, + "learning_rate": 3.582264683252116e-05, + "loss": 0.4562, + "step": 22161 + }, + { + "epoch": 0.85, + "learning_rate": 3.582200564247243e-05, + "loss": 0.4888, + "step": 22162 + }, + { + "epoch": 0.85, + "learning_rate": 3.58213644524237e-05, + "loss": 0.4488, + "step": 22163 + }, + { + "epoch": 0.85, + "learning_rate": 3.582072326237497e-05, + "loss": 0.5588, + "step": 22164 + }, + { + "epoch": 0.85, + "learning_rate": 3.5820082072326236e-05, + "loss": 0.6175, + "step": 22165 + }, + { + "epoch": 0.85, + "learning_rate": 3.581944088227751e-05, + "loss": 0.4789, + "step": 22166 + }, + { + "epoch": 0.85, + "learning_rate": 3.581879969222878e-05, + "loss": 0.534, + "step": 22167 + }, + { + "epoch": 0.85, + "learning_rate": 3.581815850218005e-05, + "loss": 0.4316, + "step": 22168 + }, + { + "epoch": 0.85, + "learning_rate": 3.5817517312131316e-05, + "loss": 0.4916, + "step": 22169 + }, + { + "epoch": 0.85, + "learning_rate": 3.581687612208259e-05, + "loss": 0.5873, + "step": 22170 + }, + { + "epoch": 0.85, + "learning_rate": 3.581623493203385e-05, + "loss": 0.4549, + "step": 22171 + }, + { + "epoch": 0.85, + "learning_rate": 3.5815593741985125e-05, + "loss": 0.4306, + "step": 22172 + }, + { + "epoch": 0.85, + "learning_rate": 3.5814952551936396e-05, + "loss": 0.4138, + "step": 22173 + }, + { + "epoch": 0.85, + "learning_rate": 3.581431136188767e-05, + "loss": 0.3925, + "step": 22174 + }, + { + "epoch": 0.85, + "learning_rate": 3.581367017183894e-05, + "loss": 0.4589, + "step": 22175 + }, + { + "epoch": 0.85, + "learning_rate": 3.5813028981790205e-05, + "loss": 0.632, + "step": 22176 + }, + { + "epoch": 0.85, + "learning_rate": 3.5812387791741476e-05, + "loss": 0.5077, + "step": 22177 + }, + { + "epoch": 0.85, + "learning_rate": 3.581174660169274e-05, + "loss": 0.4002, + "step": 22178 + }, + { + "epoch": 0.85, + "learning_rate": 3.581110541164401e-05, + "loss": 0.4367, + "step": 22179 + }, + { + "epoch": 0.85, + "learning_rate": 3.581046422159528e-05, + "loss": 0.3763, + "step": 22180 + }, + { + "epoch": 0.85, + "learning_rate": 3.580982303154655e-05, + "loss": 0.5664, + "step": 22181 + }, + { + "epoch": 0.85, + "learning_rate": 3.580918184149782e-05, + "loss": 0.5368, + "step": 22182 + }, + { + "epoch": 0.85, + "learning_rate": 3.580854065144909e-05, + "loss": 0.4719, + "step": 22183 + }, + { + "epoch": 0.85, + "learning_rate": 3.5807899461400365e-05, + "loss": 0.4939, + "step": 22184 + }, + { + "epoch": 0.85, + "learning_rate": 3.580725827135163e-05, + "loss": 0.395, + "step": 22185 + }, + { + "epoch": 0.85, + "learning_rate": 3.58066170813029e-05, + "loss": 0.4624, + "step": 22186 + }, + { + "epoch": 0.85, + "learning_rate": 3.5805975891254166e-05, + "loss": 0.4796, + "step": 22187 + }, + { + "epoch": 0.85, + "learning_rate": 3.580533470120544e-05, + "loss": 0.511, + "step": 22188 + }, + { + "epoch": 0.85, + "learning_rate": 3.58046935111567e-05, + "loss": 0.5278, + "step": 22189 + }, + { + "epoch": 0.85, + "learning_rate": 3.5804052321107975e-05, + "loss": 0.577, + "step": 22190 + }, + { + "epoch": 0.85, + "learning_rate": 3.580341113105925e-05, + "loss": 0.6693, + "step": 22191 + }, + { + "epoch": 0.85, + "learning_rate": 3.580276994101052e-05, + "loss": 0.4766, + "step": 22192 + }, + { + "epoch": 0.85, + "learning_rate": 3.580212875096179e-05, + "loss": 0.6106, + "step": 22193 + }, + { + "epoch": 0.85, + "learning_rate": 3.5801487560913055e-05, + "loss": 0.4759, + "step": 22194 + }, + { + "epoch": 0.85, + "learning_rate": 3.5800846370864327e-05, + "loss": 0.58, + "step": 22195 + }, + { + "epoch": 0.85, + "learning_rate": 3.580020518081559e-05, + "loss": 0.5931, + "step": 22196 + }, + { + "epoch": 0.85, + "learning_rate": 3.579956399076686e-05, + "loss": 0.5259, + "step": 22197 + }, + { + "epoch": 0.85, + "learning_rate": 3.5798922800718135e-05, + "loss": 0.4608, + "step": 22198 + }, + { + "epoch": 0.85, + "learning_rate": 3.579828161066941e-05, + "loss": 0.4645, + "step": 22199 + }, + { + "epoch": 0.85, + "learning_rate": 3.579764042062068e-05, + "loss": 0.4876, + "step": 22200 + }, + { + "epoch": 0.85, + "learning_rate": 3.579699923057194e-05, + "loss": 0.5182, + "step": 22201 + }, + { + "epoch": 0.85, + "learning_rate": 3.5796358040523215e-05, + "loss": 0.4058, + "step": 22202 + }, + { + "epoch": 0.85, + "learning_rate": 3.579571685047448e-05, + "loss": 0.4733, + "step": 22203 + }, + { + "epoch": 0.85, + "learning_rate": 3.579507566042575e-05, + "loss": 0.444, + "step": 22204 + }, + { + "epoch": 0.85, + "learning_rate": 3.579443447037702e-05, + "loss": 0.5109, + "step": 22205 + }, + { + "epoch": 0.85, + "learning_rate": 3.579379328032829e-05, + "loss": 0.4952, + "step": 22206 + }, + { + "epoch": 0.85, + "learning_rate": 3.579315209027956e-05, + "loss": 0.4855, + "step": 22207 + }, + { + "epoch": 0.85, + "learning_rate": 3.579251090023083e-05, + "loss": 0.6015, + "step": 22208 + }, + { + "epoch": 0.85, + "learning_rate": 3.5791869710182104e-05, + "loss": 0.516, + "step": 22209 + }, + { + "epoch": 0.85, + "learning_rate": 3.579122852013337e-05, + "loss": 0.4548, + "step": 22210 + }, + { + "epoch": 0.85, + "learning_rate": 3.579058733008464e-05, + "loss": 0.3875, + "step": 22211 + }, + { + "epoch": 0.85, + "learning_rate": 3.5789946140035905e-05, + "loss": 0.5358, + "step": 22212 + }, + { + "epoch": 0.85, + "learning_rate": 3.578930494998718e-05, + "loss": 0.5342, + "step": 22213 + }, + { + "epoch": 0.85, + "learning_rate": 3.578866375993845e-05, + "loss": 0.5129, + "step": 22214 + }, + { + "epoch": 0.85, + "learning_rate": 3.578802256988972e-05, + "loss": 0.4777, + "step": 22215 + }, + { + "epoch": 0.85, + "learning_rate": 3.5787381379840985e-05, + "loss": 0.4659, + "step": 22216 + }, + { + "epoch": 0.85, + "learning_rate": 3.578674018979226e-05, + "loss": 0.5592, + "step": 22217 + }, + { + "epoch": 0.85, + "learning_rate": 3.578609899974353e-05, + "loss": 0.5415, + "step": 22218 + }, + { + "epoch": 0.85, + "learning_rate": 3.5785457809694794e-05, + "loss": 0.4659, + "step": 22219 + }, + { + "epoch": 0.85, + "learning_rate": 3.5784816619646065e-05, + "loss": 0.4905, + "step": 22220 + }, + { + "epoch": 0.85, + "learning_rate": 3.578417542959733e-05, + "loss": 0.3488, + "step": 22221 + }, + { + "epoch": 0.85, + "learning_rate": 3.57835342395486e-05, + "loss": 0.5311, + "step": 22222 + }, + { + "epoch": 0.85, + "learning_rate": 3.5782893049499874e-05, + "loss": 0.5775, + "step": 22223 + }, + { + "epoch": 0.85, + "learning_rate": 3.5782251859451145e-05, + "loss": 0.5527, + "step": 22224 + }, + { + "epoch": 0.85, + "learning_rate": 3.578161066940241e-05, + "loss": 0.4114, + "step": 22225 + }, + { + "epoch": 0.85, + "learning_rate": 3.578096947935368e-05, + "loss": 0.497, + "step": 22226 + }, + { + "epoch": 0.85, + "learning_rate": 3.5780328289304954e-05, + "loss": 0.5021, + "step": 22227 + }, + { + "epoch": 0.85, + "learning_rate": 3.577968709925622e-05, + "loss": 0.4332, + "step": 22228 + }, + { + "epoch": 0.85, + "learning_rate": 3.577904590920749e-05, + "loss": 0.4202, + "step": 22229 + }, + { + "epoch": 0.85, + "learning_rate": 3.577840471915876e-05, + "loss": 0.4936, + "step": 22230 + }, + { + "epoch": 0.85, + "learning_rate": 3.5777763529110034e-05, + "loss": 0.5055, + "step": 22231 + }, + { + "epoch": 0.85, + "learning_rate": 3.57771223390613e-05, + "loss": 0.495, + "step": 22232 + }, + { + "epoch": 0.85, + "learning_rate": 3.577648114901257e-05, + "loss": 0.4959, + "step": 22233 + }, + { + "epoch": 0.85, + "learning_rate": 3.5775839958963835e-05, + "loss": 0.5046, + "step": 22234 + }, + { + "epoch": 0.85, + "learning_rate": 3.577519876891511e-05, + "loss": 0.6257, + "step": 22235 + }, + { + "epoch": 0.85, + "learning_rate": 3.577455757886638e-05, + "loss": 0.6127, + "step": 22236 + }, + { + "epoch": 0.85, + "learning_rate": 3.5773916388817644e-05, + "loss": 0.5143, + "step": 22237 + }, + { + "epoch": 0.85, + "learning_rate": 3.5773275198768916e-05, + "loss": 0.5334, + "step": 22238 + }, + { + "epoch": 0.86, + "learning_rate": 3.577263400872019e-05, + "loss": 0.5048, + "step": 22239 + }, + { + "epoch": 0.86, + "learning_rate": 3.577199281867146e-05, + "loss": 0.5806, + "step": 22240 + }, + { + "epoch": 0.86, + "learning_rate": 3.5771351628622724e-05, + "loss": 0.4998, + "step": 22241 + }, + { + "epoch": 0.86, + "learning_rate": 3.5770710438573996e-05, + "loss": 0.5321, + "step": 22242 + }, + { + "epoch": 0.86, + "learning_rate": 3.577006924852526e-05, + "loss": 0.4777, + "step": 22243 + }, + { + "epoch": 0.86, + "learning_rate": 3.576942805847653e-05, + "loss": 0.4651, + "step": 22244 + }, + { + "epoch": 0.86, + "learning_rate": 3.5768786868427804e-05, + "loss": 0.5164, + "step": 22245 + }, + { + "epoch": 0.86, + "learning_rate": 3.5768145678379076e-05, + "loss": 0.5667, + "step": 22246 + }, + { + "epoch": 0.86, + "learning_rate": 3.576750448833035e-05, + "loss": 0.5725, + "step": 22247 + }, + { + "epoch": 0.86, + "learning_rate": 3.576686329828161e-05, + "loss": 0.4436, + "step": 22248 + }, + { + "epoch": 0.86, + "learning_rate": 3.5766222108232884e-05, + "loss": 0.4381, + "step": 22249 + }, + { + "epoch": 0.86, + "learning_rate": 3.576558091818415e-05, + "loss": 0.5192, + "step": 22250 + }, + { + "epoch": 0.86, + "learning_rate": 3.576493972813542e-05, + "loss": 0.5171, + "step": 22251 + }, + { + "epoch": 0.86, + "learning_rate": 3.5764298538086686e-05, + "loss": 0.4726, + "step": 22252 + }, + { + "epoch": 0.86, + "learning_rate": 3.576365734803796e-05, + "loss": 0.4378, + "step": 22253 + }, + { + "epoch": 0.86, + "learning_rate": 3.576301615798923e-05, + "loss": 0.4185, + "step": 22254 + }, + { + "epoch": 0.86, + "learning_rate": 3.57623749679405e-05, + "loss": 0.5847, + "step": 22255 + }, + { + "epoch": 0.86, + "learning_rate": 3.576173377789177e-05, + "loss": 0.5449, + "step": 22256 + }, + { + "epoch": 0.86, + "learning_rate": 3.576109258784304e-05, + "loss": 0.4265, + "step": 22257 + }, + { + "epoch": 0.86, + "learning_rate": 3.576045139779431e-05, + "loss": 0.5557, + "step": 22258 + }, + { + "epoch": 0.86, + "learning_rate": 3.5759810207745574e-05, + "loss": 0.5611, + "step": 22259 + }, + { + "epoch": 0.86, + "learning_rate": 3.5759169017696846e-05, + "loss": 0.4825, + "step": 22260 + }, + { + "epoch": 0.86, + "learning_rate": 3.575852782764811e-05, + "loss": 0.3844, + "step": 22261 + }, + { + "epoch": 0.86, + "learning_rate": 3.575788663759939e-05, + "loss": 0.5129, + "step": 22262 + }, + { + "epoch": 0.86, + "learning_rate": 3.575724544755066e-05, + "loss": 0.5553, + "step": 22263 + }, + { + "epoch": 0.86, + "learning_rate": 3.5756604257501926e-05, + "loss": 0.4965, + "step": 22264 + }, + { + "epoch": 0.86, + "learning_rate": 3.57559630674532e-05, + "loss": 0.4942, + "step": 22265 + }, + { + "epoch": 0.86, + "learning_rate": 3.575532187740446e-05, + "loss": 0.4138, + "step": 22266 + }, + { + "epoch": 0.86, + "learning_rate": 3.5754680687355734e-05, + "loss": 0.4877, + "step": 22267 + }, + { + "epoch": 0.86, + "learning_rate": 3.5754039497307e-05, + "loss": 0.4311, + "step": 22268 + }, + { + "epoch": 0.86, + "learning_rate": 3.575339830725827e-05, + "loss": 0.6152, + "step": 22269 + }, + { + "epoch": 0.86, + "learning_rate": 3.575275711720954e-05, + "loss": 0.4675, + "step": 22270 + }, + { + "epoch": 0.86, + "learning_rate": 3.5752115927160814e-05, + "loss": 0.4772, + "step": 22271 + }, + { + "epoch": 0.86, + "learning_rate": 3.5751474737112086e-05, + "loss": 0.5833, + "step": 22272 + }, + { + "epoch": 0.86, + "learning_rate": 3.575083354706335e-05, + "loss": 0.5734, + "step": 22273 + }, + { + "epoch": 0.86, + "learning_rate": 3.575019235701462e-05, + "loss": 0.4913, + "step": 22274 + }, + { + "epoch": 0.86, + "learning_rate": 3.574955116696589e-05, + "loss": 0.4689, + "step": 22275 + }, + { + "epoch": 0.86, + "learning_rate": 3.574890997691716e-05, + "loss": 0.5355, + "step": 22276 + }, + { + "epoch": 0.86, + "learning_rate": 3.5748268786868424e-05, + "loss": 0.451, + "step": 22277 + }, + { + "epoch": 0.86, + "learning_rate": 3.5747627596819696e-05, + "loss": 0.5208, + "step": 22278 + }, + { + "epoch": 0.86, + "learning_rate": 3.574698640677097e-05, + "loss": 0.3925, + "step": 22279 + }, + { + "epoch": 0.86, + "learning_rate": 3.574634521672224e-05, + "loss": 0.5106, + "step": 22280 + }, + { + "epoch": 0.86, + "learning_rate": 3.574570402667351e-05, + "loss": 0.534, + "step": 22281 + }, + { + "epoch": 0.86, + "learning_rate": 3.5745062836624776e-05, + "loss": 0.4476, + "step": 22282 + }, + { + "epoch": 0.86, + "learning_rate": 3.574442164657605e-05, + "loss": 0.437, + "step": 22283 + }, + { + "epoch": 0.86, + "learning_rate": 3.574378045652731e-05, + "loss": 0.5225, + "step": 22284 + }, + { + "epoch": 0.86, + "learning_rate": 3.5743139266478585e-05, + "loss": 0.5874, + "step": 22285 + }, + { + "epoch": 0.86, + "learning_rate": 3.5742498076429856e-05, + "loss": 0.4934, + "step": 22286 + }, + { + "epoch": 0.86, + "learning_rate": 3.574185688638113e-05, + "loss": 0.4375, + "step": 22287 + }, + { + "epoch": 0.86, + "learning_rate": 3.574121569633239e-05, + "loss": 0.5452, + "step": 22288 + }, + { + "epoch": 0.86, + "learning_rate": 3.5740574506283665e-05, + "loss": 0.4212, + "step": 22289 + }, + { + "epoch": 0.86, + "learning_rate": 3.5739933316234936e-05, + "loss": 0.4602, + "step": 22290 + }, + { + "epoch": 0.86, + "learning_rate": 3.57392921261862e-05, + "loss": 0.6454, + "step": 22291 + }, + { + "epoch": 0.86, + "learning_rate": 3.573865093613747e-05, + "loss": 0.585, + "step": 22292 + }, + { + "epoch": 0.86, + "learning_rate": 3.573800974608874e-05, + "loss": 0.4966, + "step": 22293 + }, + { + "epoch": 0.86, + "learning_rate": 3.573736855604001e-05, + "loss": 0.4938, + "step": 22294 + }, + { + "epoch": 0.86, + "learning_rate": 3.573672736599128e-05, + "loss": 0.509, + "step": 22295 + }, + { + "epoch": 0.86, + "learning_rate": 3.573608617594255e-05, + "loss": 0.582, + "step": 22296 + }, + { + "epoch": 0.86, + "learning_rate": 3.5735444985893825e-05, + "loss": 0.4551, + "step": 22297 + }, + { + "epoch": 0.86, + "learning_rate": 3.573480379584509e-05, + "loss": 0.5112, + "step": 22298 + }, + { + "epoch": 0.86, + "learning_rate": 3.573416260579636e-05, + "loss": 0.4823, + "step": 22299 + }, + { + "epoch": 0.86, + "learning_rate": 3.5733521415747626e-05, + "loss": 0.5488, + "step": 22300 + }, + { + "epoch": 0.86, + "learning_rate": 3.57328802256989e-05, + "loss": 0.4954, + "step": 22301 + }, + { + "epoch": 0.86, + "learning_rate": 3.573223903565017e-05, + "loss": 0.4902, + "step": 22302 + }, + { + "epoch": 0.86, + "learning_rate": 3.573159784560144e-05, + "loss": 0.4986, + "step": 22303 + }, + { + "epoch": 0.86, + "learning_rate": 3.5730956655552707e-05, + "loss": 0.4707, + "step": 22304 + }, + { + "epoch": 0.86, + "learning_rate": 3.573031546550398e-05, + "loss": 0.5053, + "step": 22305 + }, + { + "epoch": 0.86, + "learning_rate": 3.572967427545525e-05, + "loss": 0.525, + "step": 22306 + }, + { + "epoch": 0.86, + "learning_rate": 3.5729033085406515e-05, + "loss": 0.5485, + "step": 22307 + }, + { + "epoch": 0.86, + "learning_rate": 3.572839189535779e-05, + "loss": 0.5267, + "step": 22308 + }, + { + "epoch": 0.86, + "learning_rate": 3.572775070530905e-05, + "loss": 0.4862, + "step": 22309 + }, + { + "epoch": 0.86, + "learning_rate": 3.572710951526032e-05, + "loss": 0.4401, + "step": 22310 + }, + { + "epoch": 0.86, + "learning_rate": 3.5726468325211595e-05, + "loss": 0.4588, + "step": 22311 + }, + { + "epoch": 0.86, + "learning_rate": 3.572582713516287e-05, + "loss": 0.484, + "step": 22312 + }, + { + "epoch": 0.86, + "learning_rate": 3.572518594511413e-05, + "loss": 0.5029, + "step": 22313 + }, + { + "epoch": 0.86, + "learning_rate": 3.57245447550654e-05, + "loss": 0.5646, + "step": 22314 + }, + { + "epoch": 0.86, + "learning_rate": 3.5723903565016675e-05, + "loss": 0.4751, + "step": 22315 + }, + { + "epoch": 0.86, + "learning_rate": 3.572326237496794e-05, + "loss": 0.5515, + "step": 22316 + }, + { + "epoch": 0.86, + "learning_rate": 3.572262118491921e-05, + "loss": 0.4187, + "step": 22317 + }, + { + "epoch": 0.86, + "learning_rate": 3.5721979994870483e-05, + "loss": 0.5494, + "step": 22318 + }, + { + "epoch": 0.86, + "learning_rate": 3.5721338804821755e-05, + "loss": 0.4325, + "step": 22319 + }, + { + "epoch": 0.86, + "learning_rate": 3.572069761477302e-05, + "loss": 0.4679, + "step": 22320 + }, + { + "epoch": 0.86, + "learning_rate": 3.572005642472429e-05, + "loss": 0.5055, + "step": 22321 + }, + { + "epoch": 0.86, + "learning_rate": 3.571941523467556e-05, + "loss": 0.4369, + "step": 22322 + }, + { + "epoch": 0.86, + "learning_rate": 3.571877404462683e-05, + "loss": 0.4727, + "step": 22323 + }, + { + "epoch": 0.86, + "learning_rate": 3.57181328545781e-05, + "loss": 0.5619, + "step": 22324 + }, + { + "epoch": 0.86, + "learning_rate": 3.5717491664529365e-05, + "loss": 0.5207, + "step": 22325 + }, + { + "epoch": 0.86, + "learning_rate": 3.571685047448064e-05, + "loss": 0.4896, + "step": 22326 + }, + { + "epoch": 0.86, + "learning_rate": 3.571620928443191e-05, + "loss": 0.4165, + "step": 22327 + }, + { + "epoch": 0.86, + "learning_rate": 3.571556809438318e-05, + "loss": 0.5314, + "step": 22328 + }, + { + "epoch": 0.86, + "learning_rate": 3.5714926904334445e-05, + "loss": 0.4876, + "step": 22329 + }, + { + "epoch": 0.86, + "learning_rate": 3.571428571428572e-05, + "loss": 0.4557, + "step": 22330 + }, + { + "epoch": 0.86, + "learning_rate": 3.571364452423698e-05, + "loss": 0.4904, + "step": 22331 + }, + { + "epoch": 0.86, + "learning_rate": 3.5713003334188254e-05, + "loss": 0.4054, + "step": 22332 + }, + { + "epoch": 0.86, + "learning_rate": 3.5712362144139525e-05, + "loss": 0.4884, + "step": 22333 + }, + { + "epoch": 0.86, + "learning_rate": 3.57117209540908e-05, + "loss": 0.3936, + "step": 22334 + }, + { + "epoch": 0.86, + "learning_rate": 3.571107976404207e-05, + "loss": 0.5072, + "step": 22335 + }, + { + "epoch": 0.86, + "learning_rate": 3.5710438573993334e-05, + "loss": 0.4628, + "step": 22336 + }, + { + "epoch": 0.86, + "learning_rate": 3.5709797383944605e-05, + "loss": 0.53, + "step": 22337 + }, + { + "epoch": 0.86, + "learning_rate": 3.570915619389587e-05, + "loss": 0.382, + "step": 22338 + }, + { + "epoch": 0.86, + "learning_rate": 3.570851500384714e-05, + "loss": 0.5434, + "step": 22339 + }, + { + "epoch": 0.86, + "learning_rate": 3.570787381379841e-05, + "loss": 0.5449, + "step": 22340 + }, + { + "epoch": 0.86, + "learning_rate": 3.570723262374968e-05, + "loss": 0.4305, + "step": 22341 + }, + { + "epoch": 0.86, + "learning_rate": 3.570659143370095e-05, + "loss": 0.4574, + "step": 22342 + }, + { + "epoch": 0.86, + "learning_rate": 3.570595024365222e-05, + "loss": 0.5262, + "step": 22343 + }, + { + "epoch": 0.86, + "learning_rate": 3.5705309053603494e-05, + "loss": 0.5342, + "step": 22344 + }, + { + "epoch": 0.86, + "learning_rate": 3.570466786355476e-05, + "loss": 0.4658, + "step": 22345 + }, + { + "epoch": 0.86, + "learning_rate": 3.570402667350603e-05, + "loss": 0.515, + "step": 22346 + }, + { + "epoch": 0.86, + "learning_rate": 3.5703385483457295e-05, + "loss": 0.5732, + "step": 22347 + }, + { + "epoch": 0.86, + "learning_rate": 3.570274429340857e-05, + "loss": 0.4997, + "step": 22348 + }, + { + "epoch": 0.86, + "learning_rate": 3.570210310335983e-05, + "loss": 0.5437, + "step": 22349 + }, + { + "epoch": 0.86, + "learning_rate": 3.5701461913311104e-05, + "loss": 0.4055, + "step": 22350 + }, + { + "epoch": 0.86, + "learning_rate": 3.570082072326238e-05, + "loss": 0.5288, + "step": 22351 + }, + { + "epoch": 0.86, + "learning_rate": 3.570017953321365e-05, + "loss": 0.4501, + "step": 22352 + }, + { + "epoch": 0.86, + "learning_rate": 3.569953834316492e-05, + "loss": 0.4815, + "step": 22353 + }, + { + "epoch": 0.86, + "learning_rate": 3.5698897153116184e-05, + "loss": 0.4843, + "step": 22354 + }, + { + "epoch": 0.86, + "learning_rate": 3.5698255963067456e-05, + "loss": 0.4471, + "step": 22355 + }, + { + "epoch": 0.86, + "learning_rate": 3.569761477301872e-05, + "loss": 0.4644, + "step": 22356 + }, + { + "epoch": 0.86, + "learning_rate": 3.569697358296999e-05, + "loss": 0.4065, + "step": 22357 + }, + { + "epoch": 0.86, + "learning_rate": 3.5696332392921264e-05, + "loss": 0.5376, + "step": 22358 + }, + { + "epoch": 0.86, + "learning_rate": 3.5695691202872536e-05, + "loss": 0.553, + "step": 22359 + }, + { + "epoch": 0.86, + "learning_rate": 3.569505001282381e-05, + "loss": 0.6132, + "step": 22360 + }, + { + "epoch": 0.86, + "learning_rate": 3.569440882277507e-05, + "loss": 0.5977, + "step": 22361 + }, + { + "epoch": 0.86, + "learning_rate": 3.5693767632726344e-05, + "loss": 0.5571, + "step": 22362 + }, + { + "epoch": 0.86, + "learning_rate": 3.569312644267761e-05, + "loss": 0.4512, + "step": 22363 + }, + { + "epoch": 0.86, + "learning_rate": 3.569248525262888e-05, + "loss": 0.3981, + "step": 22364 + }, + { + "epoch": 0.86, + "learning_rate": 3.5691844062580146e-05, + "loss": 0.4588, + "step": 22365 + }, + { + "epoch": 0.86, + "learning_rate": 3.569120287253142e-05, + "loss": 0.5096, + "step": 22366 + }, + { + "epoch": 0.86, + "learning_rate": 3.569056168248269e-05, + "loss": 0.5335, + "step": 22367 + }, + { + "epoch": 0.86, + "learning_rate": 3.568992049243396e-05, + "loss": 0.4142, + "step": 22368 + }, + { + "epoch": 0.86, + "learning_rate": 3.568927930238523e-05, + "loss": 0.4592, + "step": 22369 + }, + { + "epoch": 0.86, + "learning_rate": 3.56886381123365e-05, + "loss": 0.4283, + "step": 22370 + }, + { + "epoch": 0.86, + "learning_rate": 3.568799692228777e-05, + "loss": 0.4091, + "step": 22371 + }, + { + "epoch": 0.86, + "learning_rate": 3.5687355732239034e-05, + "loss": 0.4696, + "step": 22372 + }, + { + "epoch": 0.86, + "learning_rate": 3.5686714542190306e-05, + "loss": 0.4348, + "step": 22373 + }, + { + "epoch": 0.86, + "learning_rate": 3.568607335214158e-05, + "loss": 0.5147, + "step": 22374 + }, + { + "epoch": 0.86, + "learning_rate": 3.568543216209285e-05, + "loss": 0.4955, + "step": 22375 + }, + { + "epoch": 0.86, + "learning_rate": 3.5684790972044114e-05, + "loss": 0.4618, + "step": 22376 + }, + { + "epoch": 0.86, + "learning_rate": 3.5684149781995386e-05, + "loss": 0.4004, + "step": 22377 + }, + { + "epoch": 0.86, + "learning_rate": 3.568350859194666e-05, + "loss": 0.6169, + "step": 22378 + }, + { + "epoch": 0.86, + "learning_rate": 3.568286740189792e-05, + "loss": 0.5243, + "step": 22379 + }, + { + "epoch": 0.86, + "learning_rate": 3.5682226211849194e-05, + "loss": 0.5987, + "step": 22380 + }, + { + "epoch": 0.86, + "learning_rate": 3.568158502180046e-05, + "loss": 0.5334, + "step": 22381 + }, + { + "epoch": 0.86, + "learning_rate": 3.568094383175173e-05, + "loss": 0.5641, + "step": 22382 + }, + { + "epoch": 0.86, + "learning_rate": 3.5680302641703e-05, + "loss": 0.5347, + "step": 22383 + }, + { + "epoch": 0.86, + "learning_rate": 3.5679661451654274e-05, + "loss": 0.4866, + "step": 22384 + }, + { + "epoch": 0.86, + "learning_rate": 3.567902026160554e-05, + "loss": 0.4083, + "step": 22385 + }, + { + "epoch": 0.86, + "learning_rate": 3.567837907155681e-05, + "loss": 0.5061, + "step": 22386 + }, + { + "epoch": 0.86, + "learning_rate": 3.567773788150808e-05, + "loss": 0.5544, + "step": 22387 + }, + { + "epoch": 0.86, + "learning_rate": 3.567709669145935e-05, + "loss": 0.5021, + "step": 22388 + }, + { + "epoch": 0.86, + "learning_rate": 3.567645550141062e-05, + "loss": 0.5547, + "step": 22389 + }, + { + "epoch": 0.86, + "learning_rate": 3.567581431136189e-05, + "loss": 0.3869, + "step": 22390 + }, + { + "epoch": 0.86, + "learning_rate": 3.567517312131316e-05, + "loss": 0.4354, + "step": 22391 + }, + { + "epoch": 0.86, + "learning_rate": 3.567453193126443e-05, + "loss": 0.5611, + "step": 22392 + }, + { + "epoch": 0.86, + "learning_rate": 3.56738907412157e-05, + "loss": 0.4485, + "step": 22393 + }, + { + "epoch": 0.86, + "learning_rate": 3.5673249551166965e-05, + "loss": 0.5403, + "step": 22394 + }, + { + "epoch": 0.86, + "learning_rate": 3.5672608361118236e-05, + "loss": 0.4981, + "step": 22395 + }, + { + "epoch": 0.86, + "learning_rate": 3.567196717106951e-05, + "loss": 0.4969, + "step": 22396 + }, + { + "epoch": 0.86, + "learning_rate": 3.567132598102077e-05, + "loss": 0.5193, + "step": 22397 + }, + { + "epoch": 0.86, + "learning_rate": 3.5670684790972045e-05, + "loss": 0.4255, + "step": 22398 + }, + { + "epoch": 0.86, + "learning_rate": 3.5670043600923316e-05, + "loss": 0.4292, + "step": 22399 + }, + { + "epoch": 0.86, + "learning_rate": 3.566940241087459e-05, + "loss": 0.5747, + "step": 22400 + }, + { + "epoch": 0.86, + "learning_rate": 3.566876122082585e-05, + "loss": 0.5559, + "step": 22401 + }, + { + "epoch": 0.86, + "learning_rate": 3.5668120030777125e-05, + "loss": 0.5756, + "step": 22402 + }, + { + "epoch": 0.86, + "learning_rate": 3.566747884072839e-05, + "loss": 0.4499, + "step": 22403 + }, + { + "epoch": 0.86, + "learning_rate": 3.566683765067966e-05, + "loss": 0.4843, + "step": 22404 + }, + { + "epoch": 0.86, + "learning_rate": 3.566619646063093e-05, + "loss": 0.5353, + "step": 22405 + }, + { + "epoch": 0.86, + "learning_rate": 3.5665555270582205e-05, + "loss": 0.4956, + "step": 22406 + }, + { + "epoch": 0.86, + "learning_rate": 3.5664914080533477e-05, + "loss": 0.6127, + "step": 22407 + }, + { + "epoch": 0.86, + "learning_rate": 3.566427289048474e-05, + "loss": 0.4415, + "step": 22408 + }, + { + "epoch": 0.86, + "learning_rate": 3.566363170043601e-05, + "loss": 0.5437, + "step": 22409 + }, + { + "epoch": 0.86, + "learning_rate": 3.566299051038728e-05, + "loss": 0.5803, + "step": 22410 + }, + { + "epoch": 0.86, + "learning_rate": 3.566234932033855e-05, + "loss": 0.4959, + "step": 22411 + }, + { + "epoch": 0.86, + "learning_rate": 3.5661708130289815e-05, + "loss": 0.5238, + "step": 22412 + }, + { + "epoch": 0.86, + "learning_rate": 3.5661066940241086e-05, + "loss": 0.5727, + "step": 22413 + }, + { + "epoch": 0.86, + "learning_rate": 3.566042575019236e-05, + "loss": 0.5943, + "step": 22414 + }, + { + "epoch": 0.86, + "learning_rate": 3.565978456014363e-05, + "loss": 0.5223, + "step": 22415 + }, + { + "epoch": 0.86, + "learning_rate": 3.56591433700949e-05, + "loss": 0.573, + "step": 22416 + }, + { + "epoch": 0.86, + "learning_rate": 3.5658502180046167e-05, + "loss": 0.5688, + "step": 22417 + }, + { + "epoch": 0.86, + "learning_rate": 3.565786098999744e-05, + "loss": 0.3956, + "step": 22418 + }, + { + "epoch": 0.86, + "learning_rate": 3.56572197999487e-05, + "loss": 0.5439, + "step": 22419 + }, + { + "epoch": 0.86, + "learning_rate": 3.5656578609899975e-05, + "loss": 0.5163, + "step": 22420 + }, + { + "epoch": 0.86, + "learning_rate": 3.565593741985124e-05, + "loss": 0.5248, + "step": 22421 + }, + { + "epoch": 0.86, + "learning_rate": 3.565529622980252e-05, + "loss": 0.5007, + "step": 22422 + }, + { + "epoch": 0.86, + "learning_rate": 3.565465503975379e-05, + "loss": 0.5429, + "step": 22423 + }, + { + "epoch": 0.86, + "learning_rate": 3.5654013849705055e-05, + "loss": 0.419, + "step": 22424 + }, + { + "epoch": 0.86, + "learning_rate": 3.565337265965633e-05, + "loss": 0.5556, + "step": 22425 + }, + { + "epoch": 0.86, + "learning_rate": 3.565273146960759e-05, + "loss": 0.5027, + "step": 22426 + }, + { + "epoch": 0.86, + "learning_rate": 3.5652090279558863e-05, + "loss": 0.4605, + "step": 22427 + }, + { + "epoch": 0.86, + "learning_rate": 3.565144908951013e-05, + "loss": 0.5569, + "step": 22428 + }, + { + "epoch": 0.86, + "learning_rate": 3.56508078994614e-05, + "loss": 0.5469, + "step": 22429 + }, + { + "epoch": 0.86, + "learning_rate": 3.565016670941267e-05, + "loss": 0.4446, + "step": 22430 + }, + { + "epoch": 0.86, + "learning_rate": 3.5649525519363944e-05, + "loss": 0.454, + "step": 22431 + }, + { + "epoch": 0.86, + "learning_rate": 3.5648884329315215e-05, + "loss": 0.5011, + "step": 22432 + }, + { + "epoch": 0.86, + "learning_rate": 3.564824313926648e-05, + "loss": 0.3732, + "step": 22433 + }, + { + "epoch": 0.86, + "learning_rate": 3.564760194921775e-05, + "loss": 0.4953, + "step": 22434 + }, + { + "epoch": 0.86, + "learning_rate": 3.564696075916902e-05, + "loss": 0.569, + "step": 22435 + }, + { + "epoch": 0.86, + "learning_rate": 3.564631956912029e-05, + "loss": 0.4707, + "step": 22436 + }, + { + "epoch": 0.86, + "learning_rate": 3.5645678379071553e-05, + "loss": 0.4172, + "step": 22437 + }, + { + "epoch": 0.86, + "learning_rate": 3.5645037189022825e-05, + "loss": 0.4447, + "step": 22438 + }, + { + "epoch": 0.86, + "learning_rate": 3.56443959989741e-05, + "loss": 0.5358, + "step": 22439 + }, + { + "epoch": 0.86, + "learning_rate": 3.564375480892537e-05, + "loss": 0.5476, + "step": 22440 + }, + { + "epoch": 0.86, + "learning_rate": 3.564311361887664e-05, + "loss": 0.4838, + "step": 22441 + }, + { + "epoch": 0.86, + "learning_rate": 3.5642472428827905e-05, + "loss": 0.5355, + "step": 22442 + }, + { + "epoch": 0.86, + "learning_rate": 3.564183123877918e-05, + "loss": 0.4773, + "step": 22443 + }, + { + "epoch": 0.86, + "learning_rate": 3.564119004873044e-05, + "loss": 0.4756, + "step": 22444 + }, + { + "epoch": 0.86, + "learning_rate": 3.5640548858681714e-05, + "loss": 0.4167, + "step": 22445 + }, + { + "epoch": 0.86, + "learning_rate": 3.5639907668632985e-05, + "loss": 0.4044, + "step": 22446 + }, + { + "epoch": 0.86, + "learning_rate": 3.563926647858426e-05, + "loss": 0.5399, + "step": 22447 + }, + { + "epoch": 0.86, + "learning_rate": 3.563862528853552e-05, + "loss": 0.4558, + "step": 22448 + }, + { + "epoch": 0.86, + "learning_rate": 3.5637984098486794e-05, + "loss": 0.3816, + "step": 22449 + }, + { + "epoch": 0.86, + "learning_rate": 3.5637342908438065e-05, + "loss": 0.4762, + "step": 22450 + }, + { + "epoch": 0.86, + "learning_rate": 3.563670171838933e-05, + "loss": 0.4138, + "step": 22451 + }, + { + "epoch": 0.86, + "learning_rate": 3.56360605283406e-05, + "loss": 0.4288, + "step": 22452 + }, + { + "epoch": 0.86, + "learning_rate": 3.563541933829187e-05, + "loss": 0.5711, + "step": 22453 + }, + { + "epoch": 0.86, + "learning_rate": 3.563477814824314e-05, + "loss": 0.4356, + "step": 22454 + }, + { + "epoch": 0.86, + "learning_rate": 3.563413695819441e-05, + "loss": 0.5089, + "step": 22455 + }, + { + "epoch": 0.86, + "learning_rate": 3.563349576814568e-05, + "loss": 0.6022, + "step": 22456 + }, + { + "epoch": 0.86, + "learning_rate": 3.563285457809695e-05, + "loss": 0.4765, + "step": 22457 + }, + { + "epoch": 0.86, + "learning_rate": 3.563221338804822e-05, + "loss": 0.52, + "step": 22458 + }, + { + "epoch": 0.86, + "learning_rate": 3.563157219799949e-05, + "loss": 0.3799, + "step": 22459 + }, + { + "epoch": 0.86, + "learning_rate": 3.5630931007950756e-05, + "loss": 0.5088, + "step": 22460 + }, + { + "epoch": 0.86, + "learning_rate": 3.563028981790203e-05, + "loss": 0.4963, + "step": 22461 + }, + { + "epoch": 0.86, + "learning_rate": 3.56296486278533e-05, + "loss": 0.4821, + "step": 22462 + }, + { + "epoch": 0.86, + "learning_rate": 3.562900743780457e-05, + "loss": 0.4803, + "step": 22463 + }, + { + "epoch": 0.86, + "learning_rate": 3.5628366247755836e-05, + "loss": 0.4795, + "step": 22464 + }, + { + "epoch": 0.86, + "learning_rate": 3.562772505770711e-05, + "loss": 0.3586, + "step": 22465 + }, + { + "epoch": 0.86, + "learning_rate": 3.562708386765837e-05, + "loss": 0.4551, + "step": 22466 + }, + { + "epoch": 0.86, + "learning_rate": 3.5626442677609644e-05, + "loss": 0.4789, + "step": 22467 + }, + { + "epoch": 0.86, + "learning_rate": 3.5625801487560916e-05, + "loss": 0.4981, + "step": 22468 + }, + { + "epoch": 0.86, + "learning_rate": 3.562516029751218e-05, + "loss": 0.6034, + "step": 22469 + }, + { + "epoch": 0.86, + "learning_rate": 3.562451910746345e-05, + "loss": 0.4452, + "step": 22470 + }, + { + "epoch": 0.86, + "learning_rate": 3.5623877917414724e-05, + "loss": 0.3788, + "step": 22471 + }, + { + "epoch": 0.86, + "learning_rate": 3.5623236727365996e-05, + "loss": 0.4142, + "step": 22472 + }, + { + "epoch": 0.86, + "learning_rate": 3.562259553731726e-05, + "loss": 0.4484, + "step": 22473 + }, + { + "epoch": 0.86, + "learning_rate": 3.562195434726853e-05, + "loss": 0.5498, + "step": 22474 + }, + { + "epoch": 0.86, + "learning_rate": 3.56213131572198e-05, + "loss": 0.561, + "step": 22475 + }, + { + "epoch": 0.86, + "learning_rate": 3.562067196717107e-05, + "loss": 0.4846, + "step": 22476 + }, + { + "epoch": 0.86, + "learning_rate": 3.562003077712234e-05, + "loss": 0.4841, + "step": 22477 + }, + { + "epoch": 0.86, + "learning_rate": 3.561938958707361e-05, + "loss": 0.6672, + "step": 22478 + }, + { + "epoch": 0.86, + "learning_rate": 3.5618748397024884e-05, + "loss": 0.618, + "step": 22479 + }, + { + "epoch": 0.86, + "learning_rate": 3.561810720697615e-05, + "loss": 0.5755, + "step": 22480 + }, + { + "epoch": 0.86, + "learning_rate": 3.561746601692742e-05, + "loss": 0.5264, + "step": 22481 + }, + { + "epoch": 0.86, + "learning_rate": 3.5616824826878686e-05, + "loss": 0.514, + "step": 22482 + }, + { + "epoch": 0.86, + "learning_rate": 3.561618363682996e-05, + "loss": 0.4614, + "step": 22483 + }, + { + "epoch": 0.86, + "learning_rate": 3.561554244678123e-05, + "loss": 0.5166, + "step": 22484 + }, + { + "epoch": 0.86, + "learning_rate": 3.5614901256732494e-05, + "loss": 0.6423, + "step": 22485 + }, + { + "epoch": 0.86, + "learning_rate": 3.5614260066683766e-05, + "loss": 0.5874, + "step": 22486 + }, + { + "epoch": 0.86, + "learning_rate": 3.561361887663504e-05, + "loss": 0.5657, + "step": 22487 + }, + { + "epoch": 0.86, + "learning_rate": 3.561297768658631e-05, + "loss": 0.5525, + "step": 22488 + }, + { + "epoch": 0.86, + "learning_rate": 3.5612336496537574e-05, + "loss": 0.5889, + "step": 22489 + }, + { + "epoch": 0.86, + "learning_rate": 3.5611695306488846e-05, + "loss": 0.398, + "step": 22490 + }, + { + "epoch": 0.86, + "learning_rate": 3.561105411644011e-05, + "loss": 0.4401, + "step": 22491 + }, + { + "epoch": 0.86, + "learning_rate": 3.561041292639138e-05, + "loss": 0.5044, + "step": 22492 + }, + { + "epoch": 0.86, + "learning_rate": 3.5609771736342654e-05, + "loss": 0.456, + "step": 22493 + }, + { + "epoch": 0.86, + "learning_rate": 3.5609130546293926e-05, + "loss": 0.5886, + "step": 22494 + }, + { + "epoch": 0.86, + "learning_rate": 3.56084893562452e-05, + "loss": 0.529, + "step": 22495 + }, + { + "epoch": 0.86, + "learning_rate": 3.560784816619646e-05, + "loss": 0.54, + "step": 22496 + }, + { + "epoch": 0.86, + "learning_rate": 3.5607206976147734e-05, + "loss": 0.4625, + "step": 22497 + }, + { + "epoch": 0.86, + "learning_rate": 3.5606565786099e-05, + "loss": 0.4707, + "step": 22498 + }, + { + "epoch": 0.87, + "learning_rate": 3.560592459605027e-05, + "loss": 0.52, + "step": 22499 + }, + { + "epoch": 0.87, + "learning_rate": 3.5605283406001536e-05, + "loss": 0.4361, + "step": 22500 + }, + { + "epoch": 0.87, + "learning_rate": 3.560464221595281e-05, + "loss": 0.4485, + "step": 22501 + }, + { + "epoch": 0.87, + "learning_rate": 3.560400102590408e-05, + "loss": 0.5027, + "step": 22502 + }, + { + "epoch": 0.87, + "learning_rate": 3.560335983585535e-05, + "loss": 0.4875, + "step": 22503 + }, + { + "epoch": 0.87, + "learning_rate": 3.560271864580662e-05, + "loss": 0.4695, + "step": 22504 + }, + { + "epoch": 0.87, + "learning_rate": 3.560207745575789e-05, + "loss": 0.525, + "step": 22505 + }, + { + "epoch": 0.87, + "learning_rate": 3.560143626570916e-05, + "loss": 0.4542, + "step": 22506 + }, + { + "epoch": 0.87, + "learning_rate": 3.5600795075660425e-05, + "loss": 0.5636, + "step": 22507 + }, + { + "epoch": 0.87, + "learning_rate": 3.5600153885611696e-05, + "loss": 0.5358, + "step": 22508 + }, + { + "epoch": 0.87, + "learning_rate": 3.559951269556296e-05, + "loss": 0.4813, + "step": 22509 + }, + { + "epoch": 0.87, + "learning_rate": 3.559887150551424e-05, + "loss": 0.5991, + "step": 22510 + }, + { + "epoch": 0.87, + "learning_rate": 3.559823031546551e-05, + "loss": 0.5107, + "step": 22511 + }, + { + "epoch": 0.87, + "learning_rate": 3.5597589125416776e-05, + "loss": 0.6058, + "step": 22512 + }, + { + "epoch": 0.87, + "learning_rate": 3.559694793536805e-05, + "loss": 0.4818, + "step": 22513 + }, + { + "epoch": 0.87, + "learning_rate": 3.559630674531931e-05, + "loss": 0.5897, + "step": 22514 + }, + { + "epoch": 0.87, + "learning_rate": 3.5595665555270585e-05, + "loss": 0.4353, + "step": 22515 + }, + { + "epoch": 0.87, + "learning_rate": 3.559502436522185e-05, + "loss": 0.5195, + "step": 22516 + }, + { + "epoch": 0.87, + "learning_rate": 3.559438317517312e-05, + "loss": 0.4428, + "step": 22517 + }, + { + "epoch": 0.87, + "learning_rate": 3.559374198512439e-05, + "loss": 0.4834, + "step": 22518 + }, + { + "epoch": 0.87, + "learning_rate": 3.5593100795075665e-05, + "loss": 0.4672, + "step": 22519 + }, + { + "epoch": 0.87, + "learning_rate": 3.5592459605026937e-05, + "loss": 0.3944, + "step": 22520 + }, + { + "epoch": 0.87, + "learning_rate": 3.55918184149782e-05, + "loss": 0.5041, + "step": 22521 + }, + { + "epoch": 0.87, + "learning_rate": 3.559117722492947e-05, + "loss": 0.4832, + "step": 22522 + }, + { + "epoch": 0.87, + "learning_rate": 3.559053603488074e-05, + "loss": 0.4463, + "step": 22523 + }, + { + "epoch": 0.87, + "learning_rate": 3.558989484483201e-05, + "loss": 0.5737, + "step": 22524 + }, + { + "epoch": 0.87, + "learning_rate": 3.5589253654783275e-05, + "loss": 0.5176, + "step": 22525 + }, + { + "epoch": 0.87, + "learning_rate": 3.5588612464734547e-05, + "loss": 0.4906, + "step": 22526 + }, + { + "epoch": 0.87, + "learning_rate": 3.558797127468582e-05, + "loss": 0.4432, + "step": 22527 + }, + { + "epoch": 0.87, + "learning_rate": 3.558733008463709e-05, + "loss": 0.5137, + "step": 22528 + }, + { + "epoch": 0.87, + "learning_rate": 3.558668889458836e-05, + "loss": 0.4318, + "step": 22529 + }, + { + "epoch": 0.87, + "learning_rate": 3.5586047704539627e-05, + "loss": 0.5598, + "step": 22530 + }, + { + "epoch": 0.87, + "learning_rate": 3.55854065144909e-05, + "loss": 0.5172, + "step": 22531 + }, + { + "epoch": 0.87, + "learning_rate": 3.558476532444216e-05, + "loss": 0.5705, + "step": 22532 + }, + { + "epoch": 0.87, + "learning_rate": 3.5584124134393435e-05, + "loss": 0.4672, + "step": 22533 + }, + { + "epoch": 0.87, + "learning_rate": 3.558348294434471e-05, + "loss": 0.6287, + "step": 22534 + }, + { + "epoch": 0.87, + "learning_rate": 3.558284175429598e-05, + "loss": 0.5016, + "step": 22535 + }, + { + "epoch": 0.87, + "learning_rate": 3.558220056424724e-05, + "loss": 0.5992, + "step": 22536 + }, + { + "epoch": 0.87, + "learning_rate": 3.5581559374198515e-05, + "loss": 0.4487, + "step": 22537 + }, + { + "epoch": 0.87, + "learning_rate": 3.558091818414979e-05, + "loss": 0.5302, + "step": 22538 + }, + { + "epoch": 0.87, + "learning_rate": 3.558027699410105e-05, + "loss": 0.5539, + "step": 22539 + }, + { + "epoch": 0.87, + "learning_rate": 3.5579635804052323e-05, + "loss": 0.466, + "step": 22540 + }, + { + "epoch": 0.87, + "learning_rate": 3.557899461400359e-05, + "loss": 0.5075, + "step": 22541 + }, + { + "epoch": 0.87, + "learning_rate": 3.557835342395486e-05, + "loss": 0.5456, + "step": 22542 + }, + { + "epoch": 0.87, + "learning_rate": 3.557771223390613e-05, + "loss": 0.5984, + "step": 22543 + }, + { + "epoch": 0.87, + "learning_rate": 3.5577071043857404e-05, + "loss": 0.4563, + "step": 22544 + }, + { + "epoch": 0.87, + "learning_rate": 3.557642985380867e-05, + "loss": 0.583, + "step": 22545 + }, + { + "epoch": 0.87, + "learning_rate": 3.557578866375994e-05, + "loss": 0.4245, + "step": 22546 + }, + { + "epoch": 0.87, + "learning_rate": 3.557514747371121e-05, + "loss": 0.445, + "step": 22547 + }, + { + "epoch": 0.87, + "learning_rate": 3.557450628366248e-05, + "loss": 0.5707, + "step": 22548 + }, + { + "epoch": 0.87, + "learning_rate": 3.557386509361375e-05, + "loss": 0.4423, + "step": 22549 + }, + { + "epoch": 0.87, + "learning_rate": 3.557322390356502e-05, + "loss": 0.4211, + "step": 22550 + }, + { + "epoch": 0.87, + "learning_rate": 3.557258271351629e-05, + "loss": 0.5703, + "step": 22551 + }, + { + "epoch": 0.87, + "learning_rate": 3.557194152346756e-05, + "loss": 0.477, + "step": 22552 + }, + { + "epoch": 0.87, + "learning_rate": 3.557130033341883e-05, + "loss": 0.3952, + "step": 22553 + }, + { + "epoch": 0.87, + "learning_rate": 3.5570659143370094e-05, + "loss": 0.5696, + "step": 22554 + }, + { + "epoch": 0.87, + "learning_rate": 3.5570017953321365e-05, + "loss": 0.4597, + "step": 22555 + }, + { + "epoch": 0.87, + "learning_rate": 3.556937676327264e-05, + "loss": 0.4657, + "step": 22556 + }, + { + "epoch": 0.87, + "learning_rate": 3.55687355732239e-05, + "loss": 0.5616, + "step": 22557 + }, + { + "epoch": 0.87, + "learning_rate": 3.5568094383175174e-05, + "loss": 0.5336, + "step": 22558 + }, + { + "epoch": 0.87, + "learning_rate": 3.5567453193126445e-05, + "loss": 0.5264, + "step": 22559 + }, + { + "epoch": 0.87, + "learning_rate": 3.556681200307772e-05, + "loss": 0.3551, + "step": 22560 + }, + { + "epoch": 0.87, + "learning_rate": 3.556617081302898e-05, + "loss": 0.4623, + "step": 22561 + }, + { + "epoch": 0.87, + "learning_rate": 3.5565529622980254e-05, + "loss": 0.5505, + "step": 22562 + }, + { + "epoch": 0.87, + "learning_rate": 3.556488843293152e-05, + "loss": 0.6317, + "step": 22563 + }, + { + "epoch": 0.87, + "learning_rate": 3.556424724288279e-05, + "loss": 0.4395, + "step": 22564 + }, + { + "epoch": 0.87, + "learning_rate": 3.556360605283406e-05, + "loss": 0.4292, + "step": 22565 + }, + { + "epoch": 0.87, + "learning_rate": 3.5562964862785334e-05, + "loss": 0.4955, + "step": 22566 + }, + { + "epoch": 0.87, + "learning_rate": 3.5562323672736606e-05, + "loss": 0.4474, + "step": 22567 + }, + { + "epoch": 0.87, + "learning_rate": 3.556168248268787e-05, + "loss": 0.5456, + "step": 22568 + }, + { + "epoch": 0.87, + "learning_rate": 3.556104129263914e-05, + "loss": 0.551, + "step": 22569 + }, + { + "epoch": 0.87, + "learning_rate": 3.556040010259041e-05, + "loss": 0.49, + "step": 22570 + }, + { + "epoch": 0.87, + "learning_rate": 3.555975891254168e-05, + "loss": 0.4487, + "step": 22571 + }, + { + "epoch": 0.87, + "learning_rate": 3.5559117722492944e-05, + "loss": 0.403, + "step": 22572 + }, + { + "epoch": 0.87, + "learning_rate": 3.5558476532444216e-05, + "loss": 0.4603, + "step": 22573 + }, + { + "epoch": 0.87, + "learning_rate": 3.555783534239549e-05, + "loss": 0.5975, + "step": 22574 + }, + { + "epoch": 0.87, + "learning_rate": 3.555719415234676e-05, + "loss": 0.54, + "step": 22575 + }, + { + "epoch": 0.87, + "learning_rate": 3.555655296229803e-05, + "loss": 0.4709, + "step": 22576 + }, + { + "epoch": 0.87, + "learning_rate": 3.5555911772249296e-05, + "loss": 0.4219, + "step": 22577 + }, + { + "epoch": 0.87, + "learning_rate": 3.555527058220057e-05, + "loss": 0.3645, + "step": 22578 + }, + { + "epoch": 0.87, + "learning_rate": 3.555462939215183e-05, + "loss": 0.4, + "step": 22579 + }, + { + "epoch": 0.87, + "learning_rate": 3.5553988202103104e-05, + "loss": 0.4622, + "step": 22580 + }, + { + "epoch": 0.87, + "learning_rate": 3.555334701205437e-05, + "loss": 0.4914, + "step": 22581 + }, + { + "epoch": 0.87, + "learning_rate": 3.555270582200565e-05, + "loss": 0.5229, + "step": 22582 + }, + { + "epoch": 0.87, + "learning_rate": 3.555206463195692e-05, + "loss": 0.5796, + "step": 22583 + }, + { + "epoch": 0.87, + "learning_rate": 3.5551423441908184e-05, + "loss": 0.4932, + "step": 22584 + }, + { + "epoch": 0.87, + "learning_rate": 3.5550782251859456e-05, + "loss": 0.4446, + "step": 22585 + }, + { + "epoch": 0.87, + "learning_rate": 3.555014106181072e-05, + "loss": 0.605, + "step": 22586 + }, + { + "epoch": 0.87, + "learning_rate": 3.554949987176199e-05, + "loss": 0.6238, + "step": 22587 + }, + { + "epoch": 0.87, + "learning_rate": 3.554885868171326e-05, + "loss": 0.4694, + "step": 22588 + }, + { + "epoch": 0.87, + "learning_rate": 3.554821749166453e-05, + "loss": 0.4687, + "step": 22589 + }, + { + "epoch": 0.87, + "learning_rate": 3.55475763016158e-05, + "loss": 0.4381, + "step": 22590 + }, + { + "epoch": 0.87, + "learning_rate": 3.554693511156707e-05, + "loss": 0.5064, + "step": 22591 + }, + { + "epoch": 0.87, + "learning_rate": 3.5546293921518344e-05, + "loss": 0.5106, + "step": 22592 + }, + { + "epoch": 0.87, + "learning_rate": 3.554565273146961e-05, + "loss": 0.5056, + "step": 22593 + }, + { + "epoch": 0.87, + "learning_rate": 3.554501154142088e-05, + "loss": 0.5325, + "step": 22594 + }, + { + "epoch": 0.87, + "learning_rate": 3.5544370351372146e-05, + "loss": 0.5329, + "step": 22595 + }, + { + "epoch": 0.87, + "learning_rate": 3.554372916132342e-05, + "loss": 0.4522, + "step": 22596 + }, + { + "epoch": 0.87, + "learning_rate": 3.554308797127468e-05, + "loss": 0.5248, + "step": 22597 + }, + { + "epoch": 0.87, + "learning_rate": 3.5542446781225954e-05, + "loss": 0.6309, + "step": 22598 + }, + { + "epoch": 0.87, + "learning_rate": 3.5541805591177226e-05, + "loss": 0.4522, + "step": 22599 + }, + { + "epoch": 0.87, + "learning_rate": 3.55411644011285e-05, + "loss": 0.473, + "step": 22600 + }, + { + "epoch": 0.87, + "learning_rate": 3.554052321107977e-05, + "loss": 0.5685, + "step": 22601 + }, + { + "epoch": 0.87, + "learning_rate": 3.5539882021031034e-05, + "loss": 0.4921, + "step": 22602 + }, + { + "epoch": 0.87, + "learning_rate": 3.5539240830982306e-05, + "loss": 0.5068, + "step": 22603 + }, + { + "epoch": 0.87, + "learning_rate": 3.553859964093357e-05, + "loss": 0.4855, + "step": 22604 + }, + { + "epoch": 0.87, + "learning_rate": 3.553795845088484e-05, + "loss": 0.5123, + "step": 22605 + }, + { + "epoch": 0.87, + "learning_rate": 3.5537317260836114e-05, + "loss": 0.5426, + "step": 22606 + }, + { + "epoch": 0.87, + "learning_rate": 3.5536676070787386e-05, + "loss": 0.5284, + "step": 22607 + }, + { + "epoch": 0.87, + "learning_rate": 3.553603488073865e-05, + "loss": 0.5253, + "step": 22608 + }, + { + "epoch": 0.87, + "learning_rate": 3.553539369068992e-05, + "loss": 0.4336, + "step": 22609 + }, + { + "epoch": 0.87, + "learning_rate": 3.5534752500641195e-05, + "loss": 0.5766, + "step": 22610 + }, + { + "epoch": 0.87, + "learning_rate": 3.553411131059246e-05, + "loss": 0.5108, + "step": 22611 + }, + { + "epoch": 0.87, + "learning_rate": 3.553347012054373e-05, + "loss": 0.4282, + "step": 22612 + }, + { + "epoch": 0.87, + "learning_rate": 3.5532828930494996e-05, + "loss": 0.4997, + "step": 22613 + }, + { + "epoch": 0.87, + "learning_rate": 3.553218774044627e-05, + "loss": 0.4009, + "step": 22614 + }, + { + "epoch": 0.87, + "learning_rate": 3.553154655039754e-05, + "loss": 0.4644, + "step": 22615 + }, + { + "epoch": 0.87, + "learning_rate": 3.553090536034881e-05, + "loss": 0.4817, + "step": 22616 + }, + { + "epoch": 0.87, + "learning_rate": 3.5530264170300076e-05, + "loss": 0.369, + "step": 22617 + }, + { + "epoch": 0.87, + "learning_rate": 3.552962298025135e-05, + "loss": 0.4415, + "step": 22618 + }, + { + "epoch": 0.87, + "learning_rate": 3.552898179020262e-05, + "loss": 0.4192, + "step": 22619 + }, + { + "epoch": 0.87, + "learning_rate": 3.5528340600153885e-05, + "loss": 0.5629, + "step": 22620 + }, + { + "epoch": 0.87, + "learning_rate": 3.5527699410105156e-05, + "loss": 0.5415, + "step": 22621 + }, + { + "epoch": 0.87, + "learning_rate": 3.552705822005643e-05, + "loss": 0.6457, + "step": 22622 + }, + { + "epoch": 0.87, + "learning_rate": 3.55264170300077e-05, + "loss": 0.5805, + "step": 22623 + }, + { + "epoch": 0.87, + "learning_rate": 3.5525775839958965e-05, + "loss": 0.4731, + "step": 22624 + }, + { + "epoch": 0.87, + "learning_rate": 3.5525134649910236e-05, + "loss": 0.4628, + "step": 22625 + }, + { + "epoch": 0.87, + "learning_rate": 3.55244934598615e-05, + "loss": 0.4183, + "step": 22626 + }, + { + "epoch": 0.87, + "learning_rate": 3.552385226981277e-05, + "loss": 0.5163, + "step": 22627 + }, + { + "epoch": 0.87, + "learning_rate": 3.5523211079764045e-05, + "loss": 0.5078, + "step": 22628 + }, + { + "epoch": 0.87, + "learning_rate": 3.552256988971531e-05, + "loss": 0.5285, + "step": 22629 + }, + { + "epoch": 0.87, + "learning_rate": 3.552192869966658e-05, + "loss": 0.4922, + "step": 22630 + }, + { + "epoch": 0.87, + "learning_rate": 3.552128750961785e-05, + "loss": 0.4911, + "step": 22631 + }, + { + "epoch": 0.87, + "learning_rate": 3.5520646319569125e-05, + "loss": 0.475, + "step": 22632 + }, + { + "epoch": 0.87, + "learning_rate": 3.552000512952039e-05, + "loss": 0.4973, + "step": 22633 + }, + { + "epoch": 0.87, + "learning_rate": 3.551936393947166e-05, + "loss": 0.5708, + "step": 22634 + }, + { + "epoch": 0.87, + "learning_rate": 3.5518722749422926e-05, + "loss": 0.466, + "step": 22635 + }, + { + "epoch": 0.87, + "learning_rate": 3.55180815593742e-05, + "loss": 0.5177, + "step": 22636 + }, + { + "epoch": 0.87, + "learning_rate": 3.551744036932547e-05, + "loss": 0.608, + "step": 22637 + }, + { + "epoch": 0.87, + "learning_rate": 3.551679917927674e-05, + "loss": 0.5218, + "step": 22638 + }, + { + "epoch": 0.87, + "learning_rate": 3.551615798922801e-05, + "loss": 0.4421, + "step": 22639 + }, + { + "epoch": 0.87, + "learning_rate": 3.551551679917928e-05, + "loss": 0.5688, + "step": 22640 + }, + { + "epoch": 0.87, + "learning_rate": 3.551487560913055e-05, + "loss": 0.3702, + "step": 22641 + }, + { + "epoch": 0.87, + "learning_rate": 3.5514234419081815e-05, + "loss": 0.5053, + "step": 22642 + }, + { + "epoch": 0.87, + "learning_rate": 3.551359322903309e-05, + "loss": 0.4492, + "step": 22643 + }, + { + "epoch": 0.87, + "learning_rate": 3.551295203898435e-05, + "loss": 0.4659, + "step": 22644 + }, + { + "epoch": 0.87, + "learning_rate": 3.551231084893562e-05, + "loss": 0.4423, + "step": 22645 + }, + { + "epoch": 0.87, + "learning_rate": 3.5511669658886895e-05, + "loss": 0.513, + "step": 22646 + }, + { + "epoch": 0.87, + "learning_rate": 3.551102846883817e-05, + "loss": 0.482, + "step": 22647 + }, + { + "epoch": 0.87, + "learning_rate": 3.551038727878944e-05, + "loss": 0.4558, + "step": 22648 + }, + { + "epoch": 0.87, + "learning_rate": 3.5509746088740703e-05, + "loss": 0.5047, + "step": 22649 + }, + { + "epoch": 0.87, + "learning_rate": 3.5509104898691975e-05, + "loss": 0.5601, + "step": 22650 + }, + { + "epoch": 0.87, + "learning_rate": 3.550846370864324e-05, + "loss": 0.4962, + "step": 22651 + }, + { + "epoch": 0.87, + "learning_rate": 3.550782251859451e-05, + "loss": 0.4621, + "step": 22652 + }, + { + "epoch": 0.87, + "learning_rate": 3.550718132854578e-05, + "loss": 0.5403, + "step": 22653 + }, + { + "epoch": 0.87, + "learning_rate": 3.5506540138497055e-05, + "loss": 0.4761, + "step": 22654 + }, + { + "epoch": 0.87, + "learning_rate": 3.550589894844833e-05, + "loss": 0.6824, + "step": 22655 + }, + { + "epoch": 0.87, + "learning_rate": 3.550525775839959e-05, + "loss": 0.4126, + "step": 22656 + }, + { + "epoch": 0.87, + "learning_rate": 3.5504616568350864e-05, + "loss": 0.5043, + "step": 22657 + }, + { + "epoch": 0.87, + "learning_rate": 3.550397537830213e-05, + "loss": 0.493, + "step": 22658 + }, + { + "epoch": 0.87, + "learning_rate": 3.55033341882534e-05, + "loss": 0.3937, + "step": 22659 + }, + { + "epoch": 0.87, + "learning_rate": 3.5502692998204665e-05, + "loss": 0.496, + "step": 22660 + }, + { + "epoch": 0.87, + "learning_rate": 3.550205180815594e-05, + "loss": 0.5164, + "step": 22661 + }, + { + "epoch": 0.87, + "learning_rate": 3.550141061810721e-05, + "loss": 0.4862, + "step": 22662 + }, + { + "epoch": 0.87, + "learning_rate": 3.550076942805848e-05, + "loss": 0.6011, + "step": 22663 + }, + { + "epoch": 0.87, + "learning_rate": 3.550012823800975e-05, + "loss": 0.4505, + "step": 22664 + }, + { + "epoch": 0.87, + "learning_rate": 3.549948704796102e-05, + "loss": 0.3958, + "step": 22665 + }, + { + "epoch": 0.87, + "learning_rate": 3.549884585791229e-05, + "loss": 0.4728, + "step": 22666 + }, + { + "epoch": 0.87, + "learning_rate": 3.5498204667863554e-05, + "loss": 0.4986, + "step": 22667 + }, + { + "epoch": 0.87, + "learning_rate": 3.5497563477814825e-05, + "loss": 0.3943, + "step": 22668 + }, + { + "epoch": 0.87, + "learning_rate": 3.549692228776609e-05, + "loss": 0.4981, + "step": 22669 + }, + { + "epoch": 0.87, + "learning_rate": 3.549628109771737e-05, + "loss": 0.5602, + "step": 22670 + }, + { + "epoch": 0.87, + "learning_rate": 3.549563990766864e-05, + "loss": 0.4077, + "step": 22671 + }, + { + "epoch": 0.87, + "learning_rate": 3.5494998717619905e-05, + "loss": 0.4105, + "step": 22672 + }, + { + "epoch": 0.87, + "learning_rate": 3.549435752757118e-05, + "loss": 0.4735, + "step": 22673 + }, + { + "epoch": 0.87, + "learning_rate": 3.549371633752244e-05, + "loss": 0.4755, + "step": 22674 + }, + { + "epoch": 0.87, + "learning_rate": 3.5493075147473714e-05, + "loss": 0.4839, + "step": 22675 + }, + { + "epoch": 0.87, + "learning_rate": 3.549243395742498e-05, + "loss": 0.5638, + "step": 22676 + }, + { + "epoch": 0.87, + "learning_rate": 3.549179276737625e-05, + "loss": 0.4836, + "step": 22677 + }, + { + "epoch": 0.87, + "learning_rate": 3.549115157732752e-05, + "loss": 0.4073, + "step": 22678 + }, + { + "epoch": 0.87, + "learning_rate": 3.5490510387278794e-05, + "loss": 0.5021, + "step": 22679 + }, + { + "epoch": 0.87, + "learning_rate": 3.5489869197230066e-05, + "loss": 0.4409, + "step": 22680 + }, + { + "epoch": 0.87, + "learning_rate": 3.548922800718133e-05, + "loss": 0.553, + "step": 22681 + }, + { + "epoch": 0.87, + "learning_rate": 3.54885868171326e-05, + "loss": 0.6161, + "step": 22682 + }, + { + "epoch": 0.87, + "learning_rate": 3.548794562708387e-05, + "loss": 0.4734, + "step": 22683 + }, + { + "epoch": 0.87, + "learning_rate": 3.548730443703514e-05, + "loss": 0.5117, + "step": 22684 + }, + { + "epoch": 0.87, + "learning_rate": 3.5486663246986404e-05, + "loss": 0.4445, + "step": 22685 + }, + { + "epoch": 0.87, + "learning_rate": 3.5486022056937676e-05, + "loss": 0.5298, + "step": 22686 + }, + { + "epoch": 0.87, + "learning_rate": 3.548538086688895e-05, + "loss": 0.4353, + "step": 22687 + }, + { + "epoch": 0.87, + "learning_rate": 3.548473967684022e-05, + "loss": 0.4782, + "step": 22688 + }, + { + "epoch": 0.87, + "learning_rate": 3.548409848679149e-05, + "loss": 0.4856, + "step": 22689 + }, + { + "epoch": 0.87, + "learning_rate": 3.5483457296742756e-05, + "loss": 0.4723, + "step": 22690 + }, + { + "epoch": 0.87, + "learning_rate": 3.548281610669403e-05, + "loss": 0.4755, + "step": 22691 + }, + { + "epoch": 0.87, + "learning_rate": 3.548217491664529e-05, + "loss": 0.5098, + "step": 22692 + }, + { + "epoch": 0.87, + "learning_rate": 3.5481533726596564e-05, + "loss": 0.3836, + "step": 22693 + }, + { + "epoch": 0.87, + "learning_rate": 3.5480892536547836e-05, + "loss": 0.416, + "step": 22694 + }, + { + "epoch": 0.87, + "learning_rate": 3.548025134649911e-05, + "loss": 0.4408, + "step": 22695 + }, + { + "epoch": 0.87, + "learning_rate": 3.547961015645037e-05, + "loss": 0.5725, + "step": 22696 + }, + { + "epoch": 0.87, + "learning_rate": 3.5478968966401644e-05, + "loss": 0.457, + "step": 22697 + }, + { + "epoch": 0.87, + "learning_rate": 3.5478327776352916e-05, + "loss": 0.3719, + "step": 22698 + }, + { + "epoch": 0.87, + "learning_rate": 3.547768658630418e-05, + "loss": 0.4704, + "step": 22699 + }, + { + "epoch": 0.87, + "learning_rate": 3.547704539625545e-05, + "loss": 0.4224, + "step": 22700 + }, + { + "epoch": 0.87, + "learning_rate": 3.547640420620672e-05, + "loss": 0.5585, + "step": 22701 + }, + { + "epoch": 0.87, + "learning_rate": 3.547576301615799e-05, + "loss": 0.5157, + "step": 22702 + }, + { + "epoch": 0.87, + "learning_rate": 3.547512182610926e-05, + "loss": 0.6623, + "step": 22703 + }, + { + "epoch": 0.87, + "learning_rate": 3.547448063606053e-05, + "loss": 0.4143, + "step": 22704 + }, + { + "epoch": 0.87, + "learning_rate": 3.54738394460118e-05, + "loss": 0.456, + "step": 22705 + }, + { + "epoch": 0.87, + "learning_rate": 3.547319825596307e-05, + "loss": 0.4988, + "step": 22706 + }, + { + "epoch": 0.87, + "learning_rate": 3.547255706591434e-05, + "loss": 0.4872, + "step": 22707 + }, + { + "epoch": 0.87, + "learning_rate": 3.5471915875865606e-05, + "loss": 0.5407, + "step": 22708 + }, + { + "epoch": 0.87, + "learning_rate": 3.547127468581688e-05, + "loss": 0.5273, + "step": 22709 + }, + { + "epoch": 0.87, + "learning_rate": 3.547063349576815e-05, + "loss": 0.6564, + "step": 22710 + }, + { + "epoch": 0.87, + "learning_rate": 3.546999230571942e-05, + "loss": 0.4659, + "step": 22711 + }, + { + "epoch": 0.87, + "learning_rate": 3.5469351115670686e-05, + "loss": 0.5412, + "step": 22712 + }, + { + "epoch": 0.87, + "learning_rate": 3.546870992562196e-05, + "loss": 0.5171, + "step": 22713 + }, + { + "epoch": 0.87, + "learning_rate": 3.546806873557322e-05, + "loss": 0.3686, + "step": 22714 + }, + { + "epoch": 0.87, + "learning_rate": 3.5467427545524494e-05, + "loss": 0.5025, + "step": 22715 + }, + { + "epoch": 0.87, + "learning_rate": 3.5466786355475766e-05, + "loss": 0.5051, + "step": 22716 + }, + { + "epoch": 0.87, + "learning_rate": 3.546614516542703e-05, + "loss": 0.4393, + "step": 22717 + }, + { + "epoch": 0.87, + "learning_rate": 3.54655039753783e-05, + "loss": 0.4587, + "step": 22718 + }, + { + "epoch": 0.87, + "learning_rate": 3.5464862785329574e-05, + "loss": 0.5149, + "step": 22719 + }, + { + "epoch": 0.87, + "learning_rate": 3.5464221595280846e-05, + "loss": 0.5345, + "step": 22720 + }, + { + "epoch": 0.87, + "learning_rate": 3.546358040523211e-05, + "loss": 0.5529, + "step": 22721 + }, + { + "epoch": 0.87, + "learning_rate": 3.546293921518338e-05, + "loss": 0.5161, + "step": 22722 + }, + { + "epoch": 0.87, + "learning_rate": 3.546229802513465e-05, + "loss": 0.472, + "step": 22723 + }, + { + "epoch": 0.87, + "learning_rate": 3.546165683508592e-05, + "loss": 0.5276, + "step": 22724 + }, + { + "epoch": 0.87, + "learning_rate": 3.546101564503719e-05, + "loss": 0.4284, + "step": 22725 + }, + { + "epoch": 0.87, + "learning_rate": 3.546037445498846e-05, + "loss": 0.487, + "step": 22726 + }, + { + "epoch": 0.87, + "learning_rate": 3.5459733264939735e-05, + "loss": 0.3341, + "step": 22727 + }, + { + "epoch": 0.87, + "learning_rate": 3.5459092074891e-05, + "loss": 0.4557, + "step": 22728 + }, + { + "epoch": 0.87, + "learning_rate": 3.545845088484227e-05, + "loss": 0.5688, + "step": 22729 + }, + { + "epoch": 0.87, + "learning_rate": 3.5457809694793536e-05, + "loss": 0.4725, + "step": 22730 + }, + { + "epoch": 0.87, + "learning_rate": 3.545716850474481e-05, + "loss": 0.6405, + "step": 22731 + }, + { + "epoch": 0.87, + "learning_rate": 3.545652731469607e-05, + "loss": 0.561, + "step": 22732 + }, + { + "epoch": 0.87, + "learning_rate": 3.5455886124647345e-05, + "loss": 0.4196, + "step": 22733 + }, + { + "epoch": 0.87, + "learning_rate": 3.5455244934598616e-05, + "loss": 0.5102, + "step": 22734 + }, + { + "epoch": 0.87, + "learning_rate": 3.545460374454989e-05, + "loss": 0.5318, + "step": 22735 + }, + { + "epoch": 0.87, + "learning_rate": 3.545396255450116e-05, + "loss": 0.5367, + "step": 22736 + }, + { + "epoch": 0.87, + "learning_rate": 3.5453321364452425e-05, + "loss": 0.4566, + "step": 22737 + }, + { + "epoch": 0.87, + "learning_rate": 3.5452680174403696e-05, + "loss": 0.5222, + "step": 22738 + }, + { + "epoch": 0.87, + "learning_rate": 3.545203898435496e-05, + "loss": 0.4879, + "step": 22739 + }, + { + "epoch": 0.87, + "learning_rate": 3.545139779430623e-05, + "loss": 0.491, + "step": 22740 + }, + { + "epoch": 0.87, + "learning_rate": 3.54507566042575e-05, + "loss": 0.5136, + "step": 22741 + }, + { + "epoch": 0.87, + "learning_rate": 3.5450115414208777e-05, + "loss": 0.3932, + "step": 22742 + }, + { + "epoch": 0.87, + "learning_rate": 3.544947422416005e-05, + "loss": 0.4747, + "step": 22743 + }, + { + "epoch": 0.87, + "learning_rate": 3.544883303411131e-05, + "loss": 0.4995, + "step": 22744 + }, + { + "epoch": 0.87, + "learning_rate": 3.5448191844062585e-05, + "loss": 0.4501, + "step": 22745 + }, + { + "epoch": 0.87, + "learning_rate": 3.544755065401385e-05, + "loss": 0.5284, + "step": 22746 + }, + { + "epoch": 0.87, + "learning_rate": 3.544690946396512e-05, + "loss": 0.5194, + "step": 22747 + }, + { + "epoch": 0.87, + "learning_rate": 3.5446268273916386e-05, + "loss": 0.5293, + "step": 22748 + }, + { + "epoch": 0.87, + "learning_rate": 3.544562708386766e-05, + "loss": 0.5731, + "step": 22749 + }, + { + "epoch": 0.87, + "learning_rate": 3.544498589381893e-05, + "loss": 0.514, + "step": 22750 + }, + { + "epoch": 0.87, + "learning_rate": 3.54443447037702e-05, + "loss": 0.5782, + "step": 22751 + }, + { + "epoch": 0.87, + "learning_rate": 3.544370351372147e-05, + "loss": 0.5016, + "step": 22752 + }, + { + "epoch": 0.87, + "learning_rate": 3.544306232367274e-05, + "loss": 0.6087, + "step": 22753 + }, + { + "epoch": 0.87, + "learning_rate": 3.544242113362401e-05, + "loss": 0.502, + "step": 22754 + }, + { + "epoch": 0.87, + "learning_rate": 3.5441779943575275e-05, + "loss": 0.459, + "step": 22755 + }, + { + "epoch": 0.87, + "learning_rate": 3.544113875352655e-05, + "loss": 0.5164, + "step": 22756 + }, + { + "epoch": 0.87, + "learning_rate": 3.544049756347781e-05, + "loss": 0.543, + "step": 22757 + }, + { + "epoch": 0.87, + "learning_rate": 3.543985637342908e-05, + "loss": 0.5228, + "step": 22758 + }, + { + "epoch": 0.88, + "learning_rate": 3.5439215183380355e-05, + "loss": 0.5278, + "step": 22759 + }, + { + "epoch": 0.88, + "learning_rate": 3.543857399333163e-05, + "loss": 0.3913, + "step": 22760 + }, + { + "epoch": 0.88, + "learning_rate": 3.54379328032829e-05, + "loss": 0.5269, + "step": 22761 + }, + { + "epoch": 0.88, + "learning_rate": 3.5437291613234163e-05, + "loss": 0.5, + "step": 22762 + }, + { + "epoch": 0.88, + "learning_rate": 3.5436650423185435e-05, + "loss": 0.5918, + "step": 22763 + }, + { + "epoch": 0.88, + "learning_rate": 3.54360092331367e-05, + "loss": 0.4356, + "step": 22764 + }, + { + "epoch": 0.88, + "learning_rate": 3.543536804308797e-05, + "loss": 0.5261, + "step": 22765 + }, + { + "epoch": 0.88, + "learning_rate": 3.5434726853039244e-05, + "loss": 0.5128, + "step": 22766 + }, + { + "epoch": 0.88, + "learning_rate": 3.5434085662990515e-05, + "loss": 0.5251, + "step": 22767 + }, + { + "epoch": 0.88, + "learning_rate": 3.543344447294178e-05, + "loss": 0.5555, + "step": 22768 + }, + { + "epoch": 0.88, + "learning_rate": 3.543280328289305e-05, + "loss": 0.5741, + "step": 22769 + }, + { + "epoch": 0.88, + "learning_rate": 3.5432162092844324e-05, + "loss": 0.4889, + "step": 22770 + }, + { + "epoch": 0.88, + "learning_rate": 3.543152090279559e-05, + "loss": 0.4353, + "step": 22771 + }, + { + "epoch": 0.88, + "learning_rate": 3.543087971274686e-05, + "loss": 0.456, + "step": 22772 + }, + { + "epoch": 0.88, + "learning_rate": 3.5430238522698125e-05, + "loss": 0.6156, + "step": 22773 + }, + { + "epoch": 0.88, + "learning_rate": 3.54295973326494e-05, + "loss": 0.5801, + "step": 22774 + }, + { + "epoch": 0.88, + "learning_rate": 3.542895614260067e-05, + "loss": 0.5254, + "step": 22775 + }, + { + "epoch": 0.88, + "learning_rate": 3.542831495255194e-05, + "loss": 0.4448, + "step": 22776 + }, + { + "epoch": 0.88, + "learning_rate": 3.5427673762503205e-05, + "loss": 0.5489, + "step": 22777 + }, + { + "epoch": 0.88, + "learning_rate": 3.542703257245448e-05, + "loss": 0.4296, + "step": 22778 + }, + { + "epoch": 0.88, + "learning_rate": 3.542639138240575e-05, + "loss": 0.5396, + "step": 22779 + }, + { + "epoch": 0.88, + "learning_rate": 3.5425750192357014e-05, + "loss": 0.5286, + "step": 22780 + }, + { + "epoch": 0.88, + "learning_rate": 3.5425109002308285e-05, + "loss": 0.4931, + "step": 22781 + }, + { + "epoch": 0.88, + "learning_rate": 3.542446781225956e-05, + "loss": 0.4621, + "step": 22782 + }, + { + "epoch": 0.88, + "learning_rate": 3.542382662221083e-05, + "loss": 0.5373, + "step": 22783 + }, + { + "epoch": 0.88, + "learning_rate": 3.5423185432162094e-05, + "loss": 0.5356, + "step": 22784 + }, + { + "epoch": 0.88, + "learning_rate": 3.5422544242113365e-05, + "loss": 0.4114, + "step": 22785 + }, + { + "epoch": 0.88, + "learning_rate": 3.542190305206463e-05, + "loss": 0.4028, + "step": 22786 + }, + { + "epoch": 0.88, + "learning_rate": 3.54212618620159e-05, + "loss": 0.507, + "step": 22787 + }, + { + "epoch": 0.88, + "learning_rate": 3.5420620671967174e-05, + "loss": 0.4953, + "step": 22788 + }, + { + "epoch": 0.88, + "learning_rate": 3.541997948191844e-05, + "loss": 0.452, + "step": 22789 + }, + { + "epoch": 0.88, + "learning_rate": 3.541933829186971e-05, + "loss": 0.5572, + "step": 22790 + }, + { + "epoch": 0.88, + "learning_rate": 3.541869710182098e-05, + "loss": 0.5763, + "step": 22791 + }, + { + "epoch": 0.88, + "learning_rate": 3.5418055911772254e-05, + "loss": 0.6569, + "step": 22792 + }, + { + "epoch": 0.88, + "learning_rate": 3.541741472172352e-05, + "loss": 0.4583, + "step": 22793 + }, + { + "epoch": 0.88, + "learning_rate": 3.541677353167479e-05, + "loss": 0.5408, + "step": 22794 + }, + { + "epoch": 0.88, + "learning_rate": 3.5416132341626056e-05, + "loss": 0.4206, + "step": 22795 + }, + { + "epoch": 0.88, + "learning_rate": 3.541549115157733e-05, + "loss": 0.5046, + "step": 22796 + }, + { + "epoch": 0.88, + "learning_rate": 3.54148499615286e-05, + "loss": 0.5383, + "step": 22797 + }, + { + "epoch": 0.88, + "learning_rate": 3.541420877147987e-05, + "loss": 0.635, + "step": 22798 + }, + { + "epoch": 0.88, + "learning_rate": 3.541356758143114e-05, + "loss": 0.571, + "step": 22799 + }, + { + "epoch": 0.88, + "learning_rate": 3.541292639138241e-05, + "loss": 0.4563, + "step": 22800 + }, + { + "epoch": 0.88, + "learning_rate": 3.541228520133368e-05, + "loss": 0.4639, + "step": 22801 + }, + { + "epoch": 0.88, + "learning_rate": 3.5411644011284944e-05, + "loss": 0.4479, + "step": 22802 + }, + { + "epoch": 0.88, + "learning_rate": 3.5411002821236216e-05, + "loss": 0.5511, + "step": 22803 + }, + { + "epoch": 0.88, + "learning_rate": 3.541036163118748e-05, + "loss": 0.4726, + "step": 22804 + }, + { + "epoch": 0.88, + "learning_rate": 3.540972044113875e-05, + "loss": 0.5346, + "step": 22805 + }, + { + "epoch": 0.88, + "learning_rate": 3.5409079251090024e-05, + "loss": 0.4217, + "step": 22806 + }, + { + "epoch": 0.88, + "learning_rate": 3.5408438061041296e-05, + "loss": 0.5044, + "step": 22807 + }, + { + "epoch": 0.88, + "learning_rate": 3.540779687099257e-05, + "loss": 0.5159, + "step": 22808 + }, + { + "epoch": 0.88, + "learning_rate": 3.540715568094383e-05, + "loss": 0.4842, + "step": 22809 + }, + { + "epoch": 0.88, + "learning_rate": 3.5406514490895104e-05, + "loss": 0.5806, + "step": 22810 + }, + { + "epoch": 0.88, + "learning_rate": 3.540587330084637e-05, + "loss": 0.3982, + "step": 22811 + }, + { + "epoch": 0.88, + "learning_rate": 3.540523211079764e-05, + "loss": 0.5415, + "step": 22812 + }, + { + "epoch": 0.88, + "learning_rate": 3.5404590920748906e-05, + "loss": 0.5888, + "step": 22813 + }, + { + "epoch": 0.88, + "learning_rate": 3.5403949730700184e-05, + "loss": 0.5635, + "step": 22814 + }, + { + "epoch": 0.88, + "learning_rate": 3.5403308540651456e-05, + "loss": 0.4083, + "step": 22815 + }, + { + "epoch": 0.88, + "learning_rate": 3.540266735060272e-05, + "loss": 0.4782, + "step": 22816 + }, + { + "epoch": 0.88, + "learning_rate": 3.540202616055399e-05, + "loss": 0.6501, + "step": 22817 + }, + { + "epoch": 0.88, + "learning_rate": 3.540138497050526e-05, + "loss": 0.5384, + "step": 22818 + }, + { + "epoch": 0.88, + "learning_rate": 3.540074378045653e-05, + "loss": 0.3736, + "step": 22819 + }, + { + "epoch": 0.88, + "learning_rate": 3.5400102590407794e-05, + "loss": 0.5861, + "step": 22820 + }, + { + "epoch": 0.88, + "learning_rate": 3.5399461400359066e-05, + "loss": 0.5106, + "step": 22821 + }, + { + "epoch": 0.88, + "learning_rate": 3.539882021031034e-05, + "loss": 0.4428, + "step": 22822 + }, + { + "epoch": 0.88, + "learning_rate": 3.539817902026161e-05, + "loss": 0.5407, + "step": 22823 + }, + { + "epoch": 0.88, + "learning_rate": 3.539753783021288e-05, + "loss": 0.4454, + "step": 22824 + }, + { + "epoch": 0.88, + "learning_rate": 3.5396896640164146e-05, + "loss": 0.529, + "step": 22825 + }, + { + "epoch": 0.88, + "learning_rate": 3.539625545011542e-05, + "loss": 0.403, + "step": 22826 + }, + { + "epoch": 0.88, + "learning_rate": 3.539561426006668e-05, + "loss": 0.4104, + "step": 22827 + }, + { + "epoch": 0.88, + "learning_rate": 3.5394973070017954e-05, + "loss": 0.4866, + "step": 22828 + }, + { + "epoch": 0.88, + "learning_rate": 3.539433187996922e-05, + "loss": 0.5112, + "step": 22829 + }, + { + "epoch": 0.88, + "learning_rate": 3.53936906899205e-05, + "loss": 0.5728, + "step": 22830 + }, + { + "epoch": 0.88, + "learning_rate": 3.539304949987176e-05, + "loss": 0.4985, + "step": 22831 + }, + { + "epoch": 0.88, + "learning_rate": 3.5392408309823035e-05, + "loss": 0.5082, + "step": 22832 + }, + { + "epoch": 0.88, + "learning_rate": 3.5391767119774306e-05, + "loss": 0.5316, + "step": 22833 + }, + { + "epoch": 0.88, + "learning_rate": 3.539112592972557e-05, + "loss": 0.6441, + "step": 22834 + }, + { + "epoch": 0.88, + "learning_rate": 3.539048473967684e-05, + "loss": 0.4836, + "step": 22835 + }, + { + "epoch": 0.88, + "learning_rate": 3.538984354962811e-05, + "loss": 0.6336, + "step": 22836 + }, + { + "epoch": 0.88, + "learning_rate": 3.538920235957938e-05, + "loss": 0.477, + "step": 22837 + }, + { + "epoch": 0.88, + "learning_rate": 3.538856116953065e-05, + "loss": 0.4756, + "step": 22838 + }, + { + "epoch": 0.88, + "learning_rate": 3.538791997948192e-05, + "loss": 0.5025, + "step": 22839 + }, + { + "epoch": 0.88, + "learning_rate": 3.538727878943319e-05, + "loss": 0.4456, + "step": 22840 + }, + { + "epoch": 0.88, + "learning_rate": 3.538663759938446e-05, + "loss": 0.5406, + "step": 22841 + }, + { + "epoch": 0.88, + "learning_rate": 3.538599640933573e-05, + "loss": 0.4223, + "step": 22842 + }, + { + "epoch": 0.88, + "learning_rate": 3.5385355219286996e-05, + "loss": 0.493, + "step": 22843 + }, + { + "epoch": 0.88, + "learning_rate": 3.538471402923827e-05, + "loss": 0.4576, + "step": 22844 + }, + { + "epoch": 0.88, + "learning_rate": 3.538407283918953e-05, + "loss": 0.5024, + "step": 22845 + }, + { + "epoch": 0.88, + "learning_rate": 3.5383431649140805e-05, + "loss": 0.5531, + "step": 22846 + }, + { + "epoch": 0.88, + "learning_rate": 3.5382790459092076e-05, + "loss": 0.4147, + "step": 22847 + }, + { + "epoch": 0.88, + "learning_rate": 3.538214926904335e-05, + "loss": 0.4516, + "step": 22848 + }, + { + "epoch": 0.88, + "learning_rate": 3.538150807899461e-05, + "loss": 0.4856, + "step": 22849 + }, + { + "epoch": 0.88, + "learning_rate": 3.5380866888945885e-05, + "loss": 0.4785, + "step": 22850 + }, + { + "epoch": 0.88, + "learning_rate": 3.5380225698897156e-05, + "loss": 0.4662, + "step": 22851 + }, + { + "epoch": 0.88, + "learning_rate": 3.537958450884842e-05, + "loss": 0.4758, + "step": 22852 + }, + { + "epoch": 0.88, + "learning_rate": 3.537894331879969e-05, + "loss": 0.6002, + "step": 22853 + }, + { + "epoch": 0.88, + "learning_rate": 3.5378302128750965e-05, + "loss": 0.4617, + "step": 22854 + }, + { + "epoch": 0.88, + "learning_rate": 3.5377660938702237e-05, + "loss": 0.5007, + "step": 22855 + }, + { + "epoch": 0.88, + "learning_rate": 3.53770197486535e-05, + "loss": 0.4631, + "step": 22856 + }, + { + "epoch": 0.88, + "learning_rate": 3.537637855860477e-05, + "loss": 0.5615, + "step": 22857 + }, + { + "epoch": 0.88, + "learning_rate": 3.5375737368556045e-05, + "loss": 0.4528, + "step": 22858 + }, + { + "epoch": 0.88, + "learning_rate": 3.537509617850731e-05, + "loss": 0.4539, + "step": 22859 + }, + { + "epoch": 0.88, + "learning_rate": 3.537445498845858e-05, + "loss": 0.5283, + "step": 22860 + }, + { + "epoch": 0.88, + "learning_rate": 3.5373813798409847e-05, + "loss": 0.5486, + "step": 22861 + }, + { + "epoch": 0.88, + "learning_rate": 3.537317260836112e-05, + "loss": 0.5569, + "step": 22862 + }, + { + "epoch": 0.88, + "learning_rate": 3.537253141831239e-05, + "loss": 0.4856, + "step": 22863 + }, + { + "epoch": 0.88, + "learning_rate": 3.537189022826366e-05, + "loss": 0.5755, + "step": 22864 + }, + { + "epoch": 0.88, + "learning_rate": 3.537124903821493e-05, + "loss": 0.5316, + "step": 22865 + }, + { + "epoch": 0.88, + "learning_rate": 3.53706078481662e-05, + "loss": 0.45, + "step": 22866 + }, + { + "epoch": 0.88, + "learning_rate": 3.536996665811747e-05, + "loss": 0.5119, + "step": 22867 + }, + { + "epoch": 0.88, + "learning_rate": 3.5369325468068735e-05, + "loss": 0.5276, + "step": 22868 + }, + { + "epoch": 0.88, + "learning_rate": 3.536868427802001e-05, + "loss": 0.5271, + "step": 22869 + }, + { + "epoch": 0.88, + "learning_rate": 3.536804308797128e-05, + "loss": 0.5431, + "step": 22870 + }, + { + "epoch": 0.88, + "learning_rate": 3.536740189792255e-05, + "loss": 0.3734, + "step": 22871 + }, + { + "epoch": 0.88, + "learning_rate": 3.5366760707873815e-05, + "loss": 0.4586, + "step": 22872 + }, + { + "epoch": 0.88, + "learning_rate": 3.536611951782509e-05, + "loss": 0.5985, + "step": 22873 + }, + { + "epoch": 0.88, + "learning_rate": 3.536547832777635e-05, + "loss": 0.5297, + "step": 22874 + }, + { + "epoch": 0.88, + "learning_rate": 3.5364837137727623e-05, + "loss": 0.5899, + "step": 22875 + }, + { + "epoch": 0.88, + "learning_rate": 3.5364195947678895e-05, + "loss": 0.3989, + "step": 22876 + }, + { + "epoch": 0.88, + "learning_rate": 3.536355475763016e-05, + "loss": 0.4161, + "step": 22877 + }, + { + "epoch": 0.88, + "learning_rate": 3.536291356758143e-05, + "loss": 0.4594, + "step": 22878 + }, + { + "epoch": 0.88, + "learning_rate": 3.5362272377532704e-05, + "loss": 0.465, + "step": 22879 + }, + { + "epoch": 0.88, + "learning_rate": 3.5361631187483975e-05, + "loss": 0.4322, + "step": 22880 + }, + { + "epoch": 0.88, + "learning_rate": 3.536098999743524e-05, + "loss": 0.5258, + "step": 22881 + }, + { + "epoch": 0.88, + "learning_rate": 3.536034880738651e-05, + "loss": 0.4841, + "step": 22882 + }, + { + "epoch": 0.88, + "learning_rate": 3.535970761733778e-05, + "loss": 0.5676, + "step": 22883 + }, + { + "epoch": 0.88, + "learning_rate": 3.535906642728905e-05, + "loss": 0.5214, + "step": 22884 + }, + { + "epoch": 0.88, + "learning_rate": 3.535842523724032e-05, + "loss": 0.4876, + "step": 22885 + }, + { + "epoch": 0.88, + "learning_rate": 3.535778404719159e-05, + "loss": 0.5698, + "step": 22886 + }, + { + "epoch": 0.88, + "learning_rate": 3.5357142857142864e-05, + "loss": 0.3442, + "step": 22887 + }, + { + "epoch": 0.88, + "learning_rate": 3.535650166709413e-05, + "loss": 0.5265, + "step": 22888 + }, + { + "epoch": 0.88, + "learning_rate": 3.53558604770454e-05, + "loss": 0.5312, + "step": 22889 + }, + { + "epoch": 0.88, + "learning_rate": 3.5355219286996665e-05, + "loss": 0.4673, + "step": 22890 + }, + { + "epoch": 0.88, + "learning_rate": 3.535457809694794e-05, + "loss": 0.3167, + "step": 22891 + }, + { + "epoch": 0.88, + "learning_rate": 3.53539369068992e-05, + "loss": 0.547, + "step": 22892 + }, + { + "epoch": 0.88, + "learning_rate": 3.5353295716850474e-05, + "loss": 0.3986, + "step": 22893 + }, + { + "epoch": 0.88, + "learning_rate": 3.5352654526801745e-05, + "loss": 0.4096, + "step": 22894 + }, + { + "epoch": 0.88, + "learning_rate": 3.535201333675302e-05, + "loss": 0.4497, + "step": 22895 + }, + { + "epoch": 0.88, + "learning_rate": 3.535137214670429e-05, + "loss": 0.363, + "step": 22896 + }, + { + "epoch": 0.88, + "learning_rate": 3.5350730956655554e-05, + "loss": 0.4719, + "step": 22897 + }, + { + "epoch": 0.88, + "learning_rate": 3.5350089766606826e-05, + "loss": 0.5234, + "step": 22898 + }, + { + "epoch": 0.88, + "learning_rate": 3.534944857655809e-05, + "loss": 0.4278, + "step": 22899 + }, + { + "epoch": 0.88, + "learning_rate": 3.534880738650936e-05, + "loss": 0.451, + "step": 22900 + }, + { + "epoch": 0.88, + "learning_rate": 3.534816619646063e-05, + "loss": 0.5349, + "step": 22901 + }, + { + "epoch": 0.88, + "learning_rate": 3.5347525006411906e-05, + "loss": 0.5583, + "step": 22902 + }, + { + "epoch": 0.88, + "learning_rate": 3.534688381636318e-05, + "loss": 0.4055, + "step": 22903 + }, + { + "epoch": 0.88, + "learning_rate": 3.534624262631444e-05, + "loss": 0.5605, + "step": 22904 + }, + { + "epoch": 0.88, + "learning_rate": 3.5345601436265714e-05, + "loss": 0.5736, + "step": 22905 + }, + { + "epoch": 0.88, + "learning_rate": 3.534496024621698e-05, + "loss": 0.4471, + "step": 22906 + }, + { + "epoch": 0.88, + "learning_rate": 3.534431905616825e-05, + "loss": 0.451, + "step": 22907 + }, + { + "epoch": 0.88, + "learning_rate": 3.5343677866119516e-05, + "loss": 0.4814, + "step": 22908 + }, + { + "epoch": 0.88, + "learning_rate": 3.534303667607079e-05, + "loss": 0.3637, + "step": 22909 + }, + { + "epoch": 0.88, + "learning_rate": 3.534239548602206e-05, + "loss": 0.5077, + "step": 22910 + }, + { + "epoch": 0.88, + "learning_rate": 3.534175429597333e-05, + "loss": 0.4618, + "step": 22911 + }, + { + "epoch": 0.88, + "learning_rate": 3.53411131059246e-05, + "loss": 0.4185, + "step": 22912 + }, + { + "epoch": 0.88, + "learning_rate": 3.534047191587587e-05, + "loss": 0.5396, + "step": 22913 + }, + { + "epoch": 0.88, + "learning_rate": 3.533983072582714e-05, + "loss": 0.5212, + "step": 22914 + }, + { + "epoch": 0.88, + "learning_rate": 3.5339189535778404e-05, + "loss": 0.3981, + "step": 22915 + }, + { + "epoch": 0.88, + "learning_rate": 3.5338548345729676e-05, + "loss": 0.4856, + "step": 22916 + }, + { + "epoch": 0.88, + "learning_rate": 3.533790715568094e-05, + "loss": 0.5026, + "step": 22917 + }, + { + "epoch": 0.88, + "learning_rate": 3.533726596563222e-05, + "loss": 0.5305, + "step": 22918 + }, + { + "epoch": 0.88, + "learning_rate": 3.5336624775583484e-05, + "loss": 0.4875, + "step": 22919 + }, + { + "epoch": 0.88, + "learning_rate": 3.5335983585534756e-05, + "loss": 0.484, + "step": 22920 + }, + { + "epoch": 0.88, + "learning_rate": 3.533534239548603e-05, + "loss": 0.595, + "step": 22921 + }, + { + "epoch": 0.88, + "learning_rate": 3.533470120543729e-05, + "loss": 0.4586, + "step": 22922 + }, + { + "epoch": 0.88, + "learning_rate": 3.5334060015388564e-05, + "loss": 0.3729, + "step": 22923 + }, + { + "epoch": 0.88, + "learning_rate": 3.533341882533983e-05, + "loss": 0.4784, + "step": 22924 + }, + { + "epoch": 0.88, + "learning_rate": 3.53327776352911e-05, + "loss": 0.6003, + "step": 22925 + }, + { + "epoch": 0.88, + "learning_rate": 3.533213644524237e-05, + "loss": 0.4813, + "step": 22926 + }, + { + "epoch": 0.88, + "learning_rate": 3.5331495255193644e-05, + "loss": 0.551, + "step": 22927 + }, + { + "epoch": 0.88, + "learning_rate": 3.533085406514491e-05, + "loss": 0.4144, + "step": 22928 + }, + { + "epoch": 0.88, + "learning_rate": 3.533021287509618e-05, + "loss": 0.5712, + "step": 22929 + }, + { + "epoch": 0.88, + "learning_rate": 3.532957168504745e-05, + "loss": 0.4444, + "step": 22930 + }, + { + "epoch": 0.88, + "learning_rate": 3.532893049499872e-05, + "loss": 0.6027, + "step": 22931 + }, + { + "epoch": 0.88, + "learning_rate": 3.532828930494999e-05, + "loss": 0.448, + "step": 22932 + }, + { + "epoch": 0.88, + "learning_rate": 3.5327648114901254e-05, + "loss": 0.5655, + "step": 22933 + }, + { + "epoch": 0.88, + "learning_rate": 3.5327006924852526e-05, + "loss": 0.5674, + "step": 22934 + }, + { + "epoch": 0.88, + "learning_rate": 3.53263657348038e-05, + "loss": 0.517, + "step": 22935 + }, + { + "epoch": 0.88, + "learning_rate": 3.532572454475507e-05, + "loss": 0.4715, + "step": 22936 + }, + { + "epoch": 0.88, + "learning_rate": 3.5325083354706334e-05, + "loss": 0.5293, + "step": 22937 + }, + { + "epoch": 0.88, + "learning_rate": 3.5324442164657606e-05, + "loss": 0.5372, + "step": 22938 + }, + { + "epoch": 0.88, + "learning_rate": 3.532380097460888e-05, + "loss": 0.4115, + "step": 22939 + }, + { + "epoch": 0.88, + "learning_rate": 3.532315978456014e-05, + "loss": 0.4484, + "step": 22940 + }, + { + "epoch": 0.88, + "learning_rate": 3.5322518594511414e-05, + "loss": 0.4835, + "step": 22941 + }, + { + "epoch": 0.88, + "learning_rate": 3.5321877404462686e-05, + "loss": 0.5211, + "step": 22942 + }, + { + "epoch": 0.88, + "learning_rate": 3.532123621441396e-05, + "loss": 0.5443, + "step": 22943 + }, + { + "epoch": 0.88, + "learning_rate": 3.532059502436522e-05, + "loss": 0.4044, + "step": 22944 + }, + { + "epoch": 0.88, + "learning_rate": 3.5319953834316495e-05, + "loss": 0.5351, + "step": 22945 + }, + { + "epoch": 0.88, + "learning_rate": 3.531931264426776e-05, + "loss": 0.5702, + "step": 22946 + }, + { + "epoch": 0.88, + "learning_rate": 3.531867145421903e-05, + "loss": 0.5812, + "step": 22947 + }, + { + "epoch": 0.88, + "learning_rate": 3.53180302641703e-05, + "loss": 0.5668, + "step": 22948 + }, + { + "epoch": 0.88, + "learning_rate": 3.531738907412157e-05, + "loss": 0.5444, + "step": 22949 + }, + { + "epoch": 0.88, + "learning_rate": 3.531674788407284e-05, + "loss": 0.5428, + "step": 22950 + }, + { + "epoch": 0.88, + "learning_rate": 3.531610669402411e-05, + "loss": 0.4639, + "step": 22951 + }, + { + "epoch": 0.88, + "learning_rate": 3.531546550397538e-05, + "loss": 0.6148, + "step": 22952 + }, + { + "epoch": 0.88, + "learning_rate": 3.531482431392665e-05, + "loss": 0.5887, + "step": 22953 + }, + { + "epoch": 0.88, + "learning_rate": 3.531418312387792e-05, + "loss": 0.4947, + "step": 22954 + }, + { + "epoch": 0.88, + "learning_rate": 3.5313541933829185e-05, + "loss": 0.5771, + "step": 22955 + }, + { + "epoch": 0.88, + "learning_rate": 3.5312900743780456e-05, + "loss": 0.4312, + "step": 22956 + }, + { + "epoch": 0.88, + "learning_rate": 3.531225955373173e-05, + "loss": 0.5727, + "step": 22957 + }, + { + "epoch": 0.88, + "learning_rate": 3.5311618363683e-05, + "loss": 0.5134, + "step": 22958 + }, + { + "epoch": 0.88, + "learning_rate": 3.531097717363427e-05, + "loss": 0.4365, + "step": 22959 + }, + { + "epoch": 0.88, + "learning_rate": 3.5310335983585536e-05, + "loss": 0.533, + "step": 22960 + }, + { + "epoch": 0.88, + "learning_rate": 3.530969479353681e-05, + "loss": 0.5178, + "step": 22961 + }, + { + "epoch": 0.88, + "learning_rate": 3.530905360348807e-05, + "loss": 0.4232, + "step": 22962 + }, + { + "epoch": 0.88, + "learning_rate": 3.5308412413439345e-05, + "loss": 0.573, + "step": 22963 + }, + { + "epoch": 0.88, + "learning_rate": 3.530777122339061e-05, + "loss": 0.5668, + "step": 22964 + }, + { + "epoch": 0.88, + "learning_rate": 3.530713003334188e-05, + "loss": 0.5069, + "step": 22965 + }, + { + "epoch": 0.88, + "learning_rate": 3.530648884329315e-05, + "loss": 0.4435, + "step": 22966 + }, + { + "epoch": 0.88, + "learning_rate": 3.5305847653244425e-05, + "loss": 0.5632, + "step": 22967 + }, + { + "epoch": 0.88, + "learning_rate": 3.5305206463195697e-05, + "loss": 0.5133, + "step": 22968 + }, + { + "epoch": 0.88, + "learning_rate": 3.530456527314696e-05, + "loss": 0.4968, + "step": 22969 + }, + { + "epoch": 0.88, + "learning_rate": 3.530392408309823e-05, + "loss": 0.432, + "step": 22970 + }, + { + "epoch": 0.88, + "learning_rate": 3.53032828930495e-05, + "loss": 0.4485, + "step": 22971 + }, + { + "epoch": 0.88, + "learning_rate": 3.530264170300077e-05, + "loss": 0.4902, + "step": 22972 + }, + { + "epoch": 0.88, + "learning_rate": 3.5302000512952035e-05, + "loss": 0.5086, + "step": 22973 + }, + { + "epoch": 0.88, + "learning_rate": 3.530135932290331e-05, + "loss": 0.4487, + "step": 22974 + }, + { + "epoch": 0.88, + "learning_rate": 3.5300718132854585e-05, + "loss": 0.3956, + "step": 22975 + }, + { + "epoch": 0.88, + "learning_rate": 3.530007694280585e-05, + "loss": 0.503, + "step": 22976 + }, + { + "epoch": 0.88, + "learning_rate": 3.529943575275712e-05, + "loss": 0.5427, + "step": 22977 + }, + { + "epoch": 0.88, + "learning_rate": 3.529879456270839e-05, + "loss": 0.4747, + "step": 22978 + }, + { + "epoch": 0.88, + "learning_rate": 3.529815337265966e-05, + "loss": 0.4573, + "step": 22979 + }, + { + "epoch": 0.88, + "learning_rate": 3.529751218261092e-05, + "loss": 0.5293, + "step": 22980 + }, + { + "epoch": 0.88, + "learning_rate": 3.5296870992562195e-05, + "loss": 0.4678, + "step": 22981 + }, + { + "epoch": 0.88, + "learning_rate": 3.529622980251347e-05, + "loss": 0.5549, + "step": 22982 + }, + { + "epoch": 0.88, + "learning_rate": 3.529558861246474e-05, + "loss": 0.5179, + "step": 22983 + }, + { + "epoch": 0.88, + "learning_rate": 3.529494742241601e-05, + "loss": 0.478, + "step": 22984 + }, + { + "epoch": 0.88, + "learning_rate": 3.5294306232367275e-05, + "loss": 0.5108, + "step": 22985 + }, + { + "epoch": 0.88, + "learning_rate": 3.529366504231855e-05, + "loss": 0.4951, + "step": 22986 + }, + { + "epoch": 0.88, + "learning_rate": 3.529302385226981e-05, + "loss": 0.5321, + "step": 22987 + }, + { + "epoch": 0.88, + "learning_rate": 3.5292382662221083e-05, + "loss": 0.4439, + "step": 22988 + }, + { + "epoch": 0.88, + "learning_rate": 3.529174147217235e-05, + "loss": 0.5501, + "step": 22989 + }, + { + "epoch": 0.88, + "learning_rate": 3.529110028212363e-05, + "loss": 0.5036, + "step": 22990 + }, + { + "epoch": 0.88, + "learning_rate": 3.529045909207489e-05, + "loss": 0.6323, + "step": 22991 + }, + { + "epoch": 0.88, + "learning_rate": 3.5289817902026164e-05, + "loss": 0.4234, + "step": 22992 + }, + { + "epoch": 0.88, + "learning_rate": 3.5289176711977435e-05, + "loss": 0.2996, + "step": 22993 + }, + { + "epoch": 0.88, + "learning_rate": 3.52885355219287e-05, + "loss": 0.5841, + "step": 22994 + }, + { + "epoch": 0.88, + "learning_rate": 3.528789433187997e-05, + "loss": 0.5238, + "step": 22995 + }, + { + "epoch": 0.88, + "learning_rate": 3.528725314183124e-05, + "loss": 0.4829, + "step": 22996 + }, + { + "epoch": 0.88, + "learning_rate": 3.528661195178251e-05, + "loss": 0.4607, + "step": 22997 + }, + { + "epoch": 0.88, + "learning_rate": 3.528597076173378e-05, + "loss": 0.541, + "step": 22998 + }, + { + "epoch": 0.88, + "learning_rate": 3.528532957168505e-05, + "loss": 0.5127, + "step": 22999 + }, + { + "epoch": 0.88, + "learning_rate": 3.528468838163632e-05, + "loss": 0.4184, + "step": 23000 + }, + { + "epoch": 0.88, + "learning_rate": 3.528404719158759e-05, + "loss": 0.4776, + "step": 23001 + }, + { + "epoch": 0.88, + "learning_rate": 3.528340600153886e-05, + "loss": 0.5096, + "step": 23002 + }, + { + "epoch": 0.88, + "learning_rate": 3.5282764811490125e-05, + "loss": 0.4868, + "step": 23003 + }, + { + "epoch": 0.88, + "learning_rate": 3.52821236214414e-05, + "loss": 0.5473, + "step": 23004 + }, + { + "epoch": 0.88, + "learning_rate": 3.528148243139266e-05, + "loss": 0.5429, + "step": 23005 + }, + { + "epoch": 0.88, + "learning_rate": 3.5280841241343934e-05, + "loss": 0.5249, + "step": 23006 + }, + { + "epoch": 0.88, + "learning_rate": 3.5280200051295205e-05, + "loss": 0.4271, + "step": 23007 + }, + { + "epoch": 0.88, + "learning_rate": 3.527955886124648e-05, + "loss": 0.3823, + "step": 23008 + }, + { + "epoch": 0.88, + "learning_rate": 3.527891767119774e-05, + "loss": 0.3392, + "step": 23009 + }, + { + "epoch": 0.88, + "learning_rate": 3.5278276481149014e-05, + "loss": 0.4589, + "step": 23010 + }, + { + "epoch": 0.88, + "learning_rate": 3.5277635291100286e-05, + "loss": 0.4436, + "step": 23011 + }, + { + "epoch": 0.88, + "learning_rate": 3.527699410105155e-05, + "loss": 0.4291, + "step": 23012 + }, + { + "epoch": 0.88, + "learning_rate": 3.527635291100282e-05, + "loss": 0.5397, + "step": 23013 + }, + { + "epoch": 0.88, + "learning_rate": 3.5275711720954094e-05, + "loss": 0.344, + "step": 23014 + }, + { + "epoch": 0.88, + "learning_rate": 3.5275070530905366e-05, + "loss": 0.43, + "step": 23015 + }, + { + "epoch": 0.88, + "learning_rate": 3.527442934085663e-05, + "loss": 0.3873, + "step": 23016 + }, + { + "epoch": 0.88, + "learning_rate": 3.52737881508079e-05, + "loss": 0.4571, + "step": 23017 + }, + { + "epoch": 0.88, + "learning_rate": 3.527314696075917e-05, + "loss": 0.504, + "step": 23018 + }, + { + "epoch": 0.89, + "learning_rate": 3.527250577071044e-05, + "loss": 0.415, + "step": 23019 + }, + { + "epoch": 0.89, + "learning_rate": 3.527186458066171e-05, + "loss": 0.5376, + "step": 23020 + }, + { + "epoch": 0.89, + "learning_rate": 3.5271223390612976e-05, + "loss": 0.5068, + "step": 23021 + }, + { + "epoch": 0.89, + "learning_rate": 3.527058220056425e-05, + "loss": 0.5748, + "step": 23022 + }, + { + "epoch": 0.89, + "learning_rate": 3.526994101051552e-05, + "loss": 0.4927, + "step": 23023 + }, + { + "epoch": 0.89, + "learning_rate": 3.526929982046679e-05, + "loss": 0.5717, + "step": 23024 + }, + { + "epoch": 0.89, + "learning_rate": 3.5268658630418056e-05, + "loss": 0.5574, + "step": 23025 + }, + { + "epoch": 0.89, + "learning_rate": 3.526801744036933e-05, + "loss": 0.5327, + "step": 23026 + }, + { + "epoch": 0.89, + "learning_rate": 3.526737625032059e-05, + "loss": 0.4045, + "step": 23027 + }, + { + "epoch": 0.89, + "learning_rate": 3.5266735060271864e-05, + "loss": 0.5521, + "step": 23028 + }, + { + "epoch": 0.89, + "learning_rate": 3.5266093870223136e-05, + "loss": 0.4514, + "step": 23029 + }, + { + "epoch": 0.89, + "learning_rate": 3.526545268017441e-05, + "loss": 0.5316, + "step": 23030 + }, + { + "epoch": 0.89, + "learning_rate": 3.526481149012568e-05, + "loss": 0.4744, + "step": 23031 + }, + { + "epoch": 0.89, + "learning_rate": 3.5264170300076944e-05, + "loss": 0.4954, + "step": 23032 + }, + { + "epoch": 0.89, + "learning_rate": 3.5263529110028216e-05, + "loss": 0.4939, + "step": 23033 + }, + { + "epoch": 0.89, + "learning_rate": 3.526288791997948e-05, + "loss": 0.5718, + "step": 23034 + }, + { + "epoch": 0.89, + "learning_rate": 3.526224672993075e-05, + "loss": 0.4188, + "step": 23035 + }, + { + "epoch": 0.89, + "learning_rate": 3.526160553988202e-05, + "loss": 0.5307, + "step": 23036 + }, + { + "epoch": 0.89, + "learning_rate": 3.526096434983329e-05, + "loss": 0.4166, + "step": 23037 + }, + { + "epoch": 0.89, + "learning_rate": 3.526032315978456e-05, + "loss": 0.4691, + "step": 23038 + }, + { + "epoch": 0.89, + "learning_rate": 3.525968196973583e-05, + "loss": 0.558, + "step": 23039 + }, + { + "epoch": 0.89, + "learning_rate": 3.5259040779687104e-05, + "loss": 0.5307, + "step": 23040 + }, + { + "epoch": 0.89, + "learning_rate": 3.525839958963837e-05, + "loss": 0.4123, + "step": 23041 + }, + { + "epoch": 0.89, + "learning_rate": 3.525775839958964e-05, + "loss": 0.5004, + "step": 23042 + }, + { + "epoch": 0.89, + "learning_rate": 3.5257117209540906e-05, + "loss": 0.5418, + "step": 23043 + }, + { + "epoch": 0.89, + "learning_rate": 3.525647601949218e-05, + "loss": 0.4718, + "step": 23044 + }, + { + "epoch": 0.89, + "learning_rate": 3.525583482944344e-05, + "loss": 0.4165, + "step": 23045 + }, + { + "epoch": 0.89, + "learning_rate": 3.525519363939472e-05, + "loss": 0.5605, + "step": 23046 + }, + { + "epoch": 0.89, + "learning_rate": 3.525455244934599e-05, + "loss": 0.4499, + "step": 23047 + }, + { + "epoch": 0.89, + "learning_rate": 3.525391125929726e-05, + "loss": 0.4217, + "step": 23048 + }, + { + "epoch": 0.89, + "learning_rate": 3.525327006924853e-05, + "loss": 0.5432, + "step": 23049 + }, + { + "epoch": 0.89, + "learning_rate": 3.5252628879199794e-05, + "loss": 0.4099, + "step": 23050 + }, + { + "epoch": 0.89, + "learning_rate": 3.5251987689151066e-05, + "loss": 0.4858, + "step": 23051 + }, + { + "epoch": 0.89, + "learning_rate": 3.525134649910233e-05, + "loss": 0.4591, + "step": 23052 + }, + { + "epoch": 0.89, + "learning_rate": 3.52507053090536e-05, + "loss": 0.4783, + "step": 23053 + }, + { + "epoch": 0.89, + "learning_rate": 3.5250064119004874e-05, + "loss": 0.4999, + "step": 23054 + }, + { + "epoch": 0.89, + "learning_rate": 3.5249422928956146e-05, + "loss": 0.5947, + "step": 23055 + }, + { + "epoch": 0.89, + "learning_rate": 3.524878173890742e-05, + "loss": 0.4868, + "step": 23056 + }, + { + "epoch": 0.89, + "learning_rate": 3.524814054885868e-05, + "loss": 0.5302, + "step": 23057 + }, + { + "epoch": 0.89, + "learning_rate": 3.5247499358809955e-05, + "loss": 0.4813, + "step": 23058 + }, + { + "epoch": 0.89, + "learning_rate": 3.524685816876122e-05, + "loss": 0.6136, + "step": 23059 + }, + { + "epoch": 0.89, + "learning_rate": 3.524621697871249e-05, + "loss": 0.48, + "step": 23060 + }, + { + "epoch": 0.89, + "learning_rate": 3.5245575788663756e-05, + "loss": 0.5812, + "step": 23061 + }, + { + "epoch": 0.89, + "learning_rate": 3.5244934598615035e-05, + "loss": 0.5482, + "step": 23062 + }, + { + "epoch": 0.89, + "learning_rate": 3.5244293408566306e-05, + "loss": 0.5483, + "step": 23063 + }, + { + "epoch": 0.89, + "learning_rate": 3.524365221851757e-05, + "loss": 0.4263, + "step": 23064 + }, + { + "epoch": 0.89, + "learning_rate": 3.524301102846884e-05, + "loss": 0.4771, + "step": 23065 + }, + { + "epoch": 0.89, + "learning_rate": 3.524236983842011e-05, + "loss": 0.4848, + "step": 23066 + }, + { + "epoch": 0.89, + "learning_rate": 3.524172864837138e-05, + "loss": 0.4749, + "step": 23067 + }, + { + "epoch": 0.89, + "learning_rate": 3.5241087458322645e-05, + "loss": 0.4331, + "step": 23068 + }, + { + "epoch": 0.89, + "learning_rate": 3.5240446268273916e-05, + "loss": 0.5109, + "step": 23069 + }, + { + "epoch": 0.89, + "learning_rate": 3.523980507822519e-05, + "loss": 0.5474, + "step": 23070 + }, + { + "epoch": 0.89, + "learning_rate": 3.523916388817646e-05, + "loss": 0.5584, + "step": 23071 + }, + { + "epoch": 0.89, + "learning_rate": 3.523852269812773e-05, + "loss": 0.5468, + "step": 23072 + }, + { + "epoch": 0.89, + "learning_rate": 3.5237881508078996e-05, + "loss": 0.4071, + "step": 23073 + }, + { + "epoch": 0.89, + "learning_rate": 3.523724031803027e-05, + "loss": 0.4908, + "step": 23074 + }, + { + "epoch": 0.89, + "learning_rate": 3.523659912798153e-05, + "loss": 0.5191, + "step": 23075 + }, + { + "epoch": 0.89, + "learning_rate": 3.5235957937932805e-05, + "loss": 0.4699, + "step": 23076 + }, + { + "epoch": 0.89, + "learning_rate": 3.523531674788407e-05, + "loss": 0.5917, + "step": 23077 + }, + { + "epoch": 0.89, + "learning_rate": 3.523467555783535e-05, + "loss": 0.4186, + "step": 23078 + }, + { + "epoch": 0.89, + "learning_rate": 3.523403436778661e-05, + "loss": 0.4506, + "step": 23079 + }, + { + "epoch": 0.89, + "learning_rate": 3.5233393177737885e-05, + "loss": 0.5047, + "step": 23080 + }, + { + "epoch": 0.89, + "learning_rate": 3.523275198768916e-05, + "loss": 0.5178, + "step": 23081 + }, + { + "epoch": 0.89, + "learning_rate": 3.523211079764042e-05, + "loss": 0.5224, + "step": 23082 + }, + { + "epoch": 0.89, + "learning_rate": 3.523146960759169e-05, + "loss": 0.5161, + "step": 23083 + }, + { + "epoch": 0.89, + "learning_rate": 3.523082841754296e-05, + "loss": 0.5665, + "step": 23084 + }, + { + "epoch": 0.89, + "learning_rate": 3.523018722749423e-05, + "loss": 0.5386, + "step": 23085 + }, + { + "epoch": 0.89, + "learning_rate": 3.52295460374455e-05, + "loss": 0.5782, + "step": 23086 + }, + { + "epoch": 0.89, + "learning_rate": 3.522890484739677e-05, + "loss": 0.413, + "step": 23087 + }, + { + "epoch": 0.89, + "learning_rate": 3.522826365734804e-05, + "loss": 0.435, + "step": 23088 + }, + { + "epoch": 0.89, + "learning_rate": 3.522762246729931e-05, + "loss": 0.5381, + "step": 23089 + }, + { + "epoch": 0.89, + "learning_rate": 3.522698127725058e-05, + "loss": 0.5883, + "step": 23090 + }, + { + "epoch": 0.89, + "learning_rate": 3.522634008720185e-05, + "loss": 0.5251, + "step": 23091 + }, + { + "epoch": 0.89, + "learning_rate": 3.522569889715312e-05, + "loss": 0.3864, + "step": 23092 + }, + { + "epoch": 0.89, + "learning_rate": 3.522505770710438e-05, + "loss": 0.5912, + "step": 23093 + }, + { + "epoch": 0.89, + "learning_rate": 3.5224416517055655e-05, + "loss": 0.4757, + "step": 23094 + }, + { + "epoch": 0.89, + "learning_rate": 3.522377532700693e-05, + "loss": 0.5745, + "step": 23095 + }, + { + "epoch": 0.89, + "learning_rate": 3.52231341369582e-05, + "loss": 0.5078, + "step": 23096 + }, + { + "epoch": 0.89, + "learning_rate": 3.5222492946909463e-05, + "loss": 0.5496, + "step": 23097 + }, + { + "epoch": 0.89, + "learning_rate": 3.5221851756860735e-05, + "loss": 0.325, + "step": 23098 + }, + { + "epoch": 0.89, + "learning_rate": 3.522121056681201e-05, + "loss": 0.4322, + "step": 23099 + }, + { + "epoch": 0.89, + "learning_rate": 3.522056937676327e-05, + "loss": 0.5168, + "step": 23100 + }, + { + "epoch": 0.89, + "learning_rate": 3.5219928186714544e-05, + "loss": 0.4449, + "step": 23101 + }, + { + "epoch": 0.89, + "learning_rate": 3.5219286996665815e-05, + "loss": 0.4748, + "step": 23102 + }, + { + "epoch": 0.89, + "learning_rate": 3.521864580661709e-05, + "loss": 0.4824, + "step": 23103 + }, + { + "epoch": 0.89, + "learning_rate": 3.521800461656835e-05, + "loss": 0.4731, + "step": 23104 + }, + { + "epoch": 0.89, + "learning_rate": 3.5217363426519624e-05, + "loss": 0.4599, + "step": 23105 + }, + { + "epoch": 0.89, + "learning_rate": 3.521672223647089e-05, + "loss": 0.5195, + "step": 23106 + }, + { + "epoch": 0.89, + "learning_rate": 3.521608104642216e-05, + "loss": 0.5245, + "step": 23107 + }, + { + "epoch": 0.89, + "learning_rate": 3.521543985637343e-05, + "loss": 0.5268, + "step": 23108 + }, + { + "epoch": 0.89, + "learning_rate": 3.52147986663247e-05, + "loss": 0.5762, + "step": 23109 + }, + { + "epoch": 0.89, + "learning_rate": 3.521415747627597e-05, + "loss": 0.4971, + "step": 23110 + }, + { + "epoch": 0.89, + "learning_rate": 3.521351628622724e-05, + "loss": 0.5675, + "step": 23111 + }, + { + "epoch": 0.89, + "learning_rate": 3.521287509617851e-05, + "loss": 0.5204, + "step": 23112 + }, + { + "epoch": 0.89, + "learning_rate": 3.521223390612978e-05, + "loss": 0.4787, + "step": 23113 + }, + { + "epoch": 0.89, + "learning_rate": 3.521159271608105e-05, + "loss": 0.4588, + "step": 23114 + }, + { + "epoch": 0.89, + "learning_rate": 3.5210951526032314e-05, + "loss": 0.4118, + "step": 23115 + }, + { + "epoch": 0.89, + "learning_rate": 3.5210310335983585e-05, + "loss": 0.4636, + "step": 23116 + }, + { + "epoch": 0.89, + "learning_rate": 3.520966914593486e-05, + "loss": 0.4524, + "step": 23117 + }, + { + "epoch": 0.89, + "learning_rate": 3.520902795588613e-05, + "loss": 0.4548, + "step": 23118 + }, + { + "epoch": 0.89, + "learning_rate": 3.52083867658374e-05, + "loss": 0.4923, + "step": 23119 + }, + { + "epoch": 0.89, + "learning_rate": 3.5207745575788665e-05, + "loss": 0.4298, + "step": 23120 + }, + { + "epoch": 0.89, + "learning_rate": 3.520710438573994e-05, + "loss": 0.4741, + "step": 23121 + }, + { + "epoch": 0.89, + "learning_rate": 3.52064631956912e-05, + "loss": 0.4741, + "step": 23122 + }, + { + "epoch": 0.89, + "learning_rate": 3.5205822005642474e-05, + "loss": 0.4915, + "step": 23123 + }, + { + "epoch": 0.89, + "learning_rate": 3.520518081559374e-05, + "loss": 0.5901, + "step": 23124 + }, + { + "epoch": 0.89, + "learning_rate": 3.520453962554501e-05, + "loss": 0.5569, + "step": 23125 + }, + { + "epoch": 0.89, + "learning_rate": 3.520389843549628e-05, + "loss": 0.5108, + "step": 23126 + }, + { + "epoch": 0.89, + "learning_rate": 3.5203257245447554e-05, + "loss": 0.6174, + "step": 23127 + }, + { + "epoch": 0.89, + "learning_rate": 3.5202616055398826e-05, + "loss": 0.3811, + "step": 23128 + }, + { + "epoch": 0.89, + "learning_rate": 3.520197486535009e-05, + "loss": 0.6184, + "step": 23129 + }, + { + "epoch": 0.89, + "learning_rate": 3.520133367530136e-05, + "loss": 0.3436, + "step": 23130 + }, + { + "epoch": 0.89, + "learning_rate": 3.520069248525263e-05, + "loss": 0.539, + "step": 23131 + }, + { + "epoch": 0.89, + "learning_rate": 3.52000512952039e-05, + "loss": 0.6045, + "step": 23132 + }, + { + "epoch": 0.89, + "learning_rate": 3.5199410105155164e-05, + "loss": 0.526, + "step": 23133 + }, + { + "epoch": 0.89, + "learning_rate": 3.519876891510644e-05, + "loss": 0.48, + "step": 23134 + }, + { + "epoch": 0.89, + "learning_rate": 3.5198127725057714e-05, + "loss": 0.4864, + "step": 23135 + }, + { + "epoch": 0.89, + "learning_rate": 3.519748653500898e-05, + "loss": 0.5664, + "step": 23136 + }, + { + "epoch": 0.89, + "learning_rate": 3.519684534496025e-05, + "loss": 0.4419, + "step": 23137 + }, + { + "epoch": 0.89, + "learning_rate": 3.5196204154911516e-05, + "loss": 0.5593, + "step": 23138 + }, + { + "epoch": 0.89, + "learning_rate": 3.519556296486279e-05, + "loss": 0.5422, + "step": 23139 + }, + { + "epoch": 0.89, + "learning_rate": 3.519492177481405e-05, + "loss": 0.5642, + "step": 23140 + }, + { + "epoch": 0.89, + "learning_rate": 3.5194280584765324e-05, + "loss": 0.5439, + "step": 23141 + }, + { + "epoch": 0.89, + "learning_rate": 3.5193639394716596e-05, + "loss": 0.5044, + "step": 23142 + }, + { + "epoch": 0.89, + "learning_rate": 3.519299820466787e-05, + "loss": 0.5304, + "step": 23143 + }, + { + "epoch": 0.89, + "learning_rate": 3.519235701461914e-05, + "loss": 0.5009, + "step": 23144 + }, + { + "epoch": 0.89, + "learning_rate": 3.5191715824570404e-05, + "loss": 0.4354, + "step": 23145 + }, + { + "epoch": 0.89, + "learning_rate": 3.5191074634521676e-05, + "loss": 0.5251, + "step": 23146 + }, + { + "epoch": 0.89, + "learning_rate": 3.519043344447294e-05, + "loss": 0.4542, + "step": 23147 + }, + { + "epoch": 0.89, + "learning_rate": 3.518979225442421e-05, + "loss": 0.5514, + "step": 23148 + }, + { + "epoch": 0.89, + "learning_rate": 3.518915106437548e-05, + "loss": 0.5093, + "step": 23149 + }, + { + "epoch": 0.89, + "learning_rate": 3.5188509874326756e-05, + "loss": 0.6223, + "step": 23150 + }, + { + "epoch": 0.89, + "learning_rate": 3.518786868427802e-05, + "loss": 0.4717, + "step": 23151 + }, + { + "epoch": 0.89, + "learning_rate": 3.518722749422929e-05, + "loss": 0.5215, + "step": 23152 + }, + { + "epoch": 0.89, + "learning_rate": 3.5186586304180564e-05, + "loss": 0.4851, + "step": 23153 + }, + { + "epoch": 0.89, + "learning_rate": 3.518594511413183e-05, + "loss": 0.3916, + "step": 23154 + }, + { + "epoch": 0.89, + "learning_rate": 3.51853039240831e-05, + "loss": 0.5252, + "step": 23155 + }, + { + "epoch": 0.89, + "learning_rate": 3.5184662734034366e-05, + "loss": 0.3829, + "step": 23156 + }, + { + "epoch": 0.89, + "learning_rate": 3.518402154398564e-05, + "loss": 0.5286, + "step": 23157 + }, + { + "epoch": 0.89, + "learning_rate": 3.518338035393691e-05, + "loss": 0.4144, + "step": 23158 + }, + { + "epoch": 0.89, + "learning_rate": 3.518273916388818e-05, + "loss": 0.5162, + "step": 23159 + }, + { + "epoch": 0.89, + "learning_rate": 3.5182097973839446e-05, + "loss": 0.4888, + "step": 23160 + }, + { + "epoch": 0.89, + "learning_rate": 3.518145678379072e-05, + "loss": 0.5326, + "step": 23161 + }, + { + "epoch": 0.89, + "learning_rate": 3.518081559374199e-05, + "loss": 0.501, + "step": 23162 + }, + { + "epoch": 0.89, + "learning_rate": 3.5180174403693254e-05, + "loss": 0.4656, + "step": 23163 + }, + { + "epoch": 0.89, + "learning_rate": 3.5179533213644526e-05, + "loss": 0.4687, + "step": 23164 + }, + { + "epoch": 0.89, + "learning_rate": 3.517889202359579e-05, + "loss": 0.4698, + "step": 23165 + }, + { + "epoch": 0.89, + "learning_rate": 3.517825083354706e-05, + "loss": 0.5151, + "step": 23166 + }, + { + "epoch": 0.89, + "learning_rate": 3.5177609643498335e-05, + "loss": 0.5718, + "step": 23167 + }, + { + "epoch": 0.89, + "learning_rate": 3.5176968453449606e-05, + "loss": 0.5273, + "step": 23168 + }, + { + "epoch": 0.89, + "learning_rate": 3.517632726340087e-05, + "loss": 0.5299, + "step": 23169 + }, + { + "epoch": 0.89, + "learning_rate": 3.517568607335214e-05, + "loss": 0.4184, + "step": 23170 + }, + { + "epoch": 0.89, + "learning_rate": 3.5175044883303415e-05, + "loss": 0.5136, + "step": 23171 + }, + { + "epoch": 0.89, + "learning_rate": 3.517440369325468e-05, + "loss": 0.4398, + "step": 23172 + }, + { + "epoch": 0.89, + "learning_rate": 3.517376250320595e-05, + "loss": 0.4204, + "step": 23173 + }, + { + "epoch": 0.89, + "learning_rate": 3.517312131315722e-05, + "loss": 0.4792, + "step": 23174 + }, + { + "epoch": 0.89, + "learning_rate": 3.5172480123108495e-05, + "loss": 0.3965, + "step": 23175 + }, + { + "epoch": 0.89, + "learning_rate": 3.517183893305976e-05, + "loss": 0.4091, + "step": 23176 + }, + { + "epoch": 0.89, + "learning_rate": 3.517119774301103e-05, + "loss": 0.463, + "step": 23177 + }, + { + "epoch": 0.89, + "learning_rate": 3.5170556552962296e-05, + "loss": 0.5629, + "step": 23178 + }, + { + "epoch": 0.89, + "learning_rate": 3.516991536291357e-05, + "loss": 0.4531, + "step": 23179 + }, + { + "epoch": 0.89, + "learning_rate": 3.516927417286484e-05, + "loss": 0.3626, + "step": 23180 + }, + { + "epoch": 0.89, + "learning_rate": 3.5168632982816105e-05, + "loss": 0.4919, + "step": 23181 + }, + { + "epoch": 0.89, + "learning_rate": 3.5167991792767376e-05, + "loss": 0.4476, + "step": 23182 + }, + { + "epoch": 0.89, + "learning_rate": 3.516735060271865e-05, + "loss": 0.5278, + "step": 23183 + }, + { + "epoch": 0.89, + "learning_rate": 3.516670941266992e-05, + "loss": 0.4982, + "step": 23184 + }, + { + "epoch": 0.89, + "learning_rate": 3.5166068222621185e-05, + "loss": 0.4313, + "step": 23185 + }, + { + "epoch": 0.89, + "learning_rate": 3.5165427032572456e-05, + "loss": 0.4321, + "step": 23186 + }, + { + "epoch": 0.89, + "learning_rate": 3.516478584252372e-05, + "loss": 0.466, + "step": 23187 + }, + { + "epoch": 0.89, + "learning_rate": 3.516414465247499e-05, + "loss": 0.479, + "step": 23188 + }, + { + "epoch": 0.89, + "learning_rate": 3.5163503462426265e-05, + "loss": 0.5728, + "step": 23189 + }, + { + "epoch": 0.89, + "learning_rate": 3.5162862272377537e-05, + "loss": 0.6758, + "step": 23190 + }, + { + "epoch": 0.89, + "learning_rate": 3.516222108232881e-05, + "loss": 0.5328, + "step": 23191 + }, + { + "epoch": 0.89, + "learning_rate": 3.516157989228007e-05, + "loss": 0.4289, + "step": 23192 + }, + { + "epoch": 0.89, + "learning_rate": 3.5160938702231345e-05, + "loss": 0.5732, + "step": 23193 + }, + { + "epoch": 0.89, + "learning_rate": 3.516029751218261e-05, + "loss": 0.5387, + "step": 23194 + }, + { + "epoch": 0.89, + "learning_rate": 3.515965632213388e-05, + "loss": 0.4929, + "step": 23195 + }, + { + "epoch": 0.89, + "learning_rate": 3.5159015132085147e-05, + "loss": 0.469, + "step": 23196 + }, + { + "epoch": 0.89, + "learning_rate": 3.515837394203642e-05, + "loss": 0.47, + "step": 23197 + }, + { + "epoch": 0.89, + "learning_rate": 3.515773275198769e-05, + "loss": 0.425, + "step": 23198 + }, + { + "epoch": 0.89, + "learning_rate": 3.515709156193896e-05, + "loss": 0.468, + "step": 23199 + }, + { + "epoch": 0.89, + "learning_rate": 3.5156450371890233e-05, + "loss": 0.4946, + "step": 23200 + }, + { + "epoch": 0.89, + "learning_rate": 3.51558091818415e-05, + "loss": 0.4842, + "step": 23201 + }, + { + "epoch": 0.89, + "learning_rate": 3.515516799179277e-05, + "loss": 0.5246, + "step": 23202 + }, + { + "epoch": 0.89, + "learning_rate": 3.5154526801744035e-05, + "loss": 0.3943, + "step": 23203 + }, + { + "epoch": 0.89, + "learning_rate": 3.515388561169531e-05, + "loss": 0.5002, + "step": 23204 + }, + { + "epoch": 0.89, + "learning_rate": 3.515324442164658e-05, + "loss": 0.3873, + "step": 23205 + }, + { + "epoch": 0.89, + "learning_rate": 3.515260323159785e-05, + "loss": 0.4608, + "step": 23206 + }, + { + "epoch": 0.89, + "learning_rate": 3.515196204154912e-05, + "loss": 0.4494, + "step": 23207 + }, + { + "epoch": 0.89, + "learning_rate": 3.515132085150039e-05, + "loss": 0.4257, + "step": 23208 + }, + { + "epoch": 0.89, + "learning_rate": 3.515067966145166e-05, + "loss": 0.558, + "step": 23209 + }, + { + "epoch": 0.89, + "learning_rate": 3.5150038471402923e-05, + "loss": 0.3853, + "step": 23210 + }, + { + "epoch": 0.89, + "learning_rate": 3.5149397281354195e-05, + "loss": 0.6376, + "step": 23211 + }, + { + "epoch": 0.89, + "learning_rate": 3.514875609130546e-05, + "loss": 0.4276, + "step": 23212 + }, + { + "epoch": 0.89, + "learning_rate": 3.514811490125673e-05, + "loss": 0.4066, + "step": 23213 + }, + { + "epoch": 0.89, + "learning_rate": 3.5147473711208004e-05, + "loss": 0.5052, + "step": 23214 + }, + { + "epoch": 0.89, + "learning_rate": 3.5146832521159275e-05, + "loss": 0.5998, + "step": 23215 + }, + { + "epoch": 0.89, + "learning_rate": 3.514619133111055e-05, + "loss": 0.4125, + "step": 23216 + }, + { + "epoch": 0.89, + "learning_rate": 3.514555014106181e-05, + "loss": 0.5148, + "step": 23217 + }, + { + "epoch": 0.89, + "learning_rate": 3.5144908951013084e-05, + "loss": 0.5265, + "step": 23218 + }, + { + "epoch": 0.89, + "learning_rate": 3.514426776096435e-05, + "loss": 0.467, + "step": 23219 + }, + { + "epoch": 0.89, + "learning_rate": 3.514362657091562e-05, + "loss": 0.4738, + "step": 23220 + }, + { + "epoch": 0.89, + "learning_rate": 3.5142985380866885e-05, + "loss": 0.4442, + "step": 23221 + }, + { + "epoch": 0.89, + "learning_rate": 3.5142344190818164e-05, + "loss": 0.3991, + "step": 23222 + }, + { + "epoch": 0.89, + "learning_rate": 3.514170300076943e-05, + "loss": 0.573, + "step": 23223 + }, + { + "epoch": 0.89, + "learning_rate": 3.51410618107207e-05, + "loss": 0.4646, + "step": 23224 + }, + { + "epoch": 0.89, + "learning_rate": 3.514042062067197e-05, + "loss": 0.4205, + "step": 23225 + }, + { + "epoch": 0.89, + "learning_rate": 3.513977943062324e-05, + "loss": 0.5885, + "step": 23226 + }, + { + "epoch": 0.89, + "learning_rate": 3.513913824057451e-05, + "loss": 0.4824, + "step": 23227 + }, + { + "epoch": 0.89, + "learning_rate": 3.5138497050525774e-05, + "loss": 0.5632, + "step": 23228 + }, + { + "epoch": 0.89, + "learning_rate": 3.5137855860477045e-05, + "loss": 0.51, + "step": 23229 + }, + { + "epoch": 0.89, + "learning_rate": 3.513721467042832e-05, + "loss": 0.5281, + "step": 23230 + }, + { + "epoch": 0.89, + "learning_rate": 3.513657348037959e-05, + "loss": 0.596, + "step": 23231 + }, + { + "epoch": 0.89, + "learning_rate": 3.5135932290330854e-05, + "loss": 0.3845, + "step": 23232 + }, + { + "epoch": 0.89, + "learning_rate": 3.5135291100282126e-05, + "loss": 0.4138, + "step": 23233 + }, + { + "epoch": 0.89, + "learning_rate": 3.51346499102334e-05, + "loss": 0.5394, + "step": 23234 + }, + { + "epoch": 0.89, + "learning_rate": 3.513400872018466e-05, + "loss": 0.6017, + "step": 23235 + }, + { + "epoch": 0.89, + "learning_rate": 3.5133367530135934e-05, + "loss": 0.4483, + "step": 23236 + }, + { + "epoch": 0.89, + "learning_rate": 3.51327263400872e-05, + "loss": 0.5395, + "step": 23237 + }, + { + "epoch": 0.89, + "learning_rate": 3.513208515003848e-05, + "loss": 0.3532, + "step": 23238 + }, + { + "epoch": 0.89, + "learning_rate": 3.513144395998974e-05, + "loss": 0.4019, + "step": 23239 + }, + { + "epoch": 0.89, + "learning_rate": 3.5130802769941014e-05, + "loss": 0.5968, + "step": 23240 + }, + { + "epoch": 0.89, + "learning_rate": 3.5130161579892286e-05, + "loss": 0.6697, + "step": 23241 + }, + { + "epoch": 0.89, + "learning_rate": 3.512952038984355e-05, + "loss": 0.3966, + "step": 23242 + }, + { + "epoch": 0.89, + "learning_rate": 3.512887919979482e-05, + "loss": 0.4252, + "step": 23243 + }, + { + "epoch": 0.89, + "learning_rate": 3.512823800974609e-05, + "loss": 0.4678, + "step": 23244 + }, + { + "epoch": 0.89, + "learning_rate": 3.512759681969736e-05, + "loss": 0.5369, + "step": 23245 + }, + { + "epoch": 0.89, + "learning_rate": 3.512695562964863e-05, + "loss": 0.4178, + "step": 23246 + }, + { + "epoch": 0.89, + "learning_rate": 3.51263144395999e-05, + "loss": 0.5251, + "step": 23247 + }, + { + "epoch": 0.89, + "learning_rate": 3.512567324955117e-05, + "loss": 0.4982, + "step": 23248 + }, + { + "epoch": 0.89, + "learning_rate": 3.512503205950244e-05, + "loss": 0.5465, + "step": 23249 + }, + { + "epoch": 0.89, + "learning_rate": 3.512439086945371e-05, + "loss": 0.4539, + "step": 23250 + }, + { + "epoch": 0.89, + "learning_rate": 3.5123749679404976e-05, + "loss": 0.3857, + "step": 23251 + }, + { + "epoch": 0.89, + "learning_rate": 3.512310848935625e-05, + "loss": 0.5323, + "step": 23252 + }, + { + "epoch": 0.89, + "learning_rate": 3.512246729930751e-05, + "loss": 0.4605, + "step": 23253 + }, + { + "epoch": 0.89, + "learning_rate": 3.5121826109258784e-05, + "loss": 0.4663, + "step": 23254 + }, + { + "epoch": 0.89, + "learning_rate": 3.5121184919210056e-05, + "loss": 0.4543, + "step": 23255 + }, + { + "epoch": 0.89, + "learning_rate": 3.512054372916133e-05, + "loss": 0.4762, + "step": 23256 + }, + { + "epoch": 0.89, + "learning_rate": 3.511990253911259e-05, + "loss": 0.5037, + "step": 23257 + }, + { + "epoch": 0.89, + "learning_rate": 3.5119261349063864e-05, + "loss": 0.4269, + "step": 23258 + }, + { + "epoch": 0.89, + "learning_rate": 3.5118620159015136e-05, + "loss": 0.4301, + "step": 23259 + }, + { + "epoch": 0.89, + "learning_rate": 3.51179789689664e-05, + "loss": 0.4523, + "step": 23260 + }, + { + "epoch": 0.89, + "learning_rate": 3.511733777891767e-05, + "loss": 0.493, + "step": 23261 + }, + { + "epoch": 0.89, + "learning_rate": 3.5116696588868944e-05, + "loss": 0.5217, + "step": 23262 + }, + { + "epoch": 0.89, + "learning_rate": 3.5116055398820216e-05, + "loss": 0.425, + "step": 23263 + }, + { + "epoch": 0.89, + "learning_rate": 3.511541420877148e-05, + "loss": 0.4355, + "step": 23264 + }, + { + "epoch": 0.89, + "learning_rate": 3.511477301872275e-05, + "loss": 0.7961, + "step": 23265 + }, + { + "epoch": 0.89, + "learning_rate": 3.511413182867402e-05, + "loss": 0.4994, + "step": 23266 + }, + { + "epoch": 0.89, + "learning_rate": 3.511349063862529e-05, + "loss": 0.5416, + "step": 23267 + }, + { + "epoch": 0.89, + "learning_rate": 3.511284944857656e-05, + "loss": 0.3718, + "step": 23268 + }, + { + "epoch": 0.89, + "learning_rate": 3.5112208258527826e-05, + "loss": 0.5269, + "step": 23269 + }, + { + "epoch": 0.89, + "learning_rate": 3.51115670684791e-05, + "loss": 0.4242, + "step": 23270 + }, + { + "epoch": 0.89, + "learning_rate": 3.511092587843037e-05, + "loss": 0.4764, + "step": 23271 + }, + { + "epoch": 0.89, + "learning_rate": 3.511028468838164e-05, + "loss": 0.5845, + "step": 23272 + }, + { + "epoch": 0.89, + "learning_rate": 3.5109643498332906e-05, + "loss": 0.5361, + "step": 23273 + }, + { + "epoch": 0.89, + "learning_rate": 3.510900230828418e-05, + "loss": 0.4754, + "step": 23274 + }, + { + "epoch": 0.89, + "learning_rate": 3.510836111823544e-05, + "loss": 0.5312, + "step": 23275 + }, + { + "epoch": 0.89, + "learning_rate": 3.5107719928186714e-05, + "loss": 0.476, + "step": 23276 + }, + { + "epoch": 0.89, + "learning_rate": 3.5107078738137986e-05, + "loss": 0.5362, + "step": 23277 + }, + { + "epoch": 0.89, + "learning_rate": 3.510643754808926e-05, + "loss": 0.455, + "step": 23278 + }, + { + "epoch": 0.9, + "learning_rate": 3.510579635804053e-05, + "loss": 0.4785, + "step": 23279 + }, + { + "epoch": 0.9, + "learning_rate": 3.5105155167991795e-05, + "loss": 0.5668, + "step": 23280 + }, + { + "epoch": 0.9, + "learning_rate": 3.5104513977943066e-05, + "loss": 0.4041, + "step": 23281 + }, + { + "epoch": 0.9, + "learning_rate": 3.510387278789433e-05, + "loss": 0.474, + "step": 23282 + }, + { + "epoch": 0.9, + "learning_rate": 3.51032315978456e-05, + "loss": 0.4937, + "step": 23283 + }, + { + "epoch": 0.9, + "learning_rate": 3.510259040779687e-05, + "loss": 0.4487, + "step": 23284 + }, + { + "epoch": 0.9, + "learning_rate": 3.510194921774814e-05, + "loss": 0.4824, + "step": 23285 + }, + { + "epoch": 0.9, + "learning_rate": 3.510130802769941e-05, + "loss": 0.5437, + "step": 23286 + }, + { + "epoch": 0.9, + "learning_rate": 3.510066683765068e-05, + "loss": 0.5665, + "step": 23287 + }, + { + "epoch": 0.9, + "learning_rate": 3.5100025647601955e-05, + "loss": 0.3989, + "step": 23288 + }, + { + "epoch": 0.9, + "learning_rate": 3.509938445755322e-05, + "loss": 0.3758, + "step": 23289 + }, + { + "epoch": 0.9, + "learning_rate": 3.509874326750449e-05, + "loss": 0.56, + "step": 23290 + }, + { + "epoch": 0.9, + "learning_rate": 3.5098102077455756e-05, + "loss": 0.3754, + "step": 23291 + }, + { + "epoch": 0.9, + "learning_rate": 3.509746088740703e-05, + "loss": 0.4721, + "step": 23292 + }, + { + "epoch": 0.9, + "learning_rate": 3.509681969735829e-05, + "loss": 0.4642, + "step": 23293 + }, + { + "epoch": 0.9, + "learning_rate": 3.509617850730957e-05, + "loss": 0.468, + "step": 23294 + }, + { + "epoch": 0.9, + "learning_rate": 3.509553731726084e-05, + "loss": 0.5783, + "step": 23295 + }, + { + "epoch": 0.9, + "learning_rate": 3.509489612721211e-05, + "loss": 0.57, + "step": 23296 + }, + { + "epoch": 0.9, + "learning_rate": 3.509425493716338e-05, + "loss": 0.5107, + "step": 23297 + }, + { + "epoch": 0.9, + "learning_rate": 3.5093613747114645e-05, + "loss": 0.4848, + "step": 23298 + }, + { + "epoch": 0.9, + "learning_rate": 3.5092972557065917e-05, + "loss": 0.5167, + "step": 23299 + }, + { + "epoch": 0.9, + "learning_rate": 3.509233136701718e-05, + "loss": 0.5889, + "step": 23300 + }, + { + "epoch": 0.9, + "learning_rate": 3.509169017696845e-05, + "loss": 0.4483, + "step": 23301 + }, + { + "epoch": 0.9, + "learning_rate": 3.5091048986919725e-05, + "loss": 0.4961, + "step": 23302 + }, + { + "epoch": 0.9, + "learning_rate": 3.5090407796870997e-05, + "loss": 0.5486, + "step": 23303 + }, + { + "epoch": 0.9, + "learning_rate": 3.508976660682227e-05, + "loss": 0.5308, + "step": 23304 + }, + { + "epoch": 0.9, + "learning_rate": 3.508912541677353e-05, + "loss": 0.4329, + "step": 23305 + }, + { + "epoch": 0.9, + "learning_rate": 3.5088484226724805e-05, + "loss": 0.5274, + "step": 23306 + }, + { + "epoch": 0.9, + "learning_rate": 3.508784303667607e-05, + "loss": 0.5597, + "step": 23307 + }, + { + "epoch": 0.9, + "learning_rate": 3.508720184662734e-05, + "loss": 0.4735, + "step": 23308 + }, + { + "epoch": 0.9, + "learning_rate": 3.5086560656578607e-05, + "loss": 0.5342, + "step": 23309 + }, + { + "epoch": 0.9, + "learning_rate": 3.5085919466529885e-05, + "loss": 0.4919, + "step": 23310 + }, + { + "epoch": 0.9, + "learning_rate": 3.508527827648115e-05, + "loss": 0.4862, + "step": 23311 + }, + { + "epoch": 0.9, + "learning_rate": 3.508463708643242e-05, + "loss": 0.5251, + "step": 23312 + }, + { + "epoch": 0.9, + "learning_rate": 3.5083995896383693e-05, + "loss": 0.3805, + "step": 23313 + }, + { + "epoch": 0.9, + "learning_rate": 3.508335470633496e-05, + "loss": 0.479, + "step": 23314 + }, + { + "epoch": 0.9, + "learning_rate": 3.508271351628623e-05, + "loss": 0.5134, + "step": 23315 + }, + { + "epoch": 0.9, + "learning_rate": 3.5082072326237495e-05, + "loss": 0.4947, + "step": 23316 + }, + { + "epoch": 0.9, + "learning_rate": 3.508143113618877e-05, + "loss": 0.5461, + "step": 23317 + }, + { + "epoch": 0.9, + "learning_rate": 3.508078994614004e-05, + "loss": 0.4311, + "step": 23318 + }, + { + "epoch": 0.9, + "learning_rate": 3.508014875609131e-05, + "loss": 0.4936, + "step": 23319 + }, + { + "epoch": 0.9, + "learning_rate": 3.5079507566042575e-05, + "loss": 0.4783, + "step": 23320 + }, + { + "epoch": 0.9, + "learning_rate": 3.507886637599385e-05, + "loss": 0.4523, + "step": 23321 + }, + { + "epoch": 0.9, + "learning_rate": 3.507822518594512e-05, + "loss": 0.5281, + "step": 23322 + }, + { + "epoch": 0.9, + "learning_rate": 3.5077583995896383e-05, + "loss": 0.4412, + "step": 23323 + }, + { + "epoch": 0.9, + "learning_rate": 3.5076942805847655e-05, + "loss": 0.4021, + "step": 23324 + }, + { + "epoch": 0.9, + "learning_rate": 3.507630161579892e-05, + "loss": 0.593, + "step": 23325 + }, + { + "epoch": 0.9, + "learning_rate": 3.50756604257502e-05, + "loss": 0.5563, + "step": 23326 + }, + { + "epoch": 0.9, + "learning_rate": 3.5075019235701464e-05, + "loss": 0.5501, + "step": 23327 + }, + { + "epoch": 0.9, + "learning_rate": 3.5074378045652735e-05, + "loss": 0.4617, + "step": 23328 + }, + { + "epoch": 0.9, + "learning_rate": 3.5073736855604e-05, + "loss": 0.4551, + "step": 23329 + }, + { + "epoch": 0.9, + "learning_rate": 3.507309566555527e-05, + "loss": 0.5727, + "step": 23330 + }, + { + "epoch": 0.9, + "learning_rate": 3.5072454475506544e-05, + "loss": 0.4643, + "step": 23331 + }, + { + "epoch": 0.9, + "learning_rate": 3.507181328545781e-05, + "loss": 0.4385, + "step": 23332 + }, + { + "epoch": 0.9, + "learning_rate": 3.507117209540908e-05, + "loss": 0.5703, + "step": 23333 + }, + { + "epoch": 0.9, + "learning_rate": 3.507053090536035e-05, + "loss": 0.5351, + "step": 23334 + }, + { + "epoch": 0.9, + "learning_rate": 3.5069889715311624e-05, + "loss": 0.4407, + "step": 23335 + }, + { + "epoch": 0.9, + "learning_rate": 3.506924852526289e-05, + "loss": 0.6224, + "step": 23336 + }, + { + "epoch": 0.9, + "learning_rate": 3.506860733521416e-05, + "loss": 0.5991, + "step": 23337 + }, + { + "epoch": 0.9, + "learning_rate": 3.5067966145165425e-05, + "loss": 0.5023, + "step": 23338 + }, + { + "epoch": 0.9, + "learning_rate": 3.50673249551167e-05, + "loss": 0.3718, + "step": 23339 + }, + { + "epoch": 0.9, + "learning_rate": 3.506668376506797e-05, + "loss": 0.5376, + "step": 23340 + }, + { + "epoch": 0.9, + "learning_rate": 3.5066042575019234e-05, + "loss": 0.4966, + "step": 23341 + }, + { + "epoch": 0.9, + "learning_rate": 3.5065401384970505e-05, + "loss": 0.4249, + "step": 23342 + }, + { + "epoch": 0.9, + "learning_rate": 3.506476019492178e-05, + "loss": 0.5692, + "step": 23343 + }, + { + "epoch": 0.9, + "learning_rate": 3.506411900487305e-05, + "loss": 0.5365, + "step": 23344 + }, + { + "epoch": 0.9, + "learning_rate": 3.5063477814824314e-05, + "loss": 0.5818, + "step": 23345 + }, + { + "epoch": 0.9, + "learning_rate": 3.5062836624775586e-05, + "loss": 0.5543, + "step": 23346 + }, + { + "epoch": 0.9, + "learning_rate": 3.506219543472685e-05, + "loss": 0.5178, + "step": 23347 + }, + { + "epoch": 0.9, + "learning_rate": 3.506155424467812e-05, + "loss": 0.4669, + "step": 23348 + }, + { + "epoch": 0.9, + "learning_rate": 3.5060913054629394e-05, + "loss": 0.4327, + "step": 23349 + }, + { + "epoch": 0.9, + "learning_rate": 3.5060271864580666e-05, + "loss": 0.4683, + "step": 23350 + }, + { + "epoch": 0.9, + "learning_rate": 3.505963067453194e-05, + "loss": 0.5813, + "step": 23351 + }, + { + "epoch": 0.9, + "learning_rate": 3.50589894844832e-05, + "loss": 0.4702, + "step": 23352 + }, + { + "epoch": 0.9, + "learning_rate": 3.5058348294434474e-05, + "loss": 0.5559, + "step": 23353 + }, + { + "epoch": 0.9, + "learning_rate": 3.505770710438574e-05, + "loss": 0.4771, + "step": 23354 + }, + { + "epoch": 0.9, + "learning_rate": 3.505706591433701e-05, + "loss": 0.4098, + "step": 23355 + }, + { + "epoch": 0.9, + "learning_rate": 3.5056424724288276e-05, + "loss": 0.5709, + "step": 23356 + }, + { + "epoch": 0.9, + "learning_rate": 3.505578353423955e-05, + "loss": 0.6598, + "step": 23357 + }, + { + "epoch": 0.9, + "learning_rate": 3.505514234419082e-05, + "loss": 0.5744, + "step": 23358 + }, + { + "epoch": 0.9, + "learning_rate": 3.505450115414209e-05, + "loss": 0.4889, + "step": 23359 + }, + { + "epoch": 0.9, + "learning_rate": 3.505385996409336e-05, + "loss": 0.4759, + "step": 23360 + }, + { + "epoch": 0.9, + "learning_rate": 3.505321877404463e-05, + "loss": 0.5832, + "step": 23361 + }, + { + "epoch": 0.9, + "learning_rate": 3.50525775839959e-05, + "loss": 0.4362, + "step": 23362 + }, + { + "epoch": 0.9, + "learning_rate": 3.5051936393947164e-05, + "loss": 0.4451, + "step": 23363 + }, + { + "epoch": 0.9, + "learning_rate": 3.5051295203898436e-05, + "loss": 0.4914, + "step": 23364 + }, + { + "epoch": 0.9, + "learning_rate": 3.505065401384971e-05, + "loss": 0.4824, + "step": 23365 + }, + { + "epoch": 0.9, + "learning_rate": 3.505001282380098e-05, + "loss": 0.493, + "step": 23366 + }, + { + "epoch": 0.9, + "learning_rate": 3.504937163375225e-05, + "loss": 0.4735, + "step": 23367 + }, + { + "epoch": 0.9, + "learning_rate": 3.5048730443703516e-05, + "loss": 0.643, + "step": 23368 + }, + { + "epoch": 0.9, + "learning_rate": 3.504808925365479e-05, + "loss": 0.5601, + "step": 23369 + }, + { + "epoch": 0.9, + "learning_rate": 3.504744806360605e-05, + "loss": 0.5271, + "step": 23370 + }, + { + "epoch": 0.9, + "learning_rate": 3.5046806873557324e-05, + "loss": 0.5435, + "step": 23371 + }, + { + "epoch": 0.9, + "learning_rate": 3.504616568350859e-05, + "loss": 0.6014, + "step": 23372 + }, + { + "epoch": 0.9, + "learning_rate": 3.504552449345986e-05, + "loss": 0.4348, + "step": 23373 + }, + { + "epoch": 0.9, + "learning_rate": 3.504488330341113e-05, + "loss": 0.5006, + "step": 23374 + }, + { + "epoch": 0.9, + "learning_rate": 3.5044242113362404e-05, + "loss": 0.5259, + "step": 23375 + }, + { + "epoch": 0.9, + "learning_rate": 3.5043600923313676e-05, + "loss": 0.4946, + "step": 23376 + }, + { + "epoch": 0.9, + "learning_rate": 3.504295973326494e-05, + "loss": 0.5873, + "step": 23377 + }, + { + "epoch": 0.9, + "learning_rate": 3.504231854321621e-05, + "loss": 0.3948, + "step": 23378 + }, + { + "epoch": 0.9, + "learning_rate": 3.504167735316748e-05, + "loss": 0.4895, + "step": 23379 + }, + { + "epoch": 0.9, + "learning_rate": 3.504103616311875e-05, + "loss": 0.4493, + "step": 23380 + }, + { + "epoch": 0.9, + "learning_rate": 3.5040394973070014e-05, + "loss": 0.4257, + "step": 23381 + }, + { + "epoch": 0.9, + "learning_rate": 3.503975378302129e-05, + "loss": 0.4001, + "step": 23382 + }, + { + "epoch": 0.9, + "learning_rate": 3.503911259297256e-05, + "loss": 0.4428, + "step": 23383 + }, + { + "epoch": 0.9, + "learning_rate": 3.503847140292383e-05, + "loss": 0.5374, + "step": 23384 + }, + { + "epoch": 0.9, + "learning_rate": 3.50378302128751e-05, + "loss": 0.4114, + "step": 23385 + }, + { + "epoch": 0.9, + "learning_rate": 3.5037189022826366e-05, + "loss": 0.5881, + "step": 23386 + }, + { + "epoch": 0.9, + "learning_rate": 3.503654783277764e-05, + "loss": 0.5245, + "step": 23387 + }, + { + "epoch": 0.9, + "learning_rate": 3.50359066427289e-05, + "loss": 0.4524, + "step": 23388 + }, + { + "epoch": 0.9, + "learning_rate": 3.5035265452680174e-05, + "loss": 0.4426, + "step": 23389 + }, + { + "epoch": 0.9, + "learning_rate": 3.5034624262631446e-05, + "loss": 0.574, + "step": 23390 + }, + { + "epoch": 0.9, + "learning_rate": 3.503398307258272e-05, + "loss": 0.5932, + "step": 23391 + }, + { + "epoch": 0.9, + "learning_rate": 3.503334188253398e-05, + "loss": 0.4299, + "step": 23392 + }, + { + "epoch": 0.9, + "learning_rate": 3.5032700692485255e-05, + "loss": 0.5485, + "step": 23393 + }, + { + "epoch": 0.9, + "learning_rate": 3.5032059502436526e-05, + "loss": 0.4519, + "step": 23394 + }, + { + "epoch": 0.9, + "learning_rate": 3.503141831238779e-05, + "loss": 0.575, + "step": 23395 + }, + { + "epoch": 0.9, + "learning_rate": 3.503077712233906e-05, + "loss": 0.5003, + "step": 23396 + }, + { + "epoch": 0.9, + "learning_rate": 3.503013593229033e-05, + "loss": 0.504, + "step": 23397 + }, + { + "epoch": 0.9, + "learning_rate": 3.5029494742241606e-05, + "loss": 0.4747, + "step": 23398 + }, + { + "epoch": 0.9, + "learning_rate": 3.502885355219287e-05, + "loss": 0.6083, + "step": 23399 + }, + { + "epoch": 0.9, + "learning_rate": 3.502821236214414e-05, + "loss": 0.5496, + "step": 23400 + }, + { + "epoch": 0.9, + "learning_rate": 3.502757117209541e-05, + "loss": 0.5507, + "step": 23401 + }, + { + "epoch": 0.9, + "learning_rate": 3.502692998204668e-05, + "loss": 0.5284, + "step": 23402 + }, + { + "epoch": 0.9, + "learning_rate": 3.502628879199795e-05, + "loss": 0.4885, + "step": 23403 + }, + { + "epoch": 0.9, + "learning_rate": 3.5025647601949216e-05, + "loss": 0.4547, + "step": 23404 + }, + { + "epoch": 0.9, + "learning_rate": 3.502500641190049e-05, + "loss": 0.4282, + "step": 23405 + }, + { + "epoch": 0.9, + "learning_rate": 3.502436522185176e-05, + "loss": 0.4203, + "step": 23406 + }, + { + "epoch": 0.9, + "learning_rate": 3.502372403180303e-05, + "loss": 0.4862, + "step": 23407 + }, + { + "epoch": 0.9, + "learning_rate": 3.5023082841754296e-05, + "loss": 0.4763, + "step": 23408 + }, + { + "epoch": 0.9, + "learning_rate": 3.502244165170557e-05, + "loss": 0.5737, + "step": 23409 + }, + { + "epoch": 0.9, + "learning_rate": 3.502180046165683e-05, + "loss": 0.3697, + "step": 23410 + }, + { + "epoch": 0.9, + "learning_rate": 3.5021159271608105e-05, + "loss": 0.5724, + "step": 23411 + }, + { + "epoch": 0.9, + "learning_rate": 3.5020518081559377e-05, + "loss": 0.4099, + "step": 23412 + }, + { + "epoch": 0.9, + "learning_rate": 3.501987689151064e-05, + "loss": 0.4113, + "step": 23413 + }, + { + "epoch": 0.9, + "learning_rate": 3.501923570146191e-05, + "loss": 0.4998, + "step": 23414 + }, + { + "epoch": 0.9, + "learning_rate": 3.5018594511413185e-05, + "loss": 0.402, + "step": 23415 + }, + { + "epoch": 0.9, + "learning_rate": 3.501795332136446e-05, + "loss": 0.4903, + "step": 23416 + }, + { + "epoch": 0.9, + "learning_rate": 3.501731213131572e-05, + "loss": 0.5168, + "step": 23417 + }, + { + "epoch": 0.9, + "learning_rate": 3.501667094126699e-05, + "loss": 0.551, + "step": 23418 + }, + { + "epoch": 0.9, + "learning_rate": 3.501602975121826e-05, + "loss": 0.3584, + "step": 23419 + }, + { + "epoch": 0.9, + "learning_rate": 3.501538856116953e-05, + "loss": 0.4564, + "step": 23420 + }, + { + "epoch": 0.9, + "learning_rate": 3.50147473711208e-05, + "loss": 0.5599, + "step": 23421 + }, + { + "epoch": 0.9, + "learning_rate": 3.501410618107207e-05, + "loss": 0.3959, + "step": 23422 + }, + { + "epoch": 0.9, + "learning_rate": 3.5013464991023345e-05, + "loss": 0.4931, + "step": 23423 + }, + { + "epoch": 0.9, + "learning_rate": 3.501282380097461e-05, + "loss": 0.5861, + "step": 23424 + }, + { + "epoch": 0.9, + "learning_rate": 3.501218261092588e-05, + "loss": 0.4264, + "step": 23425 + }, + { + "epoch": 0.9, + "learning_rate": 3.501154142087715e-05, + "loss": 0.4506, + "step": 23426 + }, + { + "epoch": 0.9, + "learning_rate": 3.501090023082842e-05, + "loss": 0.4382, + "step": 23427 + }, + { + "epoch": 0.9, + "learning_rate": 3.501025904077969e-05, + "loss": 0.5112, + "step": 23428 + }, + { + "epoch": 0.9, + "learning_rate": 3.5009617850730955e-05, + "loss": 0.5221, + "step": 23429 + }, + { + "epoch": 0.9, + "learning_rate": 3.500897666068223e-05, + "loss": 0.4233, + "step": 23430 + }, + { + "epoch": 0.9, + "learning_rate": 3.50083354706335e-05, + "loss": 0.5901, + "step": 23431 + }, + { + "epoch": 0.9, + "learning_rate": 3.500769428058477e-05, + "loss": 0.4703, + "step": 23432 + }, + { + "epoch": 0.9, + "learning_rate": 3.5007053090536035e-05, + "loss": 0.4138, + "step": 23433 + }, + { + "epoch": 0.9, + "learning_rate": 3.500641190048731e-05, + "loss": 0.4596, + "step": 23434 + }, + { + "epoch": 0.9, + "learning_rate": 3.500577071043857e-05, + "loss": 0.442, + "step": 23435 + }, + { + "epoch": 0.9, + "learning_rate": 3.5005129520389844e-05, + "loss": 0.5431, + "step": 23436 + }, + { + "epoch": 0.9, + "learning_rate": 3.5004488330341115e-05, + "loss": 0.4902, + "step": 23437 + }, + { + "epoch": 0.9, + "learning_rate": 3.500384714029239e-05, + "loss": 0.4638, + "step": 23438 + }, + { + "epoch": 0.9, + "learning_rate": 3.500320595024366e-05, + "loss": 0.5338, + "step": 23439 + }, + { + "epoch": 0.9, + "learning_rate": 3.5002564760194924e-05, + "loss": 0.5446, + "step": 23440 + }, + { + "epoch": 0.9, + "learning_rate": 3.5001923570146195e-05, + "loss": 0.4697, + "step": 23441 + }, + { + "epoch": 0.9, + "learning_rate": 3.500128238009746e-05, + "loss": 0.4974, + "step": 23442 + }, + { + "epoch": 0.9, + "learning_rate": 3.500064119004873e-05, + "loss": 0.5344, + "step": 23443 + }, + { + "epoch": 0.9, + "learning_rate": 3.5e-05, + "loss": 0.4287, + "step": 23444 + }, + { + "epoch": 0.9, + "learning_rate": 3.499935880995127e-05, + "loss": 0.4938, + "step": 23445 + }, + { + "epoch": 0.9, + "learning_rate": 3.499871761990254e-05, + "loss": 0.4096, + "step": 23446 + }, + { + "epoch": 0.9, + "learning_rate": 3.499807642985381e-05, + "loss": 0.5736, + "step": 23447 + }, + { + "epoch": 0.9, + "learning_rate": 3.4997435239805084e-05, + "loss": 0.4979, + "step": 23448 + }, + { + "epoch": 0.9, + "learning_rate": 3.499679404975635e-05, + "loss": 0.5458, + "step": 23449 + }, + { + "epoch": 0.9, + "learning_rate": 3.499615285970762e-05, + "loss": 0.4644, + "step": 23450 + }, + { + "epoch": 0.9, + "learning_rate": 3.4995511669658885e-05, + "loss": 0.4972, + "step": 23451 + }, + { + "epoch": 0.9, + "learning_rate": 3.499487047961016e-05, + "loss": 0.4655, + "step": 23452 + }, + { + "epoch": 0.9, + "learning_rate": 3.499422928956142e-05, + "loss": 0.4513, + "step": 23453 + }, + { + "epoch": 0.9, + "learning_rate": 3.49935880995127e-05, + "loss": 0.4712, + "step": 23454 + }, + { + "epoch": 0.9, + "learning_rate": 3.499294690946397e-05, + "loss": 0.3723, + "step": 23455 + }, + { + "epoch": 0.9, + "learning_rate": 3.499230571941524e-05, + "loss": 0.4952, + "step": 23456 + }, + { + "epoch": 0.9, + "learning_rate": 3.499166452936651e-05, + "loss": 0.4082, + "step": 23457 + }, + { + "epoch": 0.9, + "learning_rate": 3.4991023339317774e-05, + "loss": 0.6532, + "step": 23458 + }, + { + "epoch": 0.9, + "learning_rate": 3.4990382149269046e-05, + "loss": 0.7212, + "step": 23459 + }, + { + "epoch": 0.9, + "learning_rate": 3.498974095922031e-05, + "loss": 0.4969, + "step": 23460 + }, + { + "epoch": 0.9, + "learning_rate": 3.498909976917158e-05, + "loss": 0.5267, + "step": 23461 + }, + { + "epoch": 0.9, + "learning_rate": 3.4988458579122854e-05, + "loss": 0.5424, + "step": 23462 + }, + { + "epoch": 0.9, + "learning_rate": 3.4987817389074126e-05, + "loss": 0.5624, + "step": 23463 + }, + { + "epoch": 0.9, + "learning_rate": 3.49871761990254e-05, + "loss": 0.4603, + "step": 23464 + }, + { + "epoch": 0.9, + "learning_rate": 3.498653500897666e-05, + "loss": 0.4138, + "step": 23465 + }, + { + "epoch": 0.9, + "learning_rate": 3.4985893818927934e-05, + "loss": 0.3961, + "step": 23466 + }, + { + "epoch": 0.9, + "learning_rate": 3.49852526288792e-05, + "loss": 0.4611, + "step": 23467 + }, + { + "epoch": 0.9, + "learning_rate": 3.498461143883047e-05, + "loss": 0.5471, + "step": 23468 + }, + { + "epoch": 0.9, + "learning_rate": 3.4983970248781736e-05, + "loss": 0.5543, + "step": 23469 + }, + { + "epoch": 0.9, + "learning_rate": 3.4983329058733014e-05, + "loss": 0.4996, + "step": 23470 + }, + { + "epoch": 0.9, + "learning_rate": 3.498268786868428e-05, + "loss": 0.4282, + "step": 23471 + }, + { + "epoch": 0.9, + "learning_rate": 3.498204667863555e-05, + "loss": 0.5117, + "step": 23472 + }, + { + "epoch": 0.9, + "learning_rate": 3.498140548858682e-05, + "loss": 0.4201, + "step": 23473 + }, + { + "epoch": 0.9, + "learning_rate": 3.498076429853809e-05, + "loss": 0.5903, + "step": 23474 + }, + { + "epoch": 0.9, + "learning_rate": 3.498012310848936e-05, + "loss": 0.5389, + "step": 23475 + }, + { + "epoch": 0.9, + "learning_rate": 3.4979481918440624e-05, + "loss": 0.4705, + "step": 23476 + }, + { + "epoch": 0.9, + "learning_rate": 3.4978840728391896e-05, + "loss": 0.5279, + "step": 23477 + }, + { + "epoch": 0.9, + "learning_rate": 3.497819953834317e-05, + "loss": 0.4081, + "step": 23478 + }, + { + "epoch": 0.9, + "learning_rate": 3.497755834829444e-05, + "loss": 0.4746, + "step": 23479 + }, + { + "epoch": 0.9, + "learning_rate": 3.4976917158245704e-05, + "loss": 0.4675, + "step": 23480 + }, + { + "epoch": 0.9, + "learning_rate": 3.4976275968196976e-05, + "loss": 0.4541, + "step": 23481 + }, + { + "epoch": 0.9, + "learning_rate": 3.497563477814825e-05, + "loss": 0.6093, + "step": 23482 + }, + { + "epoch": 0.9, + "learning_rate": 3.497499358809951e-05, + "loss": 0.4917, + "step": 23483 + }, + { + "epoch": 0.9, + "learning_rate": 3.4974352398050784e-05, + "loss": 0.4744, + "step": 23484 + }, + { + "epoch": 0.9, + "learning_rate": 3.497371120800205e-05, + "loss": 0.537, + "step": 23485 + }, + { + "epoch": 0.9, + "learning_rate": 3.497307001795333e-05, + "loss": 0.4491, + "step": 23486 + }, + { + "epoch": 0.9, + "learning_rate": 3.497242882790459e-05, + "loss": 0.4624, + "step": 23487 + }, + { + "epoch": 0.9, + "learning_rate": 3.4971787637855864e-05, + "loss": 0.4936, + "step": 23488 + }, + { + "epoch": 0.9, + "learning_rate": 3.497114644780713e-05, + "loss": 0.5432, + "step": 23489 + }, + { + "epoch": 0.9, + "learning_rate": 3.49705052577584e-05, + "loss": 0.4924, + "step": 23490 + }, + { + "epoch": 0.9, + "learning_rate": 3.496986406770967e-05, + "loss": 0.4713, + "step": 23491 + }, + { + "epoch": 0.9, + "learning_rate": 3.496922287766094e-05, + "loss": 0.4269, + "step": 23492 + }, + { + "epoch": 0.9, + "learning_rate": 3.496858168761221e-05, + "loss": 0.4692, + "step": 23493 + }, + { + "epoch": 0.9, + "learning_rate": 3.496794049756348e-05, + "loss": 0.6302, + "step": 23494 + }, + { + "epoch": 0.9, + "learning_rate": 3.496729930751475e-05, + "loss": 0.4352, + "step": 23495 + }, + { + "epoch": 0.9, + "learning_rate": 3.496665811746602e-05, + "loss": 0.5113, + "step": 23496 + }, + { + "epoch": 0.9, + "learning_rate": 3.496601692741729e-05, + "loss": 0.4704, + "step": 23497 + }, + { + "epoch": 0.9, + "learning_rate": 3.4965375737368554e-05, + "loss": 0.4796, + "step": 23498 + }, + { + "epoch": 0.9, + "learning_rate": 3.4964734547319826e-05, + "loss": 0.7008, + "step": 23499 + }, + { + "epoch": 0.9, + "learning_rate": 3.49640933572711e-05, + "loss": 0.3277, + "step": 23500 + }, + { + "epoch": 0.9, + "learning_rate": 3.496345216722236e-05, + "loss": 0.4781, + "step": 23501 + }, + { + "epoch": 0.9, + "learning_rate": 3.4962810977173635e-05, + "loss": 0.3996, + "step": 23502 + }, + { + "epoch": 0.9, + "learning_rate": 3.4962169787124906e-05, + "loss": 0.4302, + "step": 23503 + }, + { + "epoch": 0.9, + "learning_rate": 3.496152859707618e-05, + "loss": 0.4717, + "step": 23504 + }, + { + "epoch": 0.9, + "learning_rate": 3.496088740702744e-05, + "loss": 0.4739, + "step": 23505 + }, + { + "epoch": 0.9, + "learning_rate": 3.4960246216978715e-05, + "loss": 0.5028, + "step": 23506 + }, + { + "epoch": 0.9, + "learning_rate": 3.495960502692998e-05, + "loss": 0.4363, + "step": 23507 + }, + { + "epoch": 0.9, + "learning_rate": 3.495896383688125e-05, + "loss": 0.4701, + "step": 23508 + }, + { + "epoch": 0.9, + "learning_rate": 3.495832264683252e-05, + "loss": 0.3468, + "step": 23509 + }, + { + "epoch": 0.9, + "learning_rate": 3.4957681456783795e-05, + "loss": 0.6928, + "step": 23510 + }, + { + "epoch": 0.9, + "learning_rate": 3.4957040266735066e-05, + "loss": 0.4758, + "step": 23511 + }, + { + "epoch": 0.9, + "learning_rate": 3.495639907668633e-05, + "loss": 0.4649, + "step": 23512 + }, + { + "epoch": 0.9, + "learning_rate": 3.49557578866376e-05, + "loss": 0.3651, + "step": 23513 + }, + { + "epoch": 0.9, + "learning_rate": 3.495511669658887e-05, + "loss": 0.411, + "step": 23514 + }, + { + "epoch": 0.9, + "learning_rate": 3.495447550654014e-05, + "loss": 0.5584, + "step": 23515 + }, + { + "epoch": 0.9, + "learning_rate": 3.4953834316491405e-05, + "loss": 0.4304, + "step": 23516 + }, + { + "epoch": 0.9, + "learning_rate": 3.4953193126442676e-05, + "loss": 0.508, + "step": 23517 + }, + { + "epoch": 0.9, + "learning_rate": 3.495255193639395e-05, + "loss": 0.4611, + "step": 23518 + }, + { + "epoch": 0.9, + "learning_rate": 3.495191074634522e-05, + "loss": 0.4415, + "step": 23519 + }, + { + "epoch": 0.9, + "learning_rate": 3.495126955629649e-05, + "loss": 0.5666, + "step": 23520 + }, + { + "epoch": 0.9, + "learning_rate": 3.4950628366247756e-05, + "loss": 0.3722, + "step": 23521 + }, + { + "epoch": 0.9, + "learning_rate": 3.494998717619903e-05, + "loss": 0.3986, + "step": 23522 + }, + { + "epoch": 0.9, + "learning_rate": 3.494934598615029e-05, + "loss": 0.4512, + "step": 23523 + }, + { + "epoch": 0.9, + "learning_rate": 3.4948704796101565e-05, + "loss": 0.4595, + "step": 23524 + }, + { + "epoch": 0.9, + "learning_rate": 3.4948063606052837e-05, + "loss": 0.5215, + "step": 23525 + }, + { + "epoch": 0.9, + "learning_rate": 3.494742241600411e-05, + "loss": 0.4919, + "step": 23526 + }, + { + "epoch": 0.9, + "learning_rate": 3.494678122595538e-05, + "loss": 0.4162, + "step": 23527 + }, + { + "epoch": 0.9, + "learning_rate": 3.4946140035906645e-05, + "loss": 0.3911, + "step": 23528 + }, + { + "epoch": 0.9, + "learning_rate": 3.494549884585792e-05, + "loss": 0.397, + "step": 23529 + }, + { + "epoch": 0.9, + "learning_rate": 3.494485765580918e-05, + "loss": 0.5098, + "step": 23530 + }, + { + "epoch": 0.9, + "learning_rate": 3.494421646576045e-05, + "loss": 0.4866, + "step": 23531 + }, + { + "epoch": 0.9, + "learning_rate": 3.494357527571172e-05, + "loss": 0.4519, + "step": 23532 + }, + { + "epoch": 0.9, + "learning_rate": 3.494293408566299e-05, + "loss": 0.4282, + "step": 23533 + }, + { + "epoch": 0.9, + "learning_rate": 3.494229289561426e-05, + "loss": 0.5456, + "step": 23534 + }, + { + "epoch": 0.9, + "learning_rate": 3.4941651705565533e-05, + "loss": 0.5665, + "step": 23535 + }, + { + "epoch": 0.9, + "learning_rate": 3.4941010515516805e-05, + "loss": 0.401, + "step": 23536 + }, + { + "epoch": 0.9, + "learning_rate": 3.494036932546807e-05, + "loss": 0.4332, + "step": 23537 + }, + { + "epoch": 0.9, + "learning_rate": 3.493972813541934e-05, + "loss": 0.4893, + "step": 23538 + }, + { + "epoch": 0.9, + "learning_rate": 3.493908694537061e-05, + "loss": 0.4073, + "step": 23539 + }, + { + "epoch": 0.91, + "learning_rate": 3.493844575532188e-05, + "loss": 0.3557, + "step": 23540 + }, + { + "epoch": 0.91, + "learning_rate": 3.493780456527314e-05, + "loss": 0.4351, + "step": 23541 + }, + { + "epoch": 0.91, + "learning_rate": 3.493716337522442e-05, + "loss": 0.5316, + "step": 23542 + }, + { + "epoch": 0.91, + "learning_rate": 3.493652218517569e-05, + "loss": 0.3706, + "step": 23543 + }, + { + "epoch": 0.91, + "learning_rate": 3.493588099512696e-05, + "loss": 0.4643, + "step": 23544 + }, + { + "epoch": 0.91, + "learning_rate": 3.493523980507823e-05, + "loss": 0.4237, + "step": 23545 + }, + { + "epoch": 0.91, + "learning_rate": 3.4934598615029495e-05, + "loss": 0.4561, + "step": 23546 + }, + { + "epoch": 0.91, + "learning_rate": 3.493395742498077e-05, + "loss": 0.4762, + "step": 23547 + }, + { + "epoch": 0.91, + "learning_rate": 3.493331623493203e-05, + "loss": 0.5029, + "step": 23548 + }, + { + "epoch": 0.91, + "learning_rate": 3.4932675044883304e-05, + "loss": 0.5673, + "step": 23549 + }, + { + "epoch": 0.91, + "learning_rate": 3.4932033854834575e-05, + "loss": 0.6615, + "step": 23550 + }, + { + "epoch": 0.91, + "learning_rate": 3.493139266478585e-05, + "loss": 0.4409, + "step": 23551 + }, + { + "epoch": 0.91, + "learning_rate": 3.493075147473711e-05, + "loss": 0.4492, + "step": 23552 + }, + { + "epoch": 0.91, + "learning_rate": 3.4930110284688384e-05, + "loss": 0.5499, + "step": 23553 + }, + { + "epoch": 0.91, + "learning_rate": 3.4929469094639655e-05, + "loss": 0.4339, + "step": 23554 + }, + { + "epoch": 0.91, + "learning_rate": 3.492882790459092e-05, + "loss": 0.3939, + "step": 23555 + }, + { + "epoch": 0.91, + "learning_rate": 3.492818671454219e-05, + "loss": 0.4554, + "step": 23556 + }, + { + "epoch": 0.91, + "learning_rate": 3.492754552449346e-05, + "loss": 0.4935, + "step": 23557 + }, + { + "epoch": 0.91, + "learning_rate": 3.4926904334444735e-05, + "loss": 0.5541, + "step": 23558 + }, + { + "epoch": 0.91, + "learning_rate": 3.4926263144396e-05, + "loss": 0.5769, + "step": 23559 + }, + { + "epoch": 0.91, + "learning_rate": 3.492562195434727e-05, + "loss": 0.5834, + "step": 23560 + }, + { + "epoch": 0.91, + "learning_rate": 3.492498076429854e-05, + "loss": 0.3745, + "step": 23561 + }, + { + "epoch": 0.91, + "learning_rate": 3.492433957424981e-05, + "loss": 0.4691, + "step": 23562 + }, + { + "epoch": 0.91, + "learning_rate": 3.492369838420108e-05, + "loss": 0.4003, + "step": 23563 + }, + { + "epoch": 0.91, + "learning_rate": 3.4923057194152345e-05, + "loss": 0.4665, + "step": 23564 + }, + { + "epoch": 0.91, + "learning_rate": 3.492241600410362e-05, + "loss": 0.4285, + "step": 23565 + }, + { + "epoch": 0.91, + "learning_rate": 3.492177481405489e-05, + "loss": 0.5558, + "step": 23566 + }, + { + "epoch": 0.91, + "learning_rate": 3.492113362400616e-05, + "loss": 0.5579, + "step": 23567 + }, + { + "epoch": 0.91, + "learning_rate": 3.4920492433957426e-05, + "loss": 0.6228, + "step": 23568 + }, + { + "epoch": 0.91, + "learning_rate": 3.49198512439087e-05, + "loss": 0.4868, + "step": 23569 + }, + { + "epoch": 0.91, + "learning_rate": 3.491921005385996e-05, + "loss": 0.5527, + "step": 23570 + }, + { + "epoch": 0.91, + "learning_rate": 3.4918568863811234e-05, + "loss": 0.4136, + "step": 23571 + }, + { + "epoch": 0.91, + "learning_rate": 3.4917927673762506e-05, + "loss": 0.4545, + "step": 23572 + }, + { + "epoch": 0.91, + "learning_rate": 3.491728648371377e-05, + "loss": 0.3736, + "step": 23573 + }, + { + "epoch": 0.91, + "learning_rate": 3.491664529366505e-05, + "loss": 0.4038, + "step": 23574 + }, + { + "epoch": 0.91, + "learning_rate": 3.4916004103616314e-05, + "loss": 0.4979, + "step": 23575 + }, + { + "epoch": 0.91, + "learning_rate": 3.4915362913567586e-05, + "loss": 0.4721, + "step": 23576 + }, + { + "epoch": 0.91, + "learning_rate": 3.491472172351885e-05, + "loss": 0.586, + "step": 23577 + }, + { + "epoch": 0.91, + "learning_rate": 3.491408053347012e-05, + "loss": 0.4241, + "step": 23578 + }, + { + "epoch": 0.91, + "learning_rate": 3.491343934342139e-05, + "loss": 0.588, + "step": 23579 + }, + { + "epoch": 0.91, + "learning_rate": 3.491279815337266e-05, + "loss": 0.4677, + "step": 23580 + }, + { + "epoch": 0.91, + "learning_rate": 3.491215696332393e-05, + "loss": 0.5569, + "step": 23581 + }, + { + "epoch": 0.91, + "learning_rate": 3.49115157732752e-05, + "loss": 0.5936, + "step": 23582 + }, + { + "epoch": 0.91, + "learning_rate": 3.4910874583226474e-05, + "loss": 0.4826, + "step": 23583 + }, + { + "epoch": 0.91, + "learning_rate": 3.491023339317774e-05, + "loss": 0.3722, + "step": 23584 + }, + { + "epoch": 0.91, + "learning_rate": 3.490959220312901e-05, + "loss": 0.4396, + "step": 23585 + }, + { + "epoch": 0.91, + "learning_rate": 3.4908951013080276e-05, + "loss": 0.3664, + "step": 23586 + }, + { + "epoch": 0.91, + "learning_rate": 3.490830982303155e-05, + "loss": 0.4011, + "step": 23587 + }, + { + "epoch": 0.91, + "learning_rate": 3.490766863298281e-05, + "loss": 0.5567, + "step": 23588 + }, + { + "epoch": 0.91, + "learning_rate": 3.4907027442934084e-05, + "loss": 0.402, + "step": 23589 + }, + { + "epoch": 0.91, + "learning_rate": 3.4906386252885356e-05, + "loss": 0.572, + "step": 23590 + }, + { + "epoch": 0.91, + "learning_rate": 3.490574506283663e-05, + "loss": 0.4583, + "step": 23591 + }, + { + "epoch": 0.91, + "learning_rate": 3.49051038727879e-05, + "loss": 0.406, + "step": 23592 + }, + { + "epoch": 0.91, + "learning_rate": 3.4904462682739164e-05, + "loss": 0.3934, + "step": 23593 + }, + { + "epoch": 0.91, + "learning_rate": 3.4903821492690436e-05, + "loss": 0.4331, + "step": 23594 + }, + { + "epoch": 0.91, + "learning_rate": 3.49031803026417e-05, + "loss": 0.4907, + "step": 23595 + }, + { + "epoch": 0.91, + "learning_rate": 3.490253911259297e-05, + "loss": 0.5481, + "step": 23596 + }, + { + "epoch": 0.91, + "learning_rate": 3.4901897922544244e-05, + "loss": 0.5327, + "step": 23597 + }, + { + "epoch": 0.91, + "learning_rate": 3.4901256732495516e-05, + "loss": 0.5507, + "step": 23598 + }, + { + "epoch": 0.91, + "learning_rate": 3.490061554244679e-05, + "loss": 0.4859, + "step": 23599 + }, + { + "epoch": 0.91, + "learning_rate": 3.489997435239805e-05, + "loss": 0.4938, + "step": 23600 + }, + { + "epoch": 0.91, + "learning_rate": 3.4899333162349324e-05, + "loss": 0.3962, + "step": 23601 + }, + { + "epoch": 0.91, + "learning_rate": 3.489869197230059e-05, + "loss": 0.5831, + "step": 23602 + }, + { + "epoch": 0.91, + "learning_rate": 3.489805078225186e-05, + "loss": 0.4118, + "step": 23603 + }, + { + "epoch": 0.91, + "learning_rate": 3.4897409592203126e-05, + "loss": 0.5242, + "step": 23604 + }, + { + "epoch": 0.91, + "learning_rate": 3.48967684021544e-05, + "loss": 0.4999, + "step": 23605 + }, + { + "epoch": 0.91, + "learning_rate": 3.489612721210567e-05, + "loss": 0.4575, + "step": 23606 + }, + { + "epoch": 0.91, + "learning_rate": 3.489548602205694e-05, + "loss": 0.3831, + "step": 23607 + }, + { + "epoch": 0.91, + "learning_rate": 3.489484483200821e-05, + "loss": 0.4978, + "step": 23608 + }, + { + "epoch": 0.91, + "learning_rate": 3.489420364195948e-05, + "loss": 0.4762, + "step": 23609 + }, + { + "epoch": 0.91, + "learning_rate": 3.489356245191075e-05, + "loss": 0.464, + "step": 23610 + }, + { + "epoch": 0.91, + "learning_rate": 3.4892921261862014e-05, + "loss": 0.4018, + "step": 23611 + }, + { + "epoch": 0.91, + "learning_rate": 3.4892280071813286e-05, + "loss": 0.6131, + "step": 23612 + }, + { + "epoch": 0.91, + "learning_rate": 3.489163888176456e-05, + "loss": 0.386, + "step": 23613 + }, + { + "epoch": 0.91, + "learning_rate": 3.489099769171583e-05, + "loss": 0.5549, + "step": 23614 + }, + { + "epoch": 0.91, + "learning_rate": 3.48903565016671e-05, + "loss": 0.3631, + "step": 23615 + }, + { + "epoch": 0.91, + "learning_rate": 3.4889715311618366e-05, + "loss": 0.4688, + "step": 23616 + }, + { + "epoch": 0.91, + "learning_rate": 3.488907412156964e-05, + "loss": 0.5206, + "step": 23617 + }, + { + "epoch": 0.91, + "learning_rate": 3.48884329315209e-05, + "loss": 0.5737, + "step": 23618 + }, + { + "epoch": 0.91, + "learning_rate": 3.4887791741472175e-05, + "loss": 0.5895, + "step": 23619 + }, + { + "epoch": 0.91, + "learning_rate": 3.488715055142344e-05, + "loss": 0.513, + "step": 23620 + }, + { + "epoch": 0.91, + "learning_rate": 3.488650936137471e-05, + "loss": 0.4537, + "step": 23621 + }, + { + "epoch": 0.91, + "learning_rate": 3.488586817132598e-05, + "loss": 0.561, + "step": 23622 + }, + { + "epoch": 0.91, + "learning_rate": 3.4885226981277255e-05, + "loss": 0.5301, + "step": 23623 + }, + { + "epoch": 0.91, + "learning_rate": 3.4884585791228526e-05, + "loss": 0.4215, + "step": 23624 + }, + { + "epoch": 0.91, + "learning_rate": 3.488394460117979e-05, + "loss": 0.6145, + "step": 23625 + }, + { + "epoch": 0.91, + "learning_rate": 3.488330341113106e-05, + "loss": 0.5444, + "step": 23626 + }, + { + "epoch": 0.91, + "learning_rate": 3.488266222108233e-05, + "loss": 0.454, + "step": 23627 + }, + { + "epoch": 0.91, + "learning_rate": 3.48820210310336e-05, + "loss": 0.4123, + "step": 23628 + }, + { + "epoch": 0.91, + "learning_rate": 3.4881379840984865e-05, + "loss": 0.5322, + "step": 23629 + }, + { + "epoch": 0.91, + "learning_rate": 3.488073865093614e-05, + "loss": 0.4403, + "step": 23630 + }, + { + "epoch": 0.91, + "learning_rate": 3.488009746088741e-05, + "loss": 0.5329, + "step": 23631 + }, + { + "epoch": 0.91, + "learning_rate": 3.487945627083868e-05, + "loss": 0.4664, + "step": 23632 + }, + { + "epoch": 0.91, + "learning_rate": 3.487881508078995e-05, + "loss": 0.4618, + "step": 23633 + }, + { + "epoch": 0.91, + "learning_rate": 3.4878173890741217e-05, + "loss": 0.5086, + "step": 23634 + }, + { + "epoch": 0.91, + "learning_rate": 3.487753270069249e-05, + "loss": 0.5919, + "step": 23635 + }, + { + "epoch": 0.91, + "learning_rate": 3.487689151064375e-05, + "loss": 0.4742, + "step": 23636 + }, + { + "epoch": 0.91, + "learning_rate": 3.4876250320595025e-05, + "loss": 0.4668, + "step": 23637 + }, + { + "epoch": 0.91, + "learning_rate": 3.4875609130546297e-05, + "loss": 0.496, + "step": 23638 + }, + { + "epoch": 0.91, + "learning_rate": 3.487496794049757e-05, + "loss": 0.5594, + "step": 23639 + }, + { + "epoch": 0.91, + "learning_rate": 3.487432675044883e-05, + "loss": 0.5451, + "step": 23640 + }, + { + "epoch": 0.91, + "learning_rate": 3.4873685560400105e-05, + "loss": 0.3655, + "step": 23641 + }, + { + "epoch": 0.91, + "learning_rate": 3.487304437035138e-05, + "loss": 0.5391, + "step": 23642 + }, + { + "epoch": 0.91, + "learning_rate": 3.487240318030264e-05, + "loss": 0.5424, + "step": 23643 + }, + { + "epoch": 0.91, + "learning_rate": 3.487176199025391e-05, + "loss": 0.4768, + "step": 23644 + }, + { + "epoch": 0.91, + "learning_rate": 3.487112080020518e-05, + "loss": 0.4704, + "step": 23645 + }, + { + "epoch": 0.91, + "learning_rate": 3.487047961015646e-05, + "loss": 0.5468, + "step": 23646 + }, + { + "epoch": 0.91, + "learning_rate": 3.486983842010772e-05, + "loss": 0.5597, + "step": 23647 + }, + { + "epoch": 0.91, + "learning_rate": 3.4869197230058993e-05, + "loss": 0.3418, + "step": 23648 + }, + { + "epoch": 0.91, + "learning_rate": 3.486855604001026e-05, + "loss": 0.484, + "step": 23649 + }, + { + "epoch": 0.91, + "learning_rate": 3.486791484996153e-05, + "loss": 0.4823, + "step": 23650 + }, + { + "epoch": 0.91, + "learning_rate": 3.48672736599128e-05, + "loss": 0.4571, + "step": 23651 + }, + { + "epoch": 0.91, + "learning_rate": 3.486663246986407e-05, + "loss": 0.5113, + "step": 23652 + }, + { + "epoch": 0.91, + "learning_rate": 3.486599127981534e-05, + "loss": 0.4368, + "step": 23653 + }, + { + "epoch": 0.91, + "learning_rate": 3.486535008976661e-05, + "loss": 0.5854, + "step": 23654 + }, + { + "epoch": 0.91, + "learning_rate": 3.486470889971788e-05, + "loss": 0.4678, + "step": 23655 + }, + { + "epoch": 0.91, + "learning_rate": 3.486406770966915e-05, + "loss": 0.6399, + "step": 23656 + }, + { + "epoch": 0.91, + "learning_rate": 3.486342651962042e-05, + "loss": 0.4676, + "step": 23657 + }, + { + "epoch": 0.91, + "learning_rate": 3.4862785329571684e-05, + "loss": 0.3705, + "step": 23658 + }, + { + "epoch": 0.91, + "learning_rate": 3.4862144139522955e-05, + "loss": 0.4284, + "step": 23659 + }, + { + "epoch": 0.91, + "learning_rate": 3.486150294947423e-05, + "loss": 0.5436, + "step": 23660 + }, + { + "epoch": 0.91, + "learning_rate": 3.486086175942549e-05, + "loss": 0.554, + "step": 23661 + }, + { + "epoch": 0.91, + "learning_rate": 3.4860220569376764e-05, + "loss": 0.5546, + "step": 23662 + }, + { + "epoch": 0.91, + "learning_rate": 3.4859579379328035e-05, + "loss": 0.6115, + "step": 23663 + }, + { + "epoch": 0.91, + "learning_rate": 3.485893818927931e-05, + "loss": 0.4629, + "step": 23664 + }, + { + "epoch": 0.91, + "learning_rate": 3.485829699923057e-05, + "loss": 0.4754, + "step": 23665 + }, + { + "epoch": 0.91, + "learning_rate": 3.4857655809181844e-05, + "loss": 0.4563, + "step": 23666 + }, + { + "epoch": 0.91, + "learning_rate": 3.485701461913311e-05, + "loss": 0.4012, + "step": 23667 + }, + { + "epoch": 0.91, + "learning_rate": 3.485637342908438e-05, + "loss": 0.483, + "step": 23668 + }, + { + "epoch": 0.91, + "learning_rate": 3.485573223903565e-05, + "loss": 0.4717, + "step": 23669 + }, + { + "epoch": 0.91, + "learning_rate": 3.4855091048986924e-05, + "loss": 0.4106, + "step": 23670 + }, + { + "epoch": 0.91, + "learning_rate": 3.4854449858938195e-05, + "loss": 0.5371, + "step": 23671 + }, + { + "epoch": 0.91, + "learning_rate": 3.485380866888946e-05, + "loss": 0.4435, + "step": 23672 + }, + { + "epoch": 0.91, + "learning_rate": 3.485316747884073e-05, + "loss": 0.4215, + "step": 23673 + }, + { + "epoch": 0.91, + "learning_rate": 3.4852526288792e-05, + "loss": 0.3606, + "step": 23674 + }, + { + "epoch": 0.91, + "learning_rate": 3.485188509874327e-05, + "loss": 0.5139, + "step": 23675 + }, + { + "epoch": 0.91, + "learning_rate": 3.4851243908694534e-05, + "loss": 0.5208, + "step": 23676 + }, + { + "epoch": 0.91, + "learning_rate": 3.4850602718645805e-05, + "loss": 0.4653, + "step": 23677 + }, + { + "epoch": 0.91, + "learning_rate": 3.484996152859708e-05, + "loss": 0.4983, + "step": 23678 + }, + { + "epoch": 0.91, + "learning_rate": 3.484932033854835e-05, + "loss": 0.3784, + "step": 23679 + }, + { + "epoch": 0.91, + "learning_rate": 3.484867914849962e-05, + "loss": 0.5142, + "step": 23680 + }, + { + "epoch": 0.91, + "learning_rate": 3.4848037958450886e-05, + "loss": 0.5726, + "step": 23681 + }, + { + "epoch": 0.91, + "learning_rate": 3.484739676840216e-05, + "loss": 0.4747, + "step": 23682 + }, + { + "epoch": 0.91, + "learning_rate": 3.484675557835342e-05, + "loss": 0.5357, + "step": 23683 + }, + { + "epoch": 0.91, + "learning_rate": 3.4846114388304694e-05, + "loss": 0.4589, + "step": 23684 + }, + { + "epoch": 0.91, + "learning_rate": 3.4845473198255966e-05, + "loss": 0.4397, + "step": 23685 + }, + { + "epoch": 0.91, + "learning_rate": 3.484483200820724e-05, + "loss": 0.4937, + "step": 23686 + }, + { + "epoch": 0.91, + "learning_rate": 3.484419081815851e-05, + "loss": 0.4542, + "step": 23687 + }, + { + "epoch": 0.91, + "learning_rate": 3.4843549628109774e-05, + "loss": 0.5581, + "step": 23688 + }, + { + "epoch": 0.91, + "learning_rate": 3.4842908438061046e-05, + "loss": 0.5192, + "step": 23689 + }, + { + "epoch": 0.91, + "learning_rate": 3.484226724801231e-05, + "loss": 0.544, + "step": 23690 + }, + { + "epoch": 0.91, + "learning_rate": 3.484162605796358e-05, + "loss": 0.6155, + "step": 23691 + }, + { + "epoch": 0.91, + "learning_rate": 3.484098486791485e-05, + "loss": 0.5746, + "step": 23692 + }, + { + "epoch": 0.91, + "learning_rate": 3.484034367786612e-05, + "loss": 0.4899, + "step": 23693 + }, + { + "epoch": 0.91, + "learning_rate": 3.483970248781739e-05, + "loss": 0.4734, + "step": 23694 + }, + { + "epoch": 0.91, + "learning_rate": 3.483906129776866e-05, + "loss": 0.4971, + "step": 23695 + }, + { + "epoch": 0.91, + "learning_rate": 3.4838420107719934e-05, + "loss": 0.5322, + "step": 23696 + }, + { + "epoch": 0.91, + "learning_rate": 3.48377789176712e-05, + "loss": 0.594, + "step": 23697 + }, + { + "epoch": 0.91, + "learning_rate": 3.483713772762247e-05, + "loss": 0.4636, + "step": 23698 + }, + { + "epoch": 0.91, + "learning_rate": 3.4836496537573736e-05, + "loss": 0.4676, + "step": 23699 + }, + { + "epoch": 0.91, + "learning_rate": 3.483585534752501e-05, + "loss": 0.4885, + "step": 23700 + }, + { + "epoch": 0.91, + "learning_rate": 3.483521415747627e-05, + "loss": 0.4338, + "step": 23701 + }, + { + "epoch": 0.91, + "learning_rate": 3.483457296742755e-05, + "loss": 0.5592, + "step": 23702 + }, + { + "epoch": 0.91, + "learning_rate": 3.4833931777378816e-05, + "loss": 0.5186, + "step": 23703 + }, + { + "epoch": 0.91, + "learning_rate": 3.483329058733009e-05, + "loss": 0.4506, + "step": 23704 + }, + { + "epoch": 0.91, + "learning_rate": 3.483264939728136e-05, + "loss": 0.4974, + "step": 23705 + }, + { + "epoch": 0.91, + "learning_rate": 3.4832008207232624e-05, + "loss": 0.3994, + "step": 23706 + }, + { + "epoch": 0.91, + "learning_rate": 3.4831367017183896e-05, + "loss": 0.5113, + "step": 23707 + }, + { + "epoch": 0.91, + "learning_rate": 3.483072582713516e-05, + "loss": 0.5106, + "step": 23708 + }, + { + "epoch": 0.91, + "learning_rate": 3.483008463708643e-05, + "loss": 0.4829, + "step": 23709 + }, + { + "epoch": 0.91, + "learning_rate": 3.4829443447037704e-05, + "loss": 0.5215, + "step": 23710 + }, + { + "epoch": 0.91, + "learning_rate": 3.4828802256988976e-05, + "loss": 0.5503, + "step": 23711 + }, + { + "epoch": 0.91, + "learning_rate": 3.482816106694024e-05, + "loss": 0.4475, + "step": 23712 + }, + { + "epoch": 0.91, + "learning_rate": 3.482751987689151e-05, + "loss": 0.3729, + "step": 23713 + }, + { + "epoch": 0.91, + "learning_rate": 3.4826878686842784e-05, + "loss": 0.4339, + "step": 23714 + }, + { + "epoch": 0.91, + "learning_rate": 3.482623749679405e-05, + "loss": 0.51, + "step": 23715 + }, + { + "epoch": 0.91, + "learning_rate": 3.482559630674532e-05, + "loss": 0.4993, + "step": 23716 + }, + { + "epoch": 0.91, + "learning_rate": 3.4824955116696586e-05, + "loss": 0.5487, + "step": 23717 + }, + { + "epoch": 0.91, + "learning_rate": 3.4824313926647865e-05, + "loss": 0.4264, + "step": 23718 + }, + { + "epoch": 0.91, + "learning_rate": 3.482367273659913e-05, + "loss": 0.5187, + "step": 23719 + }, + { + "epoch": 0.91, + "learning_rate": 3.48230315465504e-05, + "loss": 0.4689, + "step": 23720 + }, + { + "epoch": 0.91, + "learning_rate": 3.4822390356501666e-05, + "loss": 0.5635, + "step": 23721 + }, + { + "epoch": 0.91, + "learning_rate": 3.482174916645294e-05, + "loss": 0.3609, + "step": 23722 + }, + { + "epoch": 0.91, + "learning_rate": 3.482110797640421e-05, + "loss": 0.4947, + "step": 23723 + }, + { + "epoch": 0.91, + "learning_rate": 3.4820466786355475e-05, + "loss": 0.5958, + "step": 23724 + }, + { + "epoch": 0.91, + "learning_rate": 3.4819825596306746e-05, + "loss": 0.3832, + "step": 23725 + }, + { + "epoch": 0.91, + "learning_rate": 3.481918440625802e-05, + "loss": 0.3911, + "step": 23726 + }, + { + "epoch": 0.91, + "learning_rate": 3.481854321620929e-05, + "loss": 0.6073, + "step": 23727 + }, + { + "epoch": 0.91, + "learning_rate": 3.4817902026160555e-05, + "loss": 0.4678, + "step": 23728 + }, + { + "epoch": 0.91, + "learning_rate": 3.4817260836111826e-05, + "loss": 0.4215, + "step": 23729 + }, + { + "epoch": 0.91, + "learning_rate": 3.481661964606309e-05, + "loss": 0.5558, + "step": 23730 + }, + { + "epoch": 0.91, + "learning_rate": 3.481597845601436e-05, + "loss": 0.4164, + "step": 23731 + }, + { + "epoch": 0.91, + "learning_rate": 3.4815337265965635e-05, + "loss": 0.4803, + "step": 23732 + }, + { + "epoch": 0.91, + "learning_rate": 3.48146960759169e-05, + "loss": 0.412, + "step": 23733 + }, + { + "epoch": 0.91, + "learning_rate": 3.481405488586818e-05, + "loss": 0.4555, + "step": 23734 + }, + { + "epoch": 0.91, + "learning_rate": 3.481341369581944e-05, + "loss": 0.5392, + "step": 23735 + }, + { + "epoch": 0.91, + "learning_rate": 3.4812772505770715e-05, + "loss": 0.5739, + "step": 23736 + }, + { + "epoch": 0.91, + "learning_rate": 3.481213131572198e-05, + "loss": 0.5013, + "step": 23737 + }, + { + "epoch": 0.91, + "learning_rate": 3.481149012567325e-05, + "loss": 0.45, + "step": 23738 + }, + { + "epoch": 0.91, + "learning_rate": 3.4810848935624516e-05, + "loss": 0.6459, + "step": 23739 + }, + { + "epoch": 0.91, + "learning_rate": 3.481020774557579e-05, + "loss": 0.572, + "step": 23740 + }, + { + "epoch": 0.91, + "learning_rate": 3.480956655552706e-05, + "loss": 0.3603, + "step": 23741 + }, + { + "epoch": 0.91, + "learning_rate": 3.480892536547833e-05, + "loss": 0.4764, + "step": 23742 + }, + { + "epoch": 0.91, + "learning_rate": 3.48082841754296e-05, + "loss": 0.5423, + "step": 23743 + }, + { + "epoch": 0.91, + "learning_rate": 3.480764298538087e-05, + "loss": 0.5573, + "step": 23744 + }, + { + "epoch": 0.91, + "learning_rate": 3.480700179533214e-05, + "loss": 0.4918, + "step": 23745 + }, + { + "epoch": 0.91, + "learning_rate": 3.4806360605283405e-05, + "loss": 0.5151, + "step": 23746 + }, + { + "epoch": 0.91, + "learning_rate": 3.4805719415234677e-05, + "loss": 0.4724, + "step": 23747 + }, + { + "epoch": 0.91, + "learning_rate": 3.480507822518594e-05, + "loss": 0.4651, + "step": 23748 + }, + { + "epoch": 0.91, + "learning_rate": 3.480443703513721e-05, + "loss": 0.5205, + "step": 23749 + }, + { + "epoch": 0.91, + "learning_rate": 3.4803795845088485e-05, + "loss": 0.4198, + "step": 23750 + }, + { + "epoch": 0.91, + "learning_rate": 3.480315465503976e-05, + "loss": 0.5389, + "step": 23751 + }, + { + "epoch": 0.91, + "learning_rate": 3.480251346499103e-05, + "loss": 0.5991, + "step": 23752 + }, + { + "epoch": 0.91, + "learning_rate": 3.480187227494229e-05, + "loss": 0.4844, + "step": 23753 + }, + { + "epoch": 0.91, + "learning_rate": 3.4801231084893565e-05, + "loss": 0.5147, + "step": 23754 + }, + { + "epoch": 0.91, + "learning_rate": 3.480058989484483e-05, + "loss": 0.4956, + "step": 23755 + }, + { + "epoch": 0.91, + "learning_rate": 3.47999487047961e-05, + "loss": 0.4617, + "step": 23756 + }, + { + "epoch": 0.91, + "learning_rate": 3.479930751474737e-05, + "loss": 0.5429, + "step": 23757 + }, + { + "epoch": 0.91, + "learning_rate": 3.4798666324698645e-05, + "loss": 0.4627, + "step": 23758 + }, + { + "epoch": 0.91, + "learning_rate": 3.479802513464992e-05, + "loss": 0.4896, + "step": 23759 + }, + { + "epoch": 0.91, + "learning_rate": 3.479738394460118e-05, + "loss": 0.5073, + "step": 23760 + }, + { + "epoch": 0.91, + "learning_rate": 3.4796742754552453e-05, + "loss": 0.5331, + "step": 23761 + }, + { + "epoch": 0.91, + "learning_rate": 3.479610156450372e-05, + "loss": 0.5457, + "step": 23762 + }, + { + "epoch": 0.91, + "learning_rate": 3.479546037445499e-05, + "loss": 0.4615, + "step": 23763 + }, + { + "epoch": 0.91, + "learning_rate": 3.4794819184406255e-05, + "loss": 0.4077, + "step": 23764 + }, + { + "epoch": 0.91, + "learning_rate": 3.479417799435753e-05, + "loss": 0.5166, + "step": 23765 + }, + { + "epoch": 0.91, + "learning_rate": 3.47935368043088e-05, + "loss": 0.4724, + "step": 23766 + }, + { + "epoch": 0.91, + "learning_rate": 3.479289561426007e-05, + "loss": 0.4164, + "step": 23767 + }, + { + "epoch": 0.91, + "learning_rate": 3.479225442421134e-05, + "loss": 0.524, + "step": 23768 + }, + { + "epoch": 0.91, + "learning_rate": 3.479161323416261e-05, + "loss": 0.5134, + "step": 23769 + }, + { + "epoch": 0.91, + "learning_rate": 3.479097204411388e-05, + "loss": 0.3637, + "step": 23770 + }, + { + "epoch": 0.91, + "learning_rate": 3.4790330854065144e-05, + "loss": 0.4569, + "step": 23771 + }, + { + "epoch": 0.91, + "learning_rate": 3.4789689664016415e-05, + "loss": 0.4935, + "step": 23772 + }, + { + "epoch": 0.91, + "learning_rate": 3.478904847396769e-05, + "loss": 0.4064, + "step": 23773 + }, + { + "epoch": 0.91, + "learning_rate": 3.478840728391896e-05, + "loss": 0.4996, + "step": 23774 + }, + { + "epoch": 0.91, + "learning_rate": 3.4787766093870224e-05, + "loss": 0.6057, + "step": 23775 + }, + { + "epoch": 0.91, + "learning_rate": 3.4787124903821495e-05, + "loss": 0.4459, + "step": 23776 + }, + { + "epoch": 0.91, + "learning_rate": 3.478648371377277e-05, + "loss": 0.6724, + "step": 23777 + }, + { + "epoch": 0.91, + "learning_rate": 3.478584252372403e-05, + "loss": 0.5202, + "step": 23778 + }, + { + "epoch": 0.91, + "learning_rate": 3.4785201333675304e-05, + "loss": 0.481, + "step": 23779 + }, + { + "epoch": 0.91, + "learning_rate": 3.478456014362657e-05, + "loss": 0.3822, + "step": 23780 + }, + { + "epoch": 0.91, + "learning_rate": 3.478391895357784e-05, + "loss": 0.4371, + "step": 23781 + }, + { + "epoch": 0.91, + "learning_rate": 3.478327776352911e-05, + "loss": 0.5054, + "step": 23782 + }, + { + "epoch": 0.91, + "learning_rate": 3.4782636573480384e-05, + "loss": 0.5027, + "step": 23783 + }, + { + "epoch": 0.91, + "learning_rate": 3.478199538343165e-05, + "loss": 0.5398, + "step": 23784 + }, + { + "epoch": 0.91, + "learning_rate": 3.478135419338292e-05, + "loss": 0.5501, + "step": 23785 + }, + { + "epoch": 0.91, + "learning_rate": 3.478071300333419e-05, + "loss": 0.5363, + "step": 23786 + }, + { + "epoch": 0.91, + "learning_rate": 3.478007181328546e-05, + "loss": 0.4118, + "step": 23787 + }, + { + "epoch": 0.91, + "learning_rate": 3.477943062323673e-05, + "loss": 0.3233, + "step": 23788 + }, + { + "epoch": 0.91, + "learning_rate": 3.4778789433187994e-05, + "loss": 0.4395, + "step": 23789 + }, + { + "epoch": 0.91, + "learning_rate": 3.477814824313927e-05, + "loss": 0.5264, + "step": 23790 + }, + { + "epoch": 0.91, + "learning_rate": 3.477750705309054e-05, + "loss": 0.5187, + "step": 23791 + }, + { + "epoch": 0.91, + "learning_rate": 3.477686586304181e-05, + "loss": 0.4363, + "step": 23792 + }, + { + "epoch": 0.91, + "learning_rate": 3.4776224672993074e-05, + "loss": 0.4741, + "step": 23793 + }, + { + "epoch": 0.91, + "learning_rate": 3.4775583482944346e-05, + "loss": 0.5307, + "step": 23794 + }, + { + "epoch": 0.91, + "learning_rate": 3.477494229289562e-05, + "loss": 0.4913, + "step": 23795 + }, + { + "epoch": 0.91, + "learning_rate": 3.477430110284688e-05, + "loss": 0.4418, + "step": 23796 + }, + { + "epoch": 0.91, + "learning_rate": 3.4773659912798154e-05, + "loss": 0.5231, + "step": 23797 + }, + { + "epoch": 0.91, + "learning_rate": 3.4773018722749426e-05, + "loss": 0.4235, + "step": 23798 + }, + { + "epoch": 0.91, + "learning_rate": 3.47723775327007e-05, + "loss": 0.4713, + "step": 23799 + }, + { + "epoch": 0.92, + "learning_rate": 3.477173634265196e-05, + "loss": 0.4577, + "step": 23800 + }, + { + "epoch": 0.92, + "learning_rate": 3.4771095152603234e-05, + "loss": 0.4396, + "step": 23801 + }, + { + "epoch": 0.92, + "learning_rate": 3.47704539625545e-05, + "loss": 0.4258, + "step": 23802 + }, + { + "epoch": 0.92, + "learning_rate": 3.476981277250577e-05, + "loss": 0.5691, + "step": 23803 + }, + { + "epoch": 0.92, + "learning_rate": 3.476917158245704e-05, + "loss": 0.5883, + "step": 23804 + }, + { + "epoch": 0.92, + "learning_rate": 3.476853039240831e-05, + "loss": 0.4568, + "step": 23805 + }, + { + "epoch": 0.92, + "learning_rate": 3.4767889202359586e-05, + "loss": 0.5739, + "step": 23806 + }, + { + "epoch": 0.92, + "learning_rate": 3.476724801231085e-05, + "loss": 0.528, + "step": 23807 + }, + { + "epoch": 0.92, + "learning_rate": 3.476660682226212e-05, + "loss": 0.4871, + "step": 23808 + }, + { + "epoch": 0.92, + "learning_rate": 3.476596563221339e-05, + "loss": 0.5642, + "step": 23809 + }, + { + "epoch": 0.92, + "learning_rate": 3.476532444216466e-05, + "loss": 0.3995, + "step": 23810 + }, + { + "epoch": 0.92, + "learning_rate": 3.476468325211593e-05, + "loss": 0.5209, + "step": 23811 + }, + { + "epoch": 0.92, + "learning_rate": 3.4764042062067196e-05, + "loss": 0.482, + "step": 23812 + }, + { + "epoch": 0.92, + "learning_rate": 3.476340087201847e-05, + "loss": 0.4011, + "step": 23813 + }, + { + "epoch": 0.92, + "learning_rate": 3.476275968196974e-05, + "loss": 0.4798, + "step": 23814 + }, + { + "epoch": 0.92, + "learning_rate": 3.476211849192101e-05, + "loss": 0.4656, + "step": 23815 + }, + { + "epoch": 0.92, + "learning_rate": 3.4761477301872276e-05, + "loss": 0.5015, + "step": 23816 + }, + { + "epoch": 0.92, + "learning_rate": 3.476083611182355e-05, + "loss": 0.3282, + "step": 23817 + }, + { + "epoch": 0.92, + "learning_rate": 3.476019492177481e-05, + "loss": 0.5088, + "step": 23818 + }, + { + "epoch": 0.92, + "learning_rate": 3.4759553731726084e-05, + "loss": 0.6152, + "step": 23819 + }, + { + "epoch": 0.92, + "learning_rate": 3.4758912541677356e-05, + "loss": 0.4754, + "step": 23820 + }, + { + "epoch": 0.92, + "learning_rate": 3.475827135162862e-05, + "loss": 0.4807, + "step": 23821 + }, + { + "epoch": 0.92, + "learning_rate": 3.475763016157989e-05, + "loss": 0.5267, + "step": 23822 + }, + { + "epoch": 0.92, + "learning_rate": 3.4756988971531164e-05, + "loss": 0.5512, + "step": 23823 + }, + { + "epoch": 0.92, + "learning_rate": 3.4756347781482436e-05, + "loss": 0.4897, + "step": 23824 + }, + { + "epoch": 0.92, + "learning_rate": 3.47557065914337e-05, + "loss": 0.4891, + "step": 23825 + }, + { + "epoch": 0.92, + "learning_rate": 3.475506540138497e-05, + "loss": 0.4631, + "step": 23826 + }, + { + "epoch": 0.92, + "learning_rate": 3.475442421133624e-05, + "loss": 0.5331, + "step": 23827 + }, + { + "epoch": 0.92, + "learning_rate": 3.475378302128751e-05, + "loss": 0.5388, + "step": 23828 + }, + { + "epoch": 0.92, + "learning_rate": 3.475314183123878e-05, + "loss": 0.4137, + "step": 23829 + }, + { + "epoch": 0.92, + "learning_rate": 3.475250064119005e-05, + "loss": 0.4691, + "step": 23830 + }, + { + "epoch": 0.92, + "learning_rate": 3.4751859451141325e-05, + "loss": 0.4408, + "step": 23831 + }, + { + "epoch": 0.92, + "learning_rate": 3.475121826109259e-05, + "loss": 0.5153, + "step": 23832 + }, + { + "epoch": 0.92, + "learning_rate": 3.475057707104386e-05, + "loss": 0.4005, + "step": 23833 + }, + { + "epoch": 0.92, + "learning_rate": 3.4749935880995126e-05, + "loss": 0.4653, + "step": 23834 + }, + { + "epoch": 0.92, + "learning_rate": 3.47492946909464e-05, + "loss": 0.5525, + "step": 23835 + }, + { + "epoch": 0.92, + "learning_rate": 3.474865350089766e-05, + "loss": 0.5515, + "step": 23836 + }, + { + "epoch": 0.92, + "learning_rate": 3.4748012310848935e-05, + "loss": 0.4553, + "step": 23837 + }, + { + "epoch": 0.92, + "learning_rate": 3.4747371120800206e-05, + "loss": 0.4347, + "step": 23838 + }, + { + "epoch": 0.92, + "learning_rate": 3.474672993075148e-05, + "loss": 0.5799, + "step": 23839 + }, + { + "epoch": 0.92, + "learning_rate": 3.474608874070275e-05, + "loss": 0.5384, + "step": 23840 + }, + { + "epoch": 0.92, + "learning_rate": 3.4745447550654015e-05, + "loss": 0.4887, + "step": 23841 + }, + { + "epoch": 0.92, + "learning_rate": 3.4744806360605286e-05, + "loss": 0.5422, + "step": 23842 + }, + { + "epoch": 0.92, + "learning_rate": 3.474416517055655e-05, + "loss": 0.5305, + "step": 23843 + }, + { + "epoch": 0.92, + "learning_rate": 3.474352398050782e-05, + "loss": 0.5668, + "step": 23844 + }, + { + "epoch": 0.92, + "learning_rate": 3.4742882790459095e-05, + "loss": 0.5335, + "step": 23845 + }, + { + "epoch": 0.92, + "learning_rate": 3.4742241600410366e-05, + "loss": 0.471, + "step": 23846 + }, + { + "epoch": 0.92, + "learning_rate": 3.474160041036164e-05, + "loss": 0.4879, + "step": 23847 + }, + { + "epoch": 0.92, + "learning_rate": 3.47409592203129e-05, + "loss": 0.3806, + "step": 23848 + }, + { + "epoch": 0.92, + "learning_rate": 3.4740318030264175e-05, + "loss": 0.4432, + "step": 23849 + }, + { + "epoch": 0.92, + "learning_rate": 3.473967684021544e-05, + "loss": 0.4205, + "step": 23850 + }, + { + "epoch": 0.92, + "learning_rate": 3.473903565016671e-05, + "loss": 0.5874, + "step": 23851 + }, + { + "epoch": 0.92, + "learning_rate": 3.4738394460117976e-05, + "loss": 0.3771, + "step": 23852 + }, + { + "epoch": 0.92, + "learning_rate": 3.473775327006925e-05, + "loss": 0.5561, + "step": 23853 + }, + { + "epoch": 0.92, + "learning_rate": 3.473711208002052e-05, + "loss": 0.5124, + "step": 23854 + }, + { + "epoch": 0.92, + "learning_rate": 3.473647088997179e-05, + "loss": 0.6006, + "step": 23855 + }, + { + "epoch": 0.92, + "learning_rate": 3.473582969992306e-05, + "loss": 0.4477, + "step": 23856 + }, + { + "epoch": 0.92, + "learning_rate": 3.473518850987433e-05, + "loss": 0.5058, + "step": 23857 + }, + { + "epoch": 0.92, + "learning_rate": 3.47345473198256e-05, + "loss": 0.5192, + "step": 23858 + }, + { + "epoch": 0.92, + "learning_rate": 3.4733906129776865e-05, + "loss": 0.3717, + "step": 23859 + }, + { + "epoch": 0.92, + "learning_rate": 3.4733264939728137e-05, + "loss": 0.499, + "step": 23860 + }, + { + "epoch": 0.92, + "learning_rate": 3.473262374967941e-05, + "loss": 0.5408, + "step": 23861 + }, + { + "epoch": 0.92, + "learning_rate": 3.473198255963068e-05, + "loss": 0.4454, + "step": 23862 + }, + { + "epoch": 0.92, + "learning_rate": 3.4731341369581945e-05, + "loss": 0.4005, + "step": 23863 + }, + { + "epoch": 0.92, + "learning_rate": 3.473070017953322e-05, + "loss": 0.3457, + "step": 23864 + }, + { + "epoch": 0.92, + "learning_rate": 3.473005898948449e-05, + "loss": 0.6263, + "step": 23865 + }, + { + "epoch": 0.92, + "learning_rate": 3.472941779943575e-05, + "loss": 0.3797, + "step": 23866 + }, + { + "epoch": 0.92, + "learning_rate": 3.4728776609387025e-05, + "loss": 0.411, + "step": 23867 + }, + { + "epoch": 0.92, + "learning_rate": 3.472813541933829e-05, + "loss": 0.4975, + "step": 23868 + }, + { + "epoch": 0.92, + "learning_rate": 3.472749422928956e-05, + "loss": 0.5013, + "step": 23869 + }, + { + "epoch": 0.92, + "learning_rate": 3.4726853039240833e-05, + "loss": 0.3722, + "step": 23870 + }, + { + "epoch": 0.92, + "learning_rate": 3.4726211849192105e-05, + "loss": 0.4372, + "step": 23871 + }, + { + "epoch": 0.92, + "learning_rate": 3.472557065914337e-05, + "loss": 0.4282, + "step": 23872 + }, + { + "epoch": 0.92, + "learning_rate": 3.472492946909464e-05, + "loss": 0.3552, + "step": 23873 + }, + { + "epoch": 0.92, + "learning_rate": 3.4724288279045914e-05, + "loss": 0.3917, + "step": 23874 + }, + { + "epoch": 0.92, + "learning_rate": 3.472364708899718e-05, + "loss": 0.5283, + "step": 23875 + }, + { + "epoch": 0.92, + "learning_rate": 3.472300589894845e-05, + "loss": 0.4544, + "step": 23876 + }, + { + "epoch": 0.92, + "learning_rate": 3.4722364708899715e-05, + "loss": 0.4074, + "step": 23877 + }, + { + "epoch": 0.92, + "learning_rate": 3.4721723518850994e-05, + "loss": 0.4174, + "step": 23878 + }, + { + "epoch": 0.92, + "learning_rate": 3.472108232880226e-05, + "loss": 0.3564, + "step": 23879 + }, + { + "epoch": 0.92, + "learning_rate": 3.472044113875353e-05, + "loss": 0.5082, + "step": 23880 + }, + { + "epoch": 0.92, + "learning_rate": 3.4719799948704795e-05, + "loss": 0.4081, + "step": 23881 + }, + { + "epoch": 0.92, + "learning_rate": 3.471915875865607e-05, + "loss": 0.4119, + "step": 23882 + }, + { + "epoch": 0.92, + "learning_rate": 3.471851756860734e-05, + "loss": 0.353, + "step": 23883 + }, + { + "epoch": 0.92, + "learning_rate": 3.4717876378558604e-05, + "loss": 0.5637, + "step": 23884 + }, + { + "epoch": 0.92, + "learning_rate": 3.4717235188509875e-05, + "loss": 0.561, + "step": 23885 + }, + { + "epoch": 0.92, + "learning_rate": 3.471659399846115e-05, + "loss": 0.4756, + "step": 23886 + }, + { + "epoch": 0.92, + "learning_rate": 3.471595280841242e-05, + "loss": 0.5436, + "step": 23887 + }, + { + "epoch": 0.92, + "learning_rate": 3.4715311618363684e-05, + "loss": 0.3691, + "step": 23888 + }, + { + "epoch": 0.92, + "learning_rate": 3.4714670428314955e-05, + "loss": 0.5126, + "step": 23889 + }, + { + "epoch": 0.92, + "learning_rate": 3.471402923826622e-05, + "loss": 0.4644, + "step": 23890 + }, + { + "epoch": 0.92, + "learning_rate": 3.471338804821749e-05, + "loss": 0.5946, + "step": 23891 + }, + { + "epoch": 0.92, + "learning_rate": 3.4712746858168764e-05, + "loss": 0.4441, + "step": 23892 + }, + { + "epoch": 0.92, + "learning_rate": 3.471210566812003e-05, + "loss": 0.4473, + "step": 23893 + }, + { + "epoch": 0.92, + "learning_rate": 3.471146447807131e-05, + "loss": 0.4042, + "step": 23894 + }, + { + "epoch": 0.92, + "learning_rate": 3.471082328802257e-05, + "loss": 0.4987, + "step": 23895 + }, + { + "epoch": 0.92, + "learning_rate": 3.4710182097973844e-05, + "loss": 0.4828, + "step": 23896 + }, + { + "epoch": 0.92, + "learning_rate": 3.470954090792511e-05, + "loss": 0.6446, + "step": 23897 + }, + { + "epoch": 0.92, + "learning_rate": 3.470889971787638e-05, + "loss": 0.743, + "step": 23898 + }, + { + "epoch": 0.92, + "learning_rate": 3.4708258527827645e-05, + "loss": 0.3833, + "step": 23899 + }, + { + "epoch": 0.92, + "learning_rate": 3.470761733777892e-05, + "loss": 0.5885, + "step": 23900 + }, + { + "epoch": 0.92, + "learning_rate": 3.470697614773019e-05, + "loss": 0.4605, + "step": 23901 + }, + { + "epoch": 0.92, + "learning_rate": 3.470633495768146e-05, + "loss": 0.3981, + "step": 23902 + }, + { + "epoch": 0.92, + "learning_rate": 3.470569376763273e-05, + "loss": 0.4181, + "step": 23903 + }, + { + "epoch": 0.92, + "learning_rate": 3.4705052577584e-05, + "loss": 0.4747, + "step": 23904 + }, + { + "epoch": 0.92, + "learning_rate": 3.470441138753527e-05, + "loss": 0.4109, + "step": 23905 + }, + { + "epoch": 0.92, + "learning_rate": 3.4703770197486534e-05, + "loss": 0.4457, + "step": 23906 + }, + { + "epoch": 0.92, + "learning_rate": 3.4703129007437806e-05, + "loss": 0.4674, + "step": 23907 + }, + { + "epoch": 0.92, + "learning_rate": 3.470248781738907e-05, + "loss": 0.4582, + "step": 23908 + }, + { + "epoch": 0.92, + "learning_rate": 3.470184662734034e-05, + "loss": 0.5108, + "step": 23909 + }, + { + "epoch": 0.92, + "learning_rate": 3.4701205437291614e-05, + "loss": 0.5409, + "step": 23910 + }, + { + "epoch": 0.92, + "learning_rate": 3.4700564247242886e-05, + "loss": 0.4142, + "step": 23911 + }, + { + "epoch": 0.92, + "learning_rate": 3.469992305719416e-05, + "loss": 0.5741, + "step": 23912 + }, + { + "epoch": 0.92, + "learning_rate": 3.469928186714542e-05, + "loss": 0.534, + "step": 23913 + }, + { + "epoch": 0.92, + "learning_rate": 3.4698640677096694e-05, + "loss": 0.4795, + "step": 23914 + }, + { + "epoch": 0.92, + "learning_rate": 3.469799948704796e-05, + "loss": 0.5097, + "step": 23915 + }, + { + "epoch": 0.92, + "learning_rate": 3.469735829699923e-05, + "loss": 0.609, + "step": 23916 + }, + { + "epoch": 0.92, + "learning_rate": 3.46967171069505e-05, + "loss": 0.439, + "step": 23917 + }, + { + "epoch": 0.92, + "learning_rate": 3.4696075916901774e-05, + "loss": 0.3886, + "step": 23918 + }, + { + "epoch": 0.92, + "learning_rate": 3.4695434726853046e-05, + "loss": 0.443, + "step": 23919 + }, + { + "epoch": 0.92, + "learning_rate": 3.469479353680431e-05, + "loss": 0.4498, + "step": 23920 + }, + { + "epoch": 0.92, + "learning_rate": 3.469415234675558e-05, + "loss": 0.488, + "step": 23921 + }, + { + "epoch": 0.92, + "learning_rate": 3.469351115670685e-05, + "loss": 0.5793, + "step": 23922 + }, + { + "epoch": 0.92, + "learning_rate": 3.469286996665812e-05, + "loss": 0.5069, + "step": 23923 + }, + { + "epoch": 0.92, + "learning_rate": 3.4692228776609384e-05, + "loss": 0.4649, + "step": 23924 + }, + { + "epoch": 0.92, + "learning_rate": 3.4691587586560656e-05, + "loss": 0.4783, + "step": 23925 + }, + { + "epoch": 0.92, + "learning_rate": 3.469094639651193e-05, + "loss": 0.5528, + "step": 23926 + }, + { + "epoch": 0.92, + "learning_rate": 3.46903052064632e-05, + "loss": 0.5011, + "step": 23927 + }, + { + "epoch": 0.92, + "learning_rate": 3.468966401641447e-05, + "loss": 0.5465, + "step": 23928 + }, + { + "epoch": 0.92, + "learning_rate": 3.4689022826365736e-05, + "loss": 0.4707, + "step": 23929 + }, + { + "epoch": 0.92, + "learning_rate": 3.468838163631701e-05, + "loss": 0.5052, + "step": 23930 + }, + { + "epoch": 0.92, + "learning_rate": 3.468774044626827e-05, + "loss": 0.4295, + "step": 23931 + }, + { + "epoch": 0.92, + "learning_rate": 3.4687099256219544e-05, + "loss": 0.395, + "step": 23932 + }, + { + "epoch": 0.92, + "learning_rate": 3.4686458066170816e-05, + "loss": 0.4952, + "step": 23933 + }, + { + "epoch": 0.92, + "learning_rate": 3.468581687612209e-05, + "loss": 0.3994, + "step": 23934 + }, + { + "epoch": 0.92, + "learning_rate": 3.468517568607335e-05, + "loss": 0.4574, + "step": 23935 + }, + { + "epoch": 0.92, + "learning_rate": 3.4684534496024624e-05, + "loss": 0.5548, + "step": 23936 + }, + { + "epoch": 0.92, + "learning_rate": 3.4683893305975896e-05, + "loss": 0.4974, + "step": 23937 + }, + { + "epoch": 0.92, + "learning_rate": 3.468325211592716e-05, + "loss": 0.4961, + "step": 23938 + }, + { + "epoch": 0.92, + "learning_rate": 3.468261092587843e-05, + "loss": 0.5076, + "step": 23939 + }, + { + "epoch": 0.92, + "learning_rate": 3.46819697358297e-05, + "loss": 0.5181, + "step": 23940 + }, + { + "epoch": 0.92, + "learning_rate": 3.468132854578097e-05, + "loss": 0.4396, + "step": 23941 + }, + { + "epoch": 0.92, + "learning_rate": 3.468068735573224e-05, + "loss": 0.4173, + "step": 23942 + }, + { + "epoch": 0.92, + "learning_rate": 3.468004616568351e-05, + "loss": 0.4629, + "step": 23943 + }, + { + "epoch": 0.92, + "learning_rate": 3.467940497563478e-05, + "loss": 0.575, + "step": 23944 + }, + { + "epoch": 0.92, + "learning_rate": 3.467876378558605e-05, + "loss": 0.5809, + "step": 23945 + }, + { + "epoch": 0.92, + "learning_rate": 3.467812259553732e-05, + "loss": 0.5249, + "step": 23946 + }, + { + "epoch": 0.92, + "learning_rate": 3.4677481405488586e-05, + "loss": 0.4787, + "step": 23947 + }, + { + "epoch": 0.92, + "learning_rate": 3.467684021543986e-05, + "loss": 0.482, + "step": 23948 + }, + { + "epoch": 0.92, + "learning_rate": 3.467619902539112e-05, + "loss": 0.4245, + "step": 23949 + }, + { + "epoch": 0.92, + "learning_rate": 3.46755578353424e-05, + "loss": 0.3951, + "step": 23950 + }, + { + "epoch": 0.92, + "learning_rate": 3.4674916645293666e-05, + "loss": 0.4543, + "step": 23951 + }, + { + "epoch": 0.92, + "learning_rate": 3.467427545524494e-05, + "loss": 0.3963, + "step": 23952 + }, + { + "epoch": 0.92, + "learning_rate": 3.46736342651962e-05, + "loss": 0.4769, + "step": 23953 + }, + { + "epoch": 0.92, + "learning_rate": 3.4672993075147475e-05, + "loss": 0.5059, + "step": 23954 + }, + { + "epoch": 0.92, + "learning_rate": 3.4672351885098746e-05, + "loss": 0.6343, + "step": 23955 + }, + { + "epoch": 0.92, + "learning_rate": 3.467171069505001e-05, + "loss": 0.4895, + "step": 23956 + }, + { + "epoch": 0.92, + "learning_rate": 3.467106950500128e-05, + "loss": 0.4594, + "step": 23957 + }, + { + "epoch": 0.92, + "learning_rate": 3.4670428314952555e-05, + "loss": 0.4527, + "step": 23958 + }, + { + "epoch": 0.92, + "learning_rate": 3.4669787124903826e-05, + "loss": 0.4602, + "step": 23959 + }, + { + "epoch": 0.92, + "learning_rate": 3.466914593485509e-05, + "loss": 0.4176, + "step": 23960 + }, + { + "epoch": 0.92, + "learning_rate": 3.466850474480636e-05, + "loss": 0.3996, + "step": 23961 + }, + { + "epoch": 0.92, + "learning_rate": 3.466786355475763e-05, + "loss": 0.5078, + "step": 23962 + }, + { + "epoch": 0.92, + "learning_rate": 3.46672223647089e-05, + "loss": 0.4652, + "step": 23963 + }, + { + "epoch": 0.92, + "learning_rate": 3.466658117466017e-05, + "loss": 0.5409, + "step": 23964 + }, + { + "epoch": 0.92, + "learning_rate": 3.4665939984611436e-05, + "loss": 0.5794, + "step": 23965 + }, + { + "epoch": 0.92, + "learning_rate": 3.4665298794562715e-05, + "loss": 0.5236, + "step": 23966 + }, + { + "epoch": 0.92, + "learning_rate": 3.466465760451398e-05, + "loss": 0.626, + "step": 23967 + }, + { + "epoch": 0.92, + "learning_rate": 3.466401641446525e-05, + "loss": 0.4661, + "step": 23968 + }, + { + "epoch": 0.92, + "learning_rate": 3.4663375224416517e-05, + "loss": 0.3414, + "step": 23969 + }, + { + "epoch": 0.92, + "learning_rate": 3.466273403436779e-05, + "loss": 0.5547, + "step": 23970 + }, + { + "epoch": 0.92, + "learning_rate": 3.466209284431905e-05, + "loss": 0.4622, + "step": 23971 + }, + { + "epoch": 0.92, + "learning_rate": 3.4661451654270325e-05, + "loss": 0.469, + "step": 23972 + }, + { + "epoch": 0.92, + "learning_rate": 3.46608104642216e-05, + "loss": 0.503, + "step": 23973 + }, + { + "epoch": 0.92, + "learning_rate": 3.466016927417287e-05, + "loss": 0.4426, + "step": 23974 + }, + { + "epoch": 0.92, + "learning_rate": 3.465952808412414e-05, + "loss": 0.328, + "step": 23975 + }, + { + "epoch": 0.92, + "learning_rate": 3.4658886894075405e-05, + "loss": 0.5163, + "step": 23976 + }, + { + "epoch": 0.92, + "learning_rate": 3.465824570402668e-05, + "loss": 0.5641, + "step": 23977 + }, + { + "epoch": 0.92, + "learning_rate": 3.465760451397794e-05, + "loss": 0.4847, + "step": 23978 + }, + { + "epoch": 0.92, + "learning_rate": 3.465696332392921e-05, + "loss": 0.4658, + "step": 23979 + }, + { + "epoch": 0.92, + "learning_rate": 3.465632213388048e-05, + "loss": 0.5126, + "step": 23980 + }, + { + "epoch": 0.92, + "learning_rate": 3.465568094383175e-05, + "loss": 0.4284, + "step": 23981 + }, + { + "epoch": 0.92, + "learning_rate": 3.465503975378303e-05, + "loss": 0.492, + "step": 23982 + }, + { + "epoch": 0.92, + "learning_rate": 3.4654398563734293e-05, + "loss": 0.3962, + "step": 23983 + }, + { + "epoch": 0.92, + "learning_rate": 3.4653757373685565e-05, + "loss": 0.4767, + "step": 23984 + }, + { + "epoch": 0.92, + "learning_rate": 3.465311618363683e-05, + "loss": 0.5224, + "step": 23985 + }, + { + "epoch": 0.92, + "learning_rate": 3.46524749935881e-05, + "loss": 0.4934, + "step": 23986 + }, + { + "epoch": 0.92, + "learning_rate": 3.465183380353937e-05, + "loss": 0.4015, + "step": 23987 + }, + { + "epoch": 0.92, + "learning_rate": 3.465119261349064e-05, + "loss": 0.5895, + "step": 23988 + }, + { + "epoch": 0.92, + "learning_rate": 3.465055142344191e-05, + "loss": 0.4356, + "step": 23989 + }, + { + "epoch": 0.92, + "learning_rate": 3.464991023339318e-05, + "loss": 0.5813, + "step": 23990 + }, + { + "epoch": 0.92, + "learning_rate": 3.4649269043344454e-05, + "loss": 0.4627, + "step": 23991 + }, + { + "epoch": 0.92, + "learning_rate": 3.464862785329572e-05, + "loss": 0.5327, + "step": 23992 + }, + { + "epoch": 0.92, + "learning_rate": 3.464798666324699e-05, + "loss": 0.6922, + "step": 23993 + }, + { + "epoch": 0.92, + "learning_rate": 3.4647345473198255e-05, + "loss": 0.6458, + "step": 23994 + }, + { + "epoch": 0.92, + "learning_rate": 3.464670428314953e-05, + "loss": 0.4702, + "step": 23995 + }, + { + "epoch": 0.92, + "learning_rate": 3.464606309310079e-05, + "loss": 0.5032, + "step": 23996 + }, + { + "epoch": 0.92, + "learning_rate": 3.4645421903052064e-05, + "loss": 0.4393, + "step": 23997 + }, + { + "epoch": 0.92, + "learning_rate": 3.4644780713003335e-05, + "loss": 0.4459, + "step": 23998 + }, + { + "epoch": 0.92, + "learning_rate": 3.464413952295461e-05, + "loss": 0.4747, + "step": 23999 + }, + { + "epoch": 0.92, + "learning_rate": 3.464349833290588e-05, + "loss": 0.5039, + "step": 24000 + }, + { + "epoch": 0.92, + "learning_rate": 3.4642857142857144e-05, + "loss": 0.3746, + "step": 24001 + }, + { + "epoch": 0.92, + "learning_rate": 3.4642215952808415e-05, + "loss": 0.4505, + "step": 24002 + }, + { + "epoch": 0.92, + "learning_rate": 3.464157476275968e-05, + "loss": 0.4402, + "step": 24003 + }, + { + "epoch": 0.92, + "learning_rate": 3.464093357271095e-05, + "loss": 0.527, + "step": 24004 + }, + { + "epoch": 0.92, + "learning_rate": 3.4640292382662224e-05, + "loss": 0.5296, + "step": 24005 + }, + { + "epoch": 0.92, + "learning_rate": 3.4639651192613496e-05, + "loss": 0.4339, + "step": 24006 + }, + { + "epoch": 0.92, + "learning_rate": 3.463901000256477e-05, + "loss": 0.5531, + "step": 24007 + }, + { + "epoch": 0.92, + "learning_rate": 3.463836881251603e-05, + "loss": 0.6758, + "step": 24008 + }, + { + "epoch": 0.92, + "learning_rate": 3.4637727622467304e-05, + "loss": 0.4041, + "step": 24009 + }, + { + "epoch": 0.92, + "learning_rate": 3.463708643241857e-05, + "loss": 0.574, + "step": 24010 + }, + { + "epoch": 0.92, + "learning_rate": 3.463644524236984e-05, + "loss": 0.5369, + "step": 24011 + }, + { + "epoch": 0.92, + "learning_rate": 3.4635804052321105e-05, + "loss": 0.4653, + "step": 24012 + }, + { + "epoch": 0.92, + "learning_rate": 3.463516286227238e-05, + "loss": 0.4954, + "step": 24013 + }, + { + "epoch": 0.92, + "learning_rate": 3.463452167222365e-05, + "loss": 0.6168, + "step": 24014 + }, + { + "epoch": 0.92, + "learning_rate": 3.463388048217492e-05, + "loss": 0.4127, + "step": 24015 + }, + { + "epoch": 0.92, + "learning_rate": 3.463323929212619e-05, + "loss": 0.4895, + "step": 24016 + }, + { + "epoch": 0.92, + "learning_rate": 3.463259810207746e-05, + "loss": 0.4064, + "step": 24017 + }, + { + "epoch": 0.92, + "learning_rate": 3.463195691202873e-05, + "loss": 0.4482, + "step": 24018 + }, + { + "epoch": 0.92, + "learning_rate": 3.4631315721979994e-05, + "loss": 0.5728, + "step": 24019 + }, + { + "epoch": 0.92, + "learning_rate": 3.4630674531931266e-05, + "loss": 0.5266, + "step": 24020 + }, + { + "epoch": 0.92, + "learning_rate": 3.463003334188254e-05, + "loss": 0.6015, + "step": 24021 + }, + { + "epoch": 0.92, + "learning_rate": 3.462939215183381e-05, + "loss": 0.4306, + "step": 24022 + }, + { + "epoch": 0.92, + "learning_rate": 3.4628750961785074e-05, + "loss": 0.5749, + "step": 24023 + }, + { + "epoch": 0.92, + "learning_rate": 3.4628109771736346e-05, + "loss": 0.3833, + "step": 24024 + }, + { + "epoch": 0.92, + "learning_rate": 3.462746858168762e-05, + "loss": 0.4863, + "step": 24025 + }, + { + "epoch": 0.92, + "learning_rate": 3.462682739163888e-05, + "loss": 0.5691, + "step": 24026 + }, + { + "epoch": 0.92, + "learning_rate": 3.4626186201590154e-05, + "loss": 0.5403, + "step": 24027 + }, + { + "epoch": 0.92, + "learning_rate": 3.462554501154142e-05, + "loss": 0.4186, + "step": 24028 + }, + { + "epoch": 0.92, + "learning_rate": 3.462490382149269e-05, + "loss": 0.4423, + "step": 24029 + }, + { + "epoch": 0.92, + "learning_rate": 3.462426263144396e-05, + "loss": 0.5798, + "step": 24030 + }, + { + "epoch": 0.92, + "learning_rate": 3.4623621441395234e-05, + "loss": 0.6509, + "step": 24031 + }, + { + "epoch": 0.92, + "learning_rate": 3.46229802513465e-05, + "loss": 0.4826, + "step": 24032 + }, + { + "epoch": 0.92, + "learning_rate": 3.462233906129777e-05, + "loss": 0.5344, + "step": 24033 + }, + { + "epoch": 0.92, + "learning_rate": 3.462169787124904e-05, + "loss": 0.5106, + "step": 24034 + }, + { + "epoch": 0.92, + "learning_rate": 3.462105668120031e-05, + "loss": 0.5415, + "step": 24035 + }, + { + "epoch": 0.92, + "learning_rate": 3.462041549115158e-05, + "loss": 0.5619, + "step": 24036 + }, + { + "epoch": 0.92, + "learning_rate": 3.4619774301102844e-05, + "loss": 0.5103, + "step": 24037 + }, + { + "epoch": 0.92, + "learning_rate": 3.461913311105412e-05, + "loss": 0.5316, + "step": 24038 + }, + { + "epoch": 0.92, + "learning_rate": 3.461849192100539e-05, + "loss": 0.4004, + "step": 24039 + }, + { + "epoch": 0.92, + "learning_rate": 3.461785073095666e-05, + "loss": 0.6056, + "step": 24040 + }, + { + "epoch": 0.92, + "learning_rate": 3.4617209540907924e-05, + "loss": 0.4219, + "step": 24041 + }, + { + "epoch": 0.92, + "learning_rate": 3.4616568350859196e-05, + "loss": 0.5405, + "step": 24042 + }, + { + "epoch": 0.92, + "learning_rate": 3.461592716081047e-05, + "loss": 0.5468, + "step": 24043 + }, + { + "epoch": 0.92, + "learning_rate": 3.461528597076173e-05, + "loss": 0.4137, + "step": 24044 + }, + { + "epoch": 0.92, + "learning_rate": 3.4614644780713004e-05, + "loss": 0.4554, + "step": 24045 + }, + { + "epoch": 0.92, + "learning_rate": 3.4614003590664276e-05, + "loss": 0.4623, + "step": 24046 + }, + { + "epoch": 0.92, + "learning_rate": 3.461336240061555e-05, + "loss": 0.5304, + "step": 24047 + }, + { + "epoch": 0.92, + "learning_rate": 3.461272121056681e-05, + "loss": 0.4322, + "step": 24048 + }, + { + "epoch": 0.92, + "learning_rate": 3.4612080020518084e-05, + "loss": 0.5461, + "step": 24049 + }, + { + "epoch": 0.92, + "learning_rate": 3.461143883046935e-05, + "loss": 0.453, + "step": 24050 + }, + { + "epoch": 0.92, + "learning_rate": 3.461079764042062e-05, + "loss": 0.4666, + "step": 24051 + }, + { + "epoch": 0.92, + "learning_rate": 3.461015645037189e-05, + "loss": 0.3882, + "step": 24052 + }, + { + "epoch": 0.92, + "learning_rate": 3.460951526032316e-05, + "loss": 0.5523, + "step": 24053 + }, + { + "epoch": 0.92, + "learning_rate": 3.4608874070274436e-05, + "loss": 0.5164, + "step": 24054 + }, + { + "epoch": 0.92, + "learning_rate": 3.46082328802257e-05, + "loss": 0.4865, + "step": 24055 + }, + { + "epoch": 0.92, + "learning_rate": 3.460759169017697e-05, + "loss": 0.4788, + "step": 24056 + }, + { + "epoch": 0.92, + "learning_rate": 3.460695050012824e-05, + "loss": 0.6477, + "step": 24057 + }, + { + "epoch": 0.92, + "learning_rate": 3.460630931007951e-05, + "loss": 0.5276, + "step": 24058 + }, + { + "epoch": 0.92, + "learning_rate": 3.4605668120030775e-05, + "loss": 0.4526, + "step": 24059 + }, + { + "epoch": 0.93, + "learning_rate": 3.4605026929982046e-05, + "loss": 0.5302, + "step": 24060 + }, + { + "epoch": 0.93, + "learning_rate": 3.460438573993332e-05, + "loss": 0.5605, + "step": 24061 + }, + { + "epoch": 0.93, + "learning_rate": 3.460374454988459e-05, + "loss": 0.4483, + "step": 24062 + }, + { + "epoch": 0.93, + "learning_rate": 3.460310335983586e-05, + "loss": 0.4683, + "step": 24063 + }, + { + "epoch": 0.93, + "learning_rate": 3.4602462169787126e-05, + "loss": 0.4563, + "step": 24064 + }, + { + "epoch": 0.93, + "learning_rate": 3.46018209797384e-05, + "loss": 0.4364, + "step": 24065 + }, + { + "epoch": 0.93, + "learning_rate": 3.460117978968966e-05, + "loss": 0.4245, + "step": 24066 + }, + { + "epoch": 0.93, + "learning_rate": 3.4600538599640935e-05, + "loss": 0.4681, + "step": 24067 + }, + { + "epoch": 0.93, + "learning_rate": 3.45998974095922e-05, + "loss": 0.4258, + "step": 24068 + }, + { + "epoch": 0.93, + "learning_rate": 3.459925621954347e-05, + "loss": 0.5881, + "step": 24069 + }, + { + "epoch": 0.93, + "learning_rate": 3.459861502949474e-05, + "loss": 0.4773, + "step": 24070 + }, + { + "epoch": 0.93, + "learning_rate": 3.4597973839446015e-05, + "loss": 0.4413, + "step": 24071 + }, + { + "epoch": 0.93, + "learning_rate": 3.4597332649397286e-05, + "loss": 0.584, + "step": 24072 + }, + { + "epoch": 0.93, + "learning_rate": 3.459669145934855e-05, + "loss": 0.4467, + "step": 24073 + }, + { + "epoch": 0.93, + "learning_rate": 3.459605026929982e-05, + "loss": 0.4748, + "step": 24074 + }, + { + "epoch": 0.93, + "learning_rate": 3.459540907925109e-05, + "loss": 0.4617, + "step": 24075 + }, + { + "epoch": 0.93, + "learning_rate": 3.459476788920236e-05, + "loss": 0.5442, + "step": 24076 + }, + { + "epoch": 0.93, + "learning_rate": 3.459412669915363e-05, + "loss": 0.4809, + "step": 24077 + }, + { + "epoch": 0.93, + "learning_rate": 3.45934855091049e-05, + "loss": 0.4847, + "step": 24078 + }, + { + "epoch": 0.93, + "learning_rate": 3.4592844319056175e-05, + "loss": 0.5163, + "step": 24079 + }, + { + "epoch": 0.93, + "learning_rate": 3.459220312900744e-05, + "loss": 0.4512, + "step": 24080 + }, + { + "epoch": 0.93, + "learning_rate": 3.459156193895871e-05, + "loss": 0.4852, + "step": 24081 + }, + { + "epoch": 0.93, + "learning_rate": 3.4590920748909977e-05, + "loss": 0.3235, + "step": 24082 + }, + { + "epoch": 0.93, + "learning_rate": 3.459027955886125e-05, + "loss": 0.4694, + "step": 24083 + }, + { + "epoch": 0.93, + "learning_rate": 3.458963836881251e-05, + "loss": 0.5161, + "step": 24084 + }, + { + "epoch": 0.93, + "learning_rate": 3.4588997178763785e-05, + "loss": 0.4753, + "step": 24085 + }, + { + "epoch": 0.93, + "learning_rate": 3.458835598871506e-05, + "loss": 0.5208, + "step": 24086 + }, + { + "epoch": 0.93, + "learning_rate": 3.458771479866633e-05, + "loss": 0.5328, + "step": 24087 + }, + { + "epoch": 0.93, + "learning_rate": 3.45870736086176e-05, + "loss": 0.4339, + "step": 24088 + }, + { + "epoch": 0.93, + "learning_rate": 3.4586432418568865e-05, + "loss": 0.4657, + "step": 24089 + }, + { + "epoch": 0.93, + "learning_rate": 3.458579122852014e-05, + "loss": 0.5833, + "step": 24090 + }, + { + "epoch": 0.93, + "learning_rate": 3.45851500384714e-05, + "loss": 0.4443, + "step": 24091 + }, + { + "epoch": 0.93, + "learning_rate": 3.4584508848422673e-05, + "loss": 0.4278, + "step": 24092 + }, + { + "epoch": 0.93, + "learning_rate": 3.4583867658373945e-05, + "loss": 0.4677, + "step": 24093 + }, + { + "epoch": 0.93, + "learning_rate": 3.458322646832522e-05, + "loss": 0.6457, + "step": 24094 + }, + { + "epoch": 0.93, + "learning_rate": 3.458258527827648e-05, + "loss": 0.4486, + "step": 24095 + }, + { + "epoch": 0.93, + "learning_rate": 3.4581944088227753e-05, + "loss": 0.5142, + "step": 24096 + }, + { + "epoch": 0.93, + "learning_rate": 3.4581302898179025e-05, + "loss": 0.4276, + "step": 24097 + }, + { + "epoch": 0.93, + "learning_rate": 3.458066170813029e-05, + "loss": 0.5134, + "step": 24098 + }, + { + "epoch": 0.93, + "learning_rate": 3.458002051808156e-05, + "loss": 0.5554, + "step": 24099 + }, + { + "epoch": 0.93, + "learning_rate": 3.457937932803283e-05, + "loss": 0.4567, + "step": 24100 + }, + { + "epoch": 0.93, + "learning_rate": 3.45787381379841e-05, + "loss": 0.4922, + "step": 24101 + }, + { + "epoch": 0.93, + "learning_rate": 3.457809694793537e-05, + "loss": 0.539, + "step": 24102 + }, + { + "epoch": 0.93, + "learning_rate": 3.457745575788664e-05, + "loss": 0.5896, + "step": 24103 + }, + { + "epoch": 0.93, + "learning_rate": 3.457681456783791e-05, + "loss": 0.5175, + "step": 24104 + }, + { + "epoch": 0.93, + "learning_rate": 3.457617337778918e-05, + "loss": 0.4867, + "step": 24105 + }, + { + "epoch": 0.93, + "learning_rate": 3.457553218774045e-05, + "loss": 0.4613, + "step": 24106 + }, + { + "epoch": 0.93, + "learning_rate": 3.4574890997691715e-05, + "loss": 0.4208, + "step": 24107 + }, + { + "epoch": 0.93, + "learning_rate": 3.457424980764299e-05, + "loss": 0.4575, + "step": 24108 + }, + { + "epoch": 0.93, + "learning_rate": 3.457360861759425e-05, + "loss": 0.3985, + "step": 24109 + }, + { + "epoch": 0.93, + "learning_rate": 3.457296742754553e-05, + "loss": 0.4241, + "step": 24110 + }, + { + "epoch": 0.93, + "learning_rate": 3.4572326237496795e-05, + "loss": 0.5204, + "step": 24111 + }, + { + "epoch": 0.93, + "learning_rate": 3.457168504744807e-05, + "loss": 0.42, + "step": 24112 + }, + { + "epoch": 0.93, + "learning_rate": 3.457104385739933e-05, + "loss": 0.4825, + "step": 24113 + }, + { + "epoch": 0.93, + "learning_rate": 3.4570402667350604e-05, + "loss": 0.5606, + "step": 24114 + }, + { + "epoch": 0.93, + "learning_rate": 3.4569761477301875e-05, + "loss": 0.534, + "step": 24115 + }, + { + "epoch": 0.93, + "learning_rate": 3.456912028725314e-05, + "loss": 0.6644, + "step": 24116 + }, + { + "epoch": 0.93, + "learning_rate": 3.456847909720441e-05, + "loss": 0.6599, + "step": 24117 + }, + { + "epoch": 0.93, + "learning_rate": 3.4567837907155684e-05, + "loss": 0.4515, + "step": 24118 + }, + { + "epoch": 0.93, + "learning_rate": 3.4567196717106956e-05, + "loss": 0.4685, + "step": 24119 + }, + { + "epoch": 0.93, + "learning_rate": 3.456655552705822e-05, + "loss": 0.566, + "step": 24120 + }, + { + "epoch": 0.93, + "learning_rate": 3.456591433700949e-05, + "loss": 0.6215, + "step": 24121 + }, + { + "epoch": 0.93, + "learning_rate": 3.456527314696076e-05, + "loss": 0.5136, + "step": 24122 + }, + { + "epoch": 0.93, + "learning_rate": 3.456463195691203e-05, + "loss": 0.4703, + "step": 24123 + }, + { + "epoch": 0.93, + "learning_rate": 3.45639907668633e-05, + "loss": 0.5105, + "step": 24124 + }, + { + "epoch": 0.93, + "learning_rate": 3.4563349576814566e-05, + "loss": 0.5074, + "step": 24125 + }, + { + "epoch": 0.93, + "learning_rate": 3.4562708386765844e-05, + "loss": 0.4581, + "step": 24126 + }, + { + "epoch": 0.93, + "learning_rate": 3.456206719671711e-05, + "loss": 0.4407, + "step": 24127 + }, + { + "epoch": 0.93, + "learning_rate": 3.456142600666838e-05, + "loss": 0.4492, + "step": 24128 + }, + { + "epoch": 0.93, + "learning_rate": 3.4560784816619646e-05, + "loss": 0.4949, + "step": 24129 + }, + { + "epoch": 0.93, + "learning_rate": 3.456014362657092e-05, + "loss": 0.5483, + "step": 24130 + }, + { + "epoch": 0.93, + "learning_rate": 3.455950243652218e-05, + "loss": 0.4512, + "step": 24131 + }, + { + "epoch": 0.93, + "learning_rate": 3.4558861246473454e-05, + "loss": 0.4468, + "step": 24132 + }, + { + "epoch": 0.93, + "learning_rate": 3.4558220056424726e-05, + "loss": 0.5238, + "step": 24133 + }, + { + "epoch": 0.93, + "learning_rate": 3.4557578866376e-05, + "loss": 0.6196, + "step": 24134 + }, + { + "epoch": 0.93, + "learning_rate": 3.455693767632727e-05, + "loss": 0.5494, + "step": 24135 + }, + { + "epoch": 0.93, + "learning_rate": 3.4556296486278534e-05, + "loss": 0.525, + "step": 24136 + }, + { + "epoch": 0.93, + "learning_rate": 3.4555655296229806e-05, + "loss": 0.4896, + "step": 24137 + }, + { + "epoch": 0.93, + "learning_rate": 3.455501410618107e-05, + "loss": 0.4467, + "step": 24138 + }, + { + "epoch": 0.93, + "learning_rate": 3.455437291613234e-05, + "loss": 0.4443, + "step": 24139 + }, + { + "epoch": 0.93, + "learning_rate": 3.455373172608361e-05, + "loss": 0.4327, + "step": 24140 + }, + { + "epoch": 0.93, + "learning_rate": 3.455309053603488e-05, + "loss": 0.5366, + "step": 24141 + }, + { + "epoch": 0.93, + "learning_rate": 3.455244934598616e-05, + "loss": 0.3717, + "step": 24142 + }, + { + "epoch": 0.93, + "learning_rate": 3.455180815593742e-05, + "loss": 0.3668, + "step": 24143 + }, + { + "epoch": 0.93, + "learning_rate": 3.4551166965888694e-05, + "loss": 0.5517, + "step": 24144 + }, + { + "epoch": 0.93, + "learning_rate": 3.455052577583996e-05, + "loss": 0.5814, + "step": 24145 + }, + { + "epoch": 0.93, + "learning_rate": 3.454988458579123e-05, + "loss": 0.5195, + "step": 24146 + }, + { + "epoch": 0.93, + "learning_rate": 3.4549243395742496e-05, + "loss": 0.5306, + "step": 24147 + }, + { + "epoch": 0.93, + "learning_rate": 3.454860220569377e-05, + "loss": 0.4418, + "step": 24148 + }, + { + "epoch": 0.93, + "learning_rate": 3.454796101564504e-05, + "loss": 0.4473, + "step": 24149 + }, + { + "epoch": 0.93, + "learning_rate": 3.454731982559631e-05, + "loss": 0.5156, + "step": 24150 + }, + { + "epoch": 0.93, + "learning_rate": 3.454667863554758e-05, + "loss": 0.4748, + "step": 24151 + }, + { + "epoch": 0.93, + "learning_rate": 3.454603744549885e-05, + "loss": 0.6099, + "step": 24152 + }, + { + "epoch": 0.93, + "learning_rate": 3.454539625545012e-05, + "loss": 0.4567, + "step": 24153 + }, + { + "epoch": 0.93, + "learning_rate": 3.4544755065401384e-05, + "loss": 0.3817, + "step": 24154 + }, + { + "epoch": 0.93, + "learning_rate": 3.4544113875352656e-05, + "loss": 0.5129, + "step": 24155 + }, + { + "epoch": 0.93, + "learning_rate": 3.454347268530392e-05, + "loss": 0.4328, + "step": 24156 + }, + { + "epoch": 0.93, + "learning_rate": 3.454283149525519e-05, + "loss": 0.4876, + "step": 24157 + }, + { + "epoch": 0.93, + "learning_rate": 3.4542190305206464e-05, + "loss": 0.4498, + "step": 24158 + }, + { + "epoch": 0.93, + "learning_rate": 3.4541549115157736e-05, + "loss": 0.3735, + "step": 24159 + }, + { + "epoch": 0.93, + "learning_rate": 3.454090792510901e-05, + "loss": 0.4906, + "step": 24160 + }, + { + "epoch": 0.93, + "learning_rate": 3.454026673506027e-05, + "loss": 0.569, + "step": 24161 + }, + { + "epoch": 0.93, + "learning_rate": 3.4539625545011544e-05, + "loss": 0.4557, + "step": 24162 + }, + { + "epoch": 0.93, + "learning_rate": 3.453898435496281e-05, + "loss": 0.6386, + "step": 24163 + }, + { + "epoch": 0.93, + "learning_rate": 3.453834316491408e-05, + "loss": 0.5482, + "step": 24164 + }, + { + "epoch": 0.93, + "learning_rate": 3.453770197486535e-05, + "loss": 0.5136, + "step": 24165 + }, + { + "epoch": 0.93, + "learning_rate": 3.4537060784816625e-05, + "loss": 0.4759, + "step": 24166 + }, + { + "epoch": 0.93, + "learning_rate": 3.453641959476789e-05, + "loss": 0.4398, + "step": 24167 + }, + { + "epoch": 0.93, + "learning_rate": 3.453577840471916e-05, + "loss": 0.5377, + "step": 24168 + }, + { + "epoch": 0.93, + "learning_rate": 3.453513721467043e-05, + "loss": 0.432, + "step": 24169 + }, + { + "epoch": 0.93, + "learning_rate": 3.45344960246217e-05, + "loss": 0.5139, + "step": 24170 + }, + { + "epoch": 0.93, + "learning_rate": 3.453385483457297e-05, + "loss": 0.4798, + "step": 24171 + }, + { + "epoch": 0.93, + "learning_rate": 3.4533213644524235e-05, + "loss": 0.5798, + "step": 24172 + }, + { + "epoch": 0.93, + "learning_rate": 3.4532572454475506e-05, + "loss": 0.483, + "step": 24173 + }, + { + "epoch": 0.93, + "learning_rate": 3.453193126442678e-05, + "loss": 0.3694, + "step": 24174 + }, + { + "epoch": 0.93, + "learning_rate": 3.453129007437805e-05, + "loss": 0.4654, + "step": 24175 + }, + { + "epoch": 0.93, + "learning_rate": 3.4530648884329315e-05, + "loss": 0.5249, + "step": 24176 + }, + { + "epoch": 0.93, + "learning_rate": 3.4530007694280586e-05, + "loss": 0.4788, + "step": 24177 + }, + { + "epoch": 0.93, + "learning_rate": 3.452936650423186e-05, + "loss": 0.4287, + "step": 24178 + }, + { + "epoch": 0.93, + "learning_rate": 3.452872531418312e-05, + "loss": 0.4848, + "step": 24179 + }, + { + "epoch": 0.93, + "learning_rate": 3.4528084124134395e-05, + "loss": 0.461, + "step": 24180 + }, + { + "epoch": 0.93, + "learning_rate": 3.4527442934085666e-05, + "loss": 0.4684, + "step": 24181 + }, + { + "epoch": 0.93, + "learning_rate": 3.452680174403694e-05, + "loss": 0.4531, + "step": 24182 + }, + { + "epoch": 0.93, + "learning_rate": 3.45261605539882e-05, + "loss": 0.4011, + "step": 24183 + }, + { + "epoch": 0.93, + "learning_rate": 3.4525519363939475e-05, + "loss": 0.4754, + "step": 24184 + }, + { + "epoch": 0.93, + "learning_rate": 3.4524878173890747e-05, + "loss": 0.5599, + "step": 24185 + }, + { + "epoch": 0.93, + "learning_rate": 3.452423698384201e-05, + "loss": 0.5425, + "step": 24186 + }, + { + "epoch": 0.93, + "learning_rate": 3.452359579379328e-05, + "loss": 0.3867, + "step": 24187 + }, + { + "epoch": 0.93, + "learning_rate": 3.452295460374455e-05, + "loss": 0.4711, + "step": 24188 + }, + { + "epoch": 0.93, + "learning_rate": 3.452231341369582e-05, + "loss": 0.6084, + "step": 24189 + }, + { + "epoch": 0.93, + "learning_rate": 3.452167222364709e-05, + "loss": 0.525, + "step": 24190 + }, + { + "epoch": 0.93, + "learning_rate": 3.452103103359836e-05, + "loss": 0.4992, + "step": 24191 + }, + { + "epoch": 0.93, + "learning_rate": 3.452038984354963e-05, + "loss": 0.4446, + "step": 24192 + }, + { + "epoch": 0.93, + "learning_rate": 3.45197486535009e-05, + "loss": 0.5205, + "step": 24193 + }, + { + "epoch": 0.93, + "learning_rate": 3.451910746345217e-05, + "loss": 0.5429, + "step": 24194 + }, + { + "epoch": 0.93, + "learning_rate": 3.4518466273403437e-05, + "loss": 0.5649, + "step": 24195 + }, + { + "epoch": 0.93, + "learning_rate": 3.451782508335471e-05, + "loss": 0.4681, + "step": 24196 + }, + { + "epoch": 0.93, + "learning_rate": 3.451718389330597e-05, + "loss": 0.5891, + "step": 24197 + }, + { + "epoch": 0.93, + "learning_rate": 3.451654270325725e-05, + "loss": 0.5621, + "step": 24198 + }, + { + "epoch": 0.93, + "learning_rate": 3.451590151320852e-05, + "loss": 0.4291, + "step": 24199 + }, + { + "epoch": 0.93, + "learning_rate": 3.451526032315979e-05, + "loss": 0.4444, + "step": 24200 + }, + { + "epoch": 0.93, + "learning_rate": 3.451461913311105e-05, + "loss": 0.5175, + "step": 24201 + }, + { + "epoch": 0.93, + "learning_rate": 3.4513977943062325e-05, + "loss": 0.4619, + "step": 24202 + }, + { + "epoch": 0.93, + "learning_rate": 3.45133367530136e-05, + "loss": 0.5167, + "step": 24203 + }, + { + "epoch": 0.93, + "learning_rate": 3.451269556296486e-05, + "loss": 0.5825, + "step": 24204 + }, + { + "epoch": 0.93, + "learning_rate": 3.4512054372916133e-05, + "loss": 0.4656, + "step": 24205 + }, + { + "epoch": 0.93, + "learning_rate": 3.4511413182867405e-05, + "loss": 0.4342, + "step": 24206 + }, + { + "epoch": 0.93, + "learning_rate": 3.451077199281868e-05, + "loss": 0.413, + "step": 24207 + }, + { + "epoch": 0.93, + "learning_rate": 3.451013080276994e-05, + "loss": 0.5395, + "step": 24208 + }, + { + "epoch": 0.93, + "learning_rate": 3.4509489612721214e-05, + "loss": 0.4356, + "step": 24209 + }, + { + "epoch": 0.93, + "learning_rate": 3.450884842267248e-05, + "loss": 0.3698, + "step": 24210 + }, + { + "epoch": 0.93, + "learning_rate": 3.450820723262375e-05, + "loss": 0.5114, + "step": 24211 + }, + { + "epoch": 0.93, + "learning_rate": 3.450756604257502e-05, + "loss": 0.5249, + "step": 24212 + }, + { + "epoch": 0.93, + "learning_rate": 3.450692485252629e-05, + "loss": 0.456, + "step": 24213 + }, + { + "epoch": 0.93, + "learning_rate": 3.4506283662477565e-05, + "loss": 0.5863, + "step": 24214 + }, + { + "epoch": 0.93, + "learning_rate": 3.450564247242883e-05, + "loss": 0.4951, + "step": 24215 + }, + { + "epoch": 0.93, + "learning_rate": 3.45050012823801e-05, + "loss": 0.4346, + "step": 24216 + }, + { + "epoch": 0.93, + "learning_rate": 3.450436009233137e-05, + "loss": 0.3242, + "step": 24217 + }, + { + "epoch": 0.93, + "learning_rate": 3.450371890228264e-05, + "loss": 0.4507, + "step": 24218 + }, + { + "epoch": 0.93, + "learning_rate": 3.4503077712233904e-05, + "loss": 0.5257, + "step": 24219 + }, + { + "epoch": 0.93, + "learning_rate": 3.4502436522185175e-05, + "loss": 0.5134, + "step": 24220 + }, + { + "epoch": 0.93, + "learning_rate": 3.450179533213645e-05, + "loss": 0.4351, + "step": 24221 + }, + { + "epoch": 0.93, + "learning_rate": 3.450115414208772e-05, + "loss": 0.4102, + "step": 24222 + }, + { + "epoch": 0.93, + "learning_rate": 3.450051295203899e-05, + "loss": 0.5202, + "step": 24223 + }, + { + "epoch": 0.93, + "learning_rate": 3.4499871761990255e-05, + "loss": 0.4747, + "step": 24224 + }, + { + "epoch": 0.93, + "learning_rate": 3.449923057194153e-05, + "loss": 0.3819, + "step": 24225 + }, + { + "epoch": 0.93, + "learning_rate": 3.449858938189279e-05, + "loss": 0.5514, + "step": 24226 + }, + { + "epoch": 0.93, + "learning_rate": 3.4497948191844064e-05, + "loss": 0.4738, + "step": 24227 + }, + { + "epoch": 0.93, + "learning_rate": 3.449730700179533e-05, + "loss": 0.4153, + "step": 24228 + }, + { + "epoch": 0.93, + "learning_rate": 3.44966658117466e-05, + "loss": 0.615, + "step": 24229 + }, + { + "epoch": 0.93, + "learning_rate": 3.449602462169787e-05, + "loss": 0.4059, + "step": 24230 + }, + { + "epoch": 0.93, + "learning_rate": 3.4495383431649144e-05, + "loss": 0.5058, + "step": 24231 + }, + { + "epoch": 0.93, + "learning_rate": 3.4494742241600416e-05, + "loss": 0.452, + "step": 24232 + }, + { + "epoch": 0.93, + "learning_rate": 3.449410105155168e-05, + "loss": 0.428, + "step": 24233 + }, + { + "epoch": 0.93, + "learning_rate": 3.449345986150295e-05, + "loss": 0.5157, + "step": 24234 + }, + { + "epoch": 0.93, + "learning_rate": 3.449281867145422e-05, + "loss": 0.4031, + "step": 24235 + }, + { + "epoch": 0.93, + "learning_rate": 3.449217748140549e-05, + "loss": 0.3954, + "step": 24236 + }, + { + "epoch": 0.93, + "learning_rate": 3.449153629135676e-05, + "loss": 0.4551, + "step": 24237 + }, + { + "epoch": 0.93, + "learning_rate": 3.449089510130803e-05, + "loss": 0.4215, + "step": 24238 + }, + { + "epoch": 0.93, + "learning_rate": 3.4490253911259304e-05, + "loss": 0.4716, + "step": 24239 + }, + { + "epoch": 0.93, + "learning_rate": 3.448961272121057e-05, + "loss": 0.355, + "step": 24240 + }, + { + "epoch": 0.93, + "learning_rate": 3.448897153116184e-05, + "loss": 0.4243, + "step": 24241 + }, + { + "epoch": 0.93, + "learning_rate": 3.4488330341113106e-05, + "loss": 0.4992, + "step": 24242 + }, + { + "epoch": 0.93, + "learning_rate": 3.448768915106438e-05, + "loss": 0.3609, + "step": 24243 + }, + { + "epoch": 0.93, + "learning_rate": 3.448704796101564e-05, + "loss": 0.4508, + "step": 24244 + }, + { + "epoch": 0.93, + "learning_rate": 3.4486406770966914e-05, + "loss": 0.5274, + "step": 24245 + }, + { + "epoch": 0.93, + "learning_rate": 3.4485765580918186e-05, + "loss": 0.6148, + "step": 24246 + }, + { + "epoch": 0.93, + "learning_rate": 3.448512439086946e-05, + "loss": 0.3755, + "step": 24247 + }, + { + "epoch": 0.93, + "learning_rate": 3.448448320082073e-05, + "loss": 0.4449, + "step": 24248 + }, + { + "epoch": 0.93, + "learning_rate": 3.4483842010771994e-05, + "loss": 0.5478, + "step": 24249 + }, + { + "epoch": 0.93, + "learning_rate": 3.4483200820723266e-05, + "loss": 0.5618, + "step": 24250 + }, + { + "epoch": 0.93, + "learning_rate": 3.448255963067453e-05, + "loss": 0.4717, + "step": 24251 + }, + { + "epoch": 0.93, + "learning_rate": 3.44819184406258e-05, + "loss": 0.4922, + "step": 24252 + }, + { + "epoch": 0.93, + "learning_rate": 3.4481277250577074e-05, + "loss": 0.5427, + "step": 24253 + }, + { + "epoch": 0.93, + "learning_rate": 3.4480636060528346e-05, + "loss": 0.601, + "step": 24254 + }, + { + "epoch": 0.93, + "learning_rate": 3.447999487047961e-05, + "loss": 0.5504, + "step": 24255 + }, + { + "epoch": 0.93, + "learning_rate": 3.447935368043088e-05, + "loss": 0.5322, + "step": 24256 + }, + { + "epoch": 0.93, + "learning_rate": 3.4478712490382154e-05, + "loss": 0.5618, + "step": 24257 + }, + { + "epoch": 0.93, + "learning_rate": 3.447807130033342e-05, + "loss": 0.6077, + "step": 24258 + }, + { + "epoch": 0.93, + "learning_rate": 3.447743011028469e-05, + "loss": 0.4589, + "step": 24259 + }, + { + "epoch": 0.93, + "learning_rate": 3.4476788920235956e-05, + "loss": 0.4674, + "step": 24260 + }, + { + "epoch": 0.93, + "learning_rate": 3.447614773018723e-05, + "loss": 0.4696, + "step": 24261 + }, + { + "epoch": 0.93, + "learning_rate": 3.44755065401385e-05, + "loss": 0.5377, + "step": 24262 + }, + { + "epoch": 0.93, + "learning_rate": 3.447486535008977e-05, + "loss": 0.4471, + "step": 24263 + }, + { + "epoch": 0.93, + "learning_rate": 3.4474224160041036e-05, + "loss": 0.4882, + "step": 24264 + }, + { + "epoch": 0.93, + "learning_rate": 3.447358296999231e-05, + "loss": 0.4399, + "step": 24265 + }, + { + "epoch": 0.93, + "learning_rate": 3.447294177994358e-05, + "loss": 0.5039, + "step": 24266 + }, + { + "epoch": 0.93, + "learning_rate": 3.4472300589894844e-05, + "loss": 0.4097, + "step": 24267 + }, + { + "epoch": 0.93, + "learning_rate": 3.4471659399846116e-05, + "loss": 0.4853, + "step": 24268 + }, + { + "epoch": 0.93, + "learning_rate": 3.447101820979739e-05, + "loss": 0.4283, + "step": 24269 + }, + { + "epoch": 0.93, + "learning_rate": 3.447037701974866e-05, + "loss": 0.5411, + "step": 24270 + }, + { + "epoch": 0.93, + "learning_rate": 3.4469735829699924e-05, + "loss": 0.3788, + "step": 24271 + }, + { + "epoch": 0.93, + "learning_rate": 3.4469094639651196e-05, + "loss": 0.5625, + "step": 24272 + }, + { + "epoch": 0.93, + "learning_rate": 3.446845344960246e-05, + "loss": 0.4227, + "step": 24273 + }, + { + "epoch": 0.93, + "learning_rate": 3.446781225955373e-05, + "loss": 0.5459, + "step": 24274 + }, + { + "epoch": 0.93, + "learning_rate": 3.4467171069505005e-05, + "loss": 0.4586, + "step": 24275 + }, + { + "epoch": 0.93, + "learning_rate": 3.446652987945627e-05, + "loss": 0.3891, + "step": 24276 + }, + { + "epoch": 0.93, + "learning_rate": 3.446588868940754e-05, + "loss": 0.3844, + "step": 24277 + }, + { + "epoch": 0.93, + "learning_rate": 3.446524749935881e-05, + "loss": 0.4405, + "step": 24278 + }, + { + "epoch": 0.93, + "learning_rate": 3.4464606309310085e-05, + "loss": 0.4295, + "step": 24279 + }, + { + "epoch": 0.93, + "learning_rate": 3.446396511926135e-05, + "loss": 0.4336, + "step": 24280 + }, + { + "epoch": 0.93, + "learning_rate": 3.446332392921262e-05, + "loss": 0.3923, + "step": 24281 + }, + { + "epoch": 0.93, + "learning_rate": 3.4462682739163886e-05, + "loss": 0.5133, + "step": 24282 + }, + { + "epoch": 0.93, + "learning_rate": 3.446204154911516e-05, + "loss": 0.5556, + "step": 24283 + }, + { + "epoch": 0.93, + "learning_rate": 3.446140035906643e-05, + "loss": 0.4971, + "step": 24284 + }, + { + "epoch": 0.93, + "learning_rate": 3.4460759169017695e-05, + "loss": 0.3648, + "step": 24285 + }, + { + "epoch": 0.93, + "learning_rate": 3.446011797896897e-05, + "loss": 0.5542, + "step": 24286 + }, + { + "epoch": 0.93, + "learning_rate": 3.445947678892024e-05, + "loss": 0.4446, + "step": 24287 + }, + { + "epoch": 0.93, + "learning_rate": 3.445883559887151e-05, + "loss": 0.475, + "step": 24288 + }, + { + "epoch": 0.93, + "learning_rate": 3.4458194408822775e-05, + "loss": 0.4021, + "step": 24289 + }, + { + "epoch": 0.93, + "learning_rate": 3.4457553218774046e-05, + "loss": 0.4679, + "step": 24290 + }, + { + "epoch": 0.93, + "learning_rate": 3.445691202872531e-05, + "loss": 0.3844, + "step": 24291 + }, + { + "epoch": 0.93, + "learning_rate": 3.445627083867658e-05, + "loss": 0.4873, + "step": 24292 + }, + { + "epoch": 0.93, + "learning_rate": 3.4455629648627855e-05, + "loss": 0.5075, + "step": 24293 + }, + { + "epoch": 0.93, + "learning_rate": 3.4454988458579126e-05, + "loss": 0.523, + "step": 24294 + }, + { + "epoch": 0.93, + "learning_rate": 3.44543472685304e-05, + "loss": 0.4817, + "step": 24295 + }, + { + "epoch": 0.93, + "learning_rate": 3.445370607848166e-05, + "loss": 0.4472, + "step": 24296 + }, + { + "epoch": 0.93, + "learning_rate": 3.4453064888432935e-05, + "loss": 0.4738, + "step": 24297 + }, + { + "epoch": 0.93, + "learning_rate": 3.44524236983842e-05, + "loss": 0.5519, + "step": 24298 + }, + { + "epoch": 0.93, + "learning_rate": 3.445178250833547e-05, + "loss": 0.534, + "step": 24299 + }, + { + "epoch": 0.93, + "learning_rate": 3.4451141318286736e-05, + "loss": 0.4881, + "step": 24300 + }, + { + "epoch": 0.93, + "learning_rate": 3.445050012823801e-05, + "loss": 0.2966, + "step": 24301 + }, + { + "epoch": 0.93, + "learning_rate": 3.444985893818929e-05, + "loss": 0.5174, + "step": 24302 + }, + { + "epoch": 0.93, + "learning_rate": 3.444921774814055e-05, + "loss": 0.5151, + "step": 24303 + }, + { + "epoch": 0.93, + "learning_rate": 3.444857655809182e-05, + "loss": 0.5041, + "step": 24304 + }, + { + "epoch": 0.93, + "learning_rate": 3.444793536804309e-05, + "loss": 0.4855, + "step": 24305 + }, + { + "epoch": 0.93, + "learning_rate": 3.444729417799436e-05, + "loss": 0.4081, + "step": 24306 + }, + { + "epoch": 0.93, + "learning_rate": 3.4446652987945625e-05, + "loss": 0.5671, + "step": 24307 + }, + { + "epoch": 0.93, + "learning_rate": 3.44460117978969e-05, + "loss": 0.3884, + "step": 24308 + }, + { + "epoch": 0.93, + "learning_rate": 3.444537060784817e-05, + "loss": 0.363, + "step": 24309 + }, + { + "epoch": 0.93, + "learning_rate": 3.444472941779944e-05, + "loss": 0.5739, + "step": 24310 + }, + { + "epoch": 0.93, + "learning_rate": 3.444408822775071e-05, + "loss": 0.5054, + "step": 24311 + }, + { + "epoch": 0.93, + "learning_rate": 3.444344703770198e-05, + "loss": 0.5038, + "step": 24312 + }, + { + "epoch": 0.93, + "learning_rate": 3.444280584765325e-05, + "loss": 0.391, + "step": 24313 + }, + { + "epoch": 0.93, + "learning_rate": 3.444216465760451e-05, + "loss": 0.4924, + "step": 24314 + }, + { + "epoch": 0.93, + "learning_rate": 3.4441523467555785e-05, + "loss": 0.4158, + "step": 24315 + }, + { + "epoch": 0.93, + "learning_rate": 3.444088227750705e-05, + "loss": 0.5759, + "step": 24316 + }, + { + "epoch": 0.93, + "learning_rate": 3.444024108745832e-05, + "loss": 0.5032, + "step": 24317 + }, + { + "epoch": 0.93, + "learning_rate": 3.4439599897409593e-05, + "loss": 0.4754, + "step": 24318 + }, + { + "epoch": 0.93, + "learning_rate": 3.4438958707360865e-05, + "loss": 0.3681, + "step": 24319 + }, + { + "epoch": 0.94, + "learning_rate": 3.443831751731214e-05, + "loss": 0.5007, + "step": 24320 + }, + { + "epoch": 0.94, + "learning_rate": 3.44376763272634e-05, + "loss": 0.4971, + "step": 24321 + }, + { + "epoch": 0.94, + "learning_rate": 3.4437035137214674e-05, + "loss": 0.5395, + "step": 24322 + }, + { + "epoch": 0.94, + "learning_rate": 3.443639394716594e-05, + "loss": 0.4211, + "step": 24323 + }, + { + "epoch": 0.94, + "learning_rate": 3.443575275711721e-05, + "loss": 0.5376, + "step": 24324 + }, + { + "epoch": 0.94, + "learning_rate": 3.443511156706848e-05, + "loss": 0.4982, + "step": 24325 + }, + { + "epoch": 0.94, + "learning_rate": 3.4434470377019754e-05, + "loss": 0.5468, + "step": 24326 + }, + { + "epoch": 0.94, + "learning_rate": 3.443382918697102e-05, + "loss": 0.3912, + "step": 24327 + }, + { + "epoch": 0.94, + "learning_rate": 3.443318799692229e-05, + "loss": 0.5094, + "step": 24328 + }, + { + "epoch": 0.94, + "learning_rate": 3.443254680687356e-05, + "loss": 0.4012, + "step": 24329 + }, + { + "epoch": 0.94, + "learning_rate": 3.443190561682483e-05, + "loss": 0.4473, + "step": 24330 + }, + { + "epoch": 0.94, + "learning_rate": 3.44312644267761e-05, + "loss": 0.5017, + "step": 24331 + }, + { + "epoch": 0.94, + "learning_rate": 3.4430623236727364e-05, + "loss": 0.5148, + "step": 24332 + }, + { + "epoch": 0.94, + "learning_rate": 3.4429982046678635e-05, + "loss": 0.5105, + "step": 24333 + }, + { + "epoch": 0.94, + "learning_rate": 3.442934085662991e-05, + "loss": 0.5294, + "step": 24334 + }, + { + "epoch": 0.94, + "learning_rate": 3.442869966658118e-05, + "loss": 0.43, + "step": 24335 + }, + { + "epoch": 0.94, + "learning_rate": 3.4428058476532444e-05, + "loss": 0.5302, + "step": 24336 + }, + { + "epoch": 0.94, + "learning_rate": 3.4427417286483715e-05, + "loss": 0.5161, + "step": 24337 + }, + { + "epoch": 0.94, + "learning_rate": 3.442677609643499e-05, + "loss": 0.4694, + "step": 24338 + }, + { + "epoch": 0.94, + "learning_rate": 3.442613490638625e-05, + "loss": 0.4508, + "step": 24339 + }, + { + "epoch": 0.94, + "learning_rate": 3.4425493716337524e-05, + "loss": 0.5667, + "step": 24340 + }, + { + "epoch": 0.94, + "learning_rate": 3.4424852526288796e-05, + "loss": 0.4597, + "step": 24341 + }, + { + "epoch": 0.94, + "learning_rate": 3.442421133624007e-05, + "loss": 0.5003, + "step": 24342 + }, + { + "epoch": 0.94, + "learning_rate": 3.442357014619133e-05, + "loss": 0.4356, + "step": 24343 + }, + { + "epoch": 0.94, + "learning_rate": 3.4422928956142604e-05, + "loss": 0.4649, + "step": 24344 + }, + { + "epoch": 0.94, + "learning_rate": 3.442228776609387e-05, + "loss": 0.4219, + "step": 24345 + }, + { + "epoch": 0.94, + "learning_rate": 3.442164657604514e-05, + "loss": 0.4987, + "step": 24346 + }, + { + "epoch": 0.94, + "learning_rate": 3.442100538599641e-05, + "loss": 0.6608, + "step": 24347 + }, + { + "epoch": 0.94, + "learning_rate": 3.442036419594768e-05, + "loss": 0.4901, + "step": 24348 + }, + { + "epoch": 0.94, + "learning_rate": 3.441972300589895e-05, + "loss": 0.6103, + "step": 24349 + }, + { + "epoch": 0.94, + "learning_rate": 3.441908181585022e-05, + "loss": 0.4341, + "step": 24350 + }, + { + "epoch": 0.94, + "learning_rate": 3.441844062580149e-05, + "loss": 0.6226, + "step": 24351 + }, + { + "epoch": 0.94, + "learning_rate": 3.441779943575276e-05, + "loss": 0.4055, + "step": 24352 + }, + { + "epoch": 0.94, + "learning_rate": 3.441715824570403e-05, + "loss": 0.5485, + "step": 24353 + }, + { + "epoch": 0.94, + "learning_rate": 3.4416517055655294e-05, + "loss": 0.4444, + "step": 24354 + }, + { + "epoch": 0.94, + "learning_rate": 3.4415875865606566e-05, + "loss": 0.4858, + "step": 24355 + }, + { + "epoch": 0.94, + "learning_rate": 3.441523467555784e-05, + "loss": 0.4804, + "step": 24356 + }, + { + "epoch": 0.94, + "learning_rate": 3.44145934855091e-05, + "loss": 0.4452, + "step": 24357 + }, + { + "epoch": 0.94, + "learning_rate": 3.441395229546038e-05, + "loss": 0.4069, + "step": 24358 + }, + { + "epoch": 0.94, + "learning_rate": 3.4413311105411646e-05, + "loss": 0.5386, + "step": 24359 + }, + { + "epoch": 0.94, + "learning_rate": 3.441266991536292e-05, + "loss": 0.5388, + "step": 24360 + }, + { + "epoch": 0.94, + "learning_rate": 3.441202872531418e-05, + "loss": 0.4009, + "step": 24361 + }, + { + "epoch": 0.94, + "learning_rate": 3.4411387535265454e-05, + "loss": 0.4509, + "step": 24362 + }, + { + "epoch": 0.94, + "learning_rate": 3.441074634521672e-05, + "loss": 0.4547, + "step": 24363 + }, + { + "epoch": 0.94, + "learning_rate": 3.441010515516799e-05, + "loss": 0.558, + "step": 24364 + }, + { + "epoch": 0.94, + "learning_rate": 3.440946396511926e-05, + "loss": 0.5241, + "step": 24365 + }, + { + "epoch": 0.94, + "learning_rate": 3.4408822775070534e-05, + "loss": 0.4129, + "step": 24366 + }, + { + "epoch": 0.94, + "learning_rate": 3.4408181585021806e-05, + "loss": 0.5351, + "step": 24367 + }, + { + "epoch": 0.94, + "learning_rate": 3.440754039497307e-05, + "loss": 0.4258, + "step": 24368 + }, + { + "epoch": 0.94, + "learning_rate": 3.440689920492434e-05, + "loss": 0.4347, + "step": 24369 + }, + { + "epoch": 0.94, + "learning_rate": 3.440625801487561e-05, + "loss": 0.5238, + "step": 24370 + }, + { + "epoch": 0.94, + "learning_rate": 3.440561682482688e-05, + "loss": 0.5234, + "step": 24371 + }, + { + "epoch": 0.94, + "learning_rate": 3.440497563477815e-05, + "loss": 0.5275, + "step": 24372 + }, + { + "epoch": 0.94, + "learning_rate": 3.4404334444729416e-05, + "loss": 0.5351, + "step": 24373 + }, + { + "epoch": 0.94, + "learning_rate": 3.4403693254680694e-05, + "loss": 0.5333, + "step": 24374 + }, + { + "epoch": 0.94, + "learning_rate": 3.440305206463196e-05, + "loss": 0.518, + "step": 24375 + }, + { + "epoch": 0.94, + "learning_rate": 3.440241087458323e-05, + "loss": 0.5659, + "step": 24376 + }, + { + "epoch": 0.94, + "learning_rate": 3.4401769684534496e-05, + "loss": 0.4401, + "step": 24377 + }, + { + "epoch": 0.94, + "learning_rate": 3.440112849448577e-05, + "loss": 0.493, + "step": 24378 + }, + { + "epoch": 0.94, + "learning_rate": 3.440048730443703e-05, + "loss": 0.4549, + "step": 24379 + }, + { + "epoch": 0.94, + "learning_rate": 3.4399846114388304e-05, + "loss": 0.5148, + "step": 24380 + }, + { + "epoch": 0.94, + "learning_rate": 3.4399204924339576e-05, + "loss": 0.4566, + "step": 24381 + }, + { + "epoch": 0.94, + "learning_rate": 3.439856373429085e-05, + "loss": 0.4426, + "step": 24382 + }, + { + "epoch": 0.94, + "learning_rate": 3.439792254424212e-05, + "loss": 0.5124, + "step": 24383 + }, + { + "epoch": 0.94, + "learning_rate": 3.4397281354193384e-05, + "loss": 0.5901, + "step": 24384 + }, + { + "epoch": 0.94, + "learning_rate": 3.4396640164144656e-05, + "loss": 0.4182, + "step": 24385 + }, + { + "epoch": 0.94, + "learning_rate": 3.439599897409592e-05, + "loss": 0.5629, + "step": 24386 + }, + { + "epoch": 0.94, + "learning_rate": 3.439535778404719e-05, + "loss": 0.3867, + "step": 24387 + }, + { + "epoch": 0.94, + "learning_rate": 3.439471659399846e-05, + "loss": 0.4448, + "step": 24388 + }, + { + "epoch": 0.94, + "learning_rate": 3.439407540394973e-05, + "loss": 0.529, + "step": 24389 + }, + { + "epoch": 0.94, + "learning_rate": 3.439343421390101e-05, + "loss": 0.5251, + "step": 24390 + }, + { + "epoch": 0.94, + "learning_rate": 3.439279302385227e-05, + "loss": 0.3995, + "step": 24391 + }, + { + "epoch": 0.94, + "learning_rate": 3.4392151833803545e-05, + "loss": 0.5438, + "step": 24392 + }, + { + "epoch": 0.94, + "learning_rate": 3.439151064375481e-05, + "loss": 0.4678, + "step": 24393 + }, + { + "epoch": 0.94, + "learning_rate": 3.439086945370608e-05, + "loss": 0.4697, + "step": 24394 + }, + { + "epoch": 0.94, + "learning_rate": 3.4390228263657346e-05, + "loss": 0.4284, + "step": 24395 + }, + { + "epoch": 0.94, + "learning_rate": 3.438958707360862e-05, + "loss": 0.3841, + "step": 24396 + }, + { + "epoch": 0.94, + "learning_rate": 3.438894588355989e-05, + "loss": 0.6362, + "step": 24397 + }, + { + "epoch": 0.94, + "learning_rate": 3.438830469351116e-05, + "loss": 0.4807, + "step": 24398 + }, + { + "epoch": 0.94, + "learning_rate": 3.438766350346243e-05, + "loss": 0.6555, + "step": 24399 + }, + { + "epoch": 0.94, + "learning_rate": 3.43870223134137e-05, + "loss": 0.4414, + "step": 24400 + }, + { + "epoch": 0.94, + "learning_rate": 3.438638112336497e-05, + "loss": 0.566, + "step": 24401 + }, + { + "epoch": 0.94, + "learning_rate": 3.4385739933316235e-05, + "loss": 0.5167, + "step": 24402 + }, + { + "epoch": 0.94, + "learning_rate": 3.4385098743267506e-05, + "loss": 0.5942, + "step": 24403 + }, + { + "epoch": 0.94, + "learning_rate": 3.438445755321877e-05, + "loss": 0.4711, + "step": 24404 + }, + { + "epoch": 0.94, + "learning_rate": 3.438381636317004e-05, + "loss": 0.5001, + "step": 24405 + }, + { + "epoch": 0.94, + "learning_rate": 3.4383175173121315e-05, + "loss": 0.4232, + "step": 24406 + }, + { + "epoch": 0.94, + "learning_rate": 3.4382533983072587e-05, + "loss": 0.6175, + "step": 24407 + }, + { + "epoch": 0.94, + "learning_rate": 3.438189279302386e-05, + "loss": 0.4493, + "step": 24408 + }, + { + "epoch": 0.94, + "learning_rate": 3.438125160297512e-05, + "loss": 0.5522, + "step": 24409 + }, + { + "epoch": 0.94, + "learning_rate": 3.4380610412926395e-05, + "loss": 0.4976, + "step": 24410 + }, + { + "epoch": 0.94, + "learning_rate": 3.437996922287766e-05, + "loss": 0.4436, + "step": 24411 + }, + { + "epoch": 0.94, + "learning_rate": 3.437932803282893e-05, + "loss": 0.5464, + "step": 24412 + }, + { + "epoch": 0.94, + "learning_rate": 3.43786868427802e-05, + "loss": 0.4929, + "step": 24413 + }, + { + "epoch": 0.94, + "learning_rate": 3.4378045652731475e-05, + "loss": 0.5176, + "step": 24414 + }, + { + "epoch": 0.94, + "learning_rate": 3.437740446268274e-05, + "loss": 0.5302, + "step": 24415 + }, + { + "epoch": 0.94, + "learning_rate": 3.437676327263401e-05, + "loss": 0.4594, + "step": 24416 + }, + { + "epoch": 0.94, + "learning_rate": 3.437612208258528e-05, + "loss": 0.5347, + "step": 24417 + }, + { + "epoch": 0.94, + "learning_rate": 3.437548089253655e-05, + "loss": 0.4439, + "step": 24418 + }, + { + "epoch": 0.94, + "learning_rate": 3.437483970248782e-05, + "loss": 0.5052, + "step": 24419 + }, + { + "epoch": 0.94, + "learning_rate": 3.4374198512439085e-05, + "loss": 0.4251, + "step": 24420 + }, + { + "epoch": 0.94, + "learning_rate": 3.437355732239036e-05, + "loss": 0.4905, + "step": 24421 + }, + { + "epoch": 0.94, + "learning_rate": 3.437291613234163e-05, + "loss": 0.5133, + "step": 24422 + }, + { + "epoch": 0.94, + "learning_rate": 3.43722749422929e-05, + "loss": 0.3589, + "step": 24423 + }, + { + "epoch": 0.94, + "learning_rate": 3.4371633752244165e-05, + "loss": 0.5353, + "step": 24424 + }, + { + "epoch": 0.94, + "learning_rate": 3.437099256219544e-05, + "loss": 0.4603, + "step": 24425 + }, + { + "epoch": 0.94, + "learning_rate": 3.437035137214671e-05, + "loss": 0.5635, + "step": 24426 + }, + { + "epoch": 0.94, + "learning_rate": 3.4369710182097973e-05, + "loss": 0.4129, + "step": 24427 + }, + { + "epoch": 0.94, + "learning_rate": 3.4369068992049245e-05, + "loss": 0.5184, + "step": 24428 + }, + { + "epoch": 0.94, + "learning_rate": 3.436842780200052e-05, + "loss": 0.4491, + "step": 24429 + }, + { + "epoch": 0.94, + "learning_rate": 3.436778661195179e-05, + "loss": 0.3643, + "step": 24430 + }, + { + "epoch": 0.94, + "learning_rate": 3.4367145421903053e-05, + "loss": 0.5556, + "step": 24431 + }, + { + "epoch": 0.94, + "learning_rate": 3.4366504231854325e-05, + "loss": 0.4425, + "step": 24432 + }, + { + "epoch": 0.94, + "learning_rate": 3.436586304180559e-05, + "loss": 0.5845, + "step": 24433 + }, + { + "epoch": 0.94, + "learning_rate": 3.436522185175686e-05, + "loss": 0.5408, + "step": 24434 + }, + { + "epoch": 0.94, + "learning_rate": 3.4364580661708134e-05, + "loss": 0.3838, + "step": 24435 + }, + { + "epoch": 0.94, + "learning_rate": 3.43639394716594e-05, + "loss": 0.4937, + "step": 24436 + }, + { + "epoch": 0.94, + "learning_rate": 3.436329828161067e-05, + "loss": 0.447, + "step": 24437 + }, + { + "epoch": 0.94, + "learning_rate": 3.436265709156194e-05, + "loss": 0.6174, + "step": 24438 + }, + { + "epoch": 0.94, + "learning_rate": 3.4362015901513214e-05, + "loss": 0.5896, + "step": 24439 + }, + { + "epoch": 0.94, + "learning_rate": 3.436137471146448e-05, + "loss": 0.417, + "step": 24440 + }, + { + "epoch": 0.94, + "learning_rate": 3.436073352141575e-05, + "loss": 0.5038, + "step": 24441 + }, + { + "epoch": 0.94, + "learning_rate": 3.4360092331367015e-05, + "loss": 0.5408, + "step": 24442 + }, + { + "epoch": 0.94, + "learning_rate": 3.435945114131829e-05, + "loss": 0.4442, + "step": 24443 + }, + { + "epoch": 0.94, + "learning_rate": 3.435880995126956e-05, + "loss": 0.4717, + "step": 24444 + }, + { + "epoch": 0.94, + "learning_rate": 3.4358168761220824e-05, + "loss": 0.4197, + "step": 24445 + }, + { + "epoch": 0.94, + "learning_rate": 3.43575275711721e-05, + "loss": 0.4616, + "step": 24446 + }, + { + "epoch": 0.94, + "learning_rate": 3.435688638112337e-05, + "loss": 0.2978, + "step": 24447 + }, + { + "epoch": 0.94, + "learning_rate": 3.435624519107464e-05, + "loss": 0.6055, + "step": 24448 + }, + { + "epoch": 0.94, + "learning_rate": 3.4355604001025904e-05, + "loss": 0.4379, + "step": 24449 + }, + { + "epoch": 0.94, + "learning_rate": 3.4354962810977175e-05, + "loss": 0.438, + "step": 24450 + }, + { + "epoch": 0.94, + "learning_rate": 3.435432162092844e-05, + "loss": 0.5742, + "step": 24451 + }, + { + "epoch": 0.94, + "learning_rate": 3.435368043087971e-05, + "loss": 0.4547, + "step": 24452 + }, + { + "epoch": 0.94, + "learning_rate": 3.4353039240830984e-05, + "loss": 0.3947, + "step": 24453 + }, + { + "epoch": 0.94, + "learning_rate": 3.4352398050782256e-05, + "loss": 0.5393, + "step": 24454 + }, + { + "epoch": 0.94, + "learning_rate": 3.435175686073353e-05, + "loss": 0.5121, + "step": 24455 + }, + { + "epoch": 0.94, + "learning_rate": 3.435111567068479e-05, + "loss": 0.4744, + "step": 24456 + }, + { + "epoch": 0.94, + "learning_rate": 3.4350474480636064e-05, + "loss": 0.6152, + "step": 24457 + }, + { + "epoch": 0.94, + "learning_rate": 3.434983329058733e-05, + "loss": 0.6814, + "step": 24458 + }, + { + "epoch": 0.94, + "learning_rate": 3.43491921005386e-05, + "loss": 0.4988, + "step": 24459 + }, + { + "epoch": 0.94, + "learning_rate": 3.4348550910489866e-05, + "loss": 0.3639, + "step": 24460 + }, + { + "epoch": 0.94, + "learning_rate": 3.434790972044114e-05, + "loss": 0.5041, + "step": 24461 + }, + { + "epoch": 0.94, + "learning_rate": 3.4347268530392416e-05, + "loss": 0.451, + "step": 24462 + }, + { + "epoch": 0.94, + "learning_rate": 3.434662734034368e-05, + "loss": 0.419, + "step": 24463 + }, + { + "epoch": 0.94, + "learning_rate": 3.434598615029495e-05, + "loss": 0.5324, + "step": 24464 + }, + { + "epoch": 0.94, + "learning_rate": 3.434534496024622e-05, + "loss": 0.4512, + "step": 24465 + }, + { + "epoch": 0.94, + "learning_rate": 3.434470377019749e-05, + "loss": 0.3754, + "step": 24466 + }, + { + "epoch": 0.94, + "learning_rate": 3.4344062580148754e-05, + "loss": 0.4942, + "step": 24467 + }, + { + "epoch": 0.94, + "learning_rate": 3.4343421390100026e-05, + "loss": 0.4976, + "step": 24468 + }, + { + "epoch": 0.94, + "learning_rate": 3.43427802000513e-05, + "loss": 0.4325, + "step": 24469 + }, + { + "epoch": 0.94, + "learning_rate": 3.434213901000257e-05, + "loss": 0.578, + "step": 24470 + }, + { + "epoch": 0.94, + "learning_rate": 3.434149781995384e-05, + "loss": 0.4159, + "step": 24471 + }, + { + "epoch": 0.94, + "learning_rate": 3.4340856629905106e-05, + "loss": 0.4459, + "step": 24472 + }, + { + "epoch": 0.94, + "learning_rate": 3.434021543985638e-05, + "loss": 0.4798, + "step": 24473 + }, + { + "epoch": 0.94, + "learning_rate": 3.433957424980764e-05, + "loss": 0.5264, + "step": 24474 + }, + { + "epoch": 0.94, + "learning_rate": 3.4338933059758914e-05, + "loss": 0.4508, + "step": 24475 + }, + { + "epoch": 0.94, + "learning_rate": 3.433829186971018e-05, + "loss": 0.5528, + "step": 24476 + }, + { + "epoch": 0.94, + "learning_rate": 3.433765067966145e-05, + "loss": 0.668, + "step": 24477 + }, + { + "epoch": 0.94, + "learning_rate": 3.433700948961272e-05, + "loss": 0.5433, + "step": 24478 + }, + { + "epoch": 0.94, + "learning_rate": 3.4336368299563994e-05, + "loss": 0.5067, + "step": 24479 + }, + { + "epoch": 0.94, + "learning_rate": 3.4335727109515266e-05, + "loss": 0.4979, + "step": 24480 + }, + { + "epoch": 0.94, + "learning_rate": 3.433508591946653e-05, + "loss": 0.4891, + "step": 24481 + }, + { + "epoch": 0.94, + "learning_rate": 3.43344447294178e-05, + "loss": 0.4585, + "step": 24482 + }, + { + "epoch": 0.94, + "learning_rate": 3.433380353936907e-05, + "loss": 0.4349, + "step": 24483 + }, + { + "epoch": 0.94, + "learning_rate": 3.433316234932034e-05, + "loss": 0.375, + "step": 24484 + }, + { + "epoch": 0.94, + "learning_rate": 3.433252115927161e-05, + "loss": 0.4977, + "step": 24485 + }, + { + "epoch": 0.94, + "learning_rate": 3.433187996922288e-05, + "loss": 0.5186, + "step": 24486 + }, + { + "epoch": 0.94, + "learning_rate": 3.433123877917415e-05, + "loss": 0.4424, + "step": 24487 + }, + { + "epoch": 0.94, + "learning_rate": 3.433059758912542e-05, + "loss": 0.461, + "step": 24488 + }, + { + "epoch": 0.94, + "learning_rate": 3.432995639907669e-05, + "loss": 0.4324, + "step": 24489 + }, + { + "epoch": 0.94, + "learning_rate": 3.4329315209027956e-05, + "loss": 0.4605, + "step": 24490 + }, + { + "epoch": 0.94, + "learning_rate": 3.432867401897923e-05, + "loss": 0.4684, + "step": 24491 + }, + { + "epoch": 0.94, + "learning_rate": 3.432803282893049e-05, + "loss": 0.4289, + "step": 24492 + }, + { + "epoch": 0.94, + "learning_rate": 3.4327391638881764e-05, + "loss": 0.5636, + "step": 24493 + }, + { + "epoch": 0.94, + "learning_rate": 3.4326750448833036e-05, + "loss": 0.4331, + "step": 24494 + }, + { + "epoch": 0.94, + "learning_rate": 3.432610925878431e-05, + "loss": 0.4913, + "step": 24495 + }, + { + "epoch": 0.94, + "learning_rate": 3.432546806873557e-05, + "loss": 0.575, + "step": 24496 + }, + { + "epoch": 0.94, + "learning_rate": 3.4324826878686844e-05, + "loss": 0.5013, + "step": 24497 + }, + { + "epoch": 0.94, + "learning_rate": 3.4324185688638116e-05, + "loss": 0.4956, + "step": 24498 + }, + { + "epoch": 0.94, + "learning_rate": 3.432354449858938e-05, + "loss": 0.5713, + "step": 24499 + }, + { + "epoch": 0.94, + "learning_rate": 3.432290330854065e-05, + "loss": 0.4901, + "step": 24500 + }, + { + "epoch": 0.94, + "learning_rate": 3.4322262118491925e-05, + "loss": 0.4547, + "step": 24501 + }, + { + "epoch": 0.94, + "learning_rate": 3.4321620928443196e-05, + "loss": 0.4985, + "step": 24502 + }, + { + "epoch": 0.94, + "learning_rate": 3.432097973839446e-05, + "loss": 0.4491, + "step": 24503 + }, + { + "epoch": 0.94, + "learning_rate": 3.432033854834573e-05, + "loss": 0.5295, + "step": 24504 + }, + { + "epoch": 0.94, + "learning_rate": 3.4319697358297e-05, + "loss": 0.5141, + "step": 24505 + }, + { + "epoch": 0.94, + "learning_rate": 3.431905616824827e-05, + "loss": 0.489, + "step": 24506 + }, + { + "epoch": 0.94, + "learning_rate": 3.431841497819954e-05, + "loss": 0.5145, + "step": 24507 + }, + { + "epoch": 0.94, + "learning_rate": 3.4317773788150806e-05, + "loss": 0.5503, + "step": 24508 + }, + { + "epoch": 0.94, + "learning_rate": 3.431713259810208e-05, + "loss": 0.5527, + "step": 24509 + }, + { + "epoch": 0.94, + "learning_rate": 3.431649140805335e-05, + "loss": 0.61, + "step": 24510 + }, + { + "epoch": 0.94, + "learning_rate": 3.431585021800462e-05, + "loss": 0.4675, + "step": 24511 + }, + { + "epoch": 0.94, + "learning_rate": 3.4315209027955886e-05, + "loss": 0.3957, + "step": 24512 + }, + { + "epoch": 0.94, + "learning_rate": 3.431456783790716e-05, + "loss": 0.4604, + "step": 24513 + }, + { + "epoch": 0.94, + "learning_rate": 3.431392664785842e-05, + "loss": 0.5188, + "step": 24514 + }, + { + "epoch": 0.94, + "learning_rate": 3.4313285457809695e-05, + "loss": 0.4927, + "step": 24515 + }, + { + "epoch": 0.94, + "learning_rate": 3.4312644267760966e-05, + "loss": 0.505, + "step": 24516 + }, + { + "epoch": 0.94, + "learning_rate": 3.431200307771223e-05, + "loss": 0.4733, + "step": 24517 + }, + { + "epoch": 0.94, + "learning_rate": 3.431136188766351e-05, + "loss": 0.4308, + "step": 24518 + }, + { + "epoch": 0.94, + "learning_rate": 3.4310720697614775e-05, + "loss": 0.4704, + "step": 24519 + }, + { + "epoch": 0.94, + "learning_rate": 3.4310079507566047e-05, + "loss": 0.4622, + "step": 24520 + }, + { + "epoch": 0.94, + "learning_rate": 3.430943831751731e-05, + "loss": 0.4223, + "step": 24521 + }, + { + "epoch": 0.94, + "learning_rate": 3.430879712746858e-05, + "loss": 0.4148, + "step": 24522 + }, + { + "epoch": 0.94, + "learning_rate": 3.430815593741985e-05, + "loss": 0.4953, + "step": 24523 + }, + { + "epoch": 0.94, + "learning_rate": 3.430751474737112e-05, + "loss": 0.4782, + "step": 24524 + }, + { + "epoch": 0.94, + "learning_rate": 3.430687355732239e-05, + "loss": 0.4419, + "step": 24525 + }, + { + "epoch": 0.94, + "learning_rate": 3.430623236727366e-05, + "loss": 0.4414, + "step": 24526 + }, + { + "epoch": 0.94, + "learning_rate": 3.4305591177224935e-05, + "loss": 0.6034, + "step": 24527 + }, + { + "epoch": 0.94, + "learning_rate": 3.43049499871762e-05, + "loss": 0.4244, + "step": 24528 + }, + { + "epoch": 0.94, + "learning_rate": 3.430430879712747e-05, + "loss": 0.6395, + "step": 24529 + }, + { + "epoch": 0.94, + "learning_rate": 3.4303667607078737e-05, + "loss": 0.4362, + "step": 24530 + }, + { + "epoch": 0.94, + "learning_rate": 3.430302641703001e-05, + "loss": 0.4983, + "step": 24531 + }, + { + "epoch": 0.94, + "learning_rate": 3.430238522698127e-05, + "loss": 0.4619, + "step": 24532 + }, + { + "epoch": 0.94, + "learning_rate": 3.4301744036932545e-05, + "loss": 0.4912, + "step": 24533 + }, + { + "epoch": 0.94, + "learning_rate": 3.4301102846883823e-05, + "loss": 0.4334, + "step": 24534 + }, + { + "epoch": 0.94, + "learning_rate": 3.430046165683509e-05, + "loss": 0.5174, + "step": 24535 + }, + { + "epoch": 0.94, + "learning_rate": 3.429982046678636e-05, + "loss": 0.4811, + "step": 24536 + }, + { + "epoch": 0.94, + "learning_rate": 3.4299179276737625e-05, + "loss": 0.5194, + "step": 24537 + }, + { + "epoch": 0.94, + "learning_rate": 3.42985380866889e-05, + "loss": 0.3833, + "step": 24538 + }, + { + "epoch": 0.94, + "learning_rate": 3.429789689664016e-05, + "loss": 0.498, + "step": 24539 + }, + { + "epoch": 0.94, + "learning_rate": 3.4297255706591433e-05, + "loss": 0.5763, + "step": 24540 + }, + { + "epoch": 0.94, + "learning_rate": 3.4296614516542705e-05, + "loss": 0.3635, + "step": 24541 + }, + { + "epoch": 0.94, + "learning_rate": 3.429597332649398e-05, + "loss": 0.4598, + "step": 24542 + }, + { + "epoch": 0.94, + "learning_rate": 3.429533213644525e-05, + "loss": 0.5288, + "step": 24543 + }, + { + "epoch": 0.94, + "learning_rate": 3.4294690946396514e-05, + "loss": 0.4576, + "step": 24544 + }, + { + "epoch": 0.94, + "learning_rate": 3.4294049756347785e-05, + "loss": 0.4096, + "step": 24545 + }, + { + "epoch": 0.94, + "learning_rate": 3.429340856629905e-05, + "loss": 0.5667, + "step": 24546 + }, + { + "epoch": 0.94, + "learning_rate": 3.429276737625032e-05, + "loss": 0.461, + "step": 24547 + }, + { + "epoch": 0.94, + "learning_rate": 3.429212618620159e-05, + "loss": 0.3966, + "step": 24548 + }, + { + "epoch": 0.94, + "learning_rate": 3.429148499615286e-05, + "loss": 0.4576, + "step": 24549 + }, + { + "epoch": 0.94, + "learning_rate": 3.429084380610413e-05, + "loss": 0.4686, + "step": 24550 + }, + { + "epoch": 0.94, + "learning_rate": 3.42902026160554e-05, + "loss": 0.5482, + "step": 24551 + }, + { + "epoch": 0.94, + "learning_rate": 3.4289561426006674e-05, + "loss": 0.4802, + "step": 24552 + }, + { + "epoch": 0.94, + "learning_rate": 3.428892023595794e-05, + "loss": 0.4142, + "step": 24553 + }, + { + "epoch": 0.94, + "learning_rate": 3.428827904590921e-05, + "loss": 0.4329, + "step": 24554 + }, + { + "epoch": 0.94, + "learning_rate": 3.4287637855860475e-05, + "loss": 0.4261, + "step": 24555 + }, + { + "epoch": 0.94, + "learning_rate": 3.428699666581175e-05, + "loss": 0.4439, + "step": 24556 + }, + { + "epoch": 0.94, + "learning_rate": 3.428635547576302e-05, + "loss": 0.5403, + "step": 24557 + }, + { + "epoch": 0.94, + "learning_rate": 3.428571428571429e-05, + "loss": 0.5838, + "step": 24558 + }, + { + "epoch": 0.94, + "learning_rate": 3.4285073095665555e-05, + "loss": 0.3696, + "step": 24559 + }, + { + "epoch": 0.94, + "learning_rate": 3.428443190561683e-05, + "loss": 0.4454, + "step": 24560 + }, + { + "epoch": 0.94, + "learning_rate": 3.42837907155681e-05, + "loss": 0.465, + "step": 24561 + }, + { + "epoch": 0.94, + "learning_rate": 3.4283149525519364e-05, + "loss": 0.5764, + "step": 24562 + }, + { + "epoch": 0.94, + "learning_rate": 3.4282508335470635e-05, + "loss": 0.4063, + "step": 24563 + }, + { + "epoch": 0.94, + "learning_rate": 3.42818671454219e-05, + "loss": 0.5652, + "step": 24564 + }, + { + "epoch": 0.94, + "learning_rate": 3.428122595537317e-05, + "loss": 0.3755, + "step": 24565 + }, + { + "epoch": 0.94, + "learning_rate": 3.4280584765324444e-05, + "loss": 0.3981, + "step": 24566 + }, + { + "epoch": 0.94, + "learning_rate": 3.4279943575275716e-05, + "loss": 0.4142, + "step": 24567 + }, + { + "epoch": 0.94, + "learning_rate": 3.427930238522699e-05, + "loss": 0.5017, + "step": 24568 + }, + { + "epoch": 0.94, + "learning_rate": 3.427866119517825e-05, + "loss": 0.5568, + "step": 24569 + }, + { + "epoch": 0.94, + "learning_rate": 3.4278020005129524e-05, + "loss": 0.6094, + "step": 24570 + }, + { + "epoch": 0.94, + "learning_rate": 3.427737881508079e-05, + "loss": 0.5229, + "step": 24571 + }, + { + "epoch": 0.94, + "learning_rate": 3.427673762503206e-05, + "loss": 0.443, + "step": 24572 + }, + { + "epoch": 0.94, + "learning_rate": 3.427609643498333e-05, + "loss": 0.4001, + "step": 24573 + }, + { + "epoch": 0.94, + "learning_rate": 3.4275455244934604e-05, + "loss": 0.4777, + "step": 24574 + }, + { + "epoch": 0.94, + "learning_rate": 3.427481405488587e-05, + "loss": 0.4659, + "step": 24575 + }, + { + "epoch": 0.94, + "learning_rate": 3.427417286483714e-05, + "loss": 0.5115, + "step": 24576 + }, + { + "epoch": 0.94, + "learning_rate": 3.427353167478841e-05, + "loss": 0.4666, + "step": 24577 + }, + { + "epoch": 0.94, + "learning_rate": 3.427289048473968e-05, + "loss": 0.5298, + "step": 24578 + }, + { + "epoch": 0.94, + "learning_rate": 3.427224929469095e-05, + "loss": 0.3734, + "step": 24579 + }, + { + "epoch": 0.95, + "learning_rate": 3.4271608104642214e-05, + "loss": 0.5029, + "step": 24580 + }, + { + "epoch": 0.95, + "learning_rate": 3.4270966914593486e-05, + "loss": 0.4415, + "step": 24581 + }, + { + "epoch": 0.95, + "learning_rate": 3.427032572454476e-05, + "loss": 0.3732, + "step": 24582 + }, + { + "epoch": 0.95, + "learning_rate": 3.426968453449603e-05, + "loss": 0.4307, + "step": 24583 + }, + { + "epoch": 0.95, + "learning_rate": 3.4269043344447294e-05, + "loss": 0.5175, + "step": 24584 + }, + { + "epoch": 0.95, + "learning_rate": 3.4268402154398566e-05, + "loss": 0.3734, + "step": 24585 + }, + { + "epoch": 0.95, + "learning_rate": 3.426776096434984e-05, + "loss": 0.526, + "step": 24586 + }, + { + "epoch": 0.95, + "learning_rate": 3.42671197743011e-05, + "loss": 0.4492, + "step": 24587 + }, + { + "epoch": 0.95, + "learning_rate": 3.4266478584252374e-05, + "loss": 0.6304, + "step": 24588 + }, + { + "epoch": 0.95, + "learning_rate": 3.4265837394203646e-05, + "loss": 0.4106, + "step": 24589 + }, + { + "epoch": 0.95, + "learning_rate": 3.426519620415492e-05, + "loss": 0.3923, + "step": 24590 + }, + { + "epoch": 0.95, + "learning_rate": 3.426455501410618e-05, + "loss": 0.4072, + "step": 24591 + }, + { + "epoch": 0.95, + "learning_rate": 3.4263913824057454e-05, + "loss": 0.4809, + "step": 24592 + }, + { + "epoch": 0.95, + "learning_rate": 3.426327263400872e-05, + "loss": 0.5257, + "step": 24593 + }, + { + "epoch": 0.95, + "learning_rate": 3.426263144395999e-05, + "loss": 0.4814, + "step": 24594 + }, + { + "epoch": 0.95, + "learning_rate": 3.426199025391126e-05, + "loss": 0.4648, + "step": 24595 + }, + { + "epoch": 0.95, + "learning_rate": 3.426134906386253e-05, + "loss": 0.4126, + "step": 24596 + }, + { + "epoch": 0.95, + "learning_rate": 3.42607078738138e-05, + "loss": 0.5656, + "step": 24597 + }, + { + "epoch": 0.95, + "learning_rate": 3.426006668376507e-05, + "loss": 0.6817, + "step": 24598 + }, + { + "epoch": 0.95, + "learning_rate": 3.425942549371634e-05, + "loss": 0.4371, + "step": 24599 + }, + { + "epoch": 0.95, + "learning_rate": 3.425878430366761e-05, + "loss": 0.4633, + "step": 24600 + }, + { + "epoch": 0.95, + "learning_rate": 3.425814311361888e-05, + "loss": 0.4943, + "step": 24601 + }, + { + "epoch": 0.95, + "learning_rate": 3.4257501923570144e-05, + "loss": 0.5477, + "step": 24602 + }, + { + "epoch": 0.95, + "learning_rate": 3.4256860733521416e-05, + "loss": 0.4609, + "step": 24603 + }, + { + "epoch": 0.95, + "learning_rate": 3.425621954347269e-05, + "loss": 0.4417, + "step": 24604 + }, + { + "epoch": 0.95, + "learning_rate": 3.425557835342395e-05, + "loss": 0.4528, + "step": 24605 + }, + { + "epoch": 0.95, + "learning_rate": 3.425493716337523e-05, + "loss": 0.618, + "step": 24606 + }, + { + "epoch": 0.95, + "learning_rate": 3.4254295973326496e-05, + "loss": 0.4497, + "step": 24607 + }, + { + "epoch": 0.95, + "learning_rate": 3.425365478327777e-05, + "loss": 0.498, + "step": 24608 + }, + { + "epoch": 0.95, + "learning_rate": 3.425301359322903e-05, + "loss": 0.4217, + "step": 24609 + }, + { + "epoch": 0.95, + "learning_rate": 3.4252372403180305e-05, + "loss": 0.5973, + "step": 24610 + }, + { + "epoch": 0.95, + "learning_rate": 3.425173121313157e-05, + "loss": 0.4136, + "step": 24611 + }, + { + "epoch": 0.95, + "learning_rate": 3.425109002308284e-05, + "loss": 0.5005, + "step": 24612 + }, + { + "epoch": 0.95, + "learning_rate": 3.425044883303411e-05, + "loss": 0.5311, + "step": 24613 + }, + { + "epoch": 0.95, + "learning_rate": 3.4249807642985385e-05, + "loss": 0.4541, + "step": 24614 + }, + { + "epoch": 0.95, + "learning_rate": 3.4249166452936656e-05, + "loss": 0.5373, + "step": 24615 + }, + { + "epoch": 0.95, + "learning_rate": 3.424852526288792e-05, + "loss": 0.484, + "step": 24616 + }, + { + "epoch": 0.95, + "learning_rate": 3.424788407283919e-05, + "loss": 0.4771, + "step": 24617 + }, + { + "epoch": 0.95, + "learning_rate": 3.424724288279046e-05, + "loss": 0.5751, + "step": 24618 + }, + { + "epoch": 0.95, + "learning_rate": 3.424660169274173e-05, + "loss": 0.6178, + "step": 24619 + }, + { + "epoch": 0.95, + "learning_rate": 3.4245960502692995e-05, + "loss": 0.4675, + "step": 24620 + }, + { + "epoch": 0.95, + "learning_rate": 3.4245319312644266e-05, + "loss": 0.5171, + "step": 24621 + }, + { + "epoch": 0.95, + "learning_rate": 3.4244678122595545e-05, + "loss": 0.4678, + "step": 24622 + }, + { + "epoch": 0.95, + "learning_rate": 3.424403693254681e-05, + "loss": 0.4089, + "step": 24623 + }, + { + "epoch": 0.95, + "learning_rate": 3.424339574249808e-05, + "loss": 0.454, + "step": 24624 + }, + { + "epoch": 0.95, + "learning_rate": 3.4242754552449346e-05, + "loss": 0.4222, + "step": 24625 + }, + { + "epoch": 0.95, + "learning_rate": 3.424211336240062e-05, + "loss": 0.4342, + "step": 24626 + }, + { + "epoch": 0.95, + "learning_rate": 3.424147217235188e-05, + "loss": 0.4078, + "step": 24627 + }, + { + "epoch": 0.95, + "learning_rate": 3.4240830982303155e-05, + "loss": 0.4309, + "step": 24628 + }, + { + "epoch": 0.95, + "learning_rate": 3.4240189792254426e-05, + "loss": 0.4014, + "step": 24629 + }, + { + "epoch": 0.95, + "learning_rate": 3.42395486022057e-05, + "loss": 0.4585, + "step": 24630 + }, + { + "epoch": 0.95, + "learning_rate": 3.423890741215697e-05, + "loss": 0.4266, + "step": 24631 + }, + { + "epoch": 0.95, + "learning_rate": 3.4238266222108235e-05, + "loss": 0.4952, + "step": 24632 + }, + { + "epoch": 0.95, + "learning_rate": 3.4237625032059507e-05, + "loss": 0.5663, + "step": 24633 + }, + { + "epoch": 0.95, + "learning_rate": 3.423698384201077e-05, + "loss": 0.4828, + "step": 24634 + }, + { + "epoch": 0.95, + "learning_rate": 3.423634265196204e-05, + "loss": 0.4133, + "step": 24635 + }, + { + "epoch": 0.95, + "learning_rate": 3.423570146191331e-05, + "loss": 0.5698, + "step": 24636 + }, + { + "epoch": 0.95, + "learning_rate": 3.423506027186458e-05, + "loss": 0.4637, + "step": 24637 + }, + { + "epoch": 0.95, + "learning_rate": 3.423441908181585e-05, + "loss": 0.5356, + "step": 24638 + }, + { + "epoch": 0.95, + "learning_rate": 3.423377789176712e-05, + "loss": 0.5026, + "step": 24639 + }, + { + "epoch": 0.95, + "learning_rate": 3.4233136701718395e-05, + "loss": 0.4272, + "step": 24640 + }, + { + "epoch": 0.95, + "learning_rate": 3.423249551166966e-05, + "loss": 0.4694, + "step": 24641 + }, + { + "epoch": 0.95, + "learning_rate": 3.423185432162093e-05, + "loss": 0.5806, + "step": 24642 + }, + { + "epoch": 0.95, + "learning_rate": 3.42312131315722e-05, + "loss": 0.4483, + "step": 24643 + }, + { + "epoch": 0.95, + "learning_rate": 3.423057194152347e-05, + "loss": 0.4544, + "step": 24644 + }, + { + "epoch": 0.95, + "learning_rate": 3.422993075147474e-05, + "loss": 0.443, + "step": 24645 + }, + { + "epoch": 0.95, + "learning_rate": 3.422928956142601e-05, + "loss": 0.4799, + "step": 24646 + }, + { + "epoch": 0.95, + "learning_rate": 3.422864837137728e-05, + "loss": 0.39, + "step": 24647 + }, + { + "epoch": 0.95, + "learning_rate": 3.422800718132855e-05, + "loss": 0.433, + "step": 24648 + }, + { + "epoch": 0.95, + "learning_rate": 3.422736599127982e-05, + "loss": 0.6093, + "step": 24649 + }, + { + "epoch": 0.95, + "learning_rate": 3.4226724801231085e-05, + "loss": 0.3986, + "step": 24650 + }, + { + "epoch": 0.95, + "learning_rate": 3.422608361118236e-05, + "loss": 0.4981, + "step": 24651 + }, + { + "epoch": 0.95, + "learning_rate": 3.422544242113362e-05, + "loss": 0.521, + "step": 24652 + }, + { + "epoch": 0.95, + "learning_rate": 3.4224801231084893e-05, + "loss": 0.5311, + "step": 24653 + }, + { + "epoch": 0.95, + "learning_rate": 3.4224160041036165e-05, + "loss": 0.3946, + "step": 24654 + }, + { + "epoch": 0.95, + "learning_rate": 3.422351885098744e-05, + "loss": 0.3782, + "step": 24655 + }, + { + "epoch": 0.95, + "learning_rate": 3.42228776609387e-05, + "loss": 0.489, + "step": 24656 + }, + { + "epoch": 0.95, + "learning_rate": 3.4222236470889974e-05, + "loss": 0.4855, + "step": 24657 + }, + { + "epoch": 0.95, + "learning_rate": 3.4221595280841245e-05, + "loss": 0.474, + "step": 24658 + }, + { + "epoch": 0.95, + "learning_rate": 3.422095409079251e-05, + "loss": 0.4483, + "step": 24659 + }, + { + "epoch": 0.95, + "learning_rate": 3.422031290074378e-05, + "loss": 0.3965, + "step": 24660 + }, + { + "epoch": 0.95, + "learning_rate": 3.4219671710695054e-05, + "loss": 0.4908, + "step": 24661 + }, + { + "epoch": 0.95, + "learning_rate": 3.4219030520646325e-05, + "loss": 0.4705, + "step": 24662 + }, + { + "epoch": 0.95, + "learning_rate": 3.421838933059759e-05, + "loss": 0.4041, + "step": 24663 + }, + { + "epoch": 0.95, + "learning_rate": 3.421774814054886e-05, + "loss": 0.5353, + "step": 24664 + }, + { + "epoch": 0.95, + "learning_rate": 3.421710695050013e-05, + "loss": 0.4804, + "step": 24665 + }, + { + "epoch": 0.95, + "learning_rate": 3.42164657604514e-05, + "loss": 0.3543, + "step": 24666 + }, + { + "epoch": 0.95, + "learning_rate": 3.421582457040267e-05, + "loss": 0.4151, + "step": 24667 + }, + { + "epoch": 0.95, + "learning_rate": 3.4215183380353935e-05, + "loss": 0.5362, + "step": 24668 + }, + { + "epoch": 0.95, + "learning_rate": 3.421454219030521e-05, + "loss": 0.4226, + "step": 24669 + }, + { + "epoch": 0.95, + "learning_rate": 3.421390100025648e-05, + "loss": 0.408, + "step": 24670 + }, + { + "epoch": 0.95, + "learning_rate": 3.421325981020775e-05, + "loss": 0.5514, + "step": 24671 + }, + { + "epoch": 0.95, + "learning_rate": 3.4212618620159015e-05, + "loss": 0.4618, + "step": 24672 + }, + { + "epoch": 0.95, + "learning_rate": 3.421197743011029e-05, + "loss": 0.5663, + "step": 24673 + }, + { + "epoch": 0.95, + "learning_rate": 3.421133624006155e-05, + "loss": 0.5103, + "step": 24674 + }, + { + "epoch": 0.95, + "learning_rate": 3.4210695050012824e-05, + "loss": 0.4907, + "step": 24675 + }, + { + "epoch": 0.95, + "learning_rate": 3.4210053859964096e-05, + "loss": 0.5557, + "step": 24676 + }, + { + "epoch": 0.95, + "learning_rate": 3.420941266991537e-05, + "loss": 0.4722, + "step": 24677 + }, + { + "epoch": 0.95, + "learning_rate": 3.420877147986664e-05, + "loss": 0.659, + "step": 24678 + }, + { + "epoch": 0.95, + "learning_rate": 3.4208130289817904e-05, + "loss": 0.5417, + "step": 24679 + }, + { + "epoch": 0.95, + "learning_rate": 3.4207489099769176e-05, + "loss": 0.3023, + "step": 24680 + }, + { + "epoch": 0.95, + "learning_rate": 3.420684790972044e-05, + "loss": 0.438, + "step": 24681 + }, + { + "epoch": 0.95, + "learning_rate": 3.420620671967171e-05, + "loss": 0.5718, + "step": 24682 + }, + { + "epoch": 0.95, + "learning_rate": 3.420556552962298e-05, + "loss": 0.447, + "step": 24683 + }, + { + "epoch": 0.95, + "learning_rate": 3.420492433957425e-05, + "loss": 0.5173, + "step": 24684 + }, + { + "epoch": 0.95, + "learning_rate": 3.420428314952552e-05, + "loss": 0.4291, + "step": 24685 + }, + { + "epoch": 0.95, + "learning_rate": 3.420364195947679e-05, + "loss": 0.4562, + "step": 24686 + }, + { + "epoch": 0.95, + "learning_rate": 3.4203000769428064e-05, + "loss": 0.4974, + "step": 24687 + }, + { + "epoch": 0.95, + "learning_rate": 3.420235957937933e-05, + "loss": 0.5427, + "step": 24688 + }, + { + "epoch": 0.95, + "learning_rate": 3.42017183893306e-05, + "loss": 0.4236, + "step": 24689 + }, + { + "epoch": 0.95, + "learning_rate": 3.4201077199281866e-05, + "loss": 0.5936, + "step": 24690 + }, + { + "epoch": 0.95, + "learning_rate": 3.420043600923314e-05, + "loss": 0.3464, + "step": 24691 + }, + { + "epoch": 0.95, + "learning_rate": 3.41997948191844e-05, + "loss": 0.5255, + "step": 24692 + }, + { + "epoch": 0.95, + "learning_rate": 3.4199153629135674e-05, + "loss": 0.5583, + "step": 24693 + }, + { + "epoch": 0.95, + "learning_rate": 3.419851243908695e-05, + "loss": 0.5105, + "step": 24694 + }, + { + "epoch": 0.95, + "learning_rate": 3.419787124903822e-05, + "loss": 0.5295, + "step": 24695 + }, + { + "epoch": 0.95, + "learning_rate": 3.419723005898949e-05, + "loss": 0.4613, + "step": 24696 + }, + { + "epoch": 0.95, + "learning_rate": 3.4196588868940754e-05, + "loss": 0.457, + "step": 24697 + }, + { + "epoch": 0.95, + "learning_rate": 3.4195947678892026e-05, + "loss": 0.4795, + "step": 24698 + }, + { + "epoch": 0.95, + "learning_rate": 3.419530648884329e-05, + "loss": 0.5036, + "step": 24699 + }, + { + "epoch": 0.95, + "learning_rate": 3.419466529879456e-05, + "loss": 0.4482, + "step": 24700 + }, + { + "epoch": 0.95, + "learning_rate": 3.4194024108745834e-05, + "loss": 0.4222, + "step": 24701 + }, + { + "epoch": 0.95, + "learning_rate": 3.4193382918697106e-05, + "loss": 0.4227, + "step": 24702 + }, + { + "epoch": 0.95, + "learning_rate": 3.419274172864838e-05, + "loss": 0.4907, + "step": 24703 + }, + { + "epoch": 0.95, + "learning_rate": 3.419210053859964e-05, + "loss": 0.5888, + "step": 24704 + }, + { + "epoch": 0.95, + "learning_rate": 3.4191459348550914e-05, + "loss": 0.3728, + "step": 24705 + }, + { + "epoch": 0.95, + "learning_rate": 3.419081815850218e-05, + "loss": 0.4011, + "step": 24706 + }, + { + "epoch": 0.95, + "learning_rate": 3.419017696845345e-05, + "loss": 0.4337, + "step": 24707 + }, + { + "epoch": 0.95, + "learning_rate": 3.4189535778404716e-05, + "loss": 0.4612, + "step": 24708 + }, + { + "epoch": 0.95, + "learning_rate": 3.418889458835599e-05, + "loss": 0.4243, + "step": 24709 + }, + { + "epoch": 0.95, + "learning_rate": 3.418825339830726e-05, + "loss": 0.4822, + "step": 24710 + }, + { + "epoch": 0.95, + "learning_rate": 3.418761220825853e-05, + "loss": 0.5241, + "step": 24711 + }, + { + "epoch": 0.95, + "learning_rate": 3.41869710182098e-05, + "loss": 0.6086, + "step": 24712 + }, + { + "epoch": 0.95, + "learning_rate": 3.418632982816107e-05, + "loss": 0.4495, + "step": 24713 + }, + { + "epoch": 0.95, + "learning_rate": 3.418568863811234e-05, + "loss": 0.4382, + "step": 24714 + }, + { + "epoch": 0.95, + "learning_rate": 3.4185047448063604e-05, + "loss": 0.4993, + "step": 24715 + }, + { + "epoch": 0.95, + "learning_rate": 3.4184406258014876e-05, + "loss": 0.4782, + "step": 24716 + }, + { + "epoch": 0.95, + "learning_rate": 3.418376506796615e-05, + "loss": 0.4236, + "step": 24717 + }, + { + "epoch": 0.95, + "learning_rate": 3.418312387791742e-05, + "loss": 0.4814, + "step": 24718 + }, + { + "epoch": 0.95, + "learning_rate": 3.4182482687868684e-05, + "loss": 0.5086, + "step": 24719 + }, + { + "epoch": 0.95, + "learning_rate": 3.4181841497819956e-05, + "loss": 0.6608, + "step": 24720 + }, + { + "epoch": 0.95, + "learning_rate": 3.418120030777123e-05, + "loss": 0.535, + "step": 24721 + }, + { + "epoch": 0.95, + "learning_rate": 3.418055911772249e-05, + "loss": 0.3857, + "step": 24722 + }, + { + "epoch": 0.95, + "learning_rate": 3.4179917927673765e-05, + "loss": 0.4764, + "step": 24723 + }, + { + "epoch": 0.95, + "learning_rate": 3.417927673762503e-05, + "loss": 0.5054, + "step": 24724 + }, + { + "epoch": 0.95, + "learning_rate": 3.41786355475763e-05, + "loss": 0.5222, + "step": 24725 + }, + { + "epoch": 0.95, + "learning_rate": 3.417799435752757e-05, + "loss": 0.4238, + "step": 24726 + }, + { + "epoch": 0.95, + "learning_rate": 3.4177353167478845e-05, + "loss": 0.4256, + "step": 24727 + }, + { + "epoch": 0.95, + "learning_rate": 3.417671197743011e-05, + "loss": 0.6077, + "step": 24728 + }, + { + "epoch": 0.95, + "learning_rate": 3.417607078738138e-05, + "loss": 0.5389, + "step": 24729 + }, + { + "epoch": 0.95, + "learning_rate": 3.417542959733265e-05, + "loss": 0.5731, + "step": 24730 + }, + { + "epoch": 0.95, + "learning_rate": 3.417478840728392e-05, + "loss": 0.5435, + "step": 24731 + }, + { + "epoch": 0.95, + "learning_rate": 3.417414721723519e-05, + "loss": 0.4777, + "step": 24732 + }, + { + "epoch": 0.95, + "learning_rate": 3.417350602718646e-05, + "loss": 0.5447, + "step": 24733 + }, + { + "epoch": 0.95, + "learning_rate": 3.417286483713773e-05, + "loss": 0.4152, + "step": 24734 + }, + { + "epoch": 0.95, + "learning_rate": 3.4172223647089e-05, + "loss": 0.462, + "step": 24735 + }, + { + "epoch": 0.95, + "learning_rate": 3.417158245704027e-05, + "loss": 0.4158, + "step": 24736 + }, + { + "epoch": 0.95, + "learning_rate": 3.4170941266991535e-05, + "loss": 0.4119, + "step": 24737 + }, + { + "epoch": 0.95, + "learning_rate": 3.4170300076942806e-05, + "loss": 0.4036, + "step": 24738 + }, + { + "epoch": 0.95, + "learning_rate": 3.416965888689408e-05, + "loss": 0.408, + "step": 24739 + }, + { + "epoch": 0.95, + "learning_rate": 3.416901769684534e-05, + "loss": 0.5046, + "step": 24740 + }, + { + "epoch": 0.95, + "learning_rate": 3.4168376506796615e-05, + "loss": 0.5031, + "step": 24741 + }, + { + "epoch": 0.95, + "learning_rate": 3.4167735316747887e-05, + "loss": 0.5227, + "step": 24742 + }, + { + "epoch": 0.95, + "learning_rate": 3.416709412669916e-05, + "loss": 0.4455, + "step": 24743 + }, + { + "epoch": 0.95, + "learning_rate": 3.416645293665042e-05, + "loss": 0.4728, + "step": 24744 + }, + { + "epoch": 0.95, + "learning_rate": 3.4165811746601695e-05, + "loss": 0.4874, + "step": 24745 + }, + { + "epoch": 0.95, + "learning_rate": 3.416517055655296e-05, + "loss": 0.5303, + "step": 24746 + }, + { + "epoch": 0.95, + "learning_rate": 3.416452936650423e-05, + "loss": 0.5612, + "step": 24747 + }, + { + "epoch": 0.95, + "learning_rate": 3.41638881764555e-05, + "loss": 0.4803, + "step": 24748 + }, + { + "epoch": 0.95, + "learning_rate": 3.4163246986406775e-05, + "loss": 0.515, + "step": 24749 + }, + { + "epoch": 0.95, + "learning_rate": 3.416260579635805e-05, + "loss": 0.4589, + "step": 24750 + }, + { + "epoch": 0.95, + "learning_rate": 3.416196460630931e-05, + "loss": 0.4636, + "step": 24751 + }, + { + "epoch": 0.95, + "learning_rate": 3.416132341626058e-05, + "loss": 0.5055, + "step": 24752 + }, + { + "epoch": 0.95, + "learning_rate": 3.416068222621185e-05, + "loss": 0.4998, + "step": 24753 + }, + { + "epoch": 0.95, + "learning_rate": 3.416004103616312e-05, + "loss": 0.4721, + "step": 24754 + }, + { + "epoch": 0.95, + "learning_rate": 3.415939984611439e-05, + "loss": 0.4458, + "step": 24755 + }, + { + "epoch": 0.95, + "learning_rate": 3.415875865606566e-05, + "loss": 0.4663, + "step": 24756 + }, + { + "epoch": 0.95, + "learning_rate": 3.415811746601693e-05, + "loss": 0.5299, + "step": 24757 + }, + { + "epoch": 0.95, + "learning_rate": 3.41574762759682e-05, + "loss": 0.6337, + "step": 24758 + }, + { + "epoch": 0.95, + "learning_rate": 3.415683508591947e-05, + "loss": 0.4255, + "step": 24759 + }, + { + "epoch": 0.95, + "learning_rate": 3.415619389587074e-05, + "loss": 0.5852, + "step": 24760 + }, + { + "epoch": 0.95, + "learning_rate": 3.415555270582201e-05, + "loss": 0.5338, + "step": 24761 + }, + { + "epoch": 0.95, + "learning_rate": 3.4154911515773273e-05, + "loss": 0.5263, + "step": 24762 + }, + { + "epoch": 0.95, + "learning_rate": 3.4154270325724545e-05, + "loss": 0.4505, + "step": 24763 + }, + { + "epoch": 0.95, + "learning_rate": 3.415362913567582e-05, + "loss": 0.5118, + "step": 24764 + }, + { + "epoch": 0.95, + "learning_rate": 3.415298794562708e-05, + "loss": 0.4974, + "step": 24765 + }, + { + "epoch": 0.95, + "learning_rate": 3.415234675557836e-05, + "loss": 0.496, + "step": 24766 + }, + { + "epoch": 0.95, + "learning_rate": 3.4151705565529625e-05, + "loss": 0.4258, + "step": 24767 + }, + { + "epoch": 0.95, + "learning_rate": 3.41510643754809e-05, + "loss": 0.4431, + "step": 24768 + }, + { + "epoch": 0.95, + "learning_rate": 3.415042318543216e-05, + "loss": 0.5018, + "step": 24769 + }, + { + "epoch": 0.95, + "learning_rate": 3.4149781995383434e-05, + "loss": 0.4995, + "step": 24770 + }, + { + "epoch": 0.95, + "learning_rate": 3.41491408053347e-05, + "loss": 0.5254, + "step": 24771 + }, + { + "epoch": 0.95, + "learning_rate": 3.414849961528597e-05, + "loss": 0.4855, + "step": 24772 + }, + { + "epoch": 0.95, + "learning_rate": 3.414785842523724e-05, + "loss": 0.4706, + "step": 24773 + }, + { + "epoch": 0.95, + "learning_rate": 3.4147217235188514e-05, + "loss": 0.5374, + "step": 24774 + }, + { + "epoch": 0.95, + "learning_rate": 3.4146576045139785e-05, + "loss": 0.5199, + "step": 24775 + }, + { + "epoch": 0.95, + "learning_rate": 3.414593485509105e-05, + "loss": 0.4707, + "step": 24776 + }, + { + "epoch": 0.95, + "learning_rate": 3.414529366504232e-05, + "loss": 0.4473, + "step": 24777 + }, + { + "epoch": 0.95, + "learning_rate": 3.414465247499359e-05, + "loss": 0.431, + "step": 24778 + }, + { + "epoch": 0.95, + "learning_rate": 3.414401128494486e-05, + "loss": 0.4343, + "step": 24779 + }, + { + "epoch": 0.95, + "learning_rate": 3.4143370094896124e-05, + "loss": 0.5068, + "step": 24780 + }, + { + "epoch": 0.95, + "learning_rate": 3.4142728904847395e-05, + "loss": 0.4566, + "step": 24781 + }, + { + "epoch": 0.95, + "learning_rate": 3.4142087714798674e-05, + "loss": 0.4835, + "step": 24782 + }, + { + "epoch": 0.95, + "learning_rate": 3.414144652474994e-05, + "loss": 0.4914, + "step": 24783 + }, + { + "epoch": 0.95, + "learning_rate": 3.414080533470121e-05, + "loss": 0.3703, + "step": 24784 + }, + { + "epoch": 0.95, + "learning_rate": 3.4140164144652475e-05, + "loss": 0.3869, + "step": 24785 + }, + { + "epoch": 0.95, + "learning_rate": 3.413952295460375e-05, + "loss": 0.4571, + "step": 24786 + }, + { + "epoch": 0.95, + "learning_rate": 3.413888176455501e-05, + "loss": 0.4587, + "step": 24787 + }, + { + "epoch": 0.95, + "learning_rate": 3.4138240574506284e-05, + "loss": 0.4773, + "step": 24788 + }, + { + "epoch": 0.95, + "learning_rate": 3.4137599384457556e-05, + "loss": 0.4695, + "step": 24789 + }, + { + "epoch": 0.95, + "learning_rate": 3.413695819440883e-05, + "loss": 0.4403, + "step": 24790 + }, + { + "epoch": 0.95, + "learning_rate": 3.41363170043601e-05, + "loss": 0.406, + "step": 24791 + }, + { + "epoch": 0.95, + "learning_rate": 3.4135675814311364e-05, + "loss": 0.4862, + "step": 24792 + }, + { + "epoch": 0.95, + "learning_rate": 3.4135034624262636e-05, + "loss": 0.4283, + "step": 24793 + }, + { + "epoch": 0.95, + "learning_rate": 3.41343934342139e-05, + "loss": 0.5227, + "step": 24794 + }, + { + "epoch": 0.95, + "learning_rate": 3.413375224416517e-05, + "loss": 0.468, + "step": 24795 + }, + { + "epoch": 0.95, + "learning_rate": 3.413311105411644e-05, + "loss": 0.4954, + "step": 24796 + }, + { + "epoch": 0.95, + "learning_rate": 3.413246986406771e-05, + "loss": 0.4702, + "step": 24797 + }, + { + "epoch": 0.95, + "learning_rate": 3.413182867401898e-05, + "loss": 0.5805, + "step": 24798 + }, + { + "epoch": 0.95, + "learning_rate": 3.413118748397025e-05, + "loss": 0.4541, + "step": 24799 + }, + { + "epoch": 0.95, + "learning_rate": 3.4130546293921524e-05, + "loss": 0.4106, + "step": 24800 + }, + { + "epoch": 0.95, + "learning_rate": 3.412990510387279e-05, + "loss": 0.6173, + "step": 24801 + }, + { + "epoch": 0.95, + "learning_rate": 3.412926391382406e-05, + "loss": 0.5313, + "step": 24802 + }, + { + "epoch": 0.95, + "learning_rate": 3.4128622723775326e-05, + "loss": 0.5171, + "step": 24803 + }, + { + "epoch": 0.95, + "learning_rate": 3.41279815337266e-05, + "loss": 0.574, + "step": 24804 + }, + { + "epoch": 0.95, + "learning_rate": 3.412734034367787e-05, + "loss": 0.5684, + "step": 24805 + }, + { + "epoch": 0.95, + "learning_rate": 3.412669915362914e-05, + "loss": 0.4229, + "step": 24806 + }, + { + "epoch": 0.95, + "learning_rate": 3.4126057963580406e-05, + "loss": 0.4646, + "step": 24807 + }, + { + "epoch": 0.95, + "learning_rate": 3.412541677353168e-05, + "loss": 0.3578, + "step": 24808 + }, + { + "epoch": 0.95, + "learning_rate": 3.412477558348295e-05, + "loss": 0.5918, + "step": 24809 + }, + { + "epoch": 0.95, + "learning_rate": 3.4124134393434214e-05, + "loss": 0.4529, + "step": 24810 + }, + { + "epoch": 0.95, + "learning_rate": 3.4123493203385486e-05, + "loss": 0.52, + "step": 24811 + }, + { + "epoch": 0.95, + "learning_rate": 3.412285201333675e-05, + "loss": 0.5806, + "step": 24812 + }, + { + "epoch": 0.95, + "learning_rate": 3.412221082328802e-05, + "loss": 0.4449, + "step": 24813 + }, + { + "epoch": 0.95, + "learning_rate": 3.4121569633239294e-05, + "loss": 0.4547, + "step": 24814 + }, + { + "epoch": 0.95, + "learning_rate": 3.4120928443190566e-05, + "loss": 0.568, + "step": 24815 + }, + { + "epoch": 0.95, + "learning_rate": 3.412028725314183e-05, + "loss": 0.4054, + "step": 24816 + }, + { + "epoch": 0.95, + "learning_rate": 3.41196460630931e-05, + "loss": 0.4439, + "step": 24817 + }, + { + "epoch": 0.95, + "learning_rate": 3.4119004873044374e-05, + "loss": 0.4202, + "step": 24818 + }, + { + "epoch": 0.95, + "learning_rate": 3.411836368299564e-05, + "loss": 0.4948, + "step": 24819 + }, + { + "epoch": 0.95, + "learning_rate": 3.411772249294691e-05, + "loss": 0.4158, + "step": 24820 + }, + { + "epoch": 0.95, + "learning_rate": 3.411708130289818e-05, + "loss": 0.4486, + "step": 24821 + }, + { + "epoch": 0.95, + "learning_rate": 3.4116440112849454e-05, + "loss": 0.5005, + "step": 24822 + }, + { + "epoch": 0.95, + "learning_rate": 3.411579892280072e-05, + "loss": 0.4118, + "step": 24823 + }, + { + "epoch": 0.95, + "learning_rate": 3.411515773275199e-05, + "loss": 0.5006, + "step": 24824 + }, + { + "epoch": 0.95, + "learning_rate": 3.4114516542703256e-05, + "loss": 0.5061, + "step": 24825 + }, + { + "epoch": 0.95, + "learning_rate": 3.411387535265453e-05, + "loss": 0.4511, + "step": 24826 + }, + { + "epoch": 0.95, + "learning_rate": 3.41132341626058e-05, + "loss": 0.4365, + "step": 24827 + }, + { + "epoch": 0.95, + "learning_rate": 3.4112592972557064e-05, + "loss": 0.429, + "step": 24828 + }, + { + "epoch": 0.95, + "learning_rate": 3.4111951782508336e-05, + "loss": 0.398, + "step": 24829 + }, + { + "epoch": 0.95, + "learning_rate": 3.411131059245961e-05, + "loss": 0.5349, + "step": 24830 + }, + { + "epoch": 0.95, + "learning_rate": 3.411066940241088e-05, + "loss": 0.4296, + "step": 24831 + }, + { + "epoch": 0.95, + "learning_rate": 3.4110028212362144e-05, + "loss": 0.3997, + "step": 24832 + }, + { + "epoch": 0.95, + "learning_rate": 3.4109387022313416e-05, + "loss": 0.406, + "step": 24833 + }, + { + "epoch": 0.95, + "learning_rate": 3.410874583226468e-05, + "loss": 0.5112, + "step": 24834 + }, + { + "epoch": 0.95, + "learning_rate": 3.410810464221595e-05, + "loss": 0.4918, + "step": 24835 + }, + { + "epoch": 0.95, + "learning_rate": 3.4107463452167225e-05, + "loss": 0.5948, + "step": 24836 + }, + { + "epoch": 0.95, + "learning_rate": 3.4106822262118496e-05, + "loss": 0.4802, + "step": 24837 + }, + { + "epoch": 0.95, + "learning_rate": 3.410618107206977e-05, + "loss": 0.6003, + "step": 24838 + }, + { + "epoch": 0.95, + "learning_rate": 3.410553988202103e-05, + "loss": 0.5843, + "step": 24839 + }, + { + "epoch": 0.96, + "learning_rate": 3.4104898691972305e-05, + "loss": 0.4164, + "step": 24840 + }, + { + "epoch": 0.96, + "learning_rate": 3.410425750192357e-05, + "loss": 0.3841, + "step": 24841 + }, + { + "epoch": 0.96, + "learning_rate": 3.410361631187484e-05, + "loss": 0.478, + "step": 24842 + }, + { + "epoch": 0.96, + "learning_rate": 3.4102975121826106e-05, + "loss": 0.3682, + "step": 24843 + }, + { + "epoch": 0.96, + "learning_rate": 3.410233393177738e-05, + "loss": 0.4552, + "step": 24844 + }, + { + "epoch": 0.96, + "learning_rate": 3.410169274172865e-05, + "loss": 0.4815, + "step": 24845 + }, + { + "epoch": 0.96, + "learning_rate": 3.410105155167992e-05, + "loss": 0.5018, + "step": 24846 + }, + { + "epoch": 0.96, + "learning_rate": 3.410041036163119e-05, + "loss": 0.3973, + "step": 24847 + }, + { + "epoch": 0.96, + "learning_rate": 3.409976917158246e-05, + "loss": 0.5482, + "step": 24848 + }, + { + "epoch": 0.96, + "learning_rate": 3.409912798153373e-05, + "loss": 0.5461, + "step": 24849 + }, + { + "epoch": 0.96, + "learning_rate": 3.4098486791484995e-05, + "loss": 0.4907, + "step": 24850 + }, + { + "epoch": 0.96, + "learning_rate": 3.4097845601436266e-05, + "loss": 0.4909, + "step": 24851 + }, + { + "epoch": 0.96, + "learning_rate": 3.409720441138753e-05, + "loss": 0.4261, + "step": 24852 + }, + { + "epoch": 0.96, + "learning_rate": 3.40965632213388e-05, + "loss": 0.4361, + "step": 24853 + }, + { + "epoch": 0.96, + "learning_rate": 3.409592203129008e-05, + "loss": 0.5266, + "step": 24854 + }, + { + "epoch": 0.96, + "learning_rate": 3.4095280841241347e-05, + "loss": 0.4854, + "step": 24855 + }, + { + "epoch": 0.96, + "learning_rate": 3.409463965119262e-05, + "loss": 0.5307, + "step": 24856 + }, + { + "epoch": 0.96, + "learning_rate": 3.409399846114388e-05, + "loss": 0.473, + "step": 24857 + }, + { + "epoch": 0.96, + "learning_rate": 3.4093357271095155e-05, + "loss": 0.4743, + "step": 24858 + }, + { + "epoch": 0.96, + "learning_rate": 3.409271608104642e-05, + "loss": 0.5491, + "step": 24859 + }, + { + "epoch": 0.96, + "learning_rate": 3.409207489099769e-05, + "loss": 0.5637, + "step": 24860 + }, + { + "epoch": 0.96, + "learning_rate": 3.409143370094896e-05, + "loss": 0.5103, + "step": 24861 + }, + { + "epoch": 0.96, + "learning_rate": 3.4090792510900235e-05, + "loss": 0.519, + "step": 24862 + }, + { + "epoch": 0.96, + "learning_rate": 3.409015132085151e-05, + "loss": 0.532, + "step": 24863 + }, + { + "epoch": 0.96, + "learning_rate": 3.408951013080277e-05, + "loss": 0.6028, + "step": 24864 + }, + { + "epoch": 0.96, + "learning_rate": 3.408886894075404e-05, + "loss": 0.5608, + "step": 24865 + }, + { + "epoch": 0.96, + "learning_rate": 3.408822775070531e-05, + "loss": 0.3851, + "step": 24866 + }, + { + "epoch": 0.96, + "learning_rate": 3.408758656065658e-05, + "loss": 0.4933, + "step": 24867 + }, + { + "epoch": 0.96, + "learning_rate": 3.4086945370607845e-05, + "loss": 0.5072, + "step": 24868 + }, + { + "epoch": 0.96, + "learning_rate": 3.408630418055912e-05, + "loss": 0.4893, + "step": 24869 + }, + { + "epoch": 0.96, + "learning_rate": 3.408566299051039e-05, + "loss": 0.4253, + "step": 24870 + }, + { + "epoch": 0.96, + "learning_rate": 3.408502180046166e-05, + "loss": 0.5377, + "step": 24871 + }, + { + "epoch": 0.96, + "learning_rate": 3.408438061041293e-05, + "loss": 0.4899, + "step": 24872 + }, + { + "epoch": 0.96, + "learning_rate": 3.40837394203642e-05, + "loss": 0.5091, + "step": 24873 + }, + { + "epoch": 0.96, + "learning_rate": 3.408309823031547e-05, + "loss": 0.5229, + "step": 24874 + }, + { + "epoch": 0.96, + "learning_rate": 3.4082457040266733e-05, + "loss": 0.4988, + "step": 24875 + }, + { + "epoch": 0.96, + "learning_rate": 3.4081815850218005e-05, + "loss": 0.4916, + "step": 24876 + }, + { + "epoch": 0.96, + "learning_rate": 3.408117466016928e-05, + "loss": 0.4748, + "step": 24877 + }, + { + "epoch": 0.96, + "learning_rate": 3.408053347012055e-05, + "loss": 0.4816, + "step": 24878 + }, + { + "epoch": 0.96, + "learning_rate": 3.4079892280071814e-05, + "loss": 0.5318, + "step": 24879 + }, + { + "epoch": 0.96, + "learning_rate": 3.4079251090023085e-05, + "loss": 0.6076, + "step": 24880 + }, + { + "epoch": 0.96, + "learning_rate": 3.407860989997436e-05, + "loss": 0.5355, + "step": 24881 + }, + { + "epoch": 0.96, + "learning_rate": 3.407796870992562e-05, + "loss": 0.4432, + "step": 24882 + }, + { + "epoch": 0.96, + "learning_rate": 3.4077327519876894e-05, + "loss": 0.4904, + "step": 24883 + }, + { + "epoch": 0.96, + "learning_rate": 3.407668632982816e-05, + "loss": 0.4446, + "step": 24884 + }, + { + "epoch": 0.96, + "learning_rate": 3.407604513977943e-05, + "loss": 0.4774, + "step": 24885 + }, + { + "epoch": 0.96, + "learning_rate": 3.40754039497307e-05, + "loss": 0.4351, + "step": 24886 + }, + { + "epoch": 0.96, + "learning_rate": 3.4074762759681974e-05, + "loss": 0.4785, + "step": 24887 + }, + { + "epoch": 0.96, + "learning_rate": 3.407412156963324e-05, + "loss": 0.5235, + "step": 24888 + }, + { + "epoch": 0.96, + "learning_rate": 3.407348037958451e-05, + "loss": 0.4221, + "step": 24889 + }, + { + "epoch": 0.96, + "learning_rate": 3.407283918953578e-05, + "loss": 0.4841, + "step": 24890 + }, + { + "epoch": 0.96, + "learning_rate": 3.407219799948705e-05, + "loss": 0.4717, + "step": 24891 + }, + { + "epoch": 0.96, + "learning_rate": 3.407155680943832e-05, + "loss": 0.5317, + "step": 24892 + }, + { + "epoch": 0.96, + "learning_rate": 3.407091561938959e-05, + "loss": 0.3658, + "step": 24893 + }, + { + "epoch": 0.96, + "learning_rate": 3.407027442934086e-05, + "loss": 0.3831, + "step": 24894 + }, + { + "epoch": 0.96, + "learning_rate": 3.406963323929213e-05, + "loss": 0.4758, + "step": 24895 + }, + { + "epoch": 0.96, + "learning_rate": 3.40689920492434e-05, + "loss": 0.5078, + "step": 24896 + }, + { + "epoch": 0.96, + "learning_rate": 3.4068350859194664e-05, + "loss": 0.4279, + "step": 24897 + }, + { + "epoch": 0.96, + "learning_rate": 3.4067709669145935e-05, + "loss": 0.3524, + "step": 24898 + }, + { + "epoch": 0.96, + "learning_rate": 3.406706847909721e-05, + "loss": 0.5066, + "step": 24899 + }, + { + "epoch": 0.96, + "learning_rate": 3.406642728904847e-05, + "loss": 0.4217, + "step": 24900 + }, + { + "epoch": 0.96, + "learning_rate": 3.4065786098999744e-05, + "loss": 0.4399, + "step": 24901 + }, + { + "epoch": 0.96, + "learning_rate": 3.4065144908951016e-05, + "loss": 0.4739, + "step": 24902 + }, + { + "epoch": 0.96, + "learning_rate": 3.406450371890229e-05, + "loss": 0.4578, + "step": 24903 + }, + { + "epoch": 0.96, + "learning_rate": 3.406386252885355e-05, + "loss": 0.4883, + "step": 24904 + }, + { + "epoch": 0.96, + "learning_rate": 3.4063221338804824e-05, + "loss": 0.6489, + "step": 24905 + }, + { + "epoch": 0.96, + "learning_rate": 3.406258014875609e-05, + "loss": 0.6378, + "step": 24906 + }, + { + "epoch": 0.96, + "learning_rate": 3.406193895870736e-05, + "loss": 0.3985, + "step": 24907 + }, + { + "epoch": 0.96, + "learning_rate": 3.406129776865863e-05, + "loss": 0.6133, + "step": 24908 + }, + { + "epoch": 0.96, + "learning_rate": 3.4060656578609904e-05, + "loss": 0.3887, + "step": 24909 + }, + { + "epoch": 0.96, + "learning_rate": 3.4060015388561176e-05, + "loss": 0.4367, + "step": 24910 + }, + { + "epoch": 0.96, + "learning_rate": 3.405937419851244e-05, + "loss": 0.4096, + "step": 24911 + }, + { + "epoch": 0.96, + "learning_rate": 3.405873300846371e-05, + "loss": 0.579, + "step": 24912 + }, + { + "epoch": 0.96, + "learning_rate": 3.405809181841498e-05, + "loss": 0.4869, + "step": 24913 + }, + { + "epoch": 0.96, + "learning_rate": 3.405745062836625e-05, + "loss": 0.5219, + "step": 24914 + }, + { + "epoch": 0.96, + "learning_rate": 3.4056809438317514e-05, + "loss": 0.4533, + "step": 24915 + }, + { + "epoch": 0.96, + "learning_rate": 3.4056168248268786e-05, + "loss": 0.5353, + "step": 24916 + }, + { + "epoch": 0.96, + "learning_rate": 3.405552705822006e-05, + "loss": 0.5514, + "step": 24917 + }, + { + "epoch": 0.96, + "learning_rate": 3.405488586817133e-05, + "loss": 0.5024, + "step": 24918 + }, + { + "epoch": 0.96, + "learning_rate": 3.40542446781226e-05, + "loss": 0.4791, + "step": 24919 + }, + { + "epoch": 0.96, + "learning_rate": 3.4053603488073866e-05, + "loss": 0.4254, + "step": 24920 + }, + { + "epoch": 0.96, + "learning_rate": 3.405296229802514e-05, + "loss": 0.5905, + "step": 24921 + }, + { + "epoch": 0.96, + "learning_rate": 3.40523211079764e-05, + "loss": 0.4331, + "step": 24922 + }, + { + "epoch": 0.96, + "learning_rate": 3.4051679917927674e-05, + "loss": 0.5721, + "step": 24923 + }, + { + "epoch": 0.96, + "learning_rate": 3.405103872787894e-05, + "loss": 0.5426, + "step": 24924 + }, + { + "epoch": 0.96, + "learning_rate": 3.405039753783022e-05, + "loss": 0.467, + "step": 24925 + }, + { + "epoch": 0.96, + "learning_rate": 3.404975634778149e-05, + "loss": 0.4859, + "step": 24926 + }, + { + "epoch": 0.96, + "learning_rate": 3.4049115157732754e-05, + "loss": 0.3754, + "step": 24927 + }, + { + "epoch": 0.96, + "learning_rate": 3.4048473967684026e-05, + "loss": 0.4811, + "step": 24928 + }, + { + "epoch": 0.96, + "learning_rate": 3.404783277763529e-05, + "loss": 0.5565, + "step": 24929 + }, + { + "epoch": 0.96, + "learning_rate": 3.404719158758656e-05, + "loss": 0.4996, + "step": 24930 + }, + { + "epoch": 0.96, + "learning_rate": 3.404655039753783e-05, + "loss": 0.6596, + "step": 24931 + }, + { + "epoch": 0.96, + "learning_rate": 3.40459092074891e-05, + "loss": 0.4449, + "step": 24932 + }, + { + "epoch": 0.96, + "learning_rate": 3.404526801744037e-05, + "loss": 0.4635, + "step": 24933 + }, + { + "epoch": 0.96, + "learning_rate": 3.404462682739164e-05, + "loss": 0.4983, + "step": 24934 + }, + { + "epoch": 0.96, + "learning_rate": 3.4043985637342914e-05, + "loss": 0.4672, + "step": 24935 + }, + { + "epoch": 0.96, + "learning_rate": 3.404334444729418e-05, + "loss": 0.3771, + "step": 24936 + }, + { + "epoch": 0.96, + "learning_rate": 3.404270325724545e-05, + "loss": 0.5187, + "step": 24937 + }, + { + "epoch": 0.96, + "learning_rate": 3.4042062067196716e-05, + "loss": 0.5169, + "step": 24938 + }, + { + "epoch": 0.96, + "learning_rate": 3.404142087714799e-05, + "loss": 0.573, + "step": 24939 + }, + { + "epoch": 0.96, + "learning_rate": 3.404077968709925e-05, + "loss": 0.5252, + "step": 24940 + }, + { + "epoch": 0.96, + "learning_rate": 3.4040138497050524e-05, + "loss": 0.5663, + "step": 24941 + }, + { + "epoch": 0.96, + "learning_rate": 3.40394973070018e-05, + "loss": 0.4174, + "step": 24942 + }, + { + "epoch": 0.96, + "learning_rate": 3.403885611695307e-05, + "loss": 0.4813, + "step": 24943 + }, + { + "epoch": 0.96, + "learning_rate": 3.403821492690434e-05, + "loss": 0.5663, + "step": 24944 + }, + { + "epoch": 0.96, + "learning_rate": 3.4037573736855605e-05, + "loss": 0.4704, + "step": 24945 + }, + { + "epoch": 0.96, + "learning_rate": 3.4036932546806876e-05, + "loss": 0.4283, + "step": 24946 + }, + { + "epoch": 0.96, + "learning_rate": 3.403629135675814e-05, + "loss": 0.4161, + "step": 24947 + }, + { + "epoch": 0.96, + "learning_rate": 3.403565016670941e-05, + "loss": 0.528, + "step": 24948 + }, + { + "epoch": 0.96, + "learning_rate": 3.4035008976660685e-05, + "loss": 0.4732, + "step": 24949 + }, + { + "epoch": 0.96, + "learning_rate": 3.4034367786611956e-05, + "loss": 0.5448, + "step": 24950 + }, + { + "epoch": 0.96, + "learning_rate": 3.403372659656323e-05, + "loss": 0.5107, + "step": 24951 + }, + { + "epoch": 0.96, + "learning_rate": 3.403308540651449e-05, + "loss": 0.4935, + "step": 24952 + }, + { + "epoch": 0.96, + "learning_rate": 3.4032444216465765e-05, + "loss": 0.3343, + "step": 24953 + }, + { + "epoch": 0.96, + "learning_rate": 3.403180302641703e-05, + "loss": 0.4351, + "step": 24954 + }, + { + "epoch": 0.96, + "learning_rate": 3.40311618363683e-05, + "loss": 0.4866, + "step": 24955 + }, + { + "epoch": 0.96, + "learning_rate": 3.4030520646319566e-05, + "loss": 0.6012, + "step": 24956 + }, + { + "epoch": 0.96, + "learning_rate": 3.402987945627084e-05, + "loss": 0.5237, + "step": 24957 + }, + { + "epoch": 0.96, + "learning_rate": 3.402923826622211e-05, + "loss": 0.4824, + "step": 24958 + }, + { + "epoch": 0.96, + "learning_rate": 3.402859707617338e-05, + "loss": 0.4353, + "step": 24959 + }, + { + "epoch": 0.96, + "learning_rate": 3.402795588612465e-05, + "loss": 0.4884, + "step": 24960 + }, + { + "epoch": 0.96, + "learning_rate": 3.402731469607592e-05, + "loss": 0.55, + "step": 24961 + }, + { + "epoch": 0.96, + "learning_rate": 3.402667350602719e-05, + "loss": 0.4605, + "step": 24962 + }, + { + "epoch": 0.96, + "learning_rate": 3.4026032315978455e-05, + "loss": 0.3968, + "step": 24963 + }, + { + "epoch": 0.96, + "learning_rate": 3.4025391125929726e-05, + "loss": 0.4092, + "step": 24964 + }, + { + "epoch": 0.96, + "learning_rate": 3.4024749935881e-05, + "loss": 0.4839, + "step": 24965 + }, + { + "epoch": 0.96, + "learning_rate": 3.402410874583227e-05, + "loss": 0.4948, + "step": 24966 + }, + { + "epoch": 0.96, + "learning_rate": 3.4023467555783535e-05, + "loss": 0.4517, + "step": 24967 + }, + { + "epoch": 0.96, + "learning_rate": 3.4022826365734807e-05, + "loss": 0.4542, + "step": 24968 + }, + { + "epoch": 0.96, + "learning_rate": 3.402218517568608e-05, + "loss": 0.5315, + "step": 24969 + }, + { + "epoch": 0.96, + "learning_rate": 3.402154398563734e-05, + "loss": 0.4513, + "step": 24970 + }, + { + "epoch": 0.96, + "learning_rate": 3.4020902795588615e-05, + "loss": 0.4757, + "step": 24971 + }, + { + "epoch": 0.96, + "learning_rate": 3.402026160553988e-05, + "loss": 0.5537, + "step": 24972 + }, + { + "epoch": 0.96, + "learning_rate": 3.401962041549115e-05, + "loss": 0.5417, + "step": 24973 + }, + { + "epoch": 0.96, + "learning_rate": 3.401897922544242e-05, + "loss": 0.5131, + "step": 24974 + }, + { + "epoch": 0.96, + "learning_rate": 3.4018338035393695e-05, + "loss": 0.4734, + "step": 24975 + }, + { + "epoch": 0.96, + "learning_rate": 3.401769684534496e-05, + "loss": 0.3937, + "step": 24976 + }, + { + "epoch": 0.96, + "learning_rate": 3.401705565529623e-05, + "loss": 0.4712, + "step": 24977 + }, + { + "epoch": 0.96, + "learning_rate": 3.4016414465247503e-05, + "loss": 0.3753, + "step": 24978 + }, + { + "epoch": 0.96, + "learning_rate": 3.401577327519877e-05, + "loss": 0.5621, + "step": 24979 + }, + { + "epoch": 0.96, + "learning_rate": 3.401513208515004e-05, + "loss": 0.3602, + "step": 24980 + }, + { + "epoch": 0.96, + "learning_rate": 3.401449089510131e-05, + "loss": 0.4332, + "step": 24981 + }, + { + "epoch": 0.96, + "learning_rate": 3.4013849705052584e-05, + "loss": 0.4206, + "step": 24982 + }, + { + "epoch": 0.96, + "learning_rate": 3.401320851500385e-05, + "loss": 0.6004, + "step": 24983 + }, + { + "epoch": 0.96, + "learning_rate": 3.401256732495512e-05, + "loss": 0.4468, + "step": 24984 + }, + { + "epoch": 0.96, + "learning_rate": 3.4011926134906385e-05, + "loss": 0.3626, + "step": 24985 + }, + { + "epoch": 0.96, + "learning_rate": 3.401128494485766e-05, + "loss": 0.4207, + "step": 24986 + }, + { + "epoch": 0.96, + "learning_rate": 3.401064375480893e-05, + "loss": 0.443, + "step": 24987 + }, + { + "epoch": 0.96, + "learning_rate": 3.4010002564760193e-05, + "loss": 0.5112, + "step": 24988 + }, + { + "epoch": 0.96, + "learning_rate": 3.4009361374711465e-05, + "loss": 0.4821, + "step": 24989 + }, + { + "epoch": 0.96, + "learning_rate": 3.400872018466274e-05, + "loss": 0.5111, + "step": 24990 + }, + { + "epoch": 0.96, + "learning_rate": 3.400807899461401e-05, + "loss": 0.4566, + "step": 24991 + }, + { + "epoch": 0.96, + "learning_rate": 3.4007437804565274e-05, + "loss": 0.4702, + "step": 24992 + }, + { + "epoch": 0.96, + "learning_rate": 3.4006796614516545e-05, + "loss": 0.6974, + "step": 24993 + }, + { + "epoch": 0.96, + "learning_rate": 3.400615542446781e-05, + "loss": 0.4408, + "step": 24994 + }, + { + "epoch": 0.96, + "learning_rate": 3.400551423441908e-05, + "loss": 0.6026, + "step": 24995 + }, + { + "epoch": 0.96, + "learning_rate": 3.4004873044370354e-05, + "loss": 0.4237, + "step": 24996 + }, + { + "epoch": 0.96, + "learning_rate": 3.4004231854321625e-05, + "loss": 0.4989, + "step": 24997 + }, + { + "epoch": 0.96, + "learning_rate": 3.40035906642729e-05, + "loss": 0.6686, + "step": 24998 + }, + { + "epoch": 0.96, + "learning_rate": 3.400294947422416e-05, + "loss": 0.4972, + "step": 24999 + }, + { + "epoch": 0.96, + "learning_rate": 3.4002308284175434e-05, + "loss": 0.4729, + "step": 25000 + }, + { + "epoch": 0.96, + "learning_rate": 3.40016670941267e-05, + "loss": 0.341, + "step": 25001 + }, + { + "epoch": 0.96, + "learning_rate": 3.400102590407797e-05, + "loss": 0.5067, + "step": 25002 + }, + { + "epoch": 0.96, + "learning_rate": 3.4000384714029235e-05, + "loss": 0.4749, + "step": 25003 + }, + { + "epoch": 0.96, + "learning_rate": 3.399974352398051e-05, + "loss": 0.4913, + "step": 25004 + }, + { + "epoch": 0.96, + "learning_rate": 3.399910233393178e-05, + "loss": 0.4161, + "step": 25005 + }, + { + "epoch": 0.96, + "learning_rate": 3.399846114388305e-05, + "loss": 0.5173, + "step": 25006 + }, + { + "epoch": 0.96, + "learning_rate": 3.399781995383432e-05, + "loss": 0.5367, + "step": 25007 + }, + { + "epoch": 0.96, + "learning_rate": 3.399717876378559e-05, + "loss": 0.5398, + "step": 25008 + }, + { + "epoch": 0.96, + "learning_rate": 3.399653757373686e-05, + "loss": 0.5417, + "step": 25009 + }, + { + "epoch": 0.96, + "learning_rate": 3.3995896383688124e-05, + "loss": 0.4994, + "step": 25010 + }, + { + "epoch": 0.96, + "learning_rate": 3.3995255193639396e-05, + "loss": 0.5275, + "step": 25011 + }, + { + "epoch": 0.96, + "learning_rate": 3.399461400359066e-05, + "loss": 0.5698, + "step": 25012 + }, + { + "epoch": 0.96, + "learning_rate": 3.399397281354193e-05, + "loss": 0.4471, + "step": 25013 + }, + { + "epoch": 0.96, + "learning_rate": 3.399333162349321e-05, + "loss": 0.4116, + "step": 25014 + }, + { + "epoch": 0.96, + "learning_rate": 3.3992690433444476e-05, + "loss": 0.4521, + "step": 25015 + }, + { + "epoch": 0.96, + "learning_rate": 3.399204924339575e-05, + "loss": 0.4426, + "step": 25016 + }, + { + "epoch": 0.96, + "learning_rate": 3.399140805334701e-05, + "loss": 0.3506, + "step": 25017 + }, + { + "epoch": 0.96, + "learning_rate": 3.3990766863298284e-05, + "loss": 0.4788, + "step": 25018 + }, + { + "epoch": 0.96, + "learning_rate": 3.399012567324955e-05, + "loss": 0.4067, + "step": 25019 + }, + { + "epoch": 0.96, + "learning_rate": 3.398948448320082e-05, + "loss": 0.4455, + "step": 25020 + }, + { + "epoch": 0.96, + "learning_rate": 3.398884329315209e-05, + "loss": 0.4711, + "step": 25021 + }, + { + "epoch": 0.96, + "learning_rate": 3.3988202103103364e-05, + "loss": 0.465, + "step": 25022 + }, + { + "epoch": 0.96, + "learning_rate": 3.3987560913054636e-05, + "loss": 0.5185, + "step": 25023 + }, + { + "epoch": 0.96, + "learning_rate": 3.39869197230059e-05, + "loss": 0.5597, + "step": 25024 + }, + { + "epoch": 0.96, + "learning_rate": 3.398627853295717e-05, + "loss": 0.4618, + "step": 25025 + }, + { + "epoch": 0.96, + "learning_rate": 3.398563734290844e-05, + "loss": 0.5254, + "step": 25026 + }, + { + "epoch": 0.96, + "learning_rate": 3.398499615285971e-05, + "loss": 0.5907, + "step": 25027 + }, + { + "epoch": 0.96, + "learning_rate": 3.3984354962810974e-05, + "loss": 0.4816, + "step": 25028 + }, + { + "epoch": 0.96, + "learning_rate": 3.3983713772762246e-05, + "loss": 0.4423, + "step": 25029 + }, + { + "epoch": 0.96, + "learning_rate": 3.398307258271352e-05, + "loss": 0.4786, + "step": 25030 + }, + { + "epoch": 0.96, + "learning_rate": 3.398243139266479e-05, + "loss": 0.5156, + "step": 25031 + }, + { + "epoch": 0.96, + "learning_rate": 3.398179020261606e-05, + "loss": 0.4258, + "step": 25032 + }, + { + "epoch": 0.96, + "learning_rate": 3.3981149012567326e-05, + "loss": 0.4723, + "step": 25033 + }, + { + "epoch": 0.96, + "learning_rate": 3.39805078225186e-05, + "loss": 0.6233, + "step": 25034 + }, + { + "epoch": 0.96, + "learning_rate": 3.397986663246986e-05, + "loss": 0.5879, + "step": 25035 + }, + { + "epoch": 0.96, + "learning_rate": 3.3979225442421134e-05, + "loss": 0.5249, + "step": 25036 + }, + { + "epoch": 0.96, + "learning_rate": 3.3978584252372406e-05, + "loss": 0.424, + "step": 25037 + }, + { + "epoch": 0.96, + "learning_rate": 3.397794306232368e-05, + "loss": 0.461, + "step": 25038 + }, + { + "epoch": 0.96, + "learning_rate": 3.397730187227494e-05, + "loss": 0.5147, + "step": 25039 + }, + { + "epoch": 0.96, + "learning_rate": 3.3976660682226214e-05, + "loss": 0.4194, + "step": 25040 + }, + { + "epoch": 0.96, + "learning_rate": 3.3976019492177486e-05, + "loss": 0.4621, + "step": 25041 + }, + { + "epoch": 0.96, + "learning_rate": 3.397537830212875e-05, + "loss": 0.4983, + "step": 25042 + }, + { + "epoch": 0.96, + "learning_rate": 3.397473711208002e-05, + "loss": 0.3217, + "step": 25043 + }, + { + "epoch": 0.96, + "learning_rate": 3.397409592203129e-05, + "loss": 0.4009, + "step": 25044 + }, + { + "epoch": 0.96, + "learning_rate": 3.397345473198256e-05, + "loss": 0.397, + "step": 25045 + }, + { + "epoch": 0.96, + "learning_rate": 3.397281354193383e-05, + "loss": 0.3175, + "step": 25046 + }, + { + "epoch": 0.96, + "learning_rate": 3.39721723518851e-05, + "loss": 0.3941, + "step": 25047 + }, + { + "epoch": 0.96, + "learning_rate": 3.397153116183637e-05, + "loss": 0.6058, + "step": 25048 + }, + { + "epoch": 0.96, + "learning_rate": 3.397088997178764e-05, + "loss": 0.5934, + "step": 25049 + }, + { + "epoch": 0.96, + "learning_rate": 3.397024878173891e-05, + "loss": 0.4645, + "step": 25050 + }, + { + "epoch": 0.96, + "learning_rate": 3.3969607591690176e-05, + "loss": 0.5373, + "step": 25051 + }, + { + "epoch": 0.96, + "learning_rate": 3.396896640164145e-05, + "loss": 0.4203, + "step": 25052 + }, + { + "epoch": 0.96, + "learning_rate": 3.396832521159272e-05, + "loss": 0.5636, + "step": 25053 + }, + { + "epoch": 0.96, + "learning_rate": 3.396768402154399e-05, + "loss": 0.5464, + "step": 25054 + }, + { + "epoch": 0.96, + "learning_rate": 3.3967042831495256e-05, + "loss": 0.3711, + "step": 25055 + }, + { + "epoch": 0.96, + "learning_rate": 3.396640164144653e-05, + "loss": 0.3925, + "step": 25056 + }, + { + "epoch": 0.96, + "learning_rate": 3.396576045139779e-05, + "loss": 0.4047, + "step": 25057 + }, + { + "epoch": 0.96, + "learning_rate": 3.3965119261349065e-05, + "loss": 0.5251, + "step": 25058 + }, + { + "epoch": 0.96, + "learning_rate": 3.3964478071300336e-05, + "loss": 0.453, + "step": 25059 + }, + { + "epoch": 0.96, + "learning_rate": 3.39638368812516e-05, + "loss": 0.4062, + "step": 25060 + }, + { + "epoch": 0.96, + "learning_rate": 3.396319569120287e-05, + "loss": 0.5281, + "step": 25061 + }, + { + "epoch": 0.96, + "learning_rate": 3.3962554501154145e-05, + "loss": 0.4608, + "step": 25062 + }, + { + "epoch": 0.96, + "learning_rate": 3.3961913311105416e-05, + "loss": 0.3645, + "step": 25063 + }, + { + "epoch": 0.96, + "learning_rate": 3.396127212105668e-05, + "loss": 0.4884, + "step": 25064 + }, + { + "epoch": 0.96, + "learning_rate": 3.396063093100795e-05, + "loss": 0.6272, + "step": 25065 + }, + { + "epoch": 0.96, + "learning_rate": 3.395998974095922e-05, + "loss": 0.4994, + "step": 25066 + }, + { + "epoch": 0.96, + "learning_rate": 3.395934855091049e-05, + "loss": 0.6017, + "step": 25067 + }, + { + "epoch": 0.96, + "learning_rate": 3.395870736086176e-05, + "loss": 0.4551, + "step": 25068 + }, + { + "epoch": 0.96, + "learning_rate": 3.395806617081303e-05, + "loss": 0.3934, + "step": 25069 + }, + { + "epoch": 0.96, + "learning_rate": 3.3957424980764305e-05, + "loss": 0.412, + "step": 25070 + }, + { + "epoch": 0.96, + "learning_rate": 3.395678379071557e-05, + "loss": 0.4253, + "step": 25071 + }, + { + "epoch": 0.96, + "learning_rate": 3.395614260066684e-05, + "loss": 0.4761, + "step": 25072 + }, + { + "epoch": 0.96, + "learning_rate": 3.3955501410618106e-05, + "loss": 0.5285, + "step": 25073 + }, + { + "epoch": 0.96, + "learning_rate": 3.395486022056938e-05, + "loss": 0.4543, + "step": 25074 + }, + { + "epoch": 0.96, + "learning_rate": 3.395421903052064e-05, + "loss": 0.5602, + "step": 25075 + }, + { + "epoch": 0.96, + "learning_rate": 3.3953577840471915e-05, + "loss": 0.6188, + "step": 25076 + }, + { + "epoch": 0.96, + "learning_rate": 3.3952936650423187e-05, + "loss": 0.3898, + "step": 25077 + }, + { + "epoch": 0.96, + "learning_rate": 3.395229546037446e-05, + "loss": 0.5273, + "step": 25078 + }, + { + "epoch": 0.96, + "learning_rate": 3.395165427032573e-05, + "loss": 0.4452, + "step": 25079 + }, + { + "epoch": 0.96, + "learning_rate": 3.3951013080276995e-05, + "loss": 0.4911, + "step": 25080 + }, + { + "epoch": 0.96, + "learning_rate": 3.395037189022827e-05, + "loss": 0.4738, + "step": 25081 + }, + { + "epoch": 0.96, + "learning_rate": 3.394973070017953e-05, + "loss": 0.3935, + "step": 25082 + }, + { + "epoch": 0.96, + "learning_rate": 3.39490895101308e-05, + "loss": 0.3948, + "step": 25083 + }, + { + "epoch": 0.96, + "learning_rate": 3.394844832008207e-05, + "loss": 0.3197, + "step": 25084 + }, + { + "epoch": 0.96, + "learning_rate": 3.394780713003335e-05, + "loss": 0.5157, + "step": 25085 + }, + { + "epoch": 0.96, + "learning_rate": 3.394716593998462e-05, + "loss": 0.4471, + "step": 25086 + }, + { + "epoch": 0.96, + "learning_rate": 3.394652474993588e-05, + "loss": 0.4542, + "step": 25087 + }, + { + "epoch": 0.96, + "learning_rate": 3.3945883559887155e-05, + "loss": 0.6041, + "step": 25088 + }, + { + "epoch": 0.96, + "learning_rate": 3.394524236983842e-05, + "loss": 0.4037, + "step": 25089 + }, + { + "epoch": 0.96, + "learning_rate": 3.394460117978969e-05, + "loss": 0.489, + "step": 25090 + }, + { + "epoch": 0.96, + "learning_rate": 3.394395998974096e-05, + "loss": 0.3736, + "step": 25091 + }, + { + "epoch": 0.96, + "learning_rate": 3.394331879969223e-05, + "loss": 0.4479, + "step": 25092 + }, + { + "epoch": 0.96, + "learning_rate": 3.39426776096435e-05, + "loss": 0.5221, + "step": 25093 + }, + { + "epoch": 0.96, + "learning_rate": 3.394203641959477e-05, + "loss": 0.3698, + "step": 25094 + }, + { + "epoch": 0.96, + "learning_rate": 3.3941395229546044e-05, + "loss": 0.4617, + "step": 25095 + }, + { + "epoch": 0.96, + "learning_rate": 3.394075403949731e-05, + "loss": 0.5438, + "step": 25096 + }, + { + "epoch": 0.96, + "learning_rate": 3.394011284944858e-05, + "loss": 0.4206, + "step": 25097 + }, + { + "epoch": 0.96, + "learning_rate": 3.3939471659399845e-05, + "loss": 0.577, + "step": 25098 + }, + { + "epoch": 0.96, + "learning_rate": 3.393883046935112e-05, + "loss": 0.5619, + "step": 25099 + }, + { + "epoch": 0.97, + "learning_rate": 3.393818927930238e-05, + "loss": 0.3596, + "step": 25100 + }, + { + "epoch": 0.97, + "learning_rate": 3.3937548089253654e-05, + "loss": 0.5355, + "step": 25101 + }, + { + "epoch": 0.97, + "learning_rate": 3.3936906899204925e-05, + "loss": 0.5098, + "step": 25102 + }, + { + "epoch": 0.97, + "learning_rate": 3.39362657091562e-05, + "loss": 0.3744, + "step": 25103 + }, + { + "epoch": 0.97, + "learning_rate": 3.393562451910747e-05, + "loss": 0.6627, + "step": 25104 + }, + { + "epoch": 0.97, + "learning_rate": 3.3934983329058734e-05, + "loss": 0.6093, + "step": 25105 + }, + { + "epoch": 0.97, + "learning_rate": 3.3934342139010005e-05, + "loss": 0.479, + "step": 25106 + }, + { + "epoch": 0.97, + "learning_rate": 3.393370094896127e-05, + "loss": 0.4484, + "step": 25107 + }, + { + "epoch": 0.97, + "learning_rate": 3.393305975891254e-05, + "loss": 0.4068, + "step": 25108 + }, + { + "epoch": 0.97, + "learning_rate": 3.3932418568863814e-05, + "loss": 0.6071, + "step": 25109 + }, + { + "epoch": 0.97, + "learning_rate": 3.3931777378815085e-05, + "loss": 0.5145, + "step": 25110 + }, + { + "epoch": 0.97, + "learning_rate": 3.393113618876635e-05, + "loss": 0.4393, + "step": 25111 + }, + { + "epoch": 0.97, + "learning_rate": 3.393049499871762e-05, + "loss": 0.4531, + "step": 25112 + }, + { + "epoch": 0.97, + "learning_rate": 3.3929853808668894e-05, + "loss": 0.4916, + "step": 25113 + }, + { + "epoch": 0.97, + "learning_rate": 3.392921261862016e-05, + "loss": 0.5335, + "step": 25114 + }, + { + "epoch": 0.97, + "learning_rate": 3.392857142857143e-05, + "loss": 0.4386, + "step": 25115 + }, + { + "epoch": 0.97, + "learning_rate": 3.3927930238522695e-05, + "loss": 0.5333, + "step": 25116 + }, + { + "epoch": 0.97, + "learning_rate": 3.392728904847397e-05, + "loss": 0.5702, + "step": 25117 + }, + { + "epoch": 0.97, + "learning_rate": 3.392664785842524e-05, + "loss": 0.4474, + "step": 25118 + }, + { + "epoch": 0.97, + "learning_rate": 3.392600666837651e-05, + "loss": 0.4629, + "step": 25119 + }, + { + "epoch": 0.97, + "learning_rate": 3.3925365478327775e-05, + "loss": 0.4069, + "step": 25120 + }, + { + "epoch": 0.97, + "learning_rate": 3.392472428827905e-05, + "loss": 0.444, + "step": 25121 + }, + { + "epoch": 0.97, + "learning_rate": 3.392408309823032e-05, + "loss": 0.5803, + "step": 25122 + }, + { + "epoch": 0.97, + "learning_rate": 3.3923441908181584e-05, + "loss": 0.5108, + "step": 25123 + }, + { + "epoch": 0.97, + "learning_rate": 3.3922800718132856e-05, + "loss": 0.5384, + "step": 25124 + }, + { + "epoch": 0.97, + "learning_rate": 3.392215952808413e-05, + "loss": 0.5259, + "step": 25125 + }, + { + "epoch": 0.97, + "learning_rate": 3.39215183380354e-05, + "loss": 0.4213, + "step": 25126 + }, + { + "epoch": 0.97, + "learning_rate": 3.3920877147986664e-05, + "loss": 0.4011, + "step": 25127 + }, + { + "epoch": 0.97, + "learning_rate": 3.3920235957937936e-05, + "loss": 0.3706, + "step": 25128 + }, + { + "epoch": 0.97, + "learning_rate": 3.391959476788921e-05, + "loss": 0.3903, + "step": 25129 + }, + { + "epoch": 0.97, + "learning_rate": 3.391895357784047e-05, + "loss": 0.4825, + "step": 25130 + }, + { + "epoch": 0.97, + "learning_rate": 3.3918312387791744e-05, + "loss": 0.4356, + "step": 25131 + }, + { + "epoch": 0.97, + "learning_rate": 3.391767119774301e-05, + "loss": 0.354, + "step": 25132 + }, + { + "epoch": 0.97, + "learning_rate": 3.391703000769428e-05, + "loss": 0.6034, + "step": 25133 + }, + { + "epoch": 0.97, + "learning_rate": 3.391638881764555e-05, + "loss": 0.4397, + "step": 25134 + }, + { + "epoch": 0.97, + "learning_rate": 3.3915747627596824e-05, + "loss": 0.447, + "step": 25135 + }, + { + "epoch": 0.97, + "learning_rate": 3.391510643754809e-05, + "loss": 0.4223, + "step": 25136 + }, + { + "epoch": 0.97, + "learning_rate": 3.391446524749936e-05, + "loss": 0.5868, + "step": 25137 + }, + { + "epoch": 0.97, + "learning_rate": 3.391382405745063e-05, + "loss": 0.4472, + "step": 25138 + }, + { + "epoch": 0.97, + "learning_rate": 3.39131828674019e-05, + "loss": 0.5101, + "step": 25139 + }, + { + "epoch": 0.97, + "learning_rate": 3.391254167735317e-05, + "loss": 0.5895, + "step": 25140 + }, + { + "epoch": 0.97, + "learning_rate": 3.391190048730444e-05, + "loss": 0.4721, + "step": 25141 + }, + { + "epoch": 0.97, + "learning_rate": 3.391125929725571e-05, + "loss": 0.4202, + "step": 25142 + }, + { + "epoch": 0.97, + "learning_rate": 3.391061810720698e-05, + "loss": 0.6033, + "step": 25143 + }, + { + "epoch": 0.97, + "learning_rate": 3.390997691715825e-05, + "loss": 0.4493, + "step": 25144 + }, + { + "epoch": 0.97, + "learning_rate": 3.3909335727109514e-05, + "loss": 0.5889, + "step": 25145 + }, + { + "epoch": 0.97, + "learning_rate": 3.3908694537060786e-05, + "loss": 0.5457, + "step": 25146 + }, + { + "epoch": 0.97, + "learning_rate": 3.390805334701206e-05, + "loss": 0.4917, + "step": 25147 + }, + { + "epoch": 0.97, + "learning_rate": 3.390741215696332e-05, + "loss": 0.3853, + "step": 25148 + }, + { + "epoch": 0.97, + "learning_rate": 3.3906770966914594e-05, + "loss": 0.4876, + "step": 25149 + }, + { + "epoch": 0.97, + "learning_rate": 3.3906129776865866e-05, + "loss": 0.455, + "step": 25150 + }, + { + "epoch": 0.97, + "learning_rate": 3.390548858681714e-05, + "loss": 0.4282, + "step": 25151 + }, + { + "epoch": 0.97, + "learning_rate": 3.39048473967684e-05, + "loss": 0.4676, + "step": 25152 + }, + { + "epoch": 0.97, + "learning_rate": 3.3904206206719674e-05, + "loss": 0.5479, + "step": 25153 + }, + { + "epoch": 0.97, + "learning_rate": 3.390356501667094e-05, + "loss": 0.4878, + "step": 25154 + }, + { + "epoch": 0.97, + "learning_rate": 3.390292382662221e-05, + "loss": 0.4747, + "step": 25155 + }, + { + "epoch": 0.97, + "learning_rate": 3.390228263657348e-05, + "loss": 0.4923, + "step": 25156 + }, + { + "epoch": 0.97, + "learning_rate": 3.3901641446524754e-05, + "loss": 0.5554, + "step": 25157 + }, + { + "epoch": 0.97, + "learning_rate": 3.3901000256476026e-05, + "loss": 0.493, + "step": 25158 + }, + { + "epoch": 0.97, + "learning_rate": 3.390035906642729e-05, + "loss": 0.4216, + "step": 25159 + }, + { + "epoch": 0.97, + "learning_rate": 3.389971787637856e-05, + "loss": 0.4257, + "step": 25160 + }, + { + "epoch": 0.97, + "learning_rate": 3.389907668632983e-05, + "loss": 0.544, + "step": 25161 + }, + { + "epoch": 0.97, + "learning_rate": 3.38984354962811e-05, + "loss": 0.4357, + "step": 25162 + }, + { + "epoch": 0.97, + "learning_rate": 3.3897794306232364e-05, + "loss": 0.5388, + "step": 25163 + }, + { + "epoch": 0.97, + "learning_rate": 3.3897153116183636e-05, + "loss": 0.4729, + "step": 25164 + }, + { + "epoch": 0.97, + "learning_rate": 3.389651192613491e-05, + "loss": 0.4985, + "step": 25165 + }, + { + "epoch": 0.97, + "learning_rate": 3.389587073608618e-05, + "loss": 0.4229, + "step": 25166 + }, + { + "epoch": 0.97, + "learning_rate": 3.389522954603745e-05, + "loss": 0.6303, + "step": 25167 + }, + { + "epoch": 0.97, + "learning_rate": 3.3894588355988716e-05, + "loss": 0.5698, + "step": 25168 + }, + { + "epoch": 0.97, + "learning_rate": 3.389394716593999e-05, + "loss": 0.4908, + "step": 25169 + }, + { + "epoch": 0.97, + "learning_rate": 3.389330597589125e-05, + "loss": 0.4682, + "step": 25170 + }, + { + "epoch": 0.97, + "learning_rate": 3.3892664785842525e-05, + "loss": 0.5323, + "step": 25171 + }, + { + "epoch": 0.97, + "learning_rate": 3.389202359579379e-05, + "loss": 0.6348, + "step": 25172 + }, + { + "epoch": 0.97, + "learning_rate": 3.389138240574506e-05, + "loss": 0.5301, + "step": 25173 + }, + { + "epoch": 0.97, + "learning_rate": 3.389074121569634e-05, + "loss": 0.4442, + "step": 25174 + }, + { + "epoch": 0.97, + "learning_rate": 3.3890100025647605e-05, + "loss": 0.5024, + "step": 25175 + }, + { + "epoch": 0.97, + "learning_rate": 3.3889458835598876e-05, + "loss": 0.5249, + "step": 25176 + }, + { + "epoch": 0.97, + "learning_rate": 3.388881764555014e-05, + "loss": 0.5661, + "step": 25177 + }, + { + "epoch": 0.97, + "learning_rate": 3.388817645550141e-05, + "loss": 0.5405, + "step": 25178 + }, + { + "epoch": 0.97, + "learning_rate": 3.388753526545268e-05, + "loss": 0.441, + "step": 25179 + }, + { + "epoch": 0.97, + "learning_rate": 3.388689407540395e-05, + "loss": 0.4705, + "step": 25180 + }, + { + "epoch": 0.97, + "learning_rate": 3.388625288535522e-05, + "loss": 0.5021, + "step": 25181 + }, + { + "epoch": 0.97, + "learning_rate": 3.388561169530649e-05, + "loss": 0.4297, + "step": 25182 + }, + { + "epoch": 0.97, + "learning_rate": 3.3884970505257765e-05, + "loss": 0.5741, + "step": 25183 + }, + { + "epoch": 0.97, + "learning_rate": 3.388432931520903e-05, + "loss": 0.5745, + "step": 25184 + }, + { + "epoch": 0.97, + "learning_rate": 3.38836881251603e-05, + "loss": 0.4594, + "step": 25185 + }, + { + "epoch": 0.97, + "learning_rate": 3.3883046935111566e-05, + "loss": 0.4807, + "step": 25186 + }, + { + "epoch": 0.97, + "learning_rate": 3.388240574506284e-05, + "loss": 0.5119, + "step": 25187 + }, + { + "epoch": 0.97, + "learning_rate": 3.38817645550141e-05, + "loss": 0.4815, + "step": 25188 + }, + { + "epoch": 0.97, + "learning_rate": 3.3881123364965375e-05, + "loss": 0.4149, + "step": 25189 + }, + { + "epoch": 0.97, + "learning_rate": 3.3880482174916647e-05, + "loss": 0.3345, + "step": 25190 + }, + { + "epoch": 0.97, + "learning_rate": 3.387984098486792e-05, + "loss": 0.4294, + "step": 25191 + }, + { + "epoch": 0.97, + "learning_rate": 3.387919979481919e-05, + "loss": 0.5977, + "step": 25192 + }, + { + "epoch": 0.97, + "learning_rate": 3.3878558604770455e-05, + "loss": 0.4624, + "step": 25193 + }, + { + "epoch": 0.97, + "learning_rate": 3.387791741472173e-05, + "loss": 0.5103, + "step": 25194 + }, + { + "epoch": 0.97, + "learning_rate": 3.387727622467299e-05, + "loss": 0.3962, + "step": 25195 + }, + { + "epoch": 0.97, + "learning_rate": 3.387663503462426e-05, + "loss": 0.5552, + "step": 25196 + }, + { + "epoch": 0.97, + "learning_rate": 3.3875993844575535e-05, + "loss": 0.4405, + "step": 25197 + }, + { + "epoch": 0.97, + "learning_rate": 3.387535265452681e-05, + "loss": 0.477, + "step": 25198 + }, + { + "epoch": 0.97, + "learning_rate": 3.387471146447807e-05, + "loss": 0.5457, + "step": 25199 + }, + { + "epoch": 0.97, + "learning_rate": 3.3874070274429343e-05, + "loss": 0.4631, + "step": 25200 + }, + { + "epoch": 0.97, + "learning_rate": 3.3873429084380615e-05, + "loss": 0.5287, + "step": 25201 + }, + { + "epoch": 0.97, + "learning_rate": 3.387278789433188e-05, + "loss": 0.4833, + "step": 25202 + }, + { + "epoch": 0.97, + "learning_rate": 3.387214670428315e-05, + "loss": 0.4127, + "step": 25203 + }, + { + "epoch": 0.97, + "learning_rate": 3.387150551423442e-05, + "loss": 0.536, + "step": 25204 + }, + { + "epoch": 0.97, + "learning_rate": 3.387086432418569e-05, + "loss": 0.5542, + "step": 25205 + }, + { + "epoch": 0.97, + "learning_rate": 3.387022313413696e-05, + "loss": 0.4861, + "step": 25206 + }, + { + "epoch": 0.97, + "learning_rate": 3.386958194408823e-05, + "loss": 0.5243, + "step": 25207 + }, + { + "epoch": 0.97, + "learning_rate": 3.38689407540395e-05, + "loss": 0.4522, + "step": 25208 + }, + { + "epoch": 0.97, + "learning_rate": 3.386829956399077e-05, + "loss": 0.5066, + "step": 25209 + }, + { + "epoch": 0.97, + "learning_rate": 3.386765837394204e-05, + "loss": 0.5254, + "step": 25210 + }, + { + "epoch": 0.97, + "learning_rate": 3.3867017183893305e-05, + "loss": 0.5026, + "step": 25211 + }, + { + "epoch": 0.97, + "learning_rate": 3.386637599384458e-05, + "loss": 0.4536, + "step": 25212 + }, + { + "epoch": 0.97, + "learning_rate": 3.386573480379585e-05, + "loss": 0.456, + "step": 25213 + }, + { + "epoch": 0.97, + "learning_rate": 3.386509361374712e-05, + "loss": 0.4539, + "step": 25214 + }, + { + "epoch": 0.97, + "learning_rate": 3.3864452423698385e-05, + "loss": 0.5406, + "step": 25215 + }, + { + "epoch": 0.97, + "learning_rate": 3.386381123364966e-05, + "loss": 0.5087, + "step": 25216 + }, + { + "epoch": 0.97, + "learning_rate": 3.386317004360092e-05, + "loss": 0.4285, + "step": 25217 + }, + { + "epoch": 0.97, + "learning_rate": 3.3862528853552194e-05, + "loss": 0.4417, + "step": 25218 + }, + { + "epoch": 0.97, + "learning_rate": 3.3861887663503465e-05, + "loss": 0.5412, + "step": 25219 + }, + { + "epoch": 0.97, + "learning_rate": 3.386124647345473e-05, + "loss": 0.5718, + "step": 25220 + }, + { + "epoch": 0.97, + "learning_rate": 3.3860605283406e-05, + "loss": 0.4231, + "step": 25221 + }, + { + "epoch": 0.97, + "learning_rate": 3.3859964093357274e-05, + "loss": 0.5644, + "step": 25222 + }, + { + "epoch": 0.97, + "learning_rate": 3.3859322903308545e-05, + "loss": 0.4926, + "step": 25223 + }, + { + "epoch": 0.97, + "learning_rate": 3.385868171325981e-05, + "loss": 0.4772, + "step": 25224 + }, + { + "epoch": 0.97, + "learning_rate": 3.385804052321108e-05, + "loss": 0.5183, + "step": 25225 + }, + { + "epoch": 0.97, + "learning_rate": 3.385739933316235e-05, + "loss": 0.497, + "step": 25226 + }, + { + "epoch": 0.97, + "learning_rate": 3.385675814311362e-05, + "loss": 0.5325, + "step": 25227 + }, + { + "epoch": 0.97, + "learning_rate": 3.385611695306489e-05, + "loss": 0.4842, + "step": 25228 + }, + { + "epoch": 0.97, + "learning_rate": 3.385547576301616e-05, + "loss": 0.4243, + "step": 25229 + }, + { + "epoch": 0.97, + "learning_rate": 3.3854834572967434e-05, + "loss": 0.5239, + "step": 25230 + }, + { + "epoch": 0.97, + "learning_rate": 3.38541933829187e-05, + "loss": 0.4742, + "step": 25231 + }, + { + "epoch": 0.97, + "learning_rate": 3.385355219286997e-05, + "loss": 0.3713, + "step": 25232 + }, + { + "epoch": 0.97, + "learning_rate": 3.3852911002821236e-05, + "loss": 0.4263, + "step": 25233 + }, + { + "epoch": 0.97, + "learning_rate": 3.385226981277251e-05, + "loss": 0.4609, + "step": 25234 + }, + { + "epoch": 0.97, + "learning_rate": 3.385162862272377e-05, + "loss": 0.4362, + "step": 25235 + }, + { + "epoch": 0.97, + "learning_rate": 3.3850987432675044e-05, + "loss": 0.4565, + "step": 25236 + }, + { + "epoch": 0.97, + "learning_rate": 3.3850346242626316e-05, + "loss": 0.584, + "step": 25237 + }, + { + "epoch": 0.97, + "learning_rate": 3.384970505257759e-05, + "loss": 0.5129, + "step": 25238 + }, + { + "epoch": 0.97, + "learning_rate": 3.384906386252886e-05, + "loss": 0.5691, + "step": 25239 + }, + { + "epoch": 0.97, + "learning_rate": 3.3848422672480124e-05, + "loss": 0.4681, + "step": 25240 + }, + { + "epoch": 0.97, + "learning_rate": 3.3847781482431396e-05, + "loss": 0.419, + "step": 25241 + }, + { + "epoch": 0.97, + "learning_rate": 3.384714029238266e-05, + "loss": 0.5552, + "step": 25242 + }, + { + "epoch": 0.97, + "learning_rate": 3.384649910233393e-05, + "loss": 0.5188, + "step": 25243 + }, + { + "epoch": 0.97, + "learning_rate": 3.38458579122852e-05, + "loss": 0.4688, + "step": 25244 + }, + { + "epoch": 0.97, + "learning_rate": 3.3845216722236476e-05, + "loss": 0.4721, + "step": 25245 + }, + { + "epoch": 0.97, + "learning_rate": 3.384457553218775e-05, + "loss": 0.4803, + "step": 25246 + }, + { + "epoch": 0.97, + "learning_rate": 3.384393434213901e-05, + "loss": 0.4932, + "step": 25247 + }, + { + "epoch": 0.97, + "learning_rate": 3.3843293152090284e-05, + "loss": 0.4985, + "step": 25248 + }, + { + "epoch": 0.97, + "learning_rate": 3.384265196204155e-05, + "loss": 0.5492, + "step": 25249 + }, + { + "epoch": 0.97, + "learning_rate": 3.384201077199282e-05, + "loss": 0.3906, + "step": 25250 + }, + { + "epoch": 0.97, + "learning_rate": 3.3841369581944086e-05, + "loss": 0.385, + "step": 25251 + }, + { + "epoch": 0.97, + "learning_rate": 3.384072839189536e-05, + "loss": 0.4595, + "step": 25252 + }, + { + "epoch": 0.97, + "learning_rate": 3.384008720184663e-05, + "loss": 0.3967, + "step": 25253 + }, + { + "epoch": 0.97, + "learning_rate": 3.38394460117979e-05, + "loss": 0.505, + "step": 25254 + }, + { + "epoch": 0.97, + "learning_rate": 3.383880482174917e-05, + "loss": 0.4904, + "step": 25255 + }, + { + "epoch": 0.97, + "learning_rate": 3.383816363170044e-05, + "loss": 0.5358, + "step": 25256 + }, + { + "epoch": 0.97, + "learning_rate": 3.383752244165171e-05, + "loss": 0.4307, + "step": 25257 + }, + { + "epoch": 0.97, + "learning_rate": 3.3836881251602974e-05, + "loss": 0.4757, + "step": 25258 + }, + { + "epoch": 0.97, + "learning_rate": 3.3836240061554246e-05, + "loss": 0.5961, + "step": 25259 + }, + { + "epoch": 0.97, + "learning_rate": 3.383559887150551e-05, + "loss": 0.4477, + "step": 25260 + }, + { + "epoch": 0.97, + "learning_rate": 3.383495768145678e-05, + "loss": 0.457, + "step": 25261 + }, + { + "epoch": 0.97, + "learning_rate": 3.3834316491408054e-05, + "loss": 0.4283, + "step": 25262 + }, + { + "epoch": 0.97, + "learning_rate": 3.3833675301359326e-05, + "loss": 0.3943, + "step": 25263 + }, + { + "epoch": 0.97, + "learning_rate": 3.38330341113106e-05, + "loss": 0.4964, + "step": 25264 + }, + { + "epoch": 0.97, + "learning_rate": 3.383239292126186e-05, + "loss": 0.4783, + "step": 25265 + }, + { + "epoch": 0.97, + "learning_rate": 3.3831751731213134e-05, + "loss": 0.3813, + "step": 25266 + }, + { + "epoch": 0.97, + "learning_rate": 3.38311105411644e-05, + "loss": 0.356, + "step": 25267 + }, + { + "epoch": 0.97, + "learning_rate": 3.383046935111567e-05, + "loss": 0.6152, + "step": 25268 + }, + { + "epoch": 0.97, + "learning_rate": 3.382982816106694e-05, + "loss": 0.5119, + "step": 25269 + }, + { + "epoch": 0.97, + "learning_rate": 3.3829186971018214e-05, + "loss": 0.4837, + "step": 25270 + }, + { + "epoch": 0.97, + "learning_rate": 3.382854578096948e-05, + "loss": 0.4434, + "step": 25271 + }, + { + "epoch": 0.97, + "learning_rate": 3.382790459092075e-05, + "loss": 0.3968, + "step": 25272 + }, + { + "epoch": 0.97, + "learning_rate": 3.382726340087202e-05, + "loss": 0.4792, + "step": 25273 + }, + { + "epoch": 0.97, + "learning_rate": 3.382662221082329e-05, + "loss": 0.456, + "step": 25274 + }, + { + "epoch": 0.97, + "learning_rate": 3.382598102077456e-05, + "loss": 0.4107, + "step": 25275 + }, + { + "epoch": 0.97, + "learning_rate": 3.3825339830725824e-05, + "loss": 0.4887, + "step": 25276 + }, + { + "epoch": 0.97, + "learning_rate": 3.3824698640677096e-05, + "loss": 0.4728, + "step": 25277 + }, + { + "epoch": 0.97, + "learning_rate": 3.382405745062837e-05, + "loss": 0.4272, + "step": 25278 + }, + { + "epoch": 0.97, + "learning_rate": 3.382341626057964e-05, + "loss": 0.4218, + "step": 25279 + }, + { + "epoch": 0.97, + "learning_rate": 3.3822775070530905e-05, + "loss": 0.5686, + "step": 25280 + }, + { + "epoch": 0.97, + "learning_rate": 3.3822133880482176e-05, + "loss": 0.5586, + "step": 25281 + }, + { + "epoch": 0.97, + "learning_rate": 3.382149269043345e-05, + "loss": 0.5196, + "step": 25282 + }, + { + "epoch": 0.97, + "learning_rate": 3.382085150038471e-05, + "loss": 0.3749, + "step": 25283 + }, + { + "epoch": 0.97, + "learning_rate": 3.3820210310335985e-05, + "loss": 0.3482, + "step": 25284 + }, + { + "epoch": 0.97, + "learning_rate": 3.3819569120287256e-05, + "loss": 0.399, + "step": 25285 + }, + { + "epoch": 0.97, + "learning_rate": 3.381892793023853e-05, + "loss": 0.5307, + "step": 25286 + }, + { + "epoch": 0.97, + "learning_rate": 3.381828674018979e-05, + "loss": 0.4536, + "step": 25287 + }, + { + "epoch": 0.97, + "learning_rate": 3.3817645550141065e-05, + "loss": 0.6465, + "step": 25288 + }, + { + "epoch": 0.97, + "learning_rate": 3.381700436009233e-05, + "loss": 0.3267, + "step": 25289 + }, + { + "epoch": 0.97, + "learning_rate": 3.38163631700436e-05, + "loss": 0.5672, + "step": 25290 + }, + { + "epoch": 0.97, + "learning_rate": 3.381572197999487e-05, + "loss": 0.5165, + "step": 25291 + }, + { + "epoch": 0.97, + "learning_rate": 3.381508078994614e-05, + "loss": 0.4415, + "step": 25292 + }, + { + "epoch": 0.97, + "learning_rate": 3.381443959989741e-05, + "loss": 0.4784, + "step": 25293 + }, + { + "epoch": 0.97, + "learning_rate": 3.381379840984868e-05, + "loss": 0.545, + "step": 25294 + }, + { + "epoch": 0.97, + "learning_rate": 3.381315721979995e-05, + "loss": 0.5692, + "step": 25295 + }, + { + "epoch": 0.97, + "learning_rate": 3.381251602975122e-05, + "loss": 0.395, + "step": 25296 + }, + { + "epoch": 0.97, + "learning_rate": 3.381187483970249e-05, + "loss": 0.6064, + "step": 25297 + }, + { + "epoch": 0.97, + "learning_rate": 3.3811233649653755e-05, + "loss": 0.4368, + "step": 25298 + }, + { + "epoch": 0.97, + "learning_rate": 3.3810592459605026e-05, + "loss": 0.4888, + "step": 25299 + }, + { + "epoch": 0.97, + "learning_rate": 3.38099512695563e-05, + "loss": 0.5017, + "step": 25300 + }, + { + "epoch": 0.97, + "learning_rate": 3.380931007950757e-05, + "loss": 0.4386, + "step": 25301 + }, + { + "epoch": 0.97, + "learning_rate": 3.380866888945884e-05, + "loss": 0.4618, + "step": 25302 + }, + { + "epoch": 0.97, + "learning_rate": 3.3808027699410107e-05, + "loss": 0.4379, + "step": 25303 + }, + { + "epoch": 0.97, + "learning_rate": 3.380738650936138e-05, + "loss": 0.4392, + "step": 25304 + }, + { + "epoch": 0.97, + "learning_rate": 3.380674531931264e-05, + "loss": 0.4467, + "step": 25305 + }, + { + "epoch": 0.97, + "learning_rate": 3.3806104129263915e-05, + "loss": 0.4916, + "step": 25306 + }, + { + "epoch": 0.97, + "learning_rate": 3.380546293921518e-05, + "loss": 0.5522, + "step": 25307 + }, + { + "epoch": 0.97, + "learning_rate": 3.380482174916645e-05, + "loss": 0.4514, + "step": 25308 + }, + { + "epoch": 0.97, + "learning_rate": 3.380418055911772e-05, + "loss": 0.4074, + "step": 25309 + }, + { + "epoch": 0.97, + "learning_rate": 3.3803539369068995e-05, + "loss": 0.4148, + "step": 25310 + }, + { + "epoch": 0.97, + "learning_rate": 3.380289817902027e-05, + "loss": 0.4573, + "step": 25311 + }, + { + "epoch": 0.97, + "learning_rate": 3.380225698897153e-05, + "loss": 0.4839, + "step": 25312 + }, + { + "epoch": 0.97, + "learning_rate": 3.3801615798922803e-05, + "loss": 0.4053, + "step": 25313 + }, + { + "epoch": 0.97, + "learning_rate": 3.380097460887407e-05, + "loss": 0.4784, + "step": 25314 + }, + { + "epoch": 0.97, + "learning_rate": 3.380033341882534e-05, + "loss": 0.4034, + "step": 25315 + }, + { + "epoch": 0.97, + "learning_rate": 3.3799692228776605e-05, + "loss": 0.4717, + "step": 25316 + }, + { + "epoch": 0.97, + "learning_rate": 3.3799051038727884e-05, + "loss": 0.5348, + "step": 25317 + }, + { + "epoch": 0.97, + "learning_rate": 3.3798409848679155e-05, + "loss": 0.6502, + "step": 25318 + }, + { + "epoch": 0.97, + "learning_rate": 3.379776865863042e-05, + "loss": 0.5355, + "step": 25319 + }, + { + "epoch": 0.97, + "learning_rate": 3.379712746858169e-05, + "loss": 0.4059, + "step": 25320 + }, + { + "epoch": 0.97, + "learning_rate": 3.379648627853296e-05, + "loss": 0.4671, + "step": 25321 + }, + { + "epoch": 0.97, + "learning_rate": 3.379584508848423e-05, + "loss": 0.5041, + "step": 25322 + }, + { + "epoch": 0.97, + "learning_rate": 3.3795203898435493e-05, + "loss": 0.5345, + "step": 25323 + }, + { + "epoch": 0.97, + "learning_rate": 3.3794562708386765e-05, + "loss": 0.4491, + "step": 25324 + }, + { + "epoch": 0.97, + "learning_rate": 3.379392151833804e-05, + "loss": 0.5334, + "step": 25325 + }, + { + "epoch": 0.97, + "learning_rate": 3.379328032828931e-05, + "loss": 0.4485, + "step": 25326 + }, + { + "epoch": 0.97, + "learning_rate": 3.379263913824058e-05, + "loss": 0.4271, + "step": 25327 + }, + { + "epoch": 0.97, + "learning_rate": 3.3791997948191845e-05, + "loss": 0.4469, + "step": 25328 + }, + { + "epoch": 0.97, + "learning_rate": 3.379135675814312e-05, + "loss": 0.4587, + "step": 25329 + }, + { + "epoch": 0.97, + "learning_rate": 3.379071556809438e-05, + "loss": 0.4609, + "step": 25330 + }, + { + "epoch": 0.97, + "learning_rate": 3.3790074378045654e-05, + "loss": 0.4993, + "step": 25331 + }, + { + "epoch": 0.97, + "learning_rate": 3.378943318799692e-05, + "loss": 0.4763, + "step": 25332 + }, + { + "epoch": 0.97, + "learning_rate": 3.37887919979482e-05, + "loss": 0.5768, + "step": 25333 + }, + { + "epoch": 0.97, + "learning_rate": 3.378815080789947e-05, + "loss": 0.5832, + "step": 25334 + }, + { + "epoch": 0.97, + "learning_rate": 3.3787509617850734e-05, + "loss": 0.4833, + "step": 25335 + }, + { + "epoch": 0.97, + "learning_rate": 3.3786868427802005e-05, + "loss": 0.4808, + "step": 25336 + }, + { + "epoch": 0.97, + "learning_rate": 3.378622723775327e-05, + "loss": 0.5759, + "step": 25337 + }, + { + "epoch": 0.97, + "learning_rate": 3.378558604770454e-05, + "loss": 0.5626, + "step": 25338 + }, + { + "epoch": 0.97, + "learning_rate": 3.378494485765581e-05, + "loss": 0.5506, + "step": 25339 + }, + { + "epoch": 0.97, + "learning_rate": 3.378430366760708e-05, + "loss": 0.4425, + "step": 25340 + }, + { + "epoch": 0.97, + "learning_rate": 3.378366247755835e-05, + "loss": 0.4606, + "step": 25341 + }, + { + "epoch": 0.97, + "learning_rate": 3.378302128750962e-05, + "loss": 0.4064, + "step": 25342 + }, + { + "epoch": 0.97, + "learning_rate": 3.3782380097460894e-05, + "loss": 0.4711, + "step": 25343 + }, + { + "epoch": 0.97, + "learning_rate": 3.378173890741216e-05, + "loss": 0.4904, + "step": 25344 + }, + { + "epoch": 0.97, + "learning_rate": 3.378109771736343e-05, + "loss": 0.5926, + "step": 25345 + }, + { + "epoch": 0.97, + "learning_rate": 3.3780456527314696e-05, + "loss": 0.5327, + "step": 25346 + }, + { + "epoch": 0.97, + "learning_rate": 3.377981533726597e-05, + "loss": 0.5075, + "step": 25347 + }, + { + "epoch": 0.97, + "learning_rate": 3.377917414721723e-05, + "loss": 0.4103, + "step": 25348 + }, + { + "epoch": 0.97, + "learning_rate": 3.3778532957168504e-05, + "loss": 0.4598, + "step": 25349 + }, + { + "epoch": 0.97, + "learning_rate": 3.3777891767119776e-05, + "loss": 0.5085, + "step": 25350 + }, + { + "epoch": 0.97, + "learning_rate": 3.377725057707105e-05, + "loss": 0.4625, + "step": 25351 + }, + { + "epoch": 0.97, + "learning_rate": 3.377660938702232e-05, + "loss": 0.5394, + "step": 25352 + }, + { + "epoch": 0.97, + "learning_rate": 3.3775968196973584e-05, + "loss": 0.4524, + "step": 25353 + }, + { + "epoch": 0.97, + "learning_rate": 3.3775327006924856e-05, + "loss": 0.4583, + "step": 25354 + }, + { + "epoch": 0.97, + "learning_rate": 3.377468581687612e-05, + "loss": 0.4679, + "step": 25355 + }, + { + "epoch": 0.97, + "learning_rate": 3.377404462682739e-05, + "loss": 0.5222, + "step": 25356 + }, + { + "epoch": 0.97, + "learning_rate": 3.3773403436778664e-05, + "loss": 0.5588, + "step": 25357 + }, + { + "epoch": 0.97, + "learning_rate": 3.3772762246729936e-05, + "loss": 0.4935, + "step": 25358 + }, + { + "epoch": 0.97, + "learning_rate": 3.37721210566812e-05, + "loss": 0.4884, + "step": 25359 + }, + { + "epoch": 0.98, + "learning_rate": 3.377147986663247e-05, + "loss": 0.5279, + "step": 25360 + }, + { + "epoch": 0.98, + "learning_rate": 3.3770838676583744e-05, + "loss": 0.4311, + "step": 25361 + }, + { + "epoch": 0.98, + "learning_rate": 3.377019748653501e-05, + "loss": 0.5083, + "step": 25362 + }, + { + "epoch": 0.98, + "learning_rate": 3.376955629648628e-05, + "loss": 0.4228, + "step": 25363 + }, + { + "epoch": 0.98, + "learning_rate": 3.3768915106437546e-05, + "loss": 0.5195, + "step": 25364 + }, + { + "epoch": 0.98, + "learning_rate": 3.376827391638882e-05, + "loss": 0.4458, + "step": 25365 + }, + { + "epoch": 0.98, + "learning_rate": 3.376763272634009e-05, + "loss": 0.4763, + "step": 25366 + }, + { + "epoch": 0.98, + "learning_rate": 3.376699153629136e-05, + "loss": 0.501, + "step": 25367 + }, + { + "epoch": 0.98, + "learning_rate": 3.3766350346242626e-05, + "loss": 0.4228, + "step": 25368 + }, + { + "epoch": 0.98, + "learning_rate": 3.37657091561939e-05, + "loss": 0.5558, + "step": 25369 + }, + { + "epoch": 0.98, + "learning_rate": 3.376506796614517e-05, + "loss": 0.4587, + "step": 25370 + }, + { + "epoch": 0.98, + "learning_rate": 3.3764426776096434e-05, + "loss": 0.5104, + "step": 25371 + }, + { + "epoch": 0.98, + "learning_rate": 3.3763785586047706e-05, + "loss": 0.4756, + "step": 25372 + }, + { + "epoch": 0.98, + "learning_rate": 3.376314439599898e-05, + "loss": 0.4934, + "step": 25373 + }, + { + "epoch": 0.98, + "learning_rate": 3.376250320595025e-05, + "loss": 0.6275, + "step": 25374 + }, + { + "epoch": 0.98, + "learning_rate": 3.3761862015901514e-05, + "loss": 0.5245, + "step": 25375 + }, + { + "epoch": 0.98, + "learning_rate": 3.3761220825852786e-05, + "loss": 0.5933, + "step": 25376 + }, + { + "epoch": 0.98, + "learning_rate": 3.376057963580405e-05, + "loss": 0.5472, + "step": 25377 + }, + { + "epoch": 0.98, + "learning_rate": 3.375993844575532e-05, + "loss": 0.5107, + "step": 25378 + }, + { + "epoch": 0.98, + "learning_rate": 3.3759297255706594e-05, + "loss": 0.4739, + "step": 25379 + }, + { + "epoch": 0.98, + "learning_rate": 3.375865606565786e-05, + "loss": 0.4444, + "step": 25380 + }, + { + "epoch": 0.98, + "learning_rate": 3.375801487560913e-05, + "loss": 0.5368, + "step": 25381 + }, + { + "epoch": 0.98, + "learning_rate": 3.37573736855604e-05, + "loss": 0.419, + "step": 25382 + }, + { + "epoch": 0.98, + "learning_rate": 3.3756732495511675e-05, + "loss": 0.4551, + "step": 25383 + }, + { + "epoch": 0.98, + "learning_rate": 3.375609130546294e-05, + "loss": 0.5086, + "step": 25384 + }, + { + "epoch": 0.98, + "learning_rate": 3.375545011541421e-05, + "loss": 0.5315, + "step": 25385 + }, + { + "epoch": 0.98, + "learning_rate": 3.3754808925365476e-05, + "loss": 0.4406, + "step": 25386 + }, + { + "epoch": 0.98, + "learning_rate": 3.375416773531675e-05, + "loss": 0.4799, + "step": 25387 + }, + { + "epoch": 0.98, + "learning_rate": 3.375352654526802e-05, + "loss": 0.5668, + "step": 25388 + }, + { + "epoch": 0.98, + "learning_rate": 3.375288535521929e-05, + "loss": 0.3994, + "step": 25389 + }, + { + "epoch": 0.98, + "learning_rate": 3.375224416517056e-05, + "loss": 0.4764, + "step": 25390 + }, + { + "epoch": 0.98, + "learning_rate": 3.375160297512183e-05, + "loss": 0.5685, + "step": 25391 + }, + { + "epoch": 0.98, + "learning_rate": 3.37509617850731e-05, + "loss": 0.5664, + "step": 25392 + }, + { + "epoch": 0.98, + "learning_rate": 3.3750320595024365e-05, + "loss": 0.4887, + "step": 25393 + }, + { + "epoch": 0.98, + "learning_rate": 3.3749679404975636e-05, + "loss": 0.4787, + "step": 25394 + }, + { + "epoch": 0.98, + "learning_rate": 3.37490382149269e-05, + "loss": 0.5318, + "step": 25395 + }, + { + "epoch": 0.98, + "learning_rate": 3.374839702487817e-05, + "loss": 0.4005, + "step": 25396 + }, + { + "epoch": 0.98, + "learning_rate": 3.3747755834829445e-05, + "loss": 0.446, + "step": 25397 + }, + { + "epoch": 0.98, + "learning_rate": 3.3747114644780716e-05, + "loss": 0.4462, + "step": 25398 + }, + { + "epoch": 0.98, + "learning_rate": 3.374647345473199e-05, + "loss": 0.4268, + "step": 25399 + }, + { + "epoch": 0.98, + "learning_rate": 3.374583226468325e-05, + "loss": 0.3866, + "step": 25400 + }, + { + "epoch": 0.98, + "learning_rate": 3.3745191074634525e-05, + "loss": 0.5146, + "step": 25401 + }, + { + "epoch": 0.98, + "learning_rate": 3.374454988458579e-05, + "loss": 0.512, + "step": 25402 + }, + { + "epoch": 0.98, + "learning_rate": 3.374390869453706e-05, + "loss": 0.4586, + "step": 25403 + }, + { + "epoch": 0.98, + "learning_rate": 3.3743267504488326e-05, + "loss": 0.4651, + "step": 25404 + }, + { + "epoch": 0.98, + "learning_rate": 3.3742626314439605e-05, + "loss": 0.3829, + "step": 25405 + }, + { + "epoch": 0.98, + "learning_rate": 3.3741985124390877e-05, + "loss": 0.4714, + "step": 25406 + }, + { + "epoch": 0.98, + "learning_rate": 3.374134393434214e-05, + "loss": 0.4057, + "step": 25407 + }, + { + "epoch": 0.98, + "learning_rate": 3.374070274429341e-05, + "loss": 0.5046, + "step": 25408 + }, + { + "epoch": 0.98, + "learning_rate": 3.374006155424468e-05, + "loss": 0.5558, + "step": 25409 + }, + { + "epoch": 0.98, + "learning_rate": 3.373942036419595e-05, + "loss": 0.4398, + "step": 25410 + }, + { + "epoch": 0.98, + "learning_rate": 3.3738779174147215e-05, + "loss": 0.524, + "step": 25411 + }, + { + "epoch": 0.98, + "learning_rate": 3.3738137984098487e-05, + "loss": 0.5236, + "step": 25412 + }, + { + "epoch": 0.98, + "learning_rate": 3.373749679404976e-05, + "loss": 0.5974, + "step": 25413 + }, + { + "epoch": 0.98, + "learning_rate": 3.373685560400103e-05, + "loss": 0.5657, + "step": 25414 + }, + { + "epoch": 0.98, + "learning_rate": 3.37362144139523e-05, + "loss": 0.4586, + "step": 25415 + }, + { + "epoch": 0.98, + "learning_rate": 3.373557322390357e-05, + "loss": 0.5094, + "step": 25416 + }, + { + "epoch": 0.98, + "learning_rate": 3.373493203385484e-05, + "loss": 0.4925, + "step": 25417 + }, + { + "epoch": 0.98, + "learning_rate": 3.37342908438061e-05, + "loss": 0.5191, + "step": 25418 + }, + { + "epoch": 0.98, + "learning_rate": 3.3733649653757375e-05, + "loss": 0.53, + "step": 25419 + }, + { + "epoch": 0.98, + "learning_rate": 3.373300846370864e-05, + "loss": 0.5136, + "step": 25420 + }, + { + "epoch": 0.98, + "learning_rate": 3.373236727365991e-05, + "loss": 0.5409, + "step": 25421 + }, + { + "epoch": 0.98, + "learning_rate": 3.373172608361118e-05, + "loss": 0.5062, + "step": 25422 + }, + { + "epoch": 0.98, + "learning_rate": 3.3731084893562455e-05, + "loss": 0.5249, + "step": 25423 + }, + { + "epoch": 0.98, + "learning_rate": 3.373044370351373e-05, + "loss": 0.4453, + "step": 25424 + }, + { + "epoch": 0.98, + "learning_rate": 3.372980251346499e-05, + "loss": 0.3787, + "step": 25425 + }, + { + "epoch": 0.98, + "learning_rate": 3.3729161323416263e-05, + "loss": 0.4543, + "step": 25426 + }, + { + "epoch": 0.98, + "learning_rate": 3.372852013336753e-05, + "loss": 0.5392, + "step": 25427 + }, + { + "epoch": 0.98, + "learning_rate": 3.37278789433188e-05, + "loss": 0.4734, + "step": 25428 + }, + { + "epoch": 0.98, + "learning_rate": 3.372723775327007e-05, + "loss": 0.5566, + "step": 25429 + }, + { + "epoch": 0.98, + "learning_rate": 3.3726596563221344e-05, + "loss": 0.5016, + "step": 25430 + }, + { + "epoch": 0.98, + "learning_rate": 3.372595537317261e-05, + "loss": 0.3943, + "step": 25431 + }, + { + "epoch": 0.98, + "learning_rate": 3.372531418312388e-05, + "loss": 0.4617, + "step": 25432 + }, + { + "epoch": 0.98, + "learning_rate": 3.372467299307515e-05, + "loss": 0.5303, + "step": 25433 + }, + { + "epoch": 0.98, + "learning_rate": 3.372403180302642e-05, + "loss": 0.5444, + "step": 25434 + }, + { + "epoch": 0.98, + "learning_rate": 3.372339061297769e-05, + "loss": 0.6627, + "step": 25435 + }, + { + "epoch": 0.98, + "learning_rate": 3.3722749422928954e-05, + "loss": 0.4685, + "step": 25436 + }, + { + "epoch": 0.98, + "learning_rate": 3.3722108232880225e-05, + "loss": 0.4463, + "step": 25437 + }, + { + "epoch": 0.98, + "learning_rate": 3.37214670428315e-05, + "loss": 0.4848, + "step": 25438 + }, + { + "epoch": 0.98, + "learning_rate": 3.372082585278277e-05, + "loss": 0.442, + "step": 25439 + }, + { + "epoch": 0.98, + "learning_rate": 3.3720184662734034e-05, + "loss": 0.4799, + "step": 25440 + }, + { + "epoch": 0.98, + "learning_rate": 3.3719543472685305e-05, + "loss": 0.4912, + "step": 25441 + }, + { + "epoch": 0.98, + "learning_rate": 3.371890228263658e-05, + "loss": 0.4708, + "step": 25442 + }, + { + "epoch": 0.98, + "learning_rate": 3.371826109258784e-05, + "loss": 0.5124, + "step": 25443 + }, + { + "epoch": 0.98, + "learning_rate": 3.3717619902539114e-05, + "loss": 0.4775, + "step": 25444 + }, + { + "epoch": 0.98, + "learning_rate": 3.3716978712490385e-05, + "loss": 0.4617, + "step": 25445 + }, + { + "epoch": 0.98, + "learning_rate": 3.371633752244166e-05, + "loss": 0.4087, + "step": 25446 + }, + { + "epoch": 0.98, + "learning_rate": 3.371569633239292e-05, + "loss": 0.4569, + "step": 25447 + }, + { + "epoch": 0.98, + "learning_rate": 3.3715055142344194e-05, + "loss": 0.4862, + "step": 25448 + }, + { + "epoch": 0.98, + "learning_rate": 3.371441395229546e-05, + "loss": 0.4484, + "step": 25449 + }, + { + "epoch": 0.98, + "learning_rate": 3.371377276224673e-05, + "loss": 0.5415, + "step": 25450 + }, + { + "epoch": 0.98, + "learning_rate": 3.3713131572198e-05, + "loss": 0.4199, + "step": 25451 + }, + { + "epoch": 0.98, + "learning_rate": 3.371249038214927e-05, + "loss": 0.4536, + "step": 25452 + }, + { + "epoch": 0.98, + "learning_rate": 3.371184919210054e-05, + "loss": 0.488, + "step": 25453 + }, + { + "epoch": 0.98, + "learning_rate": 3.371120800205181e-05, + "loss": 0.5523, + "step": 25454 + }, + { + "epoch": 0.98, + "learning_rate": 3.371056681200308e-05, + "loss": 0.6406, + "step": 25455 + }, + { + "epoch": 0.98, + "learning_rate": 3.370992562195435e-05, + "loss": 0.4743, + "step": 25456 + }, + { + "epoch": 0.98, + "learning_rate": 3.370928443190562e-05, + "loss": 0.3712, + "step": 25457 + }, + { + "epoch": 0.98, + "learning_rate": 3.3708643241856884e-05, + "loss": 0.4662, + "step": 25458 + }, + { + "epoch": 0.98, + "learning_rate": 3.3708002051808156e-05, + "loss": 0.5588, + "step": 25459 + }, + { + "epoch": 0.98, + "learning_rate": 3.370736086175943e-05, + "loss": 0.5334, + "step": 25460 + }, + { + "epoch": 0.98, + "learning_rate": 3.37067196717107e-05, + "loss": 0.4548, + "step": 25461 + }, + { + "epoch": 0.98, + "learning_rate": 3.370607848166197e-05, + "loss": 0.4714, + "step": 25462 + }, + { + "epoch": 0.98, + "learning_rate": 3.3705437291613236e-05, + "loss": 0.4742, + "step": 25463 + }, + { + "epoch": 0.98, + "learning_rate": 3.370479610156451e-05, + "loss": 0.5599, + "step": 25464 + }, + { + "epoch": 0.98, + "learning_rate": 3.370415491151577e-05, + "loss": 0.5721, + "step": 25465 + }, + { + "epoch": 0.98, + "learning_rate": 3.3703513721467044e-05, + "loss": 0.462, + "step": 25466 + }, + { + "epoch": 0.98, + "learning_rate": 3.370287253141831e-05, + "loss": 0.486, + "step": 25467 + }, + { + "epoch": 0.98, + "learning_rate": 3.370223134136958e-05, + "loss": 0.4429, + "step": 25468 + }, + { + "epoch": 0.98, + "learning_rate": 3.370159015132085e-05, + "loss": 0.381, + "step": 25469 + }, + { + "epoch": 0.98, + "learning_rate": 3.3700948961272124e-05, + "loss": 0.4593, + "step": 25470 + }, + { + "epoch": 0.98, + "learning_rate": 3.3700307771223396e-05, + "loss": 0.6097, + "step": 25471 + }, + { + "epoch": 0.98, + "learning_rate": 3.369966658117466e-05, + "loss": 0.529, + "step": 25472 + }, + { + "epoch": 0.98, + "learning_rate": 3.369902539112593e-05, + "loss": 0.4584, + "step": 25473 + }, + { + "epoch": 0.98, + "learning_rate": 3.36983842010772e-05, + "loss": 0.505, + "step": 25474 + }, + { + "epoch": 0.98, + "learning_rate": 3.369774301102847e-05, + "loss": 0.5279, + "step": 25475 + }, + { + "epoch": 0.98, + "learning_rate": 3.3697101820979734e-05, + "loss": 0.5228, + "step": 25476 + }, + { + "epoch": 0.98, + "learning_rate": 3.369646063093101e-05, + "loss": 0.3372, + "step": 25477 + }, + { + "epoch": 0.98, + "learning_rate": 3.3695819440882284e-05, + "loss": 0.5201, + "step": 25478 + }, + { + "epoch": 0.98, + "learning_rate": 3.369517825083355e-05, + "loss": 0.4368, + "step": 25479 + }, + { + "epoch": 0.98, + "learning_rate": 3.369453706078482e-05, + "loss": 0.4225, + "step": 25480 + }, + { + "epoch": 0.98, + "learning_rate": 3.3693895870736086e-05, + "loss": 0.4649, + "step": 25481 + }, + { + "epoch": 0.98, + "learning_rate": 3.369325468068736e-05, + "loss": 0.5609, + "step": 25482 + }, + { + "epoch": 0.98, + "learning_rate": 3.369261349063862e-05, + "loss": 0.6058, + "step": 25483 + }, + { + "epoch": 0.98, + "learning_rate": 3.3691972300589894e-05, + "loss": 0.599, + "step": 25484 + }, + { + "epoch": 0.98, + "learning_rate": 3.3691331110541166e-05, + "loss": 0.597, + "step": 25485 + }, + { + "epoch": 0.98, + "learning_rate": 3.369068992049244e-05, + "loss": 0.5341, + "step": 25486 + }, + { + "epoch": 0.98, + "learning_rate": 3.369004873044371e-05, + "loss": 0.4191, + "step": 25487 + }, + { + "epoch": 0.98, + "learning_rate": 3.3689407540394974e-05, + "loss": 0.451, + "step": 25488 + }, + { + "epoch": 0.98, + "learning_rate": 3.3688766350346246e-05, + "loss": 0.5026, + "step": 25489 + }, + { + "epoch": 0.98, + "learning_rate": 3.368812516029751e-05, + "loss": 0.4899, + "step": 25490 + }, + { + "epoch": 0.98, + "learning_rate": 3.368748397024878e-05, + "loss": 0.5148, + "step": 25491 + }, + { + "epoch": 0.98, + "learning_rate": 3.368684278020005e-05, + "loss": 0.4596, + "step": 25492 + }, + { + "epoch": 0.98, + "learning_rate": 3.3686201590151326e-05, + "loss": 0.4404, + "step": 25493 + }, + { + "epoch": 0.98, + "learning_rate": 3.368556040010259e-05, + "loss": 0.4682, + "step": 25494 + }, + { + "epoch": 0.98, + "learning_rate": 3.368491921005386e-05, + "loss": 0.4767, + "step": 25495 + }, + { + "epoch": 0.98, + "learning_rate": 3.3684278020005135e-05, + "loss": 0.5865, + "step": 25496 + }, + { + "epoch": 0.98, + "learning_rate": 3.36836368299564e-05, + "loss": 0.5169, + "step": 25497 + }, + { + "epoch": 0.98, + "learning_rate": 3.368299563990767e-05, + "loss": 0.7067, + "step": 25498 + }, + { + "epoch": 0.98, + "learning_rate": 3.3682354449858936e-05, + "loss": 0.5324, + "step": 25499 + }, + { + "epoch": 0.98, + "learning_rate": 3.368171325981021e-05, + "loss": 0.5146, + "step": 25500 + }, + { + "epoch": 0.98, + "learning_rate": 3.368107206976148e-05, + "loss": 0.4942, + "step": 25501 + }, + { + "epoch": 0.98, + "learning_rate": 3.368043087971275e-05, + "loss": 0.4034, + "step": 25502 + }, + { + "epoch": 0.98, + "learning_rate": 3.3679789689664016e-05, + "loss": 0.5416, + "step": 25503 + }, + { + "epoch": 0.98, + "learning_rate": 3.367914849961529e-05, + "loss": 0.4258, + "step": 25504 + }, + { + "epoch": 0.98, + "learning_rate": 3.367850730956656e-05, + "loss": 0.4693, + "step": 25505 + }, + { + "epoch": 0.98, + "learning_rate": 3.3677866119517825e-05, + "loss": 0.5338, + "step": 25506 + }, + { + "epoch": 0.98, + "learning_rate": 3.3677224929469096e-05, + "loss": 0.4275, + "step": 25507 + }, + { + "epoch": 0.98, + "learning_rate": 3.367658373942036e-05, + "loss": 0.443, + "step": 25508 + }, + { + "epoch": 0.98, + "learning_rate": 3.367594254937163e-05, + "loss": 0.4842, + "step": 25509 + }, + { + "epoch": 0.98, + "learning_rate": 3.3675301359322905e-05, + "loss": 0.5212, + "step": 25510 + }, + { + "epoch": 0.98, + "learning_rate": 3.3674660169274176e-05, + "loss": 0.5734, + "step": 25511 + }, + { + "epoch": 0.98, + "learning_rate": 3.367401897922545e-05, + "loss": 0.4273, + "step": 25512 + }, + { + "epoch": 0.98, + "learning_rate": 3.367337778917671e-05, + "loss": 0.4256, + "step": 25513 + }, + { + "epoch": 0.98, + "learning_rate": 3.3672736599127985e-05, + "loss": 0.3936, + "step": 25514 + }, + { + "epoch": 0.98, + "learning_rate": 3.367209540907925e-05, + "loss": 0.4884, + "step": 25515 + }, + { + "epoch": 0.98, + "learning_rate": 3.367145421903052e-05, + "loss": 0.5292, + "step": 25516 + }, + { + "epoch": 0.98, + "learning_rate": 3.367081302898179e-05, + "loss": 0.5954, + "step": 25517 + }, + { + "epoch": 0.98, + "learning_rate": 3.3670171838933065e-05, + "loss": 0.5151, + "step": 25518 + }, + { + "epoch": 0.98, + "learning_rate": 3.366953064888433e-05, + "loss": 0.4497, + "step": 25519 + }, + { + "epoch": 0.98, + "learning_rate": 3.36688894588356e-05, + "loss": 0.5447, + "step": 25520 + }, + { + "epoch": 0.98, + "learning_rate": 3.366824826878687e-05, + "loss": 0.4679, + "step": 25521 + }, + { + "epoch": 0.98, + "learning_rate": 3.366760707873814e-05, + "loss": 0.4427, + "step": 25522 + }, + { + "epoch": 0.98, + "learning_rate": 3.366696588868941e-05, + "loss": 0.5107, + "step": 25523 + }, + { + "epoch": 0.98, + "learning_rate": 3.3666324698640675e-05, + "loss": 0.3774, + "step": 25524 + }, + { + "epoch": 0.98, + "learning_rate": 3.3665683508591947e-05, + "loss": 0.5168, + "step": 25525 + }, + { + "epoch": 0.98, + "learning_rate": 3.366504231854322e-05, + "loss": 0.2846, + "step": 25526 + }, + { + "epoch": 0.98, + "learning_rate": 3.366440112849449e-05, + "loss": 0.4411, + "step": 25527 + }, + { + "epoch": 0.98, + "learning_rate": 3.3663759938445755e-05, + "loss": 0.4772, + "step": 25528 + }, + { + "epoch": 0.98, + "learning_rate": 3.366311874839703e-05, + "loss": 0.4019, + "step": 25529 + }, + { + "epoch": 0.98, + "learning_rate": 3.36624775583483e-05, + "loss": 0.4985, + "step": 25530 + }, + { + "epoch": 0.98, + "learning_rate": 3.366183636829956e-05, + "loss": 0.5418, + "step": 25531 + }, + { + "epoch": 0.98, + "learning_rate": 3.3661195178250835e-05, + "loss": 0.484, + "step": 25532 + }, + { + "epoch": 0.98, + "learning_rate": 3.366055398820211e-05, + "loss": 0.5226, + "step": 25533 + }, + { + "epoch": 0.98, + "learning_rate": 3.365991279815338e-05, + "loss": 0.4205, + "step": 25534 + }, + { + "epoch": 0.98, + "learning_rate": 3.3659271608104643e-05, + "loss": 0.4518, + "step": 25535 + }, + { + "epoch": 0.98, + "learning_rate": 3.3658630418055915e-05, + "loss": 0.4729, + "step": 25536 + }, + { + "epoch": 0.98, + "learning_rate": 3.365798922800718e-05, + "loss": 0.4464, + "step": 25537 + }, + { + "epoch": 0.98, + "learning_rate": 3.365734803795845e-05, + "loss": 0.4582, + "step": 25538 + }, + { + "epoch": 0.98, + "learning_rate": 3.3656706847909723e-05, + "loss": 0.4576, + "step": 25539 + }, + { + "epoch": 0.98, + "learning_rate": 3.365606565786099e-05, + "loss": 0.4446, + "step": 25540 + }, + { + "epoch": 0.98, + "learning_rate": 3.365542446781226e-05, + "loss": 0.429, + "step": 25541 + }, + { + "epoch": 0.98, + "learning_rate": 3.365478327776353e-05, + "loss": 0.5683, + "step": 25542 + }, + { + "epoch": 0.98, + "learning_rate": 3.3654142087714804e-05, + "loss": 0.5434, + "step": 25543 + }, + { + "epoch": 0.98, + "learning_rate": 3.365350089766607e-05, + "loss": 0.4623, + "step": 25544 + }, + { + "epoch": 0.98, + "learning_rate": 3.365285970761734e-05, + "loss": 0.5163, + "step": 25545 + }, + { + "epoch": 0.98, + "learning_rate": 3.3652218517568605e-05, + "loss": 0.5724, + "step": 25546 + }, + { + "epoch": 0.98, + "learning_rate": 3.365157732751988e-05, + "loss": 0.5635, + "step": 25547 + }, + { + "epoch": 0.98, + "learning_rate": 3.365093613747115e-05, + "loss": 0.4251, + "step": 25548 + }, + { + "epoch": 0.98, + "learning_rate": 3.365029494742242e-05, + "loss": 0.5921, + "step": 25549 + }, + { + "epoch": 0.98, + "learning_rate": 3.364965375737369e-05, + "loss": 0.4159, + "step": 25550 + }, + { + "epoch": 0.98, + "learning_rate": 3.364901256732496e-05, + "loss": 0.4468, + "step": 25551 + }, + { + "epoch": 0.98, + "learning_rate": 3.364837137727623e-05, + "loss": 0.5763, + "step": 25552 + }, + { + "epoch": 0.98, + "learning_rate": 3.3647730187227494e-05, + "loss": 0.5508, + "step": 25553 + }, + { + "epoch": 0.98, + "learning_rate": 3.3647088997178765e-05, + "loss": 0.4402, + "step": 25554 + }, + { + "epoch": 0.98, + "learning_rate": 3.364644780713003e-05, + "loss": 0.457, + "step": 25555 + }, + { + "epoch": 0.98, + "learning_rate": 3.36458066170813e-05, + "loss": 0.471, + "step": 25556 + }, + { + "epoch": 0.98, + "learning_rate": 3.3645165427032574e-05, + "loss": 0.4318, + "step": 25557 + }, + { + "epoch": 0.98, + "learning_rate": 3.3644524236983845e-05, + "loss": 0.507, + "step": 25558 + }, + { + "epoch": 0.98, + "learning_rate": 3.364388304693512e-05, + "loss": 0.4377, + "step": 25559 + }, + { + "epoch": 0.98, + "learning_rate": 3.364324185688638e-05, + "loss": 0.4863, + "step": 25560 + }, + { + "epoch": 0.98, + "learning_rate": 3.3642600666837654e-05, + "loss": 0.45, + "step": 25561 + }, + { + "epoch": 0.98, + "learning_rate": 3.364195947678892e-05, + "loss": 0.418, + "step": 25562 + }, + { + "epoch": 0.98, + "learning_rate": 3.364131828674019e-05, + "loss": 0.5859, + "step": 25563 + }, + { + "epoch": 0.98, + "learning_rate": 3.3640677096691455e-05, + "loss": 0.4324, + "step": 25564 + }, + { + "epoch": 0.98, + "learning_rate": 3.3640035906642734e-05, + "loss": 0.5444, + "step": 25565 + }, + { + "epoch": 0.98, + "learning_rate": 3.3639394716594006e-05, + "loss": 0.6352, + "step": 25566 + }, + { + "epoch": 0.98, + "learning_rate": 3.363875352654527e-05, + "loss": 0.4175, + "step": 25567 + }, + { + "epoch": 0.98, + "learning_rate": 3.363811233649654e-05, + "loss": 0.3633, + "step": 25568 + }, + { + "epoch": 0.98, + "learning_rate": 3.363747114644781e-05, + "loss": 0.5098, + "step": 25569 + }, + { + "epoch": 0.98, + "learning_rate": 3.363682995639908e-05, + "loss": 0.4507, + "step": 25570 + }, + { + "epoch": 0.98, + "learning_rate": 3.3636188766350344e-05, + "loss": 0.5263, + "step": 25571 + }, + { + "epoch": 0.98, + "learning_rate": 3.3635547576301616e-05, + "loss": 0.4565, + "step": 25572 + }, + { + "epoch": 0.98, + "learning_rate": 3.363490638625289e-05, + "loss": 0.5589, + "step": 25573 + }, + { + "epoch": 0.98, + "learning_rate": 3.363426519620416e-05, + "loss": 0.4971, + "step": 25574 + }, + { + "epoch": 0.98, + "learning_rate": 3.363362400615543e-05, + "loss": 0.4327, + "step": 25575 + }, + { + "epoch": 0.98, + "learning_rate": 3.3632982816106696e-05, + "loss": 0.5015, + "step": 25576 + }, + { + "epoch": 0.98, + "learning_rate": 3.363234162605797e-05, + "loss": 0.4862, + "step": 25577 + }, + { + "epoch": 0.98, + "learning_rate": 3.363170043600923e-05, + "loss": 0.4793, + "step": 25578 + }, + { + "epoch": 0.98, + "learning_rate": 3.3631059245960504e-05, + "loss": 0.4205, + "step": 25579 + }, + { + "epoch": 0.98, + "learning_rate": 3.363041805591177e-05, + "loss": 0.4947, + "step": 25580 + }, + { + "epoch": 0.98, + "learning_rate": 3.362977686586304e-05, + "loss": 0.4341, + "step": 25581 + }, + { + "epoch": 0.98, + "learning_rate": 3.362913567581431e-05, + "loss": 0.6476, + "step": 25582 + }, + { + "epoch": 0.98, + "learning_rate": 3.3628494485765584e-05, + "loss": 0.4912, + "step": 25583 + }, + { + "epoch": 0.98, + "learning_rate": 3.3627853295716856e-05, + "loss": 0.5042, + "step": 25584 + }, + { + "epoch": 0.98, + "learning_rate": 3.362721210566812e-05, + "loss": 0.5621, + "step": 25585 + }, + { + "epoch": 0.98, + "learning_rate": 3.362657091561939e-05, + "loss": 0.5273, + "step": 25586 + }, + { + "epoch": 0.98, + "learning_rate": 3.362592972557066e-05, + "loss": 0.5329, + "step": 25587 + }, + { + "epoch": 0.98, + "learning_rate": 3.362528853552193e-05, + "loss": 0.4707, + "step": 25588 + }, + { + "epoch": 0.98, + "learning_rate": 3.36246473454732e-05, + "loss": 0.3916, + "step": 25589 + }, + { + "epoch": 0.98, + "learning_rate": 3.362400615542447e-05, + "loss": 0.6777, + "step": 25590 + }, + { + "epoch": 0.98, + "learning_rate": 3.362336496537574e-05, + "loss": 0.5798, + "step": 25591 + }, + { + "epoch": 0.98, + "learning_rate": 3.362272377532701e-05, + "loss": 0.4018, + "step": 25592 + }, + { + "epoch": 0.98, + "learning_rate": 3.362208258527828e-05, + "loss": 0.5091, + "step": 25593 + }, + { + "epoch": 0.98, + "learning_rate": 3.3621441395229546e-05, + "loss": 0.428, + "step": 25594 + }, + { + "epoch": 0.98, + "learning_rate": 3.362080020518082e-05, + "loss": 0.4625, + "step": 25595 + }, + { + "epoch": 0.98, + "learning_rate": 3.362015901513208e-05, + "loss": 0.4196, + "step": 25596 + }, + { + "epoch": 0.98, + "learning_rate": 3.3619517825083354e-05, + "loss": 0.4801, + "step": 25597 + }, + { + "epoch": 0.98, + "learning_rate": 3.3618876635034626e-05, + "loss": 0.486, + "step": 25598 + }, + { + "epoch": 0.98, + "learning_rate": 3.36182354449859e-05, + "loss": 0.3606, + "step": 25599 + }, + { + "epoch": 0.98, + "learning_rate": 3.361759425493716e-05, + "loss": 0.5063, + "step": 25600 + }, + { + "epoch": 0.98, + "learning_rate": 3.3616953064888434e-05, + "loss": 0.5311, + "step": 25601 + }, + { + "epoch": 0.98, + "learning_rate": 3.3616311874839706e-05, + "loss": 0.3989, + "step": 25602 + }, + { + "epoch": 0.98, + "learning_rate": 3.361567068479097e-05, + "loss": 0.5563, + "step": 25603 + }, + { + "epoch": 0.98, + "learning_rate": 3.361502949474224e-05, + "loss": 0.5417, + "step": 25604 + }, + { + "epoch": 0.98, + "learning_rate": 3.3614388304693514e-05, + "loss": 0.5456, + "step": 25605 + }, + { + "epoch": 0.98, + "learning_rate": 3.3613747114644786e-05, + "loss": 0.405, + "step": 25606 + }, + { + "epoch": 0.98, + "learning_rate": 3.361310592459605e-05, + "loss": 0.4937, + "step": 25607 + }, + { + "epoch": 0.98, + "learning_rate": 3.361246473454732e-05, + "loss": 0.4618, + "step": 25608 + }, + { + "epoch": 0.98, + "learning_rate": 3.361182354449859e-05, + "loss": 0.4866, + "step": 25609 + }, + { + "epoch": 0.98, + "learning_rate": 3.361118235444986e-05, + "loss": 0.4363, + "step": 25610 + }, + { + "epoch": 0.98, + "learning_rate": 3.361054116440113e-05, + "loss": 0.4467, + "step": 25611 + }, + { + "epoch": 0.98, + "learning_rate": 3.3609899974352396e-05, + "loss": 0.4358, + "step": 25612 + }, + { + "epoch": 0.98, + "learning_rate": 3.360925878430367e-05, + "loss": 0.4238, + "step": 25613 + }, + { + "epoch": 0.98, + "learning_rate": 3.360861759425494e-05, + "loss": 0.5656, + "step": 25614 + }, + { + "epoch": 0.98, + "learning_rate": 3.360797640420621e-05, + "loss": 0.4136, + "step": 25615 + }, + { + "epoch": 0.98, + "learning_rate": 3.3607335214157476e-05, + "loss": 0.5106, + "step": 25616 + }, + { + "epoch": 0.98, + "learning_rate": 3.360669402410875e-05, + "loss": 0.6268, + "step": 25617 + }, + { + "epoch": 0.98, + "learning_rate": 3.360605283406001e-05, + "loss": 0.4967, + "step": 25618 + }, + { + "epoch": 0.98, + "learning_rate": 3.3605411644011285e-05, + "loss": 0.5864, + "step": 25619 + }, + { + "epoch": 0.99, + "learning_rate": 3.3604770453962556e-05, + "loss": 0.4804, + "step": 25620 + }, + { + "epoch": 0.99, + "learning_rate": 3.360412926391383e-05, + "loss": 0.3961, + "step": 25621 + }, + { + "epoch": 0.99, + "learning_rate": 3.36034880738651e-05, + "loss": 0.55, + "step": 25622 + }, + { + "epoch": 0.99, + "learning_rate": 3.3602846883816365e-05, + "loss": 0.5528, + "step": 25623 + }, + { + "epoch": 0.99, + "learning_rate": 3.3602205693767636e-05, + "loss": 0.5423, + "step": 25624 + }, + { + "epoch": 0.99, + "learning_rate": 3.36015645037189e-05, + "loss": 0.5739, + "step": 25625 + }, + { + "epoch": 0.99, + "learning_rate": 3.360092331367017e-05, + "loss": 0.4318, + "step": 25626 + }, + { + "epoch": 0.99, + "learning_rate": 3.360028212362144e-05, + "loss": 0.6169, + "step": 25627 + }, + { + "epoch": 0.99, + "learning_rate": 3.359964093357271e-05, + "loss": 0.4488, + "step": 25628 + }, + { + "epoch": 0.99, + "learning_rate": 3.359899974352398e-05, + "loss": 0.4338, + "step": 25629 + }, + { + "epoch": 0.99, + "learning_rate": 3.359835855347525e-05, + "loss": 0.5375, + "step": 25630 + }, + { + "epoch": 0.99, + "learning_rate": 3.3597717363426525e-05, + "loss": 0.4824, + "step": 25631 + }, + { + "epoch": 0.99, + "learning_rate": 3.359707617337779e-05, + "loss": 0.4985, + "step": 25632 + }, + { + "epoch": 0.99, + "learning_rate": 3.359643498332906e-05, + "loss": 0.3921, + "step": 25633 + }, + { + "epoch": 0.99, + "learning_rate": 3.3595793793280327e-05, + "loss": 0.4297, + "step": 25634 + }, + { + "epoch": 0.99, + "learning_rate": 3.35951526032316e-05, + "loss": 0.3872, + "step": 25635 + }, + { + "epoch": 0.99, + "learning_rate": 3.359451141318286e-05, + "loss": 0.5291, + "step": 25636 + }, + { + "epoch": 0.99, + "learning_rate": 3.359387022313414e-05, + "loss": 0.4381, + "step": 25637 + }, + { + "epoch": 0.99, + "learning_rate": 3.359322903308541e-05, + "loss": 0.4797, + "step": 25638 + }, + { + "epoch": 0.99, + "learning_rate": 3.359258784303668e-05, + "loss": 0.5338, + "step": 25639 + }, + { + "epoch": 0.99, + "learning_rate": 3.359194665298795e-05, + "loss": 0.6074, + "step": 25640 + }, + { + "epoch": 0.99, + "learning_rate": 3.3591305462939215e-05, + "loss": 0.54, + "step": 25641 + }, + { + "epoch": 0.99, + "learning_rate": 3.359066427289049e-05, + "loss": 0.4138, + "step": 25642 + }, + { + "epoch": 0.99, + "learning_rate": 3.359002308284175e-05, + "loss": 0.5542, + "step": 25643 + }, + { + "epoch": 0.99, + "learning_rate": 3.358938189279302e-05, + "loss": 0.5145, + "step": 25644 + }, + { + "epoch": 0.99, + "learning_rate": 3.3588740702744295e-05, + "loss": 0.4124, + "step": 25645 + }, + { + "epoch": 0.99, + "learning_rate": 3.358809951269557e-05, + "loss": 0.619, + "step": 25646 + }, + { + "epoch": 0.99, + "learning_rate": 3.358745832264684e-05, + "loss": 0.3809, + "step": 25647 + }, + { + "epoch": 0.99, + "learning_rate": 3.3586817132598103e-05, + "loss": 0.5077, + "step": 25648 + }, + { + "epoch": 0.99, + "learning_rate": 3.3586175942549375e-05, + "loss": 0.4435, + "step": 25649 + }, + { + "epoch": 0.99, + "learning_rate": 3.358553475250064e-05, + "loss": 0.4064, + "step": 25650 + }, + { + "epoch": 0.99, + "learning_rate": 3.358489356245191e-05, + "loss": 0.4916, + "step": 25651 + }, + { + "epoch": 0.99, + "learning_rate": 3.358425237240318e-05, + "loss": 0.4598, + "step": 25652 + }, + { + "epoch": 0.99, + "learning_rate": 3.3583611182354455e-05, + "loss": 0.4069, + "step": 25653 + }, + { + "epoch": 0.99, + "learning_rate": 3.358296999230572e-05, + "loss": 0.396, + "step": 25654 + }, + { + "epoch": 0.99, + "learning_rate": 3.358232880225699e-05, + "loss": 0.5606, + "step": 25655 + }, + { + "epoch": 0.99, + "learning_rate": 3.3581687612208264e-05, + "loss": 0.4478, + "step": 25656 + }, + { + "epoch": 0.99, + "learning_rate": 3.358104642215953e-05, + "loss": 0.4971, + "step": 25657 + }, + { + "epoch": 0.99, + "learning_rate": 3.35804052321108e-05, + "loss": 0.457, + "step": 25658 + }, + { + "epoch": 0.99, + "learning_rate": 3.3579764042062065e-05, + "loss": 0.5348, + "step": 25659 + }, + { + "epoch": 0.99, + "learning_rate": 3.357912285201334e-05, + "loss": 0.4503, + "step": 25660 + }, + { + "epoch": 0.99, + "learning_rate": 3.357848166196461e-05, + "loss": 0.4318, + "step": 25661 + }, + { + "epoch": 0.99, + "learning_rate": 3.357784047191588e-05, + "loss": 0.4342, + "step": 25662 + }, + { + "epoch": 0.99, + "learning_rate": 3.3577199281867145e-05, + "loss": 0.4974, + "step": 25663 + }, + { + "epoch": 0.99, + "learning_rate": 3.357655809181842e-05, + "loss": 0.4334, + "step": 25664 + }, + { + "epoch": 0.99, + "learning_rate": 3.357591690176969e-05, + "loss": 0.4704, + "step": 25665 + }, + { + "epoch": 0.99, + "learning_rate": 3.3575275711720954e-05, + "loss": 0.4695, + "step": 25666 + }, + { + "epoch": 0.99, + "learning_rate": 3.3574634521672225e-05, + "loss": 0.3999, + "step": 25667 + }, + { + "epoch": 0.99, + "learning_rate": 3.357399333162349e-05, + "loss": 0.4609, + "step": 25668 + }, + { + "epoch": 0.99, + "learning_rate": 3.357335214157476e-05, + "loss": 0.3851, + "step": 25669 + }, + { + "epoch": 0.99, + "learning_rate": 3.3572710951526034e-05, + "loss": 0.3406, + "step": 25670 + }, + { + "epoch": 0.99, + "learning_rate": 3.3572069761477305e-05, + "loss": 0.4546, + "step": 25671 + }, + { + "epoch": 0.99, + "learning_rate": 3.357142857142857e-05, + "loss": 0.5464, + "step": 25672 + }, + { + "epoch": 0.99, + "learning_rate": 3.357078738137984e-05, + "loss": 0.493, + "step": 25673 + }, + { + "epoch": 0.99, + "learning_rate": 3.3570146191331114e-05, + "loss": 0.4303, + "step": 25674 + }, + { + "epoch": 0.99, + "learning_rate": 3.356950500128238e-05, + "loss": 0.4443, + "step": 25675 + }, + { + "epoch": 0.99, + "learning_rate": 3.356886381123365e-05, + "loss": 0.441, + "step": 25676 + }, + { + "epoch": 0.99, + "learning_rate": 3.356822262118492e-05, + "loss": 0.3819, + "step": 25677 + }, + { + "epoch": 0.99, + "learning_rate": 3.3567581431136194e-05, + "loss": 0.413, + "step": 25678 + }, + { + "epoch": 0.99, + "learning_rate": 3.356694024108746e-05, + "loss": 0.5788, + "step": 25679 + }, + { + "epoch": 0.99, + "learning_rate": 3.356629905103873e-05, + "loss": 0.6073, + "step": 25680 + }, + { + "epoch": 0.99, + "learning_rate": 3.3565657860989996e-05, + "loss": 0.5397, + "step": 25681 + }, + { + "epoch": 0.99, + "learning_rate": 3.356501667094127e-05, + "loss": 0.3434, + "step": 25682 + }, + { + "epoch": 0.99, + "learning_rate": 3.356437548089254e-05, + "loss": 0.5962, + "step": 25683 + }, + { + "epoch": 0.99, + "learning_rate": 3.3563734290843804e-05, + "loss": 0.5864, + "step": 25684 + }, + { + "epoch": 0.99, + "learning_rate": 3.3563093100795076e-05, + "loss": 0.5534, + "step": 25685 + }, + { + "epoch": 0.99, + "learning_rate": 3.356245191074635e-05, + "loss": 0.4344, + "step": 25686 + }, + { + "epoch": 0.99, + "learning_rate": 3.356181072069762e-05, + "loss": 0.4641, + "step": 25687 + }, + { + "epoch": 0.99, + "learning_rate": 3.3561169530648884e-05, + "loss": 0.4478, + "step": 25688 + }, + { + "epoch": 0.99, + "learning_rate": 3.3560528340600156e-05, + "loss": 0.416, + "step": 25689 + }, + { + "epoch": 0.99, + "learning_rate": 3.355988715055142e-05, + "loss": 0.4255, + "step": 25690 + }, + { + "epoch": 0.99, + "learning_rate": 3.355924596050269e-05, + "loss": 0.6253, + "step": 25691 + }, + { + "epoch": 0.99, + "learning_rate": 3.3558604770453964e-05, + "loss": 0.4748, + "step": 25692 + }, + { + "epoch": 0.99, + "learning_rate": 3.3557963580405236e-05, + "loss": 0.5903, + "step": 25693 + }, + { + "epoch": 0.99, + "learning_rate": 3.355732239035651e-05, + "loss": 0.4745, + "step": 25694 + }, + { + "epoch": 0.99, + "learning_rate": 3.355668120030777e-05, + "loss": 0.4686, + "step": 25695 + }, + { + "epoch": 0.99, + "learning_rate": 3.3556040010259044e-05, + "loss": 0.4724, + "step": 25696 + }, + { + "epoch": 0.99, + "learning_rate": 3.355539882021031e-05, + "loss": 0.5054, + "step": 25697 + }, + { + "epoch": 0.99, + "learning_rate": 3.355475763016158e-05, + "loss": 0.5971, + "step": 25698 + }, + { + "epoch": 0.99, + "learning_rate": 3.355411644011285e-05, + "loss": 0.4881, + "step": 25699 + }, + { + "epoch": 0.99, + "learning_rate": 3.355347525006412e-05, + "loss": 0.4935, + "step": 25700 + }, + { + "epoch": 0.99, + "learning_rate": 3.355283406001539e-05, + "loss": 0.4356, + "step": 25701 + }, + { + "epoch": 0.99, + "learning_rate": 3.355219286996666e-05, + "loss": 0.5889, + "step": 25702 + }, + { + "epoch": 0.99, + "learning_rate": 3.355155167991793e-05, + "loss": 0.4373, + "step": 25703 + }, + { + "epoch": 0.99, + "learning_rate": 3.35509104898692e-05, + "loss": 0.4762, + "step": 25704 + }, + { + "epoch": 0.99, + "learning_rate": 3.355026929982047e-05, + "loss": 0.433, + "step": 25705 + }, + { + "epoch": 0.99, + "learning_rate": 3.3549628109771734e-05, + "loss": 0.5177, + "step": 25706 + }, + { + "epoch": 0.99, + "learning_rate": 3.3548986919723006e-05, + "loss": 0.4501, + "step": 25707 + }, + { + "epoch": 0.99, + "learning_rate": 3.354834572967428e-05, + "loss": 0.3506, + "step": 25708 + }, + { + "epoch": 0.99, + "learning_rate": 3.354770453962555e-05, + "loss": 0.4489, + "step": 25709 + }, + { + "epoch": 0.99, + "learning_rate": 3.354706334957682e-05, + "loss": 0.4655, + "step": 25710 + }, + { + "epoch": 0.99, + "learning_rate": 3.3546422159528086e-05, + "loss": 0.469, + "step": 25711 + }, + { + "epoch": 0.99, + "learning_rate": 3.354578096947936e-05, + "loss": 0.5538, + "step": 25712 + }, + { + "epoch": 0.99, + "learning_rate": 3.354513977943062e-05, + "loss": 0.5487, + "step": 25713 + }, + { + "epoch": 0.99, + "learning_rate": 3.3544498589381894e-05, + "loss": 0.5362, + "step": 25714 + }, + { + "epoch": 0.99, + "learning_rate": 3.354385739933316e-05, + "loss": 0.4339, + "step": 25715 + }, + { + "epoch": 0.99, + "learning_rate": 3.354321620928443e-05, + "loss": 0.608, + "step": 25716 + }, + { + "epoch": 0.99, + "learning_rate": 3.35425750192357e-05, + "loss": 0.4423, + "step": 25717 + }, + { + "epoch": 0.99, + "learning_rate": 3.3541933829186975e-05, + "loss": 0.5302, + "step": 25718 + }, + { + "epoch": 0.99, + "learning_rate": 3.3541292639138246e-05, + "loss": 0.362, + "step": 25719 + }, + { + "epoch": 0.99, + "learning_rate": 3.354065144908951e-05, + "loss": 0.5096, + "step": 25720 + }, + { + "epoch": 0.99, + "learning_rate": 3.354001025904078e-05, + "loss": 0.5915, + "step": 25721 + }, + { + "epoch": 0.99, + "learning_rate": 3.353936906899205e-05, + "loss": 0.5002, + "step": 25722 + }, + { + "epoch": 0.99, + "learning_rate": 3.353872787894332e-05, + "loss": 0.3756, + "step": 25723 + }, + { + "epoch": 0.99, + "learning_rate": 3.3538086688894584e-05, + "loss": 0.4433, + "step": 25724 + }, + { + "epoch": 0.99, + "learning_rate": 3.353744549884586e-05, + "loss": 0.5534, + "step": 25725 + }, + { + "epoch": 0.99, + "learning_rate": 3.3536804308797135e-05, + "loss": 0.4215, + "step": 25726 + }, + { + "epoch": 0.99, + "learning_rate": 3.35361631187484e-05, + "loss": 0.4579, + "step": 25727 + }, + { + "epoch": 0.99, + "learning_rate": 3.353552192869967e-05, + "loss": 0.6385, + "step": 25728 + }, + { + "epoch": 0.99, + "learning_rate": 3.3534880738650936e-05, + "loss": 0.5129, + "step": 25729 + }, + { + "epoch": 0.99, + "learning_rate": 3.353423954860221e-05, + "loss": 0.4702, + "step": 25730 + }, + { + "epoch": 0.99, + "learning_rate": 3.353359835855347e-05, + "loss": 0.4543, + "step": 25731 + }, + { + "epoch": 0.99, + "learning_rate": 3.3532957168504745e-05, + "loss": 0.4816, + "step": 25732 + }, + { + "epoch": 0.99, + "learning_rate": 3.3532315978456016e-05, + "loss": 0.4858, + "step": 25733 + }, + { + "epoch": 0.99, + "learning_rate": 3.353167478840729e-05, + "loss": 0.3748, + "step": 25734 + }, + { + "epoch": 0.99, + "learning_rate": 3.353103359835856e-05, + "loss": 0.5017, + "step": 25735 + }, + { + "epoch": 0.99, + "learning_rate": 3.3530392408309825e-05, + "loss": 0.5205, + "step": 25736 + }, + { + "epoch": 0.99, + "learning_rate": 3.3529751218261096e-05, + "loss": 0.5383, + "step": 25737 + }, + { + "epoch": 0.99, + "learning_rate": 3.352911002821236e-05, + "loss": 0.4951, + "step": 25738 + }, + { + "epoch": 0.99, + "learning_rate": 3.352846883816363e-05, + "loss": 0.4744, + "step": 25739 + }, + { + "epoch": 0.99, + "learning_rate": 3.35278276481149e-05, + "loss": 0.4925, + "step": 25740 + }, + { + "epoch": 0.99, + "learning_rate": 3.3527186458066177e-05, + "loss": 0.4506, + "step": 25741 + }, + { + "epoch": 0.99, + "learning_rate": 3.352654526801744e-05, + "loss": 0.4568, + "step": 25742 + }, + { + "epoch": 0.99, + "learning_rate": 3.352590407796871e-05, + "loss": 0.5376, + "step": 25743 + }, + { + "epoch": 0.99, + "learning_rate": 3.3525262887919985e-05, + "loss": 0.5058, + "step": 25744 + }, + { + "epoch": 0.99, + "learning_rate": 3.352462169787125e-05, + "loss": 0.4325, + "step": 25745 + }, + { + "epoch": 0.99, + "learning_rate": 3.352398050782252e-05, + "loss": 0.4437, + "step": 25746 + }, + { + "epoch": 0.99, + "learning_rate": 3.3523339317773787e-05, + "loss": 0.4284, + "step": 25747 + }, + { + "epoch": 0.99, + "learning_rate": 3.352269812772506e-05, + "loss": 0.4806, + "step": 25748 + }, + { + "epoch": 0.99, + "learning_rate": 3.352205693767633e-05, + "loss": 0.471, + "step": 25749 + }, + { + "epoch": 0.99, + "learning_rate": 3.35214157476276e-05, + "loss": 0.4651, + "step": 25750 + }, + { + "epoch": 0.99, + "learning_rate": 3.352077455757887e-05, + "loss": 0.4657, + "step": 25751 + }, + { + "epoch": 0.99, + "learning_rate": 3.352013336753014e-05, + "loss": 0.4779, + "step": 25752 + }, + { + "epoch": 0.99, + "learning_rate": 3.351949217748141e-05, + "loss": 0.5089, + "step": 25753 + }, + { + "epoch": 0.99, + "learning_rate": 3.3518850987432675e-05, + "loss": 0.5289, + "step": 25754 + }, + { + "epoch": 0.99, + "learning_rate": 3.351820979738395e-05, + "loss": 0.5819, + "step": 25755 + }, + { + "epoch": 0.99, + "learning_rate": 3.351756860733521e-05, + "loss": 0.433, + "step": 25756 + }, + { + "epoch": 0.99, + "learning_rate": 3.351692741728648e-05, + "loss": 0.4857, + "step": 25757 + }, + { + "epoch": 0.99, + "learning_rate": 3.3516286227237755e-05, + "loss": 0.3904, + "step": 25758 + }, + { + "epoch": 0.99, + "learning_rate": 3.351564503718903e-05, + "loss": 0.59, + "step": 25759 + }, + { + "epoch": 0.99, + "learning_rate": 3.351500384714029e-05, + "loss": 0.483, + "step": 25760 + }, + { + "epoch": 0.99, + "learning_rate": 3.3514362657091563e-05, + "loss": 0.478, + "step": 25761 + }, + { + "epoch": 0.99, + "learning_rate": 3.3513721467042835e-05, + "loss": 0.5479, + "step": 25762 + }, + { + "epoch": 0.99, + "learning_rate": 3.35130802769941e-05, + "loss": 0.4812, + "step": 25763 + }, + { + "epoch": 0.99, + "learning_rate": 3.351243908694537e-05, + "loss": 0.4724, + "step": 25764 + }, + { + "epoch": 0.99, + "learning_rate": 3.3511797896896644e-05, + "loss": 0.4387, + "step": 25765 + }, + { + "epoch": 0.99, + "learning_rate": 3.3511156706847915e-05, + "loss": 0.5653, + "step": 25766 + }, + { + "epoch": 0.99, + "learning_rate": 3.351051551679918e-05, + "loss": 0.3517, + "step": 25767 + }, + { + "epoch": 0.99, + "learning_rate": 3.350987432675045e-05, + "loss": 0.4906, + "step": 25768 + }, + { + "epoch": 0.99, + "learning_rate": 3.350923313670172e-05, + "loss": 0.597, + "step": 25769 + }, + { + "epoch": 0.99, + "learning_rate": 3.350859194665299e-05, + "loss": 0.4415, + "step": 25770 + }, + { + "epoch": 0.99, + "learning_rate": 3.350795075660426e-05, + "loss": 0.5182, + "step": 25771 + }, + { + "epoch": 0.99, + "learning_rate": 3.3507309566555525e-05, + "loss": 0.4243, + "step": 25772 + }, + { + "epoch": 0.99, + "learning_rate": 3.35066683765068e-05, + "loss": 0.7354, + "step": 25773 + }, + { + "epoch": 0.99, + "learning_rate": 3.350602718645807e-05, + "loss": 0.4355, + "step": 25774 + }, + { + "epoch": 0.99, + "learning_rate": 3.350538599640934e-05, + "loss": 0.5678, + "step": 25775 + }, + { + "epoch": 0.99, + "learning_rate": 3.3504744806360605e-05, + "loss": 0.5036, + "step": 25776 + }, + { + "epoch": 0.99, + "learning_rate": 3.350410361631188e-05, + "loss": 0.5559, + "step": 25777 + }, + { + "epoch": 0.99, + "learning_rate": 3.350346242626314e-05, + "loss": 0.5225, + "step": 25778 + }, + { + "epoch": 0.99, + "learning_rate": 3.3502821236214414e-05, + "loss": 0.5115, + "step": 25779 + }, + { + "epoch": 0.99, + "learning_rate": 3.3502180046165685e-05, + "loss": 0.4368, + "step": 25780 + }, + { + "epoch": 0.99, + "learning_rate": 3.350153885611696e-05, + "loss": 0.5701, + "step": 25781 + }, + { + "epoch": 0.99, + "learning_rate": 3.350089766606823e-05, + "loss": 0.5063, + "step": 25782 + }, + { + "epoch": 0.99, + "learning_rate": 3.3500256476019494e-05, + "loss": 0.4685, + "step": 25783 + }, + { + "epoch": 0.99, + "learning_rate": 3.3499615285970766e-05, + "loss": 0.4487, + "step": 25784 + }, + { + "epoch": 0.99, + "learning_rate": 3.349897409592203e-05, + "loss": 0.6072, + "step": 25785 + }, + { + "epoch": 0.99, + "learning_rate": 3.34983329058733e-05, + "loss": 0.4761, + "step": 25786 + }, + { + "epoch": 0.99, + "learning_rate": 3.349769171582457e-05, + "loss": 0.4039, + "step": 25787 + }, + { + "epoch": 0.99, + "learning_rate": 3.349705052577584e-05, + "loss": 0.5086, + "step": 25788 + }, + { + "epoch": 0.99, + "learning_rate": 3.349640933572711e-05, + "loss": 0.5269, + "step": 25789 + }, + { + "epoch": 0.99, + "learning_rate": 3.349576814567838e-05, + "loss": 0.4409, + "step": 25790 + }, + { + "epoch": 0.99, + "learning_rate": 3.3495126955629654e-05, + "loss": 0.5412, + "step": 25791 + }, + { + "epoch": 0.99, + "learning_rate": 3.349448576558092e-05, + "loss": 0.4749, + "step": 25792 + }, + { + "epoch": 0.99, + "learning_rate": 3.349384457553219e-05, + "loss": 0.4837, + "step": 25793 + }, + { + "epoch": 0.99, + "learning_rate": 3.3493203385483456e-05, + "loss": 0.6548, + "step": 25794 + }, + { + "epoch": 0.99, + "learning_rate": 3.349256219543473e-05, + "loss": 0.4855, + "step": 25795 + }, + { + "epoch": 0.99, + "learning_rate": 3.349192100538599e-05, + "loss": 0.4757, + "step": 25796 + }, + { + "epoch": 0.99, + "learning_rate": 3.349127981533727e-05, + "loss": 0.336, + "step": 25797 + }, + { + "epoch": 0.99, + "learning_rate": 3.349063862528854e-05, + "loss": 0.4695, + "step": 25798 + }, + { + "epoch": 0.99, + "learning_rate": 3.348999743523981e-05, + "loss": 0.4756, + "step": 25799 + }, + { + "epoch": 0.99, + "learning_rate": 3.348935624519108e-05, + "loss": 0.5018, + "step": 25800 + }, + { + "epoch": 0.99, + "learning_rate": 3.3488715055142344e-05, + "loss": 0.546, + "step": 25801 + }, + { + "epoch": 0.99, + "learning_rate": 3.3488073865093616e-05, + "loss": 0.589, + "step": 25802 + }, + { + "epoch": 0.99, + "learning_rate": 3.348743267504488e-05, + "loss": 0.4939, + "step": 25803 + }, + { + "epoch": 0.99, + "learning_rate": 3.348679148499615e-05, + "loss": 0.5122, + "step": 25804 + }, + { + "epoch": 0.99, + "learning_rate": 3.3486150294947424e-05, + "loss": 0.4831, + "step": 25805 + }, + { + "epoch": 0.99, + "learning_rate": 3.3485509104898696e-05, + "loss": 0.4299, + "step": 25806 + }, + { + "epoch": 0.99, + "learning_rate": 3.348486791484997e-05, + "loss": 0.5469, + "step": 25807 + }, + { + "epoch": 0.99, + "learning_rate": 3.348422672480123e-05, + "loss": 0.4938, + "step": 25808 + }, + { + "epoch": 0.99, + "learning_rate": 3.3483585534752504e-05, + "loss": 0.5737, + "step": 25809 + }, + { + "epoch": 0.99, + "learning_rate": 3.348294434470377e-05, + "loss": 0.4388, + "step": 25810 + }, + { + "epoch": 0.99, + "learning_rate": 3.348230315465504e-05, + "loss": 0.5324, + "step": 25811 + }, + { + "epoch": 0.99, + "learning_rate": 3.3481661964606306e-05, + "loss": 0.4455, + "step": 25812 + }, + { + "epoch": 0.99, + "learning_rate": 3.3481020774557584e-05, + "loss": 0.4086, + "step": 25813 + }, + { + "epoch": 0.99, + "learning_rate": 3.348037958450885e-05, + "loss": 0.3942, + "step": 25814 + }, + { + "epoch": 0.99, + "learning_rate": 3.347973839446012e-05, + "loss": 0.4733, + "step": 25815 + }, + { + "epoch": 0.99, + "learning_rate": 3.347909720441139e-05, + "loss": 0.5041, + "step": 25816 + }, + { + "epoch": 0.99, + "learning_rate": 3.347845601436266e-05, + "loss": 0.6219, + "step": 25817 + }, + { + "epoch": 0.99, + "learning_rate": 3.347781482431393e-05, + "loss": 0.336, + "step": 25818 + }, + { + "epoch": 0.99, + "learning_rate": 3.3477173634265194e-05, + "loss": 0.4554, + "step": 25819 + }, + { + "epoch": 0.99, + "learning_rate": 3.3476532444216466e-05, + "loss": 0.5, + "step": 25820 + }, + { + "epoch": 0.99, + "learning_rate": 3.347589125416774e-05, + "loss": 0.595, + "step": 25821 + }, + { + "epoch": 0.99, + "learning_rate": 3.347525006411901e-05, + "loss": 0.4788, + "step": 25822 + }, + { + "epoch": 0.99, + "learning_rate": 3.3474608874070274e-05, + "loss": 0.5409, + "step": 25823 + }, + { + "epoch": 0.99, + "learning_rate": 3.3473967684021546e-05, + "loss": 0.4085, + "step": 25824 + }, + { + "epoch": 0.99, + "learning_rate": 3.347332649397282e-05, + "loss": 0.5516, + "step": 25825 + }, + { + "epoch": 0.99, + "learning_rate": 3.347268530392408e-05, + "loss": 0.4061, + "step": 25826 + }, + { + "epoch": 0.99, + "learning_rate": 3.3472044113875354e-05, + "loss": 0.5435, + "step": 25827 + }, + { + "epoch": 0.99, + "learning_rate": 3.347140292382662e-05, + "loss": 0.499, + "step": 25828 + }, + { + "epoch": 0.99, + "learning_rate": 3.347076173377789e-05, + "loss": 0.3684, + "step": 25829 + }, + { + "epoch": 0.99, + "learning_rate": 3.347012054372916e-05, + "loss": 0.5068, + "step": 25830 + }, + { + "epoch": 0.99, + "learning_rate": 3.3469479353680435e-05, + "loss": 0.4673, + "step": 25831 + }, + { + "epoch": 0.99, + "learning_rate": 3.34688381636317e-05, + "loss": 0.4204, + "step": 25832 + }, + { + "epoch": 0.99, + "learning_rate": 3.346819697358297e-05, + "loss": 0.4926, + "step": 25833 + }, + { + "epoch": 0.99, + "learning_rate": 3.346755578353424e-05, + "loss": 0.517, + "step": 25834 + }, + { + "epoch": 0.99, + "learning_rate": 3.346691459348551e-05, + "loss": 0.4076, + "step": 25835 + }, + { + "epoch": 0.99, + "learning_rate": 3.346627340343678e-05, + "loss": 0.3819, + "step": 25836 + }, + { + "epoch": 0.99, + "learning_rate": 3.346563221338805e-05, + "loss": 0.4544, + "step": 25837 + }, + { + "epoch": 0.99, + "learning_rate": 3.346499102333932e-05, + "loss": 0.3871, + "step": 25838 + }, + { + "epoch": 0.99, + "learning_rate": 3.346434983329059e-05, + "loss": 0.5594, + "step": 25839 + }, + { + "epoch": 0.99, + "learning_rate": 3.346370864324186e-05, + "loss": 0.281, + "step": 25840 + }, + { + "epoch": 0.99, + "learning_rate": 3.3463067453193125e-05, + "loss": 0.4091, + "step": 25841 + }, + { + "epoch": 0.99, + "learning_rate": 3.3462426263144396e-05, + "loss": 0.4626, + "step": 25842 + }, + { + "epoch": 0.99, + "learning_rate": 3.346178507309567e-05, + "loss": 0.4323, + "step": 25843 + }, + { + "epoch": 0.99, + "learning_rate": 3.346114388304693e-05, + "loss": 0.5678, + "step": 25844 + }, + { + "epoch": 0.99, + "learning_rate": 3.3460502692998205e-05, + "loss": 0.6113, + "step": 25845 + }, + { + "epoch": 0.99, + "learning_rate": 3.3459861502949476e-05, + "loss": 0.4142, + "step": 25846 + }, + { + "epoch": 0.99, + "learning_rate": 3.345922031290075e-05, + "loss": 0.4645, + "step": 25847 + }, + { + "epoch": 0.99, + "learning_rate": 3.345857912285201e-05, + "loss": 0.4012, + "step": 25848 + }, + { + "epoch": 0.99, + "learning_rate": 3.3457937932803285e-05, + "loss": 0.4823, + "step": 25849 + }, + { + "epoch": 0.99, + "learning_rate": 3.345729674275455e-05, + "loss": 0.6323, + "step": 25850 + }, + { + "epoch": 0.99, + "learning_rate": 3.345665555270582e-05, + "loss": 0.4429, + "step": 25851 + }, + { + "epoch": 0.99, + "learning_rate": 3.345601436265709e-05, + "loss": 0.4456, + "step": 25852 + }, + { + "epoch": 0.99, + "learning_rate": 3.3455373172608365e-05, + "loss": 0.4809, + "step": 25853 + }, + { + "epoch": 0.99, + "learning_rate": 3.3454731982559637e-05, + "loss": 0.3457, + "step": 25854 + }, + { + "epoch": 0.99, + "learning_rate": 3.34540907925109e-05, + "loss": 0.5741, + "step": 25855 + }, + { + "epoch": 0.99, + "learning_rate": 3.345344960246217e-05, + "loss": 0.3799, + "step": 25856 + }, + { + "epoch": 0.99, + "learning_rate": 3.345280841241344e-05, + "loss": 0.5749, + "step": 25857 + }, + { + "epoch": 0.99, + "learning_rate": 3.345216722236471e-05, + "loss": 0.4459, + "step": 25858 + }, + { + "epoch": 0.99, + "learning_rate": 3.3451526032315975e-05, + "loss": 0.4078, + "step": 25859 + }, + { + "epoch": 0.99, + "learning_rate": 3.3450884842267247e-05, + "loss": 0.5159, + "step": 25860 + }, + { + "epoch": 0.99, + "learning_rate": 3.345024365221852e-05, + "loss": 0.4133, + "step": 25861 + }, + { + "epoch": 0.99, + "learning_rate": 3.344960246216979e-05, + "loss": 0.4483, + "step": 25862 + }, + { + "epoch": 0.99, + "learning_rate": 3.344896127212106e-05, + "loss": 0.5526, + "step": 25863 + }, + { + "epoch": 0.99, + "learning_rate": 3.344832008207233e-05, + "loss": 0.3414, + "step": 25864 + }, + { + "epoch": 0.99, + "learning_rate": 3.34476788920236e-05, + "loss": 0.4045, + "step": 25865 + }, + { + "epoch": 0.99, + "learning_rate": 3.344703770197486e-05, + "loss": 0.498, + "step": 25866 + }, + { + "epoch": 0.99, + "learning_rate": 3.3446396511926135e-05, + "loss": 0.4104, + "step": 25867 + }, + { + "epoch": 0.99, + "learning_rate": 3.34457553218774e-05, + "loss": 0.5295, + "step": 25868 + }, + { + "epoch": 0.99, + "learning_rate": 3.344511413182868e-05, + "loss": 0.3848, + "step": 25869 + }, + { + "epoch": 0.99, + "learning_rate": 3.344447294177995e-05, + "loss": 0.4527, + "step": 25870 + }, + { + "epoch": 0.99, + "learning_rate": 3.3443831751731215e-05, + "loss": 0.3998, + "step": 25871 + }, + { + "epoch": 0.99, + "learning_rate": 3.344319056168249e-05, + "loss": 0.5074, + "step": 25872 + }, + { + "epoch": 0.99, + "learning_rate": 3.344254937163375e-05, + "loss": 0.5617, + "step": 25873 + }, + { + "epoch": 0.99, + "learning_rate": 3.3441908181585024e-05, + "loss": 0.5039, + "step": 25874 + }, + { + "epoch": 0.99, + "learning_rate": 3.344126699153629e-05, + "loss": 0.4015, + "step": 25875 + }, + { + "epoch": 0.99, + "learning_rate": 3.344062580148756e-05, + "loss": 0.4781, + "step": 25876 + }, + { + "epoch": 0.99, + "learning_rate": 3.343998461143883e-05, + "loss": 0.6236, + "step": 25877 + }, + { + "epoch": 0.99, + "learning_rate": 3.3439343421390104e-05, + "loss": 0.4664, + "step": 25878 + }, + { + "epoch": 0.99, + "learning_rate": 3.3438702231341375e-05, + "loss": 0.5582, + "step": 25879 + }, + { + "epoch": 1.0, + "learning_rate": 3.343806104129264e-05, + "loss": 0.4997, + "step": 25880 + }, + { + "epoch": 1.0, + "learning_rate": 3.343741985124391e-05, + "loss": 0.4522, + "step": 25881 + }, + { + "epoch": 1.0, + "learning_rate": 3.343677866119518e-05, + "loss": 0.4954, + "step": 25882 + }, + { + "epoch": 1.0, + "learning_rate": 3.343613747114645e-05, + "loss": 0.3097, + "step": 25883 + }, + { + "epoch": 1.0, + "learning_rate": 3.3435496281097714e-05, + "loss": 0.5275, + "step": 25884 + }, + { + "epoch": 1.0, + "learning_rate": 3.343485509104899e-05, + "loss": 0.4886, + "step": 25885 + }, + { + "epoch": 1.0, + "learning_rate": 3.3434213901000264e-05, + "loss": 0.4949, + "step": 25886 + }, + { + "epoch": 1.0, + "learning_rate": 3.343357271095153e-05, + "loss": 0.4406, + "step": 25887 + }, + { + "epoch": 1.0, + "learning_rate": 3.34329315209028e-05, + "loss": 0.5978, + "step": 25888 + }, + { + "epoch": 1.0, + "learning_rate": 3.3432290330854065e-05, + "loss": 0.5106, + "step": 25889 + }, + { + "epoch": 1.0, + "learning_rate": 3.343164914080534e-05, + "loss": 0.4331, + "step": 25890 + }, + { + "epoch": 1.0, + "learning_rate": 3.34310079507566e-05, + "loss": 0.4236, + "step": 25891 + }, + { + "epoch": 1.0, + "learning_rate": 3.3430366760707874e-05, + "loss": 0.4545, + "step": 25892 + }, + { + "epoch": 1.0, + "learning_rate": 3.3429725570659145e-05, + "loss": 0.4572, + "step": 25893 + }, + { + "epoch": 1.0, + "learning_rate": 3.342908438061042e-05, + "loss": 0.5276, + "step": 25894 + }, + { + "epoch": 1.0, + "learning_rate": 3.342844319056169e-05, + "loss": 0.4715, + "step": 25895 + }, + { + "epoch": 1.0, + "learning_rate": 3.3427802000512954e-05, + "loss": 0.4739, + "step": 25896 + }, + { + "epoch": 1.0, + "learning_rate": 3.3427160810464226e-05, + "loss": 0.4971, + "step": 25897 + }, + { + "epoch": 1.0, + "learning_rate": 3.342651962041549e-05, + "loss": 0.4111, + "step": 25898 + }, + { + "epoch": 1.0, + "learning_rate": 3.342587843036676e-05, + "loss": 0.5114, + "step": 25899 + }, + { + "epoch": 1.0, + "learning_rate": 3.342523724031803e-05, + "loss": 0.5077, + "step": 25900 + }, + { + "epoch": 1.0, + "learning_rate": 3.3424596050269306e-05, + "loss": 0.353, + "step": 25901 + }, + { + "epoch": 1.0, + "learning_rate": 3.342395486022057e-05, + "loss": 0.5426, + "step": 25902 + }, + { + "epoch": 1.0, + "learning_rate": 3.342331367017184e-05, + "loss": 0.4349, + "step": 25903 + }, + { + "epoch": 1.0, + "learning_rate": 3.3422672480123114e-05, + "loss": 0.5108, + "step": 25904 + }, + { + "epoch": 1.0, + "learning_rate": 3.342203129007438e-05, + "loss": 0.4587, + "step": 25905 + }, + { + "epoch": 1.0, + "learning_rate": 3.342139010002565e-05, + "loss": 0.5464, + "step": 25906 + }, + { + "epoch": 1.0, + "learning_rate": 3.3420748909976916e-05, + "loss": 0.5012, + "step": 25907 + }, + { + "epoch": 1.0, + "learning_rate": 3.342010771992819e-05, + "loss": 0.4069, + "step": 25908 + }, + { + "epoch": 1.0, + "learning_rate": 3.341946652987946e-05, + "loss": 0.4376, + "step": 25909 + }, + { + "epoch": 1.0, + "learning_rate": 3.341882533983073e-05, + "loss": 0.4301, + "step": 25910 + }, + { + "epoch": 1.0, + "learning_rate": 3.3418184149781996e-05, + "loss": 0.4309, + "step": 25911 + }, + { + "epoch": 1.0, + "learning_rate": 3.341754295973327e-05, + "loss": 0.4733, + "step": 25912 + }, + { + "epoch": 1.0, + "learning_rate": 3.341690176968454e-05, + "loss": 0.4759, + "step": 25913 + }, + { + "epoch": 1.0, + "learning_rate": 3.3416260579635804e-05, + "loss": 0.4754, + "step": 25914 + }, + { + "epoch": 1.0, + "learning_rate": 3.3415619389587076e-05, + "loss": 0.4511, + "step": 25915 + }, + { + "epoch": 1.0, + "learning_rate": 3.341497819953834e-05, + "loss": 0.5945, + "step": 25916 + }, + { + "epoch": 1.0, + "learning_rate": 3.341433700948961e-05, + "loss": 0.4678, + "step": 25917 + }, + { + "epoch": 1.0, + "learning_rate": 3.3413695819440884e-05, + "loss": 0.4646, + "step": 25918 + }, + { + "epoch": 1.0, + "learning_rate": 3.3413054629392156e-05, + "loss": 0.5248, + "step": 25919 + }, + { + "epoch": 1.0, + "learning_rate": 3.341241343934342e-05, + "loss": 0.5231, + "step": 25920 + }, + { + "epoch": 1.0, + "learning_rate": 3.341177224929469e-05, + "loss": 0.436, + "step": 25921 + }, + { + "epoch": 1.0, + "learning_rate": 3.3411131059245964e-05, + "loss": 0.4001, + "step": 25922 + }, + { + "epoch": 1.0, + "learning_rate": 3.341048986919723e-05, + "loss": 0.4447, + "step": 25923 + }, + { + "epoch": 1.0, + "learning_rate": 3.34098486791485e-05, + "loss": 0.5235, + "step": 25924 + }, + { + "epoch": 1.0, + "learning_rate": 3.340920748909977e-05, + "loss": 0.3627, + "step": 25925 + }, + { + "epoch": 1.0, + "learning_rate": 3.3408566299051044e-05, + "loss": 0.4043, + "step": 25926 + }, + { + "epoch": 1.0, + "learning_rate": 3.340792510900231e-05, + "loss": 0.4972, + "step": 25927 + }, + { + "epoch": 1.0, + "learning_rate": 3.340728391895358e-05, + "loss": 0.5997, + "step": 25928 + }, + { + "epoch": 1.0, + "learning_rate": 3.3406642728904846e-05, + "loss": 0.5966, + "step": 25929 + }, + { + "epoch": 1.0, + "learning_rate": 3.340600153885612e-05, + "loss": 0.5955, + "step": 25930 + }, + { + "epoch": 1.0, + "learning_rate": 3.340536034880739e-05, + "loss": 0.5346, + "step": 25931 + }, + { + "epoch": 1.0, + "learning_rate": 3.3404719158758654e-05, + "loss": 0.4382, + "step": 25932 + }, + { + "epoch": 1.0, + "learning_rate": 3.3404077968709926e-05, + "loss": 0.4973, + "step": 25933 + }, + { + "epoch": 1.0, + "learning_rate": 3.34034367786612e-05, + "loss": 0.5414, + "step": 25934 + }, + { + "epoch": 1.0, + "learning_rate": 3.340279558861247e-05, + "loss": 0.584, + "step": 25935 + }, + { + "epoch": 1.0, + "learning_rate": 3.3402154398563734e-05, + "loss": 0.5359, + "step": 25936 + }, + { + "epoch": 1.0, + "learning_rate": 3.3401513208515006e-05, + "loss": 0.414, + "step": 25937 + }, + { + "epoch": 1.0, + "learning_rate": 3.340087201846627e-05, + "loss": 0.5887, + "step": 25938 + }, + { + "epoch": 1.0, + "learning_rate": 3.340023082841754e-05, + "loss": 0.4098, + "step": 25939 + }, + { + "epoch": 1.0, + "learning_rate": 3.3399589638368814e-05, + "loss": 0.5431, + "step": 25940 + }, + { + "epoch": 1.0, + "learning_rate": 3.3398948448320086e-05, + "loss": 0.4309, + "step": 25941 + }, + { + "epoch": 1.0, + "learning_rate": 3.339830725827136e-05, + "loss": 0.3694, + "step": 25942 + }, + { + "epoch": 1.0, + "learning_rate": 3.339766606822262e-05, + "loss": 0.5393, + "step": 25943 + }, + { + "epoch": 1.0, + "learning_rate": 3.3397024878173895e-05, + "loss": 0.541, + "step": 25944 + }, + { + "epoch": 1.0, + "learning_rate": 3.339638368812516e-05, + "loss": 0.4894, + "step": 25945 + }, + { + "epoch": 1.0, + "learning_rate": 3.339574249807643e-05, + "loss": 0.4596, + "step": 25946 + }, + { + "epoch": 1.0, + "learning_rate": 3.3395101308027696e-05, + "loss": 0.4026, + "step": 25947 + }, + { + "epoch": 1.0, + "learning_rate": 3.339446011797897e-05, + "loss": 0.4658, + "step": 25948 + }, + { + "epoch": 1.0, + "learning_rate": 3.339381892793024e-05, + "loss": 0.4899, + "step": 25949 + }, + { + "epoch": 1.0, + "learning_rate": 3.339317773788151e-05, + "loss": 0.5362, + "step": 25950 + }, + { + "epoch": 1.0, + "learning_rate": 3.339253654783278e-05, + "loss": 0.5493, + "step": 25951 + }, + { + "epoch": 1.0, + "learning_rate": 3.339189535778405e-05, + "loss": 0.5043, + "step": 25952 + }, + { + "epoch": 1.0, + "learning_rate": 3.339125416773532e-05, + "loss": 0.3974, + "step": 25953 + }, + { + "epoch": 1.0, + "learning_rate": 3.3390612977686585e-05, + "loss": 0.4951, + "step": 25954 + }, + { + "epoch": 1.0, + "learning_rate": 3.3389971787637856e-05, + "loss": 0.4489, + "step": 25955 + }, + { + "epoch": 1.0, + "learning_rate": 3.338933059758912e-05, + "loss": 0.4151, + "step": 25956 + }, + { + "epoch": 1.0, + "learning_rate": 3.33886894075404e-05, + "loss": 0.6462, + "step": 25957 + }, + { + "epoch": 1.0, + "learning_rate": 3.338804821749167e-05, + "loss": 0.4915, + "step": 25958 + }, + { + "epoch": 1.0, + "learning_rate": 3.3387407027442936e-05, + "loss": 0.5255, + "step": 25959 + }, + { + "epoch": 1.0, + "learning_rate": 3.338676583739421e-05, + "loss": 0.3522, + "step": 25960 + }, + { + "epoch": 1.0, + "learning_rate": 3.338612464734547e-05, + "loss": 0.4539, + "step": 25961 + }, + { + "epoch": 1.0, + "learning_rate": 3.3385483457296745e-05, + "loss": 0.4514, + "step": 25962 + }, + { + "epoch": 1.0, + "learning_rate": 3.338484226724801e-05, + "loss": 0.4128, + "step": 25963 + }, + { + "epoch": 1.0, + "learning_rate": 3.338420107719928e-05, + "loss": 0.3999, + "step": 25964 + }, + { + "epoch": 1.0, + "learning_rate": 3.338355988715055e-05, + "loss": 0.4969, + "step": 25965 + }, + { + "epoch": 1.0, + "learning_rate": 3.3382918697101825e-05, + "loss": 0.6494, + "step": 25966 + }, + { + "epoch": 1.0, + "learning_rate": 3.33822775070531e-05, + "loss": 0.5017, + "step": 25967 + }, + { + "epoch": 1.0, + "learning_rate": 3.338163631700436e-05, + "loss": 0.5236, + "step": 25968 + }, + { + "epoch": 1.0, + "learning_rate": 3.338099512695563e-05, + "loss": 0.5284, + "step": 25969 + }, + { + "epoch": 1.0, + "learning_rate": 3.33803539369069e-05, + "loss": 0.4898, + "step": 25970 + }, + { + "epoch": 1.0, + "learning_rate": 3.337971274685817e-05, + "loss": 0.4122, + "step": 25971 + }, + { + "epoch": 1.0, + "learning_rate": 3.3379071556809435e-05, + "loss": 0.4746, + "step": 25972 + }, + { + "epoch": 1.0, + "learning_rate": 3.337843036676071e-05, + "loss": 0.3761, + "step": 25973 + }, + { + "epoch": 1.0, + "learning_rate": 3.337778917671198e-05, + "loss": 0.4609, + "step": 25974 + }, + { + "epoch": 1.0, + "learning_rate": 3.337714798666325e-05, + "loss": 0.4074, + "step": 25975 + }, + { + "epoch": 1.0, + "learning_rate": 3.337650679661452e-05, + "loss": 0.4322, + "step": 25976 + }, + { + "epoch": 1.0, + "learning_rate": 3.337586560656579e-05, + "loss": 0.4587, + "step": 25977 + }, + { + "epoch": 1.0, + "learning_rate": 3.337522441651706e-05, + "loss": 0.445, + "step": 25978 + }, + { + "epoch": 1.0, + "learning_rate": 3.337458322646832e-05, + "loss": 0.4328, + "step": 25979 + }, + { + "epoch": 1.0, + "learning_rate": 3.3373942036419595e-05, + "loss": 0.4853, + "step": 25980 + }, + { + "epoch": 1.0, + "learning_rate": 3.337330084637087e-05, + "loss": 0.4386, + "step": 25981 + }, + { + "epoch": 1.0, + "learning_rate": 3.337265965632214e-05, + "loss": 0.4172, + "step": 25982 + }, + { + "epoch": 1.0, + "learning_rate": 3.3372018466273403e-05, + "loss": 0.4833, + "step": 25983 + }, + { + "epoch": 1.0, + "learning_rate": 3.3371377276224675e-05, + "loss": 0.4601, + "step": 25984 + }, + { + "epoch": 1.0, + "learning_rate": 3.337073608617595e-05, + "loss": 0.4269, + "step": 25985 + }, + { + "epoch": 1.0, + "learning_rate": 3.337009489612721e-05, + "loss": 0.4914, + "step": 25986 + }, + { + "epoch": 1.0, + "learning_rate": 3.3369453706078484e-05, + "loss": 0.473, + "step": 25987 + }, + { + "epoch": 1.0, + "learning_rate": 3.336881251602975e-05, + "loss": 0.4426, + "step": 25988 + }, + { + "epoch": 1.0, + "learning_rate": 3.336817132598102e-05, + "loss": 0.5226, + "step": 25989 + }, + { + "epoch": 1.0, + "learning_rate": 3.336753013593229e-05, + "loss": 0.5545, + "step": 25990 + }, + { + "epoch": 1.0, + "learning_rate": 3.3366888945883564e-05, + "loss": 0.5614, + "step": 25991 + }, + { + "epoch": 1.0, + "learning_rate": 3.336624775583483e-05, + "loss": 0.3646, + "step": 25992 + }, + { + "epoch": 1.0, + "learning_rate": 3.33656065657861e-05, + "loss": 0.432, + "step": 25993 + }, + { + "epoch": 1.0, + "learning_rate": 3.336496537573737e-05, + "loss": 0.4132, + "step": 25994 + }, + { + "epoch": 1.0, + "learning_rate": 3.336432418568864e-05, + "loss": 0.4139, + "step": 25995 + }, + { + "epoch": 1.0, + "learning_rate": 3.336368299563991e-05, + "loss": 0.5048, + "step": 25996 + }, + { + "epoch": 1.0, + "learning_rate": 3.336304180559118e-05, + "loss": 0.3742, + "step": 25997 + }, + { + "epoch": 1.0, + "learning_rate": 3.336240061554245e-05, + "loss": 0.4237, + "step": 25998 + }, + { + "epoch": 1.0, + "learning_rate": 3.336175942549372e-05, + "loss": 0.44, + "step": 25999 + }, + { + "epoch": 1.0, + "learning_rate": 3.336111823544499e-05, + "loss": 0.4601, + "step": 26000 + }, + { + "epoch": 1.0, + "learning_rate": 3.3360477045396254e-05, + "loss": 0.4061, + "step": 26001 + }, + { + "epoch": 1.0, + "learning_rate": 3.3359835855347525e-05, + "loss": 0.4697, + "step": 26002 + }, + { + "epoch": 1.0, + "learning_rate": 3.33591946652988e-05, + "loss": 0.605, + "step": 26003 + }, + { + "epoch": 1.0, + "learning_rate": 3.335855347525006e-05, + "loss": 0.4035, + "step": 26004 + }, + { + "epoch": 1.0, + "learning_rate": 3.3357912285201334e-05, + "loss": 0.4186, + "step": 26005 + }, + { + "epoch": 1.0, + "learning_rate": 3.3357271095152605e-05, + "loss": 0.415, + "step": 26006 + }, + { + "epoch": 1.0, + "learning_rate": 3.335662990510388e-05, + "loss": 0.4975, + "step": 26007 + }, + { + "epoch": 1.0, + "learning_rate": 3.335598871505514e-05, + "loss": 0.4009, + "step": 26008 + }, + { + "epoch": 1.0, + "learning_rate": 3.3355347525006414e-05, + "loss": 0.3853, + "step": 26009 + }, + { + "epoch": 1.0, + "learning_rate": 3.335470633495768e-05, + "loss": 0.4749, + "step": 26010 + }, + { + "epoch": 1.0, + "eval_accuracy": 0.7866152119742694, + "eval_loss": 0.45247501134872437, + "eval_runtime": 262.4569, + "eval_samples_per_second": 1585.628, + "eval_steps_per_second": 24.777, + "step": 26010 + }, + { + "epoch": 1.0, + "learning_rate": 3.335406514490895e-05, + "loss": 0.4162, + "step": 26011 + }, + { + "epoch": 1.0, + "learning_rate": 3.335342395486022e-05, + "loss": 0.4573, + "step": 26012 + }, + { + "epoch": 1.0, + "learning_rate": 3.3352782764811494e-05, + "loss": 0.4374, + "step": 26013 + }, + { + "epoch": 1.0, + "learning_rate": 3.3352141574762766e-05, + "loss": 0.3372, + "step": 26014 + }, + { + "epoch": 1.0, + "learning_rate": 3.335150038471403e-05, + "loss": 0.3033, + "step": 26015 + }, + { + "epoch": 1.0, + "learning_rate": 3.33508591946653e-05, + "loss": 0.4528, + "step": 26016 + }, + { + "epoch": 1.0, + "learning_rate": 3.335021800461657e-05, + "loss": 0.4604, + "step": 26017 + }, + { + "epoch": 1.0, + "learning_rate": 3.334957681456784e-05, + "loss": 0.3888, + "step": 26018 + }, + { + "epoch": 1.0, + "learning_rate": 3.3348935624519104e-05, + "loss": 0.4879, + "step": 26019 + }, + { + "epoch": 1.0, + "learning_rate": 3.3348294434470376e-05, + "loss": 0.4003, + "step": 26020 + }, + { + "epoch": 1.0, + "learning_rate": 3.334765324442165e-05, + "loss": 0.542, + "step": 26021 + }, + { + "epoch": 1.0, + "learning_rate": 3.334701205437292e-05, + "loss": 0.4212, + "step": 26022 + }, + { + "epoch": 1.0, + "learning_rate": 3.334637086432419e-05, + "loss": 0.4589, + "step": 26023 + }, + { + "epoch": 1.0, + "learning_rate": 3.3345729674275456e-05, + "loss": 0.3418, + "step": 26024 + }, + { + "epoch": 1.0, + "learning_rate": 3.334508848422673e-05, + "loss": 0.478, + "step": 26025 + }, + { + "epoch": 1.0, + "learning_rate": 3.334444729417799e-05, + "loss": 0.3811, + "step": 26026 + }, + { + "epoch": 1.0, + "learning_rate": 3.3343806104129264e-05, + "loss": 0.453, + "step": 26027 + }, + { + "epoch": 1.0, + "learning_rate": 3.3343164914080536e-05, + "loss": 0.456, + "step": 26028 + }, + { + "epoch": 1.0, + "learning_rate": 3.334252372403181e-05, + "loss": 0.4224, + "step": 26029 + }, + { + "epoch": 1.0, + "learning_rate": 3.334188253398308e-05, + "loss": 0.3282, + "step": 26030 + }, + { + "epoch": 1.0, + "learning_rate": 3.3341241343934344e-05, + "loss": 0.3732, + "step": 26031 + }, + { + "epoch": 1.0, + "learning_rate": 3.3340600153885616e-05, + "loss": 0.5047, + "step": 26032 + }, + { + "epoch": 1.0, + "learning_rate": 3.333995896383688e-05, + "loss": 0.3866, + "step": 26033 + }, + { + "epoch": 1.0, + "learning_rate": 3.333931777378815e-05, + "loss": 0.3813, + "step": 26034 + }, + { + "epoch": 1.0, + "learning_rate": 3.333867658373942e-05, + "loss": 0.319, + "step": 26035 + }, + { + "epoch": 1.0, + "learning_rate": 3.333803539369069e-05, + "loss": 0.321, + "step": 26036 + }, + { + "epoch": 1.0, + "learning_rate": 3.333739420364196e-05, + "loss": 0.5355, + "step": 26037 + }, + { + "epoch": 1.0, + "learning_rate": 3.333675301359323e-05, + "loss": 0.5341, + "step": 26038 + }, + { + "epoch": 1.0, + "learning_rate": 3.3336111823544504e-05, + "loss": 0.3793, + "step": 26039 + }, + { + "epoch": 1.0, + "learning_rate": 3.333547063349577e-05, + "loss": 0.3554, + "step": 26040 + }, + { + "epoch": 1.0, + "learning_rate": 3.333482944344704e-05, + "loss": 0.502, + "step": 26041 + }, + { + "epoch": 1.0, + "learning_rate": 3.3334188253398306e-05, + "loss": 0.4794, + "step": 26042 + }, + { + "epoch": 1.0, + "learning_rate": 3.333354706334958e-05, + "loss": 0.4477, + "step": 26043 + }, + { + "epoch": 1.0, + "learning_rate": 3.333290587330084e-05, + "loss": 0.3338, + "step": 26044 + }, + { + "epoch": 1.0, + "learning_rate": 3.333226468325212e-05, + "loss": 0.3367, + "step": 26045 + }, + { + "epoch": 1.0, + "learning_rate": 3.3331623493203386e-05, + "loss": 0.4162, + "step": 26046 + }, + { + "epoch": 1.0, + "learning_rate": 3.333098230315466e-05, + "loss": 0.356, + "step": 26047 + }, + { + "epoch": 1.0, + "learning_rate": 3.333034111310593e-05, + "loss": 0.3435, + "step": 26048 + }, + { + "epoch": 1.0, + "learning_rate": 3.3329699923057194e-05, + "loss": 0.39, + "step": 26049 + }, + { + "epoch": 1.0, + "learning_rate": 3.3329058733008466e-05, + "loss": 0.3843, + "step": 26050 + }, + { + "epoch": 1.0, + "learning_rate": 3.332841754295973e-05, + "loss": 0.3882, + "step": 26051 + }, + { + "epoch": 1.0, + "learning_rate": 3.3327776352911e-05, + "loss": 0.3942, + "step": 26052 + }, + { + "epoch": 1.0, + "learning_rate": 3.3327135162862275e-05, + "loss": 0.4317, + "step": 26053 + }, + { + "epoch": 1.0, + "learning_rate": 3.3326493972813546e-05, + "loss": 0.4885, + "step": 26054 + }, + { + "epoch": 1.0, + "learning_rate": 3.332585278276481e-05, + "loss": 0.4124, + "step": 26055 + }, + { + "epoch": 1.0, + "learning_rate": 3.332521159271608e-05, + "loss": 0.3336, + "step": 26056 + }, + { + "epoch": 1.0, + "learning_rate": 3.3324570402667355e-05, + "loss": 0.3271, + "step": 26057 + }, + { + "epoch": 1.0, + "learning_rate": 3.332392921261862e-05, + "loss": 0.4706, + "step": 26058 + }, + { + "epoch": 1.0, + "learning_rate": 3.332328802256989e-05, + "loss": 0.4606, + "step": 26059 + }, + { + "epoch": 1.0, + "learning_rate": 3.3322646832521156e-05, + "loss": 0.4178, + "step": 26060 + }, + { + "epoch": 1.0, + "learning_rate": 3.3322005642472435e-05, + "loss": 0.5614, + "step": 26061 + }, + { + "epoch": 1.0, + "learning_rate": 3.33213644524237e-05, + "loss": 0.4298, + "step": 26062 + }, + { + "epoch": 1.0, + "learning_rate": 3.332072326237497e-05, + "loss": 0.3586, + "step": 26063 + }, + { + "epoch": 1.0, + "learning_rate": 3.3320082072326236e-05, + "loss": 0.3667, + "step": 26064 + }, + { + "epoch": 1.0, + "learning_rate": 3.331944088227751e-05, + "loss": 0.2859, + "step": 26065 + }, + { + "epoch": 1.0, + "learning_rate": 3.331879969222878e-05, + "loss": 0.4209, + "step": 26066 + }, + { + "epoch": 1.0, + "learning_rate": 3.3318158502180045e-05, + "loss": 0.3922, + "step": 26067 + }, + { + "epoch": 1.0, + "learning_rate": 3.3317517312131316e-05, + "loss": 0.4601, + "step": 26068 + }, + { + "epoch": 1.0, + "learning_rate": 3.331687612208259e-05, + "loss": 0.4725, + "step": 26069 + }, + { + "epoch": 1.0, + "learning_rate": 3.331623493203386e-05, + "loss": 0.4761, + "step": 26070 + }, + { + "epoch": 1.0, + "learning_rate": 3.3315593741985125e-05, + "loss": 0.4308, + "step": 26071 + }, + { + "epoch": 1.0, + "learning_rate": 3.3314952551936396e-05, + "loss": 0.3908, + "step": 26072 + }, + { + "epoch": 1.0, + "learning_rate": 3.331431136188766e-05, + "loss": 0.4684, + "step": 26073 + }, + { + "epoch": 1.0, + "learning_rate": 3.331367017183893e-05, + "loss": 0.4467, + "step": 26074 + }, + { + "epoch": 1.0, + "learning_rate": 3.3313028981790205e-05, + "loss": 0.4551, + "step": 26075 + }, + { + "epoch": 1.0, + "learning_rate": 3.331238779174147e-05, + "loss": 0.4616, + "step": 26076 + }, + { + "epoch": 1.0, + "learning_rate": 3.331174660169274e-05, + "loss": 0.395, + "step": 26077 + }, + { + "epoch": 1.0, + "learning_rate": 3.331110541164401e-05, + "loss": 0.5785, + "step": 26078 + }, + { + "epoch": 1.0, + "learning_rate": 3.3310464221595285e-05, + "loss": 0.405, + "step": 26079 + }, + { + "epoch": 1.0, + "learning_rate": 3.330982303154655e-05, + "loss": 0.3934, + "step": 26080 + }, + { + "epoch": 1.0, + "learning_rate": 3.330918184149782e-05, + "loss": 0.3561, + "step": 26081 + }, + { + "epoch": 1.0, + "learning_rate": 3.330854065144909e-05, + "loss": 0.4518, + "step": 26082 + }, + { + "epoch": 1.0, + "learning_rate": 3.330789946140036e-05, + "loss": 0.4654, + "step": 26083 + }, + { + "epoch": 1.0, + "learning_rate": 3.330725827135163e-05, + "loss": 0.6239, + "step": 26084 + }, + { + "epoch": 1.0, + "learning_rate": 3.33066170813029e-05, + "loss": 0.4843, + "step": 26085 + }, + { + "epoch": 1.0, + "learning_rate": 3.3305975891254173e-05, + "loss": 0.4971, + "step": 26086 + }, + { + "epoch": 1.0, + "learning_rate": 3.330533470120544e-05, + "loss": 0.3205, + "step": 26087 + }, + { + "epoch": 1.0, + "learning_rate": 3.330469351115671e-05, + "loss": 0.3124, + "step": 26088 + }, + { + "epoch": 1.0, + "learning_rate": 3.3304052321107975e-05, + "loss": 0.3515, + "step": 26089 + }, + { + "epoch": 1.0, + "learning_rate": 3.330341113105925e-05, + "loss": 0.4306, + "step": 26090 + }, + { + "epoch": 1.0, + "learning_rate": 3.330276994101052e-05, + "loss": 0.4362, + "step": 26091 + }, + { + "epoch": 1.0, + "learning_rate": 3.330212875096178e-05, + "loss": 0.5166, + "step": 26092 + }, + { + "epoch": 1.0, + "learning_rate": 3.3301487560913055e-05, + "loss": 0.4214, + "step": 26093 + }, + { + "epoch": 1.0, + "learning_rate": 3.330084637086433e-05, + "loss": 0.3561, + "step": 26094 + }, + { + "epoch": 1.0, + "learning_rate": 3.33002051808156e-05, + "loss": 0.4889, + "step": 26095 + }, + { + "epoch": 1.0, + "learning_rate": 3.3299563990766863e-05, + "loss": 0.3402, + "step": 26096 + }, + { + "epoch": 1.0, + "learning_rate": 3.3298922800718135e-05, + "loss": 0.4736, + "step": 26097 + }, + { + "epoch": 1.0, + "learning_rate": 3.32982816106694e-05, + "loss": 0.3183, + "step": 26098 + }, + { + "epoch": 1.0, + "learning_rate": 3.329764042062067e-05, + "loss": 0.4209, + "step": 26099 + }, + { + "epoch": 1.0, + "learning_rate": 3.3296999230571944e-05, + "loss": 0.3823, + "step": 26100 + }, + { + "epoch": 1.0, + "learning_rate": 3.3296358040523215e-05, + "loss": 0.3918, + "step": 26101 + }, + { + "epoch": 1.0, + "learning_rate": 3.329571685047449e-05, + "loss": 0.4206, + "step": 26102 + }, + { + "epoch": 1.0, + "learning_rate": 3.329507566042575e-05, + "loss": 0.4127, + "step": 26103 + }, + { + "epoch": 1.0, + "learning_rate": 3.3294434470377024e-05, + "loss": 0.4304, + "step": 26104 + }, + { + "epoch": 1.0, + "learning_rate": 3.329379328032829e-05, + "loss": 0.4635, + "step": 26105 + }, + { + "epoch": 1.0, + "learning_rate": 3.329315209027956e-05, + "loss": 0.4314, + "step": 26106 + }, + { + "epoch": 1.0, + "learning_rate": 3.3292510900230825e-05, + "loss": 0.3387, + "step": 26107 + }, + { + "epoch": 1.0, + "learning_rate": 3.32918697101821e-05, + "loss": 0.5293, + "step": 26108 + }, + { + "epoch": 1.0, + "learning_rate": 3.329122852013337e-05, + "loss": 0.4009, + "step": 26109 + }, + { + "epoch": 1.0, + "learning_rate": 3.329058733008464e-05, + "loss": 0.4445, + "step": 26110 + }, + { + "epoch": 1.0, + "learning_rate": 3.328994614003591e-05, + "loss": 0.4161, + "step": 26111 + }, + { + "epoch": 1.0, + "learning_rate": 3.328930494998718e-05, + "loss": 0.3748, + "step": 26112 + }, + { + "epoch": 1.0, + "learning_rate": 3.328866375993845e-05, + "loss": 0.4897, + "step": 26113 + }, + { + "epoch": 1.0, + "learning_rate": 3.3288022569889714e-05, + "loss": 0.4461, + "step": 26114 + }, + { + "epoch": 1.0, + "learning_rate": 3.3287381379840985e-05, + "loss": 0.2986, + "step": 26115 + }, + { + "epoch": 1.0, + "learning_rate": 3.328674018979225e-05, + "loss": 0.419, + "step": 26116 + }, + { + "epoch": 1.0, + "learning_rate": 3.328609899974353e-05, + "loss": 0.433, + "step": 26117 + }, + { + "epoch": 1.0, + "learning_rate": 3.32854578096948e-05, + "loss": 0.3568, + "step": 26118 + }, + { + "epoch": 1.0, + "learning_rate": 3.3284816619646066e-05, + "loss": 0.3632, + "step": 26119 + }, + { + "epoch": 1.0, + "learning_rate": 3.328417542959734e-05, + "loss": 0.4053, + "step": 26120 + }, + { + "epoch": 1.0, + "learning_rate": 3.32835342395486e-05, + "loss": 0.3564, + "step": 26121 + }, + { + "epoch": 1.0, + "learning_rate": 3.3282893049499874e-05, + "loss": 0.4566, + "step": 26122 + }, + { + "epoch": 1.0, + "learning_rate": 3.328225185945114e-05, + "loss": 0.5776, + "step": 26123 + }, + { + "epoch": 1.0, + "learning_rate": 3.328161066940241e-05, + "loss": 0.5166, + "step": 26124 + }, + { + "epoch": 1.0, + "learning_rate": 3.328096947935368e-05, + "loss": 0.5154, + "step": 26125 + }, + { + "epoch": 1.0, + "learning_rate": 3.3280328289304954e-05, + "loss": 0.4341, + "step": 26126 + }, + { + "epoch": 1.0, + "learning_rate": 3.3279687099256226e-05, + "loss": 0.3884, + "step": 26127 + }, + { + "epoch": 1.0, + "learning_rate": 3.327904590920749e-05, + "loss": 0.4704, + "step": 26128 + }, + { + "epoch": 1.0, + "learning_rate": 3.327840471915876e-05, + "loss": 0.4838, + "step": 26129 + }, + { + "epoch": 1.0, + "learning_rate": 3.327776352911003e-05, + "loss": 0.3693, + "step": 26130 + }, + { + "epoch": 1.0, + "learning_rate": 3.32771223390613e-05, + "loss": 0.4219, + "step": 26131 + }, + { + "epoch": 1.0, + "learning_rate": 3.3276481149012564e-05, + "loss": 0.581, + "step": 26132 + }, + { + "epoch": 1.0, + "learning_rate": 3.327583995896384e-05, + "loss": 0.4178, + "step": 26133 + }, + { + "epoch": 1.0, + "learning_rate": 3.327519876891511e-05, + "loss": 0.4309, + "step": 26134 + }, + { + "epoch": 1.0, + "learning_rate": 3.327455757886638e-05, + "loss": 0.4779, + "step": 26135 + }, + { + "epoch": 1.0, + "learning_rate": 3.327391638881765e-05, + "loss": 0.4311, + "step": 26136 + }, + { + "epoch": 1.0, + "learning_rate": 3.3273275198768916e-05, + "loss": 0.4016, + "step": 26137 + }, + { + "epoch": 1.0, + "learning_rate": 3.327263400872019e-05, + "loss": 0.5078, + "step": 26138 + }, + { + "epoch": 1.0, + "learning_rate": 3.327199281867145e-05, + "loss": 0.3837, + "step": 26139 + }, + { + "epoch": 1.0, + "learning_rate": 3.3271351628622724e-05, + "loss": 0.38, + "step": 26140 + }, + { + "epoch": 1.01, + "learning_rate": 3.3270710438573996e-05, + "loss": 0.4415, + "step": 26141 + }, + { + "epoch": 1.01, + "learning_rate": 3.327006924852527e-05, + "loss": 0.5654, + "step": 26142 + }, + { + "epoch": 1.01, + "learning_rate": 3.326942805847653e-05, + "loss": 0.4291, + "step": 26143 + }, + { + "epoch": 1.01, + "learning_rate": 3.3268786868427804e-05, + "loss": 0.3659, + "step": 26144 + }, + { + "epoch": 1.01, + "learning_rate": 3.3268145678379076e-05, + "loss": 0.2937, + "step": 26145 + }, + { + "epoch": 1.01, + "learning_rate": 3.326750448833034e-05, + "loss": 0.4169, + "step": 26146 + }, + { + "epoch": 1.01, + "learning_rate": 3.326686329828161e-05, + "loss": 0.3651, + "step": 26147 + }, + { + "epoch": 1.01, + "learning_rate": 3.326622210823288e-05, + "loss": 0.4593, + "step": 26148 + }, + { + "epoch": 1.01, + "learning_rate": 3.3265580918184156e-05, + "loss": 0.3511, + "step": 26149 + }, + { + "epoch": 1.01, + "learning_rate": 3.326493972813542e-05, + "loss": 0.428, + "step": 26150 + }, + { + "epoch": 1.01, + "learning_rate": 3.326429853808669e-05, + "loss": 0.559, + "step": 26151 + }, + { + "epoch": 1.01, + "learning_rate": 3.326365734803796e-05, + "loss": 0.434, + "step": 26152 + }, + { + "epoch": 1.01, + "learning_rate": 3.326301615798923e-05, + "loss": 0.4866, + "step": 26153 + }, + { + "epoch": 1.01, + "learning_rate": 3.32623749679405e-05, + "loss": 0.5441, + "step": 26154 + }, + { + "epoch": 1.01, + "learning_rate": 3.3261733777891766e-05, + "loss": 0.3292, + "step": 26155 + }, + { + "epoch": 1.01, + "learning_rate": 3.326109258784304e-05, + "loss": 0.3989, + "step": 26156 + }, + { + "epoch": 1.01, + "learning_rate": 3.326045139779431e-05, + "loss": 0.4149, + "step": 26157 + }, + { + "epoch": 1.01, + "learning_rate": 3.325981020774558e-05, + "loss": 0.3539, + "step": 26158 + }, + { + "epoch": 1.01, + "learning_rate": 3.3259169017696846e-05, + "loss": 0.3466, + "step": 26159 + }, + { + "epoch": 1.01, + "learning_rate": 3.325852782764812e-05, + "loss": 0.4294, + "step": 26160 + }, + { + "epoch": 1.01, + "learning_rate": 3.325788663759938e-05, + "loss": 0.3289, + "step": 26161 + }, + { + "epoch": 1.01, + "learning_rate": 3.3257245447550654e-05, + "loss": 0.388, + "step": 26162 + }, + { + "epoch": 1.01, + "learning_rate": 3.3256604257501926e-05, + "loss": 0.3744, + "step": 26163 + }, + { + "epoch": 1.01, + "learning_rate": 3.325596306745319e-05, + "loss": 0.4971, + "step": 26164 + }, + { + "epoch": 1.01, + "learning_rate": 3.325532187740446e-05, + "loss": 0.3829, + "step": 26165 + }, + { + "epoch": 1.01, + "learning_rate": 3.3254680687355735e-05, + "loss": 0.3814, + "step": 26166 + }, + { + "epoch": 1.01, + "learning_rate": 3.3254039497307006e-05, + "loss": 0.3873, + "step": 26167 + }, + { + "epoch": 1.01, + "learning_rate": 3.325339830725827e-05, + "loss": 0.4964, + "step": 26168 + }, + { + "epoch": 1.01, + "learning_rate": 3.325275711720954e-05, + "loss": 0.2943, + "step": 26169 + }, + { + "epoch": 1.01, + "learning_rate": 3.325211592716081e-05, + "loss": 0.3564, + "step": 26170 + }, + { + "epoch": 1.01, + "learning_rate": 3.325147473711208e-05, + "loss": 0.4337, + "step": 26171 + }, + { + "epoch": 1.01, + "learning_rate": 3.325083354706335e-05, + "loss": 0.3411, + "step": 26172 + }, + { + "epoch": 1.01, + "learning_rate": 3.325019235701462e-05, + "loss": 0.4432, + "step": 26173 + }, + { + "epoch": 1.01, + "learning_rate": 3.3249551166965895e-05, + "loss": 0.5359, + "step": 26174 + }, + { + "epoch": 1.01, + "learning_rate": 3.324890997691716e-05, + "loss": 0.4443, + "step": 26175 + }, + { + "epoch": 1.01, + "learning_rate": 3.324826878686843e-05, + "loss": 0.4438, + "step": 26176 + }, + { + "epoch": 1.01, + "learning_rate": 3.3247627596819696e-05, + "loss": 0.4781, + "step": 26177 + }, + { + "epoch": 1.01, + "learning_rate": 3.324698640677097e-05, + "loss": 0.3907, + "step": 26178 + }, + { + "epoch": 1.01, + "learning_rate": 3.324634521672223e-05, + "loss": 0.3853, + "step": 26179 + }, + { + "epoch": 1.01, + "learning_rate": 3.3245704026673505e-05, + "loss": 0.3656, + "step": 26180 + }, + { + "epoch": 1.01, + "learning_rate": 3.3245062836624776e-05, + "loss": 0.4419, + "step": 26181 + }, + { + "epoch": 1.01, + "learning_rate": 3.324442164657605e-05, + "loss": 0.3333, + "step": 26182 + }, + { + "epoch": 1.01, + "learning_rate": 3.324378045652732e-05, + "loss": 0.3804, + "step": 26183 + }, + { + "epoch": 1.01, + "learning_rate": 3.3243139266478585e-05, + "loss": 0.4032, + "step": 26184 + }, + { + "epoch": 1.01, + "learning_rate": 3.3242498076429857e-05, + "loss": 0.395, + "step": 26185 + }, + { + "epoch": 1.01, + "learning_rate": 3.324185688638112e-05, + "loss": 0.2834, + "step": 26186 + }, + { + "epoch": 1.01, + "learning_rate": 3.324121569633239e-05, + "loss": 0.454, + "step": 26187 + }, + { + "epoch": 1.01, + "learning_rate": 3.3240574506283665e-05, + "loss": 0.3177, + "step": 26188 + }, + { + "epoch": 1.01, + "learning_rate": 3.3239933316234937e-05, + "loss": 0.4506, + "step": 26189 + }, + { + "epoch": 1.01, + "learning_rate": 3.323929212618621e-05, + "loss": 0.431, + "step": 26190 + }, + { + "epoch": 1.01, + "learning_rate": 3.323865093613747e-05, + "loss": 0.4375, + "step": 26191 + }, + { + "epoch": 1.01, + "learning_rate": 3.3238009746088745e-05, + "loss": 0.5423, + "step": 26192 + }, + { + "epoch": 1.01, + "learning_rate": 3.323736855604001e-05, + "loss": 0.4208, + "step": 26193 + }, + { + "epoch": 1.01, + "learning_rate": 3.323672736599128e-05, + "loss": 0.4536, + "step": 26194 + }, + { + "epoch": 1.01, + "learning_rate": 3.3236086175942547e-05, + "loss": 0.4522, + "step": 26195 + }, + { + "epoch": 1.01, + "learning_rate": 3.323544498589382e-05, + "loss": 0.3144, + "step": 26196 + }, + { + "epoch": 1.01, + "learning_rate": 3.323480379584509e-05, + "loss": 0.3816, + "step": 26197 + }, + { + "epoch": 1.01, + "learning_rate": 3.323416260579636e-05, + "loss": 0.5344, + "step": 26198 + }, + { + "epoch": 1.01, + "learning_rate": 3.3233521415747633e-05, + "loss": 0.482, + "step": 26199 + }, + { + "epoch": 1.01, + "learning_rate": 3.32328802256989e-05, + "loss": 0.4938, + "step": 26200 + }, + { + "epoch": 1.01, + "learning_rate": 3.323223903565017e-05, + "loss": 0.4063, + "step": 26201 + }, + { + "epoch": 1.01, + "learning_rate": 3.3231597845601435e-05, + "loss": 0.3976, + "step": 26202 + }, + { + "epoch": 1.01, + "learning_rate": 3.323095665555271e-05, + "loss": 0.4698, + "step": 26203 + }, + { + "epoch": 1.01, + "learning_rate": 3.323031546550397e-05, + "loss": 0.4441, + "step": 26204 + }, + { + "epoch": 1.01, + "learning_rate": 3.322967427545525e-05, + "loss": 0.4691, + "step": 26205 + }, + { + "epoch": 1.01, + "learning_rate": 3.3229033085406515e-05, + "loss": 0.3708, + "step": 26206 + }, + { + "epoch": 1.01, + "learning_rate": 3.322839189535779e-05, + "loss": 0.424, + "step": 26207 + }, + { + "epoch": 1.01, + "learning_rate": 3.322775070530906e-05, + "loss": 0.3223, + "step": 26208 + }, + { + "epoch": 1.01, + "learning_rate": 3.3227109515260324e-05, + "loss": 0.466, + "step": 26209 + }, + { + "epoch": 1.01, + "learning_rate": 3.3226468325211595e-05, + "loss": 0.2798, + "step": 26210 + }, + { + "epoch": 1.01, + "learning_rate": 3.322582713516286e-05, + "loss": 0.4037, + "step": 26211 + }, + { + "epoch": 1.01, + "learning_rate": 3.322518594511413e-05, + "loss": 0.4566, + "step": 26212 + }, + { + "epoch": 1.01, + "learning_rate": 3.3224544755065404e-05, + "loss": 0.4272, + "step": 26213 + }, + { + "epoch": 1.01, + "learning_rate": 3.3223903565016675e-05, + "loss": 0.3617, + "step": 26214 + }, + { + "epoch": 1.01, + "learning_rate": 3.322326237496794e-05, + "loss": 0.4128, + "step": 26215 + }, + { + "epoch": 1.01, + "learning_rate": 3.322262118491921e-05, + "loss": 0.4025, + "step": 26216 + }, + { + "epoch": 1.01, + "learning_rate": 3.3221979994870484e-05, + "loss": 0.3224, + "step": 26217 + }, + { + "epoch": 1.01, + "learning_rate": 3.322133880482175e-05, + "loss": 0.4343, + "step": 26218 + }, + { + "epoch": 1.01, + "learning_rate": 3.322069761477302e-05, + "loss": 0.4476, + "step": 26219 + }, + { + "epoch": 1.01, + "learning_rate": 3.3220056424724285e-05, + "loss": 0.3986, + "step": 26220 + }, + { + "epoch": 1.01, + "learning_rate": 3.3219415234675564e-05, + "loss": 0.4432, + "step": 26221 + }, + { + "epoch": 1.01, + "learning_rate": 3.321877404462683e-05, + "loss": 0.3531, + "step": 26222 + }, + { + "epoch": 1.01, + "learning_rate": 3.32181328545781e-05, + "loss": 0.4435, + "step": 26223 + }, + { + "epoch": 1.01, + "learning_rate": 3.3217491664529365e-05, + "loss": 0.3695, + "step": 26224 + }, + { + "epoch": 1.01, + "learning_rate": 3.321685047448064e-05, + "loss": 0.3303, + "step": 26225 + }, + { + "epoch": 1.01, + "learning_rate": 3.321620928443191e-05, + "loss": 0.3769, + "step": 26226 + }, + { + "epoch": 1.01, + "learning_rate": 3.3215568094383174e-05, + "loss": 0.3863, + "step": 26227 + }, + { + "epoch": 1.01, + "learning_rate": 3.3214926904334445e-05, + "loss": 0.4838, + "step": 26228 + }, + { + "epoch": 1.01, + "learning_rate": 3.321428571428572e-05, + "loss": 0.3901, + "step": 26229 + }, + { + "epoch": 1.01, + "learning_rate": 3.321364452423699e-05, + "loss": 0.5906, + "step": 26230 + }, + { + "epoch": 1.01, + "learning_rate": 3.3213003334188254e-05, + "loss": 0.5071, + "step": 26231 + }, + { + "epoch": 1.01, + "learning_rate": 3.3212362144139526e-05, + "loss": 0.469, + "step": 26232 + }, + { + "epoch": 1.01, + "learning_rate": 3.321172095409079e-05, + "loss": 0.2525, + "step": 26233 + }, + { + "epoch": 1.01, + "learning_rate": 3.321107976404206e-05, + "loss": 0.4603, + "step": 26234 + }, + { + "epoch": 1.01, + "learning_rate": 3.3210438573993334e-05, + "loss": 0.4109, + "step": 26235 + }, + { + "epoch": 1.01, + "learning_rate": 3.32097973839446e-05, + "loss": 0.3544, + "step": 26236 + }, + { + "epoch": 1.01, + "learning_rate": 3.320915619389587e-05, + "loss": 0.554, + "step": 26237 + }, + { + "epoch": 1.01, + "learning_rate": 3.320851500384714e-05, + "loss": 0.5827, + "step": 26238 + }, + { + "epoch": 1.01, + "learning_rate": 3.3207873813798414e-05, + "loss": 0.3308, + "step": 26239 + }, + { + "epoch": 1.01, + "learning_rate": 3.320723262374968e-05, + "loss": 0.5753, + "step": 26240 + }, + { + "epoch": 1.01, + "learning_rate": 3.320659143370095e-05, + "loss": 0.3369, + "step": 26241 + }, + { + "epoch": 1.01, + "learning_rate": 3.3205950243652216e-05, + "loss": 0.4548, + "step": 26242 + }, + { + "epoch": 1.01, + "learning_rate": 3.320530905360349e-05, + "loss": 0.3973, + "step": 26243 + }, + { + "epoch": 1.01, + "learning_rate": 3.320466786355476e-05, + "loss": 0.3432, + "step": 26244 + }, + { + "epoch": 1.01, + "learning_rate": 3.320402667350603e-05, + "loss": 0.4163, + "step": 26245 + }, + { + "epoch": 1.01, + "learning_rate": 3.32033854834573e-05, + "loss": 0.4422, + "step": 26246 + }, + { + "epoch": 1.01, + "learning_rate": 3.320274429340857e-05, + "loss": 0.4657, + "step": 26247 + }, + { + "epoch": 1.01, + "learning_rate": 3.320210310335984e-05, + "loss": 0.4201, + "step": 26248 + }, + { + "epoch": 1.01, + "learning_rate": 3.3201461913311104e-05, + "loss": 0.4476, + "step": 26249 + }, + { + "epoch": 1.01, + "learning_rate": 3.3200820723262376e-05, + "loss": 0.4557, + "step": 26250 + }, + { + "epoch": 1.01, + "learning_rate": 3.320017953321364e-05, + "loss": 0.4144, + "step": 26251 + }, + { + "epoch": 1.01, + "learning_rate": 3.319953834316491e-05, + "loss": 0.4126, + "step": 26252 + }, + { + "epoch": 1.01, + "learning_rate": 3.3198897153116184e-05, + "loss": 0.411, + "step": 26253 + }, + { + "epoch": 1.01, + "learning_rate": 3.3198255963067456e-05, + "loss": 0.4764, + "step": 26254 + }, + { + "epoch": 1.01, + "learning_rate": 3.319761477301873e-05, + "loss": 0.4044, + "step": 26255 + }, + { + "epoch": 1.01, + "learning_rate": 3.319697358296999e-05, + "loss": 0.4515, + "step": 26256 + }, + { + "epoch": 1.01, + "learning_rate": 3.3196332392921264e-05, + "loss": 0.4372, + "step": 26257 + }, + { + "epoch": 1.01, + "learning_rate": 3.319569120287253e-05, + "loss": 0.3796, + "step": 26258 + }, + { + "epoch": 1.01, + "learning_rate": 3.31950500128238e-05, + "loss": 0.4354, + "step": 26259 + }, + { + "epoch": 1.01, + "learning_rate": 3.319440882277507e-05, + "loss": 0.2951, + "step": 26260 + }, + { + "epoch": 1.01, + "learning_rate": 3.3193767632726344e-05, + "loss": 0.3366, + "step": 26261 + }, + { + "epoch": 1.01, + "learning_rate": 3.3193126442677616e-05, + "loss": 0.4141, + "step": 26262 + }, + { + "epoch": 1.01, + "learning_rate": 3.319248525262888e-05, + "loss": 0.388, + "step": 26263 + }, + { + "epoch": 1.01, + "learning_rate": 3.319184406258015e-05, + "loss": 0.4114, + "step": 26264 + }, + { + "epoch": 1.01, + "learning_rate": 3.319120287253142e-05, + "loss": 0.276, + "step": 26265 + }, + { + "epoch": 1.01, + "learning_rate": 3.319056168248269e-05, + "loss": 0.3387, + "step": 26266 + }, + { + "epoch": 1.01, + "learning_rate": 3.3189920492433954e-05, + "loss": 0.3579, + "step": 26267 + }, + { + "epoch": 1.01, + "learning_rate": 3.3189279302385226e-05, + "loss": 0.511, + "step": 26268 + }, + { + "epoch": 1.01, + "learning_rate": 3.31886381123365e-05, + "loss": 0.5154, + "step": 26269 + }, + { + "epoch": 1.01, + "learning_rate": 3.318799692228777e-05, + "loss": 0.5002, + "step": 26270 + }, + { + "epoch": 1.01, + "learning_rate": 3.318735573223904e-05, + "loss": 0.4684, + "step": 26271 + }, + { + "epoch": 1.01, + "learning_rate": 3.3186714542190306e-05, + "loss": 0.5015, + "step": 26272 + }, + { + "epoch": 1.01, + "learning_rate": 3.318607335214158e-05, + "loss": 0.3579, + "step": 26273 + }, + { + "epoch": 1.01, + "learning_rate": 3.318543216209284e-05, + "loss": 0.5057, + "step": 26274 + }, + { + "epoch": 1.01, + "learning_rate": 3.3184790972044115e-05, + "loss": 0.3897, + "step": 26275 + }, + { + "epoch": 1.01, + "learning_rate": 3.318414978199538e-05, + "loss": 0.4593, + "step": 26276 + }, + { + "epoch": 1.01, + "learning_rate": 3.318350859194666e-05, + "loss": 0.4113, + "step": 26277 + }, + { + "epoch": 1.01, + "learning_rate": 3.318286740189793e-05, + "loss": 0.2822, + "step": 26278 + }, + { + "epoch": 1.01, + "learning_rate": 3.3182226211849195e-05, + "loss": 0.3676, + "step": 26279 + }, + { + "epoch": 1.01, + "learning_rate": 3.3181585021800466e-05, + "loss": 0.5437, + "step": 26280 + }, + { + "epoch": 1.01, + "learning_rate": 3.318094383175173e-05, + "loss": 0.3278, + "step": 26281 + }, + { + "epoch": 1.01, + "learning_rate": 3.3180302641703e-05, + "loss": 0.5888, + "step": 26282 + }, + { + "epoch": 1.01, + "learning_rate": 3.317966145165427e-05, + "loss": 0.4765, + "step": 26283 + }, + { + "epoch": 1.01, + "learning_rate": 3.317902026160554e-05, + "loss": 0.4538, + "step": 26284 + }, + { + "epoch": 1.01, + "learning_rate": 3.317837907155681e-05, + "loss": 0.4324, + "step": 26285 + }, + { + "epoch": 1.01, + "learning_rate": 3.317773788150808e-05, + "loss": 0.2837, + "step": 26286 + }, + { + "epoch": 1.01, + "learning_rate": 3.3177096691459355e-05, + "loss": 0.4103, + "step": 26287 + }, + { + "epoch": 1.01, + "learning_rate": 3.317645550141062e-05, + "loss": 0.4452, + "step": 26288 + }, + { + "epoch": 1.01, + "learning_rate": 3.317581431136189e-05, + "loss": 0.3811, + "step": 26289 + }, + { + "epoch": 1.01, + "learning_rate": 3.3175173121313156e-05, + "loss": 0.3903, + "step": 26290 + }, + { + "epoch": 1.01, + "learning_rate": 3.317453193126443e-05, + "loss": 0.4793, + "step": 26291 + }, + { + "epoch": 1.01, + "learning_rate": 3.317389074121569e-05, + "loss": 0.4502, + "step": 26292 + }, + { + "epoch": 1.01, + "learning_rate": 3.317324955116697e-05, + "loss": 0.3007, + "step": 26293 + }, + { + "epoch": 1.01, + "learning_rate": 3.3172608361118236e-05, + "loss": 0.4391, + "step": 26294 + }, + { + "epoch": 1.01, + "learning_rate": 3.317196717106951e-05, + "loss": 0.3629, + "step": 26295 + }, + { + "epoch": 1.01, + "learning_rate": 3.317132598102078e-05, + "loss": 0.4429, + "step": 26296 + }, + { + "epoch": 1.01, + "learning_rate": 3.3170684790972045e-05, + "loss": 0.3721, + "step": 26297 + }, + { + "epoch": 1.01, + "learning_rate": 3.3170043600923317e-05, + "loss": 0.2754, + "step": 26298 + }, + { + "epoch": 1.01, + "learning_rate": 3.316940241087458e-05, + "loss": 0.4135, + "step": 26299 + }, + { + "epoch": 1.01, + "learning_rate": 3.316876122082585e-05, + "loss": 0.4347, + "step": 26300 + }, + { + "epoch": 1.01, + "learning_rate": 3.3168120030777125e-05, + "loss": 0.3694, + "step": 26301 + }, + { + "epoch": 1.01, + "learning_rate": 3.31674788407284e-05, + "loss": 0.4105, + "step": 26302 + }, + { + "epoch": 1.01, + "learning_rate": 3.316683765067966e-05, + "loss": 0.2663, + "step": 26303 + }, + { + "epoch": 1.01, + "learning_rate": 3.316619646063093e-05, + "loss": 0.3909, + "step": 26304 + }, + { + "epoch": 1.01, + "learning_rate": 3.3165555270582205e-05, + "loss": 0.3087, + "step": 26305 + }, + { + "epoch": 1.01, + "learning_rate": 3.316491408053347e-05, + "loss": 0.4676, + "step": 26306 + }, + { + "epoch": 1.01, + "learning_rate": 3.316427289048474e-05, + "loss": 0.4849, + "step": 26307 + }, + { + "epoch": 1.01, + "learning_rate": 3.316363170043601e-05, + "loss": 0.3621, + "step": 26308 + }, + { + "epoch": 1.01, + "learning_rate": 3.3162990510387285e-05, + "loss": 0.3084, + "step": 26309 + }, + { + "epoch": 1.01, + "learning_rate": 3.316234932033855e-05, + "loss": 0.4755, + "step": 26310 + }, + { + "epoch": 1.01, + "learning_rate": 3.316170813028982e-05, + "loss": 0.4132, + "step": 26311 + }, + { + "epoch": 1.01, + "learning_rate": 3.316106694024109e-05, + "loss": 0.325, + "step": 26312 + }, + { + "epoch": 1.01, + "learning_rate": 3.316042575019236e-05, + "loss": 0.2366, + "step": 26313 + }, + { + "epoch": 1.01, + "learning_rate": 3.315978456014363e-05, + "loss": 0.3995, + "step": 26314 + }, + { + "epoch": 1.01, + "learning_rate": 3.3159143370094895e-05, + "loss": 0.2789, + "step": 26315 + }, + { + "epoch": 1.01, + "learning_rate": 3.315850218004617e-05, + "loss": 0.4, + "step": 26316 + }, + { + "epoch": 1.01, + "learning_rate": 3.315786098999744e-05, + "loss": 0.4119, + "step": 26317 + }, + { + "epoch": 1.01, + "learning_rate": 3.315721979994871e-05, + "loss": 0.4436, + "step": 26318 + }, + { + "epoch": 1.01, + "learning_rate": 3.3156578609899975e-05, + "loss": 0.3349, + "step": 26319 + }, + { + "epoch": 1.01, + "learning_rate": 3.315593741985125e-05, + "loss": 0.3766, + "step": 26320 + }, + { + "epoch": 1.01, + "learning_rate": 3.315529622980251e-05, + "loss": 0.5428, + "step": 26321 + }, + { + "epoch": 1.01, + "learning_rate": 3.3154655039753784e-05, + "loss": 0.3885, + "step": 26322 + }, + { + "epoch": 1.01, + "learning_rate": 3.3154013849705055e-05, + "loss": 0.4645, + "step": 26323 + }, + { + "epoch": 1.01, + "learning_rate": 3.315337265965632e-05, + "loss": 0.4455, + "step": 26324 + }, + { + "epoch": 1.01, + "learning_rate": 3.315273146960759e-05, + "loss": 0.4462, + "step": 26325 + }, + { + "epoch": 1.01, + "learning_rate": 3.3152090279558864e-05, + "loss": 0.3479, + "step": 26326 + }, + { + "epoch": 1.01, + "learning_rate": 3.3151449089510135e-05, + "loss": 0.3477, + "step": 26327 + }, + { + "epoch": 1.01, + "learning_rate": 3.31508078994614e-05, + "loss": 0.3186, + "step": 26328 + }, + { + "epoch": 1.01, + "learning_rate": 3.315016670941267e-05, + "loss": 0.3231, + "step": 26329 + }, + { + "epoch": 1.01, + "learning_rate": 3.314952551936394e-05, + "loss": 0.2531, + "step": 26330 + }, + { + "epoch": 1.01, + "learning_rate": 3.314888432931521e-05, + "loss": 0.5133, + "step": 26331 + }, + { + "epoch": 1.01, + "learning_rate": 3.314824313926648e-05, + "loss": 0.5223, + "step": 26332 + }, + { + "epoch": 1.01, + "learning_rate": 3.314760194921775e-05, + "loss": 0.4562, + "step": 26333 + }, + { + "epoch": 1.01, + "learning_rate": 3.3146960759169024e-05, + "loss": 0.5386, + "step": 26334 + }, + { + "epoch": 1.01, + "learning_rate": 3.314631956912029e-05, + "loss": 0.4123, + "step": 26335 + }, + { + "epoch": 1.01, + "learning_rate": 3.314567837907156e-05, + "loss": 0.3978, + "step": 26336 + }, + { + "epoch": 1.01, + "learning_rate": 3.3145037189022825e-05, + "loss": 0.5577, + "step": 26337 + }, + { + "epoch": 1.01, + "learning_rate": 3.31443959989741e-05, + "loss": 0.4773, + "step": 26338 + }, + { + "epoch": 1.01, + "learning_rate": 3.314375480892536e-05, + "loss": 0.4269, + "step": 26339 + }, + { + "epoch": 1.01, + "learning_rate": 3.3143113618876634e-05, + "loss": 0.3297, + "step": 26340 + }, + { + "epoch": 1.01, + "learning_rate": 3.3142472428827906e-05, + "loss": 0.519, + "step": 26341 + }, + { + "epoch": 1.01, + "learning_rate": 3.314183123877918e-05, + "loss": 0.4177, + "step": 26342 + }, + { + "epoch": 1.01, + "learning_rate": 3.314119004873045e-05, + "loss": 0.4983, + "step": 26343 + }, + { + "epoch": 1.01, + "learning_rate": 3.3140548858681714e-05, + "loss": 0.3249, + "step": 26344 + }, + { + "epoch": 1.01, + "learning_rate": 3.3139907668632986e-05, + "loss": 0.433, + "step": 26345 + }, + { + "epoch": 1.01, + "learning_rate": 3.313926647858425e-05, + "loss": 0.39, + "step": 26346 + }, + { + "epoch": 1.01, + "learning_rate": 3.313862528853552e-05, + "loss": 0.4212, + "step": 26347 + }, + { + "epoch": 1.01, + "learning_rate": 3.3137984098486794e-05, + "loss": 0.368, + "step": 26348 + }, + { + "epoch": 1.01, + "learning_rate": 3.3137342908438066e-05, + "loss": 0.4075, + "step": 26349 + }, + { + "epoch": 1.01, + "learning_rate": 3.313670171838934e-05, + "loss": 0.4153, + "step": 26350 + }, + { + "epoch": 1.01, + "learning_rate": 3.31360605283406e-05, + "loss": 0.3092, + "step": 26351 + }, + { + "epoch": 1.01, + "learning_rate": 3.3135419338291874e-05, + "loss": 0.3752, + "step": 26352 + }, + { + "epoch": 1.01, + "learning_rate": 3.313477814824314e-05, + "loss": 0.3961, + "step": 26353 + }, + { + "epoch": 1.01, + "learning_rate": 3.313413695819441e-05, + "loss": 0.3482, + "step": 26354 + }, + { + "epoch": 1.01, + "learning_rate": 3.3133495768145676e-05, + "loss": 0.4781, + "step": 26355 + }, + { + "epoch": 1.01, + "learning_rate": 3.313285457809695e-05, + "loss": 0.3797, + "step": 26356 + }, + { + "epoch": 1.01, + "learning_rate": 3.313221338804822e-05, + "loss": 0.4405, + "step": 26357 + }, + { + "epoch": 1.01, + "learning_rate": 3.313157219799949e-05, + "loss": 0.5653, + "step": 26358 + }, + { + "epoch": 1.01, + "learning_rate": 3.313093100795076e-05, + "loss": 0.4367, + "step": 26359 + }, + { + "epoch": 1.01, + "learning_rate": 3.313028981790203e-05, + "loss": 0.4213, + "step": 26360 + }, + { + "epoch": 1.01, + "learning_rate": 3.31296486278533e-05, + "loss": 0.3539, + "step": 26361 + }, + { + "epoch": 1.01, + "learning_rate": 3.3129007437804564e-05, + "loss": 0.3376, + "step": 26362 + }, + { + "epoch": 1.01, + "learning_rate": 3.3128366247755836e-05, + "loss": 0.3827, + "step": 26363 + }, + { + "epoch": 1.01, + "learning_rate": 3.31277250577071e-05, + "loss": 0.4732, + "step": 26364 + }, + { + "epoch": 1.01, + "learning_rate": 3.312708386765838e-05, + "loss": 0.4315, + "step": 26365 + }, + { + "epoch": 1.01, + "learning_rate": 3.3126442677609644e-05, + "loss": 0.2893, + "step": 26366 + }, + { + "epoch": 1.01, + "learning_rate": 3.3125801487560916e-05, + "loss": 0.4549, + "step": 26367 + }, + { + "epoch": 1.01, + "learning_rate": 3.312516029751219e-05, + "loss": 0.4542, + "step": 26368 + }, + { + "epoch": 1.01, + "learning_rate": 3.312451910746345e-05, + "loss": 0.3197, + "step": 26369 + }, + { + "epoch": 1.01, + "learning_rate": 3.3123877917414724e-05, + "loss": 0.4127, + "step": 26370 + }, + { + "epoch": 1.01, + "learning_rate": 3.312323672736599e-05, + "loss": 0.282, + "step": 26371 + }, + { + "epoch": 1.01, + "learning_rate": 3.312259553731726e-05, + "loss": 0.3966, + "step": 26372 + }, + { + "epoch": 1.01, + "learning_rate": 3.312195434726853e-05, + "loss": 0.2468, + "step": 26373 + }, + { + "epoch": 1.01, + "learning_rate": 3.3121313157219804e-05, + "loss": 0.2928, + "step": 26374 + }, + { + "epoch": 1.01, + "learning_rate": 3.312067196717107e-05, + "loss": 0.422, + "step": 26375 + }, + { + "epoch": 1.01, + "learning_rate": 3.312003077712234e-05, + "loss": 0.35, + "step": 26376 + }, + { + "epoch": 1.01, + "learning_rate": 3.311938958707361e-05, + "loss": 0.5716, + "step": 26377 + }, + { + "epoch": 1.01, + "learning_rate": 3.311874839702488e-05, + "loss": 0.3952, + "step": 26378 + }, + { + "epoch": 1.01, + "learning_rate": 3.311810720697615e-05, + "loss": 0.3637, + "step": 26379 + }, + { + "epoch": 1.01, + "learning_rate": 3.3117466016927414e-05, + "loss": 0.4466, + "step": 26380 + }, + { + "epoch": 1.01, + "learning_rate": 3.311682482687869e-05, + "loss": 0.306, + "step": 26381 + }, + { + "epoch": 1.01, + "learning_rate": 3.311618363682996e-05, + "loss": 0.4642, + "step": 26382 + }, + { + "epoch": 1.01, + "learning_rate": 3.311554244678123e-05, + "loss": 0.4341, + "step": 26383 + }, + { + "epoch": 1.01, + "learning_rate": 3.3114901256732494e-05, + "loss": 0.4013, + "step": 26384 + }, + { + "epoch": 1.01, + "learning_rate": 3.3114260066683766e-05, + "loss": 0.3103, + "step": 26385 + }, + { + "epoch": 1.01, + "learning_rate": 3.311361887663504e-05, + "loss": 0.4093, + "step": 26386 + }, + { + "epoch": 1.01, + "learning_rate": 3.31129776865863e-05, + "loss": 0.4575, + "step": 26387 + }, + { + "epoch": 1.01, + "learning_rate": 3.3112336496537575e-05, + "loss": 0.3909, + "step": 26388 + }, + { + "epoch": 1.01, + "learning_rate": 3.3111695306488846e-05, + "loss": 0.4013, + "step": 26389 + }, + { + "epoch": 1.01, + "learning_rate": 3.311105411644012e-05, + "loss": 0.4128, + "step": 26390 + }, + { + "epoch": 1.01, + "learning_rate": 3.311041292639138e-05, + "loss": 0.4434, + "step": 26391 + }, + { + "epoch": 1.01, + "learning_rate": 3.3109771736342655e-05, + "loss": 0.3864, + "step": 26392 + }, + { + "epoch": 1.01, + "learning_rate": 3.310913054629392e-05, + "loss": 0.2871, + "step": 26393 + }, + { + "epoch": 1.01, + "learning_rate": 3.310848935624519e-05, + "loss": 0.5159, + "step": 26394 + }, + { + "epoch": 1.01, + "learning_rate": 3.310784816619646e-05, + "loss": 0.315, + "step": 26395 + }, + { + "epoch": 1.01, + "learning_rate": 3.310720697614773e-05, + "loss": 0.4858, + "step": 26396 + }, + { + "epoch": 1.01, + "learning_rate": 3.3106565786099006e-05, + "loss": 0.3668, + "step": 26397 + }, + { + "epoch": 1.01, + "learning_rate": 3.310592459605027e-05, + "loss": 0.3549, + "step": 26398 + }, + { + "epoch": 1.01, + "learning_rate": 3.310528340600154e-05, + "loss": 0.4962, + "step": 26399 + }, + { + "epoch": 1.01, + "learning_rate": 3.310464221595281e-05, + "loss": 0.5323, + "step": 26400 + }, + { + "epoch": 1.02, + "learning_rate": 3.310400102590408e-05, + "loss": 0.2874, + "step": 26401 + }, + { + "epoch": 1.02, + "learning_rate": 3.3103359835855345e-05, + "loss": 0.3619, + "step": 26402 + }, + { + "epoch": 1.02, + "learning_rate": 3.3102718645806616e-05, + "loss": 0.3656, + "step": 26403 + }, + { + "epoch": 1.02, + "learning_rate": 3.310207745575789e-05, + "loss": 0.3996, + "step": 26404 + }, + { + "epoch": 1.02, + "learning_rate": 3.310143626570916e-05, + "loss": 0.5249, + "step": 26405 + }, + { + "epoch": 1.02, + "learning_rate": 3.310079507566043e-05, + "loss": 0.3752, + "step": 26406 + }, + { + "epoch": 1.02, + "learning_rate": 3.3100153885611696e-05, + "loss": 0.3688, + "step": 26407 + }, + { + "epoch": 1.02, + "learning_rate": 3.309951269556297e-05, + "loss": 0.3989, + "step": 26408 + }, + { + "epoch": 1.02, + "learning_rate": 3.309887150551423e-05, + "loss": 0.412, + "step": 26409 + }, + { + "epoch": 1.02, + "learning_rate": 3.3098230315465505e-05, + "loss": 0.4662, + "step": 26410 + }, + { + "epoch": 1.02, + "learning_rate": 3.309758912541677e-05, + "loss": 0.4416, + "step": 26411 + }, + { + "epoch": 1.02, + "learning_rate": 3.309694793536804e-05, + "loss": 0.3619, + "step": 26412 + }, + { + "epoch": 1.02, + "learning_rate": 3.309630674531931e-05, + "loss": 0.4996, + "step": 26413 + }, + { + "epoch": 1.02, + "learning_rate": 3.3095665555270585e-05, + "loss": 0.4428, + "step": 26414 + }, + { + "epoch": 1.02, + "learning_rate": 3.309502436522186e-05, + "loss": 0.4177, + "step": 26415 + }, + { + "epoch": 1.02, + "learning_rate": 3.309438317517312e-05, + "loss": 0.2697, + "step": 26416 + }, + { + "epoch": 1.02, + "learning_rate": 3.309374198512439e-05, + "loss": 0.319, + "step": 26417 + }, + { + "epoch": 1.02, + "learning_rate": 3.309310079507566e-05, + "loss": 0.4222, + "step": 26418 + }, + { + "epoch": 1.02, + "learning_rate": 3.309245960502693e-05, + "loss": 0.4216, + "step": 26419 + }, + { + "epoch": 1.02, + "learning_rate": 3.30918184149782e-05, + "loss": 0.4192, + "step": 26420 + }, + { + "epoch": 1.02, + "learning_rate": 3.3091177224929473e-05, + "loss": 0.5391, + "step": 26421 + }, + { + "epoch": 1.02, + "learning_rate": 3.3090536034880745e-05, + "loss": 0.4722, + "step": 26422 + }, + { + "epoch": 1.02, + "learning_rate": 3.308989484483201e-05, + "loss": 0.3573, + "step": 26423 + }, + { + "epoch": 1.02, + "learning_rate": 3.308925365478328e-05, + "loss": 0.3443, + "step": 26424 + }, + { + "epoch": 1.02, + "learning_rate": 3.308861246473455e-05, + "loss": 0.2684, + "step": 26425 + }, + { + "epoch": 1.02, + "learning_rate": 3.308797127468582e-05, + "loss": 0.2987, + "step": 26426 + }, + { + "epoch": 1.02, + "learning_rate": 3.3087330084637083e-05, + "loss": 0.5515, + "step": 26427 + }, + { + "epoch": 1.02, + "learning_rate": 3.3086688894588355e-05, + "loss": 0.3831, + "step": 26428 + }, + { + "epoch": 1.02, + "learning_rate": 3.308604770453963e-05, + "loss": 0.5456, + "step": 26429 + }, + { + "epoch": 1.02, + "learning_rate": 3.30854065144909e-05, + "loss": 0.4304, + "step": 26430 + }, + { + "epoch": 1.02, + "learning_rate": 3.308476532444217e-05, + "loss": 0.3799, + "step": 26431 + }, + { + "epoch": 1.02, + "learning_rate": 3.3084124134393435e-05, + "loss": 0.4935, + "step": 26432 + }, + { + "epoch": 1.02, + "learning_rate": 3.308348294434471e-05, + "loss": 0.3878, + "step": 26433 + }, + { + "epoch": 1.02, + "learning_rate": 3.308284175429597e-05, + "loss": 0.4248, + "step": 26434 + }, + { + "epoch": 1.02, + "learning_rate": 3.3082200564247244e-05, + "loss": 0.3804, + "step": 26435 + }, + { + "epoch": 1.02, + "learning_rate": 3.3081559374198515e-05, + "loss": 0.3785, + "step": 26436 + }, + { + "epoch": 1.02, + "learning_rate": 3.308091818414979e-05, + "loss": 0.3952, + "step": 26437 + }, + { + "epoch": 1.02, + "learning_rate": 3.308027699410105e-05, + "loss": 0.344, + "step": 26438 + }, + { + "epoch": 1.02, + "learning_rate": 3.3079635804052324e-05, + "loss": 0.2614, + "step": 26439 + }, + { + "epoch": 1.02, + "learning_rate": 3.3078994614003595e-05, + "loss": 0.3919, + "step": 26440 + }, + { + "epoch": 1.02, + "learning_rate": 3.307835342395486e-05, + "loss": 0.2774, + "step": 26441 + }, + { + "epoch": 1.02, + "learning_rate": 3.307771223390613e-05, + "loss": 0.2877, + "step": 26442 + }, + { + "epoch": 1.02, + "learning_rate": 3.30770710438574e-05, + "loss": 0.3996, + "step": 26443 + }, + { + "epoch": 1.02, + "learning_rate": 3.307642985380867e-05, + "loss": 0.3492, + "step": 26444 + }, + { + "epoch": 1.02, + "learning_rate": 3.307578866375994e-05, + "loss": 0.2858, + "step": 26445 + }, + { + "epoch": 1.02, + "learning_rate": 3.307514747371121e-05, + "loss": 0.3314, + "step": 26446 + }, + { + "epoch": 1.02, + "learning_rate": 3.307450628366248e-05, + "loss": 0.5654, + "step": 26447 + }, + { + "epoch": 1.02, + "learning_rate": 3.307386509361375e-05, + "loss": 0.4019, + "step": 26448 + }, + { + "epoch": 1.02, + "learning_rate": 3.307322390356502e-05, + "loss": 0.3118, + "step": 26449 + }, + { + "epoch": 1.02, + "learning_rate": 3.3072582713516285e-05, + "loss": 0.5124, + "step": 26450 + }, + { + "epoch": 1.02, + "learning_rate": 3.307194152346756e-05, + "loss": 0.3807, + "step": 26451 + }, + { + "epoch": 1.02, + "learning_rate": 3.307130033341882e-05, + "loss": 0.3284, + "step": 26452 + }, + { + "epoch": 1.02, + "learning_rate": 3.30706591433701e-05, + "loss": 0.4634, + "step": 26453 + }, + { + "epoch": 1.02, + "learning_rate": 3.3070017953321366e-05, + "loss": 0.3089, + "step": 26454 + }, + { + "epoch": 1.02, + "learning_rate": 3.306937676327264e-05, + "loss": 0.4682, + "step": 26455 + }, + { + "epoch": 1.02, + "learning_rate": 3.306873557322391e-05, + "loss": 0.4389, + "step": 26456 + }, + { + "epoch": 1.02, + "learning_rate": 3.3068094383175174e-05, + "loss": 0.4895, + "step": 26457 + }, + { + "epoch": 1.02, + "learning_rate": 3.3067453193126446e-05, + "loss": 0.441, + "step": 26458 + }, + { + "epoch": 1.02, + "learning_rate": 3.306681200307771e-05, + "loss": 0.4268, + "step": 26459 + }, + { + "epoch": 1.02, + "learning_rate": 3.306617081302898e-05, + "loss": 0.4603, + "step": 26460 + }, + { + "epoch": 1.02, + "learning_rate": 3.3065529622980254e-05, + "loss": 0.3875, + "step": 26461 + }, + { + "epoch": 1.02, + "learning_rate": 3.3064888432931526e-05, + "loss": 0.5434, + "step": 26462 + }, + { + "epoch": 1.02, + "learning_rate": 3.306424724288279e-05, + "loss": 0.3542, + "step": 26463 + }, + { + "epoch": 1.02, + "learning_rate": 3.306360605283406e-05, + "loss": 0.3532, + "step": 26464 + }, + { + "epoch": 1.02, + "learning_rate": 3.3062964862785334e-05, + "loss": 0.3254, + "step": 26465 + }, + { + "epoch": 1.02, + "learning_rate": 3.30623236727366e-05, + "loss": 0.3691, + "step": 26466 + }, + { + "epoch": 1.02, + "learning_rate": 3.306168248268787e-05, + "loss": 0.3899, + "step": 26467 + }, + { + "epoch": 1.02, + "learning_rate": 3.3061041292639136e-05, + "loss": 0.4892, + "step": 26468 + }, + { + "epoch": 1.02, + "learning_rate": 3.3060400102590414e-05, + "loss": 0.3815, + "step": 26469 + }, + { + "epoch": 1.02, + "learning_rate": 3.305975891254168e-05, + "loss": 0.4002, + "step": 26470 + }, + { + "epoch": 1.02, + "learning_rate": 3.305911772249295e-05, + "loss": 0.3961, + "step": 26471 + }, + { + "epoch": 1.02, + "learning_rate": 3.3058476532444216e-05, + "loss": 0.3216, + "step": 26472 + }, + { + "epoch": 1.02, + "learning_rate": 3.305783534239549e-05, + "loss": 0.3446, + "step": 26473 + }, + { + "epoch": 1.02, + "learning_rate": 3.305719415234676e-05, + "loss": 0.332, + "step": 26474 + }, + { + "epoch": 1.02, + "learning_rate": 3.3056552962298024e-05, + "loss": 0.3798, + "step": 26475 + }, + { + "epoch": 1.02, + "learning_rate": 3.3055911772249296e-05, + "loss": 0.398, + "step": 26476 + }, + { + "epoch": 1.02, + "learning_rate": 3.305527058220057e-05, + "loss": 0.3564, + "step": 26477 + }, + { + "epoch": 1.02, + "learning_rate": 3.305462939215184e-05, + "loss": 0.3633, + "step": 26478 + }, + { + "epoch": 1.02, + "learning_rate": 3.3053988202103104e-05, + "loss": 0.3775, + "step": 26479 + }, + { + "epoch": 1.02, + "learning_rate": 3.3053347012054376e-05, + "loss": 0.3975, + "step": 26480 + }, + { + "epoch": 1.02, + "learning_rate": 3.305270582200564e-05, + "loss": 0.3042, + "step": 26481 + }, + { + "epoch": 1.02, + "learning_rate": 3.305206463195691e-05, + "loss": 0.4075, + "step": 26482 + }, + { + "epoch": 1.02, + "learning_rate": 3.3051423441908184e-05, + "loss": 0.4032, + "step": 26483 + }, + { + "epoch": 1.02, + "learning_rate": 3.305078225185945e-05, + "loss": 0.2539, + "step": 26484 + }, + { + "epoch": 1.02, + "learning_rate": 3.305014106181072e-05, + "loss": 0.3988, + "step": 26485 + }, + { + "epoch": 1.02, + "learning_rate": 3.304949987176199e-05, + "loss": 0.4426, + "step": 26486 + }, + { + "epoch": 1.02, + "learning_rate": 3.3048858681713264e-05, + "loss": 0.4302, + "step": 26487 + }, + { + "epoch": 1.02, + "learning_rate": 3.304821749166453e-05, + "loss": 0.4113, + "step": 26488 + }, + { + "epoch": 1.02, + "learning_rate": 3.30475763016158e-05, + "loss": 0.3813, + "step": 26489 + }, + { + "epoch": 1.02, + "learning_rate": 3.3046935111567066e-05, + "loss": 0.3073, + "step": 26490 + }, + { + "epoch": 1.02, + "learning_rate": 3.304629392151834e-05, + "loss": 0.3753, + "step": 26491 + }, + { + "epoch": 1.02, + "learning_rate": 3.304565273146961e-05, + "loss": 0.3386, + "step": 26492 + }, + { + "epoch": 1.02, + "learning_rate": 3.304501154142088e-05, + "loss": 0.3788, + "step": 26493 + }, + { + "epoch": 1.02, + "learning_rate": 3.304437035137215e-05, + "loss": 0.454, + "step": 26494 + }, + { + "epoch": 1.02, + "learning_rate": 3.304372916132342e-05, + "loss": 0.5557, + "step": 26495 + }, + { + "epoch": 1.02, + "learning_rate": 3.304308797127469e-05, + "loss": 0.3758, + "step": 26496 + }, + { + "epoch": 1.02, + "learning_rate": 3.3042446781225954e-05, + "loss": 0.3601, + "step": 26497 + }, + { + "epoch": 1.02, + "learning_rate": 3.3041805591177226e-05, + "loss": 0.4311, + "step": 26498 + }, + { + "epoch": 1.02, + "learning_rate": 3.304116440112849e-05, + "loss": 0.3325, + "step": 26499 + }, + { + "epoch": 1.02, + "learning_rate": 3.304052321107976e-05, + "loss": 0.4638, + "step": 26500 + }, + { + "epoch": 1.02, + "learning_rate": 3.3039882021031035e-05, + "loss": 0.3938, + "step": 26501 + }, + { + "epoch": 1.02, + "learning_rate": 3.3039240830982306e-05, + "loss": 0.3234, + "step": 26502 + }, + { + "epoch": 1.02, + "learning_rate": 3.303859964093358e-05, + "loss": 0.3909, + "step": 26503 + }, + { + "epoch": 1.02, + "learning_rate": 3.303795845088484e-05, + "loss": 0.4657, + "step": 26504 + }, + { + "epoch": 1.02, + "learning_rate": 3.3037317260836115e-05, + "loss": 0.4193, + "step": 26505 + }, + { + "epoch": 1.02, + "learning_rate": 3.303667607078738e-05, + "loss": 0.3404, + "step": 26506 + }, + { + "epoch": 1.02, + "learning_rate": 3.303603488073865e-05, + "loss": 0.2919, + "step": 26507 + }, + { + "epoch": 1.02, + "learning_rate": 3.303539369068992e-05, + "loss": 0.3, + "step": 26508 + }, + { + "epoch": 1.02, + "learning_rate": 3.3034752500641195e-05, + "loss": 0.4358, + "step": 26509 + }, + { + "epoch": 1.02, + "learning_rate": 3.3034111310592466e-05, + "loss": 0.4548, + "step": 26510 + }, + { + "epoch": 1.02, + "learning_rate": 3.303347012054373e-05, + "loss": 0.3631, + "step": 26511 + }, + { + "epoch": 1.02, + "learning_rate": 3.3032828930495e-05, + "loss": 0.2689, + "step": 26512 + }, + { + "epoch": 1.02, + "learning_rate": 3.303218774044627e-05, + "loss": 0.3869, + "step": 26513 + }, + { + "epoch": 1.02, + "learning_rate": 3.303154655039754e-05, + "loss": 0.3958, + "step": 26514 + }, + { + "epoch": 1.02, + "learning_rate": 3.3030905360348805e-05, + "loss": 0.474, + "step": 26515 + }, + { + "epoch": 1.02, + "learning_rate": 3.3030264170300076e-05, + "loss": 0.3824, + "step": 26516 + }, + { + "epoch": 1.02, + "learning_rate": 3.302962298025135e-05, + "loss": 0.4331, + "step": 26517 + }, + { + "epoch": 1.02, + "learning_rate": 3.302898179020262e-05, + "loss": 0.4089, + "step": 26518 + }, + { + "epoch": 1.02, + "learning_rate": 3.302834060015389e-05, + "loss": 0.3551, + "step": 26519 + }, + { + "epoch": 1.02, + "learning_rate": 3.3027699410105157e-05, + "loss": 0.4117, + "step": 26520 + }, + { + "epoch": 1.02, + "learning_rate": 3.302705822005643e-05, + "loss": 0.206, + "step": 26521 + }, + { + "epoch": 1.02, + "learning_rate": 3.302641703000769e-05, + "loss": 0.3593, + "step": 26522 + }, + { + "epoch": 1.02, + "learning_rate": 3.3025775839958965e-05, + "loss": 0.3615, + "step": 26523 + }, + { + "epoch": 1.02, + "learning_rate": 3.302513464991023e-05, + "loss": 0.401, + "step": 26524 + }, + { + "epoch": 1.02, + "learning_rate": 3.302449345986151e-05, + "loss": 0.4495, + "step": 26525 + }, + { + "epoch": 1.02, + "learning_rate": 3.302385226981277e-05, + "loss": 0.4114, + "step": 26526 + }, + { + "epoch": 1.02, + "learning_rate": 3.3023211079764045e-05, + "loss": 0.3464, + "step": 26527 + }, + { + "epoch": 1.02, + "learning_rate": 3.302256988971532e-05, + "loss": 0.4774, + "step": 26528 + }, + { + "epoch": 1.02, + "learning_rate": 3.302192869966658e-05, + "loss": 0.3579, + "step": 26529 + }, + { + "epoch": 1.02, + "learning_rate": 3.302128750961785e-05, + "loss": 0.424, + "step": 26530 + }, + { + "epoch": 1.02, + "learning_rate": 3.302064631956912e-05, + "loss": 0.3921, + "step": 26531 + }, + { + "epoch": 1.02, + "learning_rate": 3.302000512952039e-05, + "loss": 0.4939, + "step": 26532 + }, + { + "epoch": 1.02, + "learning_rate": 3.301936393947166e-05, + "loss": 0.3722, + "step": 26533 + }, + { + "epoch": 1.02, + "learning_rate": 3.3018722749422933e-05, + "loss": 0.43, + "step": 26534 + }, + { + "epoch": 1.02, + "learning_rate": 3.30180815593742e-05, + "loss": 0.4072, + "step": 26535 + }, + { + "epoch": 1.02, + "learning_rate": 3.301744036932547e-05, + "loss": 0.4526, + "step": 26536 + }, + { + "epoch": 1.02, + "learning_rate": 3.301679917927674e-05, + "loss": 0.292, + "step": 26537 + }, + { + "epoch": 1.02, + "learning_rate": 3.301615798922801e-05, + "loss": 0.3941, + "step": 26538 + }, + { + "epoch": 1.02, + "learning_rate": 3.301551679917928e-05, + "loss": 0.4007, + "step": 26539 + }, + { + "epoch": 1.02, + "learning_rate": 3.3014875609130543e-05, + "loss": 0.3855, + "step": 26540 + }, + { + "epoch": 1.02, + "learning_rate": 3.301423441908182e-05, + "loss": 0.4797, + "step": 26541 + }, + { + "epoch": 1.02, + "learning_rate": 3.301359322903309e-05, + "loss": 0.3425, + "step": 26542 + }, + { + "epoch": 1.02, + "learning_rate": 3.301295203898436e-05, + "loss": 0.3396, + "step": 26543 + }, + { + "epoch": 1.02, + "learning_rate": 3.3012310848935624e-05, + "loss": 0.5246, + "step": 26544 + }, + { + "epoch": 1.02, + "learning_rate": 3.3011669658886895e-05, + "loss": 0.4304, + "step": 26545 + }, + { + "epoch": 1.02, + "learning_rate": 3.301102846883817e-05, + "loss": 0.3632, + "step": 26546 + }, + { + "epoch": 1.02, + "learning_rate": 3.301038727878943e-05, + "loss": 0.3668, + "step": 26547 + }, + { + "epoch": 1.02, + "learning_rate": 3.3009746088740704e-05, + "loss": 0.4407, + "step": 26548 + }, + { + "epoch": 1.02, + "learning_rate": 3.3009104898691975e-05, + "loss": 0.331, + "step": 26549 + }, + { + "epoch": 1.02, + "learning_rate": 3.300846370864325e-05, + "loss": 0.3338, + "step": 26550 + }, + { + "epoch": 1.02, + "learning_rate": 3.300782251859451e-05, + "loss": 0.4944, + "step": 26551 + }, + { + "epoch": 1.02, + "learning_rate": 3.3007181328545784e-05, + "loss": 0.2132, + "step": 26552 + }, + { + "epoch": 1.02, + "learning_rate": 3.300654013849705e-05, + "loss": 0.4088, + "step": 26553 + }, + { + "epoch": 1.02, + "learning_rate": 3.300589894844832e-05, + "loss": 0.2769, + "step": 26554 + }, + { + "epoch": 1.02, + "learning_rate": 3.300525775839959e-05, + "loss": 0.3859, + "step": 26555 + }, + { + "epoch": 1.02, + "learning_rate": 3.300461656835086e-05, + "loss": 0.4301, + "step": 26556 + }, + { + "epoch": 1.02, + "learning_rate": 3.3003975378302136e-05, + "loss": 0.4425, + "step": 26557 + }, + { + "epoch": 1.02, + "learning_rate": 3.30033341882534e-05, + "loss": 0.4434, + "step": 26558 + }, + { + "epoch": 1.02, + "learning_rate": 3.300269299820467e-05, + "loss": 0.3935, + "step": 26559 + }, + { + "epoch": 1.02, + "learning_rate": 3.300205180815594e-05, + "loss": 0.2435, + "step": 26560 + }, + { + "epoch": 1.02, + "learning_rate": 3.300141061810721e-05, + "loss": 0.2856, + "step": 26561 + }, + { + "epoch": 1.02, + "learning_rate": 3.3000769428058474e-05, + "loss": 0.4248, + "step": 26562 + }, + { + "epoch": 1.02, + "learning_rate": 3.3000128238009745e-05, + "loss": 0.3539, + "step": 26563 + }, + { + "epoch": 1.02, + "learning_rate": 3.299948704796102e-05, + "loss": 0.3958, + "step": 26564 + }, + { + "epoch": 1.02, + "learning_rate": 3.299884585791229e-05, + "loss": 0.38, + "step": 26565 + }, + { + "epoch": 1.02, + "learning_rate": 3.299820466786356e-05, + "loss": 0.4176, + "step": 26566 + }, + { + "epoch": 1.02, + "learning_rate": 3.2997563477814826e-05, + "loss": 0.3186, + "step": 26567 + }, + { + "epoch": 1.02, + "learning_rate": 3.29969222877661e-05, + "loss": 0.4228, + "step": 26568 + }, + { + "epoch": 1.02, + "learning_rate": 3.299628109771736e-05, + "loss": 0.3748, + "step": 26569 + }, + { + "epoch": 1.02, + "learning_rate": 3.2995639907668634e-05, + "loss": 0.3049, + "step": 26570 + }, + { + "epoch": 1.02, + "learning_rate": 3.29949987176199e-05, + "loss": 0.5414, + "step": 26571 + }, + { + "epoch": 1.02, + "learning_rate": 3.299435752757117e-05, + "loss": 0.4488, + "step": 26572 + }, + { + "epoch": 1.02, + "learning_rate": 3.299371633752244e-05, + "loss": 0.3372, + "step": 26573 + }, + { + "epoch": 1.02, + "learning_rate": 3.2993075147473714e-05, + "loss": 0.4777, + "step": 26574 + }, + { + "epoch": 1.02, + "learning_rate": 3.2992433957424986e-05, + "loss": 0.443, + "step": 26575 + }, + { + "epoch": 1.02, + "learning_rate": 3.299179276737625e-05, + "loss": 0.3194, + "step": 26576 + }, + { + "epoch": 1.02, + "learning_rate": 3.299115157732752e-05, + "loss": 0.3, + "step": 26577 + }, + { + "epoch": 1.02, + "learning_rate": 3.299051038727879e-05, + "loss": 0.4458, + "step": 26578 + }, + { + "epoch": 1.02, + "learning_rate": 3.298986919723006e-05, + "loss": 0.4609, + "step": 26579 + }, + { + "epoch": 1.02, + "learning_rate": 3.298922800718133e-05, + "loss": 0.4112, + "step": 26580 + }, + { + "epoch": 1.02, + "learning_rate": 3.29885868171326e-05, + "loss": 0.4685, + "step": 26581 + }, + { + "epoch": 1.02, + "learning_rate": 3.2987945627083874e-05, + "loss": 0.4327, + "step": 26582 + }, + { + "epoch": 1.02, + "learning_rate": 3.298730443703514e-05, + "loss": 0.374, + "step": 26583 + }, + { + "epoch": 1.02, + "learning_rate": 3.298666324698641e-05, + "loss": 0.4133, + "step": 26584 + }, + { + "epoch": 1.02, + "learning_rate": 3.2986022056937676e-05, + "loss": 0.4127, + "step": 26585 + }, + { + "epoch": 1.02, + "learning_rate": 3.298538086688895e-05, + "loss": 0.5078, + "step": 26586 + }, + { + "epoch": 1.02, + "learning_rate": 3.298473967684021e-05, + "loss": 0.3345, + "step": 26587 + }, + { + "epoch": 1.02, + "learning_rate": 3.2984098486791484e-05, + "loss": 0.3945, + "step": 26588 + }, + { + "epoch": 1.02, + "learning_rate": 3.2983457296742756e-05, + "loss": 0.4149, + "step": 26589 + }, + { + "epoch": 1.02, + "learning_rate": 3.298281610669403e-05, + "loss": 0.3733, + "step": 26590 + }, + { + "epoch": 1.02, + "learning_rate": 3.29821749166453e-05, + "loss": 0.4881, + "step": 26591 + }, + { + "epoch": 1.02, + "learning_rate": 3.2981533726596564e-05, + "loss": 0.4191, + "step": 26592 + }, + { + "epoch": 1.02, + "learning_rate": 3.2980892536547836e-05, + "loss": 0.4738, + "step": 26593 + }, + { + "epoch": 1.02, + "learning_rate": 3.29802513464991e-05, + "loss": 0.3671, + "step": 26594 + }, + { + "epoch": 1.02, + "learning_rate": 3.297961015645037e-05, + "loss": 0.2988, + "step": 26595 + }, + { + "epoch": 1.02, + "learning_rate": 3.2978968966401644e-05, + "loss": 0.4394, + "step": 26596 + }, + { + "epoch": 1.02, + "learning_rate": 3.2978327776352916e-05, + "loss": 0.4316, + "step": 26597 + }, + { + "epoch": 1.02, + "learning_rate": 3.297768658630418e-05, + "loss": 0.4342, + "step": 26598 + }, + { + "epoch": 1.02, + "learning_rate": 3.297704539625545e-05, + "loss": 0.4389, + "step": 26599 + }, + { + "epoch": 1.02, + "learning_rate": 3.2976404206206724e-05, + "loss": 0.3518, + "step": 26600 + }, + { + "epoch": 1.02, + "learning_rate": 3.297576301615799e-05, + "loss": 0.4629, + "step": 26601 + }, + { + "epoch": 1.02, + "learning_rate": 3.297512182610926e-05, + "loss": 0.3875, + "step": 26602 + }, + { + "epoch": 1.02, + "learning_rate": 3.2974480636060526e-05, + "loss": 0.4486, + "step": 26603 + }, + { + "epoch": 1.02, + "learning_rate": 3.29738394460118e-05, + "loss": 0.3597, + "step": 26604 + }, + { + "epoch": 1.02, + "learning_rate": 3.297319825596307e-05, + "loss": 0.3539, + "step": 26605 + }, + { + "epoch": 1.02, + "learning_rate": 3.297255706591434e-05, + "loss": 0.3137, + "step": 26606 + }, + { + "epoch": 1.02, + "learning_rate": 3.2971915875865606e-05, + "loss": 0.4336, + "step": 26607 + }, + { + "epoch": 1.02, + "learning_rate": 3.297127468581688e-05, + "loss": 0.3765, + "step": 26608 + }, + { + "epoch": 1.02, + "learning_rate": 3.297063349576815e-05, + "loss": 0.4015, + "step": 26609 + }, + { + "epoch": 1.02, + "learning_rate": 3.2969992305719415e-05, + "loss": 0.4072, + "step": 26610 + }, + { + "epoch": 1.02, + "learning_rate": 3.2969351115670686e-05, + "loss": 0.4196, + "step": 26611 + }, + { + "epoch": 1.02, + "learning_rate": 3.296870992562195e-05, + "loss": 0.3201, + "step": 26612 + }, + { + "epoch": 1.02, + "learning_rate": 3.296806873557323e-05, + "loss": 0.38, + "step": 26613 + }, + { + "epoch": 1.02, + "learning_rate": 3.2967427545524495e-05, + "loss": 0.4388, + "step": 26614 + }, + { + "epoch": 1.02, + "learning_rate": 3.2966786355475766e-05, + "loss": 0.4431, + "step": 26615 + }, + { + "epoch": 1.02, + "learning_rate": 3.296614516542703e-05, + "loss": 0.4107, + "step": 26616 + }, + { + "epoch": 1.02, + "learning_rate": 3.29655039753783e-05, + "loss": 0.3969, + "step": 26617 + }, + { + "epoch": 1.02, + "learning_rate": 3.2964862785329575e-05, + "loss": 0.4368, + "step": 26618 + }, + { + "epoch": 1.02, + "learning_rate": 3.296422159528084e-05, + "loss": 0.5258, + "step": 26619 + }, + { + "epoch": 1.02, + "learning_rate": 3.296358040523211e-05, + "loss": 0.5184, + "step": 26620 + }, + { + "epoch": 1.02, + "learning_rate": 3.296293921518338e-05, + "loss": 0.6243, + "step": 26621 + }, + { + "epoch": 1.02, + "learning_rate": 3.2962298025134655e-05, + "loss": 0.4522, + "step": 26622 + }, + { + "epoch": 1.02, + "learning_rate": 3.296165683508592e-05, + "loss": 0.2603, + "step": 26623 + }, + { + "epoch": 1.02, + "learning_rate": 3.296101564503719e-05, + "loss": 0.3918, + "step": 26624 + }, + { + "epoch": 1.02, + "learning_rate": 3.2960374454988456e-05, + "loss": 0.3807, + "step": 26625 + }, + { + "epoch": 1.02, + "learning_rate": 3.295973326493973e-05, + "loss": 0.4744, + "step": 26626 + }, + { + "epoch": 1.02, + "learning_rate": 3.2959092074891e-05, + "loss": 0.3533, + "step": 26627 + }, + { + "epoch": 1.02, + "learning_rate": 3.2958450884842265e-05, + "loss": 0.4344, + "step": 26628 + }, + { + "epoch": 1.02, + "learning_rate": 3.295780969479354e-05, + "loss": 0.3023, + "step": 26629 + }, + { + "epoch": 1.02, + "learning_rate": 3.295716850474481e-05, + "loss": 0.428, + "step": 26630 + }, + { + "epoch": 1.02, + "learning_rate": 3.295652731469608e-05, + "loss": 0.333, + "step": 26631 + }, + { + "epoch": 1.02, + "learning_rate": 3.2955886124647345e-05, + "loss": 0.4767, + "step": 26632 + }, + { + "epoch": 1.02, + "learning_rate": 3.2955244934598617e-05, + "loss": 0.4875, + "step": 26633 + }, + { + "epoch": 1.02, + "learning_rate": 3.295460374454988e-05, + "loss": 0.4306, + "step": 26634 + }, + { + "epoch": 1.02, + "learning_rate": 3.295396255450115e-05, + "loss": 0.2926, + "step": 26635 + }, + { + "epoch": 1.02, + "learning_rate": 3.2953321364452425e-05, + "loss": 0.3099, + "step": 26636 + }, + { + "epoch": 1.02, + "learning_rate": 3.29526801744037e-05, + "loss": 0.3931, + "step": 26637 + }, + { + "epoch": 1.02, + "learning_rate": 3.295203898435497e-05, + "loss": 0.5332, + "step": 26638 + }, + { + "epoch": 1.02, + "learning_rate": 3.295139779430623e-05, + "loss": 0.5183, + "step": 26639 + }, + { + "epoch": 1.02, + "learning_rate": 3.2950756604257505e-05, + "loss": 0.4231, + "step": 26640 + }, + { + "epoch": 1.02, + "learning_rate": 3.295011541420877e-05, + "loss": 0.4374, + "step": 26641 + }, + { + "epoch": 1.02, + "learning_rate": 3.294947422416004e-05, + "loss": 0.4693, + "step": 26642 + }, + { + "epoch": 1.02, + "learning_rate": 3.2948833034111313e-05, + "loss": 0.3077, + "step": 26643 + }, + { + "epoch": 1.02, + "learning_rate": 3.294819184406258e-05, + "loss": 0.4697, + "step": 26644 + }, + { + "epoch": 1.02, + "learning_rate": 3.294755065401385e-05, + "loss": 0.2567, + "step": 26645 + }, + { + "epoch": 1.02, + "learning_rate": 3.294690946396512e-05, + "loss": 0.2818, + "step": 26646 + }, + { + "epoch": 1.02, + "learning_rate": 3.2946268273916393e-05, + "loss": 0.4494, + "step": 26647 + }, + { + "epoch": 1.02, + "learning_rate": 3.294562708386766e-05, + "loss": 0.4661, + "step": 26648 + }, + { + "epoch": 1.02, + "learning_rate": 3.294498589381893e-05, + "loss": 0.4567, + "step": 26649 + }, + { + "epoch": 1.02, + "learning_rate": 3.2944344703770195e-05, + "loss": 0.4538, + "step": 26650 + }, + { + "epoch": 1.02, + "learning_rate": 3.294370351372147e-05, + "loss": 0.3852, + "step": 26651 + }, + { + "epoch": 1.02, + "learning_rate": 3.294306232367274e-05, + "loss": 0.4726, + "step": 26652 + }, + { + "epoch": 1.02, + "learning_rate": 3.294242113362401e-05, + "loss": 0.3927, + "step": 26653 + }, + { + "epoch": 1.02, + "learning_rate": 3.294177994357528e-05, + "loss": 0.5216, + "step": 26654 + }, + { + "epoch": 1.02, + "learning_rate": 3.294113875352655e-05, + "loss": 0.4871, + "step": 26655 + }, + { + "epoch": 1.02, + "learning_rate": 3.294049756347782e-05, + "loss": 0.4887, + "step": 26656 + }, + { + "epoch": 1.02, + "learning_rate": 3.2939856373429084e-05, + "loss": 0.3722, + "step": 26657 + }, + { + "epoch": 1.02, + "learning_rate": 3.2939215183380355e-05, + "loss": 0.4656, + "step": 26658 + }, + { + "epoch": 1.02, + "learning_rate": 3.293857399333162e-05, + "loss": 0.3681, + "step": 26659 + }, + { + "epoch": 1.02, + "learning_rate": 3.293793280328289e-05, + "loss": 0.3696, + "step": 26660 + }, + { + "epoch": 1.03, + "learning_rate": 3.2937291613234164e-05, + "loss": 0.485, + "step": 26661 + }, + { + "epoch": 1.03, + "learning_rate": 3.2936650423185435e-05, + "loss": 0.4241, + "step": 26662 + }, + { + "epoch": 1.03, + "learning_rate": 3.293600923313671e-05, + "loss": 0.4666, + "step": 26663 + }, + { + "epoch": 1.03, + "learning_rate": 3.293536804308797e-05, + "loss": 0.4855, + "step": 26664 + }, + { + "epoch": 1.03, + "learning_rate": 3.2934726853039244e-05, + "loss": 0.4182, + "step": 26665 + }, + { + "epoch": 1.03, + "learning_rate": 3.293408566299051e-05, + "loss": 0.3059, + "step": 26666 + }, + { + "epoch": 1.03, + "learning_rate": 3.293344447294178e-05, + "loss": 0.3754, + "step": 26667 + }, + { + "epoch": 1.03, + "learning_rate": 3.293280328289305e-05, + "loss": 0.3633, + "step": 26668 + }, + { + "epoch": 1.03, + "learning_rate": 3.2932162092844324e-05, + "loss": 0.4001, + "step": 26669 + }, + { + "epoch": 1.03, + "learning_rate": 3.2931520902795596e-05, + "loss": 0.4248, + "step": 26670 + }, + { + "epoch": 1.03, + "learning_rate": 3.293087971274686e-05, + "loss": 0.2813, + "step": 26671 + }, + { + "epoch": 1.03, + "learning_rate": 3.293023852269813e-05, + "loss": 0.4499, + "step": 26672 + }, + { + "epoch": 1.03, + "learning_rate": 3.29295973326494e-05, + "loss": 0.299, + "step": 26673 + }, + { + "epoch": 1.03, + "learning_rate": 3.292895614260067e-05, + "loss": 0.4043, + "step": 26674 + }, + { + "epoch": 1.03, + "learning_rate": 3.2928314952551934e-05, + "loss": 0.5013, + "step": 26675 + }, + { + "epoch": 1.03, + "learning_rate": 3.2927673762503206e-05, + "loss": 0.4809, + "step": 26676 + }, + { + "epoch": 1.03, + "learning_rate": 3.292703257245448e-05, + "loss": 0.5012, + "step": 26677 + }, + { + "epoch": 1.03, + "learning_rate": 3.292639138240575e-05, + "loss": 0.3585, + "step": 26678 + }, + { + "epoch": 1.03, + "learning_rate": 3.292575019235702e-05, + "loss": 0.3807, + "step": 26679 + }, + { + "epoch": 1.03, + "learning_rate": 3.2925109002308286e-05, + "loss": 0.4375, + "step": 26680 + }, + { + "epoch": 1.03, + "learning_rate": 3.292446781225956e-05, + "loss": 0.4302, + "step": 26681 + }, + { + "epoch": 1.03, + "learning_rate": 3.292382662221082e-05, + "loss": 0.4159, + "step": 26682 + }, + { + "epoch": 1.03, + "learning_rate": 3.2923185432162094e-05, + "loss": 0.4113, + "step": 26683 + }, + { + "epoch": 1.03, + "learning_rate": 3.2922544242113366e-05, + "loss": 0.2851, + "step": 26684 + }, + { + "epoch": 1.03, + "learning_rate": 3.292190305206464e-05, + "loss": 0.445, + "step": 26685 + }, + { + "epoch": 1.03, + "learning_rate": 3.29212618620159e-05, + "loss": 0.3476, + "step": 26686 + }, + { + "epoch": 1.03, + "learning_rate": 3.2920620671967174e-05, + "loss": 0.3468, + "step": 26687 + }, + { + "epoch": 1.03, + "learning_rate": 3.2919979481918446e-05, + "loss": 0.3285, + "step": 26688 + }, + { + "epoch": 1.03, + "learning_rate": 3.291933829186971e-05, + "loss": 0.3223, + "step": 26689 + }, + { + "epoch": 1.03, + "learning_rate": 3.291869710182098e-05, + "loss": 0.4198, + "step": 26690 + }, + { + "epoch": 1.03, + "learning_rate": 3.291805591177225e-05, + "loss": 0.4271, + "step": 26691 + }, + { + "epoch": 1.03, + "learning_rate": 3.291741472172352e-05, + "loss": 0.4323, + "step": 26692 + }, + { + "epoch": 1.03, + "learning_rate": 3.291677353167479e-05, + "loss": 0.4223, + "step": 26693 + }, + { + "epoch": 1.03, + "learning_rate": 3.291613234162606e-05, + "loss": 0.3653, + "step": 26694 + }, + { + "epoch": 1.03, + "learning_rate": 3.291549115157733e-05, + "loss": 0.334, + "step": 26695 + }, + { + "epoch": 1.03, + "learning_rate": 3.29148499615286e-05, + "loss": 0.5074, + "step": 26696 + }, + { + "epoch": 1.03, + "learning_rate": 3.291420877147987e-05, + "loss": 0.3918, + "step": 26697 + }, + { + "epoch": 1.03, + "learning_rate": 3.2913567581431136e-05, + "loss": 0.5022, + "step": 26698 + }, + { + "epoch": 1.03, + "learning_rate": 3.291292639138241e-05, + "loss": 0.383, + "step": 26699 + }, + { + "epoch": 1.03, + "learning_rate": 3.291228520133367e-05, + "loss": 0.3378, + "step": 26700 + }, + { + "epoch": 1.03, + "learning_rate": 3.291164401128495e-05, + "loss": 0.3253, + "step": 26701 + }, + { + "epoch": 1.03, + "learning_rate": 3.2911002821236216e-05, + "loss": 0.3804, + "step": 26702 + }, + { + "epoch": 1.03, + "learning_rate": 3.291036163118749e-05, + "loss": 0.487, + "step": 26703 + }, + { + "epoch": 1.03, + "learning_rate": 3.290972044113875e-05, + "loss": 0.352, + "step": 26704 + }, + { + "epoch": 1.03, + "learning_rate": 3.2909079251090024e-05, + "loss": 0.4148, + "step": 26705 + }, + { + "epoch": 1.03, + "learning_rate": 3.2908438061041296e-05, + "loss": 0.4765, + "step": 26706 + }, + { + "epoch": 1.03, + "learning_rate": 3.290779687099256e-05, + "loss": 0.4027, + "step": 26707 + }, + { + "epoch": 1.03, + "learning_rate": 3.290715568094383e-05, + "loss": 0.3906, + "step": 26708 + }, + { + "epoch": 1.03, + "learning_rate": 3.2906514490895104e-05, + "loss": 0.3261, + "step": 26709 + }, + { + "epoch": 1.03, + "learning_rate": 3.2905873300846376e-05, + "loss": 0.2795, + "step": 26710 + }, + { + "epoch": 1.03, + "learning_rate": 3.290523211079764e-05, + "loss": 0.4397, + "step": 26711 + }, + { + "epoch": 1.03, + "learning_rate": 3.290459092074891e-05, + "loss": 0.4469, + "step": 26712 + }, + { + "epoch": 1.03, + "learning_rate": 3.290394973070018e-05, + "loss": 0.3304, + "step": 26713 + }, + { + "epoch": 1.03, + "learning_rate": 3.290330854065145e-05, + "loss": 0.4327, + "step": 26714 + }, + { + "epoch": 1.03, + "learning_rate": 3.290266735060272e-05, + "loss": 0.3933, + "step": 26715 + }, + { + "epoch": 1.03, + "learning_rate": 3.2902026160553986e-05, + "loss": 0.4661, + "step": 26716 + }, + { + "epoch": 1.03, + "learning_rate": 3.2901384970505265e-05, + "loss": 0.4233, + "step": 26717 + }, + { + "epoch": 1.03, + "learning_rate": 3.290074378045653e-05, + "loss": 0.4802, + "step": 26718 + }, + { + "epoch": 1.03, + "learning_rate": 3.29001025904078e-05, + "loss": 0.2953, + "step": 26719 + }, + { + "epoch": 1.03, + "learning_rate": 3.2899461400359066e-05, + "loss": 0.3317, + "step": 26720 + }, + { + "epoch": 1.03, + "learning_rate": 3.289882021031034e-05, + "loss": 0.463, + "step": 26721 + }, + { + "epoch": 1.03, + "learning_rate": 3.28981790202616e-05, + "loss": 0.4675, + "step": 26722 + }, + { + "epoch": 1.03, + "learning_rate": 3.2897537830212875e-05, + "loss": 0.3188, + "step": 26723 + }, + { + "epoch": 1.03, + "learning_rate": 3.2896896640164146e-05, + "loss": 0.2878, + "step": 26724 + }, + { + "epoch": 1.03, + "learning_rate": 3.289625545011542e-05, + "loss": 0.3064, + "step": 26725 + }, + { + "epoch": 1.03, + "learning_rate": 3.289561426006669e-05, + "loss": 0.4506, + "step": 26726 + }, + { + "epoch": 1.03, + "learning_rate": 3.2894973070017955e-05, + "loss": 0.4762, + "step": 26727 + }, + { + "epoch": 1.03, + "learning_rate": 3.2894331879969226e-05, + "loss": 0.4148, + "step": 26728 + }, + { + "epoch": 1.03, + "learning_rate": 3.289369068992049e-05, + "loss": 0.412, + "step": 26729 + }, + { + "epoch": 1.03, + "learning_rate": 3.289304949987176e-05, + "loss": 0.4865, + "step": 26730 + }, + { + "epoch": 1.03, + "learning_rate": 3.289240830982303e-05, + "loss": 0.3421, + "step": 26731 + }, + { + "epoch": 1.03, + "learning_rate": 3.28917671197743e-05, + "loss": 0.4559, + "step": 26732 + }, + { + "epoch": 1.03, + "learning_rate": 3.289112592972557e-05, + "loss": 0.4769, + "step": 26733 + }, + { + "epoch": 1.03, + "learning_rate": 3.289048473967684e-05, + "loss": 0.4676, + "step": 26734 + }, + { + "epoch": 1.03, + "learning_rate": 3.2889843549628115e-05, + "loss": 0.383, + "step": 26735 + }, + { + "epoch": 1.03, + "learning_rate": 3.288920235957938e-05, + "loss": 0.4088, + "step": 26736 + }, + { + "epoch": 1.03, + "learning_rate": 3.288856116953065e-05, + "loss": 0.3927, + "step": 26737 + }, + { + "epoch": 1.03, + "learning_rate": 3.2887919979481916e-05, + "loss": 0.4052, + "step": 26738 + }, + { + "epoch": 1.03, + "learning_rate": 3.288727878943319e-05, + "loss": 0.5232, + "step": 26739 + }, + { + "epoch": 1.03, + "learning_rate": 3.288663759938446e-05, + "loss": 0.3281, + "step": 26740 + }, + { + "epoch": 1.03, + "learning_rate": 3.288599640933573e-05, + "loss": 0.533, + "step": 26741 + }, + { + "epoch": 1.03, + "learning_rate": 3.2885355219287e-05, + "loss": 0.4188, + "step": 26742 + }, + { + "epoch": 1.03, + "learning_rate": 3.288471402923827e-05, + "loss": 0.5383, + "step": 26743 + }, + { + "epoch": 1.03, + "learning_rate": 3.288407283918954e-05, + "loss": 0.4424, + "step": 26744 + }, + { + "epoch": 1.03, + "learning_rate": 3.2883431649140805e-05, + "loss": 0.5168, + "step": 26745 + }, + { + "epoch": 1.03, + "learning_rate": 3.2882790459092077e-05, + "loss": 0.4133, + "step": 26746 + }, + { + "epoch": 1.03, + "learning_rate": 3.288214926904334e-05, + "loss": 0.3791, + "step": 26747 + }, + { + "epoch": 1.03, + "learning_rate": 3.288150807899461e-05, + "loss": 0.4372, + "step": 26748 + }, + { + "epoch": 1.03, + "learning_rate": 3.2880866888945885e-05, + "loss": 0.3917, + "step": 26749 + }, + { + "epoch": 1.03, + "learning_rate": 3.288022569889716e-05, + "loss": 0.4968, + "step": 26750 + }, + { + "epoch": 1.03, + "learning_rate": 3.287958450884843e-05, + "loss": 0.4482, + "step": 26751 + }, + { + "epoch": 1.03, + "learning_rate": 3.287894331879969e-05, + "loss": 0.4291, + "step": 26752 + }, + { + "epoch": 1.03, + "learning_rate": 3.2878302128750965e-05, + "loss": 0.4335, + "step": 26753 + }, + { + "epoch": 1.03, + "learning_rate": 3.287766093870223e-05, + "loss": 0.3369, + "step": 26754 + }, + { + "epoch": 1.03, + "learning_rate": 3.28770197486535e-05, + "loss": 0.3734, + "step": 26755 + }, + { + "epoch": 1.03, + "learning_rate": 3.2876378558604773e-05, + "loss": 0.3309, + "step": 26756 + }, + { + "epoch": 1.03, + "learning_rate": 3.2875737368556045e-05, + "loss": 0.5033, + "step": 26757 + }, + { + "epoch": 1.03, + "learning_rate": 3.287509617850731e-05, + "loss": 0.4131, + "step": 26758 + }, + { + "epoch": 1.03, + "learning_rate": 3.287445498845858e-05, + "loss": 0.4153, + "step": 26759 + }, + { + "epoch": 1.03, + "learning_rate": 3.2873813798409854e-05, + "loss": 0.2885, + "step": 26760 + }, + { + "epoch": 1.03, + "learning_rate": 3.287317260836112e-05, + "loss": 0.3977, + "step": 26761 + }, + { + "epoch": 1.03, + "learning_rate": 3.287253141831239e-05, + "loss": 0.4824, + "step": 26762 + }, + { + "epoch": 1.03, + "learning_rate": 3.2871890228263655e-05, + "loss": 0.2704, + "step": 26763 + }, + { + "epoch": 1.03, + "learning_rate": 3.287124903821493e-05, + "loss": 0.3796, + "step": 26764 + }, + { + "epoch": 1.03, + "learning_rate": 3.28706078481662e-05, + "loss": 0.405, + "step": 26765 + }, + { + "epoch": 1.03, + "learning_rate": 3.286996665811747e-05, + "loss": 0.3532, + "step": 26766 + }, + { + "epoch": 1.03, + "learning_rate": 3.2869325468068735e-05, + "loss": 0.5142, + "step": 26767 + }, + { + "epoch": 1.03, + "learning_rate": 3.286868427802001e-05, + "loss": 0.3391, + "step": 26768 + }, + { + "epoch": 1.03, + "learning_rate": 3.286804308797128e-05, + "loss": 0.3604, + "step": 26769 + }, + { + "epoch": 1.03, + "learning_rate": 3.2867401897922544e-05, + "loss": 0.3547, + "step": 26770 + }, + { + "epoch": 1.03, + "learning_rate": 3.2866760707873815e-05, + "loss": 0.3473, + "step": 26771 + }, + { + "epoch": 1.03, + "learning_rate": 3.286611951782508e-05, + "loss": 0.3298, + "step": 26772 + }, + { + "epoch": 1.03, + "learning_rate": 3.286547832777636e-05, + "loss": 0.4211, + "step": 26773 + }, + { + "epoch": 1.03, + "learning_rate": 3.2864837137727624e-05, + "loss": 0.4456, + "step": 26774 + }, + { + "epoch": 1.03, + "learning_rate": 3.2864195947678895e-05, + "loss": 0.3777, + "step": 26775 + }, + { + "epoch": 1.03, + "learning_rate": 3.286355475763016e-05, + "loss": 0.4568, + "step": 26776 + }, + { + "epoch": 1.03, + "learning_rate": 3.286291356758143e-05, + "loss": 0.4341, + "step": 26777 + }, + { + "epoch": 1.03, + "learning_rate": 3.2862272377532704e-05, + "loss": 0.2942, + "step": 26778 + }, + { + "epoch": 1.03, + "learning_rate": 3.286163118748397e-05, + "loss": 0.4175, + "step": 26779 + }, + { + "epoch": 1.03, + "learning_rate": 3.286098999743524e-05, + "loss": 0.3961, + "step": 26780 + }, + { + "epoch": 1.03, + "learning_rate": 3.286034880738651e-05, + "loss": 0.3815, + "step": 26781 + }, + { + "epoch": 1.03, + "learning_rate": 3.2859707617337784e-05, + "loss": 0.421, + "step": 26782 + }, + { + "epoch": 1.03, + "learning_rate": 3.285906642728905e-05, + "loss": 0.395, + "step": 26783 + }, + { + "epoch": 1.03, + "learning_rate": 3.285842523724032e-05, + "loss": 0.494, + "step": 26784 + }, + { + "epoch": 1.03, + "learning_rate": 3.2857784047191585e-05, + "loss": 0.3877, + "step": 26785 + }, + { + "epoch": 1.03, + "learning_rate": 3.285714285714286e-05, + "loss": 0.4376, + "step": 26786 + }, + { + "epoch": 1.03, + "learning_rate": 3.285650166709413e-05, + "loss": 0.3202, + "step": 26787 + }, + { + "epoch": 1.03, + "learning_rate": 3.2855860477045394e-05, + "loss": 0.4464, + "step": 26788 + }, + { + "epoch": 1.03, + "learning_rate": 3.285521928699667e-05, + "loss": 0.3206, + "step": 26789 + }, + { + "epoch": 1.03, + "learning_rate": 3.285457809694794e-05, + "loss": 0.3704, + "step": 26790 + }, + { + "epoch": 1.03, + "learning_rate": 3.285393690689921e-05, + "loss": 0.3887, + "step": 26791 + }, + { + "epoch": 1.03, + "learning_rate": 3.2853295716850474e-05, + "loss": 0.4695, + "step": 26792 + }, + { + "epoch": 1.03, + "learning_rate": 3.2852654526801746e-05, + "loss": 0.4618, + "step": 26793 + }, + { + "epoch": 1.03, + "learning_rate": 3.285201333675301e-05, + "loss": 0.3194, + "step": 26794 + }, + { + "epoch": 1.03, + "learning_rate": 3.285137214670428e-05, + "loss": 0.4125, + "step": 26795 + }, + { + "epoch": 1.03, + "learning_rate": 3.2850730956655554e-05, + "loss": 0.4599, + "step": 26796 + }, + { + "epoch": 1.03, + "learning_rate": 3.2850089766606826e-05, + "loss": 0.2636, + "step": 26797 + }, + { + "epoch": 1.03, + "learning_rate": 3.28494485765581e-05, + "loss": 0.4156, + "step": 26798 + }, + { + "epoch": 1.03, + "learning_rate": 3.284880738650936e-05, + "loss": 0.4318, + "step": 26799 + }, + { + "epoch": 1.03, + "learning_rate": 3.2848166196460634e-05, + "loss": 0.4805, + "step": 26800 + }, + { + "epoch": 1.03, + "learning_rate": 3.28475250064119e-05, + "loss": 0.4023, + "step": 26801 + }, + { + "epoch": 1.03, + "learning_rate": 3.284688381636317e-05, + "loss": 0.3093, + "step": 26802 + }, + { + "epoch": 1.03, + "learning_rate": 3.2846242626314436e-05, + "loss": 0.4058, + "step": 26803 + }, + { + "epoch": 1.03, + "learning_rate": 3.284560143626571e-05, + "loss": 0.4315, + "step": 26804 + }, + { + "epoch": 1.03, + "learning_rate": 3.2844960246216986e-05, + "loss": 0.3825, + "step": 26805 + }, + { + "epoch": 1.03, + "learning_rate": 3.284431905616825e-05, + "loss": 0.3383, + "step": 26806 + }, + { + "epoch": 1.03, + "learning_rate": 3.284367786611952e-05, + "loss": 0.497, + "step": 26807 + }, + { + "epoch": 1.03, + "learning_rate": 3.284303667607079e-05, + "loss": 0.5166, + "step": 26808 + }, + { + "epoch": 1.03, + "learning_rate": 3.284239548602206e-05, + "loss": 0.3382, + "step": 26809 + }, + { + "epoch": 1.03, + "learning_rate": 3.2841754295973324e-05, + "loss": 0.3666, + "step": 26810 + }, + { + "epoch": 1.03, + "learning_rate": 3.2841113105924596e-05, + "loss": 0.4187, + "step": 26811 + }, + { + "epoch": 1.03, + "learning_rate": 3.284047191587587e-05, + "loss": 0.4737, + "step": 26812 + }, + { + "epoch": 1.03, + "learning_rate": 3.283983072582714e-05, + "loss": 0.3678, + "step": 26813 + }, + { + "epoch": 1.03, + "learning_rate": 3.283918953577841e-05, + "loss": 0.3982, + "step": 26814 + }, + { + "epoch": 1.03, + "learning_rate": 3.2838548345729676e-05, + "loss": 0.4629, + "step": 26815 + }, + { + "epoch": 1.03, + "learning_rate": 3.283790715568095e-05, + "loss": 0.3882, + "step": 26816 + }, + { + "epoch": 1.03, + "learning_rate": 3.283726596563221e-05, + "loss": 0.5143, + "step": 26817 + }, + { + "epoch": 1.03, + "learning_rate": 3.2836624775583484e-05, + "loss": 0.5714, + "step": 26818 + }, + { + "epoch": 1.03, + "learning_rate": 3.283598358553475e-05, + "loss": 0.3777, + "step": 26819 + }, + { + "epoch": 1.03, + "learning_rate": 3.283534239548602e-05, + "loss": 0.4806, + "step": 26820 + }, + { + "epoch": 1.03, + "learning_rate": 3.283470120543729e-05, + "loss": 0.3488, + "step": 26821 + }, + { + "epoch": 1.03, + "learning_rate": 3.2834060015388564e-05, + "loss": 0.3976, + "step": 26822 + }, + { + "epoch": 1.03, + "learning_rate": 3.2833418825339836e-05, + "loss": 0.3602, + "step": 26823 + }, + { + "epoch": 1.03, + "learning_rate": 3.28327776352911e-05, + "loss": 0.4231, + "step": 26824 + }, + { + "epoch": 1.03, + "learning_rate": 3.283213644524237e-05, + "loss": 0.419, + "step": 26825 + }, + { + "epoch": 1.03, + "learning_rate": 3.283149525519364e-05, + "loss": 0.3728, + "step": 26826 + }, + { + "epoch": 1.03, + "learning_rate": 3.283085406514491e-05, + "loss": 0.4016, + "step": 26827 + }, + { + "epoch": 1.03, + "learning_rate": 3.283021287509618e-05, + "loss": 0.4826, + "step": 26828 + }, + { + "epoch": 1.03, + "learning_rate": 3.282957168504745e-05, + "loss": 0.4385, + "step": 26829 + }, + { + "epoch": 1.03, + "learning_rate": 3.282893049499872e-05, + "loss": 0.5214, + "step": 26830 + }, + { + "epoch": 1.03, + "learning_rate": 3.282828930494999e-05, + "loss": 0.4348, + "step": 26831 + }, + { + "epoch": 1.03, + "learning_rate": 3.282764811490126e-05, + "loss": 0.3633, + "step": 26832 + }, + { + "epoch": 1.03, + "learning_rate": 3.2827006924852526e-05, + "loss": 0.4055, + "step": 26833 + }, + { + "epoch": 1.03, + "learning_rate": 3.28263657348038e-05, + "loss": 0.4604, + "step": 26834 + }, + { + "epoch": 1.03, + "learning_rate": 3.282572454475506e-05, + "loss": 0.3593, + "step": 26835 + }, + { + "epoch": 1.03, + "learning_rate": 3.2825083354706335e-05, + "loss": 0.3604, + "step": 26836 + }, + { + "epoch": 1.03, + "learning_rate": 3.2824442164657606e-05, + "loss": 0.4807, + "step": 26837 + }, + { + "epoch": 1.03, + "learning_rate": 3.282380097460888e-05, + "loss": 0.4927, + "step": 26838 + }, + { + "epoch": 1.03, + "learning_rate": 3.282315978456015e-05, + "loss": 0.5092, + "step": 26839 + }, + { + "epoch": 1.03, + "learning_rate": 3.2822518594511415e-05, + "loss": 0.4208, + "step": 26840 + }, + { + "epoch": 1.03, + "learning_rate": 3.2821877404462686e-05, + "loss": 0.447, + "step": 26841 + }, + { + "epoch": 1.03, + "learning_rate": 3.282123621441395e-05, + "loss": 0.3984, + "step": 26842 + }, + { + "epoch": 1.03, + "learning_rate": 3.282059502436522e-05, + "loss": 0.4147, + "step": 26843 + }, + { + "epoch": 1.03, + "learning_rate": 3.2819953834316495e-05, + "loss": 0.4101, + "step": 26844 + }, + { + "epoch": 1.03, + "learning_rate": 3.2819312644267766e-05, + "loss": 0.367, + "step": 26845 + }, + { + "epoch": 1.03, + "learning_rate": 3.281867145421903e-05, + "loss": 0.4505, + "step": 26846 + }, + { + "epoch": 1.03, + "learning_rate": 3.28180302641703e-05, + "loss": 0.5194, + "step": 26847 + }, + { + "epoch": 1.03, + "learning_rate": 3.2817389074121575e-05, + "loss": 0.2573, + "step": 26848 + }, + { + "epoch": 1.03, + "learning_rate": 3.281674788407284e-05, + "loss": 0.586, + "step": 26849 + }, + { + "epoch": 1.03, + "learning_rate": 3.281610669402411e-05, + "loss": 0.3373, + "step": 26850 + }, + { + "epoch": 1.03, + "learning_rate": 3.2815465503975376e-05, + "loss": 0.3561, + "step": 26851 + }, + { + "epoch": 1.03, + "learning_rate": 3.281482431392665e-05, + "loss": 0.4165, + "step": 26852 + }, + { + "epoch": 1.03, + "learning_rate": 3.281418312387792e-05, + "loss": 0.3674, + "step": 26853 + }, + { + "epoch": 1.03, + "learning_rate": 3.281354193382919e-05, + "loss": 0.3825, + "step": 26854 + }, + { + "epoch": 1.03, + "learning_rate": 3.2812900743780457e-05, + "loss": 0.3498, + "step": 26855 + }, + { + "epoch": 1.03, + "learning_rate": 3.281225955373173e-05, + "loss": 0.3121, + "step": 26856 + }, + { + "epoch": 1.03, + "learning_rate": 3.2811618363683e-05, + "loss": 0.343, + "step": 26857 + }, + { + "epoch": 1.03, + "learning_rate": 3.2810977173634265e-05, + "loss": 0.507, + "step": 26858 + }, + { + "epoch": 1.03, + "learning_rate": 3.281033598358554e-05, + "loss": 0.4029, + "step": 26859 + }, + { + "epoch": 1.03, + "learning_rate": 3.28096947935368e-05, + "loss": 0.3965, + "step": 26860 + }, + { + "epoch": 1.03, + "learning_rate": 3.280905360348808e-05, + "loss": 0.3561, + "step": 26861 + }, + { + "epoch": 1.03, + "learning_rate": 3.2808412413439345e-05, + "loss": 0.358, + "step": 26862 + }, + { + "epoch": 1.03, + "learning_rate": 3.280777122339062e-05, + "loss": 0.5536, + "step": 26863 + }, + { + "epoch": 1.03, + "learning_rate": 3.280713003334188e-05, + "loss": 0.2682, + "step": 26864 + }, + { + "epoch": 1.03, + "learning_rate": 3.280648884329315e-05, + "loss": 0.4268, + "step": 26865 + }, + { + "epoch": 1.03, + "learning_rate": 3.2805847653244425e-05, + "loss": 0.3951, + "step": 26866 + }, + { + "epoch": 1.03, + "learning_rate": 3.280520646319569e-05, + "loss": 0.3811, + "step": 26867 + }, + { + "epoch": 1.03, + "learning_rate": 3.280456527314696e-05, + "loss": 0.357, + "step": 26868 + }, + { + "epoch": 1.03, + "learning_rate": 3.2803924083098233e-05, + "loss": 0.4344, + "step": 26869 + }, + { + "epoch": 1.03, + "learning_rate": 3.2803282893049505e-05, + "loss": 0.4481, + "step": 26870 + }, + { + "epoch": 1.03, + "learning_rate": 3.280264170300077e-05, + "loss": 0.4526, + "step": 26871 + }, + { + "epoch": 1.03, + "learning_rate": 3.280200051295204e-05, + "loss": 0.3113, + "step": 26872 + }, + { + "epoch": 1.03, + "learning_rate": 3.280135932290331e-05, + "loss": 0.4288, + "step": 26873 + }, + { + "epoch": 1.03, + "learning_rate": 3.280071813285458e-05, + "loss": 0.3926, + "step": 26874 + }, + { + "epoch": 1.03, + "learning_rate": 3.280007694280585e-05, + "loss": 0.404, + "step": 26875 + }, + { + "epoch": 1.03, + "learning_rate": 3.2799435752757115e-05, + "loss": 0.4342, + "step": 26876 + }, + { + "epoch": 1.03, + "learning_rate": 3.2798794562708394e-05, + "loss": 0.3392, + "step": 26877 + }, + { + "epoch": 1.03, + "learning_rate": 3.279815337265966e-05, + "loss": 0.4444, + "step": 26878 + }, + { + "epoch": 1.03, + "learning_rate": 3.279751218261093e-05, + "loss": 0.6062, + "step": 26879 + }, + { + "epoch": 1.03, + "learning_rate": 3.2796870992562195e-05, + "loss": 0.4184, + "step": 26880 + }, + { + "epoch": 1.03, + "learning_rate": 3.279622980251347e-05, + "loss": 0.3375, + "step": 26881 + }, + { + "epoch": 1.03, + "learning_rate": 3.279558861246473e-05, + "loss": 0.4213, + "step": 26882 + }, + { + "epoch": 1.03, + "learning_rate": 3.2794947422416004e-05, + "loss": 0.4269, + "step": 26883 + }, + { + "epoch": 1.03, + "learning_rate": 3.2794306232367275e-05, + "loss": 0.3689, + "step": 26884 + }, + { + "epoch": 1.03, + "learning_rate": 3.279366504231855e-05, + "loss": 0.5679, + "step": 26885 + }, + { + "epoch": 1.03, + "learning_rate": 3.279302385226982e-05, + "loss": 0.5241, + "step": 26886 + }, + { + "epoch": 1.03, + "learning_rate": 3.2792382662221084e-05, + "loss": 0.4083, + "step": 26887 + }, + { + "epoch": 1.03, + "learning_rate": 3.2791741472172355e-05, + "loss": 0.4483, + "step": 26888 + }, + { + "epoch": 1.03, + "learning_rate": 3.279110028212362e-05, + "loss": 0.4276, + "step": 26889 + }, + { + "epoch": 1.03, + "learning_rate": 3.279045909207489e-05, + "loss": 0.4816, + "step": 26890 + }, + { + "epoch": 1.03, + "learning_rate": 3.278981790202616e-05, + "loss": 0.6377, + "step": 26891 + }, + { + "epoch": 1.03, + "learning_rate": 3.278917671197743e-05, + "loss": 0.3172, + "step": 26892 + }, + { + "epoch": 1.03, + "learning_rate": 3.27885355219287e-05, + "loss": 0.3425, + "step": 26893 + }, + { + "epoch": 1.03, + "learning_rate": 3.278789433187997e-05, + "loss": 0.4243, + "step": 26894 + }, + { + "epoch": 1.03, + "learning_rate": 3.2787253141831244e-05, + "loss": 0.3823, + "step": 26895 + }, + { + "epoch": 1.03, + "learning_rate": 3.278661195178251e-05, + "loss": 0.4901, + "step": 26896 + }, + { + "epoch": 1.03, + "learning_rate": 3.278597076173378e-05, + "loss": 0.6127, + "step": 26897 + }, + { + "epoch": 1.03, + "learning_rate": 3.2785329571685045e-05, + "loss": 0.4376, + "step": 26898 + }, + { + "epoch": 1.03, + "learning_rate": 3.278468838163632e-05, + "loss": 0.3381, + "step": 26899 + }, + { + "epoch": 1.03, + "learning_rate": 3.278404719158759e-05, + "loss": 0.4463, + "step": 26900 + }, + { + "epoch": 1.03, + "learning_rate": 3.278340600153886e-05, + "loss": 0.4951, + "step": 26901 + }, + { + "epoch": 1.03, + "learning_rate": 3.278276481149013e-05, + "loss": 0.3131, + "step": 26902 + }, + { + "epoch": 1.03, + "learning_rate": 3.27821236214414e-05, + "loss": 0.4388, + "step": 26903 + }, + { + "epoch": 1.03, + "learning_rate": 3.278148243139267e-05, + "loss": 0.438, + "step": 26904 + }, + { + "epoch": 1.03, + "learning_rate": 3.2780841241343934e-05, + "loss": 0.3296, + "step": 26905 + }, + { + "epoch": 1.03, + "learning_rate": 3.2780200051295206e-05, + "loss": 0.3296, + "step": 26906 + }, + { + "epoch": 1.03, + "learning_rate": 3.277955886124647e-05, + "loss": 0.343, + "step": 26907 + }, + { + "epoch": 1.03, + "learning_rate": 3.277891767119774e-05, + "loss": 0.4247, + "step": 26908 + }, + { + "epoch": 1.03, + "learning_rate": 3.2778276481149014e-05, + "loss": 0.434, + "step": 26909 + }, + { + "epoch": 1.03, + "learning_rate": 3.2777635291100286e-05, + "loss": 0.513, + "step": 26910 + }, + { + "epoch": 1.03, + "learning_rate": 3.277699410105156e-05, + "loss": 0.4439, + "step": 26911 + }, + { + "epoch": 1.03, + "learning_rate": 3.277635291100282e-05, + "loss": 0.4372, + "step": 26912 + }, + { + "epoch": 1.03, + "learning_rate": 3.2775711720954094e-05, + "loss": 0.4511, + "step": 26913 + }, + { + "epoch": 1.03, + "learning_rate": 3.277507053090536e-05, + "loss": 0.477, + "step": 26914 + }, + { + "epoch": 1.03, + "learning_rate": 3.277442934085663e-05, + "loss": 0.471, + "step": 26915 + }, + { + "epoch": 1.03, + "learning_rate": 3.27737881508079e-05, + "loss": 0.3805, + "step": 26916 + }, + { + "epoch": 1.03, + "learning_rate": 3.2773146960759174e-05, + "loss": 0.3775, + "step": 26917 + }, + { + "epoch": 1.03, + "learning_rate": 3.277250577071044e-05, + "loss": 0.3852, + "step": 26918 + }, + { + "epoch": 1.03, + "learning_rate": 3.277186458066171e-05, + "loss": 0.4174, + "step": 26919 + }, + { + "epoch": 1.03, + "learning_rate": 3.277122339061298e-05, + "loss": 0.3489, + "step": 26920 + }, + { + "epoch": 1.04, + "learning_rate": 3.277058220056425e-05, + "loss": 0.4671, + "step": 26921 + }, + { + "epoch": 1.04, + "learning_rate": 3.276994101051552e-05, + "loss": 0.38, + "step": 26922 + }, + { + "epoch": 1.04, + "learning_rate": 3.2769299820466784e-05, + "loss": 0.4414, + "step": 26923 + }, + { + "epoch": 1.04, + "learning_rate": 3.2768658630418056e-05, + "loss": 0.4648, + "step": 26924 + }, + { + "epoch": 1.04, + "learning_rate": 3.276801744036933e-05, + "loss": 0.4614, + "step": 26925 + }, + { + "epoch": 1.04, + "learning_rate": 3.27673762503206e-05, + "loss": 0.3857, + "step": 26926 + }, + { + "epoch": 1.04, + "learning_rate": 3.2766735060271864e-05, + "loss": 0.365, + "step": 26927 + }, + { + "epoch": 1.04, + "learning_rate": 3.2766093870223136e-05, + "loss": 0.4254, + "step": 26928 + }, + { + "epoch": 1.04, + "learning_rate": 3.276545268017441e-05, + "loss": 0.4215, + "step": 26929 + }, + { + "epoch": 1.04, + "learning_rate": 3.276481149012567e-05, + "loss": 0.4122, + "step": 26930 + }, + { + "epoch": 1.04, + "learning_rate": 3.2764170300076944e-05, + "loss": 0.4942, + "step": 26931 + }, + { + "epoch": 1.04, + "learning_rate": 3.276352911002821e-05, + "loss": 0.3827, + "step": 26932 + }, + { + "epoch": 1.04, + "learning_rate": 3.276288791997949e-05, + "loss": 0.344, + "step": 26933 + }, + { + "epoch": 1.04, + "learning_rate": 3.276224672993075e-05, + "loss": 0.4065, + "step": 26934 + }, + { + "epoch": 1.04, + "learning_rate": 3.2761605539882024e-05, + "loss": 0.4088, + "step": 26935 + }, + { + "epoch": 1.04, + "learning_rate": 3.276096434983329e-05, + "loss": 0.2651, + "step": 26936 + }, + { + "epoch": 1.04, + "learning_rate": 3.276032315978456e-05, + "loss": 0.3481, + "step": 26937 + }, + { + "epoch": 1.04, + "learning_rate": 3.275968196973583e-05, + "loss": 0.4072, + "step": 26938 + }, + { + "epoch": 1.04, + "learning_rate": 3.27590407796871e-05, + "loss": 0.4157, + "step": 26939 + }, + { + "epoch": 1.04, + "learning_rate": 3.275839958963837e-05, + "loss": 0.4351, + "step": 26940 + }, + { + "epoch": 1.04, + "learning_rate": 3.275775839958964e-05, + "loss": 0.3941, + "step": 26941 + }, + { + "epoch": 1.04, + "learning_rate": 3.275711720954091e-05, + "loss": 0.417, + "step": 26942 + }, + { + "epoch": 1.04, + "learning_rate": 3.275647601949218e-05, + "loss": 0.395, + "step": 26943 + }, + { + "epoch": 1.04, + "learning_rate": 3.275583482944345e-05, + "loss": 0.4441, + "step": 26944 + }, + { + "epoch": 1.04, + "learning_rate": 3.2755193639394715e-05, + "loss": 0.4112, + "step": 26945 + }, + { + "epoch": 1.04, + "learning_rate": 3.2754552449345986e-05, + "loss": 0.4974, + "step": 26946 + }, + { + "epoch": 1.04, + "learning_rate": 3.275391125929726e-05, + "loss": 0.4121, + "step": 26947 + }, + { + "epoch": 1.04, + "learning_rate": 3.275327006924852e-05, + "loss": 0.4368, + "step": 26948 + }, + { + "epoch": 1.04, + "learning_rate": 3.27526288791998e-05, + "loss": 0.4153, + "step": 26949 + }, + { + "epoch": 1.04, + "learning_rate": 3.2751987689151066e-05, + "loss": 0.427, + "step": 26950 + }, + { + "epoch": 1.04, + "learning_rate": 3.275134649910234e-05, + "loss": 0.3489, + "step": 26951 + }, + { + "epoch": 1.04, + "learning_rate": 3.27507053090536e-05, + "loss": 0.3719, + "step": 26952 + }, + { + "epoch": 1.04, + "learning_rate": 3.2750064119004875e-05, + "loss": 0.4866, + "step": 26953 + }, + { + "epoch": 1.04, + "learning_rate": 3.274942292895614e-05, + "loss": 0.5014, + "step": 26954 + }, + { + "epoch": 1.04, + "learning_rate": 3.274878173890741e-05, + "loss": 0.4586, + "step": 26955 + }, + { + "epoch": 1.04, + "learning_rate": 3.274814054885868e-05, + "loss": 0.4213, + "step": 26956 + }, + { + "epoch": 1.04, + "learning_rate": 3.2747499358809955e-05, + "loss": 0.5767, + "step": 26957 + }, + { + "epoch": 1.04, + "learning_rate": 3.2746858168761227e-05, + "loss": 0.297, + "step": 26958 + }, + { + "epoch": 1.04, + "learning_rate": 3.274621697871249e-05, + "loss": 0.3754, + "step": 26959 + }, + { + "epoch": 1.04, + "learning_rate": 3.274557578866376e-05, + "loss": 0.3995, + "step": 26960 + }, + { + "epoch": 1.04, + "learning_rate": 3.274493459861503e-05, + "loss": 0.5063, + "step": 26961 + }, + { + "epoch": 1.04, + "learning_rate": 3.27442934085663e-05, + "loss": 0.363, + "step": 26962 + }, + { + "epoch": 1.04, + "learning_rate": 3.2743652218517565e-05, + "loss": 0.3967, + "step": 26963 + }, + { + "epoch": 1.04, + "learning_rate": 3.2743011028468836e-05, + "loss": 0.2848, + "step": 26964 + }, + { + "epoch": 1.04, + "learning_rate": 3.2742369838420115e-05, + "loss": 0.3362, + "step": 26965 + }, + { + "epoch": 1.04, + "learning_rate": 3.274172864837138e-05, + "loss": 0.3666, + "step": 26966 + }, + { + "epoch": 1.04, + "learning_rate": 3.274108745832265e-05, + "loss": 0.3385, + "step": 26967 + }, + { + "epoch": 1.04, + "learning_rate": 3.2740446268273917e-05, + "loss": 0.4335, + "step": 26968 + }, + { + "epoch": 1.04, + "learning_rate": 3.273980507822519e-05, + "loss": 0.5218, + "step": 26969 + }, + { + "epoch": 1.04, + "learning_rate": 3.273916388817645e-05, + "loss": 0.445, + "step": 26970 + }, + { + "epoch": 1.04, + "learning_rate": 3.2738522698127725e-05, + "loss": 0.4129, + "step": 26971 + }, + { + "epoch": 1.04, + "learning_rate": 3.2737881508079e-05, + "loss": 0.3798, + "step": 26972 + }, + { + "epoch": 1.04, + "learning_rate": 3.273724031803027e-05, + "loss": 0.4402, + "step": 26973 + }, + { + "epoch": 1.04, + "learning_rate": 3.273659912798154e-05, + "loss": 0.3605, + "step": 26974 + }, + { + "epoch": 1.04, + "learning_rate": 3.2735957937932805e-05, + "loss": 0.3721, + "step": 26975 + }, + { + "epoch": 1.04, + "learning_rate": 3.273531674788408e-05, + "loss": 0.3845, + "step": 26976 + }, + { + "epoch": 1.04, + "learning_rate": 3.273467555783534e-05, + "loss": 0.4856, + "step": 26977 + }, + { + "epoch": 1.04, + "learning_rate": 3.2734034367786613e-05, + "loss": 0.401, + "step": 26978 + }, + { + "epoch": 1.04, + "learning_rate": 3.273339317773788e-05, + "loss": 0.4966, + "step": 26979 + }, + { + "epoch": 1.04, + "learning_rate": 3.273275198768915e-05, + "loss": 0.3786, + "step": 26980 + }, + { + "epoch": 1.04, + "learning_rate": 3.273211079764042e-05, + "loss": 0.354, + "step": 26981 + }, + { + "epoch": 1.04, + "learning_rate": 3.2731469607591693e-05, + "loss": 0.436, + "step": 26982 + }, + { + "epoch": 1.04, + "learning_rate": 3.2730828417542965e-05, + "loss": 0.4799, + "step": 26983 + }, + { + "epoch": 1.04, + "learning_rate": 3.273018722749423e-05, + "loss": 0.3764, + "step": 26984 + }, + { + "epoch": 1.04, + "learning_rate": 3.27295460374455e-05, + "loss": 0.2489, + "step": 26985 + }, + { + "epoch": 1.04, + "learning_rate": 3.272890484739677e-05, + "loss": 0.4712, + "step": 26986 + }, + { + "epoch": 1.04, + "learning_rate": 3.272826365734804e-05, + "loss": 0.3781, + "step": 26987 + }, + { + "epoch": 1.04, + "learning_rate": 3.272762246729931e-05, + "loss": 0.4499, + "step": 26988 + }, + { + "epoch": 1.04, + "learning_rate": 3.272698127725058e-05, + "loss": 0.2946, + "step": 26989 + }, + { + "epoch": 1.04, + "learning_rate": 3.272634008720185e-05, + "loss": 0.3168, + "step": 26990 + }, + { + "epoch": 1.04, + "learning_rate": 3.272569889715312e-05, + "loss": 0.3772, + "step": 26991 + }, + { + "epoch": 1.04, + "learning_rate": 3.272505770710439e-05, + "loss": 0.3967, + "step": 26992 + }, + { + "epoch": 1.04, + "learning_rate": 3.2724416517055655e-05, + "loss": 0.3992, + "step": 26993 + }, + { + "epoch": 1.04, + "learning_rate": 3.272377532700693e-05, + "loss": 0.3612, + "step": 26994 + }, + { + "epoch": 1.04, + "learning_rate": 3.272313413695819e-05, + "loss": 0.4126, + "step": 26995 + }, + { + "epoch": 1.04, + "learning_rate": 3.2722492946909464e-05, + "loss": 0.3271, + "step": 26996 + }, + { + "epoch": 1.04, + "learning_rate": 3.2721851756860735e-05, + "loss": 0.5988, + "step": 26997 + }, + { + "epoch": 1.04, + "learning_rate": 3.272121056681201e-05, + "loss": 0.273, + "step": 26998 + }, + { + "epoch": 1.04, + "learning_rate": 3.272056937676327e-05, + "loss": 0.383, + "step": 26999 + }, + { + "epoch": 1.04, + "learning_rate": 3.2719928186714544e-05, + "loss": 0.3812, + "step": 27000 + }, + { + "epoch": 1.04, + "learning_rate": 3.2719286996665815e-05, + "loss": 0.4111, + "step": 27001 + }, + { + "epoch": 1.04, + "learning_rate": 3.271864580661708e-05, + "loss": 0.3717, + "step": 27002 + }, + { + "epoch": 1.04, + "learning_rate": 3.271800461656835e-05, + "loss": 0.3849, + "step": 27003 + }, + { + "epoch": 1.04, + "learning_rate": 3.2717363426519624e-05, + "loss": 0.4063, + "step": 27004 + }, + { + "epoch": 1.04, + "learning_rate": 3.2716722236470896e-05, + "loss": 0.4391, + "step": 27005 + }, + { + "epoch": 1.04, + "learning_rate": 3.271608104642216e-05, + "loss": 0.4931, + "step": 27006 + }, + { + "epoch": 1.04, + "learning_rate": 3.271543985637343e-05, + "loss": 0.3999, + "step": 27007 + }, + { + "epoch": 1.04, + "learning_rate": 3.27147986663247e-05, + "loss": 0.5024, + "step": 27008 + }, + { + "epoch": 1.04, + "learning_rate": 3.271415747627597e-05, + "loss": 0.5278, + "step": 27009 + }, + { + "epoch": 1.04, + "learning_rate": 3.271351628622724e-05, + "loss": 0.4839, + "step": 27010 + }, + { + "epoch": 1.04, + "learning_rate": 3.2712875096178506e-05, + "loss": 0.439, + "step": 27011 + }, + { + "epoch": 1.04, + "learning_rate": 3.271223390612978e-05, + "loss": 0.4163, + "step": 27012 + }, + { + "epoch": 1.04, + "learning_rate": 3.271159271608105e-05, + "loss": 0.3858, + "step": 27013 + }, + { + "epoch": 1.04, + "learning_rate": 3.271095152603232e-05, + "loss": 0.2512, + "step": 27014 + }, + { + "epoch": 1.04, + "learning_rate": 3.2710310335983586e-05, + "loss": 0.5233, + "step": 27015 + }, + { + "epoch": 1.04, + "learning_rate": 3.270966914593486e-05, + "loss": 0.4716, + "step": 27016 + }, + { + "epoch": 1.04, + "learning_rate": 3.270902795588612e-05, + "loss": 0.4545, + "step": 27017 + }, + { + "epoch": 1.04, + "learning_rate": 3.2708386765837394e-05, + "loss": 0.49, + "step": 27018 + }, + { + "epoch": 1.04, + "learning_rate": 3.2707745575788666e-05, + "loss": 0.3205, + "step": 27019 + }, + { + "epoch": 1.04, + "learning_rate": 3.270710438573993e-05, + "loss": 0.3569, + "step": 27020 + }, + { + "epoch": 1.04, + "learning_rate": 3.270646319569121e-05, + "loss": 0.4328, + "step": 27021 + }, + { + "epoch": 1.04, + "learning_rate": 3.2705822005642474e-05, + "loss": 0.4393, + "step": 27022 + }, + { + "epoch": 1.04, + "learning_rate": 3.2705180815593746e-05, + "loss": 0.3399, + "step": 27023 + }, + { + "epoch": 1.04, + "learning_rate": 3.270453962554501e-05, + "loss": 0.3897, + "step": 27024 + }, + { + "epoch": 1.04, + "learning_rate": 3.270389843549628e-05, + "loss": 0.4119, + "step": 27025 + }, + { + "epoch": 1.04, + "learning_rate": 3.2703257245447554e-05, + "loss": 0.4221, + "step": 27026 + }, + { + "epoch": 1.04, + "learning_rate": 3.270261605539882e-05, + "loss": 0.3447, + "step": 27027 + }, + { + "epoch": 1.04, + "learning_rate": 3.270197486535009e-05, + "loss": 0.4467, + "step": 27028 + }, + { + "epoch": 1.04, + "learning_rate": 3.270133367530136e-05, + "loss": 0.4253, + "step": 27029 + }, + { + "epoch": 1.04, + "learning_rate": 3.2700692485252634e-05, + "loss": 0.5185, + "step": 27030 + }, + { + "epoch": 1.04, + "learning_rate": 3.27000512952039e-05, + "loss": 0.374, + "step": 27031 + }, + { + "epoch": 1.04, + "learning_rate": 3.269941010515517e-05, + "loss": 0.3788, + "step": 27032 + }, + { + "epoch": 1.04, + "learning_rate": 3.2698768915106436e-05, + "loss": 0.3285, + "step": 27033 + }, + { + "epoch": 1.04, + "learning_rate": 3.269812772505771e-05, + "loss": 0.423, + "step": 27034 + }, + { + "epoch": 1.04, + "learning_rate": 3.269748653500898e-05, + "loss": 0.4526, + "step": 27035 + }, + { + "epoch": 1.04, + "learning_rate": 3.2696845344960244e-05, + "loss": 0.508, + "step": 27036 + }, + { + "epoch": 1.04, + "learning_rate": 3.269620415491152e-05, + "loss": 0.3798, + "step": 27037 + }, + { + "epoch": 1.04, + "learning_rate": 3.269556296486279e-05, + "loss": 0.4095, + "step": 27038 + }, + { + "epoch": 1.04, + "learning_rate": 3.269492177481406e-05, + "loss": 0.4305, + "step": 27039 + }, + { + "epoch": 1.04, + "learning_rate": 3.2694280584765324e-05, + "loss": 0.3681, + "step": 27040 + }, + { + "epoch": 1.04, + "learning_rate": 3.2693639394716596e-05, + "loss": 0.384, + "step": 27041 + }, + { + "epoch": 1.04, + "learning_rate": 3.269299820466786e-05, + "loss": 0.4366, + "step": 27042 + }, + { + "epoch": 1.04, + "learning_rate": 3.269235701461913e-05, + "loss": 0.3974, + "step": 27043 + }, + { + "epoch": 1.04, + "learning_rate": 3.2691715824570404e-05, + "loss": 0.5111, + "step": 27044 + }, + { + "epoch": 1.04, + "learning_rate": 3.2691074634521676e-05, + "loss": 0.3356, + "step": 27045 + }, + { + "epoch": 1.04, + "learning_rate": 3.269043344447295e-05, + "loss": 0.3568, + "step": 27046 + }, + { + "epoch": 1.04, + "learning_rate": 3.268979225442421e-05, + "loss": 0.4387, + "step": 27047 + }, + { + "epoch": 1.04, + "learning_rate": 3.2689151064375484e-05, + "loss": 0.4019, + "step": 27048 + }, + { + "epoch": 1.04, + "learning_rate": 3.268850987432675e-05, + "loss": 0.4186, + "step": 27049 + }, + { + "epoch": 1.04, + "learning_rate": 3.268786868427802e-05, + "loss": 0.4865, + "step": 27050 + }, + { + "epoch": 1.04, + "learning_rate": 3.2687227494229286e-05, + "loss": 0.3746, + "step": 27051 + }, + { + "epoch": 1.04, + "learning_rate": 3.268658630418056e-05, + "loss": 0.3776, + "step": 27052 + }, + { + "epoch": 1.04, + "learning_rate": 3.268594511413183e-05, + "loss": 0.3231, + "step": 27053 + }, + { + "epoch": 1.04, + "learning_rate": 3.26853039240831e-05, + "loss": 0.4206, + "step": 27054 + }, + { + "epoch": 1.04, + "learning_rate": 3.268466273403437e-05, + "loss": 0.5008, + "step": 27055 + }, + { + "epoch": 1.04, + "learning_rate": 3.268402154398564e-05, + "loss": 0.4061, + "step": 27056 + }, + { + "epoch": 1.04, + "learning_rate": 3.268338035393691e-05, + "loss": 0.527, + "step": 27057 + }, + { + "epoch": 1.04, + "learning_rate": 3.2682739163888175e-05, + "loss": 0.4127, + "step": 27058 + }, + { + "epoch": 1.04, + "learning_rate": 3.2682097973839446e-05, + "loss": 0.3627, + "step": 27059 + }, + { + "epoch": 1.04, + "learning_rate": 3.268145678379072e-05, + "loss": 0.3517, + "step": 27060 + }, + { + "epoch": 1.04, + "learning_rate": 3.268081559374199e-05, + "loss": 0.4983, + "step": 27061 + }, + { + "epoch": 1.04, + "learning_rate": 3.268017440369326e-05, + "loss": 0.3349, + "step": 27062 + }, + { + "epoch": 1.04, + "learning_rate": 3.2679533213644526e-05, + "loss": 0.4339, + "step": 27063 + }, + { + "epoch": 1.04, + "learning_rate": 3.26788920235958e-05, + "loss": 0.4323, + "step": 27064 + }, + { + "epoch": 1.04, + "learning_rate": 3.267825083354706e-05, + "loss": 0.3392, + "step": 27065 + }, + { + "epoch": 1.04, + "learning_rate": 3.2677609643498335e-05, + "loss": 0.4459, + "step": 27066 + }, + { + "epoch": 1.04, + "learning_rate": 3.26769684534496e-05, + "loss": 0.3603, + "step": 27067 + }, + { + "epoch": 1.04, + "learning_rate": 3.267632726340087e-05, + "loss": 0.2958, + "step": 27068 + }, + { + "epoch": 1.04, + "learning_rate": 3.267568607335214e-05, + "loss": 0.5743, + "step": 27069 + }, + { + "epoch": 1.04, + "learning_rate": 3.2675044883303415e-05, + "loss": 0.3403, + "step": 27070 + }, + { + "epoch": 1.04, + "learning_rate": 3.2674403693254687e-05, + "loss": 0.4547, + "step": 27071 + }, + { + "epoch": 1.04, + "learning_rate": 3.267376250320595e-05, + "loss": 0.4081, + "step": 27072 + }, + { + "epoch": 1.04, + "learning_rate": 3.267312131315722e-05, + "loss": 0.4488, + "step": 27073 + }, + { + "epoch": 1.04, + "learning_rate": 3.267248012310849e-05, + "loss": 0.4213, + "step": 27074 + }, + { + "epoch": 1.04, + "learning_rate": 3.267183893305976e-05, + "loss": 0.3432, + "step": 27075 + }, + { + "epoch": 1.04, + "learning_rate": 3.267119774301103e-05, + "loss": 0.3833, + "step": 27076 + }, + { + "epoch": 1.04, + "learning_rate": 3.26705565529623e-05, + "loss": 0.3598, + "step": 27077 + }, + { + "epoch": 1.04, + "learning_rate": 3.266991536291357e-05, + "loss": 0.344, + "step": 27078 + }, + { + "epoch": 1.04, + "learning_rate": 3.266927417286484e-05, + "loss": 0.4216, + "step": 27079 + }, + { + "epoch": 1.04, + "learning_rate": 3.266863298281611e-05, + "loss": 0.4831, + "step": 27080 + }, + { + "epoch": 1.04, + "learning_rate": 3.2667991792767377e-05, + "loss": 0.4118, + "step": 27081 + }, + { + "epoch": 1.04, + "learning_rate": 3.266735060271865e-05, + "loss": 0.3619, + "step": 27082 + }, + { + "epoch": 1.04, + "learning_rate": 3.266670941266991e-05, + "loss": 0.3947, + "step": 27083 + }, + { + "epoch": 1.04, + "learning_rate": 3.2666068222621185e-05, + "loss": 0.4355, + "step": 27084 + }, + { + "epoch": 1.04, + "learning_rate": 3.266542703257246e-05, + "loss": 0.4543, + "step": 27085 + }, + { + "epoch": 1.04, + "learning_rate": 3.266478584252373e-05, + "loss": 0.4038, + "step": 27086 + }, + { + "epoch": 1.04, + "learning_rate": 3.266414465247499e-05, + "loss": 0.6548, + "step": 27087 + }, + { + "epoch": 1.04, + "learning_rate": 3.2663503462426265e-05, + "loss": 0.3625, + "step": 27088 + }, + { + "epoch": 1.04, + "learning_rate": 3.266286227237754e-05, + "loss": 0.4762, + "step": 27089 + }, + { + "epoch": 1.04, + "learning_rate": 3.26622210823288e-05, + "loss": 0.3281, + "step": 27090 + }, + { + "epoch": 1.04, + "learning_rate": 3.2661579892280073e-05, + "loss": 0.4277, + "step": 27091 + }, + { + "epoch": 1.04, + "learning_rate": 3.2660938702231345e-05, + "loss": 0.3628, + "step": 27092 + }, + { + "epoch": 1.04, + "learning_rate": 3.266029751218262e-05, + "loss": 0.4704, + "step": 27093 + }, + { + "epoch": 1.04, + "learning_rate": 3.265965632213388e-05, + "loss": 0.4216, + "step": 27094 + }, + { + "epoch": 1.04, + "learning_rate": 3.2659015132085154e-05, + "loss": 0.3874, + "step": 27095 + }, + { + "epoch": 1.04, + "learning_rate": 3.265837394203642e-05, + "loss": 0.3946, + "step": 27096 + }, + { + "epoch": 1.04, + "learning_rate": 3.265773275198769e-05, + "loss": 0.4819, + "step": 27097 + }, + { + "epoch": 1.04, + "learning_rate": 3.265709156193896e-05, + "loss": 0.5474, + "step": 27098 + }, + { + "epoch": 1.04, + "learning_rate": 3.265645037189023e-05, + "loss": 0.4728, + "step": 27099 + }, + { + "epoch": 1.04, + "learning_rate": 3.26558091818415e-05, + "loss": 0.4122, + "step": 27100 + }, + { + "epoch": 1.04, + "learning_rate": 3.265516799179277e-05, + "loss": 0.4027, + "step": 27101 + }, + { + "epoch": 1.04, + "learning_rate": 3.265452680174404e-05, + "loss": 0.4755, + "step": 27102 + }, + { + "epoch": 1.04, + "learning_rate": 3.265388561169531e-05, + "loss": 0.5627, + "step": 27103 + }, + { + "epoch": 1.04, + "learning_rate": 3.265324442164658e-05, + "loss": 0.333, + "step": 27104 + }, + { + "epoch": 1.04, + "learning_rate": 3.2652603231597844e-05, + "loss": 0.4861, + "step": 27105 + }, + { + "epoch": 1.04, + "learning_rate": 3.2651962041549115e-05, + "loss": 0.4047, + "step": 27106 + }, + { + "epoch": 1.04, + "learning_rate": 3.265132085150039e-05, + "loss": 0.376, + "step": 27107 + }, + { + "epoch": 1.04, + "learning_rate": 3.265067966145165e-05, + "loss": 0.4696, + "step": 27108 + }, + { + "epoch": 1.04, + "learning_rate": 3.265003847140293e-05, + "loss": 0.4666, + "step": 27109 + }, + { + "epoch": 1.04, + "learning_rate": 3.2649397281354195e-05, + "loss": 0.3239, + "step": 27110 + }, + { + "epoch": 1.04, + "learning_rate": 3.264875609130547e-05, + "loss": 0.4039, + "step": 27111 + }, + { + "epoch": 1.04, + "learning_rate": 3.264811490125673e-05, + "loss": 0.3637, + "step": 27112 + }, + { + "epoch": 1.04, + "learning_rate": 3.2647473711208004e-05, + "loss": 0.4168, + "step": 27113 + }, + { + "epoch": 1.04, + "learning_rate": 3.264683252115927e-05, + "loss": 0.3777, + "step": 27114 + }, + { + "epoch": 1.04, + "learning_rate": 3.264619133111054e-05, + "loss": 0.4013, + "step": 27115 + }, + { + "epoch": 1.04, + "learning_rate": 3.264555014106181e-05, + "loss": 0.4155, + "step": 27116 + }, + { + "epoch": 1.04, + "learning_rate": 3.2644908951013084e-05, + "loss": 0.3595, + "step": 27117 + }, + { + "epoch": 1.04, + "learning_rate": 3.2644267760964356e-05, + "loss": 0.4664, + "step": 27118 + }, + { + "epoch": 1.04, + "learning_rate": 3.264362657091562e-05, + "loss": 0.3991, + "step": 27119 + }, + { + "epoch": 1.04, + "learning_rate": 3.264298538086689e-05, + "loss": 0.4258, + "step": 27120 + }, + { + "epoch": 1.04, + "learning_rate": 3.264234419081816e-05, + "loss": 0.4005, + "step": 27121 + }, + { + "epoch": 1.04, + "learning_rate": 3.264170300076943e-05, + "loss": 0.4804, + "step": 27122 + }, + { + "epoch": 1.04, + "learning_rate": 3.2641061810720694e-05, + "loss": 0.3901, + "step": 27123 + }, + { + "epoch": 1.04, + "learning_rate": 3.2640420620671966e-05, + "loss": 0.4242, + "step": 27124 + }, + { + "epoch": 1.04, + "learning_rate": 3.2639779430623244e-05, + "loss": 0.3367, + "step": 27125 + }, + { + "epoch": 1.04, + "learning_rate": 3.263913824057451e-05, + "loss": 0.2788, + "step": 27126 + }, + { + "epoch": 1.04, + "learning_rate": 3.263849705052578e-05, + "loss": 0.2705, + "step": 27127 + }, + { + "epoch": 1.04, + "learning_rate": 3.2637855860477046e-05, + "loss": 0.4133, + "step": 27128 + }, + { + "epoch": 1.04, + "learning_rate": 3.263721467042832e-05, + "loss": 0.3536, + "step": 27129 + }, + { + "epoch": 1.04, + "learning_rate": 3.263657348037958e-05, + "loss": 0.4791, + "step": 27130 + }, + { + "epoch": 1.04, + "learning_rate": 3.2635932290330854e-05, + "loss": 0.3489, + "step": 27131 + }, + { + "epoch": 1.04, + "learning_rate": 3.2635291100282126e-05, + "loss": 0.594, + "step": 27132 + }, + { + "epoch": 1.04, + "learning_rate": 3.26346499102334e-05, + "loss": 0.3761, + "step": 27133 + }, + { + "epoch": 1.04, + "learning_rate": 3.263400872018467e-05, + "loss": 0.5291, + "step": 27134 + }, + { + "epoch": 1.04, + "learning_rate": 3.2633367530135934e-05, + "loss": 0.4001, + "step": 27135 + }, + { + "epoch": 1.04, + "learning_rate": 3.2632726340087206e-05, + "loss": 0.4935, + "step": 27136 + }, + { + "epoch": 1.04, + "learning_rate": 3.263208515003847e-05, + "loss": 0.3034, + "step": 27137 + }, + { + "epoch": 1.04, + "learning_rate": 3.263144395998974e-05, + "loss": 0.4628, + "step": 27138 + }, + { + "epoch": 1.04, + "learning_rate": 3.263080276994101e-05, + "loss": 0.3547, + "step": 27139 + }, + { + "epoch": 1.04, + "learning_rate": 3.263016157989228e-05, + "loss": 0.4756, + "step": 27140 + }, + { + "epoch": 1.04, + "learning_rate": 3.262952038984355e-05, + "loss": 0.4899, + "step": 27141 + }, + { + "epoch": 1.04, + "learning_rate": 3.262887919979482e-05, + "loss": 0.5146, + "step": 27142 + }, + { + "epoch": 1.04, + "learning_rate": 3.2628238009746094e-05, + "loss": 0.462, + "step": 27143 + }, + { + "epoch": 1.04, + "learning_rate": 3.262759681969736e-05, + "loss": 0.4295, + "step": 27144 + }, + { + "epoch": 1.04, + "learning_rate": 3.262695562964863e-05, + "loss": 0.3562, + "step": 27145 + }, + { + "epoch": 1.04, + "learning_rate": 3.2626314439599896e-05, + "loss": 0.5153, + "step": 27146 + }, + { + "epoch": 1.04, + "learning_rate": 3.262567324955117e-05, + "loss": 0.3415, + "step": 27147 + }, + { + "epoch": 1.04, + "learning_rate": 3.262503205950244e-05, + "loss": 0.375, + "step": 27148 + }, + { + "epoch": 1.04, + "learning_rate": 3.262439086945371e-05, + "loss": 0.2922, + "step": 27149 + }, + { + "epoch": 1.04, + "learning_rate": 3.2623749679404976e-05, + "loss": 0.3693, + "step": 27150 + }, + { + "epoch": 1.04, + "learning_rate": 3.262310848935625e-05, + "loss": 0.4085, + "step": 27151 + }, + { + "epoch": 1.04, + "learning_rate": 3.262246729930752e-05, + "loss": 0.5484, + "step": 27152 + }, + { + "epoch": 1.04, + "learning_rate": 3.2621826109258784e-05, + "loss": 0.4591, + "step": 27153 + }, + { + "epoch": 1.04, + "learning_rate": 3.2621184919210056e-05, + "loss": 0.3633, + "step": 27154 + }, + { + "epoch": 1.04, + "learning_rate": 3.262054372916132e-05, + "loss": 0.2771, + "step": 27155 + }, + { + "epoch": 1.04, + "learning_rate": 3.261990253911259e-05, + "loss": 0.2354, + "step": 27156 + }, + { + "epoch": 1.04, + "learning_rate": 3.2619261349063864e-05, + "loss": 0.3261, + "step": 27157 + }, + { + "epoch": 1.04, + "learning_rate": 3.2618620159015136e-05, + "loss": 0.3826, + "step": 27158 + }, + { + "epoch": 1.04, + "learning_rate": 3.26179789689664e-05, + "loss": 0.4102, + "step": 27159 + }, + { + "epoch": 1.04, + "learning_rate": 3.261733777891767e-05, + "loss": 0.386, + "step": 27160 + }, + { + "epoch": 1.04, + "learning_rate": 3.2616696588868945e-05, + "loss": 0.3105, + "step": 27161 + }, + { + "epoch": 1.04, + "learning_rate": 3.261605539882021e-05, + "loss": 0.4417, + "step": 27162 + }, + { + "epoch": 1.04, + "learning_rate": 3.261541420877148e-05, + "loss": 0.4839, + "step": 27163 + }, + { + "epoch": 1.04, + "learning_rate": 3.261477301872275e-05, + "loss": 0.4481, + "step": 27164 + }, + { + "epoch": 1.04, + "learning_rate": 3.2614131828674025e-05, + "loss": 0.3869, + "step": 27165 + }, + { + "epoch": 1.04, + "learning_rate": 3.261349063862529e-05, + "loss": 0.4333, + "step": 27166 + }, + { + "epoch": 1.04, + "learning_rate": 3.261284944857656e-05, + "loss": 0.572, + "step": 27167 + }, + { + "epoch": 1.04, + "learning_rate": 3.2612208258527826e-05, + "loss": 0.3734, + "step": 27168 + }, + { + "epoch": 1.04, + "learning_rate": 3.26115670684791e-05, + "loss": 0.3979, + "step": 27169 + }, + { + "epoch": 1.04, + "learning_rate": 3.261092587843037e-05, + "loss": 0.4403, + "step": 27170 + }, + { + "epoch": 1.04, + "learning_rate": 3.2610284688381635e-05, + "loss": 0.3774, + "step": 27171 + }, + { + "epoch": 1.04, + "learning_rate": 3.2609643498332906e-05, + "loss": 0.4534, + "step": 27172 + }, + { + "epoch": 1.04, + "learning_rate": 3.260900230828418e-05, + "loss": 0.3569, + "step": 27173 + }, + { + "epoch": 1.04, + "learning_rate": 3.260836111823545e-05, + "loss": 0.3838, + "step": 27174 + }, + { + "epoch": 1.04, + "learning_rate": 3.2607719928186715e-05, + "loss": 0.4207, + "step": 27175 + }, + { + "epoch": 1.04, + "learning_rate": 3.2607078738137986e-05, + "loss": 0.356, + "step": 27176 + }, + { + "epoch": 1.04, + "learning_rate": 3.260643754808925e-05, + "loss": 0.302, + "step": 27177 + }, + { + "epoch": 1.04, + "learning_rate": 3.260579635804052e-05, + "loss": 0.3598, + "step": 27178 + }, + { + "epoch": 1.04, + "learning_rate": 3.2605155167991795e-05, + "loss": 0.4986, + "step": 27179 + }, + { + "epoch": 1.04, + "learning_rate": 3.260451397794306e-05, + "loss": 0.3338, + "step": 27180 + }, + { + "epoch": 1.05, + "learning_rate": 3.260387278789434e-05, + "loss": 0.4288, + "step": 27181 + }, + { + "epoch": 1.05, + "learning_rate": 3.26032315978456e-05, + "loss": 0.4494, + "step": 27182 + }, + { + "epoch": 1.05, + "learning_rate": 3.2602590407796875e-05, + "loss": 0.5296, + "step": 27183 + }, + { + "epoch": 1.05, + "learning_rate": 3.260194921774814e-05, + "loss": 0.4642, + "step": 27184 + }, + { + "epoch": 1.05, + "learning_rate": 3.260130802769941e-05, + "loss": 0.4252, + "step": 27185 + }, + { + "epoch": 1.05, + "learning_rate": 3.2600666837650676e-05, + "loss": 0.3976, + "step": 27186 + }, + { + "epoch": 1.05, + "learning_rate": 3.260002564760195e-05, + "loss": 0.5429, + "step": 27187 + }, + { + "epoch": 1.05, + "learning_rate": 3.259938445755322e-05, + "loss": 0.3548, + "step": 27188 + }, + { + "epoch": 1.05, + "learning_rate": 3.259874326750449e-05, + "loss": 0.3675, + "step": 27189 + }, + { + "epoch": 1.05, + "learning_rate": 3.259810207745576e-05, + "loss": 0.5147, + "step": 27190 + }, + { + "epoch": 1.05, + "learning_rate": 3.259746088740703e-05, + "loss": 0.4053, + "step": 27191 + }, + { + "epoch": 1.05, + "learning_rate": 3.25968196973583e-05, + "loss": 0.47, + "step": 27192 + }, + { + "epoch": 1.05, + "learning_rate": 3.2596178507309565e-05, + "loss": 0.4214, + "step": 27193 + }, + { + "epoch": 1.05, + "learning_rate": 3.259553731726084e-05, + "loss": 0.4066, + "step": 27194 + }, + { + "epoch": 1.05, + "learning_rate": 3.25948961272121e-05, + "loss": 0.4012, + "step": 27195 + }, + { + "epoch": 1.05, + "learning_rate": 3.259425493716337e-05, + "loss": 0.5175, + "step": 27196 + }, + { + "epoch": 1.05, + "learning_rate": 3.259361374711465e-05, + "loss": 0.4263, + "step": 27197 + }, + { + "epoch": 1.05, + "learning_rate": 3.259297255706592e-05, + "loss": 0.3617, + "step": 27198 + }, + { + "epoch": 1.05, + "learning_rate": 3.259233136701719e-05, + "loss": 0.4226, + "step": 27199 + }, + { + "epoch": 1.05, + "learning_rate": 3.259169017696845e-05, + "loss": 0.4383, + "step": 27200 + }, + { + "epoch": 1.05, + "learning_rate": 3.2591048986919725e-05, + "loss": 0.3442, + "step": 27201 + }, + { + "epoch": 1.05, + "learning_rate": 3.259040779687099e-05, + "loss": 0.4124, + "step": 27202 + }, + { + "epoch": 1.05, + "learning_rate": 3.258976660682226e-05, + "loss": 0.5072, + "step": 27203 + }, + { + "epoch": 1.05, + "learning_rate": 3.2589125416773533e-05, + "loss": 0.4488, + "step": 27204 + }, + { + "epoch": 1.05, + "learning_rate": 3.2588484226724805e-05, + "loss": 0.4323, + "step": 27205 + }, + { + "epoch": 1.05, + "learning_rate": 3.258784303667608e-05, + "loss": 0.5721, + "step": 27206 + }, + { + "epoch": 1.05, + "learning_rate": 3.258720184662734e-05, + "loss": 0.3513, + "step": 27207 + }, + { + "epoch": 1.05, + "learning_rate": 3.2586560656578614e-05, + "loss": 0.5096, + "step": 27208 + }, + { + "epoch": 1.05, + "learning_rate": 3.258591946652988e-05, + "loss": 0.3729, + "step": 27209 + }, + { + "epoch": 1.05, + "learning_rate": 3.258527827648115e-05, + "loss": 0.3806, + "step": 27210 + }, + { + "epoch": 1.05, + "learning_rate": 3.2584637086432415e-05, + "loss": 0.4341, + "step": 27211 + }, + { + "epoch": 1.05, + "learning_rate": 3.258399589638369e-05, + "loss": 0.3233, + "step": 27212 + }, + { + "epoch": 1.05, + "learning_rate": 3.2583354706334965e-05, + "loss": 0.5745, + "step": 27213 + }, + { + "epoch": 1.05, + "learning_rate": 3.258271351628623e-05, + "loss": 0.4578, + "step": 27214 + }, + { + "epoch": 1.05, + "learning_rate": 3.25820723262375e-05, + "loss": 0.409, + "step": 27215 + }, + { + "epoch": 1.05, + "learning_rate": 3.258143113618877e-05, + "loss": 0.4847, + "step": 27216 + }, + { + "epoch": 1.05, + "learning_rate": 3.258078994614004e-05, + "loss": 0.4829, + "step": 27217 + }, + { + "epoch": 1.05, + "learning_rate": 3.2580148756091304e-05, + "loss": 0.3915, + "step": 27218 + }, + { + "epoch": 1.05, + "learning_rate": 3.2579507566042575e-05, + "loss": 0.3805, + "step": 27219 + }, + { + "epoch": 1.05, + "learning_rate": 3.257886637599385e-05, + "loss": 0.4246, + "step": 27220 + }, + { + "epoch": 1.05, + "learning_rate": 3.257822518594512e-05, + "loss": 0.4268, + "step": 27221 + }, + { + "epoch": 1.05, + "learning_rate": 3.257758399589639e-05, + "loss": 0.3497, + "step": 27222 + }, + { + "epoch": 1.05, + "learning_rate": 3.2576942805847655e-05, + "loss": 0.3601, + "step": 27223 + }, + { + "epoch": 1.05, + "learning_rate": 3.257630161579893e-05, + "loss": 0.4384, + "step": 27224 + }, + { + "epoch": 1.05, + "learning_rate": 3.257566042575019e-05, + "loss": 0.3415, + "step": 27225 + }, + { + "epoch": 1.05, + "learning_rate": 3.2575019235701464e-05, + "loss": 0.3541, + "step": 27226 + }, + { + "epoch": 1.05, + "learning_rate": 3.257437804565273e-05, + "loss": 0.3997, + "step": 27227 + }, + { + "epoch": 1.05, + "learning_rate": 3.2573736855604e-05, + "loss": 0.4377, + "step": 27228 + }, + { + "epoch": 1.05, + "learning_rate": 3.257309566555527e-05, + "loss": 0.5213, + "step": 27229 + }, + { + "epoch": 1.05, + "learning_rate": 3.2572454475506544e-05, + "loss": 0.3476, + "step": 27230 + }, + { + "epoch": 1.05, + "learning_rate": 3.2571813285457816e-05, + "loss": 0.4972, + "step": 27231 + }, + { + "epoch": 1.05, + "learning_rate": 3.257117209540908e-05, + "loss": 0.3123, + "step": 27232 + }, + { + "epoch": 1.05, + "learning_rate": 3.257053090536035e-05, + "loss": 0.5461, + "step": 27233 + }, + { + "epoch": 1.05, + "learning_rate": 3.256988971531162e-05, + "loss": 0.3801, + "step": 27234 + }, + { + "epoch": 1.05, + "learning_rate": 3.256924852526289e-05, + "loss": 0.3912, + "step": 27235 + }, + { + "epoch": 1.05, + "learning_rate": 3.256860733521416e-05, + "loss": 0.4033, + "step": 27236 + }, + { + "epoch": 1.05, + "learning_rate": 3.256796614516543e-05, + "loss": 0.4898, + "step": 27237 + }, + { + "epoch": 1.05, + "learning_rate": 3.25673249551167e-05, + "loss": 0.2957, + "step": 27238 + }, + { + "epoch": 1.05, + "learning_rate": 3.256668376506797e-05, + "loss": 0.4539, + "step": 27239 + }, + { + "epoch": 1.05, + "learning_rate": 3.256604257501924e-05, + "loss": 0.4197, + "step": 27240 + }, + { + "epoch": 1.05, + "learning_rate": 3.2565401384970506e-05, + "loss": 0.441, + "step": 27241 + }, + { + "epoch": 1.05, + "learning_rate": 3.256476019492178e-05, + "loss": 0.3503, + "step": 27242 + }, + { + "epoch": 1.05, + "learning_rate": 3.256411900487304e-05, + "loss": 0.5067, + "step": 27243 + }, + { + "epoch": 1.05, + "learning_rate": 3.2563477814824314e-05, + "loss": 0.33, + "step": 27244 + }, + { + "epoch": 1.05, + "learning_rate": 3.2562836624775586e-05, + "loss": 0.3475, + "step": 27245 + }, + { + "epoch": 1.05, + "learning_rate": 3.256219543472686e-05, + "loss": 0.4568, + "step": 27246 + }, + { + "epoch": 1.05, + "learning_rate": 3.256155424467812e-05, + "loss": 0.3541, + "step": 27247 + }, + { + "epoch": 1.05, + "learning_rate": 3.2560913054629394e-05, + "loss": 0.3959, + "step": 27248 + }, + { + "epoch": 1.05, + "learning_rate": 3.2560271864580666e-05, + "loss": 0.3442, + "step": 27249 + }, + { + "epoch": 1.05, + "learning_rate": 3.255963067453193e-05, + "loss": 0.3434, + "step": 27250 + }, + { + "epoch": 1.05, + "learning_rate": 3.25589894844832e-05, + "loss": 0.2341, + "step": 27251 + }, + { + "epoch": 1.05, + "learning_rate": 3.2558348294434474e-05, + "loss": 0.3215, + "step": 27252 + }, + { + "epoch": 1.05, + "learning_rate": 3.2557707104385746e-05, + "loss": 0.3512, + "step": 27253 + }, + { + "epoch": 1.05, + "learning_rate": 3.255706591433701e-05, + "loss": 0.3826, + "step": 27254 + }, + { + "epoch": 1.05, + "learning_rate": 3.255642472428828e-05, + "loss": 0.345, + "step": 27255 + }, + { + "epoch": 1.05, + "learning_rate": 3.255578353423955e-05, + "loss": 0.4433, + "step": 27256 + }, + { + "epoch": 1.05, + "learning_rate": 3.255514234419082e-05, + "loss": 0.3944, + "step": 27257 + }, + { + "epoch": 1.05, + "learning_rate": 3.255450115414209e-05, + "loss": 0.3721, + "step": 27258 + }, + { + "epoch": 1.05, + "learning_rate": 3.2553859964093356e-05, + "loss": 0.2666, + "step": 27259 + }, + { + "epoch": 1.05, + "learning_rate": 3.255321877404463e-05, + "loss": 0.4356, + "step": 27260 + }, + { + "epoch": 1.05, + "learning_rate": 3.25525775839959e-05, + "loss": 0.3508, + "step": 27261 + }, + { + "epoch": 1.05, + "learning_rate": 3.255193639394717e-05, + "loss": 0.5363, + "step": 27262 + }, + { + "epoch": 1.05, + "learning_rate": 3.2551295203898436e-05, + "loss": 0.3722, + "step": 27263 + }, + { + "epoch": 1.05, + "learning_rate": 3.255065401384971e-05, + "loss": 0.4279, + "step": 27264 + }, + { + "epoch": 1.05, + "learning_rate": 3.255001282380097e-05, + "loss": 0.356, + "step": 27265 + }, + { + "epoch": 1.05, + "learning_rate": 3.2549371633752244e-05, + "loss": 0.3706, + "step": 27266 + }, + { + "epoch": 1.05, + "learning_rate": 3.2548730443703516e-05, + "loss": 0.2582, + "step": 27267 + }, + { + "epoch": 1.05, + "learning_rate": 3.254808925365478e-05, + "loss": 0.3744, + "step": 27268 + }, + { + "epoch": 1.05, + "learning_rate": 3.254744806360606e-05, + "loss": 0.3432, + "step": 27269 + }, + { + "epoch": 1.05, + "learning_rate": 3.2546806873557324e-05, + "loss": 0.3182, + "step": 27270 + }, + { + "epoch": 1.05, + "learning_rate": 3.2546165683508596e-05, + "loss": 0.3901, + "step": 27271 + }, + { + "epoch": 1.05, + "learning_rate": 3.254552449345986e-05, + "loss": 0.5241, + "step": 27272 + }, + { + "epoch": 1.05, + "learning_rate": 3.254488330341113e-05, + "loss": 0.5188, + "step": 27273 + }, + { + "epoch": 1.05, + "learning_rate": 3.25442421133624e-05, + "loss": 0.5211, + "step": 27274 + }, + { + "epoch": 1.05, + "learning_rate": 3.254360092331367e-05, + "loss": 0.4493, + "step": 27275 + }, + { + "epoch": 1.05, + "learning_rate": 3.254295973326494e-05, + "loss": 0.4376, + "step": 27276 + }, + { + "epoch": 1.05, + "learning_rate": 3.254231854321621e-05, + "loss": 0.3652, + "step": 27277 + }, + { + "epoch": 1.05, + "learning_rate": 3.2541677353167485e-05, + "loss": 0.3165, + "step": 27278 + }, + { + "epoch": 1.05, + "learning_rate": 3.254103616311875e-05, + "loss": 0.319, + "step": 27279 + }, + { + "epoch": 1.05, + "learning_rate": 3.254039497307002e-05, + "loss": 0.3838, + "step": 27280 + }, + { + "epoch": 1.05, + "learning_rate": 3.2539753783021286e-05, + "loss": 0.3618, + "step": 27281 + }, + { + "epoch": 1.05, + "learning_rate": 3.253911259297256e-05, + "loss": 0.3386, + "step": 27282 + }, + { + "epoch": 1.05, + "learning_rate": 3.253847140292382e-05, + "loss": 0.4813, + "step": 27283 + }, + { + "epoch": 1.05, + "learning_rate": 3.2537830212875095e-05, + "loss": 0.4084, + "step": 27284 + }, + { + "epoch": 1.05, + "learning_rate": 3.253718902282637e-05, + "loss": 0.455, + "step": 27285 + }, + { + "epoch": 1.05, + "learning_rate": 3.253654783277764e-05, + "loss": 0.4312, + "step": 27286 + }, + { + "epoch": 1.05, + "learning_rate": 3.253590664272891e-05, + "loss": 0.332, + "step": 27287 + }, + { + "epoch": 1.05, + "learning_rate": 3.2535265452680175e-05, + "loss": 0.416, + "step": 27288 + }, + { + "epoch": 1.05, + "learning_rate": 3.2534624262631446e-05, + "loss": 0.2738, + "step": 27289 + }, + { + "epoch": 1.05, + "learning_rate": 3.253398307258271e-05, + "loss": 0.3175, + "step": 27290 + }, + { + "epoch": 1.05, + "learning_rate": 3.253334188253398e-05, + "loss": 0.3797, + "step": 27291 + }, + { + "epoch": 1.05, + "learning_rate": 3.2532700692485255e-05, + "loss": 0.469, + "step": 27292 + }, + { + "epoch": 1.05, + "learning_rate": 3.2532059502436527e-05, + "loss": 0.4339, + "step": 27293 + }, + { + "epoch": 1.05, + "learning_rate": 3.25314183123878e-05, + "loss": 0.4191, + "step": 27294 + }, + { + "epoch": 1.05, + "learning_rate": 3.253077712233906e-05, + "loss": 0.3713, + "step": 27295 + }, + { + "epoch": 1.05, + "learning_rate": 3.2530135932290335e-05, + "loss": 0.3594, + "step": 27296 + }, + { + "epoch": 1.05, + "learning_rate": 3.25294947422416e-05, + "loss": 0.4059, + "step": 27297 + }, + { + "epoch": 1.05, + "learning_rate": 3.252885355219287e-05, + "loss": 0.3706, + "step": 27298 + }, + { + "epoch": 1.05, + "learning_rate": 3.2528212362144136e-05, + "loss": 0.5524, + "step": 27299 + }, + { + "epoch": 1.05, + "learning_rate": 3.252757117209541e-05, + "loss": 0.3877, + "step": 27300 + }, + { + "epoch": 1.05, + "learning_rate": 3.252692998204668e-05, + "loss": 0.5079, + "step": 27301 + }, + { + "epoch": 1.05, + "learning_rate": 3.252628879199795e-05, + "loss": 0.4203, + "step": 27302 + }, + { + "epoch": 1.05, + "learning_rate": 3.252564760194922e-05, + "loss": 0.4544, + "step": 27303 + }, + { + "epoch": 1.05, + "learning_rate": 3.252500641190049e-05, + "loss": 0.3969, + "step": 27304 + }, + { + "epoch": 1.05, + "learning_rate": 3.252436522185176e-05, + "loss": 0.3482, + "step": 27305 + }, + { + "epoch": 1.05, + "learning_rate": 3.2523724031803025e-05, + "loss": 0.4232, + "step": 27306 + }, + { + "epoch": 1.05, + "learning_rate": 3.25230828417543e-05, + "loss": 0.3508, + "step": 27307 + }, + { + "epoch": 1.05, + "learning_rate": 3.252244165170557e-05, + "loss": 0.4886, + "step": 27308 + }, + { + "epoch": 1.05, + "learning_rate": 3.252180046165684e-05, + "loss": 0.3969, + "step": 27309 + }, + { + "epoch": 1.05, + "learning_rate": 3.2521159271608105e-05, + "loss": 0.4333, + "step": 27310 + }, + { + "epoch": 1.05, + "learning_rate": 3.252051808155938e-05, + "loss": 0.3625, + "step": 27311 + }, + { + "epoch": 1.05, + "learning_rate": 3.251987689151065e-05, + "loss": 0.2819, + "step": 27312 + }, + { + "epoch": 1.05, + "learning_rate": 3.2519235701461913e-05, + "loss": 0.4256, + "step": 27313 + }, + { + "epoch": 1.05, + "learning_rate": 3.2518594511413185e-05, + "loss": 0.4304, + "step": 27314 + }, + { + "epoch": 1.05, + "learning_rate": 3.251795332136445e-05, + "loss": 0.3333, + "step": 27315 + }, + { + "epoch": 1.05, + "learning_rate": 3.251731213131572e-05, + "loss": 0.4951, + "step": 27316 + }, + { + "epoch": 1.05, + "learning_rate": 3.2516670941266994e-05, + "loss": 0.5283, + "step": 27317 + }, + { + "epoch": 1.05, + "learning_rate": 3.2516029751218265e-05, + "loss": 0.4232, + "step": 27318 + }, + { + "epoch": 1.05, + "learning_rate": 3.251538856116953e-05, + "loss": 0.377, + "step": 27319 + }, + { + "epoch": 1.05, + "learning_rate": 3.25147473711208e-05, + "loss": 0.3176, + "step": 27320 + }, + { + "epoch": 1.05, + "learning_rate": 3.2514106181072074e-05, + "loss": 0.4154, + "step": 27321 + }, + { + "epoch": 1.05, + "learning_rate": 3.251346499102334e-05, + "loss": 0.5168, + "step": 27322 + }, + { + "epoch": 1.05, + "learning_rate": 3.251282380097461e-05, + "loss": 0.3593, + "step": 27323 + }, + { + "epoch": 1.05, + "learning_rate": 3.251218261092588e-05, + "loss": 0.5123, + "step": 27324 + }, + { + "epoch": 1.05, + "learning_rate": 3.2511541420877154e-05, + "loss": 0.391, + "step": 27325 + }, + { + "epoch": 1.05, + "learning_rate": 3.251090023082842e-05, + "loss": 0.3855, + "step": 27326 + }, + { + "epoch": 1.05, + "learning_rate": 3.251025904077969e-05, + "loss": 0.4245, + "step": 27327 + }, + { + "epoch": 1.05, + "learning_rate": 3.2509617850730955e-05, + "loss": 0.3274, + "step": 27328 + }, + { + "epoch": 1.05, + "learning_rate": 3.250897666068223e-05, + "loss": 0.4584, + "step": 27329 + }, + { + "epoch": 1.05, + "learning_rate": 3.25083354706335e-05, + "loss": 0.5078, + "step": 27330 + }, + { + "epoch": 1.05, + "learning_rate": 3.2507694280584764e-05, + "loss": 0.4299, + "step": 27331 + }, + { + "epoch": 1.05, + "learning_rate": 3.2507053090536035e-05, + "loss": 0.4629, + "step": 27332 + }, + { + "epoch": 1.05, + "learning_rate": 3.250641190048731e-05, + "loss": 0.5029, + "step": 27333 + }, + { + "epoch": 1.05, + "learning_rate": 3.250577071043858e-05, + "loss": 0.4903, + "step": 27334 + }, + { + "epoch": 1.05, + "learning_rate": 3.2505129520389844e-05, + "loss": 0.2384, + "step": 27335 + }, + { + "epoch": 1.05, + "learning_rate": 3.2504488330341115e-05, + "loss": 0.4623, + "step": 27336 + }, + { + "epoch": 1.05, + "learning_rate": 3.250384714029238e-05, + "loss": 0.3447, + "step": 27337 + }, + { + "epoch": 1.05, + "learning_rate": 3.250320595024365e-05, + "loss": 0.3624, + "step": 27338 + }, + { + "epoch": 1.05, + "learning_rate": 3.2502564760194924e-05, + "loss": 0.4075, + "step": 27339 + }, + { + "epoch": 1.05, + "learning_rate": 3.250192357014619e-05, + "loss": 0.3759, + "step": 27340 + }, + { + "epoch": 1.05, + "learning_rate": 3.250128238009747e-05, + "loss": 0.442, + "step": 27341 + }, + { + "epoch": 1.05, + "learning_rate": 3.250064119004873e-05, + "loss": 0.4154, + "step": 27342 + }, + { + "epoch": 1.05, + "learning_rate": 3.2500000000000004e-05, + "loss": 0.3613, + "step": 27343 + }, + { + "epoch": 1.05, + "learning_rate": 3.249935880995127e-05, + "loss": 0.3506, + "step": 27344 + }, + { + "epoch": 1.05, + "learning_rate": 3.249871761990254e-05, + "loss": 0.3761, + "step": 27345 + }, + { + "epoch": 1.05, + "learning_rate": 3.2498076429853806e-05, + "loss": 0.4688, + "step": 27346 + }, + { + "epoch": 1.05, + "learning_rate": 3.249743523980508e-05, + "loss": 0.3924, + "step": 27347 + }, + { + "epoch": 1.05, + "learning_rate": 3.249679404975635e-05, + "loss": 0.3979, + "step": 27348 + }, + { + "epoch": 1.05, + "learning_rate": 3.249615285970762e-05, + "loss": 0.5066, + "step": 27349 + }, + { + "epoch": 1.05, + "learning_rate": 3.249551166965889e-05, + "loss": 0.5061, + "step": 27350 + }, + { + "epoch": 1.05, + "learning_rate": 3.249487047961016e-05, + "loss": 0.4627, + "step": 27351 + }, + { + "epoch": 1.05, + "learning_rate": 3.249422928956143e-05, + "loss": 0.4599, + "step": 27352 + }, + { + "epoch": 1.05, + "learning_rate": 3.2493588099512694e-05, + "loss": 0.4195, + "step": 27353 + }, + { + "epoch": 1.05, + "learning_rate": 3.2492946909463966e-05, + "loss": 0.3846, + "step": 27354 + }, + { + "epoch": 1.05, + "learning_rate": 3.249230571941523e-05, + "loss": 0.3972, + "step": 27355 + }, + { + "epoch": 1.05, + "learning_rate": 3.24916645293665e-05, + "loss": 0.4228, + "step": 27356 + }, + { + "epoch": 1.05, + "learning_rate": 3.249102333931778e-05, + "loss": 0.4653, + "step": 27357 + }, + { + "epoch": 1.05, + "learning_rate": 3.2490382149269046e-05, + "loss": 0.332, + "step": 27358 + }, + { + "epoch": 1.05, + "learning_rate": 3.248974095922032e-05, + "loss": 0.3225, + "step": 27359 + }, + { + "epoch": 1.05, + "learning_rate": 3.248909976917158e-05, + "loss": 0.4191, + "step": 27360 + }, + { + "epoch": 1.05, + "learning_rate": 3.2488458579122854e-05, + "loss": 0.6365, + "step": 27361 + }, + { + "epoch": 1.05, + "learning_rate": 3.248781738907412e-05, + "loss": 0.4819, + "step": 27362 + }, + { + "epoch": 1.05, + "learning_rate": 3.248717619902539e-05, + "loss": 0.4071, + "step": 27363 + }, + { + "epoch": 1.05, + "learning_rate": 3.248653500897666e-05, + "loss": 0.4109, + "step": 27364 + }, + { + "epoch": 1.05, + "learning_rate": 3.2485893818927934e-05, + "loss": 0.3267, + "step": 27365 + }, + { + "epoch": 1.05, + "learning_rate": 3.2485252628879206e-05, + "loss": 0.4537, + "step": 27366 + }, + { + "epoch": 1.05, + "learning_rate": 3.248461143883047e-05, + "loss": 0.5018, + "step": 27367 + }, + { + "epoch": 1.05, + "learning_rate": 3.248397024878174e-05, + "loss": 0.4568, + "step": 27368 + }, + { + "epoch": 1.05, + "learning_rate": 3.248332905873301e-05, + "loss": 0.399, + "step": 27369 + }, + { + "epoch": 1.05, + "learning_rate": 3.248268786868428e-05, + "loss": 0.4127, + "step": 27370 + }, + { + "epoch": 1.05, + "learning_rate": 3.2482046678635544e-05, + "loss": 0.283, + "step": 27371 + }, + { + "epoch": 1.05, + "learning_rate": 3.2481405488586816e-05, + "loss": 0.5291, + "step": 27372 + }, + { + "epoch": 1.05, + "learning_rate": 3.248076429853809e-05, + "loss": 0.4241, + "step": 27373 + }, + { + "epoch": 1.05, + "learning_rate": 3.248012310848936e-05, + "loss": 0.3126, + "step": 27374 + }, + { + "epoch": 1.05, + "learning_rate": 3.247948191844063e-05, + "loss": 0.5699, + "step": 27375 + }, + { + "epoch": 1.05, + "learning_rate": 3.2478840728391896e-05, + "loss": 0.5476, + "step": 27376 + }, + { + "epoch": 1.05, + "learning_rate": 3.247819953834317e-05, + "loss": 0.4417, + "step": 27377 + }, + { + "epoch": 1.05, + "learning_rate": 3.247755834829443e-05, + "loss": 0.3659, + "step": 27378 + }, + { + "epoch": 1.05, + "learning_rate": 3.2476917158245704e-05, + "loss": 0.3916, + "step": 27379 + }, + { + "epoch": 1.05, + "learning_rate": 3.2476275968196976e-05, + "loss": 0.3419, + "step": 27380 + }, + { + "epoch": 1.05, + "learning_rate": 3.247563477814825e-05, + "loss": 0.4163, + "step": 27381 + }, + { + "epoch": 1.05, + "learning_rate": 3.247499358809951e-05, + "loss": 0.4441, + "step": 27382 + }, + { + "epoch": 1.05, + "learning_rate": 3.2474352398050785e-05, + "loss": 0.4012, + "step": 27383 + }, + { + "epoch": 1.05, + "learning_rate": 3.2473711208002056e-05, + "loss": 0.4574, + "step": 27384 + }, + { + "epoch": 1.05, + "learning_rate": 3.247307001795332e-05, + "loss": 0.4426, + "step": 27385 + }, + { + "epoch": 1.05, + "learning_rate": 3.247242882790459e-05, + "loss": 0.3372, + "step": 27386 + }, + { + "epoch": 1.05, + "learning_rate": 3.247178763785586e-05, + "loss": 0.3868, + "step": 27387 + }, + { + "epoch": 1.05, + "learning_rate": 3.247114644780713e-05, + "loss": 0.421, + "step": 27388 + }, + { + "epoch": 1.05, + "learning_rate": 3.24705052577584e-05, + "loss": 0.3938, + "step": 27389 + }, + { + "epoch": 1.05, + "learning_rate": 3.246986406770967e-05, + "loss": 0.4388, + "step": 27390 + }, + { + "epoch": 1.05, + "learning_rate": 3.246922287766094e-05, + "loss": 0.3877, + "step": 27391 + }, + { + "epoch": 1.05, + "learning_rate": 3.246858168761221e-05, + "loss": 0.3883, + "step": 27392 + }, + { + "epoch": 1.05, + "learning_rate": 3.246794049756348e-05, + "loss": 0.3831, + "step": 27393 + }, + { + "epoch": 1.05, + "learning_rate": 3.2467299307514746e-05, + "loss": 0.4764, + "step": 27394 + }, + { + "epoch": 1.05, + "learning_rate": 3.246665811746602e-05, + "loss": 0.3582, + "step": 27395 + }, + { + "epoch": 1.05, + "learning_rate": 3.246601692741729e-05, + "loss": 0.3268, + "step": 27396 + }, + { + "epoch": 1.05, + "learning_rate": 3.246537573736856e-05, + "loss": 0.2897, + "step": 27397 + }, + { + "epoch": 1.05, + "learning_rate": 3.2464734547319826e-05, + "loss": 0.3549, + "step": 27398 + }, + { + "epoch": 1.05, + "learning_rate": 3.24640933572711e-05, + "loss": 0.2979, + "step": 27399 + }, + { + "epoch": 1.05, + "learning_rate": 3.246345216722237e-05, + "loss": 0.3155, + "step": 27400 + }, + { + "epoch": 1.05, + "learning_rate": 3.2462810977173635e-05, + "loss": 0.3026, + "step": 27401 + }, + { + "epoch": 1.05, + "learning_rate": 3.2462169787124906e-05, + "loss": 0.4213, + "step": 27402 + }, + { + "epoch": 1.05, + "learning_rate": 3.246152859707617e-05, + "loss": 0.4097, + "step": 27403 + }, + { + "epoch": 1.05, + "learning_rate": 3.246088740702744e-05, + "loss": 0.4846, + "step": 27404 + }, + { + "epoch": 1.05, + "learning_rate": 3.2460246216978715e-05, + "loss": 0.5438, + "step": 27405 + }, + { + "epoch": 1.05, + "learning_rate": 3.2459605026929987e-05, + "loss": 0.4854, + "step": 27406 + }, + { + "epoch": 1.05, + "learning_rate": 3.245896383688125e-05, + "loss": 0.328, + "step": 27407 + }, + { + "epoch": 1.05, + "learning_rate": 3.245832264683252e-05, + "loss": 0.3858, + "step": 27408 + }, + { + "epoch": 1.05, + "learning_rate": 3.2457681456783795e-05, + "loss": 0.4938, + "step": 27409 + }, + { + "epoch": 1.05, + "learning_rate": 3.245704026673506e-05, + "loss": 0.316, + "step": 27410 + }, + { + "epoch": 1.05, + "learning_rate": 3.245639907668633e-05, + "loss": 0.525, + "step": 27411 + }, + { + "epoch": 1.05, + "learning_rate": 3.24557578866376e-05, + "loss": 0.3636, + "step": 27412 + }, + { + "epoch": 1.05, + "learning_rate": 3.2455116696588875e-05, + "loss": 0.4679, + "step": 27413 + }, + { + "epoch": 1.05, + "learning_rate": 3.245447550654014e-05, + "loss": 0.4228, + "step": 27414 + }, + { + "epoch": 1.05, + "learning_rate": 3.245383431649141e-05, + "loss": 0.3625, + "step": 27415 + }, + { + "epoch": 1.05, + "learning_rate": 3.245319312644268e-05, + "loss": 0.4868, + "step": 27416 + }, + { + "epoch": 1.05, + "learning_rate": 3.245255193639395e-05, + "loss": 0.4302, + "step": 27417 + }, + { + "epoch": 1.05, + "learning_rate": 3.245191074634522e-05, + "loss": 0.5076, + "step": 27418 + }, + { + "epoch": 1.05, + "learning_rate": 3.2451269556296485e-05, + "loss": 0.2801, + "step": 27419 + }, + { + "epoch": 1.05, + "learning_rate": 3.245062836624776e-05, + "loss": 0.258, + "step": 27420 + }, + { + "epoch": 1.05, + "learning_rate": 3.244998717619903e-05, + "loss": 0.4237, + "step": 27421 + }, + { + "epoch": 1.05, + "learning_rate": 3.24493459861503e-05, + "loss": 0.4921, + "step": 27422 + }, + { + "epoch": 1.05, + "learning_rate": 3.2448704796101565e-05, + "loss": 0.4211, + "step": 27423 + }, + { + "epoch": 1.05, + "learning_rate": 3.244806360605284e-05, + "loss": 0.3125, + "step": 27424 + }, + { + "epoch": 1.05, + "learning_rate": 3.24474224160041e-05, + "loss": 0.448, + "step": 27425 + }, + { + "epoch": 1.05, + "learning_rate": 3.2446781225955373e-05, + "loss": 0.4012, + "step": 27426 + }, + { + "epoch": 1.05, + "learning_rate": 3.2446140035906645e-05, + "loss": 0.273, + "step": 27427 + }, + { + "epoch": 1.05, + "learning_rate": 3.244549884585791e-05, + "loss": 0.4453, + "step": 27428 + }, + { + "epoch": 1.05, + "learning_rate": 3.244485765580919e-05, + "loss": 0.3005, + "step": 27429 + }, + { + "epoch": 1.05, + "learning_rate": 3.2444216465760454e-05, + "loss": 0.3781, + "step": 27430 + }, + { + "epoch": 1.05, + "learning_rate": 3.2443575275711725e-05, + "loss": 0.4915, + "step": 27431 + }, + { + "epoch": 1.05, + "learning_rate": 3.244293408566299e-05, + "loss": 0.5337, + "step": 27432 + }, + { + "epoch": 1.05, + "learning_rate": 3.244229289561426e-05, + "loss": 0.3599, + "step": 27433 + }, + { + "epoch": 1.05, + "learning_rate": 3.244165170556553e-05, + "loss": 0.359, + "step": 27434 + }, + { + "epoch": 1.05, + "learning_rate": 3.24410105155168e-05, + "loss": 0.378, + "step": 27435 + }, + { + "epoch": 1.05, + "learning_rate": 3.244036932546807e-05, + "loss": 0.3846, + "step": 27436 + }, + { + "epoch": 1.05, + "learning_rate": 3.243972813541934e-05, + "loss": 0.4425, + "step": 27437 + }, + { + "epoch": 1.05, + "learning_rate": 3.2439086945370614e-05, + "loss": 0.3707, + "step": 27438 + }, + { + "epoch": 1.05, + "learning_rate": 3.243844575532188e-05, + "loss": 0.4484, + "step": 27439 + }, + { + "epoch": 1.05, + "learning_rate": 3.243780456527315e-05, + "loss": 0.4881, + "step": 27440 + }, + { + "epoch": 1.06, + "learning_rate": 3.2437163375224415e-05, + "loss": 0.3231, + "step": 27441 + }, + { + "epoch": 1.06, + "learning_rate": 3.243652218517569e-05, + "loss": 0.3815, + "step": 27442 + }, + { + "epoch": 1.06, + "learning_rate": 3.243588099512695e-05, + "loss": 0.4542, + "step": 27443 + }, + { + "epoch": 1.06, + "learning_rate": 3.2435239805078224e-05, + "loss": 0.4309, + "step": 27444 + }, + { + "epoch": 1.06, + "learning_rate": 3.24345986150295e-05, + "loss": 0.4347, + "step": 27445 + }, + { + "epoch": 1.06, + "learning_rate": 3.243395742498077e-05, + "loss": 0.3412, + "step": 27446 + }, + { + "epoch": 1.06, + "learning_rate": 3.243331623493204e-05, + "loss": 0.3757, + "step": 27447 + }, + { + "epoch": 1.06, + "learning_rate": 3.2432675044883304e-05, + "loss": 0.3649, + "step": 27448 + }, + { + "epoch": 1.06, + "learning_rate": 3.2432033854834575e-05, + "loss": 0.314, + "step": 27449 + }, + { + "epoch": 1.06, + "learning_rate": 3.243139266478584e-05, + "loss": 0.5446, + "step": 27450 + }, + { + "epoch": 1.06, + "learning_rate": 3.243075147473711e-05, + "loss": 0.4225, + "step": 27451 + }, + { + "epoch": 1.06, + "learning_rate": 3.2430110284688384e-05, + "loss": 0.4491, + "step": 27452 + }, + { + "epoch": 1.06, + "learning_rate": 3.2429469094639656e-05, + "loss": 0.4573, + "step": 27453 + }, + { + "epoch": 1.06, + "learning_rate": 3.242882790459093e-05, + "loss": 0.3902, + "step": 27454 + }, + { + "epoch": 1.06, + "learning_rate": 3.242818671454219e-05, + "loss": 0.5075, + "step": 27455 + }, + { + "epoch": 1.06, + "learning_rate": 3.2427545524493464e-05, + "loss": 0.4178, + "step": 27456 + }, + { + "epoch": 1.06, + "learning_rate": 3.242690433444473e-05, + "loss": 0.4776, + "step": 27457 + }, + { + "epoch": 1.06, + "learning_rate": 3.2426263144396e-05, + "loss": 0.421, + "step": 27458 + }, + { + "epoch": 1.06, + "learning_rate": 3.2425621954347266e-05, + "loss": 0.4661, + "step": 27459 + }, + { + "epoch": 1.06, + "learning_rate": 3.242498076429854e-05, + "loss": 0.4479, + "step": 27460 + }, + { + "epoch": 1.06, + "learning_rate": 3.242433957424981e-05, + "loss": 0.4233, + "step": 27461 + }, + { + "epoch": 1.06, + "learning_rate": 3.242369838420108e-05, + "loss": 0.4262, + "step": 27462 + }, + { + "epoch": 1.06, + "learning_rate": 3.242305719415235e-05, + "loss": 0.296, + "step": 27463 + }, + { + "epoch": 1.06, + "learning_rate": 3.242241600410362e-05, + "loss": 0.329, + "step": 27464 + }, + { + "epoch": 1.06, + "learning_rate": 3.242177481405489e-05, + "loss": 0.4342, + "step": 27465 + }, + { + "epoch": 1.06, + "learning_rate": 3.2421133624006154e-05, + "loss": 0.4021, + "step": 27466 + }, + { + "epoch": 1.06, + "learning_rate": 3.2420492433957426e-05, + "loss": 0.4187, + "step": 27467 + }, + { + "epoch": 1.06, + "learning_rate": 3.24198512439087e-05, + "loss": 0.4002, + "step": 27468 + }, + { + "epoch": 1.06, + "learning_rate": 3.241921005385997e-05, + "loss": 0.4119, + "step": 27469 + }, + { + "epoch": 1.06, + "learning_rate": 3.2418568863811234e-05, + "loss": 0.3321, + "step": 27470 + }, + { + "epoch": 1.06, + "learning_rate": 3.2417927673762506e-05, + "loss": 0.4324, + "step": 27471 + }, + { + "epoch": 1.06, + "learning_rate": 3.241728648371378e-05, + "loss": 0.5037, + "step": 27472 + }, + { + "epoch": 1.06, + "learning_rate": 3.241664529366504e-05, + "loss": 0.4314, + "step": 27473 + }, + { + "epoch": 1.06, + "learning_rate": 3.2416004103616314e-05, + "loss": 0.3555, + "step": 27474 + }, + { + "epoch": 1.06, + "learning_rate": 3.241536291356758e-05, + "loss": 0.3974, + "step": 27475 + }, + { + "epoch": 1.06, + "learning_rate": 3.241472172351885e-05, + "loss": 0.5087, + "step": 27476 + }, + { + "epoch": 1.06, + "learning_rate": 3.241408053347012e-05, + "loss": 0.395, + "step": 27477 + }, + { + "epoch": 1.06, + "learning_rate": 3.2413439343421394e-05, + "loss": 0.3291, + "step": 27478 + }, + { + "epoch": 1.06, + "learning_rate": 3.241279815337266e-05, + "loss": 0.4857, + "step": 27479 + }, + { + "epoch": 1.06, + "learning_rate": 3.241215696332393e-05, + "loss": 0.2824, + "step": 27480 + }, + { + "epoch": 1.06, + "learning_rate": 3.24115157732752e-05, + "loss": 0.4496, + "step": 27481 + }, + { + "epoch": 1.06, + "learning_rate": 3.241087458322647e-05, + "loss": 0.39, + "step": 27482 + }, + { + "epoch": 1.06, + "learning_rate": 3.241023339317774e-05, + "loss": 0.4087, + "step": 27483 + }, + { + "epoch": 1.06, + "learning_rate": 3.240959220312901e-05, + "loss": 0.4296, + "step": 27484 + }, + { + "epoch": 1.06, + "learning_rate": 3.240895101308028e-05, + "loss": 0.4298, + "step": 27485 + }, + { + "epoch": 1.06, + "learning_rate": 3.240830982303155e-05, + "loss": 0.3433, + "step": 27486 + }, + { + "epoch": 1.06, + "learning_rate": 3.240766863298282e-05, + "loss": 0.3088, + "step": 27487 + }, + { + "epoch": 1.06, + "learning_rate": 3.2407027442934084e-05, + "loss": 0.3936, + "step": 27488 + }, + { + "epoch": 1.06, + "learning_rate": 3.2406386252885356e-05, + "loss": 0.4005, + "step": 27489 + }, + { + "epoch": 1.06, + "learning_rate": 3.240574506283663e-05, + "loss": 0.3604, + "step": 27490 + }, + { + "epoch": 1.06, + "learning_rate": 3.240510387278789e-05, + "loss": 0.4275, + "step": 27491 + }, + { + "epoch": 1.06, + "learning_rate": 3.2404462682739164e-05, + "loss": 0.559, + "step": 27492 + }, + { + "epoch": 1.06, + "learning_rate": 3.2403821492690436e-05, + "loss": 0.3771, + "step": 27493 + }, + { + "epoch": 1.06, + "learning_rate": 3.240318030264171e-05, + "loss": 0.3037, + "step": 27494 + }, + { + "epoch": 1.06, + "learning_rate": 3.240253911259297e-05, + "loss": 0.3023, + "step": 27495 + }, + { + "epoch": 1.06, + "learning_rate": 3.2401897922544245e-05, + "loss": 0.3284, + "step": 27496 + }, + { + "epoch": 1.06, + "learning_rate": 3.240125673249551e-05, + "loss": 0.4706, + "step": 27497 + }, + { + "epoch": 1.06, + "learning_rate": 3.240061554244678e-05, + "loss": 0.3268, + "step": 27498 + }, + { + "epoch": 1.06, + "learning_rate": 3.239997435239805e-05, + "loss": 0.4384, + "step": 27499 + }, + { + "epoch": 1.06, + "learning_rate": 3.2399333162349325e-05, + "loss": 0.438, + "step": 27500 + }, + { + "epoch": 1.06, + "learning_rate": 3.2398691972300596e-05, + "loss": 0.3312, + "step": 27501 + }, + { + "epoch": 1.06, + "learning_rate": 3.239805078225186e-05, + "loss": 0.4725, + "step": 27502 + }, + { + "epoch": 1.06, + "learning_rate": 3.239740959220313e-05, + "loss": 0.396, + "step": 27503 + }, + { + "epoch": 1.06, + "learning_rate": 3.23967684021544e-05, + "loss": 0.328, + "step": 27504 + }, + { + "epoch": 1.06, + "learning_rate": 3.239612721210567e-05, + "loss": 0.4122, + "step": 27505 + }, + { + "epoch": 1.06, + "learning_rate": 3.2395486022056935e-05, + "loss": 0.4611, + "step": 27506 + }, + { + "epoch": 1.06, + "learning_rate": 3.2394844832008206e-05, + "loss": 0.329, + "step": 27507 + }, + { + "epoch": 1.06, + "learning_rate": 3.239420364195948e-05, + "loss": 0.5583, + "step": 27508 + }, + { + "epoch": 1.06, + "learning_rate": 3.239356245191075e-05, + "loss": 0.3137, + "step": 27509 + }, + { + "epoch": 1.06, + "learning_rate": 3.239292126186202e-05, + "loss": 0.5247, + "step": 27510 + }, + { + "epoch": 1.06, + "learning_rate": 3.2392280071813286e-05, + "loss": 0.3855, + "step": 27511 + }, + { + "epoch": 1.06, + "learning_rate": 3.239163888176456e-05, + "loss": 0.5289, + "step": 27512 + }, + { + "epoch": 1.06, + "learning_rate": 3.239099769171582e-05, + "loss": 0.304, + "step": 27513 + }, + { + "epoch": 1.06, + "learning_rate": 3.2390356501667095e-05, + "loss": 0.3433, + "step": 27514 + }, + { + "epoch": 1.06, + "learning_rate": 3.238971531161836e-05, + "loss": 0.4045, + "step": 27515 + }, + { + "epoch": 1.06, + "learning_rate": 3.238907412156963e-05, + "loss": 0.3665, + "step": 27516 + }, + { + "epoch": 1.06, + "learning_rate": 3.238843293152091e-05, + "loss": 0.2971, + "step": 27517 + }, + { + "epoch": 1.06, + "learning_rate": 3.2387791741472175e-05, + "loss": 0.5378, + "step": 27518 + }, + { + "epoch": 1.06, + "learning_rate": 3.2387150551423447e-05, + "loss": 0.4143, + "step": 27519 + }, + { + "epoch": 1.06, + "learning_rate": 3.238650936137471e-05, + "loss": 0.4703, + "step": 27520 + }, + { + "epoch": 1.06, + "learning_rate": 3.238586817132598e-05, + "loss": 0.2941, + "step": 27521 + }, + { + "epoch": 1.06, + "learning_rate": 3.238522698127725e-05, + "loss": 0.3703, + "step": 27522 + }, + { + "epoch": 1.06, + "learning_rate": 3.238458579122852e-05, + "loss": 0.4476, + "step": 27523 + }, + { + "epoch": 1.06, + "learning_rate": 3.238394460117979e-05, + "loss": 0.3229, + "step": 27524 + }, + { + "epoch": 1.06, + "learning_rate": 3.238330341113106e-05, + "loss": 0.4408, + "step": 27525 + }, + { + "epoch": 1.06, + "learning_rate": 3.2382662221082335e-05, + "loss": 0.4179, + "step": 27526 + }, + { + "epoch": 1.06, + "learning_rate": 3.23820210310336e-05, + "loss": 0.37, + "step": 27527 + }, + { + "epoch": 1.06, + "learning_rate": 3.238137984098487e-05, + "loss": 0.2998, + "step": 27528 + }, + { + "epoch": 1.06, + "learning_rate": 3.238073865093614e-05, + "loss": 0.2902, + "step": 27529 + }, + { + "epoch": 1.06, + "learning_rate": 3.238009746088741e-05, + "loss": 0.4746, + "step": 27530 + }, + { + "epoch": 1.06, + "learning_rate": 3.237945627083867e-05, + "loss": 0.3365, + "step": 27531 + }, + { + "epoch": 1.06, + "learning_rate": 3.2378815080789945e-05, + "loss": 0.4647, + "step": 27532 + }, + { + "epoch": 1.06, + "learning_rate": 3.237817389074122e-05, + "loss": 0.3747, + "step": 27533 + }, + { + "epoch": 1.06, + "learning_rate": 3.237753270069249e-05, + "loss": 0.2601, + "step": 27534 + }, + { + "epoch": 1.06, + "learning_rate": 3.237689151064376e-05, + "loss": 0.3557, + "step": 27535 + }, + { + "epoch": 1.06, + "learning_rate": 3.2376250320595025e-05, + "loss": 0.4119, + "step": 27536 + }, + { + "epoch": 1.06, + "learning_rate": 3.23756091305463e-05, + "loss": 0.4996, + "step": 27537 + }, + { + "epoch": 1.06, + "learning_rate": 3.237496794049756e-05, + "loss": 0.3268, + "step": 27538 + }, + { + "epoch": 1.06, + "learning_rate": 3.2374326750448833e-05, + "loss": 0.3872, + "step": 27539 + }, + { + "epoch": 1.06, + "learning_rate": 3.2373685560400105e-05, + "loss": 0.5646, + "step": 27540 + }, + { + "epoch": 1.06, + "learning_rate": 3.237304437035138e-05, + "loss": 0.3854, + "step": 27541 + }, + { + "epoch": 1.06, + "learning_rate": 3.237240318030264e-05, + "loss": 0.3901, + "step": 27542 + }, + { + "epoch": 1.06, + "learning_rate": 3.2371761990253914e-05, + "loss": 0.4918, + "step": 27543 + }, + { + "epoch": 1.06, + "learning_rate": 3.2371120800205185e-05, + "loss": 0.3446, + "step": 27544 + }, + { + "epoch": 1.06, + "learning_rate": 3.237047961015645e-05, + "loss": 0.4647, + "step": 27545 + }, + { + "epoch": 1.06, + "learning_rate": 3.236983842010772e-05, + "loss": 0.6111, + "step": 27546 + }, + { + "epoch": 1.06, + "learning_rate": 3.236919723005899e-05, + "loss": 0.4597, + "step": 27547 + }, + { + "epoch": 1.06, + "learning_rate": 3.236855604001026e-05, + "loss": 0.443, + "step": 27548 + }, + { + "epoch": 1.06, + "learning_rate": 3.236791484996153e-05, + "loss": 0.3336, + "step": 27549 + }, + { + "epoch": 1.06, + "learning_rate": 3.23672736599128e-05, + "loss": 0.4571, + "step": 27550 + }, + { + "epoch": 1.06, + "learning_rate": 3.236663246986407e-05, + "loss": 0.4755, + "step": 27551 + }, + { + "epoch": 1.06, + "learning_rate": 3.236599127981534e-05, + "loss": 0.3957, + "step": 27552 + }, + { + "epoch": 1.06, + "learning_rate": 3.236535008976661e-05, + "loss": 0.4207, + "step": 27553 + }, + { + "epoch": 1.06, + "learning_rate": 3.2364708899717875e-05, + "loss": 0.3116, + "step": 27554 + }, + { + "epoch": 1.06, + "learning_rate": 3.236406770966915e-05, + "loss": 0.4094, + "step": 27555 + }, + { + "epoch": 1.06, + "learning_rate": 3.236342651962042e-05, + "loss": 0.4625, + "step": 27556 + }, + { + "epoch": 1.06, + "learning_rate": 3.236278532957169e-05, + "loss": 0.3783, + "step": 27557 + }, + { + "epoch": 1.06, + "learning_rate": 3.2362144139522955e-05, + "loss": 0.3865, + "step": 27558 + }, + { + "epoch": 1.06, + "learning_rate": 3.236150294947423e-05, + "loss": 0.4867, + "step": 27559 + }, + { + "epoch": 1.06, + "learning_rate": 3.236086175942549e-05, + "loss": 0.3858, + "step": 27560 + }, + { + "epoch": 1.06, + "learning_rate": 3.2360220569376764e-05, + "loss": 0.3246, + "step": 27561 + }, + { + "epoch": 1.06, + "learning_rate": 3.2359579379328036e-05, + "loss": 0.4094, + "step": 27562 + }, + { + "epoch": 1.06, + "learning_rate": 3.23589381892793e-05, + "loss": 0.5184, + "step": 27563 + }, + { + "epoch": 1.06, + "learning_rate": 3.235829699923057e-05, + "loss": 0.4498, + "step": 27564 + }, + { + "epoch": 1.06, + "learning_rate": 3.2357655809181844e-05, + "loss": 0.5132, + "step": 27565 + }, + { + "epoch": 1.06, + "learning_rate": 3.2357014619133116e-05, + "loss": 0.3974, + "step": 27566 + }, + { + "epoch": 1.06, + "learning_rate": 3.235637342908438e-05, + "loss": 0.384, + "step": 27567 + }, + { + "epoch": 1.06, + "learning_rate": 3.235573223903565e-05, + "loss": 0.3202, + "step": 27568 + }, + { + "epoch": 1.06, + "learning_rate": 3.235509104898692e-05, + "loss": 0.406, + "step": 27569 + }, + { + "epoch": 1.06, + "learning_rate": 3.235444985893819e-05, + "loss": 0.4858, + "step": 27570 + }, + { + "epoch": 1.06, + "learning_rate": 3.235380866888946e-05, + "loss": 0.4512, + "step": 27571 + }, + { + "epoch": 1.06, + "learning_rate": 3.235316747884073e-05, + "loss": 0.4449, + "step": 27572 + }, + { + "epoch": 1.06, + "learning_rate": 3.2352526288792004e-05, + "loss": 0.4179, + "step": 27573 + }, + { + "epoch": 1.06, + "learning_rate": 3.235188509874327e-05, + "loss": 0.4709, + "step": 27574 + }, + { + "epoch": 1.06, + "learning_rate": 3.235124390869454e-05, + "loss": 0.3418, + "step": 27575 + }, + { + "epoch": 1.06, + "learning_rate": 3.2350602718645806e-05, + "loss": 0.3972, + "step": 27576 + }, + { + "epoch": 1.06, + "learning_rate": 3.234996152859708e-05, + "loss": 0.4016, + "step": 27577 + }, + { + "epoch": 1.06, + "learning_rate": 3.234932033854834e-05, + "loss": 0.2603, + "step": 27578 + }, + { + "epoch": 1.06, + "learning_rate": 3.2348679148499614e-05, + "loss": 0.3881, + "step": 27579 + }, + { + "epoch": 1.06, + "learning_rate": 3.2348037958450886e-05, + "loss": 0.4289, + "step": 27580 + }, + { + "epoch": 1.06, + "learning_rate": 3.234739676840216e-05, + "loss": 0.4447, + "step": 27581 + }, + { + "epoch": 1.06, + "learning_rate": 3.234675557835343e-05, + "loss": 0.3878, + "step": 27582 + }, + { + "epoch": 1.06, + "learning_rate": 3.2346114388304694e-05, + "loss": 0.3804, + "step": 27583 + }, + { + "epoch": 1.06, + "learning_rate": 3.2345473198255966e-05, + "loss": 0.3162, + "step": 27584 + }, + { + "epoch": 1.06, + "learning_rate": 3.234483200820723e-05, + "loss": 0.4365, + "step": 27585 + }, + { + "epoch": 1.06, + "learning_rate": 3.23441908181585e-05, + "loss": 0.4425, + "step": 27586 + }, + { + "epoch": 1.06, + "learning_rate": 3.234354962810977e-05, + "loss": 0.3246, + "step": 27587 + }, + { + "epoch": 1.06, + "learning_rate": 3.234290843806104e-05, + "loss": 0.4702, + "step": 27588 + }, + { + "epoch": 1.06, + "learning_rate": 3.234226724801232e-05, + "loss": 0.2464, + "step": 27589 + }, + { + "epoch": 1.06, + "learning_rate": 3.234162605796358e-05, + "loss": 0.3829, + "step": 27590 + }, + { + "epoch": 1.06, + "learning_rate": 3.2340984867914854e-05, + "loss": 0.2683, + "step": 27591 + }, + { + "epoch": 1.06, + "learning_rate": 3.234034367786612e-05, + "loss": 0.6213, + "step": 27592 + }, + { + "epoch": 1.06, + "learning_rate": 3.233970248781739e-05, + "loss": 0.3552, + "step": 27593 + }, + { + "epoch": 1.06, + "learning_rate": 3.2339061297768656e-05, + "loss": 0.3563, + "step": 27594 + }, + { + "epoch": 1.06, + "learning_rate": 3.233842010771993e-05, + "loss": 0.4544, + "step": 27595 + }, + { + "epoch": 1.06, + "learning_rate": 3.23377789176712e-05, + "loss": 0.4753, + "step": 27596 + }, + { + "epoch": 1.06, + "learning_rate": 3.233713772762247e-05, + "loss": 0.421, + "step": 27597 + }, + { + "epoch": 1.06, + "learning_rate": 3.233649653757374e-05, + "loss": 0.5438, + "step": 27598 + }, + { + "epoch": 1.06, + "learning_rate": 3.233585534752501e-05, + "loss": 0.4711, + "step": 27599 + }, + { + "epoch": 1.06, + "learning_rate": 3.233521415747628e-05, + "loss": 0.2795, + "step": 27600 + }, + { + "epoch": 1.06, + "learning_rate": 3.2334572967427544e-05, + "loss": 0.3609, + "step": 27601 + }, + { + "epoch": 1.06, + "learning_rate": 3.2333931777378816e-05, + "loss": 0.4562, + "step": 27602 + }, + { + "epoch": 1.06, + "learning_rate": 3.233329058733008e-05, + "loss": 0.3048, + "step": 27603 + }, + { + "epoch": 1.06, + "learning_rate": 3.233264939728135e-05, + "loss": 0.3694, + "step": 27604 + }, + { + "epoch": 1.06, + "learning_rate": 3.233200820723263e-05, + "loss": 0.4169, + "step": 27605 + }, + { + "epoch": 1.06, + "learning_rate": 3.2331367017183896e-05, + "loss": 0.4542, + "step": 27606 + }, + { + "epoch": 1.06, + "learning_rate": 3.233072582713517e-05, + "loss": 0.411, + "step": 27607 + }, + { + "epoch": 1.06, + "learning_rate": 3.233008463708643e-05, + "loss": 0.4991, + "step": 27608 + }, + { + "epoch": 1.06, + "learning_rate": 3.2329443447037705e-05, + "loss": 0.4748, + "step": 27609 + }, + { + "epoch": 1.06, + "learning_rate": 3.232880225698897e-05, + "loss": 0.4438, + "step": 27610 + }, + { + "epoch": 1.06, + "learning_rate": 3.232816106694024e-05, + "loss": 0.3514, + "step": 27611 + }, + { + "epoch": 1.06, + "learning_rate": 3.232751987689151e-05, + "loss": 0.4241, + "step": 27612 + }, + { + "epoch": 1.06, + "learning_rate": 3.2326878686842785e-05, + "loss": 0.36, + "step": 27613 + }, + { + "epoch": 1.06, + "learning_rate": 3.2326237496794056e-05, + "loss": 0.451, + "step": 27614 + }, + { + "epoch": 1.06, + "learning_rate": 3.232559630674532e-05, + "loss": 0.4622, + "step": 27615 + }, + { + "epoch": 1.06, + "learning_rate": 3.232495511669659e-05, + "loss": 0.3546, + "step": 27616 + }, + { + "epoch": 1.06, + "learning_rate": 3.232431392664786e-05, + "loss": 0.3513, + "step": 27617 + }, + { + "epoch": 1.06, + "learning_rate": 3.232367273659913e-05, + "loss": 0.3888, + "step": 27618 + }, + { + "epoch": 1.06, + "learning_rate": 3.2323031546550395e-05, + "loss": 0.3326, + "step": 27619 + }, + { + "epoch": 1.06, + "learning_rate": 3.2322390356501666e-05, + "loss": 0.3279, + "step": 27620 + }, + { + "epoch": 1.06, + "learning_rate": 3.232174916645294e-05, + "loss": 0.5426, + "step": 27621 + }, + { + "epoch": 1.06, + "learning_rate": 3.232110797640421e-05, + "loss": 0.322, + "step": 27622 + }, + { + "epoch": 1.06, + "learning_rate": 3.232046678635548e-05, + "loss": 0.4171, + "step": 27623 + }, + { + "epoch": 1.06, + "learning_rate": 3.2319825596306746e-05, + "loss": 0.3712, + "step": 27624 + }, + { + "epoch": 1.06, + "learning_rate": 3.231918440625802e-05, + "loss": 0.3555, + "step": 27625 + }, + { + "epoch": 1.06, + "learning_rate": 3.231854321620928e-05, + "loss": 0.4521, + "step": 27626 + }, + { + "epoch": 1.06, + "learning_rate": 3.2317902026160555e-05, + "loss": 0.3941, + "step": 27627 + }, + { + "epoch": 1.06, + "learning_rate": 3.2317260836111827e-05, + "loss": 0.4326, + "step": 27628 + }, + { + "epoch": 1.06, + "learning_rate": 3.23166196460631e-05, + "loss": 0.4018, + "step": 27629 + }, + { + "epoch": 1.06, + "learning_rate": 3.231597845601436e-05, + "loss": 0.321, + "step": 27630 + }, + { + "epoch": 1.06, + "learning_rate": 3.2315337265965635e-05, + "loss": 0.439, + "step": 27631 + }, + { + "epoch": 1.06, + "learning_rate": 3.231469607591691e-05, + "loss": 0.393, + "step": 27632 + }, + { + "epoch": 1.06, + "learning_rate": 3.231405488586817e-05, + "loss": 0.3871, + "step": 27633 + }, + { + "epoch": 1.06, + "learning_rate": 3.231341369581944e-05, + "loss": 0.3061, + "step": 27634 + }, + { + "epoch": 1.06, + "learning_rate": 3.231277250577071e-05, + "loss": 0.3905, + "step": 27635 + }, + { + "epoch": 1.06, + "learning_rate": 3.231213131572198e-05, + "loss": 0.4334, + "step": 27636 + }, + { + "epoch": 1.06, + "learning_rate": 3.231149012567325e-05, + "loss": 0.4467, + "step": 27637 + }, + { + "epoch": 1.06, + "learning_rate": 3.231084893562452e-05, + "loss": 0.2371, + "step": 27638 + }, + { + "epoch": 1.06, + "learning_rate": 3.231020774557579e-05, + "loss": 0.4043, + "step": 27639 + }, + { + "epoch": 1.06, + "learning_rate": 3.230956655552706e-05, + "loss": 0.3856, + "step": 27640 + }, + { + "epoch": 1.06, + "learning_rate": 3.230892536547833e-05, + "loss": 0.3403, + "step": 27641 + }, + { + "epoch": 1.06, + "learning_rate": 3.23082841754296e-05, + "loss": 0.2794, + "step": 27642 + }, + { + "epoch": 1.06, + "learning_rate": 3.230764298538087e-05, + "loss": 0.442, + "step": 27643 + }, + { + "epoch": 1.06, + "learning_rate": 3.230700179533214e-05, + "loss": 0.3275, + "step": 27644 + }, + { + "epoch": 1.06, + "learning_rate": 3.230636060528341e-05, + "loss": 0.2873, + "step": 27645 + }, + { + "epoch": 1.06, + "learning_rate": 3.230571941523468e-05, + "loss": 0.3569, + "step": 27646 + }, + { + "epoch": 1.06, + "learning_rate": 3.230507822518595e-05, + "loss": 0.3251, + "step": 27647 + }, + { + "epoch": 1.06, + "learning_rate": 3.2304437035137213e-05, + "loss": 0.4426, + "step": 27648 + }, + { + "epoch": 1.06, + "learning_rate": 3.2303795845088485e-05, + "loss": 0.4144, + "step": 27649 + }, + { + "epoch": 1.06, + "learning_rate": 3.230315465503976e-05, + "loss": 0.2935, + "step": 27650 + }, + { + "epoch": 1.06, + "learning_rate": 3.230251346499102e-05, + "loss": 0.4, + "step": 27651 + }, + { + "epoch": 1.06, + "learning_rate": 3.2301872274942294e-05, + "loss": 0.2746, + "step": 27652 + }, + { + "epoch": 1.06, + "learning_rate": 3.2301231084893565e-05, + "loss": 0.3965, + "step": 27653 + }, + { + "epoch": 1.06, + "learning_rate": 3.230058989484484e-05, + "loss": 0.4516, + "step": 27654 + }, + { + "epoch": 1.06, + "learning_rate": 3.22999487047961e-05, + "loss": 0.3128, + "step": 27655 + }, + { + "epoch": 1.06, + "learning_rate": 3.2299307514747374e-05, + "loss": 0.4837, + "step": 27656 + }, + { + "epoch": 1.06, + "learning_rate": 3.229866632469864e-05, + "loss": 0.3911, + "step": 27657 + }, + { + "epoch": 1.06, + "learning_rate": 3.229802513464991e-05, + "loss": 0.3953, + "step": 27658 + }, + { + "epoch": 1.06, + "learning_rate": 3.229738394460118e-05, + "loss": 0.4452, + "step": 27659 + }, + { + "epoch": 1.06, + "learning_rate": 3.2296742754552454e-05, + "loss": 0.3809, + "step": 27660 + }, + { + "epoch": 1.06, + "learning_rate": 3.2296101564503725e-05, + "loss": 0.4312, + "step": 27661 + }, + { + "epoch": 1.06, + "learning_rate": 3.229546037445499e-05, + "loss": 0.3848, + "step": 27662 + }, + { + "epoch": 1.06, + "learning_rate": 3.229481918440626e-05, + "loss": 0.3576, + "step": 27663 + }, + { + "epoch": 1.06, + "learning_rate": 3.229417799435753e-05, + "loss": 0.4266, + "step": 27664 + }, + { + "epoch": 1.06, + "learning_rate": 3.22935368043088e-05, + "loss": 0.3474, + "step": 27665 + }, + { + "epoch": 1.06, + "learning_rate": 3.2292895614260064e-05, + "loss": 0.3614, + "step": 27666 + }, + { + "epoch": 1.06, + "learning_rate": 3.2292254424211335e-05, + "loss": 0.3906, + "step": 27667 + }, + { + "epoch": 1.06, + "learning_rate": 3.229161323416261e-05, + "loss": 0.4215, + "step": 27668 + }, + { + "epoch": 1.06, + "learning_rate": 3.229097204411388e-05, + "loss": 0.3831, + "step": 27669 + }, + { + "epoch": 1.06, + "learning_rate": 3.229033085406515e-05, + "loss": 0.4345, + "step": 27670 + }, + { + "epoch": 1.06, + "learning_rate": 3.2289689664016415e-05, + "loss": 0.3288, + "step": 27671 + }, + { + "epoch": 1.06, + "learning_rate": 3.228904847396769e-05, + "loss": 0.3149, + "step": 27672 + }, + { + "epoch": 1.06, + "learning_rate": 3.228840728391895e-05, + "loss": 0.3075, + "step": 27673 + }, + { + "epoch": 1.06, + "learning_rate": 3.2287766093870224e-05, + "loss": 0.4118, + "step": 27674 + }, + { + "epoch": 1.06, + "learning_rate": 3.228712490382149e-05, + "loss": 0.3847, + "step": 27675 + }, + { + "epoch": 1.06, + "learning_rate": 3.228648371377276e-05, + "loss": 0.4032, + "step": 27676 + }, + { + "epoch": 1.06, + "learning_rate": 3.228584252372404e-05, + "loss": 0.4128, + "step": 27677 + }, + { + "epoch": 1.06, + "learning_rate": 3.2285201333675304e-05, + "loss": 0.3713, + "step": 27678 + }, + { + "epoch": 1.06, + "learning_rate": 3.2284560143626576e-05, + "loss": 0.5126, + "step": 27679 + }, + { + "epoch": 1.06, + "learning_rate": 3.228391895357784e-05, + "loss": 0.5349, + "step": 27680 + }, + { + "epoch": 1.06, + "learning_rate": 3.228327776352911e-05, + "loss": 0.3667, + "step": 27681 + }, + { + "epoch": 1.06, + "learning_rate": 3.228263657348038e-05, + "loss": 0.3354, + "step": 27682 + }, + { + "epoch": 1.06, + "learning_rate": 3.228199538343165e-05, + "loss": 0.601, + "step": 27683 + }, + { + "epoch": 1.06, + "learning_rate": 3.228135419338292e-05, + "loss": 0.4165, + "step": 27684 + }, + { + "epoch": 1.06, + "learning_rate": 3.228071300333419e-05, + "loss": 0.4001, + "step": 27685 + }, + { + "epoch": 1.06, + "learning_rate": 3.2280071813285464e-05, + "loss": 0.5024, + "step": 27686 + }, + { + "epoch": 1.06, + "learning_rate": 3.227943062323673e-05, + "loss": 0.3697, + "step": 27687 + }, + { + "epoch": 1.06, + "learning_rate": 3.2278789433188e-05, + "loss": 0.3098, + "step": 27688 + }, + { + "epoch": 1.06, + "learning_rate": 3.2278148243139266e-05, + "loss": 0.3316, + "step": 27689 + }, + { + "epoch": 1.06, + "learning_rate": 3.227750705309054e-05, + "loss": 0.4245, + "step": 27690 + }, + { + "epoch": 1.06, + "learning_rate": 3.22768658630418e-05, + "loss": 0.3256, + "step": 27691 + }, + { + "epoch": 1.06, + "learning_rate": 3.2276224672993074e-05, + "loss": 0.4266, + "step": 27692 + }, + { + "epoch": 1.06, + "learning_rate": 3.2275583482944346e-05, + "loss": 0.3877, + "step": 27693 + }, + { + "epoch": 1.06, + "learning_rate": 3.227494229289562e-05, + "loss": 0.2817, + "step": 27694 + }, + { + "epoch": 1.06, + "learning_rate": 3.227430110284689e-05, + "loss": 0.5179, + "step": 27695 + }, + { + "epoch": 1.06, + "learning_rate": 3.2273659912798154e-05, + "loss": 0.4403, + "step": 27696 + }, + { + "epoch": 1.06, + "learning_rate": 3.2273018722749426e-05, + "loss": 0.3613, + "step": 27697 + }, + { + "epoch": 1.06, + "learning_rate": 3.227237753270069e-05, + "loss": 0.4824, + "step": 27698 + }, + { + "epoch": 1.06, + "learning_rate": 3.227173634265196e-05, + "loss": 0.4217, + "step": 27699 + }, + { + "epoch": 1.06, + "learning_rate": 3.2271095152603234e-05, + "loss": 0.419, + "step": 27700 + }, + { + "epoch": 1.07, + "learning_rate": 3.2270453962554506e-05, + "loss": 0.3086, + "step": 27701 + }, + { + "epoch": 1.07, + "learning_rate": 3.226981277250577e-05, + "loss": 0.366, + "step": 27702 + }, + { + "epoch": 1.07, + "learning_rate": 3.226917158245704e-05, + "loss": 0.3635, + "step": 27703 + }, + { + "epoch": 1.07, + "learning_rate": 3.2268530392408314e-05, + "loss": 0.4975, + "step": 27704 + }, + { + "epoch": 1.07, + "learning_rate": 3.226788920235958e-05, + "loss": 0.3355, + "step": 27705 + }, + { + "epoch": 1.07, + "learning_rate": 3.226724801231085e-05, + "loss": 0.4119, + "step": 27706 + }, + { + "epoch": 1.07, + "learning_rate": 3.2266606822262116e-05, + "loss": 0.3812, + "step": 27707 + }, + { + "epoch": 1.07, + "learning_rate": 3.226596563221339e-05, + "loss": 0.3199, + "step": 27708 + }, + { + "epoch": 1.07, + "learning_rate": 3.226532444216466e-05, + "loss": 0.412, + "step": 27709 + }, + { + "epoch": 1.07, + "learning_rate": 3.226468325211593e-05, + "loss": 0.3484, + "step": 27710 + }, + { + "epoch": 1.07, + "learning_rate": 3.2264042062067196e-05, + "loss": 0.4481, + "step": 27711 + }, + { + "epoch": 1.07, + "learning_rate": 3.226340087201847e-05, + "loss": 0.3818, + "step": 27712 + }, + { + "epoch": 1.07, + "learning_rate": 3.226275968196974e-05, + "loss": 0.6352, + "step": 27713 + }, + { + "epoch": 1.07, + "learning_rate": 3.2262118491921004e-05, + "loss": 0.3853, + "step": 27714 + }, + { + "epoch": 1.07, + "learning_rate": 3.2261477301872276e-05, + "loss": 0.3034, + "step": 27715 + }, + { + "epoch": 1.07, + "learning_rate": 3.226083611182355e-05, + "loss": 0.2651, + "step": 27716 + }, + { + "epoch": 1.07, + "learning_rate": 3.226019492177482e-05, + "loss": 0.3063, + "step": 27717 + }, + { + "epoch": 1.07, + "learning_rate": 3.2259553731726085e-05, + "loss": 0.3675, + "step": 27718 + }, + { + "epoch": 1.07, + "learning_rate": 3.2258912541677356e-05, + "loss": 0.382, + "step": 27719 + }, + { + "epoch": 1.07, + "learning_rate": 3.225827135162862e-05, + "loss": 0.5012, + "step": 27720 + }, + { + "epoch": 1.07, + "learning_rate": 3.225763016157989e-05, + "loss": 0.3394, + "step": 27721 + }, + { + "epoch": 1.07, + "learning_rate": 3.2256988971531165e-05, + "loss": 0.4856, + "step": 27722 + }, + { + "epoch": 1.07, + "learning_rate": 3.225634778148243e-05, + "loss": 0.4471, + "step": 27723 + }, + { + "epoch": 1.07, + "learning_rate": 3.22557065914337e-05, + "loss": 0.496, + "step": 27724 + }, + { + "epoch": 1.07, + "learning_rate": 3.225506540138497e-05, + "loss": 0.2883, + "step": 27725 + }, + { + "epoch": 1.07, + "learning_rate": 3.2254424211336245e-05, + "loss": 0.4135, + "step": 27726 + }, + { + "epoch": 1.07, + "learning_rate": 3.225378302128751e-05, + "loss": 0.4622, + "step": 27727 + }, + { + "epoch": 1.07, + "learning_rate": 3.225314183123878e-05, + "loss": 0.5049, + "step": 27728 + }, + { + "epoch": 1.07, + "learning_rate": 3.2252500641190046e-05, + "loss": 0.4799, + "step": 27729 + }, + { + "epoch": 1.07, + "learning_rate": 3.225185945114132e-05, + "loss": 0.4558, + "step": 27730 + }, + { + "epoch": 1.07, + "learning_rate": 3.225121826109259e-05, + "loss": 0.3347, + "step": 27731 + }, + { + "epoch": 1.07, + "learning_rate": 3.225057707104386e-05, + "loss": 0.3532, + "step": 27732 + }, + { + "epoch": 1.07, + "learning_rate": 3.224993588099513e-05, + "loss": 0.3695, + "step": 27733 + }, + { + "epoch": 1.07, + "learning_rate": 3.22492946909464e-05, + "loss": 0.407, + "step": 27734 + }, + { + "epoch": 1.07, + "learning_rate": 3.224865350089767e-05, + "loss": 0.5968, + "step": 27735 + }, + { + "epoch": 1.07, + "learning_rate": 3.2248012310848935e-05, + "loss": 0.5182, + "step": 27736 + }, + { + "epoch": 1.07, + "learning_rate": 3.2247371120800206e-05, + "loss": 0.346, + "step": 27737 + }, + { + "epoch": 1.07, + "learning_rate": 3.224672993075147e-05, + "loss": 0.4343, + "step": 27738 + }, + { + "epoch": 1.07, + "learning_rate": 3.224608874070274e-05, + "loss": 0.4323, + "step": 27739 + }, + { + "epoch": 1.07, + "learning_rate": 3.2245447550654015e-05, + "loss": 0.5561, + "step": 27740 + }, + { + "epoch": 1.07, + "learning_rate": 3.2244806360605287e-05, + "loss": 0.4127, + "step": 27741 + }, + { + "epoch": 1.07, + "learning_rate": 3.224416517055656e-05, + "loss": 0.3967, + "step": 27742 + }, + { + "epoch": 1.07, + "learning_rate": 3.224352398050782e-05, + "loss": 0.341, + "step": 27743 + }, + { + "epoch": 1.07, + "learning_rate": 3.2242882790459095e-05, + "loss": 0.3857, + "step": 27744 + }, + { + "epoch": 1.07, + "learning_rate": 3.224224160041036e-05, + "loss": 0.3059, + "step": 27745 + }, + { + "epoch": 1.07, + "learning_rate": 3.224160041036163e-05, + "loss": 0.3002, + "step": 27746 + }, + { + "epoch": 1.07, + "learning_rate": 3.2240959220312897e-05, + "loss": 0.4703, + "step": 27747 + }, + { + "epoch": 1.07, + "learning_rate": 3.2240318030264175e-05, + "loss": 0.3757, + "step": 27748 + }, + { + "epoch": 1.07, + "learning_rate": 3.223967684021545e-05, + "loss": 0.3937, + "step": 27749 + }, + { + "epoch": 1.07, + "learning_rate": 3.223903565016671e-05, + "loss": 0.398, + "step": 27750 + }, + { + "epoch": 1.07, + "learning_rate": 3.2238394460117983e-05, + "loss": 0.3328, + "step": 27751 + }, + { + "epoch": 1.07, + "learning_rate": 3.223775327006925e-05, + "loss": 0.3997, + "step": 27752 + }, + { + "epoch": 1.07, + "learning_rate": 3.223711208002052e-05, + "loss": 0.3463, + "step": 27753 + }, + { + "epoch": 1.07, + "learning_rate": 3.2236470889971785e-05, + "loss": 0.3209, + "step": 27754 + }, + { + "epoch": 1.07, + "learning_rate": 3.223582969992306e-05, + "loss": 0.3463, + "step": 27755 + }, + { + "epoch": 1.07, + "learning_rate": 3.223518850987433e-05, + "loss": 0.4581, + "step": 27756 + }, + { + "epoch": 1.07, + "learning_rate": 3.22345473198256e-05, + "loss": 0.4316, + "step": 27757 + }, + { + "epoch": 1.07, + "learning_rate": 3.223390612977687e-05, + "loss": 0.5454, + "step": 27758 + }, + { + "epoch": 1.07, + "learning_rate": 3.223326493972814e-05, + "loss": 0.4544, + "step": 27759 + }, + { + "epoch": 1.07, + "learning_rate": 3.223262374967941e-05, + "loss": 0.4329, + "step": 27760 + }, + { + "epoch": 1.07, + "learning_rate": 3.2231982559630673e-05, + "loss": 0.4038, + "step": 27761 + }, + { + "epoch": 1.07, + "learning_rate": 3.2231341369581945e-05, + "loss": 0.3636, + "step": 27762 + }, + { + "epoch": 1.07, + "learning_rate": 3.223070017953321e-05, + "loss": 0.417, + "step": 27763 + }, + { + "epoch": 1.07, + "learning_rate": 3.223005898948448e-05, + "loss": 0.3761, + "step": 27764 + }, + { + "epoch": 1.07, + "learning_rate": 3.2229417799435754e-05, + "loss": 0.5312, + "step": 27765 + }, + { + "epoch": 1.07, + "learning_rate": 3.2228776609387025e-05, + "loss": 0.3293, + "step": 27766 + }, + { + "epoch": 1.07, + "learning_rate": 3.22281354193383e-05, + "loss": 0.4616, + "step": 27767 + }, + { + "epoch": 1.07, + "learning_rate": 3.222749422928956e-05, + "loss": 0.4167, + "step": 27768 + }, + { + "epoch": 1.07, + "learning_rate": 3.2226853039240834e-05, + "loss": 0.4073, + "step": 27769 + }, + { + "epoch": 1.07, + "learning_rate": 3.22262118491921e-05, + "loss": 0.4837, + "step": 27770 + }, + { + "epoch": 1.07, + "learning_rate": 3.222557065914337e-05, + "loss": 0.4787, + "step": 27771 + }, + { + "epoch": 1.07, + "learning_rate": 3.222492946909464e-05, + "loss": 0.4061, + "step": 27772 + }, + { + "epoch": 1.07, + "learning_rate": 3.2224288279045914e-05, + "loss": 0.3756, + "step": 27773 + }, + { + "epoch": 1.07, + "learning_rate": 3.222364708899718e-05, + "loss": 0.428, + "step": 27774 + }, + { + "epoch": 1.07, + "learning_rate": 3.222300589894845e-05, + "loss": 0.4561, + "step": 27775 + }, + { + "epoch": 1.07, + "learning_rate": 3.222236470889972e-05, + "loss": 0.422, + "step": 27776 + }, + { + "epoch": 1.07, + "learning_rate": 3.222172351885099e-05, + "loss": 0.4912, + "step": 27777 + }, + { + "epoch": 1.07, + "learning_rate": 3.222108232880226e-05, + "loss": 0.4004, + "step": 27778 + }, + { + "epoch": 1.07, + "learning_rate": 3.2220441138753524e-05, + "loss": 0.372, + "step": 27779 + }, + { + "epoch": 1.07, + "learning_rate": 3.2219799948704795e-05, + "loss": 0.4679, + "step": 27780 + }, + { + "epoch": 1.07, + "learning_rate": 3.221915875865607e-05, + "loss": 0.4804, + "step": 27781 + }, + { + "epoch": 1.07, + "learning_rate": 3.221851756860734e-05, + "loss": 0.3725, + "step": 27782 + }, + { + "epoch": 1.07, + "learning_rate": 3.221787637855861e-05, + "loss": 0.5073, + "step": 27783 + }, + { + "epoch": 1.07, + "learning_rate": 3.2217235188509876e-05, + "loss": 0.3713, + "step": 27784 + }, + { + "epoch": 1.07, + "learning_rate": 3.221659399846115e-05, + "loss": 0.3519, + "step": 27785 + }, + { + "epoch": 1.07, + "learning_rate": 3.221595280841241e-05, + "loss": 0.4482, + "step": 27786 + }, + { + "epoch": 1.07, + "learning_rate": 3.2215311618363684e-05, + "loss": 0.3223, + "step": 27787 + }, + { + "epoch": 1.07, + "learning_rate": 3.2214670428314956e-05, + "loss": 0.4053, + "step": 27788 + }, + { + "epoch": 1.07, + "learning_rate": 3.221402923826623e-05, + "loss": 0.3598, + "step": 27789 + }, + { + "epoch": 1.07, + "learning_rate": 3.221338804821749e-05, + "loss": 0.4429, + "step": 27790 + }, + { + "epoch": 1.07, + "learning_rate": 3.2212746858168764e-05, + "loss": 0.4241, + "step": 27791 + }, + { + "epoch": 1.07, + "learning_rate": 3.2212105668120036e-05, + "loss": 0.4604, + "step": 27792 + }, + { + "epoch": 1.07, + "learning_rate": 3.22114644780713e-05, + "loss": 0.2566, + "step": 27793 + }, + { + "epoch": 1.07, + "learning_rate": 3.221082328802257e-05, + "loss": 0.4889, + "step": 27794 + }, + { + "epoch": 1.07, + "learning_rate": 3.221018209797384e-05, + "loss": 0.4152, + "step": 27795 + }, + { + "epoch": 1.07, + "learning_rate": 3.220954090792511e-05, + "loss": 0.4037, + "step": 27796 + }, + { + "epoch": 1.07, + "learning_rate": 3.220889971787638e-05, + "loss": 0.4152, + "step": 27797 + }, + { + "epoch": 1.07, + "learning_rate": 3.220825852782765e-05, + "loss": 0.4421, + "step": 27798 + }, + { + "epoch": 1.07, + "learning_rate": 3.220761733777892e-05, + "loss": 0.3626, + "step": 27799 + }, + { + "epoch": 1.07, + "learning_rate": 3.220697614773019e-05, + "loss": 0.531, + "step": 27800 + }, + { + "epoch": 1.07, + "learning_rate": 3.220633495768146e-05, + "loss": 0.3778, + "step": 27801 + }, + { + "epoch": 1.07, + "learning_rate": 3.2205693767632726e-05, + "loss": 0.4976, + "step": 27802 + }, + { + "epoch": 1.07, + "learning_rate": 3.2205052577584e-05, + "loss": 0.4582, + "step": 27803 + }, + { + "epoch": 1.07, + "learning_rate": 3.220441138753527e-05, + "loss": 0.4175, + "step": 27804 + }, + { + "epoch": 1.07, + "learning_rate": 3.220377019748654e-05, + "loss": 0.3195, + "step": 27805 + }, + { + "epoch": 1.07, + "learning_rate": 3.2203129007437806e-05, + "loss": 0.5191, + "step": 27806 + }, + { + "epoch": 1.07, + "learning_rate": 3.220248781738908e-05, + "loss": 0.3676, + "step": 27807 + }, + { + "epoch": 1.07, + "learning_rate": 3.220184662734034e-05, + "loss": 0.3417, + "step": 27808 + }, + { + "epoch": 1.07, + "learning_rate": 3.2201205437291614e-05, + "loss": 0.4225, + "step": 27809 + }, + { + "epoch": 1.07, + "learning_rate": 3.2200564247242886e-05, + "loss": 0.3394, + "step": 27810 + }, + { + "epoch": 1.07, + "learning_rate": 3.219992305719415e-05, + "loss": 0.571, + "step": 27811 + }, + { + "epoch": 1.07, + "learning_rate": 3.219928186714542e-05, + "loss": 0.5489, + "step": 27812 + }, + { + "epoch": 1.07, + "learning_rate": 3.2198640677096694e-05, + "loss": 0.4499, + "step": 27813 + }, + { + "epoch": 1.07, + "learning_rate": 3.2197999487047966e-05, + "loss": 0.3817, + "step": 27814 + }, + { + "epoch": 1.07, + "learning_rate": 3.219735829699923e-05, + "loss": 0.4806, + "step": 27815 + }, + { + "epoch": 1.07, + "learning_rate": 3.21967171069505e-05, + "loss": 0.2848, + "step": 27816 + }, + { + "epoch": 1.07, + "learning_rate": 3.219607591690177e-05, + "loss": 0.3683, + "step": 27817 + }, + { + "epoch": 1.07, + "learning_rate": 3.219543472685304e-05, + "loss": 0.3687, + "step": 27818 + }, + { + "epoch": 1.07, + "learning_rate": 3.219479353680431e-05, + "loss": 0.4274, + "step": 27819 + }, + { + "epoch": 1.07, + "learning_rate": 3.219415234675558e-05, + "loss": 0.4, + "step": 27820 + }, + { + "epoch": 1.07, + "learning_rate": 3.2193511156706854e-05, + "loss": 0.3655, + "step": 27821 + }, + { + "epoch": 1.07, + "learning_rate": 3.219286996665812e-05, + "loss": 0.4012, + "step": 27822 + }, + { + "epoch": 1.07, + "learning_rate": 3.219222877660939e-05, + "loss": 0.3204, + "step": 27823 + }, + { + "epoch": 1.07, + "learning_rate": 3.2191587586560656e-05, + "loss": 0.4, + "step": 27824 + }, + { + "epoch": 1.07, + "learning_rate": 3.219094639651193e-05, + "loss": 0.3541, + "step": 27825 + }, + { + "epoch": 1.07, + "learning_rate": 3.219030520646319e-05, + "loss": 0.4233, + "step": 27826 + }, + { + "epoch": 1.07, + "learning_rate": 3.2189664016414464e-05, + "loss": 0.5042, + "step": 27827 + }, + { + "epoch": 1.07, + "learning_rate": 3.2189022826365736e-05, + "loss": 0.4016, + "step": 27828 + }, + { + "epoch": 1.07, + "learning_rate": 3.218838163631701e-05, + "loss": 0.3843, + "step": 27829 + }, + { + "epoch": 1.07, + "learning_rate": 3.218774044626828e-05, + "loss": 0.2921, + "step": 27830 + }, + { + "epoch": 1.07, + "learning_rate": 3.2187099256219545e-05, + "loss": 0.2993, + "step": 27831 + }, + { + "epoch": 1.07, + "learning_rate": 3.2186458066170816e-05, + "loss": 0.499, + "step": 27832 + }, + { + "epoch": 1.07, + "learning_rate": 3.218581687612208e-05, + "loss": 0.3577, + "step": 27833 + }, + { + "epoch": 1.07, + "learning_rate": 3.218517568607335e-05, + "loss": 0.4271, + "step": 27834 + }, + { + "epoch": 1.07, + "learning_rate": 3.218453449602462e-05, + "loss": 0.5271, + "step": 27835 + }, + { + "epoch": 1.07, + "learning_rate": 3.218389330597589e-05, + "loss": 0.4592, + "step": 27836 + }, + { + "epoch": 1.07, + "learning_rate": 3.218325211592717e-05, + "loss": 0.3412, + "step": 27837 + }, + { + "epoch": 1.07, + "learning_rate": 3.218261092587843e-05, + "loss": 0.4653, + "step": 27838 + }, + { + "epoch": 1.07, + "learning_rate": 3.2181969735829705e-05, + "loss": 0.3164, + "step": 27839 + }, + { + "epoch": 1.07, + "learning_rate": 3.218132854578097e-05, + "loss": 0.3905, + "step": 27840 + }, + { + "epoch": 1.07, + "learning_rate": 3.218068735573224e-05, + "loss": 0.385, + "step": 27841 + }, + { + "epoch": 1.07, + "learning_rate": 3.2180046165683506e-05, + "loss": 0.4213, + "step": 27842 + }, + { + "epoch": 1.07, + "learning_rate": 3.217940497563478e-05, + "loss": 0.31, + "step": 27843 + }, + { + "epoch": 1.07, + "learning_rate": 3.217876378558605e-05, + "loss": 0.3768, + "step": 27844 + }, + { + "epoch": 1.07, + "learning_rate": 3.217812259553732e-05, + "loss": 0.3578, + "step": 27845 + }, + { + "epoch": 1.07, + "learning_rate": 3.217748140548859e-05, + "loss": 0.312, + "step": 27846 + }, + { + "epoch": 1.07, + "learning_rate": 3.217684021543986e-05, + "loss": 0.49, + "step": 27847 + }, + { + "epoch": 1.07, + "learning_rate": 3.217619902539113e-05, + "loss": 0.394, + "step": 27848 + }, + { + "epoch": 1.07, + "learning_rate": 3.2175557835342395e-05, + "loss": 0.3753, + "step": 27849 + }, + { + "epoch": 1.07, + "learning_rate": 3.2174916645293667e-05, + "loss": 0.5353, + "step": 27850 + }, + { + "epoch": 1.07, + "learning_rate": 3.217427545524493e-05, + "loss": 0.394, + "step": 27851 + }, + { + "epoch": 1.07, + "learning_rate": 3.21736342651962e-05, + "loss": 0.4608, + "step": 27852 + }, + { + "epoch": 1.07, + "learning_rate": 3.2172993075147475e-05, + "loss": 0.2964, + "step": 27853 + }, + { + "epoch": 1.07, + "learning_rate": 3.2172351885098747e-05, + "loss": 0.4325, + "step": 27854 + }, + { + "epoch": 1.07, + "learning_rate": 3.217171069505002e-05, + "loss": 0.3665, + "step": 27855 + }, + { + "epoch": 1.07, + "learning_rate": 3.217106950500128e-05, + "loss": 0.3971, + "step": 27856 + }, + { + "epoch": 1.07, + "learning_rate": 3.2170428314952555e-05, + "loss": 0.3451, + "step": 27857 + }, + { + "epoch": 1.07, + "learning_rate": 3.216978712490382e-05, + "loss": 0.4548, + "step": 27858 + }, + { + "epoch": 1.07, + "learning_rate": 3.216914593485509e-05, + "loss": 0.3596, + "step": 27859 + }, + { + "epoch": 1.07, + "learning_rate": 3.216850474480636e-05, + "loss": 0.4408, + "step": 27860 + }, + { + "epoch": 1.07, + "learning_rate": 3.2167863554757635e-05, + "loss": 0.3034, + "step": 27861 + }, + { + "epoch": 1.07, + "learning_rate": 3.21672223647089e-05, + "loss": 0.3826, + "step": 27862 + }, + { + "epoch": 1.07, + "learning_rate": 3.216658117466017e-05, + "loss": 0.5281, + "step": 27863 + }, + { + "epoch": 1.07, + "learning_rate": 3.2165939984611443e-05, + "loss": 0.351, + "step": 27864 + }, + { + "epoch": 1.07, + "learning_rate": 3.216529879456271e-05, + "loss": 0.361, + "step": 27865 + }, + { + "epoch": 1.07, + "learning_rate": 3.216465760451398e-05, + "loss": 0.4331, + "step": 27866 + }, + { + "epoch": 1.07, + "learning_rate": 3.2164016414465245e-05, + "loss": 0.4848, + "step": 27867 + }, + { + "epoch": 1.07, + "learning_rate": 3.216337522441652e-05, + "loss": 0.5047, + "step": 27868 + }, + { + "epoch": 1.07, + "learning_rate": 3.216273403436779e-05, + "loss": 0.4804, + "step": 27869 + }, + { + "epoch": 1.07, + "learning_rate": 3.216209284431906e-05, + "loss": 0.2966, + "step": 27870 + }, + { + "epoch": 1.07, + "learning_rate": 3.2161451654270325e-05, + "loss": 0.5059, + "step": 27871 + }, + { + "epoch": 1.07, + "learning_rate": 3.21608104642216e-05, + "loss": 0.4345, + "step": 27872 + }, + { + "epoch": 1.07, + "learning_rate": 3.216016927417287e-05, + "loss": 0.4158, + "step": 27873 + }, + { + "epoch": 1.07, + "learning_rate": 3.2159528084124133e-05, + "loss": 0.4537, + "step": 27874 + }, + { + "epoch": 1.07, + "learning_rate": 3.2158886894075405e-05, + "loss": 0.337, + "step": 27875 + }, + { + "epoch": 1.07, + "learning_rate": 3.215824570402668e-05, + "loss": 0.4966, + "step": 27876 + }, + { + "epoch": 1.07, + "learning_rate": 3.215760451397795e-05, + "loss": 0.4028, + "step": 27877 + }, + { + "epoch": 1.07, + "learning_rate": 3.2156963323929214e-05, + "loss": 0.5057, + "step": 27878 + }, + { + "epoch": 1.07, + "learning_rate": 3.2156322133880485e-05, + "loss": 0.4193, + "step": 27879 + }, + { + "epoch": 1.07, + "learning_rate": 3.215568094383175e-05, + "loss": 0.4561, + "step": 27880 + }, + { + "epoch": 1.07, + "learning_rate": 3.215503975378302e-05, + "loss": 0.4424, + "step": 27881 + }, + { + "epoch": 1.07, + "learning_rate": 3.2154398563734294e-05, + "loss": 0.4002, + "step": 27882 + }, + { + "epoch": 1.07, + "learning_rate": 3.215375737368556e-05, + "loss": 0.3186, + "step": 27883 + }, + { + "epoch": 1.07, + "learning_rate": 3.215311618363683e-05, + "loss": 0.4661, + "step": 27884 + }, + { + "epoch": 1.07, + "learning_rate": 3.21524749935881e-05, + "loss": 0.3634, + "step": 27885 + }, + { + "epoch": 1.07, + "learning_rate": 3.2151833803539374e-05, + "loss": 0.3799, + "step": 27886 + }, + { + "epoch": 1.07, + "learning_rate": 3.215119261349064e-05, + "loss": 0.3117, + "step": 27887 + }, + { + "epoch": 1.07, + "learning_rate": 3.215055142344191e-05, + "loss": 0.5198, + "step": 27888 + }, + { + "epoch": 1.07, + "learning_rate": 3.2149910233393175e-05, + "loss": 0.3808, + "step": 27889 + }, + { + "epoch": 1.07, + "learning_rate": 3.214926904334445e-05, + "loss": 0.3005, + "step": 27890 + }, + { + "epoch": 1.07, + "learning_rate": 3.214862785329572e-05, + "loss": 0.4651, + "step": 27891 + }, + { + "epoch": 1.07, + "learning_rate": 3.214798666324699e-05, + "loss": 0.4771, + "step": 27892 + }, + { + "epoch": 1.07, + "learning_rate": 3.214734547319826e-05, + "loss": 0.4641, + "step": 27893 + }, + { + "epoch": 1.07, + "learning_rate": 3.214670428314953e-05, + "loss": 0.3689, + "step": 27894 + }, + { + "epoch": 1.07, + "learning_rate": 3.21460630931008e-05, + "loss": 0.4812, + "step": 27895 + }, + { + "epoch": 1.07, + "learning_rate": 3.2145421903052064e-05, + "loss": 0.3252, + "step": 27896 + }, + { + "epoch": 1.07, + "learning_rate": 3.2144780713003336e-05, + "loss": 0.2621, + "step": 27897 + }, + { + "epoch": 1.07, + "learning_rate": 3.21441395229546e-05, + "loss": 0.3899, + "step": 27898 + }, + { + "epoch": 1.07, + "learning_rate": 3.214349833290587e-05, + "loss": 0.4039, + "step": 27899 + }, + { + "epoch": 1.07, + "learning_rate": 3.2142857142857144e-05, + "loss": 0.4161, + "step": 27900 + }, + { + "epoch": 1.07, + "learning_rate": 3.2142215952808416e-05, + "loss": 0.3797, + "step": 27901 + }, + { + "epoch": 1.07, + "learning_rate": 3.214157476275969e-05, + "loss": 0.4577, + "step": 27902 + }, + { + "epoch": 1.07, + "learning_rate": 3.214093357271095e-05, + "loss": 0.3561, + "step": 27903 + }, + { + "epoch": 1.07, + "learning_rate": 3.2140292382662224e-05, + "loss": 0.3934, + "step": 27904 + }, + { + "epoch": 1.07, + "learning_rate": 3.213965119261349e-05, + "loss": 0.2728, + "step": 27905 + }, + { + "epoch": 1.07, + "learning_rate": 3.213901000256476e-05, + "loss": 0.3889, + "step": 27906 + }, + { + "epoch": 1.07, + "learning_rate": 3.2138368812516026e-05, + "loss": 0.5051, + "step": 27907 + }, + { + "epoch": 1.07, + "learning_rate": 3.2137727622467304e-05, + "loss": 0.3876, + "step": 27908 + }, + { + "epoch": 1.07, + "learning_rate": 3.2137086432418576e-05, + "loss": 0.2744, + "step": 27909 + }, + { + "epoch": 1.07, + "learning_rate": 3.213644524236984e-05, + "loss": 0.4871, + "step": 27910 + }, + { + "epoch": 1.07, + "learning_rate": 3.213580405232111e-05, + "loss": 0.487, + "step": 27911 + }, + { + "epoch": 1.07, + "learning_rate": 3.213516286227238e-05, + "loss": 0.3655, + "step": 27912 + }, + { + "epoch": 1.07, + "learning_rate": 3.213452167222365e-05, + "loss": 0.4812, + "step": 27913 + }, + { + "epoch": 1.07, + "learning_rate": 3.2133880482174914e-05, + "loss": 0.4265, + "step": 27914 + }, + { + "epoch": 1.07, + "learning_rate": 3.2133239292126186e-05, + "loss": 0.4671, + "step": 27915 + }, + { + "epoch": 1.07, + "learning_rate": 3.213259810207746e-05, + "loss": 0.4327, + "step": 27916 + }, + { + "epoch": 1.07, + "learning_rate": 3.213195691202873e-05, + "loss": 0.3805, + "step": 27917 + }, + { + "epoch": 1.07, + "learning_rate": 3.213131572198e-05, + "loss": 0.4415, + "step": 27918 + }, + { + "epoch": 1.07, + "learning_rate": 3.2130674531931266e-05, + "loss": 0.3606, + "step": 27919 + }, + { + "epoch": 1.07, + "learning_rate": 3.213003334188254e-05, + "loss": 0.3958, + "step": 27920 + }, + { + "epoch": 1.07, + "learning_rate": 3.21293921518338e-05, + "loss": 0.4319, + "step": 27921 + }, + { + "epoch": 1.07, + "learning_rate": 3.2128750961785074e-05, + "loss": 0.3666, + "step": 27922 + }, + { + "epoch": 1.07, + "learning_rate": 3.212810977173634e-05, + "loss": 0.3222, + "step": 27923 + }, + { + "epoch": 1.07, + "learning_rate": 3.212746858168761e-05, + "loss": 0.5461, + "step": 27924 + }, + { + "epoch": 1.07, + "learning_rate": 3.212682739163888e-05, + "loss": 0.4379, + "step": 27925 + }, + { + "epoch": 1.07, + "learning_rate": 3.2126186201590154e-05, + "loss": 0.3204, + "step": 27926 + }, + { + "epoch": 1.07, + "learning_rate": 3.2125545011541426e-05, + "loss": 0.366, + "step": 27927 + }, + { + "epoch": 1.07, + "learning_rate": 3.212490382149269e-05, + "loss": 0.5133, + "step": 27928 + }, + { + "epoch": 1.07, + "learning_rate": 3.212426263144396e-05, + "loss": 0.5188, + "step": 27929 + }, + { + "epoch": 1.07, + "learning_rate": 3.212362144139523e-05, + "loss": 0.3159, + "step": 27930 + }, + { + "epoch": 1.07, + "learning_rate": 3.21229802513465e-05, + "loss": 0.3905, + "step": 27931 + }, + { + "epoch": 1.07, + "learning_rate": 3.212233906129777e-05, + "loss": 0.4134, + "step": 27932 + }, + { + "epoch": 1.07, + "learning_rate": 3.212169787124904e-05, + "loss": 0.4624, + "step": 27933 + }, + { + "epoch": 1.07, + "learning_rate": 3.212105668120031e-05, + "loss": 0.5844, + "step": 27934 + }, + { + "epoch": 1.07, + "learning_rate": 3.212041549115158e-05, + "loss": 0.3539, + "step": 27935 + }, + { + "epoch": 1.07, + "learning_rate": 3.211977430110285e-05, + "loss": 0.3418, + "step": 27936 + }, + { + "epoch": 1.07, + "learning_rate": 3.2119133111054116e-05, + "loss": 0.4503, + "step": 27937 + }, + { + "epoch": 1.07, + "learning_rate": 3.211849192100539e-05, + "loss": 0.4785, + "step": 27938 + }, + { + "epoch": 1.07, + "learning_rate": 3.211785073095665e-05, + "loss": 0.4545, + "step": 27939 + }, + { + "epoch": 1.07, + "learning_rate": 3.2117209540907924e-05, + "loss": 0.3394, + "step": 27940 + }, + { + "epoch": 1.07, + "learning_rate": 3.2116568350859196e-05, + "loss": 0.4248, + "step": 27941 + }, + { + "epoch": 1.07, + "learning_rate": 3.211592716081047e-05, + "loss": 0.3072, + "step": 27942 + }, + { + "epoch": 1.07, + "learning_rate": 3.211528597076173e-05, + "loss": 0.3707, + "step": 27943 + }, + { + "epoch": 1.07, + "learning_rate": 3.2114644780713005e-05, + "loss": 0.3938, + "step": 27944 + }, + { + "epoch": 1.07, + "learning_rate": 3.2114003590664276e-05, + "loss": 0.3918, + "step": 27945 + }, + { + "epoch": 1.07, + "learning_rate": 3.211336240061554e-05, + "loss": 0.4882, + "step": 27946 + }, + { + "epoch": 1.07, + "learning_rate": 3.211272121056681e-05, + "loss": 0.4247, + "step": 27947 + }, + { + "epoch": 1.07, + "learning_rate": 3.2112080020518085e-05, + "loss": 0.3478, + "step": 27948 + }, + { + "epoch": 1.07, + "learning_rate": 3.2111438830469356e-05, + "loss": 0.5214, + "step": 27949 + }, + { + "epoch": 1.07, + "learning_rate": 3.211079764042062e-05, + "loss": 0.3864, + "step": 27950 + }, + { + "epoch": 1.07, + "learning_rate": 3.211015645037189e-05, + "loss": 0.3795, + "step": 27951 + }, + { + "epoch": 1.07, + "learning_rate": 3.210951526032316e-05, + "loss": 0.3443, + "step": 27952 + }, + { + "epoch": 1.07, + "learning_rate": 3.210887407027443e-05, + "loss": 0.3764, + "step": 27953 + }, + { + "epoch": 1.07, + "learning_rate": 3.21082328802257e-05, + "loss": 0.4142, + "step": 27954 + }, + { + "epoch": 1.07, + "learning_rate": 3.2107591690176966e-05, + "loss": 0.4184, + "step": 27955 + }, + { + "epoch": 1.07, + "learning_rate": 3.210695050012824e-05, + "loss": 0.3021, + "step": 27956 + }, + { + "epoch": 1.07, + "learning_rate": 3.210630931007951e-05, + "loss": 0.4306, + "step": 27957 + }, + { + "epoch": 1.07, + "learning_rate": 3.210566812003078e-05, + "loss": 0.3884, + "step": 27958 + }, + { + "epoch": 1.07, + "learning_rate": 3.2105026929982046e-05, + "loss": 0.3922, + "step": 27959 + }, + { + "epoch": 1.07, + "learning_rate": 3.210438573993332e-05, + "loss": 0.4268, + "step": 27960 + }, + { + "epoch": 1.08, + "learning_rate": 3.210374454988458e-05, + "loss": 0.3862, + "step": 27961 + }, + { + "epoch": 1.08, + "learning_rate": 3.2103103359835855e-05, + "loss": 0.3767, + "step": 27962 + }, + { + "epoch": 1.08, + "learning_rate": 3.2102462169787127e-05, + "loss": 0.4211, + "step": 27963 + }, + { + "epoch": 1.08, + "learning_rate": 3.21018209797384e-05, + "loss": 0.4055, + "step": 27964 + }, + { + "epoch": 1.08, + "learning_rate": 3.210117978968967e-05, + "loss": 0.3248, + "step": 27965 + }, + { + "epoch": 1.08, + "learning_rate": 3.2100538599640935e-05, + "loss": 0.3427, + "step": 27966 + }, + { + "epoch": 1.08, + "learning_rate": 3.209989740959221e-05, + "loss": 0.4495, + "step": 27967 + }, + { + "epoch": 1.08, + "learning_rate": 3.209925621954347e-05, + "loss": 0.477, + "step": 27968 + }, + { + "epoch": 1.08, + "learning_rate": 3.209861502949474e-05, + "loss": 0.3956, + "step": 27969 + }, + { + "epoch": 1.08, + "learning_rate": 3.2097973839446015e-05, + "loss": 0.3422, + "step": 27970 + }, + { + "epoch": 1.08, + "learning_rate": 3.209733264939728e-05, + "loss": 0.4715, + "step": 27971 + }, + { + "epoch": 1.08, + "learning_rate": 3.209669145934855e-05, + "loss": 0.2762, + "step": 27972 + }, + { + "epoch": 1.08, + "learning_rate": 3.209605026929982e-05, + "loss": 0.5217, + "step": 27973 + }, + { + "epoch": 1.08, + "learning_rate": 3.2095409079251095e-05, + "loss": 0.4676, + "step": 27974 + }, + { + "epoch": 1.08, + "learning_rate": 3.209476788920236e-05, + "loss": 0.3658, + "step": 27975 + }, + { + "epoch": 1.08, + "learning_rate": 3.209412669915363e-05, + "loss": 0.3781, + "step": 27976 + }, + { + "epoch": 1.08, + "learning_rate": 3.20934855091049e-05, + "loss": 0.3784, + "step": 27977 + }, + { + "epoch": 1.08, + "learning_rate": 3.209284431905617e-05, + "loss": 0.4556, + "step": 27978 + }, + { + "epoch": 1.08, + "learning_rate": 3.209220312900744e-05, + "loss": 0.3911, + "step": 27979 + }, + { + "epoch": 1.08, + "learning_rate": 3.209156193895871e-05, + "loss": 0.4602, + "step": 27980 + }, + { + "epoch": 1.08, + "learning_rate": 3.2090920748909984e-05, + "loss": 0.3626, + "step": 27981 + }, + { + "epoch": 1.08, + "learning_rate": 3.209027955886125e-05, + "loss": 0.4051, + "step": 27982 + }, + { + "epoch": 1.08, + "learning_rate": 3.208963836881252e-05, + "loss": 0.3792, + "step": 27983 + }, + { + "epoch": 1.08, + "learning_rate": 3.2088997178763785e-05, + "loss": 0.3904, + "step": 27984 + }, + { + "epoch": 1.08, + "learning_rate": 3.208835598871506e-05, + "loss": 0.3417, + "step": 27985 + }, + { + "epoch": 1.08, + "learning_rate": 3.208771479866632e-05, + "loss": 0.4277, + "step": 27986 + }, + { + "epoch": 1.08, + "learning_rate": 3.2087073608617594e-05, + "loss": 0.3599, + "step": 27987 + }, + { + "epoch": 1.08, + "learning_rate": 3.2086432418568865e-05, + "loss": 0.3847, + "step": 27988 + }, + { + "epoch": 1.08, + "learning_rate": 3.208579122852014e-05, + "loss": 0.4056, + "step": 27989 + }, + { + "epoch": 1.08, + "learning_rate": 3.208515003847141e-05, + "loss": 0.4864, + "step": 27990 + }, + { + "epoch": 1.08, + "learning_rate": 3.2084508848422674e-05, + "loss": 0.2793, + "step": 27991 + }, + { + "epoch": 1.08, + "learning_rate": 3.2083867658373945e-05, + "loss": 0.4523, + "step": 27992 + }, + { + "epoch": 1.08, + "learning_rate": 3.208322646832521e-05, + "loss": 0.3447, + "step": 27993 + }, + { + "epoch": 1.08, + "learning_rate": 3.208258527827648e-05, + "loss": 0.4184, + "step": 27994 + }, + { + "epoch": 1.08, + "learning_rate": 3.208194408822775e-05, + "loss": 0.4474, + "step": 27995 + }, + { + "epoch": 1.08, + "learning_rate": 3.208130289817902e-05, + "loss": 0.3623, + "step": 27996 + }, + { + "epoch": 1.08, + "learning_rate": 3.20806617081303e-05, + "loss": 0.4389, + "step": 27997 + }, + { + "epoch": 1.08, + "learning_rate": 3.208002051808156e-05, + "loss": 0.3713, + "step": 27998 + }, + { + "epoch": 1.08, + "learning_rate": 3.2079379328032834e-05, + "loss": 0.3394, + "step": 27999 + }, + { + "epoch": 1.08, + "learning_rate": 3.20787381379841e-05, + "loss": 0.4112, + "step": 28000 + }, + { + "epoch": 1.08, + "learning_rate": 3.207809694793537e-05, + "loss": 0.384, + "step": 28001 + }, + { + "epoch": 1.08, + "learning_rate": 3.2077455757886635e-05, + "loss": 0.5164, + "step": 28002 + }, + { + "epoch": 1.08, + "learning_rate": 3.207681456783791e-05, + "loss": 0.5079, + "step": 28003 + }, + { + "epoch": 1.08, + "learning_rate": 3.207617337778918e-05, + "loss": 0.2852, + "step": 28004 + }, + { + "epoch": 1.08, + "learning_rate": 3.207553218774045e-05, + "loss": 0.355, + "step": 28005 + }, + { + "epoch": 1.08, + "learning_rate": 3.207489099769172e-05, + "loss": 0.314, + "step": 28006 + }, + { + "epoch": 1.08, + "learning_rate": 3.207424980764299e-05, + "loss": 0.473, + "step": 28007 + }, + { + "epoch": 1.08, + "learning_rate": 3.207360861759426e-05, + "loss": 0.3946, + "step": 28008 + }, + { + "epoch": 1.08, + "learning_rate": 3.2072967427545524e-05, + "loss": 0.4832, + "step": 28009 + }, + { + "epoch": 1.08, + "learning_rate": 3.2072326237496796e-05, + "loss": 0.4924, + "step": 28010 + }, + { + "epoch": 1.08, + "learning_rate": 3.207168504744806e-05, + "loss": 0.4199, + "step": 28011 + }, + { + "epoch": 1.08, + "learning_rate": 3.207104385739933e-05, + "loss": 0.3761, + "step": 28012 + }, + { + "epoch": 1.08, + "learning_rate": 3.2070402667350604e-05, + "loss": 0.3359, + "step": 28013 + }, + { + "epoch": 1.08, + "learning_rate": 3.2069761477301876e-05, + "loss": 0.2455, + "step": 28014 + }, + { + "epoch": 1.08, + "learning_rate": 3.206912028725315e-05, + "loss": 0.4879, + "step": 28015 + }, + { + "epoch": 1.08, + "learning_rate": 3.206847909720441e-05, + "loss": 0.568, + "step": 28016 + }, + { + "epoch": 1.08, + "learning_rate": 3.2067837907155684e-05, + "loss": 0.4613, + "step": 28017 + }, + { + "epoch": 1.08, + "learning_rate": 3.206719671710695e-05, + "loss": 0.4505, + "step": 28018 + }, + { + "epoch": 1.08, + "learning_rate": 3.206655552705822e-05, + "loss": 0.3036, + "step": 28019 + }, + { + "epoch": 1.08, + "learning_rate": 3.206591433700949e-05, + "loss": 0.3454, + "step": 28020 + }, + { + "epoch": 1.08, + "learning_rate": 3.2065273146960764e-05, + "loss": 0.4393, + "step": 28021 + }, + { + "epoch": 1.08, + "learning_rate": 3.206463195691203e-05, + "loss": 0.3912, + "step": 28022 + }, + { + "epoch": 1.08, + "learning_rate": 3.20639907668633e-05, + "loss": 0.3969, + "step": 28023 + }, + { + "epoch": 1.08, + "learning_rate": 3.206334957681457e-05, + "loss": 0.2491, + "step": 28024 + }, + { + "epoch": 1.08, + "learning_rate": 3.206270838676584e-05, + "loss": 0.311, + "step": 28025 + }, + { + "epoch": 1.08, + "learning_rate": 3.206206719671711e-05, + "loss": 0.5076, + "step": 28026 + }, + { + "epoch": 1.08, + "learning_rate": 3.2061426006668374e-05, + "loss": 0.4339, + "step": 28027 + }, + { + "epoch": 1.08, + "learning_rate": 3.2060784816619646e-05, + "loss": 0.4098, + "step": 28028 + }, + { + "epoch": 1.08, + "learning_rate": 3.206014362657092e-05, + "loss": 0.4258, + "step": 28029 + }, + { + "epoch": 1.08, + "learning_rate": 3.205950243652219e-05, + "loss": 0.3924, + "step": 28030 + }, + { + "epoch": 1.08, + "learning_rate": 3.2058861246473454e-05, + "loss": 0.2887, + "step": 28031 + }, + { + "epoch": 1.08, + "learning_rate": 3.2058220056424726e-05, + "loss": 0.4581, + "step": 28032 + }, + { + "epoch": 1.08, + "learning_rate": 3.2057578866376e-05, + "loss": 0.4605, + "step": 28033 + }, + { + "epoch": 1.08, + "learning_rate": 3.205693767632726e-05, + "loss": 0.3653, + "step": 28034 + }, + { + "epoch": 1.08, + "learning_rate": 3.2056296486278534e-05, + "loss": 0.3003, + "step": 28035 + }, + { + "epoch": 1.08, + "learning_rate": 3.2055655296229806e-05, + "loss": 0.3568, + "step": 28036 + }, + { + "epoch": 1.08, + "learning_rate": 3.205501410618108e-05, + "loss": 0.4718, + "step": 28037 + }, + { + "epoch": 1.08, + "learning_rate": 3.205437291613234e-05, + "loss": 0.3374, + "step": 28038 + }, + { + "epoch": 1.08, + "learning_rate": 3.2053731726083614e-05, + "loss": 0.4071, + "step": 28039 + }, + { + "epoch": 1.08, + "learning_rate": 3.205309053603488e-05, + "loss": 0.4756, + "step": 28040 + }, + { + "epoch": 1.08, + "learning_rate": 3.205244934598615e-05, + "loss": 0.3924, + "step": 28041 + }, + { + "epoch": 1.08, + "learning_rate": 3.205180815593742e-05, + "loss": 0.4272, + "step": 28042 + }, + { + "epoch": 1.08, + "learning_rate": 3.205116696588869e-05, + "loss": 0.2654, + "step": 28043 + }, + { + "epoch": 1.08, + "learning_rate": 3.205052577583996e-05, + "loss": 0.3395, + "step": 28044 + }, + { + "epoch": 1.08, + "learning_rate": 3.204988458579123e-05, + "loss": 0.4065, + "step": 28045 + }, + { + "epoch": 1.08, + "learning_rate": 3.20492433957425e-05, + "loss": 0.3843, + "step": 28046 + }, + { + "epoch": 1.08, + "learning_rate": 3.204860220569377e-05, + "loss": 0.4551, + "step": 28047 + }, + { + "epoch": 1.08, + "learning_rate": 3.204796101564504e-05, + "loss": 0.6642, + "step": 28048 + }, + { + "epoch": 1.08, + "learning_rate": 3.2047319825596304e-05, + "loss": 0.535, + "step": 28049 + }, + { + "epoch": 1.08, + "learning_rate": 3.2046678635547576e-05, + "loss": 0.3648, + "step": 28050 + }, + { + "epoch": 1.08, + "learning_rate": 3.204603744549885e-05, + "loss": 0.3599, + "step": 28051 + }, + { + "epoch": 1.08, + "learning_rate": 3.204539625545012e-05, + "loss": 0.4053, + "step": 28052 + }, + { + "epoch": 1.08, + "learning_rate": 3.204475506540139e-05, + "loss": 0.2915, + "step": 28053 + }, + { + "epoch": 1.08, + "learning_rate": 3.2044113875352656e-05, + "loss": 0.4434, + "step": 28054 + }, + { + "epoch": 1.08, + "learning_rate": 3.204347268530393e-05, + "loss": 0.4136, + "step": 28055 + }, + { + "epoch": 1.08, + "learning_rate": 3.204283149525519e-05, + "loss": 0.2852, + "step": 28056 + }, + { + "epoch": 1.08, + "learning_rate": 3.2042190305206465e-05, + "loss": 0.5046, + "step": 28057 + }, + { + "epoch": 1.08, + "learning_rate": 3.204154911515773e-05, + "loss": 0.3331, + "step": 28058 + }, + { + "epoch": 1.08, + "learning_rate": 3.2040907925109e-05, + "loss": 0.3775, + "step": 28059 + }, + { + "epoch": 1.08, + "learning_rate": 3.204026673506027e-05, + "loss": 0.2915, + "step": 28060 + }, + { + "epoch": 1.08, + "learning_rate": 3.2039625545011545e-05, + "loss": 0.4258, + "step": 28061 + }, + { + "epoch": 1.08, + "learning_rate": 3.2038984354962816e-05, + "loss": 0.3332, + "step": 28062 + }, + { + "epoch": 1.08, + "learning_rate": 3.203834316491408e-05, + "loss": 0.343, + "step": 28063 + }, + { + "epoch": 1.08, + "learning_rate": 3.203770197486535e-05, + "loss": 0.5866, + "step": 28064 + }, + { + "epoch": 1.08, + "learning_rate": 3.203706078481662e-05, + "loss": 0.3619, + "step": 28065 + }, + { + "epoch": 1.08, + "learning_rate": 3.203641959476789e-05, + "loss": 0.349, + "step": 28066 + }, + { + "epoch": 1.08, + "learning_rate": 3.2035778404719155e-05, + "loss": 0.446, + "step": 28067 + }, + { + "epoch": 1.08, + "learning_rate": 3.203513721467043e-05, + "loss": 0.3225, + "step": 28068 + }, + { + "epoch": 1.08, + "learning_rate": 3.2034496024621705e-05, + "loss": 0.4436, + "step": 28069 + }, + { + "epoch": 1.08, + "learning_rate": 3.203385483457297e-05, + "loss": 0.3942, + "step": 28070 + }, + { + "epoch": 1.08, + "learning_rate": 3.203321364452424e-05, + "loss": 0.3899, + "step": 28071 + }, + { + "epoch": 1.08, + "learning_rate": 3.2032572454475506e-05, + "loss": 0.424, + "step": 28072 + }, + { + "epoch": 1.08, + "learning_rate": 3.203193126442678e-05, + "loss": 0.2915, + "step": 28073 + }, + { + "epoch": 1.08, + "learning_rate": 3.203129007437804e-05, + "loss": 0.317, + "step": 28074 + }, + { + "epoch": 1.08, + "learning_rate": 3.2030648884329315e-05, + "loss": 0.517, + "step": 28075 + }, + { + "epoch": 1.08, + "learning_rate": 3.2030007694280587e-05, + "loss": 0.4275, + "step": 28076 + }, + { + "epoch": 1.08, + "learning_rate": 3.202936650423186e-05, + "loss": 0.3964, + "step": 28077 + }, + { + "epoch": 1.08, + "learning_rate": 3.202872531418313e-05, + "loss": 0.3394, + "step": 28078 + }, + { + "epoch": 1.08, + "learning_rate": 3.2028084124134395e-05, + "loss": 0.3762, + "step": 28079 + }, + { + "epoch": 1.08, + "learning_rate": 3.202744293408567e-05, + "loss": 0.4291, + "step": 28080 + }, + { + "epoch": 1.08, + "learning_rate": 3.202680174403693e-05, + "loss": 0.419, + "step": 28081 + }, + { + "epoch": 1.08, + "learning_rate": 3.20261605539882e-05, + "loss": 0.3881, + "step": 28082 + }, + { + "epoch": 1.08, + "learning_rate": 3.202551936393947e-05, + "loss": 0.3449, + "step": 28083 + }, + { + "epoch": 1.08, + "learning_rate": 3.202487817389074e-05, + "loss": 0.3222, + "step": 28084 + }, + { + "epoch": 1.08, + "learning_rate": 3.202423698384201e-05, + "loss": 0.4026, + "step": 28085 + }, + { + "epoch": 1.08, + "learning_rate": 3.2023595793793283e-05, + "loss": 0.3514, + "step": 28086 + }, + { + "epoch": 1.08, + "learning_rate": 3.2022954603744555e-05, + "loss": 0.3666, + "step": 28087 + }, + { + "epoch": 1.08, + "learning_rate": 3.202231341369582e-05, + "loss": 0.4892, + "step": 28088 + }, + { + "epoch": 1.08, + "learning_rate": 3.202167222364709e-05, + "loss": 0.3281, + "step": 28089 + }, + { + "epoch": 1.08, + "learning_rate": 3.202103103359836e-05, + "loss": 0.3775, + "step": 28090 + }, + { + "epoch": 1.08, + "learning_rate": 3.202038984354963e-05, + "loss": 0.2697, + "step": 28091 + }, + { + "epoch": 1.08, + "learning_rate": 3.20197486535009e-05, + "loss": 0.3739, + "step": 28092 + }, + { + "epoch": 1.08, + "learning_rate": 3.201910746345217e-05, + "loss": 0.2382, + "step": 28093 + }, + { + "epoch": 1.08, + "learning_rate": 3.201846627340344e-05, + "loss": 0.4102, + "step": 28094 + }, + { + "epoch": 1.08, + "learning_rate": 3.201782508335471e-05, + "loss": 0.315, + "step": 28095 + }, + { + "epoch": 1.08, + "learning_rate": 3.201718389330598e-05, + "loss": 0.3257, + "step": 28096 + }, + { + "epoch": 1.08, + "learning_rate": 3.2016542703257245e-05, + "loss": 0.3393, + "step": 28097 + }, + { + "epoch": 1.08, + "learning_rate": 3.201590151320852e-05, + "loss": 0.3562, + "step": 28098 + }, + { + "epoch": 1.08, + "learning_rate": 3.201526032315978e-05, + "loss": 0.6029, + "step": 28099 + }, + { + "epoch": 1.08, + "learning_rate": 3.2014619133111054e-05, + "loss": 0.5611, + "step": 28100 + }, + { + "epoch": 1.08, + "learning_rate": 3.2013977943062325e-05, + "loss": 0.3877, + "step": 28101 + }, + { + "epoch": 1.08, + "learning_rate": 3.20133367530136e-05, + "loss": 0.4634, + "step": 28102 + }, + { + "epoch": 1.08, + "learning_rate": 3.201269556296486e-05, + "loss": 0.3159, + "step": 28103 + }, + { + "epoch": 1.08, + "learning_rate": 3.2012054372916134e-05, + "loss": 0.3664, + "step": 28104 + }, + { + "epoch": 1.08, + "learning_rate": 3.2011413182867405e-05, + "loss": 0.4293, + "step": 28105 + }, + { + "epoch": 1.08, + "learning_rate": 3.201077199281867e-05, + "loss": 0.4338, + "step": 28106 + }, + { + "epoch": 1.08, + "learning_rate": 3.201013080276994e-05, + "loss": 0.4399, + "step": 28107 + }, + { + "epoch": 1.08, + "learning_rate": 3.2009489612721214e-05, + "loss": 0.4943, + "step": 28108 + }, + { + "epoch": 1.08, + "learning_rate": 3.2008848422672485e-05, + "loss": 0.2956, + "step": 28109 + }, + { + "epoch": 1.08, + "learning_rate": 3.200820723262375e-05, + "loss": 0.3965, + "step": 28110 + }, + { + "epoch": 1.08, + "learning_rate": 3.200756604257502e-05, + "loss": 0.2936, + "step": 28111 + }, + { + "epoch": 1.08, + "learning_rate": 3.200692485252629e-05, + "loss": 0.3725, + "step": 28112 + }, + { + "epoch": 1.08, + "learning_rate": 3.200628366247756e-05, + "loss": 0.4423, + "step": 28113 + }, + { + "epoch": 1.08, + "learning_rate": 3.200564247242883e-05, + "loss": 0.2591, + "step": 28114 + }, + { + "epoch": 1.08, + "learning_rate": 3.2005001282380095e-05, + "loss": 0.561, + "step": 28115 + }, + { + "epoch": 1.08, + "learning_rate": 3.200436009233137e-05, + "loss": 0.3563, + "step": 28116 + }, + { + "epoch": 1.08, + "learning_rate": 3.200371890228264e-05, + "loss": 0.4728, + "step": 28117 + }, + { + "epoch": 1.08, + "learning_rate": 3.200307771223391e-05, + "loss": 0.2828, + "step": 28118 + }, + { + "epoch": 1.08, + "learning_rate": 3.2002436522185176e-05, + "loss": 0.4144, + "step": 28119 + }, + { + "epoch": 1.08, + "learning_rate": 3.200179533213645e-05, + "loss": 0.3779, + "step": 28120 + }, + { + "epoch": 1.08, + "learning_rate": 3.200115414208771e-05, + "loss": 0.4048, + "step": 28121 + }, + { + "epoch": 1.08, + "learning_rate": 3.2000512952038984e-05, + "loss": 0.2966, + "step": 28122 + }, + { + "epoch": 1.08, + "learning_rate": 3.1999871761990256e-05, + "loss": 0.4786, + "step": 28123 + }, + { + "epoch": 1.08, + "learning_rate": 3.199923057194153e-05, + "loss": 0.3539, + "step": 28124 + }, + { + "epoch": 1.08, + "learning_rate": 3.19985893818928e-05, + "loss": 0.4733, + "step": 28125 + }, + { + "epoch": 1.08, + "learning_rate": 3.1997948191844064e-05, + "loss": 0.4401, + "step": 28126 + }, + { + "epoch": 1.08, + "learning_rate": 3.1997307001795336e-05, + "loss": 0.4496, + "step": 28127 + }, + { + "epoch": 1.08, + "learning_rate": 3.19966658117466e-05, + "loss": 0.377, + "step": 28128 + }, + { + "epoch": 1.08, + "learning_rate": 3.199602462169787e-05, + "loss": 0.3211, + "step": 28129 + }, + { + "epoch": 1.08, + "learning_rate": 3.199538343164914e-05, + "loss": 0.457, + "step": 28130 + }, + { + "epoch": 1.08, + "learning_rate": 3.199474224160041e-05, + "loss": 0.3499, + "step": 28131 + }, + { + "epoch": 1.08, + "learning_rate": 3.199410105155168e-05, + "loss": 0.4084, + "step": 28132 + }, + { + "epoch": 1.08, + "learning_rate": 3.199345986150295e-05, + "loss": 0.4144, + "step": 28133 + }, + { + "epoch": 1.08, + "learning_rate": 3.1992818671454224e-05, + "loss": 0.3136, + "step": 28134 + }, + { + "epoch": 1.08, + "learning_rate": 3.199217748140549e-05, + "loss": 0.3431, + "step": 28135 + }, + { + "epoch": 1.08, + "learning_rate": 3.199153629135676e-05, + "loss": 0.4075, + "step": 28136 + }, + { + "epoch": 1.08, + "learning_rate": 3.1990895101308026e-05, + "loss": 0.3931, + "step": 28137 + }, + { + "epoch": 1.08, + "learning_rate": 3.19902539112593e-05, + "loss": 0.3723, + "step": 28138 + }, + { + "epoch": 1.08, + "learning_rate": 3.198961272121056e-05, + "loss": 0.4207, + "step": 28139 + }, + { + "epoch": 1.08, + "learning_rate": 3.198897153116184e-05, + "loss": 0.3732, + "step": 28140 + }, + { + "epoch": 1.08, + "learning_rate": 3.198833034111311e-05, + "loss": 0.4041, + "step": 28141 + }, + { + "epoch": 1.08, + "learning_rate": 3.198768915106438e-05, + "loss": 0.3469, + "step": 28142 + }, + { + "epoch": 1.08, + "learning_rate": 3.198704796101565e-05, + "loss": 0.4894, + "step": 28143 + }, + { + "epoch": 1.08, + "learning_rate": 3.1986406770966914e-05, + "loss": 0.3811, + "step": 28144 + }, + { + "epoch": 1.08, + "learning_rate": 3.1985765580918186e-05, + "loss": 0.4319, + "step": 28145 + }, + { + "epoch": 1.08, + "learning_rate": 3.198512439086945e-05, + "loss": 0.375, + "step": 28146 + }, + { + "epoch": 1.08, + "learning_rate": 3.198448320082072e-05, + "loss": 0.3532, + "step": 28147 + }, + { + "epoch": 1.08, + "learning_rate": 3.1983842010771994e-05, + "loss": 0.2544, + "step": 28148 + }, + { + "epoch": 1.08, + "learning_rate": 3.1983200820723266e-05, + "loss": 0.4212, + "step": 28149 + }, + { + "epoch": 1.08, + "learning_rate": 3.198255963067454e-05, + "loss": 0.4167, + "step": 28150 + }, + { + "epoch": 1.08, + "learning_rate": 3.19819184406258e-05, + "loss": 0.4428, + "step": 28151 + }, + { + "epoch": 1.08, + "learning_rate": 3.1981277250577074e-05, + "loss": 0.4716, + "step": 28152 + }, + { + "epoch": 1.08, + "learning_rate": 3.198063606052834e-05, + "loss": 0.4165, + "step": 28153 + }, + { + "epoch": 1.08, + "learning_rate": 3.197999487047961e-05, + "loss": 0.3959, + "step": 28154 + }, + { + "epoch": 1.08, + "learning_rate": 3.1979353680430876e-05, + "loss": 0.2821, + "step": 28155 + }, + { + "epoch": 1.08, + "learning_rate": 3.1978712490382154e-05, + "loss": 0.3985, + "step": 28156 + }, + { + "epoch": 1.08, + "learning_rate": 3.1978071300333426e-05, + "loss": 0.4974, + "step": 28157 + }, + { + "epoch": 1.08, + "learning_rate": 3.197743011028469e-05, + "loss": 0.407, + "step": 28158 + }, + { + "epoch": 1.08, + "learning_rate": 3.197678892023596e-05, + "loss": 0.4944, + "step": 28159 + }, + { + "epoch": 1.08, + "learning_rate": 3.197614773018723e-05, + "loss": 0.3054, + "step": 28160 + }, + { + "epoch": 1.08, + "learning_rate": 3.19755065401385e-05, + "loss": 0.3699, + "step": 28161 + }, + { + "epoch": 1.08, + "learning_rate": 3.1974865350089764e-05, + "loss": 0.3175, + "step": 28162 + }, + { + "epoch": 1.08, + "learning_rate": 3.1974224160041036e-05, + "loss": 0.3537, + "step": 28163 + }, + { + "epoch": 1.08, + "learning_rate": 3.197358296999231e-05, + "loss": 0.3898, + "step": 28164 + }, + { + "epoch": 1.08, + "learning_rate": 3.197294177994358e-05, + "loss": 0.4429, + "step": 28165 + }, + { + "epoch": 1.08, + "learning_rate": 3.197230058989485e-05, + "loss": 0.3631, + "step": 28166 + }, + { + "epoch": 1.08, + "learning_rate": 3.1971659399846116e-05, + "loss": 0.3715, + "step": 28167 + }, + { + "epoch": 1.08, + "learning_rate": 3.197101820979739e-05, + "loss": 0.4076, + "step": 28168 + }, + { + "epoch": 1.08, + "learning_rate": 3.197037701974865e-05, + "loss": 0.3655, + "step": 28169 + }, + { + "epoch": 1.08, + "learning_rate": 3.1969735829699925e-05, + "loss": 0.3313, + "step": 28170 + }, + { + "epoch": 1.08, + "learning_rate": 3.196909463965119e-05, + "loss": 0.4555, + "step": 28171 + }, + { + "epoch": 1.08, + "learning_rate": 3.196845344960246e-05, + "loss": 0.2946, + "step": 28172 + }, + { + "epoch": 1.08, + "learning_rate": 3.196781225955373e-05, + "loss": 0.4859, + "step": 28173 + }, + { + "epoch": 1.08, + "learning_rate": 3.1967171069505005e-05, + "loss": 0.3925, + "step": 28174 + }, + { + "epoch": 1.08, + "learning_rate": 3.1966529879456276e-05, + "loss": 0.3089, + "step": 28175 + }, + { + "epoch": 1.08, + "learning_rate": 3.196588868940754e-05, + "loss": 0.4371, + "step": 28176 + }, + { + "epoch": 1.08, + "learning_rate": 3.196524749935881e-05, + "loss": 0.5747, + "step": 28177 + }, + { + "epoch": 1.08, + "learning_rate": 3.196460630931008e-05, + "loss": 0.3707, + "step": 28178 + }, + { + "epoch": 1.08, + "learning_rate": 3.196396511926135e-05, + "loss": 0.4247, + "step": 28179 + }, + { + "epoch": 1.08, + "learning_rate": 3.196332392921262e-05, + "loss": 0.3293, + "step": 28180 + }, + { + "epoch": 1.08, + "learning_rate": 3.196268273916389e-05, + "loss": 0.418, + "step": 28181 + }, + { + "epoch": 1.08, + "learning_rate": 3.196204154911516e-05, + "loss": 0.3839, + "step": 28182 + }, + { + "epoch": 1.08, + "learning_rate": 3.196140035906643e-05, + "loss": 0.3955, + "step": 28183 + }, + { + "epoch": 1.08, + "learning_rate": 3.19607591690177e-05, + "loss": 0.5212, + "step": 28184 + }, + { + "epoch": 1.08, + "learning_rate": 3.1960117978968967e-05, + "loss": 0.5113, + "step": 28185 + }, + { + "epoch": 1.08, + "learning_rate": 3.195947678892024e-05, + "loss": 0.4764, + "step": 28186 + }, + { + "epoch": 1.08, + "learning_rate": 3.19588355988715e-05, + "loss": 0.3722, + "step": 28187 + }, + { + "epoch": 1.08, + "learning_rate": 3.1958194408822775e-05, + "loss": 0.4683, + "step": 28188 + }, + { + "epoch": 1.08, + "learning_rate": 3.1957553218774047e-05, + "loss": 0.3572, + "step": 28189 + }, + { + "epoch": 1.08, + "learning_rate": 3.195691202872532e-05, + "loss": 0.3725, + "step": 28190 + }, + { + "epoch": 1.08, + "learning_rate": 3.195627083867658e-05, + "loss": 0.4111, + "step": 28191 + }, + { + "epoch": 1.08, + "learning_rate": 3.1955629648627855e-05, + "loss": 0.5378, + "step": 28192 + }, + { + "epoch": 1.08, + "learning_rate": 3.195498845857913e-05, + "loss": 0.5242, + "step": 28193 + }, + { + "epoch": 1.08, + "learning_rate": 3.195434726853039e-05, + "loss": 0.3257, + "step": 28194 + }, + { + "epoch": 1.08, + "learning_rate": 3.195370607848166e-05, + "loss": 0.4671, + "step": 28195 + }, + { + "epoch": 1.08, + "learning_rate": 3.1953064888432935e-05, + "loss": 0.3465, + "step": 28196 + }, + { + "epoch": 1.08, + "learning_rate": 3.195242369838421e-05, + "loss": 0.3917, + "step": 28197 + }, + { + "epoch": 1.08, + "learning_rate": 3.195178250833547e-05, + "loss": 0.4105, + "step": 28198 + }, + { + "epoch": 1.08, + "learning_rate": 3.1951141318286743e-05, + "loss": 0.3714, + "step": 28199 + }, + { + "epoch": 1.08, + "learning_rate": 3.195050012823801e-05, + "loss": 0.4325, + "step": 28200 + }, + { + "epoch": 1.08, + "learning_rate": 3.194985893818928e-05, + "loss": 0.442, + "step": 28201 + }, + { + "epoch": 1.08, + "learning_rate": 3.194921774814055e-05, + "loss": 0.298, + "step": 28202 + }, + { + "epoch": 1.08, + "learning_rate": 3.194857655809182e-05, + "loss": 0.3016, + "step": 28203 + }, + { + "epoch": 1.08, + "learning_rate": 3.194793536804309e-05, + "loss": 0.3466, + "step": 28204 + }, + { + "epoch": 1.08, + "learning_rate": 3.194729417799436e-05, + "loss": 0.247, + "step": 28205 + }, + { + "epoch": 1.08, + "learning_rate": 3.194665298794563e-05, + "loss": 0.3551, + "step": 28206 + }, + { + "epoch": 1.08, + "learning_rate": 3.19460117978969e-05, + "loss": 0.3425, + "step": 28207 + }, + { + "epoch": 1.08, + "learning_rate": 3.194537060784817e-05, + "loss": 0.4295, + "step": 28208 + }, + { + "epoch": 1.08, + "learning_rate": 3.1944729417799434e-05, + "loss": 0.3603, + "step": 28209 + }, + { + "epoch": 1.08, + "learning_rate": 3.1944088227750705e-05, + "loss": 0.4235, + "step": 28210 + }, + { + "epoch": 1.08, + "learning_rate": 3.194344703770198e-05, + "loss": 0.5705, + "step": 28211 + }, + { + "epoch": 1.08, + "learning_rate": 3.194280584765325e-05, + "loss": 0.3887, + "step": 28212 + }, + { + "epoch": 1.08, + "learning_rate": 3.194216465760452e-05, + "loss": 0.3379, + "step": 28213 + }, + { + "epoch": 1.08, + "learning_rate": 3.1941523467555785e-05, + "loss": 0.4269, + "step": 28214 + }, + { + "epoch": 1.08, + "learning_rate": 3.194088227750706e-05, + "loss": 0.3394, + "step": 28215 + }, + { + "epoch": 1.08, + "learning_rate": 3.194024108745832e-05, + "loss": 0.4143, + "step": 28216 + }, + { + "epoch": 1.08, + "learning_rate": 3.1939599897409594e-05, + "loss": 0.4852, + "step": 28217 + }, + { + "epoch": 1.08, + "learning_rate": 3.193895870736086e-05, + "loss": 0.3352, + "step": 28218 + }, + { + "epoch": 1.08, + "learning_rate": 3.193831751731213e-05, + "loss": 0.3966, + "step": 28219 + }, + { + "epoch": 1.08, + "learning_rate": 3.19376763272634e-05, + "loss": 0.3656, + "step": 28220 + }, + { + "epoch": 1.09, + "learning_rate": 3.1937035137214674e-05, + "loss": 0.3539, + "step": 28221 + }, + { + "epoch": 1.09, + "learning_rate": 3.1936393947165945e-05, + "loss": 0.5516, + "step": 28222 + }, + { + "epoch": 1.09, + "learning_rate": 3.193575275711721e-05, + "loss": 0.3124, + "step": 28223 + }, + { + "epoch": 1.09, + "learning_rate": 3.193511156706848e-05, + "loss": 0.5159, + "step": 28224 + }, + { + "epoch": 1.09, + "learning_rate": 3.193447037701975e-05, + "loss": 0.3069, + "step": 28225 + }, + { + "epoch": 1.09, + "learning_rate": 3.193382918697102e-05, + "loss": 0.3336, + "step": 28226 + }, + { + "epoch": 1.09, + "learning_rate": 3.1933187996922284e-05, + "loss": 0.3884, + "step": 28227 + }, + { + "epoch": 1.09, + "learning_rate": 3.193254680687356e-05, + "loss": 0.4252, + "step": 28228 + }, + { + "epoch": 1.09, + "learning_rate": 3.1931905616824834e-05, + "loss": 0.4019, + "step": 28229 + }, + { + "epoch": 1.09, + "learning_rate": 3.19312644267761e-05, + "loss": 0.4253, + "step": 28230 + }, + { + "epoch": 1.09, + "learning_rate": 3.193062323672737e-05, + "loss": 0.3104, + "step": 28231 + }, + { + "epoch": 1.09, + "learning_rate": 3.1929982046678636e-05, + "loss": 0.438, + "step": 28232 + }, + { + "epoch": 1.09, + "learning_rate": 3.192934085662991e-05, + "loss": 0.3592, + "step": 28233 + }, + { + "epoch": 1.09, + "learning_rate": 3.192869966658117e-05, + "loss": 0.4295, + "step": 28234 + }, + { + "epoch": 1.09, + "learning_rate": 3.1928058476532444e-05, + "loss": 0.4751, + "step": 28235 + }, + { + "epoch": 1.09, + "learning_rate": 3.1927417286483716e-05, + "loss": 0.3449, + "step": 28236 + }, + { + "epoch": 1.09, + "learning_rate": 3.192677609643499e-05, + "loss": 0.308, + "step": 28237 + }, + { + "epoch": 1.09, + "learning_rate": 3.192613490638626e-05, + "loss": 0.3679, + "step": 28238 + }, + { + "epoch": 1.09, + "learning_rate": 3.1925493716337524e-05, + "loss": 0.3196, + "step": 28239 + }, + { + "epoch": 1.09, + "learning_rate": 3.1924852526288796e-05, + "loss": 0.4468, + "step": 28240 + }, + { + "epoch": 1.09, + "learning_rate": 3.192421133624006e-05, + "loss": 0.4924, + "step": 28241 + }, + { + "epoch": 1.09, + "learning_rate": 3.192357014619133e-05, + "loss": 0.4665, + "step": 28242 + }, + { + "epoch": 1.09, + "learning_rate": 3.19229289561426e-05, + "loss": 0.2716, + "step": 28243 + }, + { + "epoch": 1.09, + "learning_rate": 3.192228776609387e-05, + "loss": 0.4223, + "step": 28244 + }, + { + "epoch": 1.09, + "learning_rate": 3.192164657604514e-05, + "loss": 0.4682, + "step": 28245 + }, + { + "epoch": 1.09, + "learning_rate": 3.192100538599641e-05, + "loss": 0.2705, + "step": 28246 + }, + { + "epoch": 1.09, + "learning_rate": 3.1920364195947684e-05, + "loss": 0.4405, + "step": 28247 + }, + { + "epoch": 1.09, + "learning_rate": 3.191972300589895e-05, + "loss": 0.4228, + "step": 28248 + }, + { + "epoch": 1.09, + "learning_rate": 3.191908181585022e-05, + "loss": 0.5426, + "step": 28249 + }, + { + "epoch": 1.09, + "learning_rate": 3.1918440625801486e-05, + "loss": 0.5285, + "step": 28250 + }, + { + "epoch": 1.09, + "learning_rate": 3.191779943575276e-05, + "loss": 0.4086, + "step": 28251 + }, + { + "epoch": 1.09, + "learning_rate": 3.191715824570403e-05, + "loss": 0.5226, + "step": 28252 + }, + { + "epoch": 1.09, + "learning_rate": 3.19165170556553e-05, + "loss": 0.5069, + "step": 28253 + }, + { + "epoch": 1.09, + "learning_rate": 3.1915875865606566e-05, + "loss": 0.4026, + "step": 28254 + }, + { + "epoch": 1.09, + "learning_rate": 3.191523467555784e-05, + "loss": 0.3655, + "step": 28255 + }, + { + "epoch": 1.09, + "learning_rate": 3.191459348550911e-05, + "loss": 0.4252, + "step": 28256 + }, + { + "epoch": 1.09, + "learning_rate": 3.1913952295460374e-05, + "loss": 0.3982, + "step": 28257 + }, + { + "epoch": 1.09, + "learning_rate": 3.1913311105411646e-05, + "loss": 0.4748, + "step": 28258 + }, + { + "epoch": 1.09, + "learning_rate": 3.191266991536291e-05, + "loss": 0.4038, + "step": 28259 + }, + { + "epoch": 1.09, + "learning_rate": 3.191202872531418e-05, + "loss": 0.4138, + "step": 28260 + }, + { + "epoch": 1.09, + "learning_rate": 3.1911387535265454e-05, + "loss": 0.2871, + "step": 28261 + }, + { + "epoch": 1.09, + "learning_rate": 3.1910746345216726e-05, + "loss": 0.4151, + "step": 28262 + }, + { + "epoch": 1.09, + "learning_rate": 3.191010515516799e-05, + "loss": 0.3081, + "step": 28263 + }, + { + "epoch": 1.09, + "learning_rate": 3.190946396511926e-05, + "loss": 0.3129, + "step": 28264 + }, + { + "epoch": 1.09, + "learning_rate": 3.1908822775070534e-05, + "loss": 0.3106, + "step": 28265 + }, + { + "epoch": 1.09, + "learning_rate": 3.19081815850218e-05, + "loss": 0.3489, + "step": 28266 + }, + { + "epoch": 1.09, + "learning_rate": 3.190754039497307e-05, + "loss": 0.4048, + "step": 28267 + }, + { + "epoch": 1.09, + "learning_rate": 3.190689920492434e-05, + "loss": 0.4436, + "step": 28268 + }, + { + "epoch": 1.09, + "learning_rate": 3.1906258014875615e-05, + "loss": 0.4168, + "step": 28269 + }, + { + "epoch": 1.09, + "learning_rate": 3.190561682482688e-05, + "loss": 0.3915, + "step": 28270 + }, + { + "epoch": 1.09, + "learning_rate": 3.190497563477815e-05, + "loss": 0.3871, + "step": 28271 + }, + { + "epoch": 1.09, + "learning_rate": 3.1904334444729416e-05, + "loss": 0.373, + "step": 28272 + }, + { + "epoch": 1.09, + "learning_rate": 3.190369325468069e-05, + "loss": 0.2807, + "step": 28273 + }, + { + "epoch": 1.09, + "learning_rate": 3.190305206463196e-05, + "loss": 0.3855, + "step": 28274 + }, + { + "epoch": 1.09, + "learning_rate": 3.1902410874583224e-05, + "loss": 0.4097, + "step": 28275 + }, + { + "epoch": 1.09, + "learning_rate": 3.1901769684534496e-05, + "loss": 0.3254, + "step": 28276 + }, + { + "epoch": 1.09, + "learning_rate": 3.190112849448577e-05, + "loss": 0.3443, + "step": 28277 + }, + { + "epoch": 1.09, + "learning_rate": 3.190048730443704e-05, + "loss": 0.2977, + "step": 28278 + }, + { + "epoch": 1.09, + "learning_rate": 3.1899846114388305e-05, + "loss": 0.4841, + "step": 28279 + }, + { + "epoch": 1.09, + "learning_rate": 3.1899204924339576e-05, + "loss": 0.3176, + "step": 28280 + }, + { + "epoch": 1.09, + "learning_rate": 3.189856373429084e-05, + "loss": 0.4885, + "step": 28281 + }, + { + "epoch": 1.09, + "learning_rate": 3.189792254424211e-05, + "loss": 0.3066, + "step": 28282 + }, + { + "epoch": 1.09, + "learning_rate": 3.1897281354193385e-05, + "loss": 0.4222, + "step": 28283 + }, + { + "epoch": 1.09, + "learning_rate": 3.1896640164144656e-05, + "loss": 0.4588, + "step": 28284 + }, + { + "epoch": 1.09, + "learning_rate": 3.189599897409593e-05, + "loss": 0.354, + "step": 28285 + }, + { + "epoch": 1.09, + "learning_rate": 3.189535778404719e-05, + "loss": 0.4548, + "step": 28286 + }, + { + "epoch": 1.09, + "learning_rate": 3.1894716593998465e-05, + "loss": 0.4302, + "step": 28287 + }, + { + "epoch": 1.09, + "learning_rate": 3.189407540394973e-05, + "loss": 0.3574, + "step": 28288 + }, + { + "epoch": 1.09, + "learning_rate": 3.1893434213901e-05, + "loss": 0.383, + "step": 28289 + }, + { + "epoch": 1.09, + "learning_rate": 3.1892793023852266e-05, + "loss": 0.381, + "step": 28290 + }, + { + "epoch": 1.09, + "learning_rate": 3.189215183380354e-05, + "loss": 0.4082, + "step": 28291 + }, + { + "epoch": 1.09, + "learning_rate": 3.189151064375481e-05, + "loss": 0.3311, + "step": 28292 + }, + { + "epoch": 1.09, + "learning_rate": 3.189086945370608e-05, + "loss": 0.4602, + "step": 28293 + }, + { + "epoch": 1.09, + "learning_rate": 3.189022826365735e-05, + "loss": 0.6143, + "step": 28294 + }, + { + "epoch": 1.09, + "learning_rate": 3.188958707360862e-05, + "loss": 0.3351, + "step": 28295 + }, + { + "epoch": 1.09, + "learning_rate": 3.188894588355989e-05, + "loss": 0.3656, + "step": 28296 + }, + { + "epoch": 1.09, + "learning_rate": 3.1888304693511155e-05, + "loss": 0.5768, + "step": 28297 + }, + { + "epoch": 1.09, + "learning_rate": 3.1887663503462427e-05, + "loss": 0.3369, + "step": 28298 + }, + { + "epoch": 1.09, + "learning_rate": 3.188702231341369e-05, + "loss": 0.4186, + "step": 28299 + }, + { + "epoch": 1.09, + "learning_rate": 3.188638112336497e-05, + "loss": 0.2847, + "step": 28300 + }, + { + "epoch": 1.09, + "learning_rate": 3.188573993331624e-05, + "loss": 0.3583, + "step": 28301 + }, + { + "epoch": 1.09, + "learning_rate": 3.188509874326751e-05, + "loss": 0.2292, + "step": 28302 + }, + { + "epoch": 1.09, + "learning_rate": 3.188445755321878e-05, + "loss": 0.3419, + "step": 28303 + }, + { + "epoch": 1.09, + "learning_rate": 3.188381636317004e-05, + "loss": 0.3656, + "step": 28304 + }, + { + "epoch": 1.09, + "learning_rate": 3.1883175173121315e-05, + "loss": 0.5367, + "step": 28305 + }, + { + "epoch": 1.09, + "learning_rate": 3.188253398307258e-05, + "loss": 0.4175, + "step": 28306 + }, + { + "epoch": 1.09, + "learning_rate": 3.188189279302385e-05, + "loss": 0.5639, + "step": 28307 + }, + { + "epoch": 1.09, + "learning_rate": 3.188125160297512e-05, + "loss": 0.3845, + "step": 28308 + }, + { + "epoch": 1.09, + "learning_rate": 3.1880610412926395e-05, + "loss": 0.3013, + "step": 28309 + }, + { + "epoch": 1.09, + "learning_rate": 3.187996922287767e-05, + "loss": 0.4457, + "step": 28310 + }, + { + "epoch": 1.09, + "learning_rate": 3.187932803282893e-05, + "loss": 0.4613, + "step": 28311 + }, + { + "epoch": 1.09, + "learning_rate": 3.1878686842780203e-05, + "loss": 0.4141, + "step": 28312 + }, + { + "epoch": 1.09, + "learning_rate": 3.187804565273147e-05, + "loss": 0.3544, + "step": 28313 + }, + { + "epoch": 1.09, + "learning_rate": 3.187740446268274e-05, + "loss": 0.284, + "step": 28314 + }, + { + "epoch": 1.09, + "learning_rate": 3.1876763272634005e-05, + "loss": 0.4443, + "step": 28315 + }, + { + "epoch": 1.09, + "learning_rate": 3.1876122082585284e-05, + "loss": 0.3217, + "step": 28316 + }, + { + "epoch": 1.09, + "learning_rate": 3.187548089253655e-05, + "loss": 0.5655, + "step": 28317 + }, + { + "epoch": 1.09, + "learning_rate": 3.187483970248782e-05, + "loss": 0.298, + "step": 28318 + }, + { + "epoch": 1.09, + "learning_rate": 3.187419851243909e-05, + "loss": 0.3597, + "step": 28319 + }, + { + "epoch": 1.09, + "learning_rate": 3.187355732239036e-05, + "loss": 0.3462, + "step": 28320 + }, + { + "epoch": 1.09, + "learning_rate": 3.187291613234163e-05, + "loss": 0.3931, + "step": 28321 + }, + { + "epoch": 1.09, + "learning_rate": 3.1872274942292894e-05, + "loss": 0.372, + "step": 28322 + }, + { + "epoch": 1.09, + "learning_rate": 3.1871633752244165e-05, + "loss": 0.4262, + "step": 28323 + }, + { + "epoch": 1.09, + "learning_rate": 3.187099256219544e-05, + "loss": 0.3569, + "step": 28324 + }, + { + "epoch": 1.09, + "learning_rate": 3.187035137214671e-05, + "loss": 0.4027, + "step": 28325 + }, + { + "epoch": 1.09, + "learning_rate": 3.1869710182097974e-05, + "loss": 0.377, + "step": 28326 + }, + { + "epoch": 1.09, + "learning_rate": 3.1869068992049245e-05, + "loss": 0.408, + "step": 28327 + }, + { + "epoch": 1.09, + "learning_rate": 3.186842780200052e-05, + "loss": 0.4524, + "step": 28328 + }, + { + "epoch": 1.09, + "learning_rate": 3.186778661195178e-05, + "loss": 0.5026, + "step": 28329 + }, + { + "epoch": 1.09, + "learning_rate": 3.1867145421903054e-05, + "loss": 0.4168, + "step": 28330 + }, + { + "epoch": 1.09, + "learning_rate": 3.186650423185432e-05, + "loss": 0.5277, + "step": 28331 + }, + { + "epoch": 1.09, + "learning_rate": 3.186586304180559e-05, + "loss": 0.4671, + "step": 28332 + }, + { + "epoch": 1.09, + "learning_rate": 3.186522185175686e-05, + "loss": 0.4183, + "step": 28333 + }, + { + "epoch": 1.09, + "learning_rate": 3.1864580661708134e-05, + "loss": 0.2823, + "step": 28334 + }, + { + "epoch": 1.09, + "learning_rate": 3.18639394716594e-05, + "loss": 0.4317, + "step": 28335 + }, + { + "epoch": 1.09, + "learning_rate": 3.186329828161067e-05, + "loss": 0.3394, + "step": 28336 + }, + { + "epoch": 1.09, + "learning_rate": 3.186265709156194e-05, + "loss": 0.3692, + "step": 28337 + }, + { + "epoch": 1.09, + "learning_rate": 3.186201590151321e-05, + "loss": 0.4368, + "step": 28338 + }, + { + "epoch": 1.09, + "learning_rate": 3.186137471146448e-05, + "loss": 0.3524, + "step": 28339 + }, + { + "epoch": 1.09, + "learning_rate": 3.186073352141575e-05, + "loss": 0.4835, + "step": 28340 + }, + { + "epoch": 1.09, + "learning_rate": 3.186009233136702e-05, + "loss": 0.4348, + "step": 28341 + }, + { + "epoch": 1.09, + "learning_rate": 3.185945114131829e-05, + "loss": 0.3663, + "step": 28342 + }, + { + "epoch": 1.09, + "learning_rate": 3.185880995126956e-05, + "loss": 0.3155, + "step": 28343 + }, + { + "epoch": 1.09, + "learning_rate": 3.1858168761220824e-05, + "loss": 0.5661, + "step": 28344 + }, + { + "epoch": 1.09, + "learning_rate": 3.1857527571172096e-05, + "loss": 0.4732, + "step": 28345 + }, + { + "epoch": 1.09, + "learning_rate": 3.185688638112337e-05, + "loss": 0.3327, + "step": 28346 + }, + { + "epoch": 1.09, + "learning_rate": 3.185624519107463e-05, + "loss": 0.3709, + "step": 28347 + }, + { + "epoch": 1.09, + "learning_rate": 3.1855604001025904e-05, + "loss": 0.3893, + "step": 28348 + }, + { + "epoch": 1.09, + "learning_rate": 3.1854962810977176e-05, + "loss": 0.3402, + "step": 28349 + }, + { + "epoch": 1.09, + "learning_rate": 3.185432162092845e-05, + "loss": 0.4179, + "step": 28350 + }, + { + "epoch": 1.09, + "learning_rate": 3.185368043087971e-05, + "loss": 0.3553, + "step": 28351 + }, + { + "epoch": 1.09, + "learning_rate": 3.1853039240830984e-05, + "loss": 0.4371, + "step": 28352 + }, + { + "epoch": 1.09, + "learning_rate": 3.1852398050782256e-05, + "loss": 0.4292, + "step": 28353 + }, + { + "epoch": 1.09, + "learning_rate": 3.185175686073352e-05, + "loss": 0.3928, + "step": 28354 + }, + { + "epoch": 1.09, + "learning_rate": 3.185111567068479e-05, + "loss": 0.3728, + "step": 28355 + }, + { + "epoch": 1.09, + "learning_rate": 3.1850474480636064e-05, + "loss": 0.5405, + "step": 28356 + }, + { + "epoch": 1.09, + "learning_rate": 3.1849833290587336e-05, + "loss": 0.3899, + "step": 28357 + }, + { + "epoch": 1.09, + "learning_rate": 3.18491921005386e-05, + "loss": 0.4391, + "step": 28358 + }, + { + "epoch": 1.09, + "learning_rate": 3.184855091048987e-05, + "loss": 0.3134, + "step": 28359 + }, + { + "epoch": 1.09, + "learning_rate": 3.184790972044114e-05, + "loss": 0.3552, + "step": 28360 + }, + { + "epoch": 1.09, + "learning_rate": 3.184726853039241e-05, + "loss": 0.3896, + "step": 28361 + }, + { + "epoch": 1.09, + "learning_rate": 3.184662734034368e-05, + "loss": 0.3903, + "step": 28362 + }, + { + "epoch": 1.09, + "learning_rate": 3.1845986150294946e-05, + "loss": 0.38, + "step": 28363 + }, + { + "epoch": 1.09, + "learning_rate": 3.184534496024622e-05, + "loss": 0.3242, + "step": 28364 + }, + { + "epoch": 1.09, + "learning_rate": 3.184470377019749e-05, + "loss": 0.4892, + "step": 28365 + }, + { + "epoch": 1.09, + "learning_rate": 3.184406258014876e-05, + "loss": 0.4146, + "step": 28366 + }, + { + "epoch": 1.09, + "learning_rate": 3.1843421390100026e-05, + "loss": 0.4811, + "step": 28367 + }, + { + "epoch": 1.09, + "learning_rate": 3.18427802000513e-05, + "loss": 0.3825, + "step": 28368 + }, + { + "epoch": 1.09, + "learning_rate": 3.184213901000256e-05, + "loss": 0.3959, + "step": 28369 + }, + { + "epoch": 1.09, + "learning_rate": 3.1841497819953834e-05, + "loss": 0.4298, + "step": 28370 + }, + { + "epoch": 1.09, + "learning_rate": 3.1840856629905106e-05, + "loss": 0.3735, + "step": 28371 + }, + { + "epoch": 1.09, + "learning_rate": 3.184021543985638e-05, + "loss": 0.418, + "step": 28372 + }, + { + "epoch": 1.09, + "learning_rate": 3.183957424980765e-05, + "loss": 0.4119, + "step": 28373 + }, + { + "epoch": 1.09, + "learning_rate": 3.1838933059758914e-05, + "loss": 0.4572, + "step": 28374 + }, + { + "epoch": 1.09, + "learning_rate": 3.1838291869710186e-05, + "loss": 0.4182, + "step": 28375 + }, + { + "epoch": 1.09, + "learning_rate": 3.183765067966145e-05, + "loss": 0.4453, + "step": 28376 + }, + { + "epoch": 1.09, + "learning_rate": 3.183700948961272e-05, + "loss": 0.3552, + "step": 28377 + }, + { + "epoch": 1.09, + "learning_rate": 3.183636829956399e-05, + "loss": 0.3801, + "step": 28378 + }, + { + "epoch": 1.09, + "learning_rate": 3.183572710951526e-05, + "loss": 0.393, + "step": 28379 + }, + { + "epoch": 1.09, + "learning_rate": 3.183508591946653e-05, + "loss": 0.4083, + "step": 28380 + }, + { + "epoch": 1.09, + "learning_rate": 3.18344447294178e-05, + "loss": 0.3868, + "step": 28381 + }, + { + "epoch": 1.09, + "learning_rate": 3.1833803539369075e-05, + "loss": 0.4382, + "step": 28382 + }, + { + "epoch": 1.09, + "learning_rate": 3.183316234932034e-05, + "loss": 0.3508, + "step": 28383 + }, + { + "epoch": 1.09, + "learning_rate": 3.183252115927161e-05, + "loss": 0.3985, + "step": 28384 + }, + { + "epoch": 1.09, + "learning_rate": 3.1831879969222876e-05, + "loss": 0.3976, + "step": 28385 + }, + { + "epoch": 1.09, + "learning_rate": 3.183123877917415e-05, + "loss": 0.2748, + "step": 28386 + }, + { + "epoch": 1.09, + "learning_rate": 3.183059758912541e-05, + "loss": 0.4744, + "step": 28387 + }, + { + "epoch": 1.09, + "learning_rate": 3.182995639907669e-05, + "loss": 0.3625, + "step": 28388 + }, + { + "epoch": 1.09, + "learning_rate": 3.182931520902796e-05, + "loss": 0.4367, + "step": 28389 + }, + { + "epoch": 1.09, + "learning_rate": 3.182867401897923e-05, + "loss": 0.5231, + "step": 28390 + }, + { + "epoch": 1.09, + "learning_rate": 3.18280328289305e-05, + "loss": 0.4028, + "step": 28391 + }, + { + "epoch": 1.09, + "learning_rate": 3.1827391638881765e-05, + "loss": 0.3066, + "step": 28392 + }, + { + "epoch": 1.09, + "learning_rate": 3.1826750448833036e-05, + "loss": 0.4078, + "step": 28393 + }, + { + "epoch": 1.09, + "learning_rate": 3.18261092587843e-05, + "loss": 0.5249, + "step": 28394 + }, + { + "epoch": 1.09, + "learning_rate": 3.182546806873557e-05, + "loss": 0.3421, + "step": 28395 + }, + { + "epoch": 1.09, + "learning_rate": 3.1824826878686845e-05, + "loss": 0.3359, + "step": 28396 + }, + { + "epoch": 1.09, + "learning_rate": 3.1824185688638116e-05, + "loss": 0.3646, + "step": 28397 + }, + { + "epoch": 1.09, + "learning_rate": 3.182354449858939e-05, + "loss": 0.4076, + "step": 28398 + }, + { + "epoch": 1.09, + "learning_rate": 3.182290330854065e-05, + "loss": 0.3736, + "step": 28399 + }, + { + "epoch": 1.09, + "learning_rate": 3.1822262118491925e-05, + "loss": 0.4149, + "step": 28400 + }, + { + "epoch": 1.09, + "learning_rate": 3.182162092844319e-05, + "loss": 0.4543, + "step": 28401 + }, + { + "epoch": 1.09, + "learning_rate": 3.182097973839446e-05, + "loss": 0.3956, + "step": 28402 + }, + { + "epoch": 1.09, + "learning_rate": 3.1820338548345726e-05, + "loss": 0.3644, + "step": 28403 + }, + { + "epoch": 1.09, + "learning_rate": 3.1819697358297e-05, + "loss": 0.3682, + "step": 28404 + }, + { + "epoch": 1.09, + "learning_rate": 3.181905616824827e-05, + "loss": 0.4317, + "step": 28405 + }, + { + "epoch": 1.09, + "learning_rate": 3.181841497819954e-05, + "loss": 0.4607, + "step": 28406 + }, + { + "epoch": 1.09, + "learning_rate": 3.181777378815081e-05, + "loss": 0.5648, + "step": 28407 + }, + { + "epoch": 1.09, + "learning_rate": 3.181713259810208e-05, + "loss": 0.3649, + "step": 28408 + }, + { + "epoch": 1.09, + "learning_rate": 3.181649140805335e-05, + "loss": 0.4179, + "step": 28409 + }, + { + "epoch": 1.09, + "learning_rate": 3.1815850218004615e-05, + "loss": 0.3771, + "step": 28410 + }, + { + "epoch": 1.09, + "learning_rate": 3.1815209027955887e-05, + "loss": 0.3773, + "step": 28411 + }, + { + "epoch": 1.09, + "learning_rate": 3.181456783790716e-05, + "loss": 0.2948, + "step": 28412 + }, + { + "epoch": 1.09, + "learning_rate": 3.181392664785843e-05, + "loss": 0.2833, + "step": 28413 + }, + { + "epoch": 1.09, + "learning_rate": 3.1813285457809695e-05, + "loss": 0.4063, + "step": 28414 + }, + { + "epoch": 1.09, + "learning_rate": 3.181264426776097e-05, + "loss": 0.3371, + "step": 28415 + }, + { + "epoch": 1.09, + "learning_rate": 3.181200307771224e-05, + "loss": 0.4202, + "step": 28416 + }, + { + "epoch": 1.09, + "learning_rate": 3.18113618876635e-05, + "loss": 0.4765, + "step": 28417 + }, + { + "epoch": 1.09, + "learning_rate": 3.1810720697614775e-05, + "loss": 0.4319, + "step": 28418 + }, + { + "epoch": 1.09, + "learning_rate": 3.181007950756604e-05, + "loss": 0.4269, + "step": 28419 + }, + { + "epoch": 1.09, + "learning_rate": 3.180943831751731e-05, + "loss": 0.4468, + "step": 28420 + }, + { + "epoch": 1.09, + "learning_rate": 3.1808797127468583e-05, + "loss": 0.3709, + "step": 28421 + }, + { + "epoch": 1.09, + "learning_rate": 3.1808155937419855e-05, + "loss": 0.408, + "step": 28422 + }, + { + "epoch": 1.09, + "learning_rate": 3.180751474737112e-05, + "loss": 0.3008, + "step": 28423 + }, + { + "epoch": 1.09, + "learning_rate": 3.180687355732239e-05, + "loss": 0.3376, + "step": 28424 + }, + { + "epoch": 1.09, + "learning_rate": 3.1806232367273664e-05, + "loss": 0.4401, + "step": 28425 + }, + { + "epoch": 1.09, + "learning_rate": 3.180559117722493e-05, + "loss": 0.3194, + "step": 28426 + }, + { + "epoch": 1.09, + "learning_rate": 3.18049499871762e-05, + "loss": 0.3714, + "step": 28427 + }, + { + "epoch": 1.09, + "learning_rate": 3.180430879712747e-05, + "loss": 0.3339, + "step": 28428 + }, + { + "epoch": 1.09, + "learning_rate": 3.1803667607078744e-05, + "loss": 0.3069, + "step": 28429 + }, + { + "epoch": 1.09, + "learning_rate": 3.180302641703001e-05, + "loss": 0.4141, + "step": 28430 + }, + { + "epoch": 1.09, + "learning_rate": 3.180238522698128e-05, + "loss": 0.4101, + "step": 28431 + }, + { + "epoch": 1.09, + "learning_rate": 3.1801744036932545e-05, + "loss": 0.3746, + "step": 28432 + }, + { + "epoch": 1.09, + "learning_rate": 3.180110284688382e-05, + "loss": 0.5543, + "step": 28433 + }, + { + "epoch": 1.09, + "learning_rate": 3.180046165683509e-05, + "loss": 0.3384, + "step": 28434 + }, + { + "epoch": 1.09, + "learning_rate": 3.1799820466786354e-05, + "loss": 0.3031, + "step": 28435 + }, + { + "epoch": 1.09, + "learning_rate": 3.1799179276737625e-05, + "loss": 0.3836, + "step": 28436 + }, + { + "epoch": 1.09, + "learning_rate": 3.17985380866889e-05, + "loss": 0.3284, + "step": 28437 + }, + { + "epoch": 1.09, + "learning_rate": 3.179789689664017e-05, + "loss": 0.5569, + "step": 28438 + }, + { + "epoch": 1.09, + "learning_rate": 3.1797255706591434e-05, + "loss": 0.5449, + "step": 28439 + }, + { + "epoch": 1.09, + "learning_rate": 3.1796614516542705e-05, + "loss": 0.5148, + "step": 28440 + }, + { + "epoch": 1.09, + "learning_rate": 3.179597332649397e-05, + "loss": 0.4719, + "step": 28441 + }, + { + "epoch": 1.09, + "learning_rate": 3.179533213644524e-05, + "loss": 0.3585, + "step": 28442 + }, + { + "epoch": 1.09, + "learning_rate": 3.1794690946396514e-05, + "loss": 0.476, + "step": 28443 + }, + { + "epoch": 1.09, + "learning_rate": 3.1794049756347785e-05, + "loss": 0.5019, + "step": 28444 + }, + { + "epoch": 1.09, + "learning_rate": 3.179340856629906e-05, + "loss": 0.2913, + "step": 28445 + }, + { + "epoch": 1.09, + "learning_rate": 3.179276737625032e-05, + "loss": 0.4934, + "step": 28446 + }, + { + "epoch": 1.09, + "learning_rate": 3.1792126186201594e-05, + "loss": 0.3867, + "step": 28447 + }, + { + "epoch": 1.09, + "learning_rate": 3.179148499615286e-05, + "loss": 0.3865, + "step": 28448 + }, + { + "epoch": 1.09, + "learning_rate": 3.179084380610413e-05, + "loss": 0.5343, + "step": 28449 + }, + { + "epoch": 1.09, + "learning_rate": 3.1790202616055395e-05, + "loss": 0.4438, + "step": 28450 + }, + { + "epoch": 1.09, + "learning_rate": 3.178956142600667e-05, + "loss": 0.454, + "step": 28451 + }, + { + "epoch": 1.09, + "learning_rate": 3.178892023595794e-05, + "loss": 0.3384, + "step": 28452 + }, + { + "epoch": 1.09, + "learning_rate": 3.178827904590921e-05, + "loss": 0.3701, + "step": 28453 + }, + { + "epoch": 1.09, + "learning_rate": 3.178763785586048e-05, + "loss": 0.3266, + "step": 28454 + }, + { + "epoch": 1.09, + "learning_rate": 3.178699666581175e-05, + "loss": 0.342, + "step": 28455 + }, + { + "epoch": 1.09, + "learning_rate": 3.178635547576302e-05, + "loss": 0.4222, + "step": 28456 + }, + { + "epoch": 1.09, + "learning_rate": 3.1785714285714284e-05, + "loss": 0.4682, + "step": 28457 + }, + { + "epoch": 1.09, + "learning_rate": 3.1785073095665556e-05, + "loss": 0.5392, + "step": 28458 + }, + { + "epoch": 1.09, + "learning_rate": 3.178443190561682e-05, + "loss": 0.4166, + "step": 28459 + }, + { + "epoch": 1.09, + "learning_rate": 3.17837907155681e-05, + "loss": 0.3819, + "step": 28460 + }, + { + "epoch": 1.09, + "learning_rate": 3.178314952551937e-05, + "loss": 0.3347, + "step": 28461 + }, + { + "epoch": 1.09, + "learning_rate": 3.1782508335470636e-05, + "loss": 0.397, + "step": 28462 + }, + { + "epoch": 1.09, + "learning_rate": 3.178186714542191e-05, + "loss": 0.3922, + "step": 28463 + }, + { + "epoch": 1.09, + "learning_rate": 3.178122595537317e-05, + "loss": 0.3141, + "step": 28464 + }, + { + "epoch": 1.09, + "learning_rate": 3.1780584765324444e-05, + "loss": 0.311, + "step": 28465 + }, + { + "epoch": 1.09, + "learning_rate": 3.177994357527571e-05, + "loss": 0.425, + "step": 28466 + }, + { + "epoch": 1.09, + "learning_rate": 3.177930238522698e-05, + "loss": 0.3672, + "step": 28467 + }, + { + "epoch": 1.09, + "learning_rate": 3.177866119517825e-05, + "loss": 0.4098, + "step": 28468 + }, + { + "epoch": 1.09, + "learning_rate": 3.1778020005129524e-05, + "loss": 0.3891, + "step": 28469 + }, + { + "epoch": 1.09, + "learning_rate": 3.1777378815080796e-05, + "loss": 0.3612, + "step": 28470 + }, + { + "epoch": 1.09, + "learning_rate": 3.177673762503206e-05, + "loss": 0.3948, + "step": 28471 + }, + { + "epoch": 1.09, + "learning_rate": 3.177609643498333e-05, + "loss": 0.4223, + "step": 28472 + }, + { + "epoch": 1.09, + "learning_rate": 3.17754552449346e-05, + "loss": 0.2822, + "step": 28473 + }, + { + "epoch": 1.09, + "learning_rate": 3.177481405488587e-05, + "loss": 0.5054, + "step": 28474 + }, + { + "epoch": 1.09, + "learning_rate": 3.1774172864837134e-05, + "loss": 0.3099, + "step": 28475 + }, + { + "epoch": 1.09, + "learning_rate": 3.177353167478841e-05, + "loss": 0.3357, + "step": 28476 + }, + { + "epoch": 1.09, + "learning_rate": 3.177289048473968e-05, + "loss": 0.2886, + "step": 28477 + }, + { + "epoch": 1.09, + "learning_rate": 3.177224929469095e-05, + "loss": 0.2999, + "step": 28478 + }, + { + "epoch": 1.09, + "learning_rate": 3.177160810464222e-05, + "loss": 0.4615, + "step": 28479 + }, + { + "epoch": 1.09, + "learning_rate": 3.1770966914593486e-05, + "loss": 0.3569, + "step": 28480 + }, + { + "epoch": 1.1, + "learning_rate": 3.177032572454476e-05, + "loss": 0.3553, + "step": 28481 + }, + { + "epoch": 1.1, + "learning_rate": 3.176968453449602e-05, + "loss": 0.3604, + "step": 28482 + }, + { + "epoch": 1.1, + "learning_rate": 3.1769043344447294e-05, + "loss": 0.4116, + "step": 28483 + }, + { + "epoch": 1.1, + "learning_rate": 3.1768402154398566e-05, + "loss": 0.3968, + "step": 28484 + }, + { + "epoch": 1.1, + "learning_rate": 3.176776096434984e-05, + "loss": 0.4311, + "step": 28485 + }, + { + "epoch": 1.1, + "learning_rate": 3.17671197743011e-05, + "loss": 0.5411, + "step": 28486 + }, + { + "epoch": 1.1, + "learning_rate": 3.1766478584252374e-05, + "loss": 0.5487, + "step": 28487 + }, + { + "epoch": 1.1, + "learning_rate": 3.1765837394203646e-05, + "loss": 0.2938, + "step": 28488 + }, + { + "epoch": 1.1, + "learning_rate": 3.176519620415491e-05, + "loss": 0.4556, + "step": 28489 + }, + { + "epoch": 1.1, + "learning_rate": 3.176455501410618e-05, + "loss": 0.5206, + "step": 28490 + }, + { + "epoch": 1.1, + "learning_rate": 3.176391382405745e-05, + "loss": 0.2168, + "step": 28491 + }, + { + "epoch": 1.1, + "learning_rate": 3.176327263400872e-05, + "loss": 0.3346, + "step": 28492 + }, + { + "epoch": 1.1, + "learning_rate": 3.176263144395999e-05, + "loss": 0.3515, + "step": 28493 + }, + { + "epoch": 1.1, + "learning_rate": 3.176199025391126e-05, + "loss": 0.4156, + "step": 28494 + }, + { + "epoch": 1.1, + "learning_rate": 3.176134906386253e-05, + "loss": 0.3815, + "step": 28495 + }, + { + "epoch": 1.1, + "learning_rate": 3.17607078738138e-05, + "loss": 0.3575, + "step": 28496 + }, + { + "epoch": 1.1, + "learning_rate": 3.176006668376507e-05, + "loss": 0.3423, + "step": 28497 + }, + { + "epoch": 1.1, + "learning_rate": 3.1759425493716336e-05, + "loss": 0.3113, + "step": 28498 + }, + { + "epoch": 1.1, + "learning_rate": 3.175878430366761e-05, + "loss": 0.3614, + "step": 28499 + }, + { + "epoch": 1.1, + "learning_rate": 3.175814311361888e-05, + "loss": 0.5507, + "step": 28500 + }, + { + "epoch": 1.1, + "learning_rate": 3.175750192357015e-05, + "loss": 0.5006, + "step": 28501 + }, + { + "epoch": 1.1, + "learning_rate": 3.1756860733521416e-05, + "loss": 0.4357, + "step": 28502 + }, + { + "epoch": 1.1, + "learning_rate": 3.175621954347269e-05, + "loss": 0.3982, + "step": 28503 + }, + { + "epoch": 1.1, + "learning_rate": 3.175557835342395e-05, + "loss": 0.4772, + "step": 28504 + }, + { + "epoch": 1.1, + "learning_rate": 3.1754937163375225e-05, + "loss": 0.4942, + "step": 28505 + }, + { + "epoch": 1.1, + "learning_rate": 3.1754295973326496e-05, + "loss": 0.3844, + "step": 28506 + }, + { + "epoch": 1.1, + "learning_rate": 3.175365478327776e-05, + "loss": 0.3568, + "step": 28507 + }, + { + "epoch": 1.1, + "learning_rate": 3.175301359322903e-05, + "loss": 0.5534, + "step": 28508 + }, + { + "epoch": 1.1, + "learning_rate": 3.1752372403180305e-05, + "loss": 0.3577, + "step": 28509 + }, + { + "epoch": 1.1, + "learning_rate": 3.1751731213131576e-05, + "loss": 0.3084, + "step": 28510 + }, + { + "epoch": 1.1, + "learning_rate": 3.175109002308284e-05, + "loss": 0.3818, + "step": 28511 + }, + { + "epoch": 1.1, + "learning_rate": 3.175044883303411e-05, + "loss": 0.3586, + "step": 28512 + }, + { + "epoch": 1.1, + "learning_rate": 3.174980764298538e-05, + "loss": 0.3603, + "step": 28513 + }, + { + "epoch": 1.1, + "learning_rate": 3.174916645293665e-05, + "loss": 0.3837, + "step": 28514 + }, + { + "epoch": 1.1, + "learning_rate": 3.174852526288792e-05, + "loss": 0.4391, + "step": 28515 + }, + { + "epoch": 1.1, + "learning_rate": 3.174788407283919e-05, + "loss": 0.3683, + "step": 28516 + }, + { + "epoch": 1.1, + "learning_rate": 3.1747242882790465e-05, + "loss": 0.3358, + "step": 28517 + }, + { + "epoch": 1.1, + "learning_rate": 3.174660169274173e-05, + "loss": 0.5301, + "step": 28518 + }, + { + "epoch": 1.1, + "learning_rate": 3.1745960502693e-05, + "loss": 0.4088, + "step": 28519 + }, + { + "epoch": 1.1, + "learning_rate": 3.1745319312644267e-05, + "loss": 0.4207, + "step": 28520 + }, + { + "epoch": 1.1, + "learning_rate": 3.174467812259554e-05, + "loss": 0.4282, + "step": 28521 + }, + { + "epoch": 1.1, + "learning_rate": 3.17440369325468e-05, + "loss": 0.368, + "step": 28522 + }, + { + "epoch": 1.1, + "learning_rate": 3.1743395742498075e-05, + "loss": 0.4725, + "step": 28523 + }, + { + "epoch": 1.1, + "learning_rate": 3.1742754552449347e-05, + "loss": 0.4313, + "step": 28524 + }, + { + "epoch": 1.1, + "learning_rate": 3.174211336240062e-05, + "loss": 0.4041, + "step": 28525 + }, + { + "epoch": 1.1, + "learning_rate": 3.174147217235189e-05, + "loss": 0.4249, + "step": 28526 + }, + { + "epoch": 1.1, + "learning_rate": 3.1740830982303155e-05, + "loss": 0.4185, + "step": 28527 + }, + { + "epoch": 1.1, + "learning_rate": 3.174018979225443e-05, + "loss": 0.4864, + "step": 28528 + }, + { + "epoch": 1.1, + "learning_rate": 3.173954860220569e-05, + "loss": 0.3321, + "step": 28529 + }, + { + "epoch": 1.1, + "learning_rate": 3.173890741215696e-05, + "loss": 0.462, + "step": 28530 + }, + { + "epoch": 1.1, + "learning_rate": 3.173826622210823e-05, + "loss": 0.4694, + "step": 28531 + }, + { + "epoch": 1.1, + "learning_rate": 3.173762503205951e-05, + "loss": 0.4797, + "step": 28532 + }, + { + "epoch": 1.1, + "learning_rate": 3.173698384201078e-05, + "loss": 0.4011, + "step": 28533 + }, + { + "epoch": 1.1, + "learning_rate": 3.1736342651962043e-05, + "loss": 0.4272, + "step": 28534 + }, + { + "epoch": 1.1, + "learning_rate": 3.1735701461913315e-05, + "loss": 0.4848, + "step": 28535 + }, + { + "epoch": 1.1, + "learning_rate": 3.173506027186458e-05, + "loss": 0.3413, + "step": 28536 + }, + { + "epoch": 1.1, + "learning_rate": 3.173441908181585e-05, + "loss": 0.3786, + "step": 28537 + }, + { + "epoch": 1.1, + "learning_rate": 3.173377789176712e-05, + "loss": 0.4699, + "step": 28538 + }, + { + "epoch": 1.1, + "learning_rate": 3.173313670171839e-05, + "loss": 0.4838, + "step": 28539 + }, + { + "epoch": 1.1, + "learning_rate": 3.173249551166966e-05, + "loss": 0.3944, + "step": 28540 + }, + { + "epoch": 1.1, + "learning_rate": 3.173185432162093e-05, + "loss": 0.41, + "step": 28541 + }, + { + "epoch": 1.1, + "learning_rate": 3.1731213131572204e-05, + "loss": 0.3571, + "step": 28542 + }, + { + "epoch": 1.1, + "learning_rate": 3.173057194152347e-05, + "loss": 0.5512, + "step": 28543 + }, + { + "epoch": 1.1, + "learning_rate": 3.172993075147474e-05, + "loss": 0.4211, + "step": 28544 + }, + { + "epoch": 1.1, + "learning_rate": 3.1729289561426005e-05, + "loss": 0.2911, + "step": 28545 + }, + { + "epoch": 1.1, + "learning_rate": 3.172864837137728e-05, + "loss": 0.3156, + "step": 28546 + }, + { + "epoch": 1.1, + "learning_rate": 3.172800718132854e-05, + "loss": 0.4015, + "step": 28547 + }, + { + "epoch": 1.1, + "learning_rate": 3.172736599127982e-05, + "loss": 0.379, + "step": 28548 + }, + { + "epoch": 1.1, + "learning_rate": 3.172672480123109e-05, + "loss": 0.3684, + "step": 28549 + }, + { + "epoch": 1.1, + "learning_rate": 3.172608361118236e-05, + "loss": 0.3588, + "step": 28550 + }, + { + "epoch": 1.1, + "learning_rate": 3.172544242113363e-05, + "loss": 0.3227, + "step": 28551 + }, + { + "epoch": 1.1, + "learning_rate": 3.1724801231084894e-05, + "loss": 0.2949, + "step": 28552 + }, + { + "epoch": 1.1, + "learning_rate": 3.1724160041036165e-05, + "loss": 0.3638, + "step": 28553 + }, + { + "epoch": 1.1, + "learning_rate": 3.172351885098743e-05, + "loss": 0.3368, + "step": 28554 + }, + { + "epoch": 1.1, + "learning_rate": 3.17228776609387e-05, + "loss": 0.5195, + "step": 28555 + }, + { + "epoch": 1.1, + "learning_rate": 3.1722236470889974e-05, + "loss": 0.3548, + "step": 28556 + }, + { + "epoch": 1.1, + "learning_rate": 3.1721595280841245e-05, + "loss": 0.3786, + "step": 28557 + }, + { + "epoch": 1.1, + "learning_rate": 3.172095409079252e-05, + "loss": 0.3894, + "step": 28558 + }, + { + "epoch": 1.1, + "learning_rate": 3.172031290074378e-05, + "loss": 0.4168, + "step": 28559 + }, + { + "epoch": 1.1, + "learning_rate": 3.1719671710695054e-05, + "loss": 0.5704, + "step": 28560 + }, + { + "epoch": 1.1, + "learning_rate": 3.171903052064632e-05, + "loss": 0.3986, + "step": 28561 + }, + { + "epoch": 1.1, + "learning_rate": 3.171838933059759e-05, + "loss": 0.4865, + "step": 28562 + }, + { + "epoch": 1.1, + "learning_rate": 3.1717748140548855e-05, + "loss": 0.3908, + "step": 28563 + }, + { + "epoch": 1.1, + "learning_rate": 3.1717106950500134e-05, + "loss": 0.5395, + "step": 28564 + }, + { + "epoch": 1.1, + "learning_rate": 3.17164657604514e-05, + "loss": 0.3275, + "step": 28565 + }, + { + "epoch": 1.1, + "learning_rate": 3.171582457040267e-05, + "loss": 0.3931, + "step": 28566 + }, + { + "epoch": 1.1, + "learning_rate": 3.171518338035394e-05, + "loss": 0.3763, + "step": 28567 + }, + { + "epoch": 1.1, + "learning_rate": 3.171454219030521e-05, + "loss": 0.4076, + "step": 28568 + }, + { + "epoch": 1.1, + "learning_rate": 3.171390100025648e-05, + "loss": 0.3817, + "step": 28569 + }, + { + "epoch": 1.1, + "learning_rate": 3.1713259810207744e-05, + "loss": 0.3169, + "step": 28570 + }, + { + "epoch": 1.1, + "learning_rate": 3.1712618620159016e-05, + "loss": 0.4609, + "step": 28571 + }, + { + "epoch": 1.1, + "learning_rate": 3.171197743011029e-05, + "loss": 0.4394, + "step": 28572 + }, + { + "epoch": 1.1, + "learning_rate": 3.171133624006156e-05, + "loss": 0.4756, + "step": 28573 + }, + { + "epoch": 1.1, + "learning_rate": 3.1710695050012824e-05, + "loss": 0.4582, + "step": 28574 + }, + { + "epoch": 1.1, + "learning_rate": 3.1710053859964096e-05, + "loss": 0.3758, + "step": 28575 + }, + { + "epoch": 1.1, + "learning_rate": 3.170941266991537e-05, + "loss": 0.328, + "step": 28576 + }, + { + "epoch": 1.1, + "learning_rate": 3.170877147986663e-05, + "loss": 0.418, + "step": 28577 + }, + { + "epoch": 1.1, + "learning_rate": 3.1708130289817904e-05, + "loss": 0.3541, + "step": 28578 + }, + { + "epoch": 1.1, + "learning_rate": 3.170748909976917e-05, + "loss": 0.2988, + "step": 28579 + }, + { + "epoch": 1.1, + "learning_rate": 3.170684790972044e-05, + "loss": 0.3096, + "step": 28580 + }, + { + "epoch": 1.1, + "learning_rate": 3.170620671967171e-05, + "loss": 0.3246, + "step": 28581 + }, + { + "epoch": 1.1, + "learning_rate": 3.1705565529622984e-05, + "loss": 0.3397, + "step": 28582 + }, + { + "epoch": 1.1, + "learning_rate": 3.170492433957425e-05, + "loss": 0.4427, + "step": 28583 + }, + { + "epoch": 1.1, + "learning_rate": 3.170428314952552e-05, + "loss": 0.3851, + "step": 28584 + }, + { + "epoch": 1.1, + "learning_rate": 3.170364195947679e-05, + "loss": 0.3784, + "step": 28585 + }, + { + "epoch": 1.1, + "learning_rate": 3.170300076942806e-05, + "loss": 0.508, + "step": 28586 + }, + { + "epoch": 1.1, + "learning_rate": 3.170235957937933e-05, + "loss": 0.3918, + "step": 28587 + }, + { + "epoch": 1.1, + "learning_rate": 3.17017183893306e-05, + "loss": 0.4207, + "step": 28588 + }, + { + "epoch": 1.1, + "learning_rate": 3.170107719928187e-05, + "loss": 0.4724, + "step": 28589 + }, + { + "epoch": 1.1, + "learning_rate": 3.170043600923314e-05, + "loss": 0.3166, + "step": 28590 + }, + { + "epoch": 1.1, + "learning_rate": 3.169979481918441e-05, + "loss": 0.4004, + "step": 28591 + }, + { + "epoch": 1.1, + "learning_rate": 3.1699153629135674e-05, + "loss": 0.3859, + "step": 28592 + }, + { + "epoch": 1.1, + "learning_rate": 3.1698512439086946e-05, + "loss": 0.4677, + "step": 28593 + }, + { + "epoch": 1.1, + "learning_rate": 3.169787124903822e-05, + "loss": 0.5166, + "step": 28594 + }, + { + "epoch": 1.1, + "learning_rate": 3.169723005898948e-05, + "loss": 0.3761, + "step": 28595 + }, + { + "epoch": 1.1, + "learning_rate": 3.1696588868940754e-05, + "loss": 0.4082, + "step": 28596 + }, + { + "epoch": 1.1, + "learning_rate": 3.1695947678892026e-05, + "loss": 0.4846, + "step": 28597 + }, + { + "epoch": 1.1, + "learning_rate": 3.16953064888433e-05, + "loss": 0.4124, + "step": 28598 + }, + { + "epoch": 1.1, + "learning_rate": 3.169466529879456e-05, + "loss": 0.353, + "step": 28599 + }, + { + "epoch": 1.1, + "learning_rate": 3.1694024108745834e-05, + "loss": 0.5142, + "step": 28600 + }, + { + "epoch": 1.1, + "learning_rate": 3.16933829186971e-05, + "loss": 0.484, + "step": 28601 + }, + { + "epoch": 1.1, + "learning_rate": 3.169274172864837e-05, + "loss": 0.4869, + "step": 28602 + }, + { + "epoch": 1.1, + "learning_rate": 3.169210053859964e-05, + "loss": 0.3731, + "step": 28603 + }, + { + "epoch": 1.1, + "learning_rate": 3.1691459348550915e-05, + "loss": 0.436, + "step": 28604 + }, + { + "epoch": 1.1, + "learning_rate": 3.1690818158502186e-05, + "loss": 0.3241, + "step": 28605 + }, + { + "epoch": 1.1, + "learning_rate": 3.169017696845345e-05, + "loss": 0.4616, + "step": 28606 + }, + { + "epoch": 1.1, + "learning_rate": 3.168953577840472e-05, + "loss": 0.4734, + "step": 28607 + }, + { + "epoch": 1.1, + "learning_rate": 3.168889458835599e-05, + "loss": 0.4076, + "step": 28608 + }, + { + "epoch": 1.1, + "learning_rate": 3.168825339830726e-05, + "loss": 0.4122, + "step": 28609 + }, + { + "epoch": 1.1, + "learning_rate": 3.1687612208258525e-05, + "loss": 0.4061, + "step": 28610 + }, + { + "epoch": 1.1, + "learning_rate": 3.1686971018209796e-05, + "loss": 0.4132, + "step": 28611 + }, + { + "epoch": 1.1, + "learning_rate": 3.168632982816107e-05, + "loss": 0.5288, + "step": 28612 + }, + { + "epoch": 1.1, + "learning_rate": 3.168568863811234e-05, + "loss": 0.3585, + "step": 28613 + }, + { + "epoch": 1.1, + "learning_rate": 3.168504744806361e-05, + "loss": 0.3622, + "step": 28614 + }, + { + "epoch": 1.1, + "learning_rate": 3.1684406258014876e-05, + "loss": 0.4126, + "step": 28615 + }, + { + "epoch": 1.1, + "learning_rate": 3.168376506796615e-05, + "loss": 0.4266, + "step": 28616 + }, + { + "epoch": 1.1, + "learning_rate": 3.168312387791741e-05, + "loss": 0.4102, + "step": 28617 + }, + { + "epoch": 1.1, + "learning_rate": 3.1682482687868685e-05, + "loss": 0.49, + "step": 28618 + }, + { + "epoch": 1.1, + "learning_rate": 3.168184149781995e-05, + "loss": 0.3936, + "step": 28619 + }, + { + "epoch": 1.1, + "learning_rate": 3.168120030777123e-05, + "loss": 0.3795, + "step": 28620 + }, + { + "epoch": 1.1, + "learning_rate": 3.16805591177225e-05, + "loss": 0.4724, + "step": 28621 + }, + { + "epoch": 1.1, + "learning_rate": 3.1679917927673765e-05, + "loss": 0.5001, + "step": 28622 + }, + { + "epoch": 1.1, + "learning_rate": 3.1679276737625036e-05, + "loss": 0.361, + "step": 28623 + }, + { + "epoch": 1.1, + "learning_rate": 3.16786355475763e-05, + "loss": 0.4625, + "step": 28624 + }, + { + "epoch": 1.1, + "learning_rate": 3.167799435752757e-05, + "loss": 0.439, + "step": 28625 + }, + { + "epoch": 1.1, + "learning_rate": 3.167735316747884e-05, + "loss": 0.395, + "step": 28626 + }, + { + "epoch": 1.1, + "learning_rate": 3.167671197743011e-05, + "loss": 0.3923, + "step": 28627 + }, + { + "epoch": 1.1, + "learning_rate": 3.167607078738138e-05, + "loss": 0.3823, + "step": 28628 + }, + { + "epoch": 1.1, + "learning_rate": 3.167542959733265e-05, + "loss": 0.3922, + "step": 28629 + }, + { + "epoch": 1.1, + "learning_rate": 3.1674788407283925e-05, + "loss": 0.4527, + "step": 28630 + }, + { + "epoch": 1.1, + "learning_rate": 3.167414721723519e-05, + "loss": 0.4116, + "step": 28631 + }, + { + "epoch": 1.1, + "learning_rate": 3.167350602718646e-05, + "loss": 0.3844, + "step": 28632 + }, + { + "epoch": 1.1, + "learning_rate": 3.1672864837137727e-05, + "loss": 0.2632, + "step": 28633 + }, + { + "epoch": 1.1, + "learning_rate": 3.1672223647089e-05, + "loss": 0.307, + "step": 28634 + }, + { + "epoch": 1.1, + "learning_rate": 3.167158245704026e-05, + "loss": 0.3411, + "step": 28635 + }, + { + "epoch": 1.1, + "learning_rate": 3.167094126699154e-05, + "loss": 0.4837, + "step": 28636 + }, + { + "epoch": 1.1, + "learning_rate": 3.167030007694281e-05, + "loss": 0.4199, + "step": 28637 + }, + { + "epoch": 1.1, + "learning_rate": 3.166965888689408e-05, + "loss": 0.3373, + "step": 28638 + }, + { + "epoch": 1.1, + "learning_rate": 3.166901769684535e-05, + "loss": 0.4607, + "step": 28639 + }, + { + "epoch": 1.1, + "learning_rate": 3.1668376506796615e-05, + "loss": 0.4181, + "step": 28640 + }, + { + "epoch": 1.1, + "learning_rate": 3.166773531674789e-05, + "loss": 0.4164, + "step": 28641 + }, + { + "epoch": 1.1, + "learning_rate": 3.166709412669915e-05, + "loss": 0.4446, + "step": 28642 + }, + { + "epoch": 1.1, + "learning_rate": 3.1666452936650423e-05, + "loss": 0.3881, + "step": 28643 + }, + { + "epoch": 1.1, + "learning_rate": 3.1665811746601695e-05, + "loss": 0.3681, + "step": 28644 + }, + { + "epoch": 1.1, + "learning_rate": 3.166517055655297e-05, + "loss": 0.476, + "step": 28645 + }, + { + "epoch": 1.1, + "learning_rate": 3.166452936650423e-05, + "loss": 0.4742, + "step": 28646 + }, + { + "epoch": 1.1, + "learning_rate": 3.1663888176455503e-05, + "loss": 0.4129, + "step": 28647 + }, + { + "epoch": 1.1, + "learning_rate": 3.1663246986406775e-05, + "loss": 0.3659, + "step": 28648 + }, + { + "epoch": 1.1, + "learning_rate": 3.166260579635804e-05, + "loss": 0.4746, + "step": 28649 + }, + { + "epoch": 1.1, + "learning_rate": 3.166196460630931e-05, + "loss": 0.2809, + "step": 28650 + }, + { + "epoch": 1.1, + "learning_rate": 3.166132341626058e-05, + "loss": 0.3083, + "step": 28651 + }, + { + "epoch": 1.1, + "learning_rate": 3.166068222621185e-05, + "loss": 0.3049, + "step": 28652 + }, + { + "epoch": 1.1, + "learning_rate": 3.166004103616312e-05, + "loss": 0.3184, + "step": 28653 + }, + { + "epoch": 1.1, + "learning_rate": 3.165939984611439e-05, + "loss": 0.4526, + "step": 28654 + }, + { + "epoch": 1.1, + "learning_rate": 3.165875865606566e-05, + "loss": 0.33, + "step": 28655 + }, + { + "epoch": 1.1, + "learning_rate": 3.165811746601693e-05, + "loss": 0.3454, + "step": 28656 + }, + { + "epoch": 1.1, + "learning_rate": 3.16574762759682e-05, + "loss": 0.338, + "step": 28657 + }, + { + "epoch": 1.1, + "learning_rate": 3.1656835085919465e-05, + "loss": 0.4992, + "step": 28658 + }, + { + "epoch": 1.1, + "learning_rate": 3.165619389587074e-05, + "loss": 0.4331, + "step": 28659 + }, + { + "epoch": 1.1, + "learning_rate": 3.165555270582201e-05, + "loss": 0.2962, + "step": 28660 + }, + { + "epoch": 1.1, + "learning_rate": 3.165491151577328e-05, + "loss": 0.4089, + "step": 28661 + }, + { + "epoch": 1.1, + "learning_rate": 3.1654270325724545e-05, + "loss": 0.4086, + "step": 28662 + }, + { + "epoch": 1.1, + "learning_rate": 3.165362913567582e-05, + "loss": 0.3735, + "step": 28663 + }, + { + "epoch": 1.1, + "learning_rate": 3.165298794562708e-05, + "loss": 0.3993, + "step": 28664 + }, + { + "epoch": 1.1, + "learning_rate": 3.1652346755578354e-05, + "loss": 0.3417, + "step": 28665 + }, + { + "epoch": 1.1, + "learning_rate": 3.1651705565529625e-05, + "loss": 0.3673, + "step": 28666 + }, + { + "epoch": 1.1, + "learning_rate": 3.165106437548089e-05, + "loss": 0.3414, + "step": 28667 + }, + { + "epoch": 1.1, + "learning_rate": 3.165042318543216e-05, + "loss": 0.5015, + "step": 28668 + }, + { + "epoch": 1.1, + "learning_rate": 3.1649781995383434e-05, + "loss": 0.4125, + "step": 28669 + }, + { + "epoch": 1.1, + "learning_rate": 3.1649140805334706e-05, + "loss": 0.4605, + "step": 28670 + }, + { + "epoch": 1.1, + "learning_rate": 3.164849961528597e-05, + "loss": 0.4473, + "step": 28671 + }, + { + "epoch": 1.1, + "learning_rate": 3.164785842523724e-05, + "loss": 0.4241, + "step": 28672 + }, + { + "epoch": 1.1, + "learning_rate": 3.164721723518851e-05, + "loss": 0.3265, + "step": 28673 + }, + { + "epoch": 1.1, + "learning_rate": 3.164657604513978e-05, + "loss": 0.4827, + "step": 28674 + }, + { + "epoch": 1.1, + "learning_rate": 3.164593485509105e-05, + "loss": 0.3613, + "step": 28675 + }, + { + "epoch": 1.1, + "learning_rate": 3.164529366504232e-05, + "loss": 0.5107, + "step": 28676 + }, + { + "epoch": 1.1, + "learning_rate": 3.1644652474993594e-05, + "loss": 0.3424, + "step": 28677 + }, + { + "epoch": 1.1, + "learning_rate": 3.164401128494486e-05, + "loss": 0.4634, + "step": 28678 + }, + { + "epoch": 1.1, + "learning_rate": 3.164337009489613e-05, + "loss": 0.3572, + "step": 28679 + }, + { + "epoch": 1.1, + "learning_rate": 3.1642728904847396e-05, + "loss": 0.4553, + "step": 28680 + }, + { + "epoch": 1.1, + "learning_rate": 3.164208771479867e-05, + "loss": 0.3499, + "step": 28681 + }, + { + "epoch": 1.1, + "learning_rate": 3.164144652474993e-05, + "loss": 0.5213, + "step": 28682 + }, + { + "epoch": 1.1, + "learning_rate": 3.1640805334701204e-05, + "loss": 0.3528, + "step": 28683 + }, + { + "epoch": 1.1, + "learning_rate": 3.1640164144652476e-05, + "loss": 0.4068, + "step": 28684 + }, + { + "epoch": 1.1, + "learning_rate": 3.163952295460375e-05, + "loss": 0.3458, + "step": 28685 + }, + { + "epoch": 1.1, + "learning_rate": 3.163888176455502e-05, + "loss": 0.2591, + "step": 28686 + }, + { + "epoch": 1.1, + "learning_rate": 3.1638240574506284e-05, + "loss": 0.277, + "step": 28687 + }, + { + "epoch": 1.1, + "learning_rate": 3.1637599384457556e-05, + "loss": 0.2904, + "step": 28688 + }, + { + "epoch": 1.1, + "learning_rate": 3.163695819440882e-05, + "loss": 0.3384, + "step": 28689 + }, + { + "epoch": 1.1, + "learning_rate": 3.163631700436009e-05, + "loss": 0.3954, + "step": 28690 + }, + { + "epoch": 1.1, + "learning_rate": 3.163567581431136e-05, + "loss": 0.3421, + "step": 28691 + }, + { + "epoch": 1.1, + "learning_rate": 3.1635034624262636e-05, + "loss": 0.5933, + "step": 28692 + }, + { + "epoch": 1.1, + "learning_rate": 3.163439343421391e-05, + "loss": 0.5096, + "step": 28693 + }, + { + "epoch": 1.1, + "learning_rate": 3.163375224416517e-05, + "loss": 0.4207, + "step": 28694 + }, + { + "epoch": 1.1, + "learning_rate": 3.1633111054116444e-05, + "loss": 0.358, + "step": 28695 + }, + { + "epoch": 1.1, + "learning_rate": 3.163246986406771e-05, + "loss": 0.3904, + "step": 28696 + }, + { + "epoch": 1.1, + "learning_rate": 3.163182867401898e-05, + "loss": 0.3785, + "step": 28697 + }, + { + "epoch": 1.1, + "learning_rate": 3.1631187483970246e-05, + "loss": 0.4234, + "step": 28698 + }, + { + "epoch": 1.1, + "learning_rate": 3.163054629392152e-05, + "loss": 0.3678, + "step": 28699 + }, + { + "epoch": 1.1, + "learning_rate": 3.162990510387279e-05, + "loss": 0.3826, + "step": 28700 + }, + { + "epoch": 1.1, + "learning_rate": 3.162926391382406e-05, + "loss": 0.4103, + "step": 28701 + }, + { + "epoch": 1.1, + "learning_rate": 3.162862272377533e-05, + "loss": 0.5105, + "step": 28702 + }, + { + "epoch": 1.1, + "learning_rate": 3.16279815337266e-05, + "loss": 0.4136, + "step": 28703 + }, + { + "epoch": 1.1, + "learning_rate": 3.162734034367787e-05, + "loss": 0.5602, + "step": 28704 + }, + { + "epoch": 1.1, + "learning_rate": 3.1626699153629134e-05, + "loss": 0.408, + "step": 28705 + }, + { + "epoch": 1.1, + "learning_rate": 3.1626057963580406e-05, + "loss": 0.5668, + "step": 28706 + }, + { + "epoch": 1.1, + "learning_rate": 3.162541677353167e-05, + "loss": 0.4687, + "step": 28707 + }, + { + "epoch": 1.1, + "learning_rate": 3.162477558348295e-05, + "loss": 0.3817, + "step": 28708 + }, + { + "epoch": 1.1, + "learning_rate": 3.1624134393434214e-05, + "loss": 0.4123, + "step": 28709 + }, + { + "epoch": 1.1, + "learning_rate": 3.1623493203385486e-05, + "loss": 0.3559, + "step": 28710 + }, + { + "epoch": 1.1, + "learning_rate": 3.162285201333676e-05, + "loss": 0.3743, + "step": 28711 + }, + { + "epoch": 1.1, + "learning_rate": 3.162221082328802e-05, + "loss": 0.4133, + "step": 28712 + }, + { + "epoch": 1.1, + "learning_rate": 3.1621569633239294e-05, + "loss": 0.4542, + "step": 28713 + }, + { + "epoch": 1.1, + "learning_rate": 3.162092844319056e-05, + "loss": 0.3649, + "step": 28714 + }, + { + "epoch": 1.1, + "learning_rate": 3.162028725314183e-05, + "loss": 0.3566, + "step": 28715 + }, + { + "epoch": 1.1, + "learning_rate": 3.16196460630931e-05, + "loss": 0.4585, + "step": 28716 + }, + { + "epoch": 1.1, + "learning_rate": 3.1619004873044375e-05, + "loss": 0.4053, + "step": 28717 + }, + { + "epoch": 1.1, + "learning_rate": 3.161836368299564e-05, + "loss": 0.4587, + "step": 28718 + }, + { + "epoch": 1.1, + "learning_rate": 3.161772249294691e-05, + "loss": 0.4295, + "step": 28719 + }, + { + "epoch": 1.1, + "learning_rate": 3.161708130289818e-05, + "loss": 0.4222, + "step": 28720 + }, + { + "epoch": 1.1, + "learning_rate": 3.161644011284945e-05, + "loss": 0.3794, + "step": 28721 + }, + { + "epoch": 1.1, + "learning_rate": 3.161579892280072e-05, + "loss": 0.4556, + "step": 28722 + }, + { + "epoch": 1.1, + "learning_rate": 3.1615157732751985e-05, + "loss": 0.3304, + "step": 28723 + }, + { + "epoch": 1.1, + "learning_rate": 3.161451654270326e-05, + "loss": 0.3758, + "step": 28724 + }, + { + "epoch": 1.1, + "learning_rate": 3.161387535265453e-05, + "loss": 0.377, + "step": 28725 + }, + { + "epoch": 1.1, + "learning_rate": 3.16132341626058e-05, + "loss": 0.5007, + "step": 28726 + }, + { + "epoch": 1.1, + "learning_rate": 3.161259297255707e-05, + "loss": 0.3665, + "step": 28727 + }, + { + "epoch": 1.1, + "learning_rate": 3.1611951782508336e-05, + "loss": 0.3453, + "step": 28728 + }, + { + "epoch": 1.1, + "learning_rate": 3.161131059245961e-05, + "loss": 0.3956, + "step": 28729 + }, + { + "epoch": 1.1, + "learning_rate": 3.161066940241087e-05, + "loss": 0.5102, + "step": 28730 + }, + { + "epoch": 1.1, + "learning_rate": 3.1610028212362145e-05, + "loss": 0.4972, + "step": 28731 + }, + { + "epoch": 1.1, + "learning_rate": 3.1609387022313416e-05, + "loss": 0.5004, + "step": 28732 + }, + { + "epoch": 1.1, + "learning_rate": 3.160874583226469e-05, + "loss": 0.4807, + "step": 28733 + }, + { + "epoch": 1.1, + "learning_rate": 3.160810464221595e-05, + "loss": 0.4181, + "step": 28734 + }, + { + "epoch": 1.1, + "learning_rate": 3.1607463452167225e-05, + "loss": 0.3694, + "step": 28735 + }, + { + "epoch": 1.1, + "learning_rate": 3.1606822262118497e-05, + "loss": 0.4677, + "step": 28736 + }, + { + "epoch": 1.1, + "learning_rate": 3.160618107206976e-05, + "loss": 0.4258, + "step": 28737 + }, + { + "epoch": 1.1, + "learning_rate": 3.160553988202103e-05, + "loss": 0.4833, + "step": 28738 + }, + { + "epoch": 1.1, + "learning_rate": 3.16048986919723e-05, + "loss": 0.4516, + "step": 28739 + }, + { + "epoch": 1.1, + "learning_rate": 3.160425750192357e-05, + "loss": 0.2937, + "step": 28740 + }, + { + "epoch": 1.1, + "learning_rate": 3.160361631187484e-05, + "loss": 0.4992, + "step": 28741 + }, + { + "epoch": 1.11, + "learning_rate": 3.160297512182611e-05, + "loss": 0.259, + "step": 28742 + }, + { + "epoch": 1.11, + "learning_rate": 3.160233393177738e-05, + "loss": 0.4544, + "step": 28743 + }, + { + "epoch": 1.11, + "learning_rate": 3.160169274172865e-05, + "loss": 0.4471, + "step": 28744 + }, + { + "epoch": 1.11, + "learning_rate": 3.160105155167992e-05, + "loss": 0.4954, + "step": 28745 + }, + { + "epoch": 1.11, + "learning_rate": 3.1600410361631187e-05, + "loss": 0.3923, + "step": 28746 + }, + { + "epoch": 1.11, + "learning_rate": 3.159976917158246e-05, + "loss": 0.2701, + "step": 28747 + }, + { + "epoch": 1.11, + "learning_rate": 3.159912798153373e-05, + "loss": 0.3911, + "step": 28748 + }, + { + "epoch": 1.11, + "learning_rate": 3.1598486791485e-05, + "loss": 0.3088, + "step": 28749 + }, + { + "epoch": 1.11, + "learning_rate": 3.159784560143627e-05, + "loss": 0.4484, + "step": 28750 + }, + { + "epoch": 1.11, + "learning_rate": 3.159720441138754e-05, + "loss": 0.5024, + "step": 28751 + }, + { + "epoch": 1.11, + "learning_rate": 3.15965632213388e-05, + "loss": 0.2924, + "step": 28752 + }, + { + "epoch": 1.11, + "learning_rate": 3.1595922031290075e-05, + "loss": 0.4351, + "step": 28753 + }, + { + "epoch": 1.11, + "learning_rate": 3.159528084124135e-05, + "loss": 0.5672, + "step": 28754 + }, + { + "epoch": 1.11, + "learning_rate": 3.159463965119261e-05, + "loss": 0.3579, + "step": 28755 + }, + { + "epoch": 1.11, + "learning_rate": 3.1593998461143883e-05, + "loss": 0.261, + "step": 28756 + }, + { + "epoch": 1.11, + "learning_rate": 3.1593357271095155e-05, + "loss": 0.3723, + "step": 28757 + }, + { + "epoch": 1.11, + "learning_rate": 3.159271608104643e-05, + "loss": 0.4151, + "step": 28758 + }, + { + "epoch": 1.11, + "learning_rate": 3.159207489099769e-05, + "loss": 0.3494, + "step": 28759 + }, + { + "epoch": 1.11, + "learning_rate": 3.1591433700948964e-05, + "loss": 0.4532, + "step": 28760 + }, + { + "epoch": 1.11, + "learning_rate": 3.159079251090023e-05, + "loss": 0.3351, + "step": 28761 + }, + { + "epoch": 1.11, + "learning_rate": 3.15901513208515e-05, + "loss": 0.5369, + "step": 28762 + }, + { + "epoch": 1.11, + "learning_rate": 3.158951013080277e-05, + "loss": 0.272, + "step": 28763 + }, + { + "epoch": 1.11, + "learning_rate": 3.1588868940754044e-05, + "loss": 0.3679, + "step": 28764 + }, + { + "epoch": 1.11, + "learning_rate": 3.1588227750705315e-05, + "loss": 0.3774, + "step": 28765 + }, + { + "epoch": 1.11, + "learning_rate": 3.158758656065658e-05, + "loss": 0.5221, + "step": 28766 + }, + { + "epoch": 1.11, + "learning_rate": 3.158694537060785e-05, + "loss": 0.4238, + "step": 28767 + }, + { + "epoch": 1.11, + "learning_rate": 3.158630418055912e-05, + "loss": 0.4509, + "step": 28768 + }, + { + "epoch": 1.11, + "learning_rate": 3.158566299051039e-05, + "loss": 0.383, + "step": 28769 + }, + { + "epoch": 1.11, + "learning_rate": 3.1585021800461654e-05, + "loss": 0.4844, + "step": 28770 + }, + { + "epoch": 1.11, + "learning_rate": 3.1584380610412925e-05, + "loss": 0.4797, + "step": 28771 + }, + { + "epoch": 1.11, + "learning_rate": 3.15837394203642e-05, + "loss": 0.4647, + "step": 28772 + }, + { + "epoch": 1.11, + "learning_rate": 3.158309823031547e-05, + "loss": 0.349, + "step": 28773 + }, + { + "epoch": 1.11, + "learning_rate": 3.158245704026674e-05, + "loss": 0.3595, + "step": 28774 + }, + { + "epoch": 1.11, + "learning_rate": 3.1581815850218005e-05, + "loss": 0.2814, + "step": 28775 + }, + { + "epoch": 1.11, + "learning_rate": 3.158117466016928e-05, + "loss": 0.5341, + "step": 28776 + }, + { + "epoch": 1.11, + "learning_rate": 3.158053347012054e-05, + "loss": 0.5212, + "step": 28777 + }, + { + "epoch": 1.11, + "learning_rate": 3.1579892280071814e-05, + "loss": 0.4065, + "step": 28778 + }, + { + "epoch": 1.11, + "learning_rate": 3.157925109002308e-05, + "loss": 0.4538, + "step": 28779 + }, + { + "epoch": 1.11, + "learning_rate": 3.157860989997436e-05, + "loss": 0.4724, + "step": 28780 + }, + { + "epoch": 1.11, + "learning_rate": 3.157796870992563e-05, + "loss": 0.424, + "step": 28781 + }, + { + "epoch": 1.11, + "learning_rate": 3.1577327519876894e-05, + "loss": 0.2997, + "step": 28782 + }, + { + "epoch": 1.11, + "learning_rate": 3.1576686329828166e-05, + "loss": 0.4573, + "step": 28783 + }, + { + "epoch": 1.11, + "learning_rate": 3.157604513977943e-05, + "loss": 0.3702, + "step": 28784 + }, + { + "epoch": 1.11, + "learning_rate": 3.15754039497307e-05, + "loss": 0.4933, + "step": 28785 + }, + { + "epoch": 1.11, + "learning_rate": 3.157476275968197e-05, + "loss": 0.3825, + "step": 28786 + }, + { + "epoch": 1.11, + "learning_rate": 3.157412156963324e-05, + "loss": 0.5429, + "step": 28787 + }, + { + "epoch": 1.11, + "learning_rate": 3.157348037958451e-05, + "loss": 0.4375, + "step": 28788 + }, + { + "epoch": 1.11, + "learning_rate": 3.157283918953578e-05, + "loss": 0.3469, + "step": 28789 + }, + { + "epoch": 1.11, + "learning_rate": 3.1572197999487054e-05, + "loss": 0.487, + "step": 28790 + }, + { + "epoch": 1.11, + "learning_rate": 3.157155680943832e-05, + "loss": 0.4553, + "step": 28791 + }, + { + "epoch": 1.11, + "learning_rate": 3.157091561938959e-05, + "loss": 0.327, + "step": 28792 + }, + { + "epoch": 1.11, + "learning_rate": 3.1570274429340856e-05, + "loss": 0.3633, + "step": 28793 + }, + { + "epoch": 1.11, + "learning_rate": 3.156963323929213e-05, + "loss": 0.3214, + "step": 28794 + }, + { + "epoch": 1.11, + "learning_rate": 3.156899204924339e-05, + "loss": 0.3673, + "step": 28795 + }, + { + "epoch": 1.11, + "learning_rate": 3.156835085919467e-05, + "loss": 0.2777, + "step": 28796 + }, + { + "epoch": 1.11, + "learning_rate": 3.1567709669145936e-05, + "loss": 0.4171, + "step": 28797 + }, + { + "epoch": 1.11, + "learning_rate": 3.156706847909721e-05, + "loss": 0.3917, + "step": 28798 + }, + { + "epoch": 1.11, + "learning_rate": 3.156642728904848e-05, + "loss": 0.3769, + "step": 28799 + }, + { + "epoch": 1.11, + "learning_rate": 3.1565786098999744e-05, + "loss": 0.4024, + "step": 28800 + }, + { + "epoch": 1.11, + "learning_rate": 3.1565144908951016e-05, + "loss": 0.5831, + "step": 28801 + }, + { + "epoch": 1.11, + "learning_rate": 3.156450371890228e-05, + "loss": 0.311, + "step": 28802 + }, + { + "epoch": 1.11, + "learning_rate": 3.156386252885355e-05, + "loss": 0.501, + "step": 28803 + }, + { + "epoch": 1.11, + "learning_rate": 3.1563221338804824e-05, + "loss": 0.534, + "step": 28804 + }, + { + "epoch": 1.11, + "learning_rate": 3.1562580148756096e-05, + "loss": 0.4663, + "step": 28805 + }, + { + "epoch": 1.11, + "learning_rate": 3.156193895870736e-05, + "loss": 0.4173, + "step": 28806 + }, + { + "epoch": 1.11, + "learning_rate": 3.156129776865863e-05, + "loss": 0.3985, + "step": 28807 + }, + { + "epoch": 1.11, + "learning_rate": 3.1560656578609904e-05, + "loss": 0.3559, + "step": 28808 + }, + { + "epoch": 1.11, + "learning_rate": 3.156001538856117e-05, + "loss": 0.422, + "step": 28809 + }, + { + "epoch": 1.11, + "learning_rate": 3.155937419851244e-05, + "loss": 0.354, + "step": 28810 + }, + { + "epoch": 1.11, + "learning_rate": 3.1558733008463706e-05, + "loss": 0.5461, + "step": 28811 + }, + { + "epoch": 1.11, + "learning_rate": 3.155809181841498e-05, + "loss": 0.2758, + "step": 28812 + }, + { + "epoch": 1.11, + "learning_rate": 3.155745062836625e-05, + "loss": 0.3977, + "step": 28813 + }, + { + "epoch": 1.11, + "learning_rate": 3.155680943831752e-05, + "loss": 0.4484, + "step": 28814 + }, + { + "epoch": 1.11, + "learning_rate": 3.1556168248268786e-05, + "loss": 0.5216, + "step": 28815 + }, + { + "epoch": 1.11, + "learning_rate": 3.155552705822006e-05, + "loss": 0.4682, + "step": 28816 + }, + { + "epoch": 1.11, + "learning_rate": 3.155488586817133e-05, + "loss": 0.5356, + "step": 28817 + }, + { + "epoch": 1.11, + "learning_rate": 3.1554244678122594e-05, + "loss": 0.2339, + "step": 28818 + }, + { + "epoch": 1.11, + "learning_rate": 3.1553603488073866e-05, + "loss": 0.4868, + "step": 28819 + }, + { + "epoch": 1.11, + "learning_rate": 3.155296229802514e-05, + "loss": 0.3956, + "step": 28820 + }, + { + "epoch": 1.11, + "learning_rate": 3.155232110797641e-05, + "loss": 0.4636, + "step": 28821 + }, + { + "epoch": 1.11, + "learning_rate": 3.1551679917927674e-05, + "loss": 0.4558, + "step": 28822 + }, + { + "epoch": 1.11, + "learning_rate": 3.1551038727878946e-05, + "loss": 0.4022, + "step": 28823 + }, + { + "epoch": 1.11, + "learning_rate": 3.155039753783021e-05, + "loss": 0.3723, + "step": 28824 + }, + { + "epoch": 1.11, + "learning_rate": 3.154975634778148e-05, + "loss": 0.3621, + "step": 28825 + }, + { + "epoch": 1.11, + "learning_rate": 3.1549115157732755e-05, + "loss": 0.4383, + "step": 28826 + }, + { + "epoch": 1.11, + "learning_rate": 3.154847396768402e-05, + "loss": 0.4391, + "step": 28827 + }, + { + "epoch": 1.11, + "learning_rate": 3.154783277763529e-05, + "loss": 0.5759, + "step": 28828 + }, + { + "epoch": 1.11, + "learning_rate": 3.154719158758656e-05, + "loss": 0.3202, + "step": 28829 + }, + { + "epoch": 1.11, + "learning_rate": 3.1546550397537835e-05, + "loss": 0.5164, + "step": 28830 + }, + { + "epoch": 1.11, + "learning_rate": 3.15459092074891e-05, + "loss": 0.4112, + "step": 28831 + }, + { + "epoch": 1.11, + "learning_rate": 3.154526801744037e-05, + "loss": 0.4902, + "step": 28832 + }, + { + "epoch": 1.11, + "learning_rate": 3.1544626827391636e-05, + "loss": 0.3894, + "step": 28833 + }, + { + "epoch": 1.11, + "learning_rate": 3.154398563734291e-05, + "loss": 0.425, + "step": 28834 + }, + { + "epoch": 1.11, + "learning_rate": 3.154334444729418e-05, + "loss": 0.4611, + "step": 28835 + }, + { + "epoch": 1.11, + "learning_rate": 3.154270325724545e-05, + "loss": 0.5233, + "step": 28836 + }, + { + "epoch": 1.11, + "learning_rate": 3.154206206719672e-05, + "loss": 0.3089, + "step": 28837 + }, + { + "epoch": 1.11, + "learning_rate": 3.154142087714799e-05, + "loss": 0.3363, + "step": 28838 + }, + { + "epoch": 1.11, + "learning_rate": 3.154077968709926e-05, + "loss": 0.5209, + "step": 28839 + }, + { + "epoch": 1.11, + "learning_rate": 3.1540138497050525e-05, + "loss": 0.2776, + "step": 28840 + }, + { + "epoch": 1.11, + "learning_rate": 3.1539497307001796e-05, + "loss": 0.2686, + "step": 28841 + }, + { + "epoch": 1.11, + "learning_rate": 3.153885611695306e-05, + "loss": 0.4311, + "step": 28842 + }, + { + "epoch": 1.11, + "learning_rate": 3.153821492690433e-05, + "loss": 0.3936, + "step": 28843 + }, + { + "epoch": 1.11, + "learning_rate": 3.1537573736855605e-05, + "loss": 0.2921, + "step": 28844 + }, + { + "epoch": 1.11, + "learning_rate": 3.1536932546806876e-05, + "loss": 0.3065, + "step": 28845 + }, + { + "epoch": 1.11, + "learning_rate": 3.153629135675815e-05, + "loss": 0.4865, + "step": 28846 + }, + { + "epoch": 1.11, + "learning_rate": 3.153565016670941e-05, + "loss": 0.2963, + "step": 28847 + }, + { + "epoch": 1.11, + "learning_rate": 3.1535008976660685e-05, + "loss": 0.4614, + "step": 28848 + }, + { + "epoch": 1.11, + "learning_rate": 3.153436778661195e-05, + "loss": 0.3829, + "step": 28849 + }, + { + "epoch": 1.11, + "learning_rate": 3.153372659656322e-05, + "loss": 0.3472, + "step": 28850 + }, + { + "epoch": 1.11, + "learning_rate": 3.153308540651449e-05, + "loss": 0.3579, + "step": 28851 + }, + { + "epoch": 1.11, + "learning_rate": 3.1532444216465765e-05, + "loss": 0.3314, + "step": 28852 + }, + { + "epoch": 1.11, + "learning_rate": 3.153180302641704e-05, + "loss": 0.3467, + "step": 28853 + }, + { + "epoch": 1.11, + "learning_rate": 3.15311618363683e-05, + "loss": 0.4232, + "step": 28854 + }, + { + "epoch": 1.11, + "learning_rate": 3.153052064631957e-05, + "loss": 0.2989, + "step": 28855 + }, + { + "epoch": 1.11, + "learning_rate": 3.152987945627084e-05, + "loss": 0.4681, + "step": 28856 + }, + { + "epoch": 1.11, + "learning_rate": 3.152923826622211e-05, + "loss": 0.3751, + "step": 28857 + }, + { + "epoch": 1.11, + "learning_rate": 3.1528597076173375e-05, + "loss": 0.4554, + "step": 28858 + }, + { + "epoch": 1.11, + "learning_rate": 3.152795588612465e-05, + "loss": 0.3922, + "step": 28859 + }, + { + "epoch": 1.11, + "learning_rate": 3.152731469607592e-05, + "loss": 0.4566, + "step": 28860 + }, + { + "epoch": 1.11, + "learning_rate": 3.152667350602719e-05, + "loss": 0.4713, + "step": 28861 + }, + { + "epoch": 1.11, + "learning_rate": 3.152603231597846e-05, + "loss": 0.4192, + "step": 28862 + }, + { + "epoch": 1.11, + "learning_rate": 3.152539112592973e-05, + "loss": 0.5911, + "step": 28863 + }, + { + "epoch": 1.11, + "learning_rate": 3.1524749935881e-05, + "loss": 0.515, + "step": 28864 + }, + { + "epoch": 1.11, + "learning_rate": 3.152410874583226e-05, + "loss": 0.4731, + "step": 28865 + }, + { + "epoch": 1.11, + "learning_rate": 3.1523467555783535e-05, + "loss": 0.3187, + "step": 28866 + }, + { + "epoch": 1.11, + "learning_rate": 3.15228263657348e-05, + "loss": 0.3442, + "step": 28867 + }, + { + "epoch": 1.11, + "learning_rate": 3.152218517568608e-05, + "loss": 0.3447, + "step": 28868 + }, + { + "epoch": 1.11, + "learning_rate": 3.1521543985637343e-05, + "loss": 0.4511, + "step": 28869 + }, + { + "epoch": 1.11, + "learning_rate": 3.1520902795588615e-05, + "loss": 0.4505, + "step": 28870 + }, + { + "epoch": 1.11, + "learning_rate": 3.152026160553989e-05, + "loss": 0.5562, + "step": 28871 + }, + { + "epoch": 1.11, + "learning_rate": 3.151962041549115e-05, + "loss": 0.3919, + "step": 28872 + }, + { + "epoch": 1.11, + "learning_rate": 3.1518979225442424e-05, + "loss": 0.4396, + "step": 28873 + }, + { + "epoch": 1.11, + "learning_rate": 3.151833803539369e-05, + "loss": 0.4084, + "step": 28874 + }, + { + "epoch": 1.11, + "learning_rate": 3.151769684534496e-05, + "loss": 0.5303, + "step": 28875 + }, + { + "epoch": 1.11, + "learning_rate": 3.151705565529623e-05, + "loss": 0.3926, + "step": 28876 + }, + { + "epoch": 1.11, + "learning_rate": 3.1516414465247504e-05, + "loss": 0.4889, + "step": 28877 + }, + { + "epoch": 1.11, + "learning_rate": 3.151577327519877e-05, + "loss": 0.4291, + "step": 28878 + }, + { + "epoch": 1.11, + "learning_rate": 3.151513208515004e-05, + "loss": 0.3225, + "step": 28879 + }, + { + "epoch": 1.11, + "learning_rate": 3.151449089510131e-05, + "loss": 0.3862, + "step": 28880 + }, + { + "epoch": 1.11, + "learning_rate": 3.151384970505258e-05, + "loss": 0.3182, + "step": 28881 + }, + { + "epoch": 1.11, + "learning_rate": 3.151320851500385e-05, + "loss": 0.3787, + "step": 28882 + }, + { + "epoch": 1.11, + "learning_rate": 3.1512567324955114e-05, + "loss": 0.4858, + "step": 28883 + }, + { + "epoch": 1.11, + "learning_rate": 3.151192613490639e-05, + "loss": 0.4133, + "step": 28884 + }, + { + "epoch": 1.11, + "learning_rate": 3.151128494485766e-05, + "loss": 0.3235, + "step": 28885 + }, + { + "epoch": 1.11, + "learning_rate": 3.151064375480893e-05, + "loss": 0.4465, + "step": 28886 + }, + { + "epoch": 1.11, + "learning_rate": 3.1510002564760194e-05, + "loss": 0.4351, + "step": 28887 + }, + { + "epoch": 1.11, + "learning_rate": 3.1509361374711465e-05, + "loss": 0.4862, + "step": 28888 + }, + { + "epoch": 1.11, + "learning_rate": 3.150872018466274e-05, + "loss": 0.3643, + "step": 28889 + }, + { + "epoch": 1.11, + "learning_rate": 3.1508078994614e-05, + "loss": 0.3037, + "step": 28890 + }, + { + "epoch": 1.11, + "learning_rate": 3.1507437804565274e-05, + "loss": 0.3092, + "step": 28891 + }, + { + "epoch": 1.11, + "learning_rate": 3.1506796614516546e-05, + "loss": 0.4736, + "step": 28892 + }, + { + "epoch": 1.11, + "learning_rate": 3.150615542446782e-05, + "loss": 0.3982, + "step": 28893 + }, + { + "epoch": 1.11, + "learning_rate": 3.150551423441908e-05, + "loss": 0.3637, + "step": 28894 + }, + { + "epoch": 1.11, + "learning_rate": 3.1504873044370354e-05, + "loss": 0.259, + "step": 28895 + }, + { + "epoch": 1.11, + "learning_rate": 3.150423185432162e-05, + "loss": 0.4628, + "step": 28896 + }, + { + "epoch": 1.11, + "learning_rate": 3.150359066427289e-05, + "loss": 0.3757, + "step": 28897 + }, + { + "epoch": 1.11, + "learning_rate": 3.150294947422416e-05, + "loss": 0.3542, + "step": 28898 + }, + { + "epoch": 1.11, + "learning_rate": 3.150230828417543e-05, + "loss": 0.6187, + "step": 28899 + }, + { + "epoch": 1.11, + "learning_rate": 3.15016670941267e-05, + "loss": 0.4119, + "step": 28900 + }, + { + "epoch": 1.11, + "learning_rate": 3.150102590407797e-05, + "loss": 0.3934, + "step": 28901 + }, + { + "epoch": 1.11, + "learning_rate": 3.150038471402924e-05, + "loss": 0.468, + "step": 28902 + }, + { + "epoch": 1.11, + "learning_rate": 3.149974352398051e-05, + "loss": 0.4642, + "step": 28903 + }, + { + "epoch": 1.11, + "learning_rate": 3.149910233393178e-05, + "loss": 0.3378, + "step": 28904 + }, + { + "epoch": 1.11, + "learning_rate": 3.1498461143883044e-05, + "loss": 0.4961, + "step": 28905 + }, + { + "epoch": 1.11, + "learning_rate": 3.1497819953834316e-05, + "loss": 0.4811, + "step": 28906 + }, + { + "epoch": 1.11, + "learning_rate": 3.149717876378559e-05, + "loss": 0.4389, + "step": 28907 + }, + { + "epoch": 1.11, + "learning_rate": 3.149653757373686e-05, + "loss": 0.3913, + "step": 28908 + }, + { + "epoch": 1.11, + "learning_rate": 3.149589638368813e-05, + "loss": 0.4352, + "step": 28909 + }, + { + "epoch": 1.11, + "learning_rate": 3.1495255193639396e-05, + "loss": 0.5323, + "step": 28910 + }, + { + "epoch": 1.11, + "learning_rate": 3.149461400359067e-05, + "loss": 0.3699, + "step": 28911 + }, + { + "epoch": 1.11, + "learning_rate": 3.149397281354193e-05, + "loss": 0.4539, + "step": 28912 + }, + { + "epoch": 1.11, + "learning_rate": 3.1493331623493204e-05, + "loss": 0.3659, + "step": 28913 + }, + { + "epoch": 1.11, + "learning_rate": 3.1492690433444476e-05, + "loss": 0.3471, + "step": 28914 + }, + { + "epoch": 1.11, + "learning_rate": 3.149204924339574e-05, + "loss": 0.3632, + "step": 28915 + }, + { + "epoch": 1.11, + "learning_rate": 3.149140805334701e-05, + "loss": 0.3842, + "step": 28916 + }, + { + "epoch": 1.11, + "learning_rate": 3.1490766863298284e-05, + "loss": 0.4371, + "step": 28917 + }, + { + "epoch": 1.11, + "learning_rate": 3.1490125673249556e-05, + "loss": 0.4647, + "step": 28918 + }, + { + "epoch": 1.11, + "learning_rate": 3.148948448320082e-05, + "loss": 0.2632, + "step": 28919 + }, + { + "epoch": 1.11, + "learning_rate": 3.148884329315209e-05, + "loss": 0.6442, + "step": 28920 + }, + { + "epoch": 1.11, + "learning_rate": 3.148820210310336e-05, + "loss": 0.4599, + "step": 28921 + }, + { + "epoch": 1.11, + "learning_rate": 3.148756091305463e-05, + "loss": 0.3765, + "step": 28922 + }, + { + "epoch": 1.11, + "learning_rate": 3.14869197230059e-05, + "loss": 0.5454, + "step": 28923 + }, + { + "epoch": 1.11, + "learning_rate": 3.148627853295717e-05, + "loss": 0.2867, + "step": 28924 + }, + { + "epoch": 1.11, + "learning_rate": 3.1485637342908444e-05, + "loss": 0.4219, + "step": 28925 + }, + { + "epoch": 1.11, + "learning_rate": 3.148499615285971e-05, + "loss": 0.4496, + "step": 28926 + }, + { + "epoch": 1.11, + "learning_rate": 3.148435496281098e-05, + "loss": 0.4863, + "step": 28927 + }, + { + "epoch": 1.11, + "learning_rate": 3.1483713772762246e-05, + "loss": 0.2961, + "step": 28928 + }, + { + "epoch": 1.11, + "learning_rate": 3.148307258271352e-05, + "loss": 0.5041, + "step": 28929 + }, + { + "epoch": 1.11, + "learning_rate": 3.148243139266478e-05, + "loss": 0.3887, + "step": 28930 + }, + { + "epoch": 1.11, + "learning_rate": 3.1481790202616054e-05, + "loss": 0.4164, + "step": 28931 + }, + { + "epoch": 1.11, + "learning_rate": 3.1481149012567326e-05, + "loss": 0.318, + "step": 28932 + }, + { + "epoch": 1.11, + "learning_rate": 3.14805078225186e-05, + "loss": 0.3872, + "step": 28933 + }, + { + "epoch": 1.11, + "learning_rate": 3.147986663246987e-05, + "loss": 0.3518, + "step": 28934 + }, + { + "epoch": 1.11, + "learning_rate": 3.1479225442421134e-05, + "loss": 0.3871, + "step": 28935 + }, + { + "epoch": 1.11, + "learning_rate": 3.1478584252372406e-05, + "loss": 0.5366, + "step": 28936 + }, + { + "epoch": 1.11, + "learning_rate": 3.147794306232367e-05, + "loss": 0.3358, + "step": 28937 + }, + { + "epoch": 1.11, + "learning_rate": 3.147730187227494e-05, + "loss": 0.4236, + "step": 28938 + }, + { + "epoch": 1.11, + "learning_rate": 3.147666068222621e-05, + "loss": 0.3859, + "step": 28939 + }, + { + "epoch": 1.11, + "learning_rate": 3.1476019492177486e-05, + "loss": 0.4692, + "step": 28940 + }, + { + "epoch": 1.11, + "learning_rate": 3.147537830212876e-05, + "loss": 0.4392, + "step": 28941 + }, + { + "epoch": 1.11, + "learning_rate": 3.147473711208002e-05, + "loss": 0.3318, + "step": 28942 + }, + { + "epoch": 1.11, + "learning_rate": 3.1474095922031295e-05, + "loss": 0.3087, + "step": 28943 + }, + { + "epoch": 1.11, + "learning_rate": 3.147345473198256e-05, + "loss": 0.3387, + "step": 28944 + }, + { + "epoch": 1.11, + "learning_rate": 3.147281354193383e-05, + "loss": 0.4276, + "step": 28945 + }, + { + "epoch": 1.11, + "learning_rate": 3.1472172351885096e-05, + "loss": 0.4433, + "step": 28946 + }, + { + "epoch": 1.11, + "learning_rate": 3.147153116183637e-05, + "loss": 0.3498, + "step": 28947 + }, + { + "epoch": 1.11, + "learning_rate": 3.147088997178764e-05, + "loss": 0.4735, + "step": 28948 + }, + { + "epoch": 1.11, + "learning_rate": 3.147024878173891e-05, + "loss": 0.4358, + "step": 28949 + }, + { + "epoch": 1.11, + "learning_rate": 3.146960759169018e-05, + "loss": 0.369, + "step": 28950 + }, + { + "epoch": 1.11, + "learning_rate": 3.146896640164145e-05, + "loss": 0.3308, + "step": 28951 + }, + { + "epoch": 1.11, + "learning_rate": 3.146832521159272e-05, + "loss": 0.3427, + "step": 28952 + }, + { + "epoch": 1.11, + "learning_rate": 3.1467684021543985e-05, + "loss": 0.2846, + "step": 28953 + }, + { + "epoch": 1.11, + "learning_rate": 3.1467042831495256e-05, + "loss": 0.3182, + "step": 28954 + }, + { + "epoch": 1.11, + "learning_rate": 3.146640164144652e-05, + "loss": 0.3555, + "step": 28955 + }, + { + "epoch": 1.11, + "learning_rate": 3.14657604513978e-05, + "loss": 0.4679, + "step": 28956 + }, + { + "epoch": 1.11, + "learning_rate": 3.1465119261349065e-05, + "loss": 0.3742, + "step": 28957 + }, + { + "epoch": 1.11, + "learning_rate": 3.1464478071300337e-05, + "loss": 0.3143, + "step": 28958 + }, + { + "epoch": 1.11, + "learning_rate": 3.146383688125161e-05, + "loss": 0.2926, + "step": 28959 + }, + { + "epoch": 1.11, + "learning_rate": 3.146319569120287e-05, + "loss": 0.3884, + "step": 28960 + }, + { + "epoch": 1.11, + "learning_rate": 3.1462554501154145e-05, + "loss": 0.4957, + "step": 28961 + }, + { + "epoch": 1.11, + "learning_rate": 3.146191331110541e-05, + "loss": 0.6164, + "step": 28962 + }, + { + "epoch": 1.11, + "learning_rate": 3.146127212105668e-05, + "loss": 0.3521, + "step": 28963 + }, + { + "epoch": 1.11, + "learning_rate": 3.146063093100795e-05, + "loss": 0.347, + "step": 28964 + }, + { + "epoch": 1.11, + "learning_rate": 3.1459989740959225e-05, + "loss": 0.4726, + "step": 28965 + }, + { + "epoch": 1.11, + "learning_rate": 3.145934855091049e-05, + "loss": 0.4593, + "step": 28966 + }, + { + "epoch": 1.11, + "learning_rate": 3.145870736086176e-05, + "loss": 0.4047, + "step": 28967 + }, + { + "epoch": 1.11, + "learning_rate": 3.145806617081303e-05, + "loss": 0.4348, + "step": 28968 + }, + { + "epoch": 1.11, + "learning_rate": 3.14574249807643e-05, + "loss": 0.587, + "step": 28969 + }, + { + "epoch": 1.11, + "learning_rate": 3.145678379071557e-05, + "loss": 0.3218, + "step": 28970 + }, + { + "epoch": 1.11, + "learning_rate": 3.1456142600666835e-05, + "loss": 0.2559, + "step": 28971 + }, + { + "epoch": 1.11, + "learning_rate": 3.1455501410618113e-05, + "loss": 0.4108, + "step": 28972 + }, + { + "epoch": 1.11, + "learning_rate": 3.145486022056938e-05, + "loss": 0.4598, + "step": 28973 + }, + { + "epoch": 1.11, + "learning_rate": 3.145421903052065e-05, + "loss": 0.3881, + "step": 28974 + }, + { + "epoch": 1.11, + "learning_rate": 3.1453577840471915e-05, + "loss": 0.4751, + "step": 28975 + }, + { + "epoch": 1.11, + "learning_rate": 3.145293665042319e-05, + "loss": 0.4124, + "step": 28976 + }, + { + "epoch": 1.11, + "learning_rate": 3.145229546037446e-05, + "loss": 0.4437, + "step": 28977 + }, + { + "epoch": 1.11, + "learning_rate": 3.1451654270325723e-05, + "loss": 0.436, + "step": 28978 + }, + { + "epoch": 1.11, + "learning_rate": 3.1451013080276995e-05, + "loss": 0.4304, + "step": 28979 + }, + { + "epoch": 1.11, + "learning_rate": 3.145037189022827e-05, + "loss": 0.4941, + "step": 28980 + }, + { + "epoch": 1.11, + "learning_rate": 3.144973070017954e-05, + "loss": 0.429, + "step": 28981 + }, + { + "epoch": 1.11, + "learning_rate": 3.1449089510130803e-05, + "loss": 0.4804, + "step": 28982 + }, + { + "epoch": 1.11, + "learning_rate": 3.1448448320082075e-05, + "loss": 0.525, + "step": 28983 + }, + { + "epoch": 1.11, + "learning_rate": 3.144780713003334e-05, + "loss": 0.399, + "step": 28984 + }, + { + "epoch": 1.11, + "learning_rate": 3.144716593998461e-05, + "loss": 0.2845, + "step": 28985 + }, + { + "epoch": 1.11, + "learning_rate": 3.1446524749935884e-05, + "loss": 0.3492, + "step": 28986 + }, + { + "epoch": 1.11, + "learning_rate": 3.144588355988715e-05, + "loss": 0.3874, + "step": 28987 + }, + { + "epoch": 1.11, + "learning_rate": 3.144524236983842e-05, + "loss": 0.3847, + "step": 28988 + }, + { + "epoch": 1.11, + "learning_rate": 3.144460117978969e-05, + "loss": 0.434, + "step": 28989 + }, + { + "epoch": 1.11, + "learning_rate": 3.1443959989740964e-05, + "loss": 0.3414, + "step": 28990 + }, + { + "epoch": 1.11, + "learning_rate": 3.144331879969223e-05, + "loss": 0.4679, + "step": 28991 + }, + { + "epoch": 1.11, + "learning_rate": 3.14426776096435e-05, + "loss": 0.3538, + "step": 28992 + }, + { + "epoch": 1.11, + "learning_rate": 3.1442036419594765e-05, + "loss": 0.4049, + "step": 28993 + }, + { + "epoch": 1.11, + "learning_rate": 3.144139522954604e-05, + "loss": 0.4361, + "step": 28994 + }, + { + "epoch": 1.11, + "learning_rate": 3.144075403949731e-05, + "loss": 0.3962, + "step": 28995 + }, + { + "epoch": 1.11, + "learning_rate": 3.144011284944858e-05, + "loss": 0.3807, + "step": 28996 + }, + { + "epoch": 1.11, + "learning_rate": 3.143947165939985e-05, + "loss": 0.4762, + "step": 28997 + }, + { + "epoch": 1.11, + "learning_rate": 3.143883046935112e-05, + "loss": 0.3015, + "step": 28998 + }, + { + "epoch": 1.11, + "learning_rate": 3.143818927930239e-05, + "loss": 0.3264, + "step": 28999 + }, + { + "epoch": 1.11, + "learning_rate": 3.1437548089253654e-05, + "loss": 0.4564, + "step": 29000 + }, + { + "epoch": 1.11, + "learning_rate": 3.1436906899204925e-05, + "loss": 0.4988, + "step": 29001 + }, + { + "epoch": 1.12, + "learning_rate": 3.143626570915619e-05, + "loss": 0.4654, + "step": 29002 + }, + { + "epoch": 1.12, + "learning_rate": 3.143562451910746e-05, + "loss": 0.376, + "step": 29003 + }, + { + "epoch": 1.12, + "learning_rate": 3.1434983329058734e-05, + "loss": 0.465, + "step": 29004 + }, + { + "epoch": 1.12, + "learning_rate": 3.1434342139010006e-05, + "loss": 0.443, + "step": 29005 + }, + { + "epoch": 1.12, + "learning_rate": 3.143370094896128e-05, + "loss": 0.4467, + "step": 29006 + }, + { + "epoch": 1.12, + "learning_rate": 3.143305975891254e-05, + "loss": 0.3161, + "step": 29007 + }, + { + "epoch": 1.12, + "learning_rate": 3.1432418568863814e-05, + "loss": 0.4144, + "step": 29008 + }, + { + "epoch": 1.12, + "learning_rate": 3.143177737881508e-05, + "loss": 0.4415, + "step": 29009 + }, + { + "epoch": 1.12, + "learning_rate": 3.143113618876635e-05, + "loss": 0.3892, + "step": 29010 + }, + { + "epoch": 1.12, + "learning_rate": 3.143049499871762e-05, + "loss": 0.3966, + "step": 29011 + }, + { + "epoch": 1.12, + "learning_rate": 3.1429853808668894e-05, + "loss": 0.4189, + "step": 29012 + }, + { + "epoch": 1.12, + "learning_rate": 3.1429212618620166e-05, + "loss": 0.3551, + "step": 29013 + }, + { + "epoch": 1.12, + "learning_rate": 3.142857142857143e-05, + "loss": 0.3247, + "step": 29014 + }, + { + "epoch": 1.12, + "learning_rate": 3.14279302385227e-05, + "loss": 0.4454, + "step": 29015 + }, + { + "epoch": 1.12, + "learning_rate": 3.142728904847397e-05, + "loss": 0.3149, + "step": 29016 + }, + { + "epoch": 1.12, + "learning_rate": 3.142664785842524e-05, + "loss": 0.4319, + "step": 29017 + }, + { + "epoch": 1.12, + "learning_rate": 3.1426006668376504e-05, + "loss": 0.4522, + "step": 29018 + }, + { + "epoch": 1.12, + "learning_rate": 3.1425365478327776e-05, + "loss": 0.3163, + "step": 29019 + }, + { + "epoch": 1.12, + "learning_rate": 3.142472428827905e-05, + "loss": 0.3356, + "step": 29020 + }, + { + "epoch": 1.12, + "learning_rate": 3.142408309823032e-05, + "loss": 0.2686, + "step": 29021 + }, + { + "epoch": 1.12, + "learning_rate": 3.142344190818159e-05, + "loss": 0.4079, + "step": 29022 + }, + { + "epoch": 1.12, + "learning_rate": 3.1422800718132856e-05, + "loss": 0.5407, + "step": 29023 + }, + { + "epoch": 1.12, + "learning_rate": 3.142215952808413e-05, + "loss": 0.5512, + "step": 29024 + }, + { + "epoch": 1.12, + "learning_rate": 3.142151833803539e-05, + "loss": 0.4126, + "step": 29025 + }, + { + "epoch": 1.12, + "learning_rate": 3.1420877147986664e-05, + "loss": 0.4437, + "step": 29026 + }, + { + "epoch": 1.12, + "learning_rate": 3.142023595793793e-05, + "loss": 0.4093, + "step": 29027 + }, + { + "epoch": 1.12, + "learning_rate": 3.141959476788921e-05, + "loss": 0.3626, + "step": 29028 + }, + { + "epoch": 1.12, + "learning_rate": 3.141895357784047e-05, + "loss": 0.4244, + "step": 29029 + }, + { + "epoch": 1.12, + "learning_rate": 3.1418312387791744e-05, + "loss": 0.32, + "step": 29030 + }, + { + "epoch": 1.12, + "learning_rate": 3.1417671197743016e-05, + "loss": 0.3214, + "step": 29031 + }, + { + "epoch": 1.12, + "learning_rate": 3.141703000769428e-05, + "loss": 0.3488, + "step": 29032 + }, + { + "epoch": 1.12, + "learning_rate": 3.141638881764555e-05, + "loss": 0.3509, + "step": 29033 + }, + { + "epoch": 1.12, + "learning_rate": 3.141574762759682e-05, + "loss": 0.4131, + "step": 29034 + }, + { + "epoch": 1.12, + "learning_rate": 3.141510643754809e-05, + "loss": 0.536, + "step": 29035 + }, + { + "epoch": 1.12, + "learning_rate": 3.141446524749936e-05, + "loss": 0.3539, + "step": 29036 + }, + { + "epoch": 1.12, + "learning_rate": 3.141382405745063e-05, + "loss": 0.3382, + "step": 29037 + }, + { + "epoch": 1.12, + "learning_rate": 3.14131828674019e-05, + "loss": 0.455, + "step": 29038 + }, + { + "epoch": 1.12, + "learning_rate": 3.141254167735317e-05, + "loss": 0.451, + "step": 29039 + }, + { + "epoch": 1.12, + "learning_rate": 3.141190048730444e-05, + "loss": 0.4255, + "step": 29040 + }, + { + "epoch": 1.12, + "learning_rate": 3.1411259297255706e-05, + "loss": 0.4312, + "step": 29041 + }, + { + "epoch": 1.12, + "learning_rate": 3.141061810720698e-05, + "loss": 0.3351, + "step": 29042 + }, + { + "epoch": 1.12, + "learning_rate": 3.140997691715824e-05, + "loss": 0.4754, + "step": 29043 + }, + { + "epoch": 1.12, + "learning_rate": 3.140933572710952e-05, + "loss": 0.4555, + "step": 29044 + }, + { + "epoch": 1.12, + "learning_rate": 3.1408694537060786e-05, + "loss": 0.3911, + "step": 29045 + }, + { + "epoch": 1.12, + "learning_rate": 3.140805334701206e-05, + "loss": 0.3277, + "step": 29046 + }, + { + "epoch": 1.12, + "learning_rate": 3.140741215696332e-05, + "loss": 0.3332, + "step": 29047 + }, + { + "epoch": 1.12, + "learning_rate": 3.1406770966914594e-05, + "loss": 0.3674, + "step": 29048 + }, + { + "epoch": 1.12, + "learning_rate": 3.1406129776865866e-05, + "loss": 0.3993, + "step": 29049 + }, + { + "epoch": 1.12, + "learning_rate": 3.140548858681713e-05, + "loss": 0.4552, + "step": 29050 + }, + { + "epoch": 1.12, + "learning_rate": 3.14048473967684e-05, + "loss": 0.5935, + "step": 29051 + }, + { + "epoch": 1.12, + "learning_rate": 3.1404206206719675e-05, + "loss": 0.4149, + "step": 29052 + }, + { + "epoch": 1.12, + "learning_rate": 3.1403565016670946e-05, + "loss": 0.3983, + "step": 29053 + }, + { + "epoch": 1.12, + "learning_rate": 3.140292382662221e-05, + "loss": 0.5368, + "step": 29054 + }, + { + "epoch": 1.12, + "learning_rate": 3.140228263657348e-05, + "loss": 0.3752, + "step": 29055 + }, + { + "epoch": 1.12, + "learning_rate": 3.140164144652475e-05, + "loss": 0.4332, + "step": 29056 + }, + { + "epoch": 1.12, + "learning_rate": 3.140100025647602e-05, + "loss": 0.3962, + "step": 29057 + }, + { + "epoch": 1.12, + "learning_rate": 3.140035906642729e-05, + "loss": 0.3792, + "step": 29058 + }, + { + "epoch": 1.12, + "learning_rate": 3.1399717876378556e-05, + "loss": 0.3739, + "step": 29059 + }, + { + "epoch": 1.12, + "learning_rate": 3.139907668632983e-05, + "loss": 0.359, + "step": 29060 + }, + { + "epoch": 1.12, + "learning_rate": 3.13984354962811e-05, + "loss": 0.3975, + "step": 29061 + }, + { + "epoch": 1.12, + "learning_rate": 3.139779430623237e-05, + "loss": 0.4317, + "step": 29062 + }, + { + "epoch": 1.12, + "learning_rate": 3.1397153116183636e-05, + "loss": 0.3659, + "step": 29063 + }, + { + "epoch": 1.12, + "learning_rate": 3.139651192613491e-05, + "loss": 0.4307, + "step": 29064 + }, + { + "epoch": 1.12, + "learning_rate": 3.139587073608617e-05, + "loss": 0.3434, + "step": 29065 + }, + { + "epoch": 1.12, + "learning_rate": 3.1395229546037445e-05, + "loss": 0.3448, + "step": 29066 + }, + { + "epoch": 1.12, + "learning_rate": 3.1394588355988716e-05, + "loss": 0.436, + "step": 29067 + }, + { + "epoch": 1.12, + "learning_rate": 3.139394716593999e-05, + "loss": 0.3441, + "step": 29068 + }, + { + "epoch": 1.12, + "learning_rate": 3.139330597589126e-05, + "loss": 0.2892, + "step": 29069 + }, + { + "epoch": 1.12, + "learning_rate": 3.1392664785842525e-05, + "loss": 0.4, + "step": 29070 + }, + { + "epoch": 1.12, + "learning_rate": 3.1392023595793797e-05, + "loss": 0.3451, + "step": 29071 + }, + { + "epoch": 1.12, + "learning_rate": 3.139138240574506e-05, + "loss": 0.3917, + "step": 29072 + }, + { + "epoch": 1.12, + "learning_rate": 3.139074121569633e-05, + "loss": 0.2971, + "step": 29073 + }, + { + "epoch": 1.12, + "learning_rate": 3.13901000256476e-05, + "loss": 0.3445, + "step": 29074 + }, + { + "epoch": 1.12, + "learning_rate": 3.138945883559887e-05, + "loss": 0.2825, + "step": 29075 + }, + { + "epoch": 1.12, + "learning_rate": 3.138881764555014e-05, + "loss": 0.3686, + "step": 29076 + }, + { + "epoch": 1.12, + "learning_rate": 3.138817645550141e-05, + "loss": 0.5241, + "step": 29077 + }, + { + "epoch": 1.12, + "learning_rate": 3.1387535265452685e-05, + "loss": 0.2842, + "step": 29078 + }, + { + "epoch": 1.12, + "learning_rate": 3.138689407540395e-05, + "loss": 0.4572, + "step": 29079 + }, + { + "epoch": 1.12, + "learning_rate": 3.138625288535522e-05, + "loss": 0.355, + "step": 29080 + }, + { + "epoch": 1.12, + "learning_rate": 3.1385611695306487e-05, + "loss": 0.2653, + "step": 29081 + }, + { + "epoch": 1.12, + "learning_rate": 3.138497050525776e-05, + "loss": 0.3218, + "step": 29082 + }, + { + "epoch": 1.12, + "learning_rate": 3.138432931520903e-05, + "loss": 0.3028, + "step": 29083 + }, + { + "epoch": 1.12, + "learning_rate": 3.13836881251603e-05, + "loss": 0.4713, + "step": 29084 + }, + { + "epoch": 1.12, + "learning_rate": 3.1383046935111573e-05, + "loss": 0.3148, + "step": 29085 + }, + { + "epoch": 1.12, + "learning_rate": 3.138240574506284e-05, + "loss": 0.3076, + "step": 29086 + }, + { + "epoch": 1.12, + "learning_rate": 3.138176455501411e-05, + "loss": 0.4508, + "step": 29087 + }, + { + "epoch": 1.12, + "learning_rate": 3.1381123364965375e-05, + "loss": 0.3577, + "step": 29088 + }, + { + "epoch": 1.12, + "learning_rate": 3.138048217491665e-05, + "loss": 0.2899, + "step": 29089 + }, + { + "epoch": 1.12, + "learning_rate": 3.137984098486791e-05, + "loss": 0.5065, + "step": 29090 + }, + { + "epoch": 1.12, + "learning_rate": 3.1379199794819183e-05, + "loss": 0.3746, + "step": 29091 + }, + { + "epoch": 1.12, + "learning_rate": 3.1378558604770455e-05, + "loss": 0.3261, + "step": 29092 + }, + { + "epoch": 1.12, + "learning_rate": 3.137791741472173e-05, + "loss": 0.3533, + "step": 29093 + }, + { + "epoch": 1.12, + "learning_rate": 3.1377276224673e-05, + "loss": 0.4114, + "step": 29094 + }, + { + "epoch": 1.12, + "learning_rate": 3.1376635034624264e-05, + "loss": 0.4668, + "step": 29095 + }, + { + "epoch": 1.12, + "learning_rate": 3.1375993844575535e-05, + "loss": 0.2934, + "step": 29096 + }, + { + "epoch": 1.12, + "learning_rate": 3.13753526545268e-05, + "loss": 0.5092, + "step": 29097 + }, + { + "epoch": 1.12, + "learning_rate": 3.137471146447807e-05, + "loss": 0.4218, + "step": 29098 + }, + { + "epoch": 1.12, + "learning_rate": 3.137407027442934e-05, + "loss": 0.4559, + "step": 29099 + }, + { + "epoch": 1.12, + "learning_rate": 3.1373429084380615e-05, + "loss": 0.3839, + "step": 29100 + }, + { + "epoch": 1.12, + "learning_rate": 3.137278789433188e-05, + "loss": 0.5312, + "step": 29101 + }, + { + "epoch": 1.12, + "learning_rate": 3.137214670428315e-05, + "loss": 0.2636, + "step": 29102 + }, + { + "epoch": 1.12, + "learning_rate": 3.1371505514234424e-05, + "loss": 0.3811, + "step": 29103 + }, + { + "epoch": 1.12, + "learning_rate": 3.137086432418569e-05, + "loss": 0.5363, + "step": 29104 + }, + { + "epoch": 1.12, + "learning_rate": 3.137022313413696e-05, + "loss": 0.3132, + "step": 29105 + }, + { + "epoch": 1.12, + "learning_rate": 3.1369581944088225e-05, + "loss": 0.4044, + "step": 29106 + }, + { + "epoch": 1.12, + "learning_rate": 3.13689407540395e-05, + "loss": 0.35, + "step": 29107 + }, + { + "epoch": 1.12, + "learning_rate": 3.136829956399077e-05, + "loss": 0.4421, + "step": 29108 + }, + { + "epoch": 1.12, + "learning_rate": 3.136765837394204e-05, + "loss": 0.4637, + "step": 29109 + }, + { + "epoch": 1.12, + "learning_rate": 3.136701718389331e-05, + "loss": 0.3, + "step": 29110 + }, + { + "epoch": 1.12, + "learning_rate": 3.136637599384458e-05, + "loss": 0.3987, + "step": 29111 + }, + { + "epoch": 1.12, + "learning_rate": 3.136573480379585e-05, + "loss": 0.5215, + "step": 29112 + }, + { + "epoch": 1.12, + "learning_rate": 3.1365093613747114e-05, + "loss": 0.2931, + "step": 29113 + }, + { + "epoch": 1.12, + "learning_rate": 3.1364452423698385e-05, + "loss": 0.4414, + "step": 29114 + }, + { + "epoch": 1.12, + "learning_rate": 3.136381123364965e-05, + "loss": 0.4315, + "step": 29115 + }, + { + "epoch": 1.12, + "learning_rate": 3.136317004360093e-05, + "loss": 0.4146, + "step": 29116 + }, + { + "epoch": 1.12, + "learning_rate": 3.1362528853552194e-05, + "loss": 0.41, + "step": 29117 + }, + { + "epoch": 1.12, + "learning_rate": 3.1361887663503466e-05, + "loss": 0.4571, + "step": 29118 + }, + { + "epoch": 1.12, + "learning_rate": 3.136124647345474e-05, + "loss": 0.4099, + "step": 29119 + }, + { + "epoch": 1.12, + "learning_rate": 3.1360605283406e-05, + "loss": 0.3681, + "step": 29120 + }, + { + "epoch": 1.12, + "learning_rate": 3.1359964093357274e-05, + "loss": 0.3959, + "step": 29121 + }, + { + "epoch": 1.12, + "learning_rate": 3.135932290330854e-05, + "loss": 0.3678, + "step": 29122 + }, + { + "epoch": 1.12, + "learning_rate": 3.135868171325981e-05, + "loss": 0.2503, + "step": 29123 + }, + { + "epoch": 1.12, + "learning_rate": 3.135804052321108e-05, + "loss": 0.4128, + "step": 29124 + }, + { + "epoch": 1.12, + "learning_rate": 3.1357399333162354e-05, + "loss": 0.4279, + "step": 29125 + }, + { + "epoch": 1.12, + "learning_rate": 3.135675814311362e-05, + "loss": 0.379, + "step": 29126 + }, + { + "epoch": 1.12, + "learning_rate": 3.135611695306489e-05, + "loss": 0.4343, + "step": 29127 + }, + { + "epoch": 1.12, + "learning_rate": 3.135547576301616e-05, + "loss": 0.2142, + "step": 29128 + }, + { + "epoch": 1.12, + "learning_rate": 3.135483457296743e-05, + "loss": 0.396, + "step": 29129 + }, + { + "epoch": 1.12, + "learning_rate": 3.13541933829187e-05, + "loss": 0.2901, + "step": 29130 + }, + { + "epoch": 1.12, + "learning_rate": 3.1353552192869964e-05, + "loss": 0.4019, + "step": 29131 + }, + { + "epoch": 1.12, + "learning_rate": 3.135291100282124e-05, + "loss": 0.4163, + "step": 29132 + }, + { + "epoch": 1.12, + "learning_rate": 3.135226981277251e-05, + "loss": 0.2798, + "step": 29133 + }, + { + "epoch": 1.12, + "learning_rate": 3.135162862272378e-05, + "loss": 0.5383, + "step": 29134 + }, + { + "epoch": 1.12, + "learning_rate": 3.1350987432675044e-05, + "loss": 0.3962, + "step": 29135 + }, + { + "epoch": 1.12, + "learning_rate": 3.1350346242626316e-05, + "loss": 0.3781, + "step": 29136 + }, + { + "epoch": 1.12, + "learning_rate": 3.134970505257759e-05, + "loss": 0.4235, + "step": 29137 + }, + { + "epoch": 1.12, + "learning_rate": 3.134906386252885e-05, + "loss": 0.3569, + "step": 29138 + }, + { + "epoch": 1.12, + "learning_rate": 3.1348422672480124e-05, + "loss": 0.591, + "step": 29139 + }, + { + "epoch": 1.12, + "learning_rate": 3.1347781482431396e-05, + "loss": 0.517, + "step": 29140 + }, + { + "epoch": 1.12, + "learning_rate": 3.134714029238267e-05, + "loss": 0.2789, + "step": 29141 + }, + { + "epoch": 1.12, + "learning_rate": 3.134649910233393e-05, + "loss": 0.3375, + "step": 29142 + }, + { + "epoch": 1.12, + "learning_rate": 3.1345857912285204e-05, + "loss": 0.4304, + "step": 29143 + }, + { + "epoch": 1.12, + "learning_rate": 3.134521672223647e-05, + "loss": 0.3719, + "step": 29144 + }, + { + "epoch": 1.12, + "learning_rate": 3.134457553218774e-05, + "loss": 0.3997, + "step": 29145 + }, + { + "epoch": 1.12, + "learning_rate": 3.134393434213901e-05, + "loss": 0.459, + "step": 29146 + }, + { + "epoch": 1.12, + "learning_rate": 3.134329315209028e-05, + "loss": 0.3493, + "step": 29147 + }, + { + "epoch": 1.12, + "learning_rate": 3.134265196204155e-05, + "loss": 0.4705, + "step": 29148 + }, + { + "epoch": 1.12, + "learning_rate": 3.134201077199282e-05, + "loss": 0.4872, + "step": 29149 + }, + { + "epoch": 1.12, + "learning_rate": 3.134136958194409e-05, + "loss": 0.4713, + "step": 29150 + }, + { + "epoch": 1.12, + "learning_rate": 3.134072839189536e-05, + "loss": 0.3944, + "step": 29151 + }, + { + "epoch": 1.12, + "learning_rate": 3.134008720184663e-05, + "loss": 0.5291, + "step": 29152 + }, + { + "epoch": 1.12, + "learning_rate": 3.1339446011797894e-05, + "loss": 0.5284, + "step": 29153 + }, + { + "epoch": 1.12, + "learning_rate": 3.1338804821749166e-05, + "loss": 0.3743, + "step": 29154 + }, + { + "epoch": 1.12, + "learning_rate": 3.133816363170044e-05, + "loss": 0.4487, + "step": 29155 + }, + { + "epoch": 1.12, + "learning_rate": 3.133752244165171e-05, + "loss": 0.3664, + "step": 29156 + }, + { + "epoch": 1.12, + "learning_rate": 3.133688125160298e-05, + "loss": 0.3755, + "step": 29157 + }, + { + "epoch": 1.12, + "learning_rate": 3.1336240061554246e-05, + "loss": 0.4319, + "step": 29158 + }, + { + "epoch": 1.12, + "learning_rate": 3.133559887150552e-05, + "loss": 0.3606, + "step": 29159 + }, + { + "epoch": 1.12, + "learning_rate": 3.133495768145678e-05, + "loss": 0.4241, + "step": 29160 + }, + { + "epoch": 1.12, + "learning_rate": 3.1334316491408055e-05, + "loss": 0.497, + "step": 29161 + }, + { + "epoch": 1.12, + "learning_rate": 3.133367530135932e-05, + "loss": 0.2873, + "step": 29162 + }, + { + "epoch": 1.12, + "learning_rate": 3.133303411131059e-05, + "loss": 0.3741, + "step": 29163 + }, + { + "epoch": 1.12, + "learning_rate": 3.133239292126186e-05, + "loss": 0.4166, + "step": 29164 + }, + { + "epoch": 1.12, + "learning_rate": 3.1331751731213135e-05, + "loss": 0.3426, + "step": 29165 + }, + { + "epoch": 1.12, + "learning_rate": 3.1331110541164406e-05, + "loss": 0.238, + "step": 29166 + }, + { + "epoch": 1.12, + "learning_rate": 3.133046935111567e-05, + "loss": 0.3568, + "step": 29167 + }, + { + "epoch": 1.12, + "learning_rate": 3.132982816106694e-05, + "loss": 0.4052, + "step": 29168 + }, + { + "epoch": 1.12, + "learning_rate": 3.132918697101821e-05, + "loss": 0.4398, + "step": 29169 + }, + { + "epoch": 1.12, + "learning_rate": 3.132854578096948e-05, + "loss": 0.4996, + "step": 29170 + }, + { + "epoch": 1.12, + "learning_rate": 3.132790459092075e-05, + "loss": 0.4711, + "step": 29171 + }, + { + "epoch": 1.12, + "learning_rate": 3.132726340087202e-05, + "loss": 0.2929, + "step": 29172 + }, + { + "epoch": 1.12, + "learning_rate": 3.1326622210823295e-05, + "loss": 0.5043, + "step": 29173 + }, + { + "epoch": 1.12, + "learning_rate": 3.132598102077456e-05, + "loss": 0.4439, + "step": 29174 + }, + { + "epoch": 1.12, + "learning_rate": 3.132533983072583e-05, + "loss": 0.3713, + "step": 29175 + }, + { + "epoch": 1.12, + "learning_rate": 3.1324698640677096e-05, + "loss": 0.4551, + "step": 29176 + }, + { + "epoch": 1.12, + "learning_rate": 3.132405745062837e-05, + "loss": 0.4328, + "step": 29177 + }, + { + "epoch": 1.12, + "learning_rate": 3.132341626057963e-05, + "loss": 0.2946, + "step": 29178 + }, + { + "epoch": 1.12, + "learning_rate": 3.1322775070530905e-05, + "loss": 0.323, + "step": 29179 + }, + { + "epoch": 1.12, + "learning_rate": 3.1322133880482176e-05, + "loss": 0.3934, + "step": 29180 + }, + { + "epoch": 1.12, + "learning_rate": 3.132149269043345e-05, + "loss": 0.3611, + "step": 29181 + }, + { + "epoch": 1.12, + "learning_rate": 3.132085150038472e-05, + "loss": 0.3889, + "step": 29182 + }, + { + "epoch": 1.12, + "learning_rate": 3.1320210310335985e-05, + "loss": 0.3569, + "step": 29183 + }, + { + "epoch": 1.12, + "learning_rate": 3.1319569120287257e-05, + "loss": 0.394, + "step": 29184 + }, + { + "epoch": 1.12, + "learning_rate": 3.131892793023852e-05, + "loss": 0.3333, + "step": 29185 + }, + { + "epoch": 1.12, + "learning_rate": 3.131828674018979e-05, + "loss": 0.3559, + "step": 29186 + }, + { + "epoch": 1.12, + "learning_rate": 3.131764555014106e-05, + "loss": 0.4989, + "step": 29187 + }, + { + "epoch": 1.12, + "learning_rate": 3.131700436009234e-05, + "loss": 0.3771, + "step": 29188 + }, + { + "epoch": 1.12, + "learning_rate": 3.13163631700436e-05, + "loss": 0.392, + "step": 29189 + }, + { + "epoch": 1.12, + "learning_rate": 3.131572197999487e-05, + "loss": 0.3312, + "step": 29190 + }, + { + "epoch": 1.12, + "learning_rate": 3.1315080789946145e-05, + "loss": 0.2809, + "step": 29191 + }, + { + "epoch": 1.12, + "learning_rate": 3.131443959989741e-05, + "loss": 0.4172, + "step": 29192 + }, + { + "epoch": 1.12, + "learning_rate": 3.131379840984868e-05, + "loss": 0.3386, + "step": 29193 + }, + { + "epoch": 1.12, + "learning_rate": 3.131315721979995e-05, + "loss": 0.4639, + "step": 29194 + }, + { + "epoch": 1.12, + "learning_rate": 3.131251602975122e-05, + "loss": 0.3548, + "step": 29195 + }, + { + "epoch": 1.12, + "learning_rate": 3.131187483970249e-05, + "loss": 0.5353, + "step": 29196 + }, + { + "epoch": 1.12, + "learning_rate": 3.131123364965376e-05, + "loss": 0.4822, + "step": 29197 + }, + { + "epoch": 1.12, + "learning_rate": 3.131059245960503e-05, + "loss": 0.4037, + "step": 29198 + }, + { + "epoch": 1.12, + "learning_rate": 3.13099512695563e-05, + "loss": 0.339, + "step": 29199 + }, + { + "epoch": 1.12, + "learning_rate": 3.130931007950757e-05, + "loss": 0.3694, + "step": 29200 + }, + { + "epoch": 1.12, + "learning_rate": 3.1308668889458835e-05, + "loss": 0.3633, + "step": 29201 + }, + { + "epoch": 1.12, + "learning_rate": 3.130802769941011e-05, + "loss": 0.3738, + "step": 29202 + }, + { + "epoch": 1.12, + "learning_rate": 3.130738650936137e-05, + "loss": 0.3304, + "step": 29203 + }, + { + "epoch": 1.12, + "learning_rate": 3.130674531931265e-05, + "loss": 0.3494, + "step": 29204 + }, + { + "epoch": 1.12, + "learning_rate": 3.1306104129263915e-05, + "loss": 0.4132, + "step": 29205 + }, + { + "epoch": 1.12, + "learning_rate": 3.130546293921519e-05, + "loss": 0.4186, + "step": 29206 + }, + { + "epoch": 1.12, + "learning_rate": 3.130482174916645e-05, + "loss": 0.4638, + "step": 29207 + }, + { + "epoch": 1.12, + "learning_rate": 3.1304180559117724e-05, + "loss": 0.424, + "step": 29208 + }, + { + "epoch": 1.12, + "learning_rate": 3.1303539369068995e-05, + "loss": 0.401, + "step": 29209 + }, + { + "epoch": 1.12, + "learning_rate": 3.130289817902026e-05, + "loss": 0.4547, + "step": 29210 + }, + { + "epoch": 1.12, + "learning_rate": 3.130225698897153e-05, + "loss": 0.4178, + "step": 29211 + }, + { + "epoch": 1.12, + "learning_rate": 3.1301615798922804e-05, + "loss": 0.3308, + "step": 29212 + }, + { + "epoch": 1.12, + "learning_rate": 3.1300974608874075e-05, + "loss": 0.2688, + "step": 29213 + }, + { + "epoch": 1.12, + "learning_rate": 3.130033341882534e-05, + "loss": 0.3215, + "step": 29214 + }, + { + "epoch": 1.12, + "learning_rate": 3.129969222877661e-05, + "loss": 0.2413, + "step": 29215 + }, + { + "epoch": 1.12, + "learning_rate": 3.129905103872788e-05, + "loss": 0.3645, + "step": 29216 + }, + { + "epoch": 1.12, + "learning_rate": 3.129840984867915e-05, + "loss": 0.3511, + "step": 29217 + }, + { + "epoch": 1.12, + "learning_rate": 3.129776865863042e-05, + "loss": 0.3901, + "step": 29218 + }, + { + "epoch": 1.12, + "learning_rate": 3.1297127468581685e-05, + "loss": 0.6767, + "step": 29219 + }, + { + "epoch": 1.12, + "learning_rate": 3.1296486278532964e-05, + "loss": 0.5103, + "step": 29220 + }, + { + "epoch": 1.12, + "learning_rate": 3.129584508848423e-05, + "loss": 0.3923, + "step": 29221 + }, + { + "epoch": 1.12, + "learning_rate": 3.12952038984355e-05, + "loss": 0.459, + "step": 29222 + }, + { + "epoch": 1.12, + "learning_rate": 3.1294562708386765e-05, + "loss": 0.2893, + "step": 29223 + }, + { + "epoch": 1.12, + "learning_rate": 3.129392151833804e-05, + "loss": 0.3929, + "step": 29224 + }, + { + "epoch": 1.12, + "learning_rate": 3.12932803282893e-05, + "loss": 0.3179, + "step": 29225 + }, + { + "epoch": 1.12, + "learning_rate": 3.1292639138240574e-05, + "loss": 0.3631, + "step": 29226 + }, + { + "epoch": 1.12, + "learning_rate": 3.1291997948191846e-05, + "loss": 0.3166, + "step": 29227 + }, + { + "epoch": 1.12, + "learning_rate": 3.129135675814312e-05, + "loss": 0.3956, + "step": 29228 + }, + { + "epoch": 1.12, + "learning_rate": 3.129071556809439e-05, + "loss": 0.4346, + "step": 29229 + }, + { + "epoch": 1.12, + "learning_rate": 3.1290074378045654e-05, + "loss": 0.4927, + "step": 29230 + }, + { + "epoch": 1.12, + "learning_rate": 3.1289433187996926e-05, + "loss": 0.4491, + "step": 29231 + }, + { + "epoch": 1.12, + "learning_rate": 3.128879199794819e-05, + "loss": 0.4297, + "step": 29232 + }, + { + "epoch": 1.12, + "learning_rate": 3.128815080789946e-05, + "loss": 0.399, + "step": 29233 + }, + { + "epoch": 1.12, + "learning_rate": 3.128750961785073e-05, + "loss": 0.2943, + "step": 29234 + }, + { + "epoch": 1.12, + "learning_rate": 3.1286868427802e-05, + "loss": 0.3557, + "step": 29235 + }, + { + "epoch": 1.12, + "learning_rate": 3.128622723775327e-05, + "loss": 0.3227, + "step": 29236 + }, + { + "epoch": 1.12, + "learning_rate": 3.128558604770454e-05, + "loss": 0.5343, + "step": 29237 + }, + { + "epoch": 1.12, + "learning_rate": 3.1284944857655814e-05, + "loss": 0.3412, + "step": 29238 + }, + { + "epoch": 1.12, + "learning_rate": 3.128430366760708e-05, + "loss": 0.4333, + "step": 29239 + }, + { + "epoch": 1.12, + "learning_rate": 3.128366247755835e-05, + "loss": 0.4016, + "step": 29240 + }, + { + "epoch": 1.12, + "learning_rate": 3.1283021287509616e-05, + "loss": 0.4728, + "step": 29241 + }, + { + "epoch": 1.12, + "learning_rate": 3.128238009746089e-05, + "loss": 0.3787, + "step": 29242 + }, + { + "epoch": 1.12, + "learning_rate": 3.128173890741216e-05, + "loss": 0.5852, + "step": 29243 + }, + { + "epoch": 1.12, + "learning_rate": 3.128109771736343e-05, + "loss": 0.3946, + "step": 29244 + }, + { + "epoch": 1.12, + "learning_rate": 3.12804565273147e-05, + "loss": 0.3101, + "step": 29245 + }, + { + "epoch": 1.12, + "learning_rate": 3.127981533726597e-05, + "loss": 0.4533, + "step": 29246 + }, + { + "epoch": 1.12, + "learning_rate": 3.127917414721724e-05, + "loss": 0.4145, + "step": 29247 + }, + { + "epoch": 1.12, + "learning_rate": 3.1278532957168504e-05, + "loss": 0.4993, + "step": 29248 + }, + { + "epoch": 1.12, + "learning_rate": 3.1277891767119776e-05, + "loss": 0.447, + "step": 29249 + }, + { + "epoch": 1.12, + "learning_rate": 3.127725057707104e-05, + "loss": 0.4383, + "step": 29250 + }, + { + "epoch": 1.12, + "learning_rate": 3.127660938702231e-05, + "loss": 0.3172, + "step": 29251 + }, + { + "epoch": 1.12, + "learning_rate": 3.1275968196973584e-05, + "loss": 0.4399, + "step": 29252 + }, + { + "epoch": 1.12, + "learning_rate": 3.1275327006924856e-05, + "loss": 0.5084, + "step": 29253 + }, + { + "epoch": 1.12, + "learning_rate": 3.127468581687613e-05, + "loss": 0.2973, + "step": 29254 + }, + { + "epoch": 1.12, + "learning_rate": 3.127404462682739e-05, + "loss": 0.3822, + "step": 29255 + }, + { + "epoch": 1.12, + "learning_rate": 3.1273403436778664e-05, + "loss": 0.4339, + "step": 29256 + }, + { + "epoch": 1.12, + "learning_rate": 3.127276224672993e-05, + "loss": 0.357, + "step": 29257 + }, + { + "epoch": 1.12, + "learning_rate": 3.12721210566812e-05, + "loss": 0.3706, + "step": 29258 + }, + { + "epoch": 1.12, + "learning_rate": 3.127147986663247e-05, + "loss": 0.3969, + "step": 29259 + }, + { + "epoch": 1.12, + "learning_rate": 3.1270838676583744e-05, + "loss": 0.3549, + "step": 29260 + }, + { + "epoch": 1.12, + "learning_rate": 3.127019748653501e-05, + "loss": 0.4018, + "step": 29261 + }, + { + "epoch": 1.13, + "learning_rate": 3.126955629648628e-05, + "loss": 0.4435, + "step": 29262 + }, + { + "epoch": 1.13, + "learning_rate": 3.126891510643755e-05, + "loss": 0.4537, + "step": 29263 + }, + { + "epoch": 1.13, + "learning_rate": 3.126827391638882e-05, + "loss": 0.3668, + "step": 29264 + }, + { + "epoch": 1.13, + "learning_rate": 3.126763272634009e-05, + "loss": 0.3741, + "step": 29265 + }, + { + "epoch": 1.13, + "learning_rate": 3.1266991536291354e-05, + "loss": 0.323, + "step": 29266 + }, + { + "epoch": 1.13, + "learning_rate": 3.1266350346242626e-05, + "loss": 0.3795, + "step": 29267 + }, + { + "epoch": 1.13, + "learning_rate": 3.12657091561939e-05, + "loss": 0.4391, + "step": 29268 + }, + { + "epoch": 1.13, + "learning_rate": 3.126506796614517e-05, + "loss": 0.3889, + "step": 29269 + }, + { + "epoch": 1.13, + "learning_rate": 3.1264426776096434e-05, + "loss": 0.4215, + "step": 29270 + }, + { + "epoch": 1.13, + "learning_rate": 3.1263785586047706e-05, + "loss": 0.398, + "step": 29271 + }, + { + "epoch": 1.13, + "learning_rate": 3.126314439599898e-05, + "loss": 0.2776, + "step": 29272 + }, + { + "epoch": 1.13, + "learning_rate": 3.126250320595024e-05, + "loss": 0.4566, + "step": 29273 + }, + { + "epoch": 1.13, + "learning_rate": 3.1261862015901515e-05, + "loss": 0.4634, + "step": 29274 + }, + { + "epoch": 1.13, + "learning_rate": 3.126122082585278e-05, + "loss": 0.4237, + "step": 29275 + }, + { + "epoch": 1.13, + "learning_rate": 3.126057963580406e-05, + "loss": 0.3631, + "step": 29276 + }, + { + "epoch": 1.13, + "learning_rate": 3.125993844575532e-05, + "loss": 0.2883, + "step": 29277 + }, + { + "epoch": 1.13, + "learning_rate": 3.1259297255706595e-05, + "loss": 0.373, + "step": 29278 + }, + { + "epoch": 1.13, + "learning_rate": 3.125865606565786e-05, + "loss": 0.2902, + "step": 29279 + }, + { + "epoch": 1.13, + "learning_rate": 3.125801487560913e-05, + "loss": 0.3536, + "step": 29280 + }, + { + "epoch": 1.13, + "learning_rate": 3.12573736855604e-05, + "loss": 0.3037, + "step": 29281 + }, + { + "epoch": 1.13, + "learning_rate": 3.125673249551167e-05, + "loss": 0.3695, + "step": 29282 + }, + { + "epoch": 1.13, + "learning_rate": 3.125609130546294e-05, + "loss": 0.409, + "step": 29283 + }, + { + "epoch": 1.13, + "learning_rate": 3.125545011541421e-05, + "loss": 0.4333, + "step": 29284 + }, + { + "epoch": 1.13, + "learning_rate": 3.125480892536548e-05, + "loss": 0.4673, + "step": 29285 + }, + { + "epoch": 1.13, + "learning_rate": 3.125416773531675e-05, + "loss": 0.3672, + "step": 29286 + }, + { + "epoch": 1.13, + "learning_rate": 3.125352654526802e-05, + "loss": 0.4959, + "step": 29287 + }, + { + "epoch": 1.13, + "learning_rate": 3.1252885355219285e-05, + "loss": 0.4123, + "step": 29288 + }, + { + "epoch": 1.13, + "learning_rate": 3.1252244165170556e-05, + "loss": 0.398, + "step": 29289 + }, + { + "epoch": 1.13, + "learning_rate": 3.125160297512183e-05, + "loss": 0.4177, + "step": 29290 + }, + { + "epoch": 1.13, + "learning_rate": 3.125096178507309e-05, + "loss": 0.4404, + "step": 29291 + }, + { + "epoch": 1.13, + "learning_rate": 3.125032059502437e-05, + "loss": 0.5374, + "step": 29292 + }, + { + "epoch": 1.13, + "learning_rate": 3.1249679404975637e-05, + "loss": 0.377, + "step": 29293 + }, + { + "epoch": 1.13, + "learning_rate": 3.124903821492691e-05, + "loss": 0.3839, + "step": 29294 + }, + { + "epoch": 1.13, + "learning_rate": 3.124839702487817e-05, + "loss": 0.3385, + "step": 29295 + }, + { + "epoch": 1.13, + "learning_rate": 3.1247755834829445e-05, + "loss": 0.4105, + "step": 29296 + }, + { + "epoch": 1.13, + "learning_rate": 3.1247114644780717e-05, + "loss": 0.4423, + "step": 29297 + }, + { + "epoch": 1.13, + "learning_rate": 3.124647345473198e-05, + "loss": 0.4224, + "step": 29298 + }, + { + "epoch": 1.13, + "learning_rate": 3.124583226468325e-05, + "loss": 0.4038, + "step": 29299 + }, + { + "epoch": 1.13, + "learning_rate": 3.1245191074634525e-05, + "loss": 0.2996, + "step": 29300 + }, + { + "epoch": 1.13, + "learning_rate": 3.12445498845858e-05, + "loss": 0.4964, + "step": 29301 + }, + { + "epoch": 1.13, + "learning_rate": 3.124390869453706e-05, + "loss": 0.3189, + "step": 29302 + }, + { + "epoch": 1.13, + "learning_rate": 3.124326750448833e-05, + "loss": 0.3769, + "step": 29303 + }, + { + "epoch": 1.13, + "learning_rate": 3.12426263144396e-05, + "loss": 0.3347, + "step": 29304 + }, + { + "epoch": 1.13, + "learning_rate": 3.124198512439087e-05, + "loss": 0.5451, + "step": 29305 + }, + { + "epoch": 1.13, + "learning_rate": 3.124134393434214e-05, + "loss": 0.3947, + "step": 29306 + }, + { + "epoch": 1.13, + "learning_rate": 3.124070274429341e-05, + "loss": 0.4171, + "step": 29307 + }, + { + "epoch": 1.13, + "learning_rate": 3.124006155424468e-05, + "loss": 0.4596, + "step": 29308 + }, + { + "epoch": 1.13, + "learning_rate": 3.123942036419595e-05, + "loss": 0.3589, + "step": 29309 + }, + { + "epoch": 1.13, + "learning_rate": 3.123877917414722e-05, + "loss": 0.4166, + "step": 29310 + }, + { + "epoch": 1.13, + "learning_rate": 3.123813798409849e-05, + "loss": 0.3888, + "step": 29311 + }, + { + "epoch": 1.13, + "learning_rate": 3.123749679404976e-05, + "loss": 0.4112, + "step": 29312 + }, + { + "epoch": 1.13, + "learning_rate": 3.1236855604001023e-05, + "loss": 0.4129, + "step": 29313 + }, + { + "epoch": 1.13, + "learning_rate": 3.1236214413952295e-05, + "loss": 0.2627, + "step": 29314 + }, + { + "epoch": 1.13, + "learning_rate": 3.123557322390357e-05, + "loss": 0.3918, + "step": 29315 + }, + { + "epoch": 1.13, + "learning_rate": 3.123493203385484e-05, + "loss": 0.4846, + "step": 29316 + }, + { + "epoch": 1.13, + "learning_rate": 3.123429084380611e-05, + "loss": 0.4199, + "step": 29317 + }, + { + "epoch": 1.13, + "learning_rate": 3.1233649653757375e-05, + "loss": 0.3894, + "step": 29318 + }, + { + "epoch": 1.13, + "learning_rate": 3.123300846370865e-05, + "loss": 0.3259, + "step": 29319 + }, + { + "epoch": 1.13, + "learning_rate": 3.123236727365991e-05, + "loss": 0.5216, + "step": 29320 + }, + { + "epoch": 1.13, + "learning_rate": 3.1231726083611184e-05, + "loss": 0.3482, + "step": 29321 + }, + { + "epoch": 1.13, + "learning_rate": 3.123108489356245e-05, + "loss": 0.3697, + "step": 29322 + }, + { + "epoch": 1.13, + "learning_rate": 3.123044370351372e-05, + "loss": 0.2895, + "step": 29323 + }, + { + "epoch": 1.13, + "learning_rate": 3.122980251346499e-05, + "loss": 0.4157, + "step": 29324 + }, + { + "epoch": 1.13, + "learning_rate": 3.1229161323416264e-05, + "loss": 0.5087, + "step": 29325 + }, + { + "epoch": 1.13, + "learning_rate": 3.1228520133367535e-05, + "loss": 0.4173, + "step": 29326 + }, + { + "epoch": 1.13, + "learning_rate": 3.12278789433188e-05, + "loss": 0.3709, + "step": 29327 + }, + { + "epoch": 1.13, + "learning_rate": 3.122723775327007e-05, + "loss": 0.4622, + "step": 29328 + }, + { + "epoch": 1.13, + "learning_rate": 3.122659656322134e-05, + "loss": 0.3525, + "step": 29329 + }, + { + "epoch": 1.13, + "learning_rate": 3.122595537317261e-05, + "loss": 0.4637, + "step": 29330 + }, + { + "epoch": 1.13, + "learning_rate": 3.122531418312388e-05, + "loss": 0.5299, + "step": 29331 + }, + { + "epoch": 1.13, + "learning_rate": 3.122467299307515e-05, + "loss": 0.3789, + "step": 29332 + }, + { + "epoch": 1.13, + "learning_rate": 3.1224031803026424e-05, + "loss": 0.3688, + "step": 29333 + }, + { + "epoch": 1.13, + "learning_rate": 3.122339061297769e-05, + "loss": 0.3247, + "step": 29334 + }, + { + "epoch": 1.13, + "learning_rate": 3.122274942292896e-05, + "loss": 0.4081, + "step": 29335 + }, + { + "epoch": 1.13, + "learning_rate": 3.1222108232880225e-05, + "loss": 0.3009, + "step": 29336 + }, + { + "epoch": 1.13, + "learning_rate": 3.12214670428315e-05, + "loss": 0.4491, + "step": 29337 + }, + { + "epoch": 1.13, + "learning_rate": 3.122082585278276e-05, + "loss": 0.2827, + "step": 29338 + }, + { + "epoch": 1.13, + "learning_rate": 3.1220184662734034e-05, + "loss": 0.4282, + "step": 29339 + }, + { + "epoch": 1.13, + "learning_rate": 3.1219543472685306e-05, + "loss": 0.463, + "step": 29340 + }, + { + "epoch": 1.13, + "learning_rate": 3.121890228263658e-05, + "loss": 0.345, + "step": 29341 + }, + { + "epoch": 1.13, + "learning_rate": 3.121826109258785e-05, + "loss": 0.4303, + "step": 29342 + }, + { + "epoch": 1.13, + "learning_rate": 3.1217619902539114e-05, + "loss": 0.3584, + "step": 29343 + }, + { + "epoch": 1.13, + "learning_rate": 3.1216978712490386e-05, + "loss": 0.5016, + "step": 29344 + }, + { + "epoch": 1.13, + "learning_rate": 3.121633752244165e-05, + "loss": 0.3184, + "step": 29345 + }, + { + "epoch": 1.13, + "learning_rate": 3.121569633239292e-05, + "loss": 0.3738, + "step": 29346 + }, + { + "epoch": 1.13, + "learning_rate": 3.121505514234419e-05, + "loss": 0.4078, + "step": 29347 + }, + { + "epoch": 1.13, + "learning_rate": 3.1214413952295466e-05, + "loss": 0.4565, + "step": 29348 + }, + { + "epoch": 1.13, + "learning_rate": 3.121377276224673e-05, + "loss": 0.2912, + "step": 29349 + }, + { + "epoch": 1.13, + "learning_rate": 3.1213131572198e-05, + "loss": 0.4453, + "step": 29350 + }, + { + "epoch": 1.13, + "learning_rate": 3.1212490382149274e-05, + "loss": 0.3261, + "step": 29351 + }, + { + "epoch": 1.13, + "learning_rate": 3.121184919210054e-05, + "loss": 0.3589, + "step": 29352 + }, + { + "epoch": 1.13, + "learning_rate": 3.121120800205181e-05, + "loss": 0.526, + "step": 29353 + }, + { + "epoch": 1.13, + "learning_rate": 3.1210566812003076e-05, + "loss": 0.5009, + "step": 29354 + }, + { + "epoch": 1.13, + "learning_rate": 3.120992562195435e-05, + "loss": 0.4068, + "step": 29355 + }, + { + "epoch": 1.13, + "learning_rate": 3.120928443190562e-05, + "loss": 0.3082, + "step": 29356 + }, + { + "epoch": 1.13, + "learning_rate": 3.120864324185689e-05, + "loss": 0.4216, + "step": 29357 + }, + { + "epoch": 1.13, + "learning_rate": 3.1208002051808156e-05, + "loss": 0.4256, + "step": 29358 + }, + { + "epoch": 1.13, + "learning_rate": 3.120736086175943e-05, + "loss": 0.3668, + "step": 29359 + }, + { + "epoch": 1.13, + "learning_rate": 3.12067196717107e-05, + "loss": 0.3577, + "step": 29360 + }, + { + "epoch": 1.13, + "learning_rate": 3.1206078481661964e-05, + "loss": 0.4282, + "step": 29361 + }, + { + "epoch": 1.13, + "learning_rate": 3.1205437291613236e-05, + "loss": 0.5138, + "step": 29362 + }, + { + "epoch": 1.13, + "learning_rate": 3.12047961015645e-05, + "loss": 0.5003, + "step": 29363 + }, + { + "epoch": 1.13, + "learning_rate": 3.120415491151578e-05, + "loss": 0.3232, + "step": 29364 + }, + { + "epoch": 1.13, + "learning_rate": 3.1203513721467044e-05, + "loss": 0.439, + "step": 29365 + }, + { + "epoch": 1.13, + "learning_rate": 3.1202872531418316e-05, + "loss": 0.4192, + "step": 29366 + }, + { + "epoch": 1.13, + "learning_rate": 3.120223134136958e-05, + "loss": 0.3351, + "step": 29367 + }, + { + "epoch": 1.13, + "learning_rate": 3.120159015132085e-05, + "loss": 0.3815, + "step": 29368 + }, + { + "epoch": 1.13, + "learning_rate": 3.1200948961272124e-05, + "loss": 0.4212, + "step": 29369 + }, + { + "epoch": 1.13, + "learning_rate": 3.120030777122339e-05, + "loss": 0.4329, + "step": 29370 + }, + { + "epoch": 1.13, + "learning_rate": 3.119966658117466e-05, + "loss": 0.362, + "step": 29371 + }, + { + "epoch": 1.13, + "learning_rate": 3.119902539112593e-05, + "loss": 0.3316, + "step": 29372 + }, + { + "epoch": 1.13, + "learning_rate": 3.1198384201077204e-05, + "loss": 0.3246, + "step": 29373 + }, + { + "epoch": 1.13, + "learning_rate": 3.119774301102847e-05, + "loss": 0.3227, + "step": 29374 + }, + { + "epoch": 1.13, + "learning_rate": 3.119710182097974e-05, + "loss": 0.4342, + "step": 29375 + }, + { + "epoch": 1.13, + "learning_rate": 3.1196460630931006e-05, + "loss": 0.3435, + "step": 29376 + }, + { + "epoch": 1.13, + "learning_rate": 3.119581944088228e-05, + "loss": 0.2892, + "step": 29377 + }, + { + "epoch": 1.13, + "learning_rate": 3.119517825083355e-05, + "loss": 0.479, + "step": 29378 + }, + { + "epoch": 1.13, + "learning_rate": 3.1194537060784814e-05, + "loss": 0.5167, + "step": 29379 + }, + { + "epoch": 1.13, + "learning_rate": 3.119389587073609e-05, + "loss": 0.3936, + "step": 29380 + }, + { + "epoch": 1.13, + "learning_rate": 3.119325468068736e-05, + "loss": 0.4509, + "step": 29381 + }, + { + "epoch": 1.13, + "learning_rate": 3.119261349063863e-05, + "loss": 0.5025, + "step": 29382 + }, + { + "epoch": 1.13, + "learning_rate": 3.1191972300589894e-05, + "loss": 0.5699, + "step": 29383 + }, + { + "epoch": 1.13, + "learning_rate": 3.1191331110541166e-05, + "loss": 0.4978, + "step": 29384 + }, + { + "epoch": 1.13, + "learning_rate": 3.119068992049243e-05, + "loss": 0.3342, + "step": 29385 + }, + { + "epoch": 1.13, + "learning_rate": 3.11900487304437e-05, + "loss": 0.3231, + "step": 29386 + }, + { + "epoch": 1.13, + "learning_rate": 3.1189407540394975e-05, + "loss": 0.507, + "step": 29387 + }, + { + "epoch": 1.13, + "learning_rate": 3.1188766350346246e-05, + "loss": 0.4734, + "step": 29388 + }, + { + "epoch": 1.13, + "learning_rate": 3.118812516029752e-05, + "loss": 0.3598, + "step": 29389 + }, + { + "epoch": 1.13, + "learning_rate": 3.118748397024878e-05, + "loss": 0.4593, + "step": 29390 + }, + { + "epoch": 1.13, + "learning_rate": 3.1186842780200055e-05, + "loss": 0.6637, + "step": 29391 + }, + { + "epoch": 1.13, + "learning_rate": 3.118620159015132e-05, + "loss": 0.4055, + "step": 29392 + }, + { + "epoch": 1.13, + "learning_rate": 3.118556040010259e-05, + "loss": 0.3478, + "step": 29393 + }, + { + "epoch": 1.13, + "learning_rate": 3.1184919210053856e-05, + "loss": 0.4248, + "step": 29394 + }, + { + "epoch": 1.13, + "learning_rate": 3.118427802000513e-05, + "loss": 0.3081, + "step": 29395 + }, + { + "epoch": 1.13, + "learning_rate": 3.11836368299564e-05, + "loss": 0.4017, + "step": 29396 + }, + { + "epoch": 1.13, + "learning_rate": 3.118299563990767e-05, + "loss": 0.3923, + "step": 29397 + }, + { + "epoch": 1.13, + "learning_rate": 3.118235444985894e-05, + "loss": 0.3327, + "step": 29398 + }, + { + "epoch": 1.13, + "learning_rate": 3.118171325981021e-05, + "loss": 0.3743, + "step": 29399 + }, + { + "epoch": 1.13, + "learning_rate": 3.118107206976148e-05, + "loss": 0.4035, + "step": 29400 + }, + { + "epoch": 1.13, + "learning_rate": 3.1180430879712745e-05, + "loss": 0.4307, + "step": 29401 + }, + { + "epoch": 1.13, + "learning_rate": 3.1179789689664016e-05, + "loss": 0.4283, + "step": 29402 + }, + { + "epoch": 1.13, + "learning_rate": 3.117914849961529e-05, + "loss": 0.4219, + "step": 29403 + }, + { + "epoch": 1.13, + "learning_rate": 3.117850730956656e-05, + "loss": 0.569, + "step": 29404 + }, + { + "epoch": 1.13, + "learning_rate": 3.117786611951783e-05, + "loss": 0.3973, + "step": 29405 + }, + { + "epoch": 1.13, + "learning_rate": 3.1177224929469097e-05, + "loss": 0.4128, + "step": 29406 + }, + { + "epoch": 1.13, + "learning_rate": 3.117658373942037e-05, + "loss": 0.4561, + "step": 29407 + }, + { + "epoch": 1.13, + "learning_rate": 3.117594254937163e-05, + "loss": 0.3638, + "step": 29408 + }, + { + "epoch": 1.13, + "learning_rate": 3.1175301359322905e-05, + "loss": 0.3941, + "step": 29409 + }, + { + "epoch": 1.13, + "learning_rate": 3.117466016927417e-05, + "loss": 0.452, + "step": 29410 + }, + { + "epoch": 1.13, + "learning_rate": 3.117401897922544e-05, + "loss": 0.4386, + "step": 29411 + }, + { + "epoch": 1.13, + "learning_rate": 3.117337778917671e-05, + "loss": 0.3492, + "step": 29412 + }, + { + "epoch": 1.13, + "learning_rate": 3.1172736599127985e-05, + "loss": 0.3937, + "step": 29413 + }, + { + "epoch": 1.13, + "learning_rate": 3.117209540907926e-05, + "loss": 0.4034, + "step": 29414 + }, + { + "epoch": 1.13, + "learning_rate": 3.117145421903052e-05, + "loss": 0.3489, + "step": 29415 + }, + { + "epoch": 1.13, + "learning_rate": 3.117081302898179e-05, + "loss": 0.3226, + "step": 29416 + }, + { + "epoch": 1.13, + "learning_rate": 3.117017183893306e-05, + "loss": 0.3252, + "step": 29417 + }, + { + "epoch": 1.13, + "learning_rate": 3.116953064888433e-05, + "loss": 0.4434, + "step": 29418 + }, + { + "epoch": 1.13, + "learning_rate": 3.11688894588356e-05, + "loss": 0.2513, + "step": 29419 + }, + { + "epoch": 1.13, + "learning_rate": 3.1168248268786873e-05, + "loss": 0.4461, + "step": 29420 + }, + { + "epoch": 1.13, + "learning_rate": 3.116760707873814e-05, + "loss": 0.3577, + "step": 29421 + }, + { + "epoch": 1.13, + "learning_rate": 3.116696588868941e-05, + "loss": 0.4253, + "step": 29422 + }, + { + "epoch": 1.13, + "learning_rate": 3.116632469864068e-05, + "loss": 0.3936, + "step": 29423 + }, + { + "epoch": 1.13, + "learning_rate": 3.116568350859195e-05, + "loss": 0.4279, + "step": 29424 + }, + { + "epoch": 1.13, + "learning_rate": 3.116504231854322e-05, + "loss": 0.4524, + "step": 29425 + }, + { + "epoch": 1.13, + "learning_rate": 3.1164401128494483e-05, + "loss": 0.3413, + "step": 29426 + }, + { + "epoch": 1.13, + "learning_rate": 3.1163759938445755e-05, + "loss": 0.2209, + "step": 29427 + }, + { + "epoch": 1.13, + "learning_rate": 3.116311874839703e-05, + "loss": 0.3264, + "step": 29428 + }, + { + "epoch": 1.13, + "learning_rate": 3.11624775583483e-05, + "loss": 0.5377, + "step": 29429 + }, + { + "epoch": 1.13, + "learning_rate": 3.1161836368299564e-05, + "loss": 0.3722, + "step": 29430 + }, + { + "epoch": 1.13, + "learning_rate": 3.1161195178250835e-05, + "loss": 0.4177, + "step": 29431 + }, + { + "epoch": 1.13, + "learning_rate": 3.116055398820211e-05, + "loss": 0.5465, + "step": 29432 + }, + { + "epoch": 1.13, + "learning_rate": 3.115991279815337e-05, + "loss": 0.5143, + "step": 29433 + }, + { + "epoch": 1.13, + "learning_rate": 3.1159271608104644e-05, + "loss": 0.3321, + "step": 29434 + }, + { + "epoch": 1.13, + "learning_rate": 3.115863041805591e-05, + "loss": 0.3988, + "step": 29435 + }, + { + "epoch": 1.13, + "learning_rate": 3.115798922800719e-05, + "loss": 0.4092, + "step": 29436 + }, + { + "epoch": 1.13, + "learning_rate": 3.115734803795845e-05, + "loss": 0.3271, + "step": 29437 + }, + { + "epoch": 1.13, + "learning_rate": 3.1156706847909724e-05, + "loss": 0.344, + "step": 29438 + }, + { + "epoch": 1.13, + "learning_rate": 3.115606565786099e-05, + "loss": 0.4309, + "step": 29439 + }, + { + "epoch": 1.13, + "learning_rate": 3.115542446781226e-05, + "loss": 0.3251, + "step": 29440 + }, + { + "epoch": 1.13, + "learning_rate": 3.115478327776353e-05, + "loss": 0.4077, + "step": 29441 + }, + { + "epoch": 1.13, + "learning_rate": 3.11541420877148e-05, + "loss": 0.4896, + "step": 29442 + }, + { + "epoch": 1.13, + "learning_rate": 3.115350089766607e-05, + "loss": 0.5293, + "step": 29443 + }, + { + "epoch": 1.13, + "learning_rate": 3.115285970761734e-05, + "loss": 0.4119, + "step": 29444 + }, + { + "epoch": 1.13, + "learning_rate": 3.115221851756861e-05, + "loss": 0.3531, + "step": 29445 + }, + { + "epoch": 1.13, + "learning_rate": 3.115157732751988e-05, + "loss": 0.3341, + "step": 29446 + }, + { + "epoch": 1.13, + "learning_rate": 3.115093613747115e-05, + "loss": 0.5059, + "step": 29447 + }, + { + "epoch": 1.13, + "learning_rate": 3.1150294947422414e-05, + "loss": 0.3992, + "step": 29448 + }, + { + "epoch": 1.13, + "learning_rate": 3.1149653757373685e-05, + "loss": 0.285, + "step": 29449 + }, + { + "epoch": 1.13, + "learning_rate": 3.114901256732496e-05, + "loss": 0.3926, + "step": 29450 + }, + { + "epoch": 1.13, + "learning_rate": 3.114837137727622e-05, + "loss": 0.43, + "step": 29451 + }, + { + "epoch": 1.13, + "learning_rate": 3.11477301872275e-05, + "loss": 0.3203, + "step": 29452 + }, + { + "epoch": 1.13, + "learning_rate": 3.1147088997178766e-05, + "loss": 0.3734, + "step": 29453 + }, + { + "epoch": 1.13, + "learning_rate": 3.114644780713004e-05, + "loss": 0.4231, + "step": 29454 + }, + { + "epoch": 1.13, + "learning_rate": 3.11458066170813e-05, + "loss": 0.4889, + "step": 29455 + }, + { + "epoch": 1.13, + "learning_rate": 3.1145165427032574e-05, + "loss": 0.4381, + "step": 29456 + }, + { + "epoch": 1.13, + "learning_rate": 3.114452423698384e-05, + "loss": 0.4708, + "step": 29457 + }, + { + "epoch": 1.13, + "learning_rate": 3.114388304693511e-05, + "loss": 0.4617, + "step": 29458 + }, + { + "epoch": 1.13, + "learning_rate": 3.114324185688638e-05, + "loss": 0.4676, + "step": 29459 + }, + { + "epoch": 1.13, + "learning_rate": 3.1142600666837654e-05, + "loss": 0.2463, + "step": 29460 + }, + { + "epoch": 1.13, + "learning_rate": 3.1141959476788926e-05, + "loss": 0.5238, + "step": 29461 + }, + { + "epoch": 1.13, + "learning_rate": 3.114131828674019e-05, + "loss": 0.3134, + "step": 29462 + }, + { + "epoch": 1.13, + "learning_rate": 3.114067709669146e-05, + "loss": 0.4477, + "step": 29463 + }, + { + "epoch": 1.13, + "learning_rate": 3.114003590664273e-05, + "loss": 0.3716, + "step": 29464 + }, + { + "epoch": 1.13, + "learning_rate": 3.1139394716594e-05, + "loss": 0.4054, + "step": 29465 + }, + { + "epoch": 1.13, + "learning_rate": 3.1138753526545264e-05, + "loss": 0.3722, + "step": 29466 + }, + { + "epoch": 1.13, + "learning_rate": 3.1138112336496536e-05, + "loss": 0.3888, + "step": 29467 + }, + { + "epoch": 1.13, + "learning_rate": 3.113747114644781e-05, + "loss": 0.3703, + "step": 29468 + }, + { + "epoch": 1.13, + "learning_rate": 3.113682995639908e-05, + "loss": 0.4395, + "step": 29469 + }, + { + "epoch": 1.13, + "learning_rate": 3.113618876635035e-05, + "loss": 0.3915, + "step": 29470 + }, + { + "epoch": 1.13, + "learning_rate": 3.1135547576301616e-05, + "loss": 0.4583, + "step": 29471 + }, + { + "epoch": 1.13, + "learning_rate": 3.113490638625289e-05, + "loss": 0.4069, + "step": 29472 + }, + { + "epoch": 1.13, + "learning_rate": 3.113426519620415e-05, + "loss": 0.3992, + "step": 29473 + }, + { + "epoch": 1.13, + "learning_rate": 3.1133624006155424e-05, + "loss": 0.3621, + "step": 29474 + }, + { + "epoch": 1.13, + "learning_rate": 3.1132982816106696e-05, + "loss": 0.3425, + "step": 29475 + }, + { + "epoch": 1.13, + "learning_rate": 3.113234162605797e-05, + "loss": 0.3155, + "step": 29476 + }, + { + "epoch": 1.13, + "learning_rate": 3.113170043600924e-05, + "loss": 0.2813, + "step": 29477 + }, + { + "epoch": 1.13, + "learning_rate": 3.1131059245960504e-05, + "loss": 0.488, + "step": 29478 + }, + { + "epoch": 1.13, + "learning_rate": 3.1130418055911776e-05, + "loss": 0.4745, + "step": 29479 + }, + { + "epoch": 1.13, + "learning_rate": 3.112977686586304e-05, + "loss": 0.3277, + "step": 29480 + }, + { + "epoch": 1.13, + "learning_rate": 3.112913567581431e-05, + "loss": 0.4684, + "step": 29481 + }, + { + "epoch": 1.13, + "learning_rate": 3.112849448576558e-05, + "loss": 0.3496, + "step": 29482 + }, + { + "epoch": 1.13, + "learning_rate": 3.112785329571685e-05, + "loss": 0.419, + "step": 29483 + }, + { + "epoch": 1.13, + "learning_rate": 3.112721210566812e-05, + "loss": 0.3689, + "step": 29484 + }, + { + "epoch": 1.13, + "learning_rate": 3.112657091561939e-05, + "loss": 0.3129, + "step": 29485 + }, + { + "epoch": 1.13, + "learning_rate": 3.1125929725570664e-05, + "loss": 0.4137, + "step": 29486 + }, + { + "epoch": 1.13, + "learning_rate": 3.112528853552193e-05, + "loss": 0.4038, + "step": 29487 + }, + { + "epoch": 1.13, + "learning_rate": 3.11246473454732e-05, + "loss": 0.3736, + "step": 29488 + }, + { + "epoch": 1.13, + "learning_rate": 3.1124006155424466e-05, + "loss": 0.56, + "step": 29489 + }, + { + "epoch": 1.13, + "learning_rate": 3.112336496537574e-05, + "loss": 0.4418, + "step": 29490 + }, + { + "epoch": 1.13, + "learning_rate": 3.112272377532701e-05, + "loss": 0.4957, + "step": 29491 + }, + { + "epoch": 1.13, + "learning_rate": 3.112208258527828e-05, + "loss": 0.5053, + "step": 29492 + }, + { + "epoch": 1.13, + "learning_rate": 3.112144139522955e-05, + "loss": 0.4072, + "step": 29493 + }, + { + "epoch": 1.13, + "learning_rate": 3.112080020518082e-05, + "loss": 0.3258, + "step": 29494 + }, + { + "epoch": 1.13, + "learning_rate": 3.112015901513209e-05, + "loss": 0.4702, + "step": 29495 + }, + { + "epoch": 1.13, + "learning_rate": 3.1119517825083355e-05, + "loss": 0.3033, + "step": 29496 + }, + { + "epoch": 1.13, + "learning_rate": 3.1118876635034626e-05, + "loss": 0.2573, + "step": 29497 + }, + { + "epoch": 1.13, + "learning_rate": 3.111823544498589e-05, + "loss": 0.5158, + "step": 29498 + }, + { + "epoch": 1.13, + "learning_rate": 3.111759425493716e-05, + "loss": 0.4452, + "step": 29499 + }, + { + "epoch": 1.13, + "learning_rate": 3.1116953064888435e-05, + "loss": 0.4453, + "step": 29500 + }, + { + "epoch": 1.13, + "learning_rate": 3.1116311874839706e-05, + "loss": 0.4524, + "step": 29501 + }, + { + "epoch": 1.13, + "learning_rate": 3.111567068479098e-05, + "loss": 0.4807, + "step": 29502 + }, + { + "epoch": 1.13, + "learning_rate": 3.111502949474224e-05, + "loss": 0.4197, + "step": 29503 + }, + { + "epoch": 1.13, + "learning_rate": 3.1114388304693515e-05, + "loss": 0.3237, + "step": 29504 + }, + { + "epoch": 1.13, + "learning_rate": 3.111374711464478e-05, + "loss": 0.39, + "step": 29505 + }, + { + "epoch": 1.13, + "learning_rate": 3.111310592459605e-05, + "loss": 0.3008, + "step": 29506 + }, + { + "epoch": 1.13, + "learning_rate": 3.111246473454732e-05, + "loss": 0.4265, + "step": 29507 + }, + { + "epoch": 1.13, + "learning_rate": 3.1111823544498595e-05, + "loss": 0.4051, + "step": 29508 + }, + { + "epoch": 1.13, + "learning_rate": 3.111118235444986e-05, + "loss": 0.3931, + "step": 29509 + }, + { + "epoch": 1.13, + "learning_rate": 3.111054116440113e-05, + "loss": 0.2571, + "step": 29510 + }, + { + "epoch": 1.13, + "learning_rate": 3.11098999743524e-05, + "loss": 0.3593, + "step": 29511 + }, + { + "epoch": 1.13, + "learning_rate": 3.110925878430367e-05, + "loss": 0.3569, + "step": 29512 + }, + { + "epoch": 1.13, + "learning_rate": 3.110861759425494e-05, + "loss": 0.3809, + "step": 29513 + }, + { + "epoch": 1.13, + "learning_rate": 3.1107976404206205e-05, + "loss": 0.3461, + "step": 29514 + }, + { + "epoch": 1.13, + "learning_rate": 3.1107335214157476e-05, + "loss": 0.4679, + "step": 29515 + }, + { + "epoch": 1.13, + "learning_rate": 3.110669402410875e-05, + "loss": 0.4678, + "step": 29516 + }, + { + "epoch": 1.13, + "learning_rate": 3.110605283406002e-05, + "loss": 0.4271, + "step": 29517 + }, + { + "epoch": 1.13, + "learning_rate": 3.1105411644011285e-05, + "loss": 0.4078, + "step": 29518 + }, + { + "epoch": 1.13, + "learning_rate": 3.1104770453962557e-05, + "loss": 0.4495, + "step": 29519 + }, + { + "epoch": 1.13, + "learning_rate": 3.110412926391383e-05, + "loss": 0.3564, + "step": 29520 + }, + { + "epoch": 1.13, + "learning_rate": 3.110348807386509e-05, + "loss": 0.3555, + "step": 29521 + }, + { + "epoch": 1.14, + "learning_rate": 3.1102846883816365e-05, + "loss": 0.3471, + "step": 29522 + }, + { + "epoch": 1.14, + "learning_rate": 3.110220569376763e-05, + "loss": 0.3431, + "step": 29523 + }, + { + "epoch": 1.14, + "learning_rate": 3.110156450371891e-05, + "loss": 0.4494, + "step": 29524 + }, + { + "epoch": 1.14, + "learning_rate": 3.110092331367017e-05, + "loss": 0.4016, + "step": 29525 + }, + { + "epoch": 1.14, + "learning_rate": 3.1100282123621445e-05, + "loss": 0.345, + "step": 29526 + }, + { + "epoch": 1.14, + "learning_rate": 3.109964093357271e-05, + "loss": 0.3495, + "step": 29527 + }, + { + "epoch": 1.14, + "learning_rate": 3.109899974352398e-05, + "loss": 0.3836, + "step": 29528 + }, + { + "epoch": 1.14, + "learning_rate": 3.1098358553475253e-05, + "loss": 0.3348, + "step": 29529 + }, + { + "epoch": 1.14, + "learning_rate": 3.109771736342652e-05, + "loss": 0.3765, + "step": 29530 + }, + { + "epoch": 1.14, + "learning_rate": 3.109707617337779e-05, + "loss": 0.3969, + "step": 29531 + }, + { + "epoch": 1.14, + "learning_rate": 3.109643498332906e-05, + "loss": 0.3937, + "step": 29532 + }, + { + "epoch": 1.14, + "learning_rate": 3.1095793793280334e-05, + "loss": 0.3223, + "step": 29533 + }, + { + "epoch": 1.14, + "learning_rate": 3.10951526032316e-05, + "loss": 0.2714, + "step": 29534 + }, + { + "epoch": 1.14, + "learning_rate": 3.109451141318287e-05, + "loss": 0.4453, + "step": 29535 + }, + { + "epoch": 1.14, + "learning_rate": 3.1093870223134135e-05, + "loss": 0.4385, + "step": 29536 + }, + { + "epoch": 1.14, + "learning_rate": 3.109322903308541e-05, + "loss": 0.357, + "step": 29537 + }, + { + "epoch": 1.14, + "learning_rate": 3.109258784303668e-05, + "loss": 0.3115, + "step": 29538 + }, + { + "epoch": 1.14, + "learning_rate": 3.1091946652987943e-05, + "loss": 0.3032, + "step": 29539 + }, + { + "epoch": 1.14, + "learning_rate": 3.109130546293922e-05, + "loss": 0.3828, + "step": 29540 + }, + { + "epoch": 1.14, + "learning_rate": 3.109066427289049e-05, + "loss": 0.4197, + "step": 29541 + }, + { + "epoch": 1.14, + "learning_rate": 3.109002308284176e-05, + "loss": 0.6155, + "step": 29542 + }, + { + "epoch": 1.14, + "learning_rate": 3.1089381892793024e-05, + "loss": 0.341, + "step": 29543 + }, + { + "epoch": 1.14, + "learning_rate": 3.1088740702744295e-05, + "loss": 0.4304, + "step": 29544 + }, + { + "epoch": 1.14, + "learning_rate": 3.108809951269556e-05, + "loss": 0.3703, + "step": 29545 + }, + { + "epoch": 1.14, + "learning_rate": 3.108745832264683e-05, + "loss": 0.335, + "step": 29546 + }, + { + "epoch": 1.14, + "learning_rate": 3.1086817132598104e-05, + "loss": 0.3955, + "step": 29547 + }, + { + "epoch": 1.14, + "learning_rate": 3.1086175942549375e-05, + "loss": 0.4652, + "step": 29548 + }, + { + "epoch": 1.14, + "learning_rate": 3.108553475250065e-05, + "loss": 0.2893, + "step": 29549 + }, + { + "epoch": 1.14, + "learning_rate": 3.108489356245191e-05, + "loss": 0.3113, + "step": 29550 + }, + { + "epoch": 1.14, + "learning_rate": 3.1084252372403184e-05, + "loss": 0.446, + "step": 29551 + }, + { + "epoch": 1.14, + "learning_rate": 3.108361118235445e-05, + "loss": 0.47, + "step": 29552 + }, + { + "epoch": 1.14, + "learning_rate": 3.108296999230572e-05, + "loss": 0.4666, + "step": 29553 + }, + { + "epoch": 1.14, + "learning_rate": 3.1082328802256985e-05, + "loss": 0.3844, + "step": 29554 + }, + { + "epoch": 1.14, + "learning_rate": 3.108168761220826e-05, + "loss": 0.3823, + "step": 29555 + }, + { + "epoch": 1.14, + "learning_rate": 3.108104642215953e-05, + "loss": 0.3096, + "step": 29556 + }, + { + "epoch": 1.14, + "learning_rate": 3.10804052321108e-05, + "loss": 0.3832, + "step": 29557 + }, + { + "epoch": 1.14, + "learning_rate": 3.107976404206207e-05, + "loss": 0.2675, + "step": 29558 + }, + { + "epoch": 1.14, + "learning_rate": 3.107912285201334e-05, + "loss": 0.2968, + "step": 29559 + }, + { + "epoch": 1.14, + "learning_rate": 3.107848166196461e-05, + "loss": 0.4139, + "step": 29560 + }, + { + "epoch": 1.14, + "learning_rate": 3.1077840471915874e-05, + "loss": 0.4831, + "step": 29561 + }, + { + "epoch": 1.14, + "learning_rate": 3.1077199281867146e-05, + "loss": 0.402, + "step": 29562 + }, + { + "epoch": 1.14, + "learning_rate": 3.107655809181842e-05, + "loss": 0.3285, + "step": 29563 + }, + { + "epoch": 1.14, + "learning_rate": 3.107591690176969e-05, + "loss": 0.3735, + "step": 29564 + }, + { + "epoch": 1.14, + "learning_rate": 3.107527571172096e-05, + "loss": 0.3683, + "step": 29565 + }, + { + "epoch": 1.14, + "learning_rate": 3.1074634521672226e-05, + "loss": 0.476, + "step": 29566 + }, + { + "epoch": 1.14, + "learning_rate": 3.10739933316235e-05, + "loss": 0.4153, + "step": 29567 + }, + { + "epoch": 1.14, + "learning_rate": 3.107335214157476e-05, + "loss": 0.3734, + "step": 29568 + }, + { + "epoch": 1.14, + "learning_rate": 3.1072710951526034e-05, + "loss": 0.3599, + "step": 29569 + }, + { + "epoch": 1.14, + "learning_rate": 3.10720697614773e-05, + "loss": 0.3321, + "step": 29570 + }, + { + "epoch": 1.14, + "learning_rate": 3.107142857142857e-05, + "loss": 0.3835, + "step": 29571 + }, + { + "epoch": 1.14, + "learning_rate": 3.107078738137984e-05, + "loss": 0.5372, + "step": 29572 + }, + { + "epoch": 1.14, + "learning_rate": 3.1070146191331114e-05, + "loss": 0.402, + "step": 29573 + }, + { + "epoch": 1.14, + "learning_rate": 3.1069505001282386e-05, + "loss": 0.3654, + "step": 29574 + }, + { + "epoch": 1.14, + "learning_rate": 3.106886381123365e-05, + "loss": 0.3319, + "step": 29575 + }, + { + "epoch": 1.14, + "learning_rate": 3.106822262118492e-05, + "loss": 0.5243, + "step": 29576 + }, + { + "epoch": 1.14, + "learning_rate": 3.106758143113619e-05, + "loss": 0.4333, + "step": 29577 + }, + { + "epoch": 1.14, + "learning_rate": 3.106694024108746e-05, + "loss": 0.3907, + "step": 29578 + }, + { + "epoch": 1.14, + "learning_rate": 3.106629905103873e-05, + "loss": 0.304, + "step": 29579 + }, + { + "epoch": 1.14, + "learning_rate": 3.106565786099e-05, + "loss": 0.3289, + "step": 29580 + }, + { + "epoch": 1.14, + "learning_rate": 3.106501667094127e-05, + "loss": 0.3985, + "step": 29581 + }, + { + "epoch": 1.14, + "learning_rate": 3.106437548089254e-05, + "loss": 0.6537, + "step": 29582 + }, + { + "epoch": 1.14, + "learning_rate": 3.106373429084381e-05, + "loss": 0.3395, + "step": 29583 + }, + { + "epoch": 1.14, + "learning_rate": 3.1063093100795076e-05, + "loss": 0.3349, + "step": 29584 + }, + { + "epoch": 1.14, + "learning_rate": 3.106245191074635e-05, + "loss": 0.3538, + "step": 29585 + }, + { + "epoch": 1.14, + "learning_rate": 3.106181072069761e-05, + "loss": 0.4102, + "step": 29586 + }, + { + "epoch": 1.14, + "learning_rate": 3.1061169530648884e-05, + "loss": 0.2678, + "step": 29587 + }, + { + "epoch": 1.14, + "learning_rate": 3.1060528340600156e-05, + "loss": 0.3665, + "step": 29588 + }, + { + "epoch": 1.14, + "learning_rate": 3.105988715055143e-05, + "loss": 0.453, + "step": 29589 + }, + { + "epoch": 1.14, + "learning_rate": 3.105924596050269e-05, + "loss": 0.4311, + "step": 29590 + }, + { + "epoch": 1.14, + "learning_rate": 3.1058604770453964e-05, + "loss": 0.2903, + "step": 29591 + }, + { + "epoch": 1.14, + "learning_rate": 3.1057963580405236e-05, + "loss": 0.3557, + "step": 29592 + }, + { + "epoch": 1.14, + "learning_rate": 3.10573223903565e-05, + "loss": 0.4306, + "step": 29593 + }, + { + "epoch": 1.14, + "learning_rate": 3.105668120030777e-05, + "loss": 0.3925, + "step": 29594 + }, + { + "epoch": 1.14, + "learning_rate": 3.105604001025904e-05, + "loss": 0.3627, + "step": 29595 + }, + { + "epoch": 1.14, + "learning_rate": 3.1055398820210316e-05, + "loss": 0.3389, + "step": 29596 + }, + { + "epoch": 1.14, + "learning_rate": 3.105475763016158e-05, + "loss": 0.5966, + "step": 29597 + }, + { + "epoch": 1.14, + "learning_rate": 3.105411644011285e-05, + "loss": 0.3813, + "step": 29598 + }, + { + "epoch": 1.14, + "learning_rate": 3.105347525006412e-05, + "loss": 0.4442, + "step": 29599 + }, + { + "epoch": 1.14, + "learning_rate": 3.105283406001539e-05, + "loss": 0.431, + "step": 29600 + }, + { + "epoch": 1.14, + "learning_rate": 3.105219286996666e-05, + "loss": 0.3842, + "step": 29601 + }, + { + "epoch": 1.14, + "learning_rate": 3.1051551679917926e-05, + "loss": 0.4073, + "step": 29602 + }, + { + "epoch": 1.14, + "learning_rate": 3.10509104898692e-05, + "loss": 0.4594, + "step": 29603 + }, + { + "epoch": 1.14, + "learning_rate": 3.105026929982047e-05, + "loss": 0.4049, + "step": 29604 + }, + { + "epoch": 1.14, + "learning_rate": 3.104962810977174e-05, + "loss": 0.408, + "step": 29605 + }, + { + "epoch": 1.14, + "learning_rate": 3.1048986919723006e-05, + "loss": 0.4011, + "step": 29606 + }, + { + "epoch": 1.14, + "learning_rate": 3.104834572967428e-05, + "loss": 0.3766, + "step": 29607 + }, + { + "epoch": 1.14, + "learning_rate": 3.104770453962554e-05, + "loss": 0.3462, + "step": 29608 + }, + { + "epoch": 1.14, + "learning_rate": 3.1047063349576815e-05, + "loss": 0.3081, + "step": 29609 + }, + { + "epoch": 1.14, + "learning_rate": 3.1046422159528086e-05, + "loss": 0.3396, + "step": 29610 + }, + { + "epoch": 1.14, + "learning_rate": 3.104578096947935e-05, + "loss": 0.383, + "step": 29611 + }, + { + "epoch": 1.14, + "learning_rate": 3.104513977943063e-05, + "loss": 0.4496, + "step": 29612 + }, + { + "epoch": 1.14, + "learning_rate": 3.1044498589381895e-05, + "loss": 0.2817, + "step": 29613 + }, + { + "epoch": 1.14, + "learning_rate": 3.1043857399333166e-05, + "loss": 0.3174, + "step": 29614 + }, + { + "epoch": 1.14, + "learning_rate": 3.104321620928443e-05, + "loss": 0.3657, + "step": 29615 + }, + { + "epoch": 1.14, + "learning_rate": 3.10425750192357e-05, + "loss": 0.3146, + "step": 29616 + }, + { + "epoch": 1.14, + "learning_rate": 3.104193382918697e-05, + "loss": 0.3654, + "step": 29617 + }, + { + "epoch": 1.14, + "learning_rate": 3.104129263913824e-05, + "loss": 0.4491, + "step": 29618 + }, + { + "epoch": 1.14, + "learning_rate": 3.104065144908951e-05, + "loss": 0.3774, + "step": 29619 + }, + { + "epoch": 1.14, + "learning_rate": 3.104001025904078e-05, + "loss": 0.4996, + "step": 29620 + }, + { + "epoch": 1.14, + "learning_rate": 3.1039369068992055e-05, + "loss": 0.5673, + "step": 29621 + }, + { + "epoch": 1.14, + "learning_rate": 3.103872787894332e-05, + "loss": 0.3184, + "step": 29622 + }, + { + "epoch": 1.14, + "learning_rate": 3.103808668889459e-05, + "loss": 0.3751, + "step": 29623 + }, + { + "epoch": 1.14, + "learning_rate": 3.1037445498845856e-05, + "loss": 0.3238, + "step": 29624 + }, + { + "epoch": 1.14, + "learning_rate": 3.103680430879713e-05, + "loss": 0.423, + "step": 29625 + }, + { + "epoch": 1.14, + "learning_rate": 3.103616311874839e-05, + "loss": 0.4456, + "step": 29626 + }, + { + "epoch": 1.14, + "learning_rate": 3.1035521928699665e-05, + "loss": 0.333, + "step": 29627 + }, + { + "epoch": 1.14, + "learning_rate": 3.103488073865094e-05, + "loss": 0.3892, + "step": 29628 + }, + { + "epoch": 1.14, + "learning_rate": 3.103423954860221e-05, + "loss": 0.3735, + "step": 29629 + }, + { + "epoch": 1.14, + "learning_rate": 3.103359835855348e-05, + "loss": 0.2998, + "step": 29630 + }, + { + "epoch": 1.14, + "learning_rate": 3.1032957168504745e-05, + "loss": 0.3484, + "step": 29631 + }, + { + "epoch": 1.14, + "learning_rate": 3.1032315978456017e-05, + "loss": 0.4099, + "step": 29632 + }, + { + "epoch": 1.14, + "learning_rate": 3.103167478840728e-05, + "loss": 0.415, + "step": 29633 + }, + { + "epoch": 1.14, + "learning_rate": 3.103103359835855e-05, + "loss": 0.4559, + "step": 29634 + }, + { + "epoch": 1.14, + "learning_rate": 3.1030392408309825e-05, + "loss": 0.4404, + "step": 29635 + }, + { + "epoch": 1.14, + "learning_rate": 3.10297512182611e-05, + "loss": 0.4251, + "step": 29636 + }, + { + "epoch": 1.14, + "learning_rate": 3.102911002821237e-05, + "loss": 0.5453, + "step": 29637 + }, + { + "epoch": 1.14, + "learning_rate": 3.102846883816363e-05, + "loss": 0.4859, + "step": 29638 + }, + { + "epoch": 1.14, + "learning_rate": 3.1027827648114905e-05, + "loss": 0.294, + "step": 29639 + }, + { + "epoch": 1.14, + "learning_rate": 3.102718645806617e-05, + "loss": 0.4181, + "step": 29640 + }, + { + "epoch": 1.14, + "learning_rate": 3.102654526801744e-05, + "loss": 0.4242, + "step": 29641 + }, + { + "epoch": 1.14, + "learning_rate": 3.102590407796871e-05, + "loss": 0.4308, + "step": 29642 + }, + { + "epoch": 1.14, + "learning_rate": 3.102526288791998e-05, + "loss": 0.3038, + "step": 29643 + }, + { + "epoch": 1.14, + "learning_rate": 3.102462169787125e-05, + "loss": 0.4426, + "step": 29644 + }, + { + "epoch": 1.14, + "learning_rate": 3.102398050782252e-05, + "loss": 0.4585, + "step": 29645 + }, + { + "epoch": 1.14, + "learning_rate": 3.1023339317773794e-05, + "loss": 0.4865, + "step": 29646 + }, + { + "epoch": 1.14, + "learning_rate": 3.102269812772506e-05, + "loss": 0.4685, + "step": 29647 + }, + { + "epoch": 1.14, + "learning_rate": 3.102205693767633e-05, + "loss": 0.3342, + "step": 29648 + }, + { + "epoch": 1.14, + "learning_rate": 3.1021415747627595e-05, + "loss": 0.4605, + "step": 29649 + }, + { + "epoch": 1.14, + "learning_rate": 3.102077455757887e-05, + "loss": 0.4334, + "step": 29650 + }, + { + "epoch": 1.14, + "learning_rate": 3.102013336753014e-05, + "loss": 0.3133, + "step": 29651 + }, + { + "epoch": 1.14, + "learning_rate": 3.101949217748141e-05, + "loss": 0.2841, + "step": 29652 + }, + { + "epoch": 1.14, + "learning_rate": 3.1018850987432675e-05, + "loss": 0.2479, + "step": 29653 + }, + { + "epoch": 1.14, + "learning_rate": 3.101820979738395e-05, + "loss": 0.3688, + "step": 29654 + }, + { + "epoch": 1.14, + "learning_rate": 3.101756860733522e-05, + "loss": 0.4213, + "step": 29655 + }, + { + "epoch": 1.14, + "learning_rate": 3.1016927417286484e-05, + "loss": 0.2375, + "step": 29656 + }, + { + "epoch": 1.14, + "learning_rate": 3.1016286227237755e-05, + "loss": 0.4717, + "step": 29657 + }, + { + "epoch": 1.14, + "learning_rate": 3.101564503718902e-05, + "loss": 0.3457, + "step": 29658 + }, + { + "epoch": 1.14, + "learning_rate": 3.101500384714029e-05, + "loss": 0.4692, + "step": 29659 + }, + { + "epoch": 1.14, + "learning_rate": 3.1014362657091564e-05, + "loss": 0.3449, + "step": 29660 + }, + { + "epoch": 1.14, + "learning_rate": 3.1013721467042835e-05, + "loss": 0.3867, + "step": 29661 + }, + { + "epoch": 1.14, + "learning_rate": 3.10130802769941e-05, + "loss": 0.476, + "step": 29662 + }, + { + "epoch": 1.14, + "learning_rate": 3.101243908694537e-05, + "loss": 0.3848, + "step": 29663 + }, + { + "epoch": 1.14, + "learning_rate": 3.1011797896896644e-05, + "loss": 0.4972, + "step": 29664 + }, + { + "epoch": 1.14, + "learning_rate": 3.101115670684791e-05, + "loss": 0.3696, + "step": 29665 + }, + { + "epoch": 1.14, + "learning_rate": 3.101051551679918e-05, + "loss": 0.4253, + "step": 29666 + }, + { + "epoch": 1.14, + "learning_rate": 3.100987432675045e-05, + "loss": 0.4086, + "step": 29667 + }, + { + "epoch": 1.14, + "learning_rate": 3.1009233136701724e-05, + "loss": 0.2435, + "step": 29668 + }, + { + "epoch": 1.14, + "learning_rate": 3.100859194665299e-05, + "loss": 0.3373, + "step": 29669 + }, + { + "epoch": 1.14, + "learning_rate": 3.100795075660426e-05, + "loss": 0.4115, + "step": 29670 + }, + { + "epoch": 1.14, + "learning_rate": 3.1007309566555525e-05, + "loss": 0.3356, + "step": 29671 + }, + { + "epoch": 1.14, + "learning_rate": 3.10066683765068e-05, + "loss": 0.3436, + "step": 29672 + }, + { + "epoch": 1.14, + "learning_rate": 3.100602718645807e-05, + "loss": 0.3705, + "step": 29673 + }, + { + "epoch": 1.14, + "learning_rate": 3.1005385996409334e-05, + "loss": 0.3313, + "step": 29674 + }, + { + "epoch": 1.14, + "learning_rate": 3.1004744806360606e-05, + "loss": 0.4252, + "step": 29675 + }, + { + "epoch": 1.14, + "learning_rate": 3.100410361631188e-05, + "loss": 0.3453, + "step": 29676 + }, + { + "epoch": 1.14, + "learning_rate": 3.100346242626315e-05, + "loss": 0.3105, + "step": 29677 + }, + { + "epoch": 1.14, + "learning_rate": 3.1002821236214414e-05, + "loss": 0.2763, + "step": 29678 + }, + { + "epoch": 1.14, + "learning_rate": 3.1002180046165686e-05, + "loss": 0.4957, + "step": 29679 + }, + { + "epoch": 1.14, + "learning_rate": 3.100153885611696e-05, + "loss": 0.2703, + "step": 29680 + }, + { + "epoch": 1.14, + "learning_rate": 3.100089766606822e-05, + "loss": 0.2858, + "step": 29681 + }, + { + "epoch": 1.14, + "learning_rate": 3.1000256476019494e-05, + "loss": 0.4447, + "step": 29682 + }, + { + "epoch": 1.14, + "learning_rate": 3.099961528597076e-05, + "loss": 0.3824, + "step": 29683 + }, + { + "epoch": 1.14, + "learning_rate": 3.099897409592204e-05, + "loss": 0.2805, + "step": 29684 + }, + { + "epoch": 1.14, + "learning_rate": 3.09983329058733e-05, + "loss": 0.5373, + "step": 29685 + }, + { + "epoch": 1.14, + "learning_rate": 3.0997691715824574e-05, + "loss": 0.5429, + "step": 29686 + }, + { + "epoch": 1.14, + "learning_rate": 3.099705052577584e-05, + "loss": 0.4024, + "step": 29687 + }, + { + "epoch": 1.14, + "learning_rate": 3.099640933572711e-05, + "loss": 0.3135, + "step": 29688 + }, + { + "epoch": 1.14, + "learning_rate": 3.099576814567838e-05, + "loss": 0.4807, + "step": 29689 + }, + { + "epoch": 1.14, + "learning_rate": 3.099512695562965e-05, + "loss": 0.4631, + "step": 29690 + }, + { + "epoch": 1.14, + "learning_rate": 3.099448576558092e-05, + "loss": 0.3826, + "step": 29691 + }, + { + "epoch": 1.14, + "learning_rate": 3.099384457553219e-05, + "loss": 0.4264, + "step": 29692 + }, + { + "epoch": 1.14, + "learning_rate": 3.099320338548346e-05, + "loss": 0.327, + "step": 29693 + }, + { + "epoch": 1.14, + "learning_rate": 3.099256219543473e-05, + "loss": 0.406, + "step": 29694 + }, + { + "epoch": 1.14, + "learning_rate": 3.0991921005386e-05, + "loss": 0.4116, + "step": 29695 + }, + { + "epoch": 1.14, + "learning_rate": 3.0991279815337264e-05, + "loss": 0.2973, + "step": 29696 + }, + { + "epoch": 1.14, + "learning_rate": 3.0990638625288536e-05, + "loss": 0.271, + "step": 29697 + }, + { + "epoch": 1.14, + "learning_rate": 3.098999743523981e-05, + "loss": 0.3462, + "step": 29698 + }, + { + "epoch": 1.14, + "learning_rate": 3.098935624519107e-05, + "loss": 0.5003, + "step": 29699 + }, + { + "epoch": 1.14, + "learning_rate": 3.098871505514235e-05, + "loss": 0.5117, + "step": 29700 + }, + { + "epoch": 1.14, + "learning_rate": 3.0988073865093616e-05, + "loss": 0.3695, + "step": 29701 + }, + { + "epoch": 1.14, + "learning_rate": 3.098743267504489e-05, + "loss": 0.3304, + "step": 29702 + }, + { + "epoch": 1.14, + "learning_rate": 3.098679148499615e-05, + "loss": 0.3563, + "step": 29703 + }, + { + "epoch": 1.14, + "learning_rate": 3.0986150294947424e-05, + "loss": 0.4689, + "step": 29704 + }, + { + "epoch": 1.14, + "learning_rate": 3.098550910489869e-05, + "loss": 0.4153, + "step": 29705 + }, + { + "epoch": 1.14, + "learning_rate": 3.098486791484996e-05, + "loss": 0.3652, + "step": 29706 + }, + { + "epoch": 1.14, + "learning_rate": 3.098422672480123e-05, + "loss": 0.5558, + "step": 29707 + }, + { + "epoch": 1.14, + "learning_rate": 3.0983585534752504e-05, + "loss": 0.3908, + "step": 29708 + }, + { + "epoch": 1.14, + "learning_rate": 3.0982944344703776e-05, + "loss": 0.3614, + "step": 29709 + }, + { + "epoch": 1.14, + "learning_rate": 3.098230315465504e-05, + "loss": 0.3946, + "step": 29710 + }, + { + "epoch": 1.14, + "learning_rate": 3.098166196460631e-05, + "loss": 0.3213, + "step": 29711 + }, + { + "epoch": 1.14, + "learning_rate": 3.098102077455758e-05, + "loss": 0.3465, + "step": 29712 + }, + { + "epoch": 1.14, + "learning_rate": 3.098037958450885e-05, + "loss": 0.418, + "step": 29713 + }, + { + "epoch": 1.14, + "learning_rate": 3.0979738394460114e-05, + "loss": 0.3438, + "step": 29714 + }, + { + "epoch": 1.14, + "learning_rate": 3.0979097204411386e-05, + "loss": 0.3611, + "step": 29715 + }, + { + "epoch": 1.14, + "learning_rate": 3.097845601436266e-05, + "loss": 0.3486, + "step": 29716 + }, + { + "epoch": 1.14, + "learning_rate": 3.097781482431393e-05, + "loss": 0.3593, + "step": 29717 + }, + { + "epoch": 1.14, + "learning_rate": 3.09771736342652e-05, + "loss": 0.4398, + "step": 29718 + }, + { + "epoch": 1.14, + "learning_rate": 3.0976532444216466e-05, + "loss": 0.4926, + "step": 29719 + }, + { + "epoch": 1.14, + "learning_rate": 3.097589125416774e-05, + "loss": 0.3225, + "step": 29720 + }, + { + "epoch": 1.14, + "learning_rate": 3.0975250064119e-05, + "loss": 0.3529, + "step": 29721 + }, + { + "epoch": 1.14, + "learning_rate": 3.0974608874070275e-05, + "loss": 0.3804, + "step": 29722 + }, + { + "epoch": 1.14, + "learning_rate": 3.0973967684021546e-05, + "loss": 0.3636, + "step": 29723 + }, + { + "epoch": 1.14, + "learning_rate": 3.097332649397282e-05, + "loss": 0.3085, + "step": 29724 + }, + { + "epoch": 1.14, + "learning_rate": 3.097268530392409e-05, + "loss": 0.4182, + "step": 29725 + }, + { + "epoch": 1.14, + "learning_rate": 3.0972044113875355e-05, + "loss": 0.5849, + "step": 29726 + }, + { + "epoch": 1.14, + "learning_rate": 3.0971402923826626e-05, + "loss": 0.3239, + "step": 29727 + }, + { + "epoch": 1.14, + "learning_rate": 3.097076173377789e-05, + "loss": 0.3112, + "step": 29728 + }, + { + "epoch": 1.14, + "learning_rate": 3.097012054372916e-05, + "loss": 0.3664, + "step": 29729 + }, + { + "epoch": 1.14, + "learning_rate": 3.096947935368043e-05, + "loss": 0.419, + "step": 29730 + }, + { + "epoch": 1.14, + "learning_rate": 3.09688381636317e-05, + "loss": 0.4793, + "step": 29731 + }, + { + "epoch": 1.14, + "learning_rate": 3.096819697358297e-05, + "loss": 0.5634, + "step": 29732 + }, + { + "epoch": 1.14, + "learning_rate": 3.096755578353424e-05, + "loss": 0.3025, + "step": 29733 + }, + { + "epoch": 1.14, + "learning_rate": 3.0966914593485515e-05, + "loss": 0.3523, + "step": 29734 + }, + { + "epoch": 1.14, + "learning_rate": 3.096627340343678e-05, + "loss": 0.5028, + "step": 29735 + }, + { + "epoch": 1.14, + "learning_rate": 3.096563221338805e-05, + "loss": 0.3653, + "step": 29736 + }, + { + "epoch": 1.14, + "learning_rate": 3.0964991023339316e-05, + "loss": 0.337, + "step": 29737 + }, + { + "epoch": 1.14, + "learning_rate": 3.096434983329059e-05, + "loss": 0.353, + "step": 29738 + }, + { + "epoch": 1.14, + "learning_rate": 3.096370864324186e-05, + "loss": 0.3906, + "step": 29739 + }, + { + "epoch": 1.14, + "learning_rate": 3.096306745319313e-05, + "loss": 0.5217, + "step": 29740 + }, + { + "epoch": 1.14, + "learning_rate": 3.0962426263144397e-05, + "loss": 0.41, + "step": 29741 + }, + { + "epoch": 1.14, + "learning_rate": 3.096178507309567e-05, + "loss": 0.3279, + "step": 29742 + }, + { + "epoch": 1.14, + "learning_rate": 3.096114388304694e-05, + "loss": 0.3606, + "step": 29743 + }, + { + "epoch": 1.14, + "learning_rate": 3.0960502692998205e-05, + "loss": 0.3138, + "step": 29744 + }, + { + "epoch": 1.14, + "learning_rate": 3.095986150294948e-05, + "loss": 0.4032, + "step": 29745 + }, + { + "epoch": 1.14, + "learning_rate": 3.095922031290074e-05, + "loss": 0.3763, + "step": 29746 + }, + { + "epoch": 1.14, + "learning_rate": 3.095857912285201e-05, + "loss": 0.3948, + "step": 29747 + }, + { + "epoch": 1.14, + "learning_rate": 3.0957937932803285e-05, + "loss": 0.396, + "step": 29748 + }, + { + "epoch": 1.14, + "learning_rate": 3.095729674275456e-05, + "loss": 0.2792, + "step": 29749 + }, + { + "epoch": 1.14, + "learning_rate": 3.095665555270582e-05, + "loss": 0.3751, + "step": 29750 + }, + { + "epoch": 1.14, + "learning_rate": 3.095601436265709e-05, + "loss": 0.3823, + "step": 29751 + }, + { + "epoch": 1.14, + "learning_rate": 3.0955373172608365e-05, + "loss": 0.4253, + "step": 29752 + }, + { + "epoch": 1.14, + "learning_rate": 3.095473198255963e-05, + "loss": 0.3853, + "step": 29753 + }, + { + "epoch": 1.14, + "learning_rate": 3.09540907925109e-05, + "loss": 0.3481, + "step": 29754 + }, + { + "epoch": 1.14, + "learning_rate": 3.095344960246217e-05, + "loss": 0.2453, + "step": 29755 + }, + { + "epoch": 1.14, + "learning_rate": 3.0952808412413445e-05, + "loss": 0.2697, + "step": 29756 + }, + { + "epoch": 1.14, + "learning_rate": 3.095216722236471e-05, + "loss": 0.4334, + "step": 29757 + }, + { + "epoch": 1.14, + "learning_rate": 3.095152603231598e-05, + "loss": 0.3574, + "step": 29758 + }, + { + "epoch": 1.14, + "learning_rate": 3.095088484226725e-05, + "loss": 0.5104, + "step": 29759 + }, + { + "epoch": 1.14, + "learning_rate": 3.095024365221852e-05, + "loss": 0.3518, + "step": 29760 + }, + { + "epoch": 1.14, + "learning_rate": 3.094960246216979e-05, + "loss": 0.4544, + "step": 29761 + }, + { + "epoch": 1.14, + "learning_rate": 3.0948961272121055e-05, + "loss": 0.2744, + "step": 29762 + }, + { + "epoch": 1.14, + "learning_rate": 3.094832008207233e-05, + "loss": 0.4095, + "step": 29763 + }, + { + "epoch": 1.14, + "learning_rate": 3.09476788920236e-05, + "loss": 0.3714, + "step": 29764 + }, + { + "epoch": 1.14, + "learning_rate": 3.094703770197487e-05, + "loss": 0.4838, + "step": 29765 + }, + { + "epoch": 1.14, + "learning_rate": 3.0946396511926135e-05, + "loss": 0.4473, + "step": 29766 + }, + { + "epoch": 1.14, + "learning_rate": 3.094575532187741e-05, + "loss": 0.4319, + "step": 29767 + }, + { + "epoch": 1.14, + "learning_rate": 3.094511413182867e-05, + "loss": 0.391, + "step": 29768 + }, + { + "epoch": 1.14, + "learning_rate": 3.0944472941779944e-05, + "loss": 0.4795, + "step": 29769 + }, + { + "epoch": 1.14, + "learning_rate": 3.0943831751731215e-05, + "loss": 0.4433, + "step": 29770 + }, + { + "epoch": 1.14, + "learning_rate": 3.094319056168248e-05, + "loss": 0.2989, + "step": 29771 + }, + { + "epoch": 1.14, + "learning_rate": 3.094254937163376e-05, + "loss": 0.396, + "step": 29772 + }, + { + "epoch": 1.14, + "learning_rate": 3.0941908181585024e-05, + "loss": 0.314, + "step": 29773 + }, + { + "epoch": 1.14, + "learning_rate": 3.0941266991536295e-05, + "loss": 0.3604, + "step": 29774 + }, + { + "epoch": 1.14, + "learning_rate": 3.094062580148756e-05, + "loss": 0.3999, + "step": 29775 + }, + { + "epoch": 1.14, + "learning_rate": 3.093998461143883e-05, + "loss": 0.2859, + "step": 29776 + }, + { + "epoch": 1.14, + "learning_rate": 3.09393434213901e-05, + "loss": 0.3145, + "step": 29777 + }, + { + "epoch": 1.14, + "learning_rate": 3.093870223134137e-05, + "loss": 0.4219, + "step": 29778 + }, + { + "epoch": 1.14, + "learning_rate": 3.093806104129264e-05, + "loss": 0.4116, + "step": 29779 + }, + { + "epoch": 1.14, + "learning_rate": 3.093741985124391e-05, + "loss": 0.3664, + "step": 29780 + }, + { + "epoch": 1.14, + "learning_rate": 3.0936778661195184e-05, + "loss": 0.4455, + "step": 29781 + }, + { + "epoch": 1.15, + "learning_rate": 3.093613747114645e-05, + "loss": 0.4049, + "step": 29782 + }, + { + "epoch": 1.15, + "learning_rate": 3.093549628109772e-05, + "loss": 0.3884, + "step": 29783 + }, + { + "epoch": 1.15, + "learning_rate": 3.0934855091048985e-05, + "loss": 0.4309, + "step": 29784 + }, + { + "epoch": 1.15, + "learning_rate": 3.093421390100026e-05, + "loss": 0.3347, + "step": 29785 + }, + { + "epoch": 1.15, + "learning_rate": 3.093357271095152e-05, + "loss": 0.3535, + "step": 29786 + }, + { + "epoch": 1.15, + "learning_rate": 3.0932931520902794e-05, + "loss": 0.3651, + "step": 29787 + }, + { + "epoch": 1.15, + "learning_rate": 3.093229033085407e-05, + "loss": 0.4036, + "step": 29788 + }, + { + "epoch": 1.15, + "learning_rate": 3.093164914080534e-05, + "loss": 0.424, + "step": 29789 + }, + { + "epoch": 1.15, + "learning_rate": 3.093100795075661e-05, + "loss": 0.4718, + "step": 29790 + }, + { + "epoch": 1.15, + "learning_rate": 3.0930366760707874e-05, + "loss": 0.3956, + "step": 29791 + }, + { + "epoch": 1.15, + "learning_rate": 3.0929725570659146e-05, + "loss": 0.3556, + "step": 29792 + }, + { + "epoch": 1.15, + "learning_rate": 3.092908438061041e-05, + "loss": 0.3791, + "step": 29793 + }, + { + "epoch": 1.15, + "learning_rate": 3.092844319056168e-05, + "loss": 0.319, + "step": 29794 + }, + { + "epoch": 1.15, + "learning_rate": 3.0927802000512954e-05, + "loss": 0.3635, + "step": 29795 + }, + { + "epoch": 1.15, + "learning_rate": 3.0927160810464226e-05, + "loss": 0.438, + "step": 29796 + }, + { + "epoch": 1.15, + "learning_rate": 3.09265196204155e-05, + "loss": 0.4221, + "step": 29797 + }, + { + "epoch": 1.15, + "learning_rate": 3.092587843036676e-05, + "loss": 0.4197, + "step": 29798 + }, + { + "epoch": 1.15, + "learning_rate": 3.0925237240318034e-05, + "loss": 0.4447, + "step": 29799 + }, + { + "epoch": 1.15, + "learning_rate": 3.09245960502693e-05, + "loss": 0.4064, + "step": 29800 + }, + { + "epoch": 1.15, + "learning_rate": 3.092395486022057e-05, + "loss": 0.4077, + "step": 29801 + }, + { + "epoch": 1.15, + "learning_rate": 3.0923313670171836e-05, + "loss": 0.42, + "step": 29802 + }, + { + "epoch": 1.15, + "learning_rate": 3.092267248012311e-05, + "loss": 0.4208, + "step": 29803 + }, + { + "epoch": 1.15, + "learning_rate": 3.092203129007438e-05, + "loss": 0.4974, + "step": 29804 + }, + { + "epoch": 1.15, + "learning_rate": 3.092139010002565e-05, + "loss": 0.4954, + "step": 29805 + }, + { + "epoch": 1.15, + "learning_rate": 3.092074890997692e-05, + "loss": 0.3786, + "step": 29806 + }, + { + "epoch": 1.15, + "learning_rate": 3.092010771992819e-05, + "loss": 0.4116, + "step": 29807 + }, + { + "epoch": 1.15, + "learning_rate": 3.091946652987946e-05, + "loss": 0.3727, + "step": 29808 + }, + { + "epoch": 1.15, + "learning_rate": 3.0918825339830724e-05, + "loss": 0.3916, + "step": 29809 + }, + { + "epoch": 1.15, + "learning_rate": 3.0918184149781996e-05, + "loss": 0.3386, + "step": 29810 + }, + { + "epoch": 1.15, + "learning_rate": 3.091754295973327e-05, + "loss": 0.3913, + "step": 29811 + }, + { + "epoch": 1.15, + "learning_rate": 3.091690176968454e-05, + "loss": 0.32, + "step": 29812 + }, + { + "epoch": 1.15, + "learning_rate": 3.0916260579635804e-05, + "loss": 0.4482, + "step": 29813 + }, + { + "epoch": 1.15, + "learning_rate": 3.0915619389587076e-05, + "loss": 0.3187, + "step": 29814 + }, + { + "epoch": 1.15, + "learning_rate": 3.091497819953835e-05, + "loss": 0.4188, + "step": 29815 + }, + { + "epoch": 1.15, + "learning_rate": 3.091433700948961e-05, + "loss": 0.4403, + "step": 29816 + }, + { + "epoch": 1.15, + "learning_rate": 3.0913695819440884e-05, + "loss": 0.5049, + "step": 29817 + }, + { + "epoch": 1.15, + "learning_rate": 3.091305462939215e-05, + "loss": 0.4611, + "step": 29818 + }, + { + "epoch": 1.15, + "learning_rate": 3.091241343934342e-05, + "loss": 0.3221, + "step": 29819 + }, + { + "epoch": 1.15, + "learning_rate": 3.091177224929469e-05, + "loss": 0.537, + "step": 29820 + }, + { + "epoch": 1.15, + "learning_rate": 3.0911131059245964e-05, + "loss": 0.3341, + "step": 29821 + }, + { + "epoch": 1.15, + "learning_rate": 3.091048986919723e-05, + "loss": 0.3389, + "step": 29822 + }, + { + "epoch": 1.15, + "learning_rate": 3.09098486791485e-05, + "loss": 0.4253, + "step": 29823 + }, + { + "epoch": 1.15, + "learning_rate": 3.090920748909977e-05, + "loss": 0.3096, + "step": 29824 + }, + { + "epoch": 1.15, + "learning_rate": 3.090856629905104e-05, + "loss": 0.4314, + "step": 29825 + }, + { + "epoch": 1.15, + "learning_rate": 3.090792510900231e-05, + "loss": 0.5381, + "step": 29826 + }, + { + "epoch": 1.15, + "learning_rate": 3.090728391895358e-05, + "loss": 0.3288, + "step": 29827 + }, + { + "epoch": 1.15, + "learning_rate": 3.090664272890485e-05, + "loss": 0.3584, + "step": 29828 + }, + { + "epoch": 1.15, + "learning_rate": 3.090600153885612e-05, + "loss": 0.4985, + "step": 29829 + }, + { + "epoch": 1.15, + "learning_rate": 3.090536034880739e-05, + "loss": 0.4979, + "step": 29830 + }, + { + "epoch": 1.15, + "learning_rate": 3.0904719158758655e-05, + "loss": 0.4207, + "step": 29831 + }, + { + "epoch": 1.15, + "learning_rate": 3.0904077968709926e-05, + "loss": 0.4062, + "step": 29832 + }, + { + "epoch": 1.15, + "learning_rate": 3.09034367786612e-05, + "loss": 0.4702, + "step": 29833 + }, + { + "epoch": 1.15, + "learning_rate": 3.090279558861246e-05, + "loss": 0.4473, + "step": 29834 + }, + { + "epoch": 1.15, + "learning_rate": 3.0902154398563735e-05, + "loss": 0.4047, + "step": 29835 + }, + { + "epoch": 1.15, + "learning_rate": 3.0901513208515006e-05, + "loss": 0.3909, + "step": 29836 + }, + { + "epoch": 1.15, + "learning_rate": 3.090087201846628e-05, + "loss": 0.4359, + "step": 29837 + }, + { + "epoch": 1.15, + "learning_rate": 3.090023082841754e-05, + "loss": 0.3995, + "step": 29838 + }, + { + "epoch": 1.15, + "learning_rate": 3.0899589638368815e-05, + "loss": 0.3047, + "step": 29839 + }, + { + "epoch": 1.15, + "learning_rate": 3.089894844832008e-05, + "loss": 0.5164, + "step": 29840 + }, + { + "epoch": 1.15, + "learning_rate": 3.089830725827135e-05, + "loss": 0.3318, + "step": 29841 + }, + { + "epoch": 1.15, + "learning_rate": 3.089766606822262e-05, + "loss": 0.3668, + "step": 29842 + }, + { + "epoch": 1.15, + "learning_rate": 3.089702487817389e-05, + "loss": 0.5753, + "step": 29843 + }, + { + "epoch": 1.15, + "learning_rate": 3.0896383688125167e-05, + "loss": 0.3616, + "step": 29844 + }, + { + "epoch": 1.15, + "learning_rate": 3.089574249807643e-05, + "loss": 0.4695, + "step": 29845 + }, + { + "epoch": 1.15, + "learning_rate": 3.08951013080277e-05, + "loss": 0.5447, + "step": 29846 + }, + { + "epoch": 1.15, + "learning_rate": 3.089446011797897e-05, + "loss": 0.4416, + "step": 29847 + }, + { + "epoch": 1.15, + "learning_rate": 3.089381892793024e-05, + "loss": 0.3588, + "step": 29848 + }, + { + "epoch": 1.15, + "learning_rate": 3.0893177737881505e-05, + "loss": 0.4844, + "step": 29849 + }, + { + "epoch": 1.15, + "learning_rate": 3.0892536547832776e-05, + "loss": 0.3614, + "step": 29850 + }, + { + "epoch": 1.15, + "learning_rate": 3.089189535778405e-05, + "loss": 0.4999, + "step": 29851 + }, + { + "epoch": 1.15, + "learning_rate": 3.089125416773532e-05, + "loss": 0.449, + "step": 29852 + }, + { + "epoch": 1.15, + "learning_rate": 3.089061297768659e-05, + "loss": 0.4325, + "step": 29853 + }, + { + "epoch": 1.15, + "learning_rate": 3.0889971787637857e-05, + "loss": 0.3414, + "step": 29854 + }, + { + "epoch": 1.15, + "learning_rate": 3.088933059758913e-05, + "loss": 0.4318, + "step": 29855 + }, + { + "epoch": 1.15, + "learning_rate": 3.088868940754039e-05, + "loss": 0.3317, + "step": 29856 + }, + { + "epoch": 1.15, + "learning_rate": 3.0888048217491665e-05, + "loss": 0.479, + "step": 29857 + }, + { + "epoch": 1.15, + "learning_rate": 3.088740702744293e-05, + "loss": 0.3105, + "step": 29858 + }, + { + "epoch": 1.15, + "learning_rate": 3.08867658373942e-05, + "loss": 0.3879, + "step": 29859 + }, + { + "epoch": 1.15, + "learning_rate": 3.088612464734548e-05, + "loss": 0.4577, + "step": 29860 + }, + { + "epoch": 1.15, + "learning_rate": 3.0885483457296745e-05, + "loss": 0.3904, + "step": 29861 + }, + { + "epoch": 1.15, + "learning_rate": 3.088484226724802e-05, + "loss": 0.4275, + "step": 29862 + }, + { + "epoch": 1.15, + "learning_rate": 3.088420107719928e-05, + "loss": 0.4305, + "step": 29863 + }, + { + "epoch": 1.15, + "learning_rate": 3.0883559887150553e-05, + "loss": 0.4344, + "step": 29864 + }, + { + "epoch": 1.15, + "learning_rate": 3.088291869710182e-05, + "loss": 0.447, + "step": 29865 + }, + { + "epoch": 1.15, + "learning_rate": 3.088227750705309e-05, + "loss": 0.4807, + "step": 29866 + }, + { + "epoch": 1.15, + "learning_rate": 3.088163631700436e-05, + "loss": 0.3745, + "step": 29867 + }, + { + "epoch": 1.15, + "learning_rate": 3.0880995126955634e-05, + "loss": 0.4507, + "step": 29868 + }, + { + "epoch": 1.15, + "learning_rate": 3.0880353936906905e-05, + "loss": 0.4045, + "step": 29869 + }, + { + "epoch": 1.15, + "learning_rate": 3.087971274685817e-05, + "loss": 0.4382, + "step": 29870 + }, + { + "epoch": 1.15, + "learning_rate": 3.087907155680944e-05, + "loss": 0.2741, + "step": 29871 + }, + { + "epoch": 1.15, + "learning_rate": 3.087843036676071e-05, + "loss": 0.4033, + "step": 29872 + }, + { + "epoch": 1.15, + "learning_rate": 3.087778917671198e-05, + "loss": 0.3551, + "step": 29873 + }, + { + "epoch": 1.15, + "learning_rate": 3.0877147986663243e-05, + "loss": 0.3249, + "step": 29874 + }, + { + "epoch": 1.15, + "learning_rate": 3.0876506796614515e-05, + "loss": 0.3017, + "step": 29875 + }, + { + "epoch": 1.15, + "learning_rate": 3.087586560656579e-05, + "loss": 0.3794, + "step": 29876 + }, + { + "epoch": 1.15, + "learning_rate": 3.087522441651706e-05, + "loss": 0.4045, + "step": 29877 + }, + { + "epoch": 1.15, + "learning_rate": 3.087458322646833e-05, + "loss": 0.431, + "step": 29878 + }, + { + "epoch": 1.15, + "learning_rate": 3.0873942036419595e-05, + "loss": 0.3777, + "step": 29879 + }, + { + "epoch": 1.15, + "learning_rate": 3.087330084637087e-05, + "loss": 0.3636, + "step": 29880 + }, + { + "epoch": 1.15, + "learning_rate": 3.087265965632213e-05, + "loss": 0.4594, + "step": 29881 + }, + { + "epoch": 1.15, + "learning_rate": 3.0872018466273404e-05, + "loss": 0.5277, + "step": 29882 + }, + { + "epoch": 1.15, + "learning_rate": 3.0871377276224675e-05, + "loss": 0.5416, + "step": 29883 + }, + { + "epoch": 1.15, + "learning_rate": 3.087073608617595e-05, + "loss": 0.3803, + "step": 29884 + }, + { + "epoch": 1.15, + "learning_rate": 3.087009489612722e-05, + "loss": 0.3428, + "step": 29885 + }, + { + "epoch": 1.15, + "learning_rate": 3.0869453706078484e-05, + "loss": 0.4014, + "step": 29886 + }, + { + "epoch": 1.15, + "learning_rate": 3.0868812516029755e-05, + "loss": 0.3281, + "step": 29887 + }, + { + "epoch": 1.15, + "learning_rate": 3.086817132598102e-05, + "loss": 0.5446, + "step": 29888 + }, + { + "epoch": 1.15, + "learning_rate": 3.086753013593229e-05, + "loss": 0.3975, + "step": 29889 + }, + { + "epoch": 1.15, + "learning_rate": 3.086688894588356e-05, + "loss": 0.2762, + "step": 29890 + }, + { + "epoch": 1.15, + "learning_rate": 3.086624775583483e-05, + "loss": 0.5163, + "step": 29891 + }, + { + "epoch": 1.15, + "learning_rate": 3.08656065657861e-05, + "loss": 0.4639, + "step": 29892 + }, + { + "epoch": 1.15, + "learning_rate": 3.086496537573737e-05, + "loss": 0.4979, + "step": 29893 + }, + { + "epoch": 1.15, + "learning_rate": 3.0864324185688644e-05, + "loss": 0.5518, + "step": 29894 + }, + { + "epoch": 1.15, + "learning_rate": 3.086368299563991e-05, + "loss": 0.4156, + "step": 29895 + }, + { + "epoch": 1.15, + "learning_rate": 3.086304180559118e-05, + "loss": 0.3646, + "step": 29896 + }, + { + "epoch": 1.15, + "learning_rate": 3.0862400615542446e-05, + "loss": 0.3112, + "step": 29897 + }, + { + "epoch": 1.15, + "learning_rate": 3.086175942549372e-05, + "loss": 0.4146, + "step": 29898 + }, + { + "epoch": 1.15, + "learning_rate": 3.086111823544499e-05, + "loss": 0.3963, + "step": 29899 + }, + { + "epoch": 1.15, + "learning_rate": 3.086047704539626e-05, + "loss": 0.4008, + "step": 29900 + }, + { + "epoch": 1.15, + "learning_rate": 3.0859835855347526e-05, + "loss": 0.3511, + "step": 29901 + }, + { + "epoch": 1.15, + "learning_rate": 3.08591946652988e-05, + "loss": 0.4489, + "step": 29902 + }, + { + "epoch": 1.15, + "learning_rate": 3.085855347525007e-05, + "loss": 0.314, + "step": 29903 + }, + { + "epoch": 1.15, + "learning_rate": 3.0857912285201334e-05, + "loss": 0.4608, + "step": 29904 + }, + { + "epoch": 1.15, + "learning_rate": 3.0857271095152606e-05, + "loss": 0.4558, + "step": 29905 + }, + { + "epoch": 1.15, + "learning_rate": 3.085662990510387e-05, + "loss": 0.332, + "step": 29906 + }, + { + "epoch": 1.15, + "learning_rate": 3.085598871505514e-05, + "loss": 0.3576, + "step": 29907 + }, + { + "epoch": 1.15, + "learning_rate": 3.0855347525006414e-05, + "loss": 0.2953, + "step": 29908 + }, + { + "epoch": 1.15, + "learning_rate": 3.0854706334957686e-05, + "loss": 0.3564, + "step": 29909 + }, + { + "epoch": 1.15, + "learning_rate": 3.085406514490895e-05, + "loss": 0.3708, + "step": 29910 + }, + { + "epoch": 1.15, + "learning_rate": 3.085342395486022e-05, + "loss": 0.391, + "step": 29911 + }, + { + "epoch": 1.15, + "learning_rate": 3.0852782764811494e-05, + "loss": 0.4528, + "step": 29912 + }, + { + "epoch": 1.15, + "learning_rate": 3.085214157476276e-05, + "loss": 0.3753, + "step": 29913 + }, + { + "epoch": 1.15, + "learning_rate": 3.085150038471403e-05, + "loss": 0.3972, + "step": 29914 + }, + { + "epoch": 1.15, + "learning_rate": 3.08508591946653e-05, + "loss": 0.3182, + "step": 29915 + }, + { + "epoch": 1.15, + "learning_rate": 3.0850218004616574e-05, + "loss": 0.3596, + "step": 29916 + }, + { + "epoch": 1.15, + "learning_rate": 3.084957681456784e-05, + "loss": 0.4822, + "step": 29917 + }, + { + "epoch": 1.15, + "learning_rate": 3.084893562451911e-05, + "loss": 0.3995, + "step": 29918 + }, + { + "epoch": 1.15, + "learning_rate": 3.0848294434470376e-05, + "loss": 0.4002, + "step": 29919 + }, + { + "epoch": 1.15, + "learning_rate": 3.084765324442165e-05, + "loss": 0.3879, + "step": 29920 + }, + { + "epoch": 1.15, + "learning_rate": 3.084701205437292e-05, + "loss": 0.277, + "step": 29921 + }, + { + "epoch": 1.15, + "learning_rate": 3.0846370864324184e-05, + "loss": 0.4175, + "step": 29922 + }, + { + "epoch": 1.15, + "learning_rate": 3.0845729674275456e-05, + "loss": 0.4024, + "step": 29923 + }, + { + "epoch": 1.15, + "learning_rate": 3.084508848422673e-05, + "loss": 0.5271, + "step": 29924 + }, + { + "epoch": 1.15, + "learning_rate": 3.0844447294178e-05, + "loss": 0.4062, + "step": 29925 + }, + { + "epoch": 1.15, + "learning_rate": 3.0843806104129264e-05, + "loss": 0.2777, + "step": 29926 + }, + { + "epoch": 1.15, + "learning_rate": 3.0843164914080536e-05, + "loss": 0.3575, + "step": 29927 + }, + { + "epoch": 1.15, + "learning_rate": 3.08425237240318e-05, + "loss": 0.48, + "step": 29928 + }, + { + "epoch": 1.15, + "learning_rate": 3.084188253398307e-05, + "loss": 0.3588, + "step": 29929 + }, + { + "epoch": 1.15, + "learning_rate": 3.0841241343934344e-05, + "loss": 0.3331, + "step": 29930 + }, + { + "epoch": 1.15, + "learning_rate": 3.084060015388561e-05, + "loss": 0.433, + "step": 29931 + }, + { + "epoch": 1.15, + "learning_rate": 3.083995896383689e-05, + "loss": 0.3066, + "step": 29932 + }, + { + "epoch": 1.15, + "learning_rate": 3.083931777378815e-05, + "loss": 0.4397, + "step": 29933 + }, + { + "epoch": 1.15, + "learning_rate": 3.0838676583739425e-05, + "loss": 0.4029, + "step": 29934 + }, + { + "epoch": 1.15, + "learning_rate": 3.083803539369069e-05, + "loss": 0.3917, + "step": 29935 + }, + { + "epoch": 1.15, + "learning_rate": 3.083739420364196e-05, + "loss": 0.445, + "step": 29936 + }, + { + "epoch": 1.15, + "learning_rate": 3.0836753013593226e-05, + "loss": 0.2627, + "step": 29937 + }, + { + "epoch": 1.15, + "learning_rate": 3.08361118235445e-05, + "loss": 0.3108, + "step": 29938 + }, + { + "epoch": 1.15, + "learning_rate": 3.083547063349577e-05, + "loss": 0.3376, + "step": 29939 + }, + { + "epoch": 1.15, + "learning_rate": 3.083482944344704e-05, + "loss": 0.3533, + "step": 29940 + }, + { + "epoch": 1.15, + "learning_rate": 3.083418825339831e-05, + "loss": 0.373, + "step": 29941 + }, + { + "epoch": 1.15, + "learning_rate": 3.083354706334958e-05, + "loss": 0.3653, + "step": 29942 + }, + { + "epoch": 1.15, + "learning_rate": 3.083290587330085e-05, + "loss": 0.3153, + "step": 29943 + }, + { + "epoch": 1.15, + "learning_rate": 3.0832264683252115e-05, + "loss": 0.4178, + "step": 29944 + }, + { + "epoch": 1.15, + "learning_rate": 3.0831623493203386e-05, + "loss": 0.3022, + "step": 29945 + }, + { + "epoch": 1.15, + "learning_rate": 3.083098230315465e-05, + "loss": 0.338, + "step": 29946 + }, + { + "epoch": 1.15, + "learning_rate": 3.083034111310592e-05, + "loss": 0.3735, + "step": 29947 + }, + { + "epoch": 1.15, + "learning_rate": 3.08296999230572e-05, + "loss": 0.3668, + "step": 29948 + }, + { + "epoch": 1.15, + "learning_rate": 3.0829058733008466e-05, + "loss": 0.39, + "step": 29949 + }, + { + "epoch": 1.15, + "learning_rate": 3.082841754295974e-05, + "loss": 0.3079, + "step": 29950 + }, + { + "epoch": 1.15, + "learning_rate": 3.0827776352911e-05, + "loss": 0.2704, + "step": 29951 + }, + { + "epoch": 1.15, + "learning_rate": 3.0827135162862275e-05, + "loss": 0.3639, + "step": 29952 + }, + { + "epoch": 1.15, + "learning_rate": 3.082649397281354e-05, + "loss": 0.5142, + "step": 29953 + }, + { + "epoch": 1.15, + "learning_rate": 3.082585278276481e-05, + "loss": 0.354, + "step": 29954 + }, + { + "epoch": 1.15, + "learning_rate": 3.082521159271608e-05, + "loss": 0.3424, + "step": 29955 + }, + { + "epoch": 1.15, + "learning_rate": 3.0824570402667355e-05, + "loss": 0.4913, + "step": 29956 + }, + { + "epoch": 1.15, + "learning_rate": 3.0823929212618627e-05, + "loss": 0.4416, + "step": 29957 + }, + { + "epoch": 1.15, + "learning_rate": 3.082328802256989e-05, + "loss": 0.2803, + "step": 29958 + }, + { + "epoch": 1.15, + "learning_rate": 3.082264683252116e-05, + "loss": 0.3834, + "step": 29959 + }, + { + "epoch": 1.15, + "learning_rate": 3.082200564247243e-05, + "loss": 0.3604, + "step": 29960 + }, + { + "epoch": 1.15, + "learning_rate": 3.08213644524237e-05, + "loss": 0.3852, + "step": 29961 + }, + { + "epoch": 1.15, + "learning_rate": 3.0820723262374965e-05, + "loss": 0.341, + "step": 29962 + }, + { + "epoch": 1.15, + "learning_rate": 3.0820082072326237e-05, + "loss": 0.331, + "step": 29963 + }, + { + "epoch": 1.15, + "learning_rate": 3.081944088227751e-05, + "loss": 0.4784, + "step": 29964 + }, + { + "epoch": 1.15, + "learning_rate": 3.081879969222878e-05, + "loss": 0.439, + "step": 29965 + }, + { + "epoch": 1.15, + "learning_rate": 3.081815850218005e-05, + "loss": 0.4055, + "step": 29966 + }, + { + "epoch": 1.15, + "learning_rate": 3.081751731213132e-05, + "loss": 0.4051, + "step": 29967 + }, + { + "epoch": 1.15, + "learning_rate": 3.081687612208259e-05, + "loss": 0.3754, + "step": 29968 + }, + { + "epoch": 1.15, + "learning_rate": 3.081623493203385e-05, + "loss": 0.4829, + "step": 29969 + }, + { + "epoch": 1.15, + "learning_rate": 3.0815593741985125e-05, + "loss": 0.2926, + "step": 29970 + }, + { + "epoch": 1.15, + "learning_rate": 3.08149525519364e-05, + "loss": 0.4023, + "step": 29971 + }, + { + "epoch": 1.15, + "learning_rate": 3.081431136188767e-05, + "loss": 0.4573, + "step": 29972 + }, + { + "epoch": 1.15, + "learning_rate": 3.081367017183893e-05, + "loss": 0.4809, + "step": 29973 + }, + { + "epoch": 1.15, + "learning_rate": 3.0813028981790205e-05, + "loss": 0.4528, + "step": 29974 + }, + { + "epoch": 1.15, + "learning_rate": 3.081238779174148e-05, + "loss": 0.5156, + "step": 29975 + }, + { + "epoch": 1.15, + "learning_rate": 3.081174660169274e-05, + "loss": 0.4988, + "step": 29976 + }, + { + "epoch": 1.15, + "learning_rate": 3.0811105411644013e-05, + "loss": 0.4056, + "step": 29977 + }, + { + "epoch": 1.15, + "learning_rate": 3.081046422159528e-05, + "loss": 0.466, + "step": 29978 + }, + { + "epoch": 1.15, + "learning_rate": 3.080982303154655e-05, + "loss": 0.5155, + "step": 29979 + }, + { + "epoch": 1.15, + "learning_rate": 3.080918184149782e-05, + "loss": 0.4535, + "step": 29980 + }, + { + "epoch": 1.15, + "learning_rate": 3.0808540651449094e-05, + "loss": 0.4415, + "step": 29981 + }, + { + "epoch": 1.15, + "learning_rate": 3.080789946140036e-05, + "loss": 0.4192, + "step": 29982 + }, + { + "epoch": 1.15, + "learning_rate": 3.080725827135163e-05, + "loss": 0.3478, + "step": 29983 + }, + { + "epoch": 1.15, + "learning_rate": 3.08066170813029e-05, + "loss": 0.3946, + "step": 29984 + }, + { + "epoch": 1.15, + "learning_rate": 3.080597589125417e-05, + "loss": 0.4223, + "step": 29985 + }, + { + "epoch": 1.15, + "learning_rate": 3.080533470120544e-05, + "loss": 0.3191, + "step": 29986 + }, + { + "epoch": 1.15, + "learning_rate": 3.080469351115671e-05, + "loss": 0.3853, + "step": 29987 + }, + { + "epoch": 1.15, + "learning_rate": 3.080405232110798e-05, + "loss": 0.5277, + "step": 29988 + }, + { + "epoch": 1.15, + "learning_rate": 3.080341113105925e-05, + "loss": 0.416, + "step": 29989 + }, + { + "epoch": 1.15, + "learning_rate": 3.080276994101052e-05, + "loss": 0.5592, + "step": 29990 + }, + { + "epoch": 1.15, + "learning_rate": 3.0802128750961784e-05, + "loss": 0.373, + "step": 29991 + }, + { + "epoch": 1.15, + "learning_rate": 3.0801487560913055e-05, + "loss": 0.5196, + "step": 29992 + }, + { + "epoch": 1.15, + "learning_rate": 3.080084637086433e-05, + "loss": 0.305, + "step": 29993 + }, + { + "epoch": 1.15, + "learning_rate": 3.080020518081559e-05, + "loss": 0.3716, + "step": 29994 + }, + { + "epoch": 1.15, + "learning_rate": 3.0799563990766864e-05, + "loss": 0.3189, + "step": 29995 + }, + { + "epoch": 1.15, + "learning_rate": 3.0798922800718135e-05, + "loss": 0.426, + "step": 29996 + }, + { + "epoch": 1.15, + "learning_rate": 3.079828161066941e-05, + "loss": 0.4364, + "step": 29997 + }, + { + "epoch": 1.15, + "learning_rate": 3.079764042062067e-05, + "loss": 0.3964, + "step": 29998 + }, + { + "epoch": 1.15, + "learning_rate": 3.0796999230571944e-05, + "loss": 0.5173, + "step": 29999 + }, + { + "epoch": 1.15, + "learning_rate": 3.079635804052321e-05, + "loss": 0.4289, + "step": 30000 + }, + { + "epoch": 1.15, + "learning_rate": 3.079571685047448e-05, + "loss": 0.3711, + "step": 30001 + }, + { + "epoch": 1.15, + "learning_rate": 3.079507566042575e-05, + "loss": 0.4422, + "step": 30002 + }, + { + "epoch": 1.15, + "learning_rate": 3.079443447037702e-05, + "loss": 0.3737, + "step": 30003 + }, + { + "epoch": 1.15, + "learning_rate": 3.0793793280328296e-05, + "loss": 0.4534, + "step": 30004 + }, + { + "epoch": 1.15, + "learning_rate": 3.079315209027956e-05, + "loss": 0.327, + "step": 30005 + }, + { + "epoch": 1.15, + "learning_rate": 3.079251090023083e-05, + "loss": 0.2824, + "step": 30006 + }, + { + "epoch": 1.15, + "learning_rate": 3.07918697101821e-05, + "loss": 0.3548, + "step": 30007 + }, + { + "epoch": 1.15, + "learning_rate": 3.079122852013337e-05, + "loss": 0.3138, + "step": 30008 + }, + { + "epoch": 1.15, + "learning_rate": 3.0790587330084634e-05, + "loss": 0.406, + "step": 30009 + }, + { + "epoch": 1.15, + "learning_rate": 3.0789946140035906e-05, + "loss": 0.4294, + "step": 30010 + }, + { + "epoch": 1.15, + "learning_rate": 3.078930494998718e-05, + "loss": 0.4665, + "step": 30011 + }, + { + "epoch": 1.15, + "learning_rate": 3.078866375993845e-05, + "loss": 0.2824, + "step": 30012 + }, + { + "epoch": 1.15, + "learning_rate": 3.078802256988972e-05, + "loss": 0.3551, + "step": 30013 + }, + { + "epoch": 1.15, + "learning_rate": 3.0787381379840986e-05, + "loss": 0.4658, + "step": 30014 + }, + { + "epoch": 1.15, + "learning_rate": 3.078674018979226e-05, + "loss": 0.4154, + "step": 30015 + }, + { + "epoch": 1.15, + "learning_rate": 3.078609899974352e-05, + "loss": 0.4854, + "step": 30016 + }, + { + "epoch": 1.15, + "learning_rate": 3.0785457809694794e-05, + "loss": 0.4073, + "step": 30017 + }, + { + "epoch": 1.15, + "learning_rate": 3.078481661964606e-05, + "loss": 0.3632, + "step": 30018 + }, + { + "epoch": 1.15, + "learning_rate": 3.078417542959733e-05, + "loss": 0.4618, + "step": 30019 + }, + { + "epoch": 1.15, + "learning_rate": 3.078353423954861e-05, + "loss": 0.3335, + "step": 30020 + }, + { + "epoch": 1.15, + "learning_rate": 3.0782893049499874e-05, + "loss": 0.284, + "step": 30021 + }, + { + "epoch": 1.15, + "learning_rate": 3.0782251859451146e-05, + "loss": 0.3166, + "step": 30022 + }, + { + "epoch": 1.15, + "learning_rate": 3.078161066940241e-05, + "loss": 0.3933, + "step": 30023 + }, + { + "epoch": 1.15, + "learning_rate": 3.078096947935368e-05, + "loss": 0.3724, + "step": 30024 + }, + { + "epoch": 1.15, + "learning_rate": 3.078032828930495e-05, + "loss": 0.4156, + "step": 30025 + }, + { + "epoch": 1.15, + "learning_rate": 3.077968709925622e-05, + "loss": 0.4051, + "step": 30026 + }, + { + "epoch": 1.15, + "learning_rate": 3.077904590920749e-05, + "loss": 0.3368, + "step": 30027 + }, + { + "epoch": 1.15, + "learning_rate": 3.077840471915876e-05, + "loss": 0.4457, + "step": 30028 + }, + { + "epoch": 1.15, + "learning_rate": 3.0777763529110034e-05, + "loss": 0.4333, + "step": 30029 + }, + { + "epoch": 1.15, + "learning_rate": 3.07771223390613e-05, + "loss": 0.3571, + "step": 30030 + }, + { + "epoch": 1.15, + "learning_rate": 3.077648114901257e-05, + "loss": 0.3681, + "step": 30031 + }, + { + "epoch": 1.15, + "learning_rate": 3.0775839958963836e-05, + "loss": 0.5803, + "step": 30032 + }, + { + "epoch": 1.15, + "learning_rate": 3.077519876891511e-05, + "loss": 0.4385, + "step": 30033 + }, + { + "epoch": 1.15, + "learning_rate": 3.077455757886637e-05, + "loss": 0.5593, + "step": 30034 + }, + { + "epoch": 1.15, + "learning_rate": 3.0773916388817644e-05, + "loss": 0.4533, + "step": 30035 + }, + { + "epoch": 1.15, + "learning_rate": 3.0773275198768916e-05, + "loss": 0.4234, + "step": 30036 + }, + { + "epoch": 1.15, + "learning_rate": 3.077263400872019e-05, + "loss": 0.4537, + "step": 30037 + }, + { + "epoch": 1.15, + "learning_rate": 3.077199281867146e-05, + "loss": 0.2328, + "step": 30038 + }, + { + "epoch": 1.15, + "learning_rate": 3.0771351628622724e-05, + "loss": 0.3722, + "step": 30039 + }, + { + "epoch": 1.15, + "learning_rate": 3.0770710438573996e-05, + "loss": 0.4405, + "step": 30040 + }, + { + "epoch": 1.15, + "learning_rate": 3.077006924852526e-05, + "loss": 0.3699, + "step": 30041 + }, + { + "epoch": 1.16, + "learning_rate": 3.076942805847653e-05, + "loss": 0.4338, + "step": 30042 + }, + { + "epoch": 1.16, + "learning_rate": 3.0768786868427804e-05, + "loss": 0.347, + "step": 30043 + }, + { + "epoch": 1.16, + "learning_rate": 3.0768145678379076e-05, + "loss": 0.3663, + "step": 30044 + }, + { + "epoch": 1.16, + "learning_rate": 3.076750448833034e-05, + "loss": 0.3368, + "step": 30045 + }, + { + "epoch": 1.16, + "learning_rate": 3.076686329828161e-05, + "loss": 0.4644, + "step": 30046 + }, + { + "epoch": 1.16, + "learning_rate": 3.0766222108232885e-05, + "loss": 0.3588, + "step": 30047 + }, + { + "epoch": 1.16, + "learning_rate": 3.076558091818415e-05, + "loss": 0.3165, + "step": 30048 + }, + { + "epoch": 1.16, + "learning_rate": 3.076493972813542e-05, + "loss": 0.4639, + "step": 30049 + }, + { + "epoch": 1.16, + "learning_rate": 3.0764298538086686e-05, + "loss": 0.3373, + "step": 30050 + }, + { + "epoch": 1.16, + "learning_rate": 3.076365734803796e-05, + "loss": 0.4722, + "step": 30051 + }, + { + "epoch": 1.16, + "learning_rate": 3.076301615798923e-05, + "loss": 0.3788, + "step": 30052 + }, + { + "epoch": 1.16, + "learning_rate": 3.07623749679405e-05, + "loss": 0.3363, + "step": 30053 + }, + { + "epoch": 1.16, + "learning_rate": 3.076173377789177e-05, + "loss": 0.3882, + "step": 30054 + }, + { + "epoch": 1.16, + "learning_rate": 3.076109258784304e-05, + "loss": 0.2463, + "step": 30055 + }, + { + "epoch": 1.16, + "learning_rate": 3.076045139779431e-05, + "loss": 0.5154, + "step": 30056 + }, + { + "epoch": 1.16, + "learning_rate": 3.0759810207745575e-05, + "loss": 0.3574, + "step": 30057 + }, + { + "epoch": 1.16, + "learning_rate": 3.0759169017696846e-05, + "loss": 0.5862, + "step": 30058 + }, + { + "epoch": 1.16, + "learning_rate": 3.075852782764812e-05, + "loss": 0.367, + "step": 30059 + }, + { + "epoch": 1.16, + "learning_rate": 3.075788663759939e-05, + "loss": 0.5358, + "step": 30060 + }, + { + "epoch": 1.16, + "learning_rate": 3.0757245447550655e-05, + "loss": 0.3473, + "step": 30061 + }, + { + "epoch": 1.16, + "learning_rate": 3.0756604257501926e-05, + "loss": 0.3069, + "step": 30062 + }, + { + "epoch": 1.16, + "learning_rate": 3.07559630674532e-05, + "loss": 0.5913, + "step": 30063 + }, + { + "epoch": 1.16, + "learning_rate": 3.075532187740446e-05, + "loss": 0.419, + "step": 30064 + }, + { + "epoch": 1.16, + "learning_rate": 3.0754680687355735e-05, + "loss": 0.3972, + "step": 30065 + }, + { + "epoch": 1.16, + "learning_rate": 3.0754039497307e-05, + "loss": 0.3709, + "step": 30066 + }, + { + "epoch": 1.16, + "learning_rate": 3.075339830725827e-05, + "loss": 0.4433, + "step": 30067 + }, + { + "epoch": 1.16, + "learning_rate": 3.075275711720954e-05, + "loss": 0.3273, + "step": 30068 + }, + { + "epoch": 1.16, + "learning_rate": 3.0752115927160815e-05, + "loss": 0.314, + "step": 30069 + }, + { + "epoch": 1.16, + "learning_rate": 3.075147473711208e-05, + "loss": 0.4226, + "step": 30070 + }, + { + "epoch": 1.16, + "learning_rate": 3.075083354706335e-05, + "loss": 0.3828, + "step": 30071 + }, + { + "epoch": 1.16, + "learning_rate": 3.075019235701462e-05, + "loss": 0.3316, + "step": 30072 + }, + { + "epoch": 1.16, + "learning_rate": 3.074955116696589e-05, + "loss": 0.4303, + "step": 30073 + }, + { + "epoch": 1.16, + "learning_rate": 3.074890997691716e-05, + "loss": 0.4534, + "step": 30074 + }, + { + "epoch": 1.16, + "learning_rate": 3.074826878686843e-05, + "loss": 0.3051, + "step": 30075 + }, + { + "epoch": 1.16, + "learning_rate": 3.07476275968197e-05, + "loss": 0.3949, + "step": 30076 + }, + { + "epoch": 1.16, + "learning_rate": 3.074698640677097e-05, + "loss": 0.3694, + "step": 30077 + }, + { + "epoch": 1.16, + "learning_rate": 3.074634521672224e-05, + "loss": 0.4055, + "step": 30078 + }, + { + "epoch": 1.16, + "learning_rate": 3.0745704026673505e-05, + "loss": 0.3872, + "step": 30079 + }, + { + "epoch": 1.16, + "learning_rate": 3.074506283662478e-05, + "loss": 0.4405, + "step": 30080 + }, + { + "epoch": 1.16, + "learning_rate": 3.074442164657605e-05, + "loss": 0.3448, + "step": 30081 + }, + { + "epoch": 1.16, + "learning_rate": 3.074378045652731e-05, + "loss": 0.4293, + "step": 30082 + }, + { + "epoch": 1.16, + "learning_rate": 3.0743139266478585e-05, + "loss": 0.4418, + "step": 30083 + }, + { + "epoch": 1.16, + "learning_rate": 3.074249807642986e-05, + "loss": 0.5385, + "step": 30084 + }, + { + "epoch": 1.16, + "learning_rate": 3.074185688638113e-05, + "loss": 0.4343, + "step": 30085 + }, + { + "epoch": 1.16, + "learning_rate": 3.0741215696332393e-05, + "loss": 0.2996, + "step": 30086 + }, + { + "epoch": 1.16, + "learning_rate": 3.0740574506283665e-05, + "loss": 0.439, + "step": 30087 + }, + { + "epoch": 1.16, + "learning_rate": 3.073993331623493e-05, + "loss": 0.4432, + "step": 30088 + }, + { + "epoch": 1.16, + "learning_rate": 3.07392921261862e-05, + "loss": 0.4445, + "step": 30089 + }, + { + "epoch": 1.16, + "learning_rate": 3.0738650936137473e-05, + "loss": 0.4224, + "step": 30090 + }, + { + "epoch": 1.16, + "learning_rate": 3.073800974608874e-05, + "loss": 0.3016, + "step": 30091 + }, + { + "epoch": 1.16, + "learning_rate": 3.073736855604002e-05, + "loss": 0.4434, + "step": 30092 + }, + { + "epoch": 1.16, + "learning_rate": 3.073672736599128e-05, + "loss": 0.4495, + "step": 30093 + }, + { + "epoch": 1.16, + "learning_rate": 3.0736086175942554e-05, + "loss": 0.4212, + "step": 30094 + }, + { + "epoch": 1.16, + "learning_rate": 3.073544498589382e-05, + "loss": 0.337, + "step": 30095 + }, + { + "epoch": 1.16, + "learning_rate": 3.073480379584509e-05, + "loss": 0.4896, + "step": 30096 + }, + { + "epoch": 1.16, + "learning_rate": 3.0734162605796355e-05, + "loss": 0.3067, + "step": 30097 + }, + { + "epoch": 1.16, + "learning_rate": 3.073352141574763e-05, + "loss": 0.2888, + "step": 30098 + }, + { + "epoch": 1.16, + "learning_rate": 3.07328802256989e-05, + "loss": 0.5097, + "step": 30099 + }, + { + "epoch": 1.16, + "learning_rate": 3.073223903565017e-05, + "loss": 0.3819, + "step": 30100 + }, + { + "epoch": 1.16, + "learning_rate": 3.073159784560144e-05, + "loss": 0.4905, + "step": 30101 + }, + { + "epoch": 1.16, + "learning_rate": 3.073095665555271e-05, + "loss": 0.5424, + "step": 30102 + }, + { + "epoch": 1.16, + "learning_rate": 3.073031546550398e-05, + "loss": 0.4389, + "step": 30103 + }, + { + "epoch": 1.16, + "learning_rate": 3.0729674275455244e-05, + "loss": 0.438, + "step": 30104 + }, + { + "epoch": 1.16, + "learning_rate": 3.0729033085406515e-05, + "loss": 0.4384, + "step": 30105 + }, + { + "epoch": 1.16, + "learning_rate": 3.072839189535778e-05, + "loss": 0.3559, + "step": 30106 + }, + { + "epoch": 1.16, + "learning_rate": 3.072775070530905e-05, + "loss": 0.4632, + "step": 30107 + }, + { + "epoch": 1.16, + "learning_rate": 3.072710951526033e-05, + "loss": 0.31, + "step": 30108 + }, + { + "epoch": 1.16, + "learning_rate": 3.0726468325211595e-05, + "loss": 0.4115, + "step": 30109 + }, + { + "epoch": 1.16, + "learning_rate": 3.072582713516287e-05, + "loss": 0.4636, + "step": 30110 + }, + { + "epoch": 1.16, + "learning_rate": 3.072518594511413e-05, + "loss": 0.4571, + "step": 30111 + }, + { + "epoch": 1.16, + "learning_rate": 3.0724544755065404e-05, + "loss": 0.4698, + "step": 30112 + }, + { + "epoch": 1.16, + "learning_rate": 3.072390356501667e-05, + "loss": 0.4112, + "step": 30113 + }, + { + "epoch": 1.16, + "learning_rate": 3.072326237496794e-05, + "loss": 0.3678, + "step": 30114 + }, + { + "epoch": 1.16, + "learning_rate": 3.072262118491921e-05, + "loss": 0.3906, + "step": 30115 + }, + { + "epoch": 1.16, + "learning_rate": 3.0721979994870484e-05, + "loss": 0.4687, + "step": 30116 + }, + { + "epoch": 1.16, + "learning_rate": 3.0721338804821756e-05, + "loss": 0.2524, + "step": 30117 + }, + { + "epoch": 1.16, + "learning_rate": 3.072069761477302e-05, + "loss": 0.4, + "step": 30118 + }, + { + "epoch": 1.16, + "learning_rate": 3.072005642472429e-05, + "loss": 0.4099, + "step": 30119 + }, + { + "epoch": 1.16, + "learning_rate": 3.071941523467556e-05, + "loss": 0.3964, + "step": 30120 + }, + { + "epoch": 1.16, + "learning_rate": 3.071877404462683e-05, + "loss": 0.3362, + "step": 30121 + }, + { + "epoch": 1.16, + "learning_rate": 3.0718132854578094e-05, + "loss": 0.4708, + "step": 30122 + }, + { + "epoch": 1.16, + "learning_rate": 3.0717491664529366e-05, + "loss": 0.4087, + "step": 30123 + }, + { + "epoch": 1.16, + "learning_rate": 3.071685047448064e-05, + "loss": 0.3265, + "step": 30124 + }, + { + "epoch": 1.16, + "learning_rate": 3.071620928443191e-05, + "loss": 0.3462, + "step": 30125 + }, + { + "epoch": 1.16, + "learning_rate": 3.071556809438318e-05, + "loss": 0.3844, + "step": 30126 + }, + { + "epoch": 1.16, + "learning_rate": 3.0714926904334446e-05, + "loss": 0.3188, + "step": 30127 + }, + { + "epoch": 1.16, + "learning_rate": 3.071428571428572e-05, + "loss": 0.4783, + "step": 30128 + }, + { + "epoch": 1.16, + "learning_rate": 3.071364452423698e-05, + "loss": 0.3645, + "step": 30129 + }, + { + "epoch": 1.16, + "learning_rate": 3.0713003334188254e-05, + "loss": 0.3664, + "step": 30130 + }, + { + "epoch": 1.16, + "learning_rate": 3.0712362144139526e-05, + "loss": 0.3864, + "step": 30131 + }, + { + "epoch": 1.16, + "learning_rate": 3.07117209540908e-05, + "loss": 0.3992, + "step": 30132 + }, + { + "epoch": 1.16, + "learning_rate": 3.071107976404206e-05, + "loss": 0.4798, + "step": 30133 + }, + { + "epoch": 1.16, + "learning_rate": 3.0710438573993334e-05, + "loss": 0.3167, + "step": 30134 + }, + { + "epoch": 1.16, + "learning_rate": 3.0709797383944606e-05, + "loss": 0.3252, + "step": 30135 + }, + { + "epoch": 1.16, + "learning_rate": 3.070915619389587e-05, + "loss": 0.3051, + "step": 30136 + }, + { + "epoch": 1.16, + "learning_rate": 3.070851500384714e-05, + "loss": 0.4297, + "step": 30137 + }, + { + "epoch": 1.16, + "learning_rate": 3.070787381379841e-05, + "loss": 0.5142, + "step": 30138 + }, + { + "epoch": 1.16, + "learning_rate": 3.070723262374968e-05, + "loss": 0.4713, + "step": 30139 + }, + { + "epoch": 1.16, + "learning_rate": 3.070659143370095e-05, + "loss": 0.3851, + "step": 30140 + }, + { + "epoch": 1.16, + "learning_rate": 3.070595024365222e-05, + "loss": 0.4052, + "step": 30141 + }, + { + "epoch": 1.16, + "learning_rate": 3.070530905360349e-05, + "loss": 0.2633, + "step": 30142 + }, + { + "epoch": 1.16, + "learning_rate": 3.070466786355476e-05, + "loss": 0.4494, + "step": 30143 + }, + { + "epoch": 1.16, + "learning_rate": 3.070402667350603e-05, + "loss": 0.3933, + "step": 30144 + }, + { + "epoch": 1.16, + "learning_rate": 3.0703385483457296e-05, + "loss": 0.3878, + "step": 30145 + }, + { + "epoch": 1.16, + "learning_rate": 3.070274429340857e-05, + "loss": 0.3981, + "step": 30146 + }, + { + "epoch": 1.16, + "learning_rate": 3.070210310335984e-05, + "loss": 0.4845, + "step": 30147 + }, + { + "epoch": 1.16, + "learning_rate": 3.070146191331111e-05, + "loss": 0.3918, + "step": 30148 + }, + { + "epoch": 1.16, + "learning_rate": 3.0700820723262376e-05, + "loss": 0.4005, + "step": 30149 + }, + { + "epoch": 1.16, + "learning_rate": 3.070017953321365e-05, + "loss": 0.458, + "step": 30150 + }, + { + "epoch": 1.16, + "learning_rate": 3.069953834316491e-05, + "loss": 0.3837, + "step": 30151 + }, + { + "epoch": 1.16, + "learning_rate": 3.0698897153116184e-05, + "loss": 0.3949, + "step": 30152 + }, + { + "epoch": 1.16, + "learning_rate": 3.0698255963067456e-05, + "loss": 0.3546, + "step": 30153 + }, + { + "epoch": 1.16, + "learning_rate": 3.069761477301872e-05, + "loss": 0.4248, + "step": 30154 + }, + { + "epoch": 1.16, + "learning_rate": 3.069697358296999e-05, + "loss": 0.3837, + "step": 30155 + }, + { + "epoch": 1.16, + "learning_rate": 3.0696332392921264e-05, + "loss": 0.5123, + "step": 30156 + }, + { + "epoch": 1.16, + "learning_rate": 3.0695691202872536e-05, + "loss": 0.3933, + "step": 30157 + }, + { + "epoch": 1.16, + "learning_rate": 3.06950500128238e-05, + "loss": 0.4101, + "step": 30158 + }, + { + "epoch": 1.16, + "learning_rate": 3.069440882277507e-05, + "loss": 0.409, + "step": 30159 + }, + { + "epoch": 1.16, + "learning_rate": 3.069376763272634e-05, + "loss": 0.3918, + "step": 30160 + }, + { + "epoch": 1.16, + "learning_rate": 3.069312644267761e-05, + "loss": 0.4798, + "step": 30161 + }, + { + "epoch": 1.16, + "learning_rate": 3.069248525262888e-05, + "loss": 0.4846, + "step": 30162 + }, + { + "epoch": 1.16, + "learning_rate": 3.0691844062580146e-05, + "loss": 0.4329, + "step": 30163 + }, + { + "epoch": 1.16, + "learning_rate": 3.0691202872531425e-05, + "loss": 0.4282, + "step": 30164 + }, + { + "epoch": 1.16, + "learning_rate": 3.069056168248269e-05, + "loss": 0.406, + "step": 30165 + }, + { + "epoch": 1.16, + "learning_rate": 3.068992049243396e-05, + "loss": 0.4218, + "step": 30166 + }, + { + "epoch": 1.16, + "learning_rate": 3.0689279302385226e-05, + "loss": 0.458, + "step": 30167 + }, + { + "epoch": 1.16, + "learning_rate": 3.06886381123365e-05, + "loss": 0.502, + "step": 30168 + }, + { + "epoch": 1.16, + "learning_rate": 3.068799692228776e-05, + "loss": 0.4035, + "step": 30169 + }, + { + "epoch": 1.16, + "learning_rate": 3.0687355732239035e-05, + "loss": 0.4642, + "step": 30170 + }, + { + "epoch": 1.16, + "learning_rate": 3.0686714542190306e-05, + "loss": 0.4131, + "step": 30171 + }, + { + "epoch": 1.16, + "learning_rate": 3.068607335214158e-05, + "loss": 0.4386, + "step": 30172 + }, + { + "epoch": 1.16, + "learning_rate": 3.068543216209285e-05, + "loss": 0.4399, + "step": 30173 + }, + { + "epoch": 1.16, + "learning_rate": 3.0684790972044115e-05, + "loss": 0.6448, + "step": 30174 + }, + { + "epoch": 1.16, + "learning_rate": 3.0684149781995386e-05, + "loss": 0.3697, + "step": 30175 + }, + { + "epoch": 1.16, + "learning_rate": 3.068350859194665e-05, + "loss": 0.4116, + "step": 30176 + }, + { + "epoch": 1.16, + "learning_rate": 3.068286740189792e-05, + "loss": 0.3373, + "step": 30177 + }, + { + "epoch": 1.16, + "learning_rate": 3.068222621184919e-05, + "loss": 0.4352, + "step": 30178 + }, + { + "epoch": 1.16, + "learning_rate": 3.068158502180046e-05, + "loss": 0.4001, + "step": 30179 + }, + { + "epoch": 1.16, + "learning_rate": 3.068094383175174e-05, + "loss": 0.3871, + "step": 30180 + }, + { + "epoch": 1.16, + "learning_rate": 3.0680302641703e-05, + "loss": 0.2475, + "step": 30181 + }, + { + "epoch": 1.16, + "learning_rate": 3.0679661451654275e-05, + "loss": 0.3646, + "step": 30182 + }, + { + "epoch": 1.16, + "learning_rate": 3.067902026160554e-05, + "loss": 0.4719, + "step": 30183 + }, + { + "epoch": 1.16, + "learning_rate": 3.067837907155681e-05, + "loss": 0.3655, + "step": 30184 + }, + { + "epoch": 1.16, + "learning_rate": 3.0677737881508077e-05, + "loss": 0.3378, + "step": 30185 + }, + { + "epoch": 1.16, + "learning_rate": 3.067709669145935e-05, + "loss": 0.3736, + "step": 30186 + }, + { + "epoch": 1.16, + "learning_rate": 3.067645550141062e-05, + "loss": 0.3984, + "step": 30187 + }, + { + "epoch": 1.16, + "learning_rate": 3.067581431136189e-05, + "loss": 0.4675, + "step": 30188 + }, + { + "epoch": 1.16, + "learning_rate": 3.067517312131316e-05, + "loss": 0.5326, + "step": 30189 + }, + { + "epoch": 1.16, + "learning_rate": 3.067453193126443e-05, + "loss": 0.3697, + "step": 30190 + }, + { + "epoch": 1.16, + "learning_rate": 3.06738907412157e-05, + "loss": 0.4302, + "step": 30191 + }, + { + "epoch": 1.16, + "learning_rate": 3.0673249551166965e-05, + "loss": 0.2754, + "step": 30192 + }, + { + "epoch": 1.16, + "learning_rate": 3.067260836111824e-05, + "loss": 0.3757, + "step": 30193 + }, + { + "epoch": 1.16, + "learning_rate": 3.06719671710695e-05, + "loss": 0.3379, + "step": 30194 + }, + { + "epoch": 1.16, + "learning_rate": 3.067132598102077e-05, + "loss": 0.3973, + "step": 30195 + }, + { + "epoch": 1.16, + "learning_rate": 3.0670684790972045e-05, + "loss": 0.3118, + "step": 30196 + }, + { + "epoch": 1.16, + "learning_rate": 3.067004360092332e-05, + "loss": 0.4575, + "step": 30197 + }, + { + "epoch": 1.16, + "learning_rate": 3.066940241087459e-05, + "loss": 0.3676, + "step": 30198 + }, + { + "epoch": 1.16, + "learning_rate": 3.0668761220825853e-05, + "loss": 0.4258, + "step": 30199 + }, + { + "epoch": 1.16, + "learning_rate": 3.0668120030777125e-05, + "loss": 0.4153, + "step": 30200 + }, + { + "epoch": 1.16, + "learning_rate": 3.066747884072839e-05, + "loss": 0.4521, + "step": 30201 + }, + { + "epoch": 1.16, + "learning_rate": 3.066683765067966e-05, + "loss": 0.4213, + "step": 30202 + }, + { + "epoch": 1.16, + "learning_rate": 3.0666196460630934e-05, + "loss": 0.2643, + "step": 30203 + }, + { + "epoch": 1.16, + "learning_rate": 3.0665555270582205e-05, + "loss": 0.3581, + "step": 30204 + }, + { + "epoch": 1.16, + "learning_rate": 3.066491408053347e-05, + "loss": 0.34, + "step": 30205 + }, + { + "epoch": 1.16, + "learning_rate": 3.066427289048474e-05, + "loss": 0.3314, + "step": 30206 + }, + { + "epoch": 1.16, + "learning_rate": 3.0663631700436014e-05, + "loss": 0.4058, + "step": 30207 + }, + { + "epoch": 1.16, + "learning_rate": 3.066299051038728e-05, + "loss": 0.4163, + "step": 30208 + }, + { + "epoch": 1.16, + "learning_rate": 3.066234932033855e-05, + "loss": 0.3862, + "step": 30209 + }, + { + "epoch": 1.16, + "learning_rate": 3.0661708130289815e-05, + "loss": 0.401, + "step": 30210 + }, + { + "epoch": 1.16, + "learning_rate": 3.066106694024109e-05, + "loss": 0.514, + "step": 30211 + }, + { + "epoch": 1.16, + "learning_rate": 3.066042575019236e-05, + "loss": 0.3484, + "step": 30212 + }, + { + "epoch": 1.16, + "learning_rate": 3.065978456014363e-05, + "loss": 0.3817, + "step": 30213 + }, + { + "epoch": 1.16, + "learning_rate": 3.0659143370094895e-05, + "loss": 0.2496, + "step": 30214 + }, + { + "epoch": 1.16, + "learning_rate": 3.065850218004617e-05, + "loss": 0.5106, + "step": 30215 + }, + { + "epoch": 1.16, + "learning_rate": 3.065786098999744e-05, + "loss": 0.3948, + "step": 30216 + }, + { + "epoch": 1.16, + "learning_rate": 3.0657219799948704e-05, + "loss": 0.4574, + "step": 30217 + }, + { + "epoch": 1.16, + "learning_rate": 3.0656578609899975e-05, + "loss": 0.3615, + "step": 30218 + }, + { + "epoch": 1.16, + "learning_rate": 3.065593741985125e-05, + "loss": 0.3278, + "step": 30219 + }, + { + "epoch": 1.16, + "learning_rate": 3.065529622980252e-05, + "loss": 0.5256, + "step": 30220 + }, + { + "epoch": 1.16, + "learning_rate": 3.0654655039753784e-05, + "loss": 0.418, + "step": 30221 + }, + { + "epoch": 1.16, + "learning_rate": 3.0654013849705055e-05, + "loss": 0.3595, + "step": 30222 + }, + { + "epoch": 1.16, + "learning_rate": 3.065337265965632e-05, + "loss": 0.3623, + "step": 30223 + }, + { + "epoch": 1.16, + "learning_rate": 3.065273146960759e-05, + "loss": 0.3802, + "step": 30224 + }, + { + "epoch": 1.16, + "learning_rate": 3.0652090279558864e-05, + "loss": 0.4548, + "step": 30225 + }, + { + "epoch": 1.16, + "learning_rate": 3.065144908951013e-05, + "loss": 0.408, + "step": 30226 + }, + { + "epoch": 1.16, + "learning_rate": 3.06508078994614e-05, + "loss": 0.3649, + "step": 30227 + }, + { + "epoch": 1.16, + "learning_rate": 3.065016670941267e-05, + "loss": 0.3569, + "step": 30228 + }, + { + "epoch": 1.16, + "learning_rate": 3.0649525519363944e-05, + "loss": 0.3367, + "step": 30229 + }, + { + "epoch": 1.16, + "learning_rate": 3.064888432931521e-05, + "loss": 0.5193, + "step": 30230 + }, + { + "epoch": 1.16, + "learning_rate": 3.064824313926648e-05, + "loss": 0.4564, + "step": 30231 + }, + { + "epoch": 1.16, + "learning_rate": 3.0647601949217746e-05, + "loss": 0.2872, + "step": 30232 + }, + { + "epoch": 1.16, + "learning_rate": 3.064696075916902e-05, + "loss": 0.4174, + "step": 30233 + }, + { + "epoch": 1.16, + "learning_rate": 3.064631956912029e-05, + "loss": 0.377, + "step": 30234 + }, + { + "epoch": 1.16, + "learning_rate": 3.064567837907156e-05, + "loss": 0.3619, + "step": 30235 + }, + { + "epoch": 1.16, + "learning_rate": 3.064503718902283e-05, + "loss": 0.3892, + "step": 30236 + }, + { + "epoch": 1.16, + "learning_rate": 3.06443959989741e-05, + "loss": 0.3459, + "step": 30237 + }, + { + "epoch": 1.16, + "learning_rate": 3.064375480892537e-05, + "loss": 0.3728, + "step": 30238 + }, + { + "epoch": 1.16, + "learning_rate": 3.0643113618876634e-05, + "loss": 0.3604, + "step": 30239 + }, + { + "epoch": 1.16, + "learning_rate": 3.0642472428827906e-05, + "loss": 0.3512, + "step": 30240 + }, + { + "epoch": 1.16, + "learning_rate": 3.064183123877918e-05, + "loss": 0.3456, + "step": 30241 + }, + { + "epoch": 1.16, + "learning_rate": 3.064119004873044e-05, + "loss": 0.4517, + "step": 30242 + }, + { + "epoch": 1.16, + "learning_rate": 3.0640548858681714e-05, + "loss": 0.5322, + "step": 30243 + }, + { + "epoch": 1.16, + "learning_rate": 3.0639907668632986e-05, + "loss": 0.4159, + "step": 30244 + }, + { + "epoch": 1.16, + "learning_rate": 3.063926647858426e-05, + "loss": 0.4521, + "step": 30245 + }, + { + "epoch": 1.16, + "learning_rate": 3.063862528853552e-05, + "loss": 0.4125, + "step": 30246 + }, + { + "epoch": 1.16, + "learning_rate": 3.0637984098486794e-05, + "loss": 0.5082, + "step": 30247 + }, + { + "epoch": 1.16, + "learning_rate": 3.063734290843806e-05, + "loss": 0.3106, + "step": 30248 + }, + { + "epoch": 1.16, + "learning_rate": 3.063670171838933e-05, + "loss": 0.328, + "step": 30249 + }, + { + "epoch": 1.16, + "learning_rate": 3.06360605283406e-05, + "loss": 0.3594, + "step": 30250 + }, + { + "epoch": 1.16, + "learning_rate": 3.063541933829187e-05, + "loss": 0.425, + "step": 30251 + }, + { + "epoch": 1.16, + "learning_rate": 3.0634778148243146e-05, + "loss": 0.4214, + "step": 30252 + }, + { + "epoch": 1.16, + "learning_rate": 3.063413695819441e-05, + "loss": 0.287, + "step": 30253 + }, + { + "epoch": 1.16, + "learning_rate": 3.063349576814568e-05, + "loss": 0.512, + "step": 30254 + }, + { + "epoch": 1.16, + "learning_rate": 3.063285457809695e-05, + "loss": 0.4498, + "step": 30255 + }, + { + "epoch": 1.16, + "learning_rate": 3.063221338804822e-05, + "loss": 0.3636, + "step": 30256 + }, + { + "epoch": 1.16, + "learning_rate": 3.0631572197999484e-05, + "loss": 0.5197, + "step": 30257 + }, + { + "epoch": 1.16, + "learning_rate": 3.0630931007950756e-05, + "loss": 0.3628, + "step": 30258 + }, + { + "epoch": 1.16, + "learning_rate": 3.063028981790203e-05, + "loss": 0.3842, + "step": 30259 + }, + { + "epoch": 1.16, + "learning_rate": 3.06296486278533e-05, + "loss": 0.3117, + "step": 30260 + }, + { + "epoch": 1.16, + "learning_rate": 3.062900743780457e-05, + "loss": 0.4392, + "step": 30261 + }, + { + "epoch": 1.16, + "learning_rate": 3.0628366247755836e-05, + "loss": 0.6041, + "step": 30262 + }, + { + "epoch": 1.16, + "learning_rate": 3.062772505770711e-05, + "loss": 0.3811, + "step": 30263 + }, + { + "epoch": 1.16, + "learning_rate": 3.062708386765837e-05, + "loss": 0.4044, + "step": 30264 + }, + { + "epoch": 1.16, + "learning_rate": 3.0626442677609644e-05, + "loss": 0.4186, + "step": 30265 + }, + { + "epoch": 1.16, + "learning_rate": 3.062580148756091e-05, + "loss": 0.3593, + "step": 30266 + }, + { + "epoch": 1.16, + "learning_rate": 3.062516029751218e-05, + "loss": 0.4619, + "step": 30267 + }, + { + "epoch": 1.16, + "learning_rate": 3.062451910746346e-05, + "loss": 0.3276, + "step": 30268 + }, + { + "epoch": 1.16, + "learning_rate": 3.0623877917414725e-05, + "loss": 0.3785, + "step": 30269 + }, + { + "epoch": 1.16, + "learning_rate": 3.0623236727365996e-05, + "loss": 0.3839, + "step": 30270 + }, + { + "epoch": 1.16, + "learning_rate": 3.062259553731726e-05, + "loss": 0.4723, + "step": 30271 + }, + { + "epoch": 1.16, + "learning_rate": 3.062195434726853e-05, + "loss": 0.4252, + "step": 30272 + }, + { + "epoch": 1.16, + "learning_rate": 3.06213131572198e-05, + "loss": 0.4005, + "step": 30273 + }, + { + "epoch": 1.16, + "learning_rate": 3.062067196717107e-05, + "loss": 0.4155, + "step": 30274 + }, + { + "epoch": 1.16, + "learning_rate": 3.062003077712234e-05, + "loss": 0.3232, + "step": 30275 + }, + { + "epoch": 1.16, + "learning_rate": 3.061938958707361e-05, + "loss": 0.3198, + "step": 30276 + }, + { + "epoch": 1.16, + "learning_rate": 3.0618748397024885e-05, + "loss": 0.5105, + "step": 30277 + }, + { + "epoch": 1.16, + "learning_rate": 3.061810720697615e-05, + "loss": 0.3588, + "step": 30278 + }, + { + "epoch": 1.16, + "learning_rate": 3.061746601692742e-05, + "loss": 0.2709, + "step": 30279 + }, + { + "epoch": 1.16, + "learning_rate": 3.0616824826878686e-05, + "loss": 0.334, + "step": 30280 + }, + { + "epoch": 1.16, + "learning_rate": 3.061618363682996e-05, + "loss": 0.4628, + "step": 30281 + }, + { + "epoch": 1.16, + "learning_rate": 3.061554244678122e-05, + "loss": 0.4332, + "step": 30282 + }, + { + "epoch": 1.16, + "learning_rate": 3.0614901256732495e-05, + "loss": 0.37, + "step": 30283 + }, + { + "epoch": 1.16, + "learning_rate": 3.0614260066683766e-05, + "loss": 0.4038, + "step": 30284 + }, + { + "epoch": 1.16, + "learning_rate": 3.061361887663504e-05, + "loss": 0.3741, + "step": 30285 + }, + { + "epoch": 1.16, + "learning_rate": 3.061297768658631e-05, + "loss": 0.3464, + "step": 30286 + }, + { + "epoch": 1.16, + "learning_rate": 3.0612336496537575e-05, + "loss": 0.3935, + "step": 30287 + }, + { + "epoch": 1.16, + "learning_rate": 3.0611695306488846e-05, + "loss": 0.3102, + "step": 30288 + }, + { + "epoch": 1.16, + "learning_rate": 3.061105411644011e-05, + "loss": 0.3598, + "step": 30289 + }, + { + "epoch": 1.16, + "learning_rate": 3.061041292639138e-05, + "loss": 0.4241, + "step": 30290 + }, + { + "epoch": 1.16, + "learning_rate": 3.0609771736342655e-05, + "loss": 0.3767, + "step": 30291 + }, + { + "epoch": 1.16, + "learning_rate": 3.0609130546293927e-05, + "loss": 0.3132, + "step": 30292 + }, + { + "epoch": 1.16, + "learning_rate": 3.060848935624519e-05, + "loss": 0.3419, + "step": 30293 + }, + { + "epoch": 1.16, + "learning_rate": 3.060784816619646e-05, + "loss": 0.4693, + "step": 30294 + }, + { + "epoch": 1.16, + "learning_rate": 3.0607206976147735e-05, + "loss": 0.4052, + "step": 30295 + }, + { + "epoch": 1.16, + "learning_rate": 3.0606565786099e-05, + "loss": 0.3739, + "step": 30296 + }, + { + "epoch": 1.16, + "learning_rate": 3.060592459605027e-05, + "loss": 0.3648, + "step": 30297 + }, + { + "epoch": 1.16, + "learning_rate": 3.0605283406001537e-05, + "loss": 0.2611, + "step": 30298 + }, + { + "epoch": 1.16, + "learning_rate": 3.060464221595281e-05, + "loss": 0.3489, + "step": 30299 + }, + { + "epoch": 1.16, + "learning_rate": 3.060400102590408e-05, + "loss": 0.2233, + "step": 30300 + }, + { + "epoch": 1.16, + "learning_rate": 3.060335983585535e-05, + "loss": 0.3971, + "step": 30301 + }, + { + "epoch": 1.17, + "learning_rate": 3.060271864580662e-05, + "loss": 0.4139, + "step": 30302 + }, + { + "epoch": 1.17, + "learning_rate": 3.060207745575789e-05, + "loss": 0.3513, + "step": 30303 + }, + { + "epoch": 1.17, + "learning_rate": 3.060143626570916e-05, + "loss": 0.5095, + "step": 30304 + }, + { + "epoch": 1.17, + "learning_rate": 3.0600795075660425e-05, + "loss": 0.4812, + "step": 30305 + }, + { + "epoch": 1.17, + "learning_rate": 3.06001538856117e-05, + "loss": 0.3646, + "step": 30306 + }, + { + "epoch": 1.17, + "learning_rate": 3.059951269556297e-05, + "loss": 0.2769, + "step": 30307 + }, + { + "epoch": 1.17, + "learning_rate": 3.059887150551424e-05, + "loss": 0.4385, + "step": 30308 + }, + { + "epoch": 1.17, + "learning_rate": 3.0598230315465505e-05, + "loss": 0.3674, + "step": 30309 + }, + { + "epoch": 1.17, + "learning_rate": 3.059758912541678e-05, + "loss": 0.3841, + "step": 30310 + }, + { + "epoch": 1.17, + "learning_rate": 3.059694793536804e-05, + "loss": 0.3541, + "step": 30311 + }, + { + "epoch": 1.17, + "learning_rate": 3.0596306745319313e-05, + "loss": 0.4451, + "step": 30312 + }, + { + "epoch": 1.17, + "learning_rate": 3.0595665555270585e-05, + "loss": 0.4035, + "step": 30313 + }, + { + "epoch": 1.17, + "learning_rate": 3.059502436522185e-05, + "loss": 0.4002, + "step": 30314 + }, + { + "epoch": 1.17, + "learning_rate": 3.059438317517312e-05, + "loss": 0.3501, + "step": 30315 + }, + { + "epoch": 1.17, + "learning_rate": 3.0593741985124394e-05, + "loss": 0.4172, + "step": 30316 + }, + { + "epoch": 1.17, + "learning_rate": 3.0593100795075665e-05, + "loss": 0.4329, + "step": 30317 + }, + { + "epoch": 1.17, + "learning_rate": 3.059245960502693e-05, + "loss": 0.4473, + "step": 30318 + }, + { + "epoch": 1.17, + "learning_rate": 3.05918184149782e-05, + "loss": 0.3636, + "step": 30319 + }, + { + "epoch": 1.17, + "learning_rate": 3.059117722492947e-05, + "loss": 0.4412, + "step": 30320 + }, + { + "epoch": 1.17, + "learning_rate": 3.059053603488074e-05, + "loss": 0.3954, + "step": 30321 + }, + { + "epoch": 1.17, + "learning_rate": 3.058989484483201e-05, + "loss": 0.5592, + "step": 30322 + }, + { + "epoch": 1.17, + "learning_rate": 3.058925365478328e-05, + "loss": 0.3506, + "step": 30323 + }, + { + "epoch": 1.17, + "learning_rate": 3.0588612464734554e-05, + "loss": 0.4528, + "step": 30324 + }, + { + "epoch": 1.17, + "learning_rate": 3.058797127468582e-05, + "loss": 0.3582, + "step": 30325 + }, + { + "epoch": 1.17, + "learning_rate": 3.058733008463709e-05, + "loss": 0.3304, + "step": 30326 + }, + { + "epoch": 1.17, + "learning_rate": 3.0586688894588355e-05, + "loss": 0.3148, + "step": 30327 + }, + { + "epoch": 1.17, + "learning_rate": 3.058604770453963e-05, + "loss": 0.4102, + "step": 30328 + }, + { + "epoch": 1.17, + "learning_rate": 3.058540651449089e-05, + "loss": 0.4329, + "step": 30329 + }, + { + "epoch": 1.17, + "learning_rate": 3.0584765324442164e-05, + "loss": 0.3727, + "step": 30330 + }, + { + "epoch": 1.17, + "learning_rate": 3.0584124134393435e-05, + "loss": 0.3279, + "step": 30331 + }, + { + "epoch": 1.17, + "learning_rate": 3.058348294434471e-05, + "loss": 0.4935, + "step": 30332 + }, + { + "epoch": 1.17, + "learning_rate": 3.058284175429598e-05, + "loss": 0.3478, + "step": 30333 + }, + { + "epoch": 1.17, + "learning_rate": 3.0582200564247244e-05, + "loss": 0.4196, + "step": 30334 + }, + { + "epoch": 1.17, + "learning_rate": 3.0581559374198516e-05, + "loss": 0.3396, + "step": 30335 + }, + { + "epoch": 1.17, + "learning_rate": 3.058091818414978e-05, + "loss": 0.2964, + "step": 30336 + }, + { + "epoch": 1.17, + "learning_rate": 3.058027699410105e-05, + "loss": 0.3679, + "step": 30337 + }, + { + "epoch": 1.17, + "learning_rate": 3.057963580405232e-05, + "loss": 0.4131, + "step": 30338 + }, + { + "epoch": 1.17, + "learning_rate": 3.057899461400359e-05, + "loss": 0.4957, + "step": 30339 + }, + { + "epoch": 1.17, + "learning_rate": 3.057835342395487e-05, + "loss": 0.3487, + "step": 30340 + }, + { + "epoch": 1.17, + "learning_rate": 3.057771223390613e-05, + "loss": 0.4658, + "step": 30341 + }, + { + "epoch": 1.17, + "learning_rate": 3.0577071043857404e-05, + "loss": 0.4493, + "step": 30342 + }, + { + "epoch": 1.17, + "learning_rate": 3.057642985380867e-05, + "loss": 0.4048, + "step": 30343 + }, + { + "epoch": 1.17, + "learning_rate": 3.057578866375994e-05, + "loss": 0.544, + "step": 30344 + }, + { + "epoch": 1.17, + "learning_rate": 3.0575147473711206e-05, + "loss": 0.3514, + "step": 30345 + }, + { + "epoch": 1.17, + "learning_rate": 3.057450628366248e-05, + "loss": 0.3137, + "step": 30346 + }, + { + "epoch": 1.17, + "learning_rate": 3.057386509361375e-05, + "loss": 0.5618, + "step": 30347 + }, + { + "epoch": 1.17, + "learning_rate": 3.057322390356502e-05, + "loss": 0.3576, + "step": 30348 + }, + { + "epoch": 1.17, + "learning_rate": 3.057258271351629e-05, + "loss": 0.3708, + "step": 30349 + }, + { + "epoch": 1.17, + "learning_rate": 3.057194152346756e-05, + "loss": 0.3847, + "step": 30350 + }, + { + "epoch": 1.17, + "learning_rate": 3.057130033341883e-05, + "loss": 0.4549, + "step": 30351 + }, + { + "epoch": 1.17, + "learning_rate": 3.0570659143370094e-05, + "loss": 0.4226, + "step": 30352 + }, + { + "epoch": 1.17, + "learning_rate": 3.0570017953321366e-05, + "loss": 0.3425, + "step": 30353 + }, + { + "epoch": 1.17, + "learning_rate": 3.056937676327263e-05, + "loss": 0.3489, + "step": 30354 + }, + { + "epoch": 1.17, + "learning_rate": 3.05687355732239e-05, + "loss": 0.4535, + "step": 30355 + }, + { + "epoch": 1.17, + "learning_rate": 3.0568094383175174e-05, + "loss": 0.4586, + "step": 30356 + }, + { + "epoch": 1.17, + "learning_rate": 3.0567453193126446e-05, + "loss": 0.4736, + "step": 30357 + }, + { + "epoch": 1.17, + "learning_rate": 3.056681200307772e-05, + "loss": 0.3238, + "step": 30358 + }, + { + "epoch": 1.17, + "learning_rate": 3.056617081302898e-05, + "loss": 0.3228, + "step": 30359 + }, + { + "epoch": 1.17, + "learning_rate": 3.0565529622980254e-05, + "loss": 0.385, + "step": 30360 + }, + { + "epoch": 1.17, + "learning_rate": 3.056488843293152e-05, + "loss": 0.3867, + "step": 30361 + }, + { + "epoch": 1.17, + "learning_rate": 3.056424724288279e-05, + "loss": 0.3548, + "step": 30362 + }, + { + "epoch": 1.17, + "learning_rate": 3.056360605283406e-05, + "loss": 0.3298, + "step": 30363 + }, + { + "epoch": 1.17, + "learning_rate": 3.0562964862785334e-05, + "loss": 0.3524, + "step": 30364 + }, + { + "epoch": 1.17, + "learning_rate": 3.05623236727366e-05, + "loss": 0.3366, + "step": 30365 + }, + { + "epoch": 1.17, + "learning_rate": 3.056168248268787e-05, + "loss": 0.3399, + "step": 30366 + }, + { + "epoch": 1.17, + "learning_rate": 3.056104129263914e-05, + "loss": 0.5642, + "step": 30367 + }, + { + "epoch": 1.17, + "learning_rate": 3.056040010259041e-05, + "loss": 0.3884, + "step": 30368 + }, + { + "epoch": 1.17, + "learning_rate": 3.055975891254168e-05, + "loss": 0.4866, + "step": 30369 + }, + { + "epoch": 1.17, + "learning_rate": 3.0559117722492944e-05, + "loss": 0.3389, + "step": 30370 + }, + { + "epoch": 1.17, + "learning_rate": 3.0558476532444216e-05, + "loss": 0.478, + "step": 30371 + }, + { + "epoch": 1.17, + "learning_rate": 3.055783534239549e-05, + "loss": 0.4382, + "step": 30372 + }, + { + "epoch": 1.17, + "learning_rate": 3.055719415234676e-05, + "loss": 0.4853, + "step": 30373 + }, + { + "epoch": 1.17, + "learning_rate": 3.0556552962298024e-05, + "loss": 0.3386, + "step": 30374 + }, + { + "epoch": 1.17, + "learning_rate": 3.0555911772249296e-05, + "loss": 0.4303, + "step": 30375 + }, + { + "epoch": 1.17, + "learning_rate": 3.055527058220057e-05, + "loss": 0.4848, + "step": 30376 + }, + { + "epoch": 1.17, + "learning_rate": 3.055462939215183e-05, + "loss": 0.2893, + "step": 30377 + }, + { + "epoch": 1.17, + "learning_rate": 3.0553988202103104e-05, + "loss": 0.4712, + "step": 30378 + }, + { + "epoch": 1.17, + "learning_rate": 3.0553347012054376e-05, + "loss": 0.2813, + "step": 30379 + }, + { + "epoch": 1.17, + "learning_rate": 3.055270582200565e-05, + "loss": 0.4088, + "step": 30380 + }, + { + "epoch": 1.17, + "learning_rate": 3.055206463195691e-05, + "loss": 0.4035, + "step": 30381 + }, + { + "epoch": 1.17, + "learning_rate": 3.0551423441908185e-05, + "loss": 0.5086, + "step": 30382 + }, + { + "epoch": 1.17, + "learning_rate": 3.055078225185945e-05, + "loss": 0.3489, + "step": 30383 + }, + { + "epoch": 1.17, + "learning_rate": 3.055014106181072e-05, + "loss": 0.342, + "step": 30384 + }, + { + "epoch": 1.17, + "learning_rate": 3.054949987176199e-05, + "loss": 0.2888, + "step": 30385 + }, + { + "epoch": 1.17, + "learning_rate": 3.054885868171326e-05, + "loss": 0.3401, + "step": 30386 + }, + { + "epoch": 1.17, + "learning_rate": 3.054821749166453e-05, + "loss": 0.3227, + "step": 30387 + }, + { + "epoch": 1.17, + "learning_rate": 3.05475763016158e-05, + "loss": 0.3969, + "step": 30388 + }, + { + "epoch": 1.17, + "learning_rate": 3.054693511156707e-05, + "loss": 0.3158, + "step": 30389 + }, + { + "epoch": 1.17, + "learning_rate": 3.054629392151834e-05, + "loss": 0.3865, + "step": 30390 + }, + { + "epoch": 1.17, + "learning_rate": 3.054565273146961e-05, + "loss": 0.381, + "step": 30391 + }, + { + "epoch": 1.17, + "learning_rate": 3.0545011541420875e-05, + "loss": 0.5364, + "step": 30392 + }, + { + "epoch": 1.17, + "learning_rate": 3.0544370351372146e-05, + "loss": 0.3553, + "step": 30393 + }, + { + "epoch": 1.17, + "learning_rate": 3.054372916132342e-05, + "loss": 0.4847, + "step": 30394 + }, + { + "epoch": 1.17, + "learning_rate": 3.054308797127469e-05, + "loss": 0.5754, + "step": 30395 + }, + { + "epoch": 1.17, + "learning_rate": 3.054244678122596e-05, + "loss": 0.4025, + "step": 30396 + }, + { + "epoch": 1.17, + "learning_rate": 3.0541805591177226e-05, + "loss": 0.4548, + "step": 30397 + }, + { + "epoch": 1.17, + "learning_rate": 3.05411644011285e-05, + "loss": 0.4053, + "step": 30398 + }, + { + "epoch": 1.17, + "learning_rate": 3.054052321107976e-05, + "loss": 0.4873, + "step": 30399 + }, + { + "epoch": 1.17, + "learning_rate": 3.0539882021031035e-05, + "loss": 0.4096, + "step": 30400 + }, + { + "epoch": 1.17, + "learning_rate": 3.05392408309823e-05, + "loss": 0.414, + "step": 30401 + }, + { + "epoch": 1.17, + "learning_rate": 3.053859964093357e-05, + "loss": 0.3595, + "step": 30402 + }, + { + "epoch": 1.17, + "learning_rate": 3.053795845088484e-05, + "loss": 0.3263, + "step": 30403 + }, + { + "epoch": 1.17, + "learning_rate": 3.0537317260836115e-05, + "loss": 0.3935, + "step": 30404 + }, + { + "epoch": 1.17, + "learning_rate": 3.0536676070787387e-05, + "loss": 0.4239, + "step": 30405 + }, + { + "epoch": 1.17, + "learning_rate": 3.053603488073865e-05, + "loss": 0.4202, + "step": 30406 + }, + { + "epoch": 1.17, + "learning_rate": 3.053539369068992e-05, + "loss": 0.3942, + "step": 30407 + }, + { + "epoch": 1.17, + "learning_rate": 3.053475250064119e-05, + "loss": 0.3604, + "step": 30408 + }, + { + "epoch": 1.17, + "learning_rate": 3.053411131059246e-05, + "loss": 0.3077, + "step": 30409 + }, + { + "epoch": 1.17, + "learning_rate": 3.0533470120543725e-05, + "loss": 0.4088, + "step": 30410 + }, + { + "epoch": 1.17, + "learning_rate": 3.0532828930494997e-05, + "loss": 0.4263, + "step": 30411 + }, + { + "epoch": 1.17, + "learning_rate": 3.0532187740446275e-05, + "loss": 0.3897, + "step": 30412 + }, + { + "epoch": 1.17, + "learning_rate": 3.053154655039754e-05, + "loss": 0.3244, + "step": 30413 + }, + { + "epoch": 1.17, + "learning_rate": 3.053090536034881e-05, + "loss": 0.4126, + "step": 30414 + }, + { + "epoch": 1.17, + "learning_rate": 3.053026417030008e-05, + "loss": 0.3956, + "step": 30415 + }, + { + "epoch": 1.17, + "learning_rate": 3.052962298025135e-05, + "loss": 0.4075, + "step": 30416 + }, + { + "epoch": 1.17, + "learning_rate": 3.052898179020261e-05, + "loss": 0.5837, + "step": 30417 + }, + { + "epoch": 1.17, + "learning_rate": 3.0528340600153885e-05, + "loss": 0.4565, + "step": 30418 + }, + { + "epoch": 1.17, + "learning_rate": 3.052769941010516e-05, + "loss": 0.3964, + "step": 30419 + }, + { + "epoch": 1.17, + "learning_rate": 3.052705822005643e-05, + "loss": 0.3753, + "step": 30420 + }, + { + "epoch": 1.17, + "learning_rate": 3.05264170300077e-05, + "loss": 0.358, + "step": 30421 + }, + { + "epoch": 1.17, + "learning_rate": 3.0525775839958965e-05, + "loss": 0.2774, + "step": 30422 + }, + { + "epoch": 1.17, + "learning_rate": 3.052513464991024e-05, + "loss": 0.2279, + "step": 30423 + }, + { + "epoch": 1.17, + "learning_rate": 3.05244934598615e-05, + "loss": 0.4031, + "step": 30424 + }, + { + "epoch": 1.17, + "learning_rate": 3.0523852269812773e-05, + "loss": 0.3196, + "step": 30425 + }, + { + "epoch": 1.17, + "learning_rate": 3.052321107976404e-05, + "loss": 0.2825, + "step": 30426 + }, + { + "epoch": 1.17, + "learning_rate": 3.052256988971531e-05, + "loss": 0.3986, + "step": 30427 + }, + { + "epoch": 1.17, + "learning_rate": 3.052192869966658e-05, + "loss": 0.2879, + "step": 30428 + }, + { + "epoch": 1.17, + "learning_rate": 3.0521287509617854e-05, + "loss": 0.4684, + "step": 30429 + }, + { + "epoch": 1.17, + "learning_rate": 3.0520646319569125e-05, + "loss": 0.3741, + "step": 30430 + }, + { + "epoch": 1.17, + "learning_rate": 3.052000512952039e-05, + "loss": 0.4456, + "step": 30431 + }, + { + "epoch": 1.17, + "learning_rate": 3.051936393947166e-05, + "loss": 0.4496, + "step": 30432 + }, + { + "epoch": 1.17, + "learning_rate": 3.051872274942293e-05, + "loss": 0.2871, + "step": 30433 + }, + { + "epoch": 1.17, + "learning_rate": 3.05180815593742e-05, + "loss": 0.2959, + "step": 30434 + }, + { + "epoch": 1.17, + "learning_rate": 3.0517440369325467e-05, + "loss": 0.338, + "step": 30435 + }, + { + "epoch": 1.17, + "learning_rate": 3.051679917927674e-05, + "loss": 0.3088, + "step": 30436 + }, + { + "epoch": 1.17, + "learning_rate": 3.051615798922801e-05, + "loss": 0.407, + "step": 30437 + }, + { + "epoch": 1.17, + "learning_rate": 3.051551679917928e-05, + "loss": 0.4652, + "step": 30438 + }, + { + "epoch": 1.17, + "learning_rate": 3.051487560913055e-05, + "loss": 0.4142, + "step": 30439 + }, + { + "epoch": 1.17, + "learning_rate": 3.0514234419081815e-05, + "loss": 0.3291, + "step": 30440 + }, + { + "epoch": 1.17, + "learning_rate": 3.0513593229033087e-05, + "loss": 0.453, + "step": 30441 + }, + { + "epoch": 1.17, + "learning_rate": 3.0512952038984355e-05, + "loss": 0.456, + "step": 30442 + }, + { + "epoch": 1.17, + "learning_rate": 3.0512310848935627e-05, + "loss": 0.298, + "step": 30443 + }, + { + "epoch": 1.17, + "learning_rate": 3.0511669658886892e-05, + "loss": 0.2641, + "step": 30444 + }, + { + "epoch": 1.17, + "learning_rate": 3.0511028468838164e-05, + "loss": 0.3347, + "step": 30445 + }, + { + "epoch": 1.17, + "learning_rate": 3.051038727878944e-05, + "loss": 0.2768, + "step": 30446 + }, + { + "epoch": 1.17, + "learning_rate": 3.0509746088740704e-05, + "loss": 0.3803, + "step": 30447 + }, + { + "epoch": 1.17, + "learning_rate": 3.0509104898691976e-05, + "loss": 0.4126, + "step": 30448 + }, + { + "epoch": 1.17, + "learning_rate": 3.0508463708643244e-05, + "loss": 0.3742, + "step": 30449 + }, + { + "epoch": 1.17, + "learning_rate": 3.0507822518594516e-05, + "loss": 0.3654, + "step": 30450 + }, + { + "epoch": 1.17, + "learning_rate": 3.050718132854578e-05, + "loss": 0.496, + "step": 30451 + }, + { + "epoch": 1.17, + "learning_rate": 3.0506540138497052e-05, + "loss": 0.361, + "step": 30452 + }, + { + "epoch": 1.17, + "learning_rate": 3.050589894844832e-05, + "loss": 0.5475, + "step": 30453 + }, + { + "epoch": 1.17, + "learning_rate": 3.0505257758399592e-05, + "loss": 0.407, + "step": 30454 + }, + { + "epoch": 1.17, + "learning_rate": 3.0504616568350864e-05, + "loss": 0.3289, + "step": 30455 + }, + { + "epoch": 1.17, + "learning_rate": 3.050397537830213e-05, + "loss": 0.5087, + "step": 30456 + }, + { + "epoch": 1.17, + "learning_rate": 3.05033341882534e-05, + "loss": 0.489, + "step": 30457 + }, + { + "epoch": 1.17, + "learning_rate": 3.050269299820467e-05, + "loss": 0.5382, + "step": 30458 + }, + { + "epoch": 1.17, + "learning_rate": 3.050205180815594e-05, + "loss": 0.4243, + "step": 30459 + }, + { + "epoch": 1.17, + "learning_rate": 3.0501410618107206e-05, + "loss": 0.4607, + "step": 30460 + }, + { + "epoch": 1.17, + "learning_rate": 3.0500769428058477e-05, + "loss": 0.3518, + "step": 30461 + }, + { + "epoch": 1.17, + "learning_rate": 3.0500128238009746e-05, + "loss": 0.416, + "step": 30462 + }, + { + "epoch": 1.17, + "learning_rate": 3.0499487047961017e-05, + "loss": 0.5026, + "step": 30463 + }, + { + "epoch": 1.17, + "learning_rate": 3.049884585791229e-05, + "loss": 0.3714, + "step": 30464 + }, + { + "epoch": 1.17, + "learning_rate": 3.0498204667863557e-05, + "loss": 0.348, + "step": 30465 + }, + { + "epoch": 1.17, + "learning_rate": 3.049756347781483e-05, + "loss": 0.4407, + "step": 30466 + }, + { + "epoch": 1.17, + "learning_rate": 3.0496922287766094e-05, + "loss": 0.4834, + "step": 30467 + }, + { + "epoch": 1.17, + "learning_rate": 3.0496281097717366e-05, + "loss": 0.2714, + "step": 30468 + }, + { + "epoch": 1.17, + "learning_rate": 3.0495639907668634e-05, + "loss": 0.5542, + "step": 30469 + }, + { + "epoch": 1.17, + "learning_rate": 3.0494998717619906e-05, + "loss": 0.3103, + "step": 30470 + }, + { + "epoch": 1.17, + "learning_rate": 3.049435752757117e-05, + "loss": 0.2928, + "step": 30471 + }, + { + "epoch": 1.17, + "learning_rate": 3.0493716337522443e-05, + "loss": 0.4471, + "step": 30472 + }, + { + "epoch": 1.17, + "learning_rate": 3.0493075147473714e-05, + "loss": 0.468, + "step": 30473 + }, + { + "epoch": 1.17, + "learning_rate": 3.0492433957424983e-05, + "loss": 0.3499, + "step": 30474 + }, + { + "epoch": 1.17, + "learning_rate": 3.0491792767376254e-05, + "loss": 0.4032, + "step": 30475 + }, + { + "epoch": 1.17, + "learning_rate": 3.049115157732752e-05, + "loss": 0.3682, + "step": 30476 + }, + { + "epoch": 1.17, + "learning_rate": 3.049051038727879e-05, + "loss": 0.2924, + "step": 30477 + }, + { + "epoch": 1.17, + "learning_rate": 3.048986919723006e-05, + "loss": 0.4564, + "step": 30478 + }, + { + "epoch": 1.17, + "learning_rate": 3.048922800718133e-05, + "loss": 0.2642, + "step": 30479 + }, + { + "epoch": 1.17, + "learning_rate": 3.0488586817132596e-05, + "loss": 0.4755, + "step": 30480 + }, + { + "epoch": 1.17, + "learning_rate": 3.0487945627083868e-05, + "loss": 0.4288, + "step": 30481 + }, + { + "epoch": 1.17, + "learning_rate": 3.0487304437035143e-05, + "loss": 0.3358, + "step": 30482 + }, + { + "epoch": 1.17, + "learning_rate": 3.0486663246986408e-05, + "loss": 0.2935, + "step": 30483 + }, + { + "epoch": 1.17, + "learning_rate": 3.048602205693768e-05, + "loss": 0.5242, + "step": 30484 + }, + { + "epoch": 1.17, + "learning_rate": 3.0485380866888948e-05, + "loss": 0.584, + "step": 30485 + }, + { + "epoch": 1.17, + "learning_rate": 3.048473967684022e-05, + "loss": 0.3671, + "step": 30486 + }, + { + "epoch": 1.17, + "learning_rate": 3.0484098486791484e-05, + "loss": 0.4761, + "step": 30487 + }, + { + "epoch": 1.17, + "learning_rate": 3.0483457296742756e-05, + "loss": 0.249, + "step": 30488 + }, + { + "epoch": 1.17, + "learning_rate": 3.0482816106694024e-05, + "loss": 0.3487, + "step": 30489 + }, + { + "epoch": 1.17, + "learning_rate": 3.0482174916645296e-05, + "loss": 0.4825, + "step": 30490 + }, + { + "epoch": 1.17, + "learning_rate": 3.0481533726596568e-05, + "loss": 0.2942, + "step": 30491 + }, + { + "epoch": 1.17, + "learning_rate": 3.0480892536547833e-05, + "loss": 0.3066, + "step": 30492 + }, + { + "epoch": 1.17, + "learning_rate": 3.0480251346499105e-05, + "loss": 0.4778, + "step": 30493 + }, + { + "epoch": 1.17, + "learning_rate": 3.0479610156450373e-05, + "loss": 0.4723, + "step": 30494 + }, + { + "epoch": 1.17, + "learning_rate": 3.0478968966401645e-05, + "loss": 0.4162, + "step": 30495 + }, + { + "epoch": 1.17, + "learning_rate": 3.047832777635291e-05, + "loss": 0.4199, + "step": 30496 + }, + { + "epoch": 1.17, + "learning_rate": 3.047768658630418e-05, + "loss": 0.4997, + "step": 30497 + }, + { + "epoch": 1.17, + "learning_rate": 3.047704539625545e-05, + "loss": 0.3971, + "step": 30498 + }, + { + "epoch": 1.17, + "learning_rate": 3.047640420620672e-05, + "loss": 0.5926, + "step": 30499 + }, + { + "epoch": 1.17, + "learning_rate": 3.0475763016157993e-05, + "loss": 0.4427, + "step": 30500 + }, + { + "epoch": 1.17, + "learning_rate": 3.047512182610926e-05, + "loss": 0.4553, + "step": 30501 + }, + { + "epoch": 1.17, + "learning_rate": 3.0474480636060533e-05, + "loss": 0.4383, + "step": 30502 + }, + { + "epoch": 1.17, + "learning_rate": 3.0473839446011798e-05, + "loss": 0.3792, + "step": 30503 + }, + { + "epoch": 1.17, + "learning_rate": 3.047319825596307e-05, + "loss": 0.3448, + "step": 30504 + }, + { + "epoch": 1.17, + "learning_rate": 3.0472557065914338e-05, + "loss": 0.3624, + "step": 30505 + }, + { + "epoch": 1.17, + "learning_rate": 3.047191587586561e-05, + "loss": 0.3598, + "step": 30506 + }, + { + "epoch": 1.17, + "learning_rate": 3.0471274685816875e-05, + "loss": 0.3941, + "step": 30507 + }, + { + "epoch": 1.17, + "learning_rate": 3.0470633495768146e-05, + "loss": 0.3645, + "step": 30508 + }, + { + "epoch": 1.17, + "learning_rate": 3.0469992305719418e-05, + "loss": 0.2805, + "step": 30509 + }, + { + "epoch": 1.17, + "learning_rate": 3.0469351115670686e-05, + "loss": 0.3577, + "step": 30510 + }, + { + "epoch": 1.17, + "learning_rate": 3.0468709925621958e-05, + "loss": 0.446, + "step": 30511 + }, + { + "epoch": 1.17, + "learning_rate": 3.0468068735573223e-05, + "loss": 0.4113, + "step": 30512 + }, + { + "epoch": 1.17, + "learning_rate": 3.0467427545524495e-05, + "loss": 0.4284, + "step": 30513 + }, + { + "epoch": 1.17, + "learning_rate": 3.0466786355475763e-05, + "loss": 0.4215, + "step": 30514 + }, + { + "epoch": 1.17, + "learning_rate": 3.0466145165427035e-05, + "loss": 0.5017, + "step": 30515 + }, + { + "epoch": 1.17, + "learning_rate": 3.04655039753783e-05, + "loss": 0.4614, + "step": 30516 + }, + { + "epoch": 1.17, + "learning_rate": 3.046486278532957e-05, + "loss": 0.3581, + "step": 30517 + }, + { + "epoch": 1.17, + "learning_rate": 3.0464221595280847e-05, + "loss": 0.3622, + "step": 30518 + }, + { + "epoch": 1.17, + "learning_rate": 3.046358040523211e-05, + "loss": 0.3543, + "step": 30519 + }, + { + "epoch": 1.17, + "learning_rate": 3.0462939215183383e-05, + "loss": 0.3423, + "step": 30520 + }, + { + "epoch": 1.17, + "learning_rate": 3.046229802513465e-05, + "loss": 0.4228, + "step": 30521 + }, + { + "epoch": 1.17, + "learning_rate": 3.0461656835085923e-05, + "loss": 0.4184, + "step": 30522 + }, + { + "epoch": 1.17, + "learning_rate": 3.0461015645037188e-05, + "loss": 0.4165, + "step": 30523 + }, + { + "epoch": 1.17, + "learning_rate": 3.046037445498846e-05, + "loss": 0.4333, + "step": 30524 + }, + { + "epoch": 1.17, + "learning_rate": 3.045973326493973e-05, + "loss": 0.4691, + "step": 30525 + }, + { + "epoch": 1.17, + "learning_rate": 3.0459092074891e-05, + "loss": 0.4182, + "step": 30526 + }, + { + "epoch": 1.17, + "learning_rate": 3.0458450884842272e-05, + "loss": 0.3716, + "step": 30527 + }, + { + "epoch": 1.17, + "learning_rate": 3.0457809694793537e-05, + "loss": 0.4157, + "step": 30528 + }, + { + "epoch": 1.17, + "learning_rate": 3.045716850474481e-05, + "loss": 0.4068, + "step": 30529 + }, + { + "epoch": 1.17, + "learning_rate": 3.0456527314696077e-05, + "loss": 0.4251, + "step": 30530 + }, + { + "epoch": 1.17, + "learning_rate": 3.045588612464735e-05, + "loss": 0.2732, + "step": 30531 + }, + { + "epoch": 1.17, + "learning_rate": 3.0455244934598613e-05, + "loss": 0.5138, + "step": 30532 + }, + { + "epoch": 1.17, + "learning_rate": 3.0454603744549885e-05, + "loss": 0.3134, + "step": 30533 + }, + { + "epoch": 1.17, + "learning_rate": 3.0453962554501153e-05, + "loss": 0.4868, + "step": 30534 + }, + { + "epoch": 1.17, + "learning_rate": 3.0453321364452425e-05, + "loss": 0.4306, + "step": 30535 + }, + { + "epoch": 1.17, + "learning_rate": 3.0452680174403697e-05, + "loss": 0.3596, + "step": 30536 + }, + { + "epoch": 1.17, + "learning_rate": 3.0452038984354965e-05, + "loss": 0.4026, + "step": 30537 + }, + { + "epoch": 1.17, + "learning_rate": 3.0451397794306237e-05, + "loss": 0.3866, + "step": 30538 + }, + { + "epoch": 1.17, + "learning_rate": 3.0450756604257502e-05, + "loss": 0.3997, + "step": 30539 + }, + { + "epoch": 1.17, + "learning_rate": 3.0450115414208774e-05, + "loss": 0.4577, + "step": 30540 + }, + { + "epoch": 1.17, + "learning_rate": 3.0449474224160042e-05, + "loss": 0.4003, + "step": 30541 + }, + { + "epoch": 1.17, + "learning_rate": 3.0448833034111314e-05, + "loss": 0.3842, + "step": 30542 + }, + { + "epoch": 1.17, + "learning_rate": 3.044819184406258e-05, + "loss": 0.4116, + "step": 30543 + }, + { + "epoch": 1.17, + "learning_rate": 3.044755065401385e-05, + "loss": 0.3079, + "step": 30544 + }, + { + "epoch": 1.17, + "learning_rate": 3.0446909463965122e-05, + "loss": 0.3913, + "step": 30545 + }, + { + "epoch": 1.17, + "learning_rate": 3.044626827391639e-05, + "loss": 0.5106, + "step": 30546 + }, + { + "epoch": 1.17, + "learning_rate": 3.0445627083867662e-05, + "loss": 0.3814, + "step": 30547 + }, + { + "epoch": 1.17, + "learning_rate": 3.0444985893818927e-05, + "loss": 0.4155, + "step": 30548 + }, + { + "epoch": 1.17, + "learning_rate": 3.04443447037702e-05, + "loss": 0.3405, + "step": 30549 + }, + { + "epoch": 1.17, + "learning_rate": 3.0443703513721467e-05, + "loss": 0.3345, + "step": 30550 + }, + { + "epoch": 1.17, + "learning_rate": 3.044306232367274e-05, + "loss": 0.3452, + "step": 30551 + }, + { + "epoch": 1.17, + "learning_rate": 3.0442421133624004e-05, + "loss": 0.3768, + "step": 30552 + }, + { + "epoch": 1.17, + "learning_rate": 3.0441779943575275e-05, + "loss": 0.3884, + "step": 30553 + }, + { + "epoch": 1.17, + "learning_rate": 3.044113875352655e-05, + "loss": 0.4016, + "step": 30554 + }, + { + "epoch": 1.17, + "learning_rate": 3.0440497563477815e-05, + "loss": 0.3641, + "step": 30555 + }, + { + "epoch": 1.17, + "learning_rate": 3.0439856373429087e-05, + "loss": 0.468, + "step": 30556 + }, + { + "epoch": 1.17, + "learning_rate": 3.0439215183380355e-05, + "loss": 0.5186, + "step": 30557 + }, + { + "epoch": 1.17, + "learning_rate": 3.0438573993331627e-05, + "loss": 0.4849, + "step": 30558 + }, + { + "epoch": 1.17, + "learning_rate": 3.0437932803282892e-05, + "loss": 0.4798, + "step": 30559 + }, + { + "epoch": 1.17, + "learning_rate": 3.0437291613234164e-05, + "loss": 0.4499, + "step": 30560 + }, + { + "epoch": 1.17, + "learning_rate": 3.0436650423185432e-05, + "loss": 0.5458, + "step": 30561 + }, + { + "epoch": 1.18, + "learning_rate": 3.0436009233136704e-05, + "loss": 0.3246, + "step": 30562 + }, + { + "epoch": 1.18, + "learning_rate": 3.0435368043087976e-05, + "loss": 0.4157, + "step": 30563 + }, + { + "epoch": 1.18, + "learning_rate": 3.043472685303924e-05, + "loss": 0.4848, + "step": 30564 + }, + { + "epoch": 1.18, + "learning_rate": 3.0434085662990512e-05, + "loss": 0.3401, + "step": 30565 + }, + { + "epoch": 1.18, + "learning_rate": 3.043344447294178e-05, + "loss": 0.4513, + "step": 30566 + }, + { + "epoch": 1.18, + "learning_rate": 3.0432803282893052e-05, + "loss": 0.3409, + "step": 30567 + }, + { + "epoch": 1.18, + "learning_rate": 3.0432162092844317e-05, + "loss": 0.4755, + "step": 30568 + }, + { + "epoch": 1.18, + "learning_rate": 3.043152090279559e-05, + "loss": 0.4152, + "step": 30569 + }, + { + "epoch": 1.18, + "learning_rate": 3.0430879712746857e-05, + "loss": 0.3567, + "step": 30570 + }, + { + "epoch": 1.18, + "learning_rate": 3.043023852269813e-05, + "loss": 0.4484, + "step": 30571 + }, + { + "epoch": 1.18, + "learning_rate": 3.04295973326494e-05, + "loss": 0.3776, + "step": 30572 + }, + { + "epoch": 1.18, + "learning_rate": 3.042895614260067e-05, + "loss": 0.2606, + "step": 30573 + }, + { + "epoch": 1.18, + "learning_rate": 3.042831495255194e-05, + "loss": 0.3352, + "step": 30574 + }, + { + "epoch": 1.18, + "learning_rate": 3.0427673762503206e-05, + "loss": 0.4212, + "step": 30575 + }, + { + "epoch": 1.18, + "learning_rate": 3.0427032572454477e-05, + "loss": 0.4143, + "step": 30576 + }, + { + "epoch": 1.18, + "learning_rate": 3.0426391382405746e-05, + "loss": 0.3735, + "step": 30577 + }, + { + "epoch": 1.18, + "learning_rate": 3.0425750192357017e-05, + "loss": 0.4798, + "step": 30578 + }, + { + "epoch": 1.18, + "learning_rate": 3.0425109002308282e-05, + "loss": 0.4647, + "step": 30579 + }, + { + "epoch": 1.18, + "learning_rate": 3.0424467812259554e-05, + "loss": 0.5376, + "step": 30580 + }, + { + "epoch": 1.18, + "learning_rate": 3.0423826622210826e-05, + "loss": 0.362, + "step": 30581 + }, + { + "epoch": 1.18, + "learning_rate": 3.0423185432162094e-05, + "loss": 0.4301, + "step": 30582 + }, + { + "epoch": 1.18, + "learning_rate": 3.0422544242113366e-05, + "loss": 0.3926, + "step": 30583 + }, + { + "epoch": 1.18, + "learning_rate": 3.042190305206463e-05, + "loss": 0.4555, + "step": 30584 + }, + { + "epoch": 1.18, + "learning_rate": 3.0421261862015903e-05, + "loss": 0.5168, + "step": 30585 + }, + { + "epoch": 1.18, + "learning_rate": 3.042062067196717e-05, + "loss": 0.4688, + "step": 30586 + }, + { + "epoch": 1.18, + "learning_rate": 3.0419979481918443e-05, + "loss": 0.445, + "step": 30587 + }, + { + "epoch": 1.18, + "learning_rate": 3.0419338291869708e-05, + "loss": 0.3063, + "step": 30588 + }, + { + "epoch": 1.18, + "learning_rate": 3.0418697101820983e-05, + "loss": 0.3169, + "step": 30589 + }, + { + "epoch": 1.18, + "learning_rate": 3.0418055911772254e-05, + "loss": 0.3339, + "step": 30590 + }, + { + "epoch": 1.18, + "learning_rate": 3.041741472172352e-05, + "loss": 0.4515, + "step": 30591 + }, + { + "epoch": 1.18, + "learning_rate": 3.041677353167479e-05, + "loss": 0.3699, + "step": 30592 + }, + { + "epoch": 1.18, + "learning_rate": 3.041613234162606e-05, + "loss": 0.3395, + "step": 30593 + }, + { + "epoch": 1.18, + "learning_rate": 3.041549115157733e-05, + "loss": 0.4287, + "step": 30594 + }, + { + "epoch": 1.18, + "learning_rate": 3.0414849961528596e-05, + "loss": 0.4835, + "step": 30595 + }, + { + "epoch": 1.18, + "learning_rate": 3.0414208771479868e-05, + "loss": 0.3327, + "step": 30596 + }, + { + "epoch": 1.18, + "learning_rate": 3.0413567581431136e-05, + "loss": 0.3539, + "step": 30597 + }, + { + "epoch": 1.18, + "learning_rate": 3.0412926391382408e-05, + "loss": 0.3096, + "step": 30598 + }, + { + "epoch": 1.18, + "learning_rate": 3.041228520133368e-05, + "loss": 0.3923, + "step": 30599 + }, + { + "epoch": 1.18, + "learning_rate": 3.0411644011284944e-05, + "loss": 0.2788, + "step": 30600 + }, + { + "epoch": 1.18, + "learning_rate": 3.0411002821236216e-05, + "loss": 0.3665, + "step": 30601 + }, + { + "epoch": 1.18, + "learning_rate": 3.0410361631187484e-05, + "loss": 0.3036, + "step": 30602 + }, + { + "epoch": 1.18, + "learning_rate": 3.0409720441138756e-05, + "loss": 0.3507, + "step": 30603 + }, + { + "epoch": 1.18, + "learning_rate": 3.040907925109002e-05, + "loss": 0.3449, + "step": 30604 + }, + { + "epoch": 1.18, + "learning_rate": 3.0408438061041293e-05, + "loss": 0.3972, + "step": 30605 + }, + { + "epoch": 1.18, + "learning_rate": 3.040779687099256e-05, + "loss": 0.3712, + "step": 30606 + }, + { + "epoch": 1.18, + "learning_rate": 3.0407155680943833e-05, + "loss": 0.3096, + "step": 30607 + }, + { + "epoch": 1.18, + "learning_rate": 3.0406514490895105e-05, + "loss": 0.2616, + "step": 30608 + }, + { + "epoch": 1.18, + "learning_rate": 3.0405873300846373e-05, + "loss": 0.3718, + "step": 30609 + }, + { + "epoch": 1.18, + "learning_rate": 3.0405232110797645e-05, + "loss": 0.438, + "step": 30610 + }, + { + "epoch": 1.18, + "learning_rate": 3.040459092074891e-05, + "loss": 0.4578, + "step": 30611 + }, + { + "epoch": 1.18, + "learning_rate": 3.040394973070018e-05, + "loss": 0.3799, + "step": 30612 + }, + { + "epoch": 1.18, + "learning_rate": 3.040330854065145e-05, + "loss": 0.4341, + "step": 30613 + }, + { + "epoch": 1.18, + "learning_rate": 3.040266735060272e-05, + "loss": 0.4656, + "step": 30614 + }, + { + "epoch": 1.18, + "learning_rate": 3.0402026160553986e-05, + "loss": 0.4175, + "step": 30615 + }, + { + "epoch": 1.18, + "learning_rate": 3.0401384970505258e-05, + "loss": 0.2803, + "step": 30616 + }, + { + "epoch": 1.18, + "learning_rate": 3.040074378045653e-05, + "loss": 0.4597, + "step": 30617 + }, + { + "epoch": 1.18, + "learning_rate": 3.0400102590407798e-05, + "loss": 0.5247, + "step": 30618 + }, + { + "epoch": 1.18, + "learning_rate": 3.039946140035907e-05, + "loss": 0.4817, + "step": 30619 + }, + { + "epoch": 1.18, + "learning_rate": 3.0398820210310335e-05, + "loss": 0.3786, + "step": 30620 + }, + { + "epoch": 1.18, + "learning_rate": 3.0398179020261606e-05, + "loss": 0.3589, + "step": 30621 + }, + { + "epoch": 1.18, + "learning_rate": 3.0397537830212875e-05, + "loss": 0.3909, + "step": 30622 + }, + { + "epoch": 1.18, + "learning_rate": 3.0396896640164146e-05, + "loss": 0.422, + "step": 30623 + }, + { + "epoch": 1.18, + "learning_rate": 3.0396255450115418e-05, + "loss": 0.3566, + "step": 30624 + }, + { + "epoch": 1.18, + "learning_rate": 3.0395614260066687e-05, + "loss": 0.5417, + "step": 30625 + }, + { + "epoch": 1.18, + "learning_rate": 3.0394973070017958e-05, + "loss": 0.3837, + "step": 30626 + }, + { + "epoch": 1.18, + "learning_rate": 3.0394331879969223e-05, + "loss": 0.3229, + "step": 30627 + }, + { + "epoch": 1.18, + "learning_rate": 3.0393690689920495e-05, + "loss": 0.3893, + "step": 30628 + }, + { + "epoch": 1.18, + "learning_rate": 3.0393049499871763e-05, + "loss": 0.5747, + "step": 30629 + }, + { + "epoch": 1.18, + "learning_rate": 3.0392408309823035e-05, + "loss": 0.3655, + "step": 30630 + }, + { + "epoch": 1.18, + "learning_rate": 3.03917671197743e-05, + "loss": 0.344, + "step": 30631 + }, + { + "epoch": 1.18, + "learning_rate": 3.039112592972557e-05, + "loss": 0.3513, + "step": 30632 + }, + { + "epoch": 1.18, + "learning_rate": 3.0390484739676843e-05, + "loss": 0.4215, + "step": 30633 + }, + { + "epoch": 1.18, + "learning_rate": 3.038984354962811e-05, + "loss": 0.2947, + "step": 30634 + }, + { + "epoch": 1.18, + "learning_rate": 3.0389202359579383e-05, + "loss": 0.5045, + "step": 30635 + }, + { + "epoch": 1.18, + "learning_rate": 3.0388561169530648e-05, + "loss": 0.2809, + "step": 30636 + }, + { + "epoch": 1.18, + "learning_rate": 3.038791997948192e-05, + "loss": 0.3693, + "step": 30637 + }, + { + "epoch": 1.18, + "learning_rate": 3.038727878943319e-05, + "loss": 0.5922, + "step": 30638 + }, + { + "epoch": 1.18, + "learning_rate": 3.038663759938446e-05, + "loss": 0.3457, + "step": 30639 + }, + { + "epoch": 1.18, + "learning_rate": 3.0385996409335725e-05, + "loss": 0.3852, + "step": 30640 + }, + { + "epoch": 1.18, + "learning_rate": 3.0385355219286997e-05, + "loss": 0.4009, + "step": 30641 + }, + { + "epoch": 1.18, + "learning_rate": 3.0384714029238272e-05, + "loss": 0.4103, + "step": 30642 + }, + { + "epoch": 1.18, + "learning_rate": 3.0384072839189537e-05, + "loss": 0.4651, + "step": 30643 + }, + { + "epoch": 1.18, + "learning_rate": 3.038343164914081e-05, + "loss": 0.3265, + "step": 30644 + }, + { + "epoch": 1.18, + "learning_rate": 3.0382790459092077e-05, + "loss": 0.4228, + "step": 30645 + }, + { + "epoch": 1.18, + "learning_rate": 3.038214926904335e-05, + "loss": 0.3325, + "step": 30646 + }, + { + "epoch": 1.18, + "learning_rate": 3.0381508078994613e-05, + "loss": 0.3746, + "step": 30647 + }, + { + "epoch": 1.18, + "learning_rate": 3.0380866888945885e-05, + "loss": 0.2501, + "step": 30648 + }, + { + "epoch": 1.18, + "learning_rate": 3.0380225698897154e-05, + "loss": 0.3093, + "step": 30649 + }, + { + "epoch": 1.18, + "learning_rate": 3.0379584508848425e-05, + "loss": 0.3934, + "step": 30650 + }, + { + "epoch": 1.18, + "learning_rate": 3.0378943318799697e-05, + "loss": 0.3785, + "step": 30651 + }, + { + "epoch": 1.18, + "learning_rate": 3.0378302128750962e-05, + "loss": 0.2963, + "step": 30652 + }, + { + "epoch": 1.18, + "learning_rate": 3.0377660938702234e-05, + "loss": 0.4576, + "step": 30653 + }, + { + "epoch": 1.18, + "learning_rate": 3.0377019748653502e-05, + "loss": 0.3008, + "step": 30654 + }, + { + "epoch": 1.18, + "learning_rate": 3.0376378558604774e-05, + "loss": 0.407, + "step": 30655 + }, + { + "epoch": 1.18, + "learning_rate": 3.037573736855604e-05, + "loss": 0.5901, + "step": 30656 + }, + { + "epoch": 1.18, + "learning_rate": 3.037509617850731e-05, + "loss": 0.4373, + "step": 30657 + }, + { + "epoch": 1.18, + "learning_rate": 3.037445498845858e-05, + "loss": 0.373, + "step": 30658 + }, + { + "epoch": 1.18, + "learning_rate": 3.037381379840985e-05, + "loss": 0.4218, + "step": 30659 + }, + { + "epoch": 1.18, + "learning_rate": 3.0373172608361122e-05, + "loss": 0.3394, + "step": 30660 + }, + { + "epoch": 1.18, + "learning_rate": 3.037253141831239e-05, + "loss": 0.324, + "step": 30661 + }, + { + "epoch": 1.18, + "learning_rate": 3.0371890228263662e-05, + "loss": 0.3146, + "step": 30662 + }, + { + "epoch": 1.18, + "learning_rate": 3.0371249038214927e-05, + "loss": 0.3945, + "step": 30663 + }, + { + "epoch": 1.18, + "learning_rate": 3.03706078481662e-05, + "loss": 0.3929, + "step": 30664 + }, + { + "epoch": 1.18, + "learning_rate": 3.0369966658117467e-05, + "loss": 0.3259, + "step": 30665 + }, + { + "epoch": 1.18, + "learning_rate": 3.036932546806874e-05, + "loss": 0.4832, + "step": 30666 + }, + { + "epoch": 1.18, + "learning_rate": 3.0368684278020004e-05, + "loss": 0.4039, + "step": 30667 + }, + { + "epoch": 1.18, + "learning_rate": 3.0368043087971275e-05, + "loss": 0.4711, + "step": 30668 + }, + { + "epoch": 1.18, + "learning_rate": 3.0367401897922547e-05, + "loss": 0.5081, + "step": 30669 + }, + { + "epoch": 1.18, + "learning_rate": 3.0366760707873816e-05, + "loss": 0.4506, + "step": 30670 + }, + { + "epoch": 1.18, + "learning_rate": 3.0366119517825087e-05, + "loss": 0.4827, + "step": 30671 + }, + { + "epoch": 1.18, + "learning_rate": 3.0365478327776352e-05, + "loss": 0.3293, + "step": 30672 + }, + { + "epoch": 1.18, + "learning_rate": 3.0364837137727624e-05, + "loss": 0.4488, + "step": 30673 + }, + { + "epoch": 1.18, + "learning_rate": 3.0364195947678892e-05, + "loss": 0.45, + "step": 30674 + }, + { + "epoch": 1.18, + "learning_rate": 3.0363554757630164e-05, + "loss": 0.4363, + "step": 30675 + }, + { + "epoch": 1.18, + "learning_rate": 3.036291356758143e-05, + "loss": 0.4475, + "step": 30676 + }, + { + "epoch": 1.18, + "learning_rate": 3.03622723775327e-05, + "loss": 0.4594, + "step": 30677 + }, + { + "epoch": 1.18, + "learning_rate": 3.0361631187483976e-05, + "loss": 0.3266, + "step": 30678 + }, + { + "epoch": 1.18, + "learning_rate": 3.036098999743524e-05, + "loss": 0.3145, + "step": 30679 + }, + { + "epoch": 1.18, + "learning_rate": 3.0360348807386512e-05, + "loss": 0.3411, + "step": 30680 + }, + { + "epoch": 1.18, + "learning_rate": 3.035970761733778e-05, + "loss": 0.2846, + "step": 30681 + }, + { + "epoch": 1.18, + "learning_rate": 3.0359066427289052e-05, + "loss": 0.4442, + "step": 30682 + }, + { + "epoch": 1.18, + "learning_rate": 3.0358425237240317e-05, + "loss": 0.4197, + "step": 30683 + }, + { + "epoch": 1.18, + "learning_rate": 3.035778404719159e-05, + "loss": 0.5106, + "step": 30684 + }, + { + "epoch": 1.18, + "learning_rate": 3.0357142857142857e-05, + "loss": 0.3747, + "step": 30685 + }, + { + "epoch": 1.18, + "learning_rate": 3.035650166709413e-05, + "loss": 0.3834, + "step": 30686 + }, + { + "epoch": 1.18, + "learning_rate": 3.03558604770454e-05, + "loss": 0.4395, + "step": 30687 + }, + { + "epoch": 1.18, + "learning_rate": 3.0355219286996666e-05, + "loss": 0.3778, + "step": 30688 + }, + { + "epoch": 1.18, + "learning_rate": 3.0354578096947937e-05, + "loss": 0.4858, + "step": 30689 + }, + { + "epoch": 1.18, + "learning_rate": 3.0353936906899206e-05, + "loss": 0.431, + "step": 30690 + }, + { + "epoch": 1.18, + "learning_rate": 3.0353295716850478e-05, + "loss": 0.3505, + "step": 30691 + }, + { + "epoch": 1.18, + "learning_rate": 3.0352654526801742e-05, + "loss": 0.3191, + "step": 30692 + }, + { + "epoch": 1.18, + "learning_rate": 3.0352013336753014e-05, + "loss": 0.4481, + "step": 30693 + }, + { + "epoch": 1.18, + "learning_rate": 3.0351372146704283e-05, + "loss": 0.2846, + "step": 30694 + }, + { + "epoch": 1.18, + "learning_rate": 3.0350730956655554e-05, + "loss": 0.3191, + "step": 30695 + }, + { + "epoch": 1.18, + "learning_rate": 3.0350089766606826e-05, + "loss": 0.4713, + "step": 30696 + }, + { + "epoch": 1.18, + "learning_rate": 3.0349448576558094e-05, + "loss": 0.4066, + "step": 30697 + }, + { + "epoch": 1.18, + "learning_rate": 3.0348807386509366e-05, + "loss": 0.2618, + "step": 30698 + }, + { + "epoch": 1.18, + "learning_rate": 3.034816619646063e-05, + "loss": 0.4379, + "step": 30699 + }, + { + "epoch": 1.18, + "learning_rate": 3.0347525006411903e-05, + "loss": 0.4585, + "step": 30700 + }, + { + "epoch": 1.18, + "learning_rate": 3.034688381636317e-05, + "loss": 0.3816, + "step": 30701 + }, + { + "epoch": 1.18, + "learning_rate": 3.0346242626314443e-05, + "loss": 0.3446, + "step": 30702 + }, + { + "epoch": 1.18, + "learning_rate": 3.0345601436265708e-05, + "loss": 0.4498, + "step": 30703 + }, + { + "epoch": 1.18, + "learning_rate": 3.034496024621698e-05, + "loss": 0.4231, + "step": 30704 + }, + { + "epoch": 1.18, + "learning_rate": 3.034431905616825e-05, + "loss": 0.3465, + "step": 30705 + }, + { + "epoch": 1.18, + "learning_rate": 3.034367786611952e-05, + "loss": 0.3643, + "step": 30706 + }, + { + "epoch": 1.18, + "learning_rate": 3.034303667607079e-05, + "loss": 0.5303, + "step": 30707 + }, + { + "epoch": 1.18, + "learning_rate": 3.0342395486022056e-05, + "loss": 0.3807, + "step": 30708 + }, + { + "epoch": 1.18, + "learning_rate": 3.0341754295973328e-05, + "loss": 0.3735, + "step": 30709 + }, + { + "epoch": 1.18, + "learning_rate": 3.0341113105924596e-05, + "loss": 0.4332, + "step": 30710 + }, + { + "epoch": 1.18, + "learning_rate": 3.0340471915875868e-05, + "loss": 0.4208, + "step": 30711 + }, + { + "epoch": 1.18, + "learning_rate": 3.0339830725827133e-05, + "loss": 0.4529, + "step": 30712 + }, + { + "epoch": 1.18, + "learning_rate": 3.0339189535778408e-05, + "loss": 0.3828, + "step": 30713 + }, + { + "epoch": 1.18, + "learning_rate": 3.033854834572968e-05, + "loss": 0.347, + "step": 30714 + }, + { + "epoch": 1.18, + "learning_rate": 3.0337907155680945e-05, + "loss": 0.4962, + "step": 30715 + }, + { + "epoch": 1.18, + "learning_rate": 3.0337265965632216e-05, + "loss": 0.471, + "step": 30716 + }, + { + "epoch": 1.18, + "learning_rate": 3.0336624775583485e-05, + "loss": 0.5008, + "step": 30717 + }, + { + "epoch": 1.18, + "learning_rate": 3.0335983585534756e-05, + "loss": 0.5604, + "step": 30718 + }, + { + "epoch": 1.18, + "learning_rate": 3.033534239548602e-05, + "loss": 0.4368, + "step": 30719 + }, + { + "epoch": 1.18, + "learning_rate": 3.0334701205437293e-05, + "loss": 0.3639, + "step": 30720 + }, + { + "epoch": 1.18, + "learning_rate": 3.033406001538856e-05, + "loss": 0.4139, + "step": 30721 + }, + { + "epoch": 1.18, + "learning_rate": 3.0333418825339833e-05, + "loss": 0.5772, + "step": 30722 + }, + { + "epoch": 1.18, + "learning_rate": 3.0332777635291105e-05, + "loss": 0.5025, + "step": 30723 + }, + { + "epoch": 1.18, + "learning_rate": 3.033213644524237e-05, + "loss": 0.3613, + "step": 30724 + }, + { + "epoch": 1.18, + "learning_rate": 3.033149525519364e-05, + "loss": 0.4146, + "step": 30725 + }, + { + "epoch": 1.18, + "learning_rate": 3.033085406514491e-05, + "loss": 0.4303, + "step": 30726 + }, + { + "epoch": 1.18, + "learning_rate": 3.033021287509618e-05, + "loss": 0.2827, + "step": 30727 + }, + { + "epoch": 1.18, + "learning_rate": 3.0329571685047446e-05, + "loss": 0.3625, + "step": 30728 + }, + { + "epoch": 1.18, + "learning_rate": 3.0328930494998718e-05, + "loss": 0.4577, + "step": 30729 + }, + { + "epoch": 1.18, + "learning_rate": 3.0328289304949986e-05, + "loss": 0.4711, + "step": 30730 + }, + { + "epoch": 1.18, + "learning_rate": 3.0327648114901258e-05, + "loss": 0.4139, + "step": 30731 + }, + { + "epoch": 1.18, + "learning_rate": 3.032700692485253e-05, + "loss": 0.3428, + "step": 30732 + }, + { + "epoch": 1.18, + "learning_rate": 3.0326365734803798e-05, + "loss": 0.4207, + "step": 30733 + }, + { + "epoch": 1.18, + "learning_rate": 3.032572454475507e-05, + "loss": 0.4536, + "step": 30734 + }, + { + "epoch": 1.18, + "learning_rate": 3.0325083354706335e-05, + "loss": 0.4187, + "step": 30735 + }, + { + "epoch": 1.18, + "learning_rate": 3.0324442164657607e-05, + "loss": 0.4376, + "step": 30736 + }, + { + "epoch": 1.18, + "learning_rate": 3.0323800974608875e-05, + "loss": 0.3559, + "step": 30737 + }, + { + "epoch": 1.18, + "learning_rate": 3.0323159784560147e-05, + "loss": 0.4874, + "step": 30738 + }, + { + "epoch": 1.18, + "learning_rate": 3.032251859451141e-05, + "loss": 0.5242, + "step": 30739 + }, + { + "epoch": 1.18, + "learning_rate": 3.0321877404462683e-05, + "loss": 0.4403, + "step": 30740 + }, + { + "epoch": 1.18, + "learning_rate": 3.0321236214413955e-05, + "loss": 0.3504, + "step": 30741 + }, + { + "epoch": 1.18, + "learning_rate": 3.0320595024365223e-05, + "loss": 0.4246, + "step": 30742 + }, + { + "epoch": 1.18, + "learning_rate": 3.0319953834316495e-05, + "loss": 0.4043, + "step": 30743 + }, + { + "epoch": 1.18, + "learning_rate": 3.031931264426776e-05, + "loss": 0.3659, + "step": 30744 + }, + { + "epoch": 1.18, + "learning_rate": 3.031867145421903e-05, + "loss": 0.3954, + "step": 30745 + }, + { + "epoch": 1.18, + "learning_rate": 3.03180302641703e-05, + "loss": 0.2967, + "step": 30746 + }, + { + "epoch": 1.18, + "learning_rate": 3.031738907412157e-05, + "loss": 0.4095, + "step": 30747 + }, + { + "epoch": 1.18, + "learning_rate": 3.0316747884072837e-05, + "loss": 0.2713, + "step": 30748 + }, + { + "epoch": 1.18, + "learning_rate": 3.0316106694024112e-05, + "loss": 0.3625, + "step": 30749 + }, + { + "epoch": 1.18, + "learning_rate": 3.0315465503975383e-05, + "loss": 0.3933, + "step": 30750 + }, + { + "epoch": 1.18, + "learning_rate": 3.031482431392665e-05, + "loss": 0.4893, + "step": 30751 + }, + { + "epoch": 1.18, + "learning_rate": 3.031418312387792e-05, + "loss": 0.2683, + "step": 30752 + }, + { + "epoch": 1.18, + "learning_rate": 3.031354193382919e-05, + "loss": 0.4175, + "step": 30753 + }, + { + "epoch": 1.18, + "learning_rate": 3.031290074378046e-05, + "loss": 0.3888, + "step": 30754 + }, + { + "epoch": 1.18, + "learning_rate": 3.0312259553731725e-05, + "loss": 0.2426, + "step": 30755 + }, + { + "epoch": 1.18, + "learning_rate": 3.0311618363682997e-05, + "loss": 0.3026, + "step": 30756 + }, + { + "epoch": 1.18, + "learning_rate": 3.0310977173634265e-05, + "loss": 0.3835, + "step": 30757 + }, + { + "epoch": 1.18, + "learning_rate": 3.0310335983585537e-05, + "loss": 0.5052, + "step": 30758 + }, + { + "epoch": 1.18, + "learning_rate": 3.030969479353681e-05, + "loss": 0.4676, + "step": 30759 + }, + { + "epoch": 1.18, + "learning_rate": 3.0309053603488074e-05, + "loss": 0.4904, + "step": 30760 + }, + { + "epoch": 1.18, + "learning_rate": 3.0308412413439345e-05, + "loss": 0.4295, + "step": 30761 + }, + { + "epoch": 1.18, + "learning_rate": 3.0307771223390614e-05, + "loss": 0.2995, + "step": 30762 + }, + { + "epoch": 1.18, + "learning_rate": 3.0307130033341885e-05, + "loss": 0.5183, + "step": 30763 + }, + { + "epoch": 1.18, + "learning_rate": 3.030648884329315e-05, + "loss": 0.3564, + "step": 30764 + }, + { + "epoch": 1.18, + "learning_rate": 3.0305847653244422e-05, + "loss": 0.4104, + "step": 30765 + }, + { + "epoch": 1.18, + "learning_rate": 3.030520646319569e-05, + "loss": 0.3722, + "step": 30766 + }, + { + "epoch": 1.18, + "learning_rate": 3.0304565273146962e-05, + "loss": 0.4007, + "step": 30767 + }, + { + "epoch": 1.18, + "learning_rate": 3.0303924083098234e-05, + "loss": 0.4144, + "step": 30768 + }, + { + "epoch": 1.18, + "learning_rate": 3.0303282893049502e-05, + "loss": 0.3079, + "step": 30769 + }, + { + "epoch": 1.18, + "learning_rate": 3.0302641703000774e-05, + "loss": 0.247, + "step": 30770 + }, + { + "epoch": 1.18, + "learning_rate": 3.030200051295204e-05, + "loss": 0.3503, + "step": 30771 + }, + { + "epoch": 1.18, + "learning_rate": 3.030135932290331e-05, + "loss": 0.4075, + "step": 30772 + }, + { + "epoch": 1.18, + "learning_rate": 3.030071813285458e-05, + "loss": 0.3662, + "step": 30773 + }, + { + "epoch": 1.18, + "learning_rate": 3.030007694280585e-05, + "loss": 0.3072, + "step": 30774 + }, + { + "epoch": 1.18, + "learning_rate": 3.0299435752757115e-05, + "loss": 0.3772, + "step": 30775 + }, + { + "epoch": 1.18, + "learning_rate": 3.0298794562708387e-05, + "loss": 0.3826, + "step": 30776 + }, + { + "epoch": 1.18, + "learning_rate": 3.029815337265966e-05, + "loss": 0.4276, + "step": 30777 + }, + { + "epoch": 1.18, + "learning_rate": 3.0297512182610927e-05, + "loss": 0.4609, + "step": 30778 + }, + { + "epoch": 1.18, + "learning_rate": 3.02968709925622e-05, + "loss": 0.4076, + "step": 30779 + }, + { + "epoch": 1.18, + "learning_rate": 3.0296229802513464e-05, + "loss": 0.4388, + "step": 30780 + }, + { + "epoch": 1.18, + "learning_rate": 3.0295588612464736e-05, + "loss": 0.4544, + "step": 30781 + }, + { + "epoch": 1.18, + "learning_rate": 3.0294947422416004e-05, + "loss": 0.4529, + "step": 30782 + }, + { + "epoch": 1.18, + "learning_rate": 3.0294306232367276e-05, + "loss": 0.3381, + "step": 30783 + }, + { + "epoch": 1.18, + "learning_rate": 3.029366504231854e-05, + "loss": 0.3916, + "step": 30784 + }, + { + "epoch": 1.18, + "learning_rate": 3.0293023852269816e-05, + "loss": 0.3497, + "step": 30785 + }, + { + "epoch": 1.18, + "learning_rate": 3.0292382662221087e-05, + "loss": 0.3463, + "step": 30786 + }, + { + "epoch": 1.18, + "learning_rate": 3.0291741472172352e-05, + "loss": 0.3833, + "step": 30787 + }, + { + "epoch": 1.18, + "learning_rate": 3.0291100282123624e-05, + "loss": 0.3561, + "step": 30788 + }, + { + "epoch": 1.18, + "learning_rate": 3.0290459092074892e-05, + "loss": 0.4346, + "step": 30789 + }, + { + "epoch": 1.18, + "learning_rate": 3.0289817902026164e-05, + "loss": 0.4289, + "step": 30790 + }, + { + "epoch": 1.18, + "learning_rate": 3.028917671197743e-05, + "loss": 0.3685, + "step": 30791 + }, + { + "epoch": 1.18, + "learning_rate": 3.02885355219287e-05, + "loss": 0.4033, + "step": 30792 + }, + { + "epoch": 1.18, + "learning_rate": 3.028789433187997e-05, + "loss": 0.3439, + "step": 30793 + }, + { + "epoch": 1.18, + "learning_rate": 3.028725314183124e-05, + "loss": 0.3552, + "step": 30794 + }, + { + "epoch": 1.18, + "learning_rate": 3.0286611951782512e-05, + "loss": 0.3783, + "step": 30795 + }, + { + "epoch": 1.18, + "learning_rate": 3.0285970761733777e-05, + "loss": 0.3729, + "step": 30796 + }, + { + "epoch": 1.18, + "learning_rate": 3.028532957168505e-05, + "loss": 0.436, + "step": 30797 + }, + { + "epoch": 1.18, + "learning_rate": 3.0284688381636317e-05, + "loss": 0.3506, + "step": 30798 + }, + { + "epoch": 1.18, + "learning_rate": 3.028404719158759e-05, + "loss": 0.4029, + "step": 30799 + }, + { + "epoch": 1.18, + "learning_rate": 3.0283406001538854e-05, + "loss": 0.4377, + "step": 30800 + }, + { + "epoch": 1.18, + "learning_rate": 3.0282764811490126e-05, + "loss": 0.4228, + "step": 30801 + }, + { + "epoch": 1.18, + "learning_rate": 3.0282123621441394e-05, + "loss": 0.3604, + "step": 30802 + }, + { + "epoch": 1.18, + "learning_rate": 3.0281482431392666e-05, + "loss": 0.226, + "step": 30803 + }, + { + "epoch": 1.18, + "learning_rate": 3.0280841241343938e-05, + "loss": 0.454, + "step": 30804 + }, + { + "epoch": 1.18, + "learning_rate": 3.0280200051295206e-05, + "loss": 0.445, + "step": 30805 + }, + { + "epoch": 1.18, + "learning_rate": 3.0279558861246478e-05, + "loss": 0.3195, + "step": 30806 + }, + { + "epoch": 1.18, + "learning_rate": 3.0278917671197743e-05, + "loss": 0.2982, + "step": 30807 + }, + { + "epoch": 1.18, + "learning_rate": 3.0278276481149014e-05, + "loss": 0.5024, + "step": 30808 + }, + { + "epoch": 1.18, + "learning_rate": 3.0277635291100283e-05, + "loss": 0.3578, + "step": 30809 + }, + { + "epoch": 1.18, + "learning_rate": 3.0276994101051554e-05, + "loss": 0.4006, + "step": 30810 + }, + { + "epoch": 1.18, + "learning_rate": 3.0276352911002826e-05, + "loss": 0.3458, + "step": 30811 + }, + { + "epoch": 1.18, + "learning_rate": 3.027571172095409e-05, + "loss": 0.4024, + "step": 30812 + }, + { + "epoch": 1.18, + "learning_rate": 3.0275070530905363e-05, + "loss": 0.3424, + "step": 30813 + }, + { + "epoch": 1.18, + "learning_rate": 3.027442934085663e-05, + "loss": 0.3961, + "step": 30814 + }, + { + "epoch": 1.18, + "learning_rate": 3.0273788150807903e-05, + "loss": 0.3186, + "step": 30815 + }, + { + "epoch": 1.18, + "learning_rate": 3.0273146960759168e-05, + "loss": 0.3898, + "step": 30816 + }, + { + "epoch": 1.18, + "learning_rate": 3.027250577071044e-05, + "loss": 0.4501, + "step": 30817 + }, + { + "epoch": 1.18, + "learning_rate": 3.0271864580661708e-05, + "loss": 0.3343, + "step": 30818 + }, + { + "epoch": 1.18, + "learning_rate": 3.027122339061298e-05, + "loss": 0.36, + "step": 30819 + }, + { + "epoch": 1.18, + "learning_rate": 3.027058220056425e-05, + "loss": 0.3949, + "step": 30820 + }, + { + "epoch": 1.18, + "learning_rate": 3.026994101051552e-05, + "loss": 0.35, + "step": 30821 + }, + { + "epoch": 1.19, + "learning_rate": 3.026929982046679e-05, + "loss": 0.453, + "step": 30822 + }, + { + "epoch": 1.19, + "learning_rate": 3.0268658630418056e-05, + "loss": 0.4588, + "step": 30823 + }, + { + "epoch": 1.19, + "learning_rate": 3.0268017440369328e-05, + "loss": 0.4098, + "step": 30824 + }, + { + "epoch": 1.19, + "learning_rate": 3.0267376250320596e-05, + "loss": 0.3157, + "step": 30825 + }, + { + "epoch": 1.19, + "learning_rate": 3.0266735060271868e-05, + "loss": 0.4444, + "step": 30826 + }, + { + "epoch": 1.19, + "learning_rate": 3.0266093870223133e-05, + "loss": 0.6165, + "step": 30827 + }, + { + "epoch": 1.19, + "learning_rate": 3.0265452680174405e-05, + "loss": 0.3431, + "step": 30828 + }, + { + "epoch": 1.19, + "learning_rate": 3.0264811490125676e-05, + "loss": 0.418, + "step": 30829 + }, + { + "epoch": 1.19, + "learning_rate": 3.0264170300076945e-05, + "loss": 0.3801, + "step": 30830 + }, + { + "epoch": 1.19, + "learning_rate": 3.0263529110028216e-05, + "loss": 0.5004, + "step": 30831 + }, + { + "epoch": 1.19, + "learning_rate": 3.026288791997948e-05, + "loss": 0.3076, + "step": 30832 + }, + { + "epoch": 1.19, + "learning_rate": 3.0262246729930753e-05, + "loss": 0.4368, + "step": 30833 + }, + { + "epoch": 1.19, + "learning_rate": 3.026160553988202e-05, + "loss": 0.4112, + "step": 30834 + }, + { + "epoch": 1.19, + "learning_rate": 3.0260964349833293e-05, + "loss": 0.4197, + "step": 30835 + }, + { + "epoch": 1.19, + "learning_rate": 3.0260323159784558e-05, + "loss": 0.4313, + "step": 30836 + }, + { + "epoch": 1.19, + "learning_rate": 3.025968196973583e-05, + "loss": 0.354, + "step": 30837 + }, + { + "epoch": 1.19, + "learning_rate": 3.0259040779687105e-05, + "loss": 0.3939, + "step": 30838 + }, + { + "epoch": 1.19, + "learning_rate": 3.025839958963837e-05, + "loss": 0.423, + "step": 30839 + }, + { + "epoch": 1.19, + "learning_rate": 3.025775839958964e-05, + "loss": 0.3054, + "step": 30840 + }, + { + "epoch": 1.19, + "learning_rate": 3.025711720954091e-05, + "loss": 0.3817, + "step": 30841 + }, + { + "epoch": 1.19, + "learning_rate": 3.025647601949218e-05, + "loss": 0.3108, + "step": 30842 + }, + { + "epoch": 1.19, + "learning_rate": 3.0255834829443446e-05, + "loss": 0.3086, + "step": 30843 + }, + { + "epoch": 1.19, + "learning_rate": 3.0255193639394718e-05, + "loss": 0.3841, + "step": 30844 + }, + { + "epoch": 1.19, + "learning_rate": 3.0254552449345986e-05, + "loss": 0.4313, + "step": 30845 + }, + { + "epoch": 1.19, + "learning_rate": 3.0253911259297258e-05, + "loss": 0.4608, + "step": 30846 + }, + { + "epoch": 1.19, + "learning_rate": 3.025327006924853e-05, + "loss": 0.3449, + "step": 30847 + }, + { + "epoch": 1.19, + "learning_rate": 3.0252628879199795e-05, + "loss": 0.4079, + "step": 30848 + }, + { + "epoch": 1.19, + "learning_rate": 3.0251987689151067e-05, + "loss": 0.4221, + "step": 30849 + }, + { + "epoch": 1.19, + "learning_rate": 3.0251346499102335e-05, + "loss": 0.3667, + "step": 30850 + }, + { + "epoch": 1.19, + "learning_rate": 3.0250705309053607e-05, + "loss": 0.4464, + "step": 30851 + }, + { + "epoch": 1.19, + "learning_rate": 3.025006411900487e-05, + "loss": 0.3353, + "step": 30852 + }, + { + "epoch": 1.19, + "learning_rate": 3.0249422928956143e-05, + "loss": 0.3165, + "step": 30853 + }, + { + "epoch": 1.19, + "learning_rate": 3.024878173890741e-05, + "loss": 0.5983, + "step": 30854 + }, + { + "epoch": 1.19, + "learning_rate": 3.0248140548858683e-05, + "loss": 0.4873, + "step": 30855 + }, + { + "epoch": 1.19, + "learning_rate": 3.0247499358809955e-05, + "loss": 0.4951, + "step": 30856 + }, + { + "epoch": 1.19, + "learning_rate": 3.0246858168761223e-05, + "loss": 0.3316, + "step": 30857 + }, + { + "epoch": 1.19, + "learning_rate": 3.0246216978712495e-05, + "loss": 0.2845, + "step": 30858 + }, + { + "epoch": 1.19, + "learning_rate": 3.024557578866376e-05, + "loss": 0.3365, + "step": 30859 + }, + { + "epoch": 1.19, + "learning_rate": 3.0244934598615032e-05, + "loss": 0.37, + "step": 30860 + }, + { + "epoch": 1.19, + "learning_rate": 3.02442934085663e-05, + "loss": 0.3768, + "step": 30861 + }, + { + "epoch": 1.19, + "learning_rate": 3.0243652218517572e-05, + "loss": 0.5263, + "step": 30862 + }, + { + "epoch": 1.19, + "learning_rate": 3.0243011028468837e-05, + "loss": 0.3728, + "step": 30863 + }, + { + "epoch": 1.19, + "learning_rate": 3.024236983842011e-05, + "loss": 0.4773, + "step": 30864 + }, + { + "epoch": 1.19, + "learning_rate": 3.024172864837138e-05, + "loss": 0.3706, + "step": 30865 + }, + { + "epoch": 1.19, + "learning_rate": 3.024108745832265e-05, + "loss": 0.3295, + "step": 30866 + }, + { + "epoch": 1.19, + "learning_rate": 3.024044626827392e-05, + "loss": 0.4759, + "step": 30867 + }, + { + "epoch": 1.19, + "learning_rate": 3.0239805078225185e-05, + "loss": 0.3965, + "step": 30868 + }, + { + "epoch": 1.19, + "learning_rate": 3.0239163888176457e-05, + "loss": 0.3319, + "step": 30869 + }, + { + "epoch": 1.19, + "learning_rate": 3.0238522698127725e-05, + "loss": 0.4646, + "step": 30870 + }, + { + "epoch": 1.19, + "learning_rate": 3.0237881508078997e-05, + "loss": 0.3169, + "step": 30871 + }, + { + "epoch": 1.19, + "learning_rate": 3.0237240318030262e-05, + "loss": 0.5621, + "step": 30872 + }, + { + "epoch": 1.19, + "learning_rate": 3.0236599127981537e-05, + "loss": 0.3807, + "step": 30873 + }, + { + "epoch": 1.19, + "learning_rate": 3.023595793793281e-05, + "loss": 0.3519, + "step": 30874 + }, + { + "epoch": 1.19, + "learning_rate": 3.0235316747884074e-05, + "loss": 0.3548, + "step": 30875 + }, + { + "epoch": 1.19, + "learning_rate": 3.0234675557835345e-05, + "loss": 0.476, + "step": 30876 + }, + { + "epoch": 1.19, + "learning_rate": 3.0234034367786614e-05, + "loss": 0.2644, + "step": 30877 + }, + { + "epoch": 1.19, + "learning_rate": 3.0233393177737885e-05, + "loss": 0.3398, + "step": 30878 + }, + { + "epoch": 1.19, + "learning_rate": 3.023275198768915e-05, + "loss": 0.29, + "step": 30879 + }, + { + "epoch": 1.19, + "learning_rate": 3.0232110797640422e-05, + "loss": 0.3548, + "step": 30880 + }, + { + "epoch": 1.19, + "learning_rate": 3.023146960759169e-05, + "loss": 0.3984, + "step": 30881 + }, + { + "epoch": 1.19, + "learning_rate": 3.0230828417542962e-05, + "loss": 0.4823, + "step": 30882 + }, + { + "epoch": 1.19, + "learning_rate": 3.0230187227494234e-05, + "loss": 0.3521, + "step": 30883 + }, + { + "epoch": 1.19, + "learning_rate": 3.02295460374455e-05, + "loss": 0.4748, + "step": 30884 + }, + { + "epoch": 1.19, + "learning_rate": 3.022890484739677e-05, + "loss": 0.3571, + "step": 30885 + }, + { + "epoch": 1.19, + "learning_rate": 3.022826365734804e-05, + "loss": 0.3314, + "step": 30886 + }, + { + "epoch": 1.19, + "learning_rate": 3.022762246729931e-05, + "loss": 0.4625, + "step": 30887 + }, + { + "epoch": 1.19, + "learning_rate": 3.0226981277250575e-05, + "loss": 0.4584, + "step": 30888 + }, + { + "epoch": 1.19, + "learning_rate": 3.0226340087201847e-05, + "loss": 0.3917, + "step": 30889 + }, + { + "epoch": 1.19, + "learning_rate": 3.0225698897153115e-05, + "loss": 0.4239, + "step": 30890 + }, + { + "epoch": 1.19, + "learning_rate": 3.0225057707104387e-05, + "loss": 0.4252, + "step": 30891 + }, + { + "epoch": 1.19, + "learning_rate": 3.022441651705566e-05, + "loss": 0.4079, + "step": 30892 + }, + { + "epoch": 1.19, + "learning_rate": 3.0223775327006927e-05, + "loss": 0.3138, + "step": 30893 + }, + { + "epoch": 1.19, + "learning_rate": 3.02231341369582e-05, + "loss": 0.4399, + "step": 30894 + }, + { + "epoch": 1.19, + "learning_rate": 3.0222492946909464e-05, + "loss": 0.4249, + "step": 30895 + }, + { + "epoch": 1.19, + "learning_rate": 3.0221851756860736e-05, + "loss": 0.4914, + "step": 30896 + }, + { + "epoch": 1.19, + "learning_rate": 3.0221210566812004e-05, + "loss": 0.2433, + "step": 30897 + }, + { + "epoch": 1.19, + "learning_rate": 3.0220569376763276e-05, + "loss": 0.3111, + "step": 30898 + }, + { + "epoch": 1.19, + "learning_rate": 3.021992818671454e-05, + "loss": 0.4057, + "step": 30899 + }, + { + "epoch": 1.19, + "learning_rate": 3.0219286996665812e-05, + "loss": 0.4314, + "step": 30900 + }, + { + "epoch": 1.19, + "learning_rate": 3.0218645806617084e-05, + "loss": 0.3778, + "step": 30901 + }, + { + "epoch": 1.19, + "learning_rate": 3.0218004616568352e-05, + "loss": 0.3843, + "step": 30902 + }, + { + "epoch": 1.19, + "learning_rate": 3.0217363426519624e-05, + "loss": 0.4887, + "step": 30903 + }, + { + "epoch": 1.19, + "learning_rate": 3.021672223647089e-05, + "loss": 0.3096, + "step": 30904 + }, + { + "epoch": 1.19, + "learning_rate": 3.021608104642216e-05, + "loss": 0.331, + "step": 30905 + }, + { + "epoch": 1.19, + "learning_rate": 3.021543985637343e-05, + "loss": 0.4637, + "step": 30906 + }, + { + "epoch": 1.19, + "learning_rate": 3.02147986663247e-05, + "loss": 0.3646, + "step": 30907 + }, + { + "epoch": 1.19, + "learning_rate": 3.0214157476275966e-05, + "loss": 0.3941, + "step": 30908 + }, + { + "epoch": 1.19, + "learning_rate": 3.021351628622724e-05, + "loss": 0.3174, + "step": 30909 + }, + { + "epoch": 1.19, + "learning_rate": 3.0212875096178513e-05, + "loss": 0.394, + "step": 30910 + }, + { + "epoch": 1.19, + "learning_rate": 3.0212233906129777e-05, + "loss": 0.3802, + "step": 30911 + }, + { + "epoch": 1.19, + "learning_rate": 3.021159271608105e-05, + "loss": 0.4225, + "step": 30912 + }, + { + "epoch": 1.19, + "learning_rate": 3.0210951526032317e-05, + "loss": 0.3699, + "step": 30913 + }, + { + "epoch": 1.19, + "learning_rate": 3.021031033598359e-05, + "loss": 0.3442, + "step": 30914 + }, + { + "epoch": 1.19, + "learning_rate": 3.0209669145934854e-05, + "loss": 0.3018, + "step": 30915 + }, + { + "epoch": 1.19, + "learning_rate": 3.0209027955886126e-05, + "loss": 0.349, + "step": 30916 + }, + { + "epoch": 1.19, + "learning_rate": 3.0208386765837394e-05, + "loss": 0.2829, + "step": 30917 + }, + { + "epoch": 1.19, + "learning_rate": 3.0207745575788666e-05, + "loss": 0.3382, + "step": 30918 + }, + { + "epoch": 1.19, + "learning_rate": 3.0207104385739938e-05, + "loss": 0.3766, + "step": 30919 + }, + { + "epoch": 1.19, + "learning_rate": 3.0206463195691203e-05, + "loss": 0.2961, + "step": 30920 + }, + { + "epoch": 1.19, + "learning_rate": 3.0205822005642474e-05, + "loss": 0.3997, + "step": 30921 + }, + { + "epoch": 1.19, + "learning_rate": 3.0205180815593743e-05, + "loss": 0.3501, + "step": 30922 + }, + { + "epoch": 1.19, + "learning_rate": 3.0204539625545014e-05, + "loss": 0.3985, + "step": 30923 + }, + { + "epoch": 1.19, + "learning_rate": 3.020389843549628e-05, + "loss": 0.2566, + "step": 30924 + }, + { + "epoch": 1.19, + "learning_rate": 3.020325724544755e-05, + "loss": 0.3702, + "step": 30925 + }, + { + "epoch": 1.19, + "learning_rate": 3.020261605539882e-05, + "loss": 0.3154, + "step": 30926 + }, + { + "epoch": 1.19, + "learning_rate": 3.020197486535009e-05, + "loss": 0.4771, + "step": 30927 + }, + { + "epoch": 1.19, + "learning_rate": 3.0201333675301363e-05, + "loss": 0.4477, + "step": 30928 + }, + { + "epoch": 1.19, + "learning_rate": 3.020069248525263e-05, + "loss": 0.6011, + "step": 30929 + }, + { + "epoch": 1.19, + "learning_rate": 3.0200051295203903e-05, + "loss": 0.3755, + "step": 30930 + }, + { + "epoch": 1.19, + "learning_rate": 3.0199410105155168e-05, + "loss": 0.3197, + "step": 30931 + }, + { + "epoch": 1.19, + "learning_rate": 3.019876891510644e-05, + "loss": 0.3742, + "step": 30932 + }, + { + "epoch": 1.19, + "learning_rate": 3.0198127725057708e-05, + "loss": 0.384, + "step": 30933 + }, + { + "epoch": 1.19, + "learning_rate": 3.019748653500898e-05, + "loss": 0.3336, + "step": 30934 + }, + { + "epoch": 1.19, + "learning_rate": 3.0196845344960244e-05, + "loss": 0.3227, + "step": 30935 + }, + { + "epoch": 1.19, + "learning_rate": 3.0196204154911516e-05, + "loss": 0.6145, + "step": 30936 + }, + { + "epoch": 1.19, + "learning_rate": 3.0195562964862788e-05, + "loss": 0.4773, + "step": 30937 + }, + { + "epoch": 1.19, + "learning_rate": 3.0194921774814056e-05, + "loss": 0.442, + "step": 30938 + }, + { + "epoch": 1.19, + "learning_rate": 3.0194280584765328e-05, + "loss": 0.4232, + "step": 30939 + }, + { + "epoch": 1.19, + "learning_rate": 3.0193639394716593e-05, + "loss": 0.3847, + "step": 30940 + }, + { + "epoch": 1.19, + "learning_rate": 3.0192998204667865e-05, + "loss": 0.3954, + "step": 30941 + }, + { + "epoch": 1.19, + "learning_rate": 3.0192357014619133e-05, + "loss": 0.3631, + "step": 30942 + }, + { + "epoch": 1.19, + "learning_rate": 3.0191715824570405e-05, + "loss": 0.3945, + "step": 30943 + }, + { + "epoch": 1.19, + "learning_rate": 3.019107463452167e-05, + "loss": 0.4375, + "step": 30944 + }, + { + "epoch": 1.19, + "learning_rate": 3.0190433444472945e-05, + "loss": 0.3517, + "step": 30945 + }, + { + "epoch": 1.19, + "learning_rate": 3.0189792254424216e-05, + "loss": 0.3527, + "step": 30946 + }, + { + "epoch": 1.19, + "learning_rate": 3.018915106437548e-05, + "loss": 0.4703, + "step": 30947 + }, + { + "epoch": 1.19, + "learning_rate": 3.0188509874326753e-05, + "loss": 0.3621, + "step": 30948 + }, + { + "epoch": 1.19, + "learning_rate": 3.018786868427802e-05, + "loss": 0.3722, + "step": 30949 + }, + { + "epoch": 1.19, + "learning_rate": 3.0187227494229293e-05, + "loss": 0.4117, + "step": 30950 + }, + { + "epoch": 1.19, + "learning_rate": 3.0186586304180558e-05, + "loss": 0.4821, + "step": 30951 + }, + { + "epoch": 1.19, + "learning_rate": 3.018594511413183e-05, + "loss": 0.4396, + "step": 30952 + }, + { + "epoch": 1.19, + "learning_rate": 3.0185303924083098e-05, + "loss": 0.4034, + "step": 30953 + }, + { + "epoch": 1.19, + "learning_rate": 3.018466273403437e-05, + "loss": 0.464, + "step": 30954 + }, + { + "epoch": 1.19, + "learning_rate": 3.018402154398564e-05, + "loss": 0.3495, + "step": 30955 + }, + { + "epoch": 1.19, + "learning_rate": 3.0183380353936906e-05, + "loss": 0.3785, + "step": 30956 + }, + { + "epoch": 1.19, + "learning_rate": 3.0182739163888178e-05, + "loss": 0.413, + "step": 30957 + }, + { + "epoch": 1.19, + "learning_rate": 3.0182097973839446e-05, + "loss": 0.325, + "step": 30958 + }, + { + "epoch": 1.19, + "learning_rate": 3.0181456783790718e-05, + "loss": 0.3846, + "step": 30959 + }, + { + "epoch": 1.19, + "learning_rate": 3.0180815593741983e-05, + "loss": 0.4867, + "step": 30960 + }, + { + "epoch": 1.19, + "learning_rate": 3.0180174403693255e-05, + "loss": 0.4479, + "step": 30961 + }, + { + "epoch": 1.19, + "learning_rate": 3.0179533213644523e-05, + "loss": 0.4417, + "step": 30962 + }, + { + "epoch": 1.19, + "learning_rate": 3.0178892023595795e-05, + "loss": 0.3508, + "step": 30963 + }, + { + "epoch": 1.19, + "learning_rate": 3.0178250833547067e-05, + "loss": 0.4081, + "step": 30964 + }, + { + "epoch": 1.19, + "learning_rate": 3.0177609643498335e-05, + "loss": 0.4082, + "step": 30965 + }, + { + "epoch": 1.19, + "learning_rate": 3.0176968453449607e-05, + "loss": 0.3522, + "step": 30966 + }, + { + "epoch": 1.19, + "learning_rate": 3.017632726340087e-05, + "loss": 0.4669, + "step": 30967 + }, + { + "epoch": 1.19, + "learning_rate": 3.0175686073352143e-05, + "loss": 0.4244, + "step": 30968 + }, + { + "epoch": 1.19, + "learning_rate": 3.017504488330341e-05, + "loss": 0.4516, + "step": 30969 + }, + { + "epoch": 1.19, + "learning_rate": 3.0174403693254683e-05, + "loss": 0.4225, + "step": 30970 + }, + { + "epoch": 1.19, + "learning_rate": 3.017376250320595e-05, + "loss": 0.3724, + "step": 30971 + }, + { + "epoch": 1.19, + "learning_rate": 3.017312131315722e-05, + "loss": 0.3628, + "step": 30972 + }, + { + "epoch": 1.19, + "learning_rate": 3.0172480123108492e-05, + "loss": 0.4331, + "step": 30973 + }, + { + "epoch": 1.19, + "learning_rate": 3.017183893305976e-05, + "loss": 0.3242, + "step": 30974 + }, + { + "epoch": 1.19, + "learning_rate": 3.0171197743011032e-05, + "loss": 0.4065, + "step": 30975 + }, + { + "epoch": 1.19, + "learning_rate": 3.0170556552962297e-05, + "loss": 0.3307, + "step": 30976 + }, + { + "epoch": 1.19, + "learning_rate": 3.016991536291357e-05, + "loss": 0.3216, + "step": 30977 + }, + { + "epoch": 1.19, + "learning_rate": 3.0169274172864837e-05, + "loss": 0.3776, + "step": 30978 + }, + { + "epoch": 1.19, + "learning_rate": 3.016863298281611e-05, + "loss": 0.3694, + "step": 30979 + }, + { + "epoch": 1.19, + "learning_rate": 3.0167991792767373e-05, + "loss": 0.4068, + "step": 30980 + }, + { + "epoch": 1.19, + "learning_rate": 3.016735060271865e-05, + "loss": 0.3814, + "step": 30981 + }, + { + "epoch": 1.19, + "learning_rate": 3.016670941266992e-05, + "loss": 0.3794, + "step": 30982 + }, + { + "epoch": 1.19, + "learning_rate": 3.0166068222621185e-05, + "loss": 0.3635, + "step": 30983 + }, + { + "epoch": 1.19, + "learning_rate": 3.0165427032572457e-05, + "loss": 0.4707, + "step": 30984 + }, + { + "epoch": 1.19, + "learning_rate": 3.0164785842523725e-05, + "loss": 0.4144, + "step": 30985 + }, + { + "epoch": 1.19, + "learning_rate": 3.0164144652474997e-05, + "loss": 0.3494, + "step": 30986 + }, + { + "epoch": 1.19, + "learning_rate": 3.0163503462426262e-05, + "loss": 0.3359, + "step": 30987 + }, + { + "epoch": 1.19, + "learning_rate": 3.0162862272377534e-05, + "loss": 0.286, + "step": 30988 + }, + { + "epoch": 1.19, + "learning_rate": 3.0162221082328802e-05, + "loss": 0.3841, + "step": 30989 + }, + { + "epoch": 1.19, + "learning_rate": 3.0161579892280074e-05, + "loss": 0.3257, + "step": 30990 + }, + { + "epoch": 1.19, + "learning_rate": 3.0160938702231345e-05, + "loss": 0.3849, + "step": 30991 + }, + { + "epoch": 1.19, + "learning_rate": 3.016029751218261e-05, + "loss": 0.4228, + "step": 30992 + }, + { + "epoch": 1.19, + "learning_rate": 3.0159656322133882e-05, + "loss": 0.3502, + "step": 30993 + }, + { + "epoch": 1.19, + "learning_rate": 3.015901513208515e-05, + "loss": 0.4233, + "step": 30994 + }, + { + "epoch": 1.19, + "learning_rate": 3.0158373942036422e-05, + "loss": 0.4853, + "step": 30995 + }, + { + "epoch": 1.19, + "learning_rate": 3.0157732751987687e-05, + "loss": 0.3979, + "step": 30996 + }, + { + "epoch": 1.19, + "learning_rate": 3.0157091561938962e-05, + "loss": 0.356, + "step": 30997 + }, + { + "epoch": 1.19, + "learning_rate": 3.0156450371890234e-05, + "loss": 0.5184, + "step": 30998 + }, + { + "epoch": 1.19, + "learning_rate": 3.01558091818415e-05, + "loss": 0.4065, + "step": 30999 + }, + { + "epoch": 1.19, + "learning_rate": 3.015516799179277e-05, + "loss": 0.2911, + "step": 31000 + }, + { + "epoch": 1.19, + "learning_rate": 3.015452680174404e-05, + "loss": 0.57, + "step": 31001 + }, + { + "epoch": 1.19, + "learning_rate": 3.015388561169531e-05, + "loss": 0.3829, + "step": 31002 + }, + { + "epoch": 1.19, + "learning_rate": 3.0153244421646575e-05, + "loss": 0.494, + "step": 31003 + }, + { + "epoch": 1.19, + "learning_rate": 3.0152603231597847e-05, + "loss": 0.4095, + "step": 31004 + }, + { + "epoch": 1.19, + "learning_rate": 3.0151962041549116e-05, + "loss": 0.5106, + "step": 31005 + }, + { + "epoch": 1.19, + "learning_rate": 3.0151320851500387e-05, + "loss": 0.3448, + "step": 31006 + }, + { + "epoch": 1.19, + "learning_rate": 3.015067966145166e-05, + "loss": 0.4217, + "step": 31007 + }, + { + "epoch": 1.19, + "learning_rate": 3.0150038471402924e-05, + "loss": 0.3074, + "step": 31008 + }, + { + "epoch": 1.19, + "learning_rate": 3.0149397281354196e-05, + "loss": 0.412, + "step": 31009 + }, + { + "epoch": 1.19, + "learning_rate": 3.0148756091305464e-05, + "loss": 0.4664, + "step": 31010 + }, + { + "epoch": 1.19, + "learning_rate": 3.0148114901256736e-05, + "loss": 0.5302, + "step": 31011 + }, + { + "epoch": 1.19, + "learning_rate": 3.0147473711208e-05, + "loss": 0.2717, + "step": 31012 + }, + { + "epoch": 1.19, + "learning_rate": 3.0146832521159272e-05, + "loss": 0.3774, + "step": 31013 + }, + { + "epoch": 1.19, + "learning_rate": 3.014619133111054e-05, + "loss": 0.3748, + "step": 31014 + }, + { + "epoch": 1.19, + "learning_rate": 3.0145550141061812e-05, + "loss": 0.3985, + "step": 31015 + }, + { + "epoch": 1.19, + "learning_rate": 3.0144908951013084e-05, + "loss": 0.4344, + "step": 31016 + }, + { + "epoch": 1.19, + "learning_rate": 3.0144267760964352e-05, + "loss": 0.3924, + "step": 31017 + }, + { + "epoch": 1.19, + "learning_rate": 3.0143626570915624e-05, + "loss": 0.375, + "step": 31018 + }, + { + "epoch": 1.19, + "learning_rate": 3.014298538086689e-05, + "loss": 0.37, + "step": 31019 + }, + { + "epoch": 1.19, + "learning_rate": 3.014234419081816e-05, + "loss": 0.4206, + "step": 31020 + }, + { + "epoch": 1.19, + "learning_rate": 3.014170300076943e-05, + "loss": 0.3407, + "step": 31021 + }, + { + "epoch": 1.19, + "learning_rate": 3.01410618107207e-05, + "loss": 0.3615, + "step": 31022 + }, + { + "epoch": 1.19, + "learning_rate": 3.0140420620671966e-05, + "loss": 0.4431, + "step": 31023 + }, + { + "epoch": 1.19, + "learning_rate": 3.0139779430623237e-05, + "loss": 0.3316, + "step": 31024 + }, + { + "epoch": 1.19, + "learning_rate": 3.013913824057451e-05, + "loss": 0.3621, + "step": 31025 + }, + { + "epoch": 1.19, + "learning_rate": 3.0138497050525778e-05, + "loss": 0.3215, + "step": 31026 + }, + { + "epoch": 1.19, + "learning_rate": 3.013785586047705e-05, + "loss": 0.3305, + "step": 31027 + }, + { + "epoch": 1.19, + "learning_rate": 3.0137214670428314e-05, + "loss": 0.2164, + "step": 31028 + }, + { + "epoch": 1.19, + "learning_rate": 3.0136573480379586e-05, + "loss": 0.5161, + "step": 31029 + }, + { + "epoch": 1.19, + "learning_rate": 3.0135932290330854e-05, + "loss": 0.3288, + "step": 31030 + }, + { + "epoch": 1.19, + "learning_rate": 3.0135291100282126e-05, + "loss": 0.3694, + "step": 31031 + }, + { + "epoch": 1.19, + "learning_rate": 3.013464991023339e-05, + "loss": 0.4663, + "step": 31032 + }, + { + "epoch": 1.19, + "learning_rate": 3.0134008720184666e-05, + "loss": 0.2935, + "step": 31033 + }, + { + "epoch": 1.19, + "learning_rate": 3.0133367530135938e-05, + "loss": 0.5559, + "step": 31034 + }, + { + "epoch": 1.19, + "learning_rate": 3.0132726340087203e-05, + "loss": 0.3926, + "step": 31035 + }, + { + "epoch": 1.19, + "learning_rate": 3.0132085150038474e-05, + "loss": 0.4374, + "step": 31036 + }, + { + "epoch": 1.19, + "learning_rate": 3.0131443959989743e-05, + "loss": 0.4801, + "step": 31037 + }, + { + "epoch": 1.19, + "learning_rate": 3.0130802769941014e-05, + "loss": 0.4565, + "step": 31038 + }, + { + "epoch": 1.19, + "learning_rate": 3.013016157989228e-05, + "loss": 0.3762, + "step": 31039 + }, + { + "epoch": 1.19, + "learning_rate": 3.012952038984355e-05, + "loss": 0.2355, + "step": 31040 + }, + { + "epoch": 1.19, + "learning_rate": 3.012887919979482e-05, + "loss": 0.297, + "step": 31041 + }, + { + "epoch": 1.19, + "learning_rate": 3.012823800974609e-05, + "loss": 0.3219, + "step": 31042 + }, + { + "epoch": 1.19, + "learning_rate": 3.0127596819697363e-05, + "loss": 0.356, + "step": 31043 + }, + { + "epoch": 1.19, + "learning_rate": 3.0126955629648628e-05, + "loss": 0.4553, + "step": 31044 + }, + { + "epoch": 1.19, + "learning_rate": 3.01263144395999e-05, + "loss": 0.407, + "step": 31045 + }, + { + "epoch": 1.19, + "learning_rate": 3.0125673249551168e-05, + "loss": 0.4155, + "step": 31046 + }, + { + "epoch": 1.19, + "learning_rate": 3.012503205950244e-05, + "loss": 0.4817, + "step": 31047 + }, + { + "epoch": 1.19, + "learning_rate": 3.0124390869453704e-05, + "loss": 0.3576, + "step": 31048 + }, + { + "epoch": 1.19, + "learning_rate": 3.0123749679404976e-05, + "loss": 0.3617, + "step": 31049 + }, + { + "epoch": 1.19, + "learning_rate": 3.0123108489356245e-05, + "loss": 0.5001, + "step": 31050 + }, + { + "epoch": 1.19, + "learning_rate": 3.0122467299307516e-05, + "loss": 0.4948, + "step": 31051 + }, + { + "epoch": 1.19, + "learning_rate": 3.0121826109258788e-05, + "loss": 0.3468, + "step": 31052 + }, + { + "epoch": 1.19, + "learning_rate": 3.0121184919210056e-05, + "loss": 0.3733, + "step": 31053 + }, + { + "epoch": 1.19, + "learning_rate": 3.0120543729161328e-05, + "loss": 0.3797, + "step": 31054 + }, + { + "epoch": 1.19, + "learning_rate": 3.0119902539112593e-05, + "loss": 0.5051, + "step": 31055 + }, + { + "epoch": 1.19, + "learning_rate": 3.0119261349063865e-05, + "loss": 0.4, + "step": 31056 + }, + { + "epoch": 1.19, + "learning_rate": 3.0118620159015133e-05, + "loss": 0.4996, + "step": 31057 + }, + { + "epoch": 1.19, + "learning_rate": 3.0117978968966405e-05, + "loss": 0.3648, + "step": 31058 + }, + { + "epoch": 1.19, + "learning_rate": 3.011733777891767e-05, + "loss": 0.4221, + "step": 31059 + }, + { + "epoch": 1.19, + "learning_rate": 3.011669658886894e-05, + "loss": 0.4013, + "step": 31060 + }, + { + "epoch": 1.19, + "learning_rate": 3.0116055398820213e-05, + "loss": 0.3703, + "step": 31061 + }, + { + "epoch": 1.19, + "learning_rate": 3.011541420877148e-05, + "loss": 0.3636, + "step": 31062 + }, + { + "epoch": 1.19, + "learning_rate": 3.0114773018722753e-05, + "loss": 0.4711, + "step": 31063 + }, + { + "epoch": 1.19, + "learning_rate": 3.0114131828674018e-05, + "loss": 0.3177, + "step": 31064 + }, + { + "epoch": 1.19, + "learning_rate": 3.011349063862529e-05, + "loss": 0.3951, + "step": 31065 + }, + { + "epoch": 1.19, + "learning_rate": 3.0112849448576558e-05, + "loss": 0.3323, + "step": 31066 + }, + { + "epoch": 1.19, + "learning_rate": 3.011220825852783e-05, + "loss": 0.4136, + "step": 31067 + }, + { + "epoch": 1.19, + "learning_rate": 3.0111567068479095e-05, + "loss": 0.6209, + "step": 31068 + }, + { + "epoch": 1.19, + "learning_rate": 3.011092587843037e-05, + "loss": 0.4291, + "step": 31069 + }, + { + "epoch": 1.19, + "learning_rate": 3.011028468838164e-05, + "loss": 0.3568, + "step": 31070 + }, + { + "epoch": 1.19, + "learning_rate": 3.0109643498332907e-05, + "loss": 0.4903, + "step": 31071 + }, + { + "epoch": 1.19, + "learning_rate": 3.0109002308284178e-05, + "loss": 0.4544, + "step": 31072 + }, + { + "epoch": 1.19, + "learning_rate": 3.0108361118235447e-05, + "loss": 0.3824, + "step": 31073 + }, + { + "epoch": 1.19, + "learning_rate": 3.0107719928186718e-05, + "loss": 0.3844, + "step": 31074 + }, + { + "epoch": 1.19, + "learning_rate": 3.0107078738137983e-05, + "loss": 0.2529, + "step": 31075 + }, + { + "epoch": 1.19, + "learning_rate": 3.0106437548089255e-05, + "loss": 0.4322, + "step": 31076 + }, + { + "epoch": 1.19, + "learning_rate": 3.0105796358040523e-05, + "loss": 0.4451, + "step": 31077 + }, + { + "epoch": 1.19, + "learning_rate": 3.0105155167991795e-05, + "loss": 0.4327, + "step": 31078 + }, + { + "epoch": 1.19, + "learning_rate": 3.0104513977943067e-05, + "loss": 0.3381, + "step": 31079 + }, + { + "epoch": 1.19, + "learning_rate": 3.010387278789433e-05, + "loss": 0.43, + "step": 31080 + }, + { + "epoch": 1.19, + "learning_rate": 3.0103231597845603e-05, + "loss": 0.42, + "step": 31081 + }, + { + "epoch": 1.2, + "learning_rate": 3.010259040779687e-05, + "loss": 0.2724, + "step": 31082 + }, + { + "epoch": 1.2, + "learning_rate": 3.0101949217748143e-05, + "loss": 0.4033, + "step": 31083 + }, + { + "epoch": 1.2, + "learning_rate": 3.010130802769941e-05, + "loss": 0.3991, + "step": 31084 + }, + { + "epoch": 1.2, + "learning_rate": 3.010066683765068e-05, + "loss": 0.3969, + "step": 31085 + }, + { + "epoch": 1.2, + "learning_rate": 3.010002564760195e-05, + "loss": 0.2733, + "step": 31086 + }, + { + "epoch": 1.2, + "learning_rate": 3.009938445755322e-05, + "loss": 0.3369, + "step": 31087 + }, + { + "epoch": 1.2, + "learning_rate": 3.0098743267504492e-05, + "loss": 0.3932, + "step": 31088 + }, + { + "epoch": 1.2, + "learning_rate": 3.009810207745576e-05, + "loss": 0.3381, + "step": 31089 + }, + { + "epoch": 1.2, + "learning_rate": 3.0097460887407032e-05, + "loss": 0.4633, + "step": 31090 + }, + { + "epoch": 1.2, + "learning_rate": 3.0096819697358297e-05, + "loss": 0.4284, + "step": 31091 + }, + { + "epoch": 1.2, + "learning_rate": 3.009617850730957e-05, + "loss": 0.2841, + "step": 31092 + }, + { + "epoch": 1.2, + "learning_rate": 3.0095537317260837e-05, + "loss": 0.3332, + "step": 31093 + }, + { + "epoch": 1.2, + "learning_rate": 3.009489612721211e-05, + "loss": 0.3777, + "step": 31094 + }, + { + "epoch": 1.2, + "learning_rate": 3.0094254937163374e-05, + "loss": 0.332, + "step": 31095 + }, + { + "epoch": 1.2, + "learning_rate": 3.0093613747114645e-05, + "loss": 0.3424, + "step": 31096 + }, + { + "epoch": 1.2, + "learning_rate": 3.0092972557065917e-05, + "loss": 0.3149, + "step": 31097 + }, + { + "epoch": 1.2, + "learning_rate": 3.0092331367017185e-05, + "loss": 0.3157, + "step": 31098 + }, + { + "epoch": 1.2, + "learning_rate": 3.0091690176968457e-05, + "loss": 0.3009, + "step": 31099 + }, + { + "epoch": 1.2, + "learning_rate": 3.0091048986919722e-05, + "loss": 0.263, + "step": 31100 + }, + { + "epoch": 1.2, + "learning_rate": 3.0090407796870994e-05, + "loss": 0.395, + "step": 31101 + }, + { + "epoch": 1.2, + "learning_rate": 3.0089766606822262e-05, + "loss": 0.293, + "step": 31102 + }, + { + "epoch": 1.2, + "learning_rate": 3.0089125416773534e-05, + "loss": 0.4255, + "step": 31103 + }, + { + "epoch": 1.2, + "learning_rate": 3.00884842267248e-05, + "loss": 0.617, + "step": 31104 + }, + { + "epoch": 1.2, + "learning_rate": 3.0087843036676074e-05, + "loss": 0.3812, + "step": 31105 + }, + { + "epoch": 1.2, + "learning_rate": 3.0087201846627345e-05, + "loss": 0.4282, + "step": 31106 + }, + { + "epoch": 1.2, + "learning_rate": 3.008656065657861e-05, + "loss": 0.2319, + "step": 31107 + }, + { + "epoch": 1.2, + "learning_rate": 3.0085919466529882e-05, + "loss": 0.6438, + "step": 31108 + }, + { + "epoch": 1.2, + "learning_rate": 3.008527827648115e-05, + "loss": 0.3665, + "step": 31109 + }, + { + "epoch": 1.2, + "learning_rate": 3.0084637086432422e-05, + "loss": 0.3944, + "step": 31110 + }, + { + "epoch": 1.2, + "learning_rate": 3.0083995896383687e-05, + "loss": 0.3588, + "step": 31111 + }, + { + "epoch": 1.2, + "learning_rate": 3.008335470633496e-05, + "loss": 0.4046, + "step": 31112 + }, + { + "epoch": 1.2, + "learning_rate": 3.0082713516286227e-05, + "loss": 0.3264, + "step": 31113 + }, + { + "epoch": 1.2, + "learning_rate": 3.00820723262375e-05, + "loss": 0.363, + "step": 31114 + }, + { + "epoch": 1.2, + "learning_rate": 3.008143113618877e-05, + "loss": 0.3475, + "step": 31115 + }, + { + "epoch": 1.2, + "learning_rate": 3.0080789946140036e-05, + "loss": 0.4043, + "step": 31116 + }, + { + "epoch": 1.2, + "learning_rate": 3.0080148756091307e-05, + "loss": 0.3509, + "step": 31117 + }, + { + "epoch": 1.2, + "learning_rate": 3.0079507566042576e-05, + "loss": 0.3004, + "step": 31118 + }, + { + "epoch": 1.2, + "learning_rate": 3.0078866375993847e-05, + "loss": 0.4065, + "step": 31119 + }, + { + "epoch": 1.2, + "learning_rate": 3.0078225185945112e-05, + "loss": 0.4593, + "step": 31120 + }, + { + "epoch": 1.2, + "learning_rate": 3.0077583995896387e-05, + "loss": 0.3842, + "step": 31121 + }, + { + "epoch": 1.2, + "learning_rate": 3.0076942805847652e-05, + "loss": 0.3517, + "step": 31122 + }, + { + "epoch": 1.2, + "learning_rate": 3.0076301615798924e-05, + "loss": 0.4818, + "step": 31123 + }, + { + "epoch": 1.2, + "learning_rate": 3.0075660425750196e-05, + "loss": 0.419, + "step": 31124 + }, + { + "epoch": 1.2, + "learning_rate": 3.0075019235701464e-05, + "loss": 0.3837, + "step": 31125 + }, + { + "epoch": 1.2, + "learning_rate": 3.0074378045652736e-05, + "loss": 0.4652, + "step": 31126 + }, + { + "epoch": 1.2, + "learning_rate": 3.0073736855604e-05, + "loss": 0.2753, + "step": 31127 + }, + { + "epoch": 1.2, + "learning_rate": 3.0073095665555272e-05, + "loss": 0.5754, + "step": 31128 + }, + { + "epoch": 1.2, + "learning_rate": 3.007245447550654e-05, + "loss": 0.3973, + "step": 31129 + }, + { + "epoch": 1.2, + "learning_rate": 3.0071813285457812e-05, + "loss": 0.601, + "step": 31130 + }, + { + "epoch": 1.2, + "learning_rate": 3.0071172095409077e-05, + "loss": 0.4287, + "step": 31131 + }, + { + "epoch": 1.2, + "learning_rate": 3.007053090536035e-05, + "loss": 0.5073, + "step": 31132 + }, + { + "epoch": 1.2, + "learning_rate": 3.006988971531162e-05, + "loss": 0.3581, + "step": 31133 + }, + { + "epoch": 1.2, + "learning_rate": 3.006924852526289e-05, + "loss": 0.4498, + "step": 31134 + }, + { + "epoch": 1.2, + "learning_rate": 3.006860733521416e-05, + "loss": 0.3385, + "step": 31135 + }, + { + "epoch": 1.2, + "learning_rate": 3.0067966145165426e-05, + "loss": 0.4821, + "step": 31136 + }, + { + "epoch": 1.2, + "learning_rate": 3.0067324955116698e-05, + "loss": 0.3296, + "step": 31137 + }, + { + "epoch": 1.2, + "learning_rate": 3.0066683765067966e-05, + "loss": 0.3492, + "step": 31138 + }, + { + "epoch": 1.2, + "learning_rate": 3.0066042575019238e-05, + "loss": 0.4932, + "step": 31139 + }, + { + "epoch": 1.2, + "learning_rate": 3.0065401384970503e-05, + "loss": 0.322, + "step": 31140 + }, + { + "epoch": 1.2, + "learning_rate": 3.0064760194921778e-05, + "loss": 0.4859, + "step": 31141 + }, + { + "epoch": 1.2, + "learning_rate": 3.006411900487305e-05, + "loss": 0.4336, + "step": 31142 + }, + { + "epoch": 1.2, + "learning_rate": 3.0063477814824314e-05, + "loss": 0.3177, + "step": 31143 + }, + { + "epoch": 1.2, + "learning_rate": 3.0062836624775586e-05, + "loss": 0.3541, + "step": 31144 + }, + { + "epoch": 1.2, + "learning_rate": 3.0062195434726854e-05, + "loss": 0.3641, + "step": 31145 + }, + { + "epoch": 1.2, + "learning_rate": 3.0061554244678126e-05, + "loss": 0.4348, + "step": 31146 + }, + { + "epoch": 1.2, + "learning_rate": 3.006091305462939e-05, + "loss": 0.2937, + "step": 31147 + }, + { + "epoch": 1.2, + "learning_rate": 3.0060271864580663e-05, + "loss": 0.4692, + "step": 31148 + }, + { + "epoch": 1.2, + "learning_rate": 3.005963067453193e-05, + "loss": 0.332, + "step": 31149 + }, + { + "epoch": 1.2, + "learning_rate": 3.0058989484483203e-05, + "loss": 0.483, + "step": 31150 + }, + { + "epoch": 1.2, + "learning_rate": 3.0058348294434474e-05, + "loss": 0.4347, + "step": 31151 + }, + { + "epoch": 1.2, + "learning_rate": 3.005770710438574e-05, + "loss": 0.3492, + "step": 31152 + }, + { + "epoch": 1.2, + "learning_rate": 3.005706591433701e-05, + "loss": 0.3708, + "step": 31153 + }, + { + "epoch": 1.2, + "learning_rate": 3.005642472428828e-05, + "loss": 0.3508, + "step": 31154 + }, + { + "epoch": 1.2, + "learning_rate": 3.005578353423955e-05, + "loss": 0.4551, + "step": 31155 + }, + { + "epoch": 1.2, + "learning_rate": 3.0055142344190816e-05, + "loss": 0.3707, + "step": 31156 + }, + { + "epoch": 1.2, + "learning_rate": 3.005450115414209e-05, + "loss": 0.1838, + "step": 31157 + }, + { + "epoch": 1.2, + "learning_rate": 3.0053859964093356e-05, + "loss": 0.4288, + "step": 31158 + }, + { + "epoch": 1.2, + "learning_rate": 3.0053218774044628e-05, + "loss": 0.2955, + "step": 31159 + }, + { + "epoch": 1.2, + "learning_rate": 3.00525775839959e-05, + "loss": 0.4654, + "step": 31160 + }, + { + "epoch": 1.2, + "learning_rate": 3.0051936393947168e-05, + "loss": 0.4393, + "step": 31161 + }, + { + "epoch": 1.2, + "learning_rate": 3.005129520389844e-05, + "loss": 0.3909, + "step": 31162 + }, + { + "epoch": 1.2, + "learning_rate": 3.0050654013849705e-05, + "loss": 0.4268, + "step": 31163 + }, + { + "epoch": 1.2, + "learning_rate": 3.0050012823800976e-05, + "loss": 0.3481, + "step": 31164 + }, + { + "epoch": 1.2, + "learning_rate": 3.0049371633752245e-05, + "loss": 0.4012, + "step": 31165 + }, + { + "epoch": 1.2, + "learning_rate": 3.0048730443703516e-05, + "loss": 0.4942, + "step": 31166 + }, + { + "epoch": 1.2, + "learning_rate": 3.004808925365478e-05, + "loss": 0.3132, + "step": 31167 + }, + { + "epoch": 1.2, + "learning_rate": 3.0047448063606053e-05, + "loss": 0.2705, + "step": 31168 + }, + { + "epoch": 1.2, + "learning_rate": 3.0046806873557325e-05, + "loss": 0.3183, + "step": 31169 + }, + { + "epoch": 1.2, + "learning_rate": 3.0046165683508593e-05, + "loss": 0.3419, + "step": 31170 + }, + { + "epoch": 1.2, + "learning_rate": 3.0045524493459865e-05, + "loss": 0.4396, + "step": 31171 + }, + { + "epoch": 1.2, + "learning_rate": 3.004488330341113e-05, + "loss": 0.4265, + "step": 31172 + }, + { + "epoch": 1.2, + "learning_rate": 3.00442421133624e-05, + "loss": 0.3947, + "step": 31173 + }, + { + "epoch": 1.2, + "learning_rate": 3.004360092331367e-05, + "loss": 0.4911, + "step": 31174 + }, + { + "epoch": 1.2, + "learning_rate": 3.004295973326494e-05, + "loss": 0.4348, + "step": 31175 + }, + { + "epoch": 1.2, + "learning_rate": 3.0042318543216206e-05, + "loss": 0.3615, + "step": 31176 + }, + { + "epoch": 1.2, + "learning_rate": 3.004167735316748e-05, + "loss": 0.4169, + "step": 31177 + }, + { + "epoch": 1.2, + "learning_rate": 3.0041036163118753e-05, + "loss": 0.6074, + "step": 31178 + }, + { + "epoch": 1.2, + "learning_rate": 3.0040394973070018e-05, + "loss": 0.447, + "step": 31179 + }, + { + "epoch": 1.2, + "learning_rate": 3.003975378302129e-05, + "loss": 0.3376, + "step": 31180 + }, + { + "epoch": 1.2, + "learning_rate": 3.0039112592972558e-05, + "loss": 0.3786, + "step": 31181 + }, + { + "epoch": 1.2, + "learning_rate": 3.003847140292383e-05, + "loss": 0.3437, + "step": 31182 + }, + { + "epoch": 1.2, + "learning_rate": 3.0037830212875095e-05, + "loss": 0.436, + "step": 31183 + }, + { + "epoch": 1.2, + "learning_rate": 3.0037189022826367e-05, + "loss": 0.4149, + "step": 31184 + }, + { + "epoch": 1.2, + "learning_rate": 3.0036547832777635e-05, + "loss": 0.3222, + "step": 31185 + }, + { + "epoch": 1.2, + "learning_rate": 3.0035906642728907e-05, + "loss": 0.4053, + "step": 31186 + }, + { + "epoch": 1.2, + "learning_rate": 3.003526545268018e-05, + "loss": 0.3903, + "step": 31187 + }, + { + "epoch": 1.2, + "learning_rate": 3.0034624262631443e-05, + "loss": 0.3141, + "step": 31188 + }, + { + "epoch": 1.2, + "learning_rate": 3.0033983072582715e-05, + "loss": 0.4831, + "step": 31189 + }, + { + "epoch": 1.2, + "learning_rate": 3.0033341882533983e-05, + "loss": 0.5167, + "step": 31190 + }, + { + "epoch": 1.2, + "learning_rate": 3.0032700692485255e-05, + "loss": 0.3933, + "step": 31191 + }, + { + "epoch": 1.2, + "learning_rate": 3.003205950243652e-05, + "loss": 0.4686, + "step": 31192 + }, + { + "epoch": 1.2, + "learning_rate": 3.0031418312387795e-05, + "loss": 0.4312, + "step": 31193 + }, + { + "epoch": 1.2, + "learning_rate": 3.0030777122339067e-05, + "loss": 0.4665, + "step": 31194 + }, + { + "epoch": 1.2, + "learning_rate": 3.0030135932290332e-05, + "loss": 0.4631, + "step": 31195 + }, + { + "epoch": 1.2, + "learning_rate": 3.0029494742241603e-05, + "loss": 0.4752, + "step": 31196 + }, + { + "epoch": 1.2, + "learning_rate": 3.0028853552192872e-05, + "loss": 0.3921, + "step": 31197 + }, + { + "epoch": 1.2, + "learning_rate": 3.0028212362144143e-05, + "loss": 0.4198, + "step": 31198 + }, + { + "epoch": 1.2, + "learning_rate": 3.002757117209541e-05, + "loss": 0.4247, + "step": 31199 + }, + { + "epoch": 1.2, + "learning_rate": 3.002692998204668e-05, + "loss": 0.4605, + "step": 31200 + }, + { + "epoch": 1.2, + "learning_rate": 3.002628879199795e-05, + "loss": 0.3911, + "step": 31201 + }, + { + "epoch": 1.2, + "learning_rate": 3.002564760194922e-05, + "loss": 0.5089, + "step": 31202 + }, + { + "epoch": 1.2, + "learning_rate": 3.0025006411900492e-05, + "loss": 0.4465, + "step": 31203 + }, + { + "epoch": 1.2, + "learning_rate": 3.0024365221851757e-05, + "loss": 0.4061, + "step": 31204 + }, + { + "epoch": 1.2, + "learning_rate": 3.002372403180303e-05, + "loss": 0.2913, + "step": 31205 + }, + { + "epoch": 1.2, + "learning_rate": 3.0023082841754297e-05, + "loss": 0.4672, + "step": 31206 + }, + { + "epoch": 1.2, + "learning_rate": 3.002244165170557e-05, + "loss": 0.5059, + "step": 31207 + }, + { + "epoch": 1.2, + "learning_rate": 3.0021800461656834e-05, + "loss": 0.4683, + "step": 31208 + }, + { + "epoch": 1.2, + "learning_rate": 3.0021159271608105e-05, + "loss": 0.519, + "step": 31209 + }, + { + "epoch": 1.2, + "learning_rate": 3.0020518081559374e-05, + "loss": 0.5031, + "step": 31210 + }, + { + "epoch": 1.2, + "learning_rate": 3.0019876891510645e-05, + "loss": 0.2868, + "step": 31211 + }, + { + "epoch": 1.2, + "learning_rate": 3.0019235701461917e-05, + "loss": 0.3654, + "step": 31212 + }, + { + "epoch": 1.2, + "learning_rate": 3.0018594511413185e-05, + "loss": 0.3725, + "step": 31213 + }, + { + "epoch": 1.2, + "learning_rate": 3.0017953321364457e-05, + "loss": 0.3774, + "step": 31214 + }, + { + "epoch": 1.2, + "learning_rate": 3.0017312131315722e-05, + "loss": 0.3367, + "step": 31215 + }, + { + "epoch": 1.2, + "learning_rate": 3.0016670941266994e-05, + "loss": 0.3376, + "step": 31216 + }, + { + "epoch": 1.2, + "learning_rate": 3.0016029751218262e-05, + "loss": 0.2773, + "step": 31217 + }, + { + "epoch": 1.2, + "learning_rate": 3.0015388561169534e-05, + "loss": 0.3631, + "step": 31218 + }, + { + "epoch": 1.2, + "learning_rate": 3.00147473711208e-05, + "loss": 0.4851, + "step": 31219 + }, + { + "epoch": 1.2, + "learning_rate": 3.001410618107207e-05, + "loss": 0.3444, + "step": 31220 + }, + { + "epoch": 1.2, + "learning_rate": 3.0013464991023342e-05, + "loss": 0.294, + "step": 31221 + }, + { + "epoch": 1.2, + "learning_rate": 3.001282380097461e-05, + "loss": 0.4755, + "step": 31222 + }, + { + "epoch": 1.2, + "learning_rate": 3.0012182610925882e-05, + "loss": 0.3272, + "step": 31223 + }, + { + "epoch": 1.2, + "learning_rate": 3.0011541420877147e-05, + "loss": 0.4277, + "step": 31224 + }, + { + "epoch": 1.2, + "learning_rate": 3.001090023082842e-05, + "loss": 0.3644, + "step": 31225 + }, + { + "epoch": 1.2, + "learning_rate": 3.0010259040779687e-05, + "loss": 0.3048, + "step": 31226 + }, + { + "epoch": 1.2, + "learning_rate": 3.000961785073096e-05, + "loss": 0.3949, + "step": 31227 + }, + { + "epoch": 1.2, + "learning_rate": 3.0008976660682224e-05, + "loss": 0.4523, + "step": 31228 + }, + { + "epoch": 1.2, + "learning_rate": 3.00083354706335e-05, + "loss": 0.3613, + "step": 31229 + }, + { + "epoch": 1.2, + "learning_rate": 3.000769428058477e-05, + "loss": 0.2947, + "step": 31230 + }, + { + "epoch": 1.2, + "learning_rate": 3.0007053090536036e-05, + "loss": 0.3056, + "step": 31231 + }, + { + "epoch": 1.2, + "learning_rate": 3.0006411900487307e-05, + "loss": 0.3176, + "step": 31232 + }, + { + "epoch": 1.2, + "learning_rate": 3.0005770710438576e-05, + "loss": 0.4304, + "step": 31233 + }, + { + "epoch": 1.2, + "learning_rate": 3.0005129520389847e-05, + "loss": 0.2916, + "step": 31234 + }, + { + "epoch": 1.2, + "learning_rate": 3.0004488330341112e-05, + "loss": 0.48, + "step": 31235 + }, + { + "epoch": 1.2, + "learning_rate": 3.0003847140292384e-05, + "loss": 0.4264, + "step": 31236 + }, + { + "epoch": 1.2, + "learning_rate": 3.0003205950243652e-05, + "loss": 0.3987, + "step": 31237 + }, + { + "epoch": 1.2, + "learning_rate": 3.0002564760194924e-05, + "loss": 0.3193, + "step": 31238 + }, + { + "epoch": 1.2, + "learning_rate": 3.0001923570146196e-05, + "loss": 0.4253, + "step": 31239 + }, + { + "epoch": 1.2, + "learning_rate": 3.000128238009746e-05, + "loss": 0.5501, + "step": 31240 + }, + { + "epoch": 1.2, + "learning_rate": 3.0000641190048732e-05, + "loss": 0.3063, + "step": 31241 + }, + { + "epoch": 1.2, + "learning_rate": 3e-05, + "loss": 0.3743, + "step": 31242 + }, + { + "epoch": 1.2, + "learning_rate": 2.9999358809951272e-05, + "loss": 0.3476, + "step": 31243 + }, + { + "epoch": 1.2, + "learning_rate": 2.9998717619902537e-05, + "loss": 0.4234, + "step": 31244 + }, + { + "epoch": 1.2, + "learning_rate": 2.9998076429853813e-05, + "loss": 0.3011, + "step": 31245 + }, + { + "epoch": 1.2, + "learning_rate": 2.9997435239805077e-05, + "loss": 0.4286, + "step": 31246 + }, + { + "epoch": 1.2, + "learning_rate": 2.999679404975635e-05, + "loss": 0.6136, + "step": 31247 + }, + { + "epoch": 1.2, + "learning_rate": 2.999615285970762e-05, + "loss": 0.3307, + "step": 31248 + }, + { + "epoch": 1.2, + "learning_rate": 2.999551166965889e-05, + "loss": 0.5391, + "step": 31249 + }, + { + "epoch": 1.2, + "learning_rate": 2.999487047961016e-05, + "loss": 0.4131, + "step": 31250 + }, + { + "epoch": 1.2, + "learning_rate": 2.9994229289561426e-05, + "loss": 0.456, + "step": 31251 + }, + { + "epoch": 1.2, + "learning_rate": 2.9993588099512698e-05, + "loss": 0.3735, + "step": 31252 + }, + { + "epoch": 1.2, + "learning_rate": 2.9992946909463966e-05, + "loss": 0.2303, + "step": 31253 + }, + { + "epoch": 1.2, + "learning_rate": 2.9992305719415238e-05, + "loss": 0.5311, + "step": 31254 + }, + { + "epoch": 1.2, + "learning_rate": 2.9991664529366503e-05, + "loss": 0.4408, + "step": 31255 + }, + { + "epoch": 1.2, + "learning_rate": 2.9991023339317774e-05, + "loss": 0.3745, + "step": 31256 + }, + { + "epoch": 1.2, + "learning_rate": 2.9990382149269046e-05, + "loss": 0.321, + "step": 31257 + }, + { + "epoch": 1.2, + "learning_rate": 2.9989740959220314e-05, + "loss": 0.344, + "step": 31258 + }, + { + "epoch": 1.2, + "learning_rate": 2.9989099769171586e-05, + "loss": 0.3757, + "step": 31259 + }, + { + "epoch": 1.2, + "learning_rate": 2.998845857912285e-05, + "loss": 0.3749, + "step": 31260 + }, + { + "epoch": 1.2, + "learning_rate": 2.9987817389074123e-05, + "loss": 0.3993, + "step": 31261 + }, + { + "epoch": 1.2, + "learning_rate": 2.998717619902539e-05, + "loss": 0.3884, + "step": 31262 + }, + { + "epoch": 1.2, + "learning_rate": 2.9986535008976663e-05, + "loss": 0.3512, + "step": 31263 + }, + { + "epoch": 1.2, + "learning_rate": 2.9985893818927928e-05, + "loss": 0.5229, + "step": 31264 + }, + { + "epoch": 1.2, + "learning_rate": 2.9985252628879203e-05, + "loss": 0.4135, + "step": 31265 + }, + { + "epoch": 1.2, + "learning_rate": 2.9984611438830475e-05, + "loss": 0.4464, + "step": 31266 + }, + { + "epoch": 1.2, + "learning_rate": 2.998397024878174e-05, + "loss": 0.3658, + "step": 31267 + }, + { + "epoch": 1.2, + "learning_rate": 2.998332905873301e-05, + "loss": 0.5096, + "step": 31268 + }, + { + "epoch": 1.2, + "learning_rate": 2.998268786868428e-05, + "loss": 0.471, + "step": 31269 + }, + { + "epoch": 1.2, + "learning_rate": 2.998204667863555e-05, + "loss": 0.3784, + "step": 31270 + }, + { + "epoch": 1.2, + "learning_rate": 2.9981405488586816e-05, + "loss": 0.3693, + "step": 31271 + }, + { + "epoch": 1.2, + "learning_rate": 2.9980764298538088e-05, + "loss": 0.4689, + "step": 31272 + }, + { + "epoch": 1.2, + "learning_rate": 2.9980123108489356e-05, + "loss": 0.3654, + "step": 31273 + }, + { + "epoch": 1.2, + "learning_rate": 2.9979481918440628e-05, + "loss": 0.3367, + "step": 31274 + }, + { + "epoch": 1.2, + "learning_rate": 2.99788407283919e-05, + "loss": 0.412, + "step": 31275 + }, + { + "epoch": 1.2, + "learning_rate": 2.9978199538343165e-05, + "loss": 0.3699, + "step": 31276 + }, + { + "epoch": 1.2, + "learning_rate": 2.9977558348294436e-05, + "loss": 0.4023, + "step": 31277 + }, + { + "epoch": 1.2, + "learning_rate": 2.9976917158245705e-05, + "loss": 0.3103, + "step": 31278 + }, + { + "epoch": 1.2, + "learning_rate": 2.9976275968196976e-05, + "loss": 0.3463, + "step": 31279 + }, + { + "epoch": 1.2, + "learning_rate": 2.997563477814824e-05, + "loss": 0.3705, + "step": 31280 + }, + { + "epoch": 1.2, + "learning_rate": 2.9974993588099516e-05, + "loss": 0.3326, + "step": 31281 + }, + { + "epoch": 1.2, + "learning_rate": 2.997435239805078e-05, + "loss": 0.4364, + "step": 31282 + }, + { + "epoch": 1.2, + "learning_rate": 2.9973711208002053e-05, + "loss": 0.427, + "step": 31283 + }, + { + "epoch": 1.2, + "learning_rate": 2.9973070017953325e-05, + "loss": 0.4415, + "step": 31284 + }, + { + "epoch": 1.2, + "learning_rate": 2.9972428827904593e-05, + "loss": 0.4847, + "step": 31285 + }, + { + "epoch": 1.2, + "learning_rate": 2.9971787637855865e-05, + "loss": 0.3315, + "step": 31286 + }, + { + "epoch": 1.2, + "learning_rate": 2.997114644780713e-05, + "loss": 0.468, + "step": 31287 + }, + { + "epoch": 1.2, + "learning_rate": 2.99705052577584e-05, + "loss": 0.3491, + "step": 31288 + }, + { + "epoch": 1.2, + "learning_rate": 2.996986406770967e-05, + "loss": 0.2932, + "step": 31289 + }, + { + "epoch": 1.2, + "learning_rate": 2.996922287766094e-05, + "loss": 0.3633, + "step": 31290 + }, + { + "epoch": 1.2, + "learning_rate": 2.9968581687612206e-05, + "loss": 0.42, + "step": 31291 + }, + { + "epoch": 1.2, + "learning_rate": 2.9967940497563478e-05, + "loss": 0.3875, + "step": 31292 + }, + { + "epoch": 1.2, + "learning_rate": 2.996729930751475e-05, + "loss": 0.4775, + "step": 31293 + }, + { + "epoch": 1.2, + "learning_rate": 2.9966658117466018e-05, + "loss": 0.5213, + "step": 31294 + }, + { + "epoch": 1.2, + "learning_rate": 2.996601692741729e-05, + "loss": 0.3078, + "step": 31295 + }, + { + "epoch": 1.2, + "learning_rate": 2.9965375737368555e-05, + "loss": 0.3253, + "step": 31296 + }, + { + "epoch": 1.2, + "learning_rate": 2.9964734547319827e-05, + "loss": 0.3597, + "step": 31297 + }, + { + "epoch": 1.2, + "learning_rate": 2.9964093357271095e-05, + "loss": 0.3326, + "step": 31298 + }, + { + "epoch": 1.2, + "learning_rate": 2.9963452167222367e-05, + "loss": 0.479, + "step": 31299 + }, + { + "epoch": 1.2, + "learning_rate": 2.996281097717363e-05, + "loss": 0.4078, + "step": 31300 + }, + { + "epoch": 1.2, + "learning_rate": 2.9962169787124907e-05, + "loss": 0.3484, + "step": 31301 + }, + { + "epoch": 1.2, + "learning_rate": 2.996152859707618e-05, + "loss": 0.36, + "step": 31302 + }, + { + "epoch": 1.2, + "learning_rate": 2.9960887407027443e-05, + "loss": 0.482, + "step": 31303 + }, + { + "epoch": 1.2, + "learning_rate": 2.9960246216978715e-05, + "loss": 0.2637, + "step": 31304 + }, + { + "epoch": 1.2, + "learning_rate": 2.9959605026929983e-05, + "loss": 0.417, + "step": 31305 + }, + { + "epoch": 1.2, + "learning_rate": 2.9958963836881255e-05, + "loss": 0.3307, + "step": 31306 + }, + { + "epoch": 1.2, + "learning_rate": 2.995832264683252e-05, + "loss": 0.3642, + "step": 31307 + }, + { + "epoch": 1.2, + "learning_rate": 2.9957681456783792e-05, + "loss": 0.3567, + "step": 31308 + }, + { + "epoch": 1.2, + "learning_rate": 2.995704026673506e-05, + "loss": 0.3662, + "step": 31309 + }, + { + "epoch": 1.2, + "learning_rate": 2.9956399076686332e-05, + "loss": 0.4077, + "step": 31310 + }, + { + "epoch": 1.2, + "learning_rate": 2.9955757886637604e-05, + "loss": 0.5102, + "step": 31311 + }, + { + "epoch": 1.2, + "learning_rate": 2.995511669658887e-05, + "loss": 0.3722, + "step": 31312 + }, + { + "epoch": 1.2, + "learning_rate": 2.995447550654014e-05, + "loss": 0.3915, + "step": 31313 + }, + { + "epoch": 1.2, + "learning_rate": 2.995383431649141e-05, + "loss": 0.3938, + "step": 31314 + }, + { + "epoch": 1.2, + "learning_rate": 2.995319312644268e-05, + "loss": 0.3886, + "step": 31315 + }, + { + "epoch": 1.2, + "learning_rate": 2.9952551936393945e-05, + "loss": 0.4171, + "step": 31316 + }, + { + "epoch": 1.2, + "learning_rate": 2.995191074634522e-05, + "loss": 0.3793, + "step": 31317 + }, + { + "epoch": 1.2, + "learning_rate": 2.9951269556296485e-05, + "loss": 0.4059, + "step": 31318 + }, + { + "epoch": 1.2, + "learning_rate": 2.9950628366247757e-05, + "loss": 0.3947, + "step": 31319 + }, + { + "epoch": 1.2, + "learning_rate": 2.994998717619903e-05, + "loss": 0.4268, + "step": 31320 + }, + { + "epoch": 1.2, + "learning_rate": 2.9949345986150297e-05, + "loss": 0.3697, + "step": 31321 + }, + { + "epoch": 1.2, + "learning_rate": 2.994870479610157e-05, + "loss": 0.4397, + "step": 31322 + }, + { + "epoch": 1.2, + "learning_rate": 2.9948063606052834e-05, + "loss": 0.3681, + "step": 31323 + }, + { + "epoch": 1.2, + "learning_rate": 2.9947422416004105e-05, + "loss": 0.2945, + "step": 31324 + }, + { + "epoch": 1.2, + "learning_rate": 2.9946781225955374e-05, + "loss": 0.3698, + "step": 31325 + }, + { + "epoch": 1.2, + "learning_rate": 2.9946140035906645e-05, + "loss": 0.3921, + "step": 31326 + }, + { + "epoch": 1.2, + "learning_rate": 2.994549884585791e-05, + "loss": 0.4145, + "step": 31327 + }, + { + "epoch": 1.2, + "learning_rate": 2.9944857655809182e-05, + "loss": 0.4111, + "step": 31328 + }, + { + "epoch": 1.2, + "learning_rate": 2.9944216465760454e-05, + "loss": 0.3206, + "step": 31329 + }, + { + "epoch": 1.2, + "learning_rate": 2.9943575275711722e-05, + "loss": 0.4102, + "step": 31330 + }, + { + "epoch": 1.2, + "learning_rate": 2.9942934085662994e-05, + "loss": 0.381, + "step": 31331 + }, + { + "epoch": 1.2, + "learning_rate": 2.994229289561426e-05, + "loss": 0.3884, + "step": 31332 + }, + { + "epoch": 1.2, + "learning_rate": 2.994165170556553e-05, + "loss": 0.3732, + "step": 31333 + }, + { + "epoch": 1.2, + "learning_rate": 2.99410105155168e-05, + "loss": 0.4967, + "step": 31334 + }, + { + "epoch": 1.2, + "learning_rate": 2.994036932546807e-05, + "loss": 0.4406, + "step": 31335 + }, + { + "epoch": 1.2, + "learning_rate": 2.9939728135419335e-05, + "loss": 0.3426, + "step": 31336 + }, + { + "epoch": 1.2, + "learning_rate": 2.993908694537061e-05, + "loss": 0.4896, + "step": 31337 + }, + { + "epoch": 1.2, + "learning_rate": 2.9938445755321882e-05, + "loss": 0.4192, + "step": 31338 + }, + { + "epoch": 1.2, + "learning_rate": 2.9937804565273147e-05, + "loss": 0.4128, + "step": 31339 + }, + { + "epoch": 1.2, + "learning_rate": 2.993716337522442e-05, + "loss": 0.3603, + "step": 31340 + }, + { + "epoch": 1.2, + "learning_rate": 2.9936522185175687e-05, + "loss": 0.3128, + "step": 31341 + }, + { + "epoch": 1.2, + "learning_rate": 2.993588099512696e-05, + "loss": 0.2672, + "step": 31342 + }, + { + "epoch": 1.21, + "learning_rate": 2.9935239805078224e-05, + "loss": 0.3962, + "step": 31343 + }, + { + "epoch": 1.21, + "learning_rate": 2.9934598615029496e-05, + "loss": 0.4449, + "step": 31344 + }, + { + "epoch": 1.21, + "learning_rate": 2.9933957424980764e-05, + "loss": 0.3535, + "step": 31345 + }, + { + "epoch": 1.21, + "learning_rate": 2.9933316234932036e-05, + "loss": 0.4263, + "step": 31346 + }, + { + "epoch": 1.21, + "learning_rate": 2.9932675044883307e-05, + "loss": 0.3849, + "step": 31347 + }, + { + "epoch": 1.21, + "learning_rate": 2.9932033854834572e-05, + "loss": 0.2987, + "step": 31348 + }, + { + "epoch": 1.21, + "learning_rate": 2.9931392664785844e-05, + "loss": 0.5326, + "step": 31349 + }, + { + "epoch": 1.21, + "learning_rate": 2.9930751474737112e-05, + "loss": 0.5181, + "step": 31350 + }, + { + "epoch": 1.21, + "learning_rate": 2.9930110284688384e-05, + "loss": 0.3561, + "step": 31351 + }, + { + "epoch": 1.21, + "learning_rate": 2.992946909463965e-05, + "loss": 0.5138, + "step": 31352 + }, + { + "epoch": 1.21, + "learning_rate": 2.9928827904590924e-05, + "loss": 0.4996, + "step": 31353 + }, + { + "epoch": 1.21, + "learning_rate": 2.992818671454219e-05, + "loss": 0.4267, + "step": 31354 + }, + { + "epoch": 1.21, + "learning_rate": 2.992754552449346e-05, + "loss": 0.444, + "step": 31355 + }, + { + "epoch": 1.21, + "learning_rate": 2.9926904334444733e-05, + "loss": 0.3574, + "step": 31356 + }, + { + "epoch": 1.21, + "learning_rate": 2.9926263144396e-05, + "loss": 0.4123, + "step": 31357 + }, + { + "epoch": 1.21, + "learning_rate": 2.9925621954347273e-05, + "loss": 0.3475, + "step": 31358 + }, + { + "epoch": 1.21, + "learning_rate": 2.9924980764298537e-05, + "loss": 0.3331, + "step": 31359 + }, + { + "epoch": 1.21, + "learning_rate": 2.992433957424981e-05, + "loss": 0.4751, + "step": 31360 + }, + { + "epoch": 1.21, + "learning_rate": 2.9923698384201078e-05, + "loss": 0.5148, + "step": 31361 + }, + { + "epoch": 1.21, + "learning_rate": 2.992305719415235e-05, + "loss": 0.3148, + "step": 31362 + }, + { + "epoch": 1.21, + "learning_rate": 2.9922416004103614e-05, + "loss": 0.3552, + "step": 31363 + }, + { + "epoch": 1.21, + "learning_rate": 2.9921774814054886e-05, + "loss": 0.5032, + "step": 31364 + }, + { + "epoch": 1.21, + "learning_rate": 2.9921133624006158e-05, + "loss": 0.4196, + "step": 31365 + }, + { + "epoch": 1.21, + "learning_rate": 2.9920492433957426e-05, + "loss": 0.3409, + "step": 31366 + }, + { + "epoch": 1.21, + "learning_rate": 2.9919851243908698e-05, + "loss": 0.4246, + "step": 31367 + }, + { + "epoch": 1.21, + "learning_rate": 2.9919210053859963e-05, + "loss": 0.3917, + "step": 31368 + }, + { + "epoch": 1.21, + "learning_rate": 2.9918568863811234e-05, + "loss": 0.4716, + "step": 31369 + }, + { + "epoch": 1.21, + "learning_rate": 2.9917927673762503e-05, + "loss": 0.4031, + "step": 31370 + }, + { + "epoch": 1.21, + "learning_rate": 2.9917286483713774e-05, + "loss": 0.4991, + "step": 31371 + }, + { + "epoch": 1.21, + "learning_rate": 2.991664529366504e-05, + "loss": 0.412, + "step": 31372 + }, + { + "epoch": 1.21, + "learning_rate": 2.9916004103616314e-05, + "loss": 0.3954, + "step": 31373 + }, + { + "epoch": 1.21, + "learning_rate": 2.9915362913567586e-05, + "loss": 0.483, + "step": 31374 + }, + { + "epoch": 1.21, + "learning_rate": 2.991472172351885e-05, + "loss": 0.3187, + "step": 31375 + }, + { + "epoch": 1.21, + "learning_rate": 2.9914080533470123e-05, + "loss": 0.3162, + "step": 31376 + }, + { + "epoch": 1.21, + "learning_rate": 2.991343934342139e-05, + "loss": 0.269, + "step": 31377 + }, + { + "epoch": 1.21, + "learning_rate": 2.9912798153372663e-05, + "loss": 0.4895, + "step": 31378 + }, + { + "epoch": 1.21, + "learning_rate": 2.9912156963323928e-05, + "loss": 0.372, + "step": 31379 + }, + { + "epoch": 1.21, + "learning_rate": 2.99115157732752e-05, + "loss": 0.3832, + "step": 31380 + }, + { + "epoch": 1.21, + "learning_rate": 2.991087458322647e-05, + "loss": 0.4066, + "step": 31381 + }, + { + "epoch": 1.21, + "learning_rate": 2.991023339317774e-05, + "loss": 0.3855, + "step": 31382 + }, + { + "epoch": 1.21, + "learning_rate": 2.990959220312901e-05, + "loss": 0.3768, + "step": 31383 + }, + { + "epoch": 1.21, + "learning_rate": 2.9908951013080276e-05, + "loss": 0.4007, + "step": 31384 + }, + { + "epoch": 1.21, + "learning_rate": 2.9908309823031548e-05, + "loss": 0.4136, + "step": 31385 + }, + { + "epoch": 1.21, + "learning_rate": 2.9907668632982816e-05, + "loss": 0.4683, + "step": 31386 + }, + { + "epoch": 1.21, + "learning_rate": 2.9907027442934088e-05, + "loss": 0.4101, + "step": 31387 + }, + { + "epoch": 1.21, + "learning_rate": 2.9906386252885353e-05, + "loss": 0.3659, + "step": 31388 + }, + { + "epoch": 1.21, + "learning_rate": 2.9905745062836628e-05, + "loss": 0.4513, + "step": 31389 + }, + { + "epoch": 1.21, + "learning_rate": 2.99051038727879e-05, + "loss": 0.5118, + "step": 31390 + }, + { + "epoch": 1.21, + "learning_rate": 2.9904462682739165e-05, + "loss": 0.3551, + "step": 31391 + }, + { + "epoch": 1.21, + "learning_rate": 2.9903821492690436e-05, + "loss": 0.4197, + "step": 31392 + }, + { + "epoch": 1.21, + "learning_rate": 2.9903180302641705e-05, + "loss": 0.3727, + "step": 31393 + }, + { + "epoch": 1.21, + "learning_rate": 2.9902539112592976e-05, + "loss": 0.5018, + "step": 31394 + }, + { + "epoch": 1.21, + "learning_rate": 2.990189792254424e-05, + "loss": 0.4489, + "step": 31395 + }, + { + "epoch": 1.21, + "learning_rate": 2.9901256732495513e-05, + "loss": 0.3906, + "step": 31396 + }, + { + "epoch": 1.21, + "learning_rate": 2.990061554244678e-05, + "loss": 0.3112, + "step": 31397 + }, + { + "epoch": 1.21, + "learning_rate": 2.9899974352398053e-05, + "loss": 0.3623, + "step": 31398 + }, + { + "epoch": 1.21, + "learning_rate": 2.9899333162349325e-05, + "loss": 0.372, + "step": 31399 + }, + { + "epoch": 1.21, + "learning_rate": 2.989869197230059e-05, + "loss": 0.3107, + "step": 31400 + }, + { + "epoch": 1.21, + "learning_rate": 2.989805078225186e-05, + "loss": 0.3963, + "step": 31401 + }, + { + "epoch": 1.21, + "learning_rate": 2.989740959220313e-05, + "loss": 0.3669, + "step": 31402 + }, + { + "epoch": 1.21, + "learning_rate": 2.98967684021544e-05, + "loss": 0.2942, + "step": 31403 + }, + { + "epoch": 1.21, + "learning_rate": 2.9896127212105666e-05, + "loss": 0.4471, + "step": 31404 + }, + { + "epoch": 1.21, + "learning_rate": 2.989548602205694e-05, + "loss": 0.4537, + "step": 31405 + }, + { + "epoch": 1.21, + "learning_rate": 2.9894844832008207e-05, + "loss": 0.3647, + "step": 31406 + }, + { + "epoch": 1.21, + "learning_rate": 2.9894203641959478e-05, + "loss": 0.4229, + "step": 31407 + }, + { + "epoch": 1.21, + "learning_rate": 2.989356245191075e-05, + "loss": 0.5096, + "step": 31408 + }, + { + "epoch": 1.21, + "learning_rate": 2.9892921261862018e-05, + "loss": 0.4032, + "step": 31409 + }, + { + "epoch": 1.21, + "learning_rate": 2.989228007181329e-05, + "loss": 0.3128, + "step": 31410 + }, + { + "epoch": 1.21, + "learning_rate": 2.9891638881764555e-05, + "loss": 0.2291, + "step": 31411 + }, + { + "epoch": 1.21, + "learning_rate": 2.9890997691715827e-05, + "loss": 0.4051, + "step": 31412 + }, + { + "epoch": 1.21, + "learning_rate": 2.9890356501667095e-05, + "loss": 0.404, + "step": 31413 + }, + { + "epoch": 1.21, + "learning_rate": 2.9889715311618367e-05, + "loss": 0.4003, + "step": 31414 + }, + { + "epoch": 1.21, + "learning_rate": 2.988907412156963e-05, + "loss": 0.3844, + "step": 31415 + }, + { + "epoch": 1.21, + "learning_rate": 2.9888432931520903e-05, + "loss": 0.4429, + "step": 31416 + }, + { + "epoch": 1.21, + "learning_rate": 2.9887791741472175e-05, + "loss": 0.4621, + "step": 31417 + }, + { + "epoch": 1.21, + "learning_rate": 2.9887150551423443e-05, + "loss": 0.4812, + "step": 31418 + }, + { + "epoch": 1.21, + "learning_rate": 2.9886509361374715e-05, + "loss": 0.4785, + "step": 31419 + }, + { + "epoch": 1.21, + "learning_rate": 2.988586817132598e-05, + "loss": 0.3984, + "step": 31420 + }, + { + "epoch": 1.21, + "learning_rate": 2.9885226981277252e-05, + "loss": 0.321, + "step": 31421 + }, + { + "epoch": 1.21, + "learning_rate": 2.988458579122852e-05, + "loss": 0.2621, + "step": 31422 + }, + { + "epoch": 1.21, + "learning_rate": 2.9883944601179792e-05, + "loss": 0.2837, + "step": 31423 + }, + { + "epoch": 1.21, + "learning_rate": 2.9883303411131057e-05, + "loss": 0.3637, + "step": 31424 + }, + { + "epoch": 1.21, + "learning_rate": 2.9882662221082332e-05, + "loss": 0.2782, + "step": 31425 + }, + { + "epoch": 1.21, + "learning_rate": 2.9882021031033604e-05, + "loss": 0.4437, + "step": 31426 + }, + { + "epoch": 1.21, + "learning_rate": 2.988137984098487e-05, + "loss": 0.4568, + "step": 31427 + }, + { + "epoch": 1.21, + "learning_rate": 2.988073865093614e-05, + "loss": 0.3531, + "step": 31428 + }, + { + "epoch": 1.21, + "learning_rate": 2.988009746088741e-05, + "loss": 0.4044, + "step": 31429 + }, + { + "epoch": 1.21, + "learning_rate": 2.987945627083868e-05, + "loss": 0.2956, + "step": 31430 + }, + { + "epoch": 1.21, + "learning_rate": 2.9878815080789945e-05, + "loss": 0.3336, + "step": 31431 + }, + { + "epoch": 1.21, + "learning_rate": 2.9878173890741217e-05, + "loss": 0.2792, + "step": 31432 + }, + { + "epoch": 1.21, + "learning_rate": 2.9877532700692485e-05, + "loss": 0.4321, + "step": 31433 + }, + { + "epoch": 1.21, + "learning_rate": 2.9876891510643757e-05, + "loss": 0.278, + "step": 31434 + }, + { + "epoch": 1.21, + "learning_rate": 2.987625032059503e-05, + "loss": 0.3517, + "step": 31435 + }, + { + "epoch": 1.21, + "learning_rate": 2.9875609130546294e-05, + "loss": 0.2777, + "step": 31436 + }, + { + "epoch": 1.21, + "learning_rate": 2.9874967940497565e-05, + "loss": 0.5834, + "step": 31437 + }, + { + "epoch": 1.21, + "learning_rate": 2.9874326750448834e-05, + "loss": 0.223, + "step": 31438 + }, + { + "epoch": 1.21, + "learning_rate": 2.9873685560400105e-05, + "loss": 0.3393, + "step": 31439 + }, + { + "epoch": 1.21, + "learning_rate": 2.987304437035137e-05, + "loss": 0.5527, + "step": 31440 + }, + { + "epoch": 1.21, + "learning_rate": 2.9872403180302645e-05, + "loss": 0.4308, + "step": 31441 + }, + { + "epoch": 1.21, + "learning_rate": 2.987176199025391e-05, + "loss": 0.3474, + "step": 31442 + }, + { + "epoch": 1.21, + "learning_rate": 2.9871120800205182e-05, + "loss": 0.3468, + "step": 31443 + }, + { + "epoch": 1.21, + "learning_rate": 2.9870479610156454e-05, + "loss": 0.2878, + "step": 31444 + }, + { + "epoch": 1.21, + "learning_rate": 2.9869838420107722e-05, + "loss": 0.4356, + "step": 31445 + }, + { + "epoch": 1.21, + "learning_rate": 2.9869197230058994e-05, + "loss": 0.4212, + "step": 31446 + }, + { + "epoch": 1.21, + "learning_rate": 2.986855604001026e-05, + "loss": 0.5208, + "step": 31447 + }, + { + "epoch": 1.21, + "learning_rate": 2.986791484996153e-05, + "loss": 0.3949, + "step": 31448 + }, + { + "epoch": 1.21, + "learning_rate": 2.98672736599128e-05, + "loss": 0.3639, + "step": 31449 + }, + { + "epoch": 1.21, + "learning_rate": 2.986663246986407e-05, + "loss": 0.4337, + "step": 31450 + }, + { + "epoch": 1.21, + "learning_rate": 2.9865991279815336e-05, + "loss": 0.4295, + "step": 31451 + }, + { + "epoch": 1.21, + "learning_rate": 2.9865350089766607e-05, + "loss": 0.4395, + "step": 31452 + }, + { + "epoch": 1.21, + "learning_rate": 2.986470889971788e-05, + "loss": 0.4546, + "step": 31453 + }, + { + "epoch": 1.21, + "learning_rate": 2.9864067709669147e-05, + "loss": 0.3369, + "step": 31454 + }, + { + "epoch": 1.21, + "learning_rate": 2.986342651962042e-05, + "loss": 0.3599, + "step": 31455 + }, + { + "epoch": 1.21, + "learning_rate": 2.9862785329571684e-05, + "loss": 0.3922, + "step": 31456 + }, + { + "epoch": 1.21, + "learning_rate": 2.9862144139522956e-05, + "loss": 0.3511, + "step": 31457 + }, + { + "epoch": 1.21, + "learning_rate": 2.9861502949474224e-05, + "loss": 0.415, + "step": 31458 + }, + { + "epoch": 1.21, + "learning_rate": 2.9860861759425496e-05, + "loss": 0.3276, + "step": 31459 + }, + { + "epoch": 1.21, + "learning_rate": 2.986022056937676e-05, + "loss": 0.3105, + "step": 31460 + }, + { + "epoch": 1.21, + "learning_rate": 2.9859579379328036e-05, + "loss": 0.3882, + "step": 31461 + }, + { + "epoch": 1.21, + "learning_rate": 2.9858938189279307e-05, + "loss": 0.3613, + "step": 31462 + }, + { + "epoch": 1.21, + "learning_rate": 2.9858296999230572e-05, + "loss": 0.4789, + "step": 31463 + }, + { + "epoch": 1.21, + "learning_rate": 2.9857655809181844e-05, + "loss": 0.5241, + "step": 31464 + }, + { + "epoch": 1.21, + "learning_rate": 2.9857014619133112e-05, + "loss": 0.3219, + "step": 31465 + }, + { + "epoch": 1.21, + "learning_rate": 2.9856373429084384e-05, + "loss": 0.501, + "step": 31466 + }, + { + "epoch": 1.21, + "learning_rate": 2.985573223903565e-05, + "loss": 0.5617, + "step": 31467 + }, + { + "epoch": 1.21, + "learning_rate": 2.985509104898692e-05, + "loss": 0.5668, + "step": 31468 + }, + { + "epoch": 1.21, + "learning_rate": 2.985444985893819e-05, + "loss": 0.3149, + "step": 31469 + }, + { + "epoch": 1.21, + "learning_rate": 2.985380866888946e-05, + "loss": 0.3409, + "step": 31470 + }, + { + "epoch": 1.21, + "learning_rate": 2.9853167478840733e-05, + "loss": 0.462, + "step": 31471 + }, + { + "epoch": 1.21, + "learning_rate": 2.9852526288791998e-05, + "loss": 0.4172, + "step": 31472 + }, + { + "epoch": 1.21, + "learning_rate": 2.985188509874327e-05, + "loss": 0.489, + "step": 31473 + }, + { + "epoch": 1.21, + "learning_rate": 2.9851243908694538e-05, + "loss": 0.2979, + "step": 31474 + }, + { + "epoch": 1.21, + "learning_rate": 2.985060271864581e-05, + "loss": 0.3745, + "step": 31475 + }, + { + "epoch": 1.21, + "learning_rate": 2.9849961528597074e-05, + "loss": 0.3208, + "step": 31476 + }, + { + "epoch": 1.21, + "learning_rate": 2.984932033854835e-05, + "loss": 0.4987, + "step": 31477 + }, + { + "epoch": 1.21, + "learning_rate": 2.9848679148499614e-05, + "loss": 0.6162, + "step": 31478 + }, + { + "epoch": 1.21, + "learning_rate": 2.9848037958450886e-05, + "loss": 0.3192, + "step": 31479 + }, + { + "epoch": 1.21, + "learning_rate": 2.9847396768402158e-05, + "loss": 0.3806, + "step": 31480 + }, + { + "epoch": 1.21, + "learning_rate": 2.9846755578353426e-05, + "loss": 0.4375, + "step": 31481 + }, + { + "epoch": 1.21, + "learning_rate": 2.9846114388304698e-05, + "loss": 0.4626, + "step": 31482 + }, + { + "epoch": 1.21, + "learning_rate": 2.9845473198255963e-05, + "loss": 0.4202, + "step": 31483 + }, + { + "epoch": 1.21, + "learning_rate": 2.9844832008207234e-05, + "loss": 0.3744, + "step": 31484 + }, + { + "epoch": 1.21, + "learning_rate": 2.9844190818158503e-05, + "loss": 0.3647, + "step": 31485 + }, + { + "epoch": 1.21, + "learning_rate": 2.9843549628109774e-05, + "loss": 0.2831, + "step": 31486 + }, + { + "epoch": 1.21, + "learning_rate": 2.984290843806104e-05, + "loss": 0.2254, + "step": 31487 + }, + { + "epoch": 1.21, + "learning_rate": 2.984226724801231e-05, + "loss": 0.4186, + "step": 31488 + }, + { + "epoch": 1.21, + "learning_rate": 2.9841626057963583e-05, + "loss": 0.316, + "step": 31489 + }, + { + "epoch": 1.21, + "learning_rate": 2.984098486791485e-05, + "loss": 0.3486, + "step": 31490 + }, + { + "epoch": 1.21, + "learning_rate": 2.9840343677866123e-05, + "loss": 0.3683, + "step": 31491 + }, + { + "epoch": 1.21, + "learning_rate": 2.9839702487817388e-05, + "loss": 0.4413, + "step": 31492 + }, + { + "epoch": 1.21, + "learning_rate": 2.983906129776866e-05, + "loss": 0.3547, + "step": 31493 + }, + { + "epoch": 1.21, + "learning_rate": 2.9838420107719928e-05, + "loss": 0.5733, + "step": 31494 + }, + { + "epoch": 1.21, + "learning_rate": 2.98377789176712e-05, + "loss": 0.4871, + "step": 31495 + }, + { + "epoch": 1.21, + "learning_rate": 2.9837137727622465e-05, + "loss": 0.5931, + "step": 31496 + }, + { + "epoch": 1.21, + "learning_rate": 2.983649653757374e-05, + "loss": 0.4636, + "step": 31497 + }, + { + "epoch": 1.21, + "learning_rate": 2.983585534752501e-05, + "loss": 0.5133, + "step": 31498 + }, + { + "epoch": 1.21, + "learning_rate": 2.9835214157476276e-05, + "loss": 0.3272, + "step": 31499 + }, + { + "epoch": 1.21, + "learning_rate": 2.9834572967427548e-05, + "loss": 0.5051, + "step": 31500 + }, + { + "epoch": 1.21, + "learning_rate": 2.9833931777378816e-05, + "loss": 0.4539, + "step": 31501 + }, + { + "epoch": 1.21, + "learning_rate": 2.9833290587330088e-05, + "loss": 0.4477, + "step": 31502 + }, + { + "epoch": 1.21, + "learning_rate": 2.9832649397281353e-05, + "loss": 0.4312, + "step": 31503 + }, + { + "epoch": 1.21, + "learning_rate": 2.9832008207232625e-05, + "loss": 0.4109, + "step": 31504 + }, + { + "epoch": 1.21, + "learning_rate": 2.9831367017183893e-05, + "loss": 0.4537, + "step": 31505 + }, + { + "epoch": 1.21, + "learning_rate": 2.9830725827135165e-05, + "loss": 0.4308, + "step": 31506 + }, + { + "epoch": 1.21, + "learning_rate": 2.9830084637086436e-05, + "loss": 0.2917, + "step": 31507 + }, + { + "epoch": 1.21, + "learning_rate": 2.98294434470377e-05, + "loss": 0.4117, + "step": 31508 + }, + { + "epoch": 1.21, + "learning_rate": 2.9828802256988973e-05, + "loss": 0.3746, + "step": 31509 + }, + { + "epoch": 1.21, + "learning_rate": 2.982816106694024e-05, + "loss": 0.3412, + "step": 31510 + }, + { + "epoch": 1.21, + "learning_rate": 2.9827519876891513e-05, + "loss": 0.3766, + "step": 31511 + }, + { + "epoch": 1.21, + "learning_rate": 2.9826878686842778e-05, + "loss": 0.4698, + "step": 31512 + }, + { + "epoch": 1.21, + "learning_rate": 2.9826237496794053e-05, + "loss": 0.3098, + "step": 31513 + }, + { + "epoch": 1.21, + "learning_rate": 2.9825596306745318e-05, + "loss": 0.4474, + "step": 31514 + }, + { + "epoch": 1.21, + "learning_rate": 2.982495511669659e-05, + "loss": 0.3129, + "step": 31515 + }, + { + "epoch": 1.21, + "learning_rate": 2.982431392664786e-05, + "loss": 0.4351, + "step": 31516 + }, + { + "epoch": 1.21, + "learning_rate": 2.982367273659913e-05, + "loss": 0.3322, + "step": 31517 + }, + { + "epoch": 1.21, + "learning_rate": 2.98230315465504e-05, + "loss": 0.4161, + "step": 31518 + }, + { + "epoch": 1.21, + "learning_rate": 2.9822390356501667e-05, + "loss": 0.5283, + "step": 31519 + }, + { + "epoch": 1.21, + "learning_rate": 2.9821749166452938e-05, + "loss": 0.3984, + "step": 31520 + }, + { + "epoch": 1.21, + "learning_rate": 2.9821107976404207e-05, + "loss": 0.4837, + "step": 31521 + }, + { + "epoch": 1.21, + "learning_rate": 2.982046678635548e-05, + "loss": 0.3707, + "step": 31522 + }, + { + "epoch": 1.21, + "learning_rate": 2.9819825596306743e-05, + "loss": 0.3214, + "step": 31523 + }, + { + "epoch": 1.21, + "learning_rate": 2.9819184406258015e-05, + "loss": 0.4464, + "step": 31524 + }, + { + "epoch": 1.21, + "learning_rate": 2.9818543216209287e-05, + "loss": 0.4077, + "step": 31525 + }, + { + "epoch": 1.21, + "learning_rate": 2.9817902026160555e-05, + "loss": 0.3997, + "step": 31526 + }, + { + "epoch": 1.21, + "learning_rate": 2.9817260836111827e-05, + "loss": 0.3219, + "step": 31527 + }, + { + "epoch": 1.21, + "learning_rate": 2.981661964606309e-05, + "loss": 0.35, + "step": 31528 + }, + { + "epoch": 1.21, + "learning_rate": 2.9815978456014367e-05, + "loss": 0.5375, + "step": 31529 + }, + { + "epoch": 1.21, + "learning_rate": 2.9815337265965632e-05, + "loss": 0.3615, + "step": 31530 + }, + { + "epoch": 1.21, + "learning_rate": 2.9814696075916903e-05, + "loss": 0.409, + "step": 31531 + }, + { + "epoch": 1.21, + "learning_rate": 2.9814054885868172e-05, + "loss": 0.494, + "step": 31532 + }, + { + "epoch": 1.21, + "learning_rate": 2.9813413695819443e-05, + "loss": 0.3728, + "step": 31533 + }, + { + "epoch": 1.21, + "learning_rate": 2.9812772505770715e-05, + "loss": 0.4671, + "step": 31534 + }, + { + "epoch": 1.21, + "learning_rate": 2.981213131572198e-05, + "loss": 0.4211, + "step": 31535 + }, + { + "epoch": 1.21, + "learning_rate": 2.9811490125673252e-05, + "loss": 0.3669, + "step": 31536 + }, + { + "epoch": 1.21, + "learning_rate": 2.981084893562452e-05, + "loss": 0.4068, + "step": 31537 + }, + { + "epoch": 1.21, + "learning_rate": 2.9810207745575792e-05, + "loss": 0.4307, + "step": 31538 + }, + { + "epoch": 1.21, + "learning_rate": 2.9809566555527057e-05, + "loss": 0.4487, + "step": 31539 + }, + { + "epoch": 1.21, + "learning_rate": 2.980892536547833e-05, + "loss": 0.3898, + "step": 31540 + }, + { + "epoch": 1.21, + "learning_rate": 2.9808284175429597e-05, + "loss": 0.4009, + "step": 31541 + }, + { + "epoch": 1.21, + "learning_rate": 2.980764298538087e-05, + "loss": 0.3674, + "step": 31542 + }, + { + "epoch": 1.21, + "learning_rate": 2.980700179533214e-05, + "loss": 0.4193, + "step": 31543 + }, + { + "epoch": 1.21, + "learning_rate": 2.9806360605283405e-05, + "loss": 0.5344, + "step": 31544 + }, + { + "epoch": 1.21, + "learning_rate": 2.9805719415234677e-05, + "loss": 0.2955, + "step": 31545 + }, + { + "epoch": 1.21, + "learning_rate": 2.9805078225185945e-05, + "loss": 0.4195, + "step": 31546 + }, + { + "epoch": 1.21, + "learning_rate": 2.9804437035137217e-05, + "loss": 0.4411, + "step": 31547 + }, + { + "epoch": 1.21, + "learning_rate": 2.9803795845088482e-05, + "loss": 0.4428, + "step": 31548 + }, + { + "epoch": 1.21, + "learning_rate": 2.9803154655039757e-05, + "loss": 0.4493, + "step": 31549 + }, + { + "epoch": 1.21, + "learning_rate": 2.9802513464991022e-05, + "loss": 0.5096, + "step": 31550 + }, + { + "epoch": 1.21, + "learning_rate": 2.9801872274942294e-05, + "loss": 0.2667, + "step": 31551 + }, + { + "epoch": 1.21, + "learning_rate": 2.9801231084893565e-05, + "loss": 0.3495, + "step": 31552 + }, + { + "epoch": 1.21, + "learning_rate": 2.9800589894844834e-05, + "loss": 0.497, + "step": 31553 + }, + { + "epoch": 1.21, + "learning_rate": 2.9799948704796105e-05, + "loss": 0.4885, + "step": 31554 + }, + { + "epoch": 1.21, + "learning_rate": 2.979930751474737e-05, + "loss": 0.4587, + "step": 31555 + }, + { + "epoch": 1.21, + "learning_rate": 2.9798666324698642e-05, + "loss": 0.404, + "step": 31556 + }, + { + "epoch": 1.21, + "learning_rate": 2.979802513464991e-05, + "loss": 0.3432, + "step": 31557 + }, + { + "epoch": 1.21, + "learning_rate": 2.9797383944601182e-05, + "loss": 0.4015, + "step": 31558 + }, + { + "epoch": 1.21, + "learning_rate": 2.9796742754552447e-05, + "loss": 0.3602, + "step": 31559 + }, + { + "epoch": 1.21, + "learning_rate": 2.979610156450372e-05, + "loss": 0.3358, + "step": 31560 + }, + { + "epoch": 1.21, + "learning_rate": 2.979546037445499e-05, + "loss": 0.395, + "step": 31561 + }, + { + "epoch": 1.21, + "learning_rate": 2.979481918440626e-05, + "loss": 0.4761, + "step": 31562 + }, + { + "epoch": 1.21, + "learning_rate": 2.979417799435753e-05, + "loss": 0.393, + "step": 31563 + }, + { + "epoch": 1.21, + "learning_rate": 2.9793536804308796e-05, + "loss": 0.2773, + "step": 31564 + }, + { + "epoch": 1.21, + "learning_rate": 2.979289561426007e-05, + "loss": 0.3179, + "step": 31565 + }, + { + "epoch": 1.21, + "learning_rate": 2.9792254424211336e-05, + "loss": 0.4543, + "step": 31566 + }, + { + "epoch": 1.21, + "learning_rate": 2.9791613234162607e-05, + "loss": 0.3684, + "step": 31567 + }, + { + "epoch": 1.21, + "learning_rate": 2.979097204411388e-05, + "loss": 0.349, + "step": 31568 + }, + { + "epoch": 1.21, + "learning_rate": 2.9790330854065147e-05, + "loss": 0.3301, + "step": 31569 + }, + { + "epoch": 1.21, + "learning_rate": 2.978968966401642e-05, + "loss": 0.2746, + "step": 31570 + }, + { + "epoch": 1.21, + "learning_rate": 2.9789048473967684e-05, + "loss": 0.3416, + "step": 31571 + }, + { + "epoch": 1.21, + "learning_rate": 2.9788407283918956e-05, + "loss": 0.4832, + "step": 31572 + }, + { + "epoch": 1.21, + "learning_rate": 2.9787766093870224e-05, + "loss": 0.3188, + "step": 31573 + }, + { + "epoch": 1.21, + "learning_rate": 2.9787124903821496e-05, + "loss": 0.3597, + "step": 31574 + }, + { + "epoch": 1.21, + "learning_rate": 2.978648371377276e-05, + "loss": 0.3984, + "step": 31575 + }, + { + "epoch": 1.21, + "learning_rate": 2.9785842523724032e-05, + "loss": 0.3035, + "step": 31576 + }, + { + "epoch": 1.21, + "learning_rate": 2.9785201333675304e-05, + "loss": 0.3153, + "step": 31577 + }, + { + "epoch": 1.21, + "learning_rate": 2.9784560143626572e-05, + "loss": 0.3769, + "step": 31578 + }, + { + "epoch": 1.21, + "learning_rate": 2.9783918953577844e-05, + "loss": 0.4413, + "step": 31579 + }, + { + "epoch": 1.21, + "learning_rate": 2.978327776352911e-05, + "loss": 0.3916, + "step": 31580 + }, + { + "epoch": 1.21, + "learning_rate": 2.978263657348038e-05, + "loss": 0.4477, + "step": 31581 + }, + { + "epoch": 1.21, + "learning_rate": 2.978199538343165e-05, + "loss": 0.4942, + "step": 31582 + }, + { + "epoch": 1.21, + "learning_rate": 2.978135419338292e-05, + "loss": 0.3759, + "step": 31583 + }, + { + "epoch": 1.21, + "learning_rate": 2.9780713003334186e-05, + "loss": 0.3654, + "step": 31584 + }, + { + "epoch": 1.21, + "learning_rate": 2.978007181328546e-05, + "loss": 0.3818, + "step": 31585 + }, + { + "epoch": 1.21, + "learning_rate": 2.9779430623236733e-05, + "loss": 0.3474, + "step": 31586 + }, + { + "epoch": 1.21, + "learning_rate": 2.9778789433187998e-05, + "loss": 0.3184, + "step": 31587 + }, + { + "epoch": 1.21, + "learning_rate": 2.977814824313927e-05, + "loss": 0.3858, + "step": 31588 + }, + { + "epoch": 1.21, + "learning_rate": 2.9777507053090538e-05, + "loss": 0.3117, + "step": 31589 + }, + { + "epoch": 1.21, + "learning_rate": 2.977686586304181e-05, + "loss": 0.5664, + "step": 31590 + }, + { + "epoch": 1.21, + "learning_rate": 2.9776224672993074e-05, + "loss": 0.5335, + "step": 31591 + }, + { + "epoch": 1.21, + "learning_rate": 2.9775583482944346e-05, + "loss": 0.3425, + "step": 31592 + }, + { + "epoch": 1.21, + "learning_rate": 2.9774942292895614e-05, + "loss": 0.4348, + "step": 31593 + }, + { + "epoch": 1.21, + "learning_rate": 2.9774301102846886e-05, + "loss": 0.4802, + "step": 31594 + }, + { + "epoch": 1.21, + "learning_rate": 2.9773659912798158e-05, + "loss": 0.4543, + "step": 31595 + }, + { + "epoch": 1.21, + "learning_rate": 2.9773018722749423e-05, + "loss": 0.3494, + "step": 31596 + }, + { + "epoch": 1.21, + "learning_rate": 2.9772377532700694e-05, + "loss": 0.3398, + "step": 31597 + }, + { + "epoch": 1.21, + "learning_rate": 2.9771736342651963e-05, + "loss": 0.3002, + "step": 31598 + }, + { + "epoch": 1.21, + "learning_rate": 2.9771095152603234e-05, + "loss": 0.5449, + "step": 31599 + }, + { + "epoch": 1.21, + "learning_rate": 2.97704539625545e-05, + "loss": 0.4832, + "step": 31600 + }, + { + "epoch": 1.21, + "learning_rate": 2.9769812772505775e-05, + "loss": 0.3704, + "step": 31601 + }, + { + "epoch": 1.21, + "learning_rate": 2.976917158245704e-05, + "loss": 0.3895, + "step": 31602 + }, + { + "epoch": 1.22, + "learning_rate": 2.976853039240831e-05, + "loss": 0.4013, + "step": 31603 + }, + { + "epoch": 1.22, + "learning_rate": 2.9767889202359583e-05, + "loss": 0.4703, + "step": 31604 + }, + { + "epoch": 1.22, + "learning_rate": 2.976724801231085e-05, + "loss": 0.2511, + "step": 31605 + }, + { + "epoch": 1.22, + "learning_rate": 2.9766606822262123e-05, + "loss": 0.3531, + "step": 31606 + }, + { + "epoch": 1.22, + "learning_rate": 2.9765965632213388e-05, + "loss": 0.4624, + "step": 31607 + }, + { + "epoch": 1.22, + "learning_rate": 2.976532444216466e-05, + "loss": 0.3393, + "step": 31608 + }, + { + "epoch": 1.22, + "learning_rate": 2.9764683252115928e-05, + "loss": 0.2807, + "step": 31609 + }, + { + "epoch": 1.22, + "learning_rate": 2.97640420620672e-05, + "loss": 0.3666, + "step": 31610 + }, + { + "epoch": 1.22, + "learning_rate": 2.9763400872018465e-05, + "loss": 0.4105, + "step": 31611 + }, + { + "epoch": 1.22, + "learning_rate": 2.9762759681969736e-05, + "loss": 0.3173, + "step": 31612 + }, + { + "epoch": 1.22, + "learning_rate": 2.9762118491921008e-05, + "loss": 0.4982, + "step": 31613 + }, + { + "epoch": 1.22, + "learning_rate": 2.9761477301872276e-05, + "loss": 0.4144, + "step": 31614 + }, + { + "epoch": 1.22, + "learning_rate": 2.9760836111823548e-05, + "loss": 0.4329, + "step": 31615 + }, + { + "epoch": 1.22, + "learning_rate": 2.9760194921774813e-05, + "loss": 0.4987, + "step": 31616 + }, + { + "epoch": 1.22, + "learning_rate": 2.9759553731726085e-05, + "loss": 0.4619, + "step": 31617 + }, + { + "epoch": 1.22, + "learning_rate": 2.9758912541677353e-05, + "loss": 0.4681, + "step": 31618 + }, + { + "epoch": 1.22, + "learning_rate": 2.9758271351628625e-05, + "loss": 0.3463, + "step": 31619 + }, + { + "epoch": 1.22, + "learning_rate": 2.975763016157989e-05, + "loss": 0.3187, + "step": 31620 + }, + { + "epoch": 1.22, + "learning_rate": 2.9756988971531165e-05, + "loss": 0.3003, + "step": 31621 + }, + { + "epoch": 1.22, + "learning_rate": 2.9756347781482437e-05, + "loss": 0.4101, + "step": 31622 + }, + { + "epoch": 1.22, + "learning_rate": 2.97557065914337e-05, + "loss": 0.3486, + "step": 31623 + }, + { + "epoch": 1.22, + "learning_rate": 2.9755065401384973e-05, + "loss": 0.4692, + "step": 31624 + }, + { + "epoch": 1.22, + "learning_rate": 2.975442421133624e-05, + "loss": 0.4928, + "step": 31625 + }, + { + "epoch": 1.22, + "learning_rate": 2.9753783021287513e-05, + "loss": 0.3344, + "step": 31626 + }, + { + "epoch": 1.22, + "learning_rate": 2.9753141831238778e-05, + "loss": 0.3637, + "step": 31627 + }, + { + "epoch": 1.22, + "learning_rate": 2.975250064119005e-05, + "loss": 0.3848, + "step": 31628 + }, + { + "epoch": 1.22, + "learning_rate": 2.9751859451141318e-05, + "loss": 0.4983, + "step": 31629 + }, + { + "epoch": 1.22, + "learning_rate": 2.975121826109259e-05, + "loss": 0.3427, + "step": 31630 + }, + { + "epoch": 1.22, + "learning_rate": 2.975057707104386e-05, + "loss": 0.4545, + "step": 31631 + }, + { + "epoch": 1.22, + "learning_rate": 2.9749935880995127e-05, + "loss": 0.3705, + "step": 31632 + }, + { + "epoch": 1.22, + "learning_rate": 2.97492946909464e-05, + "loss": 0.3154, + "step": 31633 + }, + { + "epoch": 1.22, + "learning_rate": 2.9748653500897667e-05, + "loss": 0.5027, + "step": 31634 + }, + { + "epoch": 1.22, + "learning_rate": 2.974801231084894e-05, + "loss": 0.3401, + "step": 31635 + }, + { + "epoch": 1.22, + "learning_rate": 2.9747371120800203e-05, + "loss": 0.485, + "step": 31636 + }, + { + "epoch": 1.22, + "learning_rate": 2.974672993075148e-05, + "loss": 0.3322, + "step": 31637 + }, + { + "epoch": 1.22, + "learning_rate": 2.9746088740702743e-05, + "loss": 0.3654, + "step": 31638 + }, + { + "epoch": 1.22, + "learning_rate": 2.9745447550654015e-05, + "loss": 0.3852, + "step": 31639 + }, + { + "epoch": 1.22, + "learning_rate": 2.9744806360605287e-05, + "loss": 0.3263, + "step": 31640 + }, + { + "epoch": 1.22, + "learning_rate": 2.9744165170556555e-05, + "loss": 0.515, + "step": 31641 + }, + { + "epoch": 1.22, + "learning_rate": 2.9743523980507827e-05, + "loss": 0.3368, + "step": 31642 + }, + { + "epoch": 1.22, + "learning_rate": 2.9742882790459092e-05, + "loss": 0.4315, + "step": 31643 + }, + { + "epoch": 1.22, + "learning_rate": 2.9742241600410363e-05, + "loss": 0.3736, + "step": 31644 + }, + { + "epoch": 1.22, + "learning_rate": 2.9741600410361632e-05, + "loss": 0.3539, + "step": 31645 + }, + { + "epoch": 1.22, + "learning_rate": 2.9740959220312904e-05, + "loss": 0.5026, + "step": 31646 + }, + { + "epoch": 1.22, + "learning_rate": 2.974031803026417e-05, + "loss": 0.3453, + "step": 31647 + }, + { + "epoch": 1.22, + "learning_rate": 2.973967684021544e-05, + "loss": 0.4797, + "step": 31648 + }, + { + "epoch": 1.22, + "learning_rate": 2.9739035650166712e-05, + "loss": 0.3767, + "step": 31649 + }, + { + "epoch": 1.22, + "learning_rate": 2.973839446011798e-05, + "loss": 0.3886, + "step": 31650 + }, + { + "epoch": 1.22, + "learning_rate": 2.9737753270069252e-05, + "loss": 0.3406, + "step": 31651 + }, + { + "epoch": 1.22, + "learning_rate": 2.9737112080020517e-05, + "loss": 0.4526, + "step": 31652 + }, + { + "epoch": 1.22, + "learning_rate": 2.9736470889971792e-05, + "loss": 0.415, + "step": 31653 + }, + { + "epoch": 1.22, + "learning_rate": 2.9735829699923057e-05, + "loss": 0.4009, + "step": 31654 + }, + { + "epoch": 1.22, + "learning_rate": 2.973518850987433e-05, + "loss": 0.3429, + "step": 31655 + }, + { + "epoch": 1.22, + "learning_rate": 2.9734547319825594e-05, + "loss": 0.3305, + "step": 31656 + }, + { + "epoch": 1.22, + "learning_rate": 2.973390612977687e-05, + "loss": 0.4616, + "step": 31657 + }, + { + "epoch": 1.22, + "learning_rate": 2.973326493972814e-05, + "loss": 0.5188, + "step": 31658 + }, + { + "epoch": 1.22, + "learning_rate": 2.9732623749679405e-05, + "loss": 0.2479, + "step": 31659 + }, + { + "epoch": 1.22, + "learning_rate": 2.9731982559630677e-05, + "loss": 0.24, + "step": 31660 + }, + { + "epoch": 1.22, + "learning_rate": 2.9731341369581945e-05, + "loss": 0.426, + "step": 31661 + }, + { + "epoch": 1.22, + "learning_rate": 2.9730700179533217e-05, + "loss": 0.3323, + "step": 31662 + }, + { + "epoch": 1.22, + "learning_rate": 2.9730058989484482e-05, + "loss": 0.3268, + "step": 31663 + }, + { + "epoch": 1.22, + "learning_rate": 2.9729417799435754e-05, + "loss": 0.3377, + "step": 31664 + }, + { + "epoch": 1.22, + "learning_rate": 2.9728776609387022e-05, + "loss": 0.4304, + "step": 31665 + }, + { + "epoch": 1.22, + "learning_rate": 2.9728135419338294e-05, + "loss": 0.4135, + "step": 31666 + }, + { + "epoch": 1.22, + "learning_rate": 2.9727494229289566e-05, + "loss": 0.3628, + "step": 31667 + }, + { + "epoch": 1.22, + "learning_rate": 2.972685303924083e-05, + "loss": 0.3427, + "step": 31668 + }, + { + "epoch": 1.22, + "learning_rate": 2.9726211849192102e-05, + "loss": 0.4201, + "step": 31669 + }, + { + "epoch": 1.22, + "learning_rate": 2.972557065914337e-05, + "loss": 0.3834, + "step": 31670 + }, + { + "epoch": 1.22, + "learning_rate": 2.9724929469094642e-05, + "loss": 0.3741, + "step": 31671 + }, + { + "epoch": 1.22, + "learning_rate": 2.9724288279045907e-05, + "loss": 0.3574, + "step": 31672 + }, + { + "epoch": 1.22, + "learning_rate": 2.9723647088997182e-05, + "loss": 0.3714, + "step": 31673 + }, + { + "epoch": 1.22, + "learning_rate": 2.9723005898948447e-05, + "loss": 0.4271, + "step": 31674 + }, + { + "epoch": 1.22, + "learning_rate": 2.972236470889972e-05, + "loss": 0.3174, + "step": 31675 + }, + { + "epoch": 1.22, + "learning_rate": 2.972172351885099e-05, + "loss": 0.4203, + "step": 31676 + }, + { + "epoch": 1.22, + "learning_rate": 2.972108232880226e-05, + "loss": 0.3105, + "step": 31677 + }, + { + "epoch": 1.22, + "learning_rate": 2.972044113875353e-05, + "loss": 0.3564, + "step": 31678 + }, + { + "epoch": 1.22, + "learning_rate": 2.9719799948704796e-05, + "loss": 0.406, + "step": 31679 + }, + { + "epoch": 1.22, + "learning_rate": 2.9719158758656067e-05, + "loss": 0.4287, + "step": 31680 + }, + { + "epoch": 1.22, + "learning_rate": 2.9718517568607336e-05, + "loss": 0.2896, + "step": 31681 + }, + { + "epoch": 1.22, + "learning_rate": 2.9717876378558607e-05, + "loss": 0.4836, + "step": 31682 + }, + { + "epoch": 1.22, + "learning_rate": 2.9717235188509872e-05, + "loss": 0.4483, + "step": 31683 + }, + { + "epoch": 1.22, + "learning_rate": 2.9716593998461144e-05, + "loss": 0.4401, + "step": 31684 + }, + { + "epoch": 1.22, + "learning_rate": 2.9715952808412416e-05, + "loss": 0.3355, + "step": 31685 + }, + { + "epoch": 1.22, + "learning_rate": 2.9715311618363684e-05, + "loss": 0.3237, + "step": 31686 + }, + { + "epoch": 1.22, + "learning_rate": 2.9714670428314956e-05, + "loss": 0.3225, + "step": 31687 + }, + { + "epoch": 1.22, + "learning_rate": 2.971402923826622e-05, + "loss": 0.513, + "step": 31688 + }, + { + "epoch": 1.22, + "learning_rate": 2.9713388048217496e-05, + "loss": 0.3426, + "step": 31689 + }, + { + "epoch": 1.22, + "learning_rate": 2.971274685816876e-05, + "loss": 0.4376, + "step": 31690 + }, + { + "epoch": 1.22, + "learning_rate": 2.9712105668120033e-05, + "loss": 0.2941, + "step": 31691 + }, + { + "epoch": 1.22, + "learning_rate": 2.97114644780713e-05, + "loss": 0.4365, + "step": 31692 + }, + { + "epoch": 1.22, + "learning_rate": 2.9710823288022573e-05, + "loss": 0.3228, + "step": 31693 + }, + { + "epoch": 1.22, + "learning_rate": 2.9710182097973844e-05, + "loss": 0.2713, + "step": 31694 + }, + { + "epoch": 1.22, + "learning_rate": 2.970954090792511e-05, + "loss": 0.3549, + "step": 31695 + }, + { + "epoch": 1.22, + "learning_rate": 2.970889971787638e-05, + "loss": 0.4328, + "step": 31696 + }, + { + "epoch": 1.22, + "learning_rate": 2.970825852782765e-05, + "loss": 0.359, + "step": 31697 + }, + { + "epoch": 1.22, + "learning_rate": 2.970761733777892e-05, + "loss": 0.4783, + "step": 31698 + }, + { + "epoch": 1.22, + "learning_rate": 2.9706976147730186e-05, + "loss": 0.4746, + "step": 31699 + }, + { + "epoch": 1.22, + "learning_rate": 2.9706334957681458e-05, + "loss": 0.5726, + "step": 31700 + }, + { + "epoch": 1.22, + "learning_rate": 2.9705693767632726e-05, + "loss": 0.3825, + "step": 31701 + }, + { + "epoch": 1.22, + "learning_rate": 2.9705052577583998e-05, + "loss": 0.3665, + "step": 31702 + }, + { + "epoch": 1.22, + "learning_rate": 2.970441138753527e-05, + "loss": 0.3098, + "step": 31703 + }, + { + "epoch": 1.22, + "learning_rate": 2.9703770197486534e-05, + "loss": 0.4599, + "step": 31704 + }, + { + "epoch": 1.22, + "learning_rate": 2.9703129007437806e-05, + "loss": 0.5441, + "step": 31705 + }, + { + "epoch": 1.22, + "learning_rate": 2.9702487817389074e-05, + "loss": 0.3718, + "step": 31706 + }, + { + "epoch": 1.22, + "learning_rate": 2.9701846627340346e-05, + "loss": 0.3413, + "step": 31707 + }, + { + "epoch": 1.22, + "learning_rate": 2.970120543729161e-05, + "loss": 0.3934, + "step": 31708 + }, + { + "epoch": 1.22, + "learning_rate": 2.9700564247242886e-05, + "loss": 0.3699, + "step": 31709 + }, + { + "epoch": 1.22, + "learning_rate": 2.969992305719415e-05, + "loss": 0.5256, + "step": 31710 + }, + { + "epoch": 1.22, + "learning_rate": 2.9699281867145423e-05, + "loss": 0.5131, + "step": 31711 + }, + { + "epoch": 1.22, + "learning_rate": 2.9698640677096695e-05, + "loss": 0.4735, + "step": 31712 + }, + { + "epoch": 1.22, + "learning_rate": 2.9697999487047963e-05, + "loss": 0.3599, + "step": 31713 + }, + { + "epoch": 1.22, + "learning_rate": 2.9697358296999235e-05, + "loss": 0.424, + "step": 31714 + }, + { + "epoch": 1.22, + "learning_rate": 2.96967171069505e-05, + "loss": 0.4116, + "step": 31715 + }, + { + "epoch": 1.22, + "learning_rate": 2.969607591690177e-05, + "loss": 0.4879, + "step": 31716 + }, + { + "epoch": 1.22, + "learning_rate": 2.969543472685304e-05, + "loss": 0.3593, + "step": 31717 + }, + { + "epoch": 1.22, + "learning_rate": 2.969479353680431e-05, + "loss": 0.4249, + "step": 31718 + }, + { + "epoch": 1.22, + "learning_rate": 2.9694152346755576e-05, + "loss": 0.3978, + "step": 31719 + }, + { + "epoch": 1.22, + "learning_rate": 2.9693511156706848e-05, + "loss": 0.3945, + "step": 31720 + }, + { + "epoch": 1.22, + "learning_rate": 2.969286996665812e-05, + "loss": 0.3243, + "step": 31721 + }, + { + "epoch": 1.22, + "learning_rate": 2.9692228776609388e-05, + "loss": 0.3617, + "step": 31722 + }, + { + "epoch": 1.22, + "learning_rate": 2.969158758656066e-05, + "loss": 0.4115, + "step": 31723 + }, + { + "epoch": 1.22, + "learning_rate": 2.9690946396511925e-05, + "loss": 0.3542, + "step": 31724 + }, + { + "epoch": 1.22, + "learning_rate": 2.96903052064632e-05, + "loss": 0.4123, + "step": 31725 + }, + { + "epoch": 1.22, + "learning_rate": 2.9689664016414465e-05, + "loss": 0.3258, + "step": 31726 + }, + { + "epoch": 1.22, + "learning_rate": 2.9689022826365736e-05, + "loss": 0.4386, + "step": 31727 + }, + { + "epoch": 1.22, + "learning_rate": 2.9688381636317005e-05, + "loss": 0.3707, + "step": 31728 + }, + { + "epoch": 1.22, + "learning_rate": 2.9687740446268276e-05, + "loss": 0.3253, + "step": 31729 + }, + { + "epoch": 1.22, + "learning_rate": 2.9687099256219548e-05, + "loss": 0.3704, + "step": 31730 + }, + { + "epoch": 1.22, + "learning_rate": 2.9686458066170813e-05, + "loss": 0.3598, + "step": 31731 + }, + { + "epoch": 1.22, + "learning_rate": 2.9685816876122085e-05, + "loss": 0.4414, + "step": 31732 + }, + { + "epoch": 1.22, + "learning_rate": 2.9685175686073353e-05, + "loss": 0.3585, + "step": 31733 + }, + { + "epoch": 1.22, + "learning_rate": 2.9684534496024625e-05, + "loss": 0.3404, + "step": 31734 + }, + { + "epoch": 1.22, + "learning_rate": 2.968389330597589e-05, + "loss": 0.3521, + "step": 31735 + }, + { + "epoch": 1.22, + "learning_rate": 2.968325211592716e-05, + "loss": 0.4059, + "step": 31736 + }, + { + "epoch": 1.22, + "learning_rate": 2.968261092587843e-05, + "loss": 0.4344, + "step": 31737 + }, + { + "epoch": 1.22, + "learning_rate": 2.96819697358297e-05, + "loss": 0.3137, + "step": 31738 + }, + { + "epoch": 1.22, + "learning_rate": 2.9681328545780973e-05, + "loss": 0.4645, + "step": 31739 + }, + { + "epoch": 1.22, + "learning_rate": 2.9680687355732238e-05, + "loss": 0.4207, + "step": 31740 + }, + { + "epoch": 1.22, + "learning_rate": 2.968004616568351e-05, + "loss": 0.2778, + "step": 31741 + }, + { + "epoch": 1.22, + "learning_rate": 2.9679404975634778e-05, + "loss": 0.3882, + "step": 31742 + }, + { + "epoch": 1.22, + "learning_rate": 2.967876378558605e-05, + "loss": 0.3042, + "step": 31743 + }, + { + "epoch": 1.22, + "learning_rate": 2.9678122595537315e-05, + "loss": 0.4303, + "step": 31744 + }, + { + "epoch": 1.22, + "learning_rate": 2.967748140548859e-05, + "loss": 0.4561, + "step": 31745 + }, + { + "epoch": 1.22, + "learning_rate": 2.9676840215439855e-05, + "loss": 0.4264, + "step": 31746 + }, + { + "epoch": 1.22, + "learning_rate": 2.9676199025391127e-05, + "loss": 0.313, + "step": 31747 + }, + { + "epoch": 1.22, + "learning_rate": 2.96755578353424e-05, + "loss": 0.4175, + "step": 31748 + }, + { + "epoch": 1.22, + "learning_rate": 2.9674916645293667e-05, + "loss": 0.4709, + "step": 31749 + }, + { + "epoch": 1.22, + "learning_rate": 2.967427545524494e-05, + "loss": 0.4217, + "step": 31750 + }, + { + "epoch": 1.22, + "learning_rate": 2.9673634265196203e-05, + "loss": 0.5393, + "step": 31751 + }, + { + "epoch": 1.22, + "learning_rate": 2.9672993075147475e-05, + "loss": 0.51, + "step": 31752 + }, + { + "epoch": 1.22, + "learning_rate": 2.9672351885098743e-05, + "loss": 0.3829, + "step": 31753 + }, + { + "epoch": 1.22, + "learning_rate": 2.9671710695050015e-05, + "loss": 0.4534, + "step": 31754 + }, + { + "epoch": 1.22, + "learning_rate": 2.9671069505001287e-05, + "loss": 0.3976, + "step": 31755 + }, + { + "epoch": 1.22, + "learning_rate": 2.9670428314952552e-05, + "loss": 0.3801, + "step": 31756 + }, + { + "epoch": 1.22, + "learning_rate": 2.9669787124903824e-05, + "loss": 0.3396, + "step": 31757 + }, + { + "epoch": 1.22, + "learning_rate": 2.9669145934855092e-05, + "loss": 0.3997, + "step": 31758 + }, + { + "epoch": 1.22, + "learning_rate": 2.9668504744806364e-05, + "loss": 0.3699, + "step": 31759 + }, + { + "epoch": 1.22, + "learning_rate": 2.966786355475763e-05, + "loss": 0.3507, + "step": 31760 + }, + { + "epoch": 1.22, + "learning_rate": 2.9667222364708904e-05, + "loss": 0.3516, + "step": 31761 + }, + { + "epoch": 1.22, + "learning_rate": 2.966658117466017e-05, + "loss": 0.3734, + "step": 31762 + }, + { + "epoch": 1.22, + "learning_rate": 2.966593998461144e-05, + "loss": 0.4987, + "step": 31763 + }, + { + "epoch": 1.22, + "learning_rate": 2.9665298794562712e-05, + "loss": 0.3025, + "step": 31764 + }, + { + "epoch": 1.22, + "learning_rate": 2.966465760451398e-05, + "loss": 0.4989, + "step": 31765 + }, + { + "epoch": 1.22, + "learning_rate": 2.9664016414465252e-05, + "loss": 0.3438, + "step": 31766 + }, + { + "epoch": 1.22, + "learning_rate": 2.9663375224416517e-05, + "loss": 0.3461, + "step": 31767 + }, + { + "epoch": 1.22, + "learning_rate": 2.966273403436779e-05, + "loss": 0.3685, + "step": 31768 + }, + { + "epoch": 1.22, + "learning_rate": 2.9662092844319057e-05, + "loss": 0.3715, + "step": 31769 + }, + { + "epoch": 1.22, + "learning_rate": 2.966145165427033e-05, + "loss": 0.2117, + "step": 31770 + }, + { + "epoch": 1.22, + "learning_rate": 2.9660810464221594e-05, + "loss": 0.4135, + "step": 31771 + }, + { + "epoch": 1.22, + "learning_rate": 2.9660169274172865e-05, + "loss": 0.3446, + "step": 31772 + }, + { + "epoch": 1.22, + "learning_rate": 2.9659528084124137e-05, + "loss": 0.5879, + "step": 31773 + }, + { + "epoch": 1.22, + "learning_rate": 2.9658886894075405e-05, + "loss": 0.2503, + "step": 31774 + }, + { + "epoch": 1.22, + "learning_rate": 2.9658245704026677e-05, + "loss": 0.4018, + "step": 31775 + }, + { + "epoch": 1.22, + "learning_rate": 2.9657604513977942e-05, + "loss": 0.4291, + "step": 31776 + }, + { + "epoch": 1.22, + "learning_rate": 2.9656963323929217e-05, + "loss": 0.4241, + "step": 31777 + }, + { + "epoch": 1.22, + "learning_rate": 2.9656322133880482e-05, + "loss": 0.3524, + "step": 31778 + }, + { + "epoch": 1.22, + "learning_rate": 2.9655680943831754e-05, + "loss": 0.2966, + "step": 31779 + }, + { + "epoch": 1.22, + "learning_rate": 2.965503975378302e-05, + "loss": 0.5078, + "step": 31780 + }, + { + "epoch": 1.22, + "learning_rate": 2.9654398563734294e-05, + "loss": 0.2912, + "step": 31781 + }, + { + "epoch": 1.22, + "learning_rate": 2.9653757373685566e-05, + "loss": 0.3091, + "step": 31782 + }, + { + "epoch": 1.22, + "learning_rate": 2.965311618363683e-05, + "loss": 0.2657, + "step": 31783 + }, + { + "epoch": 1.22, + "learning_rate": 2.9652474993588102e-05, + "loss": 0.43, + "step": 31784 + }, + { + "epoch": 1.22, + "learning_rate": 2.965183380353937e-05, + "loss": 0.4195, + "step": 31785 + }, + { + "epoch": 1.22, + "learning_rate": 2.9651192613490642e-05, + "loss": 0.3899, + "step": 31786 + }, + { + "epoch": 1.22, + "learning_rate": 2.9650551423441907e-05, + "loss": 0.2792, + "step": 31787 + }, + { + "epoch": 1.22, + "learning_rate": 2.964991023339318e-05, + "loss": 0.4343, + "step": 31788 + }, + { + "epoch": 1.22, + "learning_rate": 2.9649269043344447e-05, + "loss": 0.3809, + "step": 31789 + }, + { + "epoch": 1.22, + "learning_rate": 2.964862785329572e-05, + "loss": 0.3264, + "step": 31790 + }, + { + "epoch": 1.22, + "learning_rate": 2.964798666324699e-05, + "loss": 0.3232, + "step": 31791 + }, + { + "epoch": 1.22, + "learning_rate": 2.9647345473198256e-05, + "loss": 0.4076, + "step": 31792 + }, + { + "epoch": 1.22, + "learning_rate": 2.9646704283149527e-05, + "loss": 0.3825, + "step": 31793 + }, + { + "epoch": 1.22, + "learning_rate": 2.9646063093100796e-05, + "loss": 0.291, + "step": 31794 + }, + { + "epoch": 1.22, + "learning_rate": 2.9645421903052067e-05, + "loss": 0.5332, + "step": 31795 + }, + { + "epoch": 1.22, + "learning_rate": 2.9644780713003332e-05, + "loss": 0.3993, + "step": 31796 + }, + { + "epoch": 1.22, + "learning_rate": 2.9644139522954607e-05, + "loss": 0.309, + "step": 31797 + }, + { + "epoch": 1.22, + "learning_rate": 2.9643498332905872e-05, + "loss": 0.3727, + "step": 31798 + }, + { + "epoch": 1.22, + "learning_rate": 2.9642857142857144e-05, + "loss": 0.3535, + "step": 31799 + }, + { + "epoch": 1.22, + "learning_rate": 2.9642215952808416e-05, + "loss": 0.3747, + "step": 31800 + }, + { + "epoch": 1.22, + "learning_rate": 2.9641574762759684e-05, + "loss": 0.4325, + "step": 31801 + }, + { + "epoch": 1.22, + "learning_rate": 2.9640933572710956e-05, + "loss": 0.36, + "step": 31802 + }, + { + "epoch": 1.22, + "learning_rate": 2.964029238266222e-05, + "loss": 0.3869, + "step": 31803 + }, + { + "epoch": 1.22, + "learning_rate": 2.9639651192613493e-05, + "loss": 0.3511, + "step": 31804 + }, + { + "epoch": 1.22, + "learning_rate": 2.963901000256476e-05, + "loss": 0.3949, + "step": 31805 + }, + { + "epoch": 1.22, + "learning_rate": 2.9638368812516033e-05, + "loss": 0.4857, + "step": 31806 + }, + { + "epoch": 1.22, + "learning_rate": 2.9637727622467298e-05, + "loss": 0.426, + "step": 31807 + }, + { + "epoch": 1.22, + "learning_rate": 2.963708643241857e-05, + "loss": 0.402, + "step": 31808 + }, + { + "epoch": 1.22, + "learning_rate": 2.963644524236984e-05, + "loss": 0.2957, + "step": 31809 + }, + { + "epoch": 1.22, + "learning_rate": 2.963580405232111e-05, + "loss": 0.3192, + "step": 31810 + }, + { + "epoch": 1.22, + "learning_rate": 2.963516286227238e-05, + "loss": 0.4494, + "step": 31811 + }, + { + "epoch": 1.22, + "learning_rate": 2.9634521672223646e-05, + "loss": 0.4403, + "step": 31812 + }, + { + "epoch": 1.22, + "learning_rate": 2.963388048217492e-05, + "loss": 0.4052, + "step": 31813 + }, + { + "epoch": 1.22, + "learning_rate": 2.9633239292126186e-05, + "loss": 0.4055, + "step": 31814 + }, + { + "epoch": 1.22, + "learning_rate": 2.9632598102077458e-05, + "loss": 0.4652, + "step": 31815 + }, + { + "epoch": 1.22, + "learning_rate": 2.9631956912028726e-05, + "loss": 0.4958, + "step": 31816 + }, + { + "epoch": 1.22, + "learning_rate": 2.9631315721979998e-05, + "loss": 0.3194, + "step": 31817 + }, + { + "epoch": 1.22, + "learning_rate": 2.963067453193127e-05, + "loss": 0.4121, + "step": 31818 + }, + { + "epoch": 1.22, + "learning_rate": 2.9630033341882534e-05, + "loss": 0.3655, + "step": 31819 + }, + { + "epoch": 1.22, + "learning_rate": 2.9629392151833806e-05, + "loss": 0.4543, + "step": 31820 + }, + { + "epoch": 1.22, + "learning_rate": 2.9628750961785074e-05, + "loss": 0.414, + "step": 31821 + }, + { + "epoch": 1.22, + "learning_rate": 2.9628109771736346e-05, + "loss": 0.357, + "step": 31822 + }, + { + "epoch": 1.22, + "learning_rate": 2.962746858168761e-05, + "loss": 0.4284, + "step": 31823 + }, + { + "epoch": 1.22, + "learning_rate": 2.9626827391638883e-05, + "loss": 0.4089, + "step": 31824 + }, + { + "epoch": 1.22, + "learning_rate": 2.962618620159015e-05, + "loss": 0.3018, + "step": 31825 + }, + { + "epoch": 1.22, + "learning_rate": 2.9625545011541423e-05, + "loss": 0.3418, + "step": 31826 + }, + { + "epoch": 1.22, + "learning_rate": 2.9624903821492695e-05, + "loss": 0.4912, + "step": 31827 + }, + { + "epoch": 1.22, + "learning_rate": 2.962426263144396e-05, + "loss": 0.4844, + "step": 31828 + }, + { + "epoch": 1.22, + "learning_rate": 2.962362144139523e-05, + "loss": 0.485, + "step": 31829 + }, + { + "epoch": 1.22, + "learning_rate": 2.96229802513465e-05, + "loss": 0.4246, + "step": 31830 + }, + { + "epoch": 1.22, + "learning_rate": 2.962233906129777e-05, + "loss": 0.3649, + "step": 31831 + }, + { + "epoch": 1.22, + "learning_rate": 2.9621697871249036e-05, + "loss": 0.3637, + "step": 31832 + }, + { + "epoch": 1.22, + "learning_rate": 2.962105668120031e-05, + "loss": 0.4087, + "step": 31833 + }, + { + "epoch": 1.22, + "learning_rate": 2.9620415491151576e-05, + "loss": 0.4374, + "step": 31834 + }, + { + "epoch": 1.22, + "learning_rate": 2.9619774301102848e-05, + "loss": 0.2924, + "step": 31835 + }, + { + "epoch": 1.22, + "learning_rate": 2.961913311105412e-05, + "loss": 0.4333, + "step": 31836 + }, + { + "epoch": 1.22, + "learning_rate": 2.9618491921005388e-05, + "loss": 0.4297, + "step": 31837 + }, + { + "epoch": 1.22, + "learning_rate": 2.961785073095666e-05, + "loss": 0.4681, + "step": 31838 + }, + { + "epoch": 1.22, + "learning_rate": 2.9617209540907925e-05, + "loss": 0.3369, + "step": 31839 + }, + { + "epoch": 1.22, + "learning_rate": 2.9616568350859196e-05, + "loss": 0.4363, + "step": 31840 + }, + { + "epoch": 1.22, + "learning_rate": 2.9615927160810465e-05, + "loss": 0.3538, + "step": 31841 + }, + { + "epoch": 1.22, + "learning_rate": 2.9615285970761736e-05, + "loss": 0.3171, + "step": 31842 + }, + { + "epoch": 1.22, + "learning_rate": 2.9614644780713e-05, + "loss": 0.4111, + "step": 31843 + }, + { + "epoch": 1.22, + "learning_rate": 2.9614003590664273e-05, + "loss": 0.3612, + "step": 31844 + }, + { + "epoch": 1.22, + "learning_rate": 2.9613362400615545e-05, + "loss": 0.4439, + "step": 31845 + }, + { + "epoch": 1.22, + "learning_rate": 2.9612721210566813e-05, + "loss": 0.4514, + "step": 31846 + }, + { + "epoch": 1.22, + "learning_rate": 2.9612080020518085e-05, + "loss": 0.409, + "step": 31847 + }, + { + "epoch": 1.22, + "learning_rate": 2.961143883046935e-05, + "loss": 0.4047, + "step": 31848 + }, + { + "epoch": 1.22, + "learning_rate": 2.9610797640420625e-05, + "loss": 0.3452, + "step": 31849 + }, + { + "epoch": 1.22, + "learning_rate": 2.961015645037189e-05, + "loss": 0.3382, + "step": 31850 + }, + { + "epoch": 1.22, + "learning_rate": 2.960951526032316e-05, + "loss": 0.4627, + "step": 31851 + }, + { + "epoch": 1.22, + "learning_rate": 2.960887407027443e-05, + "loss": 0.3874, + "step": 31852 + }, + { + "epoch": 1.22, + "learning_rate": 2.96082328802257e-05, + "loss": 0.4292, + "step": 31853 + }, + { + "epoch": 1.22, + "learning_rate": 2.9607591690176973e-05, + "loss": 0.3254, + "step": 31854 + }, + { + "epoch": 1.22, + "learning_rate": 2.9606950500128238e-05, + "loss": 0.3155, + "step": 31855 + }, + { + "epoch": 1.22, + "learning_rate": 2.960630931007951e-05, + "loss": 0.2706, + "step": 31856 + }, + { + "epoch": 1.22, + "learning_rate": 2.960566812003078e-05, + "loss": 0.4413, + "step": 31857 + }, + { + "epoch": 1.22, + "learning_rate": 2.960502692998205e-05, + "loss": 0.3034, + "step": 31858 + }, + { + "epoch": 1.22, + "learning_rate": 2.9604385739933315e-05, + "loss": 0.2947, + "step": 31859 + }, + { + "epoch": 1.22, + "learning_rate": 2.9603744549884587e-05, + "loss": 0.4816, + "step": 31860 + }, + { + "epoch": 1.22, + "learning_rate": 2.9603103359835855e-05, + "loss": 0.4664, + "step": 31861 + }, + { + "epoch": 1.22, + "learning_rate": 2.9602462169787127e-05, + "loss": 0.4889, + "step": 31862 + }, + { + "epoch": 1.23, + "learning_rate": 2.96018209797384e-05, + "loss": 0.442, + "step": 31863 + }, + { + "epoch": 1.23, + "learning_rate": 2.9601179789689663e-05, + "loss": 0.3716, + "step": 31864 + }, + { + "epoch": 1.23, + "learning_rate": 2.9600538599640935e-05, + "loss": 0.4348, + "step": 31865 + }, + { + "epoch": 1.23, + "learning_rate": 2.9599897409592203e-05, + "loss": 0.2862, + "step": 31866 + }, + { + "epoch": 1.23, + "learning_rate": 2.9599256219543475e-05, + "loss": 0.3089, + "step": 31867 + }, + { + "epoch": 1.23, + "learning_rate": 2.959861502949474e-05, + "loss": 0.3642, + "step": 31868 + }, + { + "epoch": 1.23, + "learning_rate": 2.9597973839446015e-05, + "loss": 0.3376, + "step": 31869 + }, + { + "epoch": 1.23, + "learning_rate": 2.959733264939728e-05, + "loss": 0.3559, + "step": 31870 + }, + { + "epoch": 1.23, + "learning_rate": 2.9596691459348552e-05, + "loss": 0.3407, + "step": 31871 + }, + { + "epoch": 1.23, + "learning_rate": 2.9596050269299824e-05, + "loss": 0.3479, + "step": 31872 + }, + { + "epoch": 1.23, + "learning_rate": 2.9595409079251092e-05, + "loss": 0.2794, + "step": 31873 + }, + { + "epoch": 1.23, + "learning_rate": 2.9594767889202364e-05, + "loss": 0.4199, + "step": 31874 + }, + { + "epoch": 1.23, + "learning_rate": 2.959412669915363e-05, + "loss": 0.4426, + "step": 31875 + }, + { + "epoch": 1.23, + "learning_rate": 2.95934855091049e-05, + "loss": 0.3449, + "step": 31876 + }, + { + "epoch": 1.23, + "learning_rate": 2.959284431905617e-05, + "loss": 0.3827, + "step": 31877 + }, + { + "epoch": 1.23, + "learning_rate": 2.959220312900744e-05, + "loss": 0.3466, + "step": 31878 + }, + { + "epoch": 1.23, + "learning_rate": 2.9591561938958705e-05, + "loss": 0.4782, + "step": 31879 + }, + { + "epoch": 1.23, + "learning_rate": 2.9590920748909977e-05, + "loss": 0.4225, + "step": 31880 + }, + { + "epoch": 1.23, + "learning_rate": 2.959027955886125e-05, + "loss": 0.5051, + "step": 31881 + }, + { + "epoch": 1.23, + "learning_rate": 2.9589638368812517e-05, + "loss": 0.4308, + "step": 31882 + }, + { + "epoch": 1.23, + "learning_rate": 2.958899717876379e-05, + "loss": 0.35, + "step": 31883 + }, + { + "epoch": 1.23, + "learning_rate": 2.9588355988715054e-05, + "loss": 0.3546, + "step": 31884 + }, + { + "epoch": 1.23, + "learning_rate": 2.958771479866633e-05, + "loss": 0.4188, + "step": 31885 + }, + { + "epoch": 1.23, + "learning_rate": 2.9587073608617594e-05, + "loss": 0.5061, + "step": 31886 + }, + { + "epoch": 1.23, + "learning_rate": 2.9586432418568865e-05, + "loss": 0.3963, + "step": 31887 + }, + { + "epoch": 1.23, + "learning_rate": 2.9585791228520134e-05, + "loss": 0.4195, + "step": 31888 + }, + { + "epoch": 1.23, + "learning_rate": 2.9585150038471406e-05, + "loss": 0.3195, + "step": 31889 + }, + { + "epoch": 1.23, + "learning_rate": 2.9584508848422677e-05, + "loss": 0.3148, + "step": 31890 + }, + { + "epoch": 1.23, + "learning_rate": 2.9583867658373942e-05, + "loss": 0.4994, + "step": 31891 + }, + { + "epoch": 1.23, + "learning_rate": 2.9583226468325214e-05, + "loss": 0.4002, + "step": 31892 + }, + { + "epoch": 1.23, + "learning_rate": 2.9582585278276482e-05, + "loss": 0.3576, + "step": 31893 + }, + { + "epoch": 1.23, + "learning_rate": 2.9581944088227754e-05, + "loss": 0.3332, + "step": 31894 + }, + { + "epoch": 1.23, + "learning_rate": 2.958130289817902e-05, + "loss": 0.384, + "step": 31895 + }, + { + "epoch": 1.23, + "learning_rate": 2.958066170813029e-05, + "loss": 0.3805, + "step": 31896 + }, + { + "epoch": 1.23, + "learning_rate": 2.958002051808156e-05, + "loss": 0.3337, + "step": 31897 + }, + { + "epoch": 1.23, + "learning_rate": 2.957937932803283e-05, + "loss": 0.3935, + "step": 31898 + }, + { + "epoch": 1.23, + "learning_rate": 2.9578738137984102e-05, + "loss": 0.3967, + "step": 31899 + }, + { + "epoch": 1.23, + "learning_rate": 2.9578096947935367e-05, + "loss": 0.2886, + "step": 31900 + }, + { + "epoch": 1.23, + "learning_rate": 2.957745575788664e-05, + "loss": 0.3429, + "step": 31901 + }, + { + "epoch": 1.23, + "learning_rate": 2.9576814567837907e-05, + "loss": 0.2809, + "step": 31902 + }, + { + "epoch": 1.23, + "learning_rate": 2.957617337778918e-05, + "loss": 0.3964, + "step": 31903 + }, + { + "epoch": 1.23, + "learning_rate": 2.9575532187740444e-05, + "loss": 0.4595, + "step": 31904 + }, + { + "epoch": 1.23, + "learning_rate": 2.957489099769172e-05, + "loss": 0.3388, + "step": 31905 + }, + { + "epoch": 1.23, + "learning_rate": 2.9574249807642984e-05, + "loss": 0.4257, + "step": 31906 + }, + { + "epoch": 1.23, + "learning_rate": 2.9573608617594256e-05, + "loss": 0.3604, + "step": 31907 + }, + { + "epoch": 1.23, + "learning_rate": 2.9572967427545527e-05, + "loss": 0.359, + "step": 31908 + }, + { + "epoch": 1.23, + "learning_rate": 2.9572326237496796e-05, + "loss": 0.3073, + "step": 31909 + }, + { + "epoch": 1.23, + "learning_rate": 2.9571685047448068e-05, + "loss": 0.4554, + "step": 31910 + }, + { + "epoch": 1.23, + "learning_rate": 2.9571043857399332e-05, + "loss": 0.3563, + "step": 31911 + }, + { + "epoch": 1.23, + "learning_rate": 2.9570402667350604e-05, + "loss": 0.465, + "step": 31912 + }, + { + "epoch": 1.23, + "learning_rate": 2.9569761477301872e-05, + "loss": 0.3184, + "step": 31913 + }, + { + "epoch": 1.23, + "learning_rate": 2.9569120287253144e-05, + "loss": 0.3917, + "step": 31914 + }, + { + "epoch": 1.23, + "learning_rate": 2.956847909720441e-05, + "loss": 0.5575, + "step": 31915 + }, + { + "epoch": 1.23, + "learning_rate": 2.956783790715568e-05, + "loss": 0.311, + "step": 31916 + }, + { + "epoch": 1.23, + "learning_rate": 2.9567196717106953e-05, + "loss": 0.4177, + "step": 31917 + }, + { + "epoch": 1.23, + "learning_rate": 2.956655552705822e-05, + "loss": 0.2645, + "step": 31918 + }, + { + "epoch": 1.23, + "learning_rate": 2.9565914337009493e-05, + "loss": 0.3555, + "step": 31919 + }, + { + "epoch": 1.23, + "learning_rate": 2.9565273146960758e-05, + "loss": 0.3394, + "step": 31920 + }, + { + "epoch": 1.23, + "learning_rate": 2.9564631956912033e-05, + "loss": 0.5089, + "step": 31921 + }, + { + "epoch": 1.23, + "learning_rate": 2.9563990766863298e-05, + "loss": 0.324, + "step": 31922 + }, + { + "epoch": 1.23, + "learning_rate": 2.956334957681457e-05, + "loss": 0.5048, + "step": 31923 + }, + { + "epoch": 1.23, + "learning_rate": 2.9562708386765838e-05, + "loss": 0.4897, + "step": 31924 + }, + { + "epoch": 1.23, + "learning_rate": 2.956206719671711e-05, + "loss": 0.3375, + "step": 31925 + }, + { + "epoch": 1.23, + "learning_rate": 2.956142600666838e-05, + "loss": 0.3148, + "step": 31926 + }, + { + "epoch": 1.23, + "learning_rate": 2.9560784816619646e-05, + "loss": 0.2988, + "step": 31927 + }, + { + "epoch": 1.23, + "learning_rate": 2.9560143626570918e-05, + "loss": 0.4167, + "step": 31928 + }, + { + "epoch": 1.23, + "learning_rate": 2.9559502436522186e-05, + "loss": 0.4113, + "step": 31929 + }, + { + "epoch": 1.23, + "learning_rate": 2.9558861246473458e-05, + "loss": 0.3633, + "step": 31930 + }, + { + "epoch": 1.23, + "learning_rate": 2.9558220056424723e-05, + "loss": 0.3843, + "step": 31931 + }, + { + "epoch": 1.23, + "learning_rate": 2.9557578866375994e-05, + "loss": 0.3491, + "step": 31932 + }, + { + "epoch": 1.23, + "learning_rate": 2.9556937676327263e-05, + "loss": 0.3819, + "step": 31933 + }, + { + "epoch": 1.23, + "learning_rate": 2.9556296486278534e-05, + "loss": 0.4166, + "step": 31934 + }, + { + "epoch": 1.23, + "learning_rate": 2.9555655296229806e-05, + "loss": 0.4991, + "step": 31935 + }, + { + "epoch": 1.23, + "learning_rate": 2.955501410618107e-05, + "loss": 0.3616, + "step": 31936 + }, + { + "epoch": 1.23, + "learning_rate": 2.9554372916132346e-05, + "loss": 0.4264, + "step": 31937 + }, + { + "epoch": 1.23, + "learning_rate": 2.955373172608361e-05, + "loss": 0.4105, + "step": 31938 + }, + { + "epoch": 1.23, + "learning_rate": 2.9553090536034883e-05, + "loss": 0.4397, + "step": 31939 + }, + { + "epoch": 1.23, + "learning_rate": 2.955244934598615e-05, + "loss": 0.3433, + "step": 31940 + }, + { + "epoch": 1.23, + "learning_rate": 2.9551808155937423e-05, + "loss": 0.3706, + "step": 31941 + }, + { + "epoch": 1.23, + "learning_rate": 2.9551166965888688e-05, + "loss": 0.3282, + "step": 31942 + }, + { + "epoch": 1.23, + "learning_rate": 2.955052577583996e-05, + "loss": 0.5208, + "step": 31943 + }, + { + "epoch": 1.23, + "learning_rate": 2.954988458579123e-05, + "loss": 0.4974, + "step": 31944 + }, + { + "epoch": 1.23, + "learning_rate": 2.95492433957425e-05, + "loss": 0.4544, + "step": 31945 + }, + { + "epoch": 1.23, + "learning_rate": 2.954860220569377e-05, + "loss": 0.4034, + "step": 31946 + }, + { + "epoch": 1.23, + "learning_rate": 2.9547961015645036e-05, + "loss": 0.3468, + "step": 31947 + }, + { + "epoch": 1.23, + "learning_rate": 2.9547319825596308e-05, + "loss": 0.3891, + "step": 31948 + }, + { + "epoch": 1.23, + "learning_rate": 2.9546678635547576e-05, + "loss": 0.3231, + "step": 31949 + }, + { + "epoch": 1.23, + "learning_rate": 2.9546037445498848e-05, + "loss": 0.4541, + "step": 31950 + }, + { + "epoch": 1.23, + "learning_rate": 2.954539625545012e-05, + "loss": 0.3789, + "step": 31951 + }, + { + "epoch": 1.23, + "learning_rate": 2.9544755065401385e-05, + "loss": 0.3546, + "step": 31952 + }, + { + "epoch": 1.23, + "learning_rate": 2.9544113875352656e-05, + "loss": 0.4053, + "step": 31953 + }, + { + "epoch": 1.23, + "learning_rate": 2.9543472685303925e-05, + "loss": 0.3107, + "step": 31954 + }, + { + "epoch": 1.23, + "learning_rate": 2.9542831495255197e-05, + "loss": 0.341, + "step": 31955 + }, + { + "epoch": 1.23, + "learning_rate": 2.954219030520646e-05, + "loss": 0.3524, + "step": 31956 + }, + { + "epoch": 1.23, + "learning_rate": 2.9541549115157737e-05, + "loss": 0.3063, + "step": 31957 + }, + { + "epoch": 1.23, + "learning_rate": 2.9540907925109e-05, + "loss": 0.3548, + "step": 31958 + }, + { + "epoch": 1.23, + "learning_rate": 2.9540266735060273e-05, + "loss": 0.3526, + "step": 31959 + }, + { + "epoch": 1.23, + "learning_rate": 2.9539625545011545e-05, + "loss": 0.3264, + "step": 31960 + }, + { + "epoch": 1.23, + "learning_rate": 2.9538984354962813e-05, + "loss": 0.3478, + "step": 31961 + }, + { + "epoch": 1.23, + "learning_rate": 2.9538343164914085e-05, + "loss": 0.4265, + "step": 31962 + }, + { + "epoch": 1.23, + "learning_rate": 2.953770197486535e-05, + "loss": 0.3644, + "step": 31963 + }, + { + "epoch": 1.23, + "learning_rate": 2.953706078481662e-05, + "loss": 0.4068, + "step": 31964 + }, + { + "epoch": 1.23, + "learning_rate": 2.953641959476789e-05, + "loss": 0.4147, + "step": 31965 + }, + { + "epoch": 1.23, + "learning_rate": 2.953577840471916e-05, + "loss": 0.4454, + "step": 31966 + }, + { + "epoch": 1.23, + "learning_rate": 2.9535137214670427e-05, + "loss": 0.4206, + "step": 31967 + }, + { + "epoch": 1.23, + "learning_rate": 2.95344960246217e-05, + "loss": 0.3654, + "step": 31968 + }, + { + "epoch": 1.23, + "learning_rate": 2.953385483457297e-05, + "loss": 0.4032, + "step": 31969 + }, + { + "epoch": 1.23, + "learning_rate": 2.953321364452424e-05, + "loss": 0.2778, + "step": 31970 + }, + { + "epoch": 1.23, + "learning_rate": 2.953257245447551e-05, + "loss": 0.3412, + "step": 31971 + }, + { + "epoch": 1.23, + "learning_rate": 2.9531931264426775e-05, + "loss": 0.4495, + "step": 31972 + }, + { + "epoch": 1.23, + "learning_rate": 2.953129007437805e-05, + "loss": 0.3596, + "step": 31973 + }, + { + "epoch": 1.23, + "learning_rate": 2.9530648884329315e-05, + "loss": 0.3611, + "step": 31974 + }, + { + "epoch": 1.23, + "learning_rate": 2.9530007694280587e-05, + "loss": 0.3475, + "step": 31975 + }, + { + "epoch": 1.23, + "learning_rate": 2.9529366504231855e-05, + "loss": 0.3516, + "step": 31976 + }, + { + "epoch": 1.23, + "learning_rate": 2.9528725314183127e-05, + "loss": 0.3378, + "step": 31977 + }, + { + "epoch": 1.23, + "learning_rate": 2.95280841241344e-05, + "loss": 0.3096, + "step": 31978 + }, + { + "epoch": 1.23, + "learning_rate": 2.9527442934085663e-05, + "loss": 0.4312, + "step": 31979 + }, + { + "epoch": 1.23, + "learning_rate": 2.9526801744036935e-05, + "loss": 0.4272, + "step": 31980 + }, + { + "epoch": 1.23, + "learning_rate": 2.9526160553988204e-05, + "loss": 0.3944, + "step": 31981 + }, + { + "epoch": 1.23, + "learning_rate": 2.9525519363939475e-05, + "loss": 0.3996, + "step": 31982 + }, + { + "epoch": 1.23, + "learning_rate": 2.952487817389074e-05, + "loss": 0.4553, + "step": 31983 + }, + { + "epoch": 1.23, + "learning_rate": 2.9524236983842012e-05, + "loss": 0.3378, + "step": 31984 + }, + { + "epoch": 1.23, + "learning_rate": 2.952359579379328e-05, + "loss": 0.3233, + "step": 31985 + }, + { + "epoch": 1.23, + "learning_rate": 2.9522954603744552e-05, + "loss": 0.4017, + "step": 31986 + }, + { + "epoch": 1.23, + "learning_rate": 2.9522313413695824e-05, + "loss": 0.4245, + "step": 31987 + }, + { + "epoch": 1.23, + "learning_rate": 2.952167222364709e-05, + "loss": 0.3119, + "step": 31988 + }, + { + "epoch": 1.23, + "learning_rate": 2.952103103359836e-05, + "loss": 0.4714, + "step": 31989 + }, + { + "epoch": 1.23, + "learning_rate": 2.952038984354963e-05, + "loss": 0.4523, + "step": 31990 + }, + { + "epoch": 1.23, + "learning_rate": 2.95197486535009e-05, + "loss": 0.2879, + "step": 31991 + }, + { + "epoch": 1.23, + "learning_rate": 2.9519107463452165e-05, + "loss": 0.3499, + "step": 31992 + }, + { + "epoch": 1.23, + "learning_rate": 2.951846627340344e-05, + "loss": 0.4368, + "step": 31993 + }, + { + "epoch": 1.23, + "learning_rate": 2.9517825083354705e-05, + "loss": 0.2945, + "step": 31994 + }, + { + "epoch": 1.23, + "learning_rate": 2.9517183893305977e-05, + "loss": 0.3789, + "step": 31995 + }, + { + "epoch": 1.23, + "learning_rate": 2.951654270325725e-05, + "loss": 0.2665, + "step": 31996 + }, + { + "epoch": 1.23, + "learning_rate": 2.9515901513208517e-05, + "loss": 0.4199, + "step": 31997 + }, + { + "epoch": 1.23, + "learning_rate": 2.951526032315979e-05, + "loss": 0.2643, + "step": 31998 + }, + { + "epoch": 1.23, + "learning_rate": 2.9514619133111054e-05, + "loss": 0.2823, + "step": 31999 + }, + { + "epoch": 1.23, + "learning_rate": 2.9513977943062325e-05, + "loss": 0.5048, + "step": 32000 + }, + { + "epoch": 1.23, + "learning_rate": 2.9513336753013594e-05, + "loss": 0.452, + "step": 32001 + }, + { + "epoch": 1.23, + "learning_rate": 2.9512695562964866e-05, + "loss": 0.3855, + "step": 32002 + }, + { + "epoch": 1.23, + "learning_rate": 2.951205437291613e-05, + "loss": 0.4858, + "step": 32003 + }, + { + "epoch": 1.23, + "learning_rate": 2.9511413182867402e-05, + "loss": 0.3789, + "step": 32004 + }, + { + "epoch": 1.23, + "learning_rate": 2.9510771992818674e-05, + "loss": 0.3212, + "step": 32005 + }, + { + "epoch": 1.23, + "learning_rate": 2.9510130802769942e-05, + "loss": 0.4373, + "step": 32006 + }, + { + "epoch": 1.23, + "learning_rate": 2.9509489612721214e-05, + "loss": 0.4874, + "step": 32007 + }, + { + "epoch": 1.23, + "learning_rate": 2.950884842267248e-05, + "loss": 0.3415, + "step": 32008 + }, + { + "epoch": 1.23, + "learning_rate": 2.9508207232623754e-05, + "loss": 0.4033, + "step": 32009 + }, + { + "epoch": 1.23, + "learning_rate": 2.950756604257502e-05, + "loss": 0.4184, + "step": 32010 + }, + { + "epoch": 1.23, + "learning_rate": 2.950692485252629e-05, + "loss": 0.4466, + "step": 32011 + }, + { + "epoch": 1.23, + "learning_rate": 2.950628366247756e-05, + "loss": 0.2668, + "step": 32012 + }, + { + "epoch": 1.23, + "learning_rate": 2.950564247242883e-05, + "loss": 0.3522, + "step": 32013 + }, + { + "epoch": 1.23, + "learning_rate": 2.9505001282380102e-05, + "loss": 0.2765, + "step": 32014 + }, + { + "epoch": 1.23, + "learning_rate": 2.9504360092331367e-05, + "loss": 0.38, + "step": 32015 + }, + { + "epoch": 1.23, + "learning_rate": 2.950371890228264e-05, + "loss": 0.3996, + "step": 32016 + }, + { + "epoch": 1.23, + "learning_rate": 2.9503077712233907e-05, + "loss": 0.4873, + "step": 32017 + }, + { + "epoch": 1.23, + "learning_rate": 2.950243652218518e-05, + "loss": 0.4395, + "step": 32018 + }, + { + "epoch": 1.23, + "learning_rate": 2.9501795332136444e-05, + "loss": 0.4379, + "step": 32019 + }, + { + "epoch": 1.23, + "learning_rate": 2.9501154142087716e-05, + "loss": 0.4146, + "step": 32020 + }, + { + "epoch": 1.23, + "learning_rate": 2.9500512952038984e-05, + "loss": 0.5484, + "step": 32021 + }, + { + "epoch": 1.23, + "learning_rate": 2.9499871761990256e-05, + "loss": 0.4074, + "step": 32022 + }, + { + "epoch": 1.23, + "learning_rate": 2.9499230571941528e-05, + "loss": 0.3708, + "step": 32023 + }, + { + "epoch": 1.23, + "learning_rate": 2.9498589381892792e-05, + "loss": 0.276, + "step": 32024 + }, + { + "epoch": 1.23, + "learning_rate": 2.9497948191844064e-05, + "loss": 0.4256, + "step": 32025 + }, + { + "epoch": 1.23, + "learning_rate": 2.9497307001795333e-05, + "loss": 0.4016, + "step": 32026 + }, + { + "epoch": 1.23, + "learning_rate": 2.9496665811746604e-05, + "loss": 0.3531, + "step": 32027 + }, + { + "epoch": 1.23, + "learning_rate": 2.949602462169787e-05, + "loss": 0.4639, + "step": 32028 + }, + { + "epoch": 1.23, + "learning_rate": 2.9495383431649144e-05, + "loss": 0.3632, + "step": 32029 + }, + { + "epoch": 1.23, + "learning_rate": 2.949474224160041e-05, + "loss": 0.3344, + "step": 32030 + }, + { + "epoch": 1.23, + "learning_rate": 2.949410105155168e-05, + "loss": 0.5283, + "step": 32031 + }, + { + "epoch": 1.23, + "learning_rate": 2.9493459861502953e-05, + "loss": 0.3854, + "step": 32032 + }, + { + "epoch": 1.23, + "learning_rate": 2.949281867145422e-05, + "loss": 0.3532, + "step": 32033 + }, + { + "epoch": 1.23, + "learning_rate": 2.9492177481405493e-05, + "loss": 0.3614, + "step": 32034 + }, + { + "epoch": 1.23, + "learning_rate": 2.9491536291356758e-05, + "loss": 0.3687, + "step": 32035 + }, + { + "epoch": 1.23, + "learning_rate": 2.949089510130803e-05, + "loss": 0.3281, + "step": 32036 + }, + { + "epoch": 1.23, + "learning_rate": 2.9490253911259298e-05, + "loss": 0.4078, + "step": 32037 + }, + { + "epoch": 1.23, + "learning_rate": 2.948961272121057e-05, + "loss": 0.4418, + "step": 32038 + }, + { + "epoch": 1.23, + "learning_rate": 2.9488971531161834e-05, + "loss": 0.4373, + "step": 32039 + }, + { + "epoch": 1.23, + "learning_rate": 2.9488330341113106e-05, + "loss": 0.2855, + "step": 32040 + }, + { + "epoch": 1.23, + "learning_rate": 2.9487689151064378e-05, + "loss": 0.455, + "step": 32041 + }, + { + "epoch": 1.23, + "learning_rate": 2.9487047961015646e-05, + "loss": 0.3888, + "step": 32042 + }, + { + "epoch": 1.23, + "learning_rate": 2.9486406770966918e-05, + "loss": 0.4386, + "step": 32043 + }, + { + "epoch": 1.23, + "learning_rate": 2.9485765580918183e-05, + "loss": 0.356, + "step": 32044 + }, + { + "epoch": 1.23, + "learning_rate": 2.9485124390869458e-05, + "loss": 0.3584, + "step": 32045 + }, + { + "epoch": 1.23, + "learning_rate": 2.9484483200820723e-05, + "loss": 0.312, + "step": 32046 + }, + { + "epoch": 1.23, + "learning_rate": 2.9483842010771995e-05, + "loss": 0.4496, + "step": 32047 + }, + { + "epoch": 1.23, + "learning_rate": 2.9483200820723263e-05, + "loss": 0.4008, + "step": 32048 + }, + { + "epoch": 1.23, + "learning_rate": 2.9482559630674535e-05, + "loss": 0.4212, + "step": 32049 + }, + { + "epoch": 1.23, + "learning_rate": 2.9481918440625806e-05, + "loss": 0.3565, + "step": 32050 + }, + { + "epoch": 1.23, + "learning_rate": 2.948127725057707e-05, + "loss": 0.3956, + "step": 32051 + }, + { + "epoch": 1.23, + "learning_rate": 2.9480636060528343e-05, + "loss": 0.5336, + "step": 32052 + }, + { + "epoch": 1.23, + "learning_rate": 2.947999487047961e-05, + "loss": 0.4145, + "step": 32053 + }, + { + "epoch": 1.23, + "learning_rate": 2.9479353680430883e-05, + "loss": 0.3451, + "step": 32054 + }, + { + "epoch": 1.23, + "learning_rate": 2.9478712490382148e-05, + "loss": 0.4298, + "step": 32055 + }, + { + "epoch": 1.23, + "learning_rate": 2.947807130033342e-05, + "loss": 0.3851, + "step": 32056 + }, + { + "epoch": 1.23, + "learning_rate": 2.9477430110284688e-05, + "loss": 0.3589, + "step": 32057 + }, + { + "epoch": 1.23, + "learning_rate": 2.947678892023596e-05, + "loss": 0.2876, + "step": 32058 + }, + { + "epoch": 1.23, + "learning_rate": 2.947614773018723e-05, + "loss": 0.2951, + "step": 32059 + }, + { + "epoch": 1.23, + "learning_rate": 2.9475506540138496e-05, + "loss": 0.437, + "step": 32060 + }, + { + "epoch": 1.23, + "learning_rate": 2.947486535008977e-05, + "loss": 0.3489, + "step": 32061 + }, + { + "epoch": 1.23, + "learning_rate": 2.9474224160041036e-05, + "loss": 0.4311, + "step": 32062 + }, + { + "epoch": 1.23, + "learning_rate": 2.9473582969992308e-05, + "loss": 0.363, + "step": 32063 + }, + { + "epoch": 1.23, + "learning_rate": 2.9472941779943576e-05, + "loss": 0.3515, + "step": 32064 + }, + { + "epoch": 1.23, + "learning_rate": 2.9472300589894848e-05, + "loss": 0.4031, + "step": 32065 + }, + { + "epoch": 1.23, + "learning_rate": 2.9471659399846113e-05, + "loss": 0.4537, + "step": 32066 + }, + { + "epoch": 1.23, + "learning_rate": 2.9471018209797385e-05, + "loss": 0.3417, + "step": 32067 + }, + { + "epoch": 1.23, + "learning_rate": 2.9470377019748657e-05, + "loss": 0.5069, + "step": 32068 + }, + { + "epoch": 1.23, + "learning_rate": 2.9469735829699925e-05, + "loss": 0.4506, + "step": 32069 + }, + { + "epoch": 1.23, + "learning_rate": 2.9469094639651197e-05, + "loss": 0.3181, + "step": 32070 + }, + { + "epoch": 1.23, + "learning_rate": 2.946845344960246e-05, + "loss": 0.206, + "step": 32071 + }, + { + "epoch": 1.23, + "learning_rate": 2.9467812259553733e-05, + "loss": 0.3886, + "step": 32072 + }, + { + "epoch": 1.23, + "learning_rate": 2.9467171069505e-05, + "loss": 0.4358, + "step": 32073 + }, + { + "epoch": 1.23, + "learning_rate": 2.9466529879456273e-05, + "loss": 0.4243, + "step": 32074 + }, + { + "epoch": 1.23, + "learning_rate": 2.9465888689407538e-05, + "loss": 0.3523, + "step": 32075 + }, + { + "epoch": 1.23, + "learning_rate": 2.946524749935881e-05, + "loss": 0.401, + "step": 32076 + }, + { + "epoch": 1.23, + "learning_rate": 2.946460630931008e-05, + "loss": 0.4673, + "step": 32077 + }, + { + "epoch": 1.23, + "learning_rate": 2.946396511926135e-05, + "loss": 0.3468, + "step": 32078 + }, + { + "epoch": 1.23, + "learning_rate": 2.9463323929212622e-05, + "loss": 0.4409, + "step": 32079 + }, + { + "epoch": 1.23, + "learning_rate": 2.9462682739163887e-05, + "loss": 0.3953, + "step": 32080 + }, + { + "epoch": 1.23, + "learning_rate": 2.9462041549115162e-05, + "loss": 0.3618, + "step": 32081 + }, + { + "epoch": 1.23, + "learning_rate": 2.9461400359066427e-05, + "loss": 0.3897, + "step": 32082 + }, + { + "epoch": 1.23, + "learning_rate": 2.94607591690177e-05, + "loss": 0.3992, + "step": 32083 + }, + { + "epoch": 1.23, + "learning_rate": 2.9460117978968967e-05, + "loss": 0.5014, + "step": 32084 + }, + { + "epoch": 1.23, + "learning_rate": 2.945947678892024e-05, + "loss": 0.497, + "step": 32085 + }, + { + "epoch": 1.23, + "learning_rate": 2.945883559887151e-05, + "loss": 0.4161, + "step": 32086 + }, + { + "epoch": 1.23, + "learning_rate": 2.9458194408822775e-05, + "loss": 0.4847, + "step": 32087 + }, + { + "epoch": 1.23, + "learning_rate": 2.9457553218774047e-05, + "loss": 0.3789, + "step": 32088 + }, + { + "epoch": 1.23, + "learning_rate": 2.9456912028725315e-05, + "loss": 0.4239, + "step": 32089 + }, + { + "epoch": 1.23, + "learning_rate": 2.9456270838676587e-05, + "loss": 0.3981, + "step": 32090 + }, + { + "epoch": 1.23, + "learning_rate": 2.9455629648627852e-05, + "loss": 0.3479, + "step": 32091 + }, + { + "epoch": 1.23, + "learning_rate": 2.9454988458579124e-05, + "loss": 0.3537, + "step": 32092 + }, + { + "epoch": 1.23, + "learning_rate": 2.9454347268530392e-05, + "loss": 0.3902, + "step": 32093 + }, + { + "epoch": 1.23, + "learning_rate": 2.9453706078481664e-05, + "loss": 0.4804, + "step": 32094 + }, + { + "epoch": 1.23, + "learning_rate": 2.9453064888432935e-05, + "loss": 0.2901, + "step": 32095 + }, + { + "epoch": 1.23, + "learning_rate": 2.94524236983842e-05, + "loss": 0.3959, + "step": 32096 + }, + { + "epoch": 1.23, + "learning_rate": 2.9451782508335475e-05, + "loss": 0.3247, + "step": 32097 + }, + { + "epoch": 1.23, + "learning_rate": 2.945114131828674e-05, + "loss": 0.3771, + "step": 32098 + }, + { + "epoch": 1.23, + "learning_rate": 2.9450500128238012e-05, + "loss": 0.3731, + "step": 32099 + }, + { + "epoch": 1.23, + "learning_rate": 2.944985893818928e-05, + "loss": 0.6166, + "step": 32100 + }, + { + "epoch": 1.23, + "learning_rate": 2.9449217748140552e-05, + "loss": 0.5107, + "step": 32101 + }, + { + "epoch": 1.23, + "learning_rate": 2.9448576558091817e-05, + "loss": 0.3921, + "step": 32102 + }, + { + "epoch": 1.23, + "learning_rate": 2.944793536804309e-05, + "loss": 0.353, + "step": 32103 + }, + { + "epoch": 1.23, + "learning_rate": 2.944729417799436e-05, + "loss": 0.3195, + "step": 32104 + }, + { + "epoch": 1.23, + "learning_rate": 2.944665298794563e-05, + "loss": 0.3555, + "step": 32105 + }, + { + "epoch": 1.23, + "learning_rate": 2.94460117978969e-05, + "loss": 0.3222, + "step": 32106 + }, + { + "epoch": 1.23, + "learning_rate": 2.9445370607848165e-05, + "loss": 0.3932, + "step": 32107 + }, + { + "epoch": 1.23, + "learning_rate": 2.9444729417799437e-05, + "loss": 0.5495, + "step": 32108 + }, + { + "epoch": 1.23, + "learning_rate": 2.9444088227750705e-05, + "loss": 0.3484, + "step": 32109 + }, + { + "epoch": 1.23, + "learning_rate": 2.9443447037701977e-05, + "loss": 0.2379, + "step": 32110 + }, + { + "epoch": 1.23, + "learning_rate": 2.9442805847653242e-05, + "loss": 0.416, + "step": 32111 + }, + { + "epoch": 1.23, + "learning_rate": 2.9442164657604514e-05, + "loss": 0.4403, + "step": 32112 + }, + { + "epoch": 1.23, + "learning_rate": 2.9441523467555786e-05, + "loss": 0.418, + "step": 32113 + }, + { + "epoch": 1.23, + "learning_rate": 2.9440882277507054e-05, + "loss": 0.4964, + "step": 32114 + }, + { + "epoch": 1.23, + "learning_rate": 2.9440241087458326e-05, + "loss": 0.3381, + "step": 32115 + }, + { + "epoch": 1.23, + "learning_rate": 2.943959989740959e-05, + "loss": 0.2635, + "step": 32116 + }, + { + "epoch": 1.23, + "learning_rate": 2.9438958707360866e-05, + "loss": 0.2697, + "step": 32117 + }, + { + "epoch": 1.23, + "learning_rate": 2.943831751731213e-05, + "loss": 0.4831, + "step": 32118 + }, + { + "epoch": 1.23, + "learning_rate": 2.9437676327263402e-05, + "loss": 0.34, + "step": 32119 + }, + { + "epoch": 1.23, + "learning_rate": 2.943703513721467e-05, + "loss": 0.5528, + "step": 32120 + }, + { + "epoch": 1.23, + "learning_rate": 2.9436393947165942e-05, + "loss": 0.5236, + "step": 32121 + }, + { + "epoch": 1.23, + "learning_rate": 2.9435752757117214e-05, + "loss": 0.3136, + "step": 32122 + }, + { + "epoch": 1.24, + "learning_rate": 2.943511156706848e-05, + "loss": 0.4454, + "step": 32123 + }, + { + "epoch": 1.24, + "learning_rate": 2.943447037701975e-05, + "loss": 0.3219, + "step": 32124 + }, + { + "epoch": 1.24, + "learning_rate": 2.943382918697102e-05, + "loss": 0.451, + "step": 32125 + }, + { + "epoch": 1.24, + "learning_rate": 2.943318799692229e-05, + "loss": 0.3071, + "step": 32126 + }, + { + "epoch": 1.24, + "learning_rate": 2.9432546806873556e-05, + "loss": 0.3725, + "step": 32127 + }, + { + "epoch": 1.24, + "learning_rate": 2.9431905616824827e-05, + "loss": 0.3301, + "step": 32128 + }, + { + "epoch": 1.24, + "learning_rate": 2.9431264426776096e-05, + "loss": 0.3752, + "step": 32129 + }, + { + "epoch": 1.24, + "learning_rate": 2.9430623236727367e-05, + "loss": 0.3299, + "step": 32130 + }, + { + "epoch": 1.24, + "learning_rate": 2.942998204667864e-05, + "loss": 0.3978, + "step": 32131 + }, + { + "epoch": 1.24, + "learning_rate": 2.9429340856629904e-05, + "loss": 0.2602, + "step": 32132 + }, + { + "epoch": 1.24, + "learning_rate": 2.942869966658118e-05, + "loss": 0.4315, + "step": 32133 + }, + { + "epoch": 1.24, + "learning_rate": 2.9428058476532444e-05, + "loss": 0.3596, + "step": 32134 + }, + { + "epoch": 1.24, + "learning_rate": 2.9427417286483716e-05, + "loss": 0.4548, + "step": 32135 + }, + { + "epoch": 1.24, + "learning_rate": 2.9426776096434984e-05, + "loss": 0.2499, + "step": 32136 + }, + { + "epoch": 1.24, + "learning_rate": 2.9426134906386256e-05, + "loss": 0.3998, + "step": 32137 + }, + { + "epoch": 1.24, + "learning_rate": 2.9425493716337528e-05, + "loss": 0.3285, + "step": 32138 + }, + { + "epoch": 1.24, + "learning_rate": 2.9424852526288793e-05, + "loss": 0.3636, + "step": 32139 + }, + { + "epoch": 1.24, + "learning_rate": 2.9424211336240064e-05, + "loss": 0.4571, + "step": 32140 + }, + { + "epoch": 1.24, + "learning_rate": 2.9423570146191333e-05, + "loss": 0.445, + "step": 32141 + }, + { + "epoch": 1.24, + "learning_rate": 2.9422928956142604e-05, + "loss": 0.4437, + "step": 32142 + }, + { + "epoch": 1.24, + "learning_rate": 2.942228776609387e-05, + "loss": 0.3564, + "step": 32143 + }, + { + "epoch": 1.24, + "learning_rate": 2.942164657604514e-05, + "loss": 0.3796, + "step": 32144 + }, + { + "epoch": 1.24, + "learning_rate": 2.942100538599641e-05, + "loss": 0.3785, + "step": 32145 + }, + { + "epoch": 1.24, + "learning_rate": 2.942036419594768e-05, + "loss": 0.3265, + "step": 32146 + }, + { + "epoch": 1.24, + "learning_rate": 2.9419723005898953e-05, + "loss": 0.3767, + "step": 32147 + }, + { + "epoch": 1.24, + "learning_rate": 2.9419081815850218e-05, + "loss": 0.5136, + "step": 32148 + }, + { + "epoch": 1.24, + "learning_rate": 2.941844062580149e-05, + "loss": 0.5289, + "step": 32149 + }, + { + "epoch": 1.24, + "learning_rate": 2.9417799435752758e-05, + "loss": 0.3419, + "step": 32150 + }, + { + "epoch": 1.24, + "learning_rate": 2.941715824570403e-05, + "loss": 0.5026, + "step": 32151 + }, + { + "epoch": 1.24, + "learning_rate": 2.9416517055655294e-05, + "loss": 0.4729, + "step": 32152 + }, + { + "epoch": 1.24, + "learning_rate": 2.941587586560657e-05, + "loss": 0.355, + "step": 32153 + }, + { + "epoch": 1.24, + "learning_rate": 2.9415234675557834e-05, + "loss": 0.3846, + "step": 32154 + }, + { + "epoch": 1.24, + "learning_rate": 2.9414593485509106e-05, + "loss": 0.2667, + "step": 32155 + }, + { + "epoch": 1.24, + "learning_rate": 2.9413952295460378e-05, + "loss": 0.4153, + "step": 32156 + }, + { + "epoch": 1.24, + "learning_rate": 2.9413311105411646e-05, + "loss": 0.3682, + "step": 32157 + }, + { + "epoch": 1.24, + "learning_rate": 2.9412669915362918e-05, + "loss": 0.4882, + "step": 32158 + }, + { + "epoch": 1.24, + "learning_rate": 2.9412028725314183e-05, + "loss": 0.4315, + "step": 32159 + }, + { + "epoch": 1.24, + "learning_rate": 2.9411387535265455e-05, + "loss": 0.4903, + "step": 32160 + }, + { + "epoch": 1.24, + "learning_rate": 2.9410746345216723e-05, + "loss": 0.2862, + "step": 32161 + }, + { + "epoch": 1.24, + "learning_rate": 2.9410105155167995e-05, + "loss": 0.4831, + "step": 32162 + }, + { + "epoch": 1.24, + "learning_rate": 2.940946396511926e-05, + "loss": 0.366, + "step": 32163 + }, + { + "epoch": 1.24, + "learning_rate": 2.940882277507053e-05, + "loss": 0.2851, + "step": 32164 + }, + { + "epoch": 1.24, + "learning_rate": 2.9408181585021803e-05, + "loss": 0.5441, + "step": 32165 + }, + { + "epoch": 1.24, + "learning_rate": 2.940754039497307e-05, + "loss": 0.2972, + "step": 32166 + }, + { + "epoch": 1.24, + "learning_rate": 2.9406899204924343e-05, + "loss": 0.3659, + "step": 32167 + }, + { + "epoch": 1.24, + "learning_rate": 2.9406258014875608e-05, + "loss": 0.3096, + "step": 32168 + }, + { + "epoch": 1.24, + "learning_rate": 2.9405616824826883e-05, + "loss": 0.3443, + "step": 32169 + }, + { + "epoch": 1.24, + "learning_rate": 2.9404975634778148e-05, + "loss": 0.4013, + "step": 32170 + }, + { + "epoch": 1.24, + "learning_rate": 2.940433444472942e-05, + "loss": 0.6037, + "step": 32171 + }, + { + "epoch": 1.24, + "learning_rate": 2.9403693254680688e-05, + "loss": 0.3025, + "step": 32172 + }, + { + "epoch": 1.24, + "learning_rate": 2.940305206463196e-05, + "loss": 0.3541, + "step": 32173 + }, + { + "epoch": 1.24, + "learning_rate": 2.940241087458323e-05, + "loss": 0.272, + "step": 32174 + }, + { + "epoch": 1.24, + "learning_rate": 2.9401769684534496e-05, + "loss": 0.3772, + "step": 32175 + }, + { + "epoch": 1.24, + "learning_rate": 2.9401128494485768e-05, + "loss": 0.5267, + "step": 32176 + }, + { + "epoch": 1.24, + "learning_rate": 2.9400487304437036e-05, + "loss": 0.4604, + "step": 32177 + }, + { + "epoch": 1.24, + "learning_rate": 2.9399846114388308e-05, + "loss": 0.426, + "step": 32178 + }, + { + "epoch": 1.24, + "learning_rate": 2.9399204924339573e-05, + "loss": 0.3116, + "step": 32179 + }, + { + "epoch": 1.24, + "learning_rate": 2.9398563734290845e-05, + "loss": 0.4798, + "step": 32180 + }, + { + "epoch": 1.24, + "learning_rate": 2.9397922544242113e-05, + "loss": 0.4145, + "step": 32181 + }, + { + "epoch": 1.24, + "learning_rate": 2.9397281354193385e-05, + "loss": 0.4102, + "step": 32182 + }, + { + "epoch": 1.24, + "learning_rate": 2.9396640164144657e-05, + "loss": 0.2895, + "step": 32183 + }, + { + "epoch": 1.24, + "learning_rate": 2.939599897409592e-05, + "loss": 0.3816, + "step": 32184 + }, + { + "epoch": 1.24, + "learning_rate": 2.9395357784047197e-05, + "loss": 0.324, + "step": 32185 + }, + { + "epoch": 1.24, + "learning_rate": 2.939471659399846e-05, + "loss": 0.66, + "step": 32186 + }, + { + "epoch": 1.24, + "learning_rate": 2.9394075403949733e-05, + "loss": 0.3181, + "step": 32187 + }, + { + "epoch": 1.24, + "learning_rate": 2.9393434213900998e-05, + "loss": 0.3503, + "step": 32188 + }, + { + "epoch": 1.24, + "learning_rate": 2.9392793023852273e-05, + "loss": 0.4128, + "step": 32189 + }, + { + "epoch": 1.24, + "learning_rate": 2.9392151833803538e-05, + "loss": 0.4773, + "step": 32190 + }, + { + "epoch": 1.24, + "learning_rate": 2.939151064375481e-05, + "loss": 0.4809, + "step": 32191 + }, + { + "epoch": 1.24, + "learning_rate": 2.9390869453706082e-05, + "loss": 0.3725, + "step": 32192 + }, + { + "epoch": 1.24, + "learning_rate": 2.939022826365735e-05, + "loss": 0.3748, + "step": 32193 + }, + { + "epoch": 1.24, + "learning_rate": 2.9389587073608622e-05, + "loss": 0.2684, + "step": 32194 + }, + { + "epoch": 1.24, + "learning_rate": 2.9388945883559887e-05, + "loss": 0.2867, + "step": 32195 + }, + { + "epoch": 1.24, + "learning_rate": 2.938830469351116e-05, + "loss": 0.424, + "step": 32196 + }, + { + "epoch": 1.24, + "learning_rate": 2.9387663503462427e-05, + "loss": 0.3922, + "step": 32197 + }, + { + "epoch": 1.24, + "learning_rate": 2.93870223134137e-05, + "loss": 0.2771, + "step": 32198 + }, + { + "epoch": 1.24, + "learning_rate": 2.9386381123364963e-05, + "loss": 0.2931, + "step": 32199 + }, + { + "epoch": 1.24, + "learning_rate": 2.9385739933316235e-05, + "loss": 0.461, + "step": 32200 + }, + { + "epoch": 1.24, + "learning_rate": 2.9385098743267507e-05, + "loss": 0.3534, + "step": 32201 + }, + { + "epoch": 1.24, + "learning_rate": 2.9384457553218775e-05, + "loss": 0.3852, + "step": 32202 + }, + { + "epoch": 1.24, + "learning_rate": 2.9383816363170047e-05, + "loss": 0.5393, + "step": 32203 + }, + { + "epoch": 1.24, + "learning_rate": 2.9383175173121312e-05, + "loss": 0.2592, + "step": 32204 + }, + { + "epoch": 1.24, + "learning_rate": 2.9382533983072587e-05, + "loss": 0.2661, + "step": 32205 + }, + { + "epoch": 1.24, + "learning_rate": 2.9381892793023852e-05, + "loss": 0.3153, + "step": 32206 + }, + { + "epoch": 1.24, + "learning_rate": 2.9381251602975124e-05, + "loss": 0.2308, + "step": 32207 + }, + { + "epoch": 1.24, + "learning_rate": 2.9380610412926392e-05, + "loss": 0.3404, + "step": 32208 + }, + { + "epoch": 1.24, + "learning_rate": 2.9379969222877664e-05, + "loss": 0.5455, + "step": 32209 + }, + { + "epoch": 1.24, + "learning_rate": 2.9379328032828935e-05, + "loss": 0.4932, + "step": 32210 + }, + { + "epoch": 1.24, + "learning_rate": 2.93786868427802e-05, + "loss": 0.4082, + "step": 32211 + }, + { + "epoch": 1.24, + "learning_rate": 2.9378045652731472e-05, + "loss": 0.3812, + "step": 32212 + }, + { + "epoch": 1.24, + "learning_rate": 2.937740446268274e-05, + "loss": 0.5006, + "step": 32213 + }, + { + "epoch": 1.24, + "learning_rate": 2.9376763272634012e-05, + "loss": 0.5341, + "step": 32214 + }, + { + "epoch": 1.24, + "learning_rate": 2.9376122082585277e-05, + "loss": 0.3639, + "step": 32215 + }, + { + "epoch": 1.24, + "learning_rate": 2.937548089253655e-05, + "loss": 0.2974, + "step": 32216 + }, + { + "epoch": 1.24, + "learning_rate": 2.9374839702487817e-05, + "loss": 0.4502, + "step": 32217 + }, + { + "epoch": 1.24, + "learning_rate": 2.937419851243909e-05, + "loss": 0.2909, + "step": 32218 + }, + { + "epoch": 1.24, + "learning_rate": 2.937355732239036e-05, + "loss": 0.366, + "step": 32219 + }, + { + "epoch": 1.24, + "learning_rate": 2.9372916132341625e-05, + "loss": 0.3291, + "step": 32220 + }, + { + "epoch": 1.24, + "learning_rate": 2.93722749422929e-05, + "loss": 0.5026, + "step": 32221 + }, + { + "epoch": 1.24, + "learning_rate": 2.9371633752244165e-05, + "loss": 0.5188, + "step": 32222 + }, + { + "epoch": 1.24, + "learning_rate": 2.9370992562195437e-05, + "loss": 0.4181, + "step": 32223 + }, + { + "epoch": 1.24, + "learning_rate": 2.9370351372146706e-05, + "loss": 0.4017, + "step": 32224 + }, + { + "epoch": 1.24, + "learning_rate": 2.9369710182097977e-05, + "loss": 0.4154, + "step": 32225 + }, + { + "epoch": 1.24, + "learning_rate": 2.9369068992049242e-05, + "loss": 0.3766, + "step": 32226 + }, + { + "epoch": 1.24, + "learning_rate": 2.9368427802000514e-05, + "loss": 0.4433, + "step": 32227 + }, + { + "epoch": 1.24, + "learning_rate": 2.9367786611951786e-05, + "loss": 0.2785, + "step": 32228 + }, + { + "epoch": 1.24, + "learning_rate": 2.9367145421903054e-05, + "loss": 0.3178, + "step": 32229 + }, + { + "epoch": 1.24, + "learning_rate": 2.9366504231854326e-05, + "loss": 0.3081, + "step": 32230 + }, + { + "epoch": 1.24, + "learning_rate": 2.936586304180559e-05, + "loss": 0.3463, + "step": 32231 + }, + { + "epoch": 1.24, + "learning_rate": 2.9365221851756862e-05, + "loss": 0.4224, + "step": 32232 + }, + { + "epoch": 1.24, + "learning_rate": 2.936458066170813e-05, + "loss": 0.258, + "step": 32233 + }, + { + "epoch": 1.24, + "learning_rate": 2.9363939471659402e-05, + "loss": 0.4075, + "step": 32234 + }, + { + "epoch": 1.24, + "learning_rate": 2.9363298281610667e-05, + "loss": 0.3172, + "step": 32235 + }, + { + "epoch": 1.24, + "learning_rate": 2.936265709156194e-05, + "loss": 0.4395, + "step": 32236 + }, + { + "epoch": 1.24, + "learning_rate": 2.936201590151321e-05, + "loss": 0.3545, + "step": 32237 + }, + { + "epoch": 1.24, + "learning_rate": 2.936137471146448e-05, + "loss": 0.4695, + "step": 32238 + }, + { + "epoch": 1.24, + "learning_rate": 2.936073352141575e-05, + "loss": 0.3775, + "step": 32239 + }, + { + "epoch": 1.24, + "learning_rate": 2.9360092331367016e-05, + "loss": 0.3309, + "step": 32240 + }, + { + "epoch": 1.24, + "learning_rate": 2.935945114131829e-05, + "loss": 0.2902, + "step": 32241 + }, + { + "epoch": 1.24, + "learning_rate": 2.9358809951269556e-05, + "loss": 0.3827, + "step": 32242 + }, + { + "epoch": 1.24, + "learning_rate": 2.9358168761220827e-05, + "loss": 0.2709, + "step": 32243 + }, + { + "epoch": 1.24, + "learning_rate": 2.9357527571172096e-05, + "loss": 0.341, + "step": 32244 + }, + { + "epoch": 1.24, + "learning_rate": 2.9356886381123368e-05, + "loss": 0.3902, + "step": 32245 + }, + { + "epoch": 1.24, + "learning_rate": 2.935624519107464e-05, + "loss": 0.2677, + "step": 32246 + }, + { + "epoch": 1.24, + "learning_rate": 2.9355604001025904e-05, + "loss": 0.5124, + "step": 32247 + }, + { + "epoch": 1.24, + "learning_rate": 2.9354962810977176e-05, + "loss": 0.3933, + "step": 32248 + }, + { + "epoch": 1.24, + "learning_rate": 2.9354321620928444e-05, + "loss": 0.4056, + "step": 32249 + }, + { + "epoch": 1.24, + "learning_rate": 2.9353680430879716e-05, + "loss": 0.379, + "step": 32250 + }, + { + "epoch": 1.24, + "learning_rate": 2.935303924083098e-05, + "loss": 0.4624, + "step": 32251 + }, + { + "epoch": 1.24, + "learning_rate": 2.9352398050782253e-05, + "loss": 0.3034, + "step": 32252 + }, + { + "epoch": 1.24, + "learning_rate": 2.935175686073352e-05, + "loss": 0.4607, + "step": 32253 + }, + { + "epoch": 1.24, + "learning_rate": 2.9351115670684793e-05, + "loss": 0.4962, + "step": 32254 + }, + { + "epoch": 1.24, + "learning_rate": 2.9350474480636064e-05, + "loss": 0.3818, + "step": 32255 + }, + { + "epoch": 1.24, + "learning_rate": 2.934983329058733e-05, + "loss": 0.3847, + "step": 32256 + }, + { + "epoch": 1.24, + "learning_rate": 2.9349192100538604e-05, + "loss": 0.3659, + "step": 32257 + }, + { + "epoch": 1.24, + "learning_rate": 2.934855091048987e-05, + "loss": 0.5538, + "step": 32258 + }, + { + "epoch": 1.24, + "learning_rate": 2.934790972044114e-05, + "loss": 0.3244, + "step": 32259 + }, + { + "epoch": 1.24, + "learning_rate": 2.934726853039241e-05, + "loss": 0.3752, + "step": 32260 + }, + { + "epoch": 1.24, + "learning_rate": 2.934662734034368e-05, + "loss": 0.3847, + "step": 32261 + }, + { + "epoch": 1.24, + "learning_rate": 2.9345986150294946e-05, + "loss": 0.4544, + "step": 32262 + }, + { + "epoch": 1.24, + "learning_rate": 2.9345344960246218e-05, + "loss": 0.4423, + "step": 32263 + }, + { + "epoch": 1.24, + "learning_rate": 2.934470377019749e-05, + "loss": 0.3715, + "step": 32264 + }, + { + "epoch": 1.24, + "learning_rate": 2.9344062580148758e-05, + "loss": 0.4428, + "step": 32265 + }, + { + "epoch": 1.24, + "learning_rate": 2.934342139010003e-05, + "loss": 0.3014, + "step": 32266 + }, + { + "epoch": 1.24, + "learning_rate": 2.9342780200051294e-05, + "loss": 0.3187, + "step": 32267 + }, + { + "epoch": 1.24, + "learning_rate": 2.9342139010002566e-05, + "loss": 0.4013, + "step": 32268 + }, + { + "epoch": 1.24, + "learning_rate": 2.9341497819953835e-05, + "loss": 0.4543, + "step": 32269 + }, + { + "epoch": 1.24, + "learning_rate": 2.9340856629905106e-05, + "loss": 0.3849, + "step": 32270 + }, + { + "epoch": 1.24, + "learning_rate": 2.934021543985637e-05, + "loss": 0.4472, + "step": 32271 + }, + { + "epoch": 1.24, + "learning_rate": 2.9339574249807643e-05, + "loss": 0.3865, + "step": 32272 + }, + { + "epoch": 1.24, + "learning_rate": 2.9338933059758915e-05, + "loss": 0.4395, + "step": 32273 + }, + { + "epoch": 1.24, + "learning_rate": 2.9338291869710183e-05, + "loss": 0.3794, + "step": 32274 + }, + { + "epoch": 1.24, + "learning_rate": 2.9337650679661455e-05, + "loss": 0.3604, + "step": 32275 + }, + { + "epoch": 1.24, + "learning_rate": 2.933700948961272e-05, + "loss": 0.4046, + "step": 32276 + }, + { + "epoch": 1.24, + "learning_rate": 2.9336368299563995e-05, + "loss": 0.3722, + "step": 32277 + }, + { + "epoch": 1.24, + "learning_rate": 2.933572710951526e-05, + "loss": 0.4627, + "step": 32278 + }, + { + "epoch": 1.24, + "learning_rate": 2.933508591946653e-05, + "loss": 0.4805, + "step": 32279 + }, + { + "epoch": 1.24, + "learning_rate": 2.93344447294178e-05, + "loss": 0.3496, + "step": 32280 + }, + { + "epoch": 1.24, + "learning_rate": 2.933380353936907e-05, + "loss": 0.4627, + "step": 32281 + }, + { + "epoch": 1.24, + "learning_rate": 2.9333162349320343e-05, + "loss": 0.3681, + "step": 32282 + }, + { + "epoch": 1.24, + "learning_rate": 2.9332521159271608e-05, + "loss": 0.3325, + "step": 32283 + }, + { + "epoch": 1.24, + "learning_rate": 2.933187996922288e-05, + "loss": 0.4903, + "step": 32284 + }, + { + "epoch": 1.24, + "learning_rate": 2.9331238779174148e-05, + "loss": 0.4585, + "step": 32285 + }, + { + "epoch": 1.24, + "learning_rate": 2.933059758912542e-05, + "loss": 0.4387, + "step": 32286 + }, + { + "epoch": 1.24, + "learning_rate": 2.9329956399076685e-05, + "loss": 0.3807, + "step": 32287 + }, + { + "epoch": 1.24, + "learning_rate": 2.9329315209027956e-05, + "loss": 0.4341, + "step": 32288 + }, + { + "epoch": 1.24, + "learning_rate": 2.9328674018979225e-05, + "loss": 0.4724, + "step": 32289 + }, + { + "epoch": 1.24, + "learning_rate": 2.9328032828930497e-05, + "loss": 0.3565, + "step": 32290 + }, + { + "epoch": 1.24, + "learning_rate": 2.9327391638881768e-05, + "loss": 0.4032, + "step": 32291 + }, + { + "epoch": 1.24, + "learning_rate": 2.9326750448833033e-05, + "loss": 0.4766, + "step": 32292 + }, + { + "epoch": 1.24, + "learning_rate": 2.9326109258784308e-05, + "loss": 0.49, + "step": 32293 + }, + { + "epoch": 1.24, + "learning_rate": 2.9325468068735573e-05, + "loss": 0.4626, + "step": 32294 + }, + { + "epoch": 1.24, + "learning_rate": 2.9324826878686845e-05, + "loss": 0.4078, + "step": 32295 + }, + { + "epoch": 1.24, + "learning_rate": 2.9324185688638113e-05, + "loss": 0.3841, + "step": 32296 + }, + { + "epoch": 1.24, + "learning_rate": 2.9323544498589385e-05, + "loss": 0.3201, + "step": 32297 + }, + { + "epoch": 1.24, + "learning_rate": 2.932290330854065e-05, + "loss": 0.3248, + "step": 32298 + }, + { + "epoch": 1.24, + "learning_rate": 2.932226211849192e-05, + "loss": 0.3022, + "step": 32299 + }, + { + "epoch": 1.24, + "learning_rate": 2.9321620928443193e-05, + "loss": 0.4551, + "step": 32300 + }, + { + "epoch": 1.24, + "learning_rate": 2.932097973839446e-05, + "loss": 0.3975, + "step": 32301 + }, + { + "epoch": 1.24, + "learning_rate": 2.9320338548345733e-05, + "loss": 0.4056, + "step": 32302 + }, + { + "epoch": 1.24, + "learning_rate": 2.9319697358297e-05, + "loss": 0.4546, + "step": 32303 + }, + { + "epoch": 1.24, + "learning_rate": 2.931905616824827e-05, + "loss": 0.3533, + "step": 32304 + }, + { + "epoch": 1.24, + "learning_rate": 2.931841497819954e-05, + "loss": 0.3349, + "step": 32305 + }, + { + "epoch": 1.24, + "learning_rate": 2.931777378815081e-05, + "loss": 0.4536, + "step": 32306 + }, + { + "epoch": 1.24, + "learning_rate": 2.9317132598102075e-05, + "loss": 0.3397, + "step": 32307 + }, + { + "epoch": 1.24, + "learning_rate": 2.9316491408053347e-05, + "loss": 0.4196, + "step": 32308 + }, + { + "epoch": 1.24, + "learning_rate": 2.9315850218004622e-05, + "loss": 0.484, + "step": 32309 + }, + { + "epoch": 1.24, + "learning_rate": 2.9315209027955887e-05, + "loss": 0.3711, + "step": 32310 + }, + { + "epoch": 1.24, + "learning_rate": 2.931456783790716e-05, + "loss": 0.2926, + "step": 32311 + }, + { + "epoch": 1.24, + "learning_rate": 2.9313926647858423e-05, + "loss": 0.417, + "step": 32312 + }, + { + "epoch": 1.24, + "learning_rate": 2.93132854578097e-05, + "loss": 0.3638, + "step": 32313 + }, + { + "epoch": 1.24, + "learning_rate": 2.9312644267760963e-05, + "loss": 0.478, + "step": 32314 + }, + { + "epoch": 1.24, + "learning_rate": 2.9312003077712235e-05, + "loss": 0.4793, + "step": 32315 + }, + { + "epoch": 1.24, + "learning_rate": 2.9311361887663504e-05, + "loss": 0.3368, + "step": 32316 + }, + { + "epoch": 1.24, + "learning_rate": 2.9310720697614775e-05, + "loss": 0.2839, + "step": 32317 + }, + { + "epoch": 1.24, + "learning_rate": 2.9310079507566047e-05, + "loss": 0.4298, + "step": 32318 + }, + { + "epoch": 1.24, + "learning_rate": 2.9309438317517312e-05, + "loss": 0.4123, + "step": 32319 + }, + { + "epoch": 1.24, + "learning_rate": 2.9308797127468584e-05, + "loss": 0.371, + "step": 32320 + }, + { + "epoch": 1.24, + "learning_rate": 2.9308155937419852e-05, + "loss": 0.3224, + "step": 32321 + }, + { + "epoch": 1.24, + "learning_rate": 2.9307514747371124e-05, + "loss": 0.4219, + "step": 32322 + }, + { + "epoch": 1.24, + "learning_rate": 2.930687355732239e-05, + "loss": 0.4269, + "step": 32323 + }, + { + "epoch": 1.24, + "learning_rate": 2.930623236727366e-05, + "loss": 0.3633, + "step": 32324 + }, + { + "epoch": 1.24, + "learning_rate": 2.9305591177224932e-05, + "loss": 0.4353, + "step": 32325 + }, + { + "epoch": 1.24, + "learning_rate": 2.93049499871762e-05, + "loss": 0.4704, + "step": 32326 + }, + { + "epoch": 1.24, + "learning_rate": 2.9304308797127472e-05, + "loss": 0.3072, + "step": 32327 + }, + { + "epoch": 1.24, + "learning_rate": 2.9303667607078737e-05, + "loss": 0.3764, + "step": 32328 + }, + { + "epoch": 1.24, + "learning_rate": 2.9303026417030012e-05, + "loss": 0.4, + "step": 32329 + }, + { + "epoch": 1.24, + "learning_rate": 2.9302385226981277e-05, + "loss": 0.3643, + "step": 32330 + }, + { + "epoch": 1.24, + "learning_rate": 2.930174403693255e-05, + "loss": 0.4289, + "step": 32331 + }, + { + "epoch": 1.24, + "learning_rate": 2.9301102846883817e-05, + "loss": 0.4553, + "step": 32332 + }, + { + "epoch": 1.24, + "learning_rate": 2.930046165683509e-05, + "loss": 0.3866, + "step": 32333 + }, + { + "epoch": 1.24, + "learning_rate": 2.929982046678636e-05, + "loss": 0.3023, + "step": 32334 + }, + { + "epoch": 1.24, + "learning_rate": 2.9299179276737626e-05, + "loss": 0.4799, + "step": 32335 + }, + { + "epoch": 1.24, + "learning_rate": 2.9298538086688897e-05, + "loss": 0.3508, + "step": 32336 + }, + { + "epoch": 1.24, + "learning_rate": 2.9297896896640166e-05, + "loss": 0.2766, + "step": 32337 + }, + { + "epoch": 1.24, + "learning_rate": 2.9297255706591437e-05, + "loss": 0.3097, + "step": 32338 + }, + { + "epoch": 1.24, + "learning_rate": 2.9296614516542702e-05, + "loss": 0.3357, + "step": 32339 + }, + { + "epoch": 1.24, + "learning_rate": 2.9295973326493974e-05, + "loss": 0.4527, + "step": 32340 + }, + { + "epoch": 1.24, + "learning_rate": 2.9295332136445242e-05, + "loss": 0.3234, + "step": 32341 + }, + { + "epoch": 1.24, + "learning_rate": 2.9294690946396514e-05, + "loss": 0.3413, + "step": 32342 + }, + { + "epoch": 1.24, + "learning_rate": 2.9294049756347786e-05, + "loss": 0.3873, + "step": 32343 + }, + { + "epoch": 1.24, + "learning_rate": 2.929340856629905e-05, + "loss": 0.4165, + "step": 32344 + }, + { + "epoch": 1.24, + "learning_rate": 2.9292767376250326e-05, + "loss": 0.3406, + "step": 32345 + }, + { + "epoch": 1.24, + "learning_rate": 2.929212618620159e-05, + "loss": 0.2957, + "step": 32346 + }, + { + "epoch": 1.24, + "learning_rate": 2.9291484996152862e-05, + "loss": 0.4177, + "step": 32347 + }, + { + "epoch": 1.24, + "learning_rate": 2.929084380610413e-05, + "loss": 0.4119, + "step": 32348 + }, + { + "epoch": 1.24, + "learning_rate": 2.9290202616055402e-05, + "loss": 0.3611, + "step": 32349 + }, + { + "epoch": 1.24, + "learning_rate": 2.9289561426006667e-05, + "loss": 0.4206, + "step": 32350 + }, + { + "epoch": 1.24, + "learning_rate": 2.928892023595794e-05, + "loss": 0.397, + "step": 32351 + }, + { + "epoch": 1.24, + "learning_rate": 2.928827904590921e-05, + "loss": 0.3364, + "step": 32352 + }, + { + "epoch": 1.24, + "learning_rate": 2.928763785586048e-05, + "loss": 0.3324, + "step": 32353 + }, + { + "epoch": 1.24, + "learning_rate": 2.928699666581175e-05, + "loss": 0.333, + "step": 32354 + }, + { + "epoch": 1.24, + "learning_rate": 2.9286355475763016e-05, + "loss": 0.4831, + "step": 32355 + }, + { + "epoch": 1.24, + "learning_rate": 2.9285714285714288e-05, + "loss": 0.337, + "step": 32356 + }, + { + "epoch": 1.24, + "learning_rate": 2.9285073095665556e-05, + "loss": 0.3355, + "step": 32357 + }, + { + "epoch": 1.24, + "learning_rate": 2.9284431905616828e-05, + "loss": 0.3828, + "step": 32358 + }, + { + "epoch": 1.24, + "learning_rate": 2.9283790715568092e-05, + "loss": 0.295, + "step": 32359 + }, + { + "epoch": 1.24, + "learning_rate": 2.9283149525519364e-05, + "loss": 0.3702, + "step": 32360 + }, + { + "epoch": 1.24, + "learning_rate": 2.9282508335470636e-05, + "loss": 0.4159, + "step": 32361 + }, + { + "epoch": 1.24, + "learning_rate": 2.9281867145421904e-05, + "loss": 0.266, + "step": 32362 + }, + { + "epoch": 1.24, + "learning_rate": 2.9281225955373176e-05, + "loss": 0.3656, + "step": 32363 + }, + { + "epoch": 1.24, + "learning_rate": 2.928058476532444e-05, + "loss": 0.4006, + "step": 32364 + }, + { + "epoch": 1.24, + "learning_rate": 2.9279943575275716e-05, + "loss": 0.2828, + "step": 32365 + }, + { + "epoch": 1.24, + "learning_rate": 2.927930238522698e-05, + "loss": 0.3168, + "step": 32366 + }, + { + "epoch": 1.24, + "learning_rate": 2.9278661195178253e-05, + "loss": 0.3017, + "step": 32367 + }, + { + "epoch": 1.24, + "learning_rate": 2.927802000512952e-05, + "loss": 0.3444, + "step": 32368 + }, + { + "epoch": 1.24, + "learning_rate": 2.9277378815080793e-05, + "loss": 0.2764, + "step": 32369 + }, + { + "epoch": 1.24, + "learning_rate": 2.9276737625032064e-05, + "loss": 0.3529, + "step": 32370 + }, + { + "epoch": 1.24, + "learning_rate": 2.927609643498333e-05, + "loss": 0.3437, + "step": 32371 + }, + { + "epoch": 1.24, + "learning_rate": 2.92754552449346e-05, + "loss": 0.3364, + "step": 32372 + }, + { + "epoch": 1.24, + "learning_rate": 2.927481405488587e-05, + "loss": 0.5096, + "step": 32373 + }, + { + "epoch": 1.24, + "learning_rate": 2.927417286483714e-05, + "loss": 0.3652, + "step": 32374 + }, + { + "epoch": 1.24, + "learning_rate": 2.9273531674788406e-05, + "loss": 0.5649, + "step": 32375 + }, + { + "epoch": 1.24, + "learning_rate": 2.9272890484739678e-05, + "loss": 0.4555, + "step": 32376 + }, + { + "epoch": 1.24, + "learning_rate": 2.9272249294690946e-05, + "loss": 0.3917, + "step": 32377 + }, + { + "epoch": 1.24, + "learning_rate": 2.9271608104642218e-05, + "loss": 0.3758, + "step": 32378 + }, + { + "epoch": 1.24, + "learning_rate": 2.927096691459349e-05, + "loss": 0.4441, + "step": 32379 + }, + { + "epoch": 1.24, + "learning_rate": 2.9270325724544754e-05, + "loss": 0.372, + "step": 32380 + }, + { + "epoch": 1.24, + "learning_rate": 2.926968453449603e-05, + "loss": 0.4038, + "step": 32381 + }, + { + "epoch": 1.24, + "learning_rate": 2.9269043344447295e-05, + "loss": 0.363, + "step": 32382 + }, + { + "epoch": 1.25, + "learning_rate": 2.9268402154398566e-05, + "loss": 0.3995, + "step": 32383 + }, + { + "epoch": 1.25, + "learning_rate": 2.9267760964349835e-05, + "loss": 0.4087, + "step": 32384 + }, + { + "epoch": 1.25, + "learning_rate": 2.9267119774301106e-05, + "loss": 0.4394, + "step": 32385 + }, + { + "epoch": 1.25, + "learning_rate": 2.926647858425237e-05, + "loss": 0.5274, + "step": 32386 + }, + { + "epoch": 1.25, + "learning_rate": 2.9265837394203643e-05, + "loss": 0.3615, + "step": 32387 + }, + { + "epoch": 1.25, + "learning_rate": 2.9265196204154915e-05, + "loss": 0.5098, + "step": 32388 + }, + { + "epoch": 1.25, + "learning_rate": 2.9264555014106183e-05, + "loss": 0.3918, + "step": 32389 + }, + { + "epoch": 1.25, + "learning_rate": 2.9263913824057455e-05, + "loss": 0.3889, + "step": 32390 + }, + { + "epoch": 1.25, + "learning_rate": 2.926327263400872e-05, + "loss": 0.4416, + "step": 32391 + }, + { + "epoch": 1.25, + "learning_rate": 2.926263144395999e-05, + "loss": 0.4973, + "step": 32392 + }, + { + "epoch": 1.25, + "learning_rate": 2.926199025391126e-05, + "loss": 0.3488, + "step": 32393 + }, + { + "epoch": 1.25, + "learning_rate": 2.926134906386253e-05, + "loss": 0.4637, + "step": 32394 + }, + { + "epoch": 1.25, + "learning_rate": 2.9260707873813796e-05, + "loss": 0.4627, + "step": 32395 + }, + { + "epoch": 1.25, + "learning_rate": 2.9260066683765068e-05, + "loss": 0.4006, + "step": 32396 + }, + { + "epoch": 1.25, + "learning_rate": 2.925942549371634e-05, + "loss": 0.4032, + "step": 32397 + }, + { + "epoch": 1.25, + "learning_rate": 2.9258784303667608e-05, + "loss": 0.2868, + "step": 32398 + }, + { + "epoch": 1.25, + "learning_rate": 2.925814311361888e-05, + "loss": 0.3755, + "step": 32399 + }, + { + "epoch": 1.25, + "learning_rate": 2.9257501923570145e-05, + "loss": 0.3942, + "step": 32400 + }, + { + "epoch": 1.25, + "learning_rate": 2.925686073352142e-05, + "loss": 0.3743, + "step": 32401 + }, + { + "epoch": 1.25, + "learning_rate": 2.9256219543472685e-05, + "loss": 0.3574, + "step": 32402 + }, + { + "epoch": 1.25, + "learning_rate": 2.9255578353423957e-05, + "loss": 0.2842, + "step": 32403 + }, + { + "epoch": 1.25, + "learning_rate": 2.9254937163375225e-05, + "loss": 0.3823, + "step": 32404 + }, + { + "epoch": 1.25, + "learning_rate": 2.9254295973326497e-05, + "loss": 0.2913, + "step": 32405 + }, + { + "epoch": 1.25, + "learning_rate": 2.9253654783277768e-05, + "loss": 0.3636, + "step": 32406 + }, + { + "epoch": 1.25, + "learning_rate": 2.9253013593229033e-05, + "loss": 0.3564, + "step": 32407 + }, + { + "epoch": 1.25, + "learning_rate": 2.9252372403180305e-05, + "loss": 0.4063, + "step": 32408 + }, + { + "epoch": 1.25, + "learning_rate": 2.9251731213131573e-05, + "loss": 0.4427, + "step": 32409 + }, + { + "epoch": 1.25, + "learning_rate": 2.9251090023082845e-05, + "loss": 0.3794, + "step": 32410 + }, + { + "epoch": 1.25, + "learning_rate": 2.925044883303411e-05, + "loss": 0.3655, + "step": 32411 + }, + { + "epoch": 1.25, + "learning_rate": 2.924980764298538e-05, + "loss": 0.5017, + "step": 32412 + }, + { + "epoch": 1.25, + "learning_rate": 2.924916645293665e-05, + "loss": 0.3457, + "step": 32413 + }, + { + "epoch": 1.25, + "learning_rate": 2.9248525262887922e-05, + "loss": 0.3397, + "step": 32414 + }, + { + "epoch": 1.25, + "learning_rate": 2.9247884072839193e-05, + "loss": 0.3462, + "step": 32415 + }, + { + "epoch": 1.25, + "learning_rate": 2.924724288279046e-05, + "loss": 0.3888, + "step": 32416 + }, + { + "epoch": 1.25, + "learning_rate": 2.9246601692741733e-05, + "loss": 0.331, + "step": 32417 + }, + { + "epoch": 1.25, + "learning_rate": 2.9245960502693e-05, + "loss": 0.2893, + "step": 32418 + }, + { + "epoch": 1.25, + "learning_rate": 2.924531931264427e-05, + "loss": 0.3282, + "step": 32419 + }, + { + "epoch": 1.25, + "learning_rate": 2.924467812259554e-05, + "loss": 0.3982, + "step": 32420 + }, + { + "epoch": 1.25, + "learning_rate": 2.924403693254681e-05, + "loss": 0.3915, + "step": 32421 + }, + { + "epoch": 1.25, + "learning_rate": 2.9243395742498075e-05, + "loss": 0.3558, + "step": 32422 + }, + { + "epoch": 1.25, + "learning_rate": 2.9242754552449347e-05, + "loss": 0.4735, + "step": 32423 + }, + { + "epoch": 1.25, + "learning_rate": 2.924211336240062e-05, + "loss": 0.4102, + "step": 32424 + }, + { + "epoch": 1.25, + "learning_rate": 2.9241472172351887e-05, + "loss": 0.4017, + "step": 32425 + }, + { + "epoch": 1.25, + "learning_rate": 2.924083098230316e-05, + "loss": 0.3215, + "step": 32426 + }, + { + "epoch": 1.25, + "learning_rate": 2.9240189792254424e-05, + "loss": 0.4424, + "step": 32427 + }, + { + "epoch": 1.25, + "learning_rate": 2.9239548602205695e-05, + "loss": 0.4671, + "step": 32428 + }, + { + "epoch": 1.25, + "learning_rate": 2.9238907412156964e-05, + "loss": 0.3157, + "step": 32429 + }, + { + "epoch": 1.25, + "learning_rate": 2.9238266222108235e-05, + "loss": 0.4105, + "step": 32430 + }, + { + "epoch": 1.25, + "learning_rate": 2.92376250320595e-05, + "loss": 0.4621, + "step": 32431 + }, + { + "epoch": 1.25, + "learning_rate": 2.9236983842010772e-05, + "loss": 0.4391, + "step": 32432 + }, + { + "epoch": 1.25, + "learning_rate": 2.9236342651962044e-05, + "loss": 0.3366, + "step": 32433 + }, + { + "epoch": 1.25, + "learning_rate": 2.9235701461913312e-05, + "loss": 0.2631, + "step": 32434 + }, + { + "epoch": 1.25, + "learning_rate": 2.9235060271864584e-05, + "loss": 0.379, + "step": 32435 + }, + { + "epoch": 1.25, + "learning_rate": 2.923441908181585e-05, + "loss": 0.5262, + "step": 32436 + }, + { + "epoch": 1.25, + "learning_rate": 2.9233777891767124e-05, + "loss": 0.3242, + "step": 32437 + }, + { + "epoch": 1.25, + "learning_rate": 2.923313670171839e-05, + "loss": 0.4168, + "step": 32438 + }, + { + "epoch": 1.25, + "learning_rate": 2.923249551166966e-05, + "loss": 0.333, + "step": 32439 + }, + { + "epoch": 1.25, + "learning_rate": 2.923185432162093e-05, + "loss": 0.4029, + "step": 32440 + }, + { + "epoch": 1.25, + "learning_rate": 2.92312131315722e-05, + "loss": 0.37, + "step": 32441 + }, + { + "epoch": 1.25, + "learning_rate": 2.9230571941523472e-05, + "loss": 0.421, + "step": 32442 + }, + { + "epoch": 1.25, + "learning_rate": 2.9229930751474737e-05, + "loss": 0.3344, + "step": 32443 + }, + { + "epoch": 1.25, + "learning_rate": 2.922928956142601e-05, + "loss": 0.4729, + "step": 32444 + }, + { + "epoch": 1.25, + "learning_rate": 2.9228648371377277e-05, + "loss": 0.5364, + "step": 32445 + }, + { + "epoch": 1.25, + "learning_rate": 2.922800718132855e-05, + "loss": 0.4223, + "step": 32446 + }, + { + "epoch": 1.25, + "learning_rate": 2.9227365991279814e-05, + "loss": 0.3867, + "step": 32447 + }, + { + "epoch": 1.25, + "learning_rate": 2.9226724801231086e-05, + "loss": 0.3423, + "step": 32448 + }, + { + "epoch": 1.25, + "learning_rate": 2.9226083611182354e-05, + "loss": 0.4576, + "step": 32449 + }, + { + "epoch": 1.25, + "learning_rate": 2.9225442421133626e-05, + "loss": 0.3046, + "step": 32450 + }, + { + "epoch": 1.25, + "learning_rate": 2.9224801231084897e-05, + "loss": 0.3946, + "step": 32451 + }, + { + "epoch": 1.25, + "learning_rate": 2.9224160041036162e-05, + "loss": 0.3525, + "step": 32452 + }, + { + "epoch": 1.25, + "learning_rate": 2.9223518850987437e-05, + "loss": 0.264, + "step": 32453 + }, + { + "epoch": 1.25, + "learning_rate": 2.9222877660938702e-05, + "loss": 0.319, + "step": 32454 + }, + { + "epoch": 1.25, + "learning_rate": 2.9222236470889974e-05, + "loss": 0.3088, + "step": 32455 + }, + { + "epoch": 1.25, + "learning_rate": 2.9221595280841242e-05, + "loss": 0.3197, + "step": 32456 + }, + { + "epoch": 1.25, + "learning_rate": 2.9220954090792514e-05, + "loss": 0.3497, + "step": 32457 + }, + { + "epoch": 1.25, + "learning_rate": 2.922031290074378e-05, + "loss": 0.4041, + "step": 32458 + }, + { + "epoch": 1.25, + "learning_rate": 2.921967171069505e-05, + "loss": 0.2716, + "step": 32459 + }, + { + "epoch": 1.25, + "learning_rate": 2.9219030520646322e-05, + "loss": 0.4225, + "step": 32460 + }, + { + "epoch": 1.25, + "learning_rate": 2.921838933059759e-05, + "loss": 0.4741, + "step": 32461 + }, + { + "epoch": 1.25, + "learning_rate": 2.9217748140548862e-05, + "loss": 0.4783, + "step": 32462 + }, + { + "epoch": 1.25, + "learning_rate": 2.9217106950500127e-05, + "loss": 0.378, + "step": 32463 + }, + { + "epoch": 1.25, + "learning_rate": 2.92164657604514e-05, + "loss": 0.2756, + "step": 32464 + }, + { + "epoch": 1.25, + "learning_rate": 2.9215824570402667e-05, + "loss": 0.3475, + "step": 32465 + }, + { + "epoch": 1.25, + "learning_rate": 2.921518338035394e-05, + "loss": 0.341, + "step": 32466 + }, + { + "epoch": 1.25, + "learning_rate": 2.9214542190305204e-05, + "loss": 0.3635, + "step": 32467 + }, + { + "epoch": 1.25, + "learning_rate": 2.9213901000256476e-05, + "loss": 0.3994, + "step": 32468 + }, + { + "epoch": 1.25, + "learning_rate": 2.921325981020775e-05, + "loss": 0.3838, + "step": 32469 + }, + { + "epoch": 1.25, + "learning_rate": 2.9212618620159016e-05, + "loss": 0.2735, + "step": 32470 + }, + { + "epoch": 1.25, + "learning_rate": 2.9211977430110288e-05, + "loss": 0.383, + "step": 32471 + }, + { + "epoch": 1.25, + "learning_rate": 2.9211336240061556e-05, + "loss": 0.4409, + "step": 32472 + }, + { + "epoch": 1.25, + "learning_rate": 2.9210695050012828e-05, + "loss": 0.4874, + "step": 32473 + }, + { + "epoch": 1.25, + "learning_rate": 2.9210053859964093e-05, + "loss": 0.3815, + "step": 32474 + }, + { + "epoch": 1.25, + "learning_rate": 2.9209412669915364e-05, + "loss": 0.3457, + "step": 32475 + }, + { + "epoch": 1.25, + "learning_rate": 2.9208771479866633e-05, + "loss": 0.2943, + "step": 32476 + }, + { + "epoch": 1.25, + "learning_rate": 2.9208130289817904e-05, + "loss": 0.301, + "step": 32477 + }, + { + "epoch": 1.25, + "learning_rate": 2.9207489099769176e-05, + "loss": 0.3622, + "step": 32478 + }, + { + "epoch": 1.25, + "learning_rate": 2.920684790972044e-05, + "loss": 0.3707, + "step": 32479 + }, + { + "epoch": 1.25, + "learning_rate": 2.9206206719671713e-05, + "loss": 0.3505, + "step": 32480 + }, + { + "epoch": 1.25, + "learning_rate": 2.920556552962298e-05, + "loss": 0.3486, + "step": 32481 + }, + { + "epoch": 1.25, + "learning_rate": 2.9204924339574253e-05, + "loss": 0.3223, + "step": 32482 + }, + { + "epoch": 1.25, + "learning_rate": 2.9204283149525518e-05, + "loss": 0.2039, + "step": 32483 + }, + { + "epoch": 1.25, + "learning_rate": 2.920364195947679e-05, + "loss": 0.4256, + "step": 32484 + }, + { + "epoch": 1.25, + "learning_rate": 2.9203000769428058e-05, + "loss": 0.3392, + "step": 32485 + }, + { + "epoch": 1.25, + "learning_rate": 2.920235957937933e-05, + "loss": 0.4351, + "step": 32486 + }, + { + "epoch": 1.25, + "learning_rate": 2.92017183893306e-05, + "loss": 0.3106, + "step": 32487 + }, + { + "epoch": 1.25, + "learning_rate": 2.9201077199281866e-05, + "loss": 0.4215, + "step": 32488 + }, + { + "epoch": 1.25, + "learning_rate": 2.920043600923314e-05, + "loss": 0.249, + "step": 32489 + }, + { + "epoch": 1.25, + "learning_rate": 2.9199794819184406e-05, + "loss": 0.2917, + "step": 32490 + }, + { + "epoch": 1.25, + "learning_rate": 2.9199153629135678e-05, + "loss": 0.4787, + "step": 32491 + }, + { + "epoch": 1.25, + "learning_rate": 2.9198512439086946e-05, + "loss": 0.3893, + "step": 32492 + }, + { + "epoch": 1.25, + "learning_rate": 2.9197871249038218e-05, + "loss": 0.3755, + "step": 32493 + }, + { + "epoch": 1.25, + "learning_rate": 2.9197230058989483e-05, + "loss": 0.4016, + "step": 32494 + }, + { + "epoch": 1.25, + "learning_rate": 2.9196588868940755e-05, + "loss": 0.3738, + "step": 32495 + }, + { + "epoch": 1.25, + "learning_rate": 2.9195947678892026e-05, + "loss": 0.5175, + "step": 32496 + }, + { + "epoch": 1.25, + "learning_rate": 2.9195306488843295e-05, + "loss": 0.4355, + "step": 32497 + }, + { + "epoch": 1.25, + "learning_rate": 2.9194665298794566e-05, + "loss": 0.2529, + "step": 32498 + }, + { + "epoch": 1.25, + "learning_rate": 2.919402410874583e-05, + "loss": 0.288, + "step": 32499 + }, + { + "epoch": 1.25, + "learning_rate": 2.9193382918697103e-05, + "loss": 0.3907, + "step": 32500 + }, + { + "epoch": 1.25, + "learning_rate": 2.919274172864837e-05, + "loss": 0.3794, + "step": 32501 + }, + { + "epoch": 1.25, + "learning_rate": 2.9192100538599643e-05, + "loss": 0.3511, + "step": 32502 + }, + { + "epoch": 1.25, + "learning_rate": 2.9191459348550908e-05, + "loss": 0.4038, + "step": 32503 + }, + { + "epoch": 1.25, + "learning_rate": 2.919081815850218e-05, + "loss": 0.6021, + "step": 32504 + }, + { + "epoch": 1.25, + "learning_rate": 2.9190176968453455e-05, + "loss": 0.4031, + "step": 32505 + }, + { + "epoch": 1.25, + "learning_rate": 2.918953577840472e-05, + "loss": 0.3014, + "step": 32506 + }, + { + "epoch": 1.25, + "learning_rate": 2.918889458835599e-05, + "loss": 0.3579, + "step": 32507 + }, + { + "epoch": 1.25, + "learning_rate": 2.918825339830726e-05, + "loss": 0.3708, + "step": 32508 + }, + { + "epoch": 1.25, + "learning_rate": 2.918761220825853e-05, + "loss": 0.4581, + "step": 32509 + }, + { + "epoch": 1.25, + "learning_rate": 2.9186971018209796e-05, + "loss": 0.3693, + "step": 32510 + }, + { + "epoch": 1.25, + "learning_rate": 2.9186329828161068e-05, + "loss": 0.5304, + "step": 32511 + }, + { + "epoch": 1.25, + "learning_rate": 2.918568863811234e-05, + "loss": 0.312, + "step": 32512 + }, + { + "epoch": 1.25, + "learning_rate": 2.9185047448063608e-05, + "loss": 0.341, + "step": 32513 + }, + { + "epoch": 1.25, + "learning_rate": 2.918440625801488e-05, + "loss": 0.3247, + "step": 32514 + }, + { + "epoch": 1.25, + "learning_rate": 2.9183765067966145e-05, + "loss": 0.3348, + "step": 32515 + }, + { + "epoch": 1.25, + "learning_rate": 2.9183123877917417e-05, + "loss": 0.34, + "step": 32516 + }, + { + "epoch": 1.25, + "learning_rate": 2.9182482687868685e-05, + "loss": 0.4271, + "step": 32517 + }, + { + "epoch": 1.25, + "learning_rate": 2.9181841497819957e-05, + "loss": 0.4579, + "step": 32518 + }, + { + "epoch": 1.25, + "learning_rate": 2.918120030777122e-05, + "loss": 0.3278, + "step": 32519 + }, + { + "epoch": 1.25, + "learning_rate": 2.9180559117722493e-05, + "loss": 0.3505, + "step": 32520 + }, + { + "epoch": 1.25, + "learning_rate": 2.9179917927673765e-05, + "loss": 0.3518, + "step": 32521 + }, + { + "epoch": 1.25, + "learning_rate": 2.9179276737625033e-05, + "loss": 0.2863, + "step": 32522 + }, + { + "epoch": 1.25, + "learning_rate": 2.9178635547576305e-05, + "loss": 0.3482, + "step": 32523 + }, + { + "epoch": 1.25, + "learning_rate": 2.917799435752757e-05, + "loss": 0.4296, + "step": 32524 + }, + { + "epoch": 1.25, + "learning_rate": 2.9177353167478845e-05, + "loss": 0.4163, + "step": 32525 + }, + { + "epoch": 1.25, + "learning_rate": 2.917671197743011e-05, + "loss": 0.3855, + "step": 32526 + }, + { + "epoch": 1.25, + "learning_rate": 2.9176070787381382e-05, + "loss": 0.4085, + "step": 32527 + }, + { + "epoch": 1.25, + "learning_rate": 2.917542959733265e-05, + "loss": 0.3348, + "step": 32528 + }, + { + "epoch": 1.25, + "learning_rate": 2.9174788407283922e-05, + "loss": 0.3391, + "step": 32529 + }, + { + "epoch": 1.25, + "learning_rate": 2.9174147217235194e-05, + "loss": 0.4331, + "step": 32530 + }, + { + "epoch": 1.25, + "learning_rate": 2.917350602718646e-05, + "loss": 0.3159, + "step": 32531 + }, + { + "epoch": 1.25, + "learning_rate": 2.917286483713773e-05, + "loss": 0.407, + "step": 32532 + }, + { + "epoch": 1.25, + "learning_rate": 2.9172223647089e-05, + "loss": 0.3998, + "step": 32533 + }, + { + "epoch": 1.25, + "learning_rate": 2.917158245704027e-05, + "loss": 0.5474, + "step": 32534 + }, + { + "epoch": 1.25, + "learning_rate": 2.9170941266991535e-05, + "loss": 0.271, + "step": 32535 + }, + { + "epoch": 1.25, + "learning_rate": 2.9170300076942807e-05, + "loss": 0.3949, + "step": 32536 + }, + { + "epoch": 1.25, + "learning_rate": 2.9169658886894075e-05, + "loss": 0.4497, + "step": 32537 + }, + { + "epoch": 1.25, + "learning_rate": 2.9169017696845347e-05, + "loss": 0.3925, + "step": 32538 + }, + { + "epoch": 1.25, + "learning_rate": 2.916837650679662e-05, + "loss": 0.2389, + "step": 32539 + }, + { + "epoch": 1.25, + "learning_rate": 2.9167735316747884e-05, + "loss": 0.3795, + "step": 32540 + }, + { + "epoch": 1.25, + "learning_rate": 2.916709412669916e-05, + "loss": 0.3759, + "step": 32541 + }, + { + "epoch": 1.25, + "learning_rate": 2.9166452936650424e-05, + "loss": 0.4185, + "step": 32542 + }, + { + "epoch": 1.25, + "learning_rate": 2.9165811746601695e-05, + "loss": 0.4311, + "step": 32543 + }, + { + "epoch": 1.25, + "learning_rate": 2.9165170556552964e-05, + "loss": 0.3687, + "step": 32544 + }, + { + "epoch": 1.25, + "learning_rate": 2.9164529366504235e-05, + "loss": 0.3072, + "step": 32545 + }, + { + "epoch": 1.25, + "learning_rate": 2.91638881764555e-05, + "loss": 0.344, + "step": 32546 + }, + { + "epoch": 1.25, + "learning_rate": 2.9163246986406772e-05, + "loss": 0.3293, + "step": 32547 + }, + { + "epoch": 1.25, + "learning_rate": 2.9162605796358044e-05, + "loss": 0.381, + "step": 32548 + }, + { + "epoch": 1.25, + "learning_rate": 2.9161964606309312e-05, + "loss": 0.388, + "step": 32549 + }, + { + "epoch": 1.25, + "learning_rate": 2.9161323416260584e-05, + "loss": 0.3637, + "step": 32550 + }, + { + "epoch": 1.25, + "learning_rate": 2.916068222621185e-05, + "loss": 0.3527, + "step": 32551 + }, + { + "epoch": 1.25, + "learning_rate": 2.916004103616312e-05, + "loss": 0.409, + "step": 32552 + }, + { + "epoch": 1.25, + "learning_rate": 2.915939984611439e-05, + "loss": 0.5151, + "step": 32553 + }, + { + "epoch": 1.25, + "learning_rate": 2.915875865606566e-05, + "loss": 0.3286, + "step": 32554 + }, + { + "epoch": 1.25, + "learning_rate": 2.9158117466016925e-05, + "loss": 0.5316, + "step": 32555 + }, + { + "epoch": 1.25, + "learning_rate": 2.9157476275968197e-05, + "loss": 0.345, + "step": 32556 + }, + { + "epoch": 1.25, + "learning_rate": 2.915683508591947e-05, + "loss": 0.4487, + "step": 32557 + }, + { + "epoch": 1.25, + "learning_rate": 2.9156193895870737e-05, + "loss": 0.4352, + "step": 32558 + }, + { + "epoch": 1.25, + "learning_rate": 2.915555270582201e-05, + "loss": 0.3739, + "step": 32559 + }, + { + "epoch": 1.25, + "learning_rate": 2.9154911515773274e-05, + "loss": 0.4615, + "step": 32560 + }, + { + "epoch": 1.25, + "learning_rate": 2.915427032572455e-05, + "loss": 0.5132, + "step": 32561 + }, + { + "epoch": 1.25, + "learning_rate": 2.9153629135675814e-05, + "loss": 0.338, + "step": 32562 + }, + { + "epoch": 1.25, + "learning_rate": 2.9152987945627086e-05, + "loss": 0.3109, + "step": 32563 + }, + { + "epoch": 1.25, + "learning_rate": 2.9152346755578354e-05, + "loss": 0.4176, + "step": 32564 + }, + { + "epoch": 1.25, + "learning_rate": 2.9151705565529626e-05, + "loss": 0.3509, + "step": 32565 + }, + { + "epoch": 1.25, + "learning_rate": 2.9151064375480897e-05, + "loss": 0.3389, + "step": 32566 + }, + { + "epoch": 1.25, + "learning_rate": 2.9150423185432162e-05, + "loss": 0.4112, + "step": 32567 + }, + { + "epoch": 1.25, + "learning_rate": 2.9149781995383434e-05, + "loss": 0.4608, + "step": 32568 + }, + { + "epoch": 1.25, + "learning_rate": 2.9149140805334702e-05, + "loss": 0.3265, + "step": 32569 + }, + { + "epoch": 1.25, + "learning_rate": 2.9148499615285974e-05, + "loss": 0.4478, + "step": 32570 + }, + { + "epoch": 1.25, + "learning_rate": 2.914785842523724e-05, + "loss": 0.2357, + "step": 32571 + }, + { + "epoch": 1.25, + "learning_rate": 2.914721723518851e-05, + "loss": 0.3706, + "step": 32572 + }, + { + "epoch": 1.25, + "learning_rate": 2.914657604513978e-05, + "loss": 0.4597, + "step": 32573 + }, + { + "epoch": 1.25, + "learning_rate": 2.914593485509105e-05, + "loss": 0.4862, + "step": 32574 + }, + { + "epoch": 1.25, + "learning_rate": 2.9145293665042322e-05, + "loss": 0.2813, + "step": 32575 + }, + { + "epoch": 1.25, + "learning_rate": 2.9144652474993587e-05, + "loss": 0.3018, + "step": 32576 + }, + { + "epoch": 1.25, + "learning_rate": 2.9144011284944863e-05, + "loss": 0.4247, + "step": 32577 + }, + { + "epoch": 1.25, + "learning_rate": 2.9143370094896127e-05, + "loss": 0.3968, + "step": 32578 + }, + { + "epoch": 1.25, + "learning_rate": 2.91427289048474e-05, + "loss": 0.471, + "step": 32579 + }, + { + "epoch": 1.25, + "learning_rate": 2.9142087714798668e-05, + "loss": 0.3139, + "step": 32580 + }, + { + "epoch": 1.25, + "learning_rate": 2.914144652474994e-05, + "loss": 0.3361, + "step": 32581 + }, + { + "epoch": 1.25, + "learning_rate": 2.9140805334701204e-05, + "loss": 0.3997, + "step": 32582 + }, + { + "epoch": 1.25, + "learning_rate": 2.9140164144652476e-05, + "loss": 0.4064, + "step": 32583 + }, + { + "epoch": 1.25, + "learning_rate": 2.9139522954603748e-05, + "loss": 0.4925, + "step": 32584 + }, + { + "epoch": 1.25, + "learning_rate": 2.9138881764555016e-05, + "loss": 0.3257, + "step": 32585 + }, + { + "epoch": 1.25, + "learning_rate": 2.9138240574506288e-05, + "loss": 0.4299, + "step": 32586 + }, + { + "epoch": 1.25, + "learning_rate": 2.9137599384457553e-05, + "loss": 0.3402, + "step": 32587 + }, + { + "epoch": 1.25, + "learning_rate": 2.9136958194408824e-05, + "loss": 0.2926, + "step": 32588 + }, + { + "epoch": 1.25, + "learning_rate": 2.9136317004360093e-05, + "loss": 0.3247, + "step": 32589 + }, + { + "epoch": 1.25, + "learning_rate": 2.9135675814311364e-05, + "loss": 0.3916, + "step": 32590 + }, + { + "epoch": 1.25, + "learning_rate": 2.913503462426263e-05, + "loss": 0.4193, + "step": 32591 + }, + { + "epoch": 1.25, + "learning_rate": 2.91343934342139e-05, + "loss": 0.3297, + "step": 32592 + }, + { + "epoch": 1.25, + "learning_rate": 2.9133752244165176e-05, + "loss": 0.3037, + "step": 32593 + }, + { + "epoch": 1.25, + "learning_rate": 2.913311105411644e-05, + "loss": 0.4957, + "step": 32594 + }, + { + "epoch": 1.25, + "learning_rate": 2.9132469864067713e-05, + "loss": 0.344, + "step": 32595 + }, + { + "epoch": 1.25, + "learning_rate": 2.913182867401898e-05, + "loss": 0.3054, + "step": 32596 + }, + { + "epoch": 1.25, + "learning_rate": 2.9131187483970253e-05, + "loss": 0.2757, + "step": 32597 + }, + { + "epoch": 1.25, + "learning_rate": 2.9130546293921518e-05, + "loss": 0.4967, + "step": 32598 + }, + { + "epoch": 1.25, + "learning_rate": 2.912990510387279e-05, + "loss": 0.3915, + "step": 32599 + }, + { + "epoch": 1.25, + "learning_rate": 2.9129263913824058e-05, + "loss": 0.3931, + "step": 32600 + }, + { + "epoch": 1.25, + "learning_rate": 2.912862272377533e-05, + "loss": 0.3028, + "step": 32601 + }, + { + "epoch": 1.25, + "learning_rate": 2.91279815337266e-05, + "loss": 0.2674, + "step": 32602 + }, + { + "epoch": 1.25, + "learning_rate": 2.9127340343677866e-05, + "loss": 0.3934, + "step": 32603 + }, + { + "epoch": 1.25, + "learning_rate": 2.9126699153629138e-05, + "loss": 0.4382, + "step": 32604 + }, + { + "epoch": 1.25, + "learning_rate": 2.9126057963580406e-05, + "loss": 0.3908, + "step": 32605 + }, + { + "epoch": 1.25, + "learning_rate": 2.9125416773531678e-05, + "loss": 0.3393, + "step": 32606 + }, + { + "epoch": 1.25, + "learning_rate": 2.9124775583482943e-05, + "loss": 0.4993, + "step": 32607 + }, + { + "epoch": 1.25, + "learning_rate": 2.9124134393434215e-05, + "loss": 0.3919, + "step": 32608 + }, + { + "epoch": 1.25, + "learning_rate": 2.9123493203385483e-05, + "loss": 0.3124, + "step": 32609 + }, + { + "epoch": 1.25, + "learning_rate": 2.9122852013336755e-05, + "loss": 0.2953, + "step": 32610 + }, + { + "epoch": 1.25, + "learning_rate": 2.9122210823288026e-05, + "loss": 0.5378, + "step": 32611 + }, + { + "epoch": 1.25, + "learning_rate": 2.912156963323929e-05, + "loss": 0.3659, + "step": 32612 + }, + { + "epoch": 1.25, + "learning_rate": 2.9120928443190566e-05, + "loss": 0.4488, + "step": 32613 + }, + { + "epoch": 1.25, + "learning_rate": 2.912028725314183e-05, + "loss": 0.3681, + "step": 32614 + }, + { + "epoch": 1.25, + "learning_rate": 2.9119646063093103e-05, + "loss": 0.3472, + "step": 32615 + }, + { + "epoch": 1.25, + "learning_rate": 2.911900487304437e-05, + "loss": 0.302, + "step": 32616 + }, + { + "epoch": 1.25, + "learning_rate": 2.9118363682995643e-05, + "loss": 0.2546, + "step": 32617 + }, + { + "epoch": 1.25, + "learning_rate": 2.9117722492946908e-05, + "loss": 0.3401, + "step": 32618 + }, + { + "epoch": 1.25, + "learning_rate": 2.911708130289818e-05, + "loss": 0.2736, + "step": 32619 + }, + { + "epoch": 1.25, + "learning_rate": 2.911644011284945e-05, + "loss": 0.3751, + "step": 32620 + }, + { + "epoch": 1.25, + "learning_rate": 2.911579892280072e-05, + "loss": 0.4215, + "step": 32621 + }, + { + "epoch": 1.25, + "learning_rate": 2.911515773275199e-05, + "loss": 0.3961, + "step": 32622 + }, + { + "epoch": 1.25, + "learning_rate": 2.9114516542703256e-05, + "loss": 0.3453, + "step": 32623 + }, + { + "epoch": 1.25, + "learning_rate": 2.9113875352654528e-05, + "loss": 0.4015, + "step": 32624 + }, + { + "epoch": 1.25, + "learning_rate": 2.9113234162605797e-05, + "loss": 0.3553, + "step": 32625 + }, + { + "epoch": 1.25, + "learning_rate": 2.9112592972557068e-05, + "loss": 0.3124, + "step": 32626 + }, + { + "epoch": 1.25, + "learning_rate": 2.9111951782508333e-05, + "loss": 0.401, + "step": 32627 + }, + { + "epoch": 1.25, + "learning_rate": 2.9111310592459605e-05, + "loss": 0.4242, + "step": 32628 + }, + { + "epoch": 1.25, + "learning_rate": 2.911066940241088e-05, + "loss": 0.3343, + "step": 32629 + }, + { + "epoch": 1.25, + "learning_rate": 2.9110028212362145e-05, + "loss": 0.3921, + "step": 32630 + }, + { + "epoch": 1.25, + "learning_rate": 2.9109387022313417e-05, + "loss": 0.3565, + "step": 32631 + }, + { + "epoch": 1.25, + "learning_rate": 2.9108745832264685e-05, + "loss": 0.4263, + "step": 32632 + }, + { + "epoch": 1.25, + "learning_rate": 2.9108104642215957e-05, + "loss": 0.4705, + "step": 32633 + }, + { + "epoch": 1.25, + "learning_rate": 2.910746345216722e-05, + "loss": 0.415, + "step": 32634 + }, + { + "epoch": 1.25, + "learning_rate": 2.9106822262118493e-05, + "loss": 0.4428, + "step": 32635 + }, + { + "epoch": 1.25, + "learning_rate": 2.910618107206976e-05, + "loss": 0.3501, + "step": 32636 + }, + { + "epoch": 1.25, + "learning_rate": 2.9105539882021033e-05, + "loss": 0.4971, + "step": 32637 + }, + { + "epoch": 1.25, + "learning_rate": 2.9104898691972305e-05, + "loss": 0.408, + "step": 32638 + }, + { + "epoch": 1.25, + "learning_rate": 2.910425750192357e-05, + "loss": 0.3736, + "step": 32639 + }, + { + "epoch": 1.25, + "learning_rate": 2.9103616311874842e-05, + "loss": 0.2904, + "step": 32640 + }, + { + "epoch": 1.25, + "learning_rate": 2.910297512182611e-05, + "loss": 0.5596, + "step": 32641 + }, + { + "epoch": 1.25, + "learning_rate": 2.9102333931777382e-05, + "loss": 0.3051, + "step": 32642 + }, + { + "epoch": 1.26, + "learning_rate": 2.9101692741728647e-05, + "loss": 0.489, + "step": 32643 + }, + { + "epoch": 1.26, + "learning_rate": 2.910105155167992e-05, + "loss": 0.3762, + "step": 32644 + }, + { + "epoch": 1.26, + "learning_rate": 2.9100410361631187e-05, + "loss": 0.4135, + "step": 32645 + }, + { + "epoch": 1.26, + "learning_rate": 2.909976917158246e-05, + "loss": 0.3359, + "step": 32646 + }, + { + "epoch": 1.26, + "learning_rate": 2.909912798153373e-05, + "loss": 0.4037, + "step": 32647 + }, + { + "epoch": 1.26, + "learning_rate": 2.9098486791484995e-05, + "loss": 0.4501, + "step": 32648 + }, + { + "epoch": 1.26, + "learning_rate": 2.909784560143627e-05, + "loss": 0.3363, + "step": 32649 + }, + { + "epoch": 1.26, + "learning_rate": 2.9097204411387535e-05, + "loss": 0.4635, + "step": 32650 + }, + { + "epoch": 1.26, + "learning_rate": 2.9096563221338807e-05, + "loss": 0.3931, + "step": 32651 + }, + { + "epoch": 1.26, + "learning_rate": 2.9095922031290075e-05, + "loss": 0.3051, + "step": 32652 + }, + { + "epoch": 1.26, + "learning_rate": 2.9095280841241347e-05, + "loss": 0.2713, + "step": 32653 + }, + { + "epoch": 1.26, + "learning_rate": 2.9094639651192612e-05, + "loss": 0.3971, + "step": 32654 + }, + { + "epoch": 1.26, + "learning_rate": 2.9093998461143884e-05, + "loss": 0.3326, + "step": 32655 + }, + { + "epoch": 1.26, + "learning_rate": 2.9093357271095155e-05, + "loss": 0.4095, + "step": 32656 + }, + { + "epoch": 1.26, + "learning_rate": 2.9092716081046424e-05, + "loss": 0.2827, + "step": 32657 + }, + { + "epoch": 1.26, + "learning_rate": 2.9092074890997695e-05, + "loss": 0.368, + "step": 32658 + }, + { + "epoch": 1.26, + "learning_rate": 2.909143370094896e-05, + "loss": 0.3075, + "step": 32659 + }, + { + "epoch": 1.26, + "learning_rate": 2.9090792510900232e-05, + "loss": 0.5206, + "step": 32660 + }, + { + "epoch": 1.26, + "learning_rate": 2.90901513208515e-05, + "loss": 0.2785, + "step": 32661 + }, + { + "epoch": 1.26, + "learning_rate": 2.9089510130802772e-05, + "loss": 0.4363, + "step": 32662 + }, + { + "epoch": 1.26, + "learning_rate": 2.9088868940754037e-05, + "loss": 0.3835, + "step": 32663 + }, + { + "epoch": 1.26, + "learning_rate": 2.908822775070531e-05, + "loss": 0.4399, + "step": 32664 + }, + { + "epoch": 1.26, + "learning_rate": 2.9087586560656584e-05, + "loss": 0.2824, + "step": 32665 + }, + { + "epoch": 1.26, + "learning_rate": 2.908694537060785e-05, + "loss": 0.2508, + "step": 32666 + }, + { + "epoch": 1.26, + "learning_rate": 2.908630418055912e-05, + "loss": 0.3258, + "step": 32667 + }, + { + "epoch": 1.26, + "learning_rate": 2.908566299051039e-05, + "loss": 0.3548, + "step": 32668 + }, + { + "epoch": 1.26, + "learning_rate": 2.908502180046166e-05, + "loss": 0.3827, + "step": 32669 + }, + { + "epoch": 1.26, + "learning_rate": 2.9084380610412926e-05, + "loss": 0.389, + "step": 32670 + }, + { + "epoch": 1.26, + "learning_rate": 2.9083739420364197e-05, + "loss": 0.4574, + "step": 32671 + }, + { + "epoch": 1.26, + "learning_rate": 2.9083098230315466e-05, + "loss": 0.4303, + "step": 32672 + }, + { + "epoch": 1.26, + "learning_rate": 2.9082457040266737e-05, + "loss": 0.4456, + "step": 32673 + }, + { + "epoch": 1.26, + "learning_rate": 2.908181585021801e-05, + "loss": 0.2225, + "step": 32674 + }, + { + "epoch": 1.26, + "learning_rate": 2.9081174660169274e-05, + "loss": 0.3964, + "step": 32675 + }, + { + "epoch": 1.26, + "learning_rate": 2.9080533470120546e-05, + "loss": 0.4364, + "step": 32676 + }, + { + "epoch": 1.26, + "learning_rate": 2.9079892280071814e-05, + "loss": 0.3457, + "step": 32677 + }, + { + "epoch": 1.26, + "learning_rate": 2.9079251090023086e-05, + "loss": 0.4248, + "step": 32678 + }, + { + "epoch": 1.26, + "learning_rate": 2.907860989997435e-05, + "loss": 0.4303, + "step": 32679 + }, + { + "epoch": 1.26, + "learning_rate": 2.9077968709925622e-05, + "loss": 0.4242, + "step": 32680 + }, + { + "epoch": 1.26, + "learning_rate": 2.907732751987689e-05, + "loss": 0.4248, + "step": 32681 + }, + { + "epoch": 1.26, + "learning_rate": 2.9076686329828162e-05, + "loss": 0.3982, + "step": 32682 + }, + { + "epoch": 1.26, + "learning_rate": 2.9076045139779434e-05, + "loss": 0.346, + "step": 32683 + }, + { + "epoch": 1.26, + "learning_rate": 2.90754039497307e-05, + "loss": 0.3984, + "step": 32684 + }, + { + "epoch": 1.26, + "learning_rate": 2.9074762759681974e-05, + "loss": 0.4448, + "step": 32685 + }, + { + "epoch": 1.26, + "learning_rate": 2.907412156963324e-05, + "loss": 0.3288, + "step": 32686 + }, + { + "epoch": 1.26, + "learning_rate": 2.907348037958451e-05, + "loss": 0.4514, + "step": 32687 + }, + { + "epoch": 1.26, + "learning_rate": 2.907283918953578e-05, + "loss": 0.4385, + "step": 32688 + }, + { + "epoch": 1.26, + "learning_rate": 2.907219799948705e-05, + "loss": 0.3944, + "step": 32689 + }, + { + "epoch": 1.26, + "learning_rate": 2.9071556809438316e-05, + "loss": 0.4987, + "step": 32690 + }, + { + "epoch": 1.26, + "learning_rate": 2.9070915619389588e-05, + "loss": 0.4452, + "step": 32691 + }, + { + "epoch": 1.26, + "learning_rate": 2.907027442934086e-05, + "loss": 0.4736, + "step": 32692 + }, + { + "epoch": 1.26, + "learning_rate": 2.9069633239292128e-05, + "loss": 0.2917, + "step": 32693 + }, + { + "epoch": 1.26, + "learning_rate": 2.90689920492434e-05, + "loss": 0.5327, + "step": 32694 + }, + { + "epoch": 1.26, + "learning_rate": 2.9068350859194664e-05, + "loss": 0.3858, + "step": 32695 + }, + { + "epoch": 1.26, + "learning_rate": 2.9067709669145936e-05, + "loss": 0.3818, + "step": 32696 + }, + { + "epoch": 1.26, + "learning_rate": 2.9067068479097204e-05, + "loss": 0.3425, + "step": 32697 + }, + { + "epoch": 1.26, + "learning_rate": 2.9066427289048476e-05, + "loss": 0.4341, + "step": 32698 + }, + { + "epoch": 1.26, + "learning_rate": 2.906578609899974e-05, + "loss": 0.4532, + "step": 32699 + }, + { + "epoch": 1.26, + "learning_rate": 2.9065144908951013e-05, + "loss": 0.4023, + "step": 32700 + }, + { + "epoch": 1.26, + "learning_rate": 2.9064503718902288e-05, + "loss": 0.4412, + "step": 32701 + }, + { + "epoch": 1.26, + "learning_rate": 2.9063862528853553e-05, + "loss": 0.4908, + "step": 32702 + }, + { + "epoch": 1.26, + "learning_rate": 2.9063221338804824e-05, + "loss": 0.3029, + "step": 32703 + }, + { + "epoch": 1.26, + "learning_rate": 2.9062580148756093e-05, + "loss": 0.3165, + "step": 32704 + }, + { + "epoch": 1.26, + "learning_rate": 2.9061938958707364e-05, + "loss": 0.2993, + "step": 32705 + }, + { + "epoch": 1.26, + "learning_rate": 2.906129776865863e-05, + "loss": 0.4051, + "step": 32706 + }, + { + "epoch": 1.26, + "learning_rate": 2.90606565786099e-05, + "loss": 0.4462, + "step": 32707 + }, + { + "epoch": 1.26, + "learning_rate": 2.9060015388561173e-05, + "loss": 0.4219, + "step": 32708 + }, + { + "epoch": 1.26, + "learning_rate": 2.905937419851244e-05, + "loss": 0.3677, + "step": 32709 + }, + { + "epoch": 1.26, + "learning_rate": 2.9058733008463713e-05, + "loss": 0.4249, + "step": 32710 + }, + { + "epoch": 1.26, + "learning_rate": 2.9058091818414978e-05, + "loss": 0.3803, + "step": 32711 + }, + { + "epoch": 1.26, + "learning_rate": 2.905745062836625e-05, + "loss": 0.5284, + "step": 32712 + }, + { + "epoch": 1.26, + "learning_rate": 2.9056809438317518e-05, + "loss": 0.3933, + "step": 32713 + }, + { + "epoch": 1.26, + "learning_rate": 2.905616824826879e-05, + "loss": 0.4036, + "step": 32714 + }, + { + "epoch": 1.26, + "learning_rate": 2.9055527058220055e-05, + "loss": 0.3491, + "step": 32715 + }, + { + "epoch": 1.26, + "learning_rate": 2.9054885868171326e-05, + "loss": 0.336, + "step": 32716 + }, + { + "epoch": 1.26, + "learning_rate": 2.90542446781226e-05, + "loss": 0.3855, + "step": 32717 + }, + { + "epoch": 1.26, + "learning_rate": 2.9053603488073866e-05, + "loss": 0.4159, + "step": 32718 + }, + { + "epoch": 1.26, + "learning_rate": 2.9052962298025138e-05, + "loss": 0.3243, + "step": 32719 + }, + { + "epoch": 1.26, + "learning_rate": 2.9052321107976403e-05, + "loss": 0.404, + "step": 32720 + }, + { + "epoch": 1.26, + "learning_rate": 2.9051679917927678e-05, + "loss": 0.4074, + "step": 32721 + }, + { + "epoch": 1.26, + "learning_rate": 2.9051038727878943e-05, + "loss": 0.3121, + "step": 32722 + }, + { + "epoch": 1.26, + "learning_rate": 2.9050397537830215e-05, + "loss": 0.2945, + "step": 32723 + }, + { + "epoch": 1.26, + "learning_rate": 2.9049756347781483e-05, + "loss": 0.3683, + "step": 32724 + }, + { + "epoch": 1.26, + "learning_rate": 2.9049115157732755e-05, + "loss": 0.3163, + "step": 32725 + }, + { + "epoch": 1.26, + "learning_rate": 2.9048473967684026e-05, + "loss": 0.4403, + "step": 32726 + }, + { + "epoch": 1.26, + "learning_rate": 2.904783277763529e-05, + "loss": 0.3772, + "step": 32727 + }, + { + "epoch": 1.26, + "learning_rate": 2.9047191587586563e-05, + "loss": 0.4354, + "step": 32728 + }, + { + "epoch": 1.26, + "learning_rate": 2.904655039753783e-05, + "loss": 0.4389, + "step": 32729 + }, + { + "epoch": 1.26, + "learning_rate": 2.9045909207489103e-05, + "loss": 0.4161, + "step": 32730 + }, + { + "epoch": 1.26, + "learning_rate": 2.9045268017440368e-05, + "loss": 0.3876, + "step": 32731 + }, + { + "epoch": 1.26, + "learning_rate": 2.904462682739164e-05, + "loss": 0.4342, + "step": 32732 + }, + { + "epoch": 1.26, + "learning_rate": 2.9043985637342908e-05, + "loss": 0.3253, + "step": 32733 + }, + { + "epoch": 1.26, + "learning_rate": 2.904334444729418e-05, + "loss": 0.3907, + "step": 32734 + }, + { + "epoch": 1.26, + "learning_rate": 2.904270325724545e-05, + "loss": 0.3537, + "step": 32735 + }, + { + "epoch": 1.26, + "learning_rate": 2.9042062067196717e-05, + "loss": 0.5008, + "step": 32736 + }, + { + "epoch": 1.26, + "learning_rate": 2.904142087714799e-05, + "loss": 0.4065, + "step": 32737 + }, + { + "epoch": 1.26, + "learning_rate": 2.9040779687099257e-05, + "loss": 0.3329, + "step": 32738 + }, + { + "epoch": 1.26, + "learning_rate": 2.9040138497050528e-05, + "loss": 0.3107, + "step": 32739 + }, + { + "epoch": 1.26, + "learning_rate": 2.9039497307001797e-05, + "loss": 0.3356, + "step": 32740 + }, + { + "epoch": 1.26, + "learning_rate": 2.903885611695307e-05, + "loss": 0.3439, + "step": 32741 + }, + { + "epoch": 1.26, + "learning_rate": 2.9038214926904333e-05, + "loss": 0.3643, + "step": 32742 + }, + { + "epoch": 1.26, + "learning_rate": 2.9037573736855605e-05, + "loss": 0.3637, + "step": 32743 + }, + { + "epoch": 1.26, + "learning_rate": 2.9036932546806877e-05, + "loss": 0.3806, + "step": 32744 + }, + { + "epoch": 1.26, + "learning_rate": 2.9036291356758145e-05, + "loss": 0.4619, + "step": 32745 + }, + { + "epoch": 1.26, + "learning_rate": 2.9035650166709417e-05, + "loss": 0.3024, + "step": 32746 + }, + { + "epoch": 1.26, + "learning_rate": 2.903500897666068e-05, + "loss": 0.2781, + "step": 32747 + }, + { + "epoch": 1.26, + "learning_rate": 2.9034367786611953e-05, + "loss": 0.4821, + "step": 32748 + }, + { + "epoch": 1.26, + "learning_rate": 2.9033726596563222e-05, + "loss": 0.375, + "step": 32749 + }, + { + "epoch": 1.26, + "learning_rate": 2.9033085406514493e-05, + "loss": 0.3696, + "step": 32750 + }, + { + "epoch": 1.26, + "learning_rate": 2.903244421646576e-05, + "loss": 0.4044, + "step": 32751 + }, + { + "epoch": 1.26, + "learning_rate": 2.903180302641703e-05, + "loss": 0.4116, + "step": 32752 + }, + { + "epoch": 1.26, + "learning_rate": 2.9031161836368305e-05, + "loss": 0.4744, + "step": 32753 + }, + { + "epoch": 1.26, + "learning_rate": 2.903052064631957e-05, + "loss": 0.4129, + "step": 32754 + }, + { + "epoch": 1.26, + "learning_rate": 2.9029879456270842e-05, + "loss": 0.5416, + "step": 32755 + }, + { + "epoch": 1.26, + "learning_rate": 2.902923826622211e-05, + "loss": 0.4436, + "step": 32756 + }, + { + "epoch": 1.26, + "learning_rate": 2.9028597076173382e-05, + "loss": 0.3493, + "step": 32757 + }, + { + "epoch": 1.26, + "learning_rate": 2.9027955886124647e-05, + "loss": 0.4256, + "step": 32758 + }, + { + "epoch": 1.26, + "learning_rate": 2.902731469607592e-05, + "loss": 0.3531, + "step": 32759 + }, + { + "epoch": 1.26, + "learning_rate": 2.9026673506027187e-05, + "loss": 0.3232, + "step": 32760 + }, + { + "epoch": 1.26, + "learning_rate": 2.902603231597846e-05, + "loss": 0.3522, + "step": 32761 + }, + { + "epoch": 1.26, + "learning_rate": 2.902539112592973e-05, + "loss": 0.5148, + "step": 32762 + }, + { + "epoch": 1.26, + "learning_rate": 2.9024749935880995e-05, + "loss": 0.3443, + "step": 32763 + }, + { + "epoch": 1.26, + "learning_rate": 2.9024108745832267e-05, + "loss": 0.3181, + "step": 32764 + }, + { + "epoch": 1.26, + "learning_rate": 2.9023467555783535e-05, + "loss": 0.4934, + "step": 32765 + }, + { + "epoch": 1.26, + "learning_rate": 2.9022826365734807e-05, + "loss": 0.3546, + "step": 32766 + }, + { + "epoch": 1.26, + "learning_rate": 2.9022185175686072e-05, + "loss": 0.4273, + "step": 32767 + }, + { + "epoch": 1.26, + "learning_rate": 2.9021543985637344e-05, + "loss": 0.3484, + "step": 32768 + }, + { + "epoch": 1.26, + "learning_rate": 2.9020902795588612e-05, + "loss": 0.4002, + "step": 32769 + }, + { + "epoch": 1.26, + "learning_rate": 2.9020261605539884e-05, + "loss": 0.4134, + "step": 32770 + }, + { + "epoch": 1.26, + "learning_rate": 2.9019620415491155e-05, + "loss": 0.3139, + "step": 32771 + }, + { + "epoch": 1.26, + "learning_rate": 2.901897922544242e-05, + "loss": 0.3137, + "step": 32772 + }, + { + "epoch": 1.26, + "learning_rate": 2.9018338035393695e-05, + "loss": 0.3835, + "step": 32773 + }, + { + "epoch": 1.26, + "learning_rate": 2.901769684534496e-05, + "loss": 0.3462, + "step": 32774 + }, + { + "epoch": 1.26, + "learning_rate": 2.9017055655296232e-05, + "loss": 0.38, + "step": 32775 + }, + { + "epoch": 1.26, + "learning_rate": 2.90164144652475e-05, + "loss": 0.3488, + "step": 32776 + }, + { + "epoch": 1.26, + "learning_rate": 2.9015773275198772e-05, + "loss": 0.3972, + "step": 32777 + }, + { + "epoch": 1.26, + "learning_rate": 2.9015132085150037e-05, + "loss": 0.4147, + "step": 32778 + }, + { + "epoch": 1.26, + "learning_rate": 2.901449089510131e-05, + "loss": 0.428, + "step": 32779 + }, + { + "epoch": 1.26, + "learning_rate": 2.901384970505258e-05, + "loss": 0.3208, + "step": 32780 + }, + { + "epoch": 1.26, + "learning_rate": 2.901320851500385e-05, + "loss": 0.4915, + "step": 32781 + }, + { + "epoch": 1.26, + "learning_rate": 2.901256732495512e-05, + "loss": 0.3055, + "step": 32782 + }, + { + "epoch": 1.26, + "learning_rate": 2.9011926134906386e-05, + "loss": 0.4261, + "step": 32783 + }, + { + "epoch": 1.26, + "learning_rate": 2.9011284944857657e-05, + "loss": 0.294, + "step": 32784 + }, + { + "epoch": 1.26, + "learning_rate": 2.9010643754808926e-05, + "loss": 0.3223, + "step": 32785 + }, + { + "epoch": 1.26, + "learning_rate": 2.9010002564760197e-05, + "loss": 0.3887, + "step": 32786 + }, + { + "epoch": 1.26, + "learning_rate": 2.9009361374711462e-05, + "loss": 0.4006, + "step": 32787 + }, + { + "epoch": 1.26, + "learning_rate": 2.9008720184662734e-05, + "loss": 0.3564, + "step": 32788 + }, + { + "epoch": 1.26, + "learning_rate": 2.900807899461401e-05, + "loss": 0.3747, + "step": 32789 + }, + { + "epoch": 1.26, + "learning_rate": 2.9007437804565274e-05, + "loss": 0.5433, + "step": 32790 + }, + { + "epoch": 1.26, + "learning_rate": 2.9006796614516546e-05, + "loss": 0.2932, + "step": 32791 + }, + { + "epoch": 1.26, + "learning_rate": 2.9006155424467814e-05, + "loss": 0.4008, + "step": 32792 + }, + { + "epoch": 1.26, + "learning_rate": 2.9005514234419086e-05, + "loss": 0.2572, + "step": 32793 + }, + { + "epoch": 1.26, + "learning_rate": 2.900487304437035e-05, + "loss": 0.4277, + "step": 32794 + }, + { + "epoch": 1.26, + "learning_rate": 2.9004231854321622e-05, + "loss": 0.5326, + "step": 32795 + }, + { + "epoch": 1.26, + "learning_rate": 2.900359066427289e-05, + "loss": 0.4989, + "step": 32796 + }, + { + "epoch": 1.26, + "learning_rate": 2.9002949474224162e-05, + "loss": 0.3191, + "step": 32797 + }, + { + "epoch": 1.26, + "learning_rate": 2.9002308284175434e-05, + "loss": 0.5248, + "step": 32798 + }, + { + "epoch": 1.26, + "learning_rate": 2.90016670941267e-05, + "loss": 0.2862, + "step": 32799 + }, + { + "epoch": 1.26, + "learning_rate": 2.900102590407797e-05, + "loss": 0.3485, + "step": 32800 + }, + { + "epoch": 1.26, + "learning_rate": 2.900038471402924e-05, + "loss": 0.387, + "step": 32801 + }, + { + "epoch": 1.26, + "learning_rate": 2.899974352398051e-05, + "loss": 0.3906, + "step": 32802 + }, + { + "epoch": 1.26, + "learning_rate": 2.8999102333931776e-05, + "loss": 0.3917, + "step": 32803 + }, + { + "epoch": 1.26, + "learning_rate": 2.8998461143883048e-05, + "loss": 0.426, + "step": 32804 + }, + { + "epoch": 1.26, + "learning_rate": 2.8997819953834316e-05, + "loss": 0.2524, + "step": 32805 + }, + { + "epoch": 1.26, + "learning_rate": 2.8997178763785588e-05, + "loss": 0.4888, + "step": 32806 + }, + { + "epoch": 1.26, + "learning_rate": 2.899653757373686e-05, + "loss": 0.406, + "step": 32807 + }, + { + "epoch": 1.26, + "learning_rate": 2.8995896383688124e-05, + "loss": 0.5147, + "step": 32808 + }, + { + "epoch": 1.26, + "learning_rate": 2.89952551936394e-05, + "loss": 0.3836, + "step": 32809 + }, + { + "epoch": 1.26, + "learning_rate": 2.8994614003590664e-05, + "loss": 0.4905, + "step": 32810 + }, + { + "epoch": 1.26, + "learning_rate": 2.8993972813541936e-05, + "loss": 0.5384, + "step": 32811 + }, + { + "epoch": 1.26, + "learning_rate": 2.8993331623493204e-05, + "loss": 0.4756, + "step": 32812 + }, + { + "epoch": 1.26, + "learning_rate": 2.8992690433444476e-05, + "loss": 0.3128, + "step": 32813 + }, + { + "epoch": 1.26, + "learning_rate": 2.899204924339574e-05, + "loss": 0.3276, + "step": 32814 + }, + { + "epoch": 1.26, + "learning_rate": 2.8991408053347013e-05, + "loss": 0.3424, + "step": 32815 + }, + { + "epoch": 1.26, + "learning_rate": 2.8990766863298284e-05, + "loss": 0.2924, + "step": 32816 + }, + { + "epoch": 1.26, + "learning_rate": 2.8990125673249553e-05, + "loss": 0.3449, + "step": 32817 + }, + { + "epoch": 1.26, + "learning_rate": 2.8989484483200824e-05, + "loss": 0.3023, + "step": 32818 + }, + { + "epoch": 1.26, + "learning_rate": 2.898884329315209e-05, + "loss": 0.3752, + "step": 32819 + }, + { + "epoch": 1.26, + "learning_rate": 2.898820210310336e-05, + "loss": 0.3804, + "step": 32820 + }, + { + "epoch": 1.26, + "learning_rate": 2.898756091305463e-05, + "loss": 0.3825, + "step": 32821 + }, + { + "epoch": 1.26, + "learning_rate": 2.89869197230059e-05, + "loss": 0.2982, + "step": 32822 + }, + { + "epoch": 1.26, + "learning_rate": 2.8986278532957166e-05, + "loss": 0.3206, + "step": 32823 + }, + { + "epoch": 1.26, + "learning_rate": 2.8985637342908438e-05, + "loss": 0.3604, + "step": 32824 + }, + { + "epoch": 1.26, + "learning_rate": 2.8984996152859713e-05, + "loss": 0.4412, + "step": 32825 + }, + { + "epoch": 1.26, + "learning_rate": 2.8984354962810978e-05, + "loss": 0.3393, + "step": 32826 + }, + { + "epoch": 1.26, + "learning_rate": 2.898371377276225e-05, + "loss": 0.5138, + "step": 32827 + }, + { + "epoch": 1.26, + "learning_rate": 2.8983072582713518e-05, + "loss": 0.3063, + "step": 32828 + }, + { + "epoch": 1.26, + "learning_rate": 2.898243139266479e-05, + "loss": 0.4709, + "step": 32829 + }, + { + "epoch": 1.26, + "learning_rate": 2.8981790202616055e-05, + "loss": 0.4191, + "step": 32830 + }, + { + "epoch": 1.26, + "learning_rate": 2.8981149012567326e-05, + "loss": 0.2436, + "step": 32831 + }, + { + "epoch": 1.26, + "learning_rate": 2.8980507822518595e-05, + "loss": 0.3518, + "step": 32832 + }, + { + "epoch": 1.26, + "learning_rate": 2.8979866632469866e-05, + "loss": 0.3613, + "step": 32833 + }, + { + "epoch": 1.26, + "learning_rate": 2.8979225442421138e-05, + "loss": 0.4777, + "step": 32834 + }, + { + "epoch": 1.26, + "learning_rate": 2.8978584252372403e-05, + "loss": 0.4338, + "step": 32835 + }, + { + "epoch": 1.26, + "learning_rate": 2.8977943062323675e-05, + "loss": 0.3943, + "step": 32836 + }, + { + "epoch": 1.26, + "learning_rate": 2.8977301872274943e-05, + "loss": 0.4853, + "step": 32837 + }, + { + "epoch": 1.26, + "learning_rate": 2.8976660682226215e-05, + "loss": 0.2772, + "step": 32838 + }, + { + "epoch": 1.26, + "learning_rate": 2.897601949217748e-05, + "loss": 0.4125, + "step": 32839 + }, + { + "epoch": 1.26, + "learning_rate": 2.897537830212875e-05, + "loss": 0.3769, + "step": 32840 + }, + { + "epoch": 1.26, + "learning_rate": 2.897473711208002e-05, + "loss": 0.474, + "step": 32841 + }, + { + "epoch": 1.26, + "learning_rate": 2.897409592203129e-05, + "loss": 0.4068, + "step": 32842 + }, + { + "epoch": 1.26, + "learning_rate": 2.8973454731982563e-05, + "loss": 0.3196, + "step": 32843 + }, + { + "epoch": 1.26, + "learning_rate": 2.8972813541933828e-05, + "loss": 0.3628, + "step": 32844 + }, + { + "epoch": 1.26, + "learning_rate": 2.8972172351885103e-05, + "loss": 0.4298, + "step": 32845 + }, + { + "epoch": 1.26, + "learning_rate": 2.8971531161836368e-05, + "loss": 0.4401, + "step": 32846 + }, + { + "epoch": 1.26, + "learning_rate": 2.897088997178764e-05, + "loss": 0.2881, + "step": 32847 + }, + { + "epoch": 1.26, + "learning_rate": 2.8970248781738908e-05, + "loss": 0.308, + "step": 32848 + }, + { + "epoch": 1.26, + "learning_rate": 2.896960759169018e-05, + "loss": 0.4263, + "step": 32849 + }, + { + "epoch": 1.26, + "learning_rate": 2.8968966401641445e-05, + "loss": 0.4725, + "step": 32850 + }, + { + "epoch": 1.26, + "learning_rate": 2.8968325211592717e-05, + "loss": 0.4852, + "step": 32851 + }, + { + "epoch": 1.26, + "learning_rate": 2.8967684021543988e-05, + "loss": 0.3726, + "step": 32852 + }, + { + "epoch": 1.26, + "learning_rate": 2.8967042831495257e-05, + "loss": 0.5191, + "step": 32853 + }, + { + "epoch": 1.26, + "learning_rate": 2.896640164144653e-05, + "loss": 0.3742, + "step": 32854 + }, + { + "epoch": 1.26, + "learning_rate": 2.8965760451397793e-05, + "loss": 0.3865, + "step": 32855 + }, + { + "epoch": 1.26, + "learning_rate": 2.8965119261349065e-05, + "loss": 0.4337, + "step": 32856 + }, + { + "epoch": 1.26, + "learning_rate": 2.8964478071300333e-05, + "loss": 0.4598, + "step": 32857 + }, + { + "epoch": 1.26, + "learning_rate": 2.8963836881251605e-05, + "loss": 0.3956, + "step": 32858 + }, + { + "epoch": 1.26, + "learning_rate": 2.896319569120287e-05, + "loss": 0.3971, + "step": 32859 + }, + { + "epoch": 1.26, + "learning_rate": 2.8962554501154142e-05, + "loss": 0.305, + "step": 32860 + }, + { + "epoch": 1.26, + "learning_rate": 2.8961913311105417e-05, + "loss": 0.516, + "step": 32861 + }, + { + "epoch": 1.26, + "learning_rate": 2.8961272121056682e-05, + "loss": 0.4959, + "step": 32862 + }, + { + "epoch": 1.26, + "learning_rate": 2.8960630931007953e-05, + "loss": 0.3327, + "step": 32863 + }, + { + "epoch": 1.26, + "learning_rate": 2.8959989740959222e-05, + "loss": 0.4587, + "step": 32864 + }, + { + "epoch": 1.26, + "learning_rate": 2.8959348550910494e-05, + "loss": 0.4083, + "step": 32865 + }, + { + "epoch": 1.26, + "learning_rate": 2.895870736086176e-05, + "loss": 0.4145, + "step": 32866 + }, + { + "epoch": 1.26, + "learning_rate": 2.895806617081303e-05, + "loss": 0.3557, + "step": 32867 + }, + { + "epoch": 1.26, + "learning_rate": 2.89574249807643e-05, + "loss": 0.4872, + "step": 32868 + }, + { + "epoch": 1.26, + "learning_rate": 2.895678379071557e-05, + "loss": 0.3084, + "step": 32869 + }, + { + "epoch": 1.26, + "learning_rate": 2.8956142600666842e-05, + "loss": 0.3991, + "step": 32870 + }, + { + "epoch": 1.26, + "learning_rate": 2.8955501410618107e-05, + "loss": 0.4, + "step": 32871 + }, + { + "epoch": 1.26, + "learning_rate": 2.895486022056938e-05, + "loss": 0.3527, + "step": 32872 + }, + { + "epoch": 1.26, + "learning_rate": 2.8954219030520647e-05, + "loss": 0.4191, + "step": 32873 + }, + { + "epoch": 1.26, + "learning_rate": 2.895357784047192e-05, + "loss": 0.3603, + "step": 32874 + }, + { + "epoch": 1.26, + "learning_rate": 2.8952936650423184e-05, + "loss": 0.3696, + "step": 32875 + }, + { + "epoch": 1.26, + "learning_rate": 2.8952295460374455e-05, + "loss": 0.3558, + "step": 32876 + }, + { + "epoch": 1.26, + "learning_rate": 2.8951654270325724e-05, + "loss": 0.4542, + "step": 32877 + }, + { + "epoch": 1.26, + "learning_rate": 2.8951013080276995e-05, + "loss": 0.4208, + "step": 32878 + }, + { + "epoch": 1.26, + "learning_rate": 2.8950371890228267e-05, + "loss": 0.5567, + "step": 32879 + }, + { + "epoch": 1.26, + "learning_rate": 2.8949730700179535e-05, + "loss": 0.3124, + "step": 32880 + }, + { + "epoch": 1.26, + "learning_rate": 2.8949089510130807e-05, + "loss": 0.4518, + "step": 32881 + }, + { + "epoch": 1.26, + "learning_rate": 2.8948448320082072e-05, + "loss": 0.4097, + "step": 32882 + }, + { + "epoch": 1.26, + "learning_rate": 2.8947807130033344e-05, + "loss": 0.3646, + "step": 32883 + }, + { + "epoch": 1.26, + "learning_rate": 2.8947165939984612e-05, + "loss": 0.3654, + "step": 32884 + }, + { + "epoch": 1.26, + "learning_rate": 2.8946524749935884e-05, + "loss": 0.4225, + "step": 32885 + }, + { + "epoch": 1.26, + "learning_rate": 2.894588355988715e-05, + "loss": 0.4935, + "step": 32886 + }, + { + "epoch": 1.26, + "learning_rate": 2.894524236983842e-05, + "loss": 0.4379, + "step": 32887 + }, + { + "epoch": 1.26, + "learning_rate": 2.8944601179789692e-05, + "loss": 0.319, + "step": 32888 + }, + { + "epoch": 1.26, + "learning_rate": 2.894395998974096e-05, + "loss": 0.3755, + "step": 32889 + }, + { + "epoch": 1.26, + "learning_rate": 2.8943318799692232e-05, + "loss": 0.2802, + "step": 32890 + }, + { + "epoch": 1.26, + "learning_rate": 2.8942677609643497e-05, + "loss": 0.317, + "step": 32891 + }, + { + "epoch": 1.26, + "learning_rate": 2.894203641959477e-05, + "loss": 0.3824, + "step": 32892 + }, + { + "epoch": 1.26, + "learning_rate": 2.8941395229546037e-05, + "loss": 0.441, + "step": 32893 + }, + { + "epoch": 1.26, + "learning_rate": 2.894075403949731e-05, + "loss": 0.4145, + "step": 32894 + }, + { + "epoch": 1.26, + "learning_rate": 2.894011284944858e-05, + "loss": 0.4738, + "step": 32895 + }, + { + "epoch": 1.26, + "learning_rate": 2.8939471659399846e-05, + "loss": 0.3955, + "step": 32896 + }, + { + "epoch": 1.26, + "learning_rate": 2.893883046935112e-05, + "loss": 0.2901, + "step": 32897 + }, + { + "epoch": 1.26, + "learning_rate": 2.8938189279302386e-05, + "loss": 0.3757, + "step": 32898 + }, + { + "epoch": 1.26, + "learning_rate": 2.8937548089253657e-05, + "loss": 0.3376, + "step": 32899 + }, + { + "epoch": 1.26, + "learning_rate": 2.8936906899204926e-05, + "loss": 0.3437, + "step": 32900 + }, + { + "epoch": 1.26, + "learning_rate": 2.8936265709156197e-05, + "loss": 0.3145, + "step": 32901 + }, + { + "epoch": 1.26, + "learning_rate": 2.8935624519107462e-05, + "loss": 0.4113, + "step": 32902 + }, + { + "epoch": 1.27, + "learning_rate": 2.8934983329058734e-05, + "loss": 0.391, + "step": 32903 + }, + { + "epoch": 1.27, + "learning_rate": 2.8934342139010006e-05, + "loss": 0.4059, + "step": 32904 + }, + { + "epoch": 1.27, + "learning_rate": 2.8933700948961274e-05, + "loss": 0.4727, + "step": 32905 + }, + { + "epoch": 1.27, + "learning_rate": 2.8933059758912546e-05, + "loss": 0.4032, + "step": 32906 + }, + { + "epoch": 1.27, + "learning_rate": 2.893241856886381e-05, + "loss": 0.4747, + "step": 32907 + }, + { + "epoch": 1.27, + "learning_rate": 2.8931777378815082e-05, + "loss": 0.3348, + "step": 32908 + }, + { + "epoch": 1.27, + "learning_rate": 2.893113618876635e-05, + "loss": 0.5819, + "step": 32909 + }, + { + "epoch": 1.27, + "learning_rate": 2.8930494998717623e-05, + "loss": 0.3766, + "step": 32910 + }, + { + "epoch": 1.27, + "learning_rate": 2.8929853808668887e-05, + "loss": 0.3296, + "step": 32911 + }, + { + "epoch": 1.27, + "learning_rate": 2.892921261862016e-05, + "loss": 0.4062, + "step": 32912 + }, + { + "epoch": 1.27, + "learning_rate": 2.8928571428571434e-05, + "loss": 0.2886, + "step": 32913 + }, + { + "epoch": 1.27, + "learning_rate": 2.89279302385227e-05, + "loss": 0.3931, + "step": 32914 + }, + { + "epoch": 1.27, + "learning_rate": 2.892728904847397e-05, + "loss": 0.4952, + "step": 32915 + }, + { + "epoch": 1.27, + "learning_rate": 2.892664785842524e-05, + "loss": 0.44, + "step": 32916 + }, + { + "epoch": 1.27, + "learning_rate": 2.892600666837651e-05, + "loss": 0.461, + "step": 32917 + }, + { + "epoch": 1.27, + "learning_rate": 2.8925365478327776e-05, + "loss": 0.4039, + "step": 32918 + }, + { + "epoch": 1.27, + "learning_rate": 2.8924724288279048e-05, + "loss": 0.2249, + "step": 32919 + }, + { + "epoch": 1.27, + "learning_rate": 2.8924083098230316e-05, + "loss": 0.2776, + "step": 32920 + }, + { + "epoch": 1.27, + "learning_rate": 2.8923441908181588e-05, + "loss": 0.4401, + "step": 32921 + }, + { + "epoch": 1.27, + "learning_rate": 2.892280071813286e-05, + "loss": 0.3472, + "step": 32922 + }, + { + "epoch": 1.27, + "learning_rate": 2.8922159528084124e-05, + "loss": 0.4387, + "step": 32923 + }, + { + "epoch": 1.27, + "learning_rate": 2.8921518338035396e-05, + "loss": 0.3144, + "step": 32924 + }, + { + "epoch": 1.27, + "learning_rate": 2.8920877147986664e-05, + "loss": 0.3978, + "step": 32925 + }, + { + "epoch": 1.27, + "learning_rate": 2.8920235957937936e-05, + "loss": 0.3885, + "step": 32926 + }, + { + "epoch": 1.27, + "learning_rate": 2.89195947678892e-05, + "loss": 0.3814, + "step": 32927 + }, + { + "epoch": 1.27, + "learning_rate": 2.8918953577840473e-05, + "loss": 0.3344, + "step": 32928 + }, + { + "epoch": 1.27, + "learning_rate": 2.891831238779174e-05, + "loss": 0.455, + "step": 32929 + }, + { + "epoch": 1.27, + "learning_rate": 2.8917671197743013e-05, + "loss": 0.3722, + "step": 32930 + }, + { + "epoch": 1.27, + "learning_rate": 2.8917030007694285e-05, + "loss": 0.318, + "step": 32931 + }, + { + "epoch": 1.27, + "learning_rate": 2.891638881764555e-05, + "loss": 0.5749, + "step": 32932 + }, + { + "epoch": 1.27, + "learning_rate": 2.8915747627596825e-05, + "loss": 0.3067, + "step": 32933 + }, + { + "epoch": 1.27, + "learning_rate": 2.891510643754809e-05, + "loss": 0.3184, + "step": 32934 + }, + { + "epoch": 1.27, + "learning_rate": 2.891446524749936e-05, + "loss": 0.3528, + "step": 32935 + }, + { + "epoch": 1.27, + "learning_rate": 2.891382405745063e-05, + "loss": 0.2425, + "step": 32936 + }, + { + "epoch": 1.27, + "learning_rate": 2.89131828674019e-05, + "loss": 0.3585, + "step": 32937 + }, + { + "epoch": 1.27, + "learning_rate": 2.8912541677353166e-05, + "loss": 0.4465, + "step": 32938 + }, + { + "epoch": 1.27, + "learning_rate": 2.8911900487304438e-05, + "loss": 0.4554, + "step": 32939 + }, + { + "epoch": 1.27, + "learning_rate": 2.891125929725571e-05, + "loss": 0.3576, + "step": 32940 + }, + { + "epoch": 1.27, + "learning_rate": 2.8910618107206978e-05, + "loss": 0.3109, + "step": 32941 + }, + { + "epoch": 1.27, + "learning_rate": 2.890997691715825e-05, + "loss": 0.3959, + "step": 32942 + }, + { + "epoch": 1.27, + "learning_rate": 2.8909335727109515e-05, + "loss": 0.3622, + "step": 32943 + }, + { + "epoch": 1.27, + "learning_rate": 2.8908694537060786e-05, + "loss": 0.4537, + "step": 32944 + }, + { + "epoch": 1.27, + "learning_rate": 2.8908053347012055e-05, + "loss": 0.4288, + "step": 32945 + }, + { + "epoch": 1.27, + "learning_rate": 2.8907412156963326e-05, + "loss": 0.4787, + "step": 32946 + }, + { + "epoch": 1.27, + "learning_rate": 2.890677096691459e-05, + "loss": 0.3562, + "step": 32947 + }, + { + "epoch": 1.27, + "learning_rate": 2.8906129776865863e-05, + "loss": 0.4365, + "step": 32948 + }, + { + "epoch": 1.27, + "learning_rate": 2.8905488586817138e-05, + "loss": 0.4099, + "step": 32949 + }, + { + "epoch": 1.27, + "learning_rate": 2.8904847396768403e-05, + "loss": 0.2694, + "step": 32950 + }, + { + "epoch": 1.27, + "learning_rate": 2.8904206206719675e-05, + "loss": 0.4149, + "step": 32951 + }, + { + "epoch": 1.27, + "learning_rate": 2.8903565016670943e-05, + "loss": 0.2234, + "step": 32952 + }, + { + "epoch": 1.27, + "learning_rate": 2.8902923826622215e-05, + "loss": 0.3474, + "step": 32953 + }, + { + "epoch": 1.27, + "learning_rate": 2.890228263657348e-05, + "loss": 0.3391, + "step": 32954 + }, + { + "epoch": 1.27, + "learning_rate": 2.890164144652475e-05, + "loss": 0.5519, + "step": 32955 + }, + { + "epoch": 1.27, + "learning_rate": 2.890100025647602e-05, + "loss": 0.4276, + "step": 32956 + }, + { + "epoch": 1.27, + "learning_rate": 2.890035906642729e-05, + "loss": 0.4891, + "step": 32957 + }, + { + "epoch": 1.27, + "learning_rate": 2.8899717876378563e-05, + "loss": 0.4289, + "step": 32958 + }, + { + "epoch": 1.27, + "learning_rate": 2.8899076686329828e-05, + "loss": 0.4292, + "step": 32959 + }, + { + "epoch": 1.27, + "learning_rate": 2.88984354962811e-05, + "loss": 0.321, + "step": 32960 + }, + { + "epoch": 1.27, + "learning_rate": 2.8897794306232368e-05, + "loss": 0.353, + "step": 32961 + }, + { + "epoch": 1.27, + "learning_rate": 2.889715311618364e-05, + "loss": 0.273, + "step": 32962 + }, + { + "epoch": 1.27, + "learning_rate": 2.8896511926134905e-05, + "loss": 0.3085, + "step": 32963 + }, + { + "epoch": 1.27, + "learning_rate": 2.8895870736086177e-05, + "loss": 0.3357, + "step": 32964 + }, + { + "epoch": 1.27, + "learning_rate": 2.8895229546037445e-05, + "loss": 0.4025, + "step": 32965 + }, + { + "epoch": 1.27, + "learning_rate": 2.8894588355988717e-05, + "loss": 0.3483, + "step": 32966 + }, + { + "epoch": 1.27, + "learning_rate": 2.889394716593999e-05, + "loss": 0.4364, + "step": 32967 + }, + { + "epoch": 1.27, + "learning_rate": 2.8893305975891253e-05, + "loss": 0.4288, + "step": 32968 + }, + { + "epoch": 1.27, + "learning_rate": 2.889266478584253e-05, + "loss": 0.4298, + "step": 32969 + }, + { + "epoch": 1.27, + "learning_rate": 2.8892023595793793e-05, + "loss": 0.5396, + "step": 32970 + }, + { + "epoch": 1.27, + "learning_rate": 2.8891382405745065e-05, + "loss": 0.4688, + "step": 32971 + }, + { + "epoch": 1.27, + "learning_rate": 2.8890741215696333e-05, + "loss": 0.4125, + "step": 32972 + }, + { + "epoch": 1.27, + "learning_rate": 2.8890100025647605e-05, + "loss": 0.3198, + "step": 32973 + }, + { + "epoch": 1.27, + "learning_rate": 2.888945883559887e-05, + "loss": 0.5414, + "step": 32974 + }, + { + "epoch": 1.27, + "learning_rate": 2.8888817645550142e-05, + "loss": 0.3555, + "step": 32975 + }, + { + "epoch": 1.27, + "learning_rate": 2.8888176455501413e-05, + "loss": 0.403, + "step": 32976 + }, + { + "epoch": 1.27, + "learning_rate": 2.8887535265452682e-05, + "loss": 0.3795, + "step": 32977 + }, + { + "epoch": 1.27, + "learning_rate": 2.8886894075403954e-05, + "loss": 0.3943, + "step": 32978 + }, + { + "epoch": 1.27, + "learning_rate": 2.888625288535522e-05, + "loss": 0.4505, + "step": 32979 + }, + { + "epoch": 1.27, + "learning_rate": 2.888561169530649e-05, + "loss": 0.341, + "step": 32980 + }, + { + "epoch": 1.27, + "learning_rate": 2.888497050525776e-05, + "loss": 0.362, + "step": 32981 + }, + { + "epoch": 1.27, + "learning_rate": 2.888432931520903e-05, + "loss": 0.3385, + "step": 32982 + }, + { + "epoch": 1.27, + "learning_rate": 2.8883688125160295e-05, + "loss": 0.3592, + "step": 32983 + }, + { + "epoch": 1.27, + "learning_rate": 2.8883046935111567e-05, + "loss": 0.4982, + "step": 32984 + }, + { + "epoch": 1.27, + "learning_rate": 2.8882405745062842e-05, + "loss": 0.4484, + "step": 32985 + }, + { + "epoch": 1.27, + "learning_rate": 2.8881764555014107e-05, + "loss": 0.3244, + "step": 32986 + }, + { + "epoch": 1.27, + "learning_rate": 2.888112336496538e-05, + "loss": 0.4183, + "step": 32987 + }, + { + "epoch": 1.27, + "learning_rate": 2.8880482174916647e-05, + "loss": 0.3194, + "step": 32988 + }, + { + "epoch": 1.27, + "learning_rate": 2.887984098486792e-05, + "loss": 0.4931, + "step": 32989 + }, + { + "epoch": 1.27, + "learning_rate": 2.8879199794819184e-05, + "loss": 0.4084, + "step": 32990 + }, + { + "epoch": 1.27, + "learning_rate": 2.8878558604770455e-05, + "loss": 0.3381, + "step": 32991 + }, + { + "epoch": 1.27, + "learning_rate": 2.8877917414721724e-05, + "loss": 0.4518, + "step": 32992 + }, + { + "epoch": 1.27, + "learning_rate": 2.8877276224672995e-05, + "loss": 0.5118, + "step": 32993 + }, + { + "epoch": 1.27, + "learning_rate": 2.8876635034624267e-05, + "loss": 0.3752, + "step": 32994 + }, + { + "epoch": 1.27, + "learning_rate": 2.8875993844575532e-05, + "loss": 0.3595, + "step": 32995 + }, + { + "epoch": 1.27, + "learning_rate": 2.8875352654526804e-05, + "loss": 0.4158, + "step": 32996 + }, + { + "epoch": 1.27, + "learning_rate": 2.8874711464478072e-05, + "loss": 0.3541, + "step": 32997 + }, + { + "epoch": 1.27, + "learning_rate": 2.8874070274429344e-05, + "loss": 0.2977, + "step": 32998 + }, + { + "epoch": 1.27, + "learning_rate": 2.887342908438061e-05, + "loss": 0.2786, + "step": 32999 + }, + { + "epoch": 1.27, + "learning_rate": 2.887278789433188e-05, + "loss": 0.3261, + "step": 33000 + }, + { + "epoch": 1.27, + "learning_rate": 2.887214670428315e-05, + "loss": 0.2615, + "step": 33001 + }, + { + "epoch": 1.27, + "learning_rate": 2.887150551423442e-05, + "loss": 0.4113, + "step": 33002 + }, + { + "epoch": 1.27, + "learning_rate": 2.8870864324185692e-05, + "loss": 0.4373, + "step": 33003 + }, + { + "epoch": 1.27, + "learning_rate": 2.887022313413696e-05, + "loss": 0.5236, + "step": 33004 + }, + { + "epoch": 1.27, + "learning_rate": 2.8869581944088232e-05, + "loss": 0.5257, + "step": 33005 + }, + { + "epoch": 1.27, + "learning_rate": 2.8868940754039497e-05, + "loss": 0.4039, + "step": 33006 + }, + { + "epoch": 1.27, + "learning_rate": 2.886829956399077e-05, + "loss": 0.4068, + "step": 33007 + }, + { + "epoch": 1.27, + "learning_rate": 2.8867658373942037e-05, + "loss": 0.385, + "step": 33008 + }, + { + "epoch": 1.27, + "learning_rate": 2.886701718389331e-05, + "loss": 0.436, + "step": 33009 + }, + { + "epoch": 1.27, + "learning_rate": 2.8866375993844574e-05, + "loss": 0.494, + "step": 33010 + }, + { + "epoch": 1.27, + "learning_rate": 2.8865734803795846e-05, + "loss": 0.4185, + "step": 33011 + }, + { + "epoch": 1.27, + "learning_rate": 2.8865093613747117e-05, + "loss": 0.5114, + "step": 33012 + }, + { + "epoch": 1.27, + "learning_rate": 2.8864452423698386e-05, + "loss": 0.3373, + "step": 33013 + }, + { + "epoch": 1.27, + "learning_rate": 2.8863811233649657e-05, + "loss": 0.1859, + "step": 33014 + }, + { + "epoch": 1.27, + "learning_rate": 2.8863170043600922e-05, + "loss": 0.4441, + "step": 33015 + }, + { + "epoch": 1.27, + "learning_rate": 2.8862528853552194e-05, + "loss": 0.2092, + "step": 33016 + }, + { + "epoch": 1.27, + "learning_rate": 2.8861887663503462e-05, + "loss": 0.3475, + "step": 33017 + }, + { + "epoch": 1.27, + "learning_rate": 2.8861246473454734e-05, + "loss": 0.4001, + "step": 33018 + }, + { + "epoch": 1.27, + "learning_rate": 2.8860605283406e-05, + "loss": 0.3613, + "step": 33019 + }, + { + "epoch": 1.27, + "learning_rate": 2.885996409335727e-05, + "loss": 0.3812, + "step": 33020 + }, + { + "epoch": 1.27, + "learning_rate": 2.8859322903308546e-05, + "loss": 0.3505, + "step": 33021 + }, + { + "epoch": 1.27, + "learning_rate": 2.885868171325981e-05, + "loss": 0.3475, + "step": 33022 + }, + { + "epoch": 1.27, + "learning_rate": 2.8858040523211083e-05, + "loss": 0.351, + "step": 33023 + }, + { + "epoch": 1.27, + "learning_rate": 2.885739933316235e-05, + "loss": 0.4258, + "step": 33024 + }, + { + "epoch": 1.27, + "learning_rate": 2.8856758143113623e-05, + "loss": 0.1945, + "step": 33025 + }, + { + "epoch": 1.27, + "learning_rate": 2.8856116953064888e-05, + "loss": 0.4324, + "step": 33026 + }, + { + "epoch": 1.27, + "learning_rate": 2.885547576301616e-05, + "loss": 0.347, + "step": 33027 + }, + { + "epoch": 1.27, + "learning_rate": 2.8854834572967428e-05, + "loss": 0.3762, + "step": 33028 + }, + { + "epoch": 1.27, + "learning_rate": 2.88541933829187e-05, + "loss": 0.4899, + "step": 33029 + }, + { + "epoch": 1.27, + "learning_rate": 2.885355219286997e-05, + "loss": 0.3217, + "step": 33030 + }, + { + "epoch": 1.27, + "learning_rate": 2.8852911002821236e-05, + "loss": 0.3237, + "step": 33031 + }, + { + "epoch": 1.27, + "learning_rate": 2.8852269812772508e-05, + "loss": 0.4976, + "step": 33032 + }, + { + "epoch": 1.27, + "learning_rate": 2.8851628622723776e-05, + "loss": 0.3834, + "step": 33033 + }, + { + "epoch": 1.27, + "learning_rate": 2.8850987432675048e-05, + "loss": 0.4864, + "step": 33034 + }, + { + "epoch": 1.27, + "learning_rate": 2.8850346242626313e-05, + "loss": 0.2846, + "step": 33035 + }, + { + "epoch": 1.27, + "learning_rate": 2.8849705052577584e-05, + "loss": 0.4827, + "step": 33036 + }, + { + "epoch": 1.27, + "learning_rate": 2.8849063862528853e-05, + "loss": 0.4957, + "step": 33037 + }, + { + "epoch": 1.27, + "learning_rate": 2.8848422672480124e-05, + "loss": 0.3398, + "step": 33038 + }, + { + "epoch": 1.27, + "learning_rate": 2.8847781482431396e-05, + "loss": 0.3708, + "step": 33039 + }, + { + "epoch": 1.27, + "learning_rate": 2.8847140292382664e-05, + "loss": 0.3118, + "step": 33040 + }, + { + "epoch": 1.27, + "learning_rate": 2.8846499102333936e-05, + "loss": 0.3201, + "step": 33041 + }, + { + "epoch": 1.27, + "learning_rate": 2.88458579122852e-05, + "loss": 0.3669, + "step": 33042 + }, + { + "epoch": 1.27, + "learning_rate": 2.8845216722236473e-05, + "loss": 0.4225, + "step": 33043 + }, + { + "epoch": 1.27, + "learning_rate": 2.884457553218774e-05, + "loss": 0.2754, + "step": 33044 + }, + { + "epoch": 1.27, + "learning_rate": 2.8843934342139013e-05, + "loss": 0.322, + "step": 33045 + }, + { + "epoch": 1.27, + "learning_rate": 2.8843293152090278e-05, + "loss": 0.3449, + "step": 33046 + }, + { + "epoch": 1.27, + "learning_rate": 2.884265196204155e-05, + "loss": 0.3272, + "step": 33047 + }, + { + "epoch": 1.27, + "learning_rate": 2.884201077199282e-05, + "loss": 0.3581, + "step": 33048 + }, + { + "epoch": 1.27, + "learning_rate": 2.884136958194409e-05, + "loss": 0.2587, + "step": 33049 + }, + { + "epoch": 1.27, + "learning_rate": 2.884072839189536e-05, + "loss": 0.3856, + "step": 33050 + }, + { + "epoch": 1.27, + "learning_rate": 2.8840087201846626e-05, + "loss": 0.3278, + "step": 33051 + }, + { + "epoch": 1.27, + "learning_rate": 2.8839446011797898e-05, + "loss": 0.3959, + "step": 33052 + }, + { + "epoch": 1.27, + "learning_rate": 2.8838804821749166e-05, + "loss": 0.273, + "step": 33053 + }, + { + "epoch": 1.27, + "learning_rate": 2.8838163631700438e-05, + "loss": 0.3557, + "step": 33054 + }, + { + "epoch": 1.27, + "learning_rate": 2.8837522441651703e-05, + "loss": 0.3798, + "step": 33055 + }, + { + "epoch": 1.27, + "learning_rate": 2.8836881251602975e-05, + "loss": 0.502, + "step": 33056 + }, + { + "epoch": 1.27, + "learning_rate": 2.883624006155425e-05, + "loss": 0.301, + "step": 33057 + }, + { + "epoch": 1.27, + "learning_rate": 2.8835598871505515e-05, + "loss": 0.5588, + "step": 33058 + }, + { + "epoch": 1.27, + "learning_rate": 2.8834957681456786e-05, + "loss": 0.4014, + "step": 33059 + }, + { + "epoch": 1.27, + "learning_rate": 2.8834316491408055e-05, + "loss": 0.372, + "step": 33060 + }, + { + "epoch": 1.27, + "learning_rate": 2.8833675301359326e-05, + "loss": 0.4762, + "step": 33061 + }, + { + "epoch": 1.27, + "learning_rate": 2.883303411131059e-05, + "loss": 0.4496, + "step": 33062 + }, + { + "epoch": 1.27, + "learning_rate": 2.8832392921261863e-05, + "loss": 0.5045, + "step": 33063 + }, + { + "epoch": 1.27, + "learning_rate": 2.883175173121313e-05, + "loss": 0.3678, + "step": 33064 + }, + { + "epoch": 1.27, + "learning_rate": 2.8831110541164403e-05, + "loss": 0.2364, + "step": 33065 + }, + { + "epoch": 1.27, + "learning_rate": 2.8830469351115675e-05, + "loss": 0.3229, + "step": 33066 + }, + { + "epoch": 1.27, + "learning_rate": 2.882982816106694e-05, + "loss": 0.3058, + "step": 33067 + }, + { + "epoch": 1.27, + "learning_rate": 2.882918697101821e-05, + "loss": 0.403, + "step": 33068 + }, + { + "epoch": 1.27, + "learning_rate": 2.882854578096948e-05, + "loss": 0.3793, + "step": 33069 + }, + { + "epoch": 1.27, + "learning_rate": 2.882790459092075e-05, + "loss": 0.5707, + "step": 33070 + }, + { + "epoch": 1.27, + "learning_rate": 2.8827263400872017e-05, + "loss": 0.3841, + "step": 33071 + }, + { + "epoch": 1.27, + "learning_rate": 2.8826622210823288e-05, + "loss": 0.4913, + "step": 33072 + }, + { + "epoch": 1.27, + "learning_rate": 2.8825981020774557e-05, + "loss": 0.4549, + "step": 33073 + }, + { + "epoch": 1.27, + "learning_rate": 2.8825339830725828e-05, + "loss": 0.4944, + "step": 33074 + }, + { + "epoch": 1.27, + "learning_rate": 2.88246986406771e-05, + "loss": 0.386, + "step": 33075 + }, + { + "epoch": 1.27, + "learning_rate": 2.882405745062837e-05, + "loss": 0.3849, + "step": 33076 + }, + { + "epoch": 1.27, + "learning_rate": 2.882341626057964e-05, + "loss": 0.2389, + "step": 33077 + }, + { + "epoch": 1.27, + "learning_rate": 2.8822775070530905e-05, + "loss": 0.3271, + "step": 33078 + }, + { + "epoch": 1.27, + "learning_rate": 2.8822133880482177e-05, + "loss": 0.3035, + "step": 33079 + }, + { + "epoch": 1.27, + "learning_rate": 2.8821492690433445e-05, + "loss": 0.3639, + "step": 33080 + }, + { + "epoch": 1.27, + "learning_rate": 2.8820851500384717e-05, + "loss": 0.4475, + "step": 33081 + }, + { + "epoch": 1.27, + "learning_rate": 2.882021031033599e-05, + "loss": 0.4695, + "step": 33082 + }, + { + "epoch": 1.27, + "learning_rate": 2.8819569120287253e-05, + "loss": 0.4374, + "step": 33083 + }, + { + "epoch": 1.27, + "learning_rate": 2.8818927930238525e-05, + "loss": 0.3359, + "step": 33084 + }, + { + "epoch": 1.27, + "learning_rate": 2.8818286740189793e-05, + "loss": 0.2889, + "step": 33085 + }, + { + "epoch": 1.27, + "learning_rate": 2.8817645550141065e-05, + "loss": 0.3548, + "step": 33086 + }, + { + "epoch": 1.27, + "learning_rate": 2.881700436009233e-05, + "loss": 0.4019, + "step": 33087 + }, + { + "epoch": 1.27, + "learning_rate": 2.8816363170043602e-05, + "loss": 0.3063, + "step": 33088 + }, + { + "epoch": 1.27, + "learning_rate": 2.881572197999487e-05, + "loss": 0.3308, + "step": 33089 + }, + { + "epoch": 1.27, + "learning_rate": 2.8815080789946142e-05, + "loss": 0.4716, + "step": 33090 + }, + { + "epoch": 1.27, + "learning_rate": 2.8814439599897414e-05, + "loss": 0.3694, + "step": 33091 + }, + { + "epoch": 1.27, + "learning_rate": 2.881379840984868e-05, + "loss": 0.2594, + "step": 33092 + }, + { + "epoch": 1.27, + "learning_rate": 2.8813157219799954e-05, + "loss": 0.429, + "step": 33093 + }, + { + "epoch": 1.27, + "learning_rate": 2.881251602975122e-05, + "loss": 0.3898, + "step": 33094 + }, + { + "epoch": 1.27, + "learning_rate": 2.881187483970249e-05, + "loss": 0.3096, + "step": 33095 + }, + { + "epoch": 1.27, + "learning_rate": 2.881123364965376e-05, + "loss": 0.4248, + "step": 33096 + }, + { + "epoch": 1.27, + "learning_rate": 2.881059245960503e-05, + "loss": 0.3515, + "step": 33097 + }, + { + "epoch": 1.27, + "learning_rate": 2.8809951269556295e-05, + "loss": 0.2841, + "step": 33098 + }, + { + "epoch": 1.27, + "learning_rate": 2.8809310079507567e-05, + "loss": 0.4387, + "step": 33099 + }, + { + "epoch": 1.27, + "learning_rate": 2.880866888945884e-05, + "loss": 0.3848, + "step": 33100 + }, + { + "epoch": 1.27, + "learning_rate": 2.8808027699410107e-05, + "loss": 0.452, + "step": 33101 + }, + { + "epoch": 1.27, + "learning_rate": 2.880738650936138e-05, + "loss": 0.4075, + "step": 33102 + }, + { + "epoch": 1.27, + "learning_rate": 2.8806745319312644e-05, + "loss": 0.3146, + "step": 33103 + }, + { + "epoch": 1.27, + "learning_rate": 2.8806104129263915e-05, + "loss": 0.3282, + "step": 33104 + }, + { + "epoch": 1.27, + "learning_rate": 2.8805462939215184e-05, + "loss": 0.4086, + "step": 33105 + }, + { + "epoch": 1.27, + "learning_rate": 2.8804821749166455e-05, + "loss": 0.3567, + "step": 33106 + }, + { + "epoch": 1.27, + "learning_rate": 2.880418055911772e-05, + "loss": 0.2756, + "step": 33107 + }, + { + "epoch": 1.27, + "learning_rate": 2.8803539369068992e-05, + "loss": 0.3028, + "step": 33108 + }, + { + "epoch": 1.27, + "learning_rate": 2.8802898179020267e-05, + "loss": 0.2709, + "step": 33109 + }, + { + "epoch": 1.27, + "learning_rate": 2.8802256988971532e-05, + "loss": 0.4482, + "step": 33110 + }, + { + "epoch": 1.27, + "learning_rate": 2.8801615798922804e-05, + "loss": 0.4402, + "step": 33111 + }, + { + "epoch": 1.27, + "learning_rate": 2.8800974608874072e-05, + "loss": 0.5099, + "step": 33112 + }, + { + "epoch": 1.27, + "learning_rate": 2.8800333418825344e-05, + "loss": 0.4116, + "step": 33113 + }, + { + "epoch": 1.27, + "learning_rate": 2.879969222877661e-05, + "loss": 0.3897, + "step": 33114 + }, + { + "epoch": 1.27, + "learning_rate": 2.879905103872788e-05, + "loss": 0.6499, + "step": 33115 + }, + { + "epoch": 1.27, + "learning_rate": 2.879840984867915e-05, + "loss": 0.3243, + "step": 33116 + }, + { + "epoch": 1.27, + "learning_rate": 2.879776865863042e-05, + "loss": 0.3496, + "step": 33117 + }, + { + "epoch": 1.27, + "learning_rate": 2.8797127468581692e-05, + "loss": 0.4112, + "step": 33118 + }, + { + "epoch": 1.27, + "learning_rate": 2.8796486278532957e-05, + "loss": 0.4682, + "step": 33119 + }, + { + "epoch": 1.27, + "learning_rate": 2.879584508848423e-05, + "loss": 0.5175, + "step": 33120 + }, + { + "epoch": 1.27, + "learning_rate": 2.8795203898435497e-05, + "loss": 0.3088, + "step": 33121 + }, + { + "epoch": 1.27, + "learning_rate": 2.879456270838677e-05, + "loss": 0.3131, + "step": 33122 + }, + { + "epoch": 1.27, + "learning_rate": 2.8793921518338034e-05, + "loss": 0.3084, + "step": 33123 + }, + { + "epoch": 1.27, + "learning_rate": 2.8793280328289306e-05, + "loss": 0.3699, + "step": 33124 + }, + { + "epoch": 1.27, + "learning_rate": 2.8792639138240574e-05, + "loss": 0.3038, + "step": 33125 + }, + { + "epoch": 1.27, + "learning_rate": 2.8791997948191846e-05, + "loss": 0.4625, + "step": 33126 + }, + { + "epoch": 1.27, + "learning_rate": 2.8791356758143117e-05, + "loss": 0.4109, + "step": 33127 + }, + { + "epoch": 1.27, + "learning_rate": 2.8790715568094386e-05, + "loss": 0.5599, + "step": 33128 + }, + { + "epoch": 1.27, + "learning_rate": 2.8790074378045657e-05, + "loss": 0.4131, + "step": 33129 + }, + { + "epoch": 1.27, + "learning_rate": 2.8789433187996922e-05, + "loss": 0.435, + "step": 33130 + }, + { + "epoch": 1.27, + "learning_rate": 2.8788791997948194e-05, + "loss": 0.3949, + "step": 33131 + }, + { + "epoch": 1.27, + "learning_rate": 2.8788150807899462e-05, + "loss": 0.3088, + "step": 33132 + }, + { + "epoch": 1.27, + "learning_rate": 2.8787509617850734e-05, + "loss": 0.4425, + "step": 33133 + }, + { + "epoch": 1.27, + "learning_rate": 2.8786868427802e-05, + "loss": 0.3586, + "step": 33134 + }, + { + "epoch": 1.27, + "learning_rate": 2.878622723775327e-05, + "loss": 0.5032, + "step": 33135 + }, + { + "epoch": 1.27, + "learning_rate": 2.8785586047704543e-05, + "loss": 0.3451, + "step": 33136 + }, + { + "epoch": 1.27, + "learning_rate": 2.878494485765581e-05, + "loss": 0.3609, + "step": 33137 + }, + { + "epoch": 1.27, + "learning_rate": 2.8784303667607083e-05, + "loss": 0.3484, + "step": 33138 + }, + { + "epoch": 1.27, + "learning_rate": 2.8783662477558348e-05, + "loss": 0.4233, + "step": 33139 + }, + { + "epoch": 1.27, + "learning_rate": 2.878302128750962e-05, + "loss": 0.3632, + "step": 33140 + }, + { + "epoch": 1.27, + "learning_rate": 2.8782380097460888e-05, + "loss": 0.4532, + "step": 33141 + }, + { + "epoch": 1.27, + "learning_rate": 2.878173890741216e-05, + "loss": 0.5111, + "step": 33142 + }, + { + "epoch": 1.27, + "learning_rate": 2.8781097717363424e-05, + "loss": 0.3801, + "step": 33143 + }, + { + "epoch": 1.27, + "learning_rate": 2.8780456527314696e-05, + "loss": 0.3162, + "step": 33144 + }, + { + "epoch": 1.27, + "learning_rate": 2.877981533726597e-05, + "loss": 0.3951, + "step": 33145 + }, + { + "epoch": 1.27, + "learning_rate": 2.8779174147217236e-05, + "loss": 0.4291, + "step": 33146 + }, + { + "epoch": 1.27, + "learning_rate": 2.8778532957168508e-05, + "loss": 0.3948, + "step": 33147 + }, + { + "epoch": 1.27, + "learning_rate": 2.8777891767119776e-05, + "loss": 0.3751, + "step": 33148 + }, + { + "epoch": 1.27, + "learning_rate": 2.8777250577071048e-05, + "loss": 0.4617, + "step": 33149 + }, + { + "epoch": 1.27, + "learning_rate": 2.8776609387022313e-05, + "loss": 0.3891, + "step": 33150 + }, + { + "epoch": 1.27, + "learning_rate": 2.8775968196973584e-05, + "loss": 0.3928, + "step": 33151 + }, + { + "epoch": 1.27, + "learning_rate": 2.8775327006924853e-05, + "loss": 0.3667, + "step": 33152 + }, + { + "epoch": 1.27, + "learning_rate": 2.8774685816876124e-05, + "loss": 0.3345, + "step": 33153 + }, + { + "epoch": 1.27, + "learning_rate": 2.8774044626827396e-05, + "loss": 0.3582, + "step": 33154 + }, + { + "epoch": 1.27, + "learning_rate": 2.877340343677866e-05, + "loss": 0.4632, + "step": 33155 + }, + { + "epoch": 1.27, + "learning_rate": 2.8772762246729933e-05, + "loss": 0.4555, + "step": 33156 + }, + { + "epoch": 1.27, + "learning_rate": 2.87721210566812e-05, + "loss": 0.2186, + "step": 33157 + }, + { + "epoch": 1.27, + "learning_rate": 2.8771479866632473e-05, + "loss": 0.3462, + "step": 33158 + }, + { + "epoch": 1.27, + "learning_rate": 2.8770838676583738e-05, + "loss": 0.3233, + "step": 33159 + }, + { + "epoch": 1.27, + "learning_rate": 2.877019748653501e-05, + "loss": 0.4267, + "step": 33160 + }, + { + "epoch": 1.27, + "learning_rate": 2.8769556296486278e-05, + "loss": 0.2562, + "step": 33161 + }, + { + "epoch": 1.27, + "learning_rate": 2.876891510643755e-05, + "loss": 0.3249, + "step": 33162 + }, + { + "epoch": 1.28, + "learning_rate": 2.876827391638882e-05, + "loss": 0.3575, + "step": 33163 + }, + { + "epoch": 1.28, + "learning_rate": 2.876763272634009e-05, + "loss": 0.4832, + "step": 33164 + }, + { + "epoch": 1.28, + "learning_rate": 2.876699153629136e-05, + "loss": 0.443, + "step": 33165 + }, + { + "epoch": 1.28, + "learning_rate": 2.8766350346242626e-05, + "loss": 0.4735, + "step": 33166 + }, + { + "epoch": 1.28, + "learning_rate": 2.8765709156193898e-05, + "loss": 0.3441, + "step": 33167 + }, + { + "epoch": 1.28, + "learning_rate": 2.8765067966145166e-05, + "loss": 0.4717, + "step": 33168 + }, + { + "epoch": 1.28, + "learning_rate": 2.8764426776096438e-05, + "loss": 0.2434, + "step": 33169 + }, + { + "epoch": 1.28, + "learning_rate": 2.8763785586047703e-05, + "loss": 0.2847, + "step": 33170 + }, + { + "epoch": 1.28, + "learning_rate": 2.8763144395998975e-05, + "loss": 0.484, + "step": 33171 + }, + { + "epoch": 1.28, + "learning_rate": 2.8762503205950246e-05, + "loss": 0.4189, + "step": 33172 + }, + { + "epoch": 1.28, + "learning_rate": 2.8761862015901515e-05, + "loss": 0.5539, + "step": 33173 + }, + { + "epoch": 1.28, + "learning_rate": 2.8761220825852786e-05, + "loss": 0.3455, + "step": 33174 + }, + { + "epoch": 1.28, + "learning_rate": 2.876057963580405e-05, + "loss": 0.3742, + "step": 33175 + }, + { + "epoch": 1.28, + "learning_rate": 2.8759938445755323e-05, + "loss": 0.3279, + "step": 33176 + }, + { + "epoch": 1.28, + "learning_rate": 2.875929725570659e-05, + "loss": 0.3527, + "step": 33177 + }, + { + "epoch": 1.28, + "learning_rate": 2.8758656065657863e-05, + "loss": 0.3499, + "step": 33178 + }, + { + "epoch": 1.28, + "learning_rate": 2.8758014875609128e-05, + "loss": 0.3852, + "step": 33179 + }, + { + "epoch": 1.28, + "learning_rate": 2.87573736855604e-05, + "loss": 0.3311, + "step": 33180 + }, + { + "epoch": 1.28, + "learning_rate": 2.8756732495511675e-05, + "loss": 0.3652, + "step": 33181 + }, + { + "epoch": 1.28, + "learning_rate": 2.875609130546294e-05, + "loss": 0.3701, + "step": 33182 + }, + { + "epoch": 1.28, + "learning_rate": 2.875545011541421e-05, + "loss": 0.4338, + "step": 33183 + }, + { + "epoch": 1.28, + "learning_rate": 2.875480892536548e-05, + "loss": 0.5042, + "step": 33184 + }, + { + "epoch": 1.28, + "learning_rate": 2.875416773531675e-05, + "loss": 0.3839, + "step": 33185 + }, + { + "epoch": 1.28, + "learning_rate": 2.8753526545268017e-05, + "loss": 0.3768, + "step": 33186 + }, + { + "epoch": 1.28, + "learning_rate": 2.875288535521929e-05, + "loss": 0.4648, + "step": 33187 + }, + { + "epoch": 1.28, + "learning_rate": 2.8752244165170557e-05, + "loss": 0.4333, + "step": 33188 + }, + { + "epoch": 1.28, + "learning_rate": 2.875160297512183e-05, + "loss": 0.2206, + "step": 33189 + }, + { + "epoch": 1.28, + "learning_rate": 2.87509617850731e-05, + "loss": 0.3787, + "step": 33190 + }, + { + "epoch": 1.28, + "learning_rate": 2.8750320595024365e-05, + "loss": 0.5668, + "step": 33191 + }, + { + "epoch": 1.28, + "learning_rate": 2.8749679404975637e-05, + "loss": 0.3995, + "step": 33192 + }, + { + "epoch": 1.28, + "learning_rate": 2.8749038214926905e-05, + "loss": 0.3172, + "step": 33193 + }, + { + "epoch": 1.28, + "learning_rate": 2.8748397024878177e-05, + "loss": 0.3779, + "step": 33194 + }, + { + "epoch": 1.28, + "learning_rate": 2.8747755834829442e-05, + "loss": 0.3664, + "step": 33195 + }, + { + "epoch": 1.28, + "learning_rate": 2.8747114644780713e-05, + "loss": 0.5798, + "step": 33196 + }, + { + "epoch": 1.28, + "learning_rate": 2.8746473454731982e-05, + "loss": 0.4333, + "step": 33197 + }, + { + "epoch": 1.28, + "learning_rate": 2.8745832264683253e-05, + "loss": 0.4526, + "step": 33198 + }, + { + "epoch": 1.28, + "learning_rate": 2.8745191074634525e-05, + "loss": 0.3529, + "step": 33199 + }, + { + "epoch": 1.28, + "learning_rate": 2.8744549884585794e-05, + "loss": 0.3426, + "step": 33200 + }, + { + "epoch": 1.28, + "learning_rate": 2.8743908694537065e-05, + "loss": 0.4565, + "step": 33201 + }, + { + "epoch": 1.28, + "learning_rate": 2.874326750448833e-05, + "loss": 0.3447, + "step": 33202 + }, + { + "epoch": 1.28, + "learning_rate": 2.8742626314439602e-05, + "loss": 0.346, + "step": 33203 + }, + { + "epoch": 1.28, + "learning_rate": 2.874198512439087e-05, + "loss": 0.4051, + "step": 33204 + }, + { + "epoch": 1.28, + "learning_rate": 2.8741343934342142e-05, + "loss": 0.4013, + "step": 33205 + }, + { + "epoch": 1.28, + "learning_rate": 2.8740702744293407e-05, + "loss": 0.2441, + "step": 33206 + }, + { + "epoch": 1.28, + "learning_rate": 2.874006155424468e-05, + "loss": 0.2903, + "step": 33207 + }, + { + "epoch": 1.28, + "learning_rate": 2.873942036419595e-05, + "loss": 0.2868, + "step": 33208 + }, + { + "epoch": 1.28, + "learning_rate": 2.873877917414722e-05, + "loss": 0.3611, + "step": 33209 + }, + { + "epoch": 1.28, + "learning_rate": 2.873813798409849e-05, + "loss": 0.3413, + "step": 33210 + }, + { + "epoch": 1.28, + "learning_rate": 2.8737496794049755e-05, + "loss": 0.3849, + "step": 33211 + }, + { + "epoch": 1.28, + "learning_rate": 2.8736855604001027e-05, + "loss": 0.4314, + "step": 33212 + }, + { + "epoch": 1.28, + "learning_rate": 2.8736214413952295e-05, + "loss": 0.278, + "step": 33213 + }, + { + "epoch": 1.28, + "learning_rate": 2.8735573223903567e-05, + "loss": 0.5168, + "step": 33214 + }, + { + "epoch": 1.28, + "learning_rate": 2.8734932033854832e-05, + "loss": 0.3595, + "step": 33215 + }, + { + "epoch": 1.28, + "learning_rate": 2.8734290843806104e-05, + "loss": 0.3502, + "step": 33216 + }, + { + "epoch": 1.28, + "learning_rate": 2.873364965375738e-05, + "loss": 0.395, + "step": 33217 + }, + { + "epoch": 1.28, + "learning_rate": 2.8733008463708644e-05, + "loss": 0.423, + "step": 33218 + }, + { + "epoch": 1.28, + "learning_rate": 2.8732367273659915e-05, + "loss": 0.4094, + "step": 33219 + }, + { + "epoch": 1.28, + "learning_rate": 2.8731726083611184e-05, + "loss": 0.4374, + "step": 33220 + }, + { + "epoch": 1.28, + "learning_rate": 2.8731084893562456e-05, + "loss": 0.5013, + "step": 33221 + }, + { + "epoch": 1.28, + "learning_rate": 2.873044370351372e-05, + "loss": 0.506, + "step": 33222 + }, + { + "epoch": 1.28, + "learning_rate": 2.8729802513464992e-05, + "loss": 0.3627, + "step": 33223 + }, + { + "epoch": 1.28, + "learning_rate": 2.872916132341626e-05, + "loss": 0.4279, + "step": 33224 + }, + { + "epoch": 1.28, + "learning_rate": 2.8728520133367532e-05, + "loss": 0.3531, + "step": 33225 + }, + { + "epoch": 1.28, + "learning_rate": 2.8727878943318804e-05, + "loss": 0.4337, + "step": 33226 + }, + { + "epoch": 1.28, + "learning_rate": 2.872723775327007e-05, + "loss": 0.4851, + "step": 33227 + }, + { + "epoch": 1.28, + "learning_rate": 2.872659656322134e-05, + "loss": 0.4105, + "step": 33228 + }, + { + "epoch": 1.28, + "learning_rate": 2.872595537317261e-05, + "loss": 0.3317, + "step": 33229 + }, + { + "epoch": 1.28, + "learning_rate": 2.872531418312388e-05, + "loss": 0.3233, + "step": 33230 + }, + { + "epoch": 1.28, + "learning_rate": 2.8724672993075146e-05, + "loss": 0.4716, + "step": 33231 + }, + { + "epoch": 1.28, + "learning_rate": 2.8724031803026417e-05, + "loss": 0.3811, + "step": 33232 + }, + { + "epoch": 1.28, + "learning_rate": 2.8723390612977686e-05, + "loss": 0.2963, + "step": 33233 + }, + { + "epoch": 1.28, + "learning_rate": 2.8722749422928957e-05, + "loss": 0.4154, + "step": 33234 + }, + { + "epoch": 1.28, + "learning_rate": 2.872210823288023e-05, + "loss": 0.488, + "step": 33235 + }, + { + "epoch": 1.28, + "learning_rate": 2.8721467042831497e-05, + "loss": 0.4286, + "step": 33236 + }, + { + "epoch": 1.28, + "learning_rate": 2.872082585278277e-05, + "loss": 0.5016, + "step": 33237 + }, + { + "epoch": 1.28, + "learning_rate": 2.8720184662734034e-05, + "loss": 0.2952, + "step": 33238 + }, + { + "epoch": 1.28, + "learning_rate": 2.8719543472685306e-05, + "loss": 0.3029, + "step": 33239 + }, + { + "epoch": 1.28, + "learning_rate": 2.8718902282636574e-05, + "loss": 0.3693, + "step": 33240 + }, + { + "epoch": 1.28, + "learning_rate": 2.8718261092587846e-05, + "loss": 0.2282, + "step": 33241 + }, + { + "epoch": 1.28, + "learning_rate": 2.871761990253911e-05, + "loss": 0.3577, + "step": 33242 + }, + { + "epoch": 1.28, + "learning_rate": 2.8716978712490382e-05, + "loss": 0.277, + "step": 33243 + }, + { + "epoch": 1.28, + "learning_rate": 2.8716337522441654e-05, + "loss": 0.4801, + "step": 33244 + }, + { + "epoch": 1.28, + "learning_rate": 2.8715696332392923e-05, + "loss": 0.3671, + "step": 33245 + }, + { + "epoch": 1.28, + "learning_rate": 2.8715055142344194e-05, + "loss": 0.3861, + "step": 33246 + }, + { + "epoch": 1.28, + "learning_rate": 2.871441395229546e-05, + "loss": 0.4763, + "step": 33247 + }, + { + "epoch": 1.28, + "learning_rate": 2.871377276224673e-05, + "loss": 0.4231, + "step": 33248 + }, + { + "epoch": 1.28, + "learning_rate": 2.8713131572198e-05, + "loss": 0.3369, + "step": 33249 + }, + { + "epoch": 1.28, + "learning_rate": 2.871249038214927e-05, + "loss": 0.4168, + "step": 33250 + }, + { + "epoch": 1.28, + "learning_rate": 2.8711849192100536e-05, + "loss": 0.4566, + "step": 33251 + }, + { + "epoch": 1.28, + "learning_rate": 2.8711208002051808e-05, + "loss": 0.3765, + "step": 33252 + }, + { + "epoch": 1.28, + "learning_rate": 2.8710566812003083e-05, + "loss": 0.2629, + "step": 33253 + }, + { + "epoch": 1.28, + "learning_rate": 2.8709925621954348e-05, + "loss": 0.3191, + "step": 33254 + }, + { + "epoch": 1.28, + "learning_rate": 2.870928443190562e-05, + "loss": 0.3573, + "step": 33255 + }, + { + "epoch": 1.28, + "learning_rate": 2.8708643241856888e-05, + "loss": 0.4319, + "step": 33256 + }, + { + "epoch": 1.28, + "learning_rate": 2.870800205180816e-05, + "loss": 0.3267, + "step": 33257 + }, + { + "epoch": 1.28, + "learning_rate": 2.8707360861759424e-05, + "loss": 0.4234, + "step": 33258 + }, + { + "epoch": 1.28, + "learning_rate": 2.8706719671710696e-05, + "loss": 0.4363, + "step": 33259 + }, + { + "epoch": 1.28, + "learning_rate": 2.8706078481661964e-05, + "loss": 0.4622, + "step": 33260 + }, + { + "epoch": 1.28, + "learning_rate": 2.8705437291613236e-05, + "loss": 0.3502, + "step": 33261 + }, + { + "epoch": 1.28, + "learning_rate": 2.8704796101564508e-05, + "loss": 0.4738, + "step": 33262 + }, + { + "epoch": 1.28, + "learning_rate": 2.8704154911515773e-05, + "loss": 0.4206, + "step": 33263 + }, + { + "epoch": 1.28, + "learning_rate": 2.8703513721467044e-05, + "loss": 0.44, + "step": 33264 + }, + { + "epoch": 1.28, + "learning_rate": 2.8702872531418313e-05, + "loss": 0.4064, + "step": 33265 + }, + { + "epoch": 1.28, + "learning_rate": 2.8702231341369585e-05, + "loss": 0.4661, + "step": 33266 + }, + { + "epoch": 1.28, + "learning_rate": 2.870159015132085e-05, + "loss": 0.3481, + "step": 33267 + }, + { + "epoch": 1.28, + "learning_rate": 2.870094896127212e-05, + "loss": 0.4431, + "step": 33268 + }, + { + "epoch": 1.28, + "learning_rate": 2.8700307771223396e-05, + "loss": 0.4903, + "step": 33269 + }, + { + "epoch": 1.28, + "learning_rate": 2.869966658117466e-05, + "loss": 0.3387, + "step": 33270 + }, + { + "epoch": 1.28, + "learning_rate": 2.8699025391125933e-05, + "loss": 0.497, + "step": 33271 + }, + { + "epoch": 1.28, + "learning_rate": 2.86983842010772e-05, + "loss": 0.4131, + "step": 33272 + }, + { + "epoch": 1.28, + "learning_rate": 2.8697743011028473e-05, + "loss": 0.5178, + "step": 33273 + }, + { + "epoch": 1.28, + "learning_rate": 2.8697101820979738e-05, + "loss": 0.3624, + "step": 33274 + }, + { + "epoch": 1.28, + "learning_rate": 2.869646063093101e-05, + "loss": 0.4516, + "step": 33275 + }, + { + "epoch": 1.28, + "learning_rate": 2.8695819440882278e-05, + "loss": 0.5313, + "step": 33276 + }, + { + "epoch": 1.28, + "learning_rate": 2.869517825083355e-05, + "loss": 0.3067, + "step": 33277 + }, + { + "epoch": 1.28, + "learning_rate": 2.869453706078482e-05, + "loss": 0.5041, + "step": 33278 + }, + { + "epoch": 1.28, + "learning_rate": 2.8693895870736086e-05, + "loss": 0.4002, + "step": 33279 + }, + { + "epoch": 1.28, + "learning_rate": 2.8693254680687358e-05, + "loss": 0.3715, + "step": 33280 + }, + { + "epoch": 1.28, + "learning_rate": 2.8692613490638626e-05, + "loss": 0.5472, + "step": 33281 + }, + { + "epoch": 1.28, + "learning_rate": 2.8691972300589898e-05, + "loss": 0.3791, + "step": 33282 + }, + { + "epoch": 1.28, + "learning_rate": 2.8691331110541163e-05, + "loss": 0.2851, + "step": 33283 + }, + { + "epoch": 1.28, + "learning_rate": 2.8690689920492435e-05, + "loss": 0.3746, + "step": 33284 + }, + { + "epoch": 1.28, + "learning_rate": 2.8690048730443703e-05, + "loss": 0.4271, + "step": 33285 + }, + { + "epoch": 1.28, + "learning_rate": 2.8689407540394975e-05, + "loss": 0.4504, + "step": 33286 + }, + { + "epoch": 1.28, + "learning_rate": 2.8688766350346247e-05, + "loss": 0.4101, + "step": 33287 + }, + { + "epoch": 1.28, + "learning_rate": 2.8688125160297515e-05, + "loss": 0.3726, + "step": 33288 + }, + { + "epoch": 1.28, + "learning_rate": 2.8687483970248787e-05, + "loss": 0.3791, + "step": 33289 + }, + { + "epoch": 1.28, + "learning_rate": 2.868684278020005e-05, + "loss": 0.3985, + "step": 33290 + }, + { + "epoch": 1.28, + "learning_rate": 2.8686201590151323e-05, + "loss": 0.3232, + "step": 33291 + }, + { + "epoch": 1.28, + "learning_rate": 2.868556040010259e-05, + "loss": 0.3104, + "step": 33292 + }, + { + "epoch": 1.28, + "learning_rate": 2.8684919210053863e-05, + "loss": 0.4771, + "step": 33293 + }, + { + "epoch": 1.28, + "learning_rate": 2.8684278020005128e-05, + "loss": 0.4064, + "step": 33294 + }, + { + "epoch": 1.28, + "learning_rate": 2.86836368299564e-05, + "loss": 0.3679, + "step": 33295 + }, + { + "epoch": 1.28, + "learning_rate": 2.868299563990767e-05, + "loss": 0.4439, + "step": 33296 + }, + { + "epoch": 1.28, + "learning_rate": 2.868235444985894e-05, + "loss": 0.3148, + "step": 33297 + }, + { + "epoch": 1.28, + "learning_rate": 2.868171325981021e-05, + "loss": 0.4552, + "step": 33298 + }, + { + "epoch": 1.28, + "learning_rate": 2.8681072069761477e-05, + "loss": 0.354, + "step": 33299 + }, + { + "epoch": 1.28, + "learning_rate": 2.868043087971275e-05, + "loss": 0.3138, + "step": 33300 + }, + { + "epoch": 1.28, + "learning_rate": 2.8679789689664017e-05, + "loss": 0.2934, + "step": 33301 + }, + { + "epoch": 1.28, + "learning_rate": 2.867914849961529e-05, + "loss": 0.3482, + "step": 33302 + }, + { + "epoch": 1.28, + "learning_rate": 2.8678507309566553e-05, + "loss": 0.2695, + "step": 33303 + }, + { + "epoch": 1.28, + "learning_rate": 2.8677866119517825e-05, + "loss": 0.4474, + "step": 33304 + }, + { + "epoch": 1.28, + "learning_rate": 2.86772249294691e-05, + "loss": 0.2721, + "step": 33305 + }, + { + "epoch": 1.28, + "learning_rate": 2.8676583739420365e-05, + "loss": 0.3037, + "step": 33306 + }, + { + "epoch": 1.28, + "learning_rate": 2.8675942549371637e-05, + "loss": 0.2703, + "step": 33307 + }, + { + "epoch": 1.28, + "learning_rate": 2.8675301359322905e-05, + "loss": 0.3941, + "step": 33308 + }, + { + "epoch": 1.28, + "learning_rate": 2.8674660169274177e-05, + "loss": 0.5717, + "step": 33309 + }, + { + "epoch": 1.28, + "learning_rate": 2.8674018979225442e-05, + "loss": 0.1919, + "step": 33310 + }, + { + "epoch": 1.28, + "learning_rate": 2.8673377789176714e-05, + "loss": 0.4456, + "step": 33311 + }, + { + "epoch": 1.28, + "learning_rate": 2.8672736599127982e-05, + "loss": 0.3606, + "step": 33312 + }, + { + "epoch": 1.28, + "learning_rate": 2.8672095409079254e-05, + "loss": 0.3647, + "step": 33313 + }, + { + "epoch": 1.28, + "learning_rate": 2.8671454219030525e-05, + "loss": 0.4259, + "step": 33314 + }, + { + "epoch": 1.28, + "learning_rate": 2.867081302898179e-05, + "loss": 0.3788, + "step": 33315 + }, + { + "epoch": 1.28, + "learning_rate": 2.8670171838933062e-05, + "loss": 0.3407, + "step": 33316 + }, + { + "epoch": 1.28, + "learning_rate": 2.866953064888433e-05, + "loss": 0.4048, + "step": 33317 + }, + { + "epoch": 1.28, + "learning_rate": 2.8668889458835602e-05, + "loss": 0.4035, + "step": 33318 + }, + { + "epoch": 1.28, + "learning_rate": 2.8668248268786867e-05, + "loss": 0.5271, + "step": 33319 + }, + { + "epoch": 1.28, + "learning_rate": 2.866760707873814e-05, + "loss": 0.3307, + "step": 33320 + }, + { + "epoch": 1.28, + "learning_rate": 2.8666965888689407e-05, + "loss": 0.2918, + "step": 33321 + }, + { + "epoch": 1.28, + "learning_rate": 2.866632469864068e-05, + "loss": 0.3761, + "step": 33322 + }, + { + "epoch": 1.28, + "learning_rate": 2.866568350859195e-05, + "loss": 0.3866, + "step": 33323 + }, + { + "epoch": 1.28, + "learning_rate": 2.866504231854322e-05, + "loss": 0.3577, + "step": 33324 + }, + { + "epoch": 1.28, + "learning_rate": 2.866440112849449e-05, + "loss": 0.3641, + "step": 33325 + }, + { + "epoch": 1.28, + "learning_rate": 2.8663759938445755e-05, + "loss": 0.3774, + "step": 33326 + }, + { + "epoch": 1.28, + "learning_rate": 2.8663118748397027e-05, + "loss": 0.3472, + "step": 33327 + }, + { + "epoch": 1.28, + "learning_rate": 2.8662477558348295e-05, + "loss": 0.4341, + "step": 33328 + }, + { + "epoch": 1.28, + "learning_rate": 2.8661836368299567e-05, + "loss": 0.4235, + "step": 33329 + }, + { + "epoch": 1.28, + "learning_rate": 2.8661195178250832e-05, + "loss": 0.4466, + "step": 33330 + }, + { + "epoch": 1.28, + "learning_rate": 2.8660553988202104e-05, + "loss": 0.3313, + "step": 33331 + }, + { + "epoch": 1.28, + "learning_rate": 2.8659912798153376e-05, + "loss": 0.4339, + "step": 33332 + }, + { + "epoch": 1.28, + "learning_rate": 2.8659271608104644e-05, + "loss": 0.4733, + "step": 33333 + }, + { + "epoch": 1.28, + "learning_rate": 2.8658630418055916e-05, + "loss": 0.4284, + "step": 33334 + }, + { + "epoch": 1.28, + "learning_rate": 2.865798922800718e-05, + "loss": 0.4083, + "step": 33335 + }, + { + "epoch": 1.28, + "learning_rate": 2.8657348037958452e-05, + "loss": 0.3996, + "step": 33336 + }, + { + "epoch": 1.28, + "learning_rate": 2.865670684790972e-05, + "loss": 0.2624, + "step": 33337 + }, + { + "epoch": 1.28, + "learning_rate": 2.8656065657860992e-05, + "loss": 0.3337, + "step": 33338 + }, + { + "epoch": 1.28, + "learning_rate": 2.8655424467812257e-05, + "loss": 0.3161, + "step": 33339 + }, + { + "epoch": 1.28, + "learning_rate": 2.865478327776353e-05, + "loss": 0.5146, + "step": 33340 + }, + { + "epoch": 1.28, + "learning_rate": 2.8654142087714804e-05, + "loss": 0.3743, + "step": 33341 + }, + { + "epoch": 1.28, + "learning_rate": 2.865350089766607e-05, + "loss": 0.5012, + "step": 33342 + }, + { + "epoch": 1.28, + "learning_rate": 2.865285970761734e-05, + "loss": 0.3236, + "step": 33343 + }, + { + "epoch": 1.28, + "learning_rate": 2.865221851756861e-05, + "loss": 0.4384, + "step": 33344 + }, + { + "epoch": 1.28, + "learning_rate": 2.865157732751988e-05, + "loss": 0.2932, + "step": 33345 + }, + { + "epoch": 1.28, + "learning_rate": 2.8650936137471146e-05, + "loss": 0.3541, + "step": 33346 + }, + { + "epoch": 1.28, + "learning_rate": 2.8650294947422417e-05, + "loss": 0.4586, + "step": 33347 + }, + { + "epoch": 1.28, + "learning_rate": 2.8649653757373686e-05, + "loss": 0.3646, + "step": 33348 + }, + { + "epoch": 1.28, + "learning_rate": 2.8649012567324957e-05, + "loss": 0.3925, + "step": 33349 + }, + { + "epoch": 1.28, + "learning_rate": 2.864837137727623e-05, + "loss": 0.4321, + "step": 33350 + }, + { + "epoch": 1.28, + "learning_rate": 2.8647730187227494e-05, + "loss": 0.3619, + "step": 33351 + }, + { + "epoch": 1.28, + "learning_rate": 2.8647088997178766e-05, + "loss": 0.3519, + "step": 33352 + }, + { + "epoch": 1.28, + "learning_rate": 2.8646447807130034e-05, + "loss": 0.3944, + "step": 33353 + }, + { + "epoch": 1.28, + "learning_rate": 2.8645806617081306e-05, + "loss": 0.38, + "step": 33354 + }, + { + "epoch": 1.28, + "learning_rate": 2.864516542703257e-05, + "loss": 0.4064, + "step": 33355 + }, + { + "epoch": 1.28, + "learning_rate": 2.8644524236983843e-05, + "loss": 0.348, + "step": 33356 + }, + { + "epoch": 1.28, + "learning_rate": 2.864388304693511e-05, + "loss": 0.4839, + "step": 33357 + }, + { + "epoch": 1.28, + "learning_rate": 2.8643241856886383e-05, + "loss": 0.3382, + "step": 33358 + }, + { + "epoch": 1.28, + "learning_rate": 2.8642600666837654e-05, + "loss": 0.3041, + "step": 33359 + }, + { + "epoch": 1.28, + "learning_rate": 2.8641959476788923e-05, + "loss": 0.3799, + "step": 33360 + }, + { + "epoch": 1.28, + "learning_rate": 2.8641318286740194e-05, + "loss": 0.3788, + "step": 33361 + }, + { + "epoch": 1.28, + "learning_rate": 2.864067709669146e-05, + "loss": 0.3155, + "step": 33362 + }, + { + "epoch": 1.28, + "learning_rate": 2.864003590664273e-05, + "loss": 0.3182, + "step": 33363 + }, + { + "epoch": 1.28, + "learning_rate": 2.8639394716594e-05, + "loss": 0.4082, + "step": 33364 + }, + { + "epoch": 1.28, + "learning_rate": 2.863875352654527e-05, + "loss": 0.2674, + "step": 33365 + }, + { + "epoch": 1.28, + "learning_rate": 2.8638112336496536e-05, + "loss": 0.4461, + "step": 33366 + }, + { + "epoch": 1.28, + "learning_rate": 2.8637471146447808e-05, + "loss": 0.3671, + "step": 33367 + }, + { + "epoch": 1.28, + "learning_rate": 2.863682995639908e-05, + "loss": 0.4741, + "step": 33368 + }, + { + "epoch": 1.28, + "learning_rate": 2.8636188766350348e-05, + "loss": 0.3391, + "step": 33369 + }, + { + "epoch": 1.28, + "learning_rate": 2.863554757630162e-05, + "loss": 0.5034, + "step": 33370 + }, + { + "epoch": 1.28, + "learning_rate": 2.8634906386252884e-05, + "loss": 0.3553, + "step": 33371 + }, + { + "epoch": 1.28, + "learning_rate": 2.8634265196204156e-05, + "loss": 0.4612, + "step": 33372 + }, + { + "epoch": 1.28, + "learning_rate": 2.8633624006155424e-05, + "loss": 0.4449, + "step": 33373 + }, + { + "epoch": 1.28, + "learning_rate": 2.8632982816106696e-05, + "loss": 0.4005, + "step": 33374 + }, + { + "epoch": 1.28, + "learning_rate": 2.863234162605796e-05, + "loss": 0.3604, + "step": 33375 + }, + { + "epoch": 1.28, + "learning_rate": 2.8631700436009233e-05, + "loss": 0.4188, + "step": 33376 + }, + { + "epoch": 1.28, + "learning_rate": 2.8631059245960508e-05, + "loss": 0.2522, + "step": 33377 + }, + { + "epoch": 1.28, + "learning_rate": 2.8630418055911773e-05, + "loss": 0.4433, + "step": 33378 + }, + { + "epoch": 1.28, + "learning_rate": 2.8629776865863045e-05, + "loss": 0.3688, + "step": 33379 + }, + { + "epoch": 1.28, + "learning_rate": 2.8629135675814313e-05, + "loss": 0.4475, + "step": 33380 + }, + { + "epoch": 1.28, + "learning_rate": 2.8628494485765585e-05, + "loss": 0.3072, + "step": 33381 + }, + { + "epoch": 1.28, + "learning_rate": 2.862785329571685e-05, + "loss": 0.3522, + "step": 33382 + }, + { + "epoch": 1.28, + "learning_rate": 2.862721210566812e-05, + "loss": 0.3695, + "step": 33383 + }, + { + "epoch": 1.28, + "learning_rate": 2.862657091561939e-05, + "loss": 0.4077, + "step": 33384 + }, + { + "epoch": 1.28, + "learning_rate": 2.862592972557066e-05, + "loss": 0.3946, + "step": 33385 + }, + { + "epoch": 1.28, + "learning_rate": 2.8625288535521933e-05, + "loss": 0.5189, + "step": 33386 + }, + { + "epoch": 1.28, + "learning_rate": 2.8624647345473198e-05, + "loss": 0.3679, + "step": 33387 + }, + { + "epoch": 1.28, + "learning_rate": 2.862400615542447e-05, + "loss": 0.2504, + "step": 33388 + }, + { + "epoch": 1.28, + "learning_rate": 2.8623364965375738e-05, + "loss": 0.2663, + "step": 33389 + }, + { + "epoch": 1.28, + "learning_rate": 2.862272377532701e-05, + "loss": 0.415, + "step": 33390 + }, + { + "epoch": 1.28, + "learning_rate": 2.8622082585278275e-05, + "loss": 0.3979, + "step": 33391 + }, + { + "epoch": 1.28, + "learning_rate": 2.8621441395229546e-05, + "loss": 0.2397, + "step": 33392 + }, + { + "epoch": 1.28, + "learning_rate": 2.8620800205180815e-05, + "loss": 0.5161, + "step": 33393 + }, + { + "epoch": 1.28, + "learning_rate": 2.8620159015132086e-05, + "loss": 0.4924, + "step": 33394 + }, + { + "epoch": 1.28, + "learning_rate": 2.8619517825083358e-05, + "loss": 0.4262, + "step": 33395 + }, + { + "epoch": 1.28, + "learning_rate": 2.8618876635034626e-05, + "loss": 0.372, + "step": 33396 + }, + { + "epoch": 1.28, + "learning_rate": 2.8618235444985898e-05, + "loss": 0.3757, + "step": 33397 + }, + { + "epoch": 1.28, + "learning_rate": 2.8617594254937163e-05, + "loss": 0.403, + "step": 33398 + }, + { + "epoch": 1.28, + "learning_rate": 2.8616953064888435e-05, + "loss": 0.3477, + "step": 33399 + }, + { + "epoch": 1.28, + "learning_rate": 2.8616311874839703e-05, + "loss": 0.3874, + "step": 33400 + }, + { + "epoch": 1.28, + "learning_rate": 2.8615670684790975e-05, + "loss": 0.3198, + "step": 33401 + }, + { + "epoch": 1.28, + "learning_rate": 2.861502949474224e-05, + "loss": 0.2541, + "step": 33402 + }, + { + "epoch": 1.28, + "learning_rate": 2.861438830469351e-05, + "loss": 0.4654, + "step": 33403 + }, + { + "epoch": 1.28, + "learning_rate": 2.8613747114644783e-05, + "loss": 0.4812, + "step": 33404 + }, + { + "epoch": 1.28, + "learning_rate": 2.861310592459605e-05, + "loss": 0.3072, + "step": 33405 + }, + { + "epoch": 1.28, + "learning_rate": 2.8612464734547323e-05, + "loss": 0.3222, + "step": 33406 + }, + { + "epoch": 1.28, + "learning_rate": 2.8611823544498588e-05, + "loss": 0.3944, + "step": 33407 + }, + { + "epoch": 1.28, + "learning_rate": 2.861118235444986e-05, + "loss": 0.2655, + "step": 33408 + }, + { + "epoch": 1.28, + "learning_rate": 2.8610541164401128e-05, + "loss": 0.3366, + "step": 33409 + }, + { + "epoch": 1.28, + "learning_rate": 2.86098999743524e-05, + "loss": 0.4413, + "step": 33410 + }, + { + "epoch": 1.28, + "learning_rate": 2.8609258784303665e-05, + "loss": 0.3396, + "step": 33411 + }, + { + "epoch": 1.28, + "learning_rate": 2.860861759425494e-05, + "loss": 0.4585, + "step": 33412 + }, + { + "epoch": 1.28, + "learning_rate": 2.8607976404206212e-05, + "loss": 0.3173, + "step": 33413 + }, + { + "epoch": 1.28, + "learning_rate": 2.8607335214157477e-05, + "loss": 0.3079, + "step": 33414 + }, + { + "epoch": 1.28, + "learning_rate": 2.860669402410875e-05, + "loss": 0.3085, + "step": 33415 + }, + { + "epoch": 1.28, + "learning_rate": 2.8606052834060017e-05, + "loss": 0.4113, + "step": 33416 + }, + { + "epoch": 1.28, + "learning_rate": 2.860541164401129e-05, + "loss": 0.4298, + "step": 33417 + }, + { + "epoch": 1.28, + "learning_rate": 2.8604770453962553e-05, + "loss": 0.2988, + "step": 33418 + }, + { + "epoch": 1.28, + "learning_rate": 2.8604129263913825e-05, + "loss": 0.4677, + "step": 33419 + }, + { + "epoch": 1.28, + "learning_rate": 2.8603488073865093e-05, + "loss": 0.2881, + "step": 33420 + }, + { + "epoch": 1.28, + "learning_rate": 2.8602846883816365e-05, + "loss": 0.2663, + "step": 33421 + }, + { + "epoch": 1.28, + "learning_rate": 2.8602205693767637e-05, + "loss": 0.2958, + "step": 33422 + }, + { + "epoch": 1.29, + "learning_rate": 2.8601564503718902e-05, + "loss": 0.4184, + "step": 33423 + }, + { + "epoch": 1.29, + "learning_rate": 2.8600923313670174e-05, + "loss": 0.3733, + "step": 33424 + }, + { + "epoch": 1.29, + "learning_rate": 2.8600282123621442e-05, + "loss": 0.344, + "step": 33425 + }, + { + "epoch": 1.29, + "learning_rate": 2.8599640933572714e-05, + "loss": 0.5282, + "step": 33426 + }, + { + "epoch": 1.29, + "learning_rate": 2.859899974352398e-05, + "loss": 0.4673, + "step": 33427 + }, + { + "epoch": 1.29, + "learning_rate": 2.859835855347525e-05, + "loss": 0.4797, + "step": 33428 + }, + { + "epoch": 1.29, + "learning_rate": 2.859771736342652e-05, + "loss": 0.4877, + "step": 33429 + }, + { + "epoch": 1.29, + "learning_rate": 2.859707617337779e-05, + "loss": 0.4259, + "step": 33430 + }, + { + "epoch": 1.29, + "learning_rate": 2.8596434983329062e-05, + "loss": 0.4479, + "step": 33431 + }, + { + "epoch": 1.29, + "learning_rate": 2.859579379328033e-05, + "loss": 0.5074, + "step": 33432 + }, + { + "epoch": 1.29, + "learning_rate": 2.8595152603231602e-05, + "loss": 0.4012, + "step": 33433 + }, + { + "epoch": 1.29, + "learning_rate": 2.8594511413182867e-05, + "loss": 0.3727, + "step": 33434 + }, + { + "epoch": 1.29, + "learning_rate": 2.859387022313414e-05, + "loss": 0.3609, + "step": 33435 + }, + { + "epoch": 1.29, + "learning_rate": 2.8593229033085407e-05, + "loss": 0.265, + "step": 33436 + }, + { + "epoch": 1.29, + "learning_rate": 2.859258784303668e-05, + "loss": 0.3665, + "step": 33437 + }, + { + "epoch": 1.29, + "learning_rate": 2.8591946652987944e-05, + "loss": 0.3556, + "step": 33438 + }, + { + "epoch": 1.29, + "learning_rate": 2.8591305462939215e-05, + "loss": 0.2937, + "step": 33439 + }, + { + "epoch": 1.29, + "learning_rate": 2.8590664272890487e-05, + "loss": 0.5089, + "step": 33440 + }, + { + "epoch": 1.29, + "learning_rate": 2.8590023082841755e-05, + "loss": 0.41, + "step": 33441 + }, + { + "epoch": 1.29, + "learning_rate": 2.8589381892793027e-05, + "loss": 0.3289, + "step": 33442 + }, + { + "epoch": 1.29, + "learning_rate": 2.8588740702744292e-05, + "loss": 0.3481, + "step": 33443 + }, + { + "epoch": 1.29, + "learning_rate": 2.8588099512695564e-05, + "loss": 0.3033, + "step": 33444 + }, + { + "epoch": 1.29, + "learning_rate": 2.8587458322646832e-05, + "loss": 0.476, + "step": 33445 + }, + { + "epoch": 1.29, + "learning_rate": 2.8586817132598104e-05, + "loss": 0.548, + "step": 33446 + }, + { + "epoch": 1.29, + "learning_rate": 2.858617594254937e-05, + "loss": 0.4721, + "step": 33447 + }, + { + "epoch": 1.29, + "learning_rate": 2.8585534752500644e-05, + "loss": 0.2285, + "step": 33448 + }, + { + "epoch": 1.29, + "learning_rate": 2.8584893562451916e-05, + "loss": 0.2555, + "step": 33449 + }, + { + "epoch": 1.29, + "learning_rate": 2.858425237240318e-05, + "loss": 0.3461, + "step": 33450 + }, + { + "epoch": 1.29, + "learning_rate": 2.8583611182354452e-05, + "loss": 0.2907, + "step": 33451 + }, + { + "epoch": 1.29, + "learning_rate": 2.858296999230572e-05, + "loss": 0.3964, + "step": 33452 + }, + { + "epoch": 1.29, + "learning_rate": 2.8582328802256992e-05, + "loss": 0.2246, + "step": 33453 + }, + { + "epoch": 1.29, + "learning_rate": 2.8581687612208257e-05, + "loss": 0.4607, + "step": 33454 + }, + { + "epoch": 1.29, + "learning_rate": 2.858104642215953e-05, + "loss": 0.5077, + "step": 33455 + }, + { + "epoch": 1.29, + "learning_rate": 2.8580405232110797e-05, + "loss": 0.4594, + "step": 33456 + }, + { + "epoch": 1.29, + "learning_rate": 2.857976404206207e-05, + "loss": 0.339, + "step": 33457 + }, + { + "epoch": 1.29, + "learning_rate": 2.857912285201334e-05, + "loss": 0.3372, + "step": 33458 + }, + { + "epoch": 1.29, + "learning_rate": 2.8578481661964606e-05, + "loss": 0.4884, + "step": 33459 + }, + { + "epoch": 1.29, + "learning_rate": 2.8577840471915877e-05, + "loss": 0.3682, + "step": 33460 + }, + { + "epoch": 1.29, + "learning_rate": 2.8577199281867146e-05, + "loss": 0.2517, + "step": 33461 + }, + { + "epoch": 1.29, + "learning_rate": 2.8576558091818417e-05, + "loss": 0.3193, + "step": 33462 + }, + { + "epoch": 1.29, + "learning_rate": 2.8575916901769682e-05, + "loss": 0.3285, + "step": 33463 + }, + { + "epoch": 1.29, + "learning_rate": 2.8575275711720954e-05, + "loss": 0.5396, + "step": 33464 + }, + { + "epoch": 1.29, + "learning_rate": 2.857463452167223e-05, + "loss": 0.541, + "step": 33465 + }, + { + "epoch": 1.29, + "learning_rate": 2.8573993331623494e-05, + "loss": 0.4392, + "step": 33466 + }, + { + "epoch": 1.29, + "learning_rate": 2.8573352141574766e-05, + "loss": 0.2913, + "step": 33467 + }, + { + "epoch": 1.29, + "learning_rate": 2.8572710951526034e-05, + "loss": 0.3799, + "step": 33468 + }, + { + "epoch": 1.29, + "learning_rate": 2.8572069761477306e-05, + "loss": 0.4601, + "step": 33469 + }, + { + "epoch": 1.29, + "learning_rate": 2.857142857142857e-05, + "loss": 0.3561, + "step": 33470 + }, + { + "epoch": 1.29, + "learning_rate": 2.8570787381379843e-05, + "loss": 0.4196, + "step": 33471 + }, + { + "epoch": 1.29, + "learning_rate": 2.857014619133111e-05, + "loss": 0.3754, + "step": 33472 + }, + { + "epoch": 1.29, + "learning_rate": 2.8569505001282383e-05, + "loss": 0.3312, + "step": 33473 + }, + { + "epoch": 1.29, + "learning_rate": 2.8568863811233654e-05, + "loss": 0.47, + "step": 33474 + }, + { + "epoch": 1.29, + "learning_rate": 2.856822262118492e-05, + "loss": 0.4011, + "step": 33475 + }, + { + "epoch": 1.29, + "learning_rate": 2.856758143113619e-05, + "loss": 0.4472, + "step": 33476 + }, + { + "epoch": 1.29, + "learning_rate": 2.856694024108746e-05, + "loss": 0.4116, + "step": 33477 + }, + { + "epoch": 1.29, + "learning_rate": 2.856629905103873e-05, + "loss": 0.3441, + "step": 33478 + }, + { + "epoch": 1.29, + "learning_rate": 2.8565657860989996e-05, + "loss": 0.2683, + "step": 33479 + }, + { + "epoch": 1.29, + "learning_rate": 2.8565016670941268e-05, + "loss": 0.3194, + "step": 33480 + }, + { + "epoch": 1.29, + "learning_rate": 2.8564375480892536e-05, + "loss": 0.3469, + "step": 33481 + }, + { + "epoch": 1.29, + "learning_rate": 2.8563734290843808e-05, + "loss": 0.3306, + "step": 33482 + }, + { + "epoch": 1.29, + "learning_rate": 2.856309310079508e-05, + "loss": 0.2752, + "step": 33483 + }, + { + "epoch": 1.29, + "learning_rate": 2.8562451910746348e-05, + "loss": 0.4545, + "step": 33484 + }, + { + "epoch": 1.29, + "learning_rate": 2.856181072069762e-05, + "loss": 0.3633, + "step": 33485 + }, + { + "epoch": 1.29, + "learning_rate": 2.8561169530648884e-05, + "loss": 0.3891, + "step": 33486 + }, + { + "epoch": 1.29, + "learning_rate": 2.8560528340600156e-05, + "loss": 0.4042, + "step": 33487 + }, + { + "epoch": 1.29, + "learning_rate": 2.8559887150551424e-05, + "loss": 0.3601, + "step": 33488 + }, + { + "epoch": 1.29, + "learning_rate": 2.8559245960502696e-05, + "loss": 0.4114, + "step": 33489 + }, + { + "epoch": 1.29, + "learning_rate": 2.855860477045396e-05, + "loss": 0.314, + "step": 33490 + }, + { + "epoch": 1.29, + "learning_rate": 2.8557963580405233e-05, + "loss": 0.2632, + "step": 33491 + }, + { + "epoch": 1.29, + "learning_rate": 2.8557322390356505e-05, + "loss": 0.3263, + "step": 33492 + }, + { + "epoch": 1.29, + "learning_rate": 2.8556681200307773e-05, + "loss": 0.3464, + "step": 33493 + }, + { + "epoch": 1.29, + "learning_rate": 2.8556040010259045e-05, + "loss": 0.4334, + "step": 33494 + }, + { + "epoch": 1.29, + "learning_rate": 2.855539882021031e-05, + "loss": 0.4282, + "step": 33495 + }, + { + "epoch": 1.29, + "learning_rate": 2.855475763016158e-05, + "loss": 0.2947, + "step": 33496 + }, + { + "epoch": 1.29, + "learning_rate": 2.855411644011285e-05, + "loss": 0.3826, + "step": 33497 + }, + { + "epoch": 1.29, + "learning_rate": 2.855347525006412e-05, + "loss": 0.2693, + "step": 33498 + }, + { + "epoch": 1.29, + "learning_rate": 2.8552834060015386e-05, + "loss": 0.4742, + "step": 33499 + }, + { + "epoch": 1.29, + "learning_rate": 2.8552192869966658e-05, + "loss": 0.3431, + "step": 33500 + }, + { + "epoch": 1.29, + "learning_rate": 2.8551551679917933e-05, + "loss": 0.3426, + "step": 33501 + }, + { + "epoch": 1.29, + "learning_rate": 2.8550910489869198e-05, + "loss": 0.3558, + "step": 33502 + }, + { + "epoch": 1.29, + "learning_rate": 2.855026929982047e-05, + "loss": 0.4618, + "step": 33503 + }, + { + "epoch": 1.29, + "learning_rate": 2.8549628109771738e-05, + "loss": 0.3133, + "step": 33504 + }, + { + "epoch": 1.29, + "learning_rate": 2.854898691972301e-05, + "loss": 0.3424, + "step": 33505 + }, + { + "epoch": 1.29, + "learning_rate": 2.8548345729674275e-05, + "loss": 0.4656, + "step": 33506 + }, + { + "epoch": 1.29, + "learning_rate": 2.8547704539625546e-05, + "loss": 0.4659, + "step": 33507 + }, + { + "epoch": 1.29, + "learning_rate": 2.8547063349576815e-05, + "loss": 0.4017, + "step": 33508 + }, + { + "epoch": 1.29, + "learning_rate": 2.8546422159528086e-05, + "loss": 0.4089, + "step": 33509 + }, + { + "epoch": 1.29, + "learning_rate": 2.8545780969479358e-05, + "loss": 0.3682, + "step": 33510 + }, + { + "epoch": 1.29, + "learning_rate": 2.8545139779430623e-05, + "loss": 0.2636, + "step": 33511 + }, + { + "epoch": 1.29, + "learning_rate": 2.8544498589381895e-05, + "loss": 0.3321, + "step": 33512 + }, + { + "epoch": 1.29, + "learning_rate": 2.8543857399333163e-05, + "loss": 0.275, + "step": 33513 + }, + { + "epoch": 1.29, + "learning_rate": 2.8543216209284435e-05, + "loss": 0.3064, + "step": 33514 + }, + { + "epoch": 1.29, + "learning_rate": 2.85425750192357e-05, + "loss": 0.4409, + "step": 33515 + }, + { + "epoch": 1.29, + "learning_rate": 2.854193382918697e-05, + "loss": 0.32, + "step": 33516 + }, + { + "epoch": 1.29, + "learning_rate": 2.854129263913824e-05, + "loss": 0.4102, + "step": 33517 + }, + { + "epoch": 1.29, + "learning_rate": 2.854065144908951e-05, + "loss": 0.389, + "step": 33518 + }, + { + "epoch": 1.29, + "learning_rate": 2.8540010259040783e-05, + "loss": 0.3271, + "step": 33519 + }, + { + "epoch": 1.29, + "learning_rate": 2.853936906899205e-05, + "loss": 0.3501, + "step": 33520 + }, + { + "epoch": 1.29, + "learning_rate": 2.8538727878943323e-05, + "loss": 0.368, + "step": 33521 + }, + { + "epoch": 1.29, + "learning_rate": 2.853808668889459e-05, + "loss": 0.475, + "step": 33522 + }, + { + "epoch": 1.29, + "learning_rate": 2.853744549884586e-05, + "loss": 0.4288, + "step": 33523 + }, + { + "epoch": 1.29, + "learning_rate": 2.853680430879713e-05, + "loss": 0.3152, + "step": 33524 + }, + { + "epoch": 1.29, + "learning_rate": 2.85361631187484e-05, + "loss": 0.2978, + "step": 33525 + }, + { + "epoch": 1.29, + "learning_rate": 2.8535521928699665e-05, + "loss": 0.359, + "step": 33526 + }, + { + "epoch": 1.29, + "learning_rate": 2.8534880738650937e-05, + "loss": 0.3167, + "step": 33527 + }, + { + "epoch": 1.29, + "learning_rate": 2.853423954860221e-05, + "loss": 0.3325, + "step": 33528 + }, + { + "epoch": 1.29, + "learning_rate": 2.8533598358553477e-05, + "loss": 0.3663, + "step": 33529 + }, + { + "epoch": 1.29, + "learning_rate": 2.853295716850475e-05, + "loss": 0.3917, + "step": 33530 + }, + { + "epoch": 1.29, + "learning_rate": 2.8532315978456013e-05, + "loss": 0.3394, + "step": 33531 + }, + { + "epoch": 1.29, + "learning_rate": 2.8531674788407285e-05, + "loss": 0.3733, + "step": 33532 + }, + { + "epoch": 1.29, + "learning_rate": 2.8531033598358553e-05, + "loss": 0.5562, + "step": 33533 + }, + { + "epoch": 1.29, + "learning_rate": 2.8530392408309825e-05, + "loss": 0.3402, + "step": 33534 + }, + { + "epoch": 1.29, + "learning_rate": 2.852975121826109e-05, + "loss": 0.3508, + "step": 33535 + }, + { + "epoch": 1.29, + "learning_rate": 2.8529110028212365e-05, + "loss": 0.3997, + "step": 33536 + }, + { + "epoch": 1.29, + "learning_rate": 2.8528468838163637e-05, + "loss": 0.3616, + "step": 33537 + }, + { + "epoch": 1.29, + "learning_rate": 2.8527827648114902e-05, + "loss": 0.3836, + "step": 33538 + }, + { + "epoch": 1.29, + "learning_rate": 2.8527186458066174e-05, + "loss": 0.3609, + "step": 33539 + }, + { + "epoch": 1.29, + "learning_rate": 2.8526545268017442e-05, + "loss": 0.3591, + "step": 33540 + }, + { + "epoch": 1.29, + "learning_rate": 2.8525904077968714e-05, + "loss": 0.3321, + "step": 33541 + }, + { + "epoch": 1.29, + "learning_rate": 2.852526288791998e-05, + "loss": 0.3913, + "step": 33542 + }, + { + "epoch": 1.29, + "learning_rate": 2.852462169787125e-05, + "loss": 0.4509, + "step": 33543 + }, + { + "epoch": 1.29, + "learning_rate": 2.852398050782252e-05, + "loss": 0.2863, + "step": 33544 + }, + { + "epoch": 1.29, + "learning_rate": 2.852333931777379e-05, + "loss": 0.4997, + "step": 33545 + }, + { + "epoch": 1.29, + "learning_rate": 2.8522698127725062e-05, + "loss": 0.3905, + "step": 33546 + }, + { + "epoch": 1.29, + "learning_rate": 2.8522056937676327e-05, + "loss": 0.3662, + "step": 33547 + }, + { + "epoch": 1.29, + "learning_rate": 2.85214157476276e-05, + "loss": 0.4049, + "step": 33548 + }, + { + "epoch": 1.29, + "learning_rate": 2.8520774557578867e-05, + "loss": 0.586, + "step": 33549 + }, + { + "epoch": 1.29, + "learning_rate": 2.852013336753014e-05, + "loss": 0.3805, + "step": 33550 + }, + { + "epoch": 1.29, + "learning_rate": 2.8519492177481404e-05, + "loss": 0.3224, + "step": 33551 + }, + { + "epoch": 1.29, + "learning_rate": 2.8518850987432675e-05, + "loss": 0.5158, + "step": 33552 + }, + { + "epoch": 1.29, + "learning_rate": 2.8518209797383944e-05, + "loss": 0.4035, + "step": 33553 + }, + { + "epoch": 1.29, + "learning_rate": 2.8517568607335215e-05, + "loss": 0.428, + "step": 33554 + }, + { + "epoch": 1.29, + "learning_rate": 2.8516927417286487e-05, + "loss": 0.3775, + "step": 33555 + }, + { + "epoch": 1.29, + "learning_rate": 2.8516286227237756e-05, + "loss": 0.3561, + "step": 33556 + }, + { + "epoch": 1.29, + "learning_rate": 2.8515645037189027e-05, + "loss": 0.4622, + "step": 33557 + }, + { + "epoch": 1.29, + "learning_rate": 2.8515003847140292e-05, + "loss": 0.3987, + "step": 33558 + }, + { + "epoch": 1.29, + "learning_rate": 2.8514362657091564e-05, + "loss": 0.3012, + "step": 33559 + }, + { + "epoch": 1.29, + "learning_rate": 2.8513721467042832e-05, + "loss": 0.4212, + "step": 33560 + }, + { + "epoch": 1.29, + "learning_rate": 2.8513080276994104e-05, + "loss": 0.4127, + "step": 33561 + }, + { + "epoch": 1.29, + "learning_rate": 2.851243908694537e-05, + "loss": 0.397, + "step": 33562 + }, + { + "epoch": 1.29, + "learning_rate": 2.851179789689664e-05, + "loss": 0.3369, + "step": 33563 + }, + { + "epoch": 1.29, + "learning_rate": 2.8511156706847912e-05, + "loss": 0.2875, + "step": 33564 + }, + { + "epoch": 1.29, + "learning_rate": 2.851051551679918e-05, + "loss": 0.2875, + "step": 33565 + }, + { + "epoch": 1.29, + "learning_rate": 2.8509874326750452e-05, + "loss": 0.4554, + "step": 33566 + }, + { + "epoch": 1.29, + "learning_rate": 2.8509233136701717e-05, + "loss": 0.413, + "step": 33567 + }, + { + "epoch": 1.29, + "learning_rate": 2.850859194665299e-05, + "loss": 0.3709, + "step": 33568 + }, + { + "epoch": 1.29, + "learning_rate": 2.8507950756604257e-05, + "loss": 0.3763, + "step": 33569 + }, + { + "epoch": 1.29, + "learning_rate": 2.850730956655553e-05, + "loss": 0.2876, + "step": 33570 + }, + { + "epoch": 1.29, + "learning_rate": 2.8506668376506794e-05, + "loss": 0.439, + "step": 33571 + }, + { + "epoch": 1.29, + "learning_rate": 2.850602718645807e-05, + "loss": 0.326, + "step": 33572 + }, + { + "epoch": 1.29, + "learning_rate": 2.850538599640934e-05, + "loss": 0.3166, + "step": 33573 + }, + { + "epoch": 1.29, + "learning_rate": 2.8504744806360606e-05, + "loss": 0.3215, + "step": 33574 + }, + { + "epoch": 1.29, + "learning_rate": 2.8504103616311877e-05, + "loss": 0.4391, + "step": 33575 + }, + { + "epoch": 1.29, + "learning_rate": 2.8503462426263146e-05, + "loss": 0.4115, + "step": 33576 + }, + { + "epoch": 1.29, + "learning_rate": 2.8502821236214418e-05, + "loss": 0.2663, + "step": 33577 + }, + { + "epoch": 1.29, + "learning_rate": 2.8502180046165682e-05, + "loss": 0.3866, + "step": 33578 + }, + { + "epoch": 1.29, + "learning_rate": 2.8501538856116954e-05, + "loss": 0.3865, + "step": 33579 + }, + { + "epoch": 1.29, + "learning_rate": 2.8500897666068223e-05, + "loss": 0.3627, + "step": 33580 + }, + { + "epoch": 1.29, + "learning_rate": 2.8500256476019494e-05, + "loss": 0.4578, + "step": 33581 + }, + { + "epoch": 1.29, + "learning_rate": 2.8499615285970766e-05, + "loss": 0.4203, + "step": 33582 + }, + { + "epoch": 1.29, + "learning_rate": 2.849897409592203e-05, + "loss": 0.3817, + "step": 33583 + }, + { + "epoch": 1.29, + "learning_rate": 2.8498332905873303e-05, + "loss": 0.3469, + "step": 33584 + }, + { + "epoch": 1.29, + "learning_rate": 2.849769171582457e-05, + "loss": 0.3218, + "step": 33585 + }, + { + "epoch": 1.29, + "learning_rate": 2.8497050525775843e-05, + "loss": 0.4948, + "step": 33586 + }, + { + "epoch": 1.29, + "learning_rate": 2.8496409335727108e-05, + "loss": 0.3845, + "step": 33587 + }, + { + "epoch": 1.29, + "learning_rate": 2.849576814567838e-05, + "loss": 0.4337, + "step": 33588 + }, + { + "epoch": 1.29, + "learning_rate": 2.8495126955629648e-05, + "loss": 0.3847, + "step": 33589 + }, + { + "epoch": 1.29, + "learning_rate": 2.849448576558092e-05, + "loss": 0.3669, + "step": 33590 + }, + { + "epoch": 1.29, + "learning_rate": 2.849384457553219e-05, + "loss": 0.4832, + "step": 33591 + }, + { + "epoch": 1.29, + "learning_rate": 2.849320338548346e-05, + "loss": 0.5428, + "step": 33592 + }, + { + "epoch": 1.29, + "learning_rate": 2.849256219543473e-05, + "loss": 0.5672, + "step": 33593 + }, + { + "epoch": 1.29, + "learning_rate": 2.8491921005385996e-05, + "loss": 0.4129, + "step": 33594 + }, + { + "epoch": 1.29, + "learning_rate": 2.8491279815337268e-05, + "loss": 0.3592, + "step": 33595 + }, + { + "epoch": 1.29, + "learning_rate": 2.8490638625288536e-05, + "loss": 0.4562, + "step": 33596 + }, + { + "epoch": 1.29, + "learning_rate": 2.8489997435239808e-05, + "loss": 0.4296, + "step": 33597 + }, + { + "epoch": 1.29, + "learning_rate": 2.8489356245191073e-05, + "loss": 0.3384, + "step": 33598 + }, + { + "epoch": 1.29, + "learning_rate": 2.8488715055142344e-05, + "loss": 0.4784, + "step": 33599 + }, + { + "epoch": 1.29, + "learning_rate": 2.8488073865093616e-05, + "loss": 0.331, + "step": 33600 + }, + { + "epoch": 1.29, + "learning_rate": 2.8487432675044885e-05, + "loss": 0.2787, + "step": 33601 + }, + { + "epoch": 1.29, + "learning_rate": 2.8486791484996156e-05, + "loss": 0.3444, + "step": 33602 + }, + { + "epoch": 1.29, + "learning_rate": 2.848615029494742e-05, + "loss": 0.4004, + "step": 33603 + }, + { + "epoch": 1.29, + "learning_rate": 2.8485509104898693e-05, + "loss": 0.4439, + "step": 33604 + }, + { + "epoch": 1.29, + "learning_rate": 2.848486791484996e-05, + "loss": 0.4077, + "step": 33605 + }, + { + "epoch": 1.29, + "learning_rate": 2.8484226724801233e-05, + "loss": 0.4984, + "step": 33606 + }, + { + "epoch": 1.29, + "learning_rate": 2.8483585534752498e-05, + "loss": 0.3601, + "step": 33607 + }, + { + "epoch": 1.29, + "learning_rate": 2.8482944344703773e-05, + "loss": 0.3855, + "step": 33608 + }, + { + "epoch": 1.29, + "learning_rate": 2.8482303154655045e-05, + "loss": 0.2354, + "step": 33609 + }, + { + "epoch": 1.29, + "learning_rate": 2.848166196460631e-05, + "loss": 0.373, + "step": 33610 + }, + { + "epoch": 1.29, + "learning_rate": 2.848102077455758e-05, + "loss": 0.4924, + "step": 33611 + }, + { + "epoch": 1.29, + "learning_rate": 2.848037958450885e-05, + "loss": 0.2633, + "step": 33612 + }, + { + "epoch": 1.29, + "learning_rate": 2.847973839446012e-05, + "loss": 0.3001, + "step": 33613 + }, + { + "epoch": 1.29, + "learning_rate": 2.8479097204411386e-05, + "loss": 0.344, + "step": 33614 + }, + { + "epoch": 1.29, + "learning_rate": 2.8478456014362658e-05, + "loss": 0.3209, + "step": 33615 + }, + { + "epoch": 1.29, + "learning_rate": 2.8477814824313926e-05, + "loss": 0.3811, + "step": 33616 + }, + { + "epoch": 1.29, + "learning_rate": 2.8477173634265198e-05, + "loss": 0.3507, + "step": 33617 + }, + { + "epoch": 1.29, + "learning_rate": 2.847653244421647e-05, + "loss": 0.5101, + "step": 33618 + }, + { + "epoch": 1.29, + "learning_rate": 2.8475891254167735e-05, + "loss": 0.4457, + "step": 33619 + }, + { + "epoch": 1.29, + "learning_rate": 2.8475250064119006e-05, + "loss": 0.483, + "step": 33620 + }, + { + "epoch": 1.29, + "learning_rate": 2.8474608874070275e-05, + "loss": 0.2998, + "step": 33621 + }, + { + "epoch": 1.29, + "learning_rate": 2.8473967684021547e-05, + "loss": 0.3432, + "step": 33622 + }, + { + "epoch": 1.29, + "learning_rate": 2.847332649397281e-05, + "loss": 0.3567, + "step": 33623 + }, + { + "epoch": 1.29, + "learning_rate": 2.8472685303924083e-05, + "loss": 0.4144, + "step": 33624 + }, + { + "epoch": 1.29, + "learning_rate": 2.847204411387535e-05, + "loss": 0.2552, + "step": 33625 + }, + { + "epoch": 1.29, + "learning_rate": 2.8471402923826623e-05, + "loss": 0.3174, + "step": 33626 + }, + { + "epoch": 1.29, + "learning_rate": 2.8470761733777895e-05, + "loss": 0.4476, + "step": 33627 + }, + { + "epoch": 1.29, + "learning_rate": 2.8470120543729163e-05, + "loss": 0.3739, + "step": 33628 + }, + { + "epoch": 1.29, + "learning_rate": 2.8469479353680435e-05, + "loss": 0.4421, + "step": 33629 + }, + { + "epoch": 1.29, + "learning_rate": 2.84688381636317e-05, + "loss": 0.294, + "step": 33630 + }, + { + "epoch": 1.29, + "learning_rate": 2.846819697358297e-05, + "loss": 0.3391, + "step": 33631 + }, + { + "epoch": 1.29, + "learning_rate": 2.846755578353424e-05, + "loss": 0.3334, + "step": 33632 + }, + { + "epoch": 1.29, + "learning_rate": 2.846691459348551e-05, + "loss": 0.3371, + "step": 33633 + }, + { + "epoch": 1.29, + "learning_rate": 2.8466273403436777e-05, + "loss": 0.4175, + "step": 33634 + }, + { + "epoch": 1.29, + "learning_rate": 2.846563221338805e-05, + "loss": 0.4031, + "step": 33635 + }, + { + "epoch": 1.29, + "learning_rate": 2.846499102333932e-05, + "loss": 0.4527, + "step": 33636 + }, + { + "epoch": 1.29, + "learning_rate": 2.846434983329059e-05, + "loss": 0.3511, + "step": 33637 + }, + { + "epoch": 1.29, + "learning_rate": 2.846370864324186e-05, + "loss": 0.4947, + "step": 33638 + }, + { + "epoch": 1.29, + "learning_rate": 2.8463067453193125e-05, + "loss": 0.3956, + "step": 33639 + }, + { + "epoch": 1.29, + "learning_rate": 2.8462426263144397e-05, + "loss": 0.315, + "step": 33640 + }, + { + "epoch": 1.29, + "learning_rate": 2.8461785073095665e-05, + "loss": 0.5056, + "step": 33641 + }, + { + "epoch": 1.29, + "learning_rate": 2.8461143883046937e-05, + "loss": 0.3939, + "step": 33642 + }, + { + "epoch": 1.29, + "learning_rate": 2.8460502692998202e-05, + "loss": 0.4034, + "step": 33643 + }, + { + "epoch": 1.29, + "learning_rate": 2.8459861502949477e-05, + "loss": 0.5086, + "step": 33644 + }, + { + "epoch": 1.29, + "learning_rate": 2.845922031290075e-05, + "loss": 0.3417, + "step": 33645 + }, + { + "epoch": 1.29, + "learning_rate": 2.8458579122852014e-05, + "loss": 0.3649, + "step": 33646 + }, + { + "epoch": 1.29, + "learning_rate": 2.8457937932803285e-05, + "loss": 0.4083, + "step": 33647 + }, + { + "epoch": 1.29, + "learning_rate": 2.8457296742754554e-05, + "loss": 0.5391, + "step": 33648 + }, + { + "epoch": 1.29, + "learning_rate": 2.8456655552705825e-05, + "loss": 0.2556, + "step": 33649 + }, + { + "epoch": 1.29, + "learning_rate": 2.845601436265709e-05, + "loss": 0.4406, + "step": 33650 + }, + { + "epoch": 1.29, + "learning_rate": 2.8455373172608362e-05, + "loss": 0.4721, + "step": 33651 + }, + { + "epoch": 1.29, + "learning_rate": 2.8454731982559634e-05, + "loss": 0.3319, + "step": 33652 + }, + { + "epoch": 1.29, + "learning_rate": 2.8454090792510902e-05, + "loss": 0.4633, + "step": 33653 + }, + { + "epoch": 1.29, + "learning_rate": 2.8453449602462174e-05, + "loss": 0.3523, + "step": 33654 + }, + { + "epoch": 1.29, + "learning_rate": 2.845280841241344e-05, + "loss": 0.3111, + "step": 33655 + }, + { + "epoch": 1.29, + "learning_rate": 2.845216722236471e-05, + "loss": 0.4045, + "step": 33656 + }, + { + "epoch": 1.29, + "learning_rate": 2.845152603231598e-05, + "loss": 0.3477, + "step": 33657 + }, + { + "epoch": 1.29, + "learning_rate": 2.845088484226725e-05, + "loss": 0.4747, + "step": 33658 + }, + { + "epoch": 1.29, + "learning_rate": 2.8450243652218515e-05, + "loss": 0.4302, + "step": 33659 + }, + { + "epoch": 1.29, + "learning_rate": 2.8449602462169787e-05, + "loss": 0.3279, + "step": 33660 + }, + { + "epoch": 1.29, + "learning_rate": 2.8448961272121062e-05, + "loss": 0.3591, + "step": 33661 + }, + { + "epoch": 1.29, + "learning_rate": 2.8448320082072327e-05, + "loss": 0.3341, + "step": 33662 + }, + { + "epoch": 1.29, + "learning_rate": 2.84476788920236e-05, + "loss": 0.3435, + "step": 33663 + }, + { + "epoch": 1.29, + "learning_rate": 2.8447037701974867e-05, + "loss": 0.3056, + "step": 33664 + }, + { + "epoch": 1.29, + "learning_rate": 2.844639651192614e-05, + "loss": 0.3712, + "step": 33665 + }, + { + "epoch": 1.29, + "learning_rate": 2.8445755321877404e-05, + "loss": 0.2973, + "step": 33666 + }, + { + "epoch": 1.29, + "learning_rate": 2.8445114131828676e-05, + "loss": 0.3968, + "step": 33667 + }, + { + "epoch": 1.29, + "learning_rate": 2.8444472941779944e-05, + "loss": 0.4291, + "step": 33668 + }, + { + "epoch": 1.29, + "learning_rate": 2.8443831751731216e-05, + "loss": 0.2846, + "step": 33669 + }, + { + "epoch": 1.29, + "learning_rate": 2.8443190561682487e-05, + "loss": 0.2443, + "step": 33670 + }, + { + "epoch": 1.29, + "learning_rate": 2.8442549371633752e-05, + "loss": 0.3853, + "step": 33671 + }, + { + "epoch": 1.29, + "learning_rate": 2.8441908181585024e-05, + "loss": 0.3588, + "step": 33672 + }, + { + "epoch": 1.29, + "learning_rate": 2.8441266991536292e-05, + "loss": 0.2528, + "step": 33673 + }, + { + "epoch": 1.29, + "learning_rate": 2.8440625801487564e-05, + "loss": 0.2741, + "step": 33674 + }, + { + "epoch": 1.29, + "learning_rate": 2.843998461143883e-05, + "loss": 0.4161, + "step": 33675 + }, + { + "epoch": 1.29, + "learning_rate": 2.84393434213901e-05, + "loss": 0.3153, + "step": 33676 + }, + { + "epoch": 1.29, + "learning_rate": 2.843870223134137e-05, + "loss": 0.4465, + "step": 33677 + }, + { + "epoch": 1.29, + "learning_rate": 2.843806104129264e-05, + "loss": 0.3937, + "step": 33678 + }, + { + "epoch": 1.29, + "learning_rate": 2.8437419851243912e-05, + "loss": 0.4584, + "step": 33679 + }, + { + "epoch": 1.29, + "learning_rate": 2.843677866119518e-05, + "loss": 0.5107, + "step": 33680 + }, + { + "epoch": 1.29, + "learning_rate": 2.8436137471146452e-05, + "loss": 0.3516, + "step": 33681 + }, + { + "epoch": 1.29, + "learning_rate": 2.8435496281097717e-05, + "loss": 0.3179, + "step": 33682 + }, + { + "epoch": 1.3, + "learning_rate": 2.843485509104899e-05, + "loss": 0.361, + "step": 33683 + }, + { + "epoch": 1.3, + "learning_rate": 2.8434213901000257e-05, + "loss": 0.3602, + "step": 33684 + }, + { + "epoch": 1.3, + "learning_rate": 2.843357271095153e-05, + "loss": 0.3324, + "step": 33685 + }, + { + "epoch": 1.3, + "learning_rate": 2.8432931520902794e-05, + "loss": 0.3689, + "step": 33686 + }, + { + "epoch": 1.3, + "learning_rate": 2.8432290330854066e-05, + "loss": 0.3901, + "step": 33687 + }, + { + "epoch": 1.3, + "learning_rate": 2.8431649140805338e-05, + "loss": 0.3989, + "step": 33688 + }, + { + "epoch": 1.3, + "learning_rate": 2.8431007950756606e-05, + "loss": 0.3307, + "step": 33689 + }, + { + "epoch": 1.3, + "learning_rate": 2.8430366760707878e-05, + "loss": 0.3406, + "step": 33690 + }, + { + "epoch": 1.3, + "learning_rate": 2.8429725570659143e-05, + "loss": 0.4063, + "step": 33691 + }, + { + "epoch": 1.3, + "learning_rate": 2.8429084380610414e-05, + "loss": 0.3172, + "step": 33692 + }, + { + "epoch": 1.3, + "learning_rate": 2.8428443190561683e-05, + "loss": 0.4219, + "step": 33693 + }, + { + "epoch": 1.3, + "learning_rate": 2.8427802000512954e-05, + "loss": 0.3947, + "step": 33694 + }, + { + "epoch": 1.3, + "learning_rate": 2.842716081046422e-05, + "loss": 0.2845, + "step": 33695 + }, + { + "epoch": 1.3, + "learning_rate": 2.8426519620415494e-05, + "loss": 0.348, + "step": 33696 + }, + { + "epoch": 1.3, + "learning_rate": 2.8425878430366766e-05, + "loss": 0.3756, + "step": 33697 + }, + { + "epoch": 1.3, + "learning_rate": 2.842523724031803e-05, + "loss": 0.3813, + "step": 33698 + }, + { + "epoch": 1.3, + "learning_rate": 2.8424596050269303e-05, + "loss": 0.3756, + "step": 33699 + }, + { + "epoch": 1.3, + "learning_rate": 2.842395486022057e-05, + "loss": 0.3728, + "step": 33700 + }, + { + "epoch": 1.3, + "learning_rate": 2.8423313670171843e-05, + "loss": 0.5127, + "step": 33701 + }, + { + "epoch": 1.3, + "learning_rate": 2.8422672480123108e-05, + "loss": 0.4005, + "step": 33702 + }, + { + "epoch": 1.3, + "learning_rate": 2.842203129007438e-05, + "loss": 0.2759, + "step": 33703 + }, + { + "epoch": 1.3, + "learning_rate": 2.8421390100025648e-05, + "loss": 0.3954, + "step": 33704 + }, + { + "epoch": 1.3, + "learning_rate": 2.842074890997692e-05, + "loss": 0.463, + "step": 33705 + }, + { + "epoch": 1.3, + "learning_rate": 2.842010771992819e-05, + "loss": 0.4165, + "step": 33706 + }, + { + "epoch": 1.3, + "learning_rate": 2.8419466529879456e-05, + "loss": 0.3577, + "step": 33707 + }, + { + "epoch": 1.3, + "learning_rate": 2.8418825339830728e-05, + "loss": 0.5023, + "step": 33708 + }, + { + "epoch": 1.3, + "learning_rate": 2.8418184149781996e-05, + "loss": 0.508, + "step": 33709 + }, + { + "epoch": 1.3, + "learning_rate": 2.8417542959733268e-05, + "loss": 0.4552, + "step": 33710 + }, + { + "epoch": 1.3, + "learning_rate": 2.8416901769684533e-05, + "loss": 0.4866, + "step": 33711 + }, + { + "epoch": 1.3, + "learning_rate": 2.8416260579635805e-05, + "loss": 0.4763, + "step": 33712 + }, + { + "epoch": 1.3, + "learning_rate": 2.8415619389587073e-05, + "loss": 0.3069, + "step": 33713 + }, + { + "epoch": 1.3, + "learning_rate": 2.8414978199538345e-05, + "loss": 0.3104, + "step": 33714 + }, + { + "epoch": 1.3, + "learning_rate": 2.8414337009489616e-05, + "loss": 0.4128, + "step": 33715 + }, + { + "epoch": 1.3, + "learning_rate": 2.8413695819440885e-05, + "loss": 0.4253, + "step": 33716 + }, + { + "epoch": 1.3, + "learning_rate": 2.8413054629392156e-05, + "loss": 0.3512, + "step": 33717 + }, + { + "epoch": 1.3, + "learning_rate": 2.841241343934342e-05, + "loss": 0.3894, + "step": 33718 + }, + { + "epoch": 1.3, + "learning_rate": 2.8411772249294693e-05, + "loss": 0.2429, + "step": 33719 + }, + { + "epoch": 1.3, + "learning_rate": 2.841113105924596e-05, + "loss": 0.4383, + "step": 33720 + }, + { + "epoch": 1.3, + "learning_rate": 2.8410489869197233e-05, + "loss": 0.336, + "step": 33721 + }, + { + "epoch": 1.3, + "learning_rate": 2.8409848679148498e-05, + "loss": 0.3843, + "step": 33722 + }, + { + "epoch": 1.3, + "learning_rate": 2.840920748909977e-05, + "loss": 0.3944, + "step": 33723 + }, + { + "epoch": 1.3, + "learning_rate": 2.840856629905104e-05, + "loss": 0.3345, + "step": 33724 + }, + { + "epoch": 1.3, + "learning_rate": 2.840792510900231e-05, + "loss": 0.4184, + "step": 33725 + }, + { + "epoch": 1.3, + "learning_rate": 2.840728391895358e-05, + "loss": 0.4347, + "step": 33726 + }, + { + "epoch": 1.3, + "learning_rate": 2.8406642728904846e-05, + "loss": 0.4607, + "step": 33727 + }, + { + "epoch": 1.3, + "learning_rate": 2.8406001538856118e-05, + "loss": 0.4375, + "step": 33728 + }, + { + "epoch": 1.3, + "learning_rate": 2.8405360348807386e-05, + "loss": 0.4428, + "step": 33729 + }, + { + "epoch": 1.3, + "learning_rate": 2.8404719158758658e-05, + "loss": 0.5149, + "step": 33730 + }, + { + "epoch": 1.3, + "learning_rate": 2.8404077968709923e-05, + "loss": 0.4064, + "step": 33731 + }, + { + "epoch": 1.3, + "learning_rate": 2.8403436778661198e-05, + "loss": 0.3015, + "step": 33732 + }, + { + "epoch": 1.3, + "learning_rate": 2.840279558861247e-05, + "loss": 0.3148, + "step": 33733 + }, + { + "epoch": 1.3, + "learning_rate": 2.8402154398563735e-05, + "loss": 0.3073, + "step": 33734 + }, + { + "epoch": 1.3, + "learning_rate": 2.8401513208515007e-05, + "loss": 0.4253, + "step": 33735 + }, + { + "epoch": 1.3, + "learning_rate": 2.8400872018466275e-05, + "loss": 0.413, + "step": 33736 + }, + { + "epoch": 1.3, + "learning_rate": 2.8400230828417547e-05, + "loss": 0.3968, + "step": 33737 + }, + { + "epoch": 1.3, + "learning_rate": 2.839958963836881e-05, + "loss": 0.3847, + "step": 33738 + }, + { + "epoch": 1.3, + "learning_rate": 2.8398948448320083e-05, + "loss": 0.2495, + "step": 33739 + }, + { + "epoch": 1.3, + "learning_rate": 2.839830725827135e-05, + "loss": 0.3248, + "step": 33740 + }, + { + "epoch": 1.3, + "learning_rate": 2.8397666068222623e-05, + "loss": 0.4647, + "step": 33741 + }, + { + "epoch": 1.3, + "learning_rate": 2.8397024878173895e-05, + "loss": 0.3388, + "step": 33742 + }, + { + "epoch": 1.3, + "learning_rate": 2.839638368812516e-05, + "loss": 0.4339, + "step": 33743 + }, + { + "epoch": 1.3, + "learning_rate": 2.839574249807643e-05, + "loss": 0.2683, + "step": 33744 + }, + { + "epoch": 1.3, + "learning_rate": 2.83951013080277e-05, + "loss": 0.2881, + "step": 33745 + }, + { + "epoch": 1.3, + "learning_rate": 2.8394460117978972e-05, + "loss": 0.3383, + "step": 33746 + }, + { + "epoch": 1.3, + "learning_rate": 2.8393818927930237e-05, + "loss": 0.2618, + "step": 33747 + }, + { + "epoch": 1.3, + "learning_rate": 2.839317773788151e-05, + "loss": 0.4414, + "step": 33748 + }, + { + "epoch": 1.3, + "learning_rate": 2.8392536547832777e-05, + "loss": 0.4592, + "step": 33749 + }, + { + "epoch": 1.3, + "learning_rate": 2.839189535778405e-05, + "loss": 0.4417, + "step": 33750 + }, + { + "epoch": 1.3, + "learning_rate": 2.839125416773532e-05, + "loss": 0.5466, + "step": 33751 + }, + { + "epoch": 1.3, + "learning_rate": 2.839061297768659e-05, + "loss": 0.3634, + "step": 33752 + }, + { + "epoch": 1.3, + "learning_rate": 2.838997178763786e-05, + "loss": 0.434, + "step": 33753 + }, + { + "epoch": 1.3, + "learning_rate": 2.8389330597589125e-05, + "loss": 0.3383, + "step": 33754 + }, + { + "epoch": 1.3, + "learning_rate": 2.8388689407540397e-05, + "loss": 0.3552, + "step": 33755 + }, + { + "epoch": 1.3, + "learning_rate": 2.8388048217491665e-05, + "loss": 0.3517, + "step": 33756 + }, + { + "epoch": 1.3, + "learning_rate": 2.8387407027442937e-05, + "loss": 0.4489, + "step": 33757 + }, + { + "epoch": 1.3, + "learning_rate": 2.8386765837394202e-05, + "loss": 0.3513, + "step": 33758 + }, + { + "epoch": 1.3, + "learning_rate": 2.8386124647345474e-05, + "loss": 0.4787, + "step": 33759 + }, + { + "epoch": 1.3, + "learning_rate": 2.8385483457296745e-05, + "loss": 0.3224, + "step": 33760 + }, + { + "epoch": 1.3, + "learning_rate": 2.8384842267248014e-05, + "loss": 0.4626, + "step": 33761 + }, + { + "epoch": 1.3, + "learning_rate": 2.8384201077199285e-05, + "loss": 0.5343, + "step": 33762 + }, + { + "epoch": 1.3, + "learning_rate": 2.838355988715055e-05, + "loss": 0.437, + "step": 33763 + }, + { + "epoch": 1.3, + "learning_rate": 2.8382918697101822e-05, + "loss": 0.3896, + "step": 33764 + }, + { + "epoch": 1.3, + "learning_rate": 2.838227750705309e-05, + "loss": 0.3457, + "step": 33765 + }, + { + "epoch": 1.3, + "learning_rate": 2.8381636317004362e-05, + "loss": 0.4158, + "step": 33766 + }, + { + "epoch": 1.3, + "learning_rate": 2.8380995126955627e-05, + "loss": 0.4343, + "step": 33767 + }, + { + "epoch": 1.3, + "learning_rate": 2.8380353936906902e-05, + "loss": 0.3622, + "step": 33768 + }, + { + "epoch": 1.3, + "learning_rate": 2.8379712746858174e-05, + "loss": 0.3675, + "step": 33769 + }, + { + "epoch": 1.3, + "learning_rate": 2.837907155680944e-05, + "loss": 0.3588, + "step": 33770 + }, + { + "epoch": 1.3, + "learning_rate": 2.837843036676071e-05, + "loss": 0.5355, + "step": 33771 + }, + { + "epoch": 1.3, + "learning_rate": 2.837778917671198e-05, + "loss": 0.5436, + "step": 33772 + }, + { + "epoch": 1.3, + "learning_rate": 2.837714798666325e-05, + "loss": 0.5667, + "step": 33773 + }, + { + "epoch": 1.3, + "learning_rate": 2.8376506796614515e-05, + "loss": 0.4063, + "step": 33774 + }, + { + "epoch": 1.3, + "learning_rate": 2.8375865606565787e-05, + "loss": 0.3849, + "step": 33775 + }, + { + "epoch": 1.3, + "learning_rate": 2.8375224416517055e-05, + "loss": 0.3949, + "step": 33776 + }, + { + "epoch": 1.3, + "learning_rate": 2.8374583226468327e-05, + "loss": 0.4316, + "step": 33777 + }, + { + "epoch": 1.3, + "learning_rate": 2.83739420364196e-05, + "loss": 0.4161, + "step": 33778 + }, + { + "epoch": 1.3, + "learning_rate": 2.8373300846370864e-05, + "loss": 0.4581, + "step": 33779 + }, + { + "epoch": 1.3, + "learning_rate": 2.8372659656322136e-05, + "loss": 0.2667, + "step": 33780 + }, + { + "epoch": 1.3, + "learning_rate": 2.8372018466273404e-05, + "loss": 0.418, + "step": 33781 + }, + { + "epoch": 1.3, + "learning_rate": 2.8371377276224676e-05, + "loss": 0.3523, + "step": 33782 + }, + { + "epoch": 1.3, + "learning_rate": 2.837073608617594e-05, + "loss": 0.2985, + "step": 33783 + }, + { + "epoch": 1.3, + "learning_rate": 2.8370094896127212e-05, + "loss": 0.3702, + "step": 33784 + }, + { + "epoch": 1.3, + "learning_rate": 2.836945370607848e-05, + "loss": 0.3874, + "step": 33785 + }, + { + "epoch": 1.3, + "learning_rate": 2.8368812516029752e-05, + "loss": 0.5308, + "step": 33786 + }, + { + "epoch": 1.3, + "learning_rate": 2.8368171325981024e-05, + "loss": 0.37, + "step": 33787 + }, + { + "epoch": 1.3, + "learning_rate": 2.8367530135932292e-05, + "loss": 0.3434, + "step": 33788 + }, + { + "epoch": 1.3, + "learning_rate": 2.8366888945883564e-05, + "loss": 0.3857, + "step": 33789 + }, + { + "epoch": 1.3, + "learning_rate": 2.836624775583483e-05, + "loss": 0.3424, + "step": 33790 + }, + { + "epoch": 1.3, + "learning_rate": 2.83656065657861e-05, + "loss": 0.432, + "step": 33791 + }, + { + "epoch": 1.3, + "learning_rate": 2.836496537573737e-05, + "loss": 0.2839, + "step": 33792 + }, + { + "epoch": 1.3, + "learning_rate": 2.836432418568864e-05, + "loss": 0.4145, + "step": 33793 + }, + { + "epoch": 1.3, + "learning_rate": 2.8363682995639906e-05, + "loss": 0.3413, + "step": 33794 + }, + { + "epoch": 1.3, + "learning_rate": 2.8363041805591177e-05, + "loss": 0.4714, + "step": 33795 + }, + { + "epoch": 1.3, + "learning_rate": 2.836240061554245e-05, + "loss": 0.3601, + "step": 33796 + }, + { + "epoch": 1.3, + "learning_rate": 2.8361759425493717e-05, + "loss": 0.4431, + "step": 33797 + }, + { + "epoch": 1.3, + "learning_rate": 2.836111823544499e-05, + "loss": 0.2794, + "step": 33798 + }, + { + "epoch": 1.3, + "learning_rate": 2.8360477045396254e-05, + "loss": 0.5702, + "step": 33799 + }, + { + "epoch": 1.3, + "learning_rate": 2.8359835855347526e-05, + "loss": 0.2757, + "step": 33800 + }, + { + "epoch": 1.3, + "learning_rate": 2.8359194665298794e-05, + "loss": 0.3632, + "step": 33801 + }, + { + "epoch": 1.3, + "learning_rate": 2.8358553475250066e-05, + "loss": 0.4017, + "step": 33802 + }, + { + "epoch": 1.3, + "learning_rate": 2.835791228520133e-05, + "loss": 0.4203, + "step": 33803 + }, + { + "epoch": 1.3, + "learning_rate": 2.8357271095152606e-05, + "loss": 0.3152, + "step": 33804 + }, + { + "epoch": 1.3, + "learning_rate": 2.8356629905103878e-05, + "loss": 0.3686, + "step": 33805 + }, + { + "epoch": 1.3, + "learning_rate": 2.8355988715055143e-05, + "loss": 0.3786, + "step": 33806 + }, + { + "epoch": 1.3, + "learning_rate": 2.8355347525006414e-05, + "loss": 0.3555, + "step": 33807 + }, + { + "epoch": 1.3, + "learning_rate": 2.8354706334957683e-05, + "loss": 0.2988, + "step": 33808 + }, + { + "epoch": 1.3, + "learning_rate": 2.8354065144908954e-05, + "loss": 0.3998, + "step": 33809 + }, + { + "epoch": 1.3, + "learning_rate": 2.835342395486022e-05, + "loss": 0.3752, + "step": 33810 + }, + { + "epoch": 1.3, + "learning_rate": 2.835278276481149e-05, + "loss": 0.4773, + "step": 33811 + }, + { + "epoch": 1.3, + "learning_rate": 2.835214157476276e-05, + "loss": 0.3589, + "step": 33812 + }, + { + "epoch": 1.3, + "learning_rate": 2.835150038471403e-05, + "loss": 0.3017, + "step": 33813 + }, + { + "epoch": 1.3, + "learning_rate": 2.8350859194665303e-05, + "loss": 0.4412, + "step": 33814 + }, + { + "epoch": 1.3, + "learning_rate": 2.8350218004616568e-05, + "loss": 0.3965, + "step": 33815 + }, + { + "epoch": 1.3, + "learning_rate": 2.834957681456784e-05, + "loss": 0.3157, + "step": 33816 + }, + { + "epoch": 1.3, + "learning_rate": 2.8348935624519108e-05, + "loss": 0.391, + "step": 33817 + }, + { + "epoch": 1.3, + "learning_rate": 2.834829443447038e-05, + "loss": 0.3005, + "step": 33818 + }, + { + "epoch": 1.3, + "learning_rate": 2.8347653244421644e-05, + "loss": 0.2974, + "step": 33819 + }, + { + "epoch": 1.3, + "learning_rate": 2.834701205437292e-05, + "loss": 0.3566, + "step": 33820 + }, + { + "epoch": 1.3, + "learning_rate": 2.8346370864324184e-05, + "loss": 0.3605, + "step": 33821 + }, + { + "epoch": 1.3, + "learning_rate": 2.8345729674275456e-05, + "loss": 0.4201, + "step": 33822 + }, + { + "epoch": 1.3, + "learning_rate": 2.8345088484226728e-05, + "loss": 0.3382, + "step": 33823 + }, + { + "epoch": 1.3, + "learning_rate": 2.8344447294177996e-05, + "loss": 0.3144, + "step": 33824 + }, + { + "epoch": 1.3, + "learning_rate": 2.8343806104129268e-05, + "loss": 0.4686, + "step": 33825 + }, + { + "epoch": 1.3, + "learning_rate": 2.8343164914080533e-05, + "loss": 0.2971, + "step": 33826 + }, + { + "epoch": 1.3, + "learning_rate": 2.8342523724031805e-05, + "loss": 0.4003, + "step": 33827 + }, + { + "epoch": 1.3, + "learning_rate": 2.8341882533983073e-05, + "loss": 0.3509, + "step": 33828 + }, + { + "epoch": 1.3, + "learning_rate": 2.8341241343934345e-05, + "loss": 0.361, + "step": 33829 + }, + { + "epoch": 1.3, + "learning_rate": 2.834060015388561e-05, + "loss": 0.4048, + "step": 33830 + }, + { + "epoch": 1.3, + "learning_rate": 2.833995896383688e-05, + "loss": 0.4685, + "step": 33831 + }, + { + "epoch": 1.3, + "learning_rate": 2.8339317773788153e-05, + "loss": 0.5326, + "step": 33832 + }, + { + "epoch": 1.3, + "learning_rate": 2.833867658373942e-05, + "loss": 0.5103, + "step": 33833 + }, + { + "epoch": 1.3, + "learning_rate": 2.8338035393690693e-05, + "loss": 0.4419, + "step": 33834 + }, + { + "epoch": 1.3, + "learning_rate": 2.8337394203641958e-05, + "loss": 0.3324, + "step": 33835 + }, + { + "epoch": 1.3, + "learning_rate": 2.833675301359323e-05, + "loss": 0.3419, + "step": 33836 + }, + { + "epoch": 1.3, + "learning_rate": 2.8336111823544498e-05, + "loss": 0.4281, + "step": 33837 + }, + { + "epoch": 1.3, + "learning_rate": 2.833547063349577e-05, + "loss": 0.3951, + "step": 33838 + }, + { + "epoch": 1.3, + "learning_rate": 2.833482944344704e-05, + "loss": 0.5136, + "step": 33839 + }, + { + "epoch": 1.3, + "learning_rate": 2.833418825339831e-05, + "loss": 0.3105, + "step": 33840 + }, + { + "epoch": 1.3, + "learning_rate": 2.833354706334958e-05, + "loss": 0.4683, + "step": 33841 + }, + { + "epoch": 1.3, + "learning_rate": 2.8332905873300846e-05, + "loss": 0.3795, + "step": 33842 + }, + { + "epoch": 1.3, + "learning_rate": 2.8332264683252118e-05, + "loss": 0.3772, + "step": 33843 + }, + { + "epoch": 1.3, + "learning_rate": 2.8331623493203387e-05, + "loss": 0.3956, + "step": 33844 + }, + { + "epoch": 1.3, + "learning_rate": 2.8330982303154658e-05, + "loss": 0.3039, + "step": 33845 + }, + { + "epoch": 1.3, + "learning_rate": 2.8330341113105923e-05, + "loss": 0.497, + "step": 33846 + }, + { + "epoch": 1.3, + "learning_rate": 2.8329699923057195e-05, + "loss": 0.4005, + "step": 33847 + }, + { + "epoch": 1.3, + "learning_rate": 2.8329058733008467e-05, + "loss": 0.3145, + "step": 33848 + }, + { + "epoch": 1.3, + "learning_rate": 2.8328417542959735e-05, + "loss": 0.3682, + "step": 33849 + }, + { + "epoch": 1.3, + "learning_rate": 2.8327776352911007e-05, + "loss": 0.3366, + "step": 33850 + }, + { + "epoch": 1.3, + "learning_rate": 2.832713516286227e-05, + "loss": 0.4263, + "step": 33851 + }, + { + "epoch": 1.3, + "learning_rate": 2.8326493972813543e-05, + "loss": 0.3099, + "step": 33852 + }, + { + "epoch": 1.3, + "learning_rate": 2.832585278276481e-05, + "loss": 0.3539, + "step": 33853 + }, + { + "epoch": 1.3, + "learning_rate": 2.8325211592716083e-05, + "loss": 0.2824, + "step": 33854 + }, + { + "epoch": 1.3, + "learning_rate": 2.8324570402667348e-05, + "loss": 0.3305, + "step": 33855 + }, + { + "epoch": 1.3, + "learning_rate": 2.8323929212618623e-05, + "loss": 0.278, + "step": 33856 + }, + { + "epoch": 1.3, + "learning_rate": 2.8323288022569895e-05, + "loss": 0.3716, + "step": 33857 + }, + { + "epoch": 1.3, + "learning_rate": 2.832264683252116e-05, + "loss": 0.4946, + "step": 33858 + }, + { + "epoch": 1.3, + "learning_rate": 2.8322005642472432e-05, + "loss": 0.3827, + "step": 33859 + }, + { + "epoch": 1.3, + "learning_rate": 2.83213644524237e-05, + "loss": 0.3383, + "step": 33860 + }, + { + "epoch": 1.3, + "learning_rate": 2.8320723262374972e-05, + "loss": 0.3973, + "step": 33861 + }, + { + "epoch": 1.3, + "learning_rate": 2.8320082072326237e-05, + "loss": 0.3448, + "step": 33862 + }, + { + "epoch": 1.3, + "learning_rate": 2.831944088227751e-05, + "loss": 0.398, + "step": 33863 + }, + { + "epoch": 1.3, + "learning_rate": 2.8318799692228777e-05, + "loss": 0.499, + "step": 33864 + }, + { + "epoch": 1.3, + "learning_rate": 2.831815850218005e-05, + "loss": 0.2643, + "step": 33865 + }, + { + "epoch": 1.3, + "learning_rate": 2.831751731213132e-05, + "loss": 0.408, + "step": 33866 + }, + { + "epoch": 1.3, + "learning_rate": 2.8316876122082585e-05, + "loss": 0.4732, + "step": 33867 + }, + { + "epoch": 1.3, + "learning_rate": 2.8316234932033857e-05, + "loss": 0.4683, + "step": 33868 + }, + { + "epoch": 1.3, + "learning_rate": 2.8315593741985125e-05, + "loss": 0.3524, + "step": 33869 + }, + { + "epoch": 1.3, + "learning_rate": 2.8314952551936397e-05, + "loss": 0.3784, + "step": 33870 + }, + { + "epoch": 1.3, + "learning_rate": 2.8314311361887662e-05, + "loss": 0.3161, + "step": 33871 + }, + { + "epoch": 1.3, + "learning_rate": 2.8313670171838934e-05, + "loss": 0.3591, + "step": 33872 + }, + { + "epoch": 1.3, + "learning_rate": 2.8313028981790202e-05, + "loss": 0.3544, + "step": 33873 + }, + { + "epoch": 1.3, + "learning_rate": 2.8312387791741474e-05, + "loss": 0.4046, + "step": 33874 + }, + { + "epoch": 1.3, + "learning_rate": 2.8311746601692745e-05, + "loss": 0.3517, + "step": 33875 + }, + { + "epoch": 1.3, + "learning_rate": 2.8311105411644014e-05, + "loss": 0.5347, + "step": 33876 + }, + { + "epoch": 1.3, + "learning_rate": 2.8310464221595285e-05, + "loss": 0.4151, + "step": 33877 + }, + { + "epoch": 1.3, + "learning_rate": 2.830982303154655e-05, + "loss": 0.3754, + "step": 33878 + }, + { + "epoch": 1.3, + "learning_rate": 2.8309181841497822e-05, + "loss": 0.4129, + "step": 33879 + }, + { + "epoch": 1.3, + "learning_rate": 2.830854065144909e-05, + "loss": 0.5088, + "step": 33880 + }, + { + "epoch": 1.3, + "learning_rate": 2.8307899461400362e-05, + "loss": 0.5235, + "step": 33881 + }, + { + "epoch": 1.3, + "learning_rate": 2.8307258271351627e-05, + "loss": 0.3815, + "step": 33882 + }, + { + "epoch": 1.3, + "learning_rate": 2.83066170813029e-05, + "loss": 0.3792, + "step": 33883 + }, + { + "epoch": 1.3, + "learning_rate": 2.830597589125417e-05, + "loss": 0.3657, + "step": 33884 + }, + { + "epoch": 1.3, + "learning_rate": 2.830533470120544e-05, + "loss": 0.3923, + "step": 33885 + }, + { + "epoch": 1.3, + "learning_rate": 2.830469351115671e-05, + "loss": 0.3799, + "step": 33886 + }, + { + "epoch": 1.3, + "learning_rate": 2.8304052321107975e-05, + "loss": 0.3626, + "step": 33887 + }, + { + "epoch": 1.3, + "learning_rate": 2.8303411131059247e-05, + "loss": 0.518, + "step": 33888 + }, + { + "epoch": 1.3, + "learning_rate": 2.8302769941010515e-05, + "loss": 0.3004, + "step": 33889 + }, + { + "epoch": 1.3, + "learning_rate": 2.8302128750961787e-05, + "loss": 0.3489, + "step": 33890 + }, + { + "epoch": 1.3, + "learning_rate": 2.8301487560913052e-05, + "loss": 0.2847, + "step": 33891 + }, + { + "epoch": 1.3, + "learning_rate": 2.8300846370864327e-05, + "loss": 0.3241, + "step": 33892 + }, + { + "epoch": 1.3, + "learning_rate": 2.83002051808156e-05, + "loss": 0.3761, + "step": 33893 + }, + { + "epoch": 1.3, + "learning_rate": 2.8299563990766864e-05, + "loss": 0.2841, + "step": 33894 + }, + { + "epoch": 1.3, + "learning_rate": 2.8298922800718136e-05, + "loss": 0.2827, + "step": 33895 + }, + { + "epoch": 1.3, + "learning_rate": 2.8298281610669404e-05, + "loss": 0.3075, + "step": 33896 + }, + { + "epoch": 1.3, + "learning_rate": 2.8297640420620676e-05, + "loss": 0.3257, + "step": 33897 + }, + { + "epoch": 1.3, + "learning_rate": 2.829699923057194e-05, + "loss": 0.4369, + "step": 33898 + }, + { + "epoch": 1.3, + "learning_rate": 2.8296358040523212e-05, + "loss": 0.3223, + "step": 33899 + }, + { + "epoch": 1.3, + "learning_rate": 2.829571685047448e-05, + "loss": 0.3927, + "step": 33900 + }, + { + "epoch": 1.3, + "learning_rate": 2.8295075660425752e-05, + "loss": 0.2796, + "step": 33901 + }, + { + "epoch": 1.3, + "learning_rate": 2.8294434470377024e-05, + "loss": 0.4011, + "step": 33902 + }, + { + "epoch": 1.3, + "learning_rate": 2.829379328032829e-05, + "loss": 0.2795, + "step": 33903 + }, + { + "epoch": 1.3, + "learning_rate": 2.829315209027956e-05, + "loss": 0.4585, + "step": 33904 + }, + { + "epoch": 1.3, + "learning_rate": 2.829251090023083e-05, + "loss": 0.2485, + "step": 33905 + }, + { + "epoch": 1.3, + "learning_rate": 2.82918697101821e-05, + "loss": 0.3968, + "step": 33906 + }, + { + "epoch": 1.3, + "learning_rate": 2.8291228520133366e-05, + "loss": 0.2829, + "step": 33907 + }, + { + "epoch": 1.3, + "learning_rate": 2.8290587330084637e-05, + "loss": 0.5356, + "step": 33908 + }, + { + "epoch": 1.3, + "learning_rate": 2.8289946140035906e-05, + "loss": 0.2559, + "step": 33909 + }, + { + "epoch": 1.3, + "learning_rate": 2.8289304949987177e-05, + "loss": 0.4625, + "step": 33910 + }, + { + "epoch": 1.3, + "learning_rate": 2.828866375993845e-05, + "loss": 0.4137, + "step": 33911 + }, + { + "epoch": 1.3, + "learning_rate": 2.8288022569889718e-05, + "loss": 0.4475, + "step": 33912 + }, + { + "epoch": 1.3, + "learning_rate": 2.828738137984099e-05, + "loss": 0.4078, + "step": 33913 + }, + { + "epoch": 1.3, + "learning_rate": 2.8286740189792254e-05, + "loss": 0.3783, + "step": 33914 + }, + { + "epoch": 1.3, + "learning_rate": 2.8286098999743526e-05, + "loss": 0.3041, + "step": 33915 + }, + { + "epoch": 1.3, + "learning_rate": 2.8285457809694794e-05, + "loss": 0.4101, + "step": 33916 + }, + { + "epoch": 1.3, + "learning_rate": 2.8284816619646066e-05, + "loss": 0.3647, + "step": 33917 + }, + { + "epoch": 1.3, + "learning_rate": 2.828417542959733e-05, + "loss": 0.3092, + "step": 33918 + }, + { + "epoch": 1.3, + "learning_rate": 2.8283534239548603e-05, + "loss": 0.466, + "step": 33919 + }, + { + "epoch": 1.3, + "learning_rate": 2.8282893049499874e-05, + "loss": 0.43, + "step": 33920 + }, + { + "epoch": 1.3, + "learning_rate": 2.8282251859451143e-05, + "loss": 0.5546, + "step": 33921 + }, + { + "epoch": 1.3, + "learning_rate": 2.8281610669402414e-05, + "loss": 0.3004, + "step": 33922 + }, + { + "epoch": 1.3, + "learning_rate": 2.828096947935368e-05, + "loss": 0.4052, + "step": 33923 + }, + { + "epoch": 1.3, + "learning_rate": 2.828032828930495e-05, + "loss": 0.3553, + "step": 33924 + }, + { + "epoch": 1.3, + "learning_rate": 2.827968709925622e-05, + "loss": 0.2614, + "step": 33925 + }, + { + "epoch": 1.3, + "learning_rate": 2.827904590920749e-05, + "loss": 0.4552, + "step": 33926 + }, + { + "epoch": 1.3, + "learning_rate": 2.8278404719158756e-05, + "loss": 0.3546, + "step": 33927 + }, + { + "epoch": 1.3, + "learning_rate": 2.827776352911003e-05, + "loss": 0.3567, + "step": 33928 + }, + { + "epoch": 1.3, + "learning_rate": 2.8277122339061303e-05, + "loss": 0.4105, + "step": 33929 + }, + { + "epoch": 1.3, + "learning_rate": 2.8276481149012568e-05, + "loss": 0.3224, + "step": 33930 + }, + { + "epoch": 1.3, + "learning_rate": 2.827583995896384e-05, + "loss": 0.3775, + "step": 33931 + }, + { + "epoch": 1.3, + "learning_rate": 2.8275198768915108e-05, + "loss": 0.3459, + "step": 33932 + }, + { + "epoch": 1.3, + "learning_rate": 2.827455757886638e-05, + "loss": 0.4282, + "step": 33933 + }, + { + "epoch": 1.3, + "learning_rate": 2.8273916388817644e-05, + "loss": 0.3809, + "step": 33934 + }, + { + "epoch": 1.3, + "learning_rate": 2.8273275198768916e-05, + "loss": 0.2548, + "step": 33935 + }, + { + "epoch": 1.3, + "learning_rate": 2.8272634008720185e-05, + "loss": 0.5295, + "step": 33936 + }, + { + "epoch": 1.3, + "learning_rate": 2.8271992818671456e-05, + "loss": 0.427, + "step": 33937 + }, + { + "epoch": 1.3, + "learning_rate": 2.8271351628622728e-05, + "loss": 0.3552, + "step": 33938 + }, + { + "epoch": 1.3, + "learning_rate": 2.8270710438573993e-05, + "loss": 0.4029, + "step": 33939 + }, + { + "epoch": 1.3, + "learning_rate": 2.8270069248525265e-05, + "loss": 0.4485, + "step": 33940 + }, + { + "epoch": 1.3, + "learning_rate": 2.8269428058476533e-05, + "loss": 0.4262, + "step": 33941 + }, + { + "epoch": 1.3, + "learning_rate": 2.8268786868427805e-05, + "loss": 0.4179, + "step": 33942 + }, + { + "epoch": 1.3, + "learning_rate": 2.826814567837907e-05, + "loss": 0.4281, + "step": 33943 + }, + { + "epoch": 1.31, + "learning_rate": 2.8267504488330345e-05, + "loss": 0.4095, + "step": 33944 + }, + { + "epoch": 1.31, + "learning_rate": 2.826686329828161e-05, + "loss": 0.4785, + "step": 33945 + }, + { + "epoch": 1.31, + "learning_rate": 2.826622210823288e-05, + "loss": 0.4266, + "step": 33946 + }, + { + "epoch": 1.31, + "learning_rate": 2.8265580918184153e-05, + "loss": 0.2918, + "step": 33947 + }, + { + "epoch": 1.31, + "learning_rate": 2.826493972813542e-05, + "loss": 0.3204, + "step": 33948 + }, + { + "epoch": 1.31, + "learning_rate": 2.8264298538086693e-05, + "loss": 0.3507, + "step": 33949 + }, + { + "epoch": 1.31, + "learning_rate": 2.8263657348037958e-05, + "loss": 0.3591, + "step": 33950 + }, + { + "epoch": 1.31, + "learning_rate": 2.826301615798923e-05, + "loss": 0.36, + "step": 33951 + }, + { + "epoch": 1.31, + "learning_rate": 2.8262374967940498e-05, + "loss": 0.2838, + "step": 33952 + }, + { + "epoch": 1.31, + "learning_rate": 2.826173377789177e-05, + "loss": 0.3673, + "step": 33953 + }, + { + "epoch": 1.31, + "learning_rate": 2.8261092587843035e-05, + "loss": 0.3172, + "step": 33954 + }, + { + "epoch": 1.31, + "learning_rate": 2.8260451397794306e-05, + "loss": 0.5599, + "step": 33955 + }, + { + "epoch": 1.31, + "learning_rate": 2.8259810207745578e-05, + "loss": 0.3484, + "step": 33956 + }, + { + "epoch": 1.31, + "learning_rate": 2.8259169017696847e-05, + "loss": 0.2909, + "step": 33957 + }, + { + "epoch": 1.31, + "learning_rate": 2.8258527827648118e-05, + "loss": 0.3863, + "step": 33958 + }, + { + "epoch": 1.31, + "learning_rate": 2.8257886637599383e-05, + "loss": 0.3679, + "step": 33959 + }, + { + "epoch": 1.31, + "learning_rate": 2.8257245447550655e-05, + "loss": 0.3, + "step": 33960 + }, + { + "epoch": 1.31, + "learning_rate": 2.8256604257501923e-05, + "loss": 0.39, + "step": 33961 + }, + { + "epoch": 1.31, + "learning_rate": 2.8255963067453195e-05, + "loss": 0.2279, + "step": 33962 + }, + { + "epoch": 1.31, + "learning_rate": 2.825532187740446e-05, + "loss": 0.3225, + "step": 33963 + }, + { + "epoch": 1.31, + "learning_rate": 2.8254680687355735e-05, + "loss": 0.412, + "step": 33964 + }, + { + "epoch": 1.31, + "learning_rate": 2.8254039497307007e-05, + "loss": 0.3292, + "step": 33965 + }, + { + "epoch": 1.31, + "learning_rate": 2.825339830725827e-05, + "loss": 0.3714, + "step": 33966 + }, + { + "epoch": 1.31, + "learning_rate": 2.8252757117209543e-05, + "loss": 0.3713, + "step": 33967 + }, + { + "epoch": 1.31, + "learning_rate": 2.8252115927160812e-05, + "loss": 0.4051, + "step": 33968 + }, + { + "epoch": 1.31, + "learning_rate": 2.8251474737112083e-05, + "loss": 0.4018, + "step": 33969 + }, + { + "epoch": 1.31, + "learning_rate": 2.825083354706335e-05, + "loss": 0.3386, + "step": 33970 + }, + { + "epoch": 1.31, + "learning_rate": 2.825019235701462e-05, + "loss": 0.3306, + "step": 33971 + }, + { + "epoch": 1.31, + "learning_rate": 2.824955116696589e-05, + "loss": 0.4193, + "step": 33972 + }, + { + "epoch": 1.31, + "learning_rate": 2.824890997691716e-05, + "loss": 0.2474, + "step": 33973 + }, + { + "epoch": 1.31, + "learning_rate": 2.8248268786868432e-05, + "loss": 0.4026, + "step": 33974 + }, + { + "epoch": 1.31, + "learning_rate": 2.8247627596819697e-05, + "loss": 0.4603, + "step": 33975 + }, + { + "epoch": 1.31, + "learning_rate": 2.824698640677097e-05, + "loss": 0.4286, + "step": 33976 + }, + { + "epoch": 1.31, + "learning_rate": 2.8246345216722237e-05, + "loss": 0.3454, + "step": 33977 + }, + { + "epoch": 1.31, + "learning_rate": 2.824570402667351e-05, + "loss": 0.4433, + "step": 33978 + }, + { + "epoch": 1.31, + "learning_rate": 2.8245062836624773e-05, + "loss": 0.4656, + "step": 33979 + }, + { + "epoch": 1.31, + "learning_rate": 2.824442164657605e-05, + "loss": 0.4499, + "step": 33980 + }, + { + "epoch": 1.31, + "learning_rate": 2.8243780456527314e-05, + "loss": 0.3917, + "step": 33981 + }, + { + "epoch": 1.31, + "learning_rate": 2.8243139266478585e-05, + "loss": 0.4125, + "step": 33982 + }, + { + "epoch": 1.31, + "learning_rate": 2.8242498076429857e-05, + "loss": 0.4131, + "step": 33983 + }, + { + "epoch": 1.31, + "learning_rate": 2.8241856886381125e-05, + "loss": 0.4045, + "step": 33984 + }, + { + "epoch": 1.31, + "learning_rate": 2.8241215696332397e-05, + "loss": 0.3092, + "step": 33985 + }, + { + "epoch": 1.31, + "learning_rate": 2.8240574506283662e-05, + "loss": 0.389, + "step": 33986 + }, + { + "epoch": 1.31, + "learning_rate": 2.8239933316234934e-05, + "loss": 0.4313, + "step": 33987 + }, + { + "epoch": 1.31, + "learning_rate": 2.8239292126186202e-05, + "loss": 0.3969, + "step": 33988 + }, + { + "epoch": 1.31, + "learning_rate": 2.8238650936137474e-05, + "loss": 0.4489, + "step": 33989 + }, + { + "epoch": 1.31, + "learning_rate": 2.823800974608874e-05, + "loss": 0.4431, + "step": 33990 + }, + { + "epoch": 1.31, + "learning_rate": 2.823736855604001e-05, + "loss": 0.4747, + "step": 33991 + }, + { + "epoch": 1.31, + "learning_rate": 2.8236727365991282e-05, + "loss": 0.4136, + "step": 33992 + }, + { + "epoch": 1.31, + "learning_rate": 2.823608617594255e-05, + "loss": 0.3583, + "step": 33993 + }, + { + "epoch": 1.31, + "learning_rate": 2.8235444985893822e-05, + "loss": 0.3652, + "step": 33994 + }, + { + "epoch": 1.31, + "learning_rate": 2.8234803795845087e-05, + "loss": 0.3539, + "step": 33995 + }, + { + "epoch": 1.31, + "learning_rate": 2.823416260579636e-05, + "loss": 0.5127, + "step": 33996 + }, + { + "epoch": 1.31, + "learning_rate": 2.8233521415747627e-05, + "loss": 0.4857, + "step": 33997 + }, + { + "epoch": 1.31, + "learning_rate": 2.82328802256989e-05, + "loss": 0.5199, + "step": 33998 + }, + { + "epoch": 1.31, + "learning_rate": 2.8232239035650164e-05, + "loss": 0.4232, + "step": 33999 + }, + { + "epoch": 1.31, + "learning_rate": 2.823159784560144e-05, + "loss": 0.3547, + "step": 34000 + }, + { + "epoch": 1.31, + "learning_rate": 2.823095665555271e-05, + "loss": 0.3912, + "step": 34001 + }, + { + "epoch": 1.31, + "learning_rate": 2.8230315465503976e-05, + "loss": 0.4345, + "step": 34002 + }, + { + "epoch": 1.31, + "learning_rate": 2.8229674275455247e-05, + "loss": 0.3136, + "step": 34003 + }, + { + "epoch": 1.31, + "learning_rate": 2.8229033085406516e-05, + "loss": 0.4196, + "step": 34004 + }, + { + "epoch": 1.31, + "learning_rate": 2.8228391895357787e-05, + "loss": 0.4458, + "step": 34005 + }, + { + "epoch": 1.31, + "learning_rate": 2.8227750705309052e-05, + "loss": 0.4015, + "step": 34006 + }, + { + "epoch": 1.31, + "learning_rate": 2.8227109515260324e-05, + "loss": 0.3814, + "step": 34007 + }, + { + "epoch": 1.31, + "learning_rate": 2.8226468325211592e-05, + "loss": 0.3453, + "step": 34008 + }, + { + "epoch": 1.31, + "learning_rate": 2.8225827135162864e-05, + "loss": 0.3475, + "step": 34009 + }, + { + "epoch": 1.31, + "learning_rate": 2.8225185945114136e-05, + "loss": 0.3587, + "step": 34010 + }, + { + "epoch": 1.31, + "learning_rate": 2.82245447550654e-05, + "loss": 0.4544, + "step": 34011 + }, + { + "epoch": 1.31, + "learning_rate": 2.8223903565016672e-05, + "loss": 0.4184, + "step": 34012 + }, + { + "epoch": 1.31, + "learning_rate": 2.822326237496794e-05, + "loss": 0.3395, + "step": 34013 + }, + { + "epoch": 1.31, + "learning_rate": 2.8222621184919212e-05, + "loss": 0.4013, + "step": 34014 + }, + { + "epoch": 1.31, + "learning_rate": 2.8221979994870477e-05, + "loss": 0.4262, + "step": 34015 + }, + { + "epoch": 1.31, + "learning_rate": 2.8221338804821752e-05, + "loss": 0.2813, + "step": 34016 + }, + { + "epoch": 1.31, + "learning_rate": 2.8220697614773017e-05, + "loss": 0.3466, + "step": 34017 + }, + { + "epoch": 1.31, + "learning_rate": 2.822005642472429e-05, + "loss": 0.4358, + "step": 34018 + }, + { + "epoch": 1.31, + "learning_rate": 2.821941523467556e-05, + "loss": 0.3624, + "step": 34019 + }, + { + "epoch": 1.31, + "learning_rate": 2.821877404462683e-05, + "loss": 0.4114, + "step": 34020 + }, + { + "epoch": 1.31, + "learning_rate": 2.82181328545781e-05, + "loss": 0.5538, + "step": 34021 + }, + { + "epoch": 1.31, + "learning_rate": 2.8217491664529366e-05, + "loss": 0.4843, + "step": 34022 + }, + { + "epoch": 1.31, + "learning_rate": 2.8216850474480638e-05, + "loss": 0.4075, + "step": 34023 + }, + { + "epoch": 1.31, + "learning_rate": 2.8216209284431906e-05, + "loss": 0.5178, + "step": 34024 + }, + { + "epoch": 1.31, + "learning_rate": 2.8215568094383178e-05, + "loss": 0.4459, + "step": 34025 + }, + { + "epoch": 1.31, + "learning_rate": 2.821492690433445e-05, + "loss": 0.454, + "step": 34026 + }, + { + "epoch": 1.31, + "learning_rate": 2.8214285714285714e-05, + "loss": 0.3401, + "step": 34027 + }, + { + "epoch": 1.31, + "learning_rate": 2.8213644524236986e-05, + "loss": 0.3342, + "step": 34028 + }, + { + "epoch": 1.31, + "learning_rate": 2.8213003334188254e-05, + "loss": 0.3229, + "step": 34029 + }, + { + "epoch": 1.31, + "learning_rate": 2.8212362144139526e-05, + "loss": 0.3726, + "step": 34030 + }, + { + "epoch": 1.31, + "learning_rate": 2.821172095409079e-05, + "loss": 0.2545, + "step": 34031 + }, + { + "epoch": 1.31, + "learning_rate": 2.8211079764042063e-05, + "loss": 0.3735, + "step": 34032 + }, + { + "epoch": 1.31, + "learning_rate": 2.821043857399333e-05, + "loss": 0.4049, + "step": 34033 + }, + { + "epoch": 1.31, + "learning_rate": 2.8209797383944603e-05, + "loss": 0.3333, + "step": 34034 + }, + { + "epoch": 1.31, + "learning_rate": 2.8209156193895874e-05, + "loss": 0.2921, + "step": 34035 + }, + { + "epoch": 1.31, + "learning_rate": 2.8208515003847143e-05, + "loss": 0.3785, + "step": 34036 + }, + { + "epoch": 1.31, + "learning_rate": 2.8207873813798414e-05, + "loss": 0.4918, + "step": 34037 + }, + { + "epoch": 1.31, + "learning_rate": 2.820723262374968e-05, + "loss": 0.3949, + "step": 34038 + }, + { + "epoch": 1.31, + "learning_rate": 2.820659143370095e-05, + "loss": 0.3188, + "step": 34039 + }, + { + "epoch": 1.31, + "learning_rate": 2.820595024365222e-05, + "loss": 0.2787, + "step": 34040 + }, + { + "epoch": 1.31, + "learning_rate": 2.820530905360349e-05, + "loss": 0.4662, + "step": 34041 + }, + { + "epoch": 1.31, + "learning_rate": 2.8204667863554756e-05, + "loss": 0.394, + "step": 34042 + }, + { + "epoch": 1.31, + "learning_rate": 2.8204026673506028e-05, + "loss": 0.3683, + "step": 34043 + }, + { + "epoch": 1.31, + "learning_rate": 2.82033854834573e-05, + "loss": 0.3558, + "step": 34044 + }, + { + "epoch": 1.31, + "learning_rate": 2.8202744293408568e-05, + "loss": 0.322, + "step": 34045 + }, + { + "epoch": 1.31, + "learning_rate": 2.820210310335984e-05, + "loss": 0.5107, + "step": 34046 + }, + { + "epoch": 1.31, + "learning_rate": 2.8201461913311105e-05, + "loss": 0.2662, + "step": 34047 + }, + { + "epoch": 1.31, + "learning_rate": 2.8200820723262376e-05, + "loss": 0.4114, + "step": 34048 + }, + { + "epoch": 1.31, + "learning_rate": 2.8200179533213645e-05, + "loss": 0.2745, + "step": 34049 + }, + { + "epoch": 1.31, + "learning_rate": 2.8199538343164916e-05, + "loss": 0.5131, + "step": 34050 + }, + { + "epoch": 1.31, + "learning_rate": 2.819889715311618e-05, + "loss": 0.3289, + "step": 34051 + }, + { + "epoch": 1.31, + "learning_rate": 2.8198255963067456e-05, + "loss": 0.4343, + "step": 34052 + }, + { + "epoch": 1.31, + "learning_rate": 2.8197614773018728e-05, + "loss": 0.3146, + "step": 34053 + }, + { + "epoch": 1.31, + "learning_rate": 2.8196973582969993e-05, + "loss": 0.3152, + "step": 34054 + }, + { + "epoch": 1.31, + "learning_rate": 2.8196332392921265e-05, + "loss": 0.3634, + "step": 34055 + }, + { + "epoch": 1.31, + "learning_rate": 2.8195691202872533e-05, + "loss": 0.3894, + "step": 34056 + }, + { + "epoch": 1.31, + "learning_rate": 2.8195050012823805e-05, + "loss": 0.2309, + "step": 34057 + }, + { + "epoch": 1.31, + "learning_rate": 2.819440882277507e-05, + "loss": 0.3988, + "step": 34058 + }, + { + "epoch": 1.31, + "learning_rate": 2.819376763272634e-05, + "loss": 0.3173, + "step": 34059 + }, + { + "epoch": 1.31, + "learning_rate": 2.819312644267761e-05, + "loss": 0.4856, + "step": 34060 + }, + { + "epoch": 1.31, + "learning_rate": 2.819248525262888e-05, + "loss": 0.3702, + "step": 34061 + }, + { + "epoch": 1.31, + "learning_rate": 2.8191844062580153e-05, + "loss": 0.2931, + "step": 34062 + }, + { + "epoch": 1.31, + "learning_rate": 2.8191202872531418e-05, + "loss": 0.3164, + "step": 34063 + }, + { + "epoch": 1.31, + "learning_rate": 2.819056168248269e-05, + "loss": 0.4002, + "step": 34064 + }, + { + "epoch": 1.31, + "learning_rate": 2.8189920492433958e-05, + "loss": 0.415, + "step": 34065 + }, + { + "epoch": 1.31, + "learning_rate": 2.818927930238523e-05, + "loss": 0.3684, + "step": 34066 + }, + { + "epoch": 1.31, + "learning_rate": 2.8188638112336495e-05, + "loss": 0.4207, + "step": 34067 + }, + { + "epoch": 1.31, + "learning_rate": 2.818799692228777e-05, + "loss": 0.4294, + "step": 34068 + }, + { + "epoch": 1.31, + "learning_rate": 2.8187355732239035e-05, + "loss": 0.4979, + "step": 34069 + }, + { + "epoch": 1.31, + "learning_rate": 2.8186714542190307e-05, + "loss": 0.2786, + "step": 34070 + }, + { + "epoch": 1.31, + "learning_rate": 2.8186073352141578e-05, + "loss": 0.3961, + "step": 34071 + }, + { + "epoch": 1.31, + "learning_rate": 2.8185432162092847e-05, + "loss": 0.3485, + "step": 34072 + }, + { + "epoch": 1.31, + "learning_rate": 2.818479097204412e-05, + "loss": 0.474, + "step": 34073 + }, + { + "epoch": 1.31, + "learning_rate": 2.8184149781995383e-05, + "loss": 0.3249, + "step": 34074 + }, + { + "epoch": 1.31, + "learning_rate": 2.8183508591946655e-05, + "loss": 0.3816, + "step": 34075 + }, + { + "epoch": 1.31, + "learning_rate": 2.8182867401897923e-05, + "loss": 0.2866, + "step": 34076 + }, + { + "epoch": 1.31, + "learning_rate": 2.8182226211849195e-05, + "loss": 0.5439, + "step": 34077 + }, + { + "epoch": 1.31, + "learning_rate": 2.818158502180046e-05, + "loss": 0.2915, + "step": 34078 + }, + { + "epoch": 1.31, + "learning_rate": 2.818094383175173e-05, + "loss": 0.3636, + "step": 34079 + }, + { + "epoch": 1.31, + "learning_rate": 2.8180302641703003e-05, + "loss": 0.4268, + "step": 34080 + }, + { + "epoch": 1.31, + "learning_rate": 2.8179661451654272e-05, + "loss": 0.412, + "step": 34081 + }, + { + "epoch": 1.31, + "learning_rate": 2.8179020261605543e-05, + "loss": 0.3837, + "step": 34082 + }, + { + "epoch": 1.31, + "learning_rate": 2.817837907155681e-05, + "loss": 0.5859, + "step": 34083 + }, + { + "epoch": 1.31, + "learning_rate": 2.817773788150808e-05, + "loss": 0.2757, + "step": 34084 + }, + { + "epoch": 1.31, + "learning_rate": 2.817709669145935e-05, + "loss": 0.2963, + "step": 34085 + }, + { + "epoch": 1.31, + "learning_rate": 2.817645550141062e-05, + "loss": 0.2886, + "step": 34086 + }, + { + "epoch": 1.31, + "learning_rate": 2.8175814311361885e-05, + "loss": 0.3868, + "step": 34087 + }, + { + "epoch": 1.31, + "learning_rate": 2.817517312131316e-05, + "loss": 0.37, + "step": 34088 + }, + { + "epoch": 1.31, + "learning_rate": 2.8174531931264432e-05, + "loss": 0.3378, + "step": 34089 + }, + { + "epoch": 1.31, + "learning_rate": 2.8173890741215697e-05, + "loss": 0.4518, + "step": 34090 + }, + { + "epoch": 1.31, + "learning_rate": 2.817324955116697e-05, + "loss": 0.3073, + "step": 34091 + }, + { + "epoch": 1.31, + "learning_rate": 2.8172608361118237e-05, + "loss": 0.3192, + "step": 34092 + }, + { + "epoch": 1.31, + "learning_rate": 2.817196717106951e-05, + "loss": 0.3991, + "step": 34093 + }, + { + "epoch": 1.31, + "learning_rate": 2.8171325981020774e-05, + "loss": 0.4275, + "step": 34094 + }, + { + "epoch": 1.31, + "learning_rate": 2.8170684790972045e-05, + "loss": 0.4313, + "step": 34095 + }, + { + "epoch": 1.31, + "learning_rate": 2.8170043600923314e-05, + "loss": 0.4096, + "step": 34096 + }, + { + "epoch": 1.31, + "learning_rate": 2.8169402410874585e-05, + "loss": 0.5185, + "step": 34097 + }, + { + "epoch": 1.31, + "learning_rate": 2.8168761220825857e-05, + "loss": 0.5329, + "step": 34098 + }, + { + "epoch": 1.31, + "learning_rate": 2.8168120030777122e-05, + "loss": 0.4733, + "step": 34099 + }, + { + "epoch": 1.31, + "learning_rate": 2.8167478840728394e-05, + "loss": 0.239, + "step": 34100 + }, + { + "epoch": 1.31, + "learning_rate": 2.8166837650679662e-05, + "loss": 0.3988, + "step": 34101 + }, + { + "epoch": 1.31, + "learning_rate": 2.8166196460630934e-05, + "loss": 0.3144, + "step": 34102 + }, + { + "epoch": 1.31, + "learning_rate": 2.81655552705822e-05, + "loss": 0.3043, + "step": 34103 + }, + { + "epoch": 1.31, + "learning_rate": 2.8164914080533474e-05, + "loss": 0.3628, + "step": 34104 + }, + { + "epoch": 1.31, + "learning_rate": 2.816427289048474e-05, + "loss": 0.294, + "step": 34105 + }, + { + "epoch": 1.31, + "learning_rate": 2.816363170043601e-05, + "loss": 0.3315, + "step": 34106 + }, + { + "epoch": 1.31, + "learning_rate": 2.8162990510387282e-05, + "loss": 0.3026, + "step": 34107 + }, + { + "epoch": 1.31, + "learning_rate": 2.816234932033855e-05, + "loss": 0.3754, + "step": 34108 + }, + { + "epoch": 1.31, + "learning_rate": 2.8161708130289822e-05, + "loss": 0.3785, + "step": 34109 + }, + { + "epoch": 1.31, + "learning_rate": 2.8161066940241087e-05, + "loss": 0.3944, + "step": 34110 + }, + { + "epoch": 1.31, + "learning_rate": 2.816042575019236e-05, + "loss": 0.2988, + "step": 34111 + }, + { + "epoch": 1.31, + "learning_rate": 2.8159784560143627e-05, + "loss": 0.3238, + "step": 34112 + }, + { + "epoch": 1.31, + "learning_rate": 2.81591433700949e-05, + "loss": 0.3311, + "step": 34113 + }, + { + "epoch": 1.31, + "learning_rate": 2.8158502180046164e-05, + "loss": 0.3264, + "step": 34114 + }, + { + "epoch": 1.31, + "learning_rate": 2.8157860989997436e-05, + "loss": 0.3354, + "step": 34115 + }, + { + "epoch": 1.31, + "learning_rate": 2.8157219799948707e-05, + "loss": 0.446, + "step": 34116 + }, + { + "epoch": 1.31, + "learning_rate": 2.8156578609899976e-05, + "loss": 0.2867, + "step": 34117 + }, + { + "epoch": 1.31, + "learning_rate": 2.8155937419851247e-05, + "loss": 0.3244, + "step": 34118 + }, + { + "epoch": 1.31, + "learning_rate": 2.8155296229802512e-05, + "loss": 0.5596, + "step": 34119 + }, + { + "epoch": 1.31, + "learning_rate": 2.8154655039753784e-05, + "loss": 0.2846, + "step": 34120 + }, + { + "epoch": 1.31, + "learning_rate": 2.8154013849705052e-05, + "loss": 0.4924, + "step": 34121 + }, + { + "epoch": 1.31, + "learning_rate": 2.8153372659656324e-05, + "loss": 0.2972, + "step": 34122 + }, + { + "epoch": 1.31, + "learning_rate": 2.815273146960759e-05, + "loss": 0.3706, + "step": 34123 + }, + { + "epoch": 1.31, + "learning_rate": 2.8152090279558864e-05, + "loss": 0.3869, + "step": 34124 + }, + { + "epoch": 1.31, + "learning_rate": 2.8151449089510136e-05, + "loss": 0.4752, + "step": 34125 + }, + { + "epoch": 1.31, + "learning_rate": 2.81508078994614e-05, + "loss": 0.3679, + "step": 34126 + }, + { + "epoch": 1.31, + "learning_rate": 2.8150166709412672e-05, + "loss": 0.3693, + "step": 34127 + }, + { + "epoch": 1.31, + "learning_rate": 2.814952551936394e-05, + "loss": 0.275, + "step": 34128 + }, + { + "epoch": 1.31, + "learning_rate": 2.8148884329315212e-05, + "loss": 0.3461, + "step": 34129 + }, + { + "epoch": 1.31, + "learning_rate": 2.8148243139266477e-05, + "loss": 0.3195, + "step": 34130 + }, + { + "epoch": 1.31, + "learning_rate": 2.814760194921775e-05, + "loss": 0.4104, + "step": 34131 + }, + { + "epoch": 1.31, + "learning_rate": 2.8146960759169017e-05, + "loss": 0.4243, + "step": 34132 + }, + { + "epoch": 1.31, + "learning_rate": 2.814631956912029e-05, + "loss": 0.4096, + "step": 34133 + }, + { + "epoch": 1.31, + "learning_rate": 2.814567837907156e-05, + "loss": 0.4058, + "step": 34134 + }, + { + "epoch": 1.31, + "learning_rate": 2.8145037189022826e-05, + "loss": 0.4741, + "step": 34135 + }, + { + "epoch": 1.31, + "learning_rate": 2.8144395998974098e-05, + "loss": 0.3129, + "step": 34136 + }, + { + "epoch": 1.31, + "learning_rate": 2.8143754808925366e-05, + "loss": 0.3833, + "step": 34137 + }, + { + "epoch": 1.31, + "learning_rate": 2.8143113618876638e-05, + "loss": 0.3468, + "step": 34138 + }, + { + "epoch": 1.31, + "learning_rate": 2.8142472428827903e-05, + "loss": 0.3004, + "step": 34139 + }, + { + "epoch": 1.31, + "learning_rate": 2.8141831238779178e-05, + "loss": 0.5407, + "step": 34140 + }, + { + "epoch": 1.31, + "learning_rate": 2.8141190048730443e-05, + "loss": 0.3916, + "step": 34141 + }, + { + "epoch": 1.31, + "learning_rate": 2.8140548858681714e-05, + "loss": 0.4645, + "step": 34142 + }, + { + "epoch": 1.31, + "learning_rate": 2.8139907668632986e-05, + "loss": 0.4339, + "step": 34143 + }, + { + "epoch": 1.31, + "learning_rate": 2.8139266478584254e-05, + "loss": 0.2418, + "step": 34144 + }, + { + "epoch": 1.31, + "learning_rate": 2.8138625288535526e-05, + "loss": 0.3326, + "step": 34145 + }, + { + "epoch": 1.31, + "learning_rate": 2.813798409848679e-05, + "loss": 0.3549, + "step": 34146 + }, + { + "epoch": 1.31, + "learning_rate": 2.8137342908438063e-05, + "loss": 0.4551, + "step": 34147 + }, + { + "epoch": 1.31, + "learning_rate": 2.813670171838933e-05, + "loss": 0.3701, + "step": 34148 + }, + { + "epoch": 1.31, + "learning_rate": 2.8136060528340603e-05, + "loss": 0.4052, + "step": 34149 + }, + { + "epoch": 1.31, + "learning_rate": 2.8135419338291868e-05, + "loss": 0.3628, + "step": 34150 + }, + { + "epoch": 1.31, + "learning_rate": 2.813477814824314e-05, + "loss": 0.4089, + "step": 34151 + }, + { + "epoch": 1.31, + "learning_rate": 2.813413695819441e-05, + "loss": 0.3119, + "step": 34152 + }, + { + "epoch": 1.31, + "learning_rate": 2.813349576814568e-05, + "loss": 0.4256, + "step": 34153 + }, + { + "epoch": 1.31, + "learning_rate": 2.813285457809695e-05, + "loss": 0.4579, + "step": 34154 + }, + { + "epoch": 1.31, + "learning_rate": 2.8132213388048216e-05, + "loss": 0.3301, + "step": 34155 + }, + { + "epoch": 1.31, + "learning_rate": 2.8131572197999488e-05, + "loss": 0.3598, + "step": 34156 + }, + { + "epoch": 1.31, + "learning_rate": 2.8130931007950756e-05, + "loss": 0.3898, + "step": 34157 + }, + { + "epoch": 1.31, + "learning_rate": 2.8130289817902028e-05, + "loss": 0.3595, + "step": 34158 + }, + { + "epoch": 1.31, + "learning_rate": 2.8129648627853293e-05, + "loss": 0.3432, + "step": 34159 + }, + { + "epoch": 1.31, + "learning_rate": 2.8129007437804568e-05, + "loss": 0.4434, + "step": 34160 + }, + { + "epoch": 1.31, + "learning_rate": 2.812836624775584e-05, + "loss": 0.3109, + "step": 34161 + }, + { + "epoch": 1.31, + "learning_rate": 2.8127725057707105e-05, + "loss": 0.3138, + "step": 34162 + }, + { + "epoch": 1.31, + "learning_rate": 2.8127083867658376e-05, + "loss": 0.2491, + "step": 34163 + }, + { + "epoch": 1.31, + "learning_rate": 2.8126442677609645e-05, + "loss": 0.3464, + "step": 34164 + }, + { + "epoch": 1.31, + "learning_rate": 2.8125801487560916e-05, + "loss": 0.3259, + "step": 34165 + }, + { + "epoch": 1.31, + "learning_rate": 2.812516029751218e-05, + "loss": 0.3497, + "step": 34166 + }, + { + "epoch": 1.31, + "learning_rate": 2.8124519107463453e-05, + "loss": 0.4726, + "step": 34167 + }, + { + "epoch": 1.31, + "learning_rate": 2.812387791741472e-05, + "loss": 0.4734, + "step": 34168 + }, + { + "epoch": 1.31, + "learning_rate": 2.8123236727365993e-05, + "loss": 0.4562, + "step": 34169 + }, + { + "epoch": 1.31, + "learning_rate": 2.8122595537317265e-05, + "loss": 0.3713, + "step": 34170 + }, + { + "epoch": 1.31, + "learning_rate": 2.812195434726853e-05, + "loss": 0.4322, + "step": 34171 + }, + { + "epoch": 1.31, + "learning_rate": 2.81213131572198e-05, + "loss": 0.333, + "step": 34172 + }, + { + "epoch": 1.31, + "learning_rate": 2.812067196717107e-05, + "loss": 0.3787, + "step": 34173 + }, + { + "epoch": 1.31, + "learning_rate": 2.812003077712234e-05, + "loss": 0.4327, + "step": 34174 + }, + { + "epoch": 1.31, + "learning_rate": 2.8119389587073606e-05, + "loss": 0.4678, + "step": 34175 + }, + { + "epoch": 1.31, + "learning_rate": 2.811874839702488e-05, + "loss": 0.268, + "step": 34176 + }, + { + "epoch": 1.31, + "learning_rate": 2.8118107206976146e-05, + "loss": 0.2627, + "step": 34177 + }, + { + "epoch": 1.31, + "learning_rate": 2.8117466016927418e-05, + "loss": 0.4494, + "step": 34178 + }, + { + "epoch": 1.31, + "learning_rate": 2.811682482687869e-05, + "loss": 0.3699, + "step": 34179 + }, + { + "epoch": 1.31, + "learning_rate": 2.8116183636829958e-05, + "loss": 0.518, + "step": 34180 + }, + { + "epoch": 1.31, + "learning_rate": 2.811554244678123e-05, + "loss": 0.3464, + "step": 34181 + }, + { + "epoch": 1.31, + "learning_rate": 2.8114901256732495e-05, + "loss": 0.4551, + "step": 34182 + }, + { + "epoch": 1.31, + "learning_rate": 2.8114260066683767e-05, + "loss": 0.4432, + "step": 34183 + }, + { + "epoch": 1.31, + "learning_rate": 2.8113618876635035e-05, + "loss": 0.3863, + "step": 34184 + }, + { + "epoch": 1.31, + "learning_rate": 2.8112977686586307e-05, + "loss": 0.4186, + "step": 34185 + }, + { + "epoch": 1.31, + "learning_rate": 2.811233649653757e-05, + "loss": 0.2982, + "step": 34186 + }, + { + "epoch": 1.31, + "learning_rate": 2.8111695306488843e-05, + "loss": 0.4119, + "step": 34187 + }, + { + "epoch": 1.31, + "learning_rate": 2.8111054116440115e-05, + "loss": 0.3374, + "step": 34188 + }, + { + "epoch": 1.31, + "learning_rate": 2.8110412926391383e-05, + "loss": 0.3711, + "step": 34189 + }, + { + "epoch": 1.31, + "learning_rate": 2.8109771736342655e-05, + "loss": 0.565, + "step": 34190 + }, + { + "epoch": 1.31, + "learning_rate": 2.810913054629392e-05, + "loss": 0.3647, + "step": 34191 + }, + { + "epoch": 1.31, + "learning_rate": 2.8108489356245192e-05, + "loss": 0.3747, + "step": 34192 + }, + { + "epoch": 1.31, + "learning_rate": 2.810784816619646e-05, + "loss": 0.2858, + "step": 34193 + }, + { + "epoch": 1.31, + "learning_rate": 2.8107206976147732e-05, + "loss": 0.296, + "step": 34194 + }, + { + "epoch": 1.31, + "learning_rate": 2.8106565786098997e-05, + "loss": 0.2881, + "step": 34195 + }, + { + "epoch": 1.31, + "learning_rate": 2.8105924596050272e-05, + "loss": 0.4337, + "step": 34196 + }, + { + "epoch": 1.31, + "learning_rate": 2.8105283406001544e-05, + "loss": 0.4278, + "step": 34197 + }, + { + "epoch": 1.31, + "learning_rate": 2.810464221595281e-05, + "loss": 0.3994, + "step": 34198 + }, + { + "epoch": 1.31, + "learning_rate": 2.810400102590408e-05, + "loss": 0.3364, + "step": 34199 + }, + { + "epoch": 1.31, + "learning_rate": 2.810335983585535e-05, + "loss": 0.3586, + "step": 34200 + }, + { + "epoch": 1.31, + "learning_rate": 2.810271864580662e-05, + "loss": 0.2541, + "step": 34201 + }, + { + "epoch": 1.31, + "learning_rate": 2.8102077455757885e-05, + "loss": 0.377, + "step": 34202 + }, + { + "epoch": 1.31, + "learning_rate": 2.8101436265709157e-05, + "loss": 0.3677, + "step": 34203 + }, + { + "epoch": 1.32, + "learning_rate": 2.8100795075660425e-05, + "loss": 0.4889, + "step": 34204 + }, + { + "epoch": 1.32, + "learning_rate": 2.8100153885611697e-05, + "loss": 0.313, + "step": 34205 + }, + { + "epoch": 1.32, + "learning_rate": 2.809951269556297e-05, + "loss": 0.4281, + "step": 34206 + }, + { + "epoch": 1.32, + "learning_rate": 2.8098871505514234e-05, + "loss": 0.286, + "step": 34207 + }, + { + "epoch": 1.32, + "learning_rate": 2.8098230315465505e-05, + "loss": 0.3549, + "step": 34208 + }, + { + "epoch": 1.32, + "learning_rate": 2.8097589125416774e-05, + "loss": 0.4027, + "step": 34209 + }, + { + "epoch": 1.32, + "learning_rate": 2.8096947935368045e-05, + "loss": 0.4139, + "step": 34210 + }, + { + "epoch": 1.32, + "learning_rate": 2.809630674531931e-05, + "loss": 0.3275, + "step": 34211 + }, + { + "epoch": 1.32, + "learning_rate": 2.8095665555270585e-05, + "loss": 0.3803, + "step": 34212 + }, + { + "epoch": 1.32, + "learning_rate": 2.809502436522185e-05, + "loss": 0.4026, + "step": 34213 + }, + { + "epoch": 1.32, + "learning_rate": 2.8094383175173122e-05, + "loss": 0.4383, + "step": 34214 + }, + { + "epoch": 1.32, + "learning_rate": 2.8093741985124394e-05, + "loss": 0.4188, + "step": 34215 + }, + { + "epoch": 1.32, + "learning_rate": 2.8093100795075662e-05, + "loss": 0.4188, + "step": 34216 + }, + { + "epoch": 1.32, + "learning_rate": 2.8092459605026934e-05, + "loss": 0.4373, + "step": 34217 + }, + { + "epoch": 1.32, + "learning_rate": 2.80918184149782e-05, + "loss": 0.3142, + "step": 34218 + }, + { + "epoch": 1.32, + "learning_rate": 2.809117722492947e-05, + "loss": 0.2837, + "step": 34219 + }, + { + "epoch": 1.32, + "learning_rate": 2.809053603488074e-05, + "loss": 0.4491, + "step": 34220 + }, + { + "epoch": 1.32, + "learning_rate": 2.808989484483201e-05, + "loss": 0.3257, + "step": 34221 + }, + { + "epoch": 1.32, + "learning_rate": 2.8089253654783282e-05, + "loss": 0.3153, + "step": 34222 + }, + { + "epoch": 1.32, + "learning_rate": 2.8088612464734547e-05, + "loss": 0.3448, + "step": 34223 + }, + { + "epoch": 1.32, + "learning_rate": 2.808797127468582e-05, + "loss": 0.3937, + "step": 34224 + }, + { + "epoch": 1.32, + "learning_rate": 2.8087330084637087e-05, + "loss": 0.3165, + "step": 34225 + }, + { + "epoch": 1.32, + "learning_rate": 2.808668889458836e-05, + "loss": 0.3152, + "step": 34226 + }, + { + "epoch": 1.32, + "learning_rate": 2.8086047704539624e-05, + "loss": 0.4154, + "step": 34227 + }, + { + "epoch": 1.32, + "learning_rate": 2.80854065144909e-05, + "loss": 0.3516, + "step": 34228 + }, + { + "epoch": 1.32, + "learning_rate": 2.8084765324442164e-05, + "loss": 0.4802, + "step": 34229 + }, + { + "epoch": 1.32, + "learning_rate": 2.8084124134393436e-05, + "loss": 0.448, + "step": 34230 + }, + { + "epoch": 1.32, + "learning_rate": 2.8083482944344707e-05, + "loss": 0.5011, + "step": 34231 + }, + { + "epoch": 1.32, + "learning_rate": 2.8082841754295976e-05, + "loss": 0.2803, + "step": 34232 + }, + { + "epoch": 1.32, + "learning_rate": 2.8082200564247247e-05, + "loss": 0.4543, + "step": 34233 + }, + { + "epoch": 1.32, + "learning_rate": 2.8081559374198512e-05, + "loss": 0.3953, + "step": 34234 + }, + { + "epoch": 1.32, + "learning_rate": 2.8080918184149784e-05, + "loss": 0.3426, + "step": 34235 + }, + { + "epoch": 1.32, + "learning_rate": 2.8080276994101052e-05, + "loss": 0.4771, + "step": 34236 + }, + { + "epoch": 1.32, + "learning_rate": 2.8079635804052324e-05, + "loss": 0.398, + "step": 34237 + }, + { + "epoch": 1.32, + "learning_rate": 2.807899461400359e-05, + "loss": 0.3837, + "step": 34238 + }, + { + "epoch": 1.32, + "learning_rate": 2.807835342395486e-05, + "loss": 0.3718, + "step": 34239 + }, + { + "epoch": 1.32, + "learning_rate": 2.8077712233906132e-05, + "loss": 0.2914, + "step": 34240 + }, + { + "epoch": 1.32, + "learning_rate": 2.80770710438574e-05, + "loss": 0.5149, + "step": 34241 + }, + { + "epoch": 1.32, + "learning_rate": 2.8076429853808673e-05, + "loss": 0.3598, + "step": 34242 + }, + { + "epoch": 1.32, + "learning_rate": 2.8075788663759937e-05, + "loss": 0.3349, + "step": 34243 + }, + { + "epoch": 1.32, + "learning_rate": 2.807514747371121e-05, + "loss": 0.4268, + "step": 34244 + }, + { + "epoch": 1.32, + "learning_rate": 2.8074506283662478e-05, + "loss": 0.4109, + "step": 34245 + }, + { + "epoch": 1.32, + "learning_rate": 2.807386509361375e-05, + "loss": 0.3074, + "step": 34246 + }, + { + "epoch": 1.32, + "learning_rate": 2.8073223903565014e-05, + "loss": 0.3894, + "step": 34247 + }, + { + "epoch": 1.32, + "learning_rate": 2.807258271351629e-05, + "loss": 0.3135, + "step": 34248 + }, + { + "epoch": 1.32, + "learning_rate": 2.807194152346756e-05, + "loss": 0.5002, + "step": 34249 + }, + { + "epoch": 1.32, + "learning_rate": 2.8071300333418826e-05, + "loss": 0.291, + "step": 34250 + }, + { + "epoch": 1.32, + "learning_rate": 2.8070659143370098e-05, + "loss": 0.3829, + "step": 34251 + }, + { + "epoch": 1.32, + "learning_rate": 2.8070017953321366e-05, + "loss": 0.4706, + "step": 34252 + }, + { + "epoch": 1.32, + "learning_rate": 2.8069376763272638e-05, + "loss": 0.4461, + "step": 34253 + }, + { + "epoch": 1.32, + "learning_rate": 2.8068735573223903e-05, + "loss": 0.5054, + "step": 34254 + }, + { + "epoch": 1.32, + "learning_rate": 2.8068094383175174e-05, + "loss": 0.4554, + "step": 34255 + }, + { + "epoch": 1.32, + "learning_rate": 2.8067453193126443e-05, + "loss": 0.486, + "step": 34256 + }, + { + "epoch": 1.32, + "learning_rate": 2.8066812003077714e-05, + "loss": 0.42, + "step": 34257 + }, + { + "epoch": 1.32, + "learning_rate": 2.8066170813028986e-05, + "loss": 0.3004, + "step": 34258 + }, + { + "epoch": 1.32, + "learning_rate": 2.806552962298025e-05, + "loss": 0.4676, + "step": 34259 + }, + { + "epoch": 1.32, + "learning_rate": 2.8064888432931523e-05, + "loss": 0.3705, + "step": 34260 + }, + { + "epoch": 1.32, + "learning_rate": 2.806424724288279e-05, + "loss": 0.4113, + "step": 34261 + }, + { + "epoch": 1.32, + "learning_rate": 2.8063606052834063e-05, + "loss": 0.431, + "step": 34262 + }, + { + "epoch": 1.32, + "learning_rate": 2.8062964862785328e-05, + "loss": 0.4378, + "step": 34263 + }, + { + "epoch": 1.32, + "learning_rate": 2.8062323672736603e-05, + "loss": 0.3025, + "step": 34264 + }, + { + "epoch": 1.32, + "learning_rate": 2.8061682482687868e-05, + "loss": 0.4492, + "step": 34265 + }, + { + "epoch": 1.32, + "learning_rate": 2.806104129263914e-05, + "loss": 0.5048, + "step": 34266 + }, + { + "epoch": 1.32, + "learning_rate": 2.806040010259041e-05, + "loss": 0.4806, + "step": 34267 + }, + { + "epoch": 1.32, + "learning_rate": 2.805975891254168e-05, + "loss": 0.4205, + "step": 34268 + }, + { + "epoch": 1.32, + "learning_rate": 2.805911772249295e-05, + "loss": 0.4548, + "step": 34269 + }, + { + "epoch": 1.32, + "learning_rate": 2.8058476532444216e-05, + "loss": 0.4591, + "step": 34270 + }, + { + "epoch": 1.32, + "learning_rate": 2.8057835342395488e-05, + "loss": 0.3175, + "step": 34271 + }, + { + "epoch": 1.32, + "learning_rate": 2.8057194152346756e-05, + "loss": 0.524, + "step": 34272 + }, + { + "epoch": 1.32, + "learning_rate": 2.8056552962298028e-05, + "loss": 0.4751, + "step": 34273 + }, + { + "epoch": 1.32, + "learning_rate": 2.8055911772249293e-05, + "loss": 0.4062, + "step": 34274 + }, + { + "epoch": 1.32, + "learning_rate": 2.8055270582200565e-05, + "loss": 0.4709, + "step": 34275 + }, + { + "epoch": 1.32, + "learning_rate": 2.8054629392151836e-05, + "loss": 0.4094, + "step": 34276 + }, + { + "epoch": 1.32, + "learning_rate": 2.8053988202103105e-05, + "loss": 0.3278, + "step": 34277 + }, + { + "epoch": 1.32, + "learning_rate": 2.8053347012054376e-05, + "loss": 0.3466, + "step": 34278 + }, + { + "epoch": 1.32, + "learning_rate": 2.805270582200564e-05, + "loss": 0.4369, + "step": 34279 + }, + { + "epoch": 1.32, + "learning_rate": 2.8052064631956913e-05, + "loss": 0.4138, + "step": 34280 + }, + { + "epoch": 1.32, + "learning_rate": 2.805142344190818e-05, + "loss": 0.2952, + "step": 34281 + }, + { + "epoch": 1.32, + "learning_rate": 2.8050782251859453e-05, + "loss": 0.3225, + "step": 34282 + }, + { + "epoch": 1.32, + "learning_rate": 2.8050141061810718e-05, + "loss": 0.3874, + "step": 34283 + }, + { + "epoch": 1.32, + "learning_rate": 2.8049499871761993e-05, + "loss": 0.4291, + "step": 34284 + }, + { + "epoch": 1.32, + "learning_rate": 2.8048858681713265e-05, + "loss": 0.3765, + "step": 34285 + }, + { + "epoch": 1.32, + "learning_rate": 2.804821749166453e-05, + "loss": 0.3925, + "step": 34286 + }, + { + "epoch": 1.32, + "learning_rate": 2.80475763016158e-05, + "loss": 0.4431, + "step": 34287 + }, + { + "epoch": 1.32, + "learning_rate": 2.804693511156707e-05, + "loss": 0.3524, + "step": 34288 + }, + { + "epoch": 1.32, + "learning_rate": 2.804629392151834e-05, + "loss": 0.3465, + "step": 34289 + }, + { + "epoch": 1.32, + "learning_rate": 2.8045652731469606e-05, + "loss": 0.3871, + "step": 34290 + }, + { + "epoch": 1.32, + "learning_rate": 2.8045011541420878e-05, + "loss": 0.4347, + "step": 34291 + }, + { + "epoch": 1.32, + "learning_rate": 2.8044370351372147e-05, + "loss": 0.3835, + "step": 34292 + }, + { + "epoch": 1.32, + "learning_rate": 2.8043729161323418e-05, + "loss": 0.5099, + "step": 34293 + }, + { + "epoch": 1.32, + "learning_rate": 2.804308797127469e-05, + "loss": 0.3221, + "step": 34294 + }, + { + "epoch": 1.32, + "learning_rate": 2.8042446781225955e-05, + "loss": 0.4878, + "step": 34295 + }, + { + "epoch": 1.32, + "learning_rate": 2.8041805591177227e-05, + "loss": 0.3855, + "step": 34296 + }, + { + "epoch": 1.32, + "learning_rate": 2.8041164401128495e-05, + "loss": 0.3552, + "step": 34297 + }, + { + "epoch": 1.32, + "learning_rate": 2.8040523211079767e-05, + "loss": 0.3982, + "step": 34298 + }, + { + "epoch": 1.32, + "learning_rate": 2.803988202103103e-05, + "loss": 0.3862, + "step": 34299 + }, + { + "epoch": 1.32, + "learning_rate": 2.8039240830982307e-05, + "loss": 0.2733, + "step": 34300 + }, + { + "epoch": 1.32, + "learning_rate": 2.803859964093357e-05, + "loss": 0.421, + "step": 34301 + }, + { + "epoch": 1.32, + "learning_rate": 2.8037958450884843e-05, + "loss": 0.4233, + "step": 34302 + }, + { + "epoch": 1.32, + "learning_rate": 2.8037317260836115e-05, + "loss": 0.359, + "step": 34303 + }, + { + "epoch": 1.32, + "learning_rate": 2.8036676070787383e-05, + "loss": 0.4396, + "step": 34304 + }, + { + "epoch": 1.32, + "learning_rate": 2.8036034880738655e-05, + "loss": 0.3705, + "step": 34305 + }, + { + "epoch": 1.32, + "learning_rate": 2.803539369068992e-05, + "loss": 0.4327, + "step": 34306 + }, + { + "epoch": 1.32, + "learning_rate": 2.8034752500641192e-05, + "loss": 0.3687, + "step": 34307 + }, + { + "epoch": 1.32, + "learning_rate": 2.803411131059246e-05, + "loss": 0.3326, + "step": 34308 + }, + { + "epoch": 1.32, + "learning_rate": 2.8033470120543732e-05, + "loss": 0.3965, + "step": 34309 + }, + { + "epoch": 1.32, + "learning_rate": 2.8032828930494997e-05, + "loss": 0.4353, + "step": 34310 + }, + { + "epoch": 1.32, + "learning_rate": 2.803218774044627e-05, + "loss": 0.3434, + "step": 34311 + }, + { + "epoch": 1.32, + "learning_rate": 2.803154655039754e-05, + "loss": 0.3157, + "step": 34312 + }, + { + "epoch": 1.32, + "learning_rate": 2.803090536034881e-05, + "loss": 0.6008, + "step": 34313 + }, + { + "epoch": 1.32, + "learning_rate": 2.803026417030008e-05, + "loss": 0.3215, + "step": 34314 + }, + { + "epoch": 1.32, + "learning_rate": 2.8029622980251345e-05, + "loss": 0.2992, + "step": 34315 + }, + { + "epoch": 1.32, + "learning_rate": 2.8028981790202617e-05, + "loss": 0.2711, + "step": 34316 + }, + { + "epoch": 1.32, + "learning_rate": 2.8028340600153885e-05, + "loss": 0.4201, + "step": 34317 + }, + { + "epoch": 1.32, + "learning_rate": 2.8027699410105157e-05, + "loss": 0.5769, + "step": 34318 + }, + { + "epoch": 1.32, + "learning_rate": 2.8027058220056422e-05, + "loss": 0.4339, + "step": 34319 + }, + { + "epoch": 1.32, + "learning_rate": 2.8026417030007697e-05, + "loss": 0.3306, + "step": 34320 + }, + { + "epoch": 1.32, + "learning_rate": 2.802577583995897e-05, + "loss": 0.2889, + "step": 34321 + }, + { + "epoch": 1.32, + "learning_rate": 2.8025134649910234e-05, + "loss": 0.4539, + "step": 34322 + }, + { + "epoch": 1.32, + "learning_rate": 2.8024493459861505e-05, + "loss": 0.5451, + "step": 34323 + }, + { + "epoch": 1.32, + "learning_rate": 2.8023852269812774e-05, + "loss": 0.4409, + "step": 34324 + }, + { + "epoch": 1.32, + "learning_rate": 2.8023211079764045e-05, + "loss": 0.4032, + "step": 34325 + }, + { + "epoch": 1.32, + "learning_rate": 2.802256988971531e-05, + "loss": 0.335, + "step": 34326 + }, + { + "epoch": 1.32, + "learning_rate": 2.8021928699666582e-05, + "loss": 0.5199, + "step": 34327 + }, + { + "epoch": 1.32, + "learning_rate": 2.802128750961785e-05, + "loss": 0.3523, + "step": 34328 + }, + { + "epoch": 1.32, + "learning_rate": 2.8020646319569122e-05, + "loss": 0.37, + "step": 34329 + }, + { + "epoch": 1.32, + "learning_rate": 2.8020005129520394e-05, + "loss": 0.3482, + "step": 34330 + }, + { + "epoch": 1.32, + "learning_rate": 2.801936393947166e-05, + "loss": 0.3661, + "step": 34331 + }, + { + "epoch": 1.32, + "learning_rate": 2.801872274942293e-05, + "loss": 0.3501, + "step": 34332 + }, + { + "epoch": 1.32, + "learning_rate": 2.80180815593742e-05, + "loss": 0.3489, + "step": 34333 + }, + { + "epoch": 1.32, + "learning_rate": 2.801744036932547e-05, + "loss": 0.475, + "step": 34334 + }, + { + "epoch": 1.32, + "learning_rate": 2.8016799179276735e-05, + "loss": 0.4096, + "step": 34335 + }, + { + "epoch": 1.32, + "learning_rate": 2.801615798922801e-05, + "loss": 0.4193, + "step": 34336 + }, + { + "epoch": 1.32, + "learning_rate": 2.8015516799179276e-05, + "loss": 0.3541, + "step": 34337 + }, + { + "epoch": 1.32, + "learning_rate": 2.8014875609130547e-05, + "loss": 0.4683, + "step": 34338 + }, + { + "epoch": 1.32, + "learning_rate": 2.801423441908182e-05, + "loss": 0.3105, + "step": 34339 + }, + { + "epoch": 1.32, + "learning_rate": 2.8013593229033087e-05, + "loss": 0.3221, + "step": 34340 + }, + { + "epoch": 1.32, + "learning_rate": 2.801295203898436e-05, + "loss": 0.3492, + "step": 34341 + }, + { + "epoch": 1.32, + "learning_rate": 2.8012310848935624e-05, + "loss": 0.38, + "step": 34342 + }, + { + "epoch": 1.32, + "learning_rate": 2.8011669658886896e-05, + "loss": 0.3946, + "step": 34343 + }, + { + "epoch": 1.32, + "learning_rate": 2.8011028468838164e-05, + "loss": 0.3065, + "step": 34344 + }, + { + "epoch": 1.32, + "learning_rate": 2.8010387278789436e-05, + "loss": 0.4785, + "step": 34345 + }, + { + "epoch": 1.32, + "learning_rate": 2.80097460887407e-05, + "loss": 0.4015, + "step": 34346 + }, + { + "epoch": 1.32, + "learning_rate": 2.8009104898691972e-05, + "loss": 0.3313, + "step": 34347 + }, + { + "epoch": 1.32, + "learning_rate": 2.8008463708643244e-05, + "loss": 0.4513, + "step": 34348 + }, + { + "epoch": 1.32, + "learning_rate": 2.8007822518594512e-05, + "loss": 0.472, + "step": 34349 + }, + { + "epoch": 1.32, + "learning_rate": 2.8007181328545784e-05, + "loss": 0.3001, + "step": 34350 + }, + { + "epoch": 1.32, + "learning_rate": 2.800654013849705e-05, + "loss": 0.4005, + "step": 34351 + }, + { + "epoch": 1.32, + "learning_rate": 2.8005898948448324e-05, + "loss": 0.4134, + "step": 34352 + }, + { + "epoch": 1.32, + "learning_rate": 2.800525775839959e-05, + "loss": 0.4089, + "step": 34353 + }, + { + "epoch": 1.32, + "learning_rate": 2.800461656835086e-05, + "loss": 0.348, + "step": 34354 + }, + { + "epoch": 1.32, + "learning_rate": 2.800397537830213e-05, + "loss": 0.3292, + "step": 34355 + }, + { + "epoch": 1.32, + "learning_rate": 2.80033341882534e-05, + "loss": 0.4505, + "step": 34356 + }, + { + "epoch": 1.32, + "learning_rate": 2.8002692998204673e-05, + "loss": 0.5299, + "step": 34357 + }, + { + "epoch": 1.32, + "learning_rate": 2.8002051808155938e-05, + "loss": 0.3998, + "step": 34358 + }, + { + "epoch": 1.32, + "learning_rate": 2.800141061810721e-05, + "loss": 0.5256, + "step": 34359 + }, + { + "epoch": 1.32, + "learning_rate": 2.8000769428058478e-05, + "loss": 0.5291, + "step": 34360 + }, + { + "epoch": 1.32, + "learning_rate": 2.800012823800975e-05, + "loss": 0.373, + "step": 34361 + }, + { + "epoch": 1.32, + "learning_rate": 2.7999487047961014e-05, + "loss": 0.4286, + "step": 34362 + }, + { + "epoch": 1.32, + "learning_rate": 2.7998845857912286e-05, + "loss": 0.3573, + "step": 34363 + }, + { + "epoch": 1.32, + "learning_rate": 2.7998204667863554e-05, + "loss": 0.4165, + "step": 34364 + }, + { + "epoch": 1.32, + "learning_rate": 2.7997563477814826e-05, + "loss": 0.3329, + "step": 34365 + }, + { + "epoch": 1.32, + "learning_rate": 2.7996922287766098e-05, + "loss": 0.5111, + "step": 34366 + }, + { + "epoch": 1.32, + "learning_rate": 2.7996281097717363e-05, + "loss": 0.2504, + "step": 34367 + }, + { + "epoch": 1.32, + "learning_rate": 2.7995639907668634e-05, + "loss": 0.343, + "step": 34368 + }, + { + "epoch": 1.32, + "learning_rate": 2.7994998717619903e-05, + "loss": 0.5642, + "step": 34369 + }, + { + "epoch": 1.32, + "learning_rate": 2.7994357527571174e-05, + "loss": 0.4049, + "step": 34370 + }, + { + "epoch": 1.32, + "learning_rate": 2.799371633752244e-05, + "loss": 0.4649, + "step": 34371 + }, + { + "epoch": 1.32, + "learning_rate": 2.7993075147473714e-05, + "loss": 0.28, + "step": 34372 + }, + { + "epoch": 1.32, + "learning_rate": 2.799243395742498e-05, + "loss": 0.4142, + "step": 34373 + }, + { + "epoch": 1.32, + "learning_rate": 2.799179276737625e-05, + "loss": 0.351, + "step": 34374 + }, + { + "epoch": 1.32, + "learning_rate": 2.7991151577327523e-05, + "loss": 0.4447, + "step": 34375 + }, + { + "epoch": 1.32, + "learning_rate": 2.799051038727879e-05, + "loss": 0.3405, + "step": 34376 + }, + { + "epoch": 1.32, + "learning_rate": 2.7989869197230063e-05, + "loss": 0.3096, + "step": 34377 + }, + { + "epoch": 1.32, + "learning_rate": 2.7989228007181328e-05, + "loss": 0.4092, + "step": 34378 + }, + { + "epoch": 1.32, + "learning_rate": 2.79885868171326e-05, + "loss": 0.3856, + "step": 34379 + }, + { + "epoch": 1.32, + "learning_rate": 2.7987945627083868e-05, + "loss": 0.5324, + "step": 34380 + }, + { + "epoch": 1.32, + "learning_rate": 2.798730443703514e-05, + "loss": 0.3084, + "step": 34381 + }, + { + "epoch": 1.32, + "learning_rate": 2.7986663246986405e-05, + "loss": 0.3469, + "step": 34382 + }, + { + "epoch": 1.32, + "learning_rate": 2.7986022056937676e-05, + "loss": 0.3457, + "step": 34383 + }, + { + "epoch": 1.32, + "learning_rate": 2.7985380866888948e-05, + "loss": 0.3571, + "step": 34384 + }, + { + "epoch": 1.32, + "learning_rate": 2.7984739676840216e-05, + "loss": 0.3748, + "step": 34385 + }, + { + "epoch": 1.32, + "learning_rate": 2.7984098486791488e-05, + "loss": 0.3868, + "step": 34386 + }, + { + "epoch": 1.32, + "learning_rate": 2.7983457296742753e-05, + "loss": 0.3551, + "step": 34387 + }, + { + "epoch": 1.32, + "learning_rate": 2.7982816106694028e-05, + "loss": 0.3789, + "step": 34388 + }, + { + "epoch": 1.32, + "learning_rate": 2.7982174916645293e-05, + "loss": 0.2789, + "step": 34389 + }, + { + "epoch": 1.32, + "learning_rate": 2.7981533726596565e-05, + "loss": 0.3349, + "step": 34390 + }, + { + "epoch": 1.32, + "learning_rate": 2.7980892536547833e-05, + "loss": 0.4383, + "step": 34391 + }, + { + "epoch": 1.32, + "learning_rate": 2.7980251346499105e-05, + "loss": 0.2779, + "step": 34392 + }, + { + "epoch": 1.32, + "learning_rate": 2.7979610156450376e-05, + "loss": 0.3769, + "step": 34393 + }, + { + "epoch": 1.32, + "learning_rate": 2.797896896640164e-05, + "loss": 0.3473, + "step": 34394 + }, + { + "epoch": 1.32, + "learning_rate": 2.7978327776352913e-05, + "loss": 0.4592, + "step": 34395 + }, + { + "epoch": 1.32, + "learning_rate": 2.797768658630418e-05, + "loss": 0.3942, + "step": 34396 + }, + { + "epoch": 1.32, + "learning_rate": 2.7977045396255453e-05, + "loss": 0.4173, + "step": 34397 + }, + { + "epoch": 1.32, + "learning_rate": 2.7976404206206718e-05, + "loss": 0.4811, + "step": 34398 + }, + { + "epoch": 1.32, + "learning_rate": 2.797576301615799e-05, + "loss": 0.4593, + "step": 34399 + }, + { + "epoch": 1.32, + "learning_rate": 2.7975121826109258e-05, + "loss": 0.3971, + "step": 34400 + }, + { + "epoch": 1.32, + "learning_rate": 2.797448063606053e-05, + "loss": 0.4293, + "step": 34401 + }, + { + "epoch": 1.32, + "learning_rate": 2.79738394460118e-05, + "loss": 0.4422, + "step": 34402 + }, + { + "epoch": 1.32, + "learning_rate": 2.7973198255963067e-05, + "loss": 0.3575, + "step": 34403 + }, + { + "epoch": 1.32, + "learning_rate": 2.7972557065914338e-05, + "loss": 0.483, + "step": 34404 + }, + { + "epoch": 1.32, + "learning_rate": 2.7971915875865607e-05, + "loss": 0.452, + "step": 34405 + }, + { + "epoch": 1.32, + "learning_rate": 2.7971274685816878e-05, + "loss": 0.4329, + "step": 34406 + }, + { + "epoch": 1.32, + "learning_rate": 2.7970633495768143e-05, + "loss": 0.3269, + "step": 34407 + }, + { + "epoch": 1.32, + "learning_rate": 2.796999230571942e-05, + "loss": 0.4957, + "step": 34408 + }, + { + "epoch": 1.32, + "learning_rate": 2.796935111567069e-05, + "loss": 0.3381, + "step": 34409 + }, + { + "epoch": 1.32, + "learning_rate": 2.7968709925621955e-05, + "loss": 0.2896, + "step": 34410 + }, + { + "epoch": 1.32, + "learning_rate": 2.7968068735573227e-05, + "loss": 0.3719, + "step": 34411 + }, + { + "epoch": 1.32, + "learning_rate": 2.7967427545524495e-05, + "loss": 0.408, + "step": 34412 + }, + { + "epoch": 1.32, + "learning_rate": 2.7966786355475767e-05, + "loss": 0.3984, + "step": 34413 + }, + { + "epoch": 1.32, + "learning_rate": 2.7966145165427032e-05, + "loss": 0.3731, + "step": 34414 + }, + { + "epoch": 1.32, + "learning_rate": 2.7965503975378303e-05, + "loss": 0.3286, + "step": 34415 + }, + { + "epoch": 1.32, + "learning_rate": 2.7964862785329572e-05, + "loss": 0.4728, + "step": 34416 + }, + { + "epoch": 1.32, + "learning_rate": 2.7964221595280843e-05, + "loss": 0.3662, + "step": 34417 + }, + { + "epoch": 1.32, + "learning_rate": 2.7963580405232115e-05, + "loss": 0.3572, + "step": 34418 + }, + { + "epoch": 1.32, + "learning_rate": 2.796293921518338e-05, + "loss": 0.3792, + "step": 34419 + }, + { + "epoch": 1.32, + "learning_rate": 2.7962298025134652e-05, + "loss": 0.5293, + "step": 34420 + }, + { + "epoch": 1.32, + "learning_rate": 2.796165683508592e-05, + "loss": 0.3755, + "step": 34421 + }, + { + "epoch": 1.32, + "learning_rate": 2.7961015645037192e-05, + "loss": 0.3302, + "step": 34422 + }, + { + "epoch": 1.32, + "learning_rate": 2.7960374454988457e-05, + "loss": 0.3404, + "step": 34423 + }, + { + "epoch": 1.32, + "learning_rate": 2.7959733264939732e-05, + "loss": 0.2992, + "step": 34424 + }, + { + "epoch": 1.32, + "learning_rate": 2.7959092074890997e-05, + "loss": 0.397, + "step": 34425 + }, + { + "epoch": 1.32, + "learning_rate": 2.795845088484227e-05, + "loss": 0.2819, + "step": 34426 + }, + { + "epoch": 1.32, + "learning_rate": 2.795780969479354e-05, + "loss": 0.3741, + "step": 34427 + }, + { + "epoch": 1.32, + "learning_rate": 2.795716850474481e-05, + "loss": 0.4389, + "step": 34428 + }, + { + "epoch": 1.32, + "learning_rate": 2.795652731469608e-05, + "loss": 0.2615, + "step": 34429 + }, + { + "epoch": 1.32, + "learning_rate": 2.7955886124647345e-05, + "loss": 0.3538, + "step": 34430 + }, + { + "epoch": 1.32, + "learning_rate": 2.7955244934598617e-05, + "loss": 0.4206, + "step": 34431 + }, + { + "epoch": 1.32, + "learning_rate": 2.7954603744549885e-05, + "loss": 0.3876, + "step": 34432 + }, + { + "epoch": 1.32, + "learning_rate": 2.7953962554501157e-05, + "loss": 0.3677, + "step": 34433 + }, + { + "epoch": 1.32, + "learning_rate": 2.7953321364452422e-05, + "loss": 0.3241, + "step": 34434 + }, + { + "epoch": 1.32, + "learning_rate": 2.7952680174403694e-05, + "loss": 0.4428, + "step": 34435 + }, + { + "epoch": 1.32, + "learning_rate": 2.7952038984354965e-05, + "loss": 0.3634, + "step": 34436 + }, + { + "epoch": 1.32, + "learning_rate": 2.7951397794306234e-05, + "loss": 0.4532, + "step": 34437 + }, + { + "epoch": 1.32, + "learning_rate": 2.7950756604257505e-05, + "loss": 0.4508, + "step": 34438 + }, + { + "epoch": 1.32, + "learning_rate": 2.795011541420877e-05, + "loss": 0.4398, + "step": 34439 + }, + { + "epoch": 1.32, + "learning_rate": 2.7949474224160042e-05, + "loss": 0.3513, + "step": 34440 + }, + { + "epoch": 1.32, + "learning_rate": 2.794883303411131e-05, + "loss": 0.4681, + "step": 34441 + }, + { + "epoch": 1.32, + "learning_rate": 2.7948191844062582e-05, + "loss": 0.4785, + "step": 34442 + }, + { + "epoch": 1.32, + "learning_rate": 2.7947550654013847e-05, + "loss": 0.3178, + "step": 34443 + }, + { + "epoch": 1.32, + "learning_rate": 2.7946909463965122e-05, + "loss": 0.5102, + "step": 34444 + }, + { + "epoch": 1.32, + "learning_rate": 2.7946268273916394e-05, + "loss": 0.5414, + "step": 34445 + }, + { + "epoch": 1.32, + "learning_rate": 2.794562708386766e-05, + "loss": 0.3797, + "step": 34446 + }, + { + "epoch": 1.32, + "learning_rate": 2.794498589381893e-05, + "loss": 0.4032, + "step": 34447 + }, + { + "epoch": 1.32, + "learning_rate": 2.79443447037702e-05, + "loss": 0.3526, + "step": 34448 + }, + { + "epoch": 1.32, + "learning_rate": 2.794370351372147e-05, + "loss": 0.4039, + "step": 34449 + }, + { + "epoch": 1.32, + "learning_rate": 2.7943062323672736e-05, + "loss": 0.4985, + "step": 34450 + }, + { + "epoch": 1.32, + "learning_rate": 2.7942421133624007e-05, + "loss": 0.5323, + "step": 34451 + }, + { + "epoch": 1.32, + "learning_rate": 2.7941779943575276e-05, + "loss": 0.3602, + "step": 34452 + }, + { + "epoch": 1.32, + "learning_rate": 2.7941138753526547e-05, + "loss": 0.3365, + "step": 34453 + }, + { + "epoch": 1.32, + "learning_rate": 2.794049756347782e-05, + "loss": 0.4238, + "step": 34454 + }, + { + "epoch": 1.32, + "learning_rate": 2.7939856373429084e-05, + "loss": 0.2856, + "step": 34455 + }, + { + "epoch": 1.32, + "learning_rate": 2.7939215183380356e-05, + "loss": 0.3631, + "step": 34456 + }, + { + "epoch": 1.32, + "learning_rate": 2.7938573993331624e-05, + "loss": 0.3884, + "step": 34457 + }, + { + "epoch": 1.32, + "learning_rate": 2.7937932803282896e-05, + "loss": 0.4629, + "step": 34458 + }, + { + "epoch": 1.32, + "learning_rate": 2.793729161323416e-05, + "loss": 0.3242, + "step": 34459 + }, + { + "epoch": 1.32, + "learning_rate": 2.7936650423185436e-05, + "loss": 0.3267, + "step": 34460 + }, + { + "epoch": 1.32, + "learning_rate": 2.79360092331367e-05, + "loss": 0.4376, + "step": 34461 + }, + { + "epoch": 1.32, + "learning_rate": 2.7935368043087972e-05, + "loss": 0.3315, + "step": 34462 + }, + { + "epoch": 1.32, + "learning_rate": 2.7934726853039244e-05, + "loss": 0.367, + "step": 34463 + }, + { + "epoch": 1.33, + "learning_rate": 2.7934085662990512e-05, + "loss": 0.3813, + "step": 34464 + }, + { + "epoch": 1.33, + "learning_rate": 2.7933444472941784e-05, + "loss": 0.3384, + "step": 34465 + }, + { + "epoch": 1.33, + "learning_rate": 2.793280328289305e-05, + "loss": 0.3472, + "step": 34466 + }, + { + "epoch": 1.33, + "learning_rate": 2.793216209284432e-05, + "loss": 0.4415, + "step": 34467 + }, + { + "epoch": 1.33, + "learning_rate": 2.793152090279559e-05, + "loss": 0.3529, + "step": 34468 + }, + { + "epoch": 1.33, + "learning_rate": 2.793087971274686e-05, + "loss": 0.4319, + "step": 34469 + }, + { + "epoch": 1.33, + "learning_rate": 2.7930238522698126e-05, + "loss": 0.2816, + "step": 34470 + }, + { + "epoch": 1.33, + "learning_rate": 2.7929597332649398e-05, + "loss": 0.5161, + "step": 34471 + }, + { + "epoch": 1.33, + "learning_rate": 2.792895614260067e-05, + "loss": 0.4031, + "step": 34472 + }, + { + "epoch": 1.33, + "learning_rate": 2.7928314952551938e-05, + "loss": 0.252, + "step": 34473 + }, + { + "epoch": 1.33, + "learning_rate": 2.792767376250321e-05, + "loss": 0.4537, + "step": 34474 + }, + { + "epoch": 1.33, + "learning_rate": 2.7927032572454474e-05, + "loss": 0.3902, + "step": 34475 + }, + { + "epoch": 1.33, + "learning_rate": 2.792639138240575e-05, + "loss": 0.372, + "step": 34476 + }, + { + "epoch": 1.33, + "learning_rate": 2.7925750192357014e-05, + "loss": 0.3276, + "step": 34477 + }, + { + "epoch": 1.33, + "learning_rate": 2.7925109002308286e-05, + "loss": 0.2666, + "step": 34478 + }, + { + "epoch": 1.33, + "learning_rate": 2.792446781225955e-05, + "loss": 0.2806, + "step": 34479 + }, + { + "epoch": 1.33, + "learning_rate": 2.7923826622210826e-05, + "loss": 0.4393, + "step": 34480 + }, + { + "epoch": 1.33, + "learning_rate": 2.7923185432162098e-05, + "loss": 0.3698, + "step": 34481 + }, + { + "epoch": 1.33, + "learning_rate": 2.7922544242113363e-05, + "loss": 0.5857, + "step": 34482 + }, + { + "epoch": 1.33, + "learning_rate": 2.7921903052064634e-05, + "loss": 0.2904, + "step": 34483 + }, + { + "epoch": 1.33, + "learning_rate": 2.7921261862015903e-05, + "loss": 0.3626, + "step": 34484 + }, + { + "epoch": 1.33, + "learning_rate": 2.7920620671967175e-05, + "loss": 0.3386, + "step": 34485 + }, + { + "epoch": 1.33, + "learning_rate": 2.791997948191844e-05, + "loss": 0.2857, + "step": 34486 + }, + { + "epoch": 1.33, + "learning_rate": 2.791933829186971e-05, + "loss": 0.3799, + "step": 34487 + }, + { + "epoch": 1.33, + "learning_rate": 2.791869710182098e-05, + "loss": 0.3785, + "step": 34488 + }, + { + "epoch": 1.33, + "learning_rate": 2.791805591177225e-05, + "loss": 0.3342, + "step": 34489 + }, + { + "epoch": 1.33, + "learning_rate": 2.7917414721723523e-05, + "loss": 0.4145, + "step": 34490 + }, + { + "epoch": 1.33, + "learning_rate": 2.7916773531674788e-05, + "loss": 0.3704, + "step": 34491 + }, + { + "epoch": 1.33, + "learning_rate": 2.791613234162606e-05, + "loss": 0.3522, + "step": 34492 + }, + { + "epoch": 1.33, + "learning_rate": 2.7915491151577328e-05, + "loss": 0.3683, + "step": 34493 + }, + { + "epoch": 1.33, + "learning_rate": 2.79148499615286e-05, + "loss": 0.3755, + "step": 34494 + }, + { + "epoch": 1.33, + "learning_rate": 2.7914208771479865e-05, + "loss": 0.2615, + "step": 34495 + }, + { + "epoch": 1.33, + "learning_rate": 2.791356758143114e-05, + "loss": 0.5154, + "step": 34496 + }, + { + "epoch": 1.33, + "learning_rate": 2.7912926391382405e-05, + "loss": 0.2853, + "step": 34497 + }, + { + "epoch": 1.33, + "learning_rate": 2.7912285201333676e-05, + "loss": 0.4836, + "step": 34498 + }, + { + "epoch": 1.33, + "learning_rate": 2.7911644011284948e-05, + "loss": 0.3523, + "step": 34499 + }, + { + "epoch": 1.33, + "learning_rate": 2.7911002821236216e-05, + "loss": 0.3366, + "step": 34500 + }, + { + "epoch": 1.33, + "learning_rate": 2.7910361631187488e-05, + "loss": 0.4561, + "step": 34501 + }, + { + "epoch": 1.33, + "learning_rate": 2.7909720441138753e-05, + "loss": 0.3733, + "step": 34502 + }, + { + "epoch": 1.33, + "learning_rate": 2.7909079251090025e-05, + "loss": 0.4618, + "step": 34503 + }, + { + "epoch": 1.33, + "learning_rate": 2.7908438061041293e-05, + "loss": 0.3616, + "step": 34504 + }, + { + "epoch": 1.33, + "learning_rate": 2.7907796870992565e-05, + "loss": 0.2635, + "step": 34505 + }, + { + "epoch": 1.33, + "learning_rate": 2.790715568094383e-05, + "loss": 0.4094, + "step": 34506 + }, + { + "epoch": 1.33, + "learning_rate": 2.79065144908951e-05, + "loss": 0.3814, + "step": 34507 + }, + { + "epoch": 1.33, + "learning_rate": 2.7905873300846373e-05, + "loss": 0.5047, + "step": 34508 + }, + { + "epoch": 1.33, + "learning_rate": 2.790523211079764e-05, + "loss": 0.3367, + "step": 34509 + }, + { + "epoch": 1.33, + "learning_rate": 2.7904590920748913e-05, + "loss": 0.292, + "step": 34510 + }, + { + "epoch": 1.33, + "learning_rate": 2.7903949730700178e-05, + "loss": 0.4352, + "step": 34511 + }, + { + "epoch": 1.33, + "learning_rate": 2.7903308540651453e-05, + "loss": 0.2794, + "step": 34512 + }, + { + "epoch": 1.33, + "learning_rate": 2.7902667350602718e-05, + "loss": 0.324, + "step": 34513 + }, + { + "epoch": 1.33, + "learning_rate": 2.790202616055399e-05, + "loss": 0.3796, + "step": 34514 + }, + { + "epoch": 1.33, + "learning_rate": 2.7901384970505258e-05, + "loss": 0.374, + "step": 34515 + }, + { + "epoch": 1.33, + "learning_rate": 2.790074378045653e-05, + "loss": 0.3194, + "step": 34516 + }, + { + "epoch": 1.33, + "learning_rate": 2.79001025904078e-05, + "loss": 0.3931, + "step": 34517 + }, + { + "epoch": 1.33, + "learning_rate": 2.7899461400359067e-05, + "loss": 0.3791, + "step": 34518 + }, + { + "epoch": 1.33, + "learning_rate": 2.789882021031034e-05, + "loss": 0.3482, + "step": 34519 + }, + { + "epoch": 1.33, + "learning_rate": 2.7898179020261607e-05, + "loss": 0.3517, + "step": 34520 + }, + { + "epoch": 1.33, + "learning_rate": 2.789753783021288e-05, + "loss": 0.4968, + "step": 34521 + }, + { + "epoch": 1.33, + "learning_rate": 2.7896896640164143e-05, + "loss": 0.3409, + "step": 34522 + }, + { + "epoch": 1.33, + "learning_rate": 2.7896255450115415e-05, + "loss": 0.467, + "step": 34523 + }, + { + "epoch": 1.33, + "learning_rate": 2.7895614260066683e-05, + "loss": 0.3844, + "step": 34524 + }, + { + "epoch": 1.33, + "learning_rate": 2.7894973070017955e-05, + "loss": 0.4905, + "step": 34525 + }, + { + "epoch": 1.33, + "learning_rate": 2.7894331879969227e-05, + "loss": 0.4626, + "step": 34526 + }, + { + "epoch": 1.33, + "learning_rate": 2.7893690689920492e-05, + "loss": 0.3158, + "step": 34527 + }, + { + "epoch": 1.33, + "learning_rate": 2.7893049499871763e-05, + "loss": 0.4072, + "step": 34528 + }, + { + "epoch": 1.33, + "learning_rate": 2.7892408309823032e-05, + "loss": 0.3066, + "step": 34529 + }, + { + "epoch": 1.33, + "learning_rate": 2.7891767119774303e-05, + "loss": 0.3352, + "step": 34530 + }, + { + "epoch": 1.33, + "learning_rate": 2.789112592972557e-05, + "loss": 0.3077, + "step": 34531 + }, + { + "epoch": 1.33, + "learning_rate": 2.7890484739676844e-05, + "loss": 0.4352, + "step": 34532 + }, + { + "epoch": 1.33, + "learning_rate": 2.788984354962811e-05, + "loss": 0.502, + "step": 34533 + }, + { + "epoch": 1.33, + "learning_rate": 2.788920235957938e-05, + "loss": 0.4639, + "step": 34534 + }, + { + "epoch": 1.33, + "learning_rate": 2.7888561169530652e-05, + "loss": 0.384, + "step": 34535 + }, + { + "epoch": 1.33, + "learning_rate": 2.788791997948192e-05, + "loss": 0.2852, + "step": 34536 + }, + { + "epoch": 1.33, + "learning_rate": 2.7887278789433192e-05, + "loss": 0.3992, + "step": 34537 + }, + { + "epoch": 1.33, + "learning_rate": 2.7886637599384457e-05, + "loss": 0.3858, + "step": 34538 + }, + { + "epoch": 1.33, + "learning_rate": 2.788599640933573e-05, + "loss": 0.3629, + "step": 34539 + }, + { + "epoch": 1.33, + "learning_rate": 2.7885355219286997e-05, + "loss": 0.4572, + "step": 34540 + }, + { + "epoch": 1.33, + "learning_rate": 2.788471402923827e-05, + "loss": 0.3116, + "step": 34541 + }, + { + "epoch": 1.33, + "learning_rate": 2.7884072839189534e-05, + "loss": 0.391, + "step": 34542 + }, + { + "epoch": 1.33, + "learning_rate": 2.7883431649140805e-05, + "loss": 0.35, + "step": 34543 + }, + { + "epoch": 1.33, + "learning_rate": 2.7882790459092077e-05, + "loss": 0.2933, + "step": 34544 + }, + { + "epoch": 1.33, + "learning_rate": 2.7882149269043345e-05, + "loss": 0.3242, + "step": 34545 + }, + { + "epoch": 1.33, + "learning_rate": 2.7881508078994617e-05, + "loss": 0.4413, + "step": 34546 + }, + { + "epoch": 1.33, + "learning_rate": 2.7880866888945882e-05, + "loss": 0.5088, + "step": 34547 + }, + { + "epoch": 1.33, + "learning_rate": 2.7880225698897157e-05, + "loss": 0.3986, + "step": 34548 + }, + { + "epoch": 1.33, + "learning_rate": 2.7879584508848422e-05, + "loss": 0.2985, + "step": 34549 + }, + { + "epoch": 1.33, + "learning_rate": 2.7878943318799694e-05, + "loss": 0.3016, + "step": 34550 + }, + { + "epoch": 1.33, + "learning_rate": 2.7878302128750962e-05, + "loss": 0.35, + "step": 34551 + }, + { + "epoch": 1.33, + "learning_rate": 2.7877660938702234e-05, + "loss": 0.3974, + "step": 34552 + }, + { + "epoch": 1.33, + "learning_rate": 2.7877019748653506e-05, + "loss": 0.3651, + "step": 34553 + }, + { + "epoch": 1.33, + "learning_rate": 2.787637855860477e-05, + "loss": 0.3354, + "step": 34554 + }, + { + "epoch": 1.33, + "learning_rate": 2.7875737368556042e-05, + "loss": 0.4962, + "step": 34555 + }, + { + "epoch": 1.33, + "learning_rate": 2.787509617850731e-05, + "loss": 0.4713, + "step": 34556 + }, + { + "epoch": 1.33, + "learning_rate": 2.7874454988458582e-05, + "loss": 0.5548, + "step": 34557 + }, + { + "epoch": 1.33, + "learning_rate": 2.7873813798409847e-05, + "loss": 0.4559, + "step": 34558 + }, + { + "epoch": 1.33, + "learning_rate": 2.787317260836112e-05, + "loss": 0.3242, + "step": 34559 + }, + { + "epoch": 1.33, + "learning_rate": 2.7872531418312387e-05, + "loss": 0.4182, + "step": 34560 + }, + { + "epoch": 1.33, + "learning_rate": 2.787189022826366e-05, + "loss": 0.4216, + "step": 34561 + }, + { + "epoch": 1.33, + "learning_rate": 2.787124903821493e-05, + "loss": 0.4024, + "step": 34562 + }, + { + "epoch": 1.33, + "learning_rate": 2.7870607848166196e-05, + "loss": 0.4459, + "step": 34563 + }, + { + "epoch": 1.33, + "learning_rate": 2.7869966658117467e-05, + "loss": 0.4314, + "step": 34564 + }, + { + "epoch": 1.33, + "learning_rate": 2.7869325468068736e-05, + "loss": 0.2639, + "step": 34565 + }, + { + "epoch": 1.33, + "learning_rate": 2.7868684278020007e-05, + "loss": 0.413, + "step": 34566 + }, + { + "epoch": 1.33, + "learning_rate": 2.7868043087971272e-05, + "loss": 0.2326, + "step": 34567 + }, + { + "epoch": 1.33, + "learning_rate": 2.7867401897922547e-05, + "loss": 0.3798, + "step": 34568 + }, + { + "epoch": 1.33, + "learning_rate": 2.7866760707873812e-05, + "loss": 0.2697, + "step": 34569 + }, + { + "epoch": 1.33, + "learning_rate": 2.7866119517825084e-05, + "loss": 0.4003, + "step": 34570 + }, + { + "epoch": 1.33, + "learning_rate": 2.7865478327776356e-05, + "loss": 0.2809, + "step": 34571 + }, + { + "epoch": 1.33, + "learning_rate": 2.7864837137727624e-05, + "loss": 0.3942, + "step": 34572 + }, + { + "epoch": 1.33, + "learning_rate": 2.7864195947678896e-05, + "loss": 0.3853, + "step": 34573 + }, + { + "epoch": 1.33, + "learning_rate": 2.786355475763016e-05, + "loss": 0.313, + "step": 34574 + }, + { + "epoch": 1.33, + "learning_rate": 2.7862913567581432e-05, + "loss": 0.3516, + "step": 34575 + }, + { + "epoch": 1.33, + "learning_rate": 2.78622723775327e-05, + "loss": 0.3597, + "step": 34576 + }, + { + "epoch": 1.33, + "learning_rate": 2.7861631187483973e-05, + "loss": 0.4122, + "step": 34577 + }, + { + "epoch": 1.33, + "learning_rate": 2.7860989997435237e-05, + "loss": 0.4172, + "step": 34578 + }, + { + "epoch": 1.33, + "learning_rate": 2.786034880738651e-05, + "loss": 0.2326, + "step": 34579 + }, + { + "epoch": 1.33, + "learning_rate": 2.785970761733778e-05, + "loss": 0.4877, + "step": 34580 + }, + { + "epoch": 1.33, + "learning_rate": 2.785906642728905e-05, + "loss": 0.4447, + "step": 34581 + }, + { + "epoch": 1.33, + "learning_rate": 2.785842523724032e-05, + "loss": 0.3611, + "step": 34582 + }, + { + "epoch": 1.33, + "learning_rate": 2.7857784047191586e-05, + "loss": 0.4206, + "step": 34583 + }, + { + "epoch": 1.33, + "learning_rate": 2.785714285714286e-05, + "loss": 0.3011, + "step": 34584 + }, + { + "epoch": 1.33, + "learning_rate": 2.7856501667094126e-05, + "loss": 0.4324, + "step": 34585 + }, + { + "epoch": 1.33, + "learning_rate": 2.7855860477045398e-05, + "loss": 0.3085, + "step": 34586 + }, + { + "epoch": 1.33, + "learning_rate": 2.7855219286996666e-05, + "loss": 0.5827, + "step": 34587 + }, + { + "epoch": 1.33, + "learning_rate": 2.7854578096947938e-05, + "loss": 0.3596, + "step": 34588 + }, + { + "epoch": 1.33, + "learning_rate": 2.785393690689921e-05, + "loss": 0.5202, + "step": 34589 + }, + { + "epoch": 1.33, + "learning_rate": 2.7853295716850474e-05, + "loss": 0.3472, + "step": 34590 + }, + { + "epoch": 1.33, + "learning_rate": 2.7852654526801746e-05, + "loss": 0.3607, + "step": 34591 + }, + { + "epoch": 1.33, + "learning_rate": 2.7852013336753014e-05, + "loss": 0.4234, + "step": 34592 + }, + { + "epoch": 1.33, + "learning_rate": 2.7851372146704286e-05, + "loss": 0.3661, + "step": 34593 + }, + { + "epoch": 1.33, + "learning_rate": 2.785073095665555e-05, + "loss": 0.4402, + "step": 34594 + }, + { + "epoch": 1.33, + "learning_rate": 2.7850089766606823e-05, + "loss": 0.3888, + "step": 34595 + }, + { + "epoch": 1.33, + "learning_rate": 2.7849448576558094e-05, + "loss": 0.3625, + "step": 34596 + }, + { + "epoch": 1.33, + "learning_rate": 2.7848807386509363e-05, + "loss": 0.3749, + "step": 34597 + }, + { + "epoch": 1.33, + "learning_rate": 2.7848166196460635e-05, + "loss": 0.3727, + "step": 34598 + }, + { + "epoch": 1.33, + "learning_rate": 2.78475250064119e-05, + "loss": 0.4554, + "step": 34599 + }, + { + "epoch": 1.33, + "learning_rate": 2.7846883816363175e-05, + "loss": 0.4171, + "step": 34600 + }, + { + "epoch": 1.33, + "learning_rate": 2.784624262631444e-05, + "loss": 0.3675, + "step": 34601 + }, + { + "epoch": 1.33, + "learning_rate": 2.784560143626571e-05, + "loss": 0.4477, + "step": 34602 + }, + { + "epoch": 1.33, + "learning_rate": 2.7844960246216976e-05, + "loss": 0.3944, + "step": 34603 + }, + { + "epoch": 1.33, + "learning_rate": 2.784431905616825e-05, + "loss": 0.4002, + "step": 34604 + }, + { + "epoch": 1.33, + "learning_rate": 2.7843677866119523e-05, + "loss": 0.3945, + "step": 34605 + }, + { + "epoch": 1.33, + "learning_rate": 2.7843036676070788e-05, + "loss": 0.4417, + "step": 34606 + }, + { + "epoch": 1.33, + "learning_rate": 2.784239548602206e-05, + "loss": 0.2646, + "step": 34607 + }, + { + "epoch": 1.33, + "learning_rate": 2.7841754295973328e-05, + "loss": 0.2901, + "step": 34608 + }, + { + "epoch": 1.33, + "learning_rate": 2.78411131059246e-05, + "loss": 0.4397, + "step": 34609 + }, + { + "epoch": 1.33, + "learning_rate": 2.7840471915875865e-05, + "loss": 0.2663, + "step": 34610 + }, + { + "epoch": 1.33, + "learning_rate": 2.7839830725827136e-05, + "loss": 0.4171, + "step": 34611 + }, + { + "epoch": 1.33, + "learning_rate": 2.7839189535778405e-05, + "loss": 0.3561, + "step": 34612 + }, + { + "epoch": 1.33, + "learning_rate": 2.7838548345729676e-05, + "loss": 0.3596, + "step": 34613 + }, + { + "epoch": 1.33, + "learning_rate": 2.7837907155680948e-05, + "loss": 0.4021, + "step": 34614 + }, + { + "epoch": 1.33, + "learning_rate": 2.7837265965632213e-05, + "loss": 0.4532, + "step": 34615 + }, + { + "epoch": 1.33, + "learning_rate": 2.7836624775583485e-05, + "loss": 0.5457, + "step": 34616 + }, + { + "epoch": 1.33, + "learning_rate": 2.7835983585534753e-05, + "loss": 0.3933, + "step": 34617 + }, + { + "epoch": 1.33, + "learning_rate": 2.7835342395486025e-05, + "loss": 0.4286, + "step": 34618 + }, + { + "epoch": 1.33, + "learning_rate": 2.783470120543729e-05, + "loss": 0.4287, + "step": 34619 + }, + { + "epoch": 1.33, + "learning_rate": 2.7834060015388565e-05, + "loss": 0.3588, + "step": 34620 + }, + { + "epoch": 1.33, + "learning_rate": 2.783341882533983e-05, + "loss": 0.2781, + "step": 34621 + }, + { + "epoch": 1.33, + "learning_rate": 2.78327776352911e-05, + "loss": 0.3297, + "step": 34622 + }, + { + "epoch": 1.33, + "learning_rate": 2.7832136445242373e-05, + "loss": 0.3592, + "step": 34623 + }, + { + "epoch": 1.33, + "learning_rate": 2.783149525519364e-05, + "loss": 0.3677, + "step": 34624 + }, + { + "epoch": 1.33, + "learning_rate": 2.7830854065144913e-05, + "loss": 0.4532, + "step": 34625 + }, + { + "epoch": 1.33, + "learning_rate": 2.7830212875096178e-05, + "loss": 0.3529, + "step": 34626 + }, + { + "epoch": 1.33, + "learning_rate": 2.782957168504745e-05, + "loss": 0.2751, + "step": 34627 + }, + { + "epoch": 1.33, + "learning_rate": 2.7828930494998718e-05, + "loss": 0.2914, + "step": 34628 + }, + { + "epoch": 1.33, + "learning_rate": 2.782828930494999e-05, + "loss": 0.4275, + "step": 34629 + }, + { + "epoch": 1.33, + "learning_rate": 2.7827648114901255e-05, + "loss": 0.4056, + "step": 34630 + }, + { + "epoch": 1.33, + "learning_rate": 2.7827006924852527e-05, + "loss": 0.3914, + "step": 34631 + }, + { + "epoch": 1.33, + "learning_rate": 2.78263657348038e-05, + "loss": 0.3869, + "step": 34632 + }, + { + "epoch": 1.33, + "learning_rate": 2.7825724544755067e-05, + "loss": 0.3284, + "step": 34633 + }, + { + "epoch": 1.33, + "learning_rate": 2.782508335470634e-05, + "loss": 0.258, + "step": 34634 + }, + { + "epoch": 1.33, + "learning_rate": 2.7824442164657603e-05, + "loss": 0.2816, + "step": 34635 + }, + { + "epoch": 1.33, + "learning_rate": 2.782380097460888e-05, + "loss": 0.3518, + "step": 34636 + }, + { + "epoch": 1.33, + "learning_rate": 2.7823159784560143e-05, + "loss": 0.339, + "step": 34637 + }, + { + "epoch": 1.33, + "learning_rate": 2.7822518594511415e-05, + "loss": 0.4214, + "step": 34638 + }, + { + "epoch": 1.33, + "learning_rate": 2.7821877404462683e-05, + "loss": 0.4605, + "step": 34639 + }, + { + "epoch": 1.33, + "learning_rate": 2.7821236214413955e-05, + "loss": 0.3016, + "step": 34640 + }, + { + "epoch": 1.33, + "learning_rate": 2.7820595024365227e-05, + "loss": 0.3936, + "step": 34641 + }, + { + "epoch": 1.33, + "learning_rate": 2.7819953834316492e-05, + "loss": 0.3213, + "step": 34642 + }, + { + "epoch": 1.33, + "learning_rate": 2.7819312644267764e-05, + "loss": 0.3949, + "step": 34643 + }, + { + "epoch": 1.33, + "learning_rate": 2.7818671454219032e-05, + "loss": 0.266, + "step": 34644 + }, + { + "epoch": 1.33, + "learning_rate": 2.7818030264170304e-05, + "loss": 0.3311, + "step": 34645 + }, + { + "epoch": 1.33, + "learning_rate": 2.781738907412157e-05, + "loss": 0.3118, + "step": 34646 + }, + { + "epoch": 1.33, + "learning_rate": 2.781674788407284e-05, + "loss": 0.3157, + "step": 34647 + }, + { + "epoch": 1.33, + "learning_rate": 2.781610669402411e-05, + "loss": 0.3721, + "step": 34648 + }, + { + "epoch": 1.33, + "learning_rate": 2.781546550397538e-05, + "loss": 0.4765, + "step": 34649 + }, + { + "epoch": 1.33, + "learning_rate": 2.7814824313926652e-05, + "loss": 0.4298, + "step": 34650 + }, + { + "epoch": 1.33, + "learning_rate": 2.7814183123877917e-05, + "loss": 0.3879, + "step": 34651 + }, + { + "epoch": 1.33, + "learning_rate": 2.781354193382919e-05, + "loss": 0.3551, + "step": 34652 + }, + { + "epoch": 1.33, + "learning_rate": 2.7812900743780457e-05, + "loss": 0.3758, + "step": 34653 + }, + { + "epoch": 1.33, + "learning_rate": 2.781225955373173e-05, + "loss": 0.4866, + "step": 34654 + }, + { + "epoch": 1.33, + "learning_rate": 2.7811618363682994e-05, + "loss": 0.3358, + "step": 34655 + }, + { + "epoch": 1.33, + "learning_rate": 2.781097717363427e-05, + "loss": 0.5974, + "step": 34656 + }, + { + "epoch": 1.33, + "learning_rate": 2.7810335983585534e-05, + "loss": 0.4054, + "step": 34657 + }, + { + "epoch": 1.33, + "learning_rate": 2.7809694793536805e-05, + "loss": 0.3488, + "step": 34658 + }, + { + "epoch": 1.33, + "learning_rate": 2.7809053603488077e-05, + "loss": 0.5553, + "step": 34659 + }, + { + "epoch": 1.33, + "learning_rate": 2.7808412413439345e-05, + "loss": 0.464, + "step": 34660 + }, + { + "epoch": 1.33, + "learning_rate": 2.7807771223390617e-05, + "loss": 0.4839, + "step": 34661 + }, + { + "epoch": 1.33, + "learning_rate": 2.7807130033341882e-05, + "loss": 0.3391, + "step": 34662 + }, + { + "epoch": 1.33, + "learning_rate": 2.7806488843293154e-05, + "loss": 0.2723, + "step": 34663 + }, + { + "epoch": 1.33, + "learning_rate": 2.7805847653244422e-05, + "loss": 0.3854, + "step": 34664 + }, + { + "epoch": 1.33, + "learning_rate": 2.7805206463195694e-05, + "loss": 0.5266, + "step": 34665 + }, + { + "epoch": 1.33, + "learning_rate": 2.780456527314696e-05, + "loss": 0.4347, + "step": 34666 + }, + { + "epoch": 1.33, + "learning_rate": 2.780392408309823e-05, + "loss": 0.3355, + "step": 34667 + }, + { + "epoch": 1.33, + "learning_rate": 2.7803282893049502e-05, + "loss": 0.416, + "step": 34668 + }, + { + "epoch": 1.33, + "learning_rate": 2.780264170300077e-05, + "loss": 0.3832, + "step": 34669 + }, + { + "epoch": 1.33, + "learning_rate": 2.7802000512952042e-05, + "loss": 0.3552, + "step": 34670 + }, + { + "epoch": 1.33, + "learning_rate": 2.7801359322903307e-05, + "loss": 0.4377, + "step": 34671 + }, + { + "epoch": 1.33, + "learning_rate": 2.7800718132854582e-05, + "loss": 0.2422, + "step": 34672 + }, + { + "epoch": 1.33, + "learning_rate": 2.7800076942805847e-05, + "loss": 0.4039, + "step": 34673 + }, + { + "epoch": 1.33, + "learning_rate": 2.779943575275712e-05, + "loss": 0.4298, + "step": 34674 + }, + { + "epoch": 1.33, + "learning_rate": 2.7798794562708387e-05, + "loss": 0.3724, + "step": 34675 + }, + { + "epoch": 1.33, + "learning_rate": 2.779815337265966e-05, + "loss": 0.3514, + "step": 34676 + }, + { + "epoch": 1.33, + "learning_rate": 2.779751218261093e-05, + "loss": 0.3425, + "step": 34677 + }, + { + "epoch": 1.33, + "learning_rate": 2.7796870992562196e-05, + "loss": 0.53, + "step": 34678 + }, + { + "epoch": 1.33, + "learning_rate": 2.7796229802513467e-05, + "loss": 0.5762, + "step": 34679 + }, + { + "epoch": 1.33, + "learning_rate": 2.7795588612464736e-05, + "loss": 0.354, + "step": 34680 + }, + { + "epoch": 1.33, + "learning_rate": 2.7794947422416007e-05, + "loss": 0.4048, + "step": 34681 + }, + { + "epoch": 1.33, + "learning_rate": 2.7794306232367272e-05, + "loss": 0.3227, + "step": 34682 + }, + { + "epoch": 1.33, + "learning_rate": 2.7793665042318544e-05, + "loss": 0.3151, + "step": 34683 + }, + { + "epoch": 1.33, + "learning_rate": 2.7793023852269812e-05, + "loss": 0.4485, + "step": 34684 + }, + { + "epoch": 1.33, + "learning_rate": 2.7792382662221084e-05, + "loss": 0.3887, + "step": 34685 + }, + { + "epoch": 1.33, + "learning_rate": 2.7791741472172356e-05, + "loss": 0.3208, + "step": 34686 + }, + { + "epoch": 1.33, + "learning_rate": 2.779110028212362e-05, + "loss": 0.3976, + "step": 34687 + }, + { + "epoch": 1.33, + "learning_rate": 2.7790459092074893e-05, + "loss": 0.4494, + "step": 34688 + }, + { + "epoch": 1.33, + "learning_rate": 2.778981790202616e-05, + "loss": 0.4022, + "step": 34689 + }, + { + "epoch": 1.33, + "learning_rate": 2.7789176711977433e-05, + "loss": 0.3627, + "step": 34690 + }, + { + "epoch": 1.33, + "learning_rate": 2.7788535521928698e-05, + "loss": 0.3061, + "step": 34691 + }, + { + "epoch": 1.33, + "learning_rate": 2.7787894331879973e-05, + "loss": 0.2879, + "step": 34692 + }, + { + "epoch": 1.33, + "learning_rate": 2.7787253141831238e-05, + "loss": 0.436, + "step": 34693 + }, + { + "epoch": 1.33, + "learning_rate": 2.778661195178251e-05, + "loss": 0.3577, + "step": 34694 + }, + { + "epoch": 1.33, + "learning_rate": 2.778597076173378e-05, + "loss": 0.3276, + "step": 34695 + }, + { + "epoch": 1.33, + "learning_rate": 2.778532957168505e-05, + "loss": 0.4806, + "step": 34696 + }, + { + "epoch": 1.33, + "learning_rate": 2.778468838163632e-05, + "loss": 0.298, + "step": 34697 + }, + { + "epoch": 1.33, + "learning_rate": 2.7784047191587586e-05, + "loss": 0.3629, + "step": 34698 + }, + { + "epoch": 1.33, + "learning_rate": 2.7783406001538858e-05, + "loss": 0.4536, + "step": 34699 + }, + { + "epoch": 1.33, + "learning_rate": 2.7782764811490126e-05, + "loss": 0.3521, + "step": 34700 + }, + { + "epoch": 1.33, + "learning_rate": 2.7782123621441398e-05, + "loss": 0.4051, + "step": 34701 + }, + { + "epoch": 1.33, + "learning_rate": 2.7781482431392663e-05, + "loss": 0.323, + "step": 34702 + }, + { + "epoch": 1.33, + "learning_rate": 2.7780841241343934e-05, + "loss": 0.3307, + "step": 34703 + }, + { + "epoch": 1.33, + "learning_rate": 2.7780200051295206e-05, + "loss": 0.3999, + "step": 34704 + }, + { + "epoch": 1.33, + "learning_rate": 2.7779558861246474e-05, + "loss": 0.2785, + "step": 34705 + }, + { + "epoch": 1.33, + "learning_rate": 2.7778917671197746e-05, + "loss": 0.4565, + "step": 34706 + }, + { + "epoch": 1.33, + "learning_rate": 2.777827648114901e-05, + "loss": 0.3289, + "step": 34707 + }, + { + "epoch": 1.33, + "learning_rate": 2.7777635291100286e-05, + "loss": 0.3911, + "step": 34708 + }, + { + "epoch": 1.33, + "learning_rate": 2.777699410105155e-05, + "loss": 0.4125, + "step": 34709 + }, + { + "epoch": 1.33, + "learning_rate": 2.7776352911002823e-05, + "loss": 0.3971, + "step": 34710 + }, + { + "epoch": 1.33, + "learning_rate": 2.777571172095409e-05, + "loss": 0.5061, + "step": 34711 + }, + { + "epoch": 1.33, + "learning_rate": 2.7775070530905363e-05, + "loss": 0.4761, + "step": 34712 + }, + { + "epoch": 1.33, + "learning_rate": 2.7774429340856635e-05, + "loss": 0.3479, + "step": 34713 + }, + { + "epoch": 1.33, + "learning_rate": 2.77737881508079e-05, + "loss": 0.4972, + "step": 34714 + }, + { + "epoch": 1.33, + "learning_rate": 2.777314696075917e-05, + "loss": 0.5017, + "step": 34715 + }, + { + "epoch": 1.33, + "learning_rate": 2.777250577071044e-05, + "loss": 0.3422, + "step": 34716 + }, + { + "epoch": 1.33, + "learning_rate": 2.777186458066171e-05, + "loss": 0.365, + "step": 34717 + }, + { + "epoch": 1.33, + "learning_rate": 2.7771223390612976e-05, + "loss": 0.4097, + "step": 34718 + }, + { + "epoch": 1.33, + "learning_rate": 2.7770582200564248e-05, + "loss": 0.5272, + "step": 34719 + }, + { + "epoch": 1.33, + "learning_rate": 2.7769941010515516e-05, + "loss": 0.4598, + "step": 34720 + }, + { + "epoch": 1.33, + "learning_rate": 2.7769299820466788e-05, + "loss": 0.4596, + "step": 34721 + }, + { + "epoch": 1.33, + "learning_rate": 2.776865863041806e-05, + "loss": 0.3614, + "step": 34722 + }, + { + "epoch": 1.33, + "learning_rate": 2.7768017440369325e-05, + "loss": 0.3482, + "step": 34723 + }, + { + "epoch": 1.34, + "learning_rate": 2.7767376250320596e-05, + "loss": 0.2256, + "step": 34724 + }, + { + "epoch": 1.34, + "learning_rate": 2.7766735060271865e-05, + "loss": 0.2818, + "step": 34725 + }, + { + "epoch": 1.34, + "learning_rate": 2.7766093870223136e-05, + "loss": 0.3937, + "step": 34726 + }, + { + "epoch": 1.34, + "learning_rate": 2.77654526801744e-05, + "loss": 0.321, + "step": 34727 + }, + { + "epoch": 1.34, + "learning_rate": 2.7764811490125676e-05, + "loss": 0.5148, + "step": 34728 + }, + { + "epoch": 1.34, + "learning_rate": 2.776417030007694e-05, + "loss": 0.3424, + "step": 34729 + }, + { + "epoch": 1.34, + "learning_rate": 2.7763529110028213e-05, + "loss": 0.394, + "step": 34730 + }, + { + "epoch": 1.34, + "learning_rate": 2.7762887919979485e-05, + "loss": 0.3514, + "step": 34731 + }, + { + "epoch": 1.34, + "learning_rate": 2.7762246729930753e-05, + "loss": 0.3446, + "step": 34732 + }, + { + "epoch": 1.34, + "learning_rate": 2.7761605539882025e-05, + "loss": 0.3283, + "step": 34733 + }, + { + "epoch": 1.34, + "learning_rate": 2.776096434983329e-05, + "loss": 0.3391, + "step": 34734 + }, + { + "epoch": 1.34, + "learning_rate": 2.776032315978456e-05, + "loss": 0.3351, + "step": 34735 + }, + { + "epoch": 1.34, + "learning_rate": 2.775968196973583e-05, + "loss": 0.4091, + "step": 34736 + }, + { + "epoch": 1.34, + "learning_rate": 2.77590407796871e-05, + "loss": 0.3903, + "step": 34737 + }, + { + "epoch": 1.34, + "learning_rate": 2.7758399589638367e-05, + "loss": 0.3835, + "step": 34738 + }, + { + "epoch": 1.34, + "learning_rate": 2.7757758399589638e-05, + "loss": 0.321, + "step": 34739 + }, + { + "epoch": 1.34, + "learning_rate": 2.775711720954091e-05, + "loss": 0.3932, + "step": 34740 + }, + { + "epoch": 1.34, + "learning_rate": 2.7756476019492178e-05, + "loss": 0.592, + "step": 34741 + }, + { + "epoch": 1.34, + "learning_rate": 2.775583482944345e-05, + "loss": 0.3808, + "step": 34742 + }, + { + "epoch": 1.34, + "learning_rate": 2.7755193639394715e-05, + "loss": 0.3891, + "step": 34743 + }, + { + "epoch": 1.34, + "learning_rate": 2.775455244934599e-05, + "loss": 0.3328, + "step": 34744 + }, + { + "epoch": 1.34, + "learning_rate": 2.7753911259297255e-05, + "loss": 0.4014, + "step": 34745 + }, + { + "epoch": 1.34, + "learning_rate": 2.7753270069248527e-05, + "loss": 0.2979, + "step": 34746 + }, + { + "epoch": 1.34, + "learning_rate": 2.7752628879199795e-05, + "loss": 0.3782, + "step": 34747 + }, + { + "epoch": 1.34, + "learning_rate": 2.7751987689151067e-05, + "loss": 0.3045, + "step": 34748 + }, + { + "epoch": 1.34, + "learning_rate": 2.775134649910234e-05, + "loss": 0.4573, + "step": 34749 + }, + { + "epoch": 1.34, + "learning_rate": 2.7750705309053603e-05, + "loss": 0.3413, + "step": 34750 + }, + { + "epoch": 1.34, + "learning_rate": 2.7750064119004875e-05, + "loss": 0.4652, + "step": 34751 + }, + { + "epoch": 1.34, + "learning_rate": 2.7749422928956143e-05, + "loss": 0.3509, + "step": 34752 + }, + { + "epoch": 1.34, + "learning_rate": 2.7748781738907415e-05, + "loss": 0.2971, + "step": 34753 + }, + { + "epoch": 1.34, + "learning_rate": 2.774814054885868e-05, + "loss": 0.4002, + "step": 34754 + }, + { + "epoch": 1.34, + "learning_rate": 2.7747499358809952e-05, + "loss": 0.2178, + "step": 34755 + }, + { + "epoch": 1.34, + "learning_rate": 2.774685816876122e-05, + "loss": 0.323, + "step": 34756 + }, + { + "epoch": 1.34, + "learning_rate": 2.7746216978712492e-05, + "loss": 0.351, + "step": 34757 + }, + { + "epoch": 1.34, + "learning_rate": 2.7745575788663764e-05, + "loss": 0.3843, + "step": 34758 + }, + { + "epoch": 1.34, + "learning_rate": 2.774493459861503e-05, + "loss": 0.3493, + "step": 34759 + }, + { + "epoch": 1.34, + "learning_rate": 2.7744293408566304e-05, + "loss": 0.4726, + "step": 34760 + }, + { + "epoch": 1.34, + "learning_rate": 2.774365221851757e-05, + "loss": 0.3243, + "step": 34761 + }, + { + "epoch": 1.34, + "learning_rate": 2.774301102846884e-05, + "loss": 0.3798, + "step": 34762 + }, + { + "epoch": 1.34, + "learning_rate": 2.774236983842011e-05, + "loss": 0.382, + "step": 34763 + }, + { + "epoch": 1.34, + "learning_rate": 2.774172864837138e-05, + "loss": 0.3814, + "step": 34764 + }, + { + "epoch": 1.34, + "learning_rate": 2.7741087458322645e-05, + "loss": 0.4922, + "step": 34765 + }, + { + "epoch": 1.34, + "learning_rate": 2.7740446268273917e-05, + "loss": 0.4517, + "step": 34766 + }, + { + "epoch": 1.34, + "learning_rate": 2.773980507822519e-05, + "loss": 0.3151, + "step": 34767 + }, + { + "epoch": 1.34, + "learning_rate": 2.7739163888176457e-05, + "loss": 0.4341, + "step": 34768 + }, + { + "epoch": 1.34, + "learning_rate": 2.773852269812773e-05, + "loss": 0.2845, + "step": 34769 + }, + { + "epoch": 1.34, + "learning_rate": 2.7737881508078994e-05, + "loss": 0.3104, + "step": 34770 + }, + { + "epoch": 1.34, + "learning_rate": 2.7737240318030265e-05, + "loss": 0.4818, + "step": 34771 + }, + { + "epoch": 1.34, + "learning_rate": 2.7736599127981534e-05, + "loss": 0.2768, + "step": 34772 + }, + { + "epoch": 1.34, + "learning_rate": 2.7735957937932805e-05, + "loss": 0.4979, + "step": 34773 + }, + { + "epoch": 1.34, + "learning_rate": 2.773531674788407e-05, + "loss": 0.4508, + "step": 34774 + }, + { + "epoch": 1.34, + "learning_rate": 2.7734675557835342e-05, + "loss": 0.3101, + "step": 34775 + }, + { + "epoch": 1.34, + "learning_rate": 2.7734034367786614e-05, + "loss": 0.3805, + "step": 34776 + }, + { + "epoch": 1.34, + "learning_rate": 2.7733393177737882e-05, + "loss": 0.4447, + "step": 34777 + }, + { + "epoch": 1.34, + "learning_rate": 2.7732751987689154e-05, + "loss": 0.46, + "step": 34778 + }, + { + "epoch": 1.34, + "learning_rate": 2.773211079764042e-05, + "loss": 0.3571, + "step": 34779 + }, + { + "epoch": 1.34, + "learning_rate": 2.7731469607591694e-05, + "loss": 0.3493, + "step": 34780 + }, + { + "epoch": 1.34, + "learning_rate": 2.773082841754296e-05, + "loss": 0.413, + "step": 34781 + }, + { + "epoch": 1.34, + "learning_rate": 2.773018722749423e-05, + "loss": 0.4535, + "step": 34782 + }, + { + "epoch": 1.34, + "learning_rate": 2.7729546037445502e-05, + "loss": 0.4663, + "step": 34783 + }, + { + "epoch": 1.34, + "learning_rate": 2.772890484739677e-05, + "loss": 0.3822, + "step": 34784 + }, + { + "epoch": 1.34, + "learning_rate": 2.7728263657348042e-05, + "loss": 0.3615, + "step": 34785 + }, + { + "epoch": 1.34, + "learning_rate": 2.7727622467299307e-05, + "loss": 0.3756, + "step": 34786 + }, + { + "epoch": 1.34, + "learning_rate": 2.772698127725058e-05, + "loss": 0.4, + "step": 34787 + }, + { + "epoch": 1.34, + "learning_rate": 2.7726340087201847e-05, + "loss": 0.343, + "step": 34788 + }, + { + "epoch": 1.34, + "learning_rate": 2.772569889715312e-05, + "loss": 0.2877, + "step": 34789 + }, + { + "epoch": 1.34, + "learning_rate": 2.7725057707104384e-05, + "loss": 0.4011, + "step": 34790 + }, + { + "epoch": 1.34, + "learning_rate": 2.7724416517055656e-05, + "loss": 0.6129, + "step": 34791 + }, + { + "epoch": 1.34, + "learning_rate": 2.7723775327006927e-05, + "loss": 0.2666, + "step": 34792 + }, + { + "epoch": 1.34, + "learning_rate": 2.7723134136958196e-05, + "loss": 0.3495, + "step": 34793 + }, + { + "epoch": 1.34, + "learning_rate": 2.7722492946909467e-05, + "loss": 0.3684, + "step": 34794 + }, + { + "epoch": 1.34, + "learning_rate": 2.7721851756860732e-05, + "loss": 0.3526, + "step": 34795 + }, + { + "epoch": 1.34, + "learning_rate": 2.7721210566812008e-05, + "loss": 0.3627, + "step": 34796 + }, + { + "epoch": 1.34, + "learning_rate": 2.7720569376763272e-05, + "loss": 0.3878, + "step": 34797 + }, + { + "epoch": 1.34, + "learning_rate": 2.7719928186714544e-05, + "loss": 0.3201, + "step": 34798 + }, + { + "epoch": 1.34, + "learning_rate": 2.7719286996665813e-05, + "loss": 0.3767, + "step": 34799 + }, + { + "epoch": 1.34, + "learning_rate": 2.7718645806617084e-05, + "loss": 0.502, + "step": 34800 + }, + { + "epoch": 1.34, + "learning_rate": 2.7718004616568356e-05, + "loss": 0.2816, + "step": 34801 + }, + { + "epoch": 1.34, + "learning_rate": 2.771736342651962e-05, + "loss": 0.3896, + "step": 34802 + }, + { + "epoch": 1.34, + "learning_rate": 2.7716722236470893e-05, + "loss": 0.3265, + "step": 34803 + }, + { + "epoch": 1.34, + "learning_rate": 2.771608104642216e-05, + "loss": 0.5402, + "step": 34804 + }, + { + "epoch": 1.34, + "learning_rate": 2.7715439856373433e-05, + "loss": 0.3074, + "step": 34805 + }, + { + "epoch": 1.34, + "learning_rate": 2.7714798666324698e-05, + "loss": 0.3935, + "step": 34806 + }, + { + "epoch": 1.34, + "learning_rate": 2.771415747627597e-05, + "loss": 0.2973, + "step": 34807 + }, + { + "epoch": 1.34, + "learning_rate": 2.7713516286227238e-05, + "loss": 0.4821, + "step": 34808 + }, + { + "epoch": 1.34, + "learning_rate": 2.771287509617851e-05, + "loss": 0.2706, + "step": 34809 + }, + { + "epoch": 1.34, + "learning_rate": 2.771223390612978e-05, + "loss": 0.3668, + "step": 34810 + }, + { + "epoch": 1.34, + "learning_rate": 2.7711592716081046e-05, + "loss": 0.3703, + "step": 34811 + }, + { + "epoch": 1.34, + "learning_rate": 2.7710951526032318e-05, + "loss": 0.4142, + "step": 34812 + }, + { + "epoch": 1.34, + "learning_rate": 2.7710310335983586e-05, + "loss": 0.3839, + "step": 34813 + }, + { + "epoch": 1.34, + "learning_rate": 2.7709669145934858e-05, + "loss": 0.3, + "step": 34814 + }, + { + "epoch": 1.34, + "learning_rate": 2.7709027955886123e-05, + "loss": 0.4491, + "step": 34815 + }, + { + "epoch": 1.34, + "learning_rate": 2.7708386765837398e-05, + "loss": 0.5851, + "step": 34816 + }, + { + "epoch": 1.34, + "learning_rate": 2.7707745575788663e-05, + "loss": 0.2923, + "step": 34817 + }, + { + "epoch": 1.34, + "learning_rate": 2.7707104385739934e-05, + "loss": 0.2768, + "step": 34818 + }, + { + "epoch": 1.34, + "learning_rate": 2.7706463195691206e-05, + "loss": 0.3012, + "step": 34819 + }, + { + "epoch": 1.34, + "learning_rate": 2.7705822005642475e-05, + "loss": 0.4219, + "step": 34820 + }, + { + "epoch": 1.34, + "learning_rate": 2.7705180815593746e-05, + "loss": 0.2965, + "step": 34821 + }, + { + "epoch": 1.34, + "learning_rate": 2.770453962554501e-05, + "loss": 0.5668, + "step": 34822 + }, + { + "epoch": 1.34, + "learning_rate": 2.7703898435496283e-05, + "loss": 0.3016, + "step": 34823 + }, + { + "epoch": 1.34, + "learning_rate": 2.770325724544755e-05, + "loss": 0.2789, + "step": 34824 + }, + { + "epoch": 1.34, + "learning_rate": 2.7702616055398823e-05, + "loss": 0.3494, + "step": 34825 + }, + { + "epoch": 1.34, + "learning_rate": 2.7701974865350088e-05, + "loss": 0.3088, + "step": 34826 + }, + { + "epoch": 1.34, + "learning_rate": 2.770133367530136e-05, + "loss": 0.3596, + "step": 34827 + }, + { + "epoch": 1.34, + "learning_rate": 2.770069248525263e-05, + "loss": 0.3834, + "step": 34828 + }, + { + "epoch": 1.34, + "learning_rate": 2.77000512952039e-05, + "loss": 0.307, + "step": 34829 + }, + { + "epoch": 1.34, + "learning_rate": 2.769941010515517e-05, + "loss": 0.4756, + "step": 34830 + }, + { + "epoch": 1.34, + "learning_rate": 2.7698768915106436e-05, + "loss": 0.4421, + "step": 34831 + }, + { + "epoch": 1.34, + "learning_rate": 2.769812772505771e-05, + "loss": 0.4874, + "step": 34832 + }, + { + "epoch": 1.34, + "learning_rate": 2.7697486535008976e-05, + "loss": 0.4327, + "step": 34833 + }, + { + "epoch": 1.34, + "learning_rate": 2.7696845344960248e-05, + "loss": 0.3471, + "step": 34834 + }, + { + "epoch": 1.34, + "learning_rate": 2.7696204154911516e-05, + "loss": 0.3295, + "step": 34835 + }, + { + "epoch": 1.34, + "learning_rate": 2.7695562964862788e-05, + "loss": 0.5034, + "step": 34836 + }, + { + "epoch": 1.34, + "learning_rate": 2.769492177481406e-05, + "loss": 0.2917, + "step": 34837 + }, + { + "epoch": 1.34, + "learning_rate": 2.7694280584765325e-05, + "loss": 0.3961, + "step": 34838 + }, + { + "epoch": 1.34, + "learning_rate": 2.7693639394716596e-05, + "loss": 0.3944, + "step": 34839 + }, + { + "epoch": 1.34, + "learning_rate": 2.7692998204667865e-05, + "loss": 0.3564, + "step": 34840 + }, + { + "epoch": 1.34, + "learning_rate": 2.7692357014619137e-05, + "loss": 0.3321, + "step": 34841 + }, + { + "epoch": 1.34, + "learning_rate": 2.76917158245704e-05, + "loss": 0.4108, + "step": 34842 + }, + { + "epoch": 1.34, + "learning_rate": 2.7691074634521673e-05, + "loss": 0.4496, + "step": 34843 + }, + { + "epoch": 1.34, + "learning_rate": 2.769043344447294e-05, + "loss": 0.3145, + "step": 34844 + }, + { + "epoch": 1.34, + "learning_rate": 2.7689792254424213e-05, + "loss": 0.2939, + "step": 34845 + }, + { + "epoch": 1.34, + "learning_rate": 2.7689151064375485e-05, + "loss": 0.3067, + "step": 34846 + }, + { + "epoch": 1.34, + "learning_rate": 2.768850987432675e-05, + "loss": 0.3199, + "step": 34847 + }, + { + "epoch": 1.34, + "learning_rate": 2.768786868427802e-05, + "loss": 0.408, + "step": 34848 + }, + { + "epoch": 1.34, + "learning_rate": 2.768722749422929e-05, + "loss": 0.2397, + "step": 34849 + }, + { + "epoch": 1.34, + "learning_rate": 2.768658630418056e-05, + "loss": 0.3823, + "step": 34850 + }, + { + "epoch": 1.34, + "learning_rate": 2.7685945114131827e-05, + "loss": 0.3784, + "step": 34851 + }, + { + "epoch": 1.34, + "learning_rate": 2.76853039240831e-05, + "loss": 0.4797, + "step": 34852 + }, + { + "epoch": 1.34, + "learning_rate": 2.7684662734034367e-05, + "loss": 0.3891, + "step": 34853 + }, + { + "epoch": 1.34, + "learning_rate": 2.768402154398564e-05, + "loss": 0.5568, + "step": 34854 + }, + { + "epoch": 1.34, + "learning_rate": 2.768338035393691e-05, + "loss": 0.4106, + "step": 34855 + }, + { + "epoch": 1.34, + "learning_rate": 2.768273916388818e-05, + "loss": 0.3193, + "step": 34856 + }, + { + "epoch": 1.34, + "learning_rate": 2.768209797383945e-05, + "loss": 0.3232, + "step": 34857 + }, + { + "epoch": 1.34, + "learning_rate": 2.7681456783790715e-05, + "loss": 0.3239, + "step": 34858 + }, + { + "epoch": 1.34, + "learning_rate": 2.7680815593741987e-05, + "loss": 0.3607, + "step": 34859 + }, + { + "epoch": 1.34, + "learning_rate": 2.7680174403693255e-05, + "loss": 0.4373, + "step": 34860 + }, + { + "epoch": 1.34, + "learning_rate": 2.7679533213644527e-05, + "loss": 0.4341, + "step": 34861 + }, + { + "epoch": 1.34, + "learning_rate": 2.7678892023595792e-05, + "loss": 0.4913, + "step": 34862 + }, + { + "epoch": 1.34, + "learning_rate": 2.7678250833547063e-05, + "loss": 0.3282, + "step": 34863 + }, + { + "epoch": 1.34, + "learning_rate": 2.7677609643498335e-05, + "loss": 0.3214, + "step": 34864 + }, + { + "epoch": 1.34, + "learning_rate": 2.7676968453449604e-05, + "loss": 0.3935, + "step": 34865 + }, + { + "epoch": 1.34, + "learning_rate": 2.7676327263400875e-05, + "loss": 0.2711, + "step": 34866 + }, + { + "epoch": 1.34, + "learning_rate": 2.767568607335214e-05, + "loss": 0.4904, + "step": 34867 + }, + { + "epoch": 1.34, + "learning_rate": 2.7675044883303415e-05, + "loss": 0.2978, + "step": 34868 + }, + { + "epoch": 1.34, + "learning_rate": 2.767440369325468e-05, + "loss": 0.4077, + "step": 34869 + }, + { + "epoch": 1.34, + "learning_rate": 2.7673762503205952e-05, + "loss": 0.3509, + "step": 34870 + }, + { + "epoch": 1.34, + "learning_rate": 2.767312131315722e-05, + "loss": 0.3443, + "step": 34871 + }, + { + "epoch": 1.34, + "learning_rate": 2.7672480123108492e-05, + "loss": 0.243, + "step": 34872 + }, + { + "epoch": 1.34, + "learning_rate": 2.7671838933059764e-05, + "loss": 0.4079, + "step": 34873 + }, + { + "epoch": 1.34, + "learning_rate": 2.767119774301103e-05, + "loss": 0.283, + "step": 34874 + }, + { + "epoch": 1.34, + "learning_rate": 2.76705565529623e-05, + "loss": 0.4709, + "step": 34875 + }, + { + "epoch": 1.34, + "learning_rate": 2.766991536291357e-05, + "loss": 0.3447, + "step": 34876 + }, + { + "epoch": 1.34, + "learning_rate": 2.766927417286484e-05, + "loss": 0.3742, + "step": 34877 + }, + { + "epoch": 1.34, + "learning_rate": 2.7668632982816105e-05, + "loss": 0.3774, + "step": 34878 + }, + { + "epoch": 1.34, + "learning_rate": 2.7667991792767377e-05, + "loss": 0.3059, + "step": 34879 + }, + { + "epoch": 1.34, + "learning_rate": 2.7667350602718645e-05, + "loss": 0.2639, + "step": 34880 + }, + { + "epoch": 1.34, + "learning_rate": 2.7666709412669917e-05, + "loss": 0.2804, + "step": 34881 + }, + { + "epoch": 1.34, + "learning_rate": 2.766606822262119e-05, + "loss": 0.3686, + "step": 34882 + }, + { + "epoch": 1.34, + "learning_rate": 2.7665427032572454e-05, + "loss": 0.2693, + "step": 34883 + }, + { + "epoch": 1.34, + "learning_rate": 2.766478584252373e-05, + "loss": 0.3988, + "step": 34884 + }, + { + "epoch": 1.34, + "learning_rate": 2.7664144652474994e-05, + "loss": 0.3341, + "step": 34885 + }, + { + "epoch": 1.34, + "learning_rate": 2.7663503462426266e-05, + "loss": 0.4018, + "step": 34886 + }, + { + "epoch": 1.34, + "learning_rate": 2.7662862272377534e-05, + "loss": 0.4106, + "step": 34887 + }, + { + "epoch": 1.34, + "learning_rate": 2.7662221082328806e-05, + "loss": 0.316, + "step": 34888 + }, + { + "epoch": 1.34, + "learning_rate": 2.766157989228007e-05, + "loss": 0.2959, + "step": 34889 + }, + { + "epoch": 1.34, + "learning_rate": 2.7660938702231342e-05, + "loss": 0.3384, + "step": 34890 + }, + { + "epoch": 1.34, + "learning_rate": 2.7660297512182614e-05, + "loss": 0.5452, + "step": 34891 + }, + { + "epoch": 1.34, + "learning_rate": 2.7659656322133882e-05, + "loss": 0.4103, + "step": 34892 + }, + { + "epoch": 1.34, + "learning_rate": 2.7659015132085154e-05, + "loss": 0.31, + "step": 34893 + }, + { + "epoch": 1.34, + "learning_rate": 2.765837394203642e-05, + "loss": 0.4033, + "step": 34894 + }, + { + "epoch": 1.34, + "learning_rate": 2.765773275198769e-05, + "loss": 0.3344, + "step": 34895 + }, + { + "epoch": 1.34, + "learning_rate": 2.765709156193896e-05, + "loss": 0.2537, + "step": 34896 + }, + { + "epoch": 1.34, + "learning_rate": 2.765645037189023e-05, + "loss": 0.3957, + "step": 34897 + }, + { + "epoch": 1.34, + "learning_rate": 2.7655809181841496e-05, + "loss": 0.4035, + "step": 34898 + }, + { + "epoch": 1.34, + "learning_rate": 2.7655167991792767e-05, + "loss": 0.3808, + "step": 34899 + }, + { + "epoch": 1.34, + "learning_rate": 2.765452680174404e-05, + "loss": 0.4623, + "step": 34900 + }, + { + "epoch": 1.34, + "learning_rate": 2.7653885611695307e-05, + "loss": 0.2133, + "step": 34901 + }, + { + "epoch": 1.34, + "learning_rate": 2.765324442164658e-05, + "loss": 0.4394, + "step": 34902 + }, + { + "epoch": 1.34, + "learning_rate": 2.7652603231597844e-05, + "loss": 0.3216, + "step": 34903 + }, + { + "epoch": 1.34, + "learning_rate": 2.765196204154912e-05, + "loss": 0.3542, + "step": 34904 + }, + { + "epoch": 1.34, + "learning_rate": 2.7651320851500384e-05, + "loss": 0.3721, + "step": 34905 + }, + { + "epoch": 1.34, + "learning_rate": 2.7650679661451656e-05, + "loss": 0.4512, + "step": 34906 + }, + { + "epoch": 1.34, + "learning_rate": 2.7650038471402924e-05, + "loss": 0.3967, + "step": 34907 + }, + { + "epoch": 1.34, + "learning_rate": 2.7649397281354196e-05, + "loss": 0.4389, + "step": 34908 + }, + { + "epoch": 1.34, + "learning_rate": 2.7648756091305468e-05, + "loss": 0.3631, + "step": 34909 + }, + { + "epoch": 1.34, + "learning_rate": 2.7648114901256732e-05, + "loss": 0.3839, + "step": 34910 + }, + { + "epoch": 1.34, + "learning_rate": 2.7647473711208004e-05, + "loss": 0.2746, + "step": 34911 + }, + { + "epoch": 1.34, + "learning_rate": 2.7646832521159273e-05, + "loss": 0.331, + "step": 34912 + }, + { + "epoch": 1.34, + "learning_rate": 2.7646191331110544e-05, + "loss": 0.3373, + "step": 34913 + }, + { + "epoch": 1.34, + "learning_rate": 2.764555014106181e-05, + "loss": 0.3585, + "step": 34914 + }, + { + "epoch": 1.34, + "learning_rate": 2.764490895101308e-05, + "loss": 0.2789, + "step": 34915 + }, + { + "epoch": 1.34, + "learning_rate": 2.764426776096435e-05, + "loss": 0.3732, + "step": 34916 + }, + { + "epoch": 1.34, + "learning_rate": 2.764362657091562e-05, + "loss": 0.5066, + "step": 34917 + }, + { + "epoch": 1.34, + "learning_rate": 2.7642985380866893e-05, + "loss": 0.3505, + "step": 34918 + }, + { + "epoch": 1.34, + "learning_rate": 2.7642344190818158e-05, + "loss": 0.3449, + "step": 34919 + }, + { + "epoch": 1.34, + "learning_rate": 2.7641703000769433e-05, + "loss": 0.2822, + "step": 34920 + }, + { + "epoch": 1.34, + "learning_rate": 2.7641061810720698e-05, + "loss": 0.3809, + "step": 34921 + }, + { + "epoch": 1.34, + "learning_rate": 2.764042062067197e-05, + "loss": 0.2852, + "step": 34922 + }, + { + "epoch": 1.34, + "learning_rate": 2.7639779430623238e-05, + "loss": 0.3007, + "step": 34923 + }, + { + "epoch": 1.34, + "learning_rate": 2.763913824057451e-05, + "loss": 0.3866, + "step": 34924 + }, + { + "epoch": 1.34, + "learning_rate": 2.7638497050525774e-05, + "loss": 0.3576, + "step": 34925 + }, + { + "epoch": 1.34, + "learning_rate": 2.7637855860477046e-05, + "loss": 0.4539, + "step": 34926 + }, + { + "epoch": 1.34, + "learning_rate": 2.7637214670428318e-05, + "loss": 0.3441, + "step": 34927 + }, + { + "epoch": 1.34, + "learning_rate": 2.7636573480379586e-05, + "loss": 0.3971, + "step": 34928 + }, + { + "epoch": 1.34, + "learning_rate": 2.7635932290330858e-05, + "loss": 0.3639, + "step": 34929 + }, + { + "epoch": 1.34, + "learning_rate": 2.7635291100282123e-05, + "loss": 0.2782, + "step": 34930 + }, + { + "epoch": 1.34, + "learning_rate": 2.7634649910233394e-05, + "loss": 0.3941, + "step": 34931 + }, + { + "epoch": 1.34, + "learning_rate": 2.7634008720184663e-05, + "loss": 0.4845, + "step": 34932 + }, + { + "epoch": 1.34, + "learning_rate": 2.7633367530135935e-05, + "loss": 0.4283, + "step": 34933 + }, + { + "epoch": 1.34, + "learning_rate": 2.76327263400872e-05, + "loss": 0.3632, + "step": 34934 + }, + { + "epoch": 1.34, + "learning_rate": 2.763208515003847e-05, + "loss": 0.4098, + "step": 34935 + }, + { + "epoch": 1.34, + "learning_rate": 2.7631443959989743e-05, + "loss": 0.3323, + "step": 34936 + }, + { + "epoch": 1.34, + "learning_rate": 2.763080276994101e-05, + "loss": 0.4214, + "step": 34937 + }, + { + "epoch": 1.34, + "learning_rate": 2.7630161579892283e-05, + "loss": 0.3813, + "step": 34938 + }, + { + "epoch": 1.34, + "learning_rate": 2.7629520389843548e-05, + "loss": 0.427, + "step": 34939 + }, + { + "epoch": 1.34, + "learning_rate": 2.7628879199794823e-05, + "loss": 0.3912, + "step": 34940 + }, + { + "epoch": 1.34, + "learning_rate": 2.7628238009746088e-05, + "loss": 0.3379, + "step": 34941 + }, + { + "epoch": 1.34, + "learning_rate": 2.762759681969736e-05, + "loss": 0.4459, + "step": 34942 + }, + { + "epoch": 1.34, + "learning_rate": 2.7626955629648628e-05, + "loss": 0.3163, + "step": 34943 + }, + { + "epoch": 1.34, + "learning_rate": 2.76263144395999e-05, + "loss": 0.2927, + "step": 34944 + }, + { + "epoch": 1.34, + "learning_rate": 2.762567324955117e-05, + "loss": 0.4818, + "step": 34945 + }, + { + "epoch": 1.34, + "learning_rate": 2.7625032059502436e-05, + "loss": 0.2885, + "step": 34946 + }, + { + "epoch": 1.34, + "learning_rate": 2.7624390869453708e-05, + "loss": 0.2769, + "step": 34947 + }, + { + "epoch": 1.34, + "learning_rate": 2.7623749679404976e-05, + "loss": 0.5025, + "step": 34948 + }, + { + "epoch": 1.34, + "learning_rate": 2.7623108489356248e-05, + "loss": 0.4365, + "step": 34949 + }, + { + "epoch": 1.34, + "learning_rate": 2.7622467299307513e-05, + "loss": 0.2815, + "step": 34950 + }, + { + "epoch": 1.34, + "learning_rate": 2.7621826109258785e-05, + "loss": 0.3614, + "step": 34951 + }, + { + "epoch": 1.34, + "learning_rate": 2.7621184919210053e-05, + "loss": 0.3478, + "step": 34952 + }, + { + "epoch": 1.34, + "learning_rate": 2.7620543729161325e-05, + "loss": 0.3483, + "step": 34953 + }, + { + "epoch": 1.34, + "learning_rate": 2.7619902539112597e-05, + "loss": 0.3654, + "step": 34954 + }, + { + "epoch": 1.34, + "learning_rate": 2.761926134906386e-05, + "loss": 0.5172, + "step": 34955 + }, + { + "epoch": 1.34, + "learning_rate": 2.7618620159015137e-05, + "loss": 0.3195, + "step": 34956 + }, + { + "epoch": 1.34, + "learning_rate": 2.76179789689664e-05, + "loss": 0.5223, + "step": 34957 + }, + { + "epoch": 1.34, + "learning_rate": 2.7617337778917673e-05, + "loss": 0.2183, + "step": 34958 + }, + { + "epoch": 1.34, + "learning_rate": 2.761669658886894e-05, + "loss": 0.3301, + "step": 34959 + }, + { + "epoch": 1.34, + "learning_rate": 2.7616055398820213e-05, + "loss": 0.4835, + "step": 34960 + }, + { + "epoch": 1.34, + "learning_rate": 2.7615414208771478e-05, + "loss": 0.3534, + "step": 34961 + }, + { + "epoch": 1.34, + "learning_rate": 2.761477301872275e-05, + "loss": 0.3332, + "step": 34962 + }, + { + "epoch": 1.34, + "learning_rate": 2.761413182867402e-05, + "loss": 0.3825, + "step": 34963 + }, + { + "epoch": 1.34, + "learning_rate": 2.761349063862529e-05, + "loss": 0.4038, + "step": 34964 + }, + { + "epoch": 1.34, + "learning_rate": 2.7612849448576562e-05, + "loss": 0.4418, + "step": 34965 + }, + { + "epoch": 1.34, + "learning_rate": 2.7612208258527827e-05, + "loss": 0.4268, + "step": 34966 + }, + { + "epoch": 1.34, + "learning_rate": 2.76115670684791e-05, + "loss": 0.3463, + "step": 34967 + }, + { + "epoch": 1.34, + "learning_rate": 2.7610925878430367e-05, + "loss": 0.4428, + "step": 34968 + }, + { + "epoch": 1.34, + "learning_rate": 2.761028468838164e-05, + "loss": 0.4063, + "step": 34969 + }, + { + "epoch": 1.34, + "learning_rate": 2.7609643498332903e-05, + "loss": 0.3523, + "step": 34970 + }, + { + "epoch": 1.34, + "learning_rate": 2.7609002308284175e-05, + "loss": 0.3008, + "step": 34971 + }, + { + "epoch": 1.34, + "learning_rate": 2.7608361118235447e-05, + "loss": 0.355, + "step": 34972 + }, + { + "epoch": 1.34, + "learning_rate": 2.7607719928186715e-05, + "loss": 0.4212, + "step": 34973 + }, + { + "epoch": 1.34, + "learning_rate": 2.7607078738137987e-05, + "loss": 0.452, + "step": 34974 + }, + { + "epoch": 1.34, + "learning_rate": 2.7606437548089252e-05, + "loss": 0.2938, + "step": 34975 + }, + { + "epoch": 1.34, + "learning_rate": 2.7605796358040527e-05, + "loss": 0.415, + "step": 34976 + }, + { + "epoch": 1.34, + "learning_rate": 2.7605155167991792e-05, + "loss": 0.3966, + "step": 34977 + }, + { + "epoch": 1.34, + "learning_rate": 2.7604513977943064e-05, + "loss": 0.3471, + "step": 34978 + }, + { + "epoch": 1.34, + "learning_rate": 2.7603872787894335e-05, + "loss": 0.3177, + "step": 34979 + }, + { + "epoch": 1.34, + "learning_rate": 2.7603231597845604e-05, + "loss": 0.3865, + "step": 34980 + }, + { + "epoch": 1.34, + "learning_rate": 2.7602590407796875e-05, + "loss": 0.2141, + "step": 34981 + }, + { + "epoch": 1.34, + "learning_rate": 2.760194921774814e-05, + "loss": 0.3888, + "step": 34982 + }, + { + "epoch": 1.34, + "learning_rate": 2.7601308027699412e-05, + "loss": 0.618, + "step": 34983 + }, + { + "epoch": 1.35, + "learning_rate": 2.760066683765068e-05, + "loss": 0.3448, + "step": 34984 + }, + { + "epoch": 1.35, + "learning_rate": 2.7600025647601952e-05, + "loss": 0.4966, + "step": 34985 + }, + { + "epoch": 1.35, + "learning_rate": 2.7599384457553217e-05, + "loss": 0.3898, + "step": 34986 + }, + { + "epoch": 1.35, + "learning_rate": 2.759874326750449e-05, + "loss": 0.4385, + "step": 34987 + }, + { + "epoch": 1.35, + "learning_rate": 2.759810207745576e-05, + "loss": 0.3181, + "step": 34988 + }, + { + "epoch": 1.35, + "learning_rate": 2.759746088740703e-05, + "loss": 0.4211, + "step": 34989 + }, + { + "epoch": 1.35, + "learning_rate": 2.75968196973583e-05, + "loss": 0.3153, + "step": 34990 + }, + { + "epoch": 1.35, + "learning_rate": 2.7596178507309565e-05, + "loss": 0.4068, + "step": 34991 + }, + { + "epoch": 1.35, + "learning_rate": 2.759553731726084e-05, + "loss": 0.3515, + "step": 34992 + }, + { + "epoch": 1.35, + "learning_rate": 2.7594896127212105e-05, + "loss": 0.5169, + "step": 34993 + }, + { + "epoch": 1.35, + "learning_rate": 2.7594254937163377e-05, + "loss": 0.3175, + "step": 34994 + }, + { + "epoch": 1.35, + "learning_rate": 2.7593613747114645e-05, + "loss": 0.3967, + "step": 34995 + }, + { + "epoch": 1.35, + "learning_rate": 2.7592972557065917e-05, + "loss": 0.3442, + "step": 34996 + }, + { + "epoch": 1.35, + "learning_rate": 2.759233136701719e-05, + "loss": 0.3525, + "step": 34997 + }, + { + "epoch": 1.35, + "learning_rate": 2.7591690176968454e-05, + "loss": 0.4421, + "step": 34998 + }, + { + "epoch": 1.35, + "learning_rate": 2.7591048986919726e-05, + "loss": 0.3756, + "step": 34999 + }, + { + "epoch": 1.35, + "learning_rate": 2.7590407796870994e-05, + "loss": 0.421, + "step": 35000 + }, + { + "epoch": 1.35, + "learning_rate": 2.7589766606822266e-05, + "loss": 0.3634, + "step": 35001 + }, + { + "epoch": 1.35, + "learning_rate": 2.758912541677353e-05, + "loss": 0.3502, + "step": 35002 + }, + { + "epoch": 1.35, + "learning_rate": 2.7588484226724802e-05, + "loss": 0.2871, + "step": 35003 + }, + { + "epoch": 1.35, + "learning_rate": 2.758784303667607e-05, + "loss": 0.4275, + "step": 35004 + }, + { + "epoch": 1.35, + "learning_rate": 2.7587201846627342e-05, + "loss": 0.2358, + "step": 35005 + }, + { + "epoch": 1.35, + "learning_rate": 2.7586560656578614e-05, + "loss": 0.4138, + "step": 35006 + }, + { + "epoch": 1.35, + "learning_rate": 2.758591946652988e-05, + "loss": 0.5083, + "step": 35007 + }, + { + "epoch": 1.35, + "learning_rate": 2.7585278276481154e-05, + "loss": 0.3282, + "step": 35008 + }, + { + "epoch": 1.35, + "learning_rate": 2.758463708643242e-05, + "loss": 0.3989, + "step": 35009 + }, + { + "epoch": 1.35, + "learning_rate": 2.758399589638369e-05, + "loss": 0.3477, + "step": 35010 + }, + { + "epoch": 1.35, + "learning_rate": 2.7583354706334956e-05, + "loss": 0.4882, + "step": 35011 + }, + { + "epoch": 1.35, + "learning_rate": 2.758271351628623e-05, + "loss": 0.4328, + "step": 35012 + }, + { + "epoch": 1.35, + "learning_rate": 2.7582072326237496e-05, + "loss": 0.4128, + "step": 35013 + }, + { + "epoch": 1.35, + "learning_rate": 2.7581431136188767e-05, + "loss": 0.4717, + "step": 35014 + }, + { + "epoch": 1.35, + "learning_rate": 2.758078994614004e-05, + "loss": 0.3839, + "step": 35015 + }, + { + "epoch": 1.35, + "learning_rate": 2.7580148756091307e-05, + "loss": 0.3352, + "step": 35016 + }, + { + "epoch": 1.35, + "learning_rate": 2.757950756604258e-05, + "loss": 0.3031, + "step": 35017 + }, + { + "epoch": 1.35, + "learning_rate": 2.7578866375993844e-05, + "loss": 0.3526, + "step": 35018 + }, + { + "epoch": 1.35, + "learning_rate": 2.7578225185945116e-05, + "loss": 0.4499, + "step": 35019 + }, + { + "epoch": 1.35, + "learning_rate": 2.7577583995896384e-05, + "loss": 0.3918, + "step": 35020 + }, + { + "epoch": 1.35, + "learning_rate": 2.7576942805847656e-05, + "loss": 0.3232, + "step": 35021 + }, + { + "epoch": 1.35, + "learning_rate": 2.757630161579892e-05, + "loss": 0.3985, + "step": 35022 + }, + { + "epoch": 1.35, + "learning_rate": 2.7575660425750193e-05, + "loss": 0.4113, + "step": 35023 + }, + { + "epoch": 1.35, + "learning_rate": 2.7575019235701464e-05, + "loss": 0.3977, + "step": 35024 + }, + { + "epoch": 1.35, + "learning_rate": 2.7574378045652733e-05, + "loss": 0.335, + "step": 35025 + }, + { + "epoch": 1.35, + "learning_rate": 2.7573736855604004e-05, + "loss": 0.4351, + "step": 35026 + }, + { + "epoch": 1.35, + "learning_rate": 2.757309566555527e-05, + "loss": 0.5038, + "step": 35027 + }, + { + "epoch": 1.35, + "learning_rate": 2.7572454475506544e-05, + "loss": 0.2981, + "step": 35028 + }, + { + "epoch": 1.35, + "learning_rate": 2.757181328545781e-05, + "loss": 0.348, + "step": 35029 + }, + { + "epoch": 1.35, + "learning_rate": 2.757117209540908e-05, + "loss": 0.3948, + "step": 35030 + }, + { + "epoch": 1.35, + "learning_rate": 2.757053090536035e-05, + "loss": 0.4104, + "step": 35031 + }, + { + "epoch": 1.35, + "learning_rate": 2.756988971531162e-05, + "loss": 0.3069, + "step": 35032 + }, + { + "epoch": 1.35, + "learning_rate": 2.7569248525262893e-05, + "loss": 0.3907, + "step": 35033 + }, + { + "epoch": 1.35, + "learning_rate": 2.7568607335214158e-05, + "loss": 0.3228, + "step": 35034 + }, + { + "epoch": 1.35, + "learning_rate": 2.756796614516543e-05, + "loss": 0.3032, + "step": 35035 + }, + { + "epoch": 1.35, + "learning_rate": 2.7567324955116698e-05, + "loss": 0.3035, + "step": 35036 + }, + { + "epoch": 1.35, + "learning_rate": 2.756668376506797e-05, + "loss": 0.4188, + "step": 35037 + }, + { + "epoch": 1.35, + "learning_rate": 2.7566042575019234e-05, + "loss": 0.2396, + "step": 35038 + }, + { + "epoch": 1.35, + "learning_rate": 2.7565401384970506e-05, + "loss": 0.3836, + "step": 35039 + }, + { + "epoch": 1.35, + "learning_rate": 2.7564760194921774e-05, + "loss": 0.4205, + "step": 35040 + }, + { + "epoch": 1.35, + "learning_rate": 2.7564119004873046e-05, + "loss": 0.4517, + "step": 35041 + }, + { + "epoch": 1.35, + "learning_rate": 2.7563477814824318e-05, + "loss": 0.5045, + "step": 35042 + }, + { + "epoch": 1.35, + "learning_rate": 2.7562836624775583e-05, + "loss": 0.3465, + "step": 35043 + }, + { + "epoch": 1.35, + "learning_rate": 2.7562195434726858e-05, + "loss": 0.5175, + "step": 35044 + }, + { + "epoch": 1.35, + "learning_rate": 2.7561554244678123e-05, + "loss": 0.48, + "step": 35045 + }, + { + "epoch": 1.35, + "learning_rate": 2.7560913054629395e-05, + "loss": 0.4374, + "step": 35046 + }, + { + "epoch": 1.35, + "learning_rate": 2.7560271864580663e-05, + "loss": 0.5411, + "step": 35047 + }, + { + "epoch": 1.35, + "learning_rate": 2.7559630674531935e-05, + "loss": 0.4964, + "step": 35048 + }, + { + "epoch": 1.35, + "learning_rate": 2.75589894844832e-05, + "loss": 0.4397, + "step": 35049 + }, + { + "epoch": 1.35, + "learning_rate": 2.755834829443447e-05, + "loss": 0.3312, + "step": 35050 + }, + { + "epoch": 1.35, + "learning_rate": 2.7557707104385743e-05, + "loss": 0.4214, + "step": 35051 + }, + { + "epoch": 1.35, + "learning_rate": 2.755706591433701e-05, + "loss": 0.2733, + "step": 35052 + }, + { + "epoch": 1.35, + "learning_rate": 2.7556424724288283e-05, + "loss": 0.404, + "step": 35053 + }, + { + "epoch": 1.35, + "learning_rate": 2.7555783534239548e-05, + "loss": 0.3474, + "step": 35054 + }, + { + "epoch": 1.35, + "learning_rate": 2.755514234419082e-05, + "loss": 0.2829, + "step": 35055 + }, + { + "epoch": 1.35, + "learning_rate": 2.7554501154142088e-05, + "loss": 0.4302, + "step": 35056 + }, + { + "epoch": 1.35, + "learning_rate": 2.755385996409336e-05, + "loss": 0.392, + "step": 35057 + }, + { + "epoch": 1.35, + "learning_rate": 2.7553218774044625e-05, + "loss": 0.3868, + "step": 35058 + }, + { + "epoch": 1.35, + "learning_rate": 2.7552577583995896e-05, + "loss": 0.3268, + "step": 35059 + }, + { + "epoch": 1.35, + "learning_rate": 2.7551936393947168e-05, + "loss": 0.4607, + "step": 35060 + }, + { + "epoch": 1.35, + "learning_rate": 2.7551295203898436e-05, + "loss": 0.3895, + "step": 35061 + }, + { + "epoch": 1.35, + "learning_rate": 2.7550654013849708e-05, + "loss": 0.3499, + "step": 35062 + }, + { + "epoch": 1.35, + "learning_rate": 2.7550012823800973e-05, + "loss": 0.388, + "step": 35063 + }, + { + "epoch": 1.35, + "learning_rate": 2.7549371633752248e-05, + "loss": 0.4248, + "step": 35064 + }, + { + "epoch": 1.35, + "learning_rate": 2.7548730443703513e-05, + "loss": 0.3855, + "step": 35065 + }, + { + "epoch": 1.35, + "learning_rate": 2.7548089253654785e-05, + "loss": 0.2968, + "step": 35066 + }, + { + "epoch": 1.35, + "learning_rate": 2.7547448063606053e-05, + "loss": 0.4608, + "step": 35067 + }, + { + "epoch": 1.35, + "learning_rate": 2.7546806873557325e-05, + "loss": 0.3996, + "step": 35068 + }, + { + "epoch": 1.35, + "learning_rate": 2.7546165683508597e-05, + "loss": 0.3737, + "step": 35069 + }, + { + "epoch": 1.35, + "learning_rate": 2.754552449345986e-05, + "loss": 0.4377, + "step": 35070 + }, + { + "epoch": 1.35, + "learning_rate": 2.7544883303411133e-05, + "loss": 0.4105, + "step": 35071 + }, + { + "epoch": 1.35, + "learning_rate": 2.75442421133624e-05, + "loss": 0.4259, + "step": 35072 + }, + { + "epoch": 1.35, + "learning_rate": 2.7543600923313673e-05, + "loss": 0.4, + "step": 35073 + }, + { + "epoch": 1.35, + "learning_rate": 2.7542959733264938e-05, + "loss": 0.3163, + "step": 35074 + }, + { + "epoch": 1.35, + "learning_rate": 2.754231854321621e-05, + "loss": 0.4187, + "step": 35075 + }, + { + "epoch": 1.35, + "learning_rate": 2.754167735316748e-05, + "loss": 0.3961, + "step": 35076 + }, + { + "epoch": 1.35, + "learning_rate": 2.754103616311875e-05, + "loss": 0.3619, + "step": 35077 + }, + { + "epoch": 1.35, + "learning_rate": 2.7540394973070022e-05, + "loss": 0.2794, + "step": 35078 + }, + { + "epoch": 1.35, + "learning_rate": 2.7539753783021287e-05, + "loss": 0.3956, + "step": 35079 + }, + { + "epoch": 1.35, + "learning_rate": 2.7539112592972562e-05, + "loss": 0.4679, + "step": 35080 + }, + { + "epoch": 1.35, + "learning_rate": 2.7538471402923827e-05, + "loss": 0.3435, + "step": 35081 + }, + { + "epoch": 1.35, + "learning_rate": 2.75378302128751e-05, + "loss": 0.6267, + "step": 35082 + }, + { + "epoch": 1.35, + "learning_rate": 2.7537189022826367e-05, + "loss": 0.409, + "step": 35083 + }, + { + "epoch": 1.35, + "learning_rate": 2.753654783277764e-05, + "loss": 0.3501, + "step": 35084 + }, + { + "epoch": 1.35, + "learning_rate": 2.7535906642728903e-05, + "loss": 0.3152, + "step": 35085 + }, + { + "epoch": 1.35, + "learning_rate": 2.7535265452680175e-05, + "loss": 0.2949, + "step": 35086 + }, + { + "epoch": 1.35, + "learning_rate": 2.7534624262631447e-05, + "loss": 0.4184, + "step": 35087 + }, + { + "epoch": 1.35, + "learning_rate": 2.7533983072582715e-05, + "loss": 0.4582, + "step": 35088 + }, + { + "epoch": 1.35, + "learning_rate": 2.7533341882533987e-05, + "loss": 0.2946, + "step": 35089 + }, + { + "epoch": 1.35, + "learning_rate": 2.7532700692485252e-05, + "loss": 0.4436, + "step": 35090 + }, + { + "epoch": 1.35, + "learning_rate": 2.7532059502436524e-05, + "loss": 0.3249, + "step": 35091 + }, + { + "epoch": 1.35, + "learning_rate": 2.7531418312387792e-05, + "loss": 0.5575, + "step": 35092 + }, + { + "epoch": 1.35, + "learning_rate": 2.7530777122339064e-05, + "loss": 0.3167, + "step": 35093 + }, + { + "epoch": 1.35, + "learning_rate": 2.753013593229033e-05, + "loss": 0.396, + "step": 35094 + }, + { + "epoch": 1.35, + "learning_rate": 2.75294947422416e-05, + "loss": 0.3092, + "step": 35095 + }, + { + "epoch": 1.35, + "learning_rate": 2.7528853552192872e-05, + "loss": 0.3518, + "step": 35096 + }, + { + "epoch": 1.35, + "learning_rate": 2.752821236214414e-05, + "loss": 0.3746, + "step": 35097 + }, + { + "epoch": 1.35, + "learning_rate": 2.7527571172095412e-05, + "loss": 0.5113, + "step": 35098 + }, + { + "epoch": 1.35, + "learning_rate": 2.7526929982046677e-05, + "loss": 0.5419, + "step": 35099 + }, + { + "epoch": 1.35, + "learning_rate": 2.7526288791997952e-05, + "loss": 0.4386, + "step": 35100 + }, + { + "epoch": 1.35, + "learning_rate": 2.7525647601949217e-05, + "loss": 0.3616, + "step": 35101 + }, + { + "epoch": 1.35, + "learning_rate": 2.752500641190049e-05, + "loss": 0.4892, + "step": 35102 + }, + { + "epoch": 1.35, + "learning_rate": 2.7524365221851757e-05, + "loss": 0.3752, + "step": 35103 + }, + { + "epoch": 1.35, + "learning_rate": 2.752372403180303e-05, + "loss": 0.2975, + "step": 35104 + }, + { + "epoch": 1.35, + "learning_rate": 2.75230828417543e-05, + "loss": 0.4932, + "step": 35105 + }, + { + "epoch": 1.35, + "learning_rate": 2.7522441651705565e-05, + "loss": 0.2513, + "step": 35106 + }, + { + "epoch": 1.35, + "learning_rate": 2.7521800461656837e-05, + "loss": 0.3279, + "step": 35107 + }, + { + "epoch": 1.35, + "learning_rate": 2.7521159271608105e-05, + "loss": 0.449, + "step": 35108 + }, + { + "epoch": 1.35, + "learning_rate": 2.7520518081559377e-05, + "loss": 0.3249, + "step": 35109 + }, + { + "epoch": 1.35, + "learning_rate": 2.7519876891510642e-05, + "loss": 0.3762, + "step": 35110 + }, + { + "epoch": 1.35, + "learning_rate": 2.7519235701461914e-05, + "loss": 0.3911, + "step": 35111 + }, + { + "epoch": 1.35, + "learning_rate": 2.7518594511413182e-05, + "loss": 0.324, + "step": 35112 + }, + { + "epoch": 1.35, + "learning_rate": 2.7517953321364454e-05, + "loss": 0.519, + "step": 35113 + }, + { + "epoch": 1.35, + "learning_rate": 2.7517312131315726e-05, + "loss": 0.2735, + "step": 35114 + }, + { + "epoch": 1.35, + "learning_rate": 2.751667094126699e-05, + "loss": 0.3472, + "step": 35115 + }, + { + "epoch": 1.35, + "learning_rate": 2.7516029751218266e-05, + "loss": 0.2208, + "step": 35116 + }, + { + "epoch": 1.35, + "learning_rate": 2.751538856116953e-05, + "loss": 0.4183, + "step": 35117 + }, + { + "epoch": 1.35, + "learning_rate": 2.7514747371120802e-05, + "loss": 0.4652, + "step": 35118 + }, + { + "epoch": 1.35, + "learning_rate": 2.751410618107207e-05, + "loss": 0.3849, + "step": 35119 + }, + { + "epoch": 1.35, + "learning_rate": 2.7513464991023342e-05, + "loss": 0.3753, + "step": 35120 + }, + { + "epoch": 1.35, + "learning_rate": 2.7512823800974607e-05, + "loss": 0.4119, + "step": 35121 + }, + { + "epoch": 1.35, + "learning_rate": 2.751218261092588e-05, + "loss": 0.3986, + "step": 35122 + }, + { + "epoch": 1.35, + "learning_rate": 2.751154142087715e-05, + "loss": 0.3506, + "step": 35123 + }, + { + "epoch": 1.35, + "learning_rate": 2.751090023082842e-05, + "loss": 0.444, + "step": 35124 + }, + { + "epoch": 1.35, + "learning_rate": 2.751025904077969e-05, + "loss": 0.3903, + "step": 35125 + }, + { + "epoch": 1.35, + "learning_rate": 2.7509617850730956e-05, + "loss": 0.3624, + "step": 35126 + }, + { + "epoch": 1.35, + "learning_rate": 2.7508976660682227e-05, + "loss": 0.4364, + "step": 35127 + }, + { + "epoch": 1.35, + "learning_rate": 2.7508335470633496e-05, + "loss": 0.3835, + "step": 35128 + }, + { + "epoch": 1.35, + "learning_rate": 2.7507694280584767e-05, + "loss": 0.3004, + "step": 35129 + }, + { + "epoch": 1.35, + "learning_rate": 2.7507053090536032e-05, + "loss": 0.376, + "step": 35130 + }, + { + "epoch": 1.35, + "learning_rate": 2.7506411900487304e-05, + "loss": 0.301, + "step": 35131 + }, + { + "epoch": 1.35, + "learning_rate": 2.750577071043858e-05, + "loss": 0.2764, + "step": 35132 + }, + { + "epoch": 1.35, + "learning_rate": 2.7505129520389844e-05, + "loss": 0.4631, + "step": 35133 + }, + { + "epoch": 1.35, + "learning_rate": 2.7504488330341116e-05, + "loss": 0.3221, + "step": 35134 + }, + { + "epoch": 1.35, + "learning_rate": 2.750384714029238e-05, + "loss": 0.4211, + "step": 35135 + }, + { + "epoch": 1.35, + "learning_rate": 2.7503205950243656e-05, + "loss": 0.4336, + "step": 35136 + }, + { + "epoch": 1.35, + "learning_rate": 2.750256476019492e-05, + "loss": 0.3318, + "step": 35137 + }, + { + "epoch": 1.35, + "learning_rate": 2.7501923570146193e-05, + "loss": 0.3344, + "step": 35138 + }, + { + "epoch": 1.35, + "learning_rate": 2.750128238009746e-05, + "loss": 0.3044, + "step": 35139 + }, + { + "epoch": 1.35, + "learning_rate": 2.7500641190048733e-05, + "loss": 0.412, + "step": 35140 + }, + { + "epoch": 1.35, + "learning_rate": 2.7500000000000004e-05, + "loss": 0.4039, + "step": 35141 + }, + { + "epoch": 1.35, + "learning_rate": 2.749935880995127e-05, + "loss": 0.3815, + "step": 35142 + }, + { + "epoch": 1.35, + "learning_rate": 2.749871761990254e-05, + "loss": 0.4006, + "step": 35143 + }, + { + "epoch": 1.35, + "learning_rate": 2.749807642985381e-05, + "loss": 0.3697, + "step": 35144 + }, + { + "epoch": 1.35, + "learning_rate": 2.749743523980508e-05, + "loss": 0.2718, + "step": 35145 + }, + { + "epoch": 1.35, + "learning_rate": 2.7496794049756346e-05, + "loss": 0.3307, + "step": 35146 + }, + { + "epoch": 1.35, + "learning_rate": 2.7496152859707618e-05, + "loss": 0.2589, + "step": 35147 + }, + { + "epoch": 1.35, + "learning_rate": 2.7495511669658886e-05, + "loss": 0.3304, + "step": 35148 + }, + { + "epoch": 1.35, + "learning_rate": 2.7494870479610158e-05, + "loss": 0.3157, + "step": 35149 + }, + { + "epoch": 1.35, + "learning_rate": 2.749422928956143e-05, + "loss": 0.3496, + "step": 35150 + }, + { + "epoch": 1.35, + "learning_rate": 2.7493588099512694e-05, + "loss": 0.4143, + "step": 35151 + }, + { + "epoch": 1.35, + "learning_rate": 2.749294690946397e-05, + "loss": 0.5508, + "step": 35152 + }, + { + "epoch": 1.35, + "learning_rate": 2.7492305719415234e-05, + "loss": 0.3699, + "step": 35153 + }, + { + "epoch": 1.35, + "learning_rate": 2.7491664529366506e-05, + "loss": 0.3403, + "step": 35154 + }, + { + "epoch": 1.35, + "learning_rate": 2.7491023339317775e-05, + "loss": 0.3309, + "step": 35155 + }, + { + "epoch": 1.35, + "learning_rate": 2.7490382149269046e-05, + "loss": 0.2903, + "step": 35156 + }, + { + "epoch": 1.35, + "learning_rate": 2.748974095922031e-05, + "loss": 0.3944, + "step": 35157 + }, + { + "epoch": 1.35, + "learning_rate": 2.7489099769171583e-05, + "loss": 0.3247, + "step": 35158 + }, + { + "epoch": 1.35, + "learning_rate": 2.7488458579122855e-05, + "loss": 0.4047, + "step": 35159 + }, + { + "epoch": 1.35, + "learning_rate": 2.7487817389074123e-05, + "loss": 0.3181, + "step": 35160 + }, + { + "epoch": 1.35, + "learning_rate": 2.7487176199025395e-05, + "loss": 0.3359, + "step": 35161 + }, + { + "epoch": 1.35, + "learning_rate": 2.748653500897666e-05, + "loss": 0.4158, + "step": 35162 + }, + { + "epoch": 1.35, + "learning_rate": 2.748589381892793e-05, + "loss": 0.39, + "step": 35163 + }, + { + "epoch": 1.35, + "learning_rate": 2.74852526288792e-05, + "loss": 0.3353, + "step": 35164 + }, + { + "epoch": 1.35, + "learning_rate": 2.748461143883047e-05, + "loss": 0.4277, + "step": 35165 + }, + { + "epoch": 1.35, + "learning_rate": 2.7483970248781743e-05, + "loss": 0.4554, + "step": 35166 + }, + { + "epoch": 1.35, + "learning_rate": 2.7483329058733008e-05, + "loss": 0.4169, + "step": 35167 + }, + { + "epoch": 1.35, + "learning_rate": 2.7482687868684283e-05, + "loss": 0.3104, + "step": 35168 + }, + { + "epoch": 1.35, + "learning_rate": 2.7482046678635548e-05, + "loss": 0.4387, + "step": 35169 + }, + { + "epoch": 1.35, + "learning_rate": 2.748140548858682e-05, + "loss": 0.3115, + "step": 35170 + }, + { + "epoch": 1.35, + "learning_rate": 2.7480764298538088e-05, + "loss": 0.268, + "step": 35171 + }, + { + "epoch": 1.35, + "learning_rate": 2.748012310848936e-05, + "loss": 0.3892, + "step": 35172 + }, + { + "epoch": 1.35, + "learning_rate": 2.7479481918440625e-05, + "loss": 0.4363, + "step": 35173 + }, + { + "epoch": 1.35, + "learning_rate": 2.7478840728391896e-05, + "loss": 0.4311, + "step": 35174 + }, + { + "epoch": 1.35, + "learning_rate": 2.7478199538343168e-05, + "loss": 0.3546, + "step": 35175 + }, + { + "epoch": 1.35, + "learning_rate": 2.7477558348294437e-05, + "loss": 0.4064, + "step": 35176 + }, + { + "epoch": 1.35, + "learning_rate": 2.7476917158245708e-05, + "loss": 0.3697, + "step": 35177 + }, + { + "epoch": 1.35, + "learning_rate": 2.7476275968196973e-05, + "loss": 0.3356, + "step": 35178 + }, + { + "epoch": 1.35, + "learning_rate": 2.7475634778148245e-05, + "loss": 0.3898, + "step": 35179 + }, + { + "epoch": 1.35, + "learning_rate": 2.7474993588099513e-05, + "loss": 0.3317, + "step": 35180 + }, + { + "epoch": 1.35, + "learning_rate": 2.7474352398050785e-05, + "loss": 0.447, + "step": 35181 + }, + { + "epoch": 1.35, + "learning_rate": 2.747371120800205e-05, + "loss": 0.4482, + "step": 35182 + }, + { + "epoch": 1.35, + "learning_rate": 2.747307001795332e-05, + "loss": 0.4811, + "step": 35183 + }, + { + "epoch": 1.35, + "learning_rate": 2.7472428827904593e-05, + "loss": 0.4007, + "step": 35184 + }, + { + "epoch": 1.35, + "learning_rate": 2.747178763785586e-05, + "loss": 0.4293, + "step": 35185 + }, + { + "epoch": 1.35, + "learning_rate": 2.7471146447807133e-05, + "loss": 0.3999, + "step": 35186 + }, + { + "epoch": 1.35, + "learning_rate": 2.7470505257758398e-05, + "loss": 0.3463, + "step": 35187 + }, + { + "epoch": 1.35, + "learning_rate": 2.7469864067709673e-05, + "loss": 0.3157, + "step": 35188 + }, + { + "epoch": 1.35, + "learning_rate": 2.746922287766094e-05, + "loss": 0.2627, + "step": 35189 + }, + { + "epoch": 1.35, + "learning_rate": 2.746858168761221e-05, + "loss": 0.4925, + "step": 35190 + }, + { + "epoch": 1.35, + "learning_rate": 2.746794049756348e-05, + "loss": 0.4067, + "step": 35191 + }, + { + "epoch": 1.35, + "learning_rate": 2.746729930751475e-05, + "loss": 0.3891, + "step": 35192 + }, + { + "epoch": 1.35, + "learning_rate": 2.7466658117466022e-05, + "loss": 0.2995, + "step": 35193 + }, + { + "epoch": 1.35, + "learning_rate": 2.7466016927417287e-05, + "loss": 0.3779, + "step": 35194 + }, + { + "epoch": 1.35, + "learning_rate": 2.746537573736856e-05, + "loss": 0.2866, + "step": 35195 + }, + { + "epoch": 1.35, + "learning_rate": 2.7464734547319827e-05, + "loss": 0.3558, + "step": 35196 + }, + { + "epoch": 1.35, + "learning_rate": 2.74640933572711e-05, + "loss": 0.4888, + "step": 35197 + }, + { + "epoch": 1.35, + "learning_rate": 2.7463452167222363e-05, + "loss": 0.3515, + "step": 35198 + }, + { + "epoch": 1.35, + "learning_rate": 2.7462810977173635e-05, + "loss": 0.3926, + "step": 35199 + }, + { + "epoch": 1.35, + "learning_rate": 2.7462169787124904e-05, + "loss": 0.3485, + "step": 35200 + }, + { + "epoch": 1.35, + "learning_rate": 2.7461528597076175e-05, + "loss": 0.2878, + "step": 35201 + }, + { + "epoch": 1.35, + "learning_rate": 2.7460887407027447e-05, + "loss": 0.4746, + "step": 35202 + }, + { + "epoch": 1.35, + "learning_rate": 2.7460246216978712e-05, + "loss": 0.4286, + "step": 35203 + }, + { + "epoch": 1.35, + "learning_rate": 2.7459605026929987e-05, + "loss": 0.4252, + "step": 35204 + }, + { + "epoch": 1.35, + "learning_rate": 2.7458963836881252e-05, + "loss": 0.3584, + "step": 35205 + }, + { + "epoch": 1.35, + "learning_rate": 2.7458322646832524e-05, + "loss": 0.5067, + "step": 35206 + }, + { + "epoch": 1.35, + "learning_rate": 2.7457681456783792e-05, + "loss": 0.3208, + "step": 35207 + }, + { + "epoch": 1.35, + "learning_rate": 2.7457040266735064e-05, + "loss": 0.4357, + "step": 35208 + }, + { + "epoch": 1.35, + "learning_rate": 2.745639907668633e-05, + "loss": 0.2873, + "step": 35209 + }, + { + "epoch": 1.35, + "learning_rate": 2.74557578866376e-05, + "loss": 0.454, + "step": 35210 + }, + { + "epoch": 1.35, + "learning_rate": 2.7455116696588872e-05, + "loss": 0.5275, + "step": 35211 + }, + { + "epoch": 1.35, + "learning_rate": 2.745447550654014e-05, + "loss": 0.2846, + "step": 35212 + }, + { + "epoch": 1.35, + "learning_rate": 2.7453834316491412e-05, + "loss": 0.4324, + "step": 35213 + }, + { + "epoch": 1.35, + "learning_rate": 2.7453193126442677e-05, + "loss": 0.4981, + "step": 35214 + }, + { + "epoch": 1.35, + "learning_rate": 2.745255193639395e-05, + "loss": 0.3645, + "step": 35215 + }, + { + "epoch": 1.35, + "learning_rate": 2.7451910746345217e-05, + "loss": 0.3212, + "step": 35216 + }, + { + "epoch": 1.35, + "learning_rate": 2.745126955629649e-05, + "loss": 0.3432, + "step": 35217 + }, + { + "epoch": 1.35, + "learning_rate": 2.7450628366247754e-05, + "loss": 0.3795, + "step": 35218 + }, + { + "epoch": 1.35, + "learning_rate": 2.7449987176199025e-05, + "loss": 0.4862, + "step": 35219 + }, + { + "epoch": 1.35, + "learning_rate": 2.7449345986150297e-05, + "loss": 0.3708, + "step": 35220 + }, + { + "epoch": 1.35, + "learning_rate": 2.7448704796101566e-05, + "loss": 0.3777, + "step": 35221 + }, + { + "epoch": 1.35, + "learning_rate": 2.7448063606052837e-05, + "loss": 0.3572, + "step": 35222 + }, + { + "epoch": 1.35, + "learning_rate": 2.7447422416004102e-05, + "loss": 0.317, + "step": 35223 + }, + { + "epoch": 1.35, + "learning_rate": 2.7446781225955377e-05, + "loss": 0.2842, + "step": 35224 + }, + { + "epoch": 1.35, + "learning_rate": 2.7446140035906642e-05, + "loss": 0.384, + "step": 35225 + }, + { + "epoch": 1.35, + "learning_rate": 2.7445498845857914e-05, + "loss": 0.2445, + "step": 35226 + }, + { + "epoch": 1.35, + "learning_rate": 2.7444857655809182e-05, + "loss": 0.3537, + "step": 35227 + }, + { + "epoch": 1.35, + "learning_rate": 2.7444216465760454e-05, + "loss": 0.401, + "step": 35228 + }, + { + "epoch": 1.35, + "learning_rate": 2.7443575275711726e-05, + "loss": 0.5027, + "step": 35229 + }, + { + "epoch": 1.35, + "learning_rate": 2.744293408566299e-05, + "loss": 0.4353, + "step": 35230 + }, + { + "epoch": 1.35, + "learning_rate": 2.7442292895614262e-05, + "loss": 0.3591, + "step": 35231 + }, + { + "epoch": 1.35, + "learning_rate": 2.744165170556553e-05, + "loss": 0.3259, + "step": 35232 + }, + { + "epoch": 1.35, + "learning_rate": 2.7441010515516802e-05, + "loss": 0.393, + "step": 35233 + }, + { + "epoch": 1.35, + "learning_rate": 2.7440369325468067e-05, + "loss": 0.3431, + "step": 35234 + }, + { + "epoch": 1.35, + "learning_rate": 2.743972813541934e-05, + "loss": 0.3839, + "step": 35235 + }, + { + "epoch": 1.35, + "learning_rate": 2.7439086945370607e-05, + "loss": 0.3143, + "step": 35236 + }, + { + "epoch": 1.35, + "learning_rate": 2.743844575532188e-05, + "loss": 0.2533, + "step": 35237 + }, + { + "epoch": 1.35, + "learning_rate": 2.743780456527315e-05, + "loss": 0.4567, + "step": 35238 + }, + { + "epoch": 1.35, + "learning_rate": 2.7437163375224416e-05, + "loss": 0.3567, + "step": 35239 + }, + { + "epoch": 1.35, + "learning_rate": 2.743652218517569e-05, + "loss": 0.2654, + "step": 35240 + }, + { + "epoch": 1.35, + "learning_rate": 2.7435880995126956e-05, + "loss": 0.4261, + "step": 35241 + }, + { + "epoch": 1.35, + "learning_rate": 2.7435239805078228e-05, + "loss": 0.3235, + "step": 35242 + }, + { + "epoch": 1.35, + "learning_rate": 2.7434598615029496e-05, + "loss": 0.353, + "step": 35243 + }, + { + "epoch": 1.36, + "learning_rate": 2.7433957424980768e-05, + "loss": 0.3641, + "step": 35244 + }, + { + "epoch": 1.36, + "learning_rate": 2.7433316234932033e-05, + "loss": 0.5051, + "step": 35245 + }, + { + "epoch": 1.36, + "learning_rate": 2.7432675044883304e-05, + "loss": 0.3625, + "step": 35246 + }, + { + "epoch": 1.36, + "learning_rate": 2.7432033854834576e-05, + "loss": 0.3592, + "step": 35247 + }, + { + "epoch": 1.36, + "learning_rate": 2.7431392664785844e-05, + "loss": 0.3317, + "step": 35248 + }, + { + "epoch": 1.36, + "learning_rate": 2.7430751474737116e-05, + "loss": 0.31, + "step": 35249 + }, + { + "epoch": 1.36, + "learning_rate": 2.743011028468838e-05, + "loss": 0.3414, + "step": 35250 + }, + { + "epoch": 1.36, + "learning_rate": 2.7429469094639653e-05, + "loss": 0.3897, + "step": 35251 + }, + { + "epoch": 1.36, + "learning_rate": 2.742882790459092e-05, + "loss": 0.3234, + "step": 35252 + }, + { + "epoch": 1.36, + "learning_rate": 2.7428186714542193e-05, + "loss": 0.2772, + "step": 35253 + }, + { + "epoch": 1.36, + "learning_rate": 2.7427545524493458e-05, + "loss": 0.3812, + "step": 35254 + }, + { + "epoch": 1.36, + "learning_rate": 2.742690433444473e-05, + "loss": 0.4621, + "step": 35255 + }, + { + "epoch": 1.36, + "learning_rate": 2.7426263144396e-05, + "loss": 0.434, + "step": 35256 + }, + { + "epoch": 1.36, + "learning_rate": 2.742562195434727e-05, + "loss": 0.4834, + "step": 35257 + }, + { + "epoch": 1.36, + "learning_rate": 2.742498076429854e-05, + "loss": 0.3678, + "step": 35258 + }, + { + "epoch": 1.36, + "learning_rate": 2.7424339574249806e-05, + "loss": 0.4226, + "step": 35259 + }, + { + "epoch": 1.36, + "learning_rate": 2.742369838420108e-05, + "loss": 0.3086, + "step": 35260 + }, + { + "epoch": 1.36, + "learning_rate": 2.7423057194152346e-05, + "loss": 0.4492, + "step": 35261 + }, + { + "epoch": 1.36, + "learning_rate": 2.7422416004103618e-05, + "loss": 0.4035, + "step": 35262 + }, + { + "epoch": 1.36, + "learning_rate": 2.7421774814054886e-05, + "loss": 0.2944, + "step": 35263 + }, + { + "epoch": 1.36, + "learning_rate": 2.7421133624006158e-05, + "loss": 0.2372, + "step": 35264 + }, + { + "epoch": 1.36, + "learning_rate": 2.742049243395743e-05, + "loss": 0.3054, + "step": 35265 + }, + { + "epoch": 1.36, + "learning_rate": 2.7419851243908695e-05, + "loss": 0.3681, + "step": 35266 + }, + { + "epoch": 1.36, + "learning_rate": 2.7419210053859966e-05, + "loss": 0.4999, + "step": 35267 + }, + { + "epoch": 1.36, + "learning_rate": 2.7418568863811235e-05, + "loss": 0.4101, + "step": 35268 + }, + { + "epoch": 1.36, + "learning_rate": 2.7417927673762506e-05, + "loss": 0.4385, + "step": 35269 + }, + { + "epoch": 1.36, + "learning_rate": 2.741728648371377e-05, + "loss": 0.4561, + "step": 35270 + }, + { + "epoch": 1.36, + "learning_rate": 2.7416645293665043e-05, + "loss": 0.3642, + "step": 35271 + }, + { + "epoch": 1.36, + "learning_rate": 2.741600410361631e-05, + "loss": 0.4408, + "step": 35272 + }, + { + "epoch": 1.36, + "learning_rate": 2.7415362913567583e-05, + "loss": 0.2846, + "step": 35273 + }, + { + "epoch": 1.36, + "learning_rate": 2.7414721723518855e-05, + "loss": 0.4204, + "step": 35274 + }, + { + "epoch": 1.36, + "learning_rate": 2.741408053347012e-05, + "loss": 0.3895, + "step": 35275 + }, + { + "epoch": 1.36, + "learning_rate": 2.7413439343421395e-05, + "loss": 0.3331, + "step": 35276 + }, + { + "epoch": 1.36, + "learning_rate": 2.741279815337266e-05, + "loss": 0.2772, + "step": 35277 + }, + { + "epoch": 1.36, + "learning_rate": 2.741215696332393e-05, + "loss": 0.3157, + "step": 35278 + }, + { + "epoch": 1.36, + "learning_rate": 2.74115157732752e-05, + "loss": 0.404, + "step": 35279 + }, + { + "epoch": 1.36, + "learning_rate": 2.741087458322647e-05, + "loss": 0.5563, + "step": 35280 + }, + { + "epoch": 1.36, + "learning_rate": 2.7410233393177736e-05, + "loss": 0.4346, + "step": 35281 + }, + { + "epoch": 1.36, + "learning_rate": 2.7409592203129008e-05, + "loss": 0.3647, + "step": 35282 + }, + { + "epoch": 1.36, + "learning_rate": 2.740895101308028e-05, + "loss": 0.3805, + "step": 35283 + }, + { + "epoch": 1.36, + "learning_rate": 2.7408309823031548e-05, + "loss": 0.458, + "step": 35284 + }, + { + "epoch": 1.36, + "learning_rate": 2.740766863298282e-05, + "loss": 0.3892, + "step": 35285 + }, + { + "epoch": 1.36, + "learning_rate": 2.7407027442934085e-05, + "loss": 0.3509, + "step": 35286 + }, + { + "epoch": 1.36, + "learning_rate": 2.7406386252885357e-05, + "loss": 0.2856, + "step": 35287 + }, + { + "epoch": 1.36, + "learning_rate": 2.7405745062836625e-05, + "loss": 0.2997, + "step": 35288 + }, + { + "epoch": 1.36, + "learning_rate": 2.7405103872787897e-05, + "loss": 0.2754, + "step": 35289 + }, + { + "epoch": 1.36, + "learning_rate": 2.740446268273916e-05, + "loss": 0.3124, + "step": 35290 + }, + { + "epoch": 1.36, + "learning_rate": 2.7403821492690433e-05, + "loss": 0.2643, + "step": 35291 + }, + { + "epoch": 1.36, + "learning_rate": 2.740318030264171e-05, + "loss": 0.2432, + "step": 35292 + }, + { + "epoch": 1.36, + "learning_rate": 2.7402539112592973e-05, + "loss": 0.3176, + "step": 35293 + }, + { + "epoch": 1.36, + "learning_rate": 2.7401897922544245e-05, + "loss": 0.2877, + "step": 35294 + }, + { + "epoch": 1.36, + "learning_rate": 2.7401256732495513e-05, + "loss": 0.3108, + "step": 35295 + }, + { + "epoch": 1.36, + "learning_rate": 2.7400615542446785e-05, + "loss": 0.3483, + "step": 35296 + }, + { + "epoch": 1.36, + "learning_rate": 2.739997435239805e-05, + "loss": 0.5281, + "step": 35297 + }, + { + "epoch": 1.36, + "learning_rate": 2.739933316234932e-05, + "loss": 0.3287, + "step": 35298 + }, + { + "epoch": 1.36, + "learning_rate": 2.739869197230059e-05, + "loss": 0.3147, + "step": 35299 + }, + { + "epoch": 1.36, + "learning_rate": 2.7398050782251862e-05, + "loss": 0.4976, + "step": 35300 + }, + { + "epoch": 1.36, + "learning_rate": 2.7397409592203133e-05, + "loss": 0.4138, + "step": 35301 + }, + { + "epoch": 1.36, + "learning_rate": 2.73967684021544e-05, + "loss": 0.309, + "step": 35302 + }, + { + "epoch": 1.36, + "learning_rate": 2.739612721210567e-05, + "loss": 0.3188, + "step": 35303 + }, + { + "epoch": 1.36, + "learning_rate": 2.739548602205694e-05, + "loss": 0.2727, + "step": 35304 + }, + { + "epoch": 1.36, + "learning_rate": 2.739484483200821e-05, + "loss": 0.4014, + "step": 35305 + }, + { + "epoch": 1.36, + "learning_rate": 2.7394203641959475e-05, + "loss": 0.3715, + "step": 35306 + }, + { + "epoch": 1.36, + "learning_rate": 2.7393562451910747e-05, + "loss": 0.4507, + "step": 35307 + }, + { + "epoch": 1.36, + "learning_rate": 2.7392921261862015e-05, + "loss": 0.4066, + "step": 35308 + }, + { + "epoch": 1.36, + "learning_rate": 2.7392280071813287e-05, + "loss": 0.4255, + "step": 35309 + }, + { + "epoch": 1.36, + "learning_rate": 2.739163888176456e-05, + "loss": 0.2533, + "step": 35310 + }, + { + "epoch": 1.36, + "learning_rate": 2.7390997691715823e-05, + "loss": 0.5163, + "step": 35311 + }, + { + "epoch": 1.36, + "learning_rate": 2.73903565016671e-05, + "loss": 0.4207, + "step": 35312 + }, + { + "epoch": 1.36, + "learning_rate": 2.7389715311618364e-05, + "loss": 0.259, + "step": 35313 + }, + { + "epoch": 1.36, + "learning_rate": 2.7389074121569635e-05, + "loss": 0.4232, + "step": 35314 + }, + { + "epoch": 1.36, + "learning_rate": 2.7388432931520904e-05, + "loss": 0.327, + "step": 35315 + }, + { + "epoch": 1.36, + "learning_rate": 2.7387791741472175e-05, + "loss": 0.5577, + "step": 35316 + }, + { + "epoch": 1.36, + "learning_rate": 2.738715055142344e-05, + "loss": 0.4135, + "step": 35317 + }, + { + "epoch": 1.36, + "learning_rate": 2.7386509361374712e-05, + "loss": 0.3978, + "step": 35318 + }, + { + "epoch": 1.36, + "learning_rate": 2.7385868171325984e-05, + "loss": 0.2874, + "step": 35319 + }, + { + "epoch": 1.36, + "learning_rate": 2.7385226981277252e-05, + "loss": 0.2625, + "step": 35320 + }, + { + "epoch": 1.36, + "learning_rate": 2.7384585791228524e-05, + "loss": 0.3675, + "step": 35321 + }, + { + "epoch": 1.36, + "learning_rate": 2.738394460117979e-05, + "loss": 0.3215, + "step": 35322 + }, + { + "epoch": 1.36, + "learning_rate": 2.738330341113106e-05, + "loss": 0.3963, + "step": 35323 + }, + { + "epoch": 1.36, + "learning_rate": 2.738266222108233e-05, + "loss": 0.3056, + "step": 35324 + }, + { + "epoch": 1.36, + "learning_rate": 2.73820210310336e-05, + "loss": 0.375, + "step": 35325 + }, + { + "epoch": 1.36, + "learning_rate": 2.7381379840984865e-05, + "loss": 0.3729, + "step": 35326 + }, + { + "epoch": 1.36, + "learning_rate": 2.7380738650936137e-05, + "loss": 0.4606, + "step": 35327 + }, + { + "epoch": 1.36, + "learning_rate": 2.7380097460887412e-05, + "loss": 0.3503, + "step": 35328 + }, + { + "epoch": 1.36, + "learning_rate": 2.7379456270838677e-05, + "loss": 0.4279, + "step": 35329 + }, + { + "epoch": 1.36, + "learning_rate": 2.737881508078995e-05, + "loss": 0.5422, + "step": 35330 + }, + { + "epoch": 1.36, + "learning_rate": 2.7378173890741217e-05, + "loss": 0.4756, + "step": 35331 + }, + { + "epoch": 1.36, + "learning_rate": 2.737753270069249e-05, + "loss": 0.3315, + "step": 35332 + }, + { + "epoch": 1.36, + "learning_rate": 2.7376891510643754e-05, + "loss": 0.4288, + "step": 35333 + }, + { + "epoch": 1.36, + "learning_rate": 2.7376250320595026e-05, + "loss": 0.347, + "step": 35334 + }, + { + "epoch": 1.36, + "learning_rate": 2.7375609130546294e-05, + "loss": 0.3224, + "step": 35335 + }, + { + "epoch": 1.36, + "learning_rate": 2.7374967940497566e-05, + "loss": 0.3489, + "step": 35336 + }, + { + "epoch": 1.36, + "learning_rate": 2.7374326750448837e-05, + "loss": 0.5998, + "step": 35337 + }, + { + "epoch": 1.36, + "learning_rate": 2.7373685560400102e-05, + "loss": 0.3496, + "step": 35338 + }, + { + "epoch": 1.36, + "learning_rate": 2.7373044370351374e-05, + "loss": 0.3893, + "step": 35339 + }, + { + "epoch": 1.36, + "learning_rate": 2.7372403180302642e-05, + "loss": 0.243, + "step": 35340 + }, + { + "epoch": 1.36, + "learning_rate": 2.7371761990253914e-05, + "loss": 0.2698, + "step": 35341 + }, + { + "epoch": 1.36, + "learning_rate": 2.737112080020518e-05, + "loss": 0.3325, + "step": 35342 + }, + { + "epoch": 1.36, + "learning_rate": 2.737047961015645e-05, + "loss": 0.3287, + "step": 35343 + }, + { + "epoch": 1.36, + "learning_rate": 2.736983842010772e-05, + "loss": 0.5237, + "step": 35344 + }, + { + "epoch": 1.36, + "learning_rate": 2.736919723005899e-05, + "loss": 0.4054, + "step": 35345 + }, + { + "epoch": 1.36, + "learning_rate": 2.7368556040010262e-05, + "loss": 0.4359, + "step": 35346 + }, + { + "epoch": 1.36, + "learning_rate": 2.7367914849961527e-05, + "loss": 0.3529, + "step": 35347 + }, + { + "epoch": 1.36, + "learning_rate": 2.7367273659912802e-05, + "loss": 0.2918, + "step": 35348 + }, + { + "epoch": 1.36, + "learning_rate": 2.7366632469864067e-05, + "loss": 0.3961, + "step": 35349 + }, + { + "epoch": 1.36, + "learning_rate": 2.736599127981534e-05, + "loss": 0.2751, + "step": 35350 + }, + { + "epoch": 1.36, + "learning_rate": 2.7365350089766607e-05, + "loss": 0.3356, + "step": 35351 + }, + { + "epoch": 1.36, + "learning_rate": 2.736470889971788e-05, + "loss": 0.4653, + "step": 35352 + }, + { + "epoch": 1.36, + "learning_rate": 2.736406770966915e-05, + "loss": 0.5268, + "step": 35353 + }, + { + "epoch": 1.36, + "learning_rate": 2.7363426519620416e-05, + "loss": 0.3497, + "step": 35354 + }, + { + "epoch": 1.36, + "learning_rate": 2.7362785329571688e-05, + "loss": 0.3724, + "step": 35355 + }, + { + "epoch": 1.36, + "learning_rate": 2.7362144139522956e-05, + "loss": 0.3786, + "step": 35356 + }, + { + "epoch": 1.36, + "learning_rate": 2.7361502949474228e-05, + "loss": 0.438, + "step": 35357 + }, + { + "epoch": 1.36, + "learning_rate": 2.7360861759425493e-05, + "loss": 0.4612, + "step": 35358 + }, + { + "epoch": 1.36, + "learning_rate": 2.7360220569376764e-05, + "loss": 0.3259, + "step": 35359 + }, + { + "epoch": 1.36, + "learning_rate": 2.7359579379328033e-05, + "loss": 0.2386, + "step": 35360 + }, + { + "epoch": 1.36, + "learning_rate": 2.7358938189279304e-05, + "loss": 0.4579, + "step": 35361 + }, + { + "epoch": 1.36, + "learning_rate": 2.7358296999230576e-05, + "loss": 0.4227, + "step": 35362 + }, + { + "epoch": 1.36, + "learning_rate": 2.735765580918184e-05, + "loss": 0.3379, + "step": 35363 + }, + { + "epoch": 1.36, + "learning_rate": 2.7357014619133116e-05, + "loss": 0.3838, + "step": 35364 + }, + { + "epoch": 1.36, + "learning_rate": 2.735637342908438e-05, + "loss": 0.4245, + "step": 35365 + }, + { + "epoch": 1.36, + "learning_rate": 2.7355732239035653e-05, + "loss": 0.3102, + "step": 35366 + }, + { + "epoch": 1.36, + "learning_rate": 2.735509104898692e-05, + "loss": 0.4265, + "step": 35367 + }, + { + "epoch": 1.36, + "learning_rate": 2.7354449858938193e-05, + "loss": 0.3702, + "step": 35368 + }, + { + "epoch": 1.36, + "learning_rate": 2.7353808668889458e-05, + "loss": 0.336, + "step": 35369 + }, + { + "epoch": 1.36, + "learning_rate": 2.735316747884073e-05, + "loss": 0.3234, + "step": 35370 + }, + { + "epoch": 1.36, + "learning_rate": 2.7352526288792e-05, + "loss": 0.2946, + "step": 35371 + }, + { + "epoch": 1.36, + "learning_rate": 2.735188509874327e-05, + "loss": 0.3033, + "step": 35372 + }, + { + "epoch": 1.36, + "learning_rate": 2.735124390869454e-05, + "loss": 0.3266, + "step": 35373 + }, + { + "epoch": 1.36, + "learning_rate": 2.7350602718645806e-05, + "loss": 0.2802, + "step": 35374 + }, + { + "epoch": 1.36, + "learning_rate": 2.7349961528597078e-05, + "loss": 0.3443, + "step": 35375 + }, + { + "epoch": 1.36, + "learning_rate": 2.7349320338548346e-05, + "loss": 0.3287, + "step": 35376 + }, + { + "epoch": 1.36, + "learning_rate": 2.7348679148499618e-05, + "loss": 0.3322, + "step": 35377 + }, + { + "epoch": 1.36, + "learning_rate": 2.7348037958450883e-05, + "loss": 0.4367, + "step": 35378 + }, + { + "epoch": 1.36, + "learning_rate": 2.7347396768402155e-05, + "loss": 0.4269, + "step": 35379 + }, + { + "epoch": 1.36, + "learning_rate": 2.7346755578353426e-05, + "loss": 0.2937, + "step": 35380 + }, + { + "epoch": 1.36, + "learning_rate": 2.7346114388304695e-05, + "loss": 0.3359, + "step": 35381 + }, + { + "epoch": 1.36, + "learning_rate": 2.7345473198255966e-05, + "loss": 0.3494, + "step": 35382 + }, + { + "epoch": 1.36, + "learning_rate": 2.734483200820723e-05, + "loss": 0.5438, + "step": 35383 + }, + { + "epoch": 1.36, + "learning_rate": 2.7344190818158506e-05, + "loss": 0.3332, + "step": 35384 + }, + { + "epoch": 1.36, + "learning_rate": 2.734354962810977e-05, + "loss": 0.3992, + "step": 35385 + }, + { + "epoch": 1.36, + "learning_rate": 2.7342908438061043e-05, + "loss": 0.4544, + "step": 35386 + }, + { + "epoch": 1.36, + "learning_rate": 2.734226724801231e-05, + "loss": 0.3486, + "step": 35387 + }, + { + "epoch": 1.36, + "learning_rate": 2.7341626057963583e-05, + "loss": 0.3871, + "step": 35388 + }, + { + "epoch": 1.36, + "learning_rate": 2.7340984867914855e-05, + "loss": 0.3645, + "step": 35389 + }, + { + "epoch": 1.36, + "learning_rate": 2.734034367786612e-05, + "loss": 0.4154, + "step": 35390 + }, + { + "epoch": 1.36, + "learning_rate": 2.733970248781739e-05, + "loss": 0.3041, + "step": 35391 + }, + { + "epoch": 1.36, + "learning_rate": 2.733906129776866e-05, + "loss": 0.324, + "step": 35392 + }, + { + "epoch": 1.36, + "learning_rate": 2.733842010771993e-05, + "loss": 0.3342, + "step": 35393 + }, + { + "epoch": 1.36, + "learning_rate": 2.7337778917671196e-05, + "loss": 0.1903, + "step": 35394 + }, + { + "epoch": 1.36, + "learning_rate": 2.7337137727622468e-05, + "loss": 0.5609, + "step": 35395 + }, + { + "epoch": 1.36, + "learning_rate": 2.7336496537573736e-05, + "loss": 0.3979, + "step": 35396 + }, + { + "epoch": 1.36, + "learning_rate": 2.7335855347525008e-05, + "loss": 0.448, + "step": 35397 + }, + { + "epoch": 1.36, + "learning_rate": 2.733521415747628e-05, + "loss": 0.2869, + "step": 35398 + }, + { + "epoch": 1.36, + "learning_rate": 2.7334572967427545e-05, + "loss": 0.4522, + "step": 35399 + }, + { + "epoch": 1.36, + "learning_rate": 2.733393177737882e-05, + "loss": 0.3002, + "step": 35400 + }, + { + "epoch": 1.36, + "learning_rate": 2.7333290587330085e-05, + "loss": 0.4194, + "step": 35401 + }, + { + "epoch": 1.36, + "learning_rate": 2.7332649397281357e-05, + "loss": 0.4026, + "step": 35402 + }, + { + "epoch": 1.36, + "learning_rate": 2.7332008207232625e-05, + "loss": 0.5008, + "step": 35403 + }, + { + "epoch": 1.36, + "learning_rate": 2.7331367017183897e-05, + "loss": 0.3573, + "step": 35404 + }, + { + "epoch": 1.36, + "learning_rate": 2.733072582713516e-05, + "loss": 0.3317, + "step": 35405 + }, + { + "epoch": 1.36, + "learning_rate": 2.7330084637086433e-05, + "loss": 0.4606, + "step": 35406 + }, + { + "epoch": 1.36, + "learning_rate": 2.7329443447037705e-05, + "loss": 0.3735, + "step": 35407 + }, + { + "epoch": 1.36, + "learning_rate": 2.7328802256988973e-05, + "loss": 0.3846, + "step": 35408 + }, + { + "epoch": 1.36, + "learning_rate": 2.7328161066940245e-05, + "loss": 0.4306, + "step": 35409 + }, + { + "epoch": 1.36, + "learning_rate": 2.732751987689151e-05, + "loss": 0.2689, + "step": 35410 + }, + { + "epoch": 1.36, + "learning_rate": 2.7326878686842782e-05, + "loss": 0.4347, + "step": 35411 + }, + { + "epoch": 1.36, + "learning_rate": 2.732623749679405e-05, + "loss": 0.3873, + "step": 35412 + }, + { + "epoch": 1.36, + "learning_rate": 2.7325596306745322e-05, + "loss": 0.2485, + "step": 35413 + }, + { + "epoch": 1.36, + "learning_rate": 2.7324955116696587e-05, + "loss": 0.3345, + "step": 35414 + }, + { + "epoch": 1.36, + "learning_rate": 2.732431392664786e-05, + "loss": 0.2971, + "step": 35415 + }, + { + "epoch": 1.36, + "learning_rate": 2.7323672736599134e-05, + "loss": 0.4657, + "step": 35416 + }, + { + "epoch": 1.36, + "learning_rate": 2.73230315465504e-05, + "loss": 0.2881, + "step": 35417 + }, + { + "epoch": 1.36, + "learning_rate": 2.732239035650167e-05, + "loss": 0.3873, + "step": 35418 + }, + { + "epoch": 1.36, + "learning_rate": 2.732174916645294e-05, + "loss": 0.3622, + "step": 35419 + }, + { + "epoch": 1.36, + "learning_rate": 2.732110797640421e-05, + "loss": 0.5462, + "step": 35420 + }, + { + "epoch": 1.36, + "learning_rate": 2.7320466786355475e-05, + "loss": 0.401, + "step": 35421 + }, + { + "epoch": 1.36, + "learning_rate": 2.7319825596306747e-05, + "loss": 0.4248, + "step": 35422 + }, + { + "epoch": 1.36, + "learning_rate": 2.7319184406258015e-05, + "loss": 0.343, + "step": 35423 + }, + { + "epoch": 1.36, + "learning_rate": 2.7318543216209287e-05, + "loss": 0.4547, + "step": 35424 + }, + { + "epoch": 1.36, + "learning_rate": 2.731790202616056e-05, + "loss": 0.356, + "step": 35425 + }, + { + "epoch": 1.36, + "learning_rate": 2.7317260836111824e-05, + "loss": 0.4286, + "step": 35426 + }, + { + "epoch": 1.36, + "learning_rate": 2.7316619646063095e-05, + "loss": 0.4446, + "step": 35427 + }, + { + "epoch": 1.36, + "learning_rate": 2.7315978456014364e-05, + "loss": 0.3047, + "step": 35428 + }, + { + "epoch": 1.36, + "learning_rate": 2.7315337265965635e-05, + "loss": 0.222, + "step": 35429 + }, + { + "epoch": 1.36, + "learning_rate": 2.73146960759169e-05, + "loss": 0.3286, + "step": 35430 + }, + { + "epoch": 1.36, + "learning_rate": 2.7314054885868172e-05, + "loss": 0.3321, + "step": 35431 + }, + { + "epoch": 1.36, + "learning_rate": 2.731341369581944e-05, + "loss": 0.3098, + "step": 35432 + }, + { + "epoch": 1.36, + "learning_rate": 2.7312772505770712e-05, + "loss": 0.3657, + "step": 35433 + }, + { + "epoch": 1.36, + "learning_rate": 2.7312131315721984e-05, + "loss": 0.5437, + "step": 35434 + }, + { + "epoch": 1.36, + "learning_rate": 2.731149012567325e-05, + "loss": 0.3083, + "step": 35435 + }, + { + "epoch": 1.36, + "learning_rate": 2.7310848935624524e-05, + "loss": 0.3567, + "step": 35436 + }, + { + "epoch": 1.36, + "learning_rate": 2.731020774557579e-05, + "loss": 0.2837, + "step": 35437 + }, + { + "epoch": 1.36, + "learning_rate": 2.730956655552706e-05, + "loss": 0.4268, + "step": 35438 + }, + { + "epoch": 1.36, + "learning_rate": 2.730892536547833e-05, + "loss": 0.4028, + "step": 35439 + }, + { + "epoch": 1.36, + "learning_rate": 2.73082841754296e-05, + "loss": 0.4176, + "step": 35440 + }, + { + "epoch": 1.36, + "learning_rate": 2.7307642985380865e-05, + "loss": 0.3896, + "step": 35441 + }, + { + "epoch": 1.36, + "learning_rate": 2.7307001795332137e-05, + "loss": 0.4024, + "step": 35442 + }, + { + "epoch": 1.36, + "learning_rate": 2.730636060528341e-05, + "loss": 0.261, + "step": 35443 + }, + { + "epoch": 1.36, + "learning_rate": 2.7305719415234677e-05, + "loss": 0.3925, + "step": 35444 + }, + { + "epoch": 1.36, + "learning_rate": 2.730507822518595e-05, + "loss": 0.3864, + "step": 35445 + }, + { + "epoch": 1.36, + "learning_rate": 2.7304437035137214e-05, + "loss": 0.267, + "step": 35446 + }, + { + "epoch": 1.36, + "learning_rate": 2.7303795845088486e-05, + "loss": 0.3597, + "step": 35447 + }, + { + "epoch": 1.36, + "learning_rate": 2.7303154655039754e-05, + "loss": 0.4062, + "step": 35448 + }, + { + "epoch": 1.36, + "learning_rate": 2.7302513464991026e-05, + "loss": 0.4726, + "step": 35449 + }, + { + "epoch": 1.36, + "learning_rate": 2.730187227494229e-05, + "loss": 0.4126, + "step": 35450 + }, + { + "epoch": 1.36, + "learning_rate": 2.7301231084893562e-05, + "loss": 0.4361, + "step": 35451 + }, + { + "epoch": 1.36, + "learning_rate": 2.7300589894844837e-05, + "loss": 0.4333, + "step": 35452 + }, + { + "epoch": 1.36, + "learning_rate": 2.7299948704796102e-05, + "loss": 0.4283, + "step": 35453 + }, + { + "epoch": 1.36, + "learning_rate": 2.7299307514747374e-05, + "loss": 0.3551, + "step": 35454 + }, + { + "epoch": 1.36, + "learning_rate": 2.7298666324698642e-05, + "loss": 0.3761, + "step": 35455 + }, + { + "epoch": 1.36, + "learning_rate": 2.7298025134649914e-05, + "loss": 0.4376, + "step": 35456 + }, + { + "epoch": 1.36, + "learning_rate": 2.729738394460118e-05, + "loss": 0.2233, + "step": 35457 + }, + { + "epoch": 1.36, + "learning_rate": 2.729674275455245e-05, + "loss": 0.399, + "step": 35458 + }, + { + "epoch": 1.36, + "learning_rate": 2.729610156450372e-05, + "loss": 0.4288, + "step": 35459 + }, + { + "epoch": 1.36, + "learning_rate": 2.729546037445499e-05, + "loss": 0.4064, + "step": 35460 + }, + { + "epoch": 1.36, + "learning_rate": 2.7294819184406263e-05, + "loss": 0.315, + "step": 35461 + }, + { + "epoch": 1.36, + "learning_rate": 2.7294177994357527e-05, + "loss": 0.4474, + "step": 35462 + }, + { + "epoch": 1.36, + "learning_rate": 2.72935368043088e-05, + "loss": 0.3469, + "step": 35463 + }, + { + "epoch": 1.36, + "learning_rate": 2.7292895614260067e-05, + "loss": 0.3699, + "step": 35464 + }, + { + "epoch": 1.36, + "learning_rate": 2.729225442421134e-05, + "loss": 0.3202, + "step": 35465 + }, + { + "epoch": 1.36, + "learning_rate": 2.7291613234162604e-05, + "loss": 0.3973, + "step": 35466 + }, + { + "epoch": 1.36, + "learning_rate": 2.7290972044113876e-05, + "loss": 0.4246, + "step": 35467 + }, + { + "epoch": 1.36, + "learning_rate": 2.7290330854065144e-05, + "loss": 0.3191, + "step": 35468 + }, + { + "epoch": 1.36, + "learning_rate": 2.7289689664016416e-05, + "loss": 0.4243, + "step": 35469 + }, + { + "epoch": 1.36, + "learning_rate": 2.7289048473967688e-05, + "loss": 0.4757, + "step": 35470 + }, + { + "epoch": 1.36, + "learning_rate": 2.7288407283918953e-05, + "loss": 0.4282, + "step": 35471 + }, + { + "epoch": 1.36, + "learning_rate": 2.7287766093870228e-05, + "loss": 0.4255, + "step": 35472 + }, + { + "epoch": 1.36, + "learning_rate": 2.7287124903821493e-05, + "loss": 0.366, + "step": 35473 + }, + { + "epoch": 1.36, + "learning_rate": 2.7286483713772764e-05, + "loss": 0.509, + "step": 35474 + }, + { + "epoch": 1.36, + "learning_rate": 2.7285842523724033e-05, + "loss": 0.4169, + "step": 35475 + }, + { + "epoch": 1.36, + "learning_rate": 2.7285201333675304e-05, + "loss": 0.334, + "step": 35476 + }, + { + "epoch": 1.36, + "learning_rate": 2.728456014362657e-05, + "loss": 0.4414, + "step": 35477 + }, + { + "epoch": 1.36, + "learning_rate": 2.728391895357784e-05, + "loss": 0.4514, + "step": 35478 + }, + { + "epoch": 1.36, + "learning_rate": 2.7283277763529113e-05, + "loss": 0.3626, + "step": 35479 + }, + { + "epoch": 1.36, + "learning_rate": 2.728263657348038e-05, + "loss": 0.295, + "step": 35480 + }, + { + "epoch": 1.36, + "learning_rate": 2.7281995383431653e-05, + "loss": 0.392, + "step": 35481 + }, + { + "epoch": 1.36, + "learning_rate": 2.7281354193382918e-05, + "loss": 0.3668, + "step": 35482 + }, + { + "epoch": 1.36, + "learning_rate": 2.728071300333419e-05, + "loss": 0.3864, + "step": 35483 + }, + { + "epoch": 1.36, + "learning_rate": 2.7280071813285458e-05, + "loss": 0.3749, + "step": 35484 + }, + { + "epoch": 1.36, + "learning_rate": 2.727943062323673e-05, + "loss": 0.4559, + "step": 35485 + }, + { + "epoch": 1.36, + "learning_rate": 2.7278789433187994e-05, + "loss": 0.3198, + "step": 35486 + }, + { + "epoch": 1.36, + "learning_rate": 2.7278148243139266e-05, + "loss": 0.4417, + "step": 35487 + }, + { + "epoch": 1.36, + "learning_rate": 2.727750705309054e-05, + "loss": 0.3445, + "step": 35488 + }, + { + "epoch": 1.36, + "learning_rate": 2.7276865863041806e-05, + "loss": 0.2333, + "step": 35489 + }, + { + "epoch": 1.36, + "learning_rate": 2.7276224672993078e-05, + "loss": 0.4336, + "step": 35490 + }, + { + "epoch": 1.36, + "learning_rate": 2.7275583482944346e-05, + "loss": 0.336, + "step": 35491 + }, + { + "epoch": 1.36, + "learning_rate": 2.7274942292895618e-05, + "loss": 0.3331, + "step": 35492 + }, + { + "epoch": 1.36, + "learning_rate": 2.7274301102846883e-05, + "loss": 0.3935, + "step": 35493 + }, + { + "epoch": 1.36, + "learning_rate": 2.7273659912798155e-05, + "loss": 0.4699, + "step": 35494 + }, + { + "epoch": 1.36, + "learning_rate": 2.7273018722749423e-05, + "loss": 0.3354, + "step": 35495 + }, + { + "epoch": 1.36, + "learning_rate": 2.7272377532700695e-05, + "loss": 0.3433, + "step": 35496 + }, + { + "epoch": 1.36, + "learning_rate": 2.7271736342651966e-05, + "loss": 0.5759, + "step": 35497 + }, + { + "epoch": 1.36, + "learning_rate": 2.727109515260323e-05, + "loss": 0.5078, + "step": 35498 + }, + { + "epoch": 1.36, + "learning_rate": 2.7270453962554503e-05, + "loss": 0.3639, + "step": 35499 + }, + { + "epoch": 1.36, + "learning_rate": 2.726981277250577e-05, + "loss": 0.2932, + "step": 35500 + }, + { + "epoch": 1.36, + "learning_rate": 2.7269171582457043e-05, + "loss": 0.3183, + "step": 35501 + }, + { + "epoch": 1.36, + "learning_rate": 2.7268530392408308e-05, + "loss": 0.3547, + "step": 35502 + }, + { + "epoch": 1.36, + "learning_rate": 2.726788920235958e-05, + "loss": 0.2878, + "step": 35503 + }, + { + "epoch": 1.37, + "learning_rate": 2.7267248012310848e-05, + "loss": 0.3564, + "step": 35504 + }, + { + "epoch": 1.37, + "learning_rate": 2.726660682226212e-05, + "loss": 0.3262, + "step": 35505 + }, + { + "epoch": 1.37, + "learning_rate": 2.726596563221339e-05, + "loss": 0.3452, + "step": 35506 + }, + { + "epoch": 1.37, + "learning_rate": 2.7265324442164656e-05, + "loss": 0.3856, + "step": 35507 + }, + { + "epoch": 1.37, + "learning_rate": 2.726468325211593e-05, + "loss": 0.4696, + "step": 35508 + }, + { + "epoch": 1.37, + "learning_rate": 2.7264042062067196e-05, + "loss": 0.4428, + "step": 35509 + }, + { + "epoch": 1.37, + "learning_rate": 2.7263400872018468e-05, + "loss": 0.3539, + "step": 35510 + }, + { + "epoch": 1.37, + "learning_rate": 2.7262759681969737e-05, + "loss": 0.4916, + "step": 35511 + }, + { + "epoch": 1.37, + "learning_rate": 2.7262118491921008e-05, + "loss": 0.3733, + "step": 35512 + }, + { + "epoch": 1.37, + "learning_rate": 2.7261477301872273e-05, + "loss": 0.3895, + "step": 35513 + }, + { + "epoch": 1.37, + "learning_rate": 2.7260836111823545e-05, + "loss": 0.3633, + "step": 35514 + }, + { + "epoch": 1.37, + "learning_rate": 2.7260194921774817e-05, + "loss": 0.3002, + "step": 35515 + }, + { + "epoch": 1.37, + "learning_rate": 2.7259553731726085e-05, + "loss": 0.4403, + "step": 35516 + }, + { + "epoch": 1.37, + "learning_rate": 2.7258912541677357e-05, + "loss": 0.4156, + "step": 35517 + }, + { + "epoch": 1.37, + "learning_rate": 2.725827135162862e-05, + "loss": 0.619, + "step": 35518 + }, + { + "epoch": 1.37, + "learning_rate": 2.7257630161579893e-05, + "loss": 0.3747, + "step": 35519 + }, + { + "epoch": 1.37, + "learning_rate": 2.725698897153116e-05, + "loss": 0.3387, + "step": 35520 + }, + { + "epoch": 1.37, + "learning_rate": 2.7256347781482433e-05, + "loss": 0.3687, + "step": 35521 + }, + { + "epoch": 1.37, + "learning_rate": 2.72557065914337e-05, + "loss": 0.3681, + "step": 35522 + }, + { + "epoch": 1.37, + "learning_rate": 2.725506540138497e-05, + "loss": 0.2774, + "step": 35523 + }, + { + "epoch": 1.37, + "learning_rate": 2.7254424211336245e-05, + "loss": 0.3427, + "step": 35524 + }, + { + "epoch": 1.37, + "learning_rate": 2.725378302128751e-05, + "loss": 0.3807, + "step": 35525 + }, + { + "epoch": 1.37, + "learning_rate": 2.7253141831238782e-05, + "loss": 0.3498, + "step": 35526 + }, + { + "epoch": 1.37, + "learning_rate": 2.725250064119005e-05, + "loss": 0.513, + "step": 35527 + }, + { + "epoch": 1.37, + "learning_rate": 2.7251859451141322e-05, + "loss": 0.3135, + "step": 35528 + }, + { + "epoch": 1.37, + "learning_rate": 2.7251218261092587e-05, + "loss": 0.3474, + "step": 35529 + }, + { + "epoch": 1.37, + "learning_rate": 2.725057707104386e-05, + "loss": 0.2286, + "step": 35530 + }, + { + "epoch": 1.37, + "learning_rate": 2.7249935880995127e-05, + "loss": 0.3836, + "step": 35531 + }, + { + "epoch": 1.37, + "learning_rate": 2.72492946909464e-05, + "loss": 0.3814, + "step": 35532 + }, + { + "epoch": 1.37, + "learning_rate": 2.724865350089767e-05, + "loss": 0.4478, + "step": 35533 + }, + { + "epoch": 1.37, + "learning_rate": 2.7248012310848935e-05, + "loss": 0.4035, + "step": 35534 + }, + { + "epoch": 1.37, + "learning_rate": 2.7247371120800207e-05, + "loss": 0.3801, + "step": 35535 + }, + { + "epoch": 1.37, + "learning_rate": 2.7246729930751475e-05, + "loss": 0.4288, + "step": 35536 + }, + { + "epoch": 1.37, + "learning_rate": 2.7246088740702747e-05, + "loss": 0.4901, + "step": 35537 + }, + { + "epoch": 1.37, + "learning_rate": 2.7245447550654012e-05, + "loss": 0.4411, + "step": 35538 + }, + { + "epoch": 1.37, + "learning_rate": 2.7244806360605284e-05, + "loss": 0.405, + "step": 35539 + }, + { + "epoch": 1.37, + "learning_rate": 2.724416517055656e-05, + "loss": 0.3849, + "step": 35540 + }, + { + "epoch": 1.37, + "learning_rate": 2.7243523980507824e-05, + "loss": 0.2778, + "step": 35541 + }, + { + "epoch": 1.37, + "learning_rate": 2.7242882790459095e-05, + "loss": 0.3366, + "step": 35542 + }, + { + "epoch": 1.37, + "learning_rate": 2.724224160041036e-05, + "loss": 0.3846, + "step": 35543 + }, + { + "epoch": 1.37, + "learning_rate": 2.7241600410361635e-05, + "loss": 0.5005, + "step": 35544 + }, + { + "epoch": 1.37, + "learning_rate": 2.72409592203129e-05, + "loss": 0.294, + "step": 35545 + }, + { + "epoch": 1.37, + "learning_rate": 2.7240318030264172e-05, + "loss": 0.3703, + "step": 35546 + }, + { + "epoch": 1.37, + "learning_rate": 2.723967684021544e-05, + "loss": 0.4367, + "step": 35547 + }, + { + "epoch": 1.37, + "learning_rate": 2.7239035650166712e-05, + "loss": 0.5157, + "step": 35548 + }, + { + "epoch": 1.37, + "learning_rate": 2.7238394460117984e-05, + "loss": 0.346, + "step": 35549 + }, + { + "epoch": 1.37, + "learning_rate": 2.723775327006925e-05, + "loss": 0.4386, + "step": 35550 + }, + { + "epoch": 1.37, + "learning_rate": 2.723711208002052e-05, + "loss": 0.2626, + "step": 35551 + }, + { + "epoch": 1.37, + "learning_rate": 2.723647088997179e-05, + "loss": 0.3785, + "step": 35552 + }, + { + "epoch": 1.37, + "learning_rate": 2.723582969992306e-05, + "loss": 0.3282, + "step": 35553 + }, + { + "epoch": 1.37, + "learning_rate": 2.7235188509874325e-05, + "loss": 0.3492, + "step": 35554 + }, + { + "epoch": 1.37, + "learning_rate": 2.7234547319825597e-05, + "loss": 0.3303, + "step": 35555 + }, + { + "epoch": 1.37, + "learning_rate": 2.7233906129776866e-05, + "loss": 0.2937, + "step": 35556 + }, + { + "epoch": 1.37, + "learning_rate": 2.7233264939728137e-05, + "loss": 0.2996, + "step": 35557 + }, + { + "epoch": 1.37, + "learning_rate": 2.723262374967941e-05, + "loss": 0.3706, + "step": 35558 + }, + { + "epoch": 1.37, + "learning_rate": 2.7231982559630674e-05, + "loss": 0.4669, + "step": 35559 + }, + { + "epoch": 1.37, + "learning_rate": 2.723134136958195e-05, + "loss": 0.4755, + "step": 35560 + }, + { + "epoch": 1.37, + "learning_rate": 2.7230700179533214e-05, + "loss": 0.3204, + "step": 35561 + }, + { + "epoch": 1.37, + "learning_rate": 2.7230058989484486e-05, + "loss": 0.4271, + "step": 35562 + }, + { + "epoch": 1.37, + "learning_rate": 2.7229417799435754e-05, + "loss": 0.3456, + "step": 35563 + }, + { + "epoch": 1.37, + "learning_rate": 2.7228776609387026e-05, + "loss": 0.2703, + "step": 35564 + }, + { + "epoch": 1.37, + "learning_rate": 2.722813541933829e-05, + "loss": 0.3941, + "step": 35565 + }, + { + "epoch": 1.37, + "learning_rate": 2.7227494229289562e-05, + "loss": 0.3403, + "step": 35566 + }, + { + "epoch": 1.37, + "learning_rate": 2.7226853039240834e-05, + "loss": 0.2497, + "step": 35567 + }, + { + "epoch": 1.37, + "learning_rate": 2.7226211849192102e-05, + "loss": 0.304, + "step": 35568 + }, + { + "epoch": 1.37, + "learning_rate": 2.7225570659143374e-05, + "loss": 0.3261, + "step": 35569 + }, + { + "epoch": 1.37, + "learning_rate": 2.722492946909464e-05, + "loss": 0.3102, + "step": 35570 + }, + { + "epoch": 1.37, + "learning_rate": 2.722428827904591e-05, + "loss": 0.4395, + "step": 35571 + }, + { + "epoch": 1.37, + "learning_rate": 2.722364708899718e-05, + "loss": 0.344, + "step": 35572 + }, + { + "epoch": 1.37, + "learning_rate": 2.722300589894845e-05, + "loss": 0.373, + "step": 35573 + }, + { + "epoch": 1.37, + "learning_rate": 2.7222364708899716e-05, + "loss": 0.3078, + "step": 35574 + }, + { + "epoch": 1.37, + "learning_rate": 2.7221723518850987e-05, + "loss": 0.3037, + "step": 35575 + }, + { + "epoch": 1.37, + "learning_rate": 2.7221082328802263e-05, + "loss": 0.2654, + "step": 35576 + }, + { + "epoch": 1.37, + "learning_rate": 2.7220441138753528e-05, + "loss": 0.5508, + "step": 35577 + }, + { + "epoch": 1.37, + "learning_rate": 2.72197999487048e-05, + "loss": 0.4181, + "step": 35578 + }, + { + "epoch": 1.37, + "learning_rate": 2.7219158758656068e-05, + "loss": 0.4197, + "step": 35579 + }, + { + "epoch": 1.37, + "learning_rate": 2.721851756860734e-05, + "loss": 0.3626, + "step": 35580 + }, + { + "epoch": 1.37, + "learning_rate": 2.7217876378558604e-05, + "loss": 0.4128, + "step": 35581 + }, + { + "epoch": 1.37, + "learning_rate": 2.7217235188509876e-05, + "loss": 0.3148, + "step": 35582 + }, + { + "epoch": 1.37, + "learning_rate": 2.7216593998461144e-05, + "loss": 0.2449, + "step": 35583 + }, + { + "epoch": 1.37, + "learning_rate": 2.7215952808412416e-05, + "loss": 0.3441, + "step": 35584 + }, + { + "epoch": 1.37, + "learning_rate": 2.7215311618363688e-05, + "loss": 0.254, + "step": 35585 + }, + { + "epoch": 1.37, + "learning_rate": 2.7214670428314953e-05, + "loss": 0.4305, + "step": 35586 + }, + { + "epoch": 1.37, + "learning_rate": 2.7214029238266224e-05, + "loss": 0.4525, + "step": 35587 + }, + { + "epoch": 1.37, + "learning_rate": 2.7213388048217493e-05, + "loss": 0.4804, + "step": 35588 + }, + { + "epoch": 1.37, + "learning_rate": 2.7212746858168764e-05, + "loss": 0.3646, + "step": 35589 + }, + { + "epoch": 1.37, + "learning_rate": 2.721210566812003e-05, + "loss": 0.3762, + "step": 35590 + }, + { + "epoch": 1.37, + "learning_rate": 2.72114644780713e-05, + "loss": 0.2865, + "step": 35591 + }, + { + "epoch": 1.37, + "learning_rate": 2.721082328802257e-05, + "loss": 0.266, + "step": 35592 + }, + { + "epoch": 1.37, + "learning_rate": 2.721018209797384e-05, + "loss": 0.3387, + "step": 35593 + }, + { + "epoch": 1.37, + "learning_rate": 2.7209540907925113e-05, + "loss": 0.3195, + "step": 35594 + }, + { + "epoch": 1.37, + "learning_rate": 2.7208899717876378e-05, + "loss": 0.4085, + "step": 35595 + }, + { + "epoch": 1.37, + "learning_rate": 2.7208258527827653e-05, + "loss": 0.4389, + "step": 35596 + }, + { + "epoch": 1.37, + "learning_rate": 2.7207617337778918e-05, + "loss": 0.6845, + "step": 35597 + }, + { + "epoch": 1.37, + "learning_rate": 2.720697614773019e-05, + "loss": 0.294, + "step": 35598 + }, + { + "epoch": 1.37, + "learning_rate": 2.7206334957681458e-05, + "loss": 0.4154, + "step": 35599 + }, + { + "epoch": 1.37, + "learning_rate": 2.720569376763273e-05, + "loss": 0.3844, + "step": 35600 + }, + { + "epoch": 1.37, + "learning_rate": 2.7205052577583995e-05, + "loss": 0.3995, + "step": 35601 + }, + { + "epoch": 1.37, + "learning_rate": 2.7204411387535266e-05, + "loss": 0.5562, + "step": 35602 + }, + { + "epoch": 1.37, + "learning_rate": 2.7203770197486538e-05, + "loss": 0.4201, + "step": 35603 + }, + { + "epoch": 1.37, + "learning_rate": 2.7203129007437806e-05, + "loss": 0.4132, + "step": 35604 + }, + { + "epoch": 1.37, + "learning_rate": 2.7202487817389078e-05, + "loss": 0.1906, + "step": 35605 + }, + { + "epoch": 1.37, + "learning_rate": 2.7201846627340343e-05, + "loss": 0.3536, + "step": 35606 + }, + { + "epoch": 1.37, + "learning_rate": 2.7201205437291615e-05, + "loss": 0.4255, + "step": 35607 + }, + { + "epoch": 1.37, + "learning_rate": 2.7200564247242883e-05, + "loss": 0.4458, + "step": 35608 + }, + { + "epoch": 1.37, + "learning_rate": 2.7199923057194155e-05, + "loss": 0.3279, + "step": 35609 + }, + { + "epoch": 1.37, + "learning_rate": 2.719928186714542e-05, + "loss": 0.375, + "step": 35610 + }, + { + "epoch": 1.37, + "learning_rate": 2.719864067709669e-05, + "loss": 0.392, + "step": 35611 + }, + { + "epoch": 1.37, + "learning_rate": 2.7197999487047966e-05, + "loss": 0.3301, + "step": 35612 + }, + { + "epoch": 1.37, + "learning_rate": 2.719735829699923e-05, + "loss": 0.3368, + "step": 35613 + }, + { + "epoch": 1.37, + "learning_rate": 2.7196717106950503e-05, + "loss": 0.4471, + "step": 35614 + }, + { + "epoch": 1.37, + "learning_rate": 2.719607591690177e-05, + "loss": 0.3892, + "step": 35615 + }, + { + "epoch": 1.37, + "learning_rate": 2.7195434726853043e-05, + "loss": 0.6121, + "step": 35616 + }, + { + "epoch": 1.37, + "learning_rate": 2.7194793536804308e-05, + "loss": 0.3542, + "step": 35617 + }, + { + "epoch": 1.37, + "learning_rate": 2.719415234675558e-05, + "loss": 0.3169, + "step": 35618 + }, + { + "epoch": 1.37, + "learning_rate": 2.7193511156706848e-05, + "loss": 0.4269, + "step": 35619 + }, + { + "epoch": 1.37, + "learning_rate": 2.719286996665812e-05, + "loss": 0.2769, + "step": 35620 + }, + { + "epoch": 1.37, + "learning_rate": 2.719222877660939e-05, + "loss": 0.3745, + "step": 35621 + }, + { + "epoch": 1.37, + "learning_rate": 2.7191587586560657e-05, + "loss": 0.3865, + "step": 35622 + }, + { + "epoch": 1.37, + "learning_rate": 2.7190946396511928e-05, + "loss": 0.4323, + "step": 35623 + }, + { + "epoch": 1.37, + "learning_rate": 2.7190305206463197e-05, + "loss": 0.5221, + "step": 35624 + }, + { + "epoch": 1.37, + "learning_rate": 2.7189664016414468e-05, + "loss": 0.4877, + "step": 35625 + }, + { + "epoch": 1.37, + "learning_rate": 2.7189022826365733e-05, + "loss": 0.3985, + "step": 35626 + }, + { + "epoch": 1.37, + "learning_rate": 2.7188381636317005e-05, + "loss": 0.3274, + "step": 35627 + }, + { + "epoch": 1.37, + "learning_rate": 2.7187740446268273e-05, + "loss": 0.4134, + "step": 35628 + }, + { + "epoch": 1.37, + "learning_rate": 2.7187099256219545e-05, + "loss": 0.4611, + "step": 35629 + }, + { + "epoch": 1.37, + "learning_rate": 2.7186458066170817e-05, + "loss": 0.294, + "step": 35630 + }, + { + "epoch": 1.37, + "learning_rate": 2.718581687612208e-05, + "loss": 0.3506, + "step": 35631 + }, + { + "epoch": 1.37, + "learning_rate": 2.7185175686073357e-05, + "loss": 0.4799, + "step": 35632 + }, + { + "epoch": 1.37, + "learning_rate": 2.718453449602462e-05, + "loss": 0.3189, + "step": 35633 + }, + { + "epoch": 1.37, + "learning_rate": 2.7183893305975893e-05, + "loss": 0.3422, + "step": 35634 + }, + { + "epoch": 1.37, + "learning_rate": 2.7183252115927162e-05, + "loss": 0.4049, + "step": 35635 + }, + { + "epoch": 1.37, + "learning_rate": 2.7182610925878433e-05, + "loss": 0.2692, + "step": 35636 + }, + { + "epoch": 1.37, + "learning_rate": 2.71819697358297e-05, + "loss": 0.3794, + "step": 35637 + }, + { + "epoch": 1.37, + "learning_rate": 2.718132854578097e-05, + "loss": 0.315, + "step": 35638 + }, + { + "epoch": 1.37, + "learning_rate": 2.7180687355732242e-05, + "loss": 0.3277, + "step": 35639 + }, + { + "epoch": 1.37, + "learning_rate": 2.718004616568351e-05, + "loss": 0.4442, + "step": 35640 + }, + { + "epoch": 1.37, + "learning_rate": 2.7179404975634782e-05, + "loss": 0.3494, + "step": 35641 + }, + { + "epoch": 1.37, + "learning_rate": 2.7178763785586047e-05, + "loss": 0.416, + "step": 35642 + }, + { + "epoch": 1.37, + "learning_rate": 2.717812259553732e-05, + "loss": 0.3668, + "step": 35643 + }, + { + "epoch": 1.37, + "learning_rate": 2.7177481405488587e-05, + "loss": 0.414, + "step": 35644 + }, + { + "epoch": 1.37, + "learning_rate": 2.717684021543986e-05, + "loss": 0.5309, + "step": 35645 + }, + { + "epoch": 1.37, + "learning_rate": 2.7176199025391124e-05, + "loss": 0.4265, + "step": 35646 + }, + { + "epoch": 1.37, + "learning_rate": 2.7175557835342395e-05, + "loss": 0.2968, + "step": 35647 + }, + { + "epoch": 1.37, + "learning_rate": 2.717491664529367e-05, + "loss": 0.4442, + "step": 35648 + }, + { + "epoch": 1.37, + "learning_rate": 2.7174275455244935e-05, + "loss": 0.3131, + "step": 35649 + }, + { + "epoch": 1.37, + "learning_rate": 2.7173634265196207e-05, + "loss": 0.3395, + "step": 35650 + }, + { + "epoch": 1.37, + "learning_rate": 2.7172993075147475e-05, + "loss": 0.3885, + "step": 35651 + }, + { + "epoch": 1.37, + "learning_rate": 2.7172351885098747e-05, + "loss": 0.4106, + "step": 35652 + }, + { + "epoch": 1.37, + "learning_rate": 2.7171710695050012e-05, + "loss": 0.3608, + "step": 35653 + }, + { + "epoch": 1.37, + "learning_rate": 2.7171069505001284e-05, + "loss": 0.2955, + "step": 35654 + }, + { + "epoch": 1.37, + "learning_rate": 2.7170428314952552e-05, + "loss": 0.3315, + "step": 35655 + }, + { + "epoch": 1.37, + "learning_rate": 2.7169787124903824e-05, + "loss": 0.3262, + "step": 35656 + }, + { + "epoch": 1.37, + "learning_rate": 2.7169145934855095e-05, + "loss": 0.4034, + "step": 35657 + }, + { + "epoch": 1.37, + "learning_rate": 2.716850474480636e-05, + "loss": 0.4574, + "step": 35658 + }, + { + "epoch": 1.37, + "learning_rate": 2.7167863554757632e-05, + "loss": 0.4226, + "step": 35659 + }, + { + "epoch": 1.37, + "learning_rate": 2.71672223647089e-05, + "loss": 0.275, + "step": 35660 + }, + { + "epoch": 1.37, + "learning_rate": 2.7166581174660172e-05, + "loss": 0.376, + "step": 35661 + }, + { + "epoch": 1.37, + "learning_rate": 2.7165939984611437e-05, + "loss": 0.4257, + "step": 35662 + }, + { + "epoch": 1.37, + "learning_rate": 2.716529879456271e-05, + "loss": 0.3061, + "step": 35663 + }, + { + "epoch": 1.37, + "learning_rate": 2.7164657604513977e-05, + "loss": 0.384, + "step": 35664 + }, + { + "epoch": 1.37, + "learning_rate": 2.716401641446525e-05, + "loss": 0.3462, + "step": 35665 + }, + { + "epoch": 1.37, + "learning_rate": 2.716337522441652e-05, + "loss": 0.43, + "step": 35666 + }, + { + "epoch": 1.37, + "learning_rate": 2.7162734034367786e-05, + "loss": 0.5378, + "step": 35667 + }, + { + "epoch": 1.37, + "learning_rate": 2.716209284431906e-05, + "loss": 0.3804, + "step": 35668 + }, + { + "epoch": 1.37, + "learning_rate": 2.7161451654270326e-05, + "loss": 0.3723, + "step": 35669 + }, + { + "epoch": 1.37, + "learning_rate": 2.7160810464221597e-05, + "loss": 0.3469, + "step": 35670 + }, + { + "epoch": 1.37, + "learning_rate": 2.7160169274172866e-05, + "loss": 0.3325, + "step": 35671 + }, + { + "epoch": 1.37, + "learning_rate": 2.7159528084124137e-05, + "loss": 0.4831, + "step": 35672 + }, + { + "epoch": 1.37, + "learning_rate": 2.7158886894075402e-05, + "loss": 0.2807, + "step": 35673 + }, + { + "epoch": 1.37, + "learning_rate": 2.7158245704026674e-05, + "loss": 0.5467, + "step": 35674 + }, + { + "epoch": 1.37, + "learning_rate": 2.7157604513977946e-05, + "loss": 0.3502, + "step": 35675 + }, + { + "epoch": 1.37, + "learning_rate": 2.7156963323929214e-05, + "loss": 0.4347, + "step": 35676 + }, + { + "epoch": 1.37, + "learning_rate": 2.7156322133880486e-05, + "loss": 0.2991, + "step": 35677 + }, + { + "epoch": 1.37, + "learning_rate": 2.715568094383175e-05, + "loss": 0.3578, + "step": 35678 + }, + { + "epoch": 1.37, + "learning_rate": 2.7155039753783022e-05, + "loss": 0.2343, + "step": 35679 + }, + { + "epoch": 1.37, + "learning_rate": 2.715439856373429e-05, + "loss": 0.2954, + "step": 35680 + }, + { + "epoch": 1.37, + "learning_rate": 2.7153757373685562e-05, + "loss": 0.3576, + "step": 35681 + }, + { + "epoch": 1.37, + "learning_rate": 2.7153116183636827e-05, + "loss": 0.3502, + "step": 35682 + }, + { + "epoch": 1.37, + "learning_rate": 2.71524749935881e-05, + "loss": 0.3508, + "step": 35683 + }, + { + "epoch": 1.37, + "learning_rate": 2.7151833803539374e-05, + "loss": 0.3912, + "step": 35684 + }, + { + "epoch": 1.37, + "learning_rate": 2.715119261349064e-05, + "loss": 0.3999, + "step": 35685 + }, + { + "epoch": 1.37, + "learning_rate": 2.715055142344191e-05, + "loss": 0.3914, + "step": 35686 + }, + { + "epoch": 1.37, + "learning_rate": 2.714991023339318e-05, + "loss": 0.5111, + "step": 35687 + }, + { + "epoch": 1.37, + "learning_rate": 2.714926904334445e-05, + "loss": 0.4233, + "step": 35688 + }, + { + "epoch": 1.37, + "learning_rate": 2.7148627853295716e-05, + "loss": 0.4114, + "step": 35689 + }, + { + "epoch": 1.37, + "learning_rate": 2.7147986663246988e-05, + "loss": 0.4405, + "step": 35690 + }, + { + "epoch": 1.37, + "learning_rate": 2.7147345473198256e-05, + "loss": 0.3853, + "step": 35691 + }, + { + "epoch": 1.37, + "learning_rate": 2.7146704283149528e-05, + "loss": 0.3293, + "step": 35692 + }, + { + "epoch": 1.37, + "learning_rate": 2.71460630931008e-05, + "loss": 0.3403, + "step": 35693 + }, + { + "epoch": 1.37, + "learning_rate": 2.7145421903052064e-05, + "loss": 0.3436, + "step": 35694 + }, + { + "epoch": 1.37, + "learning_rate": 2.7144780713003336e-05, + "loss": 0.3862, + "step": 35695 + }, + { + "epoch": 1.37, + "learning_rate": 2.7144139522954604e-05, + "loss": 0.3098, + "step": 35696 + }, + { + "epoch": 1.37, + "learning_rate": 2.7143498332905876e-05, + "loss": 0.2675, + "step": 35697 + }, + { + "epoch": 1.37, + "learning_rate": 2.714285714285714e-05, + "loss": 0.3957, + "step": 35698 + }, + { + "epoch": 1.37, + "learning_rate": 2.7142215952808413e-05, + "loss": 0.3239, + "step": 35699 + }, + { + "epoch": 1.37, + "learning_rate": 2.714157476275968e-05, + "loss": 0.424, + "step": 35700 + }, + { + "epoch": 1.37, + "learning_rate": 2.7140933572710953e-05, + "loss": 0.4, + "step": 35701 + }, + { + "epoch": 1.37, + "learning_rate": 2.7140292382662224e-05, + "loss": 0.439, + "step": 35702 + }, + { + "epoch": 1.37, + "learning_rate": 2.7139651192613493e-05, + "loss": 0.3716, + "step": 35703 + }, + { + "epoch": 1.37, + "learning_rate": 2.7139010002564764e-05, + "loss": 0.5196, + "step": 35704 + }, + { + "epoch": 1.37, + "learning_rate": 2.713836881251603e-05, + "loss": 0.3337, + "step": 35705 + }, + { + "epoch": 1.37, + "learning_rate": 2.71377276224673e-05, + "loss": 0.464, + "step": 35706 + }, + { + "epoch": 1.37, + "learning_rate": 2.713708643241857e-05, + "loss": 0.3783, + "step": 35707 + }, + { + "epoch": 1.37, + "learning_rate": 2.713644524236984e-05, + "loss": 0.4243, + "step": 35708 + }, + { + "epoch": 1.37, + "learning_rate": 2.7135804052321106e-05, + "loss": 0.2868, + "step": 35709 + }, + { + "epoch": 1.37, + "learning_rate": 2.7135162862272378e-05, + "loss": 0.4547, + "step": 35710 + }, + { + "epoch": 1.37, + "learning_rate": 2.713452167222365e-05, + "loss": 0.4322, + "step": 35711 + }, + { + "epoch": 1.37, + "learning_rate": 2.7133880482174918e-05, + "loss": 0.4412, + "step": 35712 + }, + { + "epoch": 1.37, + "learning_rate": 2.713323929212619e-05, + "loss": 0.2497, + "step": 35713 + }, + { + "epoch": 1.37, + "learning_rate": 2.7132598102077455e-05, + "loss": 0.5067, + "step": 35714 + }, + { + "epoch": 1.37, + "learning_rate": 2.7131956912028726e-05, + "loss": 0.3706, + "step": 35715 + }, + { + "epoch": 1.37, + "learning_rate": 2.7131315721979995e-05, + "loss": 0.3777, + "step": 35716 + }, + { + "epoch": 1.37, + "learning_rate": 2.7130674531931266e-05, + "loss": 0.3919, + "step": 35717 + }, + { + "epoch": 1.37, + "learning_rate": 2.713003334188253e-05, + "loss": 0.3519, + "step": 35718 + }, + { + "epoch": 1.37, + "learning_rate": 2.7129392151833803e-05, + "loss": 0.3702, + "step": 35719 + }, + { + "epoch": 1.37, + "learning_rate": 2.7128750961785078e-05, + "loss": 0.4924, + "step": 35720 + }, + { + "epoch": 1.37, + "learning_rate": 2.7128109771736343e-05, + "loss": 0.3821, + "step": 35721 + }, + { + "epoch": 1.37, + "learning_rate": 2.7127468581687615e-05, + "loss": 0.3283, + "step": 35722 + }, + { + "epoch": 1.37, + "learning_rate": 2.7126827391638883e-05, + "loss": 0.4062, + "step": 35723 + }, + { + "epoch": 1.37, + "learning_rate": 2.7126186201590155e-05, + "loss": 0.4126, + "step": 35724 + }, + { + "epoch": 1.37, + "learning_rate": 2.712554501154142e-05, + "loss": 0.4934, + "step": 35725 + }, + { + "epoch": 1.37, + "learning_rate": 2.712490382149269e-05, + "loss": 0.2717, + "step": 35726 + }, + { + "epoch": 1.37, + "learning_rate": 2.712426263144396e-05, + "loss": 0.4364, + "step": 35727 + }, + { + "epoch": 1.37, + "learning_rate": 2.712362144139523e-05, + "loss": 0.4006, + "step": 35728 + }, + { + "epoch": 1.37, + "learning_rate": 2.7122980251346503e-05, + "loss": 0.4051, + "step": 35729 + }, + { + "epoch": 1.37, + "learning_rate": 2.7122339061297768e-05, + "loss": 0.34, + "step": 35730 + }, + { + "epoch": 1.37, + "learning_rate": 2.712169787124904e-05, + "loss": 0.2721, + "step": 35731 + }, + { + "epoch": 1.37, + "learning_rate": 2.7121056681200308e-05, + "loss": 0.352, + "step": 35732 + }, + { + "epoch": 1.37, + "learning_rate": 2.712041549115158e-05, + "loss": 0.4152, + "step": 35733 + }, + { + "epoch": 1.37, + "learning_rate": 2.7119774301102845e-05, + "loss": 0.2714, + "step": 35734 + }, + { + "epoch": 1.37, + "learning_rate": 2.7119133111054117e-05, + "loss": 0.4397, + "step": 35735 + }, + { + "epoch": 1.37, + "learning_rate": 2.711849192100539e-05, + "loss": 0.3113, + "step": 35736 + }, + { + "epoch": 1.37, + "learning_rate": 2.7117850730956657e-05, + "loss": 0.3078, + "step": 35737 + }, + { + "epoch": 1.37, + "learning_rate": 2.711720954090793e-05, + "loss": 0.4248, + "step": 35738 + }, + { + "epoch": 1.37, + "learning_rate": 2.7116568350859197e-05, + "loss": 0.3444, + "step": 35739 + }, + { + "epoch": 1.37, + "learning_rate": 2.711592716081047e-05, + "loss": 0.2894, + "step": 35740 + }, + { + "epoch": 1.37, + "learning_rate": 2.7115285970761733e-05, + "loss": 0.3566, + "step": 35741 + }, + { + "epoch": 1.37, + "learning_rate": 2.7114644780713005e-05, + "loss": 0.4615, + "step": 35742 + }, + { + "epoch": 1.37, + "learning_rate": 2.7114003590664273e-05, + "loss": 0.4455, + "step": 35743 + }, + { + "epoch": 1.37, + "learning_rate": 2.7113362400615545e-05, + "loss": 0.3839, + "step": 35744 + }, + { + "epoch": 1.37, + "learning_rate": 2.7112721210566817e-05, + "loss": 0.3237, + "step": 35745 + }, + { + "epoch": 1.37, + "learning_rate": 2.7112080020518082e-05, + "loss": 0.5126, + "step": 35746 + }, + { + "epoch": 1.37, + "learning_rate": 2.7111438830469353e-05, + "loss": 0.3241, + "step": 35747 + }, + { + "epoch": 1.37, + "learning_rate": 2.7110797640420622e-05, + "loss": 0.344, + "step": 35748 + }, + { + "epoch": 1.37, + "learning_rate": 2.7110156450371893e-05, + "loss": 0.3988, + "step": 35749 + }, + { + "epoch": 1.37, + "learning_rate": 2.710951526032316e-05, + "loss": 0.2559, + "step": 35750 + }, + { + "epoch": 1.37, + "learning_rate": 2.710887407027443e-05, + "loss": 0.3313, + "step": 35751 + }, + { + "epoch": 1.37, + "learning_rate": 2.71082328802257e-05, + "loss": 0.4059, + "step": 35752 + }, + { + "epoch": 1.37, + "learning_rate": 2.710759169017697e-05, + "loss": 0.2856, + "step": 35753 + }, + { + "epoch": 1.37, + "learning_rate": 2.7106950500128242e-05, + "loss": 0.4257, + "step": 35754 + }, + { + "epoch": 1.37, + "learning_rate": 2.7106309310079507e-05, + "loss": 0.4205, + "step": 35755 + }, + { + "epoch": 1.37, + "learning_rate": 2.7105668120030782e-05, + "loss": 0.2674, + "step": 35756 + }, + { + "epoch": 1.37, + "learning_rate": 2.7105026929982047e-05, + "loss": 0.4469, + "step": 35757 + }, + { + "epoch": 1.37, + "learning_rate": 2.710438573993332e-05, + "loss": 0.2412, + "step": 35758 + }, + { + "epoch": 1.37, + "learning_rate": 2.7103744549884587e-05, + "loss": 0.411, + "step": 35759 + }, + { + "epoch": 1.37, + "learning_rate": 2.710310335983586e-05, + "loss": 0.3624, + "step": 35760 + }, + { + "epoch": 1.37, + "learning_rate": 2.7102462169787124e-05, + "loss": 0.5789, + "step": 35761 + }, + { + "epoch": 1.37, + "learning_rate": 2.7101820979738395e-05, + "loss": 0.3317, + "step": 35762 + }, + { + "epoch": 1.37, + "learning_rate": 2.7101179789689667e-05, + "loss": 0.3332, + "step": 35763 + }, + { + "epoch": 1.38, + "learning_rate": 2.7100538599640935e-05, + "loss": 0.3447, + "step": 35764 + }, + { + "epoch": 1.38, + "learning_rate": 2.7099897409592207e-05, + "loss": 0.3193, + "step": 35765 + }, + { + "epoch": 1.38, + "learning_rate": 2.7099256219543472e-05, + "loss": 0.3685, + "step": 35766 + }, + { + "epoch": 1.38, + "learning_rate": 2.7098615029494744e-05, + "loss": 0.5091, + "step": 35767 + }, + { + "epoch": 1.38, + "learning_rate": 2.7097973839446012e-05, + "loss": 0.3869, + "step": 35768 + }, + { + "epoch": 1.38, + "learning_rate": 2.7097332649397284e-05, + "loss": 0.3868, + "step": 35769 + }, + { + "epoch": 1.38, + "learning_rate": 2.709669145934855e-05, + "loss": 0.3536, + "step": 35770 + }, + { + "epoch": 1.38, + "learning_rate": 2.709605026929982e-05, + "loss": 0.4262, + "step": 35771 + }, + { + "epoch": 1.38, + "learning_rate": 2.7095409079251096e-05, + "loss": 0.3555, + "step": 35772 + }, + { + "epoch": 1.38, + "learning_rate": 2.709476788920236e-05, + "loss": 0.336, + "step": 35773 + }, + { + "epoch": 1.38, + "learning_rate": 2.7094126699153632e-05, + "loss": 0.2956, + "step": 35774 + }, + { + "epoch": 1.38, + "learning_rate": 2.70934855091049e-05, + "loss": 0.4292, + "step": 35775 + }, + { + "epoch": 1.38, + "learning_rate": 2.7092844319056172e-05, + "loss": 0.3474, + "step": 35776 + }, + { + "epoch": 1.38, + "learning_rate": 2.7092203129007437e-05, + "loss": 0.4018, + "step": 35777 + }, + { + "epoch": 1.38, + "learning_rate": 2.709156193895871e-05, + "loss": 0.3255, + "step": 35778 + }, + { + "epoch": 1.38, + "learning_rate": 2.7090920748909977e-05, + "loss": 0.4359, + "step": 35779 + }, + { + "epoch": 1.38, + "learning_rate": 2.709027955886125e-05, + "loss": 0.3727, + "step": 35780 + }, + { + "epoch": 1.38, + "learning_rate": 2.708963836881252e-05, + "loss": 0.3156, + "step": 35781 + }, + { + "epoch": 1.38, + "learning_rate": 2.7088997178763786e-05, + "loss": 0.4141, + "step": 35782 + }, + { + "epoch": 1.38, + "learning_rate": 2.7088355988715057e-05, + "loss": 0.4079, + "step": 35783 + }, + { + "epoch": 1.38, + "learning_rate": 2.7087714798666326e-05, + "loss": 0.367, + "step": 35784 + }, + { + "epoch": 1.38, + "learning_rate": 2.7087073608617597e-05, + "loss": 0.3626, + "step": 35785 + }, + { + "epoch": 1.38, + "learning_rate": 2.7086432418568862e-05, + "loss": 0.2619, + "step": 35786 + }, + { + "epoch": 1.38, + "learning_rate": 2.7085791228520134e-05, + "loss": 0.2922, + "step": 35787 + }, + { + "epoch": 1.38, + "learning_rate": 2.7085150038471402e-05, + "loss": 0.3062, + "step": 35788 + }, + { + "epoch": 1.38, + "learning_rate": 2.7084508848422674e-05, + "loss": 0.4794, + "step": 35789 + }, + { + "epoch": 1.38, + "learning_rate": 2.7083867658373946e-05, + "loss": 0.3571, + "step": 35790 + }, + { + "epoch": 1.38, + "learning_rate": 2.708322646832521e-05, + "loss": 0.3158, + "step": 35791 + }, + { + "epoch": 1.38, + "learning_rate": 2.7082585278276486e-05, + "loss": 0.4579, + "step": 35792 + }, + { + "epoch": 1.38, + "learning_rate": 2.708194408822775e-05, + "loss": 0.335, + "step": 35793 + }, + { + "epoch": 1.38, + "learning_rate": 2.7081302898179022e-05, + "loss": 0.4018, + "step": 35794 + }, + { + "epoch": 1.38, + "learning_rate": 2.708066170813029e-05, + "loss": 0.3474, + "step": 35795 + }, + { + "epoch": 1.38, + "learning_rate": 2.7080020518081563e-05, + "loss": 0.353, + "step": 35796 + }, + { + "epoch": 1.38, + "learning_rate": 2.7079379328032827e-05, + "loss": 0.4142, + "step": 35797 + }, + { + "epoch": 1.38, + "learning_rate": 2.70787381379841e-05, + "loss": 0.3593, + "step": 35798 + }, + { + "epoch": 1.38, + "learning_rate": 2.707809694793537e-05, + "loss": 0.5082, + "step": 35799 + }, + { + "epoch": 1.38, + "learning_rate": 2.707745575788664e-05, + "loss": 0.2857, + "step": 35800 + }, + { + "epoch": 1.38, + "learning_rate": 2.707681456783791e-05, + "loss": 0.4615, + "step": 35801 + }, + { + "epoch": 1.38, + "learning_rate": 2.7076173377789176e-05, + "loss": 0.2682, + "step": 35802 + }, + { + "epoch": 1.38, + "learning_rate": 2.7075532187740448e-05, + "loss": 0.3888, + "step": 35803 + }, + { + "epoch": 1.38, + "learning_rate": 2.7074890997691716e-05, + "loss": 0.3078, + "step": 35804 + }, + { + "epoch": 1.38, + "learning_rate": 2.7074249807642988e-05, + "loss": 0.3552, + "step": 35805 + }, + { + "epoch": 1.38, + "learning_rate": 2.7073608617594253e-05, + "loss": 0.4015, + "step": 35806 + }, + { + "epoch": 1.38, + "learning_rate": 2.7072967427545524e-05, + "loss": 0.3622, + "step": 35807 + }, + { + "epoch": 1.38, + "learning_rate": 2.70723262374968e-05, + "loss": 0.3973, + "step": 35808 + }, + { + "epoch": 1.38, + "learning_rate": 2.7071685047448064e-05, + "loss": 0.3748, + "step": 35809 + }, + { + "epoch": 1.38, + "learning_rate": 2.7071043857399336e-05, + "loss": 0.4756, + "step": 35810 + }, + { + "epoch": 1.38, + "learning_rate": 2.7070402667350604e-05, + "loss": 0.3182, + "step": 35811 + }, + { + "epoch": 1.38, + "learning_rate": 2.7069761477301876e-05, + "loss": 0.3641, + "step": 35812 + }, + { + "epoch": 1.38, + "learning_rate": 2.706912028725314e-05, + "loss": 0.3354, + "step": 35813 + }, + { + "epoch": 1.38, + "learning_rate": 2.7068479097204413e-05, + "loss": 0.2609, + "step": 35814 + }, + { + "epoch": 1.38, + "learning_rate": 2.706783790715568e-05, + "loss": 0.4564, + "step": 35815 + }, + { + "epoch": 1.38, + "learning_rate": 2.7067196717106953e-05, + "loss": 0.3319, + "step": 35816 + }, + { + "epoch": 1.38, + "learning_rate": 2.7066555527058225e-05, + "loss": 0.3544, + "step": 35817 + }, + { + "epoch": 1.38, + "learning_rate": 2.706591433700949e-05, + "loss": 0.4226, + "step": 35818 + }, + { + "epoch": 1.38, + "learning_rate": 2.706527314696076e-05, + "loss": 0.4524, + "step": 35819 + }, + { + "epoch": 1.38, + "learning_rate": 2.706463195691203e-05, + "loss": 0.3718, + "step": 35820 + }, + { + "epoch": 1.38, + "learning_rate": 2.70639907668633e-05, + "loss": 0.2622, + "step": 35821 + }, + { + "epoch": 1.38, + "learning_rate": 2.7063349576814566e-05, + "loss": 0.277, + "step": 35822 + }, + { + "epoch": 1.38, + "learning_rate": 2.7062708386765838e-05, + "loss": 0.3025, + "step": 35823 + }, + { + "epoch": 1.38, + "learning_rate": 2.7062067196717106e-05, + "loss": 0.4177, + "step": 35824 + }, + { + "epoch": 1.38, + "learning_rate": 2.7061426006668378e-05, + "loss": 0.2681, + "step": 35825 + }, + { + "epoch": 1.38, + "learning_rate": 2.706078481661965e-05, + "loss": 0.3264, + "step": 35826 + }, + { + "epoch": 1.38, + "learning_rate": 2.7060143626570918e-05, + "loss": 0.5172, + "step": 35827 + }, + { + "epoch": 1.38, + "learning_rate": 2.705950243652219e-05, + "loss": 0.2999, + "step": 35828 + }, + { + "epoch": 1.38, + "learning_rate": 2.7058861246473455e-05, + "loss": 0.5013, + "step": 35829 + }, + { + "epoch": 1.38, + "learning_rate": 2.7058220056424726e-05, + "loss": 0.5323, + "step": 35830 + }, + { + "epoch": 1.38, + "learning_rate": 2.7057578866375995e-05, + "loss": 0.2734, + "step": 35831 + }, + { + "epoch": 1.38, + "learning_rate": 2.7056937676327266e-05, + "loss": 0.3792, + "step": 35832 + }, + { + "epoch": 1.38, + "learning_rate": 2.705629648627853e-05, + "loss": 0.2452, + "step": 35833 + }, + { + "epoch": 1.38, + "learning_rate": 2.7055655296229803e-05, + "loss": 0.4343, + "step": 35834 + }, + { + "epoch": 1.38, + "learning_rate": 2.7055014106181075e-05, + "loss": 0.5047, + "step": 35835 + }, + { + "epoch": 1.38, + "learning_rate": 2.7054372916132343e-05, + "loss": 0.3733, + "step": 35836 + }, + { + "epoch": 1.38, + "learning_rate": 2.7053731726083615e-05, + "loss": 0.341, + "step": 35837 + }, + { + "epoch": 1.38, + "learning_rate": 2.705309053603488e-05, + "loss": 0.4319, + "step": 35838 + }, + { + "epoch": 1.38, + "learning_rate": 2.705244934598615e-05, + "loss": 0.3916, + "step": 35839 + }, + { + "epoch": 1.38, + "learning_rate": 2.705180815593742e-05, + "loss": 0.4475, + "step": 35840 + }, + { + "epoch": 1.38, + "learning_rate": 2.705116696588869e-05, + "loss": 0.489, + "step": 35841 + }, + { + "epoch": 1.38, + "learning_rate": 2.7050525775839956e-05, + "loss": 0.3483, + "step": 35842 + }, + { + "epoch": 1.38, + "learning_rate": 2.7049884585791228e-05, + "loss": 0.4152, + "step": 35843 + }, + { + "epoch": 1.38, + "learning_rate": 2.7049243395742503e-05, + "loss": 0.3451, + "step": 35844 + }, + { + "epoch": 1.38, + "learning_rate": 2.7048602205693768e-05, + "loss": 0.3545, + "step": 35845 + }, + { + "epoch": 1.38, + "learning_rate": 2.704796101564504e-05, + "loss": 0.2942, + "step": 35846 + }, + { + "epoch": 1.38, + "learning_rate": 2.7047319825596308e-05, + "loss": 0.3715, + "step": 35847 + }, + { + "epoch": 1.38, + "learning_rate": 2.704667863554758e-05, + "loss": 0.4877, + "step": 35848 + }, + { + "epoch": 1.38, + "learning_rate": 2.7046037445498845e-05, + "loss": 0.353, + "step": 35849 + }, + { + "epoch": 1.38, + "learning_rate": 2.7045396255450117e-05, + "loss": 0.4835, + "step": 35850 + }, + { + "epoch": 1.38, + "learning_rate": 2.7044755065401385e-05, + "loss": 0.4588, + "step": 35851 + }, + { + "epoch": 1.38, + "learning_rate": 2.7044113875352657e-05, + "loss": 0.6573, + "step": 35852 + }, + { + "epoch": 1.38, + "learning_rate": 2.704347268530393e-05, + "loss": 0.3676, + "step": 35853 + }, + { + "epoch": 1.38, + "learning_rate": 2.7042831495255193e-05, + "loss": 0.4499, + "step": 35854 + }, + { + "epoch": 1.38, + "learning_rate": 2.7042190305206465e-05, + "loss": 0.4108, + "step": 35855 + }, + { + "epoch": 1.38, + "learning_rate": 2.7041549115157733e-05, + "loss": 0.3901, + "step": 35856 + }, + { + "epoch": 1.38, + "learning_rate": 2.7040907925109005e-05, + "loss": 0.3729, + "step": 35857 + }, + { + "epoch": 1.38, + "learning_rate": 2.704026673506027e-05, + "loss": 0.5152, + "step": 35858 + }, + { + "epoch": 1.38, + "learning_rate": 2.7039625545011542e-05, + "loss": 0.3775, + "step": 35859 + }, + { + "epoch": 1.38, + "learning_rate": 2.703898435496281e-05, + "loss": 0.3554, + "step": 35860 + }, + { + "epoch": 1.38, + "learning_rate": 2.7038343164914082e-05, + "loss": 0.3445, + "step": 35861 + }, + { + "epoch": 1.38, + "learning_rate": 2.7037701974865354e-05, + "loss": 0.3984, + "step": 35862 + }, + { + "epoch": 1.38, + "learning_rate": 2.7037060784816622e-05, + "loss": 0.4049, + "step": 35863 + }, + { + "epoch": 1.38, + "learning_rate": 2.7036419594767894e-05, + "loss": 0.2899, + "step": 35864 + }, + { + "epoch": 1.38, + "learning_rate": 2.703577840471916e-05, + "loss": 0.4722, + "step": 35865 + }, + { + "epoch": 1.38, + "learning_rate": 2.703513721467043e-05, + "loss": 0.3779, + "step": 35866 + }, + { + "epoch": 1.38, + "learning_rate": 2.70344960246217e-05, + "loss": 0.3335, + "step": 35867 + }, + { + "epoch": 1.38, + "learning_rate": 2.703385483457297e-05, + "loss": 0.3376, + "step": 35868 + }, + { + "epoch": 1.38, + "learning_rate": 2.7033213644524235e-05, + "loss": 0.3269, + "step": 35869 + }, + { + "epoch": 1.38, + "learning_rate": 2.7032572454475507e-05, + "loss": 0.2935, + "step": 35870 + }, + { + "epoch": 1.38, + "learning_rate": 2.703193126442678e-05, + "loss": 0.3801, + "step": 35871 + }, + { + "epoch": 1.38, + "learning_rate": 2.7031290074378047e-05, + "loss": 0.4226, + "step": 35872 + }, + { + "epoch": 1.38, + "learning_rate": 2.703064888432932e-05, + "loss": 0.4673, + "step": 35873 + }, + { + "epoch": 1.38, + "learning_rate": 2.7030007694280584e-05, + "loss": 0.4931, + "step": 35874 + }, + { + "epoch": 1.38, + "learning_rate": 2.7029366504231855e-05, + "loss": 0.2967, + "step": 35875 + }, + { + "epoch": 1.38, + "learning_rate": 2.7028725314183124e-05, + "loss": 0.4014, + "step": 35876 + }, + { + "epoch": 1.38, + "learning_rate": 2.7028084124134395e-05, + "loss": 0.3608, + "step": 35877 + }, + { + "epoch": 1.38, + "learning_rate": 2.702744293408566e-05, + "loss": 0.3388, + "step": 35878 + }, + { + "epoch": 1.38, + "learning_rate": 2.7026801744036932e-05, + "loss": 0.3111, + "step": 35879 + }, + { + "epoch": 1.38, + "learning_rate": 2.7026160553988207e-05, + "loss": 0.3894, + "step": 35880 + }, + { + "epoch": 1.38, + "learning_rate": 2.7025519363939472e-05, + "loss": 0.3186, + "step": 35881 + }, + { + "epoch": 1.38, + "learning_rate": 2.7024878173890744e-05, + "loss": 0.3915, + "step": 35882 + }, + { + "epoch": 1.38, + "learning_rate": 2.7024236983842012e-05, + "loss": 0.3731, + "step": 35883 + }, + { + "epoch": 1.38, + "learning_rate": 2.7023595793793284e-05, + "loss": 0.4225, + "step": 35884 + }, + { + "epoch": 1.38, + "learning_rate": 2.702295460374455e-05, + "loss": 0.4021, + "step": 35885 + }, + { + "epoch": 1.38, + "learning_rate": 2.702231341369582e-05, + "loss": 0.3558, + "step": 35886 + }, + { + "epoch": 1.38, + "learning_rate": 2.702167222364709e-05, + "loss": 0.4856, + "step": 35887 + }, + { + "epoch": 1.38, + "learning_rate": 2.702103103359836e-05, + "loss": 0.3405, + "step": 35888 + }, + { + "epoch": 1.38, + "learning_rate": 2.7020389843549632e-05, + "loss": 0.4005, + "step": 35889 + }, + { + "epoch": 1.38, + "learning_rate": 2.7019748653500897e-05, + "loss": 0.5023, + "step": 35890 + }, + { + "epoch": 1.38, + "learning_rate": 2.701910746345217e-05, + "loss": 0.4034, + "step": 35891 + }, + { + "epoch": 1.38, + "learning_rate": 2.7018466273403437e-05, + "loss": 0.3202, + "step": 35892 + }, + { + "epoch": 1.38, + "learning_rate": 2.701782508335471e-05, + "loss": 0.4032, + "step": 35893 + }, + { + "epoch": 1.38, + "learning_rate": 2.7017183893305974e-05, + "loss": 0.3886, + "step": 35894 + }, + { + "epoch": 1.38, + "learning_rate": 2.7016542703257246e-05, + "loss": 0.2985, + "step": 35895 + }, + { + "epoch": 1.38, + "learning_rate": 2.7015901513208514e-05, + "loss": 0.4935, + "step": 35896 + }, + { + "epoch": 1.38, + "learning_rate": 2.7015260323159786e-05, + "loss": 0.3802, + "step": 35897 + }, + { + "epoch": 1.38, + "learning_rate": 2.7014619133111057e-05, + "loss": 0.3012, + "step": 35898 + }, + { + "epoch": 1.38, + "learning_rate": 2.7013977943062326e-05, + "loss": 0.5051, + "step": 35899 + }, + { + "epoch": 1.38, + "learning_rate": 2.7013336753013597e-05, + "loss": 0.4009, + "step": 35900 + }, + { + "epoch": 1.38, + "learning_rate": 2.7012695562964862e-05, + "loss": 0.3196, + "step": 35901 + }, + { + "epoch": 1.38, + "learning_rate": 2.7012054372916134e-05, + "loss": 0.2942, + "step": 35902 + }, + { + "epoch": 1.38, + "learning_rate": 2.7011413182867402e-05, + "loss": 0.2992, + "step": 35903 + }, + { + "epoch": 1.38, + "learning_rate": 2.7010771992818674e-05, + "loss": 0.3302, + "step": 35904 + }, + { + "epoch": 1.38, + "learning_rate": 2.701013080276994e-05, + "loss": 0.3459, + "step": 35905 + }, + { + "epoch": 1.38, + "learning_rate": 2.700948961272121e-05, + "loss": 0.3642, + "step": 35906 + }, + { + "epoch": 1.38, + "learning_rate": 2.7008848422672483e-05, + "loss": 0.4314, + "step": 35907 + }, + { + "epoch": 1.38, + "learning_rate": 2.700820723262375e-05, + "loss": 0.3244, + "step": 35908 + }, + { + "epoch": 1.38, + "learning_rate": 2.7007566042575023e-05, + "loss": 0.3997, + "step": 35909 + }, + { + "epoch": 1.38, + "learning_rate": 2.7006924852526287e-05, + "loss": 0.3079, + "step": 35910 + }, + { + "epoch": 1.38, + "learning_rate": 2.700628366247756e-05, + "loss": 0.3434, + "step": 35911 + }, + { + "epoch": 1.38, + "learning_rate": 2.7005642472428828e-05, + "loss": 0.3525, + "step": 35912 + }, + { + "epoch": 1.38, + "learning_rate": 2.70050012823801e-05, + "loss": 0.3009, + "step": 35913 + }, + { + "epoch": 1.38, + "learning_rate": 2.7004360092331364e-05, + "loss": 0.513, + "step": 35914 + }, + { + "epoch": 1.38, + "learning_rate": 2.7003718902282636e-05, + "loss": 0.4381, + "step": 35915 + }, + { + "epoch": 1.38, + "learning_rate": 2.700307771223391e-05, + "loss": 0.3791, + "step": 35916 + }, + { + "epoch": 1.38, + "learning_rate": 2.7002436522185176e-05, + "loss": 0.4527, + "step": 35917 + }, + { + "epoch": 1.38, + "learning_rate": 2.7001795332136448e-05, + "loss": 0.3274, + "step": 35918 + }, + { + "epoch": 1.38, + "learning_rate": 2.7001154142087716e-05, + "loss": 0.3781, + "step": 35919 + }, + { + "epoch": 1.38, + "learning_rate": 2.7000512952038988e-05, + "loss": 0.4082, + "step": 35920 + }, + { + "epoch": 1.38, + "learning_rate": 2.6999871761990253e-05, + "loss": 0.2891, + "step": 35921 + }, + { + "epoch": 1.38, + "learning_rate": 2.6999230571941524e-05, + "loss": 0.3308, + "step": 35922 + }, + { + "epoch": 1.38, + "learning_rate": 2.6998589381892796e-05, + "loss": 0.4007, + "step": 35923 + }, + { + "epoch": 1.38, + "learning_rate": 2.6997948191844064e-05, + "loss": 0.4347, + "step": 35924 + }, + { + "epoch": 1.38, + "learning_rate": 2.6997307001795336e-05, + "loss": 0.4002, + "step": 35925 + }, + { + "epoch": 1.38, + "learning_rate": 2.69966658117466e-05, + "loss": 0.4043, + "step": 35926 + }, + { + "epoch": 1.38, + "learning_rate": 2.6996024621697873e-05, + "loss": 0.6238, + "step": 35927 + }, + { + "epoch": 1.38, + "learning_rate": 2.699538343164914e-05, + "loss": 0.3352, + "step": 35928 + }, + { + "epoch": 1.38, + "learning_rate": 2.6994742241600413e-05, + "loss": 0.409, + "step": 35929 + }, + { + "epoch": 1.38, + "learning_rate": 2.6994101051551678e-05, + "loss": 0.3772, + "step": 35930 + }, + { + "epoch": 1.38, + "learning_rate": 2.699345986150295e-05, + "loss": 0.3391, + "step": 35931 + }, + { + "epoch": 1.38, + "learning_rate": 2.6992818671454225e-05, + "loss": 0.4377, + "step": 35932 + }, + { + "epoch": 1.38, + "learning_rate": 2.699217748140549e-05, + "loss": 0.2581, + "step": 35933 + }, + { + "epoch": 1.38, + "learning_rate": 2.699153629135676e-05, + "loss": 0.3382, + "step": 35934 + }, + { + "epoch": 1.38, + "learning_rate": 2.699089510130803e-05, + "loss": 0.3882, + "step": 35935 + }, + { + "epoch": 1.38, + "learning_rate": 2.69902539112593e-05, + "loss": 0.3114, + "step": 35936 + }, + { + "epoch": 1.38, + "learning_rate": 2.6989612721210566e-05, + "loss": 0.2715, + "step": 35937 + }, + { + "epoch": 1.38, + "learning_rate": 2.6988971531161838e-05, + "loss": 0.2428, + "step": 35938 + }, + { + "epoch": 1.38, + "learning_rate": 2.6988330341113106e-05, + "loss": 0.2938, + "step": 35939 + }, + { + "epoch": 1.38, + "learning_rate": 2.6987689151064378e-05, + "loss": 0.3888, + "step": 35940 + }, + { + "epoch": 1.38, + "learning_rate": 2.698704796101565e-05, + "loss": 0.3435, + "step": 35941 + }, + { + "epoch": 1.38, + "learning_rate": 2.6986406770966915e-05, + "loss": 0.3803, + "step": 35942 + }, + { + "epoch": 1.38, + "learning_rate": 2.6985765580918186e-05, + "loss": 0.4151, + "step": 35943 + }, + { + "epoch": 1.38, + "learning_rate": 2.6985124390869455e-05, + "loss": 0.4929, + "step": 35944 + }, + { + "epoch": 1.38, + "learning_rate": 2.6984483200820726e-05, + "loss": 0.4408, + "step": 35945 + }, + { + "epoch": 1.38, + "learning_rate": 2.698384201077199e-05, + "loss": 0.2975, + "step": 35946 + }, + { + "epoch": 1.38, + "learning_rate": 2.6983200820723263e-05, + "loss": 0.3895, + "step": 35947 + }, + { + "epoch": 1.38, + "learning_rate": 2.698255963067453e-05, + "loss": 0.4844, + "step": 35948 + }, + { + "epoch": 1.38, + "learning_rate": 2.6981918440625803e-05, + "loss": 0.2907, + "step": 35949 + }, + { + "epoch": 1.38, + "learning_rate": 2.6981277250577075e-05, + "loss": 0.4444, + "step": 35950 + }, + { + "epoch": 1.38, + "learning_rate": 2.6980636060528343e-05, + "loss": 0.3749, + "step": 35951 + }, + { + "epoch": 1.38, + "learning_rate": 2.6979994870479615e-05, + "loss": 0.2538, + "step": 35952 + }, + { + "epoch": 1.38, + "learning_rate": 2.697935368043088e-05, + "loss": 0.3981, + "step": 35953 + }, + { + "epoch": 1.38, + "learning_rate": 2.697871249038215e-05, + "loss": 0.4529, + "step": 35954 + }, + { + "epoch": 1.38, + "learning_rate": 2.697807130033342e-05, + "loss": 0.3999, + "step": 35955 + }, + { + "epoch": 1.38, + "learning_rate": 2.697743011028469e-05, + "loss": 0.3531, + "step": 35956 + }, + { + "epoch": 1.38, + "learning_rate": 2.6976788920235957e-05, + "loss": 0.3694, + "step": 35957 + }, + { + "epoch": 1.38, + "learning_rate": 2.6976147730187228e-05, + "loss": 0.3154, + "step": 35958 + }, + { + "epoch": 1.38, + "learning_rate": 2.69755065401385e-05, + "loss": 0.3316, + "step": 35959 + }, + { + "epoch": 1.38, + "learning_rate": 2.6974865350089768e-05, + "loss": 0.3106, + "step": 35960 + }, + { + "epoch": 1.38, + "learning_rate": 2.697422416004104e-05, + "loss": 0.469, + "step": 35961 + }, + { + "epoch": 1.38, + "learning_rate": 2.6973582969992305e-05, + "loss": 0.4499, + "step": 35962 + }, + { + "epoch": 1.38, + "learning_rate": 2.6972941779943577e-05, + "loss": 0.4362, + "step": 35963 + }, + { + "epoch": 1.38, + "learning_rate": 2.6972300589894845e-05, + "loss": 0.3366, + "step": 35964 + }, + { + "epoch": 1.38, + "learning_rate": 2.6971659399846117e-05, + "loss": 0.2956, + "step": 35965 + }, + { + "epoch": 1.38, + "learning_rate": 2.697101820979738e-05, + "loss": 0.3465, + "step": 35966 + }, + { + "epoch": 1.38, + "learning_rate": 2.6970377019748653e-05, + "loss": 0.2697, + "step": 35967 + }, + { + "epoch": 1.38, + "learning_rate": 2.696973582969993e-05, + "loss": 0.3693, + "step": 35968 + }, + { + "epoch": 1.38, + "learning_rate": 2.6969094639651193e-05, + "loss": 0.3122, + "step": 35969 + }, + { + "epoch": 1.38, + "learning_rate": 2.6968453449602465e-05, + "loss": 0.4371, + "step": 35970 + }, + { + "epoch": 1.38, + "learning_rate": 2.6967812259553733e-05, + "loss": 0.3215, + "step": 35971 + }, + { + "epoch": 1.38, + "learning_rate": 2.6967171069505005e-05, + "loss": 0.348, + "step": 35972 + }, + { + "epoch": 1.38, + "learning_rate": 2.696652987945627e-05, + "loss": 0.41, + "step": 35973 + }, + { + "epoch": 1.38, + "learning_rate": 2.6965888689407542e-05, + "loss": 0.5223, + "step": 35974 + }, + { + "epoch": 1.38, + "learning_rate": 2.696524749935881e-05, + "loss": 0.332, + "step": 35975 + }, + { + "epoch": 1.38, + "learning_rate": 2.6964606309310082e-05, + "loss": 0.3252, + "step": 35976 + }, + { + "epoch": 1.38, + "learning_rate": 2.6963965119261354e-05, + "loss": 0.3761, + "step": 35977 + }, + { + "epoch": 1.38, + "learning_rate": 2.696332392921262e-05, + "loss": 0.3062, + "step": 35978 + }, + { + "epoch": 1.38, + "learning_rate": 2.696268273916389e-05, + "loss": 0.3306, + "step": 35979 + }, + { + "epoch": 1.38, + "learning_rate": 2.696204154911516e-05, + "loss": 0.3301, + "step": 35980 + }, + { + "epoch": 1.38, + "learning_rate": 2.696140035906643e-05, + "loss": 0.3098, + "step": 35981 + }, + { + "epoch": 1.38, + "learning_rate": 2.6960759169017695e-05, + "loss": 0.4632, + "step": 35982 + }, + { + "epoch": 1.38, + "learning_rate": 2.6960117978968967e-05, + "loss": 0.3298, + "step": 35983 + }, + { + "epoch": 1.38, + "learning_rate": 2.6959476788920235e-05, + "loss": 0.2921, + "step": 35984 + }, + { + "epoch": 1.38, + "learning_rate": 2.6958835598871507e-05, + "loss": 0.4498, + "step": 35985 + }, + { + "epoch": 1.38, + "learning_rate": 2.695819440882278e-05, + "loss": 0.3673, + "step": 35986 + }, + { + "epoch": 1.38, + "learning_rate": 2.6957553218774047e-05, + "loss": 0.3627, + "step": 35987 + }, + { + "epoch": 1.38, + "learning_rate": 2.695691202872532e-05, + "loss": 0.3329, + "step": 35988 + }, + { + "epoch": 1.38, + "learning_rate": 2.6956270838676584e-05, + "loss": 0.5636, + "step": 35989 + }, + { + "epoch": 1.38, + "learning_rate": 2.6955629648627855e-05, + "loss": 0.3332, + "step": 35990 + }, + { + "epoch": 1.38, + "learning_rate": 2.6954988458579124e-05, + "loss": 0.3321, + "step": 35991 + }, + { + "epoch": 1.38, + "learning_rate": 2.6954347268530395e-05, + "loss": 0.5171, + "step": 35992 + }, + { + "epoch": 1.38, + "learning_rate": 2.695370607848166e-05, + "loss": 0.2568, + "step": 35993 + }, + { + "epoch": 1.38, + "learning_rate": 2.6953064888432932e-05, + "loss": 0.2932, + "step": 35994 + }, + { + "epoch": 1.38, + "learning_rate": 2.6952423698384204e-05, + "loss": 0.3587, + "step": 35995 + }, + { + "epoch": 1.38, + "learning_rate": 2.6951782508335472e-05, + "loss": 0.3773, + "step": 35996 + }, + { + "epoch": 1.38, + "learning_rate": 2.6951141318286744e-05, + "loss": 0.3975, + "step": 35997 + }, + { + "epoch": 1.38, + "learning_rate": 2.695050012823801e-05, + "loss": 0.4126, + "step": 35998 + }, + { + "epoch": 1.38, + "learning_rate": 2.694985893818928e-05, + "loss": 0.3628, + "step": 35999 + }, + { + "epoch": 1.38, + "learning_rate": 2.694921774814055e-05, + "loss": 0.4008, + "step": 36000 + }, + { + "epoch": 1.38, + "learning_rate": 2.694857655809182e-05, + "loss": 0.349, + "step": 36001 + }, + { + "epoch": 1.38, + "learning_rate": 2.6947935368043086e-05, + "loss": 0.4738, + "step": 36002 + }, + { + "epoch": 1.38, + "learning_rate": 2.6947294177994357e-05, + "loss": 0.4716, + "step": 36003 + }, + { + "epoch": 1.38, + "learning_rate": 2.6946652987945632e-05, + "loss": 0.3341, + "step": 36004 + }, + { + "epoch": 1.38, + "learning_rate": 2.6946011797896897e-05, + "loss": 0.3862, + "step": 36005 + }, + { + "epoch": 1.38, + "learning_rate": 2.694537060784817e-05, + "loss": 0.3033, + "step": 36006 + }, + { + "epoch": 1.38, + "learning_rate": 2.6944729417799437e-05, + "loss": 0.2734, + "step": 36007 + }, + { + "epoch": 1.38, + "learning_rate": 2.694408822775071e-05, + "loss": 0.3119, + "step": 36008 + }, + { + "epoch": 1.38, + "learning_rate": 2.6943447037701974e-05, + "loss": 0.2936, + "step": 36009 + }, + { + "epoch": 1.38, + "learning_rate": 2.6942805847653246e-05, + "loss": 0.2543, + "step": 36010 + }, + { + "epoch": 1.38, + "learning_rate": 2.6942164657604514e-05, + "loss": 0.2384, + "step": 36011 + }, + { + "epoch": 1.38, + "learning_rate": 2.6941523467555786e-05, + "loss": 0.3893, + "step": 36012 + }, + { + "epoch": 1.38, + "learning_rate": 2.6940882277507057e-05, + "loss": 0.3246, + "step": 36013 + }, + { + "epoch": 1.38, + "learning_rate": 2.6940241087458322e-05, + "loss": 0.2827, + "step": 36014 + }, + { + "epoch": 1.38, + "learning_rate": 2.6939599897409594e-05, + "loss": 0.3075, + "step": 36015 + }, + { + "epoch": 1.38, + "learning_rate": 2.6938958707360862e-05, + "loss": 0.4238, + "step": 36016 + }, + { + "epoch": 1.38, + "learning_rate": 2.6938317517312134e-05, + "loss": 0.4767, + "step": 36017 + }, + { + "epoch": 1.38, + "learning_rate": 2.69376763272634e-05, + "loss": 0.4442, + "step": 36018 + }, + { + "epoch": 1.38, + "learning_rate": 2.693703513721467e-05, + "loss": 0.401, + "step": 36019 + }, + { + "epoch": 1.38, + "learning_rate": 2.693639394716594e-05, + "loss": 0.3316, + "step": 36020 + }, + { + "epoch": 1.38, + "learning_rate": 2.693575275711721e-05, + "loss": 0.4553, + "step": 36021 + }, + { + "epoch": 1.38, + "learning_rate": 2.6935111567068483e-05, + "loss": 0.4952, + "step": 36022 + }, + { + "epoch": 1.38, + "learning_rate": 2.693447037701975e-05, + "loss": 0.2315, + "step": 36023 + }, + { + "epoch": 1.39, + "learning_rate": 2.6933829186971023e-05, + "loss": 0.4827, + "step": 36024 + }, + { + "epoch": 1.39, + "learning_rate": 2.6933187996922288e-05, + "loss": 0.5218, + "step": 36025 + }, + { + "epoch": 1.39, + "learning_rate": 2.693254680687356e-05, + "loss": 0.2423, + "step": 36026 + }, + { + "epoch": 1.39, + "learning_rate": 2.6931905616824828e-05, + "loss": 0.3332, + "step": 36027 + }, + { + "epoch": 1.39, + "learning_rate": 2.69312644267761e-05, + "loss": 0.4563, + "step": 36028 + }, + { + "epoch": 1.39, + "learning_rate": 2.6930623236727364e-05, + "loss": 0.4814, + "step": 36029 + }, + { + "epoch": 1.39, + "learning_rate": 2.6929982046678636e-05, + "loss": 0.4899, + "step": 36030 + }, + { + "epoch": 1.39, + "learning_rate": 2.6929340856629908e-05, + "loss": 0.3624, + "step": 36031 + }, + { + "epoch": 1.39, + "learning_rate": 2.6928699666581176e-05, + "loss": 0.2947, + "step": 36032 + }, + { + "epoch": 1.39, + "learning_rate": 2.6928058476532448e-05, + "loss": 0.2261, + "step": 36033 + }, + { + "epoch": 1.39, + "learning_rate": 2.6927417286483713e-05, + "loss": 0.2931, + "step": 36034 + }, + { + "epoch": 1.39, + "learning_rate": 2.6926776096434984e-05, + "loss": 0.3451, + "step": 36035 + }, + { + "epoch": 1.39, + "learning_rate": 2.6926134906386253e-05, + "loss": 0.4395, + "step": 36036 + }, + { + "epoch": 1.39, + "learning_rate": 2.6925493716337524e-05, + "loss": 0.3501, + "step": 36037 + }, + { + "epoch": 1.39, + "learning_rate": 2.692485252628879e-05, + "loss": 0.4126, + "step": 36038 + }, + { + "epoch": 1.39, + "learning_rate": 2.692421133624006e-05, + "loss": 0.4213, + "step": 36039 + }, + { + "epoch": 1.39, + "learning_rate": 2.6923570146191336e-05, + "loss": 0.3815, + "step": 36040 + }, + { + "epoch": 1.39, + "learning_rate": 2.69229289561426e-05, + "loss": 0.3224, + "step": 36041 + }, + { + "epoch": 1.39, + "learning_rate": 2.6922287766093873e-05, + "loss": 0.4223, + "step": 36042 + }, + { + "epoch": 1.39, + "learning_rate": 2.692164657604514e-05, + "loss": 0.3845, + "step": 36043 + }, + { + "epoch": 1.39, + "learning_rate": 2.6921005385996413e-05, + "loss": 0.2967, + "step": 36044 + }, + { + "epoch": 1.39, + "learning_rate": 2.6920364195947678e-05, + "loss": 0.4078, + "step": 36045 + }, + { + "epoch": 1.39, + "learning_rate": 2.691972300589895e-05, + "loss": 0.3325, + "step": 36046 + }, + { + "epoch": 1.39, + "learning_rate": 2.6919081815850218e-05, + "loss": 0.3562, + "step": 36047 + }, + { + "epoch": 1.39, + "learning_rate": 2.691844062580149e-05, + "loss": 0.4483, + "step": 36048 + }, + { + "epoch": 1.39, + "learning_rate": 2.691779943575276e-05, + "loss": 0.3845, + "step": 36049 + }, + { + "epoch": 1.39, + "learning_rate": 2.6917158245704026e-05, + "loss": 0.3511, + "step": 36050 + }, + { + "epoch": 1.39, + "learning_rate": 2.6916517055655298e-05, + "loss": 0.2657, + "step": 36051 + }, + { + "epoch": 1.39, + "learning_rate": 2.6915875865606566e-05, + "loss": 0.5494, + "step": 36052 + }, + { + "epoch": 1.39, + "learning_rate": 2.6915234675557838e-05, + "loss": 0.3487, + "step": 36053 + }, + { + "epoch": 1.39, + "learning_rate": 2.6914593485509103e-05, + "loss": 0.5696, + "step": 36054 + }, + { + "epoch": 1.39, + "learning_rate": 2.6913952295460375e-05, + "loss": 0.3518, + "step": 36055 + }, + { + "epoch": 1.39, + "learning_rate": 2.6913311105411643e-05, + "loss": 0.4484, + "step": 36056 + }, + { + "epoch": 1.39, + "learning_rate": 2.6912669915362915e-05, + "loss": 0.2659, + "step": 36057 + }, + { + "epoch": 1.39, + "learning_rate": 2.6912028725314186e-05, + "loss": 0.3178, + "step": 36058 + }, + { + "epoch": 1.39, + "learning_rate": 2.6911387535265455e-05, + "loss": 0.272, + "step": 36059 + }, + { + "epoch": 1.39, + "learning_rate": 2.6910746345216726e-05, + "loss": 0.3685, + "step": 36060 + }, + { + "epoch": 1.39, + "learning_rate": 2.691010515516799e-05, + "loss": 0.2667, + "step": 36061 + }, + { + "epoch": 1.39, + "learning_rate": 2.6909463965119263e-05, + "loss": 0.2859, + "step": 36062 + }, + { + "epoch": 1.39, + "learning_rate": 2.690882277507053e-05, + "loss": 0.3727, + "step": 36063 + }, + { + "epoch": 1.39, + "learning_rate": 2.6908181585021803e-05, + "loss": 0.3897, + "step": 36064 + }, + { + "epoch": 1.39, + "learning_rate": 2.6907540394973068e-05, + "loss": 0.37, + "step": 36065 + }, + { + "epoch": 1.39, + "learning_rate": 2.690689920492434e-05, + "loss": 0.4219, + "step": 36066 + }, + { + "epoch": 1.39, + "learning_rate": 2.690625801487561e-05, + "loss": 0.4764, + "step": 36067 + }, + { + "epoch": 1.39, + "learning_rate": 2.690561682482688e-05, + "loss": 0.5023, + "step": 36068 + }, + { + "epoch": 1.39, + "learning_rate": 2.690497563477815e-05, + "loss": 0.4429, + "step": 36069 + }, + { + "epoch": 1.39, + "learning_rate": 2.6904334444729417e-05, + "loss": 0.3023, + "step": 36070 + }, + { + "epoch": 1.39, + "learning_rate": 2.6903693254680688e-05, + "loss": 0.3645, + "step": 36071 + }, + { + "epoch": 1.39, + "learning_rate": 2.6903052064631957e-05, + "loss": 0.2791, + "step": 36072 + }, + { + "epoch": 1.39, + "learning_rate": 2.690241087458323e-05, + "loss": 0.4381, + "step": 36073 + }, + { + "epoch": 1.39, + "learning_rate": 2.6901769684534493e-05, + "loss": 0.3826, + "step": 36074 + }, + { + "epoch": 1.39, + "learning_rate": 2.6901128494485765e-05, + "loss": 0.3527, + "step": 36075 + }, + { + "epoch": 1.39, + "learning_rate": 2.690048730443704e-05, + "loss": 0.3795, + "step": 36076 + }, + { + "epoch": 1.39, + "learning_rate": 2.6899846114388305e-05, + "loss": 0.3407, + "step": 36077 + }, + { + "epoch": 1.39, + "learning_rate": 2.6899204924339577e-05, + "loss": 0.2726, + "step": 36078 + }, + { + "epoch": 1.39, + "learning_rate": 2.6898563734290845e-05, + "loss": 0.3667, + "step": 36079 + }, + { + "epoch": 1.39, + "learning_rate": 2.6897922544242117e-05, + "loss": 0.3677, + "step": 36080 + }, + { + "epoch": 1.39, + "learning_rate": 2.6897281354193382e-05, + "loss": 0.2528, + "step": 36081 + }, + { + "epoch": 1.39, + "learning_rate": 2.6896640164144653e-05, + "loss": 0.2895, + "step": 36082 + }, + { + "epoch": 1.39, + "learning_rate": 2.6895998974095922e-05, + "loss": 0.3201, + "step": 36083 + }, + { + "epoch": 1.39, + "learning_rate": 2.6895357784047193e-05, + "loss": 0.4606, + "step": 36084 + }, + { + "epoch": 1.39, + "learning_rate": 2.6894716593998465e-05, + "loss": 0.4139, + "step": 36085 + }, + { + "epoch": 1.39, + "learning_rate": 2.689407540394973e-05, + "loss": 0.3013, + "step": 36086 + }, + { + "epoch": 1.39, + "learning_rate": 2.6893434213901002e-05, + "loss": 0.3623, + "step": 36087 + }, + { + "epoch": 1.39, + "learning_rate": 2.689279302385227e-05, + "loss": 0.301, + "step": 36088 + }, + { + "epoch": 1.39, + "learning_rate": 2.6892151833803542e-05, + "loss": 0.3353, + "step": 36089 + }, + { + "epoch": 1.39, + "learning_rate": 2.6891510643754807e-05, + "loss": 0.296, + "step": 36090 + }, + { + "epoch": 1.39, + "learning_rate": 2.689086945370608e-05, + "loss": 0.3033, + "step": 36091 + }, + { + "epoch": 1.39, + "learning_rate": 2.6890228263657347e-05, + "loss": 0.2672, + "step": 36092 + }, + { + "epoch": 1.39, + "learning_rate": 2.688958707360862e-05, + "loss": 0.2263, + "step": 36093 + }, + { + "epoch": 1.39, + "learning_rate": 2.688894588355989e-05, + "loss": 0.4515, + "step": 36094 + }, + { + "epoch": 1.39, + "learning_rate": 2.688830469351116e-05, + "loss": 0.4591, + "step": 36095 + }, + { + "epoch": 1.39, + "learning_rate": 2.688766350346243e-05, + "loss": 0.4209, + "step": 36096 + }, + { + "epoch": 1.39, + "learning_rate": 2.6887022313413695e-05, + "loss": 0.4139, + "step": 36097 + }, + { + "epoch": 1.39, + "learning_rate": 2.6886381123364967e-05, + "loss": 0.4943, + "step": 36098 + }, + { + "epoch": 1.39, + "learning_rate": 2.6885739933316235e-05, + "loss": 0.3672, + "step": 36099 + }, + { + "epoch": 1.39, + "learning_rate": 2.6885098743267507e-05, + "loss": 0.4228, + "step": 36100 + }, + { + "epoch": 1.39, + "learning_rate": 2.6884457553218772e-05, + "loss": 0.3204, + "step": 36101 + }, + { + "epoch": 1.39, + "learning_rate": 2.6883816363170044e-05, + "loss": 0.3863, + "step": 36102 + }, + { + "epoch": 1.39, + "learning_rate": 2.6883175173121315e-05, + "loss": 0.2428, + "step": 36103 + }, + { + "epoch": 1.39, + "learning_rate": 2.6882533983072584e-05, + "loss": 0.3505, + "step": 36104 + }, + { + "epoch": 1.39, + "learning_rate": 2.6881892793023855e-05, + "loss": 0.2794, + "step": 36105 + }, + { + "epoch": 1.39, + "learning_rate": 2.688125160297512e-05, + "loss": 0.4795, + "step": 36106 + }, + { + "epoch": 1.39, + "learning_rate": 2.6880610412926392e-05, + "loss": 0.2974, + "step": 36107 + }, + { + "epoch": 1.39, + "learning_rate": 2.687996922287766e-05, + "loss": 0.3908, + "step": 36108 + }, + { + "epoch": 1.39, + "learning_rate": 2.6879328032828932e-05, + "loss": 0.3358, + "step": 36109 + }, + { + "epoch": 1.39, + "learning_rate": 2.6878686842780204e-05, + "loss": 0.3492, + "step": 36110 + }, + { + "epoch": 1.39, + "learning_rate": 2.6878045652731472e-05, + "loss": 0.2148, + "step": 36111 + }, + { + "epoch": 1.39, + "learning_rate": 2.6877404462682744e-05, + "loss": 0.4028, + "step": 36112 + }, + { + "epoch": 1.39, + "learning_rate": 2.687676327263401e-05, + "loss": 0.3476, + "step": 36113 + }, + { + "epoch": 1.39, + "learning_rate": 2.687612208258528e-05, + "loss": 0.2475, + "step": 36114 + }, + { + "epoch": 1.39, + "learning_rate": 2.687548089253655e-05, + "loss": 0.4247, + "step": 36115 + }, + { + "epoch": 1.39, + "learning_rate": 2.687483970248782e-05, + "loss": 0.4665, + "step": 36116 + }, + { + "epoch": 1.39, + "learning_rate": 2.6874198512439086e-05, + "loss": 0.3136, + "step": 36117 + }, + { + "epoch": 1.39, + "learning_rate": 2.6873557322390357e-05, + "loss": 0.3808, + "step": 36118 + }, + { + "epoch": 1.39, + "learning_rate": 2.687291613234163e-05, + "loss": 0.3599, + "step": 36119 + }, + { + "epoch": 1.39, + "learning_rate": 2.6872274942292897e-05, + "loss": 0.2945, + "step": 36120 + }, + { + "epoch": 1.39, + "learning_rate": 2.687163375224417e-05, + "loss": 0.2831, + "step": 36121 + }, + { + "epoch": 1.39, + "learning_rate": 2.6870992562195434e-05, + "loss": 0.3864, + "step": 36122 + }, + { + "epoch": 1.39, + "learning_rate": 2.6870351372146706e-05, + "loss": 0.4946, + "step": 36123 + }, + { + "epoch": 1.39, + "learning_rate": 2.6869710182097974e-05, + "loss": 0.4744, + "step": 36124 + }, + { + "epoch": 1.39, + "learning_rate": 2.6869068992049246e-05, + "loss": 0.3479, + "step": 36125 + }, + { + "epoch": 1.39, + "learning_rate": 2.686842780200051e-05, + "loss": 0.398, + "step": 36126 + }, + { + "epoch": 1.39, + "learning_rate": 2.6867786611951782e-05, + "loss": 0.3001, + "step": 36127 + }, + { + "epoch": 1.39, + "learning_rate": 2.6867145421903058e-05, + "loss": 0.3547, + "step": 36128 + }, + { + "epoch": 1.39, + "learning_rate": 2.6866504231854322e-05, + "loss": 0.3277, + "step": 36129 + }, + { + "epoch": 1.39, + "learning_rate": 2.6865863041805594e-05, + "loss": 0.4219, + "step": 36130 + }, + { + "epoch": 1.39, + "learning_rate": 2.6865221851756863e-05, + "loss": 0.3399, + "step": 36131 + }, + { + "epoch": 1.39, + "learning_rate": 2.6864580661708134e-05, + "loss": 0.5268, + "step": 36132 + }, + { + "epoch": 1.39, + "learning_rate": 2.68639394716594e-05, + "loss": 0.3722, + "step": 36133 + }, + { + "epoch": 1.39, + "learning_rate": 2.686329828161067e-05, + "loss": 0.3469, + "step": 36134 + }, + { + "epoch": 1.39, + "learning_rate": 2.686265709156194e-05, + "loss": 0.3295, + "step": 36135 + }, + { + "epoch": 1.39, + "learning_rate": 2.686201590151321e-05, + "loss": 0.4181, + "step": 36136 + }, + { + "epoch": 1.39, + "learning_rate": 2.6861374711464483e-05, + "loss": 0.4963, + "step": 36137 + }, + { + "epoch": 1.39, + "learning_rate": 2.6860733521415748e-05, + "loss": 0.3477, + "step": 36138 + }, + { + "epoch": 1.39, + "learning_rate": 2.686009233136702e-05, + "loss": 0.4395, + "step": 36139 + }, + { + "epoch": 1.39, + "learning_rate": 2.6859451141318288e-05, + "loss": 0.5648, + "step": 36140 + }, + { + "epoch": 1.39, + "learning_rate": 2.685880995126956e-05, + "loss": 0.3248, + "step": 36141 + }, + { + "epoch": 1.39, + "learning_rate": 2.6858168761220824e-05, + "loss": 0.3253, + "step": 36142 + }, + { + "epoch": 1.39, + "learning_rate": 2.6857527571172096e-05, + "loss": 0.2756, + "step": 36143 + }, + { + "epoch": 1.39, + "learning_rate": 2.6856886381123364e-05, + "loss": 0.3416, + "step": 36144 + }, + { + "epoch": 1.39, + "learning_rate": 2.6856245191074636e-05, + "loss": 0.3755, + "step": 36145 + }, + { + "epoch": 1.39, + "learning_rate": 2.6855604001025908e-05, + "loss": 0.3302, + "step": 36146 + }, + { + "epoch": 1.39, + "learning_rate": 2.6854962810977176e-05, + "loss": 0.504, + "step": 36147 + }, + { + "epoch": 1.39, + "learning_rate": 2.6854321620928448e-05, + "loss": 0.312, + "step": 36148 + }, + { + "epoch": 1.39, + "learning_rate": 2.6853680430879713e-05, + "loss": 0.3502, + "step": 36149 + }, + { + "epoch": 1.39, + "learning_rate": 2.6853039240830984e-05, + "loss": 0.3811, + "step": 36150 + }, + { + "epoch": 1.39, + "learning_rate": 2.6852398050782253e-05, + "loss": 0.403, + "step": 36151 + }, + { + "epoch": 1.39, + "learning_rate": 2.6851756860733525e-05, + "loss": 0.227, + "step": 36152 + }, + { + "epoch": 1.39, + "learning_rate": 2.685111567068479e-05, + "loss": 0.377, + "step": 36153 + }, + { + "epoch": 1.39, + "learning_rate": 2.685047448063606e-05, + "loss": 0.3783, + "step": 36154 + }, + { + "epoch": 1.39, + "learning_rate": 2.6849833290587333e-05, + "loss": 0.3678, + "step": 36155 + }, + { + "epoch": 1.39, + "learning_rate": 2.68491921005386e-05, + "loss": 0.326, + "step": 36156 + }, + { + "epoch": 1.39, + "learning_rate": 2.6848550910489873e-05, + "loss": 0.3977, + "step": 36157 + }, + { + "epoch": 1.39, + "learning_rate": 2.6847909720441138e-05, + "loss": 0.3546, + "step": 36158 + }, + { + "epoch": 1.39, + "learning_rate": 2.684726853039241e-05, + "loss": 0.2969, + "step": 36159 + }, + { + "epoch": 1.39, + "learning_rate": 2.6846627340343678e-05, + "loss": 0.3685, + "step": 36160 + }, + { + "epoch": 1.39, + "learning_rate": 2.684598615029495e-05, + "loss": 0.285, + "step": 36161 + }, + { + "epoch": 1.39, + "learning_rate": 2.6845344960246215e-05, + "loss": 0.4808, + "step": 36162 + }, + { + "epoch": 1.39, + "learning_rate": 2.6844703770197486e-05, + "loss": 0.2742, + "step": 36163 + }, + { + "epoch": 1.39, + "learning_rate": 2.684406258014876e-05, + "loss": 0.3202, + "step": 36164 + }, + { + "epoch": 1.39, + "learning_rate": 2.6843421390100026e-05, + "loss": 0.4505, + "step": 36165 + }, + { + "epoch": 1.39, + "learning_rate": 2.6842780200051298e-05, + "loss": 0.2551, + "step": 36166 + }, + { + "epoch": 1.39, + "learning_rate": 2.6842139010002566e-05, + "loss": 0.3602, + "step": 36167 + }, + { + "epoch": 1.39, + "learning_rate": 2.6841497819953838e-05, + "loss": 0.4607, + "step": 36168 + }, + { + "epoch": 1.39, + "learning_rate": 2.6840856629905103e-05, + "loss": 0.3233, + "step": 36169 + }, + { + "epoch": 1.39, + "learning_rate": 2.6840215439856375e-05, + "loss": 0.3544, + "step": 36170 + }, + { + "epoch": 1.39, + "learning_rate": 2.6839574249807643e-05, + "loss": 0.3274, + "step": 36171 + }, + { + "epoch": 1.39, + "learning_rate": 2.6838933059758915e-05, + "loss": 0.395, + "step": 36172 + }, + { + "epoch": 1.39, + "learning_rate": 2.6838291869710187e-05, + "loss": 0.3865, + "step": 36173 + }, + { + "epoch": 1.39, + "learning_rate": 2.683765067966145e-05, + "loss": 0.4243, + "step": 36174 + }, + { + "epoch": 1.39, + "learning_rate": 2.6837009489612723e-05, + "loss": 0.2908, + "step": 36175 + }, + { + "epoch": 1.39, + "learning_rate": 2.683636829956399e-05, + "loss": 0.3588, + "step": 36176 + }, + { + "epoch": 1.39, + "learning_rate": 2.6835727109515263e-05, + "loss": 0.3677, + "step": 36177 + }, + { + "epoch": 1.39, + "learning_rate": 2.6835085919466528e-05, + "loss": 0.4077, + "step": 36178 + }, + { + "epoch": 1.39, + "learning_rate": 2.68344447294178e-05, + "loss": 0.3125, + "step": 36179 + }, + { + "epoch": 1.39, + "learning_rate": 2.6833803539369068e-05, + "loss": 0.4202, + "step": 36180 + }, + { + "epoch": 1.39, + "learning_rate": 2.683316234932034e-05, + "loss": 0.4957, + "step": 36181 + }, + { + "epoch": 1.39, + "learning_rate": 2.683252115927161e-05, + "loss": 0.4498, + "step": 36182 + }, + { + "epoch": 1.39, + "learning_rate": 2.683187996922288e-05, + "loss": 0.3995, + "step": 36183 + }, + { + "epoch": 1.39, + "learning_rate": 2.6831238779174152e-05, + "loss": 0.2838, + "step": 36184 + }, + { + "epoch": 1.39, + "learning_rate": 2.6830597589125417e-05, + "loss": 0.3862, + "step": 36185 + }, + { + "epoch": 1.39, + "learning_rate": 2.682995639907669e-05, + "loss": 0.3293, + "step": 36186 + }, + { + "epoch": 1.39, + "learning_rate": 2.6829315209027957e-05, + "loss": 0.4866, + "step": 36187 + }, + { + "epoch": 1.39, + "learning_rate": 2.682867401897923e-05, + "loss": 0.4167, + "step": 36188 + }, + { + "epoch": 1.39, + "learning_rate": 2.6828032828930493e-05, + "loss": 0.4513, + "step": 36189 + }, + { + "epoch": 1.39, + "learning_rate": 2.6827391638881765e-05, + "loss": 0.3449, + "step": 36190 + }, + { + "epoch": 1.39, + "learning_rate": 2.6826750448833037e-05, + "loss": 0.4933, + "step": 36191 + }, + { + "epoch": 1.39, + "learning_rate": 2.6826109258784305e-05, + "loss": 0.362, + "step": 36192 + }, + { + "epoch": 1.39, + "learning_rate": 2.6825468068735577e-05, + "loss": 0.3612, + "step": 36193 + }, + { + "epoch": 1.39, + "learning_rate": 2.6824826878686842e-05, + "loss": 0.3452, + "step": 36194 + }, + { + "epoch": 1.39, + "learning_rate": 2.6824185688638113e-05, + "loss": 0.3926, + "step": 36195 + }, + { + "epoch": 1.39, + "learning_rate": 2.6823544498589382e-05, + "loss": 0.4224, + "step": 36196 + }, + { + "epoch": 1.39, + "learning_rate": 2.6822903308540654e-05, + "loss": 0.4487, + "step": 36197 + }, + { + "epoch": 1.39, + "learning_rate": 2.682226211849192e-05, + "loss": 0.373, + "step": 36198 + }, + { + "epoch": 1.39, + "learning_rate": 2.682162092844319e-05, + "loss": 0.4915, + "step": 36199 + }, + { + "epoch": 1.39, + "learning_rate": 2.6820979738394465e-05, + "loss": 0.3712, + "step": 36200 + }, + { + "epoch": 1.39, + "learning_rate": 2.682033854834573e-05, + "loss": 0.4759, + "step": 36201 + }, + { + "epoch": 1.39, + "learning_rate": 2.6819697358297002e-05, + "loss": 0.3905, + "step": 36202 + }, + { + "epoch": 1.39, + "learning_rate": 2.681905616824827e-05, + "loss": 0.4463, + "step": 36203 + }, + { + "epoch": 1.39, + "learning_rate": 2.6818414978199542e-05, + "loss": 0.4808, + "step": 36204 + }, + { + "epoch": 1.39, + "learning_rate": 2.6817773788150807e-05, + "loss": 0.3997, + "step": 36205 + }, + { + "epoch": 1.39, + "learning_rate": 2.681713259810208e-05, + "loss": 0.4441, + "step": 36206 + }, + { + "epoch": 1.39, + "learning_rate": 2.6816491408053347e-05, + "loss": 0.4264, + "step": 36207 + }, + { + "epoch": 1.39, + "learning_rate": 2.681585021800462e-05, + "loss": 0.3106, + "step": 36208 + }, + { + "epoch": 1.39, + "learning_rate": 2.681520902795589e-05, + "loss": 0.3379, + "step": 36209 + }, + { + "epoch": 1.39, + "learning_rate": 2.6814567837907155e-05, + "loss": 0.5245, + "step": 36210 + }, + { + "epoch": 1.39, + "learning_rate": 2.6813926647858427e-05, + "loss": 0.3961, + "step": 36211 + }, + { + "epoch": 1.39, + "learning_rate": 2.6813285457809695e-05, + "loss": 0.422, + "step": 36212 + }, + { + "epoch": 1.39, + "learning_rate": 2.6812644267760967e-05, + "loss": 0.358, + "step": 36213 + }, + { + "epoch": 1.39, + "learning_rate": 2.6812003077712232e-05, + "loss": 0.4005, + "step": 36214 + }, + { + "epoch": 1.39, + "learning_rate": 2.6811361887663504e-05, + "loss": 0.3729, + "step": 36215 + }, + { + "epoch": 1.39, + "learning_rate": 2.6810720697614772e-05, + "loss": 0.4385, + "step": 36216 + }, + { + "epoch": 1.39, + "learning_rate": 2.6810079507566044e-05, + "loss": 0.375, + "step": 36217 + }, + { + "epoch": 1.39, + "learning_rate": 2.6809438317517316e-05, + "loss": 0.3631, + "step": 36218 + }, + { + "epoch": 1.39, + "learning_rate": 2.6808797127468584e-05, + "loss": 0.3994, + "step": 36219 + }, + { + "epoch": 1.39, + "learning_rate": 2.6808155937419856e-05, + "loss": 0.3072, + "step": 36220 + }, + { + "epoch": 1.39, + "learning_rate": 2.680751474737112e-05, + "loss": 0.4077, + "step": 36221 + }, + { + "epoch": 1.39, + "learning_rate": 2.6806873557322392e-05, + "loss": 0.2946, + "step": 36222 + }, + { + "epoch": 1.39, + "learning_rate": 2.680623236727366e-05, + "loss": 0.2504, + "step": 36223 + }, + { + "epoch": 1.39, + "learning_rate": 2.6805591177224932e-05, + "loss": 0.4887, + "step": 36224 + }, + { + "epoch": 1.39, + "learning_rate": 2.6804949987176197e-05, + "loss": 0.2565, + "step": 36225 + }, + { + "epoch": 1.39, + "learning_rate": 2.680430879712747e-05, + "loss": 0.5334, + "step": 36226 + }, + { + "epoch": 1.39, + "learning_rate": 2.680366760707874e-05, + "loss": 0.3901, + "step": 36227 + }, + { + "epoch": 1.39, + "learning_rate": 2.680302641703001e-05, + "loss": 0.3392, + "step": 36228 + }, + { + "epoch": 1.39, + "learning_rate": 2.680238522698128e-05, + "loss": 0.3794, + "step": 36229 + }, + { + "epoch": 1.39, + "learning_rate": 2.6801744036932546e-05, + "loss": 0.4383, + "step": 36230 + }, + { + "epoch": 1.39, + "learning_rate": 2.6801102846883817e-05, + "loss": 0.3565, + "step": 36231 + }, + { + "epoch": 1.39, + "learning_rate": 2.6800461656835086e-05, + "loss": 0.327, + "step": 36232 + }, + { + "epoch": 1.39, + "learning_rate": 2.6799820466786357e-05, + "loss": 0.3231, + "step": 36233 + }, + { + "epoch": 1.39, + "learning_rate": 2.6799179276737622e-05, + "loss": 0.3016, + "step": 36234 + }, + { + "epoch": 1.39, + "learning_rate": 2.6798538086688897e-05, + "loss": 0.3405, + "step": 36235 + }, + { + "epoch": 1.39, + "learning_rate": 2.679789689664017e-05, + "loss": 0.3409, + "step": 36236 + }, + { + "epoch": 1.39, + "learning_rate": 2.6797255706591434e-05, + "loss": 0.3531, + "step": 36237 + }, + { + "epoch": 1.39, + "learning_rate": 2.6796614516542706e-05, + "loss": 0.3903, + "step": 36238 + }, + { + "epoch": 1.39, + "learning_rate": 2.6795973326493974e-05, + "loss": 0.3847, + "step": 36239 + }, + { + "epoch": 1.39, + "learning_rate": 2.6795332136445246e-05, + "loss": 0.3301, + "step": 36240 + }, + { + "epoch": 1.39, + "learning_rate": 2.679469094639651e-05, + "loss": 0.3354, + "step": 36241 + }, + { + "epoch": 1.39, + "learning_rate": 2.6794049756347783e-05, + "loss": 0.4147, + "step": 36242 + }, + { + "epoch": 1.39, + "learning_rate": 2.679340856629905e-05, + "loss": 0.3237, + "step": 36243 + }, + { + "epoch": 1.39, + "learning_rate": 2.6792767376250323e-05, + "loss": 0.4366, + "step": 36244 + }, + { + "epoch": 1.39, + "learning_rate": 2.6792126186201594e-05, + "loss": 0.3322, + "step": 36245 + }, + { + "epoch": 1.39, + "learning_rate": 2.679148499615286e-05, + "loss": 0.565, + "step": 36246 + }, + { + "epoch": 1.39, + "learning_rate": 2.679084380610413e-05, + "loss": 0.3114, + "step": 36247 + }, + { + "epoch": 1.39, + "learning_rate": 2.67902026160554e-05, + "loss": 0.3253, + "step": 36248 + }, + { + "epoch": 1.39, + "learning_rate": 2.678956142600667e-05, + "loss": 0.4005, + "step": 36249 + }, + { + "epoch": 1.39, + "learning_rate": 2.6788920235957936e-05, + "loss": 0.2981, + "step": 36250 + }, + { + "epoch": 1.39, + "learning_rate": 2.6788279045909208e-05, + "loss": 0.3642, + "step": 36251 + }, + { + "epoch": 1.39, + "learning_rate": 2.6787637855860476e-05, + "loss": 0.51, + "step": 36252 + }, + { + "epoch": 1.39, + "learning_rate": 2.6786996665811748e-05, + "loss": 0.2779, + "step": 36253 + }, + { + "epoch": 1.39, + "learning_rate": 2.678635547576302e-05, + "loss": 0.3848, + "step": 36254 + }, + { + "epoch": 1.39, + "learning_rate": 2.6785714285714288e-05, + "loss": 0.3013, + "step": 36255 + }, + { + "epoch": 1.39, + "learning_rate": 2.678507309566556e-05, + "loss": 0.405, + "step": 36256 + }, + { + "epoch": 1.39, + "learning_rate": 2.6784431905616824e-05, + "loss": 0.3665, + "step": 36257 + }, + { + "epoch": 1.39, + "learning_rate": 2.6783790715568096e-05, + "loss": 0.4248, + "step": 36258 + }, + { + "epoch": 1.39, + "learning_rate": 2.6783149525519364e-05, + "loss": 0.3253, + "step": 36259 + }, + { + "epoch": 1.39, + "learning_rate": 2.6782508335470636e-05, + "loss": 0.4394, + "step": 36260 + }, + { + "epoch": 1.39, + "learning_rate": 2.67818671454219e-05, + "loss": 0.3296, + "step": 36261 + }, + { + "epoch": 1.39, + "learning_rate": 2.6781225955373173e-05, + "loss": 0.3273, + "step": 36262 + }, + { + "epoch": 1.39, + "learning_rate": 2.6780584765324445e-05, + "loss": 0.3705, + "step": 36263 + }, + { + "epoch": 1.39, + "learning_rate": 2.6779943575275713e-05, + "loss": 0.4172, + "step": 36264 + }, + { + "epoch": 1.39, + "learning_rate": 2.6779302385226985e-05, + "loss": 0.3918, + "step": 36265 + }, + { + "epoch": 1.39, + "learning_rate": 2.677866119517825e-05, + "loss": 0.3708, + "step": 36266 + }, + { + "epoch": 1.39, + "learning_rate": 2.677802000512952e-05, + "loss": 0.3248, + "step": 36267 + }, + { + "epoch": 1.39, + "learning_rate": 2.677737881508079e-05, + "loss": 0.3598, + "step": 36268 + }, + { + "epoch": 1.39, + "learning_rate": 2.677673762503206e-05, + "loss": 0.3638, + "step": 36269 + }, + { + "epoch": 1.39, + "learning_rate": 2.6776096434983326e-05, + "loss": 0.3738, + "step": 36270 + }, + { + "epoch": 1.39, + "learning_rate": 2.67754552449346e-05, + "loss": 0.4916, + "step": 36271 + }, + { + "epoch": 1.39, + "learning_rate": 2.6774814054885873e-05, + "loss": 0.3333, + "step": 36272 + }, + { + "epoch": 1.39, + "learning_rate": 2.6774172864837138e-05, + "loss": 0.4819, + "step": 36273 + }, + { + "epoch": 1.39, + "learning_rate": 2.677353167478841e-05, + "loss": 0.3341, + "step": 36274 + }, + { + "epoch": 1.39, + "learning_rate": 2.6772890484739678e-05, + "loss": 0.295, + "step": 36275 + }, + { + "epoch": 1.39, + "learning_rate": 2.677224929469095e-05, + "loss": 0.2672, + "step": 36276 + }, + { + "epoch": 1.39, + "learning_rate": 2.6771608104642215e-05, + "loss": 0.2962, + "step": 36277 + }, + { + "epoch": 1.39, + "learning_rate": 2.6770966914593486e-05, + "loss": 0.502, + "step": 36278 + }, + { + "epoch": 1.39, + "learning_rate": 2.6770325724544755e-05, + "loss": 0.2916, + "step": 36279 + }, + { + "epoch": 1.39, + "learning_rate": 2.6769684534496026e-05, + "loss": 0.4753, + "step": 36280 + }, + { + "epoch": 1.39, + "learning_rate": 2.6769043344447298e-05, + "loss": 0.3875, + "step": 36281 + }, + { + "epoch": 1.39, + "learning_rate": 2.6768402154398563e-05, + "loss": 0.3997, + "step": 36282 + }, + { + "epoch": 1.39, + "learning_rate": 2.6767760964349835e-05, + "loss": 0.3428, + "step": 36283 + }, + { + "epoch": 1.4, + "learning_rate": 2.6767119774301103e-05, + "loss": 0.3112, + "step": 36284 + }, + { + "epoch": 1.4, + "learning_rate": 2.6766478584252375e-05, + "loss": 0.4046, + "step": 36285 + }, + { + "epoch": 1.4, + "learning_rate": 2.676583739420364e-05, + "loss": 0.4198, + "step": 36286 + }, + { + "epoch": 1.4, + "learning_rate": 2.676519620415491e-05, + "loss": 0.3809, + "step": 36287 + }, + { + "epoch": 1.4, + "learning_rate": 2.676455501410618e-05, + "loss": 0.3327, + "step": 36288 + }, + { + "epoch": 1.4, + "learning_rate": 2.676391382405745e-05, + "loss": 0.4169, + "step": 36289 + }, + { + "epoch": 1.4, + "learning_rate": 2.6763272634008723e-05, + "loss": 0.3485, + "step": 36290 + }, + { + "epoch": 1.4, + "learning_rate": 2.676263144395999e-05, + "loss": 0.3155, + "step": 36291 + }, + { + "epoch": 1.4, + "learning_rate": 2.6761990253911263e-05, + "loss": 0.5249, + "step": 36292 + }, + { + "epoch": 1.4, + "learning_rate": 2.6761349063862528e-05, + "loss": 0.3547, + "step": 36293 + }, + { + "epoch": 1.4, + "learning_rate": 2.67607078738138e-05, + "loss": 0.4031, + "step": 36294 + }, + { + "epoch": 1.4, + "learning_rate": 2.6760066683765068e-05, + "loss": 0.2737, + "step": 36295 + }, + { + "epoch": 1.4, + "learning_rate": 2.675942549371634e-05, + "loss": 0.4268, + "step": 36296 + }, + { + "epoch": 1.4, + "learning_rate": 2.6758784303667612e-05, + "loss": 0.3789, + "step": 36297 + }, + { + "epoch": 1.4, + "learning_rate": 2.6758143113618877e-05, + "loss": 0.3583, + "step": 36298 + }, + { + "epoch": 1.4, + "learning_rate": 2.675750192357015e-05, + "loss": 0.4457, + "step": 36299 + }, + { + "epoch": 1.4, + "learning_rate": 2.6756860733521417e-05, + "loss": 0.3659, + "step": 36300 + }, + { + "epoch": 1.4, + "learning_rate": 2.675621954347269e-05, + "loss": 0.4017, + "step": 36301 + }, + { + "epoch": 1.4, + "learning_rate": 2.6755578353423953e-05, + "loss": 0.3541, + "step": 36302 + }, + { + "epoch": 1.4, + "learning_rate": 2.6754937163375225e-05, + "loss": 0.2746, + "step": 36303 + }, + { + "epoch": 1.4, + "learning_rate": 2.6754295973326493e-05, + "loss": 0.3275, + "step": 36304 + }, + { + "epoch": 1.4, + "learning_rate": 2.6753654783277765e-05, + "loss": 0.4018, + "step": 36305 + }, + { + "epoch": 1.4, + "learning_rate": 2.6753013593229037e-05, + "loss": 0.3752, + "step": 36306 + }, + { + "epoch": 1.4, + "learning_rate": 2.6752372403180305e-05, + "loss": 0.3043, + "step": 36307 + }, + { + "epoch": 1.4, + "learning_rate": 2.6751731213131577e-05, + "loss": 0.2818, + "step": 36308 + }, + { + "epoch": 1.4, + "learning_rate": 2.6751090023082842e-05, + "loss": 0.3879, + "step": 36309 + }, + { + "epoch": 1.4, + "learning_rate": 2.6750448833034114e-05, + "loss": 0.2966, + "step": 36310 + }, + { + "epoch": 1.4, + "learning_rate": 2.6749807642985382e-05, + "loss": 0.2982, + "step": 36311 + }, + { + "epoch": 1.4, + "learning_rate": 2.6749166452936654e-05, + "loss": 0.347, + "step": 36312 + }, + { + "epoch": 1.4, + "learning_rate": 2.674852526288792e-05, + "loss": 0.4461, + "step": 36313 + }, + { + "epoch": 1.4, + "learning_rate": 2.674788407283919e-05, + "loss": 0.3631, + "step": 36314 + }, + { + "epoch": 1.4, + "learning_rate": 2.6747242882790462e-05, + "loss": 0.3063, + "step": 36315 + }, + { + "epoch": 1.4, + "learning_rate": 2.674660169274173e-05, + "loss": 0.3549, + "step": 36316 + }, + { + "epoch": 1.4, + "learning_rate": 2.6745960502693002e-05, + "loss": 0.4015, + "step": 36317 + }, + { + "epoch": 1.4, + "learning_rate": 2.6745319312644267e-05, + "loss": 0.2281, + "step": 36318 + }, + { + "epoch": 1.4, + "learning_rate": 2.674467812259554e-05, + "loss": 0.4114, + "step": 36319 + }, + { + "epoch": 1.4, + "learning_rate": 2.6744036932546807e-05, + "loss": 0.5378, + "step": 36320 + }, + { + "epoch": 1.4, + "learning_rate": 2.674339574249808e-05, + "loss": 0.5105, + "step": 36321 + }, + { + "epoch": 1.4, + "learning_rate": 2.6742754552449344e-05, + "loss": 0.4018, + "step": 36322 + }, + { + "epoch": 1.4, + "learning_rate": 2.6742113362400615e-05, + "loss": 0.4255, + "step": 36323 + }, + { + "epoch": 1.4, + "learning_rate": 2.674147217235189e-05, + "loss": 0.3924, + "step": 36324 + }, + { + "epoch": 1.4, + "learning_rate": 2.6740830982303155e-05, + "loss": 0.331, + "step": 36325 + }, + { + "epoch": 1.4, + "learning_rate": 2.6740189792254427e-05, + "loss": 0.4664, + "step": 36326 + }, + { + "epoch": 1.4, + "learning_rate": 2.6739548602205695e-05, + "loss": 0.3496, + "step": 36327 + }, + { + "epoch": 1.4, + "learning_rate": 2.6738907412156967e-05, + "loss": 0.4112, + "step": 36328 + }, + { + "epoch": 1.4, + "learning_rate": 2.6738266222108232e-05, + "loss": 0.397, + "step": 36329 + }, + { + "epoch": 1.4, + "learning_rate": 2.6737625032059504e-05, + "loss": 0.4277, + "step": 36330 + }, + { + "epoch": 1.4, + "learning_rate": 2.6736983842010772e-05, + "loss": 0.4165, + "step": 36331 + }, + { + "epoch": 1.4, + "learning_rate": 2.6736342651962044e-05, + "loss": 0.3065, + "step": 36332 + }, + { + "epoch": 1.4, + "learning_rate": 2.6735701461913316e-05, + "loss": 0.4654, + "step": 36333 + }, + { + "epoch": 1.4, + "learning_rate": 2.673506027186458e-05, + "loss": 0.2488, + "step": 36334 + }, + { + "epoch": 1.4, + "learning_rate": 2.6734419081815852e-05, + "loss": 0.4442, + "step": 36335 + }, + { + "epoch": 1.4, + "learning_rate": 2.673377789176712e-05, + "loss": 0.3973, + "step": 36336 + }, + { + "epoch": 1.4, + "learning_rate": 2.6733136701718392e-05, + "loss": 0.3451, + "step": 36337 + }, + { + "epoch": 1.4, + "learning_rate": 2.6732495511669657e-05, + "loss": 0.4053, + "step": 36338 + }, + { + "epoch": 1.4, + "learning_rate": 2.673185432162093e-05, + "loss": 0.2991, + "step": 36339 + }, + { + "epoch": 1.4, + "learning_rate": 2.6731213131572197e-05, + "loss": 0.3189, + "step": 36340 + }, + { + "epoch": 1.4, + "learning_rate": 2.673057194152347e-05, + "loss": 0.2899, + "step": 36341 + }, + { + "epoch": 1.4, + "learning_rate": 2.672993075147474e-05, + "loss": 0.3579, + "step": 36342 + }, + { + "epoch": 1.4, + "learning_rate": 2.672928956142601e-05, + "loss": 0.4268, + "step": 36343 + }, + { + "epoch": 1.4, + "learning_rate": 2.672864837137728e-05, + "loss": 0.3872, + "step": 36344 + }, + { + "epoch": 1.4, + "learning_rate": 2.6728007181328546e-05, + "loss": 0.4047, + "step": 36345 + }, + { + "epoch": 1.4, + "learning_rate": 2.6727365991279817e-05, + "loss": 0.3367, + "step": 36346 + }, + { + "epoch": 1.4, + "learning_rate": 2.6726724801231086e-05, + "loss": 0.3886, + "step": 36347 + }, + { + "epoch": 1.4, + "learning_rate": 2.6726083611182357e-05, + "loss": 0.3769, + "step": 36348 + }, + { + "epoch": 1.4, + "learning_rate": 2.6725442421133622e-05, + "loss": 0.3341, + "step": 36349 + }, + { + "epoch": 1.4, + "learning_rate": 2.6724801231084894e-05, + "loss": 0.4716, + "step": 36350 + }, + { + "epoch": 1.4, + "learning_rate": 2.6724160041036166e-05, + "loss": 0.364, + "step": 36351 + }, + { + "epoch": 1.4, + "learning_rate": 2.6723518850987434e-05, + "loss": 0.3295, + "step": 36352 + }, + { + "epoch": 1.4, + "learning_rate": 2.6722877660938706e-05, + "loss": 0.2524, + "step": 36353 + }, + { + "epoch": 1.4, + "learning_rate": 2.672223647088997e-05, + "loss": 0.3586, + "step": 36354 + }, + { + "epoch": 1.4, + "learning_rate": 2.6721595280841243e-05, + "loss": 0.3414, + "step": 36355 + }, + { + "epoch": 1.4, + "learning_rate": 2.672095409079251e-05, + "loss": 0.418, + "step": 36356 + }, + { + "epoch": 1.4, + "learning_rate": 2.6720312900743783e-05, + "loss": 0.3821, + "step": 36357 + }, + { + "epoch": 1.4, + "learning_rate": 2.6719671710695048e-05, + "loss": 0.487, + "step": 36358 + }, + { + "epoch": 1.4, + "learning_rate": 2.6719030520646323e-05, + "loss": 0.3761, + "step": 36359 + }, + { + "epoch": 1.4, + "learning_rate": 2.6718389330597594e-05, + "loss": 0.4652, + "step": 36360 + }, + { + "epoch": 1.4, + "learning_rate": 2.671774814054886e-05, + "loss": 0.4631, + "step": 36361 + }, + { + "epoch": 1.4, + "learning_rate": 2.671710695050013e-05, + "loss": 0.3982, + "step": 36362 + }, + { + "epoch": 1.4, + "learning_rate": 2.67164657604514e-05, + "loss": 0.502, + "step": 36363 + }, + { + "epoch": 1.4, + "learning_rate": 2.671582457040267e-05, + "loss": 0.4578, + "step": 36364 + }, + { + "epoch": 1.4, + "learning_rate": 2.6715183380353936e-05, + "loss": 0.2102, + "step": 36365 + }, + { + "epoch": 1.4, + "learning_rate": 2.6714542190305208e-05, + "loss": 0.2598, + "step": 36366 + }, + { + "epoch": 1.4, + "learning_rate": 2.6713901000256476e-05, + "loss": 0.5088, + "step": 36367 + }, + { + "epoch": 1.4, + "learning_rate": 2.6713259810207748e-05, + "loss": 0.5055, + "step": 36368 + }, + { + "epoch": 1.4, + "learning_rate": 2.671261862015902e-05, + "loss": 0.392, + "step": 36369 + }, + { + "epoch": 1.4, + "learning_rate": 2.6711977430110284e-05, + "loss": 0.3673, + "step": 36370 + }, + { + "epoch": 1.4, + "learning_rate": 2.6711336240061556e-05, + "loss": 0.2935, + "step": 36371 + }, + { + "epoch": 1.4, + "learning_rate": 2.6710695050012824e-05, + "loss": 0.4571, + "step": 36372 + }, + { + "epoch": 1.4, + "learning_rate": 2.6710053859964096e-05, + "loss": 0.3049, + "step": 36373 + }, + { + "epoch": 1.4, + "learning_rate": 2.670941266991536e-05, + "loss": 0.3896, + "step": 36374 + }, + { + "epoch": 1.4, + "learning_rate": 2.6708771479866633e-05, + "loss": 0.3714, + "step": 36375 + }, + { + "epoch": 1.4, + "learning_rate": 2.67081302898179e-05, + "loss": 0.4459, + "step": 36376 + }, + { + "epoch": 1.4, + "learning_rate": 2.6707489099769173e-05, + "loss": 0.432, + "step": 36377 + }, + { + "epoch": 1.4, + "learning_rate": 2.6706847909720445e-05, + "loss": 0.3088, + "step": 36378 + }, + { + "epoch": 1.4, + "learning_rate": 2.6706206719671713e-05, + "loss": 0.279, + "step": 36379 + }, + { + "epoch": 1.4, + "learning_rate": 2.6705565529622985e-05, + "loss": 0.4166, + "step": 36380 + }, + { + "epoch": 1.4, + "learning_rate": 2.670492433957425e-05, + "loss": 0.2465, + "step": 36381 + }, + { + "epoch": 1.4, + "learning_rate": 2.670428314952552e-05, + "loss": 0.4403, + "step": 36382 + }, + { + "epoch": 1.4, + "learning_rate": 2.670364195947679e-05, + "loss": 0.4719, + "step": 36383 + }, + { + "epoch": 1.4, + "learning_rate": 2.670300076942806e-05, + "loss": 0.3343, + "step": 36384 + }, + { + "epoch": 1.4, + "learning_rate": 2.6702359579379326e-05, + "loss": 0.4624, + "step": 36385 + }, + { + "epoch": 1.4, + "learning_rate": 2.6701718389330598e-05, + "loss": 0.3258, + "step": 36386 + }, + { + "epoch": 1.4, + "learning_rate": 2.670107719928187e-05, + "loss": 0.5015, + "step": 36387 + }, + { + "epoch": 1.4, + "learning_rate": 2.6700436009233138e-05, + "loss": 0.3388, + "step": 36388 + }, + { + "epoch": 1.4, + "learning_rate": 2.669979481918441e-05, + "loss": 0.3228, + "step": 36389 + }, + { + "epoch": 1.4, + "learning_rate": 2.6699153629135675e-05, + "loss": 0.3685, + "step": 36390 + }, + { + "epoch": 1.4, + "learning_rate": 2.6698512439086946e-05, + "loss": 0.3975, + "step": 36391 + }, + { + "epoch": 1.4, + "learning_rate": 2.6697871249038215e-05, + "loss": 0.2976, + "step": 36392 + }, + { + "epoch": 1.4, + "learning_rate": 2.6697230058989486e-05, + "loss": 0.3625, + "step": 36393 + }, + { + "epoch": 1.4, + "learning_rate": 2.669658886894075e-05, + "loss": 0.2555, + "step": 36394 + }, + { + "epoch": 1.4, + "learning_rate": 2.6695947678892027e-05, + "loss": 0.4635, + "step": 36395 + }, + { + "epoch": 1.4, + "learning_rate": 2.6695306488843298e-05, + "loss": 0.6226, + "step": 36396 + }, + { + "epoch": 1.4, + "learning_rate": 2.6694665298794563e-05, + "loss": 0.4683, + "step": 36397 + }, + { + "epoch": 1.4, + "learning_rate": 2.6694024108745835e-05, + "loss": 0.3318, + "step": 36398 + }, + { + "epoch": 1.4, + "learning_rate": 2.6693382918697103e-05, + "loss": 0.3447, + "step": 36399 + }, + { + "epoch": 1.4, + "learning_rate": 2.6692741728648375e-05, + "loss": 0.3691, + "step": 36400 + }, + { + "epoch": 1.4, + "learning_rate": 2.669210053859964e-05, + "loss": 0.2817, + "step": 36401 + }, + { + "epoch": 1.4, + "learning_rate": 2.669145934855091e-05, + "loss": 0.3456, + "step": 36402 + }, + { + "epoch": 1.4, + "learning_rate": 2.669081815850218e-05, + "loss": 0.4271, + "step": 36403 + }, + { + "epoch": 1.4, + "learning_rate": 2.669017696845345e-05, + "loss": 0.3419, + "step": 36404 + }, + { + "epoch": 1.4, + "learning_rate": 2.6689535778404723e-05, + "loss": 0.3228, + "step": 36405 + }, + { + "epoch": 1.4, + "learning_rate": 2.6688894588355988e-05, + "loss": 0.3277, + "step": 36406 + }, + { + "epoch": 1.4, + "learning_rate": 2.668825339830726e-05, + "loss": 0.3364, + "step": 36407 + }, + { + "epoch": 1.4, + "learning_rate": 2.668761220825853e-05, + "loss": 0.2965, + "step": 36408 + }, + { + "epoch": 1.4, + "learning_rate": 2.66869710182098e-05, + "loss": 0.3582, + "step": 36409 + }, + { + "epoch": 1.4, + "learning_rate": 2.6686329828161065e-05, + "loss": 0.494, + "step": 36410 + }, + { + "epoch": 1.4, + "learning_rate": 2.6685688638112337e-05, + "loss": 0.5073, + "step": 36411 + }, + { + "epoch": 1.4, + "learning_rate": 2.6685047448063605e-05, + "loss": 0.4475, + "step": 36412 + }, + { + "epoch": 1.4, + "learning_rate": 2.6684406258014877e-05, + "loss": 0.3117, + "step": 36413 + }, + { + "epoch": 1.4, + "learning_rate": 2.668376506796615e-05, + "loss": 0.2516, + "step": 36414 + }, + { + "epoch": 1.4, + "learning_rate": 2.6683123877917417e-05, + "loss": 0.2711, + "step": 36415 + }, + { + "epoch": 1.4, + "learning_rate": 2.668248268786869e-05, + "loss": 0.2979, + "step": 36416 + }, + { + "epoch": 1.4, + "learning_rate": 2.6681841497819953e-05, + "loss": 0.4198, + "step": 36417 + }, + { + "epoch": 1.4, + "learning_rate": 2.6681200307771225e-05, + "loss": 0.3547, + "step": 36418 + }, + { + "epoch": 1.4, + "learning_rate": 2.6680559117722493e-05, + "loss": 0.2993, + "step": 36419 + }, + { + "epoch": 1.4, + "learning_rate": 2.6679917927673765e-05, + "loss": 0.3536, + "step": 36420 + }, + { + "epoch": 1.4, + "learning_rate": 2.667927673762503e-05, + "loss": 0.3224, + "step": 36421 + }, + { + "epoch": 1.4, + "learning_rate": 2.6678635547576302e-05, + "loss": 0.4031, + "step": 36422 + }, + { + "epoch": 1.4, + "learning_rate": 2.6677994357527574e-05, + "loss": 0.384, + "step": 36423 + }, + { + "epoch": 1.4, + "learning_rate": 2.6677353167478842e-05, + "loss": 0.2968, + "step": 36424 + }, + { + "epoch": 1.4, + "learning_rate": 2.6676711977430114e-05, + "loss": 0.293, + "step": 36425 + }, + { + "epoch": 1.4, + "learning_rate": 2.667607078738138e-05, + "loss": 0.5358, + "step": 36426 + }, + { + "epoch": 1.4, + "learning_rate": 2.667542959733265e-05, + "loss": 0.4703, + "step": 36427 + }, + { + "epoch": 1.4, + "learning_rate": 2.667478840728392e-05, + "loss": 0.4039, + "step": 36428 + }, + { + "epoch": 1.4, + "learning_rate": 2.667414721723519e-05, + "loss": 0.4673, + "step": 36429 + }, + { + "epoch": 1.4, + "learning_rate": 2.6673506027186455e-05, + "loss": 0.3519, + "step": 36430 + }, + { + "epoch": 1.4, + "learning_rate": 2.667286483713773e-05, + "loss": 0.4438, + "step": 36431 + }, + { + "epoch": 1.4, + "learning_rate": 2.6672223647089002e-05, + "loss": 0.2902, + "step": 36432 + }, + { + "epoch": 1.4, + "learning_rate": 2.6671582457040267e-05, + "loss": 0.269, + "step": 36433 + }, + { + "epoch": 1.4, + "learning_rate": 2.667094126699154e-05, + "loss": 0.3248, + "step": 36434 + }, + { + "epoch": 1.4, + "learning_rate": 2.6670300076942807e-05, + "loss": 0.2639, + "step": 36435 + }, + { + "epoch": 1.4, + "learning_rate": 2.666965888689408e-05, + "loss": 0.41, + "step": 36436 + }, + { + "epoch": 1.4, + "learning_rate": 2.6669017696845344e-05, + "loss": 0.4062, + "step": 36437 + }, + { + "epoch": 1.4, + "learning_rate": 2.6668376506796615e-05, + "loss": 0.2836, + "step": 36438 + }, + { + "epoch": 1.4, + "learning_rate": 2.6667735316747884e-05, + "loss": 0.3407, + "step": 36439 + }, + { + "epoch": 1.4, + "learning_rate": 2.6667094126699155e-05, + "loss": 0.2334, + "step": 36440 + }, + { + "epoch": 1.4, + "learning_rate": 2.6666452936650427e-05, + "loss": 0.3471, + "step": 36441 + }, + { + "epoch": 1.4, + "learning_rate": 2.6665811746601692e-05, + "loss": 0.274, + "step": 36442 + }, + { + "epoch": 1.4, + "learning_rate": 2.6665170556552964e-05, + "loss": 0.3623, + "step": 36443 + }, + { + "epoch": 1.4, + "learning_rate": 2.6664529366504232e-05, + "loss": 0.3402, + "step": 36444 + }, + { + "epoch": 1.4, + "learning_rate": 2.6663888176455504e-05, + "loss": 0.3288, + "step": 36445 + }, + { + "epoch": 1.4, + "learning_rate": 2.666324698640677e-05, + "loss": 0.3868, + "step": 36446 + }, + { + "epoch": 1.4, + "learning_rate": 2.666260579635804e-05, + "loss": 0.3603, + "step": 36447 + }, + { + "epoch": 1.4, + "learning_rate": 2.666196460630931e-05, + "loss": 0.3584, + "step": 36448 + }, + { + "epoch": 1.4, + "learning_rate": 2.666132341626058e-05, + "loss": 0.3303, + "step": 36449 + }, + { + "epoch": 1.4, + "learning_rate": 2.6660682226211852e-05, + "loss": 0.5336, + "step": 36450 + }, + { + "epoch": 1.4, + "learning_rate": 2.666004103616312e-05, + "loss": 0.4822, + "step": 36451 + }, + { + "epoch": 1.4, + "learning_rate": 2.6659399846114392e-05, + "loss": 0.2815, + "step": 36452 + }, + { + "epoch": 1.4, + "learning_rate": 2.6658758656065657e-05, + "loss": 0.3098, + "step": 36453 + }, + { + "epoch": 1.4, + "learning_rate": 2.665811746601693e-05, + "loss": 0.3982, + "step": 36454 + }, + { + "epoch": 1.4, + "learning_rate": 2.6657476275968197e-05, + "loss": 0.3234, + "step": 36455 + }, + { + "epoch": 1.4, + "learning_rate": 2.665683508591947e-05, + "loss": 0.4576, + "step": 36456 + }, + { + "epoch": 1.4, + "learning_rate": 2.6656193895870734e-05, + "loss": 0.3554, + "step": 36457 + }, + { + "epoch": 1.4, + "learning_rate": 2.6655552705822006e-05, + "loss": 0.3481, + "step": 36458 + }, + { + "epoch": 1.4, + "learning_rate": 2.6654911515773277e-05, + "loss": 0.3941, + "step": 36459 + }, + { + "epoch": 1.4, + "learning_rate": 2.6654270325724546e-05, + "loss": 0.3332, + "step": 36460 + }, + { + "epoch": 1.4, + "learning_rate": 2.6653629135675818e-05, + "loss": 0.5387, + "step": 36461 + }, + { + "epoch": 1.4, + "learning_rate": 2.6652987945627082e-05, + "loss": 0.2914, + "step": 36462 + }, + { + "epoch": 1.4, + "learning_rate": 2.6652346755578354e-05, + "loss": 0.3999, + "step": 36463 + }, + { + "epoch": 1.4, + "learning_rate": 2.6651705565529622e-05, + "loss": 0.3772, + "step": 36464 + }, + { + "epoch": 1.4, + "learning_rate": 2.6651064375480894e-05, + "loss": 0.3082, + "step": 36465 + }, + { + "epoch": 1.4, + "learning_rate": 2.665042318543216e-05, + "loss": 0.3598, + "step": 36466 + }, + { + "epoch": 1.4, + "learning_rate": 2.6649781995383434e-05, + "loss": 0.2949, + "step": 36467 + }, + { + "epoch": 1.4, + "learning_rate": 2.6649140805334706e-05, + "loss": 0.3811, + "step": 36468 + }, + { + "epoch": 1.4, + "learning_rate": 2.664849961528597e-05, + "loss": 0.2757, + "step": 36469 + }, + { + "epoch": 1.4, + "learning_rate": 2.6647858425237243e-05, + "loss": 0.4719, + "step": 36470 + }, + { + "epoch": 1.4, + "learning_rate": 2.664721723518851e-05, + "loss": 0.2787, + "step": 36471 + }, + { + "epoch": 1.4, + "learning_rate": 2.6646576045139783e-05, + "loss": 0.3951, + "step": 36472 + }, + { + "epoch": 1.4, + "learning_rate": 2.6645934855091048e-05, + "loss": 0.3635, + "step": 36473 + }, + { + "epoch": 1.4, + "learning_rate": 2.664529366504232e-05, + "loss": 0.2532, + "step": 36474 + }, + { + "epoch": 1.4, + "learning_rate": 2.6644652474993588e-05, + "loss": 0.4672, + "step": 36475 + }, + { + "epoch": 1.4, + "learning_rate": 2.664401128494486e-05, + "loss": 0.3732, + "step": 36476 + }, + { + "epoch": 1.4, + "learning_rate": 2.664337009489613e-05, + "loss": 0.3801, + "step": 36477 + }, + { + "epoch": 1.4, + "learning_rate": 2.6642728904847396e-05, + "loss": 0.3537, + "step": 36478 + }, + { + "epoch": 1.4, + "learning_rate": 2.6642087714798668e-05, + "loss": 0.3971, + "step": 36479 + }, + { + "epoch": 1.4, + "learning_rate": 2.6641446524749936e-05, + "loss": 0.3013, + "step": 36480 + }, + { + "epoch": 1.4, + "learning_rate": 2.6640805334701208e-05, + "loss": 0.3646, + "step": 36481 + }, + { + "epoch": 1.4, + "learning_rate": 2.6640164144652473e-05, + "loss": 0.4143, + "step": 36482 + }, + { + "epoch": 1.4, + "learning_rate": 2.6639522954603744e-05, + "loss": 0.3913, + "step": 36483 + }, + { + "epoch": 1.4, + "learning_rate": 2.6638881764555013e-05, + "loss": 0.3687, + "step": 36484 + }, + { + "epoch": 1.4, + "learning_rate": 2.6638240574506284e-05, + "loss": 0.4891, + "step": 36485 + }, + { + "epoch": 1.4, + "learning_rate": 2.6637599384457556e-05, + "loss": 0.3936, + "step": 36486 + }, + { + "epoch": 1.4, + "learning_rate": 2.6636958194408825e-05, + "loss": 0.3397, + "step": 36487 + }, + { + "epoch": 1.4, + "learning_rate": 2.6636317004360096e-05, + "loss": 0.4599, + "step": 36488 + }, + { + "epoch": 1.4, + "learning_rate": 2.663567581431136e-05, + "loss": 0.3173, + "step": 36489 + }, + { + "epoch": 1.4, + "learning_rate": 2.6635034624262633e-05, + "loss": 0.2404, + "step": 36490 + }, + { + "epoch": 1.4, + "learning_rate": 2.66343934342139e-05, + "loss": 0.4624, + "step": 36491 + }, + { + "epoch": 1.4, + "learning_rate": 2.6633752244165173e-05, + "loss": 0.3062, + "step": 36492 + }, + { + "epoch": 1.4, + "learning_rate": 2.6633111054116445e-05, + "loss": 0.2423, + "step": 36493 + }, + { + "epoch": 1.4, + "learning_rate": 2.663246986406771e-05, + "loss": 0.3519, + "step": 36494 + }, + { + "epoch": 1.4, + "learning_rate": 2.663182867401898e-05, + "loss": 0.5377, + "step": 36495 + }, + { + "epoch": 1.4, + "learning_rate": 2.663118748397025e-05, + "loss": 0.3661, + "step": 36496 + }, + { + "epoch": 1.4, + "learning_rate": 2.663054629392152e-05, + "loss": 0.3237, + "step": 36497 + }, + { + "epoch": 1.4, + "learning_rate": 2.6629905103872786e-05, + "loss": 0.526, + "step": 36498 + }, + { + "epoch": 1.4, + "learning_rate": 2.6629263913824058e-05, + "loss": 0.444, + "step": 36499 + }, + { + "epoch": 1.4, + "learning_rate": 2.6628622723775326e-05, + "loss": 0.3735, + "step": 36500 + }, + { + "epoch": 1.4, + "learning_rate": 2.6627981533726598e-05, + "loss": 0.254, + "step": 36501 + }, + { + "epoch": 1.4, + "learning_rate": 2.662734034367787e-05, + "loss": 0.3406, + "step": 36502 + }, + { + "epoch": 1.4, + "learning_rate": 2.6626699153629138e-05, + "loss": 0.4051, + "step": 36503 + }, + { + "epoch": 1.4, + "learning_rate": 2.662605796358041e-05, + "loss": 0.3838, + "step": 36504 + }, + { + "epoch": 1.4, + "learning_rate": 2.6625416773531675e-05, + "loss": 0.3546, + "step": 36505 + }, + { + "epoch": 1.4, + "learning_rate": 2.6624775583482946e-05, + "loss": 0.3411, + "step": 36506 + }, + { + "epoch": 1.4, + "learning_rate": 2.6624134393434215e-05, + "loss": 0.3221, + "step": 36507 + }, + { + "epoch": 1.4, + "learning_rate": 2.6623493203385487e-05, + "loss": 0.5639, + "step": 36508 + }, + { + "epoch": 1.4, + "learning_rate": 2.662285201333675e-05, + "loss": 0.3493, + "step": 36509 + }, + { + "epoch": 1.4, + "learning_rate": 2.6622210823288023e-05, + "loss": 0.3165, + "step": 36510 + }, + { + "epoch": 1.4, + "learning_rate": 2.6621569633239295e-05, + "loss": 0.4497, + "step": 36511 + }, + { + "epoch": 1.4, + "learning_rate": 2.6620928443190563e-05, + "loss": 0.1816, + "step": 36512 + }, + { + "epoch": 1.4, + "learning_rate": 2.6620287253141835e-05, + "loss": 0.4946, + "step": 36513 + }, + { + "epoch": 1.4, + "learning_rate": 2.66196460630931e-05, + "loss": 0.3535, + "step": 36514 + }, + { + "epoch": 1.4, + "learning_rate": 2.661900487304437e-05, + "loss": 0.4227, + "step": 36515 + }, + { + "epoch": 1.4, + "learning_rate": 2.661836368299564e-05, + "loss": 0.2907, + "step": 36516 + }, + { + "epoch": 1.4, + "learning_rate": 2.661772249294691e-05, + "loss": 0.4005, + "step": 36517 + }, + { + "epoch": 1.4, + "learning_rate": 2.6617081302898177e-05, + "loss": 0.3297, + "step": 36518 + }, + { + "epoch": 1.4, + "learning_rate": 2.6616440112849452e-05, + "loss": 0.399, + "step": 36519 + }, + { + "epoch": 1.4, + "learning_rate": 2.6615798922800723e-05, + "loss": 0.3633, + "step": 36520 + }, + { + "epoch": 1.4, + "learning_rate": 2.661515773275199e-05, + "loss": 0.4404, + "step": 36521 + }, + { + "epoch": 1.4, + "learning_rate": 2.661451654270326e-05, + "loss": 0.3565, + "step": 36522 + }, + { + "epoch": 1.4, + "learning_rate": 2.661387535265453e-05, + "loss": 0.3188, + "step": 36523 + }, + { + "epoch": 1.4, + "learning_rate": 2.66132341626058e-05, + "loss": 0.3306, + "step": 36524 + }, + { + "epoch": 1.4, + "learning_rate": 2.6612592972557065e-05, + "loss": 0.3855, + "step": 36525 + }, + { + "epoch": 1.4, + "learning_rate": 2.6611951782508337e-05, + "loss": 0.3411, + "step": 36526 + }, + { + "epoch": 1.4, + "learning_rate": 2.6611310592459605e-05, + "loss": 0.3705, + "step": 36527 + }, + { + "epoch": 1.4, + "learning_rate": 2.6610669402410877e-05, + "loss": 0.2709, + "step": 36528 + }, + { + "epoch": 1.4, + "learning_rate": 2.661002821236215e-05, + "loss": 0.272, + "step": 36529 + }, + { + "epoch": 1.4, + "learning_rate": 2.6609387022313413e-05, + "loss": 0.4241, + "step": 36530 + }, + { + "epoch": 1.4, + "learning_rate": 2.6608745832264685e-05, + "loss": 0.4122, + "step": 36531 + }, + { + "epoch": 1.4, + "learning_rate": 2.6608104642215954e-05, + "loss": 0.2802, + "step": 36532 + }, + { + "epoch": 1.4, + "learning_rate": 2.6607463452167225e-05, + "loss": 0.4439, + "step": 36533 + }, + { + "epoch": 1.4, + "learning_rate": 2.660682226211849e-05, + "loss": 0.3178, + "step": 36534 + }, + { + "epoch": 1.4, + "learning_rate": 2.6606181072069762e-05, + "loss": 0.5237, + "step": 36535 + }, + { + "epoch": 1.4, + "learning_rate": 2.660553988202103e-05, + "loss": 0.4199, + "step": 36536 + }, + { + "epoch": 1.4, + "learning_rate": 2.6604898691972302e-05, + "loss": 0.4309, + "step": 36537 + }, + { + "epoch": 1.4, + "learning_rate": 2.6604257501923574e-05, + "loss": 0.346, + "step": 36538 + }, + { + "epoch": 1.4, + "learning_rate": 2.6603616311874842e-05, + "loss": 0.3901, + "step": 36539 + }, + { + "epoch": 1.4, + "learning_rate": 2.6602975121826114e-05, + "loss": 0.3298, + "step": 36540 + }, + { + "epoch": 1.4, + "learning_rate": 2.660233393177738e-05, + "loss": 0.3465, + "step": 36541 + }, + { + "epoch": 1.4, + "learning_rate": 2.660169274172865e-05, + "loss": 0.407, + "step": 36542 + }, + { + "epoch": 1.4, + "learning_rate": 2.660105155167992e-05, + "loss": 0.3423, + "step": 36543 + }, + { + "epoch": 1.4, + "learning_rate": 2.660041036163119e-05, + "loss": 0.3224, + "step": 36544 + }, + { + "epoch": 1.41, + "learning_rate": 2.6599769171582455e-05, + "loss": 0.4139, + "step": 36545 + }, + { + "epoch": 1.41, + "learning_rate": 2.6599127981533727e-05, + "loss": 0.3558, + "step": 36546 + }, + { + "epoch": 1.41, + "learning_rate": 2.6598486791485e-05, + "loss": 0.4428, + "step": 36547 + }, + { + "epoch": 1.41, + "learning_rate": 2.6597845601436267e-05, + "loss": 0.2565, + "step": 36548 + }, + { + "epoch": 1.41, + "learning_rate": 2.659720441138754e-05, + "loss": 0.5514, + "step": 36549 + }, + { + "epoch": 1.41, + "learning_rate": 2.6596563221338804e-05, + "loss": 0.4045, + "step": 36550 + }, + { + "epoch": 1.41, + "learning_rate": 2.6595922031290075e-05, + "loss": 0.4546, + "step": 36551 + }, + { + "epoch": 1.41, + "learning_rate": 2.6595280841241344e-05, + "loss": 0.3573, + "step": 36552 + }, + { + "epoch": 1.41, + "learning_rate": 2.6594639651192616e-05, + "loss": 0.3105, + "step": 36553 + }, + { + "epoch": 1.41, + "learning_rate": 2.659399846114388e-05, + "loss": 0.4084, + "step": 36554 + }, + { + "epoch": 1.41, + "learning_rate": 2.6593357271095156e-05, + "loss": 0.3185, + "step": 36555 + }, + { + "epoch": 1.41, + "learning_rate": 2.6592716081046427e-05, + "loss": 0.3969, + "step": 36556 + }, + { + "epoch": 1.41, + "learning_rate": 2.6592074890997692e-05, + "loss": 0.3266, + "step": 36557 + }, + { + "epoch": 1.41, + "learning_rate": 2.6591433700948964e-05, + "loss": 0.42, + "step": 36558 + }, + { + "epoch": 1.41, + "learning_rate": 2.6590792510900232e-05, + "loss": 0.2989, + "step": 36559 + }, + { + "epoch": 1.41, + "learning_rate": 2.6590151320851504e-05, + "loss": 0.3417, + "step": 36560 + }, + { + "epoch": 1.41, + "learning_rate": 2.658951013080277e-05, + "loss": 0.318, + "step": 36561 + }, + { + "epoch": 1.41, + "learning_rate": 2.658886894075404e-05, + "loss": 0.4003, + "step": 36562 + }, + { + "epoch": 1.41, + "learning_rate": 2.658822775070531e-05, + "loss": 0.2964, + "step": 36563 + }, + { + "epoch": 1.41, + "learning_rate": 2.658758656065658e-05, + "loss": 0.3309, + "step": 36564 + }, + { + "epoch": 1.41, + "learning_rate": 2.6586945370607852e-05, + "loss": 0.3536, + "step": 36565 + }, + { + "epoch": 1.41, + "learning_rate": 2.6586304180559117e-05, + "loss": 0.2831, + "step": 36566 + }, + { + "epoch": 1.41, + "learning_rate": 2.658566299051039e-05, + "loss": 0.3315, + "step": 36567 + }, + { + "epoch": 1.41, + "learning_rate": 2.6585021800461657e-05, + "loss": 0.3677, + "step": 36568 + }, + { + "epoch": 1.41, + "learning_rate": 2.658438061041293e-05, + "loss": 0.2924, + "step": 36569 + }, + { + "epoch": 1.41, + "learning_rate": 2.6583739420364194e-05, + "loss": 0.5145, + "step": 36570 + }, + { + "epoch": 1.41, + "learning_rate": 2.6583098230315466e-05, + "loss": 0.4091, + "step": 36571 + }, + { + "epoch": 1.41, + "learning_rate": 2.6582457040266734e-05, + "loss": 0.4409, + "step": 36572 + }, + { + "epoch": 1.41, + "learning_rate": 2.6581815850218006e-05, + "loss": 0.4131, + "step": 36573 + }, + { + "epoch": 1.41, + "learning_rate": 2.6581174660169278e-05, + "loss": 0.2578, + "step": 36574 + }, + { + "epoch": 1.41, + "learning_rate": 2.6580533470120546e-05, + "loss": 0.2908, + "step": 36575 + }, + { + "epoch": 1.41, + "learning_rate": 2.6579892280071818e-05, + "loss": 0.249, + "step": 36576 + }, + { + "epoch": 1.41, + "learning_rate": 2.6579251090023083e-05, + "loss": 0.4178, + "step": 36577 + }, + { + "epoch": 1.41, + "learning_rate": 2.6578609899974354e-05, + "loss": 0.5136, + "step": 36578 + }, + { + "epoch": 1.41, + "learning_rate": 2.6577968709925623e-05, + "loss": 0.3168, + "step": 36579 + }, + { + "epoch": 1.41, + "learning_rate": 2.6577327519876894e-05, + "loss": 0.3383, + "step": 36580 + }, + { + "epoch": 1.41, + "learning_rate": 2.657668632982816e-05, + "loss": 0.4078, + "step": 36581 + }, + { + "epoch": 1.41, + "learning_rate": 2.657604513977943e-05, + "loss": 0.3973, + "step": 36582 + }, + { + "epoch": 1.41, + "learning_rate": 2.6575403949730703e-05, + "loss": 0.4121, + "step": 36583 + }, + { + "epoch": 1.41, + "learning_rate": 2.657476275968197e-05, + "loss": 0.4293, + "step": 36584 + }, + { + "epoch": 1.41, + "learning_rate": 2.6574121569633243e-05, + "loss": 0.3979, + "step": 36585 + }, + { + "epoch": 1.41, + "learning_rate": 2.6573480379584508e-05, + "loss": 0.3532, + "step": 36586 + }, + { + "epoch": 1.41, + "learning_rate": 2.657283918953578e-05, + "loss": 0.3479, + "step": 36587 + }, + { + "epoch": 1.41, + "learning_rate": 2.6572197999487048e-05, + "loss": 0.432, + "step": 36588 + }, + { + "epoch": 1.41, + "learning_rate": 2.657155680943832e-05, + "loss": 0.2721, + "step": 36589 + }, + { + "epoch": 1.41, + "learning_rate": 2.6570915619389584e-05, + "loss": 0.3659, + "step": 36590 + }, + { + "epoch": 1.41, + "learning_rate": 2.657027442934086e-05, + "loss": 0.3076, + "step": 36591 + }, + { + "epoch": 1.41, + "learning_rate": 2.656963323929213e-05, + "loss": 0.4821, + "step": 36592 + }, + { + "epoch": 1.41, + "learning_rate": 2.6568992049243396e-05, + "loss": 0.3331, + "step": 36593 + }, + { + "epoch": 1.41, + "learning_rate": 2.6568350859194668e-05, + "loss": 0.3272, + "step": 36594 + }, + { + "epoch": 1.41, + "learning_rate": 2.6567709669145936e-05, + "loss": 0.3692, + "step": 36595 + }, + { + "epoch": 1.41, + "learning_rate": 2.6567068479097208e-05, + "loss": 0.2218, + "step": 36596 + }, + { + "epoch": 1.41, + "learning_rate": 2.6566427289048473e-05, + "loss": 0.392, + "step": 36597 + }, + { + "epoch": 1.41, + "learning_rate": 2.6565786098999745e-05, + "loss": 0.2601, + "step": 36598 + }, + { + "epoch": 1.41, + "learning_rate": 2.6565144908951013e-05, + "loss": 0.4126, + "step": 36599 + }, + { + "epoch": 1.41, + "learning_rate": 2.6564503718902285e-05, + "loss": 0.3191, + "step": 36600 + }, + { + "epoch": 1.41, + "learning_rate": 2.6563862528853556e-05, + "loss": 0.341, + "step": 36601 + }, + { + "epoch": 1.41, + "learning_rate": 2.656322133880482e-05, + "loss": 0.3322, + "step": 36602 + }, + { + "epoch": 1.41, + "learning_rate": 2.6562580148756093e-05, + "loss": 0.3736, + "step": 36603 + }, + { + "epoch": 1.41, + "learning_rate": 2.656193895870736e-05, + "loss": 0.259, + "step": 36604 + }, + { + "epoch": 1.41, + "learning_rate": 2.6561297768658633e-05, + "loss": 0.3642, + "step": 36605 + }, + { + "epoch": 1.41, + "learning_rate": 2.6560656578609898e-05, + "loss": 0.2687, + "step": 36606 + }, + { + "epoch": 1.41, + "learning_rate": 2.656001538856117e-05, + "loss": 0.316, + "step": 36607 + }, + { + "epoch": 1.41, + "learning_rate": 2.6559374198512438e-05, + "loss": 0.6566, + "step": 36608 + }, + { + "epoch": 1.41, + "learning_rate": 2.655873300846371e-05, + "loss": 0.2751, + "step": 36609 + }, + { + "epoch": 1.41, + "learning_rate": 2.655809181841498e-05, + "loss": 0.3051, + "step": 36610 + }, + { + "epoch": 1.41, + "learning_rate": 2.655745062836625e-05, + "loss": 0.5462, + "step": 36611 + }, + { + "epoch": 1.41, + "learning_rate": 2.655680943831752e-05, + "loss": 0.4265, + "step": 36612 + }, + { + "epoch": 1.41, + "learning_rate": 2.6556168248268786e-05, + "loss": 0.358, + "step": 36613 + }, + { + "epoch": 1.41, + "learning_rate": 2.6555527058220058e-05, + "loss": 0.4592, + "step": 36614 + }, + { + "epoch": 1.41, + "learning_rate": 2.6554885868171326e-05, + "loss": 0.4434, + "step": 36615 + }, + { + "epoch": 1.41, + "learning_rate": 2.6554244678122598e-05, + "loss": 0.3839, + "step": 36616 + }, + { + "epoch": 1.41, + "learning_rate": 2.6553603488073863e-05, + "loss": 0.3328, + "step": 36617 + }, + { + "epoch": 1.41, + "learning_rate": 2.6552962298025135e-05, + "loss": 0.2904, + "step": 36618 + }, + { + "epoch": 1.41, + "learning_rate": 2.6552321107976407e-05, + "loss": 0.5129, + "step": 36619 + }, + { + "epoch": 1.41, + "learning_rate": 2.6551679917927675e-05, + "loss": 0.3457, + "step": 36620 + }, + { + "epoch": 1.41, + "learning_rate": 2.6551038727878947e-05, + "loss": 0.2732, + "step": 36621 + }, + { + "epoch": 1.41, + "learning_rate": 2.655039753783021e-05, + "loss": 0.3779, + "step": 36622 + }, + { + "epoch": 1.41, + "learning_rate": 2.6549756347781483e-05, + "loss": 0.3674, + "step": 36623 + }, + { + "epoch": 1.41, + "learning_rate": 2.654911515773275e-05, + "loss": 0.3938, + "step": 36624 + }, + { + "epoch": 1.41, + "learning_rate": 2.6548473967684023e-05, + "loss": 0.3068, + "step": 36625 + }, + { + "epoch": 1.41, + "learning_rate": 2.6547832777635288e-05, + "loss": 0.4941, + "step": 36626 + }, + { + "epoch": 1.41, + "learning_rate": 2.6547191587586563e-05, + "loss": 0.4853, + "step": 36627 + }, + { + "epoch": 1.41, + "learning_rate": 2.6546550397537835e-05, + "loss": 0.3251, + "step": 36628 + }, + { + "epoch": 1.41, + "learning_rate": 2.65459092074891e-05, + "loss": 0.4749, + "step": 36629 + }, + { + "epoch": 1.41, + "learning_rate": 2.6545268017440372e-05, + "loss": 0.3801, + "step": 36630 + }, + { + "epoch": 1.41, + "learning_rate": 2.654462682739164e-05, + "loss": 0.3207, + "step": 36631 + }, + { + "epoch": 1.41, + "learning_rate": 2.6543985637342912e-05, + "loss": 0.4099, + "step": 36632 + }, + { + "epoch": 1.41, + "learning_rate": 2.6543344447294177e-05, + "loss": 0.4088, + "step": 36633 + }, + { + "epoch": 1.41, + "learning_rate": 2.654270325724545e-05, + "loss": 0.2315, + "step": 36634 + }, + { + "epoch": 1.41, + "learning_rate": 2.6542062067196717e-05, + "loss": 0.4125, + "step": 36635 + }, + { + "epoch": 1.41, + "learning_rate": 2.654142087714799e-05, + "loss": 0.3168, + "step": 36636 + }, + { + "epoch": 1.41, + "learning_rate": 2.654077968709926e-05, + "loss": 0.4471, + "step": 36637 + }, + { + "epoch": 1.41, + "learning_rate": 2.6540138497050525e-05, + "loss": 0.3866, + "step": 36638 + }, + { + "epoch": 1.41, + "learning_rate": 2.6539497307001797e-05, + "loss": 0.351, + "step": 36639 + }, + { + "epoch": 1.41, + "learning_rate": 2.6538856116953065e-05, + "loss": 0.3074, + "step": 36640 + }, + { + "epoch": 1.41, + "learning_rate": 2.6538214926904337e-05, + "loss": 0.4133, + "step": 36641 + }, + { + "epoch": 1.41, + "learning_rate": 2.6537573736855602e-05, + "loss": 0.3332, + "step": 36642 + }, + { + "epoch": 1.41, + "learning_rate": 2.6536932546806877e-05, + "loss": 0.376, + "step": 36643 + }, + { + "epoch": 1.41, + "learning_rate": 2.6536291356758142e-05, + "loss": 0.2829, + "step": 36644 + }, + { + "epoch": 1.41, + "learning_rate": 2.6535650166709414e-05, + "loss": 0.478, + "step": 36645 + }, + { + "epoch": 1.41, + "learning_rate": 2.6535008976660685e-05, + "loss": 0.3118, + "step": 36646 + }, + { + "epoch": 1.41, + "learning_rate": 2.6534367786611954e-05, + "loss": 0.3884, + "step": 36647 + }, + { + "epoch": 1.41, + "learning_rate": 2.6533726596563225e-05, + "loss": 0.303, + "step": 36648 + }, + { + "epoch": 1.41, + "learning_rate": 2.653308540651449e-05, + "loss": 0.3002, + "step": 36649 + }, + { + "epoch": 1.41, + "learning_rate": 2.6532444216465762e-05, + "loss": 0.3507, + "step": 36650 + }, + { + "epoch": 1.41, + "learning_rate": 2.653180302641703e-05, + "loss": 0.4238, + "step": 36651 + }, + { + "epoch": 1.41, + "learning_rate": 2.6531161836368302e-05, + "loss": 0.2639, + "step": 36652 + }, + { + "epoch": 1.41, + "learning_rate": 2.6530520646319567e-05, + "loss": 0.4997, + "step": 36653 + }, + { + "epoch": 1.41, + "learning_rate": 2.652987945627084e-05, + "loss": 0.3406, + "step": 36654 + }, + { + "epoch": 1.41, + "learning_rate": 2.652923826622211e-05, + "loss": 0.2597, + "step": 36655 + }, + { + "epoch": 1.41, + "learning_rate": 2.652859707617338e-05, + "loss": 0.3877, + "step": 36656 + }, + { + "epoch": 1.41, + "learning_rate": 2.652795588612465e-05, + "loss": 0.3217, + "step": 36657 + }, + { + "epoch": 1.41, + "learning_rate": 2.6527314696075915e-05, + "loss": 0.3537, + "step": 36658 + }, + { + "epoch": 1.41, + "learning_rate": 2.6526673506027187e-05, + "loss": 0.2814, + "step": 36659 + }, + { + "epoch": 1.41, + "learning_rate": 2.6526032315978455e-05, + "loss": 0.409, + "step": 36660 + }, + { + "epoch": 1.41, + "learning_rate": 2.6525391125929727e-05, + "loss": 0.4895, + "step": 36661 + }, + { + "epoch": 1.41, + "learning_rate": 2.6524749935880992e-05, + "loss": 0.3891, + "step": 36662 + }, + { + "epoch": 1.41, + "learning_rate": 2.6524108745832267e-05, + "loss": 0.4289, + "step": 36663 + }, + { + "epoch": 1.41, + "learning_rate": 2.652346755578354e-05, + "loss": 0.5332, + "step": 36664 + }, + { + "epoch": 1.41, + "learning_rate": 2.6522826365734804e-05, + "loss": 0.2444, + "step": 36665 + }, + { + "epoch": 1.41, + "learning_rate": 2.6522185175686076e-05, + "loss": 0.3564, + "step": 36666 + }, + { + "epoch": 1.41, + "learning_rate": 2.6521543985637344e-05, + "loss": 0.3868, + "step": 36667 + }, + { + "epoch": 1.41, + "learning_rate": 2.6520902795588616e-05, + "loss": 0.3435, + "step": 36668 + }, + { + "epoch": 1.41, + "learning_rate": 2.652026160553988e-05, + "loss": 0.2504, + "step": 36669 + }, + { + "epoch": 1.41, + "learning_rate": 2.6519620415491152e-05, + "loss": 0.301, + "step": 36670 + }, + { + "epoch": 1.41, + "learning_rate": 2.651897922544242e-05, + "loss": 0.4462, + "step": 36671 + }, + { + "epoch": 1.41, + "learning_rate": 2.6518338035393692e-05, + "loss": 0.3923, + "step": 36672 + }, + { + "epoch": 1.41, + "learning_rate": 2.6517696845344964e-05, + "loss": 0.2588, + "step": 36673 + }, + { + "epoch": 1.41, + "learning_rate": 2.651705565529623e-05, + "loss": 0.3863, + "step": 36674 + }, + { + "epoch": 1.41, + "learning_rate": 2.65164144652475e-05, + "loss": 0.4072, + "step": 36675 + }, + { + "epoch": 1.41, + "learning_rate": 2.651577327519877e-05, + "loss": 0.4039, + "step": 36676 + }, + { + "epoch": 1.41, + "learning_rate": 2.651513208515004e-05, + "loss": 0.4833, + "step": 36677 + }, + { + "epoch": 1.41, + "learning_rate": 2.6514490895101306e-05, + "loss": 0.4818, + "step": 36678 + }, + { + "epoch": 1.41, + "learning_rate": 2.651384970505258e-05, + "loss": 0.4036, + "step": 36679 + }, + { + "epoch": 1.41, + "learning_rate": 2.6513208515003852e-05, + "loss": 0.4383, + "step": 36680 + }, + { + "epoch": 1.41, + "learning_rate": 2.6512567324955117e-05, + "loss": 0.2427, + "step": 36681 + }, + { + "epoch": 1.41, + "learning_rate": 2.651192613490639e-05, + "loss": 0.535, + "step": 36682 + }, + { + "epoch": 1.41, + "learning_rate": 2.6511284944857657e-05, + "loss": 0.4371, + "step": 36683 + }, + { + "epoch": 1.41, + "learning_rate": 2.651064375480893e-05, + "loss": 0.3055, + "step": 36684 + }, + { + "epoch": 1.41, + "learning_rate": 2.6510002564760194e-05, + "loss": 0.4392, + "step": 36685 + }, + { + "epoch": 1.41, + "learning_rate": 2.6509361374711466e-05, + "loss": 0.266, + "step": 36686 + }, + { + "epoch": 1.41, + "learning_rate": 2.6508720184662734e-05, + "loss": 0.4275, + "step": 36687 + }, + { + "epoch": 1.41, + "learning_rate": 2.6508078994614006e-05, + "loss": 0.3432, + "step": 36688 + }, + { + "epoch": 1.41, + "learning_rate": 2.6507437804565278e-05, + "loss": 0.3528, + "step": 36689 + }, + { + "epoch": 1.41, + "learning_rate": 2.6506796614516543e-05, + "loss": 0.3547, + "step": 36690 + }, + { + "epoch": 1.41, + "learning_rate": 2.6506155424467814e-05, + "loss": 0.3361, + "step": 36691 + }, + { + "epoch": 1.41, + "learning_rate": 2.6505514234419083e-05, + "loss": 0.3453, + "step": 36692 + }, + { + "epoch": 1.41, + "learning_rate": 2.6504873044370354e-05, + "loss": 0.4731, + "step": 36693 + }, + { + "epoch": 1.41, + "learning_rate": 2.650423185432162e-05, + "loss": 0.3036, + "step": 36694 + }, + { + "epoch": 1.41, + "learning_rate": 2.650359066427289e-05, + "loss": 0.4543, + "step": 36695 + }, + { + "epoch": 1.41, + "learning_rate": 2.650294947422416e-05, + "loss": 0.2662, + "step": 36696 + }, + { + "epoch": 1.41, + "learning_rate": 2.650230828417543e-05, + "loss": 0.3423, + "step": 36697 + }, + { + "epoch": 1.41, + "learning_rate": 2.6501667094126703e-05, + "loss": 0.4868, + "step": 36698 + }, + { + "epoch": 1.41, + "learning_rate": 2.650102590407797e-05, + "loss": 0.3592, + "step": 36699 + }, + { + "epoch": 1.41, + "learning_rate": 2.6500384714029243e-05, + "loss": 0.4785, + "step": 36700 + }, + { + "epoch": 1.41, + "learning_rate": 2.6499743523980508e-05, + "loss": 0.4808, + "step": 36701 + }, + { + "epoch": 1.41, + "learning_rate": 2.649910233393178e-05, + "loss": 0.3402, + "step": 36702 + }, + { + "epoch": 1.41, + "learning_rate": 2.6498461143883048e-05, + "loss": 0.4345, + "step": 36703 + }, + { + "epoch": 1.41, + "learning_rate": 2.649781995383432e-05, + "loss": 0.4216, + "step": 36704 + }, + { + "epoch": 1.41, + "learning_rate": 2.6497178763785584e-05, + "loss": 0.4357, + "step": 36705 + }, + { + "epoch": 1.41, + "learning_rate": 2.6496537573736856e-05, + "loss": 0.354, + "step": 36706 + }, + { + "epoch": 1.41, + "learning_rate": 2.6495896383688128e-05, + "loss": 0.5286, + "step": 36707 + }, + { + "epoch": 1.41, + "learning_rate": 2.6495255193639396e-05, + "loss": 0.4462, + "step": 36708 + }, + { + "epoch": 1.41, + "learning_rate": 2.6494614003590668e-05, + "loss": 0.3236, + "step": 36709 + }, + { + "epoch": 1.41, + "learning_rate": 2.6493972813541933e-05, + "loss": 0.3229, + "step": 36710 + }, + { + "epoch": 1.41, + "learning_rate": 2.6493331623493205e-05, + "loss": 0.4591, + "step": 36711 + }, + { + "epoch": 1.41, + "learning_rate": 2.6492690433444473e-05, + "loss": 0.4493, + "step": 36712 + }, + { + "epoch": 1.41, + "learning_rate": 2.6492049243395745e-05, + "loss": 0.2671, + "step": 36713 + }, + { + "epoch": 1.41, + "learning_rate": 2.649140805334701e-05, + "loss": 0.3706, + "step": 36714 + }, + { + "epoch": 1.41, + "learning_rate": 2.6490766863298285e-05, + "loss": 0.2897, + "step": 36715 + }, + { + "epoch": 1.41, + "learning_rate": 2.6490125673249556e-05, + "loss": 0.3858, + "step": 36716 + }, + { + "epoch": 1.41, + "learning_rate": 2.648948448320082e-05, + "loss": 0.2768, + "step": 36717 + }, + { + "epoch": 1.41, + "learning_rate": 2.6488843293152093e-05, + "loss": 0.3325, + "step": 36718 + }, + { + "epoch": 1.41, + "learning_rate": 2.648820210310336e-05, + "loss": 0.4259, + "step": 36719 + }, + { + "epoch": 1.41, + "learning_rate": 2.6487560913054633e-05, + "loss": 0.4322, + "step": 36720 + }, + { + "epoch": 1.41, + "learning_rate": 2.6486919723005898e-05, + "loss": 0.3552, + "step": 36721 + }, + { + "epoch": 1.41, + "learning_rate": 2.648627853295717e-05, + "loss": 0.3862, + "step": 36722 + }, + { + "epoch": 1.41, + "learning_rate": 2.6485637342908438e-05, + "loss": 0.46, + "step": 36723 + }, + { + "epoch": 1.41, + "learning_rate": 2.648499615285971e-05, + "loss": 0.3247, + "step": 36724 + }, + { + "epoch": 1.41, + "learning_rate": 2.648435496281098e-05, + "loss": 0.4178, + "step": 36725 + }, + { + "epoch": 1.41, + "learning_rate": 2.6483713772762246e-05, + "loss": 0.3625, + "step": 36726 + }, + { + "epoch": 1.41, + "learning_rate": 2.6483072582713518e-05, + "loss": 0.3558, + "step": 36727 + }, + { + "epoch": 1.41, + "learning_rate": 2.6482431392664786e-05, + "loss": 0.3917, + "step": 36728 + }, + { + "epoch": 1.41, + "learning_rate": 2.6481790202616058e-05, + "loss": 0.3347, + "step": 36729 + }, + { + "epoch": 1.41, + "learning_rate": 2.6481149012567323e-05, + "loss": 0.4475, + "step": 36730 + }, + { + "epoch": 1.41, + "learning_rate": 2.6480507822518595e-05, + "loss": 0.3679, + "step": 36731 + }, + { + "epoch": 1.41, + "learning_rate": 2.6479866632469863e-05, + "loss": 0.3227, + "step": 36732 + }, + { + "epoch": 1.41, + "learning_rate": 2.6479225442421135e-05, + "loss": 0.3088, + "step": 36733 + }, + { + "epoch": 1.41, + "learning_rate": 2.6478584252372407e-05, + "loss": 0.3945, + "step": 36734 + }, + { + "epoch": 1.41, + "learning_rate": 2.6477943062323675e-05, + "loss": 0.411, + "step": 36735 + }, + { + "epoch": 1.41, + "learning_rate": 2.6477301872274947e-05, + "loss": 0.374, + "step": 36736 + }, + { + "epoch": 1.41, + "learning_rate": 2.647666068222621e-05, + "loss": 0.423, + "step": 36737 + }, + { + "epoch": 1.41, + "learning_rate": 2.6476019492177483e-05, + "loss": 0.2462, + "step": 36738 + }, + { + "epoch": 1.41, + "learning_rate": 2.647537830212875e-05, + "loss": 0.2504, + "step": 36739 + }, + { + "epoch": 1.41, + "learning_rate": 2.6474737112080023e-05, + "loss": 0.4493, + "step": 36740 + }, + { + "epoch": 1.41, + "learning_rate": 2.6474095922031288e-05, + "loss": 0.4773, + "step": 36741 + }, + { + "epoch": 1.41, + "learning_rate": 2.647345473198256e-05, + "loss": 0.2583, + "step": 36742 + }, + { + "epoch": 1.41, + "learning_rate": 2.6472813541933832e-05, + "loss": 0.3289, + "step": 36743 + }, + { + "epoch": 1.41, + "learning_rate": 2.64721723518851e-05, + "loss": 0.3833, + "step": 36744 + }, + { + "epoch": 1.41, + "learning_rate": 2.6471531161836372e-05, + "loss": 0.4778, + "step": 36745 + }, + { + "epoch": 1.41, + "learning_rate": 2.6470889971787637e-05, + "loss": 0.4932, + "step": 36746 + }, + { + "epoch": 1.41, + "learning_rate": 2.647024878173891e-05, + "loss": 0.3881, + "step": 36747 + }, + { + "epoch": 1.41, + "learning_rate": 2.6469607591690177e-05, + "loss": 0.4745, + "step": 36748 + }, + { + "epoch": 1.41, + "learning_rate": 2.646896640164145e-05, + "loss": 0.3191, + "step": 36749 + }, + { + "epoch": 1.41, + "learning_rate": 2.6468325211592713e-05, + "loss": 0.2942, + "step": 36750 + }, + { + "epoch": 1.41, + "learning_rate": 2.646768402154399e-05, + "loss": 0.3244, + "step": 36751 + }, + { + "epoch": 1.41, + "learning_rate": 2.646704283149526e-05, + "loss": 0.2521, + "step": 36752 + }, + { + "epoch": 1.41, + "learning_rate": 2.6466401641446525e-05, + "loss": 0.6089, + "step": 36753 + }, + { + "epoch": 1.41, + "learning_rate": 2.6465760451397797e-05, + "loss": 0.3304, + "step": 36754 + }, + { + "epoch": 1.41, + "learning_rate": 2.6465119261349065e-05, + "loss": 0.3926, + "step": 36755 + }, + { + "epoch": 1.41, + "learning_rate": 2.6464478071300337e-05, + "loss": 0.1925, + "step": 36756 + }, + { + "epoch": 1.41, + "learning_rate": 2.6463836881251602e-05, + "loss": 0.3709, + "step": 36757 + }, + { + "epoch": 1.41, + "learning_rate": 2.6463195691202874e-05, + "loss": 0.281, + "step": 36758 + }, + { + "epoch": 1.41, + "learning_rate": 2.6462554501154142e-05, + "loss": 0.4321, + "step": 36759 + }, + { + "epoch": 1.41, + "learning_rate": 2.6461913311105414e-05, + "loss": 0.2993, + "step": 36760 + }, + { + "epoch": 1.41, + "learning_rate": 2.6461272121056685e-05, + "loss": 0.5022, + "step": 36761 + }, + { + "epoch": 1.41, + "learning_rate": 2.646063093100795e-05, + "loss": 0.3082, + "step": 36762 + }, + { + "epoch": 1.41, + "learning_rate": 2.6459989740959222e-05, + "loss": 0.3043, + "step": 36763 + }, + { + "epoch": 1.41, + "learning_rate": 2.645934855091049e-05, + "loss": 0.3228, + "step": 36764 + }, + { + "epoch": 1.41, + "learning_rate": 2.6458707360861762e-05, + "loss": 0.3456, + "step": 36765 + }, + { + "epoch": 1.41, + "learning_rate": 2.6458066170813027e-05, + "loss": 0.2953, + "step": 36766 + }, + { + "epoch": 1.41, + "learning_rate": 2.6457424980764302e-05, + "loss": 0.5044, + "step": 36767 + }, + { + "epoch": 1.41, + "learning_rate": 2.6456783790715567e-05, + "loss": 0.2913, + "step": 36768 + }, + { + "epoch": 1.41, + "learning_rate": 2.645614260066684e-05, + "loss": 0.4276, + "step": 36769 + }, + { + "epoch": 1.41, + "learning_rate": 2.645550141061811e-05, + "loss": 0.3106, + "step": 36770 + }, + { + "epoch": 1.41, + "learning_rate": 2.645486022056938e-05, + "loss": 0.4238, + "step": 36771 + }, + { + "epoch": 1.41, + "learning_rate": 2.645421903052065e-05, + "loss": 0.3894, + "step": 36772 + }, + { + "epoch": 1.41, + "learning_rate": 2.6453577840471915e-05, + "loss": 0.3254, + "step": 36773 + }, + { + "epoch": 1.41, + "learning_rate": 2.6452936650423187e-05, + "loss": 0.3729, + "step": 36774 + }, + { + "epoch": 1.41, + "learning_rate": 2.6452295460374456e-05, + "loss": 0.4559, + "step": 36775 + }, + { + "epoch": 1.41, + "learning_rate": 2.6451654270325727e-05, + "loss": 0.2563, + "step": 36776 + }, + { + "epoch": 1.41, + "learning_rate": 2.6451013080276992e-05, + "loss": 0.3278, + "step": 36777 + }, + { + "epoch": 1.41, + "learning_rate": 2.6450371890228264e-05, + "loss": 0.2935, + "step": 36778 + }, + { + "epoch": 1.41, + "learning_rate": 2.6449730700179536e-05, + "loss": 0.2931, + "step": 36779 + }, + { + "epoch": 1.41, + "learning_rate": 2.6449089510130804e-05, + "loss": 0.3693, + "step": 36780 + }, + { + "epoch": 1.41, + "learning_rate": 2.6448448320082076e-05, + "loss": 0.3978, + "step": 36781 + }, + { + "epoch": 1.41, + "learning_rate": 2.644780713003334e-05, + "loss": 0.2989, + "step": 36782 + }, + { + "epoch": 1.41, + "learning_rate": 2.6447165939984612e-05, + "loss": 0.3556, + "step": 36783 + }, + { + "epoch": 1.41, + "learning_rate": 2.644652474993588e-05, + "loss": 0.3095, + "step": 36784 + }, + { + "epoch": 1.41, + "learning_rate": 2.6445883559887152e-05, + "loss": 0.5294, + "step": 36785 + }, + { + "epoch": 1.41, + "learning_rate": 2.6445242369838417e-05, + "loss": 0.3887, + "step": 36786 + }, + { + "epoch": 1.41, + "learning_rate": 2.6444601179789692e-05, + "loss": 0.592, + "step": 36787 + }, + { + "epoch": 1.41, + "learning_rate": 2.6443959989740964e-05, + "loss": 0.2732, + "step": 36788 + }, + { + "epoch": 1.41, + "learning_rate": 2.644331879969223e-05, + "loss": 0.2858, + "step": 36789 + }, + { + "epoch": 1.41, + "learning_rate": 2.64426776096435e-05, + "loss": 0.3616, + "step": 36790 + }, + { + "epoch": 1.41, + "learning_rate": 2.644203641959477e-05, + "loss": 0.4044, + "step": 36791 + }, + { + "epoch": 1.41, + "learning_rate": 2.644139522954604e-05, + "loss": 0.3637, + "step": 36792 + }, + { + "epoch": 1.41, + "learning_rate": 2.6440754039497306e-05, + "loss": 0.2678, + "step": 36793 + }, + { + "epoch": 1.41, + "learning_rate": 2.6440112849448577e-05, + "loss": 0.3662, + "step": 36794 + }, + { + "epoch": 1.41, + "learning_rate": 2.6439471659399846e-05, + "loss": 0.3334, + "step": 36795 + }, + { + "epoch": 1.41, + "learning_rate": 2.6438830469351118e-05, + "loss": 0.3272, + "step": 36796 + }, + { + "epoch": 1.41, + "learning_rate": 2.643818927930239e-05, + "loss": 0.4002, + "step": 36797 + }, + { + "epoch": 1.41, + "learning_rate": 2.6437548089253654e-05, + "loss": 0.4209, + "step": 36798 + }, + { + "epoch": 1.41, + "learning_rate": 2.6436906899204926e-05, + "loss": 0.3958, + "step": 36799 + }, + { + "epoch": 1.41, + "learning_rate": 2.6436265709156194e-05, + "loss": 0.4075, + "step": 36800 + }, + { + "epoch": 1.41, + "learning_rate": 2.6435624519107466e-05, + "loss": 0.3371, + "step": 36801 + }, + { + "epoch": 1.41, + "learning_rate": 2.643498332905873e-05, + "loss": 0.2667, + "step": 36802 + }, + { + "epoch": 1.41, + "learning_rate": 2.6434342139010006e-05, + "loss": 0.339, + "step": 36803 + }, + { + "epoch": 1.41, + "learning_rate": 2.643370094896127e-05, + "loss": 0.4191, + "step": 36804 + }, + { + "epoch": 1.42, + "learning_rate": 2.6433059758912543e-05, + "loss": 0.3505, + "step": 36805 + }, + { + "epoch": 1.42, + "learning_rate": 2.6432418568863814e-05, + "loss": 0.3343, + "step": 36806 + }, + { + "epoch": 1.42, + "learning_rate": 2.6431777378815083e-05, + "loss": 0.4305, + "step": 36807 + }, + { + "epoch": 1.42, + "learning_rate": 2.6431136188766354e-05, + "loss": 0.3541, + "step": 36808 + }, + { + "epoch": 1.42, + "learning_rate": 2.643049499871762e-05, + "loss": 0.4009, + "step": 36809 + }, + { + "epoch": 1.42, + "learning_rate": 2.642985380866889e-05, + "loss": 0.3515, + "step": 36810 + }, + { + "epoch": 1.42, + "learning_rate": 2.642921261862016e-05, + "loss": 0.3388, + "step": 36811 + }, + { + "epoch": 1.42, + "learning_rate": 2.642857142857143e-05, + "loss": 0.4367, + "step": 36812 + }, + { + "epoch": 1.42, + "learning_rate": 2.6427930238522696e-05, + "loss": 0.4968, + "step": 36813 + }, + { + "epoch": 1.42, + "learning_rate": 2.6427289048473968e-05, + "loss": 0.329, + "step": 36814 + }, + { + "epoch": 1.42, + "learning_rate": 2.642664785842524e-05, + "loss": 0.439, + "step": 36815 + }, + { + "epoch": 1.42, + "learning_rate": 2.6426006668376508e-05, + "loss": 0.2774, + "step": 36816 + }, + { + "epoch": 1.42, + "learning_rate": 2.642536547832778e-05, + "loss": 0.4689, + "step": 36817 + }, + { + "epoch": 1.42, + "learning_rate": 2.6424724288279044e-05, + "loss": 0.408, + "step": 36818 + }, + { + "epoch": 1.42, + "learning_rate": 2.6424083098230316e-05, + "loss": 0.3896, + "step": 36819 + }, + { + "epoch": 1.42, + "learning_rate": 2.6423441908181585e-05, + "loss": 0.4046, + "step": 36820 + }, + { + "epoch": 1.42, + "learning_rate": 2.6422800718132856e-05, + "loss": 0.4232, + "step": 36821 + }, + { + "epoch": 1.42, + "learning_rate": 2.642215952808412e-05, + "loss": 0.3073, + "step": 36822 + }, + { + "epoch": 1.42, + "learning_rate": 2.6421518338035396e-05, + "loss": 0.2559, + "step": 36823 + }, + { + "epoch": 1.42, + "learning_rate": 2.6420877147986668e-05, + "loss": 0.3759, + "step": 36824 + }, + { + "epoch": 1.42, + "learning_rate": 2.6420235957937933e-05, + "loss": 0.47, + "step": 36825 + }, + { + "epoch": 1.42, + "learning_rate": 2.6419594767889205e-05, + "loss": 0.3561, + "step": 36826 + }, + { + "epoch": 1.42, + "learning_rate": 2.6418953577840473e-05, + "loss": 0.3009, + "step": 36827 + }, + { + "epoch": 1.42, + "learning_rate": 2.6418312387791745e-05, + "loss": 0.3632, + "step": 36828 + }, + { + "epoch": 1.42, + "learning_rate": 2.641767119774301e-05, + "loss": 0.4783, + "step": 36829 + }, + { + "epoch": 1.42, + "learning_rate": 2.641703000769428e-05, + "loss": 0.4451, + "step": 36830 + }, + { + "epoch": 1.42, + "learning_rate": 2.641638881764555e-05, + "loss": 0.3701, + "step": 36831 + }, + { + "epoch": 1.42, + "learning_rate": 2.641574762759682e-05, + "loss": 0.2885, + "step": 36832 + }, + { + "epoch": 1.42, + "learning_rate": 2.6415106437548093e-05, + "loss": 0.4318, + "step": 36833 + }, + { + "epoch": 1.42, + "learning_rate": 2.6414465247499358e-05, + "loss": 0.2702, + "step": 36834 + }, + { + "epoch": 1.42, + "learning_rate": 2.641382405745063e-05, + "loss": 0.4433, + "step": 36835 + }, + { + "epoch": 1.42, + "learning_rate": 2.6413182867401898e-05, + "loss": 0.3362, + "step": 36836 + }, + { + "epoch": 1.42, + "learning_rate": 2.641254167735317e-05, + "loss": 0.3898, + "step": 36837 + }, + { + "epoch": 1.42, + "learning_rate": 2.6411900487304435e-05, + "loss": 0.3784, + "step": 36838 + }, + { + "epoch": 1.42, + "learning_rate": 2.641125929725571e-05, + "loss": 0.5287, + "step": 36839 + }, + { + "epoch": 1.42, + "learning_rate": 2.6410618107206975e-05, + "loss": 0.4927, + "step": 36840 + }, + { + "epoch": 1.42, + "learning_rate": 2.6409976917158247e-05, + "loss": 0.3637, + "step": 36841 + }, + { + "epoch": 1.42, + "learning_rate": 2.6409335727109518e-05, + "loss": 0.3974, + "step": 36842 + }, + { + "epoch": 1.42, + "learning_rate": 2.6408694537060787e-05, + "loss": 0.5066, + "step": 36843 + }, + { + "epoch": 1.42, + "learning_rate": 2.6408053347012058e-05, + "loss": 0.3169, + "step": 36844 + }, + { + "epoch": 1.42, + "learning_rate": 2.6407412156963323e-05, + "loss": 0.455, + "step": 36845 + }, + { + "epoch": 1.42, + "learning_rate": 2.6406770966914595e-05, + "loss": 0.3792, + "step": 36846 + }, + { + "epoch": 1.42, + "learning_rate": 2.6406129776865863e-05, + "loss": 0.38, + "step": 36847 + }, + { + "epoch": 1.42, + "learning_rate": 2.6405488586817135e-05, + "loss": 0.2826, + "step": 36848 + }, + { + "epoch": 1.42, + "learning_rate": 2.64048473967684e-05, + "loss": 0.3549, + "step": 36849 + }, + { + "epoch": 1.42, + "learning_rate": 2.640420620671967e-05, + "loss": 0.4465, + "step": 36850 + }, + { + "epoch": 1.42, + "learning_rate": 2.6403565016670943e-05, + "loss": 0.3253, + "step": 36851 + }, + { + "epoch": 1.42, + "learning_rate": 2.640292382662221e-05, + "loss": 0.2639, + "step": 36852 + }, + { + "epoch": 1.42, + "learning_rate": 2.6402282636573483e-05, + "loss": 0.3074, + "step": 36853 + }, + { + "epoch": 1.42, + "learning_rate": 2.640164144652475e-05, + "loss": 0.3235, + "step": 36854 + }, + { + "epoch": 1.42, + "learning_rate": 2.640100025647602e-05, + "loss": 0.387, + "step": 36855 + }, + { + "epoch": 1.42, + "learning_rate": 2.640035906642729e-05, + "loss": 0.4188, + "step": 36856 + }, + { + "epoch": 1.42, + "learning_rate": 2.639971787637856e-05, + "loss": 0.2055, + "step": 36857 + }, + { + "epoch": 1.42, + "learning_rate": 2.6399076686329825e-05, + "loss": 0.4498, + "step": 36858 + }, + { + "epoch": 1.42, + "learning_rate": 2.63984354962811e-05, + "loss": 0.269, + "step": 36859 + }, + { + "epoch": 1.42, + "learning_rate": 2.6397794306232372e-05, + "loss": 0.4891, + "step": 36860 + }, + { + "epoch": 1.42, + "learning_rate": 2.6397153116183637e-05, + "loss": 0.523, + "step": 36861 + }, + { + "epoch": 1.42, + "learning_rate": 2.639651192613491e-05, + "loss": 0.4532, + "step": 36862 + }, + { + "epoch": 1.42, + "learning_rate": 2.6395870736086177e-05, + "loss": 0.3987, + "step": 36863 + }, + { + "epoch": 1.42, + "learning_rate": 2.639522954603745e-05, + "loss": 0.4827, + "step": 36864 + }, + { + "epoch": 1.42, + "learning_rate": 2.6394588355988713e-05, + "loss": 0.5105, + "step": 36865 + }, + { + "epoch": 1.42, + "learning_rate": 2.6393947165939985e-05, + "loss": 0.3412, + "step": 36866 + }, + { + "epoch": 1.42, + "learning_rate": 2.6393305975891257e-05, + "loss": 0.2704, + "step": 36867 + }, + { + "epoch": 1.42, + "learning_rate": 2.6392664785842525e-05, + "loss": 0.3249, + "step": 36868 + }, + { + "epoch": 1.42, + "learning_rate": 2.6392023595793797e-05, + "loss": 0.3637, + "step": 36869 + }, + { + "epoch": 1.42, + "learning_rate": 2.6391382405745062e-05, + "loss": 0.5628, + "step": 36870 + }, + { + "epoch": 1.42, + "learning_rate": 2.6390741215696334e-05, + "loss": 0.4255, + "step": 36871 + }, + { + "epoch": 1.42, + "learning_rate": 2.6390100025647602e-05, + "loss": 0.4156, + "step": 36872 + }, + { + "epoch": 1.42, + "learning_rate": 2.6389458835598874e-05, + "loss": 0.2168, + "step": 36873 + }, + { + "epoch": 1.42, + "learning_rate": 2.638881764555014e-05, + "loss": 0.405, + "step": 36874 + }, + { + "epoch": 1.42, + "learning_rate": 2.6388176455501414e-05, + "loss": 0.4396, + "step": 36875 + }, + { + "epoch": 1.42, + "learning_rate": 2.6387535265452685e-05, + "loss": 0.3732, + "step": 36876 + }, + { + "epoch": 1.42, + "learning_rate": 2.638689407540395e-05, + "loss": 0.3174, + "step": 36877 + }, + { + "epoch": 1.42, + "learning_rate": 2.6386252885355222e-05, + "loss": 0.4383, + "step": 36878 + }, + { + "epoch": 1.42, + "learning_rate": 2.638561169530649e-05, + "loss": 0.4888, + "step": 36879 + }, + { + "epoch": 1.42, + "learning_rate": 2.6384970505257762e-05, + "loss": 0.3192, + "step": 36880 + }, + { + "epoch": 1.42, + "learning_rate": 2.6384329315209027e-05, + "loss": 0.4966, + "step": 36881 + }, + { + "epoch": 1.42, + "learning_rate": 2.63836881251603e-05, + "loss": 0.4144, + "step": 36882 + }, + { + "epoch": 1.42, + "learning_rate": 2.6383046935111567e-05, + "loss": 0.3897, + "step": 36883 + }, + { + "epoch": 1.42, + "learning_rate": 2.638240574506284e-05, + "loss": 0.4596, + "step": 36884 + }, + { + "epoch": 1.42, + "learning_rate": 2.638176455501411e-05, + "loss": 0.2867, + "step": 36885 + }, + { + "epoch": 1.42, + "learning_rate": 2.6381123364965375e-05, + "loss": 0.4521, + "step": 36886 + }, + { + "epoch": 1.42, + "learning_rate": 2.6380482174916647e-05, + "loss": 0.406, + "step": 36887 + }, + { + "epoch": 1.42, + "learning_rate": 2.6379840984867916e-05, + "loss": 0.5246, + "step": 36888 + }, + { + "epoch": 1.42, + "learning_rate": 2.6379199794819187e-05, + "loss": 0.3859, + "step": 36889 + }, + { + "epoch": 1.42, + "learning_rate": 2.6378558604770452e-05, + "loss": 0.3551, + "step": 36890 + }, + { + "epoch": 1.42, + "learning_rate": 2.6377917414721727e-05, + "loss": 0.5157, + "step": 36891 + }, + { + "epoch": 1.42, + "learning_rate": 2.6377276224672992e-05, + "loss": 0.4452, + "step": 36892 + }, + { + "epoch": 1.42, + "learning_rate": 2.6376635034624264e-05, + "loss": 0.3994, + "step": 36893 + }, + { + "epoch": 1.42, + "learning_rate": 2.6375993844575536e-05, + "loss": 0.2738, + "step": 36894 + }, + { + "epoch": 1.42, + "learning_rate": 2.6375352654526804e-05, + "loss": 0.3271, + "step": 36895 + }, + { + "epoch": 1.42, + "learning_rate": 2.6374711464478076e-05, + "loss": 0.3375, + "step": 36896 + }, + { + "epoch": 1.42, + "learning_rate": 2.637407027442934e-05, + "loss": 0.4467, + "step": 36897 + }, + { + "epoch": 1.42, + "learning_rate": 2.6373429084380612e-05, + "loss": 0.3899, + "step": 36898 + }, + { + "epoch": 1.42, + "learning_rate": 2.637278789433188e-05, + "loss": 0.3608, + "step": 36899 + }, + { + "epoch": 1.42, + "learning_rate": 2.6372146704283152e-05, + "loss": 0.4442, + "step": 36900 + }, + { + "epoch": 1.42, + "learning_rate": 2.6371505514234417e-05, + "loss": 0.3834, + "step": 36901 + }, + { + "epoch": 1.42, + "learning_rate": 2.637086432418569e-05, + "loss": 0.4075, + "step": 36902 + }, + { + "epoch": 1.42, + "learning_rate": 2.637022313413696e-05, + "loss": 0.3123, + "step": 36903 + }, + { + "epoch": 1.42, + "learning_rate": 2.636958194408823e-05, + "loss": 0.3449, + "step": 36904 + }, + { + "epoch": 1.42, + "learning_rate": 2.63689407540395e-05, + "loss": 0.2945, + "step": 36905 + }, + { + "epoch": 1.42, + "learning_rate": 2.6368299563990766e-05, + "loss": 0.3656, + "step": 36906 + }, + { + "epoch": 1.42, + "learning_rate": 2.6367658373942037e-05, + "loss": 0.3528, + "step": 36907 + }, + { + "epoch": 1.42, + "learning_rate": 2.6367017183893306e-05, + "loss": 0.4582, + "step": 36908 + }, + { + "epoch": 1.42, + "learning_rate": 2.6366375993844578e-05, + "loss": 0.311, + "step": 36909 + }, + { + "epoch": 1.42, + "learning_rate": 2.6365734803795842e-05, + "loss": 0.3195, + "step": 36910 + }, + { + "epoch": 1.42, + "learning_rate": 2.6365093613747118e-05, + "loss": 0.4791, + "step": 36911 + }, + { + "epoch": 1.42, + "learning_rate": 2.636445242369839e-05, + "loss": 0.3209, + "step": 36912 + }, + { + "epoch": 1.42, + "learning_rate": 2.6363811233649654e-05, + "loss": 0.3284, + "step": 36913 + }, + { + "epoch": 1.42, + "learning_rate": 2.6363170043600926e-05, + "loss": 0.3445, + "step": 36914 + }, + { + "epoch": 1.42, + "learning_rate": 2.6362528853552194e-05, + "loss": 0.3983, + "step": 36915 + }, + { + "epoch": 1.42, + "learning_rate": 2.6361887663503466e-05, + "loss": 0.3539, + "step": 36916 + }, + { + "epoch": 1.42, + "learning_rate": 2.636124647345473e-05, + "loss": 0.3282, + "step": 36917 + }, + { + "epoch": 1.42, + "learning_rate": 2.6360605283406003e-05, + "loss": 0.344, + "step": 36918 + }, + { + "epoch": 1.42, + "learning_rate": 2.635996409335727e-05, + "loss": 0.3295, + "step": 36919 + }, + { + "epoch": 1.42, + "learning_rate": 2.6359322903308543e-05, + "loss": 0.4072, + "step": 36920 + }, + { + "epoch": 1.42, + "learning_rate": 2.6358681713259814e-05, + "loss": 0.3453, + "step": 36921 + }, + { + "epoch": 1.42, + "learning_rate": 2.635804052321108e-05, + "loss": 0.4331, + "step": 36922 + }, + { + "epoch": 1.42, + "learning_rate": 2.635739933316235e-05, + "loss": 0.3158, + "step": 36923 + }, + { + "epoch": 1.42, + "learning_rate": 2.635675814311362e-05, + "loss": 0.3886, + "step": 36924 + }, + { + "epoch": 1.42, + "learning_rate": 2.635611695306489e-05, + "loss": 0.3323, + "step": 36925 + }, + { + "epoch": 1.42, + "learning_rate": 2.6355475763016156e-05, + "loss": 0.3419, + "step": 36926 + }, + { + "epoch": 1.42, + "learning_rate": 2.635483457296743e-05, + "loss": 0.4204, + "step": 36927 + }, + { + "epoch": 1.42, + "learning_rate": 2.6354193382918696e-05, + "loss": 0.3239, + "step": 36928 + }, + { + "epoch": 1.42, + "learning_rate": 2.6353552192869968e-05, + "loss": 0.4717, + "step": 36929 + }, + { + "epoch": 1.42, + "learning_rate": 2.635291100282124e-05, + "loss": 0.3476, + "step": 36930 + }, + { + "epoch": 1.42, + "learning_rate": 2.6352269812772508e-05, + "loss": 0.4297, + "step": 36931 + }, + { + "epoch": 1.42, + "learning_rate": 2.635162862272378e-05, + "loss": 0.3213, + "step": 36932 + }, + { + "epoch": 1.42, + "learning_rate": 2.6350987432675045e-05, + "loss": 0.3864, + "step": 36933 + }, + { + "epoch": 1.42, + "learning_rate": 2.6350346242626316e-05, + "loss": 0.3629, + "step": 36934 + }, + { + "epoch": 1.42, + "learning_rate": 2.6349705052577585e-05, + "loss": 0.2262, + "step": 36935 + }, + { + "epoch": 1.42, + "learning_rate": 2.6349063862528856e-05, + "loss": 0.4633, + "step": 36936 + }, + { + "epoch": 1.42, + "learning_rate": 2.634842267248012e-05, + "loss": 0.4195, + "step": 36937 + }, + { + "epoch": 1.42, + "learning_rate": 2.6347781482431393e-05, + "loss": 0.3364, + "step": 36938 + }, + { + "epoch": 1.42, + "learning_rate": 2.6347140292382665e-05, + "loss": 0.2798, + "step": 36939 + }, + { + "epoch": 1.42, + "learning_rate": 2.6346499102333933e-05, + "loss": 0.376, + "step": 36940 + }, + { + "epoch": 1.42, + "learning_rate": 2.6345857912285205e-05, + "loss": 0.4024, + "step": 36941 + }, + { + "epoch": 1.42, + "learning_rate": 2.634521672223647e-05, + "loss": 0.2598, + "step": 36942 + }, + { + "epoch": 1.42, + "learning_rate": 2.634457553218774e-05, + "loss": 0.272, + "step": 36943 + }, + { + "epoch": 1.42, + "learning_rate": 2.634393434213901e-05, + "loss": 0.33, + "step": 36944 + }, + { + "epoch": 1.42, + "learning_rate": 2.634329315209028e-05, + "loss": 0.4332, + "step": 36945 + }, + { + "epoch": 1.42, + "learning_rate": 2.6342651962041546e-05, + "loss": 0.2928, + "step": 36946 + }, + { + "epoch": 1.42, + "learning_rate": 2.634201077199282e-05, + "loss": 0.4048, + "step": 36947 + }, + { + "epoch": 1.42, + "learning_rate": 2.6341369581944093e-05, + "loss": 0.3939, + "step": 36948 + }, + { + "epoch": 1.42, + "learning_rate": 2.6340728391895358e-05, + "loss": 0.3883, + "step": 36949 + }, + { + "epoch": 1.42, + "learning_rate": 2.634008720184663e-05, + "loss": 0.435, + "step": 36950 + }, + { + "epoch": 1.42, + "learning_rate": 2.6339446011797898e-05, + "loss": 0.3691, + "step": 36951 + }, + { + "epoch": 1.42, + "learning_rate": 2.633880482174917e-05, + "loss": 0.3006, + "step": 36952 + }, + { + "epoch": 1.42, + "learning_rate": 2.6338163631700435e-05, + "loss": 0.3696, + "step": 36953 + }, + { + "epoch": 1.42, + "learning_rate": 2.6337522441651707e-05, + "loss": 0.3904, + "step": 36954 + }, + { + "epoch": 1.42, + "learning_rate": 2.6336881251602975e-05, + "loss": 0.3128, + "step": 36955 + }, + { + "epoch": 1.42, + "learning_rate": 2.6336240061554247e-05, + "loss": 0.2834, + "step": 36956 + }, + { + "epoch": 1.42, + "learning_rate": 2.6335598871505518e-05, + "loss": 0.3978, + "step": 36957 + }, + { + "epoch": 1.42, + "learning_rate": 2.6334957681456783e-05, + "loss": 0.417, + "step": 36958 + }, + { + "epoch": 1.42, + "learning_rate": 2.6334316491408055e-05, + "loss": 0.4276, + "step": 36959 + }, + { + "epoch": 1.42, + "learning_rate": 2.6333675301359323e-05, + "loss": 0.3503, + "step": 36960 + }, + { + "epoch": 1.42, + "learning_rate": 2.6333034111310595e-05, + "loss": 0.3487, + "step": 36961 + }, + { + "epoch": 1.42, + "learning_rate": 2.633239292126186e-05, + "loss": 0.4631, + "step": 36962 + }, + { + "epoch": 1.42, + "learning_rate": 2.6331751731213135e-05, + "loss": 0.3711, + "step": 36963 + }, + { + "epoch": 1.42, + "learning_rate": 2.63311105411644e-05, + "loss": 0.3587, + "step": 36964 + }, + { + "epoch": 1.42, + "learning_rate": 2.6330469351115672e-05, + "loss": 0.4133, + "step": 36965 + }, + { + "epoch": 1.42, + "learning_rate": 2.6329828161066943e-05, + "loss": 0.4221, + "step": 36966 + }, + { + "epoch": 1.42, + "learning_rate": 2.6329186971018212e-05, + "loss": 0.2554, + "step": 36967 + }, + { + "epoch": 1.42, + "learning_rate": 2.6328545780969483e-05, + "loss": 0.4242, + "step": 36968 + }, + { + "epoch": 1.42, + "learning_rate": 2.632790459092075e-05, + "loss": 0.3926, + "step": 36969 + }, + { + "epoch": 1.42, + "learning_rate": 2.632726340087202e-05, + "loss": 0.5231, + "step": 36970 + }, + { + "epoch": 1.42, + "learning_rate": 2.632662221082329e-05, + "loss": 0.3975, + "step": 36971 + }, + { + "epoch": 1.42, + "learning_rate": 2.632598102077456e-05, + "loss": 0.3984, + "step": 36972 + }, + { + "epoch": 1.42, + "learning_rate": 2.6325339830725825e-05, + "loss": 0.2759, + "step": 36973 + }, + { + "epoch": 1.42, + "learning_rate": 2.6324698640677097e-05, + "loss": 0.3474, + "step": 36974 + }, + { + "epoch": 1.42, + "learning_rate": 2.632405745062837e-05, + "loss": 0.5051, + "step": 36975 + }, + { + "epoch": 1.42, + "learning_rate": 2.6323416260579637e-05, + "loss": 0.3912, + "step": 36976 + }, + { + "epoch": 1.42, + "learning_rate": 2.632277507053091e-05, + "loss": 0.3283, + "step": 36977 + }, + { + "epoch": 1.42, + "learning_rate": 2.6322133880482174e-05, + "loss": 0.3143, + "step": 36978 + }, + { + "epoch": 1.42, + "learning_rate": 2.6321492690433445e-05, + "loss": 0.2722, + "step": 36979 + }, + { + "epoch": 1.42, + "learning_rate": 2.6320851500384714e-05, + "loss": 0.2926, + "step": 36980 + }, + { + "epoch": 1.42, + "learning_rate": 2.6320210310335985e-05, + "loss": 0.4347, + "step": 36981 + }, + { + "epoch": 1.42, + "learning_rate": 2.631956912028725e-05, + "loss": 0.367, + "step": 36982 + }, + { + "epoch": 1.42, + "learning_rate": 2.6318927930238525e-05, + "loss": 0.3608, + "step": 36983 + }, + { + "epoch": 1.42, + "learning_rate": 2.6318286740189797e-05, + "loss": 0.4186, + "step": 36984 + }, + { + "epoch": 1.42, + "learning_rate": 2.6317645550141062e-05, + "loss": 0.312, + "step": 36985 + }, + { + "epoch": 1.42, + "learning_rate": 2.6317004360092334e-05, + "loss": 0.4414, + "step": 36986 + }, + { + "epoch": 1.42, + "learning_rate": 2.6316363170043602e-05, + "loss": 0.3647, + "step": 36987 + }, + { + "epoch": 1.42, + "learning_rate": 2.6315721979994874e-05, + "loss": 0.3439, + "step": 36988 + }, + { + "epoch": 1.42, + "learning_rate": 2.631508078994614e-05, + "loss": 0.4058, + "step": 36989 + }, + { + "epoch": 1.42, + "learning_rate": 2.631443959989741e-05, + "loss": 0.3757, + "step": 36990 + }, + { + "epoch": 1.42, + "learning_rate": 2.631379840984868e-05, + "loss": 0.4347, + "step": 36991 + }, + { + "epoch": 1.42, + "learning_rate": 2.631315721979995e-05, + "loss": 0.2396, + "step": 36992 + }, + { + "epoch": 1.42, + "learning_rate": 2.6312516029751222e-05, + "loss": 0.3504, + "step": 36993 + }, + { + "epoch": 1.42, + "learning_rate": 2.6311874839702487e-05, + "loss": 0.4663, + "step": 36994 + }, + { + "epoch": 1.42, + "learning_rate": 2.631123364965376e-05, + "loss": 0.3349, + "step": 36995 + }, + { + "epoch": 1.42, + "learning_rate": 2.6310592459605027e-05, + "loss": 0.3529, + "step": 36996 + }, + { + "epoch": 1.42, + "learning_rate": 2.63099512695563e-05, + "loss": 0.3752, + "step": 36997 + }, + { + "epoch": 1.42, + "learning_rate": 2.6309310079507564e-05, + "loss": 0.3774, + "step": 36998 + }, + { + "epoch": 1.42, + "learning_rate": 2.630866888945884e-05, + "loss": 0.3674, + "step": 36999 + }, + { + "epoch": 1.42, + "learning_rate": 2.6308027699410104e-05, + "loss": 0.2929, + "step": 37000 + }, + { + "epoch": 1.42, + "learning_rate": 2.6307386509361376e-05, + "loss": 0.4475, + "step": 37001 + }, + { + "epoch": 1.42, + "learning_rate": 2.6306745319312647e-05, + "loss": 0.3829, + "step": 37002 + }, + { + "epoch": 1.42, + "learning_rate": 2.6306104129263916e-05, + "loss": 0.2988, + "step": 37003 + }, + { + "epoch": 1.42, + "learning_rate": 2.6305462939215187e-05, + "loss": 0.4108, + "step": 37004 + }, + { + "epoch": 1.42, + "learning_rate": 2.6304821749166452e-05, + "loss": 0.3863, + "step": 37005 + }, + { + "epoch": 1.42, + "learning_rate": 2.6304180559117724e-05, + "loss": 0.4441, + "step": 37006 + }, + { + "epoch": 1.42, + "learning_rate": 2.6303539369068992e-05, + "loss": 0.4942, + "step": 37007 + }, + { + "epoch": 1.42, + "learning_rate": 2.6302898179020264e-05, + "loss": 0.4393, + "step": 37008 + }, + { + "epoch": 1.42, + "learning_rate": 2.630225698897153e-05, + "loss": 0.3263, + "step": 37009 + }, + { + "epoch": 1.42, + "learning_rate": 2.63016157989228e-05, + "loss": 0.3784, + "step": 37010 + }, + { + "epoch": 1.42, + "learning_rate": 2.6300974608874072e-05, + "loss": 0.4253, + "step": 37011 + }, + { + "epoch": 1.42, + "learning_rate": 2.630033341882534e-05, + "loss": 0.299, + "step": 37012 + }, + { + "epoch": 1.42, + "learning_rate": 2.6299692228776612e-05, + "loss": 0.5264, + "step": 37013 + }, + { + "epoch": 1.42, + "learning_rate": 2.6299051038727877e-05, + "loss": 0.291, + "step": 37014 + }, + { + "epoch": 1.42, + "learning_rate": 2.629840984867915e-05, + "loss": 0.2536, + "step": 37015 + }, + { + "epoch": 1.42, + "learning_rate": 2.6297768658630417e-05, + "loss": 0.2863, + "step": 37016 + }, + { + "epoch": 1.42, + "learning_rate": 2.629712746858169e-05, + "loss": 0.3816, + "step": 37017 + }, + { + "epoch": 1.42, + "learning_rate": 2.6296486278532954e-05, + "loss": 0.4988, + "step": 37018 + }, + { + "epoch": 1.42, + "learning_rate": 2.629584508848423e-05, + "loss": 0.3817, + "step": 37019 + }, + { + "epoch": 1.42, + "learning_rate": 2.62952038984355e-05, + "loss": 0.3, + "step": 37020 + }, + { + "epoch": 1.42, + "learning_rate": 2.6294562708386766e-05, + "loss": 0.2976, + "step": 37021 + }, + { + "epoch": 1.42, + "learning_rate": 2.6293921518338038e-05, + "loss": 0.4483, + "step": 37022 + }, + { + "epoch": 1.42, + "learning_rate": 2.6293280328289306e-05, + "loss": 0.3491, + "step": 37023 + }, + { + "epoch": 1.42, + "learning_rate": 2.6292639138240578e-05, + "loss": 0.3084, + "step": 37024 + }, + { + "epoch": 1.42, + "learning_rate": 2.6291997948191843e-05, + "loss": 0.3047, + "step": 37025 + }, + { + "epoch": 1.42, + "learning_rate": 2.6291356758143114e-05, + "loss": 0.2765, + "step": 37026 + }, + { + "epoch": 1.42, + "learning_rate": 2.6290715568094383e-05, + "loss": 0.4552, + "step": 37027 + }, + { + "epoch": 1.42, + "learning_rate": 2.6290074378045654e-05, + "loss": 0.3455, + "step": 37028 + }, + { + "epoch": 1.42, + "learning_rate": 2.6289433187996926e-05, + "loss": 0.3605, + "step": 37029 + }, + { + "epoch": 1.42, + "learning_rate": 2.628879199794819e-05, + "loss": 0.2871, + "step": 37030 + }, + { + "epoch": 1.42, + "learning_rate": 2.6288150807899463e-05, + "loss": 0.377, + "step": 37031 + }, + { + "epoch": 1.42, + "learning_rate": 2.628750961785073e-05, + "loss": 0.4041, + "step": 37032 + }, + { + "epoch": 1.42, + "learning_rate": 2.6286868427802003e-05, + "loss": 0.431, + "step": 37033 + }, + { + "epoch": 1.42, + "learning_rate": 2.6286227237753268e-05, + "loss": 0.4397, + "step": 37034 + }, + { + "epoch": 1.42, + "learning_rate": 2.6285586047704543e-05, + "loss": 0.3815, + "step": 37035 + }, + { + "epoch": 1.42, + "learning_rate": 2.6284944857655808e-05, + "loss": 0.3249, + "step": 37036 + }, + { + "epoch": 1.42, + "learning_rate": 2.628430366760708e-05, + "loss": 0.3917, + "step": 37037 + }, + { + "epoch": 1.42, + "learning_rate": 2.628366247755835e-05, + "loss": 0.3187, + "step": 37038 + }, + { + "epoch": 1.42, + "learning_rate": 2.628302128750962e-05, + "loss": 0.3895, + "step": 37039 + }, + { + "epoch": 1.42, + "learning_rate": 2.628238009746089e-05, + "loss": 0.282, + "step": 37040 + }, + { + "epoch": 1.42, + "learning_rate": 2.6281738907412156e-05, + "loss": 0.3996, + "step": 37041 + }, + { + "epoch": 1.42, + "learning_rate": 2.6281097717363428e-05, + "loss": 0.3323, + "step": 37042 + }, + { + "epoch": 1.42, + "learning_rate": 2.6280456527314696e-05, + "loss": 0.4335, + "step": 37043 + }, + { + "epoch": 1.42, + "learning_rate": 2.6279815337265968e-05, + "loss": 0.3573, + "step": 37044 + }, + { + "epoch": 1.42, + "learning_rate": 2.6279174147217233e-05, + "loss": 0.4372, + "step": 37045 + }, + { + "epoch": 1.42, + "learning_rate": 2.6278532957168505e-05, + "loss": 0.3654, + "step": 37046 + }, + { + "epoch": 1.42, + "learning_rate": 2.6277891767119776e-05, + "loss": 0.2994, + "step": 37047 + }, + { + "epoch": 1.42, + "learning_rate": 2.6277250577071045e-05, + "loss": 0.3848, + "step": 37048 + }, + { + "epoch": 1.42, + "learning_rate": 2.6276609387022316e-05, + "loss": 0.2607, + "step": 37049 + }, + { + "epoch": 1.42, + "learning_rate": 2.627596819697358e-05, + "loss": 0.4275, + "step": 37050 + }, + { + "epoch": 1.42, + "learning_rate": 2.6275327006924856e-05, + "loss": 0.2577, + "step": 37051 + }, + { + "epoch": 1.42, + "learning_rate": 2.627468581687612e-05, + "loss": 0.3969, + "step": 37052 + }, + { + "epoch": 1.42, + "learning_rate": 2.6274044626827393e-05, + "loss": 0.2742, + "step": 37053 + }, + { + "epoch": 1.42, + "learning_rate": 2.6273403436778665e-05, + "loss": 0.2684, + "step": 37054 + }, + { + "epoch": 1.42, + "learning_rate": 2.6272762246729933e-05, + "loss": 0.3469, + "step": 37055 + }, + { + "epoch": 1.42, + "learning_rate": 2.6272121056681205e-05, + "loss": 0.4121, + "step": 37056 + }, + { + "epoch": 1.42, + "learning_rate": 2.627147986663247e-05, + "loss": 0.4625, + "step": 37057 + }, + { + "epoch": 1.42, + "learning_rate": 2.627083867658374e-05, + "loss": 0.3691, + "step": 37058 + }, + { + "epoch": 1.42, + "learning_rate": 2.627019748653501e-05, + "loss": 0.4106, + "step": 37059 + }, + { + "epoch": 1.42, + "learning_rate": 2.626955629648628e-05, + "loss": 0.3509, + "step": 37060 + }, + { + "epoch": 1.42, + "learning_rate": 2.6268915106437546e-05, + "loss": 0.3961, + "step": 37061 + }, + { + "epoch": 1.42, + "learning_rate": 2.6268273916388818e-05, + "loss": 0.327, + "step": 37062 + }, + { + "epoch": 1.42, + "learning_rate": 2.626763272634009e-05, + "loss": 0.5018, + "step": 37063 + }, + { + "epoch": 1.42, + "learning_rate": 2.6266991536291358e-05, + "loss": 0.3346, + "step": 37064 + }, + { + "epoch": 1.43, + "learning_rate": 2.626635034624263e-05, + "loss": 0.4158, + "step": 37065 + }, + { + "epoch": 1.43, + "learning_rate": 2.6265709156193895e-05, + "loss": 0.3852, + "step": 37066 + }, + { + "epoch": 1.43, + "learning_rate": 2.6265067966145167e-05, + "loss": 0.5222, + "step": 37067 + }, + { + "epoch": 1.43, + "learning_rate": 2.6264426776096435e-05, + "loss": 0.2923, + "step": 37068 + }, + { + "epoch": 1.43, + "learning_rate": 2.6263785586047707e-05, + "loss": 0.3515, + "step": 37069 + }, + { + "epoch": 1.43, + "learning_rate": 2.626314439599897e-05, + "loss": 0.3537, + "step": 37070 + }, + { + "epoch": 1.43, + "learning_rate": 2.6262503205950247e-05, + "loss": 0.4038, + "step": 37071 + }, + { + "epoch": 1.43, + "learning_rate": 2.626186201590152e-05, + "loss": 0.416, + "step": 37072 + }, + { + "epoch": 1.43, + "learning_rate": 2.6261220825852783e-05, + "loss": 0.3445, + "step": 37073 + }, + { + "epoch": 1.43, + "learning_rate": 2.6260579635804055e-05, + "loss": 0.4577, + "step": 37074 + }, + { + "epoch": 1.43, + "learning_rate": 2.6259938445755323e-05, + "loss": 0.4639, + "step": 37075 + }, + { + "epoch": 1.43, + "learning_rate": 2.6259297255706595e-05, + "loss": 0.4235, + "step": 37076 + }, + { + "epoch": 1.43, + "learning_rate": 2.625865606565786e-05, + "loss": 0.3109, + "step": 37077 + }, + { + "epoch": 1.43, + "learning_rate": 2.6258014875609132e-05, + "loss": 0.4434, + "step": 37078 + }, + { + "epoch": 1.43, + "learning_rate": 2.62573736855604e-05, + "loss": 0.4515, + "step": 37079 + }, + { + "epoch": 1.43, + "learning_rate": 2.6256732495511672e-05, + "loss": 0.2398, + "step": 37080 + }, + { + "epoch": 1.43, + "learning_rate": 2.6256091305462943e-05, + "loss": 0.541, + "step": 37081 + }, + { + "epoch": 1.43, + "learning_rate": 2.625545011541421e-05, + "loss": 0.4719, + "step": 37082 + }, + { + "epoch": 1.43, + "learning_rate": 2.625480892536548e-05, + "loss": 0.3306, + "step": 37083 + }, + { + "epoch": 1.43, + "learning_rate": 2.625416773531675e-05, + "loss": 0.4155, + "step": 37084 + }, + { + "epoch": 1.43, + "learning_rate": 2.625352654526802e-05, + "loss": 0.4122, + "step": 37085 + }, + { + "epoch": 1.43, + "learning_rate": 2.6252885355219285e-05, + "loss": 0.3296, + "step": 37086 + }, + { + "epoch": 1.43, + "learning_rate": 2.625224416517056e-05, + "loss": 0.4151, + "step": 37087 + }, + { + "epoch": 1.43, + "learning_rate": 2.6251602975121825e-05, + "loss": 0.2608, + "step": 37088 + }, + { + "epoch": 1.43, + "learning_rate": 2.6250961785073097e-05, + "loss": 0.2609, + "step": 37089 + }, + { + "epoch": 1.43, + "learning_rate": 2.625032059502437e-05, + "loss": 0.3908, + "step": 37090 + }, + { + "epoch": 1.43, + "learning_rate": 2.6249679404975637e-05, + "loss": 0.3557, + "step": 37091 + }, + { + "epoch": 1.43, + "learning_rate": 2.624903821492691e-05, + "loss": 0.3705, + "step": 37092 + }, + { + "epoch": 1.43, + "learning_rate": 2.6248397024878174e-05, + "loss": 0.4166, + "step": 37093 + }, + { + "epoch": 1.43, + "learning_rate": 2.6247755834829445e-05, + "loss": 0.3084, + "step": 37094 + }, + { + "epoch": 1.43, + "learning_rate": 2.6247114644780714e-05, + "loss": 0.3174, + "step": 37095 + }, + { + "epoch": 1.43, + "learning_rate": 2.6246473454731985e-05, + "loss": 0.4082, + "step": 37096 + }, + { + "epoch": 1.43, + "learning_rate": 2.624583226468325e-05, + "loss": 0.4648, + "step": 37097 + }, + { + "epoch": 1.43, + "learning_rate": 2.6245191074634522e-05, + "loss": 0.4325, + "step": 37098 + }, + { + "epoch": 1.43, + "learning_rate": 2.6244549884585794e-05, + "loss": 0.4651, + "step": 37099 + }, + { + "epoch": 1.43, + "learning_rate": 2.6243908694537062e-05, + "loss": 0.4372, + "step": 37100 + }, + { + "epoch": 1.43, + "learning_rate": 2.6243267504488334e-05, + "loss": 0.2828, + "step": 37101 + }, + { + "epoch": 1.43, + "learning_rate": 2.62426263144396e-05, + "loss": 0.3571, + "step": 37102 + }, + { + "epoch": 1.43, + "learning_rate": 2.624198512439087e-05, + "loss": 0.342, + "step": 37103 + }, + { + "epoch": 1.43, + "learning_rate": 2.624134393434214e-05, + "loss": 0.2756, + "step": 37104 + }, + { + "epoch": 1.43, + "learning_rate": 2.624070274429341e-05, + "loss": 0.2475, + "step": 37105 + }, + { + "epoch": 1.43, + "learning_rate": 2.6240061554244675e-05, + "loss": 0.4777, + "step": 37106 + }, + { + "epoch": 1.43, + "learning_rate": 2.623942036419595e-05, + "loss": 0.4047, + "step": 37107 + }, + { + "epoch": 1.43, + "learning_rate": 2.6238779174147222e-05, + "loss": 0.2944, + "step": 37108 + }, + { + "epoch": 1.43, + "learning_rate": 2.6238137984098487e-05, + "loss": 0.328, + "step": 37109 + }, + { + "epoch": 1.43, + "learning_rate": 2.623749679404976e-05, + "loss": 0.4677, + "step": 37110 + }, + { + "epoch": 1.43, + "learning_rate": 2.6236855604001027e-05, + "loss": 0.3985, + "step": 37111 + }, + { + "epoch": 1.43, + "learning_rate": 2.62362144139523e-05, + "loss": 0.3677, + "step": 37112 + }, + { + "epoch": 1.43, + "learning_rate": 2.6235573223903564e-05, + "loss": 0.4165, + "step": 37113 + }, + { + "epoch": 1.43, + "learning_rate": 2.6234932033854836e-05, + "loss": 0.3005, + "step": 37114 + }, + { + "epoch": 1.43, + "learning_rate": 2.6234290843806104e-05, + "loss": 0.3614, + "step": 37115 + }, + { + "epoch": 1.43, + "learning_rate": 2.6233649653757376e-05, + "loss": 0.518, + "step": 37116 + }, + { + "epoch": 1.43, + "learning_rate": 2.6233008463708647e-05, + "loss": 0.2804, + "step": 37117 + }, + { + "epoch": 1.43, + "learning_rate": 2.6232367273659912e-05, + "loss": 0.3988, + "step": 37118 + }, + { + "epoch": 1.43, + "learning_rate": 2.6231726083611184e-05, + "loss": 0.4243, + "step": 37119 + }, + { + "epoch": 1.43, + "learning_rate": 2.6231084893562452e-05, + "loss": 0.3304, + "step": 37120 + }, + { + "epoch": 1.43, + "learning_rate": 2.6230443703513724e-05, + "loss": 0.4149, + "step": 37121 + }, + { + "epoch": 1.43, + "learning_rate": 2.622980251346499e-05, + "loss": 0.4304, + "step": 37122 + }, + { + "epoch": 1.43, + "learning_rate": 2.6229161323416264e-05, + "loss": 0.3929, + "step": 37123 + }, + { + "epoch": 1.43, + "learning_rate": 2.622852013336753e-05, + "loss": 0.4175, + "step": 37124 + }, + { + "epoch": 1.43, + "learning_rate": 2.62278789433188e-05, + "loss": 0.3328, + "step": 37125 + }, + { + "epoch": 1.43, + "learning_rate": 2.6227237753270072e-05, + "loss": 0.3569, + "step": 37126 + }, + { + "epoch": 1.43, + "learning_rate": 2.622659656322134e-05, + "loss": 0.3373, + "step": 37127 + }, + { + "epoch": 1.43, + "learning_rate": 2.6225955373172613e-05, + "loss": 0.3422, + "step": 37128 + }, + { + "epoch": 1.43, + "learning_rate": 2.6225314183123877e-05, + "loss": 0.3308, + "step": 37129 + }, + { + "epoch": 1.43, + "learning_rate": 2.622467299307515e-05, + "loss": 0.349, + "step": 37130 + }, + { + "epoch": 1.43, + "learning_rate": 2.6224031803026418e-05, + "loss": 0.3161, + "step": 37131 + }, + { + "epoch": 1.43, + "learning_rate": 2.622339061297769e-05, + "loss": 0.3832, + "step": 37132 + }, + { + "epoch": 1.43, + "learning_rate": 2.6222749422928954e-05, + "loss": 0.2707, + "step": 37133 + }, + { + "epoch": 1.43, + "learning_rate": 2.6222108232880226e-05, + "loss": 0.5481, + "step": 37134 + }, + { + "epoch": 1.43, + "learning_rate": 2.6221467042831498e-05, + "loss": 0.3016, + "step": 37135 + }, + { + "epoch": 1.43, + "learning_rate": 2.6220825852782766e-05, + "loss": 0.6391, + "step": 37136 + }, + { + "epoch": 1.43, + "learning_rate": 2.6220184662734038e-05, + "loss": 0.3737, + "step": 37137 + }, + { + "epoch": 1.43, + "learning_rate": 2.6219543472685303e-05, + "loss": 0.3577, + "step": 37138 + }, + { + "epoch": 1.43, + "learning_rate": 2.6218902282636574e-05, + "loss": 0.351, + "step": 37139 + }, + { + "epoch": 1.43, + "learning_rate": 2.6218261092587843e-05, + "loss": 0.3565, + "step": 37140 + }, + { + "epoch": 1.43, + "learning_rate": 2.6217619902539114e-05, + "loss": 0.5056, + "step": 37141 + }, + { + "epoch": 1.43, + "learning_rate": 2.621697871249038e-05, + "loss": 0.422, + "step": 37142 + }, + { + "epoch": 1.43, + "learning_rate": 2.6216337522441654e-05, + "loss": 0.4353, + "step": 37143 + }, + { + "epoch": 1.43, + "learning_rate": 2.6215696332392926e-05, + "loss": 0.3055, + "step": 37144 + }, + { + "epoch": 1.43, + "learning_rate": 2.621505514234419e-05, + "loss": 0.3913, + "step": 37145 + }, + { + "epoch": 1.43, + "learning_rate": 2.6214413952295463e-05, + "loss": 0.3839, + "step": 37146 + }, + { + "epoch": 1.43, + "learning_rate": 2.621377276224673e-05, + "loss": 0.3965, + "step": 37147 + }, + { + "epoch": 1.43, + "learning_rate": 2.6213131572198003e-05, + "loss": 0.3938, + "step": 37148 + }, + { + "epoch": 1.43, + "learning_rate": 2.6212490382149268e-05, + "loss": 0.4061, + "step": 37149 + }, + { + "epoch": 1.43, + "learning_rate": 2.621184919210054e-05, + "loss": 0.2974, + "step": 37150 + }, + { + "epoch": 1.43, + "learning_rate": 2.6211208002051808e-05, + "loss": 0.4407, + "step": 37151 + }, + { + "epoch": 1.43, + "learning_rate": 2.621056681200308e-05, + "loss": 0.3271, + "step": 37152 + }, + { + "epoch": 1.43, + "learning_rate": 2.620992562195435e-05, + "loss": 0.3829, + "step": 37153 + }, + { + "epoch": 1.43, + "learning_rate": 2.6209284431905616e-05, + "loss": 0.3065, + "step": 37154 + }, + { + "epoch": 1.43, + "learning_rate": 2.6208643241856888e-05, + "loss": 0.3893, + "step": 37155 + }, + { + "epoch": 1.43, + "learning_rate": 2.6208002051808156e-05, + "loss": 0.3261, + "step": 37156 + }, + { + "epoch": 1.43, + "learning_rate": 2.6207360861759428e-05, + "loss": 0.5426, + "step": 37157 + }, + { + "epoch": 1.43, + "learning_rate": 2.6206719671710693e-05, + "loss": 0.4036, + "step": 37158 + }, + { + "epoch": 1.43, + "learning_rate": 2.6206078481661968e-05, + "loss": 0.3177, + "step": 37159 + }, + { + "epoch": 1.43, + "learning_rate": 2.6205437291613233e-05, + "loss": 0.4573, + "step": 37160 + }, + { + "epoch": 1.43, + "learning_rate": 2.6204796101564505e-05, + "loss": 0.3003, + "step": 37161 + }, + { + "epoch": 1.43, + "learning_rate": 2.6204154911515776e-05, + "loss": 0.3306, + "step": 37162 + }, + { + "epoch": 1.43, + "learning_rate": 2.6203513721467045e-05, + "loss": 0.4496, + "step": 37163 + }, + { + "epoch": 1.43, + "learning_rate": 2.6202872531418316e-05, + "loss": 0.2741, + "step": 37164 + }, + { + "epoch": 1.43, + "learning_rate": 2.620223134136958e-05, + "loss": 0.3749, + "step": 37165 + }, + { + "epoch": 1.43, + "learning_rate": 2.6201590151320853e-05, + "loss": 0.4698, + "step": 37166 + }, + { + "epoch": 1.43, + "learning_rate": 2.620094896127212e-05, + "loss": 0.3308, + "step": 37167 + }, + { + "epoch": 1.43, + "learning_rate": 2.6200307771223393e-05, + "loss": 0.3775, + "step": 37168 + }, + { + "epoch": 1.43, + "learning_rate": 2.6199666581174658e-05, + "loss": 0.4213, + "step": 37169 + }, + { + "epoch": 1.43, + "learning_rate": 2.619902539112593e-05, + "loss": 0.4582, + "step": 37170 + }, + { + "epoch": 1.43, + "learning_rate": 2.61983842010772e-05, + "loss": 0.3535, + "step": 37171 + }, + { + "epoch": 1.43, + "learning_rate": 2.619774301102847e-05, + "loss": 0.3604, + "step": 37172 + }, + { + "epoch": 1.43, + "learning_rate": 2.619710182097974e-05, + "loss": 0.3942, + "step": 37173 + }, + { + "epoch": 1.43, + "learning_rate": 2.6196460630931006e-05, + "loss": 0.3776, + "step": 37174 + }, + { + "epoch": 1.43, + "learning_rate": 2.619581944088228e-05, + "loss": 0.4577, + "step": 37175 + }, + { + "epoch": 1.43, + "learning_rate": 2.6195178250833547e-05, + "loss": 0.4144, + "step": 37176 + }, + { + "epoch": 1.43, + "learning_rate": 2.6194537060784818e-05, + "loss": 0.3899, + "step": 37177 + }, + { + "epoch": 1.43, + "learning_rate": 2.6193895870736087e-05, + "loss": 0.3617, + "step": 37178 + }, + { + "epoch": 1.43, + "learning_rate": 2.6193254680687358e-05, + "loss": 0.3669, + "step": 37179 + }, + { + "epoch": 1.43, + "learning_rate": 2.619261349063863e-05, + "loss": 0.3266, + "step": 37180 + }, + { + "epoch": 1.43, + "learning_rate": 2.6191972300589895e-05, + "loss": 0.4849, + "step": 37181 + }, + { + "epoch": 1.43, + "learning_rate": 2.6191331110541167e-05, + "loss": 0.3182, + "step": 37182 + }, + { + "epoch": 1.43, + "learning_rate": 2.6190689920492435e-05, + "loss": 0.3002, + "step": 37183 + }, + { + "epoch": 1.43, + "learning_rate": 2.6190048730443707e-05, + "loss": 0.3367, + "step": 37184 + }, + { + "epoch": 1.43, + "learning_rate": 2.618940754039497e-05, + "loss": 0.2336, + "step": 37185 + }, + { + "epoch": 1.43, + "learning_rate": 2.6188766350346243e-05, + "loss": 0.3232, + "step": 37186 + }, + { + "epoch": 1.43, + "learning_rate": 2.618812516029751e-05, + "loss": 0.3096, + "step": 37187 + }, + { + "epoch": 1.43, + "learning_rate": 2.6187483970248783e-05, + "loss": 0.2469, + "step": 37188 + }, + { + "epoch": 1.43, + "learning_rate": 2.6186842780200055e-05, + "loss": 0.4162, + "step": 37189 + }, + { + "epoch": 1.43, + "learning_rate": 2.618620159015132e-05, + "loss": 0.4587, + "step": 37190 + }, + { + "epoch": 1.43, + "learning_rate": 2.6185560400102592e-05, + "loss": 0.3496, + "step": 37191 + }, + { + "epoch": 1.43, + "learning_rate": 2.618491921005386e-05, + "loss": 0.5214, + "step": 37192 + }, + { + "epoch": 1.43, + "learning_rate": 2.6184278020005132e-05, + "loss": 0.2819, + "step": 37193 + }, + { + "epoch": 1.43, + "learning_rate": 2.6183636829956397e-05, + "loss": 0.3873, + "step": 37194 + }, + { + "epoch": 1.43, + "learning_rate": 2.6182995639907672e-05, + "loss": 0.3221, + "step": 37195 + }, + { + "epoch": 1.43, + "learning_rate": 2.6182354449858937e-05, + "loss": 0.4383, + "step": 37196 + }, + { + "epoch": 1.43, + "learning_rate": 2.618171325981021e-05, + "loss": 0.4226, + "step": 37197 + }, + { + "epoch": 1.43, + "learning_rate": 2.618107206976148e-05, + "loss": 0.3016, + "step": 37198 + }, + { + "epoch": 1.43, + "learning_rate": 2.618043087971275e-05, + "loss": 0.3025, + "step": 37199 + }, + { + "epoch": 1.43, + "learning_rate": 2.617978968966402e-05, + "loss": 0.4518, + "step": 37200 + }, + { + "epoch": 1.43, + "learning_rate": 2.6179148499615285e-05, + "loss": 0.3809, + "step": 37201 + }, + { + "epoch": 1.43, + "learning_rate": 2.6178507309566557e-05, + "loss": 0.2665, + "step": 37202 + }, + { + "epoch": 1.43, + "learning_rate": 2.6177866119517825e-05, + "loss": 0.3499, + "step": 37203 + }, + { + "epoch": 1.43, + "learning_rate": 2.6177224929469097e-05, + "loss": 0.4048, + "step": 37204 + }, + { + "epoch": 1.43, + "learning_rate": 2.6176583739420362e-05, + "loss": 0.4544, + "step": 37205 + }, + { + "epoch": 1.43, + "learning_rate": 2.6175942549371634e-05, + "loss": 0.4135, + "step": 37206 + }, + { + "epoch": 1.43, + "learning_rate": 2.6175301359322905e-05, + "loss": 0.2811, + "step": 37207 + }, + { + "epoch": 1.43, + "learning_rate": 2.6174660169274174e-05, + "loss": 0.3388, + "step": 37208 + }, + { + "epoch": 1.43, + "learning_rate": 2.6174018979225445e-05, + "loss": 0.3113, + "step": 37209 + }, + { + "epoch": 1.43, + "learning_rate": 2.617337778917671e-05, + "loss": 0.3991, + "step": 37210 + }, + { + "epoch": 1.43, + "learning_rate": 2.6172736599127985e-05, + "loss": 0.4163, + "step": 37211 + }, + { + "epoch": 1.43, + "learning_rate": 2.617209540907925e-05, + "loss": 0.4615, + "step": 37212 + }, + { + "epoch": 1.43, + "learning_rate": 2.6171454219030522e-05, + "loss": 0.2391, + "step": 37213 + }, + { + "epoch": 1.43, + "learning_rate": 2.617081302898179e-05, + "loss": 0.2667, + "step": 37214 + }, + { + "epoch": 1.43, + "learning_rate": 2.6170171838933062e-05, + "loss": 0.4843, + "step": 37215 + }, + { + "epoch": 1.43, + "learning_rate": 2.6169530648884334e-05, + "loss": 0.3605, + "step": 37216 + }, + { + "epoch": 1.43, + "learning_rate": 2.61688894588356e-05, + "loss": 0.3251, + "step": 37217 + }, + { + "epoch": 1.43, + "learning_rate": 2.616824826878687e-05, + "loss": 0.31, + "step": 37218 + }, + { + "epoch": 1.43, + "learning_rate": 2.616760707873814e-05, + "loss": 0.5927, + "step": 37219 + }, + { + "epoch": 1.43, + "learning_rate": 2.616696588868941e-05, + "loss": 0.3707, + "step": 37220 + }, + { + "epoch": 1.43, + "learning_rate": 2.6166324698640676e-05, + "loss": 0.3332, + "step": 37221 + }, + { + "epoch": 1.43, + "learning_rate": 2.6165683508591947e-05, + "loss": 0.3267, + "step": 37222 + }, + { + "epoch": 1.43, + "learning_rate": 2.6165042318543216e-05, + "loss": 0.3446, + "step": 37223 + }, + { + "epoch": 1.43, + "learning_rate": 2.6164401128494487e-05, + "loss": 0.4002, + "step": 37224 + }, + { + "epoch": 1.43, + "learning_rate": 2.616375993844576e-05, + "loss": 0.2587, + "step": 37225 + }, + { + "epoch": 1.43, + "learning_rate": 2.6163118748397024e-05, + "loss": 0.336, + "step": 37226 + }, + { + "epoch": 1.43, + "learning_rate": 2.6162477558348296e-05, + "loss": 0.3087, + "step": 37227 + }, + { + "epoch": 1.43, + "learning_rate": 2.6161836368299564e-05, + "loss": 0.3371, + "step": 37228 + }, + { + "epoch": 1.43, + "learning_rate": 2.6161195178250836e-05, + "loss": 0.3982, + "step": 37229 + }, + { + "epoch": 1.43, + "learning_rate": 2.61605539882021e-05, + "loss": 0.3147, + "step": 37230 + }, + { + "epoch": 1.43, + "learning_rate": 2.6159912798153376e-05, + "loss": 0.2248, + "step": 37231 + }, + { + "epoch": 1.43, + "learning_rate": 2.615927160810464e-05, + "loss": 0.396, + "step": 37232 + }, + { + "epoch": 1.43, + "learning_rate": 2.6158630418055912e-05, + "loss": 0.3457, + "step": 37233 + }, + { + "epoch": 1.43, + "learning_rate": 2.6157989228007184e-05, + "loss": 0.3818, + "step": 37234 + }, + { + "epoch": 1.43, + "learning_rate": 2.6157348037958452e-05, + "loss": 0.279, + "step": 37235 + }, + { + "epoch": 1.43, + "learning_rate": 2.6156706847909724e-05, + "loss": 0.5137, + "step": 37236 + }, + { + "epoch": 1.43, + "learning_rate": 2.615606565786099e-05, + "loss": 0.2064, + "step": 37237 + }, + { + "epoch": 1.43, + "learning_rate": 2.615542446781226e-05, + "loss": 0.2874, + "step": 37238 + }, + { + "epoch": 1.43, + "learning_rate": 2.615478327776353e-05, + "loss": 0.3686, + "step": 37239 + }, + { + "epoch": 1.43, + "learning_rate": 2.61541420877148e-05, + "loss": 0.3735, + "step": 37240 + }, + { + "epoch": 1.43, + "learning_rate": 2.6153500897666066e-05, + "loss": 0.3649, + "step": 37241 + }, + { + "epoch": 1.43, + "learning_rate": 2.6152859707617338e-05, + "loss": 0.2287, + "step": 37242 + }, + { + "epoch": 1.43, + "learning_rate": 2.615221851756861e-05, + "loss": 0.4138, + "step": 37243 + }, + { + "epoch": 1.43, + "learning_rate": 2.6151577327519878e-05, + "loss": 0.2817, + "step": 37244 + }, + { + "epoch": 1.43, + "learning_rate": 2.615093613747115e-05, + "loss": 0.3285, + "step": 37245 + }, + { + "epoch": 1.43, + "learning_rate": 2.6150294947422414e-05, + "loss": 0.4591, + "step": 37246 + }, + { + "epoch": 1.43, + "learning_rate": 2.614965375737369e-05, + "loss": 0.4276, + "step": 37247 + }, + { + "epoch": 1.43, + "learning_rate": 2.6149012567324954e-05, + "loss": 0.4681, + "step": 37248 + }, + { + "epoch": 1.43, + "learning_rate": 2.6148371377276226e-05, + "loss": 0.2813, + "step": 37249 + }, + { + "epoch": 1.43, + "learning_rate": 2.6147730187227498e-05, + "loss": 0.2512, + "step": 37250 + }, + { + "epoch": 1.43, + "learning_rate": 2.6147088997178766e-05, + "loss": 0.3372, + "step": 37251 + }, + { + "epoch": 1.43, + "learning_rate": 2.6146447807130038e-05, + "loss": 0.1759, + "step": 37252 + }, + { + "epoch": 1.43, + "learning_rate": 2.6145806617081303e-05, + "loss": 0.3248, + "step": 37253 + }, + { + "epoch": 1.43, + "learning_rate": 2.6145165427032574e-05, + "loss": 0.4645, + "step": 37254 + }, + { + "epoch": 1.43, + "learning_rate": 2.6144524236983843e-05, + "loss": 0.3301, + "step": 37255 + }, + { + "epoch": 1.43, + "learning_rate": 2.6143883046935114e-05, + "loss": 0.3237, + "step": 37256 + }, + { + "epoch": 1.43, + "learning_rate": 2.614324185688638e-05, + "loss": 0.308, + "step": 37257 + }, + { + "epoch": 1.43, + "learning_rate": 2.614260066683765e-05, + "loss": 0.3705, + "step": 37258 + }, + { + "epoch": 1.43, + "learning_rate": 2.6141959476788923e-05, + "loss": 0.4846, + "step": 37259 + }, + { + "epoch": 1.43, + "learning_rate": 2.614131828674019e-05, + "loss": 0.3815, + "step": 37260 + }, + { + "epoch": 1.43, + "learning_rate": 2.6140677096691463e-05, + "loss": 0.4357, + "step": 37261 + }, + { + "epoch": 1.43, + "learning_rate": 2.6140035906642728e-05, + "loss": 0.2196, + "step": 37262 + }, + { + "epoch": 1.43, + "learning_rate": 2.6139394716594e-05, + "loss": 0.366, + "step": 37263 + }, + { + "epoch": 1.43, + "learning_rate": 2.6138753526545268e-05, + "loss": 0.4613, + "step": 37264 + }, + { + "epoch": 1.43, + "learning_rate": 2.613811233649654e-05, + "loss": 0.4463, + "step": 37265 + }, + { + "epoch": 1.43, + "learning_rate": 2.6137471146447804e-05, + "loss": 0.3073, + "step": 37266 + }, + { + "epoch": 1.43, + "learning_rate": 2.613682995639908e-05, + "loss": 0.2681, + "step": 37267 + }, + { + "epoch": 1.43, + "learning_rate": 2.613618876635035e-05, + "loss": 0.2732, + "step": 37268 + }, + { + "epoch": 1.43, + "learning_rate": 2.6135547576301616e-05, + "loss": 0.3372, + "step": 37269 + }, + { + "epoch": 1.43, + "learning_rate": 2.6134906386252888e-05, + "loss": 0.4366, + "step": 37270 + }, + { + "epoch": 1.43, + "learning_rate": 2.6134265196204156e-05, + "loss": 0.3983, + "step": 37271 + }, + { + "epoch": 1.43, + "learning_rate": 2.6133624006155428e-05, + "loss": 0.2308, + "step": 37272 + }, + { + "epoch": 1.43, + "learning_rate": 2.6132982816106693e-05, + "loss": 0.3937, + "step": 37273 + }, + { + "epoch": 1.43, + "learning_rate": 2.6132341626057965e-05, + "loss": 0.3555, + "step": 37274 + }, + { + "epoch": 1.43, + "learning_rate": 2.6131700436009233e-05, + "loss": 0.4906, + "step": 37275 + }, + { + "epoch": 1.43, + "learning_rate": 2.6131059245960505e-05, + "loss": 0.3479, + "step": 37276 + }, + { + "epoch": 1.43, + "learning_rate": 2.6130418055911776e-05, + "loss": 0.4773, + "step": 37277 + }, + { + "epoch": 1.43, + "learning_rate": 2.612977686586304e-05, + "loss": 0.27, + "step": 37278 + }, + { + "epoch": 1.43, + "learning_rate": 2.6129135675814313e-05, + "loss": 0.3654, + "step": 37279 + }, + { + "epoch": 1.43, + "learning_rate": 2.612849448576558e-05, + "loss": 0.3657, + "step": 37280 + }, + { + "epoch": 1.43, + "learning_rate": 2.6127853295716853e-05, + "loss": 0.2884, + "step": 37281 + }, + { + "epoch": 1.43, + "learning_rate": 2.6127212105668118e-05, + "loss": 0.348, + "step": 37282 + }, + { + "epoch": 1.43, + "learning_rate": 2.6126570915619393e-05, + "loss": 0.3882, + "step": 37283 + }, + { + "epoch": 1.43, + "learning_rate": 2.6125929725570658e-05, + "loss": 0.4012, + "step": 37284 + }, + { + "epoch": 1.43, + "learning_rate": 2.612528853552193e-05, + "loss": 0.2982, + "step": 37285 + }, + { + "epoch": 1.43, + "learning_rate": 2.61246473454732e-05, + "loss": 0.3924, + "step": 37286 + }, + { + "epoch": 1.43, + "learning_rate": 2.612400615542447e-05, + "loss": 0.4286, + "step": 37287 + }, + { + "epoch": 1.43, + "learning_rate": 2.612336496537574e-05, + "loss": 0.4008, + "step": 37288 + }, + { + "epoch": 1.43, + "learning_rate": 2.6122723775327007e-05, + "loss": 0.3976, + "step": 37289 + }, + { + "epoch": 1.43, + "learning_rate": 2.6122082585278278e-05, + "loss": 0.2408, + "step": 37290 + }, + { + "epoch": 1.43, + "learning_rate": 2.6121441395229547e-05, + "loss": 0.3909, + "step": 37291 + }, + { + "epoch": 1.43, + "learning_rate": 2.612080020518082e-05, + "loss": 0.3519, + "step": 37292 + }, + { + "epoch": 1.43, + "learning_rate": 2.6120159015132083e-05, + "loss": 0.387, + "step": 37293 + }, + { + "epoch": 1.43, + "learning_rate": 2.6119517825083355e-05, + "loss": 0.4456, + "step": 37294 + }, + { + "epoch": 1.43, + "learning_rate": 2.6118876635034627e-05, + "loss": 0.3263, + "step": 37295 + }, + { + "epoch": 1.43, + "learning_rate": 2.6118235444985895e-05, + "loss": 0.4909, + "step": 37296 + }, + { + "epoch": 1.43, + "learning_rate": 2.6117594254937167e-05, + "loss": 0.3568, + "step": 37297 + }, + { + "epoch": 1.43, + "learning_rate": 2.611695306488843e-05, + "loss": 0.406, + "step": 37298 + }, + { + "epoch": 1.43, + "learning_rate": 2.6116311874839707e-05, + "loss": 0.3389, + "step": 37299 + }, + { + "epoch": 1.43, + "learning_rate": 2.6115670684790972e-05, + "loss": 0.3194, + "step": 37300 + }, + { + "epoch": 1.43, + "learning_rate": 2.6115029494742243e-05, + "loss": 0.3464, + "step": 37301 + }, + { + "epoch": 1.43, + "learning_rate": 2.611438830469351e-05, + "loss": 0.3297, + "step": 37302 + }, + { + "epoch": 1.43, + "learning_rate": 2.6113747114644783e-05, + "loss": 0.3584, + "step": 37303 + }, + { + "epoch": 1.43, + "learning_rate": 2.6113105924596055e-05, + "loss": 0.3507, + "step": 37304 + }, + { + "epoch": 1.43, + "learning_rate": 2.611246473454732e-05, + "loss": 0.2432, + "step": 37305 + }, + { + "epoch": 1.43, + "learning_rate": 2.6111823544498592e-05, + "loss": 0.34, + "step": 37306 + }, + { + "epoch": 1.43, + "learning_rate": 2.611118235444986e-05, + "loss": 0.4128, + "step": 37307 + }, + { + "epoch": 1.43, + "learning_rate": 2.6110541164401132e-05, + "loss": 0.3464, + "step": 37308 + }, + { + "epoch": 1.43, + "learning_rate": 2.6109899974352397e-05, + "loss": 0.3374, + "step": 37309 + }, + { + "epoch": 1.43, + "learning_rate": 2.610925878430367e-05, + "loss": 0.3767, + "step": 37310 + }, + { + "epoch": 1.43, + "learning_rate": 2.6108617594254937e-05, + "loss": 0.4052, + "step": 37311 + }, + { + "epoch": 1.43, + "learning_rate": 2.610797640420621e-05, + "loss": 0.4126, + "step": 37312 + }, + { + "epoch": 1.43, + "learning_rate": 2.610733521415748e-05, + "loss": 0.5018, + "step": 37313 + }, + { + "epoch": 1.43, + "learning_rate": 2.6106694024108745e-05, + "loss": 0.4008, + "step": 37314 + }, + { + "epoch": 1.43, + "learning_rate": 2.6106052834060017e-05, + "loss": 0.3522, + "step": 37315 + }, + { + "epoch": 1.43, + "learning_rate": 2.6105411644011285e-05, + "loss": 0.442, + "step": 37316 + }, + { + "epoch": 1.43, + "learning_rate": 2.6104770453962557e-05, + "loss": 0.2457, + "step": 37317 + }, + { + "epoch": 1.43, + "learning_rate": 2.6104129263913822e-05, + "loss": 0.2029, + "step": 37318 + }, + { + "epoch": 1.43, + "learning_rate": 2.6103488073865097e-05, + "loss": 0.3114, + "step": 37319 + }, + { + "epoch": 1.43, + "learning_rate": 2.6102846883816362e-05, + "loss": 0.2947, + "step": 37320 + }, + { + "epoch": 1.43, + "learning_rate": 2.6102205693767634e-05, + "loss": 0.4176, + "step": 37321 + }, + { + "epoch": 1.43, + "learning_rate": 2.6101564503718905e-05, + "loss": 0.3517, + "step": 37322 + }, + { + "epoch": 1.43, + "learning_rate": 2.6100923313670174e-05, + "loss": 0.3128, + "step": 37323 + }, + { + "epoch": 1.43, + "learning_rate": 2.6100282123621445e-05, + "loss": 0.3081, + "step": 37324 + }, + { + "epoch": 1.44, + "learning_rate": 2.609964093357271e-05, + "loss": 0.3518, + "step": 37325 + }, + { + "epoch": 1.44, + "learning_rate": 2.6098999743523982e-05, + "loss": 0.3546, + "step": 37326 + }, + { + "epoch": 1.44, + "learning_rate": 2.609835855347525e-05, + "loss": 0.5409, + "step": 37327 + }, + { + "epoch": 1.44, + "learning_rate": 2.6097717363426522e-05, + "loss": 0.3177, + "step": 37328 + }, + { + "epoch": 1.44, + "learning_rate": 2.6097076173377787e-05, + "loss": 0.3562, + "step": 37329 + }, + { + "epoch": 1.44, + "learning_rate": 2.609643498332906e-05, + "loss": 0.2557, + "step": 37330 + }, + { + "epoch": 1.44, + "learning_rate": 2.609579379328033e-05, + "loss": 0.3478, + "step": 37331 + }, + { + "epoch": 1.44, + "learning_rate": 2.60951526032316e-05, + "loss": 0.3643, + "step": 37332 + }, + { + "epoch": 1.44, + "learning_rate": 2.609451141318287e-05, + "loss": 0.3172, + "step": 37333 + }, + { + "epoch": 1.44, + "learning_rate": 2.6093870223134136e-05, + "loss": 0.3392, + "step": 37334 + }, + { + "epoch": 1.44, + "learning_rate": 2.609322903308541e-05, + "loss": 0.3642, + "step": 37335 + }, + { + "epoch": 1.44, + "learning_rate": 2.6092587843036676e-05, + "loss": 0.3833, + "step": 37336 + }, + { + "epoch": 1.44, + "learning_rate": 2.6091946652987947e-05, + "loss": 0.332, + "step": 37337 + }, + { + "epoch": 1.44, + "learning_rate": 2.6091305462939216e-05, + "loss": 0.4037, + "step": 37338 + }, + { + "epoch": 1.44, + "learning_rate": 2.6090664272890487e-05, + "loss": 0.4737, + "step": 37339 + }, + { + "epoch": 1.44, + "learning_rate": 2.609002308284176e-05, + "loss": 0.3058, + "step": 37340 + }, + { + "epoch": 1.44, + "learning_rate": 2.6089381892793024e-05, + "loss": 0.2988, + "step": 37341 + }, + { + "epoch": 1.44, + "learning_rate": 2.6088740702744296e-05, + "loss": 0.3145, + "step": 37342 + }, + { + "epoch": 1.44, + "learning_rate": 2.6088099512695564e-05, + "loss": 0.3179, + "step": 37343 + }, + { + "epoch": 1.44, + "learning_rate": 2.6087458322646836e-05, + "loss": 0.383, + "step": 37344 + }, + { + "epoch": 1.44, + "learning_rate": 2.60868171325981e-05, + "loss": 0.4003, + "step": 37345 + }, + { + "epoch": 1.44, + "learning_rate": 2.6086175942549372e-05, + "loss": 0.3043, + "step": 37346 + }, + { + "epoch": 1.44, + "learning_rate": 2.608553475250064e-05, + "loss": 0.3516, + "step": 37347 + }, + { + "epoch": 1.44, + "learning_rate": 2.6084893562451912e-05, + "loss": 0.4381, + "step": 37348 + }, + { + "epoch": 1.44, + "learning_rate": 2.6084252372403184e-05, + "loss": 0.3823, + "step": 37349 + }, + { + "epoch": 1.44, + "learning_rate": 2.608361118235445e-05, + "loss": 0.28, + "step": 37350 + }, + { + "epoch": 1.44, + "learning_rate": 2.608296999230572e-05, + "loss": 0.396, + "step": 37351 + }, + { + "epoch": 1.44, + "learning_rate": 2.608232880225699e-05, + "loss": 0.4062, + "step": 37352 + }, + { + "epoch": 1.44, + "learning_rate": 2.608168761220826e-05, + "loss": 0.3462, + "step": 37353 + }, + { + "epoch": 1.44, + "learning_rate": 2.6081046422159526e-05, + "loss": 0.3881, + "step": 37354 + }, + { + "epoch": 1.44, + "learning_rate": 2.60804052321108e-05, + "loss": 0.3902, + "step": 37355 + }, + { + "epoch": 1.44, + "learning_rate": 2.6079764042062066e-05, + "loss": 0.3215, + "step": 37356 + }, + { + "epoch": 1.44, + "learning_rate": 2.6079122852013338e-05, + "loss": 0.4843, + "step": 37357 + }, + { + "epoch": 1.44, + "learning_rate": 2.607848166196461e-05, + "loss": 0.3973, + "step": 37358 + }, + { + "epoch": 1.44, + "learning_rate": 2.6077840471915878e-05, + "loss": 0.2814, + "step": 37359 + }, + { + "epoch": 1.44, + "learning_rate": 2.607719928186715e-05, + "loss": 0.3117, + "step": 37360 + }, + { + "epoch": 1.44, + "learning_rate": 2.6076558091818414e-05, + "loss": 0.2943, + "step": 37361 + }, + { + "epoch": 1.44, + "learning_rate": 2.6075916901769686e-05, + "loss": 0.4163, + "step": 37362 + }, + { + "epoch": 1.44, + "learning_rate": 2.6075275711720954e-05, + "loss": 0.3367, + "step": 37363 + }, + { + "epoch": 1.44, + "learning_rate": 2.6074634521672226e-05, + "loss": 0.2898, + "step": 37364 + }, + { + "epoch": 1.44, + "learning_rate": 2.607399333162349e-05, + "loss": 0.2856, + "step": 37365 + }, + { + "epoch": 1.44, + "learning_rate": 2.6073352141574763e-05, + "loss": 0.3399, + "step": 37366 + }, + { + "epoch": 1.44, + "learning_rate": 2.6072710951526034e-05, + "loss": 0.3944, + "step": 37367 + }, + { + "epoch": 1.44, + "learning_rate": 2.6072069761477303e-05, + "loss": 0.3049, + "step": 37368 + }, + { + "epoch": 1.44, + "learning_rate": 2.6071428571428574e-05, + "loss": 0.388, + "step": 37369 + }, + { + "epoch": 1.44, + "learning_rate": 2.607078738137984e-05, + "loss": 0.4483, + "step": 37370 + }, + { + "epoch": 1.44, + "learning_rate": 2.6070146191331115e-05, + "loss": 0.3606, + "step": 37371 + }, + { + "epoch": 1.44, + "learning_rate": 2.606950500128238e-05, + "loss": 0.487, + "step": 37372 + }, + { + "epoch": 1.44, + "learning_rate": 2.606886381123365e-05, + "loss": 0.304, + "step": 37373 + }, + { + "epoch": 1.44, + "learning_rate": 2.606822262118492e-05, + "loss": 0.4564, + "step": 37374 + }, + { + "epoch": 1.44, + "learning_rate": 2.606758143113619e-05, + "loss": 0.3762, + "step": 37375 + }, + { + "epoch": 1.44, + "learning_rate": 2.6066940241087463e-05, + "loss": 0.3268, + "step": 37376 + }, + { + "epoch": 1.44, + "learning_rate": 2.6066299051038728e-05, + "loss": 0.4766, + "step": 37377 + }, + { + "epoch": 1.44, + "learning_rate": 2.606565786099e-05, + "loss": 0.3612, + "step": 37378 + }, + { + "epoch": 1.44, + "learning_rate": 2.6065016670941268e-05, + "loss": 0.3665, + "step": 37379 + }, + { + "epoch": 1.44, + "learning_rate": 2.606437548089254e-05, + "loss": 0.4787, + "step": 37380 + }, + { + "epoch": 1.44, + "learning_rate": 2.6063734290843805e-05, + "loss": 0.2756, + "step": 37381 + }, + { + "epoch": 1.44, + "learning_rate": 2.6063093100795076e-05, + "loss": 0.3278, + "step": 37382 + }, + { + "epoch": 1.44, + "learning_rate": 2.6062451910746345e-05, + "loss": 0.3667, + "step": 37383 + }, + { + "epoch": 1.44, + "learning_rate": 2.6061810720697616e-05, + "loss": 0.4219, + "step": 37384 + }, + { + "epoch": 1.44, + "learning_rate": 2.6061169530648888e-05, + "loss": 0.3721, + "step": 37385 + }, + { + "epoch": 1.44, + "learning_rate": 2.6060528340600153e-05, + "loss": 0.3625, + "step": 37386 + }, + { + "epoch": 1.44, + "learning_rate": 2.6059887150551425e-05, + "loss": 0.3798, + "step": 37387 + }, + { + "epoch": 1.44, + "learning_rate": 2.6059245960502693e-05, + "loss": 0.3011, + "step": 37388 + }, + { + "epoch": 1.44, + "learning_rate": 2.6058604770453965e-05, + "loss": 0.3656, + "step": 37389 + }, + { + "epoch": 1.44, + "learning_rate": 2.605796358040523e-05, + "loss": 0.3826, + "step": 37390 + }, + { + "epoch": 1.44, + "learning_rate": 2.6057322390356505e-05, + "loss": 0.3911, + "step": 37391 + }, + { + "epoch": 1.44, + "learning_rate": 2.605668120030777e-05, + "loss": 0.3856, + "step": 37392 + }, + { + "epoch": 1.44, + "learning_rate": 2.605604001025904e-05, + "loss": 0.4072, + "step": 37393 + }, + { + "epoch": 1.44, + "learning_rate": 2.6055398820210313e-05, + "loss": 0.2895, + "step": 37394 + }, + { + "epoch": 1.44, + "learning_rate": 2.605475763016158e-05, + "loss": 0.4105, + "step": 37395 + }, + { + "epoch": 1.44, + "learning_rate": 2.6054116440112853e-05, + "loss": 0.3058, + "step": 37396 + }, + { + "epoch": 1.44, + "learning_rate": 2.6053475250064118e-05, + "loss": 0.3569, + "step": 37397 + }, + { + "epoch": 1.44, + "learning_rate": 2.605283406001539e-05, + "loss": 0.352, + "step": 37398 + }, + { + "epoch": 1.44, + "learning_rate": 2.6052192869966658e-05, + "loss": 0.3403, + "step": 37399 + }, + { + "epoch": 1.44, + "learning_rate": 2.605155167991793e-05, + "loss": 0.3863, + "step": 37400 + }, + { + "epoch": 1.44, + "learning_rate": 2.6050910489869195e-05, + "loss": 0.5167, + "step": 37401 + }, + { + "epoch": 1.44, + "learning_rate": 2.6050269299820467e-05, + "loss": 0.332, + "step": 37402 + }, + { + "epoch": 1.44, + "learning_rate": 2.6049628109771738e-05, + "loss": 0.3407, + "step": 37403 + }, + { + "epoch": 1.44, + "learning_rate": 2.6048986919723007e-05, + "loss": 0.3989, + "step": 37404 + }, + { + "epoch": 1.44, + "learning_rate": 2.604834572967428e-05, + "loss": 0.4481, + "step": 37405 + }, + { + "epoch": 1.44, + "learning_rate": 2.6047704539625543e-05, + "loss": 0.4934, + "step": 37406 + }, + { + "epoch": 1.44, + "learning_rate": 2.604706334957682e-05, + "loss": 0.2781, + "step": 37407 + }, + { + "epoch": 1.44, + "learning_rate": 2.6046422159528083e-05, + "loss": 0.3768, + "step": 37408 + }, + { + "epoch": 1.44, + "learning_rate": 2.6045780969479355e-05, + "loss": 0.3723, + "step": 37409 + }, + { + "epoch": 1.44, + "learning_rate": 2.6045139779430623e-05, + "loss": 0.3829, + "step": 37410 + }, + { + "epoch": 1.44, + "learning_rate": 2.6044498589381895e-05, + "loss": 0.3027, + "step": 37411 + }, + { + "epoch": 1.44, + "learning_rate": 2.6043857399333167e-05, + "loss": 0.3271, + "step": 37412 + }, + { + "epoch": 1.44, + "learning_rate": 2.6043216209284432e-05, + "loss": 0.197, + "step": 37413 + }, + { + "epoch": 1.44, + "learning_rate": 2.6042575019235703e-05, + "loss": 0.4097, + "step": 37414 + }, + { + "epoch": 1.44, + "learning_rate": 2.6041933829186972e-05, + "loss": 0.3072, + "step": 37415 + }, + { + "epoch": 1.44, + "learning_rate": 2.6041292639138244e-05, + "loss": 0.3443, + "step": 37416 + }, + { + "epoch": 1.44, + "learning_rate": 2.604065144908951e-05, + "loss": 0.4028, + "step": 37417 + }, + { + "epoch": 1.44, + "learning_rate": 2.604001025904078e-05, + "loss": 0.421, + "step": 37418 + }, + { + "epoch": 1.44, + "learning_rate": 2.603936906899205e-05, + "loss": 0.1898, + "step": 37419 + }, + { + "epoch": 1.44, + "learning_rate": 2.603872787894332e-05, + "loss": 0.3819, + "step": 37420 + }, + { + "epoch": 1.44, + "learning_rate": 2.6038086688894592e-05, + "loss": 0.5769, + "step": 37421 + }, + { + "epoch": 1.44, + "learning_rate": 2.6037445498845857e-05, + "loss": 0.482, + "step": 37422 + }, + { + "epoch": 1.44, + "learning_rate": 2.6036804308797132e-05, + "loss": 0.2766, + "step": 37423 + }, + { + "epoch": 1.44, + "learning_rate": 2.6036163118748397e-05, + "loss": 0.433, + "step": 37424 + }, + { + "epoch": 1.44, + "learning_rate": 2.603552192869967e-05, + "loss": 0.3446, + "step": 37425 + }, + { + "epoch": 1.44, + "learning_rate": 2.6034880738650934e-05, + "loss": 0.4007, + "step": 37426 + }, + { + "epoch": 1.44, + "learning_rate": 2.603423954860221e-05, + "loss": 0.3077, + "step": 37427 + }, + { + "epoch": 1.44, + "learning_rate": 2.6033598358553474e-05, + "loss": 0.2999, + "step": 37428 + }, + { + "epoch": 1.44, + "learning_rate": 2.6032957168504745e-05, + "loss": 0.3929, + "step": 37429 + }, + { + "epoch": 1.44, + "learning_rate": 2.6032315978456017e-05, + "loss": 0.3028, + "step": 37430 + }, + { + "epoch": 1.44, + "learning_rate": 2.6031674788407285e-05, + "loss": 0.2992, + "step": 37431 + }, + { + "epoch": 1.44, + "learning_rate": 2.6031033598358557e-05, + "loss": 0.3735, + "step": 37432 + }, + { + "epoch": 1.44, + "learning_rate": 2.6030392408309822e-05, + "loss": 0.4403, + "step": 37433 + }, + { + "epoch": 1.44, + "learning_rate": 2.6029751218261094e-05, + "loss": 0.5067, + "step": 37434 + }, + { + "epoch": 1.44, + "learning_rate": 2.6029110028212362e-05, + "loss": 0.3271, + "step": 37435 + }, + { + "epoch": 1.44, + "learning_rate": 2.6028468838163634e-05, + "loss": 0.3242, + "step": 37436 + }, + { + "epoch": 1.44, + "learning_rate": 2.6027827648114906e-05, + "loss": 0.3526, + "step": 37437 + }, + { + "epoch": 1.44, + "learning_rate": 2.602718645806617e-05, + "loss": 0.3025, + "step": 37438 + }, + { + "epoch": 1.44, + "learning_rate": 2.6026545268017442e-05, + "loss": 0.3733, + "step": 37439 + }, + { + "epoch": 1.44, + "learning_rate": 2.602590407796871e-05, + "loss": 0.3172, + "step": 37440 + }, + { + "epoch": 1.44, + "learning_rate": 2.6025262887919982e-05, + "loss": 0.2975, + "step": 37441 + }, + { + "epoch": 1.44, + "learning_rate": 2.6024621697871247e-05, + "loss": 0.5452, + "step": 37442 + }, + { + "epoch": 1.44, + "learning_rate": 2.6023980507822522e-05, + "loss": 0.4201, + "step": 37443 + }, + { + "epoch": 1.44, + "learning_rate": 2.6023339317773787e-05, + "loss": 0.4389, + "step": 37444 + }, + { + "epoch": 1.44, + "learning_rate": 2.602269812772506e-05, + "loss": 0.5461, + "step": 37445 + }, + { + "epoch": 1.44, + "learning_rate": 2.602205693767633e-05, + "loss": 0.4622, + "step": 37446 + }, + { + "epoch": 1.44, + "learning_rate": 2.60214157476276e-05, + "loss": 0.3026, + "step": 37447 + }, + { + "epoch": 1.44, + "learning_rate": 2.602077455757887e-05, + "loss": 0.3554, + "step": 37448 + }, + { + "epoch": 1.44, + "learning_rate": 2.6020133367530136e-05, + "loss": 0.3138, + "step": 37449 + }, + { + "epoch": 1.44, + "learning_rate": 2.6019492177481407e-05, + "loss": 0.5189, + "step": 37450 + }, + { + "epoch": 1.44, + "learning_rate": 2.6018850987432676e-05, + "loss": 0.3762, + "step": 37451 + }, + { + "epoch": 1.44, + "learning_rate": 2.6018209797383947e-05, + "loss": 0.2641, + "step": 37452 + }, + { + "epoch": 1.44, + "learning_rate": 2.6017568607335212e-05, + "loss": 0.3215, + "step": 37453 + }, + { + "epoch": 1.44, + "learning_rate": 2.6016927417286484e-05, + "loss": 0.3603, + "step": 37454 + }, + { + "epoch": 1.44, + "learning_rate": 2.6016286227237756e-05, + "loss": 0.3566, + "step": 37455 + }, + { + "epoch": 1.44, + "learning_rate": 2.6015645037189024e-05, + "loss": 0.3645, + "step": 37456 + }, + { + "epoch": 1.44, + "learning_rate": 2.6015003847140296e-05, + "loss": 0.3615, + "step": 37457 + }, + { + "epoch": 1.44, + "learning_rate": 2.601436265709156e-05, + "loss": 0.4769, + "step": 37458 + }, + { + "epoch": 1.44, + "learning_rate": 2.6013721467042836e-05, + "loss": 0.3476, + "step": 37459 + }, + { + "epoch": 1.44, + "learning_rate": 2.60130802769941e-05, + "loss": 0.3232, + "step": 37460 + }, + { + "epoch": 1.44, + "learning_rate": 2.6012439086945372e-05, + "loss": 0.3961, + "step": 37461 + }, + { + "epoch": 1.44, + "learning_rate": 2.601179789689664e-05, + "loss": 0.2339, + "step": 37462 + }, + { + "epoch": 1.44, + "learning_rate": 2.6011156706847913e-05, + "loss": 0.3621, + "step": 37463 + }, + { + "epoch": 1.44, + "learning_rate": 2.6010515516799184e-05, + "loss": 0.3484, + "step": 37464 + }, + { + "epoch": 1.44, + "learning_rate": 2.600987432675045e-05, + "loss": 0.3103, + "step": 37465 + }, + { + "epoch": 1.44, + "learning_rate": 2.600923313670172e-05, + "loss": 0.3728, + "step": 37466 + }, + { + "epoch": 1.44, + "learning_rate": 2.600859194665299e-05, + "loss": 0.3598, + "step": 37467 + }, + { + "epoch": 1.44, + "learning_rate": 2.600795075660426e-05, + "loss": 0.3515, + "step": 37468 + }, + { + "epoch": 1.44, + "learning_rate": 2.6007309566555526e-05, + "loss": 0.3442, + "step": 37469 + }, + { + "epoch": 1.44, + "learning_rate": 2.6006668376506798e-05, + "loss": 0.3159, + "step": 37470 + }, + { + "epoch": 1.44, + "learning_rate": 2.6006027186458066e-05, + "loss": 0.3016, + "step": 37471 + }, + { + "epoch": 1.44, + "learning_rate": 2.6005385996409338e-05, + "loss": 0.3512, + "step": 37472 + }, + { + "epoch": 1.44, + "learning_rate": 2.600474480636061e-05, + "loss": 0.3585, + "step": 37473 + }, + { + "epoch": 1.44, + "learning_rate": 2.6004103616311874e-05, + "loss": 0.3046, + "step": 37474 + }, + { + "epoch": 1.44, + "learning_rate": 2.6003462426263146e-05, + "loss": 0.3937, + "step": 37475 + }, + { + "epoch": 1.44, + "learning_rate": 2.6002821236214414e-05, + "loss": 0.3127, + "step": 37476 + }, + { + "epoch": 1.44, + "learning_rate": 2.6002180046165686e-05, + "loss": 0.485, + "step": 37477 + }, + { + "epoch": 1.44, + "learning_rate": 2.600153885611695e-05, + "loss": 0.2972, + "step": 37478 + }, + { + "epoch": 1.44, + "learning_rate": 2.6000897666068226e-05, + "loss": 0.3693, + "step": 37479 + }, + { + "epoch": 1.44, + "learning_rate": 2.600025647601949e-05, + "loss": 0.4215, + "step": 37480 + }, + { + "epoch": 1.44, + "learning_rate": 2.5999615285970763e-05, + "loss": 0.4531, + "step": 37481 + }, + { + "epoch": 1.44, + "learning_rate": 2.5998974095922035e-05, + "loss": 0.2868, + "step": 37482 + }, + { + "epoch": 1.44, + "learning_rate": 2.5998332905873303e-05, + "loss": 0.4092, + "step": 37483 + }, + { + "epoch": 1.44, + "learning_rate": 2.5997691715824575e-05, + "loss": 0.3268, + "step": 37484 + }, + { + "epoch": 1.44, + "learning_rate": 2.599705052577584e-05, + "loss": 0.3521, + "step": 37485 + }, + { + "epoch": 1.44, + "learning_rate": 2.599640933572711e-05, + "loss": 0.3236, + "step": 37486 + }, + { + "epoch": 1.44, + "learning_rate": 2.599576814567838e-05, + "loss": 0.4385, + "step": 37487 + }, + { + "epoch": 1.44, + "learning_rate": 2.599512695562965e-05, + "loss": 0.3262, + "step": 37488 + }, + { + "epoch": 1.44, + "learning_rate": 2.5994485765580916e-05, + "loss": 0.3833, + "step": 37489 + }, + { + "epoch": 1.44, + "learning_rate": 2.5993844575532188e-05, + "loss": 0.2461, + "step": 37490 + }, + { + "epoch": 1.44, + "learning_rate": 2.599320338548346e-05, + "loss": 0.372, + "step": 37491 + }, + { + "epoch": 1.44, + "learning_rate": 2.5992562195434728e-05, + "loss": 0.4473, + "step": 37492 + }, + { + "epoch": 1.44, + "learning_rate": 2.5991921005386e-05, + "loss": 0.319, + "step": 37493 + }, + { + "epoch": 1.44, + "learning_rate": 2.5991279815337265e-05, + "loss": 0.2968, + "step": 37494 + }, + { + "epoch": 1.44, + "learning_rate": 2.599063862528854e-05, + "loss": 0.4727, + "step": 37495 + }, + { + "epoch": 1.44, + "learning_rate": 2.5989997435239805e-05, + "loss": 0.3645, + "step": 37496 + }, + { + "epoch": 1.44, + "learning_rate": 2.5989356245191076e-05, + "loss": 0.4277, + "step": 37497 + }, + { + "epoch": 1.44, + "learning_rate": 2.5988715055142345e-05, + "loss": 0.294, + "step": 37498 + }, + { + "epoch": 1.44, + "learning_rate": 2.5988073865093616e-05, + "loss": 0.3752, + "step": 37499 + }, + { + "epoch": 1.44, + "learning_rate": 2.5987432675044888e-05, + "loss": 0.3561, + "step": 37500 + }, + { + "epoch": 1.44, + "learning_rate": 2.5986791484996153e-05, + "loss": 0.1884, + "step": 37501 + }, + { + "epoch": 1.44, + "learning_rate": 2.5986150294947425e-05, + "loss": 0.4042, + "step": 37502 + }, + { + "epoch": 1.44, + "learning_rate": 2.5985509104898693e-05, + "loss": 0.3097, + "step": 37503 + }, + { + "epoch": 1.44, + "learning_rate": 2.5984867914849965e-05, + "loss": 0.3537, + "step": 37504 + }, + { + "epoch": 1.44, + "learning_rate": 2.598422672480123e-05, + "loss": 0.2618, + "step": 37505 + }, + { + "epoch": 1.44, + "learning_rate": 2.59835855347525e-05, + "loss": 0.3437, + "step": 37506 + }, + { + "epoch": 1.44, + "learning_rate": 2.598294434470377e-05, + "loss": 0.4862, + "step": 37507 + }, + { + "epoch": 1.44, + "learning_rate": 2.598230315465504e-05, + "loss": 0.3769, + "step": 37508 + }, + { + "epoch": 1.44, + "learning_rate": 2.5981661964606313e-05, + "loss": 0.3204, + "step": 37509 + }, + { + "epoch": 1.44, + "learning_rate": 2.5981020774557578e-05, + "loss": 0.3957, + "step": 37510 + }, + { + "epoch": 1.44, + "learning_rate": 2.598037958450885e-05, + "loss": 0.4004, + "step": 37511 + }, + { + "epoch": 1.44, + "learning_rate": 2.5979738394460118e-05, + "loss": 0.3609, + "step": 37512 + }, + { + "epoch": 1.44, + "learning_rate": 2.597909720441139e-05, + "loss": 0.3257, + "step": 37513 + }, + { + "epoch": 1.44, + "learning_rate": 2.5978456014362655e-05, + "loss": 0.3227, + "step": 37514 + }, + { + "epoch": 1.44, + "learning_rate": 2.597781482431393e-05, + "loss": 0.2004, + "step": 37515 + }, + { + "epoch": 1.44, + "learning_rate": 2.5977173634265195e-05, + "loss": 0.3178, + "step": 37516 + }, + { + "epoch": 1.44, + "learning_rate": 2.5976532444216467e-05, + "loss": 0.4606, + "step": 37517 + }, + { + "epoch": 1.44, + "learning_rate": 2.597589125416774e-05, + "loss": 0.3912, + "step": 37518 + }, + { + "epoch": 1.44, + "learning_rate": 2.5975250064119007e-05, + "loss": 0.2556, + "step": 37519 + }, + { + "epoch": 1.44, + "learning_rate": 2.597460887407028e-05, + "loss": 0.3873, + "step": 37520 + }, + { + "epoch": 1.44, + "learning_rate": 2.5973967684021543e-05, + "loss": 0.4159, + "step": 37521 + }, + { + "epoch": 1.44, + "learning_rate": 2.5973326493972815e-05, + "loss": 0.3731, + "step": 37522 + }, + { + "epoch": 1.44, + "learning_rate": 2.5972685303924083e-05, + "loss": 0.3243, + "step": 37523 + }, + { + "epoch": 1.44, + "learning_rate": 2.5972044113875355e-05, + "loss": 0.5269, + "step": 37524 + }, + { + "epoch": 1.44, + "learning_rate": 2.597140292382662e-05, + "loss": 0.2682, + "step": 37525 + }, + { + "epoch": 1.44, + "learning_rate": 2.5970761733777892e-05, + "loss": 0.4132, + "step": 37526 + }, + { + "epoch": 1.44, + "learning_rate": 2.5970120543729163e-05, + "loss": 0.4363, + "step": 37527 + }, + { + "epoch": 1.44, + "learning_rate": 2.5969479353680432e-05, + "loss": 0.3935, + "step": 37528 + }, + { + "epoch": 1.44, + "learning_rate": 2.5968838163631704e-05, + "loss": 0.3006, + "step": 37529 + }, + { + "epoch": 1.44, + "learning_rate": 2.596819697358297e-05, + "loss": 0.3236, + "step": 37530 + }, + { + "epoch": 1.44, + "learning_rate": 2.5967555783534244e-05, + "loss": 0.2914, + "step": 37531 + }, + { + "epoch": 1.44, + "learning_rate": 2.596691459348551e-05, + "loss": 0.4586, + "step": 37532 + }, + { + "epoch": 1.44, + "learning_rate": 2.596627340343678e-05, + "loss": 0.342, + "step": 37533 + }, + { + "epoch": 1.44, + "learning_rate": 2.596563221338805e-05, + "loss": 0.4585, + "step": 37534 + }, + { + "epoch": 1.44, + "learning_rate": 2.596499102333932e-05, + "loss": 0.3453, + "step": 37535 + }, + { + "epoch": 1.44, + "learning_rate": 2.5964349833290592e-05, + "loss": 0.3227, + "step": 37536 + }, + { + "epoch": 1.44, + "learning_rate": 2.5963708643241857e-05, + "loss": 0.4553, + "step": 37537 + }, + { + "epoch": 1.44, + "learning_rate": 2.596306745319313e-05, + "loss": 0.4134, + "step": 37538 + }, + { + "epoch": 1.44, + "learning_rate": 2.5962426263144397e-05, + "loss": 0.3569, + "step": 37539 + }, + { + "epoch": 1.44, + "learning_rate": 2.596178507309567e-05, + "loss": 0.3697, + "step": 37540 + }, + { + "epoch": 1.44, + "learning_rate": 2.5961143883046934e-05, + "loss": 0.4519, + "step": 37541 + }, + { + "epoch": 1.44, + "learning_rate": 2.5960502692998205e-05, + "loss": 0.3705, + "step": 37542 + }, + { + "epoch": 1.44, + "learning_rate": 2.5959861502949474e-05, + "loss": 0.4761, + "step": 37543 + }, + { + "epoch": 1.44, + "learning_rate": 2.5959220312900745e-05, + "loss": 0.3708, + "step": 37544 + }, + { + "epoch": 1.44, + "learning_rate": 2.5958579122852017e-05, + "loss": 0.3703, + "step": 37545 + }, + { + "epoch": 1.44, + "learning_rate": 2.5957937932803282e-05, + "loss": 0.204, + "step": 37546 + }, + { + "epoch": 1.44, + "learning_rate": 2.5957296742754554e-05, + "loss": 0.3659, + "step": 37547 + }, + { + "epoch": 1.44, + "learning_rate": 2.5956655552705822e-05, + "loss": 0.3398, + "step": 37548 + }, + { + "epoch": 1.44, + "learning_rate": 2.5956014362657094e-05, + "loss": 0.5144, + "step": 37549 + }, + { + "epoch": 1.44, + "learning_rate": 2.595537317260836e-05, + "loss": 0.3494, + "step": 37550 + }, + { + "epoch": 1.44, + "learning_rate": 2.5954731982559634e-05, + "loss": 0.3861, + "step": 37551 + }, + { + "epoch": 1.44, + "learning_rate": 2.59540907925109e-05, + "loss": 0.4122, + "step": 37552 + }, + { + "epoch": 1.44, + "learning_rate": 2.595344960246217e-05, + "loss": 0.4387, + "step": 37553 + }, + { + "epoch": 1.44, + "learning_rate": 2.5952808412413442e-05, + "loss": 0.3926, + "step": 37554 + }, + { + "epoch": 1.44, + "learning_rate": 2.595216722236471e-05, + "loss": 0.266, + "step": 37555 + }, + { + "epoch": 1.44, + "learning_rate": 2.5951526032315982e-05, + "loss": 0.4651, + "step": 37556 + }, + { + "epoch": 1.44, + "learning_rate": 2.5950884842267247e-05, + "loss": 0.3629, + "step": 37557 + }, + { + "epoch": 1.44, + "learning_rate": 2.595024365221852e-05, + "loss": 0.4282, + "step": 37558 + }, + { + "epoch": 1.44, + "learning_rate": 2.5949602462169787e-05, + "loss": 0.3396, + "step": 37559 + }, + { + "epoch": 1.44, + "learning_rate": 2.594896127212106e-05, + "loss": 0.2709, + "step": 37560 + }, + { + "epoch": 1.44, + "learning_rate": 2.5948320082072324e-05, + "loss": 0.3761, + "step": 37561 + }, + { + "epoch": 1.44, + "learning_rate": 2.5947678892023596e-05, + "loss": 0.3741, + "step": 37562 + }, + { + "epoch": 1.44, + "learning_rate": 2.5947037701974867e-05, + "loss": 0.2822, + "step": 37563 + }, + { + "epoch": 1.44, + "learning_rate": 2.5946396511926136e-05, + "loss": 0.3037, + "step": 37564 + }, + { + "epoch": 1.44, + "learning_rate": 2.5945755321877407e-05, + "loss": 0.4048, + "step": 37565 + }, + { + "epoch": 1.44, + "learning_rate": 2.5945114131828672e-05, + "loss": 0.4304, + "step": 37566 + }, + { + "epoch": 1.44, + "learning_rate": 2.5944472941779947e-05, + "loss": 0.2463, + "step": 37567 + }, + { + "epoch": 1.44, + "learning_rate": 2.5943831751731212e-05, + "loss": 0.4877, + "step": 37568 + }, + { + "epoch": 1.44, + "learning_rate": 2.5943190561682484e-05, + "loss": 0.3783, + "step": 37569 + }, + { + "epoch": 1.44, + "learning_rate": 2.5942549371633752e-05, + "loss": 0.3809, + "step": 37570 + }, + { + "epoch": 1.44, + "learning_rate": 2.5941908181585024e-05, + "loss": 0.2707, + "step": 37571 + }, + { + "epoch": 1.44, + "learning_rate": 2.5941266991536296e-05, + "loss": 0.333, + "step": 37572 + }, + { + "epoch": 1.44, + "learning_rate": 2.594062580148756e-05, + "loss": 0.3123, + "step": 37573 + }, + { + "epoch": 1.44, + "learning_rate": 2.5939984611438833e-05, + "loss": 0.3668, + "step": 37574 + }, + { + "epoch": 1.44, + "learning_rate": 2.59393434213901e-05, + "loss": 0.4137, + "step": 37575 + }, + { + "epoch": 1.44, + "learning_rate": 2.5938702231341373e-05, + "loss": 0.3539, + "step": 37576 + }, + { + "epoch": 1.44, + "learning_rate": 2.5938061041292638e-05, + "loss": 0.3496, + "step": 37577 + }, + { + "epoch": 1.44, + "learning_rate": 2.593741985124391e-05, + "loss": 0.454, + "step": 37578 + }, + { + "epoch": 1.44, + "learning_rate": 2.5936778661195178e-05, + "loss": 0.243, + "step": 37579 + }, + { + "epoch": 1.44, + "learning_rate": 2.593613747114645e-05, + "loss": 0.3667, + "step": 37580 + }, + { + "epoch": 1.44, + "learning_rate": 2.593549628109772e-05, + "loss": 0.507, + "step": 37581 + }, + { + "epoch": 1.44, + "learning_rate": 2.5934855091048986e-05, + "loss": 0.331, + "step": 37582 + }, + { + "epoch": 1.44, + "learning_rate": 2.593421390100026e-05, + "loss": 0.3092, + "step": 37583 + }, + { + "epoch": 1.44, + "learning_rate": 2.5933572710951526e-05, + "loss": 0.5286, + "step": 37584 + }, + { + "epoch": 1.45, + "learning_rate": 2.5932931520902798e-05, + "loss": 0.5576, + "step": 37585 + }, + { + "epoch": 1.45, + "learning_rate": 2.5932290330854066e-05, + "loss": 0.43, + "step": 37586 + }, + { + "epoch": 1.45, + "learning_rate": 2.5931649140805338e-05, + "loss": 0.4966, + "step": 37587 + }, + { + "epoch": 1.45, + "learning_rate": 2.5931007950756603e-05, + "loss": 0.3113, + "step": 37588 + }, + { + "epoch": 1.45, + "learning_rate": 2.5930366760707874e-05, + "loss": 0.3168, + "step": 37589 + }, + { + "epoch": 1.45, + "learning_rate": 2.5929725570659146e-05, + "loss": 0.329, + "step": 37590 + }, + { + "epoch": 1.45, + "learning_rate": 2.5929084380610414e-05, + "loss": 0.4578, + "step": 37591 + }, + { + "epoch": 1.45, + "learning_rate": 2.5928443190561686e-05, + "loss": 0.4869, + "step": 37592 + }, + { + "epoch": 1.45, + "learning_rate": 2.592780200051295e-05, + "loss": 0.3971, + "step": 37593 + }, + { + "epoch": 1.45, + "learning_rate": 2.5927160810464223e-05, + "loss": 0.3371, + "step": 37594 + }, + { + "epoch": 1.45, + "learning_rate": 2.592651962041549e-05, + "loss": 0.3062, + "step": 37595 + }, + { + "epoch": 1.45, + "learning_rate": 2.5925878430366763e-05, + "loss": 0.2735, + "step": 37596 + }, + { + "epoch": 1.45, + "learning_rate": 2.5925237240318028e-05, + "loss": 0.3529, + "step": 37597 + }, + { + "epoch": 1.45, + "learning_rate": 2.59245960502693e-05, + "loss": 0.3729, + "step": 37598 + }, + { + "epoch": 1.45, + "learning_rate": 2.592395486022057e-05, + "loss": 0.2813, + "step": 37599 + }, + { + "epoch": 1.45, + "learning_rate": 2.592331367017184e-05, + "loss": 0.3193, + "step": 37600 + }, + { + "epoch": 1.45, + "learning_rate": 2.592267248012311e-05, + "loss": 0.4077, + "step": 37601 + }, + { + "epoch": 1.45, + "learning_rate": 2.5922031290074376e-05, + "loss": 0.3847, + "step": 37602 + }, + { + "epoch": 1.45, + "learning_rate": 2.592139010002565e-05, + "loss": 0.3872, + "step": 37603 + }, + { + "epoch": 1.45, + "learning_rate": 2.5920748909976916e-05, + "loss": 0.4215, + "step": 37604 + }, + { + "epoch": 1.45, + "learning_rate": 2.5920107719928188e-05, + "loss": 0.4272, + "step": 37605 + }, + { + "epoch": 1.45, + "learning_rate": 2.5919466529879456e-05, + "loss": 0.3482, + "step": 37606 + }, + { + "epoch": 1.45, + "learning_rate": 2.5918825339830728e-05, + "loss": 0.3508, + "step": 37607 + }, + { + "epoch": 1.45, + "learning_rate": 2.5918184149782e-05, + "loss": 0.2939, + "step": 37608 + }, + { + "epoch": 1.45, + "learning_rate": 2.5917542959733265e-05, + "loss": 0.2989, + "step": 37609 + }, + { + "epoch": 1.45, + "learning_rate": 2.5916901769684536e-05, + "loss": 0.4241, + "step": 37610 + }, + { + "epoch": 1.45, + "learning_rate": 2.5916260579635805e-05, + "loss": 0.4244, + "step": 37611 + }, + { + "epoch": 1.45, + "learning_rate": 2.5915619389587076e-05, + "loss": 0.3717, + "step": 37612 + }, + { + "epoch": 1.45, + "learning_rate": 2.591497819953834e-05, + "loss": 0.2554, + "step": 37613 + }, + { + "epoch": 1.45, + "learning_rate": 2.5914337009489613e-05, + "loss": 0.5172, + "step": 37614 + }, + { + "epoch": 1.45, + "learning_rate": 2.591369581944088e-05, + "loss": 0.4476, + "step": 37615 + }, + { + "epoch": 1.45, + "learning_rate": 2.5913054629392153e-05, + "loss": 0.522, + "step": 37616 + }, + { + "epoch": 1.45, + "learning_rate": 2.5912413439343425e-05, + "loss": 0.4305, + "step": 37617 + }, + { + "epoch": 1.45, + "learning_rate": 2.591177224929469e-05, + "loss": 0.4237, + "step": 37618 + }, + { + "epoch": 1.45, + "learning_rate": 2.5911131059245965e-05, + "loss": 0.2978, + "step": 37619 + }, + { + "epoch": 1.45, + "learning_rate": 2.591048986919723e-05, + "loss": 0.3816, + "step": 37620 + }, + { + "epoch": 1.45, + "learning_rate": 2.59098486791485e-05, + "loss": 0.371, + "step": 37621 + }, + { + "epoch": 1.45, + "learning_rate": 2.590920748909977e-05, + "loss": 0.3742, + "step": 37622 + }, + { + "epoch": 1.45, + "learning_rate": 2.590856629905104e-05, + "loss": 0.4881, + "step": 37623 + }, + { + "epoch": 1.45, + "learning_rate": 2.5907925109002313e-05, + "loss": 0.3317, + "step": 37624 + }, + { + "epoch": 1.45, + "learning_rate": 2.5907283918953578e-05, + "loss": 0.4051, + "step": 37625 + }, + { + "epoch": 1.45, + "learning_rate": 2.590664272890485e-05, + "loss": 0.3283, + "step": 37626 + }, + { + "epoch": 1.45, + "learning_rate": 2.590600153885612e-05, + "loss": 0.3998, + "step": 37627 + }, + { + "epoch": 1.45, + "learning_rate": 2.590536034880739e-05, + "loss": 0.3222, + "step": 37628 + }, + { + "epoch": 1.45, + "learning_rate": 2.5904719158758655e-05, + "loss": 0.3289, + "step": 37629 + }, + { + "epoch": 1.45, + "learning_rate": 2.5904077968709927e-05, + "loss": 0.3496, + "step": 37630 + }, + { + "epoch": 1.45, + "learning_rate": 2.5903436778661195e-05, + "loss": 0.3943, + "step": 37631 + }, + { + "epoch": 1.45, + "learning_rate": 2.5902795588612467e-05, + "loss": 0.4252, + "step": 37632 + }, + { + "epoch": 1.45, + "learning_rate": 2.590215439856374e-05, + "loss": 0.4147, + "step": 37633 + }, + { + "epoch": 1.45, + "learning_rate": 2.5901513208515003e-05, + "loss": 0.3548, + "step": 37634 + }, + { + "epoch": 1.45, + "learning_rate": 2.5900872018466275e-05, + "loss": 0.3296, + "step": 37635 + }, + { + "epoch": 1.45, + "learning_rate": 2.5900230828417543e-05, + "loss": 0.3574, + "step": 37636 + }, + { + "epoch": 1.45, + "learning_rate": 2.5899589638368815e-05, + "loss": 0.4329, + "step": 37637 + }, + { + "epoch": 1.45, + "learning_rate": 2.589894844832008e-05, + "loss": 0.3974, + "step": 37638 + }, + { + "epoch": 1.45, + "learning_rate": 2.5898307258271355e-05, + "loss": 0.3368, + "step": 37639 + }, + { + "epoch": 1.45, + "learning_rate": 2.589766606822262e-05, + "loss": 0.3042, + "step": 37640 + }, + { + "epoch": 1.45, + "learning_rate": 2.5897024878173892e-05, + "loss": 0.2176, + "step": 37641 + }, + { + "epoch": 1.45, + "learning_rate": 2.5896383688125164e-05, + "loss": 0.4199, + "step": 37642 + }, + { + "epoch": 1.45, + "learning_rate": 2.5895742498076432e-05, + "loss": 0.4084, + "step": 37643 + }, + { + "epoch": 1.45, + "learning_rate": 2.5895101308027704e-05, + "loss": 0.3853, + "step": 37644 + }, + { + "epoch": 1.45, + "learning_rate": 2.589446011797897e-05, + "loss": 0.4333, + "step": 37645 + }, + { + "epoch": 1.45, + "learning_rate": 2.589381892793024e-05, + "loss": 0.3137, + "step": 37646 + }, + { + "epoch": 1.45, + "learning_rate": 2.589317773788151e-05, + "loss": 0.4454, + "step": 37647 + }, + { + "epoch": 1.45, + "learning_rate": 2.589253654783278e-05, + "loss": 0.3604, + "step": 37648 + }, + { + "epoch": 1.45, + "learning_rate": 2.5891895357784045e-05, + "loss": 0.3396, + "step": 37649 + }, + { + "epoch": 1.45, + "learning_rate": 2.5891254167735317e-05, + "loss": 0.4, + "step": 37650 + }, + { + "epoch": 1.45, + "learning_rate": 2.589061297768659e-05, + "loss": 0.3229, + "step": 37651 + }, + { + "epoch": 1.45, + "learning_rate": 2.5889971787637857e-05, + "loss": 0.3014, + "step": 37652 + }, + { + "epoch": 1.45, + "learning_rate": 2.588933059758913e-05, + "loss": 0.3965, + "step": 37653 + }, + { + "epoch": 1.45, + "learning_rate": 2.5888689407540394e-05, + "loss": 0.26, + "step": 37654 + }, + { + "epoch": 1.45, + "learning_rate": 2.588804821749167e-05, + "loss": 0.2366, + "step": 37655 + }, + { + "epoch": 1.45, + "learning_rate": 2.5887407027442934e-05, + "loss": 0.3156, + "step": 37656 + }, + { + "epoch": 1.45, + "learning_rate": 2.5886765837394205e-05, + "loss": 0.2982, + "step": 37657 + }, + { + "epoch": 1.45, + "learning_rate": 2.5886124647345474e-05, + "loss": 0.3209, + "step": 37658 + }, + { + "epoch": 1.45, + "learning_rate": 2.5885483457296745e-05, + "loss": 0.396, + "step": 37659 + }, + { + "epoch": 1.45, + "learning_rate": 2.5884842267248017e-05, + "loss": 0.3625, + "step": 37660 + }, + { + "epoch": 1.45, + "learning_rate": 2.5884201077199282e-05, + "loss": 0.3293, + "step": 37661 + }, + { + "epoch": 1.45, + "learning_rate": 2.5883559887150554e-05, + "loss": 0.3307, + "step": 37662 + }, + { + "epoch": 1.45, + "learning_rate": 2.5882918697101822e-05, + "loss": 0.3554, + "step": 37663 + }, + { + "epoch": 1.45, + "learning_rate": 2.5882277507053094e-05, + "loss": 0.4251, + "step": 37664 + }, + { + "epoch": 1.45, + "learning_rate": 2.588163631700436e-05, + "loss": 0.3953, + "step": 37665 + }, + { + "epoch": 1.45, + "learning_rate": 2.588099512695563e-05, + "loss": 0.3333, + "step": 37666 + }, + { + "epoch": 1.45, + "learning_rate": 2.58803539369069e-05, + "loss": 0.5075, + "step": 37667 + }, + { + "epoch": 1.45, + "learning_rate": 2.587971274685817e-05, + "loss": 0.3235, + "step": 37668 + }, + { + "epoch": 1.45, + "learning_rate": 2.5879071556809442e-05, + "loss": 0.3405, + "step": 37669 + }, + { + "epoch": 1.45, + "learning_rate": 2.5878430366760707e-05, + "loss": 0.277, + "step": 37670 + }, + { + "epoch": 1.45, + "learning_rate": 2.587778917671198e-05, + "loss": 0.4467, + "step": 37671 + }, + { + "epoch": 1.45, + "learning_rate": 2.5877147986663247e-05, + "loss": 0.5051, + "step": 37672 + }, + { + "epoch": 1.45, + "learning_rate": 2.587650679661452e-05, + "loss": 0.3612, + "step": 37673 + }, + { + "epoch": 1.45, + "learning_rate": 2.5875865606565784e-05, + "loss": 0.4164, + "step": 37674 + }, + { + "epoch": 1.45, + "learning_rate": 2.587522441651706e-05, + "loss": 0.3392, + "step": 37675 + }, + { + "epoch": 1.45, + "learning_rate": 2.5874583226468324e-05, + "loss": 0.2918, + "step": 37676 + }, + { + "epoch": 1.45, + "learning_rate": 2.5873942036419596e-05, + "loss": 0.3846, + "step": 37677 + }, + { + "epoch": 1.45, + "learning_rate": 2.5873300846370867e-05, + "loss": 0.3153, + "step": 37678 + }, + { + "epoch": 1.45, + "learning_rate": 2.5872659656322136e-05, + "loss": 0.4338, + "step": 37679 + }, + { + "epoch": 1.45, + "learning_rate": 2.5872018466273407e-05, + "loss": 0.4821, + "step": 37680 + }, + { + "epoch": 1.45, + "learning_rate": 2.5871377276224672e-05, + "loss": 0.2851, + "step": 37681 + }, + { + "epoch": 1.45, + "learning_rate": 2.5870736086175944e-05, + "loss": 0.339, + "step": 37682 + }, + { + "epoch": 1.45, + "learning_rate": 2.5870094896127212e-05, + "loss": 0.3665, + "step": 37683 + }, + { + "epoch": 1.45, + "learning_rate": 2.5869453706078484e-05, + "loss": 0.4016, + "step": 37684 + }, + { + "epoch": 1.45, + "learning_rate": 2.586881251602975e-05, + "loss": 0.4375, + "step": 37685 + }, + { + "epoch": 1.45, + "learning_rate": 2.586817132598102e-05, + "loss": 0.3269, + "step": 37686 + }, + { + "epoch": 1.45, + "learning_rate": 2.5867530135932293e-05, + "loss": 0.4816, + "step": 37687 + }, + { + "epoch": 1.45, + "learning_rate": 2.586688894588356e-05, + "loss": 0.2528, + "step": 37688 + }, + { + "epoch": 1.45, + "learning_rate": 2.5866247755834833e-05, + "loss": 0.2805, + "step": 37689 + }, + { + "epoch": 1.45, + "learning_rate": 2.5865606565786098e-05, + "loss": 0.3722, + "step": 37690 + }, + { + "epoch": 1.45, + "learning_rate": 2.5864965375737373e-05, + "loss": 0.3926, + "step": 37691 + }, + { + "epoch": 1.45, + "learning_rate": 2.5864324185688638e-05, + "loss": 0.5419, + "step": 37692 + }, + { + "epoch": 1.45, + "learning_rate": 2.586368299563991e-05, + "loss": 0.3846, + "step": 37693 + }, + { + "epoch": 1.45, + "learning_rate": 2.5863041805591178e-05, + "loss": 0.5322, + "step": 37694 + }, + { + "epoch": 1.45, + "learning_rate": 2.586240061554245e-05, + "loss": 0.3753, + "step": 37695 + }, + { + "epoch": 1.45, + "learning_rate": 2.586175942549372e-05, + "loss": 0.4284, + "step": 37696 + }, + { + "epoch": 1.45, + "learning_rate": 2.5861118235444986e-05, + "loss": 0.3257, + "step": 37697 + }, + { + "epoch": 1.45, + "learning_rate": 2.5860477045396258e-05, + "loss": 0.3344, + "step": 37698 + }, + { + "epoch": 1.45, + "learning_rate": 2.5859835855347526e-05, + "loss": 0.3274, + "step": 37699 + }, + { + "epoch": 1.45, + "learning_rate": 2.5859194665298798e-05, + "loss": 0.416, + "step": 37700 + }, + { + "epoch": 1.45, + "learning_rate": 2.5858553475250063e-05, + "loss": 0.2848, + "step": 37701 + }, + { + "epoch": 1.45, + "learning_rate": 2.5857912285201334e-05, + "loss": 0.2699, + "step": 37702 + }, + { + "epoch": 1.45, + "learning_rate": 2.5857271095152603e-05, + "loss": 0.472, + "step": 37703 + }, + { + "epoch": 1.45, + "learning_rate": 2.5856629905103874e-05, + "loss": 0.319, + "step": 37704 + }, + { + "epoch": 1.45, + "learning_rate": 2.5855988715055146e-05, + "loss": 0.2802, + "step": 37705 + }, + { + "epoch": 1.45, + "learning_rate": 2.585534752500641e-05, + "loss": 0.291, + "step": 37706 + }, + { + "epoch": 1.45, + "learning_rate": 2.5854706334957686e-05, + "loss": 0.318, + "step": 37707 + }, + { + "epoch": 1.45, + "learning_rate": 2.585406514490895e-05, + "loss": 0.3933, + "step": 37708 + }, + { + "epoch": 1.45, + "learning_rate": 2.5853423954860223e-05, + "loss": 0.3438, + "step": 37709 + }, + { + "epoch": 1.45, + "learning_rate": 2.585278276481149e-05, + "loss": 0.3936, + "step": 37710 + }, + { + "epoch": 1.45, + "learning_rate": 2.5852141574762763e-05, + "loss": 0.3151, + "step": 37711 + }, + { + "epoch": 1.45, + "learning_rate": 2.5851500384714028e-05, + "loss": 0.2606, + "step": 37712 + }, + { + "epoch": 1.45, + "learning_rate": 2.58508591946653e-05, + "loss": 0.3616, + "step": 37713 + }, + { + "epoch": 1.45, + "learning_rate": 2.585021800461657e-05, + "loss": 0.3125, + "step": 37714 + }, + { + "epoch": 1.45, + "learning_rate": 2.584957681456784e-05, + "loss": 0.4571, + "step": 37715 + }, + { + "epoch": 1.45, + "learning_rate": 2.584893562451911e-05, + "loss": 0.298, + "step": 37716 + }, + { + "epoch": 1.45, + "learning_rate": 2.5848294434470376e-05, + "loss": 0.303, + "step": 37717 + }, + { + "epoch": 1.45, + "learning_rate": 2.5847653244421648e-05, + "loss": 0.3443, + "step": 37718 + }, + { + "epoch": 1.45, + "learning_rate": 2.5847012054372916e-05, + "loss": 0.335, + "step": 37719 + }, + { + "epoch": 1.45, + "learning_rate": 2.5846370864324188e-05, + "loss": 0.3495, + "step": 37720 + }, + { + "epoch": 1.45, + "learning_rate": 2.5845729674275453e-05, + "loss": 0.3975, + "step": 37721 + }, + { + "epoch": 1.45, + "learning_rate": 2.5845088484226725e-05, + "loss": 0.3979, + "step": 37722 + }, + { + "epoch": 1.45, + "learning_rate": 2.5844447294177996e-05, + "loss": 0.2122, + "step": 37723 + }, + { + "epoch": 1.45, + "learning_rate": 2.5843806104129265e-05, + "loss": 0.4255, + "step": 37724 + }, + { + "epoch": 1.45, + "learning_rate": 2.5843164914080536e-05, + "loss": 0.419, + "step": 37725 + }, + { + "epoch": 1.45, + "learning_rate": 2.58425237240318e-05, + "loss": 0.456, + "step": 37726 + }, + { + "epoch": 1.45, + "learning_rate": 2.5841882533983077e-05, + "loss": 0.3624, + "step": 37727 + }, + { + "epoch": 1.45, + "learning_rate": 2.584124134393434e-05, + "loss": 0.4609, + "step": 37728 + }, + { + "epoch": 1.45, + "learning_rate": 2.5840600153885613e-05, + "loss": 0.3555, + "step": 37729 + }, + { + "epoch": 1.45, + "learning_rate": 2.583995896383688e-05, + "loss": 0.3093, + "step": 37730 + }, + { + "epoch": 1.45, + "learning_rate": 2.5839317773788153e-05, + "loss": 0.5105, + "step": 37731 + }, + { + "epoch": 1.45, + "learning_rate": 2.5838676583739425e-05, + "loss": 0.2395, + "step": 37732 + }, + { + "epoch": 1.45, + "learning_rate": 2.583803539369069e-05, + "loss": 0.4086, + "step": 37733 + }, + { + "epoch": 1.45, + "learning_rate": 2.583739420364196e-05, + "loss": 0.3715, + "step": 37734 + }, + { + "epoch": 1.45, + "learning_rate": 2.583675301359323e-05, + "loss": 0.304, + "step": 37735 + }, + { + "epoch": 1.45, + "learning_rate": 2.58361118235445e-05, + "loss": 0.2608, + "step": 37736 + }, + { + "epoch": 1.45, + "learning_rate": 2.5835470633495767e-05, + "loss": 0.4651, + "step": 37737 + }, + { + "epoch": 1.45, + "learning_rate": 2.5834829443447038e-05, + "loss": 0.5161, + "step": 37738 + }, + { + "epoch": 1.45, + "learning_rate": 2.5834188253398307e-05, + "loss": 0.4807, + "step": 37739 + }, + { + "epoch": 1.45, + "learning_rate": 2.583354706334958e-05, + "loss": 0.3701, + "step": 37740 + }, + { + "epoch": 1.45, + "learning_rate": 2.583290587330085e-05, + "loss": 0.4454, + "step": 37741 + }, + { + "epoch": 1.45, + "learning_rate": 2.5832264683252115e-05, + "loss": 0.2742, + "step": 37742 + }, + { + "epoch": 1.45, + "learning_rate": 2.583162349320339e-05, + "loss": 0.3317, + "step": 37743 + }, + { + "epoch": 1.45, + "learning_rate": 2.5830982303154655e-05, + "loss": 0.2856, + "step": 37744 + }, + { + "epoch": 1.45, + "learning_rate": 2.5830341113105927e-05, + "loss": 0.3134, + "step": 37745 + }, + { + "epoch": 1.45, + "learning_rate": 2.5829699923057195e-05, + "loss": 0.4254, + "step": 37746 + }, + { + "epoch": 1.45, + "learning_rate": 2.5829058733008467e-05, + "loss": 0.364, + "step": 37747 + }, + { + "epoch": 1.45, + "learning_rate": 2.5828417542959732e-05, + "loss": 0.3873, + "step": 37748 + }, + { + "epoch": 1.45, + "learning_rate": 2.5827776352911003e-05, + "loss": 0.3367, + "step": 37749 + }, + { + "epoch": 1.45, + "learning_rate": 2.5827135162862275e-05, + "loss": 0.4776, + "step": 37750 + }, + { + "epoch": 1.45, + "learning_rate": 2.5826493972813544e-05, + "loss": 0.3095, + "step": 37751 + }, + { + "epoch": 1.45, + "learning_rate": 2.5825852782764815e-05, + "loss": 0.3899, + "step": 37752 + }, + { + "epoch": 1.45, + "learning_rate": 2.582521159271608e-05, + "loss": 0.3258, + "step": 37753 + }, + { + "epoch": 1.45, + "learning_rate": 2.5824570402667352e-05, + "loss": 0.3327, + "step": 37754 + }, + { + "epoch": 1.45, + "learning_rate": 2.582392921261862e-05, + "loss": 0.4592, + "step": 37755 + }, + { + "epoch": 1.45, + "learning_rate": 2.5823288022569892e-05, + "loss": 0.3826, + "step": 37756 + }, + { + "epoch": 1.45, + "learning_rate": 2.5822646832521157e-05, + "loss": 0.239, + "step": 37757 + }, + { + "epoch": 1.45, + "learning_rate": 2.582200564247243e-05, + "loss": 0.3483, + "step": 37758 + }, + { + "epoch": 1.45, + "learning_rate": 2.58213644524237e-05, + "loss": 0.3033, + "step": 37759 + }, + { + "epoch": 1.45, + "learning_rate": 2.582072326237497e-05, + "loss": 0.2912, + "step": 37760 + }, + { + "epoch": 1.45, + "learning_rate": 2.582008207232624e-05, + "loss": 0.3926, + "step": 37761 + }, + { + "epoch": 1.45, + "learning_rate": 2.5819440882277505e-05, + "loss": 0.3958, + "step": 37762 + }, + { + "epoch": 1.45, + "learning_rate": 2.581879969222878e-05, + "loss": 0.373, + "step": 37763 + }, + { + "epoch": 1.45, + "learning_rate": 2.5818158502180045e-05, + "loss": 0.3111, + "step": 37764 + }, + { + "epoch": 1.45, + "learning_rate": 2.5817517312131317e-05, + "loss": 0.3688, + "step": 37765 + }, + { + "epoch": 1.45, + "learning_rate": 2.5816876122082585e-05, + "loss": 0.318, + "step": 37766 + }, + { + "epoch": 1.45, + "learning_rate": 2.5816234932033857e-05, + "loss": 0.4101, + "step": 37767 + }, + { + "epoch": 1.45, + "learning_rate": 2.581559374198513e-05, + "loss": 0.3278, + "step": 37768 + }, + { + "epoch": 1.45, + "learning_rate": 2.5814952551936394e-05, + "loss": 0.2455, + "step": 37769 + }, + { + "epoch": 1.45, + "learning_rate": 2.5814311361887665e-05, + "loss": 0.2752, + "step": 37770 + }, + { + "epoch": 1.45, + "learning_rate": 2.5813670171838934e-05, + "loss": 0.3777, + "step": 37771 + }, + { + "epoch": 1.45, + "learning_rate": 2.5813028981790206e-05, + "loss": 0.5495, + "step": 37772 + }, + { + "epoch": 1.45, + "learning_rate": 2.581238779174147e-05, + "loss": 0.3452, + "step": 37773 + }, + { + "epoch": 1.45, + "learning_rate": 2.5811746601692742e-05, + "loss": 0.4852, + "step": 37774 + }, + { + "epoch": 1.45, + "learning_rate": 2.581110541164401e-05, + "loss": 0.3918, + "step": 37775 + }, + { + "epoch": 1.45, + "learning_rate": 2.5810464221595282e-05, + "loss": 0.3556, + "step": 37776 + }, + { + "epoch": 1.45, + "learning_rate": 2.5809823031546554e-05, + "loss": 0.4994, + "step": 37777 + }, + { + "epoch": 1.45, + "learning_rate": 2.580918184149782e-05, + "loss": 0.5023, + "step": 37778 + }, + { + "epoch": 1.45, + "learning_rate": 2.5808540651449094e-05, + "loss": 0.2952, + "step": 37779 + }, + { + "epoch": 1.45, + "learning_rate": 2.580789946140036e-05, + "loss": 0.3205, + "step": 37780 + }, + { + "epoch": 1.45, + "learning_rate": 2.580725827135163e-05, + "loss": 0.366, + "step": 37781 + }, + { + "epoch": 1.45, + "learning_rate": 2.58066170813029e-05, + "loss": 0.2951, + "step": 37782 + }, + { + "epoch": 1.45, + "learning_rate": 2.580597589125417e-05, + "loss": 0.434, + "step": 37783 + }, + { + "epoch": 1.45, + "learning_rate": 2.5805334701205436e-05, + "loss": 0.313, + "step": 37784 + }, + { + "epoch": 1.45, + "learning_rate": 2.5804693511156707e-05, + "loss": 0.4353, + "step": 37785 + }, + { + "epoch": 1.45, + "learning_rate": 2.580405232110798e-05, + "loss": 0.2509, + "step": 37786 + }, + { + "epoch": 1.45, + "learning_rate": 2.5803411131059247e-05, + "loss": 0.4033, + "step": 37787 + }, + { + "epoch": 1.45, + "learning_rate": 2.580276994101052e-05, + "loss": 0.3436, + "step": 37788 + }, + { + "epoch": 1.45, + "learning_rate": 2.5802128750961784e-05, + "loss": 0.4128, + "step": 37789 + }, + { + "epoch": 1.45, + "learning_rate": 2.5801487560913056e-05, + "loss": 0.3434, + "step": 37790 + }, + { + "epoch": 1.45, + "learning_rate": 2.5800846370864324e-05, + "loss": 0.2871, + "step": 37791 + }, + { + "epoch": 1.45, + "learning_rate": 2.5800205180815596e-05, + "loss": 0.4366, + "step": 37792 + }, + { + "epoch": 1.45, + "learning_rate": 2.579956399076686e-05, + "loss": 0.378, + "step": 37793 + }, + { + "epoch": 1.45, + "learning_rate": 2.5798922800718132e-05, + "loss": 0.3159, + "step": 37794 + }, + { + "epoch": 1.45, + "learning_rate": 2.5798281610669404e-05, + "loss": 0.4254, + "step": 37795 + }, + { + "epoch": 1.45, + "learning_rate": 2.5797640420620673e-05, + "loss": 0.4268, + "step": 37796 + }, + { + "epoch": 1.45, + "learning_rate": 2.5796999230571944e-05, + "loss": 0.3296, + "step": 37797 + }, + { + "epoch": 1.45, + "learning_rate": 2.579635804052321e-05, + "loss": 0.3568, + "step": 37798 + }, + { + "epoch": 1.45, + "learning_rate": 2.5795716850474484e-05, + "loss": 0.3718, + "step": 37799 + }, + { + "epoch": 1.45, + "learning_rate": 2.579507566042575e-05, + "loss": 0.4119, + "step": 37800 + }, + { + "epoch": 1.45, + "learning_rate": 2.579443447037702e-05, + "loss": 0.4168, + "step": 37801 + }, + { + "epoch": 1.45, + "learning_rate": 2.579379328032829e-05, + "loss": 0.3693, + "step": 37802 + }, + { + "epoch": 1.45, + "learning_rate": 2.579315209027956e-05, + "loss": 0.3238, + "step": 37803 + }, + { + "epoch": 1.45, + "learning_rate": 2.5792510900230833e-05, + "loss": 0.3043, + "step": 37804 + }, + { + "epoch": 1.45, + "learning_rate": 2.5791869710182098e-05, + "loss": 0.4839, + "step": 37805 + }, + { + "epoch": 1.45, + "learning_rate": 2.579122852013337e-05, + "loss": 0.3704, + "step": 37806 + }, + { + "epoch": 1.45, + "learning_rate": 2.5790587330084638e-05, + "loss": 0.3387, + "step": 37807 + }, + { + "epoch": 1.45, + "learning_rate": 2.578994614003591e-05, + "loss": 0.4753, + "step": 37808 + }, + { + "epoch": 1.45, + "learning_rate": 2.5789304949987174e-05, + "loss": 0.3887, + "step": 37809 + }, + { + "epoch": 1.45, + "learning_rate": 2.5788663759938446e-05, + "loss": 0.4491, + "step": 37810 + }, + { + "epoch": 1.45, + "learning_rate": 2.5788022569889718e-05, + "loss": 0.6881, + "step": 37811 + }, + { + "epoch": 1.45, + "learning_rate": 2.5787381379840986e-05, + "loss": 0.3304, + "step": 37812 + }, + { + "epoch": 1.45, + "learning_rate": 2.5786740189792258e-05, + "loss": 0.4204, + "step": 37813 + }, + { + "epoch": 1.45, + "learning_rate": 2.5786098999743523e-05, + "loss": 0.4247, + "step": 37814 + }, + { + "epoch": 1.45, + "learning_rate": 2.5785457809694798e-05, + "loss": 0.2322, + "step": 37815 + }, + { + "epoch": 1.45, + "learning_rate": 2.5784816619646063e-05, + "loss": 0.2619, + "step": 37816 + }, + { + "epoch": 1.45, + "learning_rate": 2.5784175429597335e-05, + "loss": 0.3321, + "step": 37817 + }, + { + "epoch": 1.45, + "learning_rate": 2.5783534239548603e-05, + "loss": 0.4879, + "step": 37818 + }, + { + "epoch": 1.45, + "learning_rate": 2.5782893049499875e-05, + "loss": 0.3668, + "step": 37819 + }, + { + "epoch": 1.45, + "learning_rate": 2.5782251859451146e-05, + "loss": 0.3707, + "step": 37820 + }, + { + "epoch": 1.45, + "learning_rate": 2.578161066940241e-05, + "loss": 0.516, + "step": 37821 + }, + { + "epoch": 1.45, + "learning_rate": 2.5780969479353683e-05, + "loss": 0.2952, + "step": 37822 + }, + { + "epoch": 1.45, + "learning_rate": 2.578032828930495e-05, + "loss": 0.2545, + "step": 37823 + }, + { + "epoch": 1.45, + "learning_rate": 2.5779687099256223e-05, + "loss": 0.4152, + "step": 37824 + }, + { + "epoch": 1.45, + "learning_rate": 2.5779045909207488e-05, + "loss": 0.3726, + "step": 37825 + }, + { + "epoch": 1.45, + "learning_rate": 2.577840471915876e-05, + "loss": 0.3967, + "step": 37826 + }, + { + "epoch": 1.45, + "learning_rate": 2.5777763529110028e-05, + "loss": 0.2767, + "step": 37827 + }, + { + "epoch": 1.45, + "learning_rate": 2.57771223390613e-05, + "loss": 0.4992, + "step": 37828 + }, + { + "epoch": 1.45, + "learning_rate": 2.577648114901257e-05, + "loss": 0.3305, + "step": 37829 + }, + { + "epoch": 1.45, + "learning_rate": 2.5775839958963836e-05, + "loss": 0.3541, + "step": 37830 + }, + { + "epoch": 1.45, + "learning_rate": 2.577519876891511e-05, + "loss": 0.2899, + "step": 37831 + }, + { + "epoch": 1.45, + "learning_rate": 2.5774557578866376e-05, + "loss": 0.3748, + "step": 37832 + }, + { + "epoch": 1.45, + "learning_rate": 2.5773916388817648e-05, + "loss": 0.4317, + "step": 37833 + }, + { + "epoch": 1.45, + "learning_rate": 2.5773275198768913e-05, + "loss": 0.3809, + "step": 37834 + }, + { + "epoch": 1.45, + "learning_rate": 2.5772634008720188e-05, + "loss": 0.4836, + "step": 37835 + }, + { + "epoch": 1.45, + "learning_rate": 2.5771992818671453e-05, + "loss": 0.4975, + "step": 37836 + }, + { + "epoch": 1.45, + "learning_rate": 2.5771351628622725e-05, + "loss": 0.3573, + "step": 37837 + }, + { + "epoch": 1.45, + "learning_rate": 2.5770710438573997e-05, + "loss": 0.2956, + "step": 37838 + }, + { + "epoch": 1.45, + "learning_rate": 2.5770069248525265e-05, + "loss": 0.2583, + "step": 37839 + }, + { + "epoch": 1.45, + "learning_rate": 2.5769428058476537e-05, + "loss": 0.3969, + "step": 37840 + }, + { + "epoch": 1.45, + "learning_rate": 2.57687868684278e-05, + "loss": 0.3978, + "step": 37841 + }, + { + "epoch": 1.45, + "learning_rate": 2.5768145678379073e-05, + "loss": 0.4328, + "step": 37842 + }, + { + "epoch": 1.45, + "learning_rate": 2.576750448833034e-05, + "loss": 0.3185, + "step": 37843 + }, + { + "epoch": 1.45, + "learning_rate": 2.5766863298281613e-05, + "loss": 0.4035, + "step": 37844 + }, + { + "epoch": 1.46, + "learning_rate": 2.5766222108232878e-05, + "loss": 0.4285, + "step": 37845 + }, + { + "epoch": 1.46, + "learning_rate": 2.576558091818415e-05, + "loss": 0.3764, + "step": 37846 + }, + { + "epoch": 1.46, + "learning_rate": 2.576493972813542e-05, + "loss": 0.2288, + "step": 37847 + }, + { + "epoch": 1.46, + "learning_rate": 2.576429853808669e-05, + "loss": 0.398, + "step": 37848 + }, + { + "epoch": 1.46, + "learning_rate": 2.576365734803796e-05, + "loss": 0.4015, + "step": 37849 + }, + { + "epoch": 1.46, + "learning_rate": 2.5763016157989227e-05, + "loss": 0.3778, + "step": 37850 + }, + { + "epoch": 1.46, + "learning_rate": 2.5762374967940502e-05, + "loss": 0.2419, + "step": 37851 + }, + { + "epoch": 1.46, + "learning_rate": 2.5761733777891767e-05, + "loss": 0.4031, + "step": 37852 + }, + { + "epoch": 1.46, + "learning_rate": 2.576109258784304e-05, + "loss": 0.4618, + "step": 37853 + }, + { + "epoch": 1.46, + "learning_rate": 2.5760451397794307e-05, + "loss": 0.3155, + "step": 37854 + }, + { + "epoch": 1.46, + "learning_rate": 2.575981020774558e-05, + "loss": 0.3288, + "step": 37855 + }, + { + "epoch": 1.46, + "learning_rate": 2.575916901769685e-05, + "loss": 0.2747, + "step": 37856 + }, + { + "epoch": 1.46, + "learning_rate": 2.5758527827648115e-05, + "loss": 0.3124, + "step": 37857 + }, + { + "epoch": 1.46, + "learning_rate": 2.5757886637599387e-05, + "loss": 0.3574, + "step": 37858 + }, + { + "epoch": 1.46, + "learning_rate": 2.5757245447550655e-05, + "loss": 0.348, + "step": 37859 + }, + { + "epoch": 1.46, + "learning_rate": 2.5756604257501927e-05, + "loss": 0.2322, + "step": 37860 + }, + { + "epoch": 1.46, + "learning_rate": 2.5755963067453192e-05, + "loss": 0.2626, + "step": 37861 + }, + { + "epoch": 1.46, + "learning_rate": 2.5755321877404464e-05, + "loss": 0.2721, + "step": 37862 + }, + { + "epoch": 1.46, + "learning_rate": 2.5754680687355732e-05, + "loss": 0.3938, + "step": 37863 + }, + { + "epoch": 1.46, + "learning_rate": 2.5754039497307004e-05, + "loss": 0.2766, + "step": 37864 + }, + { + "epoch": 1.46, + "learning_rate": 2.5753398307258275e-05, + "loss": 0.2973, + "step": 37865 + }, + { + "epoch": 1.46, + "learning_rate": 2.575275711720954e-05, + "loss": 0.3856, + "step": 37866 + }, + { + "epoch": 1.46, + "learning_rate": 2.5752115927160815e-05, + "loss": 0.3591, + "step": 37867 + }, + { + "epoch": 1.46, + "learning_rate": 2.575147473711208e-05, + "loss": 0.3335, + "step": 37868 + }, + { + "epoch": 1.46, + "learning_rate": 2.5750833547063352e-05, + "loss": 0.4011, + "step": 37869 + }, + { + "epoch": 1.46, + "learning_rate": 2.575019235701462e-05, + "loss": 0.3837, + "step": 37870 + }, + { + "epoch": 1.46, + "learning_rate": 2.5749551166965892e-05, + "loss": 0.4451, + "step": 37871 + }, + { + "epoch": 1.46, + "learning_rate": 2.5748909976917157e-05, + "loss": 0.3508, + "step": 37872 + }, + { + "epoch": 1.46, + "learning_rate": 2.574826878686843e-05, + "loss": 0.3109, + "step": 37873 + }, + { + "epoch": 1.46, + "learning_rate": 2.57476275968197e-05, + "loss": 0.4068, + "step": 37874 + }, + { + "epoch": 1.46, + "learning_rate": 2.574698640677097e-05, + "loss": 0.4341, + "step": 37875 + }, + { + "epoch": 1.46, + "learning_rate": 2.574634521672224e-05, + "loss": 0.3945, + "step": 37876 + }, + { + "epoch": 1.46, + "learning_rate": 2.5745704026673505e-05, + "loss": 0.4828, + "step": 37877 + }, + { + "epoch": 1.46, + "learning_rate": 2.5745062836624777e-05, + "loss": 0.5036, + "step": 37878 + }, + { + "epoch": 1.46, + "learning_rate": 2.5744421646576045e-05, + "loss": 0.286, + "step": 37879 + }, + { + "epoch": 1.46, + "learning_rate": 2.5743780456527317e-05, + "loss": 0.284, + "step": 37880 + }, + { + "epoch": 1.46, + "learning_rate": 2.5743139266478582e-05, + "loss": 0.4142, + "step": 37881 + }, + { + "epoch": 1.46, + "learning_rate": 2.5742498076429854e-05, + "loss": 0.305, + "step": 37882 + }, + { + "epoch": 1.46, + "learning_rate": 2.5741856886381126e-05, + "loss": 0.3325, + "step": 37883 + }, + { + "epoch": 1.46, + "learning_rate": 2.5741215696332394e-05, + "loss": 0.3864, + "step": 37884 + }, + { + "epoch": 1.46, + "learning_rate": 2.5740574506283666e-05, + "loss": 0.4321, + "step": 37885 + }, + { + "epoch": 1.46, + "learning_rate": 2.573993331623493e-05, + "loss": 0.3049, + "step": 37886 + }, + { + "epoch": 1.46, + "learning_rate": 2.5739292126186206e-05, + "loss": 0.4551, + "step": 37887 + }, + { + "epoch": 1.46, + "learning_rate": 2.573865093613747e-05, + "loss": 0.3213, + "step": 37888 + }, + { + "epoch": 1.46, + "learning_rate": 2.5738009746088742e-05, + "loss": 0.3668, + "step": 37889 + }, + { + "epoch": 1.46, + "learning_rate": 2.573736855604001e-05, + "loss": 0.3462, + "step": 37890 + }, + { + "epoch": 1.46, + "learning_rate": 2.5736727365991282e-05, + "loss": 0.4407, + "step": 37891 + }, + { + "epoch": 1.46, + "learning_rate": 2.5736086175942554e-05, + "loss": 0.4619, + "step": 37892 + }, + { + "epoch": 1.46, + "learning_rate": 2.573544498589382e-05, + "loss": 0.413, + "step": 37893 + }, + { + "epoch": 1.46, + "learning_rate": 2.573480379584509e-05, + "loss": 0.2247, + "step": 37894 + }, + { + "epoch": 1.46, + "learning_rate": 2.573416260579636e-05, + "loss": 0.3612, + "step": 37895 + }, + { + "epoch": 1.46, + "learning_rate": 2.573352141574763e-05, + "loss": 0.3763, + "step": 37896 + }, + { + "epoch": 1.46, + "learning_rate": 2.5732880225698896e-05, + "loss": 0.2374, + "step": 37897 + }, + { + "epoch": 1.46, + "learning_rate": 2.5732239035650167e-05, + "loss": 0.3315, + "step": 37898 + }, + { + "epoch": 1.46, + "learning_rate": 2.5731597845601436e-05, + "loss": 0.2915, + "step": 37899 + }, + { + "epoch": 1.46, + "learning_rate": 2.5730956655552707e-05, + "loss": 0.2857, + "step": 37900 + }, + { + "epoch": 1.46, + "learning_rate": 2.573031546550398e-05, + "loss": 0.3556, + "step": 37901 + }, + { + "epoch": 1.46, + "learning_rate": 2.5729674275455244e-05, + "loss": 0.4016, + "step": 37902 + }, + { + "epoch": 1.46, + "learning_rate": 2.572903308540652e-05, + "loss": 0.4356, + "step": 37903 + }, + { + "epoch": 1.46, + "learning_rate": 2.5728391895357784e-05, + "loss": 0.2881, + "step": 37904 + }, + { + "epoch": 1.46, + "learning_rate": 2.5727750705309056e-05, + "loss": 0.2431, + "step": 37905 + }, + { + "epoch": 1.46, + "learning_rate": 2.5727109515260324e-05, + "loss": 0.3645, + "step": 37906 + }, + { + "epoch": 1.46, + "learning_rate": 2.5726468325211596e-05, + "loss": 0.2979, + "step": 37907 + }, + { + "epoch": 1.46, + "learning_rate": 2.572582713516286e-05, + "loss": 0.5256, + "step": 37908 + }, + { + "epoch": 1.46, + "learning_rate": 2.5725185945114133e-05, + "loss": 0.3309, + "step": 37909 + }, + { + "epoch": 1.46, + "learning_rate": 2.5724544755065404e-05, + "loss": 0.3371, + "step": 37910 + }, + { + "epoch": 1.46, + "learning_rate": 2.5723903565016673e-05, + "loss": 0.4712, + "step": 37911 + }, + { + "epoch": 1.46, + "learning_rate": 2.5723262374967944e-05, + "loss": 0.2947, + "step": 37912 + }, + { + "epoch": 1.46, + "learning_rate": 2.572262118491921e-05, + "loss": 0.2573, + "step": 37913 + }, + { + "epoch": 1.46, + "learning_rate": 2.572197999487048e-05, + "loss": 0.3814, + "step": 37914 + }, + { + "epoch": 1.46, + "learning_rate": 2.572133880482175e-05, + "loss": 0.4473, + "step": 37915 + }, + { + "epoch": 1.46, + "learning_rate": 2.572069761477302e-05, + "loss": 0.3378, + "step": 37916 + }, + { + "epoch": 1.46, + "learning_rate": 2.5720056424724286e-05, + "loss": 0.2621, + "step": 37917 + }, + { + "epoch": 1.46, + "learning_rate": 2.5719415234675558e-05, + "loss": 0.3788, + "step": 37918 + }, + { + "epoch": 1.46, + "learning_rate": 2.571877404462683e-05, + "loss": 0.2776, + "step": 37919 + }, + { + "epoch": 1.46, + "learning_rate": 2.5718132854578098e-05, + "loss": 0.3606, + "step": 37920 + }, + { + "epoch": 1.46, + "learning_rate": 2.571749166452937e-05, + "loss": 0.2505, + "step": 37921 + }, + { + "epoch": 1.46, + "learning_rate": 2.5716850474480634e-05, + "loss": 0.3207, + "step": 37922 + }, + { + "epoch": 1.46, + "learning_rate": 2.571620928443191e-05, + "loss": 0.3503, + "step": 37923 + }, + { + "epoch": 1.46, + "learning_rate": 2.5715568094383174e-05, + "loss": 0.5261, + "step": 37924 + }, + { + "epoch": 1.46, + "learning_rate": 2.5714926904334446e-05, + "loss": 0.3883, + "step": 37925 + }, + { + "epoch": 1.46, + "learning_rate": 2.5714285714285714e-05, + "loss": 0.2981, + "step": 37926 + }, + { + "epoch": 1.46, + "learning_rate": 2.5713644524236986e-05, + "loss": 0.4743, + "step": 37927 + }, + { + "epoch": 1.46, + "learning_rate": 2.5713003334188258e-05, + "loss": 0.3703, + "step": 37928 + }, + { + "epoch": 1.46, + "learning_rate": 2.5712362144139523e-05, + "loss": 0.3715, + "step": 37929 + }, + { + "epoch": 1.46, + "learning_rate": 2.5711720954090795e-05, + "loss": 0.223, + "step": 37930 + }, + { + "epoch": 1.46, + "learning_rate": 2.5711079764042063e-05, + "loss": 0.4483, + "step": 37931 + }, + { + "epoch": 1.46, + "learning_rate": 2.5710438573993335e-05, + "loss": 0.2859, + "step": 37932 + }, + { + "epoch": 1.46, + "learning_rate": 2.57097973839446e-05, + "loss": 0.3318, + "step": 37933 + }, + { + "epoch": 1.46, + "learning_rate": 2.570915619389587e-05, + "loss": 0.3792, + "step": 37934 + }, + { + "epoch": 1.46, + "learning_rate": 2.570851500384714e-05, + "loss": 0.3931, + "step": 37935 + }, + { + "epoch": 1.46, + "learning_rate": 2.570787381379841e-05, + "loss": 0.3128, + "step": 37936 + }, + { + "epoch": 1.46, + "learning_rate": 2.5707232623749683e-05, + "loss": 0.412, + "step": 37937 + }, + { + "epoch": 1.46, + "learning_rate": 2.5706591433700948e-05, + "loss": 0.257, + "step": 37938 + }, + { + "epoch": 1.46, + "learning_rate": 2.5705950243652223e-05, + "loss": 0.3258, + "step": 37939 + }, + { + "epoch": 1.46, + "learning_rate": 2.5705309053603488e-05, + "loss": 0.3817, + "step": 37940 + }, + { + "epoch": 1.46, + "learning_rate": 2.570466786355476e-05, + "loss": 0.4379, + "step": 37941 + }, + { + "epoch": 1.46, + "learning_rate": 2.5704026673506028e-05, + "loss": 0.3686, + "step": 37942 + }, + { + "epoch": 1.46, + "learning_rate": 2.57033854834573e-05, + "loss": 0.3855, + "step": 37943 + }, + { + "epoch": 1.46, + "learning_rate": 2.5702744293408565e-05, + "loss": 0.3388, + "step": 37944 + }, + { + "epoch": 1.46, + "learning_rate": 2.5702103103359836e-05, + "loss": 0.5078, + "step": 37945 + }, + { + "epoch": 1.46, + "learning_rate": 2.5701461913311108e-05, + "loss": 0.4484, + "step": 37946 + }, + { + "epoch": 1.46, + "learning_rate": 2.5700820723262376e-05, + "loss": 0.3487, + "step": 37947 + }, + { + "epoch": 1.46, + "learning_rate": 2.5700179533213648e-05, + "loss": 0.4131, + "step": 37948 + }, + { + "epoch": 1.46, + "learning_rate": 2.5699538343164913e-05, + "loss": 0.3559, + "step": 37949 + }, + { + "epoch": 1.46, + "learning_rate": 2.5698897153116185e-05, + "loss": 0.3464, + "step": 37950 + }, + { + "epoch": 1.46, + "learning_rate": 2.5698255963067453e-05, + "loss": 0.3482, + "step": 37951 + }, + { + "epoch": 1.46, + "learning_rate": 2.5697614773018725e-05, + "loss": 0.3016, + "step": 37952 + }, + { + "epoch": 1.46, + "learning_rate": 2.569697358296999e-05, + "loss": 0.2161, + "step": 37953 + }, + { + "epoch": 1.46, + "learning_rate": 2.569633239292126e-05, + "loss": 0.3353, + "step": 37954 + }, + { + "epoch": 1.46, + "learning_rate": 2.5695691202872537e-05, + "loss": 0.3279, + "step": 37955 + }, + { + "epoch": 1.46, + "learning_rate": 2.56950500128238e-05, + "loss": 0.2817, + "step": 37956 + }, + { + "epoch": 1.46, + "learning_rate": 2.5694408822775073e-05, + "loss": 0.4077, + "step": 37957 + }, + { + "epoch": 1.46, + "learning_rate": 2.5693767632726338e-05, + "loss": 0.3128, + "step": 37958 + }, + { + "epoch": 1.46, + "learning_rate": 2.5693126442677613e-05, + "loss": 0.3909, + "step": 37959 + }, + { + "epoch": 1.46, + "learning_rate": 2.5692485252628878e-05, + "loss": 0.3448, + "step": 37960 + }, + { + "epoch": 1.46, + "learning_rate": 2.569184406258015e-05, + "loss": 0.3496, + "step": 37961 + }, + { + "epoch": 1.46, + "learning_rate": 2.569120287253142e-05, + "loss": 0.3207, + "step": 37962 + }, + { + "epoch": 1.46, + "learning_rate": 2.569056168248269e-05, + "loss": 0.4338, + "step": 37963 + }, + { + "epoch": 1.46, + "learning_rate": 2.5689920492433962e-05, + "loss": 0.2633, + "step": 37964 + }, + { + "epoch": 1.46, + "learning_rate": 2.5689279302385227e-05, + "loss": 0.3861, + "step": 37965 + }, + { + "epoch": 1.46, + "learning_rate": 2.56886381123365e-05, + "loss": 0.3345, + "step": 37966 + }, + { + "epoch": 1.46, + "learning_rate": 2.5687996922287767e-05, + "loss": 0.29, + "step": 37967 + }, + { + "epoch": 1.46, + "learning_rate": 2.568735573223904e-05, + "loss": 0.4115, + "step": 37968 + }, + { + "epoch": 1.46, + "learning_rate": 2.5686714542190303e-05, + "loss": 0.3139, + "step": 37969 + }, + { + "epoch": 1.46, + "learning_rate": 2.5686073352141575e-05, + "loss": 0.4448, + "step": 37970 + }, + { + "epoch": 1.46, + "learning_rate": 2.5685432162092843e-05, + "loss": 0.3253, + "step": 37971 + }, + { + "epoch": 1.46, + "learning_rate": 2.5684790972044115e-05, + "loss": 0.4899, + "step": 37972 + }, + { + "epoch": 1.46, + "learning_rate": 2.5684149781995387e-05, + "loss": 0.2335, + "step": 37973 + }, + { + "epoch": 1.46, + "learning_rate": 2.5683508591946652e-05, + "loss": 0.4266, + "step": 37974 + }, + { + "epoch": 1.46, + "learning_rate": 2.5682867401897927e-05, + "loss": 0.3192, + "step": 37975 + }, + { + "epoch": 1.46, + "learning_rate": 2.5682226211849192e-05, + "loss": 0.354, + "step": 37976 + }, + { + "epoch": 1.46, + "learning_rate": 2.5681585021800464e-05, + "loss": 0.3135, + "step": 37977 + }, + { + "epoch": 1.46, + "learning_rate": 2.5680943831751732e-05, + "loss": 0.257, + "step": 37978 + }, + { + "epoch": 1.46, + "learning_rate": 2.5680302641703004e-05, + "loss": 0.4152, + "step": 37979 + }, + { + "epoch": 1.46, + "learning_rate": 2.567966145165427e-05, + "loss": 0.7295, + "step": 37980 + }, + { + "epoch": 1.46, + "learning_rate": 2.567902026160554e-05, + "loss": 0.3932, + "step": 37981 + }, + { + "epoch": 1.46, + "learning_rate": 2.5678379071556812e-05, + "loss": 0.4459, + "step": 37982 + }, + { + "epoch": 1.46, + "learning_rate": 2.567773788150808e-05, + "loss": 0.2787, + "step": 37983 + }, + { + "epoch": 1.46, + "learning_rate": 2.5677096691459352e-05, + "loss": 0.3995, + "step": 37984 + }, + { + "epoch": 1.46, + "learning_rate": 2.5676455501410617e-05, + "loss": 0.2872, + "step": 37985 + }, + { + "epoch": 1.46, + "learning_rate": 2.567581431136189e-05, + "loss": 0.3346, + "step": 37986 + }, + { + "epoch": 1.46, + "learning_rate": 2.5675173121313157e-05, + "loss": 0.3253, + "step": 37987 + }, + { + "epoch": 1.46, + "learning_rate": 2.567453193126443e-05, + "loss": 0.2743, + "step": 37988 + }, + { + "epoch": 1.46, + "learning_rate": 2.5673890741215694e-05, + "loss": 0.3232, + "step": 37989 + }, + { + "epoch": 1.46, + "learning_rate": 2.5673249551166965e-05, + "loss": 0.3755, + "step": 37990 + }, + { + "epoch": 1.46, + "learning_rate": 2.567260836111824e-05, + "loss": 0.3898, + "step": 37991 + }, + { + "epoch": 1.46, + "learning_rate": 2.5671967171069505e-05, + "loss": 0.4383, + "step": 37992 + }, + { + "epoch": 1.46, + "learning_rate": 2.5671325981020777e-05, + "loss": 0.3417, + "step": 37993 + }, + { + "epoch": 1.46, + "learning_rate": 2.5670684790972045e-05, + "loss": 0.3556, + "step": 37994 + }, + { + "epoch": 1.46, + "learning_rate": 2.5670043600923317e-05, + "loss": 0.3607, + "step": 37995 + }, + { + "epoch": 1.46, + "learning_rate": 2.5669402410874582e-05, + "loss": 0.3567, + "step": 37996 + }, + { + "epoch": 1.46, + "learning_rate": 2.5668761220825854e-05, + "loss": 0.2838, + "step": 37997 + }, + { + "epoch": 1.46, + "learning_rate": 2.5668120030777122e-05, + "loss": 0.3943, + "step": 37998 + }, + { + "epoch": 1.46, + "learning_rate": 2.5667478840728394e-05, + "loss": 0.2881, + "step": 37999 + }, + { + "epoch": 1.46, + "learning_rate": 2.5666837650679666e-05, + "loss": 0.3676, + "step": 38000 + }, + { + "epoch": 1.46, + "learning_rate": 2.566619646063093e-05, + "loss": 0.2924, + "step": 38001 + }, + { + "epoch": 1.46, + "learning_rate": 2.5665555270582202e-05, + "loss": 0.5581, + "step": 38002 + }, + { + "epoch": 1.46, + "learning_rate": 2.566491408053347e-05, + "loss": 0.3443, + "step": 38003 + }, + { + "epoch": 1.46, + "learning_rate": 2.5664272890484742e-05, + "loss": 0.344, + "step": 38004 + }, + { + "epoch": 1.46, + "learning_rate": 2.5663631700436007e-05, + "loss": 0.3885, + "step": 38005 + }, + { + "epoch": 1.46, + "learning_rate": 2.566299051038728e-05, + "loss": 0.329, + "step": 38006 + }, + { + "epoch": 1.46, + "learning_rate": 2.566234932033855e-05, + "loss": 0.4147, + "step": 38007 + }, + { + "epoch": 1.46, + "learning_rate": 2.566170813028982e-05, + "loss": 0.2558, + "step": 38008 + }, + { + "epoch": 1.46, + "learning_rate": 2.566106694024109e-05, + "loss": 0.4584, + "step": 38009 + }, + { + "epoch": 1.46, + "learning_rate": 2.5660425750192356e-05, + "loss": 0.357, + "step": 38010 + }, + { + "epoch": 1.46, + "learning_rate": 2.565978456014363e-05, + "loss": 0.3475, + "step": 38011 + }, + { + "epoch": 1.46, + "learning_rate": 2.5659143370094896e-05, + "loss": 0.3873, + "step": 38012 + }, + { + "epoch": 1.46, + "learning_rate": 2.5658502180046167e-05, + "loss": 0.4378, + "step": 38013 + }, + { + "epoch": 1.46, + "learning_rate": 2.5657860989997436e-05, + "loss": 0.3074, + "step": 38014 + }, + { + "epoch": 1.46, + "learning_rate": 2.5657219799948707e-05, + "loss": 0.3373, + "step": 38015 + }, + { + "epoch": 1.46, + "learning_rate": 2.565657860989998e-05, + "loss": 0.4655, + "step": 38016 + }, + { + "epoch": 1.46, + "learning_rate": 2.5655937419851244e-05, + "loss": 0.2815, + "step": 38017 + }, + { + "epoch": 1.46, + "learning_rate": 2.5655296229802516e-05, + "loss": 0.441, + "step": 38018 + }, + { + "epoch": 1.46, + "learning_rate": 2.5654655039753784e-05, + "loss": 0.4597, + "step": 38019 + }, + { + "epoch": 1.46, + "learning_rate": 2.5654013849705056e-05, + "loss": 0.4711, + "step": 38020 + }, + { + "epoch": 1.46, + "learning_rate": 2.565337265965632e-05, + "loss": 0.3226, + "step": 38021 + }, + { + "epoch": 1.46, + "learning_rate": 2.5652731469607593e-05, + "loss": 0.4103, + "step": 38022 + }, + { + "epoch": 1.46, + "learning_rate": 2.565209027955886e-05, + "loss": 0.3207, + "step": 38023 + }, + { + "epoch": 1.46, + "learning_rate": 2.5651449089510133e-05, + "loss": 0.3001, + "step": 38024 + }, + { + "epoch": 1.46, + "learning_rate": 2.5650807899461404e-05, + "loss": 0.3616, + "step": 38025 + }, + { + "epoch": 1.46, + "learning_rate": 2.565016670941267e-05, + "loss": 0.506, + "step": 38026 + }, + { + "epoch": 1.46, + "learning_rate": 2.5649525519363944e-05, + "loss": 0.3999, + "step": 38027 + }, + { + "epoch": 1.46, + "learning_rate": 2.564888432931521e-05, + "loss": 0.339, + "step": 38028 + }, + { + "epoch": 1.46, + "learning_rate": 2.564824313926648e-05, + "loss": 0.3569, + "step": 38029 + }, + { + "epoch": 1.46, + "learning_rate": 2.564760194921775e-05, + "loss": 0.2357, + "step": 38030 + }, + { + "epoch": 1.46, + "learning_rate": 2.564696075916902e-05, + "loss": 0.3304, + "step": 38031 + }, + { + "epoch": 1.46, + "learning_rate": 2.5646319569120286e-05, + "loss": 0.3513, + "step": 38032 + }, + { + "epoch": 1.46, + "learning_rate": 2.5645678379071558e-05, + "loss": 0.3583, + "step": 38033 + }, + { + "epoch": 1.46, + "learning_rate": 2.564503718902283e-05, + "loss": 0.5016, + "step": 38034 + }, + { + "epoch": 1.46, + "learning_rate": 2.5644395998974098e-05, + "loss": 0.3474, + "step": 38035 + }, + { + "epoch": 1.46, + "learning_rate": 2.564375480892537e-05, + "loss": 0.3669, + "step": 38036 + }, + { + "epoch": 1.46, + "learning_rate": 2.5643113618876634e-05, + "loss": 0.4362, + "step": 38037 + }, + { + "epoch": 1.46, + "learning_rate": 2.5642472428827906e-05, + "loss": 0.343, + "step": 38038 + }, + { + "epoch": 1.46, + "learning_rate": 2.5641831238779174e-05, + "loss": 0.3289, + "step": 38039 + }, + { + "epoch": 1.46, + "learning_rate": 2.5641190048730446e-05, + "loss": 0.3438, + "step": 38040 + }, + { + "epoch": 1.46, + "learning_rate": 2.564054885868171e-05, + "loss": 0.2692, + "step": 38041 + }, + { + "epoch": 1.46, + "learning_rate": 2.5639907668632983e-05, + "loss": 0.4309, + "step": 38042 + }, + { + "epoch": 1.46, + "learning_rate": 2.5639266478584255e-05, + "loss": 0.3424, + "step": 38043 + }, + { + "epoch": 1.46, + "learning_rate": 2.5638625288535523e-05, + "loss": 0.347, + "step": 38044 + }, + { + "epoch": 1.46, + "learning_rate": 2.5637984098486795e-05, + "loss": 0.3817, + "step": 38045 + }, + { + "epoch": 1.46, + "learning_rate": 2.563734290843806e-05, + "loss": 0.435, + "step": 38046 + }, + { + "epoch": 1.46, + "learning_rate": 2.5636701718389335e-05, + "loss": 0.3852, + "step": 38047 + }, + { + "epoch": 1.46, + "learning_rate": 2.56360605283406e-05, + "loss": 0.2924, + "step": 38048 + }, + { + "epoch": 1.46, + "learning_rate": 2.563541933829187e-05, + "loss": 0.3171, + "step": 38049 + }, + { + "epoch": 1.46, + "learning_rate": 2.563477814824314e-05, + "loss": 0.4548, + "step": 38050 + }, + { + "epoch": 1.46, + "learning_rate": 2.563413695819441e-05, + "loss": 0.3407, + "step": 38051 + }, + { + "epoch": 1.46, + "learning_rate": 2.5633495768145683e-05, + "loss": 0.3592, + "step": 38052 + }, + { + "epoch": 1.46, + "learning_rate": 2.5632854578096948e-05, + "loss": 0.3312, + "step": 38053 + }, + { + "epoch": 1.46, + "learning_rate": 2.563221338804822e-05, + "loss": 0.2672, + "step": 38054 + }, + { + "epoch": 1.46, + "learning_rate": 2.5631572197999488e-05, + "loss": 0.3133, + "step": 38055 + }, + { + "epoch": 1.46, + "learning_rate": 2.563093100795076e-05, + "loss": 0.4139, + "step": 38056 + }, + { + "epoch": 1.46, + "learning_rate": 2.5630289817902025e-05, + "loss": 0.3212, + "step": 38057 + }, + { + "epoch": 1.46, + "learning_rate": 2.5629648627853296e-05, + "loss": 0.2899, + "step": 38058 + }, + { + "epoch": 1.46, + "learning_rate": 2.5629007437804565e-05, + "loss": 0.4468, + "step": 38059 + }, + { + "epoch": 1.46, + "learning_rate": 2.5628366247755836e-05, + "loss": 0.4113, + "step": 38060 + }, + { + "epoch": 1.46, + "learning_rate": 2.5627725057707108e-05, + "loss": 0.39, + "step": 38061 + }, + { + "epoch": 1.46, + "learning_rate": 2.5627083867658373e-05, + "loss": 0.5029, + "step": 38062 + }, + { + "epoch": 1.46, + "learning_rate": 2.5626442677609648e-05, + "loss": 0.4158, + "step": 38063 + }, + { + "epoch": 1.46, + "learning_rate": 2.5625801487560913e-05, + "loss": 0.5031, + "step": 38064 + }, + { + "epoch": 1.46, + "learning_rate": 2.5625160297512185e-05, + "loss": 0.4509, + "step": 38065 + }, + { + "epoch": 1.46, + "learning_rate": 2.5624519107463453e-05, + "loss": 0.367, + "step": 38066 + }, + { + "epoch": 1.46, + "learning_rate": 2.5623877917414725e-05, + "loss": 0.2547, + "step": 38067 + }, + { + "epoch": 1.46, + "learning_rate": 2.562323672736599e-05, + "loss": 0.494, + "step": 38068 + }, + { + "epoch": 1.46, + "learning_rate": 2.562259553731726e-05, + "loss": 0.2705, + "step": 38069 + }, + { + "epoch": 1.46, + "learning_rate": 2.5621954347268533e-05, + "loss": 0.3645, + "step": 38070 + }, + { + "epoch": 1.46, + "learning_rate": 2.56213131572198e-05, + "loss": 0.4185, + "step": 38071 + }, + { + "epoch": 1.46, + "learning_rate": 2.5620671967171073e-05, + "loss": 0.2412, + "step": 38072 + }, + { + "epoch": 1.46, + "learning_rate": 2.562003077712234e-05, + "loss": 0.301, + "step": 38073 + }, + { + "epoch": 1.46, + "learning_rate": 2.561938958707361e-05, + "loss": 0.3519, + "step": 38074 + }, + { + "epoch": 1.46, + "learning_rate": 2.561874839702488e-05, + "loss": 0.5282, + "step": 38075 + }, + { + "epoch": 1.46, + "learning_rate": 2.561810720697615e-05, + "loss": 0.4097, + "step": 38076 + }, + { + "epoch": 1.46, + "learning_rate": 2.5617466016927415e-05, + "loss": 0.3195, + "step": 38077 + }, + { + "epoch": 1.46, + "learning_rate": 2.5616824826878687e-05, + "loss": 0.4061, + "step": 38078 + }, + { + "epoch": 1.46, + "learning_rate": 2.561618363682996e-05, + "loss": 0.381, + "step": 38079 + }, + { + "epoch": 1.46, + "learning_rate": 2.5615542446781227e-05, + "loss": 0.465, + "step": 38080 + }, + { + "epoch": 1.46, + "learning_rate": 2.56149012567325e-05, + "loss": 0.3057, + "step": 38081 + }, + { + "epoch": 1.46, + "learning_rate": 2.5614260066683763e-05, + "loss": 0.3889, + "step": 38082 + }, + { + "epoch": 1.46, + "learning_rate": 2.561361887663504e-05, + "loss": 0.4032, + "step": 38083 + }, + { + "epoch": 1.46, + "learning_rate": 2.5612977686586303e-05, + "loss": 0.3033, + "step": 38084 + }, + { + "epoch": 1.46, + "learning_rate": 2.5612336496537575e-05, + "loss": 0.3221, + "step": 38085 + }, + { + "epoch": 1.46, + "learning_rate": 2.5611695306488844e-05, + "loss": 0.2931, + "step": 38086 + }, + { + "epoch": 1.46, + "learning_rate": 2.5611054116440115e-05, + "loss": 0.3489, + "step": 38087 + }, + { + "epoch": 1.46, + "learning_rate": 2.5610412926391387e-05, + "loss": 0.3195, + "step": 38088 + }, + { + "epoch": 1.46, + "learning_rate": 2.5609771736342652e-05, + "loss": 0.4701, + "step": 38089 + }, + { + "epoch": 1.46, + "learning_rate": 2.5609130546293924e-05, + "loss": 0.4789, + "step": 38090 + }, + { + "epoch": 1.46, + "learning_rate": 2.5608489356245192e-05, + "loss": 0.3591, + "step": 38091 + }, + { + "epoch": 1.46, + "learning_rate": 2.5607848166196464e-05, + "loss": 0.3448, + "step": 38092 + }, + { + "epoch": 1.46, + "learning_rate": 2.560720697614773e-05, + "loss": 0.3408, + "step": 38093 + }, + { + "epoch": 1.46, + "learning_rate": 2.5606565786099e-05, + "loss": 0.3329, + "step": 38094 + }, + { + "epoch": 1.46, + "learning_rate": 2.560592459605027e-05, + "loss": 0.3568, + "step": 38095 + }, + { + "epoch": 1.46, + "learning_rate": 2.560528340600154e-05, + "loss": 0.2924, + "step": 38096 + }, + { + "epoch": 1.46, + "learning_rate": 2.5604642215952812e-05, + "loss": 0.2147, + "step": 38097 + }, + { + "epoch": 1.46, + "learning_rate": 2.5604001025904077e-05, + "loss": 0.3591, + "step": 38098 + }, + { + "epoch": 1.46, + "learning_rate": 2.5603359835855352e-05, + "loss": 0.4894, + "step": 38099 + }, + { + "epoch": 1.46, + "learning_rate": 2.5602718645806617e-05, + "loss": 0.3296, + "step": 38100 + }, + { + "epoch": 1.46, + "learning_rate": 2.560207745575789e-05, + "loss": 0.299, + "step": 38101 + }, + { + "epoch": 1.46, + "learning_rate": 2.5601436265709157e-05, + "loss": 0.3713, + "step": 38102 + }, + { + "epoch": 1.46, + "learning_rate": 2.560079507566043e-05, + "loss": 0.3435, + "step": 38103 + }, + { + "epoch": 1.46, + "learning_rate": 2.5600153885611694e-05, + "loss": 0.3007, + "step": 38104 + }, + { + "epoch": 1.47, + "learning_rate": 2.5599512695562965e-05, + "loss": 0.4276, + "step": 38105 + }, + { + "epoch": 1.47, + "learning_rate": 2.5598871505514237e-05, + "loss": 0.342, + "step": 38106 + }, + { + "epoch": 1.47, + "learning_rate": 2.5598230315465506e-05, + "loss": 0.2729, + "step": 38107 + }, + { + "epoch": 1.47, + "learning_rate": 2.5597589125416777e-05, + "loss": 0.4296, + "step": 38108 + }, + { + "epoch": 1.47, + "learning_rate": 2.5596947935368042e-05, + "loss": 0.3432, + "step": 38109 + }, + { + "epoch": 1.47, + "learning_rate": 2.5596306745319314e-05, + "loss": 0.3091, + "step": 38110 + }, + { + "epoch": 1.47, + "learning_rate": 2.5595665555270582e-05, + "loss": 0.3064, + "step": 38111 + }, + { + "epoch": 1.47, + "learning_rate": 2.5595024365221854e-05, + "loss": 0.3397, + "step": 38112 + }, + { + "epoch": 1.47, + "learning_rate": 2.559438317517312e-05, + "loss": 0.5128, + "step": 38113 + }, + { + "epoch": 1.47, + "learning_rate": 2.559374198512439e-05, + "loss": 0.459, + "step": 38114 + }, + { + "epoch": 1.47, + "learning_rate": 2.5593100795075666e-05, + "loss": 0.3237, + "step": 38115 + }, + { + "epoch": 1.47, + "learning_rate": 2.559245960502693e-05, + "loss": 0.4493, + "step": 38116 + }, + { + "epoch": 1.47, + "learning_rate": 2.5591818414978202e-05, + "loss": 0.5514, + "step": 38117 + }, + { + "epoch": 1.47, + "learning_rate": 2.559117722492947e-05, + "loss": 0.4466, + "step": 38118 + }, + { + "epoch": 1.47, + "learning_rate": 2.5590536034880742e-05, + "loss": 0.4064, + "step": 38119 + }, + { + "epoch": 1.47, + "learning_rate": 2.5589894844832007e-05, + "loss": 0.3193, + "step": 38120 + }, + { + "epoch": 1.47, + "learning_rate": 2.558925365478328e-05, + "loss": 0.4569, + "step": 38121 + }, + { + "epoch": 1.47, + "learning_rate": 2.5588612464734547e-05, + "loss": 0.4516, + "step": 38122 + }, + { + "epoch": 1.47, + "learning_rate": 2.558797127468582e-05, + "loss": 0.387, + "step": 38123 + }, + { + "epoch": 1.47, + "learning_rate": 2.558733008463709e-05, + "loss": 0.3108, + "step": 38124 + }, + { + "epoch": 1.47, + "learning_rate": 2.5586688894588356e-05, + "loss": 0.257, + "step": 38125 + }, + { + "epoch": 1.47, + "learning_rate": 2.5586047704539627e-05, + "loss": 0.3163, + "step": 38126 + }, + { + "epoch": 1.47, + "learning_rate": 2.5585406514490896e-05, + "loss": 0.3987, + "step": 38127 + }, + { + "epoch": 1.47, + "learning_rate": 2.5584765324442168e-05, + "loss": 0.5427, + "step": 38128 + }, + { + "epoch": 1.47, + "learning_rate": 2.5584124134393432e-05, + "loss": 0.3266, + "step": 38129 + }, + { + "epoch": 1.47, + "learning_rate": 2.5583482944344704e-05, + "loss": 0.457, + "step": 38130 + }, + { + "epoch": 1.47, + "learning_rate": 2.5582841754295973e-05, + "loss": 0.3935, + "step": 38131 + }, + { + "epoch": 1.47, + "learning_rate": 2.5582200564247244e-05, + "loss": 0.3818, + "step": 38132 + }, + { + "epoch": 1.47, + "learning_rate": 2.5581559374198516e-05, + "loss": 0.3026, + "step": 38133 + }, + { + "epoch": 1.47, + "learning_rate": 2.558091818414978e-05, + "loss": 0.2562, + "step": 38134 + }, + { + "epoch": 1.47, + "learning_rate": 2.5580276994101056e-05, + "loss": 0.4167, + "step": 38135 + }, + { + "epoch": 1.47, + "learning_rate": 2.557963580405232e-05, + "loss": 0.3999, + "step": 38136 + }, + { + "epoch": 1.47, + "learning_rate": 2.5578994614003593e-05, + "loss": 0.5261, + "step": 38137 + }, + { + "epoch": 1.47, + "learning_rate": 2.557835342395486e-05, + "loss": 0.435, + "step": 38138 + }, + { + "epoch": 1.47, + "learning_rate": 2.5577712233906133e-05, + "loss": 0.4132, + "step": 38139 + }, + { + "epoch": 1.47, + "learning_rate": 2.5577071043857398e-05, + "loss": 0.3651, + "step": 38140 + }, + { + "epoch": 1.47, + "learning_rate": 2.557642985380867e-05, + "loss": 0.3714, + "step": 38141 + }, + { + "epoch": 1.47, + "learning_rate": 2.557578866375994e-05, + "loss": 0.3307, + "step": 38142 + }, + { + "epoch": 1.47, + "learning_rate": 2.557514747371121e-05, + "loss": 0.3062, + "step": 38143 + }, + { + "epoch": 1.47, + "learning_rate": 2.557450628366248e-05, + "loss": 0.3766, + "step": 38144 + }, + { + "epoch": 1.47, + "learning_rate": 2.5573865093613746e-05, + "loss": 0.2498, + "step": 38145 + }, + { + "epoch": 1.47, + "learning_rate": 2.5573223903565018e-05, + "loss": 0.4602, + "step": 38146 + }, + { + "epoch": 1.47, + "learning_rate": 2.5572582713516286e-05, + "loss": 0.3347, + "step": 38147 + }, + { + "epoch": 1.47, + "learning_rate": 2.5571941523467558e-05, + "loss": 0.3321, + "step": 38148 + }, + { + "epoch": 1.47, + "learning_rate": 2.5571300333418823e-05, + "loss": 0.2914, + "step": 38149 + }, + { + "epoch": 1.47, + "learning_rate": 2.5570659143370094e-05, + "loss": 0.3514, + "step": 38150 + }, + { + "epoch": 1.47, + "learning_rate": 2.557001795332137e-05, + "loss": 0.407, + "step": 38151 + }, + { + "epoch": 1.47, + "learning_rate": 2.5569376763272635e-05, + "loss": 0.2569, + "step": 38152 + }, + { + "epoch": 1.47, + "learning_rate": 2.5568735573223906e-05, + "loss": 0.326, + "step": 38153 + }, + { + "epoch": 1.47, + "learning_rate": 2.5568094383175175e-05, + "loss": 0.2863, + "step": 38154 + }, + { + "epoch": 1.47, + "learning_rate": 2.5567453193126446e-05, + "loss": 0.3617, + "step": 38155 + }, + { + "epoch": 1.47, + "learning_rate": 2.556681200307771e-05, + "loss": 0.4205, + "step": 38156 + }, + { + "epoch": 1.47, + "learning_rate": 2.5566170813028983e-05, + "loss": 0.3626, + "step": 38157 + }, + { + "epoch": 1.47, + "learning_rate": 2.556552962298025e-05, + "loss": 0.4732, + "step": 38158 + }, + { + "epoch": 1.47, + "learning_rate": 2.5564888432931523e-05, + "loss": 0.4055, + "step": 38159 + }, + { + "epoch": 1.47, + "learning_rate": 2.5564247242882795e-05, + "loss": 0.3331, + "step": 38160 + }, + { + "epoch": 1.47, + "learning_rate": 2.556360605283406e-05, + "loss": 0.3773, + "step": 38161 + }, + { + "epoch": 1.47, + "learning_rate": 2.556296486278533e-05, + "loss": 0.3114, + "step": 38162 + }, + { + "epoch": 1.47, + "learning_rate": 2.55623236727366e-05, + "loss": 0.5208, + "step": 38163 + }, + { + "epoch": 1.47, + "learning_rate": 2.556168248268787e-05, + "loss": 0.2929, + "step": 38164 + }, + { + "epoch": 1.47, + "learning_rate": 2.5561041292639136e-05, + "loss": 0.4217, + "step": 38165 + }, + { + "epoch": 1.47, + "learning_rate": 2.5560400102590408e-05, + "loss": 0.3192, + "step": 38166 + }, + { + "epoch": 1.47, + "learning_rate": 2.5559758912541676e-05, + "loss": 0.3558, + "step": 38167 + }, + { + "epoch": 1.47, + "learning_rate": 2.5559117722492948e-05, + "loss": 0.2961, + "step": 38168 + }, + { + "epoch": 1.47, + "learning_rate": 2.555847653244422e-05, + "loss": 0.2821, + "step": 38169 + }, + { + "epoch": 1.47, + "learning_rate": 2.5557835342395485e-05, + "loss": 0.4094, + "step": 38170 + }, + { + "epoch": 1.47, + "learning_rate": 2.555719415234676e-05, + "loss": 0.2681, + "step": 38171 + }, + { + "epoch": 1.47, + "learning_rate": 2.5556552962298025e-05, + "loss": 0.4171, + "step": 38172 + }, + { + "epoch": 1.47, + "learning_rate": 2.5555911772249297e-05, + "loss": 0.3685, + "step": 38173 + }, + { + "epoch": 1.47, + "learning_rate": 2.5555270582200565e-05, + "loss": 0.3977, + "step": 38174 + }, + { + "epoch": 1.47, + "learning_rate": 2.5554629392151837e-05, + "loss": 0.4174, + "step": 38175 + }, + { + "epoch": 1.47, + "learning_rate": 2.55539882021031e-05, + "loss": 0.4554, + "step": 38176 + }, + { + "epoch": 1.47, + "learning_rate": 2.5553347012054373e-05, + "loss": 0.4441, + "step": 38177 + }, + { + "epoch": 1.47, + "learning_rate": 2.5552705822005645e-05, + "loss": 0.4042, + "step": 38178 + }, + { + "epoch": 1.47, + "learning_rate": 2.5552064631956913e-05, + "loss": 0.2051, + "step": 38179 + }, + { + "epoch": 1.47, + "learning_rate": 2.5551423441908185e-05, + "loss": 0.2834, + "step": 38180 + }, + { + "epoch": 1.47, + "learning_rate": 2.555078225185945e-05, + "loss": 0.334, + "step": 38181 + }, + { + "epoch": 1.47, + "learning_rate": 2.555014106181072e-05, + "loss": 0.3772, + "step": 38182 + }, + { + "epoch": 1.47, + "learning_rate": 2.554949987176199e-05, + "loss": 0.4293, + "step": 38183 + }, + { + "epoch": 1.47, + "learning_rate": 2.554885868171326e-05, + "loss": 0.3419, + "step": 38184 + }, + { + "epoch": 1.47, + "learning_rate": 2.5548217491664527e-05, + "loss": 0.3171, + "step": 38185 + }, + { + "epoch": 1.47, + "learning_rate": 2.55475763016158e-05, + "loss": 0.4242, + "step": 38186 + }, + { + "epoch": 1.47, + "learning_rate": 2.5546935111567073e-05, + "loss": 0.382, + "step": 38187 + }, + { + "epoch": 1.47, + "learning_rate": 2.554629392151834e-05, + "loss": 0.3693, + "step": 38188 + }, + { + "epoch": 1.47, + "learning_rate": 2.554565273146961e-05, + "loss": 0.4492, + "step": 38189 + }, + { + "epoch": 1.47, + "learning_rate": 2.554501154142088e-05, + "loss": 0.4314, + "step": 38190 + }, + { + "epoch": 1.47, + "learning_rate": 2.554437035137215e-05, + "loss": 0.4311, + "step": 38191 + }, + { + "epoch": 1.47, + "learning_rate": 2.5543729161323415e-05, + "loss": 0.3195, + "step": 38192 + }, + { + "epoch": 1.47, + "learning_rate": 2.5543087971274687e-05, + "loss": 0.3472, + "step": 38193 + }, + { + "epoch": 1.47, + "learning_rate": 2.554244678122596e-05, + "loss": 0.351, + "step": 38194 + }, + { + "epoch": 1.47, + "learning_rate": 2.5541805591177227e-05, + "loss": 0.3102, + "step": 38195 + }, + { + "epoch": 1.47, + "learning_rate": 2.55411644011285e-05, + "loss": 0.4027, + "step": 38196 + }, + { + "epoch": 1.47, + "learning_rate": 2.5540523211079764e-05, + "loss": 0.4078, + "step": 38197 + }, + { + "epoch": 1.47, + "learning_rate": 2.5539882021031035e-05, + "loss": 0.335, + "step": 38198 + }, + { + "epoch": 1.47, + "learning_rate": 2.5539240830982304e-05, + "loss": 0.4014, + "step": 38199 + }, + { + "epoch": 1.47, + "learning_rate": 2.5538599640933575e-05, + "loss": 0.3566, + "step": 38200 + }, + { + "epoch": 1.47, + "learning_rate": 2.553795845088484e-05, + "loss": 0.3535, + "step": 38201 + }, + { + "epoch": 1.47, + "learning_rate": 2.5537317260836112e-05, + "loss": 0.3566, + "step": 38202 + }, + { + "epoch": 1.47, + "learning_rate": 2.5536676070787384e-05, + "loss": 0.3667, + "step": 38203 + }, + { + "epoch": 1.47, + "learning_rate": 2.5536034880738652e-05, + "loss": 0.2956, + "step": 38204 + }, + { + "epoch": 1.47, + "learning_rate": 2.5535393690689924e-05, + "loss": 0.4232, + "step": 38205 + }, + { + "epoch": 1.47, + "learning_rate": 2.553475250064119e-05, + "loss": 0.3767, + "step": 38206 + }, + { + "epoch": 1.47, + "learning_rate": 2.5534111310592464e-05, + "loss": 0.3998, + "step": 38207 + }, + { + "epoch": 1.47, + "learning_rate": 2.553347012054373e-05, + "loss": 0.2738, + "step": 38208 + }, + { + "epoch": 1.47, + "learning_rate": 2.5532828930495e-05, + "loss": 0.4547, + "step": 38209 + }, + { + "epoch": 1.47, + "learning_rate": 2.553218774044627e-05, + "loss": 0.3754, + "step": 38210 + }, + { + "epoch": 1.47, + "learning_rate": 2.553154655039754e-05, + "loss": 0.3941, + "step": 38211 + }, + { + "epoch": 1.47, + "learning_rate": 2.5530905360348812e-05, + "loss": 0.4393, + "step": 38212 + }, + { + "epoch": 1.47, + "learning_rate": 2.5530264170300077e-05, + "loss": 0.2896, + "step": 38213 + }, + { + "epoch": 1.47, + "learning_rate": 2.552962298025135e-05, + "loss": 0.3409, + "step": 38214 + }, + { + "epoch": 1.47, + "learning_rate": 2.5528981790202617e-05, + "loss": 0.3114, + "step": 38215 + }, + { + "epoch": 1.47, + "learning_rate": 2.552834060015389e-05, + "loss": 0.3338, + "step": 38216 + }, + { + "epoch": 1.47, + "learning_rate": 2.5527699410105154e-05, + "loss": 0.3727, + "step": 38217 + }, + { + "epoch": 1.47, + "learning_rate": 2.5527058220056426e-05, + "loss": 0.3332, + "step": 38218 + }, + { + "epoch": 1.47, + "learning_rate": 2.5526417030007694e-05, + "loss": 0.3271, + "step": 38219 + }, + { + "epoch": 1.47, + "learning_rate": 2.5525775839958966e-05, + "loss": 0.3538, + "step": 38220 + }, + { + "epoch": 1.47, + "learning_rate": 2.5525134649910237e-05, + "loss": 0.3532, + "step": 38221 + }, + { + "epoch": 1.47, + "learning_rate": 2.5524493459861502e-05, + "loss": 0.4141, + "step": 38222 + }, + { + "epoch": 1.47, + "learning_rate": 2.5523852269812777e-05, + "loss": 0.3709, + "step": 38223 + }, + { + "epoch": 1.47, + "learning_rate": 2.5523211079764042e-05, + "loss": 0.3052, + "step": 38224 + }, + { + "epoch": 1.47, + "learning_rate": 2.5522569889715314e-05, + "loss": 0.2927, + "step": 38225 + }, + { + "epoch": 1.47, + "learning_rate": 2.5521928699666582e-05, + "loss": 0.4069, + "step": 38226 + }, + { + "epoch": 1.47, + "learning_rate": 2.5521287509617854e-05, + "loss": 0.3102, + "step": 38227 + }, + { + "epoch": 1.47, + "learning_rate": 2.552064631956912e-05, + "loss": 0.3846, + "step": 38228 + }, + { + "epoch": 1.47, + "learning_rate": 2.552000512952039e-05, + "loss": 0.3109, + "step": 38229 + }, + { + "epoch": 1.47, + "learning_rate": 2.5519363939471662e-05, + "loss": 0.3373, + "step": 38230 + }, + { + "epoch": 1.47, + "learning_rate": 2.551872274942293e-05, + "loss": 0.3807, + "step": 38231 + }, + { + "epoch": 1.47, + "learning_rate": 2.5518081559374202e-05, + "loss": 0.3489, + "step": 38232 + }, + { + "epoch": 1.47, + "learning_rate": 2.5517440369325467e-05, + "loss": 0.2961, + "step": 38233 + }, + { + "epoch": 1.47, + "learning_rate": 2.551679917927674e-05, + "loss": 0.3015, + "step": 38234 + }, + { + "epoch": 1.47, + "learning_rate": 2.5516157989228007e-05, + "loss": 0.581, + "step": 38235 + }, + { + "epoch": 1.47, + "learning_rate": 2.551551679917928e-05, + "loss": 0.2677, + "step": 38236 + }, + { + "epoch": 1.47, + "learning_rate": 2.5514875609130544e-05, + "loss": 0.5222, + "step": 38237 + }, + { + "epoch": 1.47, + "learning_rate": 2.5514234419081816e-05, + "loss": 0.2694, + "step": 38238 + }, + { + "epoch": 1.47, + "learning_rate": 2.551359322903309e-05, + "loss": 0.4052, + "step": 38239 + }, + { + "epoch": 1.47, + "learning_rate": 2.5512952038984356e-05, + "loss": 0.2665, + "step": 38240 + }, + { + "epoch": 1.47, + "learning_rate": 2.5512310848935628e-05, + "loss": 0.3688, + "step": 38241 + }, + { + "epoch": 1.47, + "learning_rate": 2.5511669658886896e-05, + "loss": 0.2775, + "step": 38242 + }, + { + "epoch": 1.47, + "learning_rate": 2.5511028468838168e-05, + "loss": 0.4212, + "step": 38243 + }, + { + "epoch": 1.47, + "learning_rate": 2.5510387278789433e-05, + "loss": 0.5762, + "step": 38244 + }, + { + "epoch": 1.47, + "learning_rate": 2.5509746088740704e-05, + "loss": 0.3517, + "step": 38245 + }, + { + "epoch": 1.47, + "learning_rate": 2.5509104898691973e-05, + "loss": 0.3579, + "step": 38246 + }, + { + "epoch": 1.47, + "learning_rate": 2.5508463708643244e-05, + "loss": 0.3313, + "step": 38247 + }, + { + "epoch": 1.47, + "learning_rate": 2.5507822518594516e-05, + "loss": 0.4133, + "step": 38248 + }, + { + "epoch": 1.47, + "learning_rate": 2.550718132854578e-05, + "loss": 0.4591, + "step": 38249 + }, + { + "epoch": 1.47, + "learning_rate": 2.5506540138497053e-05, + "loss": 0.2885, + "step": 38250 + }, + { + "epoch": 1.47, + "learning_rate": 2.550589894844832e-05, + "loss": 0.2459, + "step": 38251 + }, + { + "epoch": 1.47, + "learning_rate": 2.5505257758399593e-05, + "loss": 0.363, + "step": 38252 + }, + { + "epoch": 1.47, + "learning_rate": 2.5504616568350858e-05, + "loss": 0.3772, + "step": 38253 + }, + { + "epoch": 1.47, + "learning_rate": 2.550397537830213e-05, + "loss": 0.3595, + "step": 38254 + }, + { + "epoch": 1.47, + "learning_rate": 2.5503334188253398e-05, + "loss": 0.3458, + "step": 38255 + }, + { + "epoch": 1.47, + "learning_rate": 2.550269299820467e-05, + "loss": 0.3724, + "step": 38256 + }, + { + "epoch": 1.47, + "learning_rate": 2.550205180815594e-05, + "loss": 0.2503, + "step": 38257 + }, + { + "epoch": 1.47, + "learning_rate": 2.5501410618107206e-05, + "loss": 0.2657, + "step": 38258 + }, + { + "epoch": 1.47, + "learning_rate": 2.550076942805848e-05, + "loss": 0.4034, + "step": 38259 + }, + { + "epoch": 1.47, + "learning_rate": 2.5500128238009746e-05, + "loss": 0.3829, + "step": 38260 + }, + { + "epoch": 1.47, + "learning_rate": 2.5499487047961018e-05, + "loss": 0.4111, + "step": 38261 + }, + { + "epoch": 1.47, + "learning_rate": 2.5498845857912286e-05, + "loss": 0.4334, + "step": 38262 + }, + { + "epoch": 1.47, + "learning_rate": 2.5498204667863558e-05, + "loss": 0.3543, + "step": 38263 + }, + { + "epoch": 1.47, + "learning_rate": 2.5497563477814823e-05, + "loss": 0.3222, + "step": 38264 + }, + { + "epoch": 1.47, + "learning_rate": 2.5496922287766095e-05, + "loss": 0.3531, + "step": 38265 + }, + { + "epoch": 1.47, + "learning_rate": 2.5496281097717366e-05, + "loss": 0.369, + "step": 38266 + }, + { + "epoch": 1.47, + "learning_rate": 2.5495639907668635e-05, + "loss": 0.3995, + "step": 38267 + }, + { + "epoch": 1.47, + "learning_rate": 2.5494998717619906e-05, + "loss": 0.2957, + "step": 38268 + }, + { + "epoch": 1.47, + "learning_rate": 2.549435752757117e-05, + "loss": 0.2717, + "step": 38269 + }, + { + "epoch": 1.47, + "learning_rate": 2.5493716337522443e-05, + "loss": 0.3763, + "step": 38270 + }, + { + "epoch": 1.47, + "learning_rate": 2.549307514747371e-05, + "loss": 0.4348, + "step": 38271 + }, + { + "epoch": 1.47, + "learning_rate": 2.5492433957424983e-05, + "loss": 0.3318, + "step": 38272 + }, + { + "epoch": 1.47, + "learning_rate": 2.5491792767376248e-05, + "loss": 0.4222, + "step": 38273 + }, + { + "epoch": 1.47, + "learning_rate": 2.549115157732752e-05, + "loss": 0.4373, + "step": 38274 + }, + { + "epoch": 1.47, + "learning_rate": 2.5490510387278795e-05, + "loss": 0.3665, + "step": 38275 + }, + { + "epoch": 1.47, + "learning_rate": 2.548986919723006e-05, + "loss": 0.2833, + "step": 38276 + }, + { + "epoch": 1.47, + "learning_rate": 2.548922800718133e-05, + "loss": 0.2438, + "step": 38277 + }, + { + "epoch": 1.47, + "learning_rate": 2.54885868171326e-05, + "loss": 0.366, + "step": 38278 + }, + { + "epoch": 1.47, + "learning_rate": 2.548794562708387e-05, + "loss": 0.3497, + "step": 38279 + }, + { + "epoch": 1.47, + "learning_rate": 2.5487304437035136e-05, + "loss": 0.4268, + "step": 38280 + }, + { + "epoch": 1.47, + "learning_rate": 2.5486663246986408e-05, + "loss": 0.4357, + "step": 38281 + }, + { + "epoch": 1.47, + "learning_rate": 2.5486022056937676e-05, + "loss": 0.3935, + "step": 38282 + }, + { + "epoch": 1.47, + "learning_rate": 2.5485380866888948e-05, + "loss": 0.4922, + "step": 38283 + }, + { + "epoch": 1.47, + "learning_rate": 2.548473967684022e-05, + "loss": 0.3329, + "step": 38284 + }, + { + "epoch": 1.47, + "learning_rate": 2.5484098486791485e-05, + "loss": 0.3869, + "step": 38285 + }, + { + "epoch": 1.47, + "learning_rate": 2.5483457296742757e-05, + "loss": 0.3035, + "step": 38286 + }, + { + "epoch": 1.47, + "learning_rate": 2.5482816106694025e-05, + "loss": 0.397, + "step": 38287 + }, + { + "epoch": 1.47, + "learning_rate": 2.5482174916645297e-05, + "loss": 0.3703, + "step": 38288 + }, + { + "epoch": 1.47, + "learning_rate": 2.548153372659656e-05, + "loss": 0.3319, + "step": 38289 + }, + { + "epoch": 1.47, + "learning_rate": 2.5480892536547833e-05, + "loss": 0.3952, + "step": 38290 + }, + { + "epoch": 1.47, + "learning_rate": 2.54802513464991e-05, + "loss": 0.4975, + "step": 38291 + }, + { + "epoch": 1.47, + "learning_rate": 2.5479610156450373e-05, + "loss": 0.2787, + "step": 38292 + }, + { + "epoch": 1.47, + "learning_rate": 2.5478968966401645e-05, + "loss": 0.3385, + "step": 38293 + }, + { + "epoch": 1.47, + "learning_rate": 2.547832777635291e-05, + "loss": 0.2982, + "step": 38294 + }, + { + "epoch": 1.47, + "learning_rate": 2.5477686586304185e-05, + "loss": 0.3367, + "step": 38295 + }, + { + "epoch": 1.47, + "learning_rate": 2.547704539625545e-05, + "loss": 0.3305, + "step": 38296 + }, + { + "epoch": 1.47, + "learning_rate": 2.5476404206206722e-05, + "loss": 0.2275, + "step": 38297 + }, + { + "epoch": 1.47, + "learning_rate": 2.547576301615799e-05, + "loss": 0.3037, + "step": 38298 + }, + { + "epoch": 1.47, + "learning_rate": 2.5475121826109262e-05, + "loss": 0.3168, + "step": 38299 + }, + { + "epoch": 1.47, + "learning_rate": 2.5474480636060527e-05, + "loss": 0.4263, + "step": 38300 + }, + { + "epoch": 1.47, + "learning_rate": 2.54738394460118e-05, + "loss": 0.4397, + "step": 38301 + }, + { + "epoch": 1.47, + "learning_rate": 2.547319825596307e-05, + "loss": 0.3408, + "step": 38302 + }, + { + "epoch": 1.47, + "learning_rate": 2.547255706591434e-05, + "loss": 0.3619, + "step": 38303 + }, + { + "epoch": 1.47, + "learning_rate": 2.547191587586561e-05, + "loss": 0.24, + "step": 38304 + }, + { + "epoch": 1.47, + "learning_rate": 2.5471274685816875e-05, + "loss": 0.2474, + "step": 38305 + }, + { + "epoch": 1.47, + "learning_rate": 2.5470633495768147e-05, + "loss": 0.316, + "step": 38306 + }, + { + "epoch": 1.47, + "learning_rate": 2.5469992305719415e-05, + "loss": 0.4551, + "step": 38307 + }, + { + "epoch": 1.47, + "learning_rate": 2.5469351115670687e-05, + "loss": 0.3601, + "step": 38308 + }, + { + "epoch": 1.47, + "learning_rate": 2.5468709925621952e-05, + "loss": 0.5596, + "step": 38309 + }, + { + "epoch": 1.47, + "learning_rate": 2.5468068735573224e-05, + "loss": 0.4108, + "step": 38310 + }, + { + "epoch": 1.47, + "learning_rate": 2.54674275455245e-05, + "loss": 0.4706, + "step": 38311 + }, + { + "epoch": 1.47, + "learning_rate": 2.5466786355475764e-05, + "loss": 0.363, + "step": 38312 + }, + { + "epoch": 1.47, + "learning_rate": 2.5466145165427035e-05, + "loss": 0.3777, + "step": 38313 + }, + { + "epoch": 1.47, + "learning_rate": 2.5465503975378304e-05, + "loss": 0.4447, + "step": 38314 + }, + { + "epoch": 1.47, + "learning_rate": 2.5464862785329575e-05, + "loss": 0.4018, + "step": 38315 + }, + { + "epoch": 1.47, + "learning_rate": 2.546422159528084e-05, + "loss": 0.4518, + "step": 38316 + }, + { + "epoch": 1.47, + "learning_rate": 2.5463580405232112e-05, + "loss": 0.347, + "step": 38317 + }, + { + "epoch": 1.47, + "learning_rate": 2.546293921518338e-05, + "loss": 0.487, + "step": 38318 + }, + { + "epoch": 1.47, + "learning_rate": 2.5462298025134652e-05, + "loss": 0.3495, + "step": 38319 + }, + { + "epoch": 1.47, + "learning_rate": 2.5461656835085924e-05, + "loss": 0.3138, + "step": 38320 + }, + { + "epoch": 1.47, + "learning_rate": 2.546101564503719e-05, + "loss": 0.4053, + "step": 38321 + }, + { + "epoch": 1.47, + "learning_rate": 2.546037445498846e-05, + "loss": 0.5476, + "step": 38322 + }, + { + "epoch": 1.47, + "learning_rate": 2.545973326493973e-05, + "loss": 0.4605, + "step": 38323 + }, + { + "epoch": 1.47, + "learning_rate": 2.5459092074891e-05, + "loss": 0.3424, + "step": 38324 + }, + { + "epoch": 1.47, + "learning_rate": 2.5458450884842265e-05, + "loss": 0.2816, + "step": 38325 + }, + { + "epoch": 1.47, + "learning_rate": 2.5457809694793537e-05, + "loss": 0.4024, + "step": 38326 + }, + { + "epoch": 1.47, + "learning_rate": 2.5457168504744805e-05, + "loss": 0.5124, + "step": 38327 + }, + { + "epoch": 1.47, + "learning_rate": 2.5456527314696077e-05, + "loss": 0.4153, + "step": 38328 + }, + { + "epoch": 1.47, + "learning_rate": 2.545588612464735e-05, + "loss": 0.3953, + "step": 38329 + }, + { + "epoch": 1.47, + "learning_rate": 2.5455244934598614e-05, + "loss": 0.3953, + "step": 38330 + }, + { + "epoch": 1.47, + "learning_rate": 2.545460374454989e-05, + "loss": 0.4764, + "step": 38331 + }, + { + "epoch": 1.47, + "learning_rate": 2.5453962554501154e-05, + "loss": 0.388, + "step": 38332 + }, + { + "epoch": 1.47, + "learning_rate": 2.5453321364452426e-05, + "loss": 0.4074, + "step": 38333 + }, + { + "epoch": 1.47, + "learning_rate": 2.5452680174403694e-05, + "loss": 0.3475, + "step": 38334 + }, + { + "epoch": 1.47, + "learning_rate": 2.5452038984354966e-05, + "loss": 0.4327, + "step": 38335 + }, + { + "epoch": 1.47, + "learning_rate": 2.545139779430623e-05, + "loss": 0.3624, + "step": 38336 + }, + { + "epoch": 1.47, + "learning_rate": 2.5450756604257502e-05, + "loss": 0.3389, + "step": 38337 + }, + { + "epoch": 1.47, + "learning_rate": 2.5450115414208774e-05, + "loss": 0.5093, + "step": 38338 + }, + { + "epoch": 1.47, + "learning_rate": 2.5449474224160042e-05, + "loss": 0.271, + "step": 38339 + }, + { + "epoch": 1.47, + "learning_rate": 2.5448833034111314e-05, + "loss": 0.3062, + "step": 38340 + }, + { + "epoch": 1.47, + "learning_rate": 2.544819184406258e-05, + "loss": 0.2607, + "step": 38341 + }, + { + "epoch": 1.47, + "learning_rate": 2.544755065401385e-05, + "loss": 0.5398, + "step": 38342 + }, + { + "epoch": 1.47, + "learning_rate": 2.544690946396512e-05, + "loss": 0.3393, + "step": 38343 + }, + { + "epoch": 1.47, + "learning_rate": 2.544626827391639e-05, + "loss": 0.2787, + "step": 38344 + }, + { + "epoch": 1.47, + "learning_rate": 2.5445627083867656e-05, + "loss": 0.4154, + "step": 38345 + }, + { + "epoch": 1.47, + "learning_rate": 2.5444985893818927e-05, + "loss": 0.3296, + "step": 38346 + }, + { + "epoch": 1.47, + "learning_rate": 2.5444344703770203e-05, + "loss": 0.5446, + "step": 38347 + }, + { + "epoch": 1.47, + "learning_rate": 2.5443703513721467e-05, + "loss": 0.362, + "step": 38348 + }, + { + "epoch": 1.47, + "learning_rate": 2.544306232367274e-05, + "loss": 0.4256, + "step": 38349 + }, + { + "epoch": 1.47, + "learning_rate": 2.5442421133624008e-05, + "loss": 0.4623, + "step": 38350 + }, + { + "epoch": 1.47, + "learning_rate": 2.544177994357528e-05, + "loss": 0.3829, + "step": 38351 + }, + { + "epoch": 1.47, + "learning_rate": 2.5441138753526544e-05, + "loss": 0.3556, + "step": 38352 + }, + { + "epoch": 1.47, + "learning_rate": 2.5440497563477816e-05, + "loss": 0.4181, + "step": 38353 + }, + { + "epoch": 1.47, + "learning_rate": 2.5439856373429084e-05, + "loss": 0.3711, + "step": 38354 + }, + { + "epoch": 1.47, + "learning_rate": 2.5439215183380356e-05, + "loss": 0.3429, + "step": 38355 + }, + { + "epoch": 1.47, + "learning_rate": 2.5438573993331628e-05, + "loss": 0.3167, + "step": 38356 + }, + { + "epoch": 1.47, + "learning_rate": 2.5437932803282893e-05, + "loss": 0.2291, + "step": 38357 + }, + { + "epoch": 1.47, + "learning_rate": 2.5437291613234164e-05, + "loss": 0.3376, + "step": 38358 + }, + { + "epoch": 1.47, + "learning_rate": 2.5436650423185433e-05, + "loss": 0.3362, + "step": 38359 + }, + { + "epoch": 1.47, + "learning_rate": 2.5436009233136704e-05, + "loss": 0.372, + "step": 38360 + }, + { + "epoch": 1.47, + "learning_rate": 2.543536804308797e-05, + "loss": 0.3847, + "step": 38361 + }, + { + "epoch": 1.47, + "learning_rate": 2.543472685303924e-05, + "loss": 0.4584, + "step": 38362 + }, + { + "epoch": 1.47, + "learning_rate": 2.543408566299051e-05, + "loss": 0.5579, + "step": 38363 + }, + { + "epoch": 1.47, + "learning_rate": 2.543344447294178e-05, + "loss": 0.2933, + "step": 38364 + }, + { + "epoch": 1.48, + "learning_rate": 2.5432803282893053e-05, + "loss": 0.359, + "step": 38365 + }, + { + "epoch": 1.48, + "learning_rate": 2.5432162092844318e-05, + "loss": 0.4769, + "step": 38366 + }, + { + "epoch": 1.48, + "learning_rate": 2.5431520902795593e-05, + "loss": 0.3734, + "step": 38367 + }, + { + "epoch": 1.48, + "learning_rate": 2.5430879712746858e-05, + "loss": 0.2898, + "step": 38368 + }, + { + "epoch": 1.48, + "learning_rate": 2.543023852269813e-05, + "loss": 0.3678, + "step": 38369 + }, + { + "epoch": 1.48, + "learning_rate": 2.5429597332649398e-05, + "loss": 0.3212, + "step": 38370 + }, + { + "epoch": 1.48, + "learning_rate": 2.542895614260067e-05, + "loss": 0.4395, + "step": 38371 + }, + { + "epoch": 1.48, + "learning_rate": 2.5428314952551934e-05, + "loss": 0.2381, + "step": 38372 + }, + { + "epoch": 1.48, + "learning_rate": 2.5427673762503206e-05, + "loss": 0.3179, + "step": 38373 + }, + { + "epoch": 1.48, + "learning_rate": 2.5427032572454478e-05, + "loss": 0.3301, + "step": 38374 + }, + { + "epoch": 1.48, + "learning_rate": 2.5426391382405746e-05, + "loss": 0.3684, + "step": 38375 + }, + { + "epoch": 1.48, + "learning_rate": 2.5425750192357018e-05, + "loss": 0.3654, + "step": 38376 + }, + { + "epoch": 1.48, + "learning_rate": 2.5425109002308283e-05, + "loss": 0.4116, + "step": 38377 + }, + { + "epoch": 1.48, + "learning_rate": 2.5424467812259555e-05, + "loss": 0.4602, + "step": 38378 + }, + { + "epoch": 1.48, + "learning_rate": 2.5423826622210823e-05, + "loss": 0.3642, + "step": 38379 + }, + { + "epoch": 1.48, + "learning_rate": 2.5423185432162095e-05, + "loss": 0.3777, + "step": 38380 + }, + { + "epoch": 1.48, + "learning_rate": 2.5422544242113366e-05, + "loss": 0.4636, + "step": 38381 + }, + { + "epoch": 1.48, + "learning_rate": 2.542190305206463e-05, + "loss": 0.2441, + "step": 38382 + }, + { + "epoch": 1.48, + "learning_rate": 2.5421261862015906e-05, + "loss": 0.355, + "step": 38383 + }, + { + "epoch": 1.48, + "learning_rate": 2.542062067196717e-05, + "loss": 0.3632, + "step": 38384 + }, + { + "epoch": 1.48, + "learning_rate": 2.5419979481918443e-05, + "loss": 0.3486, + "step": 38385 + }, + { + "epoch": 1.48, + "learning_rate": 2.541933829186971e-05, + "loss": 0.4552, + "step": 38386 + }, + { + "epoch": 1.48, + "learning_rate": 2.5418697101820983e-05, + "loss": 0.483, + "step": 38387 + }, + { + "epoch": 1.48, + "learning_rate": 2.5418055911772248e-05, + "loss": 0.4491, + "step": 38388 + }, + { + "epoch": 1.48, + "learning_rate": 2.541741472172352e-05, + "loss": 0.3997, + "step": 38389 + }, + { + "epoch": 1.48, + "learning_rate": 2.541677353167479e-05, + "loss": 0.3806, + "step": 38390 + }, + { + "epoch": 1.48, + "learning_rate": 2.541613234162606e-05, + "loss": 0.3603, + "step": 38391 + }, + { + "epoch": 1.48, + "learning_rate": 2.541549115157733e-05, + "loss": 0.3195, + "step": 38392 + }, + { + "epoch": 1.48, + "learning_rate": 2.5414849961528596e-05, + "loss": 0.4146, + "step": 38393 + }, + { + "epoch": 1.48, + "learning_rate": 2.5414208771479868e-05, + "loss": 0.3292, + "step": 38394 + }, + { + "epoch": 1.48, + "learning_rate": 2.5413567581431136e-05, + "loss": 0.5013, + "step": 38395 + }, + { + "epoch": 1.48, + "learning_rate": 2.5412926391382408e-05, + "loss": 0.3509, + "step": 38396 + }, + { + "epoch": 1.48, + "learning_rate": 2.5412285201333673e-05, + "loss": 0.3602, + "step": 38397 + }, + { + "epoch": 1.48, + "learning_rate": 2.5411644011284945e-05, + "loss": 0.3123, + "step": 38398 + }, + { + "epoch": 1.48, + "learning_rate": 2.541100282123622e-05, + "loss": 0.4021, + "step": 38399 + }, + { + "epoch": 1.48, + "learning_rate": 2.5410361631187485e-05, + "loss": 0.3596, + "step": 38400 + }, + { + "epoch": 1.48, + "learning_rate": 2.5409720441138757e-05, + "loss": 0.4322, + "step": 38401 + }, + { + "epoch": 1.48, + "learning_rate": 2.5409079251090025e-05, + "loss": 0.3356, + "step": 38402 + }, + { + "epoch": 1.48, + "learning_rate": 2.5408438061041297e-05, + "loss": 0.2465, + "step": 38403 + }, + { + "epoch": 1.48, + "learning_rate": 2.540779687099256e-05, + "loss": 0.2837, + "step": 38404 + }, + { + "epoch": 1.48, + "learning_rate": 2.5407155680943833e-05, + "loss": 0.3149, + "step": 38405 + }, + { + "epoch": 1.48, + "learning_rate": 2.54065144908951e-05, + "loss": 0.4088, + "step": 38406 + }, + { + "epoch": 1.48, + "learning_rate": 2.5405873300846373e-05, + "loss": 0.4031, + "step": 38407 + }, + { + "epoch": 1.48, + "learning_rate": 2.5405232110797645e-05, + "loss": 0.3606, + "step": 38408 + }, + { + "epoch": 1.48, + "learning_rate": 2.540459092074891e-05, + "loss": 0.4632, + "step": 38409 + }, + { + "epoch": 1.48, + "learning_rate": 2.5403949730700182e-05, + "loss": 0.4299, + "step": 38410 + }, + { + "epoch": 1.48, + "learning_rate": 2.540330854065145e-05, + "loss": 0.2733, + "step": 38411 + }, + { + "epoch": 1.48, + "learning_rate": 2.5402667350602722e-05, + "loss": 0.2926, + "step": 38412 + }, + { + "epoch": 1.48, + "learning_rate": 2.5402026160553987e-05, + "loss": 0.334, + "step": 38413 + }, + { + "epoch": 1.48, + "learning_rate": 2.540138497050526e-05, + "loss": 0.4375, + "step": 38414 + }, + { + "epoch": 1.48, + "learning_rate": 2.5400743780456527e-05, + "loss": 0.3642, + "step": 38415 + }, + { + "epoch": 1.48, + "learning_rate": 2.54001025904078e-05, + "loss": 0.4979, + "step": 38416 + }, + { + "epoch": 1.48, + "learning_rate": 2.539946140035907e-05, + "loss": 0.2903, + "step": 38417 + }, + { + "epoch": 1.48, + "learning_rate": 2.5398820210310335e-05, + "loss": 0.3058, + "step": 38418 + }, + { + "epoch": 1.48, + "learning_rate": 2.539817902026161e-05, + "loss": 0.291, + "step": 38419 + }, + { + "epoch": 1.48, + "learning_rate": 2.5397537830212875e-05, + "loss": 0.422, + "step": 38420 + }, + { + "epoch": 1.48, + "learning_rate": 2.5396896640164147e-05, + "loss": 0.4215, + "step": 38421 + }, + { + "epoch": 1.48, + "learning_rate": 2.5396255450115415e-05, + "loss": 0.3587, + "step": 38422 + }, + { + "epoch": 1.48, + "learning_rate": 2.5395614260066687e-05, + "loss": 0.3949, + "step": 38423 + }, + { + "epoch": 1.48, + "learning_rate": 2.5394973070017952e-05, + "loss": 0.4148, + "step": 38424 + }, + { + "epoch": 1.48, + "learning_rate": 2.5394331879969224e-05, + "loss": 0.4198, + "step": 38425 + }, + { + "epoch": 1.48, + "learning_rate": 2.5393690689920495e-05, + "loss": 0.5698, + "step": 38426 + }, + { + "epoch": 1.48, + "learning_rate": 2.5393049499871764e-05, + "loss": 0.3924, + "step": 38427 + }, + { + "epoch": 1.48, + "learning_rate": 2.5392408309823035e-05, + "loss": 0.3391, + "step": 38428 + }, + { + "epoch": 1.48, + "learning_rate": 2.53917671197743e-05, + "loss": 0.4881, + "step": 38429 + }, + { + "epoch": 1.48, + "learning_rate": 2.5391125929725572e-05, + "loss": 0.3137, + "step": 38430 + }, + { + "epoch": 1.48, + "learning_rate": 2.539048473967684e-05, + "loss": 0.455, + "step": 38431 + }, + { + "epoch": 1.48, + "learning_rate": 2.5389843549628112e-05, + "loss": 0.3917, + "step": 38432 + }, + { + "epoch": 1.48, + "learning_rate": 2.5389202359579377e-05, + "loss": 0.3353, + "step": 38433 + }, + { + "epoch": 1.48, + "learning_rate": 2.538856116953065e-05, + "loss": 0.5269, + "step": 38434 + }, + { + "epoch": 1.48, + "learning_rate": 2.5387919979481924e-05, + "loss": 0.4427, + "step": 38435 + }, + { + "epoch": 1.48, + "learning_rate": 2.538727878943319e-05, + "loss": 0.2521, + "step": 38436 + }, + { + "epoch": 1.48, + "learning_rate": 2.538663759938446e-05, + "loss": 0.3443, + "step": 38437 + }, + { + "epoch": 1.48, + "learning_rate": 2.538599640933573e-05, + "loss": 0.3245, + "step": 38438 + }, + { + "epoch": 1.48, + "learning_rate": 2.5385355219287e-05, + "loss": 0.5455, + "step": 38439 + }, + { + "epoch": 1.48, + "learning_rate": 2.5384714029238265e-05, + "loss": 0.3722, + "step": 38440 + }, + { + "epoch": 1.48, + "learning_rate": 2.5384072839189537e-05, + "loss": 0.4146, + "step": 38441 + }, + { + "epoch": 1.48, + "learning_rate": 2.5383431649140806e-05, + "loss": 0.3264, + "step": 38442 + }, + { + "epoch": 1.48, + "learning_rate": 2.5382790459092077e-05, + "loss": 0.3675, + "step": 38443 + }, + { + "epoch": 1.48, + "learning_rate": 2.538214926904335e-05, + "loss": 0.4061, + "step": 38444 + }, + { + "epoch": 1.48, + "learning_rate": 2.5381508078994614e-05, + "loss": 0.4043, + "step": 38445 + }, + { + "epoch": 1.48, + "learning_rate": 2.5380866888945886e-05, + "loss": 0.3149, + "step": 38446 + }, + { + "epoch": 1.48, + "learning_rate": 2.5380225698897154e-05, + "loss": 0.3983, + "step": 38447 + }, + { + "epoch": 1.48, + "learning_rate": 2.5379584508848426e-05, + "loss": 0.3482, + "step": 38448 + }, + { + "epoch": 1.48, + "learning_rate": 2.537894331879969e-05, + "loss": 0.2827, + "step": 38449 + }, + { + "epoch": 1.48, + "learning_rate": 2.5378302128750962e-05, + "loss": 0.429, + "step": 38450 + }, + { + "epoch": 1.48, + "learning_rate": 2.537766093870223e-05, + "loss": 0.3021, + "step": 38451 + }, + { + "epoch": 1.48, + "learning_rate": 2.5377019748653502e-05, + "loss": 0.399, + "step": 38452 + }, + { + "epoch": 1.48, + "learning_rate": 2.5376378558604774e-05, + "loss": 0.3187, + "step": 38453 + }, + { + "epoch": 1.48, + "learning_rate": 2.537573736855604e-05, + "loss": 0.2867, + "step": 38454 + }, + { + "epoch": 1.48, + "learning_rate": 2.5375096178507314e-05, + "loss": 0.4205, + "step": 38455 + }, + { + "epoch": 1.48, + "learning_rate": 2.537445498845858e-05, + "loss": 0.2976, + "step": 38456 + }, + { + "epoch": 1.48, + "learning_rate": 2.537381379840985e-05, + "loss": 0.4714, + "step": 38457 + }, + { + "epoch": 1.48, + "learning_rate": 2.537317260836112e-05, + "loss": 0.4994, + "step": 38458 + }, + { + "epoch": 1.48, + "learning_rate": 2.537253141831239e-05, + "loss": 0.2059, + "step": 38459 + }, + { + "epoch": 1.48, + "learning_rate": 2.5371890228263656e-05, + "loss": 0.3493, + "step": 38460 + }, + { + "epoch": 1.48, + "learning_rate": 2.5371249038214927e-05, + "loss": 0.3533, + "step": 38461 + }, + { + "epoch": 1.48, + "learning_rate": 2.53706078481662e-05, + "loss": 0.4168, + "step": 38462 + }, + { + "epoch": 1.48, + "learning_rate": 2.5369966658117468e-05, + "loss": 0.3444, + "step": 38463 + }, + { + "epoch": 1.48, + "learning_rate": 2.536932546806874e-05, + "loss": 0.3883, + "step": 38464 + }, + { + "epoch": 1.48, + "learning_rate": 2.5368684278020004e-05, + "loss": 0.3423, + "step": 38465 + }, + { + "epoch": 1.48, + "learning_rate": 2.5368043087971276e-05, + "loss": 0.5333, + "step": 38466 + }, + { + "epoch": 1.48, + "learning_rate": 2.5367401897922544e-05, + "loss": 0.3975, + "step": 38467 + }, + { + "epoch": 1.48, + "learning_rate": 2.5366760707873816e-05, + "loss": 0.4432, + "step": 38468 + }, + { + "epoch": 1.48, + "learning_rate": 2.536611951782508e-05, + "loss": 0.3808, + "step": 38469 + }, + { + "epoch": 1.48, + "learning_rate": 2.5365478327776353e-05, + "loss": 0.3476, + "step": 38470 + }, + { + "epoch": 1.48, + "learning_rate": 2.5364837137727628e-05, + "loss": 0.327, + "step": 38471 + }, + { + "epoch": 1.48, + "learning_rate": 2.5364195947678893e-05, + "loss": 0.3099, + "step": 38472 + }, + { + "epoch": 1.48, + "learning_rate": 2.5363554757630164e-05, + "loss": 0.2937, + "step": 38473 + }, + { + "epoch": 1.48, + "learning_rate": 2.5362913567581433e-05, + "loss": 0.2235, + "step": 38474 + }, + { + "epoch": 1.48, + "learning_rate": 2.5362272377532704e-05, + "loss": 0.3769, + "step": 38475 + }, + { + "epoch": 1.48, + "learning_rate": 2.536163118748397e-05, + "loss": 0.3745, + "step": 38476 + }, + { + "epoch": 1.48, + "learning_rate": 2.536098999743524e-05, + "loss": 0.484, + "step": 38477 + }, + { + "epoch": 1.48, + "learning_rate": 2.536034880738651e-05, + "loss": 0.2816, + "step": 38478 + }, + { + "epoch": 1.48, + "learning_rate": 2.535970761733778e-05, + "loss": 0.3235, + "step": 38479 + }, + { + "epoch": 1.48, + "learning_rate": 2.5359066427289053e-05, + "loss": 0.5058, + "step": 38480 + }, + { + "epoch": 1.48, + "learning_rate": 2.5358425237240318e-05, + "loss": 0.4452, + "step": 38481 + }, + { + "epoch": 1.48, + "learning_rate": 2.535778404719159e-05, + "loss": 0.4493, + "step": 38482 + }, + { + "epoch": 1.48, + "learning_rate": 2.5357142857142858e-05, + "loss": 0.3431, + "step": 38483 + }, + { + "epoch": 1.48, + "learning_rate": 2.535650166709413e-05, + "loss": 0.2535, + "step": 38484 + }, + { + "epoch": 1.48, + "learning_rate": 2.5355860477045394e-05, + "loss": 0.3007, + "step": 38485 + }, + { + "epoch": 1.48, + "learning_rate": 2.5355219286996666e-05, + "loss": 0.3961, + "step": 38486 + }, + { + "epoch": 1.48, + "learning_rate": 2.5354578096947935e-05, + "loss": 0.4514, + "step": 38487 + }, + { + "epoch": 1.48, + "learning_rate": 2.5353936906899206e-05, + "loss": 0.2642, + "step": 38488 + }, + { + "epoch": 1.48, + "learning_rate": 2.5353295716850478e-05, + "loss": 0.3311, + "step": 38489 + }, + { + "epoch": 1.48, + "learning_rate": 2.5352654526801743e-05, + "loss": 0.3966, + "step": 38490 + }, + { + "epoch": 1.48, + "learning_rate": 2.5352013336753018e-05, + "loss": 0.488, + "step": 38491 + }, + { + "epoch": 1.48, + "learning_rate": 2.5351372146704283e-05, + "loss": 0.3685, + "step": 38492 + }, + { + "epoch": 1.48, + "learning_rate": 2.5350730956655555e-05, + "loss": 0.3665, + "step": 38493 + }, + { + "epoch": 1.48, + "learning_rate": 2.5350089766606823e-05, + "loss": 0.425, + "step": 38494 + }, + { + "epoch": 1.48, + "learning_rate": 2.5349448576558095e-05, + "loss": 0.3118, + "step": 38495 + }, + { + "epoch": 1.48, + "learning_rate": 2.534880738650936e-05, + "loss": 0.3653, + "step": 38496 + }, + { + "epoch": 1.48, + "learning_rate": 2.534816619646063e-05, + "loss": 0.2654, + "step": 38497 + }, + { + "epoch": 1.48, + "learning_rate": 2.5347525006411903e-05, + "loss": 0.4513, + "step": 38498 + }, + { + "epoch": 1.48, + "learning_rate": 2.534688381636317e-05, + "loss": 0.3979, + "step": 38499 + }, + { + "epoch": 1.48, + "learning_rate": 2.5346242626314443e-05, + "loss": 0.3368, + "step": 38500 + }, + { + "epoch": 1.48, + "learning_rate": 2.5345601436265708e-05, + "loss": 0.3863, + "step": 38501 + }, + { + "epoch": 1.48, + "learning_rate": 2.534496024621698e-05, + "loss": 0.324, + "step": 38502 + }, + { + "epoch": 1.48, + "learning_rate": 2.5344319056168248e-05, + "loss": 0.3884, + "step": 38503 + }, + { + "epoch": 1.48, + "learning_rate": 2.534367786611952e-05, + "loss": 0.3948, + "step": 38504 + }, + { + "epoch": 1.48, + "learning_rate": 2.5343036676070785e-05, + "loss": 0.4265, + "step": 38505 + }, + { + "epoch": 1.48, + "learning_rate": 2.5342395486022056e-05, + "loss": 0.4164, + "step": 38506 + }, + { + "epoch": 1.48, + "learning_rate": 2.534175429597333e-05, + "loss": 0.3647, + "step": 38507 + }, + { + "epoch": 1.48, + "learning_rate": 2.5341113105924597e-05, + "loss": 0.5261, + "step": 38508 + }, + { + "epoch": 1.48, + "learning_rate": 2.5340471915875868e-05, + "loss": 0.2635, + "step": 38509 + }, + { + "epoch": 1.48, + "learning_rate": 2.5339830725827137e-05, + "loss": 0.3705, + "step": 38510 + }, + { + "epoch": 1.48, + "learning_rate": 2.5339189535778408e-05, + "loss": 0.31, + "step": 38511 + }, + { + "epoch": 1.48, + "learning_rate": 2.5338548345729673e-05, + "loss": 0.3861, + "step": 38512 + }, + { + "epoch": 1.48, + "learning_rate": 2.5337907155680945e-05, + "loss": 0.3468, + "step": 38513 + }, + { + "epoch": 1.48, + "learning_rate": 2.5337265965632213e-05, + "loss": 0.2712, + "step": 38514 + }, + { + "epoch": 1.48, + "learning_rate": 2.5336624775583485e-05, + "loss": 0.3272, + "step": 38515 + }, + { + "epoch": 1.48, + "learning_rate": 2.5335983585534757e-05, + "loss": 0.3592, + "step": 38516 + }, + { + "epoch": 1.48, + "learning_rate": 2.533534239548602e-05, + "loss": 0.3129, + "step": 38517 + }, + { + "epoch": 1.48, + "learning_rate": 2.5334701205437293e-05, + "loss": 0.3628, + "step": 38518 + }, + { + "epoch": 1.48, + "learning_rate": 2.5334060015388562e-05, + "loss": 0.4641, + "step": 38519 + }, + { + "epoch": 1.48, + "learning_rate": 2.5333418825339833e-05, + "loss": 0.4538, + "step": 38520 + }, + { + "epoch": 1.48, + "learning_rate": 2.53327776352911e-05, + "loss": 0.446, + "step": 38521 + }, + { + "epoch": 1.48, + "learning_rate": 2.533213644524237e-05, + "loss": 0.338, + "step": 38522 + }, + { + "epoch": 1.48, + "learning_rate": 2.533149525519364e-05, + "loss": 0.274, + "step": 38523 + }, + { + "epoch": 1.48, + "learning_rate": 2.533085406514491e-05, + "loss": 0.2899, + "step": 38524 + }, + { + "epoch": 1.48, + "learning_rate": 2.5330212875096182e-05, + "loss": 0.521, + "step": 38525 + }, + { + "epoch": 1.48, + "learning_rate": 2.532957168504745e-05, + "loss": 0.404, + "step": 38526 + }, + { + "epoch": 1.48, + "learning_rate": 2.5328930494998722e-05, + "loss": 0.3611, + "step": 38527 + }, + { + "epoch": 1.48, + "learning_rate": 2.5328289304949987e-05, + "loss": 0.4526, + "step": 38528 + }, + { + "epoch": 1.48, + "learning_rate": 2.532764811490126e-05, + "loss": 0.3804, + "step": 38529 + }, + { + "epoch": 1.48, + "learning_rate": 2.5327006924852527e-05, + "loss": 0.4388, + "step": 38530 + }, + { + "epoch": 1.48, + "learning_rate": 2.53263657348038e-05, + "loss": 0.3387, + "step": 38531 + }, + { + "epoch": 1.48, + "learning_rate": 2.5325724544755064e-05, + "loss": 0.4391, + "step": 38532 + }, + { + "epoch": 1.48, + "learning_rate": 2.5325083354706335e-05, + "loss": 0.3968, + "step": 38533 + }, + { + "epoch": 1.48, + "learning_rate": 2.5324442164657607e-05, + "loss": 0.4726, + "step": 38534 + }, + { + "epoch": 1.48, + "learning_rate": 2.5323800974608875e-05, + "loss": 0.2909, + "step": 38535 + }, + { + "epoch": 1.48, + "learning_rate": 2.5323159784560147e-05, + "loss": 0.3384, + "step": 38536 + }, + { + "epoch": 1.48, + "learning_rate": 2.5322518594511412e-05, + "loss": 0.4442, + "step": 38537 + }, + { + "epoch": 1.48, + "learning_rate": 2.5321877404462684e-05, + "loss": 0.4887, + "step": 38538 + }, + { + "epoch": 1.48, + "learning_rate": 2.5321236214413952e-05, + "loss": 0.4011, + "step": 38539 + }, + { + "epoch": 1.48, + "learning_rate": 2.5320595024365224e-05, + "loss": 0.5283, + "step": 38540 + }, + { + "epoch": 1.48, + "learning_rate": 2.531995383431649e-05, + "loss": 0.3127, + "step": 38541 + }, + { + "epoch": 1.48, + "learning_rate": 2.531931264426776e-05, + "loss": 0.3883, + "step": 38542 + }, + { + "epoch": 1.48, + "learning_rate": 2.5318671454219035e-05, + "loss": 0.3674, + "step": 38543 + }, + { + "epoch": 1.48, + "learning_rate": 2.53180302641703e-05, + "loss": 0.3026, + "step": 38544 + }, + { + "epoch": 1.48, + "learning_rate": 2.5317389074121572e-05, + "loss": 0.5628, + "step": 38545 + }, + { + "epoch": 1.48, + "learning_rate": 2.531674788407284e-05, + "loss": 0.4354, + "step": 38546 + }, + { + "epoch": 1.48, + "learning_rate": 2.5316106694024112e-05, + "loss": 0.3447, + "step": 38547 + }, + { + "epoch": 1.48, + "learning_rate": 2.5315465503975377e-05, + "loss": 0.3674, + "step": 38548 + }, + { + "epoch": 1.48, + "learning_rate": 2.531482431392665e-05, + "loss": 0.3537, + "step": 38549 + }, + { + "epoch": 1.48, + "learning_rate": 2.5314183123877917e-05, + "loss": 0.3412, + "step": 38550 + }, + { + "epoch": 1.48, + "learning_rate": 2.531354193382919e-05, + "loss": 0.3423, + "step": 38551 + }, + { + "epoch": 1.48, + "learning_rate": 2.531290074378046e-05, + "loss": 0.2761, + "step": 38552 + }, + { + "epoch": 1.48, + "learning_rate": 2.5312259553731726e-05, + "loss": 0.2649, + "step": 38553 + }, + { + "epoch": 1.48, + "learning_rate": 2.5311618363682997e-05, + "loss": 0.393, + "step": 38554 + }, + { + "epoch": 1.48, + "learning_rate": 2.5310977173634266e-05, + "loss": 0.5779, + "step": 38555 + }, + { + "epoch": 1.48, + "learning_rate": 2.5310335983585537e-05, + "loss": 0.3283, + "step": 38556 + }, + { + "epoch": 1.48, + "learning_rate": 2.5309694793536802e-05, + "loss": 0.2614, + "step": 38557 + }, + { + "epoch": 1.48, + "learning_rate": 2.5309053603488074e-05, + "loss": 0.4146, + "step": 38558 + }, + { + "epoch": 1.48, + "learning_rate": 2.5308412413439342e-05, + "loss": 0.3447, + "step": 38559 + }, + { + "epoch": 1.48, + "learning_rate": 2.5307771223390614e-05, + "loss": 0.2692, + "step": 38560 + }, + { + "epoch": 1.48, + "learning_rate": 2.5307130033341886e-05, + "loss": 0.3755, + "step": 38561 + }, + { + "epoch": 1.48, + "learning_rate": 2.5306488843293154e-05, + "loss": 0.3534, + "step": 38562 + }, + { + "epoch": 1.48, + "learning_rate": 2.5305847653244426e-05, + "loss": 0.2649, + "step": 38563 + }, + { + "epoch": 1.48, + "learning_rate": 2.530520646319569e-05, + "loss": 0.417, + "step": 38564 + }, + { + "epoch": 1.48, + "learning_rate": 2.5304565273146962e-05, + "loss": 0.492, + "step": 38565 + }, + { + "epoch": 1.48, + "learning_rate": 2.530392408309823e-05, + "loss": 0.3791, + "step": 38566 + }, + { + "epoch": 1.48, + "learning_rate": 2.5303282893049502e-05, + "loss": 0.302, + "step": 38567 + }, + { + "epoch": 1.48, + "learning_rate": 2.5302641703000774e-05, + "loss": 0.3583, + "step": 38568 + }, + { + "epoch": 1.48, + "learning_rate": 2.530200051295204e-05, + "loss": 0.2813, + "step": 38569 + }, + { + "epoch": 1.48, + "learning_rate": 2.530135932290331e-05, + "loss": 0.4417, + "step": 38570 + }, + { + "epoch": 1.48, + "learning_rate": 2.530071813285458e-05, + "loss": 0.2866, + "step": 38571 + }, + { + "epoch": 1.48, + "learning_rate": 2.530007694280585e-05, + "loss": 0.3402, + "step": 38572 + }, + { + "epoch": 1.48, + "learning_rate": 2.5299435752757116e-05, + "loss": 0.364, + "step": 38573 + }, + { + "epoch": 1.48, + "learning_rate": 2.5298794562708388e-05, + "loss": 0.393, + "step": 38574 + }, + { + "epoch": 1.48, + "learning_rate": 2.5298153372659656e-05, + "loss": 0.3526, + "step": 38575 + }, + { + "epoch": 1.48, + "learning_rate": 2.5297512182610928e-05, + "loss": 0.3259, + "step": 38576 + }, + { + "epoch": 1.48, + "learning_rate": 2.52968709925622e-05, + "loss": 0.3731, + "step": 38577 + }, + { + "epoch": 1.48, + "learning_rate": 2.5296229802513464e-05, + "loss": 0.4061, + "step": 38578 + }, + { + "epoch": 1.48, + "learning_rate": 2.529558861246474e-05, + "loss": 0.4183, + "step": 38579 + }, + { + "epoch": 1.48, + "learning_rate": 2.5294947422416004e-05, + "loss": 0.3537, + "step": 38580 + }, + { + "epoch": 1.48, + "learning_rate": 2.5294306232367276e-05, + "loss": 0.3968, + "step": 38581 + }, + { + "epoch": 1.48, + "learning_rate": 2.5293665042318544e-05, + "loss": 0.3962, + "step": 38582 + }, + { + "epoch": 1.48, + "learning_rate": 2.5293023852269816e-05, + "loss": 0.2065, + "step": 38583 + }, + { + "epoch": 1.48, + "learning_rate": 2.529238266222108e-05, + "loss": 0.3146, + "step": 38584 + }, + { + "epoch": 1.48, + "learning_rate": 2.5291741472172353e-05, + "loss": 0.2958, + "step": 38585 + }, + { + "epoch": 1.48, + "learning_rate": 2.5291100282123624e-05, + "loss": 0.2652, + "step": 38586 + }, + { + "epoch": 1.48, + "learning_rate": 2.5290459092074893e-05, + "loss": 0.3695, + "step": 38587 + }, + { + "epoch": 1.48, + "learning_rate": 2.5289817902026164e-05, + "loss": 0.373, + "step": 38588 + }, + { + "epoch": 1.48, + "learning_rate": 2.528917671197743e-05, + "loss": 0.4181, + "step": 38589 + }, + { + "epoch": 1.48, + "learning_rate": 2.52885355219287e-05, + "loss": 0.2991, + "step": 38590 + }, + { + "epoch": 1.48, + "learning_rate": 2.528789433187997e-05, + "loss": 0.3545, + "step": 38591 + }, + { + "epoch": 1.48, + "learning_rate": 2.528725314183124e-05, + "loss": 0.4006, + "step": 38592 + }, + { + "epoch": 1.48, + "learning_rate": 2.5286611951782506e-05, + "loss": 0.3864, + "step": 38593 + }, + { + "epoch": 1.48, + "learning_rate": 2.5285970761733778e-05, + "loss": 0.3484, + "step": 38594 + }, + { + "epoch": 1.48, + "learning_rate": 2.5285329571685053e-05, + "loss": 0.3969, + "step": 38595 + }, + { + "epoch": 1.48, + "learning_rate": 2.5284688381636318e-05, + "loss": 0.276, + "step": 38596 + }, + { + "epoch": 1.48, + "learning_rate": 2.528404719158759e-05, + "loss": 0.3878, + "step": 38597 + }, + { + "epoch": 1.48, + "learning_rate": 2.5283406001538858e-05, + "loss": 0.4456, + "step": 38598 + }, + { + "epoch": 1.48, + "learning_rate": 2.528276481149013e-05, + "loss": 0.4962, + "step": 38599 + }, + { + "epoch": 1.48, + "learning_rate": 2.5282123621441395e-05, + "loss": 0.4241, + "step": 38600 + }, + { + "epoch": 1.48, + "learning_rate": 2.5281482431392666e-05, + "loss": 0.3299, + "step": 38601 + }, + { + "epoch": 1.48, + "learning_rate": 2.5280841241343935e-05, + "loss": 0.386, + "step": 38602 + }, + { + "epoch": 1.48, + "learning_rate": 2.5280200051295206e-05, + "loss": 0.4115, + "step": 38603 + }, + { + "epoch": 1.48, + "learning_rate": 2.5279558861246478e-05, + "loss": 0.4864, + "step": 38604 + }, + { + "epoch": 1.48, + "learning_rate": 2.5278917671197743e-05, + "loss": 0.41, + "step": 38605 + }, + { + "epoch": 1.48, + "learning_rate": 2.5278276481149015e-05, + "loss": 0.3016, + "step": 38606 + }, + { + "epoch": 1.48, + "learning_rate": 2.5277635291100283e-05, + "loss": 0.3153, + "step": 38607 + }, + { + "epoch": 1.48, + "learning_rate": 2.5276994101051555e-05, + "loss": 0.3609, + "step": 38608 + }, + { + "epoch": 1.48, + "learning_rate": 2.527635291100282e-05, + "loss": 0.4369, + "step": 38609 + }, + { + "epoch": 1.48, + "learning_rate": 2.527571172095409e-05, + "loss": 0.3914, + "step": 38610 + }, + { + "epoch": 1.48, + "learning_rate": 2.527507053090536e-05, + "loss": 0.2452, + "step": 38611 + }, + { + "epoch": 1.48, + "learning_rate": 2.527442934085663e-05, + "loss": 0.3653, + "step": 38612 + }, + { + "epoch": 1.48, + "learning_rate": 2.5273788150807903e-05, + "loss": 0.2841, + "step": 38613 + }, + { + "epoch": 1.48, + "learning_rate": 2.5273146960759168e-05, + "loss": 0.2874, + "step": 38614 + }, + { + "epoch": 1.48, + "learning_rate": 2.5272505770710443e-05, + "loss": 0.5213, + "step": 38615 + }, + { + "epoch": 1.48, + "learning_rate": 2.5271864580661708e-05, + "loss": 0.4537, + "step": 38616 + }, + { + "epoch": 1.48, + "learning_rate": 2.527122339061298e-05, + "loss": 0.3586, + "step": 38617 + }, + { + "epoch": 1.48, + "learning_rate": 2.5270582200564248e-05, + "loss": 0.422, + "step": 38618 + }, + { + "epoch": 1.48, + "learning_rate": 2.526994101051552e-05, + "loss": 0.3055, + "step": 38619 + }, + { + "epoch": 1.48, + "learning_rate": 2.5269299820466785e-05, + "loss": 0.3612, + "step": 38620 + }, + { + "epoch": 1.48, + "learning_rate": 2.5268658630418057e-05, + "loss": 0.4133, + "step": 38621 + }, + { + "epoch": 1.48, + "learning_rate": 2.5268017440369328e-05, + "loss": 0.4761, + "step": 38622 + }, + { + "epoch": 1.48, + "learning_rate": 2.5267376250320597e-05, + "loss": 0.3801, + "step": 38623 + }, + { + "epoch": 1.48, + "learning_rate": 2.526673506027187e-05, + "loss": 0.4009, + "step": 38624 + }, + { + "epoch": 1.49, + "learning_rate": 2.5266093870223133e-05, + "loss": 0.3987, + "step": 38625 + }, + { + "epoch": 1.49, + "learning_rate": 2.5265452680174405e-05, + "loss": 0.4598, + "step": 38626 + }, + { + "epoch": 1.49, + "learning_rate": 2.5264811490125673e-05, + "loss": 0.2994, + "step": 38627 + }, + { + "epoch": 1.49, + "learning_rate": 2.5264170300076945e-05, + "loss": 0.3101, + "step": 38628 + }, + { + "epoch": 1.49, + "learning_rate": 2.526352911002821e-05, + "loss": 0.3297, + "step": 38629 + }, + { + "epoch": 1.49, + "learning_rate": 2.526288791997948e-05, + "loss": 0.3468, + "step": 38630 + }, + { + "epoch": 1.49, + "learning_rate": 2.5262246729930757e-05, + "loss": 0.4129, + "step": 38631 + }, + { + "epoch": 1.49, + "learning_rate": 2.5261605539882022e-05, + "loss": 0.363, + "step": 38632 + }, + { + "epoch": 1.49, + "learning_rate": 2.5260964349833293e-05, + "loss": 0.4796, + "step": 38633 + }, + { + "epoch": 1.49, + "learning_rate": 2.5260323159784562e-05, + "loss": 0.3474, + "step": 38634 + }, + { + "epoch": 1.49, + "learning_rate": 2.5259681969735833e-05, + "loss": 0.4113, + "step": 38635 + }, + { + "epoch": 1.49, + "learning_rate": 2.52590407796871e-05, + "loss": 0.333, + "step": 38636 + }, + { + "epoch": 1.49, + "learning_rate": 2.525839958963837e-05, + "loss": 0.3326, + "step": 38637 + }, + { + "epoch": 1.49, + "learning_rate": 2.525775839958964e-05, + "loss": 0.3791, + "step": 38638 + }, + { + "epoch": 1.49, + "learning_rate": 2.525711720954091e-05, + "loss": 0.3578, + "step": 38639 + }, + { + "epoch": 1.49, + "learning_rate": 2.5256476019492182e-05, + "loss": 0.5047, + "step": 38640 + }, + { + "epoch": 1.49, + "learning_rate": 2.5255834829443447e-05, + "loss": 0.2839, + "step": 38641 + }, + { + "epoch": 1.49, + "learning_rate": 2.525519363939472e-05, + "loss": 0.3692, + "step": 38642 + }, + { + "epoch": 1.49, + "learning_rate": 2.5254552449345987e-05, + "loss": 0.3243, + "step": 38643 + }, + { + "epoch": 1.49, + "learning_rate": 2.525391125929726e-05, + "loss": 0.4272, + "step": 38644 + }, + { + "epoch": 1.49, + "learning_rate": 2.5253270069248524e-05, + "loss": 0.325, + "step": 38645 + }, + { + "epoch": 1.49, + "learning_rate": 2.5252628879199795e-05, + "loss": 0.328, + "step": 38646 + }, + { + "epoch": 1.49, + "learning_rate": 2.5251987689151064e-05, + "loss": 0.3576, + "step": 38647 + }, + { + "epoch": 1.49, + "learning_rate": 2.5251346499102335e-05, + "loss": 0.2827, + "step": 38648 + }, + { + "epoch": 1.49, + "learning_rate": 2.5250705309053607e-05, + "loss": 0.2821, + "step": 38649 + }, + { + "epoch": 1.49, + "learning_rate": 2.5250064119004875e-05, + "loss": 0.4237, + "step": 38650 + }, + { + "epoch": 1.49, + "learning_rate": 2.5249422928956147e-05, + "loss": 0.3533, + "step": 38651 + }, + { + "epoch": 1.49, + "learning_rate": 2.5248781738907412e-05, + "loss": 0.3818, + "step": 38652 + }, + { + "epoch": 1.49, + "learning_rate": 2.5248140548858684e-05, + "loss": 0.3898, + "step": 38653 + }, + { + "epoch": 1.49, + "learning_rate": 2.5247499358809952e-05, + "loss": 0.3067, + "step": 38654 + }, + { + "epoch": 1.49, + "learning_rate": 2.5246858168761224e-05, + "loss": 0.2815, + "step": 38655 + }, + { + "epoch": 1.49, + "learning_rate": 2.524621697871249e-05, + "loss": 0.4551, + "step": 38656 + }, + { + "epoch": 1.49, + "learning_rate": 2.524557578866376e-05, + "loss": 0.2094, + "step": 38657 + }, + { + "epoch": 1.49, + "learning_rate": 2.5244934598615032e-05, + "loss": 0.3413, + "step": 38658 + }, + { + "epoch": 1.49, + "learning_rate": 2.52442934085663e-05, + "loss": 0.4635, + "step": 38659 + }, + { + "epoch": 1.49, + "learning_rate": 2.5243652218517572e-05, + "loss": 0.3698, + "step": 38660 + }, + { + "epoch": 1.49, + "learning_rate": 2.5243011028468837e-05, + "loss": 0.3206, + "step": 38661 + }, + { + "epoch": 1.49, + "learning_rate": 2.524236983842011e-05, + "loss": 0.3438, + "step": 38662 + }, + { + "epoch": 1.49, + "learning_rate": 2.5241728648371377e-05, + "loss": 0.308, + "step": 38663 + }, + { + "epoch": 1.49, + "learning_rate": 2.524108745832265e-05, + "loss": 0.3927, + "step": 38664 + }, + { + "epoch": 1.49, + "learning_rate": 2.5240446268273914e-05, + "loss": 0.2058, + "step": 38665 + }, + { + "epoch": 1.49, + "learning_rate": 2.5239805078225186e-05, + "loss": 0.3344, + "step": 38666 + }, + { + "epoch": 1.49, + "learning_rate": 2.523916388817646e-05, + "loss": 0.4497, + "step": 38667 + }, + { + "epoch": 1.49, + "learning_rate": 2.5238522698127726e-05, + "loss": 0.3024, + "step": 38668 + }, + { + "epoch": 1.49, + "learning_rate": 2.5237881508078997e-05, + "loss": 0.3414, + "step": 38669 + }, + { + "epoch": 1.49, + "learning_rate": 2.5237240318030266e-05, + "loss": 0.2525, + "step": 38670 + }, + { + "epoch": 1.49, + "learning_rate": 2.5236599127981537e-05, + "loss": 0.3848, + "step": 38671 + }, + { + "epoch": 1.49, + "learning_rate": 2.5235957937932802e-05, + "loss": 0.4548, + "step": 38672 + }, + { + "epoch": 1.49, + "learning_rate": 2.5235316747884074e-05, + "loss": 0.2687, + "step": 38673 + }, + { + "epoch": 1.49, + "learning_rate": 2.5234675557835342e-05, + "loss": 0.1871, + "step": 38674 + }, + { + "epoch": 1.49, + "learning_rate": 2.5234034367786614e-05, + "loss": 0.3718, + "step": 38675 + }, + { + "epoch": 1.49, + "learning_rate": 2.5233393177737886e-05, + "loss": 0.3576, + "step": 38676 + }, + { + "epoch": 1.49, + "learning_rate": 2.523275198768915e-05, + "loss": 0.3735, + "step": 38677 + }, + { + "epoch": 1.49, + "learning_rate": 2.5232110797640422e-05, + "loss": 0.3921, + "step": 38678 + }, + { + "epoch": 1.49, + "learning_rate": 2.523146960759169e-05, + "loss": 0.3663, + "step": 38679 + }, + { + "epoch": 1.49, + "learning_rate": 2.5230828417542962e-05, + "loss": 0.4226, + "step": 38680 + }, + { + "epoch": 1.49, + "learning_rate": 2.5230187227494227e-05, + "loss": 0.3834, + "step": 38681 + }, + { + "epoch": 1.49, + "learning_rate": 2.52295460374455e-05, + "loss": 0.3318, + "step": 38682 + }, + { + "epoch": 1.49, + "learning_rate": 2.5228904847396767e-05, + "loss": 0.443, + "step": 38683 + }, + { + "epoch": 1.49, + "learning_rate": 2.522826365734804e-05, + "loss": 0.365, + "step": 38684 + }, + { + "epoch": 1.49, + "learning_rate": 2.522762246729931e-05, + "loss": 0.3029, + "step": 38685 + }, + { + "epoch": 1.49, + "learning_rate": 2.522698127725058e-05, + "loss": 0.4308, + "step": 38686 + }, + { + "epoch": 1.49, + "learning_rate": 2.522634008720185e-05, + "loss": 0.3125, + "step": 38687 + }, + { + "epoch": 1.49, + "learning_rate": 2.5225698897153116e-05, + "loss": 0.2391, + "step": 38688 + }, + { + "epoch": 1.49, + "learning_rate": 2.5225057707104388e-05, + "loss": 0.3277, + "step": 38689 + }, + { + "epoch": 1.49, + "learning_rate": 2.5224416517055656e-05, + "loss": 0.3075, + "step": 38690 + }, + { + "epoch": 1.49, + "learning_rate": 2.5223775327006928e-05, + "loss": 0.4797, + "step": 38691 + }, + { + "epoch": 1.49, + "learning_rate": 2.5223134136958193e-05, + "loss": 0.3522, + "step": 38692 + }, + { + "epoch": 1.49, + "learning_rate": 2.5222492946909464e-05, + "loss": 0.2988, + "step": 38693 + }, + { + "epoch": 1.49, + "learning_rate": 2.5221851756860736e-05, + "loss": 0.2301, + "step": 38694 + }, + { + "epoch": 1.49, + "learning_rate": 2.5221210566812004e-05, + "loss": 0.2754, + "step": 38695 + }, + { + "epoch": 1.49, + "learning_rate": 2.5220569376763276e-05, + "loss": 0.343, + "step": 38696 + }, + { + "epoch": 1.49, + "learning_rate": 2.521992818671454e-05, + "loss": 0.3148, + "step": 38697 + }, + { + "epoch": 1.49, + "learning_rate": 2.5219286996665813e-05, + "loss": 0.3178, + "step": 38698 + }, + { + "epoch": 1.49, + "learning_rate": 2.521864580661708e-05, + "loss": 0.3929, + "step": 38699 + }, + { + "epoch": 1.49, + "learning_rate": 2.5218004616568353e-05, + "loss": 0.528, + "step": 38700 + }, + { + "epoch": 1.49, + "learning_rate": 2.5217363426519618e-05, + "loss": 0.3211, + "step": 38701 + }, + { + "epoch": 1.49, + "learning_rate": 2.521672223647089e-05, + "loss": 0.2698, + "step": 38702 + }, + { + "epoch": 1.49, + "learning_rate": 2.5216081046422165e-05, + "loss": 0.4396, + "step": 38703 + }, + { + "epoch": 1.49, + "learning_rate": 2.521543985637343e-05, + "loss": 0.3907, + "step": 38704 + }, + { + "epoch": 1.49, + "learning_rate": 2.52147986663247e-05, + "loss": 0.2271, + "step": 38705 + }, + { + "epoch": 1.49, + "learning_rate": 2.521415747627597e-05, + "loss": 0.395, + "step": 38706 + }, + { + "epoch": 1.49, + "learning_rate": 2.521351628622724e-05, + "loss": 0.2734, + "step": 38707 + }, + { + "epoch": 1.49, + "learning_rate": 2.5212875096178506e-05, + "loss": 0.3887, + "step": 38708 + }, + { + "epoch": 1.49, + "learning_rate": 2.5212233906129778e-05, + "loss": 0.4663, + "step": 38709 + }, + { + "epoch": 1.49, + "learning_rate": 2.5211592716081046e-05, + "loss": 0.3555, + "step": 38710 + }, + { + "epoch": 1.49, + "learning_rate": 2.5210951526032318e-05, + "loss": 0.3235, + "step": 38711 + }, + { + "epoch": 1.49, + "learning_rate": 2.521031033598359e-05, + "loss": 0.3804, + "step": 38712 + }, + { + "epoch": 1.49, + "learning_rate": 2.5209669145934855e-05, + "loss": 0.4707, + "step": 38713 + }, + { + "epoch": 1.49, + "learning_rate": 2.5209027955886126e-05, + "loss": 0.3695, + "step": 38714 + }, + { + "epoch": 1.49, + "learning_rate": 2.5208386765837395e-05, + "loss": 0.3323, + "step": 38715 + }, + { + "epoch": 1.49, + "learning_rate": 2.5207745575788666e-05, + "loss": 0.3684, + "step": 38716 + }, + { + "epoch": 1.49, + "learning_rate": 2.520710438573993e-05, + "loss": 0.3847, + "step": 38717 + }, + { + "epoch": 1.49, + "learning_rate": 2.5206463195691203e-05, + "loss": 0.2905, + "step": 38718 + }, + { + "epoch": 1.49, + "learning_rate": 2.520582200564247e-05, + "loss": 0.2808, + "step": 38719 + }, + { + "epoch": 1.49, + "learning_rate": 2.5205180815593743e-05, + "loss": 0.4594, + "step": 38720 + }, + { + "epoch": 1.49, + "learning_rate": 2.5204539625545015e-05, + "loss": 0.4052, + "step": 38721 + }, + { + "epoch": 1.49, + "learning_rate": 2.5203898435496283e-05, + "loss": 0.3221, + "step": 38722 + }, + { + "epoch": 1.49, + "learning_rate": 2.5203257245447555e-05, + "loss": 0.4689, + "step": 38723 + }, + { + "epoch": 1.49, + "learning_rate": 2.520261605539882e-05, + "loss": 0.3039, + "step": 38724 + }, + { + "epoch": 1.49, + "learning_rate": 2.520197486535009e-05, + "loss": 0.3562, + "step": 38725 + }, + { + "epoch": 1.49, + "learning_rate": 2.520133367530136e-05, + "loss": 0.3733, + "step": 38726 + }, + { + "epoch": 1.49, + "learning_rate": 2.520069248525263e-05, + "loss": 0.3834, + "step": 38727 + }, + { + "epoch": 1.49, + "learning_rate": 2.5200051295203896e-05, + "loss": 0.3963, + "step": 38728 + }, + { + "epoch": 1.49, + "learning_rate": 2.5199410105155168e-05, + "loss": 0.4512, + "step": 38729 + }, + { + "epoch": 1.49, + "learning_rate": 2.519876891510644e-05, + "loss": 0.4744, + "step": 38730 + }, + { + "epoch": 1.49, + "learning_rate": 2.5198127725057708e-05, + "loss": 0.3767, + "step": 38731 + }, + { + "epoch": 1.49, + "learning_rate": 2.519748653500898e-05, + "loss": 0.4264, + "step": 38732 + }, + { + "epoch": 1.49, + "learning_rate": 2.5196845344960245e-05, + "loss": 0.3494, + "step": 38733 + }, + { + "epoch": 1.49, + "learning_rate": 2.5196204154911517e-05, + "loss": 0.4497, + "step": 38734 + }, + { + "epoch": 1.49, + "learning_rate": 2.5195562964862785e-05, + "loss": 0.3759, + "step": 38735 + }, + { + "epoch": 1.49, + "learning_rate": 2.5194921774814057e-05, + "loss": 0.4964, + "step": 38736 + }, + { + "epoch": 1.49, + "learning_rate": 2.519428058476532e-05, + "loss": 0.3715, + "step": 38737 + }, + { + "epoch": 1.49, + "learning_rate": 2.5193639394716593e-05, + "loss": 0.3242, + "step": 38738 + }, + { + "epoch": 1.49, + "learning_rate": 2.519299820466787e-05, + "loss": 0.4671, + "step": 38739 + }, + { + "epoch": 1.49, + "learning_rate": 2.5192357014619133e-05, + "loss": 0.3137, + "step": 38740 + }, + { + "epoch": 1.49, + "learning_rate": 2.5191715824570405e-05, + "loss": 0.3706, + "step": 38741 + }, + { + "epoch": 1.49, + "learning_rate": 2.5191074634521673e-05, + "loss": 0.3542, + "step": 38742 + }, + { + "epoch": 1.49, + "learning_rate": 2.5190433444472945e-05, + "loss": 0.2569, + "step": 38743 + }, + { + "epoch": 1.49, + "learning_rate": 2.518979225442421e-05, + "loss": 0.2874, + "step": 38744 + }, + { + "epoch": 1.49, + "learning_rate": 2.5189151064375482e-05, + "loss": 0.2821, + "step": 38745 + }, + { + "epoch": 1.49, + "learning_rate": 2.518850987432675e-05, + "loss": 0.2689, + "step": 38746 + }, + { + "epoch": 1.49, + "learning_rate": 2.5187868684278022e-05, + "loss": 0.3118, + "step": 38747 + }, + { + "epoch": 1.49, + "learning_rate": 2.5187227494229294e-05, + "loss": 0.483, + "step": 38748 + }, + { + "epoch": 1.49, + "learning_rate": 2.518658630418056e-05, + "loss": 0.3453, + "step": 38749 + }, + { + "epoch": 1.49, + "learning_rate": 2.518594511413183e-05, + "loss": 0.3604, + "step": 38750 + }, + { + "epoch": 1.49, + "learning_rate": 2.51853039240831e-05, + "loss": 0.3153, + "step": 38751 + }, + { + "epoch": 1.49, + "learning_rate": 2.518466273403437e-05, + "loss": 0.2925, + "step": 38752 + }, + { + "epoch": 1.49, + "learning_rate": 2.5184021543985635e-05, + "loss": 0.3548, + "step": 38753 + }, + { + "epoch": 1.49, + "learning_rate": 2.5183380353936907e-05, + "loss": 0.4721, + "step": 38754 + }, + { + "epoch": 1.49, + "learning_rate": 2.5182739163888175e-05, + "loss": 0.3432, + "step": 38755 + }, + { + "epoch": 1.49, + "learning_rate": 2.5182097973839447e-05, + "loss": 0.2667, + "step": 38756 + }, + { + "epoch": 1.49, + "learning_rate": 2.518145678379072e-05, + "loss": 0.4067, + "step": 38757 + }, + { + "epoch": 1.49, + "learning_rate": 2.5180815593741987e-05, + "loss": 0.2841, + "step": 38758 + }, + { + "epoch": 1.49, + "learning_rate": 2.518017440369326e-05, + "loss": 0.4773, + "step": 38759 + }, + { + "epoch": 1.49, + "learning_rate": 2.5179533213644524e-05, + "loss": 0.3442, + "step": 38760 + }, + { + "epoch": 1.49, + "learning_rate": 2.5178892023595795e-05, + "loss": 0.4934, + "step": 38761 + }, + { + "epoch": 1.49, + "learning_rate": 2.5178250833547064e-05, + "loss": 0.4092, + "step": 38762 + }, + { + "epoch": 1.49, + "learning_rate": 2.5177609643498335e-05, + "loss": 0.3793, + "step": 38763 + }, + { + "epoch": 1.49, + "learning_rate": 2.5176968453449607e-05, + "loss": 0.4188, + "step": 38764 + }, + { + "epoch": 1.49, + "learning_rate": 2.5176327263400872e-05, + "loss": 0.4235, + "step": 38765 + }, + { + "epoch": 1.49, + "learning_rate": 2.5175686073352144e-05, + "loss": 0.379, + "step": 38766 + }, + { + "epoch": 1.49, + "learning_rate": 2.5175044883303412e-05, + "loss": 0.334, + "step": 38767 + }, + { + "epoch": 1.49, + "learning_rate": 2.5174403693254684e-05, + "loss": 0.4841, + "step": 38768 + }, + { + "epoch": 1.49, + "learning_rate": 2.517376250320595e-05, + "loss": 0.4119, + "step": 38769 + }, + { + "epoch": 1.49, + "learning_rate": 2.517312131315722e-05, + "loss": 0.4439, + "step": 38770 + }, + { + "epoch": 1.49, + "learning_rate": 2.517248012310849e-05, + "loss": 0.4223, + "step": 38771 + }, + { + "epoch": 1.49, + "learning_rate": 2.517183893305976e-05, + "loss": 0.4381, + "step": 38772 + }, + { + "epoch": 1.49, + "learning_rate": 2.5171197743011032e-05, + "loss": 0.3619, + "step": 38773 + }, + { + "epoch": 1.49, + "learning_rate": 2.51705565529623e-05, + "loss": 0.4639, + "step": 38774 + }, + { + "epoch": 1.49, + "learning_rate": 2.5169915362913572e-05, + "loss": 0.4712, + "step": 38775 + }, + { + "epoch": 1.49, + "learning_rate": 2.5169274172864837e-05, + "loss": 0.364, + "step": 38776 + }, + { + "epoch": 1.49, + "learning_rate": 2.516863298281611e-05, + "loss": 0.4351, + "step": 38777 + }, + { + "epoch": 1.49, + "learning_rate": 2.5167991792767377e-05, + "loss": 0.239, + "step": 38778 + }, + { + "epoch": 1.49, + "learning_rate": 2.516735060271865e-05, + "loss": 0.3012, + "step": 38779 + }, + { + "epoch": 1.49, + "learning_rate": 2.5166709412669914e-05, + "loss": 0.3393, + "step": 38780 + }, + { + "epoch": 1.49, + "learning_rate": 2.5166068222621186e-05, + "loss": 0.2386, + "step": 38781 + }, + { + "epoch": 1.49, + "learning_rate": 2.5165427032572457e-05, + "loss": 0.232, + "step": 38782 + }, + { + "epoch": 1.49, + "learning_rate": 2.5164785842523726e-05, + "loss": 0.274, + "step": 38783 + }, + { + "epoch": 1.49, + "learning_rate": 2.5164144652474997e-05, + "loss": 0.2575, + "step": 38784 + }, + { + "epoch": 1.49, + "learning_rate": 2.5163503462426262e-05, + "loss": 0.2965, + "step": 38785 + }, + { + "epoch": 1.49, + "learning_rate": 2.5162862272377534e-05, + "loss": 0.3033, + "step": 38786 + }, + { + "epoch": 1.49, + "learning_rate": 2.5162221082328802e-05, + "loss": 0.4782, + "step": 38787 + }, + { + "epoch": 1.49, + "learning_rate": 2.5161579892280074e-05, + "loss": 0.4083, + "step": 38788 + }, + { + "epoch": 1.49, + "learning_rate": 2.516093870223134e-05, + "loss": 0.3529, + "step": 38789 + }, + { + "epoch": 1.49, + "learning_rate": 2.516029751218261e-05, + "loss": 0.387, + "step": 38790 + }, + { + "epoch": 1.49, + "learning_rate": 2.5159656322133886e-05, + "loss": 0.2647, + "step": 38791 + }, + { + "epoch": 1.49, + "learning_rate": 2.515901513208515e-05, + "loss": 0.3315, + "step": 38792 + }, + { + "epoch": 1.49, + "learning_rate": 2.5158373942036423e-05, + "loss": 0.3483, + "step": 38793 + }, + { + "epoch": 1.49, + "learning_rate": 2.515773275198769e-05, + "loss": 0.4787, + "step": 38794 + }, + { + "epoch": 1.49, + "learning_rate": 2.5157091561938963e-05, + "loss": 0.5407, + "step": 38795 + }, + { + "epoch": 1.49, + "learning_rate": 2.5156450371890228e-05, + "loss": 0.3853, + "step": 38796 + }, + { + "epoch": 1.49, + "learning_rate": 2.51558091818415e-05, + "loss": 0.3626, + "step": 38797 + }, + { + "epoch": 1.49, + "learning_rate": 2.5155167991792768e-05, + "loss": 0.3714, + "step": 38798 + }, + { + "epoch": 1.49, + "learning_rate": 2.515452680174404e-05, + "loss": 0.4643, + "step": 38799 + }, + { + "epoch": 1.49, + "learning_rate": 2.515388561169531e-05, + "loss": 0.3967, + "step": 38800 + }, + { + "epoch": 1.49, + "learning_rate": 2.5153244421646576e-05, + "loss": 0.327, + "step": 38801 + }, + { + "epoch": 1.49, + "learning_rate": 2.5152603231597848e-05, + "loss": 0.3849, + "step": 38802 + }, + { + "epoch": 1.49, + "learning_rate": 2.5151962041549116e-05, + "loss": 0.3098, + "step": 38803 + }, + { + "epoch": 1.49, + "learning_rate": 2.5151320851500388e-05, + "loss": 0.382, + "step": 38804 + }, + { + "epoch": 1.49, + "learning_rate": 2.5150679661451653e-05, + "loss": 0.3476, + "step": 38805 + }, + { + "epoch": 1.49, + "learning_rate": 2.5150038471402924e-05, + "loss": 0.4738, + "step": 38806 + }, + { + "epoch": 1.49, + "learning_rate": 2.5149397281354193e-05, + "loss": 0.3333, + "step": 38807 + }, + { + "epoch": 1.49, + "learning_rate": 2.5148756091305464e-05, + "loss": 0.3165, + "step": 38808 + }, + { + "epoch": 1.49, + "learning_rate": 2.5148114901256736e-05, + "loss": 0.4699, + "step": 38809 + }, + { + "epoch": 1.49, + "learning_rate": 2.5147473711208004e-05, + "loss": 0.4324, + "step": 38810 + }, + { + "epoch": 1.49, + "learning_rate": 2.5146832521159276e-05, + "loss": 0.4841, + "step": 38811 + }, + { + "epoch": 1.49, + "learning_rate": 2.514619133111054e-05, + "loss": 0.3735, + "step": 38812 + }, + { + "epoch": 1.49, + "learning_rate": 2.5145550141061813e-05, + "loss": 0.3521, + "step": 38813 + }, + { + "epoch": 1.49, + "learning_rate": 2.514490895101308e-05, + "loss": 0.365, + "step": 38814 + }, + { + "epoch": 1.49, + "learning_rate": 2.5144267760964353e-05, + "loss": 0.345, + "step": 38815 + }, + { + "epoch": 1.49, + "learning_rate": 2.5143626570915618e-05, + "loss": 0.3653, + "step": 38816 + }, + { + "epoch": 1.49, + "learning_rate": 2.514298538086689e-05, + "loss": 0.2827, + "step": 38817 + }, + { + "epoch": 1.49, + "learning_rate": 2.514234419081816e-05, + "loss": 0.3964, + "step": 38818 + }, + { + "epoch": 1.49, + "learning_rate": 2.514170300076943e-05, + "loss": 0.2666, + "step": 38819 + }, + { + "epoch": 1.49, + "learning_rate": 2.51410618107207e-05, + "loss": 0.3882, + "step": 38820 + }, + { + "epoch": 1.49, + "learning_rate": 2.5140420620671966e-05, + "loss": 0.508, + "step": 38821 + }, + { + "epoch": 1.49, + "learning_rate": 2.5139779430623238e-05, + "loss": 0.3547, + "step": 38822 + }, + { + "epoch": 1.49, + "learning_rate": 2.5139138240574506e-05, + "loss": 0.2813, + "step": 38823 + }, + { + "epoch": 1.49, + "learning_rate": 2.5138497050525778e-05, + "loss": 0.3347, + "step": 38824 + }, + { + "epoch": 1.49, + "learning_rate": 2.5137855860477043e-05, + "loss": 0.2787, + "step": 38825 + }, + { + "epoch": 1.49, + "learning_rate": 2.5137214670428315e-05, + "loss": 0.4056, + "step": 38826 + }, + { + "epoch": 1.49, + "learning_rate": 2.513657348037959e-05, + "loss": 0.3572, + "step": 38827 + }, + { + "epoch": 1.49, + "learning_rate": 2.5135932290330855e-05, + "loss": 0.5438, + "step": 38828 + }, + { + "epoch": 1.49, + "learning_rate": 2.5135291100282126e-05, + "loss": 0.258, + "step": 38829 + }, + { + "epoch": 1.49, + "learning_rate": 2.5134649910233395e-05, + "loss": 0.4579, + "step": 38830 + }, + { + "epoch": 1.49, + "learning_rate": 2.5134008720184666e-05, + "loss": 0.3141, + "step": 38831 + }, + { + "epoch": 1.49, + "learning_rate": 2.513336753013593e-05, + "loss": 0.4666, + "step": 38832 + }, + { + "epoch": 1.49, + "learning_rate": 2.5132726340087203e-05, + "loss": 0.4902, + "step": 38833 + }, + { + "epoch": 1.49, + "learning_rate": 2.513208515003847e-05, + "loss": 0.4108, + "step": 38834 + }, + { + "epoch": 1.49, + "learning_rate": 2.5131443959989743e-05, + "loss": 0.3638, + "step": 38835 + }, + { + "epoch": 1.49, + "learning_rate": 2.5130802769941015e-05, + "loss": 0.3805, + "step": 38836 + }, + { + "epoch": 1.49, + "learning_rate": 2.513016157989228e-05, + "loss": 0.3766, + "step": 38837 + }, + { + "epoch": 1.49, + "learning_rate": 2.512952038984355e-05, + "loss": 0.3838, + "step": 38838 + }, + { + "epoch": 1.49, + "learning_rate": 2.512887919979482e-05, + "loss": 0.3085, + "step": 38839 + }, + { + "epoch": 1.49, + "learning_rate": 2.512823800974609e-05, + "loss": 0.3139, + "step": 38840 + }, + { + "epoch": 1.49, + "learning_rate": 2.5127596819697356e-05, + "loss": 0.4469, + "step": 38841 + }, + { + "epoch": 1.49, + "learning_rate": 2.5126955629648628e-05, + "loss": 0.4307, + "step": 38842 + }, + { + "epoch": 1.49, + "learning_rate": 2.5126314439599897e-05, + "loss": 0.3339, + "step": 38843 + }, + { + "epoch": 1.49, + "learning_rate": 2.5125673249551168e-05, + "loss": 0.3556, + "step": 38844 + }, + { + "epoch": 1.49, + "learning_rate": 2.512503205950244e-05, + "loss": 0.3465, + "step": 38845 + }, + { + "epoch": 1.49, + "learning_rate": 2.5124390869453708e-05, + "loss": 0.4307, + "step": 38846 + }, + { + "epoch": 1.49, + "learning_rate": 2.512374967940498e-05, + "loss": 0.3883, + "step": 38847 + }, + { + "epoch": 1.49, + "learning_rate": 2.5123108489356245e-05, + "loss": 0.4015, + "step": 38848 + }, + { + "epoch": 1.49, + "learning_rate": 2.5122467299307517e-05, + "loss": 0.5112, + "step": 38849 + }, + { + "epoch": 1.49, + "learning_rate": 2.5121826109258785e-05, + "loss": 0.4365, + "step": 38850 + }, + { + "epoch": 1.49, + "learning_rate": 2.5121184919210057e-05, + "loss": 0.4739, + "step": 38851 + }, + { + "epoch": 1.49, + "learning_rate": 2.512054372916132e-05, + "loss": 0.2908, + "step": 38852 + }, + { + "epoch": 1.49, + "learning_rate": 2.5119902539112593e-05, + "loss": 0.4798, + "step": 38853 + }, + { + "epoch": 1.49, + "learning_rate": 2.5119261349063865e-05, + "loss": 0.3425, + "step": 38854 + }, + { + "epoch": 1.49, + "learning_rate": 2.5118620159015133e-05, + "loss": 0.4033, + "step": 38855 + }, + { + "epoch": 1.49, + "learning_rate": 2.5117978968966405e-05, + "loss": 0.3997, + "step": 38856 + }, + { + "epoch": 1.49, + "learning_rate": 2.511733777891767e-05, + "loss": 0.3383, + "step": 38857 + }, + { + "epoch": 1.49, + "learning_rate": 2.5116696588868942e-05, + "loss": 0.4147, + "step": 38858 + }, + { + "epoch": 1.49, + "learning_rate": 2.511605539882021e-05, + "loss": 0.5346, + "step": 38859 + }, + { + "epoch": 1.49, + "learning_rate": 2.5115414208771482e-05, + "loss": 0.3738, + "step": 38860 + }, + { + "epoch": 1.49, + "learning_rate": 2.5114773018722747e-05, + "loss": 0.3241, + "step": 38861 + }, + { + "epoch": 1.49, + "learning_rate": 2.511413182867402e-05, + "loss": 0.3271, + "step": 38862 + }, + { + "epoch": 1.49, + "learning_rate": 2.5113490638625294e-05, + "loss": 0.3718, + "step": 38863 + }, + { + "epoch": 1.49, + "learning_rate": 2.511284944857656e-05, + "loss": 0.314, + "step": 38864 + }, + { + "epoch": 1.49, + "learning_rate": 2.511220825852783e-05, + "loss": 0.3658, + "step": 38865 + }, + { + "epoch": 1.49, + "learning_rate": 2.51115670684791e-05, + "loss": 0.2799, + "step": 38866 + }, + { + "epoch": 1.49, + "learning_rate": 2.511092587843037e-05, + "loss": 0.348, + "step": 38867 + }, + { + "epoch": 1.49, + "learning_rate": 2.5110284688381635e-05, + "loss": 0.4031, + "step": 38868 + }, + { + "epoch": 1.49, + "learning_rate": 2.5109643498332907e-05, + "loss": 0.3104, + "step": 38869 + }, + { + "epoch": 1.49, + "learning_rate": 2.5109002308284175e-05, + "loss": 0.3375, + "step": 38870 + }, + { + "epoch": 1.49, + "learning_rate": 2.5108361118235447e-05, + "loss": 0.3677, + "step": 38871 + }, + { + "epoch": 1.49, + "learning_rate": 2.510771992818672e-05, + "loss": 0.3879, + "step": 38872 + }, + { + "epoch": 1.49, + "learning_rate": 2.5107078738137984e-05, + "loss": 0.3338, + "step": 38873 + }, + { + "epoch": 1.49, + "learning_rate": 2.5106437548089255e-05, + "loss": 0.3543, + "step": 38874 + }, + { + "epoch": 1.49, + "learning_rate": 2.5105796358040524e-05, + "loss": 0.2937, + "step": 38875 + }, + { + "epoch": 1.49, + "learning_rate": 2.5105155167991795e-05, + "loss": 0.4459, + "step": 38876 + }, + { + "epoch": 1.49, + "learning_rate": 2.510451397794306e-05, + "loss": 0.305, + "step": 38877 + }, + { + "epoch": 1.49, + "learning_rate": 2.5103872787894332e-05, + "loss": 0.3784, + "step": 38878 + }, + { + "epoch": 1.49, + "learning_rate": 2.51032315978456e-05, + "loss": 0.4663, + "step": 38879 + }, + { + "epoch": 1.49, + "learning_rate": 2.5102590407796872e-05, + "loss": 0.4574, + "step": 38880 + }, + { + "epoch": 1.49, + "learning_rate": 2.5101949217748144e-05, + "loss": 0.3374, + "step": 38881 + }, + { + "epoch": 1.49, + "learning_rate": 2.5101308027699412e-05, + "loss": 0.4004, + "step": 38882 + }, + { + "epoch": 1.49, + "learning_rate": 2.5100666837650684e-05, + "loss": 0.3022, + "step": 38883 + }, + { + "epoch": 1.49, + "learning_rate": 2.510002564760195e-05, + "loss": 0.3038, + "step": 38884 + }, + { + "epoch": 1.5, + "learning_rate": 2.509938445755322e-05, + "loss": 0.27, + "step": 38885 + }, + { + "epoch": 1.5, + "learning_rate": 2.509874326750449e-05, + "loss": 0.3515, + "step": 38886 + }, + { + "epoch": 1.5, + "learning_rate": 2.509810207745576e-05, + "loss": 0.3738, + "step": 38887 + }, + { + "epoch": 1.5, + "learning_rate": 2.5097460887407026e-05, + "loss": 0.3256, + "step": 38888 + }, + { + "epoch": 1.5, + "learning_rate": 2.5096819697358297e-05, + "loss": 0.3782, + "step": 38889 + }, + { + "epoch": 1.5, + "learning_rate": 2.509617850730957e-05, + "loss": 0.2199, + "step": 38890 + }, + { + "epoch": 1.5, + "learning_rate": 2.5095537317260837e-05, + "loss": 0.4286, + "step": 38891 + }, + { + "epoch": 1.5, + "learning_rate": 2.509489612721211e-05, + "loss": 0.3601, + "step": 38892 + }, + { + "epoch": 1.5, + "learning_rate": 2.5094254937163374e-05, + "loss": 0.2944, + "step": 38893 + }, + { + "epoch": 1.5, + "learning_rate": 2.5093613747114646e-05, + "loss": 0.2966, + "step": 38894 + }, + { + "epoch": 1.5, + "learning_rate": 2.5092972557065914e-05, + "loss": 0.2553, + "step": 38895 + }, + { + "epoch": 1.5, + "learning_rate": 2.5092331367017186e-05, + "loss": 0.3811, + "step": 38896 + }, + { + "epoch": 1.5, + "learning_rate": 2.509169017696845e-05, + "loss": 0.4163, + "step": 38897 + }, + { + "epoch": 1.5, + "learning_rate": 2.5091048986919722e-05, + "loss": 0.3296, + "step": 38898 + }, + { + "epoch": 1.5, + "learning_rate": 2.5090407796870997e-05, + "loss": 0.3515, + "step": 38899 + }, + { + "epoch": 1.5, + "learning_rate": 2.5089766606822262e-05, + "loss": 0.535, + "step": 38900 + }, + { + "epoch": 1.5, + "learning_rate": 2.5089125416773534e-05, + "loss": 0.4034, + "step": 38901 + }, + { + "epoch": 1.5, + "learning_rate": 2.5088484226724802e-05, + "loss": 0.4517, + "step": 38902 + }, + { + "epoch": 1.5, + "learning_rate": 2.5087843036676074e-05, + "loss": 0.3791, + "step": 38903 + }, + { + "epoch": 1.5, + "learning_rate": 2.508720184662734e-05, + "loss": 0.4004, + "step": 38904 + }, + { + "epoch": 1.5, + "learning_rate": 2.508656065657861e-05, + "loss": 0.3344, + "step": 38905 + }, + { + "epoch": 1.5, + "learning_rate": 2.508591946652988e-05, + "loss": 0.4176, + "step": 38906 + }, + { + "epoch": 1.5, + "learning_rate": 2.508527827648115e-05, + "loss": 0.3604, + "step": 38907 + }, + { + "epoch": 1.5, + "learning_rate": 2.5084637086432423e-05, + "loss": 0.3214, + "step": 38908 + }, + { + "epoch": 1.5, + "learning_rate": 2.5083995896383688e-05, + "loss": 0.2905, + "step": 38909 + }, + { + "epoch": 1.5, + "learning_rate": 2.508335470633496e-05, + "loss": 0.275, + "step": 38910 + }, + { + "epoch": 1.5, + "learning_rate": 2.5082713516286228e-05, + "loss": 0.3087, + "step": 38911 + }, + { + "epoch": 1.5, + "learning_rate": 2.50820723262375e-05, + "loss": 0.3154, + "step": 38912 + }, + { + "epoch": 1.5, + "learning_rate": 2.5081431136188764e-05, + "loss": 0.3154, + "step": 38913 + }, + { + "epoch": 1.5, + "learning_rate": 2.5080789946140036e-05, + "loss": 0.3683, + "step": 38914 + }, + { + "epoch": 1.5, + "learning_rate": 2.5080148756091304e-05, + "loss": 0.3798, + "step": 38915 + }, + { + "epoch": 1.5, + "learning_rate": 2.5079507566042576e-05, + "loss": 0.4237, + "step": 38916 + }, + { + "epoch": 1.5, + "learning_rate": 2.5078866375993848e-05, + "loss": 0.3206, + "step": 38917 + }, + { + "epoch": 1.5, + "learning_rate": 2.5078225185945116e-05, + "loss": 0.3351, + "step": 38918 + }, + { + "epoch": 1.5, + "learning_rate": 2.5077583995896388e-05, + "loss": 0.4034, + "step": 38919 + }, + { + "epoch": 1.5, + "learning_rate": 2.5076942805847653e-05, + "loss": 0.2914, + "step": 38920 + }, + { + "epoch": 1.5, + "learning_rate": 2.5076301615798924e-05, + "loss": 0.2802, + "step": 38921 + }, + { + "epoch": 1.5, + "learning_rate": 2.5075660425750193e-05, + "loss": 0.3332, + "step": 38922 + }, + { + "epoch": 1.5, + "learning_rate": 2.5075019235701464e-05, + "loss": 0.3437, + "step": 38923 + }, + { + "epoch": 1.5, + "learning_rate": 2.507437804565273e-05, + "loss": 0.4042, + "step": 38924 + }, + { + "epoch": 1.5, + "learning_rate": 2.5073736855604e-05, + "loss": 0.444, + "step": 38925 + }, + { + "epoch": 1.5, + "learning_rate": 2.5073095665555273e-05, + "loss": 0.3643, + "step": 38926 + }, + { + "epoch": 1.5, + "learning_rate": 2.507245447550654e-05, + "loss": 0.3922, + "step": 38927 + }, + { + "epoch": 1.5, + "learning_rate": 2.5071813285457813e-05, + "loss": 0.3557, + "step": 38928 + }, + { + "epoch": 1.5, + "learning_rate": 2.5071172095409078e-05, + "loss": 0.3227, + "step": 38929 + }, + { + "epoch": 1.5, + "learning_rate": 2.507053090536035e-05, + "loss": 0.3289, + "step": 38930 + }, + { + "epoch": 1.5, + "learning_rate": 2.5069889715311618e-05, + "loss": 0.216, + "step": 38931 + }, + { + "epoch": 1.5, + "learning_rate": 2.506924852526289e-05, + "loss": 0.3527, + "step": 38932 + }, + { + "epoch": 1.5, + "learning_rate": 2.5068607335214155e-05, + "loss": 0.4795, + "step": 38933 + }, + { + "epoch": 1.5, + "learning_rate": 2.506796614516543e-05, + "loss": 0.2802, + "step": 38934 + }, + { + "epoch": 1.5, + "learning_rate": 2.50673249551167e-05, + "loss": 0.3609, + "step": 38935 + }, + { + "epoch": 1.5, + "learning_rate": 2.5066683765067966e-05, + "loss": 0.317, + "step": 38936 + }, + { + "epoch": 1.5, + "learning_rate": 2.5066042575019238e-05, + "loss": 0.4107, + "step": 38937 + }, + { + "epoch": 1.5, + "learning_rate": 2.5065401384970506e-05, + "loss": 0.3273, + "step": 38938 + }, + { + "epoch": 1.5, + "learning_rate": 2.5064760194921778e-05, + "loss": 0.3426, + "step": 38939 + }, + { + "epoch": 1.5, + "learning_rate": 2.5064119004873043e-05, + "loss": 0.4319, + "step": 38940 + }, + { + "epoch": 1.5, + "learning_rate": 2.5063477814824315e-05, + "loss": 0.3593, + "step": 38941 + }, + { + "epoch": 1.5, + "learning_rate": 2.5062836624775583e-05, + "loss": 0.3165, + "step": 38942 + }, + { + "epoch": 1.5, + "learning_rate": 2.5062195434726855e-05, + "loss": 0.3748, + "step": 38943 + }, + { + "epoch": 1.5, + "learning_rate": 2.5061554244678126e-05, + "loss": 0.3184, + "step": 38944 + }, + { + "epoch": 1.5, + "learning_rate": 2.506091305462939e-05, + "loss": 0.3261, + "step": 38945 + }, + { + "epoch": 1.5, + "learning_rate": 2.5060271864580663e-05, + "loss": 0.3444, + "step": 38946 + }, + { + "epoch": 1.5, + "learning_rate": 2.505963067453193e-05, + "loss": 0.4264, + "step": 38947 + }, + { + "epoch": 1.5, + "learning_rate": 2.5058989484483203e-05, + "loss": 0.3927, + "step": 38948 + }, + { + "epoch": 1.5, + "learning_rate": 2.5058348294434468e-05, + "loss": 0.2732, + "step": 38949 + }, + { + "epoch": 1.5, + "learning_rate": 2.505770710438574e-05, + "loss": 0.4017, + "step": 38950 + }, + { + "epoch": 1.5, + "learning_rate": 2.5057065914337015e-05, + "loss": 0.3577, + "step": 38951 + }, + { + "epoch": 1.5, + "learning_rate": 2.505642472428828e-05, + "loss": 0.3325, + "step": 38952 + }, + { + "epoch": 1.5, + "learning_rate": 2.505578353423955e-05, + "loss": 0.4258, + "step": 38953 + }, + { + "epoch": 1.5, + "learning_rate": 2.505514234419082e-05, + "loss": 0.3986, + "step": 38954 + }, + { + "epoch": 1.5, + "learning_rate": 2.505450115414209e-05, + "loss": 0.4671, + "step": 38955 + }, + { + "epoch": 1.5, + "learning_rate": 2.5053859964093357e-05, + "loss": 0.3372, + "step": 38956 + }, + { + "epoch": 1.5, + "learning_rate": 2.5053218774044628e-05, + "loss": 0.4342, + "step": 38957 + }, + { + "epoch": 1.5, + "learning_rate": 2.5052577583995897e-05, + "loss": 0.285, + "step": 38958 + }, + { + "epoch": 1.5, + "learning_rate": 2.505193639394717e-05, + "loss": 0.3178, + "step": 38959 + }, + { + "epoch": 1.5, + "learning_rate": 2.505129520389844e-05, + "loss": 0.4906, + "step": 38960 + }, + { + "epoch": 1.5, + "learning_rate": 2.5050654013849705e-05, + "loss": 0.3891, + "step": 38961 + }, + { + "epoch": 1.5, + "learning_rate": 2.5050012823800977e-05, + "loss": 0.4867, + "step": 38962 + }, + { + "epoch": 1.5, + "learning_rate": 2.5049371633752245e-05, + "loss": 0.2369, + "step": 38963 + }, + { + "epoch": 1.5, + "learning_rate": 2.5048730443703517e-05, + "loss": 0.535, + "step": 38964 + }, + { + "epoch": 1.5, + "learning_rate": 2.504808925365478e-05, + "loss": 0.378, + "step": 38965 + }, + { + "epoch": 1.5, + "learning_rate": 2.5047448063606053e-05, + "loss": 0.4416, + "step": 38966 + }, + { + "epoch": 1.5, + "learning_rate": 2.5046806873557322e-05, + "loss": 0.369, + "step": 38967 + }, + { + "epoch": 1.5, + "learning_rate": 2.5046165683508593e-05, + "loss": 0.3811, + "step": 38968 + }, + { + "epoch": 1.5, + "learning_rate": 2.5045524493459865e-05, + "loss": 0.2876, + "step": 38969 + }, + { + "epoch": 1.5, + "learning_rate": 2.5044883303411134e-05, + "loss": 0.3493, + "step": 38970 + }, + { + "epoch": 1.5, + "learning_rate": 2.5044242113362405e-05, + "loss": 0.3475, + "step": 38971 + }, + { + "epoch": 1.5, + "learning_rate": 2.504360092331367e-05, + "loss": 0.4025, + "step": 38972 + }, + { + "epoch": 1.5, + "learning_rate": 2.5042959733264942e-05, + "loss": 0.4493, + "step": 38973 + }, + { + "epoch": 1.5, + "learning_rate": 2.504231854321621e-05, + "loss": 0.4909, + "step": 38974 + }, + { + "epoch": 1.5, + "learning_rate": 2.5041677353167482e-05, + "loss": 0.4028, + "step": 38975 + }, + { + "epoch": 1.5, + "learning_rate": 2.5041036163118747e-05, + "loss": 0.3172, + "step": 38976 + }, + { + "epoch": 1.5, + "learning_rate": 2.504039497307002e-05, + "loss": 0.5114, + "step": 38977 + }, + { + "epoch": 1.5, + "learning_rate": 2.503975378302129e-05, + "loss": 0.3478, + "step": 38978 + }, + { + "epoch": 1.5, + "learning_rate": 2.503911259297256e-05, + "loss": 0.4325, + "step": 38979 + }, + { + "epoch": 1.5, + "learning_rate": 2.503847140292383e-05, + "loss": 0.4664, + "step": 38980 + }, + { + "epoch": 1.5, + "learning_rate": 2.5037830212875095e-05, + "loss": 0.3317, + "step": 38981 + }, + { + "epoch": 1.5, + "learning_rate": 2.5037189022826367e-05, + "loss": 0.4559, + "step": 38982 + }, + { + "epoch": 1.5, + "learning_rate": 2.5036547832777635e-05, + "loss": 0.4946, + "step": 38983 + }, + { + "epoch": 1.5, + "learning_rate": 2.5035906642728907e-05, + "loss": 0.4369, + "step": 38984 + }, + { + "epoch": 1.5, + "learning_rate": 2.5035265452680172e-05, + "loss": 0.5073, + "step": 38985 + }, + { + "epoch": 1.5, + "learning_rate": 2.5034624262631444e-05, + "loss": 0.4061, + "step": 38986 + }, + { + "epoch": 1.5, + "learning_rate": 2.503398307258272e-05, + "loss": 0.3866, + "step": 38987 + }, + { + "epoch": 1.5, + "learning_rate": 2.5033341882533984e-05, + "loss": 0.3383, + "step": 38988 + }, + { + "epoch": 1.5, + "learning_rate": 2.5032700692485255e-05, + "loss": 0.5289, + "step": 38989 + }, + { + "epoch": 1.5, + "learning_rate": 2.5032059502436524e-05, + "loss": 0.3404, + "step": 38990 + }, + { + "epoch": 1.5, + "learning_rate": 2.5031418312387796e-05, + "loss": 0.255, + "step": 38991 + }, + { + "epoch": 1.5, + "learning_rate": 2.503077712233906e-05, + "loss": 0.3765, + "step": 38992 + }, + { + "epoch": 1.5, + "learning_rate": 2.5030135932290332e-05, + "loss": 0.2851, + "step": 38993 + }, + { + "epoch": 1.5, + "learning_rate": 2.50294947422416e-05, + "loss": 0.4036, + "step": 38994 + }, + { + "epoch": 1.5, + "learning_rate": 2.5028853552192872e-05, + "loss": 0.4227, + "step": 38995 + }, + { + "epoch": 1.5, + "learning_rate": 2.5028212362144144e-05, + "loss": 0.3583, + "step": 38996 + }, + { + "epoch": 1.5, + "learning_rate": 2.502757117209541e-05, + "loss": 0.4771, + "step": 38997 + }, + { + "epoch": 1.5, + "learning_rate": 2.502692998204668e-05, + "loss": 0.4054, + "step": 38998 + }, + { + "epoch": 1.5, + "learning_rate": 2.502628879199795e-05, + "loss": 0.3562, + "step": 38999 + }, + { + "epoch": 1.5, + "learning_rate": 2.502564760194922e-05, + "loss": 0.4485, + "step": 39000 + }, + { + "epoch": 1.5, + "learning_rate": 2.5025006411900486e-05, + "loss": 0.4675, + "step": 39001 + }, + { + "epoch": 1.5, + "learning_rate": 2.5024365221851757e-05, + "loss": 0.348, + "step": 39002 + }, + { + "epoch": 1.5, + "learning_rate": 2.5023724031803026e-05, + "loss": 0.2887, + "step": 39003 + }, + { + "epoch": 1.5, + "learning_rate": 2.5023082841754297e-05, + "loss": 0.3761, + "step": 39004 + }, + { + "epoch": 1.5, + "learning_rate": 2.502244165170557e-05, + "loss": 0.31, + "step": 39005 + }, + { + "epoch": 1.5, + "learning_rate": 2.5021800461656837e-05, + "loss": 0.2178, + "step": 39006 + }, + { + "epoch": 1.5, + "learning_rate": 2.502115927160811e-05, + "loss": 0.2805, + "step": 39007 + }, + { + "epoch": 1.5, + "learning_rate": 2.5020518081559374e-05, + "loss": 0.4842, + "step": 39008 + }, + { + "epoch": 1.5, + "learning_rate": 2.5019876891510646e-05, + "loss": 0.433, + "step": 39009 + }, + { + "epoch": 1.5, + "learning_rate": 2.5019235701461914e-05, + "loss": 0.3301, + "step": 39010 + }, + { + "epoch": 1.5, + "learning_rate": 2.5018594511413186e-05, + "loss": 0.2574, + "step": 39011 + }, + { + "epoch": 1.5, + "learning_rate": 2.501795332136445e-05, + "loss": 0.2751, + "step": 39012 + }, + { + "epoch": 1.5, + "learning_rate": 2.5017312131315722e-05, + "loss": 0.3517, + "step": 39013 + }, + { + "epoch": 1.5, + "learning_rate": 2.5016670941266994e-05, + "loss": 0.2801, + "step": 39014 + }, + { + "epoch": 1.5, + "learning_rate": 2.5016029751218262e-05, + "loss": 0.3807, + "step": 39015 + }, + { + "epoch": 1.5, + "learning_rate": 2.5015388561169534e-05, + "loss": 0.3953, + "step": 39016 + }, + { + "epoch": 1.5, + "learning_rate": 2.50147473711208e-05, + "loss": 0.4395, + "step": 39017 + }, + { + "epoch": 1.5, + "learning_rate": 2.501410618107207e-05, + "loss": 0.2482, + "step": 39018 + }, + { + "epoch": 1.5, + "learning_rate": 2.501346499102334e-05, + "loss": 0.3933, + "step": 39019 + }, + { + "epoch": 1.5, + "learning_rate": 2.501282380097461e-05, + "loss": 0.3549, + "step": 39020 + }, + { + "epoch": 1.5, + "learning_rate": 2.5012182610925876e-05, + "loss": 0.3395, + "step": 39021 + }, + { + "epoch": 1.5, + "learning_rate": 2.5011541420877148e-05, + "loss": 0.3451, + "step": 39022 + }, + { + "epoch": 1.5, + "learning_rate": 2.5010900230828423e-05, + "loss": 0.279, + "step": 39023 + }, + { + "epoch": 1.5, + "learning_rate": 2.5010259040779688e-05, + "loss": 0.3359, + "step": 39024 + }, + { + "epoch": 1.5, + "learning_rate": 2.500961785073096e-05, + "loss": 0.3156, + "step": 39025 + }, + { + "epoch": 1.5, + "learning_rate": 2.5008976660682228e-05, + "loss": 0.2235, + "step": 39026 + }, + { + "epoch": 1.5, + "learning_rate": 2.50083354706335e-05, + "loss": 0.2722, + "step": 39027 + }, + { + "epoch": 1.5, + "learning_rate": 2.5007694280584764e-05, + "loss": 0.3508, + "step": 39028 + }, + { + "epoch": 1.5, + "learning_rate": 2.5007053090536036e-05, + "loss": 0.3882, + "step": 39029 + }, + { + "epoch": 1.5, + "learning_rate": 2.5006411900487304e-05, + "loss": 0.4274, + "step": 39030 + }, + { + "epoch": 1.5, + "learning_rate": 2.5005770710438576e-05, + "loss": 0.258, + "step": 39031 + }, + { + "epoch": 1.5, + "learning_rate": 2.5005129520389848e-05, + "loss": 0.4915, + "step": 39032 + }, + { + "epoch": 1.5, + "learning_rate": 2.5004488330341113e-05, + "loss": 0.3264, + "step": 39033 + }, + { + "epoch": 1.5, + "learning_rate": 2.5003847140292384e-05, + "loss": 0.4505, + "step": 39034 + }, + { + "epoch": 1.5, + "learning_rate": 2.5003205950243653e-05, + "loss": 0.3957, + "step": 39035 + }, + { + "epoch": 1.5, + "learning_rate": 2.5002564760194924e-05, + "loss": 0.3318, + "step": 39036 + }, + { + "epoch": 1.5, + "learning_rate": 2.500192357014619e-05, + "loss": 0.4126, + "step": 39037 + }, + { + "epoch": 1.5, + "learning_rate": 2.500128238009746e-05, + "loss": 0.3403, + "step": 39038 + }, + { + "epoch": 1.5, + "learning_rate": 2.500064119004873e-05, + "loss": 0.289, + "step": 39039 + }, + { + "epoch": 1.5, + "learning_rate": 2.5e-05, + "loss": 0.3745, + "step": 39040 + }, + { + "epoch": 1.5, + "learning_rate": 2.499935880995127e-05, + "loss": 0.3338, + "step": 39041 + }, + { + "epoch": 1.5, + "learning_rate": 2.499871761990254e-05, + "loss": 0.3818, + "step": 39042 + }, + { + "epoch": 1.5, + "learning_rate": 2.499807642985381e-05, + "loss": 0.2207, + "step": 39043 + }, + { + "epoch": 1.5, + "learning_rate": 2.4997435239805078e-05, + "loss": 0.4183, + "step": 39044 + }, + { + "epoch": 1.5, + "learning_rate": 2.499679404975635e-05, + "loss": 0.2753, + "step": 39045 + }, + { + "epoch": 1.5, + "learning_rate": 2.4996152859707618e-05, + "loss": 0.3934, + "step": 39046 + }, + { + "epoch": 1.5, + "learning_rate": 2.499551166965889e-05, + "loss": 0.2794, + "step": 39047 + }, + { + "epoch": 1.5, + "learning_rate": 2.4994870479610158e-05, + "loss": 0.3499, + "step": 39048 + }, + { + "epoch": 1.5, + "learning_rate": 2.4994229289561426e-05, + "loss": 0.3212, + "step": 39049 + }, + { + "epoch": 1.5, + "learning_rate": 2.4993588099512698e-05, + "loss": 0.2824, + "step": 39050 + }, + { + "epoch": 1.5, + "learning_rate": 2.4992946909463966e-05, + "loss": 0.4237, + "step": 39051 + }, + { + "epoch": 1.5, + "learning_rate": 2.4992305719415235e-05, + "loss": 0.3295, + "step": 39052 + }, + { + "epoch": 1.5, + "learning_rate": 2.4991664529366503e-05, + "loss": 0.3325, + "step": 39053 + }, + { + "epoch": 1.5, + "learning_rate": 2.4991023339317775e-05, + "loss": 0.2959, + "step": 39054 + }, + { + "epoch": 1.5, + "learning_rate": 2.4990382149269046e-05, + "loss": 0.4487, + "step": 39055 + }, + { + "epoch": 1.5, + "learning_rate": 2.4989740959220315e-05, + "loss": 0.4145, + "step": 39056 + }, + { + "epoch": 1.5, + "learning_rate": 2.4989099769171583e-05, + "loss": 0.3728, + "step": 39057 + }, + { + "epoch": 1.5, + "learning_rate": 2.4988458579122855e-05, + "loss": 0.3823, + "step": 39058 + }, + { + "epoch": 1.5, + "learning_rate": 2.4987817389074123e-05, + "loss": 0.3673, + "step": 39059 + }, + { + "epoch": 1.5, + "learning_rate": 2.498717619902539e-05, + "loss": 0.277, + "step": 39060 + }, + { + "epoch": 1.5, + "learning_rate": 2.498653500897666e-05, + "loss": 0.3484, + "step": 39061 + }, + { + "epoch": 1.5, + "learning_rate": 2.498589381892793e-05, + "loss": 0.4872, + "step": 39062 + }, + { + "epoch": 1.5, + "learning_rate": 2.4985252628879203e-05, + "loss": 0.2582, + "step": 39063 + }, + { + "epoch": 1.5, + "learning_rate": 2.498461143883047e-05, + "loss": 0.3743, + "step": 39064 + }, + { + "epoch": 1.5, + "learning_rate": 2.498397024878174e-05, + "loss": 0.4096, + "step": 39065 + }, + { + "epoch": 1.5, + "learning_rate": 2.4983329058733008e-05, + "loss": 0.3303, + "step": 39066 + }, + { + "epoch": 1.5, + "learning_rate": 2.498268786868428e-05, + "loss": 0.4013, + "step": 39067 + }, + { + "epoch": 1.5, + "learning_rate": 2.4982046678635548e-05, + "loss": 0.378, + "step": 39068 + }, + { + "epoch": 1.5, + "learning_rate": 2.4981405488586817e-05, + "loss": 0.2144, + "step": 39069 + }, + { + "epoch": 1.5, + "learning_rate": 2.498076429853809e-05, + "loss": 0.2545, + "step": 39070 + }, + { + "epoch": 1.5, + "learning_rate": 2.4980123108489357e-05, + "loss": 0.3556, + "step": 39071 + }, + { + "epoch": 1.5, + "learning_rate": 2.497948191844063e-05, + "loss": 0.3951, + "step": 39072 + }, + { + "epoch": 1.5, + "learning_rate": 2.4978840728391897e-05, + "loss": 0.3668, + "step": 39073 + }, + { + "epoch": 1.5, + "learning_rate": 2.4978199538343165e-05, + "loss": 0.354, + "step": 39074 + }, + { + "epoch": 1.5, + "learning_rate": 2.4977558348294437e-05, + "loss": 0.3325, + "step": 39075 + }, + { + "epoch": 1.5, + "learning_rate": 2.4976917158245705e-05, + "loss": 0.3315, + "step": 39076 + }, + { + "epoch": 1.5, + "learning_rate": 2.4976275968196973e-05, + "loss": 0.4851, + "step": 39077 + }, + { + "epoch": 1.5, + "learning_rate": 2.4975634778148245e-05, + "loss": 0.3963, + "step": 39078 + }, + { + "epoch": 1.5, + "learning_rate": 2.4974993588099513e-05, + "loss": 0.3647, + "step": 39079 + }, + { + "epoch": 1.5, + "learning_rate": 2.4974352398050782e-05, + "loss": 0.4926, + "step": 39080 + }, + { + "epoch": 1.5, + "learning_rate": 2.4973711208002053e-05, + "loss": 0.2824, + "step": 39081 + }, + { + "epoch": 1.5, + "learning_rate": 2.4973070017953322e-05, + "loss": 0.3897, + "step": 39082 + }, + { + "epoch": 1.5, + "learning_rate": 2.4972428827904594e-05, + "loss": 0.516, + "step": 39083 + }, + { + "epoch": 1.5, + "learning_rate": 2.4971787637855862e-05, + "loss": 0.4963, + "step": 39084 + }, + { + "epoch": 1.5, + "learning_rate": 2.497114644780713e-05, + "loss": 0.399, + "step": 39085 + }, + { + "epoch": 1.5, + "learning_rate": 2.4970505257758402e-05, + "loss": 0.2426, + "step": 39086 + }, + { + "epoch": 1.5, + "learning_rate": 2.496986406770967e-05, + "loss": 0.2878, + "step": 39087 + }, + { + "epoch": 1.5, + "learning_rate": 2.496922287766094e-05, + "loss": 0.3405, + "step": 39088 + }, + { + "epoch": 1.5, + "learning_rate": 2.496858168761221e-05, + "loss": 0.4363, + "step": 39089 + }, + { + "epoch": 1.5, + "learning_rate": 2.496794049756348e-05, + "loss": 0.2608, + "step": 39090 + }, + { + "epoch": 1.5, + "learning_rate": 2.496729930751475e-05, + "loss": 0.3034, + "step": 39091 + }, + { + "epoch": 1.5, + "learning_rate": 2.496665811746602e-05, + "loss": 0.5246, + "step": 39092 + }, + { + "epoch": 1.5, + "learning_rate": 2.4966016927417287e-05, + "loss": 0.3935, + "step": 39093 + }, + { + "epoch": 1.5, + "learning_rate": 2.496537573736856e-05, + "loss": 0.3722, + "step": 39094 + }, + { + "epoch": 1.5, + "learning_rate": 2.4964734547319827e-05, + "loss": 0.328, + "step": 39095 + }, + { + "epoch": 1.5, + "learning_rate": 2.4964093357271095e-05, + "loss": 0.359, + "step": 39096 + }, + { + "epoch": 1.5, + "learning_rate": 2.4963452167222364e-05, + "loss": 0.3156, + "step": 39097 + }, + { + "epoch": 1.5, + "learning_rate": 2.4962810977173635e-05, + "loss": 0.3973, + "step": 39098 + }, + { + "epoch": 1.5, + "learning_rate": 2.4962169787124907e-05, + "loss": 0.3592, + "step": 39099 + }, + { + "epoch": 1.5, + "learning_rate": 2.4961528597076175e-05, + "loss": 0.4654, + "step": 39100 + }, + { + "epoch": 1.5, + "learning_rate": 2.4960887407027444e-05, + "loss": 0.2872, + "step": 39101 + }, + { + "epoch": 1.5, + "learning_rate": 2.4960246216978712e-05, + "loss": 0.2464, + "step": 39102 + }, + { + "epoch": 1.5, + "learning_rate": 2.4959605026929984e-05, + "loss": 0.3331, + "step": 39103 + }, + { + "epoch": 1.5, + "learning_rate": 2.4958963836881252e-05, + "loss": 0.488, + "step": 39104 + }, + { + "epoch": 1.5, + "learning_rate": 2.495832264683252e-05, + "loss": 0.3822, + "step": 39105 + }, + { + "epoch": 1.5, + "learning_rate": 2.4957681456783792e-05, + "loss": 0.3009, + "step": 39106 + }, + { + "epoch": 1.5, + "learning_rate": 2.4957040266735064e-05, + "loss": 0.3587, + "step": 39107 + }, + { + "epoch": 1.5, + "learning_rate": 2.4956399076686332e-05, + "loss": 0.3291, + "step": 39108 + }, + { + "epoch": 1.5, + "learning_rate": 2.49557578866376e-05, + "loss": 0.3182, + "step": 39109 + }, + { + "epoch": 1.5, + "learning_rate": 2.495511669658887e-05, + "loss": 0.3388, + "step": 39110 + }, + { + "epoch": 1.5, + "learning_rate": 2.495447550654014e-05, + "loss": 0.2817, + "step": 39111 + }, + { + "epoch": 1.5, + "learning_rate": 2.495383431649141e-05, + "loss": 0.2556, + "step": 39112 + }, + { + "epoch": 1.5, + "learning_rate": 2.4953193126442677e-05, + "loss": 0.4838, + "step": 39113 + }, + { + "epoch": 1.5, + "learning_rate": 2.495255193639395e-05, + "loss": 0.464, + "step": 39114 + }, + { + "epoch": 1.5, + "learning_rate": 2.4951910746345217e-05, + "loss": 0.4035, + "step": 39115 + }, + { + "epoch": 1.5, + "learning_rate": 2.495126955629649e-05, + "loss": 0.2949, + "step": 39116 + }, + { + "epoch": 1.5, + "learning_rate": 2.4950628366247757e-05, + "loss": 0.2435, + "step": 39117 + }, + { + "epoch": 1.5, + "learning_rate": 2.4949987176199026e-05, + "loss": 0.4519, + "step": 39118 + }, + { + "epoch": 1.5, + "learning_rate": 2.4949345986150297e-05, + "loss": 0.357, + "step": 39119 + }, + { + "epoch": 1.5, + "learning_rate": 2.4948704796101566e-05, + "loss": 0.2918, + "step": 39120 + }, + { + "epoch": 1.5, + "learning_rate": 2.4948063606052834e-05, + "loss": 0.2681, + "step": 39121 + }, + { + "epoch": 1.5, + "learning_rate": 2.4947422416004106e-05, + "loss": 0.325, + "step": 39122 + }, + { + "epoch": 1.5, + "learning_rate": 2.4946781225955374e-05, + "loss": 0.2922, + "step": 39123 + }, + { + "epoch": 1.5, + "learning_rate": 2.4946140035906642e-05, + "loss": 0.3626, + "step": 39124 + }, + { + "epoch": 1.5, + "learning_rate": 2.4945498845857914e-05, + "loss": 0.3545, + "step": 39125 + }, + { + "epoch": 1.5, + "learning_rate": 2.4944857655809182e-05, + "loss": 0.3595, + "step": 39126 + }, + { + "epoch": 1.5, + "learning_rate": 2.4944216465760454e-05, + "loss": 0.3038, + "step": 39127 + }, + { + "epoch": 1.5, + "learning_rate": 2.4943575275711723e-05, + "loss": 0.3918, + "step": 39128 + }, + { + "epoch": 1.5, + "learning_rate": 2.494293408566299e-05, + "loss": 0.3485, + "step": 39129 + }, + { + "epoch": 1.5, + "learning_rate": 2.4942292895614263e-05, + "loss": 0.2871, + "step": 39130 + }, + { + "epoch": 1.5, + "learning_rate": 2.494165170556553e-05, + "loss": 0.2997, + "step": 39131 + }, + { + "epoch": 1.5, + "learning_rate": 2.49410105155168e-05, + "loss": 0.3358, + "step": 39132 + }, + { + "epoch": 1.5, + "learning_rate": 2.4940369325468068e-05, + "loss": 0.3298, + "step": 39133 + }, + { + "epoch": 1.5, + "learning_rate": 2.493972813541934e-05, + "loss": 0.3119, + "step": 39134 + }, + { + "epoch": 1.5, + "learning_rate": 2.493908694537061e-05, + "loss": 0.4045, + "step": 39135 + }, + { + "epoch": 1.5, + "learning_rate": 2.493844575532188e-05, + "loss": 0.3513, + "step": 39136 + }, + { + "epoch": 1.5, + "learning_rate": 2.4937804565273148e-05, + "loss": 0.4315, + "step": 39137 + }, + { + "epoch": 1.5, + "learning_rate": 2.493716337522442e-05, + "loss": 0.3287, + "step": 39138 + }, + { + "epoch": 1.5, + "learning_rate": 2.4936522185175688e-05, + "loss": 0.3415, + "step": 39139 + }, + { + "epoch": 1.5, + "learning_rate": 2.4935880995126956e-05, + "loss": 0.3741, + "step": 39140 + }, + { + "epoch": 1.5, + "learning_rate": 2.4935239805078224e-05, + "loss": 0.287, + "step": 39141 + }, + { + "epoch": 1.5, + "learning_rate": 2.4934598615029496e-05, + "loss": 0.3197, + "step": 39142 + }, + { + "epoch": 1.5, + "learning_rate": 2.4933957424980768e-05, + "loss": 0.3834, + "step": 39143 + }, + { + "epoch": 1.5, + "learning_rate": 2.4933316234932036e-05, + "loss": 0.3612, + "step": 39144 + }, + { + "epoch": 1.5, + "learning_rate": 2.4932675044883304e-05, + "loss": 0.3235, + "step": 39145 + }, + { + "epoch": 1.51, + "learning_rate": 2.4932033854834573e-05, + "loss": 0.4566, + "step": 39146 + }, + { + "epoch": 1.51, + "learning_rate": 2.4931392664785844e-05, + "loss": 0.3525, + "step": 39147 + }, + { + "epoch": 1.51, + "learning_rate": 2.4930751474737113e-05, + "loss": 0.4277, + "step": 39148 + }, + { + "epoch": 1.51, + "learning_rate": 2.493011028468838e-05, + "loss": 0.4372, + "step": 39149 + }, + { + "epoch": 1.51, + "learning_rate": 2.4929469094639653e-05, + "loss": 0.4072, + "step": 39150 + }, + { + "epoch": 1.51, + "learning_rate": 2.492882790459092e-05, + "loss": 0.3149, + "step": 39151 + }, + { + "epoch": 1.51, + "learning_rate": 2.4928186714542193e-05, + "loss": 0.3103, + "step": 39152 + }, + { + "epoch": 1.51, + "learning_rate": 2.492754552449346e-05, + "loss": 0.2479, + "step": 39153 + }, + { + "epoch": 1.51, + "learning_rate": 2.492690433444473e-05, + "loss": 0.3239, + "step": 39154 + }, + { + "epoch": 1.51, + "learning_rate": 2.4926263144396e-05, + "loss": 0.3711, + "step": 39155 + }, + { + "epoch": 1.51, + "learning_rate": 2.492562195434727e-05, + "loss": 0.545, + "step": 39156 + }, + { + "epoch": 1.51, + "learning_rate": 2.4924980764298538e-05, + "loss": 0.3457, + "step": 39157 + }, + { + "epoch": 1.51, + "learning_rate": 2.492433957424981e-05, + "loss": 0.4741, + "step": 39158 + }, + { + "epoch": 1.51, + "learning_rate": 2.4923698384201078e-05, + "loss": 0.342, + "step": 39159 + }, + { + "epoch": 1.51, + "learning_rate": 2.4923057194152346e-05, + "loss": 0.3549, + "step": 39160 + }, + { + "epoch": 1.51, + "learning_rate": 2.4922416004103618e-05, + "loss": 0.388, + "step": 39161 + }, + { + "epoch": 1.51, + "learning_rate": 2.4921774814054886e-05, + "loss": 0.3247, + "step": 39162 + }, + { + "epoch": 1.51, + "learning_rate": 2.4921133624006158e-05, + "loss": 0.4414, + "step": 39163 + }, + { + "epoch": 1.51, + "learning_rate": 2.4920492433957426e-05, + "loss": 0.4835, + "step": 39164 + }, + { + "epoch": 1.51, + "learning_rate": 2.4919851243908695e-05, + "loss": 0.2605, + "step": 39165 + }, + { + "epoch": 1.51, + "learning_rate": 2.4919210053859966e-05, + "loss": 0.3146, + "step": 39166 + }, + { + "epoch": 1.51, + "learning_rate": 2.4918568863811235e-05, + "loss": 0.2588, + "step": 39167 + }, + { + "epoch": 1.51, + "learning_rate": 2.4917927673762503e-05, + "loss": 0.3593, + "step": 39168 + }, + { + "epoch": 1.51, + "learning_rate": 2.491728648371377e-05, + "loss": 0.3352, + "step": 39169 + }, + { + "epoch": 1.51, + "learning_rate": 2.4916645293665043e-05, + "loss": 0.4335, + "step": 39170 + }, + { + "epoch": 1.51, + "learning_rate": 2.4916004103616315e-05, + "loss": 0.3774, + "step": 39171 + }, + { + "epoch": 1.51, + "learning_rate": 2.4915362913567583e-05, + "loss": 0.3551, + "step": 39172 + }, + { + "epoch": 1.51, + "learning_rate": 2.491472172351885e-05, + "loss": 0.3851, + "step": 39173 + }, + { + "epoch": 1.51, + "learning_rate": 2.4914080533470123e-05, + "loss": 0.3183, + "step": 39174 + }, + { + "epoch": 1.51, + "learning_rate": 2.491343934342139e-05, + "loss": 0.3922, + "step": 39175 + }, + { + "epoch": 1.51, + "learning_rate": 2.491279815337266e-05, + "loss": 0.3211, + "step": 39176 + }, + { + "epoch": 1.51, + "learning_rate": 2.4912156963323928e-05, + "loss": 0.3788, + "step": 39177 + }, + { + "epoch": 1.51, + "learning_rate": 2.49115157732752e-05, + "loss": 0.479, + "step": 39178 + }, + { + "epoch": 1.51, + "learning_rate": 2.491087458322647e-05, + "loss": 0.3391, + "step": 39179 + }, + { + "epoch": 1.51, + "learning_rate": 2.491023339317774e-05, + "loss": 0.2748, + "step": 39180 + }, + { + "epoch": 1.51, + "learning_rate": 2.490959220312901e-05, + "loss": 0.2857, + "step": 39181 + }, + { + "epoch": 1.51, + "learning_rate": 2.490895101308028e-05, + "loss": 0.4239, + "step": 39182 + }, + { + "epoch": 1.51, + "learning_rate": 2.490830982303155e-05, + "loss": 0.3877, + "step": 39183 + }, + { + "epoch": 1.51, + "learning_rate": 2.4907668632982817e-05, + "loss": 0.3481, + "step": 39184 + }, + { + "epoch": 1.51, + "learning_rate": 2.4907027442934085e-05, + "loss": 0.4841, + "step": 39185 + }, + { + "epoch": 1.51, + "learning_rate": 2.4906386252885357e-05, + "loss": 0.3304, + "step": 39186 + }, + { + "epoch": 1.51, + "learning_rate": 2.490574506283663e-05, + "loss": 0.2989, + "step": 39187 + }, + { + "epoch": 1.51, + "learning_rate": 2.4905103872787897e-05, + "loss": 0.3129, + "step": 39188 + }, + { + "epoch": 1.51, + "learning_rate": 2.4904462682739165e-05, + "loss": 0.5029, + "step": 39189 + }, + { + "epoch": 1.51, + "learning_rate": 2.4903821492690433e-05, + "loss": 0.5007, + "step": 39190 + }, + { + "epoch": 1.51, + "learning_rate": 2.4903180302641705e-05, + "loss": 0.4627, + "step": 39191 + }, + { + "epoch": 1.51, + "learning_rate": 2.4902539112592973e-05, + "loss": 0.2692, + "step": 39192 + }, + { + "epoch": 1.51, + "learning_rate": 2.4901897922544242e-05, + "loss": 0.4153, + "step": 39193 + }, + { + "epoch": 1.51, + "learning_rate": 2.4901256732495514e-05, + "loss": 0.4311, + "step": 39194 + }, + { + "epoch": 1.51, + "learning_rate": 2.4900615542446782e-05, + "loss": 0.3862, + "step": 39195 + }, + { + "epoch": 1.51, + "learning_rate": 2.4899974352398054e-05, + "loss": 0.3758, + "step": 39196 + }, + { + "epoch": 1.51, + "learning_rate": 2.4899333162349322e-05, + "loss": 0.439, + "step": 39197 + }, + { + "epoch": 1.51, + "learning_rate": 2.489869197230059e-05, + "loss": 0.398, + "step": 39198 + }, + { + "epoch": 1.51, + "learning_rate": 2.4898050782251862e-05, + "loss": 0.471, + "step": 39199 + }, + { + "epoch": 1.51, + "learning_rate": 2.489740959220313e-05, + "loss": 0.5067, + "step": 39200 + }, + { + "epoch": 1.51, + "learning_rate": 2.48967684021544e-05, + "loss": 0.2991, + "step": 39201 + }, + { + "epoch": 1.51, + "learning_rate": 2.489612721210567e-05, + "loss": 0.4267, + "step": 39202 + }, + { + "epoch": 1.51, + "learning_rate": 2.489548602205694e-05, + "loss": 0.4143, + "step": 39203 + }, + { + "epoch": 1.51, + "learning_rate": 2.4894844832008207e-05, + "loss": 0.3365, + "step": 39204 + }, + { + "epoch": 1.51, + "learning_rate": 2.489420364195948e-05, + "loss": 0.4676, + "step": 39205 + }, + { + "epoch": 1.51, + "learning_rate": 2.4893562451910747e-05, + "loss": 0.5291, + "step": 39206 + }, + { + "epoch": 1.51, + "learning_rate": 2.489292126186202e-05, + "loss": 0.2218, + "step": 39207 + }, + { + "epoch": 1.51, + "learning_rate": 2.4892280071813287e-05, + "loss": 0.2852, + "step": 39208 + }, + { + "epoch": 1.51, + "learning_rate": 2.4891638881764555e-05, + "loss": 0.4838, + "step": 39209 + }, + { + "epoch": 1.51, + "learning_rate": 2.4890997691715827e-05, + "loss": 0.4714, + "step": 39210 + }, + { + "epoch": 1.51, + "learning_rate": 2.4890356501667095e-05, + "loss": 0.1866, + "step": 39211 + }, + { + "epoch": 1.51, + "learning_rate": 2.4889715311618364e-05, + "loss": 0.3051, + "step": 39212 + }, + { + "epoch": 1.51, + "learning_rate": 2.4889074121569632e-05, + "loss": 0.4027, + "step": 39213 + }, + { + "epoch": 1.51, + "learning_rate": 2.4888432931520904e-05, + "loss": 0.3896, + "step": 39214 + }, + { + "epoch": 1.51, + "learning_rate": 2.4887791741472176e-05, + "loss": 0.4515, + "step": 39215 + }, + { + "epoch": 1.51, + "learning_rate": 2.4887150551423444e-05, + "loss": 0.3923, + "step": 39216 + }, + { + "epoch": 1.51, + "learning_rate": 2.4886509361374712e-05, + "loss": 0.4127, + "step": 39217 + }, + { + "epoch": 1.51, + "learning_rate": 2.4885868171325984e-05, + "loss": 0.3509, + "step": 39218 + }, + { + "epoch": 1.51, + "learning_rate": 2.4885226981277252e-05, + "loss": 0.4068, + "step": 39219 + }, + { + "epoch": 1.51, + "learning_rate": 2.488458579122852e-05, + "loss": 0.3392, + "step": 39220 + }, + { + "epoch": 1.51, + "learning_rate": 2.488394460117979e-05, + "loss": 0.3712, + "step": 39221 + }, + { + "epoch": 1.51, + "learning_rate": 2.488330341113106e-05, + "loss": 0.3129, + "step": 39222 + }, + { + "epoch": 1.51, + "learning_rate": 2.4882662221082332e-05, + "loss": 0.4057, + "step": 39223 + }, + { + "epoch": 1.51, + "learning_rate": 2.48820210310336e-05, + "loss": 0.2589, + "step": 39224 + }, + { + "epoch": 1.51, + "learning_rate": 2.488137984098487e-05, + "loss": 0.3371, + "step": 39225 + }, + { + "epoch": 1.51, + "learning_rate": 2.4880738650936137e-05, + "loss": 0.3893, + "step": 39226 + }, + { + "epoch": 1.51, + "learning_rate": 2.488009746088741e-05, + "loss": 0.4763, + "step": 39227 + }, + { + "epoch": 1.51, + "learning_rate": 2.4879456270838677e-05, + "loss": 0.3103, + "step": 39228 + }, + { + "epoch": 1.51, + "learning_rate": 2.4878815080789946e-05, + "loss": 0.3989, + "step": 39229 + }, + { + "epoch": 1.51, + "learning_rate": 2.4878173890741217e-05, + "loss": 0.4936, + "step": 39230 + }, + { + "epoch": 1.51, + "learning_rate": 2.4877532700692486e-05, + "loss": 0.3014, + "step": 39231 + }, + { + "epoch": 1.51, + "learning_rate": 2.4876891510643757e-05, + "loss": 0.2245, + "step": 39232 + }, + { + "epoch": 1.51, + "learning_rate": 2.4876250320595026e-05, + "loss": 0.3647, + "step": 39233 + }, + { + "epoch": 1.51, + "learning_rate": 2.4875609130546294e-05, + "loss": 0.3896, + "step": 39234 + }, + { + "epoch": 1.51, + "learning_rate": 2.4874967940497566e-05, + "loss": 0.477, + "step": 39235 + }, + { + "epoch": 1.51, + "learning_rate": 2.4874326750448834e-05, + "loss": 0.2339, + "step": 39236 + }, + { + "epoch": 1.51, + "learning_rate": 2.4873685560400102e-05, + "loss": 0.461, + "step": 39237 + }, + { + "epoch": 1.51, + "learning_rate": 2.4873044370351374e-05, + "loss": 0.3887, + "step": 39238 + }, + { + "epoch": 1.51, + "learning_rate": 2.4872403180302643e-05, + "loss": 0.4067, + "step": 39239 + }, + { + "epoch": 1.51, + "learning_rate": 2.487176199025391e-05, + "loss": 0.3381, + "step": 39240 + }, + { + "epoch": 1.51, + "learning_rate": 2.4871120800205183e-05, + "loss": 0.3949, + "step": 39241 + }, + { + "epoch": 1.51, + "learning_rate": 2.487047961015645e-05, + "loss": 0.3504, + "step": 39242 + }, + { + "epoch": 1.51, + "learning_rate": 2.4869838420107723e-05, + "loss": 0.373, + "step": 39243 + }, + { + "epoch": 1.51, + "learning_rate": 2.486919723005899e-05, + "loss": 0.4166, + "step": 39244 + }, + { + "epoch": 1.51, + "learning_rate": 2.486855604001026e-05, + "loss": 0.2725, + "step": 39245 + }, + { + "epoch": 1.51, + "learning_rate": 2.486791484996153e-05, + "loss": 0.243, + "step": 39246 + }, + { + "epoch": 1.51, + "learning_rate": 2.48672736599128e-05, + "loss": 0.3584, + "step": 39247 + }, + { + "epoch": 1.51, + "learning_rate": 2.4866632469864068e-05, + "loss": 0.3138, + "step": 39248 + }, + { + "epoch": 1.51, + "learning_rate": 2.4865991279815336e-05, + "loss": 0.4121, + "step": 39249 + }, + { + "epoch": 1.51, + "learning_rate": 2.4865350089766608e-05, + "loss": 0.3254, + "step": 39250 + }, + { + "epoch": 1.51, + "learning_rate": 2.486470889971788e-05, + "loss": 0.302, + "step": 39251 + }, + { + "epoch": 1.51, + "learning_rate": 2.4864067709669148e-05, + "loss": 0.2791, + "step": 39252 + }, + { + "epoch": 1.51, + "learning_rate": 2.4863426519620416e-05, + "loss": 0.3509, + "step": 39253 + }, + { + "epoch": 1.51, + "learning_rate": 2.4862785329571688e-05, + "loss": 0.3339, + "step": 39254 + }, + { + "epoch": 1.51, + "learning_rate": 2.4862144139522956e-05, + "loss": 0.3384, + "step": 39255 + }, + { + "epoch": 1.51, + "learning_rate": 2.4861502949474224e-05, + "loss": 0.3349, + "step": 39256 + }, + { + "epoch": 1.51, + "learning_rate": 2.4860861759425493e-05, + "loss": 0.3173, + "step": 39257 + }, + { + "epoch": 1.51, + "learning_rate": 2.4860220569376764e-05, + "loss": 0.3434, + "step": 39258 + }, + { + "epoch": 1.51, + "learning_rate": 2.4859579379328036e-05, + "loss": 0.3721, + "step": 39259 + }, + { + "epoch": 1.51, + "learning_rate": 2.4858938189279305e-05, + "loss": 0.3769, + "step": 39260 + }, + { + "epoch": 1.51, + "learning_rate": 2.4858296999230573e-05, + "loss": 0.3496, + "step": 39261 + }, + { + "epoch": 1.51, + "learning_rate": 2.4857655809181845e-05, + "loss": 0.2392, + "step": 39262 + }, + { + "epoch": 1.51, + "learning_rate": 2.4857014619133113e-05, + "loss": 0.3263, + "step": 39263 + }, + { + "epoch": 1.51, + "learning_rate": 2.485637342908438e-05, + "loss": 0.3971, + "step": 39264 + }, + { + "epoch": 1.51, + "learning_rate": 2.485573223903565e-05, + "loss": 0.3842, + "step": 39265 + }, + { + "epoch": 1.51, + "learning_rate": 2.485509104898692e-05, + "loss": 0.2197, + "step": 39266 + }, + { + "epoch": 1.51, + "learning_rate": 2.485444985893819e-05, + "loss": 0.3084, + "step": 39267 + }, + { + "epoch": 1.51, + "learning_rate": 2.485380866888946e-05, + "loss": 0.4097, + "step": 39268 + }, + { + "epoch": 1.51, + "learning_rate": 2.485316747884073e-05, + "loss": 0.3133, + "step": 39269 + }, + { + "epoch": 1.51, + "learning_rate": 2.4852526288791998e-05, + "loss": 0.3709, + "step": 39270 + }, + { + "epoch": 1.51, + "learning_rate": 2.485188509874327e-05, + "loss": 0.2869, + "step": 39271 + }, + { + "epoch": 1.51, + "learning_rate": 2.4851243908694538e-05, + "loss": 0.4385, + "step": 39272 + }, + { + "epoch": 1.51, + "learning_rate": 2.4850602718645806e-05, + "loss": 0.3466, + "step": 39273 + }, + { + "epoch": 1.51, + "learning_rate": 2.4849961528597078e-05, + "loss": 0.4155, + "step": 39274 + }, + { + "epoch": 1.51, + "learning_rate": 2.4849320338548346e-05, + "loss": 0.4948, + "step": 39275 + }, + { + "epoch": 1.51, + "learning_rate": 2.4848679148499615e-05, + "loss": 0.2965, + "step": 39276 + }, + { + "epoch": 1.51, + "learning_rate": 2.4848037958450886e-05, + "loss": 0.3601, + "step": 39277 + }, + { + "epoch": 1.51, + "learning_rate": 2.4847396768402155e-05, + "loss": 0.4582, + "step": 39278 + }, + { + "epoch": 1.51, + "learning_rate": 2.4846755578353426e-05, + "loss": 0.3715, + "step": 39279 + }, + { + "epoch": 1.51, + "learning_rate": 2.4846114388304695e-05, + "loss": 0.4778, + "step": 39280 + }, + { + "epoch": 1.51, + "learning_rate": 2.4845473198255963e-05, + "loss": 0.3951, + "step": 39281 + }, + { + "epoch": 1.51, + "learning_rate": 2.4844832008207235e-05, + "loss": 0.3575, + "step": 39282 + }, + { + "epoch": 1.51, + "learning_rate": 2.4844190818158503e-05, + "loss": 0.4068, + "step": 39283 + }, + { + "epoch": 1.51, + "learning_rate": 2.484354962810977e-05, + "loss": 0.3541, + "step": 39284 + }, + { + "epoch": 1.51, + "learning_rate": 2.4842908438061043e-05, + "loss": 0.3352, + "step": 39285 + }, + { + "epoch": 1.51, + "learning_rate": 2.484226724801231e-05, + "loss": 0.3689, + "step": 39286 + }, + { + "epoch": 1.51, + "learning_rate": 2.4841626057963583e-05, + "loss": 0.4207, + "step": 39287 + }, + { + "epoch": 1.51, + "learning_rate": 2.484098486791485e-05, + "loss": 0.3465, + "step": 39288 + }, + { + "epoch": 1.51, + "learning_rate": 2.484034367786612e-05, + "loss": 0.34, + "step": 39289 + }, + { + "epoch": 1.51, + "learning_rate": 2.483970248781739e-05, + "loss": 0.5496, + "step": 39290 + }, + { + "epoch": 1.51, + "learning_rate": 2.483906129776866e-05, + "loss": 0.3582, + "step": 39291 + }, + { + "epoch": 1.51, + "learning_rate": 2.4838420107719928e-05, + "loss": 0.3319, + "step": 39292 + }, + { + "epoch": 1.51, + "learning_rate": 2.4837778917671197e-05, + "loss": 0.3326, + "step": 39293 + }, + { + "epoch": 1.51, + "learning_rate": 2.483713772762247e-05, + "loss": 0.3838, + "step": 39294 + }, + { + "epoch": 1.51, + "learning_rate": 2.483649653757374e-05, + "loss": 0.3197, + "step": 39295 + }, + { + "epoch": 1.51, + "learning_rate": 2.483585534752501e-05, + "loss": 0.3896, + "step": 39296 + }, + { + "epoch": 1.51, + "learning_rate": 2.4835214157476277e-05, + "loss": 0.3912, + "step": 39297 + }, + { + "epoch": 1.51, + "learning_rate": 2.483457296742755e-05, + "loss": 0.24, + "step": 39298 + }, + { + "epoch": 1.51, + "learning_rate": 2.4833931777378817e-05, + "loss": 0.368, + "step": 39299 + }, + { + "epoch": 1.51, + "learning_rate": 2.4833290587330085e-05, + "loss": 0.4205, + "step": 39300 + }, + { + "epoch": 1.51, + "learning_rate": 2.4832649397281353e-05, + "loss": 0.3258, + "step": 39301 + }, + { + "epoch": 1.51, + "learning_rate": 2.4832008207232625e-05, + "loss": 0.3386, + "step": 39302 + }, + { + "epoch": 1.51, + "learning_rate": 2.4831367017183897e-05, + "loss": 0.3421, + "step": 39303 + }, + { + "epoch": 1.51, + "learning_rate": 2.4830725827135165e-05, + "loss": 0.3303, + "step": 39304 + }, + { + "epoch": 1.51, + "learning_rate": 2.4830084637086434e-05, + "loss": 0.4043, + "step": 39305 + }, + { + "epoch": 1.51, + "learning_rate": 2.4829443447037702e-05, + "loss": 0.2101, + "step": 39306 + }, + { + "epoch": 1.51, + "learning_rate": 2.4828802256988974e-05, + "loss": 0.295, + "step": 39307 + }, + { + "epoch": 1.51, + "learning_rate": 2.4828161066940242e-05, + "loss": 0.3655, + "step": 39308 + }, + { + "epoch": 1.51, + "learning_rate": 2.482751987689151e-05, + "loss": 0.4007, + "step": 39309 + }, + { + "epoch": 1.51, + "learning_rate": 2.4826878686842782e-05, + "loss": 0.2368, + "step": 39310 + }, + { + "epoch": 1.51, + "learning_rate": 2.482623749679405e-05, + "loss": 0.446, + "step": 39311 + }, + { + "epoch": 1.51, + "learning_rate": 2.4825596306745322e-05, + "loss": 0.3453, + "step": 39312 + }, + { + "epoch": 1.51, + "learning_rate": 2.482495511669659e-05, + "loss": 0.4135, + "step": 39313 + }, + { + "epoch": 1.51, + "learning_rate": 2.482431392664786e-05, + "loss": 0.4258, + "step": 39314 + }, + { + "epoch": 1.51, + "learning_rate": 2.482367273659913e-05, + "loss": 0.2414, + "step": 39315 + }, + { + "epoch": 1.51, + "learning_rate": 2.48230315465504e-05, + "loss": 0.3179, + "step": 39316 + }, + { + "epoch": 1.51, + "learning_rate": 2.4822390356501667e-05, + "loss": 0.4541, + "step": 39317 + }, + { + "epoch": 1.51, + "learning_rate": 2.482174916645294e-05, + "loss": 0.4181, + "step": 39318 + }, + { + "epoch": 1.51, + "learning_rate": 2.4821107976404207e-05, + "loss": 0.3268, + "step": 39319 + }, + { + "epoch": 1.51, + "learning_rate": 2.4820466786355475e-05, + "loss": 0.4103, + "step": 39320 + }, + { + "epoch": 1.51, + "learning_rate": 2.4819825596306747e-05, + "loss": 0.3419, + "step": 39321 + }, + { + "epoch": 1.51, + "learning_rate": 2.4819184406258015e-05, + "loss": 0.4172, + "step": 39322 + }, + { + "epoch": 1.51, + "learning_rate": 2.4818543216209287e-05, + "loss": 0.4933, + "step": 39323 + }, + { + "epoch": 1.51, + "learning_rate": 2.4817902026160555e-05, + "loss": 0.2204, + "step": 39324 + }, + { + "epoch": 1.51, + "learning_rate": 2.4817260836111824e-05, + "loss": 0.2705, + "step": 39325 + }, + { + "epoch": 1.51, + "learning_rate": 2.4816619646063096e-05, + "loss": 0.4954, + "step": 39326 + }, + { + "epoch": 1.51, + "learning_rate": 2.4815978456014364e-05, + "loss": 0.4483, + "step": 39327 + }, + { + "epoch": 1.51, + "learning_rate": 2.4815337265965632e-05, + "loss": 0.3786, + "step": 39328 + }, + { + "epoch": 1.51, + "learning_rate": 2.48146960759169e-05, + "loss": 0.3236, + "step": 39329 + }, + { + "epoch": 1.51, + "learning_rate": 2.4814054885868172e-05, + "loss": 0.273, + "step": 39330 + }, + { + "epoch": 1.51, + "learning_rate": 2.4813413695819444e-05, + "loss": 0.4869, + "step": 39331 + }, + { + "epoch": 1.51, + "learning_rate": 2.4812772505770712e-05, + "loss": 0.3878, + "step": 39332 + }, + { + "epoch": 1.51, + "learning_rate": 2.481213131572198e-05, + "loss": 0.2309, + "step": 39333 + }, + { + "epoch": 1.51, + "learning_rate": 2.4811490125673252e-05, + "loss": 0.3368, + "step": 39334 + }, + { + "epoch": 1.51, + "learning_rate": 2.481084893562452e-05, + "loss": 0.5108, + "step": 39335 + }, + { + "epoch": 1.51, + "learning_rate": 2.481020774557579e-05, + "loss": 0.4106, + "step": 39336 + }, + { + "epoch": 1.51, + "learning_rate": 2.4809566555527057e-05, + "loss": 0.4062, + "step": 39337 + }, + { + "epoch": 1.51, + "learning_rate": 2.480892536547833e-05, + "loss": 0.4196, + "step": 39338 + }, + { + "epoch": 1.51, + "learning_rate": 2.48082841754296e-05, + "loss": 0.2873, + "step": 39339 + }, + { + "epoch": 1.51, + "learning_rate": 2.480764298538087e-05, + "loss": 0.4068, + "step": 39340 + }, + { + "epoch": 1.51, + "learning_rate": 2.4807001795332137e-05, + "loss": 0.5766, + "step": 39341 + }, + { + "epoch": 1.51, + "learning_rate": 2.480636060528341e-05, + "loss": 0.4146, + "step": 39342 + }, + { + "epoch": 1.51, + "learning_rate": 2.4805719415234677e-05, + "loss": 0.373, + "step": 39343 + }, + { + "epoch": 1.51, + "learning_rate": 2.4805078225185946e-05, + "loss": 0.4319, + "step": 39344 + }, + { + "epoch": 1.51, + "learning_rate": 2.4804437035137214e-05, + "loss": 0.4161, + "step": 39345 + }, + { + "epoch": 1.51, + "learning_rate": 2.4803795845088486e-05, + "loss": 0.3212, + "step": 39346 + }, + { + "epoch": 1.51, + "learning_rate": 2.4803154655039754e-05, + "loss": 0.4838, + "step": 39347 + }, + { + "epoch": 1.51, + "learning_rate": 2.4802513464991026e-05, + "loss": 0.4603, + "step": 39348 + }, + { + "epoch": 1.51, + "learning_rate": 2.4801872274942294e-05, + "loss": 0.384, + "step": 39349 + }, + { + "epoch": 1.51, + "learning_rate": 2.4801231084893563e-05, + "loss": 0.3312, + "step": 39350 + }, + { + "epoch": 1.51, + "learning_rate": 2.4800589894844834e-05, + "loss": 0.3738, + "step": 39351 + }, + { + "epoch": 1.51, + "learning_rate": 2.4799948704796103e-05, + "loss": 0.3209, + "step": 39352 + }, + { + "epoch": 1.51, + "learning_rate": 2.479930751474737e-05, + "loss": 0.4169, + "step": 39353 + }, + { + "epoch": 1.51, + "learning_rate": 2.4798666324698643e-05, + "loss": 0.3165, + "step": 39354 + }, + { + "epoch": 1.51, + "learning_rate": 2.479802513464991e-05, + "loss": 0.2251, + "step": 39355 + }, + { + "epoch": 1.51, + "learning_rate": 2.479738394460118e-05, + "loss": 0.363, + "step": 39356 + }, + { + "epoch": 1.51, + "learning_rate": 2.479674275455245e-05, + "loss": 0.3904, + "step": 39357 + }, + { + "epoch": 1.51, + "learning_rate": 2.479610156450372e-05, + "loss": 0.3231, + "step": 39358 + }, + { + "epoch": 1.51, + "learning_rate": 2.479546037445499e-05, + "loss": 0.399, + "step": 39359 + }, + { + "epoch": 1.51, + "learning_rate": 2.479481918440626e-05, + "loss": 0.2768, + "step": 39360 + }, + { + "epoch": 1.51, + "learning_rate": 2.4794177994357528e-05, + "loss": 0.3128, + "step": 39361 + }, + { + "epoch": 1.51, + "learning_rate": 2.47935368043088e-05, + "loss": 0.3722, + "step": 39362 + }, + { + "epoch": 1.51, + "learning_rate": 2.4792895614260068e-05, + "loss": 0.3101, + "step": 39363 + }, + { + "epoch": 1.51, + "learning_rate": 2.4792254424211336e-05, + "loss": 0.3793, + "step": 39364 + }, + { + "epoch": 1.51, + "learning_rate": 2.4791613234162604e-05, + "loss": 0.4986, + "step": 39365 + }, + { + "epoch": 1.51, + "learning_rate": 2.4790972044113876e-05, + "loss": 0.2406, + "step": 39366 + }, + { + "epoch": 1.51, + "learning_rate": 2.4790330854065148e-05, + "loss": 0.4676, + "step": 39367 + }, + { + "epoch": 1.51, + "learning_rate": 2.4789689664016416e-05, + "loss": 0.3826, + "step": 39368 + }, + { + "epoch": 1.51, + "learning_rate": 2.4789048473967684e-05, + "loss": 0.2621, + "step": 39369 + }, + { + "epoch": 1.51, + "learning_rate": 2.4788407283918956e-05, + "loss": 0.382, + "step": 39370 + }, + { + "epoch": 1.51, + "learning_rate": 2.4787766093870225e-05, + "loss": 0.4463, + "step": 39371 + }, + { + "epoch": 1.51, + "learning_rate": 2.4787124903821493e-05, + "loss": 0.3896, + "step": 39372 + }, + { + "epoch": 1.51, + "learning_rate": 2.478648371377276e-05, + "loss": 0.2753, + "step": 39373 + }, + { + "epoch": 1.51, + "learning_rate": 2.4785842523724033e-05, + "loss": 0.3088, + "step": 39374 + }, + { + "epoch": 1.51, + "learning_rate": 2.4785201333675305e-05, + "loss": 0.3449, + "step": 39375 + }, + { + "epoch": 1.51, + "learning_rate": 2.4784560143626573e-05, + "loss": 0.2862, + "step": 39376 + }, + { + "epoch": 1.51, + "learning_rate": 2.478391895357784e-05, + "loss": 0.2827, + "step": 39377 + }, + { + "epoch": 1.51, + "learning_rate": 2.4783277763529113e-05, + "loss": 0.4035, + "step": 39378 + }, + { + "epoch": 1.51, + "learning_rate": 2.478263657348038e-05, + "loss": 0.4175, + "step": 39379 + }, + { + "epoch": 1.51, + "learning_rate": 2.478199538343165e-05, + "loss": 0.4747, + "step": 39380 + }, + { + "epoch": 1.51, + "learning_rate": 2.4781354193382918e-05, + "loss": 0.3302, + "step": 39381 + }, + { + "epoch": 1.51, + "learning_rate": 2.478071300333419e-05, + "loss": 0.231, + "step": 39382 + }, + { + "epoch": 1.51, + "learning_rate": 2.478007181328546e-05, + "loss": 0.2923, + "step": 39383 + }, + { + "epoch": 1.51, + "learning_rate": 2.477943062323673e-05, + "loss": 0.4103, + "step": 39384 + }, + { + "epoch": 1.51, + "learning_rate": 2.4778789433187998e-05, + "loss": 0.3889, + "step": 39385 + }, + { + "epoch": 1.51, + "learning_rate": 2.477814824313927e-05, + "loss": 0.5216, + "step": 39386 + }, + { + "epoch": 1.51, + "learning_rate": 2.4777507053090538e-05, + "loss": 0.3726, + "step": 39387 + }, + { + "epoch": 1.51, + "learning_rate": 2.4776865863041806e-05, + "loss": 0.3148, + "step": 39388 + }, + { + "epoch": 1.51, + "learning_rate": 2.4776224672993075e-05, + "loss": 0.3999, + "step": 39389 + }, + { + "epoch": 1.51, + "learning_rate": 2.4775583482944346e-05, + "loss": 0.3056, + "step": 39390 + }, + { + "epoch": 1.51, + "learning_rate": 2.4774942292895615e-05, + "loss": 0.2443, + "step": 39391 + }, + { + "epoch": 1.51, + "learning_rate": 2.4774301102846887e-05, + "loss": 0.3972, + "step": 39392 + }, + { + "epoch": 1.51, + "learning_rate": 2.4773659912798155e-05, + "loss": 0.2667, + "step": 39393 + }, + { + "epoch": 1.51, + "learning_rate": 2.4773018722749423e-05, + "loss": 0.3637, + "step": 39394 + }, + { + "epoch": 1.51, + "learning_rate": 2.4772377532700695e-05, + "loss": 0.3648, + "step": 39395 + }, + { + "epoch": 1.51, + "learning_rate": 2.4771736342651963e-05, + "loss": 0.3808, + "step": 39396 + }, + { + "epoch": 1.51, + "learning_rate": 2.477109515260323e-05, + "loss": 0.4042, + "step": 39397 + }, + { + "epoch": 1.51, + "learning_rate": 2.4770453962554503e-05, + "loss": 0.2995, + "step": 39398 + }, + { + "epoch": 1.51, + "learning_rate": 2.476981277250577e-05, + "loss": 0.3159, + "step": 39399 + }, + { + "epoch": 1.51, + "learning_rate": 2.476917158245704e-05, + "loss": 0.29, + "step": 39400 + }, + { + "epoch": 1.51, + "learning_rate": 2.476853039240831e-05, + "loss": 0.2506, + "step": 39401 + }, + { + "epoch": 1.51, + "learning_rate": 2.476788920235958e-05, + "loss": 0.3371, + "step": 39402 + }, + { + "epoch": 1.51, + "learning_rate": 2.476724801231085e-05, + "loss": 0.3276, + "step": 39403 + }, + { + "epoch": 1.51, + "learning_rate": 2.476660682226212e-05, + "loss": 0.2591, + "step": 39404 + }, + { + "epoch": 1.51, + "learning_rate": 2.476596563221339e-05, + "loss": 0.3922, + "step": 39405 + }, + { + "epoch": 1.52, + "learning_rate": 2.476532444216466e-05, + "loss": 0.3023, + "step": 39406 + }, + { + "epoch": 1.52, + "learning_rate": 2.476468325211593e-05, + "loss": 0.4656, + "step": 39407 + }, + { + "epoch": 1.52, + "learning_rate": 2.4764042062067197e-05, + "loss": 0.4619, + "step": 39408 + }, + { + "epoch": 1.52, + "learning_rate": 2.4763400872018465e-05, + "loss": 0.303, + "step": 39409 + }, + { + "epoch": 1.52, + "learning_rate": 2.4762759681969737e-05, + "loss": 0.2783, + "step": 39410 + }, + { + "epoch": 1.52, + "learning_rate": 2.476211849192101e-05, + "loss": 0.415, + "step": 39411 + }, + { + "epoch": 1.52, + "learning_rate": 2.4761477301872277e-05, + "loss": 0.3384, + "step": 39412 + }, + { + "epoch": 1.52, + "learning_rate": 2.4760836111823545e-05, + "loss": 0.2838, + "step": 39413 + }, + { + "epoch": 1.52, + "learning_rate": 2.4760194921774817e-05, + "loss": 0.3169, + "step": 39414 + }, + { + "epoch": 1.52, + "learning_rate": 2.4759553731726085e-05, + "loss": 0.3486, + "step": 39415 + }, + { + "epoch": 1.52, + "learning_rate": 2.4758912541677353e-05, + "loss": 0.4897, + "step": 39416 + }, + { + "epoch": 1.52, + "learning_rate": 2.4758271351628622e-05, + "loss": 0.2667, + "step": 39417 + }, + { + "epoch": 1.52, + "learning_rate": 2.4757630161579894e-05, + "loss": 0.4016, + "step": 39418 + }, + { + "epoch": 1.52, + "learning_rate": 2.4756988971531165e-05, + "loss": 0.2342, + "step": 39419 + }, + { + "epoch": 1.52, + "learning_rate": 2.4756347781482434e-05, + "loss": 0.2826, + "step": 39420 + }, + { + "epoch": 1.52, + "learning_rate": 2.4755706591433702e-05, + "loss": 0.3183, + "step": 39421 + }, + { + "epoch": 1.52, + "learning_rate": 2.4755065401384974e-05, + "loss": 0.3533, + "step": 39422 + }, + { + "epoch": 1.52, + "learning_rate": 2.4754424211336242e-05, + "loss": 0.32, + "step": 39423 + }, + { + "epoch": 1.52, + "learning_rate": 2.475378302128751e-05, + "loss": 0.4559, + "step": 39424 + }, + { + "epoch": 1.52, + "learning_rate": 2.475314183123878e-05, + "loss": 0.2543, + "step": 39425 + }, + { + "epoch": 1.52, + "learning_rate": 2.475250064119005e-05, + "loss": 0.5256, + "step": 39426 + }, + { + "epoch": 1.52, + "learning_rate": 2.475185945114132e-05, + "loss": 0.3972, + "step": 39427 + }, + { + "epoch": 1.52, + "learning_rate": 2.475121826109259e-05, + "loss": 0.4496, + "step": 39428 + }, + { + "epoch": 1.52, + "learning_rate": 2.475057707104386e-05, + "loss": 0.3566, + "step": 39429 + }, + { + "epoch": 1.52, + "learning_rate": 2.4749935880995127e-05, + "loss": 0.4034, + "step": 39430 + }, + { + "epoch": 1.52, + "learning_rate": 2.47492946909464e-05, + "loss": 0.3376, + "step": 39431 + }, + { + "epoch": 1.52, + "learning_rate": 2.4748653500897667e-05, + "loss": 0.3788, + "step": 39432 + }, + { + "epoch": 1.52, + "learning_rate": 2.4748012310848935e-05, + "loss": 0.3552, + "step": 39433 + }, + { + "epoch": 1.52, + "learning_rate": 2.4747371120800207e-05, + "loss": 0.2725, + "step": 39434 + }, + { + "epoch": 1.52, + "learning_rate": 2.4746729930751475e-05, + "loss": 0.4532, + "step": 39435 + }, + { + "epoch": 1.52, + "learning_rate": 2.4746088740702744e-05, + "loss": 0.4176, + "step": 39436 + }, + { + "epoch": 1.52, + "learning_rate": 2.4745447550654016e-05, + "loss": 0.3459, + "step": 39437 + }, + { + "epoch": 1.52, + "learning_rate": 2.4744806360605284e-05, + "loss": 0.3588, + "step": 39438 + }, + { + "epoch": 1.52, + "learning_rate": 2.4744165170556556e-05, + "loss": 0.4113, + "step": 39439 + }, + { + "epoch": 1.52, + "learning_rate": 2.4743523980507824e-05, + "loss": 0.3049, + "step": 39440 + }, + { + "epoch": 1.52, + "learning_rate": 2.4742882790459092e-05, + "loss": 0.232, + "step": 39441 + }, + { + "epoch": 1.52, + "learning_rate": 2.4742241600410364e-05, + "loss": 0.2875, + "step": 39442 + }, + { + "epoch": 1.52, + "learning_rate": 2.4741600410361632e-05, + "loss": 0.3886, + "step": 39443 + }, + { + "epoch": 1.52, + "learning_rate": 2.47409592203129e-05, + "loss": 0.3581, + "step": 39444 + }, + { + "epoch": 1.52, + "learning_rate": 2.474031803026417e-05, + "loss": 0.3289, + "step": 39445 + }, + { + "epoch": 1.52, + "learning_rate": 2.473967684021544e-05, + "loss": 0.2992, + "step": 39446 + }, + { + "epoch": 1.52, + "learning_rate": 2.4739035650166712e-05, + "loss": 0.3287, + "step": 39447 + }, + { + "epoch": 1.52, + "learning_rate": 2.473839446011798e-05, + "loss": 0.2563, + "step": 39448 + }, + { + "epoch": 1.52, + "learning_rate": 2.473775327006925e-05, + "loss": 0.4314, + "step": 39449 + }, + { + "epoch": 1.52, + "learning_rate": 2.473711208002052e-05, + "loss": 0.3604, + "step": 39450 + }, + { + "epoch": 1.52, + "learning_rate": 2.473647088997179e-05, + "loss": 0.3102, + "step": 39451 + }, + { + "epoch": 1.52, + "learning_rate": 2.4735829699923057e-05, + "loss": 0.4238, + "step": 39452 + }, + { + "epoch": 1.52, + "learning_rate": 2.4735188509874326e-05, + "loss": 0.3443, + "step": 39453 + }, + { + "epoch": 1.52, + "learning_rate": 2.4734547319825597e-05, + "loss": 0.4824, + "step": 39454 + }, + { + "epoch": 1.52, + "learning_rate": 2.473390612977687e-05, + "loss": 0.3357, + "step": 39455 + }, + { + "epoch": 1.52, + "learning_rate": 2.4733264939728137e-05, + "loss": 0.3097, + "step": 39456 + }, + { + "epoch": 1.52, + "learning_rate": 2.4732623749679406e-05, + "loss": 0.3081, + "step": 39457 + }, + { + "epoch": 1.52, + "learning_rate": 2.4731982559630678e-05, + "loss": 0.3093, + "step": 39458 + }, + { + "epoch": 1.52, + "learning_rate": 2.4731341369581946e-05, + "loss": 0.5134, + "step": 39459 + }, + { + "epoch": 1.52, + "learning_rate": 2.4730700179533214e-05, + "loss": 0.2762, + "step": 39460 + }, + { + "epoch": 1.52, + "learning_rate": 2.4730058989484482e-05, + "loss": 0.2928, + "step": 39461 + }, + { + "epoch": 1.52, + "learning_rate": 2.4729417799435754e-05, + "loss": 0.2871, + "step": 39462 + }, + { + "epoch": 1.52, + "learning_rate": 2.4728776609387023e-05, + "loss": 0.3531, + "step": 39463 + }, + { + "epoch": 1.52, + "learning_rate": 2.4728135419338294e-05, + "loss": 0.217, + "step": 39464 + }, + { + "epoch": 1.52, + "learning_rate": 2.4727494229289563e-05, + "loss": 0.2327, + "step": 39465 + }, + { + "epoch": 1.52, + "learning_rate": 2.4726853039240834e-05, + "loss": 0.3424, + "step": 39466 + }, + { + "epoch": 1.52, + "learning_rate": 2.4726211849192103e-05, + "loss": 0.3757, + "step": 39467 + }, + { + "epoch": 1.52, + "learning_rate": 2.472557065914337e-05, + "loss": 0.4591, + "step": 39468 + }, + { + "epoch": 1.52, + "learning_rate": 2.472492946909464e-05, + "loss": 0.3462, + "step": 39469 + }, + { + "epoch": 1.52, + "learning_rate": 2.472428827904591e-05, + "loss": 0.4246, + "step": 39470 + }, + { + "epoch": 1.52, + "learning_rate": 2.472364708899718e-05, + "loss": 0.3208, + "step": 39471 + }, + { + "epoch": 1.52, + "learning_rate": 2.472300589894845e-05, + "loss": 0.3181, + "step": 39472 + }, + { + "epoch": 1.52, + "learning_rate": 2.472236470889972e-05, + "loss": 0.4464, + "step": 39473 + }, + { + "epoch": 1.52, + "learning_rate": 2.4721723518850988e-05, + "loss": 0.3063, + "step": 39474 + }, + { + "epoch": 1.52, + "learning_rate": 2.472108232880226e-05, + "loss": 0.344, + "step": 39475 + }, + { + "epoch": 1.52, + "learning_rate": 2.4720441138753528e-05, + "loss": 0.4343, + "step": 39476 + }, + { + "epoch": 1.52, + "learning_rate": 2.4719799948704796e-05, + "loss": 0.4845, + "step": 39477 + }, + { + "epoch": 1.52, + "learning_rate": 2.4719158758656068e-05, + "loss": 0.322, + "step": 39478 + }, + { + "epoch": 1.52, + "learning_rate": 2.4718517568607336e-05, + "loss": 0.317, + "step": 39479 + }, + { + "epoch": 1.52, + "learning_rate": 2.4717876378558604e-05, + "loss": 0.3246, + "step": 39480 + }, + { + "epoch": 1.52, + "learning_rate": 2.4717235188509876e-05, + "loss": 0.3243, + "step": 39481 + }, + { + "epoch": 1.52, + "learning_rate": 2.4716593998461144e-05, + "loss": 0.2679, + "step": 39482 + }, + { + "epoch": 1.52, + "learning_rate": 2.4715952808412416e-05, + "loss": 0.2923, + "step": 39483 + }, + { + "epoch": 1.52, + "learning_rate": 2.4715311618363685e-05, + "loss": 0.4405, + "step": 39484 + }, + { + "epoch": 1.52, + "learning_rate": 2.4714670428314953e-05, + "loss": 0.2684, + "step": 39485 + }, + { + "epoch": 1.52, + "learning_rate": 2.4714029238266225e-05, + "loss": 0.5604, + "step": 39486 + }, + { + "epoch": 1.52, + "learning_rate": 2.4713388048217493e-05, + "loss": 0.4541, + "step": 39487 + }, + { + "epoch": 1.52, + "learning_rate": 2.471274685816876e-05, + "loss": 0.3208, + "step": 39488 + }, + { + "epoch": 1.52, + "learning_rate": 2.471210566812003e-05, + "loss": 0.2887, + "step": 39489 + }, + { + "epoch": 1.52, + "learning_rate": 2.47114644780713e-05, + "loss": 0.2957, + "step": 39490 + }, + { + "epoch": 1.52, + "learning_rate": 2.4710823288022573e-05, + "loss": 0.3573, + "step": 39491 + }, + { + "epoch": 1.52, + "learning_rate": 2.471018209797384e-05, + "loss": 0.3845, + "step": 39492 + }, + { + "epoch": 1.52, + "learning_rate": 2.470954090792511e-05, + "loss": 0.3377, + "step": 39493 + }, + { + "epoch": 1.52, + "learning_rate": 2.470889971787638e-05, + "loss": 0.4444, + "step": 39494 + }, + { + "epoch": 1.52, + "learning_rate": 2.470825852782765e-05, + "loss": 0.2679, + "step": 39495 + }, + { + "epoch": 1.52, + "learning_rate": 2.4707617337778918e-05, + "loss": 0.3793, + "step": 39496 + }, + { + "epoch": 1.52, + "learning_rate": 2.4706976147730186e-05, + "loss": 0.3305, + "step": 39497 + }, + { + "epoch": 1.52, + "learning_rate": 2.4706334957681458e-05, + "loss": 0.3458, + "step": 39498 + }, + { + "epoch": 1.52, + "learning_rate": 2.470569376763273e-05, + "loss": 0.4815, + "step": 39499 + }, + { + "epoch": 1.52, + "learning_rate": 2.4705052577583998e-05, + "loss": 0.3234, + "step": 39500 + }, + { + "epoch": 1.52, + "learning_rate": 2.4704411387535266e-05, + "loss": 0.2814, + "step": 39501 + }, + { + "epoch": 1.52, + "learning_rate": 2.4703770197486538e-05, + "loss": 0.4193, + "step": 39502 + }, + { + "epoch": 1.52, + "learning_rate": 2.4703129007437806e-05, + "loss": 0.3619, + "step": 39503 + }, + { + "epoch": 1.52, + "learning_rate": 2.4702487817389075e-05, + "loss": 0.3547, + "step": 39504 + }, + { + "epoch": 1.52, + "learning_rate": 2.4701846627340343e-05, + "loss": 0.3735, + "step": 39505 + }, + { + "epoch": 1.52, + "learning_rate": 2.4701205437291615e-05, + "loss": 0.4012, + "step": 39506 + }, + { + "epoch": 1.52, + "learning_rate": 2.4700564247242883e-05, + "loss": 0.3589, + "step": 39507 + }, + { + "epoch": 1.52, + "learning_rate": 2.4699923057194155e-05, + "loss": 0.3198, + "step": 39508 + }, + { + "epoch": 1.52, + "learning_rate": 2.4699281867145423e-05, + "loss": 0.3331, + "step": 39509 + }, + { + "epoch": 1.52, + "learning_rate": 2.4698640677096695e-05, + "loss": 0.5286, + "step": 39510 + }, + { + "epoch": 1.52, + "learning_rate": 2.4697999487047963e-05, + "loss": 0.342, + "step": 39511 + }, + { + "epoch": 1.52, + "learning_rate": 2.469735829699923e-05, + "loss": 0.3475, + "step": 39512 + }, + { + "epoch": 1.52, + "learning_rate": 2.46967171069505e-05, + "loss": 0.3819, + "step": 39513 + }, + { + "epoch": 1.52, + "learning_rate": 2.469607591690177e-05, + "loss": 0.3308, + "step": 39514 + }, + { + "epoch": 1.52, + "learning_rate": 2.469543472685304e-05, + "loss": 0.3711, + "step": 39515 + }, + { + "epoch": 1.52, + "learning_rate": 2.469479353680431e-05, + "loss": 0.3406, + "step": 39516 + }, + { + "epoch": 1.52, + "learning_rate": 2.469415234675558e-05, + "loss": 0.3997, + "step": 39517 + }, + { + "epoch": 1.52, + "learning_rate": 2.469351115670685e-05, + "loss": 0.3596, + "step": 39518 + }, + { + "epoch": 1.52, + "learning_rate": 2.469286996665812e-05, + "loss": 0.388, + "step": 39519 + }, + { + "epoch": 1.52, + "learning_rate": 2.469222877660939e-05, + "loss": 0.3305, + "step": 39520 + }, + { + "epoch": 1.52, + "learning_rate": 2.4691587586560657e-05, + "loss": 0.4015, + "step": 39521 + }, + { + "epoch": 1.52, + "learning_rate": 2.469094639651193e-05, + "loss": 0.3338, + "step": 39522 + }, + { + "epoch": 1.52, + "learning_rate": 2.4690305206463197e-05, + "loss": 0.3115, + "step": 39523 + }, + { + "epoch": 1.52, + "learning_rate": 2.4689664016414465e-05, + "loss": 0.4072, + "step": 39524 + }, + { + "epoch": 1.52, + "learning_rate": 2.4689022826365733e-05, + "loss": 0.3926, + "step": 39525 + }, + { + "epoch": 1.52, + "learning_rate": 2.4688381636317005e-05, + "loss": 0.2339, + "step": 39526 + }, + { + "epoch": 1.52, + "learning_rate": 2.4687740446268277e-05, + "loss": 0.3181, + "step": 39527 + }, + { + "epoch": 1.52, + "learning_rate": 2.4687099256219545e-05, + "loss": 0.467, + "step": 39528 + }, + { + "epoch": 1.52, + "learning_rate": 2.4686458066170814e-05, + "loss": 0.3398, + "step": 39529 + }, + { + "epoch": 1.52, + "learning_rate": 2.4685816876122085e-05, + "loss": 0.4042, + "step": 39530 + }, + { + "epoch": 1.52, + "learning_rate": 2.4685175686073354e-05, + "loss": 0.408, + "step": 39531 + }, + { + "epoch": 1.52, + "learning_rate": 2.4684534496024622e-05, + "loss": 0.3393, + "step": 39532 + }, + { + "epoch": 1.52, + "learning_rate": 2.468389330597589e-05, + "loss": 0.4898, + "step": 39533 + }, + { + "epoch": 1.52, + "learning_rate": 2.4683252115927162e-05, + "loss": 0.4472, + "step": 39534 + }, + { + "epoch": 1.52, + "learning_rate": 2.4682610925878434e-05, + "loss": 0.4097, + "step": 39535 + }, + { + "epoch": 1.52, + "learning_rate": 2.4681969735829702e-05, + "loss": 0.4012, + "step": 39536 + }, + { + "epoch": 1.52, + "learning_rate": 2.468132854578097e-05, + "loss": 0.4017, + "step": 39537 + }, + { + "epoch": 1.52, + "learning_rate": 2.4680687355732242e-05, + "loss": 0.3642, + "step": 39538 + }, + { + "epoch": 1.52, + "learning_rate": 2.468004616568351e-05, + "loss": 0.3378, + "step": 39539 + }, + { + "epoch": 1.52, + "learning_rate": 2.467940497563478e-05, + "loss": 0.3841, + "step": 39540 + }, + { + "epoch": 1.52, + "learning_rate": 2.4678763785586047e-05, + "loss": 0.4313, + "step": 39541 + }, + { + "epoch": 1.52, + "learning_rate": 2.467812259553732e-05, + "loss": 0.3887, + "step": 39542 + }, + { + "epoch": 1.52, + "learning_rate": 2.4677481405488587e-05, + "loss": 0.3309, + "step": 39543 + }, + { + "epoch": 1.52, + "learning_rate": 2.467684021543986e-05, + "loss": 0.2949, + "step": 39544 + }, + { + "epoch": 1.52, + "learning_rate": 2.4676199025391127e-05, + "loss": 0.3972, + "step": 39545 + }, + { + "epoch": 1.52, + "learning_rate": 2.46755578353424e-05, + "loss": 0.1834, + "step": 39546 + }, + { + "epoch": 1.52, + "learning_rate": 2.4674916645293667e-05, + "loss": 0.3977, + "step": 39547 + }, + { + "epoch": 1.52, + "learning_rate": 2.4674275455244935e-05, + "loss": 0.5051, + "step": 39548 + }, + { + "epoch": 1.52, + "learning_rate": 2.4673634265196204e-05, + "loss": 0.3203, + "step": 39549 + }, + { + "epoch": 1.52, + "learning_rate": 2.4672993075147476e-05, + "loss": 0.3229, + "step": 39550 + }, + { + "epoch": 1.52, + "learning_rate": 2.4672351885098744e-05, + "loss": 0.3423, + "step": 39551 + }, + { + "epoch": 1.52, + "learning_rate": 2.4671710695050012e-05, + "loss": 0.339, + "step": 39552 + }, + { + "epoch": 1.52, + "learning_rate": 2.4671069505001284e-05, + "loss": 0.3804, + "step": 39553 + }, + { + "epoch": 1.52, + "learning_rate": 2.4670428314952552e-05, + "loss": 0.3385, + "step": 39554 + }, + { + "epoch": 1.52, + "learning_rate": 2.4669787124903824e-05, + "loss": 0.3071, + "step": 39555 + }, + { + "epoch": 1.52, + "learning_rate": 2.4669145934855092e-05, + "loss": 0.4777, + "step": 39556 + }, + { + "epoch": 1.52, + "learning_rate": 2.466850474480636e-05, + "loss": 0.3411, + "step": 39557 + }, + { + "epoch": 1.52, + "learning_rate": 2.4667863554757632e-05, + "loss": 0.2906, + "step": 39558 + }, + { + "epoch": 1.52, + "learning_rate": 2.46672223647089e-05, + "loss": 0.2959, + "step": 39559 + }, + { + "epoch": 1.52, + "learning_rate": 2.466658117466017e-05, + "loss": 0.313, + "step": 39560 + }, + { + "epoch": 1.52, + "learning_rate": 2.466593998461144e-05, + "loss": 0.3758, + "step": 39561 + }, + { + "epoch": 1.52, + "learning_rate": 2.466529879456271e-05, + "loss": 0.3899, + "step": 39562 + }, + { + "epoch": 1.52, + "learning_rate": 2.466465760451398e-05, + "loss": 0.388, + "step": 39563 + }, + { + "epoch": 1.52, + "learning_rate": 2.466401641446525e-05, + "loss": 0.3185, + "step": 39564 + }, + { + "epoch": 1.52, + "learning_rate": 2.4663375224416517e-05, + "loss": 0.292, + "step": 39565 + }, + { + "epoch": 1.52, + "learning_rate": 2.466273403436779e-05, + "loss": 0.4402, + "step": 39566 + }, + { + "epoch": 1.52, + "learning_rate": 2.4662092844319057e-05, + "loss": 0.389, + "step": 39567 + }, + { + "epoch": 1.52, + "learning_rate": 2.4661451654270326e-05, + "loss": 0.3789, + "step": 39568 + }, + { + "epoch": 1.52, + "learning_rate": 2.4660810464221594e-05, + "loss": 0.4235, + "step": 39569 + }, + { + "epoch": 1.52, + "learning_rate": 2.4660169274172866e-05, + "loss": 0.501, + "step": 39570 + }, + { + "epoch": 1.52, + "learning_rate": 2.4659528084124138e-05, + "loss": 0.393, + "step": 39571 + }, + { + "epoch": 1.52, + "learning_rate": 2.4658886894075406e-05, + "loss": 0.3848, + "step": 39572 + }, + { + "epoch": 1.52, + "learning_rate": 2.4658245704026674e-05, + "loss": 0.3582, + "step": 39573 + }, + { + "epoch": 1.52, + "learning_rate": 2.4657604513977946e-05, + "loss": 0.511, + "step": 39574 + }, + { + "epoch": 1.52, + "learning_rate": 2.4656963323929214e-05, + "loss": 0.3337, + "step": 39575 + }, + { + "epoch": 1.52, + "learning_rate": 2.4656322133880483e-05, + "loss": 0.3923, + "step": 39576 + }, + { + "epoch": 1.52, + "learning_rate": 2.465568094383175e-05, + "loss": 0.3653, + "step": 39577 + }, + { + "epoch": 1.52, + "learning_rate": 2.4655039753783023e-05, + "loss": 0.2932, + "step": 39578 + }, + { + "epoch": 1.52, + "learning_rate": 2.4654398563734294e-05, + "loss": 0.3646, + "step": 39579 + }, + { + "epoch": 1.52, + "learning_rate": 2.4653757373685563e-05, + "loss": 0.3789, + "step": 39580 + }, + { + "epoch": 1.52, + "learning_rate": 2.465311618363683e-05, + "loss": 0.2607, + "step": 39581 + }, + { + "epoch": 1.52, + "learning_rate": 2.4652474993588103e-05, + "loss": 0.5092, + "step": 39582 + }, + { + "epoch": 1.52, + "learning_rate": 2.465183380353937e-05, + "loss": 0.4959, + "step": 39583 + }, + { + "epoch": 1.52, + "learning_rate": 2.465119261349064e-05, + "loss": 0.3001, + "step": 39584 + }, + { + "epoch": 1.52, + "learning_rate": 2.4650551423441908e-05, + "loss": 0.4486, + "step": 39585 + }, + { + "epoch": 1.52, + "learning_rate": 2.464991023339318e-05, + "loss": 0.3645, + "step": 39586 + }, + { + "epoch": 1.52, + "learning_rate": 2.4649269043344448e-05, + "loss": 0.3601, + "step": 39587 + }, + { + "epoch": 1.52, + "learning_rate": 2.464862785329572e-05, + "loss": 0.3527, + "step": 39588 + }, + { + "epoch": 1.52, + "learning_rate": 2.4647986663246988e-05, + "loss": 0.3912, + "step": 39589 + }, + { + "epoch": 1.52, + "learning_rate": 2.464734547319826e-05, + "loss": 0.2881, + "step": 39590 + }, + { + "epoch": 1.52, + "learning_rate": 2.4646704283149528e-05, + "loss": 0.3194, + "step": 39591 + }, + { + "epoch": 1.52, + "learning_rate": 2.4646063093100796e-05, + "loss": 0.335, + "step": 39592 + }, + { + "epoch": 1.52, + "learning_rate": 2.4645421903052064e-05, + "loss": 0.2905, + "step": 39593 + }, + { + "epoch": 1.52, + "learning_rate": 2.4644780713003336e-05, + "loss": 0.27, + "step": 39594 + }, + { + "epoch": 1.52, + "learning_rate": 2.4644139522954605e-05, + "loss": 0.362, + "step": 39595 + }, + { + "epoch": 1.52, + "learning_rate": 2.4643498332905873e-05, + "loss": 0.2573, + "step": 39596 + }, + { + "epoch": 1.52, + "learning_rate": 2.4642857142857145e-05, + "loss": 0.3246, + "step": 39597 + }, + { + "epoch": 1.52, + "learning_rate": 2.4642215952808413e-05, + "loss": 0.4972, + "step": 39598 + }, + { + "epoch": 1.52, + "learning_rate": 2.4641574762759685e-05, + "loss": 0.4624, + "step": 39599 + }, + { + "epoch": 1.52, + "learning_rate": 2.4640933572710953e-05, + "loss": 0.4082, + "step": 39600 + }, + { + "epoch": 1.52, + "learning_rate": 2.464029238266222e-05, + "loss": 0.3095, + "step": 39601 + }, + { + "epoch": 1.52, + "learning_rate": 2.4639651192613493e-05, + "loss": 0.3458, + "step": 39602 + }, + { + "epoch": 1.52, + "learning_rate": 2.463901000256476e-05, + "loss": 0.3567, + "step": 39603 + }, + { + "epoch": 1.52, + "learning_rate": 2.463836881251603e-05, + "loss": 0.2584, + "step": 39604 + }, + { + "epoch": 1.52, + "learning_rate": 2.4637727622467298e-05, + "loss": 0.2778, + "step": 39605 + }, + { + "epoch": 1.52, + "learning_rate": 2.463708643241857e-05, + "loss": 0.4001, + "step": 39606 + }, + { + "epoch": 1.52, + "learning_rate": 2.463644524236984e-05, + "loss": 0.2847, + "step": 39607 + }, + { + "epoch": 1.52, + "learning_rate": 2.463580405232111e-05, + "loss": 0.4045, + "step": 39608 + }, + { + "epoch": 1.52, + "learning_rate": 2.4635162862272378e-05, + "loss": 0.2822, + "step": 39609 + }, + { + "epoch": 1.52, + "learning_rate": 2.463452167222365e-05, + "loss": 0.4365, + "step": 39610 + }, + { + "epoch": 1.52, + "learning_rate": 2.4633880482174918e-05, + "loss": 0.4935, + "step": 39611 + }, + { + "epoch": 1.52, + "learning_rate": 2.4633239292126186e-05, + "loss": 0.5284, + "step": 39612 + }, + { + "epoch": 1.52, + "learning_rate": 2.4632598102077455e-05, + "loss": 0.3526, + "step": 39613 + }, + { + "epoch": 1.52, + "learning_rate": 2.4631956912028726e-05, + "loss": 0.3665, + "step": 39614 + }, + { + "epoch": 1.52, + "learning_rate": 2.4631315721979998e-05, + "loss": 0.304, + "step": 39615 + }, + { + "epoch": 1.52, + "learning_rate": 2.4630674531931267e-05, + "loss": 0.3556, + "step": 39616 + }, + { + "epoch": 1.52, + "learning_rate": 2.4630033341882535e-05, + "loss": 0.4822, + "step": 39617 + }, + { + "epoch": 1.52, + "learning_rate": 2.4629392151833807e-05, + "loss": 0.2602, + "step": 39618 + }, + { + "epoch": 1.52, + "learning_rate": 2.4628750961785075e-05, + "loss": 0.4143, + "step": 39619 + }, + { + "epoch": 1.52, + "learning_rate": 2.4628109771736343e-05, + "loss": 0.3783, + "step": 39620 + }, + { + "epoch": 1.52, + "learning_rate": 2.462746858168761e-05, + "loss": 0.2806, + "step": 39621 + }, + { + "epoch": 1.52, + "learning_rate": 2.4626827391638883e-05, + "loss": 0.3608, + "step": 39622 + }, + { + "epoch": 1.52, + "learning_rate": 2.462618620159015e-05, + "loss": 0.3629, + "step": 39623 + }, + { + "epoch": 1.52, + "learning_rate": 2.4625545011541423e-05, + "loss": 0.5506, + "step": 39624 + }, + { + "epoch": 1.52, + "learning_rate": 2.462490382149269e-05, + "loss": 0.2731, + "step": 39625 + }, + { + "epoch": 1.52, + "learning_rate": 2.4624262631443963e-05, + "loss": 0.3925, + "step": 39626 + }, + { + "epoch": 1.52, + "learning_rate": 2.4623621441395232e-05, + "loss": 0.2737, + "step": 39627 + }, + { + "epoch": 1.52, + "learning_rate": 2.46229802513465e-05, + "loss": 0.3817, + "step": 39628 + }, + { + "epoch": 1.52, + "learning_rate": 2.462233906129777e-05, + "loss": 0.3669, + "step": 39629 + }, + { + "epoch": 1.52, + "learning_rate": 2.462169787124904e-05, + "loss": 0.2928, + "step": 39630 + }, + { + "epoch": 1.52, + "learning_rate": 2.462105668120031e-05, + "loss": 0.3822, + "step": 39631 + }, + { + "epoch": 1.52, + "learning_rate": 2.4620415491151577e-05, + "loss": 0.3584, + "step": 39632 + }, + { + "epoch": 1.52, + "learning_rate": 2.461977430110285e-05, + "loss": 0.2644, + "step": 39633 + }, + { + "epoch": 1.52, + "learning_rate": 2.4619133111054117e-05, + "loss": 0.3697, + "step": 39634 + }, + { + "epoch": 1.52, + "learning_rate": 2.461849192100539e-05, + "loss": 0.4102, + "step": 39635 + }, + { + "epoch": 1.52, + "learning_rate": 2.4617850730956657e-05, + "loss": 0.3275, + "step": 39636 + }, + { + "epoch": 1.52, + "learning_rate": 2.4617209540907925e-05, + "loss": 0.322, + "step": 39637 + }, + { + "epoch": 1.52, + "learning_rate": 2.4616568350859197e-05, + "loss": 0.4015, + "step": 39638 + }, + { + "epoch": 1.52, + "learning_rate": 2.4615927160810465e-05, + "loss": 0.3145, + "step": 39639 + }, + { + "epoch": 1.52, + "learning_rate": 2.4615285970761734e-05, + "loss": 0.3284, + "step": 39640 + }, + { + "epoch": 1.52, + "learning_rate": 2.4614644780713002e-05, + "loss": 0.4836, + "step": 39641 + }, + { + "epoch": 1.52, + "learning_rate": 2.4614003590664274e-05, + "loss": 0.2593, + "step": 39642 + }, + { + "epoch": 1.52, + "learning_rate": 2.4613362400615545e-05, + "loss": 0.4292, + "step": 39643 + }, + { + "epoch": 1.52, + "learning_rate": 2.4612721210566814e-05, + "loss": 0.3163, + "step": 39644 + }, + { + "epoch": 1.52, + "learning_rate": 2.4612080020518082e-05, + "loss": 0.3384, + "step": 39645 + }, + { + "epoch": 1.52, + "learning_rate": 2.4611438830469354e-05, + "loss": 0.3874, + "step": 39646 + }, + { + "epoch": 1.52, + "learning_rate": 2.4610797640420622e-05, + "loss": 0.331, + "step": 39647 + }, + { + "epoch": 1.52, + "learning_rate": 2.461015645037189e-05, + "loss": 0.4135, + "step": 39648 + }, + { + "epoch": 1.52, + "learning_rate": 2.460951526032316e-05, + "loss": 0.2358, + "step": 39649 + }, + { + "epoch": 1.52, + "learning_rate": 2.460887407027443e-05, + "loss": 0.315, + "step": 39650 + }, + { + "epoch": 1.52, + "learning_rate": 2.4608232880225702e-05, + "loss": 0.4728, + "step": 39651 + }, + { + "epoch": 1.52, + "learning_rate": 2.460759169017697e-05, + "loss": 0.3386, + "step": 39652 + }, + { + "epoch": 1.52, + "learning_rate": 2.460695050012824e-05, + "loss": 0.429, + "step": 39653 + }, + { + "epoch": 1.52, + "learning_rate": 2.460630931007951e-05, + "loss": 0.3337, + "step": 39654 + }, + { + "epoch": 1.52, + "learning_rate": 2.460566812003078e-05, + "loss": 0.323, + "step": 39655 + }, + { + "epoch": 1.52, + "learning_rate": 2.4605026929982047e-05, + "loss": 0.419, + "step": 39656 + }, + { + "epoch": 1.52, + "learning_rate": 2.4604385739933315e-05, + "loss": 0.3651, + "step": 39657 + }, + { + "epoch": 1.52, + "learning_rate": 2.4603744549884587e-05, + "loss": 0.3425, + "step": 39658 + }, + { + "epoch": 1.52, + "learning_rate": 2.460310335983586e-05, + "loss": 0.415, + "step": 39659 + }, + { + "epoch": 1.52, + "learning_rate": 2.4602462169787127e-05, + "loss": 0.3334, + "step": 39660 + }, + { + "epoch": 1.52, + "learning_rate": 2.4601820979738396e-05, + "loss": 0.4139, + "step": 39661 + }, + { + "epoch": 1.52, + "learning_rate": 2.4601179789689667e-05, + "loss": 0.3076, + "step": 39662 + }, + { + "epoch": 1.52, + "learning_rate": 2.4600538599640936e-05, + "loss": 0.2385, + "step": 39663 + }, + { + "epoch": 1.52, + "learning_rate": 2.4599897409592204e-05, + "loss": 0.3059, + "step": 39664 + }, + { + "epoch": 1.52, + "learning_rate": 2.4599256219543472e-05, + "loss": 0.4273, + "step": 39665 + }, + { + "epoch": 1.53, + "learning_rate": 2.4598615029494744e-05, + "loss": 0.4554, + "step": 39666 + }, + { + "epoch": 1.53, + "learning_rate": 2.4597973839446012e-05, + "loss": 0.2567, + "step": 39667 + }, + { + "epoch": 1.53, + "learning_rate": 2.4597332649397284e-05, + "loss": 0.4092, + "step": 39668 + }, + { + "epoch": 1.53, + "learning_rate": 2.4596691459348552e-05, + "loss": 0.3249, + "step": 39669 + }, + { + "epoch": 1.53, + "learning_rate": 2.4596050269299824e-05, + "loss": 0.5027, + "step": 39670 + }, + { + "epoch": 1.53, + "learning_rate": 2.4595409079251092e-05, + "loss": 0.3218, + "step": 39671 + }, + { + "epoch": 1.53, + "learning_rate": 2.459476788920236e-05, + "loss": 0.42, + "step": 39672 + }, + { + "epoch": 1.53, + "learning_rate": 2.459412669915363e-05, + "loss": 0.405, + "step": 39673 + }, + { + "epoch": 1.53, + "learning_rate": 2.45934855091049e-05, + "loss": 0.4665, + "step": 39674 + }, + { + "epoch": 1.53, + "learning_rate": 2.459284431905617e-05, + "loss": 0.3149, + "step": 39675 + }, + { + "epoch": 1.53, + "learning_rate": 2.4592203129007437e-05, + "loss": 0.3225, + "step": 39676 + }, + { + "epoch": 1.53, + "learning_rate": 2.459156193895871e-05, + "loss": 0.3351, + "step": 39677 + }, + { + "epoch": 1.53, + "learning_rate": 2.4590920748909977e-05, + "loss": 0.3177, + "step": 39678 + }, + { + "epoch": 1.53, + "learning_rate": 2.459027955886125e-05, + "loss": 0.301, + "step": 39679 + }, + { + "epoch": 1.53, + "learning_rate": 2.4589638368812517e-05, + "loss": 0.4146, + "step": 39680 + }, + { + "epoch": 1.53, + "learning_rate": 2.4588997178763786e-05, + "loss": 0.6065, + "step": 39681 + }, + { + "epoch": 1.53, + "learning_rate": 2.4588355988715058e-05, + "loss": 0.323, + "step": 39682 + }, + { + "epoch": 1.53, + "learning_rate": 2.4587714798666326e-05, + "loss": 0.3878, + "step": 39683 + }, + { + "epoch": 1.53, + "learning_rate": 2.4587073608617594e-05, + "loss": 0.4607, + "step": 39684 + }, + { + "epoch": 1.53, + "learning_rate": 2.4586432418568863e-05, + "loss": 0.4492, + "step": 39685 + }, + { + "epoch": 1.53, + "learning_rate": 2.4585791228520134e-05, + "loss": 0.3612, + "step": 39686 + }, + { + "epoch": 1.53, + "learning_rate": 2.4585150038471406e-05, + "loss": 0.2218, + "step": 39687 + }, + { + "epoch": 1.53, + "learning_rate": 2.4584508848422674e-05, + "loss": 0.4275, + "step": 39688 + }, + { + "epoch": 1.53, + "learning_rate": 2.4583867658373943e-05, + "loss": 0.4049, + "step": 39689 + }, + { + "epoch": 1.53, + "learning_rate": 2.4583226468325214e-05, + "loss": 0.2919, + "step": 39690 + }, + { + "epoch": 1.53, + "learning_rate": 2.4582585278276483e-05, + "loss": 0.3535, + "step": 39691 + }, + { + "epoch": 1.53, + "learning_rate": 2.458194408822775e-05, + "loss": 0.3795, + "step": 39692 + }, + { + "epoch": 1.53, + "learning_rate": 2.458130289817902e-05, + "loss": 0.2763, + "step": 39693 + }, + { + "epoch": 1.53, + "learning_rate": 2.458066170813029e-05, + "loss": 0.2915, + "step": 39694 + }, + { + "epoch": 1.53, + "learning_rate": 2.4580020518081563e-05, + "loss": 0.2647, + "step": 39695 + }, + { + "epoch": 1.53, + "learning_rate": 2.457937932803283e-05, + "loss": 0.4058, + "step": 39696 + }, + { + "epoch": 1.53, + "learning_rate": 2.45787381379841e-05, + "loss": 0.3872, + "step": 39697 + }, + { + "epoch": 1.53, + "learning_rate": 2.457809694793537e-05, + "loss": 0.3555, + "step": 39698 + }, + { + "epoch": 1.53, + "learning_rate": 2.457745575788664e-05, + "loss": 0.3091, + "step": 39699 + }, + { + "epoch": 1.53, + "learning_rate": 2.4576814567837908e-05, + "loss": 0.3072, + "step": 39700 + }, + { + "epoch": 1.53, + "learning_rate": 2.4576173377789176e-05, + "loss": 0.4391, + "step": 39701 + }, + { + "epoch": 1.53, + "learning_rate": 2.4575532187740448e-05, + "loss": 0.3946, + "step": 39702 + }, + { + "epoch": 1.53, + "learning_rate": 2.4574890997691716e-05, + "loss": 0.429, + "step": 39703 + }, + { + "epoch": 1.53, + "learning_rate": 2.4574249807642988e-05, + "loss": 0.3359, + "step": 39704 + }, + { + "epoch": 1.53, + "learning_rate": 2.4573608617594256e-05, + "loss": 0.3132, + "step": 39705 + }, + { + "epoch": 1.53, + "learning_rate": 2.4572967427545528e-05, + "loss": 0.327, + "step": 39706 + }, + { + "epoch": 1.53, + "learning_rate": 2.4572326237496796e-05, + "loss": 0.383, + "step": 39707 + }, + { + "epoch": 1.53, + "learning_rate": 2.4571685047448065e-05, + "loss": 0.3635, + "step": 39708 + }, + { + "epoch": 1.53, + "learning_rate": 2.4571043857399333e-05, + "loss": 0.3506, + "step": 39709 + }, + { + "epoch": 1.53, + "learning_rate": 2.4570402667350605e-05, + "loss": 0.3134, + "step": 39710 + }, + { + "epoch": 1.53, + "learning_rate": 2.4569761477301873e-05, + "loss": 0.2727, + "step": 39711 + }, + { + "epoch": 1.53, + "learning_rate": 2.456912028725314e-05, + "loss": 0.2569, + "step": 39712 + }, + { + "epoch": 1.53, + "learning_rate": 2.4568479097204413e-05, + "loss": 0.4222, + "step": 39713 + }, + { + "epoch": 1.53, + "learning_rate": 2.4567837907155685e-05, + "loss": 0.4349, + "step": 39714 + }, + { + "epoch": 1.53, + "learning_rate": 2.4567196717106953e-05, + "loss": 0.4383, + "step": 39715 + }, + { + "epoch": 1.53, + "learning_rate": 2.456655552705822e-05, + "loss": 0.5068, + "step": 39716 + }, + { + "epoch": 1.53, + "learning_rate": 2.456591433700949e-05, + "loss": 0.3105, + "step": 39717 + }, + { + "epoch": 1.53, + "learning_rate": 2.456527314696076e-05, + "loss": 0.2857, + "step": 39718 + }, + { + "epoch": 1.53, + "learning_rate": 2.456463195691203e-05, + "loss": 0.3416, + "step": 39719 + }, + { + "epoch": 1.53, + "learning_rate": 2.4563990766863298e-05, + "loss": 0.3443, + "step": 39720 + }, + { + "epoch": 1.53, + "learning_rate": 2.4563349576814566e-05, + "loss": 0.3157, + "step": 39721 + }, + { + "epoch": 1.53, + "learning_rate": 2.4562708386765838e-05, + "loss": 0.4383, + "step": 39722 + }, + { + "epoch": 1.53, + "learning_rate": 2.456206719671711e-05, + "loss": 0.3415, + "step": 39723 + }, + { + "epoch": 1.53, + "learning_rate": 2.4561426006668378e-05, + "loss": 0.4953, + "step": 39724 + }, + { + "epoch": 1.53, + "learning_rate": 2.4560784816619646e-05, + "loss": 0.316, + "step": 39725 + }, + { + "epoch": 1.53, + "learning_rate": 2.4560143626570918e-05, + "loss": 0.4024, + "step": 39726 + }, + { + "epoch": 1.53, + "learning_rate": 2.4559502436522187e-05, + "loss": 0.3032, + "step": 39727 + }, + { + "epoch": 1.53, + "learning_rate": 2.4558861246473455e-05, + "loss": 0.4232, + "step": 39728 + }, + { + "epoch": 1.53, + "learning_rate": 2.4558220056424723e-05, + "loss": 0.3862, + "step": 39729 + }, + { + "epoch": 1.53, + "learning_rate": 2.4557578866375995e-05, + "loss": 0.3569, + "step": 39730 + }, + { + "epoch": 1.53, + "learning_rate": 2.4556937676327267e-05, + "loss": 0.3005, + "step": 39731 + }, + { + "epoch": 1.53, + "learning_rate": 2.4556296486278535e-05, + "loss": 0.2146, + "step": 39732 + }, + { + "epoch": 1.53, + "learning_rate": 2.4555655296229803e-05, + "loss": 0.2724, + "step": 39733 + }, + { + "epoch": 1.53, + "learning_rate": 2.4555014106181075e-05, + "loss": 0.2666, + "step": 39734 + }, + { + "epoch": 1.53, + "learning_rate": 2.4554372916132343e-05, + "loss": 0.2552, + "step": 39735 + }, + { + "epoch": 1.53, + "learning_rate": 2.455373172608361e-05, + "loss": 0.3428, + "step": 39736 + }, + { + "epoch": 1.53, + "learning_rate": 2.455309053603488e-05, + "loss": 0.4973, + "step": 39737 + }, + { + "epoch": 1.53, + "learning_rate": 2.455244934598615e-05, + "loss": 0.3769, + "step": 39738 + }, + { + "epoch": 1.53, + "learning_rate": 2.455180815593742e-05, + "loss": 0.2818, + "step": 39739 + }, + { + "epoch": 1.53, + "learning_rate": 2.4551166965888692e-05, + "loss": 0.4791, + "step": 39740 + }, + { + "epoch": 1.53, + "learning_rate": 2.455052577583996e-05, + "loss": 0.3711, + "step": 39741 + }, + { + "epoch": 1.53, + "learning_rate": 2.4549884585791232e-05, + "loss": 0.3523, + "step": 39742 + }, + { + "epoch": 1.53, + "learning_rate": 2.45492433957425e-05, + "loss": 0.4521, + "step": 39743 + }, + { + "epoch": 1.53, + "learning_rate": 2.454860220569377e-05, + "loss": 0.3408, + "step": 39744 + }, + { + "epoch": 1.53, + "learning_rate": 2.4547961015645037e-05, + "loss": 0.2903, + "step": 39745 + }, + { + "epoch": 1.53, + "learning_rate": 2.454731982559631e-05, + "loss": 0.4744, + "step": 39746 + }, + { + "epoch": 1.53, + "learning_rate": 2.4546678635547577e-05, + "loss": 0.3147, + "step": 39747 + }, + { + "epoch": 1.53, + "learning_rate": 2.4546037445498845e-05, + "loss": 0.2739, + "step": 39748 + }, + { + "epoch": 1.53, + "learning_rate": 2.4545396255450117e-05, + "loss": 0.352, + "step": 39749 + }, + { + "epoch": 1.53, + "learning_rate": 2.454475506540139e-05, + "loss": 0.3421, + "step": 39750 + }, + { + "epoch": 1.53, + "learning_rate": 2.4544113875352657e-05, + "loss": 0.2875, + "step": 39751 + }, + { + "epoch": 1.53, + "learning_rate": 2.4543472685303925e-05, + "loss": 0.3128, + "step": 39752 + }, + { + "epoch": 1.53, + "learning_rate": 2.4542831495255194e-05, + "loss": 0.3974, + "step": 39753 + }, + { + "epoch": 1.53, + "learning_rate": 2.4542190305206465e-05, + "loss": 0.323, + "step": 39754 + }, + { + "epoch": 1.53, + "learning_rate": 2.4541549115157734e-05, + "loss": 0.2825, + "step": 39755 + }, + { + "epoch": 1.53, + "learning_rate": 2.4540907925109002e-05, + "loss": 0.349, + "step": 39756 + }, + { + "epoch": 1.53, + "learning_rate": 2.4540266735060274e-05, + "loss": 0.3401, + "step": 39757 + }, + { + "epoch": 1.53, + "learning_rate": 2.4539625545011542e-05, + "loss": 0.3401, + "step": 39758 + }, + { + "epoch": 1.53, + "learning_rate": 2.4538984354962814e-05, + "loss": 0.387, + "step": 39759 + }, + { + "epoch": 1.53, + "learning_rate": 2.4538343164914082e-05, + "loss": 0.306, + "step": 39760 + }, + { + "epoch": 1.53, + "learning_rate": 2.453770197486535e-05, + "loss": 0.3276, + "step": 39761 + }, + { + "epoch": 1.53, + "learning_rate": 2.4537060784816622e-05, + "loss": 0.3493, + "step": 39762 + }, + { + "epoch": 1.53, + "learning_rate": 2.453641959476789e-05, + "loss": 0.434, + "step": 39763 + }, + { + "epoch": 1.53, + "learning_rate": 2.453577840471916e-05, + "loss": 0.3394, + "step": 39764 + }, + { + "epoch": 1.53, + "learning_rate": 2.4535137214670427e-05, + "loss": 0.439, + "step": 39765 + }, + { + "epoch": 1.53, + "learning_rate": 2.45344960246217e-05, + "loss": 0.381, + "step": 39766 + }, + { + "epoch": 1.53, + "learning_rate": 2.453385483457297e-05, + "loss": 0.348, + "step": 39767 + }, + { + "epoch": 1.53, + "learning_rate": 2.453321364452424e-05, + "loss": 0.3549, + "step": 39768 + }, + { + "epoch": 1.53, + "learning_rate": 2.4532572454475507e-05, + "loss": 0.3535, + "step": 39769 + }, + { + "epoch": 1.53, + "learning_rate": 2.453193126442678e-05, + "loss": 0.4383, + "step": 39770 + }, + { + "epoch": 1.53, + "learning_rate": 2.4531290074378047e-05, + "loss": 0.4094, + "step": 39771 + }, + { + "epoch": 1.53, + "learning_rate": 2.4530648884329316e-05, + "loss": 0.4033, + "step": 39772 + }, + { + "epoch": 1.53, + "learning_rate": 2.4530007694280584e-05, + "loss": 0.4161, + "step": 39773 + }, + { + "epoch": 1.53, + "learning_rate": 2.4529366504231856e-05, + "loss": 0.3036, + "step": 39774 + }, + { + "epoch": 1.53, + "learning_rate": 2.4528725314183127e-05, + "loss": 0.4604, + "step": 39775 + }, + { + "epoch": 1.53, + "learning_rate": 2.4528084124134396e-05, + "loss": 0.4154, + "step": 39776 + }, + { + "epoch": 1.53, + "learning_rate": 2.4527442934085664e-05, + "loss": 0.5208, + "step": 39777 + }, + { + "epoch": 1.53, + "learning_rate": 2.4526801744036936e-05, + "loss": 0.3286, + "step": 39778 + }, + { + "epoch": 1.53, + "learning_rate": 2.4526160553988204e-05, + "loss": 0.2496, + "step": 39779 + }, + { + "epoch": 1.53, + "learning_rate": 2.4525519363939472e-05, + "loss": 0.3435, + "step": 39780 + }, + { + "epoch": 1.53, + "learning_rate": 2.452487817389074e-05, + "loss": 0.377, + "step": 39781 + }, + { + "epoch": 1.53, + "learning_rate": 2.4524236983842012e-05, + "loss": 0.2932, + "step": 39782 + }, + { + "epoch": 1.53, + "learning_rate": 2.452359579379328e-05, + "loss": 0.3267, + "step": 39783 + }, + { + "epoch": 1.53, + "learning_rate": 2.4522954603744552e-05, + "loss": 0.3314, + "step": 39784 + }, + { + "epoch": 1.53, + "learning_rate": 2.452231341369582e-05, + "loss": 0.3779, + "step": 39785 + }, + { + "epoch": 1.53, + "learning_rate": 2.4521672223647092e-05, + "loss": 0.4015, + "step": 39786 + }, + { + "epoch": 1.53, + "learning_rate": 2.452103103359836e-05, + "loss": 0.2578, + "step": 39787 + }, + { + "epoch": 1.53, + "learning_rate": 2.452038984354963e-05, + "loss": 0.3474, + "step": 39788 + }, + { + "epoch": 1.53, + "learning_rate": 2.4519748653500897e-05, + "loss": 0.4445, + "step": 39789 + }, + { + "epoch": 1.53, + "learning_rate": 2.451910746345217e-05, + "loss": 0.3112, + "step": 39790 + }, + { + "epoch": 1.53, + "learning_rate": 2.4518466273403437e-05, + "loss": 0.4083, + "step": 39791 + }, + { + "epoch": 1.53, + "learning_rate": 2.4517825083354706e-05, + "loss": 0.3122, + "step": 39792 + }, + { + "epoch": 1.53, + "learning_rate": 2.4517183893305978e-05, + "loss": 0.3131, + "step": 39793 + }, + { + "epoch": 1.53, + "learning_rate": 2.451654270325725e-05, + "loss": 0.4232, + "step": 39794 + }, + { + "epoch": 1.53, + "learning_rate": 2.4515901513208518e-05, + "loss": 0.3824, + "step": 39795 + }, + { + "epoch": 1.53, + "learning_rate": 2.4515260323159786e-05, + "loss": 0.5353, + "step": 39796 + }, + { + "epoch": 1.53, + "learning_rate": 2.4514619133111054e-05, + "loss": 0.3865, + "step": 39797 + }, + { + "epoch": 1.53, + "learning_rate": 2.4513977943062326e-05, + "loss": 0.5658, + "step": 39798 + }, + { + "epoch": 1.53, + "learning_rate": 2.4513336753013594e-05, + "loss": 0.4118, + "step": 39799 + }, + { + "epoch": 1.53, + "learning_rate": 2.4512695562964863e-05, + "loss": 0.276, + "step": 39800 + }, + { + "epoch": 1.53, + "learning_rate": 2.451205437291613e-05, + "loss": 0.3149, + "step": 39801 + }, + { + "epoch": 1.53, + "learning_rate": 2.4511413182867403e-05, + "loss": 0.4247, + "step": 39802 + }, + { + "epoch": 1.53, + "learning_rate": 2.4510771992818674e-05, + "loss": 0.2786, + "step": 39803 + }, + { + "epoch": 1.53, + "learning_rate": 2.4510130802769943e-05, + "loss": 0.3624, + "step": 39804 + }, + { + "epoch": 1.53, + "learning_rate": 2.450948961272121e-05, + "loss": 0.3758, + "step": 39805 + }, + { + "epoch": 1.53, + "learning_rate": 2.4508848422672483e-05, + "loss": 0.4558, + "step": 39806 + }, + { + "epoch": 1.53, + "learning_rate": 2.450820723262375e-05, + "loss": 0.4317, + "step": 39807 + }, + { + "epoch": 1.53, + "learning_rate": 2.450756604257502e-05, + "loss": 0.4086, + "step": 39808 + }, + { + "epoch": 1.53, + "learning_rate": 2.4506924852526288e-05, + "loss": 0.3635, + "step": 39809 + }, + { + "epoch": 1.53, + "learning_rate": 2.450628366247756e-05, + "loss": 0.4706, + "step": 39810 + }, + { + "epoch": 1.53, + "learning_rate": 2.450564247242883e-05, + "loss": 0.2622, + "step": 39811 + }, + { + "epoch": 1.53, + "learning_rate": 2.45050012823801e-05, + "loss": 0.3343, + "step": 39812 + }, + { + "epoch": 1.53, + "learning_rate": 2.4504360092331368e-05, + "loss": 0.3485, + "step": 39813 + }, + { + "epoch": 1.53, + "learning_rate": 2.450371890228264e-05, + "loss": 0.2654, + "step": 39814 + }, + { + "epoch": 1.53, + "learning_rate": 2.4503077712233908e-05, + "loss": 0.3797, + "step": 39815 + }, + { + "epoch": 1.53, + "learning_rate": 2.4502436522185176e-05, + "loss": 0.5379, + "step": 39816 + }, + { + "epoch": 1.53, + "learning_rate": 2.4501795332136445e-05, + "loss": 0.4632, + "step": 39817 + }, + { + "epoch": 1.53, + "learning_rate": 2.4501154142087716e-05, + "loss": 0.3089, + "step": 39818 + }, + { + "epoch": 1.53, + "learning_rate": 2.4500512952038985e-05, + "loss": 0.3246, + "step": 39819 + }, + { + "epoch": 1.53, + "learning_rate": 2.4499871761990256e-05, + "loss": 0.2635, + "step": 39820 + }, + { + "epoch": 1.53, + "learning_rate": 2.4499230571941525e-05, + "loss": 0.412, + "step": 39821 + }, + { + "epoch": 1.53, + "learning_rate": 2.4498589381892796e-05, + "loss": 0.4529, + "step": 39822 + }, + { + "epoch": 1.53, + "learning_rate": 2.4497948191844065e-05, + "loss": 0.3337, + "step": 39823 + }, + { + "epoch": 1.53, + "learning_rate": 2.4497307001795333e-05, + "loss": 0.3267, + "step": 39824 + }, + { + "epoch": 1.53, + "learning_rate": 2.44966658117466e-05, + "loss": 0.4009, + "step": 39825 + }, + { + "epoch": 1.53, + "learning_rate": 2.4496024621697873e-05, + "loss": 0.3381, + "step": 39826 + }, + { + "epoch": 1.53, + "learning_rate": 2.449538343164914e-05, + "loss": 0.4225, + "step": 39827 + }, + { + "epoch": 1.53, + "learning_rate": 2.449474224160041e-05, + "loss": 0.3471, + "step": 39828 + }, + { + "epoch": 1.53, + "learning_rate": 2.449410105155168e-05, + "loss": 0.3472, + "step": 39829 + }, + { + "epoch": 1.53, + "learning_rate": 2.4493459861502953e-05, + "loss": 0.2907, + "step": 39830 + }, + { + "epoch": 1.53, + "learning_rate": 2.449281867145422e-05, + "loss": 0.3045, + "step": 39831 + }, + { + "epoch": 1.53, + "learning_rate": 2.449217748140549e-05, + "loss": 0.3597, + "step": 39832 + }, + { + "epoch": 1.53, + "learning_rate": 2.4491536291356758e-05, + "loss": 0.4079, + "step": 39833 + }, + { + "epoch": 1.53, + "learning_rate": 2.449089510130803e-05, + "loss": 0.3698, + "step": 39834 + }, + { + "epoch": 1.53, + "learning_rate": 2.4490253911259298e-05, + "loss": 0.4451, + "step": 39835 + }, + { + "epoch": 1.53, + "learning_rate": 2.4489612721210566e-05, + "loss": 0.3151, + "step": 39836 + }, + { + "epoch": 1.53, + "learning_rate": 2.4488971531161835e-05, + "loss": 0.3067, + "step": 39837 + }, + { + "epoch": 1.53, + "learning_rate": 2.4488330341113107e-05, + "loss": 0.2932, + "step": 39838 + }, + { + "epoch": 1.53, + "learning_rate": 2.4487689151064378e-05, + "loss": 0.4078, + "step": 39839 + }, + { + "epoch": 1.53, + "learning_rate": 2.4487047961015647e-05, + "loss": 0.4363, + "step": 39840 + }, + { + "epoch": 1.53, + "learning_rate": 2.4486406770966915e-05, + "loss": 0.2925, + "step": 39841 + }, + { + "epoch": 1.53, + "learning_rate": 2.4485765580918187e-05, + "loss": 0.4152, + "step": 39842 + }, + { + "epoch": 1.53, + "learning_rate": 2.4485124390869455e-05, + "loss": 0.4144, + "step": 39843 + }, + { + "epoch": 1.53, + "learning_rate": 2.4484483200820723e-05, + "loss": 0.3299, + "step": 39844 + }, + { + "epoch": 1.53, + "learning_rate": 2.448384201077199e-05, + "loss": 0.4084, + "step": 39845 + }, + { + "epoch": 1.53, + "learning_rate": 2.4483200820723263e-05, + "loss": 0.3066, + "step": 39846 + }, + { + "epoch": 1.53, + "learning_rate": 2.4482559630674535e-05, + "loss": 0.3604, + "step": 39847 + }, + { + "epoch": 1.53, + "learning_rate": 2.4481918440625803e-05, + "loss": 0.3215, + "step": 39848 + }, + { + "epoch": 1.53, + "learning_rate": 2.448127725057707e-05, + "loss": 0.3398, + "step": 39849 + }, + { + "epoch": 1.53, + "learning_rate": 2.4480636060528343e-05, + "loss": 0.3847, + "step": 39850 + }, + { + "epoch": 1.53, + "learning_rate": 2.4479994870479612e-05, + "loss": 0.3161, + "step": 39851 + }, + { + "epoch": 1.53, + "learning_rate": 2.447935368043088e-05, + "loss": 0.2388, + "step": 39852 + }, + { + "epoch": 1.53, + "learning_rate": 2.447871249038215e-05, + "loss": 0.3583, + "step": 39853 + }, + { + "epoch": 1.53, + "learning_rate": 2.447807130033342e-05, + "loss": 0.2766, + "step": 39854 + }, + { + "epoch": 1.53, + "learning_rate": 2.4477430110284692e-05, + "loss": 0.4037, + "step": 39855 + }, + { + "epoch": 1.53, + "learning_rate": 2.447678892023596e-05, + "loss": 0.3792, + "step": 39856 + }, + { + "epoch": 1.53, + "learning_rate": 2.447614773018723e-05, + "loss": 0.2141, + "step": 39857 + }, + { + "epoch": 1.53, + "learning_rate": 2.44755065401385e-05, + "loss": 0.3335, + "step": 39858 + }, + { + "epoch": 1.53, + "learning_rate": 2.447486535008977e-05, + "loss": 0.3331, + "step": 39859 + }, + { + "epoch": 1.53, + "learning_rate": 2.4474224160041037e-05, + "loss": 0.2956, + "step": 39860 + }, + { + "epoch": 1.53, + "learning_rate": 2.4473582969992305e-05, + "loss": 0.4366, + "step": 39861 + }, + { + "epoch": 1.53, + "learning_rate": 2.4472941779943577e-05, + "loss": 0.4189, + "step": 39862 + }, + { + "epoch": 1.53, + "learning_rate": 2.4472300589894845e-05, + "loss": 0.3767, + "step": 39863 + }, + { + "epoch": 1.53, + "learning_rate": 2.4471659399846117e-05, + "loss": 0.3745, + "step": 39864 + }, + { + "epoch": 1.53, + "learning_rate": 2.4471018209797385e-05, + "loss": 0.3701, + "step": 39865 + }, + { + "epoch": 1.53, + "learning_rate": 2.4470377019748657e-05, + "loss": 0.2772, + "step": 39866 + }, + { + "epoch": 1.53, + "learning_rate": 2.4469735829699925e-05, + "loss": 0.332, + "step": 39867 + }, + { + "epoch": 1.53, + "learning_rate": 2.4469094639651194e-05, + "loss": 0.25, + "step": 39868 + }, + { + "epoch": 1.53, + "learning_rate": 2.4468453449602462e-05, + "loss": 0.4122, + "step": 39869 + }, + { + "epoch": 1.53, + "learning_rate": 2.4467812259553734e-05, + "loss": 0.3271, + "step": 39870 + }, + { + "epoch": 1.53, + "learning_rate": 2.4467171069505002e-05, + "loss": 0.3056, + "step": 39871 + }, + { + "epoch": 1.53, + "learning_rate": 2.446652987945627e-05, + "loss": 0.3967, + "step": 39872 + }, + { + "epoch": 1.53, + "learning_rate": 2.4465888689407542e-05, + "loss": 0.2644, + "step": 39873 + }, + { + "epoch": 1.53, + "learning_rate": 2.4465247499358814e-05, + "loss": 0.2909, + "step": 39874 + }, + { + "epoch": 1.53, + "learning_rate": 2.4464606309310082e-05, + "loss": 0.2219, + "step": 39875 + }, + { + "epoch": 1.53, + "learning_rate": 2.446396511926135e-05, + "loss": 0.4727, + "step": 39876 + }, + { + "epoch": 1.53, + "learning_rate": 2.446332392921262e-05, + "loss": 0.4215, + "step": 39877 + }, + { + "epoch": 1.53, + "learning_rate": 2.446268273916389e-05, + "loss": 0.232, + "step": 39878 + }, + { + "epoch": 1.53, + "learning_rate": 2.446204154911516e-05, + "loss": 0.4013, + "step": 39879 + }, + { + "epoch": 1.53, + "learning_rate": 2.4461400359066427e-05, + "loss": 0.5636, + "step": 39880 + }, + { + "epoch": 1.53, + "learning_rate": 2.4460759169017695e-05, + "loss": 0.4976, + "step": 39881 + }, + { + "epoch": 1.53, + "learning_rate": 2.4460117978968967e-05, + "loss": 0.3477, + "step": 39882 + }, + { + "epoch": 1.53, + "learning_rate": 2.445947678892024e-05, + "loss": 0.3216, + "step": 39883 + }, + { + "epoch": 1.53, + "learning_rate": 2.4458835598871507e-05, + "loss": 0.3753, + "step": 39884 + }, + { + "epoch": 1.53, + "learning_rate": 2.4458194408822776e-05, + "loss": 0.3875, + "step": 39885 + }, + { + "epoch": 1.53, + "learning_rate": 2.4457553218774047e-05, + "loss": 0.4246, + "step": 39886 + }, + { + "epoch": 1.53, + "learning_rate": 2.4456912028725316e-05, + "loss": 0.309, + "step": 39887 + }, + { + "epoch": 1.53, + "learning_rate": 2.4456270838676584e-05, + "loss": 0.3643, + "step": 39888 + }, + { + "epoch": 1.53, + "learning_rate": 2.4455629648627852e-05, + "loss": 0.4901, + "step": 39889 + }, + { + "epoch": 1.53, + "learning_rate": 2.4454988458579124e-05, + "loss": 0.3178, + "step": 39890 + }, + { + "epoch": 1.53, + "learning_rate": 2.4454347268530396e-05, + "loss": 0.2941, + "step": 39891 + }, + { + "epoch": 1.53, + "learning_rate": 2.4453706078481664e-05, + "loss": 0.4747, + "step": 39892 + }, + { + "epoch": 1.53, + "learning_rate": 2.4453064888432932e-05, + "loss": 0.3574, + "step": 39893 + }, + { + "epoch": 1.53, + "learning_rate": 2.4452423698384204e-05, + "loss": 0.3018, + "step": 39894 + }, + { + "epoch": 1.53, + "learning_rate": 2.4451782508335472e-05, + "loss": 0.2786, + "step": 39895 + }, + { + "epoch": 1.53, + "learning_rate": 2.445114131828674e-05, + "loss": 0.3554, + "step": 39896 + }, + { + "epoch": 1.53, + "learning_rate": 2.445050012823801e-05, + "loss": 0.4411, + "step": 39897 + }, + { + "epoch": 1.53, + "learning_rate": 2.444985893818928e-05, + "loss": 0.3796, + "step": 39898 + }, + { + "epoch": 1.53, + "learning_rate": 2.444921774814055e-05, + "loss": 0.4247, + "step": 39899 + }, + { + "epoch": 1.53, + "learning_rate": 2.444857655809182e-05, + "loss": 0.3495, + "step": 39900 + }, + { + "epoch": 1.53, + "learning_rate": 2.444793536804309e-05, + "loss": 0.4437, + "step": 39901 + }, + { + "epoch": 1.53, + "learning_rate": 2.444729417799436e-05, + "loss": 0.2749, + "step": 39902 + }, + { + "epoch": 1.53, + "learning_rate": 2.444665298794563e-05, + "loss": 0.3736, + "step": 39903 + }, + { + "epoch": 1.53, + "learning_rate": 2.4446011797896898e-05, + "loss": 0.3506, + "step": 39904 + }, + { + "epoch": 1.53, + "learning_rate": 2.4445370607848166e-05, + "loss": 0.2907, + "step": 39905 + }, + { + "epoch": 1.53, + "learning_rate": 2.4444729417799438e-05, + "loss": 0.3027, + "step": 39906 + }, + { + "epoch": 1.53, + "learning_rate": 2.4444088227750706e-05, + "loss": 0.3102, + "step": 39907 + }, + { + "epoch": 1.53, + "learning_rate": 2.4443447037701974e-05, + "loss": 0.3452, + "step": 39908 + }, + { + "epoch": 1.53, + "learning_rate": 2.4442805847653246e-05, + "loss": 0.3076, + "step": 39909 + }, + { + "epoch": 1.53, + "learning_rate": 2.4442164657604518e-05, + "loss": 0.2791, + "step": 39910 + }, + { + "epoch": 1.53, + "learning_rate": 2.4441523467555786e-05, + "loss": 0.2845, + "step": 39911 + }, + { + "epoch": 1.53, + "learning_rate": 2.4440882277507054e-05, + "loss": 0.2266, + "step": 39912 + }, + { + "epoch": 1.53, + "learning_rate": 2.4440241087458323e-05, + "loss": 0.4221, + "step": 39913 + }, + { + "epoch": 1.53, + "learning_rate": 2.4439599897409594e-05, + "loss": 0.3152, + "step": 39914 + }, + { + "epoch": 1.53, + "learning_rate": 2.4438958707360863e-05, + "loss": 0.4226, + "step": 39915 + }, + { + "epoch": 1.53, + "learning_rate": 2.443831751731213e-05, + "loss": 0.396, + "step": 39916 + }, + { + "epoch": 1.53, + "learning_rate": 2.44376763272634e-05, + "loss": 0.3475, + "step": 39917 + }, + { + "epoch": 1.53, + "learning_rate": 2.4437035137214674e-05, + "loss": 0.4556, + "step": 39918 + }, + { + "epoch": 1.53, + "learning_rate": 2.4436393947165943e-05, + "loss": 0.3418, + "step": 39919 + }, + { + "epoch": 1.53, + "learning_rate": 2.443575275711721e-05, + "loss": 0.2897, + "step": 39920 + }, + { + "epoch": 1.53, + "learning_rate": 2.443511156706848e-05, + "loss": 0.2179, + "step": 39921 + }, + { + "epoch": 1.53, + "learning_rate": 2.443447037701975e-05, + "loss": 0.3207, + "step": 39922 + }, + { + "epoch": 1.53, + "learning_rate": 2.443382918697102e-05, + "loss": 0.3321, + "step": 39923 + }, + { + "epoch": 1.53, + "learning_rate": 2.4433187996922288e-05, + "loss": 0.4086, + "step": 39924 + }, + { + "epoch": 1.53, + "learning_rate": 2.4432546806873556e-05, + "loss": 0.207, + "step": 39925 + }, + { + "epoch": 1.54, + "learning_rate": 2.4431905616824828e-05, + "loss": 0.391, + "step": 39926 + }, + { + "epoch": 1.54, + "learning_rate": 2.44312644267761e-05, + "loss": 0.3903, + "step": 39927 + }, + { + "epoch": 1.54, + "learning_rate": 2.4430623236727368e-05, + "loss": 0.3287, + "step": 39928 + }, + { + "epoch": 1.54, + "learning_rate": 2.4429982046678636e-05, + "loss": 0.4084, + "step": 39929 + }, + { + "epoch": 1.54, + "learning_rate": 2.4429340856629908e-05, + "loss": 0.4091, + "step": 39930 + }, + { + "epoch": 1.54, + "learning_rate": 2.4428699666581176e-05, + "loss": 0.3584, + "step": 39931 + }, + { + "epoch": 1.54, + "learning_rate": 2.4428058476532445e-05, + "loss": 0.4154, + "step": 39932 + }, + { + "epoch": 1.54, + "learning_rate": 2.4427417286483713e-05, + "loss": 0.2265, + "step": 39933 + }, + { + "epoch": 1.54, + "learning_rate": 2.4426776096434985e-05, + "loss": 0.382, + "step": 39934 + }, + { + "epoch": 1.54, + "learning_rate": 2.4426134906386253e-05, + "loss": 0.5097, + "step": 39935 + }, + { + "epoch": 1.54, + "learning_rate": 2.4425493716337525e-05, + "loss": 0.4021, + "step": 39936 + }, + { + "epoch": 1.54, + "learning_rate": 2.4424852526288793e-05, + "loss": 0.3647, + "step": 39937 + }, + { + "epoch": 1.54, + "learning_rate": 2.4424211336240065e-05, + "loss": 0.4928, + "step": 39938 + }, + { + "epoch": 1.54, + "learning_rate": 2.4423570146191333e-05, + "loss": 0.4071, + "step": 39939 + }, + { + "epoch": 1.54, + "learning_rate": 2.44229289561426e-05, + "loss": 0.2977, + "step": 39940 + }, + { + "epoch": 1.54, + "learning_rate": 2.442228776609387e-05, + "loss": 0.3628, + "step": 39941 + }, + { + "epoch": 1.54, + "learning_rate": 2.442164657604514e-05, + "loss": 0.347, + "step": 39942 + }, + { + "epoch": 1.54, + "learning_rate": 2.442100538599641e-05, + "loss": 0.3761, + "step": 39943 + }, + { + "epoch": 1.54, + "learning_rate": 2.442036419594768e-05, + "loss": 0.363, + "step": 39944 + }, + { + "epoch": 1.54, + "learning_rate": 2.441972300589895e-05, + "loss": 0.4048, + "step": 39945 + }, + { + "epoch": 1.54, + "learning_rate": 2.441908181585022e-05, + "loss": 0.3673, + "step": 39946 + }, + { + "epoch": 1.54, + "learning_rate": 2.441844062580149e-05, + "loss": 0.3633, + "step": 39947 + }, + { + "epoch": 1.54, + "learning_rate": 2.4417799435752758e-05, + "loss": 0.368, + "step": 39948 + }, + { + "epoch": 1.54, + "learning_rate": 2.4417158245704026e-05, + "loss": 0.3367, + "step": 39949 + }, + { + "epoch": 1.54, + "learning_rate": 2.4416517055655298e-05, + "loss": 0.3675, + "step": 39950 + }, + { + "epoch": 1.54, + "learning_rate": 2.4415875865606567e-05, + "loss": 0.469, + "step": 39951 + }, + { + "epoch": 1.54, + "learning_rate": 2.4415234675557835e-05, + "loss": 0.2809, + "step": 39952 + }, + { + "epoch": 1.54, + "learning_rate": 2.4414593485509107e-05, + "loss": 0.2378, + "step": 39953 + }, + { + "epoch": 1.54, + "learning_rate": 2.4413952295460378e-05, + "loss": 0.301, + "step": 39954 + }, + { + "epoch": 1.54, + "learning_rate": 2.4413311105411647e-05, + "loss": 0.2836, + "step": 39955 + }, + { + "epoch": 1.54, + "learning_rate": 2.4412669915362915e-05, + "loss": 0.2684, + "step": 39956 + }, + { + "epoch": 1.54, + "learning_rate": 2.4412028725314183e-05, + "loss": 0.3807, + "step": 39957 + }, + { + "epoch": 1.54, + "learning_rate": 2.4411387535265455e-05, + "loss": 0.3878, + "step": 39958 + }, + { + "epoch": 1.54, + "learning_rate": 2.4410746345216723e-05, + "loss": 0.396, + "step": 39959 + }, + { + "epoch": 1.54, + "learning_rate": 2.441010515516799e-05, + "loss": 0.4194, + "step": 39960 + }, + { + "epoch": 1.54, + "learning_rate": 2.440946396511926e-05, + "loss": 0.363, + "step": 39961 + }, + { + "epoch": 1.54, + "learning_rate": 2.4408822775070532e-05, + "loss": 0.347, + "step": 39962 + }, + { + "epoch": 1.54, + "learning_rate": 2.4408181585021803e-05, + "loss": 0.3592, + "step": 39963 + }, + { + "epoch": 1.54, + "learning_rate": 2.4407540394973072e-05, + "loss": 0.244, + "step": 39964 + }, + { + "epoch": 1.54, + "learning_rate": 2.440689920492434e-05, + "loss": 0.3993, + "step": 39965 + }, + { + "epoch": 1.54, + "learning_rate": 2.4406258014875612e-05, + "loss": 0.3289, + "step": 39966 + }, + { + "epoch": 1.54, + "learning_rate": 2.440561682482688e-05, + "loss": 0.2626, + "step": 39967 + }, + { + "epoch": 1.54, + "learning_rate": 2.440497563477815e-05, + "loss": 0.1712, + "step": 39968 + }, + { + "epoch": 1.54, + "learning_rate": 2.4404334444729417e-05, + "loss": 0.2457, + "step": 39969 + }, + { + "epoch": 1.54, + "learning_rate": 2.440369325468069e-05, + "loss": 0.332, + "step": 39970 + }, + { + "epoch": 1.54, + "learning_rate": 2.440305206463196e-05, + "loss": 0.3336, + "step": 39971 + }, + { + "epoch": 1.54, + "learning_rate": 2.440241087458323e-05, + "loss": 0.3543, + "step": 39972 + }, + { + "epoch": 1.54, + "learning_rate": 2.4401769684534497e-05, + "loss": 0.4545, + "step": 39973 + }, + { + "epoch": 1.54, + "learning_rate": 2.440112849448577e-05, + "loss": 0.3064, + "step": 39974 + }, + { + "epoch": 1.54, + "learning_rate": 2.4400487304437037e-05, + "loss": 0.3692, + "step": 39975 + }, + { + "epoch": 1.54, + "learning_rate": 2.4399846114388305e-05, + "loss": 0.3291, + "step": 39976 + }, + { + "epoch": 1.54, + "learning_rate": 2.4399204924339574e-05, + "loss": 0.273, + "step": 39977 + }, + { + "epoch": 1.54, + "learning_rate": 2.4398563734290845e-05, + "loss": 0.4383, + "step": 39978 + }, + { + "epoch": 1.54, + "learning_rate": 2.4397922544242114e-05, + "loss": 0.2814, + "step": 39979 + }, + { + "epoch": 1.54, + "learning_rate": 2.4397281354193385e-05, + "loss": 0.3755, + "step": 39980 + }, + { + "epoch": 1.54, + "learning_rate": 2.4396640164144654e-05, + "loss": 0.3586, + "step": 39981 + }, + { + "epoch": 1.54, + "learning_rate": 2.4395998974095925e-05, + "loss": 0.3519, + "step": 39982 + }, + { + "epoch": 1.54, + "learning_rate": 2.4395357784047194e-05, + "loss": 0.3932, + "step": 39983 + }, + { + "epoch": 1.54, + "learning_rate": 2.4394716593998462e-05, + "loss": 0.3304, + "step": 39984 + }, + { + "epoch": 1.54, + "learning_rate": 2.439407540394973e-05, + "loss": 0.34, + "step": 39985 + }, + { + "epoch": 1.54, + "learning_rate": 2.4393434213901002e-05, + "loss": 0.3175, + "step": 39986 + }, + { + "epoch": 1.54, + "learning_rate": 2.439279302385227e-05, + "loss": 0.3747, + "step": 39987 + }, + { + "epoch": 1.54, + "learning_rate": 2.439215183380354e-05, + "loss": 0.4548, + "step": 39988 + }, + { + "epoch": 1.54, + "learning_rate": 2.439151064375481e-05, + "loss": 0.3797, + "step": 39989 + }, + { + "epoch": 1.54, + "learning_rate": 2.4390869453706082e-05, + "loss": 0.4209, + "step": 39990 + }, + { + "epoch": 1.54, + "learning_rate": 2.439022826365735e-05, + "loss": 0.3908, + "step": 39991 + }, + { + "epoch": 1.54, + "learning_rate": 2.438958707360862e-05, + "loss": 0.425, + "step": 39992 + }, + { + "epoch": 1.54, + "learning_rate": 2.4388945883559887e-05, + "loss": 0.3018, + "step": 39993 + }, + { + "epoch": 1.54, + "learning_rate": 2.438830469351116e-05, + "loss": 0.2979, + "step": 39994 + }, + { + "epoch": 1.54, + "learning_rate": 2.4387663503462427e-05, + "loss": 0.3554, + "step": 39995 + }, + { + "epoch": 1.54, + "learning_rate": 2.4387022313413696e-05, + "loss": 0.2645, + "step": 39996 + }, + { + "epoch": 1.54, + "learning_rate": 2.4386381123364964e-05, + "loss": 0.3117, + "step": 39997 + }, + { + "epoch": 1.54, + "learning_rate": 2.438573993331624e-05, + "loss": 0.2614, + "step": 39998 + }, + { + "epoch": 1.54, + "learning_rate": 2.4385098743267507e-05, + "loss": 0.2673, + "step": 39999 + }, + { + "epoch": 1.54, + "learning_rate": 2.4384457553218776e-05, + "loss": 0.4244, + "step": 40000 + }, + { + "epoch": 1.54, + "learning_rate": 2.4383816363170044e-05, + "loss": 0.2485, + "step": 40001 + }, + { + "epoch": 1.54, + "learning_rate": 2.4383175173121316e-05, + "loss": 0.3213, + "step": 40002 + }, + { + "epoch": 1.54, + "learning_rate": 2.4382533983072584e-05, + "loss": 0.3841, + "step": 40003 + }, + { + "epoch": 1.54, + "learning_rate": 2.4381892793023852e-05, + "loss": 0.2524, + "step": 40004 + }, + { + "epoch": 1.54, + "learning_rate": 2.438125160297512e-05, + "loss": 0.4224, + "step": 40005 + }, + { + "epoch": 1.54, + "learning_rate": 2.4380610412926392e-05, + "loss": 0.3582, + "step": 40006 + }, + { + "epoch": 1.54, + "learning_rate": 2.4379969222877664e-05, + "loss": 0.3459, + "step": 40007 + }, + { + "epoch": 1.54, + "learning_rate": 2.4379328032828932e-05, + "loss": 0.3133, + "step": 40008 + }, + { + "epoch": 1.54, + "learning_rate": 2.43786868427802e-05, + "loss": 0.4657, + "step": 40009 + }, + { + "epoch": 1.54, + "learning_rate": 2.4378045652731472e-05, + "loss": 0.3777, + "step": 40010 + }, + { + "epoch": 1.54, + "learning_rate": 2.437740446268274e-05, + "loss": 0.4297, + "step": 40011 + }, + { + "epoch": 1.54, + "learning_rate": 2.437676327263401e-05, + "loss": 0.418, + "step": 40012 + }, + { + "epoch": 1.54, + "learning_rate": 2.4376122082585277e-05, + "loss": 0.2741, + "step": 40013 + }, + { + "epoch": 1.54, + "learning_rate": 2.437548089253655e-05, + "loss": 0.3525, + "step": 40014 + }, + { + "epoch": 1.54, + "learning_rate": 2.4374839702487817e-05, + "loss": 0.3346, + "step": 40015 + }, + { + "epoch": 1.54, + "learning_rate": 2.437419851243909e-05, + "loss": 0.4466, + "step": 40016 + }, + { + "epoch": 1.54, + "learning_rate": 2.4373557322390358e-05, + "loss": 0.5418, + "step": 40017 + }, + { + "epoch": 1.54, + "learning_rate": 2.437291613234163e-05, + "loss": 0.4265, + "step": 40018 + }, + { + "epoch": 1.54, + "learning_rate": 2.4372274942292898e-05, + "loss": 0.3641, + "step": 40019 + }, + { + "epoch": 1.54, + "learning_rate": 2.4371633752244166e-05, + "loss": 0.3905, + "step": 40020 + }, + { + "epoch": 1.54, + "learning_rate": 2.4370992562195434e-05, + "loss": 0.3825, + "step": 40021 + }, + { + "epoch": 1.54, + "learning_rate": 2.4370351372146706e-05, + "loss": 0.2524, + "step": 40022 + }, + { + "epoch": 1.54, + "learning_rate": 2.4369710182097974e-05, + "loss": 0.32, + "step": 40023 + }, + { + "epoch": 1.54, + "learning_rate": 2.4369068992049243e-05, + "loss": 0.5621, + "step": 40024 + }, + { + "epoch": 1.54, + "learning_rate": 2.4368427802000514e-05, + "loss": 0.2301, + "step": 40025 + }, + { + "epoch": 1.54, + "learning_rate": 2.4367786611951786e-05, + "loss": 0.3541, + "step": 40026 + }, + { + "epoch": 1.54, + "learning_rate": 2.4367145421903054e-05, + "loss": 0.3244, + "step": 40027 + }, + { + "epoch": 1.54, + "learning_rate": 2.4366504231854323e-05, + "loss": 0.4249, + "step": 40028 + }, + { + "epoch": 1.54, + "learning_rate": 2.436586304180559e-05, + "loss": 0.3019, + "step": 40029 + }, + { + "epoch": 1.54, + "learning_rate": 2.4365221851756863e-05, + "loss": 0.2826, + "step": 40030 + }, + { + "epoch": 1.54, + "learning_rate": 2.436458066170813e-05, + "loss": 0.266, + "step": 40031 + }, + { + "epoch": 1.54, + "learning_rate": 2.43639394716594e-05, + "loss": 0.3777, + "step": 40032 + }, + { + "epoch": 1.54, + "learning_rate": 2.4363298281610668e-05, + "loss": 0.348, + "step": 40033 + }, + { + "epoch": 1.54, + "learning_rate": 2.4362657091561943e-05, + "loss": 0.3042, + "step": 40034 + }, + { + "epoch": 1.54, + "learning_rate": 2.436201590151321e-05, + "loss": 0.2915, + "step": 40035 + }, + { + "epoch": 1.54, + "learning_rate": 2.436137471146448e-05, + "loss": 0.4037, + "step": 40036 + }, + { + "epoch": 1.54, + "learning_rate": 2.4360733521415748e-05, + "loss": 0.333, + "step": 40037 + }, + { + "epoch": 1.54, + "learning_rate": 2.436009233136702e-05, + "loss": 0.3621, + "step": 40038 + }, + { + "epoch": 1.54, + "learning_rate": 2.4359451141318288e-05, + "loss": 0.2573, + "step": 40039 + }, + { + "epoch": 1.54, + "learning_rate": 2.4358809951269556e-05, + "loss": 0.3628, + "step": 40040 + }, + { + "epoch": 1.54, + "learning_rate": 2.4358168761220825e-05, + "loss": 0.357, + "step": 40041 + }, + { + "epoch": 1.54, + "learning_rate": 2.43575275711721e-05, + "loss": 0.4078, + "step": 40042 + }, + { + "epoch": 1.54, + "learning_rate": 2.4356886381123368e-05, + "loss": 0.3033, + "step": 40043 + }, + { + "epoch": 1.54, + "learning_rate": 2.4356245191074636e-05, + "loss": 0.4786, + "step": 40044 + }, + { + "epoch": 1.54, + "learning_rate": 2.4355604001025905e-05, + "loss": 0.3382, + "step": 40045 + }, + { + "epoch": 1.54, + "learning_rate": 2.4354962810977176e-05, + "loss": 0.4242, + "step": 40046 + }, + { + "epoch": 1.54, + "learning_rate": 2.4354321620928445e-05, + "loss": 0.355, + "step": 40047 + }, + { + "epoch": 1.54, + "learning_rate": 2.4353680430879713e-05, + "loss": 0.2491, + "step": 40048 + }, + { + "epoch": 1.54, + "learning_rate": 2.435303924083098e-05, + "loss": 0.3108, + "step": 40049 + }, + { + "epoch": 1.54, + "learning_rate": 2.4352398050782253e-05, + "loss": 0.3601, + "step": 40050 + }, + { + "epoch": 1.54, + "learning_rate": 2.4351756860733525e-05, + "loss": 0.4128, + "step": 40051 + }, + { + "epoch": 1.54, + "learning_rate": 2.4351115670684793e-05, + "loss": 0.3098, + "step": 40052 + }, + { + "epoch": 1.54, + "learning_rate": 2.435047448063606e-05, + "loss": 0.3159, + "step": 40053 + }, + { + "epoch": 1.54, + "learning_rate": 2.4349833290587333e-05, + "loss": 0.358, + "step": 40054 + }, + { + "epoch": 1.54, + "learning_rate": 2.43491921005386e-05, + "loss": 0.4431, + "step": 40055 + }, + { + "epoch": 1.54, + "learning_rate": 2.434855091048987e-05, + "loss": 0.3577, + "step": 40056 + }, + { + "epoch": 1.54, + "learning_rate": 2.4347909720441138e-05, + "loss": 0.4306, + "step": 40057 + }, + { + "epoch": 1.54, + "learning_rate": 2.434726853039241e-05, + "loss": 0.4235, + "step": 40058 + }, + { + "epoch": 1.54, + "learning_rate": 2.4346627340343678e-05, + "loss": 0.3878, + "step": 40059 + }, + { + "epoch": 1.54, + "learning_rate": 2.434598615029495e-05, + "loss": 0.4867, + "step": 40060 + }, + { + "epoch": 1.54, + "learning_rate": 2.4345344960246218e-05, + "loss": 0.3102, + "step": 40061 + }, + { + "epoch": 1.54, + "learning_rate": 2.434470377019749e-05, + "loss": 0.2967, + "step": 40062 + }, + { + "epoch": 1.54, + "learning_rate": 2.4344062580148758e-05, + "loss": 0.2524, + "step": 40063 + }, + { + "epoch": 1.54, + "learning_rate": 2.4343421390100027e-05, + "loss": 0.2998, + "step": 40064 + }, + { + "epoch": 1.54, + "learning_rate": 2.4342780200051295e-05, + "loss": 0.3073, + "step": 40065 + }, + { + "epoch": 1.54, + "learning_rate": 2.4342139010002567e-05, + "loss": 0.2956, + "step": 40066 + }, + { + "epoch": 1.54, + "learning_rate": 2.4341497819953835e-05, + "loss": 0.3942, + "step": 40067 + }, + { + "epoch": 1.54, + "learning_rate": 2.4340856629905103e-05, + "loss": 0.2986, + "step": 40068 + }, + { + "epoch": 1.54, + "learning_rate": 2.4340215439856375e-05, + "loss": 0.3629, + "step": 40069 + }, + { + "epoch": 1.54, + "learning_rate": 2.4339574249807647e-05, + "loss": 0.5784, + "step": 40070 + }, + { + "epoch": 1.54, + "learning_rate": 2.4338933059758915e-05, + "loss": 0.4242, + "step": 40071 + }, + { + "epoch": 1.54, + "learning_rate": 2.4338291869710183e-05, + "loss": 0.3173, + "step": 40072 + }, + { + "epoch": 1.54, + "learning_rate": 2.433765067966145e-05, + "loss": 0.3813, + "step": 40073 + }, + { + "epoch": 1.54, + "learning_rate": 2.4337009489612723e-05, + "loss": 0.3315, + "step": 40074 + }, + { + "epoch": 1.54, + "learning_rate": 2.4336368299563992e-05, + "loss": 0.3741, + "step": 40075 + }, + { + "epoch": 1.54, + "learning_rate": 2.433572710951526e-05, + "loss": 0.2951, + "step": 40076 + }, + { + "epoch": 1.54, + "learning_rate": 2.433508591946653e-05, + "loss": 0.3082, + "step": 40077 + }, + { + "epoch": 1.54, + "learning_rate": 2.4334444729417803e-05, + "loss": 0.4525, + "step": 40078 + }, + { + "epoch": 1.54, + "learning_rate": 2.4333803539369072e-05, + "loss": 0.3241, + "step": 40079 + }, + { + "epoch": 1.54, + "learning_rate": 2.433316234932034e-05, + "loss": 0.3795, + "step": 40080 + }, + { + "epoch": 1.54, + "learning_rate": 2.433252115927161e-05, + "loss": 0.4157, + "step": 40081 + }, + { + "epoch": 1.54, + "learning_rate": 2.433187996922288e-05, + "loss": 0.3225, + "step": 40082 + }, + { + "epoch": 1.54, + "learning_rate": 2.433123877917415e-05, + "loss": 0.4241, + "step": 40083 + }, + { + "epoch": 1.54, + "learning_rate": 2.4330597589125417e-05, + "loss": 0.3339, + "step": 40084 + }, + { + "epoch": 1.54, + "learning_rate": 2.4329956399076685e-05, + "loss": 0.3269, + "step": 40085 + }, + { + "epoch": 1.54, + "learning_rate": 2.4329315209027957e-05, + "loss": 0.3332, + "step": 40086 + }, + { + "epoch": 1.54, + "learning_rate": 2.432867401897923e-05, + "loss": 0.3265, + "step": 40087 + }, + { + "epoch": 1.54, + "learning_rate": 2.4328032828930497e-05, + "loss": 0.3085, + "step": 40088 + }, + { + "epoch": 1.54, + "learning_rate": 2.4327391638881765e-05, + "loss": 0.3534, + "step": 40089 + }, + { + "epoch": 1.54, + "learning_rate": 2.4326750448833037e-05, + "loss": 0.324, + "step": 40090 + }, + { + "epoch": 1.54, + "learning_rate": 2.4326109258784305e-05, + "loss": 0.3144, + "step": 40091 + }, + { + "epoch": 1.54, + "learning_rate": 2.4325468068735574e-05, + "loss": 0.5754, + "step": 40092 + }, + { + "epoch": 1.54, + "learning_rate": 2.4324826878686842e-05, + "loss": 0.3504, + "step": 40093 + }, + { + "epoch": 1.54, + "learning_rate": 2.4324185688638114e-05, + "loss": 0.306, + "step": 40094 + }, + { + "epoch": 1.54, + "learning_rate": 2.4323544498589382e-05, + "loss": 0.3725, + "step": 40095 + }, + { + "epoch": 1.54, + "learning_rate": 2.4322903308540654e-05, + "loss": 0.3396, + "step": 40096 + }, + { + "epoch": 1.54, + "learning_rate": 2.4322262118491922e-05, + "loss": 0.3619, + "step": 40097 + }, + { + "epoch": 1.54, + "learning_rate": 2.4321620928443194e-05, + "loss": 0.2955, + "step": 40098 + }, + { + "epoch": 1.54, + "learning_rate": 2.4320979738394462e-05, + "loss": 0.3779, + "step": 40099 + }, + { + "epoch": 1.54, + "learning_rate": 2.432033854834573e-05, + "loss": 0.5406, + "step": 40100 + }, + { + "epoch": 1.54, + "learning_rate": 2.4319697358297e-05, + "loss": 0.3317, + "step": 40101 + }, + { + "epoch": 1.54, + "learning_rate": 2.431905616824827e-05, + "loss": 0.4293, + "step": 40102 + }, + { + "epoch": 1.54, + "learning_rate": 2.431841497819954e-05, + "loss": 0.3725, + "step": 40103 + }, + { + "epoch": 1.54, + "learning_rate": 2.4317773788150807e-05, + "loss": 0.3702, + "step": 40104 + }, + { + "epoch": 1.54, + "learning_rate": 2.431713259810208e-05, + "loss": 0.3427, + "step": 40105 + }, + { + "epoch": 1.54, + "learning_rate": 2.431649140805335e-05, + "loss": 0.3736, + "step": 40106 + }, + { + "epoch": 1.54, + "learning_rate": 2.431585021800462e-05, + "loss": 0.3315, + "step": 40107 + }, + { + "epoch": 1.54, + "learning_rate": 2.4315209027955887e-05, + "loss": 0.3572, + "step": 40108 + }, + { + "epoch": 1.54, + "learning_rate": 2.4314567837907156e-05, + "loss": 0.4086, + "step": 40109 + }, + { + "epoch": 1.54, + "learning_rate": 2.4313926647858427e-05, + "loss": 0.4003, + "step": 40110 + }, + { + "epoch": 1.54, + "learning_rate": 2.4313285457809696e-05, + "loss": 0.4079, + "step": 40111 + }, + { + "epoch": 1.54, + "learning_rate": 2.4312644267760964e-05, + "loss": 0.3439, + "step": 40112 + }, + { + "epoch": 1.54, + "learning_rate": 2.4312003077712232e-05, + "loss": 0.2828, + "step": 40113 + }, + { + "epoch": 1.54, + "learning_rate": 2.4311361887663507e-05, + "loss": 0.3591, + "step": 40114 + }, + { + "epoch": 1.54, + "learning_rate": 2.4310720697614776e-05, + "loss": 0.2776, + "step": 40115 + }, + { + "epoch": 1.54, + "learning_rate": 2.4310079507566044e-05, + "loss": 0.3809, + "step": 40116 + }, + { + "epoch": 1.54, + "learning_rate": 2.4309438317517312e-05, + "loss": 0.3628, + "step": 40117 + }, + { + "epoch": 1.54, + "learning_rate": 2.4308797127468584e-05, + "loss": 0.4113, + "step": 40118 + }, + { + "epoch": 1.54, + "learning_rate": 2.4308155937419852e-05, + "loss": 0.37, + "step": 40119 + }, + { + "epoch": 1.54, + "learning_rate": 2.430751474737112e-05, + "loss": 0.2532, + "step": 40120 + }, + { + "epoch": 1.54, + "learning_rate": 2.430687355732239e-05, + "loss": 0.3911, + "step": 40121 + }, + { + "epoch": 1.54, + "learning_rate": 2.430623236727366e-05, + "loss": 0.26, + "step": 40122 + }, + { + "epoch": 1.54, + "learning_rate": 2.4305591177224932e-05, + "loss": 0.4299, + "step": 40123 + }, + { + "epoch": 1.54, + "learning_rate": 2.43049499871762e-05, + "loss": 0.3101, + "step": 40124 + }, + { + "epoch": 1.54, + "learning_rate": 2.430430879712747e-05, + "loss": 0.2825, + "step": 40125 + }, + { + "epoch": 1.54, + "learning_rate": 2.430366760707874e-05, + "loss": 0.2875, + "step": 40126 + }, + { + "epoch": 1.54, + "learning_rate": 2.430302641703001e-05, + "loss": 0.3197, + "step": 40127 + }, + { + "epoch": 1.54, + "learning_rate": 2.4302385226981278e-05, + "loss": 0.4675, + "step": 40128 + }, + { + "epoch": 1.54, + "learning_rate": 2.4301744036932546e-05, + "loss": 0.4195, + "step": 40129 + }, + { + "epoch": 1.54, + "learning_rate": 2.4301102846883818e-05, + "loss": 0.4357, + "step": 40130 + }, + { + "epoch": 1.54, + "learning_rate": 2.430046165683509e-05, + "loss": 0.3742, + "step": 40131 + }, + { + "epoch": 1.54, + "learning_rate": 2.4299820466786358e-05, + "loss": 0.4354, + "step": 40132 + }, + { + "epoch": 1.54, + "learning_rate": 2.4299179276737626e-05, + "loss": 0.2948, + "step": 40133 + }, + { + "epoch": 1.54, + "learning_rate": 2.4298538086688898e-05, + "loss": 0.4593, + "step": 40134 + }, + { + "epoch": 1.54, + "learning_rate": 2.4297896896640166e-05, + "loss": 0.4037, + "step": 40135 + }, + { + "epoch": 1.54, + "learning_rate": 2.4297255706591434e-05, + "loss": 0.3448, + "step": 40136 + }, + { + "epoch": 1.54, + "learning_rate": 2.4296614516542703e-05, + "loss": 0.4536, + "step": 40137 + }, + { + "epoch": 1.54, + "learning_rate": 2.4295973326493974e-05, + "loss": 0.402, + "step": 40138 + }, + { + "epoch": 1.54, + "learning_rate": 2.4295332136445243e-05, + "loss": 0.3189, + "step": 40139 + }, + { + "epoch": 1.54, + "learning_rate": 2.4294690946396514e-05, + "loss": 0.3313, + "step": 40140 + }, + { + "epoch": 1.54, + "learning_rate": 2.4294049756347783e-05, + "loss": 0.416, + "step": 40141 + }, + { + "epoch": 1.54, + "learning_rate": 2.4293408566299054e-05, + "loss": 0.2912, + "step": 40142 + }, + { + "epoch": 1.54, + "learning_rate": 2.4292767376250323e-05, + "loss": 0.3792, + "step": 40143 + }, + { + "epoch": 1.54, + "learning_rate": 2.429212618620159e-05, + "loss": 0.4107, + "step": 40144 + }, + { + "epoch": 1.54, + "learning_rate": 2.429148499615286e-05, + "loss": 0.451, + "step": 40145 + }, + { + "epoch": 1.54, + "learning_rate": 2.429084380610413e-05, + "loss": 0.422, + "step": 40146 + }, + { + "epoch": 1.54, + "learning_rate": 2.42902026160554e-05, + "loss": 0.3641, + "step": 40147 + }, + { + "epoch": 1.54, + "learning_rate": 2.4289561426006668e-05, + "loss": 0.336, + "step": 40148 + }, + { + "epoch": 1.54, + "learning_rate": 2.428892023595794e-05, + "loss": 0.2885, + "step": 40149 + }, + { + "epoch": 1.54, + "learning_rate": 2.428827904590921e-05, + "loss": 0.3776, + "step": 40150 + }, + { + "epoch": 1.54, + "learning_rate": 2.428763785586048e-05, + "loss": 0.4204, + "step": 40151 + }, + { + "epoch": 1.54, + "learning_rate": 2.4286996665811748e-05, + "loss": 0.3315, + "step": 40152 + }, + { + "epoch": 1.54, + "learning_rate": 2.4286355475763016e-05, + "loss": 0.2811, + "step": 40153 + }, + { + "epoch": 1.54, + "learning_rate": 2.4285714285714288e-05, + "loss": 0.3334, + "step": 40154 + }, + { + "epoch": 1.54, + "learning_rate": 2.4285073095665556e-05, + "loss": 0.4289, + "step": 40155 + }, + { + "epoch": 1.54, + "learning_rate": 2.4284431905616825e-05, + "loss": 0.2409, + "step": 40156 + }, + { + "epoch": 1.54, + "learning_rate": 2.4283790715568093e-05, + "loss": 0.3693, + "step": 40157 + }, + { + "epoch": 1.54, + "learning_rate": 2.4283149525519368e-05, + "loss": 0.3094, + "step": 40158 + }, + { + "epoch": 1.54, + "learning_rate": 2.4282508335470636e-05, + "loss": 0.4163, + "step": 40159 + }, + { + "epoch": 1.54, + "learning_rate": 2.4281867145421905e-05, + "loss": 0.4219, + "step": 40160 + }, + { + "epoch": 1.54, + "learning_rate": 2.4281225955373173e-05, + "loss": 0.4622, + "step": 40161 + }, + { + "epoch": 1.54, + "learning_rate": 2.4280584765324445e-05, + "loss": 0.3888, + "step": 40162 + }, + { + "epoch": 1.54, + "learning_rate": 2.4279943575275713e-05, + "loss": 0.3498, + "step": 40163 + }, + { + "epoch": 1.54, + "learning_rate": 2.427930238522698e-05, + "loss": 0.4732, + "step": 40164 + }, + { + "epoch": 1.54, + "learning_rate": 2.427866119517825e-05, + "loss": 0.252, + "step": 40165 + }, + { + "epoch": 1.54, + "learning_rate": 2.427802000512952e-05, + "loss": 0.3908, + "step": 40166 + }, + { + "epoch": 1.54, + "learning_rate": 2.4277378815080793e-05, + "loss": 0.3968, + "step": 40167 + }, + { + "epoch": 1.54, + "learning_rate": 2.427673762503206e-05, + "loss": 0.3384, + "step": 40168 + }, + { + "epoch": 1.54, + "learning_rate": 2.427609643498333e-05, + "loss": 0.3214, + "step": 40169 + }, + { + "epoch": 1.54, + "learning_rate": 2.42754552449346e-05, + "loss": 0.4537, + "step": 40170 + }, + { + "epoch": 1.54, + "learning_rate": 2.427481405488587e-05, + "loss": 0.3277, + "step": 40171 + }, + { + "epoch": 1.54, + "learning_rate": 2.4274172864837138e-05, + "loss": 0.2885, + "step": 40172 + }, + { + "epoch": 1.54, + "learning_rate": 2.4273531674788407e-05, + "loss": 0.2642, + "step": 40173 + }, + { + "epoch": 1.54, + "learning_rate": 2.4272890484739678e-05, + "loss": 0.3397, + "step": 40174 + }, + { + "epoch": 1.54, + "learning_rate": 2.4272249294690947e-05, + "loss": 0.2975, + "step": 40175 + }, + { + "epoch": 1.54, + "learning_rate": 2.4271608104642218e-05, + "loss": 0.3898, + "step": 40176 + }, + { + "epoch": 1.54, + "learning_rate": 2.4270966914593487e-05, + "loss": 0.3799, + "step": 40177 + }, + { + "epoch": 1.54, + "learning_rate": 2.427032572454476e-05, + "loss": 0.3259, + "step": 40178 + }, + { + "epoch": 1.54, + "learning_rate": 2.4269684534496027e-05, + "loss": 0.2631, + "step": 40179 + }, + { + "epoch": 1.54, + "learning_rate": 2.4269043344447295e-05, + "loss": 0.3804, + "step": 40180 + }, + { + "epoch": 1.54, + "learning_rate": 2.4268402154398563e-05, + "loss": 0.3408, + "step": 40181 + }, + { + "epoch": 1.54, + "learning_rate": 2.4267760964349835e-05, + "loss": 0.285, + "step": 40182 + }, + { + "epoch": 1.54, + "learning_rate": 2.4267119774301103e-05, + "loss": 0.3831, + "step": 40183 + }, + { + "epoch": 1.54, + "learning_rate": 2.426647858425237e-05, + "loss": 0.323, + "step": 40184 + }, + { + "epoch": 1.54, + "learning_rate": 2.4265837394203643e-05, + "loss": 0.4346, + "step": 40185 + }, + { + "epoch": 1.55, + "learning_rate": 2.4265196204154915e-05, + "loss": 0.2641, + "step": 40186 + }, + { + "epoch": 1.55, + "learning_rate": 2.4264555014106183e-05, + "loss": 0.5304, + "step": 40187 + }, + { + "epoch": 1.55, + "learning_rate": 2.4263913824057452e-05, + "loss": 0.3635, + "step": 40188 + }, + { + "epoch": 1.55, + "learning_rate": 2.426327263400872e-05, + "loss": 0.3698, + "step": 40189 + }, + { + "epoch": 1.55, + "learning_rate": 2.4262631443959992e-05, + "loss": 0.337, + "step": 40190 + }, + { + "epoch": 1.55, + "learning_rate": 2.426199025391126e-05, + "loss": 0.4033, + "step": 40191 + }, + { + "epoch": 1.55, + "learning_rate": 2.426134906386253e-05, + "loss": 0.4837, + "step": 40192 + }, + { + "epoch": 1.55, + "learning_rate": 2.4260707873813797e-05, + "loss": 0.377, + "step": 40193 + }, + { + "epoch": 1.55, + "learning_rate": 2.4260066683765072e-05, + "loss": 0.4811, + "step": 40194 + }, + { + "epoch": 1.55, + "learning_rate": 2.425942549371634e-05, + "loss": 0.3393, + "step": 40195 + }, + { + "epoch": 1.55, + "learning_rate": 2.425878430366761e-05, + "loss": 0.274, + "step": 40196 + }, + { + "epoch": 1.55, + "learning_rate": 2.4258143113618877e-05, + "loss": 0.3684, + "step": 40197 + }, + { + "epoch": 1.55, + "learning_rate": 2.425750192357015e-05, + "loss": 0.4957, + "step": 40198 + }, + { + "epoch": 1.55, + "learning_rate": 2.4256860733521417e-05, + "loss": 0.3134, + "step": 40199 + }, + { + "epoch": 1.55, + "learning_rate": 2.4256219543472685e-05, + "loss": 0.3567, + "step": 40200 + }, + { + "epoch": 1.55, + "learning_rate": 2.4255578353423954e-05, + "loss": 0.3433, + "step": 40201 + }, + { + "epoch": 1.55, + "learning_rate": 2.4254937163375225e-05, + "loss": 0.4331, + "step": 40202 + }, + { + "epoch": 1.55, + "learning_rate": 2.4254295973326497e-05, + "loss": 0.5436, + "step": 40203 + }, + { + "epoch": 1.55, + "learning_rate": 2.4253654783277765e-05, + "loss": 0.3587, + "step": 40204 + }, + { + "epoch": 1.55, + "learning_rate": 2.4253013593229034e-05, + "loss": 0.3447, + "step": 40205 + }, + { + "epoch": 1.55, + "learning_rate": 2.4252372403180305e-05, + "loss": 0.3622, + "step": 40206 + }, + { + "epoch": 1.55, + "learning_rate": 2.4251731213131574e-05, + "loss": 0.4781, + "step": 40207 + }, + { + "epoch": 1.55, + "learning_rate": 2.4251090023082842e-05, + "loss": 0.2334, + "step": 40208 + }, + { + "epoch": 1.55, + "learning_rate": 2.425044883303411e-05, + "loss": 0.3826, + "step": 40209 + }, + { + "epoch": 1.55, + "learning_rate": 2.4249807642985382e-05, + "loss": 0.394, + "step": 40210 + }, + { + "epoch": 1.55, + "learning_rate": 2.424916645293665e-05, + "loss": 0.297, + "step": 40211 + }, + { + "epoch": 1.55, + "learning_rate": 2.4248525262887922e-05, + "loss": 0.3523, + "step": 40212 + }, + { + "epoch": 1.55, + "learning_rate": 2.424788407283919e-05, + "loss": 0.3451, + "step": 40213 + }, + { + "epoch": 1.55, + "learning_rate": 2.4247242882790462e-05, + "loss": 0.3724, + "step": 40214 + }, + { + "epoch": 1.55, + "learning_rate": 2.424660169274173e-05, + "loss": 0.383, + "step": 40215 + }, + { + "epoch": 1.55, + "learning_rate": 2.4245960502693e-05, + "loss": 0.3457, + "step": 40216 + }, + { + "epoch": 1.55, + "learning_rate": 2.4245319312644267e-05, + "loss": 0.3738, + "step": 40217 + }, + { + "epoch": 1.55, + "learning_rate": 2.424467812259554e-05, + "loss": 0.3563, + "step": 40218 + }, + { + "epoch": 1.55, + "learning_rate": 2.4244036932546807e-05, + "loss": 0.3855, + "step": 40219 + }, + { + "epoch": 1.55, + "learning_rate": 2.4243395742498076e-05, + "loss": 0.375, + "step": 40220 + }, + { + "epoch": 1.55, + "learning_rate": 2.4242754552449347e-05, + "loss": 0.3156, + "step": 40221 + }, + { + "epoch": 1.55, + "learning_rate": 2.424211336240062e-05, + "loss": 0.4198, + "step": 40222 + }, + { + "epoch": 1.55, + "learning_rate": 2.4241472172351887e-05, + "loss": 0.5077, + "step": 40223 + }, + { + "epoch": 1.55, + "learning_rate": 2.4240830982303156e-05, + "loss": 0.4046, + "step": 40224 + }, + { + "epoch": 1.55, + "learning_rate": 2.4240189792254424e-05, + "loss": 0.3146, + "step": 40225 + }, + { + "epoch": 1.55, + "learning_rate": 2.4239548602205696e-05, + "loss": 0.3937, + "step": 40226 + }, + { + "epoch": 1.55, + "learning_rate": 2.4238907412156964e-05, + "loss": 0.3449, + "step": 40227 + }, + { + "epoch": 1.55, + "learning_rate": 2.4238266222108232e-05, + "loss": 0.3396, + "step": 40228 + }, + { + "epoch": 1.55, + "learning_rate": 2.4237625032059504e-05, + "loss": 0.3556, + "step": 40229 + }, + { + "epoch": 1.55, + "learning_rate": 2.4236983842010776e-05, + "loss": 0.4332, + "step": 40230 + }, + { + "epoch": 1.55, + "learning_rate": 2.4236342651962044e-05, + "loss": 0.433, + "step": 40231 + }, + { + "epoch": 1.55, + "learning_rate": 2.4235701461913312e-05, + "loss": 0.3852, + "step": 40232 + }, + { + "epoch": 1.55, + "learning_rate": 2.423506027186458e-05, + "loss": 0.2354, + "step": 40233 + }, + { + "epoch": 1.55, + "learning_rate": 2.4234419081815852e-05, + "loss": 0.3232, + "step": 40234 + }, + { + "epoch": 1.55, + "learning_rate": 2.423377789176712e-05, + "loss": 0.3469, + "step": 40235 + }, + { + "epoch": 1.55, + "learning_rate": 2.423313670171839e-05, + "loss": 0.359, + "step": 40236 + }, + { + "epoch": 1.55, + "learning_rate": 2.4232495511669657e-05, + "loss": 0.4612, + "step": 40237 + }, + { + "epoch": 1.55, + "learning_rate": 2.4231854321620933e-05, + "loss": 0.3793, + "step": 40238 + }, + { + "epoch": 1.55, + "learning_rate": 2.42312131315722e-05, + "loss": 0.3437, + "step": 40239 + }, + { + "epoch": 1.55, + "learning_rate": 2.423057194152347e-05, + "loss": 0.4705, + "step": 40240 + }, + { + "epoch": 1.55, + "learning_rate": 2.4229930751474738e-05, + "loss": 0.3071, + "step": 40241 + }, + { + "epoch": 1.55, + "learning_rate": 2.422928956142601e-05, + "loss": 0.3828, + "step": 40242 + }, + { + "epoch": 1.55, + "learning_rate": 2.4228648371377278e-05, + "loss": 0.2867, + "step": 40243 + }, + { + "epoch": 1.55, + "learning_rate": 2.4228007181328546e-05, + "loss": 0.3203, + "step": 40244 + }, + { + "epoch": 1.55, + "learning_rate": 2.4227365991279814e-05, + "loss": 0.3553, + "step": 40245 + }, + { + "epoch": 1.55, + "learning_rate": 2.4226724801231086e-05, + "loss": 0.2621, + "step": 40246 + }, + { + "epoch": 1.55, + "learning_rate": 2.4226083611182358e-05, + "loss": 0.4029, + "step": 40247 + }, + { + "epoch": 1.55, + "learning_rate": 2.4225442421133626e-05, + "loss": 0.326, + "step": 40248 + }, + { + "epoch": 1.55, + "learning_rate": 2.4224801231084894e-05, + "loss": 0.3621, + "step": 40249 + }, + { + "epoch": 1.55, + "learning_rate": 2.4224160041036166e-05, + "loss": 0.3679, + "step": 40250 + }, + { + "epoch": 1.55, + "learning_rate": 2.4223518850987434e-05, + "loss": 0.3915, + "step": 40251 + }, + { + "epoch": 1.55, + "learning_rate": 2.4222877660938703e-05, + "loss": 0.3394, + "step": 40252 + }, + { + "epoch": 1.55, + "learning_rate": 2.422223647088997e-05, + "loss": 0.4351, + "step": 40253 + }, + { + "epoch": 1.55, + "learning_rate": 2.4221595280841243e-05, + "loss": 0.3351, + "step": 40254 + }, + { + "epoch": 1.55, + "learning_rate": 2.422095409079251e-05, + "loss": 0.3437, + "step": 40255 + }, + { + "epoch": 1.55, + "learning_rate": 2.4220312900743783e-05, + "loss": 0.3206, + "step": 40256 + }, + { + "epoch": 1.55, + "learning_rate": 2.421967171069505e-05, + "loss": 0.3351, + "step": 40257 + }, + { + "epoch": 1.55, + "learning_rate": 2.4219030520646323e-05, + "loss": 0.3239, + "step": 40258 + }, + { + "epoch": 1.55, + "learning_rate": 2.421838933059759e-05, + "loss": 0.4124, + "step": 40259 + }, + { + "epoch": 1.55, + "learning_rate": 2.421774814054886e-05, + "loss": 0.2309, + "step": 40260 + }, + { + "epoch": 1.55, + "learning_rate": 2.4217106950500128e-05, + "loss": 0.2902, + "step": 40261 + }, + { + "epoch": 1.55, + "learning_rate": 2.42164657604514e-05, + "loss": 0.2553, + "step": 40262 + }, + { + "epoch": 1.55, + "learning_rate": 2.4215824570402668e-05, + "loss": 0.2949, + "step": 40263 + }, + { + "epoch": 1.55, + "learning_rate": 2.4215183380353936e-05, + "loss": 0.3265, + "step": 40264 + }, + { + "epoch": 1.55, + "learning_rate": 2.4214542190305208e-05, + "loss": 0.4038, + "step": 40265 + }, + { + "epoch": 1.55, + "learning_rate": 2.421390100025648e-05, + "loss": 0.3648, + "step": 40266 + }, + { + "epoch": 1.55, + "learning_rate": 2.4213259810207748e-05, + "loss": 0.5128, + "step": 40267 + }, + { + "epoch": 1.55, + "learning_rate": 2.4212618620159016e-05, + "loss": 0.2128, + "step": 40268 + }, + { + "epoch": 1.55, + "learning_rate": 2.4211977430110285e-05, + "loss": 0.4639, + "step": 40269 + }, + { + "epoch": 1.55, + "learning_rate": 2.4211336240061556e-05, + "loss": 0.2228, + "step": 40270 + }, + { + "epoch": 1.55, + "learning_rate": 2.4210695050012825e-05, + "loss": 0.3444, + "step": 40271 + }, + { + "epoch": 1.55, + "learning_rate": 2.4210053859964093e-05, + "loss": 0.231, + "step": 40272 + }, + { + "epoch": 1.55, + "learning_rate": 2.420941266991536e-05, + "loss": 0.4405, + "step": 40273 + }, + { + "epoch": 1.55, + "learning_rate": 2.4208771479866636e-05, + "loss": 0.3249, + "step": 40274 + }, + { + "epoch": 1.55, + "learning_rate": 2.4208130289817905e-05, + "loss": 0.4068, + "step": 40275 + }, + { + "epoch": 1.55, + "learning_rate": 2.4207489099769173e-05, + "loss": 0.3201, + "step": 40276 + }, + { + "epoch": 1.55, + "learning_rate": 2.420684790972044e-05, + "loss": 0.2969, + "step": 40277 + }, + { + "epoch": 1.55, + "learning_rate": 2.4206206719671713e-05, + "loss": 0.3, + "step": 40278 + }, + { + "epoch": 1.55, + "learning_rate": 2.420556552962298e-05, + "loss": 0.4502, + "step": 40279 + }, + { + "epoch": 1.55, + "learning_rate": 2.420492433957425e-05, + "loss": 0.3096, + "step": 40280 + }, + { + "epoch": 1.55, + "learning_rate": 2.4204283149525518e-05, + "loss": 0.2671, + "step": 40281 + }, + { + "epoch": 1.55, + "learning_rate": 2.420364195947679e-05, + "loss": 0.2659, + "step": 40282 + }, + { + "epoch": 1.55, + "learning_rate": 2.420300076942806e-05, + "loss": 0.2206, + "step": 40283 + }, + { + "epoch": 1.55, + "learning_rate": 2.420235957937933e-05, + "loss": 0.2565, + "step": 40284 + }, + { + "epoch": 1.55, + "learning_rate": 2.4201718389330598e-05, + "loss": 0.355, + "step": 40285 + }, + { + "epoch": 1.55, + "learning_rate": 2.420107719928187e-05, + "loss": 0.2414, + "step": 40286 + }, + { + "epoch": 1.55, + "learning_rate": 2.4200436009233138e-05, + "loss": 0.3169, + "step": 40287 + }, + { + "epoch": 1.55, + "learning_rate": 2.4199794819184407e-05, + "loss": 0.248, + "step": 40288 + }, + { + "epoch": 1.55, + "learning_rate": 2.4199153629135675e-05, + "loss": 0.354, + "step": 40289 + }, + { + "epoch": 1.55, + "learning_rate": 2.4198512439086947e-05, + "loss": 0.3866, + "step": 40290 + }, + { + "epoch": 1.55, + "learning_rate": 2.4197871249038215e-05, + "loss": 0.3023, + "step": 40291 + }, + { + "epoch": 1.55, + "learning_rate": 2.4197230058989487e-05, + "loss": 0.3448, + "step": 40292 + }, + { + "epoch": 1.55, + "learning_rate": 2.4196588868940755e-05, + "loss": 0.4197, + "step": 40293 + }, + { + "epoch": 1.55, + "learning_rate": 2.4195947678892027e-05, + "loss": 0.3588, + "step": 40294 + }, + { + "epoch": 1.55, + "learning_rate": 2.4195306488843295e-05, + "loss": 0.6149, + "step": 40295 + }, + { + "epoch": 1.55, + "learning_rate": 2.4194665298794563e-05, + "loss": 0.3254, + "step": 40296 + }, + { + "epoch": 1.55, + "learning_rate": 2.4194024108745832e-05, + "loss": 0.3414, + "step": 40297 + }, + { + "epoch": 1.55, + "learning_rate": 2.4193382918697103e-05, + "loss": 0.4441, + "step": 40298 + }, + { + "epoch": 1.55, + "learning_rate": 2.4192741728648372e-05, + "loss": 0.3471, + "step": 40299 + }, + { + "epoch": 1.55, + "learning_rate": 2.419210053859964e-05, + "loss": 0.3722, + "step": 40300 + }, + { + "epoch": 1.55, + "learning_rate": 2.4191459348550912e-05, + "loss": 0.357, + "step": 40301 + }, + { + "epoch": 1.55, + "learning_rate": 2.4190818158502184e-05, + "loss": 0.5344, + "step": 40302 + }, + { + "epoch": 1.55, + "learning_rate": 2.4190176968453452e-05, + "loss": 0.3892, + "step": 40303 + }, + { + "epoch": 1.55, + "learning_rate": 2.418953577840472e-05, + "loss": 0.3028, + "step": 40304 + }, + { + "epoch": 1.55, + "learning_rate": 2.418889458835599e-05, + "loss": 0.3501, + "step": 40305 + }, + { + "epoch": 1.55, + "learning_rate": 2.418825339830726e-05, + "loss": 0.3388, + "step": 40306 + }, + { + "epoch": 1.55, + "learning_rate": 2.418761220825853e-05, + "loss": 0.396, + "step": 40307 + }, + { + "epoch": 1.55, + "learning_rate": 2.4186971018209797e-05, + "loss": 0.2955, + "step": 40308 + }, + { + "epoch": 1.55, + "learning_rate": 2.4186329828161065e-05, + "loss": 0.413, + "step": 40309 + }, + { + "epoch": 1.55, + "learning_rate": 2.418568863811234e-05, + "loss": 0.3921, + "step": 40310 + }, + { + "epoch": 1.55, + "learning_rate": 2.418504744806361e-05, + "loss": 0.4478, + "step": 40311 + }, + { + "epoch": 1.55, + "learning_rate": 2.4184406258014877e-05, + "loss": 0.28, + "step": 40312 + }, + { + "epoch": 1.55, + "learning_rate": 2.4183765067966145e-05, + "loss": 0.3344, + "step": 40313 + }, + { + "epoch": 1.55, + "learning_rate": 2.4183123877917417e-05, + "loss": 0.3501, + "step": 40314 + }, + { + "epoch": 1.55, + "learning_rate": 2.4182482687868685e-05, + "loss": 0.538, + "step": 40315 + }, + { + "epoch": 1.55, + "learning_rate": 2.4181841497819954e-05, + "loss": 0.3894, + "step": 40316 + }, + { + "epoch": 1.55, + "learning_rate": 2.4181200307771222e-05, + "loss": 0.328, + "step": 40317 + }, + { + "epoch": 1.55, + "learning_rate": 2.4180559117722497e-05, + "loss": 0.3845, + "step": 40318 + }, + { + "epoch": 1.55, + "learning_rate": 2.4179917927673765e-05, + "loss": 0.2777, + "step": 40319 + }, + { + "epoch": 1.55, + "learning_rate": 2.4179276737625034e-05, + "loss": 0.2269, + "step": 40320 + }, + { + "epoch": 1.55, + "learning_rate": 2.4178635547576302e-05, + "loss": 0.2152, + "step": 40321 + }, + { + "epoch": 1.55, + "learning_rate": 2.4177994357527574e-05, + "loss": 0.261, + "step": 40322 + }, + { + "epoch": 1.55, + "learning_rate": 2.4177353167478842e-05, + "loss": 0.4447, + "step": 40323 + }, + { + "epoch": 1.55, + "learning_rate": 2.417671197743011e-05, + "loss": 0.3727, + "step": 40324 + }, + { + "epoch": 1.55, + "learning_rate": 2.417607078738138e-05, + "loss": 0.2411, + "step": 40325 + }, + { + "epoch": 1.55, + "learning_rate": 2.417542959733265e-05, + "loss": 0.3255, + "step": 40326 + }, + { + "epoch": 1.55, + "learning_rate": 2.4174788407283922e-05, + "loss": 0.4155, + "step": 40327 + }, + { + "epoch": 1.55, + "learning_rate": 2.417414721723519e-05, + "loss": 0.3339, + "step": 40328 + }, + { + "epoch": 1.55, + "learning_rate": 2.417350602718646e-05, + "loss": 0.3335, + "step": 40329 + }, + { + "epoch": 1.55, + "learning_rate": 2.417286483713773e-05, + "loss": 0.4586, + "step": 40330 + }, + { + "epoch": 1.55, + "learning_rate": 2.4172223647089e-05, + "loss": 0.2711, + "step": 40331 + }, + { + "epoch": 1.55, + "learning_rate": 2.4171582457040267e-05, + "loss": 0.2988, + "step": 40332 + }, + { + "epoch": 1.55, + "learning_rate": 2.4170941266991536e-05, + "loss": 0.2634, + "step": 40333 + }, + { + "epoch": 1.55, + "learning_rate": 2.4170300076942807e-05, + "loss": 0.4427, + "step": 40334 + }, + { + "epoch": 1.55, + "learning_rate": 2.4169658886894076e-05, + "loss": 0.345, + "step": 40335 + }, + { + "epoch": 1.55, + "learning_rate": 2.4169017696845347e-05, + "loss": 0.2809, + "step": 40336 + }, + { + "epoch": 1.55, + "learning_rate": 2.4168376506796616e-05, + "loss": 0.2587, + "step": 40337 + }, + { + "epoch": 1.55, + "learning_rate": 2.4167735316747887e-05, + "loss": 0.4333, + "step": 40338 + }, + { + "epoch": 1.55, + "learning_rate": 2.4167094126699156e-05, + "loss": 0.4687, + "step": 40339 + }, + { + "epoch": 1.55, + "learning_rate": 2.4166452936650424e-05, + "loss": 0.2756, + "step": 40340 + }, + { + "epoch": 1.55, + "learning_rate": 2.4165811746601692e-05, + "loss": 0.4867, + "step": 40341 + }, + { + "epoch": 1.55, + "learning_rate": 2.4165170556552964e-05, + "loss": 0.292, + "step": 40342 + }, + { + "epoch": 1.55, + "learning_rate": 2.4164529366504232e-05, + "loss": 0.2596, + "step": 40343 + }, + { + "epoch": 1.55, + "learning_rate": 2.41638881764555e-05, + "loss": 0.3651, + "step": 40344 + }, + { + "epoch": 1.55, + "learning_rate": 2.4163246986406772e-05, + "loss": 0.4247, + "step": 40345 + }, + { + "epoch": 1.55, + "learning_rate": 2.4162605796358044e-05, + "loss": 0.3641, + "step": 40346 + }, + { + "epoch": 1.55, + "learning_rate": 2.4161964606309313e-05, + "loss": 0.4059, + "step": 40347 + }, + { + "epoch": 1.55, + "learning_rate": 2.416132341626058e-05, + "loss": 0.3667, + "step": 40348 + }, + { + "epoch": 1.55, + "learning_rate": 2.416068222621185e-05, + "loss": 0.338, + "step": 40349 + }, + { + "epoch": 1.55, + "learning_rate": 2.416004103616312e-05, + "loss": 0.2703, + "step": 40350 + }, + { + "epoch": 1.55, + "learning_rate": 2.415939984611439e-05, + "loss": 0.2978, + "step": 40351 + }, + { + "epoch": 1.55, + "learning_rate": 2.4158758656065658e-05, + "loss": 0.3198, + "step": 40352 + }, + { + "epoch": 1.55, + "learning_rate": 2.4158117466016926e-05, + "loss": 0.2766, + "step": 40353 + }, + { + "epoch": 1.55, + "learning_rate": 2.41574762759682e-05, + "loss": 0.4105, + "step": 40354 + }, + { + "epoch": 1.55, + "learning_rate": 2.415683508591947e-05, + "loss": 0.3146, + "step": 40355 + }, + { + "epoch": 1.55, + "learning_rate": 2.4156193895870738e-05, + "loss": 0.3573, + "step": 40356 + }, + { + "epoch": 1.55, + "learning_rate": 2.4155552705822006e-05, + "loss": 0.3352, + "step": 40357 + }, + { + "epoch": 1.55, + "learning_rate": 2.4154911515773278e-05, + "loss": 0.3278, + "step": 40358 + }, + { + "epoch": 1.55, + "learning_rate": 2.4154270325724546e-05, + "loss": 0.3317, + "step": 40359 + }, + { + "epoch": 1.55, + "learning_rate": 2.4153629135675814e-05, + "loss": 0.477, + "step": 40360 + }, + { + "epoch": 1.55, + "learning_rate": 2.4152987945627083e-05, + "loss": 0.2303, + "step": 40361 + }, + { + "epoch": 1.55, + "learning_rate": 2.4152346755578354e-05, + "loss": 0.2645, + "step": 40362 + }, + { + "epoch": 1.55, + "learning_rate": 2.4151705565529626e-05, + "loss": 0.397, + "step": 40363 + }, + { + "epoch": 1.55, + "learning_rate": 2.4151064375480894e-05, + "loss": 0.4302, + "step": 40364 + }, + { + "epoch": 1.55, + "learning_rate": 2.4150423185432163e-05, + "loss": 0.3649, + "step": 40365 + }, + { + "epoch": 1.55, + "learning_rate": 2.4149781995383434e-05, + "loss": 0.4315, + "step": 40366 + }, + { + "epoch": 1.55, + "learning_rate": 2.4149140805334703e-05, + "loss": 0.3316, + "step": 40367 + }, + { + "epoch": 1.55, + "learning_rate": 2.414849961528597e-05, + "loss": 0.332, + "step": 40368 + }, + { + "epoch": 1.55, + "learning_rate": 2.414785842523724e-05, + "loss": 0.4238, + "step": 40369 + }, + { + "epoch": 1.55, + "learning_rate": 2.414721723518851e-05, + "loss": 0.4032, + "step": 40370 + }, + { + "epoch": 1.55, + "learning_rate": 2.414657604513978e-05, + "loss": 0.3697, + "step": 40371 + }, + { + "epoch": 1.55, + "learning_rate": 2.414593485509105e-05, + "loss": 0.3462, + "step": 40372 + }, + { + "epoch": 1.55, + "learning_rate": 2.414529366504232e-05, + "loss": 0.3614, + "step": 40373 + }, + { + "epoch": 1.55, + "learning_rate": 2.414465247499359e-05, + "loss": 0.3941, + "step": 40374 + }, + { + "epoch": 1.55, + "learning_rate": 2.414401128494486e-05, + "loss": 0.403, + "step": 40375 + }, + { + "epoch": 1.55, + "learning_rate": 2.4143370094896128e-05, + "loss": 0.2971, + "step": 40376 + }, + { + "epoch": 1.55, + "learning_rate": 2.4142728904847396e-05, + "loss": 0.4824, + "step": 40377 + }, + { + "epoch": 1.55, + "learning_rate": 2.4142087714798668e-05, + "loss": 0.4803, + "step": 40378 + }, + { + "epoch": 1.55, + "learning_rate": 2.4141446524749936e-05, + "loss": 0.362, + "step": 40379 + }, + { + "epoch": 1.55, + "learning_rate": 2.4140805334701205e-05, + "loss": 0.3199, + "step": 40380 + }, + { + "epoch": 1.55, + "learning_rate": 2.4140164144652476e-05, + "loss": 0.3982, + "step": 40381 + }, + { + "epoch": 1.55, + "learning_rate": 2.4139522954603748e-05, + "loss": 0.2883, + "step": 40382 + }, + { + "epoch": 1.55, + "learning_rate": 2.4138881764555016e-05, + "loss": 0.2584, + "step": 40383 + }, + { + "epoch": 1.55, + "learning_rate": 2.4138240574506285e-05, + "loss": 0.3705, + "step": 40384 + }, + { + "epoch": 1.55, + "learning_rate": 2.4137599384457553e-05, + "loss": 0.418, + "step": 40385 + }, + { + "epoch": 1.55, + "learning_rate": 2.4136958194408825e-05, + "loss": 0.4737, + "step": 40386 + }, + { + "epoch": 1.55, + "learning_rate": 2.4136317004360093e-05, + "loss": 0.3732, + "step": 40387 + }, + { + "epoch": 1.55, + "learning_rate": 2.413567581431136e-05, + "loss": 0.266, + "step": 40388 + }, + { + "epoch": 1.55, + "learning_rate": 2.413503462426263e-05, + "loss": 0.3259, + "step": 40389 + }, + { + "epoch": 1.55, + "learning_rate": 2.4134393434213905e-05, + "loss": 0.3376, + "step": 40390 + }, + { + "epoch": 1.55, + "learning_rate": 2.4133752244165173e-05, + "loss": 0.2892, + "step": 40391 + }, + { + "epoch": 1.55, + "learning_rate": 2.413311105411644e-05, + "loss": 0.3836, + "step": 40392 + }, + { + "epoch": 1.55, + "learning_rate": 2.413246986406771e-05, + "loss": 0.4094, + "step": 40393 + }, + { + "epoch": 1.55, + "learning_rate": 2.413182867401898e-05, + "loss": 0.4116, + "step": 40394 + }, + { + "epoch": 1.55, + "learning_rate": 2.413118748397025e-05, + "loss": 0.3679, + "step": 40395 + }, + { + "epoch": 1.55, + "learning_rate": 2.4130546293921518e-05, + "loss": 0.3392, + "step": 40396 + }, + { + "epoch": 1.55, + "learning_rate": 2.4129905103872787e-05, + "loss": 0.4582, + "step": 40397 + }, + { + "epoch": 1.55, + "learning_rate": 2.4129263913824058e-05, + "loss": 0.2819, + "step": 40398 + }, + { + "epoch": 1.55, + "learning_rate": 2.412862272377533e-05, + "loss": 0.3175, + "step": 40399 + }, + { + "epoch": 1.55, + "learning_rate": 2.4127981533726598e-05, + "loss": 0.254, + "step": 40400 + }, + { + "epoch": 1.55, + "learning_rate": 2.4127340343677867e-05, + "loss": 0.3896, + "step": 40401 + }, + { + "epoch": 1.55, + "learning_rate": 2.412669915362914e-05, + "loss": 0.3821, + "step": 40402 + }, + { + "epoch": 1.55, + "learning_rate": 2.4126057963580407e-05, + "loss": 0.4389, + "step": 40403 + }, + { + "epoch": 1.55, + "learning_rate": 2.4125416773531675e-05, + "loss": 0.3908, + "step": 40404 + }, + { + "epoch": 1.55, + "learning_rate": 2.4124775583482943e-05, + "loss": 0.4373, + "step": 40405 + }, + { + "epoch": 1.55, + "learning_rate": 2.4124134393434215e-05, + "loss": 0.3957, + "step": 40406 + }, + { + "epoch": 1.55, + "learning_rate": 2.4123493203385483e-05, + "loss": 0.5101, + "step": 40407 + }, + { + "epoch": 1.55, + "learning_rate": 2.4122852013336755e-05, + "loss": 0.3267, + "step": 40408 + }, + { + "epoch": 1.55, + "learning_rate": 2.4122210823288023e-05, + "loss": 0.3224, + "step": 40409 + }, + { + "epoch": 1.55, + "learning_rate": 2.4121569633239295e-05, + "loss": 0.4155, + "step": 40410 + }, + { + "epoch": 1.55, + "learning_rate": 2.4120928443190563e-05, + "loss": 0.3402, + "step": 40411 + }, + { + "epoch": 1.55, + "learning_rate": 2.4120287253141832e-05, + "loss": 0.3144, + "step": 40412 + }, + { + "epoch": 1.55, + "learning_rate": 2.41196460630931e-05, + "loss": 0.4021, + "step": 40413 + }, + { + "epoch": 1.55, + "learning_rate": 2.4119004873044372e-05, + "loss": 0.2925, + "step": 40414 + }, + { + "epoch": 1.55, + "learning_rate": 2.411836368299564e-05, + "loss": 0.3921, + "step": 40415 + }, + { + "epoch": 1.55, + "learning_rate": 2.4117722492946912e-05, + "loss": 0.2014, + "step": 40416 + }, + { + "epoch": 1.55, + "learning_rate": 2.411708130289818e-05, + "loss": 0.3453, + "step": 40417 + }, + { + "epoch": 1.55, + "learning_rate": 2.4116440112849452e-05, + "loss": 0.3071, + "step": 40418 + }, + { + "epoch": 1.55, + "learning_rate": 2.411579892280072e-05, + "loss": 0.3278, + "step": 40419 + }, + { + "epoch": 1.55, + "learning_rate": 2.411515773275199e-05, + "loss": 0.3954, + "step": 40420 + }, + { + "epoch": 1.55, + "learning_rate": 2.4114516542703257e-05, + "loss": 0.3871, + "step": 40421 + }, + { + "epoch": 1.55, + "learning_rate": 2.411387535265453e-05, + "loss": 0.3184, + "step": 40422 + }, + { + "epoch": 1.55, + "learning_rate": 2.4113234162605797e-05, + "loss": 0.3563, + "step": 40423 + }, + { + "epoch": 1.55, + "learning_rate": 2.4112592972557065e-05, + "loss": 0.3959, + "step": 40424 + }, + { + "epoch": 1.55, + "learning_rate": 2.4111951782508337e-05, + "loss": 0.4125, + "step": 40425 + }, + { + "epoch": 1.55, + "learning_rate": 2.411131059245961e-05, + "loss": 0.2633, + "step": 40426 + }, + { + "epoch": 1.55, + "learning_rate": 2.4110669402410877e-05, + "loss": 0.5116, + "step": 40427 + }, + { + "epoch": 1.55, + "learning_rate": 2.4110028212362145e-05, + "loss": 0.3136, + "step": 40428 + }, + { + "epoch": 1.55, + "learning_rate": 2.4109387022313414e-05, + "loss": 0.4895, + "step": 40429 + }, + { + "epoch": 1.55, + "learning_rate": 2.4108745832264685e-05, + "loss": 0.294, + "step": 40430 + }, + { + "epoch": 1.55, + "learning_rate": 2.4108104642215954e-05, + "loss": 0.441, + "step": 40431 + }, + { + "epoch": 1.55, + "learning_rate": 2.4107463452167222e-05, + "loss": 0.412, + "step": 40432 + }, + { + "epoch": 1.55, + "learning_rate": 2.410682226211849e-05, + "loss": 0.3079, + "step": 40433 + }, + { + "epoch": 1.55, + "learning_rate": 2.4106181072069766e-05, + "loss": 0.2899, + "step": 40434 + }, + { + "epoch": 1.55, + "learning_rate": 2.4105539882021034e-05, + "loss": 0.3635, + "step": 40435 + }, + { + "epoch": 1.55, + "learning_rate": 2.4104898691972302e-05, + "loss": 0.2238, + "step": 40436 + }, + { + "epoch": 1.55, + "learning_rate": 2.410425750192357e-05, + "loss": 0.4376, + "step": 40437 + }, + { + "epoch": 1.55, + "learning_rate": 2.4103616311874842e-05, + "loss": 0.4125, + "step": 40438 + }, + { + "epoch": 1.55, + "learning_rate": 2.410297512182611e-05, + "loss": 0.381, + "step": 40439 + }, + { + "epoch": 1.55, + "learning_rate": 2.410233393177738e-05, + "loss": 0.4722, + "step": 40440 + }, + { + "epoch": 1.55, + "learning_rate": 2.4101692741728647e-05, + "loss": 0.3605, + "step": 40441 + }, + { + "epoch": 1.55, + "learning_rate": 2.410105155167992e-05, + "loss": 0.2578, + "step": 40442 + }, + { + "epoch": 1.55, + "learning_rate": 2.410041036163119e-05, + "loss": 0.4105, + "step": 40443 + }, + { + "epoch": 1.55, + "learning_rate": 2.409976917158246e-05, + "loss": 0.3415, + "step": 40444 + }, + { + "epoch": 1.55, + "learning_rate": 2.4099127981533727e-05, + "loss": 0.2941, + "step": 40445 + }, + { + "epoch": 1.56, + "learning_rate": 2.4098486791485e-05, + "loss": 0.3184, + "step": 40446 + }, + { + "epoch": 1.56, + "learning_rate": 2.4097845601436267e-05, + "loss": 0.3863, + "step": 40447 + }, + { + "epoch": 1.56, + "learning_rate": 2.4097204411387536e-05, + "loss": 0.3949, + "step": 40448 + }, + { + "epoch": 1.56, + "learning_rate": 2.4096563221338804e-05, + "loss": 0.3899, + "step": 40449 + }, + { + "epoch": 1.56, + "learning_rate": 2.4095922031290076e-05, + "loss": 0.3015, + "step": 40450 + }, + { + "epoch": 1.56, + "learning_rate": 2.4095280841241344e-05, + "loss": 0.2314, + "step": 40451 + }, + { + "epoch": 1.56, + "learning_rate": 2.4094639651192616e-05, + "loss": 0.4368, + "step": 40452 + }, + { + "epoch": 1.56, + "learning_rate": 2.4093998461143884e-05, + "loss": 0.3066, + "step": 40453 + }, + { + "epoch": 1.56, + "learning_rate": 2.4093357271095156e-05, + "loss": 0.4284, + "step": 40454 + }, + { + "epoch": 1.56, + "learning_rate": 2.4092716081046424e-05, + "loss": 0.3127, + "step": 40455 + }, + { + "epoch": 1.56, + "learning_rate": 2.4092074890997692e-05, + "loss": 0.28, + "step": 40456 + }, + { + "epoch": 1.56, + "learning_rate": 2.409143370094896e-05, + "loss": 0.373, + "step": 40457 + }, + { + "epoch": 1.56, + "learning_rate": 2.4090792510900232e-05, + "loss": 0.3809, + "step": 40458 + }, + { + "epoch": 1.56, + "learning_rate": 2.40901513208515e-05, + "loss": 0.4149, + "step": 40459 + }, + { + "epoch": 1.56, + "learning_rate": 2.408951013080277e-05, + "loss": 0.4824, + "step": 40460 + }, + { + "epoch": 1.56, + "learning_rate": 2.408886894075404e-05, + "loss": 0.5894, + "step": 40461 + }, + { + "epoch": 1.56, + "learning_rate": 2.4088227750705313e-05, + "loss": 0.332, + "step": 40462 + }, + { + "epoch": 1.56, + "learning_rate": 2.408758656065658e-05, + "loss": 0.4144, + "step": 40463 + }, + { + "epoch": 1.56, + "learning_rate": 2.408694537060785e-05, + "loss": 0.33, + "step": 40464 + }, + { + "epoch": 1.56, + "learning_rate": 2.4086304180559118e-05, + "loss": 0.3174, + "step": 40465 + }, + { + "epoch": 1.56, + "learning_rate": 2.408566299051039e-05, + "loss": 0.4634, + "step": 40466 + }, + { + "epoch": 1.56, + "learning_rate": 2.4085021800461658e-05, + "loss": 0.2858, + "step": 40467 + }, + { + "epoch": 1.56, + "learning_rate": 2.4084380610412926e-05, + "loss": 0.3779, + "step": 40468 + }, + { + "epoch": 1.56, + "learning_rate": 2.4083739420364194e-05, + "loss": 0.4697, + "step": 40469 + }, + { + "epoch": 1.56, + "learning_rate": 2.408309823031547e-05, + "loss": 0.2224, + "step": 40470 + }, + { + "epoch": 1.56, + "learning_rate": 2.4082457040266738e-05, + "loss": 0.401, + "step": 40471 + }, + { + "epoch": 1.56, + "learning_rate": 2.4081815850218006e-05, + "loss": 0.3854, + "step": 40472 + }, + { + "epoch": 1.56, + "learning_rate": 2.4081174660169274e-05, + "loss": 0.4135, + "step": 40473 + }, + { + "epoch": 1.56, + "learning_rate": 2.4080533470120546e-05, + "loss": 0.4448, + "step": 40474 + }, + { + "epoch": 1.56, + "learning_rate": 2.4079892280071814e-05, + "loss": 0.287, + "step": 40475 + }, + { + "epoch": 1.56, + "learning_rate": 2.4079251090023083e-05, + "loss": 0.3622, + "step": 40476 + }, + { + "epoch": 1.56, + "learning_rate": 2.407860989997435e-05, + "loss": 0.3693, + "step": 40477 + }, + { + "epoch": 1.56, + "learning_rate": 2.4077968709925623e-05, + "loss": 0.2756, + "step": 40478 + }, + { + "epoch": 1.56, + "learning_rate": 2.4077327519876895e-05, + "loss": 0.4837, + "step": 40479 + }, + { + "epoch": 1.56, + "learning_rate": 2.4076686329828163e-05, + "loss": 0.3347, + "step": 40480 + }, + { + "epoch": 1.56, + "learning_rate": 2.407604513977943e-05, + "loss": 0.2912, + "step": 40481 + }, + { + "epoch": 1.56, + "learning_rate": 2.4075403949730703e-05, + "loss": 0.2966, + "step": 40482 + }, + { + "epoch": 1.56, + "learning_rate": 2.407476275968197e-05, + "loss": 0.3041, + "step": 40483 + }, + { + "epoch": 1.56, + "learning_rate": 2.407412156963324e-05, + "loss": 0.2928, + "step": 40484 + }, + { + "epoch": 1.56, + "learning_rate": 2.4073480379584508e-05, + "loss": 0.2933, + "step": 40485 + }, + { + "epoch": 1.56, + "learning_rate": 2.407283918953578e-05, + "loss": 0.3617, + "step": 40486 + }, + { + "epoch": 1.56, + "learning_rate": 2.4072197999487048e-05, + "loss": 0.3412, + "step": 40487 + }, + { + "epoch": 1.56, + "learning_rate": 2.407155680943832e-05, + "loss": 0.438, + "step": 40488 + }, + { + "epoch": 1.56, + "learning_rate": 2.4070915619389588e-05, + "loss": 0.3405, + "step": 40489 + }, + { + "epoch": 1.56, + "learning_rate": 2.407027442934086e-05, + "loss": 0.2875, + "step": 40490 + }, + { + "epoch": 1.56, + "learning_rate": 2.4069633239292128e-05, + "loss": 0.2847, + "step": 40491 + }, + { + "epoch": 1.56, + "learning_rate": 2.4068992049243396e-05, + "loss": 0.4016, + "step": 40492 + }, + { + "epoch": 1.56, + "learning_rate": 2.4068350859194665e-05, + "loss": 0.3212, + "step": 40493 + }, + { + "epoch": 1.56, + "learning_rate": 2.4067709669145936e-05, + "loss": 0.3057, + "step": 40494 + }, + { + "epoch": 1.56, + "learning_rate": 2.4067068479097205e-05, + "loss": 0.3541, + "step": 40495 + }, + { + "epoch": 1.56, + "learning_rate": 2.4066427289048473e-05, + "loss": 0.3938, + "step": 40496 + }, + { + "epoch": 1.56, + "learning_rate": 2.4065786098999745e-05, + "loss": 0.3956, + "step": 40497 + }, + { + "epoch": 1.56, + "learning_rate": 2.4065144908951016e-05, + "loss": 0.3234, + "step": 40498 + }, + { + "epoch": 1.56, + "learning_rate": 2.4064503718902285e-05, + "loss": 0.3902, + "step": 40499 + }, + { + "epoch": 1.56, + "learning_rate": 2.4063862528853553e-05, + "loss": 0.2939, + "step": 40500 + }, + { + "epoch": 1.56, + "learning_rate": 2.406322133880482e-05, + "loss": 0.2826, + "step": 40501 + }, + { + "epoch": 1.56, + "learning_rate": 2.4062580148756093e-05, + "loss": 0.3993, + "step": 40502 + }, + { + "epoch": 1.56, + "learning_rate": 2.406193895870736e-05, + "loss": 0.3324, + "step": 40503 + }, + { + "epoch": 1.56, + "learning_rate": 2.406129776865863e-05, + "loss": 0.3181, + "step": 40504 + }, + { + "epoch": 1.56, + "learning_rate": 2.4060656578609898e-05, + "loss": 0.4412, + "step": 40505 + }, + { + "epoch": 1.56, + "learning_rate": 2.4060015388561173e-05, + "loss": 0.2891, + "step": 40506 + }, + { + "epoch": 1.56, + "learning_rate": 2.405937419851244e-05, + "loss": 0.364, + "step": 40507 + }, + { + "epoch": 1.56, + "learning_rate": 2.405873300846371e-05, + "loss": 0.3511, + "step": 40508 + }, + { + "epoch": 1.56, + "learning_rate": 2.4058091818414978e-05, + "loss": 0.4561, + "step": 40509 + }, + { + "epoch": 1.56, + "learning_rate": 2.405745062836625e-05, + "loss": 0.3831, + "step": 40510 + }, + { + "epoch": 1.56, + "learning_rate": 2.4056809438317518e-05, + "loss": 0.4214, + "step": 40511 + }, + { + "epoch": 1.56, + "learning_rate": 2.4056168248268787e-05, + "loss": 0.3802, + "step": 40512 + }, + { + "epoch": 1.56, + "learning_rate": 2.4055527058220055e-05, + "loss": 0.3727, + "step": 40513 + }, + { + "epoch": 1.56, + "learning_rate": 2.405488586817133e-05, + "loss": 0.4165, + "step": 40514 + }, + { + "epoch": 1.56, + "learning_rate": 2.40542446781226e-05, + "loss": 0.3646, + "step": 40515 + }, + { + "epoch": 1.56, + "learning_rate": 2.4053603488073867e-05, + "loss": 0.1929, + "step": 40516 + }, + { + "epoch": 1.56, + "learning_rate": 2.4052962298025135e-05, + "loss": 0.2738, + "step": 40517 + }, + { + "epoch": 1.56, + "learning_rate": 2.4052321107976407e-05, + "loss": 0.4157, + "step": 40518 + }, + { + "epoch": 1.56, + "learning_rate": 2.4051679917927675e-05, + "loss": 0.5562, + "step": 40519 + }, + { + "epoch": 1.56, + "learning_rate": 2.4051038727878943e-05, + "loss": 0.3459, + "step": 40520 + }, + { + "epoch": 1.56, + "learning_rate": 2.4050397537830212e-05, + "loss": 0.2435, + "step": 40521 + }, + { + "epoch": 1.56, + "learning_rate": 2.4049756347781483e-05, + "loss": 0.4599, + "step": 40522 + }, + { + "epoch": 1.56, + "learning_rate": 2.4049115157732755e-05, + "loss": 0.4585, + "step": 40523 + }, + { + "epoch": 1.56, + "learning_rate": 2.4048473967684023e-05, + "loss": 0.3015, + "step": 40524 + }, + { + "epoch": 1.56, + "learning_rate": 2.4047832777635292e-05, + "loss": 0.2097, + "step": 40525 + }, + { + "epoch": 1.56, + "learning_rate": 2.4047191587586564e-05, + "loss": 0.4133, + "step": 40526 + }, + { + "epoch": 1.56, + "learning_rate": 2.4046550397537832e-05, + "loss": 0.3067, + "step": 40527 + }, + { + "epoch": 1.56, + "learning_rate": 2.40459092074891e-05, + "loss": 0.4705, + "step": 40528 + }, + { + "epoch": 1.56, + "learning_rate": 2.404526801744037e-05, + "loss": 0.2891, + "step": 40529 + }, + { + "epoch": 1.56, + "learning_rate": 2.404462682739164e-05, + "loss": 0.3965, + "step": 40530 + }, + { + "epoch": 1.56, + "learning_rate": 2.404398563734291e-05, + "loss": 0.3951, + "step": 40531 + }, + { + "epoch": 1.56, + "learning_rate": 2.404334444729418e-05, + "loss": 0.3994, + "step": 40532 + }, + { + "epoch": 1.56, + "learning_rate": 2.404270325724545e-05, + "loss": 0.5107, + "step": 40533 + }, + { + "epoch": 1.56, + "learning_rate": 2.404206206719672e-05, + "loss": 0.4855, + "step": 40534 + }, + { + "epoch": 1.56, + "learning_rate": 2.404142087714799e-05, + "loss": 0.3928, + "step": 40535 + }, + { + "epoch": 1.56, + "learning_rate": 2.4040779687099257e-05, + "loss": 0.4295, + "step": 40536 + }, + { + "epoch": 1.56, + "learning_rate": 2.4040138497050525e-05, + "loss": 0.3239, + "step": 40537 + }, + { + "epoch": 1.56, + "learning_rate": 2.4039497307001797e-05, + "loss": 0.4217, + "step": 40538 + }, + { + "epoch": 1.56, + "learning_rate": 2.4038856116953065e-05, + "loss": 0.3819, + "step": 40539 + }, + { + "epoch": 1.56, + "learning_rate": 2.4038214926904334e-05, + "loss": 0.2729, + "step": 40540 + }, + { + "epoch": 1.56, + "learning_rate": 2.4037573736855605e-05, + "loss": 0.4315, + "step": 40541 + }, + { + "epoch": 1.56, + "learning_rate": 2.4036932546806877e-05, + "loss": 0.3255, + "step": 40542 + }, + { + "epoch": 1.56, + "learning_rate": 2.4036291356758145e-05, + "loss": 0.3579, + "step": 40543 + }, + { + "epoch": 1.56, + "learning_rate": 2.4035650166709414e-05, + "loss": 0.2995, + "step": 40544 + }, + { + "epoch": 1.56, + "learning_rate": 2.4035008976660682e-05, + "loss": 0.2316, + "step": 40545 + }, + { + "epoch": 1.56, + "learning_rate": 2.4034367786611954e-05, + "loss": 0.3066, + "step": 40546 + }, + { + "epoch": 1.56, + "learning_rate": 2.4033726596563222e-05, + "loss": 0.4877, + "step": 40547 + }, + { + "epoch": 1.56, + "learning_rate": 2.403308540651449e-05, + "loss": 0.35, + "step": 40548 + }, + { + "epoch": 1.56, + "learning_rate": 2.403244421646576e-05, + "loss": 0.4507, + "step": 40549 + }, + { + "epoch": 1.56, + "learning_rate": 2.4031803026417034e-05, + "loss": 0.3047, + "step": 40550 + }, + { + "epoch": 1.56, + "learning_rate": 2.4031161836368302e-05, + "loss": 0.3535, + "step": 40551 + }, + { + "epoch": 1.56, + "learning_rate": 2.403052064631957e-05, + "loss": 0.2815, + "step": 40552 + }, + { + "epoch": 1.56, + "learning_rate": 2.402987945627084e-05, + "loss": 0.3326, + "step": 40553 + }, + { + "epoch": 1.56, + "learning_rate": 2.402923826622211e-05, + "loss": 0.4292, + "step": 40554 + }, + { + "epoch": 1.56, + "learning_rate": 2.402859707617338e-05, + "loss": 0.4449, + "step": 40555 + }, + { + "epoch": 1.56, + "learning_rate": 2.4027955886124647e-05, + "loss": 0.3584, + "step": 40556 + }, + { + "epoch": 1.56, + "learning_rate": 2.4027314696075916e-05, + "loss": 0.396, + "step": 40557 + }, + { + "epoch": 1.56, + "learning_rate": 2.4026673506027187e-05, + "loss": 0.4522, + "step": 40558 + }, + { + "epoch": 1.56, + "learning_rate": 2.402603231597846e-05, + "loss": 0.3747, + "step": 40559 + }, + { + "epoch": 1.56, + "learning_rate": 2.4025391125929727e-05, + "loss": 0.3823, + "step": 40560 + }, + { + "epoch": 1.56, + "learning_rate": 2.4024749935880996e-05, + "loss": 0.5067, + "step": 40561 + }, + { + "epoch": 1.56, + "learning_rate": 2.4024108745832267e-05, + "loss": 0.2835, + "step": 40562 + }, + { + "epoch": 1.56, + "learning_rate": 2.4023467555783536e-05, + "loss": 0.4277, + "step": 40563 + }, + { + "epoch": 1.56, + "learning_rate": 2.4022826365734804e-05, + "loss": 0.4296, + "step": 40564 + }, + { + "epoch": 1.56, + "learning_rate": 2.4022185175686072e-05, + "loss": 0.3862, + "step": 40565 + }, + { + "epoch": 1.56, + "learning_rate": 2.4021543985637344e-05, + "loss": 0.3816, + "step": 40566 + }, + { + "epoch": 1.56, + "learning_rate": 2.4020902795588612e-05, + "loss": 0.3837, + "step": 40567 + }, + { + "epoch": 1.56, + "learning_rate": 2.4020261605539884e-05, + "loss": 0.3748, + "step": 40568 + }, + { + "epoch": 1.56, + "learning_rate": 2.4019620415491152e-05, + "loss": 0.376, + "step": 40569 + }, + { + "epoch": 1.56, + "learning_rate": 2.4018979225442424e-05, + "loss": 0.381, + "step": 40570 + }, + { + "epoch": 1.56, + "learning_rate": 2.4018338035393693e-05, + "loss": 0.3069, + "step": 40571 + }, + { + "epoch": 1.56, + "learning_rate": 2.401769684534496e-05, + "loss": 0.385, + "step": 40572 + }, + { + "epoch": 1.56, + "learning_rate": 2.401705565529623e-05, + "loss": 0.3253, + "step": 40573 + }, + { + "epoch": 1.56, + "learning_rate": 2.40164144652475e-05, + "loss": 0.2716, + "step": 40574 + }, + { + "epoch": 1.56, + "learning_rate": 2.401577327519877e-05, + "loss": 0.3267, + "step": 40575 + }, + { + "epoch": 1.56, + "learning_rate": 2.4015132085150038e-05, + "loss": 0.3942, + "step": 40576 + }, + { + "epoch": 1.56, + "learning_rate": 2.401449089510131e-05, + "loss": 0.3891, + "step": 40577 + }, + { + "epoch": 1.56, + "learning_rate": 2.401384970505258e-05, + "loss": 0.3297, + "step": 40578 + }, + { + "epoch": 1.56, + "learning_rate": 2.401320851500385e-05, + "loss": 0.5518, + "step": 40579 + }, + { + "epoch": 1.56, + "learning_rate": 2.4012567324955118e-05, + "loss": 0.3378, + "step": 40580 + }, + { + "epoch": 1.56, + "learning_rate": 2.4011926134906386e-05, + "loss": 0.3774, + "step": 40581 + }, + { + "epoch": 1.56, + "learning_rate": 2.4011284944857658e-05, + "loss": 0.3134, + "step": 40582 + }, + { + "epoch": 1.56, + "learning_rate": 2.4010643754808926e-05, + "loss": 0.4219, + "step": 40583 + }, + { + "epoch": 1.56, + "learning_rate": 2.4010002564760194e-05, + "loss": 0.3943, + "step": 40584 + }, + { + "epoch": 1.56, + "learning_rate": 2.4009361374711463e-05, + "loss": 0.3777, + "step": 40585 + }, + { + "epoch": 1.56, + "learning_rate": 2.4008720184662738e-05, + "loss": 0.2873, + "step": 40586 + }, + { + "epoch": 1.56, + "learning_rate": 2.4008078994614006e-05, + "loss": 0.2663, + "step": 40587 + }, + { + "epoch": 1.56, + "learning_rate": 2.4007437804565274e-05, + "loss": 0.4007, + "step": 40588 + }, + { + "epoch": 1.56, + "learning_rate": 2.4006796614516543e-05, + "loss": 0.3196, + "step": 40589 + }, + { + "epoch": 1.56, + "learning_rate": 2.4006155424467814e-05, + "loss": 0.375, + "step": 40590 + }, + { + "epoch": 1.56, + "learning_rate": 2.4005514234419083e-05, + "loss": 0.3415, + "step": 40591 + }, + { + "epoch": 1.56, + "learning_rate": 2.400487304437035e-05, + "loss": 0.3704, + "step": 40592 + }, + { + "epoch": 1.56, + "learning_rate": 2.400423185432162e-05, + "loss": 0.3177, + "step": 40593 + }, + { + "epoch": 1.56, + "learning_rate": 2.400359066427289e-05, + "loss": 0.3319, + "step": 40594 + }, + { + "epoch": 1.56, + "learning_rate": 2.4002949474224163e-05, + "loss": 0.3231, + "step": 40595 + }, + { + "epoch": 1.56, + "learning_rate": 2.400230828417543e-05, + "loss": 0.2619, + "step": 40596 + }, + { + "epoch": 1.56, + "learning_rate": 2.40016670941267e-05, + "loss": 0.3189, + "step": 40597 + }, + { + "epoch": 1.56, + "learning_rate": 2.400102590407797e-05, + "loss": 0.3504, + "step": 40598 + }, + { + "epoch": 1.56, + "learning_rate": 2.400038471402924e-05, + "loss": 0.4485, + "step": 40599 + }, + { + "epoch": 1.56, + "learning_rate": 2.3999743523980508e-05, + "loss": 0.4491, + "step": 40600 + }, + { + "epoch": 1.56, + "learning_rate": 2.3999102333931776e-05, + "loss": 0.5059, + "step": 40601 + }, + { + "epoch": 1.56, + "learning_rate": 2.3998461143883048e-05, + "loss": 0.3201, + "step": 40602 + }, + { + "epoch": 1.56, + "learning_rate": 2.399781995383432e-05, + "loss": 0.2811, + "step": 40603 + }, + { + "epoch": 1.56, + "learning_rate": 2.3997178763785588e-05, + "loss": 0.3021, + "step": 40604 + }, + { + "epoch": 1.56, + "learning_rate": 2.3996537573736856e-05, + "loss": 0.4366, + "step": 40605 + }, + { + "epoch": 1.56, + "learning_rate": 2.3995896383688128e-05, + "loss": 0.3159, + "step": 40606 + }, + { + "epoch": 1.56, + "learning_rate": 2.3995255193639396e-05, + "loss": 0.3448, + "step": 40607 + }, + { + "epoch": 1.56, + "learning_rate": 2.3994614003590665e-05, + "loss": 0.2981, + "step": 40608 + }, + { + "epoch": 1.56, + "learning_rate": 2.3993972813541933e-05, + "loss": 0.2511, + "step": 40609 + }, + { + "epoch": 1.56, + "learning_rate": 2.3993331623493205e-05, + "loss": 0.3664, + "step": 40610 + }, + { + "epoch": 1.56, + "learning_rate": 2.3992690433444473e-05, + "loss": 0.4724, + "step": 40611 + }, + { + "epoch": 1.56, + "learning_rate": 2.3992049243395745e-05, + "loss": 0.3318, + "step": 40612 + }, + { + "epoch": 1.56, + "learning_rate": 2.3991408053347013e-05, + "loss": 0.4746, + "step": 40613 + }, + { + "epoch": 1.56, + "learning_rate": 2.3990766863298285e-05, + "loss": 0.402, + "step": 40614 + }, + { + "epoch": 1.56, + "learning_rate": 2.3990125673249553e-05, + "loss": 0.3271, + "step": 40615 + }, + { + "epoch": 1.56, + "learning_rate": 2.398948448320082e-05, + "loss": 0.3592, + "step": 40616 + }, + { + "epoch": 1.56, + "learning_rate": 2.398884329315209e-05, + "loss": 0.245, + "step": 40617 + }, + { + "epoch": 1.56, + "learning_rate": 2.398820210310336e-05, + "loss": 0.4032, + "step": 40618 + }, + { + "epoch": 1.56, + "learning_rate": 2.398756091305463e-05, + "loss": 0.2974, + "step": 40619 + }, + { + "epoch": 1.56, + "learning_rate": 2.3986919723005898e-05, + "loss": 0.3975, + "step": 40620 + }, + { + "epoch": 1.56, + "learning_rate": 2.398627853295717e-05, + "loss": 0.2685, + "step": 40621 + }, + { + "epoch": 1.56, + "learning_rate": 2.398563734290844e-05, + "loss": 0.3744, + "step": 40622 + }, + { + "epoch": 1.56, + "learning_rate": 2.398499615285971e-05, + "loss": 0.2753, + "step": 40623 + }, + { + "epoch": 1.56, + "learning_rate": 2.398435496281098e-05, + "loss": 0.204, + "step": 40624 + }, + { + "epoch": 1.56, + "learning_rate": 2.3983713772762247e-05, + "loss": 0.449, + "step": 40625 + }, + { + "epoch": 1.56, + "learning_rate": 2.398307258271352e-05, + "loss": 0.2668, + "step": 40626 + }, + { + "epoch": 1.56, + "learning_rate": 2.3982431392664787e-05, + "loss": 0.2693, + "step": 40627 + }, + { + "epoch": 1.56, + "learning_rate": 2.3981790202616055e-05, + "loss": 0.2657, + "step": 40628 + }, + { + "epoch": 1.56, + "learning_rate": 2.3981149012567323e-05, + "loss": 0.4728, + "step": 40629 + }, + { + "epoch": 1.56, + "learning_rate": 2.39805078225186e-05, + "loss": 0.2164, + "step": 40630 + }, + { + "epoch": 1.56, + "learning_rate": 2.3979866632469867e-05, + "loss": 0.3439, + "step": 40631 + }, + { + "epoch": 1.56, + "learning_rate": 2.3979225442421135e-05, + "loss": 0.3035, + "step": 40632 + }, + { + "epoch": 1.56, + "learning_rate": 2.3978584252372403e-05, + "loss": 0.2008, + "step": 40633 + }, + { + "epoch": 1.56, + "learning_rate": 2.3977943062323675e-05, + "loss": 0.4702, + "step": 40634 + }, + { + "epoch": 1.56, + "learning_rate": 2.3977301872274943e-05, + "loss": 0.2819, + "step": 40635 + }, + { + "epoch": 1.56, + "learning_rate": 2.3976660682226212e-05, + "loss": 0.2287, + "step": 40636 + }, + { + "epoch": 1.56, + "learning_rate": 2.397601949217748e-05, + "loss": 0.4769, + "step": 40637 + }, + { + "epoch": 1.56, + "learning_rate": 2.3975378302128752e-05, + "loss": 0.3068, + "step": 40638 + }, + { + "epoch": 1.56, + "learning_rate": 2.3974737112080024e-05, + "loss": 0.401, + "step": 40639 + }, + { + "epoch": 1.56, + "learning_rate": 2.3974095922031292e-05, + "loss": 0.3906, + "step": 40640 + }, + { + "epoch": 1.56, + "learning_rate": 2.397345473198256e-05, + "loss": 0.3772, + "step": 40641 + }, + { + "epoch": 1.56, + "learning_rate": 2.3972813541933832e-05, + "loss": 0.3586, + "step": 40642 + }, + { + "epoch": 1.56, + "learning_rate": 2.39721723518851e-05, + "loss": 0.4034, + "step": 40643 + }, + { + "epoch": 1.56, + "learning_rate": 2.397153116183637e-05, + "loss": 0.2919, + "step": 40644 + }, + { + "epoch": 1.56, + "learning_rate": 2.3970889971787637e-05, + "loss": 0.2846, + "step": 40645 + }, + { + "epoch": 1.56, + "learning_rate": 2.397024878173891e-05, + "loss": 0.4987, + "step": 40646 + }, + { + "epoch": 1.56, + "learning_rate": 2.3969607591690177e-05, + "loss": 0.3416, + "step": 40647 + }, + { + "epoch": 1.56, + "learning_rate": 2.396896640164145e-05, + "loss": 0.3503, + "step": 40648 + }, + { + "epoch": 1.56, + "learning_rate": 2.3968325211592717e-05, + "loss": 0.425, + "step": 40649 + }, + { + "epoch": 1.56, + "learning_rate": 2.396768402154399e-05, + "loss": 0.3386, + "step": 40650 + }, + { + "epoch": 1.56, + "learning_rate": 2.3967042831495257e-05, + "loss": 0.3158, + "step": 40651 + }, + { + "epoch": 1.56, + "learning_rate": 2.3966401641446525e-05, + "loss": 0.2911, + "step": 40652 + }, + { + "epoch": 1.56, + "learning_rate": 2.3965760451397794e-05, + "loss": 0.3721, + "step": 40653 + }, + { + "epoch": 1.56, + "learning_rate": 2.3965119261349065e-05, + "loss": 0.256, + "step": 40654 + }, + { + "epoch": 1.56, + "learning_rate": 2.3964478071300334e-05, + "loss": 0.3704, + "step": 40655 + }, + { + "epoch": 1.56, + "learning_rate": 2.3963836881251602e-05, + "loss": 0.1921, + "step": 40656 + }, + { + "epoch": 1.56, + "learning_rate": 2.3963195691202874e-05, + "loss": 0.3789, + "step": 40657 + }, + { + "epoch": 1.56, + "learning_rate": 2.3962554501154146e-05, + "loss": 0.2404, + "step": 40658 + }, + { + "epoch": 1.56, + "learning_rate": 2.3961913311105414e-05, + "loss": 0.2108, + "step": 40659 + }, + { + "epoch": 1.56, + "learning_rate": 2.3961272121056682e-05, + "loss": 0.4222, + "step": 40660 + }, + { + "epoch": 1.56, + "learning_rate": 2.396063093100795e-05, + "loss": 0.4622, + "step": 40661 + }, + { + "epoch": 1.56, + "learning_rate": 2.3959989740959222e-05, + "loss": 0.4119, + "step": 40662 + }, + { + "epoch": 1.56, + "learning_rate": 2.395934855091049e-05, + "loss": 0.299, + "step": 40663 + }, + { + "epoch": 1.56, + "learning_rate": 2.395870736086176e-05, + "loss": 0.5466, + "step": 40664 + }, + { + "epoch": 1.56, + "learning_rate": 2.3958066170813027e-05, + "loss": 0.391, + "step": 40665 + }, + { + "epoch": 1.56, + "learning_rate": 2.3957424980764302e-05, + "loss": 0.4951, + "step": 40666 + }, + { + "epoch": 1.56, + "learning_rate": 2.395678379071557e-05, + "loss": 0.3117, + "step": 40667 + }, + { + "epoch": 1.56, + "learning_rate": 2.395614260066684e-05, + "loss": 0.5171, + "step": 40668 + }, + { + "epoch": 1.56, + "learning_rate": 2.3955501410618107e-05, + "loss": 0.3411, + "step": 40669 + }, + { + "epoch": 1.56, + "learning_rate": 2.395486022056938e-05, + "loss": 0.4437, + "step": 40670 + }, + { + "epoch": 1.56, + "learning_rate": 2.3954219030520647e-05, + "loss": 0.3775, + "step": 40671 + }, + { + "epoch": 1.56, + "learning_rate": 2.3953577840471916e-05, + "loss": 0.3311, + "step": 40672 + }, + { + "epoch": 1.56, + "learning_rate": 2.3952936650423184e-05, + "loss": 0.3945, + "step": 40673 + }, + { + "epoch": 1.56, + "learning_rate": 2.3952295460374456e-05, + "loss": 0.3562, + "step": 40674 + }, + { + "epoch": 1.56, + "learning_rate": 2.3951654270325727e-05, + "loss": 0.384, + "step": 40675 + }, + { + "epoch": 1.56, + "learning_rate": 2.3951013080276996e-05, + "loss": 0.3082, + "step": 40676 + }, + { + "epoch": 1.56, + "learning_rate": 2.3950371890228264e-05, + "loss": 0.2876, + "step": 40677 + }, + { + "epoch": 1.56, + "learning_rate": 2.3949730700179536e-05, + "loss": 0.2613, + "step": 40678 + }, + { + "epoch": 1.56, + "learning_rate": 2.3949089510130804e-05, + "loss": 0.4066, + "step": 40679 + }, + { + "epoch": 1.56, + "learning_rate": 2.3948448320082072e-05, + "loss": 0.3747, + "step": 40680 + }, + { + "epoch": 1.56, + "learning_rate": 2.394780713003334e-05, + "loss": 0.3932, + "step": 40681 + }, + { + "epoch": 1.56, + "learning_rate": 2.3947165939984613e-05, + "loss": 0.2949, + "step": 40682 + }, + { + "epoch": 1.56, + "learning_rate": 2.394652474993588e-05, + "loss": 0.4255, + "step": 40683 + }, + { + "epoch": 1.56, + "learning_rate": 2.3945883559887153e-05, + "loss": 0.3702, + "step": 40684 + }, + { + "epoch": 1.56, + "learning_rate": 2.394524236983842e-05, + "loss": 0.2721, + "step": 40685 + }, + { + "epoch": 1.56, + "learning_rate": 2.3944601179789693e-05, + "loss": 0.3201, + "step": 40686 + }, + { + "epoch": 1.56, + "learning_rate": 2.394395998974096e-05, + "loss": 0.4497, + "step": 40687 + }, + { + "epoch": 1.56, + "learning_rate": 2.394331879969223e-05, + "loss": 0.3697, + "step": 40688 + }, + { + "epoch": 1.56, + "learning_rate": 2.3942677609643498e-05, + "loss": 0.3805, + "step": 40689 + }, + { + "epoch": 1.56, + "learning_rate": 2.394203641959477e-05, + "loss": 0.3144, + "step": 40690 + }, + { + "epoch": 1.56, + "learning_rate": 2.3941395229546038e-05, + "loss": 0.3774, + "step": 40691 + }, + { + "epoch": 1.56, + "learning_rate": 2.3940754039497306e-05, + "loss": 0.3706, + "step": 40692 + }, + { + "epoch": 1.56, + "learning_rate": 2.3940112849448578e-05, + "loss": 0.3873, + "step": 40693 + }, + { + "epoch": 1.56, + "learning_rate": 2.393947165939985e-05, + "loss": 0.3202, + "step": 40694 + }, + { + "epoch": 1.56, + "learning_rate": 2.3938830469351118e-05, + "loss": 0.3183, + "step": 40695 + }, + { + "epoch": 1.56, + "learning_rate": 2.3938189279302386e-05, + "loss": 0.2584, + "step": 40696 + }, + { + "epoch": 1.56, + "learning_rate": 2.3937548089253654e-05, + "loss": 0.3549, + "step": 40697 + }, + { + "epoch": 1.56, + "learning_rate": 2.3936906899204926e-05, + "loss": 0.3051, + "step": 40698 + }, + { + "epoch": 1.56, + "learning_rate": 2.3936265709156194e-05, + "loss": 0.3715, + "step": 40699 + }, + { + "epoch": 1.56, + "learning_rate": 2.3935624519107463e-05, + "loss": 0.4644, + "step": 40700 + }, + { + "epoch": 1.56, + "learning_rate": 2.3934983329058734e-05, + "loss": 0.4612, + "step": 40701 + }, + { + "epoch": 1.56, + "learning_rate": 2.3934342139010006e-05, + "loss": 0.4902, + "step": 40702 + }, + { + "epoch": 1.56, + "learning_rate": 2.3933700948961275e-05, + "loss": 0.3613, + "step": 40703 + }, + { + "epoch": 1.56, + "learning_rate": 2.3933059758912543e-05, + "loss": 0.3401, + "step": 40704 + }, + { + "epoch": 1.56, + "learning_rate": 2.393241856886381e-05, + "loss": 0.4325, + "step": 40705 + }, + { + "epoch": 1.57, + "learning_rate": 2.3931777378815083e-05, + "loss": 0.2842, + "step": 40706 + }, + { + "epoch": 1.57, + "learning_rate": 2.393113618876635e-05, + "loss": 0.4937, + "step": 40707 + }, + { + "epoch": 1.57, + "learning_rate": 2.393049499871762e-05, + "loss": 0.2076, + "step": 40708 + }, + { + "epoch": 1.57, + "learning_rate": 2.3929853808668888e-05, + "loss": 0.3987, + "step": 40709 + }, + { + "epoch": 1.57, + "learning_rate": 2.3929212618620163e-05, + "loss": 0.3565, + "step": 40710 + }, + { + "epoch": 1.57, + "learning_rate": 2.392857142857143e-05, + "loss": 0.3377, + "step": 40711 + }, + { + "epoch": 1.57, + "learning_rate": 2.39279302385227e-05, + "loss": 0.2938, + "step": 40712 + }, + { + "epoch": 1.57, + "learning_rate": 2.3927289048473968e-05, + "loss": 0.3303, + "step": 40713 + }, + { + "epoch": 1.57, + "learning_rate": 2.392664785842524e-05, + "loss": 0.3217, + "step": 40714 + }, + { + "epoch": 1.57, + "learning_rate": 2.3926006668376508e-05, + "loss": 0.3097, + "step": 40715 + }, + { + "epoch": 1.57, + "learning_rate": 2.3925365478327776e-05, + "loss": 0.2882, + "step": 40716 + }, + { + "epoch": 1.57, + "learning_rate": 2.3924724288279045e-05, + "loss": 0.4026, + "step": 40717 + }, + { + "epoch": 1.57, + "learning_rate": 2.3924083098230316e-05, + "loss": 0.3769, + "step": 40718 + }, + { + "epoch": 1.57, + "learning_rate": 2.3923441908181588e-05, + "loss": 0.3894, + "step": 40719 + }, + { + "epoch": 1.57, + "learning_rate": 2.3922800718132856e-05, + "loss": 0.3383, + "step": 40720 + }, + { + "epoch": 1.57, + "learning_rate": 2.3922159528084125e-05, + "loss": 0.389, + "step": 40721 + }, + { + "epoch": 1.57, + "learning_rate": 2.3921518338035396e-05, + "loss": 0.3818, + "step": 40722 + }, + { + "epoch": 1.57, + "learning_rate": 2.3920877147986665e-05, + "loss": 0.4755, + "step": 40723 + }, + { + "epoch": 1.57, + "learning_rate": 2.3920235957937933e-05, + "loss": 0.4241, + "step": 40724 + }, + { + "epoch": 1.57, + "learning_rate": 2.39195947678892e-05, + "loss": 0.439, + "step": 40725 + }, + { + "epoch": 1.57, + "learning_rate": 2.3918953577840473e-05, + "loss": 0.3841, + "step": 40726 + }, + { + "epoch": 1.57, + "learning_rate": 2.391831238779174e-05, + "loss": 0.3679, + "step": 40727 + }, + { + "epoch": 1.57, + "learning_rate": 2.3917671197743013e-05, + "loss": 0.2559, + "step": 40728 + }, + { + "epoch": 1.57, + "learning_rate": 2.391703000769428e-05, + "loss": 0.3822, + "step": 40729 + }, + { + "epoch": 1.57, + "learning_rate": 2.3916388817645553e-05, + "loss": 0.539, + "step": 40730 + }, + { + "epoch": 1.57, + "learning_rate": 2.391574762759682e-05, + "loss": 0.308, + "step": 40731 + }, + { + "epoch": 1.57, + "learning_rate": 2.391510643754809e-05, + "loss": 0.2491, + "step": 40732 + }, + { + "epoch": 1.57, + "learning_rate": 2.3914465247499358e-05, + "loss": 0.3079, + "step": 40733 + }, + { + "epoch": 1.57, + "learning_rate": 2.391382405745063e-05, + "loss": 0.2423, + "step": 40734 + }, + { + "epoch": 1.57, + "learning_rate": 2.3913182867401898e-05, + "loss": 0.4007, + "step": 40735 + }, + { + "epoch": 1.57, + "learning_rate": 2.3912541677353167e-05, + "loss": 0.4002, + "step": 40736 + }, + { + "epoch": 1.57, + "learning_rate": 2.391190048730444e-05, + "loss": 0.3487, + "step": 40737 + }, + { + "epoch": 1.57, + "learning_rate": 2.391125929725571e-05, + "loss": 0.3285, + "step": 40738 + }, + { + "epoch": 1.57, + "learning_rate": 2.391061810720698e-05, + "loss": 0.3979, + "step": 40739 + }, + { + "epoch": 1.57, + "learning_rate": 2.3909976917158247e-05, + "loss": 0.4079, + "step": 40740 + }, + { + "epoch": 1.57, + "learning_rate": 2.3909335727109515e-05, + "loss": 0.284, + "step": 40741 + }, + { + "epoch": 1.57, + "learning_rate": 2.3908694537060787e-05, + "loss": 0.3666, + "step": 40742 + }, + { + "epoch": 1.57, + "learning_rate": 2.3908053347012055e-05, + "loss": 0.3486, + "step": 40743 + }, + { + "epoch": 1.57, + "learning_rate": 2.3907412156963323e-05, + "loss": 0.2983, + "step": 40744 + }, + { + "epoch": 1.57, + "learning_rate": 2.3906770966914592e-05, + "loss": 0.2742, + "step": 40745 + }, + { + "epoch": 1.57, + "learning_rate": 2.3906129776865867e-05, + "loss": 0.2389, + "step": 40746 + }, + { + "epoch": 1.57, + "learning_rate": 2.3905488586817135e-05, + "loss": 0.3159, + "step": 40747 + }, + { + "epoch": 1.57, + "learning_rate": 2.3904847396768404e-05, + "loss": 0.2677, + "step": 40748 + }, + { + "epoch": 1.57, + "learning_rate": 2.3904206206719672e-05, + "loss": 0.3861, + "step": 40749 + }, + { + "epoch": 1.57, + "learning_rate": 2.3903565016670944e-05, + "loss": 0.2968, + "step": 40750 + }, + { + "epoch": 1.57, + "learning_rate": 2.3902923826622212e-05, + "loss": 0.2915, + "step": 40751 + }, + { + "epoch": 1.57, + "learning_rate": 2.390228263657348e-05, + "loss": 0.304, + "step": 40752 + }, + { + "epoch": 1.57, + "learning_rate": 2.390164144652475e-05, + "loss": 0.3634, + "step": 40753 + }, + { + "epoch": 1.57, + "learning_rate": 2.390100025647602e-05, + "loss": 0.3207, + "step": 40754 + }, + { + "epoch": 1.57, + "learning_rate": 2.3900359066427292e-05, + "loss": 0.3573, + "step": 40755 + }, + { + "epoch": 1.57, + "learning_rate": 2.389971787637856e-05, + "loss": 0.3643, + "step": 40756 + }, + { + "epoch": 1.57, + "learning_rate": 2.389907668632983e-05, + "loss": 0.415, + "step": 40757 + }, + { + "epoch": 1.57, + "learning_rate": 2.38984354962811e-05, + "loss": 0.292, + "step": 40758 + }, + { + "epoch": 1.57, + "learning_rate": 2.389779430623237e-05, + "loss": 0.2887, + "step": 40759 + }, + { + "epoch": 1.57, + "learning_rate": 2.3897153116183637e-05, + "loss": 0.3849, + "step": 40760 + }, + { + "epoch": 1.57, + "learning_rate": 2.3896511926134905e-05, + "loss": 0.3303, + "step": 40761 + }, + { + "epoch": 1.57, + "learning_rate": 2.3895870736086177e-05, + "loss": 0.4162, + "step": 40762 + }, + { + "epoch": 1.57, + "learning_rate": 2.3895229546037445e-05, + "loss": 0.3693, + "step": 40763 + }, + { + "epoch": 1.57, + "learning_rate": 2.3894588355988717e-05, + "loss": 0.3908, + "step": 40764 + }, + { + "epoch": 1.57, + "learning_rate": 2.3893947165939985e-05, + "loss": 0.3369, + "step": 40765 + }, + { + "epoch": 1.57, + "learning_rate": 2.3893305975891257e-05, + "loss": 0.2822, + "step": 40766 + }, + { + "epoch": 1.57, + "learning_rate": 2.3892664785842525e-05, + "loss": 0.2356, + "step": 40767 + }, + { + "epoch": 1.57, + "learning_rate": 2.3892023595793794e-05, + "loss": 0.4058, + "step": 40768 + }, + { + "epoch": 1.57, + "learning_rate": 2.3891382405745062e-05, + "loss": 0.2628, + "step": 40769 + }, + { + "epoch": 1.57, + "learning_rate": 2.3890741215696334e-05, + "loss": 0.3519, + "step": 40770 + }, + { + "epoch": 1.57, + "learning_rate": 2.3890100025647602e-05, + "loss": 0.444, + "step": 40771 + }, + { + "epoch": 1.57, + "learning_rate": 2.388945883559887e-05, + "loss": 0.4288, + "step": 40772 + }, + { + "epoch": 1.57, + "learning_rate": 2.3888817645550142e-05, + "loss": 0.3708, + "step": 40773 + }, + { + "epoch": 1.57, + "learning_rate": 2.3888176455501414e-05, + "loss": 0.2781, + "step": 40774 + }, + { + "epoch": 1.57, + "learning_rate": 2.3887535265452682e-05, + "loss": 0.4459, + "step": 40775 + }, + { + "epoch": 1.57, + "learning_rate": 2.388689407540395e-05, + "loss": 0.2358, + "step": 40776 + }, + { + "epoch": 1.57, + "learning_rate": 2.388625288535522e-05, + "loss": 0.3752, + "step": 40777 + }, + { + "epoch": 1.57, + "learning_rate": 2.388561169530649e-05, + "loss": 0.2935, + "step": 40778 + }, + { + "epoch": 1.57, + "learning_rate": 2.388497050525776e-05, + "loss": 0.4033, + "step": 40779 + }, + { + "epoch": 1.57, + "learning_rate": 2.3884329315209027e-05, + "loss": 0.3297, + "step": 40780 + }, + { + "epoch": 1.57, + "learning_rate": 2.3883688125160296e-05, + "loss": 0.2825, + "step": 40781 + }, + { + "epoch": 1.57, + "learning_rate": 2.388304693511157e-05, + "loss": 0.2707, + "step": 40782 + }, + { + "epoch": 1.57, + "learning_rate": 2.388240574506284e-05, + "loss": 0.4833, + "step": 40783 + }, + { + "epoch": 1.57, + "learning_rate": 2.3881764555014107e-05, + "loss": 0.2498, + "step": 40784 + }, + { + "epoch": 1.57, + "learning_rate": 2.3881123364965376e-05, + "loss": 0.3724, + "step": 40785 + }, + { + "epoch": 1.57, + "learning_rate": 2.3880482174916647e-05, + "loss": 0.4169, + "step": 40786 + }, + { + "epoch": 1.57, + "learning_rate": 2.3879840984867916e-05, + "loss": 0.3592, + "step": 40787 + }, + { + "epoch": 1.57, + "learning_rate": 2.3879199794819184e-05, + "loss": 0.2809, + "step": 40788 + }, + { + "epoch": 1.57, + "learning_rate": 2.3878558604770452e-05, + "loss": 0.4617, + "step": 40789 + }, + { + "epoch": 1.57, + "learning_rate": 2.3877917414721724e-05, + "loss": 0.2859, + "step": 40790 + }, + { + "epoch": 1.57, + "learning_rate": 2.3877276224672996e-05, + "loss": 0.2841, + "step": 40791 + }, + { + "epoch": 1.57, + "learning_rate": 2.3876635034624264e-05, + "loss": 0.3227, + "step": 40792 + }, + { + "epoch": 1.57, + "learning_rate": 2.3875993844575533e-05, + "loss": 0.4262, + "step": 40793 + }, + { + "epoch": 1.57, + "learning_rate": 2.3875352654526804e-05, + "loss": 0.3316, + "step": 40794 + }, + { + "epoch": 1.57, + "learning_rate": 2.3874711464478073e-05, + "loss": 0.4648, + "step": 40795 + }, + { + "epoch": 1.57, + "learning_rate": 2.387407027442934e-05, + "loss": 0.361, + "step": 40796 + }, + { + "epoch": 1.57, + "learning_rate": 2.387342908438061e-05, + "loss": 0.3073, + "step": 40797 + }, + { + "epoch": 1.57, + "learning_rate": 2.387278789433188e-05, + "loss": 0.3404, + "step": 40798 + }, + { + "epoch": 1.57, + "learning_rate": 2.3872146704283153e-05, + "loss": 0.391, + "step": 40799 + }, + { + "epoch": 1.57, + "learning_rate": 2.387150551423442e-05, + "loss": 0.4807, + "step": 40800 + }, + { + "epoch": 1.57, + "learning_rate": 2.387086432418569e-05, + "loss": 0.3747, + "step": 40801 + }, + { + "epoch": 1.57, + "learning_rate": 2.387022313413696e-05, + "loss": 0.3602, + "step": 40802 + }, + { + "epoch": 1.57, + "learning_rate": 2.386958194408823e-05, + "loss": 0.4061, + "step": 40803 + }, + { + "epoch": 1.57, + "learning_rate": 2.3868940754039498e-05, + "loss": 0.4383, + "step": 40804 + }, + { + "epoch": 1.57, + "learning_rate": 2.3868299563990766e-05, + "loss": 0.3907, + "step": 40805 + }, + { + "epoch": 1.57, + "learning_rate": 2.3867658373942038e-05, + "loss": 0.3516, + "step": 40806 + }, + { + "epoch": 1.57, + "learning_rate": 2.3867017183893306e-05, + "loss": 0.2427, + "step": 40807 + }, + { + "epoch": 1.57, + "learning_rate": 2.3866375993844578e-05, + "loss": 0.2982, + "step": 40808 + }, + { + "epoch": 1.57, + "learning_rate": 2.3865734803795846e-05, + "loss": 0.3699, + "step": 40809 + }, + { + "epoch": 1.57, + "learning_rate": 2.3865093613747118e-05, + "loss": 0.3614, + "step": 40810 + }, + { + "epoch": 1.57, + "learning_rate": 2.3864452423698386e-05, + "loss": 0.4036, + "step": 40811 + }, + { + "epoch": 1.57, + "learning_rate": 2.3863811233649654e-05, + "loss": 0.3384, + "step": 40812 + }, + { + "epoch": 1.57, + "learning_rate": 2.3863170043600923e-05, + "loss": 0.3692, + "step": 40813 + }, + { + "epoch": 1.57, + "learning_rate": 2.3862528853552195e-05, + "loss": 0.1996, + "step": 40814 + }, + { + "epoch": 1.57, + "learning_rate": 2.3861887663503463e-05, + "loss": 0.3622, + "step": 40815 + }, + { + "epoch": 1.57, + "learning_rate": 2.386124647345473e-05, + "loss": 0.414, + "step": 40816 + }, + { + "epoch": 1.57, + "learning_rate": 2.3860605283406003e-05, + "loss": 0.2981, + "step": 40817 + }, + { + "epoch": 1.57, + "learning_rate": 2.3859964093357275e-05, + "loss": 0.3959, + "step": 40818 + }, + { + "epoch": 1.57, + "learning_rate": 2.3859322903308543e-05, + "loss": 0.5735, + "step": 40819 + }, + { + "epoch": 1.57, + "learning_rate": 2.385868171325981e-05, + "loss": 0.4297, + "step": 40820 + }, + { + "epoch": 1.57, + "learning_rate": 2.385804052321108e-05, + "loss": 0.3106, + "step": 40821 + }, + { + "epoch": 1.57, + "learning_rate": 2.385739933316235e-05, + "loss": 0.2233, + "step": 40822 + }, + { + "epoch": 1.57, + "learning_rate": 2.385675814311362e-05, + "loss": 0.2127, + "step": 40823 + }, + { + "epoch": 1.57, + "learning_rate": 2.3856116953064888e-05, + "loss": 0.2701, + "step": 40824 + }, + { + "epoch": 1.57, + "learning_rate": 2.3855475763016156e-05, + "loss": 0.2716, + "step": 40825 + }, + { + "epoch": 1.57, + "learning_rate": 2.385483457296743e-05, + "loss": 0.2762, + "step": 40826 + }, + { + "epoch": 1.57, + "learning_rate": 2.38541933829187e-05, + "loss": 0.3738, + "step": 40827 + }, + { + "epoch": 1.57, + "learning_rate": 2.3853552192869968e-05, + "loss": 0.2354, + "step": 40828 + }, + { + "epoch": 1.57, + "learning_rate": 2.3852911002821236e-05, + "loss": 0.4005, + "step": 40829 + }, + { + "epoch": 1.57, + "learning_rate": 2.3852269812772508e-05, + "loss": 0.4023, + "step": 40830 + }, + { + "epoch": 1.57, + "learning_rate": 2.3851628622723776e-05, + "loss": 0.4267, + "step": 40831 + }, + { + "epoch": 1.57, + "learning_rate": 2.3850987432675045e-05, + "loss": 0.3347, + "step": 40832 + }, + { + "epoch": 1.57, + "learning_rate": 2.3850346242626313e-05, + "loss": 0.4221, + "step": 40833 + }, + { + "epoch": 1.57, + "learning_rate": 2.3849705052577585e-05, + "loss": 0.3534, + "step": 40834 + }, + { + "epoch": 1.57, + "learning_rate": 2.3849063862528857e-05, + "loss": 0.3232, + "step": 40835 + }, + { + "epoch": 1.57, + "learning_rate": 2.3848422672480125e-05, + "loss": 0.3425, + "step": 40836 + }, + { + "epoch": 1.57, + "learning_rate": 2.3847781482431393e-05, + "loss": 0.3244, + "step": 40837 + }, + { + "epoch": 1.57, + "learning_rate": 2.3847140292382665e-05, + "loss": 0.1696, + "step": 40838 + }, + { + "epoch": 1.57, + "learning_rate": 2.3846499102333933e-05, + "loss": 0.4901, + "step": 40839 + }, + { + "epoch": 1.57, + "learning_rate": 2.38458579122852e-05, + "loss": 0.3086, + "step": 40840 + }, + { + "epoch": 1.57, + "learning_rate": 2.384521672223647e-05, + "loss": 0.5102, + "step": 40841 + }, + { + "epoch": 1.57, + "learning_rate": 2.384457553218774e-05, + "loss": 0.3881, + "step": 40842 + }, + { + "epoch": 1.57, + "learning_rate": 2.384393434213901e-05, + "loss": 0.3399, + "step": 40843 + }, + { + "epoch": 1.57, + "learning_rate": 2.384329315209028e-05, + "loss": 0.4375, + "step": 40844 + }, + { + "epoch": 1.57, + "learning_rate": 2.384265196204155e-05, + "loss": 0.4728, + "step": 40845 + }, + { + "epoch": 1.57, + "learning_rate": 2.384201077199282e-05, + "loss": 0.4143, + "step": 40846 + }, + { + "epoch": 1.57, + "learning_rate": 2.384136958194409e-05, + "loss": 0.2843, + "step": 40847 + }, + { + "epoch": 1.57, + "learning_rate": 2.384072839189536e-05, + "loss": 0.4574, + "step": 40848 + }, + { + "epoch": 1.57, + "learning_rate": 2.3840087201846627e-05, + "loss": 0.3747, + "step": 40849 + }, + { + "epoch": 1.57, + "learning_rate": 2.38394460117979e-05, + "loss": 0.402, + "step": 40850 + }, + { + "epoch": 1.57, + "learning_rate": 2.3838804821749167e-05, + "loss": 0.32, + "step": 40851 + }, + { + "epoch": 1.57, + "learning_rate": 2.3838163631700435e-05, + "loss": 0.4768, + "step": 40852 + }, + { + "epoch": 1.57, + "learning_rate": 2.3837522441651707e-05, + "loss": 0.3334, + "step": 40853 + }, + { + "epoch": 1.57, + "learning_rate": 2.383688125160298e-05, + "loss": 0.3627, + "step": 40854 + }, + { + "epoch": 1.57, + "learning_rate": 2.3836240061554247e-05, + "loss": 0.2524, + "step": 40855 + }, + { + "epoch": 1.57, + "learning_rate": 2.3835598871505515e-05, + "loss": 0.3775, + "step": 40856 + }, + { + "epoch": 1.57, + "learning_rate": 2.3834957681456783e-05, + "loss": 0.3759, + "step": 40857 + }, + { + "epoch": 1.57, + "learning_rate": 2.3834316491408055e-05, + "loss": 0.411, + "step": 40858 + }, + { + "epoch": 1.57, + "learning_rate": 2.3833675301359324e-05, + "loss": 0.3983, + "step": 40859 + }, + { + "epoch": 1.57, + "learning_rate": 2.3833034111310592e-05, + "loss": 0.5504, + "step": 40860 + }, + { + "epoch": 1.57, + "learning_rate": 2.383239292126186e-05, + "loss": 0.3253, + "step": 40861 + }, + { + "epoch": 1.57, + "learning_rate": 2.3831751731213135e-05, + "loss": 0.3703, + "step": 40862 + }, + { + "epoch": 1.57, + "learning_rate": 2.3831110541164404e-05, + "loss": 0.5079, + "step": 40863 + }, + { + "epoch": 1.57, + "learning_rate": 2.3830469351115672e-05, + "loss": 0.4345, + "step": 40864 + }, + { + "epoch": 1.57, + "learning_rate": 2.382982816106694e-05, + "loss": 0.3556, + "step": 40865 + }, + { + "epoch": 1.57, + "learning_rate": 2.3829186971018212e-05, + "loss": 0.2987, + "step": 40866 + }, + { + "epoch": 1.57, + "learning_rate": 2.382854578096948e-05, + "loss": 0.3911, + "step": 40867 + }, + { + "epoch": 1.57, + "learning_rate": 2.382790459092075e-05, + "loss": 0.4448, + "step": 40868 + }, + { + "epoch": 1.57, + "learning_rate": 2.3827263400872017e-05, + "loss": 0.4053, + "step": 40869 + }, + { + "epoch": 1.57, + "learning_rate": 2.382662221082329e-05, + "loss": 0.418, + "step": 40870 + }, + { + "epoch": 1.57, + "learning_rate": 2.382598102077456e-05, + "loss": 0.403, + "step": 40871 + }, + { + "epoch": 1.57, + "learning_rate": 2.382533983072583e-05, + "loss": 0.2667, + "step": 40872 + }, + { + "epoch": 1.57, + "learning_rate": 2.3824698640677097e-05, + "loss": 0.4875, + "step": 40873 + }, + { + "epoch": 1.57, + "learning_rate": 2.382405745062837e-05, + "loss": 0.2417, + "step": 40874 + }, + { + "epoch": 1.57, + "learning_rate": 2.3823416260579637e-05, + "loss": 0.3764, + "step": 40875 + }, + { + "epoch": 1.57, + "learning_rate": 2.3822775070530905e-05, + "loss": 0.3438, + "step": 40876 + }, + { + "epoch": 1.57, + "learning_rate": 2.3822133880482174e-05, + "loss": 0.406, + "step": 40877 + }, + { + "epoch": 1.57, + "learning_rate": 2.3821492690433445e-05, + "loss": 0.3991, + "step": 40878 + }, + { + "epoch": 1.57, + "learning_rate": 2.3820851500384714e-05, + "loss": 0.3029, + "step": 40879 + }, + { + "epoch": 1.57, + "learning_rate": 2.3820210310335986e-05, + "loss": 0.4339, + "step": 40880 + }, + { + "epoch": 1.57, + "learning_rate": 2.3819569120287254e-05, + "loss": 0.4551, + "step": 40881 + }, + { + "epoch": 1.57, + "learning_rate": 2.3818927930238526e-05, + "loss": 0.3054, + "step": 40882 + }, + { + "epoch": 1.57, + "learning_rate": 2.3818286740189794e-05, + "loss": 0.3052, + "step": 40883 + }, + { + "epoch": 1.57, + "learning_rate": 2.3817645550141062e-05, + "loss": 0.3491, + "step": 40884 + }, + { + "epoch": 1.57, + "learning_rate": 2.381700436009233e-05, + "loss": 0.3243, + "step": 40885 + }, + { + "epoch": 1.57, + "learning_rate": 2.3816363170043602e-05, + "loss": 0.3913, + "step": 40886 + }, + { + "epoch": 1.57, + "learning_rate": 2.381572197999487e-05, + "loss": 0.2844, + "step": 40887 + }, + { + "epoch": 1.57, + "learning_rate": 2.3815080789946142e-05, + "loss": 0.3049, + "step": 40888 + }, + { + "epoch": 1.57, + "learning_rate": 2.381443959989741e-05, + "loss": 0.3797, + "step": 40889 + }, + { + "epoch": 1.57, + "learning_rate": 2.3813798409848682e-05, + "loss": 0.4065, + "step": 40890 + }, + { + "epoch": 1.57, + "learning_rate": 2.381315721979995e-05, + "loss": 0.4367, + "step": 40891 + }, + { + "epoch": 1.57, + "learning_rate": 2.381251602975122e-05, + "loss": 0.4085, + "step": 40892 + }, + { + "epoch": 1.57, + "learning_rate": 2.3811874839702487e-05, + "loss": 0.4253, + "step": 40893 + }, + { + "epoch": 1.57, + "learning_rate": 2.381123364965376e-05, + "loss": 0.3863, + "step": 40894 + }, + { + "epoch": 1.57, + "learning_rate": 2.3810592459605027e-05, + "loss": 0.3542, + "step": 40895 + }, + { + "epoch": 1.57, + "learning_rate": 2.3809951269556296e-05, + "loss": 0.3004, + "step": 40896 + }, + { + "epoch": 1.57, + "learning_rate": 2.3809310079507567e-05, + "loss": 0.5023, + "step": 40897 + }, + { + "epoch": 1.57, + "learning_rate": 2.380866888945884e-05, + "loss": 0.2611, + "step": 40898 + }, + { + "epoch": 1.57, + "learning_rate": 2.3808027699410107e-05, + "loss": 0.4038, + "step": 40899 + }, + { + "epoch": 1.57, + "learning_rate": 2.3807386509361376e-05, + "loss": 0.4054, + "step": 40900 + }, + { + "epoch": 1.57, + "learning_rate": 2.3806745319312644e-05, + "loss": 0.2971, + "step": 40901 + }, + { + "epoch": 1.57, + "learning_rate": 2.3806104129263916e-05, + "loss": 0.2949, + "step": 40902 + }, + { + "epoch": 1.57, + "learning_rate": 2.3805462939215184e-05, + "loss": 0.3923, + "step": 40903 + }, + { + "epoch": 1.57, + "learning_rate": 2.3804821749166452e-05, + "loss": 0.3176, + "step": 40904 + }, + { + "epoch": 1.57, + "learning_rate": 2.380418055911772e-05, + "loss": 0.2608, + "step": 40905 + }, + { + "epoch": 1.57, + "learning_rate": 2.3803539369068996e-05, + "loss": 0.3837, + "step": 40906 + }, + { + "epoch": 1.57, + "learning_rate": 2.3802898179020264e-05, + "loss": 0.5378, + "step": 40907 + }, + { + "epoch": 1.57, + "learning_rate": 2.3802256988971533e-05, + "loss": 0.3714, + "step": 40908 + }, + { + "epoch": 1.57, + "learning_rate": 2.38016157989228e-05, + "loss": 0.3387, + "step": 40909 + }, + { + "epoch": 1.57, + "learning_rate": 2.3800974608874073e-05, + "loss": 0.2806, + "step": 40910 + }, + { + "epoch": 1.57, + "learning_rate": 2.380033341882534e-05, + "loss": 0.4953, + "step": 40911 + }, + { + "epoch": 1.57, + "learning_rate": 2.379969222877661e-05, + "loss": 0.3385, + "step": 40912 + }, + { + "epoch": 1.57, + "learning_rate": 2.3799051038727878e-05, + "loss": 0.3598, + "step": 40913 + }, + { + "epoch": 1.57, + "learning_rate": 2.379840984867915e-05, + "loss": 0.4491, + "step": 40914 + }, + { + "epoch": 1.57, + "learning_rate": 2.379776865863042e-05, + "loss": 0.2615, + "step": 40915 + }, + { + "epoch": 1.57, + "learning_rate": 2.379712746858169e-05, + "loss": 0.343, + "step": 40916 + }, + { + "epoch": 1.57, + "learning_rate": 2.3796486278532958e-05, + "loss": 0.3241, + "step": 40917 + }, + { + "epoch": 1.57, + "learning_rate": 2.379584508848423e-05, + "loss": 0.3056, + "step": 40918 + }, + { + "epoch": 1.57, + "learning_rate": 2.3795203898435498e-05, + "loss": 0.3599, + "step": 40919 + }, + { + "epoch": 1.57, + "learning_rate": 2.3794562708386766e-05, + "loss": 0.4081, + "step": 40920 + }, + { + "epoch": 1.57, + "learning_rate": 2.3793921518338034e-05, + "loss": 0.442, + "step": 40921 + }, + { + "epoch": 1.57, + "learning_rate": 2.3793280328289306e-05, + "loss": 0.3285, + "step": 40922 + }, + { + "epoch": 1.57, + "learning_rate": 2.3792639138240574e-05, + "loss": 0.2788, + "step": 40923 + }, + { + "epoch": 1.57, + "learning_rate": 2.3791997948191846e-05, + "loss": 0.3125, + "step": 40924 + }, + { + "epoch": 1.57, + "learning_rate": 2.3791356758143114e-05, + "loss": 0.2923, + "step": 40925 + }, + { + "epoch": 1.57, + "learning_rate": 2.3790715568094386e-05, + "loss": 0.3711, + "step": 40926 + }, + { + "epoch": 1.57, + "learning_rate": 2.3790074378045655e-05, + "loss": 0.3438, + "step": 40927 + }, + { + "epoch": 1.57, + "learning_rate": 2.3789433187996923e-05, + "loss": 0.3748, + "step": 40928 + }, + { + "epoch": 1.57, + "learning_rate": 2.378879199794819e-05, + "loss": 0.3388, + "step": 40929 + }, + { + "epoch": 1.57, + "learning_rate": 2.3788150807899463e-05, + "loss": 0.2457, + "step": 40930 + }, + { + "epoch": 1.57, + "learning_rate": 2.378750961785073e-05, + "loss": 0.3844, + "step": 40931 + }, + { + "epoch": 1.57, + "learning_rate": 2.3786868427802e-05, + "loss": 0.2509, + "step": 40932 + }, + { + "epoch": 1.57, + "learning_rate": 2.378622723775327e-05, + "loss": 0.3722, + "step": 40933 + }, + { + "epoch": 1.57, + "learning_rate": 2.3785586047704543e-05, + "loss": 0.1515, + "step": 40934 + }, + { + "epoch": 1.57, + "learning_rate": 2.378494485765581e-05, + "loss": 0.216, + "step": 40935 + }, + { + "epoch": 1.57, + "learning_rate": 2.378430366760708e-05, + "loss": 0.5556, + "step": 40936 + }, + { + "epoch": 1.57, + "learning_rate": 2.3783662477558348e-05, + "loss": 0.4269, + "step": 40937 + }, + { + "epoch": 1.57, + "learning_rate": 2.378302128750962e-05, + "loss": 0.3516, + "step": 40938 + }, + { + "epoch": 1.57, + "learning_rate": 2.3782380097460888e-05, + "loss": 0.25, + "step": 40939 + }, + { + "epoch": 1.57, + "learning_rate": 2.3781738907412156e-05, + "loss": 0.3751, + "step": 40940 + }, + { + "epoch": 1.57, + "learning_rate": 2.3781097717363428e-05, + "loss": 0.292, + "step": 40941 + }, + { + "epoch": 1.57, + "learning_rate": 2.37804565273147e-05, + "loss": 0.4995, + "step": 40942 + }, + { + "epoch": 1.57, + "learning_rate": 2.3779815337265968e-05, + "loss": 0.3691, + "step": 40943 + }, + { + "epoch": 1.57, + "learning_rate": 2.3779174147217236e-05, + "loss": 0.304, + "step": 40944 + }, + { + "epoch": 1.57, + "learning_rate": 2.3778532957168505e-05, + "loss": 0.3061, + "step": 40945 + }, + { + "epoch": 1.57, + "learning_rate": 2.3777891767119777e-05, + "loss": 0.4874, + "step": 40946 + }, + { + "epoch": 1.57, + "learning_rate": 2.3777250577071045e-05, + "loss": 0.2902, + "step": 40947 + }, + { + "epoch": 1.57, + "learning_rate": 2.3776609387022313e-05, + "loss": 0.3391, + "step": 40948 + }, + { + "epoch": 1.57, + "learning_rate": 2.377596819697358e-05, + "loss": 0.231, + "step": 40949 + }, + { + "epoch": 1.57, + "learning_rate": 2.3775327006924853e-05, + "loss": 0.3318, + "step": 40950 + }, + { + "epoch": 1.57, + "learning_rate": 2.3774685816876125e-05, + "loss": 0.3774, + "step": 40951 + }, + { + "epoch": 1.57, + "learning_rate": 2.3774044626827393e-05, + "loss": 0.2409, + "step": 40952 + }, + { + "epoch": 1.57, + "learning_rate": 2.377340343677866e-05, + "loss": 0.3495, + "step": 40953 + }, + { + "epoch": 1.57, + "learning_rate": 2.3772762246729933e-05, + "loss": 0.2803, + "step": 40954 + }, + { + "epoch": 1.57, + "learning_rate": 2.37721210566812e-05, + "loss": 0.4212, + "step": 40955 + }, + { + "epoch": 1.57, + "learning_rate": 2.377147986663247e-05, + "loss": 0.3818, + "step": 40956 + }, + { + "epoch": 1.57, + "learning_rate": 2.3770838676583738e-05, + "loss": 0.502, + "step": 40957 + }, + { + "epoch": 1.57, + "learning_rate": 2.377019748653501e-05, + "loss": 0.435, + "step": 40958 + }, + { + "epoch": 1.57, + "learning_rate": 2.376955629648628e-05, + "loss": 0.3309, + "step": 40959 + }, + { + "epoch": 1.57, + "learning_rate": 2.376891510643755e-05, + "loss": 0.3249, + "step": 40960 + }, + { + "epoch": 1.57, + "learning_rate": 2.376827391638882e-05, + "loss": 0.3681, + "step": 40961 + }, + { + "epoch": 1.57, + "learning_rate": 2.376763272634009e-05, + "loss": 0.3775, + "step": 40962 + }, + { + "epoch": 1.57, + "learning_rate": 2.376699153629136e-05, + "loss": 0.3697, + "step": 40963 + }, + { + "epoch": 1.57, + "learning_rate": 2.3766350346242627e-05, + "loss": 0.4055, + "step": 40964 + }, + { + "epoch": 1.57, + "learning_rate": 2.3765709156193895e-05, + "loss": 0.3873, + "step": 40965 + }, + { + "epoch": 1.58, + "learning_rate": 2.3765067966145167e-05, + "loss": 0.2739, + "step": 40966 + }, + { + "epoch": 1.58, + "learning_rate": 2.3764426776096435e-05, + "loss": 0.4253, + "step": 40967 + }, + { + "epoch": 1.58, + "learning_rate": 2.3763785586047703e-05, + "loss": 0.3483, + "step": 40968 + }, + { + "epoch": 1.58, + "learning_rate": 2.3763144395998975e-05, + "loss": 0.3589, + "step": 40969 + }, + { + "epoch": 1.58, + "learning_rate": 2.3762503205950247e-05, + "loss": 0.3615, + "step": 40970 + }, + { + "epoch": 1.58, + "learning_rate": 2.3761862015901515e-05, + "loss": 0.3324, + "step": 40971 + }, + { + "epoch": 1.58, + "learning_rate": 2.3761220825852784e-05, + "loss": 0.4249, + "step": 40972 + }, + { + "epoch": 1.58, + "learning_rate": 2.3760579635804052e-05, + "loss": 0.3018, + "step": 40973 + }, + { + "epoch": 1.58, + "learning_rate": 2.3759938445755324e-05, + "loss": 0.3181, + "step": 40974 + }, + { + "epoch": 1.58, + "learning_rate": 2.3759297255706592e-05, + "loss": 0.2901, + "step": 40975 + }, + { + "epoch": 1.58, + "learning_rate": 2.375865606565786e-05, + "loss": 0.263, + "step": 40976 + }, + { + "epoch": 1.58, + "learning_rate": 2.3758014875609132e-05, + "loss": 0.3477, + "step": 40977 + }, + { + "epoch": 1.58, + "learning_rate": 2.3757373685560404e-05, + "loss": 0.3152, + "step": 40978 + }, + { + "epoch": 1.58, + "learning_rate": 2.3756732495511672e-05, + "loss": 0.3989, + "step": 40979 + }, + { + "epoch": 1.58, + "learning_rate": 2.375609130546294e-05, + "loss": 0.2442, + "step": 40980 + }, + { + "epoch": 1.58, + "learning_rate": 2.375545011541421e-05, + "loss": 0.3763, + "step": 40981 + }, + { + "epoch": 1.58, + "learning_rate": 2.375480892536548e-05, + "loss": 0.3146, + "step": 40982 + }, + { + "epoch": 1.58, + "learning_rate": 2.375416773531675e-05, + "loss": 0.3062, + "step": 40983 + }, + { + "epoch": 1.58, + "learning_rate": 2.3753526545268017e-05, + "loss": 0.304, + "step": 40984 + }, + { + "epoch": 1.58, + "learning_rate": 2.3752885355219285e-05, + "loss": 0.395, + "step": 40985 + }, + { + "epoch": 1.58, + "learning_rate": 2.375224416517056e-05, + "loss": 0.4744, + "step": 40986 + }, + { + "epoch": 1.58, + "learning_rate": 2.375160297512183e-05, + "loss": 0.2833, + "step": 40987 + }, + { + "epoch": 1.58, + "learning_rate": 2.3750961785073097e-05, + "loss": 0.3999, + "step": 40988 + }, + { + "epoch": 1.58, + "learning_rate": 2.3750320595024365e-05, + "loss": 0.3894, + "step": 40989 + }, + { + "epoch": 1.58, + "learning_rate": 2.3749679404975637e-05, + "loss": 0.4551, + "step": 40990 + }, + { + "epoch": 1.58, + "learning_rate": 2.3749038214926905e-05, + "loss": 0.2777, + "step": 40991 + }, + { + "epoch": 1.58, + "learning_rate": 2.3748397024878174e-05, + "loss": 0.3258, + "step": 40992 + }, + { + "epoch": 1.58, + "learning_rate": 2.3747755834829442e-05, + "loss": 0.3367, + "step": 40993 + }, + { + "epoch": 1.58, + "learning_rate": 2.3747114644780714e-05, + "loss": 0.4826, + "step": 40994 + }, + { + "epoch": 1.58, + "learning_rate": 2.3746473454731986e-05, + "loss": 0.3432, + "step": 40995 + }, + { + "epoch": 1.58, + "learning_rate": 2.3745832264683254e-05, + "loss": 0.3316, + "step": 40996 + }, + { + "epoch": 1.58, + "learning_rate": 2.3745191074634522e-05, + "loss": 0.415, + "step": 40997 + }, + { + "epoch": 1.58, + "learning_rate": 2.3744549884585794e-05, + "loss": 0.3749, + "step": 40998 + }, + { + "epoch": 1.58, + "learning_rate": 2.3743908694537062e-05, + "loss": 0.2999, + "step": 40999 + }, + { + "epoch": 1.58, + "learning_rate": 2.374326750448833e-05, + "loss": 0.35, + "step": 41000 + }, + { + "epoch": 1.58, + "learning_rate": 2.37426263144396e-05, + "loss": 0.3579, + "step": 41001 + }, + { + "epoch": 1.58, + "learning_rate": 2.374198512439087e-05, + "loss": 0.3019, + "step": 41002 + }, + { + "epoch": 1.58, + "learning_rate": 2.374134393434214e-05, + "loss": 0.3174, + "step": 41003 + }, + { + "epoch": 1.58, + "learning_rate": 2.374070274429341e-05, + "loss": 0.2681, + "step": 41004 + }, + { + "epoch": 1.58, + "learning_rate": 2.374006155424468e-05, + "loss": 0.2502, + "step": 41005 + }, + { + "epoch": 1.58, + "learning_rate": 2.373942036419595e-05, + "loss": 0.3784, + "step": 41006 + }, + { + "epoch": 1.58, + "learning_rate": 2.373877917414722e-05, + "loss": 0.4831, + "step": 41007 + }, + { + "epoch": 1.58, + "learning_rate": 2.3738137984098487e-05, + "loss": 0.3616, + "step": 41008 + }, + { + "epoch": 1.58, + "learning_rate": 2.3737496794049756e-05, + "loss": 0.323, + "step": 41009 + }, + { + "epoch": 1.58, + "learning_rate": 2.3736855604001027e-05, + "loss": 0.2397, + "step": 41010 + }, + { + "epoch": 1.58, + "learning_rate": 2.3736214413952296e-05, + "loss": 0.2967, + "step": 41011 + }, + { + "epoch": 1.58, + "learning_rate": 2.3735573223903564e-05, + "loss": 0.3291, + "step": 41012 + }, + { + "epoch": 1.58, + "learning_rate": 2.3734932033854836e-05, + "loss": 0.3972, + "step": 41013 + }, + { + "epoch": 1.58, + "learning_rate": 2.3734290843806108e-05, + "loss": 0.5921, + "step": 41014 + }, + { + "epoch": 1.58, + "learning_rate": 2.3733649653757376e-05, + "loss": 0.3365, + "step": 41015 + }, + { + "epoch": 1.58, + "learning_rate": 2.3733008463708644e-05, + "loss": 0.3673, + "step": 41016 + }, + { + "epoch": 1.58, + "learning_rate": 2.3732367273659913e-05, + "loss": 0.4605, + "step": 41017 + }, + { + "epoch": 1.58, + "learning_rate": 2.3731726083611184e-05, + "loss": 0.3948, + "step": 41018 + }, + { + "epoch": 1.58, + "learning_rate": 2.3731084893562453e-05, + "loss": 0.2714, + "step": 41019 + }, + { + "epoch": 1.58, + "learning_rate": 2.373044370351372e-05, + "loss": 0.3068, + "step": 41020 + }, + { + "epoch": 1.58, + "learning_rate": 2.3729802513464993e-05, + "loss": 0.2955, + "step": 41021 + }, + { + "epoch": 1.58, + "learning_rate": 2.3729161323416264e-05, + "loss": 0.3816, + "step": 41022 + }, + { + "epoch": 1.58, + "learning_rate": 2.3728520133367533e-05, + "loss": 0.3725, + "step": 41023 + }, + { + "epoch": 1.58, + "learning_rate": 2.37278789433188e-05, + "loss": 0.2785, + "step": 41024 + }, + { + "epoch": 1.58, + "learning_rate": 2.372723775327007e-05, + "loss": 0.2947, + "step": 41025 + }, + { + "epoch": 1.58, + "learning_rate": 2.372659656322134e-05, + "loss": 0.3605, + "step": 41026 + }, + { + "epoch": 1.58, + "learning_rate": 2.372595537317261e-05, + "loss": 0.3426, + "step": 41027 + }, + { + "epoch": 1.58, + "learning_rate": 2.3725314183123878e-05, + "loss": 0.316, + "step": 41028 + }, + { + "epoch": 1.58, + "learning_rate": 2.3724672993075146e-05, + "loss": 0.4337, + "step": 41029 + }, + { + "epoch": 1.58, + "learning_rate": 2.3724031803026418e-05, + "loss": 0.279, + "step": 41030 + }, + { + "epoch": 1.58, + "learning_rate": 2.372339061297769e-05, + "loss": 0.3838, + "step": 41031 + }, + { + "epoch": 1.58, + "learning_rate": 2.3722749422928958e-05, + "loss": 0.4366, + "step": 41032 + }, + { + "epoch": 1.58, + "learning_rate": 2.3722108232880226e-05, + "loss": 0.3151, + "step": 41033 + }, + { + "epoch": 1.58, + "learning_rate": 2.3721467042831498e-05, + "loss": 0.3581, + "step": 41034 + }, + { + "epoch": 1.58, + "learning_rate": 2.3720825852782766e-05, + "loss": 0.3351, + "step": 41035 + }, + { + "epoch": 1.58, + "learning_rate": 2.3720184662734034e-05, + "loss": 0.4204, + "step": 41036 + }, + { + "epoch": 1.58, + "learning_rate": 2.3719543472685303e-05, + "loss": 0.2227, + "step": 41037 + }, + { + "epoch": 1.58, + "learning_rate": 2.3718902282636575e-05, + "loss": 0.4451, + "step": 41038 + }, + { + "epoch": 1.58, + "learning_rate": 2.3718261092587843e-05, + "loss": 0.371, + "step": 41039 + }, + { + "epoch": 1.58, + "learning_rate": 2.3717619902539115e-05, + "loss": 0.3229, + "step": 41040 + }, + { + "epoch": 1.58, + "learning_rate": 2.3716978712490383e-05, + "loss": 0.3518, + "step": 41041 + }, + { + "epoch": 1.58, + "learning_rate": 2.3716337522441655e-05, + "loss": 0.2782, + "step": 41042 + }, + { + "epoch": 1.58, + "learning_rate": 2.3715696332392923e-05, + "loss": 0.3029, + "step": 41043 + }, + { + "epoch": 1.58, + "learning_rate": 2.371505514234419e-05, + "loss": 0.3541, + "step": 41044 + }, + { + "epoch": 1.58, + "learning_rate": 2.371441395229546e-05, + "loss": 0.2542, + "step": 41045 + }, + { + "epoch": 1.58, + "learning_rate": 2.371377276224673e-05, + "loss": 0.3096, + "step": 41046 + }, + { + "epoch": 1.58, + "learning_rate": 2.3713131572198e-05, + "loss": 0.3403, + "step": 41047 + }, + { + "epoch": 1.58, + "learning_rate": 2.3712490382149268e-05, + "loss": 0.2981, + "step": 41048 + }, + { + "epoch": 1.58, + "learning_rate": 2.371184919210054e-05, + "loss": 0.3019, + "step": 41049 + }, + { + "epoch": 1.58, + "learning_rate": 2.371120800205181e-05, + "loss": 0.2955, + "step": 41050 + }, + { + "epoch": 1.58, + "learning_rate": 2.371056681200308e-05, + "loss": 0.2477, + "step": 41051 + }, + { + "epoch": 1.58, + "learning_rate": 2.3709925621954348e-05, + "loss": 0.3995, + "step": 41052 + }, + { + "epoch": 1.58, + "learning_rate": 2.3709284431905616e-05, + "loss": 0.3735, + "step": 41053 + }, + { + "epoch": 1.58, + "learning_rate": 2.3708643241856888e-05, + "loss": 0.4454, + "step": 41054 + }, + { + "epoch": 1.58, + "learning_rate": 2.3708002051808156e-05, + "loss": 0.2947, + "step": 41055 + }, + { + "epoch": 1.58, + "learning_rate": 2.3707360861759425e-05, + "loss": 0.445, + "step": 41056 + }, + { + "epoch": 1.58, + "learning_rate": 2.3706719671710696e-05, + "loss": 0.3789, + "step": 41057 + }, + { + "epoch": 1.58, + "learning_rate": 2.3706078481661968e-05, + "loss": 0.3487, + "step": 41058 + }, + { + "epoch": 1.58, + "learning_rate": 2.3705437291613237e-05, + "loss": 0.3444, + "step": 41059 + }, + { + "epoch": 1.58, + "learning_rate": 2.3704796101564505e-05, + "loss": 0.3923, + "step": 41060 + }, + { + "epoch": 1.58, + "learning_rate": 2.3704154911515773e-05, + "loss": 0.334, + "step": 41061 + }, + { + "epoch": 1.58, + "learning_rate": 2.3703513721467045e-05, + "loss": 0.4044, + "step": 41062 + }, + { + "epoch": 1.58, + "learning_rate": 2.3702872531418313e-05, + "loss": 0.2879, + "step": 41063 + }, + { + "epoch": 1.58, + "learning_rate": 2.370223134136958e-05, + "loss": 0.3732, + "step": 41064 + }, + { + "epoch": 1.58, + "learning_rate": 2.3701590151320853e-05, + "loss": 0.2513, + "step": 41065 + }, + { + "epoch": 1.58, + "learning_rate": 2.370094896127212e-05, + "loss": 0.2651, + "step": 41066 + }, + { + "epoch": 1.58, + "learning_rate": 2.3700307771223393e-05, + "loss": 0.5883, + "step": 41067 + }, + { + "epoch": 1.58, + "learning_rate": 2.369966658117466e-05, + "loss": 0.2952, + "step": 41068 + }, + { + "epoch": 1.58, + "learning_rate": 2.369902539112593e-05, + "loss": 0.3783, + "step": 41069 + }, + { + "epoch": 1.58, + "learning_rate": 2.3698384201077202e-05, + "loss": 0.3514, + "step": 41070 + }, + { + "epoch": 1.58, + "learning_rate": 2.369774301102847e-05, + "loss": 0.3073, + "step": 41071 + }, + { + "epoch": 1.58, + "learning_rate": 2.369710182097974e-05, + "loss": 0.4389, + "step": 41072 + }, + { + "epoch": 1.58, + "learning_rate": 2.3696460630931007e-05, + "loss": 0.3243, + "step": 41073 + }, + { + "epoch": 1.58, + "learning_rate": 2.369581944088228e-05, + "loss": 0.3731, + "step": 41074 + }, + { + "epoch": 1.58, + "learning_rate": 2.369517825083355e-05, + "loss": 0.3752, + "step": 41075 + }, + { + "epoch": 1.58, + "learning_rate": 2.369453706078482e-05, + "loss": 0.4467, + "step": 41076 + }, + { + "epoch": 1.58, + "learning_rate": 2.3693895870736087e-05, + "loss": 0.4481, + "step": 41077 + }, + { + "epoch": 1.58, + "learning_rate": 2.369325468068736e-05, + "loss": 0.3289, + "step": 41078 + }, + { + "epoch": 1.58, + "learning_rate": 2.3692613490638627e-05, + "loss": 0.2914, + "step": 41079 + }, + { + "epoch": 1.58, + "learning_rate": 2.3691972300589895e-05, + "loss": 0.4054, + "step": 41080 + }, + { + "epoch": 1.58, + "learning_rate": 2.3691331110541163e-05, + "loss": 0.3573, + "step": 41081 + }, + { + "epoch": 1.58, + "learning_rate": 2.3690689920492435e-05, + "loss": 0.3756, + "step": 41082 + }, + { + "epoch": 1.58, + "learning_rate": 2.3690048730443704e-05, + "loss": 0.3087, + "step": 41083 + }, + { + "epoch": 1.58, + "learning_rate": 2.3689407540394975e-05, + "loss": 0.3054, + "step": 41084 + }, + { + "epoch": 1.58, + "learning_rate": 2.3688766350346244e-05, + "loss": 0.3957, + "step": 41085 + }, + { + "epoch": 1.58, + "learning_rate": 2.3688125160297515e-05, + "loss": 0.4587, + "step": 41086 + }, + { + "epoch": 1.58, + "learning_rate": 2.3687483970248784e-05, + "loss": 0.3666, + "step": 41087 + }, + { + "epoch": 1.58, + "learning_rate": 2.3686842780200052e-05, + "loss": 0.3093, + "step": 41088 + }, + { + "epoch": 1.58, + "learning_rate": 2.368620159015132e-05, + "loss": 0.2997, + "step": 41089 + }, + { + "epoch": 1.58, + "learning_rate": 2.3685560400102592e-05, + "loss": 0.4193, + "step": 41090 + }, + { + "epoch": 1.58, + "learning_rate": 2.368491921005386e-05, + "loss": 0.3181, + "step": 41091 + }, + { + "epoch": 1.58, + "learning_rate": 2.368427802000513e-05, + "loss": 0.3236, + "step": 41092 + }, + { + "epoch": 1.58, + "learning_rate": 2.36836368299564e-05, + "loss": 0.3454, + "step": 41093 + }, + { + "epoch": 1.58, + "learning_rate": 2.3682995639907672e-05, + "loss": 0.3982, + "step": 41094 + }, + { + "epoch": 1.58, + "learning_rate": 2.368235444985894e-05, + "loss": 0.3887, + "step": 41095 + }, + { + "epoch": 1.58, + "learning_rate": 2.368171325981021e-05, + "loss": 0.4006, + "step": 41096 + }, + { + "epoch": 1.58, + "learning_rate": 2.3681072069761477e-05, + "loss": 0.3451, + "step": 41097 + }, + { + "epoch": 1.58, + "learning_rate": 2.368043087971275e-05, + "loss": 0.281, + "step": 41098 + }, + { + "epoch": 1.58, + "learning_rate": 2.3679789689664017e-05, + "loss": 0.3599, + "step": 41099 + }, + { + "epoch": 1.58, + "learning_rate": 2.3679148499615285e-05, + "loss": 0.3358, + "step": 41100 + }, + { + "epoch": 1.58, + "learning_rate": 2.3678507309566557e-05, + "loss": 0.5537, + "step": 41101 + }, + { + "epoch": 1.58, + "learning_rate": 2.367786611951783e-05, + "loss": 0.3715, + "step": 41102 + }, + { + "epoch": 1.58, + "learning_rate": 2.3677224929469097e-05, + "loss": 0.4262, + "step": 41103 + }, + { + "epoch": 1.58, + "learning_rate": 2.3676583739420366e-05, + "loss": 0.466, + "step": 41104 + }, + { + "epoch": 1.58, + "learning_rate": 2.3675942549371634e-05, + "loss": 0.4206, + "step": 41105 + }, + { + "epoch": 1.58, + "learning_rate": 2.3675301359322906e-05, + "loss": 0.2679, + "step": 41106 + }, + { + "epoch": 1.58, + "learning_rate": 2.3674660169274174e-05, + "loss": 0.3571, + "step": 41107 + }, + { + "epoch": 1.58, + "learning_rate": 2.3674018979225442e-05, + "loss": 0.359, + "step": 41108 + }, + { + "epoch": 1.58, + "learning_rate": 2.367337778917671e-05, + "loss": 0.4757, + "step": 41109 + }, + { + "epoch": 1.58, + "learning_rate": 2.3672736599127982e-05, + "loss": 0.3359, + "step": 41110 + }, + { + "epoch": 1.58, + "learning_rate": 2.3672095409079254e-05, + "loss": 0.3528, + "step": 41111 + }, + { + "epoch": 1.58, + "learning_rate": 2.3671454219030522e-05, + "loss": 0.3335, + "step": 41112 + }, + { + "epoch": 1.58, + "learning_rate": 2.367081302898179e-05, + "loss": 0.2966, + "step": 41113 + }, + { + "epoch": 1.58, + "learning_rate": 2.3670171838933062e-05, + "loss": 0.3495, + "step": 41114 + }, + { + "epoch": 1.58, + "learning_rate": 2.366953064888433e-05, + "loss": 0.3764, + "step": 41115 + }, + { + "epoch": 1.58, + "learning_rate": 2.36688894588356e-05, + "loss": 0.3913, + "step": 41116 + }, + { + "epoch": 1.58, + "learning_rate": 2.3668248268786867e-05, + "loss": 0.347, + "step": 41117 + }, + { + "epoch": 1.58, + "learning_rate": 2.366760707873814e-05, + "loss": 0.277, + "step": 41118 + }, + { + "epoch": 1.58, + "learning_rate": 2.3666965888689407e-05, + "loss": 0.401, + "step": 41119 + }, + { + "epoch": 1.58, + "learning_rate": 2.366632469864068e-05, + "loss": 0.4981, + "step": 41120 + }, + { + "epoch": 1.58, + "learning_rate": 2.3665683508591947e-05, + "loss": 0.4136, + "step": 41121 + }, + { + "epoch": 1.58, + "learning_rate": 2.366504231854322e-05, + "loss": 0.4324, + "step": 41122 + }, + { + "epoch": 1.58, + "learning_rate": 2.3664401128494487e-05, + "loss": 0.3124, + "step": 41123 + }, + { + "epoch": 1.58, + "learning_rate": 2.3663759938445756e-05, + "loss": 0.3205, + "step": 41124 + }, + { + "epoch": 1.58, + "learning_rate": 2.3663118748397024e-05, + "loss": 0.3589, + "step": 41125 + }, + { + "epoch": 1.58, + "learning_rate": 2.3662477558348296e-05, + "loss": 0.3229, + "step": 41126 + }, + { + "epoch": 1.58, + "learning_rate": 2.3661836368299564e-05, + "loss": 0.2942, + "step": 41127 + }, + { + "epoch": 1.58, + "learning_rate": 2.3661195178250833e-05, + "loss": 0.2706, + "step": 41128 + }, + { + "epoch": 1.58, + "learning_rate": 2.3660553988202104e-05, + "loss": 0.4289, + "step": 41129 + }, + { + "epoch": 1.58, + "learning_rate": 2.3659912798153376e-05, + "loss": 0.4164, + "step": 41130 + }, + { + "epoch": 1.58, + "learning_rate": 2.3659271608104644e-05, + "loss": 0.4166, + "step": 41131 + }, + { + "epoch": 1.58, + "learning_rate": 2.3658630418055913e-05, + "loss": 0.2117, + "step": 41132 + }, + { + "epoch": 1.58, + "learning_rate": 2.365798922800718e-05, + "loss": 0.3934, + "step": 41133 + }, + { + "epoch": 1.58, + "learning_rate": 2.3657348037958453e-05, + "loss": 0.4197, + "step": 41134 + }, + { + "epoch": 1.58, + "learning_rate": 2.365670684790972e-05, + "loss": 0.2815, + "step": 41135 + }, + { + "epoch": 1.58, + "learning_rate": 2.365606565786099e-05, + "loss": 0.393, + "step": 41136 + }, + { + "epoch": 1.58, + "learning_rate": 2.365542446781226e-05, + "loss": 0.2912, + "step": 41137 + }, + { + "epoch": 1.58, + "learning_rate": 2.3654783277763533e-05, + "loss": 0.3844, + "step": 41138 + }, + { + "epoch": 1.58, + "learning_rate": 2.36541420877148e-05, + "loss": 0.338, + "step": 41139 + }, + { + "epoch": 1.58, + "learning_rate": 2.365350089766607e-05, + "loss": 0.3291, + "step": 41140 + }, + { + "epoch": 1.58, + "learning_rate": 2.3652859707617338e-05, + "loss": 0.2951, + "step": 41141 + }, + { + "epoch": 1.58, + "learning_rate": 2.365221851756861e-05, + "loss": 0.3008, + "step": 41142 + }, + { + "epoch": 1.58, + "learning_rate": 2.3651577327519878e-05, + "loss": 0.3912, + "step": 41143 + }, + { + "epoch": 1.58, + "learning_rate": 2.3650936137471146e-05, + "loss": 0.3633, + "step": 41144 + }, + { + "epoch": 1.58, + "learning_rate": 2.3650294947422418e-05, + "loss": 0.2834, + "step": 41145 + }, + { + "epoch": 1.58, + "learning_rate": 2.3649653757373686e-05, + "loss": 0.2981, + "step": 41146 + }, + { + "epoch": 1.58, + "learning_rate": 2.3649012567324958e-05, + "loss": 0.2322, + "step": 41147 + }, + { + "epoch": 1.58, + "learning_rate": 2.3648371377276226e-05, + "loss": 0.2545, + "step": 41148 + }, + { + "epoch": 1.58, + "learning_rate": 2.3647730187227495e-05, + "loss": 0.3034, + "step": 41149 + }, + { + "epoch": 1.58, + "learning_rate": 2.3647088997178766e-05, + "loss": 0.296, + "step": 41150 + }, + { + "epoch": 1.58, + "learning_rate": 2.3646447807130035e-05, + "loss": 0.3001, + "step": 41151 + }, + { + "epoch": 1.58, + "learning_rate": 2.3645806617081303e-05, + "loss": 0.4118, + "step": 41152 + }, + { + "epoch": 1.58, + "learning_rate": 2.364516542703257e-05, + "loss": 0.3388, + "step": 41153 + }, + { + "epoch": 1.58, + "learning_rate": 2.3644524236983843e-05, + "loss": 0.2806, + "step": 41154 + }, + { + "epoch": 1.58, + "learning_rate": 2.364388304693511e-05, + "loss": 0.369, + "step": 41155 + }, + { + "epoch": 1.58, + "learning_rate": 2.3643241856886383e-05, + "loss": 0.4259, + "step": 41156 + }, + { + "epoch": 1.58, + "learning_rate": 2.364260066683765e-05, + "loss": 0.3674, + "step": 41157 + }, + { + "epoch": 1.58, + "learning_rate": 2.3641959476788923e-05, + "loss": 0.2668, + "step": 41158 + }, + { + "epoch": 1.58, + "learning_rate": 2.364131828674019e-05, + "loss": 0.3869, + "step": 41159 + }, + { + "epoch": 1.58, + "learning_rate": 2.364067709669146e-05, + "loss": 0.2223, + "step": 41160 + }, + { + "epoch": 1.58, + "learning_rate": 2.3640035906642728e-05, + "loss": 0.2686, + "step": 41161 + }, + { + "epoch": 1.58, + "learning_rate": 2.3639394716594e-05, + "loss": 0.3799, + "step": 41162 + }, + { + "epoch": 1.58, + "learning_rate": 2.3638753526545268e-05, + "loss": 0.2896, + "step": 41163 + }, + { + "epoch": 1.58, + "learning_rate": 2.3638112336496536e-05, + "loss": 0.3496, + "step": 41164 + }, + { + "epoch": 1.58, + "learning_rate": 2.3637471146447808e-05, + "loss": 0.3484, + "step": 41165 + }, + { + "epoch": 1.58, + "learning_rate": 2.363682995639908e-05, + "loss": 0.4497, + "step": 41166 + }, + { + "epoch": 1.58, + "learning_rate": 2.3636188766350348e-05, + "loss": 0.3083, + "step": 41167 + }, + { + "epoch": 1.58, + "learning_rate": 2.3635547576301616e-05, + "loss": 0.4652, + "step": 41168 + }, + { + "epoch": 1.58, + "learning_rate": 2.3634906386252885e-05, + "loss": 0.308, + "step": 41169 + }, + { + "epoch": 1.58, + "learning_rate": 2.3634265196204157e-05, + "loss": 0.3502, + "step": 41170 + }, + { + "epoch": 1.58, + "learning_rate": 2.3633624006155425e-05, + "loss": 0.3695, + "step": 41171 + }, + { + "epoch": 1.58, + "learning_rate": 2.3632982816106693e-05, + "loss": 0.3404, + "step": 41172 + }, + { + "epoch": 1.58, + "learning_rate": 2.3632341626057965e-05, + "loss": 0.405, + "step": 41173 + }, + { + "epoch": 1.58, + "learning_rate": 2.3631700436009237e-05, + "loss": 0.4541, + "step": 41174 + }, + { + "epoch": 1.58, + "learning_rate": 2.3631059245960505e-05, + "loss": 0.4418, + "step": 41175 + }, + { + "epoch": 1.58, + "learning_rate": 2.3630418055911773e-05, + "loss": 0.3634, + "step": 41176 + }, + { + "epoch": 1.58, + "learning_rate": 2.362977686586304e-05, + "loss": 0.2997, + "step": 41177 + }, + { + "epoch": 1.58, + "learning_rate": 2.3629135675814313e-05, + "loss": 0.5294, + "step": 41178 + }, + { + "epoch": 1.58, + "learning_rate": 2.362849448576558e-05, + "loss": 0.2394, + "step": 41179 + }, + { + "epoch": 1.58, + "learning_rate": 2.362785329571685e-05, + "loss": 0.3498, + "step": 41180 + }, + { + "epoch": 1.58, + "learning_rate": 2.362721210566812e-05, + "loss": 0.3323, + "step": 41181 + }, + { + "epoch": 1.58, + "learning_rate": 2.3626570915619393e-05, + "loss": 0.2606, + "step": 41182 + }, + { + "epoch": 1.58, + "learning_rate": 2.3625929725570662e-05, + "loss": 0.5915, + "step": 41183 + }, + { + "epoch": 1.58, + "learning_rate": 2.362528853552193e-05, + "loss": 0.4995, + "step": 41184 + }, + { + "epoch": 1.58, + "learning_rate": 2.36246473454732e-05, + "loss": 0.3429, + "step": 41185 + }, + { + "epoch": 1.58, + "learning_rate": 2.362400615542447e-05, + "loss": 0.271, + "step": 41186 + }, + { + "epoch": 1.58, + "learning_rate": 2.362336496537574e-05, + "loss": 0.2855, + "step": 41187 + }, + { + "epoch": 1.58, + "learning_rate": 2.3622723775327007e-05, + "loss": 0.3879, + "step": 41188 + }, + { + "epoch": 1.58, + "learning_rate": 2.3622082585278275e-05, + "loss": 0.286, + "step": 41189 + }, + { + "epoch": 1.58, + "learning_rate": 2.3621441395229547e-05, + "loss": 0.318, + "step": 41190 + }, + { + "epoch": 1.58, + "learning_rate": 2.362080020518082e-05, + "loss": 0.3613, + "step": 41191 + }, + { + "epoch": 1.58, + "learning_rate": 2.3620159015132087e-05, + "loss": 0.3689, + "step": 41192 + }, + { + "epoch": 1.58, + "learning_rate": 2.3619517825083355e-05, + "loss": 0.4112, + "step": 41193 + }, + { + "epoch": 1.58, + "learning_rate": 2.3618876635034627e-05, + "loss": 0.4216, + "step": 41194 + }, + { + "epoch": 1.58, + "learning_rate": 2.3618235444985895e-05, + "loss": 0.3373, + "step": 41195 + }, + { + "epoch": 1.58, + "learning_rate": 2.3617594254937164e-05, + "loss": 0.3006, + "step": 41196 + }, + { + "epoch": 1.58, + "learning_rate": 2.3616953064888432e-05, + "loss": 0.4349, + "step": 41197 + }, + { + "epoch": 1.58, + "learning_rate": 2.3616311874839704e-05, + "loss": 0.3384, + "step": 41198 + }, + { + "epoch": 1.58, + "learning_rate": 2.3615670684790972e-05, + "loss": 0.3348, + "step": 41199 + }, + { + "epoch": 1.58, + "learning_rate": 2.3615029494742244e-05, + "loss": 0.416, + "step": 41200 + }, + { + "epoch": 1.58, + "learning_rate": 2.3614388304693512e-05, + "loss": 0.4418, + "step": 41201 + }, + { + "epoch": 1.58, + "learning_rate": 2.3613747114644784e-05, + "loss": 0.3975, + "step": 41202 + }, + { + "epoch": 1.58, + "learning_rate": 2.3613105924596052e-05, + "loss": 0.425, + "step": 41203 + }, + { + "epoch": 1.58, + "learning_rate": 2.361246473454732e-05, + "loss": 0.3563, + "step": 41204 + }, + { + "epoch": 1.58, + "learning_rate": 2.361182354449859e-05, + "loss": 0.3362, + "step": 41205 + }, + { + "epoch": 1.58, + "learning_rate": 2.361118235444986e-05, + "loss": 0.2585, + "step": 41206 + }, + { + "epoch": 1.58, + "learning_rate": 2.361054116440113e-05, + "loss": 0.3364, + "step": 41207 + }, + { + "epoch": 1.58, + "learning_rate": 2.3609899974352397e-05, + "loss": 0.5828, + "step": 41208 + }, + { + "epoch": 1.58, + "learning_rate": 2.360925878430367e-05, + "loss": 0.3106, + "step": 41209 + }, + { + "epoch": 1.58, + "learning_rate": 2.360861759425494e-05, + "loss": 0.3159, + "step": 41210 + }, + { + "epoch": 1.58, + "learning_rate": 2.360797640420621e-05, + "loss": 0.3138, + "step": 41211 + }, + { + "epoch": 1.58, + "learning_rate": 2.3607335214157477e-05, + "loss": 0.3183, + "step": 41212 + }, + { + "epoch": 1.58, + "learning_rate": 2.3606694024108745e-05, + "loss": 0.2657, + "step": 41213 + }, + { + "epoch": 1.58, + "learning_rate": 2.3606052834060017e-05, + "loss": 0.3152, + "step": 41214 + }, + { + "epoch": 1.58, + "learning_rate": 2.3605411644011286e-05, + "loss": 0.2908, + "step": 41215 + }, + { + "epoch": 1.58, + "learning_rate": 2.3604770453962554e-05, + "loss": 0.3736, + "step": 41216 + }, + { + "epoch": 1.58, + "learning_rate": 2.3604129263913826e-05, + "loss": 0.4014, + "step": 41217 + }, + { + "epoch": 1.58, + "learning_rate": 2.3603488073865097e-05, + "loss": 0.3055, + "step": 41218 + }, + { + "epoch": 1.58, + "learning_rate": 2.3602846883816366e-05, + "loss": 0.375, + "step": 41219 + }, + { + "epoch": 1.58, + "learning_rate": 2.3602205693767634e-05, + "loss": 0.431, + "step": 41220 + }, + { + "epoch": 1.58, + "learning_rate": 2.3601564503718902e-05, + "loss": 0.342, + "step": 41221 + }, + { + "epoch": 1.58, + "learning_rate": 2.3600923313670174e-05, + "loss": 0.3248, + "step": 41222 + }, + { + "epoch": 1.58, + "learning_rate": 2.3600282123621442e-05, + "loss": 0.3183, + "step": 41223 + }, + { + "epoch": 1.58, + "learning_rate": 2.359964093357271e-05, + "loss": 0.4408, + "step": 41224 + }, + { + "epoch": 1.58, + "learning_rate": 2.3598999743523982e-05, + "loss": 0.2746, + "step": 41225 + }, + { + "epoch": 1.59, + "learning_rate": 2.359835855347525e-05, + "loss": 0.3659, + "step": 41226 + }, + { + "epoch": 1.59, + "learning_rate": 2.3597717363426522e-05, + "loss": 0.3328, + "step": 41227 + }, + { + "epoch": 1.59, + "learning_rate": 2.359707617337779e-05, + "loss": 0.3478, + "step": 41228 + }, + { + "epoch": 1.59, + "learning_rate": 2.359643498332906e-05, + "loss": 0.3834, + "step": 41229 + }, + { + "epoch": 1.59, + "learning_rate": 2.359579379328033e-05, + "loss": 0.355, + "step": 41230 + }, + { + "epoch": 1.59, + "learning_rate": 2.35951526032316e-05, + "loss": 0.4776, + "step": 41231 + }, + { + "epoch": 1.59, + "learning_rate": 2.3594511413182867e-05, + "loss": 0.3272, + "step": 41232 + }, + { + "epoch": 1.59, + "learning_rate": 2.3593870223134136e-05, + "loss": 0.5239, + "step": 41233 + }, + { + "epoch": 1.59, + "learning_rate": 2.3593229033085407e-05, + "loss": 0.3484, + "step": 41234 + }, + { + "epoch": 1.59, + "learning_rate": 2.3592587843036676e-05, + "loss": 0.3513, + "step": 41235 + }, + { + "epoch": 1.59, + "learning_rate": 2.3591946652987948e-05, + "loss": 0.2869, + "step": 41236 + }, + { + "epoch": 1.59, + "learning_rate": 2.3591305462939216e-05, + "loss": 0.2594, + "step": 41237 + }, + { + "epoch": 1.59, + "learning_rate": 2.3590664272890488e-05, + "loss": 0.1972, + "step": 41238 + }, + { + "epoch": 1.59, + "learning_rate": 2.3590023082841756e-05, + "loss": 0.3015, + "step": 41239 + }, + { + "epoch": 1.59, + "learning_rate": 2.3589381892793024e-05, + "loss": 0.2554, + "step": 41240 + }, + { + "epoch": 1.59, + "learning_rate": 2.3588740702744293e-05, + "loss": 0.3897, + "step": 41241 + }, + { + "epoch": 1.59, + "learning_rate": 2.3588099512695564e-05, + "loss": 0.2622, + "step": 41242 + }, + { + "epoch": 1.59, + "learning_rate": 2.3587458322646833e-05, + "loss": 0.4102, + "step": 41243 + }, + { + "epoch": 1.59, + "learning_rate": 2.35868171325981e-05, + "loss": 0.3206, + "step": 41244 + }, + { + "epoch": 1.59, + "learning_rate": 2.3586175942549373e-05, + "loss": 0.404, + "step": 41245 + }, + { + "epoch": 1.59, + "learning_rate": 2.3585534752500644e-05, + "loss": 0.4555, + "step": 41246 + }, + { + "epoch": 1.59, + "learning_rate": 2.3584893562451913e-05, + "loss": 0.2667, + "step": 41247 + }, + { + "epoch": 1.59, + "learning_rate": 2.358425237240318e-05, + "loss": 0.3195, + "step": 41248 + }, + { + "epoch": 1.59, + "learning_rate": 2.358361118235445e-05, + "loss": 0.2255, + "step": 41249 + }, + { + "epoch": 1.59, + "learning_rate": 2.358296999230572e-05, + "loss": 0.3988, + "step": 41250 + }, + { + "epoch": 1.59, + "learning_rate": 2.358232880225699e-05, + "loss": 0.3371, + "step": 41251 + }, + { + "epoch": 1.59, + "learning_rate": 2.3581687612208258e-05, + "loss": 0.3964, + "step": 41252 + }, + { + "epoch": 1.59, + "learning_rate": 2.358104642215953e-05, + "loss": 0.4107, + "step": 41253 + }, + { + "epoch": 1.59, + "learning_rate": 2.35804052321108e-05, + "loss": 0.4787, + "step": 41254 + }, + { + "epoch": 1.59, + "learning_rate": 2.357976404206207e-05, + "loss": 0.3437, + "step": 41255 + }, + { + "epoch": 1.59, + "learning_rate": 2.3579122852013338e-05, + "loss": 0.3098, + "step": 41256 + }, + { + "epoch": 1.59, + "learning_rate": 2.3578481661964606e-05, + "loss": 0.3019, + "step": 41257 + }, + { + "epoch": 1.59, + "learning_rate": 2.3577840471915878e-05, + "loss": 0.4242, + "step": 41258 + }, + { + "epoch": 1.59, + "learning_rate": 2.3577199281867146e-05, + "loss": 0.2385, + "step": 41259 + }, + { + "epoch": 1.59, + "learning_rate": 2.3576558091818415e-05, + "loss": 0.4037, + "step": 41260 + }, + { + "epoch": 1.59, + "learning_rate": 2.3575916901769686e-05, + "loss": 0.3159, + "step": 41261 + }, + { + "epoch": 1.59, + "learning_rate": 2.3575275711720955e-05, + "loss": 0.3444, + "step": 41262 + }, + { + "epoch": 1.59, + "learning_rate": 2.3574634521672226e-05, + "loss": 0.4018, + "step": 41263 + }, + { + "epoch": 1.59, + "learning_rate": 2.3573993331623495e-05, + "loss": 0.4152, + "step": 41264 + }, + { + "epoch": 1.59, + "learning_rate": 2.3573352141574763e-05, + "loss": 0.3619, + "step": 41265 + }, + { + "epoch": 1.59, + "learning_rate": 2.3572710951526035e-05, + "loss": 0.2982, + "step": 41266 + }, + { + "epoch": 1.59, + "learning_rate": 2.3572069761477303e-05, + "loss": 0.3144, + "step": 41267 + }, + { + "epoch": 1.59, + "learning_rate": 2.357142857142857e-05, + "loss": 0.3658, + "step": 41268 + }, + { + "epoch": 1.59, + "learning_rate": 2.3570787381379843e-05, + "loss": 0.3417, + "step": 41269 + }, + { + "epoch": 1.59, + "learning_rate": 2.357014619133111e-05, + "loss": 0.3546, + "step": 41270 + }, + { + "epoch": 1.59, + "learning_rate": 2.3569505001282383e-05, + "loss": 0.3833, + "step": 41271 + }, + { + "epoch": 1.59, + "learning_rate": 2.356886381123365e-05, + "loss": 0.3293, + "step": 41272 + }, + { + "epoch": 1.59, + "learning_rate": 2.356822262118492e-05, + "loss": 0.4245, + "step": 41273 + }, + { + "epoch": 1.59, + "learning_rate": 2.356758143113619e-05, + "loss": 0.2891, + "step": 41274 + }, + { + "epoch": 1.59, + "learning_rate": 2.356694024108746e-05, + "loss": 0.403, + "step": 41275 + }, + { + "epoch": 1.59, + "learning_rate": 2.3566299051038728e-05, + "loss": 0.3924, + "step": 41276 + }, + { + "epoch": 1.59, + "learning_rate": 2.3565657860989996e-05, + "loss": 0.2912, + "step": 41277 + }, + { + "epoch": 1.59, + "learning_rate": 2.3565016670941268e-05, + "loss": 0.3858, + "step": 41278 + }, + { + "epoch": 1.59, + "learning_rate": 2.3564375480892536e-05, + "loss": 0.2858, + "step": 41279 + }, + { + "epoch": 1.59, + "learning_rate": 2.3563734290843808e-05, + "loss": 0.378, + "step": 41280 + }, + { + "epoch": 1.59, + "learning_rate": 2.3563093100795077e-05, + "loss": 0.4279, + "step": 41281 + }, + { + "epoch": 1.59, + "learning_rate": 2.3562451910746348e-05, + "loss": 0.2994, + "step": 41282 + }, + { + "epoch": 1.59, + "learning_rate": 2.3561810720697617e-05, + "loss": 0.4538, + "step": 41283 + }, + { + "epoch": 1.59, + "learning_rate": 2.3561169530648885e-05, + "loss": 0.4009, + "step": 41284 + }, + { + "epoch": 1.59, + "learning_rate": 2.3560528340600153e-05, + "loss": 0.367, + "step": 41285 + }, + { + "epoch": 1.59, + "learning_rate": 2.3559887150551425e-05, + "loss": 0.4944, + "step": 41286 + }, + { + "epoch": 1.59, + "learning_rate": 2.3559245960502693e-05, + "loss": 0.362, + "step": 41287 + }, + { + "epoch": 1.59, + "learning_rate": 2.355860477045396e-05, + "loss": 0.2405, + "step": 41288 + }, + { + "epoch": 1.59, + "learning_rate": 2.3557963580405233e-05, + "loss": 0.3122, + "step": 41289 + }, + { + "epoch": 1.59, + "learning_rate": 2.3557322390356505e-05, + "loss": 0.3571, + "step": 41290 + }, + { + "epoch": 1.59, + "learning_rate": 2.3556681200307773e-05, + "loss": 0.4801, + "step": 41291 + }, + { + "epoch": 1.59, + "learning_rate": 2.355604001025904e-05, + "loss": 0.3725, + "step": 41292 + }, + { + "epoch": 1.59, + "learning_rate": 2.355539882021031e-05, + "loss": 0.3283, + "step": 41293 + }, + { + "epoch": 1.59, + "learning_rate": 2.3554757630161582e-05, + "loss": 0.3097, + "step": 41294 + }, + { + "epoch": 1.59, + "learning_rate": 2.355411644011285e-05, + "loss": 0.3002, + "step": 41295 + }, + { + "epoch": 1.59, + "learning_rate": 2.355347525006412e-05, + "loss": 0.3423, + "step": 41296 + }, + { + "epoch": 1.59, + "learning_rate": 2.355283406001539e-05, + "loss": 0.3859, + "step": 41297 + }, + { + "epoch": 1.59, + "learning_rate": 2.3552192869966662e-05, + "loss": 0.527, + "step": 41298 + }, + { + "epoch": 1.59, + "learning_rate": 2.355155167991793e-05, + "loss": 0.3061, + "step": 41299 + }, + { + "epoch": 1.59, + "learning_rate": 2.35509104898692e-05, + "loss": 0.3401, + "step": 41300 + }, + { + "epoch": 1.59, + "learning_rate": 2.3550269299820467e-05, + "loss": 0.3256, + "step": 41301 + }, + { + "epoch": 1.59, + "learning_rate": 2.354962810977174e-05, + "loss": 0.3009, + "step": 41302 + }, + { + "epoch": 1.59, + "learning_rate": 2.3548986919723007e-05, + "loss": 0.297, + "step": 41303 + }, + { + "epoch": 1.59, + "learning_rate": 2.3548345729674275e-05, + "loss": 0.347, + "step": 41304 + }, + { + "epoch": 1.59, + "learning_rate": 2.3547704539625547e-05, + "loss": 0.4636, + "step": 41305 + }, + { + "epoch": 1.59, + "learning_rate": 2.3547063349576815e-05, + "loss": 0.4772, + "step": 41306 + }, + { + "epoch": 1.59, + "learning_rate": 2.3546422159528087e-05, + "loss": 0.4633, + "step": 41307 + }, + { + "epoch": 1.59, + "learning_rate": 2.3545780969479355e-05, + "loss": 0.3159, + "step": 41308 + }, + { + "epoch": 1.59, + "learning_rate": 2.3545139779430624e-05, + "loss": 0.2987, + "step": 41309 + }, + { + "epoch": 1.59, + "learning_rate": 2.3544498589381895e-05, + "loss": 0.3968, + "step": 41310 + }, + { + "epoch": 1.59, + "learning_rate": 2.3543857399333164e-05, + "loss": 0.2856, + "step": 41311 + }, + { + "epoch": 1.59, + "learning_rate": 2.3543216209284432e-05, + "loss": 0.358, + "step": 41312 + }, + { + "epoch": 1.59, + "learning_rate": 2.35425750192357e-05, + "loss": 0.3389, + "step": 41313 + }, + { + "epoch": 1.59, + "learning_rate": 2.3541933829186972e-05, + "loss": 0.3923, + "step": 41314 + }, + { + "epoch": 1.59, + "learning_rate": 2.354129263913824e-05, + "loss": 0.2882, + "step": 41315 + }, + { + "epoch": 1.59, + "learning_rate": 2.3540651449089512e-05, + "loss": 0.4439, + "step": 41316 + }, + { + "epoch": 1.59, + "learning_rate": 2.354001025904078e-05, + "loss": 0.4314, + "step": 41317 + }, + { + "epoch": 1.59, + "learning_rate": 2.3539369068992052e-05, + "loss": 0.427, + "step": 41318 + }, + { + "epoch": 1.59, + "learning_rate": 2.353872787894332e-05, + "loss": 0.3517, + "step": 41319 + }, + { + "epoch": 1.59, + "learning_rate": 2.353808668889459e-05, + "loss": 0.4811, + "step": 41320 + }, + { + "epoch": 1.59, + "learning_rate": 2.3537445498845857e-05, + "loss": 0.375, + "step": 41321 + }, + { + "epoch": 1.59, + "learning_rate": 2.353680430879713e-05, + "loss": 0.331, + "step": 41322 + }, + { + "epoch": 1.59, + "learning_rate": 2.3536163118748397e-05, + "loss": 0.34, + "step": 41323 + }, + { + "epoch": 1.59, + "learning_rate": 2.3535521928699665e-05, + "loss": 0.3249, + "step": 41324 + }, + { + "epoch": 1.59, + "learning_rate": 2.3534880738650937e-05, + "loss": 0.2669, + "step": 41325 + }, + { + "epoch": 1.59, + "learning_rate": 2.353423954860221e-05, + "loss": 0.2291, + "step": 41326 + }, + { + "epoch": 1.59, + "learning_rate": 2.3533598358553477e-05, + "loss": 0.2833, + "step": 41327 + }, + { + "epoch": 1.59, + "learning_rate": 2.3532957168504746e-05, + "loss": 0.4076, + "step": 41328 + }, + { + "epoch": 1.59, + "learning_rate": 2.3532315978456014e-05, + "loss": 0.3579, + "step": 41329 + }, + { + "epoch": 1.59, + "learning_rate": 2.3531674788407286e-05, + "loss": 0.3308, + "step": 41330 + }, + { + "epoch": 1.59, + "learning_rate": 2.3531033598358554e-05, + "loss": 0.3366, + "step": 41331 + }, + { + "epoch": 1.59, + "learning_rate": 2.3530392408309822e-05, + "loss": 0.3741, + "step": 41332 + }, + { + "epoch": 1.59, + "learning_rate": 2.3529751218261094e-05, + "loss": 0.3761, + "step": 41333 + }, + { + "epoch": 1.59, + "learning_rate": 2.3529110028212366e-05, + "loss": 0.2541, + "step": 41334 + }, + { + "epoch": 1.59, + "learning_rate": 2.3528468838163634e-05, + "loss": 0.3058, + "step": 41335 + }, + { + "epoch": 1.59, + "learning_rate": 2.3527827648114902e-05, + "loss": 0.3934, + "step": 41336 + }, + { + "epoch": 1.59, + "learning_rate": 2.352718645806617e-05, + "loss": 0.2706, + "step": 41337 + }, + { + "epoch": 1.59, + "learning_rate": 2.3526545268017442e-05, + "loss": 0.26, + "step": 41338 + }, + { + "epoch": 1.59, + "learning_rate": 2.352590407796871e-05, + "loss": 0.3635, + "step": 41339 + }, + { + "epoch": 1.59, + "learning_rate": 2.352526288791998e-05, + "loss": 0.3827, + "step": 41340 + }, + { + "epoch": 1.59, + "learning_rate": 2.352462169787125e-05, + "loss": 0.3439, + "step": 41341 + }, + { + "epoch": 1.59, + "learning_rate": 2.352398050782252e-05, + "loss": 0.3467, + "step": 41342 + }, + { + "epoch": 1.59, + "learning_rate": 2.352333931777379e-05, + "loss": 0.3423, + "step": 41343 + }, + { + "epoch": 1.59, + "learning_rate": 2.352269812772506e-05, + "loss": 0.4624, + "step": 41344 + }, + { + "epoch": 1.59, + "learning_rate": 2.3522056937676327e-05, + "loss": 0.5265, + "step": 41345 + }, + { + "epoch": 1.59, + "learning_rate": 2.35214157476276e-05, + "loss": 0.4421, + "step": 41346 + }, + { + "epoch": 1.59, + "learning_rate": 2.3520774557578868e-05, + "loss": 0.2951, + "step": 41347 + }, + { + "epoch": 1.59, + "learning_rate": 2.3520133367530136e-05, + "loss": 0.3618, + "step": 41348 + }, + { + "epoch": 1.59, + "learning_rate": 2.3519492177481408e-05, + "loss": 0.3701, + "step": 41349 + }, + { + "epoch": 1.59, + "learning_rate": 2.3518850987432676e-05, + "loss": 0.3089, + "step": 41350 + }, + { + "epoch": 1.59, + "learning_rate": 2.3518209797383944e-05, + "loss": 0.2489, + "step": 41351 + }, + { + "epoch": 1.59, + "learning_rate": 2.3517568607335216e-05, + "loss": 0.4108, + "step": 41352 + }, + { + "epoch": 1.59, + "learning_rate": 2.3516927417286484e-05, + "loss": 0.3415, + "step": 41353 + }, + { + "epoch": 1.59, + "learning_rate": 2.3516286227237756e-05, + "loss": 0.1851, + "step": 41354 + }, + { + "epoch": 1.59, + "learning_rate": 2.3515645037189024e-05, + "loss": 0.3312, + "step": 41355 + }, + { + "epoch": 1.59, + "learning_rate": 2.3515003847140293e-05, + "loss": 0.3441, + "step": 41356 + }, + { + "epoch": 1.59, + "learning_rate": 2.351436265709156e-05, + "loss": 0.2283, + "step": 41357 + }, + { + "epoch": 1.59, + "learning_rate": 2.3513721467042833e-05, + "loss": 0.2965, + "step": 41358 + }, + { + "epoch": 1.59, + "learning_rate": 2.35130802769941e-05, + "loss": 0.3028, + "step": 41359 + }, + { + "epoch": 1.59, + "learning_rate": 2.3512439086945373e-05, + "loss": 0.2897, + "step": 41360 + }, + { + "epoch": 1.59, + "learning_rate": 2.351179789689664e-05, + "loss": 0.3272, + "step": 41361 + }, + { + "epoch": 1.59, + "learning_rate": 2.3511156706847913e-05, + "loss": 0.4333, + "step": 41362 + }, + { + "epoch": 1.59, + "learning_rate": 2.351051551679918e-05, + "loss": 0.4463, + "step": 41363 + }, + { + "epoch": 1.59, + "learning_rate": 2.350987432675045e-05, + "loss": 0.3142, + "step": 41364 + }, + { + "epoch": 1.59, + "learning_rate": 2.3509233136701718e-05, + "loss": 0.2784, + "step": 41365 + }, + { + "epoch": 1.59, + "learning_rate": 2.350859194665299e-05, + "loss": 0.3308, + "step": 41366 + }, + { + "epoch": 1.59, + "learning_rate": 2.3507950756604258e-05, + "loss": 0.5076, + "step": 41367 + }, + { + "epoch": 1.59, + "learning_rate": 2.3507309566555526e-05, + "loss": 0.3676, + "step": 41368 + }, + { + "epoch": 1.59, + "learning_rate": 2.3506668376506798e-05, + "loss": 0.4202, + "step": 41369 + }, + { + "epoch": 1.59, + "learning_rate": 2.350602718645807e-05, + "loss": 0.3121, + "step": 41370 + }, + { + "epoch": 1.59, + "learning_rate": 2.3505385996409338e-05, + "loss": 0.378, + "step": 41371 + }, + { + "epoch": 1.59, + "learning_rate": 2.3504744806360606e-05, + "loss": 0.2596, + "step": 41372 + }, + { + "epoch": 1.59, + "learning_rate": 2.3504103616311875e-05, + "loss": 0.3874, + "step": 41373 + }, + { + "epoch": 1.59, + "learning_rate": 2.3503462426263146e-05, + "loss": 0.3609, + "step": 41374 + }, + { + "epoch": 1.59, + "learning_rate": 2.3502821236214415e-05, + "loss": 0.3988, + "step": 41375 + }, + { + "epoch": 1.59, + "learning_rate": 2.3502180046165683e-05, + "loss": 0.3116, + "step": 41376 + }, + { + "epoch": 1.59, + "learning_rate": 2.3501538856116955e-05, + "loss": 0.3957, + "step": 41377 + }, + { + "epoch": 1.59, + "learning_rate": 2.3500897666068226e-05, + "loss": 0.3466, + "step": 41378 + }, + { + "epoch": 1.59, + "learning_rate": 2.3500256476019495e-05, + "loss": 0.386, + "step": 41379 + }, + { + "epoch": 1.59, + "learning_rate": 2.3499615285970763e-05, + "loss": 0.3449, + "step": 41380 + }, + { + "epoch": 1.59, + "learning_rate": 2.349897409592203e-05, + "loss": 0.3404, + "step": 41381 + }, + { + "epoch": 1.59, + "learning_rate": 2.3498332905873303e-05, + "loss": 0.2613, + "step": 41382 + }, + { + "epoch": 1.59, + "learning_rate": 2.349769171582457e-05, + "loss": 0.4049, + "step": 41383 + }, + { + "epoch": 1.59, + "learning_rate": 2.349705052577584e-05, + "loss": 0.3234, + "step": 41384 + }, + { + "epoch": 1.59, + "learning_rate": 2.349640933572711e-05, + "loss": 0.3595, + "step": 41385 + }, + { + "epoch": 1.59, + "learning_rate": 2.349576814567838e-05, + "loss": 0.4266, + "step": 41386 + }, + { + "epoch": 1.59, + "learning_rate": 2.349512695562965e-05, + "loss": 0.4459, + "step": 41387 + }, + { + "epoch": 1.59, + "learning_rate": 2.349448576558092e-05, + "loss": 0.2297, + "step": 41388 + }, + { + "epoch": 1.59, + "learning_rate": 2.3493844575532188e-05, + "loss": 0.4124, + "step": 41389 + }, + { + "epoch": 1.59, + "learning_rate": 2.349320338548346e-05, + "loss": 0.3184, + "step": 41390 + }, + { + "epoch": 1.59, + "learning_rate": 2.3492562195434728e-05, + "loss": 0.3264, + "step": 41391 + }, + { + "epoch": 1.59, + "learning_rate": 2.3491921005385996e-05, + "loss": 0.4809, + "step": 41392 + }, + { + "epoch": 1.59, + "learning_rate": 2.3491279815337265e-05, + "loss": 0.2537, + "step": 41393 + }, + { + "epoch": 1.59, + "learning_rate": 2.3490638625288537e-05, + "loss": 0.2804, + "step": 41394 + }, + { + "epoch": 1.59, + "learning_rate": 2.3489997435239805e-05, + "loss": 0.4636, + "step": 41395 + }, + { + "epoch": 1.59, + "learning_rate": 2.3489356245191077e-05, + "loss": 0.4991, + "step": 41396 + }, + { + "epoch": 1.59, + "learning_rate": 2.3488715055142345e-05, + "loss": 0.3808, + "step": 41397 + }, + { + "epoch": 1.59, + "learning_rate": 2.3488073865093617e-05, + "loss": 0.4214, + "step": 41398 + }, + { + "epoch": 1.59, + "learning_rate": 2.3487432675044885e-05, + "loss": 0.259, + "step": 41399 + }, + { + "epoch": 1.59, + "learning_rate": 2.3486791484996153e-05, + "loss": 0.2126, + "step": 41400 + }, + { + "epoch": 1.59, + "learning_rate": 2.348615029494742e-05, + "loss": 0.3114, + "step": 41401 + }, + { + "epoch": 1.59, + "learning_rate": 2.3485509104898693e-05, + "loss": 0.3504, + "step": 41402 + }, + { + "epoch": 1.59, + "learning_rate": 2.348486791484996e-05, + "loss": 0.3414, + "step": 41403 + }, + { + "epoch": 1.59, + "learning_rate": 2.348422672480123e-05, + "loss": 0.3935, + "step": 41404 + }, + { + "epoch": 1.59, + "learning_rate": 2.3483585534752502e-05, + "loss": 0.3883, + "step": 41405 + }, + { + "epoch": 1.59, + "learning_rate": 2.3482944344703773e-05, + "loss": 0.3799, + "step": 41406 + }, + { + "epoch": 1.59, + "learning_rate": 2.3482303154655042e-05, + "loss": 0.2827, + "step": 41407 + }, + { + "epoch": 1.59, + "learning_rate": 2.348166196460631e-05, + "loss": 0.2203, + "step": 41408 + }, + { + "epoch": 1.59, + "learning_rate": 2.348102077455758e-05, + "loss": 0.4152, + "step": 41409 + }, + { + "epoch": 1.59, + "learning_rate": 2.348037958450885e-05, + "loss": 0.498, + "step": 41410 + }, + { + "epoch": 1.59, + "learning_rate": 2.347973839446012e-05, + "loss": 0.3392, + "step": 41411 + }, + { + "epoch": 1.59, + "learning_rate": 2.3479097204411387e-05, + "loss": 0.4696, + "step": 41412 + }, + { + "epoch": 1.59, + "learning_rate": 2.347845601436266e-05, + "loss": 0.3887, + "step": 41413 + }, + { + "epoch": 1.59, + "learning_rate": 2.347781482431393e-05, + "loss": 0.3055, + "step": 41414 + }, + { + "epoch": 1.59, + "learning_rate": 2.34771736342652e-05, + "loss": 0.4193, + "step": 41415 + }, + { + "epoch": 1.59, + "learning_rate": 2.3476532444216467e-05, + "loss": 0.4137, + "step": 41416 + }, + { + "epoch": 1.59, + "learning_rate": 2.3475891254167735e-05, + "loss": 0.3528, + "step": 41417 + }, + { + "epoch": 1.59, + "learning_rate": 2.3475250064119007e-05, + "loss": 0.268, + "step": 41418 + }, + { + "epoch": 1.59, + "learning_rate": 2.3474608874070275e-05, + "loss": 0.4, + "step": 41419 + }, + { + "epoch": 1.59, + "learning_rate": 2.3473967684021544e-05, + "loss": 0.299, + "step": 41420 + }, + { + "epoch": 1.59, + "learning_rate": 2.3473326493972815e-05, + "loss": 0.2863, + "step": 41421 + }, + { + "epoch": 1.59, + "learning_rate": 2.3472685303924084e-05, + "loss": 0.3736, + "step": 41422 + }, + { + "epoch": 1.59, + "learning_rate": 2.3472044113875355e-05, + "loss": 0.554, + "step": 41423 + }, + { + "epoch": 1.59, + "learning_rate": 2.3471402923826624e-05, + "loss": 0.3305, + "step": 41424 + }, + { + "epoch": 1.59, + "learning_rate": 2.3470761733777892e-05, + "loss": 0.3365, + "step": 41425 + }, + { + "epoch": 1.59, + "learning_rate": 2.3470120543729164e-05, + "loss": 0.2062, + "step": 41426 + }, + { + "epoch": 1.59, + "learning_rate": 2.3469479353680432e-05, + "loss": 0.3386, + "step": 41427 + }, + { + "epoch": 1.59, + "learning_rate": 2.34688381636317e-05, + "loss": 0.3584, + "step": 41428 + }, + { + "epoch": 1.59, + "learning_rate": 2.3468196973582972e-05, + "loss": 0.2929, + "step": 41429 + }, + { + "epoch": 1.59, + "learning_rate": 2.346755578353424e-05, + "loss": 0.3149, + "step": 41430 + }, + { + "epoch": 1.59, + "learning_rate": 2.346691459348551e-05, + "loss": 0.3789, + "step": 41431 + }, + { + "epoch": 1.59, + "learning_rate": 2.346627340343678e-05, + "loss": 0.4367, + "step": 41432 + }, + { + "epoch": 1.59, + "learning_rate": 2.346563221338805e-05, + "loss": 0.2893, + "step": 41433 + }, + { + "epoch": 1.59, + "learning_rate": 2.346499102333932e-05, + "loss": 0.32, + "step": 41434 + }, + { + "epoch": 1.59, + "learning_rate": 2.346434983329059e-05, + "loss": 0.2698, + "step": 41435 + }, + { + "epoch": 1.59, + "learning_rate": 2.3463708643241857e-05, + "loss": 0.3039, + "step": 41436 + }, + { + "epoch": 1.59, + "learning_rate": 2.3463067453193125e-05, + "loss": 0.3794, + "step": 41437 + }, + { + "epoch": 1.59, + "learning_rate": 2.3462426263144397e-05, + "loss": 0.2205, + "step": 41438 + }, + { + "epoch": 1.59, + "learning_rate": 2.3461785073095666e-05, + "loss": 0.4403, + "step": 41439 + }, + { + "epoch": 1.59, + "learning_rate": 2.3461143883046934e-05, + "loss": 0.2525, + "step": 41440 + }, + { + "epoch": 1.59, + "learning_rate": 2.3460502692998206e-05, + "loss": 0.4679, + "step": 41441 + }, + { + "epoch": 1.59, + "learning_rate": 2.3459861502949477e-05, + "loss": 0.423, + "step": 41442 + }, + { + "epoch": 1.59, + "learning_rate": 2.3459220312900746e-05, + "loss": 0.325, + "step": 41443 + }, + { + "epoch": 1.59, + "learning_rate": 2.3458579122852014e-05, + "loss": 0.4315, + "step": 41444 + }, + { + "epoch": 1.59, + "learning_rate": 2.3457937932803282e-05, + "loss": 0.3361, + "step": 41445 + }, + { + "epoch": 1.59, + "learning_rate": 2.3457296742754554e-05, + "loss": 0.4048, + "step": 41446 + }, + { + "epoch": 1.59, + "learning_rate": 2.3456655552705822e-05, + "loss": 0.3917, + "step": 41447 + }, + { + "epoch": 1.59, + "learning_rate": 2.345601436265709e-05, + "loss": 0.4281, + "step": 41448 + }, + { + "epoch": 1.59, + "learning_rate": 2.3455373172608362e-05, + "loss": 0.2967, + "step": 41449 + }, + { + "epoch": 1.59, + "learning_rate": 2.3454731982559634e-05, + "loss": 0.3608, + "step": 41450 + }, + { + "epoch": 1.59, + "learning_rate": 2.3454090792510902e-05, + "loss": 0.2995, + "step": 41451 + }, + { + "epoch": 1.59, + "learning_rate": 2.345344960246217e-05, + "loss": 0.4516, + "step": 41452 + }, + { + "epoch": 1.59, + "learning_rate": 2.345280841241344e-05, + "loss": 0.3997, + "step": 41453 + }, + { + "epoch": 1.59, + "learning_rate": 2.345216722236471e-05, + "loss": 0.2495, + "step": 41454 + }, + { + "epoch": 1.59, + "learning_rate": 2.345152603231598e-05, + "loss": 0.3459, + "step": 41455 + }, + { + "epoch": 1.59, + "learning_rate": 2.3450884842267247e-05, + "loss": 0.3864, + "step": 41456 + }, + { + "epoch": 1.59, + "learning_rate": 2.345024365221852e-05, + "loss": 0.2453, + "step": 41457 + }, + { + "epoch": 1.59, + "learning_rate": 2.344960246216979e-05, + "loss": 0.357, + "step": 41458 + }, + { + "epoch": 1.59, + "learning_rate": 2.344896127212106e-05, + "loss": 0.3029, + "step": 41459 + }, + { + "epoch": 1.59, + "learning_rate": 2.3448320082072328e-05, + "loss": 0.434, + "step": 41460 + }, + { + "epoch": 1.59, + "learning_rate": 2.3447678892023596e-05, + "loss": 0.3326, + "step": 41461 + }, + { + "epoch": 1.59, + "learning_rate": 2.3447037701974868e-05, + "loss": 0.265, + "step": 41462 + }, + { + "epoch": 1.59, + "learning_rate": 2.3446396511926136e-05, + "loss": 0.2484, + "step": 41463 + }, + { + "epoch": 1.59, + "learning_rate": 2.3445755321877404e-05, + "loss": 0.3782, + "step": 41464 + }, + { + "epoch": 1.59, + "learning_rate": 2.3445114131828676e-05, + "loss": 0.311, + "step": 41465 + }, + { + "epoch": 1.59, + "learning_rate": 2.3444472941779944e-05, + "loss": 0.4303, + "step": 41466 + }, + { + "epoch": 1.59, + "learning_rate": 2.3443831751731216e-05, + "loss": 0.2438, + "step": 41467 + }, + { + "epoch": 1.59, + "learning_rate": 2.3443190561682484e-05, + "loss": 0.372, + "step": 41468 + }, + { + "epoch": 1.59, + "learning_rate": 2.3442549371633753e-05, + "loss": 0.4727, + "step": 41469 + }, + { + "epoch": 1.59, + "learning_rate": 2.3441908181585024e-05, + "loss": 0.333, + "step": 41470 + }, + { + "epoch": 1.59, + "learning_rate": 2.3441266991536293e-05, + "loss": 0.2519, + "step": 41471 + }, + { + "epoch": 1.59, + "learning_rate": 2.344062580148756e-05, + "loss": 0.3418, + "step": 41472 + }, + { + "epoch": 1.59, + "learning_rate": 2.3439984611438833e-05, + "loss": 0.3726, + "step": 41473 + }, + { + "epoch": 1.59, + "learning_rate": 2.34393434213901e-05, + "loss": 0.4014, + "step": 41474 + }, + { + "epoch": 1.59, + "learning_rate": 2.343870223134137e-05, + "loss": 0.364, + "step": 41475 + }, + { + "epoch": 1.59, + "learning_rate": 2.343806104129264e-05, + "loss": 0.4117, + "step": 41476 + }, + { + "epoch": 1.59, + "learning_rate": 2.343741985124391e-05, + "loss": 0.4045, + "step": 41477 + }, + { + "epoch": 1.59, + "learning_rate": 2.343677866119518e-05, + "loss": 0.2925, + "step": 41478 + }, + { + "epoch": 1.59, + "learning_rate": 2.343613747114645e-05, + "loss": 0.4833, + "step": 41479 + }, + { + "epoch": 1.59, + "learning_rate": 2.3435496281097718e-05, + "loss": 0.453, + "step": 41480 + }, + { + "epoch": 1.59, + "learning_rate": 2.3434855091048986e-05, + "loss": 0.3959, + "step": 41481 + }, + { + "epoch": 1.59, + "learning_rate": 2.3434213901000258e-05, + "loss": 0.353, + "step": 41482 + }, + { + "epoch": 1.59, + "learning_rate": 2.3433572710951526e-05, + "loss": 0.3771, + "step": 41483 + }, + { + "epoch": 1.59, + "learning_rate": 2.3432931520902795e-05, + "loss": 0.3294, + "step": 41484 + }, + { + "epoch": 1.59, + "learning_rate": 2.3432290330854066e-05, + "loss": 0.3906, + "step": 41485 + }, + { + "epoch": 1.6, + "learning_rate": 2.3431649140805338e-05, + "loss": 0.361, + "step": 41486 + }, + { + "epoch": 1.6, + "learning_rate": 2.3431007950756606e-05, + "loss": 0.3921, + "step": 41487 + }, + { + "epoch": 1.6, + "learning_rate": 2.3430366760707875e-05, + "loss": 0.5412, + "step": 41488 + }, + { + "epoch": 1.6, + "learning_rate": 2.3429725570659143e-05, + "loss": 0.4507, + "step": 41489 + }, + { + "epoch": 1.6, + "learning_rate": 2.3429084380610415e-05, + "loss": 0.4734, + "step": 41490 + }, + { + "epoch": 1.6, + "learning_rate": 2.3428443190561683e-05, + "loss": 0.2187, + "step": 41491 + }, + { + "epoch": 1.6, + "learning_rate": 2.342780200051295e-05, + "loss": 0.2577, + "step": 41492 + }, + { + "epoch": 1.6, + "learning_rate": 2.3427160810464223e-05, + "loss": 0.3558, + "step": 41493 + }, + { + "epoch": 1.6, + "learning_rate": 2.3426519620415495e-05, + "loss": 0.3224, + "step": 41494 + }, + { + "epoch": 1.6, + "learning_rate": 2.3425878430366763e-05, + "loss": 0.2754, + "step": 41495 + }, + { + "epoch": 1.6, + "learning_rate": 2.342523724031803e-05, + "loss": 0.4273, + "step": 41496 + }, + { + "epoch": 1.6, + "learning_rate": 2.34245960502693e-05, + "loss": 0.4509, + "step": 41497 + }, + { + "epoch": 1.6, + "learning_rate": 2.342395486022057e-05, + "loss": 0.3414, + "step": 41498 + }, + { + "epoch": 1.6, + "learning_rate": 2.342331367017184e-05, + "loss": 0.3513, + "step": 41499 + }, + { + "epoch": 1.6, + "learning_rate": 2.3422672480123108e-05, + "loss": 0.297, + "step": 41500 + }, + { + "epoch": 1.6, + "learning_rate": 2.342203129007438e-05, + "loss": 0.4089, + "step": 41501 + }, + { + "epoch": 1.6, + "learning_rate": 2.3421390100025648e-05, + "loss": 0.4085, + "step": 41502 + }, + { + "epoch": 1.6, + "learning_rate": 2.342074890997692e-05, + "loss": 0.3605, + "step": 41503 + }, + { + "epoch": 1.6, + "learning_rate": 2.3420107719928188e-05, + "loss": 0.4914, + "step": 41504 + }, + { + "epoch": 1.6, + "learning_rate": 2.3419466529879457e-05, + "loss": 0.2861, + "step": 41505 + }, + { + "epoch": 1.6, + "learning_rate": 2.3418825339830728e-05, + "loss": 0.3799, + "step": 41506 + }, + { + "epoch": 1.6, + "learning_rate": 2.3418184149781997e-05, + "loss": 0.3044, + "step": 41507 + }, + { + "epoch": 1.6, + "learning_rate": 2.3417542959733265e-05, + "loss": 0.3322, + "step": 41508 + }, + { + "epoch": 1.6, + "learning_rate": 2.3416901769684537e-05, + "loss": 0.3897, + "step": 41509 + }, + { + "epoch": 1.6, + "learning_rate": 2.3416260579635805e-05, + "loss": 0.3197, + "step": 41510 + }, + { + "epoch": 1.6, + "learning_rate": 2.3415619389587073e-05, + "loss": 0.2999, + "step": 41511 + }, + { + "epoch": 1.6, + "learning_rate": 2.3414978199538345e-05, + "loss": 0.488, + "step": 41512 + }, + { + "epoch": 1.6, + "learning_rate": 2.3414337009489613e-05, + "loss": 0.2594, + "step": 41513 + }, + { + "epoch": 1.6, + "learning_rate": 2.3413695819440885e-05, + "loss": 0.401, + "step": 41514 + }, + { + "epoch": 1.6, + "learning_rate": 2.3413054629392153e-05, + "loss": 0.4767, + "step": 41515 + }, + { + "epoch": 1.6, + "learning_rate": 2.3412413439343422e-05, + "loss": 0.305, + "step": 41516 + }, + { + "epoch": 1.6, + "learning_rate": 2.341177224929469e-05, + "loss": 0.3352, + "step": 41517 + }, + { + "epoch": 1.6, + "learning_rate": 2.3411131059245962e-05, + "loss": 0.3165, + "step": 41518 + }, + { + "epoch": 1.6, + "learning_rate": 2.341048986919723e-05, + "loss": 0.2862, + "step": 41519 + }, + { + "epoch": 1.6, + "learning_rate": 2.34098486791485e-05, + "loss": 0.2732, + "step": 41520 + }, + { + "epoch": 1.6, + "learning_rate": 2.340920748909977e-05, + "loss": 0.2151, + "step": 41521 + }, + { + "epoch": 1.6, + "learning_rate": 2.3408566299051042e-05, + "loss": 0.3669, + "step": 41522 + }, + { + "epoch": 1.6, + "learning_rate": 2.340792510900231e-05, + "loss": 0.5788, + "step": 41523 + }, + { + "epoch": 1.6, + "learning_rate": 2.340728391895358e-05, + "loss": 0.3079, + "step": 41524 + }, + { + "epoch": 1.6, + "learning_rate": 2.3406642728904847e-05, + "loss": 0.2969, + "step": 41525 + }, + { + "epoch": 1.6, + "learning_rate": 2.340600153885612e-05, + "loss": 0.3882, + "step": 41526 + }, + { + "epoch": 1.6, + "learning_rate": 2.3405360348807387e-05, + "loss": 0.3403, + "step": 41527 + }, + { + "epoch": 1.6, + "learning_rate": 2.3404719158758655e-05, + "loss": 0.334, + "step": 41528 + }, + { + "epoch": 1.6, + "learning_rate": 2.3404077968709927e-05, + "loss": 0.3385, + "step": 41529 + }, + { + "epoch": 1.6, + "learning_rate": 2.34034367786612e-05, + "loss": 0.3359, + "step": 41530 + }, + { + "epoch": 1.6, + "learning_rate": 2.3402795588612467e-05, + "loss": 0.4099, + "step": 41531 + }, + { + "epoch": 1.6, + "learning_rate": 2.3402154398563735e-05, + "loss": 0.3683, + "step": 41532 + }, + { + "epoch": 1.6, + "learning_rate": 2.3401513208515004e-05, + "loss": 0.6036, + "step": 41533 + }, + { + "epoch": 1.6, + "learning_rate": 2.3400872018466275e-05, + "loss": 0.2838, + "step": 41534 + }, + { + "epoch": 1.6, + "learning_rate": 2.3400230828417544e-05, + "loss": 0.2062, + "step": 41535 + }, + { + "epoch": 1.6, + "learning_rate": 2.3399589638368812e-05, + "loss": 0.3627, + "step": 41536 + }, + { + "epoch": 1.6, + "learning_rate": 2.3398948448320084e-05, + "loss": 0.5394, + "step": 41537 + }, + { + "epoch": 1.6, + "learning_rate": 2.3398307258271352e-05, + "loss": 0.2859, + "step": 41538 + }, + { + "epoch": 1.6, + "learning_rate": 2.3397666068222624e-05, + "loss": 0.298, + "step": 41539 + }, + { + "epoch": 1.6, + "learning_rate": 2.3397024878173892e-05, + "loss": 0.4233, + "step": 41540 + }, + { + "epoch": 1.6, + "learning_rate": 2.339638368812516e-05, + "loss": 0.3799, + "step": 41541 + }, + { + "epoch": 1.6, + "learning_rate": 2.3395742498076432e-05, + "loss": 0.3332, + "step": 41542 + }, + { + "epoch": 1.6, + "learning_rate": 2.33951013080277e-05, + "loss": 0.5814, + "step": 41543 + }, + { + "epoch": 1.6, + "learning_rate": 2.339446011797897e-05, + "loss": 0.3466, + "step": 41544 + }, + { + "epoch": 1.6, + "learning_rate": 2.339381892793024e-05, + "loss": 0.2881, + "step": 41545 + }, + { + "epoch": 1.6, + "learning_rate": 2.339317773788151e-05, + "loss": 0.297, + "step": 41546 + }, + { + "epoch": 1.6, + "learning_rate": 2.3392536547832777e-05, + "loss": 0.3814, + "step": 41547 + }, + { + "epoch": 1.6, + "learning_rate": 2.339189535778405e-05, + "loss": 0.2999, + "step": 41548 + }, + { + "epoch": 1.6, + "learning_rate": 2.3391254167735317e-05, + "loss": 0.3692, + "step": 41549 + }, + { + "epoch": 1.6, + "learning_rate": 2.339061297768659e-05, + "loss": 0.3224, + "step": 41550 + }, + { + "epoch": 1.6, + "learning_rate": 2.3389971787637857e-05, + "loss": 0.286, + "step": 41551 + }, + { + "epoch": 1.6, + "learning_rate": 2.3389330597589126e-05, + "loss": 0.3046, + "step": 41552 + }, + { + "epoch": 1.6, + "learning_rate": 2.3388689407540397e-05, + "loss": 0.2243, + "step": 41553 + }, + { + "epoch": 1.6, + "learning_rate": 2.3388048217491666e-05, + "loss": 0.2978, + "step": 41554 + }, + { + "epoch": 1.6, + "learning_rate": 2.3387407027442934e-05, + "loss": 0.2617, + "step": 41555 + }, + { + "epoch": 1.6, + "learning_rate": 2.3386765837394206e-05, + "loss": 0.3441, + "step": 41556 + }, + { + "epoch": 1.6, + "learning_rate": 2.3386124647345474e-05, + "loss": 0.2712, + "step": 41557 + }, + { + "epoch": 1.6, + "learning_rate": 2.3385483457296746e-05, + "loss": 0.3707, + "step": 41558 + }, + { + "epoch": 1.6, + "learning_rate": 2.3384842267248014e-05, + "loss": 0.3421, + "step": 41559 + }, + { + "epoch": 1.6, + "learning_rate": 2.3384201077199282e-05, + "loss": 0.287, + "step": 41560 + }, + { + "epoch": 1.6, + "learning_rate": 2.338355988715055e-05, + "loss": 0.3643, + "step": 41561 + }, + { + "epoch": 1.6, + "learning_rate": 2.3382918697101822e-05, + "loss": 0.432, + "step": 41562 + }, + { + "epoch": 1.6, + "learning_rate": 2.338227750705309e-05, + "loss": 0.3191, + "step": 41563 + }, + { + "epoch": 1.6, + "learning_rate": 2.338163631700436e-05, + "loss": 0.2665, + "step": 41564 + }, + { + "epoch": 1.6, + "learning_rate": 2.338099512695563e-05, + "loss": 0.3527, + "step": 41565 + }, + { + "epoch": 1.6, + "learning_rate": 2.3380353936906902e-05, + "loss": 0.3552, + "step": 41566 + }, + { + "epoch": 1.6, + "learning_rate": 2.337971274685817e-05, + "loss": 0.3434, + "step": 41567 + }, + { + "epoch": 1.6, + "learning_rate": 2.337907155680944e-05, + "loss": 0.435, + "step": 41568 + }, + { + "epoch": 1.6, + "learning_rate": 2.3378430366760707e-05, + "loss": 0.307, + "step": 41569 + }, + { + "epoch": 1.6, + "learning_rate": 2.337778917671198e-05, + "loss": 0.3346, + "step": 41570 + }, + { + "epoch": 1.6, + "learning_rate": 2.3377147986663248e-05, + "loss": 0.2878, + "step": 41571 + }, + { + "epoch": 1.6, + "learning_rate": 2.3376506796614516e-05, + "loss": 0.4858, + "step": 41572 + }, + { + "epoch": 1.6, + "learning_rate": 2.3375865606565788e-05, + "loss": 0.2528, + "step": 41573 + }, + { + "epoch": 1.6, + "learning_rate": 2.337522441651706e-05, + "loss": 0.4904, + "step": 41574 + }, + { + "epoch": 1.6, + "learning_rate": 2.3374583226468328e-05, + "loss": 0.3567, + "step": 41575 + }, + { + "epoch": 1.6, + "learning_rate": 2.3373942036419596e-05, + "loss": 0.4974, + "step": 41576 + }, + { + "epoch": 1.6, + "learning_rate": 2.3373300846370864e-05, + "loss": 0.2917, + "step": 41577 + }, + { + "epoch": 1.6, + "learning_rate": 2.3372659656322136e-05, + "loss": 0.496, + "step": 41578 + }, + { + "epoch": 1.6, + "learning_rate": 2.3372018466273404e-05, + "loss": 0.3913, + "step": 41579 + }, + { + "epoch": 1.6, + "learning_rate": 2.3371377276224673e-05, + "loss": 0.3329, + "step": 41580 + }, + { + "epoch": 1.6, + "learning_rate": 2.3370736086175944e-05, + "loss": 0.378, + "step": 41581 + }, + { + "epoch": 1.6, + "learning_rate": 2.3370094896127213e-05, + "loss": 0.3894, + "step": 41582 + }, + { + "epoch": 1.6, + "learning_rate": 2.3369453706078484e-05, + "loss": 0.3999, + "step": 41583 + }, + { + "epoch": 1.6, + "learning_rate": 2.3368812516029753e-05, + "loss": 0.4388, + "step": 41584 + }, + { + "epoch": 1.6, + "learning_rate": 2.336817132598102e-05, + "loss": 0.3982, + "step": 41585 + }, + { + "epoch": 1.6, + "learning_rate": 2.3367530135932293e-05, + "loss": 0.2442, + "step": 41586 + }, + { + "epoch": 1.6, + "learning_rate": 2.336688894588356e-05, + "loss": 0.3149, + "step": 41587 + }, + { + "epoch": 1.6, + "learning_rate": 2.336624775583483e-05, + "loss": 0.3059, + "step": 41588 + }, + { + "epoch": 1.6, + "learning_rate": 2.33656065657861e-05, + "loss": 0.3438, + "step": 41589 + }, + { + "epoch": 1.6, + "learning_rate": 2.336496537573737e-05, + "loss": 0.3002, + "step": 41590 + }, + { + "epoch": 1.6, + "learning_rate": 2.3364324185688638e-05, + "loss": 0.3818, + "step": 41591 + }, + { + "epoch": 1.6, + "learning_rate": 2.336368299563991e-05, + "loss": 0.3187, + "step": 41592 + }, + { + "epoch": 1.6, + "learning_rate": 2.3363041805591178e-05, + "loss": 0.2861, + "step": 41593 + }, + { + "epoch": 1.6, + "learning_rate": 2.336240061554245e-05, + "loss": 0.4187, + "step": 41594 + }, + { + "epoch": 1.6, + "learning_rate": 2.3361759425493718e-05, + "loss": 0.2119, + "step": 41595 + }, + { + "epoch": 1.6, + "learning_rate": 2.3361118235444986e-05, + "loss": 0.4727, + "step": 41596 + }, + { + "epoch": 1.6, + "learning_rate": 2.3360477045396258e-05, + "loss": 0.2864, + "step": 41597 + }, + { + "epoch": 1.6, + "learning_rate": 2.3359835855347526e-05, + "loss": 0.326, + "step": 41598 + }, + { + "epoch": 1.6, + "learning_rate": 2.3359194665298795e-05, + "loss": 0.4163, + "step": 41599 + }, + { + "epoch": 1.6, + "learning_rate": 2.3358553475250063e-05, + "loss": 0.3115, + "step": 41600 + }, + { + "epoch": 1.6, + "learning_rate": 2.3357912285201335e-05, + "loss": 0.2861, + "step": 41601 + }, + { + "epoch": 1.6, + "learning_rate": 2.3357271095152606e-05, + "loss": 0.313, + "step": 41602 + }, + { + "epoch": 1.6, + "learning_rate": 2.3356629905103875e-05, + "loss": 0.2212, + "step": 41603 + }, + { + "epoch": 1.6, + "learning_rate": 2.3355988715055143e-05, + "loss": 0.3984, + "step": 41604 + }, + { + "epoch": 1.6, + "learning_rate": 2.335534752500641e-05, + "loss": 0.2912, + "step": 41605 + }, + { + "epoch": 1.6, + "learning_rate": 2.3354706334957683e-05, + "loss": 0.2595, + "step": 41606 + }, + { + "epoch": 1.6, + "learning_rate": 2.335406514490895e-05, + "loss": 0.3861, + "step": 41607 + }, + { + "epoch": 1.6, + "learning_rate": 2.335342395486022e-05, + "loss": 0.3516, + "step": 41608 + }, + { + "epoch": 1.6, + "learning_rate": 2.335278276481149e-05, + "loss": 0.3107, + "step": 41609 + }, + { + "epoch": 1.6, + "learning_rate": 2.3352141574762763e-05, + "loss": 0.3671, + "step": 41610 + }, + { + "epoch": 1.6, + "learning_rate": 2.335150038471403e-05, + "loss": 0.3217, + "step": 41611 + }, + { + "epoch": 1.6, + "learning_rate": 2.33508591946653e-05, + "loss": 0.3792, + "step": 41612 + }, + { + "epoch": 1.6, + "learning_rate": 2.3350218004616568e-05, + "loss": 0.3736, + "step": 41613 + }, + { + "epoch": 1.6, + "learning_rate": 2.334957681456784e-05, + "loss": 0.3832, + "step": 41614 + }, + { + "epoch": 1.6, + "learning_rate": 2.3348935624519108e-05, + "loss": 0.3211, + "step": 41615 + }, + { + "epoch": 1.6, + "learning_rate": 2.3348294434470377e-05, + "loss": 0.3747, + "step": 41616 + }, + { + "epoch": 1.6, + "learning_rate": 2.3347653244421648e-05, + "loss": 0.2491, + "step": 41617 + }, + { + "epoch": 1.6, + "learning_rate": 2.3347012054372917e-05, + "loss": 0.3176, + "step": 41618 + }, + { + "epoch": 1.6, + "learning_rate": 2.3346370864324188e-05, + "loss": 0.3526, + "step": 41619 + }, + { + "epoch": 1.6, + "learning_rate": 2.3345729674275457e-05, + "loss": 0.2847, + "step": 41620 + }, + { + "epoch": 1.6, + "learning_rate": 2.3345088484226725e-05, + "loss": 0.2928, + "step": 41621 + }, + { + "epoch": 1.6, + "learning_rate": 2.3344447294177997e-05, + "loss": 0.2848, + "step": 41622 + }, + { + "epoch": 1.6, + "learning_rate": 2.3343806104129265e-05, + "loss": 0.3744, + "step": 41623 + }, + { + "epoch": 1.6, + "learning_rate": 2.3343164914080533e-05, + "loss": 0.3965, + "step": 41624 + }, + { + "epoch": 1.6, + "learning_rate": 2.3342523724031805e-05, + "loss": 0.3263, + "step": 41625 + }, + { + "epoch": 1.6, + "learning_rate": 2.3341882533983073e-05, + "loss": 0.2584, + "step": 41626 + }, + { + "epoch": 1.6, + "learning_rate": 2.334124134393434e-05, + "loss": 0.2339, + "step": 41627 + }, + { + "epoch": 1.6, + "learning_rate": 2.3340600153885613e-05, + "loss": 0.4512, + "step": 41628 + }, + { + "epoch": 1.6, + "learning_rate": 2.3339958963836882e-05, + "loss": 0.2486, + "step": 41629 + }, + { + "epoch": 1.6, + "learning_rate": 2.3339317773788153e-05, + "loss": 0.4086, + "step": 41630 + }, + { + "epoch": 1.6, + "learning_rate": 2.3338676583739422e-05, + "loss": 0.1973, + "step": 41631 + }, + { + "epoch": 1.6, + "learning_rate": 2.333803539369069e-05, + "loss": 0.3169, + "step": 41632 + }, + { + "epoch": 1.6, + "learning_rate": 2.3337394203641962e-05, + "loss": 0.2895, + "step": 41633 + }, + { + "epoch": 1.6, + "learning_rate": 2.333675301359323e-05, + "loss": 0.4294, + "step": 41634 + }, + { + "epoch": 1.6, + "learning_rate": 2.33361118235445e-05, + "loss": 0.3419, + "step": 41635 + }, + { + "epoch": 1.6, + "learning_rate": 2.3335470633495767e-05, + "loss": 0.2616, + "step": 41636 + }, + { + "epoch": 1.6, + "learning_rate": 2.333482944344704e-05, + "loss": 0.4324, + "step": 41637 + }, + { + "epoch": 1.6, + "learning_rate": 2.333418825339831e-05, + "loss": 0.3213, + "step": 41638 + }, + { + "epoch": 1.6, + "learning_rate": 2.333354706334958e-05, + "loss": 0.3411, + "step": 41639 + }, + { + "epoch": 1.6, + "learning_rate": 2.3332905873300847e-05, + "loss": 0.4213, + "step": 41640 + }, + { + "epoch": 1.6, + "learning_rate": 2.3332264683252115e-05, + "loss": 0.2809, + "step": 41641 + }, + { + "epoch": 1.6, + "learning_rate": 2.3331623493203387e-05, + "loss": 0.3726, + "step": 41642 + }, + { + "epoch": 1.6, + "learning_rate": 2.3330982303154655e-05, + "loss": 0.468, + "step": 41643 + }, + { + "epoch": 1.6, + "learning_rate": 2.3330341113105924e-05, + "loss": 0.2765, + "step": 41644 + }, + { + "epoch": 1.6, + "learning_rate": 2.3329699923057195e-05, + "loss": 0.2521, + "step": 41645 + }, + { + "epoch": 1.6, + "learning_rate": 2.3329058733008467e-05, + "loss": 0.3432, + "step": 41646 + }, + { + "epoch": 1.6, + "learning_rate": 2.3328417542959735e-05, + "loss": 0.3145, + "step": 41647 + }, + { + "epoch": 1.6, + "learning_rate": 2.3327776352911004e-05, + "loss": 0.2065, + "step": 41648 + }, + { + "epoch": 1.6, + "learning_rate": 2.3327135162862272e-05, + "loss": 0.2631, + "step": 41649 + }, + { + "epoch": 1.6, + "learning_rate": 2.3326493972813544e-05, + "loss": 0.2634, + "step": 41650 + }, + { + "epoch": 1.6, + "learning_rate": 2.3325852782764812e-05, + "loss": 0.2647, + "step": 41651 + }, + { + "epoch": 1.6, + "learning_rate": 2.332521159271608e-05, + "loss": 0.445, + "step": 41652 + }, + { + "epoch": 1.6, + "learning_rate": 2.3324570402667352e-05, + "loss": 0.2902, + "step": 41653 + }, + { + "epoch": 1.6, + "learning_rate": 2.3323929212618624e-05, + "loss": 0.2465, + "step": 41654 + }, + { + "epoch": 1.6, + "learning_rate": 2.3323288022569892e-05, + "loss": 0.4249, + "step": 41655 + }, + { + "epoch": 1.6, + "learning_rate": 2.332264683252116e-05, + "loss": 0.3397, + "step": 41656 + }, + { + "epoch": 1.6, + "learning_rate": 2.332200564247243e-05, + "loss": 0.4754, + "step": 41657 + }, + { + "epoch": 1.6, + "learning_rate": 2.33213644524237e-05, + "loss": 0.31, + "step": 41658 + }, + { + "epoch": 1.6, + "learning_rate": 2.332072326237497e-05, + "loss": 0.473, + "step": 41659 + }, + { + "epoch": 1.6, + "learning_rate": 2.3320082072326237e-05, + "loss": 0.3629, + "step": 41660 + }, + { + "epoch": 1.6, + "learning_rate": 2.331944088227751e-05, + "loss": 0.283, + "step": 41661 + }, + { + "epoch": 1.6, + "learning_rate": 2.3318799692228777e-05, + "loss": 0.2833, + "step": 41662 + }, + { + "epoch": 1.6, + "learning_rate": 2.331815850218005e-05, + "loss": 0.449, + "step": 41663 + }, + { + "epoch": 1.6, + "learning_rate": 2.3317517312131317e-05, + "loss": 0.309, + "step": 41664 + }, + { + "epoch": 1.6, + "learning_rate": 2.3316876122082586e-05, + "loss": 0.1824, + "step": 41665 + }, + { + "epoch": 1.6, + "learning_rate": 2.3316234932033857e-05, + "loss": 0.4113, + "step": 41666 + }, + { + "epoch": 1.6, + "learning_rate": 2.3315593741985126e-05, + "loss": 0.3297, + "step": 41667 + }, + { + "epoch": 1.6, + "learning_rate": 2.3314952551936394e-05, + "loss": 0.4349, + "step": 41668 + }, + { + "epoch": 1.6, + "learning_rate": 2.3314311361887666e-05, + "loss": 0.3709, + "step": 41669 + }, + { + "epoch": 1.6, + "learning_rate": 2.3313670171838934e-05, + "loss": 0.2827, + "step": 41670 + }, + { + "epoch": 1.6, + "learning_rate": 2.3313028981790202e-05, + "loss": 0.3509, + "step": 41671 + }, + { + "epoch": 1.6, + "learning_rate": 2.3312387791741474e-05, + "loss": 0.293, + "step": 41672 + }, + { + "epoch": 1.6, + "learning_rate": 2.3311746601692742e-05, + "loss": 0.3139, + "step": 41673 + }, + { + "epoch": 1.6, + "learning_rate": 2.3311105411644014e-05, + "loss": 0.482, + "step": 41674 + }, + { + "epoch": 1.6, + "learning_rate": 2.3310464221595282e-05, + "loss": 0.3403, + "step": 41675 + }, + { + "epoch": 1.6, + "learning_rate": 2.330982303154655e-05, + "loss": 0.3356, + "step": 41676 + }, + { + "epoch": 1.6, + "learning_rate": 2.3309181841497822e-05, + "loss": 0.4216, + "step": 41677 + }, + { + "epoch": 1.6, + "learning_rate": 2.330854065144909e-05, + "loss": 0.3568, + "step": 41678 + }, + { + "epoch": 1.6, + "learning_rate": 2.330789946140036e-05, + "loss": 0.3374, + "step": 41679 + }, + { + "epoch": 1.6, + "learning_rate": 2.3307258271351627e-05, + "loss": 0.2544, + "step": 41680 + }, + { + "epoch": 1.6, + "learning_rate": 2.33066170813029e-05, + "loss": 0.4221, + "step": 41681 + }, + { + "epoch": 1.6, + "learning_rate": 2.330597589125417e-05, + "loss": 0.2741, + "step": 41682 + }, + { + "epoch": 1.6, + "learning_rate": 2.330533470120544e-05, + "loss": 0.2908, + "step": 41683 + }, + { + "epoch": 1.6, + "learning_rate": 2.3304693511156708e-05, + "loss": 0.243, + "step": 41684 + }, + { + "epoch": 1.6, + "learning_rate": 2.3304052321107976e-05, + "loss": 0.4233, + "step": 41685 + }, + { + "epoch": 1.6, + "learning_rate": 2.3303411131059248e-05, + "loss": 0.2696, + "step": 41686 + }, + { + "epoch": 1.6, + "learning_rate": 2.3302769941010516e-05, + "loss": 0.3655, + "step": 41687 + }, + { + "epoch": 1.6, + "learning_rate": 2.3302128750961784e-05, + "loss": 0.4134, + "step": 41688 + }, + { + "epoch": 1.6, + "learning_rate": 2.3301487560913056e-05, + "loss": 0.2974, + "step": 41689 + }, + { + "epoch": 1.6, + "learning_rate": 2.3300846370864328e-05, + "loss": 0.3345, + "step": 41690 + }, + { + "epoch": 1.6, + "learning_rate": 2.3300205180815596e-05, + "loss": 0.3116, + "step": 41691 + }, + { + "epoch": 1.6, + "learning_rate": 2.3299563990766864e-05, + "loss": 0.3761, + "step": 41692 + }, + { + "epoch": 1.6, + "learning_rate": 2.3298922800718133e-05, + "loss": 0.2854, + "step": 41693 + }, + { + "epoch": 1.6, + "learning_rate": 2.3298281610669404e-05, + "loss": 0.2876, + "step": 41694 + }, + { + "epoch": 1.6, + "learning_rate": 2.3297640420620673e-05, + "loss": 0.2822, + "step": 41695 + }, + { + "epoch": 1.6, + "learning_rate": 2.329699923057194e-05, + "loss": 0.2658, + "step": 41696 + }, + { + "epoch": 1.6, + "learning_rate": 2.3296358040523213e-05, + "loss": 0.2835, + "step": 41697 + }, + { + "epoch": 1.6, + "learning_rate": 2.329571685047448e-05, + "loss": 0.2434, + "step": 41698 + }, + { + "epoch": 1.6, + "learning_rate": 2.3295075660425753e-05, + "loss": 0.364, + "step": 41699 + }, + { + "epoch": 1.6, + "learning_rate": 2.329443447037702e-05, + "loss": 0.2511, + "step": 41700 + }, + { + "epoch": 1.6, + "learning_rate": 2.329379328032829e-05, + "loss": 0.3711, + "step": 41701 + }, + { + "epoch": 1.6, + "learning_rate": 2.329315209027956e-05, + "loss": 0.3342, + "step": 41702 + }, + { + "epoch": 1.6, + "learning_rate": 2.329251090023083e-05, + "loss": 0.3557, + "step": 41703 + }, + { + "epoch": 1.6, + "learning_rate": 2.3291869710182098e-05, + "loss": 0.4133, + "step": 41704 + }, + { + "epoch": 1.6, + "learning_rate": 2.329122852013337e-05, + "loss": 0.4452, + "step": 41705 + }, + { + "epoch": 1.6, + "learning_rate": 2.3290587330084638e-05, + "loss": 0.3123, + "step": 41706 + }, + { + "epoch": 1.6, + "learning_rate": 2.3289946140035906e-05, + "loss": 0.3183, + "step": 41707 + }, + { + "epoch": 1.6, + "learning_rate": 2.3289304949987178e-05, + "loss": 0.2612, + "step": 41708 + }, + { + "epoch": 1.6, + "learning_rate": 2.3288663759938446e-05, + "loss": 0.4123, + "step": 41709 + }, + { + "epoch": 1.6, + "learning_rate": 2.3288022569889718e-05, + "loss": 0.465, + "step": 41710 + }, + { + "epoch": 1.6, + "learning_rate": 2.3287381379840986e-05, + "loss": 0.2349, + "step": 41711 + }, + { + "epoch": 1.6, + "learning_rate": 2.3286740189792255e-05, + "loss": 0.4307, + "step": 41712 + }, + { + "epoch": 1.6, + "learning_rate": 2.3286098999743526e-05, + "loss": 0.2916, + "step": 41713 + }, + { + "epoch": 1.6, + "learning_rate": 2.3285457809694795e-05, + "loss": 0.4794, + "step": 41714 + }, + { + "epoch": 1.6, + "learning_rate": 2.3284816619646063e-05, + "loss": 0.4343, + "step": 41715 + }, + { + "epoch": 1.6, + "learning_rate": 2.328417542959733e-05, + "loss": 0.4226, + "step": 41716 + }, + { + "epoch": 1.6, + "learning_rate": 2.3283534239548603e-05, + "loss": 0.4493, + "step": 41717 + }, + { + "epoch": 1.6, + "learning_rate": 2.3282893049499875e-05, + "loss": 0.3611, + "step": 41718 + }, + { + "epoch": 1.6, + "learning_rate": 2.3282251859451143e-05, + "loss": 0.4073, + "step": 41719 + }, + { + "epoch": 1.6, + "learning_rate": 2.328161066940241e-05, + "loss": 0.2967, + "step": 41720 + }, + { + "epoch": 1.6, + "learning_rate": 2.328096947935368e-05, + "loss": 0.331, + "step": 41721 + }, + { + "epoch": 1.6, + "learning_rate": 2.328032828930495e-05, + "loss": 0.3385, + "step": 41722 + }, + { + "epoch": 1.6, + "learning_rate": 2.327968709925622e-05, + "loss": 0.5649, + "step": 41723 + }, + { + "epoch": 1.6, + "learning_rate": 2.3279045909207488e-05, + "loss": 0.4036, + "step": 41724 + }, + { + "epoch": 1.6, + "learning_rate": 2.327840471915876e-05, + "loss": 0.3242, + "step": 41725 + }, + { + "epoch": 1.6, + "learning_rate": 2.327776352911003e-05, + "loss": 0.3781, + "step": 41726 + }, + { + "epoch": 1.6, + "learning_rate": 2.32771223390613e-05, + "loss": 0.2574, + "step": 41727 + }, + { + "epoch": 1.6, + "learning_rate": 2.3276481149012568e-05, + "loss": 0.3053, + "step": 41728 + }, + { + "epoch": 1.6, + "learning_rate": 2.3275839958963837e-05, + "loss": 0.4341, + "step": 41729 + }, + { + "epoch": 1.6, + "learning_rate": 2.3275198768915108e-05, + "loss": 0.2249, + "step": 41730 + }, + { + "epoch": 1.6, + "learning_rate": 2.3274557578866377e-05, + "loss": 0.2674, + "step": 41731 + }, + { + "epoch": 1.6, + "learning_rate": 2.3273916388817645e-05, + "loss": 0.4462, + "step": 41732 + }, + { + "epoch": 1.6, + "learning_rate": 2.3273275198768917e-05, + "loss": 0.3521, + "step": 41733 + }, + { + "epoch": 1.6, + "learning_rate": 2.3272634008720185e-05, + "loss": 0.3123, + "step": 41734 + }, + { + "epoch": 1.6, + "learning_rate": 2.3271992818671457e-05, + "loss": 0.3759, + "step": 41735 + }, + { + "epoch": 1.6, + "learning_rate": 2.3271351628622725e-05, + "loss": 0.4653, + "step": 41736 + }, + { + "epoch": 1.6, + "learning_rate": 2.3270710438573993e-05, + "loss": 0.4731, + "step": 41737 + }, + { + "epoch": 1.6, + "learning_rate": 2.3270069248525265e-05, + "loss": 0.4574, + "step": 41738 + }, + { + "epoch": 1.6, + "learning_rate": 2.3269428058476533e-05, + "loss": 0.2554, + "step": 41739 + }, + { + "epoch": 1.6, + "learning_rate": 2.3268786868427802e-05, + "loss": 0.3026, + "step": 41740 + }, + { + "epoch": 1.6, + "learning_rate": 2.3268145678379073e-05, + "loss": 0.3344, + "step": 41741 + }, + { + "epoch": 1.6, + "learning_rate": 2.3267504488330342e-05, + "loss": 0.2903, + "step": 41742 + }, + { + "epoch": 1.6, + "learning_rate": 2.3266863298281613e-05, + "loss": 0.2938, + "step": 41743 + }, + { + "epoch": 1.6, + "learning_rate": 2.3266222108232882e-05, + "loss": 0.3293, + "step": 41744 + }, + { + "epoch": 1.6, + "learning_rate": 2.326558091818415e-05, + "loss": 0.2641, + "step": 41745 + }, + { + "epoch": 1.6, + "learning_rate": 2.3264939728135422e-05, + "loss": 0.2987, + "step": 41746 + }, + { + "epoch": 1.61, + "learning_rate": 2.326429853808669e-05, + "loss": 0.2515, + "step": 41747 + }, + { + "epoch": 1.61, + "learning_rate": 2.326365734803796e-05, + "loss": 0.2778, + "step": 41748 + }, + { + "epoch": 1.61, + "learning_rate": 2.326301615798923e-05, + "loss": 0.3907, + "step": 41749 + }, + { + "epoch": 1.61, + "learning_rate": 2.32623749679405e-05, + "loss": 0.346, + "step": 41750 + }, + { + "epoch": 1.61, + "learning_rate": 2.3261733777891767e-05, + "loss": 0.4787, + "step": 41751 + }, + { + "epoch": 1.61, + "learning_rate": 2.326109258784304e-05, + "loss": 0.3799, + "step": 41752 + }, + { + "epoch": 1.61, + "learning_rate": 2.3260451397794307e-05, + "loss": 0.4357, + "step": 41753 + }, + { + "epoch": 1.61, + "learning_rate": 2.325981020774558e-05, + "loss": 0.3471, + "step": 41754 + }, + { + "epoch": 1.61, + "learning_rate": 2.3259169017696847e-05, + "loss": 0.2337, + "step": 41755 + }, + { + "epoch": 1.61, + "learning_rate": 2.3258527827648115e-05, + "loss": 0.4114, + "step": 41756 + }, + { + "epoch": 1.61, + "learning_rate": 2.3257886637599387e-05, + "loss": 0.4092, + "step": 41757 + }, + { + "epoch": 1.61, + "learning_rate": 2.3257245447550655e-05, + "loss": 0.4189, + "step": 41758 + }, + { + "epoch": 1.61, + "learning_rate": 2.3256604257501924e-05, + "loss": 0.339, + "step": 41759 + }, + { + "epoch": 1.61, + "learning_rate": 2.3255963067453192e-05, + "loss": 0.3036, + "step": 41760 + }, + { + "epoch": 1.61, + "learning_rate": 2.3255321877404464e-05, + "loss": 0.5271, + "step": 41761 + }, + { + "epoch": 1.61, + "learning_rate": 2.3254680687355735e-05, + "loss": 0.3336, + "step": 41762 + }, + { + "epoch": 1.61, + "learning_rate": 2.3254039497307004e-05, + "loss": 0.3015, + "step": 41763 + }, + { + "epoch": 1.61, + "learning_rate": 2.3253398307258272e-05, + "loss": 0.3677, + "step": 41764 + }, + { + "epoch": 1.61, + "learning_rate": 2.325275711720954e-05, + "loss": 0.2925, + "step": 41765 + }, + { + "epoch": 1.61, + "learning_rate": 2.3252115927160812e-05, + "loss": 0.3004, + "step": 41766 + }, + { + "epoch": 1.61, + "learning_rate": 2.325147473711208e-05, + "loss": 0.3119, + "step": 41767 + }, + { + "epoch": 1.61, + "learning_rate": 2.325083354706335e-05, + "loss": 0.3846, + "step": 41768 + }, + { + "epoch": 1.61, + "learning_rate": 2.325019235701462e-05, + "loss": 0.3336, + "step": 41769 + }, + { + "epoch": 1.61, + "learning_rate": 2.3249551166965892e-05, + "loss": 0.2298, + "step": 41770 + }, + { + "epoch": 1.61, + "learning_rate": 2.324890997691716e-05, + "loss": 0.3747, + "step": 41771 + }, + { + "epoch": 1.61, + "learning_rate": 2.324826878686843e-05, + "loss": 0.4631, + "step": 41772 + }, + { + "epoch": 1.61, + "learning_rate": 2.3247627596819697e-05, + "loss": 0.2565, + "step": 41773 + }, + { + "epoch": 1.61, + "learning_rate": 2.324698640677097e-05, + "loss": 0.3634, + "step": 41774 + }, + { + "epoch": 1.61, + "learning_rate": 2.3246345216722237e-05, + "loss": 0.289, + "step": 41775 + }, + { + "epoch": 1.61, + "learning_rate": 2.3245704026673506e-05, + "loss": 0.4408, + "step": 41776 + }, + { + "epoch": 1.61, + "learning_rate": 2.3245062836624777e-05, + "loss": 0.469, + "step": 41777 + }, + { + "epoch": 1.61, + "learning_rate": 2.3244421646576046e-05, + "loss": 0.4003, + "step": 41778 + }, + { + "epoch": 1.61, + "learning_rate": 2.3243780456527317e-05, + "loss": 0.3839, + "step": 41779 + }, + { + "epoch": 1.61, + "learning_rate": 2.3243139266478586e-05, + "loss": 0.32, + "step": 41780 + }, + { + "epoch": 1.61, + "learning_rate": 2.3242498076429854e-05, + "loss": 0.2813, + "step": 41781 + }, + { + "epoch": 1.61, + "learning_rate": 2.3241856886381126e-05, + "loss": 0.2981, + "step": 41782 + }, + { + "epoch": 1.61, + "learning_rate": 2.3241215696332394e-05, + "loss": 0.2823, + "step": 41783 + }, + { + "epoch": 1.61, + "learning_rate": 2.3240574506283662e-05, + "loss": 0.4517, + "step": 41784 + }, + { + "epoch": 1.61, + "learning_rate": 2.3239933316234934e-05, + "loss": 0.452, + "step": 41785 + }, + { + "epoch": 1.61, + "learning_rate": 2.3239292126186202e-05, + "loss": 0.4105, + "step": 41786 + }, + { + "epoch": 1.61, + "learning_rate": 2.323865093613747e-05, + "loss": 0.3667, + "step": 41787 + }, + { + "epoch": 1.61, + "learning_rate": 2.3238009746088742e-05, + "loss": 0.2371, + "step": 41788 + }, + { + "epoch": 1.61, + "learning_rate": 2.323736855604001e-05, + "loss": 0.2328, + "step": 41789 + }, + { + "epoch": 1.61, + "learning_rate": 2.3236727365991283e-05, + "loss": 0.3216, + "step": 41790 + }, + { + "epoch": 1.61, + "learning_rate": 2.323608617594255e-05, + "loss": 0.3711, + "step": 41791 + }, + { + "epoch": 1.61, + "learning_rate": 2.323544498589382e-05, + "loss": 0.4406, + "step": 41792 + }, + { + "epoch": 1.61, + "learning_rate": 2.323480379584509e-05, + "loss": 0.2138, + "step": 41793 + }, + { + "epoch": 1.61, + "learning_rate": 2.323416260579636e-05, + "loss": 0.2282, + "step": 41794 + }, + { + "epoch": 1.61, + "learning_rate": 2.3233521415747628e-05, + "loss": 0.2858, + "step": 41795 + }, + { + "epoch": 1.61, + "learning_rate": 2.3232880225698896e-05, + "loss": 0.436, + "step": 41796 + }, + { + "epoch": 1.61, + "learning_rate": 2.3232239035650168e-05, + "loss": 0.4474, + "step": 41797 + }, + { + "epoch": 1.61, + "learning_rate": 2.323159784560144e-05, + "loss": 0.4436, + "step": 41798 + }, + { + "epoch": 1.61, + "learning_rate": 2.3230956655552708e-05, + "loss": 0.3505, + "step": 41799 + }, + { + "epoch": 1.61, + "learning_rate": 2.3230315465503976e-05, + "loss": 0.3528, + "step": 41800 + }, + { + "epoch": 1.61, + "learning_rate": 2.3229674275455248e-05, + "loss": 0.4097, + "step": 41801 + }, + { + "epoch": 1.61, + "learning_rate": 2.3229033085406516e-05, + "loss": 0.5261, + "step": 41802 + }, + { + "epoch": 1.61, + "learning_rate": 2.3228391895357784e-05, + "loss": 0.3873, + "step": 41803 + }, + { + "epoch": 1.61, + "learning_rate": 2.3227750705309053e-05, + "loss": 0.3806, + "step": 41804 + }, + { + "epoch": 1.61, + "learning_rate": 2.3227109515260324e-05, + "loss": 0.4252, + "step": 41805 + }, + { + "epoch": 1.61, + "learning_rate": 2.3226468325211596e-05, + "loss": 0.3023, + "step": 41806 + }, + { + "epoch": 1.61, + "learning_rate": 2.3225827135162864e-05, + "loss": 0.3759, + "step": 41807 + }, + { + "epoch": 1.61, + "learning_rate": 2.3225185945114133e-05, + "loss": 0.5073, + "step": 41808 + }, + { + "epoch": 1.61, + "learning_rate": 2.32245447550654e-05, + "loss": 0.3292, + "step": 41809 + }, + { + "epoch": 1.61, + "learning_rate": 2.3223903565016673e-05, + "loss": 0.2859, + "step": 41810 + }, + { + "epoch": 1.61, + "learning_rate": 2.322326237496794e-05, + "loss": 0.302, + "step": 41811 + }, + { + "epoch": 1.61, + "learning_rate": 2.322262118491921e-05, + "loss": 0.2637, + "step": 41812 + }, + { + "epoch": 1.61, + "learning_rate": 2.322197999487048e-05, + "loss": 0.4777, + "step": 41813 + }, + { + "epoch": 1.61, + "learning_rate": 2.322133880482175e-05, + "loss": 0.3391, + "step": 41814 + }, + { + "epoch": 1.61, + "learning_rate": 2.322069761477302e-05, + "loss": 0.4029, + "step": 41815 + }, + { + "epoch": 1.61, + "learning_rate": 2.322005642472429e-05, + "loss": 0.3609, + "step": 41816 + }, + { + "epoch": 1.61, + "learning_rate": 2.3219415234675558e-05, + "loss": 0.3367, + "step": 41817 + }, + { + "epoch": 1.61, + "learning_rate": 2.321877404462683e-05, + "loss": 0.4337, + "step": 41818 + }, + { + "epoch": 1.61, + "learning_rate": 2.3218132854578098e-05, + "loss": 0.3067, + "step": 41819 + }, + { + "epoch": 1.61, + "learning_rate": 2.3217491664529366e-05, + "loss": 0.3411, + "step": 41820 + }, + { + "epoch": 1.61, + "learning_rate": 2.3216850474480638e-05, + "loss": 0.2428, + "step": 41821 + }, + { + "epoch": 1.61, + "learning_rate": 2.3216209284431906e-05, + "loss": 0.4767, + "step": 41822 + }, + { + "epoch": 1.61, + "learning_rate": 2.3215568094383175e-05, + "loss": 0.3214, + "step": 41823 + }, + { + "epoch": 1.61, + "learning_rate": 2.3214926904334446e-05, + "loss": 0.2684, + "step": 41824 + }, + { + "epoch": 1.61, + "learning_rate": 2.3214285714285715e-05, + "loss": 0.3906, + "step": 41825 + }, + { + "epoch": 1.61, + "learning_rate": 2.3213644524236986e-05, + "loss": 0.3437, + "step": 41826 + }, + { + "epoch": 1.61, + "learning_rate": 2.3213003334188255e-05, + "loss": 0.3508, + "step": 41827 + }, + { + "epoch": 1.61, + "learning_rate": 2.3212362144139523e-05, + "loss": 0.4203, + "step": 41828 + }, + { + "epoch": 1.61, + "learning_rate": 2.3211720954090795e-05, + "loss": 0.2846, + "step": 41829 + }, + { + "epoch": 1.61, + "learning_rate": 2.3211079764042063e-05, + "loss": 0.3663, + "step": 41830 + }, + { + "epoch": 1.61, + "learning_rate": 2.321043857399333e-05, + "loss": 0.3874, + "step": 41831 + }, + { + "epoch": 1.61, + "learning_rate": 2.3209797383944603e-05, + "loss": 0.3831, + "step": 41832 + }, + { + "epoch": 1.61, + "learning_rate": 2.320915619389587e-05, + "loss": 0.3241, + "step": 41833 + }, + { + "epoch": 1.61, + "learning_rate": 2.3208515003847143e-05, + "loss": 0.4395, + "step": 41834 + }, + { + "epoch": 1.61, + "learning_rate": 2.320787381379841e-05, + "loss": 0.3879, + "step": 41835 + }, + { + "epoch": 1.61, + "learning_rate": 2.320723262374968e-05, + "loss": 0.4059, + "step": 41836 + }, + { + "epoch": 1.61, + "learning_rate": 2.320659143370095e-05, + "loss": 0.3347, + "step": 41837 + }, + { + "epoch": 1.61, + "learning_rate": 2.320595024365222e-05, + "loss": 0.2608, + "step": 41838 + }, + { + "epoch": 1.61, + "learning_rate": 2.3205309053603488e-05, + "loss": 0.3654, + "step": 41839 + }, + { + "epoch": 1.61, + "learning_rate": 2.3204667863554757e-05, + "loss": 0.4402, + "step": 41840 + }, + { + "epoch": 1.61, + "learning_rate": 2.3204026673506028e-05, + "loss": 0.3284, + "step": 41841 + }, + { + "epoch": 1.61, + "learning_rate": 2.32033854834573e-05, + "loss": 0.2621, + "step": 41842 + }, + { + "epoch": 1.61, + "learning_rate": 2.3202744293408568e-05, + "loss": 0.3117, + "step": 41843 + }, + { + "epoch": 1.61, + "learning_rate": 2.3202103103359837e-05, + "loss": 0.3525, + "step": 41844 + }, + { + "epoch": 1.61, + "learning_rate": 2.3201461913311105e-05, + "loss": 0.36, + "step": 41845 + }, + { + "epoch": 1.61, + "learning_rate": 2.3200820723262377e-05, + "loss": 0.463, + "step": 41846 + }, + { + "epoch": 1.61, + "learning_rate": 2.3200179533213645e-05, + "loss": 0.5078, + "step": 41847 + }, + { + "epoch": 1.61, + "learning_rate": 2.3199538343164913e-05, + "loss": 0.2818, + "step": 41848 + }, + { + "epoch": 1.61, + "learning_rate": 2.3198897153116185e-05, + "loss": 0.3037, + "step": 41849 + }, + { + "epoch": 1.61, + "learning_rate": 2.3198255963067457e-05, + "loss": 0.4194, + "step": 41850 + }, + { + "epoch": 1.61, + "learning_rate": 2.3197614773018725e-05, + "loss": 0.3212, + "step": 41851 + }, + { + "epoch": 1.61, + "learning_rate": 2.3196973582969993e-05, + "loss": 0.4606, + "step": 41852 + }, + { + "epoch": 1.61, + "learning_rate": 2.3196332392921262e-05, + "loss": 0.2415, + "step": 41853 + }, + { + "epoch": 1.61, + "learning_rate": 2.3195691202872533e-05, + "loss": 0.3133, + "step": 41854 + }, + { + "epoch": 1.61, + "learning_rate": 2.3195050012823802e-05, + "loss": 0.3932, + "step": 41855 + }, + { + "epoch": 1.61, + "learning_rate": 2.319440882277507e-05, + "loss": 0.4415, + "step": 41856 + }, + { + "epoch": 1.61, + "learning_rate": 2.3193767632726342e-05, + "loss": 0.4082, + "step": 41857 + }, + { + "epoch": 1.61, + "learning_rate": 2.319312644267761e-05, + "loss": 0.3807, + "step": 41858 + }, + { + "epoch": 1.61, + "learning_rate": 2.3192485252628882e-05, + "loss": 0.2752, + "step": 41859 + }, + { + "epoch": 1.61, + "learning_rate": 2.319184406258015e-05, + "loss": 0.3556, + "step": 41860 + }, + { + "epoch": 1.61, + "learning_rate": 2.319120287253142e-05, + "loss": 0.2853, + "step": 41861 + }, + { + "epoch": 1.61, + "learning_rate": 2.319056168248269e-05, + "loss": 0.4178, + "step": 41862 + }, + { + "epoch": 1.61, + "learning_rate": 2.318992049243396e-05, + "loss": 0.3107, + "step": 41863 + }, + { + "epoch": 1.61, + "learning_rate": 2.3189279302385227e-05, + "loss": 0.3966, + "step": 41864 + }, + { + "epoch": 1.61, + "learning_rate": 2.31886381123365e-05, + "loss": 0.3348, + "step": 41865 + }, + { + "epoch": 1.61, + "learning_rate": 2.3187996922287767e-05, + "loss": 0.3066, + "step": 41866 + }, + { + "epoch": 1.61, + "learning_rate": 2.3187355732239035e-05, + "loss": 0.2381, + "step": 41867 + }, + { + "epoch": 1.61, + "learning_rate": 2.3186714542190307e-05, + "loss": 0.4426, + "step": 41868 + }, + { + "epoch": 1.61, + "learning_rate": 2.3186073352141575e-05, + "loss": 0.4186, + "step": 41869 + }, + { + "epoch": 1.61, + "learning_rate": 2.3185432162092847e-05, + "loss": 0.3934, + "step": 41870 + }, + { + "epoch": 1.61, + "learning_rate": 2.3184790972044115e-05, + "loss": 0.464, + "step": 41871 + }, + { + "epoch": 1.61, + "learning_rate": 2.3184149781995384e-05, + "loss": 0.2972, + "step": 41872 + }, + { + "epoch": 1.61, + "learning_rate": 2.3183508591946655e-05, + "loss": 0.231, + "step": 41873 + }, + { + "epoch": 1.61, + "learning_rate": 2.3182867401897924e-05, + "loss": 0.4048, + "step": 41874 + }, + { + "epoch": 1.61, + "learning_rate": 2.3182226211849192e-05, + "loss": 0.2853, + "step": 41875 + }, + { + "epoch": 1.61, + "learning_rate": 2.318158502180046e-05, + "loss": 0.2858, + "step": 41876 + }, + { + "epoch": 1.61, + "learning_rate": 2.3180943831751732e-05, + "loss": 0.2996, + "step": 41877 + }, + { + "epoch": 1.61, + "learning_rate": 2.3180302641703004e-05, + "loss": 0.4411, + "step": 41878 + }, + { + "epoch": 1.61, + "learning_rate": 2.3179661451654272e-05, + "loss": 0.2705, + "step": 41879 + }, + { + "epoch": 1.61, + "learning_rate": 2.317902026160554e-05, + "loss": 0.3396, + "step": 41880 + }, + { + "epoch": 1.61, + "learning_rate": 2.3178379071556812e-05, + "loss": 0.2641, + "step": 41881 + }, + { + "epoch": 1.61, + "learning_rate": 2.317773788150808e-05, + "loss": 0.4382, + "step": 41882 + }, + { + "epoch": 1.61, + "learning_rate": 2.317709669145935e-05, + "loss": 0.3479, + "step": 41883 + }, + { + "epoch": 1.61, + "learning_rate": 2.3176455501410617e-05, + "loss": 0.2004, + "step": 41884 + }, + { + "epoch": 1.61, + "learning_rate": 2.317581431136189e-05, + "loss": 0.2567, + "step": 41885 + }, + { + "epoch": 1.61, + "learning_rate": 2.317517312131316e-05, + "loss": 0.2045, + "step": 41886 + }, + { + "epoch": 1.61, + "learning_rate": 2.317453193126443e-05, + "loss": 0.354, + "step": 41887 + }, + { + "epoch": 1.61, + "learning_rate": 2.3173890741215697e-05, + "loss": 0.4679, + "step": 41888 + }, + { + "epoch": 1.61, + "learning_rate": 2.3173249551166966e-05, + "loss": 0.2826, + "step": 41889 + }, + { + "epoch": 1.61, + "learning_rate": 2.3172608361118237e-05, + "loss": 0.4059, + "step": 41890 + }, + { + "epoch": 1.61, + "learning_rate": 2.3171967171069506e-05, + "loss": 0.3093, + "step": 41891 + }, + { + "epoch": 1.61, + "learning_rate": 2.3171325981020774e-05, + "loss": 0.2044, + "step": 41892 + }, + { + "epoch": 1.61, + "learning_rate": 2.3170684790972046e-05, + "loss": 0.2815, + "step": 41893 + }, + { + "epoch": 1.61, + "learning_rate": 2.3170043600923314e-05, + "loss": 0.295, + "step": 41894 + }, + { + "epoch": 1.61, + "learning_rate": 2.3169402410874586e-05, + "loss": 0.4144, + "step": 41895 + }, + { + "epoch": 1.61, + "learning_rate": 2.3168761220825854e-05, + "loss": 0.4062, + "step": 41896 + }, + { + "epoch": 1.61, + "learning_rate": 2.3168120030777122e-05, + "loss": 0.3845, + "step": 41897 + }, + { + "epoch": 1.61, + "learning_rate": 2.3167478840728394e-05, + "loss": 0.332, + "step": 41898 + }, + { + "epoch": 1.61, + "learning_rate": 2.3166837650679662e-05, + "loss": 0.2879, + "step": 41899 + }, + { + "epoch": 1.61, + "learning_rate": 2.316619646063093e-05, + "loss": 0.3035, + "step": 41900 + }, + { + "epoch": 1.61, + "learning_rate": 2.3165555270582203e-05, + "loss": 0.2351, + "step": 41901 + }, + { + "epoch": 1.61, + "learning_rate": 2.316491408053347e-05, + "loss": 0.3106, + "step": 41902 + }, + { + "epoch": 1.61, + "learning_rate": 2.316427289048474e-05, + "loss": 0.3814, + "step": 41903 + }, + { + "epoch": 1.61, + "learning_rate": 2.316363170043601e-05, + "loss": 0.4301, + "step": 41904 + }, + { + "epoch": 1.61, + "learning_rate": 2.316299051038728e-05, + "loss": 0.4553, + "step": 41905 + }, + { + "epoch": 1.61, + "learning_rate": 2.316234932033855e-05, + "loss": 0.3579, + "step": 41906 + }, + { + "epoch": 1.61, + "learning_rate": 2.316170813028982e-05, + "loss": 0.4001, + "step": 41907 + }, + { + "epoch": 1.61, + "learning_rate": 2.3161066940241088e-05, + "loss": 0.4803, + "step": 41908 + }, + { + "epoch": 1.61, + "learning_rate": 2.316042575019236e-05, + "loss": 0.222, + "step": 41909 + }, + { + "epoch": 1.61, + "learning_rate": 2.3159784560143628e-05, + "loss": 0.3308, + "step": 41910 + }, + { + "epoch": 1.61, + "learning_rate": 2.3159143370094896e-05, + "loss": 0.3882, + "step": 41911 + }, + { + "epoch": 1.61, + "learning_rate": 2.3158502180046164e-05, + "loss": 0.3796, + "step": 41912 + }, + { + "epoch": 1.61, + "learning_rate": 2.3157860989997436e-05, + "loss": 0.2508, + "step": 41913 + }, + { + "epoch": 1.61, + "learning_rate": 2.3157219799948708e-05, + "loss": 0.2204, + "step": 41914 + }, + { + "epoch": 1.61, + "learning_rate": 2.3156578609899976e-05, + "loss": 0.3307, + "step": 41915 + }, + { + "epoch": 1.61, + "learning_rate": 2.3155937419851244e-05, + "loss": 0.1943, + "step": 41916 + }, + { + "epoch": 1.61, + "learning_rate": 2.3155296229802516e-05, + "loss": 0.3698, + "step": 41917 + }, + { + "epoch": 1.61, + "learning_rate": 2.3154655039753784e-05, + "loss": 0.4371, + "step": 41918 + }, + { + "epoch": 1.61, + "learning_rate": 2.3154013849705053e-05, + "loss": 0.4406, + "step": 41919 + }, + { + "epoch": 1.61, + "learning_rate": 2.315337265965632e-05, + "loss": 0.4173, + "step": 41920 + }, + { + "epoch": 1.61, + "learning_rate": 2.3152731469607593e-05, + "loss": 0.466, + "step": 41921 + }, + { + "epoch": 1.61, + "learning_rate": 2.3152090279558865e-05, + "loss": 0.2727, + "step": 41922 + }, + { + "epoch": 1.61, + "learning_rate": 2.3151449089510133e-05, + "loss": 0.3769, + "step": 41923 + }, + { + "epoch": 1.61, + "learning_rate": 2.31508078994614e-05, + "loss": 0.3963, + "step": 41924 + }, + { + "epoch": 1.61, + "learning_rate": 2.315016670941267e-05, + "loss": 0.3555, + "step": 41925 + }, + { + "epoch": 1.61, + "learning_rate": 2.314952551936394e-05, + "loss": 0.2868, + "step": 41926 + }, + { + "epoch": 1.61, + "learning_rate": 2.314888432931521e-05, + "loss": 0.3782, + "step": 41927 + }, + { + "epoch": 1.61, + "learning_rate": 2.3148243139266478e-05, + "loss": 0.3751, + "step": 41928 + }, + { + "epoch": 1.61, + "learning_rate": 2.314760194921775e-05, + "loss": 0.2905, + "step": 41929 + }, + { + "epoch": 1.61, + "learning_rate": 2.314696075916902e-05, + "loss": 0.4569, + "step": 41930 + }, + { + "epoch": 1.61, + "learning_rate": 2.314631956912029e-05, + "loss": 0.2939, + "step": 41931 + }, + { + "epoch": 1.61, + "learning_rate": 2.3145678379071558e-05, + "loss": 0.3008, + "step": 41932 + }, + { + "epoch": 1.61, + "learning_rate": 2.3145037189022826e-05, + "loss": 0.3789, + "step": 41933 + }, + { + "epoch": 1.61, + "learning_rate": 2.3144395998974098e-05, + "loss": 0.3464, + "step": 41934 + }, + { + "epoch": 1.61, + "learning_rate": 2.3143754808925366e-05, + "loss": 0.3718, + "step": 41935 + }, + { + "epoch": 1.61, + "learning_rate": 2.3143113618876635e-05, + "loss": 0.2662, + "step": 41936 + }, + { + "epoch": 1.61, + "learning_rate": 2.3142472428827906e-05, + "loss": 0.3361, + "step": 41937 + }, + { + "epoch": 1.61, + "learning_rate": 2.3141831238779175e-05, + "loss": 0.2346, + "step": 41938 + }, + { + "epoch": 1.61, + "learning_rate": 2.3141190048730446e-05, + "loss": 0.3102, + "step": 41939 + }, + { + "epoch": 1.61, + "learning_rate": 2.3140548858681715e-05, + "loss": 0.3995, + "step": 41940 + }, + { + "epoch": 1.61, + "learning_rate": 2.3139907668632983e-05, + "loss": 0.3585, + "step": 41941 + }, + { + "epoch": 1.61, + "learning_rate": 2.3139266478584255e-05, + "loss": 0.3332, + "step": 41942 + }, + { + "epoch": 1.61, + "learning_rate": 2.3138625288535523e-05, + "loss": 0.4477, + "step": 41943 + }, + { + "epoch": 1.61, + "learning_rate": 2.313798409848679e-05, + "loss": 0.4748, + "step": 41944 + }, + { + "epoch": 1.61, + "learning_rate": 2.3137342908438063e-05, + "loss": 0.3069, + "step": 41945 + }, + { + "epoch": 1.61, + "learning_rate": 2.313670171838933e-05, + "loss": 0.3467, + "step": 41946 + }, + { + "epoch": 1.61, + "learning_rate": 2.31360605283406e-05, + "loss": 0.377, + "step": 41947 + }, + { + "epoch": 1.61, + "learning_rate": 2.313541933829187e-05, + "loss": 0.2202, + "step": 41948 + }, + { + "epoch": 1.61, + "learning_rate": 2.313477814824314e-05, + "loss": 0.29, + "step": 41949 + }, + { + "epoch": 1.61, + "learning_rate": 2.313413695819441e-05, + "loss": 0.3732, + "step": 41950 + }, + { + "epoch": 1.61, + "learning_rate": 2.313349576814568e-05, + "loss": 0.2726, + "step": 41951 + }, + { + "epoch": 1.61, + "learning_rate": 2.3132854578096948e-05, + "loss": 0.3096, + "step": 41952 + }, + { + "epoch": 1.61, + "learning_rate": 2.313221338804822e-05, + "loss": 0.2914, + "step": 41953 + }, + { + "epoch": 1.61, + "learning_rate": 2.3131572197999488e-05, + "loss": 0.3965, + "step": 41954 + }, + { + "epoch": 1.61, + "learning_rate": 2.3130931007950757e-05, + "loss": 0.3036, + "step": 41955 + }, + { + "epoch": 1.61, + "learning_rate": 2.3130289817902025e-05, + "loss": 0.2912, + "step": 41956 + }, + { + "epoch": 1.61, + "learning_rate": 2.3129648627853297e-05, + "loss": 0.4386, + "step": 41957 + }, + { + "epoch": 1.61, + "learning_rate": 2.312900743780457e-05, + "loss": 0.2235, + "step": 41958 + }, + { + "epoch": 1.61, + "learning_rate": 2.3128366247755837e-05, + "loss": 0.5209, + "step": 41959 + }, + { + "epoch": 1.61, + "learning_rate": 2.3127725057707105e-05, + "loss": 0.3796, + "step": 41960 + }, + { + "epoch": 1.61, + "learning_rate": 2.3127083867658377e-05, + "loss": 0.3614, + "step": 41961 + }, + { + "epoch": 1.61, + "learning_rate": 2.3126442677609645e-05, + "loss": 0.4474, + "step": 41962 + }, + { + "epoch": 1.61, + "learning_rate": 2.3125801487560913e-05, + "loss": 0.1979, + "step": 41963 + }, + { + "epoch": 1.61, + "learning_rate": 2.3125160297512182e-05, + "loss": 0.3941, + "step": 41964 + }, + { + "epoch": 1.61, + "learning_rate": 2.3124519107463453e-05, + "loss": 0.3689, + "step": 41965 + }, + { + "epoch": 1.61, + "learning_rate": 2.3123877917414725e-05, + "loss": 0.3042, + "step": 41966 + }, + { + "epoch": 1.61, + "learning_rate": 2.3123236727365994e-05, + "loss": 0.3779, + "step": 41967 + }, + { + "epoch": 1.61, + "learning_rate": 2.3122595537317262e-05, + "loss": 0.4032, + "step": 41968 + }, + { + "epoch": 1.61, + "learning_rate": 2.312195434726853e-05, + "loss": 0.4587, + "step": 41969 + }, + { + "epoch": 1.61, + "learning_rate": 2.3121313157219802e-05, + "loss": 0.2987, + "step": 41970 + }, + { + "epoch": 1.61, + "learning_rate": 2.312067196717107e-05, + "loss": 0.2156, + "step": 41971 + }, + { + "epoch": 1.61, + "learning_rate": 2.312003077712234e-05, + "loss": 0.3615, + "step": 41972 + }, + { + "epoch": 1.61, + "learning_rate": 2.311938958707361e-05, + "loss": 0.2564, + "step": 41973 + }, + { + "epoch": 1.61, + "learning_rate": 2.311874839702488e-05, + "loss": 0.3996, + "step": 41974 + }, + { + "epoch": 1.61, + "learning_rate": 2.311810720697615e-05, + "loss": 0.264, + "step": 41975 + }, + { + "epoch": 1.61, + "learning_rate": 2.311746601692742e-05, + "loss": 0.3318, + "step": 41976 + }, + { + "epoch": 1.61, + "learning_rate": 2.3116824826878687e-05, + "loss": 0.2882, + "step": 41977 + }, + { + "epoch": 1.61, + "learning_rate": 2.311618363682996e-05, + "loss": 0.3008, + "step": 41978 + }, + { + "epoch": 1.61, + "learning_rate": 2.3115542446781227e-05, + "loss": 0.3304, + "step": 41979 + }, + { + "epoch": 1.61, + "learning_rate": 2.3114901256732495e-05, + "loss": 0.4538, + "step": 41980 + }, + { + "epoch": 1.61, + "learning_rate": 2.3114260066683767e-05, + "loss": 0.4583, + "step": 41981 + }, + { + "epoch": 1.61, + "learning_rate": 2.3113618876635035e-05, + "loss": 0.3579, + "step": 41982 + }, + { + "epoch": 1.61, + "learning_rate": 2.3112977686586304e-05, + "loss": 0.2616, + "step": 41983 + }, + { + "epoch": 1.61, + "learning_rate": 2.3112336496537575e-05, + "loss": 0.3043, + "step": 41984 + }, + { + "epoch": 1.61, + "learning_rate": 2.3111695306488844e-05, + "loss": 0.3099, + "step": 41985 + }, + { + "epoch": 1.61, + "learning_rate": 2.3111054116440115e-05, + "loss": 0.3353, + "step": 41986 + }, + { + "epoch": 1.61, + "learning_rate": 2.3110412926391384e-05, + "loss": 0.3897, + "step": 41987 + }, + { + "epoch": 1.61, + "learning_rate": 2.3109771736342652e-05, + "loss": 0.3563, + "step": 41988 + }, + { + "epoch": 1.61, + "learning_rate": 2.3109130546293924e-05, + "loss": 0.397, + "step": 41989 + }, + { + "epoch": 1.61, + "learning_rate": 2.3108489356245192e-05, + "loss": 0.295, + "step": 41990 + }, + { + "epoch": 1.61, + "learning_rate": 2.310784816619646e-05, + "loss": 0.3609, + "step": 41991 + }, + { + "epoch": 1.61, + "learning_rate": 2.310720697614773e-05, + "loss": 0.2708, + "step": 41992 + }, + { + "epoch": 1.61, + "learning_rate": 2.3106565786099e-05, + "loss": 0.371, + "step": 41993 + }, + { + "epoch": 1.61, + "learning_rate": 2.3105924596050272e-05, + "loss": 0.2565, + "step": 41994 + }, + { + "epoch": 1.61, + "learning_rate": 2.310528340600154e-05, + "loss": 0.462, + "step": 41995 + }, + { + "epoch": 1.61, + "learning_rate": 2.310464221595281e-05, + "loss": 0.388, + "step": 41996 + }, + { + "epoch": 1.61, + "learning_rate": 2.310400102590408e-05, + "loss": 0.3153, + "step": 41997 + }, + { + "epoch": 1.61, + "learning_rate": 2.310335983585535e-05, + "loss": 0.3004, + "step": 41998 + }, + { + "epoch": 1.61, + "learning_rate": 2.3102718645806617e-05, + "loss": 0.3319, + "step": 41999 + }, + { + "epoch": 1.61, + "learning_rate": 2.3102077455757886e-05, + "loss": 0.3856, + "step": 42000 + }, + { + "epoch": 1.61, + "learning_rate": 2.3101436265709157e-05, + "loss": 0.5766, + "step": 42001 + }, + { + "epoch": 1.61, + "learning_rate": 2.310079507566043e-05, + "loss": 0.4307, + "step": 42002 + }, + { + "epoch": 1.61, + "learning_rate": 2.3100153885611697e-05, + "loss": 0.4677, + "step": 42003 + }, + { + "epoch": 1.61, + "learning_rate": 2.3099512695562966e-05, + "loss": 0.4163, + "step": 42004 + }, + { + "epoch": 1.61, + "learning_rate": 2.3098871505514237e-05, + "loss": 0.4064, + "step": 42005 + }, + { + "epoch": 1.61, + "learning_rate": 2.3098230315465506e-05, + "loss": 0.3746, + "step": 42006 + }, + { + "epoch": 1.62, + "learning_rate": 2.3097589125416774e-05, + "loss": 0.3106, + "step": 42007 + }, + { + "epoch": 1.62, + "learning_rate": 2.3096947935368042e-05, + "loss": 0.3467, + "step": 42008 + }, + { + "epoch": 1.62, + "learning_rate": 2.3096306745319314e-05, + "loss": 0.2106, + "step": 42009 + }, + { + "epoch": 1.62, + "learning_rate": 2.3095665555270582e-05, + "loss": 0.3244, + "step": 42010 + }, + { + "epoch": 1.62, + "learning_rate": 2.3095024365221854e-05, + "loss": 0.3533, + "step": 42011 + }, + { + "epoch": 1.62, + "learning_rate": 2.3094383175173122e-05, + "loss": 0.4504, + "step": 42012 + }, + { + "epoch": 1.62, + "learning_rate": 2.309374198512439e-05, + "loss": 0.3213, + "step": 42013 + }, + { + "epoch": 1.62, + "learning_rate": 2.3093100795075663e-05, + "loss": 0.3091, + "step": 42014 + }, + { + "epoch": 1.62, + "learning_rate": 2.309245960502693e-05, + "loss": 0.4258, + "step": 42015 + }, + { + "epoch": 1.62, + "learning_rate": 2.30918184149782e-05, + "loss": 0.4268, + "step": 42016 + }, + { + "epoch": 1.62, + "learning_rate": 2.309117722492947e-05, + "loss": 0.4196, + "step": 42017 + }, + { + "epoch": 1.62, + "learning_rate": 2.309053603488074e-05, + "loss": 0.3032, + "step": 42018 + }, + { + "epoch": 1.62, + "learning_rate": 2.3089894844832008e-05, + "loss": 0.3455, + "step": 42019 + }, + { + "epoch": 1.62, + "learning_rate": 2.308925365478328e-05, + "loss": 0.3193, + "step": 42020 + }, + { + "epoch": 1.62, + "learning_rate": 2.3088612464734548e-05, + "loss": 0.3841, + "step": 42021 + }, + { + "epoch": 1.62, + "learning_rate": 2.308797127468582e-05, + "loss": 0.3198, + "step": 42022 + }, + { + "epoch": 1.62, + "learning_rate": 2.3087330084637088e-05, + "loss": 0.3394, + "step": 42023 + }, + { + "epoch": 1.62, + "learning_rate": 2.3086688894588356e-05, + "loss": 0.2796, + "step": 42024 + }, + { + "epoch": 1.62, + "learning_rate": 2.3086047704539628e-05, + "loss": 0.2013, + "step": 42025 + }, + { + "epoch": 1.62, + "learning_rate": 2.3085406514490896e-05, + "loss": 0.3819, + "step": 42026 + }, + { + "epoch": 1.62, + "learning_rate": 2.3084765324442164e-05, + "loss": 0.4236, + "step": 42027 + }, + { + "epoch": 1.62, + "learning_rate": 2.3084124134393436e-05, + "loss": 0.382, + "step": 42028 + }, + { + "epoch": 1.62, + "learning_rate": 2.3083482944344704e-05, + "loss": 0.4207, + "step": 42029 + }, + { + "epoch": 1.62, + "learning_rate": 2.3082841754295976e-05, + "loss": 0.3313, + "step": 42030 + }, + { + "epoch": 1.62, + "learning_rate": 2.3082200564247244e-05, + "loss": 0.4913, + "step": 42031 + }, + { + "epoch": 1.62, + "learning_rate": 2.3081559374198513e-05, + "loss": 0.3769, + "step": 42032 + }, + { + "epoch": 1.62, + "learning_rate": 2.3080918184149784e-05, + "loss": 0.322, + "step": 42033 + }, + { + "epoch": 1.62, + "learning_rate": 2.3080276994101053e-05, + "loss": 0.4339, + "step": 42034 + }, + { + "epoch": 1.62, + "learning_rate": 2.307963580405232e-05, + "loss": 0.4163, + "step": 42035 + }, + { + "epoch": 1.62, + "learning_rate": 2.307899461400359e-05, + "loss": 0.3583, + "step": 42036 + }, + { + "epoch": 1.62, + "learning_rate": 2.307835342395486e-05, + "loss": 0.3888, + "step": 42037 + }, + { + "epoch": 1.62, + "learning_rate": 2.3077712233906133e-05, + "loss": 0.3745, + "step": 42038 + }, + { + "epoch": 1.62, + "learning_rate": 2.30770710438574e-05, + "loss": 0.3766, + "step": 42039 + }, + { + "epoch": 1.62, + "learning_rate": 2.307642985380867e-05, + "loss": 0.3115, + "step": 42040 + }, + { + "epoch": 1.62, + "learning_rate": 2.307578866375994e-05, + "loss": 0.306, + "step": 42041 + }, + { + "epoch": 1.62, + "learning_rate": 2.307514747371121e-05, + "loss": 0.3749, + "step": 42042 + }, + { + "epoch": 1.62, + "learning_rate": 2.3074506283662478e-05, + "loss": 0.3208, + "step": 42043 + }, + { + "epoch": 1.62, + "learning_rate": 2.3073865093613746e-05, + "loss": 0.3662, + "step": 42044 + }, + { + "epoch": 1.62, + "learning_rate": 2.3073223903565018e-05, + "loss": 0.2994, + "step": 42045 + }, + { + "epoch": 1.62, + "learning_rate": 2.307258271351629e-05, + "loss": 0.2928, + "step": 42046 + }, + { + "epoch": 1.62, + "learning_rate": 2.3071941523467558e-05, + "loss": 0.3803, + "step": 42047 + }, + { + "epoch": 1.62, + "learning_rate": 2.3071300333418826e-05, + "loss": 0.3631, + "step": 42048 + }, + { + "epoch": 1.62, + "learning_rate": 2.3070659143370095e-05, + "loss": 0.4424, + "step": 42049 + }, + { + "epoch": 1.62, + "learning_rate": 2.3070017953321366e-05, + "loss": 0.3605, + "step": 42050 + }, + { + "epoch": 1.62, + "learning_rate": 2.3069376763272635e-05, + "loss": 0.3511, + "step": 42051 + }, + { + "epoch": 1.62, + "learning_rate": 2.3068735573223903e-05, + "loss": 0.4292, + "step": 42052 + }, + { + "epoch": 1.62, + "learning_rate": 2.3068094383175175e-05, + "loss": 0.449, + "step": 42053 + }, + { + "epoch": 1.62, + "learning_rate": 2.3067453193126443e-05, + "loss": 0.3226, + "step": 42054 + }, + { + "epoch": 1.62, + "learning_rate": 2.3066812003077715e-05, + "loss": 0.3161, + "step": 42055 + }, + { + "epoch": 1.62, + "learning_rate": 2.3066170813028983e-05, + "loss": 0.3115, + "step": 42056 + }, + { + "epoch": 1.62, + "learning_rate": 2.306552962298025e-05, + "loss": 0.3384, + "step": 42057 + }, + { + "epoch": 1.62, + "learning_rate": 2.3064888432931523e-05, + "loss": 0.3156, + "step": 42058 + }, + { + "epoch": 1.62, + "learning_rate": 2.306424724288279e-05, + "loss": 0.4086, + "step": 42059 + }, + { + "epoch": 1.62, + "learning_rate": 2.306360605283406e-05, + "loss": 0.3884, + "step": 42060 + }, + { + "epoch": 1.62, + "learning_rate": 2.306296486278533e-05, + "loss": 0.4132, + "step": 42061 + }, + { + "epoch": 1.62, + "learning_rate": 2.30623236727366e-05, + "loss": 0.4812, + "step": 42062 + }, + { + "epoch": 1.62, + "learning_rate": 2.3061682482687868e-05, + "loss": 0.3572, + "step": 42063 + }, + { + "epoch": 1.62, + "learning_rate": 2.306104129263914e-05, + "loss": 0.3046, + "step": 42064 + }, + { + "epoch": 1.62, + "learning_rate": 2.3060400102590408e-05, + "loss": 0.3066, + "step": 42065 + }, + { + "epoch": 1.62, + "learning_rate": 2.305975891254168e-05, + "loss": 0.3556, + "step": 42066 + }, + { + "epoch": 1.62, + "learning_rate": 2.305911772249295e-05, + "loss": 0.4427, + "step": 42067 + }, + { + "epoch": 1.62, + "learning_rate": 2.3058476532444217e-05, + "loss": 0.3366, + "step": 42068 + }, + { + "epoch": 1.62, + "learning_rate": 2.305783534239549e-05, + "loss": 0.423, + "step": 42069 + }, + { + "epoch": 1.62, + "learning_rate": 2.3057194152346757e-05, + "loss": 0.3948, + "step": 42070 + }, + { + "epoch": 1.62, + "learning_rate": 2.3056552962298025e-05, + "loss": 0.309, + "step": 42071 + }, + { + "epoch": 1.62, + "learning_rate": 2.3055911772249293e-05, + "loss": 0.3777, + "step": 42072 + }, + { + "epoch": 1.62, + "learning_rate": 2.3055270582200565e-05, + "loss": 0.3092, + "step": 42073 + }, + { + "epoch": 1.62, + "learning_rate": 2.3054629392151837e-05, + "loss": 0.3924, + "step": 42074 + }, + { + "epoch": 1.62, + "learning_rate": 2.3053988202103105e-05, + "loss": 0.2844, + "step": 42075 + }, + { + "epoch": 1.62, + "learning_rate": 2.3053347012054373e-05, + "loss": 0.3276, + "step": 42076 + }, + { + "epoch": 1.62, + "learning_rate": 2.3052705822005645e-05, + "loss": 0.3218, + "step": 42077 + }, + { + "epoch": 1.62, + "learning_rate": 2.3052064631956913e-05, + "loss": 0.3963, + "step": 42078 + }, + { + "epoch": 1.62, + "learning_rate": 2.3051423441908182e-05, + "loss": 0.3645, + "step": 42079 + }, + { + "epoch": 1.62, + "learning_rate": 2.305078225185945e-05, + "loss": 0.3383, + "step": 42080 + }, + { + "epoch": 1.62, + "learning_rate": 2.3050141061810722e-05, + "loss": 0.3611, + "step": 42081 + }, + { + "epoch": 1.62, + "learning_rate": 2.3049499871761994e-05, + "loss": 0.3748, + "step": 42082 + }, + { + "epoch": 1.62, + "learning_rate": 2.3048858681713262e-05, + "loss": 0.3806, + "step": 42083 + }, + { + "epoch": 1.62, + "learning_rate": 2.304821749166453e-05, + "loss": 0.2592, + "step": 42084 + }, + { + "epoch": 1.62, + "learning_rate": 2.3047576301615802e-05, + "loss": 0.2379, + "step": 42085 + }, + { + "epoch": 1.62, + "learning_rate": 2.304693511156707e-05, + "loss": 0.2912, + "step": 42086 + }, + { + "epoch": 1.62, + "learning_rate": 2.304629392151834e-05, + "loss": 0.425, + "step": 42087 + }, + { + "epoch": 1.62, + "learning_rate": 2.3045652731469607e-05, + "loss": 0.1957, + "step": 42088 + }, + { + "epoch": 1.62, + "learning_rate": 2.304501154142088e-05, + "loss": 0.3872, + "step": 42089 + }, + { + "epoch": 1.62, + "learning_rate": 2.3044370351372147e-05, + "loss": 0.4056, + "step": 42090 + }, + { + "epoch": 1.62, + "learning_rate": 2.304372916132342e-05, + "loss": 0.4234, + "step": 42091 + }, + { + "epoch": 1.62, + "learning_rate": 2.3043087971274687e-05, + "loss": 0.3045, + "step": 42092 + }, + { + "epoch": 1.62, + "learning_rate": 2.3042446781225955e-05, + "loss": 0.4073, + "step": 42093 + }, + { + "epoch": 1.62, + "learning_rate": 2.3041805591177227e-05, + "loss": 0.4014, + "step": 42094 + }, + { + "epoch": 1.62, + "learning_rate": 2.3041164401128495e-05, + "loss": 0.2717, + "step": 42095 + }, + { + "epoch": 1.62, + "learning_rate": 2.3040523211079764e-05, + "loss": 0.311, + "step": 42096 + }, + { + "epoch": 1.62, + "learning_rate": 2.3039882021031035e-05, + "loss": 0.3776, + "step": 42097 + }, + { + "epoch": 1.62, + "learning_rate": 2.3039240830982304e-05, + "loss": 0.4916, + "step": 42098 + }, + { + "epoch": 1.62, + "learning_rate": 2.3038599640933572e-05, + "loss": 0.3881, + "step": 42099 + }, + { + "epoch": 1.62, + "learning_rate": 2.3037958450884844e-05, + "loss": 0.4063, + "step": 42100 + }, + { + "epoch": 1.62, + "learning_rate": 2.3037317260836112e-05, + "loss": 0.2784, + "step": 42101 + }, + { + "epoch": 1.62, + "learning_rate": 2.3036676070787384e-05, + "loss": 0.5744, + "step": 42102 + }, + { + "epoch": 1.62, + "learning_rate": 2.3036034880738652e-05, + "loss": 0.3816, + "step": 42103 + }, + { + "epoch": 1.62, + "learning_rate": 2.303539369068992e-05, + "loss": 0.3693, + "step": 42104 + }, + { + "epoch": 1.62, + "learning_rate": 2.3034752500641192e-05, + "loss": 0.2993, + "step": 42105 + }, + { + "epoch": 1.62, + "learning_rate": 2.303411131059246e-05, + "loss": 0.3047, + "step": 42106 + }, + { + "epoch": 1.62, + "learning_rate": 2.303347012054373e-05, + "loss": 0.487, + "step": 42107 + }, + { + "epoch": 1.62, + "learning_rate": 2.3032828930494997e-05, + "loss": 0.3251, + "step": 42108 + }, + { + "epoch": 1.62, + "learning_rate": 2.303218774044627e-05, + "loss": 0.2122, + "step": 42109 + }, + { + "epoch": 1.62, + "learning_rate": 2.303154655039754e-05, + "loss": 0.177, + "step": 42110 + }, + { + "epoch": 1.62, + "learning_rate": 2.303090536034881e-05, + "loss": 0.4197, + "step": 42111 + }, + { + "epoch": 1.62, + "learning_rate": 2.3030264170300077e-05, + "loss": 0.3842, + "step": 42112 + }, + { + "epoch": 1.62, + "learning_rate": 2.302962298025135e-05, + "loss": 0.4992, + "step": 42113 + }, + { + "epoch": 1.62, + "learning_rate": 2.3028981790202617e-05, + "loss": 0.3308, + "step": 42114 + }, + { + "epoch": 1.62, + "learning_rate": 2.3028340600153886e-05, + "loss": 0.3881, + "step": 42115 + }, + { + "epoch": 1.62, + "learning_rate": 2.3027699410105154e-05, + "loss": 0.3019, + "step": 42116 + }, + { + "epoch": 1.62, + "learning_rate": 2.3027058220056426e-05, + "loss": 0.407, + "step": 42117 + }, + { + "epoch": 1.62, + "learning_rate": 2.3026417030007697e-05, + "loss": 0.41, + "step": 42118 + }, + { + "epoch": 1.62, + "learning_rate": 2.3025775839958966e-05, + "loss": 0.3606, + "step": 42119 + }, + { + "epoch": 1.62, + "learning_rate": 2.3025134649910234e-05, + "loss": 0.3082, + "step": 42120 + }, + { + "epoch": 1.62, + "learning_rate": 2.3024493459861506e-05, + "loss": 0.4049, + "step": 42121 + }, + { + "epoch": 1.62, + "learning_rate": 2.3023852269812774e-05, + "loss": 0.4197, + "step": 42122 + }, + { + "epoch": 1.62, + "learning_rate": 2.3023211079764042e-05, + "loss": 0.3624, + "step": 42123 + }, + { + "epoch": 1.62, + "learning_rate": 2.302256988971531e-05, + "loss": 0.2772, + "step": 42124 + }, + { + "epoch": 1.62, + "learning_rate": 2.3021928699666583e-05, + "loss": 0.2993, + "step": 42125 + }, + { + "epoch": 1.62, + "learning_rate": 2.3021287509617854e-05, + "loss": 0.2577, + "step": 42126 + }, + { + "epoch": 1.62, + "learning_rate": 2.3020646319569123e-05, + "loss": 0.4056, + "step": 42127 + }, + { + "epoch": 1.62, + "learning_rate": 2.302000512952039e-05, + "loss": 0.2789, + "step": 42128 + }, + { + "epoch": 1.62, + "learning_rate": 2.301936393947166e-05, + "loss": 0.3514, + "step": 42129 + }, + { + "epoch": 1.62, + "learning_rate": 2.301872274942293e-05, + "loss": 0.3655, + "step": 42130 + }, + { + "epoch": 1.62, + "learning_rate": 2.30180815593742e-05, + "loss": 0.3446, + "step": 42131 + }, + { + "epoch": 1.62, + "learning_rate": 2.3017440369325468e-05, + "loss": 0.3039, + "step": 42132 + }, + { + "epoch": 1.62, + "learning_rate": 2.301679917927674e-05, + "loss": 0.3136, + "step": 42133 + }, + { + "epoch": 1.62, + "learning_rate": 2.3016157989228008e-05, + "loss": 0.3737, + "step": 42134 + }, + { + "epoch": 1.62, + "learning_rate": 2.301551679917928e-05, + "loss": 0.2454, + "step": 42135 + }, + { + "epoch": 1.62, + "learning_rate": 2.3014875609130548e-05, + "loss": 0.3672, + "step": 42136 + }, + { + "epoch": 1.62, + "learning_rate": 2.3014234419081816e-05, + "loss": 0.3137, + "step": 42137 + }, + { + "epoch": 1.62, + "learning_rate": 2.3013593229033088e-05, + "loss": 0.3319, + "step": 42138 + }, + { + "epoch": 1.62, + "learning_rate": 2.3012952038984356e-05, + "loss": 0.4277, + "step": 42139 + }, + { + "epoch": 1.62, + "learning_rate": 2.3012310848935624e-05, + "loss": 0.3971, + "step": 42140 + }, + { + "epoch": 1.62, + "learning_rate": 2.3011669658886896e-05, + "loss": 0.3364, + "step": 42141 + }, + { + "epoch": 1.62, + "learning_rate": 2.3011028468838164e-05, + "loss": 0.2958, + "step": 42142 + }, + { + "epoch": 1.62, + "learning_rate": 2.3010387278789433e-05, + "loss": 0.4732, + "step": 42143 + }, + { + "epoch": 1.62, + "learning_rate": 2.3009746088740704e-05, + "loss": 0.4734, + "step": 42144 + }, + { + "epoch": 1.62, + "learning_rate": 2.3009104898691973e-05, + "loss": 0.3795, + "step": 42145 + }, + { + "epoch": 1.62, + "learning_rate": 2.3008463708643245e-05, + "loss": 0.4374, + "step": 42146 + }, + { + "epoch": 1.62, + "learning_rate": 2.3007822518594513e-05, + "loss": 0.3005, + "step": 42147 + }, + { + "epoch": 1.62, + "learning_rate": 2.300718132854578e-05, + "loss": 0.3329, + "step": 42148 + }, + { + "epoch": 1.62, + "learning_rate": 2.3006540138497053e-05, + "loss": 0.3335, + "step": 42149 + }, + { + "epoch": 1.62, + "learning_rate": 2.300589894844832e-05, + "loss": 0.4217, + "step": 42150 + }, + { + "epoch": 1.62, + "learning_rate": 2.300525775839959e-05, + "loss": 0.2293, + "step": 42151 + }, + { + "epoch": 1.62, + "learning_rate": 2.3004616568350858e-05, + "loss": 0.292, + "step": 42152 + }, + { + "epoch": 1.62, + "learning_rate": 2.300397537830213e-05, + "loss": 0.2928, + "step": 42153 + }, + { + "epoch": 1.62, + "learning_rate": 2.30033341882534e-05, + "loss": 0.311, + "step": 42154 + }, + { + "epoch": 1.62, + "learning_rate": 2.300269299820467e-05, + "loss": 0.3325, + "step": 42155 + }, + { + "epoch": 1.62, + "learning_rate": 2.3002051808155938e-05, + "loss": 0.3562, + "step": 42156 + }, + { + "epoch": 1.62, + "learning_rate": 2.300141061810721e-05, + "loss": 0.4705, + "step": 42157 + }, + { + "epoch": 1.62, + "learning_rate": 2.3000769428058478e-05, + "loss": 0.2603, + "step": 42158 + }, + { + "epoch": 1.62, + "learning_rate": 2.3000128238009746e-05, + "loss": 0.4261, + "step": 42159 + }, + { + "epoch": 1.62, + "learning_rate": 2.2999487047961015e-05, + "loss": 0.3643, + "step": 42160 + }, + { + "epoch": 1.62, + "learning_rate": 2.2998845857912286e-05, + "loss": 0.3742, + "step": 42161 + }, + { + "epoch": 1.62, + "learning_rate": 2.2998204667863558e-05, + "loss": 0.3993, + "step": 42162 + }, + { + "epoch": 1.62, + "learning_rate": 2.2997563477814826e-05, + "loss": 0.5639, + "step": 42163 + }, + { + "epoch": 1.62, + "learning_rate": 2.2996922287766095e-05, + "loss": 0.5294, + "step": 42164 + }, + { + "epoch": 1.62, + "learning_rate": 2.2996281097717366e-05, + "loss": 0.3058, + "step": 42165 + }, + { + "epoch": 1.62, + "learning_rate": 2.2995639907668635e-05, + "loss": 0.3909, + "step": 42166 + }, + { + "epoch": 1.62, + "learning_rate": 2.2994998717619903e-05, + "loss": 0.2595, + "step": 42167 + }, + { + "epoch": 1.62, + "learning_rate": 2.299435752757117e-05, + "loss": 0.2599, + "step": 42168 + }, + { + "epoch": 1.62, + "learning_rate": 2.2993716337522443e-05, + "loss": 0.3622, + "step": 42169 + }, + { + "epoch": 1.62, + "learning_rate": 2.299307514747371e-05, + "loss": 0.3071, + "step": 42170 + }, + { + "epoch": 1.62, + "learning_rate": 2.2992433957424983e-05, + "loss": 0.2744, + "step": 42171 + }, + { + "epoch": 1.62, + "learning_rate": 2.299179276737625e-05, + "loss": 0.3072, + "step": 42172 + }, + { + "epoch": 1.62, + "learning_rate": 2.299115157732752e-05, + "loss": 0.4822, + "step": 42173 + }, + { + "epoch": 1.62, + "learning_rate": 2.299051038727879e-05, + "loss": 0.2878, + "step": 42174 + }, + { + "epoch": 1.62, + "learning_rate": 2.298986919723006e-05, + "loss": 0.3898, + "step": 42175 + }, + { + "epoch": 1.62, + "learning_rate": 2.2989228007181328e-05, + "loss": 0.4718, + "step": 42176 + }, + { + "epoch": 1.62, + "learning_rate": 2.29885868171326e-05, + "loss": 0.3779, + "step": 42177 + }, + { + "epoch": 1.62, + "learning_rate": 2.298794562708387e-05, + "loss": 0.4999, + "step": 42178 + }, + { + "epoch": 1.62, + "learning_rate": 2.2987304437035137e-05, + "loss": 0.3272, + "step": 42179 + }, + { + "epoch": 1.62, + "learning_rate": 2.298666324698641e-05, + "loss": 0.3851, + "step": 42180 + }, + { + "epoch": 1.62, + "learning_rate": 2.2986022056937677e-05, + "loss": 0.3328, + "step": 42181 + }, + { + "epoch": 1.62, + "learning_rate": 2.298538086688895e-05, + "loss": 0.4391, + "step": 42182 + }, + { + "epoch": 1.62, + "learning_rate": 2.2984739676840217e-05, + "loss": 0.3264, + "step": 42183 + }, + { + "epoch": 1.62, + "learning_rate": 2.2984098486791485e-05, + "loss": 0.248, + "step": 42184 + }, + { + "epoch": 1.62, + "learning_rate": 2.2983457296742757e-05, + "loss": 0.3147, + "step": 42185 + }, + { + "epoch": 1.62, + "learning_rate": 2.2982816106694025e-05, + "loss": 0.4683, + "step": 42186 + }, + { + "epoch": 1.62, + "learning_rate": 2.2982174916645293e-05, + "loss": 0.2668, + "step": 42187 + }, + { + "epoch": 1.62, + "learning_rate": 2.2981533726596562e-05, + "loss": 0.3594, + "step": 42188 + }, + { + "epoch": 1.62, + "learning_rate": 2.2980892536547833e-05, + "loss": 0.3208, + "step": 42189 + }, + { + "epoch": 1.62, + "learning_rate": 2.2980251346499105e-05, + "loss": 0.3457, + "step": 42190 + }, + { + "epoch": 1.62, + "learning_rate": 2.2979610156450374e-05, + "loss": 0.3664, + "step": 42191 + }, + { + "epoch": 1.62, + "learning_rate": 2.2978968966401642e-05, + "loss": 0.264, + "step": 42192 + }, + { + "epoch": 1.62, + "learning_rate": 2.2978327776352914e-05, + "loss": 0.3471, + "step": 42193 + }, + { + "epoch": 1.62, + "learning_rate": 2.2977686586304182e-05, + "loss": 0.2564, + "step": 42194 + }, + { + "epoch": 1.62, + "learning_rate": 2.297704539625545e-05, + "loss": 0.4077, + "step": 42195 + }, + { + "epoch": 1.62, + "learning_rate": 2.297640420620672e-05, + "loss": 0.3822, + "step": 42196 + }, + { + "epoch": 1.62, + "learning_rate": 2.297576301615799e-05, + "loss": 0.324, + "step": 42197 + }, + { + "epoch": 1.62, + "learning_rate": 2.2975121826109262e-05, + "loss": 0.4091, + "step": 42198 + }, + { + "epoch": 1.62, + "learning_rate": 2.297448063606053e-05, + "loss": 0.3151, + "step": 42199 + }, + { + "epoch": 1.62, + "learning_rate": 2.29738394460118e-05, + "loss": 0.4216, + "step": 42200 + }, + { + "epoch": 1.62, + "learning_rate": 2.297319825596307e-05, + "loss": 0.3006, + "step": 42201 + }, + { + "epoch": 1.62, + "learning_rate": 2.297255706591434e-05, + "loss": 0.5598, + "step": 42202 + }, + { + "epoch": 1.62, + "learning_rate": 2.2971915875865607e-05, + "loss": 0.4115, + "step": 42203 + }, + { + "epoch": 1.62, + "learning_rate": 2.2971274685816875e-05, + "loss": 0.3649, + "step": 42204 + }, + { + "epoch": 1.62, + "learning_rate": 2.2970633495768147e-05, + "loss": 0.314, + "step": 42205 + }, + { + "epoch": 1.62, + "learning_rate": 2.2969992305719415e-05, + "loss": 0.3757, + "step": 42206 + }, + { + "epoch": 1.62, + "learning_rate": 2.2969351115670687e-05, + "loss": 0.4807, + "step": 42207 + }, + { + "epoch": 1.62, + "learning_rate": 2.2968709925621955e-05, + "loss": 0.2911, + "step": 42208 + }, + { + "epoch": 1.62, + "learning_rate": 2.2968068735573227e-05, + "loss": 0.4839, + "step": 42209 + }, + { + "epoch": 1.62, + "learning_rate": 2.2967427545524495e-05, + "loss": 0.3274, + "step": 42210 + }, + { + "epoch": 1.62, + "learning_rate": 2.2966786355475764e-05, + "loss": 0.4276, + "step": 42211 + }, + { + "epoch": 1.62, + "learning_rate": 2.2966145165427032e-05, + "loss": 0.3577, + "step": 42212 + }, + { + "epoch": 1.62, + "learning_rate": 2.2965503975378304e-05, + "loss": 0.4709, + "step": 42213 + }, + { + "epoch": 1.62, + "learning_rate": 2.2964862785329572e-05, + "loss": 0.2386, + "step": 42214 + }, + { + "epoch": 1.62, + "learning_rate": 2.2964221595280844e-05, + "loss": 0.4014, + "step": 42215 + }, + { + "epoch": 1.62, + "learning_rate": 2.2963580405232112e-05, + "loss": 0.3409, + "step": 42216 + }, + { + "epoch": 1.62, + "learning_rate": 2.296293921518338e-05, + "loss": 0.2987, + "step": 42217 + }, + { + "epoch": 1.62, + "learning_rate": 2.2962298025134652e-05, + "loss": 0.3439, + "step": 42218 + }, + { + "epoch": 1.62, + "learning_rate": 2.296165683508592e-05, + "loss": 0.3914, + "step": 42219 + }, + { + "epoch": 1.62, + "learning_rate": 2.296101564503719e-05, + "loss": 0.3184, + "step": 42220 + }, + { + "epoch": 1.62, + "learning_rate": 2.296037445498846e-05, + "loss": 0.3233, + "step": 42221 + }, + { + "epoch": 1.62, + "learning_rate": 2.295973326493973e-05, + "loss": 0.402, + "step": 42222 + }, + { + "epoch": 1.62, + "learning_rate": 2.2959092074890997e-05, + "loss": 0.494, + "step": 42223 + }, + { + "epoch": 1.62, + "learning_rate": 2.295845088484227e-05, + "loss": 0.5019, + "step": 42224 + }, + { + "epoch": 1.62, + "learning_rate": 2.2957809694793537e-05, + "loss": 0.2993, + "step": 42225 + }, + { + "epoch": 1.62, + "learning_rate": 2.295716850474481e-05, + "loss": 0.2703, + "step": 42226 + }, + { + "epoch": 1.62, + "learning_rate": 2.2956527314696077e-05, + "loss": 0.3577, + "step": 42227 + }, + { + "epoch": 1.62, + "learning_rate": 2.2955886124647346e-05, + "loss": 0.4278, + "step": 42228 + }, + { + "epoch": 1.62, + "learning_rate": 2.2955244934598617e-05, + "loss": 0.3091, + "step": 42229 + }, + { + "epoch": 1.62, + "learning_rate": 2.2954603744549886e-05, + "loss": 0.3579, + "step": 42230 + }, + { + "epoch": 1.62, + "learning_rate": 2.2953962554501154e-05, + "loss": 0.3177, + "step": 42231 + }, + { + "epoch": 1.62, + "learning_rate": 2.2953321364452422e-05, + "loss": 0.3834, + "step": 42232 + }, + { + "epoch": 1.62, + "learning_rate": 2.2952680174403694e-05, + "loss": 0.3457, + "step": 42233 + }, + { + "epoch": 1.62, + "learning_rate": 2.2952038984354966e-05, + "loss": 0.4833, + "step": 42234 + }, + { + "epoch": 1.62, + "learning_rate": 2.2951397794306234e-05, + "loss": 0.4573, + "step": 42235 + }, + { + "epoch": 1.62, + "learning_rate": 2.2950756604257503e-05, + "loss": 0.3105, + "step": 42236 + }, + { + "epoch": 1.62, + "learning_rate": 2.2950115414208774e-05, + "loss": 0.3158, + "step": 42237 + }, + { + "epoch": 1.62, + "learning_rate": 2.2949474224160043e-05, + "loss": 0.3464, + "step": 42238 + }, + { + "epoch": 1.62, + "learning_rate": 2.294883303411131e-05, + "loss": 0.3574, + "step": 42239 + }, + { + "epoch": 1.62, + "learning_rate": 2.294819184406258e-05, + "loss": 0.2177, + "step": 42240 + }, + { + "epoch": 1.62, + "learning_rate": 2.294755065401385e-05, + "loss": 0.3343, + "step": 42241 + }, + { + "epoch": 1.62, + "learning_rate": 2.2946909463965123e-05, + "loss": 0.3812, + "step": 42242 + }, + { + "epoch": 1.62, + "learning_rate": 2.294626827391639e-05, + "loss": 0.2541, + "step": 42243 + }, + { + "epoch": 1.62, + "learning_rate": 2.294562708386766e-05, + "loss": 0.3299, + "step": 42244 + }, + { + "epoch": 1.62, + "learning_rate": 2.294498589381893e-05, + "loss": 0.3915, + "step": 42245 + }, + { + "epoch": 1.62, + "learning_rate": 2.29443447037702e-05, + "loss": 0.4427, + "step": 42246 + }, + { + "epoch": 1.62, + "learning_rate": 2.2943703513721468e-05, + "loss": 0.427, + "step": 42247 + }, + { + "epoch": 1.62, + "learning_rate": 2.2943062323672736e-05, + "loss": 0.3574, + "step": 42248 + }, + { + "epoch": 1.62, + "learning_rate": 2.2942421133624008e-05, + "loss": 0.3911, + "step": 42249 + }, + { + "epoch": 1.62, + "learning_rate": 2.2941779943575276e-05, + "loss": 0.3756, + "step": 42250 + }, + { + "epoch": 1.62, + "learning_rate": 2.2941138753526548e-05, + "loss": 0.3639, + "step": 42251 + }, + { + "epoch": 1.62, + "learning_rate": 2.2940497563477816e-05, + "loss": 0.4704, + "step": 42252 + }, + { + "epoch": 1.62, + "learning_rate": 2.2939856373429084e-05, + "loss": 0.4886, + "step": 42253 + }, + { + "epoch": 1.62, + "learning_rate": 2.2939215183380356e-05, + "loss": 0.2724, + "step": 42254 + }, + { + "epoch": 1.62, + "learning_rate": 2.2938573993331624e-05, + "loss": 0.5131, + "step": 42255 + }, + { + "epoch": 1.62, + "learning_rate": 2.2937932803282893e-05, + "loss": 0.3983, + "step": 42256 + }, + { + "epoch": 1.62, + "learning_rate": 2.2937291613234165e-05, + "loss": 0.4295, + "step": 42257 + }, + { + "epoch": 1.62, + "learning_rate": 2.2936650423185433e-05, + "loss": 0.3103, + "step": 42258 + }, + { + "epoch": 1.62, + "learning_rate": 2.29360092331367e-05, + "loss": 0.391, + "step": 42259 + }, + { + "epoch": 1.62, + "learning_rate": 2.2935368043087973e-05, + "loss": 0.2787, + "step": 42260 + }, + { + "epoch": 1.62, + "learning_rate": 2.293472685303924e-05, + "loss": 0.322, + "step": 42261 + }, + { + "epoch": 1.62, + "learning_rate": 2.2934085662990513e-05, + "loss": 0.4449, + "step": 42262 + }, + { + "epoch": 1.62, + "learning_rate": 2.293344447294178e-05, + "loss": 0.2796, + "step": 42263 + }, + { + "epoch": 1.62, + "learning_rate": 2.293280328289305e-05, + "loss": 0.3505, + "step": 42264 + }, + { + "epoch": 1.62, + "learning_rate": 2.293216209284432e-05, + "loss": 0.2601, + "step": 42265 + }, + { + "epoch": 1.62, + "learning_rate": 2.293152090279559e-05, + "loss": 0.2235, + "step": 42266 + }, + { + "epoch": 1.63, + "learning_rate": 2.2930879712746858e-05, + "loss": 0.3918, + "step": 42267 + }, + { + "epoch": 1.63, + "learning_rate": 2.2930238522698126e-05, + "loss": 0.4833, + "step": 42268 + }, + { + "epoch": 1.63, + "learning_rate": 2.2929597332649398e-05, + "loss": 0.3216, + "step": 42269 + }, + { + "epoch": 1.63, + "learning_rate": 2.292895614260067e-05, + "loss": 0.2552, + "step": 42270 + }, + { + "epoch": 1.63, + "learning_rate": 2.2928314952551938e-05, + "loss": 0.3687, + "step": 42271 + }, + { + "epoch": 1.63, + "learning_rate": 2.2927673762503206e-05, + "loss": 0.2503, + "step": 42272 + }, + { + "epoch": 1.63, + "learning_rate": 2.2927032572454478e-05, + "loss": 0.3527, + "step": 42273 + }, + { + "epoch": 1.63, + "learning_rate": 2.2926391382405746e-05, + "loss": 0.367, + "step": 42274 + }, + { + "epoch": 1.63, + "learning_rate": 2.2925750192357015e-05, + "loss": 0.4089, + "step": 42275 + }, + { + "epoch": 1.63, + "learning_rate": 2.2925109002308283e-05, + "loss": 0.4353, + "step": 42276 + }, + { + "epoch": 1.63, + "learning_rate": 2.2924467812259555e-05, + "loss": 0.3312, + "step": 42277 + }, + { + "epoch": 1.63, + "learning_rate": 2.2923826622210827e-05, + "loss": 0.4394, + "step": 42278 + }, + { + "epoch": 1.63, + "learning_rate": 2.2923185432162095e-05, + "loss": 0.3155, + "step": 42279 + }, + { + "epoch": 1.63, + "learning_rate": 2.2922544242113363e-05, + "loss": 0.3019, + "step": 42280 + }, + { + "epoch": 1.63, + "learning_rate": 2.2921903052064635e-05, + "loss": 0.5669, + "step": 42281 + }, + { + "epoch": 1.63, + "learning_rate": 2.2921261862015903e-05, + "loss": 0.3031, + "step": 42282 + }, + { + "epoch": 1.63, + "learning_rate": 2.292062067196717e-05, + "loss": 0.37, + "step": 42283 + }, + { + "epoch": 1.63, + "learning_rate": 2.291997948191844e-05, + "loss": 0.3547, + "step": 42284 + }, + { + "epoch": 1.63, + "learning_rate": 2.291933829186971e-05, + "loss": 0.3906, + "step": 42285 + }, + { + "epoch": 1.63, + "learning_rate": 2.291869710182098e-05, + "loss": 0.3769, + "step": 42286 + }, + { + "epoch": 1.63, + "learning_rate": 2.291805591177225e-05, + "loss": 0.3292, + "step": 42287 + }, + { + "epoch": 1.63, + "learning_rate": 2.291741472172352e-05, + "loss": 0.3483, + "step": 42288 + }, + { + "epoch": 1.63, + "learning_rate": 2.291677353167479e-05, + "loss": 0.4771, + "step": 42289 + }, + { + "epoch": 1.63, + "learning_rate": 2.291613234162606e-05, + "loss": 0.3801, + "step": 42290 + }, + { + "epoch": 1.63, + "learning_rate": 2.291549115157733e-05, + "loss": 0.3759, + "step": 42291 + }, + { + "epoch": 1.63, + "learning_rate": 2.2914849961528597e-05, + "loss": 0.3817, + "step": 42292 + }, + { + "epoch": 1.63, + "learning_rate": 2.291420877147987e-05, + "loss": 0.3145, + "step": 42293 + }, + { + "epoch": 1.63, + "learning_rate": 2.2913567581431137e-05, + "loss": 0.3943, + "step": 42294 + }, + { + "epoch": 1.63, + "learning_rate": 2.2912926391382405e-05, + "loss": 0.2858, + "step": 42295 + }, + { + "epoch": 1.63, + "learning_rate": 2.2912285201333677e-05, + "loss": 0.3016, + "step": 42296 + }, + { + "epoch": 1.63, + "learning_rate": 2.2911644011284945e-05, + "loss": 0.4415, + "step": 42297 + }, + { + "epoch": 1.63, + "learning_rate": 2.2911002821236217e-05, + "loss": 0.4578, + "step": 42298 + }, + { + "epoch": 1.63, + "learning_rate": 2.2910361631187485e-05, + "loss": 0.3742, + "step": 42299 + }, + { + "epoch": 1.63, + "learning_rate": 2.2909720441138753e-05, + "loss": 0.2917, + "step": 42300 + }, + { + "epoch": 1.63, + "learning_rate": 2.2909079251090025e-05, + "loss": 0.2921, + "step": 42301 + }, + { + "epoch": 1.63, + "learning_rate": 2.2908438061041294e-05, + "loss": 0.3585, + "step": 42302 + }, + { + "epoch": 1.63, + "learning_rate": 2.2907796870992562e-05, + "loss": 0.3206, + "step": 42303 + }, + { + "epoch": 1.63, + "learning_rate": 2.290715568094383e-05, + "loss": 0.3402, + "step": 42304 + }, + { + "epoch": 1.63, + "learning_rate": 2.2906514490895102e-05, + "loss": 0.3104, + "step": 42305 + }, + { + "epoch": 1.63, + "learning_rate": 2.2905873300846374e-05, + "loss": 0.4231, + "step": 42306 + }, + { + "epoch": 1.63, + "learning_rate": 2.2905232110797642e-05, + "loss": 0.2289, + "step": 42307 + }, + { + "epoch": 1.63, + "learning_rate": 2.290459092074891e-05, + "loss": 0.3406, + "step": 42308 + }, + { + "epoch": 1.63, + "learning_rate": 2.2903949730700182e-05, + "loss": 0.2938, + "step": 42309 + }, + { + "epoch": 1.63, + "learning_rate": 2.290330854065145e-05, + "loss": 0.3573, + "step": 42310 + }, + { + "epoch": 1.63, + "learning_rate": 2.290266735060272e-05, + "loss": 0.3529, + "step": 42311 + }, + { + "epoch": 1.63, + "learning_rate": 2.2902026160553987e-05, + "loss": 0.3175, + "step": 42312 + }, + { + "epoch": 1.63, + "learning_rate": 2.290138497050526e-05, + "loss": 0.3102, + "step": 42313 + }, + { + "epoch": 1.63, + "learning_rate": 2.290074378045653e-05, + "loss": 0.269, + "step": 42314 + }, + { + "epoch": 1.63, + "learning_rate": 2.29001025904078e-05, + "loss": 0.3292, + "step": 42315 + }, + { + "epoch": 1.63, + "learning_rate": 2.2899461400359067e-05, + "loss": 0.4126, + "step": 42316 + }, + { + "epoch": 1.63, + "learning_rate": 2.289882021031034e-05, + "loss": 0.4056, + "step": 42317 + }, + { + "epoch": 1.63, + "learning_rate": 2.2898179020261607e-05, + "loss": 0.3425, + "step": 42318 + }, + { + "epoch": 1.63, + "learning_rate": 2.2897537830212875e-05, + "loss": 0.3369, + "step": 42319 + }, + { + "epoch": 1.63, + "learning_rate": 2.2896896640164144e-05, + "loss": 0.3095, + "step": 42320 + }, + { + "epoch": 1.63, + "learning_rate": 2.2896255450115415e-05, + "loss": 0.3765, + "step": 42321 + }, + { + "epoch": 1.63, + "learning_rate": 2.2895614260066687e-05, + "loss": 0.394, + "step": 42322 + }, + { + "epoch": 1.63, + "learning_rate": 2.2894973070017956e-05, + "loss": 0.2991, + "step": 42323 + }, + { + "epoch": 1.63, + "learning_rate": 2.2894331879969224e-05, + "loss": 0.2958, + "step": 42324 + }, + { + "epoch": 1.63, + "learning_rate": 2.2893690689920496e-05, + "loss": 0.3489, + "step": 42325 + }, + { + "epoch": 1.63, + "learning_rate": 2.2893049499871764e-05, + "loss": 0.488, + "step": 42326 + }, + { + "epoch": 1.63, + "learning_rate": 2.2892408309823032e-05, + "loss": 0.3853, + "step": 42327 + }, + { + "epoch": 1.63, + "learning_rate": 2.28917671197743e-05, + "loss": 0.34, + "step": 42328 + }, + { + "epoch": 1.63, + "learning_rate": 2.2891125929725572e-05, + "loss": 0.2722, + "step": 42329 + }, + { + "epoch": 1.63, + "learning_rate": 2.289048473967684e-05, + "loss": 0.3339, + "step": 42330 + }, + { + "epoch": 1.63, + "learning_rate": 2.2889843549628112e-05, + "loss": 0.4056, + "step": 42331 + }, + { + "epoch": 1.63, + "learning_rate": 2.288920235957938e-05, + "loss": 0.3031, + "step": 42332 + }, + { + "epoch": 1.63, + "learning_rate": 2.2888561169530652e-05, + "loss": 0.2612, + "step": 42333 + }, + { + "epoch": 1.63, + "learning_rate": 2.288791997948192e-05, + "loss": 0.4164, + "step": 42334 + }, + { + "epoch": 1.63, + "learning_rate": 2.288727878943319e-05, + "loss": 0.2853, + "step": 42335 + }, + { + "epoch": 1.63, + "learning_rate": 2.2886637599384457e-05, + "loss": 0.3549, + "step": 42336 + }, + { + "epoch": 1.63, + "learning_rate": 2.288599640933573e-05, + "loss": 0.1674, + "step": 42337 + }, + { + "epoch": 1.63, + "learning_rate": 2.2885355219286997e-05, + "loss": 0.3257, + "step": 42338 + }, + { + "epoch": 1.63, + "learning_rate": 2.2884714029238266e-05, + "loss": 0.4444, + "step": 42339 + }, + { + "epoch": 1.63, + "learning_rate": 2.2884072839189537e-05, + "loss": 0.2953, + "step": 42340 + }, + { + "epoch": 1.63, + "learning_rate": 2.2883431649140806e-05, + "loss": 0.4476, + "step": 42341 + }, + { + "epoch": 1.63, + "learning_rate": 2.2882790459092077e-05, + "loss": 0.4057, + "step": 42342 + }, + { + "epoch": 1.63, + "learning_rate": 2.2882149269043346e-05, + "loss": 0.2957, + "step": 42343 + }, + { + "epoch": 1.63, + "learning_rate": 2.2881508078994614e-05, + "loss": 0.3441, + "step": 42344 + }, + { + "epoch": 1.63, + "learning_rate": 2.2880866888945886e-05, + "loss": 0.4232, + "step": 42345 + }, + { + "epoch": 1.63, + "learning_rate": 2.2880225698897154e-05, + "loss": 0.2729, + "step": 42346 + }, + { + "epoch": 1.63, + "learning_rate": 2.2879584508848423e-05, + "loss": 0.3532, + "step": 42347 + }, + { + "epoch": 1.63, + "learning_rate": 2.287894331879969e-05, + "loss": 0.4045, + "step": 42348 + }, + { + "epoch": 1.63, + "learning_rate": 2.2878302128750963e-05, + "loss": 0.2659, + "step": 42349 + }, + { + "epoch": 1.63, + "learning_rate": 2.2877660938702234e-05, + "loss": 0.3315, + "step": 42350 + }, + { + "epoch": 1.63, + "learning_rate": 2.2877019748653503e-05, + "loss": 0.2827, + "step": 42351 + }, + { + "epoch": 1.63, + "learning_rate": 2.287637855860477e-05, + "loss": 0.2709, + "step": 42352 + }, + { + "epoch": 1.63, + "learning_rate": 2.2875737368556043e-05, + "loss": 0.3515, + "step": 42353 + }, + { + "epoch": 1.63, + "learning_rate": 2.287509617850731e-05, + "loss": 0.3467, + "step": 42354 + }, + { + "epoch": 1.63, + "learning_rate": 2.287445498845858e-05, + "loss": 0.3392, + "step": 42355 + }, + { + "epoch": 1.63, + "learning_rate": 2.2873813798409848e-05, + "loss": 0.4094, + "step": 42356 + }, + { + "epoch": 1.63, + "learning_rate": 2.287317260836112e-05, + "loss": 0.3182, + "step": 42357 + }, + { + "epoch": 1.63, + "learning_rate": 2.287253141831239e-05, + "loss": 0.5074, + "step": 42358 + }, + { + "epoch": 1.63, + "learning_rate": 2.287189022826366e-05, + "loss": 0.3723, + "step": 42359 + }, + { + "epoch": 1.63, + "learning_rate": 2.2871249038214928e-05, + "loss": 0.436, + "step": 42360 + }, + { + "epoch": 1.63, + "learning_rate": 2.28706078481662e-05, + "loss": 0.3211, + "step": 42361 + }, + { + "epoch": 1.63, + "learning_rate": 2.2869966658117468e-05, + "loss": 0.2954, + "step": 42362 + }, + { + "epoch": 1.63, + "learning_rate": 2.2869325468068736e-05, + "loss": 0.3917, + "step": 42363 + }, + { + "epoch": 1.63, + "learning_rate": 2.2868684278020004e-05, + "loss": 0.3342, + "step": 42364 + }, + { + "epoch": 1.63, + "learning_rate": 2.2868043087971276e-05, + "loss": 0.2837, + "step": 42365 + }, + { + "epoch": 1.63, + "learning_rate": 2.2867401897922544e-05, + "loss": 0.291, + "step": 42366 + }, + { + "epoch": 1.63, + "learning_rate": 2.2866760707873816e-05, + "loss": 0.3529, + "step": 42367 + }, + { + "epoch": 1.63, + "learning_rate": 2.2866119517825085e-05, + "loss": 0.3475, + "step": 42368 + }, + { + "epoch": 1.63, + "learning_rate": 2.2865478327776356e-05, + "loss": 0.3526, + "step": 42369 + }, + { + "epoch": 1.63, + "learning_rate": 2.2864837137727625e-05, + "loss": 0.268, + "step": 42370 + }, + { + "epoch": 1.63, + "learning_rate": 2.2864195947678893e-05, + "loss": 0.4197, + "step": 42371 + }, + { + "epoch": 1.63, + "learning_rate": 2.286355475763016e-05, + "loss": 0.3884, + "step": 42372 + }, + { + "epoch": 1.63, + "learning_rate": 2.2862913567581433e-05, + "loss": 0.4893, + "step": 42373 + }, + { + "epoch": 1.63, + "learning_rate": 2.28622723775327e-05, + "loss": 0.4015, + "step": 42374 + }, + { + "epoch": 1.63, + "learning_rate": 2.286163118748397e-05, + "loss": 0.3806, + "step": 42375 + }, + { + "epoch": 1.63, + "learning_rate": 2.286098999743524e-05, + "loss": 0.4602, + "step": 42376 + }, + { + "epoch": 1.63, + "learning_rate": 2.286034880738651e-05, + "loss": 0.3518, + "step": 42377 + }, + { + "epoch": 1.63, + "learning_rate": 2.285970761733778e-05, + "loss": 0.2236, + "step": 42378 + }, + { + "epoch": 1.63, + "learning_rate": 2.285906642728905e-05, + "loss": 0.2145, + "step": 42379 + }, + { + "epoch": 1.63, + "learning_rate": 2.2858425237240318e-05, + "loss": 0.2852, + "step": 42380 + }, + { + "epoch": 1.63, + "learning_rate": 2.285778404719159e-05, + "loss": 0.4042, + "step": 42381 + }, + { + "epoch": 1.63, + "learning_rate": 2.2857142857142858e-05, + "loss": 0.2473, + "step": 42382 + }, + { + "epoch": 1.63, + "learning_rate": 2.2856501667094126e-05, + "loss": 0.3503, + "step": 42383 + }, + { + "epoch": 1.63, + "learning_rate": 2.2855860477045395e-05, + "loss": 0.4717, + "step": 42384 + }, + { + "epoch": 1.63, + "learning_rate": 2.2855219286996666e-05, + "loss": 0.3159, + "step": 42385 + }, + { + "epoch": 1.63, + "learning_rate": 2.2854578096947938e-05, + "loss": 0.3474, + "step": 42386 + }, + { + "epoch": 1.63, + "learning_rate": 2.2853936906899206e-05, + "loss": 0.3474, + "step": 42387 + }, + { + "epoch": 1.63, + "learning_rate": 2.2853295716850475e-05, + "loss": 0.2606, + "step": 42388 + }, + { + "epoch": 1.63, + "learning_rate": 2.2852654526801747e-05, + "loss": 0.3472, + "step": 42389 + }, + { + "epoch": 1.63, + "learning_rate": 2.2852013336753015e-05, + "loss": 0.3784, + "step": 42390 + }, + { + "epoch": 1.63, + "learning_rate": 2.2851372146704283e-05, + "loss": 0.4222, + "step": 42391 + }, + { + "epoch": 1.63, + "learning_rate": 2.285073095665555e-05, + "loss": 0.3721, + "step": 42392 + }, + { + "epoch": 1.63, + "learning_rate": 2.2850089766606823e-05, + "loss": 0.3429, + "step": 42393 + }, + { + "epoch": 1.63, + "learning_rate": 2.2849448576558095e-05, + "loss": 0.2111, + "step": 42394 + }, + { + "epoch": 1.63, + "learning_rate": 2.2848807386509363e-05, + "loss": 0.3713, + "step": 42395 + }, + { + "epoch": 1.63, + "learning_rate": 2.284816619646063e-05, + "loss": 0.3975, + "step": 42396 + }, + { + "epoch": 1.63, + "learning_rate": 2.2847525006411903e-05, + "loss": 0.402, + "step": 42397 + }, + { + "epoch": 1.63, + "learning_rate": 2.284688381636317e-05, + "loss": 0.2944, + "step": 42398 + }, + { + "epoch": 1.63, + "learning_rate": 2.284624262631444e-05, + "loss": 0.2826, + "step": 42399 + }, + { + "epoch": 1.63, + "learning_rate": 2.2845601436265708e-05, + "loss": 0.3335, + "step": 42400 + }, + { + "epoch": 1.63, + "learning_rate": 2.284496024621698e-05, + "loss": 0.2784, + "step": 42401 + }, + { + "epoch": 1.63, + "learning_rate": 2.2844319056168252e-05, + "loss": 0.4796, + "step": 42402 + }, + { + "epoch": 1.63, + "learning_rate": 2.284367786611952e-05, + "loss": 0.235, + "step": 42403 + }, + { + "epoch": 1.63, + "learning_rate": 2.284303667607079e-05, + "loss": 0.3319, + "step": 42404 + }, + { + "epoch": 1.63, + "learning_rate": 2.284239548602206e-05, + "loss": 0.3022, + "step": 42405 + }, + { + "epoch": 1.63, + "learning_rate": 2.284175429597333e-05, + "loss": 0.4965, + "step": 42406 + }, + { + "epoch": 1.63, + "learning_rate": 2.2841113105924597e-05, + "loss": 0.4183, + "step": 42407 + }, + { + "epoch": 1.63, + "learning_rate": 2.2840471915875865e-05, + "loss": 0.2969, + "step": 42408 + }, + { + "epoch": 1.63, + "learning_rate": 2.2839830725827137e-05, + "loss": 0.4237, + "step": 42409 + }, + { + "epoch": 1.63, + "learning_rate": 2.2839189535778405e-05, + "loss": 0.2604, + "step": 42410 + }, + { + "epoch": 1.63, + "learning_rate": 2.2838548345729677e-05, + "loss": 0.4413, + "step": 42411 + }, + { + "epoch": 1.63, + "learning_rate": 2.2837907155680945e-05, + "loss": 0.3224, + "step": 42412 + }, + { + "epoch": 1.63, + "learning_rate": 2.2837265965632217e-05, + "loss": 0.2683, + "step": 42413 + }, + { + "epoch": 1.63, + "learning_rate": 2.2836624775583485e-05, + "loss": 0.4018, + "step": 42414 + }, + { + "epoch": 1.63, + "learning_rate": 2.2835983585534754e-05, + "loss": 0.3576, + "step": 42415 + }, + { + "epoch": 1.63, + "learning_rate": 2.2835342395486022e-05, + "loss": 0.4398, + "step": 42416 + }, + { + "epoch": 1.63, + "learning_rate": 2.2834701205437294e-05, + "loss": 0.4658, + "step": 42417 + }, + { + "epoch": 1.63, + "learning_rate": 2.2834060015388562e-05, + "loss": 0.3561, + "step": 42418 + }, + { + "epoch": 1.63, + "learning_rate": 2.283341882533983e-05, + "loss": 0.3981, + "step": 42419 + }, + { + "epoch": 1.63, + "learning_rate": 2.2832777635291102e-05, + "loss": 0.4751, + "step": 42420 + }, + { + "epoch": 1.63, + "learning_rate": 2.283213644524237e-05, + "loss": 0.3216, + "step": 42421 + }, + { + "epoch": 1.63, + "learning_rate": 2.2831495255193642e-05, + "loss": 0.4772, + "step": 42422 + }, + { + "epoch": 1.63, + "learning_rate": 2.283085406514491e-05, + "loss": 0.2893, + "step": 42423 + }, + { + "epoch": 1.63, + "learning_rate": 2.283021287509618e-05, + "loss": 0.2333, + "step": 42424 + }, + { + "epoch": 1.63, + "learning_rate": 2.282957168504745e-05, + "loss": 0.283, + "step": 42425 + }, + { + "epoch": 1.63, + "learning_rate": 2.282893049499872e-05, + "loss": 0.5324, + "step": 42426 + }, + { + "epoch": 1.63, + "learning_rate": 2.2828289304949987e-05, + "loss": 0.3881, + "step": 42427 + }, + { + "epoch": 1.63, + "learning_rate": 2.2827648114901255e-05, + "loss": 0.1991, + "step": 42428 + }, + { + "epoch": 1.63, + "learning_rate": 2.2827006924852527e-05, + "loss": 0.4313, + "step": 42429 + }, + { + "epoch": 1.63, + "learning_rate": 2.28263657348038e-05, + "loss": 0.3273, + "step": 42430 + }, + { + "epoch": 1.63, + "learning_rate": 2.2825724544755067e-05, + "loss": 0.3614, + "step": 42431 + }, + { + "epoch": 1.63, + "learning_rate": 2.2825083354706335e-05, + "loss": 0.3939, + "step": 42432 + }, + { + "epoch": 1.63, + "learning_rate": 2.2824442164657607e-05, + "loss": 0.4059, + "step": 42433 + }, + { + "epoch": 1.63, + "learning_rate": 2.2823800974608876e-05, + "loss": 0.3319, + "step": 42434 + }, + { + "epoch": 1.63, + "learning_rate": 2.2823159784560144e-05, + "loss": 0.3934, + "step": 42435 + }, + { + "epoch": 1.63, + "learning_rate": 2.2822518594511412e-05, + "loss": 0.3588, + "step": 42436 + }, + { + "epoch": 1.63, + "learning_rate": 2.2821877404462684e-05, + "loss": 0.2478, + "step": 42437 + }, + { + "epoch": 1.63, + "learning_rate": 2.2821236214413956e-05, + "loss": 0.4044, + "step": 42438 + }, + { + "epoch": 1.63, + "learning_rate": 2.2820595024365224e-05, + "loss": 0.3221, + "step": 42439 + }, + { + "epoch": 1.63, + "learning_rate": 2.2819953834316492e-05, + "loss": 0.2265, + "step": 42440 + }, + { + "epoch": 1.63, + "learning_rate": 2.2819312644267764e-05, + "loss": 0.3706, + "step": 42441 + }, + { + "epoch": 1.63, + "learning_rate": 2.2818671454219032e-05, + "loss": 0.4662, + "step": 42442 + }, + { + "epoch": 1.63, + "learning_rate": 2.28180302641703e-05, + "loss": 0.1948, + "step": 42443 + }, + { + "epoch": 1.63, + "learning_rate": 2.281738907412157e-05, + "loss": 0.3378, + "step": 42444 + }, + { + "epoch": 1.63, + "learning_rate": 2.281674788407284e-05, + "loss": 0.2169, + "step": 42445 + }, + { + "epoch": 1.63, + "learning_rate": 2.281610669402411e-05, + "loss": 0.4038, + "step": 42446 + }, + { + "epoch": 1.63, + "learning_rate": 2.281546550397538e-05, + "loss": 0.2243, + "step": 42447 + }, + { + "epoch": 1.63, + "learning_rate": 2.281482431392665e-05, + "loss": 0.4713, + "step": 42448 + }, + { + "epoch": 1.63, + "learning_rate": 2.281418312387792e-05, + "loss": 0.2274, + "step": 42449 + }, + { + "epoch": 1.63, + "learning_rate": 2.281354193382919e-05, + "loss": 0.4009, + "step": 42450 + }, + { + "epoch": 1.63, + "learning_rate": 2.2812900743780457e-05, + "loss": 0.2517, + "step": 42451 + }, + { + "epoch": 1.63, + "learning_rate": 2.2812259553731726e-05, + "loss": 0.4258, + "step": 42452 + }, + { + "epoch": 1.63, + "learning_rate": 2.2811618363682997e-05, + "loss": 0.2843, + "step": 42453 + }, + { + "epoch": 1.63, + "learning_rate": 2.2810977173634266e-05, + "loss": 0.2694, + "step": 42454 + }, + { + "epoch": 1.63, + "learning_rate": 2.2810335983585534e-05, + "loss": 0.2528, + "step": 42455 + }, + { + "epoch": 1.63, + "learning_rate": 2.2809694793536806e-05, + "loss": 0.2575, + "step": 42456 + }, + { + "epoch": 1.63, + "learning_rate": 2.2809053603488074e-05, + "loss": 0.4442, + "step": 42457 + }, + { + "epoch": 1.63, + "learning_rate": 2.2808412413439346e-05, + "loss": 0.3078, + "step": 42458 + }, + { + "epoch": 1.63, + "learning_rate": 2.2807771223390614e-05, + "loss": 0.331, + "step": 42459 + }, + { + "epoch": 1.63, + "learning_rate": 2.2807130033341883e-05, + "loss": 0.2784, + "step": 42460 + }, + { + "epoch": 1.63, + "learning_rate": 2.2806488843293154e-05, + "loss": 0.3719, + "step": 42461 + }, + { + "epoch": 1.63, + "learning_rate": 2.2805847653244423e-05, + "loss": 0.3465, + "step": 42462 + }, + { + "epoch": 1.63, + "learning_rate": 2.280520646319569e-05, + "loss": 0.3198, + "step": 42463 + }, + { + "epoch": 1.63, + "learning_rate": 2.280456527314696e-05, + "loss": 0.3491, + "step": 42464 + }, + { + "epoch": 1.63, + "learning_rate": 2.280392408309823e-05, + "loss": 0.2519, + "step": 42465 + }, + { + "epoch": 1.63, + "learning_rate": 2.2803282893049503e-05, + "loss": 0.2469, + "step": 42466 + }, + { + "epoch": 1.63, + "learning_rate": 2.280264170300077e-05, + "loss": 0.2701, + "step": 42467 + }, + { + "epoch": 1.63, + "learning_rate": 2.280200051295204e-05, + "loss": 0.3535, + "step": 42468 + }, + { + "epoch": 1.63, + "learning_rate": 2.280135932290331e-05, + "loss": 0.2806, + "step": 42469 + }, + { + "epoch": 1.63, + "learning_rate": 2.280071813285458e-05, + "loss": 0.3846, + "step": 42470 + }, + { + "epoch": 1.63, + "learning_rate": 2.2800076942805848e-05, + "loss": 0.4173, + "step": 42471 + }, + { + "epoch": 1.63, + "learning_rate": 2.2799435752757116e-05, + "loss": 0.4054, + "step": 42472 + }, + { + "epoch": 1.63, + "learning_rate": 2.2798794562708388e-05, + "loss": 0.5272, + "step": 42473 + }, + { + "epoch": 1.63, + "learning_rate": 2.279815337265966e-05, + "loss": 0.4186, + "step": 42474 + }, + { + "epoch": 1.63, + "learning_rate": 2.2797512182610928e-05, + "loss": 0.4335, + "step": 42475 + }, + { + "epoch": 1.63, + "learning_rate": 2.2796870992562196e-05, + "loss": 0.3142, + "step": 42476 + }, + { + "epoch": 1.63, + "learning_rate": 2.2796229802513468e-05, + "loss": 0.2645, + "step": 42477 + }, + { + "epoch": 1.63, + "learning_rate": 2.2795588612464736e-05, + "loss": 0.3311, + "step": 42478 + }, + { + "epoch": 1.63, + "learning_rate": 2.2794947422416004e-05, + "loss": 0.3511, + "step": 42479 + }, + { + "epoch": 1.63, + "learning_rate": 2.2794306232367273e-05, + "loss": 0.2983, + "step": 42480 + }, + { + "epoch": 1.63, + "learning_rate": 2.2793665042318545e-05, + "loss": 0.345, + "step": 42481 + }, + { + "epoch": 1.63, + "learning_rate": 2.2793023852269813e-05, + "loss": 0.3223, + "step": 42482 + }, + { + "epoch": 1.63, + "learning_rate": 2.2792382662221085e-05, + "loss": 0.4068, + "step": 42483 + }, + { + "epoch": 1.63, + "learning_rate": 2.2791741472172353e-05, + "loss": 0.3782, + "step": 42484 + }, + { + "epoch": 1.63, + "learning_rate": 2.2791100282123625e-05, + "loss": 0.4863, + "step": 42485 + }, + { + "epoch": 1.63, + "learning_rate": 2.2790459092074893e-05, + "loss": 0.2604, + "step": 42486 + }, + { + "epoch": 1.63, + "learning_rate": 2.278981790202616e-05, + "loss": 0.3619, + "step": 42487 + }, + { + "epoch": 1.63, + "learning_rate": 2.278917671197743e-05, + "loss": 0.3125, + "step": 42488 + }, + { + "epoch": 1.63, + "learning_rate": 2.27885355219287e-05, + "loss": 0.2524, + "step": 42489 + }, + { + "epoch": 1.63, + "learning_rate": 2.278789433187997e-05, + "loss": 0.3495, + "step": 42490 + }, + { + "epoch": 1.63, + "learning_rate": 2.2787253141831238e-05, + "loss": 0.3294, + "step": 42491 + }, + { + "epoch": 1.63, + "learning_rate": 2.278661195178251e-05, + "loss": 0.2944, + "step": 42492 + }, + { + "epoch": 1.63, + "learning_rate": 2.278597076173378e-05, + "loss": 0.2853, + "step": 42493 + }, + { + "epoch": 1.63, + "learning_rate": 2.278532957168505e-05, + "loss": 0.388, + "step": 42494 + }, + { + "epoch": 1.63, + "learning_rate": 2.2784688381636318e-05, + "loss": 0.3397, + "step": 42495 + }, + { + "epoch": 1.63, + "learning_rate": 2.2784047191587586e-05, + "loss": 0.364, + "step": 42496 + }, + { + "epoch": 1.63, + "learning_rate": 2.2783406001538858e-05, + "loss": 0.3209, + "step": 42497 + }, + { + "epoch": 1.63, + "learning_rate": 2.2782764811490126e-05, + "loss": 0.292, + "step": 42498 + }, + { + "epoch": 1.63, + "learning_rate": 2.2782123621441395e-05, + "loss": 0.2009, + "step": 42499 + }, + { + "epoch": 1.63, + "learning_rate": 2.2781482431392666e-05, + "loss": 0.4651, + "step": 42500 + }, + { + "epoch": 1.63, + "learning_rate": 2.2780841241343935e-05, + "loss": 0.3577, + "step": 42501 + }, + { + "epoch": 1.63, + "learning_rate": 2.2780200051295207e-05, + "loss": 0.3838, + "step": 42502 + }, + { + "epoch": 1.63, + "learning_rate": 2.2779558861246475e-05, + "loss": 0.3495, + "step": 42503 + }, + { + "epoch": 1.63, + "learning_rate": 2.2778917671197743e-05, + "loss": 0.351, + "step": 42504 + }, + { + "epoch": 1.63, + "learning_rate": 2.2778276481149015e-05, + "loss": 0.5578, + "step": 42505 + }, + { + "epoch": 1.63, + "learning_rate": 2.2777635291100283e-05, + "loss": 0.4748, + "step": 42506 + }, + { + "epoch": 1.63, + "learning_rate": 2.277699410105155e-05, + "loss": 0.3541, + "step": 42507 + }, + { + "epoch": 1.63, + "learning_rate": 2.277635291100282e-05, + "loss": 0.2411, + "step": 42508 + }, + { + "epoch": 1.63, + "learning_rate": 2.277571172095409e-05, + "loss": 0.3438, + "step": 42509 + }, + { + "epoch": 1.63, + "learning_rate": 2.2775070530905363e-05, + "loss": 0.3606, + "step": 42510 + }, + { + "epoch": 1.63, + "learning_rate": 2.277442934085663e-05, + "loss": 0.3002, + "step": 42511 + }, + { + "epoch": 1.63, + "learning_rate": 2.27737881508079e-05, + "loss": 0.2748, + "step": 42512 + }, + { + "epoch": 1.63, + "learning_rate": 2.2773146960759172e-05, + "loss": 0.3841, + "step": 42513 + }, + { + "epoch": 1.63, + "learning_rate": 2.277250577071044e-05, + "loss": 0.221, + "step": 42514 + }, + { + "epoch": 1.63, + "learning_rate": 2.277186458066171e-05, + "loss": 0.3657, + "step": 42515 + }, + { + "epoch": 1.63, + "learning_rate": 2.2771223390612977e-05, + "loss": 0.4141, + "step": 42516 + }, + { + "epoch": 1.63, + "learning_rate": 2.277058220056425e-05, + "loss": 0.3217, + "step": 42517 + }, + { + "epoch": 1.63, + "learning_rate": 2.276994101051552e-05, + "loss": 0.4079, + "step": 42518 + }, + { + "epoch": 1.63, + "learning_rate": 2.276929982046679e-05, + "loss": 0.4347, + "step": 42519 + }, + { + "epoch": 1.63, + "learning_rate": 2.2768658630418057e-05, + "loss": 0.3065, + "step": 42520 + }, + { + "epoch": 1.63, + "learning_rate": 2.276801744036933e-05, + "loss": 0.3423, + "step": 42521 + }, + { + "epoch": 1.63, + "learning_rate": 2.2767376250320597e-05, + "loss": 0.3301, + "step": 42522 + }, + { + "epoch": 1.63, + "learning_rate": 2.2766735060271865e-05, + "loss": 0.3156, + "step": 42523 + }, + { + "epoch": 1.63, + "learning_rate": 2.2766093870223133e-05, + "loss": 0.3747, + "step": 42524 + }, + { + "epoch": 1.63, + "learning_rate": 2.2765452680174405e-05, + "loss": 0.3745, + "step": 42525 + }, + { + "epoch": 1.63, + "learning_rate": 2.2764811490125674e-05, + "loss": 0.3528, + "step": 42526 + }, + { + "epoch": 1.64, + "learning_rate": 2.2764170300076945e-05, + "loss": 0.3706, + "step": 42527 + }, + { + "epoch": 1.64, + "learning_rate": 2.2763529110028214e-05, + "loss": 0.3368, + "step": 42528 + }, + { + "epoch": 1.64, + "learning_rate": 2.2762887919979485e-05, + "loss": 0.4216, + "step": 42529 + }, + { + "epoch": 1.64, + "learning_rate": 2.2762246729930754e-05, + "loss": 0.3453, + "step": 42530 + }, + { + "epoch": 1.64, + "learning_rate": 2.2761605539882022e-05, + "loss": 0.4315, + "step": 42531 + }, + { + "epoch": 1.64, + "learning_rate": 2.276096434983329e-05, + "loss": 0.397, + "step": 42532 + }, + { + "epoch": 1.64, + "learning_rate": 2.2760323159784562e-05, + "loss": 0.2866, + "step": 42533 + }, + { + "epoch": 1.64, + "learning_rate": 2.275968196973583e-05, + "loss": 0.4621, + "step": 42534 + }, + { + "epoch": 1.64, + "learning_rate": 2.27590407796871e-05, + "loss": 0.3688, + "step": 42535 + }, + { + "epoch": 1.64, + "learning_rate": 2.275839958963837e-05, + "loss": 0.3789, + "step": 42536 + }, + { + "epoch": 1.64, + "learning_rate": 2.2757758399589642e-05, + "loss": 0.3882, + "step": 42537 + }, + { + "epoch": 1.64, + "learning_rate": 2.275711720954091e-05, + "loss": 0.2977, + "step": 42538 + }, + { + "epoch": 1.64, + "learning_rate": 2.275647601949218e-05, + "loss": 0.3583, + "step": 42539 + }, + { + "epoch": 1.64, + "learning_rate": 2.2755834829443447e-05, + "loss": 0.4436, + "step": 42540 + }, + { + "epoch": 1.64, + "learning_rate": 2.275519363939472e-05, + "loss": 0.5778, + "step": 42541 + }, + { + "epoch": 1.64, + "learning_rate": 2.2754552449345987e-05, + "loss": 0.4764, + "step": 42542 + }, + { + "epoch": 1.64, + "learning_rate": 2.2753911259297255e-05, + "loss": 0.265, + "step": 42543 + }, + { + "epoch": 1.64, + "learning_rate": 2.2753270069248524e-05, + "loss": 0.4095, + "step": 42544 + }, + { + "epoch": 1.64, + "learning_rate": 2.2752628879199795e-05, + "loss": 0.4079, + "step": 42545 + }, + { + "epoch": 1.64, + "learning_rate": 2.2751987689151067e-05, + "loss": 0.359, + "step": 42546 + }, + { + "epoch": 1.64, + "learning_rate": 2.2751346499102336e-05, + "loss": 0.3736, + "step": 42547 + }, + { + "epoch": 1.64, + "learning_rate": 2.2750705309053604e-05, + "loss": 0.3784, + "step": 42548 + }, + { + "epoch": 1.64, + "learning_rate": 2.2750064119004876e-05, + "loss": 0.3325, + "step": 42549 + }, + { + "epoch": 1.64, + "learning_rate": 2.2749422928956144e-05, + "loss": 0.3766, + "step": 42550 + }, + { + "epoch": 1.64, + "learning_rate": 2.2748781738907412e-05, + "loss": 0.3464, + "step": 42551 + }, + { + "epoch": 1.64, + "learning_rate": 2.274814054885868e-05, + "loss": 0.2991, + "step": 42552 + }, + { + "epoch": 1.64, + "learning_rate": 2.2747499358809952e-05, + "loss": 0.2774, + "step": 42553 + }, + { + "epoch": 1.64, + "learning_rate": 2.2746858168761224e-05, + "loss": 0.4162, + "step": 42554 + }, + { + "epoch": 1.64, + "learning_rate": 2.2746216978712492e-05, + "loss": 0.3639, + "step": 42555 + }, + { + "epoch": 1.64, + "learning_rate": 2.274557578866376e-05, + "loss": 0.4335, + "step": 42556 + }, + { + "epoch": 1.64, + "learning_rate": 2.2744934598615032e-05, + "loss": 0.3457, + "step": 42557 + }, + { + "epoch": 1.64, + "learning_rate": 2.27442934085663e-05, + "loss": 0.3216, + "step": 42558 + }, + { + "epoch": 1.64, + "learning_rate": 2.274365221851757e-05, + "loss": 0.328, + "step": 42559 + }, + { + "epoch": 1.64, + "learning_rate": 2.2743011028468837e-05, + "loss": 0.3164, + "step": 42560 + }, + { + "epoch": 1.64, + "learning_rate": 2.274236983842011e-05, + "loss": 0.2629, + "step": 42561 + }, + { + "epoch": 1.64, + "learning_rate": 2.2741728648371377e-05, + "loss": 0.3706, + "step": 42562 + }, + { + "epoch": 1.64, + "learning_rate": 2.274108745832265e-05, + "loss": 0.3149, + "step": 42563 + }, + { + "epoch": 1.64, + "learning_rate": 2.2740446268273917e-05, + "loss": 0.3343, + "step": 42564 + }, + { + "epoch": 1.64, + "learning_rate": 2.273980507822519e-05, + "loss": 0.3564, + "step": 42565 + }, + { + "epoch": 1.64, + "learning_rate": 2.2739163888176457e-05, + "loss": 0.2264, + "step": 42566 + }, + { + "epoch": 1.64, + "learning_rate": 2.2738522698127726e-05, + "loss": 0.2875, + "step": 42567 + }, + { + "epoch": 1.64, + "learning_rate": 2.2737881508078994e-05, + "loss": 0.4724, + "step": 42568 + }, + { + "epoch": 1.64, + "learning_rate": 2.2737240318030266e-05, + "loss": 0.2973, + "step": 42569 + }, + { + "epoch": 1.64, + "learning_rate": 2.2736599127981534e-05, + "loss": 0.3338, + "step": 42570 + }, + { + "epoch": 1.64, + "learning_rate": 2.2735957937932803e-05, + "loss": 0.2443, + "step": 42571 + }, + { + "epoch": 1.64, + "learning_rate": 2.2735316747884074e-05, + "loss": 0.3113, + "step": 42572 + }, + { + "epoch": 1.64, + "learning_rate": 2.2734675557835346e-05, + "loss": 0.2349, + "step": 42573 + }, + { + "epoch": 1.64, + "learning_rate": 2.2734034367786614e-05, + "loss": 0.3701, + "step": 42574 + }, + { + "epoch": 1.64, + "learning_rate": 2.2733393177737883e-05, + "loss": 0.3097, + "step": 42575 + }, + { + "epoch": 1.64, + "learning_rate": 2.273275198768915e-05, + "loss": 0.3325, + "step": 42576 + }, + { + "epoch": 1.64, + "learning_rate": 2.2732110797640423e-05, + "loss": 0.3462, + "step": 42577 + }, + { + "epoch": 1.64, + "learning_rate": 2.273146960759169e-05, + "loss": 0.2209, + "step": 42578 + }, + { + "epoch": 1.64, + "learning_rate": 2.273082841754296e-05, + "loss": 0.3901, + "step": 42579 + }, + { + "epoch": 1.64, + "learning_rate": 2.2730187227494228e-05, + "loss": 0.3681, + "step": 42580 + }, + { + "epoch": 1.64, + "learning_rate": 2.27295460374455e-05, + "loss": 0.3458, + "step": 42581 + }, + { + "epoch": 1.64, + "learning_rate": 2.272890484739677e-05, + "loss": 0.283, + "step": 42582 + }, + { + "epoch": 1.64, + "learning_rate": 2.272826365734804e-05, + "loss": 0.3289, + "step": 42583 + }, + { + "epoch": 1.64, + "learning_rate": 2.2727622467299308e-05, + "loss": 0.3009, + "step": 42584 + }, + { + "epoch": 1.64, + "learning_rate": 2.272698127725058e-05, + "loss": 0.3909, + "step": 42585 + }, + { + "epoch": 1.64, + "learning_rate": 2.2726340087201848e-05, + "loss": 0.4038, + "step": 42586 + }, + { + "epoch": 1.64, + "learning_rate": 2.2725698897153116e-05, + "loss": 0.2948, + "step": 42587 + }, + { + "epoch": 1.64, + "learning_rate": 2.2725057707104384e-05, + "loss": 0.3201, + "step": 42588 + }, + { + "epoch": 1.64, + "learning_rate": 2.2724416517055656e-05, + "loss": 0.2421, + "step": 42589 + }, + { + "epoch": 1.64, + "learning_rate": 2.2723775327006928e-05, + "loss": 0.5544, + "step": 42590 + }, + { + "epoch": 1.64, + "learning_rate": 2.2723134136958196e-05, + "loss": 0.3429, + "step": 42591 + }, + { + "epoch": 1.64, + "learning_rate": 2.2722492946909465e-05, + "loss": 0.3248, + "step": 42592 + }, + { + "epoch": 1.64, + "learning_rate": 2.2721851756860736e-05, + "loss": 0.342, + "step": 42593 + }, + { + "epoch": 1.64, + "learning_rate": 2.2721210566812005e-05, + "loss": 0.4012, + "step": 42594 + }, + { + "epoch": 1.64, + "learning_rate": 2.2720569376763273e-05, + "loss": 0.2804, + "step": 42595 + }, + { + "epoch": 1.64, + "learning_rate": 2.271992818671454e-05, + "loss": 0.4007, + "step": 42596 + }, + { + "epoch": 1.64, + "learning_rate": 2.2719286996665813e-05, + "loss": 0.3349, + "step": 42597 + }, + { + "epoch": 1.64, + "learning_rate": 2.2718645806617085e-05, + "loss": 0.3088, + "step": 42598 + }, + { + "epoch": 1.64, + "learning_rate": 2.2718004616568353e-05, + "loss": 0.3748, + "step": 42599 + }, + { + "epoch": 1.64, + "learning_rate": 2.271736342651962e-05, + "loss": 0.3067, + "step": 42600 + }, + { + "epoch": 1.64, + "learning_rate": 2.2716722236470893e-05, + "loss": 0.4817, + "step": 42601 + }, + { + "epoch": 1.64, + "learning_rate": 2.271608104642216e-05, + "loss": 0.3164, + "step": 42602 + }, + { + "epoch": 1.64, + "learning_rate": 2.271543985637343e-05, + "loss": 0.4464, + "step": 42603 + }, + { + "epoch": 1.64, + "learning_rate": 2.2714798666324698e-05, + "loss": 0.3779, + "step": 42604 + }, + { + "epoch": 1.64, + "learning_rate": 2.271415747627597e-05, + "loss": 0.3723, + "step": 42605 + }, + { + "epoch": 1.64, + "learning_rate": 2.2713516286227238e-05, + "loss": 0.3363, + "step": 42606 + }, + { + "epoch": 1.64, + "learning_rate": 2.271287509617851e-05, + "loss": 0.3681, + "step": 42607 + }, + { + "epoch": 1.64, + "learning_rate": 2.2712233906129778e-05, + "loss": 0.3625, + "step": 42608 + }, + { + "epoch": 1.64, + "learning_rate": 2.271159271608105e-05, + "loss": 0.317, + "step": 42609 + }, + { + "epoch": 1.64, + "learning_rate": 2.2710951526032318e-05, + "loss": 0.3324, + "step": 42610 + }, + { + "epoch": 1.64, + "learning_rate": 2.2710310335983586e-05, + "loss": 0.343, + "step": 42611 + }, + { + "epoch": 1.64, + "learning_rate": 2.2709669145934855e-05, + "loss": 0.3811, + "step": 42612 + }, + { + "epoch": 1.64, + "learning_rate": 2.2709027955886127e-05, + "loss": 0.3107, + "step": 42613 + }, + { + "epoch": 1.64, + "learning_rate": 2.2708386765837395e-05, + "loss": 0.3266, + "step": 42614 + }, + { + "epoch": 1.64, + "learning_rate": 2.2707745575788663e-05, + "loss": 0.3683, + "step": 42615 + }, + { + "epoch": 1.64, + "learning_rate": 2.2707104385739935e-05, + "loss": 0.4513, + "step": 42616 + }, + { + "epoch": 1.64, + "learning_rate": 2.2706463195691207e-05, + "loss": 0.3637, + "step": 42617 + }, + { + "epoch": 1.64, + "learning_rate": 2.2705822005642475e-05, + "loss": 0.2815, + "step": 42618 + }, + { + "epoch": 1.64, + "learning_rate": 2.2705180815593743e-05, + "loss": 0.3643, + "step": 42619 + }, + { + "epoch": 1.64, + "learning_rate": 2.270453962554501e-05, + "loss": 0.387, + "step": 42620 + }, + { + "epoch": 1.64, + "learning_rate": 2.2703898435496283e-05, + "loss": 0.4831, + "step": 42621 + }, + { + "epoch": 1.64, + "learning_rate": 2.270325724544755e-05, + "loss": 0.4473, + "step": 42622 + }, + { + "epoch": 1.64, + "learning_rate": 2.270261605539882e-05, + "loss": 0.2851, + "step": 42623 + }, + { + "epoch": 1.64, + "learning_rate": 2.270197486535009e-05, + "loss": 0.3155, + "step": 42624 + }, + { + "epoch": 1.64, + "learning_rate": 2.270133367530136e-05, + "loss": 0.4188, + "step": 42625 + }, + { + "epoch": 1.64, + "learning_rate": 2.2700692485252632e-05, + "loss": 0.4064, + "step": 42626 + }, + { + "epoch": 1.64, + "learning_rate": 2.27000512952039e-05, + "loss": 0.2504, + "step": 42627 + }, + { + "epoch": 1.64, + "learning_rate": 2.269941010515517e-05, + "loss": 0.2842, + "step": 42628 + }, + { + "epoch": 1.64, + "learning_rate": 2.269876891510644e-05, + "loss": 0.485, + "step": 42629 + }, + { + "epoch": 1.64, + "learning_rate": 2.269812772505771e-05, + "loss": 0.4144, + "step": 42630 + }, + { + "epoch": 1.64, + "learning_rate": 2.2697486535008977e-05, + "loss": 0.2809, + "step": 42631 + }, + { + "epoch": 1.64, + "learning_rate": 2.2696845344960245e-05, + "loss": 0.304, + "step": 42632 + }, + { + "epoch": 1.64, + "learning_rate": 2.2696204154911517e-05, + "loss": 0.2569, + "step": 42633 + }, + { + "epoch": 1.64, + "learning_rate": 2.269556296486279e-05, + "loss": 0.341, + "step": 42634 + }, + { + "epoch": 1.64, + "learning_rate": 2.2694921774814057e-05, + "loss": 0.3293, + "step": 42635 + }, + { + "epoch": 1.64, + "learning_rate": 2.2694280584765325e-05, + "loss": 0.3702, + "step": 42636 + }, + { + "epoch": 1.64, + "learning_rate": 2.2693639394716597e-05, + "loss": 0.295, + "step": 42637 + }, + { + "epoch": 1.64, + "learning_rate": 2.2692998204667865e-05, + "loss": 0.4173, + "step": 42638 + }, + { + "epoch": 1.64, + "learning_rate": 2.2692357014619134e-05, + "loss": 0.2728, + "step": 42639 + }, + { + "epoch": 1.64, + "learning_rate": 2.2691715824570402e-05, + "loss": 0.2947, + "step": 42640 + }, + { + "epoch": 1.64, + "learning_rate": 2.2691074634521674e-05, + "loss": 0.2856, + "step": 42641 + }, + { + "epoch": 1.64, + "learning_rate": 2.2690433444472942e-05, + "loss": 0.4012, + "step": 42642 + }, + { + "epoch": 1.64, + "learning_rate": 2.2689792254424214e-05, + "loss": 0.3354, + "step": 42643 + }, + { + "epoch": 1.64, + "learning_rate": 2.2689151064375482e-05, + "loss": 0.3943, + "step": 42644 + }, + { + "epoch": 1.64, + "learning_rate": 2.2688509874326754e-05, + "loss": 0.3985, + "step": 42645 + }, + { + "epoch": 1.64, + "learning_rate": 2.2687868684278022e-05, + "loss": 0.3098, + "step": 42646 + }, + { + "epoch": 1.64, + "learning_rate": 2.268722749422929e-05, + "loss": 0.2629, + "step": 42647 + }, + { + "epoch": 1.64, + "learning_rate": 2.268658630418056e-05, + "loss": 0.4271, + "step": 42648 + }, + { + "epoch": 1.64, + "learning_rate": 2.268594511413183e-05, + "loss": 0.3513, + "step": 42649 + }, + { + "epoch": 1.64, + "learning_rate": 2.26853039240831e-05, + "loss": 0.352, + "step": 42650 + }, + { + "epoch": 1.64, + "learning_rate": 2.2684662734034367e-05, + "loss": 0.4612, + "step": 42651 + }, + { + "epoch": 1.64, + "learning_rate": 2.268402154398564e-05, + "loss": 0.3027, + "step": 42652 + }, + { + "epoch": 1.64, + "learning_rate": 2.268338035393691e-05, + "loss": 0.3683, + "step": 42653 + }, + { + "epoch": 1.64, + "learning_rate": 2.268273916388818e-05, + "loss": 0.3747, + "step": 42654 + }, + { + "epoch": 1.64, + "learning_rate": 2.2682097973839447e-05, + "loss": 0.2941, + "step": 42655 + }, + { + "epoch": 1.64, + "learning_rate": 2.2681456783790715e-05, + "loss": 0.3169, + "step": 42656 + }, + { + "epoch": 1.64, + "learning_rate": 2.2680815593741987e-05, + "loss": 0.3448, + "step": 42657 + }, + { + "epoch": 1.64, + "learning_rate": 2.2680174403693256e-05, + "loss": 0.3489, + "step": 42658 + }, + { + "epoch": 1.64, + "learning_rate": 2.2679533213644524e-05, + "loss": 0.328, + "step": 42659 + }, + { + "epoch": 1.64, + "learning_rate": 2.2678892023595792e-05, + "loss": 0.4014, + "step": 42660 + }, + { + "epoch": 1.64, + "learning_rate": 2.2678250833547064e-05, + "loss": 0.3726, + "step": 42661 + }, + { + "epoch": 1.64, + "learning_rate": 2.2677609643498336e-05, + "loss": 0.3495, + "step": 42662 + }, + { + "epoch": 1.64, + "learning_rate": 2.2676968453449604e-05, + "loss": 0.3593, + "step": 42663 + }, + { + "epoch": 1.64, + "learning_rate": 2.2676327263400872e-05, + "loss": 0.4628, + "step": 42664 + }, + { + "epoch": 1.64, + "learning_rate": 2.2675686073352144e-05, + "loss": 0.3647, + "step": 42665 + }, + { + "epoch": 1.64, + "learning_rate": 2.2675044883303412e-05, + "loss": 0.2949, + "step": 42666 + }, + { + "epoch": 1.64, + "learning_rate": 2.267440369325468e-05, + "loss": 0.2317, + "step": 42667 + }, + { + "epoch": 1.64, + "learning_rate": 2.267376250320595e-05, + "loss": 0.2851, + "step": 42668 + }, + { + "epoch": 1.64, + "learning_rate": 2.267312131315722e-05, + "loss": 0.3897, + "step": 42669 + }, + { + "epoch": 1.64, + "learning_rate": 2.2672480123108492e-05, + "loss": 0.2791, + "step": 42670 + }, + { + "epoch": 1.64, + "learning_rate": 2.267183893305976e-05, + "loss": 0.4645, + "step": 42671 + }, + { + "epoch": 1.64, + "learning_rate": 2.267119774301103e-05, + "loss": 0.3602, + "step": 42672 + }, + { + "epoch": 1.64, + "learning_rate": 2.26705565529623e-05, + "loss": 0.5749, + "step": 42673 + }, + { + "epoch": 1.64, + "learning_rate": 2.266991536291357e-05, + "loss": 0.2812, + "step": 42674 + }, + { + "epoch": 1.64, + "learning_rate": 2.2669274172864837e-05, + "loss": 0.4097, + "step": 42675 + }, + { + "epoch": 1.64, + "learning_rate": 2.2668632982816106e-05, + "loss": 0.3256, + "step": 42676 + }, + { + "epoch": 1.64, + "learning_rate": 2.2667991792767377e-05, + "loss": 0.2902, + "step": 42677 + }, + { + "epoch": 1.64, + "learning_rate": 2.2667350602718646e-05, + "loss": 0.3995, + "step": 42678 + }, + { + "epoch": 1.64, + "learning_rate": 2.2666709412669918e-05, + "loss": 0.2815, + "step": 42679 + }, + { + "epoch": 1.64, + "learning_rate": 2.2666068222621186e-05, + "loss": 0.1935, + "step": 42680 + }, + { + "epoch": 1.64, + "learning_rate": 2.2665427032572458e-05, + "loss": 0.4357, + "step": 42681 + }, + { + "epoch": 1.64, + "learning_rate": 2.2664785842523726e-05, + "loss": 0.3882, + "step": 42682 + }, + { + "epoch": 1.64, + "learning_rate": 2.2664144652474994e-05, + "loss": 0.5018, + "step": 42683 + }, + { + "epoch": 1.64, + "learning_rate": 2.2663503462426263e-05, + "loss": 0.3118, + "step": 42684 + }, + { + "epoch": 1.64, + "learning_rate": 2.2662862272377534e-05, + "loss": 0.295, + "step": 42685 + }, + { + "epoch": 1.64, + "learning_rate": 2.2662221082328803e-05, + "loss": 0.326, + "step": 42686 + }, + { + "epoch": 1.64, + "learning_rate": 2.2661579892280074e-05, + "loss": 0.3515, + "step": 42687 + }, + { + "epoch": 1.64, + "learning_rate": 2.2660938702231343e-05, + "loss": 0.3076, + "step": 42688 + }, + { + "epoch": 1.64, + "learning_rate": 2.2660297512182614e-05, + "loss": 0.3413, + "step": 42689 + }, + { + "epoch": 1.64, + "learning_rate": 2.2659656322133883e-05, + "loss": 0.278, + "step": 42690 + }, + { + "epoch": 1.64, + "learning_rate": 2.265901513208515e-05, + "loss": 0.2612, + "step": 42691 + }, + { + "epoch": 1.64, + "learning_rate": 2.265837394203642e-05, + "loss": 0.348, + "step": 42692 + }, + { + "epoch": 1.64, + "learning_rate": 2.265773275198769e-05, + "loss": 0.344, + "step": 42693 + }, + { + "epoch": 1.64, + "learning_rate": 2.265709156193896e-05, + "loss": 0.4082, + "step": 42694 + }, + { + "epoch": 1.64, + "learning_rate": 2.2656450371890228e-05, + "loss": 0.4367, + "step": 42695 + }, + { + "epoch": 1.64, + "learning_rate": 2.26558091818415e-05, + "loss": 0.4152, + "step": 42696 + }, + { + "epoch": 1.64, + "learning_rate": 2.265516799179277e-05, + "loss": 0.5336, + "step": 42697 + }, + { + "epoch": 1.64, + "learning_rate": 2.265452680174404e-05, + "loss": 0.4307, + "step": 42698 + }, + { + "epoch": 1.64, + "learning_rate": 2.2653885611695308e-05, + "loss": 0.4295, + "step": 42699 + }, + { + "epoch": 1.64, + "learning_rate": 2.2653244421646576e-05, + "loss": 0.3403, + "step": 42700 + }, + { + "epoch": 1.64, + "learning_rate": 2.2652603231597848e-05, + "loss": 0.2851, + "step": 42701 + }, + { + "epoch": 1.64, + "learning_rate": 2.2651962041549116e-05, + "loss": 0.3734, + "step": 42702 + }, + { + "epoch": 1.64, + "learning_rate": 2.2651320851500385e-05, + "loss": 0.4771, + "step": 42703 + }, + { + "epoch": 1.64, + "learning_rate": 2.2650679661451653e-05, + "loss": 0.1885, + "step": 42704 + }, + { + "epoch": 1.64, + "learning_rate": 2.2650038471402925e-05, + "loss": 0.3927, + "step": 42705 + }, + { + "epoch": 1.64, + "learning_rate": 2.2649397281354196e-05, + "loss": 0.2422, + "step": 42706 + }, + { + "epoch": 1.64, + "learning_rate": 2.2648756091305465e-05, + "loss": 0.3159, + "step": 42707 + }, + { + "epoch": 1.64, + "learning_rate": 2.2648114901256733e-05, + "loss": 0.3445, + "step": 42708 + }, + { + "epoch": 1.64, + "learning_rate": 2.2647473711208005e-05, + "loss": 0.4537, + "step": 42709 + }, + { + "epoch": 1.64, + "learning_rate": 2.2646832521159273e-05, + "loss": 0.3061, + "step": 42710 + }, + { + "epoch": 1.64, + "learning_rate": 2.264619133111054e-05, + "loss": 0.3766, + "step": 42711 + }, + { + "epoch": 1.64, + "learning_rate": 2.264555014106181e-05, + "loss": 0.3619, + "step": 42712 + }, + { + "epoch": 1.64, + "learning_rate": 2.264490895101308e-05, + "loss": 0.4265, + "step": 42713 + }, + { + "epoch": 1.64, + "learning_rate": 2.2644267760964353e-05, + "loss": 0.3312, + "step": 42714 + }, + { + "epoch": 1.64, + "learning_rate": 2.264362657091562e-05, + "loss": 0.404, + "step": 42715 + }, + { + "epoch": 1.64, + "learning_rate": 2.264298538086689e-05, + "loss": 0.4944, + "step": 42716 + }, + { + "epoch": 1.64, + "learning_rate": 2.264234419081816e-05, + "loss": 0.3286, + "step": 42717 + }, + { + "epoch": 1.64, + "learning_rate": 2.264170300076943e-05, + "loss": 0.334, + "step": 42718 + }, + { + "epoch": 1.64, + "learning_rate": 2.2641061810720698e-05, + "loss": 0.3936, + "step": 42719 + }, + { + "epoch": 1.64, + "learning_rate": 2.2640420620671966e-05, + "loss": 0.4459, + "step": 42720 + }, + { + "epoch": 1.64, + "learning_rate": 2.2639779430623238e-05, + "loss": 0.3378, + "step": 42721 + }, + { + "epoch": 1.64, + "learning_rate": 2.2639138240574506e-05, + "loss": 0.3346, + "step": 42722 + }, + { + "epoch": 1.64, + "learning_rate": 2.2638497050525778e-05, + "loss": 0.3112, + "step": 42723 + }, + { + "epoch": 1.64, + "learning_rate": 2.2637855860477047e-05, + "loss": 0.2674, + "step": 42724 + }, + { + "epoch": 1.64, + "learning_rate": 2.2637214670428318e-05, + "loss": 0.387, + "step": 42725 + }, + { + "epoch": 1.64, + "learning_rate": 2.2636573480379587e-05, + "loss": 0.4326, + "step": 42726 + }, + { + "epoch": 1.64, + "learning_rate": 2.2635932290330855e-05, + "loss": 0.2846, + "step": 42727 + }, + { + "epoch": 1.64, + "learning_rate": 2.2635291100282123e-05, + "loss": 0.3143, + "step": 42728 + }, + { + "epoch": 1.64, + "learning_rate": 2.2634649910233395e-05, + "loss": 0.3557, + "step": 42729 + }, + { + "epoch": 1.64, + "learning_rate": 2.2634008720184663e-05, + "loss": 0.3726, + "step": 42730 + }, + { + "epoch": 1.64, + "learning_rate": 2.263336753013593e-05, + "loss": 0.2918, + "step": 42731 + }, + { + "epoch": 1.64, + "learning_rate": 2.2632726340087203e-05, + "loss": 0.3837, + "step": 42732 + }, + { + "epoch": 1.64, + "learning_rate": 2.2632085150038475e-05, + "loss": 0.2468, + "step": 42733 + }, + { + "epoch": 1.64, + "learning_rate": 2.2631443959989743e-05, + "loss": 0.3371, + "step": 42734 + }, + { + "epoch": 1.64, + "learning_rate": 2.263080276994101e-05, + "loss": 0.4159, + "step": 42735 + }, + { + "epoch": 1.64, + "learning_rate": 2.263016157989228e-05, + "loss": 0.4791, + "step": 42736 + }, + { + "epoch": 1.64, + "learning_rate": 2.2629520389843552e-05, + "loss": 0.3161, + "step": 42737 + }, + { + "epoch": 1.64, + "learning_rate": 2.262887919979482e-05, + "loss": 0.3838, + "step": 42738 + }, + { + "epoch": 1.64, + "learning_rate": 2.262823800974609e-05, + "loss": 0.3785, + "step": 42739 + }, + { + "epoch": 1.64, + "learning_rate": 2.2627596819697357e-05, + "loss": 0.3878, + "step": 42740 + }, + { + "epoch": 1.64, + "learning_rate": 2.2626955629648632e-05, + "loss": 0.2911, + "step": 42741 + }, + { + "epoch": 1.64, + "learning_rate": 2.26263144395999e-05, + "loss": 0.4036, + "step": 42742 + }, + { + "epoch": 1.64, + "learning_rate": 2.262567324955117e-05, + "loss": 0.3401, + "step": 42743 + }, + { + "epoch": 1.64, + "learning_rate": 2.2625032059502437e-05, + "loss": 0.4277, + "step": 42744 + }, + { + "epoch": 1.64, + "learning_rate": 2.262439086945371e-05, + "loss": 0.2806, + "step": 42745 + }, + { + "epoch": 1.64, + "learning_rate": 2.2623749679404977e-05, + "loss": 0.3424, + "step": 42746 + }, + { + "epoch": 1.64, + "learning_rate": 2.2623108489356245e-05, + "loss": 0.309, + "step": 42747 + }, + { + "epoch": 1.64, + "learning_rate": 2.2622467299307514e-05, + "loss": 0.3122, + "step": 42748 + }, + { + "epoch": 1.64, + "learning_rate": 2.2621826109258785e-05, + "loss": 0.3075, + "step": 42749 + }, + { + "epoch": 1.64, + "learning_rate": 2.2621184919210057e-05, + "loss": 0.3893, + "step": 42750 + }, + { + "epoch": 1.64, + "learning_rate": 2.2620543729161325e-05, + "loss": 0.3258, + "step": 42751 + }, + { + "epoch": 1.64, + "learning_rate": 2.2619902539112594e-05, + "loss": 0.3667, + "step": 42752 + }, + { + "epoch": 1.64, + "learning_rate": 2.2619261349063865e-05, + "loss": 0.4009, + "step": 42753 + }, + { + "epoch": 1.64, + "learning_rate": 2.2618620159015134e-05, + "loss": 0.4105, + "step": 42754 + }, + { + "epoch": 1.64, + "learning_rate": 2.2617978968966402e-05, + "loss": 0.4184, + "step": 42755 + }, + { + "epoch": 1.64, + "learning_rate": 2.261733777891767e-05, + "loss": 0.3263, + "step": 42756 + }, + { + "epoch": 1.64, + "learning_rate": 2.2616696588868942e-05, + "loss": 0.4222, + "step": 42757 + }, + { + "epoch": 1.64, + "learning_rate": 2.261605539882021e-05, + "loss": 0.2465, + "step": 42758 + }, + { + "epoch": 1.64, + "learning_rate": 2.2615414208771482e-05, + "loss": 0.3188, + "step": 42759 + }, + { + "epoch": 1.64, + "learning_rate": 2.261477301872275e-05, + "loss": 0.2845, + "step": 42760 + }, + { + "epoch": 1.64, + "learning_rate": 2.2614131828674022e-05, + "loss": 0.4555, + "step": 42761 + }, + { + "epoch": 1.64, + "learning_rate": 2.261349063862529e-05, + "loss": 0.2918, + "step": 42762 + }, + { + "epoch": 1.64, + "learning_rate": 2.261284944857656e-05, + "loss": 0.2465, + "step": 42763 + }, + { + "epoch": 1.64, + "learning_rate": 2.2612208258527827e-05, + "loss": 0.3687, + "step": 42764 + }, + { + "epoch": 1.64, + "learning_rate": 2.26115670684791e-05, + "loss": 0.3589, + "step": 42765 + }, + { + "epoch": 1.64, + "learning_rate": 2.2610925878430367e-05, + "loss": 0.2839, + "step": 42766 + }, + { + "epoch": 1.64, + "learning_rate": 2.2610284688381635e-05, + "loss": 0.5159, + "step": 42767 + }, + { + "epoch": 1.64, + "learning_rate": 2.2609643498332907e-05, + "loss": 0.4264, + "step": 42768 + }, + { + "epoch": 1.64, + "learning_rate": 2.260900230828418e-05, + "loss": 0.3043, + "step": 42769 + }, + { + "epoch": 1.64, + "learning_rate": 2.2608361118235447e-05, + "loss": 0.263, + "step": 42770 + }, + { + "epoch": 1.64, + "learning_rate": 2.2607719928186716e-05, + "loss": 0.3234, + "step": 42771 + }, + { + "epoch": 1.64, + "learning_rate": 2.2607078738137984e-05, + "loss": 0.329, + "step": 42772 + }, + { + "epoch": 1.64, + "learning_rate": 2.2606437548089256e-05, + "loss": 0.5362, + "step": 42773 + }, + { + "epoch": 1.64, + "learning_rate": 2.2605796358040524e-05, + "loss": 0.2567, + "step": 42774 + }, + { + "epoch": 1.64, + "learning_rate": 2.2605155167991792e-05, + "loss": 0.4434, + "step": 42775 + }, + { + "epoch": 1.64, + "learning_rate": 2.260451397794306e-05, + "loss": 0.2638, + "step": 42776 + }, + { + "epoch": 1.64, + "learning_rate": 2.2603872787894336e-05, + "loss": 0.4213, + "step": 42777 + }, + { + "epoch": 1.64, + "learning_rate": 2.2603231597845604e-05, + "loss": 0.345, + "step": 42778 + }, + { + "epoch": 1.64, + "learning_rate": 2.2602590407796872e-05, + "loss": 0.391, + "step": 42779 + }, + { + "epoch": 1.64, + "learning_rate": 2.260194921774814e-05, + "loss": 0.2621, + "step": 42780 + }, + { + "epoch": 1.64, + "learning_rate": 2.2601308027699412e-05, + "loss": 0.2429, + "step": 42781 + }, + { + "epoch": 1.64, + "learning_rate": 2.260066683765068e-05, + "loss": 0.27, + "step": 42782 + }, + { + "epoch": 1.64, + "learning_rate": 2.260002564760195e-05, + "loss": 0.4142, + "step": 42783 + }, + { + "epoch": 1.64, + "learning_rate": 2.2599384457553217e-05, + "loss": 0.3401, + "step": 42784 + }, + { + "epoch": 1.64, + "learning_rate": 2.259874326750449e-05, + "loss": 0.43, + "step": 42785 + }, + { + "epoch": 1.64, + "learning_rate": 2.259810207745576e-05, + "loss": 0.3171, + "step": 42786 + }, + { + "epoch": 1.65, + "learning_rate": 2.259746088740703e-05, + "loss": 0.2588, + "step": 42787 + }, + { + "epoch": 1.65, + "learning_rate": 2.2596819697358297e-05, + "loss": 0.3299, + "step": 42788 + }, + { + "epoch": 1.65, + "learning_rate": 2.259617850730957e-05, + "loss": 0.3552, + "step": 42789 + }, + { + "epoch": 1.65, + "learning_rate": 2.2595537317260838e-05, + "loss": 0.2792, + "step": 42790 + }, + { + "epoch": 1.65, + "learning_rate": 2.2594896127212106e-05, + "loss": 0.2397, + "step": 42791 + }, + { + "epoch": 1.65, + "learning_rate": 2.2594254937163374e-05, + "loss": 0.4418, + "step": 42792 + }, + { + "epoch": 1.65, + "learning_rate": 2.2593613747114646e-05, + "loss": 0.2857, + "step": 42793 + }, + { + "epoch": 1.65, + "learning_rate": 2.2592972557065918e-05, + "loss": 0.3585, + "step": 42794 + }, + { + "epoch": 1.65, + "learning_rate": 2.2592331367017186e-05, + "loss": 0.3332, + "step": 42795 + }, + { + "epoch": 1.65, + "learning_rate": 2.2591690176968454e-05, + "loss": 0.4327, + "step": 42796 + }, + { + "epoch": 1.65, + "learning_rate": 2.2591048986919726e-05, + "loss": 0.3099, + "step": 42797 + }, + { + "epoch": 1.65, + "learning_rate": 2.2590407796870994e-05, + "loss": 0.2743, + "step": 42798 + }, + { + "epoch": 1.65, + "learning_rate": 2.2589766606822263e-05, + "loss": 0.3042, + "step": 42799 + }, + { + "epoch": 1.65, + "learning_rate": 2.258912541677353e-05, + "loss": 0.3429, + "step": 42800 + }, + { + "epoch": 1.65, + "learning_rate": 2.2588484226724803e-05, + "loss": 0.3562, + "step": 42801 + }, + { + "epoch": 1.65, + "learning_rate": 2.258784303667607e-05, + "loss": 0.3374, + "step": 42802 + }, + { + "epoch": 1.65, + "learning_rate": 2.2587201846627343e-05, + "loss": 0.2928, + "step": 42803 + }, + { + "epoch": 1.65, + "learning_rate": 2.258656065657861e-05, + "loss": 0.4042, + "step": 42804 + }, + { + "epoch": 1.65, + "learning_rate": 2.2585919466529883e-05, + "loss": 0.3008, + "step": 42805 + }, + { + "epoch": 1.65, + "learning_rate": 2.258527827648115e-05, + "loss": 0.3031, + "step": 42806 + }, + { + "epoch": 1.65, + "learning_rate": 2.258463708643242e-05, + "loss": 0.299, + "step": 42807 + }, + { + "epoch": 1.65, + "learning_rate": 2.2583995896383688e-05, + "loss": 0.3582, + "step": 42808 + }, + { + "epoch": 1.65, + "learning_rate": 2.258335470633496e-05, + "loss": 0.4329, + "step": 42809 + }, + { + "epoch": 1.65, + "learning_rate": 2.2582713516286228e-05, + "loss": 0.3678, + "step": 42810 + }, + { + "epoch": 1.65, + "learning_rate": 2.2582072326237496e-05, + "loss": 0.4176, + "step": 42811 + }, + { + "epoch": 1.65, + "learning_rate": 2.2581431136188768e-05, + "loss": 0.2693, + "step": 42812 + }, + { + "epoch": 1.65, + "learning_rate": 2.258078994614004e-05, + "loss": 0.4008, + "step": 42813 + }, + { + "epoch": 1.65, + "learning_rate": 2.2580148756091308e-05, + "loss": 0.2888, + "step": 42814 + }, + { + "epoch": 1.65, + "learning_rate": 2.2579507566042576e-05, + "loss": 0.3308, + "step": 42815 + }, + { + "epoch": 1.65, + "learning_rate": 2.2578866375993845e-05, + "loss": 0.3235, + "step": 42816 + }, + { + "epoch": 1.65, + "learning_rate": 2.2578225185945116e-05, + "loss": 0.3362, + "step": 42817 + }, + { + "epoch": 1.65, + "learning_rate": 2.2577583995896385e-05, + "loss": 0.5865, + "step": 42818 + }, + { + "epoch": 1.65, + "learning_rate": 2.2576942805847653e-05, + "loss": 0.3511, + "step": 42819 + }, + { + "epoch": 1.65, + "learning_rate": 2.257630161579892e-05, + "loss": 0.3056, + "step": 42820 + }, + { + "epoch": 1.65, + "learning_rate": 2.2575660425750196e-05, + "loss": 0.4168, + "step": 42821 + }, + { + "epoch": 1.65, + "learning_rate": 2.2575019235701465e-05, + "loss": 0.3679, + "step": 42822 + }, + { + "epoch": 1.65, + "learning_rate": 2.2574378045652733e-05, + "loss": 0.3401, + "step": 42823 + }, + { + "epoch": 1.65, + "learning_rate": 2.2573736855604e-05, + "loss": 0.4206, + "step": 42824 + }, + { + "epoch": 1.65, + "learning_rate": 2.2573095665555273e-05, + "loss": 0.3171, + "step": 42825 + }, + { + "epoch": 1.65, + "learning_rate": 2.257245447550654e-05, + "loss": 0.4273, + "step": 42826 + }, + { + "epoch": 1.65, + "learning_rate": 2.257181328545781e-05, + "loss": 0.3454, + "step": 42827 + }, + { + "epoch": 1.65, + "learning_rate": 2.2571172095409078e-05, + "loss": 0.2139, + "step": 42828 + }, + { + "epoch": 1.65, + "learning_rate": 2.257053090536035e-05, + "loss": 0.3365, + "step": 42829 + }, + { + "epoch": 1.65, + "learning_rate": 2.256988971531162e-05, + "loss": 0.3556, + "step": 42830 + }, + { + "epoch": 1.65, + "learning_rate": 2.256924852526289e-05, + "loss": 0.4005, + "step": 42831 + }, + { + "epoch": 1.65, + "learning_rate": 2.2568607335214158e-05, + "loss": 0.3598, + "step": 42832 + }, + { + "epoch": 1.65, + "learning_rate": 2.256796614516543e-05, + "loss": 0.3319, + "step": 42833 + }, + { + "epoch": 1.65, + "learning_rate": 2.2567324955116698e-05, + "loss": 0.4215, + "step": 42834 + }, + { + "epoch": 1.65, + "learning_rate": 2.2566683765067967e-05, + "loss": 0.3711, + "step": 42835 + }, + { + "epoch": 1.65, + "learning_rate": 2.2566042575019235e-05, + "loss": 0.3756, + "step": 42836 + }, + { + "epoch": 1.65, + "learning_rate": 2.2565401384970507e-05, + "loss": 0.3579, + "step": 42837 + }, + { + "epoch": 1.65, + "learning_rate": 2.2564760194921775e-05, + "loss": 0.4464, + "step": 42838 + }, + { + "epoch": 1.65, + "learning_rate": 2.2564119004873047e-05, + "loss": 0.2638, + "step": 42839 + }, + { + "epoch": 1.65, + "learning_rate": 2.2563477814824315e-05, + "loss": 0.3092, + "step": 42840 + }, + { + "epoch": 1.65, + "learning_rate": 2.2562836624775587e-05, + "loss": 0.4428, + "step": 42841 + }, + { + "epoch": 1.65, + "learning_rate": 2.2562195434726855e-05, + "loss": 0.5796, + "step": 42842 + }, + { + "epoch": 1.65, + "learning_rate": 2.2561554244678123e-05, + "loss": 0.3153, + "step": 42843 + }, + { + "epoch": 1.65, + "learning_rate": 2.256091305462939e-05, + "loss": 0.3945, + "step": 42844 + }, + { + "epoch": 1.65, + "learning_rate": 2.2560271864580663e-05, + "loss": 0.3107, + "step": 42845 + }, + { + "epoch": 1.65, + "learning_rate": 2.255963067453193e-05, + "loss": 0.5407, + "step": 42846 + }, + { + "epoch": 1.65, + "learning_rate": 2.25589894844832e-05, + "loss": 0.3311, + "step": 42847 + }, + { + "epoch": 1.65, + "learning_rate": 2.2558348294434472e-05, + "loss": 0.3432, + "step": 42848 + }, + { + "epoch": 1.65, + "learning_rate": 2.2557707104385743e-05, + "loss": 0.3476, + "step": 42849 + }, + { + "epoch": 1.65, + "learning_rate": 2.2557065914337012e-05, + "loss": 0.4064, + "step": 42850 + }, + { + "epoch": 1.65, + "learning_rate": 2.255642472428828e-05, + "loss": 0.363, + "step": 42851 + }, + { + "epoch": 1.65, + "learning_rate": 2.255578353423955e-05, + "loss": 0.357, + "step": 42852 + }, + { + "epoch": 1.65, + "learning_rate": 2.255514234419082e-05, + "loss": 0.2866, + "step": 42853 + }, + { + "epoch": 1.65, + "learning_rate": 2.255450115414209e-05, + "loss": 0.3458, + "step": 42854 + }, + { + "epoch": 1.65, + "learning_rate": 2.2553859964093357e-05, + "loss": 0.2609, + "step": 42855 + }, + { + "epoch": 1.65, + "learning_rate": 2.2553218774044625e-05, + "loss": 0.3651, + "step": 42856 + }, + { + "epoch": 1.65, + "learning_rate": 2.25525775839959e-05, + "loss": 0.4934, + "step": 42857 + }, + { + "epoch": 1.65, + "learning_rate": 2.255193639394717e-05, + "loss": 0.3942, + "step": 42858 + }, + { + "epoch": 1.65, + "learning_rate": 2.2551295203898437e-05, + "loss": 0.3357, + "step": 42859 + }, + { + "epoch": 1.65, + "learning_rate": 2.2550654013849705e-05, + "loss": 0.251, + "step": 42860 + }, + { + "epoch": 1.65, + "learning_rate": 2.2550012823800977e-05, + "loss": 0.3301, + "step": 42861 + }, + { + "epoch": 1.65, + "learning_rate": 2.2549371633752245e-05, + "loss": 0.4328, + "step": 42862 + }, + { + "epoch": 1.65, + "learning_rate": 2.2548730443703514e-05, + "loss": 0.4752, + "step": 42863 + }, + { + "epoch": 1.65, + "learning_rate": 2.2548089253654782e-05, + "loss": 0.338, + "step": 42864 + }, + { + "epoch": 1.65, + "learning_rate": 2.2547448063606054e-05, + "loss": 0.3805, + "step": 42865 + }, + { + "epoch": 1.65, + "learning_rate": 2.2546806873557325e-05, + "loss": 0.3551, + "step": 42866 + }, + { + "epoch": 1.65, + "learning_rate": 2.2546165683508594e-05, + "loss": 0.4722, + "step": 42867 + }, + { + "epoch": 1.65, + "learning_rate": 2.2545524493459862e-05, + "loss": 0.2821, + "step": 42868 + }, + { + "epoch": 1.65, + "learning_rate": 2.2544883303411134e-05, + "loss": 0.2416, + "step": 42869 + }, + { + "epoch": 1.65, + "learning_rate": 2.2544242113362402e-05, + "loss": 0.3012, + "step": 42870 + }, + { + "epoch": 1.65, + "learning_rate": 2.254360092331367e-05, + "loss": 0.3741, + "step": 42871 + }, + { + "epoch": 1.65, + "learning_rate": 2.254295973326494e-05, + "loss": 0.4317, + "step": 42872 + }, + { + "epoch": 1.65, + "learning_rate": 2.254231854321621e-05, + "loss": 0.3467, + "step": 42873 + }, + { + "epoch": 1.65, + "learning_rate": 2.2541677353167482e-05, + "loss": 0.3639, + "step": 42874 + }, + { + "epoch": 1.65, + "learning_rate": 2.254103616311875e-05, + "loss": 0.2953, + "step": 42875 + }, + { + "epoch": 1.65, + "learning_rate": 2.254039497307002e-05, + "loss": 0.3813, + "step": 42876 + }, + { + "epoch": 1.65, + "learning_rate": 2.253975378302129e-05, + "loss": 0.3827, + "step": 42877 + }, + { + "epoch": 1.65, + "learning_rate": 2.253911259297256e-05, + "loss": 0.2578, + "step": 42878 + }, + { + "epoch": 1.65, + "learning_rate": 2.2538471402923827e-05, + "loss": 0.3338, + "step": 42879 + }, + { + "epoch": 1.65, + "learning_rate": 2.2537830212875095e-05, + "loss": 0.3406, + "step": 42880 + }, + { + "epoch": 1.65, + "learning_rate": 2.2537189022826367e-05, + "loss": 0.3881, + "step": 42881 + }, + { + "epoch": 1.65, + "learning_rate": 2.2536547832777636e-05, + "loss": 0.2747, + "step": 42882 + }, + { + "epoch": 1.65, + "learning_rate": 2.2535906642728907e-05, + "loss": 0.4497, + "step": 42883 + }, + { + "epoch": 1.65, + "learning_rate": 2.2535265452680176e-05, + "loss": 0.4287, + "step": 42884 + }, + { + "epoch": 1.65, + "learning_rate": 2.2534624262631447e-05, + "loss": 0.4671, + "step": 42885 + }, + { + "epoch": 1.65, + "learning_rate": 2.2533983072582716e-05, + "loss": 0.3032, + "step": 42886 + }, + { + "epoch": 1.65, + "learning_rate": 2.2533341882533984e-05, + "loss": 0.2824, + "step": 42887 + }, + { + "epoch": 1.65, + "learning_rate": 2.2532700692485252e-05, + "loss": 0.376, + "step": 42888 + }, + { + "epoch": 1.65, + "learning_rate": 2.2532059502436524e-05, + "loss": 0.4002, + "step": 42889 + }, + { + "epoch": 1.65, + "learning_rate": 2.2531418312387792e-05, + "loss": 0.4736, + "step": 42890 + }, + { + "epoch": 1.65, + "learning_rate": 2.253077712233906e-05, + "loss": 0.4102, + "step": 42891 + }, + { + "epoch": 1.65, + "learning_rate": 2.2530135932290332e-05, + "loss": 0.2943, + "step": 42892 + }, + { + "epoch": 1.65, + "learning_rate": 2.2529494742241604e-05, + "loss": 0.2765, + "step": 42893 + }, + { + "epoch": 1.65, + "learning_rate": 2.2528853552192872e-05, + "loss": 0.4482, + "step": 42894 + }, + { + "epoch": 1.65, + "learning_rate": 2.252821236214414e-05, + "loss": 0.2706, + "step": 42895 + }, + { + "epoch": 1.65, + "learning_rate": 2.252757117209541e-05, + "loss": 0.2918, + "step": 42896 + }, + { + "epoch": 1.65, + "learning_rate": 2.252692998204668e-05, + "loss": 0.2988, + "step": 42897 + }, + { + "epoch": 1.65, + "learning_rate": 2.252628879199795e-05, + "loss": 0.3489, + "step": 42898 + }, + { + "epoch": 1.65, + "learning_rate": 2.2525647601949217e-05, + "loss": 0.3142, + "step": 42899 + }, + { + "epoch": 1.65, + "learning_rate": 2.2525006411900486e-05, + "loss": 0.3259, + "step": 42900 + }, + { + "epoch": 1.65, + "learning_rate": 2.252436522185176e-05, + "loss": 0.3948, + "step": 42901 + }, + { + "epoch": 1.65, + "learning_rate": 2.252372403180303e-05, + "loss": 0.3153, + "step": 42902 + }, + { + "epoch": 1.65, + "learning_rate": 2.2523082841754298e-05, + "loss": 0.241, + "step": 42903 + }, + { + "epoch": 1.65, + "learning_rate": 2.2522441651705566e-05, + "loss": 0.3577, + "step": 42904 + }, + { + "epoch": 1.65, + "learning_rate": 2.2521800461656838e-05, + "loss": 0.311, + "step": 42905 + }, + { + "epoch": 1.65, + "learning_rate": 2.2521159271608106e-05, + "loss": 0.3164, + "step": 42906 + }, + { + "epoch": 1.65, + "learning_rate": 2.2520518081559374e-05, + "loss": 0.2478, + "step": 42907 + }, + { + "epoch": 1.65, + "learning_rate": 2.2519876891510643e-05, + "loss": 0.4113, + "step": 42908 + }, + { + "epoch": 1.65, + "learning_rate": 2.2519235701461914e-05, + "loss": 0.3099, + "step": 42909 + }, + { + "epoch": 1.65, + "learning_rate": 2.2518594511413186e-05, + "loss": 0.3709, + "step": 42910 + }, + { + "epoch": 1.65, + "learning_rate": 2.2517953321364454e-05, + "loss": 0.4461, + "step": 42911 + }, + { + "epoch": 1.65, + "learning_rate": 2.2517312131315723e-05, + "loss": 0.3326, + "step": 42912 + }, + { + "epoch": 1.65, + "learning_rate": 2.2516670941266994e-05, + "loss": 0.316, + "step": 42913 + }, + { + "epoch": 1.65, + "learning_rate": 2.2516029751218263e-05, + "loss": 0.3743, + "step": 42914 + }, + { + "epoch": 1.65, + "learning_rate": 2.251538856116953e-05, + "loss": 0.3224, + "step": 42915 + }, + { + "epoch": 1.65, + "learning_rate": 2.25147473711208e-05, + "loss": 0.3994, + "step": 42916 + }, + { + "epoch": 1.65, + "learning_rate": 2.251410618107207e-05, + "loss": 0.3702, + "step": 42917 + }, + { + "epoch": 1.65, + "learning_rate": 2.251346499102334e-05, + "loss": 0.43, + "step": 42918 + }, + { + "epoch": 1.65, + "learning_rate": 2.251282380097461e-05, + "loss": 0.2223, + "step": 42919 + }, + { + "epoch": 1.65, + "learning_rate": 2.251218261092588e-05, + "loss": 0.3368, + "step": 42920 + }, + { + "epoch": 1.65, + "learning_rate": 2.251154142087715e-05, + "loss": 0.4041, + "step": 42921 + }, + { + "epoch": 1.65, + "learning_rate": 2.251090023082842e-05, + "loss": 0.3133, + "step": 42922 + }, + { + "epoch": 1.65, + "learning_rate": 2.2510259040779688e-05, + "loss": 0.2621, + "step": 42923 + }, + { + "epoch": 1.65, + "learning_rate": 2.2509617850730956e-05, + "loss": 0.4589, + "step": 42924 + }, + { + "epoch": 1.65, + "learning_rate": 2.2508976660682228e-05, + "loss": 0.2146, + "step": 42925 + }, + { + "epoch": 1.65, + "learning_rate": 2.2508335470633496e-05, + "loss": 0.3482, + "step": 42926 + }, + { + "epoch": 1.65, + "learning_rate": 2.2507694280584765e-05, + "loss": 0.4083, + "step": 42927 + }, + { + "epoch": 1.65, + "learning_rate": 2.2507053090536036e-05, + "loss": 0.4213, + "step": 42928 + }, + { + "epoch": 1.65, + "learning_rate": 2.2506411900487308e-05, + "loss": 0.4021, + "step": 42929 + }, + { + "epoch": 1.65, + "learning_rate": 2.2505770710438576e-05, + "loss": 0.3023, + "step": 42930 + }, + { + "epoch": 1.65, + "learning_rate": 2.2505129520389845e-05, + "loss": 0.3417, + "step": 42931 + }, + { + "epoch": 1.65, + "learning_rate": 2.2504488330341113e-05, + "loss": 0.2174, + "step": 42932 + }, + { + "epoch": 1.65, + "learning_rate": 2.2503847140292385e-05, + "loss": 0.3119, + "step": 42933 + }, + { + "epoch": 1.65, + "learning_rate": 2.2503205950243653e-05, + "loss": 0.4238, + "step": 42934 + }, + { + "epoch": 1.65, + "learning_rate": 2.250256476019492e-05, + "loss": 0.3668, + "step": 42935 + }, + { + "epoch": 1.65, + "learning_rate": 2.250192357014619e-05, + "loss": 0.364, + "step": 42936 + }, + { + "epoch": 1.65, + "learning_rate": 2.2501282380097465e-05, + "loss": 0.4, + "step": 42937 + }, + { + "epoch": 1.65, + "learning_rate": 2.2500641190048733e-05, + "loss": 0.3879, + "step": 42938 + }, + { + "epoch": 1.65, + "learning_rate": 2.25e-05, + "loss": 0.4714, + "step": 42939 + }, + { + "epoch": 1.65, + "learning_rate": 2.249935880995127e-05, + "loss": 0.4378, + "step": 42940 + }, + { + "epoch": 1.65, + "learning_rate": 2.249871761990254e-05, + "loss": 0.3369, + "step": 42941 + }, + { + "epoch": 1.65, + "learning_rate": 2.249807642985381e-05, + "loss": 0.3956, + "step": 42942 + }, + { + "epoch": 1.65, + "learning_rate": 2.2497435239805078e-05, + "loss": 0.2452, + "step": 42943 + }, + { + "epoch": 1.65, + "learning_rate": 2.2496794049756346e-05, + "loss": 0.3548, + "step": 42944 + }, + { + "epoch": 1.65, + "learning_rate": 2.2496152859707618e-05, + "loss": 0.3317, + "step": 42945 + }, + { + "epoch": 1.65, + "learning_rate": 2.249551166965889e-05, + "loss": 0.5048, + "step": 42946 + }, + { + "epoch": 1.65, + "learning_rate": 2.2494870479610158e-05, + "loss": 0.3291, + "step": 42947 + }, + { + "epoch": 1.65, + "learning_rate": 2.2494229289561427e-05, + "loss": 0.2687, + "step": 42948 + }, + { + "epoch": 1.65, + "learning_rate": 2.2493588099512698e-05, + "loss": 0.2871, + "step": 42949 + }, + { + "epoch": 1.65, + "learning_rate": 2.2492946909463967e-05, + "loss": 0.2646, + "step": 42950 + }, + { + "epoch": 1.65, + "learning_rate": 2.2492305719415235e-05, + "loss": 0.4607, + "step": 42951 + }, + { + "epoch": 1.65, + "learning_rate": 2.2491664529366503e-05, + "loss": 0.2437, + "step": 42952 + }, + { + "epoch": 1.65, + "learning_rate": 2.2491023339317775e-05, + "loss": 0.3719, + "step": 42953 + }, + { + "epoch": 1.65, + "learning_rate": 2.2490382149269043e-05, + "loss": 0.3023, + "step": 42954 + }, + { + "epoch": 1.65, + "learning_rate": 2.2489740959220315e-05, + "loss": 0.4553, + "step": 42955 + }, + { + "epoch": 1.65, + "learning_rate": 2.2489099769171583e-05, + "loss": 0.3083, + "step": 42956 + }, + { + "epoch": 1.65, + "learning_rate": 2.2488458579122855e-05, + "loss": 0.3478, + "step": 42957 + }, + { + "epoch": 1.65, + "learning_rate": 2.2487817389074123e-05, + "loss": 0.3631, + "step": 42958 + }, + { + "epoch": 1.65, + "learning_rate": 2.2487176199025392e-05, + "loss": 0.3869, + "step": 42959 + }, + { + "epoch": 1.65, + "learning_rate": 2.248653500897666e-05, + "loss": 0.4025, + "step": 42960 + }, + { + "epoch": 1.65, + "learning_rate": 2.2485893818927932e-05, + "loss": 0.3432, + "step": 42961 + }, + { + "epoch": 1.65, + "learning_rate": 2.24852526288792e-05, + "loss": 0.3578, + "step": 42962 + }, + { + "epoch": 1.65, + "learning_rate": 2.248461143883047e-05, + "loss": 0.4096, + "step": 42963 + }, + { + "epoch": 1.65, + "learning_rate": 2.248397024878174e-05, + "loss": 0.2875, + "step": 42964 + }, + { + "epoch": 1.65, + "learning_rate": 2.2483329058733012e-05, + "loss": 0.373, + "step": 42965 + }, + { + "epoch": 1.65, + "learning_rate": 2.248268786868428e-05, + "loss": 0.2521, + "step": 42966 + }, + { + "epoch": 1.65, + "learning_rate": 2.248204667863555e-05, + "loss": 0.4835, + "step": 42967 + }, + { + "epoch": 1.65, + "learning_rate": 2.2481405488586817e-05, + "loss": 0.4275, + "step": 42968 + }, + { + "epoch": 1.65, + "learning_rate": 2.248076429853809e-05, + "loss": 0.3003, + "step": 42969 + }, + { + "epoch": 1.65, + "learning_rate": 2.2480123108489357e-05, + "loss": 0.3585, + "step": 42970 + }, + { + "epoch": 1.65, + "learning_rate": 2.2479481918440625e-05, + "loss": 0.3128, + "step": 42971 + }, + { + "epoch": 1.65, + "learning_rate": 2.2478840728391897e-05, + "loss": 0.3508, + "step": 42972 + }, + { + "epoch": 1.65, + "learning_rate": 2.247819953834317e-05, + "loss": 0.4066, + "step": 42973 + }, + { + "epoch": 1.65, + "learning_rate": 2.2477558348294437e-05, + "loss": 0.29, + "step": 42974 + }, + { + "epoch": 1.65, + "learning_rate": 2.2476917158245705e-05, + "loss": 0.3576, + "step": 42975 + }, + { + "epoch": 1.65, + "learning_rate": 2.2476275968196974e-05, + "loss": 0.3508, + "step": 42976 + }, + { + "epoch": 1.65, + "learning_rate": 2.2475634778148245e-05, + "loss": 0.3584, + "step": 42977 + }, + { + "epoch": 1.65, + "learning_rate": 2.2474993588099514e-05, + "loss": 0.3419, + "step": 42978 + }, + { + "epoch": 1.65, + "learning_rate": 2.2474352398050782e-05, + "loss": 0.3647, + "step": 42979 + }, + { + "epoch": 1.65, + "learning_rate": 2.247371120800205e-05, + "loss": 0.391, + "step": 42980 + }, + { + "epoch": 1.65, + "learning_rate": 2.2473070017953325e-05, + "loss": 0.3359, + "step": 42981 + }, + { + "epoch": 1.65, + "learning_rate": 2.2472428827904594e-05, + "loss": 0.3599, + "step": 42982 + }, + { + "epoch": 1.65, + "learning_rate": 2.2471787637855862e-05, + "loss": 0.4581, + "step": 42983 + }, + { + "epoch": 1.65, + "learning_rate": 2.247114644780713e-05, + "loss": 0.3377, + "step": 42984 + }, + { + "epoch": 1.65, + "learning_rate": 2.2470505257758402e-05, + "loss": 0.3881, + "step": 42985 + }, + { + "epoch": 1.65, + "learning_rate": 2.246986406770967e-05, + "loss": 0.3699, + "step": 42986 + }, + { + "epoch": 1.65, + "learning_rate": 2.246922287766094e-05, + "loss": 0.3458, + "step": 42987 + }, + { + "epoch": 1.65, + "learning_rate": 2.2468581687612207e-05, + "loss": 0.3372, + "step": 42988 + }, + { + "epoch": 1.65, + "learning_rate": 2.246794049756348e-05, + "loss": 0.2983, + "step": 42989 + }, + { + "epoch": 1.65, + "learning_rate": 2.246729930751475e-05, + "loss": 0.4073, + "step": 42990 + }, + { + "epoch": 1.65, + "learning_rate": 2.246665811746602e-05, + "loss": 0.3944, + "step": 42991 + }, + { + "epoch": 1.65, + "learning_rate": 2.2466016927417287e-05, + "loss": 0.4115, + "step": 42992 + }, + { + "epoch": 1.65, + "learning_rate": 2.246537573736856e-05, + "loss": 0.2282, + "step": 42993 + }, + { + "epoch": 1.65, + "learning_rate": 2.2464734547319827e-05, + "loss": 0.4586, + "step": 42994 + }, + { + "epoch": 1.65, + "learning_rate": 2.2464093357271096e-05, + "loss": 0.3865, + "step": 42995 + }, + { + "epoch": 1.65, + "learning_rate": 2.2463452167222364e-05, + "loss": 0.2692, + "step": 42996 + }, + { + "epoch": 1.65, + "learning_rate": 2.2462810977173636e-05, + "loss": 0.3443, + "step": 42997 + }, + { + "epoch": 1.65, + "learning_rate": 2.2462169787124904e-05, + "loss": 0.4571, + "step": 42998 + }, + { + "epoch": 1.65, + "learning_rate": 2.2461528597076176e-05, + "loss": 0.3376, + "step": 42999 + }, + { + "epoch": 1.65, + "learning_rate": 2.2460887407027444e-05, + "loss": 0.297, + "step": 43000 + }, + { + "epoch": 1.65, + "learning_rate": 2.2460246216978716e-05, + "loss": 0.3185, + "step": 43001 + }, + { + "epoch": 1.65, + "learning_rate": 2.2459605026929984e-05, + "loss": 0.3591, + "step": 43002 + }, + { + "epoch": 1.65, + "learning_rate": 2.2458963836881252e-05, + "loss": 0.4298, + "step": 43003 + }, + { + "epoch": 1.65, + "learning_rate": 2.245832264683252e-05, + "loss": 0.3849, + "step": 43004 + }, + { + "epoch": 1.65, + "learning_rate": 2.2457681456783792e-05, + "loss": 0.3815, + "step": 43005 + }, + { + "epoch": 1.65, + "learning_rate": 2.245704026673506e-05, + "loss": 0.1956, + "step": 43006 + }, + { + "epoch": 1.65, + "learning_rate": 2.245639907668633e-05, + "loss": 0.252, + "step": 43007 + }, + { + "epoch": 1.65, + "learning_rate": 2.24557578866376e-05, + "loss": 0.444, + "step": 43008 + }, + { + "epoch": 1.65, + "learning_rate": 2.2455116696588873e-05, + "loss": 0.4156, + "step": 43009 + }, + { + "epoch": 1.65, + "learning_rate": 2.245447550654014e-05, + "loss": 0.3666, + "step": 43010 + }, + { + "epoch": 1.65, + "learning_rate": 2.245383431649141e-05, + "loss": 0.3511, + "step": 43011 + }, + { + "epoch": 1.65, + "learning_rate": 2.2453193126442677e-05, + "loss": 0.468, + "step": 43012 + }, + { + "epoch": 1.65, + "learning_rate": 2.245255193639395e-05, + "loss": 0.417, + "step": 43013 + }, + { + "epoch": 1.65, + "learning_rate": 2.2451910746345218e-05, + "loss": 0.336, + "step": 43014 + }, + { + "epoch": 1.65, + "learning_rate": 2.2451269556296486e-05, + "loss": 0.4, + "step": 43015 + }, + { + "epoch": 1.65, + "learning_rate": 2.2450628366247754e-05, + "loss": 0.2638, + "step": 43016 + }, + { + "epoch": 1.65, + "learning_rate": 2.244998717619903e-05, + "loss": 0.2468, + "step": 43017 + }, + { + "epoch": 1.65, + "learning_rate": 2.2449345986150298e-05, + "loss": 0.3637, + "step": 43018 + }, + { + "epoch": 1.65, + "learning_rate": 2.2448704796101566e-05, + "loss": 0.2713, + "step": 43019 + }, + { + "epoch": 1.65, + "learning_rate": 2.2448063606052834e-05, + "loss": 0.4685, + "step": 43020 + }, + { + "epoch": 1.65, + "learning_rate": 2.2447422416004106e-05, + "loss": 0.3111, + "step": 43021 + }, + { + "epoch": 1.65, + "learning_rate": 2.2446781225955374e-05, + "loss": 0.3137, + "step": 43022 + }, + { + "epoch": 1.65, + "learning_rate": 2.2446140035906643e-05, + "loss": 0.3268, + "step": 43023 + }, + { + "epoch": 1.65, + "learning_rate": 2.244549884585791e-05, + "loss": 0.2537, + "step": 43024 + }, + { + "epoch": 1.65, + "learning_rate": 2.2444857655809183e-05, + "loss": 0.3174, + "step": 43025 + }, + { + "epoch": 1.65, + "learning_rate": 2.2444216465760454e-05, + "loss": 0.3197, + "step": 43026 + }, + { + "epoch": 1.65, + "learning_rate": 2.2443575275711723e-05, + "loss": 0.3452, + "step": 43027 + }, + { + "epoch": 1.65, + "learning_rate": 2.244293408566299e-05, + "loss": 0.2951, + "step": 43028 + }, + { + "epoch": 1.65, + "learning_rate": 2.2442292895614263e-05, + "loss": 0.351, + "step": 43029 + }, + { + "epoch": 1.65, + "learning_rate": 2.244165170556553e-05, + "loss": 0.3013, + "step": 43030 + }, + { + "epoch": 1.65, + "learning_rate": 2.24410105155168e-05, + "loss": 0.4568, + "step": 43031 + }, + { + "epoch": 1.65, + "learning_rate": 2.2440369325468068e-05, + "loss": 0.2495, + "step": 43032 + }, + { + "epoch": 1.65, + "learning_rate": 2.243972813541934e-05, + "loss": 0.3447, + "step": 43033 + }, + { + "epoch": 1.65, + "learning_rate": 2.2439086945370608e-05, + "loss": 0.3504, + "step": 43034 + }, + { + "epoch": 1.65, + "learning_rate": 2.243844575532188e-05, + "loss": 0.4739, + "step": 43035 + }, + { + "epoch": 1.65, + "learning_rate": 2.2437804565273148e-05, + "loss": 0.2413, + "step": 43036 + }, + { + "epoch": 1.65, + "learning_rate": 2.243716337522442e-05, + "loss": 0.3092, + "step": 43037 + }, + { + "epoch": 1.65, + "learning_rate": 2.2436522185175688e-05, + "loss": 0.3868, + "step": 43038 + }, + { + "epoch": 1.65, + "learning_rate": 2.2435880995126956e-05, + "loss": 0.2547, + "step": 43039 + }, + { + "epoch": 1.65, + "learning_rate": 2.2435239805078225e-05, + "loss": 0.3687, + "step": 43040 + }, + { + "epoch": 1.65, + "learning_rate": 2.2434598615029496e-05, + "loss": 0.2744, + "step": 43041 + }, + { + "epoch": 1.65, + "learning_rate": 2.2433957424980765e-05, + "loss": 0.2484, + "step": 43042 + }, + { + "epoch": 1.65, + "learning_rate": 2.2433316234932033e-05, + "loss": 0.2902, + "step": 43043 + }, + { + "epoch": 1.65, + "learning_rate": 2.2432675044883305e-05, + "loss": 0.3227, + "step": 43044 + }, + { + "epoch": 1.65, + "learning_rate": 2.2432033854834576e-05, + "loss": 0.3159, + "step": 43045 + }, + { + "epoch": 1.65, + "learning_rate": 2.2431392664785845e-05, + "loss": 0.3157, + "step": 43046 + }, + { + "epoch": 1.66, + "learning_rate": 2.2430751474737113e-05, + "loss": 0.4066, + "step": 43047 + }, + { + "epoch": 1.66, + "learning_rate": 2.243011028468838e-05, + "loss": 0.4617, + "step": 43048 + }, + { + "epoch": 1.66, + "learning_rate": 2.2429469094639653e-05, + "loss": 0.3621, + "step": 43049 + }, + { + "epoch": 1.66, + "learning_rate": 2.242882790459092e-05, + "loss": 0.3165, + "step": 43050 + }, + { + "epoch": 1.66, + "learning_rate": 2.242818671454219e-05, + "loss": 0.2844, + "step": 43051 + }, + { + "epoch": 1.66, + "learning_rate": 2.2427545524493458e-05, + "loss": 0.3186, + "step": 43052 + }, + { + "epoch": 1.66, + "learning_rate": 2.2426904334444733e-05, + "loss": 0.4665, + "step": 43053 + }, + { + "epoch": 1.66, + "learning_rate": 2.2426263144396e-05, + "loss": 0.4404, + "step": 43054 + }, + { + "epoch": 1.66, + "learning_rate": 2.242562195434727e-05, + "loss": 0.2455, + "step": 43055 + }, + { + "epoch": 1.66, + "learning_rate": 2.2424980764298538e-05, + "loss": 0.4858, + "step": 43056 + }, + { + "epoch": 1.66, + "learning_rate": 2.242433957424981e-05, + "loss": 0.2555, + "step": 43057 + }, + { + "epoch": 1.66, + "learning_rate": 2.2423698384201078e-05, + "loss": 0.512, + "step": 43058 + }, + { + "epoch": 1.66, + "learning_rate": 2.2423057194152347e-05, + "loss": 0.3264, + "step": 43059 + }, + { + "epoch": 1.66, + "learning_rate": 2.2422416004103615e-05, + "loss": 0.3404, + "step": 43060 + }, + { + "epoch": 1.66, + "learning_rate": 2.2421774814054887e-05, + "loss": 0.3524, + "step": 43061 + }, + { + "epoch": 1.66, + "learning_rate": 2.2421133624006158e-05, + "loss": 0.391, + "step": 43062 + }, + { + "epoch": 1.66, + "learning_rate": 2.2420492433957427e-05, + "loss": 0.5474, + "step": 43063 + }, + { + "epoch": 1.66, + "learning_rate": 2.2419851243908695e-05, + "loss": 0.3204, + "step": 43064 + }, + { + "epoch": 1.66, + "learning_rate": 2.2419210053859967e-05, + "loss": 0.3293, + "step": 43065 + }, + { + "epoch": 1.66, + "learning_rate": 2.2418568863811235e-05, + "loss": 0.2012, + "step": 43066 + }, + { + "epoch": 1.66, + "learning_rate": 2.2417927673762503e-05, + "loss": 0.2715, + "step": 43067 + }, + { + "epoch": 1.66, + "learning_rate": 2.241728648371377e-05, + "loss": 0.2877, + "step": 43068 + }, + { + "epoch": 1.66, + "learning_rate": 2.2416645293665043e-05, + "loss": 0.398, + "step": 43069 + }, + { + "epoch": 1.66, + "learning_rate": 2.2416004103616315e-05, + "loss": 0.2656, + "step": 43070 + }, + { + "epoch": 1.66, + "learning_rate": 2.2415362913567583e-05, + "loss": 0.5454, + "step": 43071 + }, + { + "epoch": 1.66, + "learning_rate": 2.2414721723518852e-05, + "loss": 0.3441, + "step": 43072 + }, + { + "epoch": 1.66, + "learning_rate": 2.2414080533470123e-05, + "loss": 0.4441, + "step": 43073 + }, + { + "epoch": 1.66, + "learning_rate": 2.2413439343421392e-05, + "loss": 0.319, + "step": 43074 + }, + { + "epoch": 1.66, + "learning_rate": 2.241279815337266e-05, + "loss": 0.3502, + "step": 43075 + }, + { + "epoch": 1.66, + "learning_rate": 2.241215696332393e-05, + "loss": 0.6101, + "step": 43076 + }, + { + "epoch": 1.66, + "learning_rate": 2.24115157732752e-05, + "loss": 0.3554, + "step": 43077 + }, + { + "epoch": 1.66, + "learning_rate": 2.241087458322647e-05, + "loss": 0.3441, + "step": 43078 + }, + { + "epoch": 1.66, + "learning_rate": 2.241023339317774e-05, + "loss": 0.3508, + "step": 43079 + }, + { + "epoch": 1.66, + "learning_rate": 2.240959220312901e-05, + "loss": 0.2963, + "step": 43080 + }, + { + "epoch": 1.66, + "learning_rate": 2.240895101308028e-05, + "loss": 0.3295, + "step": 43081 + }, + { + "epoch": 1.66, + "learning_rate": 2.240830982303155e-05, + "loss": 0.4393, + "step": 43082 + }, + { + "epoch": 1.66, + "learning_rate": 2.2407668632982817e-05, + "loss": 0.2842, + "step": 43083 + }, + { + "epoch": 1.66, + "learning_rate": 2.2407027442934085e-05, + "loss": 0.3112, + "step": 43084 + }, + { + "epoch": 1.66, + "learning_rate": 2.2406386252885357e-05, + "loss": 0.3315, + "step": 43085 + }, + { + "epoch": 1.66, + "learning_rate": 2.2405745062836625e-05, + "loss": 0.1869, + "step": 43086 + }, + { + "epoch": 1.66, + "learning_rate": 2.2405103872787894e-05, + "loss": 0.3341, + "step": 43087 + }, + { + "epoch": 1.66, + "learning_rate": 2.2404462682739165e-05, + "loss": 0.3315, + "step": 43088 + }, + { + "epoch": 1.66, + "learning_rate": 2.2403821492690437e-05, + "loss": 0.381, + "step": 43089 + }, + { + "epoch": 1.66, + "learning_rate": 2.2403180302641705e-05, + "loss": 0.3885, + "step": 43090 + }, + { + "epoch": 1.66, + "learning_rate": 2.2402539112592974e-05, + "loss": 0.2987, + "step": 43091 + }, + { + "epoch": 1.66, + "learning_rate": 2.2401897922544242e-05, + "loss": 0.4184, + "step": 43092 + }, + { + "epoch": 1.66, + "learning_rate": 2.2401256732495514e-05, + "loss": 0.4287, + "step": 43093 + }, + { + "epoch": 1.66, + "learning_rate": 2.2400615542446782e-05, + "loss": 0.4475, + "step": 43094 + }, + { + "epoch": 1.66, + "learning_rate": 2.239997435239805e-05, + "loss": 0.3795, + "step": 43095 + }, + { + "epoch": 1.66, + "learning_rate": 2.239933316234932e-05, + "loss": 0.3382, + "step": 43096 + }, + { + "epoch": 1.66, + "learning_rate": 2.2398691972300594e-05, + "loss": 0.4727, + "step": 43097 + }, + { + "epoch": 1.66, + "learning_rate": 2.2398050782251862e-05, + "loss": 0.429, + "step": 43098 + }, + { + "epoch": 1.66, + "learning_rate": 2.239740959220313e-05, + "loss": 0.3131, + "step": 43099 + }, + { + "epoch": 1.66, + "learning_rate": 2.23967684021544e-05, + "loss": 0.3787, + "step": 43100 + }, + { + "epoch": 1.66, + "learning_rate": 2.239612721210567e-05, + "loss": 0.2583, + "step": 43101 + }, + { + "epoch": 1.66, + "learning_rate": 2.239548602205694e-05, + "loss": 0.295, + "step": 43102 + }, + { + "epoch": 1.66, + "learning_rate": 2.2394844832008207e-05, + "loss": 0.2719, + "step": 43103 + }, + { + "epoch": 1.66, + "learning_rate": 2.2394203641959476e-05, + "loss": 0.3768, + "step": 43104 + }, + { + "epoch": 1.66, + "learning_rate": 2.2393562451910747e-05, + "loss": 0.26, + "step": 43105 + }, + { + "epoch": 1.66, + "learning_rate": 2.239292126186202e-05, + "loss": 0.3103, + "step": 43106 + }, + { + "epoch": 1.66, + "learning_rate": 2.2392280071813287e-05, + "loss": 0.243, + "step": 43107 + }, + { + "epoch": 1.66, + "learning_rate": 2.2391638881764556e-05, + "loss": 0.3868, + "step": 43108 + }, + { + "epoch": 1.66, + "learning_rate": 2.2390997691715827e-05, + "loss": 0.3086, + "step": 43109 + }, + { + "epoch": 1.66, + "learning_rate": 2.2390356501667096e-05, + "loss": 0.3655, + "step": 43110 + }, + { + "epoch": 1.66, + "learning_rate": 2.2389715311618364e-05, + "loss": 0.3761, + "step": 43111 + }, + { + "epoch": 1.66, + "learning_rate": 2.2389074121569632e-05, + "loss": 0.3258, + "step": 43112 + }, + { + "epoch": 1.66, + "learning_rate": 2.2388432931520904e-05, + "loss": 0.4709, + "step": 43113 + }, + { + "epoch": 1.66, + "learning_rate": 2.2387791741472172e-05, + "loss": 0.3386, + "step": 43114 + }, + { + "epoch": 1.66, + "learning_rate": 2.2387150551423444e-05, + "loss": 0.3084, + "step": 43115 + }, + { + "epoch": 1.66, + "learning_rate": 2.2386509361374712e-05, + "loss": 0.4351, + "step": 43116 + }, + { + "epoch": 1.66, + "learning_rate": 2.2385868171325984e-05, + "loss": 0.4658, + "step": 43117 + }, + { + "epoch": 1.66, + "learning_rate": 2.2385226981277252e-05, + "loss": 0.4245, + "step": 43118 + }, + { + "epoch": 1.66, + "learning_rate": 2.238458579122852e-05, + "loss": 0.3108, + "step": 43119 + }, + { + "epoch": 1.66, + "learning_rate": 2.238394460117979e-05, + "loss": 0.3627, + "step": 43120 + }, + { + "epoch": 1.66, + "learning_rate": 2.238330341113106e-05, + "loss": 0.281, + "step": 43121 + }, + { + "epoch": 1.66, + "learning_rate": 2.238266222108233e-05, + "loss": 0.4678, + "step": 43122 + }, + { + "epoch": 1.66, + "learning_rate": 2.2382021031033597e-05, + "loss": 0.3055, + "step": 43123 + }, + { + "epoch": 1.66, + "learning_rate": 2.238137984098487e-05, + "loss": 0.4618, + "step": 43124 + }, + { + "epoch": 1.66, + "learning_rate": 2.238073865093614e-05, + "loss": 0.3832, + "step": 43125 + }, + { + "epoch": 1.66, + "learning_rate": 2.238009746088741e-05, + "loss": 0.4319, + "step": 43126 + }, + { + "epoch": 1.66, + "learning_rate": 2.2379456270838678e-05, + "loss": 0.3905, + "step": 43127 + }, + { + "epoch": 1.66, + "learning_rate": 2.2378815080789946e-05, + "loss": 0.5811, + "step": 43128 + }, + { + "epoch": 1.66, + "learning_rate": 2.2378173890741218e-05, + "loss": 0.3712, + "step": 43129 + }, + { + "epoch": 1.66, + "learning_rate": 2.2377532700692486e-05, + "loss": 0.3615, + "step": 43130 + }, + { + "epoch": 1.66, + "learning_rate": 2.2376891510643754e-05, + "loss": 0.3946, + "step": 43131 + }, + { + "epoch": 1.66, + "learning_rate": 2.2376250320595023e-05, + "loss": 0.249, + "step": 43132 + }, + { + "epoch": 1.66, + "learning_rate": 2.2375609130546298e-05, + "loss": 0.3421, + "step": 43133 + }, + { + "epoch": 1.66, + "learning_rate": 2.2374967940497566e-05, + "loss": 0.4418, + "step": 43134 + }, + { + "epoch": 1.66, + "learning_rate": 2.2374326750448834e-05, + "loss": 0.3098, + "step": 43135 + }, + { + "epoch": 1.66, + "learning_rate": 2.2373685560400103e-05, + "loss": 0.3038, + "step": 43136 + }, + { + "epoch": 1.66, + "learning_rate": 2.2373044370351374e-05, + "loss": 0.3982, + "step": 43137 + }, + { + "epoch": 1.66, + "learning_rate": 2.2372403180302643e-05, + "loss": 0.2951, + "step": 43138 + }, + { + "epoch": 1.66, + "learning_rate": 2.237176199025391e-05, + "loss": 0.3326, + "step": 43139 + }, + { + "epoch": 1.66, + "learning_rate": 2.237112080020518e-05, + "loss": 0.4256, + "step": 43140 + }, + { + "epoch": 1.66, + "learning_rate": 2.237047961015645e-05, + "loss": 0.3115, + "step": 43141 + }, + { + "epoch": 1.66, + "learning_rate": 2.2369838420107723e-05, + "loss": 0.371, + "step": 43142 + }, + { + "epoch": 1.66, + "learning_rate": 2.236919723005899e-05, + "loss": 0.2806, + "step": 43143 + }, + { + "epoch": 1.66, + "learning_rate": 2.236855604001026e-05, + "loss": 0.4161, + "step": 43144 + }, + { + "epoch": 1.66, + "learning_rate": 2.236791484996153e-05, + "loss": 0.2908, + "step": 43145 + }, + { + "epoch": 1.66, + "learning_rate": 2.23672736599128e-05, + "loss": 0.4436, + "step": 43146 + }, + { + "epoch": 1.66, + "learning_rate": 2.2366632469864068e-05, + "loss": 0.2532, + "step": 43147 + }, + { + "epoch": 1.66, + "learning_rate": 2.2365991279815336e-05, + "loss": 0.3324, + "step": 43148 + }, + { + "epoch": 1.66, + "learning_rate": 2.2365350089766608e-05, + "loss": 0.4356, + "step": 43149 + }, + { + "epoch": 1.66, + "learning_rate": 2.2364708899717876e-05, + "loss": 0.4583, + "step": 43150 + }, + { + "epoch": 1.66, + "learning_rate": 2.2364067709669148e-05, + "loss": 0.3069, + "step": 43151 + }, + { + "epoch": 1.66, + "learning_rate": 2.2363426519620416e-05, + "loss": 0.2908, + "step": 43152 + }, + { + "epoch": 1.66, + "learning_rate": 2.2362785329571688e-05, + "loss": 0.3295, + "step": 43153 + }, + { + "epoch": 1.66, + "learning_rate": 2.2362144139522956e-05, + "loss": 0.3332, + "step": 43154 + }, + { + "epoch": 1.66, + "learning_rate": 2.2361502949474225e-05, + "loss": 0.3789, + "step": 43155 + }, + { + "epoch": 1.66, + "learning_rate": 2.2360861759425493e-05, + "loss": 0.2764, + "step": 43156 + }, + { + "epoch": 1.66, + "learning_rate": 2.2360220569376765e-05, + "loss": 0.364, + "step": 43157 + }, + { + "epoch": 1.66, + "learning_rate": 2.2359579379328033e-05, + "loss": 0.3607, + "step": 43158 + }, + { + "epoch": 1.66, + "learning_rate": 2.2358938189279305e-05, + "loss": 0.5427, + "step": 43159 + }, + { + "epoch": 1.66, + "learning_rate": 2.2358296999230573e-05, + "loss": 0.2285, + "step": 43160 + }, + { + "epoch": 1.66, + "learning_rate": 2.2357655809181845e-05, + "loss": 0.2458, + "step": 43161 + }, + { + "epoch": 1.66, + "learning_rate": 2.2357014619133113e-05, + "loss": 0.3382, + "step": 43162 + }, + { + "epoch": 1.66, + "learning_rate": 2.235637342908438e-05, + "loss": 0.3651, + "step": 43163 + }, + { + "epoch": 1.66, + "learning_rate": 2.235573223903565e-05, + "loss": 0.4477, + "step": 43164 + }, + { + "epoch": 1.66, + "learning_rate": 2.235509104898692e-05, + "loss": 0.4593, + "step": 43165 + }, + { + "epoch": 1.66, + "learning_rate": 2.235444985893819e-05, + "loss": 0.3902, + "step": 43166 + }, + { + "epoch": 1.66, + "learning_rate": 2.2353808668889458e-05, + "loss": 0.3133, + "step": 43167 + }, + { + "epoch": 1.66, + "learning_rate": 2.235316747884073e-05, + "loss": 0.254, + "step": 43168 + }, + { + "epoch": 1.66, + "learning_rate": 2.2352526288792e-05, + "loss": 0.3225, + "step": 43169 + }, + { + "epoch": 1.66, + "learning_rate": 2.235188509874327e-05, + "loss": 0.2603, + "step": 43170 + }, + { + "epoch": 1.66, + "learning_rate": 2.2351243908694538e-05, + "loss": 0.199, + "step": 43171 + }, + { + "epoch": 1.66, + "learning_rate": 2.2350602718645807e-05, + "loss": 0.2847, + "step": 43172 + }, + { + "epoch": 1.66, + "learning_rate": 2.2349961528597078e-05, + "loss": 0.3416, + "step": 43173 + }, + { + "epoch": 1.66, + "learning_rate": 2.2349320338548347e-05, + "loss": 0.4903, + "step": 43174 + }, + { + "epoch": 1.66, + "learning_rate": 2.2348679148499615e-05, + "loss": 0.2645, + "step": 43175 + }, + { + "epoch": 1.66, + "learning_rate": 2.2348037958450883e-05, + "loss": 0.3971, + "step": 43176 + }, + { + "epoch": 1.66, + "learning_rate": 2.234739676840216e-05, + "loss": 0.3039, + "step": 43177 + }, + { + "epoch": 1.66, + "learning_rate": 2.2346755578353427e-05, + "loss": 0.469, + "step": 43178 + }, + { + "epoch": 1.66, + "learning_rate": 2.2346114388304695e-05, + "loss": 0.4225, + "step": 43179 + }, + { + "epoch": 1.66, + "learning_rate": 2.2345473198255963e-05, + "loss": 0.2418, + "step": 43180 + }, + { + "epoch": 1.66, + "learning_rate": 2.2344832008207235e-05, + "loss": 0.2318, + "step": 43181 + }, + { + "epoch": 1.66, + "learning_rate": 2.2344190818158503e-05, + "loss": 0.4912, + "step": 43182 + }, + { + "epoch": 1.66, + "learning_rate": 2.2343549628109772e-05, + "loss": 0.3617, + "step": 43183 + }, + { + "epoch": 1.66, + "learning_rate": 2.234290843806104e-05, + "loss": 0.2951, + "step": 43184 + }, + { + "epoch": 1.66, + "learning_rate": 2.2342267248012312e-05, + "loss": 0.1964, + "step": 43185 + }, + { + "epoch": 1.66, + "learning_rate": 2.2341626057963583e-05, + "loss": 0.3789, + "step": 43186 + }, + { + "epoch": 1.66, + "learning_rate": 2.2340984867914852e-05, + "loss": 0.3014, + "step": 43187 + }, + { + "epoch": 1.66, + "learning_rate": 2.234034367786612e-05, + "loss": 0.3204, + "step": 43188 + }, + { + "epoch": 1.66, + "learning_rate": 2.2339702487817392e-05, + "loss": 0.3862, + "step": 43189 + }, + { + "epoch": 1.66, + "learning_rate": 2.233906129776866e-05, + "loss": 0.2815, + "step": 43190 + }, + { + "epoch": 1.66, + "learning_rate": 2.233842010771993e-05, + "loss": 0.3936, + "step": 43191 + }, + { + "epoch": 1.66, + "learning_rate": 2.2337778917671197e-05, + "loss": 0.4477, + "step": 43192 + }, + { + "epoch": 1.66, + "learning_rate": 2.233713772762247e-05, + "loss": 0.2638, + "step": 43193 + }, + { + "epoch": 1.66, + "learning_rate": 2.2336496537573737e-05, + "loss": 0.1745, + "step": 43194 + }, + { + "epoch": 1.66, + "learning_rate": 2.233585534752501e-05, + "loss": 0.51, + "step": 43195 + }, + { + "epoch": 1.66, + "learning_rate": 2.2335214157476277e-05, + "loss": 0.2546, + "step": 43196 + }, + { + "epoch": 1.66, + "learning_rate": 2.233457296742755e-05, + "loss": 0.3806, + "step": 43197 + }, + { + "epoch": 1.66, + "learning_rate": 2.2333931777378817e-05, + "loss": 0.4351, + "step": 43198 + }, + { + "epoch": 1.66, + "learning_rate": 2.2333290587330085e-05, + "loss": 0.3191, + "step": 43199 + }, + { + "epoch": 1.66, + "learning_rate": 2.2332649397281354e-05, + "loss": 0.32, + "step": 43200 + }, + { + "epoch": 1.66, + "learning_rate": 2.2332008207232625e-05, + "loss": 0.416, + "step": 43201 + }, + { + "epoch": 1.66, + "learning_rate": 2.2331367017183894e-05, + "loss": 0.5555, + "step": 43202 + }, + { + "epoch": 1.66, + "learning_rate": 2.2330725827135162e-05, + "loss": 0.3348, + "step": 43203 + }, + { + "epoch": 1.66, + "learning_rate": 2.2330084637086434e-05, + "loss": 0.2644, + "step": 43204 + }, + { + "epoch": 1.66, + "learning_rate": 2.2329443447037705e-05, + "loss": 0.2663, + "step": 43205 + }, + { + "epoch": 1.66, + "learning_rate": 2.2328802256988974e-05, + "loss": 0.3569, + "step": 43206 + }, + { + "epoch": 1.66, + "learning_rate": 2.2328161066940242e-05, + "loss": 0.4632, + "step": 43207 + }, + { + "epoch": 1.66, + "learning_rate": 2.232751987689151e-05, + "loss": 0.5855, + "step": 43208 + }, + { + "epoch": 1.66, + "learning_rate": 2.2326878686842782e-05, + "loss": 0.3539, + "step": 43209 + }, + { + "epoch": 1.66, + "learning_rate": 2.232623749679405e-05, + "loss": 0.4022, + "step": 43210 + }, + { + "epoch": 1.66, + "learning_rate": 2.232559630674532e-05, + "loss": 0.3242, + "step": 43211 + }, + { + "epoch": 1.66, + "learning_rate": 2.2324955116696587e-05, + "loss": 0.391, + "step": 43212 + }, + { + "epoch": 1.66, + "learning_rate": 2.2324313926647862e-05, + "loss": 0.3357, + "step": 43213 + }, + { + "epoch": 1.66, + "learning_rate": 2.232367273659913e-05, + "loss": 0.3194, + "step": 43214 + }, + { + "epoch": 1.66, + "learning_rate": 2.23230315465504e-05, + "loss": 0.5256, + "step": 43215 + }, + { + "epoch": 1.66, + "learning_rate": 2.2322390356501667e-05, + "loss": 0.4828, + "step": 43216 + }, + { + "epoch": 1.66, + "learning_rate": 2.232174916645294e-05, + "loss": 0.3746, + "step": 43217 + }, + { + "epoch": 1.66, + "learning_rate": 2.2321107976404207e-05, + "loss": 0.4215, + "step": 43218 + }, + { + "epoch": 1.66, + "learning_rate": 2.2320466786355476e-05, + "loss": 0.2755, + "step": 43219 + }, + { + "epoch": 1.66, + "learning_rate": 2.2319825596306744e-05, + "loss": 0.3981, + "step": 43220 + }, + { + "epoch": 1.66, + "learning_rate": 2.2319184406258016e-05, + "loss": 0.3682, + "step": 43221 + }, + { + "epoch": 1.66, + "learning_rate": 2.2318543216209287e-05, + "loss": 0.4033, + "step": 43222 + }, + { + "epoch": 1.66, + "learning_rate": 2.2317902026160556e-05, + "loss": 0.2747, + "step": 43223 + }, + { + "epoch": 1.66, + "learning_rate": 2.2317260836111824e-05, + "loss": 0.3939, + "step": 43224 + }, + { + "epoch": 1.66, + "learning_rate": 2.2316619646063096e-05, + "loss": 0.3459, + "step": 43225 + }, + { + "epoch": 1.66, + "learning_rate": 2.2315978456014364e-05, + "loss": 0.3746, + "step": 43226 + }, + { + "epoch": 1.66, + "learning_rate": 2.2315337265965632e-05, + "loss": 0.2569, + "step": 43227 + }, + { + "epoch": 1.66, + "learning_rate": 2.23146960759169e-05, + "loss": 0.3125, + "step": 43228 + }, + { + "epoch": 1.66, + "learning_rate": 2.2314054885868172e-05, + "loss": 0.4391, + "step": 43229 + }, + { + "epoch": 1.66, + "learning_rate": 2.231341369581944e-05, + "loss": 0.4518, + "step": 43230 + }, + { + "epoch": 1.66, + "learning_rate": 2.2312772505770712e-05, + "loss": 0.3792, + "step": 43231 + }, + { + "epoch": 1.66, + "learning_rate": 2.231213131572198e-05, + "loss": 0.3109, + "step": 43232 + }, + { + "epoch": 1.66, + "learning_rate": 2.2311490125673253e-05, + "loss": 0.27, + "step": 43233 + }, + { + "epoch": 1.66, + "learning_rate": 2.231084893562452e-05, + "loss": 0.4018, + "step": 43234 + }, + { + "epoch": 1.66, + "learning_rate": 2.231020774557579e-05, + "loss": 0.2309, + "step": 43235 + }, + { + "epoch": 1.66, + "learning_rate": 2.2309566555527058e-05, + "loss": 0.3337, + "step": 43236 + }, + { + "epoch": 1.66, + "learning_rate": 2.230892536547833e-05, + "loss": 0.2327, + "step": 43237 + }, + { + "epoch": 1.66, + "learning_rate": 2.2308284175429598e-05, + "loss": 0.3485, + "step": 43238 + }, + { + "epoch": 1.66, + "learning_rate": 2.2307642985380866e-05, + "loss": 0.4491, + "step": 43239 + }, + { + "epoch": 1.66, + "learning_rate": 2.2307001795332138e-05, + "loss": 0.2825, + "step": 43240 + }, + { + "epoch": 1.66, + "learning_rate": 2.230636060528341e-05, + "loss": 0.4295, + "step": 43241 + }, + { + "epoch": 1.66, + "learning_rate": 2.2305719415234678e-05, + "loss": 0.3875, + "step": 43242 + }, + { + "epoch": 1.66, + "learning_rate": 2.2305078225185946e-05, + "loss": 0.3398, + "step": 43243 + }, + { + "epoch": 1.66, + "learning_rate": 2.2304437035137214e-05, + "loss": 0.2441, + "step": 43244 + }, + { + "epoch": 1.66, + "learning_rate": 2.2303795845088486e-05, + "loss": 0.3489, + "step": 43245 + }, + { + "epoch": 1.66, + "learning_rate": 2.2303154655039754e-05, + "loss": 0.3194, + "step": 43246 + }, + { + "epoch": 1.66, + "learning_rate": 2.2302513464991023e-05, + "loss": 0.2583, + "step": 43247 + }, + { + "epoch": 1.66, + "learning_rate": 2.230187227494229e-05, + "loss": 0.3623, + "step": 43248 + }, + { + "epoch": 1.66, + "learning_rate": 2.2301231084893566e-05, + "loss": 0.3652, + "step": 43249 + }, + { + "epoch": 1.66, + "learning_rate": 2.2300589894844834e-05, + "loss": 0.3295, + "step": 43250 + }, + { + "epoch": 1.66, + "learning_rate": 2.2299948704796103e-05, + "loss": 0.4348, + "step": 43251 + }, + { + "epoch": 1.66, + "learning_rate": 2.229930751474737e-05, + "loss": 0.5153, + "step": 43252 + }, + { + "epoch": 1.66, + "learning_rate": 2.2298666324698643e-05, + "loss": 0.3315, + "step": 43253 + }, + { + "epoch": 1.66, + "learning_rate": 2.229802513464991e-05, + "loss": 0.3454, + "step": 43254 + }, + { + "epoch": 1.66, + "learning_rate": 2.229738394460118e-05, + "loss": 0.385, + "step": 43255 + }, + { + "epoch": 1.66, + "learning_rate": 2.2296742754552448e-05, + "loss": 0.4051, + "step": 43256 + }, + { + "epoch": 1.66, + "learning_rate": 2.2296101564503723e-05, + "loss": 0.4368, + "step": 43257 + }, + { + "epoch": 1.66, + "learning_rate": 2.229546037445499e-05, + "loss": 0.3061, + "step": 43258 + }, + { + "epoch": 1.66, + "learning_rate": 2.229481918440626e-05, + "loss": 0.264, + "step": 43259 + }, + { + "epoch": 1.66, + "learning_rate": 2.2294177994357528e-05, + "loss": 0.2932, + "step": 43260 + }, + { + "epoch": 1.66, + "learning_rate": 2.22935368043088e-05, + "loss": 0.4301, + "step": 43261 + }, + { + "epoch": 1.66, + "learning_rate": 2.2292895614260068e-05, + "loss": 0.3472, + "step": 43262 + }, + { + "epoch": 1.66, + "learning_rate": 2.2292254424211336e-05, + "loss": 0.2264, + "step": 43263 + }, + { + "epoch": 1.66, + "learning_rate": 2.2291613234162605e-05, + "loss": 0.3919, + "step": 43264 + }, + { + "epoch": 1.66, + "learning_rate": 2.2290972044113876e-05, + "loss": 0.3299, + "step": 43265 + }, + { + "epoch": 1.66, + "learning_rate": 2.2290330854065148e-05, + "loss": 0.4313, + "step": 43266 + }, + { + "epoch": 1.66, + "learning_rate": 2.2289689664016416e-05, + "loss": 0.3597, + "step": 43267 + }, + { + "epoch": 1.66, + "learning_rate": 2.2289048473967685e-05, + "loss": 0.4128, + "step": 43268 + }, + { + "epoch": 1.66, + "learning_rate": 2.2288407283918956e-05, + "loss": 0.2905, + "step": 43269 + }, + { + "epoch": 1.66, + "learning_rate": 2.2287766093870225e-05, + "loss": 0.3787, + "step": 43270 + }, + { + "epoch": 1.66, + "learning_rate": 2.2287124903821493e-05, + "loss": 0.3354, + "step": 43271 + }, + { + "epoch": 1.66, + "learning_rate": 2.228648371377276e-05, + "loss": 0.3691, + "step": 43272 + }, + { + "epoch": 1.66, + "learning_rate": 2.2285842523724033e-05, + "loss": 0.485, + "step": 43273 + }, + { + "epoch": 1.66, + "learning_rate": 2.22852013336753e-05, + "loss": 0.3923, + "step": 43274 + }, + { + "epoch": 1.66, + "learning_rate": 2.2284560143626573e-05, + "loss": 0.2701, + "step": 43275 + }, + { + "epoch": 1.66, + "learning_rate": 2.228391895357784e-05, + "loss": 0.4957, + "step": 43276 + }, + { + "epoch": 1.66, + "learning_rate": 2.2283277763529113e-05, + "loss": 0.3951, + "step": 43277 + }, + { + "epoch": 1.66, + "learning_rate": 2.228263657348038e-05, + "loss": 0.3812, + "step": 43278 + }, + { + "epoch": 1.66, + "learning_rate": 2.228199538343165e-05, + "loss": 0.3506, + "step": 43279 + }, + { + "epoch": 1.66, + "learning_rate": 2.2281354193382918e-05, + "loss": 0.4503, + "step": 43280 + }, + { + "epoch": 1.66, + "learning_rate": 2.228071300333419e-05, + "loss": 0.2598, + "step": 43281 + }, + { + "epoch": 1.66, + "learning_rate": 2.2280071813285458e-05, + "loss": 0.3125, + "step": 43282 + }, + { + "epoch": 1.66, + "learning_rate": 2.2279430623236727e-05, + "loss": 0.3175, + "step": 43283 + }, + { + "epoch": 1.66, + "learning_rate": 2.2278789433187998e-05, + "loss": 0.3002, + "step": 43284 + }, + { + "epoch": 1.66, + "learning_rate": 2.227814824313927e-05, + "loss": 0.2936, + "step": 43285 + }, + { + "epoch": 1.66, + "learning_rate": 2.227750705309054e-05, + "loss": 0.4334, + "step": 43286 + }, + { + "epoch": 1.66, + "learning_rate": 2.2276865863041807e-05, + "loss": 0.3353, + "step": 43287 + }, + { + "epoch": 1.66, + "learning_rate": 2.2276224672993075e-05, + "loss": 0.3996, + "step": 43288 + }, + { + "epoch": 1.66, + "learning_rate": 2.2275583482944347e-05, + "loss": 0.3659, + "step": 43289 + }, + { + "epoch": 1.66, + "learning_rate": 2.2274942292895615e-05, + "loss": 0.4296, + "step": 43290 + }, + { + "epoch": 1.66, + "learning_rate": 2.2274301102846883e-05, + "loss": 0.2766, + "step": 43291 + }, + { + "epoch": 1.66, + "learning_rate": 2.227365991279815e-05, + "loss": 0.258, + "step": 43292 + }, + { + "epoch": 1.66, + "learning_rate": 2.2273018722749427e-05, + "loss": 0.3228, + "step": 43293 + }, + { + "epoch": 1.66, + "learning_rate": 2.2272377532700695e-05, + "loss": 0.4514, + "step": 43294 + }, + { + "epoch": 1.66, + "learning_rate": 2.2271736342651963e-05, + "loss": 0.4432, + "step": 43295 + }, + { + "epoch": 1.66, + "learning_rate": 2.2271095152603232e-05, + "loss": 0.3981, + "step": 43296 + }, + { + "epoch": 1.66, + "learning_rate": 2.2270453962554503e-05, + "loss": 0.3635, + "step": 43297 + }, + { + "epoch": 1.66, + "learning_rate": 2.2269812772505772e-05, + "loss": 0.2664, + "step": 43298 + }, + { + "epoch": 1.66, + "learning_rate": 2.226917158245704e-05, + "loss": 0.4607, + "step": 43299 + }, + { + "epoch": 1.66, + "learning_rate": 2.226853039240831e-05, + "loss": 0.2595, + "step": 43300 + }, + { + "epoch": 1.66, + "learning_rate": 2.226788920235958e-05, + "loss": 0.4772, + "step": 43301 + }, + { + "epoch": 1.66, + "learning_rate": 2.2267248012310852e-05, + "loss": 0.3031, + "step": 43302 + }, + { + "epoch": 1.66, + "learning_rate": 2.226660682226212e-05, + "loss": 0.2832, + "step": 43303 + }, + { + "epoch": 1.66, + "learning_rate": 2.226596563221339e-05, + "loss": 0.4025, + "step": 43304 + }, + { + "epoch": 1.66, + "learning_rate": 2.226532444216466e-05, + "loss": 0.4089, + "step": 43305 + }, + { + "epoch": 1.66, + "learning_rate": 2.226468325211593e-05, + "loss": 0.3044, + "step": 43306 + }, + { + "epoch": 1.67, + "learning_rate": 2.2264042062067197e-05, + "loss": 0.2988, + "step": 43307 + }, + { + "epoch": 1.67, + "learning_rate": 2.2263400872018465e-05, + "loss": 0.335, + "step": 43308 + }, + { + "epoch": 1.67, + "learning_rate": 2.2262759681969737e-05, + "loss": 0.2835, + "step": 43309 + }, + { + "epoch": 1.67, + "learning_rate": 2.2262118491921005e-05, + "loss": 0.2768, + "step": 43310 + }, + { + "epoch": 1.67, + "learning_rate": 2.2261477301872277e-05, + "loss": 0.2343, + "step": 43311 + }, + { + "epoch": 1.67, + "learning_rate": 2.2260836111823545e-05, + "loss": 0.3195, + "step": 43312 + }, + { + "epoch": 1.67, + "learning_rate": 2.2260194921774817e-05, + "loss": 0.3861, + "step": 43313 + }, + { + "epoch": 1.67, + "learning_rate": 2.2259553731726085e-05, + "loss": 0.3634, + "step": 43314 + }, + { + "epoch": 1.67, + "learning_rate": 2.2258912541677354e-05, + "loss": 0.2951, + "step": 43315 + }, + { + "epoch": 1.67, + "learning_rate": 2.2258271351628622e-05, + "loss": 0.5235, + "step": 43316 + }, + { + "epoch": 1.67, + "learning_rate": 2.2257630161579894e-05, + "loss": 0.3803, + "step": 43317 + }, + { + "epoch": 1.67, + "learning_rate": 2.2256988971531162e-05, + "loss": 0.4209, + "step": 43318 + }, + { + "epoch": 1.67, + "learning_rate": 2.225634778148243e-05, + "loss": 0.344, + "step": 43319 + }, + { + "epoch": 1.67, + "learning_rate": 2.2255706591433702e-05, + "loss": 0.3812, + "step": 43320 + }, + { + "epoch": 1.67, + "learning_rate": 2.2255065401384974e-05, + "loss": 0.1923, + "step": 43321 + }, + { + "epoch": 1.67, + "learning_rate": 2.2254424211336242e-05, + "loss": 0.3812, + "step": 43322 + }, + { + "epoch": 1.67, + "learning_rate": 2.225378302128751e-05, + "loss": 0.3921, + "step": 43323 + }, + { + "epoch": 1.67, + "learning_rate": 2.225314183123878e-05, + "loss": 0.4269, + "step": 43324 + }, + { + "epoch": 1.67, + "learning_rate": 2.225250064119005e-05, + "loss": 0.2847, + "step": 43325 + }, + { + "epoch": 1.67, + "learning_rate": 2.225185945114132e-05, + "loss": 0.339, + "step": 43326 + }, + { + "epoch": 1.67, + "learning_rate": 2.2251218261092587e-05, + "loss": 0.2438, + "step": 43327 + }, + { + "epoch": 1.67, + "learning_rate": 2.2250577071043856e-05, + "loss": 0.2732, + "step": 43328 + }, + { + "epoch": 1.67, + "learning_rate": 2.224993588099513e-05, + "loss": 0.4045, + "step": 43329 + }, + { + "epoch": 1.67, + "learning_rate": 2.22492946909464e-05, + "loss": 0.3009, + "step": 43330 + }, + { + "epoch": 1.67, + "learning_rate": 2.2248653500897667e-05, + "loss": 0.2646, + "step": 43331 + }, + { + "epoch": 1.67, + "learning_rate": 2.2248012310848936e-05, + "loss": 0.3036, + "step": 43332 + }, + { + "epoch": 1.67, + "learning_rate": 2.2247371120800207e-05, + "loss": 0.3344, + "step": 43333 + }, + { + "epoch": 1.67, + "learning_rate": 2.2246729930751476e-05, + "loss": 0.3403, + "step": 43334 + }, + { + "epoch": 1.67, + "learning_rate": 2.2246088740702744e-05, + "loss": 0.2716, + "step": 43335 + }, + { + "epoch": 1.67, + "learning_rate": 2.2245447550654012e-05, + "loss": 0.2266, + "step": 43336 + }, + { + "epoch": 1.67, + "learning_rate": 2.2244806360605284e-05, + "loss": 0.3503, + "step": 43337 + }, + { + "epoch": 1.67, + "learning_rate": 2.2244165170556556e-05, + "loss": 0.3483, + "step": 43338 + }, + { + "epoch": 1.67, + "learning_rate": 2.2243523980507824e-05, + "loss": 0.3266, + "step": 43339 + }, + { + "epoch": 1.67, + "learning_rate": 2.2242882790459092e-05, + "loss": 0.3783, + "step": 43340 + }, + { + "epoch": 1.67, + "learning_rate": 2.2242241600410364e-05, + "loss": 0.3325, + "step": 43341 + }, + { + "epoch": 1.67, + "learning_rate": 2.2241600410361632e-05, + "loss": 0.4061, + "step": 43342 + }, + { + "epoch": 1.67, + "learning_rate": 2.22409592203129e-05, + "loss": 0.3358, + "step": 43343 + }, + { + "epoch": 1.67, + "learning_rate": 2.224031803026417e-05, + "loss": 0.2687, + "step": 43344 + }, + { + "epoch": 1.67, + "learning_rate": 2.223967684021544e-05, + "loss": 0.3875, + "step": 43345 + }, + { + "epoch": 1.67, + "learning_rate": 2.2239035650166713e-05, + "loss": 0.3998, + "step": 43346 + }, + { + "epoch": 1.67, + "learning_rate": 2.223839446011798e-05, + "loss": 0.3933, + "step": 43347 + }, + { + "epoch": 1.67, + "learning_rate": 2.223775327006925e-05, + "loss": 0.3124, + "step": 43348 + }, + { + "epoch": 1.67, + "learning_rate": 2.223711208002052e-05, + "loss": 0.2897, + "step": 43349 + }, + { + "epoch": 1.67, + "learning_rate": 2.223647088997179e-05, + "loss": 0.3177, + "step": 43350 + }, + { + "epoch": 1.67, + "learning_rate": 2.2235829699923058e-05, + "loss": 0.3164, + "step": 43351 + }, + { + "epoch": 1.67, + "learning_rate": 2.2235188509874326e-05, + "loss": 0.3291, + "step": 43352 + }, + { + "epoch": 1.67, + "learning_rate": 2.2234547319825598e-05, + "loss": 0.2465, + "step": 43353 + }, + { + "epoch": 1.67, + "learning_rate": 2.2233906129776866e-05, + "loss": 0.3391, + "step": 43354 + }, + { + "epoch": 1.67, + "learning_rate": 2.2233264939728138e-05, + "loss": 0.3768, + "step": 43355 + }, + { + "epoch": 1.67, + "learning_rate": 2.2232623749679406e-05, + "loss": 0.4377, + "step": 43356 + }, + { + "epoch": 1.67, + "learning_rate": 2.2231982559630678e-05, + "loss": 0.3048, + "step": 43357 + }, + { + "epoch": 1.67, + "learning_rate": 2.2231341369581946e-05, + "loss": 0.3757, + "step": 43358 + }, + { + "epoch": 1.67, + "learning_rate": 2.2230700179533214e-05, + "loss": 0.2637, + "step": 43359 + }, + { + "epoch": 1.67, + "learning_rate": 2.2230058989484483e-05, + "loss": 0.4736, + "step": 43360 + }, + { + "epoch": 1.67, + "learning_rate": 2.2229417799435754e-05, + "loss": 0.2127, + "step": 43361 + }, + { + "epoch": 1.67, + "learning_rate": 2.2228776609387023e-05, + "loss": 0.3947, + "step": 43362 + }, + { + "epoch": 1.67, + "learning_rate": 2.222813541933829e-05, + "loss": 0.2313, + "step": 43363 + }, + { + "epoch": 1.67, + "learning_rate": 2.2227494229289563e-05, + "loss": 0.3317, + "step": 43364 + }, + { + "epoch": 1.67, + "learning_rate": 2.2226853039240835e-05, + "loss": 0.4496, + "step": 43365 + }, + { + "epoch": 1.67, + "learning_rate": 2.2226211849192103e-05, + "loss": 0.3247, + "step": 43366 + }, + { + "epoch": 1.67, + "learning_rate": 2.222557065914337e-05, + "loss": 0.6, + "step": 43367 + }, + { + "epoch": 1.67, + "learning_rate": 2.222492946909464e-05, + "loss": 0.2868, + "step": 43368 + }, + { + "epoch": 1.67, + "learning_rate": 2.222428827904591e-05, + "loss": 0.5503, + "step": 43369 + }, + { + "epoch": 1.67, + "learning_rate": 2.222364708899718e-05, + "loss": 0.31, + "step": 43370 + }, + { + "epoch": 1.67, + "learning_rate": 2.2223005898948448e-05, + "loss": 0.3065, + "step": 43371 + }, + { + "epoch": 1.67, + "learning_rate": 2.2222364708899716e-05, + "loss": 0.3029, + "step": 43372 + }, + { + "epoch": 1.67, + "learning_rate": 2.222172351885099e-05, + "loss": 0.3826, + "step": 43373 + }, + { + "epoch": 1.67, + "learning_rate": 2.222108232880226e-05, + "loss": 0.2748, + "step": 43374 + }, + { + "epoch": 1.67, + "learning_rate": 2.2220441138753528e-05, + "loss": 0.3934, + "step": 43375 + }, + { + "epoch": 1.67, + "learning_rate": 2.2219799948704796e-05, + "loss": 0.4785, + "step": 43376 + }, + { + "epoch": 1.67, + "learning_rate": 2.2219158758656068e-05, + "loss": 0.2182, + "step": 43377 + }, + { + "epoch": 1.67, + "learning_rate": 2.2218517568607336e-05, + "loss": 0.2944, + "step": 43378 + }, + { + "epoch": 1.67, + "learning_rate": 2.2217876378558605e-05, + "loss": 0.269, + "step": 43379 + }, + { + "epoch": 1.67, + "learning_rate": 2.2217235188509873e-05, + "loss": 0.2046, + "step": 43380 + }, + { + "epoch": 1.67, + "learning_rate": 2.2216593998461145e-05, + "loss": 0.4434, + "step": 43381 + }, + { + "epoch": 1.67, + "learning_rate": 2.2215952808412416e-05, + "loss": 0.3035, + "step": 43382 + }, + { + "epoch": 1.67, + "learning_rate": 2.2215311618363685e-05, + "loss": 0.2954, + "step": 43383 + }, + { + "epoch": 1.67, + "learning_rate": 2.2214670428314953e-05, + "loss": 0.3556, + "step": 43384 + }, + { + "epoch": 1.67, + "learning_rate": 2.2214029238266225e-05, + "loss": 0.3707, + "step": 43385 + }, + { + "epoch": 1.67, + "learning_rate": 2.2213388048217493e-05, + "loss": 0.2749, + "step": 43386 + }, + { + "epoch": 1.67, + "learning_rate": 2.221274685816876e-05, + "loss": 0.3073, + "step": 43387 + }, + { + "epoch": 1.67, + "learning_rate": 2.221210566812003e-05, + "loss": 0.3339, + "step": 43388 + }, + { + "epoch": 1.67, + "learning_rate": 2.22114644780713e-05, + "loss": 0.4139, + "step": 43389 + }, + { + "epoch": 1.67, + "learning_rate": 2.221082328802257e-05, + "loss": 0.4624, + "step": 43390 + }, + { + "epoch": 1.67, + "learning_rate": 2.221018209797384e-05, + "loss": 0.2779, + "step": 43391 + }, + { + "epoch": 1.67, + "learning_rate": 2.220954090792511e-05, + "loss": 0.3502, + "step": 43392 + }, + { + "epoch": 1.67, + "learning_rate": 2.220889971787638e-05, + "loss": 0.3964, + "step": 43393 + }, + { + "epoch": 1.67, + "learning_rate": 2.220825852782765e-05, + "loss": 0.2779, + "step": 43394 + }, + { + "epoch": 1.67, + "learning_rate": 2.2207617337778918e-05, + "loss": 0.2733, + "step": 43395 + }, + { + "epoch": 1.67, + "learning_rate": 2.2206976147730187e-05, + "loss": 0.2864, + "step": 43396 + }, + { + "epoch": 1.67, + "learning_rate": 2.2206334957681458e-05, + "loss": 0.3486, + "step": 43397 + }, + { + "epoch": 1.67, + "learning_rate": 2.2205693767632727e-05, + "loss": 0.4293, + "step": 43398 + }, + { + "epoch": 1.67, + "learning_rate": 2.2205052577583995e-05, + "loss": 0.436, + "step": 43399 + }, + { + "epoch": 1.67, + "learning_rate": 2.2204411387535267e-05, + "loss": 0.1666, + "step": 43400 + }, + { + "epoch": 1.67, + "learning_rate": 2.220377019748654e-05, + "loss": 0.2955, + "step": 43401 + }, + { + "epoch": 1.67, + "learning_rate": 2.2203129007437807e-05, + "loss": 0.4261, + "step": 43402 + }, + { + "epoch": 1.67, + "learning_rate": 2.2202487817389075e-05, + "loss": 0.2531, + "step": 43403 + }, + { + "epoch": 1.67, + "learning_rate": 2.2201846627340343e-05, + "loss": 0.3562, + "step": 43404 + }, + { + "epoch": 1.67, + "learning_rate": 2.2201205437291615e-05, + "loss": 0.3873, + "step": 43405 + }, + { + "epoch": 1.67, + "learning_rate": 2.2200564247242883e-05, + "loss": 0.2591, + "step": 43406 + }, + { + "epoch": 1.67, + "learning_rate": 2.2199923057194152e-05, + "loss": 0.3654, + "step": 43407 + }, + { + "epoch": 1.67, + "learning_rate": 2.219928186714542e-05, + "loss": 0.2457, + "step": 43408 + }, + { + "epoch": 1.67, + "learning_rate": 2.2198640677096695e-05, + "loss": 0.3334, + "step": 43409 + }, + { + "epoch": 1.67, + "learning_rate": 2.2197999487047964e-05, + "loss": 0.3383, + "step": 43410 + }, + { + "epoch": 1.67, + "learning_rate": 2.2197358296999232e-05, + "loss": 0.4559, + "step": 43411 + }, + { + "epoch": 1.67, + "learning_rate": 2.21967171069505e-05, + "loss": 0.4405, + "step": 43412 + }, + { + "epoch": 1.67, + "learning_rate": 2.2196075916901772e-05, + "loss": 0.283, + "step": 43413 + }, + { + "epoch": 1.67, + "learning_rate": 2.219543472685304e-05, + "loss": 0.354, + "step": 43414 + }, + { + "epoch": 1.67, + "learning_rate": 2.219479353680431e-05, + "loss": 0.2959, + "step": 43415 + }, + { + "epoch": 1.67, + "learning_rate": 2.2194152346755577e-05, + "loss": 0.4253, + "step": 43416 + }, + { + "epoch": 1.67, + "learning_rate": 2.219351115670685e-05, + "loss": 0.3523, + "step": 43417 + }, + { + "epoch": 1.67, + "learning_rate": 2.219286996665812e-05, + "loss": 0.4302, + "step": 43418 + }, + { + "epoch": 1.67, + "learning_rate": 2.219222877660939e-05, + "loss": 0.3402, + "step": 43419 + }, + { + "epoch": 1.67, + "learning_rate": 2.2191587586560657e-05, + "loss": 0.2869, + "step": 43420 + }, + { + "epoch": 1.67, + "learning_rate": 2.219094639651193e-05, + "loss": 0.3005, + "step": 43421 + }, + { + "epoch": 1.67, + "learning_rate": 2.2190305206463197e-05, + "loss": 0.431, + "step": 43422 + }, + { + "epoch": 1.67, + "learning_rate": 2.2189664016414465e-05, + "loss": 0.2848, + "step": 43423 + }, + { + "epoch": 1.67, + "learning_rate": 2.2189022826365734e-05, + "loss": 0.3338, + "step": 43424 + }, + { + "epoch": 1.67, + "learning_rate": 2.2188381636317005e-05, + "loss": 0.4572, + "step": 43425 + }, + { + "epoch": 1.67, + "learning_rate": 2.2187740446268274e-05, + "loss": 0.2646, + "step": 43426 + }, + { + "epoch": 1.67, + "learning_rate": 2.2187099256219545e-05, + "loss": 0.1953, + "step": 43427 + }, + { + "epoch": 1.67, + "learning_rate": 2.2186458066170814e-05, + "loss": 0.4802, + "step": 43428 + }, + { + "epoch": 1.67, + "learning_rate": 2.2185816876122085e-05, + "loss": 0.4204, + "step": 43429 + }, + { + "epoch": 1.67, + "learning_rate": 2.2185175686073354e-05, + "loss": 0.3873, + "step": 43430 + }, + { + "epoch": 1.67, + "learning_rate": 2.2184534496024622e-05, + "loss": 0.5122, + "step": 43431 + }, + { + "epoch": 1.67, + "learning_rate": 2.218389330597589e-05, + "loss": 0.2429, + "step": 43432 + }, + { + "epoch": 1.67, + "learning_rate": 2.2183252115927162e-05, + "loss": 0.2938, + "step": 43433 + }, + { + "epoch": 1.67, + "learning_rate": 2.218261092587843e-05, + "loss": 0.3499, + "step": 43434 + }, + { + "epoch": 1.67, + "learning_rate": 2.21819697358297e-05, + "loss": 0.209, + "step": 43435 + }, + { + "epoch": 1.67, + "learning_rate": 2.218132854578097e-05, + "loss": 0.4111, + "step": 43436 + }, + { + "epoch": 1.67, + "learning_rate": 2.2180687355732242e-05, + "loss": 0.2716, + "step": 43437 + }, + { + "epoch": 1.67, + "learning_rate": 2.218004616568351e-05, + "loss": 0.4325, + "step": 43438 + }, + { + "epoch": 1.67, + "learning_rate": 2.217940497563478e-05, + "loss": 0.2414, + "step": 43439 + }, + { + "epoch": 1.67, + "learning_rate": 2.2178763785586047e-05, + "loss": 0.2389, + "step": 43440 + }, + { + "epoch": 1.67, + "learning_rate": 2.217812259553732e-05, + "loss": 0.432, + "step": 43441 + }, + { + "epoch": 1.67, + "learning_rate": 2.2177481405488587e-05, + "loss": 0.3772, + "step": 43442 + }, + { + "epoch": 1.67, + "learning_rate": 2.2176840215439856e-05, + "loss": 0.3653, + "step": 43443 + }, + { + "epoch": 1.67, + "learning_rate": 2.2176199025391127e-05, + "loss": 0.2639, + "step": 43444 + }, + { + "epoch": 1.67, + "learning_rate": 2.21755578353424e-05, + "loss": 0.3471, + "step": 43445 + }, + { + "epoch": 1.67, + "learning_rate": 2.2174916645293667e-05, + "loss": 0.5864, + "step": 43446 + }, + { + "epoch": 1.67, + "learning_rate": 2.2174275455244936e-05, + "loss": 0.4746, + "step": 43447 + }, + { + "epoch": 1.67, + "learning_rate": 2.2173634265196204e-05, + "loss": 0.3182, + "step": 43448 + }, + { + "epoch": 1.67, + "learning_rate": 2.2172993075147476e-05, + "loss": 0.3358, + "step": 43449 + }, + { + "epoch": 1.67, + "learning_rate": 2.2172351885098744e-05, + "loss": 0.391, + "step": 43450 + }, + { + "epoch": 1.67, + "learning_rate": 2.2171710695050012e-05, + "loss": 0.3541, + "step": 43451 + }, + { + "epoch": 1.67, + "learning_rate": 2.217106950500128e-05, + "loss": 0.2606, + "step": 43452 + }, + { + "epoch": 1.67, + "learning_rate": 2.2170428314952556e-05, + "loss": 0.2388, + "step": 43453 + }, + { + "epoch": 1.67, + "learning_rate": 2.2169787124903824e-05, + "loss": 0.3438, + "step": 43454 + }, + { + "epoch": 1.67, + "learning_rate": 2.2169145934855092e-05, + "loss": 0.2477, + "step": 43455 + }, + { + "epoch": 1.67, + "learning_rate": 2.216850474480636e-05, + "loss": 0.3114, + "step": 43456 + }, + { + "epoch": 1.67, + "learning_rate": 2.2167863554757633e-05, + "loss": 0.3116, + "step": 43457 + }, + { + "epoch": 1.67, + "learning_rate": 2.21672223647089e-05, + "loss": 0.3113, + "step": 43458 + }, + { + "epoch": 1.67, + "learning_rate": 2.216658117466017e-05, + "loss": 0.2756, + "step": 43459 + }, + { + "epoch": 1.67, + "learning_rate": 2.2165939984611438e-05, + "loss": 0.2823, + "step": 43460 + }, + { + "epoch": 1.67, + "learning_rate": 2.216529879456271e-05, + "loss": 0.375, + "step": 43461 + }, + { + "epoch": 1.67, + "learning_rate": 2.216465760451398e-05, + "loss": 0.31, + "step": 43462 + }, + { + "epoch": 1.67, + "learning_rate": 2.216401641446525e-05, + "loss": 0.275, + "step": 43463 + }, + { + "epoch": 1.67, + "learning_rate": 2.2163375224416518e-05, + "loss": 0.2672, + "step": 43464 + }, + { + "epoch": 1.67, + "learning_rate": 2.216273403436779e-05, + "loss": 0.3948, + "step": 43465 + }, + { + "epoch": 1.67, + "learning_rate": 2.2162092844319058e-05, + "loss": 0.3858, + "step": 43466 + }, + { + "epoch": 1.67, + "learning_rate": 2.2161451654270326e-05, + "loss": 0.3797, + "step": 43467 + }, + { + "epoch": 1.67, + "learning_rate": 2.2160810464221594e-05, + "loss": 0.3028, + "step": 43468 + }, + { + "epoch": 1.67, + "learning_rate": 2.2160169274172866e-05, + "loss": 0.3462, + "step": 43469 + }, + { + "epoch": 1.67, + "learning_rate": 2.2159528084124134e-05, + "loss": 0.3493, + "step": 43470 + }, + { + "epoch": 1.67, + "learning_rate": 2.2158886894075406e-05, + "loss": 0.347, + "step": 43471 + }, + { + "epoch": 1.67, + "learning_rate": 2.2158245704026674e-05, + "loss": 0.3418, + "step": 43472 + }, + { + "epoch": 1.67, + "learning_rate": 2.2157604513977946e-05, + "loss": 0.2909, + "step": 43473 + }, + { + "epoch": 1.67, + "learning_rate": 2.2156963323929214e-05, + "loss": 0.2846, + "step": 43474 + }, + { + "epoch": 1.67, + "learning_rate": 2.2156322133880483e-05, + "loss": 0.3631, + "step": 43475 + }, + { + "epoch": 1.67, + "learning_rate": 2.215568094383175e-05, + "loss": 0.3938, + "step": 43476 + }, + { + "epoch": 1.67, + "learning_rate": 2.2155039753783023e-05, + "loss": 0.3795, + "step": 43477 + }, + { + "epoch": 1.67, + "learning_rate": 2.215439856373429e-05, + "loss": 0.488, + "step": 43478 + }, + { + "epoch": 1.67, + "learning_rate": 2.215375737368556e-05, + "loss": 0.3655, + "step": 43479 + }, + { + "epoch": 1.67, + "learning_rate": 2.215311618363683e-05, + "loss": 0.3597, + "step": 43480 + }, + { + "epoch": 1.67, + "learning_rate": 2.2152474993588103e-05, + "loss": 0.2811, + "step": 43481 + }, + { + "epoch": 1.67, + "learning_rate": 2.215183380353937e-05, + "loss": 0.4833, + "step": 43482 + }, + { + "epoch": 1.67, + "learning_rate": 2.215119261349064e-05, + "loss": 0.3482, + "step": 43483 + }, + { + "epoch": 1.67, + "learning_rate": 2.2150551423441908e-05, + "loss": 0.4328, + "step": 43484 + }, + { + "epoch": 1.67, + "learning_rate": 2.214991023339318e-05, + "loss": 0.3069, + "step": 43485 + }, + { + "epoch": 1.67, + "learning_rate": 2.2149269043344448e-05, + "loss": 0.3858, + "step": 43486 + }, + { + "epoch": 1.67, + "learning_rate": 2.2148627853295716e-05, + "loss": 0.4006, + "step": 43487 + }, + { + "epoch": 1.67, + "learning_rate": 2.2147986663246985e-05, + "loss": 0.3781, + "step": 43488 + }, + { + "epoch": 1.67, + "learning_rate": 2.214734547319826e-05, + "loss": 0.2134, + "step": 43489 + }, + { + "epoch": 1.67, + "learning_rate": 2.2146704283149528e-05, + "loss": 0.3221, + "step": 43490 + }, + { + "epoch": 1.67, + "learning_rate": 2.2146063093100796e-05, + "loss": 0.2998, + "step": 43491 + }, + { + "epoch": 1.67, + "learning_rate": 2.2145421903052065e-05, + "loss": 0.2827, + "step": 43492 + }, + { + "epoch": 1.67, + "learning_rate": 2.2144780713003336e-05, + "loss": 0.3828, + "step": 43493 + }, + { + "epoch": 1.67, + "learning_rate": 2.2144139522954605e-05, + "loss": 0.4586, + "step": 43494 + }, + { + "epoch": 1.67, + "learning_rate": 2.2143498332905873e-05, + "loss": 0.2783, + "step": 43495 + }, + { + "epoch": 1.67, + "learning_rate": 2.214285714285714e-05, + "loss": 0.383, + "step": 43496 + }, + { + "epoch": 1.67, + "learning_rate": 2.2142215952808413e-05, + "loss": 0.3213, + "step": 43497 + }, + { + "epoch": 1.67, + "learning_rate": 2.2141574762759685e-05, + "loss": 0.3315, + "step": 43498 + }, + { + "epoch": 1.67, + "learning_rate": 2.2140933572710953e-05, + "loss": 0.4287, + "step": 43499 + }, + { + "epoch": 1.67, + "learning_rate": 2.214029238266222e-05, + "loss": 0.3267, + "step": 43500 + }, + { + "epoch": 1.67, + "learning_rate": 2.2139651192613493e-05, + "loss": 0.4388, + "step": 43501 + }, + { + "epoch": 1.67, + "learning_rate": 2.213901000256476e-05, + "loss": 0.3474, + "step": 43502 + }, + { + "epoch": 1.67, + "learning_rate": 2.213836881251603e-05, + "loss": 0.5114, + "step": 43503 + }, + { + "epoch": 1.67, + "learning_rate": 2.2137727622467298e-05, + "loss": 0.3325, + "step": 43504 + }, + { + "epoch": 1.67, + "learning_rate": 2.213708643241857e-05, + "loss": 0.2513, + "step": 43505 + }, + { + "epoch": 1.67, + "learning_rate": 2.2136445242369838e-05, + "loss": 0.3615, + "step": 43506 + }, + { + "epoch": 1.67, + "learning_rate": 2.213580405232111e-05, + "loss": 0.4488, + "step": 43507 + }, + { + "epoch": 1.67, + "learning_rate": 2.2135162862272378e-05, + "loss": 0.2887, + "step": 43508 + }, + { + "epoch": 1.67, + "learning_rate": 2.213452167222365e-05, + "loss": 0.2984, + "step": 43509 + }, + { + "epoch": 1.67, + "learning_rate": 2.213388048217492e-05, + "loss": 0.4059, + "step": 43510 + }, + { + "epoch": 1.67, + "learning_rate": 2.2133239292126187e-05, + "loss": 0.4657, + "step": 43511 + }, + { + "epoch": 1.67, + "learning_rate": 2.2132598102077455e-05, + "loss": 0.4454, + "step": 43512 + }, + { + "epoch": 1.67, + "learning_rate": 2.2131956912028727e-05, + "loss": 0.3744, + "step": 43513 + }, + { + "epoch": 1.67, + "learning_rate": 2.2131315721979995e-05, + "loss": 0.3889, + "step": 43514 + }, + { + "epoch": 1.67, + "learning_rate": 2.2130674531931263e-05, + "loss": 0.3857, + "step": 43515 + }, + { + "epoch": 1.67, + "learning_rate": 2.2130033341882535e-05, + "loss": 0.2258, + "step": 43516 + }, + { + "epoch": 1.67, + "learning_rate": 2.2129392151833807e-05, + "loss": 0.2293, + "step": 43517 + }, + { + "epoch": 1.67, + "learning_rate": 2.2128750961785075e-05, + "loss": 0.3972, + "step": 43518 + }, + { + "epoch": 1.67, + "learning_rate": 2.2128109771736343e-05, + "loss": 0.2722, + "step": 43519 + }, + { + "epoch": 1.67, + "learning_rate": 2.2127468581687612e-05, + "loss": 0.3027, + "step": 43520 + }, + { + "epoch": 1.67, + "learning_rate": 2.2126827391638883e-05, + "loss": 0.3953, + "step": 43521 + }, + { + "epoch": 1.67, + "learning_rate": 2.2126186201590152e-05, + "loss": 0.3254, + "step": 43522 + }, + { + "epoch": 1.67, + "learning_rate": 2.212554501154142e-05, + "loss": 0.3586, + "step": 43523 + }, + { + "epoch": 1.67, + "learning_rate": 2.212490382149269e-05, + "loss": 0.3792, + "step": 43524 + }, + { + "epoch": 1.67, + "learning_rate": 2.2124262631443964e-05, + "loss": 0.3935, + "step": 43525 + }, + { + "epoch": 1.67, + "learning_rate": 2.2123621441395232e-05, + "loss": 0.4137, + "step": 43526 + }, + { + "epoch": 1.67, + "learning_rate": 2.21229802513465e-05, + "loss": 0.3232, + "step": 43527 + }, + { + "epoch": 1.67, + "learning_rate": 2.212233906129777e-05, + "loss": 0.2452, + "step": 43528 + }, + { + "epoch": 1.67, + "learning_rate": 2.212169787124904e-05, + "loss": 0.3509, + "step": 43529 + }, + { + "epoch": 1.67, + "learning_rate": 2.212105668120031e-05, + "loss": 0.2856, + "step": 43530 + }, + { + "epoch": 1.67, + "learning_rate": 2.2120415491151577e-05, + "loss": 0.3421, + "step": 43531 + }, + { + "epoch": 1.67, + "learning_rate": 2.2119774301102845e-05, + "loss": 0.2956, + "step": 43532 + }, + { + "epoch": 1.67, + "learning_rate": 2.2119133111054117e-05, + "loss": 0.2825, + "step": 43533 + }, + { + "epoch": 1.67, + "learning_rate": 2.211849192100539e-05, + "loss": 0.2884, + "step": 43534 + }, + { + "epoch": 1.67, + "learning_rate": 2.2117850730956657e-05, + "loss": 0.483, + "step": 43535 + }, + { + "epoch": 1.67, + "learning_rate": 2.2117209540907925e-05, + "loss": 0.4301, + "step": 43536 + }, + { + "epoch": 1.67, + "learning_rate": 2.2116568350859197e-05, + "loss": 0.4136, + "step": 43537 + }, + { + "epoch": 1.67, + "learning_rate": 2.2115927160810465e-05, + "loss": 0.5027, + "step": 43538 + }, + { + "epoch": 1.67, + "learning_rate": 2.2115285970761734e-05, + "loss": 0.3211, + "step": 43539 + }, + { + "epoch": 1.67, + "learning_rate": 2.2114644780713002e-05, + "loss": 0.3787, + "step": 43540 + }, + { + "epoch": 1.67, + "learning_rate": 2.2114003590664274e-05, + "loss": 0.3458, + "step": 43541 + }, + { + "epoch": 1.67, + "learning_rate": 2.2113362400615545e-05, + "loss": 0.3659, + "step": 43542 + }, + { + "epoch": 1.67, + "learning_rate": 2.2112721210566814e-05, + "loss": 0.3874, + "step": 43543 + }, + { + "epoch": 1.67, + "learning_rate": 2.2112080020518082e-05, + "loss": 0.2839, + "step": 43544 + }, + { + "epoch": 1.67, + "learning_rate": 2.2111438830469354e-05, + "loss": 0.425, + "step": 43545 + }, + { + "epoch": 1.67, + "learning_rate": 2.2110797640420622e-05, + "loss": 0.3528, + "step": 43546 + }, + { + "epoch": 1.67, + "learning_rate": 2.211015645037189e-05, + "loss": 0.3486, + "step": 43547 + }, + { + "epoch": 1.67, + "learning_rate": 2.210951526032316e-05, + "loss": 0.3298, + "step": 43548 + }, + { + "epoch": 1.67, + "learning_rate": 2.210887407027443e-05, + "loss": 0.4151, + "step": 43549 + }, + { + "epoch": 1.67, + "learning_rate": 2.21082328802257e-05, + "loss": 0.3709, + "step": 43550 + }, + { + "epoch": 1.67, + "learning_rate": 2.210759169017697e-05, + "loss": 0.2207, + "step": 43551 + }, + { + "epoch": 1.67, + "learning_rate": 2.210695050012824e-05, + "loss": 0.4118, + "step": 43552 + }, + { + "epoch": 1.67, + "learning_rate": 2.210630931007951e-05, + "loss": 0.313, + "step": 43553 + }, + { + "epoch": 1.67, + "learning_rate": 2.210566812003078e-05, + "loss": 0.2948, + "step": 43554 + }, + { + "epoch": 1.67, + "learning_rate": 2.2105026929982047e-05, + "loss": 0.3135, + "step": 43555 + }, + { + "epoch": 1.67, + "learning_rate": 2.2104385739933316e-05, + "loss": 0.4466, + "step": 43556 + }, + { + "epoch": 1.67, + "learning_rate": 2.2103744549884587e-05, + "loss": 0.2741, + "step": 43557 + }, + { + "epoch": 1.67, + "learning_rate": 2.2103103359835856e-05, + "loss": 0.3522, + "step": 43558 + }, + { + "epoch": 1.67, + "learning_rate": 2.2102462169787124e-05, + "loss": 0.2767, + "step": 43559 + }, + { + "epoch": 1.67, + "learning_rate": 2.2101820979738396e-05, + "loss": 0.2266, + "step": 43560 + }, + { + "epoch": 1.67, + "learning_rate": 2.2101179789689667e-05, + "loss": 0.3421, + "step": 43561 + }, + { + "epoch": 1.67, + "learning_rate": 2.2100538599640936e-05, + "loss": 0.2584, + "step": 43562 + }, + { + "epoch": 1.67, + "learning_rate": 2.2099897409592204e-05, + "loss": 0.4972, + "step": 43563 + }, + { + "epoch": 1.67, + "learning_rate": 2.2099256219543472e-05, + "loss": 0.4493, + "step": 43564 + }, + { + "epoch": 1.67, + "learning_rate": 2.2098615029494744e-05, + "loss": 0.4127, + "step": 43565 + }, + { + "epoch": 1.67, + "learning_rate": 2.2097973839446012e-05, + "loss": 0.3604, + "step": 43566 + }, + { + "epoch": 1.68, + "learning_rate": 2.209733264939728e-05, + "loss": 0.4232, + "step": 43567 + }, + { + "epoch": 1.68, + "learning_rate": 2.209669145934855e-05, + "loss": 0.2603, + "step": 43568 + }, + { + "epoch": 1.68, + "learning_rate": 2.2096050269299824e-05, + "loss": 0.504, + "step": 43569 + }, + { + "epoch": 1.68, + "learning_rate": 2.2095409079251093e-05, + "loss": 0.3429, + "step": 43570 + }, + { + "epoch": 1.68, + "learning_rate": 2.209476788920236e-05, + "loss": 0.2812, + "step": 43571 + }, + { + "epoch": 1.68, + "learning_rate": 2.209412669915363e-05, + "loss": 0.3813, + "step": 43572 + }, + { + "epoch": 1.68, + "learning_rate": 2.20934855091049e-05, + "loss": 0.2993, + "step": 43573 + }, + { + "epoch": 1.68, + "learning_rate": 2.209284431905617e-05, + "loss": 0.266, + "step": 43574 + }, + { + "epoch": 1.68, + "learning_rate": 2.2092203129007438e-05, + "loss": 0.4234, + "step": 43575 + }, + { + "epoch": 1.68, + "learning_rate": 2.2091561938958706e-05, + "loss": 0.2786, + "step": 43576 + }, + { + "epoch": 1.68, + "learning_rate": 2.2090920748909978e-05, + "loss": 0.462, + "step": 43577 + }, + { + "epoch": 1.68, + "learning_rate": 2.209027955886125e-05, + "loss": 0.2581, + "step": 43578 + }, + { + "epoch": 1.68, + "learning_rate": 2.2089638368812518e-05, + "loss": 0.4373, + "step": 43579 + }, + { + "epoch": 1.68, + "learning_rate": 2.2088997178763786e-05, + "loss": 0.3766, + "step": 43580 + }, + { + "epoch": 1.68, + "learning_rate": 2.2088355988715058e-05, + "loss": 0.3417, + "step": 43581 + }, + { + "epoch": 1.68, + "learning_rate": 2.2087714798666326e-05, + "loss": 0.3983, + "step": 43582 + }, + { + "epoch": 1.68, + "learning_rate": 2.2087073608617594e-05, + "loss": 0.4093, + "step": 43583 + }, + { + "epoch": 1.68, + "learning_rate": 2.2086432418568863e-05, + "loss": 0.3103, + "step": 43584 + }, + { + "epoch": 1.68, + "learning_rate": 2.2085791228520134e-05, + "loss": 0.3437, + "step": 43585 + }, + { + "epoch": 1.68, + "learning_rate": 2.2085150038471403e-05, + "loss": 0.4328, + "step": 43586 + }, + { + "epoch": 1.68, + "learning_rate": 2.2084508848422674e-05, + "loss": 0.4197, + "step": 43587 + }, + { + "epoch": 1.68, + "learning_rate": 2.2083867658373943e-05, + "loss": 0.3716, + "step": 43588 + }, + { + "epoch": 1.68, + "learning_rate": 2.2083226468325215e-05, + "loss": 0.3131, + "step": 43589 + }, + { + "epoch": 1.68, + "learning_rate": 2.2082585278276483e-05, + "loss": 0.3167, + "step": 43590 + }, + { + "epoch": 1.68, + "learning_rate": 2.208194408822775e-05, + "loss": 0.299, + "step": 43591 + }, + { + "epoch": 1.68, + "learning_rate": 2.208130289817902e-05, + "loss": 0.2198, + "step": 43592 + }, + { + "epoch": 1.68, + "learning_rate": 2.208066170813029e-05, + "loss": 0.3598, + "step": 43593 + }, + { + "epoch": 1.68, + "learning_rate": 2.208002051808156e-05, + "loss": 0.361, + "step": 43594 + }, + { + "epoch": 1.68, + "learning_rate": 2.2079379328032828e-05, + "loss": 0.285, + "step": 43595 + }, + { + "epoch": 1.68, + "learning_rate": 2.20787381379841e-05, + "loss": 0.3839, + "step": 43596 + }, + { + "epoch": 1.68, + "learning_rate": 2.207809694793537e-05, + "loss": 0.4708, + "step": 43597 + }, + { + "epoch": 1.68, + "learning_rate": 2.207745575788664e-05, + "loss": 0.3001, + "step": 43598 + }, + { + "epoch": 1.68, + "learning_rate": 2.2076814567837908e-05, + "loss": 0.3935, + "step": 43599 + }, + { + "epoch": 1.68, + "learning_rate": 2.2076173377789176e-05, + "loss": 0.3667, + "step": 43600 + }, + { + "epoch": 1.68, + "learning_rate": 2.2075532187740448e-05, + "loss": 0.2403, + "step": 43601 + }, + { + "epoch": 1.68, + "learning_rate": 2.2074890997691716e-05, + "loss": 0.2576, + "step": 43602 + }, + { + "epoch": 1.68, + "learning_rate": 2.2074249807642985e-05, + "loss": 0.2461, + "step": 43603 + }, + { + "epoch": 1.68, + "learning_rate": 2.2073608617594253e-05, + "loss": 0.3757, + "step": 43604 + }, + { + "epoch": 1.68, + "learning_rate": 2.2072967427545528e-05, + "loss": 0.3507, + "step": 43605 + }, + { + "epoch": 1.68, + "learning_rate": 2.2072326237496796e-05, + "loss": 0.4023, + "step": 43606 + }, + { + "epoch": 1.68, + "learning_rate": 2.2071685047448065e-05, + "loss": 0.3461, + "step": 43607 + }, + { + "epoch": 1.68, + "learning_rate": 2.2071043857399333e-05, + "loss": 0.3251, + "step": 43608 + }, + { + "epoch": 1.68, + "learning_rate": 2.2070402667350605e-05, + "loss": 0.3172, + "step": 43609 + }, + { + "epoch": 1.68, + "learning_rate": 2.2069761477301873e-05, + "loss": 0.2936, + "step": 43610 + }, + { + "epoch": 1.68, + "learning_rate": 2.206912028725314e-05, + "loss": 0.2599, + "step": 43611 + }, + { + "epoch": 1.68, + "learning_rate": 2.206847909720441e-05, + "loss": 0.3852, + "step": 43612 + }, + { + "epoch": 1.68, + "learning_rate": 2.206783790715568e-05, + "loss": 0.3328, + "step": 43613 + }, + { + "epoch": 1.68, + "learning_rate": 2.2067196717106953e-05, + "loss": 0.3079, + "step": 43614 + }, + { + "epoch": 1.68, + "learning_rate": 2.206655552705822e-05, + "loss": 0.3196, + "step": 43615 + }, + { + "epoch": 1.68, + "learning_rate": 2.206591433700949e-05, + "loss": 0.271, + "step": 43616 + }, + { + "epoch": 1.68, + "learning_rate": 2.206527314696076e-05, + "loss": 0.2021, + "step": 43617 + }, + { + "epoch": 1.68, + "learning_rate": 2.206463195691203e-05, + "loss": 0.4055, + "step": 43618 + }, + { + "epoch": 1.68, + "learning_rate": 2.2063990766863298e-05, + "loss": 0.2752, + "step": 43619 + }, + { + "epoch": 1.68, + "learning_rate": 2.2063349576814567e-05, + "loss": 0.2925, + "step": 43620 + }, + { + "epoch": 1.68, + "learning_rate": 2.206270838676584e-05, + "loss": 0.307, + "step": 43621 + }, + { + "epoch": 1.68, + "learning_rate": 2.2062067196717107e-05, + "loss": 0.3753, + "step": 43622 + }, + { + "epoch": 1.68, + "learning_rate": 2.206142600666838e-05, + "loss": 0.3602, + "step": 43623 + }, + { + "epoch": 1.68, + "learning_rate": 2.2060784816619647e-05, + "loss": 0.2616, + "step": 43624 + }, + { + "epoch": 1.68, + "learning_rate": 2.206014362657092e-05, + "loss": 0.3412, + "step": 43625 + }, + { + "epoch": 1.68, + "learning_rate": 2.2059502436522187e-05, + "loss": 0.4492, + "step": 43626 + }, + { + "epoch": 1.68, + "learning_rate": 2.2058861246473455e-05, + "loss": 0.3487, + "step": 43627 + }, + { + "epoch": 1.68, + "learning_rate": 2.2058220056424723e-05, + "loss": 0.3036, + "step": 43628 + }, + { + "epoch": 1.68, + "learning_rate": 2.2057578866375995e-05, + "loss": 0.4495, + "step": 43629 + }, + { + "epoch": 1.68, + "learning_rate": 2.2056937676327263e-05, + "loss": 0.5461, + "step": 43630 + }, + { + "epoch": 1.68, + "learning_rate": 2.2056296486278535e-05, + "loss": 0.3442, + "step": 43631 + }, + { + "epoch": 1.68, + "learning_rate": 2.2055655296229803e-05, + "loss": 0.3168, + "step": 43632 + }, + { + "epoch": 1.68, + "learning_rate": 2.2055014106181075e-05, + "loss": 0.3305, + "step": 43633 + }, + { + "epoch": 1.68, + "learning_rate": 2.2054372916132344e-05, + "loss": 0.4371, + "step": 43634 + }, + { + "epoch": 1.68, + "learning_rate": 2.2053731726083612e-05, + "loss": 0.3465, + "step": 43635 + }, + { + "epoch": 1.68, + "learning_rate": 2.205309053603488e-05, + "loss": 0.3212, + "step": 43636 + }, + { + "epoch": 1.68, + "learning_rate": 2.2052449345986152e-05, + "loss": 0.269, + "step": 43637 + }, + { + "epoch": 1.68, + "learning_rate": 2.205180815593742e-05, + "loss": 0.2971, + "step": 43638 + }, + { + "epoch": 1.68, + "learning_rate": 2.205116696588869e-05, + "loss": 0.271, + "step": 43639 + }, + { + "epoch": 1.68, + "learning_rate": 2.205052577583996e-05, + "loss": 0.3261, + "step": 43640 + }, + { + "epoch": 1.68, + "learning_rate": 2.2049884585791232e-05, + "loss": 0.3424, + "step": 43641 + }, + { + "epoch": 1.68, + "learning_rate": 2.20492433957425e-05, + "loss": 0.2553, + "step": 43642 + }, + { + "epoch": 1.68, + "learning_rate": 2.204860220569377e-05, + "loss": 0.5477, + "step": 43643 + }, + { + "epoch": 1.68, + "learning_rate": 2.2047961015645037e-05, + "loss": 0.3747, + "step": 43644 + }, + { + "epoch": 1.68, + "learning_rate": 2.204731982559631e-05, + "loss": 0.2718, + "step": 43645 + }, + { + "epoch": 1.68, + "learning_rate": 2.2046678635547577e-05, + "loss": 0.3231, + "step": 43646 + }, + { + "epoch": 1.68, + "learning_rate": 2.2046037445498845e-05, + "loss": 0.2479, + "step": 43647 + }, + { + "epoch": 1.68, + "learning_rate": 2.2045396255450114e-05, + "loss": 0.3085, + "step": 43648 + }, + { + "epoch": 1.68, + "learning_rate": 2.204475506540139e-05, + "loss": 0.2686, + "step": 43649 + }, + { + "epoch": 1.68, + "learning_rate": 2.2044113875352657e-05, + "loss": 0.2618, + "step": 43650 + }, + { + "epoch": 1.68, + "learning_rate": 2.2043472685303925e-05, + "loss": 0.2652, + "step": 43651 + }, + { + "epoch": 1.68, + "learning_rate": 2.2042831495255194e-05, + "loss": 0.2697, + "step": 43652 + }, + { + "epoch": 1.68, + "learning_rate": 2.2042190305206465e-05, + "loss": 0.305, + "step": 43653 + }, + { + "epoch": 1.68, + "learning_rate": 2.2041549115157734e-05, + "loss": 0.4632, + "step": 43654 + }, + { + "epoch": 1.68, + "learning_rate": 2.2040907925109002e-05, + "loss": 0.2832, + "step": 43655 + }, + { + "epoch": 1.68, + "learning_rate": 2.204026673506027e-05, + "loss": 0.2466, + "step": 43656 + }, + { + "epoch": 1.68, + "learning_rate": 2.2039625545011542e-05, + "loss": 0.3697, + "step": 43657 + }, + { + "epoch": 1.68, + "learning_rate": 2.2038984354962814e-05, + "loss": 0.2951, + "step": 43658 + }, + { + "epoch": 1.68, + "learning_rate": 2.2038343164914082e-05, + "loss": 0.2656, + "step": 43659 + }, + { + "epoch": 1.68, + "learning_rate": 2.203770197486535e-05, + "loss": 0.2929, + "step": 43660 + }, + { + "epoch": 1.68, + "learning_rate": 2.2037060784816622e-05, + "loss": 0.5161, + "step": 43661 + }, + { + "epoch": 1.68, + "learning_rate": 2.203641959476789e-05, + "loss": 0.3777, + "step": 43662 + }, + { + "epoch": 1.68, + "learning_rate": 2.203577840471916e-05, + "loss": 0.2701, + "step": 43663 + }, + { + "epoch": 1.68, + "learning_rate": 2.2035137214670427e-05, + "loss": 0.3854, + "step": 43664 + }, + { + "epoch": 1.68, + "learning_rate": 2.20344960246217e-05, + "loss": 0.2547, + "step": 43665 + }, + { + "epoch": 1.68, + "learning_rate": 2.2033854834572967e-05, + "loss": 0.2933, + "step": 43666 + }, + { + "epoch": 1.68, + "learning_rate": 2.203321364452424e-05, + "loss": 0.2922, + "step": 43667 + }, + { + "epoch": 1.68, + "learning_rate": 2.2032572454475507e-05, + "loss": 0.2829, + "step": 43668 + }, + { + "epoch": 1.68, + "learning_rate": 2.203193126442678e-05, + "loss": 0.2603, + "step": 43669 + }, + { + "epoch": 1.68, + "learning_rate": 2.2031290074378047e-05, + "loss": 0.6036, + "step": 43670 + }, + { + "epoch": 1.68, + "learning_rate": 2.2030648884329316e-05, + "loss": 0.363, + "step": 43671 + }, + { + "epoch": 1.68, + "learning_rate": 2.2030007694280584e-05, + "loss": 0.3051, + "step": 43672 + }, + { + "epoch": 1.68, + "learning_rate": 2.2029366504231856e-05, + "loss": 0.3552, + "step": 43673 + }, + { + "epoch": 1.68, + "learning_rate": 2.2028725314183124e-05, + "loss": 0.3877, + "step": 43674 + }, + { + "epoch": 1.68, + "learning_rate": 2.2028084124134392e-05, + "loss": 0.3278, + "step": 43675 + }, + { + "epoch": 1.68, + "learning_rate": 2.2027442934085664e-05, + "loss": 0.2744, + "step": 43676 + }, + { + "epoch": 1.68, + "learning_rate": 2.2026801744036936e-05, + "loss": 0.4091, + "step": 43677 + }, + { + "epoch": 1.68, + "learning_rate": 2.2026160553988204e-05, + "loss": 0.1979, + "step": 43678 + }, + { + "epoch": 1.68, + "learning_rate": 2.2025519363939473e-05, + "loss": 0.2723, + "step": 43679 + }, + { + "epoch": 1.68, + "learning_rate": 2.202487817389074e-05, + "loss": 0.2641, + "step": 43680 + }, + { + "epoch": 1.68, + "learning_rate": 2.2024236983842013e-05, + "loss": 0.2238, + "step": 43681 + }, + { + "epoch": 1.68, + "learning_rate": 2.202359579379328e-05, + "loss": 0.2289, + "step": 43682 + }, + { + "epoch": 1.68, + "learning_rate": 2.202295460374455e-05, + "loss": 0.3722, + "step": 43683 + }, + { + "epoch": 1.68, + "learning_rate": 2.2022313413695818e-05, + "loss": 0.362, + "step": 43684 + }, + { + "epoch": 1.68, + "learning_rate": 2.2021672223647093e-05, + "loss": 0.2745, + "step": 43685 + }, + { + "epoch": 1.68, + "learning_rate": 2.202103103359836e-05, + "loss": 0.2587, + "step": 43686 + }, + { + "epoch": 1.68, + "learning_rate": 2.202038984354963e-05, + "loss": 0.3558, + "step": 43687 + }, + { + "epoch": 1.68, + "learning_rate": 2.2019748653500898e-05, + "loss": 0.3132, + "step": 43688 + }, + { + "epoch": 1.68, + "learning_rate": 2.201910746345217e-05, + "loss": 0.2997, + "step": 43689 + }, + { + "epoch": 1.68, + "learning_rate": 2.2018466273403438e-05, + "loss": 0.2863, + "step": 43690 + }, + { + "epoch": 1.68, + "learning_rate": 2.2017825083354706e-05, + "loss": 0.4446, + "step": 43691 + }, + { + "epoch": 1.68, + "learning_rate": 2.2017183893305974e-05, + "loss": 0.2624, + "step": 43692 + }, + { + "epoch": 1.68, + "learning_rate": 2.2016542703257246e-05, + "loss": 0.3861, + "step": 43693 + }, + { + "epoch": 1.68, + "learning_rate": 2.2015901513208518e-05, + "loss": 0.43, + "step": 43694 + }, + { + "epoch": 1.68, + "learning_rate": 2.2015260323159786e-05, + "loss": 0.338, + "step": 43695 + }, + { + "epoch": 1.68, + "learning_rate": 2.2014619133111054e-05, + "loss": 0.3397, + "step": 43696 + }, + { + "epoch": 1.68, + "learning_rate": 2.2013977943062326e-05, + "loss": 0.1871, + "step": 43697 + }, + { + "epoch": 1.68, + "learning_rate": 2.2013336753013594e-05, + "loss": 0.2423, + "step": 43698 + }, + { + "epoch": 1.68, + "learning_rate": 2.2012695562964863e-05, + "loss": 0.3002, + "step": 43699 + }, + { + "epoch": 1.68, + "learning_rate": 2.201205437291613e-05, + "loss": 0.2713, + "step": 43700 + }, + { + "epoch": 1.68, + "learning_rate": 2.2011413182867403e-05, + "loss": 0.2735, + "step": 43701 + }, + { + "epoch": 1.68, + "learning_rate": 2.201077199281867e-05, + "loss": 0.4209, + "step": 43702 + }, + { + "epoch": 1.68, + "learning_rate": 2.2010130802769943e-05, + "loss": 0.2381, + "step": 43703 + }, + { + "epoch": 1.68, + "learning_rate": 2.200948961272121e-05, + "loss": 0.3164, + "step": 43704 + }, + { + "epoch": 1.68, + "learning_rate": 2.2008848422672483e-05, + "loss": 0.307, + "step": 43705 + }, + { + "epoch": 1.68, + "learning_rate": 2.200820723262375e-05, + "loss": 0.3386, + "step": 43706 + }, + { + "epoch": 1.68, + "learning_rate": 2.200756604257502e-05, + "loss": 0.3811, + "step": 43707 + }, + { + "epoch": 1.68, + "learning_rate": 2.2006924852526288e-05, + "loss": 0.3419, + "step": 43708 + }, + { + "epoch": 1.68, + "learning_rate": 2.200628366247756e-05, + "loss": 0.2302, + "step": 43709 + }, + { + "epoch": 1.68, + "learning_rate": 2.2005642472428828e-05, + "loss": 0.3459, + "step": 43710 + }, + { + "epoch": 1.68, + "learning_rate": 2.2005001282380096e-05, + "loss": 0.3822, + "step": 43711 + }, + { + "epoch": 1.68, + "learning_rate": 2.2004360092331368e-05, + "loss": 0.3402, + "step": 43712 + }, + { + "epoch": 1.68, + "learning_rate": 2.200371890228264e-05, + "loss": 0.3296, + "step": 43713 + }, + { + "epoch": 1.68, + "learning_rate": 2.2003077712233908e-05, + "loss": 0.4271, + "step": 43714 + }, + { + "epoch": 1.68, + "learning_rate": 2.2002436522185176e-05, + "loss": 0.4942, + "step": 43715 + }, + { + "epoch": 1.68, + "learning_rate": 2.2001795332136445e-05, + "loss": 0.2871, + "step": 43716 + }, + { + "epoch": 1.68, + "learning_rate": 2.2001154142087716e-05, + "loss": 0.2755, + "step": 43717 + }, + { + "epoch": 1.68, + "learning_rate": 2.2000512952038985e-05, + "loss": 0.3725, + "step": 43718 + }, + { + "epoch": 1.68, + "learning_rate": 2.1999871761990253e-05, + "loss": 0.4399, + "step": 43719 + }, + { + "epoch": 1.68, + "learning_rate": 2.1999230571941525e-05, + "loss": 0.2605, + "step": 43720 + }, + { + "epoch": 1.68, + "learning_rate": 2.1998589381892797e-05, + "loss": 0.2418, + "step": 43721 + }, + { + "epoch": 1.68, + "learning_rate": 2.1997948191844065e-05, + "loss": 0.4667, + "step": 43722 + }, + { + "epoch": 1.68, + "learning_rate": 2.1997307001795333e-05, + "loss": 0.4789, + "step": 43723 + }, + { + "epoch": 1.68, + "learning_rate": 2.19966658117466e-05, + "loss": 0.3134, + "step": 43724 + }, + { + "epoch": 1.68, + "learning_rate": 2.1996024621697873e-05, + "loss": 0.3134, + "step": 43725 + }, + { + "epoch": 1.68, + "learning_rate": 2.199538343164914e-05, + "loss": 0.3022, + "step": 43726 + }, + { + "epoch": 1.68, + "learning_rate": 2.199474224160041e-05, + "loss": 0.3058, + "step": 43727 + }, + { + "epoch": 1.68, + "learning_rate": 2.1994101051551678e-05, + "loss": 0.2969, + "step": 43728 + }, + { + "epoch": 1.68, + "learning_rate": 2.1993459861502953e-05, + "loss": 0.3438, + "step": 43729 + }, + { + "epoch": 1.68, + "learning_rate": 2.199281867145422e-05, + "loss": 0.2965, + "step": 43730 + }, + { + "epoch": 1.68, + "learning_rate": 2.199217748140549e-05, + "loss": 0.2356, + "step": 43731 + }, + { + "epoch": 1.68, + "learning_rate": 2.1991536291356758e-05, + "loss": 0.3238, + "step": 43732 + }, + { + "epoch": 1.68, + "learning_rate": 2.199089510130803e-05, + "loss": 0.2338, + "step": 43733 + }, + { + "epoch": 1.68, + "learning_rate": 2.19902539112593e-05, + "loss": 0.3481, + "step": 43734 + }, + { + "epoch": 1.68, + "learning_rate": 2.1989612721210567e-05, + "loss": 0.3958, + "step": 43735 + }, + { + "epoch": 1.68, + "learning_rate": 2.1988971531161835e-05, + "loss": 0.4726, + "step": 43736 + }, + { + "epoch": 1.68, + "learning_rate": 2.1988330341113107e-05, + "loss": 0.3923, + "step": 43737 + }, + { + "epoch": 1.68, + "learning_rate": 2.198768915106438e-05, + "loss": 0.4259, + "step": 43738 + }, + { + "epoch": 1.68, + "learning_rate": 2.1987047961015647e-05, + "loss": 0.4853, + "step": 43739 + }, + { + "epoch": 1.68, + "learning_rate": 2.1986406770966915e-05, + "loss": 0.4895, + "step": 43740 + }, + { + "epoch": 1.68, + "learning_rate": 2.1985765580918187e-05, + "loss": 0.2508, + "step": 43741 + }, + { + "epoch": 1.68, + "learning_rate": 2.1985124390869455e-05, + "loss": 0.3947, + "step": 43742 + }, + { + "epoch": 1.68, + "learning_rate": 2.1984483200820723e-05, + "loss": 0.3243, + "step": 43743 + }, + { + "epoch": 1.68, + "learning_rate": 2.1983842010771992e-05, + "loss": 0.4048, + "step": 43744 + }, + { + "epoch": 1.68, + "learning_rate": 2.1983200820723264e-05, + "loss": 0.4592, + "step": 43745 + }, + { + "epoch": 1.68, + "learning_rate": 2.1982559630674532e-05, + "loss": 0.2765, + "step": 43746 + }, + { + "epoch": 1.68, + "learning_rate": 2.1981918440625804e-05, + "loss": 0.3145, + "step": 43747 + }, + { + "epoch": 1.68, + "learning_rate": 2.1981277250577072e-05, + "loss": 0.3841, + "step": 43748 + }, + { + "epoch": 1.68, + "learning_rate": 2.1980636060528344e-05, + "loss": 0.3055, + "step": 43749 + }, + { + "epoch": 1.68, + "learning_rate": 2.1979994870479612e-05, + "loss": 0.271, + "step": 43750 + }, + { + "epoch": 1.68, + "learning_rate": 2.197935368043088e-05, + "loss": 0.5437, + "step": 43751 + }, + { + "epoch": 1.68, + "learning_rate": 2.197871249038215e-05, + "loss": 0.4172, + "step": 43752 + }, + { + "epoch": 1.68, + "learning_rate": 2.197807130033342e-05, + "loss": 0.4557, + "step": 43753 + }, + { + "epoch": 1.68, + "learning_rate": 2.197743011028469e-05, + "loss": 0.3913, + "step": 43754 + }, + { + "epoch": 1.68, + "learning_rate": 2.1976788920235957e-05, + "loss": 0.2718, + "step": 43755 + }, + { + "epoch": 1.68, + "learning_rate": 2.197614773018723e-05, + "loss": 0.4203, + "step": 43756 + }, + { + "epoch": 1.68, + "learning_rate": 2.19755065401385e-05, + "loss": 0.3332, + "step": 43757 + }, + { + "epoch": 1.68, + "learning_rate": 2.197486535008977e-05, + "loss": 0.2174, + "step": 43758 + }, + { + "epoch": 1.68, + "learning_rate": 2.1974224160041037e-05, + "loss": 0.3358, + "step": 43759 + }, + { + "epoch": 1.68, + "learning_rate": 2.1973582969992305e-05, + "loss": 0.4775, + "step": 43760 + }, + { + "epoch": 1.68, + "learning_rate": 2.1972941779943577e-05, + "loss": 0.3234, + "step": 43761 + }, + { + "epoch": 1.68, + "learning_rate": 2.1972300589894845e-05, + "loss": 0.36, + "step": 43762 + }, + { + "epoch": 1.68, + "learning_rate": 2.1971659399846114e-05, + "loss": 0.3499, + "step": 43763 + }, + { + "epoch": 1.68, + "learning_rate": 2.1971018209797385e-05, + "loss": 0.2631, + "step": 43764 + }, + { + "epoch": 1.68, + "learning_rate": 2.1970377019748657e-05, + "loss": 0.3639, + "step": 43765 + }, + { + "epoch": 1.68, + "learning_rate": 2.1969735829699926e-05, + "loss": 0.2228, + "step": 43766 + }, + { + "epoch": 1.68, + "learning_rate": 2.1969094639651194e-05, + "loss": 0.4627, + "step": 43767 + }, + { + "epoch": 1.68, + "learning_rate": 2.1968453449602462e-05, + "loss": 0.2331, + "step": 43768 + }, + { + "epoch": 1.68, + "learning_rate": 2.1967812259553734e-05, + "loss": 0.2925, + "step": 43769 + }, + { + "epoch": 1.68, + "learning_rate": 2.1967171069505002e-05, + "loss": 0.378, + "step": 43770 + }, + { + "epoch": 1.68, + "learning_rate": 2.196652987945627e-05, + "loss": 0.3054, + "step": 43771 + }, + { + "epoch": 1.68, + "learning_rate": 2.196588868940754e-05, + "loss": 0.3407, + "step": 43772 + }, + { + "epoch": 1.68, + "learning_rate": 2.196524749935881e-05, + "loss": 0.3806, + "step": 43773 + }, + { + "epoch": 1.68, + "learning_rate": 2.1964606309310082e-05, + "loss": 0.3173, + "step": 43774 + }, + { + "epoch": 1.68, + "learning_rate": 2.196396511926135e-05, + "loss": 0.3928, + "step": 43775 + }, + { + "epoch": 1.68, + "learning_rate": 2.196332392921262e-05, + "loss": 0.2954, + "step": 43776 + }, + { + "epoch": 1.68, + "learning_rate": 2.196268273916389e-05, + "loss": 0.332, + "step": 43777 + }, + { + "epoch": 1.68, + "learning_rate": 2.196204154911516e-05, + "loss": 0.3302, + "step": 43778 + }, + { + "epoch": 1.68, + "learning_rate": 2.1961400359066427e-05, + "loss": 0.2621, + "step": 43779 + }, + { + "epoch": 1.68, + "learning_rate": 2.1960759169017696e-05, + "loss": 0.3004, + "step": 43780 + }, + { + "epoch": 1.68, + "learning_rate": 2.1960117978968967e-05, + "loss": 0.2649, + "step": 43781 + }, + { + "epoch": 1.68, + "learning_rate": 2.1959476788920236e-05, + "loss": 0.4091, + "step": 43782 + }, + { + "epoch": 1.68, + "learning_rate": 2.1958835598871507e-05, + "loss": 0.2833, + "step": 43783 + }, + { + "epoch": 1.68, + "learning_rate": 2.1958194408822776e-05, + "loss": 0.3568, + "step": 43784 + }, + { + "epoch": 1.68, + "learning_rate": 2.1957553218774047e-05, + "loss": 0.4121, + "step": 43785 + }, + { + "epoch": 1.68, + "learning_rate": 2.1956912028725316e-05, + "loss": 0.2636, + "step": 43786 + }, + { + "epoch": 1.68, + "learning_rate": 2.1956270838676584e-05, + "loss": 0.2836, + "step": 43787 + }, + { + "epoch": 1.68, + "learning_rate": 2.1955629648627852e-05, + "loss": 0.3522, + "step": 43788 + }, + { + "epoch": 1.68, + "learning_rate": 2.1954988458579124e-05, + "loss": 0.4301, + "step": 43789 + }, + { + "epoch": 1.68, + "learning_rate": 2.1954347268530393e-05, + "loss": 0.3249, + "step": 43790 + }, + { + "epoch": 1.68, + "learning_rate": 2.195370607848166e-05, + "loss": 0.4441, + "step": 43791 + }, + { + "epoch": 1.68, + "learning_rate": 2.1953064888432933e-05, + "loss": 0.2916, + "step": 43792 + }, + { + "epoch": 1.68, + "learning_rate": 2.1952423698384204e-05, + "loss": 0.4612, + "step": 43793 + }, + { + "epoch": 1.68, + "learning_rate": 2.1951782508335473e-05, + "loss": 0.3182, + "step": 43794 + }, + { + "epoch": 1.68, + "learning_rate": 2.195114131828674e-05, + "loss": 0.3355, + "step": 43795 + }, + { + "epoch": 1.68, + "learning_rate": 2.195050012823801e-05, + "loss": 0.382, + "step": 43796 + }, + { + "epoch": 1.68, + "learning_rate": 2.194985893818928e-05, + "loss": 0.4692, + "step": 43797 + }, + { + "epoch": 1.68, + "learning_rate": 2.194921774814055e-05, + "loss": 0.3293, + "step": 43798 + }, + { + "epoch": 1.68, + "learning_rate": 2.1948576558091818e-05, + "loss": 0.3857, + "step": 43799 + }, + { + "epoch": 1.68, + "learning_rate": 2.194793536804309e-05, + "loss": 0.3452, + "step": 43800 + }, + { + "epoch": 1.68, + "learning_rate": 2.194729417799436e-05, + "loss": 0.4942, + "step": 43801 + }, + { + "epoch": 1.68, + "learning_rate": 2.194665298794563e-05, + "loss": 0.3952, + "step": 43802 + }, + { + "epoch": 1.68, + "learning_rate": 2.1946011797896898e-05, + "loss": 0.5047, + "step": 43803 + }, + { + "epoch": 1.68, + "learning_rate": 2.1945370607848166e-05, + "loss": 0.4182, + "step": 43804 + }, + { + "epoch": 1.68, + "learning_rate": 2.1944729417799438e-05, + "loss": 0.2831, + "step": 43805 + }, + { + "epoch": 1.68, + "learning_rate": 2.1944088227750706e-05, + "loss": 0.2817, + "step": 43806 + }, + { + "epoch": 1.68, + "learning_rate": 2.1943447037701974e-05, + "loss": 0.3637, + "step": 43807 + }, + { + "epoch": 1.68, + "learning_rate": 2.1942805847653243e-05, + "loss": 0.3552, + "step": 43808 + }, + { + "epoch": 1.68, + "learning_rate": 2.1942164657604514e-05, + "loss": 0.3059, + "step": 43809 + }, + { + "epoch": 1.68, + "learning_rate": 2.1941523467555786e-05, + "loss": 0.2153, + "step": 43810 + }, + { + "epoch": 1.68, + "learning_rate": 2.1940882277507055e-05, + "loss": 0.3217, + "step": 43811 + }, + { + "epoch": 1.68, + "learning_rate": 2.1940241087458323e-05, + "loss": 0.3215, + "step": 43812 + }, + { + "epoch": 1.68, + "learning_rate": 2.1939599897409595e-05, + "loss": 0.2676, + "step": 43813 + }, + { + "epoch": 1.68, + "learning_rate": 2.1938958707360863e-05, + "loss": 0.3874, + "step": 43814 + }, + { + "epoch": 1.68, + "learning_rate": 2.193831751731213e-05, + "loss": 0.3057, + "step": 43815 + }, + { + "epoch": 1.68, + "learning_rate": 2.19376763272634e-05, + "loss": 0.3564, + "step": 43816 + }, + { + "epoch": 1.68, + "learning_rate": 2.193703513721467e-05, + "loss": 0.2802, + "step": 43817 + }, + { + "epoch": 1.68, + "learning_rate": 2.193639394716594e-05, + "loss": 0.3698, + "step": 43818 + }, + { + "epoch": 1.68, + "learning_rate": 2.193575275711721e-05, + "loss": 0.4044, + "step": 43819 + }, + { + "epoch": 1.68, + "learning_rate": 2.193511156706848e-05, + "loss": 0.2682, + "step": 43820 + }, + { + "epoch": 1.68, + "learning_rate": 2.193447037701975e-05, + "loss": 0.333, + "step": 43821 + }, + { + "epoch": 1.68, + "learning_rate": 2.193382918697102e-05, + "loss": 0.336, + "step": 43822 + }, + { + "epoch": 1.68, + "learning_rate": 2.1933187996922288e-05, + "loss": 0.3112, + "step": 43823 + }, + { + "epoch": 1.68, + "learning_rate": 2.1932546806873556e-05, + "loss": 0.3216, + "step": 43824 + }, + { + "epoch": 1.68, + "learning_rate": 2.1931905616824828e-05, + "loss": 0.4522, + "step": 43825 + }, + { + "epoch": 1.68, + "learning_rate": 2.1931264426776096e-05, + "loss": 0.2862, + "step": 43826 + }, + { + "epoch": 1.69, + "learning_rate": 2.1930623236727368e-05, + "loss": 0.2411, + "step": 43827 + }, + { + "epoch": 1.69, + "learning_rate": 2.1929982046678636e-05, + "loss": 0.4033, + "step": 43828 + }, + { + "epoch": 1.69, + "learning_rate": 2.1929340856629908e-05, + "loss": 0.3989, + "step": 43829 + }, + { + "epoch": 1.69, + "learning_rate": 2.1928699666581176e-05, + "loss": 0.3594, + "step": 43830 + }, + { + "epoch": 1.69, + "learning_rate": 2.1928058476532445e-05, + "loss": 0.2119, + "step": 43831 + }, + { + "epoch": 1.69, + "learning_rate": 2.1927417286483713e-05, + "loss": 0.3241, + "step": 43832 + }, + { + "epoch": 1.69, + "learning_rate": 2.1926776096434985e-05, + "loss": 0.3053, + "step": 43833 + }, + { + "epoch": 1.69, + "learning_rate": 2.1926134906386253e-05, + "loss": 0.3978, + "step": 43834 + }, + { + "epoch": 1.69, + "learning_rate": 2.192549371633752e-05, + "loss": 0.4563, + "step": 43835 + }, + { + "epoch": 1.69, + "learning_rate": 2.1924852526288793e-05, + "loss": 0.3666, + "step": 43836 + }, + { + "epoch": 1.69, + "learning_rate": 2.1924211336240065e-05, + "loss": 0.251, + "step": 43837 + }, + { + "epoch": 1.69, + "learning_rate": 2.1923570146191333e-05, + "loss": 0.3953, + "step": 43838 + }, + { + "epoch": 1.69, + "learning_rate": 2.19229289561426e-05, + "loss": 0.2222, + "step": 43839 + }, + { + "epoch": 1.69, + "learning_rate": 2.192228776609387e-05, + "loss": 0.4003, + "step": 43840 + }, + { + "epoch": 1.69, + "learning_rate": 2.192164657604514e-05, + "loss": 0.3798, + "step": 43841 + }, + { + "epoch": 1.69, + "learning_rate": 2.192100538599641e-05, + "loss": 0.3877, + "step": 43842 + }, + { + "epoch": 1.69, + "learning_rate": 2.1920364195947678e-05, + "loss": 0.2672, + "step": 43843 + }, + { + "epoch": 1.69, + "learning_rate": 2.191972300589895e-05, + "loss": 0.3821, + "step": 43844 + }, + { + "epoch": 1.69, + "learning_rate": 2.1919081815850222e-05, + "loss": 0.3425, + "step": 43845 + }, + { + "epoch": 1.69, + "learning_rate": 2.191844062580149e-05, + "loss": 0.2445, + "step": 43846 + }, + { + "epoch": 1.69, + "learning_rate": 2.191779943575276e-05, + "loss": 0.3931, + "step": 43847 + }, + { + "epoch": 1.69, + "learning_rate": 2.1917158245704027e-05, + "loss": 0.2205, + "step": 43848 + }, + { + "epoch": 1.69, + "learning_rate": 2.19165170556553e-05, + "loss": 0.2449, + "step": 43849 + }, + { + "epoch": 1.69, + "learning_rate": 2.1915875865606567e-05, + "loss": 0.472, + "step": 43850 + }, + { + "epoch": 1.69, + "learning_rate": 2.1915234675557835e-05, + "loss": 0.458, + "step": 43851 + }, + { + "epoch": 1.69, + "learning_rate": 2.1914593485509103e-05, + "loss": 0.3741, + "step": 43852 + }, + { + "epoch": 1.69, + "learning_rate": 2.1913952295460375e-05, + "loss": 0.3757, + "step": 43853 + }, + { + "epoch": 1.69, + "learning_rate": 2.1913311105411647e-05, + "loss": 0.3901, + "step": 43854 + }, + { + "epoch": 1.69, + "learning_rate": 2.1912669915362915e-05, + "loss": 0.3452, + "step": 43855 + }, + { + "epoch": 1.69, + "learning_rate": 2.1912028725314184e-05, + "loss": 0.3795, + "step": 43856 + }, + { + "epoch": 1.69, + "learning_rate": 2.1911387535265455e-05, + "loss": 0.4955, + "step": 43857 + }, + { + "epoch": 1.69, + "learning_rate": 2.1910746345216724e-05, + "loss": 0.3123, + "step": 43858 + }, + { + "epoch": 1.69, + "learning_rate": 2.1910105155167992e-05, + "loss": 0.2648, + "step": 43859 + }, + { + "epoch": 1.69, + "learning_rate": 2.190946396511926e-05, + "loss": 0.3283, + "step": 43860 + }, + { + "epoch": 1.69, + "learning_rate": 2.1908822775070532e-05, + "loss": 0.2647, + "step": 43861 + }, + { + "epoch": 1.69, + "learning_rate": 2.19081815850218e-05, + "loss": 0.282, + "step": 43862 + }, + { + "epoch": 1.69, + "learning_rate": 2.1907540394973072e-05, + "loss": 0.3535, + "step": 43863 + }, + { + "epoch": 1.69, + "learning_rate": 2.190689920492434e-05, + "loss": 0.3676, + "step": 43864 + }, + { + "epoch": 1.69, + "learning_rate": 2.1906258014875612e-05, + "loss": 0.3416, + "step": 43865 + }, + { + "epoch": 1.69, + "learning_rate": 2.190561682482688e-05, + "loss": 0.3953, + "step": 43866 + }, + { + "epoch": 1.69, + "learning_rate": 2.190497563477815e-05, + "loss": 0.4351, + "step": 43867 + }, + { + "epoch": 1.69, + "learning_rate": 2.1904334444729417e-05, + "loss": 0.4474, + "step": 43868 + }, + { + "epoch": 1.69, + "learning_rate": 2.190369325468069e-05, + "loss": 0.3329, + "step": 43869 + }, + { + "epoch": 1.69, + "learning_rate": 2.1903052064631957e-05, + "loss": 0.2672, + "step": 43870 + }, + { + "epoch": 1.69, + "learning_rate": 2.1902410874583225e-05, + "loss": 0.3002, + "step": 43871 + }, + { + "epoch": 1.69, + "learning_rate": 2.1901769684534497e-05, + "loss": 0.4488, + "step": 43872 + }, + { + "epoch": 1.69, + "learning_rate": 2.190112849448577e-05, + "loss": 0.4092, + "step": 43873 + }, + { + "epoch": 1.69, + "learning_rate": 2.1900487304437037e-05, + "loss": 0.4101, + "step": 43874 + }, + { + "epoch": 1.69, + "learning_rate": 2.1899846114388305e-05, + "loss": 0.526, + "step": 43875 + }, + { + "epoch": 1.69, + "learning_rate": 2.1899204924339574e-05, + "loss": 0.3856, + "step": 43876 + }, + { + "epoch": 1.69, + "learning_rate": 2.1898563734290846e-05, + "loss": 0.4969, + "step": 43877 + }, + { + "epoch": 1.69, + "learning_rate": 2.1897922544242114e-05, + "loss": 0.3908, + "step": 43878 + }, + { + "epoch": 1.69, + "learning_rate": 2.1897281354193382e-05, + "loss": 0.2638, + "step": 43879 + }, + { + "epoch": 1.69, + "learning_rate": 2.1896640164144654e-05, + "loss": 0.3561, + "step": 43880 + }, + { + "epoch": 1.69, + "learning_rate": 2.1895998974095926e-05, + "loss": 0.4378, + "step": 43881 + }, + { + "epoch": 1.69, + "learning_rate": 2.1895357784047194e-05, + "loss": 0.3168, + "step": 43882 + }, + { + "epoch": 1.69, + "learning_rate": 2.1894716593998462e-05, + "loss": 0.2776, + "step": 43883 + }, + { + "epoch": 1.69, + "learning_rate": 2.189407540394973e-05, + "loss": 0.4217, + "step": 43884 + }, + { + "epoch": 1.69, + "learning_rate": 2.1893434213901002e-05, + "loss": 0.3608, + "step": 43885 + }, + { + "epoch": 1.69, + "learning_rate": 2.189279302385227e-05, + "loss": 0.3448, + "step": 43886 + }, + { + "epoch": 1.69, + "learning_rate": 2.189215183380354e-05, + "loss": 0.2331, + "step": 43887 + }, + { + "epoch": 1.69, + "learning_rate": 2.189151064375481e-05, + "loss": 0.4264, + "step": 43888 + }, + { + "epoch": 1.69, + "learning_rate": 2.189086945370608e-05, + "loss": 0.4055, + "step": 43889 + }, + { + "epoch": 1.69, + "learning_rate": 2.189022826365735e-05, + "loss": 0.3927, + "step": 43890 + }, + { + "epoch": 1.69, + "learning_rate": 2.188958707360862e-05, + "loss": 0.3607, + "step": 43891 + }, + { + "epoch": 1.69, + "learning_rate": 2.1888945883559887e-05, + "loss": 0.3292, + "step": 43892 + }, + { + "epoch": 1.69, + "learning_rate": 2.188830469351116e-05, + "loss": 0.3526, + "step": 43893 + }, + { + "epoch": 1.69, + "learning_rate": 2.1887663503462427e-05, + "loss": 0.3318, + "step": 43894 + }, + { + "epoch": 1.69, + "learning_rate": 2.1887022313413696e-05, + "loss": 0.3267, + "step": 43895 + }, + { + "epoch": 1.69, + "learning_rate": 2.1886381123364964e-05, + "loss": 0.2596, + "step": 43896 + }, + { + "epoch": 1.69, + "learning_rate": 2.1885739933316236e-05, + "loss": 0.2661, + "step": 43897 + }, + { + "epoch": 1.69, + "learning_rate": 2.1885098743267504e-05, + "loss": 0.3958, + "step": 43898 + }, + { + "epoch": 1.69, + "learning_rate": 2.1884457553218776e-05, + "loss": 0.3457, + "step": 43899 + }, + { + "epoch": 1.69, + "learning_rate": 2.1883816363170044e-05, + "loss": 0.2884, + "step": 43900 + }, + { + "epoch": 1.69, + "learning_rate": 2.1883175173121316e-05, + "loss": 0.3225, + "step": 43901 + }, + { + "epoch": 1.69, + "learning_rate": 2.1882533983072584e-05, + "loss": 0.3018, + "step": 43902 + }, + { + "epoch": 1.69, + "learning_rate": 2.1881892793023853e-05, + "loss": 0.3128, + "step": 43903 + }, + { + "epoch": 1.69, + "learning_rate": 2.188125160297512e-05, + "loss": 0.3742, + "step": 43904 + }, + { + "epoch": 1.69, + "learning_rate": 2.1880610412926393e-05, + "loss": 0.3432, + "step": 43905 + }, + { + "epoch": 1.69, + "learning_rate": 2.187996922287766e-05, + "loss": 0.2557, + "step": 43906 + }, + { + "epoch": 1.69, + "learning_rate": 2.187932803282893e-05, + "loss": 0.228, + "step": 43907 + }, + { + "epoch": 1.69, + "learning_rate": 2.18786868427802e-05, + "loss": 0.326, + "step": 43908 + }, + { + "epoch": 1.69, + "learning_rate": 2.1878045652731473e-05, + "loss": 0.3312, + "step": 43909 + }, + { + "epoch": 1.69, + "learning_rate": 2.187740446268274e-05, + "loss": 0.2838, + "step": 43910 + }, + { + "epoch": 1.69, + "learning_rate": 2.187676327263401e-05, + "loss": 0.3894, + "step": 43911 + }, + { + "epoch": 1.69, + "learning_rate": 2.1876122082585278e-05, + "loss": 0.2865, + "step": 43912 + }, + { + "epoch": 1.69, + "learning_rate": 2.187548089253655e-05, + "loss": 0.3179, + "step": 43913 + }, + { + "epoch": 1.69, + "learning_rate": 2.1874839702487818e-05, + "loss": 0.3975, + "step": 43914 + }, + { + "epoch": 1.69, + "learning_rate": 2.1874198512439086e-05, + "loss": 0.2438, + "step": 43915 + }, + { + "epoch": 1.69, + "learning_rate": 2.1873557322390358e-05, + "loss": 0.2941, + "step": 43916 + }, + { + "epoch": 1.69, + "learning_rate": 2.187291613234163e-05, + "loss": 0.3242, + "step": 43917 + }, + { + "epoch": 1.69, + "learning_rate": 2.1872274942292898e-05, + "loss": 0.3217, + "step": 43918 + }, + { + "epoch": 1.69, + "learning_rate": 2.1871633752244166e-05, + "loss": 0.4313, + "step": 43919 + }, + { + "epoch": 1.69, + "learning_rate": 2.1870992562195434e-05, + "loss": 0.3222, + "step": 43920 + }, + { + "epoch": 1.69, + "learning_rate": 2.1870351372146706e-05, + "loss": 0.3163, + "step": 43921 + }, + { + "epoch": 1.69, + "learning_rate": 2.1869710182097974e-05, + "loss": 0.4144, + "step": 43922 + }, + { + "epoch": 1.69, + "learning_rate": 2.1869068992049243e-05, + "loss": 0.3771, + "step": 43923 + }, + { + "epoch": 1.69, + "learning_rate": 2.1868427802000515e-05, + "loss": 0.3841, + "step": 43924 + }, + { + "epoch": 1.69, + "learning_rate": 2.1867786611951786e-05, + "loss": 0.3846, + "step": 43925 + }, + { + "epoch": 1.69, + "learning_rate": 2.1867145421903055e-05, + "loss": 0.2678, + "step": 43926 + }, + { + "epoch": 1.69, + "learning_rate": 2.1866504231854323e-05, + "loss": 0.3505, + "step": 43927 + }, + { + "epoch": 1.69, + "learning_rate": 2.186586304180559e-05, + "loss": 0.3922, + "step": 43928 + }, + { + "epoch": 1.69, + "learning_rate": 2.1865221851756863e-05, + "loss": 0.287, + "step": 43929 + }, + { + "epoch": 1.69, + "learning_rate": 2.186458066170813e-05, + "loss": 0.2993, + "step": 43930 + }, + { + "epoch": 1.69, + "learning_rate": 2.18639394716594e-05, + "loss": 0.3821, + "step": 43931 + }, + { + "epoch": 1.69, + "learning_rate": 2.1863298281610668e-05, + "loss": 0.3445, + "step": 43932 + }, + { + "epoch": 1.69, + "learning_rate": 2.186265709156194e-05, + "loss": 0.1939, + "step": 43933 + }, + { + "epoch": 1.69, + "learning_rate": 2.186201590151321e-05, + "loss": 0.3488, + "step": 43934 + }, + { + "epoch": 1.69, + "learning_rate": 2.186137471146448e-05, + "loss": 0.6307, + "step": 43935 + }, + { + "epoch": 1.69, + "learning_rate": 2.1860733521415748e-05, + "loss": 0.4576, + "step": 43936 + }, + { + "epoch": 1.69, + "learning_rate": 2.186009233136702e-05, + "loss": 0.355, + "step": 43937 + }, + { + "epoch": 1.69, + "learning_rate": 2.1859451141318288e-05, + "loss": 0.261, + "step": 43938 + }, + { + "epoch": 1.69, + "learning_rate": 2.1858809951269556e-05, + "loss": 0.238, + "step": 43939 + }, + { + "epoch": 1.69, + "learning_rate": 2.1858168761220825e-05, + "loss": 0.4532, + "step": 43940 + }, + { + "epoch": 1.69, + "learning_rate": 2.1857527571172096e-05, + "loss": 0.4177, + "step": 43941 + }, + { + "epoch": 1.69, + "learning_rate": 2.1856886381123365e-05, + "loss": 0.289, + "step": 43942 + }, + { + "epoch": 1.69, + "learning_rate": 2.1856245191074637e-05, + "loss": 0.3928, + "step": 43943 + }, + { + "epoch": 1.69, + "learning_rate": 2.1855604001025905e-05, + "loss": 0.2524, + "step": 43944 + }, + { + "epoch": 1.69, + "learning_rate": 2.1854962810977177e-05, + "loss": 0.2391, + "step": 43945 + }, + { + "epoch": 1.69, + "learning_rate": 2.1854321620928445e-05, + "loss": 0.4482, + "step": 43946 + }, + { + "epoch": 1.69, + "learning_rate": 2.1853680430879713e-05, + "loss": 0.3697, + "step": 43947 + }, + { + "epoch": 1.69, + "learning_rate": 2.185303924083098e-05, + "loss": 0.4097, + "step": 43948 + }, + { + "epoch": 1.69, + "learning_rate": 2.1852398050782253e-05, + "loss": 0.373, + "step": 43949 + }, + { + "epoch": 1.69, + "learning_rate": 2.185175686073352e-05, + "loss": 0.2995, + "step": 43950 + }, + { + "epoch": 1.69, + "learning_rate": 2.185111567068479e-05, + "loss": 0.2868, + "step": 43951 + }, + { + "epoch": 1.69, + "learning_rate": 2.185047448063606e-05, + "loss": 0.1922, + "step": 43952 + }, + { + "epoch": 1.69, + "learning_rate": 2.1849833290587333e-05, + "loss": 0.2247, + "step": 43953 + }, + { + "epoch": 1.69, + "learning_rate": 2.18491921005386e-05, + "loss": 0.2433, + "step": 43954 + }, + { + "epoch": 1.69, + "learning_rate": 2.184855091048987e-05, + "loss": 0.3516, + "step": 43955 + }, + { + "epoch": 1.69, + "learning_rate": 2.184790972044114e-05, + "loss": 0.2996, + "step": 43956 + }, + { + "epoch": 1.69, + "learning_rate": 2.184726853039241e-05, + "loss": 0.3313, + "step": 43957 + }, + { + "epoch": 1.69, + "learning_rate": 2.184662734034368e-05, + "loss": 0.3549, + "step": 43958 + }, + { + "epoch": 1.69, + "learning_rate": 2.1845986150294947e-05, + "loss": 0.3189, + "step": 43959 + }, + { + "epoch": 1.69, + "learning_rate": 2.184534496024622e-05, + "loss": 0.3891, + "step": 43960 + }, + { + "epoch": 1.69, + "learning_rate": 2.184470377019749e-05, + "loss": 0.4669, + "step": 43961 + }, + { + "epoch": 1.69, + "learning_rate": 2.184406258014876e-05, + "loss": 0.2705, + "step": 43962 + }, + { + "epoch": 1.69, + "learning_rate": 2.1843421390100027e-05, + "loss": 0.2766, + "step": 43963 + }, + { + "epoch": 1.69, + "learning_rate": 2.1842780200051295e-05, + "loss": 0.4569, + "step": 43964 + }, + { + "epoch": 1.69, + "learning_rate": 2.1842139010002567e-05, + "loss": 0.2667, + "step": 43965 + }, + { + "epoch": 1.69, + "learning_rate": 2.1841497819953835e-05, + "loss": 0.2602, + "step": 43966 + }, + { + "epoch": 1.69, + "learning_rate": 2.1840856629905103e-05, + "loss": 0.4039, + "step": 43967 + }, + { + "epoch": 1.69, + "learning_rate": 2.1840215439856375e-05, + "loss": 0.4327, + "step": 43968 + }, + { + "epoch": 1.69, + "learning_rate": 2.1839574249807644e-05, + "loss": 0.4323, + "step": 43969 + }, + { + "epoch": 1.69, + "learning_rate": 2.1838933059758915e-05, + "loss": 0.4133, + "step": 43970 + }, + { + "epoch": 1.69, + "learning_rate": 2.1838291869710184e-05, + "loss": 0.3769, + "step": 43971 + }, + { + "epoch": 1.69, + "learning_rate": 2.1837650679661452e-05, + "loss": 0.3568, + "step": 43972 + }, + { + "epoch": 1.69, + "learning_rate": 2.1837009489612724e-05, + "loss": 0.4019, + "step": 43973 + }, + { + "epoch": 1.69, + "learning_rate": 2.1836368299563992e-05, + "loss": 0.2825, + "step": 43974 + }, + { + "epoch": 1.69, + "learning_rate": 2.183572710951526e-05, + "loss": 0.3733, + "step": 43975 + }, + { + "epoch": 1.69, + "learning_rate": 2.183508591946653e-05, + "loss": 0.2487, + "step": 43976 + }, + { + "epoch": 1.69, + "learning_rate": 2.18344447294178e-05, + "loss": 0.4574, + "step": 43977 + }, + { + "epoch": 1.69, + "learning_rate": 2.183380353936907e-05, + "loss": 0.4992, + "step": 43978 + }, + { + "epoch": 1.69, + "learning_rate": 2.183316234932034e-05, + "loss": 0.2818, + "step": 43979 + }, + { + "epoch": 1.69, + "learning_rate": 2.183252115927161e-05, + "loss": 0.2922, + "step": 43980 + }, + { + "epoch": 1.69, + "learning_rate": 2.183187996922288e-05, + "loss": 0.2371, + "step": 43981 + }, + { + "epoch": 1.69, + "learning_rate": 2.183123877917415e-05, + "loss": 0.2665, + "step": 43982 + }, + { + "epoch": 1.69, + "learning_rate": 2.1830597589125417e-05, + "loss": 0.3729, + "step": 43983 + }, + { + "epoch": 1.69, + "learning_rate": 2.1829956399076685e-05, + "loss": 0.4287, + "step": 43984 + }, + { + "epoch": 1.69, + "learning_rate": 2.1829315209027957e-05, + "loss": 0.356, + "step": 43985 + }, + { + "epoch": 1.69, + "learning_rate": 2.1828674018979225e-05, + "loss": 0.3725, + "step": 43986 + }, + { + "epoch": 1.69, + "learning_rate": 2.1828032828930494e-05, + "loss": 0.3019, + "step": 43987 + }, + { + "epoch": 1.69, + "learning_rate": 2.1827391638881765e-05, + "loss": 0.3405, + "step": 43988 + }, + { + "epoch": 1.69, + "learning_rate": 2.1826750448833037e-05, + "loss": 0.333, + "step": 43989 + }, + { + "epoch": 1.69, + "learning_rate": 2.1826109258784306e-05, + "loss": 0.2854, + "step": 43990 + }, + { + "epoch": 1.69, + "learning_rate": 2.1825468068735574e-05, + "loss": 0.3676, + "step": 43991 + }, + { + "epoch": 1.69, + "learning_rate": 2.1824826878686842e-05, + "loss": 0.4314, + "step": 43992 + }, + { + "epoch": 1.69, + "learning_rate": 2.1824185688638114e-05, + "loss": 0.2982, + "step": 43993 + }, + { + "epoch": 1.69, + "learning_rate": 2.1823544498589382e-05, + "loss": 0.327, + "step": 43994 + }, + { + "epoch": 1.69, + "learning_rate": 2.182290330854065e-05, + "loss": 0.3174, + "step": 43995 + }, + { + "epoch": 1.69, + "learning_rate": 2.1822262118491922e-05, + "loss": 0.2778, + "step": 43996 + }, + { + "epoch": 1.69, + "learning_rate": 2.1821620928443194e-05, + "loss": 0.3223, + "step": 43997 + }, + { + "epoch": 1.69, + "learning_rate": 2.1820979738394462e-05, + "loss": 0.3478, + "step": 43998 + }, + { + "epoch": 1.69, + "learning_rate": 2.182033854834573e-05, + "loss": 0.2515, + "step": 43999 + }, + { + "epoch": 1.69, + "learning_rate": 2.1819697358297e-05, + "loss": 0.3582, + "step": 44000 + }, + { + "epoch": 1.69, + "learning_rate": 2.181905616824827e-05, + "loss": 0.305, + "step": 44001 + }, + { + "epoch": 1.69, + "learning_rate": 2.181841497819954e-05, + "loss": 0.3279, + "step": 44002 + }, + { + "epoch": 1.69, + "learning_rate": 2.1817773788150807e-05, + "loss": 0.4536, + "step": 44003 + }, + { + "epoch": 1.69, + "learning_rate": 2.181713259810208e-05, + "loss": 0.3768, + "step": 44004 + }, + { + "epoch": 1.69, + "learning_rate": 2.1816491408053347e-05, + "loss": 0.3349, + "step": 44005 + }, + { + "epoch": 1.69, + "learning_rate": 2.181585021800462e-05, + "loss": 0.3129, + "step": 44006 + }, + { + "epoch": 1.69, + "learning_rate": 2.1815209027955887e-05, + "loss": 0.4025, + "step": 44007 + }, + { + "epoch": 1.69, + "learning_rate": 2.1814567837907156e-05, + "loss": 0.3088, + "step": 44008 + }, + { + "epoch": 1.69, + "learning_rate": 2.1813926647858427e-05, + "loss": 0.268, + "step": 44009 + }, + { + "epoch": 1.69, + "learning_rate": 2.1813285457809696e-05, + "loss": 0.3624, + "step": 44010 + }, + { + "epoch": 1.69, + "learning_rate": 2.1812644267760964e-05, + "loss": 0.3497, + "step": 44011 + }, + { + "epoch": 1.69, + "learning_rate": 2.1812003077712232e-05, + "loss": 0.4153, + "step": 44012 + }, + { + "epoch": 1.69, + "learning_rate": 2.1811361887663504e-05, + "loss": 0.3731, + "step": 44013 + }, + { + "epoch": 1.69, + "learning_rate": 2.1810720697614776e-05, + "loss": 0.3765, + "step": 44014 + }, + { + "epoch": 1.69, + "learning_rate": 2.1810079507566044e-05, + "loss": 0.2831, + "step": 44015 + }, + { + "epoch": 1.69, + "learning_rate": 2.1809438317517313e-05, + "loss": 0.329, + "step": 44016 + }, + { + "epoch": 1.69, + "learning_rate": 2.1808797127468584e-05, + "loss": 0.3334, + "step": 44017 + }, + { + "epoch": 1.69, + "learning_rate": 2.1808155937419853e-05, + "loss": 0.3415, + "step": 44018 + }, + { + "epoch": 1.69, + "learning_rate": 2.180751474737112e-05, + "loss": 0.4828, + "step": 44019 + }, + { + "epoch": 1.69, + "learning_rate": 2.180687355732239e-05, + "loss": 0.3818, + "step": 44020 + }, + { + "epoch": 1.69, + "learning_rate": 2.180623236727366e-05, + "loss": 0.4862, + "step": 44021 + }, + { + "epoch": 1.69, + "learning_rate": 2.180559117722493e-05, + "loss": 0.3299, + "step": 44022 + }, + { + "epoch": 1.69, + "learning_rate": 2.18049499871762e-05, + "loss": 0.5372, + "step": 44023 + }, + { + "epoch": 1.69, + "learning_rate": 2.180430879712747e-05, + "loss": 0.3923, + "step": 44024 + }, + { + "epoch": 1.69, + "learning_rate": 2.180366760707874e-05, + "loss": 0.2926, + "step": 44025 + }, + { + "epoch": 1.69, + "learning_rate": 2.180302641703001e-05, + "loss": 0.3328, + "step": 44026 + }, + { + "epoch": 1.69, + "learning_rate": 2.1802385226981278e-05, + "loss": 0.241, + "step": 44027 + }, + { + "epoch": 1.69, + "learning_rate": 2.1801744036932546e-05, + "loss": 0.355, + "step": 44028 + }, + { + "epoch": 1.69, + "learning_rate": 2.1801102846883818e-05, + "loss": 0.3659, + "step": 44029 + }, + { + "epoch": 1.69, + "learning_rate": 2.1800461656835086e-05, + "loss": 0.4681, + "step": 44030 + }, + { + "epoch": 1.69, + "learning_rate": 2.1799820466786354e-05, + "loss": 0.3713, + "step": 44031 + }, + { + "epoch": 1.69, + "learning_rate": 2.1799179276737626e-05, + "loss": 0.3313, + "step": 44032 + }, + { + "epoch": 1.69, + "learning_rate": 2.1798538086688898e-05, + "loss": 0.4178, + "step": 44033 + }, + { + "epoch": 1.69, + "learning_rate": 2.1797896896640166e-05, + "loss": 0.2399, + "step": 44034 + }, + { + "epoch": 1.69, + "learning_rate": 2.1797255706591435e-05, + "loss": 0.3194, + "step": 44035 + }, + { + "epoch": 1.69, + "learning_rate": 2.1796614516542703e-05, + "loss": 0.3595, + "step": 44036 + }, + { + "epoch": 1.69, + "learning_rate": 2.1795973326493975e-05, + "loss": 0.2957, + "step": 44037 + }, + { + "epoch": 1.69, + "learning_rate": 2.1795332136445243e-05, + "loss": 0.287, + "step": 44038 + }, + { + "epoch": 1.69, + "learning_rate": 2.179469094639651e-05, + "loss": 0.4056, + "step": 44039 + }, + { + "epoch": 1.69, + "learning_rate": 2.1794049756347783e-05, + "loss": 0.4092, + "step": 44040 + }, + { + "epoch": 1.69, + "learning_rate": 2.1793408566299055e-05, + "loss": 0.3421, + "step": 44041 + }, + { + "epoch": 1.69, + "learning_rate": 2.1792767376250323e-05, + "loss": 0.184, + "step": 44042 + }, + { + "epoch": 1.69, + "learning_rate": 2.179212618620159e-05, + "loss": 0.3745, + "step": 44043 + }, + { + "epoch": 1.69, + "learning_rate": 2.179148499615286e-05, + "loss": 0.48, + "step": 44044 + }, + { + "epoch": 1.69, + "learning_rate": 2.179084380610413e-05, + "loss": 0.325, + "step": 44045 + }, + { + "epoch": 1.69, + "learning_rate": 2.17902026160554e-05, + "loss": 0.3143, + "step": 44046 + }, + { + "epoch": 1.69, + "learning_rate": 2.1789561426006668e-05, + "loss": 0.2913, + "step": 44047 + }, + { + "epoch": 1.69, + "learning_rate": 2.178892023595794e-05, + "loss": 0.3207, + "step": 44048 + }, + { + "epoch": 1.69, + "learning_rate": 2.1788279045909208e-05, + "loss": 0.3237, + "step": 44049 + }, + { + "epoch": 1.69, + "learning_rate": 2.178763785586048e-05, + "loss": 0.2433, + "step": 44050 + }, + { + "epoch": 1.69, + "learning_rate": 2.1786996665811748e-05, + "loss": 0.3677, + "step": 44051 + }, + { + "epoch": 1.69, + "learning_rate": 2.1786355475763016e-05, + "loss": 0.271, + "step": 44052 + }, + { + "epoch": 1.69, + "learning_rate": 2.1785714285714288e-05, + "loss": 0.4104, + "step": 44053 + }, + { + "epoch": 1.69, + "learning_rate": 2.1785073095665556e-05, + "loss": 0.473, + "step": 44054 + }, + { + "epoch": 1.69, + "learning_rate": 2.1784431905616825e-05, + "loss": 0.4249, + "step": 44055 + }, + { + "epoch": 1.69, + "learning_rate": 2.1783790715568093e-05, + "loss": 0.2656, + "step": 44056 + }, + { + "epoch": 1.69, + "learning_rate": 2.1783149525519365e-05, + "loss": 0.4828, + "step": 44057 + }, + { + "epoch": 1.69, + "learning_rate": 2.1782508335470633e-05, + "loss": 0.3333, + "step": 44058 + }, + { + "epoch": 1.69, + "learning_rate": 2.1781867145421905e-05, + "loss": 0.3332, + "step": 44059 + }, + { + "epoch": 1.69, + "learning_rate": 2.1781225955373173e-05, + "loss": 0.3421, + "step": 44060 + }, + { + "epoch": 1.69, + "learning_rate": 2.1780584765324445e-05, + "loss": 0.4533, + "step": 44061 + }, + { + "epoch": 1.69, + "learning_rate": 2.1779943575275713e-05, + "loss": 0.2838, + "step": 44062 + }, + { + "epoch": 1.69, + "learning_rate": 2.177930238522698e-05, + "loss": 0.4382, + "step": 44063 + }, + { + "epoch": 1.69, + "learning_rate": 2.177866119517825e-05, + "loss": 0.339, + "step": 44064 + }, + { + "epoch": 1.69, + "learning_rate": 2.177802000512952e-05, + "loss": 0.289, + "step": 44065 + }, + { + "epoch": 1.69, + "learning_rate": 2.177737881508079e-05, + "loss": 0.3283, + "step": 44066 + }, + { + "epoch": 1.69, + "learning_rate": 2.177673762503206e-05, + "loss": 0.3998, + "step": 44067 + }, + { + "epoch": 1.69, + "learning_rate": 2.177609643498333e-05, + "loss": 0.3274, + "step": 44068 + }, + { + "epoch": 1.69, + "learning_rate": 2.1775455244934602e-05, + "loss": 0.2521, + "step": 44069 + }, + { + "epoch": 1.69, + "learning_rate": 2.177481405488587e-05, + "loss": 0.4118, + "step": 44070 + }, + { + "epoch": 1.69, + "learning_rate": 2.177417286483714e-05, + "loss": 0.3007, + "step": 44071 + }, + { + "epoch": 1.69, + "learning_rate": 2.1773531674788407e-05, + "loss": 0.2975, + "step": 44072 + }, + { + "epoch": 1.69, + "learning_rate": 2.177289048473968e-05, + "loss": 0.2823, + "step": 44073 + }, + { + "epoch": 1.69, + "learning_rate": 2.1772249294690947e-05, + "loss": 0.3287, + "step": 44074 + }, + { + "epoch": 1.69, + "learning_rate": 2.1771608104642215e-05, + "loss": 0.3547, + "step": 44075 + }, + { + "epoch": 1.69, + "learning_rate": 2.1770966914593487e-05, + "loss": 0.4266, + "step": 44076 + }, + { + "epoch": 1.69, + "learning_rate": 2.177032572454476e-05, + "loss": 0.345, + "step": 44077 + }, + { + "epoch": 1.69, + "learning_rate": 2.1769684534496027e-05, + "loss": 0.355, + "step": 44078 + }, + { + "epoch": 1.69, + "learning_rate": 2.1769043344447295e-05, + "loss": 0.3896, + "step": 44079 + }, + { + "epoch": 1.69, + "learning_rate": 2.1768402154398564e-05, + "loss": 0.3905, + "step": 44080 + }, + { + "epoch": 1.69, + "learning_rate": 2.1767760964349835e-05, + "loss": 0.3848, + "step": 44081 + }, + { + "epoch": 1.69, + "learning_rate": 2.1767119774301104e-05, + "loss": 0.479, + "step": 44082 + }, + { + "epoch": 1.69, + "learning_rate": 2.1766478584252372e-05, + "loss": 0.208, + "step": 44083 + }, + { + "epoch": 1.69, + "learning_rate": 2.1765837394203644e-05, + "loss": 0.3174, + "step": 44084 + }, + { + "epoch": 1.69, + "learning_rate": 2.1765196204154912e-05, + "loss": 0.3545, + "step": 44085 + }, + { + "epoch": 1.69, + "learning_rate": 2.1764555014106184e-05, + "loss": 0.305, + "step": 44086 + }, + { + "epoch": 1.7, + "learning_rate": 2.1763913824057452e-05, + "loss": 0.2274, + "step": 44087 + }, + { + "epoch": 1.7, + "learning_rate": 2.176327263400872e-05, + "loss": 0.29, + "step": 44088 + }, + { + "epoch": 1.7, + "learning_rate": 2.1762631443959992e-05, + "loss": 0.4023, + "step": 44089 + }, + { + "epoch": 1.7, + "learning_rate": 2.176199025391126e-05, + "loss": 0.3585, + "step": 44090 + }, + { + "epoch": 1.7, + "learning_rate": 2.176134906386253e-05, + "loss": 0.2044, + "step": 44091 + }, + { + "epoch": 1.7, + "learning_rate": 2.17607078738138e-05, + "loss": 0.3914, + "step": 44092 + }, + { + "epoch": 1.7, + "learning_rate": 2.176006668376507e-05, + "loss": 0.2718, + "step": 44093 + }, + { + "epoch": 1.7, + "learning_rate": 2.1759425493716337e-05, + "loss": 0.3352, + "step": 44094 + }, + { + "epoch": 1.7, + "learning_rate": 2.175878430366761e-05, + "loss": 0.3461, + "step": 44095 + }, + { + "epoch": 1.7, + "learning_rate": 2.1758143113618877e-05, + "loss": 0.3855, + "step": 44096 + }, + { + "epoch": 1.7, + "learning_rate": 2.175750192357015e-05, + "loss": 0.328, + "step": 44097 + }, + { + "epoch": 1.7, + "learning_rate": 2.1756860733521417e-05, + "loss": 0.305, + "step": 44098 + }, + { + "epoch": 1.7, + "learning_rate": 2.1756219543472685e-05, + "loss": 0.3834, + "step": 44099 + }, + { + "epoch": 1.7, + "learning_rate": 2.1755578353423954e-05, + "loss": 0.3238, + "step": 44100 + }, + { + "epoch": 1.7, + "learning_rate": 2.1754937163375226e-05, + "loss": 0.4408, + "step": 44101 + }, + { + "epoch": 1.7, + "learning_rate": 2.1754295973326494e-05, + "loss": 0.4096, + "step": 44102 + }, + { + "epoch": 1.7, + "learning_rate": 2.1753654783277762e-05, + "loss": 0.3729, + "step": 44103 + }, + { + "epoch": 1.7, + "learning_rate": 2.1753013593229034e-05, + "loss": 0.3078, + "step": 44104 + }, + { + "epoch": 1.7, + "learning_rate": 2.1752372403180306e-05, + "loss": 0.372, + "step": 44105 + }, + { + "epoch": 1.7, + "learning_rate": 2.1751731213131574e-05, + "loss": 0.3326, + "step": 44106 + }, + { + "epoch": 1.7, + "learning_rate": 2.1751090023082842e-05, + "loss": 0.3945, + "step": 44107 + }, + { + "epoch": 1.7, + "learning_rate": 2.175044883303411e-05, + "loss": 0.3363, + "step": 44108 + }, + { + "epoch": 1.7, + "learning_rate": 2.1749807642985382e-05, + "loss": 0.3723, + "step": 44109 + }, + { + "epoch": 1.7, + "learning_rate": 2.174916645293665e-05, + "loss": 0.3798, + "step": 44110 + }, + { + "epoch": 1.7, + "learning_rate": 2.174852526288792e-05, + "loss": 0.4158, + "step": 44111 + }, + { + "epoch": 1.7, + "learning_rate": 2.174788407283919e-05, + "loss": 0.2797, + "step": 44112 + }, + { + "epoch": 1.7, + "learning_rate": 2.1747242882790462e-05, + "loss": 0.4706, + "step": 44113 + }, + { + "epoch": 1.7, + "learning_rate": 2.174660169274173e-05, + "loss": 0.3142, + "step": 44114 + }, + { + "epoch": 1.7, + "learning_rate": 2.1745960502693e-05, + "loss": 0.4814, + "step": 44115 + }, + { + "epoch": 1.7, + "learning_rate": 2.1745319312644267e-05, + "loss": 0.3266, + "step": 44116 + }, + { + "epoch": 1.7, + "learning_rate": 2.174467812259554e-05, + "loss": 0.2451, + "step": 44117 + }, + { + "epoch": 1.7, + "learning_rate": 2.1744036932546807e-05, + "loss": 0.2759, + "step": 44118 + }, + { + "epoch": 1.7, + "learning_rate": 2.1743395742498076e-05, + "loss": 0.3625, + "step": 44119 + }, + { + "epoch": 1.7, + "learning_rate": 2.1742754552449347e-05, + "loss": 0.3055, + "step": 44120 + }, + { + "epoch": 1.7, + "learning_rate": 2.174211336240062e-05, + "loss": 0.4087, + "step": 44121 + }, + { + "epoch": 1.7, + "learning_rate": 2.1741472172351888e-05, + "loss": 0.3505, + "step": 44122 + }, + { + "epoch": 1.7, + "learning_rate": 2.1740830982303156e-05, + "loss": 0.3065, + "step": 44123 + }, + { + "epoch": 1.7, + "learning_rate": 2.1740189792254424e-05, + "loss": 0.3774, + "step": 44124 + }, + { + "epoch": 1.7, + "learning_rate": 2.1739548602205696e-05, + "loss": 0.282, + "step": 44125 + }, + { + "epoch": 1.7, + "learning_rate": 2.1738907412156964e-05, + "loss": 0.397, + "step": 44126 + }, + { + "epoch": 1.7, + "learning_rate": 2.1738266222108233e-05, + "loss": 0.3895, + "step": 44127 + }, + { + "epoch": 1.7, + "learning_rate": 2.1737625032059504e-05, + "loss": 0.3772, + "step": 44128 + }, + { + "epoch": 1.7, + "learning_rate": 2.1736983842010773e-05, + "loss": 0.493, + "step": 44129 + }, + { + "epoch": 1.7, + "learning_rate": 2.1736342651962044e-05, + "loss": 0.3636, + "step": 44130 + }, + { + "epoch": 1.7, + "learning_rate": 2.1735701461913313e-05, + "loss": 0.2921, + "step": 44131 + }, + { + "epoch": 1.7, + "learning_rate": 2.173506027186458e-05, + "loss": 0.4165, + "step": 44132 + }, + { + "epoch": 1.7, + "learning_rate": 2.1734419081815853e-05, + "loss": 0.3918, + "step": 44133 + }, + { + "epoch": 1.7, + "learning_rate": 2.173377789176712e-05, + "loss": 0.2983, + "step": 44134 + }, + { + "epoch": 1.7, + "learning_rate": 2.173313670171839e-05, + "loss": 0.2319, + "step": 44135 + }, + { + "epoch": 1.7, + "learning_rate": 2.1732495511669658e-05, + "loss": 0.2706, + "step": 44136 + }, + { + "epoch": 1.7, + "learning_rate": 2.173185432162093e-05, + "loss": 0.416, + "step": 44137 + }, + { + "epoch": 1.7, + "learning_rate": 2.1731213131572198e-05, + "loss": 0.222, + "step": 44138 + }, + { + "epoch": 1.7, + "learning_rate": 2.173057194152347e-05, + "loss": 0.3547, + "step": 44139 + }, + { + "epoch": 1.7, + "learning_rate": 2.1729930751474738e-05, + "loss": 0.3191, + "step": 44140 + }, + { + "epoch": 1.7, + "learning_rate": 2.172928956142601e-05, + "loss": 0.3321, + "step": 44141 + }, + { + "epoch": 1.7, + "learning_rate": 2.1728648371377278e-05, + "loss": 0.27, + "step": 44142 + }, + { + "epoch": 1.7, + "learning_rate": 2.1728007181328546e-05, + "loss": 0.364, + "step": 44143 + }, + { + "epoch": 1.7, + "learning_rate": 2.1727365991279814e-05, + "loss": 0.3195, + "step": 44144 + }, + { + "epoch": 1.7, + "learning_rate": 2.1726724801231086e-05, + "loss": 0.3308, + "step": 44145 + }, + { + "epoch": 1.7, + "learning_rate": 2.1726083611182355e-05, + "loss": 0.3128, + "step": 44146 + }, + { + "epoch": 1.7, + "learning_rate": 2.1725442421133623e-05, + "loss": 0.327, + "step": 44147 + }, + { + "epoch": 1.7, + "learning_rate": 2.1724801231084895e-05, + "loss": 0.2706, + "step": 44148 + }, + { + "epoch": 1.7, + "learning_rate": 2.1724160041036166e-05, + "loss": 0.3852, + "step": 44149 + }, + { + "epoch": 1.7, + "learning_rate": 2.1723518850987435e-05, + "loss": 0.2509, + "step": 44150 + }, + { + "epoch": 1.7, + "learning_rate": 2.1722877660938703e-05, + "loss": 0.426, + "step": 44151 + }, + { + "epoch": 1.7, + "learning_rate": 2.172223647088997e-05, + "loss": 0.3658, + "step": 44152 + }, + { + "epoch": 1.7, + "learning_rate": 2.1721595280841243e-05, + "loss": 0.3287, + "step": 44153 + }, + { + "epoch": 1.7, + "learning_rate": 2.172095409079251e-05, + "loss": 0.3458, + "step": 44154 + }, + { + "epoch": 1.7, + "learning_rate": 2.172031290074378e-05, + "loss": 0.613, + "step": 44155 + }, + { + "epoch": 1.7, + "learning_rate": 2.171967171069505e-05, + "loss": 0.4017, + "step": 44156 + }, + { + "epoch": 1.7, + "learning_rate": 2.1719030520646323e-05, + "loss": 0.3017, + "step": 44157 + }, + { + "epoch": 1.7, + "learning_rate": 2.171838933059759e-05, + "loss": 0.4045, + "step": 44158 + }, + { + "epoch": 1.7, + "learning_rate": 2.171774814054886e-05, + "loss": 0.418, + "step": 44159 + }, + { + "epoch": 1.7, + "learning_rate": 2.1717106950500128e-05, + "loss": 0.2926, + "step": 44160 + }, + { + "epoch": 1.7, + "learning_rate": 2.17164657604514e-05, + "loss": 0.3605, + "step": 44161 + }, + { + "epoch": 1.7, + "learning_rate": 2.1715824570402668e-05, + "loss": 0.3979, + "step": 44162 + }, + { + "epoch": 1.7, + "learning_rate": 2.1715183380353936e-05, + "loss": 0.3256, + "step": 44163 + }, + { + "epoch": 1.7, + "learning_rate": 2.1714542190305208e-05, + "loss": 0.2237, + "step": 44164 + }, + { + "epoch": 1.7, + "learning_rate": 2.1713901000256476e-05, + "loss": 0.3847, + "step": 44165 + }, + { + "epoch": 1.7, + "learning_rate": 2.1713259810207748e-05, + "loss": 0.3295, + "step": 44166 + }, + { + "epoch": 1.7, + "learning_rate": 2.1712618620159017e-05, + "loss": 0.4463, + "step": 44167 + }, + { + "epoch": 1.7, + "learning_rate": 2.1711977430110285e-05, + "loss": 0.2331, + "step": 44168 + }, + { + "epoch": 1.7, + "learning_rate": 2.1711336240061557e-05, + "loss": 0.3247, + "step": 44169 + }, + { + "epoch": 1.7, + "learning_rate": 2.1710695050012825e-05, + "loss": 0.2426, + "step": 44170 + }, + { + "epoch": 1.7, + "learning_rate": 2.1710053859964093e-05, + "loss": 0.3172, + "step": 44171 + }, + { + "epoch": 1.7, + "learning_rate": 2.1709412669915365e-05, + "loss": 0.3575, + "step": 44172 + }, + { + "epoch": 1.7, + "learning_rate": 2.1708771479866633e-05, + "loss": 0.4305, + "step": 44173 + }, + { + "epoch": 1.7, + "learning_rate": 2.17081302898179e-05, + "loss": 0.3947, + "step": 44174 + }, + { + "epoch": 1.7, + "learning_rate": 2.1707489099769173e-05, + "loss": 0.406, + "step": 44175 + }, + { + "epoch": 1.7, + "learning_rate": 2.170684790972044e-05, + "loss": 0.2625, + "step": 44176 + }, + { + "epoch": 1.7, + "learning_rate": 2.1706206719671713e-05, + "loss": 0.3686, + "step": 44177 + }, + { + "epoch": 1.7, + "learning_rate": 2.170556552962298e-05, + "loss": 0.2895, + "step": 44178 + }, + { + "epoch": 1.7, + "learning_rate": 2.170492433957425e-05, + "loss": 0.2683, + "step": 44179 + }, + { + "epoch": 1.7, + "learning_rate": 2.170428314952552e-05, + "loss": 0.3409, + "step": 44180 + }, + { + "epoch": 1.7, + "learning_rate": 2.170364195947679e-05, + "loss": 0.372, + "step": 44181 + }, + { + "epoch": 1.7, + "learning_rate": 2.170300076942806e-05, + "loss": 0.2188, + "step": 44182 + }, + { + "epoch": 1.7, + "learning_rate": 2.1702359579379327e-05, + "loss": 0.3986, + "step": 44183 + }, + { + "epoch": 1.7, + "learning_rate": 2.17017183893306e-05, + "loss": 0.2368, + "step": 44184 + }, + { + "epoch": 1.7, + "learning_rate": 2.170107719928187e-05, + "loss": 0.5489, + "step": 44185 + }, + { + "epoch": 1.7, + "learning_rate": 2.170043600923314e-05, + "loss": 0.2531, + "step": 44186 + }, + { + "epoch": 1.7, + "learning_rate": 2.1699794819184407e-05, + "loss": 0.4299, + "step": 44187 + }, + { + "epoch": 1.7, + "learning_rate": 2.1699153629135675e-05, + "loss": 0.1821, + "step": 44188 + }, + { + "epoch": 1.7, + "learning_rate": 2.1698512439086947e-05, + "loss": 0.4826, + "step": 44189 + }, + { + "epoch": 1.7, + "learning_rate": 2.1697871249038215e-05, + "loss": 0.5014, + "step": 44190 + }, + { + "epoch": 1.7, + "learning_rate": 2.1697230058989484e-05, + "loss": 0.2731, + "step": 44191 + }, + { + "epoch": 1.7, + "learning_rate": 2.1696588868940755e-05, + "loss": 0.2578, + "step": 44192 + }, + { + "epoch": 1.7, + "learning_rate": 2.1695947678892027e-05, + "loss": 0.4155, + "step": 44193 + }, + { + "epoch": 1.7, + "learning_rate": 2.1695306488843295e-05, + "loss": 0.3022, + "step": 44194 + }, + { + "epoch": 1.7, + "learning_rate": 2.1694665298794564e-05, + "loss": 0.2783, + "step": 44195 + }, + { + "epoch": 1.7, + "learning_rate": 2.1694024108745832e-05, + "loss": 0.3113, + "step": 44196 + }, + { + "epoch": 1.7, + "learning_rate": 2.1693382918697104e-05, + "loss": 0.3618, + "step": 44197 + }, + { + "epoch": 1.7, + "learning_rate": 2.1692741728648372e-05, + "loss": 0.3656, + "step": 44198 + }, + { + "epoch": 1.7, + "learning_rate": 2.169210053859964e-05, + "loss": 0.3808, + "step": 44199 + }, + { + "epoch": 1.7, + "learning_rate": 2.1691459348550912e-05, + "loss": 0.3414, + "step": 44200 + }, + { + "epoch": 1.7, + "learning_rate": 2.1690818158502184e-05, + "loss": 0.3932, + "step": 44201 + }, + { + "epoch": 1.7, + "learning_rate": 2.1690176968453452e-05, + "loss": 0.2698, + "step": 44202 + }, + { + "epoch": 1.7, + "learning_rate": 2.168953577840472e-05, + "loss": 0.3352, + "step": 44203 + }, + { + "epoch": 1.7, + "learning_rate": 2.168889458835599e-05, + "loss": 0.3636, + "step": 44204 + }, + { + "epoch": 1.7, + "learning_rate": 2.168825339830726e-05, + "loss": 0.3698, + "step": 44205 + }, + { + "epoch": 1.7, + "learning_rate": 2.168761220825853e-05, + "loss": 0.4382, + "step": 44206 + }, + { + "epoch": 1.7, + "learning_rate": 2.1686971018209797e-05, + "loss": 0.4287, + "step": 44207 + }, + { + "epoch": 1.7, + "learning_rate": 2.168632982816107e-05, + "loss": 0.3003, + "step": 44208 + }, + { + "epoch": 1.7, + "learning_rate": 2.1685688638112337e-05, + "loss": 0.3086, + "step": 44209 + }, + { + "epoch": 1.7, + "learning_rate": 2.168504744806361e-05, + "loss": 0.3977, + "step": 44210 + }, + { + "epoch": 1.7, + "learning_rate": 2.1684406258014877e-05, + "loss": 0.3874, + "step": 44211 + }, + { + "epoch": 1.7, + "learning_rate": 2.1683765067966146e-05, + "loss": 0.2533, + "step": 44212 + }, + { + "epoch": 1.7, + "learning_rate": 2.1683123877917417e-05, + "loss": 0.2994, + "step": 44213 + }, + { + "epoch": 1.7, + "learning_rate": 2.1682482687868686e-05, + "loss": 0.4148, + "step": 44214 + }, + { + "epoch": 1.7, + "learning_rate": 2.1681841497819954e-05, + "loss": 0.347, + "step": 44215 + }, + { + "epoch": 1.7, + "learning_rate": 2.1681200307771222e-05, + "loss": 0.5175, + "step": 44216 + }, + { + "epoch": 1.7, + "learning_rate": 2.1680559117722494e-05, + "loss": 0.2826, + "step": 44217 + }, + { + "epoch": 1.7, + "learning_rate": 2.1679917927673762e-05, + "loss": 0.338, + "step": 44218 + }, + { + "epoch": 1.7, + "learning_rate": 2.1679276737625034e-05, + "loss": 0.3652, + "step": 44219 + }, + { + "epoch": 1.7, + "learning_rate": 2.1678635547576302e-05, + "loss": 0.3328, + "step": 44220 + }, + { + "epoch": 1.7, + "learning_rate": 2.1677994357527574e-05, + "loss": 0.2447, + "step": 44221 + }, + { + "epoch": 1.7, + "learning_rate": 2.1677353167478842e-05, + "loss": 0.3297, + "step": 44222 + }, + { + "epoch": 1.7, + "learning_rate": 2.167671197743011e-05, + "loss": 0.3368, + "step": 44223 + }, + { + "epoch": 1.7, + "learning_rate": 2.167607078738138e-05, + "loss": 0.3247, + "step": 44224 + }, + { + "epoch": 1.7, + "learning_rate": 2.167542959733265e-05, + "loss": 0.2564, + "step": 44225 + }, + { + "epoch": 1.7, + "learning_rate": 2.167478840728392e-05, + "loss": 0.5199, + "step": 44226 + }, + { + "epoch": 1.7, + "learning_rate": 2.1674147217235187e-05, + "loss": 0.4009, + "step": 44227 + }, + { + "epoch": 1.7, + "learning_rate": 2.167350602718646e-05, + "loss": 0.3807, + "step": 44228 + }, + { + "epoch": 1.7, + "learning_rate": 2.167286483713773e-05, + "loss": 0.362, + "step": 44229 + }, + { + "epoch": 1.7, + "learning_rate": 2.1672223647089e-05, + "loss": 0.2827, + "step": 44230 + }, + { + "epoch": 1.7, + "learning_rate": 2.1671582457040267e-05, + "loss": 0.3765, + "step": 44231 + }, + { + "epoch": 1.7, + "learning_rate": 2.1670941266991536e-05, + "loss": 0.3258, + "step": 44232 + }, + { + "epoch": 1.7, + "learning_rate": 2.1670300076942808e-05, + "loss": 0.2873, + "step": 44233 + }, + { + "epoch": 1.7, + "learning_rate": 2.1669658886894076e-05, + "loss": 0.387, + "step": 44234 + }, + { + "epoch": 1.7, + "learning_rate": 2.1669017696845344e-05, + "loss": 0.3347, + "step": 44235 + }, + { + "epoch": 1.7, + "learning_rate": 2.1668376506796616e-05, + "loss": 0.3154, + "step": 44236 + }, + { + "epoch": 1.7, + "learning_rate": 2.1667735316747888e-05, + "loss": 0.3662, + "step": 44237 + }, + { + "epoch": 1.7, + "learning_rate": 2.1667094126699156e-05, + "loss": 0.4724, + "step": 44238 + }, + { + "epoch": 1.7, + "learning_rate": 2.1666452936650424e-05, + "loss": 0.4057, + "step": 44239 + }, + { + "epoch": 1.7, + "learning_rate": 2.1665811746601693e-05, + "loss": 0.2764, + "step": 44240 + }, + { + "epoch": 1.7, + "learning_rate": 2.1665170556552964e-05, + "loss": 0.3684, + "step": 44241 + }, + { + "epoch": 1.7, + "learning_rate": 2.1664529366504233e-05, + "loss": 0.3881, + "step": 44242 + }, + { + "epoch": 1.7, + "learning_rate": 2.16638881764555e-05, + "loss": 0.5235, + "step": 44243 + }, + { + "epoch": 1.7, + "learning_rate": 2.1663246986406773e-05, + "loss": 0.3768, + "step": 44244 + }, + { + "epoch": 1.7, + "learning_rate": 2.166260579635804e-05, + "loss": 0.2499, + "step": 44245 + }, + { + "epoch": 1.7, + "learning_rate": 2.1661964606309313e-05, + "loss": 0.3792, + "step": 44246 + }, + { + "epoch": 1.7, + "learning_rate": 2.166132341626058e-05, + "loss": 0.377, + "step": 44247 + }, + { + "epoch": 1.7, + "learning_rate": 2.166068222621185e-05, + "loss": 0.4955, + "step": 44248 + }, + { + "epoch": 1.7, + "learning_rate": 2.166004103616312e-05, + "loss": 0.3721, + "step": 44249 + }, + { + "epoch": 1.7, + "learning_rate": 2.165939984611439e-05, + "loss": 0.2491, + "step": 44250 + }, + { + "epoch": 1.7, + "learning_rate": 2.1658758656065658e-05, + "loss": 0.2867, + "step": 44251 + }, + { + "epoch": 1.7, + "learning_rate": 2.165811746601693e-05, + "loss": 0.5119, + "step": 44252 + }, + { + "epoch": 1.7, + "learning_rate": 2.1657476275968198e-05, + "loss": 0.2283, + "step": 44253 + }, + { + "epoch": 1.7, + "learning_rate": 2.1656835085919466e-05, + "loss": 0.2769, + "step": 44254 + }, + { + "epoch": 1.7, + "learning_rate": 2.1656193895870738e-05, + "loss": 0.3435, + "step": 44255 + }, + { + "epoch": 1.7, + "learning_rate": 2.1655552705822006e-05, + "loss": 0.3801, + "step": 44256 + }, + { + "epoch": 1.7, + "learning_rate": 2.1654911515773278e-05, + "loss": 0.3735, + "step": 44257 + }, + { + "epoch": 1.7, + "learning_rate": 2.1654270325724546e-05, + "loss": 0.2908, + "step": 44258 + }, + { + "epoch": 1.7, + "learning_rate": 2.1653629135675815e-05, + "loss": 0.2723, + "step": 44259 + }, + { + "epoch": 1.7, + "learning_rate": 2.1652987945627083e-05, + "loss": 0.4348, + "step": 44260 + }, + { + "epoch": 1.7, + "learning_rate": 2.1652346755578355e-05, + "loss": 0.3361, + "step": 44261 + }, + { + "epoch": 1.7, + "learning_rate": 2.1651705565529623e-05, + "loss": 0.3214, + "step": 44262 + }, + { + "epoch": 1.7, + "learning_rate": 2.165106437548089e-05, + "loss": 0.3527, + "step": 44263 + }, + { + "epoch": 1.7, + "learning_rate": 2.1650423185432163e-05, + "loss": 0.2754, + "step": 44264 + }, + { + "epoch": 1.7, + "learning_rate": 2.1649781995383435e-05, + "loss": 0.3344, + "step": 44265 + }, + { + "epoch": 1.7, + "learning_rate": 2.1649140805334703e-05, + "loss": 0.3307, + "step": 44266 + }, + { + "epoch": 1.7, + "learning_rate": 2.164849961528597e-05, + "loss": 0.2877, + "step": 44267 + }, + { + "epoch": 1.7, + "learning_rate": 2.164785842523724e-05, + "loss": 0.3975, + "step": 44268 + }, + { + "epoch": 1.7, + "learning_rate": 2.164721723518851e-05, + "loss": 0.2854, + "step": 44269 + }, + { + "epoch": 1.7, + "learning_rate": 2.164657604513978e-05, + "loss": 0.3255, + "step": 44270 + }, + { + "epoch": 1.7, + "learning_rate": 2.1645934855091048e-05, + "loss": 0.291, + "step": 44271 + }, + { + "epoch": 1.7, + "learning_rate": 2.164529366504232e-05, + "loss": 0.3777, + "step": 44272 + }, + { + "epoch": 1.7, + "learning_rate": 2.164465247499359e-05, + "loss": 0.1997, + "step": 44273 + }, + { + "epoch": 1.7, + "learning_rate": 2.164401128494486e-05, + "loss": 0.3, + "step": 44274 + }, + { + "epoch": 1.7, + "learning_rate": 2.1643370094896128e-05, + "loss": 0.2974, + "step": 44275 + }, + { + "epoch": 1.7, + "learning_rate": 2.1642728904847396e-05, + "loss": 0.2408, + "step": 44276 + }, + { + "epoch": 1.7, + "learning_rate": 2.1642087714798668e-05, + "loss": 0.5287, + "step": 44277 + }, + { + "epoch": 1.7, + "learning_rate": 2.1641446524749937e-05, + "loss": 0.4037, + "step": 44278 + }, + { + "epoch": 1.7, + "learning_rate": 2.1640805334701205e-05, + "loss": 0.3569, + "step": 44279 + }, + { + "epoch": 1.7, + "learning_rate": 2.1640164144652477e-05, + "loss": 0.4294, + "step": 44280 + }, + { + "epoch": 1.7, + "learning_rate": 2.1639522954603745e-05, + "loss": 0.3076, + "step": 44281 + }, + { + "epoch": 1.7, + "learning_rate": 2.1638881764555017e-05, + "loss": 0.2311, + "step": 44282 + }, + { + "epoch": 1.7, + "learning_rate": 2.1638240574506285e-05, + "loss": 0.4341, + "step": 44283 + }, + { + "epoch": 1.7, + "learning_rate": 2.1637599384457553e-05, + "loss": 0.1998, + "step": 44284 + }, + { + "epoch": 1.7, + "learning_rate": 2.1636958194408825e-05, + "loss": 0.4174, + "step": 44285 + }, + { + "epoch": 1.7, + "learning_rate": 2.1636317004360093e-05, + "loss": 0.4148, + "step": 44286 + }, + { + "epoch": 1.7, + "learning_rate": 2.163567581431136e-05, + "loss": 0.2851, + "step": 44287 + }, + { + "epoch": 1.7, + "learning_rate": 2.1635034624262633e-05, + "loss": 0.2754, + "step": 44288 + }, + { + "epoch": 1.7, + "learning_rate": 2.16343934342139e-05, + "loss": 0.3609, + "step": 44289 + }, + { + "epoch": 1.7, + "learning_rate": 2.163375224416517e-05, + "loss": 0.4064, + "step": 44290 + }, + { + "epoch": 1.7, + "learning_rate": 2.1633111054116442e-05, + "loss": 0.452, + "step": 44291 + }, + { + "epoch": 1.7, + "learning_rate": 2.163246986406771e-05, + "loss": 0.3523, + "step": 44292 + }, + { + "epoch": 1.7, + "learning_rate": 2.1631828674018982e-05, + "loss": 0.4534, + "step": 44293 + }, + { + "epoch": 1.7, + "learning_rate": 2.163118748397025e-05, + "loss": 0.4491, + "step": 44294 + }, + { + "epoch": 1.7, + "learning_rate": 2.163054629392152e-05, + "loss": 0.2598, + "step": 44295 + }, + { + "epoch": 1.7, + "learning_rate": 2.162990510387279e-05, + "loss": 0.4114, + "step": 44296 + }, + { + "epoch": 1.7, + "learning_rate": 2.162926391382406e-05, + "loss": 0.5025, + "step": 44297 + }, + { + "epoch": 1.7, + "learning_rate": 2.1628622723775327e-05, + "loss": 0.3454, + "step": 44298 + }, + { + "epoch": 1.7, + "learning_rate": 2.16279815337266e-05, + "loss": 0.3333, + "step": 44299 + }, + { + "epoch": 1.7, + "learning_rate": 2.1627340343677867e-05, + "loss": 0.3642, + "step": 44300 + }, + { + "epoch": 1.7, + "learning_rate": 2.162669915362914e-05, + "loss": 0.4179, + "step": 44301 + }, + { + "epoch": 1.7, + "learning_rate": 2.1626057963580407e-05, + "loss": 0.2922, + "step": 44302 + }, + { + "epoch": 1.7, + "learning_rate": 2.1625416773531675e-05, + "loss": 0.5189, + "step": 44303 + }, + { + "epoch": 1.7, + "learning_rate": 2.1624775583482944e-05, + "loss": 0.4727, + "step": 44304 + }, + { + "epoch": 1.7, + "learning_rate": 2.1624134393434215e-05, + "loss": 0.2207, + "step": 44305 + }, + { + "epoch": 1.7, + "learning_rate": 2.1623493203385484e-05, + "loss": 0.2459, + "step": 44306 + }, + { + "epoch": 1.7, + "learning_rate": 2.1622852013336752e-05, + "loss": 0.3176, + "step": 44307 + }, + { + "epoch": 1.7, + "learning_rate": 2.1622210823288024e-05, + "loss": 0.4067, + "step": 44308 + }, + { + "epoch": 1.7, + "learning_rate": 2.1621569633239295e-05, + "loss": 0.3168, + "step": 44309 + }, + { + "epoch": 1.7, + "learning_rate": 2.1620928443190564e-05, + "loss": 0.4171, + "step": 44310 + }, + { + "epoch": 1.7, + "learning_rate": 2.1620287253141832e-05, + "loss": 0.3939, + "step": 44311 + }, + { + "epoch": 1.7, + "learning_rate": 2.16196460630931e-05, + "loss": 0.4285, + "step": 44312 + }, + { + "epoch": 1.7, + "learning_rate": 2.1619004873044372e-05, + "loss": 0.2808, + "step": 44313 + }, + { + "epoch": 1.7, + "learning_rate": 2.161836368299564e-05, + "loss": 0.3207, + "step": 44314 + }, + { + "epoch": 1.7, + "learning_rate": 2.161772249294691e-05, + "loss": 0.2807, + "step": 44315 + }, + { + "epoch": 1.7, + "learning_rate": 2.161708130289818e-05, + "loss": 0.3378, + "step": 44316 + }, + { + "epoch": 1.7, + "learning_rate": 2.1616440112849452e-05, + "loss": 0.3243, + "step": 44317 + }, + { + "epoch": 1.7, + "learning_rate": 2.161579892280072e-05, + "loss": 0.4116, + "step": 44318 + }, + { + "epoch": 1.7, + "learning_rate": 2.161515773275199e-05, + "loss": 0.3116, + "step": 44319 + }, + { + "epoch": 1.7, + "learning_rate": 2.1614516542703257e-05, + "loss": 0.3245, + "step": 44320 + }, + { + "epoch": 1.7, + "learning_rate": 2.161387535265453e-05, + "loss": 0.2458, + "step": 44321 + }, + { + "epoch": 1.7, + "learning_rate": 2.1613234162605797e-05, + "loss": 0.2419, + "step": 44322 + }, + { + "epoch": 1.7, + "learning_rate": 2.1612592972557066e-05, + "loss": 0.1836, + "step": 44323 + }, + { + "epoch": 1.7, + "learning_rate": 2.1611951782508337e-05, + "loss": 0.2689, + "step": 44324 + }, + { + "epoch": 1.7, + "learning_rate": 2.1611310592459606e-05, + "loss": 0.4647, + "step": 44325 + }, + { + "epoch": 1.7, + "learning_rate": 2.1610669402410877e-05, + "loss": 0.4107, + "step": 44326 + }, + { + "epoch": 1.7, + "learning_rate": 2.1610028212362146e-05, + "loss": 0.3522, + "step": 44327 + }, + { + "epoch": 1.7, + "learning_rate": 2.1609387022313414e-05, + "loss": 0.4633, + "step": 44328 + }, + { + "epoch": 1.7, + "learning_rate": 2.1608745832264686e-05, + "loss": 0.5623, + "step": 44329 + }, + { + "epoch": 1.7, + "learning_rate": 2.1608104642215954e-05, + "loss": 0.5762, + "step": 44330 + }, + { + "epoch": 1.7, + "learning_rate": 2.1607463452167222e-05, + "loss": 0.3643, + "step": 44331 + }, + { + "epoch": 1.7, + "learning_rate": 2.1606822262118494e-05, + "loss": 0.4031, + "step": 44332 + }, + { + "epoch": 1.7, + "learning_rate": 2.1606181072069762e-05, + "loss": 0.3622, + "step": 44333 + }, + { + "epoch": 1.7, + "learning_rate": 2.160553988202103e-05, + "loss": 0.3861, + "step": 44334 + }, + { + "epoch": 1.7, + "learning_rate": 2.1604898691972302e-05, + "loss": 0.3549, + "step": 44335 + }, + { + "epoch": 1.7, + "learning_rate": 2.160425750192357e-05, + "loss": 0.3065, + "step": 44336 + }, + { + "epoch": 1.7, + "learning_rate": 2.1603616311874842e-05, + "loss": 0.276, + "step": 44337 + }, + { + "epoch": 1.7, + "learning_rate": 2.160297512182611e-05, + "loss": 0.2672, + "step": 44338 + }, + { + "epoch": 1.7, + "learning_rate": 2.160233393177738e-05, + "loss": 0.2562, + "step": 44339 + }, + { + "epoch": 1.7, + "learning_rate": 2.1601692741728647e-05, + "loss": 0.3478, + "step": 44340 + }, + { + "epoch": 1.7, + "learning_rate": 2.160105155167992e-05, + "loss": 0.2724, + "step": 44341 + }, + { + "epoch": 1.7, + "learning_rate": 2.1600410361631187e-05, + "loss": 0.4803, + "step": 44342 + }, + { + "epoch": 1.7, + "learning_rate": 2.1599769171582456e-05, + "loss": 0.3446, + "step": 44343 + }, + { + "epoch": 1.7, + "learning_rate": 2.1599127981533728e-05, + "loss": 0.3165, + "step": 44344 + }, + { + "epoch": 1.7, + "learning_rate": 2.1598486791485e-05, + "loss": 0.287, + "step": 44345 + }, + { + "epoch": 1.7, + "learning_rate": 2.1597845601436268e-05, + "loss": 0.2992, + "step": 44346 + }, + { + "epoch": 1.7, + "learning_rate": 2.1597204411387536e-05, + "loss": 0.5595, + "step": 44347 + }, + { + "epoch": 1.71, + "learning_rate": 2.1596563221338804e-05, + "loss": 0.2846, + "step": 44348 + }, + { + "epoch": 1.71, + "learning_rate": 2.1595922031290076e-05, + "loss": 0.4361, + "step": 44349 + }, + { + "epoch": 1.71, + "learning_rate": 2.1595280841241344e-05, + "loss": 0.1978, + "step": 44350 + }, + { + "epoch": 1.71, + "learning_rate": 2.1594639651192613e-05, + "loss": 0.3933, + "step": 44351 + }, + { + "epoch": 1.71, + "learning_rate": 2.1593998461143884e-05, + "loss": 0.2873, + "step": 44352 + }, + { + "epoch": 1.71, + "learning_rate": 2.1593357271095156e-05, + "loss": 0.4071, + "step": 44353 + }, + { + "epoch": 1.71, + "learning_rate": 2.1592716081046424e-05, + "loss": 0.336, + "step": 44354 + }, + { + "epoch": 1.71, + "learning_rate": 2.1592074890997693e-05, + "loss": 0.2841, + "step": 44355 + }, + { + "epoch": 1.71, + "learning_rate": 2.159143370094896e-05, + "loss": 0.286, + "step": 44356 + }, + { + "epoch": 1.71, + "learning_rate": 2.1590792510900233e-05, + "loss": 0.3338, + "step": 44357 + }, + { + "epoch": 1.71, + "learning_rate": 2.15901513208515e-05, + "loss": 0.3642, + "step": 44358 + }, + { + "epoch": 1.71, + "learning_rate": 2.158951013080277e-05, + "loss": 0.216, + "step": 44359 + }, + { + "epoch": 1.71, + "learning_rate": 2.158886894075404e-05, + "loss": 0.2848, + "step": 44360 + }, + { + "epoch": 1.71, + "learning_rate": 2.158822775070531e-05, + "loss": 0.3459, + "step": 44361 + }, + { + "epoch": 1.71, + "learning_rate": 2.158758656065658e-05, + "loss": 0.3839, + "step": 44362 + }, + { + "epoch": 1.71, + "learning_rate": 2.158694537060785e-05, + "loss": 0.3539, + "step": 44363 + }, + { + "epoch": 1.71, + "learning_rate": 2.1586304180559118e-05, + "loss": 0.237, + "step": 44364 + }, + { + "epoch": 1.71, + "learning_rate": 2.158566299051039e-05, + "loss": 0.3598, + "step": 44365 + }, + { + "epoch": 1.71, + "learning_rate": 2.1585021800461658e-05, + "loss": 0.4391, + "step": 44366 + }, + { + "epoch": 1.71, + "learning_rate": 2.1584380610412926e-05, + "loss": 0.2237, + "step": 44367 + }, + { + "epoch": 1.71, + "learning_rate": 2.1583739420364198e-05, + "loss": 0.3078, + "step": 44368 + }, + { + "epoch": 1.71, + "learning_rate": 2.1583098230315466e-05, + "loss": 0.2487, + "step": 44369 + }, + { + "epoch": 1.71, + "learning_rate": 2.1582457040266735e-05, + "loss": 0.4046, + "step": 44370 + }, + { + "epoch": 1.71, + "learning_rate": 2.1581815850218006e-05, + "loss": 0.4299, + "step": 44371 + }, + { + "epoch": 1.71, + "learning_rate": 2.1581174660169275e-05, + "loss": 0.2989, + "step": 44372 + }, + { + "epoch": 1.71, + "learning_rate": 2.1580533470120546e-05, + "loss": 0.327, + "step": 44373 + }, + { + "epoch": 1.71, + "learning_rate": 2.1579892280071815e-05, + "loss": 0.3674, + "step": 44374 + }, + { + "epoch": 1.71, + "learning_rate": 2.1579251090023083e-05, + "loss": 0.2612, + "step": 44375 + }, + { + "epoch": 1.71, + "learning_rate": 2.1578609899974355e-05, + "loss": 0.3511, + "step": 44376 + }, + { + "epoch": 1.71, + "learning_rate": 2.1577968709925623e-05, + "loss": 0.3855, + "step": 44377 + }, + { + "epoch": 1.71, + "learning_rate": 2.157732751987689e-05, + "loss": 0.385, + "step": 44378 + }, + { + "epoch": 1.71, + "learning_rate": 2.157668632982816e-05, + "loss": 0.5111, + "step": 44379 + }, + { + "epoch": 1.71, + "learning_rate": 2.157604513977943e-05, + "loss": 0.4044, + "step": 44380 + }, + { + "epoch": 1.71, + "learning_rate": 2.1575403949730703e-05, + "loss": 0.3986, + "step": 44381 + }, + { + "epoch": 1.71, + "learning_rate": 2.157476275968197e-05, + "loss": 0.2488, + "step": 44382 + }, + { + "epoch": 1.71, + "learning_rate": 2.157412156963324e-05, + "loss": 0.2826, + "step": 44383 + }, + { + "epoch": 1.71, + "learning_rate": 2.1573480379584508e-05, + "loss": 0.287, + "step": 44384 + }, + { + "epoch": 1.71, + "learning_rate": 2.157283918953578e-05, + "loss": 0.203, + "step": 44385 + }, + { + "epoch": 1.71, + "learning_rate": 2.1572197999487048e-05, + "loss": 0.3895, + "step": 44386 + }, + { + "epoch": 1.71, + "learning_rate": 2.1571556809438316e-05, + "loss": 0.3458, + "step": 44387 + }, + { + "epoch": 1.71, + "learning_rate": 2.1570915619389588e-05, + "loss": 0.3716, + "step": 44388 + }, + { + "epoch": 1.71, + "learning_rate": 2.157027442934086e-05, + "loss": 0.2699, + "step": 44389 + }, + { + "epoch": 1.71, + "learning_rate": 2.1569633239292128e-05, + "loss": 0.3359, + "step": 44390 + }, + { + "epoch": 1.71, + "learning_rate": 2.1568992049243397e-05, + "loss": 0.2869, + "step": 44391 + }, + { + "epoch": 1.71, + "learning_rate": 2.1568350859194665e-05, + "loss": 0.3494, + "step": 44392 + }, + { + "epoch": 1.71, + "learning_rate": 2.1567709669145937e-05, + "loss": 0.3571, + "step": 44393 + }, + { + "epoch": 1.71, + "learning_rate": 2.1567068479097205e-05, + "loss": 0.5339, + "step": 44394 + }, + { + "epoch": 1.71, + "learning_rate": 2.1566427289048473e-05, + "loss": 0.3584, + "step": 44395 + }, + { + "epoch": 1.71, + "learning_rate": 2.1565786098999745e-05, + "loss": 0.2849, + "step": 44396 + }, + { + "epoch": 1.71, + "learning_rate": 2.1565144908951017e-05, + "loss": 0.3713, + "step": 44397 + }, + { + "epoch": 1.71, + "learning_rate": 2.1564503718902285e-05, + "loss": 0.302, + "step": 44398 + }, + { + "epoch": 1.71, + "learning_rate": 2.1563862528853553e-05, + "loss": 0.3878, + "step": 44399 + }, + { + "epoch": 1.71, + "learning_rate": 2.156322133880482e-05, + "loss": 0.4045, + "step": 44400 + }, + { + "epoch": 1.71, + "learning_rate": 2.1562580148756093e-05, + "loss": 0.2889, + "step": 44401 + }, + { + "epoch": 1.71, + "learning_rate": 2.1561938958707362e-05, + "loss": 0.3024, + "step": 44402 + }, + { + "epoch": 1.71, + "learning_rate": 2.156129776865863e-05, + "loss": 0.3733, + "step": 44403 + }, + { + "epoch": 1.71, + "learning_rate": 2.1560656578609902e-05, + "loss": 0.3083, + "step": 44404 + }, + { + "epoch": 1.71, + "learning_rate": 2.156001538856117e-05, + "loss": 0.3821, + "step": 44405 + }, + { + "epoch": 1.71, + "learning_rate": 2.1559374198512442e-05, + "loss": 0.2969, + "step": 44406 + }, + { + "epoch": 1.71, + "learning_rate": 2.155873300846371e-05, + "loss": 0.3449, + "step": 44407 + }, + { + "epoch": 1.71, + "learning_rate": 2.155809181841498e-05, + "loss": 0.4146, + "step": 44408 + }, + { + "epoch": 1.71, + "learning_rate": 2.155745062836625e-05, + "loss": 0.2838, + "step": 44409 + }, + { + "epoch": 1.71, + "learning_rate": 2.155680943831752e-05, + "loss": 0.2718, + "step": 44410 + }, + { + "epoch": 1.71, + "learning_rate": 2.1556168248268787e-05, + "loss": 0.2608, + "step": 44411 + }, + { + "epoch": 1.71, + "learning_rate": 2.155552705822006e-05, + "loss": 0.2946, + "step": 44412 + }, + { + "epoch": 1.71, + "learning_rate": 2.1554885868171327e-05, + "loss": 0.3528, + "step": 44413 + }, + { + "epoch": 1.71, + "learning_rate": 2.1554244678122595e-05, + "loss": 0.2925, + "step": 44414 + }, + { + "epoch": 1.71, + "learning_rate": 2.1553603488073867e-05, + "loss": 0.2781, + "step": 44415 + }, + { + "epoch": 1.71, + "learning_rate": 2.1552962298025135e-05, + "loss": 0.2394, + "step": 44416 + }, + { + "epoch": 1.71, + "learning_rate": 2.1552321107976407e-05, + "loss": 0.294, + "step": 44417 + }, + { + "epoch": 1.71, + "learning_rate": 2.1551679917927675e-05, + "loss": 0.2704, + "step": 44418 + }, + { + "epoch": 1.71, + "learning_rate": 2.1551038727878944e-05, + "loss": 0.2573, + "step": 44419 + }, + { + "epoch": 1.71, + "learning_rate": 2.1550397537830215e-05, + "loss": 0.5138, + "step": 44420 + }, + { + "epoch": 1.71, + "learning_rate": 2.1549756347781484e-05, + "loss": 0.293, + "step": 44421 + }, + { + "epoch": 1.71, + "learning_rate": 2.1549115157732752e-05, + "loss": 0.4254, + "step": 44422 + }, + { + "epoch": 1.71, + "learning_rate": 2.154847396768402e-05, + "loss": 0.4968, + "step": 44423 + }, + { + "epoch": 1.71, + "learning_rate": 2.1547832777635292e-05, + "loss": 0.2476, + "step": 44424 + }, + { + "epoch": 1.71, + "learning_rate": 2.1547191587586564e-05, + "loss": 0.379, + "step": 44425 + }, + { + "epoch": 1.71, + "learning_rate": 2.1546550397537832e-05, + "loss": 0.3333, + "step": 44426 + }, + { + "epoch": 1.71, + "learning_rate": 2.15459092074891e-05, + "loss": 0.3029, + "step": 44427 + }, + { + "epoch": 1.71, + "learning_rate": 2.154526801744037e-05, + "loss": 0.4067, + "step": 44428 + }, + { + "epoch": 1.71, + "learning_rate": 2.154462682739164e-05, + "loss": 0.2853, + "step": 44429 + }, + { + "epoch": 1.71, + "learning_rate": 2.154398563734291e-05, + "loss": 0.3803, + "step": 44430 + }, + { + "epoch": 1.71, + "learning_rate": 2.1543344447294177e-05, + "loss": 0.2934, + "step": 44431 + }, + { + "epoch": 1.71, + "learning_rate": 2.154270325724545e-05, + "loss": 0.4287, + "step": 44432 + }, + { + "epoch": 1.71, + "learning_rate": 2.154206206719672e-05, + "loss": 0.3943, + "step": 44433 + }, + { + "epoch": 1.71, + "learning_rate": 2.154142087714799e-05, + "loss": 0.4258, + "step": 44434 + }, + { + "epoch": 1.71, + "learning_rate": 2.1540779687099257e-05, + "loss": 0.3113, + "step": 44435 + }, + { + "epoch": 1.71, + "learning_rate": 2.1540138497050526e-05, + "loss": 0.3009, + "step": 44436 + }, + { + "epoch": 1.71, + "learning_rate": 2.1539497307001797e-05, + "loss": 0.2823, + "step": 44437 + }, + { + "epoch": 1.71, + "learning_rate": 2.1538856116953066e-05, + "loss": 0.3128, + "step": 44438 + }, + { + "epoch": 1.71, + "learning_rate": 2.1538214926904334e-05, + "loss": 0.3458, + "step": 44439 + }, + { + "epoch": 1.71, + "learning_rate": 2.1537573736855606e-05, + "loss": 0.3508, + "step": 44440 + }, + { + "epoch": 1.71, + "learning_rate": 2.1536932546806874e-05, + "loss": 0.3366, + "step": 44441 + }, + { + "epoch": 1.71, + "learning_rate": 2.1536291356758146e-05, + "loss": 0.4394, + "step": 44442 + }, + { + "epoch": 1.71, + "learning_rate": 2.1535650166709414e-05, + "loss": 0.2485, + "step": 44443 + }, + { + "epoch": 1.71, + "learning_rate": 2.1535008976660682e-05, + "loss": 0.353, + "step": 44444 + }, + { + "epoch": 1.71, + "learning_rate": 2.1534367786611954e-05, + "loss": 0.3913, + "step": 44445 + }, + { + "epoch": 1.71, + "learning_rate": 2.1533726596563222e-05, + "loss": 0.4266, + "step": 44446 + }, + { + "epoch": 1.71, + "learning_rate": 2.153308540651449e-05, + "loss": 0.3475, + "step": 44447 + }, + { + "epoch": 1.71, + "learning_rate": 2.1532444216465762e-05, + "loss": 0.2844, + "step": 44448 + }, + { + "epoch": 1.71, + "learning_rate": 2.153180302641703e-05, + "loss": 0.3802, + "step": 44449 + }, + { + "epoch": 1.71, + "learning_rate": 2.15311618363683e-05, + "loss": 0.3056, + "step": 44450 + }, + { + "epoch": 1.71, + "learning_rate": 2.153052064631957e-05, + "loss": 0.2905, + "step": 44451 + }, + { + "epoch": 1.71, + "learning_rate": 2.152987945627084e-05, + "loss": 0.2845, + "step": 44452 + }, + { + "epoch": 1.71, + "learning_rate": 2.152923826622211e-05, + "loss": 0.4054, + "step": 44453 + }, + { + "epoch": 1.71, + "learning_rate": 2.152859707617338e-05, + "loss": 0.4077, + "step": 44454 + }, + { + "epoch": 1.71, + "learning_rate": 2.1527955886124647e-05, + "loss": 0.2567, + "step": 44455 + }, + { + "epoch": 1.71, + "learning_rate": 2.152731469607592e-05, + "loss": 0.3303, + "step": 44456 + }, + { + "epoch": 1.71, + "learning_rate": 2.1526673506027188e-05, + "loss": 0.3918, + "step": 44457 + }, + { + "epoch": 1.71, + "learning_rate": 2.1526032315978456e-05, + "loss": 0.2136, + "step": 44458 + }, + { + "epoch": 1.71, + "learning_rate": 2.1525391125929724e-05, + "loss": 0.309, + "step": 44459 + }, + { + "epoch": 1.71, + "learning_rate": 2.1524749935880996e-05, + "loss": 0.267, + "step": 44460 + }, + { + "epoch": 1.71, + "learning_rate": 2.1524108745832268e-05, + "loss": 0.3383, + "step": 44461 + }, + { + "epoch": 1.71, + "learning_rate": 2.1523467555783536e-05, + "loss": 0.3117, + "step": 44462 + }, + { + "epoch": 1.71, + "learning_rate": 2.1522826365734804e-05, + "loss": 0.3611, + "step": 44463 + }, + { + "epoch": 1.71, + "learning_rate": 2.1522185175686073e-05, + "loss": 0.3508, + "step": 44464 + }, + { + "epoch": 1.71, + "learning_rate": 2.1521543985637344e-05, + "loss": 0.3046, + "step": 44465 + }, + { + "epoch": 1.71, + "learning_rate": 2.1520902795588613e-05, + "loss": 0.4123, + "step": 44466 + }, + { + "epoch": 1.71, + "learning_rate": 2.152026160553988e-05, + "loss": 0.3868, + "step": 44467 + }, + { + "epoch": 1.71, + "learning_rate": 2.1519620415491153e-05, + "loss": 0.3349, + "step": 44468 + }, + { + "epoch": 1.71, + "learning_rate": 2.1518979225442424e-05, + "loss": 0.2732, + "step": 44469 + }, + { + "epoch": 1.71, + "learning_rate": 2.1518338035393693e-05, + "loss": 0.3837, + "step": 44470 + }, + { + "epoch": 1.71, + "learning_rate": 2.151769684534496e-05, + "loss": 0.5054, + "step": 44471 + }, + { + "epoch": 1.71, + "learning_rate": 2.151705565529623e-05, + "loss": 0.3706, + "step": 44472 + }, + { + "epoch": 1.71, + "learning_rate": 2.15164144652475e-05, + "loss": 0.2848, + "step": 44473 + }, + { + "epoch": 1.71, + "learning_rate": 2.151577327519877e-05, + "loss": 0.2934, + "step": 44474 + }, + { + "epoch": 1.71, + "learning_rate": 2.1515132085150038e-05, + "loss": 0.3621, + "step": 44475 + }, + { + "epoch": 1.71, + "learning_rate": 2.151449089510131e-05, + "loss": 0.3792, + "step": 44476 + }, + { + "epoch": 1.71, + "learning_rate": 2.1513849705052578e-05, + "loss": 0.3334, + "step": 44477 + }, + { + "epoch": 1.71, + "learning_rate": 2.151320851500385e-05, + "loss": 0.458, + "step": 44478 + }, + { + "epoch": 1.71, + "learning_rate": 2.1512567324955118e-05, + "loss": 0.446, + "step": 44479 + }, + { + "epoch": 1.71, + "learning_rate": 2.1511926134906386e-05, + "loss": 0.4926, + "step": 44480 + }, + { + "epoch": 1.71, + "learning_rate": 2.1511284944857658e-05, + "loss": 0.2051, + "step": 44481 + }, + { + "epoch": 1.71, + "learning_rate": 2.1510643754808926e-05, + "loss": 0.3737, + "step": 44482 + }, + { + "epoch": 1.71, + "learning_rate": 2.1510002564760195e-05, + "loss": 0.3518, + "step": 44483 + }, + { + "epoch": 1.71, + "learning_rate": 2.1509361374711466e-05, + "loss": 0.2679, + "step": 44484 + }, + { + "epoch": 1.71, + "learning_rate": 2.1508720184662735e-05, + "loss": 0.4183, + "step": 44485 + }, + { + "epoch": 1.71, + "learning_rate": 2.1508078994614006e-05, + "loss": 0.4309, + "step": 44486 + }, + { + "epoch": 1.71, + "learning_rate": 2.1507437804565275e-05, + "loss": 0.2565, + "step": 44487 + }, + { + "epoch": 1.71, + "learning_rate": 2.1506796614516543e-05, + "loss": 0.2382, + "step": 44488 + }, + { + "epoch": 1.71, + "learning_rate": 2.1506155424467815e-05, + "loss": 0.2344, + "step": 44489 + }, + { + "epoch": 1.71, + "learning_rate": 2.1505514234419083e-05, + "loss": 0.2329, + "step": 44490 + }, + { + "epoch": 1.71, + "learning_rate": 2.150487304437035e-05, + "loss": 0.2743, + "step": 44491 + }, + { + "epoch": 1.71, + "learning_rate": 2.1504231854321623e-05, + "loss": 0.2488, + "step": 44492 + }, + { + "epoch": 1.71, + "learning_rate": 2.150359066427289e-05, + "loss": 0.4101, + "step": 44493 + }, + { + "epoch": 1.71, + "learning_rate": 2.150294947422416e-05, + "loss": 0.489, + "step": 44494 + }, + { + "epoch": 1.71, + "learning_rate": 2.150230828417543e-05, + "loss": 0.3711, + "step": 44495 + }, + { + "epoch": 1.71, + "learning_rate": 2.15016670941267e-05, + "loss": 0.4291, + "step": 44496 + }, + { + "epoch": 1.71, + "learning_rate": 2.150102590407797e-05, + "loss": 0.2452, + "step": 44497 + }, + { + "epoch": 1.71, + "learning_rate": 2.150038471402924e-05, + "loss": 0.4125, + "step": 44498 + }, + { + "epoch": 1.71, + "learning_rate": 2.1499743523980508e-05, + "loss": 0.3454, + "step": 44499 + }, + { + "epoch": 1.71, + "learning_rate": 2.149910233393178e-05, + "loss": 0.2732, + "step": 44500 + }, + { + "epoch": 1.71, + "learning_rate": 2.1498461143883048e-05, + "loss": 0.2024, + "step": 44501 + }, + { + "epoch": 1.71, + "learning_rate": 2.1497819953834317e-05, + "loss": 0.2227, + "step": 44502 + }, + { + "epoch": 1.71, + "learning_rate": 2.1497178763785585e-05, + "loss": 0.2969, + "step": 44503 + }, + { + "epoch": 1.71, + "learning_rate": 2.1496537573736857e-05, + "loss": 0.3331, + "step": 44504 + }, + { + "epoch": 1.71, + "learning_rate": 2.1495896383688128e-05, + "loss": 0.3078, + "step": 44505 + }, + { + "epoch": 1.71, + "learning_rate": 2.1495255193639397e-05, + "loss": 0.2873, + "step": 44506 + }, + { + "epoch": 1.71, + "learning_rate": 2.1494614003590665e-05, + "loss": 0.3584, + "step": 44507 + }, + { + "epoch": 1.71, + "learning_rate": 2.1493972813541933e-05, + "loss": 0.2304, + "step": 44508 + }, + { + "epoch": 1.71, + "learning_rate": 2.1493331623493205e-05, + "loss": 0.2812, + "step": 44509 + }, + { + "epoch": 1.71, + "learning_rate": 2.1492690433444473e-05, + "loss": 0.3721, + "step": 44510 + }, + { + "epoch": 1.71, + "learning_rate": 2.149204924339574e-05, + "loss": 0.3127, + "step": 44511 + }, + { + "epoch": 1.71, + "learning_rate": 2.1491408053347013e-05, + "loss": 0.3316, + "step": 44512 + }, + { + "epoch": 1.71, + "learning_rate": 2.1490766863298285e-05, + "loss": 0.3306, + "step": 44513 + }, + { + "epoch": 1.71, + "learning_rate": 2.1490125673249553e-05, + "loss": 0.3716, + "step": 44514 + }, + { + "epoch": 1.71, + "learning_rate": 2.1489484483200822e-05, + "loss": 0.379, + "step": 44515 + }, + { + "epoch": 1.71, + "learning_rate": 2.148884329315209e-05, + "loss": 0.3999, + "step": 44516 + }, + { + "epoch": 1.71, + "learning_rate": 2.1488202103103362e-05, + "loss": 0.3376, + "step": 44517 + }, + { + "epoch": 1.71, + "learning_rate": 2.148756091305463e-05, + "loss": 0.3562, + "step": 44518 + }, + { + "epoch": 1.71, + "learning_rate": 2.14869197230059e-05, + "loss": 0.3093, + "step": 44519 + }, + { + "epoch": 1.71, + "learning_rate": 2.148627853295717e-05, + "loss": 0.4462, + "step": 44520 + }, + { + "epoch": 1.71, + "learning_rate": 2.148563734290844e-05, + "loss": 0.2329, + "step": 44521 + }, + { + "epoch": 1.71, + "learning_rate": 2.148499615285971e-05, + "loss": 0.3026, + "step": 44522 + }, + { + "epoch": 1.71, + "learning_rate": 2.148435496281098e-05, + "loss": 0.4528, + "step": 44523 + }, + { + "epoch": 1.71, + "learning_rate": 2.1483713772762247e-05, + "loss": 0.3762, + "step": 44524 + }, + { + "epoch": 1.71, + "learning_rate": 2.148307258271352e-05, + "loss": 0.3996, + "step": 44525 + }, + { + "epoch": 1.71, + "learning_rate": 2.1482431392664787e-05, + "loss": 0.3035, + "step": 44526 + }, + { + "epoch": 1.71, + "learning_rate": 2.1481790202616055e-05, + "loss": 0.3329, + "step": 44527 + }, + { + "epoch": 1.71, + "learning_rate": 2.1481149012567327e-05, + "loss": 0.3219, + "step": 44528 + }, + { + "epoch": 1.71, + "learning_rate": 2.1480507822518595e-05, + "loss": 0.3002, + "step": 44529 + }, + { + "epoch": 1.71, + "learning_rate": 2.1479866632469864e-05, + "loss": 0.4949, + "step": 44530 + }, + { + "epoch": 1.71, + "learning_rate": 2.1479225442421135e-05, + "loss": 0.3775, + "step": 44531 + }, + { + "epoch": 1.71, + "learning_rate": 2.1478584252372404e-05, + "loss": 0.1856, + "step": 44532 + }, + { + "epoch": 1.71, + "learning_rate": 2.1477943062323675e-05, + "loss": 0.3904, + "step": 44533 + }, + { + "epoch": 1.71, + "learning_rate": 2.1477301872274944e-05, + "loss": 0.3106, + "step": 44534 + }, + { + "epoch": 1.71, + "learning_rate": 2.1476660682226212e-05, + "loss": 0.3056, + "step": 44535 + }, + { + "epoch": 1.71, + "learning_rate": 2.1476019492177484e-05, + "loss": 0.2317, + "step": 44536 + }, + { + "epoch": 1.71, + "learning_rate": 2.1475378302128752e-05, + "loss": 0.4961, + "step": 44537 + }, + { + "epoch": 1.71, + "learning_rate": 2.147473711208002e-05, + "loss": 0.2595, + "step": 44538 + }, + { + "epoch": 1.71, + "learning_rate": 2.147409592203129e-05, + "loss": 0.4119, + "step": 44539 + }, + { + "epoch": 1.71, + "learning_rate": 2.147345473198256e-05, + "loss": 0.3443, + "step": 44540 + }, + { + "epoch": 1.71, + "learning_rate": 2.1472813541933832e-05, + "loss": 0.3482, + "step": 44541 + }, + { + "epoch": 1.71, + "learning_rate": 2.14721723518851e-05, + "loss": 0.307, + "step": 44542 + }, + { + "epoch": 1.71, + "learning_rate": 2.147153116183637e-05, + "loss": 0.2707, + "step": 44543 + }, + { + "epoch": 1.71, + "learning_rate": 2.1470889971787637e-05, + "loss": 0.3286, + "step": 44544 + }, + { + "epoch": 1.71, + "learning_rate": 2.147024878173891e-05, + "loss": 0.4352, + "step": 44545 + }, + { + "epoch": 1.71, + "learning_rate": 2.1469607591690177e-05, + "loss": 0.3803, + "step": 44546 + }, + { + "epoch": 1.71, + "learning_rate": 2.1468966401641446e-05, + "loss": 0.2683, + "step": 44547 + }, + { + "epoch": 1.71, + "learning_rate": 2.1468325211592717e-05, + "loss": 0.3992, + "step": 44548 + }, + { + "epoch": 1.71, + "learning_rate": 2.146768402154399e-05, + "loss": 0.4199, + "step": 44549 + }, + { + "epoch": 1.71, + "learning_rate": 2.1467042831495257e-05, + "loss": 0.2671, + "step": 44550 + }, + { + "epoch": 1.71, + "learning_rate": 2.1466401641446526e-05, + "loss": 0.3681, + "step": 44551 + }, + { + "epoch": 1.71, + "learning_rate": 2.1465760451397794e-05, + "loss": 0.4194, + "step": 44552 + }, + { + "epoch": 1.71, + "learning_rate": 2.1465119261349066e-05, + "loss": 0.2576, + "step": 44553 + }, + { + "epoch": 1.71, + "learning_rate": 2.1464478071300334e-05, + "loss": 0.1807, + "step": 44554 + }, + { + "epoch": 1.71, + "learning_rate": 2.1463836881251602e-05, + "loss": 0.2555, + "step": 44555 + }, + { + "epoch": 1.71, + "learning_rate": 2.1463195691202874e-05, + "loss": 0.2924, + "step": 44556 + }, + { + "epoch": 1.71, + "learning_rate": 2.1462554501154142e-05, + "loss": 0.3351, + "step": 44557 + }, + { + "epoch": 1.71, + "learning_rate": 2.1461913311105414e-05, + "loss": 0.3909, + "step": 44558 + }, + { + "epoch": 1.71, + "learning_rate": 2.1461272121056682e-05, + "loss": 0.1892, + "step": 44559 + }, + { + "epoch": 1.71, + "learning_rate": 2.146063093100795e-05, + "loss": 0.311, + "step": 44560 + }, + { + "epoch": 1.71, + "learning_rate": 2.1459989740959222e-05, + "loss": 0.3751, + "step": 44561 + }, + { + "epoch": 1.71, + "learning_rate": 2.145934855091049e-05, + "loss": 0.3957, + "step": 44562 + }, + { + "epoch": 1.71, + "learning_rate": 2.145870736086176e-05, + "loss": 0.3025, + "step": 44563 + }, + { + "epoch": 1.71, + "learning_rate": 2.145806617081303e-05, + "loss": 0.3545, + "step": 44564 + }, + { + "epoch": 1.71, + "learning_rate": 2.14574249807643e-05, + "loss": 0.2893, + "step": 44565 + }, + { + "epoch": 1.71, + "learning_rate": 2.1456783790715567e-05, + "loss": 0.4728, + "step": 44566 + }, + { + "epoch": 1.71, + "learning_rate": 2.145614260066684e-05, + "loss": 0.3947, + "step": 44567 + }, + { + "epoch": 1.71, + "learning_rate": 2.1455501410618108e-05, + "loss": 0.43, + "step": 44568 + }, + { + "epoch": 1.71, + "learning_rate": 2.145486022056938e-05, + "loss": 0.4528, + "step": 44569 + }, + { + "epoch": 1.71, + "learning_rate": 2.1454219030520648e-05, + "loss": 0.255, + "step": 44570 + }, + { + "epoch": 1.71, + "learning_rate": 2.1453577840471916e-05, + "loss": 0.2394, + "step": 44571 + }, + { + "epoch": 1.71, + "learning_rate": 2.1452936650423188e-05, + "loss": 0.4197, + "step": 44572 + }, + { + "epoch": 1.71, + "learning_rate": 2.1452295460374456e-05, + "loss": 0.3601, + "step": 44573 + }, + { + "epoch": 1.71, + "learning_rate": 2.1451654270325724e-05, + "loss": 0.4634, + "step": 44574 + }, + { + "epoch": 1.71, + "learning_rate": 2.1451013080276993e-05, + "loss": 0.316, + "step": 44575 + }, + { + "epoch": 1.71, + "learning_rate": 2.1450371890228264e-05, + "loss": 0.355, + "step": 44576 + }, + { + "epoch": 1.71, + "learning_rate": 2.1449730700179536e-05, + "loss": 0.2087, + "step": 44577 + }, + { + "epoch": 1.71, + "learning_rate": 2.1449089510130804e-05, + "loss": 0.3099, + "step": 44578 + }, + { + "epoch": 1.71, + "learning_rate": 2.1448448320082073e-05, + "loss": 0.394, + "step": 44579 + }, + { + "epoch": 1.71, + "learning_rate": 2.1447807130033344e-05, + "loss": 0.2957, + "step": 44580 + }, + { + "epoch": 1.71, + "learning_rate": 2.1447165939984613e-05, + "loss": 0.4223, + "step": 44581 + }, + { + "epoch": 1.71, + "learning_rate": 2.144652474993588e-05, + "loss": 0.3767, + "step": 44582 + }, + { + "epoch": 1.71, + "learning_rate": 2.144588355988715e-05, + "loss": 0.2519, + "step": 44583 + }, + { + "epoch": 1.71, + "learning_rate": 2.144524236983842e-05, + "loss": 0.3884, + "step": 44584 + }, + { + "epoch": 1.71, + "learning_rate": 2.1444601179789693e-05, + "loss": 0.3124, + "step": 44585 + }, + { + "epoch": 1.71, + "learning_rate": 2.144395998974096e-05, + "loss": 0.3553, + "step": 44586 + }, + { + "epoch": 1.71, + "learning_rate": 2.144331879969223e-05, + "loss": 0.2773, + "step": 44587 + }, + { + "epoch": 1.71, + "learning_rate": 2.1442677609643498e-05, + "loss": 0.3932, + "step": 44588 + }, + { + "epoch": 1.71, + "learning_rate": 2.144203641959477e-05, + "loss": 0.2988, + "step": 44589 + }, + { + "epoch": 1.71, + "learning_rate": 2.1441395229546038e-05, + "loss": 0.4391, + "step": 44590 + }, + { + "epoch": 1.71, + "learning_rate": 2.1440754039497306e-05, + "loss": 0.3284, + "step": 44591 + }, + { + "epoch": 1.71, + "learning_rate": 2.1440112849448578e-05, + "loss": 0.2803, + "step": 44592 + }, + { + "epoch": 1.71, + "learning_rate": 2.143947165939985e-05, + "loss": 0.3611, + "step": 44593 + }, + { + "epoch": 1.71, + "learning_rate": 2.1438830469351118e-05, + "loss": 0.3897, + "step": 44594 + }, + { + "epoch": 1.71, + "learning_rate": 2.1438189279302386e-05, + "loss": 0.2848, + "step": 44595 + }, + { + "epoch": 1.71, + "learning_rate": 2.1437548089253655e-05, + "loss": 0.2805, + "step": 44596 + }, + { + "epoch": 1.71, + "learning_rate": 2.1436906899204926e-05, + "loss": 0.4114, + "step": 44597 + }, + { + "epoch": 1.71, + "learning_rate": 2.1436265709156195e-05, + "loss": 0.4308, + "step": 44598 + }, + { + "epoch": 1.71, + "learning_rate": 2.1435624519107463e-05, + "loss": 0.3199, + "step": 44599 + }, + { + "epoch": 1.71, + "learning_rate": 2.1434983329058735e-05, + "loss": 0.368, + "step": 44600 + }, + { + "epoch": 1.71, + "learning_rate": 2.1434342139010003e-05, + "loss": 0.447, + "step": 44601 + }, + { + "epoch": 1.71, + "learning_rate": 2.1433700948961275e-05, + "loss": 0.3533, + "step": 44602 + }, + { + "epoch": 1.71, + "learning_rate": 2.1433059758912543e-05, + "loss": 0.2848, + "step": 44603 + }, + { + "epoch": 1.71, + "learning_rate": 2.143241856886381e-05, + "loss": 0.3985, + "step": 44604 + }, + { + "epoch": 1.71, + "learning_rate": 2.1431777378815083e-05, + "loss": 0.2534, + "step": 44605 + }, + { + "epoch": 1.71, + "learning_rate": 2.143113618876635e-05, + "loss": 0.254, + "step": 44606 + }, + { + "epoch": 1.71, + "learning_rate": 2.143049499871762e-05, + "loss": 0.2975, + "step": 44607 + }, + { + "epoch": 1.72, + "learning_rate": 2.142985380866889e-05, + "loss": 0.2785, + "step": 44608 + }, + { + "epoch": 1.72, + "learning_rate": 2.142921261862016e-05, + "loss": 0.3766, + "step": 44609 + }, + { + "epoch": 1.72, + "learning_rate": 2.1428571428571428e-05, + "loss": 0.3358, + "step": 44610 + }, + { + "epoch": 1.72, + "learning_rate": 2.14279302385227e-05, + "loss": 0.2093, + "step": 44611 + }, + { + "epoch": 1.72, + "learning_rate": 2.1427289048473968e-05, + "loss": 0.3696, + "step": 44612 + }, + { + "epoch": 1.72, + "learning_rate": 2.142664785842524e-05, + "loss": 0.2793, + "step": 44613 + }, + { + "epoch": 1.72, + "learning_rate": 2.1426006668376508e-05, + "loss": 0.3384, + "step": 44614 + }, + { + "epoch": 1.72, + "learning_rate": 2.1425365478327777e-05, + "loss": 0.256, + "step": 44615 + }, + { + "epoch": 1.72, + "learning_rate": 2.1424724288279048e-05, + "loss": 0.3257, + "step": 44616 + }, + { + "epoch": 1.72, + "learning_rate": 2.1424083098230317e-05, + "loss": 0.3294, + "step": 44617 + }, + { + "epoch": 1.72, + "learning_rate": 2.1423441908181585e-05, + "loss": 0.1931, + "step": 44618 + }, + { + "epoch": 1.72, + "learning_rate": 2.1422800718132853e-05, + "loss": 0.4543, + "step": 44619 + }, + { + "epoch": 1.72, + "learning_rate": 2.1422159528084125e-05, + "loss": 0.3492, + "step": 44620 + }, + { + "epoch": 1.72, + "learning_rate": 2.1421518338035397e-05, + "loss": 0.2632, + "step": 44621 + }, + { + "epoch": 1.72, + "learning_rate": 2.1420877147986665e-05, + "loss": 0.346, + "step": 44622 + }, + { + "epoch": 1.72, + "learning_rate": 2.1420235957937933e-05, + "loss": 0.4982, + "step": 44623 + }, + { + "epoch": 1.72, + "learning_rate": 2.1419594767889205e-05, + "loss": 0.2646, + "step": 44624 + }, + { + "epoch": 1.72, + "learning_rate": 2.1418953577840473e-05, + "loss": 0.2838, + "step": 44625 + }, + { + "epoch": 1.72, + "learning_rate": 2.1418312387791742e-05, + "loss": 0.3276, + "step": 44626 + }, + { + "epoch": 1.72, + "learning_rate": 2.141767119774301e-05, + "loss": 0.4894, + "step": 44627 + }, + { + "epoch": 1.72, + "learning_rate": 2.1417030007694282e-05, + "loss": 0.5311, + "step": 44628 + }, + { + "epoch": 1.72, + "learning_rate": 2.1416388817645553e-05, + "loss": 0.3402, + "step": 44629 + }, + { + "epoch": 1.72, + "learning_rate": 2.1415747627596822e-05, + "loss": 0.2721, + "step": 44630 + }, + { + "epoch": 1.72, + "learning_rate": 2.141510643754809e-05, + "loss": 0.3039, + "step": 44631 + }, + { + "epoch": 1.72, + "learning_rate": 2.141446524749936e-05, + "loss": 0.3593, + "step": 44632 + }, + { + "epoch": 1.72, + "learning_rate": 2.141382405745063e-05, + "loss": 0.3948, + "step": 44633 + }, + { + "epoch": 1.72, + "learning_rate": 2.14131828674019e-05, + "loss": 0.278, + "step": 44634 + }, + { + "epoch": 1.72, + "learning_rate": 2.1412541677353167e-05, + "loss": 0.4833, + "step": 44635 + }, + { + "epoch": 1.72, + "learning_rate": 2.141190048730444e-05, + "loss": 0.5636, + "step": 44636 + }, + { + "epoch": 1.72, + "learning_rate": 2.1411259297255707e-05, + "loss": 0.3884, + "step": 44637 + }, + { + "epoch": 1.72, + "learning_rate": 2.141061810720698e-05, + "loss": 0.3854, + "step": 44638 + }, + { + "epoch": 1.72, + "learning_rate": 2.1409976917158247e-05, + "loss": 0.3986, + "step": 44639 + }, + { + "epoch": 1.72, + "learning_rate": 2.1409335727109515e-05, + "loss": 0.3055, + "step": 44640 + }, + { + "epoch": 1.72, + "learning_rate": 2.1408694537060787e-05, + "loss": 0.4021, + "step": 44641 + }, + { + "epoch": 1.72, + "learning_rate": 2.1408053347012055e-05, + "loss": 0.3862, + "step": 44642 + }, + { + "epoch": 1.72, + "learning_rate": 2.1407412156963324e-05, + "loss": 0.4526, + "step": 44643 + }, + { + "epoch": 1.72, + "learning_rate": 2.1406770966914595e-05, + "loss": 0.4946, + "step": 44644 + }, + { + "epoch": 1.72, + "learning_rate": 2.1406129776865864e-05, + "loss": 0.31, + "step": 44645 + }, + { + "epoch": 1.72, + "learning_rate": 2.1405488586817132e-05, + "loss": 0.2361, + "step": 44646 + }, + { + "epoch": 1.72, + "learning_rate": 2.1404847396768404e-05, + "loss": 0.3224, + "step": 44647 + }, + { + "epoch": 1.72, + "learning_rate": 2.1404206206719672e-05, + "loss": 0.4628, + "step": 44648 + }, + { + "epoch": 1.72, + "learning_rate": 2.1403565016670944e-05, + "loss": 0.3045, + "step": 44649 + }, + { + "epoch": 1.72, + "learning_rate": 2.1402923826622212e-05, + "loss": 0.3603, + "step": 44650 + }, + { + "epoch": 1.72, + "learning_rate": 2.140228263657348e-05, + "loss": 0.2184, + "step": 44651 + }, + { + "epoch": 1.72, + "learning_rate": 2.1401641446524752e-05, + "loss": 0.316, + "step": 44652 + }, + { + "epoch": 1.72, + "learning_rate": 2.140100025647602e-05, + "loss": 0.3804, + "step": 44653 + }, + { + "epoch": 1.72, + "learning_rate": 2.140035906642729e-05, + "loss": 0.4533, + "step": 44654 + }, + { + "epoch": 1.72, + "learning_rate": 2.1399717876378557e-05, + "loss": 0.4192, + "step": 44655 + }, + { + "epoch": 1.72, + "learning_rate": 2.139907668632983e-05, + "loss": 0.3543, + "step": 44656 + }, + { + "epoch": 1.72, + "learning_rate": 2.13984354962811e-05, + "loss": 0.2329, + "step": 44657 + }, + { + "epoch": 1.72, + "learning_rate": 2.139779430623237e-05, + "loss": 0.3809, + "step": 44658 + }, + { + "epoch": 1.72, + "learning_rate": 2.1397153116183637e-05, + "loss": 0.3, + "step": 44659 + }, + { + "epoch": 1.72, + "learning_rate": 2.139651192613491e-05, + "loss": 0.3594, + "step": 44660 + }, + { + "epoch": 1.72, + "learning_rate": 2.1395870736086177e-05, + "loss": 0.2699, + "step": 44661 + }, + { + "epoch": 1.72, + "learning_rate": 2.1395229546037446e-05, + "loss": 0.3804, + "step": 44662 + }, + { + "epoch": 1.72, + "learning_rate": 2.1394588355988714e-05, + "loss": 0.2726, + "step": 44663 + }, + { + "epoch": 1.72, + "learning_rate": 2.1393947165939986e-05, + "loss": 0.2826, + "step": 44664 + }, + { + "epoch": 1.72, + "learning_rate": 2.1393305975891257e-05, + "loss": 0.2915, + "step": 44665 + }, + { + "epoch": 1.72, + "learning_rate": 2.1392664785842526e-05, + "loss": 0.2881, + "step": 44666 + }, + { + "epoch": 1.72, + "learning_rate": 2.1392023595793794e-05, + "loss": 0.3367, + "step": 44667 + }, + { + "epoch": 1.72, + "learning_rate": 2.1391382405745062e-05, + "loss": 0.3597, + "step": 44668 + }, + { + "epoch": 1.72, + "learning_rate": 2.1390741215696334e-05, + "loss": 0.2612, + "step": 44669 + }, + { + "epoch": 1.72, + "learning_rate": 2.1390100025647602e-05, + "loss": 0.32, + "step": 44670 + }, + { + "epoch": 1.72, + "learning_rate": 2.138945883559887e-05, + "loss": 0.3033, + "step": 44671 + }, + { + "epoch": 1.72, + "learning_rate": 2.1388817645550142e-05, + "loss": 0.2455, + "step": 44672 + }, + { + "epoch": 1.72, + "learning_rate": 2.1388176455501414e-05, + "loss": 0.3583, + "step": 44673 + }, + { + "epoch": 1.72, + "learning_rate": 2.1387535265452682e-05, + "loss": 0.3648, + "step": 44674 + }, + { + "epoch": 1.72, + "learning_rate": 2.138689407540395e-05, + "loss": 0.3789, + "step": 44675 + }, + { + "epoch": 1.72, + "learning_rate": 2.138625288535522e-05, + "loss": 0.2971, + "step": 44676 + }, + { + "epoch": 1.72, + "learning_rate": 2.138561169530649e-05, + "loss": 0.2914, + "step": 44677 + }, + { + "epoch": 1.72, + "learning_rate": 2.138497050525776e-05, + "loss": 0.2794, + "step": 44678 + }, + { + "epoch": 1.72, + "learning_rate": 2.1384329315209028e-05, + "loss": 0.3558, + "step": 44679 + }, + { + "epoch": 1.72, + "learning_rate": 2.13836881251603e-05, + "loss": 0.5103, + "step": 44680 + }, + { + "epoch": 1.72, + "learning_rate": 2.1383046935111568e-05, + "loss": 0.3157, + "step": 44681 + }, + { + "epoch": 1.72, + "learning_rate": 2.138240574506284e-05, + "loss": 0.3352, + "step": 44682 + }, + { + "epoch": 1.72, + "learning_rate": 2.1381764555014108e-05, + "loss": 0.3452, + "step": 44683 + }, + { + "epoch": 1.72, + "learning_rate": 2.1381123364965376e-05, + "loss": 0.4297, + "step": 44684 + }, + { + "epoch": 1.72, + "learning_rate": 2.1380482174916648e-05, + "loss": 0.2911, + "step": 44685 + }, + { + "epoch": 1.72, + "learning_rate": 2.1379840984867916e-05, + "loss": 0.2163, + "step": 44686 + }, + { + "epoch": 1.72, + "learning_rate": 2.1379199794819184e-05, + "loss": 0.3183, + "step": 44687 + }, + { + "epoch": 1.72, + "learning_rate": 2.1378558604770456e-05, + "loss": 0.3286, + "step": 44688 + }, + { + "epoch": 1.72, + "learning_rate": 2.1377917414721724e-05, + "loss": 0.3137, + "step": 44689 + }, + { + "epoch": 1.72, + "learning_rate": 2.1377276224672993e-05, + "loss": 0.3224, + "step": 44690 + }, + { + "epoch": 1.72, + "learning_rate": 2.1376635034624264e-05, + "loss": 0.1904, + "step": 44691 + }, + { + "epoch": 1.72, + "learning_rate": 2.1375993844575533e-05, + "loss": 0.3326, + "step": 44692 + }, + { + "epoch": 1.72, + "learning_rate": 2.1375352654526804e-05, + "loss": 0.2718, + "step": 44693 + }, + { + "epoch": 1.72, + "learning_rate": 2.1374711464478073e-05, + "loss": 0.3053, + "step": 44694 + }, + { + "epoch": 1.72, + "learning_rate": 2.137407027442934e-05, + "loss": 0.3495, + "step": 44695 + }, + { + "epoch": 1.72, + "learning_rate": 2.1373429084380613e-05, + "loss": 0.4408, + "step": 44696 + }, + { + "epoch": 1.72, + "learning_rate": 2.137278789433188e-05, + "loss": 0.3452, + "step": 44697 + }, + { + "epoch": 1.72, + "learning_rate": 2.137214670428315e-05, + "loss": 0.3941, + "step": 44698 + }, + { + "epoch": 1.72, + "learning_rate": 2.1371505514234418e-05, + "loss": 0.258, + "step": 44699 + }, + { + "epoch": 1.72, + "learning_rate": 2.137086432418569e-05, + "loss": 0.2974, + "step": 44700 + }, + { + "epoch": 1.72, + "learning_rate": 2.137022313413696e-05, + "loss": 0.2162, + "step": 44701 + }, + { + "epoch": 1.72, + "learning_rate": 2.136958194408823e-05, + "loss": 0.3381, + "step": 44702 + }, + { + "epoch": 1.72, + "learning_rate": 2.1368940754039498e-05, + "loss": 0.2994, + "step": 44703 + }, + { + "epoch": 1.72, + "learning_rate": 2.136829956399077e-05, + "loss": 0.3259, + "step": 44704 + }, + { + "epoch": 1.72, + "learning_rate": 2.1367658373942038e-05, + "loss": 0.4392, + "step": 44705 + }, + { + "epoch": 1.72, + "learning_rate": 2.1367017183893306e-05, + "loss": 0.2852, + "step": 44706 + }, + { + "epoch": 1.72, + "learning_rate": 2.1366375993844575e-05, + "loss": 0.3314, + "step": 44707 + }, + { + "epoch": 1.72, + "learning_rate": 2.1365734803795846e-05, + "loss": 0.3445, + "step": 44708 + }, + { + "epoch": 1.72, + "learning_rate": 2.1365093613747118e-05, + "loss": 0.3377, + "step": 44709 + }, + { + "epoch": 1.72, + "learning_rate": 2.1364452423698386e-05, + "loss": 0.4288, + "step": 44710 + }, + { + "epoch": 1.72, + "learning_rate": 2.1363811233649655e-05, + "loss": 0.2137, + "step": 44711 + }, + { + "epoch": 1.72, + "learning_rate": 2.1363170043600923e-05, + "loss": 0.2392, + "step": 44712 + }, + { + "epoch": 1.72, + "learning_rate": 2.1362528853552195e-05, + "loss": 0.323, + "step": 44713 + }, + { + "epoch": 1.72, + "learning_rate": 2.1361887663503463e-05, + "loss": 0.357, + "step": 44714 + }, + { + "epoch": 1.72, + "learning_rate": 2.136124647345473e-05, + "loss": 0.3643, + "step": 44715 + }, + { + "epoch": 1.72, + "learning_rate": 2.1360605283406003e-05, + "loss": 0.3236, + "step": 44716 + }, + { + "epoch": 1.72, + "learning_rate": 2.135996409335727e-05, + "loss": 0.359, + "step": 44717 + }, + { + "epoch": 1.72, + "learning_rate": 2.1359322903308543e-05, + "loss": 0.385, + "step": 44718 + }, + { + "epoch": 1.72, + "learning_rate": 2.135868171325981e-05, + "loss": 0.2889, + "step": 44719 + }, + { + "epoch": 1.72, + "learning_rate": 2.135804052321108e-05, + "loss": 0.2079, + "step": 44720 + }, + { + "epoch": 1.72, + "learning_rate": 2.135739933316235e-05, + "loss": 0.3868, + "step": 44721 + }, + { + "epoch": 1.72, + "learning_rate": 2.135675814311362e-05, + "loss": 0.396, + "step": 44722 + }, + { + "epoch": 1.72, + "learning_rate": 2.1356116953064888e-05, + "loss": 0.2809, + "step": 44723 + }, + { + "epoch": 1.72, + "learning_rate": 2.135547576301616e-05, + "loss": 0.228, + "step": 44724 + }, + { + "epoch": 1.72, + "learning_rate": 2.1354834572967428e-05, + "loss": 0.1998, + "step": 44725 + }, + { + "epoch": 1.72, + "learning_rate": 2.1354193382918697e-05, + "loss": 0.6081, + "step": 44726 + }, + { + "epoch": 1.72, + "learning_rate": 2.1353552192869968e-05, + "loss": 0.2974, + "step": 44727 + }, + { + "epoch": 1.72, + "learning_rate": 2.1352911002821237e-05, + "loss": 0.3519, + "step": 44728 + }, + { + "epoch": 1.72, + "learning_rate": 2.135226981277251e-05, + "loss": 0.248, + "step": 44729 + }, + { + "epoch": 1.72, + "learning_rate": 2.1351628622723777e-05, + "loss": 0.3244, + "step": 44730 + }, + { + "epoch": 1.72, + "learning_rate": 2.1350987432675045e-05, + "loss": 0.3017, + "step": 44731 + }, + { + "epoch": 1.72, + "learning_rate": 2.1350346242626317e-05, + "loss": 0.3851, + "step": 44732 + }, + { + "epoch": 1.72, + "learning_rate": 2.1349705052577585e-05, + "loss": 0.4423, + "step": 44733 + }, + { + "epoch": 1.72, + "learning_rate": 2.1349063862528853e-05, + "loss": 0.362, + "step": 44734 + }, + { + "epoch": 1.72, + "learning_rate": 2.134842267248012e-05, + "loss": 0.327, + "step": 44735 + }, + { + "epoch": 1.72, + "learning_rate": 2.1347781482431393e-05, + "loss": 0.3186, + "step": 44736 + }, + { + "epoch": 1.72, + "learning_rate": 2.1347140292382665e-05, + "loss": 0.3756, + "step": 44737 + }, + { + "epoch": 1.72, + "learning_rate": 2.1346499102333933e-05, + "loss": 0.4175, + "step": 44738 + }, + { + "epoch": 1.72, + "learning_rate": 2.1345857912285202e-05, + "loss": 0.2925, + "step": 44739 + }, + { + "epoch": 1.72, + "learning_rate": 2.1345216722236473e-05, + "loss": 0.3922, + "step": 44740 + }, + { + "epoch": 1.72, + "learning_rate": 2.1344575532187742e-05, + "loss": 0.2085, + "step": 44741 + }, + { + "epoch": 1.72, + "learning_rate": 2.134393434213901e-05, + "loss": 0.2436, + "step": 44742 + }, + { + "epoch": 1.72, + "learning_rate": 2.134329315209028e-05, + "loss": 0.2749, + "step": 44743 + }, + { + "epoch": 1.72, + "learning_rate": 2.134265196204155e-05, + "loss": 0.3327, + "step": 44744 + }, + { + "epoch": 1.72, + "learning_rate": 2.1342010771992822e-05, + "loss": 0.2732, + "step": 44745 + }, + { + "epoch": 1.72, + "learning_rate": 2.134136958194409e-05, + "loss": 0.2849, + "step": 44746 + }, + { + "epoch": 1.72, + "learning_rate": 2.134072839189536e-05, + "loss": 0.2686, + "step": 44747 + }, + { + "epoch": 1.72, + "learning_rate": 2.1340087201846627e-05, + "loss": 0.3935, + "step": 44748 + }, + { + "epoch": 1.72, + "learning_rate": 2.13394460117979e-05, + "loss": 0.3555, + "step": 44749 + }, + { + "epoch": 1.72, + "learning_rate": 2.1338804821749167e-05, + "loss": 0.3529, + "step": 44750 + }, + { + "epoch": 1.72, + "learning_rate": 2.1338163631700435e-05, + "loss": 0.2927, + "step": 44751 + }, + { + "epoch": 1.72, + "learning_rate": 2.1337522441651707e-05, + "loss": 0.4087, + "step": 44752 + }, + { + "epoch": 1.72, + "learning_rate": 2.1336881251602975e-05, + "loss": 0.3981, + "step": 44753 + }, + { + "epoch": 1.72, + "learning_rate": 2.1336240061554247e-05, + "loss": 0.3397, + "step": 44754 + }, + { + "epoch": 1.72, + "learning_rate": 2.1335598871505515e-05, + "loss": 0.4114, + "step": 44755 + }, + { + "epoch": 1.72, + "learning_rate": 2.1334957681456784e-05, + "loss": 0.2607, + "step": 44756 + }, + { + "epoch": 1.72, + "learning_rate": 2.1334316491408055e-05, + "loss": 0.3041, + "step": 44757 + }, + { + "epoch": 1.72, + "learning_rate": 2.1333675301359324e-05, + "loss": 0.2392, + "step": 44758 + }, + { + "epoch": 1.72, + "learning_rate": 2.1333034111310592e-05, + "loss": 0.2547, + "step": 44759 + }, + { + "epoch": 1.72, + "learning_rate": 2.1332392921261864e-05, + "loss": 0.2676, + "step": 44760 + }, + { + "epoch": 1.72, + "learning_rate": 2.1331751731213132e-05, + "loss": 0.3289, + "step": 44761 + }, + { + "epoch": 1.72, + "learning_rate": 2.13311105411644e-05, + "loss": 0.3118, + "step": 44762 + }, + { + "epoch": 1.72, + "learning_rate": 2.1330469351115672e-05, + "loss": 0.3798, + "step": 44763 + }, + { + "epoch": 1.72, + "learning_rate": 2.132982816106694e-05, + "loss": 0.369, + "step": 44764 + }, + { + "epoch": 1.72, + "learning_rate": 2.1329186971018212e-05, + "loss": 0.3897, + "step": 44765 + }, + { + "epoch": 1.72, + "learning_rate": 2.132854578096948e-05, + "loss": 0.4021, + "step": 44766 + }, + { + "epoch": 1.72, + "learning_rate": 2.132790459092075e-05, + "loss": 0.3779, + "step": 44767 + }, + { + "epoch": 1.72, + "learning_rate": 2.132726340087202e-05, + "loss": 0.3684, + "step": 44768 + }, + { + "epoch": 1.72, + "learning_rate": 2.132662221082329e-05, + "loss": 0.3035, + "step": 44769 + }, + { + "epoch": 1.72, + "learning_rate": 2.1325981020774557e-05, + "loss": 0.4151, + "step": 44770 + }, + { + "epoch": 1.72, + "learning_rate": 2.132533983072583e-05, + "loss": 0.4331, + "step": 44771 + }, + { + "epoch": 1.72, + "learning_rate": 2.1324698640677097e-05, + "loss": 0.2979, + "step": 44772 + }, + { + "epoch": 1.72, + "learning_rate": 2.132405745062837e-05, + "loss": 0.3003, + "step": 44773 + }, + { + "epoch": 1.72, + "learning_rate": 2.1323416260579637e-05, + "loss": 0.5828, + "step": 44774 + }, + { + "epoch": 1.72, + "learning_rate": 2.1322775070530906e-05, + "loss": 0.3067, + "step": 44775 + }, + { + "epoch": 1.72, + "learning_rate": 2.1322133880482177e-05, + "loss": 0.3128, + "step": 44776 + }, + { + "epoch": 1.72, + "learning_rate": 2.1321492690433446e-05, + "loss": 0.3867, + "step": 44777 + }, + { + "epoch": 1.72, + "learning_rate": 2.1320851500384714e-05, + "loss": 0.4933, + "step": 44778 + }, + { + "epoch": 1.72, + "learning_rate": 2.1320210310335982e-05, + "loss": 0.3431, + "step": 44779 + }, + { + "epoch": 1.72, + "learning_rate": 2.1319569120287254e-05, + "loss": 0.3417, + "step": 44780 + }, + { + "epoch": 1.72, + "learning_rate": 2.1318927930238526e-05, + "loss": 0.3209, + "step": 44781 + }, + { + "epoch": 1.72, + "learning_rate": 2.1318286740189794e-05, + "loss": 0.5264, + "step": 44782 + }, + { + "epoch": 1.72, + "learning_rate": 2.1317645550141062e-05, + "loss": 0.3485, + "step": 44783 + }, + { + "epoch": 1.72, + "learning_rate": 2.1317004360092334e-05, + "loss": 0.4984, + "step": 44784 + }, + { + "epoch": 1.72, + "learning_rate": 2.1316363170043602e-05, + "loss": 0.2987, + "step": 44785 + }, + { + "epoch": 1.72, + "learning_rate": 2.131572197999487e-05, + "loss": 0.3052, + "step": 44786 + }, + { + "epoch": 1.72, + "learning_rate": 2.131508078994614e-05, + "loss": 0.3797, + "step": 44787 + }, + { + "epoch": 1.72, + "learning_rate": 2.131443959989741e-05, + "loss": 0.2422, + "step": 44788 + }, + { + "epoch": 1.72, + "learning_rate": 2.1313798409848683e-05, + "loss": 0.3001, + "step": 44789 + }, + { + "epoch": 1.72, + "learning_rate": 2.131315721979995e-05, + "loss": 0.3724, + "step": 44790 + }, + { + "epoch": 1.72, + "learning_rate": 2.131251602975122e-05, + "loss": 0.3396, + "step": 44791 + }, + { + "epoch": 1.72, + "learning_rate": 2.1311874839702488e-05, + "loss": 0.4417, + "step": 44792 + }, + { + "epoch": 1.72, + "learning_rate": 2.131123364965376e-05, + "loss": 0.3516, + "step": 44793 + }, + { + "epoch": 1.72, + "learning_rate": 2.1310592459605028e-05, + "loss": 0.3406, + "step": 44794 + }, + { + "epoch": 1.72, + "learning_rate": 2.1309951269556296e-05, + "loss": 0.2374, + "step": 44795 + }, + { + "epoch": 1.72, + "learning_rate": 2.1309310079507568e-05, + "loss": 0.4475, + "step": 44796 + }, + { + "epoch": 1.72, + "learning_rate": 2.1308668889458836e-05, + "loss": 0.3741, + "step": 44797 + }, + { + "epoch": 1.72, + "learning_rate": 2.1308027699410108e-05, + "loss": 0.3898, + "step": 44798 + }, + { + "epoch": 1.72, + "learning_rate": 2.1307386509361376e-05, + "loss": 0.3244, + "step": 44799 + }, + { + "epoch": 1.72, + "learning_rate": 2.1306745319312644e-05, + "loss": 0.449, + "step": 44800 + }, + { + "epoch": 1.72, + "learning_rate": 2.1306104129263916e-05, + "loss": 0.3843, + "step": 44801 + }, + { + "epoch": 1.72, + "learning_rate": 2.1305462939215184e-05, + "loss": 0.4174, + "step": 44802 + }, + { + "epoch": 1.72, + "learning_rate": 2.1304821749166453e-05, + "loss": 0.2975, + "step": 44803 + }, + { + "epoch": 1.72, + "learning_rate": 2.1304180559117724e-05, + "loss": 0.4422, + "step": 44804 + }, + { + "epoch": 1.72, + "learning_rate": 2.1303539369068993e-05, + "loss": 0.3221, + "step": 44805 + }, + { + "epoch": 1.72, + "learning_rate": 2.130289817902026e-05, + "loss": 0.4154, + "step": 44806 + }, + { + "epoch": 1.72, + "learning_rate": 2.1302256988971533e-05, + "loss": 0.2355, + "step": 44807 + }, + { + "epoch": 1.72, + "learning_rate": 2.13016157989228e-05, + "loss": 0.3175, + "step": 44808 + }, + { + "epoch": 1.72, + "learning_rate": 2.1300974608874073e-05, + "loss": 0.3837, + "step": 44809 + }, + { + "epoch": 1.72, + "learning_rate": 2.130033341882534e-05, + "loss": 0.2778, + "step": 44810 + }, + { + "epoch": 1.72, + "learning_rate": 2.129969222877661e-05, + "loss": 0.3425, + "step": 44811 + }, + { + "epoch": 1.72, + "learning_rate": 2.129905103872788e-05, + "loss": 0.3345, + "step": 44812 + }, + { + "epoch": 1.72, + "learning_rate": 2.129840984867915e-05, + "loss": 0.3788, + "step": 44813 + }, + { + "epoch": 1.72, + "learning_rate": 2.1297768658630418e-05, + "loss": 0.2581, + "step": 44814 + }, + { + "epoch": 1.72, + "learning_rate": 2.1297127468581686e-05, + "loss": 0.2961, + "step": 44815 + }, + { + "epoch": 1.72, + "learning_rate": 2.1296486278532958e-05, + "loss": 0.275, + "step": 44816 + }, + { + "epoch": 1.72, + "learning_rate": 2.129584508848423e-05, + "loss": 0.3112, + "step": 44817 + }, + { + "epoch": 1.72, + "learning_rate": 2.1295203898435498e-05, + "loss": 0.3762, + "step": 44818 + }, + { + "epoch": 1.72, + "learning_rate": 2.1294562708386766e-05, + "loss": 0.3473, + "step": 44819 + }, + { + "epoch": 1.72, + "learning_rate": 2.1293921518338038e-05, + "loss": 0.3224, + "step": 44820 + }, + { + "epoch": 1.72, + "learning_rate": 2.1293280328289306e-05, + "loss": 0.3125, + "step": 44821 + }, + { + "epoch": 1.72, + "learning_rate": 2.1292639138240575e-05, + "loss": 0.4191, + "step": 44822 + }, + { + "epoch": 1.72, + "learning_rate": 2.1291997948191843e-05, + "loss": 0.3679, + "step": 44823 + }, + { + "epoch": 1.72, + "learning_rate": 2.1291356758143115e-05, + "loss": 0.4046, + "step": 44824 + }, + { + "epoch": 1.72, + "learning_rate": 2.1290715568094386e-05, + "loss": 0.2635, + "step": 44825 + }, + { + "epoch": 1.72, + "learning_rate": 2.1290074378045655e-05, + "loss": 0.2914, + "step": 44826 + }, + { + "epoch": 1.72, + "learning_rate": 2.1289433187996923e-05, + "loss": 0.2573, + "step": 44827 + }, + { + "epoch": 1.72, + "learning_rate": 2.1288791997948195e-05, + "loss": 0.3226, + "step": 44828 + }, + { + "epoch": 1.72, + "learning_rate": 2.1288150807899463e-05, + "loss": 0.328, + "step": 44829 + }, + { + "epoch": 1.72, + "learning_rate": 2.128750961785073e-05, + "loss": 0.2889, + "step": 44830 + }, + { + "epoch": 1.72, + "learning_rate": 2.1286868427802e-05, + "loss": 0.4279, + "step": 44831 + }, + { + "epoch": 1.72, + "learning_rate": 2.128622723775327e-05, + "loss": 0.3483, + "step": 44832 + }, + { + "epoch": 1.72, + "learning_rate": 2.128558604770454e-05, + "loss": 0.3091, + "step": 44833 + }, + { + "epoch": 1.72, + "learning_rate": 2.128494485765581e-05, + "loss": 0.3151, + "step": 44834 + }, + { + "epoch": 1.72, + "learning_rate": 2.128430366760708e-05, + "loss": 0.4035, + "step": 44835 + }, + { + "epoch": 1.72, + "learning_rate": 2.1283662477558348e-05, + "loss": 0.3107, + "step": 44836 + }, + { + "epoch": 1.72, + "learning_rate": 2.128302128750962e-05, + "loss": 0.2622, + "step": 44837 + }, + { + "epoch": 1.72, + "learning_rate": 2.1282380097460888e-05, + "loss": 0.383, + "step": 44838 + }, + { + "epoch": 1.72, + "learning_rate": 2.1281738907412157e-05, + "loss": 0.4636, + "step": 44839 + }, + { + "epoch": 1.72, + "learning_rate": 2.1281097717363428e-05, + "loss": 0.2209, + "step": 44840 + }, + { + "epoch": 1.72, + "learning_rate": 2.1280456527314697e-05, + "loss": 0.3946, + "step": 44841 + }, + { + "epoch": 1.72, + "learning_rate": 2.1279815337265965e-05, + "loss": 0.5016, + "step": 44842 + }, + { + "epoch": 1.72, + "learning_rate": 2.1279174147217237e-05, + "loss": 0.3617, + "step": 44843 + }, + { + "epoch": 1.72, + "learning_rate": 2.1278532957168505e-05, + "loss": 0.389, + "step": 44844 + }, + { + "epoch": 1.72, + "learning_rate": 2.1277891767119777e-05, + "loss": 0.3071, + "step": 44845 + }, + { + "epoch": 1.72, + "learning_rate": 2.1277250577071045e-05, + "loss": 0.3942, + "step": 44846 + }, + { + "epoch": 1.72, + "learning_rate": 2.1276609387022313e-05, + "loss": 0.3065, + "step": 44847 + }, + { + "epoch": 1.72, + "learning_rate": 2.1275968196973585e-05, + "loss": 0.4685, + "step": 44848 + }, + { + "epoch": 1.72, + "learning_rate": 2.1275327006924853e-05, + "loss": 0.4039, + "step": 44849 + }, + { + "epoch": 1.72, + "learning_rate": 2.1274685816876122e-05, + "loss": 0.373, + "step": 44850 + }, + { + "epoch": 1.72, + "learning_rate": 2.127404462682739e-05, + "loss": 0.2785, + "step": 44851 + }, + { + "epoch": 1.72, + "learning_rate": 2.1273403436778662e-05, + "loss": 0.3041, + "step": 44852 + }, + { + "epoch": 1.72, + "learning_rate": 2.1272762246729934e-05, + "loss": 0.2261, + "step": 44853 + }, + { + "epoch": 1.72, + "learning_rate": 2.1272121056681202e-05, + "loss": 0.4248, + "step": 44854 + }, + { + "epoch": 1.72, + "learning_rate": 2.127147986663247e-05, + "loss": 0.361, + "step": 44855 + }, + { + "epoch": 1.72, + "learning_rate": 2.1270838676583742e-05, + "loss": 0.2346, + "step": 44856 + }, + { + "epoch": 1.72, + "learning_rate": 2.127019748653501e-05, + "loss": 0.3448, + "step": 44857 + }, + { + "epoch": 1.72, + "learning_rate": 2.126955629648628e-05, + "loss": 0.3677, + "step": 44858 + }, + { + "epoch": 1.72, + "learning_rate": 2.1268915106437547e-05, + "loss": 0.2917, + "step": 44859 + }, + { + "epoch": 1.72, + "learning_rate": 2.126827391638882e-05, + "loss": 0.2622, + "step": 44860 + }, + { + "epoch": 1.72, + "learning_rate": 2.126763272634009e-05, + "loss": 0.4682, + "step": 44861 + }, + { + "epoch": 1.72, + "learning_rate": 2.126699153629136e-05, + "loss": 0.2399, + "step": 44862 + }, + { + "epoch": 1.72, + "learning_rate": 2.1266350346242627e-05, + "loss": 0.2316, + "step": 44863 + }, + { + "epoch": 1.72, + "learning_rate": 2.12657091561939e-05, + "loss": 0.4261, + "step": 44864 + }, + { + "epoch": 1.72, + "learning_rate": 2.1265067966145167e-05, + "loss": 0.2564, + "step": 44865 + }, + { + "epoch": 1.72, + "learning_rate": 2.1264426776096435e-05, + "loss": 0.3627, + "step": 44866 + }, + { + "epoch": 1.72, + "learning_rate": 2.1263785586047704e-05, + "loss": 0.3199, + "step": 44867 + }, + { + "epoch": 1.73, + "learning_rate": 2.1263144395998975e-05, + "loss": 0.3978, + "step": 44868 + }, + { + "epoch": 1.73, + "learning_rate": 2.1262503205950247e-05, + "loss": 0.3814, + "step": 44869 + }, + { + "epoch": 1.73, + "learning_rate": 2.1261862015901515e-05, + "loss": 0.3217, + "step": 44870 + }, + { + "epoch": 1.73, + "learning_rate": 2.1261220825852784e-05, + "loss": 0.3858, + "step": 44871 + }, + { + "epoch": 1.73, + "learning_rate": 2.1260579635804052e-05, + "loss": 0.3017, + "step": 44872 + }, + { + "epoch": 1.73, + "learning_rate": 2.1259938445755324e-05, + "loss": 0.3019, + "step": 44873 + }, + { + "epoch": 1.73, + "learning_rate": 2.1259297255706592e-05, + "loss": 0.2667, + "step": 44874 + }, + { + "epoch": 1.73, + "learning_rate": 2.125865606565786e-05, + "loss": 0.364, + "step": 44875 + }, + { + "epoch": 1.73, + "learning_rate": 2.1258014875609132e-05, + "loss": 0.2793, + "step": 44876 + }, + { + "epoch": 1.73, + "learning_rate": 2.12573736855604e-05, + "loss": 0.3408, + "step": 44877 + }, + { + "epoch": 1.73, + "learning_rate": 2.1256732495511672e-05, + "loss": 0.3709, + "step": 44878 + }, + { + "epoch": 1.73, + "learning_rate": 2.125609130546294e-05, + "loss": 0.3085, + "step": 44879 + }, + { + "epoch": 1.73, + "learning_rate": 2.125545011541421e-05, + "loss": 0.3975, + "step": 44880 + }, + { + "epoch": 1.73, + "learning_rate": 2.125480892536548e-05, + "loss": 0.3233, + "step": 44881 + }, + { + "epoch": 1.73, + "learning_rate": 2.125416773531675e-05, + "loss": 0.3354, + "step": 44882 + }, + { + "epoch": 1.73, + "learning_rate": 2.1253526545268017e-05, + "loss": 0.4138, + "step": 44883 + }, + { + "epoch": 1.73, + "learning_rate": 2.125288535521929e-05, + "loss": 0.2424, + "step": 44884 + }, + { + "epoch": 1.73, + "learning_rate": 2.1252244165170557e-05, + "loss": 0.3874, + "step": 44885 + }, + { + "epoch": 1.73, + "learning_rate": 2.1251602975121826e-05, + "loss": 0.3537, + "step": 44886 + }, + { + "epoch": 1.73, + "learning_rate": 2.1250961785073097e-05, + "loss": 0.4652, + "step": 44887 + }, + { + "epoch": 1.73, + "learning_rate": 2.1250320595024366e-05, + "loss": 0.306, + "step": 44888 + }, + { + "epoch": 1.73, + "learning_rate": 2.1249679404975637e-05, + "loss": 0.3379, + "step": 44889 + }, + { + "epoch": 1.73, + "learning_rate": 2.1249038214926906e-05, + "loss": 0.4496, + "step": 44890 + }, + { + "epoch": 1.73, + "learning_rate": 2.1248397024878174e-05, + "loss": 0.3, + "step": 44891 + }, + { + "epoch": 1.73, + "learning_rate": 2.1247755834829446e-05, + "loss": 0.4761, + "step": 44892 + }, + { + "epoch": 1.73, + "learning_rate": 2.1247114644780714e-05, + "loss": 0.2673, + "step": 44893 + }, + { + "epoch": 1.73, + "learning_rate": 2.1246473454731982e-05, + "loss": 0.2957, + "step": 44894 + }, + { + "epoch": 1.73, + "learning_rate": 2.124583226468325e-05, + "loss": 0.4065, + "step": 44895 + }, + { + "epoch": 1.73, + "learning_rate": 2.1245191074634522e-05, + "loss": 0.2642, + "step": 44896 + }, + { + "epoch": 1.73, + "learning_rate": 2.1244549884585794e-05, + "loss": 0.2418, + "step": 44897 + }, + { + "epoch": 1.73, + "learning_rate": 2.1243908694537063e-05, + "loss": 0.4345, + "step": 44898 + }, + { + "epoch": 1.73, + "learning_rate": 2.124326750448833e-05, + "loss": 0.2868, + "step": 44899 + }, + { + "epoch": 1.73, + "learning_rate": 2.1242626314439603e-05, + "loss": 0.5556, + "step": 44900 + }, + { + "epoch": 1.73, + "learning_rate": 2.124198512439087e-05, + "loss": 0.1877, + "step": 44901 + }, + { + "epoch": 1.73, + "learning_rate": 2.124134393434214e-05, + "loss": 0.339, + "step": 44902 + }, + { + "epoch": 1.73, + "learning_rate": 2.1240702744293408e-05, + "loss": 0.2753, + "step": 44903 + }, + { + "epoch": 1.73, + "learning_rate": 2.124006155424468e-05, + "loss": 0.272, + "step": 44904 + }, + { + "epoch": 1.73, + "learning_rate": 2.123942036419595e-05, + "loss": 0.3896, + "step": 44905 + }, + { + "epoch": 1.73, + "learning_rate": 2.123877917414722e-05, + "loss": 0.6399, + "step": 44906 + }, + { + "epoch": 1.73, + "learning_rate": 2.1238137984098488e-05, + "loss": 0.3151, + "step": 44907 + }, + { + "epoch": 1.73, + "learning_rate": 2.123749679404976e-05, + "loss": 0.3763, + "step": 44908 + }, + { + "epoch": 1.73, + "learning_rate": 2.1236855604001028e-05, + "loss": 0.2755, + "step": 44909 + }, + { + "epoch": 1.73, + "learning_rate": 2.1236214413952296e-05, + "loss": 0.3162, + "step": 44910 + }, + { + "epoch": 1.73, + "learning_rate": 2.1235573223903564e-05, + "loss": 0.2751, + "step": 44911 + }, + { + "epoch": 1.73, + "learning_rate": 2.1234932033854836e-05, + "loss": 0.2886, + "step": 44912 + }, + { + "epoch": 1.73, + "learning_rate": 2.1234290843806104e-05, + "loss": 0.4276, + "step": 44913 + }, + { + "epoch": 1.73, + "learning_rate": 2.1233649653757376e-05, + "loss": 0.2954, + "step": 44914 + }, + { + "epoch": 1.73, + "learning_rate": 2.1233008463708644e-05, + "loss": 0.2718, + "step": 44915 + }, + { + "epoch": 1.73, + "learning_rate": 2.1232367273659913e-05, + "loss": 0.478, + "step": 44916 + }, + { + "epoch": 1.73, + "learning_rate": 2.1231726083611184e-05, + "loss": 0.3871, + "step": 44917 + }, + { + "epoch": 1.73, + "learning_rate": 2.1231084893562453e-05, + "loss": 0.3002, + "step": 44918 + }, + { + "epoch": 1.73, + "learning_rate": 2.123044370351372e-05, + "loss": 0.4627, + "step": 44919 + }, + { + "epoch": 1.73, + "learning_rate": 2.1229802513464993e-05, + "loss": 0.3637, + "step": 44920 + }, + { + "epoch": 1.73, + "learning_rate": 2.122916132341626e-05, + "loss": 0.3972, + "step": 44921 + }, + { + "epoch": 1.73, + "learning_rate": 2.122852013336753e-05, + "loss": 0.4196, + "step": 44922 + }, + { + "epoch": 1.73, + "learning_rate": 2.12278789433188e-05, + "loss": 0.2551, + "step": 44923 + }, + { + "epoch": 1.73, + "learning_rate": 2.122723775327007e-05, + "loss": 0.3225, + "step": 44924 + }, + { + "epoch": 1.73, + "learning_rate": 2.122659656322134e-05, + "loss": 0.2693, + "step": 44925 + }, + { + "epoch": 1.73, + "learning_rate": 2.122595537317261e-05, + "loss": 0.36, + "step": 44926 + }, + { + "epoch": 1.73, + "learning_rate": 2.1225314183123878e-05, + "loss": 0.3535, + "step": 44927 + }, + { + "epoch": 1.73, + "learning_rate": 2.122467299307515e-05, + "loss": 0.3043, + "step": 44928 + }, + { + "epoch": 1.73, + "learning_rate": 2.1224031803026418e-05, + "loss": 0.3379, + "step": 44929 + }, + { + "epoch": 1.73, + "learning_rate": 2.1223390612977686e-05, + "loss": 0.1831, + "step": 44930 + }, + { + "epoch": 1.73, + "learning_rate": 2.1222749422928955e-05, + "loss": 0.2689, + "step": 44931 + }, + { + "epoch": 1.73, + "learning_rate": 2.1222108232880226e-05, + "loss": 0.2548, + "step": 44932 + }, + { + "epoch": 1.73, + "learning_rate": 2.1221467042831498e-05, + "loss": 0.3449, + "step": 44933 + }, + { + "epoch": 1.73, + "learning_rate": 2.1220825852782766e-05, + "loss": 0.3983, + "step": 44934 + }, + { + "epoch": 1.73, + "learning_rate": 2.1220184662734035e-05, + "loss": 0.3012, + "step": 44935 + }, + { + "epoch": 1.73, + "learning_rate": 2.1219543472685306e-05, + "loss": 0.4141, + "step": 44936 + }, + { + "epoch": 1.73, + "learning_rate": 2.1218902282636575e-05, + "loss": 0.3582, + "step": 44937 + }, + { + "epoch": 1.73, + "learning_rate": 2.1218261092587843e-05, + "loss": 0.2402, + "step": 44938 + }, + { + "epoch": 1.73, + "learning_rate": 2.121761990253911e-05, + "loss": 0.2394, + "step": 44939 + }, + { + "epoch": 1.73, + "learning_rate": 2.1216978712490383e-05, + "loss": 0.2908, + "step": 44940 + }, + { + "epoch": 1.73, + "learning_rate": 2.1216337522441655e-05, + "loss": 0.4047, + "step": 44941 + }, + { + "epoch": 1.73, + "learning_rate": 2.1215696332392923e-05, + "loss": 0.3656, + "step": 44942 + }, + { + "epoch": 1.73, + "learning_rate": 2.121505514234419e-05, + "loss": 0.2589, + "step": 44943 + }, + { + "epoch": 1.73, + "learning_rate": 2.1214413952295463e-05, + "loss": 0.428, + "step": 44944 + }, + { + "epoch": 1.73, + "learning_rate": 2.121377276224673e-05, + "loss": 0.2921, + "step": 44945 + }, + { + "epoch": 1.73, + "learning_rate": 2.1213131572198e-05, + "loss": 0.2662, + "step": 44946 + }, + { + "epoch": 1.73, + "learning_rate": 2.1212490382149268e-05, + "loss": 0.315, + "step": 44947 + }, + { + "epoch": 1.73, + "learning_rate": 2.121184919210054e-05, + "loss": 0.4308, + "step": 44948 + }, + { + "epoch": 1.73, + "learning_rate": 2.1211208002051808e-05, + "loss": 0.4507, + "step": 44949 + }, + { + "epoch": 1.73, + "learning_rate": 2.121056681200308e-05, + "loss": 0.2398, + "step": 44950 + }, + { + "epoch": 1.73, + "learning_rate": 2.1209925621954348e-05, + "loss": 0.3202, + "step": 44951 + }, + { + "epoch": 1.73, + "learning_rate": 2.1209284431905617e-05, + "loss": 0.3421, + "step": 44952 + }, + { + "epoch": 1.73, + "learning_rate": 2.120864324185689e-05, + "loss": 0.2794, + "step": 44953 + }, + { + "epoch": 1.73, + "learning_rate": 2.1208002051808157e-05, + "loss": 0.3863, + "step": 44954 + }, + { + "epoch": 1.73, + "learning_rate": 2.1207360861759425e-05, + "loss": 0.3217, + "step": 44955 + }, + { + "epoch": 1.73, + "learning_rate": 2.1206719671710697e-05, + "loss": 0.2126, + "step": 44956 + }, + { + "epoch": 1.73, + "learning_rate": 2.1206078481661965e-05, + "loss": 0.2777, + "step": 44957 + }, + { + "epoch": 1.73, + "learning_rate": 2.1205437291613237e-05, + "loss": 0.4555, + "step": 44958 + }, + { + "epoch": 1.73, + "learning_rate": 2.1204796101564505e-05, + "loss": 0.2738, + "step": 44959 + }, + { + "epoch": 1.73, + "learning_rate": 2.1204154911515773e-05, + "loss": 0.461, + "step": 44960 + }, + { + "epoch": 1.73, + "learning_rate": 2.1203513721467045e-05, + "loss": 0.3106, + "step": 44961 + }, + { + "epoch": 1.73, + "learning_rate": 2.1202872531418313e-05, + "loss": 0.3247, + "step": 44962 + }, + { + "epoch": 1.73, + "learning_rate": 2.1202231341369582e-05, + "loss": 0.2609, + "step": 44963 + }, + { + "epoch": 1.73, + "learning_rate": 2.1201590151320854e-05, + "loss": 0.3109, + "step": 44964 + }, + { + "epoch": 1.73, + "learning_rate": 2.1200948961272122e-05, + "loss": 0.2906, + "step": 44965 + }, + { + "epoch": 1.73, + "learning_rate": 2.120030777122339e-05, + "loss": 0.2973, + "step": 44966 + }, + { + "epoch": 1.73, + "learning_rate": 2.1199666581174662e-05, + "loss": 0.2138, + "step": 44967 + }, + { + "epoch": 1.73, + "learning_rate": 2.119902539112593e-05, + "loss": 0.2804, + "step": 44968 + }, + { + "epoch": 1.73, + "learning_rate": 2.1198384201077202e-05, + "loss": 0.3057, + "step": 44969 + }, + { + "epoch": 1.73, + "learning_rate": 2.119774301102847e-05, + "loss": 0.4181, + "step": 44970 + }, + { + "epoch": 1.73, + "learning_rate": 2.119710182097974e-05, + "loss": 0.2672, + "step": 44971 + }, + { + "epoch": 1.73, + "learning_rate": 2.119646063093101e-05, + "loss": 0.3137, + "step": 44972 + }, + { + "epoch": 1.73, + "learning_rate": 2.119581944088228e-05, + "loss": 0.2857, + "step": 44973 + }, + { + "epoch": 1.73, + "learning_rate": 2.1195178250833547e-05, + "loss": 0.2837, + "step": 44974 + }, + { + "epoch": 1.73, + "learning_rate": 2.1194537060784815e-05, + "loss": 0.3989, + "step": 44975 + }, + { + "epoch": 1.73, + "learning_rate": 2.1193895870736087e-05, + "loss": 0.3871, + "step": 44976 + }, + { + "epoch": 1.73, + "learning_rate": 2.119325468068736e-05, + "loss": 0.3357, + "step": 44977 + }, + { + "epoch": 1.73, + "learning_rate": 2.1192613490638627e-05, + "loss": 0.3222, + "step": 44978 + }, + { + "epoch": 1.73, + "learning_rate": 2.1191972300589895e-05, + "loss": 0.3766, + "step": 44979 + }, + { + "epoch": 1.73, + "learning_rate": 2.1191331110541167e-05, + "loss": 0.3321, + "step": 44980 + }, + { + "epoch": 1.73, + "learning_rate": 2.1190689920492435e-05, + "loss": 0.3516, + "step": 44981 + }, + { + "epoch": 1.73, + "learning_rate": 2.1190048730443704e-05, + "loss": 0.2944, + "step": 44982 + }, + { + "epoch": 1.73, + "learning_rate": 2.1189407540394972e-05, + "loss": 0.3877, + "step": 44983 + }, + { + "epoch": 1.73, + "learning_rate": 2.1188766350346244e-05, + "loss": 0.3659, + "step": 44984 + }, + { + "epoch": 1.73, + "learning_rate": 2.1188125160297516e-05, + "loss": 0.2845, + "step": 44985 + }, + { + "epoch": 1.73, + "learning_rate": 2.1187483970248784e-05, + "loss": 0.3347, + "step": 44986 + }, + { + "epoch": 1.73, + "learning_rate": 2.1186842780200052e-05, + "loss": 0.4118, + "step": 44987 + }, + { + "epoch": 1.73, + "learning_rate": 2.1186201590151324e-05, + "loss": 0.383, + "step": 44988 + }, + { + "epoch": 1.73, + "learning_rate": 2.1185560400102592e-05, + "loss": 0.3014, + "step": 44989 + }, + { + "epoch": 1.73, + "learning_rate": 2.118491921005386e-05, + "loss": 0.2827, + "step": 44990 + }, + { + "epoch": 1.73, + "learning_rate": 2.118427802000513e-05, + "loss": 0.4037, + "step": 44991 + }, + { + "epoch": 1.73, + "learning_rate": 2.11836368299564e-05, + "loss": 0.3833, + "step": 44992 + }, + { + "epoch": 1.73, + "learning_rate": 2.118299563990767e-05, + "loss": 0.3927, + "step": 44993 + }, + { + "epoch": 1.73, + "learning_rate": 2.118235444985894e-05, + "loss": 0.4064, + "step": 44994 + }, + { + "epoch": 1.73, + "learning_rate": 2.118171325981021e-05, + "loss": 0.322, + "step": 44995 + }, + { + "epoch": 1.73, + "learning_rate": 2.1181072069761477e-05, + "loss": 0.4342, + "step": 44996 + }, + { + "epoch": 1.73, + "learning_rate": 2.118043087971275e-05, + "loss": 0.2996, + "step": 44997 + }, + { + "epoch": 1.73, + "learning_rate": 2.1179789689664017e-05, + "loss": 0.3164, + "step": 44998 + }, + { + "epoch": 1.73, + "learning_rate": 2.1179148499615286e-05, + "loss": 0.2697, + "step": 44999 + }, + { + "epoch": 1.73, + "learning_rate": 2.1178507309566557e-05, + "loss": 0.3064, + "step": 45000 + }, + { + "epoch": 1.73, + "learning_rate": 2.1177866119517826e-05, + "loss": 0.3736, + "step": 45001 + }, + { + "epoch": 1.73, + "learning_rate": 2.1177224929469094e-05, + "loss": 0.3172, + "step": 45002 + }, + { + "epoch": 1.73, + "learning_rate": 2.1176583739420366e-05, + "loss": 0.4393, + "step": 45003 + }, + { + "epoch": 1.73, + "learning_rate": 2.1175942549371634e-05, + "loss": 0.3171, + "step": 45004 + }, + { + "epoch": 1.73, + "learning_rate": 2.1175301359322906e-05, + "loss": 0.2943, + "step": 45005 + }, + { + "epoch": 1.73, + "learning_rate": 2.1174660169274174e-05, + "loss": 0.1372, + "step": 45006 + }, + { + "epoch": 1.73, + "learning_rate": 2.1174018979225442e-05, + "loss": 0.408, + "step": 45007 + }, + { + "epoch": 1.73, + "learning_rate": 2.1173377789176714e-05, + "loss": 0.3697, + "step": 45008 + }, + { + "epoch": 1.73, + "learning_rate": 2.1172736599127982e-05, + "loss": 0.3757, + "step": 45009 + }, + { + "epoch": 1.73, + "learning_rate": 2.117209540907925e-05, + "loss": 0.2166, + "step": 45010 + }, + { + "epoch": 1.73, + "learning_rate": 2.117145421903052e-05, + "loss": 0.2656, + "step": 45011 + }, + { + "epoch": 1.73, + "learning_rate": 2.117081302898179e-05, + "loss": 0.3131, + "step": 45012 + }, + { + "epoch": 1.73, + "learning_rate": 2.1170171838933063e-05, + "loss": 0.2741, + "step": 45013 + }, + { + "epoch": 1.73, + "learning_rate": 2.116953064888433e-05, + "loss": 0.2968, + "step": 45014 + }, + { + "epoch": 1.73, + "learning_rate": 2.11688894588356e-05, + "loss": 0.3335, + "step": 45015 + }, + { + "epoch": 1.73, + "learning_rate": 2.116824826878687e-05, + "loss": 0.3506, + "step": 45016 + }, + { + "epoch": 1.73, + "learning_rate": 2.116760707873814e-05, + "loss": 0.2861, + "step": 45017 + }, + { + "epoch": 1.73, + "learning_rate": 2.1166965888689408e-05, + "loss": 0.3183, + "step": 45018 + }, + { + "epoch": 1.73, + "learning_rate": 2.1166324698640676e-05, + "loss": 0.2538, + "step": 45019 + }, + { + "epoch": 1.73, + "learning_rate": 2.1165683508591948e-05, + "loss": 0.3335, + "step": 45020 + }, + { + "epoch": 1.73, + "learning_rate": 2.116504231854322e-05, + "loss": 0.4085, + "step": 45021 + }, + { + "epoch": 1.73, + "learning_rate": 2.1164401128494488e-05, + "loss": 0.2375, + "step": 45022 + }, + { + "epoch": 1.73, + "learning_rate": 2.1163759938445756e-05, + "loss": 0.3226, + "step": 45023 + }, + { + "epoch": 1.73, + "learning_rate": 2.1163118748397028e-05, + "loss": 0.3126, + "step": 45024 + }, + { + "epoch": 1.73, + "learning_rate": 2.1162477558348296e-05, + "loss": 0.24, + "step": 45025 + }, + { + "epoch": 1.73, + "learning_rate": 2.1161836368299564e-05, + "loss": 0.3075, + "step": 45026 + }, + { + "epoch": 1.73, + "learning_rate": 2.1161195178250833e-05, + "loss": 0.2832, + "step": 45027 + }, + { + "epoch": 1.73, + "learning_rate": 2.1160553988202104e-05, + "loss": 0.2871, + "step": 45028 + }, + { + "epoch": 1.73, + "learning_rate": 2.1159912798153373e-05, + "loss": 0.4126, + "step": 45029 + }, + { + "epoch": 1.73, + "learning_rate": 2.1159271608104644e-05, + "loss": 0.4397, + "step": 45030 + }, + { + "epoch": 1.73, + "learning_rate": 2.1158630418055913e-05, + "loss": 0.3162, + "step": 45031 + }, + { + "epoch": 1.73, + "learning_rate": 2.1157989228007185e-05, + "loss": 0.3228, + "step": 45032 + }, + { + "epoch": 1.73, + "learning_rate": 2.1157348037958453e-05, + "loss": 0.3638, + "step": 45033 + }, + { + "epoch": 1.73, + "learning_rate": 2.115670684790972e-05, + "loss": 0.2516, + "step": 45034 + }, + { + "epoch": 1.73, + "learning_rate": 2.115606565786099e-05, + "loss": 0.3676, + "step": 45035 + }, + { + "epoch": 1.73, + "learning_rate": 2.115542446781226e-05, + "loss": 0.4093, + "step": 45036 + }, + { + "epoch": 1.73, + "learning_rate": 2.115478327776353e-05, + "loss": 0.2541, + "step": 45037 + }, + { + "epoch": 1.73, + "learning_rate": 2.1154142087714798e-05, + "loss": 0.2998, + "step": 45038 + }, + { + "epoch": 1.73, + "learning_rate": 2.115350089766607e-05, + "loss": 0.3722, + "step": 45039 + }, + { + "epoch": 1.73, + "learning_rate": 2.1152859707617338e-05, + "loss": 0.2849, + "step": 45040 + }, + { + "epoch": 1.73, + "learning_rate": 2.115221851756861e-05, + "loss": 0.2095, + "step": 45041 + }, + { + "epoch": 1.73, + "learning_rate": 2.1151577327519878e-05, + "loss": 0.2502, + "step": 45042 + }, + { + "epoch": 1.73, + "learning_rate": 2.1150936137471146e-05, + "loss": 0.242, + "step": 45043 + }, + { + "epoch": 1.73, + "learning_rate": 2.1150294947422418e-05, + "loss": 0.3847, + "step": 45044 + }, + { + "epoch": 1.73, + "learning_rate": 2.1149653757373686e-05, + "loss": 0.2399, + "step": 45045 + }, + { + "epoch": 1.73, + "learning_rate": 2.1149012567324955e-05, + "loss": 0.2246, + "step": 45046 + }, + { + "epoch": 1.73, + "learning_rate": 2.1148371377276223e-05, + "loss": 0.4039, + "step": 45047 + }, + { + "epoch": 1.73, + "learning_rate": 2.1147730187227495e-05, + "loss": 0.3066, + "step": 45048 + }, + { + "epoch": 1.73, + "learning_rate": 2.1147088997178766e-05, + "loss": 0.5024, + "step": 45049 + }, + { + "epoch": 1.73, + "learning_rate": 2.1146447807130035e-05, + "loss": 0.3143, + "step": 45050 + }, + { + "epoch": 1.73, + "learning_rate": 2.1145806617081303e-05, + "loss": 0.2752, + "step": 45051 + }, + { + "epoch": 1.73, + "learning_rate": 2.1145165427032575e-05, + "loss": 0.4038, + "step": 45052 + }, + { + "epoch": 1.73, + "learning_rate": 2.1144524236983843e-05, + "loss": 0.3024, + "step": 45053 + }, + { + "epoch": 1.73, + "learning_rate": 2.114388304693511e-05, + "loss": 0.3015, + "step": 45054 + }, + { + "epoch": 1.73, + "learning_rate": 2.114324185688638e-05, + "loss": 0.3596, + "step": 45055 + }, + { + "epoch": 1.73, + "learning_rate": 2.114260066683765e-05, + "loss": 0.5434, + "step": 45056 + }, + { + "epoch": 1.73, + "learning_rate": 2.1141959476788923e-05, + "loss": 0.2851, + "step": 45057 + }, + { + "epoch": 1.73, + "learning_rate": 2.114131828674019e-05, + "loss": 0.3977, + "step": 45058 + }, + { + "epoch": 1.73, + "learning_rate": 2.114067709669146e-05, + "loss": 0.325, + "step": 45059 + }, + { + "epoch": 1.73, + "learning_rate": 2.114003590664273e-05, + "loss": 0.259, + "step": 45060 + }, + { + "epoch": 1.73, + "learning_rate": 2.1139394716594e-05, + "loss": 0.2715, + "step": 45061 + }, + { + "epoch": 1.73, + "learning_rate": 2.1138753526545268e-05, + "loss": 0.2734, + "step": 45062 + }, + { + "epoch": 1.73, + "learning_rate": 2.1138112336496537e-05, + "loss": 0.2769, + "step": 45063 + }, + { + "epoch": 1.73, + "learning_rate": 2.113747114644781e-05, + "loss": 0.36, + "step": 45064 + }, + { + "epoch": 1.73, + "learning_rate": 2.113682995639908e-05, + "loss": 0.3714, + "step": 45065 + }, + { + "epoch": 1.73, + "learning_rate": 2.113618876635035e-05, + "loss": 0.3502, + "step": 45066 + }, + { + "epoch": 1.73, + "learning_rate": 2.1135547576301617e-05, + "loss": 0.1912, + "step": 45067 + }, + { + "epoch": 1.73, + "learning_rate": 2.113490638625289e-05, + "loss": 0.308, + "step": 45068 + }, + { + "epoch": 1.73, + "learning_rate": 2.1134265196204157e-05, + "loss": 0.2948, + "step": 45069 + }, + { + "epoch": 1.73, + "learning_rate": 2.1133624006155425e-05, + "loss": 0.421, + "step": 45070 + }, + { + "epoch": 1.73, + "learning_rate": 2.1132982816106693e-05, + "loss": 0.3902, + "step": 45071 + }, + { + "epoch": 1.73, + "learning_rate": 2.1132341626057965e-05, + "loss": 0.3592, + "step": 45072 + }, + { + "epoch": 1.73, + "learning_rate": 2.1131700436009233e-05, + "loss": 0.3516, + "step": 45073 + }, + { + "epoch": 1.73, + "learning_rate": 2.1131059245960505e-05, + "loss": 0.3284, + "step": 45074 + }, + { + "epoch": 1.73, + "learning_rate": 2.1130418055911773e-05, + "loss": 0.3344, + "step": 45075 + }, + { + "epoch": 1.73, + "learning_rate": 2.1129776865863042e-05, + "loss": 0.4457, + "step": 45076 + }, + { + "epoch": 1.73, + "learning_rate": 2.1129135675814314e-05, + "loss": 0.2821, + "step": 45077 + }, + { + "epoch": 1.73, + "learning_rate": 2.1128494485765582e-05, + "loss": 0.3126, + "step": 45078 + }, + { + "epoch": 1.73, + "learning_rate": 2.112785329571685e-05, + "loss": 0.2119, + "step": 45079 + }, + { + "epoch": 1.73, + "learning_rate": 2.1127212105668122e-05, + "loss": 0.3357, + "step": 45080 + }, + { + "epoch": 1.73, + "learning_rate": 2.112657091561939e-05, + "loss": 0.3734, + "step": 45081 + }, + { + "epoch": 1.73, + "learning_rate": 2.112592972557066e-05, + "loss": 0.4969, + "step": 45082 + }, + { + "epoch": 1.73, + "learning_rate": 2.112528853552193e-05, + "loss": 0.2953, + "step": 45083 + }, + { + "epoch": 1.73, + "learning_rate": 2.11246473454732e-05, + "loss": 0.2888, + "step": 45084 + }, + { + "epoch": 1.73, + "learning_rate": 2.112400615542447e-05, + "loss": 0.4121, + "step": 45085 + }, + { + "epoch": 1.73, + "learning_rate": 2.112336496537574e-05, + "loss": 0.3941, + "step": 45086 + }, + { + "epoch": 1.73, + "learning_rate": 2.1122723775327007e-05, + "loss": 0.375, + "step": 45087 + }, + { + "epoch": 1.73, + "learning_rate": 2.112208258527828e-05, + "loss": 0.3112, + "step": 45088 + }, + { + "epoch": 1.73, + "learning_rate": 2.1121441395229547e-05, + "loss": 0.3519, + "step": 45089 + }, + { + "epoch": 1.73, + "learning_rate": 2.1120800205180815e-05, + "loss": 0.3432, + "step": 45090 + }, + { + "epoch": 1.73, + "learning_rate": 2.1120159015132084e-05, + "loss": 0.2087, + "step": 45091 + }, + { + "epoch": 1.73, + "learning_rate": 2.1119517825083355e-05, + "loss": 0.4048, + "step": 45092 + }, + { + "epoch": 1.73, + "learning_rate": 2.1118876635034627e-05, + "loss": 0.3792, + "step": 45093 + }, + { + "epoch": 1.73, + "learning_rate": 2.1118235444985895e-05, + "loss": 0.4034, + "step": 45094 + }, + { + "epoch": 1.73, + "learning_rate": 2.1117594254937164e-05, + "loss": 0.3668, + "step": 45095 + }, + { + "epoch": 1.73, + "learning_rate": 2.1116953064888435e-05, + "loss": 0.2573, + "step": 45096 + }, + { + "epoch": 1.73, + "learning_rate": 2.1116311874839704e-05, + "loss": 0.3483, + "step": 45097 + }, + { + "epoch": 1.73, + "learning_rate": 2.1115670684790972e-05, + "loss": 0.5356, + "step": 45098 + }, + { + "epoch": 1.73, + "learning_rate": 2.111502949474224e-05, + "loss": 0.3529, + "step": 45099 + }, + { + "epoch": 1.73, + "learning_rate": 2.1114388304693512e-05, + "loss": 0.2809, + "step": 45100 + }, + { + "epoch": 1.73, + "learning_rate": 2.1113747114644784e-05, + "loss": 0.3402, + "step": 45101 + }, + { + "epoch": 1.73, + "learning_rate": 2.1113105924596052e-05, + "loss": 0.2836, + "step": 45102 + }, + { + "epoch": 1.73, + "learning_rate": 2.111246473454732e-05, + "loss": 0.3483, + "step": 45103 + }, + { + "epoch": 1.73, + "learning_rate": 2.1111823544498592e-05, + "loss": 0.2964, + "step": 45104 + }, + { + "epoch": 1.73, + "learning_rate": 2.111118235444986e-05, + "loss": 0.284, + "step": 45105 + }, + { + "epoch": 1.73, + "learning_rate": 2.111054116440113e-05, + "loss": 0.2801, + "step": 45106 + }, + { + "epoch": 1.73, + "learning_rate": 2.1109899974352397e-05, + "loss": 0.356, + "step": 45107 + }, + { + "epoch": 1.73, + "learning_rate": 2.110925878430367e-05, + "loss": 0.4291, + "step": 45108 + }, + { + "epoch": 1.73, + "learning_rate": 2.1108617594254937e-05, + "loss": 0.1472, + "step": 45109 + }, + { + "epoch": 1.73, + "learning_rate": 2.110797640420621e-05, + "loss": 0.2932, + "step": 45110 + }, + { + "epoch": 1.73, + "learning_rate": 2.1107335214157477e-05, + "loss": 0.3797, + "step": 45111 + }, + { + "epoch": 1.73, + "learning_rate": 2.110669402410875e-05, + "loss": 0.4443, + "step": 45112 + }, + { + "epoch": 1.73, + "learning_rate": 2.1106052834060017e-05, + "loss": 0.3904, + "step": 45113 + }, + { + "epoch": 1.73, + "learning_rate": 2.1105411644011286e-05, + "loss": 0.353, + "step": 45114 + }, + { + "epoch": 1.73, + "learning_rate": 2.1104770453962554e-05, + "loss": 0.2469, + "step": 45115 + }, + { + "epoch": 1.73, + "learning_rate": 2.1104129263913826e-05, + "loss": 0.2738, + "step": 45116 + }, + { + "epoch": 1.73, + "learning_rate": 2.1103488073865094e-05, + "loss": 0.3071, + "step": 45117 + }, + { + "epoch": 1.73, + "learning_rate": 2.1102846883816362e-05, + "loss": 0.2328, + "step": 45118 + }, + { + "epoch": 1.73, + "learning_rate": 2.1102205693767634e-05, + "loss": 0.4144, + "step": 45119 + }, + { + "epoch": 1.73, + "learning_rate": 2.1101564503718902e-05, + "loss": 0.3357, + "step": 45120 + }, + { + "epoch": 1.73, + "learning_rate": 2.1100923313670174e-05, + "loss": 0.297, + "step": 45121 + }, + { + "epoch": 1.73, + "learning_rate": 2.1100282123621443e-05, + "loss": 0.3707, + "step": 45122 + }, + { + "epoch": 1.73, + "learning_rate": 2.109964093357271e-05, + "loss": 0.3214, + "step": 45123 + }, + { + "epoch": 1.73, + "learning_rate": 2.1098999743523983e-05, + "loss": 0.415, + "step": 45124 + }, + { + "epoch": 1.73, + "learning_rate": 2.109835855347525e-05, + "loss": 0.311, + "step": 45125 + }, + { + "epoch": 1.73, + "learning_rate": 2.109771736342652e-05, + "loss": 0.2805, + "step": 45126 + }, + { + "epoch": 1.73, + "learning_rate": 2.1097076173377788e-05, + "loss": 0.3417, + "step": 45127 + }, + { + "epoch": 1.74, + "learning_rate": 2.109643498332906e-05, + "loss": 0.3336, + "step": 45128 + }, + { + "epoch": 1.74, + "learning_rate": 2.109579379328033e-05, + "loss": 0.2962, + "step": 45129 + }, + { + "epoch": 1.74, + "learning_rate": 2.10951526032316e-05, + "loss": 0.3583, + "step": 45130 + }, + { + "epoch": 1.74, + "learning_rate": 2.1094511413182868e-05, + "loss": 0.2602, + "step": 45131 + }, + { + "epoch": 1.74, + "learning_rate": 2.109387022313414e-05, + "loss": 0.536, + "step": 45132 + }, + { + "epoch": 1.74, + "learning_rate": 2.1093229033085408e-05, + "loss": 0.4419, + "step": 45133 + }, + { + "epoch": 1.74, + "learning_rate": 2.1092587843036676e-05, + "loss": 0.4402, + "step": 45134 + }, + { + "epoch": 1.74, + "learning_rate": 2.1091946652987944e-05, + "loss": 0.2601, + "step": 45135 + }, + { + "epoch": 1.74, + "learning_rate": 2.1091305462939216e-05, + "loss": 0.314, + "step": 45136 + }, + { + "epoch": 1.74, + "learning_rate": 2.1090664272890488e-05, + "loss": 0.3272, + "step": 45137 + }, + { + "epoch": 1.74, + "learning_rate": 2.1090023082841756e-05, + "loss": 0.5002, + "step": 45138 + }, + { + "epoch": 1.74, + "learning_rate": 2.1089381892793024e-05, + "loss": 0.2729, + "step": 45139 + }, + { + "epoch": 1.74, + "learning_rate": 2.1088740702744296e-05, + "loss": 0.3045, + "step": 45140 + }, + { + "epoch": 1.74, + "learning_rate": 2.1088099512695564e-05, + "loss": 0.1922, + "step": 45141 + }, + { + "epoch": 1.74, + "learning_rate": 2.1087458322646833e-05, + "loss": 0.3457, + "step": 45142 + }, + { + "epoch": 1.74, + "learning_rate": 2.10868171325981e-05, + "loss": 0.3903, + "step": 45143 + }, + { + "epoch": 1.74, + "learning_rate": 2.1086175942549373e-05, + "loss": 0.256, + "step": 45144 + }, + { + "epoch": 1.74, + "learning_rate": 2.1085534752500645e-05, + "loss": 0.3555, + "step": 45145 + }, + { + "epoch": 1.74, + "learning_rate": 2.1084893562451913e-05, + "loss": 0.2777, + "step": 45146 + }, + { + "epoch": 1.74, + "learning_rate": 2.108425237240318e-05, + "loss": 0.3098, + "step": 45147 + }, + { + "epoch": 1.74, + "learning_rate": 2.1083611182354453e-05, + "loss": 0.2531, + "step": 45148 + }, + { + "epoch": 1.74, + "learning_rate": 2.108296999230572e-05, + "loss": 0.3056, + "step": 45149 + }, + { + "epoch": 1.74, + "learning_rate": 2.108232880225699e-05, + "loss": 0.3671, + "step": 45150 + }, + { + "epoch": 1.74, + "learning_rate": 2.1081687612208258e-05, + "loss": 0.2594, + "step": 45151 + }, + { + "epoch": 1.74, + "learning_rate": 2.108104642215953e-05, + "loss": 0.2891, + "step": 45152 + }, + { + "epoch": 1.74, + "learning_rate": 2.1080405232110798e-05, + "loss": 0.2904, + "step": 45153 + }, + { + "epoch": 1.74, + "learning_rate": 2.107976404206207e-05, + "loss": 0.3358, + "step": 45154 + }, + { + "epoch": 1.74, + "learning_rate": 2.1079122852013338e-05, + "loss": 0.3437, + "step": 45155 + }, + { + "epoch": 1.74, + "learning_rate": 2.107848166196461e-05, + "loss": 0.4401, + "step": 45156 + }, + { + "epoch": 1.74, + "learning_rate": 2.1077840471915878e-05, + "loss": 0.1697, + "step": 45157 + }, + { + "epoch": 1.74, + "learning_rate": 2.1077199281867146e-05, + "loss": 0.4493, + "step": 45158 + }, + { + "epoch": 1.74, + "learning_rate": 2.1076558091818415e-05, + "loss": 0.2898, + "step": 45159 + }, + { + "epoch": 1.74, + "learning_rate": 2.1075916901769686e-05, + "loss": 0.3187, + "step": 45160 + }, + { + "epoch": 1.74, + "learning_rate": 2.1075275711720955e-05, + "loss": 0.2513, + "step": 45161 + }, + { + "epoch": 1.74, + "learning_rate": 2.1074634521672223e-05, + "loss": 0.4529, + "step": 45162 + }, + { + "epoch": 1.74, + "learning_rate": 2.1073993331623495e-05, + "loss": 0.279, + "step": 45163 + }, + { + "epoch": 1.74, + "learning_rate": 2.1073352141574763e-05, + "loss": 0.2539, + "step": 45164 + }, + { + "epoch": 1.74, + "learning_rate": 2.1072710951526035e-05, + "loss": 0.1632, + "step": 45165 + }, + { + "epoch": 1.74, + "learning_rate": 2.1072069761477303e-05, + "loss": 0.3123, + "step": 45166 + }, + { + "epoch": 1.74, + "learning_rate": 2.107142857142857e-05, + "loss": 0.2658, + "step": 45167 + }, + { + "epoch": 1.74, + "learning_rate": 2.1070787381379843e-05, + "loss": 0.2791, + "step": 45168 + }, + { + "epoch": 1.74, + "learning_rate": 2.107014619133111e-05, + "loss": 0.3223, + "step": 45169 + }, + { + "epoch": 1.74, + "learning_rate": 2.106950500128238e-05, + "loss": 0.3657, + "step": 45170 + }, + { + "epoch": 1.74, + "learning_rate": 2.1068863811233648e-05, + "loss": 0.2444, + "step": 45171 + }, + { + "epoch": 1.74, + "learning_rate": 2.106822262118492e-05, + "loss": 0.3813, + "step": 45172 + }, + { + "epoch": 1.74, + "learning_rate": 2.106758143113619e-05, + "loss": 0.3488, + "step": 45173 + }, + { + "epoch": 1.74, + "learning_rate": 2.106694024108746e-05, + "loss": 0.2556, + "step": 45174 + }, + { + "epoch": 1.74, + "learning_rate": 2.106629905103873e-05, + "loss": 0.2701, + "step": 45175 + }, + { + "epoch": 1.74, + "learning_rate": 2.106565786099e-05, + "loss": 0.3067, + "step": 45176 + }, + { + "epoch": 1.74, + "learning_rate": 2.106501667094127e-05, + "loss": 0.4279, + "step": 45177 + }, + { + "epoch": 1.74, + "learning_rate": 2.1064375480892537e-05, + "loss": 0.4817, + "step": 45178 + }, + { + "epoch": 1.74, + "learning_rate": 2.1063734290843805e-05, + "loss": 0.3399, + "step": 45179 + }, + { + "epoch": 1.74, + "learning_rate": 2.1063093100795077e-05, + "loss": 0.406, + "step": 45180 + }, + { + "epoch": 1.74, + "learning_rate": 2.106245191074635e-05, + "loss": 0.2831, + "step": 45181 + }, + { + "epoch": 1.74, + "learning_rate": 2.1061810720697617e-05, + "loss": 0.2314, + "step": 45182 + }, + { + "epoch": 1.74, + "learning_rate": 2.1061169530648885e-05, + "loss": 0.1774, + "step": 45183 + }, + { + "epoch": 1.74, + "learning_rate": 2.1060528340600157e-05, + "loss": 0.2377, + "step": 45184 + }, + { + "epoch": 1.74, + "learning_rate": 2.1059887150551425e-05, + "loss": 0.3898, + "step": 45185 + }, + { + "epoch": 1.74, + "learning_rate": 2.1059245960502693e-05, + "loss": 0.29, + "step": 45186 + }, + { + "epoch": 1.74, + "learning_rate": 2.1058604770453962e-05, + "loss": 0.3058, + "step": 45187 + }, + { + "epoch": 1.74, + "learning_rate": 2.1057963580405234e-05, + "loss": 0.3141, + "step": 45188 + }, + { + "epoch": 1.74, + "learning_rate": 2.1057322390356502e-05, + "loss": 0.3995, + "step": 45189 + }, + { + "epoch": 1.74, + "learning_rate": 2.1056681200307774e-05, + "loss": 0.4306, + "step": 45190 + }, + { + "epoch": 1.74, + "learning_rate": 2.1056040010259042e-05, + "loss": 0.3331, + "step": 45191 + }, + { + "epoch": 1.74, + "learning_rate": 2.1055398820210314e-05, + "loss": 0.272, + "step": 45192 + }, + { + "epoch": 1.74, + "learning_rate": 2.1054757630161582e-05, + "loss": 0.2512, + "step": 45193 + }, + { + "epoch": 1.74, + "learning_rate": 2.105411644011285e-05, + "loss": 0.4481, + "step": 45194 + }, + { + "epoch": 1.74, + "learning_rate": 2.105347525006412e-05, + "loss": 0.3105, + "step": 45195 + }, + { + "epoch": 1.74, + "learning_rate": 2.105283406001539e-05, + "loss": 0.2995, + "step": 45196 + }, + { + "epoch": 1.74, + "learning_rate": 2.105219286996666e-05, + "loss": 0.306, + "step": 45197 + }, + { + "epoch": 1.74, + "learning_rate": 2.1051551679917927e-05, + "loss": 0.365, + "step": 45198 + }, + { + "epoch": 1.74, + "learning_rate": 2.10509104898692e-05, + "loss": 0.4781, + "step": 45199 + }, + { + "epoch": 1.74, + "learning_rate": 2.1050269299820467e-05, + "loss": 0.451, + "step": 45200 + }, + { + "epoch": 1.74, + "learning_rate": 2.104962810977174e-05, + "loss": 0.2168, + "step": 45201 + }, + { + "epoch": 1.74, + "learning_rate": 2.1048986919723007e-05, + "loss": 0.3171, + "step": 45202 + }, + { + "epoch": 1.74, + "learning_rate": 2.1048345729674275e-05, + "loss": 0.3102, + "step": 45203 + }, + { + "epoch": 1.74, + "learning_rate": 2.1047704539625547e-05, + "loss": 0.2992, + "step": 45204 + }, + { + "epoch": 1.74, + "learning_rate": 2.1047063349576815e-05, + "loss": 0.3308, + "step": 45205 + }, + { + "epoch": 1.74, + "learning_rate": 2.1046422159528084e-05, + "loss": 0.3364, + "step": 45206 + }, + { + "epoch": 1.74, + "learning_rate": 2.1045780969479352e-05, + "loss": 0.4019, + "step": 45207 + }, + { + "epoch": 1.74, + "learning_rate": 2.1045139779430624e-05, + "loss": 0.3008, + "step": 45208 + }, + { + "epoch": 1.74, + "learning_rate": 2.1044498589381896e-05, + "loss": 0.287, + "step": 45209 + }, + { + "epoch": 1.74, + "learning_rate": 2.1043857399333164e-05, + "loss": 0.216, + "step": 45210 + }, + { + "epoch": 1.74, + "learning_rate": 2.1043216209284432e-05, + "loss": 0.2073, + "step": 45211 + }, + { + "epoch": 1.74, + "learning_rate": 2.1042575019235704e-05, + "loss": 0.3056, + "step": 45212 + }, + { + "epoch": 1.74, + "learning_rate": 2.1041933829186972e-05, + "loss": 0.2115, + "step": 45213 + }, + { + "epoch": 1.74, + "learning_rate": 2.104129263913824e-05, + "loss": 0.284, + "step": 45214 + }, + { + "epoch": 1.74, + "learning_rate": 2.104065144908951e-05, + "loss": 0.3814, + "step": 45215 + }, + { + "epoch": 1.74, + "learning_rate": 2.104001025904078e-05, + "loss": 0.3473, + "step": 45216 + }, + { + "epoch": 1.74, + "learning_rate": 2.1039369068992052e-05, + "loss": 0.3191, + "step": 45217 + }, + { + "epoch": 1.74, + "learning_rate": 2.103872787894332e-05, + "loss": 0.4954, + "step": 45218 + }, + { + "epoch": 1.74, + "learning_rate": 2.103808668889459e-05, + "loss": 0.3826, + "step": 45219 + }, + { + "epoch": 1.74, + "learning_rate": 2.103744549884586e-05, + "loss": 0.3663, + "step": 45220 + }, + { + "epoch": 1.74, + "learning_rate": 2.103680430879713e-05, + "loss": 0.3971, + "step": 45221 + }, + { + "epoch": 1.74, + "learning_rate": 2.1036163118748397e-05, + "loss": 0.3812, + "step": 45222 + }, + { + "epoch": 1.74, + "learning_rate": 2.1035521928699666e-05, + "loss": 0.3634, + "step": 45223 + }, + { + "epoch": 1.74, + "learning_rate": 2.1034880738650937e-05, + "loss": 0.302, + "step": 45224 + }, + { + "epoch": 1.74, + "learning_rate": 2.1034239548602206e-05, + "loss": 0.3545, + "step": 45225 + }, + { + "epoch": 1.74, + "learning_rate": 2.1033598358553477e-05, + "loss": 0.2, + "step": 45226 + }, + { + "epoch": 1.74, + "learning_rate": 2.1032957168504746e-05, + "loss": 0.4211, + "step": 45227 + }, + { + "epoch": 1.74, + "learning_rate": 2.1032315978456017e-05, + "loss": 0.383, + "step": 45228 + }, + { + "epoch": 1.74, + "learning_rate": 2.1031674788407286e-05, + "loss": 0.3812, + "step": 45229 + }, + { + "epoch": 1.74, + "learning_rate": 2.1031033598358554e-05, + "loss": 0.3928, + "step": 45230 + }, + { + "epoch": 1.74, + "learning_rate": 2.1030392408309822e-05, + "loss": 0.3869, + "step": 45231 + }, + { + "epoch": 1.74, + "learning_rate": 2.1029751218261094e-05, + "loss": 0.373, + "step": 45232 + }, + { + "epoch": 1.74, + "learning_rate": 2.1029110028212363e-05, + "loss": 0.3356, + "step": 45233 + }, + { + "epoch": 1.74, + "learning_rate": 2.102846883816363e-05, + "loss": 0.4563, + "step": 45234 + }, + { + "epoch": 1.74, + "learning_rate": 2.1027827648114903e-05, + "loss": 0.4253, + "step": 45235 + }, + { + "epoch": 1.74, + "learning_rate": 2.1027186458066174e-05, + "loss": 0.3255, + "step": 45236 + }, + { + "epoch": 1.74, + "learning_rate": 2.1026545268017443e-05, + "loss": 0.435, + "step": 45237 + }, + { + "epoch": 1.74, + "learning_rate": 2.102590407796871e-05, + "loss": 0.3659, + "step": 45238 + }, + { + "epoch": 1.74, + "learning_rate": 2.102526288791998e-05, + "loss": 0.4059, + "step": 45239 + }, + { + "epoch": 1.74, + "learning_rate": 2.102462169787125e-05, + "loss": 0.3569, + "step": 45240 + }, + { + "epoch": 1.74, + "learning_rate": 2.102398050782252e-05, + "loss": 0.278, + "step": 45241 + }, + { + "epoch": 1.74, + "learning_rate": 2.1023339317773788e-05, + "loss": 0.3332, + "step": 45242 + }, + { + "epoch": 1.74, + "learning_rate": 2.102269812772506e-05, + "loss": 0.4433, + "step": 45243 + }, + { + "epoch": 1.74, + "learning_rate": 2.1022056937676328e-05, + "loss": 0.3914, + "step": 45244 + }, + { + "epoch": 1.74, + "learning_rate": 2.10214157476276e-05, + "loss": 0.3112, + "step": 45245 + }, + { + "epoch": 1.74, + "learning_rate": 2.1020774557578868e-05, + "loss": 0.276, + "step": 45246 + }, + { + "epoch": 1.74, + "learning_rate": 2.1020133367530136e-05, + "loss": 0.4854, + "step": 45247 + }, + { + "epoch": 1.74, + "learning_rate": 2.1019492177481408e-05, + "loss": 0.3572, + "step": 45248 + }, + { + "epoch": 1.74, + "learning_rate": 2.1018850987432676e-05, + "loss": 0.395, + "step": 45249 + }, + { + "epoch": 1.74, + "learning_rate": 2.1018209797383944e-05, + "loss": 0.3813, + "step": 45250 + }, + { + "epoch": 1.74, + "learning_rate": 2.1017568607335213e-05, + "loss": 0.2405, + "step": 45251 + }, + { + "epoch": 1.74, + "learning_rate": 2.1016927417286484e-05, + "loss": 0.3201, + "step": 45252 + }, + { + "epoch": 1.74, + "learning_rate": 2.1016286227237756e-05, + "loss": 0.3669, + "step": 45253 + }, + { + "epoch": 1.74, + "learning_rate": 2.1015645037189025e-05, + "loss": 0.299, + "step": 45254 + }, + { + "epoch": 1.74, + "learning_rate": 2.1015003847140293e-05, + "loss": 0.2945, + "step": 45255 + }, + { + "epoch": 1.74, + "learning_rate": 2.1014362657091565e-05, + "loss": 0.3045, + "step": 45256 + }, + { + "epoch": 1.74, + "learning_rate": 2.1013721467042833e-05, + "loss": 0.3375, + "step": 45257 + }, + { + "epoch": 1.74, + "learning_rate": 2.10130802769941e-05, + "loss": 0.3087, + "step": 45258 + }, + { + "epoch": 1.74, + "learning_rate": 2.101243908694537e-05, + "loss": 0.3137, + "step": 45259 + }, + { + "epoch": 1.74, + "learning_rate": 2.101179789689664e-05, + "loss": 0.3226, + "step": 45260 + }, + { + "epoch": 1.74, + "learning_rate": 2.1011156706847913e-05, + "loss": 0.2493, + "step": 45261 + }, + { + "epoch": 1.74, + "learning_rate": 2.101051551679918e-05, + "loss": 0.3043, + "step": 45262 + }, + { + "epoch": 1.74, + "learning_rate": 2.100987432675045e-05, + "loss": 0.3333, + "step": 45263 + }, + { + "epoch": 1.74, + "learning_rate": 2.100923313670172e-05, + "loss": 0.4801, + "step": 45264 + }, + { + "epoch": 1.74, + "learning_rate": 2.100859194665299e-05, + "loss": 0.245, + "step": 45265 + }, + { + "epoch": 1.74, + "learning_rate": 2.1007950756604258e-05, + "loss": 0.3191, + "step": 45266 + }, + { + "epoch": 1.74, + "learning_rate": 2.1007309566555526e-05, + "loss": 0.4398, + "step": 45267 + }, + { + "epoch": 1.74, + "learning_rate": 2.1006668376506798e-05, + "loss": 0.308, + "step": 45268 + }, + { + "epoch": 1.74, + "learning_rate": 2.1006027186458066e-05, + "loss": 0.3459, + "step": 45269 + }, + { + "epoch": 1.74, + "learning_rate": 2.1005385996409338e-05, + "loss": 0.291, + "step": 45270 + }, + { + "epoch": 1.74, + "learning_rate": 2.1004744806360606e-05, + "loss": 0.3255, + "step": 45271 + }, + { + "epoch": 1.74, + "learning_rate": 2.1004103616311878e-05, + "loss": 0.4253, + "step": 45272 + }, + { + "epoch": 1.74, + "learning_rate": 2.1003462426263146e-05, + "loss": 0.2091, + "step": 45273 + }, + { + "epoch": 1.74, + "learning_rate": 2.1002821236214415e-05, + "loss": 0.3373, + "step": 45274 + }, + { + "epoch": 1.74, + "learning_rate": 2.1002180046165683e-05, + "loss": 0.2479, + "step": 45275 + }, + { + "epoch": 1.74, + "learning_rate": 2.1001538856116955e-05, + "loss": 0.2267, + "step": 45276 + }, + { + "epoch": 1.74, + "learning_rate": 2.1000897666068223e-05, + "loss": 0.2703, + "step": 45277 + }, + { + "epoch": 1.74, + "learning_rate": 2.100025647601949e-05, + "loss": 0.3702, + "step": 45278 + }, + { + "epoch": 1.74, + "learning_rate": 2.0999615285970763e-05, + "loss": 0.3752, + "step": 45279 + }, + { + "epoch": 1.74, + "learning_rate": 2.099897409592203e-05, + "loss": 0.3453, + "step": 45280 + }, + { + "epoch": 1.74, + "learning_rate": 2.0998332905873303e-05, + "loss": 0.3597, + "step": 45281 + }, + { + "epoch": 1.74, + "learning_rate": 2.099769171582457e-05, + "loss": 0.409, + "step": 45282 + }, + { + "epoch": 1.74, + "learning_rate": 2.099705052577584e-05, + "loss": 0.3511, + "step": 45283 + }, + { + "epoch": 1.74, + "learning_rate": 2.099640933572711e-05, + "loss": 0.3039, + "step": 45284 + }, + { + "epoch": 1.74, + "learning_rate": 2.099576814567838e-05, + "loss": 0.259, + "step": 45285 + }, + { + "epoch": 1.74, + "learning_rate": 2.0995126955629648e-05, + "loss": 0.3429, + "step": 45286 + }, + { + "epoch": 1.74, + "learning_rate": 2.0994485765580917e-05, + "loss": 0.4207, + "step": 45287 + }, + { + "epoch": 1.74, + "learning_rate": 2.099384457553219e-05, + "loss": 0.2412, + "step": 45288 + }, + { + "epoch": 1.74, + "learning_rate": 2.099320338548346e-05, + "loss": 0.252, + "step": 45289 + }, + { + "epoch": 1.74, + "learning_rate": 2.099256219543473e-05, + "loss": 0.3928, + "step": 45290 + }, + { + "epoch": 1.74, + "learning_rate": 2.0991921005385997e-05, + "loss": 0.2838, + "step": 45291 + }, + { + "epoch": 1.74, + "learning_rate": 2.099127981533727e-05, + "loss": 0.2853, + "step": 45292 + }, + { + "epoch": 1.74, + "learning_rate": 2.0990638625288537e-05, + "loss": 0.3352, + "step": 45293 + }, + { + "epoch": 1.74, + "learning_rate": 2.0989997435239805e-05, + "loss": 0.2911, + "step": 45294 + }, + { + "epoch": 1.74, + "learning_rate": 2.0989356245191073e-05, + "loss": 0.2855, + "step": 45295 + }, + { + "epoch": 1.74, + "learning_rate": 2.0988715055142345e-05, + "loss": 0.3185, + "step": 45296 + }, + { + "epoch": 1.74, + "learning_rate": 2.0988073865093617e-05, + "loss": 0.366, + "step": 45297 + }, + { + "epoch": 1.74, + "learning_rate": 2.0987432675044885e-05, + "loss": 0.4566, + "step": 45298 + }, + { + "epoch": 1.74, + "learning_rate": 2.0986791484996154e-05, + "loss": 0.4549, + "step": 45299 + }, + { + "epoch": 1.74, + "learning_rate": 2.0986150294947425e-05, + "loss": 0.2409, + "step": 45300 + }, + { + "epoch": 1.74, + "learning_rate": 2.0985509104898694e-05, + "loss": 0.4109, + "step": 45301 + }, + { + "epoch": 1.74, + "learning_rate": 2.0984867914849962e-05, + "loss": 0.3583, + "step": 45302 + }, + { + "epoch": 1.74, + "learning_rate": 2.098422672480123e-05, + "loss": 0.4172, + "step": 45303 + }, + { + "epoch": 1.74, + "learning_rate": 2.0983585534752502e-05, + "loss": 0.355, + "step": 45304 + }, + { + "epoch": 1.74, + "learning_rate": 2.098294434470377e-05, + "loss": 0.3887, + "step": 45305 + }, + { + "epoch": 1.74, + "learning_rate": 2.0982303154655042e-05, + "loss": 0.3197, + "step": 45306 + }, + { + "epoch": 1.74, + "learning_rate": 2.098166196460631e-05, + "loss": 0.2695, + "step": 45307 + }, + { + "epoch": 1.74, + "learning_rate": 2.0981020774557582e-05, + "loss": 0.391, + "step": 45308 + }, + { + "epoch": 1.74, + "learning_rate": 2.098037958450885e-05, + "loss": 0.4473, + "step": 45309 + }, + { + "epoch": 1.74, + "learning_rate": 2.097973839446012e-05, + "loss": 0.2937, + "step": 45310 + }, + { + "epoch": 1.74, + "learning_rate": 2.0979097204411387e-05, + "loss": 0.2666, + "step": 45311 + }, + { + "epoch": 1.74, + "learning_rate": 2.097845601436266e-05, + "loss": 0.3512, + "step": 45312 + }, + { + "epoch": 1.74, + "learning_rate": 2.0977814824313927e-05, + "loss": 0.4701, + "step": 45313 + }, + { + "epoch": 1.74, + "learning_rate": 2.0977173634265195e-05, + "loss": 0.3005, + "step": 45314 + }, + { + "epoch": 1.74, + "learning_rate": 2.0976532444216467e-05, + "loss": 0.3633, + "step": 45315 + }, + { + "epoch": 1.74, + "learning_rate": 2.097589125416774e-05, + "loss": 0.3362, + "step": 45316 + }, + { + "epoch": 1.74, + "learning_rate": 2.0975250064119007e-05, + "loss": 0.289, + "step": 45317 + }, + { + "epoch": 1.74, + "learning_rate": 2.0974608874070275e-05, + "loss": 0.3437, + "step": 45318 + }, + { + "epoch": 1.74, + "learning_rate": 2.0973967684021544e-05, + "loss": 0.4268, + "step": 45319 + }, + { + "epoch": 1.74, + "learning_rate": 2.0973326493972816e-05, + "loss": 0.3968, + "step": 45320 + }, + { + "epoch": 1.74, + "learning_rate": 2.0972685303924084e-05, + "loss": 0.3201, + "step": 45321 + }, + { + "epoch": 1.74, + "learning_rate": 2.0972044113875352e-05, + "loss": 0.2328, + "step": 45322 + }, + { + "epoch": 1.74, + "learning_rate": 2.097140292382662e-05, + "loss": 0.4537, + "step": 45323 + }, + { + "epoch": 1.74, + "learning_rate": 2.0970761733777892e-05, + "loss": 0.3125, + "step": 45324 + }, + { + "epoch": 1.74, + "learning_rate": 2.0970120543729164e-05, + "loss": 0.3622, + "step": 45325 + }, + { + "epoch": 1.74, + "learning_rate": 2.0969479353680432e-05, + "loss": 0.2999, + "step": 45326 + }, + { + "epoch": 1.74, + "learning_rate": 2.09688381636317e-05, + "loss": 0.2226, + "step": 45327 + }, + { + "epoch": 1.74, + "learning_rate": 2.0968196973582972e-05, + "loss": 0.3521, + "step": 45328 + }, + { + "epoch": 1.74, + "learning_rate": 2.096755578353424e-05, + "loss": 0.529, + "step": 45329 + }, + { + "epoch": 1.74, + "learning_rate": 2.096691459348551e-05, + "loss": 0.358, + "step": 45330 + }, + { + "epoch": 1.74, + "learning_rate": 2.0966273403436777e-05, + "loss": 0.3721, + "step": 45331 + }, + { + "epoch": 1.74, + "learning_rate": 2.096563221338805e-05, + "loss": 0.2959, + "step": 45332 + }, + { + "epoch": 1.74, + "learning_rate": 2.096499102333932e-05, + "loss": 0.3307, + "step": 45333 + }, + { + "epoch": 1.74, + "learning_rate": 2.096434983329059e-05, + "loss": 0.2614, + "step": 45334 + }, + { + "epoch": 1.74, + "learning_rate": 2.0963708643241857e-05, + "loss": 0.2955, + "step": 45335 + }, + { + "epoch": 1.74, + "learning_rate": 2.096306745319313e-05, + "loss": 0.4452, + "step": 45336 + }, + { + "epoch": 1.74, + "learning_rate": 2.0962426263144397e-05, + "loss": 0.3616, + "step": 45337 + }, + { + "epoch": 1.74, + "learning_rate": 2.0961785073095666e-05, + "loss": 0.2606, + "step": 45338 + }, + { + "epoch": 1.74, + "learning_rate": 2.0961143883046934e-05, + "loss": 0.3185, + "step": 45339 + }, + { + "epoch": 1.74, + "learning_rate": 2.0960502692998206e-05, + "loss": 0.2253, + "step": 45340 + }, + { + "epoch": 1.74, + "learning_rate": 2.0959861502949478e-05, + "loss": 0.2652, + "step": 45341 + }, + { + "epoch": 1.74, + "learning_rate": 2.0959220312900746e-05, + "loss": 0.4445, + "step": 45342 + }, + { + "epoch": 1.74, + "learning_rate": 2.0958579122852014e-05, + "loss": 0.3986, + "step": 45343 + }, + { + "epoch": 1.74, + "learning_rate": 2.0957937932803286e-05, + "loss": 0.3877, + "step": 45344 + }, + { + "epoch": 1.74, + "learning_rate": 2.0957296742754554e-05, + "loss": 0.3015, + "step": 45345 + }, + { + "epoch": 1.74, + "learning_rate": 2.0956655552705823e-05, + "loss": 0.3185, + "step": 45346 + }, + { + "epoch": 1.74, + "learning_rate": 2.095601436265709e-05, + "loss": 0.2922, + "step": 45347 + }, + { + "epoch": 1.74, + "learning_rate": 2.0955373172608363e-05, + "loss": 0.3033, + "step": 45348 + }, + { + "epoch": 1.74, + "learning_rate": 2.095473198255963e-05, + "loss": 0.3786, + "step": 45349 + }, + { + "epoch": 1.74, + "learning_rate": 2.0954090792510903e-05, + "loss": 0.3221, + "step": 45350 + }, + { + "epoch": 1.74, + "learning_rate": 2.095344960246217e-05, + "loss": 0.443, + "step": 45351 + }, + { + "epoch": 1.74, + "learning_rate": 2.0952808412413443e-05, + "loss": 0.418, + "step": 45352 + }, + { + "epoch": 1.74, + "learning_rate": 2.095216722236471e-05, + "loss": 0.4004, + "step": 45353 + }, + { + "epoch": 1.74, + "learning_rate": 2.095152603231598e-05, + "loss": 0.3974, + "step": 45354 + }, + { + "epoch": 1.74, + "learning_rate": 2.0950884842267248e-05, + "loss": 0.3108, + "step": 45355 + }, + { + "epoch": 1.74, + "learning_rate": 2.095024365221852e-05, + "loss": 0.2252, + "step": 45356 + }, + { + "epoch": 1.74, + "learning_rate": 2.0949602462169788e-05, + "loss": 0.4854, + "step": 45357 + }, + { + "epoch": 1.74, + "learning_rate": 2.0948961272121056e-05, + "loss": 0.3186, + "step": 45358 + }, + { + "epoch": 1.74, + "learning_rate": 2.0948320082072328e-05, + "loss": 0.2593, + "step": 45359 + }, + { + "epoch": 1.74, + "learning_rate": 2.09476788920236e-05, + "loss": 0.3147, + "step": 45360 + }, + { + "epoch": 1.74, + "learning_rate": 2.0947037701974868e-05, + "loss": 0.2897, + "step": 45361 + }, + { + "epoch": 1.74, + "learning_rate": 2.0946396511926136e-05, + "loss": 0.2618, + "step": 45362 + }, + { + "epoch": 1.74, + "learning_rate": 2.0945755321877404e-05, + "loss": 0.3567, + "step": 45363 + }, + { + "epoch": 1.74, + "learning_rate": 2.0945114131828676e-05, + "loss": 0.2499, + "step": 45364 + }, + { + "epoch": 1.74, + "learning_rate": 2.0944472941779945e-05, + "loss": 0.3884, + "step": 45365 + }, + { + "epoch": 1.74, + "learning_rate": 2.0943831751731213e-05, + "loss": 0.2215, + "step": 45366 + }, + { + "epoch": 1.74, + "learning_rate": 2.094319056168248e-05, + "loss": 0.2944, + "step": 45367 + }, + { + "epoch": 1.74, + "learning_rate": 2.0942549371633753e-05, + "loss": 0.4271, + "step": 45368 + }, + { + "epoch": 1.74, + "learning_rate": 2.0941908181585025e-05, + "loss": 0.4605, + "step": 45369 + }, + { + "epoch": 1.74, + "learning_rate": 2.0941266991536293e-05, + "loss": 0.2081, + "step": 45370 + }, + { + "epoch": 1.74, + "learning_rate": 2.094062580148756e-05, + "loss": 0.3369, + "step": 45371 + }, + { + "epoch": 1.74, + "learning_rate": 2.0939984611438833e-05, + "loss": 0.4726, + "step": 45372 + }, + { + "epoch": 1.74, + "learning_rate": 2.09393434213901e-05, + "loss": 0.4456, + "step": 45373 + }, + { + "epoch": 1.74, + "learning_rate": 2.093870223134137e-05, + "loss": 0.3415, + "step": 45374 + }, + { + "epoch": 1.74, + "learning_rate": 2.0938061041292638e-05, + "loss": 0.2278, + "step": 45375 + }, + { + "epoch": 1.74, + "learning_rate": 2.093741985124391e-05, + "loss": 0.2639, + "step": 45376 + }, + { + "epoch": 1.74, + "learning_rate": 2.093677866119518e-05, + "loss": 0.4148, + "step": 45377 + }, + { + "epoch": 1.74, + "learning_rate": 2.093613747114645e-05, + "loss": 0.3285, + "step": 45378 + }, + { + "epoch": 1.74, + "learning_rate": 2.0935496281097718e-05, + "loss": 0.3495, + "step": 45379 + }, + { + "epoch": 1.74, + "learning_rate": 2.093485509104899e-05, + "loss": 0.3182, + "step": 45380 + }, + { + "epoch": 1.74, + "learning_rate": 2.0934213901000258e-05, + "loss": 0.4156, + "step": 45381 + }, + { + "epoch": 1.74, + "learning_rate": 2.0933572710951526e-05, + "loss": 0.301, + "step": 45382 + }, + { + "epoch": 1.74, + "learning_rate": 2.0932931520902795e-05, + "loss": 0.3375, + "step": 45383 + }, + { + "epoch": 1.74, + "learning_rate": 2.0932290330854066e-05, + "loss": 0.2784, + "step": 45384 + }, + { + "epoch": 1.74, + "learning_rate": 2.0931649140805335e-05, + "loss": 0.4149, + "step": 45385 + }, + { + "epoch": 1.74, + "learning_rate": 2.0931007950756607e-05, + "loss": 0.2628, + "step": 45386 + }, + { + "epoch": 1.74, + "learning_rate": 2.0930366760707875e-05, + "loss": 0.3892, + "step": 45387 + }, + { + "epoch": 1.75, + "learning_rate": 2.0929725570659147e-05, + "loss": 0.3032, + "step": 45388 + }, + { + "epoch": 1.75, + "learning_rate": 2.0929084380610415e-05, + "loss": 0.3316, + "step": 45389 + }, + { + "epoch": 1.75, + "learning_rate": 2.0928443190561683e-05, + "loss": 0.3194, + "step": 45390 + }, + { + "epoch": 1.75, + "learning_rate": 2.092780200051295e-05, + "loss": 0.2052, + "step": 45391 + }, + { + "epoch": 1.75, + "learning_rate": 2.0927160810464223e-05, + "loss": 0.3729, + "step": 45392 + }, + { + "epoch": 1.75, + "learning_rate": 2.092651962041549e-05, + "loss": 0.4491, + "step": 45393 + }, + { + "epoch": 1.75, + "learning_rate": 2.092587843036676e-05, + "loss": 0.2647, + "step": 45394 + }, + { + "epoch": 1.75, + "learning_rate": 2.092523724031803e-05, + "loss": 0.3774, + "step": 45395 + }, + { + "epoch": 1.75, + "learning_rate": 2.0924596050269303e-05, + "loss": 0.2961, + "step": 45396 + }, + { + "epoch": 1.75, + "learning_rate": 2.092395486022057e-05, + "loss": 0.2247, + "step": 45397 + }, + { + "epoch": 1.75, + "learning_rate": 2.092331367017184e-05, + "loss": 0.2315, + "step": 45398 + }, + { + "epoch": 1.75, + "learning_rate": 2.092267248012311e-05, + "loss": 0.2805, + "step": 45399 + }, + { + "epoch": 1.75, + "learning_rate": 2.092203129007438e-05, + "loss": 0.2853, + "step": 45400 + }, + { + "epoch": 1.75, + "learning_rate": 2.092139010002565e-05, + "loss": 0.1846, + "step": 45401 + }, + { + "epoch": 1.75, + "learning_rate": 2.0920748909976917e-05, + "loss": 0.4352, + "step": 45402 + }, + { + "epoch": 1.75, + "learning_rate": 2.0920107719928185e-05, + "loss": 0.252, + "step": 45403 + }, + { + "epoch": 1.75, + "learning_rate": 2.0919466529879457e-05, + "loss": 0.305, + "step": 45404 + }, + { + "epoch": 1.75, + "learning_rate": 2.091882533983073e-05, + "loss": 0.3425, + "step": 45405 + }, + { + "epoch": 1.75, + "learning_rate": 2.0918184149781997e-05, + "loss": 0.3177, + "step": 45406 + }, + { + "epoch": 1.75, + "learning_rate": 2.0917542959733265e-05, + "loss": 0.3106, + "step": 45407 + }, + { + "epoch": 1.75, + "learning_rate": 2.0916901769684537e-05, + "loss": 0.3375, + "step": 45408 + }, + { + "epoch": 1.75, + "learning_rate": 2.0916260579635805e-05, + "loss": 0.2644, + "step": 45409 + }, + { + "epoch": 1.75, + "learning_rate": 2.0915619389587073e-05, + "loss": 0.374, + "step": 45410 + }, + { + "epoch": 1.75, + "learning_rate": 2.0914978199538342e-05, + "loss": 0.5173, + "step": 45411 + }, + { + "epoch": 1.75, + "learning_rate": 2.0914337009489614e-05, + "loss": 0.3532, + "step": 45412 + }, + { + "epoch": 1.75, + "learning_rate": 2.0913695819440885e-05, + "loss": 0.2618, + "step": 45413 + }, + { + "epoch": 1.75, + "learning_rate": 2.0913054629392154e-05, + "loss": 0.285, + "step": 45414 + }, + { + "epoch": 1.75, + "learning_rate": 2.0912413439343422e-05, + "loss": 0.3346, + "step": 45415 + }, + { + "epoch": 1.75, + "learning_rate": 2.0911772249294694e-05, + "loss": 0.3992, + "step": 45416 + }, + { + "epoch": 1.75, + "learning_rate": 2.0911131059245962e-05, + "loss": 0.4159, + "step": 45417 + }, + { + "epoch": 1.75, + "learning_rate": 2.091048986919723e-05, + "loss": 0.2638, + "step": 45418 + }, + { + "epoch": 1.75, + "learning_rate": 2.09098486791485e-05, + "loss": 0.2345, + "step": 45419 + }, + { + "epoch": 1.75, + "learning_rate": 2.090920748909977e-05, + "loss": 0.2449, + "step": 45420 + }, + { + "epoch": 1.75, + "learning_rate": 2.090856629905104e-05, + "loss": 0.2537, + "step": 45421 + }, + { + "epoch": 1.75, + "learning_rate": 2.090792510900231e-05, + "loss": 0.4294, + "step": 45422 + }, + { + "epoch": 1.75, + "learning_rate": 2.090728391895358e-05, + "loss": 0.3925, + "step": 45423 + }, + { + "epoch": 1.75, + "learning_rate": 2.090664272890485e-05, + "loss": 0.3198, + "step": 45424 + }, + { + "epoch": 1.75, + "learning_rate": 2.090600153885612e-05, + "loss": 0.2166, + "step": 45425 + }, + { + "epoch": 1.75, + "learning_rate": 2.0905360348807387e-05, + "loss": 0.4152, + "step": 45426 + }, + { + "epoch": 1.75, + "learning_rate": 2.0904719158758655e-05, + "loss": 0.2548, + "step": 45427 + }, + { + "epoch": 1.75, + "learning_rate": 2.0904077968709927e-05, + "loss": 0.1811, + "step": 45428 + }, + { + "epoch": 1.75, + "learning_rate": 2.0903436778661195e-05, + "loss": 0.3815, + "step": 45429 + }, + { + "epoch": 1.75, + "learning_rate": 2.0902795588612467e-05, + "loss": 0.4883, + "step": 45430 + }, + { + "epoch": 1.75, + "learning_rate": 2.0902154398563736e-05, + "loss": 0.392, + "step": 45431 + }, + { + "epoch": 1.75, + "learning_rate": 2.0901513208515007e-05, + "loss": 0.3295, + "step": 45432 + }, + { + "epoch": 1.75, + "learning_rate": 2.0900872018466276e-05, + "loss": 0.4186, + "step": 45433 + }, + { + "epoch": 1.75, + "learning_rate": 2.0900230828417544e-05, + "loss": 0.3288, + "step": 45434 + }, + { + "epoch": 1.75, + "learning_rate": 2.0899589638368812e-05, + "loss": 0.3016, + "step": 45435 + }, + { + "epoch": 1.75, + "learning_rate": 2.0898948448320084e-05, + "loss": 0.5841, + "step": 45436 + }, + { + "epoch": 1.75, + "learning_rate": 2.0898307258271352e-05, + "loss": 0.3875, + "step": 45437 + }, + { + "epoch": 1.75, + "learning_rate": 2.089766606822262e-05, + "loss": 0.3485, + "step": 45438 + }, + { + "epoch": 1.75, + "learning_rate": 2.0897024878173892e-05, + "loss": 0.2897, + "step": 45439 + }, + { + "epoch": 1.75, + "learning_rate": 2.0896383688125164e-05, + "loss": 0.3429, + "step": 45440 + }, + { + "epoch": 1.75, + "learning_rate": 2.0895742498076432e-05, + "loss": 0.3514, + "step": 45441 + }, + { + "epoch": 1.75, + "learning_rate": 2.08951013080277e-05, + "loss": 0.3343, + "step": 45442 + }, + { + "epoch": 1.75, + "learning_rate": 2.089446011797897e-05, + "loss": 0.3095, + "step": 45443 + }, + { + "epoch": 1.75, + "learning_rate": 2.089381892793024e-05, + "loss": 0.2633, + "step": 45444 + }, + { + "epoch": 1.75, + "learning_rate": 2.089317773788151e-05, + "loss": 0.2377, + "step": 45445 + }, + { + "epoch": 1.75, + "learning_rate": 2.0892536547832777e-05, + "loss": 0.3156, + "step": 45446 + }, + { + "epoch": 1.75, + "learning_rate": 2.0891895357784046e-05, + "loss": 0.3437, + "step": 45447 + }, + { + "epoch": 1.75, + "learning_rate": 2.0891254167735317e-05, + "loss": 0.3053, + "step": 45448 + }, + { + "epoch": 1.75, + "learning_rate": 2.089061297768659e-05, + "loss": 0.3638, + "step": 45449 + }, + { + "epoch": 1.75, + "learning_rate": 2.0889971787637857e-05, + "loss": 0.3985, + "step": 45450 + }, + { + "epoch": 1.75, + "learning_rate": 2.0889330597589126e-05, + "loss": 0.4162, + "step": 45451 + }, + { + "epoch": 1.75, + "learning_rate": 2.0888689407540398e-05, + "loss": 0.289, + "step": 45452 + }, + { + "epoch": 1.75, + "learning_rate": 2.0888048217491666e-05, + "loss": 0.2689, + "step": 45453 + }, + { + "epoch": 1.75, + "learning_rate": 2.0887407027442934e-05, + "loss": 0.3165, + "step": 45454 + }, + { + "epoch": 1.75, + "learning_rate": 2.0886765837394202e-05, + "loss": 0.3653, + "step": 45455 + }, + { + "epoch": 1.75, + "learning_rate": 2.0886124647345474e-05, + "loss": 0.3573, + "step": 45456 + }, + { + "epoch": 1.75, + "learning_rate": 2.0885483457296746e-05, + "loss": 0.3603, + "step": 45457 + }, + { + "epoch": 1.75, + "learning_rate": 2.0884842267248014e-05, + "loss": 0.2597, + "step": 45458 + }, + { + "epoch": 1.75, + "learning_rate": 2.0884201077199283e-05, + "loss": 0.4714, + "step": 45459 + }, + { + "epoch": 1.75, + "learning_rate": 2.0883559887150554e-05, + "loss": 0.429, + "step": 45460 + }, + { + "epoch": 1.75, + "learning_rate": 2.0882918697101823e-05, + "loss": 0.2744, + "step": 45461 + }, + { + "epoch": 1.75, + "learning_rate": 2.088227750705309e-05, + "loss": 0.2579, + "step": 45462 + }, + { + "epoch": 1.75, + "learning_rate": 2.088163631700436e-05, + "loss": 0.526, + "step": 45463 + }, + { + "epoch": 1.75, + "learning_rate": 2.088099512695563e-05, + "loss": 0.2352, + "step": 45464 + }, + { + "epoch": 1.75, + "learning_rate": 2.08803539369069e-05, + "loss": 0.3047, + "step": 45465 + }, + { + "epoch": 1.75, + "learning_rate": 2.087971274685817e-05, + "loss": 0.3188, + "step": 45466 + }, + { + "epoch": 1.75, + "learning_rate": 2.087907155680944e-05, + "loss": 0.2488, + "step": 45467 + }, + { + "epoch": 1.75, + "learning_rate": 2.087843036676071e-05, + "loss": 0.3377, + "step": 45468 + }, + { + "epoch": 1.75, + "learning_rate": 2.087778917671198e-05, + "loss": 0.3849, + "step": 45469 + }, + { + "epoch": 1.75, + "learning_rate": 2.0877147986663248e-05, + "loss": 0.4097, + "step": 45470 + }, + { + "epoch": 1.75, + "learning_rate": 2.0876506796614516e-05, + "loss": 0.408, + "step": 45471 + }, + { + "epoch": 1.75, + "learning_rate": 2.0875865606565788e-05, + "loss": 0.3661, + "step": 45472 + }, + { + "epoch": 1.75, + "learning_rate": 2.0875224416517056e-05, + "loss": 0.3374, + "step": 45473 + }, + { + "epoch": 1.75, + "learning_rate": 2.0874583226468324e-05, + "loss": 0.3137, + "step": 45474 + }, + { + "epoch": 1.75, + "learning_rate": 2.0873942036419596e-05, + "loss": 0.3534, + "step": 45475 + }, + { + "epoch": 1.75, + "learning_rate": 2.0873300846370868e-05, + "loss": 0.3775, + "step": 45476 + }, + { + "epoch": 1.75, + "learning_rate": 2.0872659656322136e-05, + "loss": 0.3802, + "step": 45477 + }, + { + "epoch": 1.75, + "learning_rate": 2.0872018466273405e-05, + "loss": 0.4414, + "step": 45478 + }, + { + "epoch": 1.75, + "learning_rate": 2.0871377276224673e-05, + "loss": 0.4926, + "step": 45479 + }, + { + "epoch": 1.75, + "learning_rate": 2.0870736086175945e-05, + "loss": 0.2956, + "step": 45480 + }, + { + "epoch": 1.75, + "learning_rate": 2.0870094896127213e-05, + "loss": 0.2953, + "step": 45481 + }, + { + "epoch": 1.75, + "learning_rate": 2.086945370607848e-05, + "loss": 0.4445, + "step": 45482 + }, + { + "epoch": 1.75, + "learning_rate": 2.086881251602975e-05, + "loss": 0.3272, + "step": 45483 + }, + { + "epoch": 1.75, + "learning_rate": 2.086817132598102e-05, + "loss": 0.3069, + "step": 45484 + }, + { + "epoch": 1.75, + "learning_rate": 2.0867530135932293e-05, + "loss": 0.3454, + "step": 45485 + }, + { + "epoch": 1.75, + "learning_rate": 2.086688894588356e-05, + "loss": 0.3855, + "step": 45486 + }, + { + "epoch": 1.75, + "learning_rate": 2.086624775583483e-05, + "loss": 0.347, + "step": 45487 + }, + { + "epoch": 1.75, + "learning_rate": 2.08656065657861e-05, + "loss": 0.2653, + "step": 45488 + }, + { + "epoch": 1.75, + "learning_rate": 2.086496537573737e-05, + "loss": 0.3199, + "step": 45489 + }, + { + "epoch": 1.75, + "learning_rate": 2.0864324185688638e-05, + "loss": 0.2949, + "step": 45490 + }, + { + "epoch": 1.75, + "learning_rate": 2.0863682995639906e-05, + "loss": 0.3506, + "step": 45491 + }, + { + "epoch": 1.75, + "learning_rate": 2.0863041805591178e-05, + "loss": 0.4092, + "step": 45492 + }, + { + "epoch": 1.75, + "learning_rate": 2.086240061554245e-05, + "loss": 0.2292, + "step": 45493 + }, + { + "epoch": 1.75, + "learning_rate": 2.0861759425493718e-05, + "loss": 0.3647, + "step": 45494 + }, + { + "epoch": 1.75, + "learning_rate": 2.0861118235444986e-05, + "loss": 0.3988, + "step": 45495 + }, + { + "epoch": 1.75, + "learning_rate": 2.0860477045396258e-05, + "loss": 0.3441, + "step": 45496 + }, + { + "epoch": 1.75, + "learning_rate": 2.0859835855347526e-05, + "loss": 0.2893, + "step": 45497 + }, + { + "epoch": 1.75, + "learning_rate": 2.0859194665298795e-05, + "loss": 0.2436, + "step": 45498 + }, + { + "epoch": 1.75, + "learning_rate": 2.0858553475250063e-05, + "loss": 0.3526, + "step": 45499 + }, + { + "epoch": 1.75, + "learning_rate": 2.0857912285201335e-05, + "loss": 0.3043, + "step": 45500 + }, + { + "epoch": 1.75, + "learning_rate": 2.0857271095152603e-05, + "loss": 0.2831, + "step": 45501 + }, + { + "epoch": 1.75, + "learning_rate": 2.0856629905103875e-05, + "loss": 0.3352, + "step": 45502 + }, + { + "epoch": 1.75, + "learning_rate": 2.0855988715055143e-05, + "loss": 0.2875, + "step": 45503 + }, + { + "epoch": 1.75, + "learning_rate": 2.0855347525006415e-05, + "loss": 0.3632, + "step": 45504 + }, + { + "epoch": 1.75, + "learning_rate": 2.0854706334957683e-05, + "loss": 0.2919, + "step": 45505 + }, + { + "epoch": 1.75, + "learning_rate": 2.085406514490895e-05, + "loss": 0.3827, + "step": 45506 + }, + { + "epoch": 1.75, + "learning_rate": 2.085342395486022e-05, + "loss": 0.3614, + "step": 45507 + }, + { + "epoch": 1.75, + "learning_rate": 2.085278276481149e-05, + "loss": 0.3491, + "step": 45508 + }, + { + "epoch": 1.75, + "learning_rate": 2.085214157476276e-05, + "loss": 0.211, + "step": 45509 + }, + { + "epoch": 1.75, + "learning_rate": 2.085150038471403e-05, + "loss": 0.3197, + "step": 45510 + }, + { + "epoch": 1.75, + "learning_rate": 2.08508591946653e-05, + "loss": 0.2775, + "step": 45511 + }, + { + "epoch": 1.75, + "learning_rate": 2.0850218004616572e-05, + "loss": 0.4344, + "step": 45512 + }, + { + "epoch": 1.75, + "learning_rate": 2.084957681456784e-05, + "loss": 0.356, + "step": 45513 + }, + { + "epoch": 1.75, + "learning_rate": 2.084893562451911e-05, + "loss": 0.2317, + "step": 45514 + }, + { + "epoch": 1.75, + "learning_rate": 2.0848294434470377e-05, + "loss": 0.2614, + "step": 45515 + }, + { + "epoch": 1.75, + "learning_rate": 2.084765324442165e-05, + "loss": 0.2696, + "step": 45516 + }, + { + "epoch": 1.75, + "learning_rate": 2.0847012054372917e-05, + "loss": 0.3942, + "step": 45517 + }, + { + "epoch": 1.75, + "learning_rate": 2.0846370864324185e-05, + "loss": 0.2676, + "step": 45518 + }, + { + "epoch": 1.75, + "learning_rate": 2.0845729674275453e-05, + "loss": 0.2846, + "step": 45519 + }, + { + "epoch": 1.75, + "learning_rate": 2.084508848422673e-05, + "loss": 0.3414, + "step": 45520 + }, + { + "epoch": 1.75, + "learning_rate": 2.0844447294177997e-05, + "loss": 0.4639, + "step": 45521 + }, + { + "epoch": 1.75, + "learning_rate": 2.0843806104129265e-05, + "loss": 0.3406, + "step": 45522 + }, + { + "epoch": 1.75, + "learning_rate": 2.0843164914080534e-05, + "loss": 0.4465, + "step": 45523 + }, + { + "epoch": 1.75, + "learning_rate": 2.0842523724031805e-05, + "loss": 0.3592, + "step": 45524 + }, + { + "epoch": 1.75, + "learning_rate": 2.0841882533983074e-05, + "loss": 0.4438, + "step": 45525 + }, + { + "epoch": 1.75, + "learning_rate": 2.0841241343934342e-05, + "loss": 0.3786, + "step": 45526 + }, + { + "epoch": 1.75, + "learning_rate": 2.084060015388561e-05, + "loss": 0.2139, + "step": 45527 + }, + { + "epoch": 1.75, + "learning_rate": 2.0839958963836882e-05, + "loss": 0.3471, + "step": 45528 + }, + { + "epoch": 1.75, + "learning_rate": 2.0839317773788154e-05, + "loss": 0.3528, + "step": 45529 + }, + { + "epoch": 1.75, + "learning_rate": 2.0838676583739422e-05, + "loss": 0.2862, + "step": 45530 + }, + { + "epoch": 1.75, + "learning_rate": 2.083803539369069e-05, + "loss": 0.3675, + "step": 45531 + }, + { + "epoch": 1.75, + "learning_rate": 2.0837394203641962e-05, + "loss": 0.3037, + "step": 45532 + }, + { + "epoch": 1.75, + "learning_rate": 2.083675301359323e-05, + "loss": 0.2675, + "step": 45533 + }, + { + "epoch": 1.75, + "learning_rate": 2.08361118235445e-05, + "loss": 0.4017, + "step": 45534 + }, + { + "epoch": 1.75, + "learning_rate": 2.0835470633495767e-05, + "loss": 0.4638, + "step": 45535 + }, + { + "epoch": 1.75, + "learning_rate": 2.083482944344704e-05, + "loss": 0.5725, + "step": 45536 + }, + { + "epoch": 1.75, + "learning_rate": 2.083418825339831e-05, + "loss": 0.2155, + "step": 45537 + }, + { + "epoch": 1.75, + "learning_rate": 2.083354706334958e-05, + "loss": 0.3229, + "step": 45538 + }, + { + "epoch": 1.75, + "learning_rate": 2.0832905873300847e-05, + "loss": 0.5318, + "step": 45539 + }, + { + "epoch": 1.75, + "learning_rate": 2.083226468325212e-05, + "loss": 0.3332, + "step": 45540 + }, + { + "epoch": 1.75, + "learning_rate": 2.0831623493203387e-05, + "loss": 0.5628, + "step": 45541 + }, + { + "epoch": 1.75, + "learning_rate": 2.0830982303154655e-05, + "loss": 0.3857, + "step": 45542 + }, + { + "epoch": 1.75, + "learning_rate": 2.0830341113105924e-05, + "loss": 0.3344, + "step": 45543 + }, + { + "epoch": 1.75, + "learning_rate": 2.0829699923057196e-05, + "loss": 0.2879, + "step": 45544 + }, + { + "epoch": 1.75, + "learning_rate": 2.0829058733008464e-05, + "loss": 0.3414, + "step": 45545 + }, + { + "epoch": 1.75, + "learning_rate": 2.0828417542959736e-05, + "loss": 0.3147, + "step": 45546 + }, + { + "epoch": 1.75, + "learning_rate": 2.0827776352911004e-05, + "loss": 0.4453, + "step": 45547 + }, + { + "epoch": 1.75, + "learning_rate": 2.0827135162862276e-05, + "loss": 0.3091, + "step": 45548 + }, + { + "epoch": 1.75, + "learning_rate": 2.0826493972813544e-05, + "loss": 0.3759, + "step": 45549 + }, + { + "epoch": 1.75, + "learning_rate": 2.0825852782764812e-05, + "loss": 0.3886, + "step": 45550 + }, + { + "epoch": 1.75, + "learning_rate": 2.082521159271608e-05, + "loss": 0.3099, + "step": 45551 + }, + { + "epoch": 1.75, + "learning_rate": 2.0824570402667352e-05, + "loss": 0.3567, + "step": 45552 + }, + { + "epoch": 1.75, + "learning_rate": 2.082392921261862e-05, + "loss": 0.3833, + "step": 45553 + }, + { + "epoch": 1.75, + "learning_rate": 2.082328802256989e-05, + "loss": 0.3251, + "step": 45554 + }, + { + "epoch": 1.75, + "learning_rate": 2.082264683252116e-05, + "loss": 0.2747, + "step": 45555 + }, + { + "epoch": 1.75, + "learning_rate": 2.0822005642472432e-05, + "loss": 0.4541, + "step": 45556 + }, + { + "epoch": 1.75, + "learning_rate": 2.08213644524237e-05, + "loss": 0.3923, + "step": 45557 + }, + { + "epoch": 1.75, + "learning_rate": 2.082072326237497e-05, + "loss": 0.4097, + "step": 45558 + }, + { + "epoch": 1.75, + "learning_rate": 2.0820082072326237e-05, + "loss": 0.4705, + "step": 45559 + }, + { + "epoch": 1.75, + "learning_rate": 2.081944088227751e-05, + "loss": 0.3172, + "step": 45560 + }, + { + "epoch": 1.75, + "learning_rate": 2.0818799692228777e-05, + "loss": 0.3196, + "step": 45561 + }, + { + "epoch": 1.75, + "learning_rate": 2.0818158502180046e-05, + "loss": 0.2454, + "step": 45562 + }, + { + "epoch": 1.75, + "learning_rate": 2.0817517312131314e-05, + "loss": 0.3395, + "step": 45563 + }, + { + "epoch": 1.75, + "learning_rate": 2.081687612208259e-05, + "loss": 0.3801, + "step": 45564 + }, + { + "epoch": 1.75, + "learning_rate": 2.0816234932033858e-05, + "loss": 0.3197, + "step": 45565 + }, + { + "epoch": 1.75, + "learning_rate": 2.0815593741985126e-05, + "loss": 0.3184, + "step": 45566 + }, + { + "epoch": 1.75, + "learning_rate": 2.0814952551936394e-05, + "loss": 0.3266, + "step": 45567 + }, + { + "epoch": 1.75, + "learning_rate": 2.0814311361887666e-05, + "loss": 0.3013, + "step": 45568 + }, + { + "epoch": 1.75, + "learning_rate": 2.0813670171838934e-05, + "loss": 0.326, + "step": 45569 + }, + { + "epoch": 1.75, + "learning_rate": 2.0813028981790203e-05, + "loss": 0.2934, + "step": 45570 + }, + { + "epoch": 1.75, + "learning_rate": 2.081238779174147e-05, + "loss": 0.3446, + "step": 45571 + }, + { + "epoch": 1.75, + "learning_rate": 2.0811746601692743e-05, + "loss": 0.3932, + "step": 45572 + }, + { + "epoch": 1.75, + "learning_rate": 2.0811105411644014e-05, + "loss": 0.2681, + "step": 45573 + }, + { + "epoch": 1.75, + "learning_rate": 2.0810464221595283e-05, + "loss": 0.3067, + "step": 45574 + }, + { + "epoch": 1.75, + "learning_rate": 2.080982303154655e-05, + "loss": 0.4325, + "step": 45575 + }, + { + "epoch": 1.75, + "learning_rate": 2.0809181841497823e-05, + "loss": 0.2601, + "step": 45576 + }, + { + "epoch": 1.75, + "learning_rate": 2.080854065144909e-05, + "loss": 0.4155, + "step": 45577 + }, + { + "epoch": 1.75, + "learning_rate": 2.080789946140036e-05, + "loss": 0.3703, + "step": 45578 + }, + { + "epoch": 1.75, + "learning_rate": 2.0807258271351628e-05, + "loss": 0.3513, + "step": 45579 + }, + { + "epoch": 1.75, + "learning_rate": 2.08066170813029e-05, + "loss": 0.3215, + "step": 45580 + }, + { + "epoch": 1.75, + "learning_rate": 2.0805975891254168e-05, + "loss": 0.2901, + "step": 45581 + }, + { + "epoch": 1.75, + "learning_rate": 2.080533470120544e-05, + "loss": 0.3197, + "step": 45582 + }, + { + "epoch": 1.75, + "learning_rate": 2.0804693511156708e-05, + "loss": 0.3604, + "step": 45583 + }, + { + "epoch": 1.75, + "learning_rate": 2.080405232110798e-05, + "loss": 0.2907, + "step": 45584 + }, + { + "epoch": 1.75, + "learning_rate": 2.0803411131059248e-05, + "loss": 0.2528, + "step": 45585 + }, + { + "epoch": 1.75, + "learning_rate": 2.0802769941010516e-05, + "loss": 0.3864, + "step": 45586 + }, + { + "epoch": 1.75, + "learning_rate": 2.0802128750961784e-05, + "loss": 0.3932, + "step": 45587 + }, + { + "epoch": 1.75, + "learning_rate": 2.0801487560913056e-05, + "loss": 0.4341, + "step": 45588 + }, + { + "epoch": 1.75, + "learning_rate": 2.0800846370864325e-05, + "loss": 0.4035, + "step": 45589 + }, + { + "epoch": 1.75, + "learning_rate": 2.0800205180815593e-05, + "loss": 0.3937, + "step": 45590 + }, + { + "epoch": 1.75, + "learning_rate": 2.0799563990766865e-05, + "loss": 0.3555, + "step": 45591 + }, + { + "epoch": 1.75, + "learning_rate": 2.0798922800718136e-05, + "loss": 0.3808, + "step": 45592 + }, + { + "epoch": 1.75, + "learning_rate": 2.0798281610669405e-05, + "loss": 0.2875, + "step": 45593 + }, + { + "epoch": 1.75, + "learning_rate": 2.0797640420620673e-05, + "loss": 0.327, + "step": 45594 + }, + { + "epoch": 1.75, + "learning_rate": 2.079699923057194e-05, + "loss": 0.2563, + "step": 45595 + }, + { + "epoch": 1.75, + "learning_rate": 2.0796358040523213e-05, + "loss": 0.2688, + "step": 45596 + }, + { + "epoch": 1.75, + "learning_rate": 2.079571685047448e-05, + "loss": 0.3317, + "step": 45597 + }, + { + "epoch": 1.75, + "learning_rate": 2.079507566042575e-05, + "loss": 0.2635, + "step": 45598 + }, + { + "epoch": 1.75, + "learning_rate": 2.0794434470377018e-05, + "loss": 0.3436, + "step": 45599 + }, + { + "epoch": 1.75, + "learning_rate": 2.0793793280328293e-05, + "loss": 0.2429, + "step": 45600 + }, + { + "epoch": 1.75, + "learning_rate": 2.079315209027956e-05, + "loss": 0.4193, + "step": 45601 + }, + { + "epoch": 1.75, + "learning_rate": 2.079251090023083e-05, + "loss": 0.251, + "step": 45602 + }, + { + "epoch": 1.75, + "learning_rate": 2.0791869710182098e-05, + "loss": 0.4381, + "step": 45603 + }, + { + "epoch": 1.75, + "learning_rate": 2.079122852013337e-05, + "loss": 0.4361, + "step": 45604 + }, + { + "epoch": 1.75, + "learning_rate": 2.0790587330084638e-05, + "loss": 0.4669, + "step": 45605 + }, + { + "epoch": 1.75, + "learning_rate": 2.0789946140035906e-05, + "loss": 0.3155, + "step": 45606 + }, + { + "epoch": 1.75, + "learning_rate": 2.0789304949987175e-05, + "loss": 0.2581, + "step": 45607 + }, + { + "epoch": 1.75, + "learning_rate": 2.0788663759938446e-05, + "loss": 0.3719, + "step": 45608 + }, + { + "epoch": 1.75, + "learning_rate": 2.0788022569889718e-05, + "loss": 0.4119, + "step": 45609 + }, + { + "epoch": 1.75, + "learning_rate": 2.0787381379840987e-05, + "loss": 0.4181, + "step": 45610 + }, + { + "epoch": 1.75, + "learning_rate": 2.0786740189792255e-05, + "loss": 0.333, + "step": 45611 + }, + { + "epoch": 1.75, + "learning_rate": 2.0786098999743527e-05, + "loss": 0.3568, + "step": 45612 + }, + { + "epoch": 1.75, + "learning_rate": 2.0785457809694795e-05, + "loss": 0.4043, + "step": 45613 + }, + { + "epoch": 1.75, + "learning_rate": 2.0784816619646063e-05, + "loss": 0.5588, + "step": 45614 + }, + { + "epoch": 1.75, + "learning_rate": 2.078417542959733e-05, + "loss": 0.3917, + "step": 45615 + }, + { + "epoch": 1.75, + "learning_rate": 2.0783534239548603e-05, + "loss": 0.3164, + "step": 45616 + }, + { + "epoch": 1.75, + "learning_rate": 2.078289304949987e-05, + "loss": 0.2279, + "step": 45617 + }, + { + "epoch": 1.75, + "learning_rate": 2.0782251859451143e-05, + "loss": 0.3922, + "step": 45618 + }, + { + "epoch": 1.75, + "learning_rate": 2.078161066940241e-05, + "loss": 0.3828, + "step": 45619 + }, + { + "epoch": 1.75, + "learning_rate": 2.0780969479353683e-05, + "loss": 0.4134, + "step": 45620 + }, + { + "epoch": 1.75, + "learning_rate": 2.0780328289304952e-05, + "loss": 0.2212, + "step": 45621 + }, + { + "epoch": 1.75, + "learning_rate": 2.077968709925622e-05, + "loss": 0.328, + "step": 45622 + }, + { + "epoch": 1.75, + "learning_rate": 2.077904590920749e-05, + "loss": 0.3354, + "step": 45623 + }, + { + "epoch": 1.75, + "learning_rate": 2.077840471915876e-05, + "loss": 0.2824, + "step": 45624 + }, + { + "epoch": 1.75, + "learning_rate": 2.077776352911003e-05, + "loss": 0.1797, + "step": 45625 + }, + { + "epoch": 1.75, + "learning_rate": 2.07771223390613e-05, + "loss": 0.2266, + "step": 45626 + }, + { + "epoch": 1.75, + "learning_rate": 2.077648114901257e-05, + "loss": 0.3184, + "step": 45627 + }, + { + "epoch": 1.75, + "learning_rate": 2.077583995896384e-05, + "loss": 0.2432, + "step": 45628 + }, + { + "epoch": 1.75, + "learning_rate": 2.077519876891511e-05, + "loss": 0.3025, + "step": 45629 + }, + { + "epoch": 1.75, + "learning_rate": 2.0774557578866377e-05, + "loss": 0.2965, + "step": 45630 + }, + { + "epoch": 1.75, + "learning_rate": 2.0773916388817645e-05, + "loss": 0.258, + "step": 45631 + }, + { + "epoch": 1.75, + "learning_rate": 2.0773275198768917e-05, + "loss": 0.3255, + "step": 45632 + }, + { + "epoch": 1.75, + "learning_rate": 2.0772634008720185e-05, + "loss": 0.2665, + "step": 45633 + }, + { + "epoch": 1.75, + "learning_rate": 2.0771992818671454e-05, + "loss": 0.4091, + "step": 45634 + }, + { + "epoch": 1.75, + "learning_rate": 2.0771351628622725e-05, + "loss": 0.4382, + "step": 45635 + }, + { + "epoch": 1.75, + "learning_rate": 2.0770710438573997e-05, + "loss": 0.3888, + "step": 45636 + }, + { + "epoch": 1.75, + "learning_rate": 2.0770069248525265e-05, + "loss": 0.363, + "step": 45637 + }, + { + "epoch": 1.75, + "learning_rate": 2.0769428058476534e-05, + "loss": 0.4052, + "step": 45638 + }, + { + "epoch": 1.75, + "learning_rate": 2.0768786868427802e-05, + "loss": 0.4454, + "step": 45639 + }, + { + "epoch": 1.75, + "learning_rate": 2.0768145678379074e-05, + "loss": 0.231, + "step": 45640 + }, + { + "epoch": 1.75, + "learning_rate": 2.0767504488330342e-05, + "loss": 0.3621, + "step": 45641 + }, + { + "epoch": 1.75, + "learning_rate": 2.076686329828161e-05, + "loss": 0.1647, + "step": 45642 + }, + { + "epoch": 1.75, + "learning_rate": 2.076622210823288e-05, + "loss": 0.3336, + "step": 45643 + }, + { + "epoch": 1.75, + "learning_rate": 2.0765580918184154e-05, + "loss": 0.2851, + "step": 45644 + }, + { + "epoch": 1.75, + "learning_rate": 2.0764939728135422e-05, + "loss": 0.5414, + "step": 45645 + }, + { + "epoch": 1.75, + "learning_rate": 2.076429853808669e-05, + "loss": 0.4288, + "step": 45646 + }, + { + "epoch": 1.75, + "learning_rate": 2.076365734803796e-05, + "loss": 0.2828, + "step": 45647 + }, + { + "epoch": 1.76, + "learning_rate": 2.076301615798923e-05, + "loss": 0.4749, + "step": 45648 + }, + { + "epoch": 1.76, + "learning_rate": 2.07623749679405e-05, + "loss": 0.2653, + "step": 45649 + }, + { + "epoch": 1.76, + "learning_rate": 2.0761733777891767e-05, + "loss": 0.2389, + "step": 45650 + }, + { + "epoch": 1.76, + "learning_rate": 2.0761092587843035e-05, + "loss": 0.2813, + "step": 45651 + }, + { + "epoch": 1.76, + "learning_rate": 2.0760451397794307e-05, + "loss": 0.3183, + "step": 45652 + }, + { + "epoch": 1.76, + "learning_rate": 2.075981020774558e-05, + "loss": 0.1674, + "step": 45653 + }, + { + "epoch": 1.76, + "learning_rate": 2.0759169017696847e-05, + "loss": 0.2774, + "step": 45654 + }, + { + "epoch": 1.76, + "learning_rate": 2.0758527827648116e-05, + "loss": 0.4171, + "step": 45655 + }, + { + "epoch": 1.76, + "learning_rate": 2.0757886637599387e-05, + "loss": 0.3489, + "step": 45656 + }, + { + "epoch": 1.76, + "learning_rate": 2.0757245447550656e-05, + "loss": 0.4661, + "step": 45657 + }, + { + "epoch": 1.76, + "learning_rate": 2.0756604257501924e-05, + "loss": 0.3686, + "step": 45658 + }, + { + "epoch": 1.76, + "learning_rate": 2.0755963067453192e-05, + "loss": 0.3087, + "step": 45659 + }, + { + "epoch": 1.76, + "learning_rate": 2.0755321877404464e-05, + "loss": 0.395, + "step": 45660 + }, + { + "epoch": 1.76, + "learning_rate": 2.0754680687355732e-05, + "loss": 0.2873, + "step": 45661 + }, + { + "epoch": 1.76, + "learning_rate": 2.0754039497307004e-05, + "loss": 0.3022, + "step": 45662 + }, + { + "epoch": 1.76, + "learning_rate": 2.0753398307258272e-05, + "loss": 0.2352, + "step": 45663 + }, + { + "epoch": 1.76, + "learning_rate": 2.0752757117209544e-05, + "loss": 0.2311, + "step": 45664 + }, + { + "epoch": 1.76, + "learning_rate": 2.0752115927160812e-05, + "loss": 0.4019, + "step": 45665 + }, + { + "epoch": 1.76, + "learning_rate": 2.075147473711208e-05, + "loss": 0.3458, + "step": 45666 + }, + { + "epoch": 1.76, + "learning_rate": 2.075083354706335e-05, + "loss": 0.4124, + "step": 45667 + }, + { + "epoch": 1.76, + "learning_rate": 2.075019235701462e-05, + "loss": 0.2412, + "step": 45668 + }, + { + "epoch": 1.76, + "learning_rate": 2.074955116696589e-05, + "loss": 0.3281, + "step": 45669 + }, + { + "epoch": 1.76, + "learning_rate": 2.0748909976917157e-05, + "loss": 0.3009, + "step": 45670 + }, + { + "epoch": 1.76, + "learning_rate": 2.074826878686843e-05, + "loss": 0.2807, + "step": 45671 + }, + { + "epoch": 1.76, + "learning_rate": 2.07476275968197e-05, + "loss": 0.4065, + "step": 45672 + }, + { + "epoch": 1.76, + "learning_rate": 2.074698640677097e-05, + "loss": 0.3072, + "step": 45673 + }, + { + "epoch": 1.76, + "learning_rate": 2.0746345216722237e-05, + "loss": 0.282, + "step": 45674 + }, + { + "epoch": 1.76, + "learning_rate": 2.0745704026673506e-05, + "loss": 0.4792, + "step": 45675 + }, + { + "epoch": 1.76, + "learning_rate": 2.0745062836624778e-05, + "loss": 0.3288, + "step": 45676 + }, + { + "epoch": 1.76, + "learning_rate": 2.0744421646576046e-05, + "loss": 0.587, + "step": 45677 + }, + { + "epoch": 1.76, + "learning_rate": 2.0743780456527314e-05, + "loss": 0.2935, + "step": 45678 + }, + { + "epoch": 1.76, + "learning_rate": 2.0743139266478583e-05, + "loss": 0.2252, + "step": 45679 + }, + { + "epoch": 1.76, + "learning_rate": 2.0742498076429858e-05, + "loss": 0.3624, + "step": 45680 + }, + { + "epoch": 1.76, + "learning_rate": 2.0741856886381126e-05, + "loss": 0.2971, + "step": 45681 + }, + { + "epoch": 1.76, + "learning_rate": 2.0741215696332394e-05, + "loss": 0.313, + "step": 45682 + }, + { + "epoch": 1.76, + "learning_rate": 2.0740574506283663e-05, + "loss": 0.2767, + "step": 45683 + }, + { + "epoch": 1.76, + "learning_rate": 2.0739933316234934e-05, + "loss": 0.2338, + "step": 45684 + }, + { + "epoch": 1.76, + "learning_rate": 2.0739292126186203e-05, + "loss": 0.3169, + "step": 45685 + }, + { + "epoch": 1.76, + "learning_rate": 2.073865093613747e-05, + "loss": 0.2766, + "step": 45686 + }, + { + "epoch": 1.76, + "learning_rate": 2.073800974608874e-05, + "loss": 0.319, + "step": 45687 + }, + { + "epoch": 1.76, + "learning_rate": 2.073736855604001e-05, + "loss": 0.2803, + "step": 45688 + }, + { + "epoch": 1.76, + "learning_rate": 2.0736727365991283e-05, + "loss": 0.2686, + "step": 45689 + }, + { + "epoch": 1.76, + "learning_rate": 2.073608617594255e-05, + "loss": 0.3271, + "step": 45690 + }, + { + "epoch": 1.76, + "learning_rate": 2.073544498589382e-05, + "loss": 0.2904, + "step": 45691 + }, + { + "epoch": 1.76, + "learning_rate": 2.073480379584509e-05, + "loss": 0.3074, + "step": 45692 + }, + { + "epoch": 1.76, + "learning_rate": 2.073416260579636e-05, + "loss": 0.2504, + "step": 45693 + }, + { + "epoch": 1.76, + "learning_rate": 2.0733521415747628e-05, + "loss": 0.3375, + "step": 45694 + }, + { + "epoch": 1.76, + "learning_rate": 2.0732880225698896e-05, + "loss": 0.4095, + "step": 45695 + }, + { + "epoch": 1.76, + "learning_rate": 2.0732239035650168e-05, + "loss": 0.3296, + "step": 45696 + }, + { + "epoch": 1.76, + "learning_rate": 2.0731597845601436e-05, + "loss": 0.37, + "step": 45697 + }, + { + "epoch": 1.76, + "learning_rate": 2.0730956655552708e-05, + "loss": 0.3398, + "step": 45698 + }, + { + "epoch": 1.76, + "learning_rate": 2.0730315465503976e-05, + "loss": 0.2863, + "step": 45699 + }, + { + "epoch": 1.76, + "learning_rate": 2.0729674275455248e-05, + "loss": 0.2879, + "step": 45700 + }, + { + "epoch": 1.76, + "learning_rate": 2.0729033085406516e-05, + "loss": 0.3744, + "step": 45701 + }, + { + "epoch": 1.76, + "learning_rate": 2.0728391895357785e-05, + "loss": 0.3164, + "step": 45702 + }, + { + "epoch": 1.76, + "learning_rate": 2.0727750705309053e-05, + "loss": 0.3131, + "step": 45703 + }, + { + "epoch": 1.76, + "learning_rate": 2.0727109515260325e-05, + "loss": 0.317, + "step": 45704 + }, + { + "epoch": 1.76, + "learning_rate": 2.0726468325211593e-05, + "loss": 0.3418, + "step": 45705 + }, + { + "epoch": 1.76, + "learning_rate": 2.072582713516286e-05, + "loss": 0.4002, + "step": 45706 + }, + { + "epoch": 1.76, + "learning_rate": 2.0725185945114133e-05, + "loss": 0.3223, + "step": 45707 + }, + { + "epoch": 1.76, + "learning_rate": 2.0724544755065405e-05, + "loss": 0.3063, + "step": 45708 + }, + { + "epoch": 1.76, + "learning_rate": 2.0723903565016673e-05, + "loss": 0.3297, + "step": 45709 + }, + { + "epoch": 1.76, + "learning_rate": 2.072326237496794e-05, + "loss": 0.4664, + "step": 45710 + }, + { + "epoch": 1.76, + "learning_rate": 2.072262118491921e-05, + "loss": 0.2366, + "step": 45711 + }, + { + "epoch": 1.76, + "learning_rate": 2.072197999487048e-05, + "loss": 0.4779, + "step": 45712 + }, + { + "epoch": 1.76, + "learning_rate": 2.072133880482175e-05, + "loss": 0.3283, + "step": 45713 + }, + { + "epoch": 1.76, + "learning_rate": 2.0720697614773018e-05, + "loss": 0.4388, + "step": 45714 + }, + { + "epoch": 1.76, + "learning_rate": 2.072005642472429e-05, + "loss": 0.3181, + "step": 45715 + }, + { + "epoch": 1.76, + "learning_rate": 2.071941523467556e-05, + "loss": 0.2802, + "step": 45716 + }, + { + "epoch": 1.76, + "learning_rate": 2.071877404462683e-05, + "loss": 0.271, + "step": 45717 + }, + { + "epoch": 1.76, + "learning_rate": 2.0718132854578098e-05, + "loss": 0.2396, + "step": 45718 + }, + { + "epoch": 1.76, + "learning_rate": 2.0717491664529366e-05, + "loss": 0.368, + "step": 45719 + }, + { + "epoch": 1.76, + "learning_rate": 2.0716850474480638e-05, + "loss": 0.2312, + "step": 45720 + }, + { + "epoch": 1.76, + "learning_rate": 2.0716209284431907e-05, + "loss": 0.2418, + "step": 45721 + }, + { + "epoch": 1.76, + "learning_rate": 2.0715568094383175e-05, + "loss": 0.3304, + "step": 45722 + }, + { + "epoch": 1.76, + "learning_rate": 2.0714926904334443e-05, + "loss": 0.3602, + "step": 45723 + }, + { + "epoch": 1.76, + "learning_rate": 2.0714285714285718e-05, + "loss": 0.262, + "step": 45724 + }, + { + "epoch": 1.76, + "learning_rate": 2.0713644524236987e-05, + "loss": 0.2247, + "step": 45725 + }, + { + "epoch": 1.76, + "learning_rate": 2.0713003334188255e-05, + "loss": 0.336, + "step": 45726 + }, + { + "epoch": 1.76, + "learning_rate": 2.0712362144139523e-05, + "loss": 0.4986, + "step": 45727 + }, + { + "epoch": 1.76, + "learning_rate": 2.0711720954090795e-05, + "loss": 0.3814, + "step": 45728 + }, + { + "epoch": 1.76, + "learning_rate": 2.0711079764042063e-05, + "loss": 0.257, + "step": 45729 + }, + { + "epoch": 1.76, + "learning_rate": 2.071043857399333e-05, + "loss": 0.3391, + "step": 45730 + }, + { + "epoch": 1.76, + "learning_rate": 2.07097973839446e-05, + "loss": 0.2509, + "step": 45731 + }, + { + "epoch": 1.76, + "learning_rate": 2.070915619389587e-05, + "loss": 0.2461, + "step": 45732 + }, + { + "epoch": 1.76, + "learning_rate": 2.0708515003847143e-05, + "loss": 0.3251, + "step": 45733 + }, + { + "epoch": 1.76, + "learning_rate": 2.0707873813798412e-05, + "loss": 0.2454, + "step": 45734 + }, + { + "epoch": 1.76, + "learning_rate": 2.070723262374968e-05, + "loss": 0.3736, + "step": 45735 + }, + { + "epoch": 1.76, + "learning_rate": 2.0706591433700952e-05, + "loss": 0.3779, + "step": 45736 + }, + { + "epoch": 1.76, + "learning_rate": 2.070595024365222e-05, + "loss": 0.3815, + "step": 45737 + }, + { + "epoch": 1.76, + "learning_rate": 2.070530905360349e-05, + "loss": 0.259, + "step": 45738 + }, + { + "epoch": 1.76, + "learning_rate": 2.0704667863554757e-05, + "loss": 0.2396, + "step": 45739 + }, + { + "epoch": 1.76, + "learning_rate": 2.070402667350603e-05, + "loss": 0.3867, + "step": 45740 + }, + { + "epoch": 1.76, + "learning_rate": 2.0703385483457297e-05, + "loss": 0.251, + "step": 45741 + }, + { + "epoch": 1.76, + "learning_rate": 2.070274429340857e-05, + "loss": 0.2503, + "step": 45742 + }, + { + "epoch": 1.76, + "learning_rate": 2.0702103103359837e-05, + "loss": 0.4475, + "step": 45743 + }, + { + "epoch": 1.76, + "learning_rate": 2.070146191331111e-05, + "loss": 0.3206, + "step": 45744 + }, + { + "epoch": 1.76, + "learning_rate": 2.0700820723262377e-05, + "loss": 0.2633, + "step": 45745 + }, + { + "epoch": 1.76, + "learning_rate": 2.0700179533213645e-05, + "loss": 0.5485, + "step": 45746 + }, + { + "epoch": 1.76, + "learning_rate": 2.0699538343164914e-05, + "loss": 0.2789, + "step": 45747 + }, + { + "epoch": 1.76, + "learning_rate": 2.0698897153116185e-05, + "loss": 0.3431, + "step": 45748 + }, + { + "epoch": 1.76, + "learning_rate": 2.0698255963067454e-05, + "loss": 0.375, + "step": 45749 + }, + { + "epoch": 1.76, + "learning_rate": 2.0697614773018722e-05, + "loss": 0.3861, + "step": 45750 + }, + { + "epoch": 1.76, + "learning_rate": 2.0696973582969994e-05, + "loss": 0.24, + "step": 45751 + }, + { + "epoch": 1.76, + "learning_rate": 2.0696332392921265e-05, + "loss": 0.2794, + "step": 45752 + }, + { + "epoch": 1.76, + "learning_rate": 2.0695691202872534e-05, + "loss": 0.3271, + "step": 45753 + }, + { + "epoch": 1.76, + "learning_rate": 2.0695050012823802e-05, + "loss": 0.4117, + "step": 45754 + }, + { + "epoch": 1.76, + "learning_rate": 2.069440882277507e-05, + "loss": 0.4945, + "step": 45755 + }, + { + "epoch": 1.76, + "learning_rate": 2.0693767632726342e-05, + "loss": 0.3795, + "step": 45756 + }, + { + "epoch": 1.76, + "learning_rate": 2.069312644267761e-05, + "loss": 0.334, + "step": 45757 + }, + { + "epoch": 1.76, + "learning_rate": 2.069248525262888e-05, + "loss": 0.3394, + "step": 45758 + }, + { + "epoch": 1.76, + "learning_rate": 2.0691844062580147e-05, + "loss": 0.406, + "step": 45759 + }, + { + "epoch": 1.76, + "learning_rate": 2.0691202872531422e-05, + "loss": 0.3053, + "step": 45760 + }, + { + "epoch": 1.76, + "learning_rate": 2.069056168248269e-05, + "loss": 0.3594, + "step": 45761 + }, + { + "epoch": 1.76, + "learning_rate": 2.068992049243396e-05, + "loss": 0.3551, + "step": 45762 + }, + { + "epoch": 1.76, + "learning_rate": 2.0689279302385227e-05, + "loss": 0.3416, + "step": 45763 + }, + { + "epoch": 1.76, + "learning_rate": 2.06886381123365e-05, + "loss": 0.2572, + "step": 45764 + }, + { + "epoch": 1.76, + "learning_rate": 2.0687996922287767e-05, + "loss": 0.32, + "step": 45765 + }, + { + "epoch": 1.76, + "learning_rate": 2.0687355732239036e-05, + "loss": 0.3689, + "step": 45766 + }, + { + "epoch": 1.76, + "learning_rate": 2.0686714542190304e-05, + "loss": 0.3683, + "step": 45767 + }, + { + "epoch": 1.76, + "learning_rate": 2.0686073352141576e-05, + "loss": 0.3472, + "step": 45768 + }, + { + "epoch": 1.76, + "learning_rate": 2.0685432162092847e-05, + "loss": 0.3548, + "step": 45769 + }, + { + "epoch": 1.76, + "learning_rate": 2.0684790972044116e-05, + "loss": 0.576, + "step": 45770 + }, + { + "epoch": 1.76, + "learning_rate": 2.0684149781995384e-05, + "loss": 0.4206, + "step": 45771 + }, + { + "epoch": 1.76, + "learning_rate": 2.0683508591946656e-05, + "loss": 0.4443, + "step": 45772 + }, + { + "epoch": 1.76, + "learning_rate": 2.0682867401897924e-05, + "loss": 0.2861, + "step": 45773 + }, + { + "epoch": 1.76, + "learning_rate": 2.0682226211849192e-05, + "loss": 0.3144, + "step": 45774 + }, + { + "epoch": 1.76, + "learning_rate": 2.068158502180046e-05, + "loss": 0.2514, + "step": 45775 + }, + { + "epoch": 1.76, + "learning_rate": 2.0680943831751732e-05, + "loss": 0.3142, + "step": 45776 + }, + { + "epoch": 1.76, + "learning_rate": 2.0680302641703e-05, + "loss": 0.2543, + "step": 45777 + }, + { + "epoch": 1.76, + "learning_rate": 2.0679661451654272e-05, + "loss": 0.3, + "step": 45778 + }, + { + "epoch": 1.76, + "learning_rate": 2.067902026160554e-05, + "loss": 0.3099, + "step": 45779 + }, + { + "epoch": 1.76, + "learning_rate": 2.0678379071556812e-05, + "loss": 0.2308, + "step": 45780 + }, + { + "epoch": 1.76, + "learning_rate": 2.067773788150808e-05, + "loss": 0.3803, + "step": 45781 + }, + { + "epoch": 1.76, + "learning_rate": 2.067709669145935e-05, + "loss": 0.4728, + "step": 45782 + }, + { + "epoch": 1.76, + "learning_rate": 2.0676455501410617e-05, + "loss": 0.3234, + "step": 45783 + }, + { + "epoch": 1.76, + "learning_rate": 2.067581431136189e-05, + "loss": 0.2437, + "step": 45784 + }, + { + "epoch": 1.76, + "learning_rate": 2.0675173121313157e-05, + "loss": 0.269, + "step": 45785 + }, + { + "epoch": 1.76, + "learning_rate": 2.0674531931264426e-05, + "loss": 0.3714, + "step": 45786 + }, + { + "epoch": 1.76, + "learning_rate": 2.0673890741215698e-05, + "loss": 0.3661, + "step": 45787 + }, + { + "epoch": 1.76, + "learning_rate": 2.067324955116697e-05, + "loss": 0.4214, + "step": 45788 + }, + { + "epoch": 1.76, + "learning_rate": 2.0672608361118238e-05, + "loss": 0.2278, + "step": 45789 + }, + { + "epoch": 1.76, + "learning_rate": 2.0671967171069506e-05, + "loss": 0.4026, + "step": 45790 + }, + { + "epoch": 1.76, + "learning_rate": 2.0671325981020774e-05, + "loss": 0.4703, + "step": 45791 + }, + { + "epoch": 1.76, + "learning_rate": 2.0670684790972046e-05, + "loss": 0.2934, + "step": 45792 + }, + { + "epoch": 1.76, + "learning_rate": 2.0670043600923314e-05, + "loss": 0.4034, + "step": 45793 + }, + { + "epoch": 1.76, + "learning_rate": 2.0669402410874583e-05, + "loss": 0.4438, + "step": 45794 + }, + { + "epoch": 1.76, + "learning_rate": 2.066876122082585e-05, + "loss": 0.428, + "step": 45795 + }, + { + "epoch": 1.76, + "learning_rate": 2.0668120030777126e-05, + "loss": 0.4158, + "step": 45796 + }, + { + "epoch": 1.76, + "learning_rate": 2.0667478840728394e-05, + "loss": 0.2514, + "step": 45797 + }, + { + "epoch": 1.76, + "learning_rate": 2.0666837650679663e-05, + "loss": 0.314, + "step": 45798 + }, + { + "epoch": 1.76, + "learning_rate": 2.066619646063093e-05, + "loss": 0.2461, + "step": 45799 + }, + { + "epoch": 1.76, + "learning_rate": 2.0665555270582203e-05, + "loss": 0.5116, + "step": 45800 + }, + { + "epoch": 1.76, + "learning_rate": 2.066491408053347e-05, + "loss": 0.443, + "step": 45801 + }, + { + "epoch": 1.76, + "learning_rate": 2.066427289048474e-05, + "loss": 0.3976, + "step": 45802 + }, + { + "epoch": 1.76, + "learning_rate": 2.0663631700436008e-05, + "loss": 0.331, + "step": 45803 + }, + { + "epoch": 1.76, + "learning_rate": 2.066299051038728e-05, + "loss": 0.3014, + "step": 45804 + }, + { + "epoch": 1.76, + "learning_rate": 2.066234932033855e-05, + "loss": 0.2737, + "step": 45805 + }, + { + "epoch": 1.76, + "learning_rate": 2.066170813028982e-05, + "loss": 0.4708, + "step": 45806 + }, + { + "epoch": 1.76, + "learning_rate": 2.0661066940241088e-05, + "loss": 0.2986, + "step": 45807 + }, + { + "epoch": 1.76, + "learning_rate": 2.066042575019236e-05, + "loss": 0.4274, + "step": 45808 + }, + { + "epoch": 1.76, + "learning_rate": 2.0659784560143628e-05, + "loss": 0.3693, + "step": 45809 + }, + { + "epoch": 1.76, + "learning_rate": 2.0659143370094896e-05, + "loss": 0.3459, + "step": 45810 + }, + { + "epoch": 1.76, + "learning_rate": 2.0658502180046165e-05, + "loss": 0.4779, + "step": 45811 + }, + { + "epoch": 1.76, + "learning_rate": 2.0657860989997436e-05, + "loss": 0.3706, + "step": 45812 + }, + { + "epoch": 1.76, + "learning_rate": 2.0657219799948708e-05, + "loss": 0.2528, + "step": 45813 + }, + { + "epoch": 1.76, + "learning_rate": 2.0656578609899976e-05, + "loss": 0.3605, + "step": 45814 + }, + { + "epoch": 1.76, + "learning_rate": 2.0655937419851245e-05, + "loss": 0.3339, + "step": 45815 + }, + { + "epoch": 1.76, + "learning_rate": 2.0655296229802516e-05, + "loss": 0.3585, + "step": 45816 + }, + { + "epoch": 1.76, + "learning_rate": 2.0654655039753785e-05, + "loss": 0.3328, + "step": 45817 + }, + { + "epoch": 1.76, + "learning_rate": 2.0654013849705053e-05, + "loss": 0.3134, + "step": 45818 + }, + { + "epoch": 1.76, + "learning_rate": 2.065337265965632e-05, + "loss": 0.2963, + "step": 45819 + }, + { + "epoch": 1.76, + "learning_rate": 2.0652731469607593e-05, + "loss": 0.2567, + "step": 45820 + }, + { + "epoch": 1.76, + "learning_rate": 2.065209027955886e-05, + "loss": 0.4024, + "step": 45821 + }, + { + "epoch": 1.76, + "learning_rate": 2.0651449089510133e-05, + "loss": 0.2916, + "step": 45822 + }, + { + "epoch": 1.76, + "learning_rate": 2.06508078994614e-05, + "loss": 0.3822, + "step": 45823 + }, + { + "epoch": 1.76, + "learning_rate": 2.0650166709412673e-05, + "loss": 0.4609, + "step": 45824 + }, + { + "epoch": 1.76, + "learning_rate": 2.064952551936394e-05, + "loss": 0.3076, + "step": 45825 + }, + { + "epoch": 1.76, + "learning_rate": 2.064888432931521e-05, + "loss": 0.2746, + "step": 45826 + }, + { + "epoch": 1.76, + "learning_rate": 2.0648243139266478e-05, + "loss": 0.359, + "step": 45827 + }, + { + "epoch": 1.76, + "learning_rate": 2.064760194921775e-05, + "loss": 0.3204, + "step": 45828 + }, + { + "epoch": 1.76, + "learning_rate": 2.0646960759169018e-05, + "loss": 0.3151, + "step": 45829 + }, + { + "epoch": 1.76, + "learning_rate": 2.0646319569120286e-05, + "loss": 0.2949, + "step": 45830 + }, + { + "epoch": 1.76, + "learning_rate": 2.0645678379071558e-05, + "loss": 0.3293, + "step": 45831 + }, + { + "epoch": 1.76, + "learning_rate": 2.064503718902283e-05, + "loss": 0.322, + "step": 45832 + }, + { + "epoch": 1.76, + "learning_rate": 2.0644395998974098e-05, + "loss": 0.3906, + "step": 45833 + }, + { + "epoch": 1.76, + "learning_rate": 2.0643754808925367e-05, + "loss": 0.3695, + "step": 45834 + }, + { + "epoch": 1.76, + "learning_rate": 2.0643113618876635e-05, + "loss": 0.4139, + "step": 45835 + }, + { + "epoch": 1.76, + "learning_rate": 2.0642472428827907e-05, + "loss": 0.2931, + "step": 45836 + }, + { + "epoch": 1.76, + "learning_rate": 2.0641831238779175e-05, + "loss": 0.4075, + "step": 45837 + }, + { + "epoch": 1.76, + "learning_rate": 2.0641190048730443e-05, + "loss": 0.3392, + "step": 45838 + }, + { + "epoch": 1.76, + "learning_rate": 2.064054885868171e-05, + "loss": 0.3664, + "step": 45839 + }, + { + "epoch": 1.76, + "learning_rate": 2.0639907668632987e-05, + "loss": 0.3357, + "step": 45840 + }, + { + "epoch": 1.76, + "learning_rate": 2.0639266478584255e-05, + "loss": 0.2943, + "step": 45841 + }, + { + "epoch": 1.76, + "learning_rate": 2.0638625288535523e-05, + "loss": 0.4864, + "step": 45842 + }, + { + "epoch": 1.76, + "learning_rate": 2.063798409848679e-05, + "loss": 0.3863, + "step": 45843 + }, + { + "epoch": 1.76, + "learning_rate": 2.0637342908438063e-05, + "loss": 0.4213, + "step": 45844 + }, + { + "epoch": 1.76, + "learning_rate": 2.0636701718389332e-05, + "loss": 0.3145, + "step": 45845 + }, + { + "epoch": 1.76, + "learning_rate": 2.06360605283406e-05, + "loss": 0.3778, + "step": 45846 + }, + { + "epoch": 1.76, + "learning_rate": 2.063541933829187e-05, + "loss": 0.4251, + "step": 45847 + }, + { + "epoch": 1.76, + "learning_rate": 2.063477814824314e-05, + "loss": 0.2446, + "step": 45848 + }, + { + "epoch": 1.76, + "learning_rate": 2.0634136958194412e-05, + "loss": 0.27, + "step": 45849 + }, + { + "epoch": 1.76, + "learning_rate": 2.063349576814568e-05, + "loss": 0.4249, + "step": 45850 + }, + { + "epoch": 1.76, + "learning_rate": 2.063285457809695e-05, + "loss": 0.3574, + "step": 45851 + }, + { + "epoch": 1.76, + "learning_rate": 2.063221338804822e-05, + "loss": 0.279, + "step": 45852 + }, + { + "epoch": 1.76, + "learning_rate": 2.063157219799949e-05, + "loss": 0.434, + "step": 45853 + }, + { + "epoch": 1.76, + "learning_rate": 2.0630931007950757e-05, + "loss": 0.2924, + "step": 45854 + }, + { + "epoch": 1.76, + "learning_rate": 2.0630289817902025e-05, + "loss": 0.403, + "step": 45855 + }, + { + "epoch": 1.76, + "learning_rate": 2.0629648627853297e-05, + "loss": 0.2535, + "step": 45856 + }, + { + "epoch": 1.76, + "learning_rate": 2.0629007437804565e-05, + "loss": 0.4099, + "step": 45857 + }, + { + "epoch": 1.76, + "learning_rate": 2.0628366247755837e-05, + "loss": 0.4812, + "step": 45858 + }, + { + "epoch": 1.76, + "learning_rate": 2.0627725057707105e-05, + "loss": 0.5428, + "step": 45859 + }, + { + "epoch": 1.76, + "learning_rate": 2.0627083867658377e-05, + "loss": 0.3104, + "step": 45860 + }, + { + "epoch": 1.76, + "learning_rate": 2.0626442677609645e-05, + "loss": 0.3265, + "step": 45861 + }, + { + "epoch": 1.76, + "learning_rate": 2.0625801487560914e-05, + "loss": 0.3215, + "step": 45862 + }, + { + "epoch": 1.76, + "learning_rate": 2.0625160297512182e-05, + "loss": 0.4034, + "step": 45863 + }, + { + "epoch": 1.76, + "learning_rate": 2.0624519107463454e-05, + "loss": 0.2965, + "step": 45864 + }, + { + "epoch": 1.76, + "learning_rate": 2.0623877917414722e-05, + "loss": 0.461, + "step": 45865 + }, + { + "epoch": 1.76, + "learning_rate": 2.062323672736599e-05, + "loss": 0.3342, + "step": 45866 + }, + { + "epoch": 1.76, + "learning_rate": 2.0622595537317262e-05, + "loss": 0.4546, + "step": 45867 + }, + { + "epoch": 1.76, + "learning_rate": 2.0621954347268534e-05, + "loss": 0.312, + "step": 45868 + }, + { + "epoch": 1.76, + "learning_rate": 2.0621313157219802e-05, + "loss": 0.3348, + "step": 45869 + }, + { + "epoch": 1.76, + "learning_rate": 2.062067196717107e-05, + "loss": 0.2498, + "step": 45870 + }, + { + "epoch": 1.76, + "learning_rate": 2.062003077712234e-05, + "loss": 0.3413, + "step": 45871 + }, + { + "epoch": 1.76, + "learning_rate": 2.061938958707361e-05, + "loss": 0.4639, + "step": 45872 + }, + { + "epoch": 1.76, + "learning_rate": 2.061874839702488e-05, + "loss": 0.3415, + "step": 45873 + }, + { + "epoch": 1.76, + "learning_rate": 2.0618107206976147e-05, + "loss": 0.3138, + "step": 45874 + }, + { + "epoch": 1.76, + "learning_rate": 2.0617466016927415e-05, + "loss": 0.3128, + "step": 45875 + }, + { + "epoch": 1.76, + "learning_rate": 2.061682482687869e-05, + "loss": 0.4377, + "step": 45876 + }, + { + "epoch": 1.76, + "learning_rate": 2.061618363682996e-05, + "loss": 0.283, + "step": 45877 + }, + { + "epoch": 1.76, + "learning_rate": 2.0615542446781227e-05, + "loss": 0.3733, + "step": 45878 + }, + { + "epoch": 1.76, + "learning_rate": 2.0614901256732496e-05, + "loss": 0.3512, + "step": 45879 + }, + { + "epoch": 1.76, + "learning_rate": 2.0614260066683767e-05, + "loss": 0.2265, + "step": 45880 + }, + { + "epoch": 1.76, + "learning_rate": 2.0613618876635036e-05, + "loss": 0.3075, + "step": 45881 + }, + { + "epoch": 1.76, + "learning_rate": 2.0612977686586304e-05, + "loss": 0.3886, + "step": 45882 + }, + { + "epoch": 1.76, + "learning_rate": 2.0612336496537572e-05, + "loss": 0.3516, + "step": 45883 + }, + { + "epoch": 1.76, + "learning_rate": 2.0611695306488844e-05, + "loss": 0.2868, + "step": 45884 + }, + { + "epoch": 1.76, + "learning_rate": 2.0611054116440116e-05, + "loss": 0.3275, + "step": 45885 + }, + { + "epoch": 1.76, + "learning_rate": 2.0610412926391384e-05, + "loss": 0.2966, + "step": 45886 + }, + { + "epoch": 1.76, + "learning_rate": 2.0609771736342652e-05, + "loss": 0.2517, + "step": 45887 + }, + { + "epoch": 1.76, + "learning_rate": 2.0609130546293924e-05, + "loss": 0.2973, + "step": 45888 + }, + { + "epoch": 1.76, + "learning_rate": 2.0608489356245192e-05, + "loss": 0.2937, + "step": 45889 + }, + { + "epoch": 1.76, + "learning_rate": 2.060784816619646e-05, + "loss": 0.2828, + "step": 45890 + }, + { + "epoch": 1.76, + "learning_rate": 2.060720697614773e-05, + "loss": 0.3321, + "step": 45891 + }, + { + "epoch": 1.76, + "learning_rate": 2.0606565786099e-05, + "loss": 0.3585, + "step": 45892 + }, + { + "epoch": 1.76, + "learning_rate": 2.060592459605027e-05, + "loss": 0.4445, + "step": 45893 + }, + { + "epoch": 1.76, + "learning_rate": 2.060528340600154e-05, + "loss": 0.2506, + "step": 45894 + }, + { + "epoch": 1.76, + "learning_rate": 2.060464221595281e-05, + "loss": 0.4016, + "step": 45895 + }, + { + "epoch": 1.76, + "learning_rate": 2.060400102590408e-05, + "loss": 0.3706, + "step": 45896 + }, + { + "epoch": 1.76, + "learning_rate": 2.060335983585535e-05, + "loss": 0.2916, + "step": 45897 + }, + { + "epoch": 1.76, + "learning_rate": 2.0602718645806618e-05, + "loss": 0.2317, + "step": 45898 + }, + { + "epoch": 1.76, + "learning_rate": 2.0602077455757886e-05, + "loss": 0.3888, + "step": 45899 + }, + { + "epoch": 1.76, + "learning_rate": 2.0601436265709158e-05, + "loss": 0.5093, + "step": 45900 + }, + { + "epoch": 1.76, + "learning_rate": 2.0600795075660426e-05, + "loss": 0.4393, + "step": 45901 + }, + { + "epoch": 1.76, + "learning_rate": 2.0600153885611698e-05, + "loss": 0.473, + "step": 45902 + }, + { + "epoch": 1.76, + "learning_rate": 2.0599512695562966e-05, + "loss": 0.5627, + "step": 45903 + }, + { + "epoch": 1.76, + "learning_rate": 2.0598871505514238e-05, + "loss": 0.5729, + "step": 45904 + }, + { + "epoch": 1.76, + "learning_rate": 2.0598230315465506e-05, + "loss": 0.3849, + "step": 45905 + }, + { + "epoch": 1.76, + "learning_rate": 2.0597589125416774e-05, + "loss": 0.4367, + "step": 45906 + }, + { + "epoch": 1.76, + "learning_rate": 2.0596947935368043e-05, + "loss": 0.3257, + "step": 45907 + }, + { + "epoch": 1.77, + "learning_rate": 2.0596306745319314e-05, + "loss": 0.3857, + "step": 45908 + }, + { + "epoch": 1.77, + "learning_rate": 2.0595665555270583e-05, + "loss": 0.2713, + "step": 45909 + }, + { + "epoch": 1.77, + "learning_rate": 2.059502436522185e-05, + "loss": 0.374, + "step": 45910 + }, + { + "epoch": 1.77, + "learning_rate": 2.0594383175173123e-05, + "loss": 0.3325, + "step": 45911 + }, + { + "epoch": 1.77, + "learning_rate": 2.0593741985124394e-05, + "loss": 0.2959, + "step": 45912 + }, + { + "epoch": 1.77, + "learning_rate": 2.0593100795075663e-05, + "loss": 0.3154, + "step": 45913 + }, + { + "epoch": 1.77, + "learning_rate": 2.059245960502693e-05, + "loss": 0.4039, + "step": 45914 + }, + { + "epoch": 1.77, + "learning_rate": 2.05918184149782e-05, + "loss": 0.3766, + "step": 45915 + }, + { + "epoch": 1.77, + "learning_rate": 2.059117722492947e-05, + "loss": 0.5309, + "step": 45916 + }, + { + "epoch": 1.77, + "learning_rate": 2.059053603488074e-05, + "loss": 0.3354, + "step": 45917 + }, + { + "epoch": 1.77, + "learning_rate": 2.0589894844832008e-05, + "loss": 0.382, + "step": 45918 + }, + { + "epoch": 1.77, + "learning_rate": 2.0589253654783276e-05, + "loss": 0.3185, + "step": 45919 + }, + { + "epoch": 1.77, + "learning_rate": 2.058861246473455e-05, + "loss": 0.3863, + "step": 45920 + }, + { + "epoch": 1.77, + "learning_rate": 2.058797127468582e-05, + "loss": 0.3539, + "step": 45921 + }, + { + "epoch": 1.77, + "learning_rate": 2.0587330084637088e-05, + "loss": 0.289, + "step": 45922 + }, + { + "epoch": 1.77, + "learning_rate": 2.0586688894588356e-05, + "loss": 0.3683, + "step": 45923 + }, + { + "epoch": 1.77, + "learning_rate": 2.0586047704539628e-05, + "loss": 0.3143, + "step": 45924 + }, + { + "epoch": 1.77, + "learning_rate": 2.0585406514490896e-05, + "loss": 0.3404, + "step": 45925 + }, + { + "epoch": 1.77, + "learning_rate": 2.0584765324442165e-05, + "loss": 0.5914, + "step": 45926 + }, + { + "epoch": 1.77, + "learning_rate": 2.0584124134393433e-05, + "loss": 0.246, + "step": 45927 + }, + { + "epoch": 1.77, + "learning_rate": 2.0583482944344705e-05, + "loss": 0.39, + "step": 45928 + }, + { + "epoch": 1.77, + "learning_rate": 2.0582841754295976e-05, + "loss": 0.3481, + "step": 45929 + }, + { + "epoch": 1.77, + "learning_rate": 2.0582200564247245e-05, + "loss": 0.2934, + "step": 45930 + }, + { + "epoch": 1.77, + "learning_rate": 2.0581559374198513e-05, + "loss": 0.3377, + "step": 45931 + }, + { + "epoch": 1.77, + "learning_rate": 2.0580918184149785e-05, + "loss": 0.2644, + "step": 45932 + }, + { + "epoch": 1.77, + "learning_rate": 2.0580276994101053e-05, + "loss": 0.3403, + "step": 45933 + }, + { + "epoch": 1.77, + "learning_rate": 2.057963580405232e-05, + "loss": 0.3888, + "step": 45934 + }, + { + "epoch": 1.77, + "learning_rate": 2.057899461400359e-05, + "loss": 0.3393, + "step": 45935 + }, + { + "epoch": 1.77, + "learning_rate": 2.057835342395486e-05, + "loss": 0.2493, + "step": 45936 + }, + { + "epoch": 1.77, + "learning_rate": 2.057771223390613e-05, + "loss": 0.3733, + "step": 45937 + }, + { + "epoch": 1.77, + "learning_rate": 2.05770710438574e-05, + "loss": 0.3827, + "step": 45938 + }, + { + "epoch": 1.77, + "learning_rate": 2.057642985380867e-05, + "loss": 0.4266, + "step": 45939 + }, + { + "epoch": 1.77, + "learning_rate": 2.057578866375994e-05, + "loss": 0.3015, + "step": 45940 + }, + { + "epoch": 1.77, + "learning_rate": 2.057514747371121e-05, + "loss": 0.2849, + "step": 45941 + }, + { + "epoch": 1.77, + "learning_rate": 2.0574506283662478e-05, + "loss": 0.4116, + "step": 45942 + }, + { + "epoch": 1.77, + "learning_rate": 2.0573865093613746e-05, + "loss": 0.4722, + "step": 45943 + }, + { + "epoch": 1.77, + "learning_rate": 2.0573223903565018e-05, + "loss": 0.3979, + "step": 45944 + }, + { + "epoch": 1.77, + "learning_rate": 2.0572582713516287e-05, + "loss": 0.3734, + "step": 45945 + }, + { + "epoch": 1.77, + "learning_rate": 2.0571941523467555e-05, + "loss": 0.3311, + "step": 45946 + }, + { + "epoch": 1.77, + "learning_rate": 2.0571300333418827e-05, + "loss": 0.3996, + "step": 45947 + }, + { + "epoch": 1.77, + "learning_rate": 2.0570659143370098e-05, + "loss": 0.4481, + "step": 45948 + }, + { + "epoch": 1.77, + "learning_rate": 2.0570017953321367e-05, + "loss": 0.2748, + "step": 45949 + }, + { + "epoch": 1.77, + "learning_rate": 2.0569376763272635e-05, + "loss": 0.4195, + "step": 45950 + }, + { + "epoch": 1.77, + "learning_rate": 2.0568735573223903e-05, + "loss": 0.2661, + "step": 45951 + }, + { + "epoch": 1.77, + "learning_rate": 2.0568094383175175e-05, + "loss": 0.412, + "step": 45952 + }, + { + "epoch": 1.77, + "learning_rate": 2.0567453193126443e-05, + "loss": 0.3642, + "step": 45953 + }, + { + "epoch": 1.77, + "learning_rate": 2.056681200307771e-05, + "loss": 0.274, + "step": 45954 + }, + { + "epoch": 1.77, + "learning_rate": 2.056617081302898e-05, + "loss": 0.3786, + "step": 45955 + }, + { + "epoch": 1.77, + "learning_rate": 2.0565529622980255e-05, + "loss": 0.3701, + "step": 45956 + }, + { + "epoch": 1.77, + "learning_rate": 2.0564888432931523e-05, + "loss": 0.3987, + "step": 45957 + }, + { + "epoch": 1.77, + "learning_rate": 2.0564247242882792e-05, + "loss": 0.3854, + "step": 45958 + }, + { + "epoch": 1.77, + "learning_rate": 2.056360605283406e-05, + "loss": 0.27, + "step": 45959 + }, + { + "epoch": 1.77, + "learning_rate": 2.0562964862785332e-05, + "loss": 0.2811, + "step": 45960 + }, + { + "epoch": 1.77, + "learning_rate": 2.05623236727366e-05, + "loss": 0.3685, + "step": 45961 + }, + { + "epoch": 1.77, + "learning_rate": 2.056168248268787e-05, + "loss": 0.503, + "step": 45962 + }, + { + "epoch": 1.77, + "learning_rate": 2.0561041292639137e-05, + "loss": 0.3458, + "step": 45963 + }, + { + "epoch": 1.77, + "learning_rate": 2.056040010259041e-05, + "loss": 0.4441, + "step": 45964 + }, + { + "epoch": 1.77, + "learning_rate": 2.055975891254168e-05, + "loss": 0.3609, + "step": 45965 + }, + { + "epoch": 1.77, + "learning_rate": 2.055911772249295e-05, + "loss": 0.374, + "step": 45966 + }, + { + "epoch": 1.77, + "learning_rate": 2.0558476532444217e-05, + "loss": 0.2355, + "step": 45967 + }, + { + "epoch": 1.77, + "learning_rate": 2.055783534239549e-05, + "loss": 0.2905, + "step": 45968 + }, + { + "epoch": 1.77, + "learning_rate": 2.0557194152346757e-05, + "loss": 0.3262, + "step": 45969 + }, + { + "epoch": 1.77, + "learning_rate": 2.0556552962298025e-05, + "loss": 0.2711, + "step": 45970 + }, + { + "epoch": 1.77, + "learning_rate": 2.0555911772249294e-05, + "loss": 0.3846, + "step": 45971 + }, + { + "epoch": 1.77, + "learning_rate": 2.0555270582200565e-05, + "loss": 0.5639, + "step": 45972 + }, + { + "epoch": 1.77, + "learning_rate": 2.0554629392151834e-05, + "loss": 0.4625, + "step": 45973 + }, + { + "epoch": 1.77, + "learning_rate": 2.0553988202103105e-05, + "loss": 0.1683, + "step": 45974 + }, + { + "epoch": 1.77, + "learning_rate": 2.0553347012054374e-05, + "loss": 0.2667, + "step": 45975 + }, + { + "epoch": 1.77, + "learning_rate": 2.0552705822005645e-05, + "loss": 0.2836, + "step": 45976 + }, + { + "epoch": 1.77, + "learning_rate": 2.0552064631956914e-05, + "loss": 0.3002, + "step": 45977 + }, + { + "epoch": 1.77, + "learning_rate": 2.0551423441908182e-05, + "loss": 0.3317, + "step": 45978 + }, + { + "epoch": 1.77, + "learning_rate": 2.055078225185945e-05, + "loss": 0.3312, + "step": 45979 + }, + { + "epoch": 1.77, + "learning_rate": 2.0550141061810722e-05, + "loss": 0.5283, + "step": 45980 + }, + { + "epoch": 1.77, + "learning_rate": 2.054949987176199e-05, + "loss": 0.2767, + "step": 45981 + }, + { + "epoch": 1.77, + "learning_rate": 2.054885868171326e-05, + "loss": 0.439, + "step": 45982 + }, + { + "epoch": 1.77, + "learning_rate": 2.054821749166453e-05, + "loss": 0.2687, + "step": 45983 + }, + { + "epoch": 1.77, + "learning_rate": 2.0547576301615802e-05, + "loss": 0.308, + "step": 45984 + }, + { + "epoch": 1.77, + "learning_rate": 2.054693511156707e-05, + "loss": 0.298, + "step": 45985 + }, + { + "epoch": 1.77, + "learning_rate": 2.054629392151834e-05, + "loss": 0.3141, + "step": 45986 + }, + { + "epoch": 1.77, + "learning_rate": 2.0545652731469607e-05, + "loss": 0.4528, + "step": 45987 + }, + { + "epoch": 1.77, + "learning_rate": 2.054501154142088e-05, + "loss": 0.4332, + "step": 45988 + }, + { + "epoch": 1.77, + "learning_rate": 2.0544370351372147e-05, + "loss": 0.3717, + "step": 45989 + }, + { + "epoch": 1.77, + "learning_rate": 2.0543729161323416e-05, + "loss": 0.3979, + "step": 45990 + }, + { + "epoch": 1.77, + "learning_rate": 2.0543087971274684e-05, + "loss": 0.268, + "step": 45991 + }, + { + "epoch": 1.77, + "learning_rate": 2.054244678122596e-05, + "loss": 0.2922, + "step": 45992 + }, + { + "epoch": 1.77, + "learning_rate": 2.0541805591177227e-05, + "loss": 0.3012, + "step": 45993 + }, + { + "epoch": 1.77, + "learning_rate": 2.0541164401128496e-05, + "loss": 0.331, + "step": 45994 + }, + { + "epoch": 1.77, + "learning_rate": 2.0540523211079764e-05, + "loss": 0.4319, + "step": 45995 + }, + { + "epoch": 1.77, + "learning_rate": 2.0539882021031036e-05, + "loss": 0.4186, + "step": 45996 + }, + { + "epoch": 1.77, + "learning_rate": 2.0539240830982304e-05, + "loss": 0.3068, + "step": 45997 + }, + { + "epoch": 1.77, + "learning_rate": 2.0538599640933572e-05, + "loss": 0.4177, + "step": 45998 + }, + { + "epoch": 1.77, + "learning_rate": 2.053795845088484e-05, + "loss": 0.2608, + "step": 45999 + }, + { + "epoch": 1.77, + "learning_rate": 2.0537317260836116e-05, + "loss": 0.3185, + "step": 46000 + }, + { + "epoch": 1.77, + "learning_rate": 2.0536676070787384e-05, + "loss": 0.2053, + "step": 46001 + }, + { + "epoch": 1.77, + "learning_rate": 2.0536034880738652e-05, + "loss": 0.2531, + "step": 46002 + }, + { + "epoch": 1.77, + "learning_rate": 2.053539369068992e-05, + "loss": 0.4025, + "step": 46003 + }, + { + "epoch": 1.77, + "learning_rate": 2.0534752500641192e-05, + "loss": 0.3527, + "step": 46004 + }, + { + "epoch": 1.77, + "learning_rate": 2.053411131059246e-05, + "loss": 0.373, + "step": 46005 + }, + { + "epoch": 1.77, + "learning_rate": 2.053347012054373e-05, + "loss": 0.2736, + "step": 46006 + }, + { + "epoch": 1.77, + "learning_rate": 2.0532828930494997e-05, + "loss": 0.3582, + "step": 46007 + }, + { + "epoch": 1.77, + "learning_rate": 2.053218774044627e-05, + "loss": 0.2249, + "step": 46008 + }, + { + "epoch": 1.77, + "learning_rate": 2.053154655039754e-05, + "loss": 0.4563, + "step": 46009 + }, + { + "epoch": 1.77, + "learning_rate": 2.053090536034881e-05, + "loss": 0.3131, + "step": 46010 + }, + { + "epoch": 1.77, + "learning_rate": 2.0530264170300078e-05, + "loss": 0.2878, + "step": 46011 + }, + { + "epoch": 1.77, + "learning_rate": 2.052962298025135e-05, + "loss": 0.4011, + "step": 46012 + }, + { + "epoch": 1.77, + "learning_rate": 2.0528981790202618e-05, + "loss": 0.3329, + "step": 46013 + }, + { + "epoch": 1.77, + "learning_rate": 2.0528340600153886e-05, + "loss": 0.3822, + "step": 46014 + }, + { + "epoch": 1.77, + "learning_rate": 2.0527699410105154e-05, + "loss": 0.4039, + "step": 46015 + }, + { + "epoch": 1.77, + "learning_rate": 2.0527058220056426e-05, + "loss": 0.3006, + "step": 46016 + }, + { + "epoch": 1.77, + "learning_rate": 2.0526417030007694e-05, + "loss": 0.3787, + "step": 46017 + }, + { + "epoch": 1.77, + "learning_rate": 2.0525775839958966e-05, + "loss": 0.2769, + "step": 46018 + }, + { + "epoch": 1.77, + "learning_rate": 2.0525134649910234e-05, + "loss": 0.248, + "step": 46019 + }, + { + "epoch": 1.77, + "learning_rate": 2.0524493459861506e-05, + "loss": 0.2011, + "step": 46020 + }, + { + "epoch": 1.77, + "learning_rate": 2.0523852269812774e-05, + "loss": 0.3426, + "step": 46021 + }, + { + "epoch": 1.77, + "learning_rate": 2.0523211079764043e-05, + "loss": 0.3271, + "step": 46022 + }, + { + "epoch": 1.77, + "learning_rate": 2.052256988971531e-05, + "loss": 0.2909, + "step": 46023 + }, + { + "epoch": 1.77, + "learning_rate": 2.0521928699666583e-05, + "loss": 0.3594, + "step": 46024 + }, + { + "epoch": 1.77, + "learning_rate": 2.052128750961785e-05, + "loss": 0.3913, + "step": 46025 + }, + { + "epoch": 1.77, + "learning_rate": 2.052064631956912e-05, + "loss": 0.4042, + "step": 46026 + }, + { + "epoch": 1.77, + "learning_rate": 2.052000512952039e-05, + "loss": 0.2935, + "step": 46027 + }, + { + "epoch": 1.77, + "learning_rate": 2.0519363939471663e-05, + "loss": 0.3512, + "step": 46028 + }, + { + "epoch": 1.77, + "learning_rate": 2.051872274942293e-05, + "loss": 0.2176, + "step": 46029 + }, + { + "epoch": 1.77, + "learning_rate": 2.05180815593742e-05, + "loss": 0.3665, + "step": 46030 + }, + { + "epoch": 1.77, + "learning_rate": 2.0517440369325468e-05, + "loss": 0.3545, + "step": 46031 + }, + { + "epoch": 1.77, + "learning_rate": 2.051679917927674e-05, + "loss": 0.4499, + "step": 46032 + }, + { + "epoch": 1.77, + "learning_rate": 2.0516157989228008e-05, + "loss": 0.2517, + "step": 46033 + }, + { + "epoch": 1.77, + "learning_rate": 2.0515516799179276e-05, + "loss": 0.2971, + "step": 46034 + }, + { + "epoch": 1.77, + "learning_rate": 2.0514875609130545e-05, + "loss": 0.3298, + "step": 46035 + }, + { + "epoch": 1.77, + "learning_rate": 2.051423441908182e-05, + "loss": 0.3693, + "step": 46036 + }, + { + "epoch": 1.77, + "learning_rate": 2.0513593229033088e-05, + "loss": 0.2383, + "step": 46037 + }, + { + "epoch": 1.77, + "learning_rate": 2.0512952038984356e-05, + "loss": 0.1978, + "step": 46038 + }, + { + "epoch": 1.77, + "learning_rate": 2.0512310848935625e-05, + "loss": 0.3254, + "step": 46039 + }, + { + "epoch": 1.77, + "learning_rate": 2.0511669658886896e-05, + "loss": 0.2559, + "step": 46040 + }, + { + "epoch": 1.77, + "learning_rate": 2.0511028468838165e-05, + "loss": 0.3301, + "step": 46041 + }, + { + "epoch": 1.77, + "learning_rate": 2.0510387278789433e-05, + "loss": 0.4484, + "step": 46042 + }, + { + "epoch": 1.77, + "learning_rate": 2.05097460887407e-05, + "loss": 0.4721, + "step": 46043 + }, + { + "epoch": 1.77, + "learning_rate": 2.0509104898691973e-05, + "loss": 0.3305, + "step": 46044 + }, + { + "epoch": 1.77, + "learning_rate": 2.0508463708643245e-05, + "loss": 0.2395, + "step": 46045 + }, + { + "epoch": 1.77, + "learning_rate": 2.0507822518594513e-05, + "loss": 0.2823, + "step": 46046 + }, + { + "epoch": 1.77, + "learning_rate": 2.050718132854578e-05, + "loss": 0.3232, + "step": 46047 + }, + { + "epoch": 1.77, + "learning_rate": 2.0506540138497053e-05, + "loss": 0.4416, + "step": 46048 + }, + { + "epoch": 1.77, + "learning_rate": 2.050589894844832e-05, + "loss": 0.3063, + "step": 46049 + }, + { + "epoch": 1.77, + "learning_rate": 2.050525775839959e-05, + "loss": 0.3321, + "step": 46050 + }, + { + "epoch": 1.77, + "learning_rate": 2.0504616568350858e-05, + "loss": 0.3898, + "step": 46051 + }, + { + "epoch": 1.77, + "learning_rate": 2.050397537830213e-05, + "loss": 0.2849, + "step": 46052 + }, + { + "epoch": 1.77, + "learning_rate": 2.0503334188253398e-05, + "loss": 0.312, + "step": 46053 + }, + { + "epoch": 1.77, + "learning_rate": 2.050269299820467e-05, + "loss": 0.288, + "step": 46054 + }, + { + "epoch": 1.77, + "learning_rate": 2.0502051808155938e-05, + "loss": 0.2656, + "step": 46055 + }, + { + "epoch": 1.77, + "learning_rate": 2.050141061810721e-05, + "loss": 0.432, + "step": 46056 + }, + { + "epoch": 1.77, + "learning_rate": 2.0500769428058478e-05, + "loss": 0.2142, + "step": 46057 + }, + { + "epoch": 1.77, + "learning_rate": 2.0500128238009747e-05, + "loss": 0.3097, + "step": 46058 + }, + { + "epoch": 1.77, + "learning_rate": 2.0499487047961015e-05, + "loss": 0.2576, + "step": 46059 + }, + { + "epoch": 1.77, + "learning_rate": 2.0498845857912287e-05, + "loss": 0.279, + "step": 46060 + }, + { + "epoch": 1.77, + "learning_rate": 2.0498204667863555e-05, + "loss": 0.4009, + "step": 46061 + }, + { + "epoch": 1.77, + "learning_rate": 2.0497563477814823e-05, + "loss": 0.2474, + "step": 46062 + }, + { + "epoch": 1.77, + "learning_rate": 2.0496922287766095e-05, + "loss": 0.1907, + "step": 46063 + }, + { + "epoch": 1.77, + "learning_rate": 2.0496281097717367e-05, + "loss": 0.3405, + "step": 46064 + }, + { + "epoch": 1.77, + "learning_rate": 2.0495639907668635e-05, + "loss": 0.3545, + "step": 46065 + }, + { + "epoch": 1.77, + "learning_rate": 2.0494998717619903e-05, + "loss": 0.4608, + "step": 46066 + }, + { + "epoch": 1.77, + "learning_rate": 2.049435752757117e-05, + "loss": 0.3723, + "step": 46067 + }, + { + "epoch": 1.77, + "learning_rate": 2.0493716337522443e-05, + "loss": 0.5005, + "step": 46068 + }, + { + "epoch": 1.77, + "learning_rate": 2.0493075147473712e-05, + "loss": 0.3416, + "step": 46069 + }, + { + "epoch": 1.77, + "learning_rate": 2.049243395742498e-05, + "loss": 0.3792, + "step": 46070 + }, + { + "epoch": 1.77, + "learning_rate": 2.049179276737625e-05, + "loss": 0.4295, + "step": 46071 + }, + { + "epoch": 1.77, + "learning_rate": 2.0491151577327523e-05, + "loss": 0.3644, + "step": 46072 + }, + { + "epoch": 1.77, + "learning_rate": 2.0490510387278792e-05, + "loss": 0.4132, + "step": 46073 + }, + { + "epoch": 1.77, + "learning_rate": 2.048986919723006e-05, + "loss": 0.2985, + "step": 46074 + }, + { + "epoch": 1.77, + "learning_rate": 2.048922800718133e-05, + "loss": 0.2584, + "step": 46075 + }, + { + "epoch": 1.77, + "learning_rate": 2.04885868171326e-05, + "loss": 0.3205, + "step": 46076 + }, + { + "epoch": 1.77, + "learning_rate": 2.048794562708387e-05, + "loss": 0.4223, + "step": 46077 + }, + { + "epoch": 1.77, + "learning_rate": 2.0487304437035137e-05, + "loss": 0.3489, + "step": 46078 + }, + { + "epoch": 1.77, + "learning_rate": 2.0486663246986405e-05, + "loss": 0.3375, + "step": 46079 + }, + { + "epoch": 1.77, + "learning_rate": 2.0486022056937677e-05, + "loss": 0.2896, + "step": 46080 + }, + { + "epoch": 1.77, + "learning_rate": 2.048538086688895e-05, + "loss": 0.4267, + "step": 46081 + }, + { + "epoch": 1.77, + "learning_rate": 2.0484739676840217e-05, + "loss": 0.2527, + "step": 46082 + }, + { + "epoch": 1.77, + "learning_rate": 2.0484098486791485e-05, + "loss": 0.3395, + "step": 46083 + }, + { + "epoch": 1.77, + "learning_rate": 2.0483457296742757e-05, + "loss": 0.3326, + "step": 46084 + }, + { + "epoch": 1.77, + "learning_rate": 2.0482816106694025e-05, + "loss": 0.2506, + "step": 46085 + }, + { + "epoch": 1.77, + "learning_rate": 2.0482174916645294e-05, + "loss": 0.2277, + "step": 46086 + }, + { + "epoch": 1.77, + "learning_rate": 2.0481533726596562e-05, + "loss": 0.3757, + "step": 46087 + }, + { + "epoch": 1.77, + "learning_rate": 2.0480892536547834e-05, + "loss": 0.4294, + "step": 46088 + }, + { + "epoch": 1.77, + "learning_rate": 2.0480251346499102e-05, + "loss": 0.3126, + "step": 46089 + }, + { + "epoch": 1.77, + "learning_rate": 2.0479610156450374e-05, + "loss": 0.2458, + "step": 46090 + }, + { + "epoch": 1.77, + "learning_rate": 2.0478968966401642e-05, + "loss": 0.297, + "step": 46091 + }, + { + "epoch": 1.77, + "learning_rate": 2.0478327776352914e-05, + "loss": 0.2772, + "step": 46092 + }, + { + "epoch": 1.77, + "learning_rate": 2.0477686586304182e-05, + "loss": 0.3364, + "step": 46093 + }, + { + "epoch": 1.77, + "learning_rate": 2.047704539625545e-05, + "loss": 0.4368, + "step": 46094 + }, + { + "epoch": 1.77, + "learning_rate": 2.047640420620672e-05, + "loss": 0.412, + "step": 46095 + }, + { + "epoch": 1.77, + "learning_rate": 2.047576301615799e-05, + "loss": 0.2413, + "step": 46096 + }, + { + "epoch": 1.77, + "learning_rate": 2.047512182610926e-05, + "loss": 0.5758, + "step": 46097 + }, + { + "epoch": 1.77, + "learning_rate": 2.047448063606053e-05, + "loss": 0.2903, + "step": 46098 + }, + { + "epoch": 1.77, + "learning_rate": 2.04738394460118e-05, + "loss": 0.3331, + "step": 46099 + }, + { + "epoch": 1.77, + "learning_rate": 2.047319825596307e-05, + "loss": 0.3544, + "step": 46100 + }, + { + "epoch": 1.77, + "learning_rate": 2.047255706591434e-05, + "loss": 0.5015, + "step": 46101 + }, + { + "epoch": 1.77, + "learning_rate": 2.0471915875865607e-05, + "loss": 0.2842, + "step": 46102 + }, + { + "epoch": 1.77, + "learning_rate": 2.0471274685816876e-05, + "loss": 0.3304, + "step": 46103 + }, + { + "epoch": 1.77, + "learning_rate": 2.0470633495768147e-05, + "loss": 0.4287, + "step": 46104 + }, + { + "epoch": 1.77, + "learning_rate": 2.0469992305719416e-05, + "loss": 0.3488, + "step": 46105 + }, + { + "epoch": 1.77, + "learning_rate": 2.0469351115670684e-05, + "loss": 0.3894, + "step": 46106 + }, + { + "epoch": 1.77, + "learning_rate": 2.0468709925621956e-05, + "loss": 0.2763, + "step": 46107 + }, + { + "epoch": 1.77, + "learning_rate": 2.0468068735573227e-05, + "loss": 0.3545, + "step": 46108 + }, + { + "epoch": 1.77, + "learning_rate": 2.0467427545524496e-05, + "loss": 0.2757, + "step": 46109 + }, + { + "epoch": 1.77, + "learning_rate": 2.0466786355475764e-05, + "loss": 0.2876, + "step": 46110 + }, + { + "epoch": 1.77, + "learning_rate": 2.0466145165427032e-05, + "loss": 0.4281, + "step": 46111 + }, + { + "epoch": 1.77, + "learning_rate": 2.0465503975378304e-05, + "loss": 0.3895, + "step": 46112 + }, + { + "epoch": 1.77, + "learning_rate": 2.0464862785329572e-05, + "loss": 0.3282, + "step": 46113 + }, + { + "epoch": 1.77, + "learning_rate": 2.046422159528084e-05, + "loss": 0.4532, + "step": 46114 + }, + { + "epoch": 1.77, + "learning_rate": 2.046358040523211e-05, + "loss": 0.4837, + "step": 46115 + }, + { + "epoch": 1.77, + "learning_rate": 2.0462939215183384e-05, + "loss": 0.2572, + "step": 46116 + }, + { + "epoch": 1.77, + "learning_rate": 2.0462298025134652e-05, + "loss": 0.553, + "step": 46117 + }, + { + "epoch": 1.77, + "learning_rate": 2.046165683508592e-05, + "loss": 0.3433, + "step": 46118 + }, + { + "epoch": 1.77, + "learning_rate": 2.046101564503719e-05, + "loss": 0.4533, + "step": 46119 + }, + { + "epoch": 1.77, + "learning_rate": 2.046037445498846e-05, + "loss": 0.3515, + "step": 46120 + }, + { + "epoch": 1.77, + "learning_rate": 2.045973326493973e-05, + "loss": 0.2521, + "step": 46121 + }, + { + "epoch": 1.77, + "learning_rate": 2.0459092074890998e-05, + "loss": 0.519, + "step": 46122 + }, + { + "epoch": 1.77, + "learning_rate": 2.0458450884842266e-05, + "loss": 0.2903, + "step": 46123 + }, + { + "epoch": 1.77, + "learning_rate": 2.0457809694793538e-05, + "loss": 0.3018, + "step": 46124 + }, + { + "epoch": 1.77, + "learning_rate": 2.045716850474481e-05, + "loss": 0.3459, + "step": 46125 + }, + { + "epoch": 1.77, + "learning_rate": 2.0456527314696078e-05, + "loss": 0.2655, + "step": 46126 + }, + { + "epoch": 1.77, + "learning_rate": 2.0455886124647346e-05, + "loss": 0.3054, + "step": 46127 + }, + { + "epoch": 1.77, + "learning_rate": 2.0455244934598618e-05, + "loss": 0.3499, + "step": 46128 + }, + { + "epoch": 1.77, + "learning_rate": 2.0454603744549886e-05, + "loss": 0.2884, + "step": 46129 + }, + { + "epoch": 1.77, + "learning_rate": 2.0453962554501154e-05, + "loss": 0.4809, + "step": 46130 + }, + { + "epoch": 1.77, + "learning_rate": 2.0453321364452423e-05, + "loss": 0.2752, + "step": 46131 + }, + { + "epoch": 1.77, + "learning_rate": 2.0452680174403694e-05, + "loss": 0.4508, + "step": 46132 + }, + { + "epoch": 1.77, + "learning_rate": 2.0452038984354963e-05, + "loss": 0.4518, + "step": 46133 + }, + { + "epoch": 1.77, + "learning_rate": 2.0451397794306234e-05, + "loss": 0.3605, + "step": 46134 + }, + { + "epoch": 1.77, + "learning_rate": 2.0450756604257503e-05, + "loss": 0.3221, + "step": 46135 + }, + { + "epoch": 1.77, + "learning_rate": 2.0450115414208774e-05, + "loss": 0.3313, + "step": 46136 + }, + { + "epoch": 1.77, + "learning_rate": 2.0449474224160043e-05, + "loss": 0.322, + "step": 46137 + }, + { + "epoch": 1.77, + "learning_rate": 2.044883303411131e-05, + "loss": 0.2902, + "step": 46138 + }, + { + "epoch": 1.77, + "learning_rate": 2.044819184406258e-05, + "loss": 0.371, + "step": 46139 + }, + { + "epoch": 1.77, + "learning_rate": 2.044755065401385e-05, + "loss": 0.3957, + "step": 46140 + }, + { + "epoch": 1.77, + "learning_rate": 2.044690946396512e-05, + "loss": 0.2824, + "step": 46141 + }, + { + "epoch": 1.77, + "learning_rate": 2.0446268273916388e-05, + "loss": 0.251, + "step": 46142 + }, + { + "epoch": 1.77, + "learning_rate": 2.044562708386766e-05, + "loss": 0.3443, + "step": 46143 + }, + { + "epoch": 1.77, + "learning_rate": 2.044498589381893e-05, + "loss": 0.3646, + "step": 46144 + }, + { + "epoch": 1.77, + "learning_rate": 2.04443447037702e-05, + "loss": 0.3795, + "step": 46145 + }, + { + "epoch": 1.77, + "learning_rate": 2.0443703513721468e-05, + "loss": 0.2546, + "step": 46146 + }, + { + "epoch": 1.77, + "learning_rate": 2.0443062323672736e-05, + "loss": 0.3712, + "step": 46147 + }, + { + "epoch": 1.77, + "learning_rate": 2.0442421133624008e-05, + "loss": 0.3243, + "step": 46148 + }, + { + "epoch": 1.77, + "learning_rate": 2.0441779943575276e-05, + "loss": 0.3759, + "step": 46149 + }, + { + "epoch": 1.77, + "learning_rate": 2.0441138753526545e-05, + "loss": 0.3212, + "step": 46150 + }, + { + "epoch": 1.77, + "learning_rate": 2.0440497563477813e-05, + "loss": 0.3617, + "step": 46151 + }, + { + "epoch": 1.77, + "learning_rate": 2.0439856373429088e-05, + "loss": 0.2947, + "step": 46152 + }, + { + "epoch": 1.77, + "learning_rate": 2.0439215183380356e-05, + "loss": 0.2843, + "step": 46153 + }, + { + "epoch": 1.77, + "learning_rate": 2.0438573993331625e-05, + "loss": 0.2703, + "step": 46154 + }, + { + "epoch": 1.77, + "learning_rate": 2.0437932803282893e-05, + "loss": 0.4671, + "step": 46155 + }, + { + "epoch": 1.77, + "learning_rate": 2.0437291613234165e-05, + "loss": 0.3626, + "step": 46156 + }, + { + "epoch": 1.77, + "learning_rate": 2.0436650423185433e-05, + "loss": 0.3082, + "step": 46157 + }, + { + "epoch": 1.77, + "learning_rate": 2.04360092331367e-05, + "loss": 0.2798, + "step": 46158 + }, + { + "epoch": 1.77, + "learning_rate": 2.043536804308797e-05, + "loss": 0.4142, + "step": 46159 + }, + { + "epoch": 1.77, + "learning_rate": 2.043472685303924e-05, + "loss": 0.3163, + "step": 46160 + }, + { + "epoch": 1.77, + "learning_rate": 2.0434085662990513e-05, + "loss": 0.2942, + "step": 46161 + }, + { + "epoch": 1.77, + "learning_rate": 2.043344447294178e-05, + "loss": 0.4168, + "step": 46162 + }, + { + "epoch": 1.77, + "learning_rate": 2.043280328289305e-05, + "loss": 0.3694, + "step": 46163 + }, + { + "epoch": 1.77, + "learning_rate": 2.043216209284432e-05, + "loss": 0.3796, + "step": 46164 + }, + { + "epoch": 1.77, + "learning_rate": 2.043152090279559e-05, + "loss": 0.3969, + "step": 46165 + }, + { + "epoch": 1.77, + "learning_rate": 2.0430879712746858e-05, + "loss": 0.3988, + "step": 46166 + }, + { + "epoch": 1.77, + "learning_rate": 2.0430238522698127e-05, + "loss": 0.4008, + "step": 46167 + }, + { + "epoch": 1.78, + "learning_rate": 2.0429597332649398e-05, + "loss": 0.3947, + "step": 46168 + }, + { + "epoch": 1.78, + "learning_rate": 2.0428956142600667e-05, + "loss": 0.2684, + "step": 46169 + }, + { + "epoch": 1.78, + "learning_rate": 2.0428314952551938e-05, + "loss": 0.3382, + "step": 46170 + }, + { + "epoch": 1.78, + "learning_rate": 2.0427673762503207e-05, + "loss": 0.2961, + "step": 46171 + }, + { + "epoch": 1.78, + "learning_rate": 2.042703257245448e-05, + "loss": 0.3508, + "step": 46172 + }, + { + "epoch": 1.78, + "learning_rate": 2.0426391382405747e-05, + "loss": 0.247, + "step": 46173 + }, + { + "epoch": 1.78, + "learning_rate": 2.0425750192357015e-05, + "loss": 0.3264, + "step": 46174 + }, + { + "epoch": 1.78, + "learning_rate": 2.0425109002308283e-05, + "loss": 0.2233, + "step": 46175 + }, + { + "epoch": 1.78, + "learning_rate": 2.0424467812259555e-05, + "loss": 0.2958, + "step": 46176 + }, + { + "epoch": 1.78, + "learning_rate": 2.0423826622210823e-05, + "loss": 0.3382, + "step": 46177 + }, + { + "epoch": 1.78, + "learning_rate": 2.042318543216209e-05, + "loss": 0.321, + "step": 46178 + }, + { + "epoch": 1.78, + "learning_rate": 2.0422544242113363e-05, + "loss": 0.317, + "step": 46179 + }, + { + "epoch": 1.78, + "learning_rate": 2.0421903052064635e-05, + "loss": 0.2996, + "step": 46180 + }, + { + "epoch": 1.78, + "learning_rate": 2.0421261862015903e-05, + "loss": 0.3879, + "step": 46181 + }, + { + "epoch": 1.78, + "learning_rate": 2.0420620671967172e-05, + "loss": 0.493, + "step": 46182 + }, + { + "epoch": 1.78, + "learning_rate": 2.041997948191844e-05, + "loss": 0.199, + "step": 46183 + }, + { + "epoch": 1.78, + "learning_rate": 2.0419338291869712e-05, + "loss": 0.3054, + "step": 46184 + }, + { + "epoch": 1.78, + "learning_rate": 2.041869710182098e-05, + "loss": 0.3214, + "step": 46185 + }, + { + "epoch": 1.78, + "learning_rate": 2.041805591177225e-05, + "loss": 0.3877, + "step": 46186 + }, + { + "epoch": 1.78, + "learning_rate": 2.041741472172352e-05, + "loss": 0.2891, + "step": 46187 + }, + { + "epoch": 1.78, + "learning_rate": 2.0416773531674792e-05, + "loss": 0.192, + "step": 46188 + }, + { + "epoch": 1.78, + "learning_rate": 2.041613234162606e-05, + "loss": 0.399, + "step": 46189 + }, + { + "epoch": 1.78, + "learning_rate": 2.041549115157733e-05, + "loss": 0.3358, + "step": 46190 + }, + { + "epoch": 1.78, + "learning_rate": 2.0414849961528597e-05, + "loss": 0.2835, + "step": 46191 + }, + { + "epoch": 1.78, + "learning_rate": 2.041420877147987e-05, + "loss": 0.3295, + "step": 46192 + }, + { + "epoch": 1.78, + "learning_rate": 2.0413567581431137e-05, + "loss": 0.3676, + "step": 46193 + }, + { + "epoch": 1.78, + "learning_rate": 2.0412926391382405e-05, + "loss": 0.3122, + "step": 46194 + }, + { + "epoch": 1.78, + "learning_rate": 2.0412285201333674e-05, + "loss": 0.2977, + "step": 46195 + }, + { + "epoch": 1.78, + "learning_rate": 2.041164401128495e-05, + "loss": 0.37, + "step": 46196 + }, + { + "epoch": 1.78, + "learning_rate": 2.0411002821236217e-05, + "loss": 0.3603, + "step": 46197 + }, + { + "epoch": 1.78, + "learning_rate": 2.0410361631187485e-05, + "loss": 0.2936, + "step": 46198 + }, + { + "epoch": 1.78, + "learning_rate": 2.0409720441138754e-05, + "loss": 0.2609, + "step": 46199 + }, + { + "epoch": 1.78, + "learning_rate": 2.0409079251090025e-05, + "loss": 0.3408, + "step": 46200 + }, + { + "epoch": 1.78, + "learning_rate": 2.0408438061041294e-05, + "loss": 0.427, + "step": 46201 + }, + { + "epoch": 1.78, + "learning_rate": 2.0407796870992562e-05, + "loss": 0.3825, + "step": 46202 + }, + { + "epoch": 1.78, + "learning_rate": 2.040715568094383e-05, + "loss": 0.2828, + "step": 46203 + }, + { + "epoch": 1.78, + "learning_rate": 2.0406514490895102e-05, + "loss": 0.1931, + "step": 46204 + }, + { + "epoch": 1.78, + "learning_rate": 2.0405873300846374e-05, + "loss": 0.2863, + "step": 46205 + }, + { + "epoch": 1.78, + "learning_rate": 2.0405232110797642e-05, + "loss": 0.5151, + "step": 46206 + }, + { + "epoch": 1.78, + "learning_rate": 2.040459092074891e-05, + "loss": 0.3267, + "step": 46207 + }, + { + "epoch": 1.78, + "learning_rate": 2.0403949730700182e-05, + "loss": 0.3776, + "step": 46208 + }, + { + "epoch": 1.78, + "learning_rate": 2.040330854065145e-05, + "loss": 0.3434, + "step": 46209 + }, + { + "epoch": 1.78, + "learning_rate": 2.040266735060272e-05, + "loss": 0.3908, + "step": 46210 + }, + { + "epoch": 1.78, + "learning_rate": 2.0402026160553987e-05, + "loss": 0.3578, + "step": 46211 + }, + { + "epoch": 1.78, + "learning_rate": 2.040138497050526e-05, + "loss": 0.3065, + "step": 46212 + }, + { + "epoch": 1.78, + "learning_rate": 2.0400743780456527e-05, + "loss": 0.3844, + "step": 46213 + }, + { + "epoch": 1.78, + "learning_rate": 2.04001025904078e-05, + "loss": 0.2718, + "step": 46214 + }, + { + "epoch": 1.78, + "learning_rate": 2.0399461400359067e-05, + "loss": 0.3673, + "step": 46215 + }, + { + "epoch": 1.78, + "learning_rate": 2.039882021031034e-05, + "loss": 0.3588, + "step": 46216 + }, + { + "epoch": 1.78, + "learning_rate": 2.0398179020261607e-05, + "loss": 0.3151, + "step": 46217 + }, + { + "epoch": 1.78, + "learning_rate": 2.0397537830212876e-05, + "loss": 0.3725, + "step": 46218 + }, + { + "epoch": 1.78, + "learning_rate": 2.0396896640164144e-05, + "loss": 0.3531, + "step": 46219 + }, + { + "epoch": 1.78, + "learning_rate": 2.0396255450115416e-05, + "loss": 0.312, + "step": 46220 + }, + { + "epoch": 1.78, + "learning_rate": 2.0395614260066684e-05, + "loss": 0.2885, + "step": 46221 + }, + { + "epoch": 1.78, + "learning_rate": 2.0394973070017952e-05, + "loss": 0.5469, + "step": 46222 + }, + { + "epoch": 1.78, + "learning_rate": 2.0394331879969224e-05, + "loss": 0.3039, + "step": 46223 + }, + { + "epoch": 1.78, + "learning_rate": 2.0393690689920496e-05, + "loss": 0.2947, + "step": 46224 + }, + { + "epoch": 1.78, + "learning_rate": 2.0393049499871764e-05, + "loss": 0.4791, + "step": 46225 + }, + { + "epoch": 1.78, + "learning_rate": 2.0392408309823032e-05, + "loss": 0.322, + "step": 46226 + }, + { + "epoch": 1.78, + "learning_rate": 2.03917671197743e-05, + "loss": 0.3025, + "step": 46227 + }, + { + "epoch": 1.78, + "learning_rate": 2.0391125929725572e-05, + "loss": 0.3559, + "step": 46228 + }, + { + "epoch": 1.78, + "learning_rate": 2.039048473967684e-05, + "loss": 0.3032, + "step": 46229 + }, + { + "epoch": 1.78, + "learning_rate": 2.038984354962811e-05, + "loss": 0.4038, + "step": 46230 + }, + { + "epoch": 1.78, + "learning_rate": 2.0389202359579377e-05, + "loss": 0.3184, + "step": 46231 + }, + { + "epoch": 1.78, + "learning_rate": 2.0388561169530653e-05, + "loss": 0.3528, + "step": 46232 + }, + { + "epoch": 1.78, + "learning_rate": 2.038791997948192e-05, + "loss": 0.2491, + "step": 46233 + }, + { + "epoch": 1.78, + "learning_rate": 2.038727878943319e-05, + "loss": 0.2922, + "step": 46234 + }, + { + "epoch": 1.78, + "learning_rate": 2.0386637599384458e-05, + "loss": 0.3716, + "step": 46235 + }, + { + "epoch": 1.78, + "learning_rate": 2.038599640933573e-05, + "loss": 0.3816, + "step": 46236 + }, + { + "epoch": 1.78, + "learning_rate": 2.0385355219286998e-05, + "loss": 0.3456, + "step": 46237 + }, + { + "epoch": 1.78, + "learning_rate": 2.0384714029238266e-05, + "loss": 0.4348, + "step": 46238 + }, + { + "epoch": 1.78, + "learning_rate": 2.0384072839189534e-05, + "loss": 0.3952, + "step": 46239 + }, + { + "epoch": 1.78, + "learning_rate": 2.0383431649140806e-05, + "loss": 0.1818, + "step": 46240 + }, + { + "epoch": 1.78, + "learning_rate": 2.0382790459092078e-05, + "loss": 0.4794, + "step": 46241 + }, + { + "epoch": 1.78, + "learning_rate": 2.0382149269043346e-05, + "loss": 0.3624, + "step": 46242 + }, + { + "epoch": 1.78, + "learning_rate": 2.0381508078994614e-05, + "loss": 0.2297, + "step": 46243 + }, + { + "epoch": 1.78, + "learning_rate": 2.0380866888945886e-05, + "loss": 0.2439, + "step": 46244 + }, + { + "epoch": 1.78, + "learning_rate": 2.0380225698897154e-05, + "loss": 0.3187, + "step": 46245 + }, + { + "epoch": 1.78, + "learning_rate": 2.0379584508848423e-05, + "loss": 0.3322, + "step": 46246 + }, + { + "epoch": 1.78, + "learning_rate": 2.037894331879969e-05, + "loss": 0.3221, + "step": 46247 + }, + { + "epoch": 1.78, + "learning_rate": 2.0378302128750963e-05, + "loss": 0.3376, + "step": 46248 + }, + { + "epoch": 1.78, + "learning_rate": 2.037766093870223e-05, + "loss": 0.3695, + "step": 46249 + }, + { + "epoch": 1.78, + "learning_rate": 2.0377019748653503e-05, + "loss": 0.526, + "step": 46250 + }, + { + "epoch": 1.78, + "learning_rate": 2.037637855860477e-05, + "loss": 0.2638, + "step": 46251 + }, + { + "epoch": 1.78, + "learning_rate": 2.0375737368556043e-05, + "loss": 0.2695, + "step": 46252 + }, + { + "epoch": 1.78, + "learning_rate": 2.037509617850731e-05, + "loss": 0.3474, + "step": 46253 + }, + { + "epoch": 1.78, + "learning_rate": 2.037445498845858e-05, + "loss": 0.3653, + "step": 46254 + }, + { + "epoch": 1.78, + "learning_rate": 2.0373813798409848e-05, + "loss": 0.3461, + "step": 46255 + }, + { + "epoch": 1.78, + "learning_rate": 2.037317260836112e-05, + "loss": 0.3042, + "step": 46256 + }, + { + "epoch": 1.78, + "learning_rate": 2.0372531418312388e-05, + "loss": 0.3087, + "step": 46257 + }, + { + "epoch": 1.78, + "learning_rate": 2.0371890228263656e-05, + "loss": 0.454, + "step": 46258 + }, + { + "epoch": 1.78, + "learning_rate": 2.0371249038214928e-05, + "loss": 0.357, + "step": 46259 + }, + { + "epoch": 1.78, + "learning_rate": 2.03706078481662e-05, + "loss": 0.511, + "step": 46260 + }, + { + "epoch": 1.78, + "learning_rate": 2.0369966658117468e-05, + "loss": 0.2678, + "step": 46261 + }, + { + "epoch": 1.78, + "learning_rate": 2.0369325468068736e-05, + "loss": 0.3367, + "step": 46262 + }, + { + "epoch": 1.78, + "learning_rate": 2.0368684278020005e-05, + "loss": 0.3612, + "step": 46263 + }, + { + "epoch": 1.78, + "learning_rate": 2.0368043087971276e-05, + "loss": 0.3554, + "step": 46264 + }, + { + "epoch": 1.78, + "learning_rate": 2.0367401897922545e-05, + "loss": 0.3067, + "step": 46265 + }, + { + "epoch": 1.78, + "learning_rate": 2.0366760707873813e-05, + "loss": 0.3062, + "step": 46266 + }, + { + "epoch": 1.78, + "learning_rate": 2.036611951782508e-05, + "loss": 0.3694, + "step": 46267 + }, + { + "epoch": 1.78, + "learning_rate": 2.0365478327776356e-05, + "loss": 0.368, + "step": 46268 + }, + { + "epoch": 1.78, + "learning_rate": 2.0364837137727625e-05, + "loss": 0.3952, + "step": 46269 + }, + { + "epoch": 1.78, + "learning_rate": 2.0364195947678893e-05, + "loss": 0.3556, + "step": 46270 + }, + { + "epoch": 1.78, + "learning_rate": 2.036355475763016e-05, + "loss": 0.252, + "step": 46271 + }, + { + "epoch": 1.78, + "learning_rate": 2.0362913567581433e-05, + "loss": 0.2942, + "step": 46272 + }, + { + "epoch": 1.78, + "learning_rate": 2.03622723775327e-05, + "loss": 0.3145, + "step": 46273 + }, + { + "epoch": 1.78, + "learning_rate": 2.036163118748397e-05, + "loss": 0.4448, + "step": 46274 + }, + { + "epoch": 1.78, + "learning_rate": 2.0360989997435238e-05, + "loss": 0.4566, + "step": 46275 + }, + { + "epoch": 1.78, + "learning_rate": 2.036034880738651e-05, + "loss": 0.3579, + "step": 46276 + }, + { + "epoch": 1.78, + "learning_rate": 2.035970761733778e-05, + "loss": 0.2341, + "step": 46277 + }, + { + "epoch": 1.78, + "learning_rate": 2.035906642728905e-05, + "loss": 0.349, + "step": 46278 + }, + { + "epoch": 1.78, + "learning_rate": 2.0358425237240318e-05, + "loss": 0.2149, + "step": 46279 + }, + { + "epoch": 1.78, + "learning_rate": 2.035778404719159e-05, + "loss": 0.4233, + "step": 46280 + }, + { + "epoch": 1.78, + "learning_rate": 2.0357142857142858e-05, + "loss": 0.3861, + "step": 46281 + }, + { + "epoch": 1.78, + "learning_rate": 2.0356501667094127e-05, + "loss": 0.3097, + "step": 46282 + }, + { + "epoch": 1.78, + "learning_rate": 2.0355860477045395e-05, + "loss": 0.4177, + "step": 46283 + }, + { + "epoch": 1.78, + "learning_rate": 2.0355219286996667e-05, + "loss": 0.3605, + "step": 46284 + }, + { + "epoch": 1.78, + "learning_rate": 2.035457809694794e-05, + "loss": 0.323, + "step": 46285 + }, + { + "epoch": 1.78, + "learning_rate": 2.0353936906899207e-05, + "loss": 0.2925, + "step": 46286 + }, + { + "epoch": 1.78, + "learning_rate": 2.0353295716850475e-05, + "loss": 0.2874, + "step": 46287 + }, + { + "epoch": 1.78, + "learning_rate": 2.0352654526801747e-05, + "loss": 0.3436, + "step": 46288 + }, + { + "epoch": 1.78, + "learning_rate": 2.0352013336753015e-05, + "loss": 0.4061, + "step": 46289 + }, + { + "epoch": 1.78, + "learning_rate": 2.0351372146704283e-05, + "loss": 0.3219, + "step": 46290 + }, + { + "epoch": 1.78, + "learning_rate": 2.0350730956655552e-05, + "loss": 0.3662, + "step": 46291 + }, + { + "epoch": 1.78, + "learning_rate": 2.0350089766606823e-05, + "loss": 0.3359, + "step": 46292 + }, + { + "epoch": 1.78, + "learning_rate": 2.0349448576558092e-05, + "loss": 0.2423, + "step": 46293 + }, + { + "epoch": 1.78, + "learning_rate": 2.0348807386509363e-05, + "loss": 0.4449, + "step": 46294 + }, + { + "epoch": 1.78, + "learning_rate": 2.0348166196460632e-05, + "loss": 0.273, + "step": 46295 + }, + { + "epoch": 1.78, + "learning_rate": 2.0347525006411904e-05, + "loss": 0.2996, + "step": 46296 + }, + { + "epoch": 1.78, + "learning_rate": 2.0346883816363172e-05, + "loss": 0.3047, + "step": 46297 + }, + { + "epoch": 1.78, + "learning_rate": 2.034624262631444e-05, + "loss": 0.4544, + "step": 46298 + }, + { + "epoch": 1.78, + "learning_rate": 2.034560143626571e-05, + "loss": 0.2853, + "step": 46299 + }, + { + "epoch": 1.78, + "learning_rate": 2.034496024621698e-05, + "loss": 0.2731, + "step": 46300 + }, + { + "epoch": 1.78, + "learning_rate": 2.034431905616825e-05, + "loss": 0.3285, + "step": 46301 + }, + { + "epoch": 1.78, + "learning_rate": 2.0343677866119517e-05, + "loss": 0.2591, + "step": 46302 + }, + { + "epoch": 1.78, + "learning_rate": 2.034303667607079e-05, + "loss": 0.3119, + "step": 46303 + }, + { + "epoch": 1.78, + "learning_rate": 2.034239548602206e-05, + "loss": 0.4085, + "step": 46304 + }, + { + "epoch": 1.78, + "learning_rate": 2.034175429597333e-05, + "loss": 0.2943, + "step": 46305 + }, + { + "epoch": 1.78, + "learning_rate": 2.0341113105924597e-05, + "loss": 0.1857, + "step": 46306 + }, + { + "epoch": 1.78, + "learning_rate": 2.0340471915875865e-05, + "loss": 0.2887, + "step": 46307 + }, + { + "epoch": 1.78, + "learning_rate": 2.0339830725827137e-05, + "loss": 0.4245, + "step": 46308 + }, + { + "epoch": 1.78, + "learning_rate": 2.0339189535778405e-05, + "loss": 0.4507, + "step": 46309 + }, + { + "epoch": 1.78, + "learning_rate": 2.0338548345729674e-05, + "loss": 0.4521, + "step": 46310 + }, + { + "epoch": 1.78, + "learning_rate": 2.0337907155680942e-05, + "loss": 0.2474, + "step": 46311 + }, + { + "epoch": 1.78, + "learning_rate": 2.0337265965632217e-05, + "loss": 0.2658, + "step": 46312 + }, + { + "epoch": 1.78, + "learning_rate": 2.0336624775583485e-05, + "loss": 0.2929, + "step": 46313 + }, + { + "epoch": 1.78, + "learning_rate": 2.0335983585534754e-05, + "loss": 0.333, + "step": 46314 + }, + { + "epoch": 1.78, + "learning_rate": 2.0335342395486022e-05, + "loss": 0.409, + "step": 46315 + }, + { + "epoch": 1.78, + "learning_rate": 2.0334701205437294e-05, + "loss": 0.2756, + "step": 46316 + }, + { + "epoch": 1.78, + "learning_rate": 2.0334060015388562e-05, + "loss": 0.2644, + "step": 46317 + }, + { + "epoch": 1.78, + "learning_rate": 2.033341882533983e-05, + "loss": 0.3806, + "step": 46318 + }, + { + "epoch": 1.78, + "learning_rate": 2.03327776352911e-05, + "loss": 0.3304, + "step": 46319 + }, + { + "epoch": 1.78, + "learning_rate": 2.033213644524237e-05, + "loss": 0.3639, + "step": 46320 + }, + { + "epoch": 1.78, + "learning_rate": 2.0331495255193642e-05, + "loss": 0.4071, + "step": 46321 + }, + { + "epoch": 1.78, + "learning_rate": 2.033085406514491e-05, + "loss": 0.3012, + "step": 46322 + }, + { + "epoch": 1.78, + "learning_rate": 2.033021287509618e-05, + "loss": 0.2749, + "step": 46323 + }, + { + "epoch": 1.78, + "learning_rate": 2.032957168504745e-05, + "loss": 0.2558, + "step": 46324 + }, + { + "epoch": 1.78, + "learning_rate": 2.032893049499872e-05, + "loss": 0.2971, + "step": 46325 + }, + { + "epoch": 1.78, + "learning_rate": 2.0328289304949987e-05, + "loss": 0.503, + "step": 46326 + }, + { + "epoch": 1.78, + "learning_rate": 2.0327648114901256e-05, + "loss": 0.3067, + "step": 46327 + }, + { + "epoch": 1.78, + "learning_rate": 2.0327006924852527e-05, + "loss": 0.2368, + "step": 46328 + }, + { + "epoch": 1.78, + "learning_rate": 2.0326365734803796e-05, + "loss": 0.3593, + "step": 46329 + }, + { + "epoch": 1.78, + "learning_rate": 2.0325724544755067e-05, + "loss": 0.3943, + "step": 46330 + }, + { + "epoch": 1.78, + "learning_rate": 2.0325083354706336e-05, + "loss": 0.2347, + "step": 46331 + }, + { + "epoch": 1.78, + "learning_rate": 2.0324442164657607e-05, + "loss": 0.3352, + "step": 46332 + }, + { + "epoch": 1.78, + "learning_rate": 2.0323800974608876e-05, + "loss": 0.2968, + "step": 46333 + }, + { + "epoch": 1.78, + "learning_rate": 2.0323159784560144e-05, + "loss": 0.3001, + "step": 46334 + }, + { + "epoch": 1.78, + "learning_rate": 2.0322518594511412e-05, + "loss": 0.2444, + "step": 46335 + }, + { + "epoch": 1.78, + "learning_rate": 2.0321877404462684e-05, + "loss": 0.2608, + "step": 46336 + }, + { + "epoch": 1.78, + "learning_rate": 2.0321236214413952e-05, + "loss": 0.302, + "step": 46337 + }, + { + "epoch": 1.78, + "learning_rate": 2.032059502436522e-05, + "loss": 0.3018, + "step": 46338 + }, + { + "epoch": 1.78, + "learning_rate": 2.0319953834316492e-05, + "loss": 0.3995, + "step": 46339 + }, + { + "epoch": 1.78, + "learning_rate": 2.0319312644267764e-05, + "loss": 0.2451, + "step": 46340 + }, + { + "epoch": 1.78, + "learning_rate": 2.0318671454219033e-05, + "loss": 0.2491, + "step": 46341 + }, + { + "epoch": 1.78, + "learning_rate": 2.03180302641703e-05, + "loss": 0.3084, + "step": 46342 + }, + { + "epoch": 1.78, + "learning_rate": 2.031738907412157e-05, + "loss": 0.3223, + "step": 46343 + }, + { + "epoch": 1.78, + "learning_rate": 2.031674788407284e-05, + "loss": 0.3038, + "step": 46344 + }, + { + "epoch": 1.78, + "learning_rate": 2.031610669402411e-05, + "loss": 0.2788, + "step": 46345 + }, + { + "epoch": 1.78, + "learning_rate": 2.0315465503975378e-05, + "loss": 0.4925, + "step": 46346 + }, + { + "epoch": 1.78, + "learning_rate": 2.0314824313926646e-05, + "loss": 0.2792, + "step": 46347 + }, + { + "epoch": 1.78, + "learning_rate": 2.031418312387792e-05, + "loss": 0.2504, + "step": 46348 + }, + { + "epoch": 1.78, + "learning_rate": 2.031354193382919e-05, + "loss": 0.3299, + "step": 46349 + }, + { + "epoch": 1.78, + "learning_rate": 2.0312900743780458e-05, + "loss": 0.276, + "step": 46350 + }, + { + "epoch": 1.78, + "learning_rate": 2.0312259553731726e-05, + "loss": 0.3619, + "step": 46351 + }, + { + "epoch": 1.78, + "learning_rate": 2.0311618363682998e-05, + "loss": 0.4704, + "step": 46352 + }, + { + "epoch": 1.78, + "learning_rate": 2.0310977173634266e-05, + "loss": 0.3213, + "step": 46353 + }, + { + "epoch": 1.78, + "learning_rate": 2.0310335983585534e-05, + "loss": 0.2637, + "step": 46354 + }, + { + "epoch": 1.78, + "learning_rate": 2.0309694793536803e-05, + "loss": 0.3674, + "step": 46355 + }, + { + "epoch": 1.78, + "learning_rate": 2.0309053603488074e-05, + "loss": 0.3208, + "step": 46356 + }, + { + "epoch": 1.78, + "learning_rate": 2.0308412413439346e-05, + "loss": 0.3294, + "step": 46357 + }, + { + "epoch": 1.78, + "learning_rate": 2.0307771223390614e-05, + "loss": 0.2694, + "step": 46358 + }, + { + "epoch": 1.78, + "learning_rate": 2.0307130033341883e-05, + "loss": 0.3709, + "step": 46359 + }, + { + "epoch": 1.78, + "learning_rate": 2.0306488843293154e-05, + "loss": 0.3144, + "step": 46360 + }, + { + "epoch": 1.78, + "learning_rate": 2.0305847653244423e-05, + "loss": 0.2815, + "step": 46361 + }, + { + "epoch": 1.78, + "learning_rate": 2.030520646319569e-05, + "loss": 0.3164, + "step": 46362 + }, + { + "epoch": 1.78, + "learning_rate": 2.030456527314696e-05, + "loss": 0.3266, + "step": 46363 + }, + { + "epoch": 1.78, + "learning_rate": 2.030392408309823e-05, + "loss": 0.3126, + "step": 46364 + }, + { + "epoch": 1.78, + "learning_rate": 2.03032828930495e-05, + "loss": 0.3279, + "step": 46365 + }, + { + "epoch": 1.78, + "learning_rate": 2.030264170300077e-05, + "loss": 0.3447, + "step": 46366 + }, + { + "epoch": 1.78, + "learning_rate": 2.030200051295204e-05, + "loss": 0.3125, + "step": 46367 + }, + { + "epoch": 1.78, + "learning_rate": 2.030135932290331e-05, + "loss": 0.272, + "step": 46368 + }, + { + "epoch": 1.78, + "learning_rate": 2.030071813285458e-05, + "loss": 0.4243, + "step": 46369 + }, + { + "epoch": 1.78, + "learning_rate": 2.0300076942805848e-05, + "loss": 0.2749, + "step": 46370 + }, + { + "epoch": 1.78, + "learning_rate": 2.0299435752757116e-05, + "loss": 0.4036, + "step": 46371 + }, + { + "epoch": 1.78, + "learning_rate": 2.0298794562708388e-05, + "loss": 0.3933, + "step": 46372 + }, + { + "epoch": 1.78, + "learning_rate": 2.0298153372659656e-05, + "loss": 0.38, + "step": 46373 + }, + { + "epoch": 1.78, + "learning_rate": 2.0297512182610925e-05, + "loss": 0.4811, + "step": 46374 + }, + { + "epoch": 1.78, + "learning_rate": 2.0296870992562196e-05, + "loss": 0.3577, + "step": 46375 + }, + { + "epoch": 1.78, + "learning_rate": 2.0296229802513468e-05, + "loss": 0.311, + "step": 46376 + }, + { + "epoch": 1.78, + "learning_rate": 2.0295588612464736e-05, + "loss": 0.2957, + "step": 46377 + }, + { + "epoch": 1.78, + "learning_rate": 2.0294947422416005e-05, + "loss": 0.2205, + "step": 46378 + }, + { + "epoch": 1.78, + "learning_rate": 2.0294306232367273e-05, + "loss": 0.2309, + "step": 46379 + }, + { + "epoch": 1.78, + "learning_rate": 2.0293665042318545e-05, + "loss": 0.3454, + "step": 46380 + }, + { + "epoch": 1.78, + "learning_rate": 2.0293023852269813e-05, + "loss": 0.384, + "step": 46381 + }, + { + "epoch": 1.78, + "learning_rate": 2.029238266222108e-05, + "loss": 0.3778, + "step": 46382 + }, + { + "epoch": 1.78, + "learning_rate": 2.0291741472172353e-05, + "loss": 0.319, + "step": 46383 + }, + { + "epoch": 1.78, + "learning_rate": 2.0291100282123625e-05, + "loss": 0.3237, + "step": 46384 + }, + { + "epoch": 1.78, + "learning_rate": 2.0290459092074893e-05, + "loss": 0.3355, + "step": 46385 + }, + { + "epoch": 1.78, + "learning_rate": 2.028981790202616e-05, + "loss": 0.2501, + "step": 46386 + }, + { + "epoch": 1.78, + "learning_rate": 2.028917671197743e-05, + "loss": 0.3314, + "step": 46387 + }, + { + "epoch": 1.78, + "learning_rate": 2.02885355219287e-05, + "loss": 0.4131, + "step": 46388 + }, + { + "epoch": 1.78, + "learning_rate": 2.028789433187997e-05, + "loss": 0.3777, + "step": 46389 + }, + { + "epoch": 1.78, + "learning_rate": 2.0287253141831238e-05, + "loss": 0.3064, + "step": 46390 + }, + { + "epoch": 1.78, + "learning_rate": 2.0286611951782507e-05, + "loss": 0.2305, + "step": 46391 + }, + { + "epoch": 1.78, + "learning_rate": 2.028597076173378e-05, + "loss": 0.3022, + "step": 46392 + }, + { + "epoch": 1.78, + "learning_rate": 2.028532957168505e-05, + "loss": 0.3858, + "step": 46393 + }, + { + "epoch": 1.78, + "learning_rate": 2.0284688381636318e-05, + "loss": 0.5029, + "step": 46394 + }, + { + "epoch": 1.78, + "learning_rate": 2.0284047191587587e-05, + "loss": 0.3463, + "step": 46395 + }, + { + "epoch": 1.78, + "learning_rate": 2.028340600153886e-05, + "loss": 0.3228, + "step": 46396 + }, + { + "epoch": 1.78, + "learning_rate": 2.0282764811490127e-05, + "loss": 0.3126, + "step": 46397 + }, + { + "epoch": 1.78, + "learning_rate": 2.0282123621441395e-05, + "loss": 0.2794, + "step": 46398 + }, + { + "epoch": 1.78, + "learning_rate": 2.0281482431392663e-05, + "loss": 0.3825, + "step": 46399 + }, + { + "epoch": 1.78, + "learning_rate": 2.0280841241343935e-05, + "loss": 0.3671, + "step": 46400 + }, + { + "epoch": 1.78, + "learning_rate": 2.0280200051295207e-05, + "loss": 0.3427, + "step": 46401 + }, + { + "epoch": 1.78, + "learning_rate": 2.0279558861246475e-05, + "loss": 0.3143, + "step": 46402 + }, + { + "epoch": 1.78, + "learning_rate": 2.0278917671197743e-05, + "loss": 0.3401, + "step": 46403 + }, + { + "epoch": 1.78, + "learning_rate": 2.0278276481149015e-05, + "loss": 0.5012, + "step": 46404 + }, + { + "epoch": 1.78, + "learning_rate": 2.0277635291100283e-05, + "loss": 0.3497, + "step": 46405 + }, + { + "epoch": 1.78, + "learning_rate": 2.0276994101051552e-05, + "loss": 0.409, + "step": 46406 + }, + { + "epoch": 1.78, + "learning_rate": 2.027635291100282e-05, + "loss": 0.3538, + "step": 46407 + }, + { + "epoch": 1.78, + "learning_rate": 2.0275711720954092e-05, + "loss": 0.2998, + "step": 46408 + }, + { + "epoch": 1.78, + "learning_rate": 2.027507053090536e-05, + "loss": 0.3117, + "step": 46409 + }, + { + "epoch": 1.78, + "learning_rate": 2.0274429340856632e-05, + "loss": 0.3373, + "step": 46410 + }, + { + "epoch": 1.78, + "learning_rate": 2.02737881508079e-05, + "loss": 0.5135, + "step": 46411 + }, + { + "epoch": 1.78, + "learning_rate": 2.0273146960759172e-05, + "loss": 0.3104, + "step": 46412 + }, + { + "epoch": 1.78, + "learning_rate": 2.027250577071044e-05, + "loss": 0.2413, + "step": 46413 + }, + { + "epoch": 1.78, + "learning_rate": 2.027186458066171e-05, + "loss": 0.2549, + "step": 46414 + }, + { + "epoch": 1.78, + "learning_rate": 2.0271223390612977e-05, + "loss": 0.3334, + "step": 46415 + }, + { + "epoch": 1.78, + "learning_rate": 2.027058220056425e-05, + "loss": 0.4578, + "step": 46416 + }, + { + "epoch": 1.78, + "learning_rate": 2.0269941010515517e-05, + "loss": 0.422, + "step": 46417 + }, + { + "epoch": 1.78, + "learning_rate": 2.0269299820466785e-05, + "loss": 0.426, + "step": 46418 + }, + { + "epoch": 1.78, + "learning_rate": 2.0268658630418057e-05, + "loss": 0.3241, + "step": 46419 + }, + { + "epoch": 1.78, + "learning_rate": 2.026801744036933e-05, + "loss": 0.411, + "step": 46420 + }, + { + "epoch": 1.78, + "learning_rate": 2.0267376250320597e-05, + "loss": 0.3837, + "step": 46421 + }, + { + "epoch": 1.78, + "learning_rate": 2.0266735060271865e-05, + "loss": 0.357, + "step": 46422 + }, + { + "epoch": 1.78, + "learning_rate": 2.0266093870223134e-05, + "loss": 0.3645, + "step": 46423 + }, + { + "epoch": 1.78, + "learning_rate": 2.0265452680174405e-05, + "loss": 0.244, + "step": 46424 + }, + { + "epoch": 1.78, + "learning_rate": 2.0264811490125674e-05, + "loss": 0.2148, + "step": 46425 + }, + { + "epoch": 1.78, + "learning_rate": 2.0264170300076942e-05, + "loss": 0.3535, + "step": 46426 + }, + { + "epoch": 1.78, + "learning_rate": 2.026352911002821e-05, + "loss": 0.287, + "step": 46427 + }, + { + "epoch": 1.79, + "learning_rate": 2.0262887919979486e-05, + "loss": 0.2643, + "step": 46428 + }, + { + "epoch": 1.79, + "learning_rate": 2.0262246729930754e-05, + "loss": 0.2431, + "step": 46429 + }, + { + "epoch": 1.79, + "learning_rate": 2.0261605539882022e-05, + "loss": 0.2751, + "step": 46430 + }, + { + "epoch": 1.79, + "learning_rate": 2.026096434983329e-05, + "loss": 0.3357, + "step": 46431 + }, + { + "epoch": 1.79, + "learning_rate": 2.0260323159784562e-05, + "loss": 0.2907, + "step": 46432 + }, + { + "epoch": 1.79, + "learning_rate": 2.025968196973583e-05, + "loss": 0.3237, + "step": 46433 + }, + { + "epoch": 1.79, + "learning_rate": 2.02590407796871e-05, + "loss": 0.37, + "step": 46434 + }, + { + "epoch": 1.79, + "learning_rate": 2.0258399589638367e-05, + "loss": 0.2793, + "step": 46435 + }, + { + "epoch": 1.79, + "learning_rate": 2.025775839958964e-05, + "loss": 0.2857, + "step": 46436 + }, + { + "epoch": 1.79, + "learning_rate": 2.025711720954091e-05, + "loss": 0.2374, + "step": 46437 + }, + { + "epoch": 1.79, + "learning_rate": 2.025647601949218e-05, + "loss": 0.3613, + "step": 46438 + }, + { + "epoch": 1.79, + "learning_rate": 2.0255834829443447e-05, + "loss": 0.4216, + "step": 46439 + }, + { + "epoch": 1.79, + "learning_rate": 2.025519363939472e-05, + "loss": 0.2951, + "step": 46440 + }, + { + "epoch": 1.79, + "learning_rate": 2.0254552449345987e-05, + "loss": 0.4437, + "step": 46441 + }, + { + "epoch": 1.79, + "learning_rate": 2.0253911259297256e-05, + "loss": 0.3736, + "step": 46442 + }, + { + "epoch": 1.79, + "learning_rate": 2.0253270069248524e-05, + "loss": 0.283, + "step": 46443 + }, + { + "epoch": 1.79, + "learning_rate": 2.0252628879199796e-05, + "loss": 0.4352, + "step": 46444 + }, + { + "epoch": 1.79, + "learning_rate": 2.0251987689151064e-05, + "loss": 0.284, + "step": 46445 + }, + { + "epoch": 1.79, + "learning_rate": 2.0251346499102336e-05, + "loss": 0.3437, + "step": 46446 + }, + { + "epoch": 1.79, + "learning_rate": 2.0250705309053604e-05, + "loss": 0.2255, + "step": 46447 + }, + { + "epoch": 1.79, + "learning_rate": 2.0250064119004876e-05, + "loss": 0.2506, + "step": 46448 + }, + { + "epoch": 1.79, + "learning_rate": 2.0249422928956144e-05, + "loss": 0.3581, + "step": 46449 + }, + { + "epoch": 1.79, + "learning_rate": 2.0248781738907412e-05, + "loss": 0.2998, + "step": 46450 + }, + { + "epoch": 1.79, + "learning_rate": 2.024814054885868e-05, + "loss": 0.3953, + "step": 46451 + }, + { + "epoch": 1.79, + "learning_rate": 2.0247499358809952e-05, + "loss": 0.4155, + "step": 46452 + }, + { + "epoch": 1.79, + "learning_rate": 2.024685816876122e-05, + "loss": 0.3593, + "step": 46453 + }, + { + "epoch": 1.79, + "learning_rate": 2.024621697871249e-05, + "loss": 0.2902, + "step": 46454 + }, + { + "epoch": 1.79, + "learning_rate": 2.024557578866376e-05, + "loss": 0.3051, + "step": 46455 + }, + { + "epoch": 1.79, + "learning_rate": 2.0244934598615033e-05, + "loss": 0.3261, + "step": 46456 + }, + { + "epoch": 1.79, + "learning_rate": 2.02442934085663e-05, + "loss": 0.3992, + "step": 46457 + }, + { + "epoch": 1.79, + "learning_rate": 2.024365221851757e-05, + "loss": 0.3081, + "step": 46458 + }, + { + "epoch": 1.79, + "learning_rate": 2.0243011028468838e-05, + "loss": 0.4231, + "step": 46459 + }, + { + "epoch": 1.79, + "learning_rate": 2.024236983842011e-05, + "loss": 0.2278, + "step": 46460 + }, + { + "epoch": 1.79, + "learning_rate": 2.0241728648371378e-05, + "loss": 0.2489, + "step": 46461 + }, + { + "epoch": 1.79, + "learning_rate": 2.0241087458322646e-05, + "loss": 0.3113, + "step": 46462 + }, + { + "epoch": 1.79, + "learning_rate": 2.0240446268273918e-05, + "loss": 0.322, + "step": 46463 + }, + { + "epoch": 1.79, + "learning_rate": 2.023980507822519e-05, + "loss": 0.2366, + "step": 46464 + }, + { + "epoch": 1.79, + "learning_rate": 2.0239163888176458e-05, + "loss": 0.2814, + "step": 46465 + }, + { + "epoch": 1.79, + "learning_rate": 2.0238522698127726e-05, + "loss": 0.3255, + "step": 46466 + }, + { + "epoch": 1.79, + "learning_rate": 2.0237881508078994e-05, + "loss": 0.3005, + "step": 46467 + }, + { + "epoch": 1.79, + "learning_rate": 2.0237240318030266e-05, + "loss": 0.3003, + "step": 46468 + }, + { + "epoch": 1.79, + "learning_rate": 2.0236599127981534e-05, + "loss": 0.3802, + "step": 46469 + }, + { + "epoch": 1.79, + "learning_rate": 2.0235957937932803e-05, + "loss": 0.2218, + "step": 46470 + }, + { + "epoch": 1.79, + "learning_rate": 2.023531674788407e-05, + "loss": 0.3081, + "step": 46471 + }, + { + "epoch": 1.79, + "learning_rate": 2.0234675557835346e-05, + "loss": 0.3111, + "step": 46472 + }, + { + "epoch": 1.79, + "learning_rate": 2.0234034367786615e-05, + "loss": 0.3711, + "step": 46473 + }, + { + "epoch": 1.79, + "learning_rate": 2.0233393177737883e-05, + "loss": 0.3532, + "step": 46474 + }, + { + "epoch": 1.79, + "learning_rate": 2.023275198768915e-05, + "loss": 0.354, + "step": 46475 + }, + { + "epoch": 1.79, + "learning_rate": 2.0232110797640423e-05, + "loss": 0.3659, + "step": 46476 + }, + { + "epoch": 1.79, + "learning_rate": 2.023146960759169e-05, + "loss": 0.3229, + "step": 46477 + }, + { + "epoch": 1.79, + "learning_rate": 2.023082841754296e-05, + "loss": 0.3495, + "step": 46478 + }, + { + "epoch": 1.79, + "learning_rate": 2.0230187227494228e-05, + "loss": 0.3925, + "step": 46479 + }, + { + "epoch": 1.79, + "learning_rate": 2.02295460374455e-05, + "loss": 0.2062, + "step": 46480 + }, + { + "epoch": 1.79, + "learning_rate": 2.022890484739677e-05, + "loss": 0.2887, + "step": 46481 + }, + { + "epoch": 1.79, + "learning_rate": 2.022826365734804e-05, + "loss": 0.3607, + "step": 46482 + }, + { + "epoch": 1.79, + "learning_rate": 2.0227622467299308e-05, + "loss": 0.3443, + "step": 46483 + }, + { + "epoch": 1.79, + "learning_rate": 2.022698127725058e-05, + "loss": 0.4382, + "step": 46484 + }, + { + "epoch": 1.79, + "learning_rate": 2.0226340087201848e-05, + "loss": 0.2607, + "step": 46485 + }, + { + "epoch": 1.79, + "learning_rate": 2.0225698897153116e-05, + "loss": 0.2861, + "step": 46486 + }, + { + "epoch": 1.79, + "learning_rate": 2.0225057707104385e-05, + "loss": 0.4277, + "step": 46487 + }, + { + "epoch": 1.79, + "learning_rate": 2.0224416517055656e-05, + "loss": 0.3913, + "step": 46488 + }, + { + "epoch": 1.79, + "learning_rate": 2.0223775327006925e-05, + "loss": 0.3191, + "step": 46489 + }, + { + "epoch": 1.79, + "learning_rate": 2.0223134136958196e-05, + "loss": 0.215, + "step": 46490 + }, + { + "epoch": 1.79, + "learning_rate": 2.0222492946909465e-05, + "loss": 0.3299, + "step": 46491 + }, + { + "epoch": 1.79, + "learning_rate": 2.0221851756860736e-05, + "loss": 0.3561, + "step": 46492 + }, + { + "epoch": 1.79, + "learning_rate": 2.0221210566812005e-05, + "loss": 0.4835, + "step": 46493 + }, + { + "epoch": 1.79, + "learning_rate": 2.0220569376763273e-05, + "loss": 0.3333, + "step": 46494 + }, + { + "epoch": 1.79, + "learning_rate": 2.021992818671454e-05, + "loss": 0.3002, + "step": 46495 + }, + { + "epoch": 1.79, + "learning_rate": 2.0219286996665813e-05, + "loss": 0.3975, + "step": 46496 + }, + { + "epoch": 1.79, + "learning_rate": 2.021864580661708e-05, + "loss": 0.3176, + "step": 46497 + }, + { + "epoch": 1.79, + "learning_rate": 2.021800461656835e-05, + "loss": 0.347, + "step": 46498 + }, + { + "epoch": 1.79, + "learning_rate": 2.021736342651962e-05, + "loss": 0.34, + "step": 46499 + }, + { + "epoch": 1.79, + "learning_rate": 2.0216722236470893e-05, + "loss": 0.2842, + "step": 46500 + }, + { + "epoch": 1.79, + "learning_rate": 2.021608104642216e-05, + "loss": 0.3058, + "step": 46501 + }, + { + "epoch": 1.79, + "learning_rate": 2.021543985637343e-05, + "loss": 0.3541, + "step": 46502 + }, + { + "epoch": 1.79, + "learning_rate": 2.0214798666324698e-05, + "loss": 0.2776, + "step": 46503 + }, + { + "epoch": 1.79, + "learning_rate": 2.021415747627597e-05, + "loss": 0.2636, + "step": 46504 + }, + { + "epoch": 1.79, + "learning_rate": 2.0213516286227238e-05, + "loss": 0.2487, + "step": 46505 + }, + { + "epoch": 1.79, + "learning_rate": 2.0212875096178507e-05, + "loss": 0.2949, + "step": 46506 + }, + { + "epoch": 1.79, + "learning_rate": 2.0212233906129775e-05, + "loss": 0.4205, + "step": 46507 + }, + { + "epoch": 1.79, + "learning_rate": 2.021159271608105e-05, + "loss": 0.2056, + "step": 46508 + }, + { + "epoch": 1.79, + "learning_rate": 2.021095152603232e-05, + "loss": 0.4169, + "step": 46509 + }, + { + "epoch": 1.79, + "learning_rate": 2.0210310335983587e-05, + "loss": 0.3774, + "step": 46510 + }, + { + "epoch": 1.79, + "learning_rate": 2.0209669145934855e-05, + "loss": 0.3329, + "step": 46511 + }, + { + "epoch": 1.79, + "learning_rate": 2.0209027955886127e-05, + "loss": 0.3278, + "step": 46512 + }, + { + "epoch": 1.79, + "learning_rate": 2.0208386765837395e-05, + "loss": 0.356, + "step": 46513 + }, + { + "epoch": 1.79, + "learning_rate": 2.0207745575788663e-05, + "loss": 0.465, + "step": 46514 + }, + { + "epoch": 1.79, + "learning_rate": 2.0207104385739932e-05, + "loss": 0.4057, + "step": 46515 + }, + { + "epoch": 1.79, + "learning_rate": 2.0206463195691203e-05, + "loss": 0.3802, + "step": 46516 + }, + { + "epoch": 1.79, + "learning_rate": 2.0205822005642475e-05, + "loss": 0.2527, + "step": 46517 + }, + { + "epoch": 1.79, + "learning_rate": 2.0205180815593743e-05, + "loss": 0.2358, + "step": 46518 + }, + { + "epoch": 1.79, + "learning_rate": 2.0204539625545012e-05, + "loss": 0.3531, + "step": 46519 + }, + { + "epoch": 1.79, + "learning_rate": 2.0203898435496284e-05, + "loss": 0.3442, + "step": 46520 + }, + { + "epoch": 1.79, + "learning_rate": 2.0203257245447552e-05, + "loss": 0.3344, + "step": 46521 + }, + { + "epoch": 1.79, + "learning_rate": 2.020261605539882e-05, + "loss": 0.3358, + "step": 46522 + }, + { + "epoch": 1.79, + "learning_rate": 2.020197486535009e-05, + "loss": 0.3971, + "step": 46523 + }, + { + "epoch": 1.79, + "learning_rate": 2.020133367530136e-05, + "loss": 0.3008, + "step": 46524 + }, + { + "epoch": 1.79, + "learning_rate": 2.020069248525263e-05, + "loss": 0.3284, + "step": 46525 + }, + { + "epoch": 1.79, + "learning_rate": 2.02000512952039e-05, + "loss": 0.3503, + "step": 46526 + }, + { + "epoch": 1.79, + "learning_rate": 2.019941010515517e-05, + "loss": 0.4125, + "step": 46527 + }, + { + "epoch": 1.79, + "learning_rate": 2.019876891510644e-05, + "loss": 0.2676, + "step": 46528 + }, + { + "epoch": 1.79, + "learning_rate": 2.019812772505771e-05, + "loss": 0.4013, + "step": 46529 + }, + { + "epoch": 1.79, + "learning_rate": 2.0197486535008977e-05, + "loss": 0.3059, + "step": 46530 + }, + { + "epoch": 1.79, + "learning_rate": 2.0196845344960245e-05, + "loss": 0.3197, + "step": 46531 + }, + { + "epoch": 1.79, + "learning_rate": 2.0196204154911517e-05, + "loss": 0.3556, + "step": 46532 + }, + { + "epoch": 1.79, + "learning_rate": 2.0195562964862785e-05, + "loss": 0.3156, + "step": 46533 + }, + { + "epoch": 1.79, + "learning_rate": 2.0194921774814054e-05, + "loss": 0.2795, + "step": 46534 + }, + { + "epoch": 1.79, + "learning_rate": 2.0194280584765325e-05, + "loss": 0.4808, + "step": 46535 + }, + { + "epoch": 1.79, + "learning_rate": 2.0193639394716597e-05, + "loss": 0.2172, + "step": 46536 + }, + { + "epoch": 1.79, + "learning_rate": 2.0192998204667865e-05, + "loss": 0.2967, + "step": 46537 + }, + { + "epoch": 1.79, + "learning_rate": 2.0192357014619134e-05, + "loss": 0.2518, + "step": 46538 + }, + { + "epoch": 1.79, + "learning_rate": 2.0191715824570402e-05, + "loss": 0.3447, + "step": 46539 + }, + { + "epoch": 1.79, + "learning_rate": 2.0191074634521674e-05, + "loss": 0.3059, + "step": 46540 + }, + { + "epoch": 1.79, + "learning_rate": 2.0190433444472942e-05, + "loss": 0.2636, + "step": 46541 + }, + { + "epoch": 1.79, + "learning_rate": 2.018979225442421e-05, + "loss": 0.2379, + "step": 46542 + }, + { + "epoch": 1.79, + "learning_rate": 2.0189151064375482e-05, + "loss": 0.3063, + "step": 46543 + }, + { + "epoch": 1.79, + "learning_rate": 2.0188509874326754e-05, + "loss": 0.3703, + "step": 46544 + }, + { + "epoch": 1.79, + "learning_rate": 2.0187868684278022e-05, + "loss": 0.3507, + "step": 46545 + }, + { + "epoch": 1.79, + "learning_rate": 2.018722749422929e-05, + "loss": 0.4348, + "step": 46546 + }, + { + "epoch": 1.79, + "learning_rate": 2.018658630418056e-05, + "loss": 0.2897, + "step": 46547 + }, + { + "epoch": 1.79, + "learning_rate": 2.018594511413183e-05, + "loss": 0.2869, + "step": 46548 + }, + { + "epoch": 1.79, + "learning_rate": 2.01853039240831e-05, + "loss": 0.2571, + "step": 46549 + }, + { + "epoch": 1.79, + "learning_rate": 2.0184662734034367e-05, + "loss": 0.2944, + "step": 46550 + }, + { + "epoch": 1.79, + "learning_rate": 2.0184021543985636e-05, + "loss": 0.3709, + "step": 46551 + }, + { + "epoch": 1.79, + "learning_rate": 2.0183380353936907e-05, + "loss": 0.3095, + "step": 46552 + }, + { + "epoch": 1.79, + "learning_rate": 2.018273916388818e-05, + "loss": 0.309, + "step": 46553 + }, + { + "epoch": 1.79, + "learning_rate": 2.0182097973839447e-05, + "loss": 0.4021, + "step": 46554 + }, + { + "epoch": 1.79, + "learning_rate": 2.0181456783790716e-05, + "loss": 0.2554, + "step": 46555 + }, + { + "epoch": 1.79, + "learning_rate": 2.0180815593741987e-05, + "loss": 0.2389, + "step": 46556 + }, + { + "epoch": 1.79, + "learning_rate": 2.0180174403693256e-05, + "loss": 0.2602, + "step": 46557 + }, + { + "epoch": 1.79, + "learning_rate": 2.0179533213644524e-05, + "loss": 0.2759, + "step": 46558 + }, + { + "epoch": 1.79, + "learning_rate": 2.0178892023595792e-05, + "loss": 0.26, + "step": 46559 + }, + { + "epoch": 1.79, + "learning_rate": 2.0178250833547064e-05, + "loss": 0.3493, + "step": 46560 + }, + { + "epoch": 1.79, + "learning_rate": 2.0177609643498332e-05, + "loss": 0.2779, + "step": 46561 + }, + { + "epoch": 1.79, + "learning_rate": 2.0176968453449604e-05, + "loss": 0.2863, + "step": 46562 + }, + { + "epoch": 1.79, + "learning_rate": 2.0176327263400872e-05, + "loss": 0.281, + "step": 46563 + }, + { + "epoch": 1.79, + "learning_rate": 2.0175686073352144e-05, + "loss": 0.2657, + "step": 46564 + }, + { + "epoch": 1.79, + "learning_rate": 2.0175044883303413e-05, + "loss": 0.3525, + "step": 46565 + }, + { + "epoch": 1.79, + "learning_rate": 2.017440369325468e-05, + "loss": 0.2855, + "step": 46566 + }, + { + "epoch": 1.79, + "learning_rate": 2.017376250320595e-05, + "loss": 0.3616, + "step": 46567 + }, + { + "epoch": 1.79, + "learning_rate": 2.017312131315722e-05, + "loss": 0.2764, + "step": 46568 + }, + { + "epoch": 1.79, + "learning_rate": 2.017248012310849e-05, + "loss": 0.3008, + "step": 46569 + }, + { + "epoch": 1.79, + "learning_rate": 2.017183893305976e-05, + "loss": 0.3376, + "step": 46570 + }, + { + "epoch": 1.79, + "learning_rate": 2.017119774301103e-05, + "loss": 0.3598, + "step": 46571 + }, + { + "epoch": 1.79, + "learning_rate": 2.01705565529623e-05, + "loss": 0.3262, + "step": 46572 + }, + { + "epoch": 1.79, + "learning_rate": 2.016991536291357e-05, + "loss": 0.3463, + "step": 46573 + }, + { + "epoch": 1.79, + "learning_rate": 2.0169274172864838e-05, + "loss": 0.3144, + "step": 46574 + }, + { + "epoch": 1.79, + "learning_rate": 2.0168632982816106e-05, + "loss": 0.3225, + "step": 46575 + }, + { + "epoch": 1.79, + "learning_rate": 2.0167991792767378e-05, + "loss": 0.2663, + "step": 46576 + }, + { + "epoch": 1.79, + "learning_rate": 2.0167350602718646e-05, + "loss": 0.3328, + "step": 46577 + }, + { + "epoch": 1.79, + "learning_rate": 2.0166709412669914e-05, + "loss": 0.2732, + "step": 46578 + }, + { + "epoch": 1.79, + "learning_rate": 2.0166068222621186e-05, + "loss": 0.3828, + "step": 46579 + }, + { + "epoch": 1.79, + "learning_rate": 2.0165427032572458e-05, + "loss": 0.3552, + "step": 46580 + }, + { + "epoch": 1.79, + "learning_rate": 2.0164785842523726e-05, + "loss": 0.2807, + "step": 46581 + }, + { + "epoch": 1.79, + "learning_rate": 2.0164144652474994e-05, + "loss": 0.4106, + "step": 46582 + }, + { + "epoch": 1.79, + "learning_rate": 2.0163503462426263e-05, + "loss": 0.5679, + "step": 46583 + }, + { + "epoch": 1.79, + "learning_rate": 2.0162862272377534e-05, + "loss": 0.4411, + "step": 46584 + }, + { + "epoch": 1.79, + "learning_rate": 2.0162221082328803e-05, + "loss": 0.3356, + "step": 46585 + }, + { + "epoch": 1.79, + "learning_rate": 2.016157989228007e-05, + "loss": 0.4666, + "step": 46586 + }, + { + "epoch": 1.79, + "learning_rate": 2.0160938702231343e-05, + "loss": 0.2634, + "step": 46587 + }, + { + "epoch": 1.79, + "learning_rate": 2.0160297512182615e-05, + "loss": 0.3109, + "step": 46588 + }, + { + "epoch": 1.79, + "learning_rate": 2.0159656322133883e-05, + "loss": 0.1845, + "step": 46589 + }, + { + "epoch": 1.79, + "learning_rate": 2.015901513208515e-05, + "loss": 0.3471, + "step": 46590 + }, + { + "epoch": 1.79, + "learning_rate": 2.015837394203642e-05, + "loss": 0.2775, + "step": 46591 + }, + { + "epoch": 1.79, + "learning_rate": 2.015773275198769e-05, + "loss": 0.2794, + "step": 46592 + }, + { + "epoch": 1.79, + "learning_rate": 2.015709156193896e-05, + "loss": 0.4235, + "step": 46593 + }, + { + "epoch": 1.79, + "learning_rate": 2.0156450371890228e-05, + "loss": 0.4633, + "step": 46594 + }, + { + "epoch": 1.79, + "learning_rate": 2.0155809181841496e-05, + "loss": 0.3902, + "step": 46595 + }, + { + "epoch": 1.79, + "learning_rate": 2.0155167991792768e-05, + "loss": 0.3361, + "step": 46596 + }, + { + "epoch": 1.79, + "learning_rate": 2.015452680174404e-05, + "loss": 0.3366, + "step": 46597 + }, + { + "epoch": 1.79, + "learning_rate": 2.0153885611695308e-05, + "loss": 0.3381, + "step": 46598 + }, + { + "epoch": 1.79, + "learning_rate": 2.0153244421646576e-05, + "loss": 0.2547, + "step": 46599 + }, + { + "epoch": 1.79, + "learning_rate": 2.0152603231597848e-05, + "loss": 0.2886, + "step": 46600 + }, + { + "epoch": 1.79, + "learning_rate": 2.0151962041549116e-05, + "loss": 0.2355, + "step": 46601 + }, + { + "epoch": 1.79, + "learning_rate": 2.0151320851500385e-05, + "loss": 0.3262, + "step": 46602 + }, + { + "epoch": 1.79, + "learning_rate": 2.0150679661451653e-05, + "loss": 0.3118, + "step": 46603 + }, + { + "epoch": 1.79, + "learning_rate": 2.0150038471402925e-05, + "loss": 0.3721, + "step": 46604 + }, + { + "epoch": 1.79, + "learning_rate": 2.0149397281354193e-05, + "loss": 0.3301, + "step": 46605 + }, + { + "epoch": 1.79, + "learning_rate": 2.0148756091305465e-05, + "loss": 0.3473, + "step": 46606 + }, + { + "epoch": 1.79, + "learning_rate": 2.0148114901256733e-05, + "loss": 0.3195, + "step": 46607 + }, + { + "epoch": 1.79, + "learning_rate": 2.0147473711208005e-05, + "loss": 0.3325, + "step": 46608 + }, + { + "epoch": 1.79, + "learning_rate": 2.0146832521159273e-05, + "loss": 0.3031, + "step": 46609 + }, + { + "epoch": 1.79, + "learning_rate": 2.014619133111054e-05, + "loss": 0.1492, + "step": 46610 + }, + { + "epoch": 1.79, + "learning_rate": 2.014555014106181e-05, + "loss": 0.4156, + "step": 46611 + }, + { + "epoch": 1.79, + "learning_rate": 2.014490895101308e-05, + "loss": 0.343, + "step": 46612 + }, + { + "epoch": 1.79, + "learning_rate": 2.014426776096435e-05, + "loss": 0.3116, + "step": 46613 + }, + { + "epoch": 1.79, + "learning_rate": 2.0143626570915618e-05, + "loss": 0.334, + "step": 46614 + }, + { + "epoch": 1.79, + "learning_rate": 2.014298538086689e-05, + "loss": 0.3009, + "step": 46615 + }, + { + "epoch": 1.79, + "learning_rate": 2.014234419081816e-05, + "loss": 0.3734, + "step": 46616 + }, + { + "epoch": 1.79, + "learning_rate": 2.014170300076943e-05, + "loss": 0.4043, + "step": 46617 + }, + { + "epoch": 1.79, + "learning_rate": 2.01410618107207e-05, + "loss": 0.318, + "step": 46618 + }, + { + "epoch": 1.79, + "learning_rate": 2.0140420620671967e-05, + "loss": 0.4473, + "step": 46619 + }, + { + "epoch": 1.79, + "learning_rate": 2.013977943062324e-05, + "loss": 0.2856, + "step": 46620 + }, + { + "epoch": 1.79, + "learning_rate": 2.0139138240574507e-05, + "loss": 0.3624, + "step": 46621 + }, + { + "epoch": 1.79, + "learning_rate": 2.0138497050525775e-05, + "loss": 0.4201, + "step": 46622 + }, + { + "epoch": 1.79, + "learning_rate": 2.0137855860477047e-05, + "loss": 0.3955, + "step": 46623 + }, + { + "epoch": 1.79, + "learning_rate": 2.013721467042832e-05, + "loss": 0.4296, + "step": 46624 + }, + { + "epoch": 1.79, + "learning_rate": 2.0136573480379587e-05, + "loss": 0.3066, + "step": 46625 + }, + { + "epoch": 1.79, + "learning_rate": 2.0135932290330855e-05, + "loss": 0.3108, + "step": 46626 + }, + { + "epoch": 1.79, + "learning_rate": 2.0135291100282123e-05, + "loss": 0.2712, + "step": 46627 + }, + { + "epoch": 1.79, + "learning_rate": 2.0134649910233395e-05, + "loss": 0.3937, + "step": 46628 + }, + { + "epoch": 1.79, + "learning_rate": 2.0134008720184663e-05, + "loss": 0.2661, + "step": 46629 + }, + { + "epoch": 1.79, + "learning_rate": 2.0133367530135932e-05, + "loss": 0.2984, + "step": 46630 + }, + { + "epoch": 1.79, + "learning_rate": 2.01327263400872e-05, + "loss": 0.4216, + "step": 46631 + }, + { + "epoch": 1.79, + "learning_rate": 2.0132085150038472e-05, + "loss": 0.3243, + "step": 46632 + }, + { + "epoch": 1.79, + "learning_rate": 2.0131443959989744e-05, + "loss": 0.1846, + "step": 46633 + }, + { + "epoch": 1.79, + "learning_rate": 2.0130802769941012e-05, + "loss": 0.3753, + "step": 46634 + }, + { + "epoch": 1.79, + "learning_rate": 2.013016157989228e-05, + "loss": 0.2727, + "step": 46635 + }, + { + "epoch": 1.79, + "learning_rate": 2.0129520389843552e-05, + "loss": 0.3124, + "step": 46636 + }, + { + "epoch": 1.79, + "learning_rate": 2.012887919979482e-05, + "loss": 0.1895, + "step": 46637 + }, + { + "epoch": 1.79, + "learning_rate": 2.012823800974609e-05, + "loss": 0.3226, + "step": 46638 + }, + { + "epoch": 1.79, + "learning_rate": 2.0127596819697357e-05, + "loss": 0.4175, + "step": 46639 + }, + { + "epoch": 1.79, + "learning_rate": 2.012695562964863e-05, + "loss": 0.3549, + "step": 46640 + }, + { + "epoch": 1.79, + "learning_rate": 2.0126314439599897e-05, + "loss": 0.2437, + "step": 46641 + }, + { + "epoch": 1.79, + "learning_rate": 2.012567324955117e-05, + "loss": 0.3795, + "step": 46642 + }, + { + "epoch": 1.79, + "learning_rate": 2.0125032059502437e-05, + "loss": 0.2733, + "step": 46643 + }, + { + "epoch": 1.79, + "learning_rate": 2.012439086945371e-05, + "loss": 0.2883, + "step": 46644 + }, + { + "epoch": 1.79, + "learning_rate": 2.0123749679404977e-05, + "loss": 0.2611, + "step": 46645 + }, + { + "epoch": 1.79, + "learning_rate": 2.0123108489356245e-05, + "loss": 0.3806, + "step": 46646 + }, + { + "epoch": 1.79, + "learning_rate": 2.0122467299307514e-05, + "loss": 0.4097, + "step": 46647 + }, + { + "epoch": 1.79, + "learning_rate": 2.0121826109258785e-05, + "loss": 0.3554, + "step": 46648 + }, + { + "epoch": 1.79, + "learning_rate": 2.0121184919210054e-05, + "loss": 0.3948, + "step": 46649 + }, + { + "epoch": 1.79, + "learning_rate": 2.0120543729161322e-05, + "loss": 0.2166, + "step": 46650 + }, + { + "epoch": 1.79, + "learning_rate": 2.0119902539112594e-05, + "loss": 0.3222, + "step": 46651 + }, + { + "epoch": 1.79, + "learning_rate": 2.0119261349063866e-05, + "loss": 0.2005, + "step": 46652 + }, + { + "epoch": 1.79, + "learning_rate": 2.0118620159015134e-05, + "loss": 0.4897, + "step": 46653 + }, + { + "epoch": 1.79, + "learning_rate": 2.0117978968966402e-05, + "loss": 0.235, + "step": 46654 + }, + { + "epoch": 1.79, + "learning_rate": 2.011733777891767e-05, + "loss": 0.1937, + "step": 46655 + }, + { + "epoch": 1.79, + "learning_rate": 2.0116696588868942e-05, + "loss": 0.3625, + "step": 46656 + }, + { + "epoch": 1.79, + "learning_rate": 2.011605539882021e-05, + "loss": 0.3234, + "step": 46657 + }, + { + "epoch": 1.79, + "learning_rate": 2.011541420877148e-05, + "loss": 0.3737, + "step": 46658 + }, + { + "epoch": 1.79, + "learning_rate": 2.011477301872275e-05, + "loss": 0.4013, + "step": 46659 + }, + { + "epoch": 1.79, + "learning_rate": 2.0114131828674022e-05, + "loss": 0.3027, + "step": 46660 + }, + { + "epoch": 1.79, + "learning_rate": 2.011349063862529e-05, + "loss": 0.3668, + "step": 46661 + }, + { + "epoch": 1.79, + "learning_rate": 2.011284944857656e-05, + "loss": 0.3452, + "step": 46662 + }, + { + "epoch": 1.79, + "learning_rate": 2.0112208258527827e-05, + "loss": 0.4212, + "step": 46663 + }, + { + "epoch": 1.79, + "learning_rate": 2.01115670684791e-05, + "loss": 0.3053, + "step": 46664 + }, + { + "epoch": 1.79, + "learning_rate": 2.0110925878430367e-05, + "loss": 0.3546, + "step": 46665 + }, + { + "epoch": 1.79, + "learning_rate": 2.0110284688381636e-05, + "loss": 0.3191, + "step": 46666 + }, + { + "epoch": 1.79, + "learning_rate": 2.0109643498332907e-05, + "loss": 0.384, + "step": 46667 + }, + { + "epoch": 1.79, + "learning_rate": 2.010900230828418e-05, + "loss": 0.2504, + "step": 46668 + }, + { + "epoch": 1.79, + "learning_rate": 2.0108361118235447e-05, + "loss": 0.3208, + "step": 46669 + }, + { + "epoch": 1.79, + "learning_rate": 2.0107719928186716e-05, + "loss": 0.5743, + "step": 46670 + }, + { + "epoch": 1.79, + "learning_rate": 2.0107078738137984e-05, + "loss": 0.3612, + "step": 46671 + }, + { + "epoch": 1.79, + "learning_rate": 2.0106437548089256e-05, + "loss": 0.4338, + "step": 46672 + }, + { + "epoch": 1.79, + "learning_rate": 2.0105796358040524e-05, + "loss": 0.3778, + "step": 46673 + }, + { + "epoch": 1.79, + "learning_rate": 2.0105155167991792e-05, + "loss": 0.3184, + "step": 46674 + }, + { + "epoch": 1.79, + "learning_rate": 2.010451397794306e-05, + "loss": 0.2211, + "step": 46675 + }, + { + "epoch": 1.79, + "learning_rate": 2.0103872787894333e-05, + "loss": 0.2764, + "step": 46676 + }, + { + "epoch": 1.79, + "learning_rate": 2.0103231597845604e-05, + "loss": 0.3795, + "step": 46677 + }, + { + "epoch": 1.79, + "learning_rate": 2.0102590407796873e-05, + "loss": 0.3986, + "step": 46678 + }, + { + "epoch": 1.79, + "learning_rate": 2.010194921774814e-05, + "loss": 0.2509, + "step": 46679 + }, + { + "epoch": 1.79, + "learning_rate": 2.0101308027699413e-05, + "loss": 0.3289, + "step": 46680 + }, + { + "epoch": 1.79, + "learning_rate": 2.010066683765068e-05, + "loss": 0.2987, + "step": 46681 + }, + { + "epoch": 1.79, + "learning_rate": 2.010002564760195e-05, + "loss": 0.2648, + "step": 46682 + }, + { + "epoch": 1.79, + "learning_rate": 2.0099384457553218e-05, + "loss": 0.4415, + "step": 46683 + }, + { + "epoch": 1.79, + "learning_rate": 2.009874326750449e-05, + "loss": 0.3842, + "step": 46684 + }, + { + "epoch": 1.79, + "learning_rate": 2.0098102077455758e-05, + "loss": 0.4278, + "step": 46685 + }, + { + "epoch": 1.79, + "learning_rate": 2.009746088740703e-05, + "loss": 0.3191, + "step": 46686 + }, + { + "epoch": 1.79, + "learning_rate": 2.0096819697358298e-05, + "loss": 0.2655, + "step": 46687 + }, + { + "epoch": 1.8, + "learning_rate": 2.009617850730957e-05, + "loss": 0.2396, + "step": 46688 + }, + { + "epoch": 1.8, + "learning_rate": 2.0095537317260838e-05, + "loss": 0.2976, + "step": 46689 + }, + { + "epoch": 1.8, + "learning_rate": 2.0094896127212106e-05, + "loss": 0.4026, + "step": 46690 + }, + { + "epoch": 1.8, + "learning_rate": 2.0094254937163374e-05, + "loss": 0.3429, + "step": 46691 + }, + { + "epoch": 1.8, + "learning_rate": 2.0093613747114646e-05, + "loss": 0.33, + "step": 46692 + }, + { + "epoch": 1.8, + "learning_rate": 2.0092972557065914e-05, + "loss": 0.3052, + "step": 46693 + }, + { + "epoch": 1.8, + "learning_rate": 2.0092331367017183e-05, + "loss": 0.3186, + "step": 46694 + }, + { + "epoch": 1.8, + "learning_rate": 2.0091690176968454e-05, + "loss": 0.3228, + "step": 46695 + }, + { + "epoch": 1.8, + "learning_rate": 2.0091048986919726e-05, + "loss": 0.4035, + "step": 46696 + }, + { + "epoch": 1.8, + "learning_rate": 2.0090407796870995e-05, + "loss": 0.3461, + "step": 46697 + }, + { + "epoch": 1.8, + "learning_rate": 2.0089766606822263e-05, + "loss": 0.3395, + "step": 46698 + }, + { + "epoch": 1.8, + "learning_rate": 2.008912541677353e-05, + "loss": 0.2709, + "step": 46699 + }, + { + "epoch": 1.8, + "learning_rate": 2.0088484226724803e-05, + "loss": 0.2832, + "step": 46700 + }, + { + "epoch": 1.8, + "learning_rate": 2.008784303667607e-05, + "loss": 0.3862, + "step": 46701 + }, + { + "epoch": 1.8, + "learning_rate": 2.008720184662734e-05, + "loss": 0.3876, + "step": 46702 + }, + { + "epoch": 1.8, + "learning_rate": 2.008656065657861e-05, + "loss": 0.2211, + "step": 46703 + }, + { + "epoch": 1.8, + "learning_rate": 2.0085919466529883e-05, + "loss": 0.2452, + "step": 46704 + }, + { + "epoch": 1.8, + "learning_rate": 2.008527827648115e-05, + "loss": 0.312, + "step": 46705 + }, + { + "epoch": 1.8, + "learning_rate": 2.008463708643242e-05, + "loss": 0.4326, + "step": 46706 + }, + { + "epoch": 1.8, + "learning_rate": 2.0083995896383688e-05, + "loss": 0.3083, + "step": 46707 + }, + { + "epoch": 1.8, + "learning_rate": 2.008335470633496e-05, + "loss": 0.2098, + "step": 46708 + }, + { + "epoch": 1.8, + "learning_rate": 2.0082713516286228e-05, + "loss": 0.2524, + "step": 46709 + }, + { + "epoch": 1.8, + "learning_rate": 2.0082072326237496e-05, + "loss": 0.4004, + "step": 46710 + }, + { + "epoch": 1.8, + "learning_rate": 2.0081431136188768e-05, + "loss": 0.2633, + "step": 46711 + }, + { + "epoch": 1.8, + "learning_rate": 2.0080789946140036e-05, + "loss": 0.2025, + "step": 46712 + }, + { + "epoch": 1.8, + "learning_rate": 2.0080148756091308e-05, + "loss": 0.2133, + "step": 46713 + }, + { + "epoch": 1.8, + "learning_rate": 2.0079507566042576e-05, + "loss": 0.3171, + "step": 46714 + }, + { + "epoch": 1.8, + "learning_rate": 2.0078866375993845e-05, + "loss": 0.3579, + "step": 46715 + }, + { + "epoch": 1.8, + "learning_rate": 2.0078225185945116e-05, + "loss": 0.3982, + "step": 46716 + }, + { + "epoch": 1.8, + "learning_rate": 2.0077583995896385e-05, + "loss": 0.3416, + "step": 46717 + }, + { + "epoch": 1.8, + "learning_rate": 2.0076942805847653e-05, + "loss": 0.388, + "step": 46718 + }, + { + "epoch": 1.8, + "learning_rate": 2.007630161579892e-05, + "loss": 0.2672, + "step": 46719 + }, + { + "epoch": 1.8, + "learning_rate": 2.0075660425750193e-05, + "loss": 0.3542, + "step": 46720 + }, + { + "epoch": 1.8, + "learning_rate": 2.007501923570146e-05, + "loss": 0.3017, + "step": 46721 + }, + { + "epoch": 1.8, + "learning_rate": 2.0074378045652733e-05, + "loss": 0.358, + "step": 46722 + }, + { + "epoch": 1.8, + "learning_rate": 2.0073736855604e-05, + "loss": 0.4415, + "step": 46723 + }, + { + "epoch": 1.8, + "learning_rate": 2.0073095665555273e-05, + "loss": 0.4111, + "step": 46724 + }, + { + "epoch": 1.8, + "learning_rate": 2.007245447550654e-05, + "loss": 0.3432, + "step": 46725 + }, + { + "epoch": 1.8, + "learning_rate": 2.007181328545781e-05, + "loss": 0.3723, + "step": 46726 + }, + { + "epoch": 1.8, + "learning_rate": 2.0071172095409078e-05, + "loss": 0.3613, + "step": 46727 + }, + { + "epoch": 1.8, + "learning_rate": 2.007053090536035e-05, + "loss": 0.3699, + "step": 46728 + }, + { + "epoch": 1.8, + "learning_rate": 2.0069889715311618e-05, + "loss": 0.2847, + "step": 46729 + }, + { + "epoch": 1.8, + "learning_rate": 2.0069248525262887e-05, + "loss": 0.4509, + "step": 46730 + }, + { + "epoch": 1.8, + "learning_rate": 2.006860733521416e-05, + "loss": 0.3745, + "step": 46731 + }, + { + "epoch": 1.8, + "learning_rate": 2.006796614516543e-05, + "loss": 0.4086, + "step": 46732 + }, + { + "epoch": 1.8, + "learning_rate": 2.00673249551167e-05, + "loss": 0.279, + "step": 46733 + }, + { + "epoch": 1.8, + "learning_rate": 2.0066683765067967e-05, + "loss": 0.3379, + "step": 46734 + }, + { + "epoch": 1.8, + "learning_rate": 2.0066042575019235e-05, + "loss": 0.4714, + "step": 46735 + }, + { + "epoch": 1.8, + "learning_rate": 2.0065401384970507e-05, + "loss": 0.3554, + "step": 46736 + }, + { + "epoch": 1.8, + "learning_rate": 2.0064760194921775e-05, + "loss": 0.247, + "step": 46737 + }, + { + "epoch": 1.8, + "learning_rate": 2.0064119004873043e-05, + "loss": 0.4625, + "step": 46738 + }, + { + "epoch": 1.8, + "learning_rate": 2.0063477814824315e-05, + "loss": 0.2641, + "step": 46739 + }, + { + "epoch": 1.8, + "learning_rate": 2.0062836624775587e-05, + "loss": 0.416, + "step": 46740 + }, + { + "epoch": 1.8, + "learning_rate": 2.0062195434726855e-05, + "loss": 0.2699, + "step": 46741 + }, + { + "epoch": 1.8, + "learning_rate": 2.0061554244678124e-05, + "loss": 0.2509, + "step": 46742 + }, + { + "epoch": 1.8, + "learning_rate": 2.0060913054629392e-05, + "loss": 0.4362, + "step": 46743 + }, + { + "epoch": 1.8, + "learning_rate": 2.0060271864580664e-05, + "loss": 0.3283, + "step": 46744 + }, + { + "epoch": 1.8, + "learning_rate": 2.0059630674531932e-05, + "loss": 0.3643, + "step": 46745 + }, + { + "epoch": 1.8, + "learning_rate": 2.00589894844832e-05, + "loss": 0.2785, + "step": 46746 + }, + { + "epoch": 1.8, + "learning_rate": 2.0058348294434472e-05, + "loss": 0.3667, + "step": 46747 + }, + { + "epoch": 1.8, + "learning_rate": 2.005770710438574e-05, + "loss": 0.4254, + "step": 46748 + }, + { + "epoch": 1.8, + "learning_rate": 2.0057065914337012e-05, + "loss": 0.5298, + "step": 46749 + }, + { + "epoch": 1.8, + "learning_rate": 2.005642472428828e-05, + "loss": 0.1844, + "step": 46750 + }, + { + "epoch": 1.8, + "learning_rate": 2.005578353423955e-05, + "loss": 0.3815, + "step": 46751 + }, + { + "epoch": 1.8, + "learning_rate": 2.005514234419082e-05, + "loss": 0.2936, + "step": 46752 + }, + { + "epoch": 1.8, + "learning_rate": 2.005450115414209e-05, + "loss": 0.3553, + "step": 46753 + }, + { + "epoch": 1.8, + "learning_rate": 2.0053859964093357e-05, + "loss": 0.5102, + "step": 46754 + }, + { + "epoch": 1.8, + "learning_rate": 2.0053218774044625e-05, + "loss": 0.2437, + "step": 46755 + }, + { + "epoch": 1.8, + "learning_rate": 2.0052577583995897e-05, + "loss": 0.3574, + "step": 46756 + }, + { + "epoch": 1.8, + "learning_rate": 2.005193639394717e-05, + "loss": 0.3926, + "step": 46757 + }, + { + "epoch": 1.8, + "learning_rate": 2.0051295203898437e-05, + "loss": 0.3767, + "step": 46758 + }, + { + "epoch": 1.8, + "learning_rate": 2.0050654013849705e-05, + "loss": 0.3862, + "step": 46759 + }, + { + "epoch": 1.8, + "learning_rate": 2.0050012823800977e-05, + "loss": 0.4015, + "step": 46760 + }, + { + "epoch": 1.8, + "learning_rate": 2.0049371633752245e-05, + "loss": 0.3547, + "step": 46761 + }, + { + "epoch": 1.8, + "learning_rate": 2.0048730443703514e-05, + "loss": 0.3001, + "step": 46762 + }, + { + "epoch": 1.8, + "learning_rate": 2.0048089253654782e-05, + "loss": 0.2193, + "step": 46763 + }, + { + "epoch": 1.8, + "learning_rate": 2.0047448063606054e-05, + "loss": 0.3419, + "step": 46764 + }, + { + "epoch": 1.8, + "learning_rate": 2.0046806873557322e-05, + "loss": 0.4328, + "step": 46765 + }, + { + "epoch": 1.8, + "learning_rate": 2.0046165683508594e-05, + "loss": 0.3369, + "step": 46766 + }, + { + "epoch": 1.8, + "learning_rate": 2.0045524493459862e-05, + "loss": 0.3251, + "step": 46767 + }, + { + "epoch": 1.8, + "learning_rate": 2.0044883303411134e-05, + "loss": 0.4021, + "step": 46768 + }, + { + "epoch": 1.8, + "learning_rate": 2.0044242113362402e-05, + "loss": 0.285, + "step": 46769 + }, + { + "epoch": 1.8, + "learning_rate": 2.004360092331367e-05, + "loss": 0.259, + "step": 46770 + }, + { + "epoch": 1.8, + "learning_rate": 2.004295973326494e-05, + "loss": 0.3842, + "step": 46771 + }, + { + "epoch": 1.8, + "learning_rate": 2.004231854321621e-05, + "loss": 0.3673, + "step": 46772 + }, + { + "epoch": 1.8, + "learning_rate": 2.004167735316748e-05, + "loss": 0.2594, + "step": 46773 + }, + { + "epoch": 1.8, + "learning_rate": 2.0041036163118747e-05, + "loss": 0.3414, + "step": 46774 + }, + { + "epoch": 1.8, + "learning_rate": 2.004039497307002e-05, + "loss": 0.2951, + "step": 46775 + }, + { + "epoch": 1.8, + "learning_rate": 2.003975378302129e-05, + "loss": 0.3164, + "step": 46776 + }, + { + "epoch": 1.8, + "learning_rate": 2.003911259297256e-05, + "loss": 0.3606, + "step": 46777 + }, + { + "epoch": 1.8, + "learning_rate": 2.0038471402923827e-05, + "loss": 0.3037, + "step": 46778 + }, + { + "epoch": 1.8, + "learning_rate": 2.0037830212875096e-05, + "loss": 0.325, + "step": 46779 + }, + { + "epoch": 1.8, + "learning_rate": 2.0037189022826367e-05, + "loss": 0.4231, + "step": 46780 + }, + { + "epoch": 1.8, + "learning_rate": 2.0036547832777636e-05, + "loss": 0.3986, + "step": 46781 + }, + { + "epoch": 1.8, + "learning_rate": 2.0035906642728904e-05, + "loss": 0.338, + "step": 46782 + }, + { + "epoch": 1.8, + "learning_rate": 2.0035265452680176e-05, + "loss": 0.3963, + "step": 46783 + }, + { + "epoch": 1.8, + "learning_rate": 2.0034624262631448e-05, + "loss": 0.3156, + "step": 46784 + }, + { + "epoch": 1.8, + "learning_rate": 2.0033983072582716e-05, + "loss": 0.2646, + "step": 46785 + }, + { + "epoch": 1.8, + "learning_rate": 2.0033341882533984e-05, + "loss": 0.247, + "step": 46786 + }, + { + "epoch": 1.8, + "learning_rate": 2.0032700692485253e-05, + "loss": 0.2665, + "step": 46787 + }, + { + "epoch": 1.8, + "learning_rate": 2.0032059502436524e-05, + "loss": 0.3683, + "step": 46788 + }, + { + "epoch": 1.8, + "learning_rate": 2.0031418312387793e-05, + "loss": 0.2513, + "step": 46789 + }, + { + "epoch": 1.8, + "learning_rate": 2.003077712233906e-05, + "loss": 0.3808, + "step": 46790 + }, + { + "epoch": 1.8, + "learning_rate": 2.0030135932290333e-05, + "loss": 0.1964, + "step": 46791 + }, + { + "epoch": 1.8, + "learning_rate": 2.00294947422416e-05, + "loss": 0.383, + "step": 46792 + }, + { + "epoch": 1.8, + "learning_rate": 2.0028853552192873e-05, + "loss": 0.4107, + "step": 46793 + }, + { + "epoch": 1.8, + "learning_rate": 2.002821236214414e-05, + "loss": 0.2663, + "step": 46794 + }, + { + "epoch": 1.8, + "learning_rate": 2.002757117209541e-05, + "loss": 0.3079, + "step": 46795 + }, + { + "epoch": 1.8, + "learning_rate": 2.002692998204668e-05, + "loss": 0.3583, + "step": 46796 + }, + { + "epoch": 1.8, + "learning_rate": 2.002628879199795e-05, + "loss": 0.4215, + "step": 46797 + }, + { + "epoch": 1.8, + "learning_rate": 2.0025647601949218e-05, + "loss": 0.251, + "step": 46798 + }, + { + "epoch": 1.8, + "learning_rate": 2.0025006411900486e-05, + "loss": 0.1802, + "step": 46799 + }, + { + "epoch": 1.8, + "learning_rate": 2.0024365221851758e-05, + "loss": 0.3938, + "step": 46800 + }, + { + "epoch": 1.8, + "learning_rate": 2.0023724031803026e-05, + "loss": 0.269, + "step": 46801 + }, + { + "epoch": 1.8, + "learning_rate": 2.0023082841754298e-05, + "loss": 0.3592, + "step": 46802 + }, + { + "epoch": 1.8, + "learning_rate": 2.0022441651705566e-05, + "loss": 0.3645, + "step": 46803 + }, + { + "epoch": 1.8, + "learning_rate": 2.0021800461656838e-05, + "loss": 0.3514, + "step": 46804 + }, + { + "epoch": 1.8, + "learning_rate": 2.0021159271608106e-05, + "loss": 0.4234, + "step": 46805 + }, + { + "epoch": 1.8, + "learning_rate": 2.0020518081559374e-05, + "loss": 0.372, + "step": 46806 + }, + { + "epoch": 1.8, + "learning_rate": 2.0019876891510643e-05, + "loss": 0.4407, + "step": 46807 + }, + { + "epoch": 1.8, + "learning_rate": 2.0019235701461915e-05, + "loss": 0.3815, + "step": 46808 + }, + { + "epoch": 1.8, + "learning_rate": 2.0018594511413183e-05, + "loss": 0.2744, + "step": 46809 + }, + { + "epoch": 1.8, + "learning_rate": 2.001795332136445e-05, + "loss": 0.2829, + "step": 46810 + }, + { + "epoch": 1.8, + "learning_rate": 2.0017312131315723e-05, + "loss": 0.4345, + "step": 46811 + }, + { + "epoch": 1.8, + "learning_rate": 2.0016670941266995e-05, + "loss": 0.3009, + "step": 46812 + }, + { + "epoch": 1.8, + "learning_rate": 2.0016029751218263e-05, + "loss": 0.3535, + "step": 46813 + }, + { + "epoch": 1.8, + "learning_rate": 2.001538856116953e-05, + "loss": 0.3985, + "step": 46814 + }, + { + "epoch": 1.8, + "learning_rate": 2.00147473711208e-05, + "loss": 0.2413, + "step": 46815 + }, + { + "epoch": 1.8, + "learning_rate": 2.001410618107207e-05, + "loss": 0.3517, + "step": 46816 + }, + { + "epoch": 1.8, + "learning_rate": 2.001346499102334e-05, + "loss": 0.3227, + "step": 46817 + }, + { + "epoch": 1.8, + "learning_rate": 2.0012823800974608e-05, + "loss": 0.4968, + "step": 46818 + }, + { + "epoch": 1.8, + "learning_rate": 2.001218261092588e-05, + "loss": 0.2424, + "step": 46819 + }, + { + "epoch": 1.8, + "learning_rate": 2.001154142087715e-05, + "loss": 0.366, + "step": 46820 + }, + { + "epoch": 1.8, + "learning_rate": 2.001090023082842e-05, + "loss": 0.3179, + "step": 46821 + }, + { + "epoch": 1.8, + "learning_rate": 2.0010259040779688e-05, + "loss": 0.3115, + "step": 46822 + }, + { + "epoch": 1.8, + "learning_rate": 2.0009617850730956e-05, + "loss": 0.2666, + "step": 46823 + }, + { + "epoch": 1.8, + "learning_rate": 2.0008976660682228e-05, + "loss": 0.2605, + "step": 46824 + }, + { + "epoch": 1.8, + "learning_rate": 2.0008335470633496e-05, + "loss": 0.4278, + "step": 46825 + }, + { + "epoch": 1.8, + "learning_rate": 2.0007694280584765e-05, + "loss": 0.3433, + "step": 46826 + }, + { + "epoch": 1.8, + "learning_rate": 2.0007053090536036e-05, + "loss": 0.274, + "step": 46827 + }, + { + "epoch": 1.8, + "learning_rate": 2.0006411900487305e-05, + "loss": 0.3969, + "step": 46828 + }, + { + "epoch": 1.8, + "learning_rate": 2.0005770710438577e-05, + "loss": 0.2718, + "step": 46829 + }, + { + "epoch": 1.8, + "learning_rate": 2.0005129520389845e-05, + "loss": 0.3842, + "step": 46830 + }, + { + "epoch": 1.8, + "learning_rate": 2.0004488330341113e-05, + "loss": 0.4187, + "step": 46831 + }, + { + "epoch": 1.8, + "learning_rate": 2.0003847140292385e-05, + "loss": 0.3362, + "step": 46832 + }, + { + "epoch": 1.8, + "learning_rate": 2.0003205950243653e-05, + "loss": 0.2336, + "step": 46833 + }, + { + "epoch": 1.8, + "learning_rate": 2.000256476019492e-05, + "loss": 0.356, + "step": 46834 + }, + { + "epoch": 1.8, + "learning_rate": 2.000192357014619e-05, + "loss": 0.4207, + "step": 46835 + }, + { + "epoch": 1.8, + "learning_rate": 2.000128238009746e-05, + "loss": 0.2843, + "step": 46836 + }, + { + "epoch": 1.8, + "learning_rate": 2.000064119004873e-05, + "loss": 0.3058, + "step": 46837 + }, + { + "epoch": 1.8, + "learning_rate": 2e-05, + "loss": 0.3135, + "step": 46838 + }, + { + "epoch": 1.8, + "learning_rate": 1.999935880995127e-05, + "loss": 0.3196, + "step": 46839 + }, + { + "epoch": 1.8, + "learning_rate": 1.999871761990254e-05, + "loss": 0.2725, + "step": 46840 + }, + { + "epoch": 1.8, + "learning_rate": 1.999807642985381e-05, + "loss": 0.3164, + "step": 46841 + }, + { + "epoch": 1.8, + "learning_rate": 1.999743523980508e-05, + "loss": 0.2797, + "step": 46842 + }, + { + "epoch": 1.8, + "learning_rate": 1.9996794049756347e-05, + "loss": 0.2926, + "step": 46843 + }, + { + "epoch": 1.8, + "learning_rate": 1.999615285970762e-05, + "loss": 0.367, + "step": 46844 + }, + { + "epoch": 1.8, + "learning_rate": 1.9995511669658887e-05, + "loss": 0.3302, + "step": 46845 + }, + { + "epoch": 1.8, + "learning_rate": 1.9994870479610155e-05, + "loss": 0.416, + "step": 46846 + }, + { + "epoch": 1.8, + "learning_rate": 1.9994229289561427e-05, + "loss": 0.3887, + "step": 46847 + }, + { + "epoch": 1.8, + "learning_rate": 1.99935880995127e-05, + "loss": 0.5168, + "step": 46848 + }, + { + "epoch": 1.8, + "learning_rate": 1.9992946909463967e-05, + "loss": 0.434, + "step": 46849 + }, + { + "epoch": 1.8, + "learning_rate": 1.9992305719415235e-05, + "loss": 0.3275, + "step": 46850 + }, + { + "epoch": 1.8, + "learning_rate": 1.9991664529366503e-05, + "loss": 0.2995, + "step": 46851 + }, + { + "epoch": 1.8, + "learning_rate": 1.9991023339317775e-05, + "loss": 0.2694, + "step": 46852 + }, + { + "epoch": 1.8, + "learning_rate": 1.9990382149269044e-05, + "loss": 0.4497, + "step": 46853 + }, + { + "epoch": 1.8, + "learning_rate": 1.9989740959220312e-05, + "loss": 0.188, + "step": 46854 + }, + { + "epoch": 1.8, + "learning_rate": 1.9989099769171584e-05, + "loss": 0.4747, + "step": 46855 + }, + { + "epoch": 1.8, + "learning_rate": 1.9988458579122855e-05, + "loss": 0.3584, + "step": 46856 + }, + { + "epoch": 1.8, + "learning_rate": 1.9987817389074124e-05, + "loss": 0.3567, + "step": 46857 + }, + { + "epoch": 1.8, + "learning_rate": 1.9987176199025392e-05, + "loss": 0.3472, + "step": 46858 + }, + { + "epoch": 1.8, + "learning_rate": 1.998653500897666e-05, + "loss": 0.3624, + "step": 46859 + }, + { + "epoch": 1.8, + "learning_rate": 1.9985893818927932e-05, + "loss": 0.3741, + "step": 46860 + }, + { + "epoch": 1.8, + "learning_rate": 1.99852526288792e-05, + "loss": 0.2844, + "step": 46861 + }, + { + "epoch": 1.8, + "learning_rate": 1.998461143883047e-05, + "loss": 0.4447, + "step": 46862 + }, + { + "epoch": 1.8, + "learning_rate": 1.998397024878174e-05, + "loss": 0.4159, + "step": 46863 + }, + { + "epoch": 1.8, + "learning_rate": 1.9983329058733012e-05, + "loss": 0.3566, + "step": 46864 + }, + { + "epoch": 1.8, + "learning_rate": 1.998268786868428e-05, + "loss": 0.3597, + "step": 46865 + }, + { + "epoch": 1.8, + "learning_rate": 1.998204667863555e-05, + "loss": 0.2587, + "step": 46866 + }, + { + "epoch": 1.8, + "learning_rate": 1.9981405488586817e-05, + "loss": 0.4673, + "step": 46867 + }, + { + "epoch": 1.8, + "learning_rate": 1.998076429853809e-05, + "loss": 0.3892, + "step": 46868 + }, + { + "epoch": 1.8, + "learning_rate": 1.9980123108489357e-05, + "loss": 0.382, + "step": 46869 + }, + { + "epoch": 1.8, + "learning_rate": 1.9979481918440625e-05, + "loss": 0.2935, + "step": 46870 + }, + { + "epoch": 1.8, + "learning_rate": 1.9978840728391897e-05, + "loss": 0.3295, + "step": 46871 + }, + { + "epoch": 1.8, + "learning_rate": 1.9978199538343165e-05, + "loss": 0.449, + "step": 46872 + }, + { + "epoch": 1.8, + "learning_rate": 1.9977558348294437e-05, + "loss": 0.4096, + "step": 46873 + }, + { + "epoch": 1.8, + "learning_rate": 1.9976917158245706e-05, + "loss": 0.2907, + "step": 46874 + }, + { + "epoch": 1.8, + "learning_rate": 1.9976275968196974e-05, + "loss": 0.3269, + "step": 46875 + }, + { + "epoch": 1.8, + "learning_rate": 1.9975634778148246e-05, + "loss": 0.3479, + "step": 46876 + }, + { + "epoch": 1.8, + "learning_rate": 1.9974993588099514e-05, + "loss": 0.2565, + "step": 46877 + }, + { + "epoch": 1.8, + "learning_rate": 1.9974352398050782e-05, + "loss": 0.2943, + "step": 46878 + }, + { + "epoch": 1.8, + "learning_rate": 1.997371120800205e-05, + "loss": 0.3315, + "step": 46879 + }, + { + "epoch": 1.8, + "learning_rate": 1.9973070017953322e-05, + "loss": 0.3334, + "step": 46880 + }, + { + "epoch": 1.8, + "learning_rate": 1.997242882790459e-05, + "loss": 0.3559, + "step": 46881 + }, + { + "epoch": 1.8, + "learning_rate": 1.9971787637855862e-05, + "loss": 0.3671, + "step": 46882 + }, + { + "epoch": 1.8, + "learning_rate": 1.997114644780713e-05, + "loss": 0.2358, + "step": 46883 + }, + { + "epoch": 1.8, + "learning_rate": 1.9970505257758402e-05, + "loss": 0.2957, + "step": 46884 + }, + { + "epoch": 1.8, + "learning_rate": 1.996986406770967e-05, + "loss": 0.3857, + "step": 46885 + }, + { + "epoch": 1.8, + "learning_rate": 1.996922287766094e-05, + "loss": 0.4169, + "step": 46886 + }, + { + "epoch": 1.8, + "learning_rate": 1.9968581687612207e-05, + "loss": 0.3342, + "step": 46887 + }, + { + "epoch": 1.8, + "learning_rate": 1.996794049756348e-05, + "loss": 0.2907, + "step": 46888 + }, + { + "epoch": 1.8, + "learning_rate": 1.9967299307514747e-05, + "loss": 0.31, + "step": 46889 + }, + { + "epoch": 1.8, + "learning_rate": 1.9966658117466016e-05, + "loss": 0.3793, + "step": 46890 + }, + { + "epoch": 1.8, + "learning_rate": 1.9966016927417287e-05, + "loss": 0.2602, + "step": 46891 + }, + { + "epoch": 1.8, + "learning_rate": 1.996537573736856e-05, + "loss": 0.4093, + "step": 46892 + }, + { + "epoch": 1.8, + "learning_rate": 1.9964734547319827e-05, + "loss": 0.3943, + "step": 46893 + }, + { + "epoch": 1.8, + "learning_rate": 1.9964093357271096e-05, + "loss": 0.2049, + "step": 46894 + }, + { + "epoch": 1.8, + "learning_rate": 1.9963452167222364e-05, + "loss": 0.3425, + "step": 46895 + }, + { + "epoch": 1.8, + "learning_rate": 1.9962810977173636e-05, + "loss": 0.4205, + "step": 46896 + }, + { + "epoch": 1.8, + "learning_rate": 1.9962169787124904e-05, + "loss": 0.2702, + "step": 46897 + }, + { + "epoch": 1.8, + "learning_rate": 1.9961528597076172e-05, + "loss": 0.2858, + "step": 46898 + }, + { + "epoch": 1.8, + "learning_rate": 1.9960887407027444e-05, + "loss": 0.3663, + "step": 46899 + }, + { + "epoch": 1.8, + "learning_rate": 1.9960246216978716e-05, + "loss": 0.329, + "step": 46900 + }, + { + "epoch": 1.8, + "learning_rate": 1.9959605026929984e-05, + "loss": 0.4515, + "step": 46901 + }, + { + "epoch": 1.8, + "learning_rate": 1.9958963836881253e-05, + "loss": 0.3072, + "step": 46902 + }, + { + "epoch": 1.8, + "learning_rate": 1.995832264683252e-05, + "loss": 0.3165, + "step": 46903 + }, + { + "epoch": 1.8, + "learning_rate": 1.9957681456783793e-05, + "loss": 0.2918, + "step": 46904 + }, + { + "epoch": 1.8, + "learning_rate": 1.995704026673506e-05, + "loss": 0.3895, + "step": 46905 + }, + { + "epoch": 1.8, + "learning_rate": 1.995639907668633e-05, + "loss": 0.2935, + "step": 46906 + }, + { + "epoch": 1.8, + "learning_rate": 1.99557578866376e-05, + "loss": 0.2535, + "step": 46907 + }, + { + "epoch": 1.8, + "learning_rate": 1.995511669658887e-05, + "loss": 0.3798, + "step": 46908 + }, + { + "epoch": 1.8, + "learning_rate": 1.995447550654014e-05, + "loss": 0.2457, + "step": 46909 + }, + { + "epoch": 1.8, + "learning_rate": 1.995383431649141e-05, + "loss": 0.3505, + "step": 46910 + }, + { + "epoch": 1.8, + "learning_rate": 1.9953193126442678e-05, + "loss": 0.2526, + "step": 46911 + }, + { + "epoch": 1.8, + "learning_rate": 1.995255193639395e-05, + "loss": 0.3372, + "step": 46912 + }, + { + "epoch": 1.8, + "learning_rate": 1.9951910746345218e-05, + "loss": 0.4139, + "step": 46913 + }, + { + "epoch": 1.8, + "learning_rate": 1.9951269556296486e-05, + "loss": 0.3078, + "step": 46914 + }, + { + "epoch": 1.8, + "learning_rate": 1.9950628366247758e-05, + "loss": 0.3679, + "step": 46915 + }, + { + "epoch": 1.8, + "learning_rate": 1.9949987176199026e-05, + "loss": 0.4368, + "step": 46916 + }, + { + "epoch": 1.8, + "learning_rate": 1.9949345986150294e-05, + "loss": 0.311, + "step": 46917 + }, + { + "epoch": 1.8, + "learning_rate": 1.9948704796101566e-05, + "loss": 0.3663, + "step": 46918 + }, + { + "epoch": 1.8, + "learning_rate": 1.9948063606052834e-05, + "loss": 0.3135, + "step": 46919 + }, + { + "epoch": 1.8, + "learning_rate": 1.9947422416004106e-05, + "loss": 0.2857, + "step": 46920 + }, + { + "epoch": 1.8, + "learning_rate": 1.9946781225955375e-05, + "loss": 0.4005, + "step": 46921 + }, + { + "epoch": 1.8, + "learning_rate": 1.9946140035906643e-05, + "loss": 0.4115, + "step": 46922 + }, + { + "epoch": 1.8, + "learning_rate": 1.994549884585791e-05, + "loss": 0.4233, + "step": 46923 + }, + { + "epoch": 1.8, + "learning_rate": 1.9944857655809183e-05, + "loss": 0.4097, + "step": 46924 + }, + { + "epoch": 1.8, + "learning_rate": 1.994421646576045e-05, + "loss": 0.3682, + "step": 46925 + }, + { + "epoch": 1.8, + "learning_rate": 1.994357527571172e-05, + "loss": 0.3385, + "step": 46926 + }, + { + "epoch": 1.8, + "learning_rate": 1.994293408566299e-05, + "loss": 0.3353, + "step": 46927 + }, + { + "epoch": 1.8, + "learning_rate": 1.9942292895614263e-05, + "loss": 0.2794, + "step": 46928 + }, + { + "epoch": 1.8, + "learning_rate": 1.994165170556553e-05, + "loss": 0.3431, + "step": 46929 + }, + { + "epoch": 1.8, + "learning_rate": 1.99410105155168e-05, + "loss": 0.4065, + "step": 46930 + }, + { + "epoch": 1.8, + "learning_rate": 1.9940369325468068e-05, + "loss": 0.2449, + "step": 46931 + }, + { + "epoch": 1.8, + "learning_rate": 1.993972813541934e-05, + "loss": 0.3576, + "step": 46932 + }, + { + "epoch": 1.8, + "learning_rate": 1.9939086945370608e-05, + "loss": 0.2966, + "step": 46933 + }, + { + "epoch": 1.8, + "learning_rate": 1.9938445755321876e-05, + "loss": 0.3699, + "step": 46934 + }, + { + "epoch": 1.8, + "learning_rate": 1.9937804565273148e-05, + "loss": 0.3647, + "step": 46935 + }, + { + "epoch": 1.8, + "learning_rate": 1.993716337522442e-05, + "loss": 0.3693, + "step": 46936 + }, + { + "epoch": 1.8, + "learning_rate": 1.9936522185175688e-05, + "loss": 0.383, + "step": 46937 + }, + { + "epoch": 1.8, + "learning_rate": 1.9935880995126956e-05, + "loss": 0.3826, + "step": 46938 + }, + { + "epoch": 1.8, + "learning_rate": 1.9935239805078225e-05, + "loss": 0.2878, + "step": 46939 + }, + { + "epoch": 1.8, + "learning_rate": 1.9934598615029497e-05, + "loss": 0.4249, + "step": 46940 + }, + { + "epoch": 1.8, + "learning_rate": 1.9933957424980765e-05, + "loss": 0.3663, + "step": 46941 + }, + { + "epoch": 1.8, + "learning_rate": 1.9933316234932033e-05, + "loss": 0.3239, + "step": 46942 + }, + { + "epoch": 1.8, + "learning_rate": 1.9932675044883305e-05, + "loss": 0.2661, + "step": 46943 + }, + { + "epoch": 1.8, + "learning_rate": 1.9932033854834577e-05, + "loss": 0.2619, + "step": 46944 + }, + { + "epoch": 1.8, + "learning_rate": 1.9931392664785845e-05, + "loss": 0.2133, + "step": 46945 + }, + { + "epoch": 1.8, + "learning_rate": 1.9930751474737113e-05, + "loss": 0.355, + "step": 46946 + }, + { + "epoch": 1.8, + "learning_rate": 1.993011028468838e-05, + "loss": 0.2987, + "step": 46947 + }, + { + "epoch": 1.8, + "learning_rate": 1.9929469094639653e-05, + "loss": 0.2567, + "step": 46948 + }, + { + "epoch": 1.81, + "learning_rate": 1.992882790459092e-05, + "loss": 0.4255, + "step": 46949 + }, + { + "epoch": 1.81, + "learning_rate": 1.992818671454219e-05, + "loss": 0.3612, + "step": 46950 + }, + { + "epoch": 1.81, + "learning_rate": 1.992754552449346e-05, + "loss": 0.4714, + "step": 46951 + }, + { + "epoch": 1.81, + "learning_rate": 1.992690433444473e-05, + "loss": 0.4103, + "step": 46952 + }, + { + "epoch": 1.81, + "learning_rate": 1.9926263144396002e-05, + "loss": 0.2945, + "step": 46953 + }, + { + "epoch": 1.81, + "learning_rate": 1.992562195434727e-05, + "loss": 0.236, + "step": 46954 + }, + { + "epoch": 1.81, + "learning_rate": 1.992498076429854e-05, + "loss": 0.3167, + "step": 46955 + }, + { + "epoch": 1.81, + "learning_rate": 1.992433957424981e-05, + "loss": 0.3309, + "step": 46956 + }, + { + "epoch": 1.81, + "learning_rate": 1.992369838420108e-05, + "loss": 0.3742, + "step": 46957 + }, + { + "epoch": 1.81, + "learning_rate": 1.9923057194152347e-05, + "loss": 0.2715, + "step": 46958 + }, + { + "epoch": 1.81, + "learning_rate": 1.9922416004103615e-05, + "loss": 0.3078, + "step": 46959 + }, + { + "epoch": 1.81, + "learning_rate": 1.9921774814054887e-05, + "loss": 0.3343, + "step": 46960 + }, + { + "epoch": 1.81, + "learning_rate": 1.9921133624006155e-05, + "loss": 0.3081, + "step": 46961 + }, + { + "epoch": 1.81, + "learning_rate": 1.9920492433957427e-05, + "loss": 0.3074, + "step": 46962 + }, + { + "epoch": 1.81, + "learning_rate": 1.9919851243908695e-05, + "loss": 0.2366, + "step": 46963 + }, + { + "epoch": 1.81, + "learning_rate": 1.9919210053859967e-05, + "loss": 0.4048, + "step": 46964 + }, + { + "epoch": 1.81, + "learning_rate": 1.9918568863811235e-05, + "loss": 0.4076, + "step": 46965 + }, + { + "epoch": 1.81, + "learning_rate": 1.9917927673762504e-05, + "loss": 0.2536, + "step": 46966 + }, + { + "epoch": 1.81, + "learning_rate": 1.9917286483713772e-05, + "loss": 0.3407, + "step": 46967 + }, + { + "epoch": 1.81, + "learning_rate": 1.9916645293665044e-05, + "loss": 0.393, + "step": 46968 + }, + { + "epoch": 1.81, + "learning_rate": 1.9916004103616312e-05, + "loss": 0.3405, + "step": 46969 + }, + { + "epoch": 1.81, + "learning_rate": 1.991536291356758e-05, + "loss": 0.399, + "step": 46970 + }, + { + "epoch": 1.81, + "learning_rate": 1.9914721723518852e-05, + "loss": 0.3264, + "step": 46971 + }, + { + "epoch": 1.81, + "learning_rate": 1.9914080533470124e-05, + "loss": 0.4732, + "step": 46972 + }, + { + "epoch": 1.81, + "learning_rate": 1.9913439343421392e-05, + "loss": 0.1919, + "step": 46973 + }, + { + "epoch": 1.81, + "learning_rate": 1.991279815337266e-05, + "loss": 0.2484, + "step": 46974 + }, + { + "epoch": 1.81, + "learning_rate": 1.991215696332393e-05, + "loss": 0.4536, + "step": 46975 + }, + { + "epoch": 1.81, + "learning_rate": 1.99115157732752e-05, + "loss": 0.3385, + "step": 46976 + }, + { + "epoch": 1.81, + "learning_rate": 1.991087458322647e-05, + "loss": 0.3209, + "step": 46977 + }, + { + "epoch": 1.81, + "learning_rate": 1.9910233393177737e-05, + "loss": 0.3069, + "step": 46978 + }, + { + "epoch": 1.81, + "learning_rate": 1.990959220312901e-05, + "loss": 0.3701, + "step": 46979 + }, + { + "epoch": 1.81, + "learning_rate": 1.990895101308028e-05, + "loss": 0.2444, + "step": 46980 + }, + { + "epoch": 1.81, + "learning_rate": 1.990830982303155e-05, + "loss": 0.3861, + "step": 46981 + }, + { + "epoch": 1.81, + "learning_rate": 1.9907668632982817e-05, + "loss": 0.4361, + "step": 46982 + }, + { + "epoch": 1.81, + "learning_rate": 1.9907027442934085e-05, + "loss": 0.4213, + "step": 46983 + }, + { + "epoch": 1.81, + "learning_rate": 1.9906386252885357e-05, + "loss": 0.3827, + "step": 46984 + }, + { + "epoch": 1.81, + "learning_rate": 1.9905745062836625e-05, + "loss": 0.3934, + "step": 46985 + }, + { + "epoch": 1.81, + "learning_rate": 1.9905103872787894e-05, + "loss": 0.259, + "step": 46986 + }, + { + "epoch": 1.81, + "learning_rate": 1.9904462682739166e-05, + "loss": 0.4561, + "step": 46987 + }, + { + "epoch": 1.81, + "learning_rate": 1.9903821492690434e-05, + "loss": 0.4018, + "step": 46988 + }, + { + "epoch": 1.81, + "learning_rate": 1.9903180302641706e-05, + "loss": 0.3998, + "step": 46989 + }, + { + "epoch": 1.81, + "learning_rate": 1.9902539112592974e-05, + "loss": 0.2944, + "step": 46990 + }, + { + "epoch": 1.81, + "learning_rate": 1.9901897922544242e-05, + "loss": 0.445, + "step": 46991 + }, + { + "epoch": 1.81, + "learning_rate": 1.9901256732495514e-05, + "loss": 0.3076, + "step": 46992 + }, + { + "epoch": 1.81, + "learning_rate": 1.9900615542446782e-05, + "loss": 0.3312, + "step": 46993 + }, + { + "epoch": 1.81, + "learning_rate": 1.989997435239805e-05, + "loss": 0.2561, + "step": 46994 + }, + { + "epoch": 1.81, + "learning_rate": 1.9899333162349322e-05, + "loss": 0.2786, + "step": 46995 + }, + { + "epoch": 1.81, + "learning_rate": 1.989869197230059e-05, + "loss": 0.3245, + "step": 46996 + }, + { + "epoch": 1.81, + "learning_rate": 1.989805078225186e-05, + "loss": 0.3657, + "step": 46997 + }, + { + "epoch": 1.81, + "learning_rate": 1.989740959220313e-05, + "loss": 0.4903, + "step": 46998 + }, + { + "epoch": 1.81, + "learning_rate": 1.98967684021544e-05, + "loss": 0.4171, + "step": 46999 + }, + { + "epoch": 1.81, + "learning_rate": 1.989612721210567e-05, + "loss": 0.3374, + "step": 47000 + }, + { + "epoch": 1.81, + "learning_rate": 1.989548602205694e-05, + "loss": 0.3233, + "step": 47001 + }, + { + "epoch": 1.81, + "learning_rate": 1.9894844832008207e-05, + "loss": 0.2583, + "step": 47002 + }, + { + "epoch": 1.81, + "learning_rate": 1.9894203641959476e-05, + "loss": 0.4384, + "step": 47003 + }, + { + "epoch": 1.81, + "learning_rate": 1.9893562451910747e-05, + "loss": 0.2743, + "step": 47004 + }, + { + "epoch": 1.81, + "learning_rate": 1.9892921261862016e-05, + "loss": 0.4047, + "step": 47005 + }, + { + "epoch": 1.81, + "learning_rate": 1.9892280071813284e-05, + "loss": 0.3319, + "step": 47006 + }, + { + "epoch": 1.81, + "learning_rate": 1.9891638881764556e-05, + "loss": 0.3158, + "step": 47007 + }, + { + "epoch": 1.81, + "learning_rate": 1.9890997691715828e-05, + "loss": 0.3607, + "step": 47008 + }, + { + "epoch": 1.81, + "learning_rate": 1.9890356501667096e-05, + "loss": 0.4573, + "step": 47009 + }, + { + "epoch": 1.81, + "learning_rate": 1.9889715311618364e-05, + "loss": 0.2842, + "step": 47010 + }, + { + "epoch": 1.81, + "learning_rate": 1.9889074121569633e-05, + "loss": 0.3337, + "step": 47011 + }, + { + "epoch": 1.81, + "learning_rate": 1.9888432931520904e-05, + "loss": 0.2461, + "step": 47012 + }, + { + "epoch": 1.81, + "learning_rate": 1.9887791741472173e-05, + "loss": 0.2488, + "step": 47013 + }, + { + "epoch": 1.81, + "learning_rate": 1.988715055142344e-05, + "loss": 0.3396, + "step": 47014 + }, + { + "epoch": 1.81, + "learning_rate": 1.9886509361374713e-05, + "loss": 0.3009, + "step": 47015 + }, + { + "epoch": 1.81, + "learning_rate": 1.9885868171325984e-05, + "loss": 0.4012, + "step": 47016 + }, + { + "epoch": 1.81, + "learning_rate": 1.9885226981277253e-05, + "loss": 0.3354, + "step": 47017 + }, + { + "epoch": 1.81, + "learning_rate": 1.988458579122852e-05, + "loss": 0.2911, + "step": 47018 + }, + { + "epoch": 1.81, + "learning_rate": 1.988394460117979e-05, + "loss": 0.2606, + "step": 47019 + }, + { + "epoch": 1.81, + "learning_rate": 1.988330341113106e-05, + "loss": 0.3188, + "step": 47020 + }, + { + "epoch": 1.81, + "learning_rate": 1.988266222108233e-05, + "loss": 0.3757, + "step": 47021 + }, + { + "epoch": 1.81, + "learning_rate": 1.9882021031033598e-05, + "loss": 0.47, + "step": 47022 + }, + { + "epoch": 1.81, + "learning_rate": 1.988137984098487e-05, + "loss": 0.3252, + "step": 47023 + }, + { + "epoch": 1.81, + "learning_rate": 1.9880738650936138e-05, + "loss": 0.3013, + "step": 47024 + }, + { + "epoch": 1.81, + "learning_rate": 1.988009746088741e-05, + "loss": 0.3509, + "step": 47025 + }, + { + "epoch": 1.81, + "learning_rate": 1.9879456270838678e-05, + "loss": 0.3247, + "step": 47026 + }, + { + "epoch": 1.81, + "learning_rate": 1.9878815080789946e-05, + "loss": 0.3299, + "step": 47027 + }, + { + "epoch": 1.81, + "learning_rate": 1.9878173890741218e-05, + "loss": 0.3239, + "step": 47028 + }, + { + "epoch": 1.81, + "learning_rate": 1.9877532700692486e-05, + "loss": 0.3458, + "step": 47029 + }, + { + "epoch": 1.81, + "learning_rate": 1.9876891510643754e-05, + "loss": 0.4116, + "step": 47030 + }, + { + "epoch": 1.81, + "learning_rate": 1.9876250320595026e-05, + "loss": 0.3648, + "step": 47031 + }, + { + "epoch": 1.81, + "learning_rate": 1.9875609130546295e-05, + "loss": 0.4465, + "step": 47032 + }, + { + "epoch": 1.81, + "learning_rate": 1.9874967940497563e-05, + "loss": 0.4161, + "step": 47033 + }, + { + "epoch": 1.81, + "learning_rate": 1.9874326750448835e-05, + "loss": 0.3838, + "step": 47034 + }, + { + "epoch": 1.81, + "learning_rate": 1.9873685560400103e-05, + "loss": 0.3017, + "step": 47035 + }, + { + "epoch": 1.81, + "learning_rate": 1.9873044370351375e-05, + "loss": 0.3544, + "step": 47036 + }, + { + "epoch": 1.81, + "learning_rate": 1.9872403180302643e-05, + "loss": 0.2929, + "step": 47037 + }, + { + "epoch": 1.81, + "learning_rate": 1.987176199025391e-05, + "loss": 0.4719, + "step": 47038 + }, + { + "epoch": 1.81, + "learning_rate": 1.987112080020518e-05, + "loss": 0.2993, + "step": 47039 + }, + { + "epoch": 1.81, + "learning_rate": 1.987047961015645e-05, + "loss": 0.3019, + "step": 47040 + }, + { + "epoch": 1.81, + "learning_rate": 1.986983842010772e-05, + "loss": 0.2532, + "step": 47041 + }, + { + "epoch": 1.81, + "learning_rate": 1.986919723005899e-05, + "loss": 0.3003, + "step": 47042 + }, + { + "epoch": 1.81, + "learning_rate": 1.986855604001026e-05, + "loss": 0.5009, + "step": 47043 + }, + { + "epoch": 1.81, + "learning_rate": 1.986791484996153e-05, + "loss": 0.4157, + "step": 47044 + }, + { + "epoch": 1.81, + "learning_rate": 1.98672736599128e-05, + "loss": 0.3227, + "step": 47045 + }, + { + "epoch": 1.81, + "learning_rate": 1.9866632469864068e-05, + "loss": 0.3542, + "step": 47046 + }, + { + "epoch": 1.81, + "learning_rate": 1.9865991279815336e-05, + "loss": 0.2884, + "step": 47047 + }, + { + "epoch": 1.81, + "learning_rate": 1.9865350089766608e-05, + "loss": 0.2672, + "step": 47048 + }, + { + "epoch": 1.81, + "learning_rate": 1.9864708899717876e-05, + "loss": 0.2751, + "step": 47049 + }, + { + "epoch": 1.81, + "learning_rate": 1.9864067709669145e-05, + "loss": 0.3572, + "step": 47050 + }, + { + "epoch": 1.81, + "learning_rate": 1.9863426519620416e-05, + "loss": 0.3636, + "step": 47051 + }, + { + "epoch": 1.81, + "learning_rate": 1.9862785329571688e-05, + "loss": 0.3087, + "step": 47052 + }, + { + "epoch": 1.81, + "learning_rate": 1.9862144139522957e-05, + "loss": 0.3761, + "step": 47053 + }, + { + "epoch": 1.81, + "learning_rate": 1.9861502949474225e-05, + "loss": 0.3123, + "step": 47054 + }, + { + "epoch": 1.81, + "learning_rate": 1.9860861759425493e-05, + "loss": 0.267, + "step": 47055 + }, + { + "epoch": 1.81, + "learning_rate": 1.9860220569376765e-05, + "loss": 0.2864, + "step": 47056 + }, + { + "epoch": 1.81, + "learning_rate": 1.9859579379328033e-05, + "loss": 0.3718, + "step": 47057 + }, + { + "epoch": 1.81, + "learning_rate": 1.98589381892793e-05, + "loss": 0.3589, + "step": 47058 + }, + { + "epoch": 1.81, + "learning_rate": 1.9858296999230573e-05, + "loss": 0.2683, + "step": 47059 + }, + { + "epoch": 1.81, + "learning_rate": 1.9857655809181845e-05, + "loss": 0.1866, + "step": 47060 + }, + { + "epoch": 1.81, + "learning_rate": 1.9857014619133113e-05, + "loss": 0.43, + "step": 47061 + }, + { + "epoch": 1.81, + "learning_rate": 1.985637342908438e-05, + "loss": 0.2845, + "step": 47062 + }, + { + "epoch": 1.81, + "learning_rate": 1.985573223903565e-05, + "loss": 0.3674, + "step": 47063 + }, + { + "epoch": 1.81, + "learning_rate": 1.9855091048986922e-05, + "loss": 0.2391, + "step": 47064 + }, + { + "epoch": 1.81, + "learning_rate": 1.985444985893819e-05, + "loss": 0.2163, + "step": 47065 + }, + { + "epoch": 1.81, + "learning_rate": 1.985380866888946e-05, + "loss": 0.5802, + "step": 47066 + }, + { + "epoch": 1.81, + "learning_rate": 1.985316747884073e-05, + "loss": 0.3795, + "step": 47067 + }, + { + "epoch": 1.81, + "learning_rate": 1.9852526288792e-05, + "loss": 0.3988, + "step": 47068 + }, + { + "epoch": 1.81, + "learning_rate": 1.985188509874327e-05, + "loss": 0.3886, + "step": 47069 + }, + { + "epoch": 1.81, + "learning_rate": 1.985124390869454e-05, + "loss": 0.3509, + "step": 47070 + }, + { + "epoch": 1.81, + "learning_rate": 1.9850602718645807e-05, + "loss": 0.3625, + "step": 47071 + }, + { + "epoch": 1.81, + "learning_rate": 1.984996152859708e-05, + "loss": 0.4533, + "step": 47072 + }, + { + "epoch": 1.81, + "learning_rate": 1.9849320338548347e-05, + "loss": 0.4297, + "step": 47073 + }, + { + "epoch": 1.81, + "learning_rate": 1.9848679148499615e-05, + "loss": 0.2954, + "step": 47074 + }, + { + "epoch": 1.81, + "learning_rate": 1.9848037958450887e-05, + "loss": 0.3312, + "step": 47075 + }, + { + "epoch": 1.81, + "learning_rate": 1.9847396768402155e-05, + "loss": 0.2503, + "step": 47076 + }, + { + "epoch": 1.81, + "learning_rate": 1.9846755578353424e-05, + "loss": 0.2628, + "step": 47077 + }, + { + "epoch": 1.81, + "learning_rate": 1.9846114388304695e-05, + "loss": 0.4348, + "step": 47078 + }, + { + "epoch": 1.81, + "learning_rate": 1.9845473198255964e-05, + "loss": 0.351, + "step": 47079 + }, + { + "epoch": 1.81, + "learning_rate": 1.9844832008207235e-05, + "loss": 0.2879, + "step": 47080 + }, + { + "epoch": 1.81, + "learning_rate": 1.9844190818158504e-05, + "loss": 0.4041, + "step": 47081 + }, + { + "epoch": 1.81, + "learning_rate": 1.9843549628109772e-05, + "loss": 0.3236, + "step": 47082 + }, + { + "epoch": 1.81, + "learning_rate": 1.984290843806104e-05, + "loss": 0.3778, + "step": 47083 + }, + { + "epoch": 1.81, + "learning_rate": 1.9842267248012312e-05, + "loss": 0.2757, + "step": 47084 + }, + { + "epoch": 1.81, + "learning_rate": 1.984162605796358e-05, + "loss": 0.3341, + "step": 47085 + }, + { + "epoch": 1.81, + "learning_rate": 1.984098486791485e-05, + "loss": 0.3282, + "step": 47086 + }, + { + "epoch": 1.81, + "learning_rate": 1.984034367786612e-05, + "loss": 0.2623, + "step": 47087 + }, + { + "epoch": 1.81, + "learning_rate": 1.9839702487817392e-05, + "loss": 0.3361, + "step": 47088 + }, + { + "epoch": 1.81, + "learning_rate": 1.983906129776866e-05, + "loss": 0.4555, + "step": 47089 + }, + { + "epoch": 1.81, + "learning_rate": 1.983842010771993e-05, + "loss": 0.5142, + "step": 47090 + }, + { + "epoch": 1.81, + "learning_rate": 1.9837778917671197e-05, + "loss": 0.3756, + "step": 47091 + }, + { + "epoch": 1.81, + "learning_rate": 1.983713772762247e-05, + "loss": 0.3348, + "step": 47092 + }, + { + "epoch": 1.81, + "learning_rate": 1.9836496537573737e-05, + "loss": 0.3236, + "step": 47093 + }, + { + "epoch": 1.81, + "learning_rate": 1.9835855347525005e-05, + "loss": 0.2988, + "step": 47094 + }, + { + "epoch": 1.81, + "learning_rate": 1.9835214157476277e-05, + "loss": 0.3501, + "step": 47095 + }, + { + "epoch": 1.81, + "learning_rate": 1.983457296742755e-05, + "loss": 0.3523, + "step": 47096 + }, + { + "epoch": 1.81, + "learning_rate": 1.9833931777378817e-05, + "loss": 0.3217, + "step": 47097 + }, + { + "epoch": 1.81, + "learning_rate": 1.9833290587330086e-05, + "loss": 0.3355, + "step": 47098 + }, + { + "epoch": 1.81, + "learning_rate": 1.9832649397281354e-05, + "loss": 0.2796, + "step": 47099 + }, + { + "epoch": 1.81, + "learning_rate": 1.9832008207232626e-05, + "loss": 0.2112, + "step": 47100 + }, + { + "epoch": 1.81, + "learning_rate": 1.9831367017183894e-05, + "loss": 0.26, + "step": 47101 + }, + { + "epoch": 1.81, + "learning_rate": 1.9830725827135162e-05, + "loss": 0.2958, + "step": 47102 + }, + { + "epoch": 1.81, + "learning_rate": 1.9830084637086434e-05, + "loss": 0.2646, + "step": 47103 + }, + { + "epoch": 1.81, + "learning_rate": 1.9829443447037702e-05, + "loss": 0.2494, + "step": 47104 + }, + { + "epoch": 1.81, + "learning_rate": 1.9828802256988974e-05, + "loss": 0.3514, + "step": 47105 + }, + { + "epoch": 1.81, + "learning_rate": 1.9828161066940242e-05, + "loss": 0.2798, + "step": 47106 + }, + { + "epoch": 1.81, + "learning_rate": 1.982751987689151e-05, + "loss": 0.3452, + "step": 47107 + }, + { + "epoch": 1.81, + "learning_rate": 1.9826878686842782e-05, + "loss": 0.3302, + "step": 47108 + }, + { + "epoch": 1.81, + "learning_rate": 1.982623749679405e-05, + "loss": 0.2949, + "step": 47109 + }, + { + "epoch": 1.81, + "learning_rate": 1.982559630674532e-05, + "loss": 0.6527, + "step": 47110 + }, + { + "epoch": 1.81, + "learning_rate": 1.982495511669659e-05, + "loss": 0.2722, + "step": 47111 + }, + { + "epoch": 1.81, + "learning_rate": 1.982431392664786e-05, + "loss": 0.2315, + "step": 47112 + }, + { + "epoch": 1.81, + "learning_rate": 1.9823672736599127e-05, + "loss": 0.1916, + "step": 47113 + }, + { + "epoch": 1.81, + "learning_rate": 1.98230315465504e-05, + "loss": 0.3495, + "step": 47114 + }, + { + "epoch": 1.81, + "learning_rate": 1.9822390356501667e-05, + "loss": 0.3338, + "step": 47115 + }, + { + "epoch": 1.81, + "learning_rate": 1.982174916645294e-05, + "loss": 0.3415, + "step": 47116 + }, + { + "epoch": 1.81, + "learning_rate": 1.9821107976404207e-05, + "loss": 0.3758, + "step": 47117 + }, + { + "epoch": 1.81, + "learning_rate": 1.9820466786355476e-05, + "loss": 0.354, + "step": 47118 + }, + { + "epoch": 1.81, + "learning_rate": 1.9819825596306748e-05, + "loss": 0.2566, + "step": 47119 + }, + { + "epoch": 1.81, + "learning_rate": 1.9819184406258016e-05, + "loss": 0.3262, + "step": 47120 + }, + { + "epoch": 1.81, + "learning_rate": 1.9818543216209284e-05, + "loss": 0.2119, + "step": 47121 + }, + { + "epoch": 1.81, + "learning_rate": 1.9817902026160553e-05, + "loss": 0.3071, + "step": 47122 + }, + { + "epoch": 1.81, + "learning_rate": 1.9817260836111824e-05, + "loss": 0.2801, + "step": 47123 + }, + { + "epoch": 1.81, + "learning_rate": 1.9816619646063096e-05, + "loss": 0.3472, + "step": 47124 + }, + { + "epoch": 1.81, + "learning_rate": 1.9815978456014364e-05, + "loss": 0.3622, + "step": 47125 + }, + { + "epoch": 1.81, + "learning_rate": 1.9815337265965633e-05, + "loss": 0.2673, + "step": 47126 + }, + { + "epoch": 1.81, + "learning_rate": 1.98146960759169e-05, + "loss": 0.2731, + "step": 47127 + }, + { + "epoch": 1.81, + "learning_rate": 1.9814054885868173e-05, + "loss": 0.3425, + "step": 47128 + }, + { + "epoch": 1.81, + "learning_rate": 1.981341369581944e-05, + "loss": 0.3522, + "step": 47129 + }, + { + "epoch": 1.81, + "learning_rate": 1.981277250577071e-05, + "loss": 0.4517, + "step": 47130 + }, + { + "epoch": 1.81, + "learning_rate": 1.981213131572198e-05, + "loss": 0.3597, + "step": 47131 + }, + { + "epoch": 1.81, + "learning_rate": 1.9811490125673253e-05, + "loss": 0.3277, + "step": 47132 + }, + { + "epoch": 1.81, + "learning_rate": 1.981084893562452e-05, + "loss": 0.2334, + "step": 47133 + }, + { + "epoch": 1.81, + "learning_rate": 1.981020774557579e-05, + "loss": 0.2742, + "step": 47134 + }, + { + "epoch": 1.81, + "learning_rate": 1.9809566555527058e-05, + "loss": 0.4276, + "step": 47135 + }, + { + "epoch": 1.81, + "learning_rate": 1.980892536547833e-05, + "loss": 0.2814, + "step": 47136 + }, + { + "epoch": 1.81, + "learning_rate": 1.9808284175429598e-05, + "loss": 0.3572, + "step": 47137 + }, + { + "epoch": 1.81, + "learning_rate": 1.9807642985380866e-05, + "loss": 0.3382, + "step": 47138 + }, + { + "epoch": 1.81, + "learning_rate": 1.9807001795332138e-05, + "loss": 0.2975, + "step": 47139 + }, + { + "epoch": 1.81, + "learning_rate": 1.980636060528341e-05, + "loss": 0.4578, + "step": 47140 + }, + { + "epoch": 1.81, + "learning_rate": 1.9805719415234678e-05, + "loss": 0.4172, + "step": 47141 + }, + { + "epoch": 1.81, + "learning_rate": 1.9805078225185946e-05, + "loss": 0.2685, + "step": 47142 + }, + { + "epoch": 1.81, + "learning_rate": 1.9804437035137215e-05, + "loss": 0.3697, + "step": 47143 + }, + { + "epoch": 1.81, + "learning_rate": 1.9803795845088486e-05, + "loss": 0.3271, + "step": 47144 + }, + { + "epoch": 1.81, + "learning_rate": 1.9803154655039755e-05, + "loss": 0.4027, + "step": 47145 + }, + { + "epoch": 1.81, + "learning_rate": 1.9802513464991023e-05, + "loss": 0.3066, + "step": 47146 + }, + { + "epoch": 1.81, + "learning_rate": 1.9801872274942295e-05, + "loss": 0.284, + "step": 47147 + }, + { + "epoch": 1.81, + "learning_rate": 1.9801231084893563e-05, + "loss": 0.4596, + "step": 47148 + }, + { + "epoch": 1.81, + "learning_rate": 1.9800589894844835e-05, + "loss": 0.3038, + "step": 47149 + }, + { + "epoch": 1.81, + "learning_rate": 1.9799948704796103e-05, + "loss": 0.2687, + "step": 47150 + }, + { + "epoch": 1.81, + "learning_rate": 1.979930751474737e-05, + "loss": 0.3019, + "step": 47151 + }, + { + "epoch": 1.81, + "learning_rate": 1.9798666324698643e-05, + "loss": 0.3444, + "step": 47152 + }, + { + "epoch": 1.81, + "learning_rate": 1.979802513464991e-05, + "loss": 0.3314, + "step": 47153 + }, + { + "epoch": 1.81, + "learning_rate": 1.979738394460118e-05, + "loss": 0.3056, + "step": 47154 + }, + { + "epoch": 1.81, + "learning_rate": 1.979674275455245e-05, + "loss": 0.2809, + "step": 47155 + }, + { + "epoch": 1.81, + "learning_rate": 1.979610156450372e-05, + "loss": 0.2825, + "step": 47156 + }, + { + "epoch": 1.81, + "learning_rate": 1.9795460374454988e-05, + "loss": 0.3065, + "step": 47157 + }, + { + "epoch": 1.81, + "learning_rate": 1.979481918440626e-05, + "loss": 0.2377, + "step": 47158 + }, + { + "epoch": 1.81, + "learning_rate": 1.9794177994357528e-05, + "loss": 0.2806, + "step": 47159 + }, + { + "epoch": 1.81, + "learning_rate": 1.97935368043088e-05, + "loss": 0.3972, + "step": 47160 + }, + { + "epoch": 1.81, + "learning_rate": 1.9792895614260068e-05, + "loss": 0.3308, + "step": 47161 + }, + { + "epoch": 1.81, + "learning_rate": 1.9792254424211336e-05, + "loss": 0.3687, + "step": 47162 + }, + { + "epoch": 1.81, + "learning_rate": 1.9791613234162605e-05, + "loss": 0.3469, + "step": 47163 + }, + { + "epoch": 1.81, + "learning_rate": 1.9790972044113877e-05, + "loss": 0.2839, + "step": 47164 + }, + { + "epoch": 1.81, + "learning_rate": 1.9790330854065145e-05, + "loss": 0.3998, + "step": 47165 + }, + { + "epoch": 1.81, + "learning_rate": 1.9789689664016413e-05, + "loss": 0.312, + "step": 47166 + }, + { + "epoch": 1.81, + "learning_rate": 1.9789048473967685e-05, + "loss": 0.5449, + "step": 47167 + }, + { + "epoch": 1.81, + "learning_rate": 1.9788407283918957e-05, + "loss": 0.4184, + "step": 47168 + }, + { + "epoch": 1.81, + "learning_rate": 1.9787766093870225e-05, + "loss": 0.3417, + "step": 47169 + }, + { + "epoch": 1.81, + "learning_rate": 1.9787124903821493e-05, + "loss": 0.3157, + "step": 47170 + }, + { + "epoch": 1.81, + "learning_rate": 1.978648371377276e-05, + "loss": 0.3181, + "step": 47171 + }, + { + "epoch": 1.81, + "learning_rate": 1.9785842523724033e-05, + "loss": 0.2033, + "step": 47172 + }, + { + "epoch": 1.81, + "learning_rate": 1.97852013336753e-05, + "loss": 0.3559, + "step": 47173 + }, + { + "epoch": 1.81, + "learning_rate": 1.978456014362657e-05, + "loss": 0.4485, + "step": 47174 + }, + { + "epoch": 1.81, + "learning_rate": 1.978391895357784e-05, + "loss": 0.4215, + "step": 47175 + }, + { + "epoch": 1.81, + "learning_rate": 1.9783277763529113e-05, + "loss": 0.3049, + "step": 47176 + }, + { + "epoch": 1.81, + "learning_rate": 1.9782636573480382e-05, + "loss": 0.2444, + "step": 47177 + }, + { + "epoch": 1.81, + "learning_rate": 1.978199538343165e-05, + "loss": 0.333, + "step": 47178 + }, + { + "epoch": 1.81, + "learning_rate": 1.978135419338292e-05, + "loss": 0.3883, + "step": 47179 + }, + { + "epoch": 1.81, + "learning_rate": 1.978071300333419e-05, + "loss": 0.4036, + "step": 47180 + }, + { + "epoch": 1.81, + "learning_rate": 1.978007181328546e-05, + "loss": 0.3786, + "step": 47181 + }, + { + "epoch": 1.81, + "learning_rate": 1.9779430623236727e-05, + "loss": 0.3078, + "step": 47182 + }, + { + "epoch": 1.81, + "learning_rate": 1.9778789433188e-05, + "loss": 0.2565, + "step": 47183 + }, + { + "epoch": 1.81, + "learning_rate": 1.9778148243139267e-05, + "loss": 0.3641, + "step": 47184 + }, + { + "epoch": 1.81, + "learning_rate": 1.977750705309054e-05, + "loss": 0.3737, + "step": 47185 + }, + { + "epoch": 1.81, + "learning_rate": 1.9776865863041807e-05, + "loss": 0.3171, + "step": 47186 + }, + { + "epoch": 1.81, + "learning_rate": 1.9776224672993075e-05, + "loss": 0.2982, + "step": 47187 + }, + { + "epoch": 1.81, + "learning_rate": 1.9775583482944347e-05, + "loss": 0.3729, + "step": 47188 + }, + { + "epoch": 1.81, + "learning_rate": 1.9774942292895615e-05, + "loss": 0.2954, + "step": 47189 + }, + { + "epoch": 1.81, + "learning_rate": 1.9774301102846884e-05, + "loss": 0.3326, + "step": 47190 + }, + { + "epoch": 1.81, + "learning_rate": 1.9773659912798155e-05, + "loss": 0.2551, + "step": 47191 + }, + { + "epoch": 1.81, + "learning_rate": 1.9773018722749424e-05, + "loss": 0.2717, + "step": 47192 + }, + { + "epoch": 1.81, + "learning_rate": 1.9772377532700692e-05, + "loss": 0.4221, + "step": 47193 + }, + { + "epoch": 1.81, + "learning_rate": 1.9771736342651964e-05, + "loss": 0.3081, + "step": 47194 + }, + { + "epoch": 1.81, + "learning_rate": 1.9771095152603232e-05, + "loss": 0.4146, + "step": 47195 + }, + { + "epoch": 1.81, + "learning_rate": 1.9770453962554504e-05, + "loss": 0.3848, + "step": 47196 + }, + { + "epoch": 1.81, + "learning_rate": 1.9769812772505772e-05, + "loss": 0.3357, + "step": 47197 + }, + { + "epoch": 1.81, + "learning_rate": 1.976917158245704e-05, + "loss": 0.2826, + "step": 47198 + }, + { + "epoch": 1.81, + "learning_rate": 1.9768530392408312e-05, + "loss": 0.2959, + "step": 47199 + }, + { + "epoch": 1.81, + "learning_rate": 1.976788920235958e-05, + "loss": 0.3126, + "step": 47200 + }, + { + "epoch": 1.81, + "learning_rate": 1.976724801231085e-05, + "loss": 0.3417, + "step": 47201 + }, + { + "epoch": 1.81, + "learning_rate": 1.9766606822262117e-05, + "loss": 0.3908, + "step": 47202 + }, + { + "epoch": 1.81, + "learning_rate": 1.976596563221339e-05, + "loss": 0.3896, + "step": 47203 + }, + { + "epoch": 1.81, + "learning_rate": 1.976532444216466e-05, + "loss": 0.3459, + "step": 47204 + }, + { + "epoch": 1.81, + "learning_rate": 1.976468325211593e-05, + "loss": 0.2683, + "step": 47205 + }, + { + "epoch": 1.81, + "learning_rate": 1.9764042062067197e-05, + "loss": 0.4301, + "step": 47206 + }, + { + "epoch": 1.81, + "learning_rate": 1.9763400872018465e-05, + "loss": 0.3654, + "step": 47207 + }, + { + "epoch": 1.81, + "learning_rate": 1.9762759681969737e-05, + "loss": 0.2263, + "step": 47208 + }, + { + "epoch": 1.82, + "learning_rate": 1.9762118491921006e-05, + "loss": 0.304, + "step": 47209 + }, + { + "epoch": 1.82, + "learning_rate": 1.9761477301872274e-05, + "loss": 0.279, + "step": 47210 + }, + { + "epoch": 1.82, + "learning_rate": 1.9760836111823546e-05, + "loss": 0.5353, + "step": 47211 + }, + { + "epoch": 1.82, + "learning_rate": 1.9760194921774817e-05, + "loss": 0.4639, + "step": 47212 + }, + { + "epoch": 1.82, + "learning_rate": 1.9759553731726086e-05, + "loss": 0.3621, + "step": 47213 + }, + { + "epoch": 1.82, + "learning_rate": 1.9758912541677354e-05, + "loss": 0.3343, + "step": 47214 + }, + { + "epoch": 1.82, + "learning_rate": 1.9758271351628622e-05, + "loss": 0.3911, + "step": 47215 + }, + { + "epoch": 1.82, + "learning_rate": 1.9757630161579894e-05, + "loss": 0.3598, + "step": 47216 + }, + { + "epoch": 1.82, + "learning_rate": 1.9756988971531162e-05, + "loss": 0.3364, + "step": 47217 + }, + { + "epoch": 1.82, + "learning_rate": 1.975634778148243e-05, + "loss": 0.2625, + "step": 47218 + }, + { + "epoch": 1.82, + "learning_rate": 1.9755706591433702e-05, + "loss": 0.5034, + "step": 47219 + }, + { + "epoch": 1.82, + "learning_rate": 1.975506540138497e-05, + "loss": 0.369, + "step": 47220 + }, + { + "epoch": 1.82, + "learning_rate": 1.9754424211336242e-05, + "loss": 0.1783, + "step": 47221 + }, + { + "epoch": 1.82, + "learning_rate": 1.975378302128751e-05, + "loss": 0.3538, + "step": 47222 + }, + { + "epoch": 1.82, + "learning_rate": 1.975314183123878e-05, + "loss": 0.306, + "step": 47223 + }, + { + "epoch": 1.82, + "learning_rate": 1.975250064119005e-05, + "loss": 0.393, + "step": 47224 + }, + { + "epoch": 1.82, + "learning_rate": 1.975185945114132e-05, + "loss": 0.3126, + "step": 47225 + }, + { + "epoch": 1.82, + "learning_rate": 1.9751218261092587e-05, + "loss": 0.2908, + "step": 47226 + }, + { + "epoch": 1.82, + "learning_rate": 1.975057707104386e-05, + "loss": 0.2365, + "step": 47227 + }, + { + "epoch": 1.82, + "learning_rate": 1.9749935880995127e-05, + "loss": 0.4196, + "step": 47228 + }, + { + "epoch": 1.82, + "learning_rate": 1.97492946909464e-05, + "loss": 0.2711, + "step": 47229 + }, + { + "epoch": 1.82, + "learning_rate": 1.9748653500897668e-05, + "loss": 0.3486, + "step": 47230 + }, + { + "epoch": 1.82, + "learning_rate": 1.9748012310848936e-05, + "loss": 0.3103, + "step": 47231 + }, + { + "epoch": 1.82, + "learning_rate": 1.9747371120800208e-05, + "loss": 0.3596, + "step": 47232 + }, + { + "epoch": 1.82, + "learning_rate": 1.9746729930751476e-05, + "loss": 0.3636, + "step": 47233 + }, + { + "epoch": 1.82, + "learning_rate": 1.9746088740702744e-05, + "loss": 0.2057, + "step": 47234 + }, + { + "epoch": 1.82, + "learning_rate": 1.9745447550654016e-05, + "loss": 0.2994, + "step": 47235 + }, + { + "epoch": 1.82, + "learning_rate": 1.9744806360605284e-05, + "loss": 0.313, + "step": 47236 + }, + { + "epoch": 1.82, + "learning_rate": 1.9744165170556553e-05, + "loss": 0.2678, + "step": 47237 + }, + { + "epoch": 1.82, + "learning_rate": 1.9743523980507824e-05, + "loss": 0.4904, + "step": 47238 + }, + { + "epoch": 1.82, + "learning_rate": 1.9742882790459093e-05, + "loss": 0.3419, + "step": 47239 + }, + { + "epoch": 1.82, + "learning_rate": 1.9742241600410364e-05, + "loss": 0.3827, + "step": 47240 + }, + { + "epoch": 1.82, + "learning_rate": 1.9741600410361633e-05, + "loss": 0.3259, + "step": 47241 + }, + { + "epoch": 1.82, + "learning_rate": 1.97409592203129e-05, + "loss": 0.437, + "step": 47242 + }, + { + "epoch": 1.82, + "learning_rate": 1.9740318030264173e-05, + "loss": 0.4092, + "step": 47243 + }, + { + "epoch": 1.82, + "learning_rate": 1.973967684021544e-05, + "loss": 0.3127, + "step": 47244 + }, + { + "epoch": 1.82, + "learning_rate": 1.973903565016671e-05, + "loss": 0.3454, + "step": 47245 + }, + { + "epoch": 1.82, + "learning_rate": 1.9738394460117978e-05, + "loss": 0.2661, + "step": 47246 + }, + { + "epoch": 1.82, + "learning_rate": 1.973775327006925e-05, + "loss": 0.2659, + "step": 47247 + }, + { + "epoch": 1.82, + "learning_rate": 1.973711208002052e-05, + "loss": 0.2827, + "step": 47248 + }, + { + "epoch": 1.82, + "learning_rate": 1.973647088997179e-05, + "loss": 0.3734, + "step": 47249 + }, + { + "epoch": 1.82, + "learning_rate": 1.9735829699923058e-05, + "loss": 0.3352, + "step": 47250 + }, + { + "epoch": 1.82, + "learning_rate": 1.9735188509874326e-05, + "loss": 0.3855, + "step": 47251 + }, + { + "epoch": 1.82, + "learning_rate": 1.9734547319825598e-05, + "loss": 0.2805, + "step": 47252 + }, + { + "epoch": 1.82, + "learning_rate": 1.9733906129776866e-05, + "loss": 0.2834, + "step": 47253 + }, + { + "epoch": 1.82, + "learning_rate": 1.9733264939728135e-05, + "loss": 0.3525, + "step": 47254 + }, + { + "epoch": 1.82, + "learning_rate": 1.9732623749679406e-05, + "loss": 0.4367, + "step": 47255 + }, + { + "epoch": 1.82, + "learning_rate": 1.9731982559630678e-05, + "loss": 0.3433, + "step": 47256 + }, + { + "epoch": 1.82, + "learning_rate": 1.9731341369581946e-05, + "loss": 0.3425, + "step": 47257 + }, + { + "epoch": 1.82, + "learning_rate": 1.9730700179533215e-05, + "loss": 0.2849, + "step": 47258 + }, + { + "epoch": 1.82, + "learning_rate": 1.9730058989484483e-05, + "loss": 0.326, + "step": 47259 + }, + { + "epoch": 1.82, + "learning_rate": 1.9729417799435755e-05, + "loss": 0.2436, + "step": 47260 + }, + { + "epoch": 1.82, + "learning_rate": 1.9728776609387023e-05, + "loss": 0.4188, + "step": 47261 + }, + { + "epoch": 1.82, + "learning_rate": 1.972813541933829e-05, + "loss": 0.4494, + "step": 47262 + }, + { + "epoch": 1.82, + "learning_rate": 1.9727494229289563e-05, + "loss": 0.3187, + "step": 47263 + }, + { + "epoch": 1.82, + "learning_rate": 1.972685303924083e-05, + "loss": 0.3799, + "step": 47264 + }, + { + "epoch": 1.82, + "learning_rate": 1.9726211849192103e-05, + "loss": 0.3921, + "step": 47265 + }, + { + "epoch": 1.82, + "learning_rate": 1.972557065914337e-05, + "loss": 0.3906, + "step": 47266 + }, + { + "epoch": 1.82, + "learning_rate": 1.972492946909464e-05, + "loss": 0.3448, + "step": 47267 + }, + { + "epoch": 1.82, + "learning_rate": 1.972428827904591e-05, + "loss": 0.4668, + "step": 47268 + }, + { + "epoch": 1.82, + "learning_rate": 1.972364708899718e-05, + "loss": 0.3853, + "step": 47269 + }, + { + "epoch": 1.82, + "learning_rate": 1.9723005898948448e-05, + "loss": 0.3836, + "step": 47270 + }, + { + "epoch": 1.82, + "learning_rate": 1.972236470889972e-05, + "loss": 0.4117, + "step": 47271 + }, + { + "epoch": 1.82, + "learning_rate": 1.9721723518850988e-05, + "loss": 0.2817, + "step": 47272 + }, + { + "epoch": 1.82, + "learning_rate": 1.9721082328802256e-05, + "loss": 0.2876, + "step": 47273 + }, + { + "epoch": 1.82, + "learning_rate": 1.9720441138753528e-05, + "loss": 0.2807, + "step": 47274 + }, + { + "epoch": 1.82, + "learning_rate": 1.9719799948704797e-05, + "loss": 0.3042, + "step": 47275 + }, + { + "epoch": 1.82, + "learning_rate": 1.9719158758656068e-05, + "loss": 0.3284, + "step": 47276 + }, + { + "epoch": 1.82, + "learning_rate": 1.9718517568607337e-05, + "loss": 0.4335, + "step": 47277 + }, + { + "epoch": 1.82, + "learning_rate": 1.9717876378558605e-05, + "loss": 0.4261, + "step": 47278 + }, + { + "epoch": 1.82, + "learning_rate": 1.9717235188509877e-05, + "loss": 0.3383, + "step": 47279 + }, + { + "epoch": 1.82, + "learning_rate": 1.9716593998461145e-05, + "loss": 0.3675, + "step": 47280 + }, + { + "epoch": 1.82, + "learning_rate": 1.9715952808412413e-05, + "loss": 0.343, + "step": 47281 + }, + { + "epoch": 1.82, + "learning_rate": 1.971531161836368e-05, + "loss": 0.4056, + "step": 47282 + }, + { + "epoch": 1.82, + "learning_rate": 1.9714670428314953e-05, + "loss": 0.4074, + "step": 47283 + }, + { + "epoch": 1.82, + "learning_rate": 1.9714029238266225e-05, + "loss": 0.3404, + "step": 47284 + }, + { + "epoch": 1.82, + "learning_rate": 1.9713388048217493e-05, + "loss": 0.3496, + "step": 47285 + }, + { + "epoch": 1.82, + "learning_rate": 1.971274685816876e-05, + "loss": 0.3504, + "step": 47286 + }, + { + "epoch": 1.82, + "learning_rate": 1.971210566812003e-05, + "loss": 0.4205, + "step": 47287 + }, + { + "epoch": 1.82, + "learning_rate": 1.9711464478071302e-05, + "loss": 0.3434, + "step": 47288 + }, + { + "epoch": 1.82, + "learning_rate": 1.971082328802257e-05, + "loss": 0.226, + "step": 47289 + }, + { + "epoch": 1.82, + "learning_rate": 1.971018209797384e-05, + "loss": 0.4194, + "step": 47290 + }, + { + "epoch": 1.82, + "learning_rate": 1.970954090792511e-05, + "loss": 0.314, + "step": 47291 + }, + { + "epoch": 1.82, + "learning_rate": 1.9708899717876382e-05, + "loss": 0.2826, + "step": 47292 + }, + { + "epoch": 1.82, + "learning_rate": 1.970825852782765e-05, + "loss": 0.3083, + "step": 47293 + }, + { + "epoch": 1.82, + "learning_rate": 1.970761733777892e-05, + "loss": 0.286, + "step": 47294 + }, + { + "epoch": 1.82, + "learning_rate": 1.9706976147730187e-05, + "loss": 0.4099, + "step": 47295 + }, + { + "epoch": 1.82, + "learning_rate": 1.970633495768146e-05, + "loss": 0.2893, + "step": 47296 + }, + { + "epoch": 1.82, + "learning_rate": 1.9705693767632727e-05, + "loss": 0.3694, + "step": 47297 + }, + { + "epoch": 1.82, + "learning_rate": 1.9705052577583995e-05, + "loss": 0.3601, + "step": 47298 + }, + { + "epoch": 1.82, + "learning_rate": 1.9704411387535267e-05, + "loss": 0.4076, + "step": 47299 + }, + { + "epoch": 1.82, + "learning_rate": 1.9703770197486535e-05, + "loss": 0.3247, + "step": 47300 + }, + { + "epoch": 1.82, + "learning_rate": 1.9703129007437807e-05, + "loss": 0.2407, + "step": 47301 + }, + { + "epoch": 1.82, + "learning_rate": 1.9702487817389075e-05, + "loss": 0.326, + "step": 47302 + }, + { + "epoch": 1.82, + "learning_rate": 1.9701846627340344e-05, + "loss": 0.4102, + "step": 47303 + }, + { + "epoch": 1.82, + "learning_rate": 1.9701205437291615e-05, + "loss": 0.2575, + "step": 47304 + }, + { + "epoch": 1.82, + "learning_rate": 1.9700564247242884e-05, + "loss": 0.3377, + "step": 47305 + }, + { + "epoch": 1.82, + "learning_rate": 1.9699923057194152e-05, + "loss": 0.3225, + "step": 47306 + }, + { + "epoch": 1.82, + "learning_rate": 1.9699281867145424e-05, + "loss": 0.3641, + "step": 47307 + }, + { + "epoch": 1.82, + "learning_rate": 1.9698640677096692e-05, + "loss": 0.5528, + "step": 47308 + }, + { + "epoch": 1.82, + "learning_rate": 1.969799948704796e-05, + "loss": 0.3171, + "step": 47309 + }, + { + "epoch": 1.82, + "learning_rate": 1.9697358296999232e-05, + "loss": 0.3116, + "step": 47310 + }, + { + "epoch": 1.82, + "learning_rate": 1.96967171069505e-05, + "loss": 0.3598, + "step": 47311 + }, + { + "epoch": 1.82, + "learning_rate": 1.9696075916901772e-05, + "loss": 0.2977, + "step": 47312 + }, + { + "epoch": 1.82, + "learning_rate": 1.969543472685304e-05, + "loss": 0.4118, + "step": 47313 + }, + { + "epoch": 1.82, + "learning_rate": 1.969479353680431e-05, + "loss": 0.3849, + "step": 47314 + }, + { + "epoch": 1.82, + "learning_rate": 1.969415234675558e-05, + "loss": 0.2794, + "step": 47315 + }, + { + "epoch": 1.82, + "learning_rate": 1.969351115670685e-05, + "loss": 0.4832, + "step": 47316 + }, + { + "epoch": 1.82, + "learning_rate": 1.9692869966658117e-05, + "loss": 0.3486, + "step": 47317 + }, + { + "epoch": 1.82, + "learning_rate": 1.9692228776609385e-05, + "loss": 0.4075, + "step": 47318 + }, + { + "epoch": 1.82, + "learning_rate": 1.9691587586560657e-05, + "loss": 0.3458, + "step": 47319 + }, + { + "epoch": 1.82, + "learning_rate": 1.969094639651193e-05, + "loss": 0.2827, + "step": 47320 + }, + { + "epoch": 1.82, + "learning_rate": 1.9690305206463197e-05, + "loss": 0.2215, + "step": 47321 + }, + { + "epoch": 1.82, + "learning_rate": 1.9689664016414466e-05, + "loss": 0.268, + "step": 47322 + }, + { + "epoch": 1.82, + "learning_rate": 1.9689022826365737e-05, + "loss": 0.3578, + "step": 47323 + }, + { + "epoch": 1.82, + "learning_rate": 1.9688381636317006e-05, + "loss": 0.3436, + "step": 47324 + }, + { + "epoch": 1.82, + "learning_rate": 1.9687740446268274e-05, + "loss": 0.5067, + "step": 47325 + }, + { + "epoch": 1.82, + "learning_rate": 1.9687099256219542e-05, + "loss": 0.2519, + "step": 47326 + }, + { + "epoch": 1.82, + "learning_rate": 1.9686458066170814e-05, + "loss": 0.2649, + "step": 47327 + }, + { + "epoch": 1.82, + "learning_rate": 1.9685816876122086e-05, + "loss": 0.4179, + "step": 47328 + }, + { + "epoch": 1.82, + "learning_rate": 1.9685175686073354e-05, + "loss": 0.3208, + "step": 47329 + }, + { + "epoch": 1.82, + "learning_rate": 1.9684534496024622e-05, + "loss": 0.4103, + "step": 47330 + }, + { + "epoch": 1.82, + "learning_rate": 1.968389330597589e-05, + "loss": 0.382, + "step": 47331 + }, + { + "epoch": 1.82, + "learning_rate": 1.9683252115927162e-05, + "loss": 0.2559, + "step": 47332 + }, + { + "epoch": 1.82, + "learning_rate": 1.968261092587843e-05, + "loss": 0.3509, + "step": 47333 + }, + { + "epoch": 1.82, + "learning_rate": 1.96819697358297e-05, + "loss": 0.2395, + "step": 47334 + }, + { + "epoch": 1.82, + "learning_rate": 1.968132854578097e-05, + "loss": 0.3173, + "step": 47335 + }, + { + "epoch": 1.82, + "learning_rate": 1.9680687355732242e-05, + "loss": 0.3927, + "step": 47336 + }, + { + "epoch": 1.82, + "learning_rate": 1.968004616568351e-05, + "loss": 0.2314, + "step": 47337 + }, + { + "epoch": 1.82, + "learning_rate": 1.967940497563478e-05, + "loss": 0.4366, + "step": 47338 + }, + { + "epoch": 1.82, + "learning_rate": 1.9678763785586047e-05, + "loss": 0.349, + "step": 47339 + }, + { + "epoch": 1.82, + "learning_rate": 1.967812259553732e-05, + "loss": 0.4482, + "step": 47340 + }, + { + "epoch": 1.82, + "learning_rate": 1.9677481405488588e-05, + "loss": 0.2606, + "step": 47341 + }, + { + "epoch": 1.82, + "learning_rate": 1.9676840215439856e-05, + "loss": 0.4426, + "step": 47342 + }, + { + "epoch": 1.82, + "learning_rate": 1.9676199025391128e-05, + "loss": 0.4251, + "step": 47343 + }, + { + "epoch": 1.82, + "learning_rate": 1.9675557835342396e-05, + "loss": 0.2262, + "step": 47344 + }, + { + "epoch": 1.82, + "learning_rate": 1.9674916645293668e-05, + "loss": 0.3314, + "step": 47345 + }, + { + "epoch": 1.82, + "learning_rate": 1.9674275455244936e-05, + "loss": 0.1781, + "step": 47346 + }, + { + "epoch": 1.82, + "learning_rate": 1.9673634265196204e-05, + "loss": 0.3535, + "step": 47347 + }, + { + "epoch": 1.82, + "learning_rate": 1.9672993075147476e-05, + "loss": 0.3463, + "step": 47348 + }, + { + "epoch": 1.82, + "learning_rate": 1.9672351885098744e-05, + "loss": 0.4882, + "step": 47349 + }, + { + "epoch": 1.82, + "learning_rate": 1.9671710695050013e-05, + "loss": 0.3641, + "step": 47350 + }, + { + "epoch": 1.82, + "learning_rate": 1.9671069505001284e-05, + "loss": 0.3035, + "step": 47351 + }, + { + "epoch": 1.82, + "learning_rate": 1.9670428314952553e-05, + "loss": 0.2855, + "step": 47352 + }, + { + "epoch": 1.82, + "learning_rate": 1.966978712490382e-05, + "loss": 0.3564, + "step": 47353 + }, + { + "epoch": 1.82, + "learning_rate": 1.9669145934855093e-05, + "loss": 0.3899, + "step": 47354 + }, + { + "epoch": 1.82, + "learning_rate": 1.966850474480636e-05, + "loss": 0.4431, + "step": 47355 + }, + { + "epoch": 1.82, + "learning_rate": 1.9667863554757633e-05, + "loss": 0.3253, + "step": 47356 + }, + { + "epoch": 1.82, + "learning_rate": 1.96672223647089e-05, + "loss": 0.2556, + "step": 47357 + }, + { + "epoch": 1.82, + "learning_rate": 1.966658117466017e-05, + "loss": 0.2042, + "step": 47358 + }, + { + "epoch": 1.82, + "learning_rate": 1.966593998461144e-05, + "loss": 0.4001, + "step": 47359 + }, + { + "epoch": 1.82, + "learning_rate": 1.966529879456271e-05, + "loss": 0.2352, + "step": 47360 + }, + { + "epoch": 1.82, + "learning_rate": 1.9664657604513978e-05, + "loss": 0.2791, + "step": 47361 + }, + { + "epoch": 1.82, + "learning_rate": 1.9664016414465246e-05, + "loss": 0.3081, + "step": 47362 + }, + { + "epoch": 1.82, + "learning_rate": 1.9663375224416518e-05, + "loss": 0.4856, + "step": 47363 + }, + { + "epoch": 1.82, + "learning_rate": 1.966273403436779e-05, + "loss": 0.3822, + "step": 47364 + }, + { + "epoch": 1.82, + "learning_rate": 1.9662092844319058e-05, + "loss": 0.4487, + "step": 47365 + }, + { + "epoch": 1.82, + "learning_rate": 1.9661451654270326e-05, + "loss": 0.28, + "step": 47366 + }, + { + "epoch": 1.82, + "learning_rate": 1.9660810464221595e-05, + "loss": 0.3584, + "step": 47367 + }, + { + "epoch": 1.82, + "learning_rate": 1.9660169274172866e-05, + "loss": 0.353, + "step": 47368 + }, + { + "epoch": 1.82, + "learning_rate": 1.9659528084124135e-05, + "loss": 0.3928, + "step": 47369 + }, + { + "epoch": 1.82, + "learning_rate": 1.9658886894075403e-05, + "loss": 0.3483, + "step": 47370 + }, + { + "epoch": 1.82, + "learning_rate": 1.9658245704026675e-05, + "loss": 0.3335, + "step": 47371 + }, + { + "epoch": 1.82, + "learning_rate": 1.9657604513977946e-05, + "loss": 0.2654, + "step": 47372 + }, + { + "epoch": 1.82, + "learning_rate": 1.9656963323929215e-05, + "loss": 0.5026, + "step": 47373 + }, + { + "epoch": 1.82, + "learning_rate": 1.9656322133880483e-05, + "loss": 0.2806, + "step": 47374 + }, + { + "epoch": 1.82, + "learning_rate": 1.965568094383175e-05, + "loss": 0.2942, + "step": 47375 + }, + { + "epoch": 1.82, + "learning_rate": 1.9655039753783023e-05, + "loss": 0.3775, + "step": 47376 + }, + { + "epoch": 1.82, + "learning_rate": 1.965439856373429e-05, + "loss": 0.2291, + "step": 47377 + }, + { + "epoch": 1.82, + "learning_rate": 1.965375737368556e-05, + "loss": 0.3359, + "step": 47378 + }, + { + "epoch": 1.82, + "learning_rate": 1.965311618363683e-05, + "loss": 0.4195, + "step": 47379 + }, + { + "epoch": 1.82, + "learning_rate": 1.96524749935881e-05, + "loss": 0.2092, + "step": 47380 + }, + { + "epoch": 1.82, + "learning_rate": 1.965183380353937e-05, + "loss": 0.2832, + "step": 47381 + }, + { + "epoch": 1.82, + "learning_rate": 1.965119261349064e-05, + "loss": 0.272, + "step": 47382 + }, + { + "epoch": 1.82, + "learning_rate": 1.9650551423441908e-05, + "loss": 0.3763, + "step": 47383 + }, + { + "epoch": 1.82, + "learning_rate": 1.964991023339318e-05, + "loss": 0.3873, + "step": 47384 + }, + { + "epoch": 1.82, + "learning_rate": 1.9649269043344448e-05, + "loss": 0.281, + "step": 47385 + }, + { + "epoch": 1.82, + "learning_rate": 1.9648627853295716e-05, + "loss": 0.3512, + "step": 47386 + }, + { + "epoch": 1.82, + "learning_rate": 1.9647986663246988e-05, + "loss": 0.3446, + "step": 47387 + }, + { + "epoch": 1.82, + "learning_rate": 1.9647345473198257e-05, + "loss": 0.368, + "step": 47388 + }, + { + "epoch": 1.82, + "learning_rate": 1.9646704283149525e-05, + "loss": 0.3332, + "step": 47389 + }, + { + "epoch": 1.82, + "learning_rate": 1.9646063093100797e-05, + "loss": 0.3378, + "step": 47390 + }, + { + "epoch": 1.82, + "learning_rate": 1.9645421903052065e-05, + "loss": 0.2545, + "step": 47391 + }, + { + "epoch": 1.82, + "learning_rate": 1.9644780713003337e-05, + "loss": 0.2905, + "step": 47392 + }, + { + "epoch": 1.82, + "learning_rate": 1.9644139522954605e-05, + "loss": 0.4455, + "step": 47393 + }, + { + "epoch": 1.82, + "learning_rate": 1.9643498332905873e-05, + "loss": 0.2814, + "step": 47394 + }, + { + "epoch": 1.82, + "learning_rate": 1.9642857142857145e-05, + "loss": 0.48, + "step": 47395 + }, + { + "epoch": 1.82, + "learning_rate": 1.9642215952808413e-05, + "loss": 0.272, + "step": 47396 + }, + { + "epoch": 1.82, + "learning_rate": 1.964157476275968e-05, + "loss": 0.2286, + "step": 47397 + }, + { + "epoch": 1.82, + "learning_rate": 1.964093357271095e-05, + "loss": 0.3631, + "step": 47398 + }, + { + "epoch": 1.82, + "learning_rate": 1.9640292382662222e-05, + "loss": 0.2295, + "step": 47399 + }, + { + "epoch": 1.82, + "learning_rate": 1.9639651192613493e-05, + "loss": 0.3401, + "step": 47400 + }, + { + "epoch": 1.82, + "learning_rate": 1.9639010002564762e-05, + "loss": 0.3179, + "step": 47401 + }, + { + "epoch": 1.82, + "learning_rate": 1.963836881251603e-05, + "loss": 0.269, + "step": 47402 + }, + { + "epoch": 1.82, + "learning_rate": 1.9637727622467302e-05, + "loss": 0.3252, + "step": 47403 + }, + { + "epoch": 1.82, + "learning_rate": 1.963708643241857e-05, + "loss": 0.2867, + "step": 47404 + }, + { + "epoch": 1.82, + "learning_rate": 1.963644524236984e-05, + "loss": 0.4064, + "step": 47405 + }, + { + "epoch": 1.82, + "learning_rate": 1.9635804052321107e-05, + "loss": 0.2177, + "step": 47406 + }, + { + "epoch": 1.82, + "learning_rate": 1.963516286227238e-05, + "loss": 0.2482, + "step": 47407 + }, + { + "epoch": 1.82, + "learning_rate": 1.963452167222365e-05, + "loss": 0.2206, + "step": 47408 + }, + { + "epoch": 1.82, + "learning_rate": 1.963388048217492e-05, + "loss": 0.3118, + "step": 47409 + }, + { + "epoch": 1.82, + "learning_rate": 1.9633239292126187e-05, + "loss": 0.3021, + "step": 47410 + }, + { + "epoch": 1.82, + "learning_rate": 1.9632598102077455e-05, + "loss": 0.271, + "step": 47411 + }, + { + "epoch": 1.82, + "learning_rate": 1.9631956912028727e-05, + "loss": 0.2718, + "step": 47412 + }, + { + "epoch": 1.82, + "learning_rate": 1.9631315721979995e-05, + "loss": 0.2918, + "step": 47413 + }, + { + "epoch": 1.82, + "learning_rate": 1.9630674531931264e-05, + "loss": 0.2458, + "step": 47414 + }, + { + "epoch": 1.82, + "learning_rate": 1.9630033341882535e-05, + "loss": 0.2346, + "step": 47415 + }, + { + "epoch": 1.82, + "learning_rate": 1.9629392151833807e-05, + "loss": 0.2649, + "step": 47416 + }, + { + "epoch": 1.82, + "learning_rate": 1.9628750961785075e-05, + "loss": 0.2283, + "step": 47417 + }, + { + "epoch": 1.82, + "learning_rate": 1.9628109771736344e-05, + "loss": 0.2853, + "step": 47418 + }, + { + "epoch": 1.82, + "learning_rate": 1.9627468581687612e-05, + "loss": 0.2108, + "step": 47419 + }, + { + "epoch": 1.82, + "learning_rate": 1.9626827391638884e-05, + "loss": 0.2819, + "step": 47420 + }, + { + "epoch": 1.82, + "learning_rate": 1.9626186201590152e-05, + "loss": 0.34, + "step": 47421 + }, + { + "epoch": 1.82, + "learning_rate": 1.962554501154142e-05, + "loss": 0.376, + "step": 47422 + }, + { + "epoch": 1.82, + "learning_rate": 1.9624903821492692e-05, + "loss": 0.2516, + "step": 47423 + }, + { + "epoch": 1.82, + "learning_rate": 1.962426263144396e-05, + "loss": 0.2794, + "step": 47424 + }, + { + "epoch": 1.82, + "learning_rate": 1.9623621441395232e-05, + "loss": 0.3466, + "step": 47425 + }, + { + "epoch": 1.82, + "learning_rate": 1.96229802513465e-05, + "loss": 0.2957, + "step": 47426 + }, + { + "epoch": 1.82, + "learning_rate": 1.962233906129777e-05, + "loss": 0.4032, + "step": 47427 + }, + { + "epoch": 1.82, + "learning_rate": 1.962169787124904e-05, + "loss": 0.2708, + "step": 47428 + }, + { + "epoch": 1.82, + "learning_rate": 1.962105668120031e-05, + "loss": 0.3922, + "step": 47429 + }, + { + "epoch": 1.82, + "learning_rate": 1.9620415491151577e-05, + "loss": 0.4017, + "step": 47430 + }, + { + "epoch": 1.82, + "learning_rate": 1.961977430110285e-05, + "loss": 0.2727, + "step": 47431 + }, + { + "epoch": 1.82, + "learning_rate": 1.9619133111054117e-05, + "loss": 0.4018, + "step": 47432 + }, + { + "epoch": 1.82, + "learning_rate": 1.9618491921005386e-05, + "loss": 0.3714, + "step": 47433 + }, + { + "epoch": 1.82, + "learning_rate": 1.9617850730956657e-05, + "loss": 0.3241, + "step": 47434 + }, + { + "epoch": 1.82, + "learning_rate": 1.9617209540907926e-05, + "loss": 0.5805, + "step": 47435 + }, + { + "epoch": 1.82, + "learning_rate": 1.9616568350859197e-05, + "loss": 0.3904, + "step": 47436 + }, + { + "epoch": 1.82, + "learning_rate": 1.9615927160810466e-05, + "loss": 0.3158, + "step": 47437 + }, + { + "epoch": 1.82, + "learning_rate": 1.9615285970761734e-05, + "loss": 0.4958, + "step": 47438 + }, + { + "epoch": 1.82, + "learning_rate": 1.9614644780713006e-05, + "loss": 0.2178, + "step": 47439 + }, + { + "epoch": 1.82, + "learning_rate": 1.9614003590664274e-05, + "loss": 0.3103, + "step": 47440 + }, + { + "epoch": 1.82, + "learning_rate": 1.9613362400615542e-05, + "loss": 0.3366, + "step": 47441 + }, + { + "epoch": 1.82, + "learning_rate": 1.961272121056681e-05, + "loss": 0.3177, + "step": 47442 + }, + { + "epoch": 1.82, + "learning_rate": 1.9612080020518082e-05, + "loss": 0.2716, + "step": 47443 + }, + { + "epoch": 1.82, + "learning_rate": 1.9611438830469354e-05, + "loss": 0.3313, + "step": 47444 + }, + { + "epoch": 1.82, + "learning_rate": 1.9610797640420622e-05, + "loss": 0.3659, + "step": 47445 + }, + { + "epoch": 1.82, + "learning_rate": 1.961015645037189e-05, + "loss": 0.4114, + "step": 47446 + }, + { + "epoch": 1.82, + "learning_rate": 1.9609515260323162e-05, + "loss": 0.3404, + "step": 47447 + }, + { + "epoch": 1.82, + "learning_rate": 1.960887407027443e-05, + "loss": 0.2478, + "step": 47448 + }, + { + "epoch": 1.82, + "learning_rate": 1.96082328802257e-05, + "loss": 0.2557, + "step": 47449 + }, + { + "epoch": 1.82, + "learning_rate": 1.9607591690176967e-05, + "loss": 0.2812, + "step": 47450 + }, + { + "epoch": 1.82, + "learning_rate": 1.960695050012824e-05, + "loss": 0.3283, + "step": 47451 + }, + { + "epoch": 1.82, + "learning_rate": 1.960630931007951e-05, + "loss": 0.2722, + "step": 47452 + }, + { + "epoch": 1.82, + "learning_rate": 1.960566812003078e-05, + "loss": 0.4114, + "step": 47453 + }, + { + "epoch": 1.82, + "learning_rate": 1.9605026929982048e-05, + "loss": 0.3487, + "step": 47454 + }, + { + "epoch": 1.82, + "learning_rate": 1.9604385739933316e-05, + "loss": 0.3379, + "step": 47455 + }, + { + "epoch": 1.82, + "learning_rate": 1.9603744549884588e-05, + "loss": 0.3495, + "step": 47456 + }, + { + "epoch": 1.82, + "learning_rate": 1.9603103359835856e-05, + "loss": 0.2651, + "step": 47457 + }, + { + "epoch": 1.82, + "learning_rate": 1.9602462169787124e-05, + "loss": 0.2689, + "step": 47458 + }, + { + "epoch": 1.82, + "learning_rate": 1.9601820979738396e-05, + "loss": 0.2993, + "step": 47459 + }, + { + "epoch": 1.82, + "learning_rate": 1.9601179789689664e-05, + "loss": 0.2284, + "step": 47460 + }, + { + "epoch": 1.82, + "learning_rate": 1.9600538599640936e-05, + "loss": 0.3286, + "step": 47461 + }, + { + "epoch": 1.82, + "learning_rate": 1.9599897409592204e-05, + "loss": 0.3055, + "step": 47462 + }, + { + "epoch": 1.82, + "learning_rate": 1.9599256219543473e-05, + "loss": 0.3794, + "step": 47463 + }, + { + "epoch": 1.82, + "learning_rate": 1.9598615029494744e-05, + "loss": 0.3936, + "step": 47464 + }, + { + "epoch": 1.82, + "learning_rate": 1.9597973839446013e-05, + "loss": 0.266, + "step": 47465 + }, + { + "epoch": 1.82, + "learning_rate": 1.959733264939728e-05, + "loss": 0.2972, + "step": 47466 + }, + { + "epoch": 1.82, + "learning_rate": 1.9596691459348553e-05, + "loss": 0.2138, + "step": 47467 + }, + { + "epoch": 1.82, + "learning_rate": 1.959605026929982e-05, + "loss": 0.3403, + "step": 47468 + }, + { + "epoch": 1.83, + "learning_rate": 1.959540907925109e-05, + "loss": 0.4551, + "step": 47469 + }, + { + "epoch": 1.83, + "learning_rate": 1.959476788920236e-05, + "loss": 0.1995, + "step": 47470 + }, + { + "epoch": 1.83, + "learning_rate": 1.959412669915363e-05, + "loss": 0.2449, + "step": 47471 + }, + { + "epoch": 1.83, + "learning_rate": 1.95934855091049e-05, + "loss": 0.2326, + "step": 47472 + }, + { + "epoch": 1.83, + "learning_rate": 1.959284431905617e-05, + "loss": 0.4185, + "step": 47473 + }, + { + "epoch": 1.83, + "learning_rate": 1.9592203129007438e-05, + "loss": 0.3528, + "step": 47474 + }, + { + "epoch": 1.83, + "learning_rate": 1.959156193895871e-05, + "loss": 0.291, + "step": 47475 + }, + { + "epoch": 1.83, + "learning_rate": 1.9590920748909978e-05, + "loss": 0.2893, + "step": 47476 + }, + { + "epoch": 1.83, + "learning_rate": 1.9590279558861246e-05, + "loss": 0.2639, + "step": 47477 + }, + { + "epoch": 1.83, + "learning_rate": 1.9589638368812515e-05, + "loss": 0.338, + "step": 47478 + }, + { + "epoch": 1.83, + "learning_rate": 1.9588997178763786e-05, + "loss": 0.4789, + "step": 47479 + }, + { + "epoch": 1.83, + "learning_rate": 1.9588355988715058e-05, + "loss": 0.2899, + "step": 47480 + }, + { + "epoch": 1.83, + "learning_rate": 1.9587714798666326e-05, + "loss": 0.3901, + "step": 47481 + }, + { + "epoch": 1.83, + "learning_rate": 1.9587073608617595e-05, + "loss": 0.4208, + "step": 47482 + }, + { + "epoch": 1.83, + "learning_rate": 1.9586432418568866e-05, + "loss": 0.3091, + "step": 47483 + }, + { + "epoch": 1.83, + "learning_rate": 1.9585791228520135e-05, + "loss": 0.2579, + "step": 47484 + }, + { + "epoch": 1.83, + "learning_rate": 1.9585150038471403e-05, + "loss": 0.3903, + "step": 47485 + }, + { + "epoch": 1.83, + "learning_rate": 1.958450884842267e-05, + "loss": 0.3543, + "step": 47486 + }, + { + "epoch": 1.83, + "learning_rate": 1.9583867658373943e-05, + "loss": 0.3166, + "step": 47487 + }, + { + "epoch": 1.83, + "learning_rate": 1.9583226468325215e-05, + "loss": 0.3203, + "step": 47488 + }, + { + "epoch": 1.83, + "learning_rate": 1.9582585278276483e-05, + "loss": 0.4154, + "step": 47489 + }, + { + "epoch": 1.83, + "learning_rate": 1.958194408822775e-05, + "loss": 0.2828, + "step": 47490 + }, + { + "epoch": 1.83, + "learning_rate": 1.958130289817902e-05, + "loss": 0.2663, + "step": 47491 + }, + { + "epoch": 1.83, + "learning_rate": 1.958066170813029e-05, + "loss": 0.299, + "step": 47492 + }, + { + "epoch": 1.83, + "learning_rate": 1.958002051808156e-05, + "loss": 0.4365, + "step": 47493 + }, + { + "epoch": 1.83, + "learning_rate": 1.9579379328032828e-05, + "loss": 0.3151, + "step": 47494 + }, + { + "epoch": 1.83, + "learning_rate": 1.95787381379841e-05, + "loss": 0.3661, + "step": 47495 + }, + { + "epoch": 1.83, + "learning_rate": 1.9578096947935368e-05, + "loss": 0.3811, + "step": 47496 + }, + { + "epoch": 1.83, + "learning_rate": 1.957745575788664e-05, + "loss": 0.3977, + "step": 47497 + }, + { + "epoch": 1.83, + "learning_rate": 1.9576814567837908e-05, + "loss": 0.3358, + "step": 47498 + }, + { + "epoch": 1.83, + "learning_rate": 1.9576173377789177e-05, + "loss": 0.3269, + "step": 47499 + }, + { + "epoch": 1.83, + "learning_rate": 1.9575532187740448e-05, + "loss": 0.3854, + "step": 47500 + }, + { + "epoch": 1.83, + "learning_rate": 1.9574890997691717e-05, + "loss": 0.3291, + "step": 47501 + }, + { + "epoch": 1.83, + "learning_rate": 1.9574249807642985e-05, + "loss": 0.3502, + "step": 47502 + }, + { + "epoch": 1.83, + "learning_rate": 1.9573608617594257e-05, + "loss": 0.3753, + "step": 47503 + }, + { + "epoch": 1.83, + "learning_rate": 1.9572967427545525e-05, + "loss": 0.3129, + "step": 47504 + }, + { + "epoch": 1.83, + "learning_rate": 1.9572326237496793e-05, + "loss": 0.478, + "step": 47505 + }, + { + "epoch": 1.83, + "learning_rate": 1.9571685047448065e-05, + "loss": 0.354, + "step": 47506 + }, + { + "epoch": 1.83, + "learning_rate": 1.9571043857399333e-05, + "loss": 0.4398, + "step": 47507 + }, + { + "epoch": 1.83, + "learning_rate": 1.9570402667350605e-05, + "loss": 0.3491, + "step": 47508 + }, + { + "epoch": 1.83, + "learning_rate": 1.9569761477301873e-05, + "loss": 0.2496, + "step": 47509 + }, + { + "epoch": 1.83, + "learning_rate": 1.9569120287253142e-05, + "loss": 0.3596, + "step": 47510 + }, + { + "epoch": 1.83, + "learning_rate": 1.9568479097204413e-05, + "loss": 0.25, + "step": 47511 + }, + { + "epoch": 1.83, + "learning_rate": 1.9567837907155682e-05, + "loss": 0.2306, + "step": 47512 + }, + { + "epoch": 1.83, + "learning_rate": 1.956719671710695e-05, + "loss": 0.3587, + "step": 47513 + }, + { + "epoch": 1.83, + "learning_rate": 1.9566555527058222e-05, + "loss": 0.4222, + "step": 47514 + }, + { + "epoch": 1.83, + "learning_rate": 1.956591433700949e-05, + "loss": 0.3066, + "step": 47515 + }, + { + "epoch": 1.83, + "learning_rate": 1.9565273146960762e-05, + "loss": 0.3709, + "step": 47516 + }, + { + "epoch": 1.83, + "learning_rate": 1.956463195691203e-05, + "loss": 0.4982, + "step": 47517 + }, + { + "epoch": 1.83, + "learning_rate": 1.95639907668633e-05, + "loss": 0.2622, + "step": 47518 + }, + { + "epoch": 1.83, + "learning_rate": 1.956334957681457e-05, + "loss": 0.3784, + "step": 47519 + }, + { + "epoch": 1.83, + "learning_rate": 1.956270838676584e-05, + "loss": 0.3867, + "step": 47520 + }, + { + "epoch": 1.83, + "learning_rate": 1.9562067196717107e-05, + "loss": 0.3652, + "step": 47521 + }, + { + "epoch": 1.83, + "learning_rate": 1.9561426006668375e-05, + "loss": 0.2803, + "step": 47522 + }, + { + "epoch": 1.83, + "learning_rate": 1.9560784816619647e-05, + "loss": 0.3051, + "step": 47523 + }, + { + "epoch": 1.83, + "learning_rate": 1.956014362657092e-05, + "loss": 0.3193, + "step": 47524 + }, + { + "epoch": 1.83, + "learning_rate": 1.9559502436522187e-05, + "loss": 0.3947, + "step": 47525 + }, + { + "epoch": 1.83, + "learning_rate": 1.9558861246473455e-05, + "loss": 0.2374, + "step": 47526 + }, + { + "epoch": 1.83, + "learning_rate": 1.9558220056424727e-05, + "loss": 0.3864, + "step": 47527 + }, + { + "epoch": 1.83, + "learning_rate": 1.9557578866375995e-05, + "loss": 0.2709, + "step": 47528 + }, + { + "epoch": 1.83, + "learning_rate": 1.9556937676327264e-05, + "loss": 0.2433, + "step": 47529 + }, + { + "epoch": 1.83, + "learning_rate": 1.9556296486278532e-05, + "loss": 0.3864, + "step": 47530 + }, + { + "epoch": 1.83, + "learning_rate": 1.9555655296229804e-05, + "loss": 0.3094, + "step": 47531 + }, + { + "epoch": 1.83, + "learning_rate": 1.9555014106181075e-05, + "loss": 0.3233, + "step": 47532 + }, + { + "epoch": 1.83, + "learning_rate": 1.9554372916132344e-05, + "loss": 0.3549, + "step": 47533 + }, + { + "epoch": 1.83, + "learning_rate": 1.9553731726083612e-05, + "loss": 0.4086, + "step": 47534 + }, + { + "epoch": 1.83, + "learning_rate": 1.955309053603488e-05, + "loss": 0.2512, + "step": 47535 + }, + { + "epoch": 1.83, + "learning_rate": 1.9552449345986152e-05, + "loss": 0.3825, + "step": 47536 + }, + { + "epoch": 1.83, + "learning_rate": 1.955180815593742e-05, + "loss": 0.3897, + "step": 47537 + }, + { + "epoch": 1.83, + "learning_rate": 1.955116696588869e-05, + "loss": 0.2615, + "step": 47538 + }, + { + "epoch": 1.83, + "learning_rate": 1.955052577583996e-05, + "loss": 0.3353, + "step": 47539 + }, + { + "epoch": 1.83, + "learning_rate": 1.954988458579123e-05, + "loss": 0.3941, + "step": 47540 + }, + { + "epoch": 1.83, + "learning_rate": 1.95492433957425e-05, + "loss": 0.5054, + "step": 47541 + }, + { + "epoch": 1.83, + "learning_rate": 1.954860220569377e-05, + "loss": 0.3521, + "step": 47542 + }, + { + "epoch": 1.83, + "learning_rate": 1.9547961015645037e-05, + "loss": 0.3271, + "step": 47543 + }, + { + "epoch": 1.83, + "learning_rate": 1.954731982559631e-05, + "loss": 0.412, + "step": 47544 + }, + { + "epoch": 1.83, + "learning_rate": 1.9546678635547577e-05, + "loss": 0.4528, + "step": 47545 + }, + { + "epoch": 1.83, + "learning_rate": 1.9546037445498846e-05, + "loss": 0.3178, + "step": 47546 + }, + { + "epoch": 1.83, + "learning_rate": 1.9545396255450117e-05, + "loss": 0.3116, + "step": 47547 + }, + { + "epoch": 1.83, + "learning_rate": 1.9544755065401386e-05, + "loss": 0.3323, + "step": 47548 + }, + { + "epoch": 1.83, + "learning_rate": 1.9544113875352654e-05, + "loss": 0.2365, + "step": 47549 + }, + { + "epoch": 1.83, + "learning_rate": 1.9543472685303926e-05, + "loss": 0.3849, + "step": 47550 + }, + { + "epoch": 1.83, + "learning_rate": 1.9542831495255194e-05, + "loss": 0.2911, + "step": 47551 + }, + { + "epoch": 1.83, + "learning_rate": 1.9542190305206466e-05, + "loss": 0.3597, + "step": 47552 + }, + { + "epoch": 1.83, + "learning_rate": 1.9541549115157734e-05, + "loss": 0.4364, + "step": 47553 + }, + { + "epoch": 1.83, + "learning_rate": 1.9540907925109002e-05, + "loss": 0.378, + "step": 47554 + }, + { + "epoch": 1.83, + "learning_rate": 1.9540266735060274e-05, + "loss": 0.2821, + "step": 47555 + }, + { + "epoch": 1.83, + "learning_rate": 1.9539625545011542e-05, + "loss": 0.275, + "step": 47556 + }, + { + "epoch": 1.83, + "learning_rate": 1.953898435496281e-05, + "loss": 0.2668, + "step": 47557 + }, + { + "epoch": 1.83, + "learning_rate": 1.953834316491408e-05, + "loss": 0.3372, + "step": 47558 + }, + { + "epoch": 1.83, + "learning_rate": 1.953770197486535e-05, + "loss": 0.2409, + "step": 47559 + }, + { + "epoch": 1.83, + "learning_rate": 1.9537060784816622e-05, + "loss": 0.2519, + "step": 47560 + }, + { + "epoch": 1.83, + "learning_rate": 1.953641959476789e-05, + "loss": 0.3018, + "step": 47561 + }, + { + "epoch": 1.83, + "learning_rate": 1.953577840471916e-05, + "loss": 0.3345, + "step": 47562 + }, + { + "epoch": 1.83, + "learning_rate": 1.953513721467043e-05, + "loss": 0.3724, + "step": 47563 + }, + { + "epoch": 1.83, + "learning_rate": 1.95344960246217e-05, + "loss": 0.2765, + "step": 47564 + }, + { + "epoch": 1.83, + "learning_rate": 1.9533854834572968e-05, + "loss": 0.3162, + "step": 47565 + }, + { + "epoch": 1.83, + "learning_rate": 1.9533213644524236e-05, + "loss": 0.3268, + "step": 47566 + }, + { + "epoch": 1.83, + "learning_rate": 1.9532572454475508e-05, + "loss": 0.3047, + "step": 47567 + }, + { + "epoch": 1.83, + "learning_rate": 1.953193126442678e-05, + "loss": 0.5044, + "step": 47568 + }, + { + "epoch": 1.83, + "learning_rate": 1.9531290074378048e-05, + "loss": 0.47, + "step": 47569 + }, + { + "epoch": 1.83, + "learning_rate": 1.9530648884329316e-05, + "loss": 0.4304, + "step": 47570 + }, + { + "epoch": 1.83, + "learning_rate": 1.9530007694280584e-05, + "loss": 0.2026, + "step": 47571 + }, + { + "epoch": 1.83, + "learning_rate": 1.9529366504231856e-05, + "loss": 0.344, + "step": 47572 + }, + { + "epoch": 1.83, + "learning_rate": 1.9528725314183124e-05, + "loss": 0.2993, + "step": 47573 + }, + { + "epoch": 1.83, + "learning_rate": 1.9528084124134393e-05, + "loss": 0.4432, + "step": 47574 + }, + { + "epoch": 1.83, + "learning_rate": 1.9527442934085664e-05, + "loss": 0.3701, + "step": 47575 + }, + { + "epoch": 1.83, + "learning_rate": 1.9526801744036933e-05, + "loss": 0.3898, + "step": 47576 + }, + { + "epoch": 1.83, + "learning_rate": 1.9526160553988204e-05, + "loss": 0.3103, + "step": 47577 + }, + { + "epoch": 1.83, + "learning_rate": 1.9525519363939473e-05, + "loss": 0.4583, + "step": 47578 + }, + { + "epoch": 1.83, + "learning_rate": 1.952487817389074e-05, + "loss": 0.2411, + "step": 47579 + }, + { + "epoch": 1.83, + "learning_rate": 1.9524236983842013e-05, + "loss": 0.3482, + "step": 47580 + }, + { + "epoch": 1.83, + "learning_rate": 1.952359579379328e-05, + "loss": 0.362, + "step": 47581 + }, + { + "epoch": 1.83, + "learning_rate": 1.952295460374455e-05, + "loss": 0.4439, + "step": 47582 + }, + { + "epoch": 1.83, + "learning_rate": 1.952231341369582e-05, + "loss": 0.2428, + "step": 47583 + }, + { + "epoch": 1.83, + "learning_rate": 1.952167222364709e-05, + "loss": 0.2814, + "step": 47584 + }, + { + "epoch": 1.83, + "learning_rate": 1.9521031033598358e-05, + "loss": 0.396, + "step": 47585 + }, + { + "epoch": 1.83, + "learning_rate": 1.952038984354963e-05, + "loss": 0.3874, + "step": 47586 + }, + { + "epoch": 1.83, + "learning_rate": 1.9519748653500898e-05, + "loss": 0.2002, + "step": 47587 + }, + { + "epoch": 1.83, + "learning_rate": 1.951910746345217e-05, + "loss": 0.2899, + "step": 47588 + }, + { + "epoch": 1.83, + "learning_rate": 1.9518466273403438e-05, + "loss": 0.3506, + "step": 47589 + }, + { + "epoch": 1.83, + "learning_rate": 1.9517825083354706e-05, + "loss": 0.1897, + "step": 47590 + }, + { + "epoch": 1.83, + "learning_rate": 1.9517183893305978e-05, + "loss": 0.3143, + "step": 47591 + }, + { + "epoch": 1.83, + "learning_rate": 1.9516542703257246e-05, + "loss": 0.3051, + "step": 47592 + }, + { + "epoch": 1.83, + "learning_rate": 1.9515901513208515e-05, + "loss": 0.2948, + "step": 47593 + }, + { + "epoch": 1.83, + "learning_rate": 1.9515260323159783e-05, + "loss": 0.3389, + "step": 47594 + }, + { + "epoch": 1.83, + "learning_rate": 1.9514619133111055e-05, + "loss": 0.2938, + "step": 47595 + }, + { + "epoch": 1.83, + "learning_rate": 1.9513977943062326e-05, + "loss": 0.4757, + "step": 47596 + }, + { + "epoch": 1.83, + "learning_rate": 1.9513336753013595e-05, + "loss": 0.2529, + "step": 47597 + }, + { + "epoch": 1.83, + "learning_rate": 1.9512695562964863e-05, + "loss": 0.4533, + "step": 47598 + }, + { + "epoch": 1.83, + "learning_rate": 1.9512054372916135e-05, + "loss": 0.4485, + "step": 47599 + }, + { + "epoch": 1.83, + "learning_rate": 1.9511413182867403e-05, + "loss": 0.3963, + "step": 47600 + }, + { + "epoch": 1.83, + "learning_rate": 1.951077199281867e-05, + "loss": 0.3044, + "step": 47601 + }, + { + "epoch": 1.83, + "learning_rate": 1.951013080276994e-05, + "loss": 0.2858, + "step": 47602 + }, + { + "epoch": 1.83, + "learning_rate": 1.950948961272121e-05, + "loss": 0.4176, + "step": 47603 + }, + { + "epoch": 1.83, + "learning_rate": 1.9508848422672483e-05, + "loss": 0.3195, + "step": 47604 + }, + { + "epoch": 1.83, + "learning_rate": 1.950820723262375e-05, + "loss": 0.4638, + "step": 47605 + }, + { + "epoch": 1.83, + "learning_rate": 1.950756604257502e-05, + "loss": 0.2903, + "step": 47606 + }, + { + "epoch": 1.83, + "learning_rate": 1.950692485252629e-05, + "loss": 0.2925, + "step": 47607 + }, + { + "epoch": 1.83, + "learning_rate": 1.950628366247756e-05, + "loss": 0.407, + "step": 47608 + }, + { + "epoch": 1.83, + "learning_rate": 1.9505642472428828e-05, + "loss": 0.2821, + "step": 47609 + }, + { + "epoch": 1.83, + "learning_rate": 1.9505001282380097e-05, + "loss": 0.3762, + "step": 47610 + }, + { + "epoch": 1.83, + "learning_rate": 1.9504360092331368e-05, + "loss": 0.3004, + "step": 47611 + }, + { + "epoch": 1.83, + "learning_rate": 1.950371890228264e-05, + "loss": 0.4084, + "step": 47612 + }, + { + "epoch": 1.83, + "learning_rate": 1.9503077712233908e-05, + "loss": 0.38, + "step": 47613 + }, + { + "epoch": 1.83, + "learning_rate": 1.9502436522185177e-05, + "loss": 0.3742, + "step": 47614 + }, + { + "epoch": 1.83, + "learning_rate": 1.9501795332136445e-05, + "loss": 0.321, + "step": 47615 + }, + { + "epoch": 1.83, + "learning_rate": 1.9501154142087717e-05, + "loss": 0.2662, + "step": 47616 + }, + { + "epoch": 1.83, + "learning_rate": 1.9500512952038985e-05, + "loss": 0.4422, + "step": 47617 + }, + { + "epoch": 1.83, + "learning_rate": 1.9499871761990253e-05, + "loss": 0.2511, + "step": 47618 + }, + { + "epoch": 1.83, + "learning_rate": 1.9499230571941525e-05, + "loss": 0.3225, + "step": 47619 + }, + { + "epoch": 1.83, + "learning_rate": 1.9498589381892793e-05, + "loss": 0.3607, + "step": 47620 + }, + { + "epoch": 1.83, + "learning_rate": 1.9497948191844065e-05, + "loss": 0.3742, + "step": 47621 + }, + { + "epoch": 1.83, + "learning_rate": 1.9497307001795333e-05, + "loss": 0.1889, + "step": 47622 + }, + { + "epoch": 1.83, + "learning_rate": 1.9496665811746602e-05, + "loss": 0.3899, + "step": 47623 + }, + { + "epoch": 1.83, + "learning_rate": 1.9496024621697873e-05, + "loss": 0.3275, + "step": 47624 + }, + { + "epoch": 1.83, + "learning_rate": 1.9495383431649142e-05, + "loss": 0.4499, + "step": 47625 + }, + { + "epoch": 1.83, + "learning_rate": 1.949474224160041e-05, + "loss": 0.4515, + "step": 47626 + }, + { + "epoch": 1.83, + "learning_rate": 1.9494101051551682e-05, + "loss": 0.3504, + "step": 47627 + }, + { + "epoch": 1.83, + "learning_rate": 1.949345986150295e-05, + "loss": 0.3752, + "step": 47628 + }, + { + "epoch": 1.83, + "learning_rate": 1.949281867145422e-05, + "loss": 0.4053, + "step": 47629 + }, + { + "epoch": 1.83, + "learning_rate": 1.949217748140549e-05, + "loss": 0.3786, + "step": 47630 + }, + { + "epoch": 1.83, + "learning_rate": 1.949153629135676e-05, + "loss": 0.2742, + "step": 47631 + }, + { + "epoch": 1.83, + "learning_rate": 1.949089510130803e-05, + "loss": 0.3454, + "step": 47632 + }, + { + "epoch": 1.83, + "learning_rate": 1.94902539112593e-05, + "loss": 0.3244, + "step": 47633 + }, + { + "epoch": 1.83, + "learning_rate": 1.9489612721210567e-05, + "loss": 0.3468, + "step": 47634 + }, + { + "epoch": 1.83, + "learning_rate": 1.948897153116184e-05, + "loss": 0.3561, + "step": 47635 + }, + { + "epoch": 1.83, + "learning_rate": 1.9488330341113107e-05, + "loss": 0.2531, + "step": 47636 + }, + { + "epoch": 1.83, + "learning_rate": 1.9487689151064375e-05, + "loss": 0.3338, + "step": 47637 + }, + { + "epoch": 1.83, + "learning_rate": 1.9487047961015644e-05, + "loss": 0.3722, + "step": 47638 + }, + { + "epoch": 1.83, + "learning_rate": 1.9486406770966915e-05, + "loss": 0.3309, + "step": 47639 + }, + { + "epoch": 1.83, + "learning_rate": 1.9485765580918187e-05, + "loss": 0.2948, + "step": 47640 + }, + { + "epoch": 1.83, + "learning_rate": 1.9485124390869455e-05, + "loss": 0.4323, + "step": 47641 + }, + { + "epoch": 1.83, + "learning_rate": 1.9484483200820724e-05, + "loss": 0.288, + "step": 47642 + }, + { + "epoch": 1.83, + "learning_rate": 1.9483842010771995e-05, + "loss": 0.2038, + "step": 47643 + }, + { + "epoch": 1.83, + "learning_rate": 1.9483200820723264e-05, + "loss": 0.3331, + "step": 47644 + }, + { + "epoch": 1.83, + "learning_rate": 1.9482559630674532e-05, + "loss": 0.4672, + "step": 47645 + }, + { + "epoch": 1.83, + "learning_rate": 1.94819184406258e-05, + "loss": 0.3263, + "step": 47646 + }, + { + "epoch": 1.83, + "learning_rate": 1.9481277250577072e-05, + "loss": 0.316, + "step": 47647 + }, + { + "epoch": 1.83, + "learning_rate": 1.9480636060528344e-05, + "loss": 0.3399, + "step": 47648 + }, + { + "epoch": 1.83, + "learning_rate": 1.9479994870479612e-05, + "loss": 0.3164, + "step": 47649 + }, + { + "epoch": 1.83, + "learning_rate": 1.947935368043088e-05, + "loss": 0.4708, + "step": 47650 + }, + { + "epoch": 1.83, + "learning_rate": 1.9478712490382152e-05, + "loss": 0.2265, + "step": 47651 + }, + { + "epoch": 1.83, + "learning_rate": 1.947807130033342e-05, + "loss": 0.4179, + "step": 47652 + }, + { + "epoch": 1.83, + "learning_rate": 1.947743011028469e-05, + "loss": 0.2675, + "step": 47653 + }, + { + "epoch": 1.83, + "learning_rate": 1.9476788920235957e-05, + "loss": 0.3138, + "step": 47654 + }, + { + "epoch": 1.83, + "learning_rate": 1.947614773018723e-05, + "loss": 0.3819, + "step": 47655 + }, + { + "epoch": 1.83, + "learning_rate": 1.9475506540138497e-05, + "loss": 0.3351, + "step": 47656 + }, + { + "epoch": 1.83, + "learning_rate": 1.947486535008977e-05, + "loss": 0.3542, + "step": 47657 + }, + { + "epoch": 1.83, + "learning_rate": 1.9474224160041037e-05, + "loss": 0.1964, + "step": 47658 + }, + { + "epoch": 1.83, + "learning_rate": 1.9473582969992306e-05, + "loss": 0.338, + "step": 47659 + }, + { + "epoch": 1.83, + "learning_rate": 1.9472941779943577e-05, + "loss": 0.3174, + "step": 47660 + }, + { + "epoch": 1.83, + "learning_rate": 1.9472300589894846e-05, + "loss": 0.5112, + "step": 47661 + }, + { + "epoch": 1.83, + "learning_rate": 1.9471659399846114e-05, + "loss": 0.3321, + "step": 47662 + }, + { + "epoch": 1.83, + "learning_rate": 1.9471018209797386e-05, + "loss": 0.2292, + "step": 47663 + }, + { + "epoch": 1.83, + "learning_rate": 1.9470377019748654e-05, + "loss": 0.4817, + "step": 47664 + }, + { + "epoch": 1.83, + "learning_rate": 1.9469735829699922e-05, + "loss": 0.2588, + "step": 47665 + }, + { + "epoch": 1.83, + "learning_rate": 1.9469094639651194e-05, + "loss": 0.4223, + "step": 47666 + }, + { + "epoch": 1.83, + "learning_rate": 1.9468453449602462e-05, + "loss": 0.3005, + "step": 47667 + }, + { + "epoch": 1.83, + "learning_rate": 1.9467812259553734e-05, + "loss": 0.2825, + "step": 47668 + }, + { + "epoch": 1.83, + "learning_rate": 1.9467171069505002e-05, + "loss": 0.2221, + "step": 47669 + }, + { + "epoch": 1.83, + "learning_rate": 1.946652987945627e-05, + "loss": 0.2628, + "step": 47670 + }, + { + "epoch": 1.83, + "learning_rate": 1.9465888689407542e-05, + "loss": 0.3435, + "step": 47671 + }, + { + "epoch": 1.83, + "learning_rate": 1.946524749935881e-05, + "loss": 0.226, + "step": 47672 + }, + { + "epoch": 1.83, + "learning_rate": 1.946460630931008e-05, + "loss": 0.16, + "step": 47673 + }, + { + "epoch": 1.83, + "learning_rate": 1.9463965119261347e-05, + "loss": 0.4582, + "step": 47674 + }, + { + "epoch": 1.83, + "learning_rate": 1.946332392921262e-05, + "loss": 0.2148, + "step": 47675 + }, + { + "epoch": 1.83, + "learning_rate": 1.946268273916389e-05, + "loss": 0.2701, + "step": 47676 + }, + { + "epoch": 1.83, + "learning_rate": 1.946204154911516e-05, + "loss": 0.2639, + "step": 47677 + }, + { + "epoch": 1.83, + "learning_rate": 1.9461400359066428e-05, + "loss": 0.3754, + "step": 47678 + }, + { + "epoch": 1.83, + "learning_rate": 1.94607591690177e-05, + "loss": 0.4636, + "step": 47679 + }, + { + "epoch": 1.83, + "learning_rate": 1.9460117978968968e-05, + "loss": 0.3899, + "step": 47680 + }, + { + "epoch": 1.83, + "learning_rate": 1.9459476788920236e-05, + "loss": 0.3921, + "step": 47681 + }, + { + "epoch": 1.83, + "learning_rate": 1.9458835598871504e-05, + "loss": 0.2752, + "step": 47682 + }, + { + "epoch": 1.83, + "learning_rate": 1.9458194408822776e-05, + "loss": 0.3284, + "step": 47683 + }, + { + "epoch": 1.83, + "learning_rate": 1.9457553218774048e-05, + "loss": 0.5435, + "step": 47684 + }, + { + "epoch": 1.83, + "learning_rate": 1.9456912028725316e-05, + "loss": 0.3059, + "step": 47685 + }, + { + "epoch": 1.83, + "learning_rate": 1.9456270838676584e-05, + "loss": 0.3284, + "step": 47686 + }, + { + "epoch": 1.83, + "learning_rate": 1.9455629648627856e-05, + "loss": 0.2466, + "step": 47687 + }, + { + "epoch": 1.83, + "learning_rate": 1.9454988458579124e-05, + "loss": 0.1776, + "step": 47688 + }, + { + "epoch": 1.83, + "learning_rate": 1.9454347268530393e-05, + "loss": 0.3534, + "step": 47689 + }, + { + "epoch": 1.83, + "learning_rate": 1.945370607848166e-05, + "loss": 0.3378, + "step": 47690 + }, + { + "epoch": 1.83, + "learning_rate": 1.9453064888432933e-05, + "loss": 0.4069, + "step": 47691 + }, + { + "epoch": 1.83, + "learning_rate": 1.94524236983842e-05, + "loss": 0.2666, + "step": 47692 + }, + { + "epoch": 1.83, + "learning_rate": 1.9451782508335473e-05, + "loss": 0.3641, + "step": 47693 + }, + { + "epoch": 1.83, + "learning_rate": 1.945114131828674e-05, + "loss": 0.2973, + "step": 47694 + }, + { + "epoch": 1.83, + "learning_rate": 1.945050012823801e-05, + "loss": 0.4411, + "step": 47695 + }, + { + "epoch": 1.83, + "learning_rate": 1.944985893818928e-05, + "loss": 0.2317, + "step": 47696 + }, + { + "epoch": 1.83, + "learning_rate": 1.944921774814055e-05, + "loss": 0.4554, + "step": 47697 + }, + { + "epoch": 1.83, + "learning_rate": 1.9448576558091818e-05, + "loss": 0.2938, + "step": 47698 + }, + { + "epoch": 1.83, + "learning_rate": 1.944793536804309e-05, + "loss": 0.2423, + "step": 47699 + }, + { + "epoch": 1.83, + "learning_rate": 1.9447294177994358e-05, + "loss": 0.326, + "step": 47700 + }, + { + "epoch": 1.83, + "learning_rate": 1.944665298794563e-05, + "loss": 0.3235, + "step": 47701 + }, + { + "epoch": 1.83, + "learning_rate": 1.9446011797896898e-05, + "loss": 0.3107, + "step": 47702 + }, + { + "epoch": 1.83, + "learning_rate": 1.9445370607848166e-05, + "loss": 0.2951, + "step": 47703 + }, + { + "epoch": 1.83, + "learning_rate": 1.9444729417799438e-05, + "loss": 0.2515, + "step": 47704 + }, + { + "epoch": 1.83, + "learning_rate": 1.9444088227750706e-05, + "loss": 0.3266, + "step": 47705 + }, + { + "epoch": 1.83, + "learning_rate": 1.9443447037701975e-05, + "loss": 0.3132, + "step": 47706 + }, + { + "epoch": 1.83, + "learning_rate": 1.9442805847653246e-05, + "loss": 0.3609, + "step": 47707 + }, + { + "epoch": 1.83, + "learning_rate": 1.9442164657604515e-05, + "loss": 0.3798, + "step": 47708 + }, + { + "epoch": 1.83, + "learning_rate": 1.9441523467555783e-05, + "loss": 0.3501, + "step": 47709 + }, + { + "epoch": 1.83, + "learning_rate": 1.9440882277507055e-05, + "loss": 0.3802, + "step": 47710 + }, + { + "epoch": 1.83, + "learning_rate": 1.9440241087458323e-05, + "loss": 0.4558, + "step": 47711 + }, + { + "epoch": 1.83, + "learning_rate": 1.9439599897409595e-05, + "loss": 0.3043, + "step": 47712 + }, + { + "epoch": 1.83, + "learning_rate": 1.9438958707360863e-05, + "loss": 0.3619, + "step": 47713 + }, + { + "epoch": 1.83, + "learning_rate": 1.943831751731213e-05, + "loss": 0.2378, + "step": 47714 + }, + { + "epoch": 1.83, + "learning_rate": 1.9437676327263403e-05, + "loss": 0.2739, + "step": 47715 + }, + { + "epoch": 1.83, + "learning_rate": 1.943703513721467e-05, + "loss": 0.2883, + "step": 47716 + }, + { + "epoch": 1.83, + "learning_rate": 1.943639394716594e-05, + "loss": 0.2709, + "step": 47717 + }, + { + "epoch": 1.83, + "learning_rate": 1.9435752757117208e-05, + "loss": 0.4548, + "step": 47718 + }, + { + "epoch": 1.83, + "learning_rate": 1.943511156706848e-05, + "loss": 0.3532, + "step": 47719 + }, + { + "epoch": 1.83, + "learning_rate": 1.943447037701975e-05, + "loss": 0.3808, + "step": 47720 + }, + { + "epoch": 1.83, + "learning_rate": 1.943382918697102e-05, + "loss": 0.3176, + "step": 47721 + }, + { + "epoch": 1.83, + "learning_rate": 1.9433187996922288e-05, + "loss": 0.3062, + "step": 47722 + }, + { + "epoch": 1.83, + "learning_rate": 1.943254680687356e-05, + "loss": 0.3425, + "step": 47723 + }, + { + "epoch": 1.83, + "learning_rate": 1.9431905616824828e-05, + "loss": 0.3701, + "step": 47724 + }, + { + "epoch": 1.83, + "learning_rate": 1.9431264426776097e-05, + "loss": 0.2677, + "step": 47725 + }, + { + "epoch": 1.83, + "learning_rate": 1.9430623236727365e-05, + "loss": 0.2424, + "step": 47726 + }, + { + "epoch": 1.83, + "learning_rate": 1.9429982046678637e-05, + "loss": 0.324, + "step": 47727 + }, + { + "epoch": 1.83, + "learning_rate": 1.942934085662991e-05, + "loss": 0.3696, + "step": 47728 + }, + { + "epoch": 1.84, + "learning_rate": 1.9428699666581177e-05, + "loss": 0.3479, + "step": 47729 + }, + { + "epoch": 1.84, + "learning_rate": 1.9428058476532445e-05, + "loss": 0.2509, + "step": 47730 + }, + { + "epoch": 1.84, + "learning_rate": 1.9427417286483717e-05, + "loss": 0.4493, + "step": 47731 + }, + { + "epoch": 1.84, + "learning_rate": 1.9426776096434985e-05, + "loss": 0.3006, + "step": 47732 + }, + { + "epoch": 1.84, + "learning_rate": 1.9426134906386253e-05, + "loss": 0.3859, + "step": 47733 + }, + { + "epoch": 1.84, + "learning_rate": 1.9425493716337522e-05, + "loss": 0.3151, + "step": 47734 + }, + { + "epoch": 1.84, + "learning_rate": 1.9424852526288793e-05, + "loss": 0.316, + "step": 47735 + }, + { + "epoch": 1.84, + "learning_rate": 1.9424211336240062e-05, + "loss": 0.3785, + "step": 47736 + }, + { + "epoch": 1.84, + "learning_rate": 1.9423570146191333e-05, + "loss": 0.3797, + "step": 47737 + }, + { + "epoch": 1.84, + "learning_rate": 1.9422928956142602e-05, + "loss": 0.349, + "step": 47738 + }, + { + "epoch": 1.84, + "learning_rate": 1.942228776609387e-05, + "loss": 0.3499, + "step": 47739 + }, + { + "epoch": 1.84, + "learning_rate": 1.9421646576045142e-05, + "loss": 0.2974, + "step": 47740 + }, + { + "epoch": 1.84, + "learning_rate": 1.942100538599641e-05, + "loss": 0.2258, + "step": 47741 + }, + { + "epoch": 1.84, + "learning_rate": 1.942036419594768e-05, + "loss": 0.4793, + "step": 47742 + }, + { + "epoch": 1.84, + "learning_rate": 1.941972300589895e-05, + "loss": 0.2969, + "step": 47743 + }, + { + "epoch": 1.84, + "learning_rate": 1.941908181585022e-05, + "loss": 0.3505, + "step": 47744 + }, + { + "epoch": 1.84, + "learning_rate": 1.9418440625801487e-05, + "loss": 0.3196, + "step": 47745 + }, + { + "epoch": 1.84, + "learning_rate": 1.941779943575276e-05, + "loss": 0.4015, + "step": 47746 + }, + { + "epoch": 1.84, + "learning_rate": 1.9417158245704027e-05, + "loss": 0.3857, + "step": 47747 + }, + { + "epoch": 1.84, + "learning_rate": 1.94165170556553e-05, + "loss": 0.3404, + "step": 47748 + }, + { + "epoch": 1.84, + "learning_rate": 1.9415875865606567e-05, + "loss": 0.3097, + "step": 47749 + }, + { + "epoch": 1.84, + "learning_rate": 1.9415234675557835e-05, + "loss": 0.2711, + "step": 47750 + }, + { + "epoch": 1.84, + "learning_rate": 1.9414593485509107e-05, + "loss": 0.2355, + "step": 47751 + }, + { + "epoch": 1.84, + "learning_rate": 1.9413952295460375e-05, + "loss": 0.4677, + "step": 47752 + }, + { + "epoch": 1.84, + "learning_rate": 1.9413311105411644e-05, + "loss": 0.4278, + "step": 47753 + }, + { + "epoch": 1.84, + "learning_rate": 1.9412669915362912e-05, + "loss": 0.341, + "step": 47754 + }, + { + "epoch": 1.84, + "learning_rate": 1.9412028725314184e-05, + "loss": 0.3976, + "step": 47755 + }, + { + "epoch": 1.84, + "learning_rate": 1.9411387535265455e-05, + "loss": 0.3019, + "step": 47756 + }, + { + "epoch": 1.84, + "learning_rate": 1.9410746345216724e-05, + "loss": 0.437, + "step": 47757 + }, + { + "epoch": 1.84, + "learning_rate": 1.9410105155167992e-05, + "loss": 0.3424, + "step": 47758 + }, + { + "epoch": 1.84, + "learning_rate": 1.9409463965119264e-05, + "loss": 0.5597, + "step": 47759 + }, + { + "epoch": 1.84, + "learning_rate": 1.9408822775070532e-05, + "loss": 0.2986, + "step": 47760 + }, + { + "epoch": 1.84, + "learning_rate": 1.94081815850218e-05, + "loss": 0.3343, + "step": 47761 + }, + { + "epoch": 1.84, + "learning_rate": 1.940754039497307e-05, + "loss": 0.2863, + "step": 47762 + }, + { + "epoch": 1.84, + "learning_rate": 1.940689920492434e-05, + "loss": 0.2972, + "step": 47763 + }, + { + "epoch": 1.84, + "learning_rate": 1.9406258014875612e-05, + "loss": 0.385, + "step": 47764 + }, + { + "epoch": 1.84, + "learning_rate": 1.940561682482688e-05, + "loss": 0.3196, + "step": 47765 + }, + { + "epoch": 1.84, + "learning_rate": 1.940497563477815e-05, + "loss": 0.3264, + "step": 47766 + }, + { + "epoch": 1.84, + "learning_rate": 1.940433444472942e-05, + "loss": 0.3655, + "step": 47767 + }, + { + "epoch": 1.84, + "learning_rate": 1.940369325468069e-05, + "loss": 0.3903, + "step": 47768 + }, + { + "epoch": 1.84, + "learning_rate": 1.9403052064631957e-05, + "loss": 0.455, + "step": 47769 + }, + { + "epoch": 1.84, + "learning_rate": 1.9402410874583226e-05, + "loss": 0.426, + "step": 47770 + }, + { + "epoch": 1.84, + "learning_rate": 1.9401769684534497e-05, + "loss": 0.204, + "step": 47771 + }, + { + "epoch": 1.84, + "learning_rate": 1.9401128494485766e-05, + "loss": 0.4153, + "step": 47772 + }, + { + "epoch": 1.84, + "learning_rate": 1.9400487304437037e-05, + "loss": 0.2804, + "step": 47773 + }, + { + "epoch": 1.84, + "learning_rate": 1.9399846114388306e-05, + "loss": 0.2797, + "step": 47774 + }, + { + "epoch": 1.84, + "learning_rate": 1.9399204924339574e-05, + "loss": 0.2725, + "step": 47775 + }, + { + "epoch": 1.84, + "learning_rate": 1.9398563734290846e-05, + "loss": 0.3242, + "step": 47776 + }, + { + "epoch": 1.84, + "learning_rate": 1.9397922544242114e-05, + "loss": 0.2607, + "step": 47777 + }, + { + "epoch": 1.84, + "learning_rate": 1.9397281354193382e-05, + "loss": 0.3047, + "step": 47778 + }, + { + "epoch": 1.84, + "learning_rate": 1.9396640164144654e-05, + "loss": 0.251, + "step": 47779 + }, + { + "epoch": 1.84, + "learning_rate": 1.9395998974095922e-05, + "loss": 0.314, + "step": 47780 + }, + { + "epoch": 1.84, + "learning_rate": 1.939535778404719e-05, + "loss": 0.4571, + "step": 47781 + }, + { + "epoch": 1.84, + "learning_rate": 1.9394716593998462e-05, + "loss": 0.2996, + "step": 47782 + }, + { + "epoch": 1.84, + "learning_rate": 1.939407540394973e-05, + "loss": 0.241, + "step": 47783 + }, + { + "epoch": 1.84, + "learning_rate": 1.9393434213901003e-05, + "loss": 0.2053, + "step": 47784 + }, + { + "epoch": 1.84, + "learning_rate": 1.939279302385227e-05, + "loss": 0.2902, + "step": 47785 + }, + { + "epoch": 1.84, + "learning_rate": 1.939215183380354e-05, + "loss": 0.362, + "step": 47786 + }, + { + "epoch": 1.84, + "learning_rate": 1.939151064375481e-05, + "loss": 0.1742, + "step": 47787 + }, + { + "epoch": 1.84, + "learning_rate": 1.939086945370608e-05, + "loss": 0.4254, + "step": 47788 + }, + { + "epoch": 1.84, + "learning_rate": 1.9390228263657348e-05, + "loss": 0.3401, + "step": 47789 + }, + { + "epoch": 1.84, + "learning_rate": 1.9389587073608616e-05, + "loss": 0.2792, + "step": 47790 + }, + { + "epoch": 1.84, + "learning_rate": 1.9388945883559888e-05, + "loss": 0.245, + "step": 47791 + }, + { + "epoch": 1.84, + "learning_rate": 1.938830469351116e-05, + "loss": 0.2761, + "step": 47792 + }, + { + "epoch": 1.84, + "learning_rate": 1.9387663503462428e-05, + "loss": 0.4355, + "step": 47793 + }, + { + "epoch": 1.84, + "learning_rate": 1.9387022313413696e-05, + "loss": 0.4513, + "step": 47794 + }, + { + "epoch": 1.84, + "learning_rate": 1.9386381123364968e-05, + "loss": 0.3738, + "step": 47795 + }, + { + "epoch": 1.84, + "learning_rate": 1.9385739933316236e-05, + "loss": 0.3263, + "step": 47796 + }, + { + "epoch": 1.84, + "learning_rate": 1.9385098743267504e-05, + "loss": 0.2886, + "step": 47797 + }, + { + "epoch": 1.84, + "learning_rate": 1.9384457553218773e-05, + "loss": 0.306, + "step": 47798 + }, + { + "epoch": 1.84, + "learning_rate": 1.9383816363170044e-05, + "loss": 0.4193, + "step": 47799 + }, + { + "epoch": 1.84, + "learning_rate": 1.9383175173121316e-05, + "loss": 0.2338, + "step": 47800 + }, + { + "epoch": 1.84, + "learning_rate": 1.9382533983072584e-05, + "loss": 0.3184, + "step": 47801 + }, + { + "epoch": 1.84, + "learning_rate": 1.9381892793023853e-05, + "loss": 0.4228, + "step": 47802 + }, + { + "epoch": 1.84, + "learning_rate": 1.9381251602975124e-05, + "loss": 0.3207, + "step": 47803 + }, + { + "epoch": 1.84, + "learning_rate": 1.9380610412926393e-05, + "loss": 0.2388, + "step": 47804 + }, + { + "epoch": 1.84, + "learning_rate": 1.937996922287766e-05, + "loss": 0.3654, + "step": 47805 + }, + { + "epoch": 1.84, + "learning_rate": 1.937932803282893e-05, + "loss": 0.2817, + "step": 47806 + }, + { + "epoch": 1.84, + "learning_rate": 1.93786868427802e-05, + "loss": 0.3292, + "step": 47807 + }, + { + "epoch": 1.84, + "learning_rate": 1.9378045652731473e-05, + "loss": 0.3385, + "step": 47808 + }, + { + "epoch": 1.84, + "learning_rate": 1.937740446268274e-05, + "loss": 0.3739, + "step": 47809 + }, + { + "epoch": 1.84, + "learning_rate": 1.937676327263401e-05, + "loss": 0.316, + "step": 47810 + }, + { + "epoch": 1.84, + "learning_rate": 1.937612208258528e-05, + "loss": 0.2911, + "step": 47811 + }, + { + "epoch": 1.84, + "learning_rate": 1.937548089253655e-05, + "loss": 0.3727, + "step": 47812 + }, + { + "epoch": 1.84, + "learning_rate": 1.9374839702487818e-05, + "loss": 0.2883, + "step": 47813 + }, + { + "epoch": 1.84, + "learning_rate": 1.9374198512439086e-05, + "loss": 0.3436, + "step": 47814 + }, + { + "epoch": 1.84, + "learning_rate": 1.9373557322390358e-05, + "loss": 0.4428, + "step": 47815 + }, + { + "epoch": 1.84, + "learning_rate": 1.9372916132341626e-05, + "loss": 0.3439, + "step": 47816 + }, + { + "epoch": 1.84, + "learning_rate": 1.9372274942292898e-05, + "loss": 0.225, + "step": 47817 + }, + { + "epoch": 1.84, + "learning_rate": 1.9371633752244166e-05, + "loss": 0.2974, + "step": 47818 + }, + { + "epoch": 1.84, + "learning_rate": 1.9370992562195435e-05, + "loss": 0.3157, + "step": 47819 + }, + { + "epoch": 1.84, + "learning_rate": 1.9370351372146706e-05, + "loss": 0.2394, + "step": 47820 + }, + { + "epoch": 1.84, + "learning_rate": 1.9369710182097975e-05, + "loss": 0.3257, + "step": 47821 + }, + { + "epoch": 1.84, + "learning_rate": 1.9369068992049243e-05, + "loss": 0.3459, + "step": 47822 + }, + { + "epoch": 1.84, + "learning_rate": 1.9368427802000515e-05, + "loss": 0.4849, + "step": 47823 + }, + { + "epoch": 1.84, + "learning_rate": 1.9367786611951783e-05, + "loss": 0.3254, + "step": 47824 + }, + { + "epoch": 1.84, + "learning_rate": 1.936714542190305e-05, + "loss": 0.2924, + "step": 47825 + }, + { + "epoch": 1.84, + "learning_rate": 1.9366504231854323e-05, + "loss": 0.2351, + "step": 47826 + }, + { + "epoch": 1.84, + "learning_rate": 1.936586304180559e-05, + "loss": 0.3084, + "step": 47827 + }, + { + "epoch": 1.84, + "learning_rate": 1.9365221851756863e-05, + "loss": 0.2942, + "step": 47828 + }, + { + "epoch": 1.84, + "learning_rate": 1.936458066170813e-05, + "loss": 0.3703, + "step": 47829 + }, + { + "epoch": 1.84, + "learning_rate": 1.93639394716594e-05, + "loss": 0.2705, + "step": 47830 + }, + { + "epoch": 1.84, + "learning_rate": 1.936329828161067e-05, + "loss": 0.3721, + "step": 47831 + }, + { + "epoch": 1.84, + "learning_rate": 1.936265709156194e-05, + "loss": 0.3751, + "step": 47832 + }, + { + "epoch": 1.84, + "learning_rate": 1.9362015901513208e-05, + "loss": 0.2981, + "step": 47833 + }, + { + "epoch": 1.84, + "learning_rate": 1.9361374711464477e-05, + "loss": 0.3215, + "step": 47834 + }, + { + "epoch": 1.84, + "learning_rate": 1.9360733521415748e-05, + "loss": 0.2529, + "step": 47835 + }, + { + "epoch": 1.84, + "learning_rate": 1.936009233136702e-05, + "loss": 0.3669, + "step": 47836 + }, + { + "epoch": 1.84, + "learning_rate": 1.9359451141318288e-05, + "loss": 0.2871, + "step": 47837 + }, + { + "epoch": 1.84, + "learning_rate": 1.9358809951269557e-05, + "loss": 0.2344, + "step": 47838 + }, + { + "epoch": 1.84, + "learning_rate": 1.935816876122083e-05, + "loss": 0.3104, + "step": 47839 + }, + { + "epoch": 1.84, + "learning_rate": 1.9357527571172097e-05, + "loss": 0.37, + "step": 47840 + }, + { + "epoch": 1.84, + "learning_rate": 1.9356886381123365e-05, + "loss": 0.3904, + "step": 47841 + }, + { + "epoch": 1.84, + "learning_rate": 1.9356245191074633e-05, + "loss": 0.3305, + "step": 47842 + }, + { + "epoch": 1.84, + "learning_rate": 1.9355604001025905e-05, + "loss": 0.4105, + "step": 47843 + }, + { + "epoch": 1.84, + "learning_rate": 1.9354962810977177e-05, + "loss": 0.4566, + "step": 47844 + }, + { + "epoch": 1.84, + "learning_rate": 1.9354321620928445e-05, + "loss": 0.4164, + "step": 47845 + }, + { + "epoch": 1.84, + "learning_rate": 1.9353680430879713e-05, + "loss": 0.2341, + "step": 47846 + }, + { + "epoch": 1.84, + "learning_rate": 1.9353039240830985e-05, + "loss": 0.216, + "step": 47847 + }, + { + "epoch": 1.84, + "learning_rate": 1.9352398050782253e-05, + "loss": 0.2185, + "step": 47848 + }, + { + "epoch": 1.84, + "learning_rate": 1.9351756860733522e-05, + "loss": 0.1981, + "step": 47849 + }, + { + "epoch": 1.84, + "learning_rate": 1.935111567068479e-05, + "loss": 0.2952, + "step": 47850 + }, + { + "epoch": 1.84, + "learning_rate": 1.9350474480636062e-05, + "loss": 0.34, + "step": 47851 + }, + { + "epoch": 1.84, + "learning_rate": 1.934983329058733e-05, + "loss": 0.3254, + "step": 47852 + }, + { + "epoch": 1.84, + "learning_rate": 1.9349192100538602e-05, + "loss": 0.3806, + "step": 47853 + }, + { + "epoch": 1.84, + "learning_rate": 1.934855091048987e-05, + "loss": 0.3116, + "step": 47854 + }, + { + "epoch": 1.84, + "learning_rate": 1.9347909720441142e-05, + "loss": 0.3008, + "step": 47855 + }, + { + "epoch": 1.84, + "learning_rate": 1.934726853039241e-05, + "loss": 0.429, + "step": 47856 + }, + { + "epoch": 1.84, + "learning_rate": 1.934662734034368e-05, + "loss": 0.447, + "step": 47857 + }, + { + "epoch": 1.84, + "learning_rate": 1.9345986150294947e-05, + "loss": 0.4453, + "step": 47858 + }, + { + "epoch": 1.84, + "learning_rate": 1.934534496024622e-05, + "loss": 0.4307, + "step": 47859 + }, + { + "epoch": 1.84, + "learning_rate": 1.9344703770197487e-05, + "loss": 0.397, + "step": 47860 + }, + { + "epoch": 1.84, + "learning_rate": 1.9344062580148755e-05, + "loss": 0.3523, + "step": 47861 + }, + { + "epoch": 1.84, + "learning_rate": 1.9343421390100027e-05, + "loss": 0.351, + "step": 47862 + }, + { + "epoch": 1.84, + "learning_rate": 1.9342780200051295e-05, + "loss": 0.2919, + "step": 47863 + }, + { + "epoch": 1.84, + "learning_rate": 1.9342139010002567e-05, + "loss": 0.4045, + "step": 47864 + }, + { + "epoch": 1.84, + "learning_rate": 1.9341497819953835e-05, + "loss": 0.3735, + "step": 47865 + }, + { + "epoch": 1.84, + "learning_rate": 1.9340856629905104e-05, + "loss": 0.4462, + "step": 47866 + }, + { + "epoch": 1.84, + "learning_rate": 1.9340215439856375e-05, + "loss": 0.4233, + "step": 47867 + }, + { + "epoch": 1.84, + "learning_rate": 1.9339574249807644e-05, + "loss": 0.204, + "step": 47868 + }, + { + "epoch": 1.84, + "learning_rate": 1.9338933059758912e-05, + "loss": 0.3152, + "step": 47869 + }, + { + "epoch": 1.84, + "learning_rate": 1.933829186971018e-05, + "loss": 0.2231, + "step": 47870 + }, + { + "epoch": 1.84, + "learning_rate": 1.9337650679661452e-05, + "loss": 0.4741, + "step": 47871 + }, + { + "epoch": 1.84, + "learning_rate": 1.9337009489612724e-05, + "loss": 0.4625, + "step": 47872 + }, + { + "epoch": 1.84, + "learning_rate": 1.9336368299563992e-05, + "loss": 0.4302, + "step": 47873 + }, + { + "epoch": 1.84, + "learning_rate": 1.933572710951526e-05, + "loss": 0.3421, + "step": 47874 + }, + { + "epoch": 1.84, + "learning_rate": 1.9335085919466532e-05, + "loss": 0.2488, + "step": 47875 + }, + { + "epoch": 1.84, + "learning_rate": 1.93344447294178e-05, + "loss": 0.3646, + "step": 47876 + }, + { + "epoch": 1.84, + "learning_rate": 1.933380353936907e-05, + "loss": 0.3026, + "step": 47877 + }, + { + "epoch": 1.84, + "learning_rate": 1.9333162349320337e-05, + "loss": 0.3179, + "step": 47878 + }, + { + "epoch": 1.84, + "learning_rate": 1.933252115927161e-05, + "loss": 0.3017, + "step": 47879 + }, + { + "epoch": 1.84, + "learning_rate": 1.933187996922288e-05, + "loss": 0.4936, + "step": 47880 + }, + { + "epoch": 1.84, + "learning_rate": 1.933123877917415e-05, + "loss": 0.3914, + "step": 47881 + }, + { + "epoch": 1.84, + "learning_rate": 1.9330597589125417e-05, + "loss": 0.3644, + "step": 47882 + }, + { + "epoch": 1.84, + "learning_rate": 1.932995639907669e-05, + "loss": 0.391, + "step": 47883 + }, + { + "epoch": 1.84, + "learning_rate": 1.9329315209027957e-05, + "loss": 0.3487, + "step": 47884 + }, + { + "epoch": 1.84, + "learning_rate": 1.9328674018979226e-05, + "loss": 0.4539, + "step": 47885 + }, + { + "epoch": 1.84, + "learning_rate": 1.9328032828930494e-05, + "loss": 0.3553, + "step": 47886 + }, + { + "epoch": 1.84, + "learning_rate": 1.9327391638881766e-05, + "loss": 0.2966, + "step": 47887 + }, + { + "epoch": 1.84, + "learning_rate": 1.9326750448833034e-05, + "loss": 0.2432, + "step": 47888 + }, + { + "epoch": 1.84, + "learning_rate": 1.9326109258784306e-05, + "loss": 0.2747, + "step": 47889 + }, + { + "epoch": 1.84, + "learning_rate": 1.9325468068735574e-05, + "loss": 0.429, + "step": 47890 + }, + { + "epoch": 1.84, + "learning_rate": 1.9324826878686846e-05, + "loss": 0.3691, + "step": 47891 + }, + { + "epoch": 1.84, + "learning_rate": 1.9324185688638114e-05, + "loss": 0.2779, + "step": 47892 + }, + { + "epoch": 1.84, + "learning_rate": 1.9323544498589382e-05, + "loss": 0.3072, + "step": 47893 + }, + { + "epoch": 1.84, + "learning_rate": 1.932290330854065e-05, + "loss": 0.3616, + "step": 47894 + }, + { + "epoch": 1.84, + "learning_rate": 1.9322262118491923e-05, + "loss": 0.293, + "step": 47895 + }, + { + "epoch": 1.84, + "learning_rate": 1.932162092844319e-05, + "loss": 0.4151, + "step": 47896 + }, + { + "epoch": 1.84, + "learning_rate": 1.9320979738394463e-05, + "loss": 0.2824, + "step": 47897 + }, + { + "epoch": 1.84, + "learning_rate": 1.932033854834573e-05, + "loss": 0.3121, + "step": 47898 + }, + { + "epoch": 1.84, + "learning_rate": 1.9319697358297e-05, + "loss": 0.4244, + "step": 47899 + }, + { + "epoch": 1.84, + "learning_rate": 1.931905616824827e-05, + "loss": 0.2796, + "step": 47900 + }, + { + "epoch": 1.84, + "learning_rate": 1.931841497819954e-05, + "loss": 0.342, + "step": 47901 + }, + { + "epoch": 1.84, + "learning_rate": 1.9317773788150808e-05, + "loss": 0.421, + "step": 47902 + }, + { + "epoch": 1.84, + "learning_rate": 1.931713259810208e-05, + "loss": 0.2671, + "step": 47903 + }, + { + "epoch": 1.84, + "learning_rate": 1.9316491408053348e-05, + "loss": 0.3453, + "step": 47904 + }, + { + "epoch": 1.84, + "learning_rate": 1.9315850218004616e-05, + "loss": 0.4122, + "step": 47905 + }, + { + "epoch": 1.84, + "learning_rate": 1.9315209027955888e-05, + "loss": 0.4005, + "step": 47906 + }, + { + "epoch": 1.84, + "learning_rate": 1.9314567837907156e-05, + "loss": 0.2393, + "step": 47907 + }, + { + "epoch": 1.84, + "learning_rate": 1.9313926647858428e-05, + "loss": 0.3205, + "step": 47908 + }, + { + "epoch": 1.84, + "learning_rate": 1.9313285457809696e-05, + "loss": 0.2512, + "step": 47909 + }, + { + "epoch": 1.84, + "learning_rate": 1.9312644267760964e-05, + "loss": 0.2919, + "step": 47910 + }, + { + "epoch": 1.84, + "learning_rate": 1.9312003077712236e-05, + "loss": 0.3413, + "step": 47911 + }, + { + "epoch": 1.84, + "learning_rate": 1.9311361887663504e-05, + "loss": 0.3092, + "step": 47912 + }, + { + "epoch": 1.84, + "learning_rate": 1.9310720697614773e-05, + "loss": 0.5166, + "step": 47913 + }, + { + "epoch": 1.84, + "learning_rate": 1.931007950756604e-05, + "loss": 0.4318, + "step": 47914 + }, + { + "epoch": 1.84, + "learning_rate": 1.9309438317517313e-05, + "loss": 0.4048, + "step": 47915 + }, + { + "epoch": 1.84, + "learning_rate": 1.9308797127468585e-05, + "loss": 0.3481, + "step": 47916 + }, + { + "epoch": 1.84, + "learning_rate": 1.9308155937419853e-05, + "loss": 0.2939, + "step": 47917 + }, + { + "epoch": 1.84, + "learning_rate": 1.930751474737112e-05, + "loss": 0.3096, + "step": 47918 + }, + { + "epoch": 1.84, + "learning_rate": 1.9306873557322393e-05, + "loss": 0.3323, + "step": 47919 + }, + { + "epoch": 1.84, + "learning_rate": 1.930623236727366e-05, + "loss": 0.3091, + "step": 47920 + }, + { + "epoch": 1.84, + "learning_rate": 1.930559117722493e-05, + "loss": 0.3988, + "step": 47921 + }, + { + "epoch": 1.84, + "learning_rate": 1.9304949987176198e-05, + "loss": 0.3149, + "step": 47922 + }, + { + "epoch": 1.84, + "learning_rate": 1.930430879712747e-05, + "loss": 0.3357, + "step": 47923 + }, + { + "epoch": 1.84, + "learning_rate": 1.930366760707874e-05, + "loss": 0.2175, + "step": 47924 + }, + { + "epoch": 1.84, + "learning_rate": 1.930302641703001e-05, + "loss": 0.3127, + "step": 47925 + }, + { + "epoch": 1.84, + "learning_rate": 1.9302385226981278e-05, + "loss": 0.2248, + "step": 47926 + }, + { + "epoch": 1.84, + "learning_rate": 1.930174403693255e-05, + "loss": 0.4141, + "step": 47927 + }, + { + "epoch": 1.84, + "learning_rate": 1.9301102846883818e-05, + "loss": 0.3209, + "step": 47928 + }, + { + "epoch": 1.84, + "learning_rate": 1.9300461656835086e-05, + "loss": 0.3723, + "step": 47929 + }, + { + "epoch": 1.84, + "learning_rate": 1.9299820466786355e-05, + "loss": 0.2681, + "step": 47930 + }, + { + "epoch": 1.84, + "learning_rate": 1.9299179276737626e-05, + "loss": 0.3452, + "step": 47931 + }, + { + "epoch": 1.84, + "learning_rate": 1.9298538086688895e-05, + "loss": 0.3919, + "step": 47932 + }, + { + "epoch": 1.84, + "learning_rate": 1.9297896896640166e-05, + "loss": 0.2694, + "step": 47933 + }, + { + "epoch": 1.84, + "learning_rate": 1.9297255706591435e-05, + "loss": 0.1837, + "step": 47934 + }, + { + "epoch": 1.84, + "learning_rate": 1.9296614516542706e-05, + "loss": 0.365, + "step": 47935 + }, + { + "epoch": 1.84, + "learning_rate": 1.9295973326493975e-05, + "loss": 0.259, + "step": 47936 + }, + { + "epoch": 1.84, + "learning_rate": 1.9295332136445243e-05, + "loss": 0.4069, + "step": 47937 + }, + { + "epoch": 1.84, + "learning_rate": 1.929469094639651e-05, + "loss": 0.2729, + "step": 47938 + }, + { + "epoch": 1.84, + "learning_rate": 1.9294049756347783e-05, + "loss": 0.3448, + "step": 47939 + }, + { + "epoch": 1.84, + "learning_rate": 1.929340856629905e-05, + "loss": 0.525, + "step": 47940 + }, + { + "epoch": 1.84, + "learning_rate": 1.929276737625032e-05, + "loss": 0.2189, + "step": 47941 + }, + { + "epoch": 1.84, + "learning_rate": 1.929212618620159e-05, + "loss": 0.3337, + "step": 47942 + }, + { + "epoch": 1.84, + "learning_rate": 1.929148499615286e-05, + "loss": 0.495, + "step": 47943 + }, + { + "epoch": 1.84, + "learning_rate": 1.929084380610413e-05, + "loss": 0.3435, + "step": 47944 + }, + { + "epoch": 1.84, + "learning_rate": 1.92902026160554e-05, + "loss": 0.3019, + "step": 47945 + }, + { + "epoch": 1.84, + "learning_rate": 1.9289561426006668e-05, + "loss": 0.3152, + "step": 47946 + }, + { + "epoch": 1.84, + "learning_rate": 1.928892023595794e-05, + "loss": 0.3355, + "step": 47947 + }, + { + "epoch": 1.84, + "learning_rate": 1.9288279045909208e-05, + "loss": 0.3251, + "step": 47948 + }, + { + "epoch": 1.84, + "learning_rate": 1.9287637855860477e-05, + "loss": 0.311, + "step": 47949 + }, + { + "epoch": 1.84, + "learning_rate": 1.9286996665811745e-05, + "loss": 0.1902, + "step": 47950 + }, + { + "epoch": 1.84, + "learning_rate": 1.9286355475763017e-05, + "loss": 0.3725, + "step": 47951 + }, + { + "epoch": 1.84, + "learning_rate": 1.928571428571429e-05, + "loss": 0.4288, + "step": 47952 + }, + { + "epoch": 1.84, + "learning_rate": 1.9285073095665557e-05, + "loss": 0.3559, + "step": 47953 + }, + { + "epoch": 1.84, + "learning_rate": 1.9284431905616825e-05, + "loss": 0.2654, + "step": 47954 + }, + { + "epoch": 1.84, + "learning_rate": 1.9283790715568097e-05, + "loss": 0.3621, + "step": 47955 + }, + { + "epoch": 1.84, + "learning_rate": 1.9283149525519365e-05, + "loss": 0.2778, + "step": 47956 + }, + { + "epoch": 1.84, + "learning_rate": 1.9282508335470633e-05, + "loss": 0.4095, + "step": 47957 + }, + { + "epoch": 1.84, + "learning_rate": 1.9281867145421902e-05, + "loss": 0.4772, + "step": 47958 + }, + { + "epoch": 1.84, + "learning_rate": 1.9281225955373173e-05, + "loss": 0.265, + "step": 47959 + }, + { + "epoch": 1.84, + "learning_rate": 1.9280584765324445e-05, + "loss": 0.3187, + "step": 47960 + }, + { + "epoch": 1.84, + "learning_rate": 1.9279943575275714e-05, + "loss": 0.4613, + "step": 47961 + }, + { + "epoch": 1.84, + "learning_rate": 1.9279302385226982e-05, + "loss": 0.4052, + "step": 47962 + }, + { + "epoch": 1.84, + "learning_rate": 1.9278661195178254e-05, + "loss": 0.249, + "step": 47963 + }, + { + "epoch": 1.84, + "learning_rate": 1.9278020005129522e-05, + "loss": 0.3471, + "step": 47964 + }, + { + "epoch": 1.84, + "learning_rate": 1.927737881508079e-05, + "loss": 0.3521, + "step": 47965 + }, + { + "epoch": 1.84, + "learning_rate": 1.927673762503206e-05, + "loss": 0.4381, + "step": 47966 + }, + { + "epoch": 1.84, + "learning_rate": 1.927609643498333e-05, + "loss": 0.4506, + "step": 47967 + }, + { + "epoch": 1.84, + "learning_rate": 1.92754552449346e-05, + "loss": 0.3837, + "step": 47968 + }, + { + "epoch": 1.84, + "learning_rate": 1.927481405488587e-05, + "loss": 0.3003, + "step": 47969 + }, + { + "epoch": 1.84, + "learning_rate": 1.927417286483714e-05, + "loss": 0.2864, + "step": 47970 + }, + { + "epoch": 1.84, + "learning_rate": 1.927353167478841e-05, + "loss": 0.3023, + "step": 47971 + }, + { + "epoch": 1.84, + "learning_rate": 1.927289048473968e-05, + "loss": 0.3487, + "step": 47972 + }, + { + "epoch": 1.84, + "learning_rate": 1.9272249294690947e-05, + "loss": 0.2713, + "step": 47973 + }, + { + "epoch": 1.84, + "learning_rate": 1.9271608104642215e-05, + "loss": 0.3934, + "step": 47974 + }, + { + "epoch": 1.84, + "learning_rate": 1.9270966914593487e-05, + "loss": 0.2806, + "step": 47975 + }, + { + "epoch": 1.84, + "learning_rate": 1.9270325724544755e-05, + "loss": 0.2456, + "step": 47976 + }, + { + "epoch": 1.84, + "learning_rate": 1.9269684534496024e-05, + "loss": 0.2512, + "step": 47977 + }, + { + "epoch": 1.84, + "learning_rate": 1.9269043344447295e-05, + "loss": 0.2572, + "step": 47978 + }, + { + "epoch": 1.84, + "learning_rate": 1.9268402154398567e-05, + "loss": 0.3281, + "step": 47979 + }, + { + "epoch": 1.84, + "learning_rate": 1.9267760964349835e-05, + "loss": 0.2786, + "step": 47980 + }, + { + "epoch": 1.84, + "learning_rate": 1.9267119774301104e-05, + "loss": 0.2957, + "step": 47981 + }, + { + "epoch": 1.84, + "learning_rate": 1.9266478584252372e-05, + "loss": 0.3041, + "step": 47982 + }, + { + "epoch": 1.84, + "learning_rate": 1.9265837394203644e-05, + "loss": 0.3751, + "step": 47983 + }, + { + "epoch": 1.84, + "learning_rate": 1.9265196204154912e-05, + "loss": 0.2864, + "step": 47984 + }, + { + "epoch": 1.84, + "learning_rate": 1.926455501410618e-05, + "loss": 0.2822, + "step": 47985 + }, + { + "epoch": 1.84, + "learning_rate": 1.9263913824057452e-05, + "loss": 0.3151, + "step": 47986 + }, + { + "epoch": 1.84, + "learning_rate": 1.926327263400872e-05, + "loss": 0.2907, + "step": 47987 + }, + { + "epoch": 1.84, + "learning_rate": 1.9262631443959992e-05, + "loss": 0.3783, + "step": 47988 + }, + { + "epoch": 1.85, + "learning_rate": 1.926199025391126e-05, + "loss": 0.2755, + "step": 47989 + }, + { + "epoch": 1.85, + "learning_rate": 1.926134906386253e-05, + "loss": 0.4776, + "step": 47990 + }, + { + "epoch": 1.85, + "learning_rate": 1.92607078738138e-05, + "loss": 0.2572, + "step": 47991 + }, + { + "epoch": 1.85, + "learning_rate": 1.926006668376507e-05, + "loss": 0.4735, + "step": 47992 + }, + { + "epoch": 1.85, + "learning_rate": 1.9259425493716337e-05, + "loss": 0.277, + "step": 47993 + }, + { + "epoch": 1.85, + "learning_rate": 1.9258784303667606e-05, + "loss": 0.501, + "step": 47994 + }, + { + "epoch": 1.85, + "learning_rate": 1.9258143113618877e-05, + "loss": 0.2535, + "step": 47995 + }, + { + "epoch": 1.85, + "learning_rate": 1.925750192357015e-05, + "loss": 0.4846, + "step": 47996 + }, + { + "epoch": 1.85, + "learning_rate": 1.9256860733521417e-05, + "loss": 0.3087, + "step": 47997 + }, + { + "epoch": 1.85, + "learning_rate": 1.9256219543472686e-05, + "loss": 0.3564, + "step": 47998 + }, + { + "epoch": 1.85, + "learning_rate": 1.9255578353423957e-05, + "loss": 0.2956, + "step": 47999 + }, + { + "epoch": 1.85, + "learning_rate": 1.9254937163375226e-05, + "loss": 0.2843, + "step": 48000 + }, + { + "epoch": 1.85, + "learning_rate": 1.9254295973326494e-05, + "loss": 0.2083, + "step": 48001 + }, + { + "epoch": 1.85, + "learning_rate": 1.9253654783277762e-05, + "loss": 0.4029, + "step": 48002 + }, + { + "epoch": 1.85, + "learning_rate": 1.9253013593229034e-05, + "loss": 0.2746, + "step": 48003 + }, + { + "epoch": 1.85, + "learning_rate": 1.9252372403180306e-05, + "loss": 0.2828, + "step": 48004 + }, + { + "epoch": 1.85, + "learning_rate": 1.9251731213131574e-05, + "loss": 0.3051, + "step": 48005 + }, + { + "epoch": 1.85, + "learning_rate": 1.9251090023082842e-05, + "loss": 0.2622, + "step": 48006 + }, + { + "epoch": 1.85, + "learning_rate": 1.9250448833034114e-05, + "loss": 0.4173, + "step": 48007 + }, + { + "epoch": 1.85, + "learning_rate": 1.9249807642985383e-05, + "loss": 0.4324, + "step": 48008 + }, + { + "epoch": 1.85, + "learning_rate": 1.924916645293665e-05, + "loss": 0.3996, + "step": 48009 + }, + { + "epoch": 1.85, + "learning_rate": 1.924852526288792e-05, + "loss": 0.3372, + "step": 48010 + }, + { + "epoch": 1.85, + "learning_rate": 1.924788407283919e-05, + "loss": 0.3432, + "step": 48011 + }, + { + "epoch": 1.85, + "learning_rate": 1.924724288279046e-05, + "loss": 0.3485, + "step": 48012 + }, + { + "epoch": 1.85, + "learning_rate": 1.924660169274173e-05, + "loss": 0.226, + "step": 48013 + }, + { + "epoch": 1.85, + "learning_rate": 1.9245960502693e-05, + "loss": 0.2662, + "step": 48014 + }, + { + "epoch": 1.85, + "learning_rate": 1.924531931264427e-05, + "loss": 0.2984, + "step": 48015 + }, + { + "epoch": 1.85, + "learning_rate": 1.924467812259554e-05, + "loss": 0.3672, + "step": 48016 + }, + { + "epoch": 1.85, + "learning_rate": 1.9244036932546808e-05, + "loss": 0.3831, + "step": 48017 + }, + { + "epoch": 1.85, + "learning_rate": 1.9243395742498076e-05, + "loss": 0.2146, + "step": 48018 + }, + { + "epoch": 1.85, + "learning_rate": 1.9242754552449348e-05, + "loss": 0.349, + "step": 48019 + }, + { + "epoch": 1.85, + "learning_rate": 1.9242113362400616e-05, + "loss": 0.3579, + "step": 48020 + }, + { + "epoch": 1.85, + "learning_rate": 1.9241472172351884e-05, + "loss": 0.2846, + "step": 48021 + }, + { + "epoch": 1.85, + "learning_rate": 1.9240830982303156e-05, + "loss": 0.2731, + "step": 48022 + }, + { + "epoch": 1.85, + "learning_rate": 1.9240189792254424e-05, + "loss": 0.2089, + "step": 48023 + }, + { + "epoch": 1.85, + "learning_rate": 1.9239548602205696e-05, + "loss": 0.3508, + "step": 48024 + }, + { + "epoch": 1.85, + "learning_rate": 1.9238907412156964e-05, + "loss": 0.2482, + "step": 48025 + }, + { + "epoch": 1.85, + "learning_rate": 1.9238266222108233e-05, + "loss": 0.4325, + "step": 48026 + }, + { + "epoch": 1.85, + "learning_rate": 1.9237625032059504e-05, + "loss": 0.2237, + "step": 48027 + }, + { + "epoch": 1.85, + "learning_rate": 1.9236983842010773e-05, + "loss": 0.4597, + "step": 48028 + }, + { + "epoch": 1.85, + "learning_rate": 1.923634265196204e-05, + "loss": 0.3667, + "step": 48029 + }, + { + "epoch": 1.85, + "learning_rate": 1.923570146191331e-05, + "loss": 0.3286, + "step": 48030 + }, + { + "epoch": 1.85, + "learning_rate": 1.923506027186458e-05, + "loss": 0.2295, + "step": 48031 + }, + { + "epoch": 1.85, + "learning_rate": 1.9234419081815853e-05, + "loss": 0.4438, + "step": 48032 + }, + { + "epoch": 1.85, + "learning_rate": 1.923377789176712e-05, + "loss": 0.3924, + "step": 48033 + }, + { + "epoch": 1.85, + "learning_rate": 1.923313670171839e-05, + "loss": 0.4116, + "step": 48034 + }, + { + "epoch": 1.85, + "learning_rate": 1.923249551166966e-05, + "loss": 0.2811, + "step": 48035 + }, + { + "epoch": 1.85, + "learning_rate": 1.923185432162093e-05, + "loss": 0.404, + "step": 48036 + }, + { + "epoch": 1.85, + "learning_rate": 1.9231213131572198e-05, + "loss": 0.2157, + "step": 48037 + }, + { + "epoch": 1.85, + "learning_rate": 1.9230571941523466e-05, + "loss": 0.4097, + "step": 48038 + }, + { + "epoch": 1.85, + "learning_rate": 1.9229930751474738e-05, + "loss": 0.2391, + "step": 48039 + }, + { + "epoch": 1.85, + "learning_rate": 1.922928956142601e-05, + "loss": 0.2574, + "step": 48040 + }, + { + "epoch": 1.85, + "learning_rate": 1.9228648371377278e-05, + "loss": 0.3034, + "step": 48041 + }, + { + "epoch": 1.85, + "learning_rate": 1.9228007181328546e-05, + "loss": 0.2389, + "step": 48042 + }, + { + "epoch": 1.85, + "learning_rate": 1.9227365991279818e-05, + "loss": 0.2869, + "step": 48043 + }, + { + "epoch": 1.85, + "learning_rate": 1.9226724801231086e-05, + "loss": 0.2558, + "step": 48044 + }, + { + "epoch": 1.85, + "learning_rate": 1.9226083611182355e-05, + "loss": 0.4171, + "step": 48045 + }, + { + "epoch": 1.85, + "learning_rate": 1.9225442421133623e-05, + "loss": 0.415, + "step": 48046 + }, + { + "epoch": 1.85, + "learning_rate": 1.9224801231084895e-05, + "loss": 0.3066, + "step": 48047 + }, + { + "epoch": 1.85, + "learning_rate": 1.9224160041036163e-05, + "loss": 0.3865, + "step": 48048 + }, + { + "epoch": 1.85, + "learning_rate": 1.9223518850987435e-05, + "loss": 0.2407, + "step": 48049 + }, + { + "epoch": 1.85, + "learning_rate": 1.9222877660938703e-05, + "loss": 0.4606, + "step": 48050 + }, + { + "epoch": 1.85, + "learning_rate": 1.9222236470889975e-05, + "loss": 0.3439, + "step": 48051 + }, + { + "epoch": 1.85, + "learning_rate": 1.9221595280841243e-05, + "loss": 0.3487, + "step": 48052 + }, + { + "epoch": 1.85, + "learning_rate": 1.922095409079251e-05, + "loss": 0.2037, + "step": 48053 + }, + { + "epoch": 1.85, + "learning_rate": 1.922031290074378e-05, + "loss": 0.2752, + "step": 48054 + }, + { + "epoch": 1.85, + "learning_rate": 1.921967171069505e-05, + "loss": 0.3073, + "step": 48055 + }, + { + "epoch": 1.85, + "learning_rate": 1.921903052064632e-05, + "loss": 0.2411, + "step": 48056 + }, + { + "epoch": 1.85, + "learning_rate": 1.9218389330597588e-05, + "loss": 0.2016, + "step": 48057 + }, + { + "epoch": 1.85, + "learning_rate": 1.921774814054886e-05, + "loss": 0.3499, + "step": 48058 + }, + { + "epoch": 1.85, + "learning_rate": 1.921710695050013e-05, + "loss": 0.2545, + "step": 48059 + }, + { + "epoch": 1.85, + "learning_rate": 1.92164657604514e-05, + "loss": 0.2949, + "step": 48060 + }, + { + "epoch": 1.85, + "learning_rate": 1.921582457040267e-05, + "loss": 0.3327, + "step": 48061 + }, + { + "epoch": 1.85, + "learning_rate": 1.9215183380353937e-05, + "loss": 0.1693, + "step": 48062 + }, + { + "epoch": 1.85, + "learning_rate": 1.921454219030521e-05, + "loss": 0.2763, + "step": 48063 + }, + { + "epoch": 1.85, + "learning_rate": 1.9213901000256477e-05, + "loss": 0.3039, + "step": 48064 + }, + { + "epoch": 1.85, + "learning_rate": 1.9213259810207745e-05, + "loss": 0.3362, + "step": 48065 + }, + { + "epoch": 1.85, + "learning_rate": 1.9212618620159013e-05, + "loss": 0.3241, + "step": 48066 + }, + { + "epoch": 1.85, + "learning_rate": 1.9211977430110285e-05, + "loss": 0.2845, + "step": 48067 + }, + { + "epoch": 1.85, + "learning_rate": 1.9211336240061557e-05, + "loss": 0.3205, + "step": 48068 + }, + { + "epoch": 1.85, + "learning_rate": 1.9210695050012825e-05, + "loss": 0.1758, + "step": 48069 + }, + { + "epoch": 1.85, + "learning_rate": 1.9210053859964093e-05, + "loss": 0.411, + "step": 48070 + }, + { + "epoch": 1.85, + "learning_rate": 1.9209412669915365e-05, + "loss": 0.216, + "step": 48071 + }, + { + "epoch": 1.85, + "learning_rate": 1.9208771479866633e-05, + "loss": 0.208, + "step": 48072 + }, + { + "epoch": 1.85, + "learning_rate": 1.9208130289817902e-05, + "loss": 0.3102, + "step": 48073 + }, + { + "epoch": 1.85, + "learning_rate": 1.920748909976917e-05, + "loss": 0.3917, + "step": 48074 + }, + { + "epoch": 1.85, + "learning_rate": 1.9206847909720442e-05, + "loss": 0.3352, + "step": 48075 + }, + { + "epoch": 1.85, + "learning_rate": 1.9206206719671714e-05, + "loss": 0.1717, + "step": 48076 + }, + { + "epoch": 1.85, + "learning_rate": 1.9205565529622982e-05, + "loss": 0.2155, + "step": 48077 + }, + { + "epoch": 1.85, + "learning_rate": 1.920492433957425e-05, + "loss": 0.3691, + "step": 48078 + }, + { + "epoch": 1.85, + "learning_rate": 1.9204283149525522e-05, + "loss": 0.3623, + "step": 48079 + }, + { + "epoch": 1.85, + "learning_rate": 1.920364195947679e-05, + "loss": 0.4172, + "step": 48080 + }, + { + "epoch": 1.85, + "learning_rate": 1.920300076942806e-05, + "loss": 0.2367, + "step": 48081 + }, + { + "epoch": 1.85, + "learning_rate": 1.9202359579379327e-05, + "loss": 0.4895, + "step": 48082 + }, + { + "epoch": 1.85, + "learning_rate": 1.92017183893306e-05, + "loss": 0.3574, + "step": 48083 + }, + { + "epoch": 1.85, + "learning_rate": 1.920107719928187e-05, + "loss": 0.2223, + "step": 48084 + }, + { + "epoch": 1.85, + "learning_rate": 1.920043600923314e-05, + "loss": 0.3186, + "step": 48085 + }, + { + "epoch": 1.85, + "learning_rate": 1.9199794819184407e-05, + "loss": 0.2794, + "step": 48086 + }, + { + "epoch": 1.85, + "learning_rate": 1.919915362913568e-05, + "loss": 0.2914, + "step": 48087 + }, + { + "epoch": 1.85, + "learning_rate": 1.9198512439086947e-05, + "loss": 0.203, + "step": 48088 + }, + { + "epoch": 1.85, + "learning_rate": 1.9197871249038215e-05, + "loss": 0.3773, + "step": 48089 + }, + { + "epoch": 1.85, + "learning_rate": 1.9197230058989484e-05, + "loss": 0.3531, + "step": 48090 + }, + { + "epoch": 1.85, + "learning_rate": 1.9196588868940755e-05, + "loss": 0.4117, + "step": 48091 + }, + { + "epoch": 1.85, + "learning_rate": 1.9195947678892024e-05, + "loss": 0.2699, + "step": 48092 + }, + { + "epoch": 1.85, + "learning_rate": 1.9195306488843295e-05, + "loss": 0.2854, + "step": 48093 + }, + { + "epoch": 1.85, + "learning_rate": 1.9194665298794564e-05, + "loss": 0.2614, + "step": 48094 + }, + { + "epoch": 1.85, + "learning_rate": 1.9194024108745836e-05, + "loss": 0.3991, + "step": 48095 + }, + { + "epoch": 1.85, + "learning_rate": 1.9193382918697104e-05, + "loss": 0.3987, + "step": 48096 + }, + { + "epoch": 1.85, + "learning_rate": 1.9192741728648372e-05, + "loss": 0.2517, + "step": 48097 + }, + { + "epoch": 1.85, + "learning_rate": 1.919210053859964e-05, + "loss": 0.3916, + "step": 48098 + }, + { + "epoch": 1.85, + "learning_rate": 1.9191459348550912e-05, + "loss": 0.3176, + "step": 48099 + }, + { + "epoch": 1.85, + "learning_rate": 1.919081815850218e-05, + "loss": 0.3513, + "step": 48100 + }, + { + "epoch": 1.85, + "learning_rate": 1.919017696845345e-05, + "loss": 0.2886, + "step": 48101 + }, + { + "epoch": 1.85, + "learning_rate": 1.918953577840472e-05, + "loss": 0.3087, + "step": 48102 + }, + { + "epoch": 1.85, + "learning_rate": 1.918889458835599e-05, + "loss": 0.2684, + "step": 48103 + }, + { + "epoch": 1.85, + "learning_rate": 1.918825339830726e-05, + "loss": 0.1791, + "step": 48104 + }, + { + "epoch": 1.85, + "learning_rate": 1.918761220825853e-05, + "loss": 0.4732, + "step": 48105 + }, + { + "epoch": 1.85, + "learning_rate": 1.9186971018209797e-05, + "loss": 0.4997, + "step": 48106 + }, + { + "epoch": 1.85, + "learning_rate": 1.918632982816107e-05, + "loss": 0.4229, + "step": 48107 + }, + { + "epoch": 1.85, + "learning_rate": 1.9185688638112337e-05, + "loss": 0.3692, + "step": 48108 + }, + { + "epoch": 1.85, + "learning_rate": 1.9185047448063606e-05, + "loss": 0.2417, + "step": 48109 + }, + { + "epoch": 1.85, + "learning_rate": 1.9184406258014874e-05, + "loss": 0.3266, + "step": 48110 + }, + { + "epoch": 1.85, + "learning_rate": 1.9183765067966146e-05, + "loss": 0.4244, + "step": 48111 + }, + { + "epoch": 1.85, + "learning_rate": 1.9183123877917417e-05, + "loss": 0.3435, + "step": 48112 + }, + { + "epoch": 1.85, + "learning_rate": 1.9182482687868686e-05, + "loss": 0.2513, + "step": 48113 + }, + { + "epoch": 1.85, + "learning_rate": 1.9181841497819954e-05, + "loss": 0.2861, + "step": 48114 + }, + { + "epoch": 1.85, + "learning_rate": 1.9181200307771226e-05, + "loss": 0.2717, + "step": 48115 + }, + { + "epoch": 1.85, + "learning_rate": 1.9180559117722494e-05, + "loss": 0.2591, + "step": 48116 + }, + { + "epoch": 1.85, + "learning_rate": 1.9179917927673762e-05, + "loss": 0.4029, + "step": 48117 + }, + { + "epoch": 1.85, + "learning_rate": 1.917927673762503e-05, + "loss": 0.223, + "step": 48118 + }, + { + "epoch": 1.85, + "learning_rate": 1.9178635547576303e-05, + "loss": 0.2387, + "step": 48119 + }, + { + "epoch": 1.85, + "learning_rate": 1.9177994357527574e-05, + "loss": 0.4104, + "step": 48120 + }, + { + "epoch": 1.85, + "learning_rate": 1.9177353167478843e-05, + "loss": 0.2906, + "step": 48121 + }, + { + "epoch": 1.85, + "learning_rate": 1.917671197743011e-05, + "loss": 0.277, + "step": 48122 + }, + { + "epoch": 1.85, + "learning_rate": 1.9176070787381383e-05, + "loss": 0.2771, + "step": 48123 + }, + { + "epoch": 1.85, + "learning_rate": 1.917542959733265e-05, + "loss": 0.279, + "step": 48124 + }, + { + "epoch": 1.85, + "learning_rate": 1.917478840728392e-05, + "loss": 0.2821, + "step": 48125 + }, + { + "epoch": 1.85, + "learning_rate": 1.9174147217235188e-05, + "loss": 0.3816, + "step": 48126 + }, + { + "epoch": 1.85, + "learning_rate": 1.917350602718646e-05, + "loss": 0.4048, + "step": 48127 + }, + { + "epoch": 1.85, + "learning_rate": 1.9172864837137728e-05, + "loss": 0.355, + "step": 48128 + }, + { + "epoch": 1.85, + "learning_rate": 1.9172223647089e-05, + "loss": 0.3609, + "step": 48129 + }, + { + "epoch": 1.85, + "learning_rate": 1.9171582457040268e-05, + "loss": 0.2461, + "step": 48130 + }, + { + "epoch": 1.85, + "learning_rate": 1.917094126699154e-05, + "loss": 0.3327, + "step": 48131 + }, + { + "epoch": 1.85, + "learning_rate": 1.9170300076942808e-05, + "loss": 0.3008, + "step": 48132 + }, + { + "epoch": 1.85, + "learning_rate": 1.9169658886894076e-05, + "loss": 0.3493, + "step": 48133 + }, + { + "epoch": 1.85, + "learning_rate": 1.9169017696845344e-05, + "loss": 0.3356, + "step": 48134 + }, + { + "epoch": 1.85, + "learning_rate": 1.9168376506796616e-05, + "loss": 0.2512, + "step": 48135 + }, + { + "epoch": 1.85, + "learning_rate": 1.9167735316747884e-05, + "loss": 0.2908, + "step": 48136 + }, + { + "epoch": 1.85, + "learning_rate": 1.9167094126699153e-05, + "loss": 0.316, + "step": 48137 + }, + { + "epoch": 1.85, + "learning_rate": 1.9166452936650424e-05, + "loss": 0.3515, + "step": 48138 + }, + { + "epoch": 1.85, + "learning_rate": 1.9165811746601696e-05, + "loss": 0.1994, + "step": 48139 + }, + { + "epoch": 1.85, + "learning_rate": 1.9165170556552965e-05, + "loss": 0.2419, + "step": 48140 + }, + { + "epoch": 1.85, + "learning_rate": 1.9164529366504233e-05, + "loss": 0.461, + "step": 48141 + }, + { + "epoch": 1.85, + "learning_rate": 1.91638881764555e-05, + "loss": 0.2559, + "step": 48142 + }, + { + "epoch": 1.85, + "learning_rate": 1.9163246986406773e-05, + "loss": 0.4234, + "step": 48143 + }, + { + "epoch": 1.85, + "learning_rate": 1.916260579635804e-05, + "loss": 0.2683, + "step": 48144 + }, + { + "epoch": 1.85, + "learning_rate": 1.916196460630931e-05, + "loss": 0.3423, + "step": 48145 + }, + { + "epoch": 1.85, + "learning_rate": 1.9161323416260578e-05, + "loss": 0.4134, + "step": 48146 + }, + { + "epoch": 1.85, + "learning_rate": 1.916068222621185e-05, + "loss": 0.4549, + "step": 48147 + }, + { + "epoch": 1.85, + "learning_rate": 1.916004103616312e-05, + "loss": 0.3882, + "step": 48148 + }, + { + "epoch": 1.85, + "learning_rate": 1.915939984611439e-05, + "loss": 0.4388, + "step": 48149 + }, + { + "epoch": 1.85, + "learning_rate": 1.9158758656065658e-05, + "loss": 0.3345, + "step": 48150 + }, + { + "epoch": 1.85, + "learning_rate": 1.915811746601693e-05, + "loss": 0.3015, + "step": 48151 + }, + { + "epoch": 1.85, + "learning_rate": 1.9157476275968198e-05, + "loss": 0.407, + "step": 48152 + }, + { + "epoch": 1.85, + "learning_rate": 1.9156835085919466e-05, + "loss": 0.2881, + "step": 48153 + }, + { + "epoch": 1.85, + "learning_rate": 1.9156193895870735e-05, + "loss": 0.33, + "step": 48154 + }, + { + "epoch": 1.85, + "learning_rate": 1.9155552705822006e-05, + "loss": 0.3739, + "step": 48155 + }, + { + "epoch": 1.85, + "learning_rate": 1.9154911515773278e-05, + "loss": 0.2622, + "step": 48156 + }, + { + "epoch": 1.85, + "learning_rate": 1.9154270325724546e-05, + "loss": 0.2885, + "step": 48157 + }, + { + "epoch": 1.85, + "learning_rate": 1.9153629135675815e-05, + "loss": 0.4853, + "step": 48158 + }, + { + "epoch": 1.85, + "learning_rate": 1.9152987945627086e-05, + "loss": 0.3576, + "step": 48159 + }, + { + "epoch": 1.85, + "learning_rate": 1.9152346755578355e-05, + "loss": 0.3653, + "step": 48160 + }, + { + "epoch": 1.85, + "learning_rate": 1.9151705565529623e-05, + "loss": 0.366, + "step": 48161 + }, + { + "epoch": 1.85, + "learning_rate": 1.915106437548089e-05, + "loss": 0.3219, + "step": 48162 + }, + { + "epoch": 1.85, + "learning_rate": 1.9150423185432163e-05, + "loss": 0.3351, + "step": 48163 + }, + { + "epoch": 1.85, + "learning_rate": 1.914978199538343e-05, + "loss": 0.2586, + "step": 48164 + }, + { + "epoch": 1.85, + "learning_rate": 1.9149140805334703e-05, + "loss": 0.3576, + "step": 48165 + }, + { + "epoch": 1.85, + "learning_rate": 1.914849961528597e-05, + "loss": 0.2983, + "step": 48166 + }, + { + "epoch": 1.85, + "learning_rate": 1.9147858425237243e-05, + "loss": 0.3227, + "step": 48167 + }, + { + "epoch": 1.85, + "learning_rate": 1.914721723518851e-05, + "loss": 0.3109, + "step": 48168 + }, + { + "epoch": 1.85, + "learning_rate": 1.914657604513978e-05, + "loss": 0.3139, + "step": 48169 + }, + { + "epoch": 1.85, + "learning_rate": 1.9145934855091048e-05, + "loss": 0.3244, + "step": 48170 + }, + { + "epoch": 1.85, + "learning_rate": 1.914529366504232e-05, + "loss": 0.3535, + "step": 48171 + }, + { + "epoch": 1.85, + "learning_rate": 1.914465247499359e-05, + "loss": 0.3286, + "step": 48172 + }, + { + "epoch": 1.85, + "learning_rate": 1.914401128494486e-05, + "loss": 0.3369, + "step": 48173 + }, + { + "epoch": 1.85, + "learning_rate": 1.914337009489613e-05, + "loss": 0.4597, + "step": 48174 + }, + { + "epoch": 1.85, + "learning_rate": 1.91427289048474e-05, + "loss": 0.489, + "step": 48175 + }, + { + "epoch": 1.85, + "learning_rate": 1.914208771479867e-05, + "loss": 0.1573, + "step": 48176 + }, + { + "epoch": 1.85, + "learning_rate": 1.9141446524749937e-05, + "loss": 0.3079, + "step": 48177 + }, + { + "epoch": 1.85, + "learning_rate": 1.9140805334701205e-05, + "loss": 0.415, + "step": 48178 + }, + { + "epoch": 1.85, + "learning_rate": 1.9140164144652477e-05, + "loss": 0.4102, + "step": 48179 + }, + { + "epoch": 1.85, + "learning_rate": 1.9139522954603745e-05, + "loss": 0.3182, + "step": 48180 + }, + { + "epoch": 1.85, + "learning_rate": 1.9138881764555013e-05, + "loss": 0.4129, + "step": 48181 + }, + { + "epoch": 1.85, + "learning_rate": 1.9138240574506285e-05, + "loss": 0.3434, + "step": 48182 + }, + { + "epoch": 1.85, + "learning_rate": 1.9137599384457557e-05, + "loss": 0.3419, + "step": 48183 + }, + { + "epoch": 1.85, + "learning_rate": 1.9136958194408825e-05, + "loss": 0.312, + "step": 48184 + }, + { + "epoch": 1.85, + "learning_rate": 1.9136317004360094e-05, + "loss": 0.3119, + "step": 48185 + }, + { + "epoch": 1.85, + "learning_rate": 1.9135675814311362e-05, + "loss": 0.3033, + "step": 48186 + }, + { + "epoch": 1.85, + "learning_rate": 1.9135034624262634e-05, + "loss": 0.3295, + "step": 48187 + }, + { + "epoch": 1.85, + "learning_rate": 1.9134393434213902e-05, + "loss": 0.4042, + "step": 48188 + }, + { + "epoch": 1.85, + "learning_rate": 1.913375224416517e-05, + "loss": 0.2191, + "step": 48189 + }, + { + "epoch": 1.85, + "learning_rate": 1.913311105411644e-05, + "loss": 0.2947, + "step": 48190 + }, + { + "epoch": 1.85, + "learning_rate": 1.913246986406771e-05, + "loss": 0.3553, + "step": 48191 + }, + { + "epoch": 1.85, + "learning_rate": 1.9131828674018982e-05, + "loss": 0.3274, + "step": 48192 + }, + { + "epoch": 1.85, + "learning_rate": 1.913118748397025e-05, + "loss": 0.3157, + "step": 48193 + }, + { + "epoch": 1.85, + "learning_rate": 1.913054629392152e-05, + "loss": 0.2001, + "step": 48194 + }, + { + "epoch": 1.85, + "learning_rate": 1.912990510387279e-05, + "loss": 0.4245, + "step": 48195 + }, + { + "epoch": 1.85, + "learning_rate": 1.912926391382406e-05, + "loss": 0.3864, + "step": 48196 + }, + { + "epoch": 1.85, + "learning_rate": 1.9128622723775327e-05, + "loss": 0.3795, + "step": 48197 + }, + { + "epoch": 1.85, + "learning_rate": 1.9127981533726595e-05, + "loss": 0.4709, + "step": 48198 + }, + { + "epoch": 1.85, + "learning_rate": 1.9127340343677867e-05, + "loss": 0.1481, + "step": 48199 + }, + { + "epoch": 1.85, + "learning_rate": 1.912669915362914e-05, + "loss": 0.5055, + "step": 48200 + }, + { + "epoch": 1.85, + "learning_rate": 1.9126057963580407e-05, + "loss": 0.2817, + "step": 48201 + }, + { + "epoch": 1.85, + "learning_rate": 1.9125416773531675e-05, + "loss": 0.2949, + "step": 48202 + }, + { + "epoch": 1.85, + "learning_rate": 1.9124775583482947e-05, + "loss": 0.3573, + "step": 48203 + }, + { + "epoch": 1.85, + "learning_rate": 1.9124134393434215e-05, + "loss": 0.2344, + "step": 48204 + }, + { + "epoch": 1.85, + "learning_rate": 1.9123493203385484e-05, + "loss": 0.3245, + "step": 48205 + }, + { + "epoch": 1.85, + "learning_rate": 1.9122852013336752e-05, + "loss": 0.4075, + "step": 48206 + }, + { + "epoch": 1.85, + "learning_rate": 1.9122210823288024e-05, + "loss": 0.2681, + "step": 48207 + }, + { + "epoch": 1.85, + "learning_rate": 1.9121569633239292e-05, + "loss": 0.4276, + "step": 48208 + }, + { + "epoch": 1.85, + "learning_rate": 1.9120928443190564e-05, + "loss": 0.2708, + "step": 48209 + }, + { + "epoch": 1.85, + "learning_rate": 1.9120287253141832e-05, + "loss": 0.3637, + "step": 48210 + }, + { + "epoch": 1.85, + "learning_rate": 1.9119646063093104e-05, + "loss": 0.2705, + "step": 48211 + }, + { + "epoch": 1.85, + "learning_rate": 1.9119004873044372e-05, + "loss": 0.4103, + "step": 48212 + }, + { + "epoch": 1.85, + "learning_rate": 1.911836368299564e-05, + "loss": 0.414, + "step": 48213 + }, + { + "epoch": 1.85, + "learning_rate": 1.911772249294691e-05, + "loss": 0.2394, + "step": 48214 + }, + { + "epoch": 1.85, + "learning_rate": 1.911708130289818e-05, + "loss": 0.3039, + "step": 48215 + }, + { + "epoch": 1.85, + "learning_rate": 1.911644011284945e-05, + "loss": 0.3129, + "step": 48216 + }, + { + "epoch": 1.85, + "learning_rate": 1.9115798922800717e-05, + "loss": 0.4607, + "step": 48217 + }, + { + "epoch": 1.85, + "learning_rate": 1.911515773275199e-05, + "loss": 0.4851, + "step": 48218 + }, + { + "epoch": 1.85, + "learning_rate": 1.911451654270326e-05, + "loss": 0.2634, + "step": 48219 + }, + { + "epoch": 1.85, + "learning_rate": 1.911387535265453e-05, + "loss": 0.3493, + "step": 48220 + }, + { + "epoch": 1.85, + "learning_rate": 1.9113234162605797e-05, + "loss": 0.1924, + "step": 48221 + }, + { + "epoch": 1.85, + "learning_rate": 1.9112592972557066e-05, + "loss": 0.2247, + "step": 48222 + }, + { + "epoch": 1.85, + "learning_rate": 1.9111951782508337e-05, + "loss": 0.3246, + "step": 48223 + }, + { + "epoch": 1.85, + "learning_rate": 1.9111310592459606e-05, + "loss": 0.3273, + "step": 48224 + }, + { + "epoch": 1.85, + "learning_rate": 1.9110669402410874e-05, + "loss": 0.2183, + "step": 48225 + }, + { + "epoch": 1.85, + "learning_rate": 1.9110028212362142e-05, + "loss": 0.4008, + "step": 48226 + }, + { + "epoch": 1.85, + "learning_rate": 1.9109387022313414e-05, + "loss": 0.1582, + "step": 48227 + }, + { + "epoch": 1.85, + "learning_rate": 1.9108745832264686e-05, + "loss": 0.2637, + "step": 48228 + }, + { + "epoch": 1.85, + "learning_rate": 1.9108104642215954e-05, + "loss": 0.2303, + "step": 48229 + }, + { + "epoch": 1.85, + "learning_rate": 1.9107463452167223e-05, + "loss": 0.3171, + "step": 48230 + }, + { + "epoch": 1.85, + "learning_rate": 1.9106822262118494e-05, + "loss": 0.3795, + "step": 48231 + }, + { + "epoch": 1.85, + "learning_rate": 1.9106181072069763e-05, + "loss": 0.3143, + "step": 48232 + }, + { + "epoch": 1.85, + "learning_rate": 1.910553988202103e-05, + "loss": 0.5725, + "step": 48233 + }, + { + "epoch": 1.85, + "learning_rate": 1.91048986919723e-05, + "loss": 0.2616, + "step": 48234 + }, + { + "epoch": 1.85, + "learning_rate": 1.910425750192357e-05, + "loss": 0.308, + "step": 48235 + }, + { + "epoch": 1.85, + "learning_rate": 1.9103616311874843e-05, + "loss": 0.2834, + "step": 48236 + }, + { + "epoch": 1.85, + "learning_rate": 1.910297512182611e-05, + "loss": 0.3519, + "step": 48237 + }, + { + "epoch": 1.85, + "learning_rate": 1.910233393177738e-05, + "loss": 0.391, + "step": 48238 + }, + { + "epoch": 1.85, + "learning_rate": 1.910169274172865e-05, + "loss": 0.4073, + "step": 48239 + }, + { + "epoch": 1.85, + "learning_rate": 1.910105155167992e-05, + "loss": 0.351, + "step": 48240 + }, + { + "epoch": 1.85, + "learning_rate": 1.9100410361631188e-05, + "loss": 0.2317, + "step": 48241 + }, + { + "epoch": 1.85, + "learning_rate": 1.9099769171582456e-05, + "loss": 0.297, + "step": 48242 + }, + { + "epoch": 1.85, + "learning_rate": 1.9099127981533728e-05, + "loss": 0.4559, + "step": 48243 + }, + { + "epoch": 1.85, + "learning_rate": 1.9098486791484996e-05, + "loss": 0.4482, + "step": 48244 + }, + { + "epoch": 1.85, + "learning_rate": 1.9097845601436268e-05, + "loss": 0.438, + "step": 48245 + }, + { + "epoch": 1.85, + "learning_rate": 1.9097204411387536e-05, + "loss": 0.3001, + "step": 48246 + }, + { + "epoch": 1.85, + "learning_rate": 1.9096563221338808e-05, + "loss": 0.429, + "step": 48247 + }, + { + "epoch": 1.85, + "learning_rate": 1.9095922031290076e-05, + "loss": 0.4158, + "step": 48248 + }, + { + "epoch": 1.86, + "learning_rate": 1.9095280841241344e-05, + "loss": 0.3884, + "step": 48249 + }, + { + "epoch": 1.86, + "learning_rate": 1.9094639651192613e-05, + "loss": 0.2639, + "step": 48250 + }, + { + "epoch": 1.86, + "learning_rate": 1.9093998461143885e-05, + "loss": 0.3103, + "step": 48251 + }, + { + "epoch": 1.86, + "learning_rate": 1.9093357271095153e-05, + "loss": 0.3603, + "step": 48252 + }, + { + "epoch": 1.86, + "learning_rate": 1.909271608104642e-05, + "loss": 0.2588, + "step": 48253 + }, + { + "epoch": 1.86, + "learning_rate": 1.9092074890997693e-05, + "loss": 0.3481, + "step": 48254 + }, + { + "epoch": 1.86, + "learning_rate": 1.9091433700948965e-05, + "loss": 0.3362, + "step": 48255 + }, + { + "epoch": 1.86, + "learning_rate": 1.9090792510900233e-05, + "loss": 0.2919, + "step": 48256 + }, + { + "epoch": 1.86, + "learning_rate": 1.90901513208515e-05, + "loss": 0.4117, + "step": 48257 + }, + { + "epoch": 1.86, + "learning_rate": 1.908951013080277e-05, + "loss": 0.3178, + "step": 48258 + }, + { + "epoch": 1.86, + "learning_rate": 1.908886894075404e-05, + "loss": 0.3036, + "step": 48259 + }, + { + "epoch": 1.86, + "learning_rate": 1.908822775070531e-05, + "loss": 0.3004, + "step": 48260 + }, + { + "epoch": 1.86, + "learning_rate": 1.9087586560656578e-05, + "loss": 0.4364, + "step": 48261 + }, + { + "epoch": 1.86, + "learning_rate": 1.9086945370607846e-05, + "loss": 0.3242, + "step": 48262 + }, + { + "epoch": 1.86, + "learning_rate": 1.908630418055912e-05, + "loss": 0.3098, + "step": 48263 + }, + { + "epoch": 1.86, + "learning_rate": 1.908566299051039e-05, + "loss": 0.2704, + "step": 48264 + }, + { + "epoch": 1.86, + "learning_rate": 1.9085021800461658e-05, + "loss": 0.3508, + "step": 48265 + }, + { + "epoch": 1.86, + "learning_rate": 1.9084380610412926e-05, + "loss": 0.3846, + "step": 48266 + }, + { + "epoch": 1.86, + "learning_rate": 1.9083739420364198e-05, + "loss": 0.3925, + "step": 48267 + }, + { + "epoch": 1.86, + "learning_rate": 1.9083098230315466e-05, + "loss": 0.2848, + "step": 48268 + }, + { + "epoch": 1.86, + "learning_rate": 1.9082457040266735e-05, + "loss": 0.3176, + "step": 48269 + }, + { + "epoch": 1.86, + "learning_rate": 1.9081815850218003e-05, + "loss": 0.4168, + "step": 48270 + }, + { + "epoch": 1.86, + "learning_rate": 1.9081174660169275e-05, + "loss": 0.2542, + "step": 48271 + }, + { + "epoch": 1.86, + "learning_rate": 1.9080533470120547e-05, + "loss": 0.3556, + "step": 48272 + }, + { + "epoch": 1.86, + "learning_rate": 1.9079892280071815e-05, + "loss": 0.363, + "step": 48273 + }, + { + "epoch": 1.86, + "learning_rate": 1.9079251090023083e-05, + "loss": 0.3396, + "step": 48274 + }, + { + "epoch": 1.86, + "learning_rate": 1.9078609899974355e-05, + "loss": 0.3575, + "step": 48275 + }, + { + "epoch": 1.86, + "learning_rate": 1.9077968709925623e-05, + "loss": 0.3991, + "step": 48276 + }, + { + "epoch": 1.86, + "learning_rate": 1.907732751987689e-05, + "loss": 0.3005, + "step": 48277 + }, + { + "epoch": 1.86, + "learning_rate": 1.907668632982816e-05, + "loss": 0.3225, + "step": 48278 + }, + { + "epoch": 1.86, + "learning_rate": 1.907604513977943e-05, + "loss": 0.263, + "step": 48279 + }, + { + "epoch": 1.86, + "learning_rate": 1.9075403949730703e-05, + "loss": 0.3444, + "step": 48280 + }, + { + "epoch": 1.86, + "learning_rate": 1.907476275968197e-05, + "loss": 0.5118, + "step": 48281 + }, + { + "epoch": 1.86, + "learning_rate": 1.907412156963324e-05, + "loss": 0.5414, + "step": 48282 + }, + { + "epoch": 1.86, + "learning_rate": 1.907348037958451e-05, + "loss": 0.3088, + "step": 48283 + }, + { + "epoch": 1.86, + "learning_rate": 1.907283918953578e-05, + "loss": 0.2596, + "step": 48284 + }, + { + "epoch": 1.86, + "learning_rate": 1.907219799948705e-05, + "loss": 0.5238, + "step": 48285 + }, + { + "epoch": 1.86, + "learning_rate": 1.9071556809438317e-05, + "loss": 0.2834, + "step": 48286 + }, + { + "epoch": 1.86, + "learning_rate": 1.907091561938959e-05, + "loss": 0.2554, + "step": 48287 + }, + { + "epoch": 1.86, + "learning_rate": 1.9070274429340857e-05, + "loss": 0.3205, + "step": 48288 + }, + { + "epoch": 1.86, + "learning_rate": 1.906963323929213e-05, + "loss": 0.357, + "step": 48289 + }, + { + "epoch": 1.86, + "learning_rate": 1.9068992049243397e-05, + "loss": 0.3556, + "step": 48290 + }, + { + "epoch": 1.86, + "learning_rate": 1.906835085919467e-05, + "loss": 0.3401, + "step": 48291 + }, + { + "epoch": 1.86, + "learning_rate": 1.9067709669145937e-05, + "loss": 0.4149, + "step": 48292 + }, + { + "epoch": 1.86, + "learning_rate": 1.9067068479097205e-05, + "loss": 0.2725, + "step": 48293 + }, + { + "epoch": 1.86, + "learning_rate": 1.9066427289048473e-05, + "loss": 0.2699, + "step": 48294 + }, + { + "epoch": 1.86, + "learning_rate": 1.9065786098999745e-05, + "loss": 0.2701, + "step": 48295 + }, + { + "epoch": 1.86, + "learning_rate": 1.9065144908951014e-05, + "loss": 0.3014, + "step": 48296 + }, + { + "epoch": 1.86, + "learning_rate": 1.9064503718902282e-05, + "loss": 0.3461, + "step": 48297 + }, + { + "epoch": 1.86, + "learning_rate": 1.9063862528853554e-05, + "loss": 0.3374, + "step": 48298 + }, + { + "epoch": 1.86, + "learning_rate": 1.9063221338804825e-05, + "loss": 0.3497, + "step": 48299 + }, + { + "epoch": 1.86, + "learning_rate": 1.9062580148756094e-05, + "loss": 0.2714, + "step": 48300 + }, + { + "epoch": 1.86, + "learning_rate": 1.9061938958707362e-05, + "loss": 0.5089, + "step": 48301 + }, + { + "epoch": 1.86, + "learning_rate": 1.906129776865863e-05, + "loss": 0.2642, + "step": 48302 + }, + { + "epoch": 1.86, + "learning_rate": 1.9060656578609902e-05, + "loss": 0.4867, + "step": 48303 + }, + { + "epoch": 1.86, + "learning_rate": 1.906001538856117e-05, + "loss": 0.3579, + "step": 48304 + }, + { + "epoch": 1.86, + "learning_rate": 1.905937419851244e-05, + "loss": 0.3985, + "step": 48305 + }, + { + "epoch": 1.86, + "learning_rate": 1.9058733008463707e-05, + "loss": 0.3414, + "step": 48306 + }, + { + "epoch": 1.86, + "learning_rate": 1.905809181841498e-05, + "loss": 0.3967, + "step": 48307 + }, + { + "epoch": 1.86, + "learning_rate": 1.905745062836625e-05, + "loss": 0.4185, + "step": 48308 + }, + { + "epoch": 1.86, + "learning_rate": 1.905680943831752e-05, + "loss": 0.4056, + "step": 48309 + }, + { + "epoch": 1.86, + "learning_rate": 1.9056168248268787e-05, + "loss": 0.2588, + "step": 48310 + }, + { + "epoch": 1.86, + "learning_rate": 1.905552705822006e-05, + "loss": 0.3189, + "step": 48311 + }, + { + "epoch": 1.86, + "learning_rate": 1.9054885868171327e-05, + "loss": 0.3384, + "step": 48312 + }, + { + "epoch": 1.86, + "learning_rate": 1.9054244678122595e-05, + "loss": 0.3556, + "step": 48313 + }, + { + "epoch": 1.86, + "learning_rate": 1.9053603488073864e-05, + "loss": 0.3218, + "step": 48314 + }, + { + "epoch": 1.86, + "learning_rate": 1.9052962298025135e-05, + "loss": 0.3454, + "step": 48315 + }, + { + "epoch": 1.86, + "learning_rate": 1.9052321107976407e-05, + "loss": 0.3842, + "step": 48316 + }, + { + "epoch": 1.86, + "learning_rate": 1.9051679917927676e-05, + "loss": 0.3648, + "step": 48317 + }, + { + "epoch": 1.86, + "learning_rate": 1.9051038727878944e-05, + "loss": 0.3133, + "step": 48318 + }, + { + "epoch": 1.86, + "learning_rate": 1.9050397537830216e-05, + "loss": 0.2971, + "step": 48319 + }, + { + "epoch": 1.86, + "learning_rate": 1.9049756347781484e-05, + "loss": 0.3916, + "step": 48320 + }, + { + "epoch": 1.86, + "learning_rate": 1.9049115157732752e-05, + "loss": 0.2173, + "step": 48321 + }, + { + "epoch": 1.86, + "learning_rate": 1.904847396768402e-05, + "loss": 0.259, + "step": 48322 + }, + { + "epoch": 1.86, + "learning_rate": 1.9047832777635292e-05, + "loss": 0.3062, + "step": 48323 + }, + { + "epoch": 1.86, + "learning_rate": 1.904719158758656e-05, + "loss": 0.3456, + "step": 48324 + }, + { + "epoch": 1.86, + "learning_rate": 1.9046550397537832e-05, + "loss": 0.2768, + "step": 48325 + }, + { + "epoch": 1.86, + "learning_rate": 1.90459092074891e-05, + "loss": 0.3067, + "step": 48326 + }, + { + "epoch": 1.86, + "learning_rate": 1.9045268017440372e-05, + "loss": 0.2121, + "step": 48327 + }, + { + "epoch": 1.86, + "learning_rate": 1.904462682739164e-05, + "loss": 0.2714, + "step": 48328 + }, + { + "epoch": 1.86, + "learning_rate": 1.904398563734291e-05, + "loss": 0.2595, + "step": 48329 + }, + { + "epoch": 1.86, + "learning_rate": 1.9043344447294177e-05, + "loss": 0.2309, + "step": 48330 + }, + { + "epoch": 1.86, + "learning_rate": 1.904270325724545e-05, + "loss": 0.4423, + "step": 48331 + }, + { + "epoch": 1.86, + "learning_rate": 1.9042062067196717e-05, + "loss": 0.2748, + "step": 48332 + }, + { + "epoch": 1.86, + "learning_rate": 1.9041420877147986e-05, + "loss": 0.2954, + "step": 48333 + }, + { + "epoch": 1.86, + "learning_rate": 1.9040779687099257e-05, + "loss": 0.3357, + "step": 48334 + }, + { + "epoch": 1.86, + "learning_rate": 1.904013849705053e-05, + "loss": 0.5219, + "step": 48335 + }, + { + "epoch": 1.86, + "learning_rate": 1.9039497307001797e-05, + "loss": 0.2847, + "step": 48336 + }, + { + "epoch": 1.86, + "learning_rate": 1.9038856116953066e-05, + "loss": 0.255, + "step": 48337 + }, + { + "epoch": 1.86, + "learning_rate": 1.9038214926904334e-05, + "loss": 0.2626, + "step": 48338 + }, + { + "epoch": 1.86, + "learning_rate": 1.9037573736855606e-05, + "loss": 0.2883, + "step": 48339 + }, + { + "epoch": 1.86, + "learning_rate": 1.9036932546806874e-05, + "loss": 0.1747, + "step": 48340 + }, + { + "epoch": 1.86, + "learning_rate": 1.9036291356758143e-05, + "loss": 0.2995, + "step": 48341 + }, + { + "epoch": 1.86, + "learning_rate": 1.903565016670941e-05, + "loss": 0.266, + "step": 48342 + }, + { + "epoch": 1.86, + "learning_rate": 1.9035008976660686e-05, + "loss": 0.288, + "step": 48343 + }, + { + "epoch": 1.86, + "learning_rate": 1.9034367786611954e-05, + "loss": 0.3031, + "step": 48344 + }, + { + "epoch": 1.86, + "learning_rate": 1.9033726596563223e-05, + "loss": 0.3605, + "step": 48345 + }, + { + "epoch": 1.86, + "learning_rate": 1.903308540651449e-05, + "loss": 0.3501, + "step": 48346 + }, + { + "epoch": 1.86, + "learning_rate": 1.9032444216465763e-05, + "loss": 0.2618, + "step": 48347 + }, + { + "epoch": 1.86, + "learning_rate": 1.903180302641703e-05, + "loss": 0.4181, + "step": 48348 + }, + { + "epoch": 1.86, + "learning_rate": 1.90311618363683e-05, + "loss": 0.3084, + "step": 48349 + }, + { + "epoch": 1.86, + "learning_rate": 1.9030520646319568e-05, + "loss": 0.4262, + "step": 48350 + }, + { + "epoch": 1.86, + "learning_rate": 1.902987945627084e-05, + "loss": 0.3437, + "step": 48351 + }, + { + "epoch": 1.86, + "learning_rate": 1.902923826622211e-05, + "loss": 0.2915, + "step": 48352 + }, + { + "epoch": 1.86, + "learning_rate": 1.902859707617338e-05, + "loss": 0.2765, + "step": 48353 + }, + { + "epoch": 1.86, + "learning_rate": 1.9027955886124648e-05, + "loss": 0.28, + "step": 48354 + }, + { + "epoch": 1.86, + "learning_rate": 1.902731469607592e-05, + "loss": 0.2844, + "step": 48355 + }, + { + "epoch": 1.86, + "learning_rate": 1.9026673506027188e-05, + "loss": 0.3848, + "step": 48356 + }, + { + "epoch": 1.86, + "learning_rate": 1.9026032315978456e-05, + "loss": 0.3988, + "step": 48357 + }, + { + "epoch": 1.86, + "learning_rate": 1.9025391125929724e-05, + "loss": 0.2484, + "step": 48358 + }, + { + "epoch": 1.86, + "learning_rate": 1.9024749935880996e-05, + "loss": 0.4232, + "step": 48359 + }, + { + "epoch": 1.86, + "learning_rate": 1.9024108745832264e-05, + "loss": 0.2364, + "step": 48360 + }, + { + "epoch": 1.86, + "learning_rate": 1.9023467555783536e-05, + "loss": 0.359, + "step": 48361 + }, + { + "epoch": 1.86, + "learning_rate": 1.9022826365734805e-05, + "loss": 0.3138, + "step": 48362 + }, + { + "epoch": 1.86, + "learning_rate": 1.9022185175686076e-05, + "loss": 0.2844, + "step": 48363 + }, + { + "epoch": 1.86, + "learning_rate": 1.9021543985637345e-05, + "loss": 0.4435, + "step": 48364 + }, + { + "epoch": 1.86, + "learning_rate": 1.9020902795588613e-05, + "loss": 0.4329, + "step": 48365 + }, + { + "epoch": 1.86, + "learning_rate": 1.902026160553988e-05, + "loss": 0.3638, + "step": 48366 + }, + { + "epoch": 1.86, + "learning_rate": 1.9019620415491153e-05, + "loss": 0.2888, + "step": 48367 + }, + { + "epoch": 1.86, + "learning_rate": 1.901897922544242e-05, + "loss": 0.3452, + "step": 48368 + }, + { + "epoch": 1.86, + "learning_rate": 1.9018338035393693e-05, + "loss": 0.4049, + "step": 48369 + }, + { + "epoch": 1.86, + "learning_rate": 1.901769684534496e-05, + "loss": 0.1851, + "step": 48370 + }, + { + "epoch": 1.86, + "learning_rate": 1.9017055655296233e-05, + "loss": 0.2952, + "step": 48371 + }, + { + "epoch": 1.86, + "learning_rate": 1.90164144652475e-05, + "loss": 0.3628, + "step": 48372 + }, + { + "epoch": 1.86, + "learning_rate": 1.901577327519877e-05, + "loss": 0.3197, + "step": 48373 + }, + { + "epoch": 1.86, + "learning_rate": 1.9015132085150038e-05, + "loss": 0.2395, + "step": 48374 + }, + { + "epoch": 1.86, + "learning_rate": 1.901449089510131e-05, + "loss": 0.3539, + "step": 48375 + }, + { + "epoch": 1.86, + "learning_rate": 1.9013849705052578e-05, + "loss": 0.3475, + "step": 48376 + }, + { + "epoch": 1.86, + "learning_rate": 1.9013208515003846e-05, + "loss": 0.2957, + "step": 48377 + }, + { + "epoch": 1.86, + "learning_rate": 1.9012567324955118e-05, + "loss": 0.1967, + "step": 48378 + }, + { + "epoch": 1.86, + "learning_rate": 1.901192613490639e-05, + "loss": 0.2683, + "step": 48379 + }, + { + "epoch": 1.86, + "learning_rate": 1.9011284944857658e-05, + "loss": 0.3034, + "step": 48380 + }, + { + "epoch": 1.86, + "learning_rate": 1.9010643754808926e-05, + "loss": 0.3947, + "step": 48381 + }, + { + "epoch": 1.86, + "learning_rate": 1.9010002564760195e-05, + "loss": 0.3313, + "step": 48382 + }, + { + "epoch": 1.86, + "learning_rate": 1.9009361374711467e-05, + "loss": 0.279, + "step": 48383 + }, + { + "epoch": 1.86, + "learning_rate": 1.9008720184662735e-05, + "loss": 0.281, + "step": 48384 + }, + { + "epoch": 1.86, + "learning_rate": 1.9008078994614003e-05, + "loss": 0.27, + "step": 48385 + }, + { + "epoch": 1.86, + "learning_rate": 1.900743780456527e-05, + "loss": 0.4148, + "step": 48386 + }, + { + "epoch": 1.86, + "learning_rate": 1.9006796614516547e-05, + "loss": 0.3102, + "step": 48387 + }, + { + "epoch": 1.86, + "learning_rate": 1.9006155424467815e-05, + "loss": 0.3622, + "step": 48388 + }, + { + "epoch": 1.86, + "learning_rate": 1.9005514234419083e-05, + "loss": 0.2842, + "step": 48389 + }, + { + "epoch": 1.86, + "learning_rate": 1.900487304437035e-05, + "loss": 0.3616, + "step": 48390 + }, + { + "epoch": 1.86, + "learning_rate": 1.9004231854321623e-05, + "loss": 0.3063, + "step": 48391 + }, + { + "epoch": 1.86, + "learning_rate": 1.900359066427289e-05, + "loss": 0.2249, + "step": 48392 + }, + { + "epoch": 1.86, + "learning_rate": 1.900294947422416e-05, + "loss": 0.3496, + "step": 48393 + }, + { + "epoch": 1.86, + "learning_rate": 1.9002308284175428e-05, + "loss": 0.2438, + "step": 48394 + }, + { + "epoch": 1.86, + "learning_rate": 1.90016670941267e-05, + "loss": 0.423, + "step": 48395 + }, + { + "epoch": 1.86, + "learning_rate": 1.9001025904077972e-05, + "loss": 0.286, + "step": 48396 + }, + { + "epoch": 1.86, + "learning_rate": 1.900038471402924e-05, + "loss": 0.2228, + "step": 48397 + }, + { + "epoch": 1.86, + "learning_rate": 1.899974352398051e-05, + "loss": 0.2589, + "step": 48398 + }, + { + "epoch": 1.86, + "learning_rate": 1.899910233393178e-05, + "loss": 0.2943, + "step": 48399 + }, + { + "epoch": 1.86, + "learning_rate": 1.899846114388305e-05, + "loss": 0.4083, + "step": 48400 + }, + { + "epoch": 1.86, + "learning_rate": 1.8997819953834317e-05, + "loss": 0.3958, + "step": 48401 + }, + { + "epoch": 1.86, + "learning_rate": 1.8997178763785585e-05, + "loss": 0.3891, + "step": 48402 + }, + { + "epoch": 1.86, + "learning_rate": 1.8996537573736857e-05, + "loss": 0.4691, + "step": 48403 + }, + { + "epoch": 1.86, + "learning_rate": 1.8995896383688125e-05, + "loss": 0.4453, + "step": 48404 + }, + { + "epoch": 1.86, + "learning_rate": 1.8995255193639397e-05, + "loss": 0.3274, + "step": 48405 + }, + { + "epoch": 1.86, + "learning_rate": 1.8994614003590665e-05, + "loss": 0.4208, + "step": 48406 + }, + { + "epoch": 1.86, + "learning_rate": 1.8993972813541937e-05, + "loss": 0.3672, + "step": 48407 + }, + { + "epoch": 1.86, + "learning_rate": 1.8993331623493205e-05, + "loss": 0.2564, + "step": 48408 + }, + { + "epoch": 1.86, + "learning_rate": 1.8992690433444474e-05, + "loss": 0.1489, + "step": 48409 + }, + { + "epoch": 1.86, + "learning_rate": 1.8992049243395742e-05, + "loss": 0.4434, + "step": 48410 + }, + { + "epoch": 1.86, + "learning_rate": 1.8991408053347014e-05, + "loss": 0.3007, + "step": 48411 + }, + { + "epoch": 1.86, + "learning_rate": 1.8990766863298282e-05, + "loss": 0.285, + "step": 48412 + }, + { + "epoch": 1.86, + "learning_rate": 1.899012567324955e-05, + "loss": 0.2805, + "step": 48413 + }, + { + "epoch": 1.86, + "learning_rate": 1.8989484483200822e-05, + "loss": 0.3072, + "step": 48414 + }, + { + "epoch": 1.86, + "learning_rate": 1.8988843293152094e-05, + "loss": 0.3823, + "step": 48415 + }, + { + "epoch": 1.86, + "learning_rate": 1.8988202103103362e-05, + "loss": 0.3473, + "step": 48416 + }, + { + "epoch": 1.86, + "learning_rate": 1.898756091305463e-05, + "loss": 0.2669, + "step": 48417 + }, + { + "epoch": 1.86, + "learning_rate": 1.89869197230059e-05, + "loss": 0.3565, + "step": 48418 + }, + { + "epoch": 1.86, + "learning_rate": 1.898627853295717e-05, + "loss": 0.3474, + "step": 48419 + }, + { + "epoch": 1.86, + "learning_rate": 1.898563734290844e-05, + "loss": 0.3074, + "step": 48420 + }, + { + "epoch": 1.86, + "learning_rate": 1.8984996152859707e-05, + "loss": 0.3447, + "step": 48421 + }, + { + "epoch": 1.86, + "learning_rate": 1.8984354962810975e-05, + "loss": 0.2209, + "step": 48422 + }, + { + "epoch": 1.86, + "learning_rate": 1.898371377276225e-05, + "loss": 0.2635, + "step": 48423 + }, + { + "epoch": 1.86, + "learning_rate": 1.898307258271352e-05, + "loss": 0.4225, + "step": 48424 + }, + { + "epoch": 1.86, + "learning_rate": 1.8982431392664787e-05, + "loss": 0.3624, + "step": 48425 + }, + { + "epoch": 1.86, + "learning_rate": 1.8981790202616055e-05, + "loss": 0.3871, + "step": 48426 + }, + { + "epoch": 1.86, + "learning_rate": 1.8981149012567327e-05, + "loss": 0.3682, + "step": 48427 + }, + { + "epoch": 1.86, + "learning_rate": 1.8980507822518596e-05, + "loss": 0.3362, + "step": 48428 + }, + { + "epoch": 1.86, + "learning_rate": 1.8979866632469864e-05, + "loss": 0.2952, + "step": 48429 + }, + { + "epoch": 1.86, + "learning_rate": 1.8979225442421132e-05, + "loss": 0.2664, + "step": 48430 + }, + { + "epoch": 1.86, + "learning_rate": 1.8978584252372404e-05, + "loss": 0.2983, + "step": 48431 + }, + { + "epoch": 1.86, + "learning_rate": 1.8977943062323676e-05, + "loss": 0.2318, + "step": 48432 + }, + { + "epoch": 1.86, + "learning_rate": 1.8977301872274944e-05, + "loss": 0.2301, + "step": 48433 + }, + { + "epoch": 1.86, + "learning_rate": 1.8976660682226212e-05, + "loss": 0.2629, + "step": 48434 + }, + { + "epoch": 1.86, + "learning_rate": 1.8976019492177484e-05, + "loss": 0.3164, + "step": 48435 + }, + { + "epoch": 1.86, + "learning_rate": 1.8975378302128752e-05, + "loss": 0.2968, + "step": 48436 + }, + { + "epoch": 1.86, + "learning_rate": 1.897473711208002e-05, + "loss": 0.2689, + "step": 48437 + }, + { + "epoch": 1.86, + "learning_rate": 1.897409592203129e-05, + "loss": 0.3532, + "step": 48438 + }, + { + "epoch": 1.86, + "learning_rate": 1.897345473198256e-05, + "loss": 0.232, + "step": 48439 + }, + { + "epoch": 1.86, + "learning_rate": 1.897281354193383e-05, + "loss": 0.3186, + "step": 48440 + }, + { + "epoch": 1.86, + "learning_rate": 1.89721723518851e-05, + "loss": 0.4265, + "step": 48441 + }, + { + "epoch": 1.86, + "learning_rate": 1.897153116183637e-05, + "loss": 0.2998, + "step": 48442 + }, + { + "epoch": 1.86, + "learning_rate": 1.897088997178764e-05, + "loss": 0.3494, + "step": 48443 + }, + { + "epoch": 1.86, + "learning_rate": 1.897024878173891e-05, + "loss": 0.2534, + "step": 48444 + }, + { + "epoch": 1.86, + "learning_rate": 1.8969607591690177e-05, + "loss": 0.2523, + "step": 48445 + }, + { + "epoch": 1.86, + "learning_rate": 1.8968966401641446e-05, + "loss": 0.3959, + "step": 48446 + }, + { + "epoch": 1.86, + "learning_rate": 1.8968325211592717e-05, + "loss": 0.3698, + "step": 48447 + }, + { + "epoch": 1.86, + "learning_rate": 1.8967684021543986e-05, + "loss": 0.2118, + "step": 48448 + }, + { + "epoch": 1.86, + "learning_rate": 1.8967042831495254e-05, + "loss": 0.5473, + "step": 48449 + }, + { + "epoch": 1.86, + "learning_rate": 1.8966401641446526e-05, + "loss": 0.4784, + "step": 48450 + }, + { + "epoch": 1.86, + "learning_rate": 1.8965760451397798e-05, + "loss": 0.2787, + "step": 48451 + }, + { + "epoch": 1.86, + "learning_rate": 1.8965119261349066e-05, + "loss": 0.348, + "step": 48452 + }, + { + "epoch": 1.86, + "learning_rate": 1.8964478071300334e-05, + "loss": 0.2602, + "step": 48453 + }, + { + "epoch": 1.86, + "learning_rate": 1.8963836881251603e-05, + "loss": 0.4649, + "step": 48454 + }, + { + "epoch": 1.86, + "learning_rate": 1.8963195691202874e-05, + "loss": 0.5161, + "step": 48455 + }, + { + "epoch": 1.86, + "learning_rate": 1.8962554501154143e-05, + "loss": 0.3608, + "step": 48456 + }, + { + "epoch": 1.86, + "learning_rate": 1.896191331110541e-05, + "loss": 0.3661, + "step": 48457 + }, + { + "epoch": 1.86, + "learning_rate": 1.8961272121056683e-05, + "loss": 0.2348, + "step": 48458 + }, + { + "epoch": 1.86, + "learning_rate": 1.8960630931007954e-05, + "loss": 0.2959, + "step": 48459 + }, + { + "epoch": 1.86, + "learning_rate": 1.8959989740959223e-05, + "loss": 0.276, + "step": 48460 + }, + { + "epoch": 1.86, + "learning_rate": 1.895934855091049e-05, + "loss": 0.4427, + "step": 48461 + }, + { + "epoch": 1.86, + "learning_rate": 1.895870736086176e-05, + "loss": 0.3232, + "step": 48462 + }, + { + "epoch": 1.86, + "learning_rate": 1.895806617081303e-05, + "loss": 0.2986, + "step": 48463 + }, + { + "epoch": 1.86, + "learning_rate": 1.89574249807643e-05, + "loss": 0.3035, + "step": 48464 + }, + { + "epoch": 1.86, + "learning_rate": 1.8956783790715568e-05, + "loss": 0.3312, + "step": 48465 + }, + { + "epoch": 1.86, + "learning_rate": 1.8956142600666836e-05, + "loss": 0.3128, + "step": 48466 + }, + { + "epoch": 1.86, + "learning_rate": 1.895550141061811e-05, + "loss": 0.2814, + "step": 48467 + }, + { + "epoch": 1.86, + "learning_rate": 1.895486022056938e-05, + "loss": 0.409, + "step": 48468 + }, + { + "epoch": 1.86, + "learning_rate": 1.8954219030520648e-05, + "loss": 0.2992, + "step": 48469 + }, + { + "epoch": 1.86, + "learning_rate": 1.8953577840471916e-05, + "loss": 0.385, + "step": 48470 + }, + { + "epoch": 1.86, + "learning_rate": 1.8952936650423188e-05, + "loss": 0.2954, + "step": 48471 + }, + { + "epoch": 1.86, + "learning_rate": 1.8952295460374456e-05, + "loss": 0.3293, + "step": 48472 + }, + { + "epoch": 1.86, + "learning_rate": 1.8951654270325724e-05, + "loss": 0.3299, + "step": 48473 + }, + { + "epoch": 1.86, + "learning_rate": 1.8951013080276993e-05, + "loss": 0.3646, + "step": 48474 + }, + { + "epoch": 1.86, + "learning_rate": 1.8950371890228265e-05, + "loss": 0.3421, + "step": 48475 + }, + { + "epoch": 1.86, + "learning_rate": 1.8949730700179536e-05, + "loss": 0.1756, + "step": 48476 + }, + { + "epoch": 1.86, + "learning_rate": 1.8949089510130805e-05, + "loss": 0.2534, + "step": 48477 + }, + { + "epoch": 1.86, + "learning_rate": 1.8948448320082073e-05, + "loss": 0.3401, + "step": 48478 + }, + { + "epoch": 1.86, + "learning_rate": 1.8947807130033345e-05, + "loss": 0.3127, + "step": 48479 + }, + { + "epoch": 1.86, + "learning_rate": 1.8947165939984613e-05, + "loss": 0.2729, + "step": 48480 + }, + { + "epoch": 1.86, + "learning_rate": 1.894652474993588e-05, + "loss": 0.2206, + "step": 48481 + }, + { + "epoch": 1.86, + "learning_rate": 1.894588355988715e-05, + "loss": 0.337, + "step": 48482 + }, + { + "epoch": 1.86, + "learning_rate": 1.894524236983842e-05, + "loss": 0.3755, + "step": 48483 + }, + { + "epoch": 1.86, + "learning_rate": 1.894460117978969e-05, + "loss": 0.2608, + "step": 48484 + }, + { + "epoch": 1.86, + "learning_rate": 1.894395998974096e-05, + "loss": 0.4837, + "step": 48485 + }, + { + "epoch": 1.86, + "learning_rate": 1.894331879969223e-05, + "loss": 0.4805, + "step": 48486 + }, + { + "epoch": 1.86, + "learning_rate": 1.89426776096435e-05, + "loss": 0.324, + "step": 48487 + }, + { + "epoch": 1.86, + "learning_rate": 1.894203641959477e-05, + "loss": 0.2106, + "step": 48488 + }, + { + "epoch": 1.86, + "learning_rate": 1.8941395229546038e-05, + "loss": 0.2319, + "step": 48489 + }, + { + "epoch": 1.86, + "learning_rate": 1.8940754039497306e-05, + "loss": 0.3042, + "step": 48490 + }, + { + "epoch": 1.86, + "learning_rate": 1.8940112849448578e-05, + "loss": 0.3465, + "step": 48491 + }, + { + "epoch": 1.86, + "learning_rate": 1.8939471659399846e-05, + "loss": 0.2306, + "step": 48492 + }, + { + "epoch": 1.86, + "learning_rate": 1.8938830469351115e-05, + "loss": 0.1861, + "step": 48493 + }, + { + "epoch": 1.86, + "learning_rate": 1.8938189279302386e-05, + "loss": 0.2113, + "step": 48494 + }, + { + "epoch": 1.86, + "learning_rate": 1.8937548089253658e-05, + "loss": 0.2197, + "step": 48495 + }, + { + "epoch": 1.86, + "learning_rate": 1.8936906899204927e-05, + "loss": 0.3249, + "step": 48496 + }, + { + "epoch": 1.86, + "learning_rate": 1.8936265709156195e-05, + "loss": 0.5235, + "step": 48497 + }, + { + "epoch": 1.86, + "learning_rate": 1.8935624519107463e-05, + "loss": 0.286, + "step": 48498 + }, + { + "epoch": 1.86, + "learning_rate": 1.8934983329058735e-05, + "loss": 0.3183, + "step": 48499 + }, + { + "epoch": 1.86, + "learning_rate": 1.8934342139010003e-05, + "loss": 0.2382, + "step": 48500 + }, + { + "epoch": 1.86, + "learning_rate": 1.893370094896127e-05, + "loss": 0.4173, + "step": 48501 + }, + { + "epoch": 1.86, + "learning_rate": 1.893305975891254e-05, + "loss": 0.2522, + "step": 48502 + }, + { + "epoch": 1.86, + "learning_rate": 1.8932418568863815e-05, + "loss": 0.3766, + "step": 48503 + }, + { + "epoch": 1.86, + "learning_rate": 1.8931777378815083e-05, + "loss": 0.3077, + "step": 48504 + }, + { + "epoch": 1.86, + "learning_rate": 1.893113618876635e-05, + "loss": 0.2841, + "step": 48505 + }, + { + "epoch": 1.86, + "learning_rate": 1.893049499871762e-05, + "loss": 0.3374, + "step": 48506 + }, + { + "epoch": 1.86, + "learning_rate": 1.8929853808668892e-05, + "loss": 0.2265, + "step": 48507 + }, + { + "epoch": 1.86, + "learning_rate": 1.892921261862016e-05, + "loss": 0.2732, + "step": 48508 + }, + { + "epoch": 1.87, + "learning_rate": 1.892857142857143e-05, + "loss": 0.4325, + "step": 48509 + }, + { + "epoch": 1.87, + "learning_rate": 1.8927930238522697e-05, + "loss": 0.2647, + "step": 48510 + }, + { + "epoch": 1.87, + "learning_rate": 1.892728904847397e-05, + "loss": 0.4054, + "step": 48511 + }, + { + "epoch": 1.87, + "learning_rate": 1.892664785842524e-05, + "loss": 0.2305, + "step": 48512 + }, + { + "epoch": 1.87, + "learning_rate": 1.892600666837651e-05, + "loss": 0.299, + "step": 48513 + }, + { + "epoch": 1.87, + "learning_rate": 1.8925365478327777e-05, + "loss": 0.5337, + "step": 48514 + }, + { + "epoch": 1.87, + "learning_rate": 1.892472428827905e-05, + "loss": 0.3754, + "step": 48515 + }, + { + "epoch": 1.87, + "learning_rate": 1.8924083098230317e-05, + "loss": 0.2263, + "step": 48516 + }, + { + "epoch": 1.87, + "learning_rate": 1.8923441908181585e-05, + "loss": 0.2737, + "step": 48517 + }, + { + "epoch": 1.87, + "learning_rate": 1.8922800718132853e-05, + "loss": 0.3949, + "step": 48518 + }, + { + "epoch": 1.87, + "learning_rate": 1.8922159528084125e-05, + "loss": 0.3224, + "step": 48519 + }, + { + "epoch": 1.87, + "learning_rate": 1.8921518338035394e-05, + "loss": 0.2361, + "step": 48520 + }, + { + "epoch": 1.87, + "learning_rate": 1.8920877147986665e-05, + "loss": 0.4045, + "step": 48521 + }, + { + "epoch": 1.87, + "learning_rate": 1.8920235957937934e-05, + "loss": 0.4226, + "step": 48522 + }, + { + "epoch": 1.87, + "learning_rate": 1.8919594767889205e-05, + "loss": 0.3643, + "step": 48523 + }, + { + "epoch": 1.87, + "learning_rate": 1.8918953577840474e-05, + "loss": 0.3586, + "step": 48524 + }, + { + "epoch": 1.87, + "learning_rate": 1.8918312387791742e-05, + "loss": 0.2229, + "step": 48525 + }, + { + "epoch": 1.87, + "learning_rate": 1.891767119774301e-05, + "loss": 0.4046, + "step": 48526 + }, + { + "epoch": 1.87, + "learning_rate": 1.8917030007694282e-05, + "loss": 0.2857, + "step": 48527 + }, + { + "epoch": 1.87, + "learning_rate": 1.891638881764555e-05, + "loss": 0.2355, + "step": 48528 + }, + { + "epoch": 1.87, + "learning_rate": 1.891574762759682e-05, + "loss": 0.4877, + "step": 48529 + }, + { + "epoch": 1.87, + "learning_rate": 1.891510643754809e-05, + "loss": 0.2564, + "step": 48530 + }, + { + "epoch": 1.87, + "learning_rate": 1.8914465247499362e-05, + "loss": 0.2867, + "step": 48531 + }, + { + "epoch": 1.87, + "learning_rate": 1.891382405745063e-05, + "loss": 0.4329, + "step": 48532 + }, + { + "epoch": 1.87, + "learning_rate": 1.89131828674019e-05, + "loss": 0.2927, + "step": 48533 + }, + { + "epoch": 1.87, + "learning_rate": 1.8912541677353167e-05, + "loss": 0.3414, + "step": 48534 + }, + { + "epoch": 1.87, + "learning_rate": 1.891190048730444e-05, + "loss": 0.4236, + "step": 48535 + }, + { + "epoch": 1.87, + "learning_rate": 1.8911259297255707e-05, + "loss": 0.404, + "step": 48536 + }, + { + "epoch": 1.87, + "learning_rate": 1.8910618107206975e-05, + "loss": 0.2292, + "step": 48537 + }, + { + "epoch": 1.87, + "learning_rate": 1.8909976917158244e-05, + "loss": 0.2828, + "step": 48538 + }, + { + "epoch": 1.87, + "learning_rate": 1.890933572710952e-05, + "loss": 0.187, + "step": 48539 + }, + { + "epoch": 1.87, + "learning_rate": 1.8908694537060787e-05, + "loss": 0.3594, + "step": 48540 + }, + { + "epoch": 1.87, + "learning_rate": 1.8908053347012056e-05, + "loss": 0.3709, + "step": 48541 + }, + { + "epoch": 1.87, + "learning_rate": 1.8907412156963324e-05, + "loss": 0.3412, + "step": 48542 + }, + { + "epoch": 1.87, + "learning_rate": 1.8906770966914596e-05, + "loss": 0.245, + "step": 48543 + }, + { + "epoch": 1.87, + "learning_rate": 1.8906129776865864e-05, + "loss": 0.4423, + "step": 48544 + }, + { + "epoch": 1.87, + "learning_rate": 1.8905488586817132e-05, + "loss": 0.3369, + "step": 48545 + }, + { + "epoch": 1.87, + "learning_rate": 1.89048473967684e-05, + "loss": 0.3819, + "step": 48546 + }, + { + "epoch": 1.87, + "learning_rate": 1.8904206206719672e-05, + "loss": 0.435, + "step": 48547 + }, + { + "epoch": 1.87, + "learning_rate": 1.8903565016670944e-05, + "loss": 0.3794, + "step": 48548 + }, + { + "epoch": 1.87, + "learning_rate": 1.8902923826622212e-05, + "loss": 0.2921, + "step": 48549 + }, + { + "epoch": 1.87, + "learning_rate": 1.890228263657348e-05, + "loss": 0.3384, + "step": 48550 + }, + { + "epoch": 1.87, + "learning_rate": 1.8901641446524752e-05, + "loss": 0.3475, + "step": 48551 + }, + { + "epoch": 1.87, + "learning_rate": 1.890100025647602e-05, + "loss": 0.3241, + "step": 48552 + }, + { + "epoch": 1.87, + "learning_rate": 1.890035906642729e-05, + "loss": 0.3063, + "step": 48553 + }, + { + "epoch": 1.87, + "learning_rate": 1.8899717876378557e-05, + "loss": 0.3163, + "step": 48554 + }, + { + "epoch": 1.87, + "learning_rate": 1.889907668632983e-05, + "loss": 0.265, + "step": 48555 + }, + { + "epoch": 1.87, + "learning_rate": 1.88984354962811e-05, + "loss": 0.42, + "step": 48556 + }, + { + "epoch": 1.87, + "learning_rate": 1.889779430623237e-05, + "loss": 0.3151, + "step": 48557 + }, + { + "epoch": 1.87, + "learning_rate": 1.8897153116183637e-05, + "loss": 0.2123, + "step": 48558 + }, + { + "epoch": 1.87, + "learning_rate": 1.889651192613491e-05, + "loss": 0.2442, + "step": 48559 + }, + { + "epoch": 1.87, + "learning_rate": 1.8895870736086177e-05, + "loss": 0.2496, + "step": 48560 + }, + { + "epoch": 1.87, + "learning_rate": 1.8895229546037446e-05, + "loss": 0.347, + "step": 48561 + }, + { + "epoch": 1.87, + "learning_rate": 1.8894588355988714e-05, + "loss": 0.2363, + "step": 48562 + }, + { + "epoch": 1.87, + "learning_rate": 1.8893947165939986e-05, + "loss": 0.4137, + "step": 48563 + }, + { + "epoch": 1.87, + "learning_rate": 1.8893305975891254e-05, + "loss": 0.3872, + "step": 48564 + }, + { + "epoch": 1.87, + "learning_rate": 1.8892664785842526e-05, + "loss": 0.3664, + "step": 48565 + }, + { + "epoch": 1.87, + "learning_rate": 1.8892023595793794e-05, + "loss": 0.1971, + "step": 48566 + }, + { + "epoch": 1.87, + "learning_rate": 1.8891382405745066e-05, + "loss": 0.3055, + "step": 48567 + }, + { + "epoch": 1.87, + "learning_rate": 1.8890741215696334e-05, + "loss": 0.2458, + "step": 48568 + }, + { + "epoch": 1.87, + "learning_rate": 1.8890100025647603e-05, + "loss": 0.3224, + "step": 48569 + }, + { + "epoch": 1.87, + "learning_rate": 1.888945883559887e-05, + "loss": 0.2085, + "step": 48570 + }, + { + "epoch": 1.87, + "learning_rate": 1.8888817645550143e-05, + "loss": 0.2087, + "step": 48571 + }, + { + "epoch": 1.87, + "learning_rate": 1.888817645550141e-05, + "loss": 0.3371, + "step": 48572 + }, + { + "epoch": 1.87, + "learning_rate": 1.888753526545268e-05, + "loss": 0.2923, + "step": 48573 + }, + { + "epoch": 1.87, + "learning_rate": 1.888689407540395e-05, + "loss": 0.4541, + "step": 48574 + }, + { + "epoch": 1.87, + "learning_rate": 1.8886252885355223e-05, + "loss": 0.2515, + "step": 48575 + }, + { + "epoch": 1.87, + "learning_rate": 1.888561169530649e-05, + "loss": 0.3224, + "step": 48576 + }, + { + "epoch": 1.87, + "learning_rate": 1.888497050525776e-05, + "loss": 0.3302, + "step": 48577 + }, + { + "epoch": 1.87, + "learning_rate": 1.8884329315209028e-05, + "loss": 0.2788, + "step": 48578 + }, + { + "epoch": 1.87, + "learning_rate": 1.88836881251603e-05, + "loss": 0.4189, + "step": 48579 + }, + { + "epoch": 1.87, + "learning_rate": 1.8883046935111568e-05, + "loss": 0.3201, + "step": 48580 + }, + { + "epoch": 1.87, + "learning_rate": 1.8882405745062836e-05, + "loss": 0.3983, + "step": 48581 + }, + { + "epoch": 1.87, + "learning_rate": 1.8881764555014104e-05, + "loss": 0.2611, + "step": 48582 + }, + { + "epoch": 1.87, + "learning_rate": 1.888112336496538e-05, + "loss": 0.3308, + "step": 48583 + }, + { + "epoch": 1.87, + "learning_rate": 1.8880482174916648e-05, + "loss": 0.2659, + "step": 48584 + }, + { + "epoch": 1.87, + "learning_rate": 1.8879840984867916e-05, + "loss": 0.3857, + "step": 48585 + }, + { + "epoch": 1.87, + "learning_rate": 1.8879199794819185e-05, + "loss": 0.442, + "step": 48586 + }, + { + "epoch": 1.87, + "learning_rate": 1.8878558604770456e-05, + "loss": 0.2591, + "step": 48587 + }, + { + "epoch": 1.87, + "learning_rate": 1.8877917414721725e-05, + "loss": 0.3771, + "step": 48588 + }, + { + "epoch": 1.87, + "learning_rate": 1.8877276224672993e-05, + "loss": 0.3075, + "step": 48589 + }, + { + "epoch": 1.87, + "learning_rate": 1.887663503462426e-05, + "loss": 0.2946, + "step": 48590 + }, + { + "epoch": 1.87, + "learning_rate": 1.8875993844575533e-05, + "loss": 0.3051, + "step": 48591 + }, + { + "epoch": 1.87, + "learning_rate": 1.8875352654526805e-05, + "loss": 0.3034, + "step": 48592 + }, + { + "epoch": 1.87, + "learning_rate": 1.8874711464478073e-05, + "loss": 0.2641, + "step": 48593 + }, + { + "epoch": 1.87, + "learning_rate": 1.887407027442934e-05, + "loss": 0.3488, + "step": 48594 + }, + { + "epoch": 1.87, + "learning_rate": 1.8873429084380613e-05, + "loss": 0.2824, + "step": 48595 + }, + { + "epoch": 1.87, + "learning_rate": 1.887278789433188e-05, + "loss": 0.3166, + "step": 48596 + }, + { + "epoch": 1.87, + "learning_rate": 1.887214670428315e-05, + "loss": 0.2383, + "step": 48597 + }, + { + "epoch": 1.87, + "learning_rate": 1.8871505514234418e-05, + "loss": 0.28, + "step": 48598 + }, + { + "epoch": 1.87, + "learning_rate": 1.887086432418569e-05, + "loss": 0.2761, + "step": 48599 + }, + { + "epoch": 1.87, + "learning_rate": 1.8870223134136958e-05, + "loss": 0.3742, + "step": 48600 + }, + { + "epoch": 1.87, + "learning_rate": 1.886958194408823e-05, + "loss": 0.2891, + "step": 48601 + }, + { + "epoch": 1.87, + "learning_rate": 1.8868940754039498e-05, + "loss": 0.3238, + "step": 48602 + }, + { + "epoch": 1.87, + "learning_rate": 1.886829956399077e-05, + "loss": 0.3055, + "step": 48603 + }, + { + "epoch": 1.87, + "learning_rate": 1.8867658373942038e-05, + "loss": 0.3668, + "step": 48604 + }, + { + "epoch": 1.87, + "learning_rate": 1.8867017183893306e-05, + "loss": 0.3741, + "step": 48605 + }, + { + "epoch": 1.87, + "learning_rate": 1.8866375993844575e-05, + "loss": 0.2704, + "step": 48606 + }, + { + "epoch": 1.87, + "learning_rate": 1.8865734803795847e-05, + "loss": 0.3246, + "step": 48607 + }, + { + "epoch": 1.87, + "learning_rate": 1.8865093613747115e-05, + "loss": 0.3421, + "step": 48608 + }, + { + "epoch": 1.87, + "learning_rate": 1.8864452423698383e-05, + "loss": 0.4491, + "step": 48609 + }, + { + "epoch": 1.87, + "learning_rate": 1.8863811233649655e-05, + "loss": 0.2479, + "step": 48610 + }, + { + "epoch": 1.87, + "learning_rate": 1.8863170043600927e-05, + "loss": 0.2915, + "step": 48611 + }, + { + "epoch": 1.87, + "learning_rate": 1.8862528853552195e-05, + "loss": 0.3059, + "step": 48612 + }, + { + "epoch": 1.87, + "learning_rate": 1.8861887663503463e-05, + "loss": 0.33, + "step": 48613 + }, + { + "epoch": 1.87, + "learning_rate": 1.886124647345473e-05, + "loss": 0.2957, + "step": 48614 + }, + { + "epoch": 1.87, + "learning_rate": 1.8860605283406003e-05, + "loss": 0.5492, + "step": 48615 + }, + { + "epoch": 1.87, + "learning_rate": 1.885996409335727e-05, + "loss": 0.2875, + "step": 48616 + }, + { + "epoch": 1.87, + "learning_rate": 1.885932290330854e-05, + "loss": 0.1349, + "step": 48617 + }, + { + "epoch": 1.87, + "learning_rate": 1.885868171325981e-05, + "loss": 0.2572, + "step": 48618 + }, + { + "epoch": 1.87, + "learning_rate": 1.8858040523211083e-05, + "loss": 0.3799, + "step": 48619 + }, + { + "epoch": 1.87, + "learning_rate": 1.8857399333162352e-05, + "loss": 0.1972, + "step": 48620 + }, + { + "epoch": 1.87, + "learning_rate": 1.885675814311362e-05, + "loss": 0.2771, + "step": 48621 + }, + { + "epoch": 1.87, + "learning_rate": 1.885611695306489e-05, + "loss": 0.3636, + "step": 48622 + }, + { + "epoch": 1.87, + "learning_rate": 1.885547576301616e-05, + "loss": 0.229, + "step": 48623 + }, + { + "epoch": 1.87, + "learning_rate": 1.885483457296743e-05, + "loss": 0.2108, + "step": 48624 + }, + { + "epoch": 1.87, + "learning_rate": 1.8854193382918697e-05, + "loss": 0.4502, + "step": 48625 + }, + { + "epoch": 1.87, + "learning_rate": 1.8853552192869965e-05, + "loss": 0.3179, + "step": 48626 + }, + { + "epoch": 1.87, + "learning_rate": 1.8852911002821237e-05, + "loss": 0.263, + "step": 48627 + }, + { + "epoch": 1.87, + "learning_rate": 1.885226981277251e-05, + "loss": 0.3988, + "step": 48628 + }, + { + "epoch": 1.87, + "learning_rate": 1.8851628622723777e-05, + "loss": 0.3773, + "step": 48629 + }, + { + "epoch": 1.87, + "learning_rate": 1.8850987432675045e-05, + "loss": 0.305, + "step": 48630 + }, + { + "epoch": 1.87, + "learning_rate": 1.8850346242626317e-05, + "loss": 0.3045, + "step": 48631 + }, + { + "epoch": 1.87, + "learning_rate": 1.8849705052577585e-05, + "loss": 0.2821, + "step": 48632 + }, + { + "epoch": 1.87, + "learning_rate": 1.8849063862528854e-05, + "loss": 0.3855, + "step": 48633 + }, + { + "epoch": 1.87, + "learning_rate": 1.8848422672480122e-05, + "loss": 0.257, + "step": 48634 + }, + { + "epoch": 1.87, + "learning_rate": 1.8847781482431394e-05, + "loss": 0.33, + "step": 48635 + }, + { + "epoch": 1.87, + "learning_rate": 1.8847140292382662e-05, + "loss": 0.2215, + "step": 48636 + }, + { + "epoch": 1.87, + "learning_rate": 1.8846499102333934e-05, + "loss": 0.2258, + "step": 48637 + }, + { + "epoch": 1.87, + "learning_rate": 1.8845857912285202e-05, + "loss": 0.3502, + "step": 48638 + }, + { + "epoch": 1.87, + "learning_rate": 1.8845216722236474e-05, + "loss": 0.2843, + "step": 48639 + }, + { + "epoch": 1.87, + "learning_rate": 1.8844575532187742e-05, + "loss": 0.2696, + "step": 48640 + }, + { + "epoch": 1.87, + "learning_rate": 1.884393434213901e-05, + "loss": 0.2984, + "step": 48641 + }, + { + "epoch": 1.87, + "learning_rate": 1.884329315209028e-05, + "loss": 0.2813, + "step": 48642 + }, + { + "epoch": 1.87, + "learning_rate": 1.884265196204155e-05, + "loss": 0.3297, + "step": 48643 + }, + { + "epoch": 1.87, + "learning_rate": 1.884201077199282e-05, + "loss": 0.2847, + "step": 48644 + }, + { + "epoch": 1.87, + "learning_rate": 1.8841369581944087e-05, + "loss": 0.2288, + "step": 48645 + }, + { + "epoch": 1.87, + "learning_rate": 1.884072839189536e-05, + "loss": 0.2721, + "step": 48646 + }, + { + "epoch": 1.87, + "learning_rate": 1.884008720184663e-05, + "loss": 0.3671, + "step": 48647 + }, + { + "epoch": 1.87, + "learning_rate": 1.88394460117979e-05, + "loss": 0.3457, + "step": 48648 + }, + { + "epoch": 1.87, + "learning_rate": 1.8838804821749167e-05, + "loss": 0.377, + "step": 48649 + }, + { + "epoch": 1.87, + "learning_rate": 1.8838163631700435e-05, + "loss": 0.4236, + "step": 48650 + }, + { + "epoch": 1.87, + "learning_rate": 1.8837522441651707e-05, + "loss": 0.2473, + "step": 48651 + }, + { + "epoch": 1.87, + "learning_rate": 1.8836881251602976e-05, + "loss": 0.3618, + "step": 48652 + }, + { + "epoch": 1.87, + "learning_rate": 1.8836240061554244e-05, + "loss": 0.3199, + "step": 48653 + }, + { + "epoch": 1.87, + "learning_rate": 1.8835598871505516e-05, + "loss": 0.286, + "step": 48654 + }, + { + "epoch": 1.87, + "learning_rate": 1.8834957681456787e-05, + "loss": 0.5492, + "step": 48655 + }, + { + "epoch": 1.87, + "learning_rate": 1.8834316491408056e-05, + "loss": 0.3784, + "step": 48656 + }, + { + "epoch": 1.87, + "learning_rate": 1.8833675301359324e-05, + "loss": 0.3501, + "step": 48657 + }, + { + "epoch": 1.87, + "learning_rate": 1.8833034111310592e-05, + "loss": 0.3269, + "step": 48658 + }, + { + "epoch": 1.87, + "learning_rate": 1.8832392921261864e-05, + "loss": 0.2913, + "step": 48659 + }, + { + "epoch": 1.87, + "learning_rate": 1.8831751731213132e-05, + "loss": 0.3572, + "step": 48660 + }, + { + "epoch": 1.87, + "learning_rate": 1.88311105411644e-05, + "loss": 0.293, + "step": 48661 + }, + { + "epoch": 1.87, + "learning_rate": 1.883046935111567e-05, + "loss": 0.3816, + "step": 48662 + }, + { + "epoch": 1.87, + "learning_rate": 1.8829828161066944e-05, + "loss": 0.281, + "step": 48663 + }, + { + "epoch": 1.87, + "learning_rate": 1.8829186971018212e-05, + "loss": 0.3249, + "step": 48664 + }, + { + "epoch": 1.87, + "learning_rate": 1.882854578096948e-05, + "loss": 0.2704, + "step": 48665 + }, + { + "epoch": 1.87, + "learning_rate": 1.882790459092075e-05, + "loss": 0.3824, + "step": 48666 + }, + { + "epoch": 1.87, + "learning_rate": 1.882726340087202e-05, + "loss": 0.3709, + "step": 48667 + }, + { + "epoch": 1.87, + "learning_rate": 1.882662221082329e-05, + "loss": 0.3829, + "step": 48668 + }, + { + "epoch": 1.87, + "learning_rate": 1.8825981020774557e-05, + "loss": 0.2549, + "step": 48669 + }, + { + "epoch": 1.87, + "learning_rate": 1.8825339830725826e-05, + "loss": 0.3388, + "step": 48670 + }, + { + "epoch": 1.87, + "learning_rate": 1.8824698640677097e-05, + "loss": 0.3961, + "step": 48671 + }, + { + "epoch": 1.87, + "learning_rate": 1.882405745062837e-05, + "loss": 0.3177, + "step": 48672 + }, + { + "epoch": 1.87, + "learning_rate": 1.8823416260579638e-05, + "loss": 0.4393, + "step": 48673 + }, + { + "epoch": 1.87, + "learning_rate": 1.8822775070530906e-05, + "loss": 0.3219, + "step": 48674 + }, + { + "epoch": 1.87, + "learning_rate": 1.8822133880482178e-05, + "loss": 0.2598, + "step": 48675 + }, + { + "epoch": 1.87, + "learning_rate": 1.8821492690433446e-05, + "loss": 0.4244, + "step": 48676 + }, + { + "epoch": 1.87, + "learning_rate": 1.8820851500384714e-05, + "loss": 0.3962, + "step": 48677 + }, + { + "epoch": 1.87, + "learning_rate": 1.8820210310335983e-05, + "loss": 0.3813, + "step": 48678 + }, + { + "epoch": 1.87, + "learning_rate": 1.8819569120287254e-05, + "loss": 0.2652, + "step": 48679 + }, + { + "epoch": 1.87, + "learning_rate": 1.8818927930238523e-05, + "loss": 0.248, + "step": 48680 + }, + { + "epoch": 1.87, + "learning_rate": 1.8818286740189794e-05, + "loss": 0.2928, + "step": 48681 + }, + { + "epoch": 1.87, + "learning_rate": 1.8817645550141063e-05, + "loss": 0.4065, + "step": 48682 + }, + { + "epoch": 1.87, + "learning_rate": 1.8817004360092334e-05, + "loss": 0.3205, + "step": 48683 + }, + { + "epoch": 1.87, + "learning_rate": 1.8816363170043603e-05, + "loss": 0.2676, + "step": 48684 + }, + { + "epoch": 1.87, + "learning_rate": 1.881572197999487e-05, + "loss": 0.3728, + "step": 48685 + }, + { + "epoch": 1.87, + "learning_rate": 1.881508078994614e-05, + "loss": 0.2767, + "step": 48686 + }, + { + "epoch": 1.87, + "learning_rate": 1.881443959989741e-05, + "loss": 0.3213, + "step": 48687 + }, + { + "epoch": 1.87, + "learning_rate": 1.881379840984868e-05, + "loss": 0.3473, + "step": 48688 + }, + { + "epoch": 1.87, + "learning_rate": 1.8813157219799948e-05, + "loss": 0.394, + "step": 48689 + }, + { + "epoch": 1.87, + "learning_rate": 1.881251602975122e-05, + "loss": 0.5565, + "step": 48690 + }, + { + "epoch": 1.87, + "learning_rate": 1.881187483970249e-05, + "loss": 0.303, + "step": 48691 + }, + { + "epoch": 1.87, + "learning_rate": 1.881123364965376e-05, + "loss": 0.3416, + "step": 48692 + }, + { + "epoch": 1.87, + "learning_rate": 1.8810592459605028e-05, + "loss": 0.3113, + "step": 48693 + }, + { + "epoch": 1.87, + "learning_rate": 1.8809951269556296e-05, + "loss": 0.4018, + "step": 48694 + }, + { + "epoch": 1.87, + "learning_rate": 1.8809310079507568e-05, + "loss": 0.2734, + "step": 48695 + }, + { + "epoch": 1.87, + "learning_rate": 1.8808668889458836e-05, + "loss": 0.3166, + "step": 48696 + }, + { + "epoch": 1.87, + "learning_rate": 1.8808027699410105e-05, + "loss": 0.2881, + "step": 48697 + }, + { + "epoch": 1.87, + "learning_rate": 1.8807386509361373e-05, + "loss": 0.41, + "step": 48698 + }, + { + "epoch": 1.87, + "learning_rate": 1.8806745319312648e-05, + "loss": 0.3607, + "step": 48699 + }, + { + "epoch": 1.87, + "learning_rate": 1.8806104129263916e-05, + "loss": 0.3942, + "step": 48700 + }, + { + "epoch": 1.87, + "learning_rate": 1.8805462939215185e-05, + "loss": 0.3651, + "step": 48701 + }, + { + "epoch": 1.87, + "learning_rate": 1.8804821749166453e-05, + "loss": 0.3024, + "step": 48702 + }, + { + "epoch": 1.87, + "learning_rate": 1.8804180559117725e-05, + "loss": 0.3415, + "step": 48703 + }, + { + "epoch": 1.87, + "learning_rate": 1.8803539369068993e-05, + "loss": 0.3019, + "step": 48704 + }, + { + "epoch": 1.87, + "learning_rate": 1.880289817902026e-05, + "loss": 0.3841, + "step": 48705 + }, + { + "epoch": 1.87, + "learning_rate": 1.880225698897153e-05, + "loss": 0.2487, + "step": 48706 + }, + { + "epoch": 1.87, + "learning_rate": 1.88016157989228e-05, + "loss": 0.3353, + "step": 48707 + }, + { + "epoch": 1.87, + "learning_rate": 1.8800974608874073e-05, + "loss": 0.3427, + "step": 48708 + }, + { + "epoch": 1.87, + "learning_rate": 1.880033341882534e-05, + "loss": 0.234, + "step": 48709 + }, + { + "epoch": 1.87, + "learning_rate": 1.879969222877661e-05, + "loss": 0.4142, + "step": 48710 + }, + { + "epoch": 1.87, + "learning_rate": 1.879905103872788e-05, + "loss": 0.3648, + "step": 48711 + }, + { + "epoch": 1.87, + "learning_rate": 1.879840984867915e-05, + "loss": 0.2226, + "step": 48712 + }, + { + "epoch": 1.87, + "learning_rate": 1.8797768658630418e-05, + "loss": 0.447, + "step": 48713 + }, + { + "epoch": 1.87, + "learning_rate": 1.8797127468581686e-05, + "loss": 0.3022, + "step": 48714 + }, + { + "epoch": 1.87, + "learning_rate": 1.8796486278532958e-05, + "loss": 0.286, + "step": 48715 + }, + { + "epoch": 1.87, + "learning_rate": 1.8795845088484226e-05, + "loss": 0.2813, + "step": 48716 + }, + { + "epoch": 1.87, + "learning_rate": 1.8795203898435498e-05, + "loss": 0.2703, + "step": 48717 + }, + { + "epoch": 1.87, + "learning_rate": 1.8794562708386767e-05, + "loss": 0.413, + "step": 48718 + }, + { + "epoch": 1.87, + "learning_rate": 1.8793921518338038e-05, + "loss": 0.2351, + "step": 48719 + }, + { + "epoch": 1.87, + "learning_rate": 1.8793280328289307e-05, + "loss": 0.4397, + "step": 48720 + }, + { + "epoch": 1.87, + "learning_rate": 1.8792639138240575e-05, + "loss": 0.444, + "step": 48721 + }, + { + "epoch": 1.87, + "learning_rate": 1.8791997948191843e-05, + "loss": 0.4357, + "step": 48722 + }, + { + "epoch": 1.87, + "learning_rate": 1.8791356758143115e-05, + "loss": 0.286, + "step": 48723 + }, + { + "epoch": 1.87, + "learning_rate": 1.8790715568094383e-05, + "loss": 0.2534, + "step": 48724 + }, + { + "epoch": 1.87, + "learning_rate": 1.879007437804565e-05, + "loss": 0.3117, + "step": 48725 + }, + { + "epoch": 1.87, + "learning_rate": 1.8789433187996923e-05, + "loss": 0.3066, + "step": 48726 + }, + { + "epoch": 1.87, + "learning_rate": 1.8788791997948195e-05, + "loss": 0.2677, + "step": 48727 + }, + { + "epoch": 1.87, + "learning_rate": 1.8788150807899463e-05, + "loss": 0.2894, + "step": 48728 + }, + { + "epoch": 1.87, + "learning_rate": 1.878750961785073e-05, + "loss": 0.3705, + "step": 48729 + }, + { + "epoch": 1.87, + "learning_rate": 1.8786868427802e-05, + "loss": 0.3725, + "step": 48730 + }, + { + "epoch": 1.87, + "learning_rate": 1.8786227237753272e-05, + "loss": 0.2503, + "step": 48731 + }, + { + "epoch": 1.87, + "learning_rate": 1.878558604770454e-05, + "loss": 0.2352, + "step": 48732 + }, + { + "epoch": 1.87, + "learning_rate": 1.878494485765581e-05, + "loss": 0.4072, + "step": 48733 + }, + { + "epoch": 1.87, + "learning_rate": 1.8784303667607077e-05, + "loss": 0.2676, + "step": 48734 + }, + { + "epoch": 1.87, + "learning_rate": 1.8783662477558352e-05, + "loss": 0.3114, + "step": 48735 + }, + { + "epoch": 1.87, + "learning_rate": 1.878302128750962e-05, + "loss": 0.3157, + "step": 48736 + }, + { + "epoch": 1.87, + "learning_rate": 1.878238009746089e-05, + "loss": 0.3385, + "step": 48737 + }, + { + "epoch": 1.87, + "learning_rate": 1.8781738907412157e-05, + "loss": 0.4187, + "step": 48738 + }, + { + "epoch": 1.87, + "learning_rate": 1.878109771736343e-05, + "loss": 0.2574, + "step": 48739 + }, + { + "epoch": 1.87, + "learning_rate": 1.8780456527314697e-05, + "loss": 0.3292, + "step": 48740 + }, + { + "epoch": 1.87, + "learning_rate": 1.8779815337265965e-05, + "loss": 0.2666, + "step": 48741 + }, + { + "epoch": 1.87, + "learning_rate": 1.8779174147217234e-05, + "loss": 0.5221, + "step": 48742 + }, + { + "epoch": 1.87, + "learning_rate": 1.877853295716851e-05, + "loss": 0.2411, + "step": 48743 + }, + { + "epoch": 1.87, + "learning_rate": 1.8777891767119777e-05, + "loss": 0.2644, + "step": 48744 + }, + { + "epoch": 1.87, + "learning_rate": 1.8777250577071045e-05, + "loss": 0.3117, + "step": 48745 + }, + { + "epoch": 1.87, + "learning_rate": 1.8776609387022314e-05, + "loss": 0.3548, + "step": 48746 + }, + { + "epoch": 1.87, + "learning_rate": 1.8775968196973585e-05, + "loss": 0.3927, + "step": 48747 + }, + { + "epoch": 1.87, + "learning_rate": 1.8775327006924854e-05, + "loss": 0.2097, + "step": 48748 + }, + { + "epoch": 1.87, + "learning_rate": 1.8774685816876122e-05, + "loss": 0.4234, + "step": 48749 + }, + { + "epoch": 1.87, + "learning_rate": 1.877404462682739e-05, + "loss": 0.2598, + "step": 48750 + }, + { + "epoch": 1.87, + "learning_rate": 1.8773403436778662e-05, + "loss": 0.2601, + "step": 48751 + }, + { + "epoch": 1.87, + "learning_rate": 1.8772762246729934e-05, + "loss": 0.3814, + "step": 48752 + }, + { + "epoch": 1.87, + "learning_rate": 1.8772121056681202e-05, + "loss": 0.2976, + "step": 48753 + }, + { + "epoch": 1.87, + "learning_rate": 1.877147986663247e-05, + "loss": 0.275, + "step": 48754 + }, + { + "epoch": 1.87, + "learning_rate": 1.8770838676583742e-05, + "loss": 0.2354, + "step": 48755 + }, + { + "epoch": 1.87, + "learning_rate": 1.877019748653501e-05, + "loss": 0.3749, + "step": 48756 + }, + { + "epoch": 1.87, + "learning_rate": 1.876955629648628e-05, + "loss": 0.2849, + "step": 48757 + }, + { + "epoch": 1.87, + "learning_rate": 1.8768915106437547e-05, + "loss": 0.3896, + "step": 48758 + }, + { + "epoch": 1.87, + "learning_rate": 1.876827391638882e-05, + "loss": 0.2449, + "step": 48759 + }, + { + "epoch": 1.87, + "learning_rate": 1.8767632726340087e-05, + "loss": 0.4335, + "step": 48760 + }, + { + "epoch": 1.87, + "learning_rate": 1.876699153629136e-05, + "loss": 0.3901, + "step": 48761 + }, + { + "epoch": 1.87, + "learning_rate": 1.8766350346242627e-05, + "loss": 0.3223, + "step": 48762 + }, + { + "epoch": 1.87, + "learning_rate": 1.87657091561939e-05, + "loss": 0.3661, + "step": 48763 + }, + { + "epoch": 1.87, + "learning_rate": 1.8765067966145167e-05, + "loss": 0.3816, + "step": 48764 + }, + { + "epoch": 1.87, + "learning_rate": 1.8764426776096436e-05, + "loss": 0.3963, + "step": 48765 + }, + { + "epoch": 1.87, + "learning_rate": 1.8763785586047704e-05, + "loss": 0.2658, + "step": 48766 + }, + { + "epoch": 1.87, + "learning_rate": 1.8763144395998976e-05, + "loss": 0.2281, + "step": 48767 + }, + { + "epoch": 1.87, + "learning_rate": 1.8762503205950244e-05, + "loss": 0.3863, + "step": 48768 + }, + { + "epoch": 1.88, + "learning_rate": 1.8761862015901512e-05, + "loss": 0.3711, + "step": 48769 + }, + { + "epoch": 1.88, + "learning_rate": 1.8761220825852784e-05, + "loss": 0.2688, + "step": 48770 + }, + { + "epoch": 1.88, + "learning_rate": 1.8760579635804056e-05, + "loss": 0.2913, + "step": 48771 + }, + { + "epoch": 1.88, + "learning_rate": 1.8759938445755324e-05, + "loss": 0.2752, + "step": 48772 + }, + { + "epoch": 1.88, + "learning_rate": 1.8759297255706592e-05, + "loss": 0.3688, + "step": 48773 + }, + { + "epoch": 1.88, + "learning_rate": 1.875865606565786e-05, + "loss": 0.249, + "step": 48774 + }, + { + "epoch": 1.88, + "learning_rate": 1.8758014875609132e-05, + "loss": 0.2862, + "step": 48775 + }, + { + "epoch": 1.88, + "learning_rate": 1.87573736855604e-05, + "loss": 0.2355, + "step": 48776 + }, + { + "epoch": 1.88, + "learning_rate": 1.875673249551167e-05, + "loss": 0.2528, + "step": 48777 + }, + { + "epoch": 1.88, + "learning_rate": 1.8756091305462937e-05, + "loss": 0.3511, + "step": 48778 + }, + { + "epoch": 1.88, + "learning_rate": 1.8755450115414212e-05, + "loss": 0.4349, + "step": 48779 + }, + { + "epoch": 1.88, + "learning_rate": 1.875480892536548e-05, + "loss": 0.2772, + "step": 48780 + }, + { + "epoch": 1.88, + "learning_rate": 1.875416773531675e-05, + "loss": 0.2359, + "step": 48781 + }, + { + "epoch": 1.88, + "learning_rate": 1.8753526545268017e-05, + "loss": 0.2244, + "step": 48782 + }, + { + "epoch": 1.88, + "learning_rate": 1.875288535521929e-05, + "loss": 0.4184, + "step": 48783 + }, + { + "epoch": 1.88, + "learning_rate": 1.8752244165170558e-05, + "loss": 0.2596, + "step": 48784 + }, + { + "epoch": 1.88, + "learning_rate": 1.8751602975121826e-05, + "loss": 0.2914, + "step": 48785 + }, + { + "epoch": 1.88, + "learning_rate": 1.8750961785073094e-05, + "loss": 0.4064, + "step": 48786 + }, + { + "epoch": 1.88, + "learning_rate": 1.8750320595024366e-05, + "loss": 0.3361, + "step": 48787 + }, + { + "epoch": 1.88, + "learning_rate": 1.8749679404975638e-05, + "loss": 0.2863, + "step": 48788 + }, + { + "epoch": 1.88, + "learning_rate": 1.8749038214926906e-05, + "loss": 0.3964, + "step": 48789 + }, + { + "epoch": 1.88, + "learning_rate": 1.8748397024878174e-05, + "loss": 0.3052, + "step": 48790 + }, + { + "epoch": 1.88, + "learning_rate": 1.8747755834829446e-05, + "loss": 0.2312, + "step": 48791 + }, + { + "epoch": 1.88, + "learning_rate": 1.8747114644780714e-05, + "loss": 0.2867, + "step": 48792 + }, + { + "epoch": 1.88, + "learning_rate": 1.8746473454731983e-05, + "loss": 0.2708, + "step": 48793 + }, + { + "epoch": 1.88, + "learning_rate": 1.874583226468325e-05, + "loss": 0.3154, + "step": 48794 + }, + { + "epoch": 1.88, + "learning_rate": 1.8745191074634523e-05, + "loss": 0.3124, + "step": 48795 + }, + { + "epoch": 1.88, + "learning_rate": 1.874454988458579e-05, + "loss": 0.2675, + "step": 48796 + }, + { + "epoch": 1.88, + "learning_rate": 1.8743908694537063e-05, + "loss": 0.4123, + "step": 48797 + }, + { + "epoch": 1.88, + "learning_rate": 1.874326750448833e-05, + "loss": 0.3801, + "step": 48798 + }, + { + "epoch": 1.88, + "learning_rate": 1.8742626314439603e-05, + "loss": 0.2563, + "step": 48799 + }, + { + "epoch": 1.88, + "learning_rate": 1.874198512439087e-05, + "loss": 0.4924, + "step": 48800 + }, + { + "epoch": 1.88, + "learning_rate": 1.874134393434214e-05, + "loss": 0.2515, + "step": 48801 + }, + { + "epoch": 1.88, + "learning_rate": 1.8740702744293408e-05, + "loss": 0.4353, + "step": 48802 + }, + { + "epoch": 1.88, + "learning_rate": 1.874006155424468e-05, + "loss": 0.2854, + "step": 48803 + }, + { + "epoch": 1.88, + "learning_rate": 1.8739420364195948e-05, + "loss": 0.3905, + "step": 48804 + }, + { + "epoch": 1.88, + "learning_rate": 1.8738779174147216e-05, + "loss": 0.2898, + "step": 48805 + }, + { + "epoch": 1.88, + "learning_rate": 1.8738137984098488e-05, + "loss": 0.2256, + "step": 48806 + }, + { + "epoch": 1.88, + "learning_rate": 1.873749679404976e-05, + "loss": 0.3395, + "step": 48807 + }, + { + "epoch": 1.88, + "learning_rate": 1.8736855604001028e-05, + "loss": 0.2537, + "step": 48808 + }, + { + "epoch": 1.88, + "learning_rate": 1.8736214413952296e-05, + "loss": 0.3416, + "step": 48809 + }, + { + "epoch": 1.88, + "learning_rate": 1.8735573223903565e-05, + "loss": 0.4332, + "step": 48810 + }, + { + "epoch": 1.88, + "learning_rate": 1.8734932033854836e-05, + "loss": 0.3019, + "step": 48811 + }, + { + "epoch": 1.88, + "learning_rate": 1.8734290843806105e-05, + "loss": 0.2709, + "step": 48812 + }, + { + "epoch": 1.88, + "learning_rate": 1.8733649653757373e-05, + "loss": 0.2597, + "step": 48813 + }, + { + "epoch": 1.88, + "learning_rate": 1.873300846370864e-05, + "loss": 0.2603, + "step": 48814 + }, + { + "epoch": 1.88, + "learning_rate": 1.8732367273659916e-05, + "loss": 0.2984, + "step": 48815 + }, + { + "epoch": 1.88, + "learning_rate": 1.8731726083611185e-05, + "loss": 0.3071, + "step": 48816 + }, + { + "epoch": 1.88, + "learning_rate": 1.8731084893562453e-05, + "loss": 0.2583, + "step": 48817 + }, + { + "epoch": 1.88, + "learning_rate": 1.873044370351372e-05, + "loss": 0.224, + "step": 48818 + }, + { + "epoch": 1.88, + "learning_rate": 1.8729802513464993e-05, + "loss": 0.3775, + "step": 48819 + }, + { + "epoch": 1.88, + "learning_rate": 1.872916132341626e-05, + "loss": 0.3155, + "step": 48820 + }, + { + "epoch": 1.88, + "learning_rate": 1.872852013336753e-05, + "loss": 0.2084, + "step": 48821 + }, + { + "epoch": 1.88, + "learning_rate": 1.8727878943318798e-05, + "loss": 0.2854, + "step": 48822 + }, + { + "epoch": 1.88, + "learning_rate": 1.872723775327007e-05, + "loss": 0.2983, + "step": 48823 + }, + { + "epoch": 1.88, + "learning_rate": 1.872659656322134e-05, + "loss": 0.2373, + "step": 48824 + }, + { + "epoch": 1.88, + "learning_rate": 1.872595537317261e-05, + "loss": 0.2579, + "step": 48825 + }, + { + "epoch": 1.88, + "learning_rate": 1.8725314183123878e-05, + "loss": 0.302, + "step": 48826 + }, + { + "epoch": 1.88, + "learning_rate": 1.872467299307515e-05, + "loss": 0.3473, + "step": 48827 + }, + { + "epoch": 1.88, + "learning_rate": 1.8724031803026418e-05, + "loss": 0.3793, + "step": 48828 + }, + { + "epoch": 1.88, + "learning_rate": 1.8723390612977687e-05, + "loss": 0.3614, + "step": 48829 + }, + { + "epoch": 1.88, + "learning_rate": 1.8722749422928955e-05, + "loss": 0.3063, + "step": 48830 + }, + { + "epoch": 1.88, + "learning_rate": 1.8722108232880227e-05, + "loss": 0.24, + "step": 48831 + }, + { + "epoch": 1.88, + "learning_rate": 1.8721467042831495e-05, + "loss": 0.3667, + "step": 48832 + }, + { + "epoch": 1.88, + "learning_rate": 1.8720825852782767e-05, + "loss": 0.3053, + "step": 48833 + }, + { + "epoch": 1.88, + "learning_rate": 1.8720184662734035e-05, + "loss": 0.4953, + "step": 48834 + }, + { + "epoch": 1.88, + "learning_rate": 1.8719543472685307e-05, + "loss": 0.2935, + "step": 48835 + }, + { + "epoch": 1.88, + "learning_rate": 1.8718902282636575e-05, + "loss": 0.3157, + "step": 48836 + }, + { + "epoch": 1.88, + "learning_rate": 1.8718261092587843e-05, + "loss": 0.2231, + "step": 48837 + }, + { + "epoch": 1.88, + "learning_rate": 1.871761990253911e-05, + "loss": 0.2637, + "step": 48838 + }, + { + "epoch": 1.88, + "learning_rate": 1.8716978712490383e-05, + "loss": 0.2451, + "step": 48839 + }, + { + "epoch": 1.88, + "learning_rate": 1.871633752244165e-05, + "loss": 0.3384, + "step": 48840 + }, + { + "epoch": 1.88, + "learning_rate": 1.8715696332392923e-05, + "loss": 0.2048, + "step": 48841 + }, + { + "epoch": 1.88, + "learning_rate": 1.8715055142344192e-05, + "loss": 0.3601, + "step": 48842 + }, + { + "epoch": 1.88, + "learning_rate": 1.8714413952295463e-05, + "loss": 0.3416, + "step": 48843 + }, + { + "epoch": 1.88, + "learning_rate": 1.8713772762246732e-05, + "loss": 0.2563, + "step": 48844 + }, + { + "epoch": 1.88, + "learning_rate": 1.8713131572198e-05, + "loss": 0.3186, + "step": 48845 + }, + { + "epoch": 1.88, + "learning_rate": 1.871249038214927e-05, + "loss": 0.332, + "step": 48846 + }, + { + "epoch": 1.88, + "learning_rate": 1.871184919210054e-05, + "loss": 0.3675, + "step": 48847 + }, + { + "epoch": 1.88, + "learning_rate": 1.871120800205181e-05, + "loss": 0.2561, + "step": 48848 + }, + { + "epoch": 1.88, + "learning_rate": 1.8710566812003077e-05, + "loss": 0.3139, + "step": 48849 + }, + { + "epoch": 1.88, + "learning_rate": 1.870992562195435e-05, + "loss": 0.4936, + "step": 48850 + }, + { + "epoch": 1.88, + "learning_rate": 1.870928443190562e-05, + "loss": 0.4511, + "step": 48851 + }, + { + "epoch": 1.88, + "learning_rate": 1.870864324185689e-05, + "loss": 0.4058, + "step": 48852 + }, + { + "epoch": 1.88, + "learning_rate": 1.8708002051808157e-05, + "loss": 0.4061, + "step": 48853 + }, + { + "epoch": 1.88, + "learning_rate": 1.8707360861759425e-05, + "loss": 0.3029, + "step": 48854 + }, + { + "epoch": 1.88, + "learning_rate": 1.8706719671710697e-05, + "loss": 0.394, + "step": 48855 + }, + { + "epoch": 1.88, + "learning_rate": 1.8706078481661965e-05, + "loss": 0.2914, + "step": 48856 + }, + { + "epoch": 1.88, + "learning_rate": 1.8705437291613234e-05, + "loss": 0.3039, + "step": 48857 + }, + { + "epoch": 1.88, + "learning_rate": 1.8704796101564502e-05, + "loss": 0.218, + "step": 48858 + }, + { + "epoch": 1.88, + "learning_rate": 1.8704154911515777e-05, + "loss": 0.3105, + "step": 48859 + }, + { + "epoch": 1.88, + "learning_rate": 1.8703513721467045e-05, + "loss": 0.3254, + "step": 48860 + }, + { + "epoch": 1.88, + "learning_rate": 1.8702872531418314e-05, + "loss": 0.2915, + "step": 48861 + }, + { + "epoch": 1.88, + "learning_rate": 1.8702231341369582e-05, + "loss": 0.3694, + "step": 48862 + }, + { + "epoch": 1.88, + "learning_rate": 1.8701590151320854e-05, + "loss": 0.2676, + "step": 48863 + }, + { + "epoch": 1.88, + "learning_rate": 1.8700948961272122e-05, + "loss": 0.3536, + "step": 48864 + }, + { + "epoch": 1.88, + "learning_rate": 1.870030777122339e-05, + "loss": 0.3035, + "step": 48865 + }, + { + "epoch": 1.88, + "learning_rate": 1.869966658117466e-05, + "loss": 0.2615, + "step": 48866 + }, + { + "epoch": 1.88, + "learning_rate": 1.869902539112593e-05, + "loss": 0.2703, + "step": 48867 + }, + { + "epoch": 1.88, + "learning_rate": 1.8698384201077202e-05, + "loss": 0.339, + "step": 48868 + }, + { + "epoch": 1.88, + "learning_rate": 1.869774301102847e-05, + "loss": 0.2964, + "step": 48869 + }, + { + "epoch": 1.88, + "learning_rate": 1.869710182097974e-05, + "loss": 0.4372, + "step": 48870 + }, + { + "epoch": 1.88, + "learning_rate": 1.869646063093101e-05, + "loss": 0.2764, + "step": 48871 + }, + { + "epoch": 1.88, + "learning_rate": 1.869581944088228e-05, + "loss": 0.2736, + "step": 48872 + }, + { + "epoch": 1.88, + "learning_rate": 1.8695178250833547e-05, + "loss": 0.3836, + "step": 48873 + }, + { + "epoch": 1.88, + "learning_rate": 1.8694537060784815e-05, + "loss": 0.344, + "step": 48874 + }, + { + "epoch": 1.88, + "learning_rate": 1.8693895870736087e-05, + "loss": 0.2721, + "step": 48875 + }, + { + "epoch": 1.88, + "learning_rate": 1.8693254680687356e-05, + "loss": 0.2551, + "step": 48876 + }, + { + "epoch": 1.88, + "learning_rate": 1.8692613490638627e-05, + "loss": 0.2809, + "step": 48877 + }, + { + "epoch": 1.88, + "learning_rate": 1.8691972300589896e-05, + "loss": 0.398, + "step": 48878 + }, + { + "epoch": 1.88, + "learning_rate": 1.8691331110541167e-05, + "loss": 0.3939, + "step": 48879 + }, + { + "epoch": 1.88, + "learning_rate": 1.8690689920492436e-05, + "loss": 0.2925, + "step": 48880 + }, + { + "epoch": 1.88, + "learning_rate": 1.8690048730443704e-05, + "loss": 0.3432, + "step": 48881 + }, + { + "epoch": 1.88, + "learning_rate": 1.8689407540394972e-05, + "loss": 0.2109, + "step": 48882 + }, + { + "epoch": 1.88, + "learning_rate": 1.8688766350346244e-05, + "loss": 0.2361, + "step": 48883 + }, + { + "epoch": 1.88, + "learning_rate": 1.8688125160297512e-05, + "loss": 0.3339, + "step": 48884 + }, + { + "epoch": 1.88, + "learning_rate": 1.868748397024878e-05, + "loss": 0.3846, + "step": 48885 + }, + { + "epoch": 1.88, + "learning_rate": 1.8686842780200052e-05, + "loss": 0.2705, + "step": 48886 + }, + { + "epoch": 1.88, + "learning_rate": 1.8686201590151324e-05, + "loss": 0.291, + "step": 48887 + }, + { + "epoch": 1.88, + "learning_rate": 1.8685560400102592e-05, + "loss": 0.4286, + "step": 48888 + }, + { + "epoch": 1.88, + "learning_rate": 1.868491921005386e-05, + "loss": 0.4573, + "step": 48889 + }, + { + "epoch": 1.88, + "learning_rate": 1.868427802000513e-05, + "loss": 0.2839, + "step": 48890 + }, + { + "epoch": 1.88, + "learning_rate": 1.86836368299564e-05, + "loss": 0.3015, + "step": 48891 + }, + { + "epoch": 1.88, + "learning_rate": 1.868299563990767e-05, + "loss": 0.3065, + "step": 48892 + }, + { + "epoch": 1.88, + "learning_rate": 1.8682354449858937e-05, + "loss": 0.2877, + "step": 48893 + }, + { + "epoch": 1.88, + "learning_rate": 1.8681713259810206e-05, + "loss": 0.348, + "step": 48894 + }, + { + "epoch": 1.88, + "learning_rate": 1.868107206976148e-05, + "loss": 0.3689, + "step": 48895 + }, + { + "epoch": 1.88, + "learning_rate": 1.868043087971275e-05, + "loss": 0.355, + "step": 48896 + }, + { + "epoch": 1.88, + "learning_rate": 1.8679789689664018e-05, + "loss": 0.3797, + "step": 48897 + }, + { + "epoch": 1.88, + "learning_rate": 1.8679148499615286e-05, + "loss": 0.3724, + "step": 48898 + }, + { + "epoch": 1.88, + "learning_rate": 1.8678507309566558e-05, + "loss": 0.2671, + "step": 48899 + }, + { + "epoch": 1.88, + "learning_rate": 1.8677866119517826e-05, + "loss": 0.3364, + "step": 48900 + }, + { + "epoch": 1.88, + "learning_rate": 1.8677224929469094e-05, + "loss": 0.3102, + "step": 48901 + }, + { + "epoch": 1.88, + "learning_rate": 1.8676583739420363e-05, + "loss": 0.2212, + "step": 48902 + }, + { + "epoch": 1.88, + "learning_rate": 1.8675942549371634e-05, + "loss": 0.4126, + "step": 48903 + }, + { + "epoch": 1.88, + "learning_rate": 1.8675301359322906e-05, + "loss": 0.3034, + "step": 48904 + }, + { + "epoch": 1.88, + "learning_rate": 1.8674660169274174e-05, + "loss": 0.3473, + "step": 48905 + }, + { + "epoch": 1.88, + "learning_rate": 1.8674018979225443e-05, + "loss": 0.3282, + "step": 48906 + }, + { + "epoch": 1.88, + "learning_rate": 1.8673377789176714e-05, + "loss": 0.3577, + "step": 48907 + }, + { + "epoch": 1.88, + "learning_rate": 1.8672736599127983e-05, + "loss": 0.4058, + "step": 48908 + }, + { + "epoch": 1.88, + "learning_rate": 1.867209540907925e-05, + "loss": 0.3213, + "step": 48909 + }, + { + "epoch": 1.88, + "learning_rate": 1.867145421903052e-05, + "loss": 0.2741, + "step": 48910 + }, + { + "epoch": 1.88, + "learning_rate": 1.867081302898179e-05, + "loss": 0.3396, + "step": 48911 + }, + { + "epoch": 1.88, + "learning_rate": 1.867017183893306e-05, + "loss": 0.2835, + "step": 48912 + }, + { + "epoch": 1.88, + "learning_rate": 1.866953064888433e-05, + "loss": 0.3831, + "step": 48913 + }, + { + "epoch": 1.88, + "learning_rate": 1.86688894588356e-05, + "loss": 0.2498, + "step": 48914 + }, + { + "epoch": 1.88, + "learning_rate": 1.866824826878687e-05, + "loss": 0.4392, + "step": 48915 + }, + { + "epoch": 1.88, + "learning_rate": 1.866760707873814e-05, + "loss": 0.3446, + "step": 48916 + }, + { + "epoch": 1.88, + "learning_rate": 1.8666965888689408e-05, + "loss": 0.2567, + "step": 48917 + }, + { + "epoch": 1.88, + "learning_rate": 1.8666324698640676e-05, + "loss": 0.3702, + "step": 48918 + }, + { + "epoch": 1.88, + "learning_rate": 1.8665683508591948e-05, + "loss": 0.2426, + "step": 48919 + }, + { + "epoch": 1.88, + "learning_rate": 1.8665042318543216e-05, + "loss": 0.2301, + "step": 48920 + }, + { + "epoch": 1.88, + "learning_rate": 1.8664401128494485e-05, + "loss": 0.3194, + "step": 48921 + }, + { + "epoch": 1.88, + "learning_rate": 1.8663759938445756e-05, + "loss": 0.2539, + "step": 48922 + }, + { + "epoch": 1.88, + "learning_rate": 1.8663118748397028e-05, + "loss": 0.396, + "step": 48923 + }, + { + "epoch": 1.88, + "learning_rate": 1.8662477558348296e-05, + "loss": 0.3256, + "step": 48924 + }, + { + "epoch": 1.88, + "learning_rate": 1.8661836368299565e-05, + "loss": 0.3995, + "step": 48925 + }, + { + "epoch": 1.88, + "learning_rate": 1.8661195178250833e-05, + "loss": 0.2745, + "step": 48926 + }, + { + "epoch": 1.88, + "learning_rate": 1.8660553988202105e-05, + "loss": 0.3624, + "step": 48927 + }, + { + "epoch": 1.88, + "learning_rate": 1.8659912798153373e-05, + "loss": 0.3803, + "step": 48928 + }, + { + "epoch": 1.88, + "learning_rate": 1.865927160810464e-05, + "loss": 0.4674, + "step": 48929 + }, + { + "epoch": 1.88, + "learning_rate": 1.8658630418055913e-05, + "loss": 0.3079, + "step": 48930 + }, + { + "epoch": 1.88, + "learning_rate": 1.8657989228007185e-05, + "loss": 0.5541, + "step": 48931 + }, + { + "epoch": 1.88, + "learning_rate": 1.8657348037958453e-05, + "loss": 0.3129, + "step": 48932 + }, + { + "epoch": 1.88, + "learning_rate": 1.865670684790972e-05, + "loss": 0.1749, + "step": 48933 + }, + { + "epoch": 1.88, + "learning_rate": 1.865606565786099e-05, + "loss": 0.2974, + "step": 48934 + }, + { + "epoch": 1.88, + "learning_rate": 1.865542446781226e-05, + "loss": 0.3994, + "step": 48935 + }, + { + "epoch": 1.88, + "learning_rate": 1.865478327776353e-05, + "loss": 0.272, + "step": 48936 + }, + { + "epoch": 1.88, + "learning_rate": 1.8654142087714798e-05, + "loss": 0.45, + "step": 48937 + }, + { + "epoch": 1.88, + "learning_rate": 1.8653500897666066e-05, + "loss": 0.4684, + "step": 48938 + }, + { + "epoch": 1.88, + "learning_rate": 1.865285970761734e-05, + "loss": 0.335, + "step": 48939 + }, + { + "epoch": 1.88, + "learning_rate": 1.865221851756861e-05, + "loss": 0.381, + "step": 48940 + }, + { + "epoch": 1.88, + "learning_rate": 1.8651577327519878e-05, + "loss": 0.2701, + "step": 48941 + }, + { + "epoch": 1.88, + "learning_rate": 1.8650936137471147e-05, + "loss": 0.3076, + "step": 48942 + }, + { + "epoch": 1.88, + "learning_rate": 1.8650294947422418e-05, + "loss": 0.3661, + "step": 48943 + }, + { + "epoch": 1.88, + "learning_rate": 1.8649653757373687e-05, + "loss": 0.4639, + "step": 48944 + }, + { + "epoch": 1.88, + "learning_rate": 1.8649012567324955e-05, + "loss": 0.3283, + "step": 48945 + }, + { + "epoch": 1.88, + "learning_rate": 1.8648371377276223e-05, + "loss": 0.38, + "step": 48946 + }, + { + "epoch": 1.88, + "learning_rate": 1.8647730187227495e-05, + "loss": 0.3178, + "step": 48947 + }, + { + "epoch": 1.88, + "learning_rate": 1.8647088997178767e-05, + "loss": 0.4751, + "step": 48948 + }, + { + "epoch": 1.88, + "learning_rate": 1.8646447807130035e-05, + "loss": 0.3182, + "step": 48949 + }, + { + "epoch": 1.88, + "learning_rate": 1.8645806617081303e-05, + "loss": 0.2688, + "step": 48950 + }, + { + "epoch": 1.88, + "learning_rate": 1.8645165427032575e-05, + "loss": 0.3318, + "step": 48951 + }, + { + "epoch": 1.88, + "learning_rate": 1.8644524236983843e-05, + "loss": 0.2593, + "step": 48952 + }, + { + "epoch": 1.88, + "learning_rate": 1.8643883046935112e-05, + "loss": 0.3558, + "step": 48953 + }, + { + "epoch": 1.88, + "learning_rate": 1.864324185688638e-05, + "loss": 0.2724, + "step": 48954 + }, + { + "epoch": 1.88, + "learning_rate": 1.8642600666837652e-05, + "loss": 0.3534, + "step": 48955 + }, + { + "epoch": 1.88, + "learning_rate": 1.864195947678892e-05, + "loss": 0.354, + "step": 48956 + }, + { + "epoch": 1.88, + "learning_rate": 1.8641318286740192e-05, + "loss": 0.2628, + "step": 48957 + }, + { + "epoch": 1.88, + "learning_rate": 1.864067709669146e-05, + "loss": 0.3745, + "step": 48958 + }, + { + "epoch": 1.88, + "learning_rate": 1.8640035906642732e-05, + "loss": 0.2965, + "step": 48959 + }, + { + "epoch": 1.88, + "learning_rate": 1.8639394716594e-05, + "loss": 0.3088, + "step": 48960 + }, + { + "epoch": 1.88, + "learning_rate": 1.863875352654527e-05, + "loss": 0.3092, + "step": 48961 + }, + { + "epoch": 1.88, + "learning_rate": 1.8638112336496537e-05, + "loss": 0.3908, + "step": 48962 + }, + { + "epoch": 1.88, + "learning_rate": 1.863747114644781e-05, + "loss": 0.3792, + "step": 48963 + }, + { + "epoch": 1.88, + "learning_rate": 1.8636829956399077e-05, + "loss": 0.3448, + "step": 48964 + }, + { + "epoch": 1.88, + "learning_rate": 1.8636188766350345e-05, + "loss": 0.5032, + "step": 48965 + }, + { + "epoch": 1.88, + "learning_rate": 1.8635547576301617e-05, + "loss": 0.3704, + "step": 48966 + }, + { + "epoch": 1.88, + "learning_rate": 1.863490638625289e-05, + "loss": 0.2899, + "step": 48967 + }, + { + "epoch": 1.88, + "learning_rate": 1.8634265196204157e-05, + "loss": 0.3323, + "step": 48968 + }, + { + "epoch": 1.88, + "learning_rate": 1.8633624006155425e-05, + "loss": 0.2969, + "step": 48969 + }, + { + "epoch": 1.88, + "learning_rate": 1.8632982816106694e-05, + "loss": 0.4834, + "step": 48970 + }, + { + "epoch": 1.88, + "learning_rate": 1.8632341626057965e-05, + "loss": 0.3728, + "step": 48971 + }, + { + "epoch": 1.88, + "learning_rate": 1.8631700436009234e-05, + "loss": 0.2903, + "step": 48972 + }, + { + "epoch": 1.88, + "learning_rate": 1.8631059245960502e-05, + "loss": 0.3235, + "step": 48973 + }, + { + "epoch": 1.88, + "learning_rate": 1.863041805591177e-05, + "loss": 0.3107, + "step": 48974 + }, + { + "epoch": 1.88, + "learning_rate": 1.8629776865863045e-05, + "loss": 0.3393, + "step": 48975 + }, + { + "epoch": 1.88, + "learning_rate": 1.8629135675814314e-05, + "loss": 0.2792, + "step": 48976 + }, + { + "epoch": 1.88, + "learning_rate": 1.8628494485765582e-05, + "loss": 0.2765, + "step": 48977 + }, + { + "epoch": 1.88, + "learning_rate": 1.862785329571685e-05, + "loss": 0.2677, + "step": 48978 + }, + { + "epoch": 1.88, + "learning_rate": 1.8627212105668122e-05, + "loss": 0.3669, + "step": 48979 + }, + { + "epoch": 1.88, + "learning_rate": 1.862657091561939e-05, + "loss": 0.4031, + "step": 48980 + }, + { + "epoch": 1.88, + "learning_rate": 1.862592972557066e-05, + "loss": 0.2159, + "step": 48981 + }, + { + "epoch": 1.88, + "learning_rate": 1.8625288535521927e-05, + "loss": 0.2369, + "step": 48982 + }, + { + "epoch": 1.88, + "learning_rate": 1.86246473454732e-05, + "loss": 0.3126, + "step": 48983 + }, + { + "epoch": 1.88, + "learning_rate": 1.862400615542447e-05, + "loss": 0.315, + "step": 48984 + }, + { + "epoch": 1.88, + "learning_rate": 1.862336496537574e-05, + "loss": 0.2282, + "step": 48985 + }, + { + "epoch": 1.88, + "learning_rate": 1.8622723775327007e-05, + "loss": 0.2428, + "step": 48986 + }, + { + "epoch": 1.88, + "learning_rate": 1.862208258527828e-05, + "loss": 0.3887, + "step": 48987 + }, + { + "epoch": 1.88, + "learning_rate": 1.8621441395229547e-05, + "loss": 0.3603, + "step": 48988 + }, + { + "epoch": 1.88, + "learning_rate": 1.8620800205180816e-05, + "loss": 0.2522, + "step": 48989 + }, + { + "epoch": 1.88, + "learning_rate": 1.8620159015132084e-05, + "loss": 0.3651, + "step": 48990 + }, + { + "epoch": 1.88, + "learning_rate": 1.8619517825083356e-05, + "loss": 0.3365, + "step": 48991 + }, + { + "epoch": 1.88, + "learning_rate": 1.8618876635034624e-05, + "loss": 0.3684, + "step": 48992 + }, + { + "epoch": 1.88, + "learning_rate": 1.8618235444985896e-05, + "loss": 0.3738, + "step": 48993 + }, + { + "epoch": 1.88, + "learning_rate": 1.8617594254937164e-05, + "loss": 0.2073, + "step": 48994 + }, + { + "epoch": 1.88, + "learning_rate": 1.8616953064888436e-05, + "loss": 0.3207, + "step": 48995 + }, + { + "epoch": 1.88, + "learning_rate": 1.8616311874839704e-05, + "loss": 0.3533, + "step": 48996 + }, + { + "epoch": 1.88, + "learning_rate": 1.8615670684790972e-05, + "loss": 0.3183, + "step": 48997 + }, + { + "epoch": 1.88, + "learning_rate": 1.861502949474224e-05, + "loss": 0.1983, + "step": 48998 + }, + { + "epoch": 1.88, + "learning_rate": 1.8614388304693512e-05, + "loss": 0.2617, + "step": 48999 + }, + { + "epoch": 1.88, + "learning_rate": 1.861374711464478e-05, + "loss": 0.3846, + "step": 49000 + }, + { + "epoch": 1.88, + "learning_rate": 1.861310592459605e-05, + "loss": 0.3308, + "step": 49001 + }, + { + "epoch": 1.88, + "learning_rate": 1.861246473454732e-05, + "loss": 0.2484, + "step": 49002 + }, + { + "epoch": 1.88, + "learning_rate": 1.8611823544498593e-05, + "loss": 0.2701, + "step": 49003 + }, + { + "epoch": 1.88, + "learning_rate": 1.861118235444986e-05, + "loss": 0.1767, + "step": 49004 + }, + { + "epoch": 1.88, + "learning_rate": 1.861054116440113e-05, + "loss": 0.4121, + "step": 49005 + }, + { + "epoch": 1.88, + "learning_rate": 1.8609899974352397e-05, + "loss": 0.2693, + "step": 49006 + }, + { + "epoch": 1.88, + "learning_rate": 1.860925878430367e-05, + "loss": 0.1739, + "step": 49007 + }, + { + "epoch": 1.88, + "learning_rate": 1.8608617594254938e-05, + "loss": 0.2712, + "step": 49008 + }, + { + "epoch": 1.88, + "learning_rate": 1.8607976404206206e-05, + "loss": 0.2326, + "step": 49009 + }, + { + "epoch": 1.88, + "learning_rate": 1.8607335214157474e-05, + "loss": 0.2191, + "step": 49010 + }, + { + "epoch": 1.88, + "learning_rate": 1.860669402410875e-05, + "loss": 0.1946, + "step": 49011 + }, + { + "epoch": 1.88, + "learning_rate": 1.8606052834060018e-05, + "loss": 0.2949, + "step": 49012 + }, + { + "epoch": 1.88, + "learning_rate": 1.8605411644011286e-05, + "loss": 0.3561, + "step": 49013 + }, + { + "epoch": 1.88, + "learning_rate": 1.8604770453962554e-05, + "loss": 0.4443, + "step": 49014 + }, + { + "epoch": 1.88, + "learning_rate": 1.8604129263913826e-05, + "loss": 0.3213, + "step": 49015 + }, + { + "epoch": 1.88, + "learning_rate": 1.8603488073865094e-05, + "loss": 0.2832, + "step": 49016 + }, + { + "epoch": 1.88, + "learning_rate": 1.8602846883816363e-05, + "loss": 0.3521, + "step": 49017 + }, + { + "epoch": 1.88, + "learning_rate": 1.860220569376763e-05, + "loss": 0.3606, + "step": 49018 + }, + { + "epoch": 1.88, + "learning_rate": 1.8601564503718903e-05, + "loss": 0.2933, + "step": 49019 + }, + { + "epoch": 1.88, + "learning_rate": 1.8600923313670174e-05, + "loss": 0.3656, + "step": 49020 + }, + { + "epoch": 1.88, + "learning_rate": 1.8600282123621443e-05, + "loss": 0.3425, + "step": 49021 + }, + { + "epoch": 1.88, + "learning_rate": 1.859964093357271e-05, + "loss": 0.3132, + "step": 49022 + }, + { + "epoch": 1.88, + "learning_rate": 1.8598999743523983e-05, + "loss": 0.4185, + "step": 49023 + }, + { + "epoch": 1.88, + "learning_rate": 1.859835855347525e-05, + "loss": 0.2435, + "step": 49024 + }, + { + "epoch": 1.88, + "learning_rate": 1.859771736342652e-05, + "loss": 0.3856, + "step": 49025 + }, + { + "epoch": 1.88, + "learning_rate": 1.8597076173377788e-05, + "loss": 0.4041, + "step": 49026 + }, + { + "epoch": 1.88, + "learning_rate": 1.859643498332906e-05, + "loss": 0.1692, + "step": 49027 + }, + { + "epoch": 1.88, + "learning_rate": 1.859579379328033e-05, + "loss": 0.3354, + "step": 49028 + }, + { + "epoch": 1.89, + "learning_rate": 1.85951526032316e-05, + "loss": 0.3425, + "step": 49029 + }, + { + "epoch": 1.89, + "learning_rate": 1.8594511413182868e-05, + "loss": 0.452, + "step": 49030 + }, + { + "epoch": 1.89, + "learning_rate": 1.859387022313414e-05, + "loss": 0.2627, + "step": 49031 + }, + { + "epoch": 1.89, + "learning_rate": 1.8593229033085408e-05, + "loss": 0.3328, + "step": 49032 + }, + { + "epoch": 1.89, + "learning_rate": 1.8592587843036676e-05, + "loss": 0.3375, + "step": 49033 + }, + { + "epoch": 1.89, + "learning_rate": 1.8591946652987945e-05, + "loss": 0.2761, + "step": 49034 + }, + { + "epoch": 1.89, + "learning_rate": 1.8591305462939216e-05, + "loss": 0.2917, + "step": 49035 + }, + { + "epoch": 1.89, + "learning_rate": 1.8590664272890485e-05, + "loss": 0.3438, + "step": 49036 + }, + { + "epoch": 1.89, + "learning_rate": 1.8590023082841756e-05, + "loss": 0.2464, + "step": 49037 + }, + { + "epoch": 1.89, + "learning_rate": 1.8589381892793025e-05, + "loss": 0.279, + "step": 49038 + }, + { + "epoch": 1.89, + "learning_rate": 1.8588740702744296e-05, + "loss": 0.3195, + "step": 49039 + }, + { + "epoch": 1.89, + "learning_rate": 1.8588099512695565e-05, + "loss": 0.4537, + "step": 49040 + }, + { + "epoch": 1.89, + "learning_rate": 1.8587458322646833e-05, + "loss": 0.2973, + "step": 49041 + }, + { + "epoch": 1.89, + "learning_rate": 1.85868171325981e-05, + "loss": 0.3206, + "step": 49042 + }, + { + "epoch": 1.89, + "learning_rate": 1.8586175942549373e-05, + "loss": 0.3415, + "step": 49043 + }, + { + "epoch": 1.89, + "learning_rate": 1.858553475250064e-05, + "loss": 0.3044, + "step": 49044 + }, + { + "epoch": 1.89, + "learning_rate": 1.858489356245191e-05, + "loss": 0.3545, + "step": 49045 + }, + { + "epoch": 1.89, + "learning_rate": 1.858425237240318e-05, + "loss": 0.449, + "step": 49046 + }, + { + "epoch": 1.89, + "learning_rate": 1.8583611182354453e-05, + "loss": 0.2296, + "step": 49047 + }, + { + "epoch": 1.89, + "learning_rate": 1.858296999230572e-05, + "loss": 0.3182, + "step": 49048 + }, + { + "epoch": 1.89, + "learning_rate": 1.858232880225699e-05, + "loss": 0.3402, + "step": 49049 + }, + { + "epoch": 1.89, + "learning_rate": 1.8581687612208258e-05, + "loss": 0.2328, + "step": 49050 + }, + { + "epoch": 1.89, + "learning_rate": 1.858104642215953e-05, + "loss": 0.2561, + "step": 49051 + }, + { + "epoch": 1.89, + "learning_rate": 1.8580405232110798e-05, + "loss": 0.305, + "step": 49052 + }, + { + "epoch": 1.89, + "learning_rate": 1.8579764042062067e-05, + "loss": 0.2883, + "step": 49053 + }, + { + "epoch": 1.89, + "learning_rate": 1.8579122852013335e-05, + "loss": 0.2994, + "step": 49054 + }, + { + "epoch": 1.89, + "learning_rate": 1.857848166196461e-05, + "loss": 0.3811, + "step": 49055 + }, + { + "epoch": 1.89, + "learning_rate": 1.8577840471915878e-05, + "loss": 0.2325, + "step": 49056 + }, + { + "epoch": 1.89, + "learning_rate": 1.8577199281867147e-05, + "loss": 0.2757, + "step": 49057 + }, + { + "epoch": 1.89, + "learning_rate": 1.8576558091818415e-05, + "loss": 0.316, + "step": 49058 + }, + { + "epoch": 1.89, + "learning_rate": 1.8575916901769687e-05, + "loss": 0.2248, + "step": 49059 + }, + { + "epoch": 1.89, + "learning_rate": 1.8575275711720955e-05, + "loss": 0.4483, + "step": 49060 + }, + { + "epoch": 1.89, + "learning_rate": 1.8574634521672223e-05, + "loss": 0.3394, + "step": 49061 + }, + { + "epoch": 1.89, + "learning_rate": 1.857399333162349e-05, + "loss": 0.3269, + "step": 49062 + }, + { + "epoch": 1.89, + "learning_rate": 1.8573352141574763e-05, + "loss": 0.3166, + "step": 49063 + }, + { + "epoch": 1.89, + "learning_rate": 1.8572710951526035e-05, + "loss": 0.3489, + "step": 49064 + }, + { + "epoch": 1.89, + "learning_rate": 1.8572069761477303e-05, + "loss": 0.2942, + "step": 49065 + }, + { + "epoch": 1.89, + "learning_rate": 1.8571428571428572e-05, + "loss": 0.5572, + "step": 49066 + }, + { + "epoch": 1.89, + "learning_rate": 1.8570787381379843e-05, + "loss": 0.3865, + "step": 49067 + }, + { + "epoch": 1.89, + "learning_rate": 1.8570146191331112e-05, + "loss": 0.3618, + "step": 49068 + }, + { + "epoch": 1.89, + "learning_rate": 1.856950500128238e-05, + "loss": 0.2162, + "step": 49069 + }, + { + "epoch": 1.89, + "learning_rate": 1.856886381123365e-05, + "loss": 0.346, + "step": 49070 + }, + { + "epoch": 1.89, + "learning_rate": 1.856822262118492e-05, + "loss": 0.1992, + "step": 49071 + }, + { + "epoch": 1.89, + "learning_rate": 1.856758143113619e-05, + "loss": 0.5229, + "step": 49072 + }, + { + "epoch": 1.89, + "learning_rate": 1.856694024108746e-05, + "loss": 0.403, + "step": 49073 + }, + { + "epoch": 1.89, + "learning_rate": 1.856629905103873e-05, + "loss": 0.3319, + "step": 49074 + }, + { + "epoch": 1.89, + "learning_rate": 1.856565786099e-05, + "loss": 0.1711, + "step": 49075 + }, + { + "epoch": 1.89, + "learning_rate": 1.856501667094127e-05, + "loss": 0.3253, + "step": 49076 + }, + { + "epoch": 1.89, + "learning_rate": 1.8564375480892537e-05, + "loss": 0.3087, + "step": 49077 + }, + { + "epoch": 1.89, + "learning_rate": 1.8563734290843805e-05, + "loss": 0.2993, + "step": 49078 + }, + { + "epoch": 1.89, + "learning_rate": 1.8563093100795077e-05, + "loss": 0.2584, + "step": 49079 + }, + { + "epoch": 1.89, + "learning_rate": 1.8562451910746345e-05, + "loss": 0.2148, + "step": 49080 + }, + { + "epoch": 1.89, + "learning_rate": 1.8561810720697614e-05, + "loss": 0.3291, + "step": 49081 + }, + { + "epoch": 1.89, + "learning_rate": 1.8561169530648885e-05, + "loss": 0.1996, + "step": 49082 + }, + { + "epoch": 1.89, + "learning_rate": 1.8560528340600157e-05, + "loss": 0.2632, + "step": 49083 + }, + { + "epoch": 1.89, + "learning_rate": 1.8559887150551425e-05, + "loss": 0.2489, + "step": 49084 + }, + { + "epoch": 1.89, + "learning_rate": 1.8559245960502694e-05, + "loss": 0.3619, + "step": 49085 + }, + { + "epoch": 1.89, + "learning_rate": 1.8558604770453962e-05, + "loss": 0.3358, + "step": 49086 + }, + { + "epoch": 1.89, + "learning_rate": 1.8557963580405234e-05, + "loss": 0.3124, + "step": 49087 + }, + { + "epoch": 1.89, + "learning_rate": 1.8557322390356502e-05, + "loss": 0.252, + "step": 49088 + }, + { + "epoch": 1.89, + "learning_rate": 1.855668120030777e-05, + "loss": 0.2122, + "step": 49089 + }, + { + "epoch": 1.89, + "learning_rate": 1.855604001025904e-05, + "loss": 0.2903, + "step": 49090 + }, + { + "epoch": 1.89, + "learning_rate": 1.8555398820210314e-05, + "loss": 0.2786, + "step": 49091 + }, + { + "epoch": 1.89, + "learning_rate": 1.8554757630161582e-05, + "loss": 0.2723, + "step": 49092 + }, + { + "epoch": 1.89, + "learning_rate": 1.855411644011285e-05, + "loss": 0.284, + "step": 49093 + }, + { + "epoch": 1.89, + "learning_rate": 1.855347525006412e-05, + "loss": 0.3556, + "step": 49094 + }, + { + "epoch": 1.89, + "learning_rate": 1.855283406001539e-05, + "loss": 0.278, + "step": 49095 + }, + { + "epoch": 1.89, + "learning_rate": 1.855219286996666e-05, + "loss": 0.4615, + "step": 49096 + }, + { + "epoch": 1.89, + "learning_rate": 1.8551551679917927e-05, + "loss": 0.2569, + "step": 49097 + }, + { + "epoch": 1.89, + "learning_rate": 1.8550910489869196e-05, + "loss": 0.2525, + "step": 49098 + }, + { + "epoch": 1.89, + "learning_rate": 1.8550269299820467e-05, + "loss": 0.3711, + "step": 49099 + }, + { + "epoch": 1.89, + "learning_rate": 1.854962810977174e-05, + "loss": 0.2711, + "step": 49100 + }, + { + "epoch": 1.89, + "learning_rate": 1.8548986919723007e-05, + "loss": 0.3593, + "step": 49101 + }, + { + "epoch": 1.89, + "learning_rate": 1.8548345729674276e-05, + "loss": 0.4449, + "step": 49102 + }, + { + "epoch": 1.89, + "learning_rate": 1.8547704539625547e-05, + "loss": 0.221, + "step": 49103 + }, + { + "epoch": 1.89, + "learning_rate": 1.8547063349576816e-05, + "loss": 0.2802, + "step": 49104 + }, + { + "epoch": 1.89, + "learning_rate": 1.8546422159528084e-05, + "loss": 0.302, + "step": 49105 + }, + { + "epoch": 1.89, + "learning_rate": 1.8545780969479352e-05, + "loss": 0.2457, + "step": 49106 + }, + { + "epoch": 1.89, + "learning_rate": 1.8545139779430624e-05, + "loss": 0.3318, + "step": 49107 + }, + { + "epoch": 1.89, + "learning_rate": 1.8544498589381892e-05, + "loss": 0.2924, + "step": 49108 + }, + { + "epoch": 1.89, + "learning_rate": 1.8543857399333164e-05, + "loss": 0.2594, + "step": 49109 + }, + { + "epoch": 1.89, + "learning_rate": 1.8543216209284432e-05, + "loss": 0.2122, + "step": 49110 + }, + { + "epoch": 1.89, + "learning_rate": 1.8542575019235704e-05, + "loss": 0.4719, + "step": 49111 + }, + { + "epoch": 1.89, + "learning_rate": 1.8541933829186972e-05, + "loss": 0.2089, + "step": 49112 + }, + { + "epoch": 1.89, + "learning_rate": 1.854129263913824e-05, + "loss": 0.1656, + "step": 49113 + }, + { + "epoch": 1.89, + "learning_rate": 1.854065144908951e-05, + "loss": 0.3143, + "step": 49114 + }, + { + "epoch": 1.89, + "learning_rate": 1.854001025904078e-05, + "loss": 0.2809, + "step": 49115 + }, + { + "epoch": 1.89, + "learning_rate": 1.853936906899205e-05, + "loss": 0.3369, + "step": 49116 + }, + { + "epoch": 1.89, + "learning_rate": 1.8538727878943317e-05, + "loss": 0.3676, + "step": 49117 + }, + { + "epoch": 1.89, + "learning_rate": 1.853808668889459e-05, + "loss": 0.318, + "step": 49118 + }, + { + "epoch": 1.89, + "learning_rate": 1.853744549884586e-05, + "loss": 0.3119, + "step": 49119 + }, + { + "epoch": 1.89, + "learning_rate": 1.853680430879713e-05, + "loss": 0.3401, + "step": 49120 + }, + { + "epoch": 1.89, + "learning_rate": 1.8536163118748398e-05, + "loss": 0.3131, + "step": 49121 + }, + { + "epoch": 1.89, + "learning_rate": 1.8535521928699666e-05, + "loss": 0.1972, + "step": 49122 + }, + { + "epoch": 1.89, + "learning_rate": 1.8534880738650938e-05, + "loss": 0.3603, + "step": 49123 + }, + { + "epoch": 1.89, + "learning_rate": 1.8534239548602206e-05, + "loss": 0.1792, + "step": 49124 + }, + { + "epoch": 1.89, + "learning_rate": 1.8533598358553474e-05, + "loss": 0.351, + "step": 49125 + }, + { + "epoch": 1.89, + "learning_rate": 1.8532957168504746e-05, + "loss": 0.4785, + "step": 49126 + }, + { + "epoch": 1.89, + "learning_rate": 1.8532315978456018e-05, + "loss": 0.2888, + "step": 49127 + }, + { + "epoch": 1.89, + "learning_rate": 1.8531674788407286e-05, + "loss": 0.3034, + "step": 49128 + }, + { + "epoch": 1.89, + "learning_rate": 1.8531033598358554e-05, + "loss": 0.4002, + "step": 49129 + }, + { + "epoch": 1.89, + "learning_rate": 1.8530392408309823e-05, + "loss": 0.3349, + "step": 49130 + }, + { + "epoch": 1.89, + "learning_rate": 1.8529751218261094e-05, + "loss": 0.2219, + "step": 49131 + }, + { + "epoch": 1.89, + "learning_rate": 1.8529110028212363e-05, + "loss": 0.2574, + "step": 49132 + }, + { + "epoch": 1.89, + "learning_rate": 1.852846883816363e-05, + "loss": 0.404, + "step": 49133 + }, + { + "epoch": 1.89, + "learning_rate": 1.85278276481149e-05, + "loss": 0.2568, + "step": 49134 + }, + { + "epoch": 1.89, + "learning_rate": 1.8527186458066174e-05, + "loss": 0.3351, + "step": 49135 + }, + { + "epoch": 1.89, + "learning_rate": 1.8526545268017443e-05, + "loss": 0.2497, + "step": 49136 + }, + { + "epoch": 1.89, + "learning_rate": 1.852590407796871e-05, + "loss": 0.4363, + "step": 49137 + }, + { + "epoch": 1.89, + "learning_rate": 1.852526288791998e-05, + "loss": 0.3429, + "step": 49138 + }, + { + "epoch": 1.89, + "learning_rate": 1.852462169787125e-05, + "loss": 0.3281, + "step": 49139 + }, + { + "epoch": 1.89, + "learning_rate": 1.852398050782252e-05, + "loss": 0.2925, + "step": 49140 + }, + { + "epoch": 1.89, + "learning_rate": 1.8523339317773788e-05, + "loss": 0.2431, + "step": 49141 + }, + { + "epoch": 1.89, + "learning_rate": 1.8522698127725056e-05, + "loss": 0.207, + "step": 49142 + }, + { + "epoch": 1.89, + "learning_rate": 1.8522056937676328e-05, + "loss": 0.3548, + "step": 49143 + }, + { + "epoch": 1.89, + "learning_rate": 1.85214157476276e-05, + "loss": 0.3056, + "step": 49144 + }, + { + "epoch": 1.89, + "learning_rate": 1.8520774557578868e-05, + "loss": 0.2436, + "step": 49145 + }, + { + "epoch": 1.89, + "learning_rate": 1.8520133367530136e-05, + "loss": 0.2692, + "step": 49146 + }, + { + "epoch": 1.89, + "learning_rate": 1.8519492177481408e-05, + "loss": 0.2875, + "step": 49147 + }, + { + "epoch": 1.89, + "learning_rate": 1.8518850987432676e-05, + "loss": 0.4024, + "step": 49148 + }, + { + "epoch": 1.89, + "learning_rate": 1.8518209797383945e-05, + "loss": 0.2797, + "step": 49149 + }, + { + "epoch": 1.89, + "learning_rate": 1.8517568607335213e-05, + "loss": 0.3269, + "step": 49150 + }, + { + "epoch": 1.89, + "learning_rate": 1.8516927417286485e-05, + "loss": 0.2072, + "step": 49151 + }, + { + "epoch": 1.89, + "learning_rate": 1.8516286227237753e-05, + "loss": 0.2927, + "step": 49152 + }, + { + "epoch": 1.89, + "learning_rate": 1.8515645037189025e-05, + "loss": 0.2272, + "step": 49153 + }, + { + "epoch": 1.89, + "learning_rate": 1.8515003847140293e-05, + "loss": 0.3128, + "step": 49154 + }, + { + "epoch": 1.89, + "learning_rate": 1.8514362657091565e-05, + "loss": 0.2594, + "step": 49155 + }, + { + "epoch": 1.89, + "learning_rate": 1.8513721467042833e-05, + "loss": 0.3267, + "step": 49156 + }, + { + "epoch": 1.89, + "learning_rate": 1.85130802769941e-05, + "loss": 0.4645, + "step": 49157 + }, + { + "epoch": 1.89, + "learning_rate": 1.851243908694537e-05, + "loss": 0.3419, + "step": 49158 + }, + { + "epoch": 1.89, + "learning_rate": 1.851179789689664e-05, + "loss": 0.2807, + "step": 49159 + }, + { + "epoch": 1.89, + "learning_rate": 1.851115670684791e-05, + "loss": 0.2884, + "step": 49160 + }, + { + "epoch": 1.89, + "learning_rate": 1.8510515516799178e-05, + "loss": 0.1919, + "step": 49161 + }, + { + "epoch": 1.89, + "learning_rate": 1.850987432675045e-05, + "loss": 0.2707, + "step": 49162 + }, + { + "epoch": 1.89, + "learning_rate": 1.850923313670172e-05, + "loss": 0.2911, + "step": 49163 + }, + { + "epoch": 1.89, + "learning_rate": 1.850859194665299e-05, + "loss": 0.2756, + "step": 49164 + }, + { + "epoch": 1.89, + "learning_rate": 1.8507950756604258e-05, + "loss": 0.34, + "step": 49165 + }, + { + "epoch": 1.89, + "learning_rate": 1.8507309566555527e-05, + "loss": 0.2786, + "step": 49166 + }, + { + "epoch": 1.89, + "learning_rate": 1.8506668376506798e-05, + "loss": 0.3188, + "step": 49167 + }, + { + "epoch": 1.89, + "learning_rate": 1.8506027186458067e-05, + "loss": 0.3471, + "step": 49168 + }, + { + "epoch": 1.89, + "learning_rate": 1.8505385996409335e-05, + "loss": 0.289, + "step": 49169 + }, + { + "epoch": 1.89, + "learning_rate": 1.8504744806360603e-05, + "loss": 0.2223, + "step": 49170 + }, + { + "epoch": 1.89, + "learning_rate": 1.850410361631188e-05, + "loss": 0.3408, + "step": 49171 + }, + { + "epoch": 1.89, + "learning_rate": 1.8503462426263147e-05, + "loss": 0.3103, + "step": 49172 + }, + { + "epoch": 1.89, + "learning_rate": 1.8502821236214415e-05, + "loss": 0.2247, + "step": 49173 + }, + { + "epoch": 1.89, + "learning_rate": 1.8502180046165683e-05, + "loss": 0.3064, + "step": 49174 + }, + { + "epoch": 1.89, + "learning_rate": 1.8501538856116955e-05, + "loss": 0.3198, + "step": 49175 + }, + { + "epoch": 1.89, + "learning_rate": 1.8500897666068223e-05, + "loss": 0.269, + "step": 49176 + }, + { + "epoch": 1.89, + "learning_rate": 1.8500256476019492e-05, + "loss": 0.4204, + "step": 49177 + }, + { + "epoch": 1.89, + "learning_rate": 1.849961528597076e-05, + "loss": 0.3758, + "step": 49178 + }, + { + "epoch": 1.89, + "learning_rate": 1.8498974095922032e-05, + "loss": 0.303, + "step": 49179 + }, + { + "epoch": 1.89, + "learning_rate": 1.8498332905873303e-05, + "loss": 0.203, + "step": 49180 + }, + { + "epoch": 1.89, + "learning_rate": 1.8497691715824572e-05, + "loss": 0.2506, + "step": 49181 + }, + { + "epoch": 1.89, + "learning_rate": 1.849705052577584e-05, + "loss": 0.2666, + "step": 49182 + }, + { + "epoch": 1.89, + "learning_rate": 1.8496409335727112e-05, + "loss": 0.3154, + "step": 49183 + }, + { + "epoch": 1.89, + "learning_rate": 1.849576814567838e-05, + "loss": 0.3066, + "step": 49184 + }, + { + "epoch": 1.89, + "learning_rate": 1.849512695562965e-05, + "loss": 0.2623, + "step": 49185 + }, + { + "epoch": 1.89, + "learning_rate": 1.8494485765580917e-05, + "loss": 0.3725, + "step": 49186 + }, + { + "epoch": 1.89, + "learning_rate": 1.849384457553219e-05, + "loss": 0.3019, + "step": 49187 + }, + { + "epoch": 1.89, + "learning_rate": 1.8493203385483457e-05, + "loss": 0.2187, + "step": 49188 + }, + { + "epoch": 1.89, + "learning_rate": 1.849256219543473e-05, + "loss": 0.2663, + "step": 49189 + }, + { + "epoch": 1.89, + "learning_rate": 1.8491921005385997e-05, + "loss": 0.3809, + "step": 49190 + }, + { + "epoch": 1.89, + "learning_rate": 1.849127981533727e-05, + "loss": 0.4193, + "step": 49191 + }, + { + "epoch": 1.89, + "learning_rate": 1.8490638625288537e-05, + "loss": 0.3509, + "step": 49192 + }, + { + "epoch": 1.89, + "learning_rate": 1.8489997435239805e-05, + "loss": 0.3338, + "step": 49193 + }, + { + "epoch": 1.89, + "learning_rate": 1.8489356245191074e-05, + "loss": 0.3997, + "step": 49194 + }, + { + "epoch": 1.89, + "learning_rate": 1.8488715055142345e-05, + "loss": 0.2419, + "step": 49195 + }, + { + "epoch": 1.89, + "learning_rate": 1.8488073865093614e-05, + "loss": 0.3849, + "step": 49196 + }, + { + "epoch": 1.89, + "learning_rate": 1.8487432675044882e-05, + "loss": 0.3188, + "step": 49197 + }, + { + "epoch": 1.89, + "learning_rate": 1.8486791484996154e-05, + "loss": 0.2555, + "step": 49198 + }, + { + "epoch": 1.89, + "learning_rate": 1.8486150294947425e-05, + "loss": 0.3269, + "step": 49199 + }, + { + "epoch": 1.89, + "learning_rate": 1.8485509104898694e-05, + "loss": 0.3861, + "step": 49200 + }, + { + "epoch": 1.89, + "learning_rate": 1.8484867914849962e-05, + "loss": 0.267, + "step": 49201 + }, + { + "epoch": 1.89, + "learning_rate": 1.848422672480123e-05, + "loss": 0.2894, + "step": 49202 + }, + { + "epoch": 1.89, + "learning_rate": 1.8483585534752502e-05, + "loss": 0.2965, + "step": 49203 + }, + { + "epoch": 1.89, + "learning_rate": 1.848294434470377e-05, + "loss": 0.4255, + "step": 49204 + }, + { + "epoch": 1.89, + "learning_rate": 1.848230315465504e-05, + "loss": 0.3957, + "step": 49205 + }, + { + "epoch": 1.89, + "learning_rate": 1.848166196460631e-05, + "loss": 0.2525, + "step": 49206 + }, + { + "epoch": 1.89, + "learning_rate": 1.8481020774557582e-05, + "loss": 0.3123, + "step": 49207 + }, + { + "epoch": 1.89, + "learning_rate": 1.848037958450885e-05, + "loss": 0.32, + "step": 49208 + }, + { + "epoch": 1.89, + "learning_rate": 1.847973839446012e-05, + "loss": 0.2566, + "step": 49209 + }, + { + "epoch": 1.89, + "learning_rate": 1.8479097204411387e-05, + "loss": 0.267, + "step": 49210 + }, + { + "epoch": 1.89, + "learning_rate": 1.847845601436266e-05, + "loss": 0.3948, + "step": 49211 + }, + { + "epoch": 1.89, + "learning_rate": 1.8477814824313927e-05, + "loss": 0.3468, + "step": 49212 + }, + { + "epoch": 1.89, + "learning_rate": 1.8477173634265196e-05, + "loss": 0.2706, + "step": 49213 + }, + { + "epoch": 1.89, + "learning_rate": 1.8476532444216464e-05, + "loss": 0.3512, + "step": 49214 + }, + { + "epoch": 1.89, + "learning_rate": 1.847589125416774e-05, + "loss": 0.4289, + "step": 49215 + }, + { + "epoch": 1.89, + "learning_rate": 1.8475250064119007e-05, + "loss": 0.3285, + "step": 49216 + }, + { + "epoch": 1.89, + "learning_rate": 1.8474608874070276e-05, + "loss": 0.3312, + "step": 49217 + }, + { + "epoch": 1.89, + "learning_rate": 1.8473967684021544e-05, + "loss": 0.267, + "step": 49218 + }, + { + "epoch": 1.89, + "learning_rate": 1.8473326493972816e-05, + "loss": 0.2482, + "step": 49219 + }, + { + "epoch": 1.89, + "learning_rate": 1.8472685303924084e-05, + "loss": 0.441, + "step": 49220 + }, + { + "epoch": 1.89, + "learning_rate": 1.8472044113875352e-05, + "loss": 0.1583, + "step": 49221 + }, + { + "epoch": 1.89, + "learning_rate": 1.847140292382662e-05, + "loss": 0.3539, + "step": 49222 + }, + { + "epoch": 1.89, + "learning_rate": 1.8470761733777892e-05, + "loss": 0.4959, + "step": 49223 + }, + { + "epoch": 1.89, + "learning_rate": 1.8470120543729164e-05, + "loss": 0.1977, + "step": 49224 + }, + { + "epoch": 1.89, + "learning_rate": 1.8469479353680432e-05, + "loss": 0.2498, + "step": 49225 + }, + { + "epoch": 1.89, + "learning_rate": 1.84688381636317e-05, + "loss": 0.3087, + "step": 49226 + }, + { + "epoch": 1.89, + "learning_rate": 1.8468196973582973e-05, + "loss": 0.3441, + "step": 49227 + }, + { + "epoch": 1.89, + "learning_rate": 1.846755578353424e-05, + "loss": 0.3541, + "step": 49228 + }, + { + "epoch": 1.89, + "learning_rate": 1.846691459348551e-05, + "loss": 0.4075, + "step": 49229 + }, + { + "epoch": 1.89, + "learning_rate": 1.8466273403436778e-05, + "loss": 0.3726, + "step": 49230 + }, + { + "epoch": 1.89, + "learning_rate": 1.846563221338805e-05, + "loss": 0.2688, + "step": 49231 + }, + { + "epoch": 1.89, + "learning_rate": 1.8464991023339318e-05, + "loss": 0.2458, + "step": 49232 + }, + { + "epoch": 1.89, + "learning_rate": 1.846434983329059e-05, + "loss": 0.3686, + "step": 49233 + }, + { + "epoch": 1.89, + "learning_rate": 1.8463708643241858e-05, + "loss": 0.2731, + "step": 49234 + }, + { + "epoch": 1.89, + "learning_rate": 1.846306745319313e-05, + "loss": 0.2923, + "step": 49235 + }, + { + "epoch": 1.89, + "learning_rate": 1.8462426263144398e-05, + "loss": 0.4914, + "step": 49236 + }, + { + "epoch": 1.89, + "learning_rate": 1.8461785073095666e-05, + "loss": 0.3164, + "step": 49237 + }, + { + "epoch": 1.89, + "learning_rate": 1.8461143883046934e-05, + "loss": 0.3844, + "step": 49238 + }, + { + "epoch": 1.89, + "learning_rate": 1.8460502692998206e-05, + "loss": 0.2546, + "step": 49239 + }, + { + "epoch": 1.89, + "learning_rate": 1.8459861502949474e-05, + "loss": 0.4015, + "step": 49240 + }, + { + "epoch": 1.89, + "learning_rate": 1.8459220312900743e-05, + "loss": 0.334, + "step": 49241 + }, + { + "epoch": 1.89, + "learning_rate": 1.8458579122852014e-05, + "loss": 0.2731, + "step": 49242 + }, + { + "epoch": 1.89, + "learning_rate": 1.8457937932803286e-05, + "loss": 0.4997, + "step": 49243 + }, + { + "epoch": 1.89, + "learning_rate": 1.8457296742754554e-05, + "loss": 0.3828, + "step": 49244 + }, + { + "epoch": 1.89, + "learning_rate": 1.8456655552705823e-05, + "loss": 0.3866, + "step": 49245 + }, + { + "epoch": 1.89, + "learning_rate": 1.845601436265709e-05, + "loss": 0.3474, + "step": 49246 + }, + { + "epoch": 1.89, + "learning_rate": 1.8455373172608363e-05, + "loss": 0.4065, + "step": 49247 + }, + { + "epoch": 1.89, + "learning_rate": 1.845473198255963e-05, + "loss": 0.3899, + "step": 49248 + }, + { + "epoch": 1.89, + "learning_rate": 1.84540907925109e-05, + "loss": 0.2712, + "step": 49249 + }, + { + "epoch": 1.89, + "learning_rate": 1.8453449602462168e-05, + "loss": 0.3377, + "step": 49250 + }, + { + "epoch": 1.89, + "learning_rate": 1.8452808412413443e-05, + "loss": 0.3411, + "step": 49251 + }, + { + "epoch": 1.89, + "learning_rate": 1.845216722236471e-05, + "loss": 0.3484, + "step": 49252 + }, + { + "epoch": 1.89, + "learning_rate": 1.845152603231598e-05, + "loss": 0.3387, + "step": 49253 + }, + { + "epoch": 1.89, + "learning_rate": 1.8450884842267248e-05, + "loss": 0.358, + "step": 49254 + }, + { + "epoch": 1.89, + "learning_rate": 1.845024365221852e-05, + "loss": 0.3109, + "step": 49255 + }, + { + "epoch": 1.89, + "learning_rate": 1.8449602462169788e-05, + "loss": 0.3366, + "step": 49256 + }, + { + "epoch": 1.89, + "learning_rate": 1.8448961272121056e-05, + "loss": 0.3646, + "step": 49257 + }, + { + "epoch": 1.89, + "learning_rate": 1.8448320082072325e-05, + "loss": 0.3564, + "step": 49258 + }, + { + "epoch": 1.89, + "learning_rate": 1.8447678892023596e-05, + "loss": 0.3877, + "step": 49259 + }, + { + "epoch": 1.89, + "learning_rate": 1.8447037701974868e-05, + "loss": 0.2085, + "step": 49260 + }, + { + "epoch": 1.89, + "learning_rate": 1.8446396511926136e-05, + "loss": 0.365, + "step": 49261 + }, + { + "epoch": 1.89, + "learning_rate": 1.8445755321877405e-05, + "loss": 0.3345, + "step": 49262 + }, + { + "epoch": 1.89, + "learning_rate": 1.8445114131828676e-05, + "loss": 0.228, + "step": 49263 + }, + { + "epoch": 1.89, + "learning_rate": 1.8444472941779945e-05, + "loss": 0.3452, + "step": 49264 + }, + { + "epoch": 1.89, + "learning_rate": 1.8443831751731213e-05, + "loss": 0.3506, + "step": 49265 + }, + { + "epoch": 1.89, + "learning_rate": 1.844319056168248e-05, + "loss": 0.3621, + "step": 49266 + }, + { + "epoch": 1.89, + "learning_rate": 1.8442549371633753e-05, + "loss": 0.346, + "step": 49267 + }, + { + "epoch": 1.89, + "learning_rate": 1.844190818158502e-05, + "loss": 0.4719, + "step": 49268 + }, + { + "epoch": 1.89, + "learning_rate": 1.8441266991536293e-05, + "loss": 0.2972, + "step": 49269 + }, + { + "epoch": 1.89, + "learning_rate": 1.844062580148756e-05, + "loss": 0.3315, + "step": 49270 + }, + { + "epoch": 1.89, + "learning_rate": 1.8439984611438833e-05, + "loss": 0.1765, + "step": 49271 + }, + { + "epoch": 1.89, + "learning_rate": 1.84393434213901e-05, + "loss": 0.4773, + "step": 49272 + }, + { + "epoch": 1.89, + "learning_rate": 1.843870223134137e-05, + "loss": 0.3023, + "step": 49273 + }, + { + "epoch": 1.89, + "learning_rate": 1.8438061041292638e-05, + "loss": 0.3026, + "step": 49274 + }, + { + "epoch": 1.89, + "learning_rate": 1.843741985124391e-05, + "loss": 0.2499, + "step": 49275 + }, + { + "epoch": 1.89, + "learning_rate": 1.8436778661195178e-05, + "loss": 0.2853, + "step": 49276 + }, + { + "epoch": 1.89, + "learning_rate": 1.8436137471146447e-05, + "loss": 0.2193, + "step": 49277 + }, + { + "epoch": 1.89, + "learning_rate": 1.8435496281097718e-05, + "loss": 0.3488, + "step": 49278 + }, + { + "epoch": 1.89, + "learning_rate": 1.843485509104899e-05, + "loss": 0.4048, + "step": 49279 + }, + { + "epoch": 1.89, + "learning_rate": 1.843421390100026e-05, + "loss": 0.4217, + "step": 49280 + }, + { + "epoch": 1.89, + "learning_rate": 1.8433572710951527e-05, + "loss": 0.2031, + "step": 49281 + }, + { + "epoch": 1.89, + "learning_rate": 1.8432931520902795e-05, + "loss": 0.2914, + "step": 49282 + }, + { + "epoch": 1.89, + "learning_rate": 1.8432290330854067e-05, + "loss": 0.2977, + "step": 49283 + }, + { + "epoch": 1.89, + "learning_rate": 1.8431649140805335e-05, + "loss": 0.3271, + "step": 49284 + }, + { + "epoch": 1.89, + "learning_rate": 1.8431007950756603e-05, + "loss": 0.2617, + "step": 49285 + }, + { + "epoch": 1.89, + "learning_rate": 1.8430366760707875e-05, + "loss": 0.3855, + "step": 49286 + }, + { + "epoch": 1.89, + "learning_rate": 1.8429725570659147e-05, + "loss": 0.353, + "step": 49287 + }, + { + "epoch": 1.89, + "learning_rate": 1.8429084380610415e-05, + "loss": 0.2861, + "step": 49288 + }, + { + "epoch": 1.9, + "learning_rate": 1.8428443190561683e-05, + "loss": 0.381, + "step": 49289 + }, + { + "epoch": 1.9, + "learning_rate": 1.8427802000512952e-05, + "loss": 0.362, + "step": 49290 + }, + { + "epoch": 1.9, + "learning_rate": 1.8427160810464223e-05, + "loss": 0.2615, + "step": 49291 + }, + { + "epoch": 1.9, + "learning_rate": 1.8426519620415492e-05, + "loss": 0.3032, + "step": 49292 + }, + { + "epoch": 1.9, + "learning_rate": 1.842587843036676e-05, + "loss": 0.3865, + "step": 49293 + }, + { + "epoch": 1.9, + "learning_rate": 1.842523724031803e-05, + "loss": 0.2897, + "step": 49294 + }, + { + "epoch": 1.9, + "learning_rate": 1.84245960502693e-05, + "loss": 0.4058, + "step": 49295 + }, + { + "epoch": 1.9, + "learning_rate": 1.8423954860220572e-05, + "loss": 0.3529, + "step": 49296 + }, + { + "epoch": 1.9, + "learning_rate": 1.842331367017184e-05, + "loss": 0.3783, + "step": 49297 + }, + { + "epoch": 1.9, + "learning_rate": 1.842267248012311e-05, + "loss": 0.351, + "step": 49298 + }, + { + "epoch": 1.9, + "learning_rate": 1.842203129007438e-05, + "loss": 0.2527, + "step": 49299 + }, + { + "epoch": 1.9, + "learning_rate": 1.842139010002565e-05, + "loss": 0.3507, + "step": 49300 + }, + { + "epoch": 1.9, + "learning_rate": 1.8420748909976917e-05, + "loss": 0.3784, + "step": 49301 + }, + { + "epoch": 1.9, + "learning_rate": 1.8420107719928185e-05, + "loss": 0.3255, + "step": 49302 + }, + { + "epoch": 1.9, + "learning_rate": 1.8419466529879457e-05, + "loss": 0.2661, + "step": 49303 + }, + { + "epoch": 1.9, + "learning_rate": 1.8418825339830725e-05, + "loss": 0.2178, + "step": 49304 + }, + { + "epoch": 1.9, + "learning_rate": 1.8418184149781997e-05, + "loss": 0.3172, + "step": 49305 + }, + { + "epoch": 1.9, + "learning_rate": 1.8417542959733265e-05, + "loss": 0.4857, + "step": 49306 + }, + { + "epoch": 1.9, + "learning_rate": 1.8416901769684537e-05, + "loss": 0.2505, + "step": 49307 + }, + { + "epoch": 1.9, + "learning_rate": 1.8416260579635805e-05, + "loss": 0.3429, + "step": 49308 + }, + { + "epoch": 1.9, + "learning_rate": 1.8415619389587074e-05, + "loss": 0.2223, + "step": 49309 + }, + { + "epoch": 1.9, + "learning_rate": 1.8414978199538342e-05, + "loss": 0.4896, + "step": 49310 + }, + { + "epoch": 1.9, + "learning_rate": 1.8414337009489614e-05, + "loss": 0.2421, + "step": 49311 + }, + { + "epoch": 1.9, + "learning_rate": 1.8413695819440882e-05, + "loss": 0.268, + "step": 49312 + }, + { + "epoch": 1.9, + "learning_rate": 1.8413054629392154e-05, + "loss": 0.28, + "step": 49313 + }, + { + "epoch": 1.9, + "learning_rate": 1.8412413439343422e-05, + "loss": 0.2559, + "step": 49314 + }, + { + "epoch": 1.9, + "learning_rate": 1.8411772249294694e-05, + "loss": 0.3493, + "step": 49315 + }, + { + "epoch": 1.9, + "learning_rate": 1.8411131059245962e-05, + "loss": 0.26, + "step": 49316 + }, + { + "epoch": 1.9, + "learning_rate": 1.841048986919723e-05, + "loss": 0.2832, + "step": 49317 + }, + { + "epoch": 1.9, + "learning_rate": 1.84098486791485e-05, + "loss": 0.4264, + "step": 49318 + }, + { + "epoch": 1.9, + "learning_rate": 1.840920748909977e-05, + "loss": 0.4427, + "step": 49319 + }, + { + "epoch": 1.9, + "learning_rate": 1.840856629905104e-05, + "loss": 0.4511, + "step": 49320 + }, + { + "epoch": 1.9, + "learning_rate": 1.8407925109002307e-05, + "loss": 0.2198, + "step": 49321 + }, + { + "epoch": 1.9, + "learning_rate": 1.840728391895358e-05, + "loss": 0.2473, + "step": 49322 + }, + { + "epoch": 1.9, + "learning_rate": 1.840664272890485e-05, + "loss": 0.373, + "step": 49323 + }, + { + "epoch": 1.9, + "learning_rate": 1.840600153885612e-05, + "loss": 0.3325, + "step": 49324 + }, + { + "epoch": 1.9, + "learning_rate": 1.8405360348807387e-05, + "loss": 0.2738, + "step": 49325 + }, + { + "epoch": 1.9, + "learning_rate": 1.8404719158758656e-05, + "loss": 0.3977, + "step": 49326 + }, + { + "epoch": 1.9, + "learning_rate": 1.8404077968709927e-05, + "loss": 0.3664, + "step": 49327 + }, + { + "epoch": 1.9, + "learning_rate": 1.8403436778661196e-05, + "loss": 0.2589, + "step": 49328 + }, + { + "epoch": 1.9, + "learning_rate": 1.8402795588612464e-05, + "loss": 0.2167, + "step": 49329 + }, + { + "epoch": 1.9, + "learning_rate": 1.8402154398563732e-05, + "loss": 0.3607, + "step": 49330 + }, + { + "epoch": 1.9, + "learning_rate": 1.8401513208515007e-05, + "loss": 0.3438, + "step": 49331 + }, + { + "epoch": 1.9, + "learning_rate": 1.8400872018466276e-05, + "loss": 0.3255, + "step": 49332 + }, + { + "epoch": 1.9, + "learning_rate": 1.8400230828417544e-05, + "loss": 0.2888, + "step": 49333 + }, + { + "epoch": 1.9, + "learning_rate": 1.8399589638368812e-05, + "loss": 0.3565, + "step": 49334 + }, + { + "epoch": 1.9, + "learning_rate": 1.8398948448320084e-05, + "loss": 0.4318, + "step": 49335 + }, + { + "epoch": 1.9, + "learning_rate": 1.8398307258271352e-05, + "loss": 0.3437, + "step": 49336 + }, + { + "epoch": 1.9, + "learning_rate": 1.839766606822262e-05, + "loss": 0.2957, + "step": 49337 + }, + { + "epoch": 1.9, + "learning_rate": 1.839702487817389e-05, + "loss": 0.3609, + "step": 49338 + }, + { + "epoch": 1.9, + "learning_rate": 1.839638368812516e-05, + "loss": 0.2334, + "step": 49339 + }, + { + "epoch": 1.9, + "learning_rate": 1.8395742498076433e-05, + "loss": 0.4687, + "step": 49340 + }, + { + "epoch": 1.9, + "learning_rate": 1.83951013080277e-05, + "loss": 0.3388, + "step": 49341 + }, + { + "epoch": 1.9, + "learning_rate": 1.839446011797897e-05, + "loss": 0.4008, + "step": 49342 + }, + { + "epoch": 1.9, + "learning_rate": 1.839381892793024e-05, + "loss": 0.3115, + "step": 49343 + }, + { + "epoch": 1.9, + "learning_rate": 1.839317773788151e-05, + "loss": 0.2348, + "step": 49344 + }, + { + "epoch": 1.9, + "learning_rate": 1.8392536547832778e-05, + "loss": 0.2929, + "step": 49345 + }, + { + "epoch": 1.9, + "learning_rate": 1.8391895357784046e-05, + "loss": 0.2418, + "step": 49346 + }, + { + "epoch": 1.9, + "learning_rate": 1.8391254167735318e-05, + "loss": 0.2928, + "step": 49347 + }, + { + "epoch": 1.9, + "learning_rate": 1.8390612977686586e-05, + "loss": 0.2752, + "step": 49348 + }, + { + "epoch": 1.9, + "learning_rate": 1.8389971787637858e-05, + "loss": 0.3391, + "step": 49349 + }, + { + "epoch": 1.9, + "learning_rate": 1.8389330597589126e-05, + "loss": 0.2734, + "step": 49350 + }, + { + "epoch": 1.9, + "learning_rate": 1.8388689407540398e-05, + "loss": 0.3007, + "step": 49351 + }, + { + "epoch": 1.9, + "learning_rate": 1.8388048217491666e-05, + "loss": 0.3979, + "step": 49352 + }, + { + "epoch": 1.9, + "learning_rate": 1.8387407027442934e-05, + "loss": 0.24, + "step": 49353 + }, + { + "epoch": 1.9, + "learning_rate": 1.8386765837394203e-05, + "loss": 0.3352, + "step": 49354 + }, + { + "epoch": 1.9, + "learning_rate": 1.8386124647345474e-05, + "loss": 0.3974, + "step": 49355 + }, + { + "epoch": 1.9, + "learning_rate": 1.8385483457296743e-05, + "loss": 0.2681, + "step": 49356 + }, + { + "epoch": 1.9, + "learning_rate": 1.838484226724801e-05, + "loss": 0.3372, + "step": 49357 + }, + { + "epoch": 1.9, + "learning_rate": 1.8384201077199283e-05, + "loss": 0.2888, + "step": 49358 + }, + { + "epoch": 1.9, + "learning_rate": 1.8383559887150555e-05, + "loss": 0.3584, + "step": 49359 + }, + { + "epoch": 1.9, + "learning_rate": 1.8382918697101823e-05, + "loss": 0.3297, + "step": 49360 + }, + { + "epoch": 1.9, + "learning_rate": 1.838227750705309e-05, + "loss": 0.2173, + "step": 49361 + }, + { + "epoch": 1.9, + "learning_rate": 1.838163631700436e-05, + "loss": 0.2357, + "step": 49362 + }, + { + "epoch": 1.9, + "learning_rate": 1.838099512695563e-05, + "loss": 0.43, + "step": 49363 + }, + { + "epoch": 1.9, + "learning_rate": 1.83803539369069e-05, + "loss": 0.2561, + "step": 49364 + }, + { + "epoch": 1.9, + "learning_rate": 1.8379712746858168e-05, + "loss": 0.2704, + "step": 49365 + }, + { + "epoch": 1.9, + "learning_rate": 1.837907155680944e-05, + "loss": 0.3228, + "step": 49366 + }, + { + "epoch": 1.9, + "learning_rate": 1.837843036676071e-05, + "loss": 0.2143, + "step": 49367 + }, + { + "epoch": 1.9, + "learning_rate": 1.837778917671198e-05, + "loss": 0.3183, + "step": 49368 + }, + { + "epoch": 1.9, + "learning_rate": 1.8377147986663248e-05, + "loss": 0.3112, + "step": 49369 + }, + { + "epoch": 1.9, + "learning_rate": 1.8376506796614516e-05, + "loss": 0.3058, + "step": 49370 + }, + { + "epoch": 1.9, + "learning_rate": 1.8375865606565788e-05, + "loss": 0.3378, + "step": 49371 + }, + { + "epoch": 1.9, + "learning_rate": 1.8375224416517056e-05, + "loss": 0.3281, + "step": 49372 + }, + { + "epoch": 1.9, + "learning_rate": 1.8374583226468325e-05, + "loss": 0.4081, + "step": 49373 + }, + { + "epoch": 1.9, + "learning_rate": 1.8373942036419593e-05, + "loss": 0.2287, + "step": 49374 + }, + { + "epoch": 1.9, + "learning_rate": 1.8373300846370865e-05, + "loss": 0.4285, + "step": 49375 + }, + { + "epoch": 1.9, + "learning_rate": 1.8372659656322136e-05, + "loss": 0.3507, + "step": 49376 + }, + { + "epoch": 1.9, + "learning_rate": 1.8372018466273405e-05, + "loss": 0.407, + "step": 49377 + }, + { + "epoch": 1.9, + "learning_rate": 1.8371377276224673e-05, + "loss": 0.2946, + "step": 49378 + }, + { + "epoch": 1.9, + "learning_rate": 1.8370736086175945e-05, + "loss": 0.4653, + "step": 49379 + }, + { + "epoch": 1.9, + "learning_rate": 1.8370094896127213e-05, + "loss": 0.3854, + "step": 49380 + }, + { + "epoch": 1.9, + "learning_rate": 1.836945370607848e-05, + "loss": 0.222, + "step": 49381 + }, + { + "epoch": 1.9, + "learning_rate": 1.836881251602975e-05, + "loss": 0.37, + "step": 49382 + }, + { + "epoch": 1.9, + "learning_rate": 1.836817132598102e-05, + "loss": 0.472, + "step": 49383 + }, + { + "epoch": 1.9, + "learning_rate": 1.836753013593229e-05, + "loss": 0.3077, + "step": 49384 + }, + { + "epoch": 1.9, + "learning_rate": 1.836688894588356e-05, + "loss": 0.2215, + "step": 49385 + }, + { + "epoch": 1.9, + "learning_rate": 1.836624775583483e-05, + "loss": 0.2913, + "step": 49386 + }, + { + "epoch": 1.9, + "learning_rate": 1.83656065657861e-05, + "loss": 0.347, + "step": 49387 + }, + { + "epoch": 1.9, + "learning_rate": 1.836496537573737e-05, + "loss": 0.2844, + "step": 49388 + }, + { + "epoch": 1.9, + "learning_rate": 1.8364324185688638e-05, + "loss": 0.3377, + "step": 49389 + }, + { + "epoch": 1.9, + "learning_rate": 1.8363682995639907e-05, + "loss": 0.3477, + "step": 49390 + }, + { + "epoch": 1.9, + "learning_rate": 1.8363041805591178e-05, + "loss": 0.3149, + "step": 49391 + }, + { + "epoch": 1.9, + "learning_rate": 1.8362400615542447e-05, + "loss": 0.2639, + "step": 49392 + }, + { + "epoch": 1.9, + "learning_rate": 1.8361759425493715e-05, + "loss": 0.4279, + "step": 49393 + }, + { + "epoch": 1.9, + "learning_rate": 1.8361118235444987e-05, + "loss": 0.2441, + "step": 49394 + }, + { + "epoch": 1.9, + "learning_rate": 1.836047704539626e-05, + "loss": 0.2967, + "step": 49395 + }, + { + "epoch": 1.9, + "learning_rate": 1.8359835855347527e-05, + "loss": 0.4967, + "step": 49396 + }, + { + "epoch": 1.9, + "learning_rate": 1.8359194665298795e-05, + "loss": 0.4613, + "step": 49397 + }, + { + "epoch": 1.9, + "learning_rate": 1.8358553475250063e-05, + "loss": 0.4728, + "step": 49398 + }, + { + "epoch": 1.9, + "learning_rate": 1.8357912285201335e-05, + "loss": 0.3037, + "step": 49399 + }, + { + "epoch": 1.9, + "learning_rate": 1.8357271095152603e-05, + "loss": 0.2888, + "step": 49400 + }, + { + "epoch": 1.9, + "learning_rate": 1.8356629905103872e-05, + "loss": 0.4417, + "step": 49401 + }, + { + "epoch": 1.9, + "learning_rate": 1.8355988715055143e-05, + "loss": 0.3317, + "step": 49402 + }, + { + "epoch": 1.9, + "learning_rate": 1.8355347525006415e-05, + "loss": 0.5404, + "step": 49403 + }, + { + "epoch": 1.9, + "learning_rate": 1.8354706334957684e-05, + "loss": 0.245, + "step": 49404 + }, + { + "epoch": 1.9, + "learning_rate": 1.8354065144908952e-05, + "loss": 0.3177, + "step": 49405 + }, + { + "epoch": 1.9, + "learning_rate": 1.835342395486022e-05, + "loss": 0.3782, + "step": 49406 + }, + { + "epoch": 1.9, + "learning_rate": 1.8352782764811492e-05, + "loss": 0.3295, + "step": 49407 + }, + { + "epoch": 1.9, + "learning_rate": 1.835214157476276e-05, + "loss": 0.3528, + "step": 49408 + }, + { + "epoch": 1.9, + "learning_rate": 1.835150038471403e-05, + "loss": 0.3423, + "step": 49409 + }, + { + "epoch": 1.9, + "learning_rate": 1.83508591946653e-05, + "loss": 0.3526, + "step": 49410 + }, + { + "epoch": 1.9, + "learning_rate": 1.8350218004616572e-05, + "loss": 0.3489, + "step": 49411 + }, + { + "epoch": 1.9, + "learning_rate": 1.834957681456784e-05, + "loss": 0.2936, + "step": 49412 + }, + { + "epoch": 1.9, + "learning_rate": 1.834893562451911e-05, + "loss": 0.3182, + "step": 49413 + }, + { + "epoch": 1.9, + "learning_rate": 1.8348294434470377e-05, + "loss": 0.3924, + "step": 49414 + }, + { + "epoch": 1.9, + "learning_rate": 1.834765324442165e-05, + "loss": 0.2651, + "step": 49415 + }, + { + "epoch": 1.9, + "learning_rate": 1.8347012054372917e-05, + "loss": 0.3671, + "step": 49416 + }, + { + "epoch": 1.9, + "learning_rate": 1.8346370864324185e-05, + "loss": 0.1819, + "step": 49417 + }, + { + "epoch": 1.9, + "learning_rate": 1.8345729674275454e-05, + "loss": 0.2677, + "step": 49418 + }, + { + "epoch": 1.9, + "learning_rate": 1.8345088484226725e-05, + "loss": 0.4582, + "step": 49419 + }, + { + "epoch": 1.9, + "learning_rate": 1.8344447294177997e-05, + "loss": 0.2781, + "step": 49420 + }, + { + "epoch": 1.9, + "learning_rate": 1.8343806104129265e-05, + "loss": 0.5632, + "step": 49421 + }, + { + "epoch": 1.9, + "learning_rate": 1.8343164914080534e-05, + "loss": 0.2615, + "step": 49422 + }, + { + "epoch": 1.9, + "learning_rate": 1.8342523724031805e-05, + "loss": 0.2816, + "step": 49423 + }, + { + "epoch": 1.9, + "learning_rate": 1.8341882533983074e-05, + "loss": 0.3724, + "step": 49424 + }, + { + "epoch": 1.9, + "learning_rate": 1.8341241343934342e-05, + "loss": 0.2212, + "step": 49425 + }, + { + "epoch": 1.9, + "learning_rate": 1.834060015388561e-05, + "loss": 0.2893, + "step": 49426 + }, + { + "epoch": 1.9, + "learning_rate": 1.8339958963836882e-05, + "loss": 0.3018, + "step": 49427 + }, + { + "epoch": 1.9, + "learning_rate": 1.833931777378815e-05, + "loss": 0.2601, + "step": 49428 + }, + { + "epoch": 1.9, + "learning_rate": 1.8338676583739422e-05, + "loss": 0.298, + "step": 49429 + }, + { + "epoch": 1.9, + "learning_rate": 1.833803539369069e-05, + "loss": 0.3232, + "step": 49430 + }, + { + "epoch": 1.9, + "learning_rate": 1.8337394203641962e-05, + "loss": 0.2765, + "step": 49431 + }, + { + "epoch": 1.9, + "learning_rate": 1.833675301359323e-05, + "loss": 0.2524, + "step": 49432 + }, + { + "epoch": 1.9, + "learning_rate": 1.83361118235445e-05, + "loss": 0.3555, + "step": 49433 + }, + { + "epoch": 1.9, + "learning_rate": 1.8335470633495767e-05, + "loss": 0.2969, + "step": 49434 + }, + { + "epoch": 1.9, + "learning_rate": 1.833482944344704e-05, + "loss": 0.3591, + "step": 49435 + }, + { + "epoch": 1.9, + "learning_rate": 1.8334188253398307e-05, + "loss": 0.3024, + "step": 49436 + }, + { + "epoch": 1.9, + "learning_rate": 1.8333547063349576e-05, + "loss": 0.4265, + "step": 49437 + }, + { + "epoch": 1.9, + "learning_rate": 1.8332905873300847e-05, + "loss": 0.235, + "step": 49438 + }, + { + "epoch": 1.9, + "learning_rate": 1.833226468325212e-05, + "loss": 0.3198, + "step": 49439 + }, + { + "epoch": 1.9, + "learning_rate": 1.8331623493203387e-05, + "loss": 0.3291, + "step": 49440 + }, + { + "epoch": 1.9, + "learning_rate": 1.8330982303154656e-05, + "loss": 0.3718, + "step": 49441 + }, + { + "epoch": 1.9, + "learning_rate": 1.8330341113105924e-05, + "loss": 0.2852, + "step": 49442 + }, + { + "epoch": 1.9, + "learning_rate": 1.8329699923057196e-05, + "loss": 0.4349, + "step": 49443 + }, + { + "epoch": 1.9, + "learning_rate": 1.8329058733008464e-05, + "loss": 0.2765, + "step": 49444 + }, + { + "epoch": 1.9, + "learning_rate": 1.8328417542959732e-05, + "loss": 0.263, + "step": 49445 + }, + { + "epoch": 1.9, + "learning_rate": 1.8327776352911004e-05, + "loss": 0.3916, + "step": 49446 + }, + { + "epoch": 1.9, + "learning_rate": 1.8327135162862276e-05, + "loss": 0.2428, + "step": 49447 + }, + { + "epoch": 1.9, + "learning_rate": 1.8326493972813544e-05, + "loss": 0.253, + "step": 49448 + }, + { + "epoch": 1.9, + "learning_rate": 1.8325852782764813e-05, + "loss": 0.3868, + "step": 49449 + }, + { + "epoch": 1.9, + "learning_rate": 1.832521159271608e-05, + "loss": 0.3066, + "step": 49450 + }, + { + "epoch": 1.9, + "learning_rate": 1.8324570402667353e-05, + "loss": 0.2283, + "step": 49451 + }, + { + "epoch": 1.9, + "learning_rate": 1.832392921261862e-05, + "loss": 0.3922, + "step": 49452 + }, + { + "epoch": 1.9, + "learning_rate": 1.832328802256989e-05, + "loss": 0.2675, + "step": 49453 + }, + { + "epoch": 1.9, + "learning_rate": 1.8322646832521158e-05, + "loss": 0.2662, + "step": 49454 + }, + { + "epoch": 1.9, + "learning_rate": 1.832200564247243e-05, + "loss": 0.2936, + "step": 49455 + }, + { + "epoch": 1.9, + "learning_rate": 1.83213644524237e-05, + "loss": 0.4488, + "step": 49456 + }, + { + "epoch": 1.9, + "learning_rate": 1.832072326237497e-05, + "loss": 0.3681, + "step": 49457 + }, + { + "epoch": 1.9, + "learning_rate": 1.8320082072326238e-05, + "loss": 0.3325, + "step": 49458 + }, + { + "epoch": 1.9, + "learning_rate": 1.831944088227751e-05, + "loss": 0.2712, + "step": 49459 + }, + { + "epoch": 1.9, + "learning_rate": 1.8318799692228778e-05, + "loss": 0.5258, + "step": 49460 + }, + { + "epoch": 1.9, + "learning_rate": 1.8318158502180046e-05, + "loss": 0.3355, + "step": 49461 + }, + { + "epoch": 1.9, + "learning_rate": 1.8317517312131314e-05, + "loss": 0.3092, + "step": 49462 + }, + { + "epoch": 1.9, + "learning_rate": 1.8316876122082586e-05, + "loss": 0.3221, + "step": 49463 + }, + { + "epoch": 1.9, + "learning_rate": 1.8316234932033854e-05, + "loss": 0.392, + "step": 49464 + }, + { + "epoch": 1.9, + "learning_rate": 1.8315593741985126e-05, + "loss": 0.2547, + "step": 49465 + }, + { + "epoch": 1.9, + "learning_rate": 1.8314952551936394e-05, + "loss": 0.2057, + "step": 49466 + }, + { + "epoch": 1.9, + "learning_rate": 1.8314311361887666e-05, + "loss": 0.2732, + "step": 49467 + }, + { + "epoch": 1.9, + "learning_rate": 1.8313670171838934e-05, + "loss": 0.307, + "step": 49468 + }, + { + "epoch": 1.9, + "learning_rate": 1.8313028981790203e-05, + "loss": 0.305, + "step": 49469 + }, + { + "epoch": 1.9, + "learning_rate": 1.831238779174147e-05, + "loss": 0.4783, + "step": 49470 + }, + { + "epoch": 1.9, + "learning_rate": 1.8311746601692743e-05, + "loss": 0.4171, + "step": 49471 + }, + { + "epoch": 1.9, + "learning_rate": 1.831110541164401e-05, + "loss": 0.3783, + "step": 49472 + }, + { + "epoch": 1.9, + "learning_rate": 1.831046422159528e-05, + "loss": 0.3407, + "step": 49473 + }, + { + "epoch": 1.9, + "learning_rate": 1.830982303154655e-05, + "loss": 0.3746, + "step": 49474 + }, + { + "epoch": 1.9, + "learning_rate": 1.8309181841497823e-05, + "loss": 0.2097, + "step": 49475 + }, + { + "epoch": 1.9, + "learning_rate": 1.830854065144909e-05, + "loss": 0.27, + "step": 49476 + }, + { + "epoch": 1.9, + "learning_rate": 1.830789946140036e-05, + "loss": 0.2262, + "step": 49477 + }, + { + "epoch": 1.9, + "learning_rate": 1.8307258271351628e-05, + "loss": 0.3903, + "step": 49478 + }, + { + "epoch": 1.9, + "learning_rate": 1.83066170813029e-05, + "loss": 0.3606, + "step": 49479 + }, + { + "epoch": 1.9, + "learning_rate": 1.8305975891254168e-05, + "loss": 0.2133, + "step": 49480 + }, + { + "epoch": 1.9, + "learning_rate": 1.8305334701205436e-05, + "loss": 0.2071, + "step": 49481 + }, + { + "epoch": 1.9, + "learning_rate": 1.8304693511156708e-05, + "loss": 0.3989, + "step": 49482 + }, + { + "epoch": 1.9, + "learning_rate": 1.830405232110798e-05, + "loss": 0.3638, + "step": 49483 + }, + { + "epoch": 1.9, + "learning_rate": 1.8303411131059248e-05, + "loss": 0.4201, + "step": 49484 + }, + { + "epoch": 1.9, + "learning_rate": 1.8302769941010516e-05, + "loss": 0.3047, + "step": 49485 + }, + { + "epoch": 1.9, + "learning_rate": 1.8302128750961785e-05, + "loss": 0.4735, + "step": 49486 + }, + { + "epoch": 1.9, + "learning_rate": 1.8301487560913056e-05, + "loss": 0.2005, + "step": 49487 + }, + { + "epoch": 1.9, + "learning_rate": 1.8300846370864325e-05, + "loss": 0.3823, + "step": 49488 + }, + { + "epoch": 1.9, + "learning_rate": 1.8300205180815593e-05, + "loss": 0.3471, + "step": 49489 + }, + { + "epoch": 1.9, + "learning_rate": 1.8299563990766865e-05, + "loss": 0.3393, + "step": 49490 + }, + { + "epoch": 1.9, + "learning_rate": 1.8298922800718133e-05, + "loss": 0.2424, + "step": 49491 + }, + { + "epoch": 1.9, + "learning_rate": 1.8298281610669405e-05, + "loss": 0.2545, + "step": 49492 + }, + { + "epoch": 1.9, + "learning_rate": 1.8297640420620673e-05, + "loss": 0.2966, + "step": 49493 + }, + { + "epoch": 1.9, + "learning_rate": 1.829699923057194e-05, + "loss": 0.3293, + "step": 49494 + }, + { + "epoch": 1.9, + "learning_rate": 1.8296358040523213e-05, + "loss": 0.2578, + "step": 49495 + }, + { + "epoch": 1.9, + "learning_rate": 1.829571685047448e-05, + "loss": 0.3215, + "step": 49496 + }, + { + "epoch": 1.9, + "learning_rate": 1.829507566042575e-05, + "loss": 0.3879, + "step": 49497 + }, + { + "epoch": 1.9, + "learning_rate": 1.8294434470377018e-05, + "loss": 0.4484, + "step": 49498 + }, + { + "epoch": 1.9, + "learning_rate": 1.829379328032829e-05, + "loss": 0.3214, + "step": 49499 + }, + { + "epoch": 1.9, + "learning_rate": 1.829315209027956e-05, + "loss": 0.224, + "step": 49500 + }, + { + "epoch": 1.9, + "learning_rate": 1.829251090023083e-05, + "loss": 0.3563, + "step": 49501 + }, + { + "epoch": 1.9, + "learning_rate": 1.8291869710182098e-05, + "loss": 0.4013, + "step": 49502 + }, + { + "epoch": 1.9, + "learning_rate": 1.829122852013337e-05, + "loss": 0.2765, + "step": 49503 + }, + { + "epoch": 1.9, + "learning_rate": 1.829058733008464e-05, + "loss": 0.1914, + "step": 49504 + }, + { + "epoch": 1.9, + "learning_rate": 1.8289946140035907e-05, + "loss": 0.2565, + "step": 49505 + }, + { + "epoch": 1.9, + "learning_rate": 1.8289304949987175e-05, + "loss": 0.3055, + "step": 49506 + }, + { + "epoch": 1.9, + "learning_rate": 1.8288663759938447e-05, + "loss": 0.3863, + "step": 49507 + }, + { + "epoch": 1.9, + "learning_rate": 1.8288022569889715e-05, + "loss": 0.3184, + "step": 49508 + }, + { + "epoch": 1.9, + "learning_rate": 1.8287381379840987e-05, + "loss": 0.1883, + "step": 49509 + }, + { + "epoch": 1.9, + "learning_rate": 1.8286740189792255e-05, + "loss": 0.3807, + "step": 49510 + }, + { + "epoch": 1.9, + "learning_rate": 1.8286098999743527e-05, + "loss": 0.3318, + "step": 49511 + }, + { + "epoch": 1.9, + "learning_rate": 1.8285457809694795e-05, + "loss": 0.4001, + "step": 49512 + }, + { + "epoch": 1.9, + "learning_rate": 1.8284816619646063e-05, + "loss": 0.2954, + "step": 49513 + }, + { + "epoch": 1.9, + "learning_rate": 1.8284175429597332e-05, + "loss": 0.2409, + "step": 49514 + }, + { + "epoch": 1.9, + "learning_rate": 1.8283534239548603e-05, + "loss": 0.3383, + "step": 49515 + }, + { + "epoch": 1.9, + "learning_rate": 1.8282893049499872e-05, + "loss": 0.2231, + "step": 49516 + }, + { + "epoch": 1.9, + "learning_rate": 1.828225185945114e-05, + "loss": 0.3258, + "step": 49517 + }, + { + "epoch": 1.9, + "learning_rate": 1.8281610669402412e-05, + "loss": 0.2508, + "step": 49518 + }, + { + "epoch": 1.9, + "learning_rate": 1.8280969479353684e-05, + "loss": 0.284, + "step": 49519 + }, + { + "epoch": 1.9, + "learning_rate": 1.8280328289304952e-05, + "loss": 0.2682, + "step": 49520 + }, + { + "epoch": 1.9, + "learning_rate": 1.827968709925622e-05, + "loss": 0.4239, + "step": 49521 + }, + { + "epoch": 1.9, + "learning_rate": 1.827904590920749e-05, + "loss": 0.3476, + "step": 49522 + }, + { + "epoch": 1.9, + "learning_rate": 1.827840471915876e-05, + "loss": 0.267, + "step": 49523 + }, + { + "epoch": 1.9, + "learning_rate": 1.827776352911003e-05, + "loss": 0.3066, + "step": 49524 + }, + { + "epoch": 1.9, + "learning_rate": 1.8277122339061297e-05, + "loss": 0.3503, + "step": 49525 + }, + { + "epoch": 1.9, + "learning_rate": 1.827648114901257e-05, + "loss": 0.3023, + "step": 49526 + }, + { + "epoch": 1.9, + "learning_rate": 1.827583995896384e-05, + "loss": 0.4181, + "step": 49527 + }, + { + "epoch": 1.9, + "learning_rate": 1.827519876891511e-05, + "loss": 0.2198, + "step": 49528 + }, + { + "epoch": 1.9, + "learning_rate": 1.8274557578866377e-05, + "loss": 0.3035, + "step": 49529 + }, + { + "epoch": 1.9, + "learning_rate": 1.8273916388817645e-05, + "loss": 0.2635, + "step": 49530 + }, + { + "epoch": 1.9, + "learning_rate": 1.8273275198768917e-05, + "loss": 0.2412, + "step": 49531 + }, + { + "epoch": 1.9, + "learning_rate": 1.8272634008720185e-05, + "loss": 0.3254, + "step": 49532 + }, + { + "epoch": 1.9, + "learning_rate": 1.8271992818671454e-05, + "loss": 0.4342, + "step": 49533 + }, + { + "epoch": 1.9, + "learning_rate": 1.8271351628622725e-05, + "loss": 0.2173, + "step": 49534 + }, + { + "epoch": 1.9, + "learning_rate": 1.8270710438573994e-05, + "loss": 0.2891, + "step": 49535 + }, + { + "epoch": 1.9, + "learning_rate": 1.8270069248525265e-05, + "loss": 0.2974, + "step": 49536 + }, + { + "epoch": 1.9, + "learning_rate": 1.8269428058476534e-05, + "loss": 0.2852, + "step": 49537 + }, + { + "epoch": 1.9, + "learning_rate": 1.8268786868427802e-05, + "loss": 0.281, + "step": 49538 + }, + { + "epoch": 1.9, + "learning_rate": 1.8268145678379074e-05, + "loss": 0.3857, + "step": 49539 + }, + { + "epoch": 1.9, + "learning_rate": 1.8267504488330342e-05, + "loss": 0.3965, + "step": 49540 + }, + { + "epoch": 1.9, + "learning_rate": 1.826686329828161e-05, + "loss": 0.2649, + "step": 49541 + }, + { + "epoch": 1.9, + "learning_rate": 1.826622210823288e-05, + "loss": 0.4822, + "step": 49542 + }, + { + "epoch": 1.9, + "learning_rate": 1.826558091818415e-05, + "loss": 0.3063, + "step": 49543 + }, + { + "epoch": 1.9, + "learning_rate": 1.826493972813542e-05, + "loss": 0.2531, + "step": 49544 + }, + { + "epoch": 1.9, + "learning_rate": 1.826429853808669e-05, + "loss": 0.2871, + "step": 49545 + }, + { + "epoch": 1.9, + "learning_rate": 1.826365734803796e-05, + "loss": 0.3517, + "step": 49546 + }, + { + "epoch": 1.9, + "learning_rate": 1.826301615798923e-05, + "loss": 0.3324, + "step": 49547 + }, + { + "epoch": 1.9, + "learning_rate": 1.82623749679405e-05, + "loss": 0.3262, + "step": 49548 + }, + { + "epoch": 1.9, + "learning_rate": 1.8261733777891767e-05, + "loss": 0.3522, + "step": 49549 + }, + { + "epoch": 1.91, + "learning_rate": 1.8261092587843036e-05, + "loss": 0.281, + "step": 49550 + }, + { + "epoch": 1.91, + "learning_rate": 1.8260451397794307e-05, + "loss": 0.3544, + "step": 49551 + }, + { + "epoch": 1.91, + "learning_rate": 1.8259810207745576e-05, + "loss": 0.3151, + "step": 49552 + }, + { + "epoch": 1.91, + "learning_rate": 1.8259169017696844e-05, + "loss": 0.1969, + "step": 49553 + }, + { + "epoch": 1.91, + "learning_rate": 1.8258527827648116e-05, + "loss": 0.2646, + "step": 49554 + }, + { + "epoch": 1.91, + "learning_rate": 1.8257886637599387e-05, + "loss": 0.3492, + "step": 49555 + }, + { + "epoch": 1.91, + "learning_rate": 1.8257245447550656e-05, + "loss": 0.2771, + "step": 49556 + }, + { + "epoch": 1.91, + "learning_rate": 1.8256604257501924e-05, + "loss": 0.2577, + "step": 49557 + }, + { + "epoch": 1.91, + "learning_rate": 1.8255963067453192e-05, + "loss": 0.2114, + "step": 49558 + }, + { + "epoch": 1.91, + "learning_rate": 1.8255321877404464e-05, + "loss": 0.2402, + "step": 49559 + }, + { + "epoch": 1.91, + "learning_rate": 1.8254680687355732e-05, + "loss": 0.3248, + "step": 49560 + }, + { + "epoch": 1.91, + "learning_rate": 1.8254039497307e-05, + "loss": 0.4093, + "step": 49561 + }, + { + "epoch": 1.91, + "learning_rate": 1.8253398307258273e-05, + "loss": 0.2842, + "step": 49562 + }, + { + "epoch": 1.91, + "learning_rate": 1.8252757117209544e-05, + "loss": 0.2304, + "step": 49563 + }, + { + "epoch": 1.91, + "learning_rate": 1.8252115927160813e-05, + "loss": 0.2816, + "step": 49564 + }, + { + "epoch": 1.91, + "learning_rate": 1.825147473711208e-05, + "loss": 0.2624, + "step": 49565 + }, + { + "epoch": 1.91, + "learning_rate": 1.825083354706335e-05, + "loss": 0.2336, + "step": 49566 + }, + { + "epoch": 1.91, + "learning_rate": 1.825019235701462e-05, + "loss": 0.295, + "step": 49567 + }, + { + "epoch": 1.91, + "learning_rate": 1.824955116696589e-05, + "loss": 0.3424, + "step": 49568 + }, + { + "epoch": 1.91, + "learning_rate": 1.8248909976917158e-05, + "loss": 0.2497, + "step": 49569 + }, + { + "epoch": 1.91, + "learning_rate": 1.824826878686843e-05, + "loss": 0.2567, + "step": 49570 + }, + { + "epoch": 1.91, + "learning_rate": 1.8247627596819698e-05, + "loss": 0.2086, + "step": 49571 + }, + { + "epoch": 1.91, + "learning_rate": 1.824698640677097e-05, + "loss": 0.3648, + "step": 49572 + }, + { + "epoch": 1.91, + "learning_rate": 1.8246345216722238e-05, + "loss": 0.2239, + "step": 49573 + }, + { + "epoch": 1.91, + "learning_rate": 1.8245704026673506e-05, + "loss": 0.2496, + "step": 49574 + }, + { + "epoch": 1.91, + "learning_rate": 1.8245062836624778e-05, + "loss": 0.1893, + "step": 49575 + }, + { + "epoch": 1.91, + "learning_rate": 1.8244421646576046e-05, + "loss": 0.3338, + "step": 49576 + }, + { + "epoch": 1.91, + "learning_rate": 1.8243780456527314e-05, + "loss": 0.2605, + "step": 49577 + }, + { + "epoch": 1.91, + "learning_rate": 1.8243139266478583e-05, + "loss": 0.4226, + "step": 49578 + }, + { + "epoch": 1.91, + "learning_rate": 1.8242498076429854e-05, + "loss": 0.2968, + "step": 49579 + }, + { + "epoch": 1.91, + "learning_rate": 1.8241856886381123e-05, + "loss": 0.2593, + "step": 49580 + }, + { + "epoch": 1.91, + "learning_rate": 1.8241215696332394e-05, + "loss": 0.1722, + "step": 49581 + }, + { + "epoch": 1.91, + "learning_rate": 1.8240574506283663e-05, + "loss": 0.2663, + "step": 49582 + }, + { + "epoch": 1.91, + "learning_rate": 1.8239933316234935e-05, + "loss": 0.3814, + "step": 49583 + }, + { + "epoch": 1.91, + "learning_rate": 1.8239292126186203e-05, + "loss": 0.3046, + "step": 49584 + }, + { + "epoch": 1.91, + "learning_rate": 1.823865093613747e-05, + "loss": 0.3123, + "step": 49585 + }, + { + "epoch": 1.91, + "learning_rate": 1.823800974608874e-05, + "loss": 0.2281, + "step": 49586 + }, + { + "epoch": 1.91, + "learning_rate": 1.823736855604001e-05, + "loss": 0.2775, + "step": 49587 + }, + { + "epoch": 1.91, + "learning_rate": 1.823672736599128e-05, + "loss": 0.2703, + "step": 49588 + }, + { + "epoch": 1.91, + "learning_rate": 1.8236086175942548e-05, + "loss": 0.2712, + "step": 49589 + }, + { + "epoch": 1.91, + "learning_rate": 1.823544498589382e-05, + "loss": 0.3052, + "step": 49590 + }, + { + "epoch": 1.91, + "learning_rate": 1.823480379584509e-05, + "loss": 0.2815, + "step": 49591 + }, + { + "epoch": 1.91, + "learning_rate": 1.823416260579636e-05, + "loss": 0.3801, + "step": 49592 + }, + { + "epoch": 1.91, + "learning_rate": 1.8233521415747628e-05, + "loss": 0.2566, + "step": 49593 + }, + { + "epoch": 1.91, + "learning_rate": 1.8232880225698896e-05, + "loss": 0.3629, + "step": 49594 + }, + { + "epoch": 1.91, + "learning_rate": 1.8232239035650168e-05, + "loss": 0.4805, + "step": 49595 + }, + { + "epoch": 1.91, + "learning_rate": 1.8231597845601436e-05, + "loss": 0.3176, + "step": 49596 + }, + { + "epoch": 1.91, + "learning_rate": 1.8230956655552705e-05, + "loss": 0.2957, + "step": 49597 + }, + { + "epoch": 1.91, + "learning_rate": 1.8230315465503976e-05, + "loss": 0.3593, + "step": 49598 + }, + { + "epoch": 1.91, + "learning_rate": 1.8229674275455248e-05, + "loss": 0.2642, + "step": 49599 + }, + { + "epoch": 1.91, + "learning_rate": 1.8229033085406516e-05, + "loss": 0.3119, + "step": 49600 + }, + { + "epoch": 1.91, + "learning_rate": 1.8228391895357785e-05, + "loss": 0.3006, + "step": 49601 + }, + { + "epoch": 1.91, + "learning_rate": 1.8227750705309053e-05, + "loss": 0.1717, + "step": 49602 + }, + { + "epoch": 1.91, + "learning_rate": 1.8227109515260325e-05, + "loss": 0.2262, + "step": 49603 + }, + { + "epoch": 1.91, + "learning_rate": 1.8226468325211593e-05, + "loss": 0.4011, + "step": 49604 + }, + { + "epoch": 1.91, + "learning_rate": 1.822582713516286e-05, + "loss": 0.2435, + "step": 49605 + }, + { + "epoch": 1.91, + "learning_rate": 1.8225185945114133e-05, + "loss": 0.3077, + "step": 49606 + }, + { + "epoch": 1.91, + "learning_rate": 1.8224544755065405e-05, + "loss": 0.3592, + "step": 49607 + }, + { + "epoch": 1.91, + "learning_rate": 1.8223903565016673e-05, + "loss": 0.2825, + "step": 49608 + }, + { + "epoch": 1.91, + "learning_rate": 1.822326237496794e-05, + "loss": 0.4198, + "step": 49609 + }, + { + "epoch": 1.91, + "learning_rate": 1.822262118491921e-05, + "loss": 0.3262, + "step": 49610 + }, + { + "epoch": 1.91, + "learning_rate": 1.822197999487048e-05, + "loss": 0.2603, + "step": 49611 + }, + { + "epoch": 1.91, + "learning_rate": 1.822133880482175e-05, + "loss": 0.2378, + "step": 49612 + }, + { + "epoch": 1.91, + "learning_rate": 1.8220697614773018e-05, + "loss": 0.2605, + "step": 49613 + }, + { + "epoch": 1.91, + "learning_rate": 1.822005642472429e-05, + "loss": 0.4142, + "step": 49614 + }, + { + "epoch": 1.91, + "learning_rate": 1.821941523467556e-05, + "loss": 0.2219, + "step": 49615 + }, + { + "epoch": 1.91, + "learning_rate": 1.821877404462683e-05, + "loss": 0.3228, + "step": 49616 + }, + { + "epoch": 1.91, + "learning_rate": 1.82181328545781e-05, + "loss": 0.2836, + "step": 49617 + }, + { + "epoch": 1.91, + "learning_rate": 1.8217491664529367e-05, + "loss": 0.1953, + "step": 49618 + }, + { + "epoch": 1.91, + "learning_rate": 1.821685047448064e-05, + "loss": 0.3406, + "step": 49619 + }, + { + "epoch": 1.91, + "learning_rate": 1.8216209284431907e-05, + "loss": 0.3351, + "step": 49620 + }, + { + "epoch": 1.91, + "learning_rate": 1.8215568094383175e-05, + "loss": 0.3614, + "step": 49621 + }, + { + "epoch": 1.91, + "learning_rate": 1.8214926904334443e-05, + "loss": 0.2489, + "step": 49622 + }, + { + "epoch": 1.91, + "learning_rate": 1.8214285714285715e-05, + "loss": 0.2645, + "step": 49623 + }, + { + "epoch": 1.91, + "learning_rate": 1.8213644524236983e-05, + "loss": 0.2037, + "step": 49624 + }, + { + "epoch": 1.91, + "learning_rate": 1.8213003334188255e-05, + "loss": 0.3422, + "step": 49625 + }, + { + "epoch": 1.91, + "learning_rate": 1.8212362144139523e-05, + "loss": 0.3514, + "step": 49626 + }, + { + "epoch": 1.91, + "learning_rate": 1.8211720954090795e-05, + "loss": 0.3245, + "step": 49627 + }, + { + "epoch": 1.91, + "learning_rate": 1.8211079764042064e-05, + "loss": 0.2947, + "step": 49628 + }, + { + "epoch": 1.91, + "learning_rate": 1.8210438573993332e-05, + "loss": 0.3193, + "step": 49629 + }, + { + "epoch": 1.91, + "learning_rate": 1.82097973839446e-05, + "loss": 0.4094, + "step": 49630 + }, + { + "epoch": 1.91, + "learning_rate": 1.8209156193895872e-05, + "loss": 0.357, + "step": 49631 + }, + { + "epoch": 1.91, + "learning_rate": 1.820851500384714e-05, + "loss": 0.2414, + "step": 49632 + }, + { + "epoch": 1.91, + "learning_rate": 1.820787381379841e-05, + "loss": 0.3779, + "step": 49633 + }, + { + "epoch": 1.91, + "learning_rate": 1.820723262374968e-05, + "loss": 0.5441, + "step": 49634 + }, + { + "epoch": 1.91, + "learning_rate": 1.8206591433700952e-05, + "loss": 0.3044, + "step": 49635 + }, + { + "epoch": 1.91, + "learning_rate": 1.820595024365222e-05, + "loss": 0.2298, + "step": 49636 + }, + { + "epoch": 1.91, + "learning_rate": 1.820530905360349e-05, + "loss": 0.2364, + "step": 49637 + }, + { + "epoch": 1.91, + "learning_rate": 1.8204667863554757e-05, + "loss": 0.3391, + "step": 49638 + }, + { + "epoch": 1.91, + "learning_rate": 1.820402667350603e-05, + "loss": 0.2887, + "step": 49639 + }, + { + "epoch": 1.91, + "learning_rate": 1.8203385483457297e-05, + "loss": 0.3522, + "step": 49640 + }, + { + "epoch": 1.91, + "learning_rate": 1.8202744293408565e-05, + "loss": 0.3208, + "step": 49641 + }, + { + "epoch": 1.91, + "learning_rate": 1.8202103103359837e-05, + "loss": 0.2581, + "step": 49642 + }, + { + "epoch": 1.91, + "learning_rate": 1.820146191331111e-05, + "loss": 0.2792, + "step": 49643 + }, + { + "epoch": 1.91, + "learning_rate": 1.8200820723262377e-05, + "loss": 0.2419, + "step": 49644 + }, + { + "epoch": 1.91, + "learning_rate": 1.8200179533213645e-05, + "loss": 0.3617, + "step": 49645 + }, + { + "epoch": 1.91, + "learning_rate": 1.8199538343164914e-05, + "loss": 0.4621, + "step": 49646 + }, + { + "epoch": 1.91, + "learning_rate": 1.8198897153116185e-05, + "loss": 0.2817, + "step": 49647 + }, + { + "epoch": 1.91, + "learning_rate": 1.8198255963067454e-05, + "loss": 0.2943, + "step": 49648 + }, + { + "epoch": 1.91, + "learning_rate": 1.8197614773018722e-05, + "loss": 0.3345, + "step": 49649 + }, + { + "epoch": 1.91, + "learning_rate": 1.8196973582969994e-05, + "loss": 0.3663, + "step": 49650 + }, + { + "epoch": 1.91, + "learning_rate": 1.8196332392921262e-05, + "loss": 0.291, + "step": 49651 + }, + { + "epoch": 1.91, + "learning_rate": 1.8195691202872534e-05, + "loss": 0.345, + "step": 49652 + }, + { + "epoch": 1.91, + "learning_rate": 1.8195050012823802e-05, + "loss": 0.2762, + "step": 49653 + }, + { + "epoch": 1.91, + "learning_rate": 1.819440882277507e-05, + "loss": 0.2763, + "step": 49654 + }, + { + "epoch": 1.91, + "learning_rate": 1.8193767632726342e-05, + "loss": 0.3015, + "step": 49655 + }, + { + "epoch": 1.91, + "learning_rate": 1.819312644267761e-05, + "loss": 0.4775, + "step": 49656 + }, + { + "epoch": 1.91, + "learning_rate": 1.819248525262888e-05, + "loss": 0.288, + "step": 49657 + }, + { + "epoch": 1.91, + "learning_rate": 1.8191844062580147e-05, + "loss": 0.3634, + "step": 49658 + }, + { + "epoch": 1.91, + "learning_rate": 1.819120287253142e-05, + "loss": 0.4427, + "step": 49659 + }, + { + "epoch": 1.91, + "learning_rate": 1.8190561682482687e-05, + "loss": 0.2184, + "step": 49660 + }, + { + "epoch": 1.91, + "learning_rate": 1.818992049243396e-05, + "loss": 0.405, + "step": 49661 + }, + { + "epoch": 1.91, + "learning_rate": 1.8189279302385227e-05, + "loss": 0.2559, + "step": 49662 + }, + { + "epoch": 1.91, + "learning_rate": 1.81886381123365e-05, + "loss": 0.425, + "step": 49663 + }, + { + "epoch": 1.91, + "learning_rate": 1.8187996922287767e-05, + "loss": 0.3356, + "step": 49664 + }, + { + "epoch": 1.91, + "learning_rate": 1.8187355732239036e-05, + "loss": 0.2676, + "step": 49665 + }, + { + "epoch": 1.91, + "learning_rate": 1.8186714542190304e-05, + "loss": 0.2307, + "step": 49666 + }, + { + "epoch": 1.91, + "learning_rate": 1.8186073352141576e-05, + "loss": 0.4644, + "step": 49667 + }, + { + "epoch": 1.91, + "learning_rate": 1.8185432162092844e-05, + "loss": 0.2577, + "step": 49668 + }, + { + "epoch": 1.91, + "learning_rate": 1.8184790972044112e-05, + "loss": 0.385, + "step": 49669 + }, + { + "epoch": 1.91, + "learning_rate": 1.8184149781995384e-05, + "loss": 0.4079, + "step": 49670 + }, + { + "epoch": 1.91, + "learning_rate": 1.8183508591946656e-05, + "loss": 0.3334, + "step": 49671 + }, + { + "epoch": 1.91, + "learning_rate": 1.8182867401897924e-05, + "loss": 0.477, + "step": 49672 + }, + { + "epoch": 1.91, + "learning_rate": 1.8182226211849193e-05, + "loss": 0.3857, + "step": 49673 + }, + { + "epoch": 1.91, + "learning_rate": 1.818158502180046e-05, + "loss": 0.2971, + "step": 49674 + }, + { + "epoch": 1.91, + "learning_rate": 1.8180943831751733e-05, + "loss": 0.292, + "step": 49675 + }, + { + "epoch": 1.91, + "learning_rate": 1.8180302641703e-05, + "loss": 0.3388, + "step": 49676 + }, + { + "epoch": 1.91, + "learning_rate": 1.817966145165427e-05, + "loss": 0.3551, + "step": 49677 + }, + { + "epoch": 1.91, + "learning_rate": 1.817902026160554e-05, + "loss": 0.3282, + "step": 49678 + }, + { + "epoch": 1.91, + "learning_rate": 1.8178379071556813e-05, + "loss": 0.3579, + "step": 49679 + }, + { + "epoch": 1.91, + "learning_rate": 1.817773788150808e-05, + "loss": 0.4794, + "step": 49680 + }, + { + "epoch": 1.91, + "learning_rate": 1.817709669145935e-05, + "loss": 0.4473, + "step": 49681 + }, + { + "epoch": 1.91, + "learning_rate": 1.8176455501410618e-05, + "loss": 0.2555, + "step": 49682 + }, + { + "epoch": 1.91, + "learning_rate": 1.817581431136189e-05, + "loss": 0.2979, + "step": 49683 + }, + { + "epoch": 1.91, + "learning_rate": 1.8175173121313158e-05, + "loss": 0.3927, + "step": 49684 + }, + { + "epoch": 1.91, + "learning_rate": 1.8174531931264426e-05, + "loss": 0.3628, + "step": 49685 + }, + { + "epoch": 1.91, + "learning_rate": 1.8173890741215698e-05, + "loss": 0.3752, + "step": 49686 + }, + { + "epoch": 1.91, + "learning_rate": 1.817324955116697e-05, + "loss": 0.3234, + "step": 49687 + }, + { + "epoch": 1.91, + "learning_rate": 1.8172608361118238e-05, + "loss": 0.2926, + "step": 49688 + }, + { + "epoch": 1.91, + "learning_rate": 1.8171967171069506e-05, + "loss": 0.3856, + "step": 49689 + }, + { + "epoch": 1.91, + "learning_rate": 1.8171325981020774e-05, + "loss": 0.3318, + "step": 49690 + }, + { + "epoch": 1.91, + "learning_rate": 1.8170684790972046e-05, + "loss": 0.4322, + "step": 49691 + }, + { + "epoch": 1.91, + "learning_rate": 1.8170043600923314e-05, + "loss": 0.3868, + "step": 49692 + }, + { + "epoch": 1.91, + "learning_rate": 1.8169402410874583e-05, + "loss": 0.3379, + "step": 49693 + }, + { + "epoch": 1.91, + "learning_rate": 1.8168761220825855e-05, + "loss": 0.3086, + "step": 49694 + }, + { + "epoch": 1.91, + "learning_rate": 1.8168120030777123e-05, + "loss": 0.3615, + "step": 49695 + }, + { + "epoch": 1.91, + "learning_rate": 1.8167478840728395e-05, + "loss": 0.4319, + "step": 49696 + }, + { + "epoch": 1.91, + "learning_rate": 1.8166837650679663e-05, + "loss": 0.2685, + "step": 49697 + }, + { + "epoch": 1.91, + "learning_rate": 1.816619646063093e-05, + "loss": 0.3481, + "step": 49698 + }, + { + "epoch": 1.91, + "learning_rate": 1.8165555270582203e-05, + "loss": 0.2957, + "step": 49699 + }, + { + "epoch": 1.91, + "learning_rate": 1.816491408053347e-05, + "loss": 0.2906, + "step": 49700 + }, + { + "epoch": 1.91, + "learning_rate": 1.816427289048474e-05, + "loss": 0.3637, + "step": 49701 + }, + { + "epoch": 1.91, + "learning_rate": 1.8163631700436008e-05, + "loss": 0.3862, + "step": 49702 + }, + { + "epoch": 1.91, + "learning_rate": 1.816299051038728e-05, + "loss": 0.2398, + "step": 49703 + }, + { + "epoch": 1.91, + "learning_rate": 1.8162349320338548e-05, + "loss": 0.3915, + "step": 49704 + }, + { + "epoch": 1.91, + "learning_rate": 1.816170813028982e-05, + "loss": 0.2364, + "step": 49705 + }, + { + "epoch": 1.91, + "learning_rate": 1.8161066940241088e-05, + "loss": 0.2605, + "step": 49706 + }, + { + "epoch": 1.91, + "learning_rate": 1.816042575019236e-05, + "loss": 0.3898, + "step": 49707 + }, + { + "epoch": 1.91, + "learning_rate": 1.8159784560143628e-05, + "loss": 0.4496, + "step": 49708 + }, + { + "epoch": 1.91, + "learning_rate": 1.8159143370094896e-05, + "loss": 0.2625, + "step": 49709 + }, + { + "epoch": 1.91, + "learning_rate": 1.8158502180046165e-05, + "loss": 0.2569, + "step": 49710 + }, + { + "epoch": 1.91, + "learning_rate": 1.8157860989997436e-05, + "loss": 0.3776, + "step": 49711 + }, + { + "epoch": 1.91, + "learning_rate": 1.8157219799948705e-05, + "loss": 0.4111, + "step": 49712 + }, + { + "epoch": 1.91, + "learning_rate": 1.8156578609899973e-05, + "loss": 0.3055, + "step": 49713 + }, + { + "epoch": 1.91, + "learning_rate": 1.8155937419851245e-05, + "loss": 0.3858, + "step": 49714 + }, + { + "epoch": 1.91, + "learning_rate": 1.8155296229802517e-05, + "loss": 0.3482, + "step": 49715 + }, + { + "epoch": 1.91, + "learning_rate": 1.8154655039753785e-05, + "loss": 0.3461, + "step": 49716 + }, + { + "epoch": 1.91, + "learning_rate": 1.8154013849705053e-05, + "loss": 0.3617, + "step": 49717 + }, + { + "epoch": 1.91, + "learning_rate": 1.815337265965632e-05, + "loss": 0.2605, + "step": 49718 + }, + { + "epoch": 1.91, + "learning_rate": 1.8152731469607593e-05, + "loss": 0.304, + "step": 49719 + }, + { + "epoch": 1.91, + "learning_rate": 1.815209027955886e-05, + "loss": 0.3748, + "step": 49720 + }, + { + "epoch": 1.91, + "learning_rate": 1.815144908951013e-05, + "loss": 0.2162, + "step": 49721 + }, + { + "epoch": 1.91, + "learning_rate": 1.81508078994614e-05, + "loss": 0.314, + "step": 49722 + }, + { + "epoch": 1.91, + "learning_rate": 1.8150166709412673e-05, + "loss": 0.2979, + "step": 49723 + }, + { + "epoch": 1.91, + "learning_rate": 1.814952551936394e-05, + "loss": 0.3169, + "step": 49724 + }, + { + "epoch": 1.91, + "learning_rate": 1.814888432931521e-05, + "loss": 0.2221, + "step": 49725 + }, + { + "epoch": 1.91, + "learning_rate": 1.8148243139266478e-05, + "loss": 0.321, + "step": 49726 + }, + { + "epoch": 1.91, + "learning_rate": 1.814760194921775e-05, + "loss": 0.3849, + "step": 49727 + }, + { + "epoch": 1.91, + "learning_rate": 1.814696075916902e-05, + "loss": 0.2123, + "step": 49728 + }, + { + "epoch": 1.91, + "learning_rate": 1.8146319569120287e-05, + "loss": 0.2452, + "step": 49729 + }, + { + "epoch": 1.91, + "learning_rate": 1.814567837907156e-05, + "loss": 0.2468, + "step": 49730 + }, + { + "epoch": 1.91, + "learning_rate": 1.8145037189022827e-05, + "loss": 0.2785, + "step": 49731 + }, + { + "epoch": 1.91, + "learning_rate": 1.81443959989741e-05, + "loss": 0.2403, + "step": 49732 + }, + { + "epoch": 1.91, + "learning_rate": 1.8143754808925367e-05, + "loss": 0.2741, + "step": 49733 + }, + { + "epoch": 1.91, + "learning_rate": 1.8143113618876635e-05, + "loss": 0.3394, + "step": 49734 + }, + { + "epoch": 1.91, + "learning_rate": 1.8142472428827907e-05, + "loss": 0.5706, + "step": 49735 + }, + { + "epoch": 1.91, + "learning_rate": 1.8141831238779175e-05, + "loss": 0.2291, + "step": 49736 + }, + { + "epoch": 1.91, + "learning_rate": 1.8141190048730443e-05, + "loss": 0.2668, + "step": 49737 + }, + { + "epoch": 1.91, + "learning_rate": 1.8140548858681715e-05, + "loss": 0.5362, + "step": 49738 + }, + { + "epoch": 1.91, + "learning_rate": 1.8139907668632984e-05, + "loss": 0.2784, + "step": 49739 + }, + { + "epoch": 1.91, + "learning_rate": 1.8139266478584252e-05, + "loss": 0.3332, + "step": 49740 + }, + { + "epoch": 1.91, + "learning_rate": 1.8138625288535524e-05, + "loss": 0.3612, + "step": 49741 + }, + { + "epoch": 1.91, + "learning_rate": 1.8137984098486792e-05, + "loss": 0.381, + "step": 49742 + }, + { + "epoch": 1.91, + "learning_rate": 1.8137342908438064e-05, + "loss": 0.2994, + "step": 49743 + }, + { + "epoch": 1.91, + "learning_rate": 1.8136701718389332e-05, + "loss": 0.2793, + "step": 49744 + }, + { + "epoch": 1.91, + "learning_rate": 1.81360605283406e-05, + "loss": 0.3062, + "step": 49745 + }, + { + "epoch": 1.91, + "learning_rate": 1.813541933829187e-05, + "loss": 0.4131, + "step": 49746 + }, + { + "epoch": 1.91, + "learning_rate": 1.813477814824314e-05, + "loss": 0.4681, + "step": 49747 + }, + { + "epoch": 1.91, + "learning_rate": 1.813413695819441e-05, + "loss": 0.2207, + "step": 49748 + }, + { + "epoch": 1.91, + "learning_rate": 1.8133495768145677e-05, + "loss": 0.5872, + "step": 49749 + }, + { + "epoch": 1.91, + "learning_rate": 1.813285457809695e-05, + "loss": 0.2582, + "step": 49750 + }, + { + "epoch": 1.91, + "learning_rate": 1.813221338804822e-05, + "loss": 0.3717, + "step": 49751 + }, + { + "epoch": 1.91, + "learning_rate": 1.813157219799949e-05, + "loss": 0.3038, + "step": 49752 + }, + { + "epoch": 1.91, + "learning_rate": 1.8130931007950757e-05, + "loss": 0.3764, + "step": 49753 + }, + { + "epoch": 1.91, + "learning_rate": 1.8130289817902025e-05, + "loss": 0.3691, + "step": 49754 + }, + { + "epoch": 1.91, + "learning_rate": 1.8129648627853297e-05, + "loss": 0.3718, + "step": 49755 + }, + { + "epoch": 1.91, + "learning_rate": 1.8129007437804565e-05, + "loss": 0.3115, + "step": 49756 + }, + { + "epoch": 1.91, + "learning_rate": 1.8128366247755834e-05, + "loss": 0.2261, + "step": 49757 + }, + { + "epoch": 1.91, + "learning_rate": 1.8127725057707105e-05, + "loss": 0.2735, + "step": 49758 + }, + { + "epoch": 1.91, + "learning_rate": 1.8127083867658377e-05, + "loss": 0.3849, + "step": 49759 + }, + { + "epoch": 1.91, + "learning_rate": 1.8126442677609646e-05, + "loss": 0.2322, + "step": 49760 + }, + { + "epoch": 1.91, + "learning_rate": 1.8125801487560914e-05, + "loss": 0.2716, + "step": 49761 + }, + { + "epoch": 1.91, + "learning_rate": 1.8125160297512182e-05, + "loss": 0.4347, + "step": 49762 + }, + { + "epoch": 1.91, + "learning_rate": 1.8124519107463454e-05, + "loss": 0.2806, + "step": 49763 + }, + { + "epoch": 1.91, + "learning_rate": 1.8123877917414722e-05, + "loss": 0.4284, + "step": 49764 + }, + { + "epoch": 1.91, + "learning_rate": 1.812323672736599e-05, + "loss": 0.2382, + "step": 49765 + }, + { + "epoch": 1.91, + "learning_rate": 1.8122595537317262e-05, + "loss": 0.2408, + "step": 49766 + }, + { + "epoch": 1.91, + "learning_rate": 1.812195434726853e-05, + "loss": 0.3243, + "step": 49767 + }, + { + "epoch": 1.91, + "learning_rate": 1.8121313157219802e-05, + "loss": 0.2312, + "step": 49768 + }, + { + "epoch": 1.91, + "learning_rate": 1.812067196717107e-05, + "loss": 0.192, + "step": 49769 + }, + { + "epoch": 1.91, + "learning_rate": 1.812003077712234e-05, + "loss": 0.211, + "step": 49770 + }, + { + "epoch": 1.91, + "learning_rate": 1.811938958707361e-05, + "loss": 0.3218, + "step": 49771 + }, + { + "epoch": 1.91, + "learning_rate": 1.811874839702488e-05, + "loss": 0.4016, + "step": 49772 + }, + { + "epoch": 1.91, + "learning_rate": 1.8118107206976147e-05, + "loss": 0.4048, + "step": 49773 + }, + { + "epoch": 1.91, + "learning_rate": 1.811746601692742e-05, + "loss": 0.2893, + "step": 49774 + }, + { + "epoch": 1.91, + "learning_rate": 1.8116824826878687e-05, + "loss": 0.324, + "step": 49775 + }, + { + "epoch": 1.91, + "learning_rate": 1.8116183636829956e-05, + "loss": 0.2651, + "step": 49776 + }, + { + "epoch": 1.91, + "learning_rate": 1.8115542446781227e-05, + "loss": 0.3477, + "step": 49777 + }, + { + "epoch": 1.91, + "learning_rate": 1.8114901256732496e-05, + "loss": 0.3257, + "step": 49778 + }, + { + "epoch": 1.91, + "learning_rate": 1.8114260066683767e-05, + "loss": 0.1871, + "step": 49779 + }, + { + "epoch": 1.91, + "learning_rate": 1.8113618876635036e-05, + "loss": 0.2931, + "step": 49780 + }, + { + "epoch": 1.91, + "learning_rate": 1.8112977686586304e-05, + "loss": 0.3113, + "step": 49781 + }, + { + "epoch": 1.91, + "learning_rate": 1.8112336496537572e-05, + "loss": 0.3261, + "step": 49782 + }, + { + "epoch": 1.91, + "learning_rate": 1.8111695306488844e-05, + "loss": 0.3452, + "step": 49783 + }, + { + "epoch": 1.91, + "learning_rate": 1.8111054116440113e-05, + "loss": 0.3365, + "step": 49784 + }, + { + "epoch": 1.91, + "learning_rate": 1.8110412926391384e-05, + "loss": 0.437, + "step": 49785 + }, + { + "epoch": 1.91, + "learning_rate": 1.8109771736342653e-05, + "loss": 0.2602, + "step": 49786 + }, + { + "epoch": 1.91, + "learning_rate": 1.8109130546293924e-05, + "loss": 0.3163, + "step": 49787 + }, + { + "epoch": 1.91, + "learning_rate": 1.8108489356245193e-05, + "loss": 0.4432, + "step": 49788 + }, + { + "epoch": 1.91, + "learning_rate": 1.810784816619646e-05, + "loss": 0.4948, + "step": 49789 + }, + { + "epoch": 1.91, + "learning_rate": 1.810720697614773e-05, + "loss": 0.4534, + "step": 49790 + }, + { + "epoch": 1.91, + "learning_rate": 1.8106565786099e-05, + "loss": 0.3624, + "step": 49791 + }, + { + "epoch": 1.91, + "learning_rate": 1.810592459605027e-05, + "loss": 0.2947, + "step": 49792 + }, + { + "epoch": 1.91, + "learning_rate": 1.8105283406001538e-05, + "loss": 0.2469, + "step": 49793 + }, + { + "epoch": 1.91, + "learning_rate": 1.810464221595281e-05, + "loss": 0.2863, + "step": 49794 + }, + { + "epoch": 1.91, + "learning_rate": 1.810400102590408e-05, + "loss": 0.2939, + "step": 49795 + }, + { + "epoch": 1.91, + "learning_rate": 1.810335983585535e-05, + "loss": 0.4026, + "step": 49796 + }, + { + "epoch": 1.91, + "learning_rate": 1.8102718645806618e-05, + "loss": 0.3374, + "step": 49797 + }, + { + "epoch": 1.91, + "learning_rate": 1.8102077455757886e-05, + "loss": 0.3865, + "step": 49798 + }, + { + "epoch": 1.91, + "learning_rate": 1.8101436265709158e-05, + "loss": 0.2189, + "step": 49799 + }, + { + "epoch": 1.91, + "learning_rate": 1.8100795075660426e-05, + "loss": 0.2482, + "step": 49800 + }, + { + "epoch": 1.91, + "learning_rate": 1.8100153885611694e-05, + "loss": 0.3623, + "step": 49801 + }, + { + "epoch": 1.91, + "learning_rate": 1.8099512695562966e-05, + "loss": 0.2426, + "step": 49802 + }, + { + "epoch": 1.91, + "learning_rate": 1.8098871505514238e-05, + "loss": 0.3507, + "step": 49803 + }, + { + "epoch": 1.91, + "learning_rate": 1.8098230315465506e-05, + "loss": 0.3401, + "step": 49804 + }, + { + "epoch": 1.91, + "learning_rate": 1.8097589125416775e-05, + "loss": 0.3663, + "step": 49805 + }, + { + "epoch": 1.91, + "learning_rate": 1.8096947935368043e-05, + "loss": 0.261, + "step": 49806 + }, + { + "epoch": 1.91, + "learning_rate": 1.8096306745319315e-05, + "loss": 0.2543, + "step": 49807 + }, + { + "epoch": 1.91, + "learning_rate": 1.8095665555270583e-05, + "loss": 0.225, + "step": 49808 + }, + { + "epoch": 1.91, + "learning_rate": 1.809502436522185e-05, + "loss": 0.2886, + "step": 49809 + }, + { + "epoch": 1.92, + "learning_rate": 1.8094383175173123e-05, + "loss": 0.272, + "step": 49810 + }, + { + "epoch": 1.92, + "learning_rate": 1.809374198512439e-05, + "loss": 0.4071, + "step": 49811 + }, + { + "epoch": 1.92, + "learning_rate": 1.8093100795075663e-05, + "loss": 0.3676, + "step": 49812 + }, + { + "epoch": 1.92, + "learning_rate": 1.809245960502693e-05, + "loss": 0.1766, + "step": 49813 + }, + { + "epoch": 1.92, + "learning_rate": 1.80918184149782e-05, + "loss": 0.3172, + "step": 49814 + }, + { + "epoch": 1.92, + "learning_rate": 1.809117722492947e-05, + "loss": 0.3506, + "step": 49815 + }, + { + "epoch": 1.92, + "learning_rate": 1.809053603488074e-05, + "loss": 0.331, + "step": 49816 + }, + { + "epoch": 1.92, + "learning_rate": 1.8089894844832008e-05, + "loss": 0.3623, + "step": 49817 + }, + { + "epoch": 1.92, + "learning_rate": 1.808925365478328e-05, + "loss": 0.2812, + "step": 49818 + }, + { + "epoch": 1.92, + "learning_rate": 1.8088612464734548e-05, + "loss": 0.3357, + "step": 49819 + }, + { + "epoch": 1.92, + "learning_rate": 1.8087971274685816e-05, + "loss": 0.1816, + "step": 49820 + }, + { + "epoch": 1.92, + "learning_rate": 1.8087330084637088e-05, + "loss": 0.4893, + "step": 49821 + }, + { + "epoch": 1.92, + "learning_rate": 1.8086688894588356e-05, + "loss": 0.365, + "step": 49822 + }, + { + "epoch": 1.92, + "learning_rate": 1.8086047704539628e-05, + "loss": 0.2978, + "step": 49823 + }, + { + "epoch": 1.92, + "learning_rate": 1.8085406514490896e-05, + "loss": 0.3693, + "step": 49824 + }, + { + "epoch": 1.92, + "learning_rate": 1.8084765324442165e-05, + "loss": 0.2918, + "step": 49825 + }, + { + "epoch": 1.92, + "learning_rate": 1.8084124134393433e-05, + "loss": 0.2145, + "step": 49826 + }, + { + "epoch": 1.92, + "learning_rate": 1.8083482944344705e-05, + "loss": 0.3069, + "step": 49827 + }, + { + "epoch": 1.92, + "learning_rate": 1.8082841754295973e-05, + "loss": 0.3856, + "step": 49828 + }, + { + "epoch": 1.92, + "learning_rate": 1.808220056424724e-05, + "loss": 0.2475, + "step": 49829 + }, + { + "epoch": 1.92, + "learning_rate": 1.8081559374198513e-05, + "loss": 0.2232, + "step": 49830 + }, + { + "epoch": 1.92, + "learning_rate": 1.8080918184149785e-05, + "loss": 0.3266, + "step": 49831 + }, + { + "epoch": 1.92, + "learning_rate": 1.8080276994101053e-05, + "loss": 0.2446, + "step": 49832 + }, + { + "epoch": 1.92, + "learning_rate": 1.807963580405232e-05, + "loss": 0.4134, + "step": 49833 + }, + { + "epoch": 1.92, + "learning_rate": 1.807899461400359e-05, + "loss": 0.4475, + "step": 49834 + }, + { + "epoch": 1.92, + "learning_rate": 1.807835342395486e-05, + "loss": 0.2502, + "step": 49835 + }, + { + "epoch": 1.92, + "learning_rate": 1.807771223390613e-05, + "loss": 0.2754, + "step": 49836 + }, + { + "epoch": 1.92, + "learning_rate": 1.8077071043857398e-05, + "loss": 0.402, + "step": 49837 + }, + { + "epoch": 1.92, + "learning_rate": 1.807642985380867e-05, + "loss": 0.3065, + "step": 49838 + }, + { + "epoch": 1.92, + "learning_rate": 1.8075788663759942e-05, + "loss": 0.3398, + "step": 49839 + }, + { + "epoch": 1.92, + "learning_rate": 1.807514747371121e-05, + "loss": 0.3697, + "step": 49840 + }, + { + "epoch": 1.92, + "learning_rate": 1.807450628366248e-05, + "loss": 0.225, + "step": 49841 + }, + { + "epoch": 1.92, + "learning_rate": 1.8073865093613747e-05, + "loss": 0.4261, + "step": 49842 + }, + { + "epoch": 1.92, + "learning_rate": 1.807322390356502e-05, + "loss": 0.3612, + "step": 49843 + }, + { + "epoch": 1.92, + "learning_rate": 1.8072582713516287e-05, + "loss": 0.302, + "step": 49844 + }, + { + "epoch": 1.92, + "learning_rate": 1.8071941523467555e-05, + "loss": 0.2638, + "step": 49845 + }, + { + "epoch": 1.92, + "learning_rate": 1.8071300333418827e-05, + "loss": 0.254, + "step": 49846 + }, + { + "epoch": 1.92, + "learning_rate": 1.8070659143370095e-05, + "loss": 0.3835, + "step": 49847 + }, + { + "epoch": 1.92, + "learning_rate": 1.8070017953321367e-05, + "loss": 0.3354, + "step": 49848 + }, + { + "epoch": 1.92, + "learning_rate": 1.8069376763272635e-05, + "loss": 0.2024, + "step": 49849 + }, + { + "epoch": 1.92, + "learning_rate": 1.8068735573223904e-05, + "loss": 0.3302, + "step": 49850 + }, + { + "epoch": 1.92, + "learning_rate": 1.8068094383175175e-05, + "loss": 0.3919, + "step": 49851 + }, + { + "epoch": 1.92, + "learning_rate": 1.8067453193126444e-05, + "loss": 0.3355, + "step": 49852 + }, + { + "epoch": 1.92, + "learning_rate": 1.8066812003077712e-05, + "loss": 0.382, + "step": 49853 + }, + { + "epoch": 1.92, + "learning_rate": 1.8066170813028984e-05, + "loss": 0.2996, + "step": 49854 + }, + { + "epoch": 1.92, + "learning_rate": 1.8065529622980252e-05, + "loss": 0.2739, + "step": 49855 + }, + { + "epoch": 1.92, + "learning_rate": 1.806488843293152e-05, + "loss": 0.3697, + "step": 49856 + }, + { + "epoch": 1.92, + "learning_rate": 1.8064247242882792e-05, + "loss": 0.2166, + "step": 49857 + }, + { + "epoch": 1.92, + "learning_rate": 1.806360605283406e-05, + "loss": 0.2223, + "step": 49858 + }, + { + "epoch": 1.92, + "learning_rate": 1.8062964862785332e-05, + "loss": 0.3106, + "step": 49859 + }, + { + "epoch": 1.92, + "learning_rate": 1.80623236727366e-05, + "loss": 0.2719, + "step": 49860 + }, + { + "epoch": 1.92, + "learning_rate": 1.806168248268787e-05, + "loss": 0.3038, + "step": 49861 + }, + { + "epoch": 1.92, + "learning_rate": 1.8061041292639137e-05, + "loss": 0.324, + "step": 49862 + }, + { + "epoch": 1.92, + "learning_rate": 1.806040010259041e-05, + "loss": 0.1847, + "step": 49863 + }, + { + "epoch": 1.92, + "learning_rate": 1.8059758912541677e-05, + "loss": 0.3062, + "step": 49864 + }, + { + "epoch": 1.92, + "learning_rate": 1.8059117722492945e-05, + "loss": 0.4862, + "step": 49865 + }, + { + "epoch": 1.92, + "learning_rate": 1.8058476532444217e-05, + "loss": 0.2455, + "step": 49866 + }, + { + "epoch": 1.92, + "learning_rate": 1.805783534239549e-05, + "loss": 0.2695, + "step": 49867 + }, + { + "epoch": 1.92, + "learning_rate": 1.8057194152346757e-05, + "loss": 0.319, + "step": 49868 + }, + { + "epoch": 1.92, + "learning_rate": 1.8056552962298025e-05, + "loss": 0.3614, + "step": 49869 + }, + { + "epoch": 1.92, + "learning_rate": 1.8055911772249294e-05, + "loss": 0.3888, + "step": 49870 + }, + { + "epoch": 1.92, + "learning_rate": 1.8055270582200566e-05, + "loss": 0.2167, + "step": 49871 + }, + { + "epoch": 1.92, + "learning_rate": 1.8054629392151834e-05, + "loss": 0.3037, + "step": 49872 + }, + { + "epoch": 1.92, + "learning_rate": 1.8053988202103102e-05, + "loss": 0.3156, + "step": 49873 + }, + { + "epoch": 1.92, + "learning_rate": 1.8053347012054374e-05, + "loss": 0.4045, + "step": 49874 + }, + { + "epoch": 1.92, + "learning_rate": 1.8052705822005646e-05, + "loss": 0.5296, + "step": 49875 + }, + { + "epoch": 1.92, + "learning_rate": 1.8052064631956914e-05, + "loss": 0.4172, + "step": 49876 + }, + { + "epoch": 1.92, + "learning_rate": 1.8051423441908182e-05, + "loss": 0.4619, + "step": 49877 + }, + { + "epoch": 1.92, + "learning_rate": 1.805078225185945e-05, + "loss": 0.3031, + "step": 49878 + }, + { + "epoch": 1.92, + "learning_rate": 1.8050141061810722e-05, + "loss": 0.4864, + "step": 49879 + }, + { + "epoch": 1.92, + "learning_rate": 1.804949987176199e-05, + "loss": 0.2582, + "step": 49880 + }, + { + "epoch": 1.92, + "learning_rate": 1.804885868171326e-05, + "loss": 0.4016, + "step": 49881 + }, + { + "epoch": 1.92, + "learning_rate": 1.804821749166453e-05, + "loss": 0.3049, + "step": 49882 + }, + { + "epoch": 1.92, + "learning_rate": 1.8047576301615802e-05, + "loss": 0.2763, + "step": 49883 + }, + { + "epoch": 1.92, + "learning_rate": 1.804693511156707e-05, + "loss": 0.3154, + "step": 49884 + }, + { + "epoch": 1.92, + "learning_rate": 1.804629392151834e-05, + "loss": 0.2225, + "step": 49885 + }, + { + "epoch": 1.92, + "learning_rate": 1.8045652731469607e-05, + "loss": 0.4216, + "step": 49886 + }, + { + "epoch": 1.92, + "learning_rate": 1.804501154142088e-05, + "loss": 0.2534, + "step": 49887 + }, + { + "epoch": 1.92, + "learning_rate": 1.8044370351372147e-05, + "loss": 0.1842, + "step": 49888 + }, + { + "epoch": 1.92, + "learning_rate": 1.8043729161323416e-05, + "loss": 0.4017, + "step": 49889 + }, + { + "epoch": 1.92, + "learning_rate": 1.8043087971274687e-05, + "loss": 0.3474, + "step": 49890 + }, + { + "epoch": 1.92, + "learning_rate": 1.8042446781225956e-05, + "loss": 0.2217, + "step": 49891 + }, + { + "epoch": 1.92, + "learning_rate": 1.8041805591177228e-05, + "loss": 0.4357, + "step": 49892 + }, + { + "epoch": 1.92, + "learning_rate": 1.8041164401128496e-05, + "loss": 0.3616, + "step": 49893 + }, + { + "epoch": 1.92, + "learning_rate": 1.8040523211079764e-05, + "loss": 0.2074, + "step": 49894 + }, + { + "epoch": 1.92, + "learning_rate": 1.8039882021031036e-05, + "loss": 0.3429, + "step": 49895 + }, + { + "epoch": 1.92, + "learning_rate": 1.8039240830982304e-05, + "loss": 0.266, + "step": 49896 + }, + { + "epoch": 1.92, + "learning_rate": 1.8038599640933573e-05, + "loss": 0.3441, + "step": 49897 + }, + { + "epoch": 1.92, + "learning_rate": 1.8037958450884844e-05, + "loss": 0.328, + "step": 49898 + }, + { + "epoch": 1.92, + "learning_rate": 1.8037317260836113e-05, + "loss": 0.3152, + "step": 49899 + }, + { + "epoch": 1.92, + "learning_rate": 1.803667607078738e-05, + "loss": 0.2714, + "step": 49900 + }, + { + "epoch": 1.92, + "learning_rate": 1.8036034880738653e-05, + "loss": 0.4284, + "step": 49901 + }, + { + "epoch": 1.92, + "learning_rate": 1.803539369068992e-05, + "loss": 0.3423, + "step": 49902 + }, + { + "epoch": 1.92, + "learning_rate": 1.8034752500641193e-05, + "loss": 0.2879, + "step": 49903 + }, + { + "epoch": 1.92, + "learning_rate": 1.803411131059246e-05, + "loss": 0.3201, + "step": 49904 + }, + { + "epoch": 1.92, + "learning_rate": 1.803347012054373e-05, + "loss": 0.3275, + "step": 49905 + }, + { + "epoch": 1.92, + "learning_rate": 1.8032828930494998e-05, + "loss": 0.2583, + "step": 49906 + }, + { + "epoch": 1.92, + "learning_rate": 1.803218774044627e-05, + "loss": 0.3362, + "step": 49907 + }, + { + "epoch": 1.92, + "learning_rate": 1.8031546550397538e-05, + "loss": 0.2346, + "step": 49908 + }, + { + "epoch": 1.92, + "learning_rate": 1.8030905360348806e-05, + "loss": 0.2696, + "step": 49909 + }, + { + "epoch": 1.92, + "learning_rate": 1.8030264170300078e-05, + "loss": 0.4726, + "step": 49910 + }, + { + "epoch": 1.92, + "learning_rate": 1.802962298025135e-05, + "loss": 0.2896, + "step": 49911 + }, + { + "epoch": 1.92, + "learning_rate": 1.8028981790202618e-05, + "loss": 0.2125, + "step": 49912 + }, + { + "epoch": 1.92, + "learning_rate": 1.8028340600153886e-05, + "loss": 0.2722, + "step": 49913 + }, + { + "epoch": 1.92, + "learning_rate": 1.8027699410105154e-05, + "loss": 0.3954, + "step": 49914 + }, + { + "epoch": 1.92, + "learning_rate": 1.8027058220056426e-05, + "loss": 0.3689, + "step": 49915 + }, + { + "epoch": 1.92, + "learning_rate": 1.8026417030007695e-05, + "loss": 0.2055, + "step": 49916 + }, + { + "epoch": 1.92, + "learning_rate": 1.8025775839958963e-05, + "loss": 0.5281, + "step": 49917 + }, + { + "epoch": 1.92, + "learning_rate": 1.8025134649910235e-05, + "loss": 0.3589, + "step": 49918 + }, + { + "epoch": 1.92, + "learning_rate": 1.8024493459861506e-05, + "loss": 0.4302, + "step": 49919 + }, + { + "epoch": 1.92, + "learning_rate": 1.8023852269812775e-05, + "loss": 0.3869, + "step": 49920 + }, + { + "epoch": 1.92, + "learning_rate": 1.8023211079764043e-05, + "loss": 0.2156, + "step": 49921 + }, + { + "epoch": 1.92, + "learning_rate": 1.802256988971531e-05, + "loss": 0.4307, + "step": 49922 + }, + { + "epoch": 1.92, + "learning_rate": 1.8021928699666583e-05, + "loss": 0.3214, + "step": 49923 + }, + { + "epoch": 1.92, + "learning_rate": 1.802128750961785e-05, + "loss": 0.4037, + "step": 49924 + }, + { + "epoch": 1.92, + "learning_rate": 1.802064631956912e-05, + "loss": 0.4614, + "step": 49925 + }, + { + "epoch": 1.92, + "learning_rate": 1.802000512952039e-05, + "loss": 0.2921, + "step": 49926 + }, + { + "epoch": 1.92, + "learning_rate": 1.801936393947166e-05, + "loss": 0.3703, + "step": 49927 + }, + { + "epoch": 1.92, + "learning_rate": 1.801872274942293e-05, + "loss": 0.2677, + "step": 49928 + }, + { + "epoch": 1.92, + "learning_rate": 1.80180815593742e-05, + "loss": 0.206, + "step": 49929 + }, + { + "epoch": 1.92, + "learning_rate": 1.8017440369325468e-05, + "loss": 0.2999, + "step": 49930 + }, + { + "epoch": 1.92, + "learning_rate": 1.801679917927674e-05, + "loss": 0.167, + "step": 49931 + }, + { + "epoch": 1.92, + "learning_rate": 1.8016157989228008e-05, + "loss": 0.4343, + "step": 49932 + }, + { + "epoch": 1.92, + "learning_rate": 1.8015516799179276e-05, + "loss": 0.2406, + "step": 49933 + }, + { + "epoch": 1.92, + "learning_rate": 1.8014875609130548e-05, + "loss": 0.3677, + "step": 49934 + }, + { + "epoch": 1.92, + "learning_rate": 1.8014234419081816e-05, + "loss": 0.2723, + "step": 49935 + }, + { + "epoch": 1.92, + "learning_rate": 1.8013593229033085e-05, + "loss": 0.4018, + "step": 49936 + }, + { + "epoch": 1.92, + "learning_rate": 1.8012952038984357e-05, + "loss": 0.3674, + "step": 49937 + }, + { + "epoch": 1.92, + "learning_rate": 1.8012310848935625e-05, + "loss": 0.4188, + "step": 49938 + }, + { + "epoch": 1.92, + "learning_rate": 1.8011669658886897e-05, + "loss": 0.3639, + "step": 49939 + }, + { + "epoch": 1.92, + "learning_rate": 1.8011028468838165e-05, + "loss": 0.3662, + "step": 49940 + }, + { + "epoch": 1.92, + "learning_rate": 1.8010387278789433e-05, + "loss": 0.2609, + "step": 49941 + }, + { + "epoch": 1.92, + "learning_rate": 1.8009746088740705e-05, + "loss": 0.3011, + "step": 49942 + }, + { + "epoch": 1.92, + "learning_rate": 1.8009104898691973e-05, + "loss": 0.2451, + "step": 49943 + }, + { + "epoch": 1.92, + "learning_rate": 1.800846370864324e-05, + "loss": 0.383, + "step": 49944 + }, + { + "epoch": 1.92, + "learning_rate": 1.800782251859451e-05, + "loss": 0.2794, + "step": 49945 + }, + { + "epoch": 1.92, + "learning_rate": 1.800718132854578e-05, + "loss": 0.3583, + "step": 49946 + }, + { + "epoch": 1.92, + "learning_rate": 1.8006540138497053e-05, + "loss": 0.4446, + "step": 49947 + }, + { + "epoch": 1.92, + "learning_rate": 1.800589894844832e-05, + "loss": 0.2157, + "step": 49948 + }, + { + "epoch": 1.92, + "learning_rate": 1.800525775839959e-05, + "loss": 0.3332, + "step": 49949 + }, + { + "epoch": 1.92, + "learning_rate": 1.800461656835086e-05, + "loss": 0.3641, + "step": 49950 + }, + { + "epoch": 1.92, + "learning_rate": 1.800397537830213e-05, + "loss": 0.4101, + "step": 49951 + }, + { + "epoch": 1.92, + "learning_rate": 1.80033341882534e-05, + "loss": 0.388, + "step": 49952 + }, + { + "epoch": 1.92, + "learning_rate": 1.8002692998204667e-05, + "loss": 0.29, + "step": 49953 + }, + { + "epoch": 1.92, + "learning_rate": 1.800205180815594e-05, + "loss": 0.3384, + "step": 49954 + }, + { + "epoch": 1.92, + "learning_rate": 1.800141061810721e-05, + "loss": 0.3357, + "step": 49955 + }, + { + "epoch": 1.92, + "learning_rate": 1.800076942805848e-05, + "loss": 0.3818, + "step": 49956 + }, + { + "epoch": 1.92, + "learning_rate": 1.8000128238009747e-05, + "loss": 0.2577, + "step": 49957 + }, + { + "epoch": 1.92, + "learning_rate": 1.7999487047961015e-05, + "loss": 0.2531, + "step": 49958 + }, + { + "epoch": 1.92, + "learning_rate": 1.7998845857912287e-05, + "loss": 0.3251, + "step": 49959 + }, + { + "epoch": 1.92, + "learning_rate": 1.7998204667863555e-05, + "loss": 0.2763, + "step": 49960 + }, + { + "epoch": 1.92, + "learning_rate": 1.7997563477814823e-05, + "loss": 0.2438, + "step": 49961 + }, + { + "epoch": 1.92, + "learning_rate": 1.7996922287766095e-05, + "loss": 0.302, + "step": 49962 + }, + { + "epoch": 1.92, + "learning_rate": 1.7996281097717364e-05, + "loss": 0.3696, + "step": 49963 + }, + { + "epoch": 1.92, + "learning_rate": 1.7995639907668635e-05, + "loss": 0.4467, + "step": 49964 + }, + { + "epoch": 1.92, + "learning_rate": 1.7994998717619904e-05, + "loss": 0.3958, + "step": 49965 + }, + { + "epoch": 1.92, + "learning_rate": 1.7994357527571172e-05, + "loss": 0.3456, + "step": 49966 + }, + { + "epoch": 1.92, + "learning_rate": 1.7993716337522444e-05, + "loss": 0.3952, + "step": 49967 + }, + { + "epoch": 1.92, + "learning_rate": 1.7993075147473712e-05, + "loss": 0.4235, + "step": 49968 + }, + { + "epoch": 1.92, + "learning_rate": 1.799243395742498e-05, + "loss": 0.3293, + "step": 49969 + }, + { + "epoch": 1.92, + "learning_rate": 1.7991792767376252e-05, + "loss": 0.332, + "step": 49970 + }, + { + "epoch": 1.92, + "learning_rate": 1.799115157732752e-05, + "loss": 0.361, + "step": 49971 + }, + { + "epoch": 1.92, + "learning_rate": 1.7990510387278792e-05, + "loss": 0.3148, + "step": 49972 + }, + { + "epoch": 1.92, + "learning_rate": 1.798986919723006e-05, + "loss": 0.3847, + "step": 49973 + }, + { + "epoch": 1.92, + "learning_rate": 1.798922800718133e-05, + "loss": 0.3045, + "step": 49974 + }, + { + "epoch": 1.92, + "learning_rate": 1.79885868171326e-05, + "loss": 0.2911, + "step": 49975 + }, + { + "epoch": 1.92, + "learning_rate": 1.798794562708387e-05, + "loss": 0.3002, + "step": 49976 + }, + { + "epoch": 1.92, + "learning_rate": 1.7987304437035137e-05, + "loss": 0.308, + "step": 49977 + }, + { + "epoch": 1.92, + "learning_rate": 1.798666324698641e-05, + "loss": 0.3261, + "step": 49978 + }, + { + "epoch": 1.92, + "learning_rate": 1.7986022056937677e-05, + "loss": 0.3623, + "step": 49979 + }, + { + "epoch": 1.92, + "learning_rate": 1.7985380866888945e-05, + "loss": 0.5678, + "step": 49980 + }, + { + "epoch": 1.92, + "learning_rate": 1.7984739676840217e-05, + "loss": 0.2955, + "step": 49981 + }, + { + "epoch": 1.92, + "learning_rate": 1.7984098486791485e-05, + "loss": 0.4026, + "step": 49982 + }, + { + "epoch": 1.92, + "learning_rate": 1.7983457296742757e-05, + "loss": 0.3203, + "step": 49983 + }, + { + "epoch": 1.92, + "learning_rate": 1.7982816106694026e-05, + "loss": 0.3563, + "step": 49984 + }, + { + "epoch": 1.92, + "learning_rate": 1.7982174916645294e-05, + "loss": 0.3279, + "step": 49985 + }, + { + "epoch": 1.92, + "learning_rate": 1.7981533726596562e-05, + "loss": 0.3282, + "step": 49986 + }, + { + "epoch": 1.92, + "learning_rate": 1.7980892536547834e-05, + "loss": 0.2422, + "step": 49987 + }, + { + "epoch": 1.92, + "learning_rate": 1.7980251346499102e-05, + "loss": 0.2008, + "step": 49988 + }, + { + "epoch": 1.92, + "learning_rate": 1.797961015645037e-05, + "loss": 0.2868, + "step": 49989 + }, + { + "epoch": 1.92, + "learning_rate": 1.7978968966401642e-05, + "loss": 0.3175, + "step": 49990 + }, + { + "epoch": 1.92, + "learning_rate": 1.7978327776352914e-05, + "loss": 0.3589, + "step": 49991 + }, + { + "epoch": 1.92, + "learning_rate": 1.7977686586304182e-05, + "loss": 0.3351, + "step": 49992 + }, + { + "epoch": 1.92, + "learning_rate": 1.797704539625545e-05, + "loss": 0.2737, + "step": 49993 + }, + { + "epoch": 1.92, + "learning_rate": 1.797640420620672e-05, + "loss": 0.3266, + "step": 49994 + }, + { + "epoch": 1.92, + "learning_rate": 1.797576301615799e-05, + "loss": 0.1833, + "step": 49995 + }, + { + "epoch": 1.92, + "learning_rate": 1.797512182610926e-05, + "loss": 0.2241, + "step": 49996 + }, + { + "epoch": 1.92, + "learning_rate": 1.7974480636060527e-05, + "loss": 0.2516, + "step": 49997 + }, + { + "epoch": 1.92, + "learning_rate": 1.79738394460118e-05, + "loss": 0.3, + "step": 49998 + }, + { + "epoch": 1.92, + "learning_rate": 1.797319825596307e-05, + "loss": 0.1641, + "step": 49999 + }, + { + "epoch": 1.92, + "learning_rate": 1.797255706591434e-05, + "loss": 0.2068, + "step": 50000 + }, + { + "epoch": 1.92, + "learning_rate": 1.7971915875865607e-05, + "loss": 0.4133, + "step": 50001 + }, + { + "epoch": 1.92, + "learning_rate": 1.7971274685816876e-05, + "loss": 0.2128, + "step": 50002 + }, + { + "epoch": 1.92, + "learning_rate": 1.7970633495768147e-05, + "loss": 0.4143, + "step": 50003 + }, + { + "epoch": 1.92, + "learning_rate": 1.7969992305719416e-05, + "loss": 0.2799, + "step": 50004 + }, + { + "epoch": 1.92, + "learning_rate": 1.7969351115670684e-05, + "loss": 0.3858, + "step": 50005 + }, + { + "epoch": 1.92, + "learning_rate": 1.7968709925621956e-05, + "loss": 0.3816, + "step": 50006 + }, + { + "epoch": 1.92, + "learning_rate": 1.7968068735573224e-05, + "loss": 0.2788, + "step": 50007 + }, + { + "epoch": 1.92, + "learning_rate": 1.7967427545524496e-05, + "loss": 0.225, + "step": 50008 + }, + { + "epoch": 1.92, + "learning_rate": 1.7966786355475764e-05, + "loss": 0.375, + "step": 50009 + }, + { + "epoch": 1.92, + "learning_rate": 1.7966145165427033e-05, + "loss": 0.3455, + "step": 50010 + }, + { + "epoch": 1.92, + "learning_rate": 1.7965503975378304e-05, + "loss": 0.3164, + "step": 50011 + }, + { + "epoch": 1.92, + "learning_rate": 1.7964862785329573e-05, + "loss": 0.2708, + "step": 50012 + }, + { + "epoch": 1.92, + "learning_rate": 1.796422159528084e-05, + "loss": 0.4609, + "step": 50013 + }, + { + "epoch": 1.92, + "learning_rate": 1.7963580405232113e-05, + "loss": 0.2041, + "step": 50014 + }, + { + "epoch": 1.92, + "learning_rate": 1.796293921518338e-05, + "loss": 0.3728, + "step": 50015 + }, + { + "epoch": 1.92, + "learning_rate": 1.796229802513465e-05, + "loss": 0.4465, + "step": 50016 + }, + { + "epoch": 1.92, + "learning_rate": 1.796165683508592e-05, + "loss": 0.4012, + "step": 50017 + }, + { + "epoch": 1.92, + "learning_rate": 1.796101564503719e-05, + "loss": 0.2785, + "step": 50018 + }, + { + "epoch": 1.92, + "learning_rate": 1.796037445498846e-05, + "loss": 0.3294, + "step": 50019 + }, + { + "epoch": 1.92, + "learning_rate": 1.795973326493973e-05, + "loss": 0.2037, + "step": 50020 + }, + { + "epoch": 1.92, + "learning_rate": 1.7959092074890998e-05, + "loss": 0.3728, + "step": 50021 + }, + { + "epoch": 1.92, + "learning_rate": 1.795845088484227e-05, + "loss": 0.2172, + "step": 50022 + }, + { + "epoch": 1.92, + "learning_rate": 1.7957809694793538e-05, + "loss": 0.2376, + "step": 50023 + }, + { + "epoch": 1.92, + "learning_rate": 1.7957168504744806e-05, + "loss": 0.3594, + "step": 50024 + }, + { + "epoch": 1.92, + "learning_rate": 1.7956527314696074e-05, + "loss": 0.397, + "step": 50025 + }, + { + "epoch": 1.92, + "learning_rate": 1.7955886124647346e-05, + "loss": 0.3186, + "step": 50026 + }, + { + "epoch": 1.92, + "learning_rate": 1.7955244934598618e-05, + "loss": 0.3274, + "step": 50027 + }, + { + "epoch": 1.92, + "learning_rate": 1.7954603744549886e-05, + "loss": 0.2379, + "step": 50028 + }, + { + "epoch": 1.92, + "learning_rate": 1.7953962554501155e-05, + "loss": 0.3418, + "step": 50029 + }, + { + "epoch": 1.92, + "learning_rate": 1.7953321364452423e-05, + "loss": 0.2807, + "step": 50030 + }, + { + "epoch": 1.92, + "learning_rate": 1.7952680174403695e-05, + "loss": 0.2344, + "step": 50031 + }, + { + "epoch": 1.92, + "learning_rate": 1.7952038984354963e-05, + "loss": 0.2442, + "step": 50032 + }, + { + "epoch": 1.92, + "learning_rate": 1.795139779430623e-05, + "loss": 0.3801, + "step": 50033 + }, + { + "epoch": 1.92, + "learning_rate": 1.7950756604257503e-05, + "loss": 0.3178, + "step": 50034 + }, + { + "epoch": 1.92, + "learning_rate": 1.7950115414208775e-05, + "loss": 0.3436, + "step": 50035 + }, + { + "epoch": 1.92, + "learning_rate": 1.7949474224160043e-05, + "loss": 0.3203, + "step": 50036 + }, + { + "epoch": 1.92, + "learning_rate": 1.794883303411131e-05, + "loss": 0.3944, + "step": 50037 + }, + { + "epoch": 1.92, + "learning_rate": 1.794819184406258e-05, + "loss": 0.326, + "step": 50038 + }, + { + "epoch": 1.92, + "learning_rate": 1.794755065401385e-05, + "loss": 0.2836, + "step": 50039 + }, + { + "epoch": 1.92, + "learning_rate": 1.794690946396512e-05, + "loss": 0.2826, + "step": 50040 + }, + { + "epoch": 1.92, + "learning_rate": 1.7946268273916388e-05, + "loss": 0.2819, + "step": 50041 + }, + { + "epoch": 1.92, + "learning_rate": 1.794562708386766e-05, + "loss": 0.428, + "step": 50042 + }, + { + "epoch": 1.92, + "learning_rate": 1.7944985893818928e-05, + "loss": 0.2794, + "step": 50043 + }, + { + "epoch": 1.92, + "learning_rate": 1.79443447037702e-05, + "loss": 0.2188, + "step": 50044 + }, + { + "epoch": 1.92, + "learning_rate": 1.7943703513721468e-05, + "loss": 0.2023, + "step": 50045 + }, + { + "epoch": 1.92, + "learning_rate": 1.7943062323672736e-05, + "loss": 0.2963, + "step": 50046 + }, + { + "epoch": 1.92, + "learning_rate": 1.7942421133624008e-05, + "loss": 0.3905, + "step": 50047 + }, + { + "epoch": 1.92, + "learning_rate": 1.7941779943575276e-05, + "loss": 0.3763, + "step": 50048 + }, + { + "epoch": 1.92, + "learning_rate": 1.7941138753526545e-05, + "loss": 0.1797, + "step": 50049 + }, + { + "epoch": 1.92, + "learning_rate": 1.7940497563477817e-05, + "loss": 0.3462, + "step": 50050 + }, + { + "epoch": 1.92, + "learning_rate": 1.7939856373429085e-05, + "loss": 0.3249, + "step": 50051 + }, + { + "epoch": 1.92, + "learning_rate": 1.7939215183380353e-05, + "loss": 0.3018, + "step": 50052 + }, + { + "epoch": 1.92, + "learning_rate": 1.7938573993331625e-05, + "loss": 0.2054, + "step": 50053 + }, + { + "epoch": 1.92, + "learning_rate": 1.7937932803282893e-05, + "loss": 0.3378, + "step": 50054 + }, + { + "epoch": 1.92, + "learning_rate": 1.7937291613234165e-05, + "loss": 0.2452, + "step": 50055 + }, + { + "epoch": 1.92, + "learning_rate": 1.7936650423185433e-05, + "loss": 0.3224, + "step": 50056 + }, + { + "epoch": 1.92, + "learning_rate": 1.79360092331367e-05, + "loss": 0.3014, + "step": 50057 + }, + { + "epoch": 1.92, + "learning_rate": 1.7935368043087973e-05, + "loss": 0.2832, + "step": 50058 + }, + { + "epoch": 1.92, + "learning_rate": 1.793472685303924e-05, + "loss": 0.2516, + "step": 50059 + }, + { + "epoch": 1.92, + "learning_rate": 1.793408566299051e-05, + "loss": 0.3392, + "step": 50060 + }, + { + "epoch": 1.92, + "learning_rate": 1.793344447294178e-05, + "loss": 0.2658, + "step": 50061 + }, + { + "epoch": 1.92, + "learning_rate": 1.793280328289305e-05, + "loss": 0.3364, + "step": 50062 + }, + { + "epoch": 1.92, + "learning_rate": 1.7932162092844322e-05, + "loss": 0.2893, + "step": 50063 + }, + { + "epoch": 1.92, + "learning_rate": 1.793152090279559e-05, + "loss": 0.2845, + "step": 50064 + }, + { + "epoch": 1.92, + "learning_rate": 1.793087971274686e-05, + "loss": 0.247, + "step": 50065 + }, + { + "epoch": 1.92, + "learning_rate": 1.793023852269813e-05, + "loss": 0.463, + "step": 50066 + }, + { + "epoch": 1.92, + "learning_rate": 1.79295973326494e-05, + "loss": 0.3239, + "step": 50067 + }, + { + "epoch": 1.92, + "learning_rate": 1.7928956142600667e-05, + "loss": 0.2644, + "step": 50068 + }, + { + "epoch": 1.92, + "learning_rate": 1.7928314952551935e-05, + "loss": 0.3205, + "step": 50069 + }, + { + "epoch": 1.93, + "learning_rate": 1.7927673762503207e-05, + "loss": 0.4087, + "step": 50070 + }, + { + "epoch": 1.93, + "learning_rate": 1.792703257245448e-05, + "loss": 0.2669, + "step": 50071 + }, + { + "epoch": 1.93, + "learning_rate": 1.7926391382405747e-05, + "loss": 0.2362, + "step": 50072 + }, + { + "epoch": 1.93, + "learning_rate": 1.7925750192357015e-05, + "loss": 0.3475, + "step": 50073 + }, + { + "epoch": 1.93, + "learning_rate": 1.7925109002308284e-05, + "loss": 0.3037, + "step": 50074 + }, + { + "epoch": 1.93, + "learning_rate": 1.7924467812259555e-05, + "loss": 0.2794, + "step": 50075 + }, + { + "epoch": 1.93, + "learning_rate": 1.7923826622210824e-05, + "loss": 0.3169, + "step": 50076 + }, + { + "epoch": 1.93, + "learning_rate": 1.7923185432162092e-05, + "loss": 0.2888, + "step": 50077 + }, + { + "epoch": 1.93, + "learning_rate": 1.7922544242113364e-05, + "loss": 0.2946, + "step": 50078 + }, + { + "epoch": 1.93, + "learning_rate": 1.7921903052064635e-05, + "loss": 0.4859, + "step": 50079 + }, + { + "epoch": 1.93, + "learning_rate": 1.7921261862015904e-05, + "loss": 0.2553, + "step": 50080 + }, + { + "epoch": 1.93, + "learning_rate": 1.7920620671967172e-05, + "loss": 0.5122, + "step": 50081 + }, + { + "epoch": 1.93, + "learning_rate": 1.791997948191844e-05, + "loss": 0.2966, + "step": 50082 + }, + { + "epoch": 1.93, + "learning_rate": 1.7919338291869712e-05, + "loss": 0.4586, + "step": 50083 + }, + { + "epoch": 1.93, + "learning_rate": 1.791869710182098e-05, + "loss": 0.352, + "step": 50084 + }, + { + "epoch": 1.93, + "learning_rate": 1.791805591177225e-05, + "loss": 0.2937, + "step": 50085 + }, + { + "epoch": 1.93, + "learning_rate": 1.791741472172352e-05, + "loss": 0.2655, + "step": 50086 + }, + { + "epoch": 1.93, + "learning_rate": 1.791677353167479e-05, + "loss": 0.298, + "step": 50087 + }, + { + "epoch": 1.93, + "learning_rate": 1.791613234162606e-05, + "loss": 0.402, + "step": 50088 + }, + { + "epoch": 1.93, + "learning_rate": 1.791549115157733e-05, + "loss": 0.3185, + "step": 50089 + }, + { + "epoch": 1.93, + "learning_rate": 1.7914849961528597e-05, + "loss": 0.2319, + "step": 50090 + }, + { + "epoch": 1.93, + "learning_rate": 1.791420877147987e-05, + "loss": 0.2246, + "step": 50091 + }, + { + "epoch": 1.93, + "learning_rate": 1.7913567581431137e-05, + "loss": 0.3538, + "step": 50092 + }, + { + "epoch": 1.93, + "learning_rate": 1.7912926391382405e-05, + "loss": 0.2176, + "step": 50093 + }, + { + "epoch": 1.93, + "learning_rate": 1.7912285201333677e-05, + "loss": 0.3187, + "step": 50094 + }, + { + "epoch": 1.93, + "learning_rate": 1.7911644011284946e-05, + "loss": 0.2233, + "step": 50095 + }, + { + "epoch": 1.93, + "learning_rate": 1.7911002821236214e-05, + "loss": 0.2789, + "step": 50096 + }, + { + "epoch": 1.93, + "learning_rate": 1.7910361631187486e-05, + "loss": 0.3867, + "step": 50097 + }, + { + "epoch": 1.93, + "learning_rate": 1.7909720441138754e-05, + "loss": 0.3203, + "step": 50098 + }, + { + "epoch": 1.93, + "learning_rate": 1.7909079251090026e-05, + "loss": 0.261, + "step": 50099 + }, + { + "epoch": 1.93, + "learning_rate": 1.7908438061041294e-05, + "loss": 0.4437, + "step": 50100 + }, + { + "epoch": 1.93, + "learning_rate": 1.7907796870992562e-05, + "loss": 0.3328, + "step": 50101 + }, + { + "epoch": 1.93, + "learning_rate": 1.7907155680943834e-05, + "loss": 0.2906, + "step": 50102 + }, + { + "epoch": 1.93, + "learning_rate": 1.7906514490895102e-05, + "loss": 0.4863, + "step": 50103 + }, + { + "epoch": 1.93, + "learning_rate": 1.790587330084637e-05, + "loss": 0.3638, + "step": 50104 + }, + { + "epoch": 1.93, + "learning_rate": 1.790523211079764e-05, + "loss": 0.2928, + "step": 50105 + }, + { + "epoch": 1.93, + "learning_rate": 1.790459092074891e-05, + "loss": 0.556, + "step": 50106 + }, + { + "epoch": 1.93, + "learning_rate": 1.7903949730700182e-05, + "loss": 0.2014, + "step": 50107 + }, + { + "epoch": 1.93, + "learning_rate": 1.790330854065145e-05, + "loss": 0.2995, + "step": 50108 + }, + { + "epoch": 1.93, + "learning_rate": 1.790266735060272e-05, + "loss": 0.1858, + "step": 50109 + }, + { + "epoch": 1.93, + "learning_rate": 1.7902026160553987e-05, + "loss": 0.2719, + "step": 50110 + }, + { + "epoch": 1.93, + "learning_rate": 1.790138497050526e-05, + "loss": 0.2863, + "step": 50111 + }, + { + "epoch": 1.93, + "learning_rate": 1.7900743780456527e-05, + "loss": 0.2629, + "step": 50112 + }, + { + "epoch": 1.93, + "learning_rate": 1.7900102590407796e-05, + "loss": 0.2556, + "step": 50113 + }, + { + "epoch": 1.93, + "learning_rate": 1.7899461400359067e-05, + "loss": 0.336, + "step": 50114 + }, + { + "epoch": 1.93, + "learning_rate": 1.789882021031034e-05, + "loss": 0.3292, + "step": 50115 + }, + { + "epoch": 1.93, + "learning_rate": 1.7898179020261608e-05, + "loss": 0.2941, + "step": 50116 + }, + { + "epoch": 1.93, + "learning_rate": 1.7897537830212876e-05, + "loss": 0.3017, + "step": 50117 + }, + { + "epoch": 1.93, + "learning_rate": 1.7896896640164144e-05, + "loss": 0.2808, + "step": 50118 + }, + { + "epoch": 1.93, + "learning_rate": 1.7896255450115416e-05, + "loss": 0.3728, + "step": 50119 + }, + { + "epoch": 1.93, + "learning_rate": 1.7895614260066684e-05, + "loss": 0.2309, + "step": 50120 + }, + { + "epoch": 1.93, + "learning_rate": 1.7894973070017953e-05, + "loss": 0.2147, + "step": 50121 + }, + { + "epoch": 1.93, + "learning_rate": 1.7894331879969224e-05, + "loss": 0.2227, + "step": 50122 + }, + { + "epoch": 1.93, + "learning_rate": 1.7893690689920493e-05, + "loss": 0.2676, + "step": 50123 + }, + { + "epoch": 1.93, + "learning_rate": 1.7893049499871764e-05, + "loss": 0.545, + "step": 50124 + }, + { + "epoch": 1.93, + "learning_rate": 1.7892408309823033e-05, + "loss": 0.4004, + "step": 50125 + }, + { + "epoch": 1.93, + "learning_rate": 1.78917671197743e-05, + "loss": 0.3789, + "step": 50126 + }, + { + "epoch": 1.93, + "learning_rate": 1.7891125929725573e-05, + "loss": 0.4421, + "step": 50127 + }, + { + "epoch": 1.93, + "learning_rate": 1.789048473967684e-05, + "loss": 0.4403, + "step": 50128 + }, + { + "epoch": 1.93, + "learning_rate": 1.788984354962811e-05, + "loss": 0.247, + "step": 50129 + }, + { + "epoch": 1.93, + "learning_rate": 1.788920235957938e-05, + "loss": 0.2227, + "step": 50130 + }, + { + "epoch": 1.93, + "learning_rate": 1.788856116953065e-05, + "loss": 0.2913, + "step": 50131 + }, + { + "epoch": 1.93, + "learning_rate": 1.7887919979481918e-05, + "loss": 0.2954, + "step": 50132 + }, + { + "epoch": 1.93, + "learning_rate": 1.788727878943319e-05, + "loss": 0.2801, + "step": 50133 + }, + { + "epoch": 1.93, + "learning_rate": 1.7886637599384458e-05, + "loss": 0.3204, + "step": 50134 + }, + { + "epoch": 1.93, + "learning_rate": 1.788599640933573e-05, + "loss": 0.2931, + "step": 50135 + }, + { + "epoch": 1.93, + "learning_rate": 1.7885355219286998e-05, + "loss": 0.188, + "step": 50136 + }, + { + "epoch": 1.93, + "learning_rate": 1.7884714029238266e-05, + "loss": 0.2798, + "step": 50137 + }, + { + "epoch": 1.93, + "learning_rate": 1.7884072839189538e-05, + "loss": 0.2834, + "step": 50138 + }, + { + "epoch": 1.93, + "learning_rate": 1.7883431649140806e-05, + "loss": 0.3949, + "step": 50139 + }, + { + "epoch": 1.93, + "learning_rate": 1.7882790459092075e-05, + "loss": 0.4173, + "step": 50140 + }, + { + "epoch": 1.93, + "learning_rate": 1.7882149269043343e-05, + "loss": 0.393, + "step": 50141 + }, + { + "epoch": 1.93, + "learning_rate": 1.7881508078994615e-05, + "loss": 0.3715, + "step": 50142 + }, + { + "epoch": 1.93, + "learning_rate": 1.7880866888945886e-05, + "loss": 0.2499, + "step": 50143 + }, + { + "epoch": 1.93, + "learning_rate": 1.7880225698897155e-05, + "loss": 0.3162, + "step": 50144 + }, + { + "epoch": 1.93, + "learning_rate": 1.7879584508848423e-05, + "loss": 0.3036, + "step": 50145 + }, + { + "epoch": 1.93, + "learning_rate": 1.7878943318799695e-05, + "loss": 0.4141, + "step": 50146 + }, + { + "epoch": 1.93, + "learning_rate": 1.7878302128750963e-05, + "loss": 0.4424, + "step": 50147 + }, + { + "epoch": 1.93, + "learning_rate": 1.787766093870223e-05, + "loss": 0.2405, + "step": 50148 + }, + { + "epoch": 1.93, + "learning_rate": 1.78770197486535e-05, + "loss": 0.2494, + "step": 50149 + }, + { + "epoch": 1.93, + "learning_rate": 1.787637855860477e-05, + "loss": 0.23, + "step": 50150 + }, + { + "epoch": 1.93, + "learning_rate": 1.7875737368556043e-05, + "loss": 0.1945, + "step": 50151 + }, + { + "epoch": 1.93, + "learning_rate": 1.787509617850731e-05, + "loss": 0.2973, + "step": 50152 + }, + { + "epoch": 1.93, + "learning_rate": 1.787445498845858e-05, + "loss": 0.447, + "step": 50153 + }, + { + "epoch": 1.93, + "learning_rate": 1.7873813798409848e-05, + "loss": 0.3317, + "step": 50154 + }, + { + "epoch": 1.93, + "learning_rate": 1.787317260836112e-05, + "loss": 0.2489, + "step": 50155 + }, + { + "epoch": 1.93, + "learning_rate": 1.7872531418312388e-05, + "loss": 0.2728, + "step": 50156 + }, + { + "epoch": 1.93, + "learning_rate": 1.7871890228263656e-05, + "loss": 0.2535, + "step": 50157 + }, + { + "epoch": 1.93, + "learning_rate": 1.7871249038214928e-05, + "loss": 0.2912, + "step": 50158 + }, + { + "epoch": 1.93, + "learning_rate": 1.7870607848166196e-05, + "loss": 0.2198, + "step": 50159 + }, + { + "epoch": 1.93, + "learning_rate": 1.7869966658117468e-05, + "loss": 0.3273, + "step": 50160 + }, + { + "epoch": 1.93, + "learning_rate": 1.7869325468068737e-05, + "loss": 0.4678, + "step": 50161 + }, + { + "epoch": 1.93, + "learning_rate": 1.7868684278020005e-05, + "loss": 0.3837, + "step": 50162 + }, + { + "epoch": 1.93, + "learning_rate": 1.7868043087971277e-05, + "loss": 0.2469, + "step": 50163 + }, + { + "epoch": 1.93, + "learning_rate": 1.7867401897922545e-05, + "loss": 0.3175, + "step": 50164 + }, + { + "epoch": 1.93, + "learning_rate": 1.7866760707873813e-05, + "loss": 0.2683, + "step": 50165 + }, + { + "epoch": 1.93, + "learning_rate": 1.7866119517825085e-05, + "loss": 0.2652, + "step": 50166 + }, + { + "epoch": 1.93, + "learning_rate": 1.7865478327776353e-05, + "loss": 0.1993, + "step": 50167 + }, + { + "epoch": 1.93, + "learning_rate": 1.7864837137727625e-05, + "loss": 0.2818, + "step": 50168 + }, + { + "epoch": 1.93, + "learning_rate": 1.7864195947678893e-05, + "loss": 0.3199, + "step": 50169 + }, + { + "epoch": 1.93, + "learning_rate": 1.786355475763016e-05, + "loss": 0.2418, + "step": 50170 + }, + { + "epoch": 1.93, + "learning_rate": 1.7862913567581433e-05, + "loss": 0.3191, + "step": 50171 + }, + { + "epoch": 1.93, + "learning_rate": 1.78622723775327e-05, + "loss": 0.2652, + "step": 50172 + }, + { + "epoch": 1.93, + "learning_rate": 1.786163118748397e-05, + "loss": 0.309, + "step": 50173 + }, + { + "epoch": 1.93, + "learning_rate": 1.7860989997435242e-05, + "loss": 0.2192, + "step": 50174 + }, + { + "epoch": 1.93, + "learning_rate": 1.786034880738651e-05, + "loss": 0.3114, + "step": 50175 + }, + { + "epoch": 1.93, + "learning_rate": 1.785970761733778e-05, + "loss": 0.279, + "step": 50176 + }, + { + "epoch": 1.93, + "learning_rate": 1.785906642728905e-05, + "loss": 0.251, + "step": 50177 + }, + { + "epoch": 1.93, + "learning_rate": 1.785842523724032e-05, + "loss": 0.4441, + "step": 50178 + }, + { + "epoch": 1.93, + "learning_rate": 1.785778404719159e-05, + "loss": 0.2872, + "step": 50179 + }, + { + "epoch": 1.93, + "learning_rate": 1.785714285714286e-05, + "loss": 0.3141, + "step": 50180 + }, + { + "epoch": 1.93, + "learning_rate": 1.7856501667094127e-05, + "loss": 0.2567, + "step": 50181 + }, + { + "epoch": 1.93, + "learning_rate": 1.78558604770454e-05, + "loss": 0.3823, + "step": 50182 + }, + { + "epoch": 1.93, + "learning_rate": 1.7855219286996667e-05, + "loss": 0.3759, + "step": 50183 + }, + { + "epoch": 1.93, + "learning_rate": 1.7854578096947935e-05, + "loss": 0.2841, + "step": 50184 + }, + { + "epoch": 1.93, + "learning_rate": 1.7853936906899204e-05, + "loss": 0.2536, + "step": 50185 + }, + { + "epoch": 1.93, + "learning_rate": 1.7853295716850475e-05, + "loss": 0.3889, + "step": 50186 + }, + { + "epoch": 1.93, + "learning_rate": 1.7852654526801747e-05, + "loss": 0.2495, + "step": 50187 + }, + { + "epoch": 1.93, + "learning_rate": 1.7852013336753015e-05, + "loss": 0.3331, + "step": 50188 + }, + { + "epoch": 1.93, + "learning_rate": 1.7851372146704284e-05, + "loss": 0.407, + "step": 50189 + }, + { + "epoch": 1.93, + "learning_rate": 1.7850730956655552e-05, + "loss": 0.4237, + "step": 50190 + }, + { + "epoch": 1.93, + "learning_rate": 1.7850089766606824e-05, + "loss": 0.2789, + "step": 50191 + }, + { + "epoch": 1.93, + "learning_rate": 1.7849448576558092e-05, + "loss": 0.3333, + "step": 50192 + }, + { + "epoch": 1.93, + "learning_rate": 1.784880738650936e-05, + "loss": 0.2809, + "step": 50193 + }, + { + "epoch": 1.93, + "learning_rate": 1.7848166196460632e-05, + "loss": 0.3667, + "step": 50194 + }, + { + "epoch": 1.93, + "learning_rate": 1.7847525006411904e-05, + "loss": 0.2774, + "step": 50195 + }, + { + "epoch": 1.93, + "learning_rate": 1.7846883816363172e-05, + "loss": 0.2269, + "step": 50196 + }, + { + "epoch": 1.93, + "learning_rate": 1.784624262631444e-05, + "loss": 0.2987, + "step": 50197 + }, + { + "epoch": 1.93, + "learning_rate": 1.784560143626571e-05, + "loss": 0.4218, + "step": 50198 + }, + { + "epoch": 1.93, + "learning_rate": 1.784496024621698e-05, + "loss": 0.2666, + "step": 50199 + }, + { + "epoch": 1.93, + "learning_rate": 1.784431905616825e-05, + "loss": 0.3251, + "step": 50200 + }, + { + "epoch": 1.93, + "learning_rate": 1.7843677866119517e-05, + "loss": 0.4654, + "step": 50201 + }, + { + "epoch": 1.93, + "learning_rate": 1.784303667607079e-05, + "loss": 0.3268, + "step": 50202 + }, + { + "epoch": 1.93, + "learning_rate": 1.7842395486022057e-05, + "loss": 0.4902, + "step": 50203 + }, + { + "epoch": 1.93, + "learning_rate": 1.784175429597333e-05, + "loss": 0.2622, + "step": 50204 + }, + { + "epoch": 1.93, + "learning_rate": 1.7841113105924597e-05, + "loss": 0.4592, + "step": 50205 + }, + { + "epoch": 1.93, + "learning_rate": 1.7840471915875866e-05, + "loss": 0.2563, + "step": 50206 + }, + { + "epoch": 1.93, + "learning_rate": 1.7839830725827137e-05, + "loss": 0.3248, + "step": 50207 + }, + { + "epoch": 1.93, + "learning_rate": 1.7839189535778406e-05, + "loss": 0.3135, + "step": 50208 + }, + { + "epoch": 1.93, + "learning_rate": 1.7838548345729674e-05, + "loss": 0.2989, + "step": 50209 + }, + { + "epoch": 1.93, + "learning_rate": 1.7837907155680946e-05, + "loss": 0.3744, + "step": 50210 + }, + { + "epoch": 1.93, + "learning_rate": 1.7837265965632214e-05, + "loss": 0.3401, + "step": 50211 + }, + { + "epoch": 1.93, + "learning_rate": 1.7836624775583482e-05, + "loss": 0.2984, + "step": 50212 + }, + { + "epoch": 1.93, + "learning_rate": 1.7835983585534754e-05, + "loss": 0.3948, + "step": 50213 + }, + { + "epoch": 1.93, + "learning_rate": 1.7835342395486022e-05, + "loss": 0.2911, + "step": 50214 + }, + { + "epoch": 1.93, + "learning_rate": 1.7834701205437294e-05, + "loss": 0.3241, + "step": 50215 + }, + { + "epoch": 1.93, + "learning_rate": 1.7834060015388562e-05, + "loss": 0.1856, + "step": 50216 + }, + { + "epoch": 1.93, + "learning_rate": 1.783341882533983e-05, + "loss": 0.3632, + "step": 50217 + }, + { + "epoch": 1.93, + "learning_rate": 1.7832777635291102e-05, + "loss": 0.5473, + "step": 50218 + }, + { + "epoch": 1.93, + "learning_rate": 1.783213644524237e-05, + "loss": 0.3272, + "step": 50219 + }, + { + "epoch": 1.93, + "learning_rate": 1.783149525519364e-05, + "loss": 0.5607, + "step": 50220 + }, + { + "epoch": 1.93, + "learning_rate": 1.7830854065144907e-05, + "loss": 0.3674, + "step": 50221 + }, + { + "epoch": 1.93, + "learning_rate": 1.783021287509618e-05, + "loss": 0.258, + "step": 50222 + }, + { + "epoch": 1.93, + "learning_rate": 1.782957168504745e-05, + "loss": 0.4618, + "step": 50223 + }, + { + "epoch": 1.93, + "learning_rate": 1.782893049499872e-05, + "loss": 0.319, + "step": 50224 + }, + { + "epoch": 1.93, + "learning_rate": 1.7828289304949987e-05, + "loss": 0.3549, + "step": 50225 + }, + { + "epoch": 1.93, + "learning_rate": 1.782764811490126e-05, + "loss": 0.2697, + "step": 50226 + }, + { + "epoch": 1.93, + "learning_rate": 1.7827006924852528e-05, + "loss": 0.3399, + "step": 50227 + }, + { + "epoch": 1.93, + "learning_rate": 1.7826365734803796e-05, + "loss": 0.1863, + "step": 50228 + }, + { + "epoch": 1.93, + "learning_rate": 1.7825724544755064e-05, + "loss": 0.2844, + "step": 50229 + }, + { + "epoch": 1.93, + "learning_rate": 1.7825083354706336e-05, + "loss": 0.2639, + "step": 50230 + }, + { + "epoch": 1.93, + "learning_rate": 1.7824442164657608e-05, + "loss": 0.2821, + "step": 50231 + }, + { + "epoch": 1.93, + "learning_rate": 1.7823800974608876e-05, + "loss": 0.2659, + "step": 50232 + }, + { + "epoch": 1.93, + "learning_rate": 1.7823159784560144e-05, + "loss": 0.363, + "step": 50233 + }, + { + "epoch": 1.93, + "learning_rate": 1.7822518594511413e-05, + "loss": 0.4232, + "step": 50234 + }, + { + "epoch": 1.93, + "learning_rate": 1.7821877404462684e-05, + "loss": 0.2797, + "step": 50235 + }, + { + "epoch": 1.93, + "learning_rate": 1.7821236214413953e-05, + "loss": 0.2767, + "step": 50236 + }, + { + "epoch": 1.93, + "learning_rate": 1.782059502436522e-05, + "loss": 0.4129, + "step": 50237 + }, + { + "epoch": 1.93, + "learning_rate": 1.7819953834316493e-05, + "loss": 0.2104, + "step": 50238 + }, + { + "epoch": 1.93, + "learning_rate": 1.781931264426776e-05, + "loss": 0.2747, + "step": 50239 + }, + { + "epoch": 1.93, + "learning_rate": 1.7818671454219033e-05, + "loss": 0.3538, + "step": 50240 + }, + { + "epoch": 1.93, + "learning_rate": 1.78180302641703e-05, + "loss": 0.224, + "step": 50241 + }, + { + "epoch": 1.93, + "learning_rate": 1.781738907412157e-05, + "loss": 0.4301, + "step": 50242 + }, + { + "epoch": 1.93, + "learning_rate": 1.781674788407284e-05, + "loss": 0.3141, + "step": 50243 + }, + { + "epoch": 1.93, + "learning_rate": 1.781610669402411e-05, + "loss": 0.2237, + "step": 50244 + }, + { + "epoch": 1.93, + "learning_rate": 1.7815465503975378e-05, + "loss": 0.3127, + "step": 50245 + }, + { + "epoch": 1.93, + "learning_rate": 1.781482431392665e-05, + "loss": 0.2596, + "step": 50246 + }, + { + "epoch": 1.93, + "learning_rate": 1.7814183123877918e-05, + "loss": 0.3286, + "step": 50247 + }, + { + "epoch": 1.93, + "learning_rate": 1.7813541933829186e-05, + "loss": 0.3863, + "step": 50248 + }, + { + "epoch": 1.93, + "learning_rate": 1.7812900743780458e-05, + "loss": 0.193, + "step": 50249 + }, + { + "epoch": 1.93, + "learning_rate": 1.7812259553731726e-05, + "loss": 0.3071, + "step": 50250 + }, + { + "epoch": 1.93, + "learning_rate": 1.7811618363682998e-05, + "loss": 0.5512, + "step": 50251 + }, + { + "epoch": 1.93, + "learning_rate": 1.7810977173634266e-05, + "loss": 0.3294, + "step": 50252 + }, + { + "epoch": 1.93, + "learning_rate": 1.7810335983585535e-05, + "loss": 0.2722, + "step": 50253 + }, + { + "epoch": 1.93, + "learning_rate": 1.7809694793536806e-05, + "loss": 0.3059, + "step": 50254 + }, + { + "epoch": 1.93, + "learning_rate": 1.7809053603488075e-05, + "loss": 0.4695, + "step": 50255 + }, + { + "epoch": 1.93, + "learning_rate": 1.7808412413439343e-05, + "loss": 0.3192, + "step": 50256 + }, + { + "epoch": 1.93, + "learning_rate": 1.7807771223390615e-05, + "loss": 0.4414, + "step": 50257 + }, + { + "epoch": 1.93, + "learning_rate": 1.7807130033341883e-05, + "loss": 0.2656, + "step": 50258 + }, + { + "epoch": 1.93, + "learning_rate": 1.7806488843293155e-05, + "loss": 0.334, + "step": 50259 + }, + { + "epoch": 1.93, + "learning_rate": 1.7805847653244423e-05, + "loss": 0.3186, + "step": 50260 + }, + { + "epoch": 1.93, + "learning_rate": 1.780520646319569e-05, + "loss": 0.2944, + "step": 50261 + }, + { + "epoch": 1.93, + "learning_rate": 1.7804565273146963e-05, + "loss": 0.2881, + "step": 50262 + }, + { + "epoch": 1.93, + "learning_rate": 1.780392408309823e-05, + "loss": 0.331, + "step": 50263 + }, + { + "epoch": 1.93, + "learning_rate": 1.78032828930495e-05, + "loss": 0.381, + "step": 50264 + }, + { + "epoch": 1.93, + "learning_rate": 1.7802641703000768e-05, + "loss": 0.3472, + "step": 50265 + }, + { + "epoch": 1.93, + "learning_rate": 1.780200051295204e-05, + "loss": 0.3074, + "step": 50266 + }, + { + "epoch": 1.93, + "learning_rate": 1.780135932290331e-05, + "loss": 0.3019, + "step": 50267 + }, + { + "epoch": 1.93, + "learning_rate": 1.780071813285458e-05, + "loss": 0.2418, + "step": 50268 + }, + { + "epoch": 1.93, + "learning_rate": 1.7800076942805848e-05, + "loss": 0.3704, + "step": 50269 + }, + { + "epoch": 1.93, + "learning_rate": 1.779943575275712e-05, + "loss": 0.2299, + "step": 50270 + }, + { + "epoch": 1.93, + "learning_rate": 1.7798794562708388e-05, + "loss": 0.2739, + "step": 50271 + }, + { + "epoch": 1.93, + "learning_rate": 1.7798153372659657e-05, + "loss": 0.3124, + "step": 50272 + }, + { + "epoch": 1.93, + "learning_rate": 1.7797512182610925e-05, + "loss": 0.3858, + "step": 50273 + }, + { + "epoch": 1.93, + "learning_rate": 1.7796870992562197e-05, + "loss": 0.2395, + "step": 50274 + }, + { + "epoch": 1.93, + "learning_rate": 1.7796229802513468e-05, + "loss": 0.289, + "step": 50275 + }, + { + "epoch": 1.93, + "learning_rate": 1.7795588612464737e-05, + "loss": 0.2558, + "step": 50276 + }, + { + "epoch": 1.93, + "learning_rate": 1.7794947422416005e-05, + "loss": 0.3538, + "step": 50277 + }, + { + "epoch": 1.93, + "learning_rate": 1.7794306232367273e-05, + "loss": 0.381, + "step": 50278 + }, + { + "epoch": 1.93, + "learning_rate": 1.7793665042318545e-05, + "loss": 0.2535, + "step": 50279 + }, + { + "epoch": 1.93, + "learning_rate": 1.7793023852269813e-05, + "loss": 0.3178, + "step": 50280 + }, + { + "epoch": 1.93, + "learning_rate": 1.779238266222108e-05, + "loss": 0.3322, + "step": 50281 + }, + { + "epoch": 1.93, + "learning_rate": 1.7791741472172353e-05, + "loss": 0.2793, + "step": 50282 + }, + { + "epoch": 1.93, + "learning_rate": 1.779110028212362e-05, + "loss": 0.3486, + "step": 50283 + }, + { + "epoch": 1.93, + "learning_rate": 1.7790459092074893e-05, + "loss": 0.3131, + "step": 50284 + }, + { + "epoch": 1.93, + "learning_rate": 1.7789817902026162e-05, + "loss": 0.3272, + "step": 50285 + }, + { + "epoch": 1.93, + "learning_rate": 1.778917671197743e-05, + "loss": 0.4146, + "step": 50286 + }, + { + "epoch": 1.93, + "learning_rate": 1.7788535521928702e-05, + "loss": 0.248, + "step": 50287 + }, + { + "epoch": 1.93, + "learning_rate": 1.778789433187997e-05, + "loss": 0.1913, + "step": 50288 + }, + { + "epoch": 1.93, + "learning_rate": 1.778725314183124e-05, + "loss": 0.3233, + "step": 50289 + }, + { + "epoch": 1.93, + "learning_rate": 1.778661195178251e-05, + "loss": 0.2828, + "step": 50290 + }, + { + "epoch": 1.93, + "learning_rate": 1.778597076173378e-05, + "loss": 0.4028, + "step": 50291 + }, + { + "epoch": 1.93, + "learning_rate": 1.7785329571685047e-05, + "loss": 0.3182, + "step": 50292 + }, + { + "epoch": 1.93, + "learning_rate": 1.778468838163632e-05, + "loss": 0.2828, + "step": 50293 + }, + { + "epoch": 1.93, + "learning_rate": 1.7784047191587587e-05, + "loss": 0.4016, + "step": 50294 + }, + { + "epoch": 1.93, + "learning_rate": 1.778340600153886e-05, + "loss": 0.3833, + "step": 50295 + }, + { + "epoch": 1.93, + "learning_rate": 1.7782764811490127e-05, + "loss": 0.2695, + "step": 50296 + }, + { + "epoch": 1.93, + "learning_rate": 1.7782123621441395e-05, + "loss": 0.1993, + "step": 50297 + }, + { + "epoch": 1.93, + "learning_rate": 1.7781482431392667e-05, + "loss": 0.3693, + "step": 50298 + }, + { + "epoch": 1.93, + "learning_rate": 1.7780841241343935e-05, + "loss": 0.3493, + "step": 50299 + }, + { + "epoch": 1.93, + "learning_rate": 1.7780200051295204e-05, + "loss": 0.4048, + "step": 50300 + }, + { + "epoch": 1.93, + "learning_rate": 1.7779558861246472e-05, + "loss": 0.4021, + "step": 50301 + }, + { + "epoch": 1.93, + "learning_rate": 1.7778917671197744e-05, + "loss": 0.2319, + "step": 50302 + }, + { + "epoch": 1.93, + "learning_rate": 1.7778276481149015e-05, + "loss": 0.2471, + "step": 50303 + }, + { + "epoch": 1.93, + "learning_rate": 1.7777635291100284e-05, + "loss": 0.3806, + "step": 50304 + }, + { + "epoch": 1.93, + "learning_rate": 1.7776994101051552e-05, + "loss": 0.3841, + "step": 50305 + }, + { + "epoch": 1.93, + "learning_rate": 1.7776352911002824e-05, + "loss": 0.2889, + "step": 50306 + }, + { + "epoch": 1.93, + "learning_rate": 1.7775711720954092e-05, + "loss": 0.2447, + "step": 50307 + }, + { + "epoch": 1.93, + "learning_rate": 1.777507053090536e-05, + "loss": 0.2587, + "step": 50308 + }, + { + "epoch": 1.93, + "learning_rate": 1.777442934085663e-05, + "loss": 0.2494, + "step": 50309 + }, + { + "epoch": 1.93, + "learning_rate": 1.77737881508079e-05, + "loss": 0.4059, + "step": 50310 + }, + { + "epoch": 1.93, + "learning_rate": 1.7773146960759172e-05, + "loss": 0.2374, + "step": 50311 + }, + { + "epoch": 1.93, + "learning_rate": 1.777250577071044e-05, + "loss": 0.4565, + "step": 50312 + }, + { + "epoch": 1.93, + "learning_rate": 1.777186458066171e-05, + "loss": 0.4037, + "step": 50313 + }, + { + "epoch": 1.93, + "learning_rate": 1.7771223390612977e-05, + "loss": 0.3956, + "step": 50314 + }, + { + "epoch": 1.93, + "learning_rate": 1.777058220056425e-05, + "loss": 0.4305, + "step": 50315 + }, + { + "epoch": 1.93, + "learning_rate": 1.7769941010515517e-05, + "loss": 0.3114, + "step": 50316 + }, + { + "epoch": 1.93, + "learning_rate": 1.7769299820466786e-05, + "loss": 0.2135, + "step": 50317 + }, + { + "epoch": 1.93, + "learning_rate": 1.7768658630418057e-05, + "loss": 0.3202, + "step": 50318 + }, + { + "epoch": 1.93, + "learning_rate": 1.7768017440369326e-05, + "loss": 0.297, + "step": 50319 + }, + { + "epoch": 1.93, + "learning_rate": 1.7767376250320597e-05, + "loss": 0.3613, + "step": 50320 + }, + { + "epoch": 1.93, + "learning_rate": 1.7766735060271866e-05, + "loss": 0.1939, + "step": 50321 + }, + { + "epoch": 1.93, + "learning_rate": 1.7766093870223134e-05, + "loss": 0.4686, + "step": 50322 + }, + { + "epoch": 1.93, + "learning_rate": 1.7765452680174406e-05, + "loss": 0.2769, + "step": 50323 + }, + { + "epoch": 1.93, + "learning_rate": 1.7764811490125674e-05, + "loss": 0.3841, + "step": 50324 + }, + { + "epoch": 1.93, + "learning_rate": 1.7764170300076942e-05, + "loss": 0.3385, + "step": 50325 + }, + { + "epoch": 1.93, + "learning_rate": 1.7763529110028214e-05, + "loss": 0.298, + "step": 50326 + }, + { + "epoch": 1.93, + "learning_rate": 1.7762887919979482e-05, + "loss": 0.466, + "step": 50327 + }, + { + "epoch": 1.93, + "learning_rate": 1.776224672993075e-05, + "loss": 0.2208, + "step": 50328 + }, + { + "epoch": 1.93, + "learning_rate": 1.7761605539882022e-05, + "loss": 0.4126, + "step": 50329 + }, + { + "epoch": 1.94, + "learning_rate": 1.776096434983329e-05, + "loss": 0.3164, + "step": 50330 + }, + { + "epoch": 1.94, + "learning_rate": 1.7760323159784562e-05, + "loss": 0.3377, + "step": 50331 + }, + { + "epoch": 1.94, + "learning_rate": 1.775968196973583e-05, + "loss": 0.2234, + "step": 50332 + }, + { + "epoch": 1.94, + "learning_rate": 1.77590407796871e-05, + "loss": 0.262, + "step": 50333 + }, + { + "epoch": 1.94, + "learning_rate": 1.775839958963837e-05, + "loss": 0.3362, + "step": 50334 + }, + { + "epoch": 1.94, + "learning_rate": 1.775775839958964e-05, + "loss": 0.3076, + "step": 50335 + }, + { + "epoch": 1.94, + "learning_rate": 1.7757117209540907e-05, + "loss": 0.4464, + "step": 50336 + }, + { + "epoch": 1.94, + "learning_rate": 1.7756476019492176e-05, + "loss": 0.2491, + "step": 50337 + }, + { + "epoch": 1.94, + "learning_rate": 1.7755834829443448e-05, + "loss": 0.2497, + "step": 50338 + }, + { + "epoch": 1.94, + "learning_rate": 1.775519363939472e-05, + "loss": 0.4365, + "step": 50339 + }, + { + "epoch": 1.94, + "learning_rate": 1.7754552449345988e-05, + "loss": 0.4377, + "step": 50340 + }, + { + "epoch": 1.94, + "learning_rate": 1.7753911259297256e-05, + "loss": 0.3984, + "step": 50341 + }, + { + "epoch": 1.94, + "learning_rate": 1.7753270069248528e-05, + "loss": 0.219, + "step": 50342 + }, + { + "epoch": 1.94, + "learning_rate": 1.7752628879199796e-05, + "loss": 0.2397, + "step": 50343 + }, + { + "epoch": 1.94, + "learning_rate": 1.7751987689151064e-05, + "loss": 0.3231, + "step": 50344 + }, + { + "epoch": 1.94, + "learning_rate": 1.7751346499102333e-05, + "loss": 0.2813, + "step": 50345 + }, + { + "epoch": 1.94, + "learning_rate": 1.7750705309053604e-05, + "loss": 0.2495, + "step": 50346 + }, + { + "epoch": 1.94, + "learning_rate": 1.7750064119004876e-05, + "loss": 0.3352, + "step": 50347 + }, + { + "epoch": 1.94, + "learning_rate": 1.7749422928956144e-05, + "loss": 0.2752, + "step": 50348 + }, + { + "epoch": 1.94, + "learning_rate": 1.7748781738907413e-05, + "loss": 0.2772, + "step": 50349 + }, + { + "epoch": 1.94, + "learning_rate": 1.7748140548858684e-05, + "loss": 0.1982, + "step": 50350 + }, + { + "epoch": 1.94, + "learning_rate": 1.7747499358809953e-05, + "loss": 0.5441, + "step": 50351 + }, + { + "epoch": 1.94, + "learning_rate": 1.774685816876122e-05, + "loss": 0.3152, + "step": 50352 + }, + { + "epoch": 1.94, + "learning_rate": 1.774621697871249e-05, + "loss": 0.2486, + "step": 50353 + }, + { + "epoch": 1.94, + "learning_rate": 1.774557578866376e-05, + "loss": 0.2382, + "step": 50354 + }, + { + "epoch": 1.94, + "learning_rate": 1.7744934598615033e-05, + "loss": 0.3427, + "step": 50355 + }, + { + "epoch": 1.94, + "learning_rate": 1.77442934085663e-05, + "loss": 0.3684, + "step": 50356 + }, + { + "epoch": 1.94, + "learning_rate": 1.774365221851757e-05, + "loss": 0.3037, + "step": 50357 + }, + { + "epoch": 1.94, + "learning_rate": 1.7743011028468838e-05, + "loss": 0.3581, + "step": 50358 + }, + { + "epoch": 1.94, + "learning_rate": 1.774236983842011e-05, + "loss": 0.2265, + "step": 50359 + }, + { + "epoch": 1.94, + "learning_rate": 1.7741728648371378e-05, + "loss": 0.2798, + "step": 50360 + }, + { + "epoch": 1.94, + "learning_rate": 1.7741087458322646e-05, + "loss": 0.3202, + "step": 50361 + }, + { + "epoch": 1.94, + "learning_rate": 1.7740446268273918e-05, + "loss": 0.2853, + "step": 50362 + }, + { + "epoch": 1.94, + "learning_rate": 1.7739805078225186e-05, + "loss": 0.3864, + "step": 50363 + }, + { + "epoch": 1.94, + "learning_rate": 1.7739163888176458e-05, + "loss": 0.2099, + "step": 50364 + }, + { + "epoch": 1.94, + "learning_rate": 1.7738522698127726e-05, + "loss": 0.2244, + "step": 50365 + }, + { + "epoch": 1.94, + "learning_rate": 1.7737881508078995e-05, + "loss": 0.4329, + "step": 50366 + }, + { + "epoch": 1.94, + "learning_rate": 1.7737240318030266e-05, + "loss": 0.3514, + "step": 50367 + }, + { + "epoch": 1.94, + "learning_rate": 1.7736599127981535e-05, + "loss": 0.3618, + "step": 50368 + }, + { + "epoch": 1.94, + "learning_rate": 1.7735957937932803e-05, + "loss": 0.2885, + "step": 50369 + }, + { + "epoch": 1.94, + "learning_rate": 1.7735316747884075e-05, + "loss": 0.3863, + "step": 50370 + }, + { + "epoch": 1.94, + "learning_rate": 1.7734675557835343e-05, + "loss": 0.4258, + "step": 50371 + }, + { + "epoch": 1.94, + "learning_rate": 1.773403436778661e-05, + "loss": 0.4514, + "step": 50372 + }, + { + "epoch": 1.94, + "learning_rate": 1.7733393177737883e-05, + "loss": 0.3437, + "step": 50373 + }, + { + "epoch": 1.94, + "learning_rate": 1.773275198768915e-05, + "loss": 0.3123, + "step": 50374 + }, + { + "epoch": 1.94, + "learning_rate": 1.7732110797640423e-05, + "loss": 0.3076, + "step": 50375 + }, + { + "epoch": 1.94, + "learning_rate": 1.773146960759169e-05, + "loss": 0.2304, + "step": 50376 + }, + { + "epoch": 1.94, + "learning_rate": 1.773082841754296e-05, + "loss": 0.3445, + "step": 50377 + }, + { + "epoch": 1.94, + "learning_rate": 1.773018722749423e-05, + "loss": 0.2468, + "step": 50378 + }, + { + "epoch": 1.94, + "learning_rate": 1.77295460374455e-05, + "loss": 0.3776, + "step": 50379 + }, + { + "epoch": 1.94, + "learning_rate": 1.7728904847396768e-05, + "loss": 0.3009, + "step": 50380 + }, + { + "epoch": 1.94, + "learning_rate": 1.7728263657348036e-05, + "loss": 0.3545, + "step": 50381 + }, + { + "epoch": 1.94, + "learning_rate": 1.7727622467299308e-05, + "loss": 0.3646, + "step": 50382 + }, + { + "epoch": 1.94, + "learning_rate": 1.772698127725058e-05, + "loss": 0.2576, + "step": 50383 + }, + { + "epoch": 1.94, + "learning_rate": 1.7726340087201848e-05, + "loss": 0.325, + "step": 50384 + }, + { + "epoch": 1.94, + "learning_rate": 1.7725698897153117e-05, + "loss": 0.3515, + "step": 50385 + }, + { + "epoch": 1.94, + "learning_rate": 1.7725057707104388e-05, + "loss": 0.2856, + "step": 50386 + }, + { + "epoch": 1.94, + "learning_rate": 1.7724416517055657e-05, + "loss": 0.3687, + "step": 50387 + }, + { + "epoch": 1.94, + "learning_rate": 1.7723775327006925e-05, + "loss": 0.4415, + "step": 50388 + }, + { + "epoch": 1.94, + "learning_rate": 1.7723134136958193e-05, + "loss": 0.4176, + "step": 50389 + }, + { + "epoch": 1.94, + "learning_rate": 1.7722492946909465e-05, + "loss": 0.252, + "step": 50390 + }, + { + "epoch": 1.94, + "learning_rate": 1.7721851756860737e-05, + "loss": 0.2627, + "step": 50391 + }, + { + "epoch": 1.94, + "learning_rate": 1.7721210566812005e-05, + "loss": 0.3109, + "step": 50392 + }, + { + "epoch": 1.94, + "learning_rate": 1.7720569376763273e-05, + "loss": 0.2447, + "step": 50393 + }, + { + "epoch": 1.94, + "learning_rate": 1.771992818671454e-05, + "loss": 0.3337, + "step": 50394 + }, + { + "epoch": 1.94, + "learning_rate": 1.7719286996665813e-05, + "loss": 0.328, + "step": 50395 + }, + { + "epoch": 1.94, + "learning_rate": 1.7718645806617082e-05, + "loss": 0.2644, + "step": 50396 + }, + { + "epoch": 1.94, + "learning_rate": 1.771800461656835e-05, + "loss": 0.3724, + "step": 50397 + }, + { + "epoch": 1.94, + "learning_rate": 1.7717363426519622e-05, + "loss": 0.2766, + "step": 50398 + }, + { + "epoch": 1.94, + "learning_rate": 1.771672223647089e-05, + "loss": 0.2588, + "step": 50399 + }, + { + "epoch": 1.94, + "learning_rate": 1.7716081046422162e-05, + "loss": 0.2851, + "step": 50400 + }, + { + "epoch": 1.94, + "learning_rate": 1.771543985637343e-05, + "loss": 0.2005, + "step": 50401 + }, + { + "epoch": 1.94, + "learning_rate": 1.77147986663247e-05, + "loss": 0.278, + "step": 50402 + }, + { + "epoch": 1.94, + "learning_rate": 1.771415747627597e-05, + "loss": 0.2453, + "step": 50403 + }, + { + "epoch": 1.94, + "learning_rate": 1.771351628622724e-05, + "loss": 0.4912, + "step": 50404 + }, + { + "epoch": 1.94, + "learning_rate": 1.7712875096178507e-05, + "loss": 0.2764, + "step": 50405 + }, + { + "epoch": 1.94, + "learning_rate": 1.771223390612978e-05, + "loss": 0.2601, + "step": 50406 + }, + { + "epoch": 1.94, + "learning_rate": 1.7711592716081047e-05, + "loss": 0.4023, + "step": 50407 + }, + { + "epoch": 1.94, + "learning_rate": 1.7710951526032315e-05, + "loss": 0.2771, + "step": 50408 + }, + { + "epoch": 1.94, + "learning_rate": 1.7710310335983587e-05, + "loss": 0.2675, + "step": 50409 + }, + { + "epoch": 1.94, + "learning_rate": 1.7709669145934855e-05, + "loss": 0.2029, + "step": 50410 + }, + { + "epoch": 1.94, + "learning_rate": 1.7709027955886127e-05, + "loss": 0.3345, + "step": 50411 + }, + { + "epoch": 1.94, + "learning_rate": 1.7708386765837395e-05, + "loss": 0.2434, + "step": 50412 + }, + { + "epoch": 1.94, + "learning_rate": 1.7707745575788664e-05, + "loss": 0.2438, + "step": 50413 + }, + { + "epoch": 1.94, + "learning_rate": 1.7707104385739935e-05, + "loss": 0.2349, + "step": 50414 + }, + { + "epoch": 1.94, + "learning_rate": 1.7706463195691204e-05, + "loss": 0.304, + "step": 50415 + }, + { + "epoch": 1.94, + "learning_rate": 1.7705822005642472e-05, + "loss": 0.3414, + "step": 50416 + }, + { + "epoch": 1.94, + "learning_rate": 1.770518081559374e-05, + "loss": 0.3427, + "step": 50417 + }, + { + "epoch": 1.94, + "learning_rate": 1.7704539625545012e-05, + "loss": 0.3135, + "step": 50418 + }, + { + "epoch": 1.94, + "learning_rate": 1.7703898435496284e-05, + "loss": 0.3246, + "step": 50419 + }, + { + "epoch": 1.94, + "learning_rate": 1.7703257245447552e-05, + "loss": 0.2304, + "step": 50420 + }, + { + "epoch": 1.94, + "learning_rate": 1.770261605539882e-05, + "loss": 0.2417, + "step": 50421 + }, + { + "epoch": 1.94, + "learning_rate": 1.7701974865350092e-05, + "loss": 0.3308, + "step": 50422 + }, + { + "epoch": 1.94, + "learning_rate": 1.770133367530136e-05, + "loss": 0.416, + "step": 50423 + }, + { + "epoch": 1.94, + "learning_rate": 1.770069248525263e-05, + "loss": 0.3306, + "step": 50424 + }, + { + "epoch": 1.94, + "learning_rate": 1.7700051295203897e-05, + "loss": 0.381, + "step": 50425 + }, + { + "epoch": 1.94, + "learning_rate": 1.769941010515517e-05, + "loss": 0.3766, + "step": 50426 + }, + { + "epoch": 1.94, + "learning_rate": 1.769876891510644e-05, + "loss": 0.3693, + "step": 50427 + }, + { + "epoch": 1.94, + "learning_rate": 1.769812772505771e-05, + "loss": 0.3744, + "step": 50428 + }, + { + "epoch": 1.94, + "learning_rate": 1.7697486535008977e-05, + "loss": 0.3294, + "step": 50429 + }, + { + "epoch": 1.94, + "learning_rate": 1.769684534496025e-05, + "loss": 0.4212, + "step": 50430 + }, + { + "epoch": 1.94, + "learning_rate": 1.7696204154911517e-05, + "loss": 0.3041, + "step": 50431 + }, + { + "epoch": 1.94, + "learning_rate": 1.7695562964862786e-05, + "loss": 0.452, + "step": 50432 + }, + { + "epoch": 1.94, + "learning_rate": 1.7694921774814054e-05, + "loss": 0.2993, + "step": 50433 + }, + { + "epoch": 1.94, + "learning_rate": 1.7694280584765326e-05, + "loss": 0.2731, + "step": 50434 + }, + { + "epoch": 1.94, + "learning_rate": 1.7693639394716594e-05, + "loss": 0.2811, + "step": 50435 + }, + { + "epoch": 1.94, + "learning_rate": 1.7692998204667866e-05, + "loss": 0.299, + "step": 50436 + }, + { + "epoch": 1.94, + "learning_rate": 1.7692357014619134e-05, + "loss": 0.3378, + "step": 50437 + }, + { + "epoch": 1.94, + "learning_rate": 1.7691715824570402e-05, + "loss": 0.3081, + "step": 50438 + }, + { + "epoch": 1.94, + "learning_rate": 1.7691074634521674e-05, + "loss": 0.4692, + "step": 50439 + }, + { + "epoch": 1.94, + "learning_rate": 1.7690433444472942e-05, + "loss": 0.451, + "step": 50440 + }, + { + "epoch": 1.94, + "learning_rate": 1.768979225442421e-05, + "loss": 0.2929, + "step": 50441 + }, + { + "epoch": 1.94, + "learning_rate": 1.7689151064375482e-05, + "loss": 0.3587, + "step": 50442 + }, + { + "epoch": 1.94, + "learning_rate": 1.768850987432675e-05, + "loss": 0.424, + "step": 50443 + }, + { + "epoch": 1.94, + "learning_rate": 1.7687868684278022e-05, + "loss": 0.4152, + "step": 50444 + }, + { + "epoch": 1.94, + "learning_rate": 1.768722749422929e-05, + "loss": 0.2018, + "step": 50445 + }, + { + "epoch": 1.94, + "learning_rate": 1.768658630418056e-05, + "loss": 0.3009, + "step": 50446 + }, + { + "epoch": 1.94, + "learning_rate": 1.768594511413183e-05, + "loss": 0.3433, + "step": 50447 + }, + { + "epoch": 1.94, + "learning_rate": 1.76853039240831e-05, + "loss": 0.2558, + "step": 50448 + }, + { + "epoch": 1.94, + "learning_rate": 1.7684662734034367e-05, + "loss": 0.2895, + "step": 50449 + }, + { + "epoch": 1.94, + "learning_rate": 1.768402154398564e-05, + "loss": 0.5388, + "step": 50450 + }, + { + "epoch": 1.94, + "learning_rate": 1.7683380353936908e-05, + "loss": 0.2674, + "step": 50451 + }, + { + "epoch": 1.94, + "learning_rate": 1.7682739163888176e-05, + "loss": 0.2402, + "step": 50452 + }, + { + "epoch": 1.94, + "learning_rate": 1.7682097973839448e-05, + "loss": 0.4759, + "step": 50453 + }, + { + "epoch": 1.94, + "learning_rate": 1.7681456783790716e-05, + "loss": 0.2688, + "step": 50454 + }, + { + "epoch": 1.94, + "learning_rate": 1.7680815593741988e-05, + "loss": 0.26, + "step": 50455 + }, + { + "epoch": 1.94, + "learning_rate": 1.7680174403693256e-05, + "loss": 0.2764, + "step": 50456 + }, + { + "epoch": 1.94, + "learning_rate": 1.7679533213644524e-05, + "loss": 0.3026, + "step": 50457 + }, + { + "epoch": 1.94, + "learning_rate": 1.7678892023595796e-05, + "loss": 0.3333, + "step": 50458 + }, + { + "epoch": 1.94, + "learning_rate": 1.7678250833547064e-05, + "loss": 0.222, + "step": 50459 + }, + { + "epoch": 1.94, + "learning_rate": 1.7677609643498333e-05, + "loss": 0.3399, + "step": 50460 + }, + { + "epoch": 1.94, + "learning_rate": 1.76769684534496e-05, + "loss": 0.2411, + "step": 50461 + }, + { + "epoch": 1.94, + "learning_rate": 1.7676327263400873e-05, + "loss": 0.3133, + "step": 50462 + }, + { + "epoch": 1.94, + "learning_rate": 1.7675686073352144e-05, + "loss": 0.3509, + "step": 50463 + }, + { + "epoch": 1.94, + "learning_rate": 1.7675044883303413e-05, + "loss": 0.2518, + "step": 50464 + }, + { + "epoch": 1.94, + "learning_rate": 1.767440369325468e-05, + "loss": 0.4004, + "step": 50465 + }, + { + "epoch": 1.94, + "learning_rate": 1.7673762503205953e-05, + "loss": 0.3488, + "step": 50466 + }, + { + "epoch": 1.94, + "learning_rate": 1.767312131315722e-05, + "loss": 0.2534, + "step": 50467 + }, + { + "epoch": 1.94, + "learning_rate": 1.767248012310849e-05, + "loss": 0.296, + "step": 50468 + }, + { + "epoch": 1.94, + "learning_rate": 1.7671838933059758e-05, + "loss": 0.3223, + "step": 50469 + }, + { + "epoch": 1.94, + "learning_rate": 1.767119774301103e-05, + "loss": 0.3355, + "step": 50470 + }, + { + "epoch": 1.94, + "learning_rate": 1.76705565529623e-05, + "loss": 0.3387, + "step": 50471 + }, + { + "epoch": 1.94, + "learning_rate": 1.766991536291357e-05, + "loss": 0.3463, + "step": 50472 + }, + { + "epoch": 1.94, + "learning_rate": 1.7669274172864838e-05, + "loss": 0.3107, + "step": 50473 + }, + { + "epoch": 1.94, + "learning_rate": 1.766863298281611e-05, + "loss": 0.3612, + "step": 50474 + }, + { + "epoch": 1.94, + "learning_rate": 1.7667991792767378e-05, + "loss": 0.4439, + "step": 50475 + }, + { + "epoch": 1.94, + "learning_rate": 1.7667350602718646e-05, + "loss": 0.3535, + "step": 50476 + }, + { + "epoch": 1.94, + "learning_rate": 1.7666709412669915e-05, + "loss": 0.25, + "step": 50477 + }, + { + "epoch": 1.94, + "learning_rate": 1.7666068222621186e-05, + "loss": 0.4287, + "step": 50478 + }, + { + "epoch": 1.94, + "learning_rate": 1.7665427032572455e-05, + "loss": 0.3678, + "step": 50479 + }, + { + "epoch": 1.94, + "learning_rate": 1.7664785842523726e-05, + "loss": 0.4557, + "step": 50480 + }, + { + "epoch": 1.94, + "learning_rate": 1.7664144652474995e-05, + "loss": 0.2311, + "step": 50481 + }, + { + "epoch": 1.94, + "learning_rate": 1.7663503462426263e-05, + "loss": 0.2622, + "step": 50482 + }, + { + "epoch": 1.94, + "learning_rate": 1.7662862272377535e-05, + "loss": 0.2766, + "step": 50483 + }, + { + "epoch": 1.94, + "learning_rate": 1.7662221082328803e-05, + "loss": 0.287, + "step": 50484 + }, + { + "epoch": 1.94, + "learning_rate": 1.766157989228007e-05, + "loss": 0.3867, + "step": 50485 + }, + { + "epoch": 1.94, + "learning_rate": 1.7660938702231343e-05, + "loss": 0.3107, + "step": 50486 + }, + { + "epoch": 1.94, + "learning_rate": 1.766029751218261e-05, + "loss": 0.5135, + "step": 50487 + }, + { + "epoch": 1.94, + "learning_rate": 1.765965632213388e-05, + "loss": 0.3443, + "step": 50488 + }, + { + "epoch": 1.94, + "learning_rate": 1.765901513208515e-05, + "loss": 0.2861, + "step": 50489 + }, + { + "epoch": 1.94, + "learning_rate": 1.765837394203642e-05, + "loss": 0.311, + "step": 50490 + }, + { + "epoch": 1.94, + "learning_rate": 1.765773275198769e-05, + "loss": 0.4439, + "step": 50491 + }, + { + "epoch": 1.94, + "learning_rate": 1.765709156193896e-05, + "loss": 0.2922, + "step": 50492 + }, + { + "epoch": 1.94, + "learning_rate": 1.7656450371890228e-05, + "loss": 0.3334, + "step": 50493 + }, + { + "epoch": 1.94, + "learning_rate": 1.76558091818415e-05, + "loss": 0.2406, + "step": 50494 + }, + { + "epoch": 1.94, + "learning_rate": 1.7655167991792768e-05, + "loss": 0.308, + "step": 50495 + }, + { + "epoch": 1.94, + "learning_rate": 1.7654526801744037e-05, + "loss": 0.3098, + "step": 50496 + }, + { + "epoch": 1.94, + "learning_rate": 1.7653885611695305e-05, + "loss": 0.3826, + "step": 50497 + }, + { + "epoch": 1.94, + "learning_rate": 1.7653244421646577e-05, + "loss": 0.2112, + "step": 50498 + }, + { + "epoch": 1.94, + "learning_rate": 1.7652603231597848e-05, + "loss": 0.4163, + "step": 50499 + }, + { + "epoch": 1.94, + "learning_rate": 1.7651962041549117e-05, + "loss": 0.3701, + "step": 50500 + }, + { + "epoch": 1.94, + "learning_rate": 1.7651320851500385e-05, + "loss": 0.3504, + "step": 50501 + }, + { + "epoch": 1.94, + "learning_rate": 1.7650679661451657e-05, + "loss": 0.353, + "step": 50502 + }, + { + "epoch": 1.94, + "learning_rate": 1.7650038471402925e-05, + "loss": 0.3063, + "step": 50503 + }, + { + "epoch": 1.94, + "learning_rate": 1.7649397281354193e-05, + "loss": 0.3134, + "step": 50504 + }, + { + "epoch": 1.94, + "learning_rate": 1.764875609130546e-05, + "loss": 0.3518, + "step": 50505 + }, + { + "epoch": 1.94, + "learning_rate": 1.7648114901256733e-05, + "loss": 0.2838, + "step": 50506 + }, + { + "epoch": 1.94, + "learning_rate": 1.7647473711208005e-05, + "loss": 0.3565, + "step": 50507 + }, + { + "epoch": 1.94, + "learning_rate": 1.7646832521159273e-05, + "loss": 0.2739, + "step": 50508 + }, + { + "epoch": 1.94, + "learning_rate": 1.7646191331110542e-05, + "loss": 0.2071, + "step": 50509 + }, + { + "epoch": 1.94, + "learning_rate": 1.7645550141061813e-05, + "loss": 0.2292, + "step": 50510 + }, + { + "epoch": 1.94, + "learning_rate": 1.7644908951013082e-05, + "loss": 0.4261, + "step": 50511 + }, + { + "epoch": 1.94, + "learning_rate": 1.764426776096435e-05, + "loss": 0.3485, + "step": 50512 + }, + { + "epoch": 1.94, + "learning_rate": 1.764362657091562e-05, + "loss": 0.3709, + "step": 50513 + }, + { + "epoch": 1.94, + "learning_rate": 1.764298538086689e-05, + "loss": 0.2807, + "step": 50514 + }, + { + "epoch": 1.94, + "learning_rate": 1.764234419081816e-05, + "loss": 0.3642, + "step": 50515 + }, + { + "epoch": 1.94, + "learning_rate": 1.764170300076943e-05, + "loss": 0.3565, + "step": 50516 + }, + { + "epoch": 1.94, + "learning_rate": 1.76410618107207e-05, + "loss": 0.2325, + "step": 50517 + }, + { + "epoch": 1.94, + "learning_rate": 1.7640420620671967e-05, + "loss": 0.2844, + "step": 50518 + }, + { + "epoch": 1.94, + "learning_rate": 1.763977943062324e-05, + "loss": 0.2598, + "step": 50519 + }, + { + "epoch": 1.94, + "learning_rate": 1.7639138240574507e-05, + "loss": 0.3705, + "step": 50520 + }, + { + "epoch": 1.94, + "learning_rate": 1.7638497050525775e-05, + "loss": 0.3024, + "step": 50521 + }, + { + "epoch": 1.94, + "learning_rate": 1.7637855860477047e-05, + "loss": 0.2491, + "step": 50522 + }, + { + "epoch": 1.94, + "learning_rate": 1.7637214670428315e-05, + "loss": 0.2692, + "step": 50523 + }, + { + "epoch": 1.94, + "learning_rate": 1.7636573480379584e-05, + "loss": 0.239, + "step": 50524 + }, + { + "epoch": 1.94, + "learning_rate": 1.7635932290330855e-05, + "loss": 0.4267, + "step": 50525 + }, + { + "epoch": 1.94, + "learning_rate": 1.7635291100282124e-05, + "loss": 0.4184, + "step": 50526 + }, + { + "epoch": 1.94, + "learning_rate": 1.7634649910233395e-05, + "loss": 0.3887, + "step": 50527 + }, + { + "epoch": 1.94, + "learning_rate": 1.7634008720184664e-05, + "loss": 0.3243, + "step": 50528 + }, + { + "epoch": 1.94, + "learning_rate": 1.7633367530135932e-05, + "loss": 0.2018, + "step": 50529 + }, + { + "epoch": 1.94, + "learning_rate": 1.7632726340087204e-05, + "loss": 0.325, + "step": 50530 + }, + { + "epoch": 1.94, + "learning_rate": 1.7632085150038472e-05, + "loss": 0.3474, + "step": 50531 + }, + { + "epoch": 1.94, + "learning_rate": 1.763144395998974e-05, + "loss": 0.335, + "step": 50532 + }, + { + "epoch": 1.94, + "learning_rate": 1.763080276994101e-05, + "loss": 0.25, + "step": 50533 + }, + { + "epoch": 1.94, + "learning_rate": 1.763016157989228e-05, + "loss": 0.3714, + "step": 50534 + }, + { + "epoch": 1.94, + "learning_rate": 1.7629520389843552e-05, + "loss": 0.277, + "step": 50535 + }, + { + "epoch": 1.94, + "learning_rate": 1.762887919979482e-05, + "loss": 0.3106, + "step": 50536 + }, + { + "epoch": 1.94, + "learning_rate": 1.762823800974609e-05, + "loss": 0.2734, + "step": 50537 + }, + { + "epoch": 1.94, + "learning_rate": 1.762759681969736e-05, + "loss": 0.3554, + "step": 50538 + }, + { + "epoch": 1.94, + "learning_rate": 1.762695562964863e-05, + "loss": 0.3544, + "step": 50539 + }, + { + "epoch": 1.94, + "learning_rate": 1.7626314439599897e-05, + "loss": 0.2416, + "step": 50540 + }, + { + "epoch": 1.94, + "learning_rate": 1.7625673249551166e-05, + "loss": 0.4181, + "step": 50541 + }, + { + "epoch": 1.94, + "learning_rate": 1.7625032059502437e-05, + "loss": 0.2912, + "step": 50542 + }, + { + "epoch": 1.94, + "learning_rate": 1.762439086945371e-05, + "loss": 0.2279, + "step": 50543 + }, + { + "epoch": 1.94, + "learning_rate": 1.7623749679404977e-05, + "loss": 0.2703, + "step": 50544 + }, + { + "epoch": 1.94, + "learning_rate": 1.7623108489356246e-05, + "loss": 0.3642, + "step": 50545 + }, + { + "epoch": 1.94, + "learning_rate": 1.7622467299307517e-05, + "loss": 0.3088, + "step": 50546 + }, + { + "epoch": 1.94, + "learning_rate": 1.7621826109258786e-05, + "loss": 0.3937, + "step": 50547 + }, + { + "epoch": 1.94, + "learning_rate": 1.7621184919210054e-05, + "loss": 0.2754, + "step": 50548 + }, + { + "epoch": 1.94, + "learning_rate": 1.7620543729161322e-05, + "loss": 0.3555, + "step": 50549 + }, + { + "epoch": 1.94, + "learning_rate": 1.7619902539112594e-05, + "loss": 0.3181, + "step": 50550 + }, + { + "epoch": 1.94, + "learning_rate": 1.7619261349063866e-05, + "loss": 0.4023, + "step": 50551 + }, + { + "epoch": 1.94, + "learning_rate": 1.7618620159015134e-05, + "loss": 0.2591, + "step": 50552 + }, + { + "epoch": 1.94, + "learning_rate": 1.7617978968966402e-05, + "loss": 0.3617, + "step": 50553 + }, + { + "epoch": 1.94, + "learning_rate": 1.7617337778917674e-05, + "loss": 0.3603, + "step": 50554 + }, + { + "epoch": 1.94, + "learning_rate": 1.7616696588868942e-05, + "loss": 0.2633, + "step": 50555 + }, + { + "epoch": 1.94, + "learning_rate": 1.761605539882021e-05, + "loss": 0.3622, + "step": 50556 + }, + { + "epoch": 1.94, + "learning_rate": 1.761541420877148e-05, + "loss": 0.297, + "step": 50557 + }, + { + "epoch": 1.94, + "learning_rate": 1.761477301872275e-05, + "loss": 0.3776, + "step": 50558 + }, + { + "epoch": 1.94, + "learning_rate": 1.761413182867402e-05, + "loss": 0.2829, + "step": 50559 + }, + { + "epoch": 1.94, + "learning_rate": 1.761349063862529e-05, + "loss": 0.2503, + "step": 50560 + }, + { + "epoch": 1.94, + "learning_rate": 1.761284944857656e-05, + "loss": 0.3247, + "step": 50561 + }, + { + "epoch": 1.94, + "learning_rate": 1.7612208258527828e-05, + "loss": 0.2729, + "step": 50562 + }, + { + "epoch": 1.94, + "learning_rate": 1.76115670684791e-05, + "loss": 0.3059, + "step": 50563 + }, + { + "epoch": 1.94, + "learning_rate": 1.7610925878430368e-05, + "loss": 0.2563, + "step": 50564 + }, + { + "epoch": 1.94, + "learning_rate": 1.7610284688381636e-05, + "loss": 0.4093, + "step": 50565 + }, + { + "epoch": 1.94, + "learning_rate": 1.7609643498332908e-05, + "loss": 0.2702, + "step": 50566 + }, + { + "epoch": 1.94, + "learning_rate": 1.7609002308284176e-05, + "loss": 0.3928, + "step": 50567 + }, + { + "epoch": 1.94, + "learning_rate": 1.7608361118235444e-05, + "loss": 0.2363, + "step": 50568 + }, + { + "epoch": 1.94, + "learning_rate": 1.7607719928186716e-05, + "loss": 0.4528, + "step": 50569 + }, + { + "epoch": 1.94, + "learning_rate": 1.7607078738137984e-05, + "loss": 0.4556, + "step": 50570 + }, + { + "epoch": 1.94, + "learning_rate": 1.7606437548089256e-05, + "loss": 0.2325, + "step": 50571 + }, + { + "epoch": 1.94, + "learning_rate": 1.7605796358040524e-05, + "loss": 0.1854, + "step": 50572 + }, + { + "epoch": 1.94, + "learning_rate": 1.7605155167991793e-05, + "loss": 0.2437, + "step": 50573 + }, + { + "epoch": 1.94, + "learning_rate": 1.7604513977943064e-05, + "loss": 0.3193, + "step": 50574 + }, + { + "epoch": 1.94, + "learning_rate": 1.7603872787894333e-05, + "loss": 0.279, + "step": 50575 + }, + { + "epoch": 1.94, + "learning_rate": 1.76032315978456e-05, + "loss": 0.3257, + "step": 50576 + }, + { + "epoch": 1.94, + "learning_rate": 1.760259040779687e-05, + "loss": 0.3006, + "step": 50577 + }, + { + "epoch": 1.94, + "learning_rate": 1.760194921774814e-05, + "loss": 0.3029, + "step": 50578 + }, + { + "epoch": 1.94, + "learning_rate": 1.7601308027699413e-05, + "loss": 0.3476, + "step": 50579 + }, + { + "epoch": 1.94, + "learning_rate": 1.760066683765068e-05, + "loss": 0.3683, + "step": 50580 + }, + { + "epoch": 1.94, + "learning_rate": 1.760002564760195e-05, + "loss": 0.3068, + "step": 50581 + }, + { + "epoch": 1.94, + "learning_rate": 1.759938445755322e-05, + "loss": 0.2756, + "step": 50582 + }, + { + "epoch": 1.94, + "learning_rate": 1.759874326750449e-05, + "loss": 0.2725, + "step": 50583 + }, + { + "epoch": 1.94, + "learning_rate": 1.7598102077455758e-05, + "loss": 0.4696, + "step": 50584 + }, + { + "epoch": 1.94, + "learning_rate": 1.7597460887407026e-05, + "loss": 0.421, + "step": 50585 + }, + { + "epoch": 1.94, + "learning_rate": 1.7596819697358298e-05, + "loss": 0.2961, + "step": 50586 + }, + { + "epoch": 1.94, + "learning_rate": 1.759617850730957e-05, + "loss": 0.2343, + "step": 50587 + }, + { + "epoch": 1.94, + "learning_rate": 1.7595537317260838e-05, + "loss": 0.1829, + "step": 50588 + }, + { + "epoch": 1.94, + "learning_rate": 1.7594896127212106e-05, + "loss": 0.3076, + "step": 50589 + }, + { + "epoch": 1.95, + "learning_rate": 1.7594254937163378e-05, + "loss": 0.509, + "step": 50590 + }, + { + "epoch": 1.95, + "learning_rate": 1.7593613747114646e-05, + "loss": 0.3591, + "step": 50591 + }, + { + "epoch": 1.95, + "learning_rate": 1.7592972557065915e-05, + "loss": 0.271, + "step": 50592 + }, + { + "epoch": 1.95, + "learning_rate": 1.7592331367017183e-05, + "loss": 0.3566, + "step": 50593 + }, + { + "epoch": 1.95, + "learning_rate": 1.7591690176968455e-05, + "loss": 0.353, + "step": 50594 + }, + { + "epoch": 1.95, + "learning_rate": 1.7591048986919723e-05, + "loss": 0.2926, + "step": 50595 + }, + { + "epoch": 1.95, + "learning_rate": 1.7590407796870995e-05, + "loss": 0.2155, + "step": 50596 + }, + { + "epoch": 1.95, + "learning_rate": 1.7589766606822263e-05, + "loss": 0.2889, + "step": 50597 + }, + { + "epoch": 1.95, + "learning_rate": 1.758912541677353e-05, + "loss": 0.3624, + "step": 50598 + }, + { + "epoch": 1.95, + "learning_rate": 1.7588484226724803e-05, + "loss": 0.2056, + "step": 50599 + }, + { + "epoch": 1.95, + "learning_rate": 1.758784303667607e-05, + "loss": 0.4405, + "step": 50600 + }, + { + "epoch": 1.95, + "learning_rate": 1.758720184662734e-05, + "loss": 0.3017, + "step": 50601 + }, + { + "epoch": 1.95, + "learning_rate": 1.758656065657861e-05, + "loss": 0.367, + "step": 50602 + }, + { + "epoch": 1.95, + "learning_rate": 1.758591946652988e-05, + "loss": 0.275, + "step": 50603 + }, + { + "epoch": 1.95, + "learning_rate": 1.7585278276481148e-05, + "loss": 0.1923, + "step": 50604 + }, + { + "epoch": 1.95, + "learning_rate": 1.758463708643242e-05, + "loss": 0.3472, + "step": 50605 + }, + { + "epoch": 1.95, + "learning_rate": 1.7583995896383688e-05, + "loss": 0.3184, + "step": 50606 + }, + { + "epoch": 1.95, + "learning_rate": 1.758335470633496e-05, + "loss": 0.4039, + "step": 50607 + }, + { + "epoch": 1.95, + "learning_rate": 1.7582713516286228e-05, + "loss": 0.3449, + "step": 50608 + }, + { + "epoch": 1.95, + "learning_rate": 1.7582072326237497e-05, + "loss": 0.3152, + "step": 50609 + }, + { + "epoch": 1.95, + "learning_rate": 1.7581431136188768e-05, + "loss": 0.3776, + "step": 50610 + }, + { + "epoch": 1.95, + "learning_rate": 1.7580789946140037e-05, + "loss": 0.3042, + "step": 50611 + }, + { + "epoch": 1.95, + "learning_rate": 1.7580148756091305e-05, + "loss": 0.3369, + "step": 50612 + }, + { + "epoch": 1.95, + "learning_rate": 1.7579507566042573e-05, + "loss": 0.3447, + "step": 50613 + }, + { + "epoch": 1.95, + "learning_rate": 1.7578866375993845e-05, + "loss": 0.2786, + "step": 50614 + }, + { + "epoch": 1.95, + "learning_rate": 1.7578225185945117e-05, + "loss": 0.3047, + "step": 50615 + }, + { + "epoch": 1.95, + "learning_rate": 1.7577583995896385e-05, + "loss": 0.2788, + "step": 50616 + }, + { + "epoch": 1.95, + "learning_rate": 1.7576942805847653e-05, + "loss": 0.3582, + "step": 50617 + }, + { + "epoch": 1.95, + "learning_rate": 1.7576301615798925e-05, + "loss": 0.3991, + "step": 50618 + }, + { + "epoch": 1.95, + "learning_rate": 1.7575660425750193e-05, + "loss": 0.3693, + "step": 50619 + }, + { + "epoch": 1.95, + "learning_rate": 1.7575019235701462e-05, + "loss": 0.2955, + "step": 50620 + }, + { + "epoch": 1.95, + "learning_rate": 1.757437804565273e-05, + "loss": 0.1694, + "step": 50621 + }, + { + "epoch": 1.95, + "learning_rate": 1.7573736855604002e-05, + "loss": 0.2865, + "step": 50622 + }, + { + "epoch": 1.95, + "learning_rate": 1.7573095665555273e-05, + "loss": 0.2839, + "step": 50623 + }, + { + "epoch": 1.95, + "learning_rate": 1.7572454475506542e-05, + "loss": 0.3099, + "step": 50624 + }, + { + "epoch": 1.95, + "learning_rate": 1.757181328545781e-05, + "loss": 0.3571, + "step": 50625 + }, + { + "epoch": 1.95, + "learning_rate": 1.7571172095409082e-05, + "loss": 0.3606, + "step": 50626 + }, + { + "epoch": 1.95, + "learning_rate": 1.757053090536035e-05, + "loss": 0.328, + "step": 50627 + }, + { + "epoch": 1.95, + "learning_rate": 1.756988971531162e-05, + "loss": 0.2411, + "step": 50628 + }, + { + "epoch": 1.95, + "learning_rate": 1.7569248525262887e-05, + "loss": 0.3589, + "step": 50629 + }, + { + "epoch": 1.95, + "learning_rate": 1.756860733521416e-05, + "loss": 0.3985, + "step": 50630 + }, + { + "epoch": 1.95, + "learning_rate": 1.7567966145165427e-05, + "loss": 0.2745, + "step": 50631 + }, + { + "epoch": 1.95, + "learning_rate": 1.75673249551167e-05, + "loss": 0.3788, + "step": 50632 + }, + { + "epoch": 1.95, + "learning_rate": 1.7566683765067967e-05, + "loss": 0.3258, + "step": 50633 + }, + { + "epoch": 1.95, + "learning_rate": 1.756604257501924e-05, + "loss": 0.2594, + "step": 50634 + }, + { + "epoch": 1.95, + "learning_rate": 1.7565401384970507e-05, + "loss": 0.237, + "step": 50635 + }, + { + "epoch": 1.95, + "learning_rate": 1.7564760194921775e-05, + "loss": 0.5114, + "step": 50636 + }, + { + "epoch": 1.95, + "learning_rate": 1.7564119004873044e-05, + "loss": 0.369, + "step": 50637 + }, + { + "epoch": 1.95, + "learning_rate": 1.7563477814824315e-05, + "loss": 0.2578, + "step": 50638 + }, + { + "epoch": 1.95, + "learning_rate": 1.7562836624775584e-05, + "loss": 0.2227, + "step": 50639 + }, + { + "epoch": 1.95, + "learning_rate": 1.7562195434726855e-05, + "loss": 0.289, + "step": 50640 + }, + { + "epoch": 1.95, + "learning_rate": 1.7561554244678124e-05, + "loss": 0.2906, + "step": 50641 + }, + { + "epoch": 1.95, + "learning_rate": 1.7560913054629392e-05, + "loss": 0.4803, + "step": 50642 + }, + { + "epoch": 1.95, + "learning_rate": 1.7560271864580664e-05, + "loss": 0.2235, + "step": 50643 + }, + { + "epoch": 1.95, + "learning_rate": 1.7559630674531932e-05, + "loss": 0.2047, + "step": 50644 + }, + { + "epoch": 1.95, + "learning_rate": 1.75589894844832e-05, + "loss": 0.2638, + "step": 50645 + }, + { + "epoch": 1.95, + "learning_rate": 1.7558348294434472e-05, + "loss": 0.2267, + "step": 50646 + }, + { + "epoch": 1.95, + "learning_rate": 1.755770710438574e-05, + "loss": 0.3687, + "step": 50647 + }, + { + "epoch": 1.95, + "learning_rate": 1.755706591433701e-05, + "loss": 0.2689, + "step": 50648 + }, + { + "epoch": 1.95, + "learning_rate": 1.755642472428828e-05, + "loss": 0.3064, + "step": 50649 + }, + { + "epoch": 1.95, + "learning_rate": 1.755578353423955e-05, + "loss": 0.3004, + "step": 50650 + }, + { + "epoch": 1.95, + "learning_rate": 1.755514234419082e-05, + "loss": 0.3565, + "step": 50651 + }, + { + "epoch": 1.95, + "learning_rate": 1.755450115414209e-05, + "loss": 0.3324, + "step": 50652 + }, + { + "epoch": 1.95, + "learning_rate": 1.7553859964093357e-05, + "loss": 0.3581, + "step": 50653 + }, + { + "epoch": 1.95, + "learning_rate": 1.755321877404463e-05, + "loss": 0.3093, + "step": 50654 + }, + { + "epoch": 1.95, + "learning_rate": 1.7552577583995897e-05, + "loss": 0.2277, + "step": 50655 + }, + { + "epoch": 1.95, + "learning_rate": 1.7551936393947166e-05, + "loss": 0.3363, + "step": 50656 + }, + { + "epoch": 1.95, + "learning_rate": 1.7551295203898434e-05, + "loss": 0.352, + "step": 50657 + }, + { + "epoch": 1.95, + "learning_rate": 1.7550654013849706e-05, + "loss": 0.2841, + "step": 50658 + }, + { + "epoch": 1.95, + "learning_rate": 1.7550012823800977e-05, + "loss": 0.2006, + "step": 50659 + }, + { + "epoch": 1.95, + "learning_rate": 1.7549371633752246e-05, + "loss": 0.282, + "step": 50660 + }, + { + "epoch": 1.95, + "learning_rate": 1.7548730443703514e-05, + "loss": 0.2668, + "step": 50661 + }, + { + "epoch": 1.95, + "learning_rate": 1.7548089253654786e-05, + "loss": 0.2911, + "step": 50662 + }, + { + "epoch": 1.95, + "learning_rate": 1.7547448063606054e-05, + "loss": 0.216, + "step": 50663 + }, + { + "epoch": 1.95, + "learning_rate": 1.7546806873557322e-05, + "loss": 0.3486, + "step": 50664 + }, + { + "epoch": 1.95, + "learning_rate": 1.754616568350859e-05, + "loss": 0.3116, + "step": 50665 + }, + { + "epoch": 1.95, + "learning_rate": 1.7545524493459862e-05, + "loss": 0.3483, + "step": 50666 + }, + { + "epoch": 1.95, + "learning_rate": 1.7544883303411134e-05, + "loss": 0.2616, + "step": 50667 + }, + { + "epoch": 1.95, + "learning_rate": 1.7544242113362402e-05, + "loss": 0.261, + "step": 50668 + }, + { + "epoch": 1.95, + "learning_rate": 1.754360092331367e-05, + "loss": 0.3128, + "step": 50669 + }, + { + "epoch": 1.95, + "learning_rate": 1.7542959733264943e-05, + "loss": 0.3219, + "step": 50670 + }, + { + "epoch": 1.95, + "learning_rate": 1.754231854321621e-05, + "loss": 0.3565, + "step": 50671 + }, + { + "epoch": 1.95, + "learning_rate": 1.754167735316748e-05, + "loss": 0.4569, + "step": 50672 + }, + { + "epoch": 1.95, + "learning_rate": 1.7541036163118748e-05, + "loss": 0.4992, + "step": 50673 + }, + { + "epoch": 1.95, + "learning_rate": 1.754039497307002e-05, + "loss": 0.3616, + "step": 50674 + }, + { + "epoch": 1.95, + "learning_rate": 1.7539753783021288e-05, + "loss": 0.1931, + "step": 50675 + }, + { + "epoch": 1.95, + "learning_rate": 1.753911259297256e-05, + "loss": 0.2936, + "step": 50676 + }, + { + "epoch": 1.95, + "learning_rate": 1.7538471402923828e-05, + "loss": 0.2414, + "step": 50677 + }, + { + "epoch": 1.95, + "learning_rate": 1.75378302128751e-05, + "loss": 0.2851, + "step": 50678 + }, + { + "epoch": 1.95, + "learning_rate": 1.7537189022826368e-05, + "loss": 0.269, + "step": 50679 + }, + { + "epoch": 1.95, + "learning_rate": 1.7536547832777636e-05, + "loss": 0.2246, + "step": 50680 + }, + { + "epoch": 1.95, + "learning_rate": 1.7535906642728904e-05, + "loss": 0.363, + "step": 50681 + }, + { + "epoch": 1.95, + "learning_rate": 1.7535265452680176e-05, + "loss": 0.3258, + "step": 50682 + }, + { + "epoch": 1.95, + "learning_rate": 1.7534624262631444e-05, + "loss": 0.4368, + "step": 50683 + }, + { + "epoch": 1.95, + "learning_rate": 1.7533983072582713e-05, + "loss": 0.4388, + "step": 50684 + }, + { + "epoch": 1.95, + "learning_rate": 1.7533341882533984e-05, + "loss": 0.3131, + "step": 50685 + }, + { + "epoch": 1.95, + "learning_rate": 1.7532700692485253e-05, + "loss": 0.3129, + "step": 50686 + }, + { + "epoch": 1.95, + "learning_rate": 1.7532059502436524e-05, + "loss": 0.2887, + "step": 50687 + }, + { + "epoch": 1.95, + "learning_rate": 1.7531418312387793e-05, + "loss": 0.3262, + "step": 50688 + }, + { + "epoch": 1.95, + "learning_rate": 1.753077712233906e-05, + "loss": 0.2151, + "step": 50689 + }, + { + "epoch": 1.95, + "learning_rate": 1.7530135932290333e-05, + "loss": 0.3059, + "step": 50690 + }, + { + "epoch": 1.95, + "learning_rate": 1.75294947422416e-05, + "loss": 0.2901, + "step": 50691 + }, + { + "epoch": 1.95, + "learning_rate": 1.752885355219287e-05, + "loss": 0.3486, + "step": 50692 + }, + { + "epoch": 1.95, + "learning_rate": 1.7528212362144138e-05, + "loss": 0.3031, + "step": 50693 + }, + { + "epoch": 1.95, + "learning_rate": 1.752757117209541e-05, + "loss": 0.2912, + "step": 50694 + }, + { + "epoch": 1.95, + "learning_rate": 1.752692998204668e-05, + "loss": 0.2626, + "step": 50695 + }, + { + "epoch": 1.95, + "learning_rate": 1.752628879199795e-05, + "loss": 0.3167, + "step": 50696 + }, + { + "epoch": 1.95, + "learning_rate": 1.7525647601949218e-05, + "loss": 0.2923, + "step": 50697 + }, + { + "epoch": 1.95, + "learning_rate": 1.752500641190049e-05, + "loss": 0.2287, + "step": 50698 + }, + { + "epoch": 1.95, + "learning_rate": 1.7524365221851758e-05, + "loss": 0.2086, + "step": 50699 + }, + { + "epoch": 1.95, + "learning_rate": 1.7523724031803026e-05, + "loss": 0.3826, + "step": 50700 + }, + { + "epoch": 1.95, + "learning_rate": 1.7523082841754295e-05, + "loss": 0.26, + "step": 50701 + }, + { + "epoch": 1.95, + "learning_rate": 1.7522441651705566e-05, + "loss": 0.3139, + "step": 50702 + }, + { + "epoch": 1.95, + "learning_rate": 1.7521800461656838e-05, + "loss": 0.4031, + "step": 50703 + }, + { + "epoch": 1.95, + "learning_rate": 1.7521159271608106e-05, + "loss": 0.3004, + "step": 50704 + }, + { + "epoch": 1.95, + "learning_rate": 1.7520518081559375e-05, + "loss": 0.1764, + "step": 50705 + }, + { + "epoch": 1.95, + "learning_rate": 1.7519876891510646e-05, + "loss": 0.3654, + "step": 50706 + }, + { + "epoch": 1.95, + "learning_rate": 1.7519235701461915e-05, + "loss": 0.3175, + "step": 50707 + }, + { + "epoch": 1.95, + "learning_rate": 1.7518594511413183e-05, + "loss": 0.2387, + "step": 50708 + }, + { + "epoch": 1.95, + "learning_rate": 1.751795332136445e-05, + "loss": 0.2892, + "step": 50709 + }, + { + "epoch": 1.95, + "learning_rate": 1.7517312131315723e-05, + "loss": 0.2126, + "step": 50710 + }, + { + "epoch": 1.95, + "learning_rate": 1.751667094126699e-05, + "loss": 0.3338, + "step": 50711 + }, + { + "epoch": 1.95, + "learning_rate": 1.7516029751218263e-05, + "loss": 0.2921, + "step": 50712 + }, + { + "epoch": 1.95, + "learning_rate": 1.751538856116953e-05, + "loss": 0.3477, + "step": 50713 + }, + { + "epoch": 1.95, + "learning_rate": 1.7514747371120803e-05, + "loss": 0.28, + "step": 50714 + }, + { + "epoch": 1.95, + "learning_rate": 1.751410618107207e-05, + "loss": 0.287, + "step": 50715 + }, + { + "epoch": 1.95, + "learning_rate": 1.751346499102334e-05, + "loss": 0.2158, + "step": 50716 + }, + { + "epoch": 1.95, + "learning_rate": 1.7512823800974608e-05, + "loss": 0.2912, + "step": 50717 + }, + { + "epoch": 1.95, + "learning_rate": 1.751218261092588e-05, + "loss": 0.2879, + "step": 50718 + }, + { + "epoch": 1.95, + "learning_rate": 1.7511541420877148e-05, + "loss": 0.2471, + "step": 50719 + }, + { + "epoch": 1.95, + "learning_rate": 1.7510900230828417e-05, + "loss": 0.2814, + "step": 50720 + }, + { + "epoch": 1.95, + "learning_rate": 1.7510259040779688e-05, + "loss": 0.4186, + "step": 50721 + }, + { + "epoch": 1.95, + "learning_rate": 1.7509617850730957e-05, + "loss": 0.1585, + "step": 50722 + }, + { + "epoch": 1.95, + "learning_rate": 1.750897666068223e-05, + "loss": 0.3545, + "step": 50723 + }, + { + "epoch": 1.95, + "learning_rate": 1.7508335470633497e-05, + "loss": 0.3833, + "step": 50724 + }, + { + "epoch": 1.95, + "learning_rate": 1.7507694280584765e-05, + "loss": 0.3593, + "step": 50725 + }, + { + "epoch": 1.95, + "learning_rate": 1.7507053090536037e-05, + "loss": 0.2879, + "step": 50726 + }, + { + "epoch": 1.95, + "learning_rate": 1.7506411900487305e-05, + "loss": 0.2884, + "step": 50727 + }, + { + "epoch": 1.95, + "learning_rate": 1.7505770710438573e-05, + "loss": 0.4615, + "step": 50728 + }, + { + "epoch": 1.95, + "learning_rate": 1.7505129520389845e-05, + "loss": 0.3787, + "step": 50729 + }, + { + "epoch": 1.95, + "learning_rate": 1.7504488330341113e-05, + "loss": 0.3356, + "step": 50730 + }, + { + "epoch": 1.95, + "learning_rate": 1.7503847140292385e-05, + "loss": 0.2751, + "step": 50731 + }, + { + "epoch": 1.95, + "learning_rate": 1.7503205950243653e-05, + "loss": 0.303, + "step": 50732 + }, + { + "epoch": 1.95, + "learning_rate": 1.7502564760194922e-05, + "loss": 0.3358, + "step": 50733 + }, + { + "epoch": 1.95, + "learning_rate": 1.7501923570146193e-05, + "loss": 0.3335, + "step": 50734 + }, + { + "epoch": 1.95, + "learning_rate": 1.7501282380097462e-05, + "loss": 0.2158, + "step": 50735 + }, + { + "epoch": 1.95, + "learning_rate": 1.750064119004873e-05, + "loss": 0.1847, + "step": 50736 + }, + { + "epoch": 1.95, + "learning_rate": 1.75e-05, + "loss": 0.3247, + "step": 50737 + }, + { + "epoch": 1.95, + "learning_rate": 1.749935880995127e-05, + "loss": 0.3219, + "step": 50738 + }, + { + "epoch": 1.95, + "learning_rate": 1.7498717619902542e-05, + "loss": 0.2275, + "step": 50739 + }, + { + "epoch": 1.95, + "learning_rate": 1.749807642985381e-05, + "loss": 0.3447, + "step": 50740 + }, + { + "epoch": 1.95, + "learning_rate": 1.749743523980508e-05, + "loss": 0.3017, + "step": 50741 + }, + { + "epoch": 1.95, + "learning_rate": 1.749679404975635e-05, + "loss": 0.3616, + "step": 50742 + }, + { + "epoch": 1.95, + "learning_rate": 1.749615285970762e-05, + "loss": 0.2529, + "step": 50743 + }, + { + "epoch": 1.95, + "learning_rate": 1.7495511669658887e-05, + "loss": 0.3092, + "step": 50744 + }, + { + "epoch": 1.95, + "learning_rate": 1.7494870479610155e-05, + "loss": 0.26, + "step": 50745 + }, + { + "epoch": 1.95, + "learning_rate": 1.7494229289561427e-05, + "loss": 0.4413, + "step": 50746 + }, + { + "epoch": 1.95, + "learning_rate": 1.74935880995127e-05, + "loss": 0.3444, + "step": 50747 + }, + { + "epoch": 1.95, + "learning_rate": 1.7492946909463967e-05, + "loss": 0.6209, + "step": 50748 + }, + { + "epoch": 1.95, + "learning_rate": 1.7492305719415235e-05, + "loss": 0.3315, + "step": 50749 + }, + { + "epoch": 1.95, + "learning_rate": 1.7491664529366507e-05, + "loss": 0.194, + "step": 50750 + }, + { + "epoch": 1.95, + "learning_rate": 1.7491023339317775e-05, + "loss": 0.3586, + "step": 50751 + }, + { + "epoch": 1.95, + "learning_rate": 1.7490382149269044e-05, + "loss": 0.3652, + "step": 50752 + }, + { + "epoch": 1.95, + "learning_rate": 1.7489740959220312e-05, + "loss": 0.2547, + "step": 50753 + }, + { + "epoch": 1.95, + "learning_rate": 1.7489099769171584e-05, + "loss": 0.3412, + "step": 50754 + }, + { + "epoch": 1.95, + "learning_rate": 1.7488458579122852e-05, + "loss": 0.3266, + "step": 50755 + }, + { + "epoch": 1.95, + "learning_rate": 1.7487817389074124e-05, + "loss": 0.2675, + "step": 50756 + }, + { + "epoch": 1.95, + "learning_rate": 1.7487176199025392e-05, + "loss": 0.3417, + "step": 50757 + }, + { + "epoch": 1.95, + "learning_rate": 1.7486535008976664e-05, + "loss": 0.3215, + "step": 50758 + }, + { + "epoch": 1.95, + "learning_rate": 1.7485893818927932e-05, + "loss": 0.4163, + "step": 50759 + }, + { + "epoch": 1.95, + "learning_rate": 1.74852526288792e-05, + "loss": 0.3082, + "step": 50760 + }, + { + "epoch": 1.95, + "learning_rate": 1.748461143883047e-05, + "loss": 0.295, + "step": 50761 + }, + { + "epoch": 1.95, + "learning_rate": 1.748397024878174e-05, + "loss": 0.2387, + "step": 50762 + }, + { + "epoch": 1.95, + "learning_rate": 1.748332905873301e-05, + "loss": 0.3344, + "step": 50763 + }, + { + "epoch": 1.95, + "learning_rate": 1.7482687868684277e-05, + "loss": 0.3165, + "step": 50764 + }, + { + "epoch": 1.95, + "learning_rate": 1.748204667863555e-05, + "loss": 0.3003, + "step": 50765 + }, + { + "epoch": 1.95, + "learning_rate": 1.7481405488586817e-05, + "loss": 0.369, + "step": 50766 + }, + { + "epoch": 1.95, + "learning_rate": 1.748076429853809e-05, + "loss": 0.236, + "step": 50767 + }, + { + "epoch": 1.95, + "learning_rate": 1.7480123108489357e-05, + "loss": 0.2689, + "step": 50768 + }, + { + "epoch": 1.95, + "learning_rate": 1.7479481918440626e-05, + "loss": 0.3483, + "step": 50769 + }, + { + "epoch": 1.95, + "learning_rate": 1.7478840728391897e-05, + "loss": 0.4934, + "step": 50770 + }, + { + "epoch": 1.95, + "learning_rate": 1.7478199538343166e-05, + "loss": 0.3133, + "step": 50771 + }, + { + "epoch": 1.95, + "learning_rate": 1.7477558348294434e-05, + "loss": 0.268, + "step": 50772 + }, + { + "epoch": 1.95, + "learning_rate": 1.7476917158245702e-05, + "loss": 0.3534, + "step": 50773 + }, + { + "epoch": 1.95, + "learning_rate": 1.7476275968196974e-05, + "loss": 0.3426, + "step": 50774 + }, + { + "epoch": 1.95, + "learning_rate": 1.7475634778148246e-05, + "loss": 0.3709, + "step": 50775 + }, + { + "epoch": 1.95, + "learning_rate": 1.7474993588099514e-05, + "loss": 0.2982, + "step": 50776 + }, + { + "epoch": 1.95, + "learning_rate": 1.7474352398050782e-05, + "loss": 0.3997, + "step": 50777 + }, + { + "epoch": 1.95, + "learning_rate": 1.7473711208002054e-05, + "loss": 0.3918, + "step": 50778 + }, + { + "epoch": 1.95, + "learning_rate": 1.7473070017953322e-05, + "loss": 0.2866, + "step": 50779 + }, + { + "epoch": 1.95, + "learning_rate": 1.747242882790459e-05, + "loss": 0.2719, + "step": 50780 + }, + { + "epoch": 1.95, + "learning_rate": 1.747178763785586e-05, + "loss": 0.2896, + "step": 50781 + }, + { + "epoch": 1.95, + "learning_rate": 1.747114644780713e-05, + "loss": 0.3837, + "step": 50782 + }, + { + "epoch": 1.95, + "learning_rate": 1.7470505257758403e-05, + "loss": 0.415, + "step": 50783 + }, + { + "epoch": 1.95, + "learning_rate": 1.746986406770967e-05, + "loss": 0.2183, + "step": 50784 + }, + { + "epoch": 1.95, + "learning_rate": 1.746922287766094e-05, + "loss": 0.4003, + "step": 50785 + }, + { + "epoch": 1.95, + "learning_rate": 1.746858168761221e-05, + "loss": 0.3176, + "step": 50786 + }, + { + "epoch": 1.95, + "learning_rate": 1.746794049756348e-05, + "loss": 0.372, + "step": 50787 + }, + { + "epoch": 1.95, + "learning_rate": 1.7467299307514748e-05, + "loss": 0.3797, + "step": 50788 + }, + { + "epoch": 1.95, + "learning_rate": 1.7466658117466016e-05, + "loss": 0.4492, + "step": 50789 + }, + { + "epoch": 1.95, + "learning_rate": 1.7466016927417288e-05, + "loss": 0.3645, + "step": 50790 + }, + { + "epoch": 1.95, + "learning_rate": 1.7465375737368556e-05, + "loss": 0.2856, + "step": 50791 + }, + { + "epoch": 1.95, + "learning_rate": 1.7464734547319828e-05, + "loss": 0.2065, + "step": 50792 + }, + { + "epoch": 1.95, + "learning_rate": 1.7464093357271096e-05, + "loss": 0.2346, + "step": 50793 + }, + { + "epoch": 1.95, + "learning_rate": 1.7463452167222368e-05, + "loss": 0.3613, + "step": 50794 + }, + { + "epoch": 1.95, + "learning_rate": 1.7462810977173636e-05, + "loss": 0.2638, + "step": 50795 + }, + { + "epoch": 1.95, + "learning_rate": 1.7462169787124904e-05, + "loss": 0.3859, + "step": 50796 + }, + { + "epoch": 1.95, + "learning_rate": 1.7461528597076173e-05, + "loss": 0.3861, + "step": 50797 + }, + { + "epoch": 1.95, + "learning_rate": 1.7460887407027444e-05, + "loss": 0.2631, + "step": 50798 + }, + { + "epoch": 1.95, + "learning_rate": 1.7460246216978713e-05, + "loss": 0.2813, + "step": 50799 + }, + { + "epoch": 1.95, + "learning_rate": 1.745960502692998e-05, + "loss": 0.2975, + "step": 50800 + }, + { + "epoch": 1.95, + "learning_rate": 1.7458963836881253e-05, + "loss": 0.3384, + "step": 50801 + }, + { + "epoch": 1.95, + "learning_rate": 1.7458322646832525e-05, + "loss": 0.3776, + "step": 50802 + }, + { + "epoch": 1.95, + "learning_rate": 1.7457681456783793e-05, + "loss": 0.4127, + "step": 50803 + }, + { + "epoch": 1.95, + "learning_rate": 1.745704026673506e-05, + "loss": 0.3716, + "step": 50804 + }, + { + "epoch": 1.95, + "learning_rate": 1.745639907668633e-05, + "loss": 0.3133, + "step": 50805 + }, + { + "epoch": 1.95, + "learning_rate": 1.74557578866376e-05, + "loss": 0.3345, + "step": 50806 + }, + { + "epoch": 1.95, + "learning_rate": 1.745511669658887e-05, + "loss": 0.3757, + "step": 50807 + }, + { + "epoch": 1.95, + "learning_rate": 1.7454475506540138e-05, + "loss": 0.3764, + "step": 50808 + }, + { + "epoch": 1.95, + "learning_rate": 1.7453834316491406e-05, + "loss": 0.29, + "step": 50809 + }, + { + "epoch": 1.95, + "learning_rate": 1.7453193126442678e-05, + "loss": 0.1546, + "step": 50810 + }, + { + "epoch": 1.95, + "learning_rate": 1.745255193639395e-05, + "loss": 0.5023, + "step": 50811 + }, + { + "epoch": 1.95, + "learning_rate": 1.7451910746345218e-05, + "loss": 0.3874, + "step": 50812 + }, + { + "epoch": 1.95, + "learning_rate": 1.7451269556296486e-05, + "loss": 0.3551, + "step": 50813 + }, + { + "epoch": 1.95, + "learning_rate": 1.7450628366247758e-05, + "loss": 0.4591, + "step": 50814 + }, + { + "epoch": 1.95, + "learning_rate": 1.7449987176199026e-05, + "loss": 0.3166, + "step": 50815 + }, + { + "epoch": 1.95, + "learning_rate": 1.7449345986150295e-05, + "loss": 0.4187, + "step": 50816 + }, + { + "epoch": 1.95, + "learning_rate": 1.7448704796101563e-05, + "loss": 0.321, + "step": 50817 + }, + { + "epoch": 1.95, + "learning_rate": 1.7448063606052835e-05, + "loss": 0.3476, + "step": 50818 + }, + { + "epoch": 1.95, + "learning_rate": 1.7447422416004106e-05, + "loss": 0.2948, + "step": 50819 + }, + { + "epoch": 1.95, + "learning_rate": 1.7446781225955375e-05, + "loss": 0.3213, + "step": 50820 + }, + { + "epoch": 1.95, + "learning_rate": 1.7446140035906643e-05, + "loss": 0.2086, + "step": 50821 + }, + { + "epoch": 1.95, + "learning_rate": 1.7445498845857915e-05, + "loss": 0.5036, + "step": 50822 + }, + { + "epoch": 1.95, + "learning_rate": 1.7444857655809183e-05, + "loss": 0.339, + "step": 50823 + }, + { + "epoch": 1.95, + "learning_rate": 1.744421646576045e-05, + "loss": 0.522, + "step": 50824 + }, + { + "epoch": 1.95, + "learning_rate": 1.744357527571172e-05, + "loss": 0.2586, + "step": 50825 + }, + { + "epoch": 1.95, + "learning_rate": 1.744293408566299e-05, + "loss": 0.2827, + "step": 50826 + }, + { + "epoch": 1.95, + "learning_rate": 1.7442292895614263e-05, + "loss": 0.3158, + "step": 50827 + }, + { + "epoch": 1.95, + "learning_rate": 1.744165170556553e-05, + "loss": 0.3596, + "step": 50828 + }, + { + "epoch": 1.95, + "learning_rate": 1.74410105155168e-05, + "loss": 0.2657, + "step": 50829 + }, + { + "epoch": 1.95, + "learning_rate": 1.744036932546807e-05, + "loss": 0.3396, + "step": 50830 + }, + { + "epoch": 1.95, + "learning_rate": 1.743972813541934e-05, + "loss": 0.2075, + "step": 50831 + }, + { + "epoch": 1.95, + "learning_rate": 1.7439086945370608e-05, + "loss": 0.298, + "step": 50832 + }, + { + "epoch": 1.95, + "learning_rate": 1.7438445755321877e-05, + "loss": 0.2343, + "step": 50833 + }, + { + "epoch": 1.95, + "learning_rate": 1.7437804565273148e-05, + "loss": 0.4387, + "step": 50834 + }, + { + "epoch": 1.95, + "learning_rate": 1.7437163375224417e-05, + "loss": 0.3895, + "step": 50835 + }, + { + "epoch": 1.95, + "learning_rate": 1.743652218517569e-05, + "loss": 0.3203, + "step": 50836 + }, + { + "epoch": 1.95, + "learning_rate": 1.7435880995126957e-05, + "loss": 0.3003, + "step": 50837 + }, + { + "epoch": 1.95, + "learning_rate": 1.743523980507823e-05, + "loss": 0.3078, + "step": 50838 + }, + { + "epoch": 1.95, + "learning_rate": 1.7434598615029497e-05, + "loss": 0.4058, + "step": 50839 + }, + { + "epoch": 1.95, + "learning_rate": 1.7433957424980765e-05, + "loss": 0.2852, + "step": 50840 + }, + { + "epoch": 1.95, + "learning_rate": 1.7433316234932033e-05, + "loss": 0.2653, + "step": 50841 + }, + { + "epoch": 1.95, + "learning_rate": 1.7432675044883305e-05, + "loss": 0.1934, + "step": 50842 + }, + { + "epoch": 1.95, + "learning_rate": 1.7432033854834573e-05, + "loss": 0.3263, + "step": 50843 + }, + { + "epoch": 1.95, + "learning_rate": 1.7431392664785842e-05, + "loss": 0.3172, + "step": 50844 + }, + { + "epoch": 1.95, + "learning_rate": 1.7430751474737113e-05, + "loss": 0.3082, + "step": 50845 + }, + { + "epoch": 1.95, + "learning_rate": 1.7430110284688382e-05, + "loss": 0.3447, + "step": 50846 + }, + { + "epoch": 1.95, + "learning_rate": 1.7429469094639654e-05, + "loss": 0.3854, + "step": 50847 + }, + { + "epoch": 1.95, + "learning_rate": 1.7428827904590922e-05, + "loss": 0.4362, + "step": 50848 + }, + { + "epoch": 1.95, + "learning_rate": 1.742818671454219e-05, + "loss": 0.3492, + "step": 50849 + }, + { + "epoch": 1.96, + "learning_rate": 1.7427545524493462e-05, + "loss": 0.2323, + "step": 50850 + }, + { + "epoch": 1.96, + "learning_rate": 1.742690433444473e-05, + "loss": 0.3354, + "step": 50851 + }, + { + "epoch": 1.96, + "learning_rate": 1.7426263144396e-05, + "loss": 0.5082, + "step": 50852 + }, + { + "epoch": 1.96, + "learning_rate": 1.7425621954347267e-05, + "loss": 0.3246, + "step": 50853 + }, + { + "epoch": 1.96, + "learning_rate": 1.742498076429854e-05, + "loss": 0.2471, + "step": 50854 + }, + { + "epoch": 1.96, + "learning_rate": 1.742433957424981e-05, + "loss": 0.2803, + "step": 50855 + }, + { + "epoch": 1.96, + "learning_rate": 1.742369838420108e-05, + "loss": 0.507, + "step": 50856 + }, + { + "epoch": 1.96, + "learning_rate": 1.7423057194152347e-05, + "loss": 0.1997, + "step": 50857 + }, + { + "epoch": 1.96, + "learning_rate": 1.742241600410362e-05, + "loss": 0.2293, + "step": 50858 + }, + { + "epoch": 1.96, + "learning_rate": 1.7421774814054887e-05, + "loss": 0.2085, + "step": 50859 + }, + { + "epoch": 1.96, + "learning_rate": 1.7421133624006155e-05, + "loss": 0.2567, + "step": 50860 + }, + { + "epoch": 1.96, + "learning_rate": 1.7420492433957424e-05, + "loss": 0.3562, + "step": 50861 + }, + { + "epoch": 1.96, + "learning_rate": 1.7419851243908695e-05, + "loss": 0.3034, + "step": 50862 + }, + { + "epoch": 1.96, + "learning_rate": 1.7419210053859967e-05, + "loss": 0.2673, + "step": 50863 + }, + { + "epoch": 1.96, + "learning_rate": 1.7418568863811235e-05, + "loss": 0.2563, + "step": 50864 + }, + { + "epoch": 1.96, + "learning_rate": 1.7417927673762504e-05, + "loss": 0.2993, + "step": 50865 + }, + { + "epoch": 1.96, + "learning_rate": 1.7417286483713775e-05, + "loss": 0.388, + "step": 50866 + }, + { + "epoch": 1.96, + "learning_rate": 1.7416645293665044e-05, + "loss": 0.3075, + "step": 50867 + }, + { + "epoch": 1.96, + "learning_rate": 1.7416004103616312e-05, + "loss": 0.2631, + "step": 50868 + }, + { + "epoch": 1.96, + "learning_rate": 1.741536291356758e-05, + "loss": 0.2607, + "step": 50869 + }, + { + "epoch": 1.96, + "learning_rate": 1.7414721723518852e-05, + "loss": 0.3407, + "step": 50870 + }, + { + "epoch": 1.96, + "learning_rate": 1.741408053347012e-05, + "loss": 0.2579, + "step": 50871 + }, + { + "epoch": 1.96, + "learning_rate": 1.7413439343421392e-05, + "loss": 0.3232, + "step": 50872 + }, + { + "epoch": 1.96, + "learning_rate": 1.741279815337266e-05, + "loss": 0.384, + "step": 50873 + }, + { + "epoch": 1.96, + "learning_rate": 1.7412156963323932e-05, + "loss": 0.3451, + "step": 50874 + }, + { + "epoch": 1.96, + "learning_rate": 1.74115157732752e-05, + "loss": 0.2355, + "step": 50875 + }, + { + "epoch": 1.96, + "learning_rate": 1.741087458322647e-05, + "loss": 0.2096, + "step": 50876 + }, + { + "epoch": 1.96, + "learning_rate": 1.7410233393177737e-05, + "loss": 0.3229, + "step": 50877 + }, + { + "epoch": 1.96, + "learning_rate": 1.740959220312901e-05, + "loss": 0.4536, + "step": 50878 + }, + { + "epoch": 1.96, + "learning_rate": 1.7408951013080277e-05, + "loss": 0.2945, + "step": 50879 + }, + { + "epoch": 1.96, + "learning_rate": 1.7408309823031546e-05, + "loss": 0.1892, + "step": 50880 + }, + { + "epoch": 1.96, + "learning_rate": 1.7407668632982817e-05, + "loss": 0.327, + "step": 50881 + }, + { + "epoch": 1.96, + "learning_rate": 1.740702744293409e-05, + "loss": 0.3474, + "step": 50882 + }, + { + "epoch": 1.96, + "learning_rate": 1.7406386252885357e-05, + "loss": 0.1826, + "step": 50883 + }, + { + "epoch": 1.96, + "learning_rate": 1.7405745062836626e-05, + "loss": 0.3835, + "step": 50884 + }, + { + "epoch": 1.96, + "learning_rate": 1.7405103872787894e-05, + "loss": 0.2619, + "step": 50885 + }, + { + "epoch": 1.96, + "learning_rate": 1.7404462682739166e-05, + "loss": 0.4733, + "step": 50886 + }, + { + "epoch": 1.96, + "learning_rate": 1.7403821492690434e-05, + "loss": 0.343, + "step": 50887 + }, + { + "epoch": 1.96, + "learning_rate": 1.7403180302641702e-05, + "loss": 0.3016, + "step": 50888 + }, + { + "epoch": 1.96, + "learning_rate": 1.740253911259297e-05, + "loss": 0.498, + "step": 50889 + }, + { + "epoch": 1.96, + "learning_rate": 1.7401897922544242e-05, + "loss": 0.2589, + "step": 50890 + }, + { + "epoch": 1.96, + "learning_rate": 1.7401256732495514e-05, + "loss": 0.4056, + "step": 50891 + }, + { + "epoch": 1.96, + "learning_rate": 1.7400615542446783e-05, + "loss": 0.2373, + "step": 50892 + }, + { + "epoch": 1.96, + "learning_rate": 1.739997435239805e-05, + "loss": 0.4726, + "step": 50893 + }, + { + "epoch": 1.96, + "learning_rate": 1.7399333162349323e-05, + "loss": 0.2963, + "step": 50894 + }, + { + "epoch": 1.96, + "learning_rate": 1.739869197230059e-05, + "loss": 0.2351, + "step": 50895 + }, + { + "epoch": 1.96, + "learning_rate": 1.739805078225186e-05, + "loss": 0.3372, + "step": 50896 + }, + { + "epoch": 1.96, + "learning_rate": 1.7397409592203128e-05, + "loss": 0.2802, + "step": 50897 + }, + { + "epoch": 1.96, + "learning_rate": 1.73967684021544e-05, + "loss": 0.4462, + "step": 50898 + }, + { + "epoch": 1.96, + "learning_rate": 1.739612721210567e-05, + "loss": 0.2111, + "step": 50899 + }, + { + "epoch": 1.96, + "learning_rate": 1.739548602205694e-05, + "loss": 0.3836, + "step": 50900 + }, + { + "epoch": 1.96, + "learning_rate": 1.7394844832008208e-05, + "loss": 0.4822, + "step": 50901 + }, + { + "epoch": 1.96, + "learning_rate": 1.739420364195948e-05, + "loss": 0.3274, + "step": 50902 + }, + { + "epoch": 1.96, + "learning_rate": 1.7393562451910748e-05, + "loss": 0.3312, + "step": 50903 + }, + { + "epoch": 1.96, + "learning_rate": 1.7392921261862016e-05, + "loss": 0.352, + "step": 50904 + }, + { + "epoch": 1.96, + "learning_rate": 1.7392280071813284e-05, + "loss": 0.3343, + "step": 50905 + }, + { + "epoch": 1.96, + "learning_rate": 1.7391638881764556e-05, + "loss": 0.3287, + "step": 50906 + }, + { + "epoch": 1.96, + "learning_rate": 1.7390997691715824e-05, + "loss": 0.4475, + "step": 50907 + }, + { + "epoch": 1.96, + "learning_rate": 1.7390356501667096e-05, + "loss": 0.3653, + "step": 50908 + }, + { + "epoch": 1.96, + "learning_rate": 1.7389715311618364e-05, + "loss": 0.222, + "step": 50909 + }, + { + "epoch": 1.96, + "learning_rate": 1.7389074121569636e-05, + "loss": 0.4278, + "step": 50910 + }, + { + "epoch": 1.96, + "learning_rate": 1.7388432931520904e-05, + "loss": 0.3715, + "step": 50911 + }, + { + "epoch": 1.96, + "learning_rate": 1.7387791741472173e-05, + "loss": 0.2101, + "step": 50912 + }, + { + "epoch": 1.96, + "learning_rate": 1.738715055142344e-05, + "loss": 0.3941, + "step": 50913 + }, + { + "epoch": 1.96, + "learning_rate": 1.7386509361374713e-05, + "loss": 0.3154, + "step": 50914 + }, + { + "epoch": 1.96, + "learning_rate": 1.738586817132598e-05, + "loss": 0.3858, + "step": 50915 + }, + { + "epoch": 1.96, + "learning_rate": 1.738522698127725e-05, + "loss": 0.3361, + "step": 50916 + }, + { + "epoch": 1.96, + "learning_rate": 1.738458579122852e-05, + "loss": 0.3278, + "step": 50917 + }, + { + "epoch": 1.96, + "learning_rate": 1.7383944601179793e-05, + "loss": 0.3588, + "step": 50918 + }, + { + "epoch": 1.96, + "learning_rate": 1.738330341113106e-05, + "loss": 0.3312, + "step": 50919 + }, + { + "epoch": 1.96, + "learning_rate": 1.738266222108233e-05, + "loss": 0.3137, + "step": 50920 + }, + { + "epoch": 1.96, + "learning_rate": 1.7382021031033598e-05, + "loss": 0.2847, + "step": 50921 + }, + { + "epoch": 1.96, + "learning_rate": 1.738137984098487e-05, + "loss": 0.2811, + "step": 50922 + }, + { + "epoch": 1.96, + "learning_rate": 1.7380738650936138e-05, + "loss": 0.2779, + "step": 50923 + }, + { + "epoch": 1.96, + "learning_rate": 1.7380097460887406e-05, + "loss": 0.1866, + "step": 50924 + }, + { + "epoch": 1.96, + "learning_rate": 1.7379456270838678e-05, + "loss": 0.3796, + "step": 50925 + }, + { + "epoch": 1.96, + "learning_rate": 1.7378815080789946e-05, + "loss": 0.2902, + "step": 50926 + }, + { + "epoch": 1.96, + "learning_rate": 1.7378173890741218e-05, + "loss": 0.4, + "step": 50927 + }, + { + "epoch": 1.96, + "learning_rate": 1.7377532700692486e-05, + "loss": 0.3891, + "step": 50928 + }, + { + "epoch": 1.96, + "learning_rate": 1.7376891510643755e-05, + "loss": 0.3941, + "step": 50929 + }, + { + "epoch": 1.96, + "learning_rate": 1.7376250320595026e-05, + "loss": 0.4083, + "step": 50930 + }, + { + "epoch": 1.96, + "learning_rate": 1.7375609130546295e-05, + "loss": 0.2498, + "step": 50931 + }, + { + "epoch": 1.96, + "learning_rate": 1.7374967940497563e-05, + "loss": 0.3071, + "step": 50932 + }, + { + "epoch": 1.96, + "learning_rate": 1.737432675044883e-05, + "loss": 0.3876, + "step": 50933 + }, + { + "epoch": 1.96, + "learning_rate": 1.7373685560400103e-05, + "loss": 0.3213, + "step": 50934 + }, + { + "epoch": 1.96, + "learning_rate": 1.7373044370351375e-05, + "loss": 0.3636, + "step": 50935 + }, + { + "epoch": 1.96, + "learning_rate": 1.7372403180302643e-05, + "loss": 0.5031, + "step": 50936 + }, + { + "epoch": 1.96, + "learning_rate": 1.737176199025391e-05, + "loss": 0.2869, + "step": 50937 + }, + { + "epoch": 1.96, + "learning_rate": 1.7371120800205183e-05, + "loss": 0.285, + "step": 50938 + }, + { + "epoch": 1.96, + "learning_rate": 1.737047961015645e-05, + "loss": 0.2305, + "step": 50939 + }, + { + "epoch": 1.96, + "learning_rate": 1.736983842010772e-05, + "loss": 0.392, + "step": 50940 + }, + { + "epoch": 1.96, + "learning_rate": 1.7369197230058988e-05, + "loss": 0.3786, + "step": 50941 + }, + { + "epoch": 1.96, + "learning_rate": 1.736855604001026e-05, + "loss": 0.3664, + "step": 50942 + }, + { + "epoch": 1.96, + "learning_rate": 1.736791484996153e-05, + "loss": 0.3195, + "step": 50943 + }, + { + "epoch": 1.96, + "learning_rate": 1.73672736599128e-05, + "loss": 0.3094, + "step": 50944 + }, + { + "epoch": 1.96, + "learning_rate": 1.7366632469864068e-05, + "loss": 0.2953, + "step": 50945 + }, + { + "epoch": 1.96, + "learning_rate": 1.736599127981534e-05, + "loss": 0.3503, + "step": 50946 + }, + { + "epoch": 1.96, + "learning_rate": 1.736535008976661e-05, + "loss": 0.2477, + "step": 50947 + }, + { + "epoch": 1.96, + "learning_rate": 1.7364708899717877e-05, + "loss": 0.2573, + "step": 50948 + }, + { + "epoch": 1.96, + "learning_rate": 1.7364067709669145e-05, + "loss": 0.1946, + "step": 50949 + }, + { + "epoch": 1.96, + "learning_rate": 1.7363426519620417e-05, + "loss": 0.2368, + "step": 50950 + }, + { + "epoch": 1.96, + "learning_rate": 1.7362785329571685e-05, + "loss": 0.325, + "step": 50951 + }, + { + "epoch": 1.96, + "learning_rate": 1.7362144139522957e-05, + "loss": 0.3731, + "step": 50952 + }, + { + "epoch": 1.96, + "learning_rate": 1.7361502949474225e-05, + "loss": 0.3476, + "step": 50953 + }, + { + "epoch": 1.96, + "learning_rate": 1.7360861759425497e-05, + "loss": 0.272, + "step": 50954 + }, + { + "epoch": 1.96, + "learning_rate": 1.7360220569376765e-05, + "loss": 0.2446, + "step": 50955 + }, + { + "epoch": 1.96, + "learning_rate": 1.7359579379328033e-05, + "loss": 0.2946, + "step": 50956 + }, + { + "epoch": 1.96, + "learning_rate": 1.7358938189279302e-05, + "loss": 0.465, + "step": 50957 + }, + { + "epoch": 1.96, + "learning_rate": 1.7358296999230574e-05, + "loss": 0.2901, + "step": 50958 + }, + { + "epoch": 1.96, + "learning_rate": 1.7357655809181842e-05, + "loss": 0.4885, + "step": 50959 + }, + { + "epoch": 1.96, + "learning_rate": 1.735701461913311e-05, + "loss": 0.2398, + "step": 50960 + }, + { + "epoch": 1.96, + "learning_rate": 1.7356373429084382e-05, + "loss": 0.4282, + "step": 50961 + }, + { + "epoch": 1.96, + "learning_rate": 1.7355732239035654e-05, + "loss": 0.2373, + "step": 50962 + }, + { + "epoch": 1.96, + "learning_rate": 1.7355091048986922e-05, + "loss": 0.2326, + "step": 50963 + }, + { + "epoch": 1.96, + "learning_rate": 1.735444985893819e-05, + "loss": 0.3205, + "step": 50964 + }, + { + "epoch": 1.96, + "learning_rate": 1.735380866888946e-05, + "loss": 0.2401, + "step": 50965 + }, + { + "epoch": 1.96, + "learning_rate": 1.735316747884073e-05, + "loss": 0.2056, + "step": 50966 + }, + { + "epoch": 1.96, + "learning_rate": 1.7352526288792e-05, + "loss": 0.179, + "step": 50967 + }, + { + "epoch": 1.96, + "learning_rate": 1.7351885098743267e-05, + "loss": 0.1993, + "step": 50968 + }, + { + "epoch": 1.96, + "learning_rate": 1.7351243908694535e-05, + "loss": 0.3548, + "step": 50969 + }, + { + "epoch": 1.96, + "learning_rate": 1.7350602718645807e-05, + "loss": 0.2899, + "step": 50970 + }, + { + "epoch": 1.96, + "learning_rate": 1.734996152859708e-05, + "loss": 0.2985, + "step": 50971 + }, + { + "epoch": 1.96, + "learning_rate": 1.7349320338548347e-05, + "loss": 0.2401, + "step": 50972 + }, + { + "epoch": 1.96, + "learning_rate": 1.7348679148499615e-05, + "loss": 0.3277, + "step": 50973 + }, + { + "epoch": 1.96, + "learning_rate": 1.7348037958450887e-05, + "loss": 0.3367, + "step": 50974 + }, + { + "epoch": 1.96, + "learning_rate": 1.7347396768402155e-05, + "loss": 0.2209, + "step": 50975 + }, + { + "epoch": 1.96, + "learning_rate": 1.7346755578353424e-05, + "loss": 0.4017, + "step": 50976 + }, + { + "epoch": 1.96, + "learning_rate": 1.7346114388304692e-05, + "loss": 0.2027, + "step": 50977 + }, + { + "epoch": 1.96, + "learning_rate": 1.7345473198255964e-05, + "loss": 0.2972, + "step": 50978 + }, + { + "epoch": 1.96, + "learning_rate": 1.7344832008207236e-05, + "loss": 0.4031, + "step": 50979 + }, + { + "epoch": 1.96, + "learning_rate": 1.7344190818158504e-05, + "loss": 0.3999, + "step": 50980 + }, + { + "epoch": 1.96, + "learning_rate": 1.7343549628109772e-05, + "loss": 0.2821, + "step": 50981 + }, + { + "epoch": 1.96, + "learning_rate": 1.7342908438061044e-05, + "loss": 0.445, + "step": 50982 + }, + { + "epoch": 1.96, + "learning_rate": 1.7342267248012312e-05, + "loss": 0.2993, + "step": 50983 + }, + { + "epoch": 1.96, + "learning_rate": 1.734162605796358e-05, + "loss": 0.3278, + "step": 50984 + }, + { + "epoch": 1.96, + "learning_rate": 1.734098486791485e-05, + "loss": 0.2601, + "step": 50985 + }, + { + "epoch": 1.96, + "learning_rate": 1.734034367786612e-05, + "loss": 0.3896, + "step": 50986 + }, + { + "epoch": 1.96, + "learning_rate": 1.733970248781739e-05, + "loss": 0.3664, + "step": 50987 + }, + { + "epoch": 1.96, + "learning_rate": 1.733906129776866e-05, + "loss": 0.4191, + "step": 50988 + }, + { + "epoch": 1.96, + "learning_rate": 1.733842010771993e-05, + "loss": 0.41, + "step": 50989 + }, + { + "epoch": 1.96, + "learning_rate": 1.73377789176712e-05, + "loss": 0.3166, + "step": 50990 + }, + { + "epoch": 1.96, + "learning_rate": 1.733713772762247e-05, + "loss": 0.2751, + "step": 50991 + }, + { + "epoch": 1.96, + "learning_rate": 1.7336496537573737e-05, + "loss": 0.4204, + "step": 50992 + }, + { + "epoch": 1.96, + "learning_rate": 1.7335855347525006e-05, + "loss": 0.1908, + "step": 50993 + }, + { + "epoch": 1.96, + "learning_rate": 1.7335214157476277e-05, + "loss": 0.2574, + "step": 50994 + }, + { + "epoch": 1.96, + "learning_rate": 1.7334572967427546e-05, + "loss": 0.307, + "step": 50995 + }, + { + "epoch": 1.96, + "learning_rate": 1.7333931777378814e-05, + "loss": 0.2657, + "step": 50996 + }, + { + "epoch": 1.96, + "learning_rate": 1.7333290587330086e-05, + "loss": 0.2366, + "step": 50997 + }, + { + "epoch": 1.96, + "learning_rate": 1.7332649397281357e-05, + "loss": 0.3326, + "step": 50998 + }, + { + "epoch": 1.96, + "learning_rate": 1.7332008207232626e-05, + "loss": 0.3484, + "step": 50999 + }, + { + "epoch": 1.96, + "learning_rate": 1.7331367017183894e-05, + "loss": 0.2957, + "step": 51000 + }, + { + "epoch": 1.96, + "learning_rate": 1.7330725827135162e-05, + "loss": 0.4307, + "step": 51001 + }, + { + "epoch": 1.96, + "learning_rate": 1.7330084637086434e-05, + "loss": 0.349, + "step": 51002 + }, + { + "epoch": 1.96, + "learning_rate": 1.7329443447037702e-05, + "loss": 0.3, + "step": 51003 + }, + { + "epoch": 1.96, + "learning_rate": 1.732880225698897e-05, + "loss": 0.2804, + "step": 51004 + }, + { + "epoch": 1.96, + "learning_rate": 1.732816106694024e-05, + "loss": 0.2277, + "step": 51005 + }, + { + "epoch": 1.96, + "learning_rate": 1.7327519876891514e-05, + "loss": 0.4345, + "step": 51006 + }, + { + "epoch": 1.96, + "learning_rate": 1.7326878686842783e-05, + "loss": 0.3389, + "step": 51007 + }, + { + "epoch": 1.96, + "learning_rate": 1.732623749679405e-05, + "loss": 0.3647, + "step": 51008 + }, + { + "epoch": 1.96, + "learning_rate": 1.732559630674532e-05, + "loss": 0.3619, + "step": 51009 + }, + { + "epoch": 1.96, + "learning_rate": 1.732495511669659e-05, + "loss": 0.2784, + "step": 51010 + }, + { + "epoch": 1.96, + "learning_rate": 1.732431392664786e-05, + "loss": 0.2385, + "step": 51011 + }, + { + "epoch": 1.96, + "learning_rate": 1.7323672736599128e-05, + "loss": 0.2273, + "step": 51012 + }, + { + "epoch": 1.96, + "learning_rate": 1.7323031546550396e-05, + "loss": 0.2774, + "step": 51013 + }, + { + "epoch": 1.96, + "learning_rate": 1.7322390356501668e-05, + "loss": 0.2249, + "step": 51014 + }, + { + "epoch": 1.96, + "learning_rate": 1.732174916645294e-05, + "loss": 0.2409, + "step": 51015 + }, + { + "epoch": 1.96, + "learning_rate": 1.7321107976404208e-05, + "loss": 0.3844, + "step": 51016 + }, + { + "epoch": 1.96, + "learning_rate": 1.7320466786355476e-05, + "loss": 0.1927, + "step": 51017 + }, + { + "epoch": 1.96, + "learning_rate": 1.7319825596306748e-05, + "loss": 0.3179, + "step": 51018 + }, + { + "epoch": 1.96, + "learning_rate": 1.7319184406258016e-05, + "loss": 0.2913, + "step": 51019 + }, + { + "epoch": 1.96, + "learning_rate": 1.7318543216209284e-05, + "loss": 0.2526, + "step": 51020 + }, + { + "epoch": 1.96, + "learning_rate": 1.7317902026160553e-05, + "loss": 0.3144, + "step": 51021 + }, + { + "epoch": 1.96, + "learning_rate": 1.7317260836111824e-05, + "loss": 0.314, + "step": 51022 + }, + { + "epoch": 1.96, + "learning_rate": 1.7316619646063096e-05, + "loss": 0.2756, + "step": 51023 + }, + { + "epoch": 1.96, + "learning_rate": 1.7315978456014364e-05, + "loss": 0.2717, + "step": 51024 + }, + { + "epoch": 1.96, + "learning_rate": 1.7315337265965633e-05, + "loss": 0.3809, + "step": 51025 + }, + { + "epoch": 1.96, + "learning_rate": 1.7314696075916905e-05, + "loss": 0.2479, + "step": 51026 + }, + { + "epoch": 1.96, + "learning_rate": 1.7314054885868173e-05, + "loss": 0.2296, + "step": 51027 + }, + { + "epoch": 1.96, + "learning_rate": 1.731341369581944e-05, + "loss": 0.2595, + "step": 51028 + }, + { + "epoch": 1.96, + "learning_rate": 1.731277250577071e-05, + "loss": 0.2538, + "step": 51029 + }, + { + "epoch": 1.96, + "learning_rate": 1.731213131572198e-05, + "loss": 0.3672, + "step": 51030 + }, + { + "epoch": 1.96, + "learning_rate": 1.731149012567325e-05, + "loss": 0.3279, + "step": 51031 + }, + { + "epoch": 1.96, + "learning_rate": 1.731084893562452e-05, + "loss": 0.3753, + "step": 51032 + }, + { + "epoch": 1.96, + "learning_rate": 1.731020774557579e-05, + "loss": 0.2188, + "step": 51033 + }, + { + "epoch": 1.96, + "learning_rate": 1.730956655552706e-05, + "loss": 0.3628, + "step": 51034 + }, + { + "epoch": 1.96, + "learning_rate": 1.730892536547833e-05, + "loss": 0.2272, + "step": 51035 + }, + { + "epoch": 1.96, + "learning_rate": 1.7308284175429598e-05, + "loss": 0.3257, + "step": 51036 + }, + { + "epoch": 1.96, + "learning_rate": 1.7307642985380866e-05, + "loss": 0.3186, + "step": 51037 + }, + { + "epoch": 1.96, + "learning_rate": 1.7307001795332138e-05, + "loss": 0.2966, + "step": 51038 + }, + { + "epoch": 1.96, + "learning_rate": 1.7306360605283406e-05, + "loss": 0.2715, + "step": 51039 + }, + { + "epoch": 1.96, + "learning_rate": 1.7305719415234675e-05, + "loss": 0.2269, + "step": 51040 + }, + { + "epoch": 1.96, + "learning_rate": 1.7305078225185946e-05, + "loss": 0.1885, + "step": 51041 + }, + { + "epoch": 1.96, + "learning_rate": 1.7304437035137218e-05, + "loss": 0.3805, + "step": 51042 + }, + { + "epoch": 1.96, + "learning_rate": 1.7303795845088486e-05, + "loss": 0.1776, + "step": 51043 + }, + { + "epoch": 1.96, + "learning_rate": 1.7303154655039755e-05, + "loss": 0.422, + "step": 51044 + }, + { + "epoch": 1.96, + "learning_rate": 1.7302513464991023e-05, + "loss": 0.3108, + "step": 51045 + }, + { + "epoch": 1.96, + "learning_rate": 1.7301872274942295e-05, + "loss": 0.2395, + "step": 51046 + }, + { + "epoch": 1.96, + "learning_rate": 1.7301231084893563e-05, + "loss": 0.2801, + "step": 51047 + }, + { + "epoch": 1.96, + "learning_rate": 1.730058989484483e-05, + "loss": 0.4736, + "step": 51048 + }, + { + "epoch": 1.96, + "learning_rate": 1.72999487047961e-05, + "loss": 0.1994, + "step": 51049 + }, + { + "epoch": 1.96, + "learning_rate": 1.729930751474737e-05, + "loss": 0.3968, + "step": 51050 + }, + { + "epoch": 1.96, + "learning_rate": 1.7298666324698643e-05, + "loss": 0.2155, + "step": 51051 + }, + { + "epoch": 1.96, + "learning_rate": 1.729802513464991e-05, + "loss": 0.3635, + "step": 51052 + }, + { + "epoch": 1.96, + "learning_rate": 1.729738394460118e-05, + "loss": 0.4177, + "step": 51053 + }, + { + "epoch": 1.96, + "learning_rate": 1.729674275455245e-05, + "loss": 0.351, + "step": 51054 + }, + { + "epoch": 1.96, + "learning_rate": 1.729610156450372e-05, + "loss": 0.3029, + "step": 51055 + }, + { + "epoch": 1.96, + "learning_rate": 1.7295460374454988e-05, + "loss": 0.1825, + "step": 51056 + }, + { + "epoch": 1.96, + "learning_rate": 1.7294819184406257e-05, + "loss": 0.3504, + "step": 51057 + }, + { + "epoch": 1.96, + "learning_rate": 1.729417799435753e-05, + "loss": 0.4892, + "step": 51058 + }, + { + "epoch": 1.96, + "learning_rate": 1.72935368043088e-05, + "loss": 0.2734, + "step": 51059 + }, + { + "epoch": 1.96, + "learning_rate": 1.729289561426007e-05, + "loss": 0.2758, + "step": 51060 + }, + { + "epoch": 1.96, + "learning_rate": 1.7292254424211337e-05, + "loss": 0.4398, + "step": 51061 + }, + { + "epoch": 1.96, + "learning_rate": 1.729161323416261e-05, + "loss": 0.1853, + "step": 51062 + }, + { + "epoch": 1.96, + "learning_rate": 1.7290972044113877e-05, + "loss": 0.3635, + "step": 51063 + }, + { + "epoch": 1.96, + "learning_rate": 1.7290330854065145e-05, + "loss": 0.3069, + "step": 51064 + }, + { + "epoch": 1.96, + "learning_rate": 1.7289689664016413e-05, + "loss": 0.2004, + "step": 51065 + }, + { + "epoch": 1.96, + "learning_rate": 1.7289048473967685e-05, + "loss": 0.2456, + "step": 51066 + }, + { + "epoch": 1.96, + "learning_rate": 1.7288407283918953e-05, + "loss": 0.3174, + "step": 51067 + }, + { + "epoch": 1.96, + "learning_rate": 1.7287766093870225e-05, + "loss": 0.4824, + "step": 51068 + }, + { + "epoch": 1.96, + "learning_rate": 1.7287124903821493e-05, + "loss": 0.2258, + "step": 51069 + }, + { + "epoch": 1.96, + "learning_rate": 1.7286483713772765e-05, + "loss": 0.2915, + "step": 51070 + }, + { + "epoch": 1.96, + "learning_rate": 1.7285842523724034e-05, + "loss": 0.269, + "step": 51071 + }, + { + "epoch": 1.96, + "learning_rate": 1.7285201333675302e-05, + "loss": 0.2259, + "step": 51072 + }, + { + "epoch": 1.96, + "learning_rate": 1.728456014362657e-05, + "loss": 0.3514, + "step": 51073 + }, + { + "epoch": 1.96, + "learning_rate": 1.7283918953577842e-05, + "loss": 0.3106, + "step": 51074 + }, + { + "epoch": 1.96, + "learning_rate": 1.728327776352911e-05, + "loss": 0.3341, + "step": 51075 + }, + { + "epoch": 1.96, + "learning_rate": 1.728263657348038e-05, + "loss": 0.4936, + "step": 51076 + }, + { + "epoch": 1.96, + "learning_rate": 1.728199538343165e-05, + "loss": 0.2502, + "step": 51077 + }, + { + "epoch": 1.96, + "learning_rate": 1.7281354193382922e-05, + "loss": 0.3734, + "step": 51078 + }, + { + "epoch": 1.96, + "learning_rate": 1.728071300333419e-05, + "loss": 0.3107, + "step": 51079 + }, + { + "epoch": 1.96, + "learning_rate": 1.728007181328546e-05, + "loss": 0.2247, + "step": 51080 + }, + { + "epoch": 1.96, + "learning_rate": 1.7279430623236727e-05, + "loss": 0.241, + "step": 51081 + }, + { + "epoch": 1.96, + "learning_rate": 1.7278789433188e-05, + "loss": 0.3046, + "step": 51082 + }, + { + "epoch": 1.96, + "learning_rate": 1.7278148243139267e-05, + "loss": 0.3323, + "step": 51083 + }, + { + "epoch": 1.96, + "learning_rate": 1.7277507053090535e-05, + "loss": 0.3442, + "step": 51084 + }, + { + "epoch": 1.96, + "learning_rate": 1.7276865863041804e-05, + "loss": 0.3307, + "step": 51085 + }, + { + "epoch": 1.96, + "learning_rate": 1.727622467299308e-05, + "loss": 0.2031, + "step": 51086 + }, + { + "epoch": 1.96, + "learning_rate": 1.7275583482944347e-05, + "loss": 0.2554, + "step": 51087 + }, + { + "epoch": 1.96, + "learning_rate": 1.7274942292895615e-05, + "loss": 0.5059, + "step": 51088 + }, + { + "epoch": 1.96, + "learning_rate": 1.7274301102846884e-05, + "loss": 0.3196, + "step": 51089 + }, + { + "epoch": 1.96, + "learning_rate": 1.7273659912798155e-05, + "loss": 0.3247, + "step": 51090 + }, + { + "epoch": 1.96, + "learning_rate": 1.7273018722749424e-05, + "loss": 0.4337, + "step": 51091 + }, + { + "epoch": 1.96, + "learning_rate": 1.7272377532700692e-05, + "loss": 0.4363, + "step": 51092 + }, + { + "epoch": 1.96, + "learning_rate": 1.727173634265196e-05, + "loss": 0.2252, + "step": 51093 + }, + { + "epoch": 1.96, + "learning_rate": 1.7271095152603232e-05, + "loss": 0.2741, + "step": 51094 + }, + { + "epoch": 1.96, + "learning_rate": 1.7270453962554504e-05, + "loss": 0.2658, + "step": 51095 + }, + { + "epoch": 1.96, + "learning_rate": 1.7269812772505772e-05, + "loss": 0.3526, + "step": 51096 + }, + { + "epoch": 1.96, + "learning_rate": 1.726917158245704e-05, + "loss": 0.4055, + "step": 51097 + }, + { + "epoch": 1.96, + "learning_rate": 1.7268530392408312e-05, + "loss": 0.4313, + "step": 51098 + }, + { + "epoch": 1.96, + "learning_rate": 1.726788920235958e-05, + "loss": 0.3438, + "step": 51099 + }, + { + "epoch": 1.96, + "learning_rate": 1.726724801231085e-05, + "loss": 0.4385, + "step": 51100 + }, + { + "epoch": 1.96, + "learning_rate": 1.7266606822262117e-05, + "loss": 0.2787, + "step": 51101 + }, + { + "epoch": 1.96, + "learning_rate": 1.726596563221339e-05, + "loss": 0.391, + "step": 51102 + }, + { + "epoch": 1.96, + "learning_rate": 1.7265324442164657e-05, + "loss": 0.453, + "step": 51103 + }, + { + "epoch": 1.96, + "learning_rate": 1.726468325211593e-05, + "loss": 0.3691, + "step": 51104 + }, + { + "epoch": 1.96, + "learning_rate": 1.7264042062067197e-05, + "loss": 0.2992, + "step": 51105 + }, + { + "epoch": 1.96, + "learning_rate": 1.726340087201847e-05, + "loss": 0.3164, + "step": 51106 + }, + { + "epoch": 1.96, + "learning_rate": 1.7262759681969737e-05, + "loss": 0.3547, + "step": 51107 + }, + { + "epoch": 1.96, + "learning_rate": 1.7262118491921006e-05, + "loss": 0.478, + "step": 51108 + }, + { + "epoch": 1.96, + "learning_rate": 1.7261477301872274e-05, + "loss": 0.2276, + "step": 51109 + }, + { + "epoch": 1.97, + "learning_rate": 1.7260836111823546e-05, + "loss": 0.1982, + "step": 51110 + }, + { + "epoch": 1.97, + "learning_rate": 1.7260194921774814e-05, + "loss": 0.2614, + "step": 51111 + }, + { + "epoch": 1.97, + "learning_rate": 1.7259553731726086e-05, + "loss": 0.2557, + "step": 51112 + }, + { + "epoch": 1.97, + "learning_rate": 1.7258912541677354e-05, + "loss": 0.4015, + "step": 51113 + }, + { + "epoch": 1.97, + "learning_rate": 1.7258271351628626e-05, + "loss": 0.3763, + "step": 51114 + }, + { + "epoch": 1.97, + "learning_rate": 1.7257630161579894e-05, + "loss": 0.2813, + "step": 51115 + }, + { + "epoch": 1.97, + "learning_rate": 1.7256988971531163e-05, + "loss": 0.294, + "step": 51116 + }, + { + "epoch": 1.97, + "learning_rate": 1.725634778148243e-05, + "loss": 0.212, + "step": 51117 + }, + { + "epoch": 1.97, + "learning_rate": 1.7255706591433703e-05, + "loss": 0.2657, + "step": 51118 + }, + { + "epoch": 1.97, + "learning_rate": 1.725506540138497e-05, + "loss": 0.3055, + "step": 51119 + }, + { + "epoch": 1.97, + "learning_rate": 1.725442421133624e-05, + "loss": 0.4466, + "step": 51120 + }, + { + "epoch": 1.97, + "learning_rate": 1.725378302128751e-05, + "loss": 0.2338, + "step": 51121 + }, + { + "epoch": 1.97, + "learning_rate": 1.7253141831238783e-05, + "loss": 0.3563, + "step": 51122 + }, + { + "epoch": 1.97, + "learning_rate": 1.725250064119005e-05, + "loss": 0.2674, + "step": 51123 + }, + { + "epoch": 1.97, + "learning_rate": 1.725185945114132e-05, + "loss": 0.2926, + "step": 51124 + }, + { + "epoch": 1.97, + "learning_rate": 1.7251218261092588e-05, + "loss": 0.476, + "step": 51125 + }, + { + "epoch": 1.97, + "learning_rate": 1.725057707104386e-05, + "loss": 0.2561, + "step": 51126 + }, + { + "epoch": 1.97, + "learning_rate": 1.7249935880995128e-05, + "loss": 0.3054, + "step": 51127 + }, + { + "epoch": 1.97, + "learning_rate": 1.7249294690946396e-05, + "loss": 0.262, + "step": 51128 + }, + { + "epoch": 1.97, + "learning_rate": 1.7248653500897664e-05, + "loss": 0.2545, + "step": 51129 + }, + { + "epoch": 1.97, + "learning_rate": 1.7248012310848936e-05, + "loss": 0.4242, + "step": 51130 + }, + { + "epoch": 1.97, + "learning_rate": 1.7247371120800208e-05, + "loss": 0.1415, + "step": 51131 + }, + { + "epoch": 1.97, + "learning_rate": 1.7246729930751476e-05, + "loss": 0.29, + "step": 51132 + }, + { + "epoch": 1.97, + "learning_rate": 1.7246088740702744e-05, + "loss": 0.2617, + "step": 51133 + }, + { + "epoch": 1.97, + "learning_rate": 1.7245447550654016e-05, + "loss": 0.3608, + "step": 51134 + }, + { + "epoch": 1.97, + "learning_rate": 1.7244806360605284e-05, + "loss": 0.3265, + "step": 51135 + }, + { + "epoch": 1.97, + "learning_rate": 1.7244165170556553e-05, + "loss": 0.4283, + "step": 51136 + }, + { + "epoch": 1.97, + "learning_rate": 1.724352398050782e-05, + "loss": 0.3138, + "step": 51137 + }, + { + "epoch": 1.97, + "learning_rate": 1.7242882790459093e-05, + "loss": 0.3747, + "step": 51138 + }, + { + "epoch": 1.97, + "learning_rate": 1.7242241600410365e-05, + "loss": 0.305, + "step": 51139 + }, + { + "epoch": 1.97, + "learning_rate": 1.7241600410361633e-05, + "loss": 0.3952, + "step": 51140 + }, + { + "epoch": 1.97, + "learning_rate": 1.72409592203129e-05, + "loss": 0.2752, + "step": 51141 + }, + { + "epoch": 1.97, + "learning_rate": 1.7240318030264173e-05, + "loss": 0.3186, + "step": 51142 + }, + { + "epoch": 1.97, + "learning_rate": 1.723967684021544e-05, + "loss": 0.3468, + "step": 51143 + }, + { + "epoch": 1.97, + "learning_rate": 1.723903565016671e-05, + "loss": 0.2751, + "step": 51144 + }, + { + "epoch": 1.97, + "learning_rate": 1.7238394460117978e-05, + "loss": 0.308, + "step": 51145 + }, + { + "epoch": 1.97, + "learning_rate": 1.723775327006925e-05, + "loss": 0.1482, + "step": 51146 + }, + { + "epoch": 1.97, + "learning_rate": 1.7237112080020518e-05, + "loss": 0.3161, + "step": 51147 + }, + { + "epoch": 1.97, + "learning_rate": 1.723647088997179e-05, + "loss": 0.3478, + "step": 51148 + }, + { + "epoch": 1.97, + "learning_rate": 1.7235829699923058e-05, + "loss": 0.3637, + "step": 51149 + }, + { + "epoch": 1.97, + "learning_rate": 1.723518850987433e-05, + "loss": 0.2934, + "step": 51150 + }, + { + "epoch": 1.97, + "learning_rate": 1.7234547319825598e-05, + "loss": 0.2409, + "step": 51151 + }, + { + "epoch": 1.97, + "learning_rate": 1.7233906129776866e-05, + "loss": 0.2887, + "step": 51152 + }, + { + "epoch": 1.97, + "learning_rate": 1.7233264939728135e-05, + "loss": 0.2497, + "step": 51153 + }, + { + "epoch": 1.97, + "learning_rate": 1.7232623749679406e-05, + "loss": 0.3085, + "step": 51154 + }, + { + "epoch": 1.97, + "learning_rate": 1.7231982559630675e-05, + "loss": 0.3678, + "step": 51155 + }, + { + "epoch": 1.97, + "learning_rate": 1.7231341369581943e-05, + "loss": 0.3942, + "step": 51156 + }, + { + "epoch": 1.97, + "learning_rate": 1.7230700179533215e-05, + "loss": 0.3928, + "step": 51157 + }, + { + "epoch": 1.97, + "learning_rate": 1.7230058989484487e-05, + "loss": 0.3211, + "step": 51158 + }, + { + "epoch": 1.97, + "learning_rate": 1.7229417799435755e-05, + "loss": 0.2763, + "step": 51159 + }, + { + "epoch": 1.97, + "learning_rate": 1.7228776609387023e-05, + "loss": 0.3635, + "step": 51160 + }, + { + "epoch": 1.97, + "learning_rate": 1.722813541933829e-05, + "loss": 0.2815, + "step": 51161 + }, + { + "epoch": 1.97, + "learning_rate": 1.7227494229289563e-05, + "loss": 0.37, + "step": 51162 + }, + { + "epoch": 1.97, + "learning_rate": 1.722685303924083e-05, + "loss": 0.3769, + "step": 51163 + }, + { + "epoch": 1.97, + "learning_rate": 1.72262118491921e-05, + "loss": 0.2556, + "step": 51164 + }, + { + "epoch": 1.97, + "learning_rate": 1.7225570659143368e-05, + "loss": 0.1696, + "step": 51165 + }, + { + "epoch": 1.97, + "learning_rate": 1.7224929469094643e-05, + "loss": 0.3028, + "step": 51166 + }, + { + "epoch": 1.97, + "learning_rate": 1.722428827904591e-05, + "loss": 0.382, + "step": 51167 + }, + { + "epoch": 1.97, + "learning_rate": 1.722364708899718e-05, + "loss": 0.2661, + "step": 51168 + }, + { + "epoch": 1.97, + "learning_rate": 1.722300589894845e-05, + "loss": 0.3183, + "step": 51169 + }, + { + "epoch": 1.97, + "learning_rate": 1.722236470889972e-05, + "loss": 0.2678, + "step": 51170 + }, + { + "epoch": 1.97, + "learning_rate": 1.722172351885099e-05, + "loss": 0.2181, + "step": 51171 + }, + { + "epoch": 1.97, + "learning_rate": 1.7221082328802257e-05, + "loss": 0.36, + "step": 51172 + }, + { + "epoch": 1.97, + "learning_rate": 1.7220441138753525e-05, + "loss": 0.3313, + "step": 51173 + }, + { + "epoch": 1.97, + "learning_rate": 1.7219799948704797e-05, + "loss": 0.3388, + "step": 51174 + }, + { + "epoch": 1.97, + "learning_rate": 1.721915875865607e-05, + "loss": 0.3753, + "step": 51175 + }, + { + "epoch": 1.97, + "learning_rate": 1.7218517568607337e-05, + "loss": 0.2982, + "step": 51176 + }, + { + "epoch": 1.97, + "learning_rate": 1.7217876378558605e-05, + "loss": 0.3847, + "step": 51177 + }, + { + "epoch": 1.97, + "learning_rate": 1.7217235188509877e-05, + "loss": 0.2808, + "step": 51178 + }, + { + "epoch": 1.97, + "learning_rate": 1.7216593998461145e-05, + "loss": 0.4036, + "step": 51179 + }, + { + "epoch": 1.97, + "learning_rate": 1.7215952808412413e-05, + "loss": 0.3048, + "step": 51180 + }, + { + "epoch": 1.97, + "learning_rate": 1.7215311618363682e-05, + "loss": 0.3759, + "step": 51181 + }, + { + "epoch": 1.97, + "learning_rate": 1.7214670428314954e-05, + "loss": 0.2675, + "step": 51182 + }, + { + "epoch": 1.97, + "learning_rate": 1.7214029238266222e-05, + "loss": 0.2159, + "step": 51183 + }, + { + "epoch": 1.97, + "learning_rate": 1.7213388048217494e-05, + "loss": 0.3186, + "step": 51184 + }, + { + "epoch": 1.97, + "learning_rate": 1.7212746858168762e-05, + "loss": 0.4924, + "step": 51185 + }, + { + "epoch": 1.97, + "learning_rate": 1.7212105668120034e-05, + "loss": 0.2344, + "step": 51186 + }, + { + "epoch": 1.97, + "learning_rate": 1.7211464478071302e-05, + "loss": 0.3959, + "step": 51187 + }, + { + "epoch": 1.97, + "learning_rate": 1.721082328802257e-05, + "loss": 0.1579, + "step": 51188 + }, + { + "epoch": 1.97, + "learning_rate": 1.721018209797384e-05, + "loss": 0.2689, + "step": 51189 + }, + { + "epoch": 1.97, + "learning_rate": 1.720954090792511e-05, + "loss": 0.3189, + "step": 51190 + }, + { + "epoch": 1.97, + "learning_rate": 1.720889971787638e-05, + "loss": 0.2724, + "step": 51191 + }, + { + "epoch": 1.97, + "learning_rate": 1.7208258527827647e-05, + "loss": 0.3425, + "step": 51192 + }, + { + "epoch": 1.97, + "learning_rate": 1.720761733777892e-05, + "loss": 0.3045, + "step": 51193 + }, + { + "epoch": 1.97, + "learning_rate": 1.720697614773019e-05, + "loss": 0.3693, + "step": 51194 + }, + { + "epoch": 1.97, + "learning_rate": 1.720633495768146e-05, + "loss": 0.2514, + "step": 51195 + }, + { + "epoch": 1.97, + "learning_rate": 1.7205693767632727e-05, + "loss": 0.3628, + "step": 51196 + }, + { + "epoch": 1.97, + "learning_rate": 1.7205052577583995e-05, + "loss": 0.4168, + "step": 51197 + }, + { + "epoch": 1.97, + "learning_rate": 1.7204411387535267e-05, + "loss": 0.2503, + "step": 51198 + }, + { + "epoch": 1.97, + "learning_rate": 1.7203770197486535e-05, + "loss": 0.3128, + "step": 51199 + }, + { + "epoch": 1.97, + "learning_rate": 1.7203129007437804e-05, + "loss": 0.2188, + "step": 51200 + }, + { + "epoch": 1.97, + "learning_rate": 1.7202487817389075e-05, + "loss": 0.2034, + "step": 51201 + }, + { + "epoch": 1.97, + "learning_rate": 1.7201846627340347e-05, + "loss": 0.187, + "step": 51202 + }, + { + "epoch": 1.97, + "learning_rate": 1.7201205437291616e-05, + "loss": 0.38, + "step": 51203 + }, + { + "epoch": 1.97, + "learning_rate": 1.7200564247242884e-05, + "loss": 0.3374, + "step": 51204 + }, + { + "epoch": 1.97, + "learning_rate": 1.7199923057194152e-05, + "loss": 0.472, + "step": 51205 + }, + { + "epoch": 1.97, + "learning_rate": 1.7199281867145424e-05, + "loss": 0.4022, + "step": 51206 + }, + { + "epoch": 1.97, + "learning_rate": 1.7198640677096692e-05, + "loss": 0.3074, + "step": 51207 + }, + { + "epoch": 1.97, + "learning_rate": 1.719799948704796e-05, + "loss": 0.2699, + "step": 51208 + }, + { + "epoch": 1.97, + "learning_rate": 1.719735829699923e-05, + "loss": 0.3229, + "step": 51209 + }, + { + "epoch": 1.97, + "learning_rate": 1.7196717106950504e-05, + "loss": 0.3959, + "step": 51210 + }, + { + "epoch": 1.97, + "learning_rate": 1.7196075916901772e-05, + "loss": 0.3893, + "step": 51211 + }, + { + "epoch": 1.97, + "learning_rate": 1.719543472685304e-05, + "loss": 0.2532, + "step": 51212 + }, + { + "epoch": 1.97, + "learning_rate": 1.719479353680431e-05, + "loss": 0.3418, + "step": 51213 + }, + { + "epoch": 1.97, + "learning_rate": 1.719415234675558e-05, + "loss": 0.2137, + "step": 51214 + }, + { + "epoch": 1.97, + "learning_rate": 1.719351115670685e-05, + "loss": 0.4678, + "step": 51215 + }, + { + "epoch": 1.97, + "learning_rate": 1.7192869966658117e-05, + "loss": 0.2066, + "step": 51216 + }, + { + "epoch": 1.97, + "learning_rate": 1.7192228776609386e-05, + "loss": 0.211, + "step": 51217 + }, + { + "epoch": 1.97, + "learning_rate": 1.7191587586560657e-05, + "loss": 0.3496, + "step": 51218 + }, + { + "epoch": 1.97, + "learning_rate": 1.719094639651193e-05, + "loss": 0.3317, + "step": 51219 + }, + { + "epoch": 1.97, + "learning_rate": 1.7190305206463197e-05, + "loss": 0.3759, + "step": 51220 + }, + { + "epoch": 1.97, + "learning_rate": 1.7189664016414466e-05, + "loss": 0.3273, + "step": 51221 + }, + { + "epoch": 1.97, + "learning_rate": 1.7189022826365737e-05, + "loss": 0.2595, + "step": 51222 + }, + { + "epoch": 1.97, + "learning_rate": 1.7188381636317006e-05, + "loss": 0.3332, + "step": 51223 + }, + { + "epoch": 1.97, + "learning_rate": 1.7187740446268274e-05, + "loss": 0.2951, + "step": 51224 + }, + { + "epoch": 1.97, + "learning_rate": 1.7187099256219542e-05, + "loss": 0.2737, + "step": 51225 + }, + { + "epoch": 1.97, + "learning_rate": 1.7186458066170814e-05, + "loss": 0.2483, + "step": 51226 + }, + { + "epoch": 1.97, + "learning_rate": 1.7185816876122083e-05, + "loss": 0.3088, + "step": 51227 + }, + { + "epoch": 1.97, + "learning_rate": 1.7185175686073354e-05, + "loss": 0.3548, + "step": 51228 + }, + { + "epoch": 1.97, + "learning_rate": 1.7184534496024623e-05, + "loss": 0.2911, + "step": 51229 + }, + { + "epoch": 1.97, + "learning_rate": 1.7183893305975894e-05, + "loss": 0.3196, + "step": 51230 + }, + { + "epoch": 1.97, + "learning_rate": 1.7183252115927163e-05, + "loss": 0.2933, + "step": 51231 + }, + { + "epoch": 1.97, + "learning_rate": 1.718261092587843e-05, + "loss": 0.2772, + "step": 51232 + }, + { + "epoch": 1.97, + "learning_rate": 1.71819697358297e-05, + "loss": 0.3303, + "step": 51233 + }, + { + "epoch": 1.97, + "learning_rate": 1.718132854578097e-05, + "loss": 0.2931, + "step": 51234 + }, + { + "epoch": 1.97, + "learning_rate": 1.718068735573224e-05, + "loss": 0.3273, + "step": 51235 + }, + { + "epoch": 1.97, + "learning_rate": 1.7180046165683508e-05, + "loss": 0.3321, + "step": 51236 + }, + { + "epoch": 1.97, + "learning_rate": 1.717940497563478e-05, + "loss": 0.4259, + "step": 51237 + }, + { + "epoch": 1.97, + "learning_rate": 1.717876378558605e-05, + "loss": 0.2008, + "step": 51238 + }, + { + "epoch": 1.97, + "learning_rate": 1.717812259553732e-05, + "loss": 0.3083, + "step": 51239 + }, + { + "epoch": 1.97, + "learning_rate": 1.7177481405488588e-05, + "loss": 0.308, + "step": 51240 + }, + { + "epoch": 1.97, + "learning_rate": 1.7176840215439856e-05, + "loss": 0.2292, + "step": 51241 + }, + { + "epoch": 1.97, + "learning_rate": 1.7176199025391128e-05, + "loss": 0.3687, + "step": 51242 + }, + { + "epoch": 1.97, + "learning_rate": 1.7175557835342396e-05, + "loss": 0.4162, + "step": 51243 + }, + { + "epoch": 1.97, + "learning_rate": 1.7174916645293664e-05, + "loss": 0.349, + "step": 51244 + }, + { + "epoch": 1.97, + "learning_rate": 1.7174275455244933e-05, + "loss": 0.1988, + "step": 51245 + }, + { + "epoch": 1.97, + "learning_rate": 1.7173634265196208e-05, + "loss": 0.2797, + "step": 51246 + }, + { + "epoch": 1.97, + "learning_rate": 1.7172993075147476e-05, + "loss": 0.3917, + "step": 51247 + }, + { + "epoch": 1.97, + "learning_rate": 1.7172351885098745e-05, + "loss": 0.3902, + "step": 51248 + }, + { + "epoch": 1.97, + "learning_rate": 1.7171710695050013e-05, + "loss": 0.439, + "step": 51249 + }, + { + "epoch": 1.97, + "learning_rate": 1.7171069505001285e-05, + "loss": 0.3451, + "step": 51250 + }, + { + "epoch": 1.97, + "learning_rate": 1.7170428314952553e-05, + "loss": 0.2755, + "step": 51251 + }, + { + "epoch": 1.97, + "learning_rate": 1.716978712490382e-05, + "loss": 0.3298, + "step": 51252 + }, + { + "epoch": 1.97, + "learning_rate": 1.716914593485509e-05, + "loss": 0.3194, + "step": 51253 + }, + { + "epoch": 1.97, + "learning_rate": 1.716850474480636e-05, + "loss": 0.269, + "step": 51254 + }, + { + "epoch": 1.97, + "learning_rate": 1.7167863554757633e-05, + "loss": 0.4538, + "step": 51255 + }, + { + "epoch": 1.97, + "learning_rate": 1.71672223647089e-05, + "loss": 0.1744, + "step": 51256 + }, + { + "epoch": 1.97, + "learning_rate": 1.716658117466017e-05, + "loss": 0.4032, + "step": 51257 + }, + { + "epoch": 1.97, + "learning_rate": 1.716593998461144e-05, + "loss": 0.3843, + "step": 51258 + }, + { + "epoch": 1.97, + "learning_rate": 1.716529879456271e-05, + "loss": 0.3281, + "step": 51259 + }, + { + "epoch": 1.97, + "learning_rate": 1.7164657604513978e-05, + "loss": 0.4172, + "step": 51260 + }, + { + "epoch": 1.97, + "learning_rate": 1.7164016414465246e-05, + "loss": 0.3005, + "step": 51261 + }, + { + "epoch": 1.97, + "learning_rate": 1.7163375224416518e-05, + "loss": 0.2971, + "step": 51262 + }, + { + "epoch": 1.97, + "learning_rate": 1.7162734034367786e-05, + "loss": 0.3906, + "step": 51263 + }, + { + "epoch": 1.97, + "learning_rate": 1.7162092844319058e-05, + "loss": 0.3397, + "step": 51264 + }, + { + "epoch": 1.97, + "learning_rate": 1.7161451654270326e-05, + "loss": 0.3251, + "step": 51265 + }, + { + "epoch": 1.97, + "learning_rate": 1.7160810464221598e-05, + "loss": 0.2841, + "step": 51266 + }, + { + "epoch": 1.97, + "learning_rate": 1.7160169274172866e-05, + "loss": 0.3802, + "step": 51267 + }, + { + "epoch": 1.97, + "learning_rate": 1.7159528084124135e-05, + "loss": 0.2637, + "step": 51268 + }, + { + "epoch": 1.97, + "learning_rate": 1.7158886894075403e-05, + "loss": 0.3578, + "step": 51269 + }, + { + "epoch": 1.97, + "learning_rate": 1.7158245704026675e-05, + "loss": 0.3914, + "step": 51270 + }, + { + "epoch": 1.97, + "learning_rate": 1.7157604513977943e-05, + "loss": 0.2929, + "step": 51271 + }, + { + "epoch": 1.97, + "learning_rate": 1.715696332392921e-05, + "loss": 0.2522, + "step": 51272 + }, + { + "epoch": 1.97, + "learning_rate": 1.7156322133880483e-05, + "loss": 0.2938, + "step": 51273 + }, + { + "epoch": 1.97, + "learning_rate": 1.7155680943831755e-05, + "loss": 0.2914, + "step": 51274 + }, + { + "epoch": 1.97, + "learning_rate": 1.7155039753783023e-05, + "loss": 0.2202, + "step": 51275 + }, + { + "epoch": 1.97, + "learning_rate": 1.715439856373429e-05, + "loss": 0.2579, + "step": 51276 + }, + { + "epoch": 1.97, + "learning_rate": 1.715375737368556e-05, + "loss": 0.3606, + "step": 51277 + }, + { + "epoch": 1.97, + "learning_rate": 1.715311618363683e-05, + "loss": 0.3173, + "step": 51278 + }, + { + "epoch": 1.97, + "learning_rate": 1.71524749935881e-05, + "loss": 0.3159, + "step": 51279 + }, + { + "epoch": 1.97, + "learning_rate": 1.7151833803539368e-05, + "loss": 0.3606, + "step": 51280 + }, + { + "epoch": 1.97, + "learning_rate": 1.7151192613490637e-05, + "loss": 0.4093, + "step": 51281 + }, + { + "epoch": 1.97, + "learning_rate": 1.7150551423441912e-05, + "loss": 0.358, + "step": 51282 + }, + { + "epoch": 1.97, + "learning_rate": 1.714991023339318e-05, + "loss": 0.3353, + "step": 51283 + }, + { + "epoch": 1.97, + "learning_rate": 1.714926904334445e-05, + "loss": 0.2638, + "step": 51284 + }, + { + "epoch": 1.97, + "learning_rate": 1.7148627853295717e-05, + "loss": 0.2873, + "step": 51285 + }, + { + "epoch": 1.97, + "learning_rate": 1.714798666324699e-05, + "loss": 0.2984, + "step": 51286 + }, + { + "epoch": 1.97, + "learning_rate": 1.7147345473198257e-05, + "loss": 0.442, + "step": 51287 + }, + { + "epoch": 1.97, + "learning_rate": 1.7146704283149525e-05, + "loss": 0.3208, + "step": 51288 + }, + { + "epoch": 1.97, + "learning_rate": 1.7146063093100793e-05, + "loss": 0.3226, + "step": 51289 + }, + { + "epoch": 1.97, + "learning_rate": 1.7145421903052065e-05, + "loss": 0.295, + "step": 51290 + }, + { + "epoch": 1.97, + "learning_rate": 1.7144780713003337e-05, + "loss": 0.2975, + "step": 51291 + }, + { + "epoch": 1.97, + "learning_rate": 1.7144139522954605e-05, + "loss": 0.2586, + "step": 51292 + }, + { + "epoch": 1.97, + "learning_rate": 1.7143498332905874e-05, + "loss": 0.1764, + "step": 51293 + }, + { + "epoch": 1.97, + "learning_rate": 1.7142857142857145e-05, + "loss": 0.3757, + "step": 51294 + }, + { + "epoch": 1.97, + "learning_rate": 1.7142215952808414e-05, + "loss": 0.235, + "step": 51295 + }, + { + "epoch": 1.97, + "learning_rate": 1.7141574762759682e-05, + "loss": 0.279, + "step": 51296 + }, + { + "epoch": 1.97, + "learning_rate": 1.714093357271095e-05, + "loss": 0.2615, + "step": 51297 + }, + { + "epoch": 1.97, + "learning_rate": 1.7140292382662222e-05, + "loss": 0.4644, + "step": 51298 + }, + { + "epoch": 1.97, + "learning_rate": 1.7139651192613494e-05, + "loss": 0.3282, + "step": 51299 + }, + { + "epoch": 1.97, + "learning_rate": 1.7139010002564762e-05, + "loss": 0.358, + "step": 51300 + }, + { + "epoch": 1.97, + "learning_rate": 1.713836881251603e-05, + "loss": 0.4244, + "step": 51301 + }, + { + "epoch": 1.97, + "learning_rate": 1.7137727622467302e-05, + "loss": 0.3425, + "step": 51302 + }, + { + "epoch": 1.97, + "learning_rate": 1.713708643241857e-05, + "loss": 0.3597, + "step": 51303 + }, + { + "epoch": 1.97, + "learning_rate": 1.713644524236984e-05, + "loss": 0.3325, + "step": 51304 + }, + { + "epoch": 1.97, + "learning_rate": 1.7135804052321107e-05, + "loss": 0.2972, + "step": 51305 + }, + { + "epoch": 1.97, + "learning_rate": 1.713516286227238e-05, + "loss": 0.3281, + "step": 51306 + }, + { + "epoch": 1.97, + "learning_rate": 1.7134521672223647e-05, + "loss": 0.2157, + "step": 51307 + }, + { + "epoch": 1.97, + "learning_rate": 1.713388048217492e-05, + "loss": 0.2944, + "step": 51308 + }, + { + "epoch": 1.97, + "learning_rate": 1.7133239292126187e-05, + "loss": 0.2591, + "step": 51309 + }, + { + "epoch": 1.97, + "learning_rate": 1.713259810207746e-05, + "loss": 0.3773, + "step": 51310 + }, + { + "epoch": 1.97, + "learning_rate": 1.7131956912028727e-05, + "loss": 0.3533, + "step": 51311 + }, + { + "epoch": 1.97, + "learning_rate": 1.7131315721979995e-05, + "loss": 0.2438, + "step": 51312 + }, + { + "epoch": 1.97, + "learning_rate": 1.7130674531931264e-05, + "loss": 0.3118, + "step": 51313 + }, + { + "epoch": 1.97, + "learning_rate": 1.7130033341882536e-05, + "loss": 0.3672, + "step": 51314 + }, + { + "epoch": 1.97, + "learning_rate": 1.7129392151833804e-05, + "loss": 0.4487, + "step": 51315 + }, + { + "epoch": 1.97, + "learning_rate": 1.7128750961785072e-05, + "loss": 0.2738, + "step": 51316 + }, + { + "epoch": 1.97, + "learning_rate": 1.7128109771736344e-05, + "loss": 0.2627, + "step": 51317 + }, + { + "epoch": 1.97, + "learning_rate": 1.7127468581687616e-05, + "loss": 0.3627, + "step": 51318 + }, + { + "epoch": 1.97, + "learning_rate": 1.7126827391638884e-05, + "loss": 0.3238, + "step": 51319 + }, + { + "epoch": 1.97, + "learning_rate": 1.7126186201590152e-05, + "loss": 0.1922, + "step": 51320 + }, + { + "epoch": 1.97, + "learning_rate": 1.712554501154142e-05, + "loss": 0.3444, + "step": 51321 + }, + { + "epoch": 1.97, + "learning_rate": 1.7124903821492692e-05, + "loss": 0.2431, + "step": 51322 + }, + { + "epoch": 1.97, + "learning_rate": 1.712426263144396e-05, + "loss": 0.2641, + "step": 51323 + }, + { + "epoch": 1.97, + "learning_rate": 1.712362144139523e-05, + "loss": 0.2579, + "step": 51324 + }, + { + "epoch": 1.97, + "learning_rate": 1.7122980251346497e-05, + "loss": 0.4007, + "step": 51325 + }, + { + "epoch": 1.97, + "learning_rate": 1.7122339061297772e-05, + "loss": 0.3397, + "step": 51326 + }, + { + "epoch": 1.97, + "learning_rate": 1.712169787124904e-05, + "loss": 0.3809, + "step": 51327 + }, + { + "epoch": 1.97, + "learning_rate": 1.712105668120031e-05, + "loss": 0.4196, + "step": 51328 + }, + { + "epoch": 1.97, + "learning_rate": 1.7120415491151577e-05, + "loss": 0.3147, + "step": 51329 + }, + { + "epoch": 1.97, + "learning_rate": 1.711977430110285e-05, + "loss": 0.4085, + "step": 51330 + }, + { + "epoch": 1.97, + "learning_rate": 1.7119133111054117e-05, + "loss": 0.2232, + "step": 51331 + }, + { + "epoch": 1.97, + "learning_rate": 1.7118491921005386e-05, + "loss": 0.2745, + "step": 51332 + }, + { + "epoch": 1.97, + "learning_rate": 1.7117850730956654e-05, + "loss": 0.4776, + "step": 51333 + }, + { + "epoch": 1.97, + "learning_rate": 1.7117209540907926e-05, + "loss": 0.2515, + "step": 51334 + }, + { + "epoch": 1.97, + "learning_rate": 1.7116568350859198e-05, + "loss": 0.3753, + "step": 51335 + }, + { + "epoch": 1.97, + "learning_rate": 1.7115927160810466e-05, + "loss": 0.3131, + "step": 51336 + }, + { + "epoch": 1.97, + "learning_rate": 1.7115285970761734e-05, + "loss": 0.3555, + "step": 51337 + }, + { + "epoch": 1.97, + "learning_rate": 1.7114644780713006e-05, + "loss": 0.308, + "step": 51338 + }, + { + "epoch": 1.97, + "learning_rate": 1.7114003590664274e-05, + "loss": 0.3113, + "step": 51339 + }, + { + "epoch": 1.97, + "learning_rate": 1.7113362400615543e-05, + "loss": 0.3619, + "step": 51340 + }, + { + "epoch": 1.97, + "learning_rate": 1.711272121056681e-05, + "loss": 0.2754, + "step": 51341 + }, + { + "epoch": 1.97, + "learning_rate": 1.7112080020518083e-05, + "loss": 0.3049, + "step": 51342 + }, + { + "epoch": 1.97, + "learning_rate": 1.711143883046935e-05, + "loss": 0.2385, + "step": 51343 + }, + { + "epoch": 1.97, + "learning_rate": 1.7110797640420623e-05, + "loss": 0.3785, + "step": 51344 + }, + { + "epoch": 1.97, + "learning_rate": 1.711015645037189e-05, + "loss": 0.3122, + "step": 51345 + }, + { + "epoch": 1.97, + "learning_rate": 1.7109515260323163e-05, + "loss": 0.2444, + "step": 51346 + }, + { + "epoch": 1.97, + "learning_rate": 1.710887407027443e-05, + "loss": 0.3542, + "step": 51347 + }, + { + "epoch": 1.97, + "learning_rate": 1.71082328802257e-05, + "loss": 0.4045, + "step": 51348 + }, + { + "epoch": 1.97, + "learning_rate": 1.7107591690176968e-05, + "loss": 0.2016, + "step": 51349 + }, + { + "epoch": 1.97, + "learning_rate": 1.710695050012824e-05, + "loss": 0.3435, + "step": 51350 + }, + { + "epoch": 1.97, + "learning_rate": 1.7106309310079508e-05, + "loss": 0.2524, + "step": 51351 + }, + { + "epoch": 1.97, + "learning_rate": 1.7105668120030776e-05, + "loss": 0.3936, + "step": 51352 + }, + { + "epoch": 1.97, + "learning_rate": 1.7105026929982048e-05, + "loss": 0.5168, + "step": 51353 + }, + { + "epoch": 1.97, + "learning_rate": 1.710438573993332e-05, + "loss": 0.4309, + "step": 51354 + }, + { + "epoch": 1.97, + "learning_rate": 1.7103744549884588e-05, + "loss": 0.3633, + "step": 51355 + }, + { + "epoch": 1.97, + "learning_rate": 1.7103103359835856e-05, + "loss": 0.2593, + "step": 51356 + }, + { + "epoch": 1.97, + "learning_rate": 1.7102462169787124e-05, + "loss": 0.242, + "step": 51357 + }, + { + "epoch": 1.97, + "learning_rate": 1.7101820979738396e-05, + "loss": 0.321, + "step": 51358 + }, + { + "epoch": 1.97, + "learning_rate": 1.7101179789689665e-05, + "loss": 0.277, + "step": 51359 + }, + { + "epoch": 1.97, + "learning_rate": 1.7100538599640933e-05, + "loss": 0.2632, + "step": 51360 + }, + { + "epoch": 1.97, + "learning_rate": 1.70998974095922e-05, + "loss": 0.3528, + "step": 51361 + }, + { + "epoch": 1.97, + "learning_rate": 1.7099256219543476e-05, + "loss": 0.2905, + "step": 51362 + }, + { + "epoch": 1.97, + "learning_rate": 1.7098615029494745e-05, + "loss": 0.4043, + "step": 51363 + }, + { + "epoch": 1.97, + "learning_rate": 1.7097973839446013e-05, + "loss": 0.4527, + "step": 51364 + }, + { + "epoch": 1.97, + "learning_rate": 1.709733264939728e-05, + "loss": 0.2637, + "step": 51365 + }, + { + "epoch": 1.97, + "learning_rate": 1.7096691459348553e-05, + "loss": 0.495, + "step": 51366 + }, + { + "epoch": 1.97, + "learning_rate": 1.709605026929982e-05, + "loss": 0.3205, + "step": 51367 + }, + { + "epoch": 1.97, + "learning_rate": 1.709540907925109e-05, + "loss": 0.2394, + "step": 51368 + }, + { + "epoch": 1.97, + "learning_rate": 1.7094767889202358e-05, + "loss": 0.3017, + "step": 51369 + }, + { + "epoch": 1.98, + "learning_rate": 1.709412669915363e-05, + "loss": 0.4632, + "step": 51370 + }, + { + "epoch": 1.98, + "learning_rate": 1.70934855091049e-05, + "loss": 0.3764, + "step": 51371 + }, + { + "epoch": 1.98, + "learning_rate": 1.709284431905617e-05, + "loss": 0.3266, + "step": 51372 + }, + { + "epoch": 1.98, + "learning_rate": 1.7092203129007438e-05, + "loss": 0.2518, + "step": 51373 + }, + { + "epoch": 1.98, + "learning_rate": 1.709156193895871e-05, + "loss": 0.2731, + "step": 51374 + }, + { + "epoch": 1.98, + "learning_rate": 1.7090920748909978e-05, + "loss": 0.3386, + "step": 51375 + }, + { + "epoch": 1.98, + "learning_rate": 1.7090279558861246e-05, + "loss": 0.311, + "step": 51376 + }, + { + "epoch": 1.98, + "learning_rate": 1.7089638368812515e-05, + "loss": 0.2779, + "step": 51377 + }, + { + "epoch": 1.98, + "learning_rate": 1.7088997178763786e-05, + "loss": 0.3024, + "step": 51378 + }, + { + "epoch": 1.98, + "learning_rate": 1.7088355988715055e-05, + "loss": 0.2649, + "step": 51379 + }, + { + "epoch": 1.98, + "learning_rate": 1.7087714798666327e-05, + "loss": 0.3428, + "step": 51380 + }, + { + "epoch": 1.98, + "learning_rate": 1.7087073608617595e-05, + "loss": 0.2271, + "step": 51381 + }, + { + "epoch": 1.98, + "learning_rate": 1.7086432418568867e-05, + "loss": 0.294, + "step": 51382 + }, + { + "epoch": 1.98, + "learning_rate": 1.7085791228520135e-05, + "loss": 0.2113, + "step": 51383 + }, + { + "epoch": 1.98, + "learning_rate": 1.7085150038471403e-05, + "loss": 0.3915, + "step": 51384 + }, + { + "epoch": 1.98, + "learning_rate": 1.708450884842267e-05, + "loss": 0.3482, + "step": 51385 + }, + { + "epoch": 1.98, + "learning_rate": 1.7083867658373943e-05, + "loss": 0.2653, + "step": 51386 + }, + { + "epoch": 1.98, + "learning_rate": 1.708322646832521e-05, + "loss": 0.3343, + "step": 51387 + }, + { + "epoch": 1.98, + "learning_rate": 1.708258527827648e-05, + "loss": 0.2345, + "step": 51388 + }, + { + "epoch": 1.98, + "learning_rate": 1.708194408822775e-05, + "loss": 0.2702, + "step": 51389 + }, + { + "epoch": 1.98, + "learning_rate": 1.7081302898179023e-05, + "loss": 0.3928, + "step": 51390 + }, + { + "epoch": 1.98, + "learning_rate": 1.708066170813029e-05, + "loss": 0.3157, + "step": 51391 + }, + { + "epoch": 1.98, + "learning_rate": 1.708002051808156e-05, + "loss": 0.29, + "step": 51392 + }, + { + "epoch": 1.98, + "learning_rate": 1.707937932803283e-05, + "loss": 0.2492, + "step": 51393 + }, + { + "epoch": 1.98, + "learning_rate": 1.70787381379841e-05, + "loss": 0.3124, + "step": 51394 + }, + { + "epoch": 1.98, + "learning_rate": 1.707809694793537e-05, + "loss": 0.3577, + "step": 51395 + }, + { + "epoch": 1.98, + "learning_rate": 1.7077455757886637e-05, + "loss": 0.271, + "step": 51396 + }, + { + "epoch": 1.98, + "learning_rate": 1.707681456783791e-05, + "loss": 0.2323, + "step": 51397 + }, + { + "epoch": 1.98, + "learning_rate": 1.707617337778918e-05, + "loss": 0.2512, + "step": 51398 + }, + { + "epoch": 1.98, + "learning_rate": 1.707553218774045e-05, + "loss": 0.2397, + "step": 51399 + }, + { + "epoch": 1.98, + "learning_rate": 1.7074890997691717e-05, + "loss": 0.2292, + "step": 51400 + }, + { + "epoch": 1.98, + "learning_rate": 1.7074249807642985e-05, + "loss": 0.2489, + "step": 51401 + }, + { + "epoch": 1.98, + "learning_rate": 1.7073608617594257e-05, + "loss": 0.1955, + "step": 51402 + }, + { + "epoch": 1.98, + "learning_rate": 1.7072967427545525e-05, + "loss": 0.3629, + "step": 51403 + }, + { + "epoch": 1.98, + "learning_rate": 1.7072326237496793e-05, + "loss": 0.2705, + "step": 51404 + }, + { + "epoch": 1.98, + "learning_rate": 1.7071685047448062e-05, + "loss": 0.3072, + "step": 51405 + }, + { + "epoch": 1.98, + "learning_rate": 1.7071043857399337e-05, + "loss": 0.2946, + "step": 51406 + }, + { + "epoch": 1.98, + "learning_rate": 1.7070402667350605e-05, + "loss": 0.274, + "step": 51407 + }, + { + "epoch": 1.98, + "learning_rate": 1.7069761477301874e-05, + "loss": 0.4054, + "step": 51408 + }, + { + "epoch": 1.98, + "learning_rate": 1.7069120287253142e-05, + "loss": 0.3314, + "step": 51409 + }, + { + "epoch": 1.98, + "learning_rate": 1.7068479097204414e-05, + "loss": 0.3946, + "step": 51410 + }, + { + "epoch": 1.98, + "learning_rate": 1.7067837907155682e-05, + "loss": 0.3221, + "step": 51411 + }, + { + "epoch": 1.98, + "learning_rate": 1.706719671710695e-05, + "loss": 0.4576, + "step": 51412 + }, + { + "epoch": 1.98, + "learning_rate": 1.706655552705822e-05, + "loss": 0.5443, + "step": 51413 + }, + { + "epoch": 1.98, + "learning_rate": 1.706591433700949e-05, + "loss": 0.5246, + "step": 51414 + }, + { + "epoch": 1.98, + "learning_rate": 1.7065273146960762e-05, + "loss": 0.2682, + "step": 51415 + }, + { + "epoch": 1.98, + "learning_rate": 1.706463195691203e-05, + "loss": 0.3347, + "step": 51416 + }, + { + "epoch": 1.98, + "learning_rate": 1.70639907668633e-05, + "loss": 0.2403, + "step": 51417 + }, + { + "epoch": 1.98, + "learning_rate": 1.706334957681457e-05, + "loss": 0.3395, + "step": 51418 + }, + { + "epoch": 1.98, + "learning_rate": 1.706270838676584e-05, + "loss": 0.278, + "step": 51419 + }, + { + "epoch": 1.98, + "learning_rate": 1.7062067196717107e-05, + "loss": 0.4006, + "step": 51420 + }, + { + "epoch": 1.98, + "learning_rate": 1.7061426006668375e-05, + "loss": 0.2494, + "step": 51421 + }, + { + "epoch": 1.98, + "learning_rate": 1.7060784816619647e-05, + "loss": 0.3183, + "step": 51422 + }, + { + "epoch": 1.98, + "learning_rate": 1.7060143626570915e-05, + "loss": 0.292, + "step": 51423 + }, + { + "epoch": 1.98, + "learning_rate": 1.7059502436522187e-05, + "loss": 0.4168, + "step": 51424 + }, + { + "epoch": 1.98, + "learning_rate": 1.7058861246473456e-05, + "loss": 0.253, + "step": 51425 + }, + { + "epoch": 1.98, + "learning_rate": 1.7058220056424727e-05, + "loss": 0.3312, + "step": 51426 + }, + { + "epoch": 1.98, + "learning_rate": 1.7057578866375996e-05, + "loss": 0.2891, + "step": 51427 + }, + { + "epoch": 1.98, + "learning_rate": 1.7056937676327264e-05, + "loss": 0.276, + "step": 51428 + }, + { + "epoch": 1.98, + "learning_rate": 1.7056296486278532e-05, + "loss": 0.4156, + "step": 51429 + }, + { + "epoch": 1.98, + "learning_rate": 1.7055655296229804e-05, + "loss": 0.3579, + "step": 51430 + }, + { + "epoch": 1.98, + "learning_rate": 1.7055014106181072e-05, + "loss": 0.4636, + "step": 51431 + }, + { + "epoch": 1.98, + "learning_rate": 1.705437291613234e-05, + "loss": 0.4835, + "step": 51432 + }, + { + "epoch": 1.98, + "learning_rate": 1.7053731726083612e-05, + "loss": 0.2329, + "step": 51433 + }, + { + "epoch": 1.98, + "learning_rate": 1.7053090536034884e-05, + "loss": 0.2997, + "step": 51434 + }, + { + "epoch": 1.98, + "learning_rate": 1.7052449345986152e-05, + "loss": 0.3214, + "step": 51435 + }, + { + "epoch": 1.98, + "learning_rate": 1.705180815593742e-05, + "loss": 0.2589, + "step": 51436 + }, + { + "epoch": 1.98, + "learning_rate": 1.705116696588869e-05, + "loss": 0.478, + "step": 51437 + }, + { + "epoch": 1.98, + "learning_rate": 1.705052577583996e-05, + "loss": 0.3905, + "step": 51438 + }, + { + "epoch": 1.98, + "learning_rate": 1.704988458579123e-05, + "loss": 0.3982, + "step": 51439 + }, + { + "epoch": 1.98, + "learning_rate": 1.7049243395742497e-05, + "loss": 0.3229, + "step": 51440 + }, + { + "epoch": 1.98, + "learning_rate": 1.7048602205693766e-05, + "loss": 0.3297, + "step": 51441 + }, + { + "epoch": 1.98, + "learning_rate": 1.704796101564504e-05, + "loss": 0.306, + "step": 51442 + }, + { + "epoch": 1.98, + "learning_rate": 1.704731982559631e-05, + "loss": 0.3092, + "step": 51443 + }, + { + "epoch": 1.98, + "learning_rate": 1.7046678635547577e-05, + "loss": 0.2936, + "step": 51444 + }, + { + "epoch": 1.98, + "learning_rate": 1.7046037445498846e-05, + "loss": 0.1961, + "step": 51445 + }, + { + "epoch": 1.98, + "learning_rate": 1.7045396255450118e-05, + "loss": 0.3323, + "step": 51446 + }, + { + "epoch": 1.98, + "learning_rate": 1.7044755065401386e-05, + "loss": 0.3347, + "step": 51447 + }, + { + "epoch": 1.98, + "learning_rate": 1.7044113875352654e-05, + "loss": 0.277, + "step": 51448 + }, + { + "epoch": 1.98, + "learning_rate": 1.7043472685303922e-05, + "loss": 0.3425, + "step": 51449 + }, + { + "epoch": 1.98, + "learning_rate": 1.7042831495255194e-05, + "loss": 0.3727, + "step": 51450 + }, + { + "epoch": 1.98, + "learning_rate": 1.7042190305206466e-05, + "loss": 0.3381, + "step": 51451 + }, + { + "epoch": 1.98, + "learning_rate": 1.7041549115157734e-05, + "loss": 0.2908, + "step": 51452 + }, + { + "epoch": 1.98, + "learning_rate": 1.7040907925109003e-05, + "loss": 0.3194, + "step": 51453 + }, + { + "epoch": 1.98, + "learning_rate": 1.7040266735060274e-05, + "loss": 0.4701, + "step": 51454 + }, + { + "epoch": 1.98, + "learning_rate": 1.7039625545011543e-05, + "loss": 0.1366, + "step": 51455 + }, + { + "epoch": 1.98, + "learning_rate": 1.703898435496281e-05, + "loss": 0.3241, + "step": 51456 + }, + { + "epoch": 1.98, + "learning_rate": 1.703834316491408e-05, + "loss": 0.3318, + "step": 51457 + }, + { + "epoch": 1.98, + "learning_rate": 1.703770197486535e-05, + "loss": 0.3936, + "step": 51458 + }, + { + "epoch": 1.98, + "learning_rate": 1.703706078481662e-05, + "loss": 0.2866, + "step": 51459 + }, + { + "epoch": 1.98, + "learning_rate": 1.703641959476789e-05, + "loss": 0.2865, + "step": 51460 + }, + { + "epoch": 1.98, + "learning_rate": 1.703577840471916e-05, + "loss": 0.389, + "step": 51461 + }, + { + "epoch": 1.98, + "learning_rate": 1.703513721467043e-05, + "loss": 0.3726, + "step": 51462 + }, + { + "epoch": 1.98, + "learning_rate": 1.70344960246217e-05, + "loss": 0.2504, + "step": 51463 + }, + { + "epoch": 1.98, + "learning_rate": 1.7033854834572968e-05, + "loss": 0.3815, + "step": 51464 + }, + { + "epoch": 1.98, + "learning_rate": 1.7033213644524236e-05, + "loss": 0.1853, + "step": 51465 + }, + { + "epoch": 1.98, + "learning_rate": 1.7032572454475508e-05, + "loss": 0.3225, + "step": 51466 + }, + { + "epoch": 1.98, + "learning_rate": 1.7031931264426776e-05, + "loss": 0.3688, + "step": 51467 + }, + { + "epoch": 1.98, + "learning_rate": 1.7031290074378044e-05, + "loss": 0.2411, + "step": 51468 + }, + { + "epoch": 1.98, + "learning_rate": 1.7030648884329316e-05, + "loss": 0.2758, + "step": 51469 + }, + { + "epoch": 1.98, + "learning_rate": 1.7030007694280588e-05, + "loss": 0.2881, + "step": 51470 + }, + { + "epoch": 1.98, + "learning_rate": 1.7029366504231856e-05, + "loss": 0.3092, + "step": 51471 + }, + { + "epoch": 1.98, + "learning_rate": 1.7028725314183125e-05, + "loss": 0.2209, + "step": 51472 + }, + { + "epoch": 1.98, + "learning_rate": 1.7028084124134393e-05, + "loss": 0.3282, + "step": 51473 + }, + { + "epoch": 1.98, + "learning_rate": 1.7027442934085665e-05, + "loss": 0.3618, + "step": 51474 + }, + { + "epoch": 1.98, + "learning_rate": 1.7026801744036933e-05, + "loss": 0.2618, + "step": 51475 + }, + { + "epoch": 1.98, + "learning_rate": 1.70261605539882e-05, + "loss": 0.2697, + "step": 51476 + }, + { + "epoch": 1.98, + "learning_rate": 1.702551936393947e-05, + "loss": 0.4058, + "step": 51477 + }, + { + "epoch": 1.98, + "learning_rate": 1.7024878173890745e-05, + "loss": 0.3989, + "step": 51478 + }, + { + "epoch": 1.98, + "learning_rate": 1.7024236983842013e-05, + "loss": 0.4129, + "step": 51479 + }, + { + "epoch": 1.98, + "learning_rate": 1.702359579379328e-05, + "loss": 0.3101, + "step": 51480 + }, + { + "epoch": 1.98, + "learning_rate": 1.702295460374455e-05, + "loss": 0.4234, + "step": 51481 + }, + { + "epoch": 1.98, + "learning_rate": 1.702231341369582e-05, + "loss": 0.2121, + "step": 51482 + }, + { + "epoch": 1.98, + "learning_rate": 1.702167222364709e-05, + "loss": 0.3093, + "step": 51483 + }, + { + "epoch": 1.98, + "learning_rate": 1.7021031033598358e-05, + "loss": 0.373, + "step": 51484 + }, + { + "epoch": 1.98, + "learning_rate": 1.7020389843549626e-05, + "loss": 0.3164, + "step": 51485 + }, + { + "epoch": 1.98, + "learning_rate": 1.70197486535009e-05, + "loss": 0.2438, + "step": 51486 + }, + { + "epoch": 1.98, + "learning_rate": 1.701910746345217e-05, + "loss": 0.3662, + "step": 51487 + }, + { + "epoch": 1.98, + "learning_rate": 1.7018466273403438e-05, + "loss": 0.2778, + "step": 51488 + }, + { + "epoch": 1.98, + "learning_rate": 1.7017825083354706e-05, + "loss": 0.265, + "step": 51489 + }, + { + "epoch": 1.98, + "learning_rate": 1.7017183893305978e-05, + "loss": 0.3111, + "step": 51490 + }, + { + "epoch": 1.98, + "learning_rate": 1.7016542703257246e-05, + "loss": 0.4251, + "step": 51491 + }, + { + "epoch": 1.98, + "learning_rate": 1.7015901513208515e-05, + "loss": 0.3299, + "step": 51492 + }, + { + "epoch": 1.98, + "learning_rate": 1.7015260323159783e-05, + "loss": 0.2606, + "step": 51493 + }, + { + "epoch": 1.98, + "learning_rate": 1.7014619133111055e-05, + "loss": 0.2268, + "step": 51494 + }, + { + "epoch": 1.98, + "learning_rate": 1.7013977943062327e-05, + "loss": 0.2899, + "step": 51495 + }, + { + "epoch": 1.98, + "learning_rate": 1.7013336753013595e-05, + "loss": 0.3427, + "step": 51496 + }, + { + "epoch": 1.98, + "learning_rate": 1.7012695562964863e-05, + "loss": 0.4016, + "step": 51497 + }, + { + "epoch": 1.98, + "learning_rate": 1.7012054372916135e-05, + "loss": 0.4812, + "step": 51498 + }, + { + "epoch": 1.98, + "learning_rate": 1.7011413182867403e-05, + "loss": 0.2688, + "step": 51499 + }, + { + "epoch": 1.98, + "learning_rate": 1.701077199281867e-05, + "loss": 0.3998, + "step": 51500 + }, + { + "epoch": 1.98, + "learning_rate": 1.701013080276994e-05, + "loss": 0.2481, + "step": 51501 + }, + { + "epoch": 1.98, + "learning_rate": 1.700948961272121e-05, + "loss": 0.2136, + "step": 51502 + }, + { + "epoch": 1.98, + "learning_rate": 1.700884842267248e-05, + "loss": 0.259, + "step": 51503 + }, + { + "epoch": 1.98, + "learning_rate": 1.7008207232623752e-05, + "loss": 0.3319, + "step": 51504 + }, + { + "epoch": 1.98, + "learning_rate": 1.700756604257502e-05, + "loss": 0.3455, + "step": 51505 + }, + { + "epoch": 1.98, + "learning_rate": 1.7006924852526292e-05, + "loss": 0.2158, + "step": 51506 + }, + { + "epoch": 1.98, + "learning_rate": 1.700628366247756e-05, + "loss": 0.1548, + "step": 51507 + }, + { + "epoch": 1.98, + "learning_rate": 1.700564247242883e-05, + "loss": 0.339, + "step": 51508 + }, + { + "epoch": 1.98, + "learning_rate": 1.7005001282380097e-05, + "loss": 0.3896, + "step": 51509 + }, + { + "epoch": 1.98, + "learning_rate": 1.700436009233137e-05, + "loss": 0.3648, + "step": 51510 + }, + { + "epoch": 1.98, + "learning_rate": 1.7003718902282637e-05, + "loss": 0.2817, + "step": 51511 + }, + { + "epoch": 1.98, + "learning_rate": 1.7003077712233905e-05, + "loss": 0.3146, + "step": 51512 + }, + { + "epoch": 1.98, + "learning_rate": 1.7002436522185177e-05, + "loss": 0.3149, + "step": 51513 + }, + { + "epoch": 1.98, + "learning_rate": 1.700179533213645e-05, + "loss": 0.3278, + "step": 51514 + }, + { + "epoch": 1.98, + "learning_rate": 1.7001154142087717e-05, + "loss": 0.3643, + "step": 51515 + }, + { + "epoch": 1.98, + "learning_rate": 1.7000512952038985e-05, + "loss": 0.2363, + "step": 51516 + }, + { + "epoch": 1.98, + "learning_rate": 1.6999871761990254e-05, + "loss": 0.3466, + "step": 51517 + }, + { + "epoch": 1.98, + "learning_rate": 1.6999230571941525e-05, + "loss": 0.3604, + "step": 51518 + }, + { + "epoch": 1.98, + "learning_rate": 1.6998589381892794e-05, + "loss": 0.3151, + "step": 51519 + }, + { + "epoch": 1.98, + "learning_rate": 1.6997948191844062e-05, + "loss": 0.3934, + "step": 51520 + }, + { + "epoch": 1.98, + "learning_rate": 1.699730700179533e-05, + "loss": 0.2034, + "step": 51521 + }, + { + "epoch": 1.98, + "learning_rate": 1.6996665811746605e-05, + "loss": 0.3397, + "step": 51522 + }, + { + "epoch": 1.98, + "learning_rate": 1.6996024621697874e-05, + "loss": 0.322, + "step": 51523 + }, + { + "epoch": 1.98, + "learning_rate": 1.6995383431649142e-05, + "loss": 0.381, + "step": 51524 + }, + { + "epoch": 1.98, + "learning_rate": 1.699474224160041e-05, + "loss": 0.2726, + "step": 51525 + }, + { + "epoch": 1.98, + "learning_rate": 1.6994101051551682e-05, + "loss": 0.3543, + "step": 51526 + }, + { + "epoch": 1.98, + "learning_rate": 1.699345986150295e-05, + "loss": 0.1791, + "step": 51527 + }, + { + "epoch": 1.98, + "learning_rate": 1.699281867145422e-05, + "loss": 0.3557, + "step": 51528 + }, + { + "epoch": 1.98, + "learning_rate": 1.6992177481405487e-05, + "loss": 0.4422, + "step": 51529 + }, + { + "epoch": 1.98, + "learning_rate": 1.699153629135676e-05, + "loss": 0.1737, + "step": 51530 + }, + { + "epoch": 1.98, + "learning_rate": 1.699089510130803e-05, + "loss": 0.4137, + "step": 51531 + }, + { + "epoch": 1.98, + "learning_rate": 1.69902539112593e-05, + "loss": 0.3268, + "step": 51532 + }, + { + "epoch": 1.98, + "learning_rate": 1.6989612721210567e-05, + "loss": 0.2785, + "step": 51533 + }, + { + "epoch": 1.98, + "learning_rate": 1.698897153116184e-05, + "loss": 0.3489, + "step": 51534 + }, + { + "epoch": 1.98, + "learning_rate": 1.6988330341113107e-05, + "loss": 0.2845, + "step": 51535 + }, + { + "epoch": 1.98, + "learning_rate": 1.6987689151064375e-05, + "loss": 0.3384, + "step": 51536 + }, + { + "epoch": 1.98, + "learning_rate": 1.6987047961015644e-05, + "loss": 0.3358, + "step": 51537 + }, + { + "epoch": 1.98, + "learning_rate": 1.6986406770966916e-05, + "loss": 0.2571, + "step": 51538 + }, + { + "epoch": 1.98, + "learning_rate": 1.6985765580918184e-05, + "loss": 0.3763, + "step": 51539 + }, + { + "epoch": 1.98, + "learning_rate": 1.6985124390869456e-05, + "loss": 0.2232, + "step": 51540 + }, + { + "epoch": 1.98, + "learning_rate": 1.6984483200820724e-05, + "loss": 0.3155, + "step": 51541 + }, + { + "epoch": 1.98, + "learning_rate": 1.6983842010771996e-05, + "loss": 0.4261, + "step": 51542 + }, + { + "epoch": 1.98, + "learning_rate": 1.6983200820723264e-05, + "loss": 0.3346, + "step": 51543 + }, + { + "epoch": 1.98, + "learning_rate": 1.6982559630674532e-05, + "loss": 0.3025, + "step": 51544 + }, + { + "epoch": 1.98, + "learning_rate": 1.69819184406258e-05, + "loss": 0.3061, + "step": 51545 + }, + { + "epoch": 1.98, + "learning_rate": 1.6981277250577072e-05, + "loss": 0.5518, + "step": 51546 + }, + { + "epoch": 1.98, + "learning_rate": 1.698063606052834e-05, + "loss": 0.4227, + "step": 51547 + }, + { + "epoch": 1.98, + "learning_rate": 1.697999487047961e-05, + "loss": 0.2958, + "step": 51548 + }, + { + "epoch": 1.98, + "learning_rate": 1.697935368043088e-05, + "loss": 0.2503, + "step": 51549 + }, + { + "epoch": 1.98, + "learning_rate": 1.6978712490382152e-05, + "loss": 0.3101, + "step": 51550 + }, + { + "epoch": 1.98, + "learning_rate": 1.697807130033342e-05, + "loss": 0.3634, + "step": 51551 + }, + { + "epoch": 1.98, + "learning_rate": 1.697743011028469e-05, + "loss": 0.22, + "step": 51552 + }, + { + "epoch": 1.98, + "learning_rate": 1.6976788920235957e-05, + "loss": 0.6298, + "step": 51553 + }, + { + "epoch": 1.98, + "learning_rate": 1.697614773018723e-05, + "loss": 0.3604, + "step": 51554 + }, + { + "epoch": 1.98, + "learning_rate": 1.6975506540138497e-05, + "loss": 0.2357, + "step": 51555 + }, + { + "epoch": 1.98, + "learning_rate": 1.6974865350089766e-05, + "loss": 0.3999, + "step": 51556 + }, + { + "epoch": 1.98, + "learning_rate": 1.6974224160041034e-05, + "loss": 0.3652, + "step": 51557 + }, + { + "epoch": 1.98, + "learning_rate": 1.697358296999231e-05, + "loss": 0.3719, + "step": 51558 + }, + { + "epoch": 1.98, + "learning_rate": 1.6972941779943578e-05, + "loss": 0.2174, + "step": 51559 + }, + { + "epoch": 1.98, + "learning_rate": 1.6972300589894846e-05, + "loss": 0.3176, + "step": 51560 + }, + { + "epoch": 1.98, + "learning_rate": 1.6971659399846114e-05, + "loss": 0.2027, + "step": 51561 + }, + { + "epoch": 1.98, + "learning_rate": 1.6971018209797386e-05, + "loss": 0.2719, + "step": 51562 + }, + { + "epoch": 1.98, + "learning_rate": 1.6970377019748654e-05, + "loss": 0.3138, + "step": 51563 + }, + { + "epoch": 1.98, + "learning_rate": 1.6969735829699923e-05, + "loss": 0.2213, + "step": 51564 + }, + { + "epoch": 1.98, + "learning_rate": 1.696909463965119e-05, + "loss": 0.3906, + "step": 51565 + }, + { + "epoch": 1.98, + "learning_rate": 1.6968453449602463e-05, + "loss": 0.2644, + "step": 51566 + }, + { + "epoch": 1.98, + "learning_rate": 1.6967812259553734e-05, + "loss": 0.3877, + "step": 51567 + }, + { + "epoch": 1.98, + "learning_rate": 1.6967171069505003e-05, + "loss": 0.322, + "step": 51568 + }, + { + "epoch": 1.98, + "learning_rate": 1.696652987945627e-05, + "loss": 0.2719, + "step": 51569 + }, + { + "epoch": 1.98, + "learning_rate": 1.6965888689407543e-05, + "loss": 0.301, + "step": 51570 + }, + { + "epoch": 1.98, + "learning_rate": 1.696524749935881e-05, + "loss": 0.2908, + "step": 51571 + }, + { + "epoch": 1.98, + "learning_rate": 1.696460630931008e-05, + "loss": 0.2818, + "step": 51572 + }, + { + "epoch": 1.98, + "learning_rate": 1.6963965119261348e-05, + "loss": 0.3028, + "step": 51573 + }, + { + "epoch": 1.98, + "learning_rate": 1.696332392921262e-05, + "loss": 0.2063, + "step": 51574 + }, + { + "epoch": 1.98, + "learning_rate": 1.6962682739163888e-05, + "loss": 0.3154, + "step": 51575 + }, + { + "epoch": 1.98, + "learning_rate": 1.696204154911516e-05, + "loss": 0.2641, + "step": 51576 + }, + { + "epoch": 1.98, + "learning_rate": 1.6961400359066428e-05, + "loss": 0.2268, + "step": 51577 + }, + { + "epoch": 1.98, + "learning_rate": 1.69607591690177e-05, + "loss": 0.3651, + "step": 51578 + }, + { + "epoch": 1.98, + "learning_rate": 1.6960117978968968e-05, + "loss": 0.3981, + "step": 51579 + }, + { + "epoch": 1.98, + "learning_rate": 1.6959476788920236e-05, + "loss": 0.3659, + "step": 51580 + }, + { + "epoch": 1.98, + "learning_rate": 1.6958835598871504e-05, + "loss": 0.271, + "step": 51581 + }, + { + "epoch": 1.98, + "learning_rate": 1.6958194408822776e-05, + "loss": 0.3728, + "step": 51582 + }, + { + "epoch": 1.98, + "learning_rate": 1.6957553218774045e-05, + "loss": 0.2921, + "step": 51583 + }, + { + "epoch": 1.98, + "learning_rate": 1.6956912028725316e-05, + "loss": 0.1941, + "step": 51584 + }, + { + "epoch": 1.98, + "learning_rate": 1.6956270838676585e-05, + "loss": 0.2717, + "step": 51585 + }, + { + "epoch": 1.98, + "learning_rate": 1.6955629648627856e-05, + "loss": 0.321, + "step": 51586 + }, + { + "epoch": 1.98, + "learning_rate": 1.6954988458579125e-05, + "loss": 0.261, + "step": 51587 + }, + { + "epoch": 1.98, + "learning_rate": 1.6954347268530393e-05, + "loss": 0.3453, + "step": 51588 + }, + { + "epoch": 1.98, + "learning_rate": 1.695370607848166e-05, + "loss": 0.2965, + "step": 51589 + }, + { + "epoch": 1.98, + "learning_rate": 1.6953064888432933e-05, + "loss": 0.3816, + "step": 51590 + }, + { + "epoch": 1.98, + "learning_rate": 1.69524236983842e-05, + "loss": 0.2355, + "step": 51591 + }, + { + "epoch": 1.98, + "learning_rate": 1.695178250833547e-05, + "loss": 0.2335, + "step": 51592 + }, + { + "epoch": 1.98, + "learning_rate": 1.695114131828674e-05, + "loss": 0.2104, + "step": 51593 + }, + { + "epoch": 1.98, + "learning_rate": 1.6950500128238013e-05, + "loss": 0.3305, + "step": 51594 + }, + { + "epoch": 1.98, + "learning_rate": 1.694985893818928e-05, + "loss": 0.2126, + "step": 51595 + }, + { + "epoch": 1.98, + "learning_rate": 1.694921774814055e-05, + "loss": 0.3024, + "step": 51596 + }, + { + "epoch": 1.98, + "learning_rate": 1.6948576558091818e-05, + "loss": 0.3959, + "step": 51597 + }, + { + "epoch": 1.98, + "learning_rate": 1.694793536804309e-05, + "loss": 0.2345, + "step": 51598 + }, + { + "epoch": 1.98, + "learning_rate": 1.6947294177994358e-05, + "loss": 0.3055, + "step": 51599 + }, + { + "epoch": 1.98, + "learning_rate": 1.6946652987945626e-05, + "loss": 0.4622, + "step": 51600 + }, + { + "epoch": 1.98, + "learning_rate": 1.6946011797896895e-05, + "loss": 0.1924, + "step": 51601 + }, + { + "epoch": 1.98, + "learning_rate": 1.694537060784817e-05, + "loss": 0.4703, + "step": 51602 + }, + { + "epoch": 1.98, + "learning_rate": 1.6944729417799438e-05, + "loss": 0.2789, + "step": 51603 + }, + { + "epoch": 1.98, + "learning_rate": 1.6944088227750707e-05, + "loss": 0.2032, + "step": 51604 + }, + { + "epoch": 1.98, + "learning_rate": 1.6943447037701975e-05, + "loss": 0.1614, + "step": 51605 + }, + { + "epoch": 1.98, + "learning_rate": 1.6942805847653247e-05, + "loss": 0.2698, + "step": 51606 + }, + { + "epoch": 1.98, + "learning_rate": 1.6942164657604515e-05, + "loss": 0.2485, + "step": 51607 + }, + { + "epoch": 1.98, + "learning_rate": 1.6941523467555783e-05, + "loss": 0.3649, + "step": 51608 + }, + { + "epoch": 1.98, + "learning_rate": 1.694088227750705e-05, + "loss": 0.3315, + "step": 51609 + }, + { + "epoch": 1.98, + "learning_rate": 1.6940241087458323e-05, + "loss": 0.2136, + "step": 51610 + }, + { + "epoch": 1.98, + "learning_rate": 1.6939599897409595e-05, + "loss": 0.4041, + "step": 51611 + }, + { + "epoch": 1.98, + "learning_rate": 1.6938958707360863e-05, + "loss": 0.2527, + "step": 51612 + }, + { + "epoch": 1.98, + "learning_rate": 1.693831751731213e-05, + "loss": 0.2619, + "step": 51613 + }, + { + "epoch": 1.98, + "learning_rate": 1.6937676327263403e-05, + "loss": 0.3969, + "step": 51614 + }, + { + "epoch": 1.98, + "learning_rate": 1.6937035137214672e-05, + "loss": 0.2229, + "step": 51615 + }, + { + "epoch": 1.98, + "learning_rate": 1.693639394716594e-05, + "loss": 0.2856, + "step": 51616 + }, + { + "epoch": 1.98, + "learning_rate": 1.693575275711721e-05, + "loss": 0.3943, + "step": 51617 + }, + { + "epoch": 1.98, + "learning_rate": 1.693511156706848e-05, + "loss": 0.4652, + "step": 51618 + }, + { + "epoch": 1.98, + "learning_rate": 1.693447037701975e-05, + "loss": 0.2673, + "step": 51619 + }, + { + "epoch": 1.98, + "learning_rate": 1.693382918697102e-05, + "loss": 0.3313, + "step": 51620 + }, + { + "epoch": 1.98, + "learning_rate": 1.693318799692229e-05, + "loss": 0.4208, + "step": 51621 + }, + { + "epoch": 1.98, + "learning_rate": 1.693254680687356e-05, + "loss": 0.3434, + "step": 51622 + }, + { + "epoch": 1.98, + "learning_rate": 1.693190561682483e-05, + "loss": 0.4071, + "step": 51623 + }, + { + "epoch": 1.98, + "learning_rate": 1.6931264426776097e-05, + "loss": 0.3306, + "step": 51624 + }, + { + "epoch": 1.98, + "learning_rate": 1.6930623236727365e-05, + "loss": 0.4851, + "step": 51625 + }, + { + "epoch": 1.98, + "learning_rate": 1.6929982046678637e-05, + "loss": 0.1854, + "step": 51626 + }, + { + "epoch": 1.98, + "learning_rate": 1.6929340856629905e-05, + "loss": 0.327, + "step": 51627 + }, + { + "epoch": 1.98, + "learning_rate": 1.6928699666581174e-05, + "loss": 0.3152, + "step": 51628 + }, + { + "epoch": 1.98, + "learning_rate": 1.6928058476532445e-05, + "loss": 0.2419, + "step": 51629 + }, + { + "epoch": 1.99, + "learning_rate": 1.6927417286483717e-05, + "loss": 0.2786, + "step": 51630 + }, + { + "epoch": 1.99, + "learning_rate": 1.6926776096434985e-05, + "loss": 0.3159, + "step": 51631 + }, + { + "epoch": 1.99, + "learning_rate": 1.6926134906386254e-05, + "loss": 0.256, + "step": 51632 + }, + { + "epoch": 1.99, + "learning_rate": 1.6925493716337522e-05, + "loss": 0.2767, + "step": 51633 + }, + { + "epoch": 1.99, + "learning_rate": 1.6924852526288794e-05, + "loss": 0.3723, + "step": 51634 + }, + { + "epoch": 1.99, + "learning_rate": 1.6924211336240062e-05, + "loss": 0.2763, + "step": 51635 + }, + { + "epoch": 1.99, + "learning_rate": 1.692357014619133e-05, + "loss": 0.3861, + "step": 51636 + }, + { + "epoch": 1.99, + "learning_rate": 1.69229289561426e-05, + "loss": 0.3583, + "step": 51637 + }, + { + "epoch": 1.99, + "learning_rate": 1.6922287766093874e-05, + "loss": 0.3675, + "step": 51638 + }, + { + "epoch": 1.99, + "learning_rate": 1.6921646576045142e-05, + "loss": 0.3178, + "step": 51639 + }, + { + "epoch": 1.99, + "learning_rate": 1.692100538599641e-05, + "loss": 0.3963, + "step": 51640 + }, + { + "epoch": 1.99, + "learning_rate": 1.692036419594768e-05, + "loss": 0.2404, + "step": 51641 + }, + { + "epoch": 1.99, + "learning_rate": 1.691972300589895e-05, + "loss": 0.2074, + "step": 51642 + }, + { + "epoch": 1.99, + "learning_rate": 1.691908181585022e-05, + "loss": 0.2505, + "step": 51643 + }, + { + "epoch": 1.99, + "learning_rate": 1.6918440625801487e-05, + "loss": 0.2611, + "step": 51644 + }, + { + "epoch": 1.99, + "learning_rate": 1.6917799435752755e-05, + "loss": 0.1912, + "step": 51645 + }, + { + "epoch": 1.99, + "learning_rate": 1.6917158245704027e-05, + "loss": 0.4196, + "step": 51646 + }, + { + "epoch": 1.99, + "learning_rate": 1.69165170556553e-05, + "loss": 0.2624, + "step": 51647 + }, + { + "epoch": 1.99, + "learning_rate": 1.6915875865606567e-05, + "loss": 0.2605, + "step": 51648 + }, + { + "epoch": 1.99, + "learning_rate": 1.6915234675557836e-05, + "loss": 0.272, + "step": 51649 + }, + { + "epoch": 1.99, + "learning_rate": 1.6914593485509107e-05, + "loss": 0.2928, + "step": 51650 + }, + { + "epoch": 1.99, + "learning_rate": 1.6913952295460376e-05, + "loss": 0.2931, + "step": 51651 + }, + { + "epoch": 1.99, + "learning_rate": 1.6913311105411644e-05, + "loss": 0.1593, + "step": 51652 + }, + { + "epoch": 1.99, + "learning_rate": 1.6912669915362912e-05, + "loss": 0.2971, + "step": 51653 + }, + { + "epoch": 1.99, + "learning_rate": 1.6912028725314184e-05, + "loss": 0.2617, + "step": 51654 + }, + { + "epoch": 1.99, + "learning_rate": 1.6911387535265452e-05, + "loss": 0.2693, + "step": 51655 + }, + { + "epoch": 1.99, + "learning_rate": 1.6910746345216724e-05, + "loss": 0.2839, + "step": 51656 + }, + { + "epoch": 1.99, + "learning_rate": 1.6910105155167992e-05, + "loss": 0.181, + "step": 51657 + }, + { + "epoch": 1.99, + "learning_rate": 1.6909463965119264e-05, + "loss": 0.2418, + "step": 51658 + }, + { + "epoch": 1.99, + "learning_rate": 1.6908822775070532e-05, + "loss": 0.3268, + "step": 51659 + }, + { + "epoch": 1.99, + "learning_rate": 1.69081815850218e-05, + "loss": 0.3273, + "step": 51660 + }, + { + "epoch": 1.99, + "learning_rate": 1.690754039497307e-05, + "loss": 0.3804, + "step": 51661 + }, + { + "epoch": 1.99, + "learning_rate": 1.690689920492434e-05, + "loss": 0.2417, + "step": 51662 + }, + { + "epoch": 1.99, + "learning_rate": 1.690625801487561e-05, + "loss": 0.3821, + "step": 51663 + }, + { + "epoch": 1.99, + "learning_rate": 1.6905616824826877e-05, + "loss": 0.2955, + "step": 51664 + }, + { + "epoch": 1.99, + "learning_rate": 1.690497563477815e-05, + "loss": 0.3175, + "step": 51665 + }, + { + "epoch": 1.99, + "learning_rate": 1.690433444472942e-05, + "loss": 0.3041, + "step": 51666 + }, + { + "epoch": 1.99, + "learning_rate": 1.690369325468069e-05, + "loss": 0.3807, + "step": 51667 + }, + { + "epoch": 1.99, + "learning_rate": 1.6903052064631957e-05, + "loss": 0.353, + "step": 51668 + }, + { + "epoch": 1.99, + "learning_rate": 1.6902410874583226e-05, + "loss": 0.5078, + "step": 51669 + }, + { + "epoch": 1.99, + "learning_rate": 1.6901769684534498e-05, + "loss": 0.3924, + "step": 51670 + }, + { + "epoch": 1.99, + "learning_rate": 1.6901128494485766e-05, + "loss": 0.3651, + "step": 51671 + }, + { + "epoch": 1.99, + "learning_rate": 1.6900487304437034e-05, + "loss": 0.4048, + "step": 51672 + }, + { + "epoch": 1.99, + "learning_rate": 1.6899846114388303e-05, + "loss": 0.4059, + "step": 51673 + }, + { + "epoch": 1.99, + "learning_rate": 1.6899204924339578e-05, + "loss": 0.3001, + "step": 51674 + }, + { + "epoch": 1.99, + "learning_rate": 1.6898563734290846e-05, + "loss": 0.2025, + "step": 51675 + }, + { + "epoch": 1.99, + "learning_rate": 1.6897922544242114e-05, + "loss": 0.3905, + "step": 51676 + }, + { + "epoch": 1.99, + "learning_rate": 1.6897281354193383e-05, + "loss": 0.3102, + "step": 51677 + }, + { + "epoch": 1.99, + "learning_rate": 1.6896640164144654e-05, + "loss": 0.2659, + "step": 51678 + }, + { + "epoch": 1.99, + "learning_rate": 1.6895998974095923e-05, + "loss": 0.512, + "step": 51679 + }, + { + "epoch": 1.99, + "learning_rate": 1.689535778404719e-05, + "loss": 0.2526, + "step": 51680 + }, + { + "epoch": 1.99, + "learning_rate": 1.689471659399846e-05, + "loss": 0.4925, + "step": 51681 + }, + { + "epoch": 1.99, + "learning_rate": 1.6894075403949734e-05, + "loss": 0.2201, + "step": 51682 + }, + { + "epoch": 1.99, + "learning_rate": 1.6893434213901003e-05, + "loss": 0.2749, + "step": 51683 + }, + { + "epoch": 1.99, + "learning_rate": 1.689279302385227e-05, + "loss": 0.2842, + "step": 51684 + }, + { + "epoch": 1.99, + "learning_rate": 1.689215183380354e-05, + "loss": 0.2341, + "step": 51685 + }, + { + "epoch": 1.99, + "learning_rate": 1.689151064375481e-05, + "loss": 0.2805, + "step": 51686 + }, + { + "epoch": 1.99, + "learning_rate": 1.689086945370608e-05, + "loss": 0.23, + "step": 51687 + }, + { + "epoch": 1.99, + "learning_rate": 1.6890228263657348e-05, + "loss": 0.2997, + "step": 51688 + }, + { + "epoch": 1.99, + "learning_rate": 1.6889587073608616e-05, + "loss": 0.2514, + "step": 51689 + }, + { + "epoch": 1.99, + "learning_rate": 1.6888945883559888e-05, + "loss": 0.2419, + "step": 51690 + }, + { + "epoch": 1.99, + "learning_rate": 1.688830469351116e-05, + "loss": 0.1961, + "step": 51691 + }, + { + "epoch": 1.99, + "learning_rate": 1.6887663503462428e-05, + "loss": 0.3473, + "step": 51692 + }, + { + "epoch": 1.99, + "learning_rate": 1.6887022313413696e-05, + "loss": 0.3053, + "step": 51693 + }, + { + "epoch": 1.99, + "learning_rate": 1.6886381123364968e-05, + "loss": 0.3835, + "step": 51694 + }, + { + "epoch": 1.99, + "learning_rate": 1.6885739933316236e-05, + "loss": 0.4394, + "step": 51695 + }, + { + "epoch": 1.99, + "learning_rate": 1.6885098743267505e-05, + "loss": 0.3763, + "step": 51696 + }, + { + "epoch": 1.99, + "learning_rate": 1.6884457553218773e-05, + "loss": 0.4064, + "step": 51697 + }, + { + "epoch": 1.99, + "learning_rate": 1.6883816363170045e-05, + "loss": 0.1811, + "step": 51698 + }, + { + "epoch": 1.99, + "learning_rate": 1.6883175173121313e-05, + "loss": 0.3639, + "step": 51699 + }, + { + "epoch": 1.99, + "learning_rate": 1.6882533983072585e-05, + "loss": 0.2533, + "step": 51700 + }, + { + "epoch": 1.99, + "learning_rate": 1.6881892793023853e-05, + "loss": 0.2596, + "step": 51701 + }, + { + "epoch": 1.99, + "learning_rate": 1.6881251602975125e-05, + "loss": 0.4883, + "step": 51702 + }, + { + "epoch": 1.99, + "learning_rate": 1.6880610412926393e-05, + "loss": 0.2715, + "step": 51703 + }, + { + "epoch": 1.99, + "learning_rate": 1.687996922287766e-05, + "loss": 0.3529, + "step": 51704 + }, + { + "epoch": 1.99, + "learning_rate": 1.687932803282893e-05, + "loss": 0.3294, + "step": 51705 + }, + { + "epoch": 1.99, + "learning_rate": 1.68786868427802e-05, + "loss": 0.4867, + "step": 51706 + }, + { + "epoch": 1.99, + "learning_rate": 1.687804565273147e-05, + "loss": 0.4362, + "step": 51707 + }, + { + "epoch": 1.99, + "learning_rate": 1.6877404462682738e-05, + "loss": 0.4035, + "step": 51708 + }, + { + "epoch": 1.99, + "learning_rate": 1.687676327263401e-05, + "loss": 0.2758, + "step": 51709 + }, + { + "epoch": 1.99, + "learning_rate": 1.687612208258528e-05, + "loss": 0.3656, + "step": 51710 + }, + { + "epoch": 1.99, + "learning_rate": 1.687548089253655e-05, + "loss": 0.3045, + "step": 51711 + }, + { + "epoch": 1.99, + "learning_rate": 1.6874839702487818e-05, + "loss": 0.3971, + "step": 51712 + }, + { + "epoch": 1.99, + "learning_rate": 1.6874198512439086e-05, + "loss": 0.4066, + "step": 51713 + }, + { + "epoch": 1.99, + "learning_rate": 1.6873557322390358e-05, + "loss": 0.221, + "step": 51714 + }, + { + "epoch": 1.99, + "learning_rate": 1.6872916132341627e-05, + "loss": 0.4221, + "step": 51715 + }, + { + "epoch": 1.99, + "learning_rate": 1.6872274942292895e-05, + "loss": 0.3683, + "step": 51716 + }, + { + "epoch": 1.99, + "learning_rate": 1.6871633752244163e-05, + "loss": 0.4225, + "step": 51717 + }, + { + "epoch": 1.99, + "learning_rate": 1.6870992562195438e-05, + "loss": 0.3089, + "step": 51718 + }, + { + "epoch": 1.99, + "learning_rate": 1.6870351372146707e-05, + "loss": 0.3145, + "step": 51719 + }, + { + "epoch": 1.99, + "learning_rate": 1.6869710182097975e-05, + "loss": 0.38, + "step": 51720 + }, + { + "epoch": 1.99, + "learning_rate": 1.6869068992049243e-05, + "loss": 0.3532, + "step": 51721 + }, + { + "epoch": 1.99, + "learning_rate": 1.6868427802000515e-05, + "loss": 0.2363, + "step": 51722 + }, + { + "epoch": 1.99, + "learning_rate": 1.6867786611951783e-05, + "loss": 0.3032, + "step": 51723 + }, + { + "epoch": 1.99, + "learning_rate": 1.686714542190305e-05, + "loss": 0.1985, + "step": 51724 + }, + { + "epoch": 1.99, + "learning_rate": 1.686650423185432e-05, + "loss": 0.2245, + "step": 51725 + }, + { + "epoch": 1.99, + "learning_rate": 1.686586304180559e-05, + "loss": 0.238, + "step": 51726 + }, + { + "epoch": 1.99, + "learning_rate": 1.6865221851756863e-05, + "loss": 0.2684, + "step": 51727 + }, + { + "epoch": 1.99, + "learning_rate": 1.6864580661708132e-05, + "loss": 0.2716, + "step": 51728 + }, + { + "epoch": 1.99, + "learning_rate": 1.68639394716594e-05, + "loss": 0.3872, + "step": 51729 + }, + { + "epoch": 1.99, + "learning_rate": 1.6863298281610672e-05, + "loss": 0.4413, + "step": 51730 + }, + { + "epoch": 1.99, + "learning_rate": 1.686265709156194e-05, + "loss": 0.2699, + "step": 51731 + }, + { + "epoch": 1.99, + "learning_rate": 1.686201590151321e-05, + "loss": 0.2995, + "step": 51732 + }, + { + "epoch": 1.99, + "learning_rate": 1.6861374711464477e-05, + "loss": 0.3607, + "step": 51733 + }, + { + "epoch": 1.99, + "learning_rate": 1.686073352141575e-05, + "loss": 0.3725, + "step": 51734 + }, + { + "epoch": 1.99, + "learning_rate": 1.6860092331367017e-05, + "loss": 0.2867, + "step": 51735 + }, + { + "epoch": 1.99, + "learning_rate": 1.685945114131829e-05, + "loss": 0.2581, + "step": 51736 + }, + { + "epoch": 1.99, + "learning_rate": 1.6858809951269557e-05, + "loss": 0.2235, + "step": 51737 + }, + { + "epoch": 1.99, + "learning_rate": 1.685816876122083e-05, + "loss": 0.3421, + "step": 51738 + }, + { + "epoch": 1.99, + "learning_rate": 1.6857527571172097e-05, + "loss": 0.2154, + "step": 51739 + }, + { + "epoch": 1.99, + "learning_rate": 1.6856886381123365e-05, + "loss": 0.2881, + "step": 51740 + }, + { + "epoch": 1.99, + "learning_rate": 1.6856245191074634e-05, + "loss": 0.3897, + "step": 51741 + }, + { + "epoch": 1.99, + "learning_rate": 1.6855604001025905e-05, + "loss": 0.2649, + "step": 51742 + }, + { + "epoch": 1.99, + "learning_rate": 1.6854962810977174e-05, + "loss": 0.2562, + "step": 51743 + }, + { + "epoch": 1.99, + "learning_rate": 1.6854321620928442e-05, + "loss": 0.2112, + "step": 51744 + }, + { + "epoch": 1.99, + "learning_rate": 1.6853680430879714e-05, + "loss": 0.1749, + "step": 51745 + }, + { + "epoch": 1.99, + "learning_rate": 1.6853039240830985e-05, + "loss": 0.2731, + "step": 51746 + }, + { + "epoch": 1.99, + "learning_rate": 1.6852398050782254e-05, + "loss": 0.2425, + "step": 51747 + }, + { + "epoch": 1.99, + "learning_rate": 1.6851756860733522e-05, + "loss": 0.2699, + "step": 51748 + }, + { + "epoch": 1.99, + "learning_rate": 1.685111567068479e-05, + "loss": 0.258, + "step": 51749 + }, + { + "epoch": 1.99, + "learning_rate": 1.6850474480636062e-05, + "loss": 0.3934, + "step": 51750 + }, + { + "epoch": 1.99, + "learning_rate": 1.684983329058733e-05, + "loss": 0.2886, + "step": 51751 + }, + { + "epoch": 1.99, + "learning_rate": 1.68491921005386e-05, + "loss": 0.3335, + "step": 51752 + }, + { + "epoch": 1.99, + "learning_rate": 1.6848550910489867e-05, + "loss": 0.3256, + "step": 51753 + }, + { + "epoch": 1.99, + "learning_rate": 1.6847909720441142e-05, + "loss": 0.4377, + "step": 51754 + }, + { + "epoch": 1.99, + "learning_rate": 1.684726853039241e-05, + "loss": 0.4876, + "step": 51755 + }, + { + "epoch": 1.99, + "learning_rate": 1.684662734034368e-05, + "loss": 0.2788, + "step": 51756 + }, + { + "epoch": 1.99, + "learning_rate": 1.6845986150294947e-05, + "loss": 0.3669, + "step": 51757 + }, + { + "epoch": 1.99, + "learning_rate": 1.684534496024622e-05, + "loss": 0.361, + "step": 51758 + }, + { + "epoch": 1.99, + "learning_rate": 1.6844703770197487e-05, + "loss": 0.3741, + "step": 51759 + }, + { + "epoch": 1.99, + "learning_rate": 1.6844062580148756e-05, + "loss": 0.4683, + "step": 51760 + }, + { + "epoch": 1.99, + "learning_rate": 1.6843421390100024e-05, + "loss": 0.4184, + "step": 51761 + }, + { + "epoch": 1.99, + "learning_rate": 1.6842780200051296e-05, + "loss": 0.375, + "step": 51762 + }, + { + "epoch": 1.99, + "learning_rate": 1.6842139010002567e-05, + "loss": 0.2198, + "step": 51763 + }, + { + "epoch": 1.99, + "learning_rate": 1.6841497819953836e-05, + "loss": 0.2817, + "step": 51764 + }, + { + "epoch": 1.99, + "learning_rate": 1.6840856629905104e-05, + "loss": 0.2819, + "step": 51765 + }, + { + "epoch": 1.99, + "learning_rate": 1.6840215439856376e-05, + "loss": 0.26, + "step": 51766 + }, + { + "epoch": 1.99, + "learning_rate": 1.6839574249807644e-05, + "loss": 0.1798, + "step": 51767 + }, + { + "epoch": 1.99, + "learning_rate": 1.6838933059758912e-05, + "loss": 0.242, + "step": 51768 + }, + { + "epoch": 1.99, + "learning_rate": 1.683829186971018e-05, + "loss": 0.434, + "step": 51769 + }, + { + "epoch": 1.99, + "learning_rate": 1.6837650679661452e-05, + "loss": 0.332, + "step": 51770 + }, + { + "epoch": 1.99, + "learning_rate": 1.6837009489612724e-05, + "loss": 0.3013, + "step": 51771 + }, + { + "epoch": 1.99, + "learning_rate": 1.6836368299563992e-05, + "loss": 0.2388, + "step": 51772 + }, + { + "epoch": 1.99, + "learning_rate": 1.683572710951526e-05, + "loss": 0.278, + "step": 51773 + }, + { + "epoch": 1.99, + "learning_rate": 1.6835085919466532e-05, + "loss": 0.368, + "step": 51774 + }, + { + "epoch": 1.99, + "learning_rate": 1.68344447294178e-05, + "loss": 0.2798, + "step": 51775 + }, + { + "epoch": 1.99, + "learning_rate": 1.683380353936907e-05, + "loss": 0.3641, + "step": 51776 + }, + { + "epoch": 1.99, + "learning_rate": 1.6833162349320337e-05, + "loss": 0.3403, + "step": 51777 + }, + { + "epoch": 1.99, + "learning_rate": 1.683252115927161e-05, + "loss": 0.3669, + "step": 51778 + }, + { + "epoch": 1.99, + "learning_rate": 1.6831879969222877e-05, + "loss": 0.2792, + "step": 51779 + }, + { + "epoch": 1.99, + "learning_rate": 1.683123877917415e-05, + "loss": 0.3215, + "step": 51780 + }, + { + "epoch": 1.99, + "learning_rate": 1.6830597589125418e-05, + "loss": 0.4074, + "step": 51781 + }, + { + "epoch": 1.99, + "learning_rate": 1.682995639907669e-05, + "loss": 0.3635, + "step": 51782 + }, + { + "epoch": 1.99, + "learning_rate": 1.6829315209027958e-05, + "loss": 0.4024, + "step": 51783 + }, + { + "epoch": 1.99, + "learning_rate": 1.6828674018979226e-05, + "loss": 0.4087, + "step": 51784 + }, + { + "epoch": 1.99, + "learning_rate": 1.6828032828930494e-05, + "loss": 0.3382, + "step": 51785 + }, + { + "epoch": 1.99, + "learning_rate": 1.6827391638881766e-05, + "loss": 0.3545, + "step": 51786 + }, + { + "epoch": 1.99, + "learning_rate": 1.6826750448833034e-05, + "loss": 0.4152, + "step": 51787 + }, + { + "epoch": 1.99, + "learning_rate": 1.6826109258784303e-05, + "loss": 0.3204, + "step": 51788 + }, + { + "epoch": 1.99, + "learning_rate": 1.6825468068735574e-05, + "loss": 0.1869, + "step": 51789 + }, + { + "epoch": 1.99, + "learning_rate": 1.6824826878686846e-05, + "loss": 0.3728, + "step": 51790 + }, + { + "epoch": 1.99, + "learning_rate": 1.6824185688638114e-05, + "loss": 0.4062, + "step": 51791 + }, + { + "epoch": 1.99, + "learning_rate": 1.6823544498589383e-05, + "loss": 0.3779, + "step": 51792 + }, + { + "epoch": 1.99, + "learning_rate": 1.682290330854065e-05, + "loss": 0.2544, + "step": 51793 + }, + { + "epoch": 1.99, + "learning_rate": 1.6822262118491923e-05, + "loss": 0.2493, + "step": 51794 + }, + { + "epoch": 1.99, + "learning_rate": 1.682162092844319e-05, + "loss": 0.2678, + "step": 51795 + }, + { + "epoch": 1.99, + "learning_rate": 1.682097973839446e-05, + "loss": 0.3244, + "step": 51796 + }, + { + "epoch": 1.99, + "learning_rate": 1.6820338548345728e-05, + "loss": 0.2769, + "step": 51797 + }, + { + "epoch": 1.99, + "learning_rate": 1.6819697358297003e-05, + "loss": 0.3277, + "step": 51798 + }, + { + "epoch": 1.99, + "learning_rate": 1.681905616824827e-05, + "loss": 0.3625, + "step": 51799 + }, + { + "epoch": 1.99, + "learning_rate": 1.681841497819954e-05, + "loss": 0.2799, + "step": 51800 + }, + { + "epoch": 1.99, + "learning_rate": 1.6817773788150808e-05, + "loss": 0.4589, + "step": 51801 + }, + { + "epoch": 1.99, + "learning_rate": 1.681713259810208e-05, + "loss": 0.3133, + "step": 51802 + }, + { + "epoch": 1.99, + "learning_rate": 1.6816491408053348e-05, + "loss": 0.2903, + "step": 51803 + }, + { + "epoch": 1.99, + "learning_rate": 1.6815850218004616e-05, + "loss": 0.2455, + "step": 51804 + }, + { + "epoch": 1.99, + "learning_rate": 1.6815209027955885e-05, + "loss": 0.4051, + "step": 51805 + }, + { + "epoch": 1.99, + "learning_rate": 1.6814567837907156e-05, + "loss": 0.2401, + "step": 51806 + }, + { + "epoch": 1.99, + "learning_rate": 1.6813926647858428e-05, + "loss": 0.2554, + "step": 51807 + }, + { + "epoch": 1.99, + "learning_rate": 1.6813285457809696e-05, + "loss": 0.2379, + "step": 51808 + }, + { + "epoch": 1.99, + "learning_rate": 1.6812644267760965e-05, + "loss": 0.2318, + "step": 51809 + }, + { + "epoch": 1.99, + "learning_rate": 1.6812003077712236e-05, + "loss": 0.4294, + "step": 51810 + }, + { + "epoch": 1.99, + "learning_rate": 1.6811361887663505e-05, + "loss": 0.2875, + "step": 51811 + }, + { + "epoch": 1.99, + "learning_rate": 1.6810720697614773e-05, + "loss": 0.1987, + "step": 51812 + }, + { + "epoch": 1.99, + "learning_rate": 1.681007950756604e-05, + "loss": 0.3952, + "step": 51813 + }, + { + "epoch": 1.99, + "learning_rate": 1.6809438317517313e-05, + "loss": 0.3018, + "step": 51814 + }, + { + "epoch": 1.99, + "learning_rate": 1.680879712746858e-05, + "loss": 0.4544, + "step": 51815 + }, + { + "epoch": 1.99, + "learning_rate": 1.6808155937419853e-05, + "loss": 0.2891, + "step": 51816 + }, + { + "epoch": 1.99, + "learning_rate": 1.680751474737112e-05, + "loss": 0.3079, + "step": 51817 + }, + { + "epoch": 1.99, + "learning_rate": 1.6806873557322393e-05, + "loss": 0.2942, + "step": 51818 + }, + { + "epoch": 1.99, + "learning_rate": 1.680623236727366e-05, + "loss": 0.2585, + "step": 51819 + }, + { + "epoch": 1.99, + "learning_rate": 1.680559117722493e-05, + "loss": 0.4276, + "step": 51820 + }, + { + "epoch": 1.99, + "learning_rate": 1.6804949987176198e-05, + "loss": 0.283, + "step": 51821 + }, + { + "epoch": 1.99, + "learning_rate": 1.680430879712747e-05, + "loss": 0.3226, + "step": 51822 + }, + { + "epoch": 1.99, + "learning_rate": 1.6803667607078738e-05, + "loss": 0.416, + "step": 51823 + }, + { + "epoch": 1.99, + "learning_rate": 1.6803026417030006e-05, + "loss": 0.4631, + "step": 51824 + }, + { + "epoch": 1.99, + "learning_rate": 1.6802385226981278e-05, + "loss": 0.2313, + "step": 51825 + }, + { + "epoch": 1.99, + "learning_rate": 1.680174403693255e-05, + "loss": 0.1929, + "step": 51826 + }, + { + "epoch": 1.99, + "learning_rate": 1.6801102846883818e-05, + "loss": 0.3162, + "step": 51827 + }, + { + "epoch": 1.99, + "learning_rate": 1.6800461656835087e-05, + "loss": 0.4109, + "step": 51828 + }, + { + "epoch": 1.99, + "learning_rate": 1.6799820466786355e-05, + "loss": 0.2843, + "step": 51829 + }, + { + "epoch": 1.99, + "learning_rate": 1.6799179276737627e-05, + "loss": 0.2338, + "step": 51830 + }, + { + "epoch": 1.99, + "learning_rate": 1.6798538086688895e-05, + "loss": 0.2744, + "step": 51831 + }, + { + "epoch": 1.99, + "learning_rate": 1.6797896896640163e-05, + "loss": 0.289, + "step": 51832 + }, + { + "epoch": 1.99, + "learning_rate": 1.679725570659143e-05, + "loss": 0.2413, + "step": 51833 + }, + { + "epoch": 1.99, + "learning_rate": 1.6796614516542707e-05, + "loss": 0.3853, + "step": 51834 + }, + { + "epoch": 1.99, + "learning_rate": 1.6795973326493975e-05, + "loss": 0.3076, + "step": 51835 + }, + { + "epoch": 1.99, + "learning_rate": 1.6795332136445243e-05, + "loss": 0.2727, + "step": 51836 + }, + { + "epoch": 1.99, + "learning_rate": 1.679469094639651e-05, + "loss": 0.3334, + "step": 51837 + }, + { + "epoch": 1.99, + "learning_rate": 1.6794049756347783e-05, + "loss": 0.2342, + "step": 51838 + }, + { + "epoch": 1.99, + "learning_rate": 1.6793408566299052e-05, + "loss": 0.378, + "step": 51839 + }, + { + "epoch": 1.99, + "learning_rate": 1.679276737625032e-05, + "loss": 0.3512, + "step": 51840 + }, + { + "epoch": 1.99, + "learning_rate": 1.679212618620159e-05, + "loss": 0.2571, + "step": 51841 + }, + { + "epoch": 1.99, + "learning_rate": 1.679148499615286e-05, + "loss": 0.3803, + "step": 51842 + }, + { + "epoch": 1.99, + "learning_rate": 1.6790843806104132e-05, + "loss": 0.3046, + "step": 51843 + }, + { + "epoch": 1.99, + "learning_rate": 1.67902026160554e-05, + "loss": 0.3739, + "step": 51844 + }, + { + "epoch": 1.99, + "learning_rate": 1.678956142600667e-05, + "loss": 0.2269, + "step": 51845 + }, + { + "epoch": 1.99, + "learning_rate": 1.678892023595794e-05, + "loss": 0.293, + "step": 51846 + }, + { + "epoch": 1.99, + "learning_rate": 1.678827904590921e-05, + "loss": 0.367, + "step": 51847 + }, + { + "epoch": 1.99, + "learning_rate": 1.6787637855860477e-05, + "loss": 0.3579, + "step": 51848 + }, + { + "epoch": 1.99, + "learning_rate": 1.6786996665811745e-05, + "loss": 0.2445, + "step": 51849 + }, + { + "epoch": 1.99, + "learning_rate": 1.6786355475763017e-05, + "loss": 0.3059, + "step": 51850 + }, + { + "epoch": 1.99, + "learning_rate": 1.6785714285714285e-05, + "loss": 0.3926, + "step": 51851 + }, + { + "epoch": 1.99, + "learning_rate": 1.6785073095665557e-05, + "loss": 0.2678, + "step": 51852 + }, + { + "epoch": 1.99, + "learning_rate": 1.6784431905616825e-05, + "loss": 0.2207, + "step": 51853 + }, + { + "epoch": 1.99, + "learning_rate": 1.6783790715568097e-05, + "loss": 0.2768, + "step": 51854 + }, + { + "epoch": 1.99, + "learning_rate": 1.6783149525519365e-05, + "loss": 0.3701, + "step": 51855 + }, + { + "epoch": 1.99, + "learning_rate": 1.6782508335470634e-05, + "loss": 0.3106, + "step": 51856 + }, + { + "epoch": 1.99, + "learning_rate": 1.6781867145421902e-05, + "loss": 0.4389, + "step": 51857 + }, + { + "epoch": 1.99, + "learning_rate": 1.6781225955373174e-05, + "loss": 0.3449, + "step": 51858 + }, + { + "epoch": 1.99, + "learning_rate": 1.6780584765324442e-05, + "loss": 0.2925, + "step": 51859 + }, + { + "epoch": 1.99, + "learning_rate": 1.677994357527571e-05, + "loss": 0.4281, + "step": 51860 + }, + { + "epoch": 1.99, + "learning_rate": 1.6779302385226982e-05, + "loss": 0.308, + "step": 51861 + }, + { + "epoch": 1.99, + "learning_rate": 1.6778661195178254e-05, + "loss": 0.311, + "step": 51862 + }, + { + "epoch": 1.99, + "learning_rate": 1.6778020005129522e-05, + "loss": 0.3237, + "step": 51863 + }, + { + "epoch": 1.99, + "learning_rate": 1.677737881508079e-05, + "loss": 0.2159, + "step": 51864 + }, + { + "epoch": 1.99, + "learning_rate": 1.677673762503206e-05, + "loss": 0.2028, + "step": 51865 + }, + { + "epoch": 1.99, + "learning_rate": 1.677609643498333e-05, + "loss": 0.2525, + "step": 51866 + }, + { + "epoch": 1.99, + "learning_rate": 1.67754552449346e-05, + "loss": 0.3375, + "step": 51867 + }, + { + "epoch": 1.99, + "learning_rate": 1.6774814054885867e-05, + "loss": 0.2633, + "step": 51868 + }, + { + "epoch": 1.99, + "learning_rate": 1.677417286483714e-05, + "loss": 0.2744, + "step": 51869 + }, + { + "epoch": 1.99, + "learning_rate": 1.677353167478841e-05, + "loss": 0.3876, + "step": 51870 + }, + { + "epoch": 1.99, + "learning_rate": 1.677289048473968e-05, + "loss": 0.4315, + "step": 51871 + }, + { + "epoch": 1.99, + "learning_rate": 1.6772249294690947e-05, + "loss": 0.3066, + "step": 51872 + }, + { + "epoch": 1.99, + "learning_rate": 1.6771608104642216e-05, + "loss": 0.2389, + "step": 51873 + }, + { + "epoch": 1.99, + "learning_rate": 1.6770966914593487e-05, + "loss": 0.4475, + "step": 51874 + }, + { + "epoch": 1.99, + "learning_rate": 1.6770325724544756e-05, + "loss": 0.3943, + "step": 51875 + }, + { + "epoch": 1.99, + "learning_rate": 1.6769684534496024e-05, + "loss": 0.2722, + "step": 51876 + }, + { + "epoch": 1.99, + "learning_rate": 1.6769043344447292e-05, + "loss": 0.2126, + "step": 51877 + }, + { + "epoch": 1.99, + "learning_rate": 1.6768402154398567e-05, + "loss": 0.1944, + "step": 51878 + }, + { + "epoch": 1.99, + "learning_rate": 1.6767760964349836e-05, + "loss": 0.3411, + "step": 51879 + }, + { + "epoch": 1.99, + "learning_rate": 1.6767119774301104e-05, + "loss": 0.3235, + "step": 51880 + }, + { + "epoch": 1.99, + "learning_rate": 1.6766478584252372e-05, + "loss": 0.3588, + "step": 51881 + }, + { + "epoch": 1.99, + "learning_rate": 1.6765837394203644e-05, + "loss": 0.3112, + "step": 51882 + }, + { + "epoch": 1.99, + "learning_rate": 1.6765196204154912e-05, + "loss": 0.2381, + "step": 51883 + }, + { + "epoch": 1.99, + "learning_rate": 1.676455501410618e-05, + "loss": 0.2939, + "step": 51884 + }, + { + "epoch": 1.99, + "learning_rate": 1.676391382405745e-05, + "loss": 0.1854, + "step": 51885 + }, + { + "epoch": 1.99, + "learning_rate": 1.676327263400872e-05, + "loss": 0.2606, + "step": 51886 + }, + { + "epoch": 1.99, + "learning_rate": 1.6762631443959992e-05, + "loss": 0.4312, + "step": 51887 + }, + { + "epoch": 1.99, + "learning_rate": 1.676199025391126e-05, + "loss": 0.2371, + "step": 51888 + }, + { + "epoch": 1.99, + "learning_rate": 1.676134906386253e-05, + "loss": 0.3384, + "step": 51889 + }, + { + "epoch": 2.0, + "learning_rate": 1.67607078738138e-05, + "loss": 0.3613, + "step": 51890 + }, + { + "epoch": 2.0, + "learning_rate": 1.676006668376507e-05, + "loss": 0.347, + "step": 51891 + }, + { + "epoch": 2.0, + "learning_rate": 1.6759425493716338e-05, + "loss": 0.208, + "step": 51892 + }, + { + "epoch": 2.0, + "learning_rate": 1.6758784303667606e-05, + "loss": 0.3643, + "step": 51893 + }, + { + "epoch": 2.0, + "learning_rate": 1.6758143113618878e-05, + "loss": 0.2113, + "step": 51894 + }, + { + "epoch": 2.0, + "learning_rate": 1.6757501923570146e-05, + "loss": 0.1914, + "step": 51895 + }, + { + "epoch": 2.0, + "learning_rate": 1.6756860733521418e-05, + "loss": 0.3405, + "step": 51896 + }, + { + "epoch": 2.0, + "learning_rate": 1.6756219543472686e-05, + "loss": 0.2583, + "step": 51897 + }, + { + "epoch": 2.0, + "learning_rate": 1.6755578353423958e-05, + "loss": 0.3564, + "step": 51898 + }, + { + "epoch": 2.0, + "learning_rate": 1.6754937163375226e-05, + "loss": 0.2083, + "step": 51899 + }, + { + "epoch": 2.0, + "learning_rate": 1.6754295973326494e-05, + "loss": 0.3421, + "step": 51900 + }, + { + "epoch": 2.0, + "learning_rate": 1.6753654783277763e-05, + "loss": 0.261, + "step": 51901 + }, + { + "epoch": 2.0, + "learning_rate": 1.6753013593229034e-05, + "loss": 0.3611, + "step": 51902 + }, + { + "epoch": 2.0, + "learning_rate": 1.6752372403180303e-05, + "loss": 0.2152, + "step": 51903 + }, + { + "epoch": 2.0, + "learning_rate": 1.675173121313157e-05, + "loss": 0.286, + "step": 51904 + }, + { + "epoch": 2.0, + "learning_rate": 1.6751090023082843e-05, + "loss": 0.2093, + "step": 51905 + }, + { + "epoch": 2.0, + "learning_rate": 1.6750448833034114e-05, + "loss": 0.3609, + "step": 51906 + }, + { + "epoch": 2.0, + "learning_rate": 1.6749807642985383e-05, + "loss": 0.4203, + "step": 51907 + }, + { + "epoch": 2.0, + "learning_rate": 1.674916645293665e-05, + "loss": 0.3753, + "step": 51908 + }, + { + "epoch": 2.0, + "learning_rate": 1.674852526288792e-05, + "loss": 0.2844, + "step": 51909 + }, + { + "epoch": 2.0, + "learning_rate": 1.674788407283919e-05, + "loss": 0.2449, + "step": 51910 + }, + { + "epoch": 2.0, + "learning_rate": 1.674724288279046e-05, + "loss": 0.3752, + "step": 51911 + }, + { + "epoch": 2.0, + "learning_rate": 1.6746601692741728e-05, + "loss": 0.3245, + "step": 51912 + }, + { + "epoch": 2.0, + "learning_rate": 1.6745960502692996e-05, + "loss": 0.2402, + "step": 51913 + }, + { + "epoch": 2.0, + "learning_rate": 1.674531931264427e-05, + "loss": 0.2696, + "step": 51914 + }, + { + "epoch": 2.0, + "learning_rate": 1.674467812259554e-05, + "loss": 0.3451, + "step": 51915 + }, + { + "epoch": 2.0, + "learning_rate": 1.6744036932546808e-05, + "loss": 0.2304, + "step": 51916 + }, + { + "epoch": 2.0, + "learning_rate": 1.6743395742498076e-05, + "loss": 0.1858, + "step": 51917 + }, + { + "epoch": 2.0, + "learning_rate": 1.6742754552449348e-05, + "loss": 0.2321, + "step": 51918 + }, + { + "epoch": 2.0, + "learning_rate": 1.6742113362400616e-05, + "loss": 0.2308, + "step": 51919 + }, + { + "epoch": 2.0, + "learning_rate": 1.6741472172351885e-05, + "loss": 0.2456, + "step": 51920 + }, + { + "epoch": 2.0, + "learning_rate": 1.6740830982303153e-05, + "loss": 0.4181, + "step": 51921 + }, + { + "epoch": 2.0, + "learning_rate": 1.6740189792254425e-05, + "loss": 0.2839, + "step": 51922 + }, + { + "epoch": 2.0, + "learning_rate": 1.6739548602205696e-05, + "loss": 0.4915, + "step": 51923 + }, + { + "epoch": 2.0, + "learning_rate": 1.6738907412156965e-05, + "loss": 0.2048, + "step": 51924 + }, + { + "epoch": 2.0, + "learning_rate": 1.6738266222108233e-05, + "loss": 0.2355, + "step": 51925 + }, + { + "epoch": 2.0, + "learning_rate": 1.6737625032059505e-05, + "loss": 0.3806, + "step": 51926 + }, + { + "epoch": 2.0, + "learning_rate": 1.6736983842010773e-05, + "loss": 0.3037, + "step": 51927 + }, + { + "epoch": 2.0, + "learning_rate": 1.673634265196204e-05, + "loss": 0.2619, + "step": 51928 + }, + { + "epoch": 2.0, + "learning_rate": 1.673570146191331e-05, + "loss": 0.3031, + "step": 51929 + }, + { + "epoch": 2.0, + "learning_rate": 1.673506027186458e-05, + "loss": 0.3424, + "step": 51930 + }, + { + "epoch": 2.0, + "learning_rate": 1.673441908181585e-05, + "loss": 0.3133, + "step": 51931 + }, + { + "epoch": 2.0, + "learning_rate": 1.673377789176712e-05, + "loss": 0.3327, + "step": 51932 + }, + { + "epoch": 2.0, + "learning_rate": 1.673313670171839e-05, + "loss": 0.3558, + "step": 51933 + }, + { + "epoch": 2.0, + "learning_rate": 1.673249551166966e-05, + "loss": 0.3085, + "step": 51934 + }, + { + "epoch": 2.0, + "learning_rate": 1.673185432162093e-05, + "loss": 0.278, + "step": 51935 + }, + { + "epoch": 2.0, + "learning_rate": 1.6731213131572198e-05, + "loss": 0.3136, + "step": 51936 + }, + { + "epoch": 2.0, + "learning_rate": 1.6730571941523466e-05, + "loss": 0.5854, + "step": 51937 + }, + { + "epoch": 2.0, + "learning_rate": 1.6729930751474738e-05, + "loss": 0.2556, + "step": 51938 + }, + { + "epoch": 2.0, + "learning_rate": 1.6729289561426007e-05, + "loss": 0.4273, + "step": 51939 + }, + { + "epoch": 2.0, + "learning_rate": 1.6728648371377275e-05, + "loss": 0.4351, + "step": 51940 + }, + { + "epoch": 2.0, + "learning_rate": 1.6728007181328547e-05, + "loss": 0.4191, + "step": 51941 + }, + { + "epoch": 2.0, + "learning_rate": 1.6727365991279818e-05, + "loss": 0.3923, + "step": 51942 + }, + { + "epoch": 2.0, + "learning_rate": 1.6726724801231087e-05, + "loss": 0.2549, + "step": 51943 + }, + { + "epoch": 2.0, + "learning_rate": 1.6726083611182355e-05, + "loss": 0.2831, + "step": 51944 + }, + { + "epoch": 2.0, + "learning_rate": 1.6725442421133623e-05, + "loss": 0.2461, + "step": 51945 + }, + { + "epoch": 2.0, + "learning_rate": 1.6724801231084895e-05, + "loss": 0.2559, + "step": 51946 + }, + { + "epoch": 2.0, + "learning_rate": 1.6724160041036163e-05, + "loss": 0.3339, + "step": 51947 + }, + { + "epoch": 2.0, + "learning_rate": 1.672351885098743e-05, + "loss": 0.2287, + "step": 51948 + }, + { + "epoch": 2.0, + "learning_rate": 1.67228776609387e-05, + "loss": 0.2265, + "step": 51949 + }, + { + "epoch": 2.0, + "learning_rate": 1.6722236470889975e-05, + "loss": 0.2442, + "step": 51950 + }, + { + "epoch": 2.0, + "learning_rate": 1.6721595280841243e-05, + "loss": 0.6401, + "step": 51951 + }, + { + "epoch": 2.0, + "learning_rate": 1.6720954090792512e-05, + "loss": 0.2099, + "step": 51952 + }, + { + "epoch": 2.0, + "learning_rate": 1.672031290074378e-05, + "loss": 0.187, + "step": 51953 + }, + { + "epoch": 2.0, + "learning_rate": 1.6719671710695052e-05, + "loss": 0.2557, + "step": 51954 + }, + { + "epoch": 2.0, + "learning_rate": 1.671903052064632e-05, + "loss": 0.3743, + "step": 51955 + }, + { + "epoch": 2.0, + "learning_rate": 1.671838933059759e-05, + "loss": 0.3796, + "step": 51956 + }, + { + "epoch": 2.0, + "learning_rate": 1.6717748140548857e-05, + "loss": 0.2584, + "step": 51957 + }, + { + "epoch": 2.0, + "learning_rate": 1.6717106950500132e-05, + "loss": 0.3126, + "step": 51958 + }, + { + "epoch": 2.0, + "learning_rate": 1.67164657604514e-05, + "loss": 0.4444, + "step": 51959 + }, + { + "epoch": 2.0, + "learning_rate": 1.671582457040267e-05, + "loss": 0.2068, + "step": 51960 + }, + { + "epoch": 2.0, + "learning_rate": 1.6715183380353937e-05, + "loss": 0.3574, + "step": 51961 + }, + { + "epoch": 2.0, + "learning_rate": 1.671454219030521e-05, + "loss": 0.2957, + "step": 51962 + }, + { + "epoch": 2.0, + "learning_rate": 1.6713901000256477e-05, + "loss": 0.2961, + "step": 51963 + }, + { + "epoch": 2.0, + "learning_rate": 1.6713259810207745e-05, + "loss": 0.2468, + "step": 51964 + }, + { + "epoch": 2.0, + "learning_rate": 1.6712618620159014e-05, + "loss": 0.3066, + "step": 51965 + }, + { + "epoch": 2.0, + "learning_rate": 1.6711977430110285e-05, + "loss": 0.186, + "step": 51966 + }, + { + "epoch": 2.0, + "learning_rate": 1.6711336240061557e-05, + "loss": 0.4259, + "step": 51967 + }, + { + "epoch": 2.0, + "learning_rate": 1.6710695050012825e-05, + "loss": 0.5518, + "step": 51968 + }, + { + "epoch": 2.0, + "learning_rate": 1.6710053859964094e-05, + "loss": 0.2648, + "step": 51969 + }, + { + "epoch": 2.0, + "learning_rate": 1.6709412669915365e-05, + "loss": 0.2397, + "step": 51970 + }, + { + "epoch": 2.0, + "learning_rate": 1.6708771479866634e-05, + "loss": 0.4061, + "step": 51971 + }, + { + "epoch": 2.0, + "learning_rate": 1.6708130289817902e-05, + "loss": 0.4008, + "step": 51972 + }, + { + "epoch": 2.0, + "learning_rate": 1.670748909976917e-05, + "loss": 0.4004, + "step": 51973 + }, + { + "epoch": 2.0, + "learning_rate": 1.6706847909720442e-05, + "loss": 0.3421, + "step": 51974 + }, + { + "epoch": 2.0, + "learning_rate": 1.670620671967171e-05, + "loss": 0.235, + "step": 51975 + }, + { + "epoch": 2.0, + "learning_rate": 1.6705565529622982e-05, + "loss": 0.2133, + "step": 51976 + }, + { + "epoch": 2.0, + "learning_rate": 1.670492433957425e-05, + "loss": 0.3651, + "step": 51977 + }, + { + "epoch": 2.0, + "learning_rate": 1.6704283149525522e-05, + "loss": 0.2895, + "step": 51978 + }, + { + "epoch": 2.0, + "learning_rate": 1.670364195947679e-05, + "loss": 0.4158, + "step": 51979 + }, + { + "epoch": 2.0, + "learning_rate": 1.670300076942806e-05, + "loss": 0.344, + "step": 51980 + }, + { + "epoch": 2.0, + "learning_rate": 1.6702359579379327e-05, + "loss": 0.2247, + "step": 51981 + }, + { + "epoch": 2.0, + "learning_rate": 1.67017183893306e-05, + "loss": 0.3354, + "step": 51982 + }, + { + "epoch": 2.0, + "learning_rate": 1.6701077199281867e-05, + "loss": 0.2123, + "step": 51983 + }, + { + "epoch": 2.0, + "learning_rate": 1.6700436009233136e-05, + "loss": 0.3285, + "step": 51984 + }, + { + "epoch": 2.0, + "learning_rate": 1.6699794819184407e-05, + "loss": 0.3071, + "step": 51985 + }, + { + "epoch": 2.0, + "learning_rate": 1.669915362913568e-05, + "loss": 0.2642, + "step": 51986 + }, + { + "epoch": 2.0, + "learning_rate": 1.6698512439086947e-05, + "loss": 0.3008, + "step": 51987 + }, + { + "epoch": 2.0, + "learning_rate": 1.6697871249038216e-05, + "loss": 0.3195, + "step": 51988 + }, + { + "epoch": 2.0, + "learning_rate": 1.6697230058989484e-05, + "loss": 0.4322, + "step": 51989 + }, + { + "epoch": 2.0, + "learning_rate": 1.6696588868940756e-05, + "loss": 0.3111, + "step": 51990 + }, + { + "epoch": 2.0, + "learning_rate": 1.6695947678892024e-05, + "loss": 0.2761, + "step": 51991 + }, + { + "epoch": 2.0, + "learning_rate": 1.6695306488843292e-05, + "loss": 0.3745, + "step": 51992 + }, + { + "epoch": 2.0, + "learning_rate": 1.669466529879456e-05, + "loss": 0.3382, + "step": 51993 + }, + { + "epoch": 2.0, + "learning_rate": 1.6694024108745836e-05, + "loss": 0.2348, + "step": 51994 + }, + { + "epoch": 2.0, + "learning_rate": 1.6693382918697104e-05, + "loss": 0.2459, + "step": 51995 + }, + { + "epoch": 2.0, + "learning_rate": 1.6692741728648372e-05, + "loss": 0.3146, + "step": 51996 + }, + { + "epoch": 2.0, + "learning_rate": 1.669210053859964e-05, + "loss": 0.4192, + "step": 51997 + }, + { + "epoch": 2.0, + "learning_rate": 1.6691459348550912e-05, + "loss": 0.2578, + "step": 51998 + }, + { + "epoch": 2.0, + "learning_rate": 1.669081815850218e-05, + "loss": 0.388, + "step": 51999 + }, + { + "epoch": 2.0, + "learning_rate": 1.669017696845345e-05, + "loss": 0.3039, + "step": 52000 + }, + { + "epoch": 2.0, + "learning_rate": 1.6689535778404717e-05, + "loss": 0.2764, + "step": 52001 + }, + { + "epoch": 2.0, + "learning_rate": 1.668889458835599e-05, + "loss": 0.3754, + "step": 52002 + }, + { + "epoch": 2.0, + "learning_rate": 1.668825339830726e-05, + "loss": 0.2854, + "step": 52003 + }, + { + "epoch": 2.0, + "learning_rate": 1.668761220825853e-05, + "loss": 0.3042, + "step": 52004 + }, + { + "epoch": 2.0, + "learning_rate": 1.6686971018209798e-05, + "loss": 0.3709, + "step": 52005 + }, + { + "epoch": 2.0, + "learning_rate": 1.668632982816107e-05, + "loss": 0.3875, + "step": 52006 + }, + { + "epoch": 2.0, + "learning_rate": 1.6685688638112338e-05, + "loss": 0.2456, + "step": 52007 + }, + { + "epoch": 2.0, + "learning_rate": 1.6685047448063606e-05, + "loss": 0.3131, + "step": 52008 + }, + { + "epoch": 2.0, + "learning_rate": 1.6684406258014874e-05, + "loss": 0.3491, + "step": 52009 + }, + { + "epoch": 2.0, + "learning_rate": 1.6683765067966146e-05, + "loss": 0.3317, + "step": 52010 + }, + { + "epoch": 2.0, + "learning_rate": 1.6683123877917414e-05, + "loss": 0.3469, + "step": 52011 + }, + { + "epoch": 2.0, + "learning_rate": 1.6682482687868686e-05, + "loss": 0.301, + "step": 52012 + }, + { + "epoch": 2.0, + "learning_rate": 1.6681841497819954e-05, + "loss": 0.3444, + "step": 52013 + }, + { + "epoch": 2.0, + "learning_rate": 1.6681200307771226e-05, + "loss": 0.2251, + "step": 52014 + }, + { + "epoch": 2.0, + "learning_rate": 1.6680559117722494e-05, + "loss": 0.4044, + "step": 52015 + }, + { + "epoch": 2.0, + "learning_rate": 1.6679917927673763e-05, + "loss": 0.179, + "step": 52016 + }, + { + "epoch": 2.0, + "learning_rate": 1.667927673762503e-05, + "loss": 0.1906, + "step": 52017 + }, + { + "epoch": 2.0, + "learning_rate": 1.6678635547576303e-05, + "loss": 0.2671, + "step": 52018 + }, + { + "epoch": 2.0, + "learning_rate": 1.667799435752757e-05, + "loss": 0.2276, + "step": 52019 + }, + { + "epoch": 2.0, + "learning_rate": 1.667735316747884e-05, + "loss": 0.2054, + "step": 52020 + }, + { + "epoch": 2.0, + "eval_accuracy": 0.8583738426899334, + "eval_loss": 0.3326870799064636, + "eval_runtime": 254.1848, + "eval_samples_per_second": 1637.23, + "eval_steps_per_second": 25.584, + "step": 52020 + }, + { + "epoch": 2.0, + "learning_rate": 1.667671197743011e-05, + "loss": 0.2617, + "step": 52021 + }, + { + "epoch": 2.0, + "learning_rate": 1.6676070787381383e-05, + "loss": 0.1911, + "step": 52022 + }, + { + "epoch": 2.0, + "learning_rate": 1.667542959733265e-05, + "loss": 0.2017, + "step": 52023 + }, + { + "epoch": 2.0, + "learning_rate": 1.667478840728392e-05, + "loss": 0.2035, + "step": 52024 + }, + { + "epoch": 2.0, + "learning_rate": 1.6674147217235188e-05, + "loss": 0.3463, + "step": 52025 + }, + { + "epoch": 2.0, + "learning_rate": 1.667350602718646e-05, + "loss": 0.1639, + "step": 52026 + }, + { + "epoch": 2.0, + "learning_rate": 1.6672864837137728e-05, + "loss": 0.3473, + "step": 52027 + }, + { + "epoch": 2.0, + "learning_rate": 1.6672223647088996e-05, + "loss": 0.3198, + "step": 52028 + }, + { + "epoch": 2.0, + "learning_rate": 1.6671582457040268e-05, + "loss": 0.2587, + "step": 52029 + }, + { + "epoch": 2.0, + "learning_rate": 1.667094126699154e-05, + "loss": 0.3567, + "step": 52030 + }, + { + "epoch": 2.0, + "learning_rate": 1.6670300076942808e-05, + "loss": 0.2572, + "step": 52031 + }, + { + "epoch": 2.0, + "learning_rate": 1.6669658886894076e-05, + "loss": 0.1505, + "step": 52032 + }, + { + "epoch": 2.0, + "learning_rate": 1.6669017696845345e-05, + "loss": 0.3574, + "step": 52033 + }, + { + "epoch": 2.0, + "learning_rate": 1.6668376506796616e-05, + "loss": 0.2953, + "step": 52034 + }, + { + "epoch": 2.0, + "learning_rate": 1.6667735316747885e-05, + "loss": 0.2512, + "step": 52035 + }, + { + "epoch": 2.0, + "learning_rate": 1.6667094126699153e-05, + "loss": 0.2213, + "step": 52036 + }, + { + "epoch": 2.0, + "learning_rate": 1.666645293665042e-05, + "loss": 0.2848, + "step": 52037 + }, + { + "epoch": 2.0, + "learning_rate": 1.6665811746601693e-05, + "loss": 0.283, + "step": 52038 + }, + { + "epoch": 2.0, + "learning_rate": 1.6665170556552965e-05, + "loss": 0.2179, + "step": 52039 + }, + { + "epoch": 2.0, + "learning_rate": 1.6664529366504233e-05, + "loss": 0.2242, + "step": 52040 + }, + { + "epoch": 2.0, + "learning_rate": 1.66638881764555e-05, + "loss": 0.2129, + "step": 52041 + }, + { + "epoch": 2.0, + "learning_rate": 1.6663246986406773e-05, + "loss": 0.2305, + "step": 52042 + }, + { + "epoch": 2.0, + "learning_rate": 1.666260579635804e-05, + "loss": 0.2353, + "step": 52043 + }, + { + "epoch": 2.0, + "learning_rate": 1.666196460630931e-05, + "loss": 0.2345, + "step": 52044 + }, + { + "epoch": 2.0, + "learning_rate": 1.6661323416260578e-05, + "loss": 0.1753, + "step": 52045 + }, + { + "epoch": 2.0, + "learning_rate": 1.666068222621185e-05, + "loss": 0.2836, + "step": 52046 + }, + { + "epoch": 2.0, + "learning_rate": 1.6660041036163118e-05, + "loss": 0.1693, + "step": 52047 + }, + { + "epoch": 2.0, + "learning_rate": 1.665939984611439e-05, + "loss": 0.3506, + "step": 52048 + }, + { + "epoch": 2.0, + "learning_rate": 1.6658758656065658e-05, + "loss": 0.1323, + "step": 52049 + }, + { + "epoch": 2.0, + "learning_rate": 1.665811746601693e-05, + "loss": 0.317, + "step": 52050 + }, + { + "epoch": 2.0, + "learning_rate": 1.6657476275968198e-05, + "loss": 0.2147, + "step": 52051 + }, + { + "epoch": 2.0, + "learning_rate": 1.6656835085919467e-05, + "loss": 0.2596, + "step": 52052 + }, + { + "epoch": 2.0, + "learning_rate": 1.6656193895870735e-05, + "loss": 0.3258, + "step": 52053 + }, + { + "epoch": 2.0, + "learning_rate": 1.6655552705822007e-05, + "loss": 0.2399, + "step": 52054 + }, + { + "epoch": 2.0, + "learning_rate": 1.6654911515773275e-05, + "loss": 0.3015, + "step": 52055 + }, + { + "epoch": 2.0, + "learning_rate": 1.6654270325724547e-05, + "loss": 0.1675, + "step": 52056 + }, + { + "epoch": 2.0, + "learning_rate": 1.6653629135675815e-05, + "loss": 0.3086, + "step": 52057 + }, + { + "epoch": 2.0, + "learning_rate": 1.6652987945627087e-05, + "loss": 0.2626, + "step": 52058 + }, + { + "epoch": 2.0, + "learning_rate": 1.6652346755578355e-05, + "loss": 0.2905, + "step": 52059 + }, + { + "epoch": 2.0, + "learning_rate": 1.6651705565529623e-05, + "loss": 0.2847, + "step": 52060 + }, + { + "epoch": 2.0, + "learning_rate": 1.665106437548089e-05, + "loss": 0.2195, + "step": 52061 + }, + { + "epoch": 2.0, + "learning_rate": 1.6650423185432163e-05, + "loss": 0.1736, + "step": 52062 + }, + { + "epoch": 2.0, + "learning_rate": 1.6649781995383432e-05, + "loss": 0.2856, + "step": 52063 + }, + { + "epoch": 2.0, + "learning_rate": 1.66491408053347e-05, + "loss": 0.1987, + "step": 52064 + }, + { + "epoch": 2.0, + "learning_rate": 1.6648499615285972e-05, + "loss": 0.2331, + "step": 52065 + }, + { + "epoch": 2.0, + "learning_rate": 1.6647858425237243e-05, + "loss": 0.242, + "step": 52066 + }, + { + "epoch": 2.0, + "learning_rate": 1.6647217235188512e-05, + "loss": 0.2327, + "step": 52067 + }, + { + "epoch": 2.0, + "learning_rate": 1.664657604513978e-05, + "loss": 0.2549, + "step": 52068 + }, + { + "epoch": 2.0, + "learning_rate": 1.664593485509105e-05, + "loss": 0.2275, + "step": 52069 + }, + { + "epoch": 2.0, + "learning_rate": 1.664529366504232e-05, + "loss": 0.1498, + "step": 52070 + }, + { + "epoch": 2.0, + "learning_rate": 1.664465247499359e-05, + "loss": 0.2896, + "step": 52071 + }, + { + "epoch": 2.0, + "learning_rate": 1.6644011284944857e-05, + "loss": 0.2093, + "step": 52072 + }, + { + "epoch": 2.0, + "learning_rate": 1.6643370094896125e-05, + "loss": 0.3377, + "step": 52073 + }, + { + "epoch": 2.0, + "learning_rate": 1.66427289048474e-05, + "loss": 0.2371, + "step": 52074 + }, + { + "epoch": 2.0, + "learning_rate": 1.664208771479867e-05, + "loss": 0.3506, + "step": 52075 + }, + { + "epoch": 2.0, + "learning_rate": 1.6641446524749937e-05, + "loss": 0.1742, + "step": 52076 + }, + { + "epoch": 2.0, + "learning_rate": 1.6640805334701205e-05, + "loss": 0.2447, + "step": 52077 + }, + { + "epoch": 2.0, + "learning_rate": 1.6640164144652477e-05, + "loss": 0.2461, + "step": 52078 + }, + { + "epoch": 2.0, + "learning_rate": 1.6639522954603745e-05, + "loss": 0.2299, + "step": 52079 + }, + { + "epoch": 2.0, + "learning_rate": 1.6638881764555014e-05, + "loss": 0.2283, + "step": 52080 + }, + { + "epoch": 2.0, + "learning_rate": 1.6638240574506282e-05, + "loss": 0.2391, + "step": 52081 + }, + { + "epoch": 2.0, + "learning_rate": 1.6637599384457554e-05, + "loss": 0.1629, + "step": 52082 + }, + { + "epoch": 2.0, + "learning_rate": 1.6636958194408825e-05, + "loss": 0.2922, + "step": 52083 + }, + { + "epoch": 2.0, + "learning_rate": 1.6636317004360094e-05, + "loss": 0.1703, + "step": 52084 + }, + { + "epoch": 2.0, + "learning_rate": 1.6635675814311362e-05, + "loss": 0.2692, + "step": 52085 + }, + { + "epoch": 2.0, + "learning_rate": 1.6635034624262634e-05, + "loss": 0.1317, + "step": 52086 + }, + { + "epoch": 2.0, + "learning_rate": 1.6634393434213902e-05, + "loss": 0.2006, + "step": 52087 + }, + { + "epoch": 2.0, + "learning_rate": 1.663375224416517e-05, + "loss": 0.2575, + "step": 52088 + }, + { + "epoch": 2.0, + "learning_rate": 1.663311105411644e-05, + "loss": 0.2762, + "step": 52089 + }, + { + "epoch": 2.0, + "learning_rate": 1.663246986406771e-05, + "loss": 0.24, + "step": 52090 + }, + { + "epoch": 2.0, + "learning_rate": 1.663182867401898e-05, + "loss": 0.2096, + "step": 52091 + }, + { + "epoch": 2.0, + "learning_rate": 1.663118748397025e-05, + "loss": 0.2377, + "step": 52092 + }, + { + "epoch": 2.0, + "learning_rate": 1.663054629392152e-05, + "loss": 0.1571, + "step": 52093 + }, + { + "epoch": 2.0, + "learning_rate": 1.662990510387279e-05, + "loss": 0.1418, + "step": 52094 + }, + { + "epoch": 2.0, + "learning_rate": 1.662926391382406e-05, + "loss": 0.1738, + "step": 52095 + }, + { + "epoch": 2.0, + "learning_rate": 1.6628622723775327e-05, + "loss": 0.2314, + "step": 52096 + }, + { + "epoch": 2.0, + "learning_rate": 1.6627981533726596e-05, + "loss": 0.2419, + "step": 52097 + }, + { + "epoch": 2.0, + "learning_rate": 1.6627340343677867e-05, + "loss": 0.3327, + "step": 52098 + }, + { + "epoch": 2.0, + "learning_rate": 1.6626699153629136e-05, + "loss": 0.1942, + "step": 52099 + }, + { + "epoch": 2.0, + "learning_rate": 1.6626057963580404e-05, + "loss": 0.1784, + "step": 52100 + }, + { + "epoch": 2.0, + "learning_rate": 1.6625416773531676e-05, + "loss": 0.1859, + "step": 52101 + }, + { + "epoch": 2.0, + "learning_rate": 1.6624775583482947e-05, + "loss": 0.2778, + "step": 52102 + }, + { + "epoch": 2.0, + "learning_rate": 1.6624134393434216e-05, + "loss": 0.2588, + "step": 52103 + }, + { + "epoch": 2.0, + "learning_rate": 1.6623493203385484e-05, + "loss": 0.2228, + "step": 52104 + }, + { + "epoch": 2.0, + "learning_rate": 1.6622852013336752e-05, + "loss": 0.2196, + "step": 52105 + }, + { + "epoch": 2.0, + "learning_rate": 1.6622210823288024e-05, + "loss": 0.1966, + "step": 52106 + }, + { + "epoch": 2.0, + "learning_rate": 1.6621569633239292e-05, + "loss": 0.2396, + "step": 52107 + }, + { + "epoch": 2.0, + "learning_rate": 1.662092844319056e-05, + "loss": 0.2783, + "step": 52108 + }, + { + "epoch": 2.0, + "learning_rate": 1.6620287253141832e-05, + "loss": 0.2561, + "step": 52109 + }, + { + "epoch": 2.0, + "learning_rate": 1.6619646063093104e-05, + "loss": 0.1998, + "step": 52110 + }, + { + "epoch": 2.0, + "learning_rate": 1.6619004873044372e-05, + "loss": 0.1988, + "step": 52111 + }, + { + "epoch": 2.0, + "learning_rate": 1.661836368299564e-05, + "loss": 0.2795, + "step": 52112 + }, + { + "epoch": 2.0, + "learning_rate": 1.661772249294691e-05, + "loss": 0.3489, + "step": 52113 + }, + { + "epoch": 2.0, + "learning_rate": 1.661708130289818e-05, + "loss": 0.326, + "step": 52114 + }, + { + "epoch": 2.0, + "learning_rate": 1.661644011284945e-05, + "loss": 0.1914, + "step": 52115 + }, + { + "epoch": 2.0, + "learning_rate": 1.6615798922800718e-05, + "loss": 0.1967, + "step": 52116 + }, + { + "epoch": 2.0, + "learning_rate": 1.6615157732751986e-05, + "loss": 0.284, + "step": 52117 + }, + { + "epoch": 2.0, + "learning_rate": 1.6614516542703258e-05, + "loss": 0.459, + "step": 52118 + }, + { + "epoch": 2.0, + "learning_rate": 1.661387535265453e-05, + "loss": 0.2823, + "step": 52119 + }, + { + "epoch": 2.0, + "learning_rate": 1.6613234162605798e-05, + "loss": 0.198, + "step": 52120 + }, + { + "epoch": 2.0, + "learning_rate": 1.6612592972557066e-05, + "loss": 0.219, + "step": 52121 + }, + { + "epoch": 2.0, + "learning_rate": 1.6611951782508338e-05, + "loss": 0.2343, + "step": 52122 + }, + { + "epoch": 2.0, + "learning_rate": 1.6611310592459606e-05, + "loss": 0.3341, + "step": 52123 + }, + { + "epoch": 2.0, + "learning_rate": 1.6610669402410874e-05, + "loss": 0.1613, + "step": 52124 + }, + { + "epoch": 2.0, + "learning_rate": 1.6610028212362143e-05, + "loss": 0.1693, + "step": 52125 + }, + { + "epoch": 2.0, + "learning_rate": 1.6609387022313414e-05, + "loss": 0.1515, + "step": 52126 + }, + { + "epoch": 2.0, + "learning_rate": 1.6608745832264683e-05, + "loss": 0.2843, + "step": 52127 + }, + { + "epoch": 2.0, + "learning_rate": 1.6608104642215954e-05, + "loss": 0.1728, + "step": 52128 + }, + { + "epoch": 2.0, + "learning_rate": 1.6607463452167223e-05, + "loss": 0.2781, + "step": 52129 + }, + { + "epoch": 2.0, + "learning_rate": 1.6606822262118494e-05, + "loss": 0.2736, + "step": 52130 + }, + { + "epoch": 2.0, + "learning_rate": 1.6606181072069763e-05, + "loss": 0.2279, + "step": 52131 + }, + { + "epoch": 2.0, + "learning_rate": 1.660553988202103e-05, + "loss": 0.3514, + "step": 52132 + }, + { + "epoch": 2.0, + "learning_rate": 1.66048986919723e-05, + "loss": 0.4183, + "step": 52133 + }, + { + "epoch": 2.0, + "learning_rate": 1.660425750192357e-05, + "loss": 0.1432, + "step": 52134 + }, + { + "epoch": 2.0, + "learning_rate": 1.660361631187484e-05, + "loss": 0.332, + "step": 52135 + }, + { + "epoch": 2.0, + "learning_rate": 1.6602975121826108e-05, + "loss": 0.3763, + "step": 52136 + }, + { + "epoch": 2.0, + "learning_rate": 1.660233393177738e-05, + "loss": 0.1665, + "step": 52137 + }, + { + "epoch": 2.0, + "learning_rate": 1.660169274172865e-05, + "loss": 0.1378, + "step": 52138 + }, + { + "epoch": 2.0, + "learning_rate": 1.660105155167992e-05, + "loss": 0.1884, + "step": 52139 + }, + { + "epoch": 2.0, + "learning_rate": 1.6600410361631188e-05, + "loss": 0.1487, + "step": 52140 + }, + { + "epoch": 2.0, + "learning_rate": 1.6599769171582456e-05, + "loss": 0.2101, + "step": 52141 + }, + { + "epoch": 2.0, + "learning_rate": 1.6599127981533728e-05, + "loss": 0.2895, + "step": 52142 + }, + { + "epoch": 2.0, + "learning_rate": 1.6598486791484996e-05, + "loss": 0.2703, + "step": 52143 + }, + { + "epoch": 2.0, + "learning_rate": 1.6597845601436265e-05, + "loss": 0.1885, + "step": 52144 + }, + { + "epoch": 2.0, + "learning_rate": 1.6597204411387536e-05, + "loss": 0.2518, + "step": 52145 + }, + { + "epoch": 2.0, + "learning_rate": 1.6596563221338808e-05, + "loss": 0.2568, + "step": 52146 + }, + { + "epoch": 2.0, + "learning_rate": 1.6595922031290076e-05, + "loss": 0.21, + "step": 52147 + }, + { + "epoch": 2.0, + "learning_rate": 1.6595280841241345e-05, + "loss": 0.2308, + "step": 52148 + }, + { + "epoch": 2.0, + "learning_rate": 1.6594639651192613e-05, + "loss": 0.2144, + "step": 52149 + }, + { + "epoch": 2.0, + "learning_rate": 1.6593998461143885e-05, + "loss": 0.1758, + "step": 52150 + }, + { + "epoch": 2.01, + "learning_rate": 1.6593357271095153e-05, + "loss": 0.2157, + "step": 52151 + }, + { + "epoch": 2.01, + "learning_rate": 1.659271608104642e-05, + "loss": 0.2658, + "step": 52152 + }, + { + "epoch": 2.01, + "learning_rate": 1.659207489099769e-05, + "loss": 0.179, + "step": 52153 + }, + { + "epoch": 2.01, + "learning_rate": 1.6591433700948965e-05, + "loss": 0.1441, + "step": 52154 + }, + { + "epoch": 2.01, + "learning_rate": 1.6590792510900233e-05, + "loss": 0.3363, + "step": 52155 + }, + { + "epoch": 2.01, + "learning_rate": 1.65901513208515e-05, + "loss": 0.3107, + "step": 52156 + }, + { + "epoch": 2.01, + "learning_rate": 1.658951013080277e-05, + "loss": 0.2127, + "step": 52157 + }, + { + "epoch": 2.01, + "learning_rate": 1.658886894075404e-05, + "loss": 0.1935, + "step": 52158 + }, + { + "epoch": 2.01, + "learning_rate": 1.658822775070531e-05, + "loss": 0.2215, + "step": 52159 + }, + { + "epoch": 2.01, + "learning_rate": 1.6587586560656578e-05, + "loss": 0.242, + "step": 52160 + }, + { + "epoch": 2.01, + "learning_rate": 1.6586945370607847e-05, + "loss": 0.1376, + "step": 52161 + }, + { + "epoch": 2.01, + "learning_rate": 1.6586304180559118e-05, + "loss": 0.3926, + "step": 52162 + }, + { + "epoch": 2.01, + "learning_rate": 1.658566299051039e-05, + "loss": 0.2251, + "step": 52163 + }, + { + "epoch": 2.01, + "learning_rate": 1.6585021800461658e-05, + "loss": 0.2906, + "step": 52164 + }, + { + "epoch": 2.01, + "learning_rate": 1.6584380610412927e-05, + "loss": 0.283, + "step": 52165 + }, + { + "epoch": 2.01, + "learning_rate": 1.65837394203642e-05, + "loss": 0.2015, + "step": 52166 + }, + { + "epoch": 2.01, + "learning_rate": 1.6583098230315467e-05, + "loss": 0.1475, + "step": 52167 + }, + { + "epoch": 2.01, + "learning_rate": 1.6582457040266735e-05, + "loss": 0.442, + "step": 52168 + }, + { + "epoch": 2.01, + "learning_rate": 1.6581815850218003e-05, + "loss": 0.3038, + "step": 52169 + }, + { + "epoch": 2.01, + "learning_rate": 1.6581174660169275e-05, + "loss": 0.3023, + "step": 52170 + }, + { + "epoch": 2.01, + "learning_rate": 1.6580533470120543e-05, + "loss": 0.2237, + "step": 52171 + }, + { + "epoch": 2.01, + "learning_rate": 1.6579892280071815e-05, + "loss": 0.2077, + "step": 52172 + }, + { + "epoch": 2.01, + "learning_rate": 1.6579251090023083e-05, + "loss": 0.2392, + "step": 52173 + }, + { + "epoch": 2.01, + "learning_rate": 1.6578609899974355e-05, + "loss": 0.2721, + "step": 52174 + }, + { + "epoch": 2.01, + "learning_rate": 1.6577968709925623e-05, + "loss": 0.2106, + "step": 52175 + }, + { + "epoch": 2.01, + "learning_rate": 1.6577327519876892e-05, + "loss": 0.173, + "step": 52176 + }, + { + "epoch": 2.01, + "learning_rate": 1.657668632982816e-05, + "loss": 0.3038, + "step": 52177 + }, + { + "epoch": 2.01, + "learning_rate": 1.6576045139779432e-05, + "loss": 0.2132, + "step": 52178 + }, + { + "epoch": 2.01, + "learning_rate": 1.65754039497307e-05, + "loss": 0.2787, + "step": 52179 + }, + { + "epoch": 2.01, + "learning_rate": 1.657476275968197e-05, + "loss": 0.2799, + "step": 52180 + }, + { + "epoch": 2.01, + "learning_rate": 1.657412156963324e-05, + "loss": 0.2617, + "step": 52181 + }, + { + "epoch": 2.01, + "learning_rate": 1.6573480379584512e-05, + "loss": 0.1578, + "step": 52182 + }, + { + "epoch": 2.01, + "learning_rate": 1.657283918953578e-05, + "loss": 0.2966, + "step": 52183 + }, + { + "epoch": 2.01, + "learning_rate": 1.657219799948705e-05, + "loss": 0.1966, + "step": 52184 + }, + { + "epoch": 2.01, + "learning_rate": 1.6571556809438317e-05, + "loss": 0.1879, + "step": 52185 + }, + { + "epoch": 2.01, + "learning_rate": 1.657091561938959e-05, + "loss": 0.2419, + "step": 52186 + }, + { + "epoch": 2.01, + "learning_rate": 1.6570274429340857e-05, + "loss": 0.2235, + "step": 52187 + }, + { + "epoch": 2.01, + "learning_rate": 1.6569633239292125e-05, + "loss": 0.199, + "step": 52188 + }, + { + "epoch": 2.01, + "learning_rate": 1.6568992049243397e-05, + "loss": 0.3044, + "step": 52189 + }, + { + "epoch": 2.01, + "learning_rate": 1.656835085919467e-05, + "loss": 0.1331, + "step": 52190 + }, + { + "epoch": 2.01, + "learning_rate": 1.6567709669145937e-05, + "loss": 0.3579, + "step": 52191 + }, + { + "epoch": 2.01, + "learning_rate": 1.6567068479097205e-05, + "loss": 0.2101, + "step": 52192 + }, + { + "epoch": 2.01, + "learning_rate": 1.6566427289048474e-05, + "loss": 0.1815, + "step": 52193 + }, + { + "epoch": 2.01, + "learning_rate": 1.6565786098999745e-05, + "loss": 0.2144, + "step": 52194 + }, + { + "epoch": 2.01, + "learning_rate": 1.6565144908951014e-05, + "loss": 0.1624, + "step": 52195 + }, + { + "epoch": 2.01, + "learning_rate": 1.6564503718902282e-05, + "loss": 0.2912, + "step": 52196 + }, + { + "epoch": 2.01, + "learning_rate": 1.656386252885355e-05, + "loss": 0.273, + "step": 52197 + }, + { + "epoch": 2.01, + "learning_rate": 1.6563221338804822e-05, + "loss": 0.423, + "step": 52198 + }, + { + "epoch": 2.01, + "learning_rate": 1.6562580148756094e-05, + "loss": 0.1273, + "step": 52199 + }, + { + "epoch": 2.01, + "learning_rate": 1.6561938958707362e-05, + "loss": 0.3272, + "step": 52200 + }, + { + "epoch": 2.01, + "learning_rate": 1.656129776865863e-05, + "loss": 0.4221, + "step": 52201 + }, + { + "epoch": 2.01, + "learning_rate": 1.6560656578609902e-05, + "loss": 0.2268, + "step": 52202 + }, + { + "epoch": 2.01, + "learning_rate": 1.656001538856117e-05, + "loss": 0.1182, + "step": 52203 + }, + { + "epoch": 2.01, + "learning_rate": 1.655937419851244e-05, + "loss": 0.2061, + "step": 52204 + }, + { + "epoch": 2.01, + "learning_rate": 1.6558733008463707e-05, + "loss": 0.1853, + "step": 52205 + }, + { + "epoch": 2.01, + "learning_rate": 1.655809181841498e-05, + "loss": 0.1225, + "step": 52206 + }, + { + "epoch": 2.01, + "learning_rate": 1.6557450628366247e-05, + "loss": 0.2942, + "step": 52207 + }, + { + "epoch": 2.01, + "learning_rate": 1.655680943831752e-05, + "loss": 0.2692, + "step": 52208 + }, + { + "epoch": 2.01, + "learning_rate": 1.6556168248268787e-05, + "loss": 0.2775, + "step": 52209 + }, + { + "epoch": 2.01, + "learning_rate": 1.655552705822006e-05, + "loss": 0.2671, + "step": 52210 + }, + { + "epoch": 2.01, + "learning_rate": 1.6554885868171327e-05, + "loss": 0.2251, + "step": 52211 + }, + { + "epoch": 2.01, + "learning_rate": 1.6554244678122596e-05, + "loss": 0.3033, + "step": 52212 + }, + { + "epoch": 2.01, + "learning_rate": 1.6553603488073864e-05, + "loss": 0.3424, + "step": 52213 + }, + { + "epoch": 2.01, + "learning_rate": 1.6552962298025136e-05, + "loss": 0.2749, + "step": 52214 + }, + { + "epoch": 2.01, + "learning_rate": 1.6552321107976404e-05, + "loss": 0.1213, + "step": 52215 + }, + { + "epoch": 2.01, + "learning_rate": 1.6551679917927672e-05, + "loss": 0.126, + "step": 52216 + }, + { + "epoch": 2.01, + "learning_rate": 1.6551038727878944e-05, + "loss": 0.257, + "step": 52217 + }, + { + "epoch": 2.01, + "learning_rate": 1.6550397537830216e-05, + "loss": 0.3855, + "step": 52218 + }, + { + "epoch": 2.01, + "learning_rate": 1.6549756347781484e-05, + "loss": 0.1912, + "step": 52219 + }, + { + "epoch": 2.01, + "learning_rate": 1.6549115157732752e-05, + "loss": 0.3741, + "step": 52220 + }, + { + "epoch": 2.01, + "learning_rate": 1.654847396768402e-05, + "loss": 0.211, + "step": 52221 + }, + { + "epoch": 2.01, + "learning_rate": 1.6547832777635292e-05, + "loss": 0.1566, + "step": 52222 + }, + { + "epoch": 2.01, + "learning_rate": 1.654719158758656e-05, + "loss": 0.3234, + "step": 52223 + }, + { + "epoch": 2.01, + "learning_rate": 1.654655039753783e-05, + "loss": 0.234, + "step": 52224 + }, + { + "epoch": 2.01, + "learning_rate": 1.65459092074891e-05, + "loss": 0.1899, + "step": 52225 + }, + { + "epoch": 2.01, + "learning_rate": 1.6545268017440373e-05, + "loss": 0.2692, + "step": 52226 + }, + { + "epoch": 2.01, + "learning_rate": 1.654462682739164e-05, + "loss": 0.2241, + "step": 52227 + }, + { + "epoch": 2.01, + "learning_rate": 1.654398563734291e-05, + "loss": 0.3816, + "step": 52228 + }, + { + "epoch": 2.01, + "learning_rate": 1.6543344447294178e-05, + "loss": 0.1961, + "step": 52229 + }, + { + "epoch": 2.01, + "learning_rate": 1.654270325724545e-05, + "loss": 0.1599, + "step": 52230 + }, + { + "epoch": 2.01, + "learning_rate": 1.6542062067196718e-05, + "loss": 0.1869, + "step": 52231 + }, + { + "epoch": 2.01, + "learning_rate": 1.6541420877147986e-05, + "loss": 0.2517, + "step": 52232 + }, + { + "epoch": 2.01, + "learning_rate": 1.6540779687099258e-05, + "loss": 0.1901, + "step": 52233 + }, + { + "epoch": 2.01, + "learning_rate": 1.6540138497050526e-05, + "loss": 0.1374, + "step": 52234 + }, + { + "epoch": 2.01, + "learning_rate": 1.6539497307001798e-05, + "loss": 0.3507, + "step": 52235 + }, + { + "epoch": 2.01, + "learning_rate": 1.6538856116953066e-05, + "loss": 0.2468, + "step": 52236 + }, + { + "epoch": 2.01, + "learning_rate": 1.6538214926904334e-05, + "loss": 0.2697, + "step": 52237 + }, + { + "epoch": 2.01, + "learning_rate": 1.6537573736855606e-05, + "loss": 0.2811, + "step": 52238 + }, + { + "epoch": 2.01, + "learning_rate": 1.6536932546806874e-05, + "loss": 0.1774, + "step": 52239 + }, + { + "epoch": 2.01, + "learning_rate": 1.6536291356758143e-05, + "loss": 0.2485, + "step": 52240 + }, + { + "epoch": 2.01, + "learning_rate": 1.653565016670941e-05, + "loss": 0.2342, + "step": 52241 + }, + { + "epoch": 2.01, + "learning_rate": 1.6535008976660683e-05, + "loss": 0.3067, + "step": 52242 + }, + { + "epoch": 2.01, + "learning_rate": 1.6534367786611954e-05, + "loss": 0.2354, + "step": 52243 + }, + { + "epoch": 2.01, + "learning_rate": 1.6533726596563223e-05, + "loss": 0.1643, + "step": 52244 + }, + { + "epoch": 2.01, + "learning_rate": 1.653308540651449e-05, + "loss": 0.1242, + "step": 52245 + }, + { + "epoch": 2.01, + "learning_rate": 1.6532444216465763e-05, + "loss": 0.3055, + "step": 52246 + }, + { + "epoch": 2.01, + "learning_rate": 1.653180302641703e-05, + "loss": 0.2186, + "step": 52247 + }, + { + "epoch": 2.01, + "learning_rate": 1.65311618363683e-05, + "loss": 0.2219, + "step": 52248 + }, + { + "epoch": 2.01, + "learning_rate": 1.6530520646319568e-05, + "loss": 0.2881, + "step": 52249 + }, + { + "epoch": 2.01, + "learning_rate": 1.652987945627084e-05, + "loss": 0.1669, + "step": 52250 + }, + { + "epoch": 2.01, + "learning_rate": 1.6529238266222108e-05, + "loss": 0.2353, + "step": 52251 + }, + { + "epoch": 2.01, + "learning_rate": 1.652859707617338e-05, + "loss": 0.2765, + "step": 52252 + }, + { + "epoch": 2.01, + "learning_rate": 1.6527955886124648e-05, + "loss": 0.1364, + "step": 52253 + }, + { + "epoch": 2.01, + "learning_rate": 1.652731469607592e-05, + "loss": 0.2101, + "step": 52254 + }, + { + "epoch": 2.01, + "learning_rate": 1.6526673506027188e-05, + "loss": 0.2026, + "step": 52255 + }, + { + "epoch": 2.01, + "learning_rate": 1.6526032315978456e-05, + "loss": 0.2543, + "step": 52256 + }, + { + "epoch": 2.01, + "learning_rate": 1.6525391125929725e-05, + "loss": 0.2002, + "step": 52257 + }, + { + "epoch": 2.01, + "learning_rate": 1.6524749935880996e-05, + "loss": 0.2102, + "step": 52258 + }, + { + "epoch": 2.01, + "learning_rate": 1.6524108745832265e-05, + "loss": 0.2361, + "step": 52259 + }, + { + "epoch": 2.01, + "learning_rate": 1.6523467555783533e-05, + "loss": 0.4007, + "step": 52260 + }, + { + "epoch": 2.01, + "learning_rate": 1.6522826365734805e-05, + "loss": 0.2383, + "step": 52261 + }, + { + "epoch": 2.01, + "learning_rate": 1.6522185175686076e-05, + "loss": 0.2565, + "step": 52262 + }, + { + "epoch": 2.01, + "learning_rate": 1.6521543985637345e-05, + "loss": 0.1791, + "step": 52263 + }, + { + "epoch": 2.01, + "learning_rate": 1.6520902795588613e-05, + "loss": 0.3471, + "step": 52264 + }, + { + "epoch": 2.01, + "learning_rate": 1.652026160553988e-05, + "loss": 0.3322, + "step": 52265 + }, + { + "epoch": 2.01, + "learning_rate": 1.6519620415491153e-05, + "loss": 0.2311, + "step": 52266 + }, + { + "epoch": 2.01, + "learning_rate": 1.651897922544242e-05, + "loss": 0.3693, + "step": 52267 + }, + { + "epoch": 2.01, + "learning_rate": 1.651833803539369e-05, + "loss": 0.2196, + "step": 52268 + }, + { + "epoch": 2.01, + "learning_rate": 1.651769684534496e-05, + "loss": 0.3392, + "step": 52269 + }, + { + "epoch": 2.01, + "learning_rate": 1.6517055655296233e-05, + "loss": 0.2322, + "step": 52270 + }, + { + "epoch": 2.01, + "learning_rate": 1.65164144652475e-05, + "loss": 0.238, + "step": 52271 + }, + { + "epoch": 2.01, + "learning_rate": 1.651577327519877e-05, + "loss": 0.1978, + "step": 52272 + }, + { + "epoch": 2.01, + "learning_rate": 1.6515132085150038e-05, + "loss": 0.173, + "step": 52273 + }, + { + "epoch": 2.01, + "learning_rate": 1.651449089510131e-05, + "loss": 0.2695, + "step": 52274 + }, + { + "epoch": 2.01, + "learning_rate": 1.6513849705052578e-05, + "loss": 0.2381, + "step": 52275 + }, + { + "epoch": 2.01, + "learning_rate": 1.6513208515003847e-05, + "loss": 0.306, + "step": 52276 + }, + { + "epoch": 2.01, + "learning_rate": 1.6512567324955115e-05, + "loss": 0.239, + "step": 52277 + }, + { + "epoch": 2.01, + "learning_rate": 1.6511926134906387e-05, + "loss": 0.3069, + "step": 52278 + }, + { + "epoch": 2.01, + "learning_rate": 1.651128494485766e-05, + "loss": 0.3074, + "step": 52279 + }, + { + "epoch": 2.01, + "learning_rate": 1.6510643754808927e-05, + "loss": 0.2022, + "step": 52280 + }, + { + "epoch": 2.01, + "learning_rate": 1.6510002564760195e-05, + "loss": 0.2435, + "step": 52281 + }, + { + "epoch": 2.01, + "learning_rate": 1.6509361374711467e-05, + "loss": 0.1119, + "step": 52282 + }, + { + "epoch": 2.01, + "learning_rate": 1.6508720184662735e-05, + "loss": 0.2203, + "step": 52283 + }, + { + "epoch": 2.01, + "learning_rate": 1.6508078994614003e-05, + "loss": 0.2122, + "step": 52284 + }, + { + "epoch": 2.01, + "learning_rate": 1.6507437804565272e-05, + "loss": 0.1168, + "step": 52285 + }, + { + "epoch": 2.01, + "learning_rate": 1.6506796614516543e-05, + "loss": 0.2344, + "step": 52286 + }, + { + "epoch": 2.01, + "learning_rate": 1.6506155424467812e-05, + "loss": 0.1773, + "step": 52287 + }, + { + "epoch": 2.01, + "learning_rate": 1.6505514234419083e-05, + "loss": 0.2158, + "step": 52288 + }, + { + "epoch": 2.01, + "learning_rate": 1.6504873044370352e-05, + "loss": 0.1161, + "step": 52289 + }, + { + "epoch": 2.01, + "learning_rate": 1.6504231854321624e-05, + "loss": 0.2267, + "step": 52290 + }, + { + "epoch": 2.01, + "learning_rate": 1.6503590664272892e-05, + "loss": 0.1634, + "step": 52291 + }, + { + "epoch": 2.01, + "learning_rate": 1.650294947422416e-05, + "loss": 0.2035, + "step": 52292 + }, + { + "epoch": 2.01, + "learning_rate": 1.650230828417543e-05, + "loss": 0.3173, + "step": 52293 + }, + { + "epoch": 2.01, + "learning_rate": 1.65016670941267e-05, + "loss": 0.2436, + "step": 52294 + }, + { + "epoch": 2.01, + "learning_rate": 1.650102590407797e-05, + "loss": 0.1659, + "step": 52295 + }, + { + "epoch": 2.01, + "learning_rate": 1.6500384714029237e-05, + "loss": 0.1952, + "step": 52296 + }, + { + "epoch": 2.01, + "learning_rate": 1.649974352398051e-05, + "loss": 0.313, + "step": 52297 + }, + { + "epoch": 2.01, + "learning_rate": 1.649910233393178e-05, + "loss": 0.2457, + "step": 52298 + }, + { + "epoch": 2.01, + "learning_rate": 1.649846114388305e-05, + "loss": 0.2362, + "step": 52299 + }, + { + "epoch": 2.01, + "learning_rate": 1.6497819953834317e-05, + "loss": 0.2668, + "step": 52300 + }, + { + "epoch": 2.01, + "learning_rate": 1.6497178763785585e-05, + "loss": 0.318, + "step": 52301 + }, + { + "epoch": 2.01, + "learning_rate": 1.6496537573736857e-05, + "loss": 0.1908, + "step": 52302 + }, + { + "epoch": 2.01, + "learning_rate": 1.6495896383688125e-05, + "loss": 0.2175, + "step": 52303 + }, + { + "epoch": 2.01, + "learning_rate": 1.6495255193639394e-05, + "loss": 0.2169, + "step": 52304 + }, + { + "epoch": 2.01, + "learning_rate": 1.6494614003590665e-05, + "loss": 0.2344, + "step": 52305 + }, + { + "epoch": 2.01, + "learning_rate": 1.6493972813541937e-05, + "loss": 0.2825, + "step": 52306 + }, + { + "epoch": 2.01, + "learning_rate": 1.6493331623493205e-05, + "loss": 0.2083, + "step": 52307 + }, + { + "epoch": 2.01, + "learning_rate": 1.6492690433444474e-05, + "loss": 0.1925, + "step": 52308 + }, + { + "epoch": 2.01, + "learning_rate": 1.6492049243395742e-05, + "loss": 0.2294, + "step": 52309 + }, + { + "epoch": 2.01, + "learning_rate": 1.6491408053347014e-05, + "loss": 0.3539, + "step": 52310 + }, + { + "epoch": 2.01, + "learning_rate": 1.6490766863298282e-05, + "loss": 0.2152, + "step": 52311 + }, + { + "epoch": 2.01, + "learning_rate": 1.649012567324955e-05, + "loss": 0.1772, + "step": 52312 + }, + { + "epoch": 2.01, + "learning_rate": 1.6489484483200822e-05, + "loss": 0.1796, + "step": 52313 + }, + { + "epoch": 2.01, + "learning_rate": 1.648884329315209e-05, + "loss": 0.2547, + "step": 52314 + }, + { + "epoch": 2.01, + "learning_rate": 1.6488202103103362e-05, + "loss": 0.1909, + "step": 52315 + }, + { + "epoch": 2.01, + "learning_rate": 1.648756091305463e-05, + "loss": 0.1475, + "step": 52316 + }, + { + "epoch": 2.01, + "learning_rate": 1.64869197230059e-05, + "loss": 0.0987, + "step": 52317 + }, + { + "epoch": 2.01, + "learning_rate": 1.648627853295717e-05, + "loss": 0.3618, + "step": 52318 + }, + { + "epoch": 2.01, + "learning_rate": 1.648563734290844e-05, + "loss": 0.1595, + "step": 52319 + }, + { + "epoch": 2.01, + "learning_rate": 1.6484996152859707e-05, + "loss": 0.1511, + "step": 52320 + }, + { + "epoch": 2.01, + "learning_rate": 1.6484354962810976e-05, + "loss": 0.2836, + "step": 52321 + }, + { + "epoch": 2.01, + "learning_rate": 1.6483713772762247e-05, + "loss": 0.2569, + "step": 52322 + }, + { + "epoch": 2.01, + "learning_rate": 1.6483072582713516e-05, + "loss": 0.2467, + "step": 52323 + }, + { + "epoch": 2.01, + "learning_rate": 1.6482431392664787e-05, + "loss": 0.1919, + "step": 52324 + }, + { + "epoch": 2.01, + "learning_rate": 1.6481790202616056e-05, + "loss": 0.2359, + "step": 52325 + }, + { + "epoch": 2.01, + "learning_rate": 1.6481149012567327e-05, + "loss": 0.2383, + "step": 52326 + }, + { + "epoch": 2.01, + "learning_rate": 1.6480507822518596e-05, + "loss": 0.3221, + "step": 52327 + }, + { + "epoch": 2.01, + "learning_rate": 1.6479866632469864e-05, + "loss": 0.2042, + "step": 52328 + }, + { + "epoch": 2.01, + "learning_rate": 1.6479225442421132e-05, + "loss": 0.2955, + "step": 52329 + }, + { + "epoch": 2.01, + "learning_rate": 1.6478584252372404e-05, + "loss": 0.1679, + "step": 52330 + }, + { + "epoch": 2.01, + "learning_rate": 1.6477943062323672e-05, + "loss": 0.1463, + "step": 52331 + }, + { + "epoch": 2.01, + "learning_rate": 1.647730187227494e-05, + "loss": 0.2282, + "step": 52332 + }, + { + "epoch": 2.01, + "learning_rate": 1.6476660682226212e-05, + "loss": 0.1685, + "step": 52333 + }, + { + "epoch": 2.01, + "learning_rate": 1.6476019492177484e-05, + "loss": 0.2518, + "step": 52334 + }, + { + "epoch": 2.01, + "learning_rate": 1.6475378302128753e-05, + "loss": 0.1741, + "step": 52335 + }, + { + "epoch": 2.01, + "learning_rate": 1.647473711208002e-05, + "loss": 0.1993, + "step": 52336 + }, + { + "epoch": 2.01, + "learning_rate": 1.647409592203129e-05, + "loss": 0.1679, + "step": 52337 + }, + { + "epoch": 2.01, + "learning_rate": 1.647345473198256e-05, + "loss": 0.2573, + "step": 52338 + }, + { + "epoch": 2.01, + "learning_rate": 1.647281354193383e-05, + "loss": 0.2645, + "step": 52339 + }, + { + "epoch": 2.01, + "learning_rate": 1.6472172351885098e-05, + "loss": 0.2786, + "step": 52340 + }, + { + "epoch": 2.01, + "learning_rate": 1.647153116183637e-05, + "loss": 0.2361, + "step": 52341 + }, + { + "epoch": 2.01, + "learning_rate": 1.647088997178764e-05, + "loss": 0.3454, + "step": 52342 + }, + { + "epoch": 2.01, + "learning_rate": 1.647024878173891e-05, + "loss": 0.241, + "step": 52343 + }, + { + "epoch": 2.01, + "learning_rate": 1.6469607591690178e-05, + "loss": 0.2845, + "step": 52344 + }, + { + "epoch": 2.01, + "learning_rate": 1.6468966401641446e-05, + "loss": 0.0994, + "step": 52345 + }, + { + "epoch": 2.01, + "learning_rate": 1.6468325211592718e-05, + "loss": 0.2298, + "step": 52346 + }, + { + "epoch": 2.01, + "learning_rate": 1.6467684021543986e-05, + "loss": 0.26, + "step": 52347 + }, + { + "epoch": 2.01, + "learning_rate": 1.6467042831495254e-05, + "loss": 0.1595, + "step": 52348 + }, + { + "epoch": 2.01, + "learning_rate": 1.6466401641446526e-05, + "loss": 0.1846, + "step": 52349 + }, + { + "epoch": 2.01, + "learning_rate": 1.6465760451397798e-05, + "loss": 0.2745, + "step": 52350 + }, + { + "epoch": 2.01, + "learning_rate": 1.6465119261349066e-05, + "loss": 0.1036, + "step": 52351 + }, + { + "epoch": 2.01, + "learning_rate": 1.6464478071300334e-05, + "loss": 0.3682, + "step": 52352 + }, + { + "epoch": 2.01, + "learning_rate": 1.6463836881251603e-05, + "loss": 0.1423, + "step": 52353 + }, + { + "epoch": 2.01, + "learning_rate": 1.6463195691202874e-05, + "loss": 0.2133, + "step": 52354 + }, + { + "epoch": 2.01, + "learning_rate": 1.6462554501154143e-05, + "loss": 0.2439, + "step": 52355 + }, + { + "epoch": 2.01, + "learning_rate": 1.646191331110541e-05, + "loss": 0.1768, + "step": 52356 + }, + { + "epoch": 2.01, + "learning_rate": 1.6461272121056683e-05, + "loss": 0.2537, + "step": 52357 + }, + { + "epoch": 2.01, + "learning_rate": 1.646063093100795e-05, + "loss": 0.3072, + "step": 52358 + }, + { + "epoch": 2.01, + "learning_rate": 1.6459989740959223e-05, + "loss": 0.1829, + "step": 52359 + }, + { + "epoch": 2.01, + "learning_rate": 1.645934855091049e-05, + "loss": 0.2053, + "step": 52360 + }, + { + "epoch": 2.01, + "learning_rate": 1.645870736086176e-05, + "loss": 0.2322, + "step": 52361 + }, + { + "epoch": 2.01, + "learning_rate": 1.645806617081303e-05, + "loss": 0.2364, + "step": 52362 + }, + { + "epoch": 2.01, + "learning_rate": 1.64574249807643e-05, + "loss": 0.1406, + "step": 52363 + }, + { + "epoch": 2.01, + "learning_rate": 1.6456783790715568e-05, + "loss": 0.2999, + "step": 52364 + }, + { + "epoch": 2.01, + "learning_rate": 1.6456142600666836e-05, + "loss": 0.1683, + "step": 52365 + }, + { + "epoch": 2.01, + "learning_rate": 1.6455501410618108e-05, + "loss": 0.2098, + "step": 52366 + }, + { + "epoch": 2.01, + "learning_rate": 1.6454860220569376e-05, + "loss": 0.2228, + "step": 52367 + }, + { + "epoch": 2.01, + "learning_rate": 1.6454219030520648e-05, + "loss": 0.3274, + "step": 52368 + }, + { + "epoch": 2.01, + "learning_rate": 1.6453577840471916e-05, + "loss": 0.2542, + "step": 52369 + }, + { + "epoch": 2.01, + "learning_rate": 1.6452936650423188e-05, + "loss": 0.3052, + "step": 52370 + }, + { + "epoch": 2.01, + "learning_rate": 1.6452295460374456e-05, + "loss": 0.1908, + "step": 52371 + }, + { + "epoch": 2.01, + "learning_rate": 1.6451654270325725e-05, + "loss": 0.1944, + "step": 52372 + }, + { + "epoch": 2.01, + "learning_rate": 1.6451013080276993e-05, + "loss": 0.2844, + "step": 52373 + }, + { + "epoch": 2.01, + "learning_rate": 1.6450371890228265e-05, + "loss": 0.1489, + "step": 52374 + }, + { + "epoch": 2.01, + "learning_rate": 1.6449730700179533e-05, + "loss": 0.267, + "step": 52375 + }, + { + "epoch": 2.01, + "learning_rate": 1.64490895101308e-05, + "loss": 0.113, + "step": 52376 + }, + { + "epoch": 2.01, + "learning_rate": 1.6448448320082073e-05, + "loss": 0.3642, + "step": 52377 + }, + { + "epoch": 2.01, + "learning_rate": 1.6447807130033345e-05, + "loss": 0.1432, + "step": 52378 + }, + { + "epoch": 2.01, + "learning_rate": 1.6447165939984613e-05, + "loss": 0.2095, + "step": 52379 + }, + { + "epoch": 2.01, + "learning_rate": 1.644652474993588e-05, + "loss": 0.2048, + "step": 52380 + }, + { + "epoch": 2.01, + "learning_rate": 1.644588355988715e-05, + "loss": 0.3261, + "step": 52381 + }, + { + "epoch": 2.01, + "learning_rate": 1.644524236983842e-05, + "loss": 0.2695, + "step": 52382 + }, + { + "epoch": 2.01, + "learning_rate": 1.644460117978969e-05, + "loss": 0.1586, + "step": 52383 + }, + { + "epoch": 2.01, + "learning_rate": 1.6443959989740958e-05, + "loss": 0.184, + "step": 52384 + }, + { + "epoch": 2.01, + "learning_rate": 1.644331879969223e-05, + "loss": 0.2225, + "step": 52385 + }, + { + "epoch": 2.01, + "learning_rate": 1.64426776096435e-05, + "loss": 0.2256, + "step": 52386 + }, + { + "epoch": 2.01, + "learning_rate": 1.644203641959477e-05, + "loss": 0.2718, + "step": 52387 + }, + { + "epoch": 2.01, + "learning_rate": 1.6441395229546038e-05, + "loss": 0.2364, + "step": 52388 + }, + { + "epoch": 2.01, + "learning_rate": 1.6440754039497307e-05, + "loss": 0.3298, + "step": 52389 + }, + { + "epoch": 2.01, + "learning_rate": 1.644011284944858e-05, + "loss": 0.1676, + "step": 52390 + }, + { + "epoch": 2.01, + "learning_rate": 1.6439471659399847e-05, + "loss": 0.1634, + "step": 52391 + }, + { + "epoch": 2.01, + "learning_rate": 1.6438830469351115e-05, + "loss": 0.4177, + "step": 52392 + }, + { + "epoch": 2.01, + "learning_rate": 1.6438189279302387e-05, + "loss": 0.2688, + "step": 52393 + }, + { + "epoch": 2.01, + "learning_rate": 1.6437548089253655e-05, + "loss": 0.1713, + "step": 52394 + }, + { + "epoch": 2.01, + "learning_rate": 1.6436906899204927e-05, + "loss": 0.3435, + "step": 52395 + }, + { + "epoch": 2.01, + "learning_rate": 1.6436265709156195e-05, + "loss": 0.1782, + "step": 52396 + }, + { + "epoch": 2.01, + "learning_rate": 1.6435624519107463e-05, + "loss": 0.2204, + "step": 52397 + }, + { + "epoch": 2.01, + "learning_rate": 1.6434983329058735e-05, + "loss": 0.2869, + "step": 52398 + }, + { + "epoch": 2.01, + "learning_rate": 1.6434342139010003e-05, + "loss": 0.324, + "step": 52399 + }, + { + "epoch": 2.01, + "learning_rate": 1.6433700948961272e-05, + "loss": 0.1548, + "step": 52400 + }, + { + "epoch": 2.01, + "learning_rate": 1.643305975891254e-05, + "loss": 0.3892, + "step": 52401 + }, + { + "epoch": 2.01, + "learning_rate": 1.6432418568863812e-05, + "loss": 0.209, + "step": 52402 + }, + { + "epoch": 2.01, + "learning_rate": 1.643177737881508e-05, + "loss": 0.2548, + "step": 52403 + }, + { + "epoch": 2.01, + "learning_rate": 1.6431136188766352e-05, + "loss": 0.3233, + "step": 52404 + }, + { + "epoch": 2.01, + "learning_rate": 1.643049499871762e-05, + "loss": 0.2332, + "step": 52405 + }, + { + "epoch": 2.01, + "learning_rate": 1.6429853808668892e-05, + "loss": 0.3638, + "step": 52406 + }, + { + "epoch": 2.01, + "learning_rate": 1.642921261862016e-05, + "loss": 0.3234, + "step": 52407 + }, + { + "epoch": 2.01, + "learning_rate": 1.642857142857143e-05, + "loss": 0.2772, + "step": 52408 + }, + { + "epoch": 2.01, + "learning_rate": 1.6427930238522697e-05, + "loss": 0.3288, + "step": 52409 + }, + { + "epoch": 2.01, + "learning_rate": 1.642728904847397e-05, + "loss": 0.1575, + "step": 52410 + }, + { + "epoch": 2.02, + "learning_rate": 1.6426647858425237e-05, + "loss": 0.2803, + "step": 52411 + }, + { + "epoch": 2.02, + "learning_rate": 1.6426006668376505e-05, + "loss": 0.1785, + "step": 52412 + }, + { + "epoch": 2.02, + "learning_rate": 1.6425365478327777e-05, + "loss": 0.2641, + "step": 52413 + }, + { + "epoch": 2.02, + "learning_rate": 1.642472428827905e-05, + "loss": 0.2249, + "step": 52414 + }, + { + "epoch": 2.02, + "learning_rate": 1.6424083098230317e-05, + "loss": 0.2943, + "step": 52415 + }, + { + "epoch": 2.02, + "learning_rate": 1.6423441908181585e-05, + "loss": 0.2272, + "step": 52416 + }, + { + "epoch": 2.02, + "learning_rate": 1.6422800718132854e-05, + "loss": 0.2913, + "step": 52417 + }, + { + "epoch": 2.02, + "learning_rate": 1.6422159528084125e-05, + "loss": 0.3143, + "step": 52418 + }, + { + "epoch": 2.02, + "learning_rate": 1.6421518338035394e-05, + "loss": 0.2482, + "step": 52419 + }, + { + "epoch": 2.02, + "learning_rate": 1.6420877147986662e-05, + "loss": 0.2177, + "step": 52420 + }, + { + "epoch": 2.02, + "learning_rate": 1.6420235957937934e-05, + "loss": 0.3585, + "step": 52421 + }, + { + "epoch": 2.02, + "learning_rate": 1.6419594767889206e-05, + "loss": 0.2192, + "step": 52422 + }, + { + "epoch": 2.02, + "learning_rate": 1.6418953577840474e-05, + "loss": 0.213, + "step": 52423 + }, + { + "epoch": 2.02, + "learning_rate": 1.6418312387791742e-05, + "loss": 0.2115, + "step": 52424 + }, + { + "epoch": 2.02, + "learning_rate": 1.641767119774301e-05, + "loss": 0.1641, + "step": 52425 + }, + { + "epoch": 2.02, + "learning_rate": 1.6417030007694282e-05, + "loss": 0.2285, + "step": 52426 + }, + { + "epoch": 2.02, + "learning_rate": 1.641638881764555e-05, + "loss": 0.2398, + "step": 52427 + }, + { + "epoch": 2.02, + "learning_rate": 1.641574762759682e-05, + "loss": 0.3518, + "step": 52428 + }, + { + "epoch": 2.02, + "learning_rate": 1.641510643754809e-05, + "loss": 0.1809, + "step": 52429 + }, + { + "epoch": 2.02, + "learning_rate": 1.641446524749936e-05, + "loss": 0.1177, + "step": 52430 + }, + { + "epoch": 2.02, + "learning_rate": 1.641382405745063e-05, + "loss": 0.1939, + "step": 52431 + }, + { + "epoch": 2.02, + "learning_rate": 1.64131828674019e-05, + "loss": 0.2208, + "step": 52432 + }, + { + "epoch": 2.02, + "learning_rate": 1.6412541677353167e-05, + "loss": 0.2443, + "step": 52433 + }, + { + "epoch": 2.02, + "learning_rate": 1.641190048730444e-05, + "loss": 0.1486, + "step": 52434 + }, + { + "epoch": 2.02, + "learning_rate": 1.6411259297255707e-05, + "loss": 0.3309, + "step": 52435 + }, + { + "epoch": 2.02, + "learning_rate": 1.6410618107206976e-05, + "loss": 0.1587, + "step": 52436 + }, + { + "epoch": 2.02, + "learning_rate": 1.6409976917158247e-05, + "loss": 0.1489, + "step": 52437 + }, + { + "epoch": 2.02, + "learning_rate": 1.6409335727109516e-05, + "loss": 0.3611, + "step": 52438 + }, + { + "epoch": 2.02, + "learning_rate": 1.6408694537060787e-05, + "loss": 0.1984, + "step": 52439 + }, + { + "epoch": 2.02, + "learning_rate": 1.6408053347012056e-05, + "loss": 0.2398, + "step": 52440 + }, + { + "epoch": 2.02, + "learning_rate": 1.6407412156963324e-05, + "loss": 0.3442, + "step": 52441 + }, + { + "epoch": 2.02, + "learning_rate": 1.6406770966914596e-05, + "loss": 0.1608, + "step": 52442 + }, + { + "epoch": 2.02, + "learning_rate": 1.6406129776865864e-05, + "loss": 0.3425, + "step": 52443 + }, + { + "epoch": 2.02, + "learning_rate": 1.6405488586817132e-05, + "loss": 0.2128, + "step": 52444 + }, + { + "epoch": 2.02, + "learning_rate": 1.64048473967684e-05, + "loss": 0.1921, + "step": 52445 + }, + { + "epoch": 2.02, + "learning_rate": 1.6404206206719673e-05, + "loss": 0.2091, + "step": 52446 + }, + { + "epoch": 2.02, + "learning_rate": 1.640356501667094e-05, + "loss": 0.2194, + "step": 52447 + }, + { + "epoch": 2.02, + "learning_rate": 1.6402923826622213e-05, + "loss": 0.1364, + "step": 52448 + }, + { + "epoch": 2.02, + "learning_rate": 1.640228263657348e-05, + "loss": 0.1961, + "step": 52449 + }, + { + "epoch": 2.02, + "learning_rate": 1.6401641446524753e-05, + "loss": 0.2204, + "step": 52450 + }, + { + "epoch": 2.02, + "learning_rate": 1.640100025647602e-05, + "loss": 0.3382, + "step": 52451 + }, + { + "epoch": 2.02, + "learning_rate": 1.640035906642729e-05, + "loss": 0.1825, + "step": 52452 + }, + { + "epoch": 2.02, + "learning_rate": 1.6399717876378558e-05, + "loss": 0.1805, + "step": 52453 + }, + { + "epoch": 2.02, + "learning_rate": 1.639907668632983e-05, + "loss": 0.3062, + "step": 52454 + }, + { + "epoch": 2.02, + "learning_rate": 1.6398435496281098e-05, + "loss": 0.2546, + "step": 52455 + }, + { + "epoch": 2.02, + "learning_rate": 1.6397794306232366e-05, + "loss": 0.1792, + "step": 52456 + }, + { + "epoch": 2.02, + "learning_rate": 1.6397153116183638e-05, + "loss": 0.2788, + "step": 52457 + }, + { + "epoch": 2.02, + "learning_rate": 1.639651192613491e-05, + "loss": 0.1641, + "step": 52458 + }, + { + "epoch": 2.02, + "learning_rate": 1.6395870736086178e-05, + "loss": 0.3163, + "step": 52459 + }, + { + "epoch": 2.02, + "learning_rate": 1.6395229546037446e-05, + "loss": 0.337, + "step": 52460 + }, + { + "epoch": 2.02, + "learning_rate": 1.6394588355988714e-05, + "loss": 0.226, + "step": 52461 + }, + { + "epoch": 2.02, + "learning_rate": 1.6393947165939986e-05, + "loss": 0.2374, + "step": 52462 + }, + { + "epoch": 2.02, + "learning_rate": 1.6393305975891254e-05, + "loss": 0.1559, + "step": 52463 + }, + { + "epoch": 2.02, + "learning_rate": 1.6392664785842523e-05, + "loss": 0.2992, + "step": 52464 + }, + { + "epoch": 2.02, + "learning_rate": 1.6392023595793794e-05, + "loss": 0.2318, + "step": 52465 + }, + { + "epoch": 2.02, + "learning_rate": 1.6391382405745066e-05, + "loss": 0.2764, + "step": 52466 + }, + { + "epoch": 2.02, + "learning_rate": 1.6390741215696335e-05, + "loss": 0.176, + "step": 52467 + }, + { + "epoch": 2.02, + "learning_rate": 1.6390100025647603e-05, + "loss": 0.166, + "step": 52468 + }, + { + "epoch": 2.02, + "learning_rate": 1.638945883559887e-05, + "loss": 0.3383, + "step": 52469 + }, + { + "epoch": 2.02, + "learning_rate": 1.6388817645550143e-05, + "loss": 0.2635, + "step": 52470 + }, + { + "epoch": 2.02, + "learning_rate": 1.638817645550141e-05, + "loss": 0.293, + "step": 52471 + }, + { + "epoch": 2.02, + "learning_rate": 1.638753526545268e-05, + "loss": 0.2806, + "step": 52472 + }, + { + "epoch": 2.02, + "learning_rate": 1.638689407540395e-05, + "loss": 0.3518, + "step": 52473 + }, + { + "epoch": 2.02, + "learning_rate": 1.638625288535522e-05, + "loss": 0.2762, + "step": 52474 + }, + { + "epoch": 2.02, + "learning_rate": 1.638561169530649e-05, + "loss": 0.2412, + "step": 52475 + }, + { + "epoch": 2.02, + "learning_rate": 1.638497050525776e-05, + "loss": 0.2291, + "step": 52476 + }, + { + "epoch": 2.02, + "learning_rate": 1.6384329315209028e-05, + "loss": 0.1856, + "step": 52477 + }, + { + "epoch": 2.02, + "learning_rate": 1.63836881251603e-05, + "loss": 0.1556, + "step": 52478 + }, + { + "epoch": 2.02, + "learning_rate": 1.6383046935111568e-05, + "loss": 0.2702, + "step": 52479 + }, + { + "epoch": 2.02, + "learning_rate": 1.6382405745062836e-05, + "loss": 0.1713, + "step": 52480 + }, + { + "epoch": 2.02, + "learning_rate": 1.6381764555014105e-05, + "loss": 0.2377, + "step": 52481 + }, + { + "epoch": 2.02, + "learning_rate": 1.6381123364965376e-05, + "loss": 0.2313, + "step": 52482 + }, + { + "epoch": 2.02, + "learning_rate": 1.6380482174916645e-05, + "loss": 0.2995, + "step": 52483 + }, + { + "epoch": 2.02, + "learning_rate": 1.6379840984867916e-05, + "loss": 0.183, + "step": 52484 + }, + { + "epoch": 2.02, + "learning_rate": 1.6379199794819185e-05, + "loss": 0.2629, + "step": 52485 + }, + { + "epoch": 2.02, + "learning_rate": 1.6378558604770456e-05, + "loss": 0.3001, + "step": 52486 + }, + { + "epoch": 2.02, + "learning_rate": 1.6377917414721725e-05, + "loss": 0.2216, + "step": 52487 + }, + { + "epoch": 2.02, + "learning_rate": 1.6377276224672993e-05, + "loss": 0.232, + "step": 52488 + }, + { + "epoch": 2.02, + "learning_rate": 1.637663503462426e-05, + "loss": 0.1911, + "step": 52489 + }, + { + "epoch": 2.02, + "learning_rate": 1.6375993844575533e-05, + "loss": 0.2157, + "step": 52490 + }, + { + "epoch": 2.02, + "learning_rate": 1.63753526545268e-05, + "loss": 0.256, + "step": 52491 + }, + { + "epoch": 2.02, + "learning_rate": 1.637471146447807e-05, + "loss": 0.2972, + "step": 52492 + }, + { + "epoch": 2.02, + "learning_rate": 1.637407027442934e-05, + "loss": 0.1627, + "step": 52493 + }, + { + "epoch": 2.02, + "learning_rate": 1.6373429084380613e-05, + "loss": 0.2691, + "step": 52494 + }, + { + "epoch": 2.02, + "learning_rate": 1.637278789433188e-05, + "loss": 0.2681, + "step": 52495 + }, + { + "epoch": 2.02, + "learning_rate": 1.637214670428315e-05, + "loss": 0.1949, + "step": 52496 + }, + { + "epoch": 2.02, + "learning_rate": 1.6371505514234418e-05, + "loss": 0.2374, + "step": 52497 + }, + { + "epoch": 2.02, + "learning_rate": 1.637086432418569e-05, + "loss": 0.3855, + "step": 52498 + }, + { + "epoch": 2.02, + "learning_rate": 1.6370223134136958e-05, + "loss": 0.2954, + "step": 52499 + }, + { + "epoch": 2.02, + "learning_rate": 1.6369581944088227e-05, + "loss": 0.2069, + "step": 52500 + }, + { + "epoch": 2.02, + "learning_rate": 1.63689407540395e-05, + "loss": 0.0992, + "step": 52501 + }, + { + "epoch": 2.02, + "learning_rate": 1.636829956399077e-05, + "loss": 0.2171, + "step": 52502 + }, + { + "epoch": 2.02, + "learning_rate": 1.636765837394204e-05, + "loss": 0.2052, + "step": 52503 + }, + { + "epoch": 2.02, + "learning_rate": 1.6367017183893307e-05, + "loss": 0.3344, + "step": 52504 + }, + { + "epoch": 2.02, + "learning_rate": 1.6366375993844575e-05, + "loss": 0.2129, + "step": 52505 + }, + { + "epoch": 2.02, + "learning_rate": 1.6365734803795847e-05, + "loss": 0.1688, + "step": 52506 + }, + { + "epoch": 2.02, + "learning_rate": 1.6365093613747115e-05, + "loss": 0.3917, + "step": 52507 + }, + { + "epoch": 2.02, + "learning_rate": 1.6364452423698383e-05, + "loss": 0.3296, + "step": 52508 + }, + { + "epoch": 2.02, + "learning_rate": 1.6363811233649655e-05, + "loss": 0.3334, + "step": 52509 + }, + { + "epoch": 2.02, + "learning_rate": 1.6363170043600923e-05, + "loss": 0.231, + "step": 52510 + }, + { + "epoch": 2.02, + "learning_rate": 1.6362528853552195e-05, + "loss": 0.2846, + "step": 52511 + }, + { + "epoch": 2.02, + "learning_rate": 1.6361887663503463e-05, + "loss": 0.3325, + "step": 52512 + }, + { + "epoch": 2.02, + "learning_rate": 1.6361246473454732e-05, + "loss": 0.1973, + "step": 52513 + }, + { + "epoch": 2.02, + "learning_rate": 1.6360605283406004e-05, + "loss": 0.1699, + "step": 52514 + }, + { + "epoch": 2.02, + "learning_rate": 1.6359964093357272e-05, + "loss": 0.2896, + "step": 52515 + }, + { + "epoch": 2.02, + "learning_rate": 1.635932290330854e-05, + "loss": 0.2758, + "step": 52516 + }, + { + "epoch": 2.02, + "learning_rate": 1.6358681713259812e-05, + "loss": 0.1742, + "step": 52517 + }, + { + "epoch": 2.02, + "learning_rate": 1.635804052321108e-05, + "loss": 0.3552, + "step": 52518 + }, + { + "epoch": 2.02, + "learning_rate": 1.635739933316235e-05, + "loss": 0.1798, + "step": 52519 + }, + { + "epoch": 2.02, + "learning_rate": 1.635675814311362e-05, + "loss": 0.1363, + "step": 52520 + }, + { + "epoch": 2.02, + "learning_rate": 1.635611695306489e-05, + "loss": 0.3277, + "step": 52521 + }, + { + "epoch": 2.02, + "learning_rate": 1.635547576301616e-05, + "loss": 0.2177, + "step": 52522 + }, + { + "epoch": 2.02, + "learning_rate": 1.635483457296743e-05, + "loss": 0.2799, + "step": 52523 + }, + { + "epoch": 2.02, + "learning_rate": 1.6354193382918697e-05, + "loss": 0.1203, + "step": 52524 + }, + { + "epoch": 2.02, + "learning_rate": 1.6353552192869965e-05, + "loss": 0.1674, + "step": 52525 + }, + { + "epoch": 2.02, + "learning_rate": 1.6352911002821237e-05, + "loss": 0.3416, + "step": 52526 + }, + { + "epoch": 2.02, + "learning_rate": 1.6352269812772505e-05, + "loss": 0.2802, + "step": 52527 + }, + { + "epoch": 2.02, + "learning_rate": 1.6351628622723777e-05, + "loss": 0.1772, + "step": 52528 + }, + { + "epoch": 2.02, + "learning_rate": 1.6350987432675045e-05, + "loss": 0.2514, + "step": 52529 + }, + { + "epoch": 2.02, + "learning_rate": 1.6350346242626317e-05, + "loss": 0.2363, + "step": 52530 + }, + { + "epoch": 2.02, + "learning_rate": 1.6349705052577585e-05, + "loss": 0.3014, + "step": 52531 + }, + { + "epoch": 2.02, + "learning_rate": 1.6349063862528854e-05, + "loss": 0.2163, + "step": 52532 + }, + { + "epoch": 2.02, + "learning_rate": 1.6348422672480122e-05, + "loss": 0.2128, + "step": 52533 + }, + { + "epoch": 2.02, + "learning_rate": 1.6347781482431394e-05, + "loss": 0.3455, + "step": 52534 + }, + { + "epoch": 2.02, + "learning_rate": 1.6347140292382662e-05, + "loss": 0.2116, + "step": 52535 + }, + { + "epoch": 2.02, + "learning_rate": 1.634649910233393e-05, + "loss": 0.2322, + "step": 52536 + }, + { + "epoch": 2.02, + "learning_rate": 1.6345857912285202e-05, + "loss": 0.322, + "step": 52537 + }, + { + "epoch": 2.02, + "learning_rate": 1.6345216722236474e-05, + "loss": 0.2064, + "step": 52538 + }, + { + "epoch": 2.02, + "learning_rate": 1.6344575532187742e-05, + "loss": 0.2563, + "step": 52539 + }, + { + "epoch": 2.02, + "learning_rate": 1.634393434213901e-05, + "loss": 0.2188, + "step": 52540 + }, + { + "epoch": 2.02, + "learning_rate": 1.634329315209028e-05, + "loss": 0.2361, + "step": 52541 + }, + { + "epoch": 2.02, + "learning_rate": 1.634265196204155e-05, + "loss": 0.2392, + "step": 52542 + }, + { + "epoch": 2.02, + "learning_rate": 1.634201077199282e-05, + "loss": 0.1904, + "step": 52543 + }, + { + "epoch": 2.02, + "learning_rate": 1.6341369581944087e-05, + "loss": 0.3204, + "step": 52544 + }, + { + "epoch": 2.02, + "learning_rate": 1.634072839189536e-05, + "loss": 0.2579, + "step": 52545 + }, + { + "epoch": 2.02, + "learning_rate": 1.634008720184663e-05, + "loss": 0.3075, + "step": 52546 + }, + { + "epoch": 2.02, + "learning_rate": 1.63394460117979e-05, + "loss": 0.3074, + "step": 52547 + }, + { + "epoch": 2.02, + "learning_rate": 1.6338804821749167e-05, + "loss": 0.1648, + "step": 52548 + }, + { + "epoch": 2.02, + "learning_rate": 1.6338163631700436e-05, + "loss": 0.2494, + "step": 52549 + }, + { + "epoch": 2.02, + "learning_rate": 1.6337522441651707e-05, + "loss": 0.2781, + "step": 52550 + }, + { + "epoch": 2.02, + "learning_rate": 1.6336881251602976e-05, + "loss": 0.2082, + "step": 52551 + }, + { + "epoch": 2.02, + "learning_rate": 1.6336240061554244e-05, + "loss": 0.1168, + "step": 52552 + }, + { + "epoch": 2.02, + "learning_rate": 1.6335598871505516e-05, + "loss": 0.2993, + "step": 52553 + }, + { + "epoch": 2.02, + "learning_rate": 1.6334957681456784e-05, + "loss": 0.2077, + "step": 52554 + }, + { + "epoch": 2.02, + "learning_rate": 1.6334316491408056e-05, + "loss": 0.3111, + "step": 52555 + }, + { + "epoch": 2.02, + "learning_rate": 1.6333675301359324e-05, + "loss": 0.3119, + "step": 52556 + }, + { + "epoch": 2.02, + "learning_rate": 1.6333034111310592e-05, + "loss": 0.2628, + "step": 52557 + }, + { + "epoch": 2.02, + "learning_rate": 1.6332392921261864e-05, + "loss": 0.3842, + "step": 52558 + }, + { + "epoch": 2.02, + "learning_rate": 1.6331751731213133e-05, + "loss": 0.2202, + "step": 52559 + }, + { + "epoch": 2.02, + "learning_rate": 1.63311105411644e-05, + "loss": 0.1768, + "step": 52560 + }, + { + "epoch": 2.02, + "learning_rate": 1.6330469351115673e-05, + "loss": 0.2531, + "step": 52561 + }, + { + "epoch": 2.02, + "learning_rate": 1.632982816106694e-05, + "loss": 0.1439, + "step": 52562 + }, + { + "epoch": 2.02, + "learning_rate": 1.632918697101821e-05, + "loss": 0.1959, + "step": 52563 + }, + { + "epoch": 2.02, + "learning_rate": 1.632854578096948e-05, + "loss": 0.2413, + "step": 52564 + }, + { + "epoch": 2.02, + "learning_rate": 1.632790459092075e-05, + "loss": 0.3232, + "step": 52565 + }, + { + "epoch": 2.02, + "learning_rate": 1.632726340087202e-05, + "loss": 0.1653, + "step": 52566 + }, + { + "epoch": 2.02, + "learning_rate": 1.632662221082329e-05, + "loss": 0.2665, + "step": 52567 + }, + { + "epoch": 2.02, + "learning_rate": 1.6325981020774558e-05, + "loss": 0.2861, + "step": 52568 + }, + { + "epoch": 2.02, + "learning_rate": 1.6325339830725826e-05, + "loss": 0.1444, + "step": 52569 + }, + { + "epoch": 2.02, + "learning_rate": 1.6324698640677098e-05, + "loss": 0.2089, + "step": 52570 + }, + { + "epoch": 2.02, + "learning_rate": 1.6324057450628366e-05, + "loss": 0.265, + "step": 52571 + }, + { + "epoch": 2.02, + "learning_rate": 1.6323416260579634e-05, + "loss": 0.3147, + "step": 52572 + }, + { + "epoch": 2.02, + "learning_rate": 1.6322775070530906e-05, + "loss": 0.2183, + "step": 52573 + }, + { + "epoch": 2.02, + "learning_rate": 1.6322133880482178e-05, + "loss": 0.3258, + "step": 52574 + }, + { + "epoch": 2.02, + "learning_rate": 1.6321492690433446e-05, + "loss": 0.1967, + "step": 52575 + }, + { + "epoch": 2.02, + "learning_rate": 1.6320851500384714e-05, + "loss": 0.2034, + "step": 52576 + }, + { + "epoch": 2.02, + "learning_rate": 1.6320210310335983e-05, + "loss": 0.1694, + "step": 52577 + }, + { + "epoch": 2.02, + "learning_rate": 1.6319569120287254e-05, + "loss": 0.2655, + "step": 52578 + }, + { + "epoch": 2.02, + "learning_rate": 1.6318927930238523e-05, + "loss": 0.234, + "step": 52579 + }, + { + "epoch": 2.02, + "learning_rate": 1.631828674018979e-05, + "loss": 0.2182, + "step": 52580 + }, + { + "epoch": 2.02, + "learning_rate": 1.6317645550141063e-05, + "loss": 0.21, + "step": 52581 + }, + { + "epoch": 2.02, + "learning_rate": 1.6317004360092335e-05, + "loss": 0.1562, + "step": 52582 + }, + { + "epoch": 2.02, + "learning_rate": 1.6316363170043603e-05, + "loss": 0.2065, + "step": 52583 + }, + { + "epoch": 2.02, + "learning_rate": 1.631572197999487e-05, + "loss": 0.111, + "step": 52584 + }, + { + "epoch": 2.02, + "learning_rate": 1.631508078994614e-05, + "loss": 0.1739, + "step": 52585 + }, + { + "epoch": 2.02, + "learning_rate": 1.631443959989741e-05, + "loss": 0.2763, + "step": 52586 + }, + { + "epoch": 2.02, + "learning_rate": 1.631379840984868e-05, + "loss": 0.2474, + "step": 52587 + }, + { + "epoch": 2.02, + "learning_rate": 1.6313157219799948e-05, + "loss": 0.3421, + "step": 52588 + }, + { + "epoch": 2.02, + "learning_rate": 1.631251602975122e-05, + "loss": 0.1469, + "step": 52589 + }, + { + "epoch": 2.02, + "learning_rate": 1.6311874839702488e-05, + "loss": 0.2837, + "step": 52590 + }, + { + "epoch": 2.02, + "learning_rate": 1.631123364965376e-05, + "loss": 0.1694, + "step": 52591 + }, + { + "epoch": 2.02, + "learning_rate": 1.6310592459605028e-05, + "loss": 0.1979, + "step": 52592 + }, + { + "epoch": 2.02, + "learning_rate": 1.6309951269556296e-05, + "loss": 0.3098, + "step": 52593 + }, + { + "epoch": 2.02, + "learning_rate": 1.6309310079507568e-05, + "loss": 0.1681, + "step": 52594 + }, + { + "epoch": 2.02, + "learning_rate": 1.6308668889458836e-05, + "loss": 0.2106, + "step": 52595 + }, + { + "epoch": 2.02, + "learning_rate": 1.6308027699410105e-05, + "loss": 0.319, + "step": 52596 + }, + { + "epoch": 2.02, + "learning_rate": 1.6307386509361376e-05, + "loss": 0.2661, + "step": 52597 + }, + { + "epoch": 2.02, + "learning_rate": 1.6306745319312645e-05, + "loss": 0.4001, + "step": 52598 + }, + { + "epoch": 2.02, + "learning_rate": 1.6306104129263913e-05, + "loss": 0.3182, + "step": 52599 + }, + { + "epoch": 2.02, + "learning_rate": 1.6305462939215185e-05, + "loss": 0.2558, + "step": 52600 + }, + { + "epoch": 2.02, + "learning_rate": 1.6304821749166453e-05, + "loss": 0.2466, + "step": 52601 + }, + { + "epoch": 2.02, + "learning_rate": 1.6304180559117725e-05, + "loss": 0.3092, + "step": 52602 + }, + { + "epoch": 2.02, + "learning_rate": 1.6303539369068993e-05, + "loss": 0.183, + "step": 52603 + }, + { + "epoch": 2.02, + "learning_rate": 1.630289817902026e-05, + "loss": 0.2502, + "step": 52604 + }, + { + "epoch": 2.02, + "learning_rate": 1.630225698897153e-05, + "loss": 0.1756, + "step": 52605 + }, + { + "epoch": 2.02, + "learning_rate": 1.63016157989228e-05, + "loss": 0.3365, + "step": 52606 + }, + { + "epoch": 2.02, + "learning_rate": 1.630097460887407e-05, + "loss": 0.2704, + "step": 52607 + }, + { + "epoch": 2.02, + "learning_rate": 1.6300333418825338e-05, + "loss": 0.1537, + "step": 52608 + }, + { + "epoch": 2.02, + "learning_rate": 1.629969222877661e-05, + "loss": 0.2199, + "step": 52609 + }, + { + "epoch": 2.02, + "learning_rate": 1.629905103872788e-05, + "loss": 0.2141, + "step": 52610 + }, + { + "epoch": 2.02, + "learning_rate": 1.629840984867915e-05, + "loss": 0.2522, + "step": 52611 + }, + { + "epoch": 2.02, + "learning_rate": 1.629776865863042e-05, + "loss": 0.2527, + "step": 52612 + }, + { + "epoch": 2.02, + "learning_rate": 1.6297127468581687e-05, + "loss": 0.2272, + "step": 52613 + }, + { + "epoch": 2.02, + "learning_rate": 1.629648627853296e-05, + "loss": 0.1689, + "step": 52614 + }, + { + "epoch": 2.02, + "learning_rate": 1.6295845088484227e-05, + "loss": 0.3, + "step": 52615 + }, + { + "epoch": 2.02, + "learning_rate": 1.6295203898435495e-05, + "loss": 0.1716, + "step": 52616 + }, + { + "epoch": 2.02, + "learning_rate": 1.6294562708386767e-05, + "loss": 0.1999, + "step": 52617 + }, + { + "epoch": 2.02, + "learning_rate": 1.629392151833804e-05, + "loss": 0.1425, + "step": 52618 + }, + { + "epoch": 2.02, + "learning_rate": 1.6293280328289307e-05, + "loss": 0.1718, + "step": 52619 + }, + { + "epoch": 2.02, + "learning_rate": 1.6292639138240575e-05, + "loss": 0.2335, + "step": 52620 + }, + { + "epoch": 2.02, + "learning_rate": 1.6291997948191843e-05, + "loss": 0.2961, + "step": 52621 + }, + { + "epoch": 2.02, + "learning_rate": 1.6291356758143115e-05, + "loss": 0.1997, + "step": 52622 + }, + { + "epoch": 2.02, + "learning_rate": 1.6290715568094383e-05, + "loss": 0.2349, + "step": 52623 + }, + { + "epoch": 2.02, + "learning_rate": 1.6290074378045652e-05, + "loss": 0.5256, + "step": 52624 + }, + { + "epoch": 2.02, + "learning_rate": 1.6289433187996924e-05, + "loss": 0.1623, + "step": 52625 + }, + { + "epoch": 2.02, + "learning_rate": 1.6288791997948195e-05, + "loss": 0.2712, + "step": 52626 + }, + { + "epoch": 2.02, + "learning_rate": 1.6288150807899464e-05, + "loss": 0.1796, + "step": 52627 + }, + { + "epoch": 2.02, + "learning_rate": 1.6287509617850732e-05, + "loss": 0.2714, + "step": 52628 + }, + { + "epoch": 2.02, + "learning_rate": 1.6286868427802e-05, + "loss": 0.321, + "step": 52629 + }, + { + "epoch": 2.02, + "learning_rate": 1.6286227237753272e-05, + "loss": 0.2895, + "step": 52630 + }, + { + "epoch": 2.02, + "learning_rate": 1.628558604770454e-05, + "loss": 0.2296, + "step": 52631 + }, + { + "epoch": 2.02, + "learning_rate": 1.628494485765581e-05, + "loss": 0.3251, + "step": 52632 + }, + { + "epoch": 2.02, + "learning_rate": 1.628430366760708e-05, + "loss": 0.2566, + "step": 52633 + }, + { + "epoch": 2.02, + "learning_rate": 1.628366247755835e-05, + "loss": 0.2577, + "step": 52634 + }, + { + "epoch": 2.02, + "learning_rate": 1.628302128750962e-05, + "loss": 0.1495, + "step": 52635 + }, + { + "epoch": 2.02, + "learning_rate": 1.628238009746089e-05, + "loss": 0.2592, + "step": 52636 + }, + { + "epoch": 2.02, + "learning_rate": 1.6281738907412157e-05, + "loss": 0.2224, + "step": 52637 + }, + { + "epoch": 2.02, + "learning_rate": 1.628109771736343e-05, + "loss": 0.2464, + "step": 52638 + }, + { + "epoch": 2.02, + "learning_rate": 1.6280456527314697e-05, + "loss": 0.2929, + "step": 52639 + }, + { + "epoch": 2.02, + "learning_rate": 1.6279815337265965e-05, + "loss": 0.4429, + "step": 52640 + }, + { + "epoch": 2.02, + "learning_rate": 1.6279174147217237e-05, + "loss": 0.3165, + "step": 52641 + }, + { + "epoch": 2.02, + "learning_rate": 1.6278532957168505e-05, + "loss": 0.3291, + "step": 52642 + }, + { + "epoch": 2.02, + "learning_rate": 1.6277891767119774e-05, + "loss": 0.2019, + "step": 52643 + }, + { + "epoch": 2.02, + "learning_rate": 1.6277250577071045e-05, + "loss": 0.1882, + "step": 52644 + }, + { + "epoch": 2.02, + "learning_rate": 1.6276609387022314e-05, + "loss": 0.2208, + "step": 52645 + }, + { + "epoch": 2.02, + "learning_rate": 1.6275968196973586e-05, + "loss": 0.1597, + "step": 52646 + }, + { + "epoch": 2.02, + "learning_rate": 1.6275327006924854e-05, + "loss": 0.1873, + "step": 52647 + }, + { + "epoch": 2.02, + "learning_rate": 1.6274685816876122e-05, + "loss": 0.1131, + "step": 52648 + }, + { + "epoch": 2.02, + "learning_rate": 1.627404462682739e-05, + "loss": 0.1915, + "step": 52649 + }, + { + "epoch": 2.02, + "learning_rate": 1.6273403436778662e-05, + "loss": 0.3183, + "step": 52650 + }, + { + "epoch": 2.02, + "learning_rate": 1.627276224672993e-05, + "loss": 0.1612, + "step": 52651 + }, + { + "epoch": 2.02, + "learning_rate": 1.62721210566812e-05, + "loss": 0.2196, + "step": 52652 + }, + { + "epoch": 2.02, + "learning_rate": 1.627147986663247e-05, + "loss": 0.0913, + "step": 52653 + }, + { + "epoch": 2.02, + "learning_rate": 1.6270838676583742e-05, + "loss": 0.17, + "step": 52654 + }, + { + "epoch": 2.02, + "learning_rate": 1.627019748653501e-05, + "loss": 0.1424, + "step": 52655 + }, + { + "epoch": 2.02, + "learning_rate": 1.626955629648628e-05, + "loss": 0.1774, + "step": 52656 + }, + { + "epoch": 2.02, + "learning_rate": 1.6268915106437547e-05, + "loss": 0.2269, + "step": 52657 + }, + { + "epoch": 2.02, + "learning_rate": 1.626827391638882e-05, + "loss": 0.3282, + "step": 52658 + }, + { + "epoch": 2.02, + "learning_rate": 1.6267632726340087e-05, + "loss": 0.2146, + "step": 52659 + }, + { + "epoch": 2.02, + "learning_rate": 1.6266991536291356e-05, + "loss": 0.2656, + "step": 52660 + }, + { + "epoch": 2.02, + "learning_rate": 1.6266350346242627e-05, + "loss": 0.4472, + "step": 52661 + }, + { + "epoch": 2.02, + "learning_rate": 1.62657091561939e-05, + "loss": 0.2057, + "step": 52662 + }, + { + "epoch": 2.02, + "learning_rate": 1.6265067966145167e-05, + "loss": 0.2001, + "step": 52663 + }, + { + "epoch": 2.02, + "learning_rate": 1.6264426776096436e-05, + "loss": 0.1455, + "step": 52664 + }, + { + "epoch": 2.02, + "learning_rate": 1.6263785586047704e-05, + "loss": 0.1316, + "step": 52665 + }, + { + "epoch": 2.02, + "learning_rate": 1.6263144395998976e-05, + "loss": 0.2601, + "step": 52666 + }, + { + "epoch": 2.02, + "learning_rate": 1.6262503205950244e-05, + "loss": 0.2963, + "step": 52667 + }, + { + "epoch": 2.02, + "learning_rate": 1.6261862015901512e-05, + "loss": 0.1496, + "step": 52668 + }, + { + "epoch": 2.02, + "learning_rate": 1.6261220825852784e-05, + "loss": 0.0906, + "step": 52669 + }, + { + "epoch": 2.02, + "learning_rate": 1.6260579635804053e-05, + "loss": 0.2004, + "step": 52670 + }, + { + "epoch": 2.03, + "learning_rate": 1.6259938445755324e-05, + "loss": 0.3053, + "step": 52671 + }, + { + "epoch": 2.03, + "learning_rate": 1.6259297255706593e-05, + "loss": 0.2356, + "step": 52672 + }, + { + "epoch": 2.03, + "learning_rate": 1.625865606565786e-05, + "loss": 0.2077, + "step": 52673 + }, + { + "epoch": 2.03, + "learning_rate": 1.6258014875609133e-05, + "loss": 0.371, + "step": 52674 + }, + { + "epoch": 2.03, + "learning_rate": 1.62573736855604e-05, + "loss": 0.2434, + "step": 52675 + }, + { + "epoch": 2.03, + "learning_rate": 1.625673249551167e-05, + "loss": 0.3485, + "step": 52676 + }, + { + "epoch": 2.03, + "learning_rate": 1.625609130546294e-05, + "loss": 0.2174, + "step": 52677 + }, + { + "epoch": 2.03, + "learning_rate": 1.625545011541421e-05, + "loss": 0.204, + "step": 52678 + }, + { + "epoch": 2.03, + "learning_rate": 1.6254808925365478e-05, + "loss": 0.2508, + "step": 52679 + }, + { + "epoch": 2.03, + "learning_rate": 1.625416773531675e-05, + "loss": 0.22, + "step": 52680 + }, + { + "epoch": 2.03, + "learning_rate": 1.6253526545268018e-05, + "loss": 0.2526, + "step": 52681 + }, + { + "epoch": 2.03, + "learning_rate": 1.625288535521929e-05, + "loss": 0.2006, + "step": 52682 + }, + { + "epoch": 2.03, + "learning_rate": 1.6252244165170558e-05, + "loss": 0.2421, + "step": 52683 + }, + { + "epoch": 2.03, + "learning_rate": 1.6251602975121826e-05, + "loss": 0.2071, + "step": 52684 + }, + { + "epoch": 2.03, + "learning_rate": 1.6250961785073094e-05, + "loss": 0.2677, + "step": 52685 + }, + { + "epoch": 2.03, + "learning_rate": 1.6250320595024366e-05, + "loss": 0.3268, + "step": 52686 + }, + { + "epoch": 2.03, + "learning_rate": 1.6249679404975634e-05, + "loss": 0.211, + "step": 52687 + }, + { + "epoch": 2.03, + "learning_rate": 1.6249038214926903e-05, + "loss": 0.3154, + "step": 52688 + }, + { + "epoch": 2.03, + "learning_rate": 1.6248397024878174e-05, + "loss": 0.2886, + "step": 52689 + }, + { + "epoch": 2.03, + "learning_rate": 1.6247755834829446e-05, + "loss": 0.2985, + "step": 52690 + }, + { + "epoch": 2.03, + "learning_rate": 1.6247114644780715e-05, + "loss": 0.3117, + "step": 52691 + }, + { + "epoch": 2.03, + "learning_rate": 1.6246473454731983e-05, + "loss": 0.3392, + "step": 52692 + }, + { + "epoch": 2.03, + "learning_rate": 1.624583226468325e-05, + "loss": 0.1902, + "step": 52693 + }, + { + "epoch": 2.03, + "learning_rate": 1.6245191074634523e-05, + "loss": 0.2156, + "step": 52694 + }, + { + "epoch": 2.03, + "learning_rate": 1.624454988458579e-05, + "loss": 0.2265, + "step": 52695 + }, + { + "epoch": 2.03, + "learning_rate": 1.624390869453706e-05, + "loss": 0.2017, + "step": 52696 + }, + { + "epoch": 2.03, + "learning_rate": 1.624326750448833e-05, + "loss": 0.2554, + "step": 52697 + }, + { + "epoch": 2.03, + "learning_rate": 1.6242626314439603e-05, + "loss": 0.3053, + "step": 52698 + }, + { + "epoch": 2.03, + "learning_rate": 1.624198512439087e-05, + "loss": 0.2332, + "step": 52699 + }, + { + "epoch": 2.03, + "learning_rate": 1.624134393434214e-05, + "loss": 0.2044, + "step": 52700 + }, + { + "epoch": 2.03, + "learning_rate": 1.6240702744293408e-05, + "loss": 0.2248, + "step": 52701 + }, + { + "epoch": 2.03, + "learning_rate": 1.624006155424468e-05, + "loss": 0.1992, + "step": 52702 + }, + { + "epoch": 2.03, + "learning_rate": 1.6239420364195948e-05, + "loss": 0.2072, + "step": 52703 + }, + { + "epoch": 2.03, + "learning_rate": 1.6238779174147216e-05, + "loss": 0.2815, + "step": 52704 + }, + { + "epoch": 2.03, + "learning_rate": 1.6238137984098488e-05, + "loss": 0.32, + "step": 52705 + }, + { + "epoch": 2.03, + "learning_rate": 1.6237496794049756e-05, + "loss": 0.1767, + "step": 52706 + }, + { + "epoch": 2.03, + "learning_rate": 1.6236855604001028e-05, + "loss": 0.3238, + "step": 52707 + }, + { + "epoch": 2.03, + "learning_rate": 1.6236214413952296e-05, + "loss": 0.3495, + "step": 52708 + }, + { + "epoch": 2.03, + "learning_rate": 1.6235573223903565e-05, + "loss": 0.118, + "step": 52709 + }, + { + "epoch": 2.03, + "learning_rate": 1.6234932033854836e-05, + "loss": 0.268, + "step": 52710 + }, + { + "epoch": 2.03, + "learning_rate": 1.6234290843806105e-05, + "loss": 0.2298, + "step": 52711 + }, + { + "epoch": 2.03, + "learning_rate": 1.6233649653757373e-05, + "loss": 0.2259, + "step": 52712 + }, + { + "epoch": 2.03, + "learning_rate": 1.6233008463708645e-05, + "loss": 0.197, + "step": 52713 + }, + { + "epoch": 2.03, + "learning_rate": 1.6232367273659913e-05, + "loss": 0.3182, + "step": 52714 + }, + { + "epoch": 2.03, + "learning_rate": 1.6231726083611185e-05, + "loss": 0.3044, + "step": 52715 + }, + { + "epoch": 2.03, + "learning_rate": 1.6231084893562453e-05, + "loss": 0.2559, + "step": 52716 + }, + { + "epoch": 2.03, + "learning_rate": 1.623044370351372e-05, + "loss": 0.2124, + "step": 52717 + }, + { + "epoch": 2.03, + "learning_rate": 1.6229802513464993e-05, + "loss": 0.1518, + "step": 52718 + }, + { + "epoch": 2.03, + "learning_rate": 1.622916132341626e-05, + "loss": 0.2046, + "step": 52719 + }, + { + "epoch": 2.03, + "learning_rate": 1.622852013336753e-05, + "loss": 0.1714, + "step": 52720 + }, + { + "epoch": 2.03, + "learning_rate": 1.62278789433188e-05, + "loss": 0.2979, + "step": 52721 + }, + { + "epoch": 2.03, + "learning_rate": 1.622723775327007e-05, + "loss": 0.3165, + "step": 52722 + }, + { + "epoch": 2.03, + "learning_rate": 1.622659656322134e-05, + "loss": 0.2942, + "step": 52723 + }, + { + "epoch": 2.03, + "learning_rate": 1.622595537317261e-05, + "loss": 0.167, + "step": 52724 + }, + { + "epoch": 2.03, + "learning_rate": 1.622531418312388e-05, + "loss": 0.3736, + "step": 52725 + }, + { + "epoch": 2.03, + "learning_rate": 1.622467299307515e-05, + "loss": 0.2297, + "step": 52726 + }, + { + "epoch": 2.03, + "learning_rate": 1.622403180302642e-05, + "loss": 0.2371, + "step": 52727 + }, + { + "epoch": 2.03, + "learning_rate": 1.6223390612977687e-05, + "loss": 0.3375, + "step": 52728 + }, + { + "epoch": 2.03, + "learning_rate": 1.6222749422928955e-05, + "loss": 0.2149, + "step": 52729 + }, + { + "epoch": 2.03, + "learning_rate": 1.6222108232880227e-05, + "loss": 0.2931, + "step": 52730 + }, + { + "epoch": 2.03, + "learning_rate": 1.6221467042831495e-05, + "loss": 0.2614, + "step": 52731 + }, + { + "epoch": 2.03, + "learning_rate": 1.6220825852782763e-05, + "loss": 0.1937, + "step": 52732 + }, + { + "epoch": 2.03, + "learning_rate": 1.6220184662734035e-05, + "loss": 0.249, + "step": 52733 + }, + { + "epoch": 2.03, + "learning_rate": 1.6219543472685307e-05, + "loss": 0.2489, + "step": 52734 + }, + { + "epoch": 2.03, + "learning_rate": 1.6218902282636575e-05, + "loss": 0.2054, + "step": 52735 + }, + { + "epoch": 2.03, + "learning_rate": 1.6218261092587844e-05, + "loss": 0.1962, + "step": 52736 + }, + { + "epoch": 2.03, + "learning_rate": 1.6217619902539112e-05, + "loss": 0.3243, + "step": 52737 + }, + { + "epoch": 2.03, + "learning_rate": 1.6216978712490384e-05, + "loss": 0.3374, + "step": 52738 + }, + { + "epoch": 2.03, + "learning_rate": 1.6216337522441652e-05, + "loss": 0.2917, + "step": 52739 + }, + { + "epoch": 2.03, + "learning_rate": 1.621569633239292e-05, + "loss": 0.3644, + "step": 52740 + }, + { + "epoch": 2.03, + "learning_rate": 1.6215055142344192e-05, + "loss": 0.2819, + "step": 52741 + }, + { + "epoch": 2.03, + "learning_rate": 1.6214413952295464e-05, + "loss": 0.1581, + "step": 52742 + }, + { + "epoch": 2.03, + "learning_rate": 1.6213772762246732e-05, + "loss": 0.1706, + "step": 52743 + }, + { + "epoch": 2.03, + "learning_rate": 1.6213131572198e-05, + "loss": 0.1787, + "step": 52744 + }, + { + "epoch": 2.03, + "learning_rate": 1.621249038214927e-05, + "loss": 0.2374, + "step": 52745 + }, + { + "epoch": 2.03, + "learning_rate": 1.621184919210054e-05, + "loss": 0.2388, + "step": 52746 + }, + { + "epoch": 2.03, + "learning_rate": 1.621120800205181e-05, + "loss": 0.2828, + "step": 52747 + }, + { + "epoch": 2.03, + "learning_rate": 1.6210566812003077e-05, + "loss": 0.2078, + "step": 52748 + }, + { + "epoch": 2.03, + "learning_rate": 1.620992562195435e-05, + "loss": 0.2887, + "step": 52749 + }, + { + "epoch": 2.03, + "learning_rate": 1.6209284431905617e-05, + "loss": 0.1961, + "step": 52750 + }, + { + "epoch": 2.03, + "learning_rate": 1.620864324185689e-05, + "loss": 0.1494, + "step": 52751 + }, + { + "epoch": 2.03, + "learning_rate": 1.6208002051808157e-05, + "loss": 0.3321, + "step": 52752 + }, + { + "epoch": 2.03, + "learning_rate": 1.6207360861759425e-05, + "loss": 0.3045, + "step": 52753 + }, + { + "epoch": 2.03, + "learning_rate": 1.6206719671710697e-05, + "loss": 0.2021, + "step": 52754 + }, + { + "epoch": 2.03, + "learning_rate": 1.6206078481661965e-05, + "loss": 0.2391, + "step": 52755 + }, + { + "epoch": 2.03, + "learning_rate": 1.6205437291613234e-05, + "loss": 0.2173, + "step": 52756 + }, + { + "epoch": 2.03, + "learning_rate": 1.6204796101564506e-05, + "loss": 0.1641, + "step": 52757 + }, + { + "epoch": 2.03, + "learning_rate": 1.6204154911515774e-05, + "loss": 0.303, + "step": 52758 + }, + { + "epoch": 2.03, + "learning_rate": 1.6203513721467042e-05, + "loss": 0.3999, + "step": 52759 + }, + { + "epoch": 2.03, + "learning_rate": 1.6202872531418314e-05, + "loss": 0.2375, + "step": 52760 + }, + { + "epoch": 2.03, + "learning_rate": 1.6202231341369582e-05, + "loss": 0.2423, + "step": 52761 + }, + { + "epoch": 2.03, + "learning_rate": 1.6201590151320854e-05, + "loss": 0.3778, + "step": 52762 + }, + { + "epoch": 2.03, + "learning_rate": 1.6200948961272122e-05, + "loss": 0.1595, + "step": 52763 + }, + { + "epoch": 2.03, + "learning_rate": 1.620030777122339e-05, + "loss": 0.2748, + "step": 52764 + }, + { + "epoch": 2.03, + "learning_rate": 1.6199666581174662e-05, + "loss": 0.1332, + "step": 52765 + }, + { + "epoch": 2.03, + "learning_rate": 1.619902539112593e-05, + "loss": 0.2932, + "step": 52766 + }, + { + "epoch": 2.03, + "learning_rate": 1.61983842010772e-05, + "loss": 0.3509, + "step": 52767 + }, + { + "epoch": 2.03, + "learning_rate": 1.6197743011028467e-05, + "loss": 0.2604, + "step": 52768 + }, + { + "epoch": 2.03, + "learning_rate": 1.619710182097974e-05, + "loss": 0.2797, + "step": 52769 + }, + { + "epoch": 2.03, + "learning_rate": 1.619646063093101e-05, + "loss": 0.2695, + "step": 52770 + }, + { + "epoch": 2.03, + "learning_rate": 1.619581944088228e-05, + "loss": 0.1835, + "step": 52771 + }, + { + "epoch": 2.03, + "learning_rate": 1.6195178250833547e-05, + "loss": 0.2856, + "step": 52772 + }, + { + "epoch": 2.03, + "learning_rate": 1.6194537060784816e-05, + "loss": 0.1577, + "step": 52773 + }, + { + "epoch": 2.03, + "learning_rate": 1.6193895870736087e-05, + "loss": 0.311, + "step": 52774 + }, + { + "epoch": 2.03, + "learning_rate": 1.6193254680687356e-05, + "loss": 0.1232, + "step": 52775 + }, + { + "epoch": 2.03, + "learning_rate": 1.6192613490638624e-05, + "loss": 0.2109, + "step": 52776 + }, + { + "epoch": 2.03, + "learning_rate": 1.6191972300589896e-05, + "loss": 0.1923, + "step": 52777 + }, + { + "epoch": 2.03, + "learning_rate": 1.6191331110541168e-05, + "loss": 0.2039, + "step": 52778 + }, + { + "epoch": 2.03, + "learning_rate": 1.6190689920492436e-05, + "loss": 0.185, + "step": 52779 + }, + { + "epoch": 2.03, + "learning_rate": 1.6190048730443704e-05, + "loss": 0.3087, + "step": 52780 + }, + { + "epoch": 2.03, + "learning_rate": 1.6189407540394973e-05, + "loss": 0.1709, + "step": 52781 + }, + { + "epoch": 2.03, + "learning_rate": 1.6188766350346244e-05, + "loss": 0.224, + "step": 52782 + }, + { + "epoch": 2.03, + "learning_rate": 1.6188125160297513e-05, + "loss": 0.3151, + "step": 52783 + }, + { + "epoch": 2.03, + "learning_rate": 1.618748397024878e-05, + "loss": 0.2102, + "step": 52784 + }, + { + "epoch": 2.03, + "learning_rate": 1.6186842780200053e-05, + "loss": 0.216, + "step": 52785 + }, + { + "epoch": 2.03, + "learning_rate": 1.618620159015132e-05, + "loss": 0.1651, + "step": 52786 + }, + { + "epoch": 2.03, + "learning_rate": 1.6185560400102593e-05, + "loss": 0.3493, + "step": 52787 + }, + { + "epoch": 2.03, + "learning_rate": 1.618491921005386e-05, + "loss": 0.2833, + "step": 52788 + }, + { + "epoch": 2.03, + "learning_rate": 1.618427802000513e-05, + "loss": 0.2075, + "step": 52789 + }, + { + "epoch": 2.03, + "learning_rate": 1.61836368299564e-05, + "loss": 0.1908, + "step": 52790 + }, + { + "epoch": 2.03, + "learning_rate": 1.618299563990767e-05, + "loss": 0.1814, + "step": 52791 + }, + { + "epoch": 2.03, + "learning_rate": 1.6182354449858938e-05, + "loss": 0.2468, + "step": 52792 + }, + { + "epoch": 2.03, + "learning_rate": 1.618171325981021e-05, + "loss": 0.3647, + "step": 52793 + }, + { + "epoch": 2.03, + "learning_rate": 1.6181072069761478e-05, + "loss": 0.2498, + "step": 52794 + }, + { + "epoch": 2.03, + "learning_rate": 1.6180430879712746e-05, + "loss": 0.3741, + "step": 52795 + }, + { + "epoch": 2.03, + "learning_rate": 1.6179789689664018e-05, + "loss": 0.1963, + "step": 52796 + }, + { + "epoch": 2.03, + "learning_rate": 1.6179148499615286e-05, + "loss": 0.2048, + "step": 52797 + }, + { + "epoch": 2.03, + "learning_rate": 1.6178507309566558e-05, + "loss": 0.306, + "step": 52798 + }, + { + "epoch": 2.03, + "learning_rate": 1.6177866119517826e-05, + "loss": 0.18, + "step": 52799 + }, + { + "epoch": 2.03, + "learning_rate": 1.6177224929469094e-05, + "loss": 0.3105, + "step": 52800 + }, + { + "epoch": 2.03, + "learning_rate": 1.6176583739420366e-05, + "loss": 0.1964, + "step": 52801 + }, + { + "epoch": 2.03, + "learning_rate": 1.6175942549371635e-05, + "loss": 0.2912, + "step": 52802 + }, + { + "epoch": 2.03, + "learning_rate": 1.6175301359322903e-05, + "loss": 0.2714, + "step": 52803 + }, + { + "epoch": 2.03, + "learning_rate": 1.617466016927417e-05, + "loss": 0.2926, + "step": 52804 + }, + { + "epoch": 2.03, + "learning_rate": 1.6174018979225443e-05, + "loss": 0.2311, + "step": 52805 + }, + { + "epoch": 2.03, + "learning_rate": 1.6173377789176715e-05, + "loss": 0.3458, + "step": 52806 + }, + { + "epoch": 2.03, + "learning_rate": 1.6172736599127983e-05, + "loss": 0.2985, + "step": 52807 + }, + { + "epoch": 2.03, + "learning_rate": 1.617209540907925e-05, + "loss": 0.2495, + "step": 52808 + }, + { + "epoch": 2.03, + "learning_rate": 1.617145421903052e-05, + "loss": 0.1391, + "step": 52809 + }, + { + "epoch": 2.03, + "learning_rate": 1.617081302898179e-05, + "loss": 0.1752, + "step": 52810 + }, + { + "epoch": 2.03, + "learning_rate": 1.617017183893306e-05, + "loss": 0.2764, + "step": 52811 + }, + { + "epoch": 2.03, + "learning_rate": 1.6169530648884328e-05, + "loss": 0.2992, + "step": 52812 + }, + { + "epoch": 2.03, + "learning_rate": 1.61688894588356e-05, + "loss": 0.1581, + "step": 52813 + }, + { + "epoch": 2.03, + "learning_rate": 1.616824826878687e-05, + "loss": 0.3385, + "step": 52814 + }, + { + "epoch": 2.03, + "learning_rate": 1.616760707873814e-05, + "loss": 0.2304, + "step": 52815 + }, + { + "epoch": 2.03, + "learning_rate": 1.6166965888689408e-05, + "loss": 0.2148, + "step": 52816 + }, + { + "epoch": 2.03, + "learning_rate": 1.6166324698640676e-05, + "loss": 0.2045, + "step": 52817 + }, + { + "epoch": 2.03, + "learning_rate": 1.6165683508591948e-05, + "loss": 0.2907, + "step": 52818 + }, + { + "epoch": 2.03, + "learning_rate": 1.6165042318543216e-05, + "loss": 0.1589, + "step": 52819 + }, + { + "epoch": 2.03, + "learning_rate": 1.6164401128494485e-05, + "loss": 0.2632, + "step": 52820 + }, + { + "epoch": 2.03, + "learning_rate": 1.6163759938445756e-05, + "loss": 0.3803, + "step": 52821 + }, + { + "epoch": 2.03, + "learning_rate": 1.6163118748397028e-05, + "loss": 0.1479, + "step": 52822 + }, + { + "epoch": 2.03, + "learning_rate": 1.6162477558348297e-05, + "loss": 0.3678, + "step": 52823 + }, + { + "epoch": 2.03, + "learning_rate": 1.6161836368299565e-05, + "loss": 0.1143, + "step": 52824 + }, + { + "epoch": 2.03, + "learning_rate": 1.6161195178250833e-05, + "loss": 0.2235, + "step": 52825 + }, + { + "epoch": 2.03, + "learning_rate": 1.6160553988202105e-05, + "loss": 0.1801, + "step": 52826 + }, + { + "epoch": 2.03, + "learning_rate": 1.6159912798153373e-05, + "loss": 0.2007, + "step": 52827 + }, + { + "epoch": 2.03, + "learning_rate": 1.615927160810464e-05, + "loss": 0.2432, + "step": 52828 + }, + { + "epoch": 2.03, + "learning_rate": 1.6158630418055913e-05, + "loss": 0.1809, + "step": 52829 + }, + { + "epoch": 2.03, + "learning_rate": 1.615798922800718e-05, + "loss": 0.2397, + "step": 52830 + }, + { + "epoch": 2.03, + "learning_rate": 1.6157348037958453e-05, + "loss": 0.2738, + "step": 52831 + }, + { + "epoch": 2.03, + "learning_rate": 1.615670684790972e-05, + "loss": 0.3078, + "step": 52832 + }, + { + "epoch": 2.03, + "learning_rate": 1.615606565786099e-05, + "loss": 0.2046, + "step": 52833 + }, + { + "epoch": 2.03, + "learning_rate": 1.615542446781226e-05, + "loss": 0.2062, + "step": 52834 + }, + { + "epoch": 2.03, + "learning_rate": 1.615478327776353e-05, + "loss": 0.2964, + "step": 52835 + }, + { + "epoch": 2.03, + "learning_rate": 1.61541420877148e-05, + "loss": 0.25, + "step": 52836 + }, + { + "epoch": 2.03, + "learning_rate": 1.615350089766607e-05, + "loss": 0.2413, + "step": 52837 + }, + { + "epoch": 2.03, + "learning_rate": 1.615285970761734e-05, + "loss": 0.3157, + "step": 52838 + }, + { + "epoch": 2.03, + "learning_rate": 1.6152218517568607e-05, + "loss": 0.2051, + "step": 52839 + }, + { + "epoch": 2.03, + "learning_rate": 1.615157732751988e-05, + "loss": 0.1551, + "step": 52840 + }, + { + "epoch": 2.03, + "learning_rate": 1.6150936137471147e-05, + "loss": 0.195, + "step": 52841 + }, + { + "epoch": 2.03, + "learning_rate": 1.615029494742242e-05, + "loss": 0.2481, + "step": 52842 + }, + { + "epoch": 2.03, + "learning_rate": 1.6149653757373687e-05, + "loss": 0.2912, + "step": 52843 + }, + { + "epoch": 2.03, + "learning_rate": 1.6149012567324955e-05, + "loss": 0.1461, + "step": 52844 + }, + { + "epoch": 2.03, + "learning_rate": 1.6148371377276227e-05, + "loss": 0.2234, + "step": 52845 + }, + { + "epoch": 2.03, + "learning_rate": 1.6147730187227495e-05, + "loss": 0.255, + "step": 52846 + }, + { + "epoch": 2.03, + "learning_rate": 1.6147088997178764e-05, + "loss": 0.2126, + "step": 52847 + }, + { + "epoch": 2.03, + "learning_rate": 1.6146447807130032e-05, + "loss": 0.1493, + "step": 52848 + }, + { + "epoch": 2.03, + "learning_rate": 1.6145806617081304e-05, + "loss": 0.2959, + "step": 52849 + }, + { + "epoch": 2.03, + "learning_rate": 1.6145165427032575e-05, + "loss": 0.2938, + "step": 52850 + }, + { + "epoch": 2.03, + "learning_rate": 1.6144524236983844e-05, + "loss": 0.1731, + "step": 52851 + }, + { + "epoch": 2.03, + "learning_rate": 1.6143883046935112e-05, + "loss": 0.2182, + "step": 52852 + }, + { + "epoch": 2.03, + "learning_rate": 1.614324185688638e-05, + "loss": 0.2525, + "step": 52853 + }, + { + "epoch": 2.03, + "learning_rate": 1.6142600666837652e-05, + "loss": 0.3711, + "step": 52854 + }, + { + "epoch": 2.03, + "learning_rate": 1.614195947678892e-05, + "loss": 0.1941, + "step": 52855 + }, + { + "epoch": 2.03, + "learning_rate": 1.614131828674019e-05, + "loss": 0.2912, + "step": 52856 + }, + { + "epoch": 2.03, + "learning_rate": 1.614067709669146e-05, + "loss": 0.2097, + "step": 52857 + }, + { + "epoch": 2.03, + "learning_rate": 1.6140035906642732e-05, + "loss": 0.1738, + "step": 52858 + }, + { + "epoch": 2.03, + "learning_rate": 1.6139394716594e-05, + "loss": 0.3477, + "step": 52859 + }, + { + "epoch": 2.03, + "learning_rate": 1.613875352654527e-05, + "loss": 0.208, + "step": 52860 + }, + { + "epoch": 2.03, + "learning_rate": 1.6138112336496537e-05, + "loss": 0.1728, + "step": 52861 + }, + { + "epoch": 2.03, + "learning_rate": 1.613747114644781e-05, + "loss": 0.1647, + "step": 52862 + }, + { + "epoch": 2.03, + "learning_rate": 1.6136829956399077e-05, + "loss": 0.1387, + "step": 52863 + }, + { + "epoch": 2.03, + "learning_rate": 1.6136188766350345e-05, + "loss": 0.2534, + "step": 52864 + }, + { + "epoch": 2.03, + "learning_rate": 1.6135547576301617e-05, + "loss": 0.2255, + "step": 52865 + }, + { + "epoch": 2.03, + "learning_rate": 1.6134906386252885e-05, + "loss": 0.3189, + "step": 52866 + }, + { + "epoch": 2.03, + "learning_rate": 1.6134265196204157e-05, + "loss": 0.3228, + "step": 52867 + }, + { + "epoch": 2.03, + "learning_rate": 1.6133624006155426e-05, + "loss": 0.2148, + "step": 52868 + }, + { + "epoch": 2.03, + "learning_rate": 1.6132982816106694e-05, + "loss": 0.22, + "step": 52869 + }, + { + "epoch": 2.03, + "learning_rate": 1.6132341626057966e-05, + "loss": 0.1928, + "step": 52870 + }, + { + "epoch": 2.03, + "learning_rate": 1.6131700436009234e-05, + "loss": 0.2973, + "step": 52871 + }, + { + "epoch": 2.03, + "learning_rate": 1.6131059245960502e-05, + "loss": 0.1856, + "step": 52872 + }, + { + "epoch": 2.03, + "learning_rate": 1.6130418055911774e-05, + "loss": 0.2141, + "step": 52873 + }, + { + "epoch": 2.03, + "learning_rate": 1.6129776865863042e-05, + "loss": 0.2121, + "step": 52874 + }, + { + "epoch": 2.03, + "learning_rate": 1.612913567581431e-05, + "loss": 0.199, + "step": 52875 + }, + { + "epoch": 2.03, + "learning_rate": 1.6128494485765582e-05, + "loss": 0.2351, + "step": 52876 + }, + { + "epoch": 2.03, + "learning_rate": 1.612785329571685e-05, + "loss": 0.2204, + "step": 52877 + }, + { + "epoch": 2.03, + "learning_rate": 1.6127212105668122e-05, + "loss": 0.1774, + "step": 52878 + }, + { + "epoch": 2.03, + "learning_rate": 1.612657091561939e-05, + "loss": 0.3812, + "step": 52879 + }, + { + "epoch": 2.03, + "learning_rate": 1.612592972557066e-05, + "loss": 0.3337, + "step": 52880 + }, + { + "epoch": 2.03, + "learning_rate": 1.612528853552193e-05, + "loss": 0.1782, + "step": 52881 + }, + { + "epoch": 2.03, + "learning_rate": 1.61246473454732e-05, + "loss": 0.3104, + "step": 52882 + }, + { + "epoch": 2.03, + "learning_rate": 1.6124006155424467e-05, + "loss": 0.1915, + "step": 52883 + }, + { + "epoch": 2.03, + "learning_rate": 1.6123364965375736e-05, + "loss": 0.2194, + "step": 52884 + }, + { + "epoch": 2.03, + "learning_rate": 1.6122723775327007e-05, + "loss": 0.2156, + "step": 52885 + }, + { + "epoch": 2.03, + "learning_rate": 1.612208258527828e-05, + "loss": 0.2506, + "step": 52886 + }, + { + "epoch": 2.03, + "learning_rate": 1.6121441395229547e-05, + "loss": 0.2138, + "step": 52887 + }, + { + "epoch": 2.03, + "learning_rate": 1.6120800205180816e-05, + "loss": 0.2417, + "step": 52888 + }, + { + "epoch": 2.03, + "learning_rate": 1.6120159015132088e-05, + "loss": 0.5176, + "step": 52889 + }, + { + "epoch": 2.03, + "learning_rate": 1.6119517825083356e-05, + "loss": 0.3031, + "step": 52890 + }, + { + "epoch": 2.03, + "learning_rate": 1.6118876635034624e-05, + "loss": 0.1557, + "step": 52891 + }, + { + "epoch": 2.03, + "learning_rate": 1.6118235444985892e-05, + "loss": 0.2178, + "step": 52892 + }, + { + "epoch": 2.03, + "learning_rate": 1.6117594254937164e-05, + "loss": 0.2069, + "step": 52893 + }, + { + "epoch": 2.03, + "learning_rate": 1.6116953064888436e-05, + "loss": 0.3082, + "step": 52894 + }, + { + "epoch": 2.03, + "learning_rate": 1.6116311874839704e-05, + "loss": 0.2456, + "step": 52895 + }, + { + "epoch": 2.03, + "learning_rate": 1.6115670684790973e-05, + "loss": 0.2462, + "step": 52896 + }, + { + "epoch": 2.03, + "learning_rate": 1.611502949474224e-05, + "loss": 0.1947, + "step": 52897 + }, + { + "epoch": 2.03, + "learning_rate": 1.6114388304693513e-05, + "loss": 0.2664, + "step": 52898 + }, + { + "epoch": 2.03, + "learning_rate": 1.611374711464478e-05, + "loss": 0.3655, + "step": 52899 + }, + { + "epoch": 2.03, + "learning_rate": 1.611310592459605e-05, + "loss": 0.2919, + "step": 52900 + }, + { + "epoch": 2.03, + "learning_rate": 1.611246473454732e-05, + "loss": 0.2508, + "step": 52901 + }, + { + "epoch": 2.03, + "learning_rate": 1.611182354449859e-05, + "loss": 0.2496, + "step": 52902 + }, + { + "epoch": 2.03, + "learning_rate": 1.611118235444986e-05, + "loss": 0.1655, + "step": 52903 + }, + { + "epoch": 2.03, + "learning_rate": 1.611054116440113e-05, + "loss": 0.1727, + "step": 52904 + }, + { + "epoch": 2.03, + "learning_rate": 1.6109899974352398e-05, + "loss": 0.2268, + "step": 52905 + }, + { + "epoch": 2.03, + "learning_rate": 1.610925878430367e-05, + "loss": 0.2447, + "step": 52906 + }, + { + "epoch": 2.03, + "learning_rate": 1.6108617594254938e-05, + "loss": 0.3644, + "step": 52907 + }, + { + "epoch": 2.03, + "learning_rate": 1.6107976404206206e-05, + "loss": 0.1102, + "step": 52908 + }, + { + "epoch": 2.03, + "learning_rate": 1.6107335214157478e-05, + "loss": 0.2191, + "step": 52909 + }, + { + "epoch": 2.03, + "learning_rate": 1.6106694024108746e-05, + "loss": 0.2587, + "step": 52910 + }, + { + "epoch": 2.03, + "learning_rate": 1.6106052834060018e-05, + "loss": 0.2708, + "step": 52911 + }, + { + "epoch": 2.03, + "learning_rate": 1.6105411644011286e-05, + "loss": 0.2597, + "step": 52912 + }, + { + "epoch": 2.03, + "learning_rate": 1.6104770453962555e-05, + "loss": 0.1436, + "step": 52913 + }, + { + "epoch": 2.03, + "learning_rate": 1.6104129263913826e-05, + "loss": 0.1572, + "step": 52914 + }, + { + "epoch": 2.03, + "learning_rate": 1.6103488073865095e-05, + "loss": 0.1786, + "step": 52915 + }, + { + "epoch": 2.03, + "learning_rate": 1.6102846883816363e-05, + "loss": 0.296, + "step": 52916 + }, + { + "epoch": 2.03, + "learning_rate": 1.6102205693767635e-05, + "loss": 0.3542, + "step": 52917 + }, + { + "epoch": 2.03, + "learning_rate": 1.6101564503718903e-05, + "loss": 0.0878, + "step": 52918 + }, + { + "epoch": 2.03, + "learning_rate": 1.610092331367017e-05, + "loss": 0.1664, + "step": 52919 + }, + { + "epoch": 2.03, + "learning_rate": 1.6100282123621443e-05, + "loss": 0.1366, + "step": 52920 + }, + { + "epoch": 2.03, + "learning_rate": 1.609964093357271e-05, + "loss": 0.2265, + "step": 52921 + }, + { + "epoch": 2.03, + "learning_rate": 1.6098999743523983e-05, + "loss": 0.283, + "step": 52922 + }, + { + "epoch": 2.03, + "learning_rate": 1.609835855347525e-05, + "loss": 0.3297, + "step": 52923 + }, + { + "epoch": 2.03, + "learning_rate": 1.609771736342652e-05, + "loss": 0.1677, + "step": 52924 + }, + { + "epoch": 2.03, + "learning_rate": 1.609707617337779e-05, + "loss": 0.2128, + "step": 52925 + }, + { + "epoch": 2.03, + "learning_rate": 1.609643498332906e-05, + "loss": 0.3018, + "step": 52926 + }, + { + "epoch": 2.03, + "learning_rate": 1.6095793793280328e-05, + "loss": 0.1776, + "step": 52927 + }, + { + "epoch": 2.03, + "learning_rate": 1.6095152603231596e-05, + "loss": 0.359, + "step": 52928 + }, + { + "epoch": 2.03, + "learning_rate": 1.6094511413182868e-05, + "loss": 0.1432, + "step": 52929 + }, + { + "epoch": 2.03, + "learning_rate": 1.609387022313414e-05, + "loss": 0.3055, + "step": 52930 + }, + { + "epoch": 2.04, + "learning_rate": 1.6093229033085408e-05, + "loss": 0.2175, + "step": 52931 + }, + { + "epoch": 2.04, + "learning_rate": 1.6092587843036676e-05, + "loss": 0.3057, + "step": 52932 + }, + { + "epoch": 2.04, + "learning_rate": 1.6091946652987945e-05, + "loss": 0.2214, + "step": 52933 + }, + { + "epoch": 2.04, + "learning_rate": 1.6091305462939217e-05, + "loss": 0.2872, + "step": 52934 + }, + { + "epoch": 2.04, + "learning_rate": 1.6090664272890485e-05, + "loss": 0.1633, + "step": 52935 + }, + { + "epoch": 2.04, + "learning_rate": 1.6090023082841753e-05, + "loss": 0.3154, + "step": 52936 + }, + { + "epoch": 2.04, + "learning_rate": 1.6089381892793025e-05, + "loss": 0.2025, + "step": 52937 + }, + { + "epoch": 2.04, + "learning_rate": 1.6088740702744297e-05, + "loss": 0.2219, + "step": 52938 + }, + { + "epoch": 2.04, + "learning_rate": 1.6088099512695565e-05, + "loss": 0.219, + "step": 52939 + }, + { + "epoch": 2.04, + "learning_rate": 1.6087458322646833e-05, + "loss": 0.299, + "step": 52940 + }, + { + "epoch": 2.04, + "learning_rate": 1.60868171325981e-05, + "loss": 0.2654, + "step": 52941 + }, + { + "epoch": 2.04, + "learning_rate": 1.6086175942549373e-05, + "loss": 0.3594, + "step": 52942 + }, + { + "epoch": 2.04, + "learning_rate": 1.608553475250064e-05, + "loss": 0.3384, + "step": 52943 + }, + { + "epoch": 2.04, + "learning_rate": 1.608489356245191e-05, + "loss": 0.2444, + "step": 52944 + }, + { + "epoch": 2.04, + "learning_rate": 1.608425237240318e-05, + "loss": 0.1545, + "step": 52945 + }, + { + "epoch": 2.04, + "learning_rate": 1.608361118235445e-05, + "loss": 0.2143, + "step": 52946 + }, + { + "epoch": 2.04, + "learning_rate": 1.6082969992305722e-05, + "loss": 0.2693, + "step": 52947 + }, + { + "epoch": 2.04, + "learning_rate": 1.608232880225699e-05, + "loss": 0.3423, + "step": 52948 + }, + { + "epoch": 2.04, + "learning_rate": 1.608168761220826e-05, + "loss": 0.2049, + "step": 52949 + }, + { + "epoch": 2.04, + "learning_rate": 1.608104642215953e-05, + "loss": 0.2932, + "step": 52950 + }, + { + "epoch": 2.04, + "learning_rate": 1.60804052321108e-05, + "loss": 0.2368, + "step": 52951 + }, + { + "epoch": 2.04, + "learning_rate": 1.6079764042062067e-05, + "loss": 0.2538, + "step": 52952 + }, + { + "epoch": 2.04, + "learning_rate": 1.607912285201334e-05, + "loss": 0.2623, + "step": 52953 + }, + { + "epoch": 2.04, + "learning_rate": 1.6078481661964607e-05, + "loss": 0.1389, + "step": 52954 + }, + { + "epoch": 2.04, + "learning_rate": 1.6077840471915875e-05, + "loss": 0.1274, + "step": 52955 + }, + { + "epoch": 2.04, + "learning_rate": 1.6077199281867147e-05, + "loss": 0.169, + "step": 52956 + }, + { + "epoch": 2.04, + "learning_rate": 1.6076558091818415e-05, + "loss": 0.1756, + "step": 52957 + }, + { + "epoch": 2.04, + "learning_rate": 1.6075916901769687e-05, + "loss": 0.2708, + "step": 52958 + }, + { + "epoch": 2.04, + "learning_rate": 1.6075275711720955e-05, + "loss": 0.2523, + "step": 52959 + }, + { + "epoch": 2.04, + "learning_rate": 1.6074634521672224e-05, + "loss": 0.3205, + "step": 52960 + }, + { + "epoch": 2.04, + "learning_rate": 1.6073993331623495e-05, + "loss": 0.1913, + "step": 52961 + }, + { + "epoch": 2.04, + "learning_rate": 1.6073352141574764e-05, + "loss": 0.2266, + "step": 52962 + }, + { + "epoch": 2.04, + "learning_rate": 1.6072710951526032e-05, + "loss": 0.2535, + "step": 52963 + }, + { + "epoch": 2.04, + "learning_rate": 1.60720697614773e-05, + "loss": 0.2848, + "step": 52964 + }, + { + "epoch": 2.04, + "learning_rate": 1.6071428571428572e-05, + "loss": 0.2063, + "step": 52965 + }, + { + "epoch": 2.04, + "learning_rate": 1.6070787381379844e-05, + "loss": 0.165, + "step": 52966 + }, + { + "epoch": 2.04, + "learning_rate": 1.6070146191331112e-05, + "loss": 0.2593, + "step": 52967 + }, + { + "epoch": 2.04, + "learning_rate": 1.606950500128238e-05, + "loss": 0.243, + "step": 52968 + }, + { + "epoch": 2.04, + "learning_rate": 1.6068863811233652e-05, + "loss": 0.3037, + "step": 52969 + }, + { + "epoch": 2.04, + "learning_rate": 1.606822262118492e-05, + "loss": 0.2628, + "step": 52970 + }, + { + "epoch": 2.04, + "learning_rate": 1.606758143113619e-05, + "loss": 0.3359, + "step": 52971 + }, + { + "epoch": 2.04, + "learning_rate": 1.6066940241087457e-05, + "loss": 0.287, + "step": 52972 + }, + { + "epoch": 2.04, + "learning_rate": 1.606629905103873e-05, + "loss": 0.3246, + "step": 52973 + }, + { + "epoch": 2.04, + "learning_rate": 1.606565786099e-05, + "loss": 0.1984, + "step": 52974 + }, + { + "epoch": 2.04, + "learning_rate": 1.606501667094127e-05, + "loss": 0.1745, + "step": 52975 + }, + { + "epoch": 2.04, + "learning_rate": 1.6064375480892537e-05, + "loss": 0.2988, + "step": 52976 + }, + { + "epoch": 2.04, + "learning_rate": 1.6063734290843805e-05, + "loss": 0.1492, + "step": 52977 + }, + { + "epoch": 2.04, + "learning_rate": 1.6063093100795077e-05, + "loss": 0.1983, + "step": 52978 + }, + { + "epoch": 2.04, + "learning_rate": 1.6062451910746345e-05, + "loss": 0.1835, + "step": 52979 + }, + { + "epoch": 2.04, + "learning_rate": 1.6061810720697614e-05, + "loss": 0.135, + "step": 52980 + }, + { + "epoch": 2.04, + "learning_rate": 1.6061169530648886e-05, + "loss": 0.1996, + "step": 52981 + }, + { + "epoch": 2.04, + "learning_rate": 1.6060528340600154e-05, + "loss": 0.2833, + "step": 52982 + }, + { + "epoch": 2.04, + "learning_rate": 1.6059887150551426e-05, + "loss": 0.3173, + "step": 52983 + }, + { + "epoch": 2.04, + "learning_rate": 1.6059245960502694e-05, + "loss": 0.1223, + "step": 52984 + }, + { + "epoch": 2.04, + "learning_rate": 1.6058604770453962e-05, + "loss": 0.316, + "step": 52985 + }, + { + "epoch": 2.04, + "learning_rate": 1.6057963580405234e-05, + "loss": 0.2293, + "step": 52986 + }, + { + "epoch": 2.04, + "learning_rate": 1.6057322390356502e-05, + "loss": 0.1757, + "step": 52987 + }, + { + "epoch": 2.04, + "learning_rate": 1.605668120030777e-05, + "loss": 0.2924, + "step": 52988 + }, + { + "epoch": 2.04, + "learning_rate": 1.6056040010259042e-05, + "loss": 0.2214, + "step": 52989 + }, + { + "epoch": 2.04, + "learning_rate": 1.605539882021031e-05, + "loss": 0.1731, + "step": 52990 + }, + { + "epoch": 2.04, + "learning_rate": 1.605475763016158e-05, + "loss": 0.1798, + "step": 52991 + }, + { + "epoch": 2.04, + "learning_rate": 1.605411644011285e-05, + "loss": 0.2135, + "step": 52992 + }, + { + "epoch": 2.04, + "learning_rate": 1.605347525006412e-05, + "loss": 0.2991, + "step": 52993 + }, + { + "epoch": 2.04, + "learning_rate": 1.605283406001539e-05, + "loss": 0.3206, + "step": 52994 + }, + { + "epoch": 2.04, + "learning_rate": 1.605219286996666e-05, + "loss": 0.2387, + "step": 52995 + }, + { + "epoch": 2.04, + "learning_rate": 1.6051551679917927e-05, + "loss": 0.2455, + "step": 52996 + }, + { + "epoch": 2.04, + "learning_rate": 1.60509104898692e-05, + "loss": 0.2333, + "step": 52997 + }, + { + "epoch": 2.04, + "learning_rate": 1.6050269299820467e-05, + "loss": 0.2118, + "step": 52998 + }, + { + "epoch": 2.04, + "learning_rate": 1.6049628109771736e-05, + "loss": 0.2825, + "step": 52999 + }, + { + "epoch": 2.04, + "learning_rate": 1.6048986919723007e-05, + "loss": 0.3265, + "step": 53000 + }, + { + "epoch": 2.04, + "learning_rate": 1.6048345729674276e-05, + "loss": 0.1642, + "step": 53001 + }, + { + "epoch": 2.04, + "learning_rate": 1.6047704539625548e-05, + "loss": 0.3359, + "step": 53002 + }, + { + "epoch": 2.04, + "learning_rate": 1.6047063349576816e-05, + "loss": 0.2788, + "step": 53003 + }, + { + "epoch": 2.04, + "learning_rate": 1.6046422159528084e-05, + "loss": 0.1867, + "step": 53004 + }, + { + "epoch": 2.04, + "learning_rate": 1.6045780969479356e-05, + "loss": 0.2142, + "step": 53005 + }, + { + "epoch": 2.04, + "learning_rate": 1.6045139779430624e-05, + "loss": 0.2346, + "step": 53006 + }, + { + "epoch": 2.04, + "learning_rate": 1.6044498589381893e-05, + "loss": 0.2976, + "step": 53007 + }, + { + "epoch": 2.04, + "learning_rate": 1.604385739933316e-05, + "loss": 0.1936, + "step": 53008 + }, + { + "epoch": 2.04, + "learning_rate": 1.6043216209284433e-05, + "loss": 0.1634, + "step": 53009 + }, + { + "epoch": 2.04, + "learning_rate": 1.6042575019235704e-05, + "loss": 0.3187, + "step": 53010 + }, + { + "epoch": 2.04, + "learning_rate": 1.6041933829186973e-05, + "loss": 0.2415, + "step": 53011 + }, + { + "epoch": 2.04, + "learning_rate": 1.604129263913824e-05, + "loss": 0.299, + "step": 53012 + }, + { + "epoch": 2.04, + "learning_rate": 1.604065144908951e-05, + "loss": 0.1455, + "step": 53013 + }, + { + "epoch": 2.04, + "learning_rate": 1.604001025904078e-05, + "loss": 0.2719, + "step": 53014 + }, + { + "epoch": 2.04, + "learning_rate": 1.603936906899205e-05, + "loss": 0.305, + "step": 53015 + }, + { + "epoch": 2.04, + "learning_rate": 1.6038727878943318e-05, + "loss": 0.2778, + "step": 53016 + }, + { + "epoch": 2.04, + "learning_rate": 1.603808668889459e-05, + "loss": 0.295, + "step": 53017 + }, + { + "epoch": 2.04, + "learning_rate": 1.603744549884586e-05, + "loss": 0.3865, + "step": 53018 + }, + { + "epoch": 2.04, + "learning_rate": 1.603680430879713e-05, + "loss": 0.1691, + "step": 53019 + }, + { + "epoch": 2.04, + "learning_rate": 1.6036163118748398e-05, + "loss": 0.2668, + "step": 53020 + }, + { + "epoch": 2.04, + "learning_rate": 1.6035521928699666e-05, + "loss": 0.4279, + "step": 53021 + }, + { + "epoch": 2.04, + "learning_rate": 1.6034880738650938e-05, + "loss": 0.1626, + "step": 53022 + }, + { + "epoch": 2.04, + "learning_rate": 1.6034239548602206e-05, + "loss": 0.2387, + "step": 53023 + }, + { + "epoch": 2.04, + "learning_rate": 1.6033598358553474e-05, + "loss": 0.2591, + "step": 53024 + }, + { + "epoch": 2.04, + "learning_rate": 1.6032957168504746e-05, + "loss": 0.3295, + "step": 53025 + }, + { + "epoch": 2.04, + "learning_rate": 1.6032315978456015e-05, + "loss": 0.1944, + "step": 53026 + }, + { + "epoch": 2.04, + "learning_rate": 1.6031674788407286e-05, + "loss": 0.2608, + "step": 53027 + }, + { + "epoch": 2.04, + "learning_rate": 1.6031033598358555e-05, + "loss": 0.2719, + "step": 53028 + }, + { + "epoch": 2.04, + "learning_rate": 1.6030392408309823e-05, + "loss": 0.2078, + "step": 53029 + }, + { + "epoch": 2.04, + "learning_rate": 1.6029751218261095e-05, + "loss": 0.2874, + "step": 53030 + }, + { + "epoch": 2.04, + "learning_rate": 1.6029110028212363e-05, + "loss": 0.3479, + "step": 53031 + }, + { + "epoch": 2.04, + "learning_rate": 1.602846883816363e-05, + "loss": 0.1711, + "step": 53032 + }, + { + "epoch": 2.04, + "learning_rate": 1.6027827648114903e-05, + "loss": 0.2596, + "step": 53033 + }, + { + "epoch": 2.04, + "learning_rate": 1.602718645806617e-05, + "loss": 0.2492, + "step": 53034 + }, + { + "epoch": 2.04, + "learning_rate": 1.602654526801744e-05, + "loss": 0.3124, + "step": 53035 + }, + { + "epoch": 2.04, + "learning_rate": 1.602590407796871e-05, + "loss": 0.2175, + "step": 53036 + }, + { + "epoch": 2.04, + "learning_rate": 1.602526288791998e-05, + "loss": 0.1637, + "step": 53037 + }, + { + "epoch": 2.04, + "learning_rate": 1.602462169787125e-05, + "loss": 0.132, + "step": 53038 + }, + { + "epoch": 2.04, + "learning_rate": 1.602398050782252e-05, + "loss": 0.3214, + "step": 53039 + }, + { + "epoch": 2.04, + "learning_rate": 1.6023339317773788e-05, + "loss": 0.283, + "step": 53040 + }, + { + "epoch": 2.04, + "learning_rate": 1.602269812772506e-05, + "loss": 0.1427, + "step": 53041 + }, + { + "epoch": 2.04, + "learning_rate": 1.6022056937676328e-05, + "loss": 0.2669, + "step": 53042 + }, + { + "epoch": 2.04, + "learning_rate": 1.6021415747627596e-05, + "loss": 0.2126, + "step": 53043 + }, + { + "epoch": 2.04, + "learning_rate": 1.6020774557578865e-05, + "loss": 0.1967, + "step": 53044 + }, + { + "epoch": 2.04, + "learning_rate": 1.6020133367530136e-05, + "loss": 0.1384, + "step": 53045 + }, + { + "epoch": 2.04, + "learning_rate": 1.6019492177481408e-05, + "loss": 0.2537, + "step": 53046 + }, + { + "epoch": 2.04, + "learning_rate": 1.6018850987432677e-05, + "loss": 0.2121, + "step": 53047 + }, + { + "epoch": 2.04, + "learning_rate": 1.6018209797383945e-05, + "loss": 0.2863, + "step": 53048 + }, + { + "epoch": 2.04, + "learning_rate": 1.6017568607335217e-05, + "loss": 0.2793, + "step": 53049 + }, + { + "epoch": 2.04, + "learning_rate": 1.6016927417286485e-05, + "loss": 0.2211, + "step": 53050 + }, + { + "epoch": 2.04, + "learning_rate": 1.6016286227237753e-05, + "loss": 0.3047, + "step": 53051 + }, + { + "epoch": 2.04, + "learning_rate": 1.601564503718902e-05, + "loss": 0.2757, + "step": 53052 + }, + { + "epoch": 2.04, + "learning_rate": 1.6015003847140293e-05, + "loss": 0.1842, + "step": 53053 + }, + { + "epoch": 2.04, + "learning_rate": 1.6014362657091565e-05, + "loss": 0.312, + "step": 53054 + }, + { + "epoch": 2.04, + "learning_rate": 1.6013721467042833e-05, + "loss": 0.1228, + "step": 53055 + }, + { + "epoch": 2.04, + "learning_rate": 1.60130802769941e-05, + "loss": 0.3075, + "step": 53056 + }, + { + "epoch": 2.04, + "learning_rate": 1.601243908694537e-05, + "loss": 0.2584, + "step": 53057 + }, + { + "epoch": 2.04, + "learning_rate": 1.6011797896896642e-05, + "loss": 0.2134, + "step": 53058 + }, + { + "epoch": 2.04, + "learning_rate": 1.601115670684791e-05, + "loss": 0.273, + "step": 53059 + }, + { + "epoch": 2.04, + "learning_rate": 1.601051551679918e-05, + "loss": 0.2291, + "step": 53060 + }, + { + "epoch": 2.04, + "learning_rate": 1.600987432675045e-05, + "loss": 0.4196, + "step": 53061 + }, + { + "epoch": 2.04, + "learning_rate": 1.600923313670172e-05, + "loss": 0.1775, + "step": 53062 + }, + { + "epoch": 2.04, + "learning_rate": 1.600859194665299e-05, + "loss": 0.2084, + "step": 53063 + }, + { + "epoch": 2.04, + "learning_rate": 1.600795075660426e-05, + "loss": 0.202, + "step": 53064 + }, + { + "epoch": 2.04, + "learning_rate": 1.6007309566555527e-05, + "loss": 0.2191, + "step": 53065 + }, + { + "epoch": 2.04, + "learning_rate": 1.60066683765068e-05, + "loss": 0.2565, + "step": 53066 + }, + { + "epoch": 2.04, + "learning_rate": 1.6006027186458067e-05, + "loss": 0.2176, + "step": 53067 + }, + { + "epoch": 2.04, + "learning_rate": 1.6005385996409335e-05, + "loss": 0.2408, + "step": 53068 + }, + { + "epoch": 2.04, + "learning_rate": 1.6004744806360607e-05, + "loss": 0.2318, + "step": 53069 + }, + { + "epoch": 2.04, + "learning_rate": 1.6004103616311875e-05, + "loss": 0.2537, + "step": 53070 + }, + { + "epoch": 2.04, + "learning_rate": 1.6003462426263144e-05, + "loss": 0.1478, + "step": 53071 + }, + { + "epoch": 2.04, + "learning_rate": 1.6002821236214415e-05, + "loss": 0.2622, + "step": 53072 + }, + { + "epoch": 2.04, + "learning_rate": 1.6002180046165684e-05, + "loss": 0.2152, + "step": 53073 + }, + { + "epoch": 2.04, + "learning_rate": 1.6001538856116955e-05, + "loss": 0.2093, + "step": 53074 + }, + { + "epoch": 2.04, + "learning_rate": 1.6000897666068224e-05, + "loss": 0.2709, + "step": 53075 + }, + { + "epoch": 2.04, + "learning_rate": 1.6000256476019492e-05, + "loss": 0.2154, + "step": 53076 + }, + { + "epoch": 2.04, + "learning_rate": 1.5999615285970764e-05, + "loss": 0.3758, + "step": 53077 + }, + { + "epoch": 2.04, + "learning_rate": 1.5998974095922032e-05, + "loss": 0.1493, + "step": 53078 + }, + { + "epoch": 2.04, + "learning_rate": 1.59983329058733e-05, + "loss": 0.2325, + "step": 53079 + }, + { + "epoch": 2.04, + "learning_rate": 1.599769171582457e-05, + "loss": 0.2049, + "step": 53080 + }, + { + "epoch": 2.04, + "learning_rate": 1.599705052577584e-05, + "loss": 0.1292, + "step": 53081 + }, + { + "epoch": 2.04, + "learning_rate": 1.5996409335727112e-05, + "loss": 0.3333, + "step": 53082 + }, + { + "epoch": 2.04, + "learning_rate": 1.599576814567838e-05, + "loss": 0.3371, + "step": 53083 + }, + { + "epoch": 2.04, + "learning_rate": 1.599512695562965e-05, + "loss": 0.4379, + "step": 53084 + }, + { + "epoch": 2.04, + "learning_rate": 1.599448576558092e-05, + "loss": 0.1959, + "step": 53085 + }, + { + "epoch": 2.04, + "learning_rate": 1.599384457553219e-05, + "loss": 0.2053, + "step": 53086 + }, + { + "epoch": 2.04, + "learning_rate": 1.5993203385483457e-05, + "loss": 0.2508, + "step": 53087 + }, + { + "epoch": 2.04, + "learning_rate": 1.5992562195434725e-05, + "loss": 0.2091, + "step": 53088 + }, + { + "epoch": 2.04, + "learning_rate": 1.5991921005385997e-05, + "loss": 0.2304, + "step": 53089 + }, + { + "epoch": 2.04, + "learning_rate": 1.599127981533727e-05, + "loss": 0.1665, + "step": 53090 + }, + { + "epoch": 2.04, + "learning_rate": 1.5990638625288537e-05, + "loss": 0.2956, + "step": 53091 + }, + { + "epoch": 2.04, + "learning_rate": 1.5989997435239806e-05, + "loss": 0.2455, + "step": 53092 + }, + { + "epoch": 2.04, + "learning_rate": 1.5989356245191077e-05, + "loss": 0.2836, + "step": 53093 + }, + { + "epoch": 2.04, + "learning_rate": 1.5988715055142346e-05, + "loss": 0.2568, + "step": 53094 + }, + { + "epoch": 2.04, + "learning_rate": 1.5988073865093614e-05, + "loss": 0.2872, + "step": 53095 + }, + { + "epoch": 2.04, + "learning_rate": 1.5987432675044882e-05, + "loss": 0.2763, + "step": 53096 + }, + { + "epoch": 2.04, + "learning_rate": 1.5986791484996154e-05, + "loss": 0.2286, + "step": 53097 + }, + { + "epoch": 2.04, + "learning_rate": 1.5986150294947426e-05, + "loss": 0.2742, + "step": 53098 + }, + { + "epoch": 2.04, + "learning_rate": 1.5985509104898694e-05, + "loss": 0.2878, + "step": 53099 + }, + { + "epoch": 2.04, + "learning_rate": 1.5984867914849962e-05, + "loss": 0.3568, + "step": 53100 + }, + { + "epoch": 2.04, + "learning_rate": 1.598422672480123e-05, + "loss": 0.2899, + "step": 53101 + }, + { + "epoch": 2.04, + "learning_rate": 1.5983585534752502e-05, + "loss": 0.2668, + "step": 53102 + }, + { + "epoch": 2.04, + "learning_rate": 1.598294434470377e-05, + "loss": 0.2866, + "step": 53103 + }, + { + "epoch": 2.04, + "learning_rate": 1.598230315465504e-05, + "loss": 0.1668, + "step": 53104 + }, + { + "epoch": 2.04, + "learning_rate": 1.598166196460631e-05, + "loss": 0.284, + "step": 53105 + }, + { + "epoch": 2.04, + "learning_rate": 1.598102077455758e-05, + "loss": 0.2608, + "step": 53106 + }, + { + "epoch": 2.04, + "learning_rate": 1.598037958450885e-05, + "loss": 0.2113, + "step": 53107 + }, + { + "epoch": 2.04, + "learning_rate": 1.597973839446012e-05, + "loss": 0.2335, + "step": 53108 + }, + { + "epoch": 2.04, + "learning_rate": 1.5979097204411387e-05, + "loss": 0.2152, + "step": 53109 + }, + { + "epoch": 2.04, + "learning_rate": 1.597845601436266e-05, + "loss": 0.2325, + "step": 53110 + }, + { + "epoch": 2.04, + "learning_rate": 1.5977814824313927e-05, + "loss": 0.3096, + "step": 53111 + }, + { + "epoch": 2.04, + "learning_rate": 1.5977173634265196e-05, + "loss": 0.3191, + "step": 53112 + }, + { + "epoch": 2.04, + "learning_rate": 1.5976532444216468e-05, + "loss": 0.2538, + "step": 53113 + }, + { + "epoch": 2.04, + "learning_rate": 1.5975891254167736e-05, + "loss": 0.2442, + "step": 53114 + }, + { + "epoch": 2.04, + "learning_rate": 1.5975250064119004e-05, + "loss": 0.2697, + "step": 53115 + }, + { + "epoch": 2.04, + "learning_rate": 1.5974608874070276e-05, + "loss": 0.2469, + "step": 53116 + }, + { + "epoch": 2.04, + "learning_rate": 1.5973967684021544e-05, + "loss": 0.3427, + "step": 53117 + }, + { + "epoch": 2.04, + "learning_rate": 1.5973326493972816e-05, + "loss": 0.2478, + "step": 53118 + }, + { + "epoch": 2.04, + "learning_rate": 1.5972685303924084e-05, + "loss": 0.3288, + "step": 53119 + }, + { + "epoch": 2.04, + "learning_rate": 1.5972044113875353e-05, + "loss": 0.1926, + "step": 53120 + }, + { + "epoch": 2.04, + "learning_rate": 1.5971402923826624e-05, + "loss": 0.1692, + "step": 53121 + }, + { + "epoch": 2.04, + "learning_rate": 1.5970761733777893e-05, + "loss": 0.2951, + "step": 53122 + }, + { + "epoch": 2.04, + "learning_rate": 1.597012054372916e-05, + "loss": 0.2058, + "step": 53123 + }, + { + "epoch": 2.04, + "learning_rate": 1.596947935368043e-05, + "loss": 0.3841, + "step": 53124 + }, + { + "epoch": 2.04, + "learning_rate": 1.59688381636317e-05, + "loss": 0.1559, + "step": 53125 + }, + { + "epoch": 2.04, + "learning_rate": 1.5968196973582973e-05, + "loss": 0.173, + "step": 53126 + }, + { + "epoch": 2.04, + "learning_rate": 1.596755578353424e-05, + "loss": 0.3041, + "step": 53127 + }, + { + "epoch": 2.04, + "learning_rate": 1.596691459348551e-05, + "loss": 0.2661, + "step": 53128 + }, + { + "epoch": 2.04, + "learning_rate": 1.596627340343678e-05, + "loss": 0.1969, + "step": 53129 + }, + { + "epoch": 2.04, + "learning_rate": 1.596563221338805e-05, + "loss": 0.2888, + "step": 53130 + }, + { + "epoch": 2.04, + "learning_rate": 1.5964991023339318e-05, + "loss": 0.1423, + "step": 53131 + }, + { + "epoch": 2.04, + "learning_rate": 1.5964349833290586e-05, + "loss": 0.2943, + "step": 53132 + }, + { + "epoch": 2.04, + "learning_rate": 1.5963708643241858e-05, + "loss": 0.1319, + "step": 53133 + }, + { + "epoch": 2.04, + "learning_rate": 1.596306745319313e-05, + "loss": 0.1961, + "step": 53134 + }, + { + "epoch": 2.04, + "learning_rate": 1.5962426263144398e-05, + "loss": 0.2082, + "step": 53135 + }, + { + "epoch": 2.04, + "learning_rate": 1.5961785073095666e-05, + "loss": 0.3653, + "step": 53136 + }, + { + "epoch": 2.04, + "learning_rate": 1.5961143883046935e-05, + "loss": 0.1758, + "step": 53137 + }, + { + "epoch": 2.04, + "learning_rate": 1.5960502692998206e-05, + "loss": 0.2077, + "step": 53138 + }, + { + "epoch": 2.04, + "learning_rate": 1.5959861502949475e-05, + "loss": 0.2922, + "step": 53139 + }, + { + "epoch": 2.04, + "learning_rate": 1.5959220312900743e-05, + "loss": 0.2236, + "step": 53140 + }, + { + "epoch": 2.04, + "learning_rate": 1.5958579122852015e-05, + "loss": 0.2894, + "step": 53141 + }, + { + "epoch": 2.04, + "learning_rate": 1.5957937932803283e-05, + "loss": 0.2044, + "step": 53142 + }, + { + "epoch": 2.04, + "learning_rate": 1.5957296742754555e-05, + "loss": 0.3246, + "step": 53143 + }, + { + "epoch": 2.04, + "learning_rate": 1.5956655552705823e-05, + "loss": 0.1836, + "step": 53144 + }, + { + "epoch": 2.04, + "learning_rate": 1.595601436265709e-05, + "loss": 0.4043, + "step": 53145 + }, + { + "epoch": 2.04, + "learning_rate": 1.5955373172608363e-05, + "loss": 0.2563, + "step": 53146 + }, + { + "epoch": 2.04, + "learning_rate": 1.595473198255963e-05, + "loss": 0.2341, + "step": 53147 + }, + { + "epoch": 2.04, + "learning_rate": 1.59540907925109e-05, + "loss": 0.2782, + "step": 53148 + }, + { + "epoch": 2.04, + "learning_rate": 1.595344960246217e-05, + "loss": 0.4646, + "step": 53149 + }, + { + "epoch": 2.04, + "learning_rate": 1.595280841241344e-05, + "loss": 0.3186, + "step": 53150 + }, + { + "epoch": 2.04, + "learning_rate": 1.5952167222364708e-05, + "loss": 0.338, + "step": 53151 + }, + { + "epoch": 2.04, + "learning_rate": 1.595152603231598e-05, + "loss": 0.261, + "step": 53152 + }, + { + "epoch": 2.04, + "learning_rate": 1.5950884842267248e-05, + "loss": 0.2545, + "step": 53153 + }, + { + "epoch": 2.04, + "learning_rate": 1.595024365221852e-05, + "loss": 0.1982, + "step": 53154 + }, + { + "epoch": 2.04, + "learning_rate": 1.5949602462169788e-05, + "loss": 0.4713, + "step": 53155 + }, + { + "epoch": 2.04, + "learning_rate": 1.5948961272121056e-05, + "loss": 0.3651, + "step": 53156 + }, + { + "epoch": 2.04, + "learning_rate": 1.5948320082072328e-05, + "loss": 0.2319, + "step": 53157 + }, + { + "epoch": 2.04, + "learning_rate": 1.5947678892023597e-05, + "loss": 0.2165, + "step": 53158 + }, + { + "epoch": 2.04, + "learning_rate": 1.5947037701974865e-05, + "loss": 0.2057, + "step": 53159 + }, + { + "epoch": 2.04, + "learning_rate": 1.5946396511926133e-05, + "loss": 0.1759, + "step": 53160 + }, + { + "epoch": 2.04, + "learning_rate": 1.5945755321877405e-05, + "loss": 0.2182, + "step": 53161 + }, + { + "epoch": 2.04, + "learning_rate": 1.5945114131828677e-05, + "loss": 0.3217, + "step": 53162 + }, + { + "epoch": 2.04, + "learning_rate": 1.5944472941779945e-05, + "loss": 0.2093, + "step": 53163 + }, + { + "epoch": 2.04, + "learning_rate": 1.5943831751731213e-05, + "loss": 0.1601, + "step": 53164 + }, + { + "epoch": 2.04, + "learning_rate": 1.5943190561682485e-05, + "loss": 0.171, + "step": 53165 + }, + { + "epoch": 2.04, + "learning_rate": 1.5942549371633753e-05, + "loss": 0.315, + "step": 53166 + }, + { + "epoch": 2.04, + "learning_rate": 1.594190818158502e-05, + "loss": 0.2126, + "step": 53167 + }, + { + "epoch": 2.04, + "learning_rate": 1.594126699153629e-05, + "loss": 0.1307, + "step": 53168 + }, + { + "epoch": 2.04, + "learning_rate": 1.594062580148756e-05, + "loss": 0.2112, + "step": 53169 + }, + { + "epoch": 2.04, + "learning_rate": 1.5939984611438833e-05, + "loss": 0.2009, + "step": 53170 + }, + { + "epoch": 2.04, + "learning_rate": 1.5939343421390102e-05, + "loss": 0.1701, + "step": 53171 + }, + { + "epoch": 2.04, + "learning_rate": 1.593870223134137e-05, + "loss": 0.3091, + "step": 53172 + }, + { + "epoch": 2.04, + "learning_rate": 1.5938061041292642e-05, + "loss": 0.2823, + "step": 53173 + }, + { + "epoch": 2.04, + "learning_rate": 1.593741985124391e-05, + "loss": 0.1626, + "step": 53174 + }, + { + "epoch": 2.04, + "learning_rate": 1.593677866119518e-05, + "loss": 0.1726, + "step": 53175 + }, + { + "epoch": 2.04, + "learning_rate": 1.5936137471146447e-05, + "loss": 0.2869, + "step": 53176 + }, + { + "epoch": 2.04, + "learning_rate": 1.593549628109772e-05, + "loss": 0.224, + "step": 53177 + }, + { + "epoch": 2.04, + "learning_rate": 1.5934855091048987e-05, + "loss": 0.3121, + "step": 53178 + }, + { + "epoch": 2.04, + "learning_rate": 1.593421390100026e-05, + "loss": 0.2159, + "step": 53179 + }, + { + "epoch": 2.04, + "learning_rate": 1.5933572710951527e-05, + "loss": 0.18, + "step": 53180 + }, + { + "epoch": 2.04, + "learning_rate": 1.5932931520902795e-05, + "loss": 0.3675, + "step": 53181 + }, + { + "epoch": 2.04, + "learning_rate": 1.5932290330854067e-05, + "loss": 0.2694, + "step": 53182 + }, + { + "epoch": 2.04, + "learning_rate": 1.5931649140805335e-05, + "loss": 0.1792, + "step": 53183 + }, + { + "epoch": 2.04, + "learning_rate": 1.5931007950756604e-05, + "loss": 0.2455, + "step": 53184 + }, + { + "epoch": 2.04, + "learning_rate": 1.5930366760707875e-05, + "loss": 0.227, + "step": 53185 + }, + { + "epoch": 2.04, + "learning_rate": 1.5929725570659144e-05, + "loss": 0.225, + "step": 53186 + }, + { + "epoch": 2.04, + "learning_rate": 1.5929084380610412e-05, + "loss": 0.1549, + "step": 53187 + }, + { + "epoch": 2.04, + "learning_rate": 1.5928443190561684e-05, + "loss": 0.2987, + "step": 53188 + }, + { + "epoch": 2.04, + "learning_rate": 1.5927802000512952e-05, + "loss": 0.37, + "step": 53189 + }, + { + "epoch": 2.04, + "learning_rate": 1.5927160810464224e-05, + "loss": 0.2718, + "step": 53190 + }, + { + "epoch": 2.05, + "learning_rate": 1.5926519620415492e-05, + "loss": 0.1876, + "step": 53191 + }, + { + "epoch": 2.05, + "learning_rate": 1.592587843036676e-05, + "loss": 0.2573, + "step": 53192 + }, + { + "epoch": 2.05, + "learning_rate": 1.5925237240318032e-05, + "loss": 0.1631, + "step": 53193 + }, + { + "epoch": 2.05, + "learning_rate": 1.59245960502693e-05, + "loss": 0.2433, + "step": 53194 + }, + { + "epoch": 2.05, + "learning_rate": 1.592395486022057e-05, + "loss": 0.2266, + "step": 53195 + }, + { + "epoch": 2.05, + "learning_rate": 1.592331367017184e-05, + "loss": 0.1987, + "step": 53196 + }, + { + "epoch": 2.05, + "learning_rate": 1.592267248012311e-05, + "loss": 0.1661, + "step": 53197 + }, + { + "epoch": 2.05, + "learning_rate": 1.592203129007438e-05, + "loss": 0.2627, + "step": 53198 + }, + { + "epoch": 2.05, + "learning_rate": 1.592139010002565e-05, + "loss": 0.2269, + "step": 53199 + }, + { + "epoch": 2.05, + "learning_rate": 1.5920748909976917e-05, + "loss": 0.205, + "step": 53200 + }, + { + "epoch": 2.05, + "learning_rate": 1.592010771992819e-05, + "loss": 0.2944, + "step": 53201 + }, + { + "epoch": 2.05, + "learning_rate": 1.5919466529879457e-05, + "loss": 0.218, + "step": 53202 + }, + { + "epoch": 2.05, + "learning_rate": 1.5918825339830726e-05, + "loss": 0.3027, + "step": 53203 + }, + { + "epoch": 2.05, + "learning_rate": 1.5918184149781994e-05, + "loss": 0.2245, + "step": 53204 + }, + { + "epoch": 2.05, + "learning_rate": 1.5917542959733266e-05, + "loss": 0.2231, + "step": 53205 + }, + { + "epoch": 2.05, + "learning_rate": 1.5916901769684537e-05, + "loss": 0.2692, + "step": 53206 + }, + { + "epoch": 2.05, + "learning_rate": 1.5916260579635806e-05, + "loss": 0.2705, + "step": 53207 + }, + { + "epoch": 2.05, + "learning_rate": 1.5915619389587074e-05, + "loss": 0.1875, + "step": 53208 + }, + { + "epoch": 2.05, + "learning_rate": 1.5914978199538346e-05, + "loss": 0.1923, + "step": 53209 + }, + { + "epoch": 2.05, + "learning_rate": 1.5914337009489614e-05, + "loss": 0.1781, + "step": 53210 + }, + { + "epoch": 2.05, + "learning_rate": 1.5913695819440882e-05, + "loss": 0.0795, + "step": 53211 + }, + { + "epoch": 2.05, + "learning_rate": 1.591305462939215e-05, + "loss": 0.343, + "step": 53212 + }, + { + "epoch": 2.05, + "learning_rate": 1.5912413439343422e-05, + "loss": 0.2558, + "step": 53213 + }, + { + "epoch": 2.05, + "learning_rate": 1.5911772249294694e-05, + "loss": 0.4048, + "step": 53214 + }, + { + "epoch": 2.05, + "learning_rate": 1.5911131059245962e-05, + "loss": 0.1951, + "step": 53215 + }, + { + "epoch": 2.05, + "learning_rate": 1.591048986919723e-05, + "loss": 0.2216, + "step": 53216 + }, + { + "epoch": 2.05, + "learning_rate": 1.59098486791485e-05, + "loss": 0.3489, + "step": 53217 + }, + { + "epoch": 2.05, + "learning_rate": 1.590920748909977e-05, + "loss": 0.2289, + "step": 53218 + }, + { + "epoch": 2.05, + "learning_rate": 1.590856629905104e-05, + "loss": 0.3391, + "step": 53219 + }, + { + "epoch": 2.05, + "learning_rate": 1.5907925109002307e-05, + "loss": 0.1706, + "step": 53220 + }, + { + "epoch": 2.05, + "learning_rate": 1.590728391895358e-05, + "loss": 0.3002, + "step": 53221 + }, + { + "epoch": 2.05, + "learning_rate": 1.5906642728904847e-05, + "loss": 0.1455, + "step": 53222 + }, + { + "epoch": 2.05, + "learning_rate": 1.590600153885612e-05, + "loss": 0.1765, + "step": 53223 + }, + { + "epoch": 2.05, + "learning_rate": 1.5905360348807388e-05, + "loss": 0.1528, + "step": 53224 + }, + { + "epoch": 2.05, + "learning_rate": 1.5904719158758656e-05, + "loss": 0.3745, + "step": 53225 + }, + { + "epoch": 2.05, + "learning_rate": 1.5904077968709928e-05, + "loss": 0.2638, + "step": 53226 + }, + { + "epoch": 2.05, + "learning_rate": 1.5903436778661196e-05, + "loss": 0.3242, + "step": 53227 + }, + { + "epoch": 2.05, + "learning_rate": 1.5902795588612464e-05, + "loss": 0.213, + "step": 53228 + }, + { + "epoch": 2.05, + "learning_rate": 1.5902154398563736e-05, + "loss": 0.2168, + "step": 53229 + }, + { + "epoch": 2.05, + "learning_rate": 1.5901513208515004e-05, + "loss": 0.2162, + "step": 53230 + }, + { + "epoch": 2.05, + "learning_rate": 1.5900872018466273e-05, + "loss": 0.2198, + "step": 53231 + }, + { + "epoch": 2.05, + "learning_rate": 1.5900230828417544e-05, + "loss": 0.1793, + "step": 53232 + }, + { + "epoch": 2.05, + "learning_rate": 1.5899589638368813e-05, + "loss": 0.286, + "step": 53233 + }, + { + "epoch": 2.05, + "learning_rate": 1.5898948448320084e-05, + "loss": 0.1648, + "step": 53234 + }, + { + "epoch": 2.05, + "learning_rate": 1.5898307258271353e-05, + "loss": 0.2293, + "step": 53235 + }, + { + "epoch": 2.05, + "learning_rate": 1.589766606822262e-05, + "loss": 0.2295, + "step": 53236 + }, + { + "epoch": 2.05, + "learning_rate": 1.5897024878173893e-05, + "loss": 0.3081, + "step": 53237 + }, + { + "epoch": 2.05, + "learning_rate": 1.589638368812516e-05, + "loss": 0.2407, + "step": 53238 + }, + { + "epoch": 2.05, + "learning_rate": 1.589574249807643e-05, + "loss": 0.2622, + "step": 53239 + }, + { + "epoch": 2.05, + "learning_rate": 1.5895101308027698e-05, + "loss": 0.1933, + "step": 53240 + }, + { + "epoch": 2.05, + "learning_rate": 1.589446011797897e-05, + "loss": 0.1751, + "step": 53241 + }, + { + "epoch": 2.05, + "learning_rate": 1.589381892793024e-05, + "loss": 0.287, + "step": 53242 + }, + { + "epoch": 2.05, + "learning_rate": 1.589317773788151e-05, + "loss": 0.2335, + "step": 53243 + }, + { + "epoch": 2.05, + "learning_rate": 1.5892536547832778e-05, + "loss": 0.2403, + "step": 53244 + }, + { + "epoch": 2.05, + "learning_rate": 1.589189535778405e-05, + "loss": 0.2388, + "step": 53245 + }, + { + "epoch": 2.05, + "learning_rate": 1.5891254167735318e-05, + "loss": 0.2514, + "step": 53246 + }, + { + "epoch": 2.05, + "learning_rate": 1.5890612977686586e-05, + "loss": 0.1543, + "step": 53247 + }, + { + "epoch": 2.05, + "learning_rate": 1.5889971787637855e-05, + "loss": 0.3489, + "step": 53248 + }, + { + "epoch": 2.05, + "learning_rate": 1.5889330597589126e-05, + "loss": 0.2125, + "step": 53249 + }, + { + "epoch": 2.05, + "learning_rate": 1.5888689407540398e-05, + "loss": 0.273, + "step": 53250 + }, + { + "epoch": 2.05, + "learning_rate": 1.5888048217491666e-05, + "loss": 0.1506, + "step": 53251 + }, + { + "epoch": 2.05, + "learning_rate": 1.5887407027442935e-05, + "loss": 0.1837, + "step": 53252 + }, + { + "epoch": 2.05, + "learning_rate": 1.5886765837394206e-05, + "loss": 0.2912, + "step": 53253 + }, + { + "epoch": 2.05, + "learning_rate": 1.5886124647345475e-05, + "loss": 0.2089, + "step": 53254 + }, + { + "epoch": 2.05, + "learning_rate": 1.5885483457296743e-05, + "loss": 0.2536, + "step": 53255 + }, + { + "epoch": 2.05, + "learning_rate": 1.588484226724801e-05, + "loss": 0.2271, + "step": 53256 + }, + { + "epoch": 2.05, + "learning_rate": 1.5884201077199283e-05, + "loss": 0.2348, + "step": 53257 + }, + { + "epoch": 2.05, + "learning_rate": 1.588355988715055e-05, + "loss": 0.1155, + "step": 53258 + }, + { + "epoch": 2.05, + "learning_rate": 1.5882918697101823e-05, + "loss": 0.1898, + "step": 53259 + }, + { + "epoch": 2.05, + "learning_rate": 1.588227750705309e-05, + "loss": 0.186, + "step": 53260 + }, + { + "epoch": 2.05, + "learning_rate": 1.588163631700436e-05, + "loss": 0.3719, + "step": 53261 + }, + { + "epoch": 2.05, + "learning_rate": 1.588099512695563e-05, + "loss": 0.2691, + "step": 53262 + }, + { + "epoch": 2.05, + "learning_rate": 1.58803539369069e-05, + "loss": 0.1227, + "step": 53263 + }, + { + "epoch": 2.05, + "learning_rate": 1.5879712746858168e-05, + "loss": 0.1754, + "step": 53264 + }, + { + "epoch": 2.05, + "learning_rate": 1.587907155680944e-05, + "loss": 0.2434, + "step": 53265 + }, + { + "epoch": 2.05, + "learning_rate": 1.5878430366760708e-05, + "loss": 0.1907, + "step": 53266 + }, + { + "epoch": 2.05, + "learning_rate": 1.5877789176711976e-05, + "loss": 0.1378, + "step": 53267 + }, + { + "epoch": 2.05, + "learning_rate": 1.5877147986663248e-05, + "loss": 0.3628, + "step": 53268 + }, + { + "epoch": 2.05, + "learning_rate": 1.5876506796614517e-05, + "loss": 0.3438, + "step": 53269 + }, + { + "epoch": 2.05, + "learning_rate": 1.5875865606565788e-05, + "loss": 0.4197, + "step": 53270 + }, + { + "epoch": 2.05, + "learning_rate": 1.5875224416517057e-05, + "loss": 0.166, + "step": 53271 + }, + { + "epoch": 2.05, + "learning_rate": 1.5874583226468325e-05, + "loss": 0.2797, + "step": 53272 + }, + { + "epoch": 2.05, + "learning_rate": 1.5873942036419597e-05, + "loss": 0.4469, + "step": 53273 + }, + { + "epoch": 2.05, + "learning_rate": 1.5873300846370865e-05, + "loss": 0.2242, + "step": 53274 + }, + { + "epoch": 2.05, + "learning_rate": 1.5872659656322133e-05, + "loss": 0.2262, + "step": 53275 + }, + { + "epoch": 2.05, + "learning_rate": 1.58720184662734e-05, + "loss": 0.3344, + "step": 53276 + }, + { + "epoch": 2.05, + "learning_rate": 1.5871377276224673e-05, + "loss": 0.2216, + "step": 53277 + }, + { + "epoch": 2.05, + "learning_rate": 1.5870736086175945e-05, + "loss": 0.229, + "step": 53278 + }, + { + "epoch": 2.05, + "learning_rate": 1.5870094896127213e-05, + "loss": 0.1688, + "step": 53279 + }, + { + "epoch": 2.05, + "learning_rate": 1.586945370607848e-05, + "loss": 0.252, + "step": 53280 + }, + { + "epoch": 2.05, + "learning_rate": 1.5868812516029753e-05, + "loss": 0.1882, + "step": 53281 + }, + { + "epoch": 2.05, + "learning_rate": 1.5868171325981022e-05, + "loss": 0.4208, + "step": 53282 + }, + { + "epoch": 2.05, + "learning_rate": 1.586753013593229e-05, + "loss": 0.0847, + "step": 53283 + }, + { + "epoch": 2.05, + "learning_rate": 1.586688894588356e-05, + "loss": 0.317, + "step": 53284 + }, + { + "epoch": 2.05, + "learning_rate": 1.586624775583483e-05, + "loss": 0.1688, + "step": 53285 + }, + { + "epoch": 2.05, + "learning_rate": 1.5865606565786102e-05, + "loss": 0.192, + "step": 53286 + }, + { + "epoch": 2.05, + "learning_rate": 1.586496537573737e-05, + "loss": 0.1909, + "step": 53287 + }, + { + "epoch": 2.05, + "learning_rate": 1.586432418568864e-05, + "loss": 0.3063, + "step": 53288 + }, + { + "epoch": 2.05, + "learning_rate": 1.586368299563991e-05, + "loss": 0.1806, + "step": 53289 + }, + { + "epoch": 2.05, + "learning_rate": 1.586304180559118e-05, + "loss": 0.3468, + "step": 53290 + }, + { + "epoch": 2.05, + "learning_rate": 1.5862400615542447e-05, + "loss": 0.2172, + "step": 53291 + }, + { + "epoch": 2.05, + "learning_rate": 1.5861759425493715e-05, + "loss": 0.192, + "step": 53292 + }, + { + "epoch": 2.05, + "learning_rate": 1.5861118235444987e-05, + "loss": 0.1538, + "step": 53293 + }, + { + "epoch": 2.05, + "learning_rate": 1.586047704539626e-05, + "loss": 0.1027, + "step": 53294 + }, + { + "epoch": 2.05, + "learning_rate": 1.5859835855347527e-05, + "loss": 0.21, + "step": 53295 + }, + { + "epoch": 2.05, + "learning_rate": 1.5859194665298795e-05, + "loss": 0.2308, + "step": 53296 + }, + { + "epoch": 2.05, + "learning_rate": 1.5858553475250067e-05, + "loss": 0.1989, + "step": 53297 + }, + { + "epoch": 2.05, + "learning_rate": 1.5857912285201335e-05, + "loss": 0.2628, + "step": 53298 + }, + { + "epoch": 2.05, + "learning_rate": 1.5857271095152604e-05, + "loss": 0.2404, + "step": 53299 + }, + { + "epoch": 2.05, + "learning_rate": 1.5856629905103872e-05, + "loss": 0.2257, + "step": 53300 + }, + { + "epoch": 2.05, + "learning_rate": 1.5855988715055144e-05, + "loss": 0.2834, + "step": 53301 + }, + { + "epoch": 2.05, + "learning_rate": 1.5855347525006412e-05, + "loss": 0.1576, + "step": 53302 + }, + { + "epoch": 2.05, + "learning_rate": 1.5854706334957684e-05, + "loss": 0.2454, + "step": 53303 + }, + { + "epoch": 2.05, + "learning_rate": 1.5854065144908952e-05, + "loss": 0.1625, + "step": 53304 + }, + { + "epoch": 2.05, + "learning_rate": 1.585342395486022e-05, + "loss": 0.2654, + "step": 53305 + }, + { + "epoch": 2.05, + "learning_rate": 1.5852782764811492e-05, + "loss": 0.1855, + "step": 53306 + }, + { + "epoch": 2.05, + "learning_rate": 1.585214157476276e-05, + "loss": 0.359, + "step": 53307 + }, + { + "epoch": 2.05, + "learning_rate": 1.585150038471403e-05, + "loss": 0.3661, + "step": 53308 + }, + { + "epoch": 2.05, + "learning_rate": 1.58508591946653e-05, + "loss": 0.3731, + "step": 53309 + }, + { + "epoch": 2.05, + "learning_rate": 1.585021800461657e-05, + "loss": 0.2147, + "step": 53310 + }, + { + "epoch": 2.05, + "learning_rate": 1.5849576814567837e-05, + "loss": 0.2625, + "step": 53311 + }, + { + "epoch": 2.05, + "learning_rate": 1.584893562451911e-05, + "loss": 0.1829, + "step": 53312 + }, + { + "epoch": 2.05, + "learning_rate": 1.5848294434470377e-05, + "loss": 0.1187, + "step": 53313 + }, + { + "epoch": 2.05, + "learning_rate": 1.584765324442165e-05, + "loss": 0.1504, + "step": 53314 + }, + { + "epoch": 2.05, + "learning_rate": 1.5847012054372917e-05, + "loss": 0.1676, + "step": 53315 + }, + { + "epoch": 2.05, + "learning_rate": 1.5846370864324186e-05, + "loss": 0.1913, + "step": 53316 + }, + { + "epoch": 2.05, + "learning_rate": 1.5845729674275457e-05, + "loss": 0.2089, + "step": 53317 + }, + { + "epoch": 2.05, + "learning_rate": 1.5845088484226726e-05, + "loss": 0.2756, + "step": 53318 + }, + { + "epoch": 2.05, + "learning_rate": 1.5844447294177994e-05, + "loss": 0.288, + "step": 53319 + }, + { + "epoch": 2.05, + "learning_rate": 1.5843806104129262e-05, + "loss": 0.3818, + "step": 53320 + }, + { + "epoch": 2.05, + "learning_rate": 1.5843164914080534e-05, + "loss": 0.1546, + "step": 53321 + }, + { + "epoch": 2.05, + "learning_rate": 1.5842523724031806e-05, + "loss": 0.1901, + "step": 53322 + }, + { + "epoch": 2.05, + "learning_rate": 1.5841882533983074e-05, + "loss": 0.2846, + "step": 53323 + }, + { + "epoch": 2.05, + "learning_rate": 1.5841241343934342e-05, + "loss": 0.2059, + "step": 53324 + }, + { + "epoch": 2.05, + "learning_rate": 1.5840600153885614e-05, + "loss": 0.153, + "step": 53325 + }, + { + "epoch": 2.05, + "learning_rate": 1.5839958963836882e-05, + "loss": 0.2623, + "step": 53326 + }, + { + "epoch": 2.05, + "learning_rate": 1.583931777378815e-05, + "loss": 0.2058, + "step": 53327 + }, + { + "epoch": 2.05, + "learning_rate": 1.583867658373942e-05, + "loss": 0.1531, + "step": 53328 + }, + { + "epoch": 2.05, + "learning_rate": 1.583803539369069e-05, + "loss": 0.1446, + "step": 53329 + }, + { + "epoch": 2.05, + "learning_rate": 1.5837394203641962e-05, + "loss": 0.2871, + "step": 53330 + }, + { + "epoch": 2.05, + "learning_rate": 1.583675301359323e-05, + "loss": 0.2592, + "step": 53331 + }, + { + "epoch": 2.05, + "learning_rate": 1.58361118235445e-05, + "loss": 0.1575, + "step": 53332 + }, + { + "epoch": 2.05, + "learning_rate": 1.583547063349577e-05, + "loss": 0.1825, + "step": 53333 + }, + { + "epoch": 2.05, + "learning_rate": 1.583482944344704e-05, + "loss": 0.2741, + "step": 53334 + }, + { + "epoch": 2.05, + "learning_rate": 1.5834188253398308e-05, + "loss": 0.3146, + "step": 53335 + }, + { + "epoch": 2.05, + "learning_rate": 1.5833547063349576e-05, + "loss": 0.3126, + "step": 53336 + }, + { + "epoch": 2.05, + "learning_rate": 1.5832905873300848e-05, + "loss": 0.2078, + "step": 53337 + }, + { + "epoch": 2.05, + "learning_rate": 1.5832264683252116e-05, + "loss": 0.3325, + "step": 53338 + }, + { + "epoch": 2.05, + "learning_rate": 1.5831623493203388e-05, + "loss": 0.2467, + "step": 53339 + }, + { + "epoch": 2.05, + "learning_rate": 1.5830982303154656e-05, + "loss": 0.349, + "step": 53340 + }, + { + "epoch": 2.05, + "learning_rate": 1.5830341113105924e-05, + "loss": 0.1555, + "step": 53341 + }, + { + "epoch": 2.05, + "learning_rate": 1.5829699923057196e-05, + "loss": 0.2935, + "step": 53342 + }, + { + "epoch": 2.05, + "learning_rate": 1.5829058733008464e-05, + "loss": 0.1842, + "step": 53343 + }, + { + "epoch": 2.05, + "learning_rate": 1.5828417542959733e-05, + "loss": 0.2338, + "step": 53344 + }, + { + "epoch": 2.05, + "learning_rate": 1.5827776352911004e-05, + "loss": 0.1905, + "step": 53345 + }, + { + "epoch": 2.05, + "learning_rate": 1.5827135162862273e-05, + "loss": 0.2085, + "step": 53346 + }, + { + "epoch": 2.05, + "learning_rate": 1.582649397281354e-05, + "loss": 0.1472, + "step": 53347 + }, + { + "epoch": 2.05, + "learning_rate": 1.5825852782764813e-05, + "loss": 0.4561, + "step": 53348 + }, + { + "epoch": 2.05, + "learning_rate": 1.582521159271608e-05, + "loss": 0.1796, + "step": 53349 + }, + { + "epoch": 2.05, + "learning_rate": 1.5824570402667353e-05, + "loss": 0.285, + "step": 53350 + }, + { + "epoch": 2.05, + "learning_rate": 1.582392921261862e-05, + "loss": 0.2778, + "step": 53351 + }, + { + "epoch": 2.05, + "learning_rate": 1.582328802256989e-05, + "loss": 0.3212, + "step": 53352 + }, + { + "epoch": 2.05, + "learning_rate": 1.582264683252116e-05, + "loss": 0.181, + "step": 53353 + }, + { + "epoch": 2.05, + "learning_rate": 1.582200564247243e-05, + "loss": 0.2818, + "step": 53354 + }, + { + "epoch": 2.05, + "learning_rate": 1.5821364452423698e-05, + "loss": 0.4301, + "step": 53355 + }, + { + "epoch": 2.05, + "learning_rate": 1.5820723262374966e-05, + "loss": 0.1885, + "step": 53356 + }, + { + "epoch": 2.05, + "learning_rate": 1.5820082072326238e-05, + "loss": 0.1554, + "step": 53357 + }, + { + "epoch": 2.05, + "learning_rate": 1.581944088227751e-05, + "loss": 0.2808, + "step": 53358 + }, + { + "epoch": 2.05, + "learning_rate": 1.5818799692228778e-05, + "loss": 0.2398, + "step": 53359 + }, + { + "epoch": 2.05, + "learning_rate": 1.5818158502180046e-05, + "loss": 0.344, + "step": 53360 + }, + { + "epoch": 2.05, + "learning_rate": 1.5817517312131318e-05, + "loss": 0.2184, + "step": 53361 + }, + { + "epoch": 2.05, + "learning_rate": 1.5816876122082586e-05, + "loss": 0.2409, + "step": 53362 + }, + { + "epoch": 2.05, + "learning_rate": 1.5816234932033855e-05, + "loss": 0.1582, + "step": 53363 + }, + { + "epoch": 2.05, + "learning_rate": 1.5815593741985123e-05, + "loss": 0.166, + "step": 53364 + }, + { + "epoch": 2.05, + "learning_rate": 1.5814952551936395e-05, + "loss": 0.0978, + "step": 53365 + }, + { + "epoch": 2.05, + "learning_rate": 1.5814311361887666e-05, + "loss": 0.3198, + "step": 53366 + }, + { + "epoch": 2.05, + "learning_rate": 1.5813670171838935e-05, + "loss": 0.2818, + "step": 53367 + }, + { + "epoch": 2.05, + "learning_rate": 1.5813028981790203e-05, + "loss": 0.174, + "step": 53368 + }, + { + "epoch": 2.05, + "learning_rate": 1.5812387791741475e-05, + "loss": 0.1516, + "step": 53369 + }, + { + "epoch": 2.05, + "learning_rate": 1.5811746601692743e-05, + "loss": 0.2092, + "step": 53370 + }, + { + "epoch": 2.05, + "learning_rate": 1.581110541164401e-05, + "loss": 0.262, + "step": 53371 + }, + { + "epoch": 2.05, + "learning_rate": 1.581046422159528e-05, + "loss": 0.2693, + "step": 53372 + }, + { + "epoch": 2.05, + "learning_rate": 1.580982303154655e-05, + "loss": 0.2764, + "step": 53373 + }, + { + "epoch": 2.05, + "learning_rate": 1.580918184149782e-05, + "loss": 0.4107, + "step": 53374 + }, + { + "epoch": 2.05, + "learning_rate": 1.580854065144909e-05, + "loss": 0.3229, + "step": 53375 + }, + { + "epoch": 2.05, + "learning_rate": 1.580789946140036e-05, + "loss": 0.26, + "step": 53376 + }, + { + "epoch": 2.05, + "learning_rate": 1.580725827135163e-05, + "loss": 0.2483, + "step": 53377 + }, + { + "epoch": 2.05, + "learning_rate": 1.58066170813029e-05, + "loss": 0.154, + "step": 53378 + }, + { + "epoch": 2.05, + "learning_rate": 1.5805975891254168e-05, + "loss": 0.177, + "step": 53379 + }, + { + "epoch": 2.05, + "learning_rate": 1.5805334701205437e-05, + "loss": 0.3003, + "step": 53380 + }, + { + "epoch": 2.05, + "learning_rate": 1.5804693511156708e-05, + "loss": 0.276, + "step": 53381 + }, + { + "epoch": 2.05, + "learning_rate": 1.5804052321107977e-05, + "loss": 0.2396, + "step": 53382 + }, + { + "epoch": 2.05, + "learning_rate": 1.5803411131059248e-05, + "loss": 0.3912, + "step": 53383 + }, + { + "epoch": 2.05, + "learning_rate": 1.5802769941010517e-05, + "loss": 0.3116, + "step": 53384 + }, + { + "epoch": 2.05, + "learning_rate": 1.5802128750961785e-05, + "loss": 0.3457, + "step": 53385 + }, + { + "epoch": 2.05, + "learning_rate": 1.5801487560913057e-05, + "loss": 0.4455, + "step": 53386 + }, + { + "epoch": 2.05, + "learning_rate": 1.5800846370864325e-05, + "loss": 0.3276, + "step": 53387 + }, + { + "epoch": 2.05, + "learning_rate": 1.5800205180815593e-05, + "loss": 0.1998, + "step": 53388 + }, + { + "epoch": 2.05, + "learning_rate": 1.5799563990766865e-05, + "loss": 0.3457, + "step": 53389 + }, + { + "epoch": 2.05, + "learning_rate": 1.5798922800718133e-05, + "loss": 0.1795, + "step": 53390 + }, + { + "epoch": 2.05, + "learning_rate": 1.57982816106694e-05, + "loss": 0.1907, + "step": 53391 + }, + { + "epoch": 2.05, + "learning_rate": 1.5797640420620673e-05, + "loss": 0.1906, + "step": 53392 + }, + { + "epoch": 2.05, + "learning_rate": 1.5796999230571942e-05, + "loss": 0.2499, + "step": 53393 + }, + { + "epoch": 2.05, + "learning_rate": 1.5796358040523213e-05, + "loss": 0.1794, + "step": 53394 + }, + { + "epoch": 2.05, + "learning_rate": 1.5795716850474482e-05, + "loss": 0.2046, + "step": 53395 + }, + { + "epoch": 2.05, + "learning_rate": 1.579507566042575e-05, + "loss": 0.2427, + "step": 53396 + }, + { + "epoch": 2.05, + "learning_rate": 1.5794434470377022e-05, + "loss": 0.2025, + "step": 53397 + }, + { + "epoch": 2.05, + "learning_rate": 1.579379328032829e-05, + "loss": 0.2842, + "step": 53398 + }, + { + "epoch": 2.05, + "learning_rate": 1.579315209027956e-05, + "loss": 0.1527, + "step": 53399 + }, + { + "epoch": 2.05, + "learning_rate": 1.5792510900230827e-05, + "loss": 0.1247, + "step": 53400 + }, + { + "epoch": 2.05, + "learning_rate": 1.57918697101821e-05, + "loss": 0.3889, + "step": 53401 + }, + { + "epoch": 2.05, + "learning_rate": 1.579122852013337e-05, + "loss": 0.3117, + "step": 53402 + }, + { + "epoch": 2.05, + "learning_rate": 1.579058733008464e-05, + "loss": 0.1914, + "step": 53403 + }, + { + "epoch": 2.05, + "learning_rate": 1.5789946140035907e-05, + "loss": 0.1984, + "step": 53404 + }, + { + "epoch": 2.05, + "learning_rate": 1.578930494998718e-05, + "loss": 0.1515, + "step": 53405 + }, + { + "epoch": 2.05, + "learning_rate": 1.5788663759938447e-05, + "loss": 0.2005, + "step": 53406 + }, + { + "epoch": 2.05, + "learning_rate": 1.5788022569889715e-05, + "loss": 0.3483, + "step": 53407 + }, + { + "epoch": 2.05, + "learning_rate": 1.5787381379840984e-05, + "loss": 0.1847, + "step": 53408 + }, + { + "epoch": 2.05, + "learning_rate": 1.5786740189792255e-05, + "loss": 0.2641, + "step": 53409 + }, + { + "epoch": 2.05, + "learning_rate": 1.5786098999743527e-05, + "loss": 0.2599, + "step": 53410 + }, + { + "epoch": 2.05, + "learning_rate": 1.5785457809694795e-05, + "loss": 0.1908, + "step": 53411 + }, + { + "epoch": 2.05, + "learning_rate": 1.5784816619646064e-05, + "loss": 0.2117, + "step": 53412 + }, + { + "epoch": 2.05, + "learning_rate": 1.5784175429597335e-05, + "loss": 0.0726, + "step": 53413 + }, + { + "epoch": 2.05, + "learning_rate": 1.5783534239548604e-05, + "loss": 0.1893, + "step": 53414 + }, + { + "epoch": 2.05, + "learning_rate": 1.5782893049499872e-05, + "loss": 0.289, + "step": 53415 + }, + { + "epoch": 2.05, + "learning_rate": 1.578225185945114e-05, + "loss": 0.2058, + "step": 53416 + }, + { + "epoch": 2.05, + "learning_rate": 1.5781610669402412e-05, + "loss": 0.2608, + "step": 53417 + }, + { + "epoch": 2.05, + "learning_rate": 1.578096947935368e-05, + "loss": 0.2192, + "step": 53418 + }, + { + "epoch": 2.05, + "learning_rate": 1.5780328289304952e-05, + "loss": 0.189, + "step": 53419 + }, + { + "epoch": 2.05, + "learning_rate": 1.577968709925622e-05, + "loss": 0.2138, + "step": 53420 + }, + { + "epoch": 2.05, + "learning_rate": 1.577904590920749e-05, + "loss": 0.3162, + "step": 53421 + }, + { + "epoch": 2.05, + "learning_rate": 1.577840471915876e-05, + "loss": 0.1986, + "step": 53422 + }, + { + "epoch": 2.05, + "learning_rate": 1.577776352911003e-05, + "loss": 0.2696, + "step": 53423 + }, + { + "epoch": 2.05, + "learning_rate": 1.5777122339061297e-05, + "loss": 0.2721, + "step": 53424 + }, + { + "epoch": 2.05, + "learning_rate": 1.577648114901257e-05, + "loss": 0.2875, + "step": 53425 + }, + { + "epoch": 2.05, + "learning_rate": 1.5775839958963837e-05, + "loss": 0.2319, + "step": 53426 + }, + { + "epoch": 2.05, + "learning_rate": 1.5775198768915106e-05, + "loss": 0.1712, + "step": 53427 + }, + { + "epoch": 2.05, + "learning_rate": 1.5774557578866377e-05, + "loss": 0.1299, + "step": 53428 + }, + { + "epoch": 2.05, + "learning_rate": 1.5773916388817646e-05, + "loss": 0.3263, + "step": 53429 + }, + { + "epoch": 2.05, + "learning_rate": 1.5773275198768917e-05, + "loss": 0.1831, + "step": 53430 + }, + { + "epoch": 2.05, + "learning_rate": 1.5772634008720186e-05, + "loss": 0.1503, + "step": 53431 + }, + { + "epoch": 2.05, + "learning_rate": 1.5771992818671454e-05, + "loss": 0.1851, + "step": 53432 + }, + { + "epoch": 2.05, + "learning_rate": 1.5771351628622726e-05, + "loss": 0.1947, + "step": 53433 + }, + { + "epoch": 2.05, + "learning_rate": 1.5770710438573994e-05, + "loss": 0.1607, + "step": 53434 + }, + { + "epoch": 2.05, + "learning_rate": 1.5770069248525262e-05, + "loss": 0.2131, + "step": 53435 + }, + { + "epoch": 2.05, + "learning_rate": 1.576942805847653e-05, + "loss": 0.2158, + "step": 53436 + }, + { + "epoch": 2.05, + "learning_rate": 1.5768786868427802e-05, + "loss": 0.3344, + "step": 53437 + }, + { + "epoch": 2.05, + "learning_rate": 1.5768145678379074e-05, + "loss": 0.1603, + "step": 53438 + }, + { + "epoch": 2.05, + "learning_rate": 1.5767504488330342e-05, + "loss": 0.3735, + "step": 53439 + }, + { + "epoch": 2.05, + "learning_rate": 1.576686329828161e-05, + "loss": 0.1328, + "step": 53440 + }, + { + "epoch": 2.05, + "learning_rate": 1.5766222108232882e-05, + "loss": 0.1854, + "step": 53441 + }, + { + "epoch": 2.05, + "learning_rate": 1.576558091818415e-05, + "loss": 0.2511, + "step": 53442 + }, + { + "epoch": 2.05, + "learning_rate": 1.576493972813542e-05, + "loss": 0.3172, + "step": 53443 + }, + { + "epoch": 2.05, + "learning_rate": 1.5764298538086687e-05, + "loss": 0.2193, + "step": 53444 + }, + { + "epoch": 2.05, + "learning_rate": 1.576365734803796e-05, + "loss": 0.2246, + "step": 53445 + }, + { + "epoch": 2.05, + "learning_rate": 1.576301615798923e-05, + "loss": 0.1684, + "step": 53446 + }, + { + "epoch": 2.05, + "learning_rate": 1.57623749679405e-05, + "loss": 0.3512, + "step": 53447 + }, + { + "epoch": 2.05, + "learning_rate": 1.5761733777891768e-05, + "loss": 0.2198, + "step": 53448 + }, + { + "epoch": 2.05, + "learning_rate": 1.576109258784304e-05, + "loss": 0.2262, + "step": 53449 + }, + { + "epoch": 2.05, + "learning_rate": 1.5760451397794308e-05, + "loss": 0.2485, + "step": 53450 + }, + { + "epoch": 2.06, + "learning_rate": 1.5759810207745576e-05, + "loss": 0.2114, + "step": 53451 + }, + { + "epoch": 2.06, + "learning_rate": 1.5759169017696844e-05, + "loss": 0.1687, + "step": 53452 + }, + { + "epoch": 2.06, + "learning_rate": 1.5758527827648116e-05, + "loss": 0.546, + "step": 53453 + }, + { + "epoch": 2.06, + "learning_rate": 1.5757886637599384e-05, + "loss": 0.1804, + "step": 53454 + }, + { + "epoch": 2.06, + "learning_rate": 1.5757245447550656e-05, + "loss": 0.3928, + "step": 53455 + }, + { + "epoch": 2.06, + "learning_rate": 1.5756604257501924e-05, + "loss": 0.1897, + "step": 53456 + }, + { + "epoch": 2.06, + "learning_rate": 1.5755963067453196e-05, + "loss": 0.212, + "step": 53457 + }, + { + "epoch": 2.06, + "learning_rate": 1.5755321877404464e-05, + "loss": 0.0833, + "step": 53458 + }, + { + "epoch": 2.06, + "learning_rate": 1.5754680687355733e-05, + "loss": 0.1744, + "step": 53459 + }, + { + "epoch": 2.06, + "learning_rate": 1.5754039497307e-05, + "loss": 0.2428, + "step": 53460 + }, + { + "epoch": 2.06, + "learning_rate": 1.5753398307258273e-05, + "loss": 0.161, + "step": 53461 + }, + { + "epoch": 2.06, + "learning_rate": 1.575275711720954e-05, + "loss": 0.1771, + "step": 53462 + }, + { + "epoch": 2.06, + "learning_rate": 1.575211592716081e-05, + "loss": 0.1636, + "step": 53463 + }, + { + "epoch": 2.06, + "learning_rate": 1.575147473711208e-05, + "loss": 0.3087, + "step": 53464 + }, + { + "epoch": 2.06, + "learning_rate": 1.575083354706335e-05, + "loss": 0.2986, + "step": 53465 + }, + { + "epoch": 2.06, + "learning_rate": 1.575019235701462e-05, + "loss": 0.1354, + "step": 53466 + }, + { + "epoch": 2.06, + "learning_rate": 1.574955116696589e-05, + "loss": 0.1763, + "step": 53467 + }, + { + "epoch": 2.06, + "learning_rate": 1.5748909976917158e-05, + "loss": 0.2502, + "step": 53468 + }, + { + "epoch": 2.06, + "learning_rate": 1.574826878686843e-05, + "loss": 0.2245, + "step": 53469 + }, + { + "epoch": 2.06, + "learning_rate": 1.5747627596819698e-05, + "loss": 0.2061, + "step": 53470 + }, + { + "epoch": 2.06, + "learning_rate": 1.5746986406770966e-05, + "loss": 0.4063, + "step": 53471 + }, + { + "epoch": 2.06, + "learning_rate": 1.5746345216722238e-05, + "loss": 0.219, + "step": 53472 + }, + { + "epoch": 2.06, + "learning_rate": 1.5745704026673506e-05, + "loss": 0.3112, + "step": 53473 + }, + { + "epoch": 2.06, + "learning_rate": 1.5745062836624778e-05, + "loss": 0.2759, + "step": 53474 + }, + { + "epoch": 2.06, + "learning_rate": 1.5744421646576046e-05, + "loss": 0.2489, + "step": 53475 + }, + { + "epoch": 2.06, + "learning_rate": 1.5743780456527315e-05, + "loss": 0.2837, + "step": 53476 + }, + { + "epoch": 2.06, + "learning_rate": 1.5743139266478586e-05, + "loss": 0.3345, + "step": 53477 + }, + { + "epoch": 2.06, + "learning_rate": 1.5742498076429855e-05, + "loss": 0.3303, + "step": 53478 + }, + { + "epoch": 2.06, + "learning_rate": 1.5741856886381123e-05, + "loss": 0.2285, + "step": 53479 + }, + { + "epoch": 2.06, + "learning_rate": 1.574121569633239e-05, + "loss": 0.3499, + "step": 53480 + }, + { + "epoch": 2.06, + "learning_rate": 1.5740574506283663e-05, + "loss": 0.2389, + "step": 53481 + }, + { + "epoch": 2.06, + "learning_rate": 1.5739933316234935e-05, + "loss": 0.3032, + "step": 53482 + }, + { + "epoch": 2.06, + "learning_rate": 1.5739292126186203e-05, + "loss": 0.1368, + "step": 53483 + }, + { + "epoch": 2.06, + "learning_rate": 1.573865093613747e-05, + "loss": 0.2451, + "step": 53484 + }, + { + "epoch": 2.06, + "learning_rate": 1.5738009746088743e-05, + "loss": 0.2162, + "step": 53485 + }, + { + "epoch": 2.06, + "learning_rate": 1.573736855604001e-05, + "loss": 0.2321, + "step": 53486 + }, + { + "epoch": 2.06, + "learning_rate": 1.573672736599128e-05, + "loss": 0.2248, + "step": 53487 + }, + { + "epoch": 2.06, + "learning_rate": 1.5736086175942548e-05, + "loss": 0.1753, + "step": 53488 + }, + { + "epoch": 2.06, + "learning_rate": 1.573544498589382e-05, + "loss": 0.2202, + "step": 53489 + }, + { + "epoch": 2.06, + "learning_rate": 1.573480379584509e-05, + "loss": 0.3465, + "step": 53490 + }, + { + "epoch": 2.06, + "learning_rate": 1.573416260579636e-05, + "loss": 0.3623, + "step": 53491 + }, + { + "epoch": 2.06, + "learning_rate": 1.5733521415747628e-05, + "loss": 0.2997, + "step": 53492 + }, + { + "epoch": 2.06, + "learning_rate": 1.57328802256989e-05, + "loss": 0.1891, + "step": 53493 + }, + { + "epoch": 2.06, + "learning_rate": 1.5732239035650168e-05, + "loss": 0.351, + "step": 53494 + }, + { + "epoch": 2.06, + "learning_rate": 1.5731597845601437e-05, + "loss": 0.1319, + "step": 53495 + }, + { + "epoch": 2.06, + "learning_rate": 1.5730956655552705e-05, + "loss": 0.1884, + "step": 53496 + }, + { + "epoch": 2.06, + "learning_rate": 1.5730315465503977e-05, + "loss": 0.2066, + "step": 53497 + }, + { + "epoch": 2.06, + "learning_rate": 1.5729674275455245e-05, + "loss": 0.1493, + "step": 53498 + }, + { + "epoch": 2.06, + "learning_rate": 1.5729033085406517e-05, + "loss": 0.187, + "step": 53499 + }, + { + "epoch": 2.06, + "learning_rate": 1.5728391895357785e-05, + "loss": 0.2125, + "step": 53500 + }, + { + "epoch": 2.06, + "learning_rate": 1.5727750705309057e-05, + "loss": 0.253, + "step": 53501 + }, + { + "epoch": 2.06, + "learning_rate": 1.5727109515260325e-05, + "loss": 0.2256, + "step": 53502 + }, + { + "epoch": 2.06, + "learning_rate": 1.5726468325211593e-05, + "loss": 0.185, + "step": 53503 + }, + { + "epoch": 2.06, + "learning_rate": 1.5725827135162862e-05, + "loss": 0.2361, + "step": 53504 + }, + { + "epoch": 2.06, + "learning_rate": 1.5725185945114133e-05, + "loss": 0.2333, + "step": 53505 + }, + { + "epoch": 2.06, + "learning_rate": 1.5724544755065402e-05, + "loss": 0.3726, + "step": 53506 + }, + { + "epoch": 2.06, + "learning_rate": 1.572390356501667e-05, + "loss": 0.256, + "step": 53507 + }, + { + "epoch": 2.06, + "learning_rate": 1.5723262374967942e-05, + "loss": 0.261, + "step": 53508 + }, + { + "epoch": 2.06, + "learning_rate": 1.572262118491921e-05, + "loss": 0.2714, + "step": 53509 + }, + { + "epoch": 2.06, + "learning_rate": 1.5721979994870482e-05, + "loss": 0.2741, + "step": 53510 + }, + { + "epoch": 2.06, + "learning_rate": 1.572133880482175e-05, + "loss": 0.2264, + "step": 53511 + }, + { + "epoch": 2.06, + "learning_rate": 1.572069761477302e-05, + "loss": 0.1244, + "step": 53512 + }, + { + "epoch": 2.06, + "learning_rate": 1.572005642472429e-05, + "loss": 0.1709, + "step": 53513 + }, + { + "epoch": 2.06, + "learning_rate": 1.571941523467556e-05, + "loss": 0.2127, + "step": 53514 + }, + { + "epoch": 2.06, + "learning_rate": 1.5718774044626827e-05, + "loss": 0.2568, + "step": 53515 + }, + { + "epoch": 2.06, + "learning_rate": 1.5718132854578095e-05, + "loss": 0.2291, + "step": 53516 + }, + { + "epoch": 2.06, + "learning_rate": 1.5717491664529367e-05, + "loss": 0.2017, + "step": 53517 + }, + { + "epoch": 2.06, + "learning_rate": 1.571685047448064e-05, + "loss": 0.2677, + "step": 53518 + }, + { + "epoch": 2.06, + "learning_rate": 1.5716209284431907e-05, + "loss": 0.1717, + "step": 53519 + }, + { + "epoch": 2.06, + "learning_rate": 1.5715568094383175e-05, + "loss": 0.2864, + "step": 53520 + }, + { + "epoch": 2.06, + "learning_rate": 1.5714926904334447e-05, + "loss": 0.1976, + "step": 53521 + }, + { + "epoch": 2.06, + "learning_rate": 1.5714285714285715e-05, + "loss": 0.224, + "step": 53522 + }, + { + "epoch": 2.06, + "learning_rate": 1.5713644524236984e-05, + "loss": 0.1312, + "step": 53523 + }, + { + "epoch": 2.06, + "learning_rate": 1.5713003334188252e-05, + "loss": 0.2736, + "step": 53524 + }, + { + "epoch": 2.06, + "learning_rate": 1.5712362144139524e-05, + "loss": 0.3015, + "step": 53525 + }, + { + "epoch": 2.06, + "learning_rate": 1.5711720954090795e-05, + "loss": 0.2096, + "step": 53526 + }, + { + "epoch": 2.06, + "learning_rate": 1.5711079764042064e-05, + "loss": 0.2444, + "step": 53527 + }, + { + "epoch": 2.06, + "learning_rate": 1.5710438573993332e-05, + "loss": 0.1542, + "step": 53528 + }, + { + "epoch": 2.06, + "learning_rate": 1.5709797383944604e-05, + "loss": 0.2874, + "step": 53529 + }, + { + "epoch": 2.06, + "learning_rate": 1.5709156193895872e-05, + "loss": 0.0847, + "step": 53530 + }, + { + "epoch": 2.06, + "learning_rate": 1.570851500384714e-05, + "loss": 0.2602, + "step": 53531 + }, + { + "epoch": 2.06, + "learning_rate": 1.570787381379841e-05, + "loss": 0.4079, + "step": 53532 + }, + { + "epoch": 2.06, + "learning_rate": 1.570723262374968e-05, + "loss": 0.318, + "step": 53533 + }, + { + "epoch": 2.06, + "learning_rate": 1.570659143370095e-05, + "loss": 0.3071, + "step": 53534 + }, + { + "epoch": 2.06, + "learning_rate": 1.570595024365222e-05, + "loss": 0.1763, + "step": 53535 + }, + { + "epoch": 2.06, + "learning_rate": 1.570530905360349e-05, + "loss": 0.2107, + "step": 53536 + }, + { + "epoch": 2.06, + "learning_rate": 1.570466786355476e-05, + "loss": 0.181, + "step": 53537 + }, + { + "epoch": 2.06, + "learning_rate": 1.570402667350603e-05, + "loss": 0.2109, + "step": 53538 + }, + { + "epoch": 2.06, + "learning_rate": 1.5703385483457297e-05, + "loss": 0.1893, + "step": 53539 + }, + { + "epoch": 2.06, + "learning_rate": 1.5702744293408566e-05, + "loss": 0.2865, + "step": 53540 + }, + { + "epoch": 2.06, + "learning_rate": 1.5702103103359837e-05, + "loss": 0.1447, + "step": 53541 + }, + { + "epoch": 2.06, + "learning_rate": 1.5701461913311106e-05, + "loss": 0.2805, + "step": 53542 + }, + { + "epoch": 2.06, + "learning_rate": 1.5700820723262374e-05, + "loss": 0.28, + "step": 53543 + }, + { + "epoch": 2.06, + "learning_rate": 1.5700179533213646e-05, + "loss": 0.1953, + "step": 53544 + }, + { + "epoch": 2.06, + "learning_rate": 1.5699538343164914e-05, + "loss": 0.3086, + "step": 53545 + }, + { + "epoch": 2.06, + "learning_rate": 1.5698897153116186e-05, + "loss": 0.2394, + "step": 53546 + }, + { + "epoch": 2.06, + "learning_rate": 1.5698255963067454e-05, + "loss": 0.2231, + "step": 53547 + }, + { + "epoch": 2.06, + "learning_rate": 1.5697614773018722e-05, + "loss": 0.1666, + "step": 53548 + }, + { + "epoch": 2.06, + "learning_rate": 1.5696973582969994e-05, + "loss": 0.1619, + "step": 53549 + }, + { + "epoch": 2.06, + "learning_rate": 1.5696332392921262e-05, + "loss": 0.2825, + "step": 53550 + }, + { + "epoch": 2.06, + "learning_rate": 1.569569120287253e-05, + "loss": 0.2505, + "step": 53551 + }, + { + "epoch": 2.06, + "learning_rate": 1.56950500128238e-05, + "loss": 0.2859, + "step": 53552 + }, + { + "epoch": 2.06, + "learning_rate": 1.569440882277507e-05, + "loss": 0.1806, + "step": 53553 + }, + { + "epoch": 2.06, + "learning_rate": 1.5693767632726342e-05, + "loss": 0.3022, + "step": 53554 + }, + { + "epoch": 2.06, + "learning_rate": 1.569312644267761e-05, + "loss": 0.3108, + "step": 53555 + }, + { + "epoch": 2.06, + "learning_rate": 1.569248525262888e-05, + "loss": 0.2744, + "step": 53556 + }, + { + "epoch": 2.06, + "learning_rate": 1.569184406258015e-05, + "loss": 0.2266, + "step": 53557 + }, + { + "epoch": 2.06, + "learning_rate": 1.569120287253142e-05, + "loss": 0.1242, + "step": 53558 + }, + { + "epoch": 2.06, + "learning_rate": 1.5690561682482688e-05, + "loss": 0.2504, + "step": 53559 + }, + { + "epoch": 2.06, + "learning_rate": 1.5689920492433956e-05, + "loss": 0.2629, + "step": 53560 + }, + { + "epoch": 2.06, + "learning_rate": 1.5689279302385228e-05, + "loss": 0.1463, + "step": 53561 + }, + { + "epoch": 2.06, + "learning_rate": 1.56886381123365e-05, + "loss": 0.1945, + "step": 53562 + }, + { + "epoch": 2.06, + "learning_rate": 1.5687996922287768e-05, + "loss": 0.4338, + "step": 53563 + }, + { + "epoch": 2.06, + "learning_rate": 1.5687355732239036e-05, + "loss": 0.2949, + "step": 53564 + }, + { + "epoch": 2.06, + "learning_rate": 1.5686714542190308e-05, + "loss": 0.3541, + "step": 53565 + }, + { + "epoch": 2.06, + "learning_rate": 1.5686073352141576e-05, + "loss": 0.2215, + "step": 53566 + }, + { + "epoch": 2.06, + "learning_rate": 1.5685432162092844e-05, + "loss": 0.2724, + "step": 53567 + }, + { + "epoch": 2.06, + "learning_rate": 1.5684790972044113e-05, + "loss": 0.1675, + "step": 53568 + }, + { + "epoch": 2.06, + "learning_rate": 1.5684149781995384e-05, + "loss": 0.1943, + "step": 53569 + }, + { + "epoch": 2.06, + "learning_rate": 1.5683508591946656e-05, + "loss": 0.2326, + "step": 53570 + }, + { + "epoch": 2.06, + "learning_rate": 1.5682867401897924e-05, + "loss": 0.3554, + "step": 53571 + }, + { + "epoch": 2.06, + "learning_rate": 1.5682226211849193e-05, + "loss": 0.1926, + "step": 53572 + }, + { + "epoch": 2.06, + "learning_rate": 1.5681585021800464e-05, + "loss": 0.1691, + "step": 53573 + }, + { + "epoch": 2.06, + "learning_rate": 1.5680943831751733e-05, + "loss": 0.3183, + "step": 53574 + }, + { + "epoch": 2.06, + "learning_rate": 1.5680302641703e-05, + "loss": 0.2206, + "step": 53575 + }, + { + "epoch": 2.06, + "learning_rate": 1.567966145165427e-05, + "loss": 0.2746, + "step": 53576 + }, + { + "epoch": 2.06, + "learning_rate": 1.567902026160554e-05, + "loss": 0.3266, + "step": 53577 + }, + { + "epoch": 2.06, + "learning_rate": 1.567837907155681e-05, + "loss": 0.2257, + "step": 53578 + }, + { + "epoch": 2.06, + "learning_rate": 1.567773788150808e-05, + "loss": 0.3031, + "step": 53579 + }, + { + "epoch": 2.06, + "learning_rate": 1.567709669145935e-05, + "loss": 0.171, + "step": 53580 + }, + { + "epoch": 2.06, + "learning_rate": 1.567645550141062e-05, + "loss": 0.229, + "step": 53581 + }, + { + "epoch": 2.06, + "learning_rate": 1.567581431136189e-05, + "loss": 0.1592, + "step": 53582 + }, + { + "epoch": 2.06, + "learning_rate": 1.5675173121313158e-05, + "loss": 0.2226, + "step": 53583 + }, + { + "epoch": 2.06, + "learning_rate": 1.5674531931264426e-05, + "loss": 0.2058, + "step": 53584 + }, + { + "epoch": 2.06, + "learning_rate": 1.5673890741215698e-05, + "loss": 0.2619, + "step": 53585 + }, + { + "epoch": 2.06, + "learning_rate": 1.5673249551166966e-05, + "loss": 0.2174, + "step": 53586 + }, + { + "epoch": 2.06, + "learning_rate": 1.5672608361118235e-05, + "loss": 0.2117, + "step": 53587 + }, + { + "epoch": 2.06, + "learning_rate": 1.5671967171069506e-05, + "loss": 0.3258, + "step": 53588 + }, + { + "epoch": 2.06, + "learning_rate": 1.5671325981020775e-05, + "loss": 0.1993, + "step": 53589 + }, + { + "epoch": 2.06, + "learning_rate": 1.5670684790972046e-05, + "loss": 0.2963, + "step": 53590 + }, + { + "epoch": 2.06, + "learning_rate": 1.5670043600923315e-05, + "loss": 0.2297, + "step": 53591 + }, + { + "epoch": 2.06, + "learning_rate": 1.5669402410874583e-05, + "loss": 0.0947, + "step": 53592 + }, + { + "epoch": 2.06, + "learning_rate": 1.5668761220825855e-05, + "loss": 0.3053, + "step": 53593 + }, + { + "epoch": 2.06, + "learning_rate": 1.5668120030777123e-05, + "loss": 0.2918, + "step": 53594 + }, + { + "epoch": 2.06, + "learning_rate": 1.566747884072839e-05, + "loss": 0.1986, + "step": 53595 + }, + { + "epoch": 2.06, + "learning_rate": 1.566683765067966e-05, + "loss": 0.2632, + "step": 53596 + }, + { + "epoch": 2.06, + "learning_rate": 1.566619646063093e-05, + "loss": 0.1766, + "step": 53597 + }, + { + "epoch": 2.06, + "learning_rate": 1.5665555270582203e-05, + "loss": 0.2238, + "step": 53598 + }, + { + "epoch": 2.06, + "learning_rate": 1.566491408053347e-05, + "loss": 0.2546, + "step": 53599 + }, + { + "epoch": 2.06, + "learning_rate": 1.566427289048474e-05, + "loss": 0.3131, + "step": 53600 + }, + { + "epoch": 2.06, + "learning_rate": 1.566363170043601e-05, + "loss": 0.169, + "step": 53601 + }, + { + "epoch": 2.06, + "learning_rate": 1.566299051038728e-05, + "loss": 0.4086, + "step": 53602 + }, + { + "epoch": 2.06, + "learning_rate": 1.5662349320338548e-05, + "loss": 0.2573, + "step": 53603 + }, + { + "epoch": 2.06, + "learning_rate": 1.5661708130289817e-05, + "loss": 0.3102, + "step": 53604 + }, + { + "epoch": 2.06, + "learning_rate": 1.5661066940241088e-05, + "loss": 0.213, + "step": 53605 + }, + { + "epoch": 2.06, + "learning_rate": 1.566042575019236e-05, + "loss": 0.18, + "step": 53606 + }, + { + "epoch": 2.06, + "learning_rate": 1.5659784560143628e-05, + "loss": 0.1457, + "step": 53607 + }, + { + "epoch": 2.06, + "learning_rate": 1.5659143370094897e-05, + "loss": 0.3517, + "step": 53608 + }, + { + "epoch": 2.06, + "learning_rate": 1.565850218004617e-05, + "loss": 0.2885, + "step": 53609 + }, + { + "epoch": 2.06, + "learning_rate": 1.5657860989997437e-05, + "loss": 0.2178, + "step": 53610 + }, + { + "epoch": 2.06, + "learning_rate": 1.5657219799948705e-05, + "loss": 0.2741, + "step": 53611 + }, + { + "epoch": 2.06, + "learning_rate": 1.5656578609899973e-05, + "loss": 0.3393, + "step": 53612 + }, + { + "epoch": 2.06, + "learning_rate": 1.5655937419851245e-05, + "loss": 0.3023, + "step": 53613 + }, + { + "epoch": 2.06, + "learning_rate": 1.5655296229802513e-05, + "loss": 0.2346, + "step": 53614 + }, + { + "epoch": 2.06, + "learning_rate": 1.5654655039753785e-05, + "loss": 0.2609, + "step": 53615 + }, + { + "epoch": 2.06, + "learning_rate": 1.5654013849705053e-05, + "loss": 0.1997, + "step": 53616 + }, + { + "epoch": 2.06, + "learning_rate": 1.5653372659656325e-05, + "loss": 0.2426, + "step": 53617 + }, + { + "epoch": 2.06, + "learning_rate": 1.5652731469607593e-05, + "loss": 0.2883, + "step": 53618 + }, + { + "epoch": 2.06, + "learning_rate": 1.5652090279558862e-05, + "loss": 0.3576, + "step": 53619 + }, + { + "epoch": 2.06, + "learning_rate": 1.565144908951013e-05, + "loss": 0.1549, + "step": 53620 + }, + { + "epoch": 2.06, + "learning_rate": 1.5650807899461402e-05, + "loss": 0.1985, + "step": 53621 + }, + { + "epoch": 2.06, + "learning_rate": 1.565016670941267e-05, + "loss": 0.2299, + "step": 53622 + }, + { + "epoch": 2.06, + "learning_rate": 1.564952551936394e-05, + "loss": 0.1329, + "step": 53623 + }, + { + "epoch": 2.06, + "learning_rate": 1.564888432931521e-05, + "loss": 0.3688, + "step": 53624 + }, + { + "epoch": 2.06, + "learning_rate": 1.5648243139266482e-05, + "loss": 0.3053, + "step": 53625 + }, + { + "epoch": 2.06, + "learning_rate": 1.564760194921775e-05, + "loss": 0.1871, + "step": 53626 + }, + { + "epoch": 2.06, + "learning_rate": 1.564696075916902e-05, + "loss": 0.2196, + "step": 53627 + }, + { + "epoch": 2.06, + "learning_rate": 1.5646319569120287e-05, + "loss": 0.2697, + "step": 53628 + }, + { + "epoch": 2.06, + "learning_rate": 1.564567837907156e-05, + "loss": 0.2513, + "step": 53629 + }, + { + "epoch": 2.06, + "learning_rate": 1.5645037189022827e-05, + "loss": 0.161, + "step": 53630 + }, + { + "epoch": 2.06, + "learning_rate": 1.5644395998974095e-05, + "loss": 0.3271, + "step": 53631 + }, + { + "epoch": 2.06, + "learning_rate": 1.5643754808925364e-05, + "loss": 0.2365, + "step": 53632 + }, + { + "epoch": 2.06, + "learning_rate": 1.5643113618876635e-05, + "loss": 0.2839, + "step": 53633 + }, + { + "epoch": 2.06, + "learning_rate": 1.5642472428827907e-05, + "loss": 0.3124, + "step": 53634 + }, + { + "epoch": 2.06, + "learning_rate": 1.5641831238779175e-05, + "loss": 0.2697, + "step": 53635 + }, + { + "epoch": 2.06, + "learning_rate": 1.5641190048730444e-05, + "loss": 0.1646, + "step": 53636 + }, + { + "epoch": 2.06, + "learning_rate": 1.5640548858681715e-05, + "loss": 0.2428, + "step": 53637 + }, + { + "epoch": 2.06, + "learning_rate": 1.5639907668632984e-05, + "loss": 0.185, + "step": 53638 + }, + { + "epoch": 2.06, + "learning_rate": 1.5639266478584252e-05, + "loss": 0.1878, + "step": 53639 + }, + { + "epoch": 2.06, + "learning_rate": 1.563862528853552e-05, + "loss": 0.3615, + "step": 53640 + }, + { + "epoch": 2.06, + "learning_rate": 1.5637984098486792e-05, + "loss": 0.2344, + "step": 53641 + }, + { + "epoch": 2.06, + "learning_rate": 1.5637342908438064e-05, + "loss": 0.2538, + "step": 53642 + }, + { + "epoch": 2.06, + "learning_rate": 1.5636701718389332e-05, + "loss": 0.2606, + "step": 53643 + }, + { + "epoch": 2.06, + "learning_rate": 1.56360605283406e-05, + "loss": 0.2687, + "step": 53644 + }, + { + "epoch": 2.06, + "learning_rate": 1.5635419338291872e-05, + "loss": 0.2886, + "step": 53645 + }, + { + "epoch": 2.06, + "learning_rate": 1.563477814824314e-05, + "loss": 0.277, + "step": 53646 + }, + { + "epoch": 2.06, + "learning_rate": 1.563413695819441e-05, + "loss": 0.2372, + "step": 53647 + }, + { + "epoch": 2.06, + "learning_rate": 1.5633495768145677e-05, + "loss": 0.1583, + "step": 53648 + }, + { + "epoch": 2.06, + "learning_rate": 1.563285457809695e-05, + "loss": 0.2291, + "step": 53649 + }, + { + "epoch": 2.06, + "learning_rate": 1.5632213388048217e-05, + "loss": 0.2767, + "step": 53650 + }, + { + "epoch": 2.06, + "learning_rate": 1.563157219799949e-05, + "loss": 0.2162, + "step": 53651 + }, + { + "epoch": 2.06, + "learning_rate": 1.5630931007950757e-05, + "loss": 0.2452, + "step": 53652 + }, + { + "epoch": 2.06, + "learning_rate": 1.563028981790203e-05, + "loss": 0.2588, + "step": 53653 + }, + { + "epoch": 2.06, + "learning_rate": 1.5629648627853297e-05, + "loss": 0.2837, + "step": 53654 + }, + { + "epoch": 2.06, + "learning_rate": 1.5629007437804566e-05, + "loss": 0.264, + "step": 53655 + }, + { + "epoch": 2.06, + "learning_rate": 1.5628366247755834e-05, + "loss": 0.1262, + "step": 53656 + }, + { + "epoch": 2.06, + "learning_rate": 1.5627725057707106e-05, + "loss": 0.2378, + "step": 53657 + }, + { + "epoch": 2.06, + "learning_rate": 1.5627083867658374e-05, + "loss": 0.3868, + "step": 53658 + }, + { + "epoch": 2.06, + "learning_rate": 1.5626442677609642e-05, + "loss": 0.1922, + "step": 53659 + }, + { + "epoch": 2.06, + "learning_rate": 1.5625801487560914e-05, + "loss": 0.2705, + "step": 53660 + }, + { + "epoch": 2.06, + "learning_rate": 1.5625160297512186e-05, + "loss": 0.2454, + "step": 53661 + }, + { + "epoch": 2.06, + "learning_rate": 1.5624519107463454e-05, + "loss": 0.1343, + "step": 53662 + }, + { + "epoch": 2.06, + "learning_rate": 1.5623877917414722e-05, + "loss": 0.3276, + "step": 53663 + }, + { + "epoch": 2.06, + "learning_rate": 1.562323672736599e-05, + "loss": 0.1569, + "step": 53664 + }, + { + "epoch": 2.06, + "learning_rate": 1.5622595537317262e-05, + "loss": 0.1437, + "step": 53665 + }, + { + "epoch": 2.06, + "learning_rate": 1.562195434726853e-05, + "loss": 0.1384, + "step": 53666 + }, + { + "epoch": 2.06, + "learning_rate": 1.56213131572198e-05, + "loss": 0.1389, + "step": 53667 + }, + { + "epoch": 2.06, + "learning_rate": 1.562067196717107e-05, + "loss": 0.2971, + "step": 53668 + }, + { + "epoch": 2.06, + "learning_rate": 1.562003077712234e-05, + "loss": 0.1667, + "step": 53669 + }, + { + "epoch": 2.06, + "learning_rate": 1.561938958707361e-05, + "loss": 0.2401, + "step": 53670 + }, + { + "epoch": 2.06, + "learning_rate": 1.561874839702488e-05, + "loss": 0.3046, + "step": 53671 + }, + { + "epoch": 2.06, + "learning_rate": 1.5618107206976148e-05, + "loss": 0.2493, + "step": 53672 + }, + { + "epoch": 2.06, + "learning_rate": 1.561746601692742e-05, + "loss": 0.1933, + "step": 53673 + }, + { + "epoch": 2.06, + "learning_rate": 1.5616824826878688e-05, + "loss": 0.1978, + "step": 53674 + }, + { + "epoch": 2.06, + "learning_rate": 1.5616183636829956e-05, + "loss": 0.2178, + "step": 53675 + }, + { + "epoch": 2.06, + "learning_rate": 1.5615542446781224e-05, + "loss": 0.3627, + "step": 53676 + }, + { + "epoch": 2.06, + "learning_rate": 1.5614901256732496e-05, + "loss": 0.1958, + "step": 53677 + }, + { + "epoch": 2.06, + "learning_rate": 1.5614260066683768e-05, + "loss": 0.329, + "step": 53678 + }, + { + "epoch": 2.06, + "learning_rate": 1.5613618876635036e-05, + "loss": 0.2207, + "step": 53679 + }, + { + "epoch": 2.06, + "learning_rate": 1.5612977686586304e-05, + "loss": 0.1334, + "step": 53680 + }, + { + "epoch": 2.06, + "learning_rate": 1.5612336496537576e-05, + "loss": 0.2172, + "step": 53681 + }, + { + "epoch": 2.06, + "learning_rate": 1.5611695306488844e-05, + "loss": 0.3556, + "step": 53682 + }, + { + "epoch": 2.06, + "learning_rate": 1.5611054116440113e-05, + "loss": 0.2658, + "step": 53683 + }, + { + "epoch": 2.06, + "learning_rate": 1.561041292639138e-05, + "loss": 0.2332, + "step": 53684 + }, + { + "epoch": 2.06, + "learning_rate": 1.5609771736342653e-05, + "loss": 0.3044, + "step": 53685 + }, + { + "epoch": 2.06, + "learning_rate": 1.5609130546293924e-05, + "loss": 0.2105, + "step": 53686 + }, + { + "epoch": 2.06, + "learning_rate": 1.5608489356245193e-05, + "loss": 0.1328, + "step": 53687 + }, + { + "epoch": 2.06, + "learning_rate": 1.560784816619646e-05, + "loss": 0.2699, + "step": 53688 + }, + { + "epoch": 2.06, + "learning_rate": 1.5607206976147733e-05, + "loss": 0.3959, + "step": 53689 + }, + { + "epoch": 2.06, + "learning_rate": 1.5606565786099e-05, + "loss": 0.225, + "step": 53690 + }, + { + "epoch": 2.06, + "learning_rate": 1.560592459605027e-05, + "loss": 0.208, + "step": 53691 + }, + { + "epoch": 2.06, + "learning_rate": 1.5605283406001538e-05, + "loss": 0.1904, + "step": 53692 + }, + { + "epoch": 2.06, + "learning_rate": 1.560464221595281e-05, + "loss": 0.4166, + "step": 53693 + }, + { + "epoch": 2.06, + "learning_rate": 1.5604001025904078e-05, + "loss": 0.1703, + "step": 53694 + }, + { + "epoch": 2.06, + "learning_rate": 1.560335983585535e-05, + "loss": 0.2124, + "step": 53695 + }, + { + "epoch": 2.06, + "learning_rate": 1.5602718645806618e-05, + "loss": 0.2127, + "step": 53696 + }, + { + "epoch": 2.06, + "learning_rate": 1.560207745575789e-05, + "loss": 0.1948, + "step": 53697 + }, + { + "epoch": 2.06, + "learning_rate": 1.5601436265709158e-05, + "loss": 0.3389, + "step": 53698 + }, + { + "epoch": 2.06, + "learning_rate": 1.5600795075660426e-05, + "loss": 0.2972, + "step": 53699 + }, + { + "epoch": 2.06, + "learning_rate": 1.5600153885611695e-05, + "loss": 0.3604, + "step": 53700 + }, + { + "epoch": 2.06, + "learning_rate": 1.5599512695562966e-05, + "loss": 0.2509, + "step": 53701 + }, + { + "epoch": 2.06, + "learning_rate": 1.5598871505514235e-05, + "loss": 0.226, + "step": 53702 + }, + { + "epoch": 2.06, + "learning_rate": 1.5598230315465503e-05, + "loss": 0.3035, + "step": 53703 + }, + { + "epoch": 2.06, + "learning_rate": 1.5597589125416775e-05, + "loss": 0.2158, + "step": 53704 + }, + { + "epoch": 2.06, + "learning_rate": 1.5596947935368046e-05, + "loss": 0.1612, + "step": 53705 + }, + { + "epoch": 2.06, + "learning_rate": 1.5596306745319315e-05, + "loss": 0.1789, + "step": 53706 + }, + { + "epoch": 2.06, + "learning_rate": 1.5595665555270583e-05, + "loss": 0.1533, + "step": 53707 + }, + { + "epoch": 2.06, + "learning_rate": 1.559502436522185e-05, + "loss": 0.2119, + "step": 53708 + }, + { + "epoch": 2.06, + "learning_rate": 1.5594383175173123e-05, + "loss": 0.1698, + "step": 53709 + }, + { + "epoch": 2.06, + "learning_rate": 1.559374198512439e-05, + "loss": 0.2745, + "step": 53710 + }, + { + "epoch": 2.07, + "learning_rate": 1.559310079507566e-05, + "loss": 0.193, + "step": 53711 + }, + { + "epoch": 2.07, + "learning_rate": 1.5592459605026928e-05, + "loss": 0.2378, + "step": 53712 + }, + { + "epoch": 2.07, + "learning_rate": 1.55918184149782e-05, + "loss": 0.2527, + "step": 53713 + }, + { + "epoch": 2.07, + "learning_rate": 1.559117722492947e-05, + "loss": 0.2765, + "step": 53714 + }, + { + "epoch": 2.07, + "learning_rate": 1.559053603488074e-05, + "loss": 0.2282, + "step": 53715 + }, + { + "epoch": 2.07, + "learning_rate": 1.5589894844832008e-05, + "loss": 0.1918, + "step": 53716 + }, + { + "epoch": 2.07, + "learning_rate": 1.558925365478328e-05, + "loss": 0.1907, + "step": 53717 + }, + { + "epoch": 2.07, + "learning_rate": 1.5588612464734548e-05, + "loss": 0.2653, + "step": 53718 + }, + { + "epoch": 2.07, + "learning_rate": 1.5587971274685817e-05, + "loss": 0.2226, + "step": 53719 + }, + { + "epoch": 2.07, + "learning_rate": 1.5587330084637085e-05, + "loss": 0.3118, + "step": 53720 + }, + { + "epoch": 2.07, + "learning_rate": 1.5586688894588357e-05, + "loss": 0.148, + "step": 53721 + }, + { + "epoch": 2.07, + "learning_rate": 1.558604770453963e-05, + "loss": 0.2343, + "step": 53722 + }, + { + "epoch": 2.07, + "learning_rate": 1.5585406514490897e-05, + "loss": 0.2014, + "step": 53723 + }, + { + "epoch": 2.07, + "learning_rate": 1.5584765324442165e-05, + "loss": 0.3223, + "step": 53724 + }, + { + "epoch": 2.07, + "learning_rate": 1.5584124134393437e-05, + "loss": 0.2255, + "step": 53725 + }, + { + "epoch": 2.07, + "learning_rate": 1.5583482944344705e-05, + "loss": 0.2486, + "step": 53726 + }, + { + "epoch": 2.07, + "learning_rate": 1.5582841754295973e-05, + "loss": 0.22, + "step": 53727 + }, + { + "epoch": 2.07, + "learning_rate": 1.5582200564247242e-05, + "loss": 0.2186, + "step": 53728 + }, + { + "epoch": 2.07, + "learning_rate": 1.5581559374198513e-05, + "loss": 0.2043, + "step": 53729 + }, + { + "epoch": 2.07, + "learning_rate": 1.5580918184149782e-05, + "loss": 0.204, + "step": 53730 + }, + { + "epoch": 2.07, + "learning_rate": 1.5580276994101053e-05, + "loss": 0.1955, + "step": 53731 + }, + { + "epoch": 2.07, + "learning_rate": 1.5579635804052322e-05, + "loss": 0.124, + "step": 53732 + }, + { + "epoch": 2.07, + "learning_rate": 1.5578994614003594e-05, + "loss": 0.2536, + "step": 53733 + }, + { + "epoch": 2.07, + "learning_rate": 1.5578353423954862e-05, + "loss": 0.1093, + "step": 53734 + }, + { + "epoch": 2.07, + "learning_rate": 1.557771223390613e-05, + "loss": 0.2345, + "step": 53735 + }, + { + "epoch": 2.07, + "learning_rate": 1.55770710438574e-05, + "loss": 0.2383, + "step": 53736 + }, + { + "epoch": 2.07, + "learning_rate": 1.557642985380867e-05, + "loss": 0.1748, + "step": 53737 + }, + { + "epoch": 2.07, + "learning_rate": 1.557578866375994e-05, + "loss": 0.2176, + "step": 53738 + }, + { + "epoch": 2.07, + "learning_rate": 1.5575147473711207e-05, + "loss": 0.1195, + "step": 53739 + }, + { + "epoch": 2.07, + "learning_rate": 1.557450628366248e-05, + "loss": 0.1911, + "step": 53740 + }, + { + "epoch": 2.07, + "learning_rate": 1.557386509361375e-05, + "loss": 0.2944, + "step": 53741 + }, + { + "epoch": 2.07, + "learning_rate": 1.557322390356502e-05, + "loss": 0.2514, + "step": 53742 + }, + { + "epoch": 2.07, + "learning_rate": 1.5572582713516287e-05, + "loss": 0.2338, + "step": 53743 + }, + { + "epoch": 2.07, + "learning_rate": 1.5571941523467555e-05, + "loss": 0.1882, + "step": 53744 + }, + { + "epoch": 2.07, + "learning_rate": 1.5571300333418827e-05, + "loss": 0.1488, + "step": 53745 + }, + { + "epoch": 2.07, + "learning_rate": 1.5570659143370095e-05, + "loss": 0.2848, + "step": 53746 + }, + { + "epoch": 2.07, + "learning_rate": 1.5570017953321364e-05, + "loss": 0.3044, + "step": 53747 + }, + { + "epoch": 2.07, + "learning_rate": 1.5569376763272632e-05, + "loss": 0.2523, + "step": 53748 + }, + { + "epoch": 2.07, + "learning_rate": 1.5568735573223904e-05, + "loss": 0.1902, + "step": 53749 + }, + { + "epoch": 2.07, + "learning_rate": 1.5568094383175175e-05, + "loss": 0.1109, + "step": 53750 + }, + { + "epoch": 2.07, + "learning_rate": 1.5567453193126444e-05, + "loss": 0.2868, + "step": 53751 + }, + { + "epoch": 2.07, + "learning_rate": 1.5566812003077712e-05, + "loss": 0.2066, + "step": 53752 + }, + { + "epoch": 2.07, + "learning_rate": 1.5566170813028984e-05, + "loss": 0.3151, + "step": 53753 + }, + { + "epoch": 2.07, + "learning_rate": 1.5565529622980252e-05, + "loss": 0.1829, + "step": 53754 + }, + { + "epoch": 2.07, + "learning_rate": 1.556488843293152e-05, + "loss": 0.2709, + "step": 53755 + }, + { + "epoch": 2.07, + "learning_rate": 1.556424724288279e-05, + "loss": 0.2353, + "step": 53756 + }, + { + "epoch": 2.07, + "learning_rate": 1.556360605283406e-05, + "loss": 0.2327, + "step": 53757 + }, + { + "epoch": 2.07, + "learning_rate": 1.5562964862785332e-05, + "loss": 0.3025, + "step": 53758 + }, + { + "epoch": 2.07, + "learning_rate": 1.55623236727366e-05, + "loss": 0.2684, + "step": 53759 + }, + { + "epoch": 2.07, + "learning_rate": 1.556168248268787e-05, + "loss": 0.2487, + "step": 53760 + }, + { + "epoch": 2.07, + "learning_rate": 1.556104129263914e-05, + "loss": 0.23, + "step": 53761 + }, + { + "epoch": 2.07, + "learning_rate": 1.556040010259041e-05, + "loss": 0.3757, + "step": 53762 + }, + { + "epoch": 2.07, + "learning_rate": 1.5559758912541677e-05, + "loss": 0.3001, + "step": 53763 + }, + { + "epoch": 2.07, + "learning_rate": 1.5559117722492946e-05, + "loss": 0.1291, + "step": 53764 + }, + { + "epoch": 2.07, + "learning_rate": 1.5558476532444217e-05, + "loss": 0.1808, + "step": 53765 + }, + { + "epoch": 2.07, + "learning_rate": 1.555783534239549e-05, + "loss": 0.2193, + "step": 53766 + }, + { + "epoch": 2.07, + "learning_rate": 1.5557194152346757e-05, + "loss": 0.2493, + "step": 53767 + }, + { + "epoch": 2.07, + "learning_rate": 1.5556552962298026e-05, + "loss": 0.2007, + "step": 53768 + }, + { + "epoch": 2.07, + "learning_rate": 1.5555911772249297e-05, + "loss": 0.0914, + "step": 53769 + }, + { + "epoch": 2.07, + "learning_rate": 1.5555270582200566e-05, + "loss": 0.2139, + "step": 53770 + }, + { + "epoch": 2.07, + "learning_rate": 1.5554629392151834e-05, + "loss": 0.2563, + "step": 53771 + }, + { + "epoch": 2.07, + "learning_rate": 1.5553988202103102e-05, + "loss": 0.3201, + "step": 53772 + }, + { + "epoch": 2.07, + "learning_rate": 1.5553347012054374e-05, + "loss": 0.2856, + "step": 53773 + }, + { + "epoch": 2.07, + "learning_rate": 1.5552705822005642e-05, + "loss": 0.2289, + "step": 53774 + }, + { + "epoch": 2.07, + "learning_rate": 1.5552064631956914e-05, + "loss": 0.3271, + "step": 53775 + }, + { + "epoch": 2.07, + "learning_rate": 1.5551423441908182e-05, + "loss": 0.2397, + "step": 53776 + }, + { + "epoch": 2.07, + "learning_rate": 1.5550782251859454e-05, + "loss": 0.2508, + "step": 53777 + }, + { + "epoch": 2.07, + "learning_rate": 1.5550141061810723e-05, + "loss": 0.1902, + "step": 53778 + }, + { + "epoch": 2.07, + "learning_rate": 1.554949987176199e-05, + "loss": 0.2136, + "step": 53779 + }, + { + "epoch": 2.07, + "learning_rate": 1.554885868171326e-05, + "loss": 0.2282, + "step": 53780 + }, + { + "epoch": 2.07, + "learning_rate": 1.554821749166453e-05, + "loss": 0.1758, + "step": 53781 + }, + { + "epoch": 2.07, + "learning_rate": 1.55475763016158e-05, + "loss": 0.3319, + "step": 53782 + }, + { + "epoch": 2.07, + "learning_rate": 1.5546935111567068e-05, + "loss": 0.1632, + "step": 53783 + }, + { + "epoch": 2.07, + "learning_rate": 1.554629392151834e-05, + "loss": 0.2355, + "step": 53784 + }, + { + "epoch": 2.07, + "learning_rate": 1.554565273146961e-05, + "loss": 0.196, + "step": 53785 + }, + { + "epoch": 2.07, + "learning_rate": 1.554501154142088e-05, + "loss": 0.2674, + "step": 53786 + }, + { + "epoch": 2.07, + "learning_rate": 1.5544370351372148e-05, + "loss": 0.268, + "step": 53787 + }, + { + "epoch": 2.07, + "learning_rate": 1.5543729161323416e-05, + "loss": 0.236, + "step": 53788 + }, + { + "epoch": 2.07, + "learning_rate": 1.5543087971274688e-05, + "loss": 0.3058, + "step": 53789 + }, + { + "epoch": 2.07, + "learning_rate": 1.5542446781225956e-05, + "loss": 0.4696, + "step": 53790 + }, + { + "epoch": 2.07, + "learning_rate": 1.5541805591177224e-05, + "loss": 0.3181, + "step": 53791 + }, + { + "epoch": 2.07, + "learning_rate": 1.5541164401128493e-05, + "loss": 0.2681, + "step": 53792 + }, + { + "epoch": 2.07, + "learning_rate": 1.5540523211079764e-05, + "loss": 0.2974, + "step": 53793 + }, + { + "epoch": 2.07, + "learning_rate": 1.5539882021031036e-05, + "loss": 0.2006, + "step": 53794 + }, + { + "epoch": 2.07, + "learning_rate": 1.5539240830982304e-05, + "loss": 0.1836, + "step": 53795 + }, + { + "epoch": 2.07, + "learning_rate": 1.5538599640933573e-05, + "loss": 0.2225, + "step": 53796 + }, + { + "epoch": 2.07, + "learning_rate": 1.5537958450884844e-05, + "loss": 0.236, + "step": 53797 + }, + { + "epoch": 2.07, + "learning_rate": 1.5537317260836113e-05, + "loss": 0.1402, + "step": 53798 + }, + { + "epoch": 2.07, + "learning_rate": 1.553667607078738e-05, + "loss": 0.2339, + "step": 53799 + }, + { + "epoch": 2.07, + "learning_rate": 1.553603488073865e-05, + "loss": 0.2104, + "step": 53800 + }, + { + "epoch": 2.07, + "learning_rate": 1.553539369068992e-05, + "loss": 0.2266, + "step": 53801 + }, + { + "epoch": 2.07, + "learning_rate": 1.5534752500641193e-05, + "loss": 0.2857, + "step": 53802 + }, + { + "epoch": 2.07, + "learning_rate": 1.553411131059246e-05, + "loss": 0.233, + "step": 53803 + }, + { + "epoch": 2.07, + "learning_rate": 1.553347012054373e-05, + "loss": 0.3462, + "step": 53804 + }, + { + "epoch": 2.07, + "learning_rate": 1.5532828930495e-05, + "loss": 0.3861, + "step": 53805 + }, + { + "epoch": 2.07, + "learning_rate": 1.553218774044627e-05, + "loss": 0.1457, + "step": 53806 + }, + { + "epoch": 2.07, + "learning_rate": 1.5531546550397538e-05, + "loss": 0.5398, + "step": 53807 + }, + { + "epoch": 2.07, + "learning_rate": 1.5530905360348806e-05, + "loss": 0.3294, + "step": 53808 + }, + { + "epoch": 2.07, + "learning_rate": 1.5530264170300078e-05, + "loss": 0.2774, + "step": 53809 + }, + { + "epoch": 2.07, + "learning_rate": 1.5529622980251346e-05, + "loss": 0.2184, + "step": 53810 + }, + { + "epoch": 2.07, + "learning_rate": 1.5528981790202618e-05, + "loss": 0.2472, + "step": 53811 + }, + { + "epoch": 2.07, + "learning_rate": 1.5528340600153886e-05, + "loss": 0.2398, + "step": 53812 + }, + { + "epoch": 2.07, + "learning_rate": 1.5527699410105158e-05, + "loss": 0.2, + "step": 53813 + }, + { + "epoch": 2.07, + "learning_rate": 1.5527058220056426e-05, + "loss": 0.2267, + "step": 53814 + }, + { + "epoch": 2.07, + "learning_rate": 1.5526417030007695e-05, + "loss": 0.1745, + "step": 53815 + }, + { + "epoch": 2.07, + "learning_rate": 1.5525775839958963e-05, + "loss": 0.2668, + "step": 53816 + }, + { + "epoch": 2.07, + "learning_rate": 1.5525134649910235e-05, + "loss": 0.1871, + "step": 53817 + }, + { + "epoch": 2.07, + "learning_rate": 1.5524493459861503e-05, + "loss": 0.2756, + "step": 53818 + }, + { + "epoch": 2.07, + "learning_rate": 1.552385226981277e-05, + "loss": 0.3585, + "step": 53819 + }, + { + "epoch": 2.07, + "learning_rate": 1.5523211079764043e-05, + "loss": 0.1405, + "step": 53820 + }, + { + "epoch": 2.07, + "learning_rate": 1.5522569889715315e-05, + "loss": 0.2409, + "step": 53821 + }, + { + "epoch": 2.07, + "learning_rate": 1.5521928699666583e-05, + "loss": 0.1389, + "step": 53822 + }, + { + "epoch": 2.07, + "learning_rate": 1.552128750961785e-05, + "loss": 0.2252, + "step": 53823 + }, + { + "epoch": 2.07, + "learning_rate": 1.552064631956912e-05, + "loss": 0.2909, + "step": 53824 + }, + { + "epoch": 2.07, + "learning_rate": 1.552000512952039e-05, + "loss": 0.2378, + "step": 53825 + }, + { + "epoch": 2.07, + "learning_rate": 1.551936393947166e-05, + "loss": 0.1393, + "step": 53826 + }, + { + "epoch": 2.07, + "learning_rate": 1.5518722749422928e-05, + "loss": 0.3264, + "step": 53827 + }, + { + "epoch": 2.07, + "learning_rate": 1.5518081559374197e-05, + "loss": 0.2197, + "step": 53828 + }, + { + "epoch": 2.07, + "learning_rate": 1.551744036932547e-05, + "loss": 0.2398, + "step": 53829 + }, + { + "epoch": 2.07, + "learning_rate": 1.551679917927674e-05, + "loss": 0.2967, + "step": 53830 + }, + { + "epoch": 2.07, + "learning_rate": 1.5516157989228008e-05, + "loss": 0.2583, + "step": 53831 + }, + { + "epoch": 2.07, + "learning_rate": 1.5515516799179277e-05, + "loss": 0.1761, + "step": 53832 + }, + { + "epoch": 2.07, + "learning_rate": 1.551487560913055e-05, + "loss": 0.2715, + "step": 53833 + }, + { + "epoch": 2.07, + "learning_rate": 1.5514234419081817e-05, + "loss": 0.1879, + "step": 53834 + }, + { + "epoch": 2.07, + "learning_rate": 1.5513593229033085e-05, + "loss": 0.3048, + "step": 53835 + }, + { + "epoch": 2.07, + "learning_rate": 1.5512952038984353e-05, + "loss": 0.2721, + "step": 53836 + }, + { + "epoch": 2.07, + "learning_rate": 1.5512310848935625e-05, + "loss": 0.172, + "step": 53837 + }, + { + "epoch": 2.07, + "learning_rate": 1.5511669658886897e-05, + "loss": 0.2151, + "step": 53838 + }, + { + "epoch": 2.07, + "learning_rate": 1.5511028468838165e-05, + "loss": 0.1623, + "step": 53839 + }, + { + "epoch": 2.07, + "learning_rate": 1.5510387278789433e-05, + "loss": 0.2571, + "step": 53840 + }, + { + "epoch": 2.07, + "learning_rate": 1.5509746088740705e-05, + "loss": 0.1539, + "step": 53841 + }, + { + "epoch": 2.07, + "learning_rate": 1.5509104898691973e-05, + "loss": 0.2355, + "step": 53842 + }, + { + "epoch": 2.07, + "learning_rate": 1.5508463708643242e-05, + "loss": 0.2243, + "step": 53843 + }, + { + "epoch": 2.07, + "learning_rate": 1.550782251859451e-05, + "loss": 0.3071, + "step": 53844 + }, + { + "epoch": 2.07, + "learning_rate": 1.5507181328545782e-05, + "loss": 0.3489, + "step": 53845 + }, + { + "epoch": 2.07, + "learning_rate": 1.550654013849705e-05, + "loss": 0.2095, + "step": 53846 + }, + { + "epoch": 2.07, + "learning_rate": 1.5505898948448322e-05, + "loss": 0.2879, + "step": 53847 + }, + { + "epoch": 2.07, + "learning_rate": 1.550525775839959e-05, + "loss": 0.3493, + "step": 53848 + }, + { + "epoch": 2.07, + "learning_rate": 1.5504616568350862e-05, + "loss": 0.2282, + "step": 53849 + }, + { + "epoch": 2.07, + "learning_rate": 1.550397537830213e-05, + "loss": 0.3227, + "step": 53850 + }, + { + "epoch": 2.07, + "learning_rate": 1.55033341882534e-05, + "loss": 0.3086, + "step": 53851 + }, + { + "epoch": 2.07, + "learning_rate": 1.5502692998204667e-05, + "loss": 0.2147, + "step": 53852 + }, + { + "epoch": 2.07, + "learning_rate": 1.550205180815594e-05, + "loss": 0.2762, + "step": 53853 + }, + { + "epoch": 2.07, + "learning_rate": 1.5501410618107207e-05, + "loss": 0.2526, + "step": 53854 + }, + { + "epoch": 2.07, + "learning_rate": 1.550076942805848e-05, + "loss": 0.2506, + "step": 53855 + }, + { + "epoch": 2.07, + "learning_rate": 1.5500128238009747e-05, + "loss": 0.1943, + "step": 53856 + }, + { + "epoch": 2.07, + "learning_rate": 1.549948704796102e-05, + "loss": 0.2279, + "step": 53857 + }, + { + "epoch": 2.07, + "learning_rate": 1.5498845857912287e-05, + "loss": 0.3094, + "step": 53858 + }, + { + "epoch": 2.07, + "learning_rate": 1.5498204667863555e-05, + "loss": 0.328, + "step": 53859 + }, + { + "epoch": 2.07, + "learning_rate": 1.5497563477814824e-05, + "loss": 0.2363, + "step": 53860 + }, + { + "epoch": 2.07, + "learning_rate": 1.5496922287766095e-05, + "loss": 0.2289, + "step": 53861 + }, + { + "epoch": 2.07, + "learning_rate": 1.5496281097717364e-05, + "loss": 0.2658, + "step": 53862 + }, + { + "epoch": 2.07, + "learning_rate": 1.5495639907668632e-05, + "loss": 0.2163, + "step": 53863 + }, + { + "epoch": 2.07, + "learning_rate": 1.5494998717619904e-05, + "loss": 0.1514, + "step": 53864 + }, + { + "epoch": 2.07, + "learning_rate": 1.5494357527571176e-05, + "loss": 0.0883, + "step": 53865 + }, + { + "epoch": 2.07, + "learning_rate": 1.5493716337522444e-05, + "loss": 0.2437, + "step": 53866 + }, + { + "epoch": 2.07, + "learning_rate": 1.5493075147473712e-05, + "loss": 0.3791, + "step": 53867 + }, + { + "epoch": 2.07, + "learning_rate": 1.549243395742498e-05, + "loss": 0.3429, + "step": 53868 + }, + { + "epoch": 2.07, + "learning_rate": 1.5491792767376252e-05, + "loss": 0.3093, + "step": 53869 + }, + { + "epoch": 2.07, + "learning_rate": 1.549115157732752e-05, + "loss": 0.2428, + "step": 53870 + }, + { + "epoch": 2.07, + "learning_rate": 1.549051038727879e-05, + "loss": 0.1912, + "step": 53871 + }, + { + "epoch": 2.07, + "learning_rate": 1.5489869197230057e-05, + "loss": 0.1336, + "step": 53872 + }, + { + "epoch": 2.07, + "learning_rate": 1.548922800718133e-05, + "loss": 0.2375, + "step": 53873 + }, + { + "epoch": 2.07, + "learning_rate": 1.54885868171326e-05, + "loss": 0.4155, + "step": 53874 + }, + { + "epoch": 2.07, + "learning_rate": 1.548794562708387e-05, + "loss": 0.2415, + "step": 53875 + }, + { + "epoch": 2.07, + "learning_rate": 1.5487304437035137e-05, + "loss": 0.1505, + "step": 53876 + }, + { + "epoch": 2.07, + "learning_rate": 1.548666324698641e-05, + "loss": 0.2, + "step": 53877 + }, + { + "epoch": 2.07, + "learning_rate": 1.5486022056937677e-05, + "loss": 0.216, + "step": 53878 + }, + { + "epoch": 2.07, + "learning_rate": 1.5485380866888946e-05, + "loss": 0.3207, + "step": 53879 + }, + { + "epoch": 2.07, + "learning_rate": 1.5484739676840214e-05, + "loss": 0.1989, + "step": 53880 + }, + { + "epoch": 2.07, + "learning_rate": 1.5484098486791486e-05, + "loss": 0.3159, + "step": 53881 + }, + { + "epoch": 2.07, + "learning_rate": 1.5483457296742757e-05, + "loss": 0.2804, + "step": 53882 + }, + { + "epoch": 2.07, + "learning_rate": 1.5482816106694026e-05, + "loss": 0.3177, + "step": 53883 + }, + { + "epoch": 2.07, + "learning_rate": 1.5482174916645294e-05, + "loss": 0.3476, + "step": 53884 + }, + { + "epoch": 2.07, + "learning_rate": 1.5481533726596566e-05, + "loss": 0.262, + "step": 53885 + }, + { + "epoch": 2.07, + "learning_rate": 1.5480892536547834e-05, + "loss": 0.1972, + "step": 53886 + }, + { + "epoch": 2.07, + "learning_rate": 1.5480251346499102e-05, + "loss": 0.2764, + "step": 53887 + }, + { + "epoch": 2.07, + "learning_rate": 1.547961015645037e-05, + "loss": 0.4184, + "step": 53888 + }, + { + "epoch": 2.07, + "learning_rate": 1.5478968966401643e-05, + "loss": 0.3388, + "step": 53889 + }, + { + "epoch": 2.07, + "learning_rate": 1.547832777635291e-05, + "loss": 0.2902, + "step": 53890 + }, + { + "epoch": 2.07, + "learning_rate": 1.5477686586304183e-05, + "loss": 0.2116, + "step": 53891 + }, + { + "epoch": 2.07, + "learning_rate": 1.547704539625545e-05, + "loss": 0.1078, + "step": 53892 + }, + { + "epoch": 2.07, + "learning_rate": 1.5476404206206723e-05, + "loss": 0.2097, + "step": 53893 + }, + { + "epoch": 2.07, + "learning_rate": 1.547576301615799e-05, + "loss": 0.1644, + "step": 53894 + }, + { + "epoch": 2.07, + "learning_rate": 1.547512182610926e-05, + "loss": 0.2161, + "step": 53895 + }, + { + "epoch": 2.07, + "learning_rate": 1.5474480636060528e-05, + "loss": 0.2102, + "step": 53896 + }, + { + "epoch": 2.07, + "learning_rate": 1.54738394460118e-05, + "loss": 0.2831, + "step": 53897 + }, + { + "epoch": 2.07, + "learning_rate": 1.5473198255963068e-05, + "loss": 0.2213, + "step": 53898 + }, + { + "epoch": 2.07, + "learning_rate": 1.5472557065914336e-05, + "loss": 0.2325, + "step": 53899 + }, + { + "epoch": 2.07, + "learning_rate": 1.5471915875865608e-05, + "loss": 0.2507, + "step": 53900 + }, + { + "epoch": 2.07, + "learning_rate": 1.547127468581688e-05, + "loss": 0.189, + "step": 53901 + }, + { + "epoch": 2.07, + "learning_rate": 1.5470633495768148e-05, + "loss": 0.1739, + "step": 53902 + }, + { + "epoch": 2.07, + "learning_rate": 1.5469992305719416e-05, + "loss": 0.2403, + "step": 53903 + }, + { + "epoch": 2.07, + "learning_rate": 1.5469351115670684e-05, + "loss": 0.2799, + "step": 53904 + }, + { + "epoch": 2.07, + "learning_rate": 1.5468709925621956e-05, + "loss": 0.2679, + "step": 53905 + }, + { + "epoch": 2.07, + "learning_rate": 1.5468068735573224e-05, + "loss": 0.1132, + "step": 53906 + }, + { + "epoch": 2.07, + "learning_rate": 1.5467427545524493e-05, + "loss": 0.1809, + "step": 53907 + }, + { + "epoch": 2.07, + "learning_rate": 1.546678635547576e-05, + "loss": 0.2062, + "step": 53908 + }, + { + "epoch": 2.07, + "learning_rate": 1.5466145165427036e-05, + "loss": 0.1525, + "step": 53909 + }, + { + "epoch": 2.07, + "learning_rate": 1.5465503975378305e-05, + "loss": 0.2182, + "step": 53910 + }, + { + "epoch": 2.07, + "learning_rate": 1.5464862785329573e-05, + "loss": 0.2946, + "step": 53911 + }, + { + "epoch": 2.07, + "learning_rate": 1.546422159528084e-05, + "loss": 0.2784, + "step": 53912 + }, + { + "epoch": 2.07, + "learning_rate": 1.5463580405232113e-05, + "loss": 0.3194, + "step": 53913 + }, + { + "epoch": 2.07, + "learning_rate": 1.546293921518338e-05, + "loss": 0.3394, + "step": 53914 + }, + { + "epoch": 2.07, + "learning_rate": 1.546229802513465e-05, + "loss": 0.1804, + "step": 53915 + }, + { + "epoch": 2.07, + "learning_rate": 1.5461656835085918e-05, + "loss": 0.185, + "step": 53916 + }, + { + "epoch": 2.07, + "learning_rate": 1.546101564503719e-05, + "loss": 0.4408, + "step": 53917 + }, + { + "epoch": 2.07, + "learning_rate": 1.546037445498846e-05, + "loss": 0.1004, + "step": 53918 + }, + { + "epoch": 2.07, + "learning_rate": 1.545973326493973e-05, + "loss": 0.283, + "step": 53919 + }, + { + "epoch": 2.07, + "learning_rate": 1.5459092074890998e-05, + "loss": 0.2453, + "step": 53920 + }, + { + "epoch": 2.07, + "learning_rate": 1.545845088484227e-05, + "loss": 0.1951, + "step": 53921 + }, + { + "epoch": 2.07, + "learning_rate": 1.5457809694793538e-05, + "loss": 0.277, + "step": 53922 + }, + { + "epoch": 2.07, + "learning_rate": 1.5457168504744806e-05, + "loss": 0.208, + "step": 53923 + }, + { + "epoch": 2.07, + "learning_rate": 1.5456527314696075e-05, + "loss": 0.3681, + "step": 53924 + }, + { + "epoch": 2.07, + "learning_rate": 1.5455886124647346e-05, + "loss": 0.1552, + "step": 53925 + }, + { + "epoch": 2.07, + "learning_rate": 1.5455244934598615e-05, + "loss": 0.2815, + "step": 53926 + }, + { + "epoch": 2.07, + "learning_rate": 1.5454603744549886e-05, + "loss": 0.2222, + "step": 53927 + }, + { + "epoch": 2.07, + "learning_rate": 1.5453962554501155e-05, + "loss": 0.1787, + "step": 53928 + }, + { + "epoch": 2.07, + "learning_rate": 1.5453321364452426e-05, + "loss": 0.2962, + "step": 53929 + }, + { + "epoch": 2.07, + "learning_rate": 1.5452680174403695e-05, + "loss": 0.2292, + "step": 53930 + }, + { + "epoch": 2.07, + "learning_rate": 1.5452038984354963e-05, + "loss": 0.2244, + "step": 53931 + }, + { + "epoch": 2.07, + "learning_rate": 1.545139779430623e-05, + "loss": 0.193, + "step": 53932 + }, + { + "epoch": 2.07, + "learning_rate": 1.5450756604257503e-05, + "loss": 0.3315, + "step": 53933 + }, + { + "epoch": 2.07, + "learning_rate": 1.545011541420877e-05, + "loss": 0.2696, + "step": 53934 + }, + { + "epoch": 2.07, + "learning_rate": 1.544947422416004e-05, + "loss": 0.226, + "step": 53935 + }, + { + "epoch": 2.07, + "learning_rate": 1.544883303411131e-05, + "loss": 0.1457, + "step": 53936 + }, + { + "epoch": 2.07, + "learning_rate": 1.5448191844062583e-05, + "loss": 0.1649, + "step": 53937 + }, + { + "epoch": 2.07, + "learning_rate": 1.544755065401385e-05, + "loss": 0.2935, + "step": 53938 + }, + { + "epoch": 2.07, + "learning_rate": 1.544690946396512e-05, + "loss": 0.2783, + "step": 53939 + }, + { + "epoch": 2.07, + "learning_rate": 1.5446268273916388e-05, + "loss": 0.2149, + "step": 53940 + }, + { + "epoch": 2.07, + "learning_rate": 1.544562708386766e-05, + "loss": 0.2677, + "step": 53941 + }, + { + "epoch": 2.07, + "learning_rate": 1.5444985893818928e-05, + "loss": 0.1296, + "step": 53942 + }, + { + "epoch": 2.07, + "learning_rate": 1.5444344703770197e-05, + "loss": 0.3021, + "step": 53943 + }, + { + "epoch": 2.07, + "learning_rate": 1.5443703513721465e-05, + "loss": 0.2803, + "step": 53944 + }, + { + "epoch": 2.07, + "learning_rate": 1.544306232367274e-05, + "loss": 0.1503, + "step": 53945 + }, + { + "epoch": 2.07, + "learning_rate": 1.544242113362401e-05, + "loss": 0.2152, + "step": 53946 + }, + { + "epoch": 2.07, + "learning_rate": 1.5441779943575277e-05, + "loss": 0.3287, + "step": 53947 + }, + { + "epoch": 2.07, + "learning_rate": 1.5441138753526545e-05, + "loss": 0.3245, + "step": 53948 + }, + { + "epoch": 2.07, + "learning_rate": 1.5440497563477817e-05, + "loss": 0.2779, + "step": 53949 + }, + { + "epoch": 2.07, + "learning_rate": 1.5439856373429085e-05, + "loss": 0.3496, + "step": 53950 + }, + { + "epoch": 2.07, + "learning_rate": 1.5439215183380353e-05, + "loss": 0.2816, + "step": 53951 + }, + { + "epoch": 2.07, + "learning_rate": 1.5438573993331622e-05, + "loss": 0.4265, + "step": 53952 + }, + { + "epoch": 2.07, + "learning_rate": 1.5437932803282893e-05, + "loss": 0.246, + "step": 53953 + }, + { + "epoch": 2.07, + "learning_rate": 1.5437291613234165e-05, + "loss": 0.2748, + "step": 53954 + }, + { + "epoch": 2.07, + "learning_rate": 1.5436650423185434e-05, + "loss": 0.1901, + "step": 53955 + }, + { + "epoch": 2.07, + "learning_rate": 1.5436009233136702e-05, + "loss": 0.301, + "step": 53956 + }, + { + "epoch": 2.07, + "learning_rate": 1.5435368043087974e-05, + "loss": 0.2573, + "step": 53957 + }, + { + "epoch": 2.07, + "learning_rate": 1.5434726853039242e-05, + "loss": 0.167, + "step": 53958 + }, + { + "epoch": 2.07, + "learning_rate": 1.543408566299051e-05, + "loss": 0.214, + "step": 53959 + }, + { + "epoch": 2.07, + "learning_rate": 1.543344447294178e-05, + "loss": 0.2558, + "step": 53960 + }, + { + "epoch": 2.07, + "learning_rate": 1.543280328289305e-05, + "loss": 0.2326, + "step": 53961 + }, + { + "epoch": 2.07, + "learning_rate": 1.5432162092844322e-05, + "loss": 0.2072, + "step": 53962 + }, + { + "epoch": 2.07, + "learning_rate": 1.543152090279559e-05, + "loss": 0.2885, + "step": 53963 + }, + { + "epoch": 2.07, + "learning_rate": 1.543087971274686e-05, + "loss": 0.4158, + "step": 53964 + }, + { + "epoch": 2.07, + "learning_rate": 1.543023852269813e-05, + "loss": 0.2806, + "step": 53965 + }, + { + "epoch": 2.07, + "learning_rate": 1.54295973326494e-05, + "loss": 0.3783, + "step": 53966 + }, + { + "epoch": 2.07, + "learning_rate": 1.5428956142600667e-05, + "loss": 0.2531, + "step": 53967 + }, + { + "epoch": 2.07, + "learning_rate": 1.5428314952551935e-05, + "loss": 0.1734, + "step": 53968 + }, + { + "epoch": 2.07, + "learning_rate": 1.5427673762503207e-05, + "loss": 0.1174, + "step": 53969 + }, + { + "epoch": 2.07, + "learning_rate": 1.5427032572454475e-05, + "loss": 0.1169, + "step": 53970 + }, + { + "epoch": 2.08, + "learning_rate": 1.5426391382405747e-05, + "loss": 0.1698, + "step": 53971 + }, + { + "epoch": 2.08, + "learning_rate": 1.5425750192357015e-05, + "loss": 0.189, + "step": 53972 + }, + { + "epoch": 2.08, + "learning_rate": 1.5425109002308287e-05, + "loss": 0.1669, + "step": 53973 + }, + { + "epoch": 2.08, + "learning_rate": 1.5424467812259555e-05, + "loss": 0.1333, + "step": 53974 + }, + { + "epoch": 2.08, + "learning_rate": 1.5423826622210824e-05, + "loss": 0.2248, + "step": 53975 + }, + { + "epoch": 2.08, + "learning_rate": 1.5423185432162092e-05, + "loss": 0.3273, + "step": 53976 + }, + { + "epoch": 2.08, + "learning_rate": 1.5422544242113364e-05, + "loss": 0.1768, + "step": 53977 + }, + { + "epoch": 2.08, + "learning_rate": 1.5421903052064632e-05, + "loss": 0.1307, + "step": 53978 + }, + { + "epoch": 2.08, + "learning_rate": 1.54212618620159e-05, + "loss": 0.2198, + "step": 53979 + }, + { + "epoch": 2.08, + "learning_rate": 1.5420620671967172e-05, + "loss": 0.2925, + "step": 53980 + }, + { + "epoch": 2.08, + "learning_rate": 1.5419979481918444e-05, + "loss": 0.1523, + "step": 53981 + }, + { + "epoch": 2.08, + "learning_rate": 1.5419338291869712e-05, + "loss": 0.3156, + "step": 53982 + }, + { + "epoch": 2.08, + "learning_rate": 1.541869710182098e-05, + "loss": 0.3379, + "step": 53983 + }, + { + "epoch": 2.08, + "learning_rate": 1.541805591177225e-05, + "loss": 0.2629, + "step": 53984 + }, + { + "epoch": 2.08, + "learning_rate": 1.541741472172352e-05, + "loss": 0.2347, + "step": 53985 + }, + { + "epoch": 2.08, + "learning_rate": 1.541677353167479e-05, + "loss": 0.2777, + "step": 53986 + }, + { + "epoch": 2.08, + "learning_rate": 1.5416132341626057e-05, + "loss": 0.1162, + "step": 53987 + }, + { + "epoch": 2.08, + "learning_rate": 1.5415491151577326e-05, + "loss": 0.1768, + "step": 53988 + }, + { + "epoch": 2.08, + "learning_rate": 1.54148499615286e-05, + "loss": 0.2045, + "step": 53989 + }, + { + "epoch": 2.08, + "learning_rate": 1.541420877147987e-05, + "loss": 0.174, + "step": 53990 + }, + { + "epoch": 2.08, + "learning_rate": 1.5413567581431137e-05, + "loss": 0.2947, + "step": 53991 + }, + { + "epoch": 2.08, + "learning_rate": 1.5412926391382406e-05, + "loss": 0.1887, + "step": 53992 + }, + { + "epoch": 2.08, + "learning_rate": 1.5412285201333677e-05, + "loss": 0.2314, + "step": 53993 + }, + { + "epoch": 2.08, + "learning_rate": 1.5411644011284946e-05, + "loss": 0.2934, + "step": 53994 + }, + { + "epoch": 2.08, + "learning_rate": 1.5411002821236214e-05, + "loss": 0.2461, + "step": 53995 + }, + { + "epoch": 2.08, + "learning_rate": 1.5410361631187482e-05, + "loss": 0.263, + "step": 53996 + }, + { + "epoch": 2.08, + "learning_rate": 1.5409720441138754e-05, + "loss": 0.2821, + "step": 53997 + }, + { + "epoch": 2.08, + "learning_rate": 1.5409079251090026e-05, + "loss": 0.2224, + "step": 53998 + }, + { + "epoch": 2.08, + "learning_rate": 1.5408438061041294e-05, + "loss": 0.2066, + "step": 53999 + }, + { + "epoch": 2.08, + "learning_rate": 1.5407796870992562e-05, + "loss": 0.3042, + "step": 54000 + }, + { + "epoch": 2.08, + "learning_rate": 1.5407155680943834e-05, + "loss": 0.307, + "step": 54001 + }, + { + "epoch": 2.08, + "learning_rate": 1.5406514490895103e-05, + "loss": 0.1857, + "step": 54002 + }, + { + "epoch": 2.08, + "learning_rate": 1.540587330084637e-05, + "loss": 0.2102, + "step": 54003 + }, + { + "epoch": 2.08, + "learning_rate": 1.540523211079764e-05, + "loss": 0.1816, + "step": 54004 + }, + { + "epoch": 2.08, + "learning_rate": 1.540459092074891e-05, + "loss": 0.2236, + "step": 54005 + }, + { + "epoch": 2.08, + "learning_rate": 1.540394973070018e-05, + "loss": 0.179, + "step": 54006 + }, + { + "epoch": 2.08, + "learning_rate": 1.540330854065145e-05, + "loss": 0.2042, + "step": 54007 + }, + { + "epoch": 2.08, + "learning_rate": 1.540266735060272e-05, + "loss": 0.2061, + "step": 54008 + }, + { + "epoch": 2.08, + "learning_rate": 1.540202616055399e-05, + "loss": 0.2593, + "step": 54009 + }, + { + "epoch": 2.08, + "learning_rate": 1.540138497050526e-05, + "loss": 0.2143, + "step": 54010 + }, + { + "epoch": 2.08, + "learning_rate": 1.5400743780456528e-05, + "loss": 0.241, + "step": 54011 + }, + { + "epoch": 2.08, + "learning_rate": 1.5400102590407796e-05, + "loss": 0.2022, + "step": 54012 + }, + { + "epoch": 2.08, + "learning_rate": 1.5399461400359068e-05, + "loss": 0.16, + "step": 54013 + }, + { + "epoch": 2.08, + "learning_rate": 1.5398820210310336e-05, + "loss": 0.1597, + "step": 54014 + }, + { + "epoch": 2.08, + "learning_rate": 1.5398179020261604e-05, + "loss": 0.3193, + "step": 54015 + }, + { + "epoch": 2.08, + "learning_rate": 1.5397537830212876e-05, + "loss": 0.3061, + "step": 54016 + }, + { + "epoch": 2.08, + "learning_rate": 1.5396896640164148e-05, + "loss": 0.1625, + "step": 54017 + }, + { + "epoch": 2.08, + "learning_rate": 1.5396255450115416e-05, + "loss": 0.1718, + "step": 54018 + }, + { + "epoch": 2.08, + "learning_rate": 1.5395614260066684e-05, + "loss": 0.2544, + "step": 54019 + }, + { + "epoch": 2.08, + "learning_rate": 1.5394973070017953e-05, + "loss": 0.2501, + "step": 54020 + }, + { + "epoch": 2.08, + "learning_rate": 1.5394331879969224e-05, + "loss": 0.2636, + "step": 54021 + }, + { + "epoch": 2.08, + "learning_rate": 1.5393690689920493e-05, + "loss": 0.2222, + "step": 54022 + }, + { + "epoch": 2.08, + "learning_rate": 1.539304949987176e-05, + "loss": 0.0656, + "step": 54023 + }, + { + "epoch": 2.08, + "learning_rate": 1.539240830982303e-05, + "loss": 0.2605, + "step": 54024 + }, + { + "epoch": 2.08, + "learning_rate": 1.5391767119774305e-05, + "loss": 0.2741, + "step": 54025 + }, + { + "epoch": 2.08, + "learning_rate": 1.5391125929725573e-05, + "loss": 0.3161, + "step": 54026 + }, + { + "epoch": 2.08, + "learning_rate": 1.539048473967684e-05, + "loss": 0.2393, + "step": 54027 + }, + { + "epoch": 2.08, + "learning_rate": 1.538984354962811e-05, + "loss": 0.2671, + "step": 54028 + }, + { + "epoch": 2.08, + "learning_rate": 1.538920235957938e-05, + "loss": 0.2522, + "step": 54029 + }, + { + "epoch": 2.08, + "learning_rate": 1.538856116953065e-05, + "loss": 0.181, + "step": 54030 + }, + { + "epoch": 2.08, + "learning_rate": 1.5387919979481918e-05, + "loss": 0.2493, + "step": 54031 + }, + { + "epoch": 2.08, + "learning_rate": 1.5387278789433186e-05, + "loss": 0.1723, + "step": 54032 + }, + { + "epoch": 2.08, + "learning_rate": 1.5386637599384458e-05, + "loss": 0.332, + "step": 54033 + }, + { + "epoch": 2.08, + "learning_rate": 1.538599640933573e-05, + "loss": 0.3149, + "step": 54034 + }, + { + "epoch": 2.08, + "learning_rate": 1.5385355219286998e-05, + "loss": 0.2712, + "step": 54035 + }, + { + "epoch": 2.08, + "learning_rate": 1.5384714029238266e-05, + "loss": 0.1848, + "step": 54036 + }, + { + "epoch": 2.08, + "learning_rate": 1.5384072839189538e-05, + "loss": 0.2249, + "step": 54037 + }, + { + "epoch": 2.08, + "learning_rate": 1.5383431649140806e-05, + "loss": 0.1314, + "step": 54038 + }, + { + "epoch": 2.08, + "learning_rate": 1.5382790459092075e-05, + "loss": 0.3175, + "step": 54039 + }, + { + "epoch": 2.08, + "learning_rate": 1.5382149269043343e-05, + "loss": 0.3293, + "step": 54040 + }, + { + "epoch": 2.08, + "learning_rate": 1.5381508078994615e-05, + "loss": 0.168, + "step": 54041 + }, + { + "epoch": 2.08, + "learning_rate": 1.5380866888945887e-05, + "loss": 0.2804, + "step": 54042 + }, + { + "epoch": 2.08, + "learning_rate": 1.5380225698897155e-05, + "loss": 0.2314, + "step": 54043 + }, + { + "epoch": 2.08, + "learning_rate": 1.5379584508848423e-05, + "loss": 0.2543, + "step": 54044 + }, + { + "epoch": 2.08, + "learning_rate": 1.5378943318799695e-05, + "loss": 0.2805, + "step": 54045 + }, + { + "epoch": 2.08, + "learning_rate": 1.5378302128750963e-05, + "loss": 0.1695, + "step": 54046 + }, + { + "epoch": 2.08, + "learning_rate": 1.537766093870223e-05, + "loss": 0.1877, + "step": 54047 + }, + { + "epoch": 2.08, + "learning_rate": 1.53770197486535e-05, + "loss": 0.2788, + "step": 54048 + }, + { + "epoch": 2.08, + "learning_rate": 1.537637855860477e-05, + "loss": 0.2339, + "step": 54049 + }, + { + "epoch": 2.08, + "learning_rate": 1.537573736855604e-05, + "loss": 0.1745, + "step": 54050 + }, + { + "epoch": 2.08, + "learning_rate": 1.537509617850731e-05, + "loss": 0.1894, + "step": 54051 + }, + { + "epoch": 2.08, + "learning_rate": 1.537445498845858e-05, + "loss": 0.2286, + "step": 54052 + }, + { + "epoch": 2.08, + "learning_rate": 1.537381379840985e-05, + "loss": 0.2354, + "step": 54053 + }, + { + "epoch": 2.08, + "learning_rate": 1.537317260836112e-05, + "loss": 0.3054, + "step": 54054 + }, + { + "epoch": 2.08, + "learning_rate": 1.537253141831239e-05, + "loss": 0.3501, + "step": 54055 + }, + { + "epoch": 2.08, + "learning_rate": 1.5371890228263657e-05, + "loss": 0.246, + "step": 54056 + }, + { + "epoch": 2.08, + "learning_rate": 1.537124903821493e-05, + "loss": 0.3162, + "step": 54057 + }, + { + "epoch": 2.08, + "learning_rate": 1.5370607848166197e-05, + "loss": 0.1002, + "step": 54058 + }, + { + "epoch": 2.08, + "learning_rate": 1.5369966658117465e-05, + "loss": 0.2804, + "step": 54059 + }, + { + "epoch": 2.08, + "learning_rate": 1.5369325468068737e-05, + "loss": 0.2898, + "step": 54060 + }, + { + "epoch": 2.08, + "learning_rate": 1.536868427802001e-05, + "loss": 0.2637, + "step": 54061 + }, + { + "epoch": 2.08, + "learning_rate": 1.5368043087971277e-05, + "loss": 0.2393, + "step": 54062 + }, + { + "epoch": 2.08, + "learning_rate": 1.5367401897922545e-05, + "loss": 0.2673, + "step": 54063 + }, + { + "epoch": 2.08, + "learning_rate": 1.5366760707873813e-05, + "loss": 0.2045, + "step": 54064 + }, + { + "epoch": 2.08, + "learning_rate": 1.5366119517825085e-05, + "loss": 0.2312, + "step": 54065 + }, + { + "epoch": 2.08, + "learning_rate": 1.5365478327776353e-05, + "loss": 0.198, + "step": 54066 + }, + { + "epoch": 2.08, + "learning_rate": 1.5364837137727622e-05, + "loss": 0.268, + "step": 54067 + }, + { + "epoch": 2.08, + "learning_rate": 1.536419594767889e-05, + "loss": 0.2074, + "step": 54068 + }, + { + "epoch": 2.08, + "learning_rate": 1.5363554757630165e-05, + "loss": 0.2665, + "step": 54069 + }, + { + "epoch": 2.08, + "learning_rate": 1.5362913567581434e-05, + "loss": 0.2081, + "step": 54070 + }, + { + "epoch": 2.08, + "learning_rate": 1.5362272377532702e-05, + "loss": 0.3031, + "step": 54071 + }, + { + "epoch": 2.08, + "learning_rate": 1.536163118748397e-05, + "loss": 0.2693, + "step": 54072 + }, + { + "epoch": 2.08, + "learning_rate": 1.5360989997435242e-05, + "loss": 0.3164, + "step": 54073 + }, + { + "epoch": 2.08, + "learning_rate": 1.536034880738651e-05, + "loss": 0.2848, + "step": 54074 + }, + { + "epoch": 2.08, + "learning_rate": 1.535970761733778e-05, + "loss": 0.3022, + "step": 54075 + }, + { + "epoch": 2.08, + "learning_rate": 1.5359066427289047e-05, + "loss": 0.1267, + "step": 54076 + }, + { + "epoch": 2.08, + "learning_rate": 1.535842523724032e-05, + "loss": 0.2496, + "step": 54077 + }, + { + "epoch": 2.08, + "learning_rate": 1.535778404719159e-05, + "loss": 0.1148, + "step": 54078 + }, + { + "epoch": 2.08, + "learning_rate": 1.535714285714286e-05, + "loss": 0.2847, + "step": 54079 + }, + { + "epoch": 2.08, + "learning_rate": 1.5356501667094127e-05, + "loss": 0.1775, + "step": 54080 + }, + { + "epoch": 2.08, + "learning_rate": 1.53558604770454e-05, + "loss": 0.1642, + "step": 54081 + }, + { + "epoch": 2.08, + "learning_rate": 1.5355219286996667e-05, + "loss": 0.194, + "step": 54082 + }, + { + "epoch": 2.08, + "learning_rate": 1.5354578096947935e-05, + "loss": 0.2636, + "step": 54083 + }, + { + "epoch": 2.08, + "learning_rate": 1.5353936906899204e-05, + "loss": 0.1793, + "step": 54084 + }, + { + "epoch": 2.08, + "learning_rate": 1.5353295716850475e-05, + "loss": 0.1778, + "step": 54085 + }, + { + "epoch": 2.08, + "learning_rate": 1.5352654526801744e-05, + "loss": 0.2586, + "step": 54086 + }, + { + "epoch": 2.08, + "learning_rate": 1.5352013336753015e-05, + "loss": 0.2434, + "step": 54087 + }, + { + "epoch": 2.08, + "learning_rate": 1.5351372146704284e-05, + "loss": 0.2133, + "step": 54088 + }, + { + "epoch": 2.08, + "learning_rate": 1.5350730956655556e-05, + "loss": 0.1935, + "step": 54089 + }, + { + "epoch": 2.08, + "learning_rate": 1.5350089766606824e-05, + "loss": 0.293, + "step": 54090 + }, + { + "epoch": 2.08, + "learning_rate": 1.5349448576558092e-05, + "loss": 0.2048, + "step": 54091 + }, + { + "epoch": 2.08, + "learning_rate": 1.534880738650936e-05, + "loss": 0.135, + "step": 54092 + }, + { + "epoch": 2.08, + "learning_rate": 1.5348166196460632e-05, + "loss": 0.204, + "step": 54093 + }, + { + "epoch": 2.08, + "learning_rate": 1.53475250064119e-05, + "loss": 0.2802, + "step": 54094 + }, + { + "epoch": 2.08, + "learning_rate": 1.534688381636317e-05, + "loss": 0.2347, + "step": 54095 + }, + { + "epoch": 2.08, + "learning_rate": 1.534624262631444e-05, + "loss": 0.2046, + "step": 54096 + }, + { + "epoch": 2.08, + "learning_rate": 1.5345601436265712e-05, + "loss": 0.2258, + "step": 54097 + }, + { + "epoch": 2.08, + "learning_rate": 1.534496024621698e-05, + "loss": 0.1266, + "step": 54098 + }, + { + "epoch": 2.08, + "learning_rate": 1.534431905616825e-05, + "loss": 0.3576, + "step": 54099 + }, + { + "epoch": 2.08, + "learning_rate": 1.5343677866119517e-05, + "loss": 0.1515, + "step": 54100 + }, + { + "epoch": 2.08, + "learning_rate": 1.534303667607079e-05, + "loss": 0.2518, + "step": 54101 + }, + { + "epoch": 2.08, + "learning_rate": 1.5342395486022057e-05, + "loss": 0.21, + "step": 54102 + }, + { + "epoch": 2.08, + "learning_rate": 1.5341754295973326e-05, + "loss": 0.3226, + "step": 54103 + }, + { + "epoch": 2.08, + "learning_rate": 1.5341113105924594e-05, + "loss": 0.2015, + "step": 54104 + }, + { + "epoch": 2.08, + "learning_rate": 1.534047191587587e-05, + "loss": 0.2249, + "step": 54105 + }, + { + "epoch": 2.08, + "learning_rate": 1.5339830725827137e-05, + "loss": 0.2551, + "step": 54106 + }, + { + "epoch": 2.08, + "learning_rate": 1.5339189535778406e-05, + "loss": 0.1518, + "step": 54107 + }, + { + "epoch": 2.08, + "learning_rate": 1.5338548345729674e-05, + "loss": 0.1432, + "step": 54108 + }, + { + "epoch": 2.08, + "learning_rate": 1.5337907155680946e-05, + "loss": 0.3888, + "step": 54109 + }, + { + "epoch": 2.08, + "learning_rate": 1.5337265965632214e-05, + "loss": 0.2209, + "step": 54110 + }, + { + "epoch": 2.08, + "learning_rate": 1.5336624775583482e-05, + "loss": 0.0945, + "step": 54111 + }, + { + "epoch": 2.08, + "learning_rate": 1.533598358553475e-05, + "loss": 0.2622, + "step": 54112 + }, + { + "epoch": 2.08, + "learning_rate": 1.5335342395486023e-05, + "loss": 0.2461, + "step": 54113 + }, + { + "epoch": 2.08, + "learning_rate": 1.5334701205437294e-05, + "loss": 0.2668, + "step": 54114 + }, + { + "epoch": 2.08, + "learning_rate": 1.5334060015388563e-05, + "loss": 0.1754, + "step": 54115 + }, + { + "epoch": 2.08, + "learning_rate": 1.533341882533983e-05, + "loss": 0.2519, + "step": 54116 + }, + { + "epoch": 2.08, + "learning_rate": 1.5332777635291103e-05, + "loss": 0.1968, + "step": 54117 + }, + { + "epoch": 2.08, + "learning_rate": 1.533213644524237e-05, + "loss": 0.2572, + "step": 54118 + }, + { + "epoch": 2.08, + "learning_rate": 1.533149525519364e-05, + "loss": 0.2273, + "step": 54119 + }, + { + "epoch": 2.08, + "learning_rate": 1.5330854065144908e-05, + "loss": 0.3789, + "step": 54120 + }, + { + "epoch": 2.08, + "learning_rate": 1.533021287509618e-05, + "loss": 0.2994, + "step": 54121 + }, + { + "epoch": 2.08, + "learning_rate": 1.5329571685047448e-05, + "loss": 0.3055, + "step": 54122 + }, + { + "epoch": 2.08, + "learning_rate": 1.532893049499872e-05, + "loss": 0.2327, + "step": 54123 + }, + { + "epoch": 2.08, + "learning_rate": 1.5328289304949988e-05, + "loss": 0.1873, + "step": 54124 + }, + { + "epoch": 2.08, + "learning_rate": 1.532764811490126e-05, + "loss": 0.3951, + "step": 54125 + }, + { + "epoch": 2.08, + "learning_rate": 1.5327006924852528e-05, + "loss": 0.2674, + "step": 54126 + }, + { + "epoch": 2.08, + "learning_rate": 1.5326365734803796e-05, + "loss": 0.259, + "step": 54127 + }, + { + "epoch": 2.08, + "learning_rate": 1.5325724544755064e-05, + "loss": 0.3128, + "step": 54128 + }, + { + "epoch": 2.08, + "learning_rate": 1.5325083354706336e-05, + "loss": 0.2674, + "step": 54129 + }, + { + "epoch": 2.08, + "learning_rate": 1.5324442164657604e-05, + "loss": 0.236, + "step": 54130 + }, + { + "epoch": 2.08, + "learning_rate": 1.5323800974608873e-05, + "loss": 0.2558, + "step": 54131 + }, + { + "epoch": 2.08, + "learning_rate": 1.5323159784560144e-05, + "loss": 0.2986, + "step": 54132 + }, + { + "epoch": 2.08, + "learning_rate": 1.5322518594511416e-05, + "loss": 0.2893, + "step": 54133 + }, + { + "epoch": 2.08, + "learning_rate": 1.5321877404462685e-05, + "loss": 0.2662, + "step": 54134 + }, + { + "epoch": 2.08, + "learning_rate": 1.5321236214413953e-05, + "loss": 0.1993, + "step": 54135 + }, + { + "epoch": 2.08, + "learning_rate": 1.532059502436522e-05, + "loss": 0.3664, + "step": 54136 + }, + { + "epoch": 2.08, + "learning_rate": 1.5319953834316493e-05, + "loss": 0.187, + "step": 54137 + }, + { + "epoch": 2.08, + "learning_rate": 1.531931264426776e-05, + "loss": 0.1221, + "step": 54138 + }, + { + "epoch": 2.08, + "learning_rate": 1.531867145421903e-05, + "loss": 0.1692, + "step": 54139 + }, + { + "epoch": 2.08, + "learning_rate": 1.53180302641703e-05, + "loss": 0.2421, + "step": 54140 + }, + { + "epoch": 2.08, + "learning_rate": 1.5317389074121573e-05, + "loss": 0.325, + "step": 54141 + }, + { + "epoch": 2.08, + "learning_rate": 1.531674788407284e-05, + "loss": 0.1998, + "step": 54142 + }, + { + "epoch": 2.08, + "learning_rate": 1.531610669402411e-05, + "loss": 0.167, + "step": 54143 + }, + { + "epoch": 2.08, + "learning_rate": 1.5315465503975378e-05, + "loss": 0.2672, + "step": 54144 + }, + { + "epoch": 2.08, + "learning_rate": 1.531482431392665e-05, + "loss": 0.192, + "step": 54145 + }, + { + "epoch": 2.08, + "learning_rate": 1.5314183123877918e-05, + "loss": 0.2948, + "step": 54146 + }, + { + "epoch": 2.08, + "learning_rate": 1.5313541933829186e-05, + "loss": 0.2322, + "step": 54147 + }, + { + "epoch": 2.08, + "learning_rate": 1.5312900743780455e-05, + "loss": 0.1668, + "step": 54148 + }, + { + "epoch": 2.08, + "learning_rate": 1.531225955373173e-05, + "loss": 0.3058, + "step": 54149 + }, + { + "epoch": 2.08, + "learning_rate": 1.5311618363682998e-05, + "loss": 0.3062, + "step": 54150 + }, + { + "epoch": 2.08, + "learning_rate": 1.5310977173634266e-05, + "loss": 0.2101, + "step": 54151 + }, + { + "epoch": 2.08, + "learning_rate": 1.5310335983585535e-05, + "loss": 0.2154, + "step": 54152 + }, + { + "epoch": 2.08, + "learning_rate": 1.5309694793536806e-05, + "loss": 0.2493, + "step": 54153 + }, + { + "epoch": 2.08, + "learning_rate": 1.5309053603488075e-05, + "loss": 0.3117, + "step": 54154 + }, + { + "epoch": 2.08, + "learning_rate": 1.5308412413439343e-05, + "loss": 0.2207, + "step": 54155 + }, + { + "epoch": 2.08, + "learning_rate": 1.530777122339061e-05, + "loss": 0.1826, + "step": 54156 + }, + { + "epoch": 2.08, + "learning_rate": 1.5307130033341883e-05, + "loss": 0.2465, + "step": 54157 + }, + { + "epoch": 2.08, + "learning_rate": 1.5306488843293155e-05, + "loss": 0.2122, + "step": 54158 + }, + { + "epoch": 2.08, + "learning_rate": 1.5305847653244423e-05, + "loss": 0.1995, + "step": 54159 + }, + { + "epoch": 2.08, + "learning_rate": 1.530520646319569e-05, + "loss": 0.1331, + "step": 54160 + }, + { + "epoch": 2.08, + "learning_rate": 1.5304565273146963e-05, + "loss": 0.2103, + "step": 54161 + }, + { + "epoch": 2.08, + "learning_rate": 1.530392408309823e-05, + "loss": 0.2678, + "step": 54162 + }, + { + "epoch": 2.08, + "learning_rate": 1.53032828930495e-05, + "loss": 0.1398, + "step": 54163 + }, + { + "epoch": 2.08, + "learning_rate": 1.5302641703000768e-05, + "loss": 0.3364, + "step": 54164 + }, + { + "epoch": 2.08, + "learning_rate": 1.530200051295204e-05, + "loss": 0.2859, + "step": 54165 + }, + { + "epoch": 2.08, + "learning_rate": 1.530135932290331e-05, + "loss": 0.2172, + "step": 54166 + }, + { + "epoch": 2.08, + "learning_rate": 1.530071813285458e-05, + "loss": 0.1248, + "step": 54167 + }, + { + "epoch": 2.08, + "learning_rate": 1.530007694280585e-05, + "loss": 0.2256, + "step": 54168 + }, + { + "epoch": 2.08, + "learning_rate": 1.529943575275712e-05, + "loss": 0.3075, + "step": 54169 + }, + { + "epoch": 2.08, + "learning_rate": 1.529879456270839e-05, + "loss": 0.2068, + "step": 54170 + }, + { + "epoch": 2.08, + "learning_rate": 1.5298153372659657e-05, + "loss": 0.2451, + "step": 54171 + }, + { + "epoch": 2.08, + "learning_rate": 1.5297512182610925e-05, + "loss": 0.3257, + "step": 54172 + }, + { + "epoch": 2.08, + "learning_rate": 1.5296870992562197e-05, + "loss": 0.1834, + "step": 54173 + }, + { + "epoch": 2.08, + "learning_rate": 1.5296229802513465e-05, + "loss": 0.2134, + "step": 54174 + }, + { + "epoch": 2.08, + "learning_rate": 1.5295588612464733e-05, + "loss": 0.2599, + "step": 54175 + }, + { + "epoch": 2.08, + "learning_rate": 1.5294947422416005e-05, + "loss": 0.2321, + "step": 54176 + }, + { + "epoch": 2.08, + "learning_rate": 1.5294306232367277e-05, + "loss": 0.2357, + "step": 54177 + }, + { + "epoch": 2.08, + "learning_rate": 1.5293665042318545e-05, + "loss": 0.2587, + "step": 54178 + }, + { + "epoch": 2.08, + "learning_rate": 1.5293023852269814e-05, + "loss": 0.1973, + "step": 54179 + }, + { + "epoch": 2.08, + "learning_rate": 1.5292382662221082e-05, + "loss": 0.414, + "step": 54180 + }, + { + "epoch": 2.08, + "learning_rate": 1.5291741472172354e-05, + "loss": 0.2016, + "step": 54181 + }, + { + "epoch": 2.08, + "learning_rate": 1.5291100282123622e-05, + "loss": 0.3563, + "step": 54182 + }, + { + "epoch": 2.08, + "learning_rate": 1.529045909207489e-05, + "loss": 0.2152, + "step": 54183 + }, + { + "epoch": 2.08, + "learning_rate": 1.528981790202616e-05, + "loss": 0.2777, + "step": 54184 + }, + { + "epoch": 2.08, + "learning_rate": 1.5289176711977434e-05, + "loss": 0.3201, + "step": 54185 + }, + { + "epoch": 2.08, + "learning_rate": 1.5288535521928702e-05, + "loss": 0.226, + "step": 54186 + }, + { + "epoch": 2.08, + "learning_rate": 1.528789433187997e-05, + "loss": 0.2001, + "step": 54187 + }, + { + "epoch": 2.08, + "learning_rate": 1.528725314183124e-05, + "loss": 0.2512, + "step": 54188 + }, + { + "epoch": 2.08, + "learning_rate": 1.528661195178251e-05, + "loss": 0.2641, + "step": 54189 + }, + { + "epoch": 2.08, + "learning_rate": 1.528597076173378e-05, + "loss": 0.1712, + "step": 54190 + }, + { + "epoch": 2.08, + "learning_rate": 1.5285329571685047e-05, + "loss": 0.2685, + "step": 54191 + }, + { + "epoch": 2.08, + "learning_rate": 1.5284688381636315e-05, + "loss": 0.2392, + "step": 54192 + }, + { + "epoch": 2.08, + "learning_rate": 1.5284047191587587e-05, + "loss": 0.1221, + "step": 54193 + }, + { + "epoch": 2.08, + "learning_rate": 1.528340600153886e-05, + "loss": 0.2515, + "step": 54194 + }, + { + "epoch": 2.08, + "learning_rate": 1.5282764811490127e-05, + "loss": 0.2383, + "step": 54195 + }, + { + "epoch": 2.08, + "learning_rate": 1.5282123621441395e-05, + "loss": 0.2684, + "step": 54196 + }, + { + "epoch": 2.08, + "learning_rate": 1.5281482431392667e-05, + "loss": 0.2864, + "step": 54197 + }, + { + "epoch": 2.08, + "learning_rate": 1.5280841241343935e-05, + "loss": 0.2345, + "step": 54198 + }, + { + "epoch": 2.08, + "learning_rate": 1.5280200051295204e-05, + "loss": 0.2882, + "step": 54199 + }, + { + "epoch": 2.08, + "learning_rate": 1.5279558861246472e-05, + "loss": 0.2358, + "step": 54200 + }, + { + "epoch": 2.08, + "learning_rate": 1.5278917671197744e-05, + "loss": 0.2527, + "step": 54201 + }, + { + "epoch": 2.08, + "learning_rate": 1.5278276481149012e-05, + "loss": 0.2513, + "step": 54202 + }, + { + "epoch": 2.08, + "learning_rate": 1.5277635291100284e-05, + "loss": 0.21, + "step": 54203 + }, + { + "epoch": 2.08, + "learning_rate": 1.5276994101051552e-05, + "loss": 0.2997, + "step": 54204 + }, + { + "epoch": 2.08, + "learning_rate": 1.5276352911002824e-05, + "loss": 0.1707, + "step": 54205 + }, + { + "epoch": 2.08, + "learning_rate": 1.5275711720954092e-05, + "loss": 0.1945, + "step": 54206 + }, + { + "epoch": 2.08, + "learning_rate": 1.527507053090536e-05, + "loss": 0.1929, + "step": 54207 + }, + { + "epoch": 2.08, + "learning_rate": 1.527442934085663e-05, + "loss": 0.1951, + "step": 54208 + }, + { + "epoch": 2.08, + "learning_rate": 1.52737881508079e-05, + "loss": 0.2841, + "step": 54209 + }, + { + "epoch": 2.08, + "learning_rate": 1.527314696075917e-05, + "loss": 0.2056, + "step": 54210 + }, + { + "epoch": 2.08, + "learning_rate": 1.5272505770710437e-05, + "loss": 0.3683, + "step": 54211 + }, + { + "epoch": 2.08, + "learning_rate": 1.527186458066171e-05, + "loss": 0.2559, + "step": 54212 + }, + { + "epoch": 2.08, + "learning_rate": 1.527122339061298e-05, + "loss": 0.3411, + "step": 54213 + }, + { + "epoch": 2.08, + "learning_rate": 1.527058220056425e-05, + "loss": 0.1692, + "step": 54214 + }, + { + "epoch": 2.08, + "learning_rate": 1.5269941010515517e-05, + "loss": 0.2602, + "step": 54215 + }, + { + "epoch": 2.08, + "learning_rate": 1.5269299820466786e-05, + "loss": 0.1891, + "step": 54216 + }, + { + "epoch": 2.08, + "learning_rate": 1.5268658630418057e-05, + "loss": 0.1359, + "step": 54217 + }, + { + "epoch": 2.08, + "learning_rate": 1.5268017440369326e-05, + "loss": 0.1192, + "step": 54218 + }, + { + "epoch": 2.08, + "learning_rate": 1.5267376250320594e-05, + "loss": 0.319, + "step": 54219 + }, + { + "epoch": 2.08, + "learning_rate": 1.5266735060271862e-05, + "loss": 0.2447, + "step": 54220 + }, + { + "epoch": 2.08, + "learning_rate": 1.5266093870223138e-05, + "loss": 0.3079, + "step": 54221 + }, + { + "epoch": 2.08, + "learning_rate": 1.5265452680174406e-05, + "loss": 0.1911, + "step": 54222 + }, + { + "epoch": 2.08, + "learning_rate": 1.5264811490125674e-05, + "loss": 0.1969, + "step": 54223 + }, + { + "epoch": 2.08, + "learning_rate": 1.5264170300076943e-05, + "loss": 0.2643, + "step": 54224 + }, + { + "epoch": 2.08, + "learning_rate": 1.5263529110028214e-05, + "loss": 0.2582, + "step": 54225 + }, + { + "epoch": 2.08, + "learning_rate": 1.5262887919979483e-05, + "loss": 0.2035, + "step": 54226 + }, + { + "epoch": 2.08, + "learning_rate": 1.526224672993075e-05, + "loss": 0.2816, + "step": 54227 + }, + { + "epoch": 2.08, + "learning_rate": 1.526160553988202e-05, + "loss": 0.2552, + "step": 54228 + }, + { + "epoch": 2.08, + "learning_rate": 1.526096434983329e-05, + "loss": 0.1296, + "step": 54229 + }, + { + "epoch": 2.08, + "learning_rate": 1.5260323159784563e-05, + "loss": 0.2456, + "step": 54230 + }, + { + "epoch": 2.09, + "learning_rate": 1.525968196973583e-05, + "loss": 0.3287, + "step": 54231 + }, + { + "epoch": 2.09, + "learning_rate": 1.52590407796871e-05, + "loss": 0.1601, + "step": 54232 + }, + { + "epoch": 2.09, + "learning_rate": 1.525839958963837e-05, + "loss": 0.1985, + "step": 54233 + }, + { + "epoch": 2.09, + "learning_rate": 1.525775839958964e-05, + "loss": 0.1954, + "step": 54234 + }, + { + "epoch": 2.09, + "learning_rate": 1.5257117209540908e-05, + "loss": 0.1535, + "step": 54235 + }, + { + "epoch": 2.09, + "learning_rate": 1.5256476019492178e-05, + "loss": 0.2958, + "step": 54236 + }, + { + "epoch": 2.09, + "learning_rate": 1.5255834829443446e-05, + "loss": 0.257, + "step": 54237 + }, + { + "epoch": 2.09, + "learning_rate": 1.525519363939472e-05, + "loss": 0.1733, + "step": 54238 + }, + { + "epoch": 2.09, + "learning_rate": 1.5254552449345988e-05, + "loss": 0.2115, + "step": 54239 + }, + { + "epoch": 2.09, + "learning_rate": 1.5253911259297258e-05, + "loss": 0.2201, + "step": 54240 + }, + { + "epoch": 2.09, + "learning_rate": 1.5253270069248526e-05, + "loss": 0.1263, + "step": 54241 + }, + { + "epoch": 2.09, + "learning_rate": 1.5252628879199796e-05, + "loss": 0.2235, + "step": 54242 + }, + { + "epoch": 2.09, + "learning_rate": 1.5251987689151064e-05, + "loss": 0.1822, + "step": 54243 + }, + { + "epoch": 2.09, + "learning_rate": 1.5251346499102334e-05, + "loss": 0.1355, + "step": 54244 + }, + { + "epoch": 2.09, + "learning_rate": 1.5250705309053603e-05, + "loss": 0.3106, + "step": 54245 + }, + { + "epoch": 2.09, + "learning_rate": 1.5250064119004873e-05, + "loss": 0.1543, + "step": 54246 + }, + { + "epoch": 2.09, + "learning_rate": 1.5249422928956145e-05, + "loss": 0.2335, + "step": 54247 + }, + { + "epoch": 2.09, + "learning_rate": 1.5248781738907415e-05, + "loss": 0.2711, + "step": 54248 + }, + { + "epoch": 2.09, + "learning_rate": 1.5248140548858683e-05, + "loss": 0.174, + "step": 54249 + }, + { + "epoch": 2.09, + "learning_rate": 1.5247499358809953e-05, + "loss": 0.164, + "step": 54250 + }, + { + "epoch": 2.09, + "learning_rate": 1.5246858168761221e-05, + "loss": 0.1612, + "step": 54251 + }, + { + "epoch": 2.09, + "learning_rate": 1.5246216978712491e-05, + "loss": 0.2313, + "step": 54252 + }, + { + "epoch": 2.09, + "learning_rate": 1.524557578866376e-05, + "loss": 0.2659, + "step": 54253 + }, + { + "epoch": 2.09, + "learning_rate": 1.524493459861503e-05, + "loss": 0.227, + "step": 54254 + }, + { + "epoch": 2.09, + "learning_rate": 1.5244293408566298e-05, + "loss": 0.3154, + "step": 54255 + }, + { + "epoch": 2.09, + "learning_rate": 1.5243652218517571e-05, + "loss": 0.399, + "step": 54256 + }, + { + "epoch": 2.09, + "learning_rate": 1.524301102846884e-05, + "loss": 0.3979, + "step": 54257 + }, + { + "epoch": 2.09, + "learning_rate": 1.524236983842011e-05, + "loss": 0.3279, + "step": 54258 + }, + { + "epoch": 2.09, + "learning_rate": 1.5241728648371378e-05, + "loss": 0.3354, + "step": 54259 + }, + { + "epoch": 2.09, + "learning_rate": 1.5241087458322648e-05, + "loss": 0.1846, + "step": 54260 + }, + { + "epoch": 2.09, + "learning_rate": 1.5240446268273916e-05, + "loss": 0.3826, + "step": 54261 + }, + { + "epoch": 2.09, + "learning_rate": 1.5239805078225186e-05, + "loss": 0.1067, + "step": 54262 + }, + { + "epoch": 2.09, + "learning_rate": 1.5239163888176455e-05, + "loss": 0.5777, + "step": 54263 + }, + { + "epoch": 2.09, + "learning_rate": 1.5238522698127725e-05, + "loss": 0.1845, + "step": 54264 + }, + { + "epoch": 2.09, + "learning_rate": 1.5237881508078996e-05, + "loss": 0.2057, + "step": 54265 + }, + { + "epoch": 2.09, + "learning_rate": 1.5237240318030267e-05, + "loss": 0.3307, + "step": 54266 + }, + { + "epoch": 2.09, + "learning_rate": 1.5236599127981535e-05, + "loss": 0.2306, + "step": 54267 + }, + { + "epoch": 2.09, + "learning_rate": 1.5235957937932805e-05, + "loss": 0.3459, + "step": 54268 + }, + { + "epoch": 2.09, + "learning_rate": 1.5235316747884073e-05, + "loss": 0.148, + "step": 54269 + }, + { + "epoch": 2.09, + "learning_rate": 1.5234675557835343e-05, + "loss": 0.1843, + "step": 54270 + }, + { + "epoch": 2.09, + "learning_rate": 1.5234034367786612e-05, + "loss": 0.3338, + "step": 54271 + }, + { + "epoch": 2.09, + "learning_rate": 1.5233393177737882e-05, + "loss": 0.3372, + "step": 54272 + }, + { + "epoch": 2.09, + "learning_rate": 1.523275198768915e-05, + "loss": 0.2438, + "step": 54273 + }, + { + "epoch": 2.09, + "learning_rate": 1.5232110797640423e-05, + "loss": 0.2209, + "step": 54274 + }, + { + "epoch": 2.09, + "learning_rate": 1.5231469607591692e-05, + "loss": 0.3427, + "step": 54275 + }, + { + "epoch": 2.09, + "learning_rate": 1.5230828417542962e-05, + "loss": 0.3148, + "step": 54276 + }, + { + "epoch": 2.09, + "learning_rate": 1.523018722749423e-05, + "loss": 0.307, + "step": 54277 + }, + { + "epoch": 2.09, + "learning_rate": 1.52295460374455e-05, + "loss": 0.2316, + "step": 54278 + }, + { + "epoch": 2.09, + "learning_rate": 1.5228904847396768e-05, + "loss": 0.2178, + "step": 54279 + }, + { + "epoch": 2.09, + "learning_rate": 1.5228263657348038e-05, + "loss": 0.2148, + "step": 54280 + }, + { + "epoch": 2.09, + "learning_rate": 1.5227622467299307e-05, + "loss": 0.1943, + "step": 54281 + }, + { + "epoch": 2.09, + "learning_rate": 1.5226981277250577e-05, + "loss": 0.3419, + "step": 54282 + }, + { + "epoch": 2.09, + "learning_rate": 1.5226340087201848e-05, + "loss": 0.1873, + "step": 54283 + }, + { + "epoch": 2.09, + "learning_rate": 1.5225698897153118e-05, + "loss": 0.3351, + "step": 54284 + }, + { + "epoch": 2.09, + "learning_rate": 1.5225057707104387e-05, + "loss": 0.1802, + "step": 54285 + }, + { + "epoch": 2.09, + "learning_rate": 1.5224416517055657e-05, + "loss": 0.2811, + "step": 54286 + }, + { + "epoch": 2.09, + "learning_rate": 1.5223775327006925e-05, + "loss": 0.3822, + "step": 54287 + }, + { + "epoch": 2.09, + "learning_rate": 1.5223134136958195e-05, + "loss": 0.2041, + "step": 54288 + }, + { + "epoch": 2.09, + "learning_rate": 1.5222492946909463e-05, + "loss": 0.1974, + "step": 54289 + }, + { + "epoch": 2.09, + "learning_rate": 1.5221851756860734e-05, + "loss": 0.1964, + "step": 54290 + }, + { + "epoch": 2.09, + "learning_rate": 1.5221210566812002e-05, + "loss": 0.3669, + "step": 54291 + }, + { + "epoch": 2.09, + "learning_rate": 1.5220569376763275e-05, + "loss": 0.2547, + "step": 54292 + }, + { + "epoch": 2.09, + "learning_rate": 1.5219928186714544e-05, + "loss": 0.1596, + "step": 54293 + }, + { + "epoch": 2.09, + "learning_rate": 1.5219286996665814e-05, + "loss": 0.2627, + "step": 54294 + }, + { + "epoch": 2.09, + "learning_rate": 1.5218645806617082e-05, + "loss": 0.2032, + "step": 54295 + }, + { + "epoch": 2.09, + "learning_rate": 1.5218004616568352e-05, + "loss": 0.2302, + "step": 54296 + }, + { + "epoch": 2.09, + "learning_rate": 1.521736342651962e-05, + "loss": 0.276, + "step": 54297 + }, + { + "epoch": 2.09, + "learning_rate": 1.521672223647089e-05, + "loss": 0.2825, + "step": 54298 + }, + { + "epoch": 2.09, + "learning_rate": 1.5216081046422159e-05, + "loss": 0.1989, + "step": 54299 + }, + { + "epoch": 2.09, + "learning_rate": 1.5215439856373429e-05, + "loss": 0.3704, + "step": 54300 + }, + { + "epoch": 2.09, + "learning_rate": 1.52147986663247e-05, + "loss": 0.193, + "step": 54301 + }, + { + "epoch": 2.09, + "learning_rate": 1.521415747627597e-05, + "loss": 0.2547, + "step": 54302 + }, + { + "epoch": 2.09, + "learning_rate": 1.5213516286227239e-05, + "loss": 0.1795, + "step": 54303 + }, + { + "epoch": 2.09, + "learning_rate": 1.5212875096178509e-05, + "loss": 0.1585, + "step": 54304 + }, + { + "epoch": 2.09, + "learning_rate": 1.5212233906129777e-05, + "loss": 0.207, + "step": 54305 + }, + { + "epoch": 2.09, + "learning_rate": 1.5211592716081047e-05, + "loss": 0.1452, + "step": 54306 + }, + { + "epoch": 2.09, + "learning_rate": 1.5210951526032315e-05, + "loss": 0.3404, + "step": 54307 + }, + { + "epoch": 2.09, + "learning_rate": 1.5210310335983585e-05, + "loss": 0.3114, + "step": 54308 + }, + { + "epoch": 2.09, + "learning_rate": 1.5209669145934854e-05, + "loss": 0.3618, + "step": 54309 + }, + { + "epoch": 2.09, + "learning_rate": 1.5209027955886127e-05, + "loss": 0.243, + "step": 54310 + }, + { + "epoch": 2.09, + "learning_rate": 1.5208386765837396e-05, + "loss": 0.2075, + "step": 54311 + }, + { + "epoch": 2.09, + "learning_rate": 1.5207745575788666e-05, + "loss": 0.1487, + "step": 54312 + }, + { + "epoch": 2.09, + "learning_rate": 1.5207104385739934e-05, + "loss": 0.1538, + "step": 54313 + }, + { + "epoch": 2.09, + "learning_rate": 1.5206463195691204e-05, + "loss": 0.1438, + "step": 54314 + }, + { + "epoch": 2.09, + "learning_rate": 1.5205822005642472e-05, + "loss": 0.1929, + "step": 54315 + }, + { + "epoch": 2.09, + "learning_rate": 1.5205180815593742e-05, + "loss": 0.255, + "step": 54316 + }, + { + "epoch": 2.09, + "learning_rate": 1.520453962554501e-05, + "loss": 0.1215, + "step": 54317 + }, + { + "epoch": 2.09, + "learning_rate": 1.520389843549628e-05, + "loss": 0.1583, + "step": 54318 + }, + { + "epoch": 2.09, + "learning_rate": 1.5203257245447552e-05, + "loss": 0.3777, + "step": 54319 + }, + { + "epoch": 2.09, + "learning_rate": 1.5202616055398822e-05, + "loss": 0.2172, + "step": 54320 + }, + { + "epoch": 2.09, + "learning_rate": 1.520197486535009e-05, + "loss": 0.225, + "step": 54321 + }, + { + "epoch": 2.09, + "learning_rate": 1.520133367530136e-05, + "loss": 0.118, + "step": 54322 + }, + { + "epoch": 2.09, + "learning_rate": 1.5200692485252629e-05, + "loss": 0.1782, + "step": 54323 + }, + { + "epoch": 2.09, + "learning_rate": 1.5200051295203899e-05, + "loss": 0.3968, + "step": 54324 + }, + { + "epoch": 2.09, + "learning_rate": 1.5199410105155167e-05, + "loss": 0.3648, + "step": 54325 + }, + { + "epoch": 2.09, + "learning_rate": 1.5198768915106437e-05, + "loss": 0.1629, + "step": 54326 + }, + { + "epoch": 2.09, + "learning_rate": 1.5198127725057709e-05, + "loss": 0.3173, + "step": 54327 + }, + { + "epoch": 2.09, + "learning_rate": 1.5197486535008979e-05, + "loss": 0.2876, + "step": 54328 + }, + { + "epoch": 2.09, + "learning_rate": 1.5196845344960247e-05, + "loss": 0.2683, + "step": 54329 + }, + { + "epoch": 2.09, + "learning_rate": 1.5196204154911517e-05, + "loss": 0.2853, + "step": 54330 + }, + { + "epoch": 2.09, + "learning_rate": 1.5195562964862786e-05, + "loss": 0.1421, + "step": 54331 + }, + { + "epoch": 2.09, + "learning_rate": 1.5194921774814056e-05, + "loss": 0.2197, + "step": 54332 + }, + { + "epoch": 2.09, + "learning_rate": 1.5194280584765324e-05, + "loss": 0.3331, + "step": 54333 + }, + { + "epoch": 2.09, + "learning_rate": 1.5193639394716594e-05, + "loss": 0.2259, + "step": 54334 + }, + { + "epoch": 2.09, + "learning_rate": 1.5192998204667863e-05, + "loss": 0.2751, + "step": 54335 + }, + { + "epoch": 2.09, + "learning_rate": 1.5192357014619136e-05, + "loss": 0.3217, + "step": 54336 + }, + { + "epoch": 2.09, + "learning_rate": 1.5191715824570404e-05, + "loss": 0.2456, + "step": 54337 + }, + { + "epoch": 2.09, + "learning_rate": 1.5191074634521674e-05, + "loss": 0.1731, + "step": 54338 + }, + { + "epoch": 2.09, + "learning_rate": 1.5190433444472943e-05, + "loss": 0.258, + "step": 54339 + }, + { + "epoch": 2.09, + "learning_rate": 1.5189792254424213e-05, + "loss": 0.2011, + "step": 54340 + }, + { + "epoch": 2.09, + "learning_rate": 1.5189151064375481e-05, + "loss": 0.213, + "step": 54341 + }, + { + "epoch": 2.09, + "learning_rate": 1.5188509874326751e-05, + "loss": 0.1767, + "step": 54342 + }, + { + "epoch": 2.09, + "learning_rate": 1.518786868427802e-05, + "loss": 0.2335, + "step": 54343 + }, + { + "epoch": 2.09, + "learning_rate": 1.518722749422929e-05, + "loss": 0.2675, + "step": 54344 + }, + { + "epoch": 2.09, + "learning_rate": 1.5186586304180561e-05, + "loss": 0.198, + "step": 54345 + }, + { + "epoch": 2.09, + "learning_rate": 1.5185945114131831e-05, + "loss": 0.1941, + "step": 54346 + }, + { + "epoch": 2.09, + "learning_rate": 1.51853039240831e-05, + "loss": 0.2668, + "step": 54347 + }, + { + "epoch": 2.09, + "learning_rate": 1.518466273403437e-05, + "loss": 0.1708, + "step": 54348 + }, + { + "epoch": 2.09, + "learning_rate": 1.5184021543985638e-05, + "loss": 0.2657, + "step": 54349 + }, + { + "epoch": 2.09, + "learning_rate": 1.5183380353936908e-05, + "loss": 0.3328, + "step": 54350 + }, + { + "epoch": 2.09, + "learning_rate": 1.5182739163888176e-05, + "loss": 0.2007, + "step": 54351 + }, + { + "epoch": 2.09, + "learning_rate": 1.5182097973839446e-05, + "loss": 0.2327, + "step": 54352 + }, + { + "epoch": 2.09, + "learning_rate": 1.5181456783790714e-05, + "loss": 0.3051, + "step": 54353 + }, + { + "epoch": 2.09, + "learning_rate": 1.5180815593741988e-05, + "loss": 0.1903, + "step": 54354 + }, + { + "epoch": 2.09, + "learning_rate": 1.5180174403693256e-05, + "loss": 0.2522, + "step": 54355 + }, + { + "epoch": 2.09, + "learning_rate": 1.5179533213644526e-05, + "loss": 0.2656, + "step": 54356 + }, + { + "epoch": 2.09, + "learning_rate": 1.5178892023595795e-05, + "loss": 0.3942, + "step": 54357 + }, + { + "epoch": 2.09, + "learning_rate": 1.5178250833547065e-05, + "loss": 0.23, + "step": 54358 + }, + { + "epoch": 2.09, + "learning_rate": 1.5177609643498333e-05, + "loss": 0.2593, + "step": 54359 + }, + { + "epoch": 2.09, + "learning_rate": 1.5176968453449603e-05, + "loss": 0.2446, + "step": 54360 + }, + { + "epoch": 2.09, + "learning_rate": 1.5176327263400871e-05, + "loss": 0.2162, + "step": 54361 + }, + { + "epoch": 2.09, + "learning_rate": 1.5175686073352141e-05, + "loss": 0.1645, + "step": 54362 + }, + { + "epoch": 2.09, + "learning_rate": 1.5175044883303413e-05, + "loss": 0.2202, + "step": 54363 + }, + { + "epoch": 2.09, + "learning_rate": 1.5174403693254683e-05, + "loss": 0.2892, + "step": 54364 + }, + { + "epoch": 2.09, + "learning_rate": 1.5173762503205951e-05, + "loss": 0.3633, + "step": 54365 + }, + { + "epoch": 2.09, + "learning_rate": 1.5173121313157221e-05, + "loss": 0.2655, + "step": 54366 + }, + { + "epoch": 2.09, + "learning_rate": 1.517248012310849e-05, + "loss": 0.3249, + "step": 54367 + }, + { + "epoch": 2.09, + "learning_rate": 1.517183893305976e-05, + "loss": 0.266, + "step": 54368 + }, + { + "epoch": 2.09, + "learning_rate": 1.5171197743011028e-05, + "loss": 0.2455, + "step": 54369 + }, + { + "epoch": 2.09, + "learning_rate": 1.5170556552962298e-05, + "loss": 0.3166, + "step": 54370 + }, + { + "epoch": 2.09, + "learning_rate": 1.5169915362913566e-05, + "loss": 0.2118, + "step": 54371 + }, + { + "epoch": 2.09, + "learning_rate": 1.516927417286484e-05, + "loss": 0.2952, + "step": 54372 + }, + { + "epoch": 2.09, + "learning_rate": 1.5168632982816108e-05, + "loss": 0.2991, + "step": 54373 + }, + { + "epoch": 2.09, + "learning_rate": 1.5167991792767378e-05, + "loss": 0.1936, + "step": 54374 + }, + { + "epoch": 2.09, + "learning_rate": 1.5167350602718646e-05, + "loss": 0.2492, + "step": 54375 + }, + { + "epoch": 2.09, + "learning_rate": 1.5166709412669916e-05, + "loss": 0.1007, + "step": 54376 + }, + { + "epoch": 2.09, + "learning_rate": 1.5166068222621185e-05, + "loss": 0.1797, + "step": 54377 + }, + { + "epoch": 2.09, + "learning_rate": 1.5165427032572455e-05, + "loss": 0.1579, + "step": 54378 + }, + { + "epoch": 2.09, + "learning_rate": 1.5164785842523723e-05, + "loss": 0.2594, + "step": 54379 + }, + { + "epoch": 2.09, + "learning_rate": 1.5164144652474993e-05, + "loss": 0.1803, + "step": 54380 + }, + { + "epoch": 2.09, + "learning_rate": 1.5163503462426265e-05, + "loss": 0.2767, + "step": 54381 + }, + { + "epoch": 2.09, + "learning_rate": 1.5162862272377535e-05, + "loss": 0.1468, + "step": 54382 + }, + { + "epoch": 2.09, + "learning_rate": 1.5162221082328803e-05, + "loss": 0.3285, + "step": 54383 + }, + { + "epoch": 2.09, + "learning_rate": 1.5161579892280073e-05, + "loss": 0.3213, + "step": 54384 + }, + { + "epoch": 2.09, + "learning_rate": 1.5160938702231342e-05, + "loss": 0.1975, + "step": 54385 + }, + { + "epoch": 2.09, + "learning_rate": 1.5160297512182612e-05, + "loss": 0.1628, + "step": 54386 + }, + { + "epoch": 2.09, + "learning_rate": 1.515965632213388e-05, + "loss": 0.2412, + "step": 54387 + }, + { + "epoch": 2.09, + "learning_rate": 1.515901513208515e-05, + "loss": 0.1513, + "step": 54388 + }, + { + "epoch": 2.09, + "learning_rate": 1.5158373942036418e-05, + "loss": 0.3514, + "step": 54389 + }, + { + "epoch": 2.09, + "learning_rate": 1.5157732751987692e-05, + "loss": 0.1874, + "step": 54390 + }, + { + "epoch": 2.09, + "learning_rate": 1.515709156193896e-05, + "loss": 0.3063, + "step": 54391 + }, + { + "epoch": 2.09, + "learning_rate": 1.515645037189023e-05, + "loss": 0.1725, + "step": 54392 + }, + { + "epoch": 2.09, + "learning_rate": 1.5155809181841498e-05, + "loss": 0.2078, + "step": 54393 + }, + { + "epoch": 2.09, + "learning_rate": 1.5155167991792768e-05, + "loss": 0.2769, + "step": 54394 + }, + { + "epoch": 2.09, + "learning_rate": 1.5154526801744037e-05, + "loss": 0.2188, + "step": 54395 + }, + { + "epoch": 2.09, + "learning_rate": 1.5153885611695307e-05, + "loss": 0.1779, + "step": 54396 + }, + { + "epoch": 2.09, + "learning_rate": 1.5153244421646575e-05, + "loss": 0.1148, + "step": 54397 + }, + { + "epoch": 2.09, + "learning_rate": 1.5152603231597845e-05, + "loss": 0.2275, + "step": 54398 + }, + { + "epoch": 2.09, + "learning_rate": 1.5151962041549117e-05, + "loss": 0.2056, + "step": 54399 + }, + { + "epoch": 2.09, + "learning_rate": 1.5151320851500387e-05, + "loss": 0.2788, + "step": 54400 + }, + { + "epoch": 2.09, + "learning_rate": 1.5150679661451655e-05, + "loss": 0.3633, + "step": 54401 + }, + { + "epoch": 2.09, + "learning_rate": 1.5150038471402925e-05, + "loss": 0.1813, + "step": 54402 + }, + { + "epoch": 2.09, + "learning_rate": 1.5149397281354194e-05, + "loss": 0.1644, + "step": 54403 + }, + { + "epoch": 2.09, + "learning_rate": 1.5148756091305464e-05, + "loss": 0.3201, + "step": 54404 + }, + { + "epoch": 2.09, + "learning_rate": 1.5148114901256732e-05, + "loss": 0.2182, + "step": 54405 + }, + { + "epoch": 2.09, + "learning_rate": 1.5147473711208002e-05, + "loss": 0.1663, + "step": 54406 + }, + { + "epoch": 2.09, + "learning_rate": 1.514683252115927e-05, + "loss": 0.297, + "step": 54407 + }, + { + "epoch": 2.09, + "learning_rate": 1.5146191331110544e-05, + "loss": 0.393, + "step": 54408 + }, + { + "epoch": 2.09, + "learning_rate": 1.5145550141061812e-05, + "loss": 0.2795, + "step": 54409 + }, + { + "epoch": 2.09, + "learning_rate": 1.5144908951013082e-05, + "loss": 0.2088, + "step": 54410 + }, + { + "epoch": 2.09, + "learning_rate": 1.514426776096435e-05, + "loss": 0.316, + "step": 54411 + }, + { + "epoch": 2.09, + "learning_rate": 1.514362657091562e-05, + "loss": 0.2384, + "step": 54412 + }, + { + "epoch": 2.09, + "learning_rate": 1.5142985380866889e-05, + "loss": 0.1723, + "step": 54413 + }, + { + "epoch": 2.09, + "learning_rate": 1.5142344190818159e-05, + "loss": 0.1725, + "step": 54414 + }, + { + "epoch": 2.09, + "learning_rate": 1.5141703000769427e-05, + "loss": 0.2128, + "step": 54415 + }, + { + "epoch": 2.09, + "learning_rate": 1.5141061810720697e-05, + "loss": 0.1855, + "step": 54416 + }, + { + "epoch": 2.09, + "learning_rate": 1.5140420620671969e-05, + "loss": 0.2576, + "step": 54417 + }, + { + "epoch": 2.09, + "learning_rate": 1.5139779430623239e-05, + "loss": 0.2055, + "step": 54418 + }, + { + "epoch": 2.09, + "learning_rate": 1.5139138240574507e-05, + "loss": 0.3669, + "step": 54419 + }, + { + "epoch": 2.09, + "learning_rate": 1.5138497050525777e-05, + "loss": 0.2434, + "step": 54420 + }, + { + "epoch": 2.09, + "learning_rate": 1.5137855860477045e-05, + "loss": 0.2605, + "step": 54421 + }, + { + "epoch": 2.09, + "learning_rate": 1.5137214670428316e-05, + "loss": 0.1939, + "step": 54422 + }, + { + "epoch": 2.09, + "learning_rate": 1.5136573480379584e-05, + "loss": 0.1428, + "step": 54423 + }, + { + "epoch": 2.09, + "learning_rate": 1.5135932290330854e-05, + "loss": 0.1627, + "step": 54424 + }, + { + "epoch": 2.09, + "learning_rate": 1.5135291100282126e-05, + "loss": 0.3401, + "step": 54425 + }, + { + "epoch": 2.09, + "learning_rate": 1.5134649910233396e-05, + "loss": 0.4717, + "step": 54426 + }, + { + "epoch": 2.09, + "learning_rate": 1.5134008720184664e-05, + "loss": 0.2504, + "step": 54427 + }, + { + "epoch": 2.09, + "learning_rate": 1.5133367530135934e-05, + "loss": 0.2464, + "step": 54428 + }, + { + "epoch": 2.09, + "learning_rate": 1.5132726340087202e-05, + "loss": 0.2575, + "step": 54429 + }, + { + "epoch": 2.09, + "learning_rate": 1.5132085150038472e-05, + "loss": 0.3649, + "step": 54430 + }, + { + "epoch": 2.09, + "learning_rate": 1.513144395998974e-05, + "loss": 0.1507, + "step": 54431 + }, + { + "epoch": 2.09, + "learning_rate": 1.513080276994101e-05, + "loss": 0.2642, + "step": 54432 + }, + { + "epoch": 2.09, + "learning_rate": 1.5130161579892279e-05, + "loss": 0.2019, + "step": 54433 + }, + { + "epoch": 2.09, + "learning_rate": 1.5129520389843552e-05, + "loss": 0.3314, + "step": 54434 + }, + { + "epoch": 2.09, + "learning_rate": 1.512887919979482e-05, + "loss": 0.2368, + "step": 54435 + }, + { + "epoch": 2.09, + "learning_rate": 1.512823800974609e-05, + "loss": 0.2198, + "step": 54436 + }, + { + "epoch": 2.09, + "learning_rate": 1.5127596819697359e-05, + "loss": 0.2533, + "step": 54437 + }, + { + "epoch": 2.09, + "learning_rate": 1.5126955629648629e-05, + "loss": 0.117, + "step": 54438 + }, + { + "epoch": 2.09, + "learning_rate": 1.5126314439599897e-05, + "loss": 0.1481, + "step": 54439 + }, + { + "epoch": 2.09, + "learning_rate": 1.5125673249551167e-05, + "loss": 0.1623, + "step": 54440 + }, + { + "epoch": 2.09, + "learning_rate": 1.5125032059502436e-05, + "loss": 0.2716, + "step": 54441 + }, + { + "epoch": 2.09, + "learning_rate": 1.5124390869453706e-05, + "loss": 0.2889, + "step": 54442 + }, + { + "epoch": 2.09, + "learning_rate": 1.5123749679404978e-05, + "loss": 0.3151, + "step": 54443 + }, + { + "epoch": 2.09, + "learning_rate": 1.5123108489356248e-05, + "loss": 0.2288, + "step": 54444 + }, + { + "epoch": 2.09, + "learning_rate": 1.5122467299307516e-05, + "loss": 0.1942, + "step": 54445 + }, + { + "epoch": 2.09, + "learning_rate": 1.5121826109258786e-05, + "loss": 0.1585, + "step": 54446 + }, + { + "epoch": 2.09, + "learning_rate": 1.5121184919210054e-05, + "loss": 0.3092, + "step": 54447 + }, + { + "epoch": 2.09, + "learning_rate": 1.5120543729161324e-05, + "loss": 0.1357, + "step": 54448 + }, + { + "epoch": 2.09, + "learning_rate": 1.5119902539112593e-05, + "loss": 0.2748, + "step": 54449 + }, + { + "epoch": 2.09, + "learning_rate": 1.5119261349063863e-05, + "loss": 0.2293, + "step": 54450 + }, + { + "epoch": 2.09, + "learning_rate": 1.5118620159015131e-05, + "loss": 0.2842, + "step": 54451 + }, + { + "epoch": 2.09, + "learning_rate": 1.5117978968966404e-05, + "loss": 0.2017, + "step": 54452 + }, + { + "epoch": 2.09, + "learning_rate": 1.5117337778917673e-05, + "loss": 0.2905, + "step": 54453 + }, + { + "epoch": 2.09, + "learning_rate": 1.5116696588868943e-05, + "loss": 0.1712, + "step": 54454 + }, + { + "epoch": 2.09, + "learning_rate": 1.5116055398820211e-05, + "loss": 0.2456, + "step": 54455 + }, + { + "epoch": 2.09, + "learning_rate": 1.5115414208771481e-05, + "loss": 0.3812, + "step": 54456 + }, + { + "epoch": 2.09, + "learning_rate": 1.511477301872275e-05, + "loss": 0.2568, + "step": 54457 + }, + { + "epoch": 2.09, + "learning_rate": 1.511413182867402e-05, + "loss": 0.14, + "step": 54458 + }, + { + "epoch": 2.09, + "learning_rate": 1.5113490638625288e-05, + "loss": 0.1964, + "step": 54459 + }, + { + "epoch": 2.09, + "learning_rate": 1.5112849448576558e-05, + "loss": 0.2837, + "step": 54460 + }, + { + "epoch": 2.09, + "learning_rate": 1.511220825852783e-05, + "loss": 0.3536, + "step": 54461 + }, + { + "epoch": 2.09, + "learning_rate": 1.51115670684791e-05, + "loss": 0.1724, + "step": 54462 + }, + { + "epoch": 2.09, + "learning_rate": 1.5110925878430368e-05, + "loss": 0.155, + "step": 54463 + }, + { + "epoch": 2.09, + "learning_rate": 1.5110284688381638e-05, + "loss": 0.1318, + "step": 54464 + }, + { + "epoch": 2.09, + "learning_rate": 1.5109643498332906e-05, + "loss": 0.2492, + "step": 54465 + }, + { + "epoch": 2.09, + "learning_rate": 1.5109002308284176e-05, + "loss": 0.2473, + "step": 54466 + }, + { + "epoch": 2.09, + "learning_rate": 1.5108361118235444e-05, + "loss": 0.1613, + "step": 54467 + }, + { + "epoch": 2.09, + "learning_rate": 1.5107719928186715e-05, + "loss": 0.2806, + "step": 54468 + }, + { + "epoch": 2.09, + "learning_rate": 1.5107078738137983e-05, + "loss": 0.2549, + "step": 54469 + }, + { + "epoch": 2.09, + "learning_rate": 1.5106437548089256e-05, + "loss": 0.3167, + "step": 54470 + }, + { + "epoch": 2.09, + "learning_rate": 1.5105796358040525e-05, + "loss": 0.231, + "step": 54471 + }, + { + "epoch": 2.09, + "learning_rate": 1.5105155167991795e-05, + "loss": 0.2826, + "step": 54472 + }, + { + "epoch": 2.09, + "learning_rate": 1.5104513977943063e-05, + "loss": 0.3984, + "step": 54473 + }, + { + "epoch": 2.09, + "learning_rate": 1.5103872787894333e-05, + "loss": 0.2657, + "step": 54474 + }, + { + "epoch": 2.09, + "learning_rate": 1.5103231597845601e-05, + "loss": 0.2753, + "step": 54475 + }, + { + "epoch": 2.09, + "learning_rate": 1.5102590407796871e-05, + "loss": 0.2833, + "step": 54476 + }, + { + "epoch": 2.09, + "learning_rate": 1.510194921774814e-05, + "loss": 0.1801, + "step": 54477 + }, + { + "epoch": 2.09, + "learning_rate": 1.510130802769941e-05, + "loss": 0.2278, + "step": 54478 + }, + { + "epoch": 2.09, + "learning_rate": 1.5100666837650681e-05, + "loss": 0.2262, + "step": 54479 + }, + { + "epoch": 2.09, + "learning_rate": 1.5100025647601951e-05, + "loss": 0.2489, + "step": 54480 + }, + { + "epoch": 2.09, + "learning_rate": 1.509938445755322e-05, + "loss": 0.2461, + "step": 54481 + }, + { + "epoch": 2.09, + "learning_rate": 1.509874326750449e-05, + "loss": 0.2888, + "step": 54482 + }, + { + "epoch": 2.09, + "learning_rate": 1.5098102077455758e-05, + "loss": 0.1716, + "step": 54483 + }, + { + "epoch": 2.09, + "learning_rate": 1.5097460887407028e-05, + "loss": 0.1253, + "step": 54484 + }, + { + "epoch": 2.09, + "learning_rate": 1.5096819697358296e-05, + "loss": 0.308, + "step": 54485 + }, + { + "epoch": 2.09, + "learning_rate": 1.5096178507309566e-05, + "loss": 0.2128, + "step": 54486 + }, + { + "epoch": 2.09, + "learning_rate": 1.5095537317260835e-05, + "loss": 0.1841, + "step": 54487 + }, + { + "epoch": 2.09, + "learning_rate": 1.5094896127212108e-05, + "loss": 0.1646, + "step": 54488 + }, + { + "epoch": 2.09, + "learning_rate": 1.5094254937163377e-05, + "loss": 0.1525, + "step": 54489 + }, + { + "epoch": 2.09, + "learning_rate": 1.5093613747114647e-05, + "loss": 0.1798, + "step": 54490 + }, + { + "epoch": 2.1, + "learning_rate": 1.5092972557065915e-05, + "loss": 0.1984, + "step": 54491 + }, + { + "epoch": 2.1, + "learning_rate": 1.5092331367017185e-05, + "loss": 0.3281, + "step": 54492 + }, + { + "epoch": 2.1, + "learning_rate": 1.5091690176968453e-05, + "loss": 0.203, + "step": 54493 + }, + { + "epoch": 2.1, + "learning_rate": 1.5091048986919723e-05, + "loss": 0.2096, + "step": 54494 + }, + { + "epoch": 2.1, + "learning_rate": 1.5090407796870992e-05, + "loss": 0.1539, + "step": 54495 + }, + { + "epoch": 2.1, + "learning_rate": 1.5089766606822262e-05, + "loss": 0.2159, + "step": 54496 + }, + { + "epoch": 2.1, + "learning_rate": 1.5089125416773533e-05, + "loss": 0.3072, + "step": 54497 + }, + { + "epoch": 2.1, + "learning_rate": 1.5088484226724803e-05, + "loss": 0.2677, + "step": 54498 + }, + { + "epoch": 2.1, + "learning_rate": 1.5087843036676072e-05, + "loss": 0.2198, + "step": 54499 + }, + { + "epoch": 2.1, + "learning_rate": 1.5087201846627342e-05, + "loss": 0.3646, + "step": 54500 + }, + { + "epoch": 2.1, + "learning_rate": 1.508656065657861e-05, + "loss": 0.2222, + "step": 54501 + }, + { + "epoch": 2.1, + "learning_rate": 1.508591946652988e-05, + "loss": 0.1359, + "step": 54502 + }, + { + "epoch": 2.1, + "learning_rate": 1.5085278276481148e-05, + "loss": 0.1741, + "step": 54503 + }, + { + "epoch": 2.1, + "learning_rate": 1.5084637086432418e-05, + "loss": 0.2131, + "step": 54504 + }, + { + "epoch": 2.1, + "learning_rate": 1.5083995896383687e-05, + "loss": 0.2035, + "step": 54505 + }, + { + "epoch": 2.1, + "learning_rate": 1.508335470633496e-05, + "loss": 0.1953, + "step": 54506 + }, + { + "epoch": 2.1, + "learning_rate": 1.5082713516286228e-05, + "loss": 0.2223, + "step": 54507 + }, + { + "epoch": 2.1, + "learning_rate": 1.5082072326237498e-05, + "loss": 0.1541, + "step": 54508 + }, + { + "epoch": 2.1, + "learning_rate": 1.5081431136188767e-05, + "loss": 0.2728, + "step": 54509 + }, + { + "epoch": 2.1, + "learning_rate": 1.5080789946140037e-05, + "loss": 0.2405, + "step": 54510 + }, + { + "epoch": 2.1, + "learning_rate": 1.5080148756091305e-05, + "loss": 0.3248, + "step": 54511 + }, + { + "epoch": 2.1, + "learning_rate": 1.5079507566042575e-05, + "loss": 0.2906, + "step": 54512 + }, + { + "epoch": 2.1, + "learning_rate": 1.5078866375993844e-05, + "loss": 0.1905, + "step": 54513 + }, + { + "epoch": 2.1, + "learning_rate": 1.5078225185945117e-05, + "loss": 0.3313, + "step": 54514 + }, + { + "epoch": 2.1, + "learning_rate": 1.5077583995896385e-05, + "loss": 0.3929, + "step": 54515 + }, + { + "epoch": 2.1, + "learning_rate": 1.5076942805847655e-05, + "loss": 0.2161, + "step": 54516 + }, + { + "epoch": 2.1, + "learning_rate": 1.5076301615798924e-05, + "loss": 0.314, + "step": 54517 + }, + { + "epoch": 2.1, + "learning_rate": 1.5075660425750194e-05, + "loss": 0.0675, + "step": 54518 + }, + { + "epoch": 2.1, + "learning_rate": 1.5075019235701462e-05, + "loss": 0.1348, + "step": 54519 + }, + { + "epoch": 2.1, + "learning_rate": 1.5074378045652732e-05, + "loss": 0.274, + "step": 54520 + }, + { + "epoch": 2.1, + "learning_rate": 1.5073736855604e-05, + "loss": 0.1779, + "step": 54521 + }, + { + "epoch": 2.1, + "learning_rate": 1.507309566555527e-05, + "loss": 0.1492, + "step": 54522 + }, + { + "epoch": 2.1, + "learning_rate": 1.5072454475506542e-05, + "loss": 0.1652, + "step": 54523 + }, + { + "epoch": 2.1, + "learning_rate": 1.5071813285457812e-05, + "loss": 0.2, + "step": 54524 + }, + { + "epoch": 2.1, + "learning_rate": 1.507117209540908e-05, + "loss": 0.2628, + "step": 54525 + }, + { + "epoch": 2.1, + "learning_rate": 1.507053090536035e-05, + "loss": 0.2598, + "step": 54526 + }, + { + "epoch": 2.1, + "learning_rate": 1.5069889715311619e-05, + "loss": 0.3224, + "step": 54527 + }, + { + "epoch": 2.1, + "learning_rate": 1.5069248525262889e-05, + "loss": 0.3334, + "step": 54528 + }, + { + "epoch": 2.1, + "learning_rate": 1.5068607335214157e-05, + "loss": 0.3619, + "step": 54529 + }, + { + "epoch": 2.1, + "learning_rate": 1.5067966145165427e-05, + "loss": 0.238, + "step": 54530 + }, + { + "epoch": 2.1, + "learning_rate": 1.5067324955116695e-05, + "loss": 0.1122, + "step": 54531 + }, + { + "epoch": 2.1, + "learning_rate": 1.5066683765067969e-05, + "loss": 0.1814, + "step": 54532 + }, + { + "epoch": 2.1, + "learning_rate": 1.5066042575019237e-05, + "loss": 0.3251, + "step": 54533 + }, + { + "epoch": 2.1, + "learning_rate": 1.5065401384970507e-05, + "loss": 0.1835, + "step": 54534 + }, + { + "epoch": 2.1, + "learning_rate": 1.5064760194921776e-05, + "loss": 0.1814, + "step": 54535 + }, + { + "epoch": 2.1, + "learning_rate": 1.5064119004873046e-05, + "loss": 0.3224, + "step": 54536 + }, + { + "epoch": 2.1, + "learning_rate": 1.5063477814824314e-05, + "loss": 0.2534, + "step": 54537 + }, + { + "epoch": 2.1, + "learning_rate": 1.5062836624775584e-05, + "loss": 0.3374, + "step": 54538 + }, + { + "epoch": 2.1, + "learning_rate": 1.5062195434726852e-05, + "loss": 0.1768, + "step": 54539 + }, + { + "epoch": 2.1, + "learning_rate": 1.5061554244678122e-05, + "loss": 0.2377, + "step": 54540 + }, + { + "epoch": 2.1, + "learning_rate": 1.5060913054629394e-05, + "loss": 0.2107, + "step": 54541 + }, + { + "epoch": 2.1, + "learning_rate": 1.5060271864580664e-05, + "loss": 0.1663, + "step": 54542 + }, + { + "epoch": 2.1, + "learning_rate": 1.5059630674531932e-05, + "loss": 0.1706, + "step": 54543 + }, + { + "epoch": 2.1, + "learning_rate": 1.5058989484483202e-05, + "loss": 0.1161, + "step": 54544 + }, + { + "epoch": 2.1, + "learning_rate": 1.505834829443447e-05, + "loss": 0.2273, + "step": 54545 + }, + { + "epoch": 2.1, + "learning_rate": 1.505770710438574e-05, + "loss": 0.2547, + "step": 54546 + }, + { + "epoch": 2.1, + "learning_rate": 1.5057065914337009e-05, + "loss": 0.2722, + "step": 54547 + }, + { + "epoch": 2.1, + "learning_rate": 1.5056424724288279e-05, + "loss": 0.1352, + "step": 54548 + }, + { + "epoch": 2.1, + "learning_rate": 1.5055783534239547e-05, + "loss": 0.2434, + "step": 54549 + }, + { + "epoch": 2.1, + "learning_rate": 1.505514234419082e-05, + "loss": 0.2155, + "step": 54550 + }, + { + "epoch": 2.1, + "learning_rate": 1.5054501154142089e-05, + "loss": 0.3385, + "step": 54551 + }, + { + "epoch": 2.1, + "learning_rate": 1.5053859964093359e-05, + "loss": 0.1521, + "step": 54552 + }, + { + "epoch": 2.1, + "learning_rate": 1.5053218774044627e-05, + "loss": 0.2154, + "step": 54553 + }, + { + "epoch": 2.1, + "learning_rate": 1.5052577583995897e-05, + "loss": 0.1969, + "step": 54554 + }, + { + "epoch": 2.1, + "learning_rate": 1.5051936393947166e-05, + "loss": 0.2064, + "step": 54555 + }, + { + "epoch": 2.1, + "learning_rate": 1.5051295203898436e-05, + "loss": 0.2669, + "step": 54556 + }, + { + "epoch": 2.1, + "learning_rate": 1.5050654013849704e-05, + "loss": 0.3043, + "step": 54557 + }, + { + "epoch": 2.1, + "learning_rate": 1.5050012823800974e-05, + "loss": 0.28, + "step": 54558 + }, + { + "epoch": 2.1, + "learning_rate": 1.5049371633752246e-05, + "loss": 0.3622, + "step": 54559 + }, + { + "epoch": 2.1, + "learning_rate": 1.5048730443703516e-05, + "loss": 0.335, + "step": 54560 + }, + { + "epoch": 2.1, + "learning_rate": 1.5048089253654784e-05, + "loss": 0.183, + "step": 54561 + }, + { + "epoch": 2.1, + "learning_rate": 1.5047448063606054e-05, + "loss": 0.2347, + "step": 54562 + }, + { + "epoch": 2.1, + "learning_rate": 1.5046806873557323e-05, + "loss": 0.1502, + "step": 54563 + }, + { + "epoch": 2.1, + "learning_rate": 1.5046165683508593e-05, + "loss": 0.1427, + "step": 54564 + }, + { + "epoch": 2.1, + "learning_rate": 1.5045524493459861e-05, + "loss": 0.2272, + "step": 54565 + }, + { + "epoch": 2.1, + "learning_rate": 1.5044883303411131e-05, + "loss": 0.1968, + "step": 54566 + }, + { + "epoch": 2.1, + "learning_rate": 1.50442421133624e-05, + "loss": 0.1649, + "step": 54567 + }, + { + "epoch": 2.1, + "learning_rate": 1.5043600923313673e-05, + "loss": 0.2253, + "step": 54568 + }, + { + "epoch": 2.1, + "learning_rate": 1.5042959733264941e-05, + "loss": 0.3132, + "step": 54569 + }, + { + "epoch": 2.1, + "learning_rate": 1.5042318543216211e-05, + "loss": 0.2289, + "step": 54570 + }, + { + "epoch": 2.1, + "learning_rate": 1.504167735316748e-05, + "loss": 0.285, + "step": 54571 + }, + { + "epoch": 2.1, + "learning_rate": 1.504103616311875e-05, + "loss": 0.1714, + "step": 54572 + }, + { + "epoch": 2.1, + "learning_rate": 1.5040394973070018e-05, + "loss": 0.2221, + "step": 54573 + }, + { + "epoch": 2.1, + "learning_rate": 1.5039753783021288e-05, + "loss": 0.2399, + "step": 54574 + }, + { + "epoch": 2.1, + "learning_rate": 1.5039112592972556e-05, + "loss": 0.261, + "step": 54575 + }, + { + "epoch": 2.1, + "learning_rate": 1.5038471402923826e-05, + "loss": 0.2346, + "step": 54576 + }, + { + "epoch": 2.1, + "learning_rate": 1.5037830212875098e-05, + "loss": 0.3076, + "step": 54577 + }, + { + "epoch": 2.1, + "learning_rate": 1.5037189022826368e-05, + "loss": 0.2069, + "step": 54578 + }, + { + "epoch": 2.1, + "learning_rate": 1.5036547832777636e-05, + "loss": 0.2024, + "step": 54579 + }, + { + "epoch": 2.1, + "learning_rate": 1.5035906642728906e-05, + "loss": 0.2303, + "step": 54580 + }, + { + "epoch": 2.1, + "learning_rate": 1.5035265452680175e-05, + "loss": 0.219, + "step": 54581 + }, + { + "epoch": 2.1, + "learning_rate": 1.5034624262631445e-05, + "loss": 0.3751, + "step": 54582 + }, + { + "epoch": 2.1, + "learning_rate": 1.5033983072582713e-05, + "loss": 0.2543, + "step": 54583 + }, + { + "epoch": 2.1, + "learning_rate": 1.5033341882533983e-05, + "loss": 0.2245, + "step": 54584 + }, + { + "epoch": 2.1, + "learning_rate": 1.5032700692485251e-05, + "loss": 0.2119, + "step": 54585 + }, + { + "epoch": 2.1, + "learning_rate": 1.5032059502436525e-05, + "loss": 0.2746, + "step": 54586 + }, + { + "epoch": 2.1, + "learning_rate": 1.5031418312387793e-05, + "loss": 0.2676, + "step": 54587 + }, + { + "epoch": 2.1, + "learning_rate": 1.5030777122339063e-05, + "loss": 0.3498, + "step": 54588 + }, + { + "epoch": 2.1, + "learning_rate": 1.5030135932290331e-05, + "loss": 0.2196, + "step": 54589 + }, + { + "epoch": 2.1, + "learning_rate": 1.5029494742241601e-05, + "loss": 0.1636, + "step": 54590 + }, + { + "epoch": 2.1, + "learning_rate": 1.502885355219287e-05, + "loss": 0.1594, + "step": 54591 + }, + { + "epoch": 2.1, + "learning_rate": 1.502821236214414e-05, + "loss": 0.2271, + "step": 54592 + }, + { + "epoch": 2.1, + "learning_rate": 1.5027571172095408e-05, + "loss": 0.2807, + "step": 54593 + }, + { + "epoch": 2.1, + "learning_rate": 1.5026929982046678e-05, + "loss": 0.1983, + "step": 54594 + }, + { + "epoch": 2.1, + "learning_rate": 1.502628879199795e-05, + "loss": 0.2888, + "step": 54595 + }, + { + "epoch": 2.1, + "learning_rate": 1.502564760194922e-05, + "loss": 0.1998, + "step": 54596 + }, + { + "epoch": 2.1, + "learning_rate": 1.5025006411900488e-05, + "loss": 0.3313, + "step": 54597 + }, + { + "epoch": 2.1, + "learning_rate": 1.5024365221851758e-05, + "loss": 0.2898, + "step": 54598 + }, + { + "epoch": 2.1, + "learning_rate": 1.5023724031803026e-05, + "loss": 0.3054, + "step": 54599 + }, + { + "epoch": 2.1, + "learning_rate": 1.5023082841754297e-05, + "loss": 0.1843, + "step": 54600 + }, + { + "epoch": 2.1, + "learning_rate": 1.5022441651705565e-05, + "loss": 0.2244, + "step": 54601 + }, + { + "epoch": 2.1, + "learning_rate": 1.5021800461656835e-05, + "loss": 0.2564, + "step": 54602 + }, + { + "epoch": 2.1, + "learning_rate": 1.5021159271608103e-05, + "loss": 0.175, + "step": 54603 + }, + { + "epoch": 2.1, + "learning_rate": 1.5020518081559377e-05, + "loss": 0.2611, + "step": 54604 + }, + { + "epoch": 2.1, + "learning_rate": 1.5019876891510645e-05, + "loss": 0.2674, + "step": 54605 + }, + { + "epoch": 2.1, + "learning_rate": 1.5019235701461915e-05, + "loss": 0.3521, + "step": 54606 + }, + { + "epoch": 2.1, + "learning_rate": 1.5018594511413183e-05, + "loss": 0.2361, + "step": 54607 + }, + { + "epoch": 2.1, + "learning_rate": 1.5017953321364453e-05, + "loss": 0.2136, + "step": 54608 + }, + { + "epoch": 2.1, + "learning_rate": 1.5017312131315722e-05, + "loss": 0.1898, + "step": 54609 + }, + { + "epoch": 2.1, + "learning_rate": 1.5016670941266992e-05, + "loss": 0.2524, + "step": 54610 + }, + { + "epoch": 2.1, + "learning_rate": 1.501602975121826e-05, + "loss": 0.2683, + "step": 54611 + }, + { + "epoch": 2.1, + "learning_rate": 1.5015388561169533e-05, + "loss": 0.1411, + "step": 54612 + }, + { + "epoch": 2.1, + "learning_rate": 1.5014747371120802e-05, + "loss": 0.2133, + "step": 54613 + }, + { + "epoch": 2.1, + "learning_rate": 1.5014106181072072e-05, + "loss": 0.2688, + "step": 54614 + }, + { + "epoch": 2.1, + "learning_rate": 1.501346499102334e-05, + "loss": 0.2727, + "step": 54615 + }, + { + "epoch": 2.1, + "learning_rate": 1.501282380097461e-05, + "loss": 0.2059, + "step": 54616 + }, + { + "epoch": 2.1, + "learning_rate": 1.5012182610925878e-05, + "loss": 0.3236, + "step": 54617 + }, + { + "epoch": 2.1, + "learning_rate": 1.5011541420877148e-05, + "loss": 0.2767, + "step": 54618 + }, + { + "epoch": 2.1, + "learning_rate": 1.5010900230828417e-05, + "loss": 0.1502, + "step": 54619 + }, + { + "epoch": 2.1, + "learning_rate": 1.5010259040779687e-05, + "loss": 0.3214, + "step": 54620 + }, + { + "epoch": 2.1, + "learning_rate": 1.5009617850730959e-05, + "loss": 0.3397, + "step": 54621 + }, + { + "epoch": 2.1, + "learning_rate": 1.5008976660682229e-05, + "loss": 0.1835, + "step": 54622 + }, + { + "epoch": 2.1, + "learning_rate": 1.5008335470633497e-05, + "loss": 0.2273, + "step": 54623 + }, + { + "epoch": 2.1, + "learning_rate": 1.5007694280584767e-05, + "loss": 0.212, + "step": 54624 + }, + { + "epoch": 2.1, + "learning_rate": 1.5007053090536035e-05, + "loss": 0.2003, + "step": 54625 + }, + { + "epoch": 2.1, + "learning_rate": 1.5006411900487305e-05, + "loss": 0.2697, + "step": 54626 + }, + { + "epoch": 2.1, + "learning_rate": 1.5005770710438574e-05, + "loss": 0.1962, + "step": 54627 + }, + { + "epoch": 2.1, + "learning_rate": 1.5005129520389844e-05, + "loss": 0.2354, + "step": 54628 + }, + { + "epoch": 2.1, + "learning_rate": 1.5004488330341112e-05, + "loss": 0.3135, + "step": 54629 + }, + { + "epoch": 2.1, + "learning_rate": 1.5003847140292385e-05, + "loss": 0.1738, + "step": 54630 + }, + { + "epoch": 2.1, + "learning_rate": 1.5003205950243654e-05, + "loss": 0.2382, + "step": 54631 + }, + { + "epoch": 2.1, + "learning_rate": 1.5002564760194924e-05, + "loss": 0.3239, + "step": 54632 + }, + { + "epoch": 2.1, + "learning_rate": 1.5001923570146192e-05, + "loss": 0.1438, + "step": 54633 + }, + { + "epoch": 2.1, + "learning_rate": 1.5001282380097462e-05, + "loss": 0.148, + "step": 54634 + }, + { + "epoch": 2.1, + "learning_rate": 1.500064119004873e-05, + "loss": 0.1528, + "step": 54635 + }, + { + "epoch": 2.1, + "learning_rate": 1.5e-05, + "loss": 0.2351, + "step": 54636 + }, + { + "epoch": 2.1, + "learning_rate": 1.4999358809951269e-05, + "loss": 0.1655, + "step": 54637 + }, + { + "epoch": 2.1, + "learning_rate": 1.4998717619902539e-05, + "loss": 0.1639, + "step": 54638 + }, + { + "epoch": 2.1, + "learning_rate": 1.499807642985381e-05, + "loss": 0.278, + "step": 54639 + }, + { + "epoch": 2.1, + "learning_rate": 1.499743523980508e-05, + "loss": 0.4498, + "step": 54640 + }, + { + "epoch": 2.1, + "learning_rate": 1.4996794049756349e-05, + "loss": 0.1661, + "step": 54641 + }, + { + "epoch": 2.1, + "learning_rate": 1.4996152859707619e-05, + "loss": 0.1565, + "step": 54642 + }, + { + "epoch": 2.1, + "learning_rate": 1.4995511669658887e-05, + "loss": 0.2185, + "step": 54643 + }, + { + "epoch": 2.1, + "learning_rate": 1.4994870479610157e-05, + "loss": 0.2378, + "step": 54644 + }, + { + "epoch": 2.1, + "learning_rate": 1.4994229289561426e-05, + "loss": 0.2863, + "step": 54645 + }, + { + "epoch": 2.1, + "learning_rate": 1.4993588099512696e-05, + "loss": 0.1953, + "step": 54646 + }, + { + "epoch": 2.1, + "learning_rate": 1.4992946909463964e-05, + "loss": 0.2711, + "step": 54647 + }, + { + "epoch": 2.1, + "learning_rate": 1.4992305719415237e-05, + "loss": 0.253, + "step": 54648 + }, + { + "epoch": 2.1, + "learning_rate": 1.4991664529366506e-05, + "loss": 0.3874, + "step": 54649 + }, + { + "epoch": 2.1, + "learning_rate": 1.4991023339317776e-05, + "loss": 0.3422, + "step": 54650 + }, + { + "epoch": 2.1, + "learning_rate": 1.4990382149269044e-05, + "loss": 0.3208, + "step": 54651 + }, + { + "epoch": 2.1, + "learning_rate": 1.4989740959220314e-05, + "loss": 0.2281, + "step": 54652 + }, + { + "epoch": 2.1, + "learning_rate": 1.4989099769171582e-05, + "loss": 0.2493, + "step": 54653 + }, + { + "epoch": 2.1, + "learning_rate": 1.4988458579122852e-05, + "loss": 0.1911, + "step": 54654 + }, + { + "epoch": 2.1, + "learning_rate": 1.498781738907412e-05, + "loss": 0.1821, + "step": 54655 + }, + { + "epoch": 2.1, + "learning_rate": 1.498717619902539e-05, + "loss": 0.241, + "step": 54656 + }, + { + "epoch": 2.1, + "learning_rate": 1.4986535008976662e-05, + "loss": 0.257, + "step": 54657 + }, + { + "epoch": 2.1, + "learning_rate": 1.4985893818927932e-05, + "loss": 0.296, + "step": 54658 + }, + { + "epoch": 2.1, + "learning_rate": 1.49852526288792e-05, + "loss": 0.2229, + "step": 54659 + }, + { + "epoch": 2.1, + "learning_rate": 1.498461143883047e-05, + "loss": 0.2025, + "step": 54660 + }, + { + "epoch": 2.1, + "learning_rate": 1.4983970248781739e-05, + "loss": 0.1076, + "step": 54661 + }, + { + "epoch": 2.1, + "learning_rate": 1.4983329058733009e-05, + "loss": 0.2558, + "step": 54662 + }, + { + "epoch": 2.1, + "learning_rate": 1.4982687868684277e-05, + "loss": 0.2422, + "step": 54663 + }, + { + "epoch": 2.1, + "learning_rate": 1.4982046678635547e-05, + "loss": 0.248, + "step": 54664 + }, + { + "epoch": 2.1, + "learning_rate": 1.4981405488586816e-05, + "loss": 0.2357, + "step": 54665 + }, + { + "epoch": 2.1, + "learning_rate": 1.498076429853809e-05, + "loss": 0.2902, + "step": 54666 + }, + { + "epoch": 2.1, + "learning_rate": 1.4980123108489358e-05, + "loss": 0.2853, + "step": 54667 + }, + { + "epoch": 2.1, + "learning_rate": 1.4979481918440628e-05, + "loss": 0.3144, + "step": 54668 + }, + { + "epoch": 2.1, + "learning_rate": 1.4978840728391896e-05, + "loss": 0.2766, + "step": 54669 + }, + { + "epoch": 2.1, + "learning_rate": 1.4978199538343166e-05, + "loss": 0.141, + "step": 54670 + }, + { + "epoch": 2.1, + "learning_rate": 1.4977558348294434e-05, + "loss": 0.2389, + "step": 54671 + }, + { + "epoch": 2.1, + "learning_rate": 1.4976917158245704e-05, + "loss": 0.2523, + "step": 54672 + }, + { + "epoch": 2.1, + "learning_rate": 1.4976275968196973e-05, + "loss": 0.3361, + "step": 54673 + }, + { + "epoch": 2.1, + "learning_rate": 1.4975634778148243e-05, + "loss": 0.3293, + "step": 54674 + }, + { + "epoch": 2.1, + "learning_rate": 1.4974993588099514e-05, + "loss": 0.2494, + "step": 54675 + }, + { + "epoch": 2.1, + "learning_rate": 1.4974352398050784e-05, + "loss": 0.2096, + "step": 54676 + }, + { + "epoch": 2.1, + "learning_rate": 1.4973711208002053e-05, + "loss": 0.3226, + "step": 54677 + }, + { + "epoch": 2.1, + "learning_rate": 1.4973070017953323e-05, + "loss": 0.272, + "step": 54678 + }, + { + "epoch": 2.1, + "learning_rate": 1.4972428827904591e-05, + "loss": 0.1988, + "step": 54679 + }, + { + "epoch": 2.1, + "learning_rate": 1.4971787637855861e-05, + "loss": 0.313, + "step": 54680 + }, + { + "epoch": 2.1, + "learning_rate": 1.497114644780713e-05, + "loss": 0.3136, + "step": 54681 + }, + { + "epoch": 2.1, + "learning_rate": 1.49705052577584e-05, + "loss": 0.2755, + "step": 54682 + }, + { + "epoch": 2.1, + "learning_rate": 1.4969864067709668e-05, + "loss": 0.2357, + "step": 54683 + }, + { + "epoch": 2.1, + "learning_rate": 1.4969222877660941e-05, + "loss": 0.2133, + "step": 54684 + }, + { + "epoch": 2.1, + "learning_rate": 1.496858168761221e-05, + "loss": 0.1882, + "step": 54685 + }, + { + "epoch": 2.1, + "learning_rate": 1.496794049756348e-05, + "loss": 0.137, + "step": 54686 + }, + { + "epoch": 2.1, + "learning_rate": 1.4967299307514748e-05, + "loss": 0.2404, + "step": 54687 + }, + { + "epoch": 2.1, + "learning_rate": 1.4966658117466018e-05, + "loss": 0.2724, + "step": 54688 + }, + { + "epoch": 2.1, + "learning_rate": 1.4966016927417286e-05, + "loss": 0.1597, + "step": 54689 + }, + { + "epoch": 2.1, + "learning_rate": 1.4965375737368556e-05, + "loss": 0.2363, + "step": 54690 + }, + { + "epoch": 2.1, + "learning_rate": 1.4964734547319825e-05, + "loss": 0.3512, + "step": 54691 + }, + { + "epoch": 2.1, + "learning_rate": 1.4964093357271095e-05, + "loss": 0.152, + "step": 54692 + }, + { + "epoch": 2.1, + "learning_rate": 1.4963452167222366e-05, + "loss": 0.2323, + "step": 54693 + }, + { + "epoch": 2.1, + "learning_rate": 1.4962810977173636e-05, + "loss": 0.3099, + "step": 54694 + }, + { + "epoch": 2.1, + "learning_rate": 1.4962169787124905e-05, + "loss": 0.2957, + "step": 54695 + }, + { + "epoch": 2.1, + "learning_rate": 1.4961528597076175e-05, + "loss": 0.2785, + "step": 54696 + }, + { + "epoch": 2.1, + "learning_rate": 1.4960887407027443e-05, + "loss": 0.1342, + "step": 54697 + }, + { + "epoch": 2.1, + "learning_rate": 1.4960246216978713e-05, + "loss": 0.2175, + "step": 54698 + }, + { + "epoch": 2.1, + "learning_rate": 1.4959605026929981e-05, + "loss": 0.272, + "step": 54699 + }, + { + "epoch": 2.1, + "learning_rate": 1.4958963836881251e-05, + "loss": 0.2108, + "step": 54700 + }, + { + "epoch": 2.1, + "learning_rate": 1.495832264683252e-05, + "loss": 0.3902, + "step": 54701 + }, + { + "epoch": 2.1, + "learning_rate": 1.4957681456783793e-05, + "loss": 0.3405, + "step": 54702 + }, + { + "epoch": 2.1, + "learning_rate": 1.4957040266735061e-05, + "loss": 0.0979, + "step": 54703 + }, + { + "epoch": 2.1, + "learning_rate": 1.4956399076686331e-05, + "loss": 0.347, + "step": 54704 + }, + { + "epoch": 2.1, + "learning_rate": 1.49557578866376e-05, + "loss": 0.355, + "step": 54705 + }, + { + "epoch": 2.1, + "learning_rate": 1.495511669658887e-05, + "loss": 0.1696, + "step": 54706 + }, + { + "epoch": 2.1, + "learning_rate": 1.4954475506540138e-05, + "loss": 0.2134, + "step": 54707 + }, + { + "epoch": 2.1, + "learning_rate": 1.4953834316491408e-05, + "loss": 0.1724, + "step": 54708 + }, + { + "epoch": 2.1, + "learning_rate": 1.4953193126442676e-05, + "loss": 0.2818, + "step": 54709 + }, + { + "epoch": 2.1, + "learning_rate": 1.495255193639395e-05, + "loss": 0.1709, + "step": 54710 + }, + { + "epoch": 2.1, + "learning_rate": 1.4951910746345218e-05, + "loss": 0.3377, + "step": 54711 + }, + { + "epoch": 2.1, + "learning_rate": 1.4951269556296488e-05, + "loss": 0.2865, + "step": 54712 + }, + { + "epoch": 2.1, + "learning_rate": 1.4950628366247757e-05, + "loss": 0.2958, + "step": 54713 + }, + { + "epoch": 2.1, + "learning_rate": 1.4949987176199027e-05, + "loss": 0.2203, + "step": 54714 + }, + { + "epoch": 2.1, + "learning_rate": 1.4949345986150295e-05, + "loss": 0.1175, + "step": 54715 + }, + { + "epoch": 2.1, + "learning_rate": 1.4948704796101565e-05, + "loss": 0.1494, + "step": 54716 + }, + { + "epoch": 2.1, + "learning_rate": 1.4948063606052833e-05, + "loss": 0.1377, + "step": 54717 + }, + { + "epoch": 2.1, + "learning_rate": 1.4947422416004103e-05, + "loss": 0.2602, + "step": 54718 + }, + { + "epoch": 2.1, + "learning_rate": 1.4946781225955375e-05, + "loss": 0.21, + "step": 54719 + }, + { + "epoch": 2.1, + "learning_rate": 1.4946140035906645e-05, + "loss": 0.2386, + "step": 54720 + }, + { + "epoch": 2.1, + "learning_rate": 1.4945498845857913e-05, + "loss": 0.1733, + "step": 54721 + }, + { + "epoch": 2.1, + "learning_rate": 1.4944857655809183e-05, + "loss": 0.2713, + "step": 54722 + }, + { + "epoch": 2.1, + "learning_rate": 1.4944216465760452e-05, + "loss": 0.2269, + "step": 54723 + }, + { + "epoch": 2.1, + "learning_rate": 1.4943575275711722e-05, + "loss": 0.2573, + "step": 54724 + }, + { + "epoch": 2.1, + "learning_rate": 1.494293408566299e-05, + "loss": 0.3019, + "step": 54725 + }, + { + "epoch": 2.1, + "learning_rate": 1.494229289561426e-05, + "loss": 0.223, + "step": 54726 + }, + { + "epoch": 2.1, + "learning_rate": 1.4941651705565528e-05, + "loss": 0.2462, + "step": 54727 + }, + { + "epoch": 2.1, + "learning_rate": 1.4941010515516802e-05, + "loss": 0.2299, + "step": 54728 + }, + { + "epoch": 2.1, + "learning_rate": 1.494036932546807e-05, + "loss": 0.3571, + "step": 54729 + }, + { + "epoch": 2.1, + "learning_rate": 1.493972813541934e-05, + "loss": 0.277, + "step": 54730 + }, + { + "epoch": 2.1, + "learning_rate": 1.4939086945370608e-05, + "loss": 0.1081, + "step": 54731 + }, + { + "epoch": 2.1, + "learning_rate": 1.4938445755321878e-05, + "loss": 0.2773, + "step": 54732 + }, + { + "epoch": 2.1, + "learning_rate": 1.4937804565273147e-05, + "loss": 0.1838, + "step": 54733 + }, + { + "epoch": 2.1, + "learning_rate": 1.4937163375224417e-05, + "loss": 0.1901, + "step": 54734 + }, + { + "epoch": 2.1, + "learning_rate": 1.4936522185175685e-05, + "loss": 0.2663, + "step": 54735 + }, + { + "epoch": 2.1, + "learning_rate": 1.4935880995126955e-05, + "loss": 0.2627, + "step": 54736 + }, + { + "epoch": 2.1, + "learning_rate": 1.4935239805078227e-05, + "loss": 0.172, + "step": 54737 + }, + { + "epoch": 2.1, + "learning_rate": 1.4934598615029497e-05, + "loss": 0.1724, + "step": 54738 + }, + { + "epoch": 2.1, + "learning_rate": 1.4933957424980765e-05, + "loss": 0.3568, + "step": 54739 + }, + { + "epoch": 2.1, + "learning_rate": 1.4933316234932035e-05, + "loss": 0.2532, + "step": 54740 + }, + { + "epoch": 2.1, + "learning_rate": 1.4932675044883304e-05, + "loss": 0.2182, + "step": 54741 + }, + { + "epoch": 2.1, + "learning_rate": 1.4932033854834574e-05, + "loss": 0.215, + "step": 54742 + }, + { + "epoch": 2.1, + "learning_rate": 1.4931392664785842e-05, + "loss": 0.2117, + "step": 54743 + }, + { + "epoch": 2.1, + "learning_rate": 1.4930751474737112e-05, + "loss": 0.19, + "step": 54744 + }, + { + "epoch": 2.1, + "learning_rate": 1.493011028468838e-05, + "loss": 0.1845, + "step": 54745 + }, + { + "epoch": 2.1, + "learning_rate": 1.4929469094639654e-05, + "loss": 0.246, + "step": 54746 + }, + { + "epoch": 2.1, + "learning_rate": 1.4928827904590922e-05, + "loss": 0.2508, + "step": 54747 + }, + { + "epoch": 2.1, + "learning_rate": 1.4928186714542192e-05, + "loss": 0.3009, + "step": 54748 + }, + { + "epoch": 2.1, + "learning_rate": 1.492754552449346e-05, + "loss": 0.2428, + "step": 54749 + }, + { + "epoch": 2.1, + "learning_rate": 1.492690433444473e-05, + "loss": 0.1616, + "step": 54750 + }, + { + "epoch": 2.1, + "learning_rate": 1.4926263144395999e-05, + "loss": 0.2176, + "step": 54751 + }, + { + "epoch": 2.11, + "learning_rate": 1.4925621954347269e-05, + "loss": 0.4015, + "step": 54752 + }, + { + "epoch": 2.11, + "learning_rate": 1.4924980764298537e-05, + "loss": 0.1541, + "step": 54753 + }, + { + "epoch": 2.11, + "learning_rate": 1.4924339574249807e-05, + "loss": 0.2341, + "step": 54754 + }, + { + "epoch": 2.11, + "learning_rate": 1.4923698384201079e-05, + "loss": 0.2081, + "step": 54755 + }, + { + "epoch": 2.11, + "learning_rate": 1.4923057194152349e-05, + "loss": 0.1921, + "step": 54756 + }, + { + "epoch": 2.11, + "learning_rate": 1.4922416004103617e-05, + "loss": 0.158, + "step": 54757 + }, + { + "epoch": 2.11, + "learning_rate": 1.4921774814054887e-05, + "loss": 0.1457, + "step": 54758 + }, + { + "epoch": 2.11, + "learning_rate": 1.4921133624006156e-05, + "loss": 0.2049, + "step": 54759 + }, + { + "epoch": 2.11, + "learning_rate": 1.4920492433957426e-05, + "loss": 0.1561, + "step": 54760 + }, + { + "epoch": 2.11, + "learning_rate": 1.4919851243908694e-05, + "loss": 0.2108, + "step": 54761 + }, + { + "epoch": 2.11, + "learning_rate": 1.4919210053859964e-05, + "loss": 0.2756, + "step": 54762 + }, + { + "epoch": 2.11, + "learning_rate": 1.4918568863811232e-05, + "loss": 0.2369, + "step": 54763 + }, + { + "epoch": 2.11, + "learning_rate": 1.4917927673762506e-05, + "loss": 0.2753, + "step": 54764 + }, + { + "epoch": 2.11, + "learning_rate": 1.4917286483713774e-05, + "loss": 0.1959, + "step": 54765 + }, + { + "epoch": 2.11, + "learning_rate": 1.4916645293665044e-05, + "loss": 0.3112, + "step": 54766 + }, + { + "epoch": 2.11, + "learning_rate": 1.4916004103616312e-05, + "loss": 0.2893, + "step": 54767 + }, + { + "epoch": 2.11, + "learning_rate": 1.4915362913567582e-05, + "loss": 0.2116, + "step": 54768 + }, + { + "epoch": 2.11, + "learning_rate": 1.491472172351885e-05, + "loss": 0.1245, + "step": 54769 + }, + { + "epoch": 2.11, + "learning_rate": 1.491408053347012e-05, + "loss": 0.2837, + "step": 54770 + }, + { + "epoch": 2.11, + "learning_rate": 1.4913439343421389e-05, + "loss": 0.2167, + "step": 54771 + }, + { + "epoch": 2.11, + "learning_rate": 1.4912798153372659e-05, + "loss": 0.2402, + "step": 54772 + }, + { + "epoch": 2.11, + "learning_rate": 1.491215696332393e-05, + "loss": 0.3911, + "step": 54773 + }, + { + "epoch": 2.11, + "learning_rate": 1.49115157732752e-05, + "loss": 0.1885, + "step": 54774 + }, + { + "epoch": 2.11, + "learning_rate": 1.4910874583226469e-05, + "loss": 0.1818, + "step": 54775 + }, + { + "epoch": 2.11, + "learning_rate": 1.491023339317774e-05, + "loss": 0.1527, + "step": 54776 + }, + { + "epoch": 2.11, + "learning_rate": 1.4909592203129007e-05, + "loss": 0.1588, + "step": 54777 + }, + { + "epoch": 2.11, + "learning_rate": 1.4908951013080278e-05, + "loss": 0.2701, + "step": 54778 + }, + { + "epoch": 2.11, + "learning_rate": 1.4908309823031546e-05, + "loss": 0.2132, + "step": 54779 + }, + { + "epoch": 2.11, + "learning_rate": 1.4907668632982816e-05, + "loss": 0.1793, + "step": 54780 + }, + { + "epoch": 2.11, + "learning_rate": 1.4907027442934086e-05, + "loss": 0.4066, + "step": 54781 + }, + { + "epoch": 2.11, + "learning_rate": 1.4906386252885358e-05, + "loss": 0.3355, + "step": 54782 + }, + { + "epoch": 2.11, + "learning_rate": 1.4905745062836626e-05, + "loss": 0.165, + "step": 54783 + }, + { + "epoch": 2.11, + "learning_rate": 1.4905103872787896e-05, + "loss": 0.3627, + "step": 54784 + }, + { + "epoch": 2.11, + "learning_rate": 1.4904462682739164e-05, + "loss": 0.2247, + "step": 54785 + }, + { + "epoch": 2.11, + "learning_rate": 1.4903821492690434e-05, + "loss": 0.2377, + "step": 54786 + }, + { + "epoch": 2.11, + "learning_rate": 1.4903180302641703e-05, + "loss": 0.2534, + "step": 54787 + }, + { + "epoch": 2.11, + "learning_rate": 1.4902539112592973e-05, + "loss": 0.2185, + "step": 54788 + }, + { + "epoch": 2.11, + "learning_rate": 1.4901897922544241e-05, + "loss": 0.296, + "step": 54789 + }, + { + "epoch": 2.11, + "learning_rate": 1.4901256732495511e-05, + "loss": 0.1597, + "step": 54790 + }, + { + "epoch": 2.11, + "learning_rate": 1.4900615542446783e-05, + "loss": 0.2412, + "step": 54791 + }, + { + "epoch": 2.11, + "learning_rate": 1.4899974352398053e-05, + "loss": 0.1345, + "step": 54792 + }, + { + "epoch": 2.11, + "learning_rate": 1.4899333162349321e-05, + "loss": 0.2337, + "step": 54793 + }, + { + "epoch": 2.11, + "learning_rate": 1.4898691972300591e-05, + "loss": 0.2167, + "step": 54794 + }, + { + "epoch": 2.11, + "learning_rate": 1.489805078225186e-05, + "loss": 0.2777, + "step": 54795 + }, + { + "epoch": 2.11, + "learning_rate": 1.489740959220313e-05, + "loss": 0.1335, + "step": 54796 + }, + { + "epoch": 2.11, + "learning_rate": 1.4896768402154398e-05, + "loss": 0.1606, + "step": 54797 + }, + { + "epoch": 2.11, + "learning_rate": 1.4896127212105668e-05, + "loss": 0.2432, + "step": 54798 + }, + { + "epoch": 2.11, + "learning_rate": 1.489548602205694e-05, + "loss": 0.2204, + "step": 54799 + }, + { + "epoch": 2.11, + "learning_rate": 1.489484483200821e-05, + "loss": 0.2933, + "step": 54800 + }, + { + "epoch": 2.11, + "learning_rate": 1.4894203641959478e-05, + "loss": 0.3162, + "step": 54801 + }, + { + "epoch": 2.11, + "learning_rate": 1.4893562451910748e-05, + "loss": 0.1567, + "step": 54802 + }, + { + "epoch": 2.11, + "learning_rate": 1.4892921261862016e-05, + "loss": 0.3816, + "step": 54803 + }, + { + "epoch": 2.11, + "learning_rate": 1.4892280071813286e-05, + "loss": 0.2374, + "step": 54804 + }, + { + "epoch": 2.11, + "learning_rate": 1.4891638881764555e-05, + "loss": 0.4521, + "step": 54805 + }, + { + "epoch": 2.11, + "learning_rate": 1.4890997691715825e-05, + "loss": 0.1996, + "step": 54806 + }, + { + "epoch": 2.11, + "learning_rate": 1.4890356501667093e-05, + "loss": 0.2607, + "step": 54807 + }, + { + "epoch": 2.11, + "learning_rate": 1.4889715311618366e-05, + "loss": 0.172, + "step": 54808 + }, + { + "epoch": 2.11, + "learning_rate": 1.4889074121569635e-05, + "loss": 0.1208, + "step": 54809 + }, + { + "epoch": 2.11, + "learning_rate": 1.4888432931520905e-05, + "loss": 0.2545, + "step": 54810 + }, + { + "epoch": 2.11, + "learning_rate": 1.4887791741472173e-05, + "loss": 0.1814, + "step": 54811 + }, + { + "epoch": 2.11, + "learning_rate": 1.4887150551423443e-05, + "loss": 0.3304, + "step": 54812 + }, + { + "epoch": 2.11, + "learning_rate": 1.4886509361374711e-05, + "loss": 0.295, + "step": 54813 + }, + { + "epoch": 2.11, + "learning_rate": 1.4885868171325981e-05, + "loss": 0.1628, + "step": 54814 + }, + { + "epoch": 2.11, + "learning_rate": 1.488522698127725e-05, + "loss": 0.331, + "step": 54815 + }, + { + "epoch": 2.11, + "learning_rate": 1.488458579122852e-05, + "loss": 0.226, + "step": 54816 + }, + { + "epoch": 2.11, + "learning_rate": 1.4883944601179791e-05, + "loss": 0.2087, + "step": 54817 + }, + { + "epoch": 2.11, + "learning_rate": 1.4883303411131061e-05, + "loss": 0.1976, + "step": 54818 + }, + { + "epoch": 2.11, + "learning_rate": 1.488266222108233e-05, + "loss": 0.3726, + "step": 54819 + }, + { + "epoch": 2.11, + "learning_rate": 1.48820210310336e-05, + "loss": 0.1626, + "step": 54820 + }, + { + "epoch": 2.11, + "learning_rate": 1.4881379840984868e-05, + "loss": 0.1904, + "step": 54821 + }, + { + "epoch": 2.11, + "learning_rate": 1.4880738650936138e-05, + "loss": 0.194, + "step": 54822 + }, + { + "epoch": 2.11, + "learning_rate": 1.4880097460887407e-05, + "loss": 0.2369, + "step": 54823 + }, + { + "epoch": 2.11, + "learning_rate": 1.4879456270838677e-05, + "loss": 0.2488, + "step": 54824 + }, + { + "epoch": 2.11, + "learning_rate": 1.4878815080789945e-05, + "loss": 0.1384, + "step": 54825 + }, + { + "epoch": 2.11, + "learning_rate": 1.4878173890741218e-05, + "loss": 0.1824, + "step": 54826 + }, + { + "epoch": 2.11, + "learning_rate": 1.4877532700692487e-05, + "loss": 0.3351, + "step": 54827 + }, + { + "epoch": 2.11, + "learning_rate": 1.4876891510643757e-05, + "loss": 0.2513, + "step": 54828 + }, + { + "epoch": 2.11, + "learning_rate": 1.4876250320595025e-05, + "loss": 0.1855, + "step": 54829 + }, + { + "epoch": 2.11, + "learning_rate": 1.4875609130546295e-05, + "loss": 0.4908, + "step": 54830 + }, + { + "epoch": 2.11, + "learning_rate": 1.4874967940497563e-05, + "loss": 0.2305, + "step": 54831 + }, + { + "epoch": 2.11, + "learning_rate": 1.4874326750448833e-05, + "loss": 0.3204, + "step": 54832 + }, + { + "epoch": 2.11, + "learning_rate": 1.4873685560400102e-05, + "loss": 0.1298, + "step": 54833 + }, + { + "epoch": 2.11, + "learning_rate": 1.4873044370351372e-05, + "loss": 0.278, + "step": 54834 + }, + { + "epoch": 2.11, + "learning_rate": 1.4872403180302643e-05, + "loss": 0.1828, + "step": 54835 + }, + { + "epoch": 2.11, + "learning_rate": 1.4871761990253913e-05, + "loss": 0.2953, + "step": 54836 + }, + { + "epoch": 2.11, + "learning_rate": 1.4871120800205182e-05, + "loss": 0.2033, + "step": 54837 + }, + { + "epoch": 2.11, + "learning_rate": 1.4870479610156452e-05, + "loss": 0.1602, + "step": 54838 + }, + { + "epoch": 2.11, + "learning_rate": 1.486983842010772e-05, + "loss": 0.2849, + "step": 54839 + }, + { + "epoch": 2.11, + "learning_rate": 1.486919723005899e-05, + "loss": 0.1802, + "step": 54840 + }, + { + "epoch": 2.11, + "learning_rate": 1.4868556040010258e-05, + "loss": 0.2315, + "step": 54841 + }, + { + "epoch": 2.11, + "learning_rate": 1.4867914849961528e-05, + "loss": 0.269, + "step": 54842 + }, + { + "epoch": 2.11, + "learning_rate": 1.4867273659912797e-05, + "loss": 0.2321, + "step": 54843 + }, + { + "epoch": 2.11, + "learning_rate": 1.486663246986407e-05, + "loss": 0.4774, + "step": 54844 + }, + { + "epoch": 2.11, + "learning_rate": 1.4865991279815339e-05, + "loss": 0.3236, + "step": 54845 + }, + { + "epoch": 2.11, + "learning_rate": 1.4865350089766609e-05, + "loss": 0.1769, + "step": 54846 + }, + { + "epoch": 2.11, + "learning_rate": 1.4864708899717877e-05, + "loss": 0.1363, + "step": 54847 + }, + { + "epoch": 2.11, + "learning_rate": 1.4864067709669147e-05, + "loss": 0.2004, + "step": 54848 + }, + { + "epoch": 2.11, + "learning_rate": 1.4863426519620415e-05, + "loss": 0.2381, + "step": 54849 + }, + { + "epoch": 2.11, + "learning_rate": 1.4862785329571685e-05, + "loss": 0.1582, + "step": 54850 + }, + { + "epoch": 2.11, + "learning_rate": 1.4862144139522954e-05, + "loss": 0.2321, + "step": 54851 + }, + { + "epoch": 2.11, + "learning_rate": 1.4861502949474224e-05, + "loss": 0.2863, + "step": 54852 + }, + { + "epoch": 2.11, + "learning_rate": 1.4860861759425495e-05, + "loss": 0.1625, + "step": 54853 + }, + { + "epoch": 2.11, + "learning_rate": 1.4860220569376765e-05, + "loss": 0.169, + "step": 54854 + }, + { + "epoch": 2.11, + "learning_rate": 1.4859579379328034e-05, + "loss": 0.1183, + "step": 54855 + }, + { + "epoch": 2.11, + "learning_rate": 1.4858938189279304e-05, + "loss": 0.2718, + "step": 54856 + }, + { + "epoch": 2.11, + "learning_rate": 1.4858296999230572e-05, + "loss": 0.2441, + "step": 54857 + }, + { + "epoch": 2.11, + "learning_rate": 1.4857655809181842e-05, + "loss": 0.1568, + "step": 54858 + }, + { + "epoch": 2.11, + "learning_rate": 1.485701461913311e-05, + "loss": 0.2505, + "step": 54859 + }, + { + "epoch": 2.11, + "learning_rate": 1.485637342908438e-05, + "loss": 0.3783, + "step": 54860 + }, + { + "epoch": 2.11, + "learning_rate": 1.485573223903565e-05, + "loss": 0.2191, + "step": 54861 + }, + { + "epoch": 2.11, + "learning_rate": 1.4855091048986922e-05, + "loss": 0.3009, + "step": 54862 + }, + { + "epoch": 2.11, + "learning_rate": 1.485444985893819e-05, + "loss": 0.1966, + "step": 54863 + }, + { + "epoch": 2.11, + "learning_rate": 1.485380866888946e-05, + "loss": 0.2916, + "step": 54864 + }, + { + "epoch": 2.11, + "learning_rate": 1.4853167478840729e-05, + "loss": 0.2302, + "step": 54865 + }, + { + "epoch": 2.11, + "learning_rate": 1.4852526288791999e-05, + "loss": 0.2934, + "step": 54866 + }, + { + "epoch": 2.11, + "learning_rate": 1.4851885098743267e-05, + "loss": 0.2133, + "step": 54867 + }, + { + "epoch": 2.11, + "learning_rate": 1.4851243908694537e-05, + "loss": 0.2467, + "step": 54868 + }, + { + "epoch": 2.11, + "learning_rate": 1.4850602718645806e-05, + "loss": 0.2386, + "step": 54869 + }, + { + "epoch": 2.11, + "learning_rate": 1.4849961528597076e-05, + "loss": 0.2254, + "step": 54870 + }, + { + "epoch": 2.11, + "learning_rate": 1.4849320338548347e-05, + "loss": 0.3067, + "step": 54871 + }, + { + "epoch": 2.11, + "learning_rate": 1.4848679148499617e-05, + "loss": 0.1352, + "step": 54872 + }, + { + "epoch": 2.11, + "learning_rate": 1.4848037958450886e-05, + "loss": 0.257, + "step": 54873 + }, + { + "epoch": 2.11, + "learning_rate": 1.4847396768402156e-05, + "loss": 0.3273, + "step": 54874 + }, + { + "epoch": 2.11, + "learning_rate": 1.4846755578353424e-05, + "loss": 0.3079, + "step": 54875 + }, + { + "epoch": 2.11, + "learning_rate": 1.4846114388304694e-05, + "loss": 0.2033, + "step": 54876 + }, + { + "epoch": 2.11, + "learning_rate": 1.4845473198255962e-05, + "loss": 0.1977, + "step": 54877 + }, + { + "epoch": 2.11, + "learning_rate": 1.4844832008207232e-05, + "loss": 0.1837, + "step": 54878 + }, + { + "epoch": 2.11, + "learning_rate": 1.4844190818158502e-05, + "loss": 0.2416, + "step": 54879 + }, + { + "epoch": 2.11, + "learning_rate": 1.4843549628109774e-05, + "loss": 0.2414, + "step": 54880 + }, + { + "epoch": 2.11, + "learning_rate": 1.4842908438061042e-05, + "loss": 0.2567, + "step": 54881 + }, + { + "epoch": 2.11, + "learning_rate": 1.4842267248012312e-05, + "loss": 0.2446, + "step": 54882 + }, + { + "epoch": 2.11, + "learning_rate": 1.484162605796358e-05, + "loss": 0.1679, + "step": 54883 + }, + { + "epoch": 2.11, + "learning_rate": 1.484098486791485e-05, + "loss": 0.3111, + "step": 54884 + }, + { + "epoch": 2.11, + "learning_rate": 1.4840343677866119e-05, + "loss": 0.3269, + "step": 54885 + }, + { + "epoch": 2.11, + "learning_rate": 1.4839702487817389e-05, + "loss": 0.2842, + "step": 54886 + }, + { + "epoch": 2.11, + "learning_rate": 1.4839061297768657e-05, + "loss": 0.2207, + "step": 54887 + }, + { + "epoch": 2.11, + "learning_rate": 1.4838420107719927e-05, + "loss": 0.3116, + "step": 54888 + }, + { + "epoch": 2.11, + "learning_rate": 1.48377789176712e-05, + "loss": 0.3507, + "step": 54889 + }, + { + "epoch": 2.11, + "learning_rate": 1.483713772762247e-05, + "loss": 0.3202, + "step": 54890 + }, + { + "epoch": 2.11, + "learning_rate": 1.4836496537573738e-05, + "loss": 0.1374, + "step": 54891 + }, + { + "epoch": 2.11, + "learning_rate": 1.4835855347525008e-05, + "loss": 0.1662, + "step": 54892 + }, + { + "epoch": 2.11, + "learning_rate": 1.4835214157476276e-05, + "loss": 0.2343, + "step": 54893 + }, + { + "epoch": 2.11, + "learning_rate": 1.4834572967427546e-05, + "loss": 0.2148, + "step": 54894 + }, + { + "epoch": 2.11, + "learning_rate": 1.4833931777378814e-05, + "loss": 0.2323, + "step": 54895 + }, + { + "epoch": 2.11, + "learning_rate": 1.4833290587330084e-05, + "loss": 0.3527, + "step": 54896 + }, + { + "epoch": 2.11, + "learning_rate": 1.4832649397281356e-05, + "loss": 0.2748, + "step": 54897 + }, + { + "epoch": 2.11, + "learning_rate": 1.4832008207232626e-05, + "loss": 0.2534, + "step": 54898 + }, + { + "epoch": 2.11, + "learning_rate": 1.4831367017183894e-05, + "loss": 0.1878, + "step": 54899 + }, + { + "epoch": 2.11, + "learning_rate": 1.4830725827135164e-05, + "loss": 0.2801, + "step": 54900 + }, + { + "epoch": 2.11, + "learning_rate": 1.4830084637086433e-05, + "loss": 0.3468, + "step": 54901 + }, + { + "epoch": 2.11, + "learning_rate": 1.4829443447037703e-05, + "loss": 0.2199, + "step": 54902 + }, + { + "epoch": 2.11, + "learning_rate": 1.4828802256988971e-05, + "loss": 0.1783, + "step": 54903 + }, + { + "epoch": 2.11, + "learning_rate": 1.4828161066940241e-05, + "loss": 0.2479, + "step": 54904 + }, + { + "epoch": 2.11, + "learning_rate": 1.482751987689151e-05, + "loss": 0.1891, + "step": 54905 + }, + { + "epoch": 2.11, + "learning_rate": 1.4826878686842783e-05, + "loss": 0.268, + "step": 54906 + }, + { + "epoch": 2.11, + "learning_rate": 1.4826237496794051e-05, + "loss": 0.2415, + "step": 54907 + }, + { + "epoch": 2.11, + "learning_rate": 1.4825596306745321e-05, + "loss": 0.2558, + "step": 54908 + }, + { + "epoch": 2.11, + "learning_rate": 1.482495511669659e-05, + "loss": 0.2191, + "step": 54909 + }, + { + "epoch": 2.11, + "learning_rate": 1.482431392664786e-05, + "loss": 0.2477, + "step": 54910 + }, + { + "epoch": 2.11, + "learning_rate": 1.4823672736599128e-05, + "loss": 0.2289, + "step": 54911 + }, + { + "epoch": 2.11, + "learning_rate": 1.4823031546550398e-05, + "loss": 0.2055, + "step": 54912 + }, + { + "epoch": 2.11, + "learning_rate": 1.4822390356501666e-05, + "loss": 0.3537, + "step": 54913 + }, + { + "epoch": 2.11, + "learning_rate": 1.4821749166452936e-05, + "loss": 0.2728, + "step": 54914 + }, + { + "epoch": 2.11, + "learning_rate": 1.4821107976404208e-05, + "loss": 0.3087, + "step": 54915 + }, + { + "epoch": 2.11, + "learning_rate": 1.4820466786355478e-05, + "loss": 0.327, + "step": 54916 + }, + { + "epoch": 2.11, + "learning_rate": 1.4819825596306746e-05, + "loss": 0.2802, + "step": 54917 + }, + { + "epoch": 2.11, + "learning_rate": 1.4819184406258016e-05, + "loss": 0.1955, + "step": 54918 + }, + { + "epoch": 2.11, + "learning_rate": 1.4818543216209285e-05, + "loss": 0.2135, + "step": 54919 + }, + { + "epoch": 2.11, + "learning_rate": 1.4817902026160555e-05, + "loss": 0.1945, + "step": 54920 + }, + { + "epoch": 2.11, + "learning_rate": 1.4817260836111823e-05, + "loss": 0.1904, + "step": 54921 + }, + { + "epoch": 2.11, + "learning_rate": 1.4816619646063093e-05, + "loss": 0.3102, + "step": 54922 + }, + { + "epoch": 2.11, + "learning_rate": 1.4815978456014363e-05, + "loss": 0.256, + "step": 54923 + }, + { + "epoch": 2.11, + "learning_rate": 1.4815337265965635e-05, + "loss": 0.1732, + "step": 54924 + }, + { + "epoch": 2.11, + "learning_rate": 1.4814696075916903e-05, + "loss": 0.3075, + "step": 54925 + }, + { + "epoch": 2.11, + "learning_rate": 1.4814054885868173e-05, + "loss": 0.1627, + "step": 54926 + }, + { + "epoch": 2.11, + "learning_rate": 1.4813413695819441e-05, + "loss": 0.2022, + "step": 54927 + }, + { + "epoch": 2.11, + "learning_rate": 1.4812772505770711e-05, + "loss": 0.1325, + "step": 54928 + }, + { + "epoch": 2.11, + "learning_rate": 1.481213131572198e-05, + "loss": 0.3169, + "step": 54929 + }, + { + "epoch": 2.11, + "learning_rate": 1.481149012567325e-05, + "loss": 0.2274, + "step": 54930 + }, + { + "epoch": 2.11, + "learning_rate": 1.4810848935624518e-05, + "loss": 0.1835, + "step": 54931 + }, + { + "epoch": 2.11, + "learning_rate": 1.4810207745575788e-05, + "loss": 0.1833, + "step": 54932 + }, + { + "epoch": 2.11, + "learning_rate": 1.480956655552706e-05, + "loss": 0.2328, + "step": 54933 + }, + { + "epoch": 2.11, + "learning_rate": 1.480892536547833e-05, + "loss": 0.257, + "step": 54934 + }, + { + "epoch": 2.11, + "learning_rate": 1.4808284175429598e-05, + "loss": 0.1936, + "step": 54935 + }, + { + "epoch": 2.11, + "learning_rate": 1.4807642985380868e-05, + "loss": 0.3087, + "step": 54936 + }, + { + "epoch": 2.11, + "learning_rate": 1.4807001795332137e-05, + "loss": 0.1936, + "step": 54937 + }, + { + "epoch": 2.11, + "learning_rate": 1.4806360605283407e-05, + "loss": 0.1396, + "step": 54938 + }, + { + "epoch": 2.11, + "learning_rate": 1.4805719415234675e-05, + "loss": 0.329, + "step": 54939 + }, + { + "epoch": 2.11, + "learning_rate": 1.4805078225185945e-05, + "loss": 0.1953, + "step": 54940 + }, + { + "epoch": 2.11, + "learning_rate": 1.4804437035137215e-05, + "loss": 0.2161, + "step": 54941 + }, + { + "epoch": 2.11, + "learning_rate": 1.4803795845088487e-05, + "loss": 0.2778, + "step": 54942 + }, + { + "epoch": 2.11, + "learning_rate": 1.4803154655039755e-05, + "loss": 0.3391, + "step": 54943 + }, + { + "epoch": 2.11, + "learning_rate": 1.4802513464991025e-05, + "loss": 0.2829, + "step": 54944 + }, + { + "epoch": 2.11, + "learning_rate": 1.4801872274942293e-05, + "loss": 0.2686, + "step": 54945 + }, + { + "epoch": 2.11, + "learning_rate": 1.4801231084893563e-05, + "loss": 0.1585, + "step": 54946 + }, + { + "epoch": 2.11, + "learning_rate": 1.4800589894844832e-05, + "loss": 0.2916, + "step": 54947 + }, + { + "epoch": 2.11, + "learning_rate": 1.4799948704796102e-05, + "loss": 0.2588, + "step": 54948 + }, + { + "epoch": 2.11, + "learning_rate": 1.479930751474737e-05, + "loss": 0.1983, + "step": 54949 + }, + { + "epoch": 2.11, + "learning_rate": 1.479866632469864e-05, + "loss": 0.4429, + "step": 54950 + }, + { + "epoch": 2.11, + "learning_rate": 1.4798025134649912e-05, + "loss": 0.333, + "step": 54951 + }, + { + "epoch": 2.11, + "learning_rate": 1.4797383944601182e-05, + "loss": 0.2167, + "step": 54952 + }, + { + "epoch": 2.11, + "learning_rate": 1.479674275455245e-05, + "loss": 0.3098, + "step": 54953 + }, + { + "epoch": 2.11, + "learning_rate": 1.479610156450372e-05, + "loss": 0.2981, + "step": 54954 + }, + { + "epoch": 2.11, + "learning_rate": 1.4795460374454988e-05, + "loss": 0.2915, + "step": 54955 + }, + { + "epoch": 2.11, + "learning_rate": 1.4794819184406259e-05, + "loss": 0.2099, + "step": 54956 + }, + { + "epoch": 2.11, + "learning_rate": 1.4794177994357527e-05, + "loss": 0.2978, + "step": 54957 + }, + { + "epoch": 2.11, + "learning_rate": 1.4793536804308797e-05, + "loss": 0.2687, + "step": 54958 + }, + { + "epoch": 2.11, + "learning_rate": 1.4792895614260067e-05, + "loss": 0.1644, + "step": 54959 + }, + { + "epoch": 2.11, + "learning_rate": 1.4792254424211339e-05, + "loss": 0.2955, + "step": 54960 + }, + { + "epoch": 2.11, + "learning_rate": 1.4791613234162607e-05, + "loss": 0.1781, + "step": 54961 + }, + { + "epoch": 2.11, + "learning_rate": 1.4790972044113877e-05, + "loss": 0.175, + "step": 54962 + }, + { + "epoch": 2.11, + "learning_rate": 1.4790330854065145e-05, + "loss": 0.1284, + "step": 54963 + }, + { + "epoch": 2.11, + "learning_rate": 1.4789689664016415e-05, + "loss": 0.3726, + "step": 54964 + }, + { + "epoch": 2.11, + "learning_rate": 1.4789048473967684e-05, + "loss": 0.2073, + "step": 54965 + }, + { + "epoch": 2.11, + "learning_rate": 1.4788407283918954e-05, + "loss": 0.2131, + "step": 54966 + }, + { + "epoch": 2.11, + "learning_rate": 1.4787766093870222e-05, + "loss": 0.2027, + "step": 54967 + }, + { + "epoch": 2.11, + "learning_rate": 1.4787124903821492e-05, + "loss": 0.2428, + "step": 54968 + }, + { + "epoch": 2.11, + "learning_rate": 1.4786483713772764e-05, + "loss": 0.2679, + "step": 54969 + }, + { + "epoch": 2.11, + "learning_rate": 1.4785842523724034e-05, + "loss": 0.1811, + "step": 54970 + }, + { + "epoch": 2.11, + "learning_rate": 1.4785201333675302e-05, + "loss": 0.3, + "step": 54971 + }, + { + "epoch": 2.11, + "learning_rate": 1.4784560143626572e-05, + "loss": 0.3036, + "step": 54972 + }, + { + "epoch": 2.11, + "learning_rate": 1.478391895357784e-05, + "loss": 0.2135, + "step": 54973 + }, + { + "epoch": 2.11, + "learning_rate": 1.478327776352911e-05, + "loss": 0.219, + "step": 54974 + }, + { + "epoch": 2.11, + "learning_rate": 1.4782636573480379e-05, + "loss": 0.2772, + "step": 54975 + }, + { + "epoch": 2.11, + "learning_rate": 1.4781995383431649e-05, + "loss": 0.1876, + "step": 54976 + }, + { + "epoch": 2.11, + "learning_rate": 1.4781354193382919e-05, + "loss": 0.4243, + "step": 54977 + }, + { + "epoch": 2.11, + "learning_rate": 1.478071300333419e-05, + "loss": 0.1139, + "step": 54978 + }, + { + "epoch": 2.11, + "learning_rate": 1.4780071813285459e-05, + "loss": 0.3012, + "step": 54979 + }, + { + "epoch": 2.11, + "learning_rate": 1.4779430623236729e-05, + "loss": 0.2998, + "step": 54980 + }, + { + "epoch": 2.11, + "learning_rate": 1.4778789433187997e-05, + "loss": 0.1966, + "step": 54981 + }, + { + "epoch": 2.11, + "learning_rate": 1.4778148243139267e-05, + "loss": 0.1337, + "step": 54982 + }, + { + "epoch": 2.11, + "learning_rate": 1.4777507053090536e-05, + "loss": 0.2174, + "step": 54983 + }, + { + "epoch": 2.11, + "learning_rate": 1.4776865863041806e-05, + "loss": 0.2514, + "step": 54984 + }, + { + "epoch": 2.11, + "learning_rate": 1.4776224672993076e-05, + "loss": 0.2177, + "step": 54985 + }, + { + "epoch": 2.11, + "learning_rate": 1.4775583482944344e-05, + "loss": 0.1559, + "step": 54986 + }, + { + "epoch": 2.11, + "learning_rate": 1.4774942292895616e-05, + "loss": 0.2229, + "step": 54987 + }, + { + "epoch": 2.11, + "learning_rate": 1.4774301102846886e-05, + "loss": 0.1847, + "step": 54988 + }, + { + "epoch": 2.11, + "learning_rate": 1.4773659912798154e-05, + "loss": 0.2697, + "step": 54989 + }, + { + "epoch": 2.11, + "learning_rate": 1.4773018722749424e-05, + "loss": 0.3123, + "step": 54990 + }, + { + "epoch": 2.11, + "learning_rate": 1.4772377532700692e-05, + "loss": 0.3111, + "step": 54991 + }, + { + "epoch": 2.11, + "learning_rate": 1.4771736342651962e-05, + "loss": 0.3253, + "step": 54992 + }, + { + "epoch": 2.11, + "learning_rate": 1.477109515260323e-05, + "loss": 0.2431, + "step": 54993 + }, + { + "epoch": 2.11, + "learning_rate": 1.47704539625545e-05, + "loss": 0.2951, + "step": 54994 + }, + { + "epoch": 2.11, + "learning_rate": 1.4769812772505772e-05, + "loss": 0.3178, + "step": 54995 + }, + { + "epoch": 2.11, + "learning_rate": 1.4769171582457042e-05, + "loss": 0.2093, + "step": 54996 + }, + { + "epoch": 2.11, + "learning_rate": 1.476853039240831e-05, + "loss": 0.1771, + "step": 54997 + }, + { + "epoch": 2.11, + "learning_rate": 1.476788920235958e-05, + "loss": 0.2195, + "step": 54998 + }, + { + "epoch": 2.11, + "learning_rate": 1.476724801231085e-05, + "loss": 0.22, + "step": 54999 + }, + { + "epoch": 2.11, + "learning_rate": 1.476660682226212e-05, + "loss": 0.2928, + "step": 55000 + }, + { + "epoch": 2.11, + "learning_rate": 1.4765965632213388e-05, + "loss": 0.2271, + "step": 55001 + }, + { + "epoch": 2.11, + "learning_rate": 1.4765324442164658e-05, + "loss": 0.2023, + "step": 55002 + }, + { + "epoch": 2.11, + "learning_rate": 1.4764683252115928e-05, + "loss": 0.1719, + "step": 55003 + }, + { + "epoch": 2.11, + "learning_rate": 1.47640420620672e-05, + "loss": 0.2376, + "step": 55004 + }, + { + "epoch": 2.11, + "learning_rate": 1.4763400872018468e-05, + "loss": 0.1439, + "step": 55005 + }, + { + "epoch": 2.11, + "learning_rate": 1.4762759681969738e-05, + "loss": 0.1685, + "step": 55006 + }, + { + "epoch": 2.11, + "learning_rate": 1.4762118491921006e-05, + "loss": 0.2391, + "step": 55007 + }, + { + "epoch": 2.11, + "learning_rate": 1.4761477301872276e-05, + "loss": 0.2789, + "step": 55008 + }, + { + "epoch": 2.11, + "learning_rate": 1.4760836111823544e-05, + "loss": 0.2572, + "step": 55009 + }, + { + "epoch": 2.11, + "learning_rate": 1.4760194921774814e-05, + "loss": 0.2872, + "step": 55010 + }, + { + "epoch": 2.11, + "learning_rate": 1.4759553731726083e-05, + "loss": 0.1638, + "step": 55011 + }, + { + "epoch": 2.12, + "learning_rate": 1.4758912541677353e-05, + "loss": 0.2397, + "step": 55012 + }, + { + "epoch": 2.12, + "learning_rate": 1.4758271351628624e-05, + "loss": 0.3155, + "step": 55013 + }, + { + "epoch": 2.12, + "learning_rate": 1.4757630161579894e-05, + "loss": 0.229, + "step": 55014 + }, + { + "epoch": 2.12, + "learning_rate": 1.4756988971531163e-05, + "loss": 0.1401, + "step": 55015 + }, + { + "epoch": 2.12, + "learning_rate": 1.4756347781482433e-05, + "loss": 0.1465, + "step": 55016 + }, + { + "epoch": 2.12, + "learning_rate": 1.4755706591433701e-05, + "loss": 0.1365, + "step": 55017 + }, + { + "epoch": 2.12, + "learning_rate": 1.4755065401384971e-05, + "loss": 0.2069, + "step": 55018 + }, + { + "epoch": 2.12, + "learning_rate": 1.475442421133624e-05, + "loss": 0.259, + "step": 55019 + }, + { + "epoch": 2.12, + "learning_rate": 1.475378302128751e-05, + "loss": 0.3553, + "step": 55020 + }, + { + "epoch": 2.12, + "learning_rate": 1.475314183123878e-05, + "loss": 0.1966, + "step": 55021 + }, + { + "epoch": 2.12, + "learning_rate": 1.4752500641190051e-05, + "loss": 0.2649, + "step": 55022 + }, + { + "epoch": 2.12, + "learning_rate": 1.475185945114132e-05, + "loss": 0.2478, + "step": 55023 + }, + { + "epoch": 2.12, + "learning_rate": 1.475121826109259e-05, + "loss": 0.1712, + "step": 55024 + }, + { + "epoch": 2.12, + "learning_rate": 1.4750577071043858e-05, + "loss": 0.1982, + "step": 55025 + }, + { + "epoch": 2.12, + "learning_rate": 1.4749935880995128e-05, + "loss": 0.1832, + "step": 55026 + }, + { + "epoch": 2.12, + "learning_rate": 1.4749294690946396e-05, + "loss": 0.2479, + "step": 55027 + }, + { + "epoch": 2.12, + "learning_rate": 1.4748653500897666e-05, + "loss": 0.1502, + "step": 55028 + }, + { + "epoch": 2.12, + "learning_rate": 1.4748012310848935e-05, + "loss": 0.2447, + "step": 55029 + }, + { + "epoch": 2.12, + "learning_rate": 1.4747371120800205e-05, + "loss": 0.4281, + "step": 55030 + }, + { + "epoch": 2.12, + "learning_rate": 1.4746729930751476e-05, + "loss": 0.2194, + "step": 55031 + }, + { + "epoch": 2.12, + "learning_rate": 1.4746088740702746e-05, + "loss": 0.1676, + "step": 55032 + }, + { + "epoch": 2.12, + "learning_rate": 1.4745447550654015e-05, + "loss": 0.178, + "step": 55033 + }, + { + "epoch": 2.12, + "learning_rate": 1.4744806360605285e-05, + "loss": 0.1471, + "step": 55034 + }, + { + "epoch": 2.12, + "learning_rate": 1.4744165170556553e-05, + "loss": 0.195, + "step": 55035 + }, + { + "epoch": 2.12, + "learning_rate": 1.4743523980507823e-05, + "loss": 0.158, + "step": 55036 + }, + { + "epoch": 2.12, + "learning_rate": 1.4742882790459091e-05, + "loss": 0.2179, + "step": 55037 + }, + { + "epoch": 2.12, + "learning_rate": 1.4742241600410361e-05, + "loss": 0.1946, + "step": 55038 + }, + { + "epoch": 2.12, + "learning_rate": 1.4741600410361631e-05, + "loss": 0.3484, + "step": 55039 + }, + { + "epoch": 2.12, + "learning_rate": 1.4740959220312903e-05, + "loss": 0.0982, + "step": 55040 + }, + { + "epoch": 2.12, + "learning_rate": 1.4740318030264171e-05, + "loss": 0.242, + "step": 55041 + }, + { + "epoch": 2.12, + "learning_rate": 1.4739676840215441e-05, + "loss": 0.2145, + "step": 55042 + }, + { + "epoch": 2.12, + "learning_rate": 1.473903565016671e-05, + "loss": 0.2918, + "step": 55043 + }, + { + "epoch": 2.12, + "learning_rate": 1.473839446011798e-05, + "loss": 0.2072, + "step": 55044 + }, + { + "epoch": 2.12, + "learning_rate": 1.4737753270069248e-05, + "loss": 0.2034, + "step": 55045 + }, + { + "epoch": 2.12, + "learning_rate": 1.4737112080020518e-05, + "loss": 0.2482, + "step": 55046 + }, + { + "epoch": 2.12, + "learning_rate": 1.4736470889971788e-05, + "loss": 0.2583, + "step": 55047 + }, + { + "epoch": 2.12, + "learning_rate": 1.4735829699923057e-05, + "loss": 0.2649, + "step": 55048 + }, + { + "epoch": 2.12, + "learning_rate": 1.4735188509874328e-05, + "loss": 0.2197, + "step": 55049 + }, + { + "epoch": 2.12, + "learning_rate": 1.4734547319825598e-05, + "loss": 0.3503, + "step": 55050 + }, + { + "epoch": 2.12, + "learning_rate": 1.4733906129776867e-05, + "loss": 0.2762, + "step": 55051 + }, + { + "epoch": 2.12, + "learning_rate": 1.4733264939728137e-05, + "loss": 0.2697, + "step": 55052 + }, + { + "epoch": 2.12, + "learning_rate": 1.4732623749679405e-05, + "loss": 0.2658, + "step": 55053 + }, + { + "epoch": 2.12, + "learning_rate": 1.4731982559630675e-05, + "loss": 0.1916, + "step": 55054 + }, + { + "epoch": 2.12, + "learning_rate": 1.4731341369581943e-05, + "loss": 0.1339, + "step": 55055 + }, + { + "epoch": 2.12, + "learning_rate": 1.4730700179533213e-05, + "loss": 0.2058, + "step": 55056 + }, + { + "epoch": 2.12, + "learning_rate": 1.4730058989484483e-05, + "loss": 0.1309, + "step": 55057 + }, + { + "epoch": 2.12, + "learning_rate": 1.4729417799435755e-05, + "loss": 0.2286, + "step": 55058 + }, + { + "epoch": 2.12, + "learning_rate": 1.4728776609387023e-05, + "loss": 0.2186, + "step": 55059 + }, + { + "epoch": 2.12, + "learning_rate": 1.4728135419338293e-05, + "loss": 0.2051, + "step": 55060 + }, + { + "epoch": 2.12, + "learning_rate": 1.4727494229289562e-05, + "loss": 0.2608, + "step": 55061 + }, + { + "epoch": 2.12, + "learning_rate": 1.4726853039240832e-05, + "loss": 0.2783, + "step": 55062 + }, + { + "epoch": 2.12, + "learning_rate": 1.47262118491921e-05, + "loss": 0.2281, + "step": 55063 + }, + { + "epoch": 2.12, + "learning_rate": 1.472557065914337e-05, + "loss": 0.2473, + "step": 55064 + }, + { + "epoch": 2.12, + "learning_rate": 1.472492946909464e-05, + "loss": 0.3222, + "step": 55065 + }, + { + "epoch": 2.12, + "learning_rate": 1.4724288279045908e-05, + "loss": 0.2226, + "step": 55066 + }, + { + "epoch": 2.12, + "learning_rate": 1.472364708899718e-05, + "loss": 0.2366, + "step": 55067 + }, + { + "epoch": 2.12, + "learning_rate": 1.472300589894845e-05, + "loss": 0.1509, + "step": 55068 + }, + { + "epoch": 2.12, + "learning_rate": 1.4722364708899719e-05, + "loss": 0.2135, + "step": 55069 + }, + { + "epoch": 2.12, + "learning_rate": 1.4721723518850989e-05, + "loss": 0.378, + "step": 55070 + }, + { + "epoch": 2.12, + "learning_rate": 1.4721082328802257e-05, + "loss": 0.2773, + "step": 55071 + }, + { + "epoch": 2.12, + "learning_rate": 1.4720441138753527e-05, + "loss": 0.3432, + "step": 55072 + }, + { + "epoch": 2.12, + "learning_rate": 1.4719799948704795e-05, + "loss": 0.1462, + "step": 55073 + }, + { + "epoch": 2.12, + "learning_rate": 1.4719158758656065e-05, + "loss": 0.1892, + "step": 55074 + }, + { + "epoch": 2.12, + "learning_rate": 1.4718517568607335e-05, + "loss": 0.3943, + "step": 55075 + }, + { + "epoch": 2.12, + "learning_rate": 1.4717876378558607e-05, + "loss": 0.3237, + "step": 55076 + }, + { + "epoch": 2.12, + "learning_rate": 1.4717235188509875e-05, + "loss": 0.3044, + "step": 55077 + }, + { + "epoch": 2.12, + "learning_rate": 1.4716593998461145e-05, + "loss": 0.1115, + "step": 55078 + }, + { + "epoch": 2.12, + "learning_rate": 1.4715952808412414e-05, + "loss": 0.1567, + "step": 55079 + }, + { + "epoch": 2.12, + "learning_rate": 1.4715311618363684e-05, + "loss": 0.3311, + "step": 55080 + }, + { + "epoch": 2.12, + "learning_rate": 1.4714670428314952e-05, + "loss": 0.1639, + "step": 55081 + }, + { + "epoch": 2.12, + "learning_rate": 1.4714029238266222e-05, + "loss": 0.221, + "step": 55082 + }, + { + "epoch": 2.12, + "learning_rate": 1.4713388048217492e-05, + "loss": 0.2664, + "step": 55083 + }, + { + "epoch": 2.12, + "learning_rate": 1.4712746858168764e-05, + "loss": 0.3436, + "step": 55084 + }, + { + "epoch": 2.12, + "learning_rate": 1.4712105668120032e-05, + "loss": 0.163, + "step": 55085 + }, + { + "epoch": 2.12, + "learning_rate": 1.4711464478071302e-05, + "loss": 0.2823, + "step": 55086 + }, + { + "epoch": 2.12, + "learning_rate": 1.471082328802257e-05, + "loss": 0.1153, + "step": 55087 + }, + { + "epoch": 2.12, + "learning_rate": 1.471018209797384e-05, + "loss": 0.2216, + "step": 55088 + }, + { + "epoch": 2.12, + "learning_rate": 1.4709540907925109e-05, + "loss": 0.2376, + "step": 55089 + }, + { + "epoch": 2.12, + "learning_rate": 1.4708899717876379e-05, + "loss": 0.2854, + "step": 55090 + }, + { + "epoch": 2.12, + "learning_rate": 1.4708258527827647e-05, + "loss": 0.214, + "step": 55091 + }, + { + "epoch": 2.12, + "learning_rate": 1.4707617337778917e-05, + "loss": 0.2631, + "step": 55092 + }, + { + "epoch": 2.12, + "learning_rate": 1.4706976147730189e-05, + "loss": 0.1765, + "step": 55093 + }, + { + "epoch": 2.12, + "learning_rate": 1.4706334957681459e-05, + "loss": 0.1983, + "step": 55094 + }, + { + "epoch": 2.12, + "learning_rate": 1.4705693767632727e-05, + "loss": 0.2469, + "step": 55095 + }, + { + "epoch": 2.12, + "learning_rate": 1.4705052577583997e-05, + "loss": 0.2132, + "step": 55096 + }, + { + "epoch": 2.12, + "learning_rate": 1.4704411387535266e-05, + "loss": 0.2133, + "step": 55097 + }, + { + "epoch": 2.12, + "learning_rate": 1.4703770197486536e-05, + "loss": 0.1844, + "step": 55098 + }, + { + "epoch": 2.12, + "learning_rate": 1.4703129007437804e-05, + "loss": 0.2766, + "step": 55099 + }, + { + "epoch": 2.12, + "learning_rate": 1.4702487817389074e-05, + "loss": 0.2406, + "step": 55100 + }, + { + "epoch": 2.12, + "learning_rate": 1.4701846627340344e-05, + "loss": 0.2618, + "step": 55101 + }, + { + "epoch": 2.12, + "learning_rate": 1.4701205437291616e-05, + "loss": 0.2944, + "step": 55102 + }, + { + "epoch": 2.12, + "learning_rate": 1.4700564247242884e-05, + "loss": 0.1758, + "step": 55103 + }, + { + "epoch": 2.12, + "learning_rate": 1.4699923057194154e-05, + "loss": 0.3173, + "step": 55104 + }, + { + "epoch": 2.12, + "learning_rate": 1.4699281867145422e-05, + "loss": 0.2725, + "step": 55105 + }, + { + "epoch": 2.12, + "learning_rate": 1.4698640677096692e-05, + "loss": 0.2816, + "step": 55106 + }, + { + "epoch": 2.12, + "learning_rate": 1.469799948704796e-05, + "loss": 0.3456, + "step": 55107 + }, + { + "epoch": 2.12, + "learning_rate": 1.469735829699923e-05, + "loss": 0.2555, + "step": 55108 + }, + { + "epoch": 2.12, + "learning_rate": 1.4696717106950499e-05, + "loss": 0.1837, + "step": 55109 + }, + { + "epoch": 2.12, + "learning_rate": 1.4696075916901769e-05, + "loss": 0.2476, + "step": 55110 + }, + { + "epoch": 2.12, + "learning_rate": 1.4695434726853041e-05, + "loss": 0.1498, + "step": 55111 + }, + { + "epoch": 2.12, + "learning_rate": 1.4694793536804311e-05, + "loss": 0.2807, + "step": 55112 + }, + { + "epoch": 2.12, + "learning_rate": 1.469415234675558e-05, + "loss": 0.2821, + "step": 55113 + }, + { + "epoch": 2.12, + "learning_rate": 1.469351115670685e-05, + "loss": 0.1587, + "step": 55114 + }, + { + "epoch": 2.12, + "learning_rate": 1.4692869966658118e-05, + "loss": 0.3088, + "step": 55115 + }, + { + "epoch": 2.12, + "learning_rate": 1.4692228776609388e-05, + "loss": 0.1778, + "step": 55116 + }, + { + "epoch": 2.12, + "learning_rate": 1.4691587586560656e-05, + "loss": 0.1412, + "step": 55117 + }, + { + "epoch": 2.12, + "learning_rate": 1.4690946396511926e-05, + "loss": 0.2599, + "step": 55118 + }, + { + "epoch": 2.12, + "learning_rate": 1.4690305206463196e-05, + "loss": 0.2074, + "step": 55119 + }, + { + "epoch": 2.12, + "learning_rate": 1.4689664016414468e-05, + "loss": 0.4586, + "step": 55120 + }, + { + "epoch": 2.12, + "learning_rate": 1.4689022826365736e-05, + "loss": 0.1406, + "step": 55121 + }, + { + "epoch": 2.12, + "learning_rate": 1.4688381636317006e-05, + "loss": 0.2413, + "step": 55122 + }, + { + "epoch": 2.12, + "learning_rate": 1.4687740446268274e-05, + "loss": 0.2393, + "step": 55123 + }, + { + "epoch": 2.12, + "learning_rate": 1.4687099256219544e-05, + "loss": 0.1662, + "step": 55124 + }, + { + "epoch": 2.12, + "learning_rate": 1.4686458066170813e-05, + "loss": 0.1785, + "step": 55125 + }, + { + "epoch": 2.12, + "learning_rate": 1.4685816876122083e-05, + "loss": 0.1584, + "step": 55126 + }, + { + "epoch": 2.12, + "learning_rate": 1.4685175686073353e-05, + "loss": 0.2747, + "step": 55127 + }, + { + "epoch": 2.12, + "learning_rate": 1.4684534496024621e-05, + "loss": 0.2174, + "step": 55128 + }, + { + "epoch": 2.12, + "learning_rate": 1.4683893305975893e-05, + "loss": 0.316, + "step": 55129 + }, + { + "epoch": 2.12, + "learning_rate": 1.4683252115927163e-05, + "loss": 0.2133, + "step": 55130 + }, + { + "epoch": 2.12, + "learning_rate": 1.4682610925878431e-05, + "loss": 0.2606, + "step": 55131 + }, + { + "epoch": 2.12, + "learning_rate": 1.4681969735829701e-05, + "loss": 0.2761, + "step": 55132 + }, + { + "epoch": 2.12, + "learning_rate": 1.468132854578097e-05, + "loss": 0.2753, + "step": 55133 + }, + { + "epoch": 2.12, + "learning_rate": 1.468068735573224e-05, + "loss": 0.15, + "step": 55134 + }, + { + "epoch": 2.12, + "learning_rate": 1.4680046165683508e-05, + "loss": 0.2779, + "step": 55135 + }, + { + "epoch": 2.12, + "learning_rate": 1.4679404975634778e-05, + "loss": 0.2595, + "step": 55136 + }, + { + "epoch": 2.12, + "learning_rate": 1.4678763785586048e-05, + "loss": 0.1945, + "step": 55137 + }, + { + "epoch": 2.12, + "learning_rate": 1.467812259553732e-05, + "loss": 0.2678, + "step": 55138 + }, + { + "epoch": 2.12, + "learning_rate": 1.4677481405488588e-05, + "loss": 0.1357, + "step": 55139 + }, + { + "epoch": 2.12, + "learning_rate": 1.4676840215439858e-05, + "loss": 0.1995, + "step": 55140 + }, + { + "epoch": 2.12, + "learning_rate": 1.4676199025391126e-05, + "loss": 0.1153, + "step": 55141 + }, + { + "epoch": 2.12, + "learning_rate": 1.4675557835342396e-05, + "loss": 0.2738, + "step": 55142 + }, + { + "epoch": 2.12, + "learning_rate": 1.4674916645293665e-05, + "loss": 0.3077, + "step": 55143 + }, + { + "epoch": 2.12, + "learning_rate": 1.4674275455244935e-05, + "loss": 0.1905, + "step": 55144 + }, + { + "epoch": 2.12, + "learning_rate": 1.4673634265196205e-05, + "loss": 0.3494, + "step": 55145 + }, + { + "epoch": 2.12, + "learning_rate": 1.4672993075147473e-05, + "loss": 0.1521, + "step": 55146 + }, + { + "epoch": 2.12, + "learning_rate": 1.4672351885098745e-05, + "loss": 0.3369, + "step": 55147 + }, + { + "epoch": 2.12, + "learning_rate": 1.4671710695050015e-05, + "loss": 0.2334, + "step": 55148 + }, + { + "epoch": 2.12, + "learning_rate": 1.4671069505001283e-05, + "loss": 0.2804, + "step": 55149 + }, + { + "epoch": 2.12, + "learning_rate": 1.4670428314952553e-05, + "loss": 0.2153, + "step": 55150 + }, + { + "epoch": 2.12, + "learning_rate": 1.4669787124903821e-05, + "loss": 0.1095, + "step": 55151 + }, + { + "epoch": 2.12, + "learning_rate": 1.4669145934855091e-05, + "loss": 0.2032, + "step": 55152 + }, + { + "epoch": 2.12, + "learning_rate": 1.466850474480636e-05, + "loss": 0.2465, + "step": 55153 + }, + { + "epoch": 2.12, + "learning_rate": 1.466786355475763e-05, + "loss": 0.1735, + "step": 55154 + }, + { + "epoch": 2.12, + "learning_rate": 1.46672223647089e-05, + "loss": 0.2928, + "step": 55155 + }, + { + "epoch": 2.12, + "learning_rate": 1.4666581174660172e-05, + "loss": 0.2985, + "step": 55156 + }, + { + "epoch": 2.12, + "learning_rate": 1.466593998461144e-05, + "loss": 0.2339, + "step": 55157 + }, + { + "epoch": 2.12, + "learning_rate": 1.466529879456271e-05, + "loss": 0.2514, + "step": 55158 + }, + { + "epoch": 2.12, + "learning_rate": 1.4664657604513978e-05, + "loss": 0.2218, + "step": 55159 + }, + { + "epoch": 2.12, + "learning_rate": 1.4664016414465248e-05, + "loss": 0.2425, + "step": 55160 + }, + { + "epoch": 2.12, + "learning_rate": 1.4663375224416517e-05, + "loss": 0.1803, + "step": 55161 + }, + { + "epoch": 2.12, + "learning_rate": 1.4662734034367787e-05, + "loss": 0.1869, + "step": 55162 + }, + { + "epoch": 2.12, + "learning_rate": 1.4662092844319057e-05, + "loss": 0.3103, + "step": 55163 + }, + { + "epoch": 2.12, + "learning_rate": 1.4661451654270325e-05, + "loss": 0.2745, + "step": 55164 + }, + { + "epoch": 2.12, + "learning_rate": 1.4660810464221597e-05, + "loss": 0.2334, + "step": 55165 + }, + { + "epoch": 2.12, + "learning_rate": 1.4660169274172867e-05, + "loss": 0.1623, + "step": 55166 + }, + { + "epoch": 2.12, + "learning_rate": 1.4659528084124135e-05, + "loss": 0.3128, + "step": 55167 + }, + { + "epoch": 2.12, + "learning_rate": 1.4658886894075405e-05, + "loss": 0.1923, + "step": 55168 + }, + { + "epoch": 2.12, + "learning_rate": 1.4658245704026673e-05, + "loss": 0.1541, + "step": 55169 + }, + { + "epoch": 2.12, + "learning_rate": 1.4657604513977943e-05, + "loss": 0.1792, + "step": 55170 + }, + { + "epoch": 2.12, + "learning_rate": 1.4656963323929212e-05, + "loss": 0.4025, + "step": 55171 + }, + { + "epoch": 2.12, + "learning_rate": 1.4656322133880482e-05, + "loss": 0.2252, + "step": 55172 + }, + { + "epoch": 2.12, + "learning_rate": 1.4655680943831752e-05, + "loss": 0.3127, + "step": 55173 + }, + { + "epoch": 2.12, + "learning_rate": 1.4655039753783023e-05, + "loss": 0.1945, + "step": 55174 + }, + { + "epoch": 2.12, + "learning_rate": 1.4654398563734292e-05, + "loss": 0.2542, + "step": 55175 + }, + { + "epoch": 2.12, + "learning_rate": 1.4653757373685562e-05, + "loss": 0.162, + "step": 55176 + }, + { + "epoch": 2.12, + "learning_rate": 1.465311618363683e-05, + "loss": 0.4074, + "step": 55177 + }, + { + "epoch": 2.12, + "learning_rate": 1.46524749935881e-05, + "loss": 0.1833, + "step": 55178 + }, + { + "epoch": 2.12, + "learning_rate": 1.4651833803539369e-05, + "loss": 0.198, + "step": 55179 + }, + { + "epoch": 2.12, + "learning_rate": 1.4651192613490639e-05, + "loss": 0.2041, + "step": 55180 + }, + { + "epoch": 2.12, + "learning_rate": 1.4650551423441909e-05, + "loss": 0.2554, + "step": 55181 + }, + { + "epoch": 2.12, + "learning_rate": 1.464991023339318e-05, + "loss": 0.1722, + "step": 55182 + }, + { + "epoch": 2.12, + "learning_rate": 1.4649269043344449e-05, + "loss": 0.1238, + "step": 55183 + }, + { + "epoch": 2.12, + "learning_rate": 1.4648627853295719e-05, + "loss": 0.2547, + "step": 55184 + }, + { + "epoch": 2.12, + "learning_rate": 1.4647986663246987e-05, + "loss": 0.2328, + "step": 55185 + }, + { + "epoch": 2.12, + "learning_rate": 1.4647345473198257e-05, + "loss": 0.186, + "step": 55186 + }, + { + "epoch": 2.12, + "learning_rate": 1.4646704283149525e-05, + "loss": 0.2598, + "step": 55187 + }, + { + "epoch": 2.12, + "learning_rate": 1.4646063093100795e-05, + "loss": 0.1194, + "step": 55188 + }, + { + "epoch": 2.12, + "learning_rate": 1.4645421903052065e-05, + "loss": 0.2673, + "step": 55189 + }, + { + "epoch": 2.12, + "learning_rate": 1.4644780713003334e-05, + "loss": 0.2897, + "step": 55190 + }, + { + "epoch": 2.12, + "learning_rate": 1.4644139522954605e-05, + "loss": 0.2134, + "step": 55191 + }, + { + "epoch": 2.12, + "learning_rate": 1.4643498332905875e-05, + "loss": 0.26, + "step": 55192 + }, + { + "epoch": 2.12, + "learning_rate": 1.4642857142857144e-05, + "loss": 0.2141, + "step": 55193 + }, + { + "epoch": 2.12, + "learning_rate": 1.4642215952808414e-05, + "loss": 0.3051, + "step": 55194 + }, + { + "epoch": 2.12, + "learning_rate": 1.4641574762759682e-05, + "loss": 0.2882, + "step": 55195 + }, + { + "epoch": 2.12, + "learning_rate": 1.4640933572710952e-05, + "loss": 0.2538, + "step": 55196 + }, + { + "epoch": 2.12, + "learning_rate": 1.464029238266222e-05, + "loss": 0.3032, + "step": 55197 + }, + { + "epoch": 2.12, + "learning_rate": 1.463965119261349e-05, + "loss": 0.295, + "step": 55198 + }, + { + "epoch": 2.12, + "learning_rate": 1.463901000256476e-05, + "loss": 0.3292, + "step": 55199 + }, + { + "epoch": 2.12, + "learning_rate": 1.4638368812516032e-05, + "loss": 0.2061, + "step": 55200 + }, + { + "epoch": 2.12, + "learning_rate": 1.46377276224673e-05, + "loss": 0.2207, + "step": 55201 + }, + { + "epoch": 2.12, + "learning_rate": 1.463708643241857e-05, + "loss": 0.2346, + "step": 55202 + }, + { + "epoch": 2.12, + "learning_rate": 1.4636445242369839e-05, + "loss": 0.2632, + "step": 55203 + }, + { + "epoch": 2.12, + "learning_rate": 1.4635804052321109e-05, + "loss": 0.1925, + "step": 55204 + }, + { + "epoch": 2.12, + "learning_rate": 1.4635162862272377e-05, + "loss": 0.3024, + "step": 55205 + }, + { + "epoch": 2.12, + "learning_rate": 1.4634521672223647e-05, + "loss": 0.1798, + "step": 55206 + }, + { + "epoch": 2.12, + "learning_rate": 1.4633880482174917e-05, + "loss": 0.2555, + "step": 55207 + }, + { + "epoch": 2.12, + "learning_rate": 1.4633239292126186e-05, + "loss": 0.2384, + "step": 55208 + }, + { + "epoch": 2.12, + "learning_rate": 1.4632598102077457e-05, + "loss": 0.28, + "step": 55209 + }, + { + "epoch": 2.12, + "learning_rate": 1.4631956912028727e-05, + "loss": 0.3142, + "step": 55210 + }, + { + "epoch": 2.12, + "learning_rate": 1.4631315721979996e-05, + "loss": 0.2662, + "step": 55211 + }, + { + "epoch": 2.12, + "learning_rate": 1.4630674531931266e-05, + "loss": 0.2283, + "step": 55212 + }, + { + "epoch": 2.12, + "learning_rate": 1.4630033341882534e-05, + "loss": 0.306, + "step": 55213 + }, + { + "epoch": 2.12, + "learning_rate": 1.4629392151833804e-05, + "loss": 0.1317, + "step": 55214 + }, + { + "epoch": 2.12, + "learning_rate": 1.4628750961785072e-05, + "loss": 0.2281, + "step": 55215 + }, + { + "epoch": 2.12, + "learning_rate": 1.4628109771736342e-05, + "loss": 0.221, + "step": 55216 + }, + { + "epoch": 2.12, + "learning_rate": 1.4627468581687612e-05, + "loss": 0.2297, + "step": 55217 + }, + { + "epoch": 2.12, + "learning_rate": 1.4626827391638884e-05, + "loss": 0.296, + "step": 55218 + }, + { + "epoch": 2.12, + "learning_rate": 1.4626186201590152e-05, + "loss": 0.3126, + "step": 55219 + }, + { + "epoch": 2.12, + "learning_rate": 1.4625545011541423e-05, + "loss": 0.2367, + "step": 55220 + }, + { + "epoch": 2.12, + "learning_rate": 1.462490382149269e-05, + "loss": 0.1876, + "step": 55221 + }, + { + "epoch": 2.12, + "learning_rate": 1.4624262631443961e-05, + "loss": 0.2388, + "step": 55222 + }, + { + "epoch": 2.12, + "learning_rate": 1.462362144139523e-05, + "loss": 0.1695, + "step": 55223 + }, + { + "epoch": 2.12, + "learning_rate": 1.46229802513465e-05, + "loss": 0.2834, + "step": 55224 + }, + { + "epoch": 2.12, + "learning_rate": 1.462233906129777e-05, + "loss": 0.2711, + "step": 55225 + }, + { + "epoch": 2.12, + "learning_rate": 1.4621697871249038e-05, + "loss": 0.2785, + "step": 55226 + }, + { + "epoch": 2.12, + "learning_rate": 1.462105668120031e-05, + "loss": 0.2604, + "step": 55227 + }, + { + "epoch": 2.12, + "learning_rate": 1.462041549115158e-05, + "loss": 0.1376, + "step": 55228 + }, + { + "epoch": 2.12, + "learning_rate": 1.4619774301102848e-05, + "loss": 0.3693, + "step": 55229 + }, + { + "epoch": 2.12, + "learning_rate": 1.4619133111054118e-05, + "loss": 0.1896, + "step": 55230 + }, + { + "epoch": 2.12, + "learning_rate": 1.4618491921005386e-05, + "loss": 0.3902, + "step": 55231 + }, + { + "epoch": 2.12, + "learning_rate": 1.4617850730956656e-05, + "loss": 0.2688, + "step": 55232 + }, + { + "epoch": 2.12, + "learning_rate": 1.4617209540907924e-05, + "loss": 0.2535, + "step": 55233 + }, + { + "epoch": 2.12, + "learning_rate": 1.4616568350859194e-05, + "loss": 0.2374, + "step": 55234 + }, + { + "epoch": 2.12, + "learning_rate": 1.4615927160810464e-05, + "loss": 0.1448, + "step": 55235 + }, + { + "epoch": 2.12, + "learning_rate": 1.4615285970761736e-05, + "loss": 0.2349, + "step": 55236 + }, + { + "epoch": 2.12, + "learning_rate": 1.4614644780713004e-05, + "loss": 0.3149, + "step": 55237 + }, + { + "epoch": 2.12, + "learning_rate": 1.4614003590664274e-05, + "loss": 0.1416, + "step": 55238 + }, + { + "epoch": 2.12, + "learning_rate": 1.4613362400615543e-05, + "loss": 0.2289, + "step": 55239 + }, + { + "epoch": 2.12, + "learning_rate": 1.4612721210566813e-05, + "loss": 0.1483, + "step": 55240 + }, + { + "epoch": 2.12, + "learning_rate": 1.4612080020518081e-05, + "loss": 0.1615, + "step": 55241 + }, + { + "epoch": 2.12, + "learning_rate": 1.4611438830469351e-05, + "loss": 0.2104, + "step": 55242 + }, + { + "epoch": 2.12, + "learning_rate": 1.4610797640420621e-05, + "loss": 0.3029, + "step": 55243 + }, + { + "epoch": 2.12, + "learning_rate": 1.461015645037189e-05, + "loss": 0.1712, + "step": 55244 + }, + { + "epoch": 2.12, + "learning_rate": 1.4609515260323161e-05, + "loss": 0.2073, + "step": 55245 + }, + { + "epoch": 2.12, + "learning_rate": 1.4608874070274431e-05, + "loss": 0.2308, + "step": 55246 + }, + { + "epoch": 2.12, + "learning_rate": 1.46082328802257e-05, + "loss": 0.2181, + "step": 55247 + }, + { + "epoch": 2.12, + "learning_rate": 1.460759169017697e-05, + "loss": 0.1699, + "step": 55248 + }, + { + "epoch": 2.12, + "learning_rate": 1.4606950500128238e-05, + "loss": 0.2736, + "step": 55249 + }, + { + "epoch": 2.12, + "learning_rate": 1.4606309310079508e-05, + "loss": 0.2899, + "step": 55250 + }, + { + "epoch": 2.12, + "learning_rate": 1.4605668120030778e-05, + "loss": 0.236, + "step": 55251 + }, + { + "epoch": 2.12, + "learning_rate": 1.4605026929982046e-05, + "loss": 0.346, + "step": 55252 + }, + { + "epoch": 2.12, + "learning_rate": 1.4604385739933316e-05, + "loss": 0.2097, + "step": 55253 + }, + { + "epoch": 2.12, + "learning_rate": 1.4603744549884588e-05, + "loss": 0.1418, + "step": 55254 + }, + { + "epoch": 2.12, + "learning_rate": 1.4603103359835856e-05, + "loss": 0.3013, + "step": 55255 + }, + { + "epoch": 2.12, + "learning_rate": 1.4602462169787126e-05, + "loss": 0.277, + "step": 55256 + }, + { + "epoch": 2.12, + "learning_rate": 1.4601820979738395e-05, + "loss": 0.1626, + "step": 55257 + }, + { + "epoch": 2.12, + "learning_rate": 1.4601179789689665e-05, + "loss": 0.2955, + "step": 55258 + }, + { + "epoch": 2.12, + "learning_rate": 1.4600538599640933e-05, + "loss": 0.2424, + "step": 55259 + }, + { + "epoch": 2.12, + "learning_rate": 1.4599897409592203e-05, + "loss": 0.209, + "step": 55260 + }, + { + "epoch": 2.12, + "learning_rate": 1.4599256219543473e-05, + "loss": 0.1644, + "step": 55261 + }, + { + "epoch": 2.12, + "learning_rate": 1.4598615029494741e-05, + "loss": 0.3019, + "step": 55262 + }, + { + "epoch": 2.12, + "learning_rate": 1.4597973839446013e-05, + "loss": 0.3461, + "step": 55263 + }, + { + "epoch": 2.12, + "learning_rate": 1.4597332649397283e-05, + "loss": 0.2557, + "step": 55264 + }, + { + "epoch": 2.12, + "learning_rate": 1.4596691459348551e-05, + "loss": 0.2804, + "step": 55265 + }, + { + "epoch": 2.12, + "learning_rate": 1.4596050269299822e-05, + "loss": 0.2535, + "step": 55266 + }, + { + "epoch": 2.12, + "learning_rate": 1.459540907925109e-05, + "loss": 0.2186, + "step": 55267 + }, + { + "epoch": 2.12, + "learning_rate": 1.459476788920236e-05, + "loss": 0.5047, + "step": 55268 + }, + { + "epoch": 2.12, + "learning_rate": 1.459412669915363e-05, + "loss": 0.2755, + "step": 55269 + }, + { + "epoch": 2.12, + "learning_rate": 1.4593485509104898e-05, + "loss": 0.1741, + "step": 55270 + }, + { + "epoch": 2.12, + "learning_rate": 1.459284431905617e-05, + "loss": 0.2433, + "step": 55271 + }, + { + "epoch": 2.13, + "learning_rate": 1.459220312900744e-05, + "loss": 0.2144, + "step": 55272 + }, + { + "epoch": 2.13, + "learning_rate": 1.4591561938958708e-05, + "loss": 0.2525, + "step": 55273 + }, + { + "epoch": 2.13, + "learning_rate": 1.4590920748909978e-05, + "loss": 0.0957, + "step": 55274 + }, + { + "epoch": 2.13, + "learning_rate": 1.4590279558861247e-05, + "loss": 0.1668, + "step": 55275 + }, + { + "epoch": 2.13, + "learning_rate": 1.4589638368812517e-05, + "loss": 0.2633, + "step": 55276 + }, + { + "epoch": 2.13, + "learning_rate": 1.4588997178763785e-05, + "loss": 0.1785, + "step": 55277 + }, + { + "epoch": 2.13, + "learning_rate": 1.4588355988715055e-05, + "loss": 0.2788, + "step": 55278 + }, + { + "epoch": 2.13, + "learning_rate": 1.4587714798666325e-05, + "loss": 0.2652, + "step": 55279 + }, + { + "epoch": 2.13, + "learning_rate": 1.4587073608617597e-05, + "loss": 0.22, + "step": 55280 + }, + { + "epoch": 2.13, + "learning_rate": 1.4586432418568865e-05, + "loss": 0.2551, + "step": 55281 + }, + { + "epoch": 2.13, + "learning_rate": 1.4585791228520135e-05, + "loss": 0.3497, + "step": 55282 + }, + { + "epoch": 2.13, + "learning_rate": 1.4585150038471403e-05, + "loss": 0.2284, + "step": 55283 + }, + { + "epoch": 2.13, + "learning_rate": 1.4584508848422673e-05, + "loss": 0.2518, + "step": 55284 + }, + { + "epoch": 2.13, + "learning_rate": 1.4583867658373942e-05, + "loss": 0.1833, + "step": 55285 + }, + { + "epoch": 2.13, + "learning_rate": 1.4583226468325212e-05, + "loss": 0.1932, + "step": 55286 + }, + { + "epoch": 2.13, + "learning_rate": 1.4582585278276482e-05, + "loss": 0.3685, + "step": 55287 + }, + { + "epoch": 2.13, + "learning_rate": 1.458194408822775e-05, + "loss": 0.2441, + "step": 55288 + }, + { + "epoch": 2.13, + "learning_rate": 1.4581302898179022e-05, + "loss": 0.1993, + "step": 55289 + }, + { + "epoch": 2.13, + "learning_rate": 1.4580661708130292e-05, + "loss": 0.144, + "step": 55290 + }, + { + "epoch": 2.13, + "learning_rate": 1.458002051808156e-05, + "loss": 0.1784, + "step": 55291 + }, + { + "epoch": 2.13, + "learning_rate": 1.457937932803283e-05, + "loss": 0.2596, + "step": 55292 + }, + { + "epoch": 2.13, + "learning_rate": 1.4578738137984099e-05, + "loss": 0.2304, + "step": 55293 + }, + { + "epoch": 2.13, + "learning_rate": 1.4578096947935369e-05, + "loss": 0.252, + "step": 55294 + }, + { + "epoch": 2.13, + "learning_rate": 1.4577455757886637e-05, + "loss": 0.2434, + "step": 55295 + }, + { + "epoch": 2.13, + "learning_rate": 1.4576814567837907e-05, + "loss": 0.2926, + "step": 55296 + }, + { + "epoch": 2.13, + "learning_rate": 1.4576173377789177e-05, + "loss": 0.2447, + "step": 55297 + }, + { + "epoch": 2.13, + "learning_rate": 1.4575532187740449e-05, + "loss": 0.16, + "step": 55298 + }, + { + "epoch": 2.13, + "learning_rate": 1.4574890997691717e-05, + "loss": 0.2672, + "step": 55299 + }, + { + "epoch": 2.13, + "learning_rate": 1.4574249807642987e-05, + "loss": 0.2594, + "step": 55300 + }, + { + "epoch": 2.13, + "learning_rate": 1.4573608617594255e-05, + "loss": 0.209, + "step": 55301 + }, + { + "epoch": 2.13, + "learning_rate": 1.4572967427545525e-05, + "loss": 0.2804, + "step": 55302 + }, + { + "epoch": 2.13, + "learning_rate": 1.4572326237496794e-05, + "loss": 0.184, + "step": 55303 + }, + { + "epoch": 2.13, + "learning_rate": 1.4571685047448064e-05, + "loss": 0.2276, + "step": 55304 + }, + { + "epoch": 2.13, + "learning_rate": 1.4571043857399334e-05, + "loss": 0.1725, + "step": 55305 + }, + { + "epoch": 2.13, + "learning_rate": 1.4570402667350602e-05, + "loss": 0.2869, + "step": 55306 + }, + { + "epoch": 2.13, + "learning_rate": 1.4569761477301874e-05, + "loss": 0.2686, + "step": 55307 + }, + { + "epoch": 2.13, + "learning_rate": 1.4569120287253144e-05, + "loss": 0.1844, + "step": 55308 + }, + { + "epoch": 2.13, + "learning_rate": 1.4568479097204412e-05, + "loss": 0.3234, + "step": 55309 + }, + { + "epoch": 2.13, + "learning_rate": 1.4567837907155682e-05, + "loss": 0.1988, + "step": 55310 + }, + { + "epoch": 2.13, + "learning_rate": 1.456719671710695e-05, + "loss": 0.2189, + "step": 55311 + }, + { + "epoch": 2.13, + "learning_rate": 1.456655552705822e-05, + "loss": 0.2283, + "step": 55312 + }, + { + "epoch": 2.13, + "learning_rate": 1.456591433700949e-05, + "loss": 0.1843, + "step": 55313 + }, + { + "epoch": 2.13, + "learning_rate": 1.4565273146960759e-05, + "loss": 0.226, + "step": 55314 + }, + { + "epoch": 2.13, + "learning_rate": 1.4564631956912029e-05, + "loss": 0.1297, + "step": 55315 + }, + { + "epoch": 2.13, + "learning_rate": 1.45639907668633e-05, + "loss": 0.2539, + "step": 55316 + }, + { + "epoch": 2.13, + "learning_rate": 1.4563349576814569e-05, + "loss": 0.2035, + "step": 55317 + }, + { + "epoch": 2.13, + "learning_rate": 1.4562708386765839e-05, + "loss": 0.3089, + "step": 55318 + }, + { + "epoch": 2.13, + "learning_rate": 1.4562067196717107e-05, + "loss": 0.2487, + "step": 55319 + }, + { + "epoch": 2.13, + "learning_rate": 1.4561426006668377e-05, + "loss": 0.2407, + "step": 55320 + }, + { + "epoch": 2.13, + "learning_rate": 1.4560784816619646e-05, + "loss": 0.1775, + "step": 55321 + }, + { + "epoch": 2.13, + "learning_rate": 1.4560143626570916e-05, + "loss": 0.2997, + "step": 55322 + }, + { + "epoch": 2.13, + "learning_rate": 1.4559502436522186e-05, + "loss": 0.169, + "step": 55323 + }, + { + "epoch": 2.13, + "learning_rate": 1.4558861246473454e-05, + "loss": 0.2561, + "step": 55324 + }, + { + "epoch": 2.13, + "learning_rate": 1.4558220056424726e-05, + "loss": 0.2587, + "step": 55325 + }, + { + "epoch": 2.13, + "learning_rate": 1.4557578866375996e-05, + "loss": 0.1895, + "step": 55326 + }, + { + "epoch": 2.13, + "learning_rate": 1.4556937676327264e-05, + "loss": 0.0916, + "step": 55327 + }, + { + "epoch": 2.13, + "learning_rate": 1.4556296486278534e-05, + "loss": 0.1796, + "step": 55328 + }, + { + "epoch": 2.13, + "learning_rate": 1.4555655296229802e-05, + "loss": 0.2666, + "step": 55329 + }, + { + "epoch": 2.13, + "learning_rate": 1.4555014106181072e-05, + "loss": 0.2128, + "step": 55330 + }, + { + "epoch": 2.13, + "learning_rate": 1.4554372916132342e-05, + "loss": 0.1087, + "step": 55331 + }, + { + "epoch": 2.13, + "learning_rate": 1.455373172608361e-05, + "loss": 0.3131, + "step": 55332 + }, + { + "epoch": 2.13, + "learning_rate": 1.455309053603488e-05, + "loss": 0.2241, + "step": 55333 + }, + { + "epoch": 2.13, + "learning_rate": 1.4552449345986153e-05, + "loss": 0.2127, + "step": 55334 + }, + { + "epoch": 2.13, + "learning_rate": 1.4551808155937421e-05, + "loss": 0.195, + "step": 55335 + }, + { + "epoch": 2.13, + "learning_rate": 1.4551166965888691e-05, + "loss": 0.3152, + "step": 55336 + }, + { + "epoch": 2.13, + "learning_rate": 1.455052577583996e-05, + "loss": 0.245, + "step": 55337 + }, + { + "epoch": 2.13, + "learning_rate": 1.454988458579123e-05, + "loss": 0.2588, + "step": 55338 + }, + { + "epoch": 2.13, + "learning_rate": 1.4549243395742498e-05, + "loss": 0.122, + "step": 55339 + }, + { + "epoch": 2.13, + "learning_rate": 1.4548602205693768e-05, + "loss": 0.1762, + "step": 55340 + }, + { + "epoch": 2.13, + "learning_rate": 1.4547961015645038e-05, + "loss": 0.2841, + "step": 55341 + }, + { + "epoch": 2.13, + "learning_rate": 1.4547319825596306e-05, + "loss": 0.1877, + "step": 55342 + }, + { + "epoch": 2.13, + "learning_rate": 1.4546678635547578e-05, + "loss": 0.3105, + "step": 55343 + }, + { + "epoch": 2.13, + "learning_rate": 1.4546037445498848e-05, + "loss": 0.1903, + "step": 55344 + }, + { + "epoch": 2.13, + "learning_rate": 1.4545396255450116e-05, + "loss": 0.2185, + "step": 55345 + }, + { + "epoch": 2.13, + "learning_rate": 1.4544755065401386e-05, + "loss": 0.2399, + "step": 55346 + }, + { + "epoch": 2.13, + "learning_rate": 1.4544113875352654e-05, + "loss": 0.1644, + "step": 55347 + }, + { + "epoch": 2.13, + "learning_rate": 1.4543472685303924e-05, + "loss": 0.3343, + "step": 55348 + }, + { + "epoch": 2.13, + "learning_rate": 1.4542831495255194e-05, + "loss": 0.3706, + "step": 55349 + }, + { + "epoch": 2.13, + "learning_rate": 1.4542190305206463e-05, + "loss": 0.4579, + "step": 55350 + }, + { + "epoch": 2.13, + "learning_rate": 1.4541549115157733e-05, + "loss": 0.238, + "step": 55351 + }, + { + "epoch": 2.13, + "learning_rate": 1.4540907925109004e-05, + "loss": 0.2467, + "step": 55352 + }, + { + "epoch": 2.13, + "learning_rate": 1.4540266735060273e-05, + "loss": 0.1863, + "step": 55353 + }, + { + "epoch": 2.13, + "learning_rate": 1.4539625545011543e-05, + "loss": 0.1437, + "step": 55354 + }, + { + "epoch": 2.13, + "learning_rate": 1.4538984354962811e-05, + "loss": 0.1992, + "step": 55355 + }, + { + "epoch": 2.13, + "learning_rate": 1.4538343164914081e-05, + "loss": 0.2487, + "step": 55356 + }, + { + "epoch": 2.13, + "learning_rate": 1.453770197486535e-05, + "loss": 0.1513, + "step": 55357 + }, + { + "epoch": 2.13, + "learning_rate": 1.453706078481662e-05, + "loss": 0.1734, + "step": 55358 + }, + { + "epoch": 2.13, + "learning_rate": 1.453641959476789e-05, + "loss": 0.1944, + "step": 55359 + }, + { + "epoch": 2.13, + "learning_rate": 1.4535778404719158e-05, + "loss": 0.2474, + "step": 55360 + }, + { + "epoch": 2.13, + "learning_rate": 1.453513721467043e-05, + "loss": 0.2033, + "step": 55361 + }, + { + "epoch": 2.13, + "learning_rate": 1.45344960246217e-05, + "loss": 0.2433, + "step": 55362 + }, + { + "epoch": 2.13, + "learning_rate": 1.4533854834572968e-05, + "loss": 0.3165, + "step": 55363 + }, + { + "epoch": 2.13, + "learning_rate": 1.4533213644524238e-05, + "loss": 0.2458, + "step": 55364 + }, + { + "epoch": 2.13, + "learning_rate": 1.4532572454475506e-05, + "loss": 0.2137, + "step": 55365 + }, + { + "epoch": 2.13, + "learning_rate": 1.4531931264426776e-05, + "loss": 0.2452, + "step": 55366 + }, + { + "epoch": 2.13, + "learning_rate": 1.4531290074378046e-05, + "loss": 0.2228, + "step": 55367 + }, + { + "epoch": 2.13, + "learning_rate": 1.4530648884329315e-05, + "loss": 0.1949, + "step": 55368 + }, + { + "epoch": 2.13, + "learning_rate": 1.4530007694280586e-05, + "loss": 0.2402, + "step": 55369 + }, + { + "epoch": 2.13, + "learning_rate": 1.4529366504231856e-05, + "loss": 0.2339, + "step": 55370 + }, + { + "epoch": 2.13, + "learning_rate": 1.4528725314183125e-05, + "loss": 0.2324, + "step": 55371 + }, + { + "epoch": 2.13, + "learning_rate": 1.4528084124134395e-05, + "loss": 0.249, + "step": 55372 + }, + { + "epoch": 2.13, + "learning_rate": 1.4527442934085663e-05, + "loss": 0.3246, + "step": 55373 + }, + { + "epoch": 2.13, + "learning_rate": 1.4526801744036933e-05, + "loss": 0.1914, + "step": 55374 + }, + { + "epoch": 2.13, + "learning_rate": 1.4526160553988201e-05, + "loss": 0.2621, + "step": 55375 + }, + { + "epoch": 2.13, + "learning_rate": 1.4525519363939471e-05, + "loss": 0.1815, + "step": 55376 + }, + { + "epoch": 2.13, + "learning_rate": 1.4524878173890742e-05, + "loss": 0.1111, + "step": 55377 + }, + { + "epoch": 2.13, + "learning_rate": 1.4524236983842013e-05, + "loss": 0.2211, + "step": 55378 + }, + { + "epoch": 2.13, + "learning_rate": 1.4523595793793282e-05, + "loss": 0.1625, + "step": 55379 + }, + { + "epoch": 2.13, + "learning_rate": 1.4522954603744552e-05, + "loss": 0.2461, + "step": 55380 + }, + { + "epoch": 2.13, + "learning_rate": 1.452231341369582e-05, + "loss": 0.2156, + "step": 55381 + }, + { + "epoch": 2.13, + "learning_rate": 1.452167222364709e-05, + "loss": 0.2239, + "step": 55382 + }, + { + "epoch": 2.13, + "learning_rate": 1.4521031033598358e-05, + "loss": 0.3147, + "step": 55383 + }, + { + "epoch": 2.13, + "learning_rate": 1.4520389843549628e-05, + "loss": 0.2951, + "step": 55384 + }, + { + "epoch": 2.13, + "learning_rate": 1.4519748653500898e-05, + "loss": 0.3941, + "step": 55385 + }, + { + "epoch": 2.13, + "learning_rate": 1.4519107463452167e-05, + "loss": 0.273, + "step": 55386 + }, + { + "epoch": 2.13, + "learning_rate": 1.4518466273403438e-05, + "loss": 0.162, + "step": 55387 + }, + { + "epoch": 2.13, + "learning_rate": 1.4517825083354708e-05, + "loss": 0.2581, + "step": 55388 + }, + { + "epoch": 2.13, + "learning_rate": 1.4517183893305977e-05, + "loss": 0.2682, + "step": 55389 + }, + { + "epoch": 2.13, + "learning_rate": 1.4516542703257247e-05, + "loss": 0.2434, + "step": 55390 + }, + { + "epoch": 2.13, + "learning_rate": 1.4515901513208515e-05, + "loss": 0.3152, + "step": 55391 + }, + { + "epoch": 2.13, + "learning_rate": 1.4515260323159785e-05, + "loss": 0.1349, + "step": 55392 + }, + { + "epoch": 2.13, + "learning_rate": 1.4514619133111055e-05, + "loss": 0.2244, + "step": 55393 + }, + { + "epoch": 2.13, + "learning_rate": 1.4513977943062323e-05, + "loss": 0.2618, + "step": 55394 + }, + { + "epoch": 2.13, + "learning_rate": 1.4513336753013593e-05, + "loss": 0.1077, + "step": 55395 + }, + { + "epoch": 2.13, + "learning_rate": 1.4512695562964865e-05, + "loss": 0.1373, + "step": 55396 + }, + { + "epoch": 2.13, + "learning_rate": 1.4512054372916133e-05, + "loss": 0.1556, + "step": 55397 + }, + { + "epoch": 2.13, + "learning_rate": 1.4511413182867404e-05, + "loss": 0.1989, + "step": 55398 + }, + { + "epoch": 2.13, + "learning_rate": 1.4510771992818672e-05, + "loss": 0.179, + "step": 55399 + }, + { + "epoch": 2.13, + "learning_rate": 1.4510130802769942e-05, + "loss": 0.2131, + "step": 55400 + }, + { + "epoch": 2.13, + "learning_rate": 1.450948961272121e-05, + "loss": 0.3263, + "step": 55401 + }, + { + "epoch": 2.13, + "learning_rate": 1.450884842267248e-05, + "loss": 0.2179, + "step": 55402 + }, + { + "epoch": 2.13, + "learning_rate": 1.450820723262375e-05, + "loss": 0.2507, + "step": 55403 + }, + { + "epoch": 2.13, + "learning_rate": 1.4507566042575019e-05, + "loss": 0.234, + "step": 55404 + }, + { + "epoch": 2.13, + "learning_rate": 1.450692485252629e-05, + "loss": 0.2067, + "step": 55405 + }, + { + "epoch": 2.13, + "learning_rate": 1.450628366247756e-05, + "loss": 0.344, + "step": 55406 + }, + { + "epoch": 2.13, + "learning_rate": 1.4505642472428829e-05, + "loss": 0.2197, + "step": 55407 + }, + { + "epoch": 2.13, + "learning_rate": 1.4505001282380099e-05, + "loss": 0.2716, + "step": 55408 + }, + { + "epoch": 2.13, + "learning_rate": 1.4504360092331367e-05, + "loss": 0.1977, + "step": 55409 + }, + { + "epoch": 2.13, + "learning_rate": 1.4503718902282637e-05, + "loss": 0.3195, + "step": 55410 + }, + { + "epoch": 2.13, + "learning_rate": 1.4503077712233907e-05, + "loss": 0.2689, + "step": 55411 + }, + { + "epoch": 2.13, + "learning_rate": 1.4502436522185175e-05, + "loss": 0.1336, + "step": 55412 + }, + { + "epoch": 2.13, + "learning_rate": 1.4501795332136445e-05, + "loss": 0.2532, + "step": 55413 + }, + { + "epoch": 2.13, + "learning_rate": 1.4501154142087717e-05, + "loss": 0.2964, + "step": 55414 + }, + { + "epoch": 2.13, + "learning_rate": 1.4500512952038985e-05, + "loss": 0.2447, + "step": 55415 + }, + { + "epoch": 2.13, + "learning_rate": 1.4499871761990255e-05, + "loss": 0.2996, + "step": 55416 + }, + { + "epoch": 2.13, + "learning_rate": 1.4499230571941524e-05, + "loss": 0.4474, + "step": 55417 + }, + { + "epoch": 2.13, + "learning_rate": 1.4498589381892794e-05, + "loss": 0.2464, + "step": 55418 + }, + { + "epoch": 2.13, + "learning_rate": 1.4497948191844062e-05, + "loss": 0.3587, + "step": 55419 + }, + { + "epoch": 2.13, + "learning_rate": 1.4497307001795332e-05, + "loss": 0.3162, + "step": 55420 + }, + { + "epoch": 2.13, + "learning_rate": 1.4496665811746602e-05, + "loss": 0.2482, + "step": 55421 + }, + { + "epoch": 2.13, + "learning_rate": 1.449602462169787e-05, + "loss": 0.1987, + "step": 55422 + }, + { + "epoch": 2.13, + "learning_rate": 1.4495383431649142e-05, + "loss": 0.2518, + "step": 55423 + }, + { + "epoch": 2.13, + "learning_rate": 1.4494742241600412e-05, + "loss": 0.3405, + "step": 55424 + }, + { + "epoch": 2.13, + "learning_rate": 1.449410105155168e-05, + "loss": 0.1755, + "step": 55425 + }, + { + "epoch": 2.13, + "learning_rate": 1.449345986150295e-05, + "loss": 0.4605, + "step": 55426 + }, + { + "epoch": 2.13, + "learning_rate": 1.4492818671454219e-05, + "loss": 0.2105, + "step": 55427 + }, + { + "epoch": 2.13, + "learning_rate": 1.4492177481405489e-05, + "loss": 0.3658, + "step": 55428 + }, + { + "epoch": 2.13, + "learning_rate": 1.4491536291356759e-05, + "loss": 0.2384, + "step": 55429 + }, + { + "epoch": 2.13, + "learning_rate": 1.4490895101308027e-05, + "loss": 0.3868, + "step": 55430 + }, + { + "epoch": 2.13, + "learning_rate": 1.4490253911259297e-05, + "loss": 0.3028, + "step": 55431 + }, + { + "epoch": 2.13, + "learning_rate": 1.4489612721210569e-05, + "loss": 0.2503, + "step": 55432 + }, + { + "epoch": 2.13, + "learning_rate": 1.4488971531161837e-05, + "loss": 0.1727, + "step": 55433 + }, + { + "epoch": 2.13, + "learning_rate": 1.4488330341113107e-05, + "loss": 0.1788, + "step": 55434 + }, + { + "epoch": 2.13, + "learning_rate": 1.4487689151064376e-05, + "loss": 0.1643, + "step": 55435 + }, + { + "epoch": 2.13, + "learning_rate": 1.4487047961015646e-05, + "loss": 0.2041, + "step": 55436 + }, + { + "epoch": 2.13, + "learning_rate": 1.4486406770966914e-05, + "loss": 0.2223, + "step": 55437 + }, + { + "epoch": 2.13, + "learning_rate": 1.4485765580918184e-05, + "loss": 0.2017, + "step": 55438 + }, + { + "epoch": 2.13, + "learning_rate": 1.4485124390869454e-05, + "loss": 0.2211, + "step": 55439 + }, + { + "epoch": 2.13, + "learning_rate": 1.4484483200820722e-05, + "loss": 0.1882, + "step": 55440 + }, + { + "epoch": 2.13, + "learning_rate": 1.4483842010771994e-05, + "loss": 0.3046, + "step": 55441 + }, + { + "epoch": 2.13, + "learning_rate": 1.4483200820723264e-05, + "loss": 0.2037, + "step": 55442 + }, + { + "epoch": 2.13, + "learning_rate": 1.4482559630674533e-05, + "loss": 0.1318, + "step": 55443 + }, + { + "epoch": 2.13, + "learning_rate": 1.4481918440625803e-05, + "loss": 0.2373, + "step": 55444 + }, + { + "epoch": 2.13, + "learning_rate": 1.4481277250577071e-05, + "loss": 0.1622, + "step": 55445 + }, + { + "epoch": 2.13, + "learning_rate": 1.4480636060528341e-05, + "loss": 0.2255, + "step": 55446 + }, + { + "epoch": 2.13, + "learning_rate": 1.4479994870479611e-05, + "loss": 0.1569, + "step": 55447 + }, + { + "epoch": 2.13, + "learning_rate": 1.447935368043088e-05, + "loss": 0.2641, + "step": 55448 + }, + { + "epoch": 2.13, + "learning_rate": 1.447871249038215e-05, + "loss": 0.2861, + "step": 55449 + }, + { + "epoch": 2.13, + "learning_rate": 1.4478071300333421e-05, + "loss": 0.1881, + "step": 55450 + }, + { + "epoch": 2.13, + "learning_rate": 1.447743011028469e-05, + "loss": 0.1612, + "step": 55451 + }, + { + "epoch": 2.13, + "learning_rate": 1.447678892023596e-05, + "loss": 0.2519, + "step": 55452 + }, + { + "epoch": 2.13, + "learning_rate": 1.4476147730187228e-05, + "loss": 0.1776, + "step": 55453 + }, + { + "epoch": 2.13, + "learning_rate": 1.4475506540138498e-05, + "loss": 0.3761, + "step": 55454 + }, + { + "epoch": 2.13, + "learning_rate": 1.4474865350089768e-05, + "loss": 0.1616, + "step": 55455 + }, + { + "epoch": 2.13, + "learning_rate": 1.4474224160041036e-05, + "loss": 0.3963, + "step": 55456 + }, + { + "epoch": 2.13, + "learning_rate": 1.4473582969992306e-05, + "loss": 0.1901, + "step": 55457 + }, + { + "epoch": 2.13, + "learning_rate": 1.4472941779943574e-05, + "loss": 0.3899, + "step": 55458 + }, + { + "epoch": 2.13, + "learning_rate": 1.4472300589894846e-05, + "loss": 0.2305, + "step": 55459 + }, + { + "epoch": 2.13, + "learning_rate": 1.4471659399846116e-05, + "loss": 0.2473, + "step": 55460 + }, + { + "epoch": 2.13, + "learning_rate": 1.4471018209797384e-05, + "loss": 0.2074, + "step": 55461 + }, + { + "epoch": 2.13, + "learning_rate": 1.4470377019748654e-05, + "loss": 0.305, + "step": 55462 + }, + { + "epoch": 2.13, + "learning_rate": 1.4469735829699923e-05, + "loss": 0.2369, + "step": 55463 + }, + { + "epoch": 2.13, + "learning_rate": 1.4469094639651193e-05, + "loss": 0.2005, + "step": 55464 + }, + { + "epoch": 2.13, + "learning_rate": 1.4468453449602463e-05, + "loss": 0.2686, + "step": 55465 + }, + { + "epoch": 2.13, + "learning_rate": 1.4467812259553731e-05, + "loss": 0.1672, + "step": 55466 + }, + { + "epoch": 2.13, + "learning_rate": 1.4467171069505003e-05, + "loss": 0.3445, + "step": 55467 + }, + { + "epoch": 2.13, + "learning_rate": 1.4466529879456273e-05, + "loss": 0.1902, + "step": 55468 + }, + { + "epoch": 2.13, + "learning_rate": 1.4465888689407541e-05, + "loss": 0.3427, + "step": 55469 + }, + { + "epoch": 2.13, + "learning_rate": 1.4465247499358811e-05, + "loss": 0.2513, + "step": 55470 + }, + { + "epoch": 2.13, + "learning_rate": 1.446460630931008e-05, + "loss": 0.2372, + "step": 55471 + }, + { + "epoch": 2.13, + "learning_rate": 1.446396511926135e-05, + "loss": 0.2052, + "step": 55472 + }, + { + "epoch": 2.13, + "learning_rate": 1.446332392921262e-05, + "loss": 0.3136, + "step": 55473 + }, + { + "epoch": 2.13, + "learning_rate": 1.4462682739163888e-05, + "loss": 0.1873, + "step": 55474 + }, + { + "epoch": 2.13, + "learning_rate": 1.4462041549115158e-05, + "loss": 0.3218, + "step": 55475 + }, + { + "epoch": 2.13, + "learning_rate": 1.446140035906643e-05, + "loss": 0.2532, + "step": 55476 + }, + { + "epoch": 2.13, + "learning_rate": 1.4460759169017698e-05, + "loss": 0.2304, + "step": 55477 + }, + { + "epoch": 2.13, + "learning_rate": 1.4460117978968968e-05, + "loss": 0.2032, + "step": 55478 + }, + { + "epoch": 2.13, + "learning_rate": 1.4459476788920236e-05, + "loss": 0.3084, + "step": 55479 + }, + { + "epoch": 2.13, + "learning_rate": 1.4458835598871506e-05, + "loss": 0.3617, + "step": 55480 + }, + { + "epoch": 2.13, + "learning_rate": 1.4458194408822775e-05, + "loss": 0.2193, + "step": 55481 + }, + { + "epoch": 2.13, + "learning_rate": 1.4457553218774045e-05, + "loss": 0.2376, + "step": 55482 + }, + { + "epoch": 2.13, + "learning_rate": 1.4456912028725315e-05, + "loss": 0.3146, + "step": 55483 + }, + { + "epoch": 2.13, + "learning_rate": 1.4456270838676583e-05, + "loss": 0.2188, + "step": 55484 + }, + { + "epoch": 2.13, + "learning_rate": 1.4455629648627855e-05, + "loss": 0.18, + "step": 55485 + }, + { + "epoch": 2.13, + "learning_rate": 1.4454988458579125e-05, + "loss": 0.2127, + "step": 55486 + }, + { + "epoch": 2.13, + "learning_rate": 1.4454347268530393e-05, + "loss": 0.1687, + "step": 55487 + }, + { + "epoch": 2.13, + "learning_rate": 1.4453706078481663e-05, + "loss": 0.2415, + "step": 55488 + }, + { + "epoch": 2.13, + "learning_rate": 1.4453064888432932e-05, + "loss": 0.315, + "step": 55489 + }, + { + "epoch": 2.13, + "learning_rate": 1.4452423698384202e-05, + "loss": 0.187, + "step": 55490 + }, + { + "epoch": 2.13, + "learning_rate": 1.4451782508335472e-05, + "loss": 0.2573, + "step": 55491 + }, + { + "epoch": 2.13, + "learning_rate": 1.445114131828674e-05, + "loss": 0.2811, + "step": 55492 + }, + { + "epoch": 2.13, + "learning_rate": 1.445050012823801e-05, + "loss": 0.1739, + "step": 55493 + }, + { + "epoch": 2.13, + "learning_rate": 1.4449858938189282e-05, + "loss": 0.2747, + "step": 55494 + }, + { + "epoch": 2.13, + "learning_rate": 1.444921774814055e-05, + "loss": 0.254, + "step": 55495 + }, + { + "epoch": 2.13, + "learning_rate": 1.444857655809182e-05, + "loss": 0.2769, + "step": 55496 + }, + { + "epoch": 2.13, + "learning_rate": 1.4447935368043088e-05, + "loss": 0.1439, + "step": 55497 + }, + { + "epoch": 2.13, + "learning_rate": 1.4447294177994358e-05, + "loss": 0.1972, + "step": 55498 + }, + { + "epoch": 2.13, + "learning_rate": 1.4446652987945627e-05, + "loss": 0.2393, + "step": 55499 + }, + { + "epoch": 2.13, + "learning_rate": 1.4446011797896897e-05, + "loss": 0.249, + "step": 55500 + }, + { + "epoch": 2.13, + "learning_rate": 1.4445370607848167e-05, + "loss": 0.2749, + "step": 55501 + }, + { + "epoch": 2.13, + "learning_rate": 1.4444729417799435e-05, + "loss": 0.2379, + "step": 55502 + }, + { + "epoch": 2.13, + "learning_rate": 1.4444088227750707e-05, + "loss": 0.2629, + "step": 55503 + }, + { + "epoch": 2.13, + "learning_rate": 1.4443447037701977e-05, + "loss": 0.1865, + "step": 55504 + }, + { + "epoch": 2.13, + "learning_rate": 1.4442805847653245e-05, + "loss": 0.3604, + "step": 55505 + }, + { + "epoch": 2.13, + "learning_rate": 1.4442164657604515e-05, + "loss": 0.2221, + "step": 55506 + }, + { + "epoch": 2.13, + "learning_rate": 1.4441523467555783e-05, + "loss": 0.2335, + "step": 55507 + }, + { + "epoch": 2.13, + "learning_rate": 1.4440882277507053e-05, + "loss": 0.2069, + "step": 55508 + }, + { + "epoch": 2.13, + "learning_rate": 1.4440241087458323e-05, + "loss": 0.1988, + "step": 55509 + }, + { + "epoch": 2.13, + "learning_rate": 1.4439599897409592e-05, + "loss": 0.2421, + "step": 55510 + }, + { + "epoch": 2.13, + "learning_rate": 1.4438958707360862e-05, + "loss": 0.378, + "step": 55511 + }, + { + "epoch": 2.13, + "learning_rate": 1.4438317517312134e-05, + "loss": 0.2732, + "step": 55512 + }, + { + "epoch": 2.13, + "learning_rate": 1.4437676327263402e-05, + "loss": 0.1852, + "step": 55513 + }, + { + "epoch": 2.13, + "learning_rate": 1.4437035137214672e-05, + "loss": 0.244, + "step": 55514 + }, + { + "epoch": 2.13, + "learning_rate": 1.443639394716594e-05, + "loss": 0.211, + "step": 55515 + }, + { + "epoch": 2.13, + "learning_rate": 1.443575275711721e-05, + "loss": 0.3282, + "step": 55516 + }, + { + "epoch": 2.13, + "learning_rate": 1.443511156706848e-05, + "loss": 0.3926, + "step": 55517 + }, + { + "epoch": 2.13, + "learning_rate": 1.4434470377019749e-05, + "loss": 0.1601, + "step": 55518 + }, + { + "epoch": 2.13, + "learning_rate": 1.4433829186971019e-05, + "loss": 0.3869, + "step": 55519 + }, + { + "epoch": 2.13, + "learning_rate": 1.4433187996922287e-05, + "loss": 0.1874, + "step": 55520 + }, + { + "epoch": 2.13, + "learning_rate": 1.4432546806873559e-05, + "loss": 0.2396, + "step": 55521 + }, + { + "epoch": 2.13, + "learning_rate": 1.4431905616824829e-05, + "loss": 0.2663, + "step": 55522 + }, + { + "epoch": 2.13, + "learning_rate": 1.4431264426776097e-05, + "loss": 0.1454, + "step": 55523 + }, + { + "epoch": 2.13, + "learning_rate": 1.4430623236727367e-05, + "loss": 0.1659, + "step": 55524 + }, + { + "epoch": 2.13, + "learning_rate": 1.4429982046678635e-05, + "loss": 0.1755, + "step": 55525 + }, + { + "epoch": 2.13, + "learning_rate": 1.4429340856629905e-05, + "loss": 0.2703, + "step": 55526 + }, + { + "epoch": 2.13, + "learning_rate": 1.4428699666581175e-05, + "loss": 0.3364, + "step": 55527 + }, + { + "epoch": 2.13, + "learning_rate": 1.4428058476532444e-05, + "loss": 0.2426, + "step": 55528 + }, + { + "epoch": 2.13, + "learning_rate": 1.4427417286483714e-05, + "loss": 0.2082, + "step": 55529 + }, + { + "epoch": 2.13, + "learning_rate": 1.4426776096434986e-05, + "loss": 0.1707, + "step": 55530 + }, + { + "epoch": 2.13, + "learning_rate": 1.4426134906386254e-05, + "loss": 0.178, + "step": 55531 + }, + { + "epoch": 2.14, + "learning_rate": 1.4425493716337524e-05, + "loss": 0.2681, + "step": 55532 + }, + { + "epoch": 2.14, + "learning_rate": 1.4424852526288792e-05, + "loss": 0.3551, + "step": 55533 + }, + { + "epoch": 2.14, + "learning_rate": 1.4424211336240062e-05, + "loss": 0.0707, + "step": 55534 + }, + { + "epoch": 2.14, + "learning_rate": 1.4423570146191332e-05, + "loss": 0.2699, + "step": 55535 + }, + { + "epoch": 2.14, + "learning_rate": 1.44229289561426e-05, + "loss": 0.2872, + "step": 55536 + }, + { + "epoch": 2.14, + "learning_rate": 1.442228776609387e-05, + "loss": 0.1295, + "step": 55537 + }, + { + "epoch": 2.14, + "learning_rate": 1.4421646576045139e-05, + "loss": 0.1358, + "step": 55538 + }, + { + "epoch": 2.14, + "learning_rate": 1.442100538599641e-05, + "loss": 0.2163, + "step": 55539 + }, + { + "epoch": 2.14, + "learning_rate": 1.442036419594768e-05, + "loss": 0.1029, + "step": 55540 + }, + { + "epoch": 2.14, + "learning_rate": 1.4419723005898949e-05, + "loss": 0.1917, + "step": 55541 + }, + { + "epoch": 2.14, + "learning_rate": 1.4419081815850219e-05, + "loss": 0.1756, + "step": 55542 + }, + { + "epoch": 2.14, + "learning_rate": 1.4418440625801487e-05, + "loss": 0.1524, + "step": 55543 + }, + { + "epoch": 2.14, + "learning_rate": 1.4417799435752757e-05, + "loss": 0.2702, + "step": 55544 + }, + { + "epoch": 2.14, + "learning_rate": 1.4417158245704027e-05, + "loss": 0.1577, + "step": 55545 + }, + { + "epoch": 2.14, + "learning_rate": 1.4416517055655296e-05, + "loss": 0.1902, + "step": 55546 + }, + { + "epoch": 2.14, + "learning_rate": 1.4415875865606566e-05, + "loss": 0.2725, + "step": 55547 + }, + { + "epoch": 2.14, + "learning_rate": 1.4415234675557837e-05, + "loss": 0.2112, + "step": 55548 + }, + { + "epoch": 2.14, + "learning_rate": 1.4414593485509106e-05, + "loss": 0.2327, + "step": 55549 + }, + { + "epoch": 2.14, + "learning_rate": 1.4413952295460376e-05, + "loss": 0.2699, + "step": 55550 + }, + { + "epoch": 2.14, + "learning_rate": 1.4413311105411644e-05, + "loss": 0.2371, + "step": 55551 + }, + { + "epoch": 2.14, + "learning_rate": 1.4412669915362914e-05, + "loss": 0.5484, + "step": 55552 + }, + { + "epoch": 2.14, + "learning_rate": 1.4412028725314184e-05, + "loss": 0.1595, + "step": 55553 + }, + { + "epoch": 2.14, + "learning_rate": 1.4411387535265452e-05, + "loss": 0.3858, + "step": 55554 + }, + { + "epoch": 2.14, + "learning_rate": 1.4410746345216723e-05, + "loss": 0.282, + "step": 55555 + }, + { + "epoch": 2.14, + "learning_rate": 1.4410105155167994e-05, + "loss": 0.2737, + "step": 55556 + }, + { + "epoch": 2.14, + "learning_rate": 1.4409463965119263e-05, + "loss": 0.3492, + "step": 55557 + }, + { + "epoch": 2.14, + "learning_rate": 1.4408822775070533e-05, + "loss": 0.1902, + "step": 55558 + }, + { + "epoch": 2.14, + "learning_rate": 1.4408181585021801e-05, + "loss": 0.3229, + "step": 55559 + }, + { + "epoch": 2.14, + "learning_rate": 1.4407540394973071e-05, + "loss": 0.1697, + "step": 55560 + }, + { + "epoch": 2.14, + "learning_rate": 1.440689920492434e-05, + "loss": 0.28, + "step": 55561 + }, + { + "epoch": 2.14, + "learning_rate": 1.440625801487561e-05, + "loss": 0.4534, + "step": 55562 + }, + { + "epoch": 2.14, + "learning_rate": 1.440561682482688e-05, + "loss": 0.26, + "step": 55563 + }, + { + "epoch": 2.14, + "learning_rate": 1.4404975634778148e-05, + "loss": 0.3256, + "step": 55564 + }, + { + "epoch": 2.14, + "learning_rate": 1.440433444472942e-05, + "loss": 0.2346, + "step": 55565 + }, + { + "epoch": 2.14, + "learning_rate": 1.440369325468069e-05, + "loss": 0.31, + "step": 55566 + }, + { + "epoch": 2.14, + "learning_rate": 1.4403052064631958e-05, + "loss": 0.3473, + "step": 55567 + }, + { + "epoch": 2.14, + "learning_rate": 1.4402410874583228e-05, + "loss": 0.4808, + "step": 55568 + }, + { + "epoch": 2.14, + "learning_rate": 1.4401769684534496e-05, + "loss": 0.1776, + "step": 55569 + }, + { + "epoch": 2.14, + "learning_rate": 1.4401128494485766e-05, + "loss": 0.1394, + "step": 55570 + }, + { + "epoch": 2.14, + "learning_rate": 1.4400487304437036e-05, + "loss": 0.3539, + "step": 55571 + }, + { + "epoch": 2.14, + "learning_rate": 1.4399846114388304e-05, + "loss": 0.2196, + "step": 55572 + }, + { + "epoch": 2.14, + "learning_rate": 1.4399204924339574e-05, + "loss": 0.2885, + "step": 55573 + }, + { + "epoch": 2.14, + "learning_rate": 1.4398563734290846e-05, + "loss": 0.1342, + "step": 55574 + }, + { + "epoch": 2.14, + "learning_rate": 1.4397922544242114e-05, + "loss": 0.2011, + "step": 55575 + }, + { + "epoch": 2.14, + "learning_rate": 1.4397281354193385e-05, + "loss": 0.2543, + "step": 55576 + }, + { + "epoch": 2.14, + "learning_rate": 1.4396640164144653e-05, + "loss": 0.2196, + "step": 55577 + }, + { + "epoch": 2.14, + "learning_rate": 1.4395998974095923e-05, + "loss": 0.1913, + "step": 55578 + }, + { + "epoch": 2.14, + "learning_rate": 1.4395357784047193e-05, + "loss": 0.1733, + "step": 55579 + }, + { + "epoch": 2.14, + "learning_rate": 1.4394716593998461e-05, + "loss": 0.2431, + "step": 55580 + }, + { + "epoch": 2.14, + "learning_rate": 1.4394075403949731e-05, + "loss": 0.1771, + "step": 55581 + }, + { + "epoch": 2.14, + "learning_rate": 1.4393434213901e-05, + "loss": 0.1713, + "step": 55582 + }, + { + "epoch": 2.14, + "learning_rate": 1.4392793023852271e-05, + "loss": 0.3041, + "step": 55583 + }, + { + "epoch": 2.14, + "learning_rate": 1.4392151833803541e-05, + "loss": 0.2489, + "step": 55584 + }, + { + "epoch": 2.14, + "learning_rate": 1.439151064375481e-05, + "loss": 0.1162, + "step": 55585 + }, + { + "epoch": 2.14, + "learning_rate": 1.439086945370608e-05, + "loss": 0.174, + "step": 55586 + }, + { + "epoch": 2.14, + "learning_rate": 1.4390228263657348e-05, + "loss": 0.3971, + "step": 55587 + }, + { + "epoch": 2.14, + "learning_rate": 1.4389587073608618e-05, + "loss": 0.2461, + "step": 55588 + }, + { + "epoch": 2.14, + "learning_rate": 1.4388945883559888e-05, + "loss": 0.2757, + "step": 55589 + }, + { + "epoch": 2.14, + "learning_rate": 1.4388304693511156e-05, + "loss": 0.2743, + "step": 55590 + }, + { + "epoch": 2.14, + "learning_rate": 1.4387663503462426e-05, + "loss": 0.2006, + "step": 55591 + }, + { + "epoch": 2.14, + "learning_rate": 1.4387022313413698e-05, + "loss": 0.259, + "step": 55592 + }, + { + "epoch": 2.14, + "learning_rate": 1.4386381123364966e-05, + "loss": 0.2684, + "step": 55593 + }, + { + "epoch": 2.14, + "learning_rate": 1.4385739933316236e-05, + "loss": 0.2764, + "step": 55594 + }, + { + "epoch": 2.14, + "learning_rate": 1.4385098743267505e-05, + "loss": 0.3241, + "step": 55595 + }, + { + "epoch": 2.14, + "learning_rate": 1.4384457553218775e-05, + "loss": 0.1956, + "step": 55596 + }, + { + "epoch": 2.14, + "learning_rate": 1.4383816363170045e-05, + "loss": 0.1528, + "step": 55597 + }, + { + "epoch": 2.14, + "learning_rate": 1.4383175173121313e-05, + "loss": 0.2826, + "step": 55598 + }, + { + "epoch": 2.14, + "learning_rate": 1.4382533983072583e-05, + "loss": 0.2594, + "step": 55599 + }, + { + "epoch": 2.14, + "learning_rate": 1.4381892793023852e-05, + "loss": 0.2007, + "step": 55600 + }, + { + "epoch": 2.14, + "learning_rate": 1.4381251602975123e-05, + "loss": 0.1366, + "step": 55601 + }, + { + "epoch": 2.14, + "learning_rate": 1.4380610412926393e-05, + "loss": 0.1646, + "step": 55602 + }, + { + "epoch": 2.14, + "learning_rate": 1.4379969222877662e-05, + "loss": 0.3181, + "step": 55603 + }, + { + "epoch": 2.14, + "learning_rate": 1.4379328032828932e-05, + "loss": 0.181, + "step": 55604 + }, + { + "epoch": 2.14, + "learning_rate": 1.43786868427802e-05, + "loss": 0.1651, + "step": 55605 + }, + { + "epoch": 2.14, + "learning_rate": 1.437804565273147e-05, + "loss": 0.2406, + "step": 55606 + }, + { + "epoch": 2.14, + "learning_rate": 1.437740446268274e-05, + "loss": 0.3381, + "step": 55607 + }, + { + "epoch": 2.14, + "learning_rate": 1.4376763272634008e-05, + "loss": 0.1432, + "step": 55608 + }, + { + "epoch": 2.14, + "learning_rate": 1.4376122082585278e-05, + "loss": 0.2068, + "step": 55609 + }, + { + "epoch": 2.14, + "learning_rate": 1.437548089253655e-05, + "loss": 0.113, + "step": 55610 + }, + { + "epoch": 2.14, + "learning_rate": 1.4374839702487818e-05, + "loss": 0.2007, + "step": 55611 + }, + { + "epoch": 2.14, + "learning_rate": 1.4374198512439088e-05, + "loss": 0.215, + "step": 55612 + }, + { + "epoch": 2.14, + "learning_rate": 1.4373557322390357e-05, + "loss": 0.2761, + "step": 55613 + }, + { + "epoch": 2.14, + "learning_rate": 1.4372916132341627e-05, + "loss": 0.1786, + "step": 55614 + }, + { + "epoch": 2.14, + "learning_rate": 1.4372274942292897e-05, + "loss": 0.2967, + "step": 55615 + }, + { + "epoch": 2.14, + "learning_rate": 1.4371633752244165e-05, + "loss": 0.2249, + "step": 55616 + }, + { + "epoch": 2.14, + "learning_rate": 1.4370992562195435e-05, + "loss": 0.1611, + "step": 55617 + }, + { + "epoch": 2.14, + "learning_rate": 1.4370351372146703e-05, + "loss": 0.186, + "step": 55618 + }, + { + "epoch": 2.14, + "learning_rate": 1.4369710182097975e-05, + "loss": 0.2565, + "step": 55619 + }, + { + "epoch": 2.14, + "learning_rate": 1.4369068992049245e-05, + "loss": 0.1734, + "step": 55620 + }, + { + "epoch": 2.14, + "learning_rate": 1.4368427802000514e-05, + "loss": 0.3001, + "step": 55621 + }, + { + "epoch": 2.14, + "learning_rate": 1.4367786611951784e-05, + "loss": 0.4167, + "step": 55622 + }, + { + "epoch": 2.14, + "learning_rate": 1.4367145421903052e-05, + "loss": 0.2391, + "step": 55623 + }, + { + "epoch": 2.14, + "learning_rate": 1.4366504231854322e-05, + "loss": 0.1393, + "step": 55624 + }, + { + "epoch": 2.14, + "learning_rate": 1.4365863041805592e-05, + "loss": 0.2548, + "step": 55625 + }, + { + "epoch": 2.14, + "learning_rate": 1.436522185175686e-05, + "loss": 0.2975, + "step": 55626 + }, + { + "epoch": 2.14, + "learning_rate": 1.436458066170813e-05, + "loss": 0.1614, + "step": 55627 + }, + { + "epoch": 2.14, + "learning_rate": 1.4363939471659402e-05, + "loss": 0.217, + "step": 55628 + }, + { + "epoch": 2.14, + "learning_rate": 1.436329828161067e-05, + "loss": 0.1822, + "step": 55629 + }, + { + "epoch": 2.14, + "learning_rate": 1.436265709156194e-05, + "loss": 0.2204, + "step": 55630 + }, + { + "epoch": 2.14, + "learning_rate": 1.4362015901513209e-05, + "loss": 0.0936, + "step": 55631 + }, + { + "epoch": 2.14, + "learning_rate": 1.4361374711464479e-05, + "loss": 0.2426, + "step": 55632 + }, + { + "epoch": 2.14, + "learning_rate": 1.4360733521415749e-05, + "loss": 0.2183, + "step": 55633 + }, + { + "epoch": 2.14, + "learning_rate": 1.4360092331367017e-05, + "loss": 0.254, + "step": 55634 + }, + { + "epoch": 2.14, + "learning_rate": 1.4359451141318287e-05, + "loss": 0.2257, + "step": 55635 + }, + { + "epoch": 2.14, + "learning_rate": 1.4358809951269555e-05, + "loss": 0.1759, + "step": 55636 + }, + { + "epoch": 2.14, + "learning_rate": 1.4358168761220827e-05, + "loss": 0.1636, + "step": 55637 + }, + { + "epoch": 2.14, + "learning_rate": 1.4357527571172097e-05, + "loss": 0.2052, + "step": 55638 + }, + { + "epoch": 2.14, + "learning_rate": 1.4356886381123365e-05, + "loss": 0.27, + "step": 55639 + }, + { + "epoch": 2.14, + "learning_rate": 1.4356245191074635e-05, + "loss": 0.2164, + "step": 55640 + }, + { + "epoch": 2.14, + "learning_rate": 1.4355604001025904e-05, + "loss": 0.1253, + "step": 55641 + }, + { + "epoch": 2.14, + "learning_rate": 1.4354962810977174e-05, + "loss": 0.2049, + "step": 55642 + }, + { + "epoch": 2.14, + "learning_rate": 1.4354321620928444e-05, + "loss": 0.2431, + "step": 55643 + }, + { + "epoch": 2.14, + "learning_rate": 1.4353680430879712e-05, + "loss": 0.2926, + "step": 55644 + }, + { + "epoch": 2.14, + "learning_rate": 1.4353039240830982e-05, + "loss": 0.2925, + "step": 55645 + }, + { + "epoch": 2.14, + "learning_rate": 1.4352398050782254e-05, + "loss": 0.2608, + "step": 55646 + }, + { + "epoch": 2.14, + "learning_rate": 1.4351756860733522e-05, + "loss": 0.1911, + "step": 55647 + }, + { + "epoch": 2.14, + "learning_rate": 1.4351115670684792e-05, + "loss": 0.2343, + "step": 55648 + }, + { + "epoch": 2.14, + "learning_rate": 1.435047448063606e-05, + "loss": 0.2291, + "step": 55649 + }, + { + "epoch": 2.14, + "learning_rate": 1.434983329058733e-05, + "loss": 0.3313, + "step": 55650 + }, + { + "epoch": 2.14, + "learning_rate": 1.43491921005386e-05, + "loss": 0.1655, + "step": 55651 + }, + { + "epoch": 2.14, + "learning_rate": 1.4348550910489869e-05, + "loss": 0.3466, + "step": 55652 + }, + { + "epoch": 2.14, + "learning_rate": 1.4347909720441139e-05, + "loss": 0.2207, + "step": 55653 + }, + { + "epoch": 2.14, + "learning_rate": 1.434726853039241e-05, + "loss": 0.2561, + "step": 55654 + }, + { + "epoch": 2.14, + "learning_rate": 1.4346627340343679e-05, + "loss": 0.1696, + "step": 55655 + }, + { + "epoch": 2.14, + "learning_rate": 1.4345986150294949e-05, + "loss": 0.3409, + "step": 55656 + }, + { + "epoch": 2.14, + "learning_rate": 1.4345344960246217e-05, + "loss": 0.2003, + "step": 55657 + }, + { + "epoch": 2.14, + "learning_rate": 1.4344703770197487e-05, + "loss": 0.1957, + "step": 55658 + }, + { + "epoch": 2.14, + "learning_rate": 1.4344062580148757e-05, + "loss": 0.2395, + "step": 55659 + }, + { + "epoch": 2.14, + "learning_rate": 1.4343421390100026e-05, + "loss": 0.2691, + "step": 55660 + }, + { + "epoch": 2.14, + "learning_rate": 1.4342780200051296e-05, + "loss": 0.202, + "step": 55661 + }, + { + "epoch": 2.14, + "learning_rate": 1.4342139010002564e-05, + "loss": 0.2206, + "step": 55662 + }, + { + "epoch": 2.14, + "learning_rate": 1.4341497819953836e-05, + "loss": 0.2903, + "step": 55663 + }, + { + "epoch": 2.14, + "learning_rate": 1.4340856629905106e-05, + "loss": 0.3159, + "step": 55664 + }, + { + "epoch": 2.14, + "learning_rate": 1.4340215439856374e-05, + "loss": 0.2762, + "step": 55665 + }, + { + "epoch": 2.14, + "learning_rate": 1.4339574249807644e-05, + "loss": 0.1795, + "step": 55666 + }, + { + "epoch": 2.14, + "learning_rate": 1.4338933059758913e-05, + "loss": 0.1543, + "step": 55667 + }, + { + "epoch": 2.14, + "learning_rate": 1.4338291869710183e-05, + "loss": 0.1787, + "step": 55668 + }, + { + "epoch": 2.14, + "learning_rate": 1.4337650679661453e-05, + "loss": 0.1733, + "step": 55669 + }, + { + "epoch": 2.14, + "learning_rate": 1.4337009489612721e-05, + "loss": 0.3496, + "step": 55670 + }, + { + "epoch": 2.14, + "learning_rate": 1.4336368299563991e-05, + "loss": 0.2839, + "step": 55671 + }, + { + "epoch": 2.14, + "learning_rate": 1.4335727109515263e-05, + "loss": 0.3042, + "step": 55672 + }, + { + "epoch": 2.14, + "learning_rate": 1.4335085919466531e-05, + "loss": 0.1773, + "step": 55673 + }, + { + "epoch": 2.14, + "learning_rate": 1.4334444729417801e-05, + "loss": 0.1846, + "step": 55674 + }, + { + "epoch": 2.14, + "learning_rate": 1.433380353936907e-05, + "loss": 0.1237, + "step": 55675 + }, + { + "epoch": 2.14, + "learning_rate": 1.433316234932034e-05, + "loss": 0.1928, + "step": 55676 + }, + { + "epoch": 2.14, + "learning_rate": 1.433252115927161e-05, + "loss": 0.2458, + "step": 55677 + }, + { + "epoch": 2.14, + "learning_rate": 1.4331879969222878e-05, + "loss": 0.2455, + "step": 55678 + }, + { + "epoch": 2.14, + "learning_rate": 1.4331238779174148e-05, + "loss": 0.2582, + "step": 55679 + }, + { + "epoch": 2.14, + "learning_rate": 1.4330597589125416e-05, + "loss": 0.3679, + "step": 55680 + }, + { + "epoch": 2.14, + "learning_rate": 1.4329956399076688e-05, + "loss": 0.3602, + "step": 55681 + }, + { + "epoch": 2.14, + "learning_rate": 1.4329315209027958e-05, + "loss": 0.2276, + "step": 55682 + }, + { + "epoch": 2.14, + "learning_rate": 1.4328674018979226e-05, + "loss": 0.1389, + "step": 55683 + }, + { + "epoch": 2.14, + "learning_rate": 1.4328032828930496e-05, + "loss": 0.2387, + "step": 55684 + }, + { + "epoch": 2.14, + "learning_rate": 1.4327391638881764e-05, + "loss": 0.3731, + "step": 55685 + }, + { + "epoch": 2.14, + "learning_rate": 1.4326750448833034e-05, + "loss": 0.1928, + "step": 55686 + }, + { + "epoch": 2.14, + "learning_rate": 1.4326109258784305e-05, + "loss": 0.2781, + "step": 55687 + }, + { + "epoch": 2.14, + "learning_rate": 1.4325468068735573e-05, + "loss": 0.2314, + "step": 55688 + }, + { + "epoch": 2.14, + "learning_rate": 1.4324826878686843e-05, + "loss": 0.1752, + "step": 55689 + }, + { + "epoch": 2.14, + "learning_rate": 1.4324185688638115e-05, + "loss": 0.2321, + "step": 55690 + }, + { + "epoch": 2.14, + "learning_rate": 1.4323544498589383e-05, + "loss": 0.1857, + "step": 55691 + }, + { + "epoch": 2.14, + "learning_rate": 1.4322903308540653e-05, + "loss": 0.2454, + "step": 55692 + }, + { + "epoch": 2.14, + "learning_rate": 1.4322262118491921e-05, + "loss": 0.294, + "step": 55693 + }, + { + "epoch": 2.14, + "learning_rate": 1.4321620928443191e-05, + "loss": 0.2153, + "step": 55694 + }, + { + "epoch": 2.14, + "learning_rate": 1.4320979738394461e-05, + "loss": 0.2214, + "step": 55695 + }, + { + "epoch": 2.14, + "learning_rate": 1.432033854834573e-05, + "loss": 0.1078, + "step": 55696 + }, + { + "epoch": 2.14, + "learning_rate": 1.4319697358297e-05, + "loss": 0.1637, + "step": 55697 + }, + { + "epoch": 2.14, + "learning_rate": 1.4319056168248268e-05, + "loss": 0.2544, + "step": 55698 + }, + { + "epoch": 2.14, + "learning_rate": 1.431841497819954e-05, + "loss": 0.1696, + "step": 55699 + }, + { + "epoch": 2.14, + "learning_rate": 1.431777378815081e-05, + "loss": 0.309, + "step": 55700 + }, + { + "epoch": 2.14, + "learning_rate": 1.4317132598102078e-05, + "loss": 0.1974, + "step": 55701 + }, + { + "epoch": 2.14, + "learning_rate": 1.4316491408053348e-05, + "loss": 0.3106, + "step": 55702 + }, + { + "epoch": 2.14, + "learning_rate": 1.4315850218004616e-05, + "loss": 0.1934, + "step": 55703 + }, + { + "epoch": 2.14, + "learning_rate": 1.4315209027955886e-05, + "loss": 0.3072, + "step": 55704 + }, + { + "epoch": 2.14, + "learning_rate": 1.4314567837907156e-05, + "loss": 0.0839, + "step": 55705 + }, + { + "epoch": 2.14, + "learning_rate": 1.4313926647858425e-05, + "loss": 0.1996, + "step": 55706 + }, + { + "epoch": 2.14, + "learning_rate": 1.4313285457809695e-05, + "loss": 0.2179, + "step": 55707 + }, + { + "epoch": 2.14, + "learning_rate": 1.4312644267760967e-05, + "loss": 0.1685, + "step": 55708 + }, + { + "epoch": 2.14, + "learning_rate": 1.4312003077712235e-05, + "loss": 0.2559, + "step": 55709 + }, + { + "epoch": 2.14, + "learning_rate": 1.4311361887663505e-05, + "loss": 0.3421, + "step": 55710 + }, + { + "epoch": 2.14, + "learning_rate": 1.4310720697614773e-05, + "loss": 0.3511, + "step": 55711 + }, + { + "epoch": 2.14, + "learning_rate": 1.4310079507566043e-05, + "loss": 0.3193, + "step": 55712 + }, + { + "epoch": 2.14, + "learning_rate": 1.4309438317517313e-05, + "loss": 0.1078, + "step": 55713 + }, + { + "epoch": 2.14, + "learning_rate": 1.4308797127468582e-05, + "loss": 0.1779, + "step": 55714 + }, + { + "epoch": 2.14, + "learning_rate": 1.4308155937419852e-05, + "loss": 0.2768, + "step": 55715 + }, + { + "epoch": 2.14, + "learning_rate": 1.430751474737112e-05, + "loss": 0.1303, + "step": 55716 + }, + { + "epoch": 2.14, + "learning_rate": 1.4306873557322392e-05, + "loss": 0.2122, + "step": 55717 + }, + { + "epoch": 2.14, + "learning_rate": 1.4306232367273662e-05, + "loss": 0.2087, + "step": 55718 + }, + { + "epoch": 2.14, + "learning_rate": 1.430559117722493e-05, + "loss": 0.2138, + "step": 55719 + }, + { + "epoch": 2.14, + "learning_rate": 1.43049499871762e-05, + "loss": 0.2071, + "step": 55720 + }, + { + "epoch": 2.14, + "learning_rate": 1.430430879712747e-05, + "loss": 0.1514, + "step": 55721 + }, + { + "epoch": 2.14, + "learning_rate": 1.4303667607078738e-05, + "loss": 0.3489, + "step": 55722 + }, + { + "epoch": 2.14, + "learning_rate": 1.4303026417030008e-05, + "loss": 0.2335, + "step": 55723 + }, + { + "epoch": 2.14, + "learning_rate": 1.4302385226981277e-05, + "loss": 0.3334, + "step": 55724 + }, + { + "epoch": 2.14, + "learning_rate": 1.4301744036932547e-05, + "loss": 0.377, + "step": 55725 + }, + { + "epoch": 2.14, + "learning_rate": 1.4301102846883818e-05, + "loss": 0.385, + "step": 55726 + }, + { + "epoch": 2.14, + "learning_rate": 1.4300461656835087e-05, + "loss": 0.1554, + "step": 55727 + }, + { + "epoch": 2.14, + "learning_rate": 1.4299820466786357e-05, + "loss": 0.1928, + "step": 55728 + }, + { + "epoch": 2.14, + "learning_rate": 1.4299179276737625e-05, + "loss": 0.1301, + "step": 55729 + }, + { + "epoch": 2.14, + "learning_rate": 1.4298538086688895e-05, + "loss": 0.2921, + "step": 55730 + }, + { + "epoch": 2.14, + "learning_rate": 1.4297896896640165e-05, + "loss": 0.2033, + "step": 55731 + }, + { + "epoch": 2.14, + "learning_rate": 1.4297255706591433e-05, + "loss": 0.1372, + "step": 55732 + }, + { + "epoch": 2.14, + "learning_rate": 1.4296614516542704e-05, + "loss": 0.1675, + "step": 55733 + }, + { + "epoch": 2.14, + "learning_rate": 1.4295973326493972e-05, + "loss": 0.2111, + "step": 55734 + }, + { + "epoch": 2.14, + "learning_rate": 1.4295332136445244e-05, + "loss": 0.2138, + "step": 55735 + }, + { + "epoch": 2.14, + "learning_rate": 1.4294690946396514e-05, + "loss": 0.1901, + "step": 55736 + }, + { + "epoch": 2.14, + "learning_rate": 1.4294049756347782e-05, + "loss": 0.2356, + "step": 55737 + }, + { + "epoch": 2.14, + "learning_rate": 1.4293408566299052e-05, + "loss": 0.3777, + "step": 55738 + }, + { + "epoch": 2.14, + "learning_rate": 1.4292767376250322e-05, + "loss": 0.3085, + "step": 55739 + }, + { + "epoch": 2.14, + "learning_rate": 1.429212618620159e-05, + "loss": 0.269, + "step": 55740 + }, + { + "epoch": 2.14, + "learning_rate": 1.429148499615286e-05, + "loss": 0.1912, + "step": 55741 + }, + { + "epoch": 2.14, + "learning_rate": 1.4290843806104129e-05, + "loss": 0.2729, + "step": 55742 + }, + { + "epoch": 2.14, + "learning_rate": 1.4290202616055399e-05, + "loss": 0.2522, + "step": 55743 + }, + { + "epoch": 2.14, + "learning_rate": 1.428956142600667e-05, + "loss": 0.3028, + "step": 55744 + }, + { + "epoch": 2.14, + "learning_rate": 1.4288920235957939e-05, + "loss": 0.2433, + "step": 55745 + }, + { + "epoch": 2.14, + "learning_rate": 1.4288279045909209e-05, + "loss": 0.0905, + "step": 55746 + }, + { + "epoch": 2.14, + "learning_rate": 1.4287637855860477e-05, + "loss": 0.1701, + "step": 55747 + }, + { + "epoch": 2.14, + "learning_rate": 1.4286996665811747e-05, + "loss": 0.1329, + "step": 55748 + }, + { + "epoch": 2.14, + "learning_rate": 1.4286355475763017e-05, + "loss": 0.1827, + "step": 55749 + }, + { + "epoch": 2.14, + "learning_rate": 1.4285714285714285e-05, + "loss": 0.1945, + "step": 55750 + }, + { + "epoch": 2.14, + "learning_rate": 1.4285073095665555e-05, + "loss": 0.1947, + "step": 55751 + }, + { + "epoch": 2.14, + "learning_rate": 1.4284431905616827e-05, + "loss": 0.3224, + "step": 55752 + }, + { + "epoch": 2.14, + "learning_rate": 1.4283790715568095e-05, + "loss": 0.258, + "step": 55753 + }, + { + "epoch": 2.14, + "learning_rate": 1.4283149525519366e-05, + "loss": 0.1503, + "step": 55754 + }, + { + "epoch": 2.14, + "learning_rate": 1.4282508335470634e-05, + "loss": 0.1319, + "step": 55755 + }, + { + "epoch": 2.14, + "learning_rate": 1.4281867145421904e-05, + "loss": 0.1846, + "step": 55756 + }, + { + "epoch": 2.14, + "learning_rate": 1.4281225955373174e-05, + "loss": 0.1475, + "step": 55757 + }, + { + "epoch": 2.14, + "learning_rate": 1.4280584765324442e-05, + "loss": 0.1911, + "step": 55758 + }, + { + "epoch": 2.14, + "learning_rate": 1.4279943575275712e-05, + "loss": 0.2192, + "step": 55759 + }, + { + "epoch": 2.14, + "learning_rate": 1.427930238522698e-05, + "loss": 0.3332, + "step": 55760 + }, + { + "epoch": 2.14, + "learning_rate": 1.4278661195178252e-05, + "loss": 0.2477, + "step": 55761 + }, + { + "epoch": 2.14, + "learning_rate": 1.4278020005129522e-05, + "loss": 0.403, + "step": 55762 + }, + { + "epoch": 2.14, + "learning_rate": 1.427737881508079e-05, + "loss": 0.341, + "step": 55763 + }, + { + "epoch": 2.14, + "learning_rate": 1.427673762503206e-05, + "loss": 0.2406, + "step": 55764 + }, + { + "epoch": 2.14, + "learning_rate": 1.4276096434983329e-05, + "loss": 0.2151, + "step": 55765 + }, + { + "epoch": 2.14, + "learning_rate": 1.4275455244934599e-05, + "loss": 0.1131, + "step": 55766 + }, + { + "epoch": 2.14, + "learning_rate": 1.4274814054885869e-05, + "loss": 0.1783, + "step": 55767 + }, + { + "epoch": 2.14, + "learning_rate": 1.4274172864837137e-05, + "loss": 0.2248, + "step": 55768 + }, + { + "epoch": 2.14, + "learning_rate": 1.4273531674788407e-05, + "loss": 0.4117, + "step": 55769 + }, + { + "epoch": 2.14, + "learning_rate": 1.4272890484739679e-05, + "loss": 0.1678, + "step": 55770 + }, + { + "epoch": 2.14, + "learning_rate": 1.4272249294690947e-05, + "loss": 0.183, + "step": 55771 + }, + { + "epoch": 2.14, + "learning_rate": 1.4271608104642217e-05, + "loss": 0.2562, + "step": 55772 + }, + { + "epoch": 2.14, + "learning_rate": 1.4270966914593486e-05, + "loss": 0.2979, + "step": 55773 + }, + { + "epoch": 2.14, + "learning_rate": 1.4270325724544756e-05, + "loss": 0.1441, + "step": 55774 + }, + { + "epoch": 2.14, + "learning_rate": 1.4269684534496026e-05, + "loss": 0.2687, + "step": 55775 + }, + { + "epoch": 2.14, + "learning_rate": 1.4269043344447294e-05, + "loss": 0.2416, + "step": 55776 + }, + { + "epoch": 2.14, + "learning_rate": 1.4268402154398564e-05, + "loss": 0.2902, + "step": 55777 + }, + { + "epoch": 2.14, + "learning_rate": 1.4267760964349833e-05, + "loss": 0.2781, + "step": 55778 + }, + { + "epoch": 2.14, + "learning_rate": 1.4267119774301104e-05, + "loss": 0.2933, + "step": 55779 + }, + { + "epoch": 2.14, + "learning_rate": 1.4266478584252374e-05, + "loss": 0.2841, + "step": 55780 + }, + { + "epoch": 2.14, + "learning_rate": 1.4265837394203643e-05, + "loss": 0.1491, + "step": 55781 + }, + { + "epoch": 2.14, + "learning_rate": 1.4265196204154913e-05, + "loss": 0.2232, + "step": 55782 + }, + { + "epoch": 2.14, + "learning_rate": 1.4264555014106183e-05, + "loss": 0.1744, + "step": 55783 + }, + { + "epoch": 2.14, + "learning_rate": 1.4263913824057451e-05, + "loss": 0.2834, + "step": 55784 + }, + { + "epoch": 2.14, + "learning_rate": 1.4263272634008721e-05, + "loss": 0.2241, + "step": 55785 + }, + { + "epoch": 2.14, + "learning_rate": 1.426263144395999e-05, + "loss": 0.2596, + "step": 55786 + }, + { + "epoch": 2.14, + "learning_rate": 1.426199025391126e-05, + "loss": 0.3455, + "step": 55787 + }, + { + "epoch": 2.14, + "learning_rate": 1.4261349063862531e-05, + "loss": 0.2142, + "step": 55788 + }, + { + "epoch": 2.14, + "learning_rate": 1.42607078738138e-05, + "loss": 0.2927, + "step": 55789 + }, + { + "epoch": 2.14, + "learning_rate": 1.426006668376507e-05, + "loss": 0.4035, + "step": 55790 + }, + { + "epoch": 2.14, + "learning_rate": 1.4259425493716338e-05, + "loss": 0.2244, + "step": 55791 + }, + { + "epoch": 2.15, + "learning_rate": 1.4258784303667608e-05, + "loss": 0.1176, + "step": 55792 + }, + { + "epoch": 2.15, + "learning_rate": 1.4258143113618878e-05, + "loss": 0.2134, + "step": 55793 + }, + { + "epoch": 2.15, + "learning_rate": 1.4257501923570146e-05, + "loss": 0.1234, + "step": 55794 + }, + { + "epoch": 2.15, + "learning_rate": 1.4256860733521416e-05, + "loss": 0.1835, + "step": 55795 + }, + { + "epoch": 2.15, + "learning_rate": 1.4256219543472684e-05, + "loss": 0.2099, + "step": 55796 + }, + { + "epoch": 2.15, + "learning_rate": 1.4255578353423956e-05, + "loss": 0.2317, + "step": 55797 + }, + { + "epoch": 2.15, + "learning_rate": 1.4254937163375226e-05, + "loss": 0.1127, + "step": 55798 + }, + { + "epoch": 2.15, + "learning_rate": 1.4254295973326495e-05, + "loss": 0.2902, + "step": 55799 + }, + { + "epoch": 2.15, + "learning_rate": 1.4253654783277765e-05, + "loss": 0.225, + "step": 55800 + }, + { + "epoch": 2.15, + "learning_rate": 1.4253013593229035e-05, + "loss": 0.241, + "step": 55801 + }, + { + "epoch": 2.15, + "learning_rate": 1.4252372403180303e-05, + "loss": 0.2241, + "step": 55802 + }, + { + "epoch": 2.15, + "learning_rate": 1.4251731213131573e-05, + "loss": 0.3006, + "step": 55803 + }, + { + "epoch": 2.15, + "learning_rate": 1.4251090023082841e-05, + "loss": 0.3908, + "step": 55804 + }, + { + "epoch": 2.15, + "learning_rate": 1.4250448833034111e-05, + "loss": 0.2515, + "step": 55805 + }, + { + "epoch": 2.15, + "learning_rate": 1.4249807642985383e-05, + "loss": 0.3208, + "step": 55806 + }, + { + "epoch": 2.15, + "learning_rate": 1.4249166452936651e-05, + "loss": 0.2514, + "step": 55807 + }, + { + "epoch": 2.15, + "learning_rate": 1.4248525262887921e-05, + "loss": 0.2921, + "step": 55808 + }, + { + "epoch": 2.15, + "learning_rate": 1.424788407283919e-05, + "loss": 0.1532, + "step": 55809 + }, + { + "epoch": 2.15, + "learning_rate": 1.424724288279046e-05, + "loss": 0.2304, + "step": 55810 + }, + { + "epoch": 2.15, + "learning_rate": 1.424660169274173e-05, + "loss": 0.3621, + "step": 55811 + }, + { + "epoch": 2.15, + "learning_rate": 1.4245960502692998e-05, + "loss": 0.3733, + "step": 55812 + }, + { + "epoch": 2.15, + "learning_rate": 1.4245319312644268e-05, + "loss": 0.2424, + "step": 55813 + }, + { + "epoch": 2.15, + "learning_rate": 1.4244678122595536e-05, + "loss": 0.2508, + "step": 55814 + }, + { + "epoch": 2.15, + "learning_rate": 1.4244036932546808e-05, + "loss": 0.2212, + "step": 55815 + }, + { + "epoch": 2.15, + "learning_rate": 1.4243395742498078e-05, + "loss": 0.2364, + "step": 55816 + }, + { + "epoch": 2.15, + "learning_rate": 1.4242754552449346e-05, + "loss": 0.1022, + "step": 55817 + }, + { + "epoch": 2.15, + "learning_rate": 1.4242113362400616e-05, + "loss": 0.2165, + "step": 55818 + }, + { + "epoch": 2.15, + "learning_rate": 1.4241472172351886e-05, + "loss": 0.2074, + "step": 55819 + }, + { + "epoch": 2.15, + "learning_rate": 1.4240830982303155e-05, + "loss": 0.2507, + "step": 55820 + }, + { + "epoch": 2.15, + "learning_rate": 1.4240189792254425e-05, + "loss": 0.2587, + "step": 55821 + }, + { + "epoch": 2.15, + "learning_rate": 1.4239548602205693e-05, + "loss": 0.1722, + "step": 55822 + }, + { + "epoch": 2.15, + "learning_rate": 1.4238907412156963e-05, + "loss": 0.1556, + "step": 55823 + }, + { + "epoch": 2.15, + "learning_rate": 1.4238266222108235e-05, + "loss": 0.2846, + "step": 55824 + }, + { + "epoch": 2.15, + "learning_rate": 1.4237625032059503e-05, + "loss": 0.2032, + "step": 55825 + }, + { + "epoch": 2.15, + "learning_rate": 1.4236983842010773e-05, + "loss": 0.1991, + "step": 55826 + }, + { + "epoch": 2.15, + "learning_rate": 1.4236342651962042e-05, + "loss": 0.1811, + "step": 55827 + }, + { + "epoch": 2.15, + "learning_rate": 1.4235701461913312e-05, + "loss": 0.1788, + "step": 55828 + }, + { + "epoch": 2.15, + "learning_rate": 1.4235060271864582e-05, + "loss": 0.1306, + "step": 55829 + }, + { + "epoch": 2.15, + "learning_rate": 1.423441908181585e-05, + "loss": 0.1843, + "step": 55830 + }, + { + "epoch": 2.15, + "learning_rate": 1.423377789176712e-05, + "loss": 0.2112, + "step": 55831 + }, + { + "epoch": 2.15, + "learning_rate": 1.4233136701718388e-05, + "loss": 0.2857, + "step": 55832 + }, + { + "epoch": 2.15, + "learning_rate": 1.423249551166966e-05, + "loss": 0.1969, + "step": 55833 + }, + { + "epoch": 2.15, + "learning_rate": 1.423185432162093e-05, + "loss": 0.1551, + "step": 55834 + }, + { + "epoch": 2.15, + "learning_rate": 1.4231213131572198e-05, + "loss": 0.2616, + "step": 55835 + }, + { + "epoch": 2.15, + "learning_rate": 1.4230571941523468e-05, + "loss": 0.1727, + "step": 55836 + }, + { + "epoch": 2.15, + "learning_rate": 1.4229930751474738e-05, + "loss": 0.1549, + "step": 55837 + }, + { + "epoch": 2.15, + "learning_rate": 1.4229289561426007e-05, + "loss": 0.3012, + "step": 55838 + }, + { + "epoch": 2.15, + "learning_rate": 1.4228648371377277e-05, + "loss": 0.2931, + "step": 55839 + }, + { + "epoch": 2.15, + "learning_rate": 1.4228007181328545e-05, + "loss": 0.2384, + "step": 55840 + }, + { + "epoch": 2.15, + "learning_rate": 1.4227365991279817e-05, + "loss": 0.1913, + "step": 55841 + }, + { + "epoch": 2.15, + "learning_rate": 1.4226724801231087e-05, + "loss": 0.1747, + "step": 55842 + }, + { + "epoch": 2.15, + "learning_rate": 1.4226083611182355e-05, + "loss": 0.2563, + "step": 55843 + }, + { + "epoch": 2.15, + "learning_rate": 1.4225442421133625e-05, + "loss": 0.3369, + "step": 55844 + }, + { + "epoch": 2.15, + "learning_rate": 1.4224801231084894e-05, + "loss": 0.1175, + "step": 55845 + }, + { + "epoch": 2.15, + "learning_rate": 1.4224160041036164e-05, + "loss": 0.1817, + "step": 55846 + }, + { + "epoch": 2.15, + "learning_rate": 1.4223518850987434e-05, + "loss": 0.2376, + "step": 55847 + }, + { + "epoch": 2.15, + "learning_rate": 1.4222877660938702e-05, + "loss": 0.1706, + "step": 55848 + }, + { + "epoch": 2.15, + "learning_rate": 1.4222236470889972e-05, + "loss": 0.254, + "step": 55849 + }, + { + "epoch": 2.15, + "learning_rate": 1.4221595280841244e-05, + "loss": 0.2106, + "step": 55850 + }, + { + "epoch": 2.15, + "learning_rate": 1.4220954090792512e-05, + "loss": 0.2586, + "step": 55851 + }, + { + "epoch": 2.15, + "learning_rate": 1.4220312900743782e-05, + "loss": 0.2281, + "step": 55852 + }, + { + "epoch": 2.15, + "learning_rate": 1.421967171069505e-05, + "loss": 0.2663, + "step": 55853 + }, + { + "epoch": 2.15, + "learning_rate": 1.421903052064632e-05, + "loss": 0.2653, + "step": 55854 + }, + { + "epoch": 2.15, + "learning_rate": 1.421838933059759e-05, + "loss": 0.3558, + "step": 55855 + }, + { + "epoch": 2.15, + "learning_rate": 1.4217748140548859e-05, + "loss": 0.3239, + "step": 55856 + }, + { + "epoch": 2.15, + "learning_rate": 1.4217106950500129e-05, + "loss": 0.2502, + "step": 55857 + }, + { + "epoch": 2.15, + "learning_rate": 1.4216465760451397e-05, + "loss": 0.1939, + "step": 55858 + }, + { + "epoch": 2.15, + "learning_rate": 1.4215824570402669e-05, + "loss": 0.4367, + "step": 55859 + }, + { + "epoch": 2.15, + "learning_rate": 1.4215183380353939e-05, + "loss": 0.3262, + "step": 55860 + }, + { + "epoch": 2.15, + "learning_rate": 1.4214542190305207e-05, + "loss": 0.1915, + "step": 55861 + }, + { + "epoch": 2.15, + "learning_rate": 1.4213901000256477e-05, + "loss": 0.2048, + "step": 55862 + }, + { + "epoch": 2.15, + "learning_rate": 1.4213259810207747e-05, + "loss": 0.1143, + "step": 55863 + }, + { + "epoch": 2.15, + "learning_rate": 1.4212618620159015e-05, + "loss": 0.3622, + "step": 55864 + }, + { + "epoch": 2.15, + "learning_rate": 1.4211977430110286e-05, + "loss": 0.3626, + "step": 55865 + }, + { + "epoch": 2.15, + "learning_rate": 1.4211336240061554e-05, + "loss": 0.246, + "step": 55866 + }, + { + "epoch": 2.15, + "learning_rate": 1.4210695050012824e-05, + "loss": 0.2807, + "step": 55867 + }, + { + "epoch": 2.15, + "learning_rate": 1.4210053859964096e-05, + "loss": 0.1016, + "step": 55868 + }, + { + "epoch": 2.15, + "learning_rate": 1.4209412669915364e-05, + "loss": 0.2977, + "step": 55869 + }, + { + "epoch": 2.15, + "learning_rate": 1.4208771479866634e-05, + "loss": 0.2362, + "step": 55870 + }, + { + "epoch": 2.15, + "learning_rate": 1.4208130289817902e-05, + "loss": 0.2004, + "step": 55871 + }, + { + "epoch": 2.15, + "learning_rate": 1.4207489099769172e-05, + "loss": 0.2001, + "step": 55872 + }, + { + "epoch": 2.15, + "learning_rate": 1.4206847909720442e-05, + "loss": 0.1584, + "step": 55873 + }, + { + "epoch": 2.15, + "learning_rate": 1.420620671967171e-05, + "loss": 0.2952, + "step": 55874 + }, + { + "epoch": 2.15, + "learning_rate": 1.420556552962298e-05, + "loss": 0.1854, + "step": 55875 + }, + { + "epoch": 2.15, + "learning_rate": 1.4204924339574249e-05, + "loss": 0.2689, + "step": 55876 + }, + { + "epoch": 2.15, + "learning_rate": 1.420428314952552e-05, + "loss": 0.2448, + "step": 55877 + }, + { + "epoch": 2.15, + "learning_rate": 1.420364195947679e-05, + "loss": 0.3303, + "step": 55878 + }, + { + "epoch": 2.15, + "learning_rate": 1.4203000769428059e-05, + "loss": 0.3138, + "step": 55879 + }, + { + "epoch": 2.15, + "learning_rate": 1.4202359579379329e-05, + "loss": 0.2802, + "step": 55880 + }, + { + "epoch": 2.15, + "learning_rate": 1.4201718389330599e-05, + "loss": 0.1995, + "step": 55881 + }, + { + "epoch": 2.15, + "learning_rate": 1.4201077199281867e-05, + "loss": 0.2185, + "step": 55882 + }, + { + "epoch": 2.15, + "learning_rate": 1.4200436009233137e-05, + "loss": 0.196, + "step": 55883 + }, + { + "epoch": 2.15, + "learning_rate": 1.4199794819184406e-05, + "loss": 0.1489, + "step": 55884 + }, + { + "epoch": 2.15, + "learning_rate": 1.4199153629135676e-05, + "loss": 0.1642, + "step": 55885 + }, + { + "epoch": 2.15, + "learning_rate": 1.4198512439086948e-05, + "loss": 0.2523, + "step": 55886 + }, + { + "epoch": 2.15, + "learning_rate": 1.4197871249038216e-05, + "loss": 0.2268, + "step": 55887 + }, + { + "epoch": 2.15, + "learning_rate": 1.4197230058989486e-05, + "loss": 0.2893, + "step": 55888 + }, + { + "epoch": 2.15, + "learning_rate": 1.4196588868940754e-05, + "loss": 0.2839, + "step": 55889 + }, + { + "epoch": 2.15, + "learning_rate": 1.4195947678892024e-05, + "loss": 0.143, + "step": 55890 + }, + { + "epoch": 2.15, + "learning_rate": 1.4195306488843294e-05, + "loss": 0.1561, + "step": 55891 + }, + { + "epoch": 2.15, + "learning_rate": 1.4194665298794563e-05, + "loss": 0.2215, + "step": 55892 + }, + { + "epoch": 2.15, + "learning_rate": 1.4194024108745833e-05, + "loss": 0.1558, + "step": 55893 + }, + { + "epoch": 2.15, + "learning_rate": 1.4193382918697101e-05, + "loss": 0.3042, + "step": 55894 + }, + { + "epoch": 2.15, + "learning_rate": 1.4192741728648373e-05, + "loss": 0.1533, + "step": 55895 + }, + { + "epoch": 2.15, + "learning_rate": 1.4192100538599643e-05, + "loss": 0.1783, + "step": 55896 + }, + { + "epoch": 2.15, + "learning_rate": 1.4191459348550911e-05, + "loss": 0.1998, + "step": 55897 + }, + { + "epoch": 2.15, + "learning_rate": 1.4190818158502181e-05, + "loss": 0.221, + "step": 55898 + }, + { + "epoch": 2.15, + "learning_rate": 1.4190176968453451e-05, + "loss": 0.3157, + "step": 55899 + }, + { + "epoch": 2.15, + "learning_rate": 1.418953577840472e-05, + "loss": 0.3409, + "step": 55900 + }, + { + "epoch": 2.15, + "learning_rate": 1.418889458835599e-05, + "loss": 0.2004, + "step": 55901 + }, + { + "epoch": 2.15, + "learning_rate": 1.4188253398307258e-05, + "loss": 0.4749, + "step": 55902 + }, + { + "epoch": 2.15, + "learning_rate": 1.4187612208258528e-05, + "loss": 0.3196, + "step": 55903 + }, + { + "epoch": 2.15, + "learning_rate": 1.41869710182098e-05, + "loss": 0.3233, + "step": 55904 + }, + { + "epoch": 2.15, + "learning_rate": 1.4186329828161068e-05, + "loss": 0.1676, + "step": 55905 + }, + { + "epoch": 2.15, + "learning_rate": 1.4185688638112338e-05, + "loss": 0.2007, + "step": 55906 + }, + { + "epoch": 2.15, + "learning_rate": 1.4185047448063606e-05, + "loss": 0.2102, + "step": 55907 + }, + { + "epoch": 2.15, + "learning_rate": 1.4184406258014876e-05, + "loss": 0.2682, + "step": 55908 + }, + { + "epoch": 2.15, + "learning_rate": 1.4183765067966146e-05, + "loss": 0.1587, + "step": 55909 + }, + { + "epoch": 2.15, + "learning_rate": 1.4183123877917415e-05, + "loss": 0.2115, + "step": 55910 + }, + { + "epoch": 2.15, + "learning_rate": 1.4182482687868685e-05, + "loss": 0.4452, + "step": 55911 + }, + { + "epoch": 2.15, + "learning_rate": 1.4181841497819953e-05, + "loss": 0.3198, + "step": 55912 + }, + { + "epoch": 2.15, + "learning_rate": 1.4181200307771225e-05, + "loss": 0.1976, + "step": 55913 + }, + { + "epoch": 2.15, + "learning_rate": 1.4180559117722495e-05, + "loss": 0.2019, + "step": 55914 + }, + { + "epoch": 2.15, + "learning_rate": 1.4179917927673763e-05, + "loss": 0.271, + "step": 55915 + }, + { + "epoch": 2.15, + "learning_rate": 1.4179276737625033e-05, + "loss": 0.1871, + "step": 55916 + }, + { + "epoch": 2.15, + "learning_rate": 1.4178635547576303e-05, + "loss": 0.1337, + "step": 55917 + }, + { + "epoch": 2.15, + "learning_rate": 1.4177994357527571e-05, + "loss": 0.2376, + "step": 55918 + }, + { + "epoch": 2.15, + "learning_rate": 1.4177353167478841e-05, + "loss": 0.1954, + "step": 55919 + }, + { + "epoch": 2.15, + "learning_rate": 1.417671197743011e-05, + "loss": 0.2552, + "step": 55920 + }, + { + "epoch": 2.15, + "learning_rate": 1.417607078738138e-05, + "loss": 0.1337, + "step": 55921 + }, + { + "epoch": 2.15, + "learning_rate": 1.4175429597332651e-05, + "loss": 0.3833, + "step": 55922 + }, + { + "epoch": 2.15, + "learning_rate": 1.417478840728392e-05, + "loss": 0.1316, + "step": 55923 + }, + { + "epoch": 2.15, + "learning_rate": 1.417414721723519e-05, + "loss": 0.1285, + "step": 55924 + }, + { + "epoch": 2.15, + "learning_rate": 1.417350602718646e-05, + "loss": 0.2067, + "step": 55925 + }, + { + "epoch": 2.15, + "learning_rate": 1.4172864837137728e-05, + "loss": 0.32, + "step": 55926 + }, + { + "epoch": 2.15, + "learning_rate": 1.4172223647088998e-05, + "loss": 0.2755, + "step": 55927 + }, + { + "epoch": 2.15, + "learning_rate": 1.4171582457040266e-05, + "loss": 0.2435, + "step": 55928 + }, + { + "epoch": 2.15, + "learning_rate": 1.4170941266991536e-05, + "loss": 0.2701, + "step": 55929 + }, + { + "epoch": 2.15, + "learning_rate": 1.4170300076942805e-05, + "loss": 0.2108, + "step": 55930 + }, + { + "epoch": 2.15, + "learning_rate": 1.4169658886894077e-05, + "loss": 0.3096, + "step": 55931 + }, + { + "epoch": 2.15, + "learning_rate": 1.4169017696845347e-05, + "loss": 0.0985, + "step": 55932 + }, + { + "epoch": 2.15, + "learning_rate": 1.4168376506796615e-05, + "loss": 0.2627, + "step": 55933 + }, + { + "epoch": 2.15, + "learning_rate": 1.4167735316747885e-05, + "loss": 0.376, + "step": 55934 + }, + { + "epoch": 2.15, + "learning_rate": 1.4167094126699155e-05, + "loss": 0.21, + "step": 55935 + }, + { + "epoch": 2.15, + "learning_rate": 1.4166452936650423e-05, + "loss": 0.2797, + "step": 55936 + }, + { + "epoch": 2.15, + "learning_rate": 1.4165811746601693e-05, + "loss": 0.3116, + "step": 55937 + }, + { + "epoch": 2.15, + "learning_rate": 1.4165170556552962e-05, + "loss": 0.3237, + "step": 55938 + }, + { + "epoch": 2.15, + "learning_rate": 1.4164529366504233e-05, + "loss": 0.2684, + "step": 55939 + }, + { + "epoch": 2.15, + "learning_rate": 1.4163888176455503e-05, + "loss": 0.2084, + "step": 55940 + }, + { + "epoch": 2.15, + "learning_rate": 1.4163246986406772e-05, + "loss": 0.1702, + "step": 55941 + }, + { + "epoch": 2.15, + "learning_rate": 1.4162605796358042e-05, + "loss": 0.1992, + "step": 55942 + }, + { + "epoch": 2.15, + "learning_rate": 1.4161964606309312e-05, + "loss": 0.1774, + "step": 55943 + }, + { + "epoch": 2.15, + "learning_rate": 1.416132341626058e-05, + "loss": 0.2217, + "step": 55944 + }, + { + "epoch": 2.15, + "learning_rate": 1.416068222621185e-05, + "loss": 0.2609, + "step": 55945 + }, + { + "epoch": 2.15, + "learning_rate": 1.4160041036163118e-05, + "loss": 0.1827, + "step": 55946 + }, + { + "epoch": 2.15, + "learning_rate": 1.4159399846114388e-05, + "loss": 0.3333, + "step": 55947 + }, + { + "epoch": 2.15, + "learning_rate": 1.415875865606566e-05, + "loss": 0.2433, + "step": 55948 + }, + { + "epoch": 2.15, + "learning_rate": 1.4158117466016928e-05, + "loss": 0.2565, + "step": 55949 + }, + { + "epoch": 2.15, + "learning_rate": 1.4157476275968198e-05, + "loss": 0.2173, + "step": 55950 + }, + { + "epoch": 2.15, + "learning_rate": 1.4156835085919467e-05, + "loss": 0.0989, + "step": 55951 + }, + { + "epoch": 2.15, + "learning_rate": 1.4156193895870737e-05, + "loss": 0.2529, + "step": 55952 + }, + { + "epoch": 2.15, + "learning_rate": 1.4155552705822007e-05, + "loss": 0.166, + "step": 55953 + }, + { + "epoch": 2.15, + "learning_rate": 1.4154911515773275e-05, + "loss": 0.2203, + "step": 55954 + }, + { + "epoch": 2.15, + "learning_rate": 1.4154270325724545e-05, + "loss": 0.239, + "step": 55955 + }, + { + "epoch": 2.15, + "learning_rate": 1.4153629135675814e-05, + "loss": 0.2618, + "step": 55956 + }, + { + "epoch": 2.15, + "learning_rate": 1.4152987945627085e-05, + "loss": 0.1293, + "step": 55957 + }, + { + "epoch": 2.15, + "learning_rate": 1.4152346755578355e-05, + "loss": 0.1506, + "step": 55958 + }, + { + "epoch": 2.15, + "learning_rate": 1.4151705565529624e-05, + "loss": 0.2575, + "step": 55959 + }, + { + "epoch": 2.15, + "learning_rate": 1.4151064375480894e-05, + "loss": 0.3114, + "step": 55960 + }, + { + "epoch": 2.15, + "learning_rate": 1.4150423185432164e-05, + "loss": 0.3287, + "step": 55961 + }, + { + "epoch": 2.15, + "learning_rate": 1.4149781995383432e-05, + "loss": 0.2056, + "step": 55962 + }, + { + "epoch": 2.15, + "learning_rate": 1.4149140805334702e-05, + "loss": 0.264, + "step": 55963 + }, + { + "epoch": 2.15, + "learning_rate": 1.414849961528597e-05, + "loss": 0.2061, + "step": 55964 + }, + { + "epoch": 2.15, + "learning_rate": 1.414785842523724e-05, + "loss": 0.315, + "step": 55965 + }, + { + "epoch": 2.15, + "learning_rate": 1.4147217235188512e-05, + "loss": 0.1498, + "step": 55966 + }, + { + "epoch": 2.15, + "learning_rate": 1.414657604513978e-05, + "loss": 0.1363, + "step": 55967 + }, + { + "epoch": 2.15, + "learning_rate": 1.414593485509105e-05, + "loss": 0.1368, + "step": 55968 + }, + { + "epoch": 2.15, + "learning_rate": 1.4145293665042319e-05, + "loss": 0.2338, + "step": 55969 + }, + { + "epoch": 2.15, + "learning_rate": 1.4144652474993589e-05, + "loss": 0.3917, + "step": 55970 + }, + { + "epoch": 2.15, + "learning_rate": 1.4144011284944859e-05, + "loss": 0.2755, + "step": 55971 + }, + { + "epoch": 2.15, + "learning_rate": 1.4143370094896127e-05, + "loss": 0.213, + "step": 55972 + }, + { + "epoch": 2.15, + "learning_rate": 1.4142728904847397e-05, + "loss": 0.2464, + "step": 55973 + }, + { + "epoch": 2.15, + "learning_rate": 1.4142087714798665e-05, + "loss": 0.239, + "step": 55974 + }, + { + "epoch": 2.15, + "learning_rate": 1.4141446524749937e-05, + "loss": 0.2421, + "step": 55975 + }, + { + "epoch": 2.15, + "learning_rate": 1.4140805334701207e-05, + "loss": 0.1647, + "step": 55976 + }, + { + "epoch": 2.15, + "learning_rate": 1.4140164144652476e-05, + "loss": 0.3113, + "step": 55977 + }, + { + "epoch": 2.15, + "learning_rate": 1.4139522954603746e-05, + "loss": 0.3152, + "step": 55978 + }, + { + "epoch": 2.15, + "learning_rate": 1.4138881764555016e-05, + "loss": 0.2415, + "step": 55979 + }, + { + "epoch": 2.15, + "learning_rate": 1.4138240574506284e-05, + "loss": 0.1859, + "step": 55980 + }, + { + "epoch": 2.15, + "learning_rate": 1.4137599384457554e-05, + "loss": 0.4062, + "step": 55981 + }, + { + "epoch": 2.15, + "learning_rate": 1.4136958194408822e-05, + "loss": 0.1131, + "step": 55982 + }, + { + "epoch": 2.15, + "learning_rate": 1.4136317004360092e-05, + "loss": 0.2159, + "step": 55983 + }, + { + "epoch": 2.15, + "learning_rate": 1.4135675814311364e-05, + "loss": 0.2902, + "step": 55984 + }, + { + "epoch": 2.15, + "learning_rate": 1.4135034624262632e-05, + "loss": 0.1696, + "step": 55985 + }, + { + "epoch": 2.15, + "learning_rate": 1.4134393434213902e-05, + "loss": 0.247, + "step": 55986 + }, + { + "epoch": 2.15, + "learning_rate": 1.4133752244165172e-05, + "loss": 0.2331, + "step": 55987 + }, + { + "epoch": 2.15, + "learning_rate": 1.413311105411644e-05, + "loss": 0.219, + "step": 55988 + }, + { + "epoch": 2.15, + "learning_rate": 1.413246986406771e-05, + "loss": 0.2364, + "step": 55989 + }, + { + "epoch": 2.15, + "learning_rate": 1.4131828674018979e-05, + "loss": 0.1852, + "step": 55990 + }, + { + "epoch": 2.15, + "learning_rate": 1.4131187483970249e-05, + "loss": 0.2613, + "step": 55991 + }, + { + "epoch": 2.15, + "learning_rate": 1.4130546293921517e-05, + "loss": 0.2689, + "step": 55992 + }, + { + "epoch": 2.15, + "learning_rate": 1.4129905103872789e-05, + "loss": 0.2387, + "step": 55993 + }, + { + "epoch": 2.15, + "learning_rate": 1.4129263913824059e-05, + "loss": 0.1328, + "step": 55994 + }, + { + "epoch": 2.15, + "learning_rate": 1.4128622723775327e-05, + "loss": 0.2646, + "step": 55995 + }, + { + "epoch": 2.15, + "learning_rate": 1.4127981533726597e-05, + "loss": 0.2809, + "step": 55996 + }, + { + "epoch": 2.15, + "learning_rate": 1.4127340343677868e-05, + "loss": 0.1607, + "step": 55997 + }, + { + "epoch": 2.15, + "learning_rate": 1.4126699153629136e-05, + "loss": 0.3218, + "step": 55998 + }, + { + "epoch": 2.15, + "learning_rate": 1.4126057963580406e-05, + "loss": 0.1206, + "step": 55999 + }, + { + "epoch": 2.15, + "learning_rate": 1.4125416773531674e-05, + "loss": 0.19, + "step": 56000 + }, + { + "epoch": 2.15, + "learning_rate": 1.4124775583482944e-05, + "loss": 0.2011, + "step": 56001 + }, + { + "epoch": 2.15, + "learning_rate": 1.4124134393434216e-05, + "loss": 0.2267, + "step": 56002 + }, + { + "epoch": 2.15, + "learning_rate": 1.4123493203385484e-05, + "loss": 0.2354, + "step": 56003 + }, + { + "epoch": 2.15, + "learning_rate": 1.4122852013336754e-05, + "loss": 0.141, + "step": 56004 + }, + { + "epoch": 2.15, + "learning_rate": 1.4122210823288024e-05, + "loss": 0.3256, + "step": 56005 + }, + { + "epoch": 2.15, + "learning_rate": 1.4121569633239293e-05, + "loss": 0.1498, + "step": 56006 + }, + { + "epoch": 2.15, + "learning_rate": 1.4120928443190563e-05, + "loss": 0.1432, + "step": 56007 + }, + { + "epoch": 2.15, + "learning_rate": 1.4120287253141831e-05, + "loss": 0.1934, + "step": 56008 + }, + { + "epoch": 2.15, + "learning_rate": 1.4119646063093101e-05, + "loss": 0.2098, + "step": 56009 + }, + { + "epoch": 2.15, + "learning_rate": 1.411900487304437e-05, + "loss": 0.1256, + "step": 56010 + }, + { + "epoch": 2.15, + "learning_rate": 1.4118363682995641e-05, + "loss": 0.2283, + "step": 56011 + }, + { + "epoch": 2.15, + "learning_rate": 1.4117722492946911e-05, + "loss": 0.2232, + "step": 56012 + }, + { + "epoch": 2.15, + "learning_rate": 1.411708130289818e-05, + "loss": 0.2536, + "step": 56013 + }, + { + "epoch": 2.15, + "learning_rate": 1.411644011284945e-05, + "loss": 0.2509, + "step": 56014 + }, + { + "epoch": 2.15, + "learning_rate": 1.411579892280072e-05, + "loss": 0.4069, + "step": 56015 + }, + { + "epoch": 2.15, + "learning_rate": 1.4115157732751988e-05, + "loss": 0.1455, + "step": 56016 + }, + { + "epoch": 2.15, + "learning_rate": 1.4114516542703258e-05, + "loss": 0.2356, + "step": 56017 + }, + { + "epoch": 2.15, + "learning_rate": 1.4113875352654526e-05, + "loss": 0.3078, + "step": 56018 + }, + { + "epoch": 2.15, + "learning_rate": 1.4113234162605796e-05, + "loss": 0.1804, + "step": 56019 + }, + { + "epoch": 2.15, + "learning_rate": 1.4112592972557068e-05, + "loss": 0.182, + "step": 56020 + }, + { + "epoch": 2.15, + "learning_rate": 1.4111951782508336e-05, + "loss": 0.5277, + "step": 56021 + }, + { + "epoch": 2.15, + "learning_rate": 1.4111310592459606e-05, + "loss": 0.1038, + "step": 56022 + }, + { + "epoch": 2.15, + "learning_rate": 1.4110669402410876e-05, + "loss": 0.2637, + "step": 56023 + }, + { + "epoch": 2.15, + "learning_rate": 1.4110028212362145e-05, + "loss": 0.1707, + "step": 56024 + }, + { + "epoch": 2.15, + "learning_rate": 1.4109387022313415e-05, + "loss": 0.2875, + "step": 56025 + }, + { + "epoch": 2.15, + "learning_rate": 1.4108745832264683e-05, + "loss": 0.2695, + "step": 56026 + }, + { + "epoch": 2.15, + "learning_rate": 1.4108104642215953e-05, + "loss": 0.1781, + "step": 56027 + }, + { + "epoch": 2.15, + "learning_rate": 1.4107463452167225e-05, + "loss": 0.2059, + "step": 56028 + }, + { + "epoch": 2.15, + "learning_rate": 1.4106822262118493e-05, + "loss": 0.194, + "step": 56029 + }, + { + "epoch": 2.15, + "learning_rate": 1.4106181072069763e-05, + "loss": 0.1046, + "step": 56030 + }, + { + "epoch": 2.15, + "learning_rate": 1.4105539882021031e-05, + "loss": 0.1538, + "step": 56031 + }, + { + "epoch": 2.15, + "learning_rate": 1.4104898691972301e-05, + "loss": 0.2009, + "step": 56032 + }, + { + "epoch": 2.15, + "learning_rate": 1.4104257501923571e-05, + "loss": 0.2094, + "step": 56033 + }, + { + "epoch": 2.15, + "learning_rate": 1.410361631187484e-05, + "loss": 0.2076, + "step": 56034 + }, + { + "epoch": 2.15, + "learning_rate": 1.410297512182611e-05, + "loss": 0.2322, + "step": 56035 + }, + { + "epoch": 2.15, + "learning_rate": 1.4102333931777378e-05, + "loss": 0.1552, + "step": 56036 + }, + { + "epoch": 2.15, + "learning_rate": 1.410169274172865e-05, + "loss": 0.2204, + "step": 56037 + }, + { + "epoch": 2.15, + "learning_rate": 1.410105155167992e-05, + "loss": 0.1205, + "step": 56038 + }, + { + "epoch": 2.15, + "learning_rate": 1.4100410361631188e-05, + "loss": 0.3235, + "step": 56039 + }, + { + "epoch": 2.15, + "learning_rate": 1.4099769171582458e-05, + "loss": 0.1191, + "step": 56040 + }, + { + "epoch": 2.15, + "learning_rate": 1.4099127981533728e-05, + "loss": 0.2138, + "step": 56041 + }, + { + "epoch": 2.15, + "learning_rate": 1.4098486791484996e-05, + "loss": 0.2392, + "step": 56042 + }, + { + "epoch": 2.15, + "learning_rate": 1.4097845601436267e-05, + "loss": 0.1929, + "step": 56043 + }, + { + "epoch": 2.15, + "learning_rate": 1.4097204411387535e-05, + "loss": 0.2117, + "step": 56044 + }, + { + "epoch": 2.15, + "learning_rate": 1.4096563221338805e-05, + "loss": 0.2364, + "step": 56045 + }, + { + "epoch": 2.15, + "learning_rate": 1.4095922031290077e-05, + "loss": 0.3024, + "step": 56046 + }, + { + "epoch": 2.15, + "learning_rate": 1.4095280841241345e-05, + "loss": 0.2979, + "step": 56047 + }, + { + "epoch": 2.15, + "learning_rate": 1.4094639651192615e-05, + "loss": 0.3035, + "step": 56048 + }, + { + "epoch": 2.15, + "learning_rate": 1.4093998461143885e-05, + "loss": 0.2738, + "step": 56049 + }, + { + "epoch": 2.15, + "learning_rate": 1.4093357271095153e-05, + "loss": 0.3682, + "step": 56050 + }, + { + "epoch": 2.15, + "learning_rate": 1.4092716081046423e-05, + "loss": 0.2526, + "step": 56051 + }, + { + "epoch": 2.16, + "learning_rate": 1.4092074890997692e-05, + "loss": 0.2196, + "step": 56052 + }, + { + "epoch": 2.16, + "learning_rate": 1.4091433700948962e-05, + "loss": 0.264, + "step": 56053 + }, + { + "epoch": 2.16, + "learning_rate": 1.409079251090023e-05, + "loss": 0.2133, + "step": 56054 + }, + { + "epoch": 2.16, + "learning_rate": 1.4090151320851502e-05, + "loss": 0.2462, + "step": 56055 + }, + { + "epoch": 2.16, + "learning_rate": 1.4089510130802772e-05, + "loss": 0.2574, + "step": 56056 + }, + { + "epoch": 2.16, + "learning_rate": 1.408886894075404e-05, + "loss": 0.3413, + "step": 56057 + }, + { + "epoch": 2.16, + "learning_rate": 1.408822775070531e-05, + "loss": 0.2164, + "step": 56058 + }, + { + "epoch": 2.16, + "learning_rate": 1.408758656065658e-05, + "loss": 0.1954, + "step": 56059 + }, + { + "epoch": 2.16, + "learning_rate": 1.4086945370607848e-05, + "loss": 0.2809, + "step": 56060 + }, + { + "epoch": 2.16, + "learning_rate": 1.4086304180559118e-05, + "loss": 0.2577, + "step": 56061 + }, + { + "epoch": 2.16, + "learning_rate": 1.4085662990510387e-05, + "loss": 0.0746, + "step": 56062 + }, + { + "epoch": 2.16, + "learning_rate": 1.4085021800461657e-05, + "loss": 0.2219, + "step": 56063 + }, + { + "epoch": 2.16, + "learning_rate": 1.4084380610412929e-05, + "loss": 0.2151, + "step": 56064 + }, + { + "epoch": 2.16, + "learning_rate": 1.4083739420364197e-05, + "loss": 0.1838, + "step": 56065 + }, + { + "epoch": 2.16, + "learning_rate": 1.4083098230315467e-05, + "loss": 0.2578, + "step": 56066 + }, + { + "epoch": 2.16, + "learning_rate": 1.4082457040266737e-05, + "loss": 0.2731, + "step": 56067 + }, + { + "epoch": 2.16, + "learning_rate": 1.4081815850218005e-05, + "loss": 0.3847, + "step": 56068 + }, + { + "epoch": 2.16, + "learning_rate": 1.4081174660169275e-05, + "loss": 0.1179, + "step": 56069 + }, + { + "epoch": 2.16, + "learning_rate": 1.4080533470120544e-05, + "loss": 0.1561, + "step": 56070 + }, + { + "epoch": 2.16, + "learning_rate": 1.4079892280071814e-05, + "loss": 0.322, + "step": 56071 + }, + { + "epoch": 2.16, + "learning_rate": 1.4079251090023082e-05, + "loss": 0.2355, + "step": 56072 + }, + { + "epoch": 2.16, + "learning_rate": 1.4078609899974354e-05, + "loss": 0.1695, + "step": 56073 + }, + { + "epoch": 2.16, + "learning_rate": 1.4077968709925624e-05, + "loss": 0.3935, + "step": 56074 + }, + { + "epoch": 2.16, + "learning_rate": 1.4077327519876892e-05, + "loss": 0.2383, + "step": 56075 + }, + { + "epoch": 2.16, + "learning_rate": 1.4076686329828162e-05, + "loss": 0.1824, + "step": 56076 + }, + { + "epoch": 2.16, + "learning_rate": 1.4076045139779432e-05, + "loss": 0.232, + "step": 56077 + }, + { + "epoch": 2.16, + "learning_rate": 1.40754039497307e-05, + "loss": 0.2673, + "step": 56078 + }, + { + "epoch": 2.16, + "learning_rate": 1.407476275968197e-05, + "loss": 0.3219, + "step": 56079 + }, + { + "epoch": 2.16, + "learning_rate": 1.4074121569633239e-05, + "loss": 0.4693, + "step": 56080 + }, + { + "epoch": 2.16, + "learning_rate": 1.4073480379584509e-05, + "loss": 0.1723, + "step": 56081 + }, + { + "epoch": 2.16, + "learning_rate": 1.407283918953578e-05, + "loss": 0.1744, + "step": 56082 + }, + { + "epoch": 2.16, + "learning_rate": 1.4072197999487049e-05, + "loss": 0.1515, + "step": 56083 + }, + { + "epoch": 2.16, + "learning_rate": 1.4071556809438319e-05, + "loss": 0.2914, + "step": 56084 + }, + { + "epoch": 2.16, + "learning_rate": 1.4070915619389589e-05, + "loss": 0.1585, + "step": 56085 + }, + { + "epoch": 2.16, + "learning_rate": 1.4070274429340857e-05, + "loss": 0.2462, + "step": 56086 + }, + { + "epoch": 2.16, + "learning_rate": 1.4069633239292127e-05, + "loss": 0.228, + "step": 56087 + }, + { + "epoch": 2.16, + "learning_rate": 1.4068992049243396e-05, + "loss": 0.3222, + "step": 56088 + }, + { + "epoch": 2.16, + "learning_rate": 1.4068350859194666e-05, + "loss": 0.1236, + "step": 56089 + }, + { + "epoch": 2.16, + "learning_rate": 1.4067709669145934e-05, + "loss": 0.1713, + "step": 56090 + }, + { + "epoch": 2.16, + "learning_rate": 1.4067068479097206e-05, + "loss": 0.3151, + "step": 56091 + }, + { + "epoch": 2.16, + "learning_rate": 1.4066427289048476e-05, + "loss": 0.2594, + "step": 56092 + }, + { + "epoch": 2.16, + "learning_rate": 1.4065786098999744e-05, + "loss": 0.245, + "step": 56093 + }, + { + "epoch": 2.16, + "learning_rate": 1.4065144908951014e-05, + "loss": 0.2233, + "step": 56094 + }, + { + "epoch": 2.16, + "learning_rate": 1.4064503718902284e-05, + "loss": 0.278, + "step": 56095 + }, + { + "epoch": 2.16, + "learning_rate": 1.4063862528853552e-05, + "loss": 0.2797, + "step": 56096 + }, + { + "epoch": 2.16, + "learning_rate": 1.4063221338804822e-05, + "loss": 0.3566, + "step": 56097 + }, + { + "epoch": 2.16, + "learning_rate": 1.406258014875609e-05, + "loss": 0.1739, + "step": 56098 + }, + { + "epoch": 2.16, + "learning_rate": 1.406193895870736e-05, + "loss": 0.2674, + "step": 56099 + }, + { + "epoch": 2.16, + "learning_rate": 1.4061297768658632e-05, + "loss": 0.2386, + "step": 56100 + }, + { + "epoch": 2.16, + "learning_rate": 1.40606565786099e-05, + "loss": 0.2675, + "step": 56101 + }, + { + "epoch": 2.16, + "learning_rate": 1.406001538856117e-05, + "loss": 0.223, + "step": 56102 + }, + { + "epoch": 2.16, + "learning_rate": 1.405937419851244e-05, + "loss": 0.2143, + "step": 56103 + }, + { + "epoch": 2.16, + "learning_rate": 1.4058733008463709e-05, + "loss": 0.17, + "step": 56104 + }, + { + "epoch": 2.16, + "learning_rate": 1.4058091818414979e-05, + "loss": 0.2325, + "step": 56105 + }, + { + "epoch": 2.16, + "learning_rate": 1.4057450628366247e-05, + "loss": 0.1914, + "step": 56106 + }, + { + "epoch": 2.16, + "learning_rate": 1.4056809438317517e-05, + "loss": 0.2976, + "step": 56107 + }, + { + "epoch": 2.16, + "learning_rate": 1.4056168248268786e-05, + "loss": 0.2538, + "step": 56108 + }, + { + "epoch": 2.16, + "learning_rate": 1.4055527058220058e-05, + "loss": 0.3104, + "step": 56109 + }, + { + "epoch": 2.16, + "learning_rate": 1.4054885868171328e-05, + "loss": 0.1334, + "step": 56110 + }, + { + "epoch": 2.16, + "learning_rate": 1.4054244678122596e-05, + "loss": 0.2019, + "step": 56111 + }, + { + "epoch": 2.16, + "learning_rate": 1.4053603488073866e-05, + "loss": 0.257, + "step": 56112 + }, + { + "epoch": 2.16, + "learning_rate": 1.4052962298025136e-05, + "loss": 0.2031, + "step": 56113 + }, + { + "epoch": 2.16, + "learning_rate": 1.4052321107976404e-05, + "loss": 0.1519, + "step": 56114 + }, + { + "epoch": 2.16, + "learning_rate": 1.4051679917927674e-05, + "loss": 0.2223, + "step": 56115 + }, + { + "epoch": 2.16, + "learning_rate": 1.4051038727878943e-05, + "loss": 0.3161, + "step": 56116 + }, + { + "epoch": 2.16, + "learning_rate": 1.4050397537830213e-05, + "loss": 0.2072, + "step": 56117 + }, + { + "epoch": 2.16, + "learning_rate": 1.4049756347781484e-05, + "loss": 0.2234, + "step": 56118 + }, + { + "epoch": 2.16, + "learning_rate": 1.4049115157732753e-05, + "loss": 0.1915, + "step": 56119 + }, + { + "epoch": 2.16, + "learning_rate": 1.4048473967684023e-05, + "loss": 0.1348, + "step": 56120 + }, + { + "epoch": 2.16, + "learning_rate": 1.4047832777635293e-05, + "loss": 0.2166, + "step": 56121 + }, + { + "epoch": 2.16, + "learning_rate": 1.4047191587586561e-05, + "loss": 0.3671, + "step": 56122 + }, + { + "epoch": 2.16, + "learning_rate": 1.4046550397537831e-05, + "loss": 0.2094, + "step": 56123 + }, + { + "epoch": 2.16, + "learning_rate": 1.40459092074891e-05, + "loss": 0.2527, + "step": 56124 + }, + { + "epoch": 2.16, + "learning_rate": 1.404526801744037e-05, + "loss": 0.1931, + "step": 56125 + }, + { + "epoch": 2.16, + "learning_rate": 1.4044626827391641e-05, + "loss": 0.1807, + "step": 56126 + }, + { + "epoch": 2.16, + "learning_rate": 1.404398563734291e-05, + "loss": 0.082, + "step": 56127 + }, + { + "epoch": 2.16, + "learning_rate": 1.404334444729418e-05, + "loss": 0.2568, + "step": 56128 + }, + { + "epoch": 2.16, + "learning_rate": 1.404270325724545e-05, + "loss": 0.3273, + "step": 56129 + }, + { + "epoch": 2.16, + "learning_rate": 1.4042062067196718e-05, + "loss": 0.4042, + "step": 56130 + }, + { + "epoch": 2.16, + "learning_rate": 1.4041420877147988e-05, + "loss": 0.2426, + "step": 56131 + }, + { + "epoch": 2.16, + "learning_rate": 1.4040779687099256e-05, + "loss": 0.2446, + "step": 56132 + }, + { + "epoch": 2.16, + "learning_rate": 1.4040138497050526e-05, + "loss": 0.2715, + "step": 56133 + }, + { + "epoch": 2.16, + "learning_rate": 1.4039497307001795e-05, + "loss": 0.1661, + "step": 56134 + }, + { + "epoch": 2.16, + "learning_rate": 1.4038856116953066e-05, + "loss": 0.3068, + "step": 56135 + }, + { + "epoch": 2.16, + "learning_rate": 1.4038214926904336e-05, + "loss": 0.2371, + "step": 56136 + }, + { + "epoch": 2.16, + "learning_rate": 1.4037573736855605e-05, + "loss": 0.1645, + "step": 56137 + }, + { + "epoch": 2.16, + "learning_rate": 1.4036932546806875e-05, + "loss": 0.3005, + "step": 56138 + }, + { + "epoch": 2.16, + "learning_rate": 1.4036291356758145e-05, + "loss": 0.1623, + "step": 56139 + }, + { + "epoch": 2.16, + "learning_rate": 1.4035650166709413e-05, + "loss": 0.3244, + "step": 56140 + }, + { + "epoch": 2.16, + "learning_rate": 1.4035008976660683e-05, + "loss": 0.2969, + "step": 56141 + }, + { + "epoch": 2.16, + "learning_rate": 1.4034367786611951e-05, + "loss": 0.212, + "step": 56142 + }, + { + "epoch": 2.16, + "learning_rate": 1.4033726596563221e-05, + "loss": 0.3528, + "step": 56143 + }, + { + "epoch": 2.16, + "learning_rate": 1.4033085406514493e-05, + "loss": 0.2595, + "step": 56144 + }, + { + "epoch": 2.16, + "learning_rate": 1.4032444216465761e-05, + "loss": 0.2426, + "step": 56145 + }, + { + "epoch": 2.16, + "learning_rate": 1.4031803026417031e-05, + "loss": 0.1719, + "step": 56146 + }, + { + "epoch": 2.16, + "learning_rate": 1.4031161836368301e-05, + "loss": 0.1919, + "step": 56147 + }, + { + "epoch": 2.16, + "learning_rate": 1.403052064631957e-05, + "loss": 0.2721, + "step": 56148 + }, + { + "epoch": 2.16, + "learning_rate": 1.402987945627084e-05, + "loss": 0.1988, + "step": 56149 + }, + { + "epoch": 2.16, + "learning_rate": 1.4029238266222108e-05, + "loss": 0.328, + "step": 56150 + }, + { + "epoch": 2.16, + "learning_rate": 1.4028597076173378e-05, + "loss": 0.3042, + "step": 56151 + }, + { + "epoch": 2.16, + "learning_rate": 1.4027955886124646e-05, + "loss": 0.189, + "step": 56152 + }, + { + "epoch": 2.16, + "learning_rate": 1.4027314696075918e-05, + "loss": 0.3492, + "step": 56153 + }, + { + "epoch": 2.16, + "learning_rate": 1.4026673506027188e-05, + "loss": 0.2018, + "step": 56154 + }, + { + "epoch": 2.16, + "learning_rate": 1.4026032315978457e-05, + "loss": 0.4203, + "step": 56155 + }, + { + "epoch": 2.16, + "learning_rate": 1.4025391125929727e-05, + "loss": 0.3845, + "step": 56156 + }, + { + "epoch": 2.16, + "learning_rate": 1.4024749935880997e-05, + "loss": 0.225, + "step": 56157 + }, + { + "epoch": 2.16, + "learning_rate": 1.4024108745832265e-05, + "loss": 0.2717, + "step": 56158 + }, + { + "epoch": 2.16, + "learning_rate": 1.4023467555783535e-05, + "loss": 0.275, + "step": 56159 + }, + { + "epoch": 2.16, + "learning_rate": 1.4022826365734803e-05, + "loss": 0.1345, + "step": 56160 + }, + { + "epoch": 2.16, + "learning_rate": 1.4022185175686073e-05, + "loss": 0.2537, + "step": 56161 + }, + { + "epoch": 2.16, + "learning_rate": 1.4021543985637345e-05, + "loss": 0.2763, + "step": 56162 + }, + { + "epoch": 2.16, + "learning_rate": 1.4020902795588613e-05, + "loss": 0.3627, + "step": 56163 + }, + { + "epoch": 2.16, + "learning_rate": 1.4020261605539883e-05, + "loss": 0.2553, + "step": 56164 + }, + { + "epoch": 2.16, + "learning_rate": 1.4019620415491153e-05, + "loss": 0.2199, + "step": 56165 + }, + { + "epoch": 2.16, + "learning_rate": 1.4018979225442422e-05, + "loss": 0.2834, + "step": 56166 + }, + { + "epoch": 2.16, + "learning_rate": 1.4018338035393692e-05, + "loss": 0.2506, + "step": 56167 + }, + { + "epoch": 2.16, + "learning_rate": 1.401769684534496e-05, + "loss": 0.3015, + "step": 56168 + }, + { + "epoch": 2.16, + "learning_rate": 1.401705565529623e-05, + "loss": 0.2217, + "step": 56169 + }, + { + "epoch": 2.16, + "learning_rate": 1.4016414465247498e-05, + "loss": 0.2936, + "step": 56170 + }, + { + "epoch": 2.16, + "learning_rate": 1.401577327519877e-05, + "loss": 0.3072, + "step": 56171 + }, + { + "epoch": 2.16, + "learning_rate": 1.401513208515004e-05, + "loss": 0.2468, + "step": 56172 + }, + { + "epoch": 2.16, + "learning_rate": 1.4014490895101308e-05, + "loss": 0.2644, + "step": 56173 + }, + { + "epoch": 2.16, + "learning_rate": 1.4013849705052578e-05, + "loss": 0.2344, + "step": 56174 + }, + { + "epoch": 2.16, + "learning_rate": 1.4013208515003849e-05, + "loss": 0.1973, + "step": 56175 + }, + { + "epoch": 2.16, + "learning_rate": 1.4012567324955117e-05, + "loss": 0.1706, + "step": 56176 + }, + { + "epoch": 2.16, + "learning_rate": 1.4011926134906387e-05, + "loss": 0.262, + "step": 56177 + }, + { + "epoch": 2.16, + "learning_rate": 1.4011284944857655e-05, + "loss": 0.2566, + "step": 56178 + }, + { + "epoch": 2.16, + "learning_rate": 1.4010643754808925e-05, + "loss": 0.2564, + "step": 56179 + }, + { + "epoch": 2.16, + "learning_rate": 1.4010002564760197e-05, + "loss": 0.1569, + "step": 56180 + }, + { + "epoch": 2.16, + "learning_rate": 1.4009361374711465e-05, + "loss": 0.3882, + "step": 56181 + }, + { + "epoch": 2.16, + "learning_rate": 1.4008720184662735e-05, + "loss": 0.2735, + "step": 56182 + }, + { + "epoch": 2.16, + "learning_rate": 1.4008078994614005e-05, + "loss": 0.2033, + "step": 56183 + }, + { + "epoch": 2.16, + "learning_rate": 1.4007437804565274e-05, + "loss": 0.1407, + "step": 56184 + }, + { + "epoch": 2.16, + "learning_rate": 1.4006796614516544e-05, + "loss": 0.2407, + "step": 56185 + }, + { + "epoch": 2.16, + "learning_rate": 1.4006155424467812e-05, + "loss": 0.2078, + "step": 56186 + }, + { + "epoch": 2.16, + "learning_rate": 1.4005514234419082e-05, + "loss": 0.1619, + "step": 56187 + }, + { + "epoch": 2.16, + "learning_rate": 1.400487304437035e-05, + "loss": 0.2586, + "step": 56188 + }, + { + "epoch": 2.16, + "learning_rate": 1.4004231854321622e-05, + "loss": 0.2143, + "step": 56189 + }, + { + "epoch": 2.16, + "learning_rate": 1.4003590664272892e-05, + "loss": 0.2658, + "step": 56190 + }, + { + "epoch": 2.16, + "learning_rate": 1.4002949474224162e-05, + "loss": 0.172, + "step": 56191 + }, + { + "epoch": 2.16, + "learning_rate": 1.400230828417543e-05, + "loss": 0.3515, + "step": 56192 + }, + { + "epoch": 2.16, + "learning_rate": 1.40016670941267e-05, + "loss": 0.2108, + "step": 56193 + }, + { + "epoch": 2.16, + "learning_rate": 1.4001025904077969e-05, + "loss": 0.2132, + "step": 56194 + }, + { + "epoch": 2.16, + "learning_rate": 1.4000384714029239e-05, + "loss": 0.2503, + "step": 56195 + }, + { + "epoch": 2.16, + "learning_rate": 1.3999743523980507e-05, + "loss": 0.2533, + "step": 56196 + }, + { + "epoch": 2.16, + "learning_rate": 1.3999102333931777e-05, + "loss": 0.1768, + "step": 56197 + }, + { + "epoch": 2.16, + "learning_rate": 1.3998461143883049e-05, + "loss": 0.1377, + "step": 56198 + }, + { + "epoch": 2.16, + "learning_rate": 1.3997819953834317e-05, + "loss": 0.2271, + "step": 56199 + }, + { + "epoch": 2.16, + "learning_rate": 1.3997178763785587e-05, + "loss": 0.2918, + "step": 56200 + }, + { + "epoch": 2.16, + "learning_rate": 1.3996537573736857e-05, + "loss": 0.1912, + "step": 56201 + }, + { + "epoch": 2.16, + "learning_rate": 1.3995896383688126e-05, + "loss": 0.2858, + "step": 56202 + }, + { + "epoch": 2.16, + "learning_rate": 1.3995255193639396e-05, + "loss": 0.1644, + "step": 56203 + }, + { + "epoch": 2.16, + "learning_rate": 1.3994614003590664e-05, + "loss": 0.2772, + "step": 56204 + }, + { + "epoch": 2.16, + "learning_rate": 1.3993972813541934e-05, + "loss": 0.3443, + "step": 56205 + }, + { + "epoch": 2.16, + "learning_rate": 1.3993331623493202e-05, + "loss": 0.2399, + "step": 56206 + }, + { + "epoch": 2.16, + "learning_rate": 1.3992690433444474e-05, + "loss": 0.2457, + "step": 56207 + }, + { + "epoch": 2.16, + "learning_rate": 1.3992049243395744e-05, + "loss": 0.2156, + "step": 56208 + }, + { + "epoch": 2.16, + "learning_rate": 1.3991408053347014e-05, + "loss": 0.3976, + "step": 56209 + }, + { + "epoch": 2.16, + "learning_rate": 1.3990766863298282e-05, + "loss": 0.4122, + "step": 56210 + }, + { + "epoch": 2.16, + "learning_rate": 1.3990125673249552e-05, + "loss": 0.1431, + "step": 56211 + }, + { + "epoch": 2.16, + "learning_rate": 1.398948448320082e-05, + "loss": 0.185, + "step": 56212 + }, + { + "epoch": 2.16, + "learning_rate": 1.398884329315209e-05, + "loss": 0.2447, + "step": 56213 + }, + { + "epoch": 2.16, + "learning_rate": 1.3988202103103359e-05, + "loss": 0.2428, + "step": 56214 + }, + { + "epoch": 2.16, + "learning_rate": 1.3987560913054629e-05, + "loss": 0.1666, + "step": 56215 + }, + { + "epoch": 2.16, + "learning_rate": 1.39869197230059e-05, + "loss": 0.2985, + "step": 56216 + }, + { + "epoch": 2.16, + "learning_rate": 1.3986278532957169e-05, + "loss": 0.189, + "step": 56217 + }, + { + "epoch": 2.16, + "learning_rate": 1.3985637342908439e-05, + "loss": 0.367, + "step": 56218 + }, + { + "epoch": 2.16, + "learning_rate": 1.398499615285971e-05, + "loss": 0.3127, + "step": 56219 + }, + { + "epoch": 2.16, + "learning_rate": 1.3984354962810977e-05, + "loss": 0.2292, + "step": 56220 + }, + { + "epoch": 2.16, + "learning_rate": 1.3983713772762248e-05, + "loss": 0.2551, + "step": 56221 + }, + { + "epoch": 2.16, + "learning_rate": 1.3983072582713516e-05, + "loss": 0.3593, + "step": 56222 + }, + { + "epoch": 2.16, + "learning_rate": 1.3982431392664786e-05, + "loss": 0.2806, + "step": 56223 + }, + { + "epoch": 2.16, + "learning_rate": 1.3981790202616058e-05, + "loss": 0.3551, + "step": 56224 + }, + { + "epoch": 2.16, + "learning_rate": 1.3981149012567326e-05, + "loss": 0.1956, + "step": 56225 + }, + { + "epoch": 2.16, + "learning_rate": 1.3980507822518596e-05, + "loss": 0.2981, + "step": 56226 + }, + { + "epoch": 2.16, + "learning_rate": 1.3979866632469866e-05, + "loss": 0.2473, + "step": 56227 + }, + { + "epoch": 2.16, + "learning_rate": 1.3979225442421134e-05, + "loss": 0.1587, + "step": 56228 + }, + { + "epoch": 2.16, + "learning_rate": 1.3978584252372404e-05, + "loss": 0.2359, + "step": 56229 + }, + { + "epoch": 2.16, + "learning_rate": 1.3977943062323673e-05, + "loss": 0.2494, + "step": 56230 + }, + { + "epoch": 2.16, + "learning_rate": 1.3977301872274943e-05, + "loss": 0.3141, + "step": 56231 + }, + { + "epoch": 2.16, + "learning_rate": 1.3976660682226211e-05, + "loss": 0.328, + "step": 56232 + }, + { + "epoch": 2.16, + "learning_rate": 1.3976019492177483e-05, + "loss": 0.2662, + "step": 56233 + }, + { + "epoch": 2.16, + "learning_rate": 1.3975378302128753e-05, + "loss": 0.3353, + "step": 56234 + }, + { + "epoch": 2.16, + "learning_rate": 1.3974737112080021e-05, + "loss": 0.1083, + "step": 56235 + }, + { + "epoch": 2.16, + "learning_rate": 1.3974095922031291e-05, + "loss": 0.1831, + "step": 56236 + }, + { + "epoch": 2.16, + "learning_rate": 1.3973454731982561e-05, + "loss": 0.2032, + "step": 56237 + }, + { + "epoch": 2.16, + "learning_rate": 1.397281354193383e-05, + "loss": 0.2686, + "step": 56238 + }, + { + "epoch": 2.16, + "learning_rate": 1.39721723518851e-05, + "loss": 0.2075, + "step": 56239 + }, + { + "epoch": 2.16, + "learning_rate": 1.3971531161836368e-05, + "loss": 0.2735, + "step": 56240 + }, + { + "epoch": 2.16, + "learning_rate": 1.3970889971787638e-05, + "loss": 0.22, + "step": 56241 + }, + { + "epoch": 2.16, + "learning_rate": 1.397024878173891e-05, + "loss": 0.2736, + "step": 56242 + }, + { + "epoch": 2.16, + "learning_rate": 1.3969607591690178e-05, + "loss": 0.3013, + "step": 56243 + }, + { + "epoch": 2.16, + "learning_rate": 1.3968966401641448e-05, + "loss": 0.1935, + "step": 56244 + }, + { + "epoch": 2.16, + "learning_rate": 1.3968325211592718e-05, + "loss": 0.1994, + "step": 56245 + }, + { + "epoch": 2.16, + "learning_rate": 1.3967684021543986e-05, + "loss": 0.2697, + "step": 56246 + }, + { + "epoch": 2.16, + "learning_rate": 1.3967042831495256e-05, + "loss": 0.1824, + "step": 56247 + }, + { + "epoch": 2.16, + "learning_rate": 1.3966401641446525e-05, + "loss": 0.2148, + "step": 56248 + }, + { + "epoch": 2.16, + "learning_rate": 1.3965760451397795e-05, + "loss": 0.2687, + "step": 56249 + }, + { + "epoch": 2.16, + "learning_rate": 1.3965119261349063e-05, + "loss": 0.1502, + "step": 56250 + }, + { + "epoch": 2.16, + "learning_rate": 1.3964478071300335e-05, + "loss": 0.2095, + "step": 56251 + }, + { + "epoch": 2.16, + "learning_rate": 1.3963836881251605e-05, + "loss": 0.3392, + "step": 56252 + }, + { + "epoch": 2.16, + "learning_rate": 1.3963195691202875e-05, + "loss": 0.2757, + "step": 56253 + }, + { + "epoch": 2.16, + "learning_rate": 1.3962554501154143e-05, + "loss": 0.3409, + "step": 56254 + }, + { + "epoch": 2.16, + "learning_rate": 1.3961913311105413e-05, + "loss": 0.3599, + "step": 56255 + }, + { + "epoch": 2.16, + "learning_rate": 1.3961272121056681e-05, + "loss": 0.2344, + "step": 56256 + }, + { + "epoch": 2.16, + "learning_rate": 1.3960630931007951e-05, + "loss": 0.3344, + "step": 56257 + }, + { + "epoch": 2.16, + "learning_rate": 1.395998974095922e-05, + "loss": 0.1816, + "step": 56258 + }, + { + "epoch": 2.16, + "learning_rate": 1.395934855091049e-05, + "loss": 0.2132, + "step": 56259 + }, + { + "epoch": 2.16, + "learning_rate": 1.3958707360861761e-05, + "loss": 0.2202, + "step": 56260 + }, + { + "epoch": 2.16, + "learning_rate": 1.395806617081303e-05, + "loss": 0.1788, + "step": 56261 + }, + { + "epoch": 2.16, + "learning_rate": 1.39574249807643e-05, + "loss": 0.2316, + "step": 56262 + }, + { + "epoch": 2.16, + "learning_rate": 1.395678379071557e-05, + "loss": 0.2199, + "step": 56263 + }, + { + "epoch": 2.16, + "learning_rate": 1.3956142600666838e-05, + "loss": 0.2094, + "step": 56264 + }, + { + "epoch": 2.16, + "learning_rate": 1.3955501410618108e-05, + "loss": 0.2581, + "step": 56265 + }, + { + "epoch": 2.16, + "learning_rate": 1.3954860220569377e-05, + "loss": 0.1894, + "step": 56266 + }, + { + "epoch": 2.16, + "learning_rate": 1.3954219030520647e-05, + "loss": 0.3107, + "step": 56267 + }, + { + "epoch": 2.16, + "learning_rate": 1.3953577840471915e-05, + "loss": 0.3001, + "step": 56268 + }, + { + "epoch": 2.16, + "learning_rate": 1.3952936650423187e-05, + "loss": 0.2768, + "step": 56269 + }, + { + "epoch": 2.16, + "learning_rate": 1.3952295460374457e-05, + "loss": 0.3048, + "step": 56270 + }, + { + "epoch": 2.16, + "learning_rate": 1.3951654270325727e-05, + "loss": 0.2886, + "step": 56271 + }, + { + "epoch": 2.16, + "learning_rate": 1.3951013080276995e-05, + "loss": 0.1877, + "step": 56272 + }, + { + "epoch": 2.16, + "learning_rate": 1.3950371890228265e-05, + "loss": 0.1843, + "step": 56273 + }, + { + "epoch": 2.16, + "learning_rate": 1.3949730700179533e-05, + "loss": 0.2693, + "step": 56274 + }, + { + "epoch": 2.16, + "learning_rate": 1.3949089510130803e-05, + "loss": 0.242, + "step": 56275 + }, + { + "epoch": 2.16, + "learning_rate": 1.3948448320082072e-05, + "loss": 0.1866, + "step": 56276 + }, + { + "epoch": 2.16, + "learning_rate": 1.3947807130033342e-05, + "loss": 0.2179, + "step": 56277 + }, + { + "epoch": 2.16, + "learning_rate": 1.3947165939984613e-05, + "loss": 0.2347, + "step": 56278 + }, + { + "epoch": 2.16, + "learning_rate": 1.3946524749935882e-05, + "loss": 0.2007, + "step": 56279 + }, + { + "epoch": 2.16, + "learning_rate": 1.3945883559887152e-05, + "loss": 0.2741, + "step": 56280 + }, + { + "epoch": 2.16, + "learning_rate": 1.3945242369838422e-05, + "loss": 0.2524, + "step": 56281 + }, + { + "epoch": 2.16, + "learning_rate": 1.394460117978969e-05, + "loss": 0.3883, + "step": 56282 + }, + { + "epoch": 2.16, + "learning_rate": 1.394395998974096e-05, + "loss": 0.1763, + "step": 56283 + }, + { + "epoch": 2.16, + "learning_rate": 1.3943318799692228e-05, + "loss": 0.2854, + "step": 56284 + }, + { + "epoch": 2.16, + "learning_rate": 1.3942677609643498e-05, + "loss": 0.2892, + "step": 56285 + }, + { + "epoch": 2.16, + "learning_rate": 1.3942036419594767e-05, + "loss": 0.281, + "step": 56286 + }, + { + "epoch": 2.16, + "learning_rate": 1.3941395229546039e-05, + "loss": 0.2434, + "step": 56287 + }, + { + "epoch": 2.16, + "learning_rate": 1.3940754039497309e-05, + "loss": 0.2155, + "step": 56288 + }, + { + "epoch": 2.16, + "learning_rate": 1.3940112849448579e-05, + "loss": 0.1891, + "step": 56289 + }, + { + "epoch": 2.16, + "learning_rate": 1.3939471659399847e-05, + "loss": 0.33, + "step": 56290 + }, + { + "epoch": 2.16, + "learning_rate": 1.3938830469351117e-05, + "loss": 0.1948, + "step": 56291 + }, + { + "epoch": 2.16, + "learning_rate": 1.3938189279302385e-05, + "loss": 0.2937, + "step": 56292 + }, + { + "epoch": 2.16, + "learning_rate": 1.3937548089253655e-05, + "loss": 0.1803, + "step": 56293 + }, + { + "epoch": 2.16, + "learning_rate": 1.3936906899204924e-05, + "loss": 0.2113, + "step": 56294 + }, + { + "epoch": 2.16, + "learning_rate": 1.3936265709156194e-05, + "loss": 0.1707, + "step": 56295 + }, + { + "epoch": 2.16, + "learning_rate": 1.3935624519107465e-05, + "loss": 0.2575, + "step": 56296 + }, + { + "epoch": 2.16, + "learning_rate": 1.3934983329058734e-05, + "loss": 0.3402, + "step": 56297 + }, + { + "epoch": 2.16, + "learning_rate": 1.3934342139010004e-05, + "loss": 0.1375, + "step": 56298 + }, + { + "epoch": 2.16, + "learning_rate": 1.3933700948961274e-05, + "loss": 0.2305, + "step": 56299 + }, + { + "epoch": 2.16, + "learning_rate": 1.3933059758912542e-05, + "loss": 0.2709, + "step": 56300 + }, + { + "epoch": 2.16, + "learning_rate": 1.3932418568863812e-05, + "loss": 0.277, + "step": 56301 + }, + { + "epoch": 2.16, + "learning_rate": 1.393177737881508e-05, + "loss": 0.2691, + "step": 56302 + }, + { + "epoch": 2.16, + "learning_rate": 1.393113618876635e-05, + "loss": 0.2232, + "step": 56303 + }, + { + "epoch": 2.16, + "learning_rate": 1.3930494998717619e-05, + "loss": 0.3599, + "step": 56304 + }, + { + "epoch": 2.16, + "learning_rate": 1.392985380866889e-05, + "loss": 0.1581, + "step": 56305 + }, + { + "epoch": 2.16, + "learning_rate": 1.392921261862016e-05, + "loss": 0.3529, + "step": 56306 + }, + { + "epoch": 2.16, + "learning_rate": 1.392857142857143e-05, + "loss": 0.1975, + "step": 56307 + }, + { + "epoch": 2.16, + "learning_rate": 1.3927930238522699e-05, + "loss": 0.2931, + "step": 56308 + }, + { + "epoch": 2.16, + "learning_rate": 1.3927289048473969e-05, + "loss": 0.2166, + "step": 56309 + }, + { + "epoch": 2.16, + "learning_rate": 1.3926647858425237e-05, + "loss": 0.3111, + "step": 56310 + }, + { + "epoch": 2.16, + "learning_rate": 1.3926006668376507e-05, + "loss": 0.2023, + "step": 56311 + }, + { + "epoch": 2.17, + "learning_rate": 1.3925365478327776e-05, + "loss": 0.211, + "step": 56312 + }, + { + "epoch": 2.17, + "learning_rate": 1.3924724288279047e-05, + "loss": 0.2241, + "step": 56313 + }, + { + "epoch": 2.17, + "learning_rate": 1.3924083098230317e-05, + "loss": 0.2933, + "step": 56314 + }, + { + "epoch": 2.17, + "learning_rate": 1.3923441908181587e-05, + "loss": 0.2406, + "step": 56315 + }, + { + "epoch": 2.17, + "learning_rate": 1.3922800718132856e-05, + "loss": 0.1973, + "step": 56316 + }, + { + "epoch": 2.17, + "learning_rate": 1.3922159528084126e-05, + "loss": 0.2863, + "step": 56317 + }, + { + "epoch": 2.17, + "learning_rate": 1.3921518338035394e-05, + "loss": 0.2706, + "step": 56318 + }, + { + "epoch": 2.17, + "learning_rate": 1.3920877147986664e-05, + "loss": 0.2308, + "step": 56319 + }, + { + "epoch": 2.17, + "learning_rate": 1.3920235957937932e-05, + "loss": 0.2312, + "step": 56320 + }, + { + "epoch": 2.17, + "learning_rate": 1.3919594767889202e-05, + "loss": 0.3088, + "step": 56321 + }, + { + "epoch": 2.17, + "learning_rate": 1.3918953577840474e-05, + "loss": 0.1827, + "step": 56322 + }, + { + "epoch": 2.17, + "learning_rate": 1.3918312387791742e-05, + "loss": 0.4477, + "step": 56323 + }, + { + "epoch": 2.17, + "learning_rate": 1.3917671197743012e-05, + "loss": 0.2636, + "step": 56324 + }, + { + "epoch": 2.17, + "learning_rate": 1.3917030007694282e-05, + "loss": 0.3248, + "step": 56325 + }, + { + "epoch": 2.17, + "learning_rate": 1.391638881764555e-05, + "loss": 0.1744, + "step": 56326 + }, + { + "epoch": 2.17, + "learning_rate": 1.391574762759682e-05, + "loss": 0.3632, + "step": 56327 + }, + { + "epoch": 2.17, + "learning_rate": 1.3915106437548089e-05, + "loss": 0.3455, + "step": 56328 + }, + { + "epoch": 2.17, + "learning_rate": 1.3914465247499359e-05, + "loss": 0.2344, + "step": 56329 + }, + { + "epoch": 2.17, + "learning_rate": 1.3913824057450627e-05, + "loss": 0.3329, + "step": 56330 + }, + { + "epoch": 2.17, + "learning_rate": 1.39131828674019e-05, + "loss": 0.1299, + "step": 56331 + }, + { + "epoch": 2.17, + "learning_rate": 1.391254167735317e-05, + "loss": 0.2753, + "step": 56332 + }, + { + "epoch": 2.17, + "learning_rate": 1.391190048730444e-05, + "loss": 0.0895, + "step": 56333 + }, + { + "epoch": 2.17, + "learning_rate": 1.3911259297255708e-05, + "loss": 0.2736, + "step": 56334 + }, + { + "epoch": 2.17, + "learning_rate": 1.3910618107206978e-05, + "loss": 0.2118, + "step": 56335 + }, + { + "epoch": 2.17, + "learning_rate": 1.3909976917158246e-05, + "loss": 0.191, + "step": 56336 + }, + { + "epoch": 2.17, + "learning_rate": 1.3909335727109516e-05, + "loss": 0.2674, + "step": 56337 + }, + { + "epoch": 2.17, + "learning_rate": 1.3908694537060784e-05, + "loss": 0.3177, + "step": 56338 + }, + { + "epoch": 2.17, + "learning_rate": 1.3908053347012054e-05, + "loss": 0.2719, + "step": 56339 + }, + { + "epoch": 2.17, + "learning_rate": 1.3907412156963326e-05, + "loss": 0.2893, + "step": 56340 + }, + { + "epoch": 2.17, + "learning_rate": 1.3906770966914594e-05, + "loss": 0.1842, + "step": 56341 + }, + { + "epoch": 2.17, + "learning_rate": 1.3906129776865864e-05, + "loss": 0.1848, + "step": 56342 + }, + { + "epoch": 2.17, + "learning_rate": 1.3905488586817134e-05, + "loss": 0.2875, + "step": 56343 + }, + { + "epoch": 2.17, + "learning_rate": 1.3904847396768403e-05, + "loss": 0.176, + "step": 56344 + }, + { + "epoch": 2.17, + "learning_rate": 1.3904206206719673e-05, + "loss": 0.3092, + "step": 56345 + }, + { + "epoch": 2.17, + "learning_rate": 1.3903565016670941e-05, + "loss": 0.2241, + "step": 56346 + }, + { + "epoch": 2.17, + "learning_rate": 1.3902923826622211e-05, + "loss": 0.3362, + "step": 56347 + }, + { + "epoch": 2.17, + "learning_rate": 1.390228263657348e-05, + "loss": 0.3824, + "step": 56348 + }, + { + "epoch": 2.17, + "learning_rate": 1.3901641446524751e-05, + "loss": 0.1804, + "step": 56349 + }, + { + "epoch": 2.17, + "learning_rate": 1.3901000256476021e-05, + "loss": 0.3828, + "step": 56350 + }, + { + "epoch": 2.17, + "learning_rate": 1.3900359066427291e-05, + "loss": 0.2094, + "step": 56351 + }, + { + "epoch": 2.17, + "learning_rate": 1.389971787637856e-05, + "loss": 0.1709, + "step": 56352 + }, + { + "epoch": 2.17, + "learning_rate": 1.389907668632983e-05, + "loss": 0.23, + "step": 56353 + }, + { + "epoch": 2.17, + "learning_rate": 1.3898435496281098e-05, + "loss": 0.1817, + "step": 56354 + }, + { + "epoch": 2.17, + "learning_rate": 1.3897794306232368e-05, + "loss": 0.2778, + "step": 56355 + }, + { + "epoch": 2.17, + "learning_rate": 1.3897153116183636e-05, + "loss": 0.1692, + "step": 56356 + }, + { + "epoch": 2.17, + "learning_rate": 1.3896511926134906e-05, + "loss": 0.0816, + "step": 56357 + }, + { + "epoch": 2.17, + "learning_rate": 1.3895870736086178e-05, + "loss": 0.3664, + "step": 56358 + }, + { + "epoch": 2.17, + "learning_rate": 1.3895229546037446e-05, + "loss": 0.2231, + "step": 56359 + }, + { + "epoch": 2.17, + "learning_rate": 1.3894588355988716e-05, + "loss": 0.3983, + "step": 56360 + }, + { + "epoch": 2.17, + "learning_rate": 1.3893947165939986e-05, + "loss": 0.2139, + "step": 56361 + }, + { + "epoch": 2.17, + "learning_rate": 1.3893305975891255e-05, + "loss": 0.2003, + "step": 56362 + }, + { + "epoch": 2.17, + "learning_rate": 1.3892664785842525e-05, + "loss": 0.3562, + "step": 56363 + }, + { + "epoch": 2.17, + "learning_rate": 1.3892023595793793e-05, + "loss": 0.2951, + "step": 56364 + }, + { + "epoch": 2.17, + "learning_rate": 1.3891382405745063e-05, + "loss": 0.1808, + "step": 56365 + }, + { + "epoch": 2.17, + "learning_rate": 1.3890741215696331e-05, + "loss": 0.3147, + "step": 56366 + }, + { + "epoch": 2.17, + "learning_rate": 1.3890100025647603e-05, + "loss": 0.1822, + "step": 56367 + }, + { + "epoch": 2.17, + "learning_rate": 1.3889458835598873e-05, + "loss": 0.2146, + "step": 56368 + }, + { + "epoch": 2.17, + "learning_rate": 1.3888817645550143e-05, + "loss": 0.1732, + "step": 56369 + }, + { + "epoch": 2.17, + "learning_rate": 1.3888176455501411e-05, + "loss": 0.224, + "step": 56370 + }, + { + "epoch": 2.17, + "learning_rate": 1.3887535265452681e-05, + "loss": 0.2558, + "step": 56371 + }, + { + "epoch": 2.17, + "learning_rate": 1.388689407540395e-05, + "loss": 0.2048, + "step": 56372 + }, + { + "epoch": 2.17, + "learning_rate": 1.388625288535522e-05, + "loss": 0.4417, + "step": 56373 + }, + { + "epoch": 2.17, + "learning_rate": 1.3885611695306488e-05, + "loss": 0.3185, + "step": 56374 + }, + { + "epoch": 2.17, + "learning_rate": 1.3884970505257758e-05, + "loss": 0.2228, + "step": 56375 + }, + { + "epoch": 2.17, + "learning_rate": 1.388432931520903e-05, + "loss": 0.2765, + "step": 56376 + }, + { + "epoch": 2.17, + "learning_rate": 1.3883688125160298e-05, + "loss": 0.1935, + "step": 56377 + }, + { + "epoch": 2.17, + "learning_rate": 1.3883046935111568e-05, + "loss": 0.2689, + "step": 56378 + }, + { + "epoch": 2.17, + "learning_rate": 1.3882405745062838e-05, + "loss": 0.2354, + "step": 56379 + }, + { + "epoch": 2.17, + "learning_rate": 1.3881764555014107e-05, + "loss": 0.2445, + "step": 56380 + }, + { + "epoch": 2.17, + "learning_rate": 1.3881123364965377e-05, + "loss": 0.2391, + "step": 56381 + }, + { + "epoch": 2.17, + "learning_rate": 1.3880482174916645e-05, + "loss": 0.2491, + "step": 56382 + }, + { + "epoch": 2.17, + "learning_rate": 1.3879840984867915e-05, + "loss": 0.2249, + "step": 56383 + }, + { + "epoch": 2.17, + "learning_rate": 1.3879199794819183e-05, + "loss": 0.2705, + "step": 56384 + }, + { + "epoch": 2.17, + "learning_rate": 1.3878558604770455e-05, + "loss": 0.1508, + "step": 56385 + }, + { + "epoch": 2.17, + "learning_rate": 1.3877917414721725e-05, + "loss": 0.1891, + "step": 56386 + }, + { + "epoch": 2.17, + "learning_rate": 1.3877276224672995e-05, + "loss": 0.3141, + "step": 56387 + }, + { + "epoch": 2.17, + "learning_rate": 1.3876635034624263e-05, + "loss": 0.3525, + "step": 56388 + }, + { + "epoch": 2.17, + "learning_rate": 1.3875993844575533e-05, + "loss": 0.1415, + "step": 56389 + }, + { + "epoch": 2.17, + "learning_rate": 1.3875352654526802e-05, + "loss": 0.1904, + "step": 56390 + }, + { + "epoch": 2.17, + "learning_rate": 1.3874711464478072e-05, + "loss": 0.2165, + "step": 56391 + }, + { + "epoch": 2.17, + "learning_rate": 1.387407027442934e-05, + "loss": 0.2687, + "step": 56392 + }, + { + "epoch": 2.17, + "learning_rate": 1.387342908438061e-05, + "loss": 0.3275, + "step": 56393 + }, + { + "epoch": 2.17, + "learning_rate": 1.3872787894331882e-05, + "loss": 0.1906, + "step": 56394 + }, + { + "epoch": 2.17, + "learning_rate": 1.3872146704283152e-05, + "loss": 0.3503, + "step": 56395 + }, + { + "epoch": 2.17, + "learning_rate": 1.387150551423442e-05, + "loss": 0.2726, + "step": 56396 + }, + { + "epoch": 2.17, + "learning_rate": 1.387086432418569e-05, + "loss": 0.1857, + "step": 56397 + }, + { + "epoch": 2.17, + "learning_rate": 1.3870223134136959e-05, + "loss": 0.2745, + "step": 56398 + }, + { + "epoch": 2.17, + "learning_rate": 1.3869581944088229e-05, + "loss": 0.2563, + "step": 56399 + }, + { + "epoch": 2.17, + "learning_rate": 1.3868940754039497e-05, + "loss": 0.2231, + "step": 56400 + }, + { + "epoch": 2.17, + "learning_rate": 1.3868299563990767e-05, + "loss": 0.2087, + "step": 56401 + }, + { + "epoch": 2.17, + "learning_rate": 1.3867658373942035e-05, + "loss": 0.3346, + "step": 56402 + }, + { + "epoch": 2.17, + "learning_rate": 1.3867017183893307e-05, + "loss": 0.1674, + "step": 56403 + }, + { + "epoch": 2.17, + "learning_rate": 1.3866375993844577e-05, + "loss": 0.1968, + "step": 56404 + }, + { + "epoch": 2.17, + "learning_rate": 1.3865734803795847e-05, + "loss": 0.2428, + "step": 56405 + }, + { + "epoch": 2.17, + "learning_rate": 1.3865093613747115e-05, + "loss": 0.1903, + "step": 56406 + }, + { + "epoch": 2.17, + "learning_rate": 1.3864452423698385e-05, + "loss": 0.2812, + "step": 56407 + }, + { + "epoch": 2.17, + "learning_rate": 1.3863811233649654e-05, + "loss": 0.3197, + "step": 56408 + }, + { + "epoch": 2.17, + "learning_rate": 1.3863170043600924e-05, + "loss": 0.2888, + "step": 56409 + }, + { + "epoch": 2.17, + "learning_rate": 1.3862528853552192e-05, + "loss": 0.1511, + "step": 56410 + }, + { + "epoch": 2.17, + "learning_rate": 1.3861887663503464e-05, + "loss": 0.2196, + "step": 56411 + }, + { + "epoch": 2.17, + "learning_rate": 1.3861246473454734e-05, + "loss": 0.1383, + "step": 56412 + }, + { + "epoch": 2.17, + "learning_rate": 1.3860605283406004e-05, + "loss": 0.1638, + "step": 56413 + }, + { + "epoch": 2.17, + "learning_rate": 1.3859964093357272e-05, + "loss": 0.299, + "step": 56414 + }, + { + "epoch": 2.17, + "learning_rate": 1.3859322903308542e-05, + "loss": 0.1843, + "step": 56415 + }, + { + "epoch": 2.17, + "learning_rate": 1.385868171325981e-05, + "loss": 0.4055, + "step": 56416 + }, + { + "epoch": 2.17, + "learning_rate": 1.385804052321108e-05, + "loss": 0.2059, + "step": 56417 + }, + { + "epoch": 2.17, + "learning_rate": 1.3857399333162349e-05, + "loss": 0.299, + "step": 56418 + }, + { + "epoch": 2.17, + "learning_rate": 1.3856758143113619e-05, + "loss": 0.2472, + "step": 56419 + }, + { + "epoch": 2.17, + "learning_rate": 1.385611695306489e-05, + "loss": 0.2758, + "step": 56420 + }, + { + "epoch": 2.17, + "learning_rate": 1.3855475763016159e-05, + "loss": 0.2414, + "step": 56421 + }, + { + "epoch": 2.17, + "learning_rate": 1.3854834572967429e-05, + "loss": 0.184, + "step": 56422 + }, + { + "epoch": 2.17, + "learning_rate": 1.3854193382918699e-05, + "loss": 0.2534, + "step": 56423 + }, + { + "epoch": 2.17, + "learning_rate": 1.3853552192869967e-05, + "loss": 0.1867, + "step": 56424 + }, + { + "epoch": 2.17, + "learning_rate": 1.3852911002821237e-05, + "loss": 0.1846, + "step": 56425 + }, + { + "epoch": 2.17, + "learning_rate": 1.3852269812772506e-05, + "loss": 0.2031, + "step": 56426 + }, + { + "epoch": 2.17, + "learning_rate": 1.3851628622723776e-05, + "loss": 0.1857, + "step": 56427 + }, + { + "epoch": 2.17, + "learning_rate": 1.3850987432675044e-05, + "loss": 0.2694, + "step": 56428 + }, + { + "epoch": 2.17, + "learning_rate": 1.3850346242626316e-05, + "loss": 0.2387, + "step": 56429 + }, + { + "epoch": 2.17, + "learning_rate": 1.3849705052577586e-05, + "loss": 0.1271, + "step": 56430 + }, + { + "epoch": 2.17, + "learning_rate": 1.3849063862528856e-05, + "loss": 0.2608, + "step": 56431 + }, + { + "epoch": 2.17, + "learning_rate": 1.3848422672480124e-05, + "loss": 0.2944, + "step": 56432 + }, + { + "epoch": 2.17, + "learning_rate": 1.3847781482431394e-05, + "loss": 0.1977, + "step": 56433 + }, + { + "epoch": 2.17, + "learning_rate": 1.3847140292382662e-05, + "loss": 0.2562, + "step": 56434 + }, + { + "epoch": 2.17, + "learning_rate": 1.3846499102333932e-05, + "loss": 0.229, + "step": 56435 + }, + { + "epoch": 2.17, + "learning_rate": 1.38458579122852e-05, + "loss": 0.262, + "step": 56436 + }, + { + "epoch": 2.17, + "learning_rate": 1.384521672223647e-05, + "loss": 0.2983, + "step": 56437 + }, + { + "epoch": 2.17, + "learning_rate": 1.3844575532187742e-05, + "loss": 0.2255, + "step": 56438 + }, + { + "epoch": 2.17, + "learning_rate": 1.384393434213901e-05, + "loss": 0.3114, + "step": 56439 + }, + { + "epoch": 2.17, + "learning_rate": 1.384329315209028e-05, + "loss": 0.1671, + "step": 56440 + }, + { + "epoch": 2.17, + "learning_rate": 1.384265196204155e-05, + "loss": 0.2797, + "step": 56441 + }, + { + "epoch": 2.17, + "learning_rate": 1.384201077199282e-05, + "loss": 0.1851, + "step": 56442 + }, + { + "epoch": 2.17, + "learning_rate": 1.384136958194409e-05, + "loss": 0.1792, + "step": 56443 + }, + { + "epoch": 2.17, + "learning_rate": 1.3840728391895358e-05, + "loss": 0.1508, + "step": 56444 + }, + { + "epoch": 2.17, + "learning_rate": 1.3840087201846628e-05, + "loss": 0.3423, + "step": 56445 + }, + { + "epoch": 2.17, + "learning_rate": 1.3839446011797896e-05, + "loss": 0.191, + "step": 56446 + }, + { + "epoch": 2.17, + "learning_rate": 1.3838804821749168e-05, + "loss": 0.2091, + "step": 56447 + }, + { + "epoch": 2.17, + "learning_rate": 1.3838163631700438e-05, + "loss": 0.2162, + "step": 56448 + }, + { + "epoch": 2.17, + "learning_rate": 1.3837522441651708e-05, + "loss": 0.3026, + "step": 56449 + }, + { + "epoch": 2.17, + "learning_rate": 1.3836881251602976e-05, + "loss": 0.1942, + "step": 56450 + }, + { + "epoch": 2.17, + "learning_rate": 1.3836240061554246e-05, + "loss": 0.2932, + "step": 56451 + }, + { + "epoch": 2.17, + "learning_rate": 1.3835598871505514e-05, + "loss": 0.1748, + "step": 56452 + }, + { + "epoch": 2.17, + "learning_rate": 1.3834957681456784e-05, + "loss": 0.4152, + "step": 56453 + }, + { + "epoch": 2.17, + "learning_rate": 1.3834316491408053e-05, + "loss": 0.2993, + "step": 56454 + }, + { + "epoch": 2.17, + "learning_rate": 1.3833675301359323e-05, + "loss": 0.1856, + "step": 56455 + }, + { + "epoch": 2.17, + "learning_rate": 1.3833034111310594e-05, + "loss": 0.3749, + "step": 56456 + }, + { + "epoch": 2.17, + "learning_rate": 1.3832392921261864e-05, + "loss": 0.2352, + "step": 56457 + }, + { + "epoch": 2.17, + "learning_rate": 1.3831751731213133e-05, + "loss": 0.2616, + "step": 56458 + }, + { + "epoch": 2.17, + "learning_rate": 1.3831110541164403e-05, + "loss": 0.2298, + "step": 56459 + }, + { + "epoch": 2.17, + "learning_rate": 1.3830469351115671e-05, + "loss": 0.1045, + "step": 56460 + }, + { + "epoch": 2.17, + "learning_rate": 1.3829828161066941e-05, + "loss": 0.1682, + "step": 56461 + }, + { + "epoch": 2.17, + "learning_rate": 1.382918697101821e-05, + "loss": 0.2425, + "step": 56462 + }, + { + "epoch": 2.17, + "learning_rate": 1.382854578096948e-05, + "loss": 0.1957, + "step": 56463 + }, + { + "epoch": 2.17, + "learning_rate": 1.3827904590920748e-05, + "loss": 0.3245, + "step": 56464 + }, + { + "epoch": 2.17, + "learning_rate": 1.382726340087202e-05, + "loss": 0.0984, + "step": 56465 + }, + { + "epoch": 2.17, + "learning_rate": 1.382662221082329e-05, + "loss": 0.2187, + "step": 56466 + }, + { + "epoch": 2.17, + "learning_rate": 1.382598102077456e-05, + "loss": 0.1993, + "step": 56467 + }, + { + "epoch": 2.17, + "learning_rate": 1.3825339830725828e-05, + "loss": 0.1934, + "step": 56468 + }, + { + "epoch": 2.17, + "learning_rate": 1.3824698640677098e-05, + "loss": 0.2215, + "step": 56469 + }, + { + "epoch": 2.17, + "learning_rate": 1.3824057450628366e-05, + "loss": 0.2165, + "step": 56470 + }, + { + "epoch": 2.17, + "learning_rate": 1.3823416260579636e-05, + "loss": 0.1141, + "step": 56471 + }, + { + "epoch": 2.17, + "learning_rate": 1.3822775070530905e-05, + "loss": 0.1736, + "step": 56472 + }, + { + "epoch": 2.17, + "learning_rate": 1.3822133880482175e-05, + "loss": 0.225, + "step": 56473 + }, + { + "epoch": 2.17, + "learning_rate": 1.3821492690433446e-05, + "loss": 0.298, + "step": 56474 + }, + { + "epoch": 2.17, + "learning_rate": 1.3820851500384716e-05, + "loss": 0.2468, + "step": 56475 + }, + { + "epoch": 2.17, + "learning_rate": 1.3820210310335985e-05, + "loss": 0.1295, + "step": 56476 + }, + { + "epoch": 2.17, + "learning_rate": 1.3819569120287255e-05, + "loss": 0.4083, + "step": 56477 + }, + { + "epoch": 2.17, + "learning_rate": 1.3818927930238523e-05, + "loss": 0.2673, + "step": 56478 + }, + { + "epoch": 2.17, + "learning_rate": 1.3818286740189793e-05, + "loss": 0.2807, + "step": 56479 + }, + { + "epoch": 2.17, + "learning_rate": 1.3817645550141061e-05, + "loss": 0.2174, + "step": 56480 + }, + { + "epoch": 2.17, + "learning_rate": 1.3817004360092331e-05, + "loss": 0.3495, + "step": 56481 + }, + { + "epoch": 2.17, + "learning_rate": 1.38163631700436e-05, + "loss": 0.3325, + "step": 56482 + }, + { + "epoch": 2.17, + "learning_rate": 1.3815721979994871e-05, + "loss": 0.2185, + "step": 56483 + }, + { + "epoch": 2.17, + "learning_rate": 1.3815080789946141e-05, + "loss": 0.1754, + "step": 56484 + }, + { + "epoch": 2.17, + "learning_rate": 1.3814439599897412e-05, + "loss": 0.3559, + "step": 56485 + }, + { + "epoch": 2.17, + "learning_rate": 1.381379840984868e-05, + "loss": 0.1264, + "step": 56486 + }, + { + "epoch": 2.17, + "learning_rate": 1.381315721979995e-05, + "loss": 0.3013, + "step": 56487 + }, + { + "epoch": 2.17, + "learning_rate": 1.3812516029751218e-05, + "loss": 0.1612, + "step": 56488 + }, + { + "epoch": 2.17, + "learning_rate": 1.3811874839702488e-05, + "loss": 0.1358, + "step": 56489 + }, + { + "epoch": 2.17, + "learning_rate": 1.3811233649653757e-05, + "loss": 0.1, + "step": 56490 + }, + { + "epoch": 2.17, + "learning_rate": 1.3810592459605027e-05, + "loss": 0.2622, + "step": 56491 + }, + { + "epoch": 2.17, + "learning_rate": 1.3809951269556298e-05, + "loss": 0.3418, + "step": 56492 + }, + { + "epoch": 2.17, + "learning_rate": 1.3809310079507568e-05, + "loss": 0.1881, + "step": 56493 + }, + { + "epoch": 2.17, + "learning_rate": 1.3808668889458837e-05, + "loss": 0.3296, + "step": 56494 + }, + { + "epoch": 2.17, + "learning_rate": 1.3808027699410107e-05, + "loss": 0.2373, + "step": 56495 + }, + { + "epoch": 2.17, + "learning_rate": 1.3807386509361375e-05, + "loss": 0.2017, + "step": 56496 + }, + { + "epoch": 2.17, + "learning_rate": 1.3806745319312645e-05, + "loss": 0.0882, + "step": 56497 + }, + { + "epoch": 2.17, + "learning_rate": 1.3806104129263913e-05, + "loss": 0.2357, + "step": 56498 + }, + { + "epoch": 2.17, + "learning_rate": 1.3805462939215183e-05, + "loss": 0.2078, + "step": 56499 + }, + { + "epoch": 2.17, + "learning_rate": 1.3804821749166452e-05, + "loss": 0.1442, + "step": 56500 + }, + { + "epoch": 2.17, + "learning_rate": 1.3804180559117723e-05, + "loss": 0.1458, + "step": 56501 + }, + { + "epoch": 2.17, + "learning_rate": 1.3803539369068993e-05, + "loss": 0.2165, + "step": 56502 + }, + { + "epoch": 2.17, + "learning_rate": 1.3802898179020263e-05, + "loss": 0.2315, + "step": 56503 + }, + { + "epoch": 2.17, + "learning_rate": 1.3802256988971532e-05, + "loss": 0.1583, + "step": 56504 + }, + { + "epoch": 2.17, + "learning_rate": 1.3801615798922802e-05, + "loss": 0.1809, + "step": 56505 + }, + { + "epoch": 2.17, + "learning_rate": 1.380097460887407e-05, + "loss": 0.2386, + "step": 56506 + }, + { + "epoch": 2.17, + "learning_rate": 1.380033341882534e-05, + "loss": 0.3089, + "step": 56507 + }, + { + "epoch": 2.17, + "learning_rate": 1.3799692228776608e-05, + "loss": 0.2647, + "step": 56508 + }, + { + "epoch": 2.17, + "learning_rate": 1.379905103872788e-05, + "loss": 0.1971, + "step": 56509 + }, + { + "epoch": 2.17, + "learning_rate": 1.379840984867915e-05, + "loss": 0.2523, + "step": 56510 + }, + { + "epoch": 2.17, + "learning_rate": 1.379776865863042e-05, + "loss": 0.3078, + "step": 56511 + }, + { + "epoch": 2.17, + "learning_rate": 1.3797127468581689e-05, + "loss": 0.1265, + "step": 56512 + }, + { + "epoch": 2.17, + "learning_rate": 1.3796486278532959e-05, + "loss": 0.3451, + "step": 56513 + }, + { + "epoch": 2.17, + "learning_rate": 1.3795845088484227e-05, + "loss": 0.2843, + "step": 56514 + }, + { + "epoch": 2.17, + "learning_rate": 1.3795203898435497e-05, + "loss": 0.2161, + "step": 56515 + }, + { + "epoch": 2.17, + "learning_rate": 1.3794562708386765e-05, + "loss": 0.2221, + "step": 56516 + }, + { + "epoch": 2.17, + "learning_rate": 1.3793921518338035e-05, + "loss": 0.191, + "step": 56517 + }, + { + "epoch": 2.17, + "learning_rate": 1.3793280328289307e-05, + "loss": 0.2346, + "step": 56518 + }, + { + "epoch": 2.17, + "learning_rate": 1.3792639138240577e-05, + "loss": 0.2016, + "step": 56519 + }, + { + "epoch": 2.17, + "learning_rate": 1.3791997948191845e-05, + "loss": 0.175, + "step": 56520 + }, + { + "epoch": 2.17, + "learning_rate": 1.3791356758143115e-05, + "loss": 0.2239, + "step": 56521 + }, + { + "epoch": 2.17, + "learning_rate": 1.3790715568094384e-05, + "loss": 0.2792, + "step": 56522 + }, + { + "epoch": 2.17, + "learning_rate": 1.3790074378045654e-05, + "loss": 0.1846, + "step": 56523 + }, + { + "epoch": 2.17, + "learning_rate": 1.3789433187996922e-05, + "loss": 0.2583, + "step": 56524 + }, + { + "epoch": 2.17, + "learning_rate": 1.3788791997948192e-05, + "loss": 0.3231, + "step": 56525 + }, + { + "epoch": 2.17, + "learning_rate": 1.378815080789946e-05, + "loss": 0.3732, + "step": 56526 + }, + { + "epoch": 2.17, + "learning_rate": 1.3787509617850732e-05, + "loss": 0.3336, + "step": 56527 + }, + { + "epoch": 2.17, + "learning_rate": 1.3786868427802002e-05, + "loss": 0.3669, + "step": 56528 + }, + { + "epoch": 2.17, + "learning_rate": 1.3786227237753272e-05, + "loss": 0.2616, + "step": 56529 + }, + { + "epoch": 2.17, + "learning_rate": 1.378558604770454e-05, + "loss": 0.24, + "step": 56530 + }, + { + "epoch": 2.17, + "learning_rate": 1.378494485765581e-05, + "loss": 0.1978, + "step": 56531 + }, + { + "epoch": 2.17, + "learning_rate": 1.3784303667607079e-05, + "loss": 0.2382, + "step": 56532 + }, + { + "epoch": 2.17, + "learning_rate": 1.3783662477558349e-05, + "loss": 0.2661, + "step": 56533 + }, + { + "epoch": 2.17, + "learning_rate": 1.3783021287509617e-05, + "loss": 0.1797, + "step": 56534 + }, + { + "epoch": 2.17, + "learning_rate": 1.3782380097460887e-05, + "loss": 0.368, + "step": 56535 + }, + { + "epoch": 2.17, + "learning_rate": 1.3781738907412159e-05, + "loss": 0.1994, + "step": 56536 + }, + { + "epoch": 2.17, + "learning_rate": 1.3781097717363429e-05, + "loss": 0.2465, + "step": 56537 + }, + { + "epoch": 2.17, + "learning_rate": 1.3780456527314697e-05, + "loss": 0.2112, + "step": 56538 + }, + { + "epoch": 2.17, + "learning_rate": 1.3779815337265967e-05, + "loss": 0.2884, + "step": 56539 + }, + { + "epoch": 2.17, + "learning_rate": 1.3779174147217236e-05, + "loss": 0.2631, + "step": 56540 + }, + { + "epoch": 2.17, + "learning_rate": 1.3778532957168506e-05, + "loss": 0.2207, + "step": 56541 + }, + { + "epoch": 2.17, + "learning_rate": 1.3777891767119774e-05, + "loss": 0.1986, + "step": 56542 + }, + { + "epoch": 2.17, + "learning_rate": 1.3777250577071044e-05, + "loss": 0.1853, + "step": 56543 + }, + { + "epoch": 2.17, + "learning_rate": 1.3776609387022312e-05, + "loss": 0.2857, + "step": 56544 + }, + { + "epoch": 2.17, + "learning_rate": 1.3775968196973584e-05, + "loss": 0.1469, + "step": 56545 + }, + { + "epoch": 2.17, + "learning_rate": 1.3775327006924854e-05, + "loss": 0.2434, + "step": 56546 + }, + { + "epoch": 2.17, + "learning_rate": 1.3774685816876124e-05, + "loss": 0.1984, + "step": 56547 + }, + { + "epoch": 2.17, + "learning_rate": 1.3774044626827392e-05, + "loss": 0.1879, + "step": 56548 + }, + { + "epoch": 2.17, + "learning_rate": 1.3773403436778662e-05, + "loss": 0.2677, + "step": 56549 + }, + { + "epoch": 2.17, + "learning_rate": 1.377276224672993e-05, + "loss": 0.2192, + "step": 56550 + }, + { + "epoch": 2.17, + "learning_rate": 1.37721210566812e-05, + "loss": 0.1883, + "step": 56551 + }, + { + "epoch": 2.17, + "learning_rate": 1.3771479866632469e-05, + "loss": 0.1687, + "step": 56552 + }, + { + "epoch": 2.17, + "learning_rate": 1.377083867658374e-05, + "loss": 0.1749, + "step": 56553 + }, + { + "epoch": 2.17, + "learning_rate": 1.3770197486535011e-05, + "loss": 0.2515, + "step": 56554 + }, + { + "epoch": 2.17, + "learning_rate": 1.3769556296486281e-05, + "loss": 0.2904, + "step": 56555 + }, + { + "epoch": 2.17, + "learning_rate": 1.376891510643755e-05, + "loss": 0.2995, + "step": 56556 + }, + { + "epoch": 2.17, + "learning_rate": 1.376827391638882e-05, + "loss": 0.2165, + "step": 56557 + }, + { + "epoch": 2.17, + "learning_rate": 1.3767632726340088e-05, + "loss": 0.1507, + "step": 56558 + }, + { + "epoch": 2.17, + "learning_rate": 1.3766991536291358e-05, + "loss": 0.1689, + "step": 56559 + }, + { + "epoch": 2.17, + "learning_rate": 1.3766350346242626e-05, + "loss": 0.2191, + "step": 56560 + }, + { + "epoch": 2.17, + "learning_rate": 1.3765709156193896e-05, + "loss": 0.1087, + "step": 56561 + }, + { + "epoch": 2.17, + "learning_rate": 1.3765067966145164e-05, + "loss": 0.2676, + "step": 56562 + }, + { + "epoch": 2.17, + "learning_rate": 1.3764426776096436e-05, + "loss": 0.1928, + "step": 56563 + }, + { + "epoch": 2.17, + "learning_rate": 1.3763785586047706e-05, + "loss": 0.2108, + "step": 56564 + }, + { + "epoch": 2.17, + "learning_rate": 1.3763144395998976e-05, + "loss": 0.1577, + "step": 56565 + }, + { + "epoch": 2.17, + "learning_rate": 1.3762503205950244e-05, + "loss": 0.1007, + "step": 56566 + }, + { + "epoch": 2.17, + "learning_rate": 1.3761862015901514e-05, + "loss": 0.2852, + "step": 56567 + }, + { + "epoch": 2.17, + "learning_rate": 1.3761220825852783e-05, + "loss": 0.2063, + "step": 56568 + }, + { + "epoch": 2.17, + "learning_rate": 1.3760579635804053e-05, + "loss": 0.2352, + "step": 56569 + }, + { + "epoch": 2.17, + "learning_rate": 1.3759938445755321e-05, + "loss": 0.3223, + "step": 56570 + }, + { + "epoch": 2.17, + "learning_rate": 1.3759297255706591e-05, + "loss": 0.2903, + "step": 56571 + }, + { + "epoch": 2.18, + "learning_rate": 1.3758656065657863e-05, + "loss": 0.2375, + "step": 56572 + }, + { + "epoch": 2.18, + "learning_rate": 1.3758014875609133e-05, + "loss": 0.2969, + "step": 56573 + }, + { + "epoch": 2.18, + "learning_rate": 1.3757373685560401e-05, + "loss": 0.1626, + "step": 56574 + }, + { + "epoch": 2.18, + "learning_rate": 1.3756732495511671e-05, + "loss": 0.1688, + "step": 56575 + }, + { + "epoch": 2.18, + "learning_rate": 1.375609130546294e-05, + "loss": 0.1966, + "step": 56576 + }, + { + "epoch": 2.18, + "learning_rate": 1.375545011541421e-05, + "loss": 0.1918, + "step": 56577 + }, + { + "epoch": 2.18, + "learning_rate": 1.3754808925365478e-05, + "loss": 0.1648, + "step": 56578 + }, + { + "epoch": 2.18, + "learning_rate": 1.3754167735316748e-05, + "loss": 0.0976, + "step": 56579 + }, + { + "epoch": 2.18, + "learning_rate": 1.3753526545268016e-05, + "loss": 0.2314, + "step": 56580 + }, + { + "epoch": 2.18, + "learning_rate": 1.375288535521929e-05, + "loss": 0.2178, + "step": 56581 + }, + { + "epoch": 2.18, + "learning_rate": 1.3752244165170558e-05, + "loss": 0.273, + "step": 56582 + }, + { + "epoch": 2.18, + "learning_rate": 1.3751602975121828e-05, + "loss": 0.2342, + "step": 56583 + }, + { + "epoch": 2.18, + "learning_rate": 1.3750961785073096e-05, + "loss": 0.2965, + "step": 56584 + }, + { + "epoch": 2.18, + "learning_rate": 1.3750320595024366e-05, + "loss": 0.2073, + "step": 56585 + }, + { + "epoch": 2.18, + "learning_rate": 1.3749679404975635e-05, + "loss": 0.2275, + "step": 56586 + }, + { + "epoch": 2.18, + "learning_rate": 1.3749038214926905e-05, + "loss": 0.1961, + "step": 56587 + }, + { + "epoch": 2.18, + "learning_rate": 1.3748397024878173e-05, + "loss": 0.1344, + "step": 56588 + }, + { + "epoch": 2.18, + "learning_rate": 1.3747755834829443e-05, + "loss": 0.0842, + "step": 56589 + }, + { + "epoch": 2.18, + "learning_rate": 1.3747114644780715e-05, + "loss": 0.3584, + "step": 56590 + }, + { + "epoch": 2.18, + "learning_rate": 1.3746473454731985e-05, + "loss": 0.379, + "step": 56591 + }, + { + "epoch": 2.18, + "learning_rate": 1.3745832264683253e-05, + "loss": 0.1366, + "step": 56592 + }, + { + "epoch": 2.18, + "learning_rate": 1.3745191074634523e-05, + "loss": 0.1491, + "step": 56593 + }, + { + "epoch": 2.18, + "learning_rate": 1.3744549884585791e-05, + "loss": 0.1897, + "step": 56594 + }, + { + "epoch": 2.18, + "learning_rate": 1.3743908694537061e-05, + "loss": 0.3039, + "step": 56595 + }, + { + "epoch": 2.18, + "learning_rate": 1.374326750448833e-05, + "loss": 0.1726, + "step": 56596 + }, + { + "epoch": 2.18, + "learning_rate": 1.37426263144396e-05, + "loss": 0.1942, + "step": 56597 + }, + { + "epoch": 2.18, + "learning_rate": 1.3741985124390872e-05, + "loss": 0.114, + "step": 56598 + }, + { + "epoch": 2.18, + "learning_rate": 1.3741343934342142e-05, + "loss": 0.2401, + "step": 56599 + }, + { + "epoch": 2.18, + "learning_rate": 1.374070274429341e-05, + "loss": 0.2954, + "step": 56600 + }, + { + "epoch": 2.18, + "learning_rate": 1.374006155424468e-05, + "loss": 0.2098, + "step": 56601 + }, + { + "epoch": 2.18, + "learning_rate": 1.3739420364195948e-05, + "loss": 0.2227, + "step": 56602 + }, + { + "epoch": 2.18, + "learning_rate": 1.3738779174147218e-05, + "loss": 0.1369, + "step": 56603 + }, + { + "epoch": 2.18, + "learning_rate": 1.3738137984098487e-05, + "loss": 0.1544, + "step": 56604 + }, + { + "epoch": 2.18, + "learning_rate": 1.3737496794049757e-05, + "loss": 0.2434, + "step": 56605 + }, + { + "epoch": 2.18, + "learning_rate": 1.3736855604001025e-05, + "loss": 0.206, + "step": 56606 + }, + { + "epoch": 2.18, + "learning_rate": 1.3736214413952297e-05, + "loss": 0.3261, + "step": 56607 + }, + { + "epoch": 2.18, + "learning_rate": 1.3735573223903567e-05, + "loss": 0.2147, + "step": 56608 + }, + { + "epoch": 2.18, + "learning_rate": 1.3734932033854837e-05, + "loss": 0.2142, + "step": 56609 + }, + { + "epoch": 2.18, + "learning_rate": 1.3734290843806105e-05, + "loss": 0.1767, + "step": 56610 + }, + { + "epoch": 2.18, + "learning_rate": 1.3733649653757375e-05, + "loss": 0.3128, + "step": 56611 + }, + { + "epoch": 2.18, + "learning_rate": 1.3733008463708643e-05, + "loss": 0.2827, + "step": 56612 + }, + { + "epoch": 2.18, + "learning_rate": 1.3732367273659913e-05, + "loss": 0.2244, + "step": 56613 + }, + { + "epoch": 2.18, + "learning_rate": 1.3731726083611182e-05, + "loss": 0.3064, + "step": 56614 + }, + { + "epoch": 2.18, + "learning_rate": 1.3731084893562452e-05, + "loss": 0.156, + "step": 56615 + }, + { + "epoch": 2.18, + "learning_rate": 1.3730443703513723e-05, + "loss": 0.3556, + "step": 56616 + }, + { + "epoch": 2.18, + "learning_rate": 1.3729802513464993e-05, + "loss": 0.458, + "step": 56617 + }, + { + "epoch": 2.18, + "learning_rate": 1.3729161323416262e-05, + "loss": 0.369, + "step": 56618 + }, + { + "epoch": 2.18, + "learning_rate": 1.3728520133367532e-05, + "loss": 0.196, + "step": 56619 + }, + { + "epoch": 2.18, + "learning_rate": 1.37278789433188e-05, + "loss": 0.2884, + "step": 56620 + }, + { + "epoch": 2.18, + "learning_rate": 1.372723775327007e-05, + "loss": 0.1652, + "step": 56621 + }, + { + "epoch": 2.18, + "learning_rate": 1.3726596563221339e-05, + "loss": 0.2321, + "step": 56622 + }, + { + "epoch": 2.18, + "learning_rate": 1.3725955373172609e-05, + "loss": 0.2265, + "step": 56623 + }, + { + "epoch": 2.18, + "learning_rate": 1.3725314183123877e-05, + "loss": 0.2136, + "step": 56624 + }, + { + "epoch": 2.18, + "learning_rate": 1.3724672993075149e-05, + "loss": 0.3175, + "step": 56625 + }, + { + "epoch": 2.18, + "learning_rate": 1.3724031803026419e-05, + "loss": 0.1753, + "step": 56626 + }, + { + "epoch": 2.18, + "learning_rate": 1.3723390612977689e-05, + "loss": 0.3729, + "step": 56627 + }, + { + "epoch": 2.18, + "learning_rate": 1.3722749422928957e-05, + "loss": 0.2948, + "step": 56628 + }, + { + "epoch": 2.18, + "learning_rate": 1.3722108232880227e-05, + "loss": 0.3183, + "step": 56629 + }, + { + "epoch": 2.18, + "learning_rate": 1.3721467042831495e-05, + "loss": 0.2199, + "step": 56630 + }, + { + "epoch": 2.18, + "learning_rate": 1.3720825852782765e-05, + "loss": 0.1988, + "step": 56631 + }, + { + "epoch": 2.18, + "learning_rate": 1.3720184662734034e-05, + "loss": 0.4057, + "step": 56632 + }, + { + "epoch": 2.18, + "learning_rate": 1.3719543472685304e-05, + "loss": 0.2306, + "step": 56633 + }, + { + "epoch": 2.18, + "learning_rate": 1.3718902282636575e-05, + "loss": 0.2174, + "step": 56634 + }, + { + "epoch": 2.18, + "learning_rate": 1.3718261092587845e-05, + "loss": 0.2822, + "step": 56635 + }, + { + "epoch": 2.18, + "learning_rate": 1.3717619902539114e-05, + "loss": 0.2743, + "step": 56636 + }, + { + "epoch": 2.18, + "learning_rate": 1.3716978712490384e-05, + "loss": 0.2836, + "step": 56637 + }, + { + "epoch": 2.18, + "learning_rate": 1.3716337522441652e-05, + "loss": 0.2663, + "step": 56638 + }, + { + "epoch": 2.18, + "learning_rate": 1.3715696332392922e-05, + "loss": 0.2451, + "step": 56639 + }, + { + "epoch": 2.18, + "learning_rate": 1.371505514234419e-05, + "loss": 0.3318, + "step": 56640 + }, + { + "epoch": 2.18, + "learning_rate": 1.371441395229546e-05, + "loss": 0.3599, + "step": 56641 + }, + { + "epoch": 2.18, + "learning_rate": 1.3713772762246729e-05, + "loss": 0.2725, + "step": 56642 + }, + { + "epoch": 2.18, + "learning_rate": 1.3713131572198e-05, + "loss": 0.3921, + "step": 56643 + }, + { + "epoch": 2.18, + "learning_rate": 1.371249038214927e-05, + "loss": 0.2193, + "step": 56644 + }, + { + "epoch": 2.18, + "learning_rate": 1.371184919210054e-05, + "loss": 0.2771, + "step": 56645 + }, + { + "epoch": 2.18, + "learning_rate": 1.3711208002051809e-05, + "loss": 0.2134, + "step": 56646 + }, + { + "epoch": 2.18, + "learning_rate": 1.3710566812003079e-05, + "loss": 0.3043, + "step": 56647 + }, + { + "epoch": 2.18, + "learning_rate": 1.3709925621954347e-05, + "loss": 0.2521, + "step": 56648 + }, + { + "epoch": 2.18, + "learning_rate": 1.3709284431905617e-05, + "loss": 0.3149, + "step": 56649 + }, + { + "epoch": 2.18, + "learning_rate": 1.3708643241856886e-05, + "loss": 0.1513, + "step": 56650 + }, + { + "epoch": 2.18, + "learning_rate": 1.3708002051808156e-05, + "loss": 0.2219, + "step": 56651 + }, + { + "epoch": 2.18, + "learning_rate": 1.3707360861759427e-05, + "loss": 0.2219, + "step": 56652 + }, + { + "epoch": 2.18, + "learning_rate": 1.3706719671710697e-05, + "loss": 0.1526, + "step": 56653 + }, + { + "epoch": 2.18, + "learning_rate": 1.3706078481661966e-05, + "loss": 0.1531, + "step": 56654 + }, + { + "epoch": 2.18, + "learning_rate": 1.3705437291613236e-05, + "loss": 0.2125, + "step": 56655 + }, + { + "epoch": 2.18, + "learning_rate": 1.3704796101564504e-05, + "loss": 0.3814, + "step": 56656 + }, + { + "epoch": 2.18, + "learning_rate": 1.3704154911515774e-05, + "loss": 0.1578, + "step": 56657 + }, + { + "epoch": 2.18, + "learning_rate": 1.3703513721467042e-05, + "loss": 0.1659, + "step": 56658 + }, + { + "epoch": 2.18, + "learning_rate": 1.3702872531418312e-05, + "loss": 0.2289, + "step": 56659 + }, + { + "epoch": 2.18, + "learning_rate": 1.370223134136958e-05, + "loss": 0.1868, + "step": 56660 + }, + { + "epoch": 2.18, + "learning_rate": 1.3701590151320854e-05, + "loss": 0.228, + "step": 56661 + }, + { + "epoch": 2.18, + "learning_rate": 1.3700948961272122e-05, + "loss": 0.3201, + "step": 56662 + }, + { + "epoch": 2.18, + "learning_rate": 1.3700307771223393e-05, + "loss": 0.1921, + "step": 56663 + }, + { + "epoch": 2.18, + "learning_rate": 1.369966658117466e-05, + "loss": 0.2161, + "step": 56664 + }, + { + "epoch": 2.18, + "learning_rate": 1.3699025391125931e-05, + "loss": 0.2128, + "step": 56665 + }, + { + "epoch": 2.18, + "learning_rate": 1.36983842010772e-05, + "loss": 0.2841, + "step": 56666 + }, + { + "epoch": 2.18, + "learning_rate": 1.369774301102847e-05, + "loss": 0.2434, + "step": 56667 + }, + { + "epoch": 2.18, + "learning_rate": 1.3697101820979738e-05, + "loss": 0.1778, + "step": 56668 + }, + { + "epoch": 2.18, + "learning_rate": 1.3696460630931008e-05, + "loss": 0.0939, + "step": 56669 + }, + { + "epoch": 2.18, + "learning_rate": 1.369581944088228e-05, + "loss": 0.2262, + "step": 56670 + }, + { + "epoch": 2.18, + "learning_rate": 1.369517825083355e-05, + "loss": 0.3161, + "step": 56671 + }, + { + "epoch": 2.18, + "learning_rate": 1.3694537060784818e-05, + "loss": 0.1493, + "step": 56672 + }, + { + "epoch": 2.18, + "learning_rate": 1.3693895870736088e-05, + "loss": 0.2175, + "step": 56673 + }, + { + "epoch": 2.18, + "learning_rate": 1.3693254680687356e-05, + "loss": 0.1634, + "step": 56674 + }, + { + "epoch": 2.18, + "learning_rate": 1.3692613490638626e-05, + "loss": 0.1382, + "step": 56675 + }, + { + "epoch": 2.18, + "learning_rate": 1.3691972300589894e-05, + "loss": 0.1249, + "step": 56676 + }, + { + "epoch": 2.18, + "learning_rate": 1.3691331110541164e-05, + "loss": 0.2584, + "step": 56677 + }, + { + "epoch": 2.18, + "learning_rate": 1.3690689920492433e-05, + "loss": 0.2305, + "step": 56678 + }, + { + "epoch": 2.18, + "learning_rate": 1.3690048730443706e-05, + "loss": 0.2465, + "step": 56679 + }, + { + "epoch": 2.18, + "learning_rate": 1.3689407540394974e-05, + "loss": 0.2714, + "step": 56680 + }, + { + "epoch": 2.18, + "learning_rate": 1.3688766350346244e-05, + "loss": 0.1075, + "step": 56681 + }, + { + "epoch": 2.18, + "learning_rate": 1.3688125160297513e-05, + "loss": 0.2254, + "step": 56682 + }, + { + "epoch": 2.18, + "learning_rate": 1.3687483970248783e-05, + "loss": 0.3556, + "step": 56683 + }, + { + "epoch": 2.18, + "learning_rate": 1.3686842780200051e-05, + "loss": 0.0963, + "step": 56684 + }, + { + "epoch": 2.18, + "learning_rate": 1.3686201590151321e-05, + "loss": 0.4155, + "step": 56685 + }, + { + "epoch": 2.18, + "learning_rate": 1.368556040010259e-05, + "loss": 0.3795, + "step": 56686 + }, + { + "epoch": 2.18, + "learning_rate": 1.368491921005386e-05, + "loss": 0.0926, + "step": 56687 + }, + { + "epoch": 2.18, + "learning_rate": 1.3684278020005131e-05, + "loss": 0.1496, + "step": 56688 + }, + { + "epoch": 2.18, + "learning_rate": 1.3683636829956401e-05, + "loss": 0.2621, + "step": 56689 + }, + { + "epoch": 2.18, + "learning_rate": 1.368299563990767e-05, + "loss": 0.2393, + "step": 56690 + }, + { + "epoch": 2.18, + "learning_rate": 1.368235444985894e-05, + "loss": 0.2283, + "step": 56691 + }, + { + "epoch": 2.18, + "learning_rate": 1.3681713259810208e-05, + "loss": 0.2449, + "step": 56692 + }, + { + "epoch": 2.18, + "learning_rate": 1.3681072069761478e-05, + "loss": 0.1525, + "step": 56693 + }, + { + "epoch": 2.18, + "learning_rate": 1.3680430879712746e-05, + "loss": 0.1898, + "step": 56694 + }, + { + "epoch": 2.18, + "learning_rate": 1.3679789689664016e-05, + "loss": 0.2498, + "step": 56695 + }, + { + "epoch": 2.18, + "learning_rate": 1.3679148499615288e-05, + "loss": 0.2758, + "step": 56696 + }, + { + "epoch": 2.18, + "learning_rate": 1.3678507309566558e-05, + "loss": 0.3112, + "step": 56697 + }, + { + "epoch": 2.18, + "learning_rate": 1.3677866119517826e-05, + "loss": 0.3353, + "step": 56698 + }, + { + "epoch": 2.18, + "learning_rate": 1.3677224929469096e-05, + "loss": 0.1314, + "step": 56699 + }, + { + "epoch": 2.18, + "learning_rate": 1.3676583739420365e-05, + "loss": 0.4, + "step": 56700 + }, + { + "epoch": 2.18, + "learning_rate": 1.3675942549371635e-05, + "loss": 0.3238, + "step": 56701 + }, + { + "epoch": 2.18, + "learning_rate": 1.3675301359322903e-05, + "loss": 0.1481, + "step": 56702 + }, + { + "epoch": 2.18, + "learning_rate": 1.3674660169274173e-05, + "loss": 0.1646, + "step": 56703 + }, + { + "epoch": 2.18, + "learning_rate": 1.3674018979225441e-05, + "loss": 0.2139, + "step": 56704 + }, + { + "epoch": 2.18, + "learning_rate": 1.3673377789176713e-05, + "loss": 0.2325, + "step": 56705 + }, + { + "epoch": 2.18, + "learning_rate": 1.3672736599127983e-05, + "loss": 0.2087, + "step": 56706 + }, + { + "epoch": 2.18, + "learning_rate": 1.3672095409079253e-05, + "loss": 0.3512, + "step": 56707 + }, + { + "epoch": 2.18, + "learning_rate": 1.3671454219030522e-05, + "loss": 0.2906, + "step": 56708 + }, + { + "epoch": 2.18, + "learning_rate": 1.3670813028981792e-05, + "loss": 0.3223, + "step": 56709 + }, + { + "epoch": 2.18, + "learning_rate": 1.367017183893306e-05, + "loss": 0.2199, + "step": 56710 + }, + { + "epoch": 2.18, + "learning_rate": 1.366953064888433e-05, + "loss": 0.2186, + "step": 56711 + }, + { + "epoch": 2.18, + "learning_rate": 1.3668889458835598e-05, + "loss": 0.301, + "step": 56712 + }, + { + "epoch": 2.18, + "learning_rate": 1.3668248268786868e-05, + "loss": 0.2403, + "step": 56713 + }, + { + "epoch": 2.18, + "learning_rate": 1.366760707873814e-05, + "loss": 0.1822, + "step": 56714 + }, + { + "epoch": 2.18, + "learning_rate": 1.366696588868941e-05, + "loss": 0.2514, + "step": 56715 + }, + { + "epoch": 2.18, + "learning_rate": 1.3666324698640678e-05, + "loss": 0.1926, + "step": 56716 + }, + { + "epoch": 2.18, + "learning_rate": 1.3665683508591948e-05, + "loss": 0.1294, + "step": 56717 + }, + { + "epoch": 2.18, + "learning_rate": 1.3665042318543217e-05, + "loss": 0.2075, + "step": 56718 + }, + { + "epoch": 2.18, + "learning_rate": 1.3664401128494487e-05, + "loss": 0.1508, + "step": 56719 + }, + { + "epoch": 2.18, + "learning_rate": 1.3663759938445755e-05, + "loss": 0.2658, + "step": 56720 + }, + { + "epoch": 2.18, + "learning_rate": 1.3663118748397025e-05, + "loss": 0.149, + "step": 56721 + }, + { + "epoch": 2.18, + "learning_rate": 1.3662477558348293e-05, + "loss": 0.2357, + "step": 56722 + }, + { + "epoch": 2.18, + "learning_rate": 1.3661836368299567e-05, + "loss": 0.2026, + "step": 56723 + }, + { + "epoch": 2.18, + "learning_rate": 1.3661195178250835e-05, + "loss": 0.1987, + "step": 56724 + }, + { + "epoch": 2.18, + "learning_rate": 1.3660553988202105e-05, + "loss": 0.1759, + "step": 56725 + }, + { + "epoch": 2.18, + "learning_rate": 1.3659912798153373e-05, + "loss": 0.1766, + "step": 56726 + }, + { + "epoch": 2.18, + "learning_rate": 1.3659271608104643e-05, + "loss": 0.2546, + "step": 56727 + }, + { + "epoch": 2.18, + "learning_rate": 1.3658630418055912e-05, + "loss": 0.1583, + "step": 56728 + }, + { + "epoch": 2.18, + "learning_rate": 1.3657989228007182e-05, + "loss": 0.1635, + "step": 56729 + }, + { + "epoch": 2.18, + "learning_rate": 1.365734803795845e-05, + "loss": 0.2496, + "step": 56730 + }, + { + "epoch": 2.18, + "learning_rate": 1.365670684790972e-05, + "loss": 0.357, + "step": 56731 + }, + { + "epoch": 2.18, + "learning_rate": 1.3656065657860992e-05, + "loss": 0.1607, + "step": 56732 + }, + { + "epoch": 2.18, + "learning_rate": 1.3655424467812262e-05, + "loss": 0.1668, + "step": 56733 + }, + { + "epoch": 2.18, + "learning_rate": 1.365478327776353e-05, + "loss": 0.3338, + "step": 56734 + }, + { + "epoch": 2.18, + "learning_rate": 1.36541420877148e-05, + "loss": 0.1155, + "step": 56735 + }, + { + "epoch": 2.18, + "learning_rate": 1.3653500897666069e-05, + "loss": 0.1278, + "step": 56736 + }, + { + "epoch": 2.18, + "learning_rate": 1.3652859707617339e-05, + "loss": 0.2835, + "step": 56737 + }, + { + "epoch": 2.18, + "learning_rate": 1.3652218517568607e-05, + "loss": 0.2572, + "step": 56738 + }, + { + "epoch": 2.18, + "learning_rate": 1.3651577327519877e-05, + "loss": 0.1762, + "step": 56739 + }, + { + "epoch": 2.18, + "learning_rate": 1.3650936137471145e-05, + "loss": 0.3423, + "step": 56740 + }, + { + "epoch": 2.18, + "learning_rate": 1.3650294947422419e-05, + "loss": 0.2214, + "step": 56741 + }, + { + "epoch": 2.18, + "learning_rate": 1.3649653757373687e-05, + "loss": 0.17, + "step": 56742 + }, + { + "epoch": 2.18, + "learning_rate": 1.3649012567324957e-05, + "loss": 0.2095, + "step": 56743 + }, + { + "epoch": 2.18, + "learning_rate": 1.3648371377276225e-05, + "loss": 0.2394, + "step": 56744 + }, + { + "epoch": 2.18, + "learning_rate": 1.3647730187227495e-05, + "loss": 0.2302, + "step": 56745 + }, + { + "epoch": 2.18, + "learning_rate": 1.3647088997178764e-05, + "loss": 0.4582, + "step": 56746 + }, + { + "epoch": 2.18, + "learning_rate": 1.3646447807130034e-05, + "loss": 0.1543, + "step": 56747 + }, + { + "epoch": 2.18, + "learning_rate": 1.3645806617081302e-05, + "loss": 0.2043, + "step": 56748 + }, + { + "epoch": 2.18, + "learning_rate": 1.3645165427032572e-05, + "loss": 0.276, + "step": 56749 + }, + { + "epoch": 2.18, + "learning_rate": 1.3644524236983844e-05, + "loss": 0.1753, + "step": 56750 + }, + { + "epoch": 2.18, + "learning_rate": 1.3643883046935114e-05, + "loss": 0.1309, + "step": 56751 + }, + { + "epoch": 2.18, + "learning_rate": 1.3643241856886382e-05, + "loss": 0.2559, + "step": 56752 + }, + { + "epoch": 2.18, + "learning_rate": 1.3642600666837652e-05, + "loss": 0.2486, + "step": 56753 + }, + { + "epoch": 2.18, + "learning_rate": 1.364195947678892e-05, + "loss": 0.2353, + "step": 56754 + }, + { + "epoch": 2.18, + "learning_rate": 1.364131828674019e-05, + "loss": 0.2593, + "step": 56755 + }, + { + "epoch": 2.18, + "learning_rate": 1.3640677096691459e-05, + "loss": 0.2267, + "step": 56756 + }, + { + "epoch": 2.18, + "learning_rate": 1.3640035906642729e-05, + "loss": 0.2343, + "step": 56757 + }, + { + "epoch": 2.18, + "learning_rate": 1.3639394716593997e-05, + "loss": 0.3694, + "step": 56758 + }, + { + "epoch": 2.18, + "learning_rate": 1.363875352654527e-05, + "loss": 0.2855, + "step": 56759 + }, + { + "epoch": 2.18, + "learning_rate": 1.3638112336496539e-05, + "loss": 0.2103, + "step": 56760 + }, + { + "epoch": 2.18, + "learning_rate": 1.3637471146447809e-05, + "loss": 0.1892, + "step": 56761 + }, + { + "epoch": 2.18, + "learning_rate": 1.3636829956399077e-05, + "loss": 0.295, + "step": 56762 + }, + { + "epoch": 2.18, + "learning_rate": 1.3636188766350347e-05, + "loss": 0.2349, + "step": 56763 + }, + { + "epoch": 2.18, + "learning_rate": 1.3635547576301616e-05, + "loss": 0.3329, + "step": 56764 + }, + { + "epoch": 2.18, + "learning_rate": 1.3634906386252886e-05, + "loss": 0.1374, + "step": 56765 + }, + { + "epoch": 2.18, + "learning_rate": 1.3634265196204154e-05, + "loss": 0.2338, + "step": 56766 + }, + { + "epoch": 2.18, + "learning_rate": 1.3633624006155424e-05, + "loss": 0.1696, + "step": 56767 + }, + { + "epoch": 2.18, + "learning_rate": 1.3632982816106696e-05, + "loss": 0.344, + "step": 56768 + }, + { + "epoch": 2.18, + "learning_rate": 1.3632341626057966e-05, + "loss": 0.1611, + "step": 56769 + }, + { + "epoch": 2.18, + "learning_rate": 1.3631700436009234e-05, + "loss": 0.1824, + "step": 56770 + }, + { + "epoch": 2.18, + "learning_rate": 1.3631059245960504e-05, + "loss": 0.2491, + "step": 56771 + }, + { + "epoch": 2.18, + "learning_rate": 1.3630418055911772e-05, + "loss": 0.1798, + "step": 56772 + }, + { + "epoch": 2.18, + "learning_rate": 1.3629776865863042e-05, + "loss": 0.2652, + "step": 56773 + }, + { + "epoch": 2.18, + "learning_rate": 1.362913567581431e-05, + "loss": 0.1721, + "step": 56774 + }, + { + "epoch": 2.18, + "learning_rate": 1.362849448576558e-05, + "loss": 0.1428, + "step": 56775 + }, + { + "epoch": 2.18, + "learning_rate": 1.362785329571685e-05, + "loss": 0.3022, + "step": 56776 + }, + { + "epoch": 2.18, + "learning_rate": 1.3627212105668123e-05, + "loss": 0.1918, + "step": 56777 + }, + { + "epoch": 2.18, + "learning_rate": 1.3626570915619391e-05, + "loss": 0.2862, + "step": 56778 + }, + { + "epoch": 2.18, + "learning_rate": 1.3625929725570661e-05, + "loss": 0.2121, + "step": 56779 + }, + { + "epoch": 2.18, + "learning_rate": 1.362528853552193e-05, + "loss": 0.249, + "step": 56780 + }, + { + "epoch": 2.18, + "learning_rate": 1.36246473454732e-05, + "loss": 0.3576, + "step": 56781 + }, + { + "epoch": 2.18, + "learning_rate": 1.3624006155424468e-05, + "loss": 0.2037, + "step": 56782 + }, + { + "epoch": 2.18, + "learning_rate": 1.3623364965375738e-05, + "loss": 0.2432, + "step": 56783 + }, + { + "epoch": 2.18, + "learning_rate": 1.3622723775327006e-05, + "loss": 0.2021, + "step": 56784 + }, + { + "epoch": 2.18, + "learning_rate": 1.362208258527828e-05, + "loss": 0.1286, + "step": 56785 + }, + { + "epoch": 2.18, + "learning_rate": 1.3621441395229548e-05, + "loss": 0.3626, + "step": 56786 + }, + { + "epoch": 2.18, + "learning_rate": 1.3620800205180818e-05, + "loss": 0.2203, + "step": 56787 + }, + { + "epoch": 2.18, + "learning_rate": 1.3620159015132086e-05, + "loss": 0.2031, + "step": 56788 + }, + { + "epoch": 2.18, + "learning_rate": 1.3619517825083356e-05, + "loss": 0.1959, + "step": 56789 + }, + { + "epoch": 2.18, + "learning_rate": 1.3618876635034624e-05, + "loss": 0.1811, + "step": 56790 + }, + { + "epoch": 2.18, + "learning_rate": 1.3618235444985894e-05, + "loss": 0.2052, + "step": 56791 + }, + { + "epoch": 2.18, + "learning_rate": 1.3617594254937163e-05, + "loss": 0.1756, + "step": 56792 + }, + { + "epoch": 2.18, + "learning_rate": 1.3616953064888433e-05, + "loss": 0.2851, + "step": 56793 + }, + { + "epoch": 2.18, + "learning_rate": 1.3616311874839704e-05, + "loss": 0.2064, + "step": 56794 + }, + { + "epoch": 2.18, + "learning_rate": 1.3615670684790975e-05, + "loss": 0.1902, + "step": 56795 + }, + { + "epoch": 2.18, + "learning_rate": 1.3615029494742243e-05, + "loss": 0.2409, + "step": 56796 + }, + { + "epoch": 2.18, + "learning_rate": 1.3614388304693513e-05, + "loss": 0.2068, + "step": 56797 + }, + { + "epoch": 2.18, + "learning_rate": 1.3613747114644781e-05, + "loss": 0.2016, + "step": 56798 + }, + { + "epoch": 2.18, + "learning_rate": 1.3613105924596051e-05, + "loss": 0.1696, + "step": 56799 + }, + { + "epoch": 2.18, + "learning_rate": 1.361246473454732e-05, + "loss": 0.4526, + "step": 56800 + }, + { + "epoch": 2.18, + "learning_rate": 1.361182354449859e-05, + "loss": 0.1508, + "step": 56801 + }, + { + "epoch": 2.18, + "learning_rate": 1.3611182354449858e-05, + "loss": 0.1197, + "step": 56802 + }, + { + "epoch": 2.18, + "learning_rate": 1.3610541164401131e-05, + "loss": 0.2517, + "step": 56803 + }, + { + "epoch": 2.18, + "learning_rate": 1.36098999743524e-05, + "loss": 0.1396, + "step": 56804 + }, + { + "epoch": 2.18, + "learning_rate": 1.360925878430367e-05, + "loss": 0.1645, + "step": 56805 + }, + { + "epoch": 2.18, + "learning_rate": 1.3608617594254938e-05, + "loss": 0.2468, + "step": 56806 + }, + { + "epoch": 2.18, + "learning_rate": 1.3607976404206208e-05, + "loss": 0.2288, + "step": 56807 + }, + { + "epoch": 2.18, + "learning_rate": 1.3607335214157476e-05, + "loss": 0.1895, + "step": 56808 + }, + { + "epoch": 2.18, + "learning_rate": 1.3606694024108746e-05, + "loss": 0.2445, + "step": 56809 + }, + { + "epoch": 2.18, + "learning_rate": 1.3606052834060015e-05, + "loss": 0.2929, + "step": 56810 + }, + { + "epoch": 2.18, + "learning_rate": 1.3605411644011285e-05, + "loss": 0.2414, + "step": 56811 + }, + { + "epoch": 2.18, + "learning_rate": 1.3604770453962556e-05, + "loss": 0.2002, + "step": 56812 + }, + { + "epoch": 2.18, + "learning_rate": 1.3604129263913826e-05, + "loss": 0.2391, + "step": 56813 + }, + { + "epoch": 2.18, + "learning_rate": 1.3603488073865095e-05, + "loss": 0.3455, + "step": 56814 + }, + { + "epoch": 2.18, + "learning_rate": 1.3602846883816365e-05, + "loss": 0.1488, + "step": 56815 + }, + { + "epoch": 2.18, + "learning_rate": 1.3602205693767633e-05, + "loss": 0.3617, + "step": 56816 + }, + { + "epoch": 2.18, + "learning_rate": 1.3601564503718903e-05, + "loss": 0.163, + "step": 56817 + }, + { + "epoch": 2.18, + "learning_rate": 1.3600923313670171e-05, + "loss": 0.2679, + "step": 56818 + }, + { + "epoch": 2.18, + "learning_rate": 1.3600282123621441e-05, + "loss": 0.1955, + "step": 56819 + }, + { + "epoch": 2.18, + "learning_rate": 1.359964093357271e-05, + "loss": 0.2635, + "step": 56820 + }, + { + "epoch": 2.18, + "learning_rate": 1.3598999743523983e-05, + "loss": 0.445, + "step": 56821 + }, + { + "epoch": 2.18, + "learning_rate": 1.3598358553475252e-05, + "loss": 0.0969, + "step": 56822 + }, + { + "epoch": 2.18, + "learning_rate": 1.3597717363426522e-05, + "loss": 0.2251, + "step": 56823 + }, + { + "epoch": 2.18, + "learning_rate": 1.359707617337779e-05, + "loss": 0.2839, + "step": 56824 + }, + { + "epoch": 2.18, + "learning_rate": 1.359643498332906e-05, + "loss": 0.5055, + "step": 56825 + }, + { + "epoch": 2.18, + "learning_rate": 1.3595793793280328e-05, + "loss": 0.2714, + "step": 56826 + }, + { + "epoch": 2.18, + "learning_rate": 1.3595152603231598e-05, + "loss": 0.2399, + "step": 56827 + }, + { + "epoch": 2.18, + "learning_rate": 1.3594511413182867e-05, + "loss": 0.1627, + "step": 56828 + }, + { + "epoch": 2.18, + "learning_rate": 1.3593870223134137e-05, + "loss": 0.1971, + "step": 56829 + }, + { + "epoch": 2.18, + "learning_rate": 1.3593229033085408e-05, + "loss": 0.2235, + "step": 56830 + }, + { + "epoch": 2.18, + "learning_rate": 1.3592587843036678e-05, + "loss": 0.2017, + "step": 56831 + }, + { + "epoch": 2.19, + "learning_rate": 1.3591946652987947e-05, + "loss": 0.3196, + "step": 56832 + }, + { + "epoch": 2.19, + "learning_rate": 1.3591305462939217e-05, + "loss": 0.2972, + "step": 56833 + }, + { + "epoch": 2.19, + "learning_rate": 1.3590664272890485e-05, + "loss": 0.2274, + "step": 56834 + }, + { + "epoch": 2.19, + "learning_rate": 1.3590023082841755e-05, + "loss": 0.2636, + "step": 56835 + }, + { + "epoch": 2.19, + "learning_rate": 1.3589381892793023e-05, + "loss": 0.2459, + "step": 56836 + }, + { + "epoch": 2.19, + "learning_rate": 1.3588740702744293e-05, + "loss": 0.2679, + "step": 56837 + }, + { + "epoch": 2.19, + "learning_rate": 1.3588099512695562e-05, + "loss": 0.1712, + "step": 56838 + }, + { + "epoch": 2.19, + "learning_rate": 1.3587458322646835e-05, + "loss": 0.2782, + "step": 56839 + }, + { + "epoch": 2.19, + "learning_rate": 1.3586817132598103e-05, + "loss": 0.1031, + "step": 56840 + }, + { + "epoch": 2.19, + "learning_rate": 1.3586175942549374e-05, + "loss": 0.2283, + "step": 56841 + }, + { + "epoch": 2.19, + "learning_rate": 1.3585534752500642e-05, + "loss": 0.2245, + "step": 56842 + }, + { + "epoch": 2.19, + "learning_rate": 1.3584893562451912e-05, + "loss": 0.3384, + "step": 56843 + }, + { + "epoch": 2.19, + "learning_rate": 1.358425237240318e-05, + "loss": 0.1977, + "step": 56844 + }, + { + "epoch": 2.19, + "learning_rate": 1.358361118235445e-05, + "loss": 0.2966, + "step": 56845 + }, + { + "epoch": 2.19, + "learning_rate": 1.3582969992305719e-05, + "loss": 0.2218, + "step": 56846 + }, + { + "epoch": 2.19, + "learning_rate": 1.3582328802256989e-05, + "loss": 0.3485, + "step": 56847 + }, + { + "epoch": 2.19, + "learning_rate": 1.358168761220826e-05, + "loss": 0.2315, + "step": 56848 + }, + { + "epoch": 2.19, + "learning_rate": 1.358104642215953e-05, + "loss": 0.2394, + "step": 56849 + }, + { + "epoch": 2.19, + "learning_rate": 1.3580405232110799e-05, + "loss": 0.3368, + "step": 56850 + }, + { + "epoch": 2.19, + "learning_rate": 1.3579764042062069e-05, + "loss": 0.2132, + "step": 56851 + }, + { + "epoch": 2.19, + "learning_rate": 1.3579122852013337e-05, + "loss": 0.3326, + "step": 56852 + }, + { + "epoch": 2.19, + "learning_rate": 1.3578481661964607e-05, + "loss": 0.2859, + "step": 56853 + }, + { + "epoch": 2.19, + "learning_rate": 1.3577840471915875e-05, + "loss": 0.1283, + "step": 56854 + }, + { + "epoch": 2.19, + "learning_rate": 1.3577199281867145e-05, + "loss": 0.2393, + "step": 56855 + }, + { + "epoch": 2.19, + "learning_rate": 1.3576558091818414e-05, + "loss": 0.2165, + "step": 56856 + }, + { + "epoch": 2.19, + "learning_rate": 1.3575916901769687e-05, + "loss": 0.1991, + "step": 56857 + }, + { + "epoch": 2.19, + "learning_rate": 1.3575275711720955e-05, + "loss": 0.2283, + "step": 56858 + }, + { + "epoch": 2.19, + "learning_rate": 1.3574634521672225e-05, + "loss": 0.2418, + "step": 56859 + }, + { + "epoch": 2.19, + "learning_rate": 1.3573993331623494e-05, + "loss": 0.2929, + "step": 56860 + }, + { + "epoch": 2.19, + "learning_rate": 1.3573352141574764e-05, + "loss": 0.2467, + "step": 56861 + }, + { + "epoch": 2.19, + "learning_rate": 1.3572710951526032e-05, + "loss": 0.1535, + "step": 56862 + }, + { + "epoch": 2.19, + "learning_rate": 1.3572069761477302e-05, + "loss": 0.2515, + "step": 56863 + }, + { + "epoch": 2.19, + "learning_rate": 1.357142857142857e-05, + "loss": 0.3059, + "step": 56864 + }, + { + "epoch": 2.19, + "learning_rate": 1.357078738137984e-05, + "loss": 0.2522, + "step": 56865 + }, + { + "epoch": 2.19, + "learning_rate": 1.3570146191331112e-05, + "loss": 0.2194, + "step": 56866 + }, + { + "epoch": 2.19, + "learning_rate": 1.3569505001282382e-05, + "loss": 0.2209, + "step": 56867 + }, + { + "epoch": 2.19, + "learning_rate": 1.356886381123365e-05, + "loss": 0.2634, + "step": 56868 + }, + { + "epoch": 2.19, + "learning_rate": 1.356822262118492e-05, + "loss": 0.2182, + "step": 56869 + }, + { + "epoch": 2.19, + "learning_rate": 1.3567581431136189e-05, + "loss": 0.2454, + "step": 56870 + }, + { + "epoch": 2.19, + "learning_rate": 1.3566940241087459e-05, + "loss": 0.217, + "step": 56871 + }, + { + "epoch": 2.19, + "learning_rate": 1.3566299051038727e-05, + "loss": 0.1868, + "step": 56872 + }, + { + "epoch": 2.19, + "learning_rate": 1.3565657860989997e-05, + "loss": 0.2147, + "step": 56873 + }, + { + "epoch": 2.19, + "learning_rate": 1.3565016670941266e-05, + "loss": 0.3403, + "step": 56874 + }, + { + "epoch": 2.19, + "learning_rate": 1.3564375480892539e-05, + "loss": 0.2597, + "step": 56875 + }, + { + "epoch": 2.19, + "learning_rate": 1.3563734290843807e-05, + "loss": 0.309, + "step": 56876 + }, + { + "epoch": 2.19, + "learning_rate": 1.3563093100795077e-05, + "loss": 0.2232, + "step": 56877 + }, + { + "epoch": 2.19, + "learning_rate": 1.3562451910746346e-05, + "loss": 0.2778, + "step": 56878 + }, + { + "epoch": 2.19, + "learning_rate": 1.3561810720697616e-05, + "loss": 0.2168, + "step": 56879 + }, + { + "epoch": 2.19, + "learning_rate": 1.3561169530648884e-05, + "loss": 0.2129, + "step": 56880 + }, + { + "epoch": 2.19, + "learning_rate": 1.3560528340600154e-05, + "loss": 0.1936, + "step": 56881 + }, + { + "epoch": 2.19, + "learning_rate": 1.3559887150551422e-05, + "loss": 0.2254, + "step": 56882 + }, + { + "epoch": 2.19, + "learning_rate": 1.3559245960502696e-05, + "loss": 0.2167, + "step": 56883 + }, + { + "epoch": 2.19, + "learning_rate": 1.3558604770453964e-05, + "loss": 0.1859, + "step": 56884 + }, + { + "epoch": 2.19, + "learning_rate": 1.3557963580405234e-05, + "loss": 0.2371, + "step": 56885 + }, + { + "epoch": 2.19, + "learning_rate": 1.3557322390356503e-05, + "loss": 0.2235, + "step": 56886 + }, + { + "epoch": 2.19, + "learning_rate": 1.3556681200307773e-05, + "loss": 0.2409, + "step": 56887 + }, + { + "epoch": 2.19, + "learning_rate": 1.3556040010259041e-05, + "loss": 0.1513, + "step": 56888 + }, + { + "epoch": 2.19, + "learning_rate": 1.3555398820210311e-05, + "loss": 0.1027, + "step": 56889 + }, + { + "epoch": 2.19, + "learning_rate": 1.355475763016158e-05, + "loss": 0.3051, + "step": 56890 + }, + { + "epoch": 2.19, + "learning_rate": 1.355411644011285e-05, + "loss": 0.2692, + "step": 56891 + }, + { + "epoch": 2.19, + "learning_rate": 1.3553475250064121e-05, + "loss": 0.1965, + "step": 56892 + }, + { + "epoch": 2.19, + "learning_rate": 1.3552834060015391e-05, + "loss": 0.2119, + "step": 56893 + }, + { + "epoch": 2.19, + "learning_rate": 1.355219286996666e-05, + "loss": 0.2187, + "step": 56894 + }, + { + "epoch": 2.19, + "learning_rate": 1.355155167991793e-05, + "loss": 0.2149, + "step": 56895 + }, + { + "epoch": 2.19, + "learning_rate": 1.3550910489869198e-05, + "loss": 0.1837, + "step": 56896 + }, + { + "epoch": 2.19, + "learning_rate": 1.3550269299820468e-05, + "loss": 0.2155, + "step": 56897 + }, + { + "epoch": 2.19, + "learning_rate": 1.3549628109771736e-05, + "loss": 0.2704, + "step": 56898 + }, + { + "epoch": 2.19, + "learning_rate": 1.3548986919723006e-05, + "loss": 0.2199, + "step": 56899 + }, + { + "epoch": 2.19, + "learning_rate": 1.3548345729674274e-05, + "loss": 0.1494, + "step": 56900 + }, + { + "epoch": 2.19, + "learning_rate": 1.3547704539625548e-05, + "loss": 0.3642, + "step": 56901 + }, + { + "epoch": 2.19, + "learning_rate": 1.3547063349576816e-05, + "loss": 0.2297, + "step": 56902 + }, + { + "epoch": 2.19, + "learning_rate": 1.3546422159528086e-05, + "loss": 0.1185, + "step": 56903 + }, + { + "epoch": 2.19, + "learning_rate": 1.3545780969479354e-05, + "loss": 0.3996, + "step": 56904 + }, + { + "epoch": 2.19, + "learning_rate": 1.3545139779430624e-05, + "loss": 0.2985, + "step": 56905 + }, + { + "epoch": 2.19, + "learning_rate": 1.3544498589381893e-05, + "loss": 0.3607, + "step": 56906 + }, + { + "epoch": 2.19, + "learning_rate": 1.3543857399333163e-05, + "loss": 0.2717, + "step": 56907 + }, + { + "epoch": 2.19, + "learning_rate": 1.3543216209284431e-05, + "loss": 0.297, + "step": 56908 + }, + { + "epoch": 2.19, + "learning_rate": 1.3542575019235701e-05, + "loss": 0.3066, + "step": 56909 + }, + { + "epoch": 2.19, + "learning_rate": 1.3541933829186973e-05, + "loss": 0.3164, + "step": 56910 + }, + { + "epoch": 2.19, + "learning_rate": 1.3541292639138243e-05, + "loss": 0.3061, + "step": 56911 + }, + { + "epoch": 2.19, + "learning_rate": 1.3540651449089511e-05, + "loss": 0.1721, + "step": 56912 + }, + { + "epoch": 2.19, + "learning_rate": 1.3540010259040781e-05, + "loss": 0.204, + "step": 56913 + }, + { + "epoch": 2.19, + "learning_rate": 1.353936906899205e-05, + "loss": 0.1559, + "step": 56914 + }, + { + "epoch": 2.19, + "learning_rate": 1.353872787894332e-05, + "loss": 0.2531, + "step": 56915 + }, + { + "epoch": 2.19, + "learning_rate": 1.3538086688894588e-05, + "loss": 0.4417, + "step": 56916 + }, + { + "epoch": 2.19, + "learning_rate": 1.3537445498845858e-05, + "loss": 0.2385, + "step": 56917 + }, + { + "epoch": 2.19, + "learning_rate": 1.3536804308797126e-05, + "loss": 0.2414, + "step": 56918 + }, + { + "epoch": 2.19, + "learning_rate": 1.35361631187484e-05, + "loss": 0.2366, + "step": 56919 + }, + { + "epoch": 2.19, + "learning_rate": 1.3535521928699668e-05, + "loss": 0.2114, + "step": 56920 + }, + { + "epoch": 2.19, + "learning_rate": 1.3534880738650938e-05, + "loss": 0.2262, + "step": 56921 + }, + { + "epoch": 2.19, + "learning_rate": 1.3534239548602206e-05, + "loss": 0.1184, + "step": 56922 + }, + { + "epoch": 2.19, + "learning_rate": 1.3533598358553476e-05, + "loss": 0.1896, + "step": 56923 + }, + { + "epoch": 2.19, + "learning_rate": 1.3532957168504745e-05, + "loss": 0.1109, + "step": 56924 + }, + { + "epoch": 2.19, + "learning_rate": 1.3532315978456015e-05, + "loss": 0.2938, + "step": 56925 + }, + { + "epoch": 2.19, + "learning_rate": 1.3531674788407283e-05, + "loss": 0.1507, + "step": 56926 + }, + { + "epoch": 2.19, + "learning_rate": 1.3531033598358553e-05, + "loss": 0.3225, + "step": 56927 + }, + { + "epoch": 2.19, + "learning_rate": 1.3530392408309825e-05, + "loss": 0.1692, + "step": 56928 + }, + { + "epoch": 2.19, + "learning_rate": 1.3529751218261095e-05, + "loss": 0.2277, + "step": 56929 + }, + { + "epoch": 2.19, + "learning_rate": 1.3529110028212363e-05, + "loss": 0.234, + "step": 56930 + }, + { + "epoch": 2.19, + "learning_rate": 1.3528468838163633e-05, + "loss": 0.2089, + "step": 56931 + }, + { + "epoch": 2.19, + "learning_rate": 1.3527827648114902e-05, + "loss": 0.1876, + "step": 56932 + }, + { + "epoch": 2.19, + "learning_rate": 1.3527186458066172e-05, + "loss": 0.2107, + "step": 56933 + }, + { + "epoch": 2.19, + "learning_rate": 1.352654526801744e-05, + "loss": 0.1692, + "step": 56934 + }, + { + "epoch": 2.19, + "learning_rate": 1.352590407796871e-05, + "loss": 0.2331, + "step": 56935 + }, + { + "epoch": 2.19, + "learning_rate": 1.3525262887919978e-05, + "loss": 0.1637, + "step": 56936 + }, + { + "epoch": 2.19, + "learning_rate": 1.3524621697871252e-05, + "loss": 0.1625, + "step": 56937 + }, + { + "epoch": 2.19, + "learning_rate": 1.352398050782252e-05, + "loss": 0.2359, + "step": 56938 + }, + { + "epoch": 2.19, + "learning_rate": 1.352333931777379e-05, + "loss": 0.2584, + "step": 56939 + }, + { + "epoch": 2.19, + "learning_rate": 1.3522698127725058e-05, + "loss": 0.1009, + "step": 56940 + }, + { + "epoch": 2.19, + "learning_rate": 1.3522056937676328e-05, + "loss": 0.1608, + "step": 56941 + }, + { + "epoch": 2.19, + "learning_rate": 1.3521415747627597e-05, + "loss": 0.1782, + "step": 56942 + }, + { + "epoch": 2.19, + "learning_rate": 1.3520774557578867e-05, + "loss": 0.1894, + "step": 56943 + }, + { + "epoch": 2.19, + "learning_rate": 1.3520133367530135e-05, + "loss": 0.1932, + "step": 56944 + }, + { + "epoch": 2.19, + "learning_rate": 1.3519492177481405e-05, + "loss": 0.2928, + "step": 56945 + }, + { + "epoch": 2.19, + "learning_rate": 1.3518850987432677e-05, + "loss": 0.2979, + "step": 56946 + }, + { + "epoch": 2.19, + "learning_rate": 1.3518209797383947e-05, + "loss": 0.4138, + "step": 56947 + }, + { + "epoch": 2.19, + "learning_rate": 1.3517568607335215e-05, + "loss": 0.316, + "step": 56948 + }, + { + "epoch": 2.19, + "learning_rate": 1.3516927417286485e-05, + "loss": 0.33, + "step": 56949 + }, + { + "epoch": 2.19, + "learning_rate": 1.3516286227237753e-05, + "loss": 0.2033, + "step": 56950 + }, + { + "epoch": 2.19, + "learning_rate": 1.3515645037189023e-05, + "loss": 0.2435, + "step": 56951 + }, + { + "epoch": 2.19, + "learning_rate": 1.3515003847140292e-05, + "loss": 0.3312, + "step": 56952 + }, + { + "epoch": 2.19, + "learning_rate": 1.3514362657091562e-05, + "loss": 0.17, + "step": 56953 + }, + { + "epoch": 2.19, + "learning_rate": 1.351372146704283e-05, + "loss": 0.2696, + "step": 56954 + }, + { + "epoch": 2.19, + "learning_rate": 1.3513080276994104e-05, + "loss": 0.2507, + "step": 56955 + }, + { + "epoch": 2.19, + "learning_rate": 1.3512439086945372e-05, + "loss": 0.2518, + "step": 56956 + }, + { + "epoch": 2.19, + "learning_rate": 1.3511797896896642e-05, + "loss": 0.3119, + "step": 56957 + }, + { + "epoch": 2.19, + "learning_rate": 1.351115670684791e-05, + "loss": 0.1776, + "step": 56958 + }, + { + "epoch": 2.19, + "learning_rate": 1.351051551679918e-05, + "loss": 0.2117, + "step": 56959 + }, + { + "epoch": 2.19, + "learning_rate": 1.3509874326750449e-05, + "loss": 0.2834, + "step": 56960 + }, + { + "epoch": 2.19, + "learning_rate": 1.3509233136701719e-05, + "loss": 0.1979, + "step": 56961 + }, + { + "epoch": 2.19, + "learning_rate": 1.3508591946652987e-05, + "loss": 0.323, + "step": 56962 + }, + { + "epoch": 2.19, + "learning_rate": 1.3507950756604257e-05, + "loss": 0.2626, + "step": 56963 + }, + { + "epoch": 2.19, + "learning_rate": 1.3507309566555529e-05, + "loss": 0.1529, + "step": 56964 + }, + { + "epoch": 2.19, + "learning_rate": 1.3506668376506799e-05, + "loss": 0.2776, + "step": 56965 + }, + { + "epoch": 2.19, + "learning_rate": 1.3506027186458067e-05, + "loss": 0.2803, + "step": 56966 + }, + { + "epoch": 2.19, + "learning_rate": 1.3505385996409337e-05, + "loss": 0.2008, + "step": 56967 + }, + { + "epoch": 2.19, + "learning_rate": 1.3504744806360605e-05, + "loss": 0.179, + "step": 56968 + }, + { + "epoch": 2.19, + "learning_rate": 1.3504103616311875e-05, + "loss": 0.2137, + "step": 56969 + }, + { + "epoch": 2.19, + "learning_rate": 1.3503462426263144e-05, + "loss": 0.2349, + "step": 56970 + }, + { + "epoch": 2.19, + "learning_rate": 1.3502821236214414e-05, + "loss": 0.1912, + "step": 56971 + }, + { + "epoch": 2.19, + "learning_rate": 1.3502180046165682e-05, + "loss": 0.1815, + "step": 56972 + }, + { + "epoch": 2.19, + "learning_rate": 1.3501538856116956e-05, + "loss": 0.1916, + "step": 56973 + }, + { + "epoch": 2.19, + "learning_rate": 1.3500897666068224e-05, + "loss": 0.251, + "step": 56974 + }, + { + "epoch": 2.19, + "learning_rate": 1.3500256476019494e-05, + "loss": 0.2547, + "step": 56975 + }, + { + "epoch": 2.19, + "learning_rate": 1.3499615285970762e-05, + "loss": 0.2197, + "step": 56976 + }, + { + "epoch": 2.19, + "learning_rate": 1.3498974095922032e-05, + "loss": 0.2405, + "step": 56977 + }, + { + "epoch": 2.19, + "learning_rate": 1.34983329058733e-05, + "loss": 0.2824, + "step": 56978 + }, + { + "epoch": 2.19, + "learning_rate": 1.349769171582457e-05, + "loss": 0.3037, + "step": 56979 + }, + { + "epoch": 2.19, + "learning_rate": 1.3497050525775839e-05, + "loss": 0.3044, + "step": 56980 + }, + { + "epoch": 2.19, + "learning_rate": 1.3496409335727112e-05, + "loss": 0.1594, + "step": 56981 + }, + { + "epoch": 2.19, + "learning_rate": 1.349576814567838e-05, + "loss": 0.1354, + "step": 56982 + }, + { + "epoch": 2.19, + "learning_rate": 1.349512695562965e-05, + "loss": 0.1264, + "step": 56983 + }, + { + "epoch": 2.19, + "learning_rate": 1.3494485765580919e-05, + "loss": 0.2799, + "step": 56984 + }, + { + "epoch": 2.19, + "learning_rate": 1.3493844575532189e-05, + "loss": 0.2609, + "step": 56985 + }, + { + "epoch": 2.19, + "learning_rate": 1.3493203385483457e-05, + "loss": 0.2694, + "step": 56986 + }, + { + "epoch": 2.19, + "learning_rate": 1.3492562195434727e-05, + "loss": 0.4248, + "step": 56987 + }, + { + "epoch": 2.19, + "learning_rate": 1.3491921005385996e-05, + "loss": 0.1728, + "step": 56988 + }, + { + "epoch": 2.19, + "learning_rate": 1.3491279815337266e-05, + "loss": 0.3128, + "step": 56989 + }, + { + "epoch": 2.19, + "learning_rate": 1.3490638625288537e-05, + "loss": 0.2689, + "step": 56990 + }, + { + "epoch": 2.19, + "learning_rate": 1.3489997435239807e-05, + "loss": 0.2403, + "step": 56991 + }, + { + "epoch": 2.19, + "learning_rate": 1.3489356245191076e-05, + "loss": 0.1749, + "step": 56992 + }, + { + "epoch": 2.19, + "learning_rate": 1.3488715055142346e-05, + "loss": 0.3339, + "step": 56993 + }, + { + "epoch": 2.19, + "learning_rate": 1.3488073865093614e-05, + "loss": 0.1509, + "step": 56994 + }, + { + "epoch": 2.19, + "learning_rate": 1.3487432675044884e-05, + "loss": 0.3023, + "step": 56995 + }, + { + "epoch": 2.19, + "learning_rate": 1.3486791484996152e-05, + "loss": 0.1909, + "step": 56996 + }, + { + "epoch": 2.19, + "learning_rate": 1.3486150294947422e-05, + "loss": 0.1999, + "step": 56997 + }, + { + "epoch": 2.19, + "learning_rate": 1.348550910489869e-05, + "loss": 0.2049, + "step": 56998 + }, + { + "epoch": 2.19, + "learning_rate": 1.3484867914849964e-05, + "loss": 0.1795, + "step": 56999 + }, + { + "epoch": 2.19, + "learning_rate": 1.3484226724801233e-05, + "loss": 0.338, + "step": 57000 + }, + { + "epoch": 2.19, + "learning_rate": 1.3483585534752503e-05, + "loss": 0.2858, + "step": 57001 + }, + { + "epoch": 2.19, + "learning_rate": 1.3482944344703771e-05, + "loss": 0.2448, + "step": 57002 + }, + { + "epoch": 2.19, + "learning_rate": 1.3482303154655041e-05, + "loss": 0.2414, + "step": 57003 + }, + { + "epoch": 2.19, + "learning_rate": 1.348166196460631e-05, + "loss": 0.1603, + "step": 57004 + }, + { + "epoch": 2.19, + "learning_rate": 1.348102077455758e-05, + "loss": 0.2778, + "step": 57005 + }, + { + "epoch": 2.19, + "learning_rate": 1.3480379584508848e-05, + "loss": 0.2824, + "step": 57006 + }, + { + "epoch": 2.19, + "learning_rate": 1.3479738394460118e-05, + "loss": 0.1699, + "step": 57007 + }, + { + "epoch": 2.19, + "learning_rate": 1.347909720441139e-05, + "loss": 0.1578, + "step": 57008 + }, + { + "epoch": 2.19, + "learning_rate": 1.347845601436266e-05, + "loss": 0.2232, + "step": 57009 + }, + { + "epoch": 2.19, + "learning_rate": 1.3477814824313928e-05, + "loss": 0.2584, + "step": 57010 + }, + { + "epoch": 2.19, + "learning_rate": 1.3477173634265198e-05, + "loss": 0.1461, + "step": 57011 + }, + { + "epoch": 2.19, + "learning_rate": 1.3476532444216466e-05, + "loss": 0.1994, + "step": 57012 + }, + { + "epoch": 2.19, + "learning_rate": 1.3475891254167736e-05, + "loss": 0.1388, + "step": 57013 + }, + { + "epoch": 2.19, + "learning_rate": 1.3475250064119004e-05, + "loss": 0.151, + "step": 57014 + }, + { + "epoch": 2.19, + "learning_rate": 1.3474608874070274e-05, + "loss": 0.2046, + "step": 57015 + }, + { + "epoch": 2.19, + "learning_rate": 1.3473967684021543e-05, + "loss": 0.2826, + "step": 57016 + }, + { + "epoch": 2.19, + "learning_rate": 1.3473326493972816e-05, + "loss": 0.2162, + "step": 57017 + }, + { + "epoch": 2.19, + "learning_rate": 1.3472685303924084e-05, + "loss": 0.282, + "step": 57018 + }, + { + "epoch": 2.19, + "learning_rate": 1.3472044113875355e-05, + "loss": 0.2593, + "step": 57019 + }, + { + "epoch": 2.19, + "learning_rate": 1.3471402923826623e-05, + "loss": 0.2254, + "step": 57020 + }, + { + "epoch": 2.19, + "learning_rate": 1.3470761733777893e-05, + "loss": 0.2457, + "step": 57021 + }, + { + "epoch": 2.19, + "learning_rate": 1.3470120543729161e-05, + "loss": 0.3204, + "step": 57022 + }, + { + "epoch": 2.19, + "learning_rate": 1.3469479353680431e-05, + "loss": 0.1941, + "step": 57023 + }, + { + "epoch": 2.19, + "learning_rate": 1.34688381636317e-05, + "loss": 0.2131, + "step": 57024 + }, + { + "epoch": 2.19, + "learning_rate": 1.346819697358297e-05, + "loss": 0.2448, + "step": 57025 + }, + { + "epoch": 2.19, + "learning_rate": 1.3467555783534241e-05, + "loss": 0.2119, + "step": 57026 + }, + { + "epoch": 2.19, + "learning_rate": 1.3466914593485511e-05, + "loss": 0.222, + "step": 57027 + }, + { + "epoch": 2.19, + "learning_rate": 1.346627340343678e-05, + "loss": 0.2194, + "step": 57028 + }, + { + "epoch": 2.19, + "learning_rate": 1.346563221338805e-05, + "loss": 0.2421, + "step": 57029 + }, + { + "epoch": 2.19, + "learning_rate": 1.3464991023339318e-05, + "loss": 0.1751, + "step": 57030 + }, + { + "epoch": 2.19, + "learning_rate": 1.3464349833290588e-05, + "loss": 0.189, + "step": 57031 + }, + { + "epoch": 2.19, + "learning_rate": 1.3463708643241856e-05, + "loss": 0.2105, + "step": 57032 + }, + { + "epoch": 2.19, + "learning_rate": 1.3463067453193126e-05, + "loss": 0.2341, + "step": 57033 + }, + { + "epoch": 2.19, + "learning_rate": 1.3462426263144395e-05, + "loss": 0.3734, + "step": 57034 + }, + { + "epoch": 2.19, + "learning_rate": 1.3461785073095668e-05, + "loss": 0.2816, + "step": 57035 + }, + { + "epoch": 2.19, + "learning_rate": 1.3461143883046936e-05, + "loss": 0.1409, + "step": 57036 + }, + { + "epoch": 2.19, + "learning_rate": 1.3460502692998206e-05, + "loss": 0.1971, + "step": 57037 + }, + { + "epoch": 2.19, + "learning_rate": 1.3459861502949475e-05, + "loss": 0.2326, + "step": 57038 + }, + { + "epoch": 2.19, + "learning_rate": 1.3459220312900745e-05, + "loss": 0.3757, + "step": 57039 + }, + { + "epoch": 2.19, + "learning_rate": 1.3458579122852013e-05, + "loss": 0.2647, + "step": 57040 + }, + { + "epoch": 2.19, + "learning_rate": 1.3457937932803283e-05, + "loss": 0.2315, + "step": 57041 + }, + { + "epoch": 2.19, + "learning_rate": 1.3457296742754551e-05, + "loss": 0.228, + "step": 57042 + }, + { + "epoch": 2.19, + "learning_rate": 1.3456655552705822e-05, + "loss": 0.2337, + "step": 57043 + }, + { + "epoch": 2.19, + "learning_rate": 1.3456014362657093e-05, + "loss": 0.2018, + "step": 57044 + }, + { + "epoch": 2.19, + "learning_rate": 1.3455373172608363e-05, + "loss": 0.1642, + "step": 57045 + }, + { + "epoch": 2.19, + "learning_rate": 1.3454731982559632e-05, + "loss": 0.3316, + "step": 57046 + }, + { + "epoch": 2.19, + "learning_rate": 1.3454090792510902e-05, + "loss": 0.1173, + "step": 57047 + }, + { + "epoch": 2.19, + "learning_rate": 1.345344960246217e-05, + "loss": 0.2248, + "step": 57048 + }, + { + "epoch": 2.19, + "learning_rate": 1.345280841241344e-05, + "loss": 0.1968, + "step": 57049 + }, + { + "epoch": 2.19, + "learning_rate": 1.3452167222364708e-05, + "loss": 0.2148, + "step": 57050 + }, + { + "epoch": 2.19, + "learning_rate": 1.3451526032315978e-05, + "loss": 0.2161, + "step": 57051 + }, + { + "epoch": 2.19, + "learning_rate": 1.3450884842267247e-05, + "loss": 0.2057, + "step": 57052 + }, + { + "epoch": 2.19, + "learning_rate": 1.345024365221852e-05, + "loss": 0.2467, + "step": 57053 + }, + { + "epoch": 2.19, + "learning_rate": 1.3449602462169788e-05, + "loss": 0.2772, + "step": 57054 + }, + { + "epoch": 2.19, + "learning_rate": 1.3448961272121058e-05, + "loss": 0.1691, + "step": 57055 + }, + { + "epoch": 2.19, + "learning_rate": 1.3448320082072327e-05, + "loss": 0.4142, + "step": 57056 + }, + { + "epoch": 2.19, + "learning_rate": 1.3447678892023597e-05, + "loss": 0.2322, + "step": 57057 + }, + { + "epoch": 2.19, + "learning_rate": 1.3447037701974865e-05, + "loss": 0.3144, + "step": 57058 + }, + { + "epoch": 2.19, + "learning_rate": 1.3446396511926135e-05, + "loss": 0.2163, + "step": 57059 + }, + { + "epoch": 2.19, + "learning_rate": 1.3445755321877403e-05, + "loss": 0.1958, + "step": 57060 + }, + { + "epoch": 2.19, + "learning_rate": 1.3445114131828673e-05, + "loss": 0.3223, + "step": 57061 + }, + { + "epoch": 2.19, + "learning_rate": 1.3444472941779945e-05, + "loss": 0.2035, + "step": 57062 + }, + { + "epoch": 2.19, + "learning_rate": 1.3443831751731215e-05, + "loss": 0.2696, + "step": 57063 + }, + { + "epoch": 2.19, + "learning_rate": 1.3443190561682484e-05, + "loss": 0.2595, + "step": 57064 + }, + { + "epoch": 2.19, + "learning_rate": 1.3442549371633754e-05, + "loss": 0.2182, + "step": 57065 + }, + { + "epoch": 2.19, + "learning_rate": 1.3441908181585022e-05, + "loss": 0.2462, + "step": 57066 + }, + { + "epoch": 2.19, + "learning_rate": 1.3441266991536292e-05, + "loss": 0.2566, + "step": 57067 + }, + { + "epoch": 2.19, + "learning_rate": 1.344062580148756e-05, + "loss": 0.262, + "step": 57068 + }, + { + "epoch": 2.19, + "learning_rate": 1.343998461143883e-05, + "loss": 0.2887, + "step": 57069 + }, + { + "epoch": 2.19, + "learning_rate": 1.3439343421390102e-05, + "loss": 0.2657, + "step": 57070 + }, + { + "epoch": 2.19, + "learning_rate": 1.3438702231341372e-05, + "loss": 0.334, + "step": 57071 + }, + { + "epoch": 2.19, + "learning_rate": 1.343806104129264e-05, + "loss": 0.23, + "step": 57072 + }, + { + "epoch": 2.19, + "learning_rate": 1.343741985124391e-05, + "loss": 0.2928, + "step": 57073 + }, + { + "epoch": 2.19, + "learning_rate": 1.3436778661195179e-05, + "loss": 0.2161, + "step": 57074 + }, + { + "epoch": 2.19, + "learning_rate": 1.3436137471146449e-05, + "loss": 0.2069, + "step": 57075 + }, + { + "epoch": 2.19, + "learning_rate": 1.3435496281097717e-05, + "loss": 0.1972, + "step": 57076 + }, + { + "epoch": 2.19, + "learning_rate": 1.3434855091048987e-05, + "loss": 0.2344, + "step": 57077 + }, + { + "epoch": 2.19, + "learning_rate": 1.3434213901000255e-05, + "loss": 0.3802, + "step": 57078 + }, + { + "epoch": 2.19, + "learning_rate": 1.3433572710951529e-05, + "loss": 0.2919, + "step": 57079 + }, + { + "epoch": 2.19, + "learning_rate": 1.3432931520902797e-05, + "loss": 0.2897, + "step": 57080 + }, + { + "epoch": 2.19, + "learning_rate": 1.3432290330854067e-05, + "loss": 0.2451, + "step": 57081 + }, + { + "epoch": 2.19, + "learning_rate": 1.3431649140805335e-05, + "loss": 0.1394, + "step": 57082 + }, + { + "epoch": 2.19, + "learning_rate": 1.3431007950756605e-05, + "loss": 0.1603, + "step": 57083 + }, + { + "epoch": 2.19, + "learning_rate": 1.3430366760707874e-05, + "loss": 0.2836, + "step": 57084 + }, + { + "epoch": 2.19, + "learning_rate": 1.3429725570659144e-05, + "loss": 0.1587, + "step": 57085 + }, + { + "epoch": 2.19, + "learning_rate": 1.3429084380610412e-05, + "loss": 0.2499, + "step": 57086 + }, + { + "epoch": 2.19, + "learning_rate": 1.3428443190561682e-05, + "loss": 0.2278, + "step": 57087 + }, + { + "epoch": 2.19, + "learning_rate": 1.3427802000512954e-05, + "loss": 0.2005, + "step": 57088 + }, + { + "epoch": 2.19, + "learning_rate": 1.3427160810464224e-05, + "loss": 0.1891, + "step": 57089 + }, + { + "epoch": 2.19, + "learning_rate": 1.3426519620415492e-05, + "loss": 0.2668, + "step": 57090 + }, + { + "epoch": 2.19, + "learning_rate": 1.3425878430366762e-05, + "loss": 0.2997, + "step": 57091 + }, + { + "epoch": 2.2, + "learning_rate": 1.342523724031803e-05, + "loss": 0.2792, + "step": 57092 + }, + { + "epoch": 2.2, + "learning_rate": 1.34245960502693e-05, + "loss": 0.2803, + "step": 57093 + }, + { + "epoch": 2.2, + "learning_rate": 1.3423954860220569e-05, + "loss": 0.1546, + "step": 57094 + }, + { + "epoch": 2.2, + "learning_rate": 1.3423313670171839e-05, + "loss": 0.2828, + "step": 57095 + }, + { + "epoch": 2.2, + "learning_rate": 1.3422672480123107e-05, + "loss": 0.2149, + "step": 57096 + }, + { + "epoch": 2.2, + "learning_rate": 1.342203129007438e-05, + "loss": 0.2338, + "step": 57097 + }, + { + "epoch": 2.2, + "learning_rate": 1.3421390100025649e-05, + "loss": 0.2408, + "step": 57098 + }, + { + "epoch": 2.2, + "learning_rate": 1.3420748909976919e-05, + "loss": 0.1806, + "step": 57099 + }, + { + "epoch": 2.2, + "learning_rate": 1.3420107719928187e-05, + "loss": 0.1944, + "step": 57100 + }, + { + "epoch": 2.2, + "learning_rate": 1.3419466529879457e-05, + "loss": 0.4377, + "step": 57101 + }, + { + "epoch": 2.2, + "learning_rate": 1.3418825339830726e-05, + "loss": 0.2258, + "step": 57102 + }, + { + "epoch": 2.2, + "learning_rate": 1.3418184149781996e-05, + "loss": 0.2512, + "step": 57103 + }, + { + "epoch": 2.2, + "learning_rate": 1.3417542959733264e-05, + "loss": 0.1559, + "step": 57104 + }, + { + "epoch": 2.2, + "learning_rate": 1.3416901769684534e-05, + "loss": 0.2096, + "step": 57105 + }, + { + "epoch": 2.2, + "learning_rate": 1.3416260579635806e-05, + "loss": 0.381, + "step": 57106 + }, + { + "epoch": 2.2, + "learning_rate": 1.3415619389587076e-05, + "loss": 0.2019, + "step": 57107 + }, + { + "epoch": 2.2, + "learning_rate": 1.3414978199538344e-05, + "loss": 0.1862, + "step": 57108 + }, + { + "epoch": 2.2, + "learning_rate": 1.3414337009489614e-05, + "loss": 0.1407, + "step": 57109 + }, + { + "epoch": 2.2, + "learning_rate": 1.3413695819440883e-05, + "loss": 0.2578, + "step": 57110 + }, + { + "epoch": 2.2, + "learning_rate": 1.3413054629392153e-05, + "loss": 0.4279, + "step": 57111 + }, + { + "epoch": 2.2, + "learning_rate": 1.3412413439343421e-05, + "loss": 0.1528, + "step": 57112 + }, + { + "epoch": 2.2, + "learning_rate": 1.3411772249294691e-05, + "loss": 0.2172, + "step": 57113 + }, + { + "epoch": 2.2, + "learning_rate": 1.341113105924596e-05, + "loss": 0.2033, + "step": 57114 + }, + { + "epoch": 2.2, + "learning_rate": 1.3410489869197233e-05, + "loss": 0.2986, + "step": 57115 + }, + { + "epoch": 2.2, + "learning_rate": 1.3409848679148501e-05, + "loss": 0.3664, + "step": 57116 + }, + { + "epoch": 2.2, + "learning_rate": 1.3409207489099771e-05, + "loss": 0.3014, + "step": 57117 + }, + { + "epoch": 2.2, + "learning_rate": 1.340856629905104e-05, + "loss": 0.2947, + "step": 57118 + }, + { + "epoch": 2.2, + "learning_rate": 1.340792510900231e-05, + "loss": 0.3055, + "step": 57119 + }, + { + "epoch": 2.2, + "learning_rate": 1.3407283918953578e-05, + "loss": 0.2426, + "step": 57120 + }, + { + "epoch": 2.2, + "learning_rate": 1.3406642728904848e-05, + "loss": 0.193, + "step": 57121 + }, + { + "epoch": 2.2, + "learning_rate": 1.3406001538856116e-05, + "loss": 0.331, + "step": 57122 + }, + { + "epoch": 2.2, + "learning_rate": 1.3405360348807386e-05, + "loss": 0.2713, + "step": 57123 + }, + { + "epoch": 2.2, + "learning_rate": 1.3404719158758658e-05, + "loss": 0.388, + "step": 57124 + }, + { + "epoch": 2.2, + "learning_rate": 1.3404077968709928e-05, + "loss": 0.2012, + "step": 57125 + }, + { + "epoch": 2.2, + "learning_rate": 1.3403436778661196e-05, + "loss": 0.1814, + "step": 57126 + }, + { + "epoch": 2.2, + "learning_rate": 1.3402795588612466e-05, + "loss": 0.2964, + "step": 57127 + }, + { + "epoch": 2.2, + "learning_rate": 1.3402154398563734e-05, + "loss": 0.1539, + "step": 57128 + }, + { + "epoch": 2.2, + "learning_rate": 1.3401513208515004e-05, + "loss": 0.1886, + "step": 57129 + }, + { + "epoch": 2.2, + "learning_rate": 1.3400872018466273e-05, + "loss": 0.3316, + "step": 57130 + }, + { + "epoch": 2.2, + "learning_rate": 1.3400230828417543e-05, + "loss": 0.1058, + "step": 57131 + }, + { + "epoch": 2.2, + "learning_rate": 1.3399589638368811e-05, + "loss": 0.1894, + "step": 57132 + }, + { + "epoch": 2.2, + "learning_rate": 1.3398948448320085e-05, + "loss": 0.2462, + "step": 57133 + }, + { + "epoch": 2.2, + "learning_rate": 1.3398307258271353e-05, + "loss": 0.1605, + "step": 57134 + }, + { + "epoch": 2.2, + "learning_rate": 1.3397666068222623e-05, + "loss": 0.2108, + "step": 57135 + }, + { + "epoch": 2.2, + "learning_rate": 1.3397024878173891e-05, + "loss": 0.2679, + "step": 57136 + }, + { + "epoch": 2.2, + "learning_rate": 1.3396383688125161e-05, + "loss": 0.293, + "step": 57137 + }, + { + "epoch": 2.2, + "learning_rate": 1.339574249807643e-05, + "loss": 0.2674, + "step": 57138 + }, + { + "epoch": 2.2, + "learning_rate": 1.33951013080277e-05, + "loss": 0.3514, + "step": 57139 + }, + { + "epoch": 2.2, + "learning_rate": 1.3394460117978968e-05, + "loss": 0.2102, + "step": 57140 + }, + { + "epoch": 2.2, + "learning_rate": 1.3393818927930238e-05, + "loss": 0.3356, + "step": 57141 + }, + { + "epoch": 2.2, + "learning_rate": 1.339317773788151e-05, + "loss": 0.305, + "step": 57142 + }, + { + "epoch": 2.2, + "learning_rate": 1.339253654783278e-05, + "loss": 0.2642, + "step": 57143 + }, + { + "epoch": 2.2, + "learning_rate": 1.3391895357784048e-05, + "loss": 0.166, + "step": 57144 + }, + { + "epoch": 2.2, + "learning_rate": 1.3391254167735318e-05, + "loss": 0.0905, + "step": 57145 + }, + { + "epoch": 2.2, + "learning_rate": 1.3390612977686586e-05, + "loss": 0.1923, + "step": 57146 + }, + { + "epoch": 2.2, + "learning_rate": 1.3389971787637856e-05, + "loss": 0.2358, + "step": 57147 + }, + { + "epoch": 2.2, + "learning_rate": 1.3389330597589125e-05, + "loss": 0.242, + "step": 57148 + }, + { + "epoch": 2.2, + "learning_rate": 1.3388689407540395e-05, + "loss": 0.1318, + "step": 57149 + }, + { + "epoch": 2.2, + "learning_rate": 1.3388048217491663e-05, + "loss": 0.2085, + "step": 57150 + }, + { + "epoch": 2.2, + "learning_rate": 1.3387407027442937e-05, + "loss": 0.2391, + "step": 57151 + }, + { + "epoch": 2.2, + "learning_rate": 1.3386765837394205e-05, + "loss": 0.2457, + "step": 57152 + }, + { + "epoch": 2.2, + "learning_rate": 1.3386124647345475e-05, + "loss": 0.2689, + "step": 57153 + }, + { + "epoch": 2.2, + "learning_rate": 1.3385483457296743e-05, + "loss": 0.2612, + "step": 57154 + }, + { + "epoch": 2.2, + "learning_rate": 1.3384842267248013e-05, + "loss": 0.2026, + "step": 57155 + }, + { + "epoch": 2.2, + "learning_rate": 1.3384201077199282e-05, + "loss": 0.2153, + "step": 57156 + }, + { + "epoch": 2.2, + "learning_rate": 1.3383559887150552e-05, + "loss": 0.298, + "step": 57157 + }, + { + "epoch": 2.2, + "learning_rate": 1.338291869710182e-05, + "loss": 0.2198, + "step": 57158 + }, + { + "epoch": 2.2, + "learning_rate": 1.338227750705309e-05, + "loss": 0.133, + "step": 57159 + }, + { + "epoch": 2.2, + "learning_rate": 1.3381636317004362e-05, + "loss": 0.2799, + "step": 57160 + }, + { + "epoch": 2.2, + "learning_rate": 1.3380995126955632e-05, + "loss": 0.4789, + "step": 57161 + }, + { + "epoch": 2.2, + "learning_rate": 1.33803539369069e-05, + "loss": 0.1765, + "step": 57162 + }, + { + "epoch": 2.2, + "learning_rate": 1.337971274685817e-05, + "loss": 0.1778, + "step": 57163 + }, + { + "epoch": 2.2, + "learning_rate": 1.3379071556809438e-05, + "loss": 0.2462, + "step": 57164 + }, + { + "epoch": 2.2, + "learning_rate": 1.3378430366760708e-05, + "loss": 0.3114, + "step": 57165 + }, + { + "epoch": 2.2, + "learning_rate": 1.3377789176711977e-05, + "loss": 0.2648, + "step": 57166 + }, + { + "epoch": 2.2, + "learning_rate": 1.3377147986663247e-05, + "loss": 0.1317, + "step": 57167 + }, + { + "epoch": 2.2, + "learning_rate": 1.3376506796614518e-05, + "loss": 0.2081, + "step": 57168 + }, + { + "epoch": 2.2, + "learning_rate": 1.3375865606565788e-05, + "loss": 0.1775, + "step": 57169 + }, + { + "epoch": 2.2, + "learning_rate": 1.3375224416517057e-05, + "loss": 0.1434, + "step": 57170 + }, + { + "epoch": 2.2, + "learning_rate": 1.3374583226468327e-05, + "loss": 0.2228, + "step": 57171 + }, + { + "epoch": 2.2, + "learning_rate": 1.3373942036419595e-05, + "loss": 0.3097, + "step": 57172 + }, + { + "epoch": 2.2, + "learning_rate": 1.3373300846370865e-05, + "loss": 0.2015, + "step": 57173 + }, + { + "epoch": 2.2, + "learning_rate": 1.3372659656322133e-05, + "loss": 0.1595, + "step": 57174 + }, + { + "epoch": 2.2, + "learning_rate": 1.3372018466273404e-05, + "loss": 0.2947, + "step": 57175 + }, + { + "epoch": 2.2, + "learning_rate": 1.3371377276224672e-05, + "loss": 0.1776, + "step": 57176 + }, + { + "epoch": 2.2, + "learning_rate": 1.3370736086175945e-05, + "loss": 0.2307, + "step": 57177 + }, + { + "epoch": 2.2, + "learning_rate": 1.3370094896127214e-05, + "loss": 0.4058, + "step": 57178 + }, + { + "epoch": 2.2, + "learning_rate": 1.3369453706078484e-05, + "loss": 0.1731, + "step": 57179 + }, + { + "epoch": 2.2, + "learning_rate": 1.3368812516029752e-05, + "loss": 0.3141, + "step": 57180 + }, + { + "epoch": 2.2, + "learning_rate": 1.3368171325981022e-05, + "loss": 0.1699, + "step": 57181 + }, + { + "epoch": 2.2, + "learning_rate": 1.336753013593229e-05, + "loss": 0.1642, + "step": 57182 + }, + { + "epoch": 2.2, + "learning_rate": 1.336688894588356e-05, + "loss": 0.2103, + "step": 57183 + }, + { + "epoch": 2.2, + "learning_rate": 1.3366247755834829e-05, + "loss": 0.174, + "step": 57184 + }, + { + "epoch": 2.2, + "learning_rate": 1.3365606565786099e-05, + "loss": 0.2606, + "step": 57185 + }, + { + "epoch": 2.2, + "learning_rate": 1.336496537573737e-05, + "loss": 0.2654, + "step": 57186 + }, + { + "epoch": 2.2, + "learning_rate": 1.336432418568864e-05, + "loss": 0.1533, + "step": 57187 + }, + { + "epoch": 2.2, + "learning_rate": 1.3363682995639909e-05, + "loss": 0.2371, + "step": 57188 + }, + { + "epoch": 2.2, + "learning_rate": 1.3363041805591179e-05, + "loss": 0.373, + "step": 57189 + }, + { + "epoch": 2.2, + "learning_rate": 1.3362400615542447e-05, + "loss": 0.1526, + "step": 57190 + }, + { + "epoch": 2.2, + "learning_rate": 1.3361759425493717e-05, + "loss": 0.2239, + "step": 57191 + }, + { + "epoch": 2.2, + "learning_rate": 1.3361118235444985e-05, + "loss": 0.2338, + "step": 57192 + }, + { + "epoch": 2.2, + "learning_rate": 1.3360477045396255e-05, + "loss": 0.1415, + "step": 57193 + }, + { + "epoch": 2.2, + "learning_rate": 1.3359835855347524e-05, + "loss": 0.3229, + "step": 57194 + }, + { + "epoch": 2.2, + "learning_rate": 1.3359194665298797e-05, + "loss": 0.2815, + "step": 57195 + }, + { + "epoch": 2.2, + "learning_rate": 1.3358553475250066e-05, + "loss": 0.2631, + "step": 57196 + }, + { + "epoch": 2.2, + "learning_rate": 1.3357912285201336e-05, + "loss": 0.4454, + "step": 57197 + }, + { + "epoch": 2.2, + "learning_rate": 1.3357271095152604e-05, + "loss": 0.2738, + "step": 57198 + }, + { + "epoch": 2.2, + "learning_rate": 1.3356629905103874e-05, + "loss": 0.2545, + "step": 57199 + }, + { + "epoch": 2.2, + "learning_rate": 1.3355988715055142e-05, + "loss": 0.2994, + "step": 57200 + }, + { + "epoch": 2.2, + "learning_rate": 1.3355347525006412e-05, + "loss": 0.1631, + "step": 57201 + }, + { + "epoch": 2.2, + "learning_rate": 1.335470633495768e-05, + "loss": 0.1422, + "step": 57202 + }, + { + "epoch": 2.2, + "learning_rate": 1.335406514490895e-05, + "loss": 0.2556, + "step": 57203 + }, + { + "epoch": 2.2, + "learning_rate": 1.3353423954860222e-05, + "loss": 0.2519, + "step": 57204 + }, + { + "epoch": 2.2, + "learning_rate": 1.3352782764811492e-05, + "loss": 0.1379, + "step": 57205 + }, + { + "epoch": 2.2, + "learning_rate": 1.335214157476276e-05, + "loss": 0.2363, + "step": 57206 + }, + { + "epoch": 2.2, + "learning_rate": 1.335150038471403e-05, + "loss": 0.1981, + "step": 57207 + }, + { + "epoch": 2.2, + "learning_rate": 1.3350859194665299e-05, + "loss": 0.2238, + "step": 57208 + }, + { + "epoch": 2.2, + "learning_rate": 1.3350218004616569e-05, + "loss": 0.2264, + "step": 57209 + }, + { + "epoch": 2.2, + "learning_rate": 1.3349576814567837e-05, + "loss": 0.2295, + "step": 57210 + }, + { + "epoch": 2.2, + "learning_rate": 1.3348935624519107e-05, + "loss": 0.1946, + "step": 57211 + }, + { + "epoch": 2.2, + "learning_rate": 1.3348294434470376e-05, + "loss": 0.2128, + "step": 57212 + }, + { + "epoch": 2.2, + "learning_rate": 1.3347653244421649e-05, + "loss": 0.2009, + "step": 57213 + }, + { + "epoch": 2.2, + "learning_rate": 1.3347012054372917e-05, + "loss": 0.2008, + "step": 57214 + }, + { + "epoch": 2.2, + "learning_rate": 1.3346370864324187e-05, + "loss": 0.146, + "step": 57215 + }, + { + "epoch": 2.2, + "learning_rate": 1.3345729674275456e-05, + "loss": 0.2025, + "step": 57216 + }, + { + "epoch": 2.2, + "learning_rate": 1.3345088484226726e-05, + "loss": 0.3786, + "step": 57217 + }, + { + "epoch": 2.2, + "learning_rate": 1.3344447294177994e-05, + "loss": 0.166, + "step": 57218 + }, + { + "epoch": 2.2, + "learning_rate": 1.3343806104129264e-05, + "loss": 0.1203, + "step": 57219 + }, + { + "epoch": 2.2, + "learning_rate": 1.3343164914080532e-05, + "loss": 0.1974, + "step": 57220 + }, + { + "epoch": 2.2, + "learning_rate": 1.3342523724031803e-05, + "loss": 0.1783, + "step": 57221 + }, + { + "epoch": 2.2, + "learning_rate": 1.3341882533983074e-05, + "loss": 0.1697, + "step": 57222 + }, + { + "epoch": 2.2, + "learning_rate": 1.3341241343934344e-05, + "loss": 0.2325, + "step": 57223 + }, + { + "epoch": 2.2, + "learning_rate": 1.3340600153885613e-05, + "loss": 0.2717, + "step": 57224 + }, + { + "epoch": 2.2, + "learning_rate": 1.3339958963836883e-05, + "loss": 0.2062, + "step": 57225 + }, + { + "epoch": 2.2, + "learning_rate": 1.3339317773788151e-05, + "loss": 0.1911, + "step": 57226 + }, + { + "epoch": 2.2, + "learning_rate": 1.3338676583739421e-05, + "loss": 0.3353, + "step": 57227 + }, + { + "epoch": 2.2, + "learning_rate": 1.333803539369069e-05, + "loss": 0.2557, + "step": 57228 + }, + { + "epoch": 2.2, + "learning_rate": 1.333739420364196e-05, + "loss": 0.1917, + "step": 57229 + }, + { + "epoch": 2.2, + "learning_rate": 1.3336753013593228e-05, + "loss": 0.2351, + "step": 57230 + }, + { + "epoch": 2.2, + "learning_rate": 1.3336111823544501e-05, + "loss": 0.1683, + "step": 57231 + }, + { + "epoch": 2.2, + "learning_rate": 1.333547063349577e-05, + "loss": 0.2985, + "step": 57232 + }, + { + "epoch": 2.2, + "learning_rate": 1.333482944344704e-05, + "loss": 0.291, + "step": 57233 + }, + { + "epoch": 2.2, + "learning_rate": 1.3334188253398308e-05, + "loss": 0.2462, + "step": 57234 + }, + { + "epoch": 2.2, + "learning_rate": 1.3333547063349578e-05, + "loss": 0.2046, + "step": 57235 + }, + { + "epoch": 2.2, + "learning_rate": 1.3332905873300846e-05, + "loss": 0.379, + "step": 57236 + }, + { + "epoch": 2.2, + "learning_rate": 1.3332264683252116e-05, + "loss": 0.3338, + "step": 57237 + }, + { + "epoch": 2.2, + "learning_rate": 1.3331623493203384e-05, + "loss": 0.3117, + "step": 57238 + }, + { + "epoch": 2.2, + "learning_rate": 1.3330982303154654e-05, + "loss": 0.26, + "step": 57239 + }, + { + "epoch": 2.2, + "learning_rate": 1.3330341113105926e-05, + "loss": 0.1416, + "step": 57240 + }, + { + "epoch": 2.2, + "learning_rate": 1.3329699923057196e-05, + "loss": 0.2877, + "step": 57241 + }, + { + "epoch": 2.2, + "learning_rate": 1.3329058733008465e-05, + "loss": 0.2661, + "step": 57242 + }, + { + "epoch": 2.2, + "learning_rate": 1.3328417542959735e-05, + "loss": 0.2996, + "step": 57243 + }, + { + "epoch": 2.2, + "learning_rate": 1.3327776352911003e-05, + "loss": 0.162, + "step": 57244 + }, + { + "epoch": 2.2, + "learning_rate": 1.3327135162862273e-05, + "loss": 0.2418, + "step": 57245 + }, + { + "epoch": 2.2, + "learning_rate": 1.3326493972813541e-05, + "loss": 0.1487, + "step": 57246 + }, + { + "epoch": 2.2, + "learning_rate": 1.3325852782764811e-05, + "loss": 0.1685, + "step": 57247 + }, + { + "epoch": 2.2, + "learning_rate": 1.332521159271608e-05, + "loss": 0.2528, + "step": 57248 + }, + { + "epoch": 2.2, + "learning_rate": 1.3324570402667353e-05, + "loss": 0.239, + "step": 57249 + }, + { + "epoch": 2.2, + "learning_rate": 1.3323929212618621e-05, + "loss": 0.2064, + "step": 57250 + }, + { + "epoch": 2.2, + "learning_rate": 1.3323288022569891e-05, + "loss": 0.2684, + "step": 57251 + }, + { + "epoch": 2.2, + "learning_rate": 1.332264683252116e-05, + "loss": 0.3559, + "step": 57252 + }, + { + "epoch": 2.2, + "learning_rate": 1.332200564247243e-05, + "loss": 0.2793, + "step": 57253 + }, + { + "epoch": 2.2, + "learning_rate": 1.3321364452423698e-05, + "loss": 0.1833, + "step": 57254 + }, + { + "epoch": 2.2, + "learning_rate": 1.3320723262374968e-05, + "loss": 0.1284, + "step": 57255 + }, + { + "epoch": 2.2, + "learning_rate": 1.3320082072326236e-05, + "loss": 0.3337, + "step": 57256 + }, + { + "epoch": 2.2, + "learning_rate": 1.3319440882277506e-05, + "loss": 0.2953, + "step": 57257 + }, + { + "epoch": 2.2, + "learning_rate": 1.3318799692228778e-05, + "loss": 0.2878, + "step": 57258 + }, + { + "epoch": 2.2, + "learning_rate": 1.3318158502180048e-05, + "loss": 0.2393, + "step": 57259 + }, + { + "epoch": 2.2, + "learning_rate": 1.3317517312131316e-05, + "loss": 0.2357, + "step": 57260 + }, + { + "epoch": 2.2, + "learning_rate": 1.3316876122082586e-05, + "loss": 0.2508, + "step": 57261 + }, + { + "epoch": 2.2, + "learning_rate": 1.3316234932033855e-05, + "loss": 0.2001, + "step": 57262 + }, + { + "epoch": 2.2, + "learning_rate": 1.3315593741985125e-05, + "loss": 0.3913, + "step": 57263 + }, + { + "epoch": 2.2, + "learning_rate": 1.3314952551936393e-05, + "loss": 0.246, + "step": 57264 + }, + { + "epoch": 2.2, + "learning_rate": 1.3314311361887663e-05, + "loss": 0.2215, + "step": 57265 + }, + { + "epoch": 2.2, + "learning_rate": 1.3313670171838935e-05, + "loss": 0.2334, + "step": 57266 + }, + { + "epoch": 2.2, + "learning_rate": 1.3313028981790205e-05, + "loss": 0.1608, + "step": 57267 + }, + { + "epoch": 2.2, + "learning_rate": 1.3312387791741473e-05, + "loss": 0.3129, + "step": 57268 + }, + { + "epoch": 2.2, + "learning_rate": 1.3311746601692743e-05, + "loss": 0.2586, + "step": 57269 + }, + { + "epoch": 2.2, + "learning_rate": 1.3311105411644012e-05, + "loss": 0.1377, + "step": 57270 + }, + { + "epoch": 2.2, + "learning_rate": 1.3310464221595282e-05, + "loss": 0.1281, + "step": 57271 + }, + { + "epoch": 2.2, + "learning_rate": 1.330982303154655e-05, + "loss": 0.1758, + "step": 57272 + }, + { + "epoch": 2.2, + "learning_rate": 1.330918184149782e-05, + "loss": 0.2115, + "step": 57273 + }, + { + "epoch": 2.2, + "learning_rate": 1.3308540651449088e-05, + "loss": 0.2402, + "step": 57274 + }, + { + "epoch": 2.2, + "learning_rate": 1.3307899461400362e-05, + "loss": 0.2677, + "step": 57275 + }, + { + "epoch": 2.2, + "learning_rate": 1.330725827135163e-05, + "loss": 0.2044, + "step": 57276 + }, + { + "epoch": 2.2, + "learning_rate": 1.33066170813029e-05, + "loss": 0.3328, + "step": 57277 + }, + { + "epoch": 2.2, + "learning_rate": 1.3305975891254168e-05, + "loss": 0.1878, + "step": 57278 + }, + { + "epoch": 2.2, + "learning_rate": 1.3305334701205438e-05, + "loss": 0.1957, + "step": 57279 + }, + { + "epoch": 2.2, + "learning_rate": 1.3304693511156707e-05, + "loss": 0.1588, + "step": 57280 + }, + { + "epoch": 2.2, + "learning_rate": 1.3304052321107977e-05, + "loss": 0.221, + "step": 57281 + }, + { + "epoch": 2.2, + "learning_rate": 1.3303411131059245e-05, + "loss": 0.2835, + "step": 57282 + }, + { + "epoch": 2.2, + "learning_rate": 1.3302769941010515e-05, + "loss": 0.2536, + "step": 57283 + }, + { + "epoch": 2.2, + "learning_rate": 1.3302128750961787e-05, + "loss": 0.2304, + "step": 57284 + }, + { + "epoch": 2.2, + "learning_rate": 1.3301487560913057e-05, + "loss": 0.1815, + "step": 57285 + }, + { + "epoch": 2.2, + "learning_rate": 1.3300846370864325e-05, + "loss": 0.2028, + "step": 57286 + }, + { + "epoch": 2.2, + "learning_rate": 1.3300205180815595e-05, + "loss": 0.2126, + "step": 57287 + }, + { + "epoch": 2.2, + "learning_rate": 1.3299563990766864e-05, + "loss": 0.2975, + "step": 57288 + }, + { + "epoch": 2.2, + "learning_rate": 1.3298922800718134e-05, + "loss": 0.2578, + "step": 57289 + }, + { + "epoch": 2.2, + "learning_rate": 1.3298281610669402e-05, + "loss": 0.1522, + "step": 57290 + }, + { + "epoch": 2.2, + "learning_rate": 1.3297640420620672e-05, + "loss": 0.255, + "step": 57291 + }, + { + "epoch": 2.2, + "learning_rate": 1.329699923057194e-05, + "loss": 0.1729, + "step": 57292 + }, + { + "epoch": 2.2, + "learning_rate": 1.3296358040523214e-05, + "loss": 0.1922, + "step": 57293 + }, + { + "epoch": 2.2, + "learning_rate": 1.3295716850474482e-05, + "loss": 0.3438, + "step": 57294 + }, + { + "epoch": 2.2, + "learning_rate": 1.3295075660425752e-05, + "loss": 0.2409, + "step": 57295 + }, + { + "epoch": 2.2, + "learning_rate": 1.329443447037702e-05, + "loss": 0.2797, + "step": 57296 + }, + { + "epoch": 2.2, + "learning_rate": 1.329379328032829e-05, + "loss": 0.2904, + "step": 57297 + }, + { + "epoch": 2.2, + "learning_rate": 1.3293152090279559e-05, + "loss": 0.133, + "step": 57298 + }, + { + "epoch": 2.2, + "learning_rate": 1.3292510900230829e-05, + "loss": 0.1713, + "step": 57299 + }, + { + "epoch": 2.2, + "learning_rate": 1.3291869710182097e-05, + "loss": 0.2488, + "step": 57300 + }, + { + "epoch": 2.2, + "learning_rate": 1.3291228520133367e-05, + "loss": 0.1588, + "step": 57301 + }, + { + "epoch": 2.2, + "learning_rate": 1.3290587330084639e-05, + "loss": 0.2239, + "step": 57302 + }, + { + "epoch": 2.2, + "learning_rate": 1.3289946140035909e-05, + "loss": 0.3301, + "step": 57303 + }, + { + "epoch": 2.2, + "learning_rate": 1.3289304949987177e-05, + "loss": 0.281, + "step": 57304 + }, + { + "epoch": 2.2, + "learning_rate": 1.3288663759938447e-05, + "loss": 0.2309, + "step": 57305 + }, + { + "epoch": 2.2, + "learning_rate": 1.3288022569889715e-05, + "loss": 0.2283, + "step": 57306 + }, + { + "epoch": 2.2, + "learning_rate": 1.3287381379840985e-05, + "loss": 0.2251, + "step": 57307 + }, + { + "epoch": 2.2, + "learning_rate": 1.3286740189792254e-05, + "loss": 0.1853, + "step": 57308 + }, + { + "epoch": 2.2, + "learning_rate": 1.3286098999743524e-05, + "loss": 0.1711, + "step": 57309 + }, + { + "epoch": 2.2, + "learning_rate": 1.3285457809694792e-05, + "loss": 0.2739, + "step": 57310 + }, + { + "epoch": 2.2, + "learning_rate": 1.3284816619646066e-05, + "loss": 0.1495, + "step": 57311 + }, + { + "epoch": 2.2, + "learning_rate": 1.3284175429597334e-05, + "loss": 0.3927, + "step": 57312 + }, + { + "epoch": 2.2, + "learning_rate": 1.3283534239548604e-05, + "loss": 0.3173, + "step": 57313 + }, + { + "epoch": 2.2, + "learning_rate": 1.3282893049499872e-05, + "loss": 0.2946, + "step": 57314 + }, + { + "epoch": 2.2, + "learning_rate": 1.3282251859451142e-05, + "loss": 0.1321, + "step": 57315 + }, + { + "epoch": 2.2, + "learning_rate": 1.328161066940241e-05, + "loss": 0.2014, + "step": 57316 + }, + { + "epoch": 2.2, + "learning_rate": 1.328096947935368e-05, + "loss": 0.3045, + "step": 57317 + }, + { + "epoch": 2.2, + "learning_rate": 1.3280328289304949e-05, + "loss": 0.2501, + "step": 57318 + }, + { + "epoch": 2.2, + "learning_rate": 1.3279687099256219e-05, + "loss": 0.3052, + "step": 57319 + }, + { + "epoch": 2.2, + "learning_rate": 1.327904590920749e-05, + "loss": 0.1836, + "step": 57320 + }, + { + "epoch": 2.2, + "learning_rate": 1.327840471915876e-05, + "loss": 0.1826, + "step": 57321 + }, + { + "epoch": 2.2, + "learning_rate": 1.3277763529110029e-05, + "loss": 0.0923, + "step": 57322 + }, + { + "epoch": 2.2, + "learning_rate": 1.3277122339061299e-05, + "loss": 0.2302, + "step": 57323 + }, + { + "epoch": 2.2, + "learning_rate": 1.3276481149012567e-05, + "loss": 0.2334, + "step": 57324 + }, + { + "epoch": 2.2, + "learning_rate": 1.3275839958963837e-05, + "loss": 0.1286, + "step": 57325 + }, + { + "epoch": 2.2, + "learning_rate": 1.3275198768915106e-05, + "loss": 0.2038, + "step": 57326 + }, + { + "epoch": 2.2, + "learning_rate": 1.3274557578866376e-05, + "loss": 0.2538, + "step": 57327 + }, + { + "epoch": 2.2, + "learning_rate": 1.3273916388817644e-05, + "loss": 0.2746, + "step": 57328 + }, + { + "epoch": 2.2, + "learning_rate": 1.3273275198768918e-05, + "loss": 0.166, + "step": 57329 + }, + { + "epoch": 2.2, + "learning_rate": 1.3272634008720186e-05, + "loss": 0.3731, + "step": 57330 + }, + { + "epoch": 2.2, + "learning_rate": 1.3271992818671456e-05, + "loss": 0.2217, + "step": 57331 + }, + { + "epoch": 2.2, + "learning_rate": 1.3271351628622724e-05, + "loss": 0.349, + "step": 57332 + }, + { + "epoch": 2.2, + "learning_rate": 1.3270710438573994e-05, + "loss": 0.247, + "step": 57333 + }, + { + "epoch": 2.2, + "learning_rate": 1.3270069248525263e-05, + "loss": 0.3066, + "step": 57334 + }, + { + "epoch": 2.2, + "learning_rate": 1.3269428058476533e-05, + "loss": 0.1698, + "step": 57335 + }, + { + "epoch": 2.2, + "learning_rate": 1.3268786868427801e-05, + "loss": 0.1779, + "step": 57336 + }, + { + "epoch": 2.2, + "learning_rate": 1.3268145678379071e-05, + "loss": 0.1663, + "step": 57337 + }, + { + "epoch": 2.2, + "learning_rate": 1.3267504488330343e-05, + "loss": 0.1836, + "step": 57338 + }, + { + "epoch": 2.2, + "learning_rate": 1.3266863298281613e-05, + "loss": 0.2203, + "step": 57339 + }, + { + "epoch": 2.2, + "learning_rate": 1.3266222108232881e-05, + "loss": 0.3181, + "step": 57340 + }, + { + "epoch": 2.2, + "learning_rate": 1.3265580918184151e-05, + "loss": 0.3664, + "step": 57341 + }, + { + "epoch": 2.2, + "learning_rate": 1.326493972813542e-05, + "loss": 0.1622, + "step": 57342 + }, + { + "epoch": 2.2, + "learning_rate": 1.326429853808669e-05, + "loss": 0.2468, + "step": 57343 + }, + { + "epoch": 2.2, + "learning_rate": 1.3263657348037958e-05, + "loss": 0.1961, + "step": 57344 + }, + { + "epoch": 2.2, + "learning_rate": 1.3263016157989228e-05, + "loss": 0.2034, + "step": 57345 + }, + { + "epoch": 2.2, + "learning_rate": 1.3262374967940496e-05, + "loss": 0.378, + "step": 57346 + }, + { + "epoch": 2.2, + "learning_rate": 1.326173377789177e-05, + "loss": 0.1801, + "step": 57347 + }, + { + "epoch": 2.2, + "learning_rate": 1.3261092587843038e-05, + "loss": 0.3571, + "step": 57348 + }, + { + "epoch": 2.2, + "learning_rate": 1.3260451397794308e-05, + "loss": 0.2793, + "step": 57349 + }, + { + "epoch": 2.2, + "learning_rate": 1.3259810207745576e-05, + "loss": 0.1274, + "step": 57350 + }, + { + "epoch": 2.2, + "learning_rate": 1.3259169017696846e-05, + "loss": 0.4245, + "step": 57351 + }, + { + "epoch": 2.2, + "learning_rate": 1.3258527827648114e-05, + "loss": 0.2342, + "step": 57352 + }, + { + "epoch": 2.21, + "learning_rate": 1.3257886637599385e-05, + "loss": 0.3665, + "step": 57353 + }, + { + "epoch": 2.21, + "learning_rate": 1.3257245447550653e-05, + "loss": 0.1938, + "step": 57354 + }, + { + "epoch": 2.21, + "learning_rate": 1.3256604257501926e-05, + "loss": 0.2727, + "step": 57355 + }, + { + "epoch": 2.21, + "learning_rate": 1.3255963067453195e-05, + "loss": 0.2302, + "step": 57356 + }, + { + "epoch": 2.21, + "learning_rate": 1.3255321877404465e-05, + "loss": 0.1447, + "step": 57357 + }, + { + "epoch": 2.21, + "learning_rate": 1.3254680687355733e-05, + "loss": 0.2557, + "step": 57358 + }, + { + "epoch": 2.21, + "learning_rate": 1.3254039497307003e-05, + "loss": 0.2153, + "step": 57359 + }, + { + "epoch": 2.21, + "learning_rate": 1.3253398307258271e-05, + "loss": 0.3323, + "step": 57360 + }, + { + "epoch": 2.21, + "learning_rate": 1.3252757117209541e-05, + "loss": 0.249, + "step": 57361 + }, + { + "epoch": 2.21, + "learning_rate": 1.325211592716081e-05, + "loss": 0.2172, + "step": 57362 + }, + { + "epoch": 2.21, + "learning_rate": 1.325147473711208e-05, + "loss": 0.1721, + "step": 57363 + }, + { + "epoch": 2.21, + "learning_rate": 1.3250833547063351e-05, + "loss": 0.2891, + "step": 57364 + }, + { + "epoch": 2.21, + "learning_rate": 1.3250192357014621e-05, + "loss": 0.2176, + "step": 57365 + }, + { + "epoch": 2.21, + "learning_rate": 1.324955116696589e-05, + "loss": 0.3117, + "step": 57366 + }, + { + "epoch": 2.21, + "learning_rate": 1.324890997691716e-05, + "loss": 0.2991, + "step": 57367 + }, + { + "epoch": 2.21, + "learning_rate": 1.3248268786868428e-05, + "loss": 0.2496, + "step": 57368 + }, + { + "epoch": 2.21, + "learning_rate": 1.3247627596819698e-05, + "loss": 0.1257, + "step": 57369 + }, + { + "epoch": 2.21, + "learning_rate": 1.3246986406770966e-05, + "loss": 0.244, + "step": 57370 + }, + { + "epoch": 2.21, + "learning_rate": 1.3246345216722236e-05, + "loss": 0.1113, + "step": 57371 + }, + { + "epoch": 2.21, + "learning_rate": 1.3245704026673505e-05, + "loss": 0.3831, + "step": 57372 + }, + { + "epoch": 2.21, + "learning_rate": 1.3245062836624778e-05, + "loss": 0.1565, + "step": 57373 + }, + { + "epoch": 2.21, + "learning_rate": 1.3244421646576047e-05, + "loss": 0.2808, + "step": 57374 + }, + { + "epoch": 2.21, + "learning_rate": 1.3243780456527317e-05, + "loss": 0.275, + "step": 57375 + }, + { + "epoch": 2.21, + "learning_rate": 1.3243139266478585e-05, + "loss": 0.2577, + "step": 57376 + }, + { + "epoch": 2.21, + "learning_rate": 1.3242498076429855e-05, + "loss": 0.2915, + "step": 57377 + }, + { + "epoch": 2.21, + "learning_rate": 1.3241856886381123e-05, + "loss": 0.29, + "step": 57378 + }, + { + "epoch": 2.21, + "learning_rate": 1.3241215696332393e-05, + "loss": 0.1939, + "step": 57379 + }, + { + "epoch": 2.21, + "learning_rate": 1.3240574506283662e-05, + "loss": 0.1758, + "step": 57380 + }, + { + "epoch": 2.21, + "learning_rate": 1.3239933316234932e-05, + "loss": 0.1988, + "step": 57381 + }, + { + "epoch": 2.21, + "learning_rate": 1.3239292126186203e-05, + "loss": 0.192, + "step": 57382 + }, + { + "epoch": 2.21, + "learning_rate": 1.3238650936137473e-05, + "loss": 0.1475, + "step": 57383 + }, + { + "epoch": 2.21, + "learning_rate": 1.3238009746088742e-05, + "loss": 0.2129, + "step": 57384 + }, + { + "epoch": 2.21, + "learning_rate": 1.3237368556040012e-05, + "loss": 0.1728, + "step": 57385 + }, + { + "epoch": 2.21, + "learning_rate": 1.323672736599128e-05, + "loss": 0.2779, + "step": 57386 + }, + { + "epoch": 2.21, + "learning_rate": 1.323608617594255e-05, + "loss": 0.3254, + "step": 57387 + }, + { + "epoch": 2.21, + "learning_rate": 1.3235444985893818e-05, + "loss": 0.4035, + "step": 57388 + }, + { + "epoch": 2.21, + "learning_rate": 1.3234803795845088e-05, + "loss": 0.2515, + "step": 57389 + }, + { + "epoch": 2.21, + "learning_rate": 1.3234162605796357e-05, + "loss": 0.1287, + "step": 57390 + }, + { + "epoch": 2.21, + "learning_rate": 1.323352141574763e-05, + "loss": 0.174, + "step": 57391 + }, + { + "epoch": 2.21, + "learning_rate": 1.3232880225698898e-05, + "loss": 0.2861, + "step": 57392 + }, + { + "epoch": 2.21, + "learning_rate": 1.3232239035650168e-05, + "loss": 0.1743, + "step": 57393 + }, + { + "epoch": 2.21, + "learning_rate": 1.3231597845601437e-05, + "loss": 0.4159, + "step": 57394 + }, + { + "epoch": 2.21, + "learning_rate": 1.3230956655552707e-05, + "loss": 0.2523, + "step": 57395 + }, + { + "epoch": 2.21, + "learning_rate": 1.3230315465503975e-05, + "loss": 0.1926, + "step": 57396 + }, + { + "epoch": 2.21, + "learning_rate": 1.3229674275455245e-05, + "loss": 0.2575, + "step": 57397 + }, + { + "epoch": 2.21, + "learning_rate": 1.3229033085406513e-05, + "loss": 0.1147, + "step": 57398 + }, + { + "epoch": 2.21, + "learning_rate": 1.3228391895357784e-05, + "loss": 0.1633, + "step": 57399 + }, + { + "epoch": 2.21, + "learning_rate": 1.3227750705309055e-05, + "loss": 0.255, + "step": 57400 + }, + { + "epoch": 2.21, + "learning_rate": 1.3227109515260325e-05, + "loss": 0.2877, + "step": 57401 + }, + { + "epoch": 2.21, + "learning_rate": 1.3226468325211594e-05, + "loss": 0.2954, + "step": 57402 + }, + { + "epoch": 2.21, + "learning_rate": 1.3225827135162864e-05, + "loss": 0.2473, + "step": 57403 + }, + { + "epoch": 2.21, + "learning_rate": 1.3225185945114132e-05, + "loss": 0.3695, + "step": 57404 + }, + { + "epoch": 2.21, + "learning_rate": 1.3224544755065402e-05, + "loss": 0.2557, + "step": 57405 + }, + { + "epoch": 2.21, + "learning_rate": 1.322390356501667e-05, + "loss": 0.2098, + "step": 57406 + }, + { + "epoch": 2.21, + "learning_rate": 1.322326237496794e-05, + "loss": 0.2672, + "step": 57407 + }, + { + "epoch": 2.21, + "learning_rate": 1.3222621184919209e-05, + "loss": 0.3049, + "step": 57408 + }, + { + "epoch": 2.21, + "learning_rate": 1.3221979994870482e-05, + "loss": 0.2769, + "step": 57409 + }, + { + "epoch": 2.21, + "learning_rate": 1.322133880482175e-05, + "loss": 0.231, + "step": 57410 + }, + { + "epoch": 2.21, + "learning_rate": 1.322069761477302e-05, + "loss": 0.2613, + "step": 57411 + }, + { + "epoch": 2.21, + "learning_rate": 1.3220056424724289e-05, + "loss": 0.1578, + "step": 57412 + }, + { + "epoch": 2.21, + "learning_rate": 1.3219415234675559e-05, + "loss": 0.2357, + "step": 57413 + }, + { + "epoch": 2.21, + "learning_rate": 1.3218774044626827e-05, + "loss": 0.1545, + "step": 57414 + }, + { + "epoch": 2.21, + "learning_rate": 1.3218132854578097e-05, + "loss": 0.2837, + "step": 57415 + }, + { + "epoch": 2.21, + "learning_rate": 1.3217491664529365e-05, + "loss": 0.2521, + "step": 57416 + }, + { + "epoch": 2.21, + "learning_rate": 1.3216850474480635e-05, + "loss": 0.2881, + "step": 57417 + }, + { + "epoch": 2.21, + "learning_rate": 1.3216209284431907e-05, + "loss": 0.1478, + "step": 57418 + }, + { + "epoch": 2.21, + "learning_rate": 1.3215568094383177e-05, + "loss": 0.1593, + "step": 57419 + }, + { + "epoch": 2.21, + "learning_rate": 1.3214926904334446e-05, + "loss": 0.2878, + "step": 57420 + }, + { + "epoch": 2.21, + "learning_rate": 1.3214285714285716e-05, + "loss": 0.1843, + "step": 57421 + }, + { + "epoch": 2.21, + "learning_rate": 1.3213644524236984e-05, + "loss": 0.3601, + "step": 57422 + }, + { + "epoch": 2.21, + "learning_rate": 1.3213003334188254e-05, + "loss": 0.1177, + "step": 57423 + }, + { + "epoch": 2.21, + "learning_rate": 1.3212362144139522e-05, + "loss": 0.1782, + "step": 57424 + }, + { + "epoch": 2.21, + "learning_rate": 1.3211720954090792e-05, + "loss": 0.3109, + "step": 57425 + }, + { + "epoch": 2.21, + "learning_rate": 1.321107976404206e-05, + "loss": 0.2042, + "step": 57426 + }, + { + "epoch": 2.21, + "learning_rate": 1.3210438573993334e-05, + "loss": 0.2655, + "step": 57427 + }, + { + "epoch": 2.21, + "learning_rate": 1.3209797383944602e-05, + "loss": 0.1835, + "step": 57428 + }, + { + "epoch": 2.21, + "learning_rate": 1.3209156193895872e-05, + "loss": 0.3303, + "step": 57429 + }, + { + "epoch": 2.21, + "learning_rate": 1.320851500384714e-05, + "loss": 0.2645, + "step": 57430 + }, + { + "epoch": 2.21, + "learning_rate": 1.320787381379841e-05, + "loss": 0.1244, + "step": 57431 + }, + { + "epoch": 2.21, + "learning_rate": 1.3207232623749679e-05, + "loss": 0.1158, + "step": 57432 + }, + { + "epoch": 2.21, + "learning_rate": 1.3206591433700949e-05, + "loss": 0.3913, + "step": 57433 + }, + { + "epoch": 2.21, + "learning_rate": 1.3205950243652217e-05, + "loss": 0.2778, + "step": 57434 + }, + { + "epoch": 2.21, + "learning_rate": 1.3205309053603487e-05, + "loss": 0.2293, + "step": 57435 + }, + { + "epoch": 2.21, + "learning_rate": 1.3204667863554759e-05, + "loss": 0.2415, + "step": 57436 + }, + { + "epoch": 2.21, + "learning_rate": 1.3204026673506029e-05, + "loss": 0.2035, + "step": 57437 + }, + { + "epoch": 2.21, + "learning_rate": 1.3203385483457297e-05, + "loss": 0.2807, + "step": 57438 + }, + { + "epoch": 2.21, + "learning_rate": 1.3202744293408567e-05, + "loss": 0.2374, + "step": 57439 + }, + { + "epoch": 2.21, + "learning_rate": 1.3202103103359836e-05, + "loss": 0.2723, + "step": 57440 + }, + { + "epoch": 2.21, + "learning_rate": 1.3201461913311106e-05, + "loss": 0.2397, + "step": 57441 + }, + { + "epoch": 2.21, + "learning_rate": 1.3200820723262374e-05, + "loss": 0.3075, + "step": 57442 + }, + { + "epoch": 2.21, + "learning_rate": 1.3200179533213644e-05, + "loss": 0.1791, + "step": 57443 + }, + { + "epoch": 2.21, + "learning_rate": 1.3199538343164913e-05, + "loss": 0.1308, + "step": 57444 + }, + { + "epoch": 2.21, + "learning_rate": 1.3198897153116186e-05, + "loss": 0.2635, + "step": 57445 + }, + { + "epoch": 2.21, + "learning_rate": 1.3198255963067454e-05, + "loss": 0.122, + "step": 57446 + }, + { + "epoch": 2.21, + "learning_rate": 1.3197614773018724e-05, + "loss": 0.3434, + "step": 57447 + }, + { + "epoch": 2.21, + "learning_rate": 1.3196973582969993e-05, + "loss": 0.2266, + "step": 57448 + }, + { + "epoch": 2.21, + "learning_rate": 1.3196332392921263e-05, + "loss": 0.1511, + "step": 57449 + }, + { + "epoch": 2.21, + "learning_rate": 1.3195691202872531e-05, + "loss": 0.1549, + "step": 57450 + }, + { + "epoch": 2.21, + "learning_rate": 1.3195050012823801e-05, + "loss": 0.2805, + "step": 57451 + }, + { + "epoch": 2.21, + "learning_rate": 1.319440882277507e-05, + "loss": 0.3002, + "step": 57452 + }, + { + "epoch": 2.21, + "learning_rate": 1.3193767632726343e-05, + "loss": 0.295, + "step": 57453 + }, + { + "epoch": 2.21, + "learning_rate": 1.3193126442677611e-05, + "loss": 0.2982, + "step": 57454 + }, + { + "epoch": 2.21, + "learning_rate": 1.3192485252628881e-05, + "loss": 0.0949, + "step": 57455 + }, + { + "epoch": 2.21, + "learning_rate": 1.319184406258015e-05, + "loss": 0.1955, + "step": 57456 + }, + { + "epoch": 2.21, + "learning_rate": 1.319120287253142e-05, + "loss": 0.2842, + "step": 57457 + }, + { + "epoch": 2.21, + "learning_rate": 1.3190561682482688e-05, + "loss": 0.1712, + "step": 57458 + }, + { + "epoch": 2.21, + "learning_rate": 1.3189920492433958e-05, + "loss": 0.1795, + "step": 57459 + }, + { + "epoch": 2.21, + "learning_rate": 1.3189279302385226e-05, + "loss": 0.2557, + "step": 57460 + }, + { + "epoch": 2.21, + "learning_rate": 1.3188638112336496e-05, + "loss": 0.2646, + "step": 57461 + }, + { + "epoch": 2.21, + "learning_rate": 1.3187996922287768e-05, + "loss": 0.2012, + "step": 57462 + }, + { + "epoch": 2.21, + "learning_rate": 1.3187355732239038e-05, + "loss": 0.2002, + "step": 57463 + }, + { + "epoch": 2.21, + "learning_rate": 1.3186714542190306e-05, + "loss": 0.229, + "step": 57464 + }, + { + "epoch": 2.21, + "learning_rate": 1.3186073352141576e-05, + "loss": 0.3524, + "step": 57465 + }, + { + "epoch": 2.21, + "learning_rate": 1.3185432162092845e-05, + "loss": 0.2904, + "step": 57466 + }, + { + "epoch": 2.21, + "learning_rate": 1.3184790972044115e-05, + "loss": 0.2123, + "step": 57467 + }, + { + "epoch": 2.21, + "learning_rate": 1.3184149781995383e-05, + "loss": 0.1562, + "step": 57468 + }, + { + "epoch": 2.21, + "learning_rate": 1.3183508591946653e-05, + "loss": 0.273, + "step": 57469 + }, + { + "epoch": 2.21, + "learning_rate": 1.3182867401897921e-05, + "loss": 0.3875, + "step": 57470 + }, + { + "epoch": 2.21, + "learning_rate": 1.3182226211849195e-05, + "loss": 0.2102, + "step": 57471 + }, + { + "epoch": 2.21, + "learning_rate": 1.3181585021800463e-05, + "loss": 0.2581, + "step": 57472 + }, + { + "epoch": 2.21, + "learning_rate": 1.3180943831751733e-05, + "loss": 0.2833, + "step": 57473 + }, + { + "epoch": 2.21, + "learning_rate": 1.3180302641703001e-05, + "loss": 0.152, + "step": 57474 + }, + { + "epoch": 2.21, + "learning_rate": 1.3179661451654271e-05, + "loss": 0.2488, + "step": 57475 + }, + { + "epoch": 2.21, + "learning_rate": 1.317902026160554e-05, + "loss": 0.2372, + "step": 57476 + }, + { + "epoch": 2.21, + "learning_rate": 1.317837907155681e-05, + "loss": 0.3204, + "step": 57477 + }, + { + "epoch": 2.21, + "learning_rate": 1.3177737881508078e-05, + "loss": 0.1094, + "step": 57478 + }, + { + "epoch": 2.21, + "learning_rate": 1.3177096691459348e-05, + "loss": 0.3377, + "step": 57479 + }, + { + "epoch": 2.21, + "learning_rate": 1.317645550141062e-05, + "loss": 0.1711, + "step": 57480 + }, + { + "epoch": 2.21, + "learning_rate": 1.317581431136189e-05, + "loss": 0.3833, + "step": 57481 + }, + { + "epoch": 2.21, + "learning_rate": 1.3175173121313158e-05, + "loss": 0.3205, + "step": 57482 + }, + { + "epoch": 2.21, + "learning_rate": 1.3174531931264428e-05, + "loss": 0.1975, + "step": 57483 + }, + { + "epoch": 2.21, + "learning_rate": 1.3173890741215696e-05, + "loss": 0.2127, + "step": 57484 + }, + { + "epoch": 2.21, + "learning_rate": 1.3173249551166966e-05, + "loss": 0.2738, + "step": 57485 + }, + { + "epoch": 2.21, + "learning_rate": 1.3172608361118235e-05, + "loss": 0.1668, + "step": 57486 + }, + { + "epoch": 2.21, + "learning_rate": 1.3171967171069505e-05, + "loss": 0.3735, + "step": 57487 + }, + { + "epoch": 2.21, + "learning_rate": 1.3171325981020773e-05, + "loss": 0.3889, + "step": 57488 + }, + { + "epoch": 2.21, + "learning_rate": 1.3170684790972047e-05, + "loss": 0.1669, + "step": 57489 + }, + { + "epoch": 2.21, + "learning_rate": 1.3170043600923315e-05, + "loss": 0.1919, + "step": 57490 + }, + { + "epoch": 2.21, + "learning_rate": 1.3169402410874585e-05, + "loss": 0.2057, + "step": 57491 + }, + { + "epoch": 2.21, + "learning_rate": 1.3168761220825853e-05, + "loss": 0.1462, + "step": 57492 + }, + { + "epoch": 2.21, + "learning_rate": 1.3168120030777123e-05, + "loss": 0.2401, + "step": 57493 + }, + { + "epoch": 2.21, + "learning_rate": 1.3167478840728392e-05, + "loss": 0.2583, + "step": 57494 + }, + { + "epoch": 2.21, + "learning_rate": 1.3166837650679662e-05, + "loss": 0.2404, + "step": 57495 + }, + { + "epoch": 2.21, + "learning_rate": 1.316619646063093e-05, + "loss": 0.2198, + "step": 57496 + }, + { + "epoch": 2.21, + "learning_rate": 1.31655552705822e-05, + "loss": 0.3271, + "step": 57497 + }, + { + "epoch": 2.21, + "learning_rate": 1.3164914080533472e-05, + "loss": 0.1734, + "step": 57498 + }, + { + "epoch": 2.21, + "learning_rate": 1.3164272890484742e-05, + "loss": 0.2263, + "step": 57499 + }, + { + "epoch": 2.21, + "learning_rate": 1.316363170043601e-05, + "loss": 0.3016, + "step": 57500 + }, + { + "epoch": 2.21, + "learning_rate": 1.316299051038728e-05, + "loss": 0.1634, + "step": 57501 + }, + { + "epoch": 2.21, + "learning_rate": 1.3162349320338548e-05, + "loss": 0.1626, + "step": 57502 + }, + { + "epoch": 2.21, + "learning_rate": 1.3161708130289818e-05, + "loss": 0.2417, + "step": 57503 + }, + { + "epoch": 2.21, + "learning_rate": 1.3161066940241087e-05, + "loss": 0.2934, + "step": 57504 + }, + { + "epoch": 2.21, + "learning_rate": 1.3160425750192357e-05, + "loss": 0.099, + "step": 57505 + }, + { + "epoch": 2.21, + "learning_rate": 1.3159784560143625e-05, + "loss": 0.3261, + "step": 57506 + }, + { + "epoch": 2.21, + "learning_rate": 1.3159143370094899e-05, + "loss": 0.1352, + "step": 57507 + }, + { + "epoch": 2.21, + "learning_rate": 1.3158502180046167e-05, + "loss": 0.313, + "step": 57508 + }, + { + "epoch": 2.21, + "learning_rate": 1.3157860989997437e-05, + "loss": 0.3263, + "step": 57509 + }, + { + "epoch": 2.21, + "learning_rate": 1.3157219799948705e-05, + "loss": 0.1833, + "step": 57510 + }, + { + "epoch": 2.21, + "learning_rate": 1.3156578609899975e-05, + "loss": 0.2441, + "step": 57511 + }, + { + "epoch": 2.21, + "learning_rate": 1.3155937419851244e-05, + "loss": 0.307, + "step": 57512 + }, + { + "epoch": 2.21, + "learning_rate": 1.3155296229802514e-05, + "loss": 0.2899, + "step": 57513 + }, + { + "epoch": 2.21, + "learning_rate": 1.3154655039753782e-05, + "loss": 0.2182, + "step": 57514 + }, + { + "epoch": 2.21, + "learning_rate": 1.3154013849705052e-05, + "loss": 0.1801, + "step": 57515 + }, + { + "epoch": 2.21, + "learning_rate": 1.3153372659656324e-05, + "loss": 0.1436, + "step": 57516 + }, + { + "epoch": 2.21, + "learning_rate": 1.3152731469607594e-05, + "loss": 0.2649, + "step": 57517 + }, + { + "epoch": 2.21, + "learning_rate": 1.3152090279558862e-05, + "loss": 0.2298, + "step": 57518 + }, + { + "epoch": 2.21, + "learning_rate": 1.3151449089510132e-05, + "loss": 0.244, + "step": 57519 + }, + { + "epoch": 2.21, + "learning_rate": 1.31508078994614e-05, + "loss": 0.3572, + "step": 57520 + }, + { + "epoch": 2.21, + "learning_rate": 1.315016670941267e-05, + "loss": 0.389, + "step": 57521 + }, + { + "epoch": 2.21, + "learning_rate": 1.3149525519363939e-05, + "loss": 0.1598, + "step": 57522 + }, + { + "epoch": 2.21, + "learning_rate": 1.3148884329315209e-05, + "loss": 0.276, + "step": 57523 + }, + { + "epoch": 2.21, + "learning_rate": 1.3148243139266477e-05, + "loss": 0.4173, + "step": 57524 + }, + { + "epoch": 2.21, + "learning_rate": 1.314760194921775e-05, + "loss": 0.1275, + "step": 57525 + }, + { + "epoch": 2.21, + "learning_rate": 1.3146960759169019e-05, + "loss": 0.2511, + "step": 57526 + }, + { + "epoch": 2.21, + "learning_rate": 1.3146319569120289e-05, + "loss": 0.3573, + "step": 57527 + }, + { + "epoch": 2.21, + "learning_rate": 1.3145678379071557e-05, + "loss": 0.2018, + "step": 57528 + }, + { + "epoch": 2.21, + "learning_rate": 1.3145037189022827e-05, + "loss": 0.1755, + "step": 57529 + }, + { + "epoch": 2.21, + "learning_rate": 1.3144395998974095e-05, + "loss": 0.2823, + "step": 57530 + }, + { + "epoch": 2.21, + "learning_rate": 1.3143754808925366e-05, + "loss": 0.1548, + "step": 57531 + }, + { + "epoch": 2.21, + "learning_rate": 1.3143113618876634e-05, + "loss": 0.2218, + "step": 57532 + }, + { + "epoch": 2.21, + "learning_rate": 1.3142472428827904e-05, + "loss": 0.1917, + "step": 57533 + }, + { + "epoch": 2.21, + "learning_rate": 1.3141831238779176e-05, + "loss": 0.3084, + "step": 57534 + }, + { + "epoch": 2.21, + "learning_rate": 1.3141190048730446e-05, + "loss": 0.2135, + "step": 57535 + }, + { + "epoch": 2.21, + "learning_rate": 1.3140548858681714e-05, + "loss": 0.1547, + "step": 57536 + }, + { + "epoch": 2.21, + "learning_rate": 1.3139907668632984e-05, + "loss": 0.2438, + "step": 57537 + }, + { + "epoch": 2.21, + "learning_rate": 1.3139266478584252e-05, + "loss": 0.2908, + "step": 57538 + }, + { + "epoch": 2.21, + "learning_rate": 1.3138625288535522e-05, + "loss": 0.2226, + "step": 57539 + }, + { + "epoch": 2.21, + "learning_rate": 1.313798409848679e-05, + "loss": 0.2563, + "step": 57540 + }, + { + "epoch": 2.21, + "learning_rate": 1.313734290843806e-05, + "loss": 0.1084, + "step": 57541 + }, + { + "epoch": 2.21, + "learning_rate": 1.3136701718389332e-05, + "loss": 0.2371, + "step": 57542 + }, + { + "epoch": 2.21, + "learning_rate": 1.3136060528340602e-05, + "loss": 0.2571, + "step": 57543 + }, + { + "epoch": 2.21, + "learning_rate": 1.313541933829187e-05, + "loss": 0.1598, + "step": 57544 + }, + { + "epoch": 2.21, + "learning_rate": 1.313477814824314e-05, + "loss": 0.3245, + "step": 57545 + }, + { + "epoch": 2.21, + "learning_rate": 1.3134136958194409e-05, + "loss": 0.2495, + "step": 57546 + }, + { + "epoch": 2.21, + "learning_rate": 1.3133495768145679e-05, + "loss": 0.2128, + "step": 57547 + }, + { + "epoch": 2.21, + "learning_rate": 1.3132854578096947e-05, + "loss": 0.2553, + "step": 57548 + }, + { + "epoch": 2.21, + "learning_rate": 1.3132213388048217e-05, + "loss": 0.2529, + "step": 57549 + }, + { + "epoch": 2.21, + "learning_rate": 1.3131572197999486e-05, + "loss": 0.3434, + "step": 57550 + }, + { + "epoch": 2.21, + "learning_rate": 1.313093100795076e-05, + "loss": 0.3163, + "step": 57551 + }, + { + "epoch": 2.21, + "learning_rate": 1.3130289817902028e-05, + "loss": 0.3328, + "step": 57552 + }, + { + "epoch": 2.21, + "learning_rate": 1.3129648627853298e-05, + "loss": 0.1866, + "step": 57553 + }, + { + "epoch": 2.21, + "learning_rate": 1.3129007437804566e-05, + "loss": 0.2519, + "step": 57554 + }, + { + "epoch": 2.21, + "learning_rate": 1.3128366247755836e-05, + "loss": 0.1929, + "step": 57555 + }, + { + "epoch": 2.21, + "learning_rate": 1.3127725057707104e-05, + "loss": 0.2927, + "step": 57556 + }, + { + "epoch": 2.21, + "learning_rate": 1.3127083867658374e-05, + "loss": 0.2686, + "step": 57557 + }, + { + "epoch": 2.21, + "learning_rate": 1.3126442677609643e-05, + "loss": 0.1479, + "step": 57558 + }, + { + "epoch": 2.21, + "learning_rate": 1.3125801487560913e-05, + "loss": 0.1932, + "step": 57559 + }, + { + "epoch": 2.21, + "learning_rate": 1.3125160297512184e-05, + "loss": 0.2108, + "step": 57560 + }, + { + "epoch": 2.21, + "learning_rate": 1.3124519107463454e-05, + "loss": 0.326, + "step": 57561 + }, + { + "epoch": 2.21, + "learning_rate": 1.3123877917414723e-05, + "loss": 0.2119, + "step": 57562 + }, + { + "epoch": 2.21, + "learning_rate": 1.3123236727365993e-05, + "loss": 0.1781, + "step": 57563 + }, + { + "epoch": 2.21, + "learning_rate": 1.3122595537317261e-05, + "loss": 0.1892, + "step": 57564 + }, + { + "epoch": 2.21, + "learning_rate": 1.3121954347268531e-05, + "loss": 0.1606, + "step": 57565 + }, + { + "epoch": 2.21, + "learning_rate": 1.31213131572198e-05, + "loss": 0.3024, + "step": 57566 + }, + { + "epoch": 2.21, + "learning_rate": 1.312067196717107e-05, + "loss": 0.2434, + "step": 57567 + }, + { + "epoch": 2.21, + "learning_rate": 1.3120030777122338e-05, + "loss": 0.3312, + "step": 57568 + }, + { + "epoch": 2.21, + "learning_rate": 1.3119389587073611e-05, + "loss": 0.1908, + "step": 57569 + }, + { + "epoch": 2.21, + "learning_rate": 1.311874839702488e-05, + "loss": 0.307, + "step": 57570 + }, + { + "epoch": 2.21, + "learning_rate": 1.311810720697615e-05, + "loss": 0.2294, + "step": 57571 + }, + { + "epoch": 2.21, + "learning_rate": 1.3117466016927418e-05, + "loss": 0.2445, + "step": 57572 + }, + { + "epoch": 2.21, + "learning_rate": 1.3116824826878688e-05, + "loss": 0.3605, + "step": 57573 + }, + { + "epoch": 2.21, + "learning_rate": 1.3116183636829956e-05, + "loss": 0.1758, + "step": 57574 + }, + { + "epoch": 2.21, + "learning_rate": 1.3115542446781226e-05, + "loss": 0.2255, + "step": 57575 + }, + { + "epoch": 2.21, + "learning_rate": 1.3114901256732495e-05, + "loss": 0.1621, + "step": 57576 + }, + { + "epoch": 2.21, + "learning_rate": 1.3114260066683765e-05, + "loss": 0.2834, + "step": 57577 + }, + { + "epoch": 2.21, + "learning_rate": 1.3113618876635036e-05, + "loss": 0.1893, + "step": 57578 + }, + { + "epoch": 2.21, + "learning_rate": 1.3112977686586306e-05, + "loss": 0.3066, + "step": 57579 + }, + { + "epoch": 2.21, + "learning_rate": 1.3112336496537575e-05, + "loss": 0.1887, + "step": 57580 + }, + { + "epoch": 2.21, + "learning_rate": 1.3111695306488845e-05, + "loss": 0.328, + "step": 57581 + }, + { + "epoch": 2.21, + "learning_rate": 1.3111054116440113e-05, + "loss": 0.2504, + "step": 57582 + }, + { + "epoch": 2.21, + "learning_rate": 1.3110412926391383e-05, + "loss": 0.1809, + "step": 57583 + }, + { + "epoch": 2.21, + "learning_rate": 1.3109771736342651e-05, + "loss": 0.2038, + "step": 57584 + }, + { + "epoch": 2.21, + "learning_rate": 1.3109130546293921e-05, + "loss": 0.2232, + "step": 57585 + }, + { + "epoch": 2.21, + "learning_rate": 1.310848935624519e-05, + "loss": 0.348, + "step": 57586 + }, + { + "epoch": 2.21, + "learning_rate": 1.3107848166196463e-05, + "loss": 0.2066, + "step": 57587 + }, + { + "epoch": 2.21, + "learning_rate": 1.3107206976147731e-05, + "loss": 0.2625, + "step": 57588 + }, + { + "epoch": 2.21, + "learning_rate": 1.3106565786099001e-05, + "loss": 0.2382, + "step": 57589 + }, + { + "epoch": 2.21, + "learning_rate": 1.310592459605027e-05, + "loss": 0.2505, + "step": 57590 + }, + { + "epoch": 2.21, + "learning_rate": 1.310528340600154e-05, + "loss": 0.2076, + "step": 57591 + }, + { + "epoch": 2.21, + "learning_rate": 1.3104642215952808e-05, + "loss": 0.2066, + "step": 57592 + }, + { + "epoch": 2.21, + "learning_rate": 1.3104001025904078e-05, + "loss": 0.1637, + "step": 57593 + }, + { + "epoch": 2.21, + "learning_rate": 1.3103359835855346e-05, + "loss": 0.2811, + "step": 57594 + }, + { + "epoch": 2.21, + "learning_rate": 1.3102718645806616e-05, + "loss": 0.1718, + "step": 57595 + }, + { + "epoch": 2.21, + "learning_rate": 1.3102077455757888e-05, + "loss": 0.2926, + "step": 57596 + }, + { + "epoch": 2.21, + "learning_rate": 1.3101436265709158e-05, + "loss": 0.3138, + "step": 57597 + }, + { + "epoch": 2.21, + "learning_rate": 1.3100795075660427e-05, + "loss": 0.2941, + "step": 57598 + }, + { + "epoch": 2.21, + "learning_rate": 1.3100153885611697e-05, + "loss": 0.2842, + "step": 57599 + }, + { + "epoch": 2.21, + "learning_rate": 1.3099512695562965e-05, + "loss": 0.1829, + "step": 57600 + }, + { + "epoch": 2.21, + "learning_rate": 1.3098871505514235e-05, + "loss": 0.1541, + "step": 57601 + }, + { + "epoch": 2.21, + "learning_rate": 1.3098230315465503e-05, + "loss": 0.2479, + "step": 57602 + }, + { + "epoch": 2.21, + "learning_rate": 1.3097589125416773e-05, + "loss": 0.1233, + "step": 57603 + }, + { + "epoch": 2.21, + "learning_rate": 1.3096947935368043e-05, + "loss": 0.3446, + "step": 57604 + }, + { + "epoch": 2.21, + "learning_rate": 1.3096306745319315e-05, + "loss": 0.2966, + "step": 57605 + }, + { + "epoch": 2.21, + "learning_rate": 1.3095665555270583e-05, + "loss": 0.1205, + "step": 57606 + }, + { + "epoch": 2.21, + "learning_rate": 1.3095024365221853e-05, + "loss": 0.226, + "step": 57607 + }, + { + "epoch": 2.21, + "learning_rate": 1.3094383175173122e-05, + "loss": 0.3821, + "step": 57608 + }, + { + "epoch": 2.21, + "learning_rate": 1.3093741985124392e-05, + "loss": 0.3066, + "step": 57609 + }, + { + "epoch": 2.21, + "learning_rate": 1.309310079507566e-05, + "loss": 0.1494, + "step": 57610 + }, + { + "epoch": 2.21, + "learning_rate": 1.309245960502693e-05, + "loss": 0.2619, + "step": 57611 + }, + { + "epoch": 2.21, + "learning_rate": 1.3091818414978198e-05, + "loss": 0.31, + "step": 57612 + }, + { + "epoch": 2.22, + "learning_rate": 1.3091177224929468e-05, + "loss": 0.4742, + "step": 57613 + }, + { + "epoch": 2.22, + "learning_rate": 1.309053603488074e-05, + "loss": 0.2411, + "step": 57614 + }, + { + "epoch": 2.22, + "learning_rate": 1.308989484483201e-05, + "loss": 0.1903, + "step": 57615 + }, + { + "epoch": 2.22, + "learning_rate": 1.3089253654783278e-05, + "loss": 0.2083, + "step": 57616 + }, + { + "epoch": 2.22, + "learning_rate": 1.3088612464734548e-05, + "loss": 0.2567, + "step": 57617 + }, + { + "epoch": 2.22, + "learning_rate": 1.3087971274685817e-05, + "loss": 0.2718, + "step": 57618 + }, + { + "epoch": 2.22, + "learning_rate": 1.3087330084637087e-05, + "loss": 0.2377, + "step": 57619 + }, + { + "epoch": 2.22, + "learning_rate": 1.3086688894588355e-05, + "loss": 0.2392, + "step": 57620 + }, + { + "epoch": 2.22, + "learning_rate": 1.3086047704539625e-05, + "loss": 0.2694, + "step": 57621 + }, + { + "epoch": 2.22, + "learning_rate": 1.3085406514490895e-05, + "loss": 0.2636, + "step": 57622 + }, + { + "epoch": 2.22, + "learning_rate": 1.3084765324442167e-05, + "loss": 0.1865, + "step": 57623 + }, + { + "epoch": 2.22, + "learning_rate": 1.3084124134393435e-05, + "loss": 0.2061, + "step": 57624 + }, + { + "epoch": 2.22, + "learning_rate": 1.3083482944344705e-05, + "loss": 0.1847, + "step": 57625 + }, + { + "epoch": 2.22, + "learning_rate": 1.3082841754295974e-05, + "loss": 0.3379, + "step": 57626 + }, + { + "epoch": 2.22, + "learning_rate": 1.3082200564247244e-05, + "loss": 0.3177, + "step": 57627 + }, + { + "epoch": 2.22, + "learning_rate": 1.3081559374198512e-05, + "loss": 0.2114, + "step": 57628 + }, + { + "epoch": 2.22, + "learning_rate": 1.3080918184149782e-05, + "loss": 0.2375, + "step": 57629 + }, + { + "epoch": 2.22, + "learning_rate": 1.308027699410105e-05, + "loss": 0.2088, + "step": 57630 + }, + { + "epoch": 2.22, + "learning_rate": 1.307963580405232e-05, + "loss": 0.217, + "step": 57631 + }, + { + "epoch": 2.22, + "learning_rate": 1.3078994614003592e-05, + "loss": 0.251, + "step": 57632 + }, + { + "epoch": 2.22, + "learning_rate": 1.3078353423954862e-05, + "loss": 0.2349, + "step": 57633 + }, + { + "epoch": 2.22, + "learning_rate": 1.307771223390613e-05, + "loss": 0.1355, + "step": 57634 + }, + { + "epoch": 2.22, + "learning_rate": 1.30770710438574e-05, + "loss": 0.2368, + "step": 57635 + }, + { + "epoch": 2.22, + "learning_rate": 1.3076429853808669e-05, + "loss": 0.3726, + "step": 57636 + }, + { + "epoch": 2.22, + "learning_rate": 1.3075788663759939e-05, + "loss": 0.2159, + "step": 57637 + }, + { + "epoch": 2.22, + "learning_rate": 1.3075147473711207e-05, + "loss": 0.1903, + "step": 57638 + }, + { + "epoch": 2.22, + "learning_rate": 1.3074506283662477e-05, + "loss": 0.2008, + "step": 57639 + }, + { + "epoch": 2.22, + "learning_rate": 1.3073865093613749e-05, + "loss": 0.2862, + "step": 57640 + }, + { + "epoch": 2.22, + "learning_rate": 1.3073223903565019e-05, + "loss": 0.2322, + "step": 57641 + }, + { + "epoch": 2.22, + "learning_rate": 1.3072582713516287e-05, + "loss": 0.28, + "step": 57642 + }, + { + "epoch": 2.22, + "learning_rate": 1.3071941523467557e-05, + "loss": 0.1762, + "step": 57643 + }, + { + "epoch": 2.22, + "learning_rate": 1.3071300333418826e-05, + "loss": 0.3223, + "step": 57644 + }, + { + "epoch": 2.22, + "learning_rate": 1.3070659143370096e-05, + "loss": 0.2131, + "step": 57645 + }, + { + "epoch": 2.22, + "learning_rate": 1.3070017953321364e-05, + "loss": 0.26, + "step": 57646 + }, + { + "epoch": 2.22, + "learning_rate": 1.3069376763272634e-05, + "loss": 0.1106, + "step": 57647 + }, + { + "epoch": 2.22, + "learning_rate": 1.3068735573223902e-05, + "loss": 0.2528, + "step": 57648 + }, + { + "epoch": 2.22, + "learning_rate": 1.3068094383175176e-05, + "loss": 0.1493, + "step": 57649 + }, + { + "epoch": 2.22, + "learning_rate": 1.3067453193126444e-05, + "loss": 0.2235, + "step": 57650 + }, + { + "epoch": 2.22, + "learning_rate": 1.3066812003077714e-05, + "loss": 0.2305, + "step": 57651 + }, + { + "epoch": 2.22, + "learning_rate": 1.3066170813028982e-05, + "loss": 0.1929, + "step": 57652 + }, + { + "epoch": 2.22, + "learning_rate": 1.3065529622980252e-05, + "loss": 0.2185, + "step": 57653 + }, + { + "epoch": 2.22, + "learning_rate": 1.306488843293152e-05, + "loss": 0.2155, + "step": 57654 + }, + { + "epoch": 2.22, + "learning_rate": 1.306424724288279e-05, + "loss": 0.2871, + "step": 57655 + }, + { + "epoch": 2.22, + "learning_rate": 1.3063606052834059e-05, + "loss": 0.1201, + "step": 57656 + }, + { + "epoch": 2.22, + "learning_rate": 1.3062964862785329e-05, + "loss": 0.1633, + "step": 57657 + }, + { + "epoch": 2.22, + "learning_rate": 1.30623236727366e-05, + "loss": 0.2607, + "step": 57658 + }, + { + "epoch": 2.22, + "learning_rate": 1.306168248268787e-05, + "loss": 0.2058, + "step": 57659 + }, + { + "epoch": 2.22, + "learning_rate": 1.3061041292639139e-05, + "loss": 0.4244, + "step": 57660 + }, + { + "epoch": 2.22, + "learning_rate": 1.306040010259041e-05, + "loss": 0.2427, + "step": 57661 + }, + { + "epoch": 2.22, + "learning_rate": 1.3059758912541677e-05, + "loss": 0.2268, + "step": 57662 + }, + { + "epoch": 2.22, + "learning_rate": 1.3059117722492948e-05, + "loss": 0.2562, + "step": 57663 + }, + { + "epoch": 2.22, + "learning_rate": 1.3058476532444216e-05, + "loss": 0.3452, + "step": 57664 + }, + { + "epoch": 2.22, + "learning_rate": 1.3057835342395486e-05, + "loss": 0.1862, + "step": 57665 + }, + { + "epoch": 2.22, + "learning_rate": 1.3057194152346754e-05, + "loss": 0.2038, + "step": 57666 + }, + { + "epoch": 2.22, + "learning_rate": 1.3056552962298028e-05, + "loss": 0.2856, + "step": 57667 + }, + { + "epoch": 2.22, + "learning_rate": 1.3055911772249296e-05, + "loss": 0.1467, + "step": 57668 + }, + { + "epoch": 2.22, + "learning_rate": 1.3055270582200566e-05, + "loss": 0.3084, + "step": 57669 + }, + { + "epoch": 2.22, + "learning_rate": 1.3054629392151834e-05, + "loss": 0.21, + "step": 57670 + }, + { + "epoch": 2.22, + "learning_rate": 1.3053988202103104e-05, + "loss": 0.3899, + "step": 57671 + }, + { + "epoch": 2.22, + "learning_rate": 1.3053347012054373e-05, + "loss": 0.2236, + "step": 57672 + }, + { + "epoch": 2.22, + "learning_rate": 1.3052705822005643e-05, + "loss": 0.3456, + "step": 57673 + }, + { + "epoch": 2.22, + "learning_rate": 1.3052064631956911e-05, + "loss": 0.2316, + "step": 57674 + }, + { + "epoch": 2.22, + "learning_rate": 1.3051423441908181e-05, + "loss": 0.2214, + "step": 57675 + }, + { + "epoch": 2.22, + "learning_rate": 1.3050782251859453e-05, + "loss": 0.2537, + "step": 57676 + }, + { + "epoch": 2.22, + "learning_rate": 1.3050141061810723e-05, + "loss": 0.3114, + "step": 57677 + }, + { + "epoch": 2.22, + "learning_rate": 1.3049499871761991e-05, + "loss": 0.2175, + "step": 57678 + }, + { + "epoch": 2.22, + "learning_rate": 1.3048858681713261e-05, + "loss": 0.1872, + "step": 57679 + }, + { + "epoch": 2.22, + "learning_rate": 1.304821749166453e-05, + "loss": 0.3087, + "step": 57680 + }, + { + "epoch": 2.22, + "learning_rate": 1.30475763016158e-05, + "loss": 0.2301, + "step": 57681 + }, + { + "epoch": 2.22, + "learning_rate": 1.3046935111567068e-05, + "loss": 0.2639, + "step": 57682 + }, + { + "epoch": 2.22, + "learning_rate": 1.3046293921518338e-05, + "loss": 0.1551, + "step": 57683 + }, + { + "epoch": 2.22, + "learning_rate": 1.3045652731469608e-05, + "loss": 0.1908, + "step": 57684 + }, + { + "epoch": 2.22, + "learning_rate": 1.304501154142088e-05, + "loss": 0.246, + "step": 57685 + }, + { + "epoch": 2.22, + "learning_rate": 1.3044370351372148e-05, + "loss": 0.2714, + "step": 57686 + }, + { + "epoch": 2.22, + "learning_rate": 1.3043729161323418e-05, + "loss": 0.173, + "step": 57687 + }, + { + "epoch": 2.22, + "learning_rate": 1.3043087971274686e-05, + "loss": 0.2471, + "step": 57688 + }, + { + "epoch": 2.22, + "learning_rate": 1.3042446781225956e-05, + "loss": 0.2141, + "step": 57689 + }, + { + "epoch": 2.22, + "learning_rate": 1.3041805591177225e-05, + "loss": 0.2181, + "step": 57690 + }, + { + "epoch": 2.22, + "learning_rate": 1.3041164401128495e-05, + "loss": 0.318, + "step": 57691 + }, + { + "epoch": 2.22, + "learning_rate": 1.3040523211079763e-05, + "loss": 0.1743, + "step": 57692 + }, + { + "epoch": 2.22, + "learning_rate": 1.3039882021031033e-05, + "loss": 0.2452, + "step": 57693 + }, + { + "epoch": 2.22, + "learning_rate": 1.3039240830982305e-05, + "loss": 0.1851, + "step": 57694 + }, + { + "epoch": 2.22, + "learning_rate": 1.3038599640933575e-05, + "loss": 0.1937, + "step": 57695 + }, + { + "epoch": 2.22, + "learning_rate": 1.3037958450884843e-05, + "loss": 0.2555, + "step": 57696 + }, + { + "epoch": 2.22, + "learning_rate": 1.3037317260836113e-05, + "loss": 0.2574, + "step": 57697 + }, + { + "epoch": 2.22, + "learning_rate": 1.3036676070787381e-05, + "loss": 0.2385, + "step": 57698 + }, + { + "epoch": 2.22, + "learning_rate": 1.3036034880738651e-05, + "loss": 0.1294, + "step": 57699 + }, + { + "epoch": 2.22, + "learning_rate": 1.303539369068992e-05, + "loss": 0.1402, + "step": 57700 + }, + { + "epoch": 2.22, + "learning_rate": 1.303475250064119e-05, + "loss": 0.161, + "step": 57701 + }, + { + "epoch": 2.22, + "learning_rate": 1.303411131059246e-05, + "loss": 0.2563, + "step": 57702 + }, + { + "epoch": 2.22, + "learning_rate": 1.3033470120543731e-05, + "loss": 0.134, + "step": 57703 + }, + { + "epoch": 2.22, + "learning_rate": 1.3032828930495e-05, + "loss": 0.2274, + "step": 57704 + }, + { + "epoch": 2.22, + "learning_rate": 1.303218774044627e-05, + "loss": 0.3447, + "step": 57705 + }, + { + "epoch": 2.22, + "learning_rate": 1.3031546550397538e-05, + "loss": 0.217, + "step": 57706 + }, + { + "epoch": 2.22, + "learning_rate": 1.3030905360348808e-05, + "loss": 0.2491, + "step": 57707 + }, + { + "epoch": 2.22, + "learning_rate": 1.3030264170300076e-05, + "loss": 0.1368, + "step": 57708 + }, + { + "epoch": 2.22, + "learning_rate": 1.3029622980251347e-05, + "loss": 0.2045, + "step": 57709 + }, + { + "epoch": 2.22, + "learning_rate": 1.3028981790202615e-05, + "loss": 0.1894, + "step": 57710 + }, + { + "epoch": 2.22, + "learning_rate": 1.3028340600153885e-05, + "loss": 0.3468, + "step": 57711 + }, + { + "epoch": 2.22, + "learning_rate": 1.3027699410105157e-05, + "loss": 0.2925, + "step": 57712 + }, + { + "epoch": 2.22, + "learning_rate": 1.3027058220056427e-05, + "loss": 0.1973, + "step": 57713 + }, + { + "epoch": 2.22, + "learning_rate": 1.3026417030007695e-05, + "loss": 0.1057, + "step": 57714 + }, + { + "epoch": 2.22, + "learning_rate": 1.3025775839958965e-05, + "loss": 0.2114, + "step": 57715 + }, + { + "epoch": 2.22, + "learning_rate": 1.3025134649910233e-05, + "loss": 0.2592, + "step": 57716 + }, + { + "epoch": 2.22, + "learning_rate": 1.3024493459861503e-05, + "loss": 0.3378, + "step": 57717 + }, + { + "epoch": 2.22, + "learning_rate": 1.3023852269812772e-05, + "loss": 0.2673, + "step": 57718 + }, + { + "epoch": 2.22, + "learning_rate": 1.3023211079764042e-05, + "loss": 0.2603, + "step": 57719 + }, + { + "epoch": 2.22, + "learning_rate": 1.3022569889715312e-05, + "loss": 0.2022, + "step": 57720 + }, + { + "epoch": 2.22, + "learning_rate": 1.3021928699666583e-05, + "loss": 0.2354, + "step": 57721 + }, + { + "epoch": 2.22, + "learning_rate": 1.3021287509617852e-05, + "loss": 0.1823, + "step": 57722 + }, + { + "epoch": 2.22, + "learning_rate": 1.3020646319569122e-05, + "loss": 0.1965, + "step": 57723 + }, + { + "epoch": 2.22, + "learning_rate": 1.302000512952039e-05, + "loss": 0.2191, + "step": 57724 + }, + { + "epoch": 2.22, + "learning_rate": 1.301936393947166e-05, + "loss": 0.2326, + "step": 57725 + }, + { + "epoch": 2.22, + "learning_rate": 1.3018722749422928e-05, + "loss": 0.2977, + "step": 57726 + }, + { + "epoch": 2.22, + "learning_rate": 1.3018081559374198e-05, + "loss": 0.2895, + "step": 57727 + }, + { + "epoch": 2.22, + "learning_rate": 1.3017440369325467e-05, + "loss": 0.3105, + "step": 57728 + }, + { + "epoch": 2.22, + "learning_rate": 1.3016799179276737e-05, + "loss": 0.2606, + "step": 57729 + }, + { + "epoch": 2.22, + "learning_rate": 1.3016157989228009e-05, + "loss": 0.226, + "step": 57730 + }, + { + "epoch": 2.22, + "learning_rate": 1.3015516799179279e-05, + "loss": 0.1357, + "step": 57731 + }, + { + "epoch": 2.22, + "learning_rate": 1.3014875609130547e-05, + "loss": 0.2047, + "step": 57732 + }, + { + "epoch": 2.22, + "learning_rate": 1.3014234419081817e-05, + "loss": 0.2324, + "step": 57733 + }, + { + "epoch": 2.22, + "learning_rate": 1.3013593229033085e-05, + "loss": 0.2704, + "step": 57734 + }, + { + "epoch": 2.22, + "learning_rate": 1.3012952038984355e-05, + "loss": 0.3223, + "step": 57735 + }, + { + "epoch": 2.22, + "learning_rate": 1.3012310848935624e-05, + "loss": 0.1791, + "step": 57736 + }, + { + "epoch": 2.22, + "learning_rate": 1.3011669658886894e-05, + "loss": 0.2411, + "step": 57737 + }, + { + "epoch": 2.22, + "learning_rate": 1.3011028468838165e-05, + "loss": 0.1777, + "step": 57738 + }, + { + "epoch": 2.22, + "learning_rate": 1.3010387278789435e-05, + "loss": 0.2636, + "step": 57739 + }, + { + "epoch": 2.22, + "learning_rate": 1.3009746088740704e-05, + "loss": 0.2558, + "step": 57740 + }, + { + "epoch": 2.22, + "learning_rate": 1.3009104898691974e-05, + "loss": 0.3412, + "step": 57741 + }, + { + "epoch": 2.22, + "learning_rate": 1.3008463708643242e-05, + "loss": 0.1944, + "step": 57742 + }, + { + "epoch": 2.22, + "learning_rate": 1.3007822518594512e-05, + "loss": 0.2767, + "step": 57743 + }, + { + "epoch": 2.22, + "learning_rate": 1.300718132854578e-05, + "loss": 0.3061, + "step": 57744 + }, + { + "epoch": 2.22, + "learning_rate": 1.300654013849705e-05, + "loss": 0.1929, + "step": 57745 + }, + { + "epoch": 2.22, + "learning_rate": 1.300589894844832e-05, + "loss": 0.2463, + "step": 57746 + }, + { + "epoch": 2.22, + "learning_rate": 1.3005257758399592e-05, + "loss": 0.3284, + "step": 57747 + }, + { + "epoch": 2.22, + "learning_rate": 1.300461656835086e-05, + "loss": 0.2304, + "step": 57748 + }, + { + "epoch": 2.22, + "learning_rate": 1.300397537830213e-05, + "loss": 0.1842, + "step": 57749 + }, + { + "epoch": 2.22, + "learning_rate": 1.3003334188253399e-05, + "loss": 0.1743, + "step": 57750 + }, + { + "epoch": 2.22, + "learning_rate": 1.3002692998204669e-05, + "loss": 0.3328, + "step": 57751 + }, + { + "epoch": 2.22, + "learning_rate": 1.3002051808155937e-05, + "loss": 0.1921, + "step": 57752 + }, + { + "epoch": 2.22, + "learning_rate": 1.3001410618107207e-05, + "loss": 0.3216, + "step": 57753 + }, + { + "epoch": 2.22, + "learning_rate": 1.3000769428058476e-05, + "loss": 0.2495, + "step": 57754 + }, + { + "epoch": 2.22, + "learning_rate": 1.3000128238009746e-05, + "loss": 0.2572, + "step": 57755 + }, + { + "epoch": 2.22, + "learning_rate": 1.2999487047961017e-05, + "loss": 0.2673, + "step": 57756 + }, + { + "epoch": 2.22, + "learning_rate": 1.2998845857912287e-05, + "loss": 0.1779, + "step": 57757 + }, + { + "epoch": 2.22, + "learning_rate": 1.2998204667863556e-05, + "loss": 0.3691, + "step": 57758 + }, + { + "epoch": 2.22, + "learning_rate": 1.2997563477814826e-05, + "loss": 0.1682, + "step": 57759 + }, + { + "epoch": 2.22, + "learning_rate": 1.2996922287766094e-05, + "loss": 0.2511, + "step": 57760 + }, + { + "epoch": 2.22, + "learning_rate": 1.2996281097717364e-05, + "loss": 0.1706, + "step": 57761 + }, + { + "epoch": 2.22, + "learning_rate": 1.2995639907668632e-05, + "loss": 0.2222, + "step": 57762 + }, + { + "epoch": 2.22, + "learning_rate": 1.2994998717619902e-05, + "loss": 0.3354, + "step": 57763 + }, + { + "epoch": 2.22, + "learning_rate": 1.2994357527571172e-05, + "loss": 0.1683, + "step": 57764 + }, + { + "epoch": 2.22, + "learning_rate": 1.2993716337522444e-05, + "loss": 0.2415, + "step": 57765 + }, + { + "epoch": 2.22, + "learning_rate": 1.2993075147473712e-05, + "loss": 0.1787, + "step": 57766 + }, + { + "epoch": 2.22, + "learning_rate": 1.2992433957424982e-05, + "loss": 0.1567, + "step": 57767 + }, + { + "epoch": 2.22, + "learning_rate": 1.299179276737625e-05, + "loss": 0.2037, + "step": 57768 + }, + { + "epoch": 2.22, + "learning_rate": 1.299115157732752e-05, + "loss": 0.3167, + "step": 57769 + }, + { + "epoch": 2.22, + "learning_rate": 1.2990510387278789e-05, + "loss": 0.1758, + "step": 57770 + }, + { + "epoch": 2.22, + "learning_rate": 1.2989869197230059e-05, + "loss": 0.1784, + "step": 57771 + }, + { + "epoch": 2.22, + "learning_rate": 1.2989228007181327e-05, + "loss": 0.2469, + "step": 57772 + }, + { + "epoch": 2.22, + "learning_rate": 1.2988586817132597e-05, + "loss": 0.2274, + "step": 57773 + }, + { + "epoch": 2.22, + "learning_rate": 1.298794562708387e-05, + "loss": 0.2364, + "step": 57774 + }, + { + "epoch": 2.22, + "learning_rate": 1.298730443703514e-05, + "loss": 0.2804, + "step": 57775 + }, + { + "epoch": 2.22, + "learning_rate": 1.2986663246986408e-05, + "loss": 0.1666, + "step": 57776 + }, + { + "epoch": 2.22, + "learning_rate": 1.2986022056937678e-05, + "loss": 0.3184, + "step": 57777 + }, + { + "epoch": 2.22, + "learning_rate": 1.2985380866888946e-05, + "loss": 0.2703, + "step": 57778 + }, + { + "epoch": 2.22, + "learning_rate": 1.2984739676840216e-05, + "loss": 0.2295, + "step": 57779 + }, + { + "epoch": 2.22, + "learning_rate": 1.2984098486791484e-05, + "loss": 0.2406, + "step": 57780 + }, + { + "epoch": 2.22, + "learning_rate": 1.2983457296742754e-05, + "loss": 0.2285, + "step": 57781 + }, + { + "epoch": 2.22, + "learning_rate": 1.2982816106694024e-05, + "loss": 0.2693, + "step": 57782 + }, + { + "epoch": 2.22, + "learning_rate": 1.2982174916645296e-05, + "loss": 0.1871, + "step": 57783 + }, + { + "epoch": 2.22, + "learning_rate": 1.2981533726596564e-05, + "loss": 0.2181, + "step": 57784 + }, + { + "epoch": 2.22, + "learning_rate": 1.2980892536547834e-05, + "loss": 0.2864, + "step": 57785 + }, + { + "epoch": 2.22, + "learning_rate": 1.2980251346499103e-05, + "loss": 0.1384, + "step": 57786 + }, + { + "epoch": 2.22, + "learning_rate": 1.2979610156450373e-05, + "loss": 0.2139, + "step": 57787 + }, + { + "epoch": 2.22, + "learning_rate": 1.2978968966401641e-05, + "loss": 0.2105, + "step": 57788 + }, + { + "epoch": 2.22, + "learning_rate": 1.2978327776352911e-05, + "loss": 0.2867, + "step": 57789 + }, + { + "epoch": 2.22, + "learning_rate": 1.297768658630418e-05, + "loss": 0.3423, + "step": 57790 + }, + { + "epoch": 2.22, + "learning_rate": 1.297704539625545e-05, + "loss": 0.1795, + "step": 57791 + }, + { + "epoch": 2.22, + "learning_rate": 1.2976404206206721e-05, + "loss": 0.2351, + "step": 57792 + }, + { + "epoch": 2.22, + "learning_rate": 1.2975763016157991e-05, + "loss": 0.118, + "step": 57793 + }, + { + "epoch": 2.22, + "learning_rate": 1.297512182610926e-05, + "loss": 0.182, + "step": 57794 + }, + { + "epoch": 2.22, + "learning_rate": 1.297448063606053e-05, + "loss": 0.3852, + "step": 57795 + }, + { + "epoch": 2.22, + "learning_rate": 1.2973839446011798e-05, + "loss": 0.2855, + "step": 57796 + }, + { + "epoch": 2.22, + "learning_rate": 1.2973198255963068e-05, + "loss": 0.1314, + "step": 57797 + }, + { + "epoch": 2.22, + "learning_rate": 1.2972557065914336e-05, + "loss": 0.2076, + "step": 57798 + }, + { + "epoch": 2.22, + "learning_rate": 1.2971915875865606e-05, + "loss": 0.2715, + "step": 57799 + }, + { + "epoch": 2.22, + "learning_rate": 1.2971274685816876e-05, + "loss": 0.2272, + "step": 57800 + }, + { + "epoch": 2.22, + "learning_rate": 1.2970633495768148e-05, + "loss": 0.2122, + "step": 57801 + }, + { + "epoch": 2.22, + "learning_rate": 1.2969992305719416e-05, + "loss": 0.3922, + "step": 57802 + }, + { + "epoch": 2.22, + "learning_rate": 1.2969351115670686e-05, + "loss": 0.2603, + "step": 57803 + }, + { + "epoch": 2.22, + "learning_rate": 1.2968709925621955e-05, + "loss": 0.2216, + "step": 57804 + }, + { + "epoch": 2.22, + "learning_rate": 1.2968068735573225e-05, + "loss": 0.2326, + "step": 57805 + }, + { + "epoch": 2.22, + "learning_rate": 1.2967427545524493e-05, + "loss": 0.2009, + "step": 57806 + }, + { + "epoch": 2.22, + "learning_rate": 1.2966786355475763e-05, + "loss": 0.1767, + "step": 57807 + }, + { + "epoch": 2.22, + "learning_rate": 1.2966145165427033e-05, + "loss": 0.2224, + "step": 57808 + }, + { + "epoch": 2.22, + "learning_rate": 1.2965503975378301e-05, + "loss": 0.3422, + "step": 57809 + }, + { + "epoch": 2.22, + "learning_rate": 1.2964862785329573e-05, + "loss": 0.118, + "step": 57810 + }, + { + "epoch": 2.22, + "learning_rate": 1.2964221595280843e-05, + "loss": 0.2784, + "step": 57811 + }, + { + "epoch": 2.22, + "learning_rate": 1.2963580405232111e-05, + "loss": 0.2237, + "step": 57812 + }, + { + "epoch": 2.22, + "learning_rate": 1.2962939215183381e-05, + "loss": 0.2541, + "step": 57813 + }, + { + "epoch": 2.22, + "learning_rate": 1.296229802513465e-05, + "loss": 0.2937, + "step": 57814 + }, + { + "epoch": 2.22, + "learning_rate": 1.296165683508592e-05, + "loss": 0.1482, + "step": 57815 + }, + { + "epoch": 2.22, + "learning_rate": 1.2961015645037188e-05, + "loss": 0.2084, + "step": 57816 + }, + { + "epoch": 2.22, + "learning_rate": 1.2960374454988458e-05, + "loss": 0.2378, + "step": 57817 + }, + { + "epoch": 2.22, + "learning_rate": 1.2959733264939728e-05, + "loss": 0.2502, + "step": 57818 + }, + { + "epoch": 2.22, + "learning_rate": 1.2959092074891e-05, + "loss": 0.3562, + "step": 57819 + }, + { + "epoch": 2.22, + "learning_rate": 1.2958450884842268e-05, + "loss": 0.3407, + "step": 57820 + }, + { + "epoch": 2.22, + "learning_rate": 1.2957809694793538e-05, + "loss": 0.2366, + "step": 57821 + }, + { + "epoch": 2.22, + "learning_rate": 1.2957168504744807e-05, + "loss": 0.1639, + "step": 57822 + }, + { + "epoch": 2.22, + "learning_rate": 1.2956527314696077e-05, + "loss": 0.3027, + "step": 57823 + }, + { + "epoch": 2.22, + "learning_rate": 1.2955886124647345e-05, + "loss": 0.1728, + "step": 57824 + }, + { + "epoch": 2.22, + "learning_rate": 1.2955244934598615e-05, + "loss": 0.2307, + "step": 57825 + }, + { + "epoch": 2.22, + "learning_rate": 1.2954603744549885e-05, + "loss": 0.333, + "step": 57826 + }, + { + "epoch": 2.22, + "learning_rate": 1.2953962554501157e-05, + "loss": 0.2398, + "step": 57827 + }, + { + "epoch": 2.22, + "learning_rate": 1.2953321364452425e-05, + "loss": 0.301, + "step": 57828 + }, + { + "epoch": 2.22, + "learning_rate": 1.2952680174403695e-05, + "loss": 0.3051, + "step": 57829 + }, + { + "epoch": 2.22, + "learning_rate": 1.2952038984354963e-05, + "loss": 0.3156, + "step": 57830 + }, + { + "epoch": 2.22, + "learning_rate": 1.2951397794306233e-05, + "loss": 0.2366, + "step": 57831 + }, + { + "epoch": 2.22, + "learning_rate": 1.2950756604257502e-05, + "loss": 0.2372, + "step": 57832 + }, + { + "epoch": 2.22, + "learning_rate": 1.2950115414208772e-05, + "loss": 0.2391, + "step": 57833 + }, + { + "epoch": 2.22, + "learning_rate": 1.294947422416004e-05, + "loss": 0.1827, + "step": 57834 + }, + { + "epoch": 2.22, + "learning_rate": 1.294883303411131e-05, + "loss": 0.2351, + "step": 57835 + }, + { + "epoch": 2.22, + "learning_rate": 1.2948191844062582e-05, + "loss": 0.2272, + "step": 57836 + }, + { + "epoch": 2.22, + "learning_rate": 1.2947550654013852e-05, + "loss": 0.2926, + "step": 57837 + }, + { + "epoch": 2.22, + "learning_rate": 1.294690946396512e-05, + "loss": 0.2603, + "step": 57838 + }, + { + "epoch": 2.22, + "learning_rate": 1.294626827391639e-05, + "loss": 0.2014, + "step": 57839 + }, + { + "epoch": 2.22, + "learning_rate": 1.2945627083867658e-05, + "loss": 0.2041, + "step": 57840 + }, + { + "epoch": 2.22, + "learning_rate": 1.2944985893818929e-05, + "loss": 0.2529, + "step": 57841 + }, + { + "epoch": 2.22, + "learning_rate": 1.2944344703770197e-05, + "loss": 0.2567, + "step": 57842 + }, + { + "epoch": 2.22, + "learning_rate": 1.2943703513721467e-05, + "loss": 0.2227, + "step": 57843 + }, + { + "epoch": 2.22, + "learning_rate": 1.2943062323672737e-05, + "loss": 0.1814, + "step": 57844 + }, + { + "epoch": 2.22, + "learning_rate": 1.2942421133624009e-05, + "loss": 0.2, + "step": 57845 + }, + { + "epoch": 2.22, + "learning_rate": 1.2941779943575277e-05, + "loss": 0.1383, + "step": 57846 + }, + { + "epoch": 2.22, + "learning_rate": 1.2941138753526547e-05, + "loss": 0.2677, + "step": 57847 + }, + { + "epoch": 2.22, + "learning_rate": 1.2940497563477815e-05, + "loss": 0.1735, + "step": 57848 + }, + { + "epoch": 2.22, + "learning_rate": 1.2939856373429085e-05, + "loss": 0.2798, + "step": 57849 + }, + { + "epoch": 2.22, + "learning_rate": 1.2939215183380354e-05, + "loss": 0.155, + "step": 57850 + }, + { + "epoch": 2.22, + "learning_rate": 1.2938573993331624e-05, + "loss": 0.2858, + "step": 57851 + }, + { + "epoch": 2.22, + "learning_rate": 1.2937932803282892e-05, + "loss": 0.1527, + "step": 57852 + }, + { + "epoch": 2.22, + "learning_rate": 1.2937291613234162e-05, + "loss": 0.2501, + "step": 57853 + }, + { + "epoch": 2.22, + "learning_rate": 1.2936650423185434e-05, + "loss": 0.2619, + "step": 57854 + }, + { + "epoch": 2.22, + "learning_rate": 1.2936009233136704e-05, + "loss": 0.2898, + "step": 57855 + }, + { + "epoch": 2.22, + "learning_rate": 1.2935368043087972e-05, + "loss": 0.2268, + "step": 57856 + }, + { + "epoch": 2.22, + "learning_rate": 1.2934726853039242e-05, + "loss": 0.2996, + "step": 57857 + }, + { + "epoch": 2.22, + "learning_rate": 1.293408566299051e-05, + "loss": 0.2266, + "step": 57858 + }, + { + "epoch": 2.22, + "learning_rate": 1.293344447294178e-05, + "loss": 0.1856, + "step": 57859 + }, + { + "epoch": 2.22, + "learning_rate": 1.2932803282893049e-05, + "loss": 0.1376, + "step": 57860 + }, + { + "epoch": 2.22, + "learning_rate": 1.2932162092844319e-05, + "loss": 0.221, + "step": 57861 + }, + { + "epoch": 2.22, + "learning_rate": 1.2931520902795589e-05, + "loss": 0.4427, + "step": 57862 + }, + { + "epoch": 2.22, + "learning_rate": 1.293087971274686e-05, + "loss": 0.193, + "step": 57863 + }, + { + "epoch": 2.22, + "learning_rate": 1.2930238522698129e-05, + "loss": 0.1837, + "step": 57864 + }, + { + "epoch": 2.22, + "learning_rate": 1.2929597332649399e-05, + "loss": 0.1519, + "step": 57865 + }, + { + "epoch": 2.22, + "learning_rate": 1.2928956142600667e-05, + "loss": 0.1932, + "step": 57866 + }, + { + "epoch": 2.22, + "learning_rate": 1.2928314952551937e-05, + "loss": 0.1461, + "step": 57867 + }, + { + "epoch": 2.22, + "learning_rate": 1.2927673762503206e-05, + "loss": 0.1998, + "step": 57868 + }, + { + "epoch": 2.22, + "learning_rate": 1.2927032572454476e-05, + "loss": 0.2212, + "step": 57869 + }, + { + "epoch": 2.22, + "learning_rate": 1.2926391382405746e-05, + "loss": 0.1729, + "step": 57870 + }, + { + "epoch": 2.22, + "learning_rate": 1.2925750192357014e-05, + "loss": 0.1367, + "step": 57871 + }, + { + "epoch": 2.22, + "learning_rate": 1.2925109002308286e-05, + "loss": 0.2899, + "step": 57872 + }, + { + "epoch": 2.23, + "learning_rate": 1.2924467812259556e-05, + "loss": 0.2203, + "step": 57873 + }, + { + "epoch": 2.23, + "learning_rate": 1.2923826622210824e-05, + "loss": 0.2198, + "step": 57874 + }, + { + "epoch": 2.23, + "learning_rate": 1.2923185432162094e-05, + "loss": 0.3489, + "step": 57875 + }, + { + "epoch": 2.23, + "learning_rate": 1.2922544242113362e-05, + "loss": 0.4576, + "step": 57876 + }, + { + "epoch": 2.23, + "learning_rate": 1.2921903052064632e-05, + "loss": 0.3855, + "step": 57877 + }, + { + "epoch": 2.23, + "learning_rate": 1.29212618620159e-05, + "loss": 0.1569, + "step": 57878 + }, + { + "epoch": 2.23, + "learning_rate": 1.292062067196717e-05, + "loss": 0.3059, + "step": 57879 + }, + { + "epoch": 2.23, + "learning_rate": 1.291997948191844e-05, + "loss": 0.2577, + "step": 57880 + }, + { + "epoch": 2.23, + "learning_rate": 1.2919338291869712e-05, + "loss": 0.2749, + "step": 57881 + }, + { + "epoch": 2.23, + "learning_rate": 1.291869710182098e-05, + "loss": 0.3088, + "step": 57882 + }, + { + "epoch": 2.23, + "learning_rate": 1.291805591177225e-05, + "loss": 0.1935, + "step": 57883 + }, + { + "epoch": 2.23, + "learning_rate": 1.2917414721723519e-05, + "loss": 0.2021, + "step": 57884 + }, + { + "epoch": 2.23, + "learning_rate": 1.291677353167479e-05, + "loss": 0.2499, + "step": 57885 + }, + { + "epoch": 2.23, + "learning_rate": 1.2916132341626058e-05, + "loss": 0.2633, + "step": 57886 + }, + { + "epoch": 2.23, + "learning_rate": 1.2915491151577328e-05, + "loss": 0.1871, + "step": 57887 + }, + { + "epoch": 2.23, + "learning_rate": 1.2914849961528598e-05, + "loss": 0.2933, + "step": 57888 + }, + { + "epoch": 2.23, + "learning_rate": 1.2914208771479866e-05, + "loss": 0.3054, + "step": 57889 + }, + { + "epoch": 2.23, + "learning_rate": 1.2913567581431138e-05, + "loss": 0.139, + "step": 57890 + }, + { + "epoch": 2.23, + "learning_rate": 1.2912926391382408e-05, + "loss": 0.124, + "step": 57891 + }, + { + "epoch": 2.23, + "learning_rate": 1.2912285201333676e-05, + "loss": 0.2807, + "step": 57892 + }, + { + "epoch": 2.23, + "learning_rate": 1.2911644011284946e-05, + "loss": 0.207, + "step": 57893 + }, + { + "epoch": 2.23, + "learning_rate": 1.2911002821236214e-05, + "loss": 0.3343, + "step": 57894 + }, + { + "epoch": 2.23, + "learning_rate": 1.2910361631187484e-05, + "loss": 0.1929, + "step": 57895 + }, + { + "epoch": 2.23, + "learning_rate": 1.2909720441138753e-05, + "loss": 0.2763, + "step": 57896 + }, + { + "epoch": 2.23, + "learning_rate": 1.2909079251090023e-05, + "loss": 0.3593, + "step": 57897 + }, + { + "epoch": 2.23, + "learning_rate": 1.2908438061041293e-05, + "loss": 0.4406, + "step": 57898 + }, + { + "epoch": 2.23, + "learning_rate": 1.2907796870992564e-05, + "loss": 0.1855, + "step": 57899 + }, + { + "epoch": 2.23, + "learning_rate": 1.2907155680943833e-05, + "loss": 0.2524, + "step": 57900 + }, + { + "epoch": 2.23, + "learning_rate": 1.2906514490895103e-05, + "loss": 0.3098, + "step": 57901 + }, + { + "epoch": 2.23, + "learning_rate": 1.2905873300846371e-05, + "loss": 0.3031, + "step": 57902 + }, + { + "epoch": 2.23, + "learning_rate": 1.2905232110797641e-05, + "loss": 0.2333, + "step": 57903 + }, + { + "epoch": 2.23, + "learning_rate": 1.290459092074891e-05, + "loss": 0.1713, + "step": 57904 + }, + { + "epoch": 2.23, + "learning_rate": 1.290394973070018e-05, + "loss": 0.1364, + "step": 57905 + }, + { + "epoch": 2.23, + "learning_rate": 1.290330854065145e-05, + "loss": 0.2758, + "step": 57906 + }, + { + "epoch": 2.23, + "learning_rate": 1.2902667350602718e-05, + "loss": 0.2352, + "step": 57907 + }, + { + "epoch": 2.23, + "learning_rate": 1.290202616055399e-05, + "loss": 0.2145, + "step": 57908 + }, + { + "epoch": 2.23, + "learning_rate": 1.290138497050526e-05, + "loss": 0.0979, + "step": 57909 + }, + { + "epoch": 2.23, + "learning_rate": 1.2900743780456528e-05, + "loss": 0.2034, + "step": 57910 + }, + { + "epoch": 2.23, + "learning_rate": 1.2900102590407798e-05, + "loss": 0.3684, + "step": 57911 + }, + { + "epoch": 2.23, + "learning_rate": 1.2899461400359066e-05, + "loss": 0.1504, + "step": 57912 + }, + { + "epoch": 2.23, + "learning_rate": 1.2898820210310336e-05, + "loss": 0.2493, + "step": 57913 + }, + { + "epoch": 2.23, + "learning_rate": 1.2898179020261605e-05, + "loss": 0.1899, + "step": 57914 + }, + { + "epoch": 2.23, + "learning_rate": 1.2897537830212875e-05, + "loss": 0.1413, + "step": 57915 + }, + { + "epoch": 2.23, + "learning_rate": 1.2896896640164145e-05, + "loss": 0.1761, + "step": 57916 + }, + { + "epoch": 2.23, + "learning_rate": 1.2896255450115416e-05, + "loss": 0.1977, + "step": 57917 + }, + { + "epoch": 2.23, + "learning_rate": 1.2895614260066685e-05, + "loss": 0.1617, + "step": 57918 + }, + { + "epoch": 2.23, + "learning_rate": 1.2894973070017955e-05, + "loss": 0.1566, + "step": 57919 + }, + { + "epoch": 2.23, + "learning_rate": 1.2894331879969223e-05, + "loss": 0.1465, + "step": 57920 + }, + { + "epoch": 2.23, + "learning_rate": 1.2893690689920493e-05, + "loss": 0.4564, + "step": 57921 + }, + { + "epoch": 2.23, + "learning_rate": 1.2893049499871761e-05, + "loss": 0.2487, + "step": 57922 + }, + { + "epoch": 2.23, + "learning_rate": 1.2892408309823031e-05, + "loss": 0.1333, + "step": 57923 + }, + { + "epoch": 2.23, + "learning_rate": 1.2891767119774301e-05, + "loss": 0.2876, + "step": 57924 + }, + { + "epoch": 2.23, + "learning_rate": 1.2891125929725573e-05, + "loss": 0.2284, + "step": 57925 + }, + { + "epoch": 2.23, + "learning_rate": 1.2890484739676841e-05, + "loss": 0.2254, + "step": 57926 + }, + { + "epoch": 2.23, + "learning_rate": 1.2889843549628111e-05, + "loss": 0.1731, + "step": 57927 + }, + { + "epoch": 2.23, + "learning_rate": 1.288920235957938e-05, + "loss": 0.2112, + "step": 57928 + }, + { + "epoch": 2.23, + "learning_rate": 1.288856116953065e-05, + "loss": 0.2102, + "step": 57929 + }, + { + "epoch": 2.23, + "learning_rate": 1.2887919979481918e-05, + "loss": 0.2403, + "step": 57930 + }, + { + "epoch": 2.23, + "learning_rate": 1.2887278789433188e-05, + "loss": 0.2782, + "step": 57931 + }, + { + "epoch": 2.23, + "learning_rate": 1.2886637599384457e-05, + "loss": 0.347, + "step": 57932 + }, + { + "epoch": 2.23, + "learning_rate": 1.2885996409335727e-05, + "loss": 0.2228, + "step": 57933 + }, + { + "epoch": 2.23, + "learning_rate": 1.2885355219286998e-05, + "loss": 0.1802, + "step": 57934 + }, + { + "epoch": 2.23, + "learning_rate": 1.2884714029238268e-05, + "loss": 0.2323, + "step": 57935 + }, + { + "epoch": 2.23, + "learning_rate": 1.2884072839189537e-05, + "loss": 0.331, + "step": 57936 + }, + { + "epoch": 2.23, + "learning_rate": 1.2883431649140807e-05, + "loss": 0.3516, + "step": 57937 + }, + { + "epoch": 2.23, + "learning_rate": 1.2882790459092075e-05, + "loss": 0.2297, + "step": 57938 + }, + { + "epoch": 2.23, + "learning_rate": 1.2882149269043345e-05, + "loss": 0.1681, + "step": 57939 + }, + { + "epoch": 2.23, + "learning_rate": 1.2881508078994613e-05, + "loss": 0.3449, + "step": 57940 + }, + { + "epoch": 2.23, + "learning_rate": 1.2880866888945883e-05, + "loss": 0.3077, + "step": 57941 + }, + { + "epoch": 2.23, + "learning_rate": 1.2880225698897153e-05, + "loss": 0.2111, + "step": 57942 + }, + { + "epoch": 2.23, + "learning_rate": 1.2879584508848425e-05, + "loss": 0.2987, + "step": 57943 + }, + { + "epoch": 2.23, + "learning_rate": 1.2878943318799693e-05, + "loss": 0.3094, + "step": 57944 + }, + { + "epoch": 2.23, + "learning_rate": 1.2878302128750963e-05, + "loss": 0.1061, + "step": 57945 + }, + { + "epoch": 2.23, + "learning_rate": 1.2877660938702232e-05, + "loss": 0.2933, + "step": 57946 + }, + { + "epoch": 2.23, + "learning_rate": 1.2877019748653502e-05, + "loss": 0.1736, + "step": 57947 + }, + { + "epoch": 2.23, + "learning_rate": 1.287637855860477e-05, + "loss": 0.2104, + "step": 57948 + }, + { + "epoch": 2.23, + "learning_rate": 1.287573736855604e-05, + "loss": 0.1757, + "step": 57949 + }, + { + "epoch": 2.23, + "learning_rate": 1.287509617850731e-05, + "loss": 0.2198, + "step": 57950 + }, + { + "epoch": 2.23, + "learning_rate": 1.2874454988458578e-05, + "loss": 0.2154, + "step": 57951 + }, + { + "epoch": 2.23, + "learning_rate": 1.287381379840985e-05, + "loss": 0.2198, + "step": 57952 + }, + { + "epoch": 2.23, + "learning_rate": 1.287317260836112e-05, + "loss": 0.1748, + "step": 57953 + }, + { + "epoch": 2.23, + "learning_rate": 1.2872531418312389e-05, + "loss": 0.2185, + "step": 57954 + }, + { + "epoch": 2.23, + "learning_rate": 1.2871890228263659e-05, + "loss": 0.2355, + "step": 57955 + }, + { + "epoch": 2.23, + "learning_rate": 1.2871249038214927e-05, + "loss": 0.2544, + "step": 57956 + }, + { + "epoch": 2.23, + "learning_rate": 1.2870607848166197e-05, + "loss": 0.2243, + "step": 57957 + }, + { + "epoch": 2.23, + "learning_rate": 1.2869966658117465e-05, + "loss": 0.3122, + "step": 57958 + }, + { + "epoch": 2.23, + "learning_rate": 1.2869325468068735e-05, + "loss": 0.2513, + "step": 57959 + }, + { + "epoch": 2.23, + "learning_rate": 1.2868684278020005e-05, + "loss": 0.1828, + "step": 57960 + }, + { + "epoch": 2.23, + "learning_rate": 1.2868043087971277e-05, + "loss": 0.2293, + "step": 57961 + }, + { + "epoch": 2.23, + "learning_rate": 1.2867401897922545e-05, + "loss": 0.3306, + "step": 57962 + }, + { + "epoch": 2.23, + "learning_rate": 1.2866760707873815e-05, + "loss": 0.2703, + "step": 57963 + }, + { + "epoch": 2.23, + "learning_rate": 1.2866119517825084e-05, + "loss": 0.2185, + "step": 57964 + }, + { + "epoch": 2.23, + "learning_rate": 1.2865478327776354e-05, + "loss": 0.3295, + "step": 57965 + }, + { + "epoch": 2.23, + "learning_rate": 1.2864837137727622e-05, + "loss": 0.1416, + "step": 57966 + }, + { + "epoch": 2.23, + "learning_rate": 1.2864195947678892e-05, + "loss": 0.3487, + "step": 57967 + }, + { + "epoch": 2.23, + "learning_rate": 1.2863554757630162e-05, + "loss": 0.277, + "step": 57968 + }, + { + "epoch": 2.23, + "learning_rate": 1.286291356758143e-05, + "loss": 0.2673, + "step": 57969 + }, + { + "epoch": 2.23, + "learning_rate": 1.2862272377532702e-05, + "loss": 0.3819, + "step": 57970 + }, + { + "epoch": 2.23, + "learning_rate": 1.2861631187483972e-05, + "loss": 0.2566, + "step": 57971 + }, + { + "epoch": 2.23, + "learning_rate": 1.286098999743524e-05, + "loss": 0.2525, + "step": 57972 + }, + { + "epoch": 2.23, + "learning_rate": 1.286034880738651e-05, + "loss": 0.2054, + "step": 57973 + }, + { + "epoch": 2.23, + "learning_rate": 1.2859707617337779e-05, + "loss": 0.2587, + "step": 57974 + }, + { + "epoch": 2.23, + "learning_rate": 1.2859066427289049e-05, + "loss": 0.1359, + "step": 57975 + }, + { + "epoch": 2.23, + "learning_rate": 1.2858425237240317e-05, + "loss": 0.2417, + "step": 57976 + }, + { + "epoch": 2.23, + "learning_rate": 1.2857784047191587e-05, + "loss": 0.2694, + "step": 57977 + }, + { + "epoch": 2.23, + "learning_rate": 1.2857142857142857e-05, + "loss": 0.3361, + "step": 57978 + }, + { + "epoch": 2.23, + "learning_rate": 1.2856501667094129e-05, + "loss": 0.3065, + "step": 57979 + }, + { + "epoch": 2.23, + "learning_rate": 1.2855860477045397e-05, + "loss": 0.1511, + "step": 57980 + }, + { + "epoch": 2.23, + "learning_rate": 1.2855219286996667e-05, + "loss": 0.2528, + "step": 57981 + }, + { + "epoch": 2.23, + "learning_rate": 1.2854578096947936e-05, + "loss": 0.29, + "step": 57982 + }, + { + "epoch": 2.23, + "learning_rate": 1.2853936906899206e-05, + "loss": 0.1897, + "step": 57983 + }, + { + "epoch": 2.23, + "learning_rate": 1.2853295716850474e-05, + "loss": 0.223, + "step": 57984 + }, + { + "epoch": 2.23, + "learning_rate": 1.2852654526801744e-05, + "loss": 0.1757, + "step": 57985 + }, + { + "epoch": 2.23, + "learning_rate": 1.2852013336753014e-05, + "loss": 0.2423, + "step": 57986 + }, + { + "epoch": 2.23, + "learning_rate": 1.2851372146704282e-05, + "loss": 0.2377, + "step": 57987 + }, + { + "epoch": 2.23, + "learning_rate": 1.2850730956655554e-05, + "loss": 0.1898, + "step": 57988 + }, + { + "epoch": 2.23, + "learning_rate": 1.2850089766606824e-05, + "loss": 0.2725, + "step": 57989 + }, + { + "epoch": 2.23, + "learning_rate": 1.2849448576558092e-05, + "loss": 0.23, + "step": 57990 + }, + { + "epoch": 2.23, + "learning_rate": 1.2848807386509362e-05, + "loss": 0.2149, + "step": 57991 + }, + { + "epoch": 2.23, + "learning_rate": 1.284816619646063e-05, + "loss": 0.2116, + "step": 57992 + }, + { + "epoch": 2.23, + "learning_rate": 1.28475250064119e-05, + "loss": 0.1962, + "step": 57993 + }, + { + "epoch": 2.23, + "learning_rate": 1.2846883816363169e-05, + "loss": 0.2508, + "step": 57994 + }, + { + "epoch": 2.23, + "learning_rate": 1.2846242626314439e-05, + "loss": 0.2471, + "step": 57995 + }, + { + "epoch": 2.23, + "learning_rate": 1.284560143626571e-05, + "loss": 0.2618, + "step": 57996 + }, + { + "epoch": 2.23, + "learning_rate": 1.2844960246216981e-05, + "loss": 0.1562, + "step": 57997 + }, + { + "epoch": 2.23, + "learning_rate": 1.284431905616825e-05, + "loss": 0.1368, + "step": 57998 + }, + { + "epoch": 2.23, + "learning_rate": 1.284367786611952e-05, + "loss": 0.1996, + "step": 57999 + }, + { + "epoch": 2.23, + "learning_rate": 1.2843036676070788e-05, + "loss": 0.288, + "step": 58000 + }, + { + "epoch": 2.23, + "learning_rate": 1.2842395486022058e-05, + "loss": 0.247, + "step": 58001 + }, + { + "epoch": 2.23, + "learning_rate": 1.2841754295973326e-05, + "loss": 0.1373, + "step": 58002 + }, + { + "epoch": 2.23, + "learning_rate": 1.2841113105924596e-05, + "loss": 0.2584, + "step": 58003 + }, + { + "epoch": 2.23, + "learning_rate": 1.2840471915875866e-05, + "loss": 0.213, + "step": 58004 + }, + { + "epoch": 2.23, + "learning_rate": 1.2839830725827134e-05, + "loss": 0.4837, + "step": 58005 + }, + { + "epoch": 2.23, + "learning_rate": 1.2839189535778406e-05, + "loss": 0.2258, + "step": 58006 + }, + { + "epoch": 2.23, + "learning_rate": 1.2838548345729676e-05, + "loss": 0.2328, + "step": 58007 + }, + { + "epoch": 2.23, + "learning_rate": 1.2837907155680944e-05, + "loss": 0.2375, + "step": 58008 + }, + { + "epoch": 2.23, + "learning_rate": 1.2837265965632214e-05, + "loss": 0.2232, + "step": 58009 + }, + { + "epoch": 2.23, + "learning_rate": 1.2836624775583483e-05, + "loss": 0.2039, + "step": 58010 + }, + { + "epoch": 2.23, + "learning_rate": 1.2835983585534753e-05, + "loss": 0.2164, + "step": 58011 + }, + { + "epoch": 2.23, + "learning_rate": 1.2835342395486023e-05, + "loss": 0.2083, + "step": 58012 + }, + { + "epoch": 2.23, + "learning_rate": 1.2834701205437291e-05, + "loss": 0.1567, + "step": 58013 + }, + { + "epoch": 2.23, + "learning_rate": 1.2834060015388561e-05, + "loss": 0.2427, + "step": 58014 + }, + { + "epoch": 2.23, + "learning_rate": 1.2833418825339833e-05, + "loss": 0.3782, + "step": 58015 + }, + { + "epoch": 2.23, + "learning_rate": 1.2832777635291101e-05, + "loss": 0.2164, + "step": 58016 + }, + { + "epoch": 2.23, + "learning_rate": 1.2832136445242371e-05, + "loss": 0.2547, + "step": 58017 + }, + { + "epoch": 2.23, + "learning_rate": 1.283149525519364e-05, + "loss": 0.2339, + "step": 58018 + }, + { + "epoch": 2.23, + "learning_rate": 1.283085406514491e-05, + "loss": 0.2865, + "step": 58019 + }, + { + "epoch": 2.23, + "learning_rate": 1.2830212875096178e-05, + "loss": 0.2201, + "step": 58020 + }, + { + "epoch": 2.23, + "learning_rate": 1.2829571685047448e-05, + "loss": 0.1906, + "step": 58021 + }, + { + "epoch": 2.23, + "learning_rate": 1.2828930494998718e-05, + "loss": 0.1574, + "step": 58022 + }, + { + "epoch": 2.23, + "learning_rate": 1.282828930494999e-05, + "loss": 0.3616, + "step": 58023 + }, + { + "epoch": 2.23, + "learning_rate": 1.2827648114901258e-05, + "loss": 0.2301, + "step": 58024 + }, + { + "epoch": 2.23, + "learning_rate": 1.2827006924852528e-05, + "loss": 0.2268, + "step": 58025 + }, + { + "epoch": 2.23, + "learning_rate": 1.2826365734803796e-05, + "loss": 0.368, + "step": 58026 + }, + { + "epoch": 2.23, + "learning_rate": 1.2825724544755066e-05, + "loss": 0.2335, + "step": 58027 + }, + { + "epoch": 2.23, + "learning_rate": 1.2825083354706335e-05, + "loss": 0.3928, + "step": 58028 + }, + { + "epoch": 2.23, + "learning_rate": 1.2824442164657605e-05, + "loss": 0.3859, + "step": 58029 + }, + { + "epoch": 2.23, + "learning_rate": 1.2823800974608875e-05, + "loss": 0.2056, + "step": 58030 + }, + { + "epoch": 2.23, + "learning_rate": 1.2823159784560143e-05, + "loss": 0.2358, + "step": 58031 + }, + { + "epoch": 2.23, + "learning_rate": 1.2822518594511415e-05, + "loss": 0.2238, + "step": 58032 + }, + { + "epoch": 2.23, + "learning_rate": 1.2821877404462685e-05, + "loss": 0.3452, + "step": 58033 + }, + { + "epoch": 2.23, + "learning_rate": 1.2821236214413953e-05, + "loss": 0.245, + "step": 58034 + }, + { + "epoch": 2.23, + "learning_rate": 1.2820595024365223e-05, + "loss": 0.1377, + "step": 58035 + }, + { + "epoch": 2.23, + "learning_rate": 1.2819953834316491e-05, + "loss": 0.2161, + "step": 58036 + }, + { + "epoch": 2.23, + "learning_rate": 1.2819312644267761e-05, + "loss": 0.1751, + "step": 58037 + }, + { + "epoch": 2.23, + "learning_rate": 1.281867145421903e-05, + "loss": 0.2141, + "step": 58038 + }, + { + "epoch": 2.23, + "learning_rate": 1.28180302641703e-05, + "loss": 0.268, + "step": 58039 + }, + { + "epoch": 2.23, + "learning_rate": 1.281738907412157e-05, + "loss": 0.3115, + "step": 58040 + }, + { + "epoch": 2.23, + "learning_rate": 1.2816747884072842e-05, + "loss": 0.1666, + "step": 58041 + }, + { + "epoch": 2.23, + "learning_rate": 1.281610669402411e-05, + "loss": 0.2193, + "step": 58042 + }, + { + "epoch": 2.23, + "learning_rate": 1.281546550397538e-05, + "loss": 0.3247, + "step": 58043 + }, + { + "epoch": 2.23, + "learning_rate": 1.2814824313926648e-05, + "loss": 0.2328, + "step": 58044 + }, + { + "epoch": 2.23, + "learning_rate": 1.2814183123877918e-05, + "loss": 0.1817, + "step": 58045 + }, + { + "epoch": 2.23, + "learning_rate": 1.2813541933829187e-05, + "loss": 0.1957, + "step": 58046 + }, + { + "epoch": 2.23, + "learning_rate": 1.2812900743780457e-05, + "loss": 0.1748, + "step": 58047 + }, + { + "epoch": 2.23, + "learning_rate": 1.2812259553731727e-05, + "loss": 0.1469, + "step": 58048 + }, + { + "epoch": 2.23, + "learning_rate": 1.2811618363682995e-05, + "loss": 0.225, + "step": 58049 + }, + { + "epoch": 2.23, + "learning_rate": 1.2810977173634267e-05, + "loss": 0.2808, + "step": 58050 + }, + { + "epoch": 2.23, + "learning_rate": 1.2810335983585537e-05, + "loss": 0.2852, + "step": 58051 + }, + { + "epoch": 2.23, + "learning_rate": 1.2809694793536805e-05, + "loss": 0.2747, + "step": 58052 + }, + { + "epoch": 2.23, + "learning_rate": 1.2809053603488075e-05, + "loss": 0.1753, + "step": 58053 + }, + { + "epoch": 2.23, + "learning_rate": 1.2808412413439343e-05, + "loss": 0.1399, + "step": 58054 + }, + { + "epoch": 2.23, + "learning_rate": 1.2807771223390613e-05, + "loss": 0.2109, + "step": 58055 + }, + { + "epoch": 2.23, + "learning_rate": 1.2807130033341882e-05, + "loss": 0.2174, + "step": 58056 + }, + { + "epoch": 2.23, + "learning_rate": 1.2806488843293152e-05, + "loss": 0.226, + "step": 58057 + }, + { + "epoch": 2.23, + "learning_rate": 1.2805847653244422e-05, + "loss": 0.232, + "step": 58058 + }, + { + "epoch": 2.23, + "learning_rate": 1.2805206463195693e-05, + "loss": 0.1543, + "step": 58059 + }, + { + "epoch": 2.23, + "learning_rate": 1.2804565273146962e-05, + "loss": 0.3545, + "step": 58060 + }, + { + "epoch": 2.23, + "learning_rate": 1.2803924083098232e-05, + "loss": 0.2912, + "step": 58061 + }, + { + "epoch": 2.23, + "learning_rate": 1.28032828930495e-05, + "loss": 0.1084, + "step": 58062 + }, + { + "epoch": 2.23, + "learning_rate": 1.280264170300077e-05, + "loss": 0.1952, + "step": 58063 + }, + { + "epoch": 2.23, + "learning_rate": 1.2802000512952039e-05, + "loss": 0.2919, + "step": 58064 + }, + { + "epoch": 2.23, + "learning_rate": 1.2801359322903309e-05, + "loss": 0.2832, + "step": 58065 + }, + { + "epoch": 2.23, + "learning_rate": 1.2800718132854579e-05, + "loss": 0.1894, + "step": 58066 + }, + { + "epoch": 2.23, + "learning_rate": 1.2800076942805847e-05, + "loss": 0.185, + "step": 58067 + }, + { + "epoch": 2.23, + "learning_rate": 1.2799435752757119e-05, + "loss": 0.168, + "step": 58068 + }, + { + "epoch": 2.23, + "learning_rate": 1.2798794562708389e-05, + "loss": 0.2613, + "step": 58069 + }, + { + "epoch": 2.23, + "learning_rate": 1.2798153372659657e-05, + "loss": 0.3228, + "step": 58070 + }, + { + "epoch": 2.23, + "learning_rate": 1.2797512182610927e-05, + "loss": 0.3207, + "step": 58071 + }, + { + "epoch": 2.23, + "learning_rate": 1.2796870992562195e-05, + "loss": 0.175, + "step": 58072 + }, + { + "epoch": 2.23, + "learning_rate": 1.2796229802513465e-05, + "loss": 0.1862, + "step": 58073 + }, + { + "epoch": 2.23, + "learning_rate": 1.2795588612464735e-05, + "loss": 0.2575, + "step": 58074 + }, + { + "epoch": 2.23, + "learning_rate": 1.2794947422416004e-05, + "loss": 0.1418, + "step": 58075 + }, + { + "epoch": 2.23, + "learning_rate": 1.2794306232367274e-05, + "loss": 0.2123, + "step": 58076 + }, + { + "epoch": 2.23, + "learning_rate": 1.2793665042318545e-05, + "loss": 0.101, + "step": 58077 + }, + { + "epoch": 2.23, + "learning_rate": 1.2793023852269814e-05, + "loss": 0.168, + "step": 58078 + }, + { + "epoch": 2.23, + "learning_rate": 1.2792382662221084e-05, + "loss": 0.2077, + "step": 58079 + }, + { + "epoch": 2.23, + "learning_rate": 1.2791741472172352e-05, + "loss": 0.3185, + "step": 58080 + }, + { + "epoch": 2.23, + "learning_rate": 1.2791100282123622e-05, + "loss": 0.4269, + "step": 58081 + }, + { + "epoch": 2.23, + "learning_rate": 1.279045909207489e-05, + "loss": 0.1959, + "step": 58082 + }, + { + "epoch": 2.23, + "learning_rate": 1.278981790202616e-05, + "loss": 0.1983, + "step": 58083 + }, + { + "epoch": 2.23, + "learning_rate": 1.278917671197743e-05, + "loss": 0.1683, + "step": 58084 + }, + { + "epoch": 2.23, + "learning_rate": 1.2788535521928699e-05, + "loss": 0.1477, + "step": 58085 + }, + { + "epoch": 2.23, + "learning_rate": 1.278789433187997e-05, + "loss": 0.2152, + "step": 58086 + }, + { + "epoch": 2.23, + "learning_rate": 1.278725314183124e-05, + "loss": 0.2212, + "step": 58087 + }, + { + "epoch": 2.23, + "learning_rate": 1.2786611951782509e-05, + "loss": 0.1404, + "step": 58088 + }, + { + "epoch": 2.23, + "learning_rate": 1.2785970761733779e-05, + "loss": 0.237, + "step": 58089 + }, + { + "epoch": 2.23, + "learning_rate": 1.2785329571685047e-05, + "loss": 0.2667, + "step": 58090 + }, + { + "epoch": 2.23, + "learning_rate": 1.2784688381636317e-05, + "loss": 0.3079, + "step": 58091 + }, + { + "epoch": 2.23, + "learning_rate": 1.2784047191587587e-05, + "loss": 0.3148, + "step": 58092 + }, + { + "epoch": 2.23, + "learning_rate": 1.2783406001538856e-05, + "loss": 0.2891, + "step": 58093 + }, + { + "epoch": 2.23, + "learning_rate": 1.2782764811490126e-05, + "loss": 0.189, + "step": 58094 + }, + { + "epoch": 2.23, + "learning_rate": 1.2782123621441397e-05, + "loss": 0.19, + "step": 58095 + }, + { + "epoch": 2.23, + "learning_rate": 1.2781482431392666e-05, + "loss": 0.2666, + "step": 58096 + }, + { + "epoch": 2.23, + "learning_rate": 1.2780841241343936e-05, + "loss": 0.2867, + "step": 58097 + }, + { + "epoch": 2.23, + "learning_rate": 1.2780200051295204e-05, + "loss": 0.244, + "step": 58098 + }, + { + "epoch": 2.23, + "learning_rate": 1.2779558861246474e-05, + "loss": 0.2311, + "step": 58099 + }, + { + "epoch": 2.23, + "learning_rate": 1.2778917671197742e-05, + "loss": 0.1387, + "step": 58100 + }, + { + "epoch": 2.23, + "learning_rate": 1.2778276481149012e-05, + "loss": 0.1313, + "step": 58101 + }, + { + "epoch": 2.23, + "learning_rate": 1.2777635291100282e-05, + "loss": 0.273, + "step": 58102 + }, + { + "epoch": 2.23, + "learning_rate": 1.277699410105155e-05, + "loss": 0.1749, + "step": 58103 + }, + { + "epoch": 2.23, + "learning_rate": 1.2776352911002822e-05, + "loss": 0.1485, + "step": 58104 + }, + { + "epoch": 2.23, + "learning_rate": 1.2775711720954092e-05, + "loss": 0.3426, + "step": 58105 + }, + { + "epoch": 2.23, + "learning_rate": 1.277507053090536e-05, + "loss": 0.2472, + "step": 58106 + }, + { + "epoch": 2.23, + "learning_rate": 1.277442934085663e-05, + "loss": 0.2848, + "step": 58107 + }, + { + "epoch": 2.23, + "learning_rate": 1.27737881508079e-05, + "loss": 0.2333, + "step": 58108 + }, + { + "epoch": 2.23, + "learning_rate": 1.277314696075917e-05, + "loss": 0.2703, + "step": 58109 + }, + { + "epoch": 2.23, + "learning_rate": 1.277250577071044e-05, + "loss": 0.1798, + "step": 58110 + }, + { + "epoch": 2.23, + "learning_rate": 1.2771864580661708e-05, + "loss": 0.3009, + "step": 58111 + }, + { + "epoch": 2.23, + "learning_rate": 1.277122339061298e-05, + "loss": 0.2144, + "step": 58112 + }, + { + "epoch": 2.23, + "learning_rate": 1.277058220056425e-05, + "loss": 0.166, + "step": 58113 + }, + { + "epoch": 2.23, + "learning_rate": 1.2769941010515518e-05, + "loss": 0.2784, + "step": 58114 + }, + { + "epoch": 2.23, + "learning_rate": 1.2769299820466788e-05, + "loss": 0.2158, + "step": 58115 + }, + { + "epoch": 2.23, + "learning_rate": 1.2768658630418056e-05, + "loss": 0.2479, + "step": 58116 + }, + { + "epoch": 2.23, + "learning_rate": 1.2768017440369326e-05, + "loss": 0.1798, + "step": 58117 + }, + { + "epoch": 2.23, + "learning_rate": 1.2767376250320594e-05, + "loss": 0.2548, + "step": 58118 + }, + { + "epoch": 2.23, + "learning_rate": 1.2766735060271864e-05, + "loss": 0.1791, + "step": 58119 + }, + { + "epoch": 2.23, + "learning_rate": 1.2766093870223134e-05, + "loss": 0.2085, + "step": 58120 + }, + { + "epoch": 2.23, + "learning_rate": 1.2765452680174406e-05, + "loss": 0.2341, + "step": 58121 + }, + { + "epoch": 2.23, + "learning_rate": 1.2764811490125674e-05, + "loss": 0.2788, + "step": 58122 + }, + { + "epoch": 2.23, + "learning_rate": 1.2764170300076944e-05, + "loss": 0.173, + "step": 58123 + }, + { + "epoch": 2.23, + "learning_rate": 1.2763529110028213e-05, + "loss": 0.4034, + "step": 58124 + }, + { + "epoch": 2.23, + "learning_rate": 1.2762887919979483e-05, + "loss": 0.2572, + "step": 58125 + }, + { + "epoch": 2.23, + "learning_rate": 1.2762246729930751e-05, + "loss": 0.1747, + "step": 58126 + }, + { + "epoch": 2.23, + "learning_rate": 1.2761605539882021e-05, + "loss": 0.2769, + "step": 58127 + }, + { + "epoch": 2.23, + "learning_rate": 1.2760964349833291e-05, + "loss": 0.3123, + "step": 58128 + }, + { + "epoch": 2.23, + "learning_rate": 1.276032315978456e-05, + "loss": 0.1504, + "step": 58129 + }, + { + "epoch": 2.23, + "learning_rate": 1.2759681969735831e-05, + "loss": 0.245, + "step": 58130 + }, + { + "epoch": 2.23, + "learning_rate": 1.2759040779687101e-05, + "loss": 0.2669, + "step": 58131 + }, + { + "epoch": 2.23, + "learning_rate": 1.275839958963837e-05, + "loss": 0.1533, + "step": 58132 + }, + { + "epoch": 2.24, + "learning_rate": 1.275775839958964e-05, + "loss": 0.1977, + "step": 58133 + }, + { + "epoch": 2.24, + "learning_rate": 1.2757117209540908e-05, + "loss": 0.1979, + "step": 58134 + }, + { + "epoch": 2.24, + "learning_rate": 1.2756476019492178e-05, + "loss": 0.3765, + "step": 58135 + }, + { + "epoch": 2.24, + "learning_rate": 1.2755834829443448e-05, + "loss": 0.319, + "step": 58136 + }, + { + "epoch": 2.24, + "learning_rate": 1.2755193639394716e-05, + "loss": 0.2505, + "step": 58137 + }, + { + "epoch": 2.24, + "learning_rate": 1.2754552449345986e-05, + "loss": 0.1104, + "step": 58138 + }, + { + "epoch": 2.24, + "learning_rate": 1.2753911259297258e-05, + "loss": 0.2182, + "step": 58139 + }, + { + "epoch": 2.24, + "learning_rate": 1.2753270069248526e-05, + "loss": 0.2293, + "step": 58140 + }, + { + "epoch": 2.24, + "learning_rate": 1.2752628879199796e-05, + "loss": 0.2682, + "step": 58141 + }, + { + "epoch": 2.24, + "learning_rate": 1.2751987689151065e-05, + "loss": 0.2803, + "step": 58142 + }, + { + "epoch": 2.24, + "learning_rate": 1.2751346499102335e-05, + "loss": 0.3514, + "step": 58143 + }, + { + "epoch": 2.24, + "learning_rate": 1.2750705309053603e-05, + "loss": 0.1462, + "step": 58144 + }, + { + "epoch": 2.24, + "learning_rate": 1.2750064119004873e-05, + "loss": 0.1456, + "step": 58145 + }, + { + "epoch": 2.24, + "learning_rate": 1.2749422928956143e-05, + "loss": 0.1568, + "step": 58146 + }, + { + "epoch": 2.24, + "learning_rate": 1.2748781738907411e-05, + "loss": 0.3885, + "step": 58147 + }, + { + "epoch": 2.24, + "learning_rate": 1.2748140548858683e-05, + "loss": 0.2803, + "step": 58148 + }, + { + "epoch": 2.24, + "learning_rate": 1.2747499358809953e-05, + "loss": 0.1911, + "step": 58149 + }, + { + "epoch": 2.24, + "learning_rate": 1.2746858168761221e-05, + "loss": 0.1987, + "step": 58150 + }, + { + "epoch": 2.24, + "learning_rate": 1.2746216978712492e-05, + "loss": 0.2857, + "step": 58151 + }, + { + "epoch": 2.24, + "learning_rate": 1.274557578866376e-05, + "loss": 0.2935, + "step": 58152 + }, + { + "epoch": 2.24, + "learning_rate": 1.274493459861503e-05, + "loss": 0.2675, + "step": 58153 + }, + { + "epoch": 2.24, + "learning_rate": 1.27442934085663e-05, + "loss": 0.2739, + "step": 58154 + }, + { + "epoch": 2.24, + "learning_rate": 1.2743652218517568e-05, + "loss": 0.2388, + "step": 58155 + }, + { + "epoch": 2.24, + "learning_rate": 1.2743011028468838e-05, + "loss": 0.2065, + "step": 58156 + }, + { + "epoch": 2.24, + "learning_rate": 1.274236983842011e-05, + "loss": 0.1578, + "step": 58157 + }, + { + "epoch": 2.24, + "learning_rate": 1.2741728648371378e-05, + "loss": 0.2275, + "step": 58158 + }, + { + "epoch": 2.24, + "learning_rate": 1.2741087458322648e-05, + "loss": 0.167, + "step": 58159 + }, + { + "epoch": 2.24, + "learning_rate": 1.2740446268273917e-05, + "loss": 0.3826, + "step": 58160 + }, + { + "epoch": 2.24, + "learning_rate": 1.2739805078225187e-05, + "loss": 0.1847, + "step": 58161 + }, + { + "epoch": 2.24, + "learning_rate": 1.2739163888176455e-05, + "loss": 0.1559, + "step": 58162 + }, + { + "epoch": 2.24, + "learning_rate": 1.2738522698127725e-05, + "loss": 0.2449, + "step": 58163 + }, + { + "epoch": 2.24, + "learning_rate": 1.2737881508078995e-05, + "loss": 0.3178, + "step": 58164 + }, + { + "epoch": 2.24, + "learning_rate": 1.2737240318030263e-05, + "loss": 0.1939, + "step": 58165 + }, + { + "epoch": 2.24, + "learning_rate": 1.2736599127981535e-05, + "loss": 0.1626, + "step": 58166 + }, + { + "epoch": 2.24, + "learning_rate": 1.2735957937932805e-05, + "loss": 0.3037, + "step": 58167 + }, + { + "epoch": 2.24, + "learning_rate": 1.2735316747884073e-05, + "loss": 0.2165, + "step": 58168 + }, + { + "epoch": 2.24, + "learning_rate": 1.2734675557835343e-05, + "loss": 0.2587, + "step": 58169 + }, + { + "epoch": 2.24, + "learning_rate": 1.2734034367786612e-05, + "loss": 0.1764, + "step": 58170 + }, + { + "epoch": 2.24, + "learning_rate": 1.2733393177737882e-05, + "loss": 0.1197, + "step": 58171 + }, + { + "epoch": 2.24, + "learning_rate": 1.2732751987689152e-05, + "loss": 0.2622, + "step": 58172 + }, + { + "epoch": 2.24, + "learning_rate": 1.273211079764042e-05, + "loss": 0.2402, + "step": 58173 + }, + { + "epoch": 2.24, + "learning_rate": 1.273146960759169e-05, + "loss": 0.1795, + "step": 58174 + }, + { + "epoch": 2.24, + "learning_rate": 1.2730828417542962e-05, + "loss": 0.1196, + "step": 58175 + }, + { + "epoch": 2.24, + "learning_rate": 1.273018722749423e-05, + "loss": 0.3006, + "step": 58176 + }, + { + "epoch": 2.24, + "learning_rate": 1.27295460374455e-05, + "loss": 0.2326, + "step": 58177 + }, + { + "epoch": 2.24, + "learning_rate": 1.2728904847396769e-05, + "loss": 0.2476, + "step": 58178 + }, + { + "epoch": 2.24, + "learning_rate": 1.2728263657348039e-05, + "loss": 0.2071, + "step": 58179 + }, + { + "epoch": 2.24, + "learning_rate": 1.2727622467299307e-05, + "loss": 0.1646, + "step": 58180 + }, + { + "epoch": 2.24, + "learning_rate": 1.2726981277250577e-05, + "loss": 0.2174, + "step": 58181 + }, + { + "epoch": 2.24, + "learning_rate": 1.2726340087201847e-05, + "loss": 0.1682, + "step": 58182 + }, + { + "epoch": 2.24, + "learning_rate": 1.2725698897153115e-05, + "loss": 0.1699, + "step": 58183 + }, + { + "epoch": 2.24, + "learning_rate": 1.2725057707104387e-05, + "loss": 0.2221, + "step": 58184 + }, + { + "epoch": 2.24, + "learning_rate": 1.2724416517055657e-05, + "loss": 0.3057, + "step": 58185 + }, + { + "epoch": 2.24, + "learning_rate": 1.2723775327006925e-05, + "loss": 0.4154, + "step": 58186 + }, + { + "epoch": 2.24, + "learning_rate": 1.2723134136958195e-05, + "loss": 0.1837, + "step": 58187 + }, + { + "epoch": 2.24, + "learning_rate": 1.2722492946909464e-05, + "loss": 0.263, + "step": 58188 + }, + { + "epoch": 2.24, + "learning_rate": 1.2721851756860734e-05, + "loss": 0.1445, + "step": 58189 + }, + { + "epoch": 2.24, + "learning_rate": 1.2721210566812004e-05, + "loss": 0.3103, + "step": 58190 + }, + { + "epoch": 2.24, + "learning_rate": 1.2720569376763272e-05, + "loss": 0.2748, + "step": 58191 + }, + { + "epoch": 2.24, + "learning_rate": 1.2719928186714542e-05, + "loss": 0.1988, + "step": 58192 + }, + { + "epoch": 2.24, + "learning_rate": 1.2719286996665814e-05, + "loss": 0.1673, + "step": 58193 + }, + { + "epoch": 2.24, + "learning_rate": 1.2718645806617082e-05, + "loss": 0.1885, + "step": 58194 + }, + { + "epoch": 2.24, + "learning_rate": 1.2718004616568352e-05, + "loss": 0.2378, + "step": 58195 + }, + { + "epoch": 2.24, + "learning_rate": 1.271736342651962e-05, + "loss": 0.275, + "step": 58196 + }, + { + "epoch": 2.24, + "learning_rate": 1.271672223647089e-05, + "loss": 0.2274, + "step": 58197 + }, + { + "epoch": 2.24, + "learning_rate": 1.2716081046422159e-05, + "loss": 0.1045, + "step": 58198 + }, + { + "epoch": 2.24, + "learning_rate": 1.2715439856373429e-05, + "loss": 0.269, + "step": 58199 + }, + { + "epoch": 2.24, + "learning_rate": 1.2714798666324699e-05, + "loss": 0.3473, + "step": 58200 + }, + { + "epoch": 2.24, + "learning_rate": 1.2714157476275967e-05, + "loss": 0.2334, + "step": 58201 + }, + { + "epoch": 2.24, + "learning_rate": 1.2713516286227239e-05, + "loss": 0.2313, + "step": 58202 + }, + { + "epoch": 2.24, + "learning_rate": 1.2712875096178509e-05, + "loss": 0.207, + "step": 58203 + }, + { + "epoch": 2.24, + "learning_rate": 1.2712233906129777e-05, + "loss": 0.2766, + "step": 58204 + }, + { + "epoch": 2.24, + "learning_rate": 1.2711592716081047e-05, + "loss": 0.1184, + "step": 58205 + }, + { + "epoch": 2.24, + "learning_rate": 1.2710951526032316e-05, + "loss": 0.2166, + "step": 58206 + }, + { + "epoch": 2.24, + "learning_rate": 1.2710310335983586e-05, + "loss": 0.1497, + "step": 58207 + }, + { + "epoch": 2.24, + "learning_rate": 1.2709669145934856e-05, + "loss": 0.1057, + "step": 58208 + }, + { + "epoch": 2.24, + "learning_rate": 1.2709027955886124e-05, + "loss": 0.2923, + "step": 58209 + }, + { + "epoch": 2.24, + "learning_rate": 1.2708386765837396e-05, + "loss": 0.1945, + "step": 58210 + }, + { + "epoch": 2.24, + "learning_rate": 1.2707745575788666e-05, + "loss": 0.3668, + "step": 58211 + }, + { + "epoch": 2.24, + "learning_rate": 1.2707104385739934e-05, + "loss": 0.2635, + "step": 58212 + }, + { + "epoch": 2.24, + "learning_rate": 1.2706463195691204e-05, + "loss": 0.2067, + "step": 58213 + }, + { + "epoch": 2.24, + "learning_rate": 1.2705822005642472e-05, + "loss": 0.2223, + "step": 58214 + }, + { + "epoch": 2.24, + "learning_rate": 1.2705180815593742e-05, + "loss": 0.1717, + "step": 58215 + }, + { + "epoch": 2.24, + "learning_rate": 1.2704539625545012e-05, + "loss": 0.2273, + "step": 58216 + }, + { + "epoch": 2.24, + "learning_rate": 1.270389843549628e-05, + "loss": 0.2448, + "step": 58217 + }, + { + "epoch": 2.24, + "learning_rate": 1.270325724544755e-05, + "loss": 0.2931, + "step": 58218 + }, + { + "epoch": 2.24, + "learning_rate": 1.2702616055398823e-05, + "loss": 0.4125, + "step": 58219 + }, + { + "epoch": 2.24, + "learning_rate": 1.2701974865350091e-05, + "loss": 0.1586, + "step": 58220 + }, + { + "epoch": 2.24, + "learning_rate": 1.2701333675301361e-05, + "loss": 0.3032, + "step": 58221 + }, + { + "epoch": 2.24, + "learning_rate": 1.270069248525263e-05, + "loss": 0.2581, + "step": 58222 + }, + { + "epoch": 2.24, + "learning_rate": 1.27000512952039e-05, + "loss": 0.2118, + "step": 58223 + }, + { + "epoch": 2.24, + "learning_rate": 1.2699410105155168e-05, + "loss": 0.1825, + "step": 58224 + }, + { + "epoch": 2.24, + "learning_rate": 1.2698768915106438e-05, + "loss": 0.1915, + "step": 58225 + }, + { + "epoch": 2.24, + "learning_rate": 1.2698127725057708e-05, + "loss": 0.2134, + "step": 58226 + }, + { + "epoch": 2.24, + "learning_rate": 1.2697486535008976e-05, + "loss": 0.1135, + "step": 58227 + }, + { + "epoch": 2.24, + "learning_rate": 1.2696845344960248e-05, + "loss": 0.1921, + "step": 58228 + }, + { + "epoch": 2.24, + "learning_rate": 1.2696204154911518e-05, + "loss": 0.2338, + "step": 58229 + }, + { + "epoch": 2.24, + "learning_rate": 1.2695562964862786e-05, + "loss": 0.1232, + "step": 58230 + }, + { + "epoch": 2.24, + "learning_rate": 1.2694921774814056e-05, + "loss": 0.3338, + "step": 58231 + }, + { + "epoch": 2.24, + "learning_rate": 1.2694280584765324e-05, + "loss": 0.1695, + "step": 58232 + }, + { + "epoch": 2.24, + "learning_rate": 1.2693639394716594e-05, + "loss": 0.2047, + "step": 58233 + }, + { + "epoch": 2.24, + "learning_rate": 1.2692998204667864e-05, + "loss": 0.2453, + "step": 58234 + }, + { + "epoch": 2.24, + "learning_rate": 1.2692357014619133e-05, + "loss": 0.1869, + "step": 58235 + }, + { + "epoch": 2.24, + "learning_rate": 1.2691715824570403e-05, + "loss": 0.2381, + "step": 58236 + }, + { + "epoch": 2.24, + "learning_rate": 1.2691074634521674e-05, + "loss": 0.208, + "step": 58237 + }, + { + "epoch": 2.24, + "learning_rate": 1.2690433444472943e-05, + "loss": 0.221, + "step": 58238 + }, + { + "epoch": 2.24, + "learning_rate": 1.2689792254424213e-05, + "loss": 0.2558, + "step": 58239 + }, + { + "epoch": 2.24, + "learning_rate": 1.2689151064375481e-05, + "loss": 0.2808, + "step": 58240 + }, + { + "epoch": 2.24, + "learning_rate": 1.2688509874326751e-05, + "loss": 0.1602, + "step": 58241 + }, + { + "epoch": 2.24, + "learning_rate": 1.268786868427802e-05, + "loss": 0.2117, + "step": 58242 + }, + { + "epoch": 2.24, + "learning_rate": 1.268722749422929e-05, + "loss": 0.1555, + "step": 58243 + }, + { + "epoch": 2.24, + "learning_rate": 1.268658630418056e-05, + "loss": 0.1495, + "step": 58244 + }, + { + "epoch": 2.24, + "learning_rate": 1.2685945114131828e-05, + "loss": 0.2293, + "step": 58245 + }, + { + "epoch": 2.24, + "learning_rate": 1.26853039240831e-05, + "loss": 0.2097, + "step": 58246 + }, + { + "epoch": 2.24, + "learning_rate": 1.268466273403437e-05, + "loss": 0.1965, + "step": 58247 + }, + { + "epoch": 2.24, + "learning_rate": 1.2684021543985638e-05, + "loss": 0.4187, + "step": 58248 + }, + { + "epoch": 2.24, + "learning_rate": 1.2683380353936908e-05, + "loss": 0.1892, + "step": 58249 + }, + { + "epoch": 2.24, + "learning_rate": 1.2682739163888176e-05, + "loss": 0.1711, + "step": 58250 + }, + { + "epoch": 2.24, + "learning_rate": 1.2682097973839446e-05, + "loss": 0.1988, + "step": 58251 + }, + { + "epoch": 2.24, + "learning_rate": 1.2681456783790716e-05, + "loss": 0.2605, + "step": 58252 + }, + { + "epoch": 2.24, + "learning_rate": 1.2680815593741985e-05, + "loss": 0.274, + "step": 58253 + }, + { + "epoch": 2.24, + "learning_rate": 1.2680174403693255e-05, + "loss": 0.1757, + "step": 58254 + }, + { + "epoch": 2.24, + "learning_rate": 1.2679533213644526e-05, + "loss": 0.2067, + "step": 58255 + }, + { + "epoch": 2.24, + "learning_rate": 1.2678892023595795e-05, + "loss": 0.3644, + "step": 58256 + }, + { + "epoch": 2.24, + "learning_rate": 1.2678250833547065e-05, + "loss": 0.1341, + "step": 58257 + }, + { + "epoch": 2.24, + "learning_rate": 1.2677609643498333e-05, + "loss": 0.2537, + "step": 58258 + }, + { + "epoch": 2.24, + "learning_rate": 1.2676968453449603e-05, + "loss": 0.2027, + "step": 58259 + }, + { + "epoch": 2.24, + "learning_rate": 1.2676327263400871e-05, + "loss": 0.1777, + "step": 58260 + }, + { + "epoch": 2.24, + "learning_rate": 1.2675686073352141e-05, + "loss": 0.1419, + "step": 58261 + }, + { + "epoch": 2.24, + "learning_rate": 1.2675044883303411e-05, + "loss": 0.2658, + "step": 58262 + }, + { + "epoch": 2.24, + "learning_rate": 1.267440369325468e-05, + "loss": 0.3983, + "step": 58263 + }, + { + "epoch": 2.24, + "learning_rate": 1.2673762503205952e-05, + "loss": 0.1806, + "step": 58264 + }, + { + "epoch": 2.24, + "learning_rate": 1.2673121313157222e-05, + "loss": 0.1577, + "step": 58265 + }, + { + "epoch": 2.24, + "learning_rate": 1.267248012310849e-05, + "loss": 0.3762, + "step": 58266 + }, + { + "epoch": 2.24, + "learning_rate": 1.267183893305976e-05, + "loss": 0.2342, + "step": 58267 + }, + { + "epoch": 2.24, + "learning_rate": 1.2671197743011028e-05, + "loss": 0.1325, + "step": 58268 + }, + { + "epoch": 2.24, + "learning_rate": 1.2670556552962298e-05, + "loss": 0.336, + "step": 58269 + }, + { + "epoch": 2.24, + "learning_rate": 1.2669915362913568e-05, + "loss": 0.2907, + "step": 58270 + }, + { + "epoch": 2.24, + "learning_rate": 1.2669274172864837e-05, + "loss": 0.2288, + "step": 58271 + }, + { + "epoch": 2.24, + "learning_rate": 1.2668632982816107e-05, + "loss": 0.3447, + "step": 58272 + }, + { + "epoch": 2.24, + "learning_rate": 1.2667991792767378e-05, + "loss": 0.1706, + "step": 58273 + }, + { + "epoch": 2.24, + "learning_rate": 1.2667350602718647e-05, + "loss": 0.3962, + "step": 58274 + }, + { + "epoch": 2.24, + "learning_rate": 1.2666709412669917e-05, + "loss": 0.2446, + "step": 58275 + }, + { + "epoch": 2.24, + "learning_rate": 1.2666068222621185e-05, + "loss": 0.2765, + "step": 58276 + }, + { + "epoch": 2.24, + "learning_rate": 1.2665427032572455e-05, + "loss": 0.222, + "step": 58277 + }, + { + "epoch": 2.24, + "learning_rate": 1.2664785842523725e-05, + "loss": 0.1962, + "step": 58278 + }, + { + "epoch": 2.24, + "learning_rate": 1.2664144652474993e-05, + "loss": 0.2349, + "step": 58279 + }, + { + "epoch": 2.24, + "learning_rate": 1.2663503462426263e-05, + "loss": 0.1739, + "step": 58280 + }, + { + "epoch": 2.24, + "learning_rate": 1.2662862272377532e-05, + "loss": 0.2025, + "step": 58281 + }, + { + "epoch": 2.24, + "learning_rate": 1.2662221082328803e-05, + "loss": 0.1641, + "step": 58282 + }, + { + "epoch": 2.24, + "learning_rate": 1.2661579892280073e-05, + "loss": 0.2671, + "step": 58283 + }, + { + "epoch": 2.24, + "learning_rate": 1.2660938702231342e-05, + "loss": 0.2252, + "step": 58284 + }, + { + "epoch": 2.24, + "learning_rate": 1.2660297512182612e-05, + "loss": 0.1306, + "step": 58285 + }, + { + "epoch": 2.24, + "learning_rate": 1.265965632213388e-05, + "loss": 0.2001, + "step": 58286 + }, + { + "epoch": 2.24, + "learning_rate": 1.265901513208515e-05, + "loss": 0.1657, + "step": 58287 + }, + { + "epoch": 2.24, + "learning_rate": 1.265837394203642e-05, + "loss": 0.1926, + "step": 58288 + }, + { + "epoch": 2.24, + "learning_rate": 1.2657732751987689e-05, + "loss": 0.178, + "step": 58289 + }, + { + "epoch": 2.24, + "learning_rate": 1.2657091561938959e-05, + "loss": 0.2328, + "step": 58290 + }, + { + "epoch": 2.24, + "learning_rate": 1.265645037189023e-05, + "loss": 0.1978, + "step": 58291 + }, + { + "epoch": 2.24, + "learning_rate": 1.2655809181841499e-05, + "loss": 0.1775, + "step": 58292 + }, + { + "epoch": 2.24, + "learning_rate": 1.2655167991792769e-05, + "loss": 0.1661, + "step": 58293 + }, + { + "epoch": 2.24, + "learning_rate": 1.2654526801744037e-05, + "loss": 0.1499, + "step": 58294 + }, + { + "epoch": 2.24, + "learning_rate": 1.2653885611695307e-05, + "loss": 0.2315, + "step": 58295 + }, + { + "epoch": 2.24, + "learning_rate": 1.2653244421646577e-05, + "loss": 0.2753, + "step": 58296 + }, + { + "epoch": 2.24, + "learning_rate": 1.2652603231597845e-05, + "loss": 0.1896, + "step": 58297 + }, + { + "epoch": 2.24, + "learning_rate": 1.2651962041549115e-05, + "loss": 0.2049, + "step": 58298 + }, + { + "epoch": 2.24, + "learning_rate": 1.2651320851500387e-05, + "loss": 0.1758, + "step": 58299 + }, + { + "epoch": 2.24, + "learning_rate": 1.2650679661451655e-05, + "loss": 0.1545, + "step": 58300 + }, + { + "epoch": 2.24, + "learning_rate": 1.2650038471402925e-05, + "loss": 0.2688, + "step": 58301 + }, + { + "epoch": 2.24, + "learning_rate": 1.2649397281354194e-05, + "loss": 0.183, + "step": 58302 + }, + { + "epoch": 2.24, + "learning_rate": 1.2648756091305464e-05, + "loss": 0.1602, + "step": 58303 + }, + { + "epoch": 2.24, + "learning_rate": 1.2648114901256732e-05, + "loss": 0.3132, + "step": 58304 + }, + { + "epoch": 2.24, + "learning_rate": 1.2647473711208002e-05, + "loss": 0.1636, + "step": 58305 + }, + { + "epoch": 2.24, + "learning_rate": 1.2646832521159272e-05, + "loss": 0.2186, + "step": 58306 + }, + { + "epoch": 2.24, + "learning_rate": 1.264619133111054e-05, + "loss": 0.2161, + "step": 58307 + }, + { + "epoch": 2.24, + "learning_rate": 1.2645550141061812e-05, + "loss": 0.2221, + "step": 58308 + }, + { + "epoch": 2.24, + "learning_rate": 1.2644908951013082e-05, + "loss": 0.2575, + "step": 58309 + }, + { + "epoch": 2.24, + "learning_rate": 1.264426776096435e-05, + "loss": 0.2348, + "step": 58310 + }, + { + "epoch": 2.24, + "learning_rate": 1.264362657091562e-05, + "loss": 0.294, + "step": 58311 + }, + { + "epoch": 2.24, + "learning_rate": 1.2642985380866889e-05, + "loss": 0.3041, + "step": 58312 + }, + { + "epoch": 2.24, + "learning_rate": 1.2642344190818159e-05, + "loss": 0.1612, + "step": 58313 + }, + { + "epoch": 2.24, + "learning_rate": 1.2641703000769429e-05, + "loss": 0.1346, + "step": 58314 + }, + { + "epoch": 2.24, + "learning_rate": 1.2641061810720697e-05, + "loss": 0.163, + "step": 58315 + }, + { + "epoch": 2.24, + "learning_rate": 1.2640420620671967e-05, + "loss": 0.2199, + "step": 58316 + }, + { + "epoch": 2.24, + "learning_rate": 1.2639779430623239e-05, + "loss": 0.2946, + "step": 58317 + }, + { + "epoch": 2.24, + "learning_rate": 1.2639138240574507e-05, + "loss": 0.3217, + "step": 58318 + }, + { + "epoch": 2.24, + "learning_rate": 1.2638497050525777e-05, + "loss": 0.2408, + "step": 58319 + }, + { + "epoch": 2.24, + "learning_rate": 1.2637855860477046e-05, + "loss": 0.2134, + "step": 58320 + }, + { + "epoch": 2.24, + "learning_rate": 1.2637214670428316e-05, + "loss": 0.1859, + "step": 58321 + }, + { + "epoch": 2.24, + "learning_rate": 1.2636573480379584e-05, + "loss": 0.2025, + "step": 58322 + }, + { + "epoch": 2.24, + "learning_rate": 1.2635932290330854e-05, + "loss": 0.2052, + "step": 58323 + }, + { + "epoch": 2.24, + "learning_rate": 1.2635291100282124e-05, + "loss": 0.2515, + "step": 58324 + }, + { + "epoch": 2.24, + "learning_rate": 1.2634649910233392e-05, + "loss": 0.2195, + "step": 58325 + }, + { + "epoch": 2.24, + "learning_rate": 1.2634008720184664e-05, + "loss": 0.1479, + "step": 58326 + }, + { + "epoch": 2.24, + "learning_rate": 1.2633367530135934e-05, + "loss": 0.2336, + "step": 58327 + }, + { + "epoch": 2.24, + "learning_rate": 1.2632726340087202e-05, + "loss": 0.2213, + "step": 58328 + }, + { + "epoch": 2.24, + "learning_rate": 1.2632085150038473e-05, + "loss": 0.1472, + "step": 58329 + }, + { + "epoch": 2.24, + "learning_rate": 1.263144395998974e-05, + "loss": 0.1735, + "step": 58330 + }, + { + "epoch": 2.24, + "learning_rate": 1.2630802769941011e-05, + "loss": 0.1478, + "step": 58331 + }, + { + "epoch": 2.24, + "learning_rate": 1.2630161579892281e-05, + "loss": 0.2107, + "step": 58332 + }, + { + "epoch": 2.24, + "learning_rate": 1.262952038984355e-05, + "loss": 0.3144, + "step": 58333 + }, + { + "epoch": 2.24, + "learning_rate": 1.262887919979482e-05, + "loss": 0.2307, + "step": 58334 + }, + { + "epoch": 2.24, + "learning_rate": 1.2628238009746091e-05, + "loss": 0.3659, + "step": 58335 + }, + { + "epoch": 2.24, + "learning_rate": 1.262759681969736e-05, + "loss": 0.2198, + "step": 58336 + }, + { + "epoch": 2.24, + "learning_rate": 1.262695562964863e-05, + "loss": 0.2952, + "step": 58337 + }, + { + "epoch": 2.24, + "learning_rate": 1.2626314439599898e-05, + "loss": 0.3564, + "step": 58338 + }, + { + "epoch": 2.24, + "learning_rate": 1.2625673249551168e-05, + "loss": 0.2971, + "step": 58339 + }, + { + "epoch": 2.24, + "learning_rate": 1.2625032059502438e-05, + "loss": 0.2071, + "step": 58340 + }, + { + "epoch": 2.24, + "learning_rate": 1.2624390869453706e-05, + "loss": 0.1459, + "step": 58341 + }, + { + "epoch": 2.24, + "learning_rate": 1.2623749679404976e-05, + "loss": 0.3182, + "step": 58342 + }, + { + "epoch": 2.24, + "learning_rate": 1.2623108489356244e-05, + "loss": 0.3391, + "step": 58343 + }, + { + "epoch": 2.24, + "learning_rate": 1.2622467299307516e-05, + "loss": 0.1619, + "step": 58344 + }, + { + "epoch": 2.24, + "learning_rate": 1.2621826109258786e-05, + "loss": 0.3023, + "step": 58345 + }, + { + "epoch": 2.24, + "learning_rate": 1.2621184919210054e-05, + "loss": 0.2208, + "step": 58346 + }, + { + "epoch": 2.24, + "learning_rate": 1.2620543729161324e-05, + "loss": 0.2051, + "step": 58347 + }, + { + "epoch": 2.24, + "learning_rate": 1.2619902539112593e-05, + "loss": 0.2509, + "step": 58348 + }, + { + "epoch": 2.24, + "learning_rate": 1.2619261349063863e-05, + "loss": 0.2602, + "step": 58349 + }, + { + "epoch": 2.24, + "learning_rate": 1.2618620159015133e-05, + "loss": 0.1605, + "step": 58350 + }, + { + "epoch": 2.24, + "learning_rate": 1.2617978968966401e-05, + "loss": 0.1313, + "step": 58351 + }, + { + "epoch": 2.24, + "learning_rate": 1.2617337778917671e-05, + "loss": 0.2954, + "step": 58352 + }, + { + "epoch": 2.24, + "learning_rate": 1.2616696588868943e-05, + "loss": 0.2172, + "step": 58353 + }, + { + "epoch": 2.24, + "learning_rate": 1.2616055398820211e-05, + "loss": 0.1102, + "step": 58354 + }, + { + "epoch": 2.24, + "learning_rate": 1.2615414208771481e-05, + "loss": 0.1624, + "step": 58355 + }, + { + "epoch": 2.24, + "learning_rate": 1.261477301872275e-05, + "loss": 0.2708, + "step": 58356 + }, + { + "epoch": 2.24, + "learning_rate": 1.261413182867402e-05, + "loss": 0.1954, + "step": 58357 + }, + { + "epoch": 2.24, + "learning_rate": 1.261349063862529e-05, + "loss": 0.2515, + "step": 58358 + }, + { + "epoch": 2.24, + "learning_rate": 1.2612849448576558e-05, + "loss": 0.2306, + "step": 58359 + }, + { + "epoch": 2.24, + "learning_rate": 1.2612208258527828e-05, + "loss": 0.1679, + "step": 58360 + }, + { + "epoch": 2.24, + "learning_rate": 1.2611567068479096e-05, + "loss": 0.2911, + "step": 58361 + }, + { + "epoch": 2.24, + "learning_rate": 1.2610925878430368e-05, + "loss": 0.2431, + "step": 58362 + }, + { + "epoch": 2.24, + "learning_rate": 1.2610284688381638e-05, + "loss": 0.1253, + "step": 58363 + }, + { + "epoch": 2.24, + "learning_rate": 1.2609643498332906e-05, + "loss": 0.2133, + "step": 58364 + }, + { + "epoch": 2.24, + "learning_rate": 1.2609002308284176e-05, + "loss": 0.3451, + "step": 58365 + }, + { + "epoch": 2.24, + "learning_rate": 1.2608361118235445e-05, + "loss": 0.205, + "step": 58366 + }, + { + "epoch": 2.24, + "learning_rate": 1.2607719928186715e-05, + "loss": 0.2131, + "step": 58367 + }, + { + "epoch": 2.24, + "learning_rate": 1.2607078738137985e-05, + "loss": 0.3348, + "step": 58368 + }, + { + "epoch": 2.24, + "learning_rate": 1.2606437548089253e-05, + "loss": 0.2071, + "step": 58369 + }, + { + "epoch": 2.24, + "learning_rate": 1.2605796358040523e-05, + "loss": 0.1215, + "step": 58370 + }, + { + "epoch": 2.24, + "learning_rate": 1.2605155167991795e-05, + "loss": 0.2267, + "step": 58371 + }, + { + "epoch": 2.24, + "learning_rate": 1.2604513977943063e-05, + "loss": 0.2503, + "step": 58372 + }, + { + "epoch": 2.24, + "learning_rate": 1.2603872787894333e-05, + "loss": 0.2502, + "step": 58373 + }, + { + "epoch": 2.24, + "learning_rate": 1.2603231597845602e-05, + "loss": 0.1694, + "step": 58374 + }, + { + "epoch": 2.24, + "learning_rate": 1.2602590407796872e-05, + "loss": 0.2342, + "step": 58375 + }, + { + "epoch": 2.24, + "learning_rate": 1.2601949217748142e-05, + "loss": 0.2662, + "step": 58376 + }, + { + "epoch": 2.24, + "learning_rate": 1.260130802769941e-05, + "loss": 0.2292, + "step": 58377 + }, + { + "epoch": 2.24, + "learning_rate": 1.260066683765068e-05, + "loss": 0.1717, + "step": 58378 + }, + { + "epoch": 2.24, + "learning_rate": 1.2600025647601948e-05, + "loss": 0.3504, + "step": 58379 + }, + { + "epoch": 2.24, + "learning_rate": 1.259938445755322e-05, + "loss": 0.201, + "step": 58380 + }, + { + "epoch": 2.24, + "learning_rate": 1.259874326750449e-05, + "loss": 0.1624, + "step": 58381 + }, + { + "epoch": 2.24, + "learning_rate": 1.2598102077455758e-05, + "loss": 0.1891, + "step": 58382 + }, + { + "epoch": 2.24, + "learning_rate": 1.2597460887407028e-05, + "loss": 0.2649, + "step": 58383 + }, + { + "epoch": 2.24, + "learning_rate": 1.2596819697358297e-05, + "loss": 0.2275, + "step": 58384 + }, + { + "epoch": 2.24, + "learning_rate": 1.2596178507309567e-05, + "loss": 0.2233, + "step": 58385 + }, + { + "epoch": 2.24, + "learning_rate": 1.2595537317260837e-05, + "loss": 0.2649, + "step": 58386 + }, + { + "epoch": 2.24, + "learning_rate": 1.2594896127212105e-05, + "loss": 0.224, + "step": 58387 + }, + { + "epoch": 2.24, + "learning_rate": 1.2594254937163375e-05, + "loss": 0.3051, + "step": 58388 + }, + { + "epoch": 2.24, + "learning_rate": 1.2593613747114647e-05, + "loss": 0.2944, + "step": 58389 + }, + { + "epoch": 2.24, + "learning_rate": 1.2592972557065915e-05, + "loss": 0.2155, + "step": 58390 + }, + { + "epoch": 2.24, + "learning_rate": 1.2592331367017185e-05, + "loss": 0.3711, + "step": 58391 + }, + { + "epoch": 2.24, + "learning_rate": 1.2591690176968453e-05, + "loss": 0.2524, + "step": 58392 + }, + { + "epoch": 2.25, + "learning_rate": 1.2591048986919723e-05, + "loss": 0.2415, + "step": 58393 + }, + { + "epoch": 2.25, + "learning_rate": 1.2590407796870993e-05, + "loss": 0.2911, + "step": 58394 + }, + { + "epoch": 2.25, + "learning_rate": 1.2589766606822262e-05, + "loss": 0.3249, + "step": 58395 + }, + { + "epoch": 2.25, + "learning_rate": 1.2589125416773532e-05, + "loss": 0.2027, + "step": 58396 + }, + { + "epoch": 2.25, + "learning_rate": 1.2588484226724804e-05, + "loss": 0.1394, + "step": 58397 + }, + { + "epoch": 2.25, + "learning_rate": 1.2587843036676072e-05, + "loss": 0.2714, + "step": 58398 + }, + { + "epoch": 2.25, + "learning_rate": 1.2587201846627342e-05, + "loss": 0.1454, + "step": 58399 + }, + { + "epoch": 2.25, + "learning_rate": 1.258656065657861e-05, + "loss": 0.3174, + "step": 58400 + }, + { + "epoch": 2.25, + "learning_rate": 1.258591946652988e-05, + "loss": 0.1889, + "step": 58401 + }, + { + "epoch": 2.25, + "learning_rate": 1.258527827648115e-05, + "loss": 0.2128, + "step": 58402 + }, + { + "epoch": 2.25, + "learning_rate": 1.2584637086432419e-05, + "loss": 0.2701, + "step": 58403 + }, + { + "epoch": 2.25, + "learning_rate": 1.2583995896383689e-05, + "loss": 0.2181, + "step": 58404 + }, + { + "epoch": 2.25, + "learning_rate": 1.2583354706334957e-05, + "loss": 0.1626, + "step": 58405 + }, + { + "epoch": 2.25, + "learning_rate": 1.2582713516286229e-05, + "loss": 0.252, + "step": 58406 + }, + { + "epoch": 2.25, + "learning_rate": 1.2582072326237499e-05, + "loss": 0.2122, + "step": 58407 + }, + { + "epoch": 2.25, + "learning_rate": 1.2581431136188767e-05, + "loss": 0.1815, + "step": 58408 + }, + { + "epoch": 2.25, + "learning_rate": 1.2580789946140037e-05, + "loss": 0.1354, + "step": 58409 + }, + { + "epoch": 2.25, + "learning_rate": 1.2580148756091305e-05, + "loss": 0.2601, + "step": 58410 + }, + { + "epoch": 2.25, + "learning_rate": 1.2579507566042575e-05, + "loss": 0.1514, + "step": 58411 + }, + { + "epoch": 2.25, + "learning_rate": 1.2578866375993845e-05, + "loss": 0.2188, + "step": 58412 + }, + { + "epoch": 2.25, + "learning_rate": 1.2578225185945114e-05, + "loss": 0.1652, + "step": 58413 + }, + { + "epoch": 2.25, + "learning_rate": 1.2577583995896384e-05, + "loss": 0.3381, + "step": 58414 + }, + { + "epoch": 2.25, + "learning_rate": 1.2576942805847655e-05, + "loss": 0.1713, + "step": 58415 + }, + { + "epoch": 2.25, + "learning_rate": 1.2576301615798924e-05, + "loss": 0.2488, + "step": 58416 + }, + { + "epoch": 2.25, + "learning_rate": 1.2575660425750194e-05, + "loss": 0.1759, + "step": 58417 + }, + { + "epoch": 2.25, + "learning_rate": 1.2575019235701462e-05, + "loss": 0.2799, + "step": 58418 + }, + { + "epoch": 2.25, + "learning_rate": 1.2574378045652732e-05, + "loss": 0.2546, + "step": 58419 + }, + { + "epoch": 2.25, + "learning_rate": 1.2573736855604002e-05, + "loss": 0.1886, + "step": 58420 + }, + { + "epoch": 2.25, + "learning_rate": 1.257309566555527e-05, + "loss": 0.1884, + "step": 58421 + }, + { + "epoch": 2.25, + "learning_rate": 1.257245447550654e-05, + "loss": 0.1686, + "step": 58422 + }, + { + "epoch": 2.25, + "learning_rate": 1.2571813285457809e-05, + "loss": 0.1287, + "step": 58423 + }, + { + "epoch": 2.25, + "learning_rate": 1.257117209540908e-05, + "loss": 0.2143, + "step": 58424 + }, + { + "epoch": 2.25, + "learning_rate": 1.257053090536035e-05, + "loss": 0.3397, + "step": 58425 + }, + { + "epoch": 2.25, + "learning_rate": 1.2569889715311619e-05, + "loss": 0.1934, + "step": 58426 + }, + { + "epoch": 2.25, + "learning_rate": 1.2569248525262889e-05, + "loss": 0.2341, + "step": 58427 + }, + { + "epoch": 2.25, + "learning_rate": 1.2568607335214157e-05, + "loss": 0.1626, + "step": 58428 + }, + { + "epoch": 2.25, + "learning_rate": 1.2567966145165427e-05, + "loss": 0.1626, + "step": 58429 + }, + { + "epoch": 2.25, + "learning_rate": 1.2567324955116697e-05, + "loss": 0.2878, + "step": 58430 + }, + { + "epoch": 2.25, + "learning_rate": 1.2566683765067966e-05, + "loss": 0.2693, + "step": 58431 + }, + { + "epoch": 2.25, + "learning_rate": 1.2566042575019236e-05, + "loss": 0.1548, + "step": 58432 + }, + { + "epoch": 2.25, + "learning_rate": 1.2565401384970507e-05, + "loss": 0.1988, + "step": 58433 + }, + { + "epoch": 2.25, + "learning_rate": 1.2564760194921776e-05, + "loss": 0.196, + "step": 58434 + }, + { + "epoch": 2.25, + "learning_rate": 1.2564119004873046e-05, + "loss": 0.2635, + "step": 58435 + }, + { + "epoch": 2.25, + "learning_rate": 1.2563477814824314e-05, + "loss": 0.2075, + "step": 58436 + }, + { + "epoch": 2.25, + "learning_rate": 1.2562836624775584e-05, + "loss": 0.2391, + "step": 58437 + }, + { + "epoch": 2.25, + "learning_rate": 1.2562195434726854e-05, + "loss": 0.2907, + "step": 58438 + }, + { + "epoch": 2.25, + "learning_rate": 1.2561554244678122e-05, + "loss": 0.3525, + "step": 58439 + }, + { + "epoch": 2.25, + "learning_rate": 1.2560913054629393e-05, + "loss": 0.1629, + "step": 58440 + }, + { + "epoch": 2.25, + "learning_rate": 1.256027186458066e-05, + "loss": 0.1818, + "step": 58441 + }, + { + "epoch": 2.25, + "learning_rate": 1.2559630674531933e-05, + "loss": 0.2868, + "step": 58442 + }, + { + "epoch": 2.25, + "learning_rate": 1.2558989484483203e-05, + "loss": 0.2479, + "step": 58443 + }, + { + "epoch": 2.25, + "learning_rate": 1.2558348294434471e-05, + "loss": 0.3001, + "step": 58444 + }, + { + "epoch": 2.25, + "learning_rate": 1.2557707104385741e-05, + "loss": 0.1874, + "step": 58445 + }, + { + "epoch": 2.25, + "learning_rate": 1.255706591433701e-05, + "loss": 0.2044, + "step": 58446 + }, + { + "epoch": 2.25, + "learning_rate": 1.255642472428828e-05, + "loss": 0.1443, + "step": 58447 + }, + { + "epoch": 2.25, + "learning_rate": 1.255578353423955e-05, + "loss": 0.1989, + "step": 58448 + }, + { + "epoch": 2.25, + "learning_rate": 1.2555142344190818e-05, + "loss": 0.2419, + "step": 58449 + }, + { + "epoch": 2.25, + "learning_rate": 1.2554501154142088e-05, + "loss": 0.1934, + "step": 58450 + }, + { + "epoch": 2.25, + "learning_rate": 1.255385996409336e-05, + "loss": 0.2373, + "step": 58451 + }, + { + "epoch": 2.25, + "learning_rate": 1.2553218774044628e-05, + "loss": 0.2407, + "step": 58452 + }, + { + "epoch": 2.25, + "learning_rate": 1.2552577583995898e-05, + "loss": 0.1521, + "step": 58453 + }, + { + "epoch": 2.25, + "learning_rate": 1.2551936393947166e-05, + "loss": 0.2556, + "step": 58454 + }, + { + "epoch": 2.25, + "learning_rate": 1.2551295203898436e-05, + "loss": 0.2672, + "step": 58455 + }, + { + "epoch": 2.25, + "learning_rate": 1.2550654013849706e-05, + "loss": 0.0872, + "step": 58456 + }, + { + "epoch": 2.25, + "learning_rate": 1.2550012823800974e-05, + "loss": 0.3487, + "step": 58457 + }, + { + "epoch": 2.25, + "learning_rate": 1.2549371633752244e-05, + "loss": 0.2876, + "step": 58458 + }, + { + "epoch": 2.25, + "learning_rate": 1.2548730443703513e-05, + "loss": 0.3312, + "step": 58459 + }, + { + "epoch": 2.25, + "learning_rate": 1.2548089253654784e-05, + "loss": 0.2961, + "step": 58460 + }, + { + "epoch": 2.25, + "learning_rate": 1.2547448063606055e-05, + "loss": 0.2622, + "step": 58461 + }, + { + "epoch": 2.25, + "learning_rate": 1.2546806873557323e-05, + "loss": 0.1793, + "step": 58462 + }, + { + "epoch": 2.25, + "learning_rate": 1.2546165683508593e-05, + "loss": 0.2686, + "step": 58463 + }, + { + "epoch": 2.25, + "learning_rate": 1.2545524493459861e-05, + "loss": 0.2532, + "step": 58464 + }, + { + "epoch": 2.25, + "learning_rate": 1.2544883303411131e-05, + "loss": 0.2357, + "step": 58465 + }, + { + "epoch": 2.25, + "learning_rate": 1.2544242113362401e-05, + "loss": 0.159, + "step": 58466 + }, + { + "epoch": 2.25, + "learning_rate": 1.254360092331367e-05, + "loss": 0.1082, + "step": 58467 + }, + { + "epoch": 2.25, + "learning_rate": 1.254295973326494e-05, + "loss": 0.2696, + "step": 58468 + }, + { + "epoch": 2.25, + "learning_rate": 1.2542318543216211e-05, + "loss": 0.2657, + "step": 58469 + }, + { + "epoch": 2.25, + "learning_rate": 1.254167735316748e-05, + "loss": 0.1437, + "step": 58470 + }, + { + "epoch": 2.25, + "learning_rate": 1.254103616311875e-05, + "loss": 0.2769, + "step": 58471 + }, + { + "epoch": 2.25, + "learning_rate": 1.2540394973070018e-05, + "loss": 0.2228, + "step": 58472 + }, + { + "epoch": 2.25, + "learning_rate": 1.2539753783021288e-05, + "loss": 0.276, + "step": 58473 + }, + { + "epoch": 2.25, + "learning_rate": 1.2539112592972558e-05, + "loss": 0.2627, + "step": 58474 + }, + { + "epoch": 2.25, + "learning_rate": 1.2538471402923826e-05, + "loss": 0.2417, + "step": 58475 + }, + { + "epoch": 2.25, + "learning_rate": 1.2537830212875096e-05, + "loss": 0.1614, + "step": 58476 + }, + { + "epoch": 2.25, + "learning_rate": 1.2537189022826365e-05, + "loss": 0.201, + "step": 58477 + }, + { + "epoch": 2.25, + "learning_rate": 1.2536547832777636e-05, + "loss": 0.3315, + "step": 58478 + }, + { + "epoch": 2.25, + "learning_rate": 1.2535906642728906e-05, + "loss": 0.1448, + "step": 58479 + }, + { + "epoch": 2.25, + "learning_rate": 1.2535265452680175e-05, + "loss": 0.1843, + "step": 58480 + }, + { + "epoch": 2.25, + "learning_rate": 1.2534624262631445e-05, + "loss": 0.2621, + "step": 58481 + }, + { + "epoch": 2.25, + "learning_rate": 1.2533983072582715e-05, + "loss": 0.2242, + "step": 58482 + }, + { + "epoch": 2.25, + "learning_rate": 1.2533341882533983e-05, + "loss": 0.2873, + "step": 58483 + }, + { + "epoch": 2.25, + "learning_rate": 1.2532700692485253e-05, + "loss": 0.2682, + "step": 58484 + }, + { + "epoch": 2.25, + "learning_rate": 1.2532059502436521e-05, + "loss": 0.2083, + "step": 58485 + }, + { + "epoch": 2.25, + "learning_rate": 1.2531418312387792e-05, + "loss": 0.149, + "step": 58486 + }, + { + "epoch": 2.25, + "learning_rate": 1.2530777122339063e-05, + "loss": 0.2155, + "step": 58487 + }, + { + "epoch": 2.25, + "learning_rate": 1.2530135932290332e-05, + "loss": 0.1509, + "step": 58488 + }, + { + "epoch": 2.25, + "learning_rate": 1.2529494742241602e-05, + "loss": 0.2062, + "step": 58489 + }, + { + "epoch": 2.25, + "learning_rate": 1.252885355219287e-05, + "loss": 0.1539, + "step": 58490 + }, + { + "epoch": 2.25, + "learning_rate": 1.252821236214414e-05, + "loss": 0.139, + "step": 58491 + }, + { + "epoch": 2.25, + "learning_rate": 1.252757117209541e-05, + "loss": 0.2234, + "step": 58492 + }, + { + "epoch": 2.25, + "learning_rate": 1.2526929982046678e-05, + "loss": 0.1277, + "step": 58493 + }, + { + "epoch": 2.25, + "learning_rate": 1.2526288791997948e-05, + "loss": 0.2069, + "step": 58494 + }, + { + "epoch": 2.25, + "learning_rate": 1.252564760194922e-05, + "loss": 0.0989, + "step": 58495 + }, + { + "epoch": 2.25, + "learning_rate": 1.2525006411900488e-05, + "loss": 0.1944, + "step": 58496 + }, + { + "epoch": 2.25, + "learning_rate": 1.2524365221851758e-05, + "loss": 0.1869, + "step": 58497 + }, + { + "epoch": 2.25, + "learning_rate": 1.2523724031803027e-05, + "loss": 0.2959, + "step": 58498 + }, + { + "epoch": 2.25, + "learning_rate": 1.2523082841754297e-05, + "loss": 0.2491, + "step": 58499 + }, + { + "epoch": 2.25, + "learning_rate": 1.2522441651705567e-05, + "loss": 0.3044, + "step": 58500 + }, + { + "epoch": 2.25, + "learning_rate": 1.2521800461656835e-05, + "loss": 0.2115, + "step": 58501 + }, + { + "epoch": 2.25, + "learning_rate": 1.2521159271608105e-05, + "loss": 0.4179, + "step": 58502 + }, + { + "epoch": 2.25, + "learning_rate": 1.2520518081559373e-05, + "loss": 0.1182, + "step": 58503 + }, + { + "epoch": 2.25, + "learning_rate": 1.2519876891510645e-05, + "loss": 0.1795, + "step": 58504 + }, + { + "epoch": 2.25, + "learning_rate": 1.2519235701461915e-05, + "loss": 0.2725, + "step": 58505 + }, + { + "epoch": 2.25, + "learning_rate": 1.2518594511413183e-05, + "loss": 0.2679, + "step": 58506 + }, + { + "epoch": 2.25, + "learning_rate": 1.2517953321364454e-05, + "loss": 0.1682, + "step": 58507 + }, + { + "epoch": 2.25, + "learning_rate": 1.2517312131315722e-05, + "loss": 0.2003, + "step": 58508 + }, + { + "epoch": 2.25, + "learning_rate": 1.2516670941266992e-05, + "loss": 0.2408, + "step": 58509 + }, + { + "epoch": 2.25, + "learning_rate": 1.2516029751218262e-05, + "loss": 0.324, + "step": 58510 + }, + { + "epoch": 2.25, + "learning_rate": 1.251538856116953e-05, + "loss": 0.2585, + "step": 58511 + }, + { + "epoch": 2.25, + "learning_rate": 1.25147473711208e-05, + "loss": 0.3599, + "step": 58512 + }, + { + "epoch": 2.25, + "learning_rate": 1.2514106181072072e-05, + "loss": 0.2061, + "step": 58513 + }, + { + "epoch": 2.25, + "learning_rate": 1.251346499102334e-05, + "loss": 0.2442, + "step": 58514 + }, + { + "epoch": 2.25, + "learning_rate": 1.251282380097461e-05, + "loss": 0.2893, + "step": 58515 + }, + { + "epoch": 2.25, + "learning_rate": 1.2512182610925879e-05, + "loss": 0.1167, + "step": 58516 + }, + { + "epoch": 2.25, + "learning_rate": 1.2511541420877149e-05, + "loss": 0.3202, + "step": 58517 + }, + { + "epoch": 2.25, + "learning_rate": 1.2510900230828419e-05, + "loss": 0.1825, + "step": 58518 + }, + { + "epoch": 2.25, + "learning_rate": 1.2510259040779687e-05, + "loss": 0.1996, + "step": 58519 + }, + { + "epoch": 2.25, + "learning_rate": 1.2509617850730957e-05, + "loss": 0.2045, + "step": 58520 + }, + { + "epoch": 2.25, + "learning_rate": 1.2508976660682225e-05, + "loss": 0.3502, + "step": 58521 + }, + { + "epoch": 2.25, + "learning_rate": 1.2508335470633497e-05, + "loss": 0.3021, + "step": 58522 + }, + { + "epoch": 2.25, + "learning_rate": 1.2507694280584767e-05, + "loss": 0.1892, + "step": 58523 + }, + { + "epoch": 2.25, + "learning_rate": 1.2507053090536035e-05, + "loss": 0.1702, + "step": 58524 + }, + { + "epoch": 2.25, + "learning_rate": 1.2506411900487305e-05, + "loss": 0.2377, + "step": 58525 + }, + { + "epoch": 2.25, + "learning_rate": 1.2505770710438574e-05, + "loss": 0.2657, + "step": 58526 + }, + { + "epoch": 2.25, + "learning_rate": 1.2505129520389844e-05, + "loss": 0.2277, + "step": 58527 + }, + { + "epoch": 2.25, + "learning_rate": 1.2504488330341114e-05, + "loss": 0.1636, + "step": 58528 + }, + { + "epoch": 2.25, + "learning_rate": 1.2503847140292382e-05, + "loss": 0.2211, + "step": 58529 + }, + { + "epoch": 2.25, + "learning_rate": 1.2503205950243652e-05, + "loss": 0.2436, + "step": 58530 + }, + { + "epoch": 2.25, + "learning_rate": 1.2502564760194924e-05, + "loss": 0.1686, + "step": 58531 + }, + { + "epoch": 2.25, + "learning_rate": 1.2501923570146192e-05, + "loss": 0.3292, + "step": 58532 + }, + { + "epoch": 2.25, + "learning_rate": 1.2501282380097462e-05, + "loss": 0.1507, + "step": 58533 + }, + { + "epoch": 2.25, + "learning_rate": 1.250064119004873e-05, + "loss": 0.2515, + "step": 58534 + }, + { + "epoch": 2.25, + "learning_rate": 1.25e-05, + "loss": 0.2947, + "step": 58535 + }, + { + "epoch": 2.25, + "learning_rate": 1.249935880995127e-05, + "loss": 0.2203, + "step": 58536 + }, + { + "epoch": 2.25, + "learning_rate": 1.2498717619902539e-05, + "loss": 0.1907, + "step": 58537 + }, + { + "epoch": 2.25, + "learning_rate": 1.2498076429853809e-05, + "loss": 0.2229, + "step": 58538 + }, + { + "epoch": 2.25, + "learning_rate": 1.2497435239805079e-05, + "loss": 0.3219, + "step": 58539 + }, + { + "epoch": 2.25, + "learning_rate": 1.2496794049756349e-05, + "loss": 0.186, + "step": 58540 + }, + { + "epoch": 2.25, + "learning_rate": 1.2496152859707617e-05, + "loss": 0.155, + "step": 58541 + }, + { + "epoch": 2.25, + "learning_rate": 1.2495511669658887e-05, + "loss": 0.1444, + "step": 58542 + }, + { + "epoch": 2.25, + "learning_rate": 1.2494870479610157e-05, + "loss": 0.0999, + "step": 58543 + }, + { + "epoch": 2.25, + "learning_rate": 1.2494229289561427e-05, + "loss": 0.1441, + "step": 58544 + }, + { + "epoch": 2.25, + "learning_rate": 1.2493588099512696e-05, + "loss": 0.1102, + "step": 58545 + }, + { + "epoch": 2.25, + "learning_rate": 1.2492946909463966e-05, + "loss": 0.3465, + "step": 58546 + }, + { + "epoch": 2.25, + "learning_rate": 1.2492305719415236e-05, + "loss": 0.2281, + "step": 58547 + }, + { + "epoch": 2.25, + "learning_rate": 1.2491664529366504e-05, + "loss": 0.1677, + "step": 58548 + }, + { + "epoch": 2.25, + "learning_rate": 1.2491023339317774e-05, + "loss": 0.1498, + "step": 58549 + }, + { + "epoch": 2.25, + "learning_rate": 1.2490382149269044e-05, + "loss": 0.3735, + "step": 58550 + }, + { + "epoch": 2.25, + "learning_rate": 1.2489740959220314e-05, + "loss": 0.2018, + "step": 58551 + }, + { + "epoch": 2.25, + "learning_rate": 1.2489099769171583e-05, + "loss": 0.3032, + "step": 58552 + }, + { + "epoch": 2.25, + "learning_rate": 1.2488458579122853e-05, + "loss": 0.2084, + "step": 58553 + }, + { + "epoch": 2.25, + "learning_rate": 1.2487817389074123e-05, + "loss": 0.1583, + "step": 58554 + }, + { + "epoch": 2.25, + "learning_rate": 1.2487176199025391e-05, + "loss": 0.245, + "step": 58555 + }, + { + "epoch": 2.25, + "learning_rate": 1.2486535008976661e-05, + "loss": 0.1826, + "step": 58556 + }, + { + "epoch": 2.25, + "learning_rate": 1.2485893818927931e-05, + "loss": 0.2455, + "step": 58557 + }, + { + "epoch": 2.25, + "learning_rate": 1.2485252628879201e-05, + "loss": 0.4726, + "step": 58558 + }, + { + "epoch": 2.25, + "learning_rate": 1.248461143883047e-05, + "loss": 0.2164, + "step": 58559 + }, + { + "epoch": 2.25, + "learning_rate": 1.248397024878174e-05, + "loss": 0.2152, + "step": 58560 + }, + { + "epoch": 2.25, + "learning_rate": 1.248332905873301e-05, + "loss": 0.3478, + "step": 58561 + }, + { + "epoch": 2.25, + "learning_rate": 1.248268786868428e-05, + "loss": 0.2144, + "step": 58562 + }, + { + "epoch": 2.25, + "learning_rate": 1.2482046678635548e-05, + "loss": 0.2031, + "step": 58563 + }, + { + "epoch": 2.25, + "learning_rate": 1.2481405488586818e-05, + "loss": 0.2267, + "step": 58564 + }, + { + "epoch": 2.25, + "learning_rate": 1.2480764298538088e-05, + "loss": 0.2224, + "step": 58565 + }, + { + "epoch": 2.25, + "learning_rate": 1.2480123108489356e-05, + "loss": 0.3325, + "step": 58566 + }, + { + "epoch": 2.25, + "learning_rate": 1.2479481918440626e-05, + "loss": 0.2834, + "step": 58567 + }, + { + "epoch": 2.25, + "learning_rate": 1.2478840728391896e-05, + "loss": 0.1546, + "step": 58568 + }, + { + "epoch": 2.25, + "learning_rate": 1.2478199538343166e-05, + "loss": 0.3074, + "step": 58569 + }, + { + "epoch": 2.25, + "learning_rate": 1.2477558348294434e-05, + "loss": 0.1683, + "step": 58570 + }, + { + "epoch": 2.25, + "learning_rate": 1.2476917158245704e-05, + "loss": 0.2801, + "step": 58571 + }, + { + "epoch": 2.25, + "learning_rate": 1.2476275968196974e-05, + "loss": 0.2845, + "step": 58572 + }, + { + "epoch": 2.25, + "learning_rate": 1.2475634778148245e-05, + "loss": 0.4637, + "step": 58573 + }, + { + "epoch": 2.25, + "learning_rate": 1.2474993588099513e-05, + "loss": 0.2814, + "step": 58574 + }, + { + "epoch": 2.25, + "learning_rate": 1.2474352398050783e-05, + "loss": 0.1918, + "step": 58575 + }, + { + "epoch": 2.25, + "learning_rate": 1.2473711208002053e-05, + "loss": 0.2481, + "step": 58576 + }, + { + "epoch": 2.25, + "learning_rate": 1.2473070017953321e-05, + "loss": 0.2005, + "step": 58577 + }, + { + "epoch": 2.25, + "learning_rate": 1.2472428827904591e-05, + "loss": 0.302, + "step": 58578 + }, + { + "epoch": 2.25, + "learning_rate": 1.2471787637855861e-05, + "loss": 0.3043, + "step": 58579 + }, + { + "epoch": 2.25, + "learning_rate": 1.2471146447807131e-05, + "loss": 0.1776, + "step": 58580 + }, + { + "epoch": 2.25, + "learning_rate": 1.24705052577584e-05, + "loss": 0.1818, + "step": 58581 + }, + { + "epoch": 2.25, + "learning_rate": 1.246986406770967e-05, + "loss": 0.3801, + "step": 58582 + }, + { + "epoch": 2.25, + "learning_rate": 1.246922287766094e-05, + "loss": 0.1808, + "step": 58583 + }, + { + "epoch": 2.25, + "learning_rate": 1.246858168761221e-05, + "loss": 0.2784, + "step": 58584 + }, + { + "epoch": 2.25, + "learning_rate": 1.2467940497563478e-05, + "loss": 0.2183, + "step": 58585 + }, + { + "epoch": 2.25, + "learning_rate": 1.2467299307514748e-05, + "loss": 0.1584, + "step": 58586 + }, + { + "epoch": 2.25, + "learning_rate": 1.2466658117466018e-05, + "loss": 0.2813, + "step": 58587 + }, + { + "epoch": 2.25, + "learning_rate": 1.2466016927417286e-05, + "loss": 0.1396, + "step": 58588 + }, + { + "epoch": 2.25, + "learning_rate": 1.2465375737368556e-05, + "loss": 0.1217, + "step": 58589 + }, + { + "epoch": 2.25, + "learning_rate": 1.2464734547319826e-05, + "loss": 0.1994, + "step": 58590 + }, + { + "epoch": 2.25, + "learning_rate": 1.2464093357271096e-05, + "loss": 0.2379, + "step": 58591 + }, + { + "epoch": 2.25, + "learning_rate": 1.2463452167222365e-05, + "loss": 0.1881, + "step": 58592 + }, + { + "epoch": 2.25, + "learning_rate": 1.2462810977173635e-05, + "loss": 0.2679, + "step": 58593 + }, + { + "epoch": 2.25, + "learning_rate": 1.2462169787124905e-05, + "loss": 0.2086, + "step": 58594 + }, + { + "epoch": 2.25, + "learning_rate": 1.2461528597076173e-05, + "loss": 0.279, + "step": 58595 + }, + { + "epoch": 2.25, + "learning_rate": 1.2460887407027443e-05, + "loss": 0.1622, + "step": 58596 + }, + { + "epoch": 2.25, + "learning_rate": 1.2460246216978713e-05, + "loss": 0.1286, + "step": 58597 + }, + { + "epoch": 2.25, + "learning_rate": 1.2459605026929983e-05, + "loss": 0.3279, + "step": 58598 + }, + { + "epoch": 2.25, + "learning_rate": 1.2458963836881252e-05, + "loss": 0.1829, + "step": 58599 + }, + { + "epoch": 2.25, + "learning_rate": 1.2458322646832522e-05, + "loss": 0.2834, + "step": 58600 + }, + { + "epoch": 2.25, + "learning_rate": 1.2457681456783792e-05, + "loss": 0.1642, + "step": 58601 + }, + { + "epoch": 2.25, + "learning_rate": 1.2457040266735062e-05, + "loss": 0.1793, + "step": 58602 + }, + { + "epoch": 2.25, + "learning_rate": 1.245639907668633e-05, + "loss": 0.2945, + "step": 58603 + }, + { + "epoch": 2.25, + "learning_rate": 1.24557578866376e-05, + "loss": 0.1663, + "step": 58604 + }, + { + "epoch": 2.25, + "learning_rate": 1.245511669658887e-05, + "loss": 0.1023, + "step": 58605 + }, + { + "epoch": 2.25, + "learning_rate": 1.245447550654014e-05, + "loss": 0.2071, + "step": 58606 + }, + { + "epoch": 2.25, + "learning_rate": 1.2453834316491408e-05, + "loss": 0.1433, + "step": 58607 + }, + { + "epoch": 2.25, + "learning_rate": 1.2453193126442678e-05, + "loss": 0.1209, + "step": 58608 + }, + { + "epoch": 2.25, + "learning_rate": 1.2452551936393948e-05, + "loss": 0.2157, + "step": 58609 + }, + { + "epoch": 2.25, + "learning_rate": 1.2451910746345217e-05, + "loss": 0.1844, + "step": 58610 + }, + { + "epoch": 2.25, + "learning_rate": 1.2451269556296487e-05, + "loss": 0.1912, + "step": 58611 + }, + { + "epoch": 2.25, + "learning_rate": 1.2450628366247757e-05, + "loss": 0.3482, + "step": 58612 + }, + { + "epoch": 2.25, + "learning_rate": 1.2449987176199027e-05, + "loss": 0.1517, + "step": 58613 + }, + { + "epoch": 2.25, + "learning_rate": 1.2449345986150295e-05, + "loss": 0.2576, + "step": 58614 + }, + { + "epoch": 2.25, + "learning_rate": 1.2448704796101565e-05, + "loss": 0.3099, + "step": 58615 + }, + { + "epoch": 2.25, + "learning_rate": 1.2448063606052835e-05, + "loss": 0.1308, + "step": 58616 + }, + { + "epoch": 2.25, + "learning_rate": 1.2447422416004103e-05, + "loss": 0.1438, + "step": 58617 + }, + { + "epoch": 2.25, + "learning_rate": 1.2446781225955374e-05, + "loss": 0.1345, + "step": 58618 + }, + { + "epoch": 2.25, + "learning_rate": 1.2446140035906644e-05, + "loss": 0.1721, + "step": 58619 + }, + { + "epoch": 2.25, + "learning_rate": 1.2445498845857914e-05, + "loss": 0.3581, + "step": 58620 + }, + { + "epoch": 2.25, + "learning_rate": 1.2444857655809182e-05, + "loss": 0.38, + "step": 58621 + }, + { + "epoch": 2.25, + "learning_rate": 1.2444216465760452e-05, + "loss": 0.1652, + "step": 58622 + }, + { + "epoch": 2.25, + "learning_rate": 1.2443575275711722e-05, + "loss": 0.144, + "step": 58623 + }, + { + "epoch": 2.25, + "learning_rate": 1.2442934085662992e-05, + "loss": 0.2228, + "step": 58624 + }, + { + "epoch": 2.25, + "learning_rate": 1.244229289561426e-05, + "loss": 0.2333, + "step": 58625 + }, + { + "epoch": 2.25, + "learning_rate": 1.244165170556553e-05, + "loss": 0.2368, + "step": 58626 + }, + { + "epoch": 2.25, + "learning_rate": 1.24410105155168e-05, + "loss": 0.2125, + "step": 58627 + }, + { + "epoch": 2.25, + "learning_rate": 1.2440369325468069e-05, + "loss": 0.3621, + "step": 58628 + }, + { + "epoch": 2.25, + "learning_rate": 1.2439728135419339e-05, + "loss": 0.3354, + "step": 58629 + }, + { + "epoch": 2.25, + "learning_rate": 1.2439086945370609e-05, + "loss": 0.0781, + "step": 58630 + }, + { + "epoch": 2.25, + "learning_rate": 1.2438445755321879e-05, + "loss": 0.2179, + "step": 58631 + }, + { + "epoch": 2.25, + "learning_rate": 1.2437804565273147e-05, + "loss": 0.1473, + "step": 58632 + }, + { + "epoch": 2.25, + "learning_rate": 1.2437163375224417e-05, + "loss": 0.1764, + "step": 58633 + }, + { + "epoch": 2.25, + "learning_rate": 1.2436522185175687e-05, + "loss": 0.2445, + "step": 58634 + }, + { + "epoch": 2.25, + "learning_rate": 1.2435880995126955e-05, + "loss": 0.2182, + "step": 58635 + }, + { + "epoch": 2.25, + "learning_rate": 1.2435239805078225e-05, + "loss": 0.2, + "step": 58636 + }, + { + "epoch": 2.25, + "learning_rate": 1.2434598615029495e-05, + "loss": 0.2318, + "step": 58637 + }, + { + "epoch": 2.25, + "learning_rate": 1.2433957424980765e-05, + "loss": 0.1967, + "step": 58638 + }, + { + "epoch": 2.25, + "learning_rate": 1.2433316234932034e-05, + "loss": 0.2593, + "step": 58639 + }, + { + "epoch": 2.25, + "learning_rate": 1.2432675044883304e-05, + "loss": 0.1769, + "step": 58640 + }, + { + "epoch": 2.25, + "learning_rate": 1.2432033854834574e-05, + "loss": 0.21, + "step": 58641 + }, + { + "epoch": 2.25, + "learning_rate": 1.2431392664785844e-05, + "loss": 0.2429, + "step": 58642 + }, + { + "epoch": 2.25, + "learning_rate": 1.2430751474737112e-05, + "loss": 0.1999, + "step": 58643 + }, + { + "epoch": 2.25, + "learning_rate": 1.2430110284688382e-05, + "loss": 0.1357, + "step": 58644 + }, + { + "epoch": 2.25, + "learning_rate": 1.2429469094639652e-05, + "loss": 0.2767, + "step": 58645 + }, + { + "epoch": 2.25, + "learning_rate": 1.2428827904590922e-05, + "loss": 0.1521, + "step": 58646 + }, + { + "epoch": 2.25, + "learning_rate": 1.242818671454219e-05, + "loss": 0.2649, + "step": 58647 + }, + { + "epoch": 2.25, + "learning_rate": 1.242754552449346e-05, + "loss": 0.2814, + "step": 58648 + }, + { + "epoch": 2.25, + "learning_rate": 1.242690433444473e-05, + "loss": 0.1343, + "step": 58649 + }, + { + "epoch": 2.25, + "learning_rate": 1.2426263144395999e-05, + "loss": 0.3079, + "step": 58650 + }, + { + "epoch": 2.25, + "learning_rate": 1.2425621954347269e-05, + "loss": 0.21, + "step": 58651 + }, + { + "epoch": 2.25, + "learning_rate": 1.2424980764298539e-05, + "loss": 0.269, + "step": 58652 + }, + { + "epoch": 2.26, + "learning_rate": 1.2424339574249807e-05, + "loss": 0.1754, + "step": 58653 + }, + { + "epoch": 2.26, + "learning_rate": 1.2423698384201077e-05, + "loss": 0.3223, + "step": 58654 + }, + { + "epoch": 2.26, + "learning_rate": 1.2423057194152347e-05, + "loss": 0.2371, + "step": 58655 + }, + { + "epoch": 2.26, + "learning_rate": 1.2422416004103617e-05, + "loss": 0.2191, + "step": 58656 + }, + { + "epoch": 2.26, + "learning_rate": 1.2421774814054886e-05, + "loss": 0.2655, + "step": 58657 + }, + { + "epoch": 2.26, + "learning_rate": 1.2421133624006156e-05, + "loss": 0.1267, + "step": 58658 + }, + { + "epoch": 2.26, + "learning_rate": 1.2420492433957426e-05, + "loss": 0.3212, + "step": 58659 + }, + { + "epoch": 2.26, + "learning_rate": 1.2419851243908696e-05, + "loss": 0.1905, + "step": 58660 + }, + { + "epoch": 2.26, + "learning_rate": 1.2419210053859964e-05, + "loss": 0.3356, + "step": 58661 + }, + { + "epoch": 2.26, + "learning_rate": 1.2418568863811234e-05, + "loss": 0.235, + "step": 58662 + }, + { + "epoch": 2.26, + "learning_rate": 1.2417927673762504e-05, + "loss": 0.2561, + "step": 58663 + }, + { + "epoch": 2.26, + "learning_rate": 1.2417286483713774e-05, + "loss": 0.1762, + "step": 58664 + }, + { + "epoch": 2.26, + "learning_rate": 1.2416645293665043e-05, + "loss": 0.2069, + "step": 58665 + }, + { + "epoch": 2.26, + "learning_rate": 1.2416004103616313e-05, + "loss": 0.2405, + "step": 58666 + }, + { + "epoch": 2.26, + "learning_rate": 1.2415362913567583e-05, + "loss": 0.2659, + "step": 58667 + }, + { + "epoch": 2.26, + "learning_rate": 1.2414721723518851e-05, + "loss": 0.1714, + "step": 58668 + }, + { + "epoch": 2.26, + "learning_rate": 1.2414080533470121e-05, + "loss": 0.2672, + "step": 58669 + }, + { + "epoch": 2.26, + "learning_rate": 1.2413439343421391e-05, + "loss": 0.3042, + "step": 58670 + }, + { + "epoch": 2.26, + "learning_rate": 1.2412798153372661e-05, + "loss": 0.1505, + "step": 58671 + }, + { + "epoch": 2.26, + "learning_rate": 1.241215696332393e-05, + "loss": 0.1975, + "step": 58672 + }, + { + "epoch": 2.26, + "learning_rate": 1.24115157732752e-05, + "loss": 0.2622, + "step": 58673 + }, + { + "epoch": 2.26, + "learning_rate": 1.241087458322647e-05, + "loss": 0.1693, + "step": 58674 + }, + { + "epoch": 2.26, + "learning_rate": 1.2410233393177738e-05, + "loss": 0.2717, + "step": 58675 + }, + { + "epoch": 2.26, + "learning_rate": 1.2409592203129008e-05, + "loss": 0.4573, + "step": 58676 + }, + { + "epoch": 2.26, + "learning_rate": 1.2408951013080278e-05, + "loss": 0.2144, + "step": 58677 + }, + { + "epoch": 2.26, + "learning_rate": 1.2408309823031548e-05, + "loss": 0.2277, + "step": 58678 + }, + { + "epoch": 2.26, + "learning_rate": 1.2407668632982816e-05, + "loss": 0.1602, + "step": 58679 + }, + { + "epoch": 2.26, + "learning_rate": 1.2407027442934086e-05, + "loss": 0.3053, + "step": 58680 + }, + { + "epoch": 2.26, + "learning_rate": 1.2406386252885356e-05, + "loss": 0.3944, + "step": 58681 + }, + { + "epoch": 2.26, + "learning_rate": 1.2405745062836626e-05, + "loss": 0.1734, + "step": 58682 + }, + { + "epoch": 2.26, + "learning_rate": 1.2405103872787894e-05, + "loss": 0.2629, + "step": 58683 + }, + { + "epoch": 2.26, + "learning_rate": 1.2404462682739165e-05, + "loss": 0.2725, + "step": 58684 + }, + { + "epoch": 2.26, + "learning_rate": 1.2403821492690435e-05, + "loss": 0.2107, + "step": 58685 + }, + { + "epoch": 2.26, + "learning_rate": 1.2403180302641705e-05, + "loss": 0.2706, + "step": 58686 + }, + { + "epoch": 2.26, + "learning_rate": 1.2402539112592973e-05, + "loss": 0.4167, + "step": 58687 + }, + { + "epoch": 2.26, + "learning_rate": 1.2401897922544243e-05, + "loss": 0.2655, + "step": 58688 + }, + { + "epoch": 2.26, + "learning_rate": 1.2401256732495513e-05, + "loss": 0.146, + "step": 58689 + }, + { + "epoch": 2.26, + "learning_rate": 1.2400615542446781e-05, + "loss": 0.2521, + "step": 58690 + }, + { + "epoch": 2.26, + "learning_rate": 1.2399974352398051e-05, + "loss": 0.2891, + "step": 58691 + }, + { + "epoch": 2.26, + "learning_rate": 1.2399333162349321e-05, + "loss": 0.2418, + "step": 58692 + }, + { + "epoch": 2.26, + "learning_rate": 1.239869197230059e-05, + "loss": 0.1541, + "step": 58693 + }, + { + "epoch": 2.26, + "learning_rate": 1.239805078225186e-05, + "loss": 0.2772, + "step": 58694 + }, + { + "epoch": 2.26, + "learning_rate": 1.239740959220313e-05, + "loss": 0.2348, + "step": 58695 + }, + { + "epoch": 2.26, + "learning_rate": 1.23967684021544e-05, + "loss": 0.297, + "step": 58696 + }, + { + "epoch": 2.26, + "learning_rate": 1.2396127212105668e-05, + "loss": 0.207, + "step": 58697 + }, + { + "epoch": 2.26, + "learning_rate": 1.2395486022056938e-05, + "loss": 0.4046, + "step": 58698 + }, + { + "epoch": 2.26, + "learning_rate": 1.2394844832008208e-05, + "loss": 0.2355, + "step": 58699 + }, + { + "epoch": 2.26, + "learning_rate": 1.2394203641959478e-05, + "loss": 0.2036, + "step": 58700 + }, + { + "epoch": 2.26, + "learning_rate": 1.2393562451910746e-05, + "loss": 0.2106, + "step": 58701 + }, + { + "epoch": 2.26, + "learning_rate": 1.2392921261862016e-05, + "loss": 0.3306, + "step": 58702 + }, + { + "epoch": 2.26, + "learning_rate": 1.2392280071813286e-05, + "loss": 0.1673, + "step": 58703 + }, + { + "epoch": 2.26, + "learning_rate": 1.2391638881764556e-05, + "loss": 0.137, + "step": 58704 + }, + { + "epoch": 2.26, + "learning_rate": 1.2390997691715825e-05, + "loss": 0.1479, + "step": 58705 + }, + { + "epoch": 2.26, + "learning_rate": 1.2390356501667095e-05, + "loss": 0.3736, + "step": 58706 + }, + { + "epoch": 2.26, + "learning_rate": 1.2389715311618365e-05, + "loss": 0.1161, + "step": 58707 + }, + { + "epoch": 2.26, + "learning_rate": 1.2389074121569635e-05, + "loss": 0.2858, + "step": 58708 + }, + { + "epoch": 2.26, + "learning_rate": 1.2388432931520903e-05, + "loss": 0.2602, + "step": 58709 + }, + { + "epoch": 2.26, + "learning_rate": 1.2387791741472173e-05, + "loss": 0.3267, + "step": 58710 + }, + { + "epoch": 2.26, + "learning_rate": 1.2387150551423443e-05, + "loss": 0.2443, + "step": 58711 + }, + { + "epoch": 2.26, + "learning_rate": 1.2386509361374712e-05, + "loss": 0.133, + "step": 58712 + }, + { + "epoch": 2.26, + "learning_rate": 1.2385868171325982e-05, + "loss": 0.1638, + "step": 58713 + }, + { + "epoch": 2.26, + "learning_rate": 1.2385226981277252e-05, + "loss": 0.2761, + "step": 58714 + }, + { + "epoch": 2.26, + "learning_rate": 1.238458579122852e-05, + "loss": 0.1801, + "step": 58715 + }, + { + "epoch": 2.26, + "learning_rate": 1.238394460117979e-05, + "loss": 0.1653, + "step": 58716 + }, + { + "epoch": 2.26, + "learning_rate": 1.238330341113106e-05, + "loss": 0.2586, + "step": 58717 + }, + { + "epoch": 2.26, + "learning_rate": 1.238266222108233e-05, + "loss": 0.3705, + "step": 58718 + }, + { + "epoch": 2.26, + "learning_rate": 1.2382021031033598e-05, + "loss": 0.2731, + "step": 58719 + }, + { + "epoch": 2.26, + "learning_rate": 1.2381379840984868e-05, + "loss": 0.2745, + "step": 58720 + }, + { + "epoch": 2.26, + "learning_rate": 1.2380738650936138e-05, + "loss": 0.2753, + "step": 58721 + }, + { + "epoch": 2.26, + "learning_rate": 1.2380097460887408e-05, + "loss": 0.1565, + "step": 58722 + }, + { + "epoch": 2.26, + "learning_rate": 1.2379456270838677e-05, + "loss": 0.3494, + "step": 58723 + }, + { + "epoch": 2.26, + "learning_rate": 1.2378815080789947e-05, + "loss": 0.2883, + "step": 58724 + }, + { + "epoch": 2.26, + "learning_rate": 1.2378173890741217e-05, + "loss": 0.3671, + "step": 58725 + }, + { + "epoch": 2.26, + "learning_rate": 1.2377532700692487e-05, + "loss": 0.2153, + "step": 58726 + }, + { + "epoch": 2.26, + "learning_rate": 1.2376891510643755e-05, + "loss": 0.2761, + "step": 58727 + }, + { + "epoch": 2.26, + "learning_rate": 1.2376250320595025e-05, + "loss": 0.1581, + "step": 58728 + }, + { + "epoch": 2.26, + "learning_rate": 1.2375609130546295e-05, + "loss": 0.2268, + "step": 58729 + }, + { + "epoch": 2.26, + "learning_rate": 1.2374967940497564e-05, + "loss": 0.3059, + "step": 58730 + }, + { + "epoch": 2.26, + "learning_rate": 1.2374326750448834e-05, + "loss": 0.2825, + "step": 58731 + }, + { + "epoch": 2.26, + "learning_rate": 1.2373685560400104e-05, + "loss": 0.223, + "step": 58732 + }, + { + "epoch": 2.26, + "learning_rate": 1.2373044370351372e-05, + "loss": 0.2195, + "step": 58733 + }, + { + "epoch": 2.26, + "learning_rate": 1.2372403180302642e-05, + "loss": 0.267, + "step": 58734 + }, + { + "epoch": 2.26, + "learning_rate": 1.2371761990253912e-05, + "loss": 0.2431, + "step": 58735 + }, + { + "epoch": 2.26, + "learning_rate": 1.2371120800205182e-05, + "loss": 0.2799, + "step": 58736 + }, + { + "epoch": 2.26, + "learning_rate": 1.237047961015645e-05, + "loss": 0.2163, + "step": 58737 + }, + { + "epoch": 2.26, + "learning_rate": 1.236983842010772e-05, + "loss": 0.1818, + "step": 58738 + }, + { + "epoch": 2.26, + "learning_rate": 1.236919723005899e-05, + "loss": 0.1173, + "step": 58739 + }, + { + "epoch": 2.26, + "learning_rate": 1.236855604001026e-05, + "loss": 0.2814, + "step": 58740 + }, + { + "epoch": 2.26, + "learning_rate": 1.2367914849961529e-05, + "loss": 0.1618, + "step": 58741 + }, + { + "epoch": 2.26, + "learning_rate": 1.2367273659912799e-05, + "loss": 0.1379, + "step": 58742 + }, + { + "epoch": 2.26, + "learning_rate": 1.2366632469864069e-05, + "loss": 0.2123, + "step": 58743 + }, + { + "epoch": 2.26, + "learning_rate": 1.2365991279815339e-05, + "loss": 0.2802, + "step": 58744 + }, + { + "epoch": 2.26, + "learning_rate": 1.2365350089766607e-05, + "loss": 0.2772, + "step": 58745 + }, + { + "epoch": 2.26, + "learning_rate": 1.2364708899717877e-05, + "loss": 0.216, + "step": 58746 + }, + { + "epoch": 2.26, + "learning_rate": 1.2364067709669147e-05, + "loss": 0.1108, + "step": 58747 + }, + { + "epoch": 2.26, + "learning_rate": 1.2363426519620417e-05, + "loss": 0.3813, + "step": 58748 + }, + { + "epoch": 2.26, + "learning_rate": 1.2362785329571685e-05, + "loss": 0.1373, + "step": 58749 + }, + { + "epoch": 2.26, + "learning_rate": 1.2362144139522955e-05, + "loss": 0.3364, + "step": 58750 + }, + { + "epoch": 2.26, + "learning_rate": 1.2361502949474226e-05, + "loss": 0.2691, + "step": 58751 + }, + { + "epoch": 2.26, + "learning_rate": 1.2360861759425494e-05, + "loss": 0.2258, + "step": 58752 + }, + { + "epoch": 2.26, + "learning_rate": 1.2360220569376764e-05, + "loss": 0.2075, + "step": 58753 + }, + { + "epoch": 2.26, + "learning_rate": 1.2359579379328034e-05, + "loss": 0.1914, + "step": 58754 + }, + { + "epoch": 2.26, + "learning_rate": 1.2358938189279302e-05, + "loss": 0.2372, + "step": 58755 + }, + { + "epoch": 2.26, + "learning_rate": 1.2358296999230572e-05, + "loss": 0.1728, + "step": 58756 + }, + { + "epoch": 2.26, + "learning_rate": 1.2357655809181842e-05, + "loss": 0.127, + "step": 58757 + }, + { + "epoch": 2.26, + "learning_rate": 1.2357014619133112e-05, + "loss": 0.1934, + "step": 58758 + }, + { + "epoch": 2.26, + "learning_rate": 1.235637342908438e-05, + "loss": 0.3692, + "step": 58759 + }, + { + "epoch": 2.26, + "learning_rate": 1.235573223903565e-05, + "loss": 0.3215, + "step": 58760 + }, + { + "epoch": 2.26, + "learning_rate": 1.235509104898692e-05, + "loss": 0.1511, + "step": 58761 + }, + { + "epoch": 2.26, + "learning_rate": 1.235444985893819e-05, + "loss": 0.1723, + "step": 58762 + }, + { + "epoch": 2.26, + "learning_rate": 1.2353808668889459e-05, + "loss": 0.3014, + "step": 58763 + }, + { + "epoch": 2.26, + "learning_rate": 1.2353167478840729e-05, + "loss": 0.32, + "step": 58764 + }, + { + "epoch": 2.26, + "learning_rate": 1.2352526288791999e-05, + "loss": 0.28, + "step": 58765 + }, + { + "epoch": 2.26, + "learning_rate": 1.2351885098743269e-05, + "loss": 0.3168, + "step": 58766 + }, + { + "epoch": 2.26, + "learning_rate": 1.2351243908694537e-05, + "loss": 0.281, + "step": 58767 + }, + { + "epoch": 2.26, + "learning_rate": 1.2350602718645807e-05, + "loss": 0.2547, + "step": 58768 + }, + { + "epoch": 2.26, + "learning_rate": 1.2349961528597077e-05, + "loss": 0.2093, + "step": 58769 + }, + { + "epoch": 2.26, + "learning_rate": 1.2349320338548347e-05, + "loss": 0.2149, + "step": 58770 + }, + { + "epoch": 2.26, + "learning_rate": 1.2348679148499616e-05, + "loss": 0.1166, + "step": 58771 + }, + { + "epoch": 2.26, + "learning_rate": 1.2348037958450886e-05, + "loss": 0.4532, + "step": 58772 + }, + { + "epoch": 2.26, + "learning_rate": 1.2347396768402154e-05, + "loss": 0.303, + "step": 58773 + }, + { + "epoch": 2.26, + "learning_rate": 1.2346755578353424e-05, + "loss": 0.1981, + "step": 58774 + }, + { + "epoch": 2.26, + "learning_rate": 1.2346114388304694e-05, + "loss": 0.2316, + "step": 58775 + }, + { + "epoch": 2.26, + "learning_rate": 1.2345473198255964e-05, + "loss": 0.3598, + "step": 58776 + }, + { + "epoch": 2.26, + "learning_rate": 1.2344832008207233e-05, + "loss": 0.201, + "step": 58777 + }, + { + "epoch": 2.26, + "learning_rate": 1.2344190818158503e-05, + "loss": 0.255, + "step": 58778 + }, + { + "epoch": 2.26, + "learning_rate": 1.2343549628109773e-05, + "loss": 0.2646, + "step": 58779 + }, + { + "epoch": 2.26, + "learning_rate": 1.2342908438061043e-05, + "loss": 0.2971, + "step": 58780 + }, + { + "epoch": 2.26, + "learning_rate": 1.2342267248012311e-05, + "loss": 0.1505, + "step": 58781 + }, + { + "epoch": 2.26, + "learning_rate": 1.2341626057963581e-05, + "loss": 0.2519, + "step": 58782 + }, + { + "epoch": 2.26, + "learning_rate": 1.2340984867914851e-05, + "loss": 0.3618, + "step": 58783 + }, + { + "epoch": 2.26, + "learning_rate": 1.2340343677866121e-05, + "loss": 0.3098, + "step": 58784 + }, + { + "epoch": 2.26, + "learning_rate": 1.233970248781739e-05, + "loss": 0.1943, + "step": 58785 + }, + { + "epoch": 2.26, + "learning_rate": 1.233906129776866e-05, + "loss": 0.1777, + "step": 58786 + }, + { + "epoch": 2.26, + "learning_rate": 1.233842010771993e-05, + "loss": 0.3707, + "step": 58787 + }, + { + "epoch": 2.26, + "learning_rate": 1.23377789176712e-05, + "loss": 0.2654, + "step": 58788 + }, + { + "epoch": 2.26, + "learning_rate": 1.2337137727622468e-05, + "loss": 0.259, + "step": 58789 + }, + { + "epoch": 2.26, + "learning_rate": 1.2336496537573738e-05, + "loss": 0.1336, + "step": 58790 + }, + { + "epoch": 2.26, + "learning_rate": 1.2335855347525006e-05, + "loss": 0.1517, + "step": 58791 + }, + { + "epoch": 2.26, + "learning_rate": 1.2335214157476276e-05, + "loss": 0.2543, + "step": 58792 + }, + { + "epoch": 2.26, + "learning_rate": 1.2334572967427546e-05, + "loss": 0.1386, + "step": 58793 + }, + { + "epoch": 2.26, + "learning_rate": 1.2333931777378816e-05, + "loss": 0.1904, + "step": 58794 + }, + { + "epoch": 2.26, + "learning_rate": 1.2333290587330084e-05, + "loss": 0.4618, + "step": 58795 + }, + { + "epoch": 2.26, + "learning_rate": 1.2332649397281355e-05, + "loss": 0.2775, + "step": 58796 + }, + { + "epoch": 2.26, + "learning_rate": 1.2332008207232625e-05, + "loss": 0.2457, + "step": 58797 + }, + { + "epoch": 2.26, + "learning_rate": 1.2331367017183895e-05, + "loss": 0.2632, + "step": 58798 + }, + { + "epoch": 2.26, + "learning_rate": 1.2330725827135163e-05, + "loss": 0.4107, + "step": 58799 + }, + { + "epoch": 2.26, + "learning_rate": 1.2330084637086433e-05, + "loss": 0.2319, + "step": 58800 + }, + { + "epoch": 2.26, + "learning_rate": 1.2329443447037703e-05, + "loss": 0.1788, + "step": 58801 + }, + { + "epoch": 2.26, + "learning_rate": 1.2328802256988973e-05, + "loss": 0.3075, + "step": 58802 + }, + { + "epoch": 2.26, + "learning_rate": 1.2328161066940241e-05, + "loss": 0.1475, + "step": 58803 + }, + { + "epoch": 2.26, + "learning_rate": 1.2327519876891511e-05, + "loss": 0.2014, + "step": 58804 + }, + { + "epoch": 2.26, + "learning_rate": 1.2326878686842781e-05, + "loss": 0.3328, + "step": 58805 + }, + { + "epoch": 2.26, + "learning_rate": 1.2326237496794051e-05, + "loss": 0.3103, + "step": 58806 + }, + { + "epoch": 2.26, + "learning_rate": 1.232559630674532e-05, + "loss": 0.3054, + "step": 58807 + }, + { + "epoch": 2.26, + "learning_rate": 1.232495511669659e-05, + "loss": 0.1785, + "step": 58808 + }, + { + "epoch": 2.26, + "learning_rate": 1.232431392664786e-05, + "loss": 0.3094, + "step": 58809 + }, + { + "epoch": 2.26, + "learning_rate": 1.232367273659913e-05, + "loss": 0.2438, + "step": 58810 + }, + { + "epoch": 2.26, + "learning_rate": 1.2323031546550398e-05, + "loss": 0.202, + "step": 58811 + }, + { + "epoch": 2.26, + "learning_rate": 1.2322390356501668e-05, + "loss": 0.165, + "step": 58812 + }, + { + "epoch": 2.26, + "learning_rate": 1.2321749166452936e-05, + "loss": 0.2941, + "step": 58813 + }, + { + "epoch": 2.26, + "learning_rate": 1.2321107976404206e-05, + "loss": 0.2324, + "step": 58814 + }, + { + "epoch": 2.26, + "learning_rate": 1.2320466786355476e-05, + "loss": 0.1963, + "step": 58815 + }, + { + "epoch": 2.26, + "learning_rate": 1.2319825596306746e-05, + "loss": 0.286, + "step": 58816 + }, + { + "epoch": 2.26, + "learning_rate": 1.2319184406258015e-05, + "loss": 0.2825, + "step": 58817 + }, + { + "epoch": 2.26, + "learning_rate": 1.2318543216209285e-05, + "loss": 0.2128, + "step": 58818 + }, + { + "epoch": 2.26, + "learning_rate": 1.2317902026160555e-05, + "loss": 0.249, + "step": 58819 + }, + { + "epoch": 2.26, + "learning_rate": 1.2317260836111825e-05, + "loss": 0.1596, + "step": 58820 + }, + { + "epoch": 2.26, + "learning_rate": 1.2316619646063093e-05, + "loss": 0.2268, + "step": 58821 + }, + { + "epoch": 2.26, + "learning_rate": 1.2315978456014363e-05, + "loss": 0.189, + "step": 58822 + }, + { + "epoch": 2.26, + "learning_rate": 1.2315337265965633e-05, + "loss": 0.1759, + "step": 58823 + }, + { + "epoch": 2.26, + "learning_rate": 1.2314696075916903e-05, + "loss": 0.262, + "step": 58824 + }, + { + "epoch": 2.26, + "learning_rate": 1.2314054885868172e-05, + "loss": 0.3047, + "step": 58825 + }, + { + "epoch": 2.26, + "learning_rate": 1.2313413695819442e-05, + "loss": 0.2717, + "step": 58826 + }, + { + "epoch": 2.26, + "learning_rate": 1.2312772505770712e-05, + "loss": 0.1428, + "step": 58827 + }, + { + "epoch": 2.26, + "learning_rate": 1.2312131315721982e-05, + "loss": 0.1261, + "step": 58828 + }, + { + "epoch": 2.26, + "learning_rate": 1.231149012567325e-05, + "loss": 0.1794, + "step": 58829 + }, + { + "epoch": 2.26, + "learning_rate": 1.231084893562452e-05, + "loss": 0.1873, + "step": 58830 + }, + { + "epoch": 2.26, + "learning_rate": 1.2310207745575788e-05, + "loss": 0.2313, + "step": 58831 + }, + { + "epoch": 2.26, + "learning_rate": 1.2309566555527058e-05, + "loss": 0.2823, + "step": 58832 + }, + { + "epoch": 2.26, + "learning_rate": 1.2308925365478328e-05, + "loss": 0.2384, + "step": 58833 + }, + { + "epoch": 2.26, + "learning_rate": 1.2308284175429598e-05, + "loss": 0.3023, + "step": 58834 + }, + { + "epoch": 2.26, + "learning_rate": 1.2307642985380867e-05, + "loss": 0.2523, + "step": 58835 + }, + { + "epoch": 2.26, + "learning_rate": 1.2307001795332137e-05, + "loss": 0.4037, + "step": 58836 + }, + { + "epoch": 2.26, + "learning_rate": 1.2306360605283407e-05, + "loss": 0.1987, + "step": 58837 + }, + { + "epoch": 2.26, + "learning_rate": 1.2305719415234677e-05, + "loss": 0.1386, + "step": 58838 + }, + { + "epoch": 2.26, + "learning_rate": 1.2305078225185945e-05, + "loss": 0.2116, + "step": 58839 + }, + { + "epoch": 2.26, + "learning_rate": 1.2304437035137215e-05, + "loss": 0.1531, + "step": 58840 + }, + { + "epoch": 2.26, + "learning_rate": 1.2303795845088485e-05, + "loss": 0.2322, + "step": 58841 + }, + { + "epoch": 2.26, + "learning_rate": 1.2303154655039755e-05, + "loss": 0.2175, + "step": 58842 + }, + { + "epoch": 2.26, + "learning_rate": 1.2302513464991024e-05, + "loss": 0.3381, + "step": 58843 + }, + { + "epoch": 2.26, + "learning_rate": 1.2301872274942294e-05, + "loss": 0.1977, + "step": 58844 + }, + { + "epoch": 2.26, + "learning_rate": 1.2301231084893564e-05, + "loss": 0.1465, + "step": 58845 + }, + { + "epoch": 2.26, + "learning_rate": 1.2300589894844834e-05, + "loss": 0.1972, + "step": 58846 + }, + { + "epoch": 2.26, + "learning_rate": 1.2299948704796102e-05, + "loss": 0.3266, + "step": 58847 + }, + { + "epoch": 2.26, + "learning_rate": 1.2299307514747372e-05, + "loss": 0.1912, + "step": 58848 + }, + { + "epoch": 2.26, + "learning_rate": 1.2298666324698642e-05, + "loss": 0.1532, + "step": 58849 + }, + { + "epoch": 2.26, + "learning_rate": 1.2298025134649912e-05, + "loss": 0.2727, + "step": 58850 + }, + { + "epoch": 2.26, + "learning_rate": 1.229738394460118e-05, + "loss": 0.1933, + "step": 58851 + }, + { + "epoch": 2.26, + "learning_rate": 1.229674275455245e-05, + "loss": 0.2588, + "step": 58852 + }, + { + "epoch": 2.26, + "learning_rate": 1.2296101564503719e-05, + "loss": 0.3048, + "step": 58853 + }, + { + "epoch": 2.26, + "learning_rate": 1.2295460374454989e-05, + "loss": 0.1916, + "step": 58854 + }, + { + "epoch": 2.26, + "learning_rate": 1.2294819184406259e-05, + "loss": 0.3076, + "step": 58855 + }, + { + "epoch": 2.26, + "learning_rate": 1.2294177994357529e-05, + "loss": 0.168, + "step": 58856 + }, + { + "epoch": 2.26, + "learning_rate": 1.2293536804308797e-05, + "loss": 0.1876, + "step": 58857 + }, + { + "epoch": 2.26, + "learning_rate": 1.2292895614260067e-05, + "loss": 0.2124, + "step": 58858 + }, + { + "epoch": 2.26, + "learning_rate": 1.2292254424211337e-05, + "loss": 0.212, + "step": 58859 + }, + { + "epoch": 2.26, + "learning_rate": 1.2291613234162607e-05, + "loss": 0.3196, + "step": 58860 + }, + { + "epoch": 2.26, + "learning_rate": 1.2290972044113875e-05, + "loss": 0.2761, + "step": 58861 + }, + { + "epoch": 2.26, + "learning_rate": 1.2290330854065146e-05, + "loss": 0.196, + "step": 58862 + }, + { + "epoch": 2.26, + "learning_rate": 1.2289689664016416e-05, + "loss": 0.333, + "step": 58863 + }, + { + "epoch": 2.26, + "learning_rate": 1.2289048473967686e-05, + "loss": 0.293, + "step": 58864 + }, + { + "epoch": 2.26, + "learning_rate": 1.2288407283918954e-05, + "loss": 0.2669, + "step": 58865 + }, + { + "epoch": 2.26, + "learning_rate": 1.2287766093870224e-05, + "loss": 0.228, + "step": 58866 + }, + { + "epoch": 2.26, + "learning_rate": 1.2287124903821494e-05, + "loss": 0.2849, + "step": 58867 + }, + { + "epoch": 2.26, + "learning_rate": 1.2286483713772764e-05, + "loss": 0.2663, + "step": 58868 + }, + { + "epoch": 2.26, + "learning_rate": 1.2285842523724032e-05, + "loss": 0.2869, + "step": 58869 + }, + { + "epoch": 2.26, + "learning_rate": 1.2285201333675302e-05, + "loss": 0.2729, + "step": 58870 + }, + { + "epoch": 2.26, + "learning_rate": 1.228456014362657e-05, + "loss": 0.2076, + "step": 58871 + }, + { + "epoch": 2.26, + "learning_rate": 1.2283918953577842e-05, + "loss": 0.1726, + "step": 58872 + }, + { + "epoch": 2.26, + "learning_rate": 1.228327776352911e-05, + "loss": 0.2094, + "step": 58873 + }, + { + "epoch": 2.26, + "learning_rate": 1.228263657348038e-05, + "loss": 0.3021, + "step": 58874 + }, + { + "epoch": 2.26, + "learning_rate": 1.2281995383431649e-05, + "loss": 0.2135, + "step": 58875 + }, + { + "epoch": 2.26, + "learning_rate": 1.2281354193382919e-05, + "loss": 0.2585, + "step": 58876 + }, + { + "epoch": 2.26, + "learning_rate": 1.2280713003334189e-05, + "loss": 0.1848, + "step": 58877 + }, + { + "epoch": 2.26, + "learning_rate": 1.2280071813285459e-05, + "loss": 0.1735, + "step": 58878 + }, + { + "epoch": 2.26, + "learning_rate": 1.2279430623236727e-05, + "loss": 0.3011, + "step": 58879 + }, + { + "epoch": 2.26, + "learning_rate": 1.2278789433187997e-05, + "loss": 0.1741, + "step": 58880 + }, + { + "epoch": 2.26, + "learning_rate": 1.2278148243139267e-05, + "loss": 0.2645, + "step": 58881 + }, + { + "epoch": 2.26, + "learning_rate": 1.2277507053090537e-05, + "loss": 0.2434, + "step": 58882 + }, + { + "epoch": 2.26, + "learning_rate": 1.2276865863041806e-05, + "loss": 0.162, + "step": 58883 + }, + { + "epoch": 2.26, + "learning_rate": 1.2276224672993076e-05, + "loss": 0.2013, + "step": 58884 + }, + { + "epoch": 2.26, + "learning_rate": 1.2275583482944346e-05, + "loss": 0.2236, + "step": 58885 + }, + { + "epoch": 2.26, + "learning_rate": 1.2274942292895616e-05, + "loss": 0.2966, + "step": 58886 + }, + { + "epoch": 2.26, + "learning_rate": 1.2274301102846884e-05, + "loss": 0.3431, + "step": 58887 + }, + { + "epoch": 2.26, + "learning_rate": 1.2273659912798154e-05, + "loss": 0.2353, + "step": 58888 + }, + { + "epoch": 2.26, + "learning_rate": 1.2273018722749423e-05, + "loss": 0.3012, + "step": 58889 + }, + { + "epoch": 2.26, + "learning_rate": 1.2272377532700694e-05, + "loss": 0.2026, + "step": 58890 + }, + { + "epoch": 2.26, + "learning_rate": 1.2271736342651963e-05, + "loss": 0.1659, + "step": 58891 + }, + { + "epoch": 2.26, + "learning_rate": 1.2271095152603233e-05, + "loss": 0.2542, + "step": 58892 + }, + { + "epoch": 2.26, + "learning_rate": 1.2270453962554501e-05, + "loss": 0.2352, + "step": 58893 + }, + { + "epoch": 2.26, + "learning_rate": 1.2269812772505771e-05, + "loss": 0.2275, + "step": 58894 + }, + { + "epoch": 2.26, + "learning_rate": 1.2269171582457041e-05, + "loss": 0.2391, + "step": 58895 + }, + { + "epoch": 2.26, + "learning_rate": 1.2268530392408311e-05, + "loss": 0.1799, + "step": 58896 + }, + { + "epoch": 2.26, + "learning_rate": 1.226788920235958e-05, + "loss": 0.1738, + "step": 58897 + }, + { + "epoch": 2.26, + "learning_rate": 1.226724801231085e-05, + "loss": 0.1112, + "step": 58898 + }, + { + "epoch": 2.26, + "learning_rate": 1.226660682226212e-05, + "loss": 0.303, + "step": 58899 + }, + { + "epoch": 2.26, + "learning_rate": 1.226596563221339e-05, + "loss": 0.1755, + "step": 58900 + }, + { + "epoch": 2.26, + "learning_rate": 1.2265324442164658e-05, + "loss": 0.121, + "step": 58901 + }, + { + "epoch": 2.26, + "learning_rate": 1.2264683252115928e-05, + "loss": 0.2047, + "step": 58902 + }, + { + "epoch": 2.26, + "learning_rate": 1.2264042062067198e-05, + "loss": 0.282, + "step": 58903 + }, + { + "epoch": 2.26, + "learning_rate": 1.2263400872018468e-05, + "loss": 0.3289, + "step": 58904 + }, + { + "epoch": 2.26, + "learning_rate": 1.2262759681969736e-05, + "loss": 0.2378, + "step": 58905 + }, + { + "epoch": 2.26, + "learning_rate": 1.2262118491921006e-05, + "loss": 0.3028, + "step": 58906 + }, + { + "epoch": 2.26, + "learning_rate": 1.2261477301872276e-05, + "loss": 0.2326, + "step": 58907 + }, + { + "epoch": 2.26, + "learning_rate": 1.2260836111823546e-05, + "loss": 0.2535, + "step": 58908 + }, + { + "epoch": 2.26, + "learning_rate": 1.2260194921774815e-05, + "loss": 0.2483, + "step": 58909 + }, + { + "epoch": 2.26, + "learning_rate": 1.2259553731726085e-05, + "loss": 0.1466, + "step": 58910 + }, + { + "epoch": 2.26, + "learning_rate": 1.2258912541677353e-05, + "loss": 0.2435, + "step": 58911 + }, + { + "epoch": 2.26, + "learning_rate": 1.2258271351628625e-05, + "loss": 0.2814, + "step": 58912 + }, + { + "epoch": 2.27, + "learning_rate": 1.2257630161579893e-05, + "loss": 0.3172, + "step": 58913 + }, + { + "epoch": 2.27, + "learning_rate": 1.2256988971531163e-05, + "loss": 0.3218, + "step": 58914 + }, + { + "epoch": 2.27, + "learning_rate": 1.2256347781482431e-05, + "loss": 0.2296, + "step": 58915 + }, + { + "epoch": 2.27, + "learning_rate": 1.2255706591433701e-05, + "loss": 0.3448, + "step": 58916 + }, + { + "epoch": 2.27, + "learning_rate": 1.2255065401384971e-05, + "loss": 0.3178, + "step": 58917 + }, + { + "epoch": 2.27, + "learning_rate": 1.2254424211336241e-05, + "loss": 0.1896, + "step": 58918 + }, + { + "epoch": 2.27, + "learning_rate": 1.225378302128751e-05, + "loss": 0.3612, + "step": 58919 + }, + { + "epoch": 2.27, + "learning_rate": 1.225314183123878e-05, + "loss": 0.2187, + "step": 58920 + }, + { + "epoch": 2.27, + "learning_rate": 1.225250064119005e-05, + "loss": 0.2374, + "step": 58921 + }, + { + "epoch": 2.27, + "learning_rate": 1.225185945114132e-05, + "loss": 0.216, + "step": 58922 + }, + { + "epoch": 2.27, + "learning_rate": 1.2251218261092588e-05, + "loss": 0.1597, + "step": 58923 + }, + { + "epoch": 2.27, + "learning_rate": 1.2250577071043858e-05, + "loss": 0.1371, + "step": 58924 + }, + { + "epoch": 2.27, + "learning_rate": 1.2249935880995128e-05, + "loss": 0.145, + "step": 58925 + }, + { + "epoch": 2.27, + "learning_rate": 1.2249294690946398e-05, + "loss": 0.202, + "step": 58926 + }, + { + "epoch": 2.27, + "learning_rate": 1.2248653500897666e-05, + "loss": 0.2643, + "step": 58927 + }, + { + "epoch": 2.27, + "learning_rate": 1.2248012310848937e-05, + "loss": 0.2787, + "step": 58928 + }, + { + "epoch": 2.27, + "learning_rate": 1.2247371120800205e-05, + "loss": 0.2036, + "step": 58929 + }, + { + "epoch": 2.27, + "learning_rate": 1.2246729930751477e-05, + "loss": 0.1678, + "step": 58930 + }, + { + "epoch": 2.27, + "learning_rate": 1.2246088740702745e-05, + "loss": 0.1768, + "step": 58931 + }, + { + "epoch": 2.27, + "learning_rate": 1.2245447550654015e-05, + "loss": 0.3105, + "step": 58932 + }, + { + "epoch": 2.27, + "learning_rate": 1.2244806360605283e-05, + "loss": 0.1971, + "step": 58933 + }, + { + "epoch": 2.27, + "learning_rate": 1.2244165170556553e-05, + "loss": 0.2481, + "step": 58934 + }, + { + "epoch": 2.27, + "learning_rate": 1.2243523980507823e-05, + "loss": 0.2107, + "step": 58935 + }, + { + "epoch": 2.27, + "learning_rate": 1.2242882790459093e-05, + "loss": 0.2871, + "step": 58936 + }, + { + "epoch": 2.27, + "learning_rate": 1.2242241600410362e-05, + "loss": 0.171, + "step": 58937 + }, + { + "epoch": 2.27, + "learning_rate": 1.2241600410361632e-05, + "loss": 0.1561, + "step": 58938 + }, + { + "epoch": 2.27, + "learning_rate": 1.2240959220312902e-05, + "loss": 0.2149, + "step": 58939 + }, + { + "epoch": 2.27, + "learning_rate": 1.2240318030264172e-05, + "loss": 0.1234, + "step": 58940 + }, + { + "epoch": 2.27, + "learning_rate": 1.223967684021544e-05, + "loss": 0.1413, + "step": 58941 + }, + { + "epoch": 2.27, + "learning_rate": 1.223903565016671e-05, + "loss": 0.2038, + "step": 58942 + }, + { + "epoch": 2.27, + "learning_rate": 1.223839446011798e-05, + "loss": 0.2083, + "step": 58943 + }, + { + "epoch": 2.27, + "learning_rate": 1.223775327006925e-05, + "loss": 0.3581, + "step": 58944 + }, + { + "epoch": 2.27, + "learning_rate": 1.2237112080020518e-05, + "loss": 0.2216, + "step": 58945 + }, + { + "epoch": 2.27, + "learning_rate": 1.2236470889971788e-05, + "loss": 0.1752, + "step": 58946 + }, + { + "epoch": 2.27, + "learning_rate": 1.2235829699923058e-05, + "loss": 0.2105, + "step": 58947 + }, + { + "epoch": 2.27, + "learning_rate": 1.2235188509874328e-05, + "loss": 0.1822, + "step": 58948 + }, + { + "epoch": 2.27, + "learning_rate": 1.2234547319825597e-05, + "loss": 0.1748, + "step": 58949 + }, + { + "epoch": 2.27, + "learning_rate": 1.2233906129776867e-05, + "loss": 0.2412, + "step": 58950 + }, + { + "epoch": 2.27, + "learning_rate": 1.2233264939728135e-05, + "loss": 0.0846, + "step": 58951 + }, + { + "epoch": 2.27, + "learning_rate": 1.2232623749679407e-05, + "loss": 0.1697, + "step": 58952 + }, + { + "epoch": 2.27, + "learning_rate": 1.2231982559630675e-05, + "loss": 0.315, + "step": 58953 + }, + { + "epoch": 2.27, + "learning_rate": 1.2231341369581945e-05, + "loss": 0.2681, + "step": 58954 + }, + { + "epoch": 2.27, + "learning_rate": 1.2230700179533214e-05, + "loss": 0.2733, + "step": 58955 + }, + { + "epoch": 2.27, + "learning_rate": 1.2230058989484484e-05, + "loss": 0.2314, + "step": 58956 + }, + { + "epoch": 2.27, + "learning_rate": 1.2229417799435754e-05, + "loss": 0.1502, + "step": 58957 + }, + { + "epoch": 2.27, + "learning_rate": 1.2228776609387024e-05, + "loss": 0.177, + "step": 58958 + }, + { + "epoch": 2.27, + "learning_rate": 1.2228135419338292e-05, + "loss": 0.2609, + "step": 58959 + }, + { + "epoch": 2.27, + "learning_rate": 1.2227494229289562e-05, + "loss": 0.201, + "step": 58960 + }, + { + "epoch": 2.27, + "learning_rate": 1.2226853039240832e-05, + "loss": 0.3573, + "step": 58961 + }, + { + "epoch": 2.27, + "learning_rate": 1.2226211849192102e-05, + "loss": 0.271, + "step": 58962 + }, + { + "epoch": 2.27, + "learning_rate": 1.222557065914337e-05, + "loss": 0.1516, + "step": 58963 + }, + { + "epoch": 2.27, + "learning_rate": 1.222492946909464e-05, + "loss": 0.2554, + "step": 58964 + }, + { + "epoch": 2.27, + "learning_rate": 1.222428827904591e-05, + "loss": 0.125, + "step": 58965 + }, + { + "epoch": 2.27, + "learning_rate": 1.222364708899718e-05, + "loss": 0.2171, + "step": 58966 + }, + { + "epoch": 2.27, + "learning_rate": 1.2223005898948449e-05, + "loss": 0.1824, + "step": 58967 + }, + { + "epoch": 2.27, + "learning_rate": 1.2222364708899719e-05, + "loss": 0.2043, + "step": 58968 + }, + { + "epoch": 2.27, + "learning_rate": 1.2221723518850987e-05, + "loss": 0.264, + "step": 58969 + }, + { + "epoch": 2.27, + "learning_rate": 1.2221082328802259e-05, + "loss": 0.1455, + "step": 58970 + }, + { + "epoch": 2.27, + "learning_rate": 1.2220441138753527e-05, + "loss": 0.1461, + "step": 58971 + }, + { + "epoch": 2.27, + "learning_rate": 1.2219799948704797e-05, + "loss": 0.2427, + "step": 58972 + }, + { + "epoch": 2.27, + "learning_rate": 1.2219158758656065e-05, + "loss": 0.1707, + "step": 58973 + }, + { + "epoch": 2.27, + "learning_rate": 1.2218517568607337e-05, + "loss": 0.2125, + "step": 58974 + }, + { + "epoch": 2.27, + "learning_rate": 1.2217876378558606e-05, + "loss": 0.2035, + "step": 58975 + }, + { + "epoch": 2.27, + "learning_rate": 1.2217235188509876e-05, + "loss": 0.325, + "step": 58976 + }, + { + "epoch": 2.27, + "learning_rate": 1.2216593998461144e-05, + "loss": 0.1825, + "step": 58977 + }, + { + "epoch": 2.27, + "learning_rate": 1.2215952808412414e-05, + "loss": 0.1433, + "step": 58978 + }, + { + "epoch": 2.27, + "learning_rate": 1.2215311618363684e-05, + "loss": 0.3526, + "step": 58979 + }, + { + "epoch": 2.27, + "learning_rate": 1.2214670428314954e-05, + "loss": 0.1886, + "step": 58980 + }, + { + "epoch": 2.27, + "learning_rate": 1.2214029238266222e-05, + "loss": 0.2556, + "step": 58981 + }, + { + "epoch": 2.27, + "learning_rate": 1.2213388048217492e-05, + "loss": 0.1019, + "step": 58982 + }, + { + "epoch": 2.27, + "learning_rate": 1.2212746858168762e-05, + "loss": 0.2524, + "step": 58983 + }, + { + "epoch": 2.27, + "learning_rate": 1.2212105668120032e-05, + "loss": 0.2986, + "step": 58984 + }, + { + "epoch": 2.27, + "learning_rate": 1.22114644780713e-05, + "loss": 0.3076, + "step": 58985 + }, + { + "epoch": 2.27, + "learning_rate": 1.221082328802257e-05, + "loss": 0.3119, + "step": 58986 + }, + { + "epoch": 2.27, + "learning_rate": 1.221018209797384e-05, + "loss": 0.2606, + "step": 58987 + }, + { + "epoch": 2.27, + "learning_rate": 1.220954090792511e-05, + "loss": 0.1228, + "step": 58988 + }, + { + "epoch": 2.27, + "learning_rate": 1.2208899717876379e-05, + "loss": 0.3264, + "step": 58989 + }, + { + "epoch": 2.27, + "learning_rate": 1.2208258527827649e-05, + "loss": 0.2117, + "step": 58990 + }, + { + "epoch": 2.27, + "learning_rate": 1.2207617337778917e-05, + "loss": 0.1374, + "step": 58991 + }, + { + "epoch": 2.27, + "learning_rate": 1.2206976147730189e-05, + "loss": 0.1909, + "step": 58992 + }, + { + "epoch": 2.27, + "learning_rate": 1.2206334957681457e-05, + "loss": 0.1594, + "step": 58993 + }, + { + "epoch": 2.27, + "learning_rate": 1.2205693767632728e-05, + "loss": 0.2143, + "step": 58994 + }, + { + "epoch": 2.27, + "learning_rate": 1.2205052577583996e-05, + "loss": 0.2591, + "step": 58995 + }, + { + "epoch": 2.27, + "learning_rate": 1.2204411387535266e-05, + "loss": 0.2475, + "step": 58996 + }, + { + "epoch": 2.27, + "learning_rate": 1.2203770197486536e-05, + "loss": 0.1993, + "step": 58997 + }, + { + "epoch": 2.27, + "learning_rate": 1.2203129007437806e-05, + "loss": 0.1685, + "step": 58998 + }, + { + "epoch": 2.27, + "learning_rate": 1.2202487817389074e-05, + "loss": 0.2594, + "step": 58999 + }, + { + "epoch": 2.27, + "learning_rate": 1.2201846627340344e-05, + "loss": 0.2341, + "step": 59000 + }, + { + "epoch": 2.27, + "learning_rate": 1.2201205437291614e-05, + "loss": 0.1401, + "step": 59001 + }, + { + "epoch": 2.27, + "learning_rate": 1.2200564247242884e-05, + "loss": 0.2317, + "step": 59002 + }, + { + "epoch": 2.27, + "learning_rate": 1.2199923057194153e-05, + "loss": 0.2614, + "step": 59003 + }, + { + "epoch": 2.27, + "learning_rate": 1.2199281867145423e-05, + "loss": 0.1708, + "step": 59004 + }, + { + "epoch": 2.27, + "learning_rate": 1.2198640677096693e-05, + "loss": 0.2383, + "step": 59005 + }, + { + "epoch": 2.27, + "learning_rate": 1.2197999487047963e-05, + "loss": 0.2166, + "step": 59006 + }, + { + "epoch": 2.27, + "learning_rate": 1.2197358296999231e-05, + "loss": 0.1508, + "step": 59007 + }, + { + "epoch": 2.27, + "learning_rate": 1.2196717106950501e-05, + "loss": 0.198, + "step": 59008 + }, + { + "epoch": 2.27, + "learning_rate": 1.219607591690177e-05, + "loss": 0.2685, + "step": 59009 + }, + { + "epoch": 2.27, + "learning_rate": 1.2195434726853041e-05, + "loss": 0.2001, + "step": 59010 + }, + { + "epoch": 2.27, + "learning_rate": 1.219479353680431e-05, + "loss": 0.1823, + "step": 59011 + }, + { + "epoch": 2.27, + "learning_rate": 1.219415234675558e-05, + "loss": 0.3053, + "step": 59012 + }, + { + "epoch": 2.27, + "learning_rate": 1.2193511156706848e-05, + "loss": 0.3033, + "step": 59013 + }, + { + "epoch": 2.27, + "learning_rate": 1.219286996665812e-05, + "loss": 0.2911, + "step": 59014 + }, + { + "epoch": 2.27, + "learning_rate": 1.2192228776609388e-05, + "loss": 0.3076, + "step": 59015 + }, + { + "epoch": 2.27, + "learning_rate": 1.2191587586560658e-05, + "loss": 0.3691, + "step": 59016 + }, + { + "epoch": 2.27, + "learning_rate": 1.2190946396511926e-05, + "loss": 0.1958, + "step": 59017 + }, + { + "epoch": 2.27, + "learning_rate": 1.2190305206463196e-05, + "loss": 0.3239, + "step": 59018 + }, + { + "epoch": 2.27, + "learning_rate": 1.2189664016414466e-05, + "loss": 0.2183, + "step": 59019 + }, + { + "epoch": 2.27, + "learning_rate": 1.2189022826365736e-05, + "loss": 0.1472, + "step": 59020 + }, + { + "epoch": 2.27, + "learning_rate": 1.2188381636317005e-05, + "loss": 0.1896, + "step": 59021 + }, + { + "epoch": 2.27, + "learning_rate": 1.2187740446268275e-05, + "loss": 0.3128, + "step": 59022 + }, + { + "epoch": 2.27, + "learning_rate": 1.2187099256219545e-05, + "loss": 0.1058, + "step": 59023 + }, + { + "epoch": 2.27, + "learning_rate": 1.2186458066170815e-05, + "loss": 0.1909, + "step": 59024 + }, + { + "epoch": 2.27, + "learning_rate": 1.2185816876122083e-05, + "loss": 0.1532, + "step": 59025 + }, + { + "epoch": 2.27, + "learning_rate": 1.2185175686073353e-05, + "loss": 0.2154, + "step": 59026 + }, + { + "epoch": 2.27, + "learning_rate": 1.2184534496024621e-05, + "loss": 0.222, + "step": 59027 + }, + { + "epoch": 2.27, + "learning_rate": 1.2183893305975893e-05, + "loss": 0.2396, + "step": 59028 + }, + { + "epoch": 2.27, + "learning_rate": 1.2183252115927161e-05, + "loss": 0.1613, + "step": 59029 + }, + { + "epoch": 2.27, + "learning_rate": 1.2182610925878431e-05, + "loss": 0.1756, + "step": 59030 + }, + { + "epoch": 2.27, + "learning_rate": 1.21819697358297e-05, + "loss": 0.2452, + "step": 59031 + }, + { + "epoch": 2.27, + "learning_rate": 1.2181328545780971e-05, + "loss": 0.2994, + "step": 59032 + }, + { + "epoch": 2.27, + "learning_rate": 1.218068735573224e-05, + "loss": 0.3718, + "step": 59033 + }, + { + "epoch": 2.27, + "learning_rate": 1.218004616568351e-05, + "loss": 0.2163, + "step": 59034 + }, + { + "epoch": 2.27, + "learning_rate": 1.2179404975634778e-05, + "loss": 0.1716, + "step": 59035 + }, + { + "epoch": 2.27, + "learning_rate": 1.217876378558605e-05, + "loss": 0.13, + "step": 59036 + }, + { + "epoch": 2.27, + "learning_rate": 1.2178122595537318e-05, + "loss": 0.2584, + "step": 59037 + }, + { + "epoch": 2.27, + "learning_rate": 1.2177481405488588e-05, + "loss": 0.2364, + "step": 59038 + }, + { + "epoch": 2.27, + "learning_rate": 1.2176840215439856e-05, + "loss": 0.2269, + "step": 59039 + }, + { + "epoch": 2.27, + "learning_rate": 1.2176199025391127e-05, + "loss": 0.2661, + "step": 59040 + }, + { + "epoch": 2.27, + "learning_rate": 1.2175557835342397e-05, + "loss": 0.3106, + "step": 59041 + }, + { + "epoch": 2.27, + "learning_rate": 1.2174916645293667e-05, + "loss": 0.3213, + "step": 59042 + }, + { + "epoch": 2.27, + "learning_rate": 1.2174275455244935e-05, + "loss": 0.3009, + "step": 59043 + }, + { + "epoch": 2.27, + "learning_rate": 1.2173634265196205e-05, + "loss": 0.2149, + "step": 59044 + }, + { + "epoch": 2.27, + "learning_rate": 1.2172993075147475e-05, + "loss": 0.3293, + "step": 59045 + }, + { + "epoch": 2.27, + "learning_rate": 1.2172351885098745e-05, + "loss": 0.1568, + "step": 59046 + }, + { + "epoch": 2.27, + "learning_rate": 1.2171710695050013e-05, + "loss": 0.1631, + "step": 59047 + }, + { + "epoch": 2.27, + "learning_rate": 1.2171069505001283e-05, + "loss": 0.3242, + "step": 59048 + }, + { + "epoch": 2.27, + "learning_rate": 1.2170428314952552e-05, + "loss": 0.3469, + "step": 59049 + }, + { + "epoch": 2.27, + "learning_rate": 1.2169787124903823e-05, + "loss": 0.1647, + "step": 59050 + }, + { + "epoch": 2.27, + "learning_rate": 1.2169145934855092e-05, + "loss": 0.1236, + "step": 59051 + }, + { + "epoch": 2.27, + "learning_rate": 1.2168504744806362e-05, + "loss": 0.2719, + "step": 59052 + }, + { + "epoch": 2.27, + "learning_rate": 1.216786355475763e-05, + "loss": 0.1262, + "step": 59053 + }, + { + "epoch": 2.27, + "learning_rate": 1.2167222364708902e-05, + "loss": 0.3213, + "step": 59054 + }, + { + "epoch": 2.27, + "learning_rate": 1.216658117466017e-05, + "loss": 0.1506, + "step": 59055 + }, + { + "epoch": 2.27, + "learning_rate": 1.216593998461144e-05, + "loss": 0.3069, + "step": 59056 + }, + { + "epoch": 2.27, + "learning_rate": 1.2165298794562708e-05, + "loss": 0.2587, + "step": 59057 + }, + { + "epoch": 2.27, + "learning_rate": 1.2164657604513978e-05, + "loss": 0.2543, + "step": 59058 + }, + { + "epoch": 2.27, + "learning_rate": 1.2164016414465248e-05, + "loss": 0.1809, + "step": 59059 + }, + { + "epoch": 2.27, + "learning_rate": 1.2163375224416518e-05, + "loss": 0.2538, + "step": 59060 + }, + { + "epoch": 2.27, + "learning_rate": 1.2162734034367787e-05, + "loss": 0.2563, + "step": 59061 + }, + { + "epoch": 2.27, + "learning_rate": 1.2162092844319057e-05, + "loss": 0.1531, + "step": 59062 + }, + { + "epoch": 2.27, + "learning_rate": 1.2161451654270327e-05, + "loss": 0.1878, + "step": 59063 + }, + { + "epoch": 2.27, + "learning_rate": 1.2160810464221597e-05, + "loss": 0.282, + "step": 59064 + }, + { + "epoch": 2.27, + "learning_rate": 1.2160169274172865e-05, + "loss": 0.1081, + "step": 59065 + }, + { + "epoch": 2.27, + "learning_rate": 1.2159528084124135e-05, + "loss": 0.1844, + "step": 59066 + }, + { + "epoch": 2.27, + "learning_rate": 1.2158886894075404e-05, + "loss": 0.1304, + "step": 59067 + }, + { + "epoch": 2.27, + "learning_rate": 1.2158245704026675e-05, + "loss": 0.1377, + "step": 59068 + }, + { + "epoch": 2.27, + "learning_rate": 1.2157604513977944e-05, + "loss": 0.1911, + "step": 59069 + }, + { + "epoch": 2.27, + "learning_rate": 1.2156963323929214e-05, + "loss": 0.1852, + "step": 59070 + }, + { + "epoch": 2.27, + "learning_rate": 1.2156322133880482e-05, + "loss": 0.2353, + "step": 59071 + }, + { + "epoch": 2.27, + "learning_rate": 1.2155680943831754e-05, + "loss": 0.2355, + "step": 59072 + }, + { + "epoch": 2.27, + "learning_rate": 1.2155039753783022e-05, + "loss": 0.0954, + "step": 59073 + }, + { + "epoch": 2.27, + "learning_rate": 1.2154398563734292e-05, + "loss": 0.2601, + "step": 59074 + }, + { + "epoch": 2.27, + "learning_rate": 1.215375737368556e-05, + "loss": 0.2112, + "step": 59075 + }, + { + "epoch": 2.27, + "learning_rate": 1.215311618363683e-05, + "loss": 0.1623, + "step": 59076 + }, + { + "epoch": 2.27, + "learning_rate": 1.21524749935881e-05, + "loss": 0.2077, + "step": 59077 + }, + { + "epoch": 2.27, + "learning_rate": 1.215183380353937e-05, + "loss": 0.287, + "step": 59078 + }, + { + "epoch": 2.27, + "learning_rate": 1.2151192613490639e-05, + "loss": 0.2803, + "step": 59079 + }, + { + "epoch": 2.27, + "learning_rate": 1.2150551423441909e-05, + "loss": 0.2818, + "step": 59080 + }, + { + "epoch": 2.27, + "learning_rate": 1.2149910233393179e-05, + "loss": 0.2597, + "step": 59081 + }, + { + "epoch": 2.27, + "learning_rate": 1.2149269043344449e-05, + "loss": 0.2858, + "step": 59082 + }, + { + "epoch": 2.27, + "learning_rate": 1.2148627853295717e-05, + "loss": 0.2235, + "step": 59083 + }, + { + "epoch": 2.27, + "learning_rate": 1.2147986663246987e-05, + "loss": 0.2614, + "step": 59084 + }, + { + "epoch": 2.27, + "learning_rate": 1.2147345473198257e-05, + "loss": 0.211, + "step": 59085 + }, + { + "epoch": 2.27, + "learning_rate": 1.2146704283149527e-05, + "loss": 0.2816, + "step": 59086 + }, + { + "epoch": 2.27, + "learning_rate": 1.2146063093100796e-05, + "loss": 0.2292, + "step": 59087 + }, + { + "epoch": 2.27, + "learning_rate": 1.2145421903052066e-05, + "loss": 0.1464, + "step": 59088 + }, + { + "epoch": 2.27, + "learning_rate": 1.2144780713003334e-05, + "loss": 0.2687, + "step": 59089 + }, + { + "epoch": 2.27, + "learning_rate": 1.2144139522954606e-05, + "loss": 0.258, + "step": 59090 + }, + { + "epoch": 2.27, + "learning_rate": 1.2143498332905874e-05, + "loss": 0.1798, + "step": 59091 + }, + { + "epoch": 2.27, + "learning_rate": 1.2142857142857144e-05, + "loss": 0.1673, + "step": 59092 + }, + { + "epoch": 2.27, + "learning_rate": 1.2142215952808412e-05, + "loss": 0.1798, + "step": 59093 + }, + { + "epoch": 2.27, + "learning_rate": 1.2141574762759684e-05, + "loss": 0.1246, + "step": 59094 + }, + { + "epoch": 2.27, + "learning_rate": 1.2140933572710952e-05, + "loss": 0.1508, + "step": 59095 + }, + { + "epoch": 2.27, + "learning_rate": 1.2140292382662222e-05, + "loss": 0.219, + "step": 59096 + }, + { + "epoch": 2.27, + "learning_rate": 1.213965119261349e-05, + "loss": 0.2557, + "step": 59097 + }, + { + "epoch": 2.27, + "learning_rate": 1.213901000256476e-05, + "loss": 0.3228, + "step": 59098 + }, + { + "epoch": 2.27, + "learning_rate": 1.213836881251603e-05, + "loss": 0.1481, + "step": 59099 + }, + { + "epoch": 2.27, + "learning_rate": 1.21377276224673e-05, + "loss": 0.3808, + "step": 59100 + }, + { + "epoch": 2.27, + "learning_rate": 1.2137086432418569e-05, + "loss": 0.208, + "step": 59101 + }, + { + "epoch": 2.27, + "learning_rate": 1.2136445242369839e-05, + "loss": 0.203, + "step": 59102 + }, + { + "epoch": 2.27, + "learning_rate": 1.2135804052321109e-05, + "loss": 0.2525, + "step": 59103 + }, + { + "epoch": 2.27, + "learning_rate": 1.213516286227238e-05, + "loss": 0.122, + "step": 59104 + }, + { + "epoch": 2.27, + "learning_rate": 1.2134521672223647e-05, + "loss": 0.1916, + "step": 59105 + }, + { + "epoch": 2.27, + "learning_rate": 1.2133880482174918e-05, + "loss": 0.1719, + "step": 59106 + }, + { + "epoch": 2.27, + "learning_rate": 1.2133239292126186e-05, + "loss": 0.1194, + "step": 59107 + }, + { + "epoch": 2.27, + "learning_rate": 1.2132598102077458e-05, + "loss": 0.1593, + "step": 59108 + }, + { + "epoch": 2.27, + "learning_rate": 1.2131956912028726e-05, + "loss": 0.2971, + "step": 59109 + }, + { + "epoch": 2.27, + "learning_rate": 1.2131315721979996e-05, + "loss": 0.205, + "step": 59110 + }, + { + "epoch": 2.27, + "learning_rate": 1.2130674531931264e-05, + "loss": 0.1397, + "step": 59111 + }, + { + "epoch": 2.27, + "learning_rate": 1.2130033341882536e-05, + "loss": 0.2424, + "step": 59112 + }, + { + "epoch": 2.27, + "learning_rate": 1.2129392151833804e-05, + "loss": 0.239, + "step": 59113 + }, + { + "epoch": 2.27, + "learning_rate": 1.2128750961785074e-05, + "loss": 0.3373, + "step": 59114 + }, + { + "epoch": 2.27, + "learning_rate": 1.2128109771736343e-05, + "loss": 0.2046, + "step": 59115 + }, + { + "epoch": 2.27, + "learning_rate": 1.2127468581687613e-05, + "loss": 0.1665, + "step": 59116 + }, + { + "epoch": 2.27, + "learning_rate": 1.2126827391638883e-05, + "loss": 0.1786, + "step": 59117 + }, + { + "epoch": 2.27, + "learning_rate": 1.2126186201590153e-05, + "loss": 0.3042, + "step": 59118 + }, + { + "epoch": 2.27, + "learning_rate": 1.2125545011541421e-05, + "loss": 0.2678, + "step": 59119 + }, + { + "epoch": 2.27, + "learning_rate": 1.2124903821492691e-05, + "loss": 0.1104, + "step": 59120 + }, + { + "epoch": 2.27, + "learning_rate": 1.2124262631443961e-05, + "loss": 0.1504, + "step": 59121 + }, + { + "epoch": 2.27, + "learning_rate": 1.2123621441395231e-05, + "loss": 0.3783, + "step": 59122 + }, + { + "epoch": 2.27, + "learning_rate": 1.21229802513465e-05, + "loss": 0.1659, + "step": 59123 + }, + { + "epoch": 2.27, + "learning_rate": 1.212233906129777e-05, + "loss": 0.2119, + "step": 59124 + }, + { + "epoch": 2.27, + "learning_rate": 1.2121697871249038e-05, + "loss": 0.1817, + "step": 59125 + }, + { + "epoch": 2.27, + "learning_rate": 1.212105668120031e-05, + "loss": 0.1611, + "step": 59126 + }, + { + "epoch": 2.27, + "learning_rate": 1.2120415491151578e-05, + "loss": 0.2185, + "step": 59127 + }, + { + "epoch": 2.27, + "learning_rate": 1.2119774301102848e-05, + "loss": 0.2079, + "step": 59128 + }, + { + "epoch": 2.27, + "learning_rate": 1.2119133111054116e-05, + "loss": 0.4028, + "step": 59129 + }, + { + "epoch": 2.27, + "learning_rate": 1.2118491921005388e-05, + "loss": 0.4141, + "step": 59130 + }, + { + "epoch": 2.27, + "learning_rate": 1.2117850730956656e-05, + "loss": 0.2461, + "step": 59131 + }, + { + "epoch": 2.27, + "learning_rate": 1.2117209540907926e-05, + "loss": 0.2164, + "step": 59132 + }, + { + "epoch": 2.27, + "learning_rate": 1.2116568350859195e-05, + "loss": 0.232, + "step": 59133 + }, + { + "epoch": 2.27, + "learning_rate": 1.2115927160810466e-05, + "loss": 0.2526, + "step": 59134 + }, + { + "epoch": 2.27, + "learning_rate": 1.2115285970761735e-05, + "loss": 0.3687, + "step": 59135 + }, + { + "epoch": 2.27, + "learning_rate": 1.2114644780713005e-05, + "loss": 0.2936, + "step": 59136 + }, + { + "epoch": 2.27, + "learning_rate": 1.2114003590664273e-05, + "loss": 0.2611, + "step": 59137 + }, + { + "epoch": 2.27, + "learning_rate": 1.2113362400615543e-05, + "loss": 0.3049, + "step": 59138 + }, + { + "epoch": 2.27, + "learning_rate": 1.2112721210566813e-05, + "loss": 0.1759, + "step": 59139 + }, + { + "epoch": 2.27, + "learning_rate": 1.2112080020518083e-05, + "loss": 0.2662, + "step": 59140 + }, + { + "epoch": 2.27, + "learning_rate": 1.2111438830469351e-05, + "loss": 0.2128, + "step": 59141 + }, + { + "epoch": 2.27, + "learning_rate": 1.2110797640420621e-05, + "loss": 0.0978, + "step": 59142 + }, + { + "epoch": 2.27, + "learning_rate": 1.2110156450371891e-05, + "loss": 0.1995, + "step": 59143 + }, + { + "epoch": 2.27, + "learning_rate": 1.2109515260323161e-05, + "loss": 0.261, + "step": 59144 + }, + { + "epoch": 2.27, + "learning_rate": 1.210887407027443e-05, + "loss": 0.1995, + "step": 59145 + }, + { + "epoch": 2.27, + "learning_rate": 1.21082328802257e-05, + "loss": 0.3139, + "step": 59146 + }, + { + "epoch": 2.27, + "learning_rate": 1.2107591690176968e-05, + "loss": 0.2555, + "step": 59147 + }, + { + "epoch": 2.27, + "learning_rate": 1.210695050012824e-05, + "loss": 0.1708, + "step": 59148 + }, + { + "epoch": 2.27, + "learning_rate": 1.2106309310079508e-05, + "loss": 0.2883, + "step": 59149 + }, + { + "epoch": 2.27, + "learning_rate": 1.2105668120030778e-05, + "loss": 0.204, + "step": 59150 + }, + { + "epoch": 2.27, + "learning_rate": 1.2105026929982047e-05, + "loss": 0.378, + "step": 59151 + }, + { + "epoch": 2.27, + "learning_rate": 1.2104385739933318e-05, + "loss": 0.1669, + "step": 59152 + }, + { + "epoch": 2.27, + "learning_rate": 1.2103744549884587e-05, + "loss": 0.1976, + "step": 59153 + }, + { + "epoch": 2.27, + "learning_rate": 1.2103103359835857e-05, + "loss": 0.225, + "step": 59154 + }, + { + "epoch": 2.27, + "learning_rate": 1.2102462169787125e-05, + "loss": 0.1992, + "step": 59155 + }, + { + "epoch": 2.27, + "learning_rate": 1.2101820979738395e-05, + "loss": 0.2316, + "step": 59156 + }, + { + "epoch": 2.27, + "learning_rate": 1.2101179789689665e-05, + "loss": 0.285, + "step": 59157 + }, + { + "epoch": 2.27, + "learning_rate": 1.2100538599640935e-05, + "loss": 0.2227, + "step": 59158 + }, + { + "epoch": 2.27, + "learning_rate": 1.2099897409592203e-05, + "loss": 0.3115, + "step": 59159 + }, + { + "epoch": 2.27, + "learning_rate": 1.2099256219543473e-05, + "loss": 0.2767, + "step": 59160 + }, + { + "epoch": 2.27, + "learning_rate": 1.2098615029494743e-05, + "loss": 0.2053, + "step": 59161 + }, + { + "epoch": 2.27, + "learning_rate": 1.2097973839446013e-05, + "loss": 0.1206, + "step": 59162 + }, + { + "epoch": 2.27, + "learning_rate": 1.2097332649397282e-05, + "loss": 0.2442, + "step": 59163 + }, + { + "epoch": 2.27, + "learning_rate": 1.2096691459348552e-05, + "loss": 0.2233, + "step": 59164 + }, + { + "epoch": 2.27, + "learning_rate": 1.209605026929982e-05, + "loss": 0.3206, + "step": 59165 + }, + { + "epoch": 2.27, + "learning_rate": 1.2095409079251092e-05, + "loss": 0.1808, + "step": 59166 + }, + { + "epoch": 2.27, + "learning_rate": 1.209476788920236e-05, + "loss": 0.1539, + "step": 59167 + }, + { + "epoch": 2.27, + "learning_rate": 1.209412669915363e-05, + "loss": 0.2197, + "step": 59168 + }, + { + "epoch": 2.27, + "learning_rate": 1.2093485509104898e-05, + "loss": 0.3112, + "step": 59169 + }, + { + "epoch": 2.27, + "learning_rate": 1.209284431905617e-05, + "loss": 0.2191, + "step": 59170 + }, + { + "epoch": 2.27, + "learning_rate": 1.2092203129007438e-05, + "loss": 0.1795, + "step": 59171 + }, + { + "epoch": 2.27, + "learning_rate": 1.2091561938958709e-05, + "loss": 0.1148, + "step": 59172 + }, + { + "epoch": 2.28, + "learning_rate": 1.2090920748909977e-05, + "loss": 0.1789, + "step": 59173 + }, + { + "epoch": 2.28, + "learning_rate": 1.2090279558861249e-05, + "loss": 0.3292, + "step": 59174 + }, + { + "epoch": 2.28, + "learning_rate": 1.2089638368812517e-05, + "loss": 0.3516, + "step": 59175 + }, + { + "epoch": 2.28, + "learning_rate": 1.2088997178763787e-05, + "loss": 0.2252, + "step": 59176 + }, + { + "epoch": 2.28, + "learning_rate": 1.2088355988715055e-05, + "loss": 0.3131, + "step": 59177 + }, + { + "epoch": 2.28, + "learning_rate": 1.2087714798666325e-05, + "loss": 0.3167, + "step": 59178 + }, + { + "epoch": 2.28, + "learning_rate": 1.2087073608617595e-05, + "loss": 0.2604, + "step": 59179 + }, + { + "epoch": 2.28, + "learning_rate": 1.2086432418568865e-05, + "loss": 0.2084, + "step": 59180 + }, + { + "epoch": 2.28, + "learning_rate": 1.2085791228520134e-05, + "loss": 0.1797, + "step": 59181 + }, + { + "epoch": 2.28, + "learning_rate": 1.2085150038471404e-05, + "loss": 0.3026, + "step": 59182 + }, + { + "epoch": 2.28, + "learning_rate": 1.2084508848422674e-05, + "loss": 0.1788, + "step": 59183 + }, + { + "epoch": 2.28, + "learning_rate": 1.2083867658373944e-05, + "loss": 0.1461, + "step": 59184 + }, + { + "epoch": 2.28, + "learning_rate": 1.2083226468325212e-05, + "loss": 0.1548, + "step": 59185 + }, + { + "epoch": 2.28, + "learning_rate": 1.2082585278276482e-05, + "loss": 0.128, + "step": 59186 + }, + { + "epoch": 2.28, + "learning_rate": 1.208194408822775e-05, + "loss": 0.1525, + "step": 59187 + }, + { + "epoch": 2.28, + "learning_rate": 1.2081302898179022e-05, + "loss": 0.2171, + "step": 59188 + }, + { + "epoch": 2.28, + "learning_rate": 1.208066170813029e-05, + "loss": 0.1182, + "step": 59189 + }, + { + "epoch": 2.28, + "learning_rate": 1.208002051808156e-05, + "loss": 0.2218, + "step": 59190 + }, + { + "epoch": 2.28, + "learning_rate": 1.2079379328032829e-05, + "loss": 0.1713, + "step": 59191 + }, + { + "epoch": 2.28, + "learning_rate": 1.20787381379841e-05, + "loss": 0.2254, + "step": 59192 + }, + { + "epoch": 2.28, + "learning_rate": 1.2078096947935369e-05, + "loss": 0.2272, + "step": 59193 + }, + { + "epoch": 2.28, + "learning_rate": 1.2077455757886639e-05, + "loss": 0.172, + "step": 59194 + }, + { + "epoch": 2.28, + "learning_rate": 1.2076814567837907e-05, + "loss": 0.2247, + "step": 59195 + }, + { + "epoch": 2.28, + "learning_rate": 1.2076173377789177e-05, + "loss": 0.3341, + "step": 59196 + }, + { + "epoch": 2.28, + "learning_rate": 1.2075532187740447e-05, + "loss": 0.1481, + "step": 59197 + }, + { + "epoch": 2.28, + "learning_rate": 1.2074890997691717e-05, + "loss": 0.1317, + "step": 59198 + }, + { + "epoch": 2.28, + "learning_rate": 1.2074249807642986e-05, + "loss": 0.2703, + "step": 59199 + }, + { + "epoch": 2.28, + "learning_rate": 1.2073608617594256e-05, + "loss": 0.2824, + "step": 59200 + }, + { + "epoch": 2.28, + "learning_rate": 1.2072967427545526e-05, + "loss": 0.1736, + "step": 59201 + }, + { + "epoch": 2.28, + "learning_rate": 1.2072326237496796e-05, + "loss": 0.1629, + "step": 59202 + }, + { + "epoch": 2.28, + "learning_rate": 1.2071685047448064e-05, + "loss": 0.2375, + "step": 59203 + }, + { + "epoch": 2.28, + "learning_rate": 1.2071043857399334e-05, + "loss": 0.2574, + "step": 59204 + }, + { + "epoch": 2.28, + "learning_rate": 1.2070402667350602e-05, + "loss": 0.2252, + "step": 59205 + }, + { + "epoch": 2.28, + "learning_rate": 1.2069761477301874e-05, + "loss": 0.2448, + "step": 59206 + }, + { + "epoch": 2.28, + "learning_rate": 1.2069120287253142e-05, + "loss": 0.3295, + "step": 59207 + }, + { + "epoch": 2.28, + "learning_rate": 1.2068479097204412e-05, + "loss": 0.2095, + "step": 59208 + }, + { + "epoch": 2.28, + "learning_rate": 1.206783790715568e-05, + "loss": 0.2351, + "step": 59209 + }, + { + "epoch": 2.28, + "learning_rate": 1.2067196717106952e-05, + "loss": 0.2146, + "step": 59210 + }, + { + "epoch": 2.28, + "learning_rate": 1.206655552705822e-05, + "loss": 0.1627, + "step": 59211 + }, + { + "epoch": 2.28, + "learning_rate": 1.206591433700949e-05, + "loss": 0.2877, + "step": 59212 + }, + { + "epoch": 2.28, + "learning_rate": 1.2065273146960759e-05, + "loss": 0.2082, + "step": 59213 + }, + { + "epoch": 2.28, + "learning_rate": 1.2064631956912029e-05, + "loss": 0.2506, + "step": 59214 + }, + { + "epoch": 2.28, + "learning_rate": 1.2063990766863299e-05, + "loss": 0.1244, + "step": 59215 + }, + { + "epoch": 2.28, + "learning_rate": 1.206334957681457e-05, + "loss": 0.1803, + "step": 59216 + }, + { + "epoch": 2.28, + "learning_rate": 1.2062708386765837e-05, + "loss": 0.1618, + "step": 59217 + }, + { + "epoch": 2.28, + "learning_rate": 1.2062067196717108e-05, + "loss": 0.2985, + "step": 59218 + }, + { + "epoch": 2.28, + "learning_rate": 1.2061426006668378e-05, + "loss": 0.1244, + "step": 59219 + }, + { + "epoch": 2.28, + "learning_rate": 1.2060784816619648e-05, + "loss": 0.1221, + "step": 59220 + }, + { + "epoch": 2.28, + "learning_rate": 1.2060143626570916e-05, + "loss": 0.3328, + "step": 59221 + }, + { + "epoch": 2.28, + "learning_rate": 1.2059502436522186e-05, + "loss": 0.2835, + "step": 59222 + }, + { + "epoch": 2.28, + "learning_rate": 1.2058861246473456e-05, + "loss": 0.2558, + "step": 59223 + }, + { + "epoch": 2.28, + "learning_rate": 1.2058220056424726e-05, + "loss": 0.2725, + "step": 59224 + }, + { + "epoch": 2.28, + "learning_rate": 1.2057578866375994e-05, + "loss": 0.3324, + "step": 59225 + }, + { + "epoch": 2.28, + "learning_rate": 1.2056937676327264e-05, + "loss": 0.2269, + "step": 59226 + }, + { + "epoch": 2.28, + "learning_rate": 1.2056296486278533e-05, + "loss": 0.3297, + "step": 59227 + }, + { + "epoch": 2.28, + "learning_rate": 1.2055655296229804e-05, + "loss": 0.1704, + "step": 59228 + }, + { + "epoch": 2.28, + "learning_rate": 1.2055014106181073e-05, + "loss": 0.1254, + "step": 59229 + }, + { + "epoch": 2.28, + "learning_rate": 1.2054372916132343e-05, + "loss": 0.2134, + "step": 59230 + }, + { + "epoch": 2.28, + "learning_rate": 1.2053731726083611e-05, + "loss": 0.3509, + "step": 59231 + }, + { + "epoch": 2.28, + "learning_rate": 1.2053090536034883e-05, + "loss": 0.1491, + "step": 59232 + }, + { + "epoch": 2.28, + "learning_rate": 1.2052449345986151e-05, + "loss": 0.2046, + "step": 59233 + }, + { + "epoch": 2.28, + "learning_rate": 1.2051808155937421e-05, + "loss": 0.2579, + "step": 59234 + }, + { + "epoch": 2.28, + "learning_rate": 1.205116696588869e-05, + "loss": 0.2493, + "step": 59235 + }, + { + "epoch": 2.28, + "learning_rate": 1.205052577583996e-05, + "loss": 0.2816, + "step": 59236 + }, + { + "epoch": 2.28, + "learning_rate": 1.204988458579123e-05, + "loss": 0.4544, + "step": 59237 + }, + { + "epoch": 2.28, + "learning_rate": 1.20492433957425e-05, + "loss": 0.2717, + "step": 59238 + }, + { + "epoch": 2.28, + "learning_rate": 1.2048602205693768e-05, + "loss": 0.2557, + "step": 59239 + }, + { + "epoch": 2.28, + "learning_rate": 1.2047961015645038e-05, + "loss": 0.2041, + "step": 59240 + }, + { + "epoch": 2.28, + "learning_rate": 1.2047319825596308e-05, + "loss": 0.2212, + "step": 59241 + }, + { + "epoch": 2.28, + "learning_rate": 1.2046678635547578e-05, + "loss": 0.1832, + "step": 59242 + }, + { + "epoch": 2.28, + "learning_rate": 1.2046037445498846e-05, + "loss": 0.207, + "step": 59243 + }, + { + "epoch": 2.28, + "learning_rate": 1.2045396255450116e-05, + "loss": 0.1771, + "step": 59244 + }, + { + "epoch": 2.28, + "learning_rate": 1.2044755065401385e-05, + "loss": 0.1748, + "step": 59245 + }, + { + "epoch": 2.28, + "learning_rate": 1.2044113875352656e-05, + "loss": 0.2607, + "step": 59246 + }, + { + "epoch": 2.28, + "learning_rate": 1.2043472685303925e-05, + "loss": 0.3283, + "step": 59247 + }, + { + "epoch": 2.28, + "learning_rate": 1.2042831495255195e-05, + "loss": 0.2567, + "step": 59248 + }, + { + "epoch": 2.28, + "learning_rate": 1.2042190305206463e-05, + "loss": 0.2621, + "step": 59249 + }, + { + "epoch": 2.28, + "learning_rate": 1.2041549115157735e-05, + "loss": 0.146, + "step": 59250 + }, + { + "epoch": 2.28, + "learning_rate": 1.2040907925109003e-05, + "loss": 0.139, + "step": 59251 + }, + { + "epoch": 2.28, + "learning_rate": 1.2040266735060273e-05, + "loss": 0.1463, + "step": 59252 + }, + { + "epoch": 2.28, + "learning_rate": 1.2039625545011541e-05, + "loss": 0.1888, + "step": 59253 + }, + { + "epoch": 2.28, + "learning_rate": 1.2038984354962811e-05, + "loss": 0.2135, + "step": 59254 + }, + { + "epoch": 2.28, + "learning_rate": 1.2038343164914081e-05, + "loss": 0.206, + "step": 59255 + }, + { + "epoch": 2.28, + "learning_rate": 1.2037701974865351e-05, + "loss": 0.239, + "step": 59256 + }, + { + "epoch": 2.28, + "learning_rate": 1.203706078481662e-05, + "loss": 0.2027, + "step": 59257 + }, + { + "epoch": 2.28, + "learning_rate": 1.203641959476789e-05, + "loss": 0.206, + "step": 59258 + }, + { + "epoch": 2.28, + "learning_rate": 1.203577840471916e-05, + "loss": 0.1784, + "step": 59259 + }, + { + "epoch": 2.28, + "learning_rate": 1.203513721467043e-05, + "loss": 0.1297, + "step": 59260 + }, + { + "epoch": 2.28, + "learning_rate": 1.2034496024621698e-05, + "loss": 0.2544, + "step": 59261 + }, + { + "epoch": 2.28, + "learning_rate": 1.2033854834572968e-05, + "loss": 0.1839, + "step": 59262 + }, + { + "epoch": 2.28, + "learning_rate": 1.2033213644524237e-05, + "loss": 0.2042, + "step": 59263 + }, + { + "epoch": 2.28, + "learning_rate": 1.2032572454475508e-05, + "loss": 0.2226, + "step": 59264 + }, + { + "epoch": 2.28, + "learning_rate": 1.2031931264426777e-05, + "loss": 0.3109, + "step": 59265 + }, + { + "epoch": 2.28, + "learning_rate": 1.2031290074378047e-05, + "loss": 0.3461, + "step": 59266 + }, + { + "epoch": 2.28, + "learning_rate": 1.2030648884329315e-05, + "loss": 0.1413, + "step": 59267 + }, + { + "epoch": 2.28, + "learning_rate": 1.2030007694280587e-05, + "loss": 0.2261, + "step": 59268 + }, + { + "epoch": 2.28, + "learning_rate": 1.2029366504231855e-05, + "loss": 0.2537, + "step": 59269 + }, + { + "epoch": 2.28, + "learning_rate": 1.2028725314183125e-05, + "loss": 0.3926, + "step": 59270 + }, + { + "epoch": 2.28, + "learning_rate": 1.2028084124134393e-05, + "loss": 0.2624, + "step": 59271 + }, + { + "epoch": 2.28, + "learning_rate": 1.2027442934085665e-05, + "loss": 0.1146, + "step": 59272 + }, + { + "epoch": 2.28, + "learning_rate": 1.2026801744036933e-05, + "loss": 0.316, + "step": 59273 + }, + { + "epoch": 2.28, + "learning_rate": 1.2026160553988203e-05, + "loss": 0.2384, + "step": 59274 + }, + { + "epoch": 2.28, + "learning_rate": 1.2025519363939472e-05, + "loss": 0.2836, + "step": 59275 + }, + { + "epoch": 2.28, + "learning_rate": 1.2024878173890742e-05, + "loss": 0.2671, + "step": 59276 + }, + { + "epoch": 2.28, + "learning_rate": 1.2024236983842012e-05, + "loss": 0.1622, + "step": 59277 + }, + { + "epoch": 2.28, + "learning_rate": 1.2023595793793282e-05, + "loss": 0.232, + "step": 59278 + }, + { + "epoch": 2.28, + "learning_rate": 1.202295460374455e-05, + "loss": 0.1301, + "step": 59279 + }, + { + "epoch": 2.28, + "learning_rate": 1.202231341369582e-05, + "loss": 0.2372, + "step": 59280 + }, + { + "epoch": 2.28, + "learning_rate": 1.202167222364709e-05, + "loss": 0.2238, + "step": 59281 + }, + { + "epoch": 2.28, + "learning_rate": 1.202103103359836e-05, + "loss": 0.2476, + "step": 59282 + }, + { + "epoch": 2.28, + "learning_rate": 1.2020389843549628e-05, + "loss": 0.2085, + "step": 59283 + }, + { + "epoch": 2.28, + "learning_rate": 1.2019748653500899e-05, + "loss": 0.224, + "step": 59284 + }, + { + "epoch": 2.28, + "learning_rate": 1.2019107463452167e-05, + "loss": 0.1456, + "step": 59285 + }, + { + "epoch": 2.28, + "learning_rate": 1.2018466273403439e-05, + "loss": 0.2443, + "step": 59286 + }, + { + "epoch": 2.28, + "learning_rate": 1.2017825083354707e-05, + "loss": 0.2479, + "step": 59287 + }, + { + "epoch": 2.28, + "learning_rate": 1.2017183893305977e-05, + "loss": 0.3577, + "step": 59288 + }, + { + "epoch": 2.28, + "learning_rate": 1.2016542703257245e-05, + "loss": 0.2505, + "step": 59289 + }, + { + "epoch": 2.28, + "learning_rate": 1.2015901513208517e-05, + "loss": 0.226, + "step": 59290 + }, + { + "epoch": 2.28, + "learning_rate": 1.2015260323159785e-05, + "loss": 0.314, + "step": 59291 + }, + { + "epoch": 2.28, + "learning_rate": 1.2014619133111055e-05, + "loss": 0.2199, + "step": 59292 + }, + { + "epoch": 2.28, + "learning_rate": 1.2013977943062324e-05, + "loss": 0.2528, + "step": 59293 + }, + { + "epoch": 2.28, + "learning_rate": 1.2013336753013594e-05, + "loss": 0.1907, + "step": 59294 + }, + { + "epoch": 2.28, + "learning_rate": 1.2012695562964864e-05, + "loss": 0.2643, + "step": 59295 + }, + { + "epoch": 2.28, + "learning_rate": 1.2012054372916134e-05, + "loss": 0.1784, + "step": 59296 + }, + { + "epoch": 2.28, + "learning_rate": 1.2011413182867402e-05, + "loss": 0.2617, + "step": 59297 + }, + { + "epoch": 2.28, + "learning_rate": 1.2010771992818672e-05, + "loss": 0.1919, + "step": 59298 + }, + { + "epoch": 2.28, + "learning_rate": 1.2010130802769942e-05, + "loss": 0.2452, + "step": 59299 + }, + { + "epoch": 2.28, + "learning_rate": 1.2009489612721212e-05, + "loss": 0.1858, + "step": 59300 + }, + { + "epoch": 2.28, + "learning_rate": 1.200884842267248e-05, + "loss": 0.091, + "step": 59301 + }, + { + "epoch": 2.28, + "learning_rate": 1.200820723262375e-05, + "loss": 0.1611, + "step": 59302 + }, + { + "epoch": 2.28, + "learning_rate": 1.2007566042575019e-05, + "loss": 0.2481, + "step": 59303 + }, + { + "epoch": 2.28, + "learning_rate": 1.200692485252629e-05, + "loss": 0.2966, + "step": 59304 + }, + { + "epoch": 2.28, + "learning_rate": 1.2006283662477559e-05, + "loss": 0.2409, + "step": 59305 + }, + { + "epoch": 2.28, + "learning_rate": 1.2005642472428829e-05, + "loss": 0.1541, + "step": 59306 + }, + { + "epoch": 2.28, + "learning_rate": 1.2005001282380097e-05, + "loss": 0.2272, + "step": 59307 + }, + { + "epoch": 2.28, + "learning_rate": 1.2004360092331369e-05, + "loss": 0.2511, + "step": 59308 + }, + { + "epoch": 2.28, + "learning_rate": 1.2003718902282637e-05, + "loss": 0.1736, + "step": 59309 + }, + { + "epoch": 2.28, + "learning_rate": 1.2003077712233907e-05, + "loss": 0.1074, + "step": 59310 + }, + { + "epoch": 2.28, + "learning_rate": 1.2002436522185176e-05, + "loss": 0.2131, + "step": 59311 + }, + { + "epoch": 2.28, + "learning_rate": 1.2001795332136446e-05, + "loss": 0.1933, + "step": 59312 + }, + { + "epoch": 2.28, + "learning_rate": 1.2001154142087716e-05, + "loss": 0.1942, + "step": 59313 + }, + { + "epoch": 2.28, + "learning_rate": 1.2000512952038986e-05, + "loss": 0.1834, + "step": 59314 + }, + { + "epoch": 2.28, + "learning_rate": 1.1999871761990254e-05, + "loss": 0.2909, + "step": 59315 + }, + { + "epoch": 2.28, + "learning_rate": 1.1999230571941524e-05, + "loss": 0.4, + "step": 59316 + }, + { + "epoch": 2.28, + "learning_rate": 1.1998589381892794e-05, + "loss": 0.2057, + "step": 59317 + }, + { + "epoch": 2.28, + "learning_rate": 1.1997948191844064e-05, + "loss": 0.1681, + "step": 59318 + }, + { + "epoch": 2.28, + "learning_rate": 1.1997307001795332e-05, + "loss": 0.2972, + "step": 59319 + }, + { + "epoch": 2.28, + "learning_rate": 1.1996665811746602e-05, + "loss": 0.4111, + "step": 59320 + }, + { + "epoch": 2.28, + "learning_rate": 1.1996024621697872e-05, + "loss": 0.2456, + "step": 59321 + }, + { + "epoch": 2.28, + "learning_rate": 1.1995383431649142e-05, + "loss": 0.1703, + "step": 59322 + }, + { + "epoch": 2.28, + "learning_rate": 1.199474224160041e-05, + "loss": 0.3012, + "step": 59323 + }, + { + "epoch": 2.28, + "learning_rate": 1.199410105155168e-05, + "loss": 0.1751, + "step": 59324 + }, + { + "epoch": 2.28, + "learning_rate": 1.1993459861502949e-05, + "loss": 0.2119, + "step": 59325 + }, + { + "epoch": 2.28, + "learning_rate": 1.199281867145422e-05, + "loss": 0.3114, + "step": 59326 + }, + { + "epoch": 2.28, + "learning_rate": 1.199217748140549e-05, + "loss": 0.288, + "step": 59327 + }, + { + "epoch": 2.28, + "learning_rate": 1.199153629135676e-05, + "loss": 0.138, + "step": 59328 + }, + { + "epoch": 2.28, + "learning_rate": 1.1990895101308028e-05, + "loss": 0.2115, + "step": 59329 + }, + { + "epoch": 2.28, + "learning_rate": 1.19902539112593e-05, + "loss": 0.1789, + "step": 59330 + }, + { + "epoch": 2.28, + "learning_rate": 1.1989612721210568e-05, + "loss": 0.3088, + "step": 59331 + }, + { + "epoch": 2.28, + "learning_rate": 1.1988971531161838e-05, + "loss": 0.2485, + "step": 59332 + }, + { + "epoch": 2.28, + "learning_rate": 1.1988330341113106e-05, + "loss": 0.1639, + "step": 59333 + }, + { + "epoch": 2.28, + "learning_rate": 1.1987689151064376e-05, + "loss": 0.123, + "step": 59334 + }, + { + "epoch": 2.28, + "learning_rate": 1.1987047961015646e-05, + "loss": 0.2493, + "step": 59335 + }, + { + "epoch": 2.28, + "learning_rate": 1.1986406770966916e-05, + "loss": 0.2218, + "step": 59336 + }, + { + "epoch": 2.28, + "learning_rate": 1.1985765580918184e-05, + "loss": 0.1747, + "step": 59337 + }, + { + "epoch": 2.28, + "learning_rate": 1.1985124390869454e-05, + "loss": 0.4549, + "step": 59338 + }, + { + "epoch": 2.28, + "learning_rate": 1.1984483200820724e-05, + "loss": 0.23, + "step": 59339 + }, + { + "epoch": 2.28, + "learning_rate": 1.1983842010771994e-05, + "loss": 0.3419, + "step": 59340 + }, + { + "epoch": 2.28, + "learning_rate": 1.1983200820723263e-05, + "loss": 0.1881, + "step": 59341 + }, + { + "epoch": 2.28, + "learning_rate": 1.1982559630674533e-05, + "loss": 0.2346, + "step": 59342 + }, + { + "epoch": 2.28, + "learning_rate": 1.1981918440625801e-05, + "loss": 0.2846, + "step": 59343 + }, + { + "epoch": 2.28, + "learning_rate": 1.1981277250577073e-05, + "loss": 0.2758, + "step": 59344 + }, + { + "epoch": 2.28, + "learning_rate": 1.1980636060528341e-05, + "loss": 0.2061, + "step": 59345 + }, + { + "epoch": 2.28, + "learning_rate": 1.1979994870479611e-05, + "loss": 0.2263, + "step": 59346 + }, + { + "epoch": 2.28, + "learning_rate": 1.197935368043088e-05, + "loss": 0.2086, + "step": 59347 + }, + { + "epoch": 2.28, + "learning_rate": 1.1978712490382151e-05, + "loss": 0.2015, + "step": 59348 + }, + { + "epoch": 2.28, + "learning_rate": 1.197807130033342e-05, + "loss": 0.381, + "step": 59349 + }, + { + "epoch": 2.28, + "learning_rate": 1.197743011028469e-05, + "loss": 0.1638, + "step": 59350 + }, + { + "epoch": 2.28, + "learning_rate": 1.1976788920235958e-05, + "loss": 0.2062, + "step": 59351 + }, + { + "epoch": 2.28, + "learning_rate": 1.1976147730187228e-05, + "loss": 0.1948, + "step": 59352 + }, + { + "epoch": 2.28, + "learning_rate": 1.1975506540138498e-05, + "loss": 0.2185, + "step": 59353 + }, + { + "epoch": 2.28, + "learning_rate": 1.1974865350089768e-05, + "loss": 0.1554, + "step": 59354 + }, + { + "epoch": 2.28, + "learning_rate": 1.1974224160041036e-05, + "loss": 0.3001, + "step": 59355 + }, + { + "epoch": 2.28, + "learning_rate": 1.1973582969992306e-05, + "loss": 0.3748, + "step": 59356 + }, + { + "epoch": 2.28, + "learning_rate": 1.1972941779943576e-05, + "loss": 0.2809, + "step": 59357 + }, + { + "epoch": 2.28, + "learning_rate": 1.1972300589894846e-05, + "loss": 0.1826, + "step": 59358 + }, + { + "epoch": 2.28, + "learning_rate": 1.1971659399846115e-05, + "loss": 0.1234, + "step": 59359 + }, + { + "epoch": 2.28, + "learning_rate": 1.1971018209797385e-05, + "loss": 0.2925, + "step": 59360 + }, + { + "epoch": 2.28, + "learning_rate": 1.1970377019748653e-05, + "loss": 0.2061, + "step": 59361 + }, + { + "epoch": 2.28, + "learning_rate": 1.1969735829699925e-05, + "loss": 0.2469, + "step": 59362 + }, + { + "epoch": 2.28, + "learning_rate": 1.1969094639651193e-05, + "loss": 0.2612, + "step": 59363 + }, + { + "epoch": 2.28, + "learning_rate": 1.1968453449602463e-05, + "loss": 0.2531, + "step": 59364 + }, + { + "epoch": 2.28, + "learning_rate": 1.1967812259553731e-05, + "loss": 0.1047, + "step": 59365 + }, + { + "epoch": 2.28, + "learning_rate": 1.1967171069505003e-05, + "loss": 0.177, + "step": 59366 + }, + { + "epoch": 2.28, + "learning_rate": 1.1966529879456271e-05, + "loss": 0.4558, + "step": 59367 + }, + { + "epoch": 2.28, + "learning_rate": 1.1965888689407541e-05, + "loss": 0.266, + "step": 59368 + }, + { + "epoch": 2.28, + "learning_rate": 1.196524749935881e-05, + "loss": 0.2177, + "step": 59369 + }, + { + "epoch": 2.28, + "learning_rate": 1.1964606309310081e-05, + "loss": 0.1821, + "step": 59370 + }, + { + "epoch": 2.28, + "learning_rate": 1.196396511926135e-05, + "loss": 0.2138, + "step": 59371 + }, + { + "epoch": 2.28, + "learning_rate": 1.196332392921262e-05, + "loss": 0.1846, + "step": 59372 + }, + { + "epoch": 2.28, + "learning_rate": 1.1962682739163888e-05, + "loss": 0.2431, + "step": 59373 + }, + { + "epoch": 2.28, + "learning_rate": 1.1962041549115158e-05, + "loss": 0.3781, + "step": 59374 + }, + { + "epoch": 2.28, + "learning_rate": 1.1961400359066428e-05, + "loss": 0.2123, + "step": 59375 + }, + { + "epoch": 2.28, + "learning_rate": 1.1960759169017698e-05, + "loss": 0.1651, + "step": 59376 + }, + { + "epoch": 2.28, + "learning_rate": 1.1960117978968967e-05, + "loss": 0.0821, + "step": 59377 + }, + { + "epoch": 2.28, + "learning_rate": 1.1959476788920237e-05, + "loss": 0.287, + "step": 59378 + }, + { + "epoch": 2.28, + "learning_rate": 1.1958835598871507e-05, + "loss": 0.2112, + "step": 59379 + }, + { + "epoch": 2.28, + "learning_rate": 1.1958194408822777e-05, + "loss": 0.1902, + "step": 59380 + }, + { + "epoch": 2.28, + "learning_rate": 1.1957553218774045e-05, + "loss": 0.1867, + "step": 59381 + }, + { + "epoch": 2.28, + "learning_rate": 1.1956912028725315e-05, + "loss": 0.2009, + "step": 59382 + }, + { + "epoch": 2.28, + "learning_rate": 1.1956270838676583e-05, + "loss": 0.1547, + "step": 59383 + }, + { + "epoch": 2.28, + "learning_rate": 1.1955629648627855e-05, + "loss": 0.3628, + "step": 59384 + }, + { + "epoch": 2.28, + "learning_rate": 1.1954988458579123e-05, + "loss": 0.2272, + "step": 59385 + }, + { + "epoch": 2.28, + "learning_rate": 1.1954347268530393e-05, + "loss": 0.2423, + "step": 59386 + }, + { + "epoch": 2.28, + "learning_rate": 1.1953706078481662e-05, + "loss": 0.1661, + "step": 59387 + }, + { + "epoch": 2.28, + "learning_rate": 1.1953064888432933e-05, + "loss": 0.196, + "step": 59388 + }, + { + "epoch": 2.28, + "learning_rate": 1.1952423698384202e-05, + "loss": 0.1698, + "step": 59389 + }, + { + "epoch": 2.28, + "learning_rate": 1.1951782508335472e-05, + "loss": 0.2664, + "step": 59390 + }, + { + "epoch": 2.28, + "learning_rate": 1.195114131828674e-05, + "loss": 0.3673, + "step": 59391 + }, + { + "epoch": 2.28, + "learning_rate": 1.195050012823801e-05, + "loss": 0.2404, + "step": 59392 + }, + { + "epoch": 2.28, + "learning_rate": 1.194985893818928e-05, + "loss": 0.3256, + "step": 59393 + }, + { + "epoch": 2.28, + "learning_rate": 1.194921774814055e-05, + "loss": 0.1096, + "step": 59394 + }, + { + "epoch": 2.28, + "learning_rate": 1.1948576558091819e-05, + "loss": 0.3566, + "step": 59395 + }, + { + "epoch": 2.28, + "learning_rate": 1.1947935368043089e-05, + "loss": 0.2758, + "step": 59396 + }, + { + "epoch": 2.28, + "learning_rate": 1.1947294177994359e-05, + "loss": 0.3926, + "step": 59397 + }, + { + "epoch": 2.28, + "learning_rate": 1.1946652987945629e-05, + "loss": 0.3089, + "step": 59398 + }, + { + "epoch": 2.28, + "learning_rate": 1.1946011797896897e-05, + "loss": 0.1426, + "step": 59399 + }, + { + "epoch": 2.28, + "learning_rate": 1.1945370607848167e-05, + "loss": 0.4364, + "step": 59400 + }, + { + "epoch": 2.28, + "learning_rate": 1.1944729417799435e-05, + "loss": 0.2544, + "step": 59401 + }, + { + "epoch": 2.28, + "learning_rate": 1.1944088227750707e-05, + "loss": 0.327, + "step": 59402 + }, + { + "epoch": 2.28, + "learning_rate": 1.1943447037701975e-05, + "loss": 0.1637, + "step": 59403 + }, + { + "epoch": 2.28, + "learning_rate": 1.1942805847653245e-05, + "loss": 0.1715, + "step": 59404 + }, + { + "epoch": 2.28, + "learning_rate": 1.1942164657604514e-05, + "loss": 0.2062, + "step": 59405 + }, + { + "epoch": 2.28, + "learning_rate": 1.1941523467555785e-05, + "loss": 0.193, + "step": 59406 + }, + { + "epoch": 2.28, + "learning_rate": 1.1940882277507054e-05, + "loss": 0.223, + "step": 59407 + }, + { + "epoch": 2.28, + "learning_rate": 1.1940241087458324e-05, + "loss": 0.232, + "step": 59408 + }, + { + "epoch": 2.28, + "learning_rate": 1.1939599897409592e-05, + "loss": 0.1574, + "step": 59409 + }, + { + "epoch": 2.28, + "learning_rate": 1.1938958707360862e-05, + "loss": 0.1854, + "step": 59410 + }, + { + "epoch": 2.28, + "learning_rate": 1.1938317517312132e-05, + "loss": 0.1897, + "step": 59411 + }, + { + "epoch": 2.28, + "learning_rate": 1.1937676327263402e-05, + "loss": 0.1822, + "step": 59412 + }, + { + "epoch": 2.28, + "learning_rate": 1.193703513721467e-05, + "loss": 0.2997, + "step": 59413 + }, + { + "epoch": 2.28, + "learning_rate": 1.193639394716594e-05, + "loss": 0.2025, + "step": 59414 + }, + { + "epoch": 2.28, + "learning_rate": 1.193575275711721e-05, + "loss": 0.1962, + "step": 59415 + }, + { + "epoch": 2.28, + "learning_rate": 1.193511156706848e-05, + "loss": 0.2582, + "step": 59416 + }, + { + "epoch": 2.28, + "learning_rate": 1.1934470377019749e-05, + "loss": 0.2473, + "step": 59417 + }, + { + "epoch": 2.28, + "learning_rate": 1.1933829186971019e-05, + "loss": 0.2837, + "step": 59418 + }, + { + "epoch": 2.28, + "learning_rate": 1.1933187996922289e-05, + "loss": 0.2908, + "step": 59419 + }, + { + "epoch": 2.28, + "learning_rate": 1.1932546806873559e-05, + "loss": 0.221, + "step": 59420 + }, + { + "epoch": 2.28, + "learning_rate": 1.1931905616824827e-05, + "loss": 0.1902, + "step": 59421 + }, + { + "epoch": 2.28, + "learning_rate": 1.1931264426776097e-05, + "loss": 0.1691, + "step": 59422 + }, + { + "epoch": 2.28, + "learning_rate": 1.1930623236727366e-05, + "loss": 0.3389, + "step": 59423 + }, + { + "epoch": 2.28, + "learning_rate": 1.1929982046678637e-05, + "loss": 0.1335, + "step": 59424 + }, + { + "epoch": 2.28, + "learning_rate": 1.1929340856629906e-05, + "loss": 0.1805, + "step": 59425 + }, + { + "epoch": 2.28, + "learning_rate": 1.1928699666581176e-05, + "loss": 0.1722, + "step": 59426 + }, + { + "epoch": 2.28, + "learning_rate": 1.1928058476532444e-05, + "loss": 0.1614, + "step": 59427 + }, + { + "epoch": 2.28, + "learning_rate": 1.1927417286483716e-05, + "loss": 0.1448, + "step": 59428 + }, + { + "epoch": 2.28, + "learning_rate": 1.1926776096434984e-05, + "loss": 0.4112, + "step": 59429 + }, + { + "epoch": 2.28, + "learning_rate": 1.1926134906386254e-05, + "loss": 0.3493, + "step": 59430 + }, + { + "epoch": 2.28, + "learning_rate": 1.1925493716337522e-05, + "loss": 0.3494, + "step": 59431 + }, + { + "epoch": 2.28, + "learning_rate": 1.1924852526288792e-05, + "loss": 0.3354, + "step": 59432 + }, + { + "epoch": 2.29, + "learning_rate": 1.1924211336240062e-05, + "loss": 0.1598, + "step": 59433 + }, + { + "epoch": 2.29, + "learning_rate": 1.1923570146191332e-05, + "loss": 0.2468, + "step": 59434 + }, + { + "epoch": 2.29, + "learning_rate": 1.19229289561426e-05, + "loss": 0.1869, + "step": 59435 + }, + { + "epoch": 2.29, + "learning_rate": 1.192228776609387e-05, + "loss": 0.1908, + "step": 59436 + }, + { + "epoch": 2.29, + "learning_rate": 1.192164657604514e-05, + "loss": 0.2649, + "step": 59437 + }, + { + "epoch": 2.29, + "learning_rate": 1.192100538599641e-05, + "loss": 0.1584, + "step": 59438 + }, + { + "epoch": 2.29, + "learning_rate": 1.192036419594768e-05, + "loss": 0.1543, + "step": 59439 + }, + { + "epoch": 2.29, + "learning_rate": 1.191972300589895e-05, + "loss": 0.1701, + "step": 59440 + }, + { + "epoch": 2.29, + "learning_rate": 1.1919081815850218e-05, + "loss": 0.1681, + "step": 59441 + }, + { + "epoch": 2.29, + "learning_rate": 1.191844062580149e-05, + "loss": 0.2472, + "step": 59442 + }, + { + "epoch": 2.29, + "learning_rate": 1.1917799435752758e-05, + "loss": 0.2686, + "step": 59443 + }, + { + "epoch": 2.29, + "learning_rate": 1.1917158245704028e-05, + "loss": 0.2117, + "step": 59444 + }, + { + "epoch": 2.29, + "learning_rate": 1.1916517055655296e-05, + "loss": 0.3207, + "step": 59445 + }, + { + "epoch": 2.29, + "learning_rate": 1.1915875865606568e-05, + "loss": 0.2352, + "step": 59446 + }, + { + "epoch": 2.29, + "learning_rate": 1.1915234675557836e-05, + "loss": 0.1733, + "step": 59447 + }, + { + "epoch": 2.29, + "learning_rate": 1.1914593485509106e-05, + "loss": 0.2084, + "step": 59448 + }, + { + "epoch": 2.29, + "learning_rate": 1.1913952295460374e-05, + "loss": 0.2476, + "step": 59449 + }, + { + "epoch": 2.29, + "learning_rate": 1.1913311105411644e-05, + "loss": 0.141, + "step": 59450 + }, + { + "epoch": 2.29, + "learning_rate": 1.1912669915362914e-05, + "loss": 0.2079, + "step": 59451 + }, + { + "epoch": 2.29, + "learning_rate": 1.1912028725314184e-05, + "loss": 0.3969, + "step": 59452 + }, + { + "epoch": 2.29, + "learning_rate": 1.1911387535265453e-05, + "loss": 0.1927, + "step": 59453 + }, + { + "epoch": 2.29, + "learning_rate": 1.1910746345216723e-05, + "loss": 0.1523, + "step": 59454 + }, + { + "epoch": 2.29, + "learning_rate": 1.1910105155167993e-05, + "loss": 0.2339, + "step": 59455 + }, + { + "epoch": 2.29, + "learning_rate": 1.1909463965119263e-05, + "loss": 0.1861, + "step": 59456 + }, + { + "epoch": 2.29, + "learning_rate": 1.1908822775070531e-05, + "loss": 0.2878, + "step": 59457 + }, + { + "epoch": 2.29, + "learning_rate": 1.1908181585021801e-05, + "loss": 0.2561, + "step": 59458 + }, + { + "epoch": 2.29, + "learning_rate": 1.1907540394973071e-05, + "loss": 0.2091, + "step": 59459 + }, + { + "epoch": 2.29, + "learning_rate": 1.1906899204924341e-05, + "loss": 0.1902, + "step": 59460 + }, + { + "epoch": 2.29, + "learning_rate": 1.190625801487561e-05, + "loss": 0.2254, + "step": 59461 + }, + { + "epoch": 2.29, + "learning_rate": 1.190561682482688e-05, + "loss": 0.219, + "step": 59462 + }, + { + "epoch": 2.29, + "learning_rate": 1.1904975634778148e-05, + "loss": 0.2036, + "step": 59463 + }, + { + "epoch": 2.29, + "learning_rate": 1.190433444472942e-05, + "loss": 0.2773, + "step": 59464 + }, + { + "epoch": 2.29, + "learning_rate": 1.1903693254680688e-05, + "loss": 0.2514, + "step": 59465 + }, + { + "epoch": 2.29, + "learning_rate": 1.1903052064631958e-05, + "loss": 0.2983, + "step": 59466 + }, + { + "epoch": 2.29, + "learning_rate": 1.1902410874583226e-05, + "loss": 0.1433, + "step": 59467 + }, + { + "epoch": 2.29, + "learning_rate": 1.1901769684534498e-05, + "loss": 0.1655, + "step": 59468 + }, + { + "epoch": 2.29, + "learning_rate": 1.1901128494485766e-05, + "loss": 0.233, + "step": 59469 + }, + { + "epoch": 2.29, + "learning_rate": 1.1900487304437036e-05, + "loss": 0.1856, + "step": 59470 + }, + { + "epoch": 2.29, + "learning_rate": 1.1899846114388305e-05, + "loss": 0.1874, + "step": 59471 + }, + { + "epoch": 2.29, + "learning_rate": 1.1899204924339575e-05, + "loss": 0.2602, + "step": 59472 + }, + { + "epoch": 2.29, + "learning_rate": 1.1898563734290845e-05, + "loss": 0.1833, + "step": 59473 + }, + { + "epoch": 2.29, + "learning_rate": 1.1897922544242115e-05, + "loss": 0.2823, + "step": 59474 + }, + { + "epoch": 2.29, + "learning_rate": 1.1897281354193383e-05, + "loss": 0.2406, + "step": 59475 + }, + { + "epoch": 2.29, + "learning_rate": 1.1896640164144653e-05, + "loss": 0.2769, + "step": 59476 + }, + { + "epoch": 2.29, + "learning_rate": 1.1895998974095923e-05, + "loss": 0.2226, + "step": 59477 + }, + { + "epoch": 2.29, + "learning_rate": 1.1895357784047193e-05, + "loss": 0.1908, + "step": 59478 + }, + { + "epoch": 2.29, + "learning_rate": 1.1894716593998461e-05, + "loss": 0.14, + "step": 59479 + }, + { + "epoch": 2.29, + "learning_rate": 1.1894075403949731e-05, + "loss": 0.2244, + "step": 59480 + }, + { + "epoch": 2.29, + "learning_rate": 1.1893434213901e-05, + "loss": 0.1998, + "step": 59481 + }, + { + "epoch": 2.29, + "learning_rate": 1.1892793023852272e-05, + "loss": 0.2129, + "step": 59482 + }, + { + "epoch": 2.29, + "learning_rate": 1.189215183380354e-05, + "loss": 0.1314, + "step": 59483 + }, + { + "epoch": 2.29, + "learning_rate": 1.189151064375481e-05, + "loss": 0.253, + "step": 59484 + }, + { + "epoch": 2.29, + "learning_rate": 1.1890869453706078e-05, + "loss": 0.1967, + "step": 59485 + }, + { + "epoch": 2.29, + "learning_rate": 1.189022826365735e-05, + "loss": 0.1174, + "step": 59486 + }, + { + "epoch": 2.29, + "learning_rate": 1.1889587073608618e-05, + "loss": 0.3085, + "step": 59487 + }, + { + "epoch": 2.29, + "learning_rate": 1.1888945883559888e-05, + "loss": 0.3114, + "step": 59488 + }, + { + "epoch": 2.29, + "learning_rate": 1.1888304693511157e-05, + "loss": 0.1811, + "step": 59489 + }, + { + "epoch": 2.29, + "learning_rate": 1.1887663503462427e-05, + "loss": 0.3211, + "step": 59490 + }, + { + "epoch": 2.29, + "learning_rate": 1.1887022313413697e-05, + "loss": 0.2129, + "step": 59491 + }, + { + "epoch": 2.29, + "learning_rate": 1.1886381123364967e-05, + "loss": 0.3771, + "step": 59492 + }, + { + "epoch": 2.29, + "learning_rate": 1.1885739933316235e-05, + "loss": 0.2664, + "step": 59493 + }, + { + "epoch": 2.29, + "learning_rate": 1.1885098743267505e-05, + "loss": 0.2049, + "step": 59494 + }, + { + "epoch": 2.29, + "learning_rate": 1.1884457553218775e-05, + "loss": 0.1674, + "step": 59495 + }, + { + "epoch": 2.29, + "learning_rate": 1.1883816363170045e-05, + "loss": 0.198, + "step": 59496 + }, + { + "epoch": 2.29, + "learning_rate": 1.1883175173121313e-05, + "loss": 0.2567, + "step": 59497 + }, + { + "epoch": 2.29, + "learning_rate": 1.1882533983072583e-05, + "loss": 0.2574, + "step": 59498 + }, + { + "epoch": 2.29, + "learning_rate": 1.1881892793023852e-05, + "loss": 0.1724, + "step": 59499 + }, + { + "epoch": 2.29, + "learning_rate": 1.1881251602975123e-05, + "loss": 0.1747, + "step": 59500 + }, + { + "epoch": 2.29, + "learning_rate": 1.1880610412926392e-05, + "loss": 0.3323, + "step": 59501 + }, + { + "epoch": 2.29, + "learning_rate": 1.1879969222877662e-05, + "loss": 0.3295, + "step": 59502 + }, + { + "epoch": 2.29, + "learning_rate": 1.187932803282893e-05, + "loss": 0.2668, + "step": 59503 + }, + { + "epoch": 2.29, + "learning_rate": 1.1878686842780202e-05, + "loss": 0.2173, + "step": 59504 + }, + { + "epoch": 2.29, + "learning_rate": 1.187804565273147e-05, + "loss": 0.2304, + "step": 59505 + }, + { + "epoch": 2.29, + "learning_rate": 1.187740446268274e-05, + "loss": 0.1099, + "step": 59506 + }, + { + "epoch": 2.29, + "learning_rate": 1.1876763272634009e-05, + "loss": 0.2627, + "step": 59507 + }, + { + "epoch": 2.29, + "learning_rate": 1.187612208258528e-05, + "loss": 0.1819, + "step": 59508 + }, + { + "epoch": 2.29, + "learning_rate": 1.1875480892536549e-05, + "loss": 0.269, + "step": 59509 + }, + { + "epoch": 2.29, + "learning_rate": 1.1874839702487819e-05, + "loss": 0.1674, + "step": 59510 + }, + { + "epoch": 2.29, + "learning_rate": 1.1874198512439087e-05, + "loss": 0.2233, + "step": 59511 + }, + { + "epoch": 2.29, + "learning_rate": 1.1873557322390357e-05, + "loss": 0.2472, + "step": 59512 + }, + { + "epoch": 2.29, + "learning_rate": 1.1872916132341627e-05, + "loss": 0.3224, + "step": 59513 + }, + { + "epoch": 2.29, + "learning_rate": 1.1872274942292897e-05, + "loss": 0.2247, + "step": 59514 + }, + { + "epoch": 2.29, + "learning_rate": 1.1871633752244165e-05, + "loss": 0.2025, + "step": 59515 + }, + { + "epoch": 2.29, + "learning_rate": 1.1870992562195435e-05, + "loss": 0.1971, + "step": 59516 + }, + { + "epoch": 2.29, + "learning_rate": 1.1870351372146705e-05, + "loss": 0.1975, + "step": 59517 + }, + { + "epoch": 2.29, + "learning_rate": 1.1869710182097975e-05, + "loss": 0.343, + "step": 59518 + }, + { + "epoch": 2.29, + "learning_rate": 1.1869068992049244e-05, + "loss": 0.2907, + "step": 59519 + }, + { + "epoch": 2.29, + "learning_rate": 1.1868427802000514e-05, + "loss": 0.3323, + "step": 59520 + }, + { + "epoch": 2.29, + "learning_rate": 1.1867786611951782e-05, + "loss": 0.1796, + "step": 59521 + }, + { + "epoch": 2.29, + "learning_rate": 1.1867145421903054e-05, + "loss": 0.1713, + "step": 59522 + }, + { + "epoch": 2.29, + "learning_rate": 1.1866504231854322e-05, + "loss": 0.1333, + "step": 59523 + }, + { + "epoch": 2.29, + "learning_rate": 1.1865863041805592e-05, + "loss": 0.1986, + "step": 59524 + }, + { + "epoch": 2.29, + "learning_rate": 1.186522185175686e-05, + "loss": 0.2455, + "step": 59525 + }, + { + "epoch": 2.29, + "learning_rate": 1.1864580661708132e-05, + "loss": 0.364, + "step": 59526 + }, + { + "epoch": 2.29, + "learning_rate": 1.18639394716594e-05, + "loss": 0.171, + "step": 59527 + }, + { + "epoch": 2.29, + "learning_rate": 1.186329828161067e-05, + "loss": 0.2205, + "step": 59528 + }, + { + "epoch": 2.29, + "learning_rate": 1.1862657091561939e-05, + "loss": 0.346, + "step": 59529 + }, + { + "epoch": 2.29, + "learning_rate": 1.1862015901513209e-05, + "loss": 0.2314, + "step": 59530 + }, + { + "epoch": 2.29, + "learning_rate": 1.1861374711464479e-05, + "loss": 0.1473, + "step": 59531 + }, + { + "epoch": 2.29, + "learning_rate": 1.1860733521415749e-05, + "loss": 0.3355, + "step": 59532 + }, + { + "epoch": 2.29, + "learning_rate": 1.1860092331367017e-05, + "loss": 0.2556, + "step": 59533 + }, + { + "epoch": 2.29, + "learning_rate": 1.1859451141318287e-05, + "loss": 0.2846, + "step": 59534 + }, + { + "epoch": 2.29, + "learning_rate": 1.1858809951269557e-05, + "loss": 0.2669, + "step": 59535 + }, + { + "epoch": 2.29, + "learning_rate": 1.1858168761220827e-05, + "loss": 0.2241, + "step": 59536 + }, + { + "epoch": 2.29, + "learning_rate": 1.1857527571172096e-05, + "loss": 0.2062, + "step": 59537 + }, + { + "epoch": 2.29, + "learning_rate": 1.1856886381123366e-05, + "loss": 0.2993, + "step": 59538 + }, + { + "epoch": 2.29, + "learning_rate": 1.1856245191074634e-05, + "loss": 0.1791, + "step": 59539 + }, + { + "epoch": 2.29, + "learning_rate": 1.1855604001025906e-05, + "loss": 0.1693, + "step": 59540 + }, + { + "epoch": 2.29, + "learning_rate": 1.1854962810977174e-05, + "loss": 0.2488, + "step": 59541 + }, + { + "epoch": 2.29, + "learning_rate": 1.1854321620928444e-05, + "loss": 0.2807, + "step": 59542 + }, + { + "epoch": 2.29, + "learning_rate": 1.1853680430879712e-05, + "loss": 0.1041, + "step": 59543 + }, + { + "epoch": 2.29, + "learning_rate": 1.1853039240830984e-05, + "loss": 0.2387, + "step": 59544 + }, + { + "epoch": 2.29, + "learning_rate": 1.1852398050782252e-05, + "loss": 0.2728, + "step": 59545 + }, + { + "epoch": 2.29, + "learning_rate": 1.1851756860733522e-05, + "loss": 0.2288, + "step": 59546 + }, + { + "epoch": 2.29, + "learning_rate": 1.185111567068479e-05, + "loss": 0.2553, + "step": 59547 + }, + { + "epoch": 2.29, + "learning_rate": 1.185047448063606e-05, + "loss": 0.2389, + "step": 59548 + }, + { + "epoch": 2.29, + "learning_rate": 1.184983329058733e-05, + "loss": 0.2132, + "step": 59549 + }, + { + "epoch": 2.29, + "learning_rate": 1.1849192100538601e-05, + "loss": 0.1577, + "step": 59550 + }, + { + "epoch": 2.29, + "learning_rate": 1.184855091048987e-05, + "loss": 0.2356, + "step": 59551 + }, + { + "epoch": 2.29, + "learning_rate": 1.184790972044114e-05, + "loss": 0.2227, + "step": 59552 + }, + { + "epoch": 2.29, + "learning_rate": 1.184726853039241e-05, + "loss": 0.2309, + "step": 59553 + }, + { + "epoch": 2.29, + "learning_rate": 1.184662734034368e-05, + "loss": 0.2087, + "step": 59554 + }, + { + "epoch": 2.29, + "learning_rate": 1.1845986150294948e-05, + "loss": 0.2394, + "step": 59555 + }, + { + "epoch": 2.29, + "learning_rate": 1.1845344960246218e-05, + "loss": 0.2145, + "step": 59556 + }, + { + "epoch": 2.29, + "learning_rate": 1.1844703770197488e-05, + "loss": 0.3212, + "step": 59557 + }, + { + "epoch": 2.29, + "learning_rate": 1.1844062580148758e-05, + "loss": 0.2797, + "step": 59558 + }, + { + "epoch": 2.29, + "learning_rate": 1.1843421390100026e-05, + "loss": 0.246, + "step": 59559 + }, + { + "epoch": 2.29, + "learning_rate": 1.1842780200051296e-05, + "loss": 0.1493, + "step": 59560 + }, + { + "epoch": 2.29, + "learning_rate": 1.1842139010002564e-05, + "loss": 0.1385, + "step": 59561 + }, + { + "epoch": 2.29, + "learning_rate": 1.1841497819953836e-05, + "loss": 0.3392, + "step": 59562 + }, + { + "epoch": 2.29, + "learning_rate": 1.1840856629905104e-05, + "loss": 0.2186, + "step": 59563 + }, + { + "epoch": 2.29, + "learning_rate": 1.1840215439856374e-05, + "loss": 0.1436, + "step": 59564 + }, + { + "epoch": 2.29, + "learning_rate": 1.1839574249807643e-05, + "loss": 0.3017, + "step": 59565 + }, + { + "epoch": 2.29, + "learning_rate": 1.1838933059758914e-05, + "loss": 0.1738, + "step": 59566 + }, + { + "epoch": 2.29, + "learning_rate": 1.1838291869710183e-05, + "loss": 0.1909, + "step": 59567 + }, + { + "epoch": 2.29, + "learning_rate": 1.1837650679661453e-05, + "loss": 0.2284, + "step": 59568 + }, + { + "epoch": 2.29, + "learning_rate": 1.1837009489612721e-05, + "loss": 0.103, + "step": 59569 + }, + { + "epoch": 2.29, + "learning_rate": 1.1836368299563991e-05, + "loss": 0.2313, + "step": 59570 + }, + { + "epoch": 2.29, + "learning_rate": 1.1835727109515261e-05, + "loss": 0.1948, + "step": 59571 + }, + { + "epoch": 2.29, + "learning_rate": 1.1835085919466531e-05, + "loss": 0.2827, + "step": 59572 + }, + { + "epoch": 2.29, + "learning_rate": 1.18344447294178e-05, + "loss": 0.2496, + "step": 59573 + }, + { + "epoch": 2.29, + "learning_rate": 1.183380353936907e-05, + "loss": 0.4096, + "step": 59574 + }, + { + "epoch": 2.29, + "learning_rate": 1.183316234932034e-05, + "loss": 0.2513, + "step": 59575 + }, + { + "epoch": 2.29, + "learning_rate": 1.183252115927161e-05, + "loss": 0.2562, + "step": 59576 + }, + { + "epoch": 2.29, + "learning_rate": 1.1831879969222878e-05, + "loss": 0.2063, + "step": 59577 + }, + { + "epoch": 2.29, + "learning_rate": 1.1831238779174148e-05, + "loss": 0.2281, + "step": 59578 + }, + { + "epoch": 2.29, + "learning_rate": 1.1830597589125416e-05, + "loss": 0.1775, + "step": 59579 + }, + { + "epoch": 2.29, + "learning_rate": 1.1829956399076688e-05, + "loss": 0.2538, + "step": 59580 + }, + { + "epoch": 2.29, + "learning_rate": 1.1829315209027956e-05, + "loss": 0.1507, + "step": 59581 + }, + { + "epoch": 2.29, + "learning_rate": 1.1828674018979226e-05, + "loss": 0.3175, + "step": 59582 + }, + { + "epoch": 2.29, + "learning_rate": 1.1828032828930495e-05, + "loss": 0.1489, + "step": 59583 + }, + { + "epoch": 2.29, + "learning_rate": 1.1827391638881766e-05, + "loss": 0.2605, + "step": 59584 + }, + { + "epoch": 2.29, + "learning_rate": 1.1826750448833035e-05, + "loss": 0.1406, + "step": 59585 + }, + { + "epoch": 2.29, + "learning_rate": 1.1826109258784305e-05, + "loss": 0.1164, + "step": 59586 + }, + { + "epoch": 2.29, + "learning_rate": 1.1825468068735573e-05, + "loss": 0.1478, + "step": 59587 + }, + { + "epoch": 2.29, + "learning_rate": 1.1824826878686843e-05, + "loss": 0.2194, + "step": 59588 + }, + { + "epoch": 2.29, + "learning_rate": 1.1824185688638113e-05, + "loss": 0.2514, + "step": 59589 + }, + { + "epoch": 2.29, + "learning_rate": 1.1823544498589383e-05, + "loss": 0.389, + "step": 59590 + }, + { + "epoch": 2.29, + "learning_rate": 1.1822903308540651e-05, + "loss": 0.2669, + "step": 59591 + }, + { + "epoch": 2.29, + "learning_rate": 1.1822262118491921e-05, + "loss": 0.2564, + "step": 59592 + }, + { + "epoch": 2.29, + "learning_rate": 1.1821620928443191e-05, + "loss": 0.3032, + "step": 59593 + }, + { + "epoch": 2.29, + "learning_rate": 1.1820979738394462e-05, + "loss": 0.3166, + "step": 59594 + }, + { + "epoch": 2.29, + "learning_rate": 1.182033854834573e-05, + "loss": 0.2171, + "step": 59595 + }, + { + "epoch": 2.29, + "learning_rate": 1.1819697358297e-05, + "loss": 0.1505, + "step": 59596 + }, + { + "epoch": 2.29, + "learning_rate": 1.1819056168248268e-05, + "loss": 0.2102, + "step": 59597 + }, + { + "epoch": 2.29, + "learning_rate": 1.181841497819954e-05, + "loss": 0.2502, + "step": 59598 + }, + { + "epoch": 2.29, + "learning_rate": 1.1817773788150808e-05, + "loss": 0.1876, + "step": 59599 + }, + { + "epoch": 2.29, + "learning_rate": 1.1817132598102078e-05, + "loss": 0.2398, + "step": 59600 + }, + { + "epoch": 2.29, + "learning_rate": 1.1816491408053347e-05, + "loss": 0.2326, + "step": 59601 + }, + { + "epoch": 2.29, + "learning_rate": 1.1815850218004618e-05, + "loss": 0.1822, + "step": 59602 + }, + { + "epoch": 2.29, + "learning_rate": 1.1815209027955887e-05, + "loss": 0.1626, + "step": 59603 + }, + { + "epoch": 2.29, + "learning_rate": 1.1814567837907157e-05, + "loss": 0.2257, + "step": 59604 + }, + { + "epoch": 2.29, + "learning_rate": 1.1813926647858425e-05, + "loss": 0.2317, + "step": 59605 + }, + { + "epoch": 2.29, + "learning_rate": 1.1813285457809697e-05, + "loss": 0.1708, + "step": 59606 + }, + { + "epoch": 2.29, + "learning_rate": 1.1812644267760965e-05, + "loss": 0.1639, + "step": 59607 + }, + { + "epoch": 2.29, + "learning_rate": 1.1812003077712235e-05, + "loss": 0.1427, + "step": 59608 + }, + { + "epoch": 2.29, + "learning_rate": 1.1811361887663503e-05, + "loss": 0.1895, + "step": 59609 + }, + { + "epoch": 2.29, + "learning_rate": 1.1810720697614773e-05, + "loss": 0.2447, + "step": 59610 + }, + { + "epoch": 2.29, + "learning_rate": 1.1810079507566043e-05, + "loss": 0.1961, + "step": 59611 + }, + { + "epoch": 2.29, + "learning_rate": 1.1809438317517313e-05, + "loss": 0.2248, + "step": 59612 + }, + { + "epoch": 2.29, + "learning_rate": 1.1808797127468582e-05, + "loss": 0.2694, + "step": 59613 + }, + { + "epoch": 2.29, + "learning_rate": 1.1808155937419852e-05, + "loss": 0.1949, + "step": 59614 + }, + { + "epoch": 2.29, + "learning_rate": 1.1807514747371122e-05, + "loss": 0.2986, + "step": 59615 + }, + { + "epoch": 2.29, + "learning_rate": 1.1806873557322392e-05, + "loss": 0.1811, + "step": 59616 + }, + { + "epoch": 2.29, + "learning_rate": 1.180623236727366e-05, + "loss": 0.1928, + "step": 59617 + }, + { + "epoch": 2.29, + "learning_rate": 1.180559117722493e-05, + "loss": 0.2155, + "step": 59618 + }, + { + "epoch": 2.29, + "learning_rate": 1.1804949987176199e-05, + "loss": 0.2785, + "step": 59619 + }, + { + "epoch": 2.29, + "learning_rate": 1.180430879712747e-05, + "loss": 0.2512, + "step": 59620 + }, + { + "epoch": 2.29, + "learning_rate": 1.1803667607078739e-05, + "loss": 0.2899, + "step": 59621 + }, + { + "epoch": 2.29, + "learning_rate": 1.1803026417030009e-05, + "loss": 0.1848, + "step": 59622 + }, + { + "epoch": 2.29, + "learning_rate": 1.1802385226981277e-05, + "loss": 0.2265, + "step": 59623 + }, + { + "epoch": 2.29, + "learning_rate": 1.1801744036932549e-05, + "loss": 0.2863, + "step": 59624 + }, + { + "epoch": 2.29, + "learning_rate": 1.1801102846883817e-05, + "loss": 0.4599, + "step": 59625 + }, + { + "epoch": 2.29, + "learning_rate": 1.1800461656835087e-05, + "loss": 0.1654, + "step": 59626 + }, + { + "epoch": 2.29, + "learning_rate": 1.1799820466786355e-05, + "loss": 0.2267, + "step": 59627 + }, + { + "epoch": 2.29, + "learning_rate": 1.1799179276737625e-05, + "loss": 0.1331, + "step": 59628 + }, + { + "epoch": 2.29, + "learning_rate": 1.1798538086688895e-05, + "loss": 0.2907, + "step": 59629 + }, + { + "epoch": 2.29, + "learning_rate": 1.1797896896640165e-05, + "loss": 0.3618, + "step": 59630 + }, + { + "epoch": 2.29, + "learning_rate": 1.1797255706591434e-05, + "loss": 0.1957, + "step": 59631 + }, + { + "epoch": 2.29, + "learning_rate": 1.1796614516542704e-05, + "loss": 0.2212, + "step": 59632 + }, + { + "epoch": 2.29, + "learning_rate": 1.1795973326493974e-05, + "loss": 0.1584, + "step": 59633 + }, + { + "epoch": 2.29, + "learning_rate": 1.1795332136445244e-05, + "loss": 0.3025, + "step": 59634 + }, + { + "epoch": 2.29, + "learning_rate": 1.1794690946396512e-05, + "loss": 0.3518, + "step": 59635 + }, + { + "epoch": 2.29, + "learning_rate": 1.1794049756347782e-05, + "loss": 0.1899, + "step": 59636 + }, + { + "epoch": 2.29, + "learning_rate": 1.179340856629905e-05, + "loss": 0.2505, + "step": 59637 + }, + { + "epoch": 2.29, + "learning_rate": 1.1792767376250322e-05, + "loss": 0.3381, + "step": 59638 + }, + { + "epoch": 2.29, + "learning_rate": 1.179212618620159e-05, + "loss": 0.1125, + "step": 59639 + }, + { + "epoch": 2.29, + "learning_rate": 1.179148499615286e-05, + "loss": 0.2505, + "step": 59640 + }, + { + "epoch": 2.29, + "learning_rate": 1.1790843806104129e-05, + "loss": 0.1809, + "step": 59641 + }, + { + "epoch": 2.29, + "learning_rate": 1.17902026160554e-05, + "loss": 0.2997, + "step": 59642 + }, + { + "epoch": 2.29, + "learning_rate": 1.1789561426006669e-05, + "loss": 0.3567, + "step": 59643 + }, + { + "epoch": 2.29, + "learning_rate": 1.1788920235957939e-05, + "loss": 0.2726, + "step": 59644 + }, + { + "epoch": 2.29, + "learning_rate": 1.1788279045909207e-05, + "loss": 0.1964, + "step": 59645 + }, + { + "epoch": 2.29, + "learning_rate": 1.1787637855860477e-05, + "loss": 0.3311, + "step": 59646 + }, + { + "epoch": 2.29, + "learning_rate": 1.1786996665811747e-05, + "loss": 0.2252, + "step": 59647 + }, + { + "epoch": 2.29, + "learning_rate": 1.1786355475763017e-05, + "loss": 0.2288, + "step": 59648 + }, + { + "epoch": 2.29, + "learning_rate": 1.1785714285714286e-05, + "loss": 0.1772, + "step": 59649 + }, + { + "epoch": 2.29, + "learning_rate": 1.1785073095665556e-05, + "loss": 0.2316, + "step": 59650 + }, + { + "epoch": 2.29, + "learning_rate": 1.1784431905616826e-05, + "loss": 0.2576, + "step": 59651 + }, + { + "epoch": 2.29, + "learning_rate": 1.1783790715568096e-05, + "loss": 0.1916, + "step": 59652 + }, + { + "epoch": 2.29, + "learning_rate": 1.1783149525519364e-05, + "loss": 0.198, + "step": 59653 + }, + { + "epoch": 2.29, + "learning_rate": 1.1782508335470634e-05, + "loss": 0.268, + "step": 59654 + }, + { + "epoch": 2.29, + "learning_rate": 1.1781867145421904e-05, + "loss": 0.2596, + "step": 59655 + }, + { + "epoch": 2.29, + "learning_rate": 1.1781225955373174e-05, + "loss": 0.2198, + "step": 59656 + }, + { + "epoch": 2.29, + "learning_rate": 1.1780584765324442e-05, + "loss": 0.2022, + "step": 59657 + }, + { + "epoch": 2.29, + "learning_rate": 1.1779943575275712e-05, + "loss": 0.2086, + "step": 59658 + }, + { + "epoch": 2.29, + "learning_rate": 1.177930238522698e-05, + "loss": 0.2699, + "step": 59659 + }, + { + "epoch": 2.29, + "learning_rate": 1.1778661195178253e-05, + "loss": 0.3003, + "step": 59660 + }, + { + "epoch": 2.29, + "learning_rate": 1.177802000512952e-05, + "loss": 0.227, + "step": 59661 + }, + { + "epoch": 2.29, + "learning_rate": 1.1777378815080791e-05, + "loss": 0.1862, + "step": 59662 + }, + { + "epoch": 2.29, + "learning_rate": 1.177673762503206e-05, + "loss": 0.2258, + "step": 59663 + }, + { + "epoch": 2.29, + "learning_rate": 1.1776096434983331e-05, + "loss": 0.2756, + "step": 59664 + }, + { + "epoch": 2.29, + "learning_rate": 1.17754552449346e-05, + "loss": 0.2989, + "step": 59665 + }, + { + "epoch": 2.29, + "learning_rate": 1.177481405488587e-05, + "loss": 0.1864, + "step": 59666 + }, + { + "epoch": 2.29, + "learning_rate": 1.1774172864837138e-05, + "loss": 0.362, + "step": 59667 + }, + { + "epoch": 2.29, + "learning_rate": 1.1773531674788408e-05, + "loss": 0.3804, + "step": 59668 + }, + { + "epoch": 2.29, + "learning_rate": 1.1772890484739678e-05, + "loss": 0.2016, + "step": 59669 + }, + { + "epoch": 2.29, + "learning_rate": 1.1772249294690948e-05, + "loss": 0.2815, + "step": 59670 + }, + { + "epoch": 2.29, + "learning_rate": 1.1771608104642216e-05, + "loss": 0.1292, + "step": 59671 + }, + { + "epoch": 2.29, + "learning_rate": 1.1770966914593486e-05, + "loss": 0.2103, + "step": 59672 + }, + { + "epoch": 2.29, + "learning_rate": 1.1770325724544756e-05, + "loss": 0.3188, + "step": 59673 + }, + { + "epoch": 2.29, + "learning_rate": 1.1769684534496026e-05, + "loss": 0.4728, + "step": 59674 + }, + { + "epoch": 2.29, + "learning_rate": 1.1769043344447294e-05, + "loss": 0.3602, + "step": 59675 + }, + { + "epoch": 2.29, + "learning_rate": 1.1768402154398564e-05, + "loss": 0.2107, + "step": 59676 + }, + { + "epoch": 2.29, + "learning_rate": 1.1767760964349833e-05, + "loss": 0.2138, + "step": 59677 + }, + { + "epoch": 2.29, + "learning_rate": 1.1767119774301104e-05, + "loss": 0.215, + "step": 59678 + }, + { + "epoch": 2.29, + "learning_rate": 1.1766478584252373e-05, + "loss": 0.2951, + "step": 59679 + }, + { + "epoch": 2.29, + "learning_rate": 1.1765837394203643e-05, + "loss": 0.3477, + "step": 59680 + }, + { + "epoch": 2.29, + "learning_rate": 1.1765196204154911e-05, + "loss": 0.288, + "step": 59681 + }, + { + "epoch": 2.29, + "learning_rate": 1.1764555014106183e-05, + "loss": 0.2511, + "step": 59682 + }, + { + "epoch": 2.29, + "learning_rate": 1.1763913824057451e-05, + "loss": 0.3686, + "step": 59683 + }, + { + "epoch": 2.29, + "learning_rate": 1.1763272634008721e-05, + "loss": 0.1789, + "step": 59684 + }, + { + "epoch": 2.29, + "learning_rate": 1.176263144395999e-05, + "loss": 0.2509, + "step": 59685 + }, + { + "epoch": 2.29, + "learning_rate": 1.176199025391126e-05, + "loss": 0.1488, + "step": 59686 + }, + { + "epoch": 2.29, + "learning_rate": 1.176134906386253e-05, + "loss": 0.2058, + "step": 59687 + }, + { + "epoch": 2.29, + "learning_rate": 1.17607078738138e-05, + "loss": 0.1922, + "step": 59688 + }, + { + "epoch": 2.29, + "learning_rate": 1.1760066683765068e-05, + "loss": 0.2227, + "step": 59689 + }, + { + "epoch": 2.29, + "learning_rate": 1.1759425493716338e-05, + "loss": 0.2398, + "step": 59690 + }, + { + "epoch": 2.29, + "learning_rate": 1.1758784303667608e-05, + "loss": 0.1844, + "step": 59691 + }, + { + "epoch": 2.29, + "learning_rate": 1.1758143113618878e-05, + "loss": 0.2193, + "step": 59692 + }, + { + "epoch": 2.3, + "learning_rate": 1.1757501923570146e-05, + "loss": 0.1875, + "step": 59693 + }, + { + "epoch": 2.3, + "learning_rate": 1.1756860733521416e-05, + "loss": 0.209, + "step": 59694 + }, + { + "epoch": 2.3, + "learning_rate": 1.1756219543472686e-05, + "loss": 0.3031, + "step": 59695 + }, + { + "epoch": 2.3, + "learning_rate": 1.1755578353423956e-05, + "loss": 0.347, + "step": 59696 + }, + { + "epoch": 2.3, + "learning_rate": 1.1754937163375225e-05, + "loss": 0.2429, + "step": 59697 + }, + { + "epoch": 2.3, + "learning_rate": 1.1754295973326495e-05, + "loss": 0.2116, + "step": 59698 + }, + { + "epoch": 2.3, + "learning_rate": 1.1753654783277763e-05, + "loss": 0.238, + "step": 59699 + }, + { + "epoch": 2.3, + "learning_rate": 1.1753013593229035e-05, + "loss": 0.1608, + "step": 59700 + }, + { + "epoch": 2.3, + "learning_rate": 1.1752372403180303e-05, + "loss": 0.2689, + "step": 59701 + }, + { + "epoch": 2.3, + "learning_rate": 1.1751731213131573e-05, + "loss": 0.1845, + "step": 59702 + }, + { + "epoch": 2.3, + "learning_rate": 1.1751090023082841e-05, + "loss": 0.2711, + "step": 59703 + }, + { + "epoch": 2.3, + "learning_rate": 1.1750448833034113e-05, + "loss": 0.1622, + "step": 59704 + }, + { + "epoch": 2.3, + "learning_rate": 1.1749807642985382e-05, + "loss": 0.4177, + "step": 59705 + }, + { + "epoch": 2.3, + "learning_rate": 1.1749166452936652e-05, + "loss": 0.2758, + "step": 59706 + }, + { + "epoch": 2.3, + "learning_rate": 1.174852526288792e-05, + "loss": 0.167, + "step": 59707 + }, + { + "epoch": 2.3, + "learning_rate": 1.174788407283919e-05, + "loss": 0.2033, + "step": 59708 + }, + { + "epoch": 2.3, + "learning_rate": 1.174724288279046e-05, + "loss": 0.2901, + "step": 59709 + }, + { + "epoch": 2.3, + "learning_rate": 1.174660169274173e-05, + "loss": 0.2967, + "step": 59710 + }, + { + "epoch": 2.3, + "learning_rate": 1.1745960502692998e-05, + "loss": 0.4242, + "step": 59711 + }, + { + "epoch": 2.3, + "learning_rate": 1.1745319312644268e-05, + "loss": 0.2961, + "step": 59712 + }, + { + "epoch": 2.3, + "learning_rate": 1.1744678122595538e-05, + "loss": 0.2012, + "step": 59713 + }, + { + "epoch": 2.3, + "learning_rate": 1.1744036932546808e-05, + "loss": 0.3261, + "step": 59714 + }, + { + "epoch": 2.3, + "learning_rate": 1.1743395742498077e-05, + "loss": 0.1138, + "step": 59715 + }, + { + "epoch": 2.3, + "learning_rate": 1.1742754552449347e-05, + "loss": 0.0685, + "step": 59716 + }, + { + "epoch": 2.3, + "learning_rate": 1.1742113362400615e-05, + "loss": 0.2176, + "step": 59717 + }, + { + "epoch": 2.3, + "learning_rate": 1.1741472172351887e-05, + "loss": 0.1368, + "step": 59718 + }, + { + "epoch": 2.3, + "learning_rate": 1.1740830982303155e-05, + "loss": 0.2774, + "step": 59719 + }, + { + "epoch": 2.3, + "learning_rate": 1.1740189792254425e-05, + "loss": 0.1461, + "step": 59720 + }, + { + "epoch": 2.3, + "learning_rate": 1.1739548602205693e-05, + "loss": 0.2603, + "step": 59721 + }, + { + "epoch": 2.3, + "learning_rate": 1.1738907412156965e-05, + "loss": 0.2513, + "step": 59722 + }, + { + "epoch": 2.3, + "learning_rate": 1.1738266222108233e-05, + "loss": 0.3007, + "step": 59723 + }, + { + "epoch": 2.3, + "learning_rate": 1.1737625032059503e-05, + "loss": 0.2076, + "step": 59724 + }, + { + "epoch": 2.3, + "learning_rate": 1.1736983842010772e-05, + "loss": 0.1998, + "step": 59725 + }, + { + "epoch": 2.3, + "learning_rate": 1.1736342651962042e-05, + "loss": 0.2934, + "step": 59726 + }, + { + "epoch": 2.3, + "learning_rate": 1.1735701461913312e-05, + "loss": 0.2181, + "step": 59727 + }, + { + "epoch": 2.3, + "learning_rate": 1.1735060271864582e-05, + "loss": 0.2103, + "step": 59728 + }, + { + "epoch": 2.3, + "learning_rate": 1.173441908181585e-05, + "loss": 0.1786, + "step": 59729 + }, + { + "epoch": 2.3, + "learning_rate": 1.173377789176712e-05, + "loss": 0.229, + "step": 59730 + }, + { + "epoch": 2.3, + "learning_rate": 1.173313670171839e-05, + "loss": 0.2942, + "step": 59731 + }, + { + "epoch": 2.3, + "learning_rate": 1.173249551166966e-05, + "loss": 0.1288, + "step": 59732 + }, + { + "epoch": 2.3, + "learning_rate": 1.1731854321620929e-05, + "loss": 0.1149, + "step": 59733 + }, + { + "epoch": 2.3, + "learning_rate": 1.1731213131572199e-05, + "loss": 0.2437, + "step": 59734 + }, + { + "epoch": 2.3, + "learning_rate": 1.1730571941523467e-05, + "loss": 0.2324, + "step": 59735 + }, + { + "epoch": 2.3, + "learning_rate": 1.1729930751474739e-05, + "loss": 0.3693, + "step": 59736 + }, + { + "epoch": 2.3, + "learning_rate": 1.1729289561426007e-05, + "loss": 0.1148, + "step": 59737 + }, + { + "epoch": 2.3, + "learning_rate": 1.1728648371377277e-05, + "loss": 0.2512, + "step": 59738 + }, + { + "epoch": 2.3, + "learning_rate": 1.1728007181328545e-05, + "loss": 0.2712, + "step": 59739 + }, + { + "epoch": 2.3, + "learning_rate": 1.1727365991279817e-05, + "loss": 0.3205, + "step": 59740 + }, + { + "epoch": 2.3, + "learning_rate": 1.1726724801231085e-05, + "loss": 0.1749, + "step": 59741 + }, + { + "epoch": 2.3, + "learning_rate": 1.1726083611182355e-05, + "loss": 0.2432, + "step": 59742 + }, + { + "epoch": 2.3, + "learning_rate": 1.1725442421133624e-05, + "loss": 0.2796, + "step": 59743 + }, + { + "epoch": 2.3, + "learning_rate": 1.1724801231084895e-05, + "loss": 0.1937, + "step": 59744 + }, + { + "epoch": 2.3, + "learning_rate": 1.1724160041036164e-05, + "loss": 0.2128, + "step": 59745 + }, + { + "epoch": 2.3, + "learning_rate": 1.1723518850987434e-05, + "loss": 0.3659, + "step": 59746 + }, + { + "epoch": 2.3, + "learning_rate": 1.1722877660938702e-05, + "loss": 0.2579, + "step": 59747 + }, + { + "epoch": 2.3, + "learning_rate": 1.1722236470889972e-05, + "loss": 0.1883, + "step": 59748 + }, + { + "epoch": 2.3, + "learning_rate": 1.1721595280841242e-05, + "loss": 0.1892, + "step": 59749 + }, + { + "epoch": 2.3, + "learning_rate": 1.1720954090792512e-05, + "loss": 0.2448, + "step": 59750 + }, + { + "epoch": 2.3, + "learning_rate": 1.172031290074378e-05, + "loss": 0.1203, + "step": 59751 + }, + { + "epoch": 2.3, + "learning_rate": 1.171967171069505e-05, + "loss": 0.2194, + "step": 59752 + }, + { + "epoch": 2.3, + "learning_rate": 1.171903052064632e-05, + "loss": 0.2122, + "step": 59753 + }, + { + "epoch": 2.3, + "learning_rate": 1.171838933059759e-05, + "loss": 0.1637, + "step": 59754 + }, + { + "epoch": 2.3, + "learning_rate": 1.1717748140548859e-05, + "loss": 0.1336, + "step": 59755 + }, + { + "epoch": 2.3, + "learning_rate": 1.1717106950500129e-05, + "loss": 0.1541, + "step": 59756 + }, + { + "epoch": 2.3, + "learning_rate": 1.1716465760451397e-05, + "loss": 0.1894, + "step": 59757 + }, + { + "epoch": 2.3, + "learning_rate": 1.1715824570402669e-05, + "loss": 0.3025, + "step": 59758 + }, + { + "epoch": 2.3, + "learning_rate": 1.1715183380353937e-05, + "loss": 0.2516, + "step": 59759 + }, + { + "epoch": 2.3, + "learning_rate": 1.1714542190305207e-05, + "loss": 0.2097, + "step": 59760 + }, + { + "epoch": 2.3, + "learning_rate": 1.1713901000256476e-05, + "loss": 0.2092, + "step": 59761 + }, + { + "epoch": 2.3, + "learning_rate": 1.1713259810207747e-05, + "loss": 0.3304, + "step": 59762 + }, + { + "epoch": 2.3, + "learning_rate": 1.1712618620159016e-05, + "loss": 0.4375, + "step": 59763 + }, + { + "epoch": 2.3, + "learning_rate": 1.1711977430110286e-05, + "loss": 0.2452, + "step": 59764 + }, + { + "epoch": 2.3, + "learning_rate": 1.1711336240061554e-05, + "loss": 0.1646, + "step": 59765 + }, + { + "epoch": 2.3, + "learning_rate": 1.1710695050012824e-05, + "loss": 0.1655, + "step": 59766 + }, + { + "epoch": 2.3, + "learning_rate": 1.1710053859964094e-05, + "loss": 0.2905, + "step": 59767 + }, + { + "epoch": 2.3, + "learning_rate": 1.1709412669915364e-05, + "loss": 0.3016, + "step": 59768 + }, + { + "epoch": 2.3, + "learning_rate": 1.1708771479866632e-05, + "loss": 0.21, + "step": 59769 + }, + { + "epoch": 2.3, + "learning_rate": 1.1708130289817902e-05, + "loss": 0.1557, + "step": 59770 + }, + { + "epoch": 2.3, + "learning_rate": 1.1707489099769172e-05, + "loss": 0.2744, + "step": 59771 + }, + { + "epoch": 2.3, + "learning_rate": 1.1706847909720443e-05, + "loss": 0.25, + "step": 59772 + }, + { + "epoch": 2.3, + "learning_rate": 1.1706206719671711e-05, + "loss": 0.1332, + "step": 59773 + }, + { + "epoch": 2.3, + "learning_rate": 1.1705565529622981e-05, + "loss": 0.2698, + "step": 59774 + }, + { + "epoch": 2.3, + "learning_rate": 1.170492433957425e-05, + "loss": 0.1937, + "step": 59775 + }, + { + "epoch": 2.3, + "learning_rate": 1.1704283149525521e-05, + "loss": 0.2427, + "step": 59776 + }, + { + "epoch": 2.3, + "learning_rate": 1.170364195947679e-05, + "loss": 0.4015, + "step": 59777 + }, + { + "epoch": 2.3, + "learning_rate": 1.170300076942806e-05, + "loss": 0.2526, + "step": 59778 + }, + { + "epoch": 2.3, + "learning_rate": 1.1702359579379328e-05, + "loss": 0.1771, + "step": 59779 + }, + { + "epoch": 2.3, + "learning_rate": 1.17017183893306e-05, + "loss": 0.1759, + "step": 59780 + }, + { + "epoch": 2.3, + "learning_rate": 1.1701077199281868e-05, + "loss": 0.2607, + "step": 59781 + }, + { + "epoch": 2.3, + "learning_rate": 1.1700436009233138e-05, + "loss": 0.3686, + "step": 59782 + }, + { + "epoch": 2.3, + "learning_rate": 1.1699794819184406e-05, + "loss": 0.3846, + "step": 59783 + }, + { + "epoch": 2.3, + "learning_rate": 1.1699153629135676e-05, + "loss": 0.2786, + "step": 59784 + }, + { + "epoch": 2.3, + "learning_rate": 1.1698512439086946e-05, + "loss": 0.2333, + "step": 59785 + }, + { + "epoch": 2.3, + "learning_rate": 1.1697871249038216e-05, + "loss": 0.2196, + "step": 59786 + }, + { + "epoch": 2.3, + "learning_rate": 1.1697230058989484e-05, + "loss": 0.2362, + "step": 59787 + }, + { + "epoch": 2.3, + "learning_rate": 1.1696588868940754e-05, + "loss": 0.2421, + "step": 59788 + }, + { + "epoch": 2.3, + "learning_rate": 1.1695947678892024e-05, + "loss": 0.322, + "step": 59789 + }, + { + "epoch": 2.3, + "learning_rate": 1.1695306488843294e-05, + "loss": 0.1939, + "step": 59790 + }, + { + "epoch": 2.3, + "learning_rate": 1.1694665298794563e-05, + "loss": 0.2408, + "step": 59791 + }, + { + "epoch": 2.3, + "learning_rate": 1.1694024108745833e-05, + "loss": 0.2916, + "step": 59792 + }, + { + "epoch": 2.3, + "learning_rate": 1.1693382918697103e-05, + "loss": 0.1664, + "step": 59793 + }, + { + "epoch": 2.3, + "learning_rate": 1.1692741728648373e-05, + "loss": 0.211, + "step": 59794 + }, + { + "epoch": 2.3, + "learning_rate": 1.1692100538599641e-05, + "loss": 0.2387, + "step": 59795 + }, + { + "epoch": 2.3, + "learning_rate": 1.1691459348550911e-05, + "loss": 0.1594, + "step": 59796 + }, + { + "epoch": 2.3, + "learning_rate": 1.169081815850218e-05, + "loss": 0.0988, + "step": 59797 + }, + { + "epoch": 2.3, + "learning_rate": 1.1690176968453451e-05, + "loss": 0.2734, + "step": 59798 + }, + { + "epoch": 2.3, + "learning_rate": 1.168953577840472e-05, + "loss": 0.2175, + "step": 59799 + }, + { + "epoch": 2.3, + "learning_rate": 1.168889458835599e-05, + "loss": 0.2666, + "step": 59800 + }, + { + "epoch": 2.3, + "learning_rate": 1.1688253398307258e-05, + "loss": 0.3103, + "step": 59801 + }, + { + "epoch": 2.3, + "learning_rate": 1.168761220825853e-05, + "loss": 0.2232, + "step": 59802 + }, + { + "epoch": 2.3, + "learning_rate": 1.1686971018209798e-05, + "loss": 0.1352, + "step": 59803 + }, + { + "epoch": 2.3, + "learning_rate": 1.1686329828161068e-05, + "loss": 0.1999, + "step": 59804 + }, + { + "epoch": 2.3, + "learning_rate": 1.1685688638112336e-05, + "loss": 0.2974, + "step": 59805 + }, + { + "epoch": 2.3, + "learning_rate": 1.1685047448063606e-05, + "loss": 0.3394, + "step": 59806 + }, + { + "epoch": 2.3, + "learning_rate": 1.1684406258014876e-05, + "loss": 0.2828, + "step": 59807 + }, + { + "epoch": 2.3, + "learning_rate": 1.1683765067966146e-05, + "loss": 0.2494, + "step": 59808 + }, + { + "epoch": 2.3, + "learning_rate": 1.1683123877917415e-05, + "loss": 0.1068, + "step": 59809 + }, + { + "epoch": 2.3, + "learning_rate": 1.1682482687868685e-05, + "loss": 0.2156, + "step": 59810 + }, + { + "epoch": 2.3, + "learning_rate": 1.1681841497819955e-05, + "loss": 0.2305, + "step": 59811 + }, + { + "epoch": 2.3, + "learning_rate": 1.1681200307771225e-05, + "loss": 0.2229, + "step": 59812 + }, + { + "epoch": 2.3, + "learning_rate": 1.1680559117722493e-05, + "loss": 0.1273, + "step": 59813 + }, + { + "epoch": 2.3, + "learning_rate": 1.1679917927673763e-05, + "loss": 0.2244, + "step": 59814 + }, + { + "epoch": 2.3, + "learning_rate": 1.1679276737625031e-05, + "loss": 0.2656, + "step": 59815 + }, + { + "epoch": 2.3, + "learning_rate": 1.1678635547576303e-05, + "loss": 0.3255, + "step": 59816 + }, + { + "epoch": 2.3, + "learning_rate": 1.1677994357527572e-05, + "loss": 0.1614, + "step": 59817 + }, + { + "epoch": 2.3, + "learning_rate": 1.1677353167478842e-05, + "loss": 0.2691, + "step": 59818 + }, + { + "epoch": 2.3, + "learning_rate": 1.167671197743011e-05, + "loss": 0.2543, + "step": 59819 + }, + { + "epoch": 2.3, + "learning_rate": 1.1676070787381382e-05, + "loss": 0.3323, + "step": 59820 + }, + { + "epoch": 2.3, + "learning_rate": 1.167542959733265e-05, + "loss": 0.3069, + "step": 59821 + }, + { + "epoch": 2.3, + "learning_rate": 1.167478840728392e-05, + "loss": 0.1439, + "step": 59822 + }, + { + "epoch": 2.3, + "learning_rate": 1.1674147217235188e-05, + "loss": 0.2441, + "step": 59823 + }, + { + "epoch": 2.3, + "learning_rate": 1.1673506027186458e-05, + "loss": 0.3744, + "step": 59824 + }, + { + "epoch": 2.3, + "learning_rate": 1.1672864837137728e-05, + "loss": 0.241, + "step": 59825 + }, + { + "epoch": 2.3, + "learning_rate": 1.1672223647088998e-05, + "loss": 0.2493, + "step": 59826 + }, + { + "epoch": 2.3, + "learning_rate": 1.1671582457040267e-05, + "loss": 0.3439, + "step": 59827 + }, + { + "epoch": 2.3, + "learning_rate": 1.1670941266991537e-05, + "loss": 0.4411, + "step": 59828 + }, + { + "epoch": 2.3, + "learning_rate": 1.1670300076942807e-05, + "loss": 0.133, + "step": 59829 + }, + { + "epoch": 2.3, + "learning_rate": 1.1669658886894077e-05, + "loss": 0.2963, + "step": 59830 + }, + { + "epoch": 2.3, + "learning_rate": 1.1669017696845345e-05, + "loss": 0.1867, + "step": 59831 + }, + { + "epoch": 2.3, + "learning_rate": 1.1668376506796615e-05, + "loss": 0.1761, + "step": 59832 + }, + { + "epoch": 2.3, + "learning_rate": 1.1667735316747883e-05, + "loss": 0.1302, + "step": 59833 + }, + { + "epoch": 2.3, + "learning_rate": 1.1667094126699155e-05, + "loss": 0.2858, + "step": 59834 + }, + { + "epoch": 2.3, + "learning_rate": 1.1666452936650423e-05, + "loss": 0.2867, + "step": 59835 + }, + { + "epoch": 2.3, + "learning_rate": 1.1665811746601693e-05, + "loss": 0.3426, + "step": 59836 + }, + { + "epoch": 2.3, + "learning_rate": 1.1665170556552962e-05, + "loss": 0.2748, + "step": 59837 + }, + { + "epoch": 2.3, + "learning_rate": 1.1664529366504234e-05, + "loss": 0.2135, + "step": 59838 + }, + { + "epoch": 2.3, + "learning_rate": 1.1663888176455502e-05, + "loss": 0.245, + "step": 59839 + }, + { + "epoch": 2.3, + "learning_rate": 1.1663246986406772e-05, + "loss": 0.2252, + "step": 59840 + }, + { + "epoch": 2.3, + "learning_rate": 1.166260579635804e-05, + "loss": 0.2839, + "step": 59841 + }, + { + "epoch": 2.3, + "learning_rate": 1.1661964606309312e-05, + "loss": 0.2507, + "step": 59842 + }, + { + "epoch": 2.3, + "learning_rate": 1.166132341626058e-05, + "loss": 0.1479, + "step": 59843 + }, + { + "epoch": 2.3, + "learning_rate": 1.166068222621185e-05, + "loss": 0.2999, + "step": 59844 + }, + { + "epoch": 2.3, + "learning_rate": 1.1660041036163119e-05, + "loss": 0.2447, + "step": 59845 + }, + { + "epoch": 2.3, + "learning_rate": 1.1659399846114389e-05, + "loss": 0.2473, + "step": 59846 + }, + { + "epoch": 2.3, + "learning_rate": 1.1658758656065659e-05, + "loss": 0.2044, + "step": 59847 + }, + { + "epoch": 2.3, + "learning_rate": 1.1658117466016929e-05, + "loss": 0.1531, + "step": 59848 + }, + { + "epoch": 2.3, + "learning_rate": 1.1657476275968197e-05, + "loss": 0.4056, + "step": 59849 + }, + { + "epoch": 2.3, + "learning_rate": 1.1656835085919467e-05, + "loss": 0.2901, + "step": 59850 + }, + { + "epoch": 2.3, + "learning_rate": 1.1656193895870737e-05, + "loss": 0.1969, + "step": 59851 + }, + { + "epoch": 2.3, + "learning_rate": 1.1655552705822007e-05, + "loss": 0.2039, + "step": 59852 + }, + { + "epoch": 2.3, + "learning_rate": 1.1654911515773275e-05, + "loss": 0.2043, + "step": 59853 + }, + { + "epoch": 2.3, + "learning_rate": 1.1654270325724545e-05, + "loss": 0.229, + "step": 59854 + }, + { + "epoch": 2.3, + "learning_rate": 1.1653629135675814e-05, + "loss": 0.3592, + "step": 59855 + }, + { + "epoch": 2.3, + "learning_rate": 1.1652987945627085e-05, + "loss": 0.3298, + "step": 59856 + }, + { + "epoch": 2.3, + "learning_rate": 1.1652346755578354e-05, + "loss": 0.3168, + "step": 59857 + }, + { + "epoch": 2.3, + "learning_rate": 1.1651705565529624e-05, + "loss": 0.4426, + "step": 59858 + }, + { + "epoch": 2.3, + "learning_rate": 1.1651064375480892e-05, + "loss": 0.279, + "step": 59859 + }, + { + "epoch": 2.3, + "learning_rate": 1.1650423185432164e-05, + "loss": 0.1671, + "step": 59860 + }, + { + "epoch": 2.3, + "learning_rate": 1.1649781995383432e-05, + "loss": 0.1678, + "step": 59861 + }, + { + "epoch": 2.3, + "learning_rate": 1.1649140805334702e-05, + "loss": 0.2724, + "step": 59862 + }, + { + "epoch": 2.3, + "learning_rate": 1.164849961528597e-05, + "loss": 0.349, + "step": 59863 + }, + { + "epoch": 2.3, + "learning_rate": 1.164785842523724e-05, + "loss": 0.3234, + "step": 59864 + }, + { + "epoch": 2.3, + "learning_rate": 1.164721723518851e-05, + "loss": 0.245, + "step": 59865 + }, + { + "epoch": 2.3, + "learning_rate": 1.164657604513978e-05, + "loss": 0.1868, + "step": 59866 + }, + { + "epoch": 2.3, + "learning_rate": 1.1645934855091049e-05, + "loss": 0.3012, + "step": 59867 + }, + { + "epoch": 2.3, + "learning_rate": 1.1645293665042319e-05, + "loss": 0.3327, + "step": 59868 + }, + { + "epoch": 2.3, + "learning_rate": 1.1644652474993589e-05, + "loss": 0.1876, + "step": 59869 + }, + { + "epoch": 2.3, + "learning_rate": 1.1644011284944859e-05, + "loss": 0.1832, + "step": 59870 + }, + { + "epoch": 2.3, + "learning_rate": 1.1643370094896127e-05, + "loss": 0.2078, + "step": 59871 + }, + { + "epoch": 2.3, + "learning_rate": 1.1642728904847397e-05, + "loss": 0.2238, + "step": 59872 + }, + { + "epoch": 2.3, + "learning_rate": 1.1642087714798666e-05, + "loss": 0.2672, + "step": 59873 + }, + { + "epoch": 2.3, + "learning_rate": 1.1641446524749937e-05, + "loss": 0.4089, + "step": 59874 + }, + { + "epoch": 2.3, + "learning_rate": 1.1640805334701206e-05, + "loss": 0.225, + "step": 59875 + }, + { + "epoch": 2.3, + "learning_rate": 1.1640164144652476e-05, + "loss": 0.1954, + "step": 59876 + }, + { + "epoch": 2.3, + "learning_rate": 1.1639522954603744e-05, + "loss": 0.2226, + "step": 59877 + }, + { + "epoch": 2.3, + "learning_rate": 1.1638881764555016e-05, + "loss": 0.2167, + "step": 59878 + }, + { + "epoch": 2.3, + "learning_rate": 1.1638240574506284e-05, + "loss": 0.1931, + "step": 59879 + }, + { + "epoch": 2.3, + "learning_rate": 1.1637599384457554e-05, + "loss": 0.1929, + "step": 59880 + }, + { + "epoch": 2.3, + "learning_rate": 1.1636958194408822e-05, + "loss": 0.27, + "step": 59881 + }, + { + "epoch": 2.3, + "learning_rate": 1.1636317004360092e-05, + "loss": 0.1782, + "step": 59882 + }, + { + "epoch": 2.3, + "learning_rate": 1.1635675814311363e-05, + "loss": 0.2024, + "step": 59883 + }, + { + "epoch": 2.3, + "learning_rate": 1.1635034624262633e-05, + "loss": 0.2273, + "step": 59884 + }, + { + "epoch": 2.3, + "learning_rate": 1.1634393434213901e-05, + "loss": 0.1995, + "step": 59885 + }, + { + "epoch": 2.3, + "learning_rate": 1.1633752244165171e-05, + "loss": 0.302, + "step": 59886 + }, + { + "epoch": 2.3, + "learning_rate": 1.1633111054116441e-05, + "loss": 0.2763, + "step": 59887 + }, + { + "epoch": 2.3, + "learning_rate": 1.1632469864067711e-05, + "loss": 0.1571, + "step": 59888 + }, + { + "epoch": 2.3, + "learning_rate": 1.163182867401898e-05, + "loss": 0.2745, + "step": 59889 + }, + { + "epoch": 2.3, + "learning_rate": 1.163118748397025e-05, + "loss": 0.2394, + "step": 59890 + }, + { + "epoch": 2.3, + "learning_rate": 1.163054629392152e-05, + "loss": 0.3147, + "step": 59891 + }, + { + "epoch": 2.3, + "learning_rate": 1.162990510387279e-05, + "loss": 0.1959, + "step": 59892 + }, + { + "epoch": 2.3, + "learning_rate": 1.1629263913824058e-05, + "loss": 0.2416, + "step": 59893 + }, + { + "epoch": 2.3, + "learning_rate": 1.1628622723775328e-05, + "loss": 0.2239, + "step": 59894 + }, + { + "epoch": 2.3, + "learning_rate": 1.1627981533726596e-05, + "loss": 0.2591, + "step": 59895 + }, + { + "epoch": 2.3, + "learning_rate": 1.1627340343677868e-05, + "loss": 0.2673, + "step": 59896 + }, + { + "epoch": 2.3, + "learning_rate": 1.1626699153629136e-05, + "loss": 0.323, + "step": 59897 + }, + { + "epoch": 2.3, + "learning_rate": 1.1626057963580406e-05, + "loss": 0.3094, + "step": 59898 + }, + { + "epoch": 2.3, + "learning_rate": 1.1625416773531674e-05, + "loss": 0.2014, + "step": 59899 + }, + { + "epoch": 2.3, + "learning_rate": 1.1624775583482946e-05, + "loss": 0.155, + "step": 59900 + }, + { + "epoch": 2.3, + "learning_rate": 1.1624134393434214e-05, + "loss": 0.2016, + "step": 59901 + }, + { + "epoch": 2.3, + "learning_rate": 1.1623493203385484e-05, + "loss": 0.1844, + "step": 59902 + }, + { + "epoch": 2.3, + "learning_rate": 1.1622852013336753e-05, + "loss": 0.239, + "step": 59903 + }, + { + "epoch": 2.3, + "learning_rate": 1.1622210823288023e-05, + "loss": 0.2207, + "step": 59904 + }, + { + "epoch": 2.3, + "learning_rate": 1.1621569633239293e-05, + "loss": 0.3473, + "step": 59905 + }, + { + "epoch": 2.3, + "learning_rate": 1.1620928443190563e-05, + "loss": 0.1692, + "step": 59906 + }, + { + "epoch": 2.3, + "learning_rate": 1.1620287253141831e-05, + "loss": 0.1969, + "step": 59907 + }, + { + "epoch": 2.3, + "learning_rate": 1.1619646063093101e-05, + "loss": 0.2533, + "step": 59908 + }, + { + "epoch": 2.3, + "learning_rate": 1.1619004873044371e-05, + "loss": 0.2077, + "step": 59909 + }, + { + "epoch": 2.3, + "learning_rate": 1.1618363682995641e-05, + "loss": 0.2167, + "step": 59910 + }, + { + "epoch": 2.3, + "learning_rate": 1.161772249294691e-05, + "loss": 0.2391, + "step": 59911 + }, + { + "epoch": 2.3, + "learning_rate": 1.161708130289818e-05, + "loss": 0.1203, + "step": 59912 + }, + { + "epoch": 2.3, + "learning_rate": 1.1616440112849448e-05, + "loss": 0.1769, + "step": 59913 + }, + { + "epoch": 2.3, + "learning_rate": 1.161579892280072e-05, + "loss": 0.2093, + "step": 59914 + }, + { + "epoch": 2.3, + "learning_rate": 1.1615157732751988e-05, + "loss": 0.2942, + "step": 59915 + }, + { + "epoch": 2.3, + "learning_rate": 1.1614516542703258e-05, + "loss": 0.2923, + "step": 59916 + }, + { + "epoch": 2.3, + "learning_rate": 1.1613875352654526e-05, + "loss": 0.275, + "step": 59917 + }, + { + "epoch": 2.3, + "learning_rate": 1.1613234162605798e-05, + "loss": 0.2143, + "step": 59918 + }, + { + "epoch": 2.3, + "learning_rate": 1.1612592972557066e-05, + "loss": 0.3501, + "step": 59919 + }, + { + "epoch": 2.3, + "learning_rate": 1.1611951782508336e-05, + "loss": 0.2381, + "step": 59920 + }, + { + "epoch": 2.3, + "learning_rate": 1.1611310592459605e-05, + "loss": 0.2218, + "step": 59921 + }, + { + "epoch": 2.3, + "learning_rate": 1.1610669402410875e-05, + "loss": 0.2561, + "step": 59922 + }, + { + "epoch": 2.3, + "learning_rate": 1.1610028212362145e-05, + "loss": 0.2379, + "step": 59923 + }, + { + "epoch": 2.3, + "learning_rate": 1.1609387022313415e-05, + "loss": 0.2049, + "step": 59924 + }, + { + "epoch": 2.3, + "learning_rate": 1.1608745832264683e-05, + "loss": 0.1679, + "step": 59925 + }, + { + "epoch": 2.3, + "learning_rate": 1.1608104642215953e-05, + "loss": 0.2064, + "step": 59926 + }, + { + "epoch": 2.3, + "learning_rate": 1.1607463452167223e-05, + "loss": 0.1706, + "step": 59927 + }, + { + "epoch": 2.3, + "learning_rate": 1.1606822262118493e-05, + "loss": 0.1228, + "step": 59928 + }, + { + "epoch": 2.3, + "learning_rate": 1.1606181072069762e-05, + "loss": 0.218, + "step": 59929 + }, + { + "epoch": 2.3, + "learning_rate": 1.1605539882021032e-05, + "loss": 0.1516, + "step": 59930 + }, + { + "epoch": 2.3, + "learning_rate": 1.1604898691972302e-05, + "loss": 0.2369, + "step": 59931 + }, + { + "epoch": 2.3, + "learning_rate": 1.1604257501923572e-05, + "loss": 0.2138, + "step": 59932 + }, + { + "epoch": 2.3, + "learning_rate": 1.160361631187484e-05, + "loss": 0.1708, + "step": 59933 + }, + { + "epoch": 2.3, + "learning_rate": 1.160297512182611e-05, + "loss": 0.1878, + "step": 59934 + }, + { + "epoch": 2.3, + "learning_rate": 1.1602333931777378e-05, + "loss": 0.1285, + "step": 59935 + }, + { + "epoch": 2.3, + "learning_rate": 1.160169274172865e-05, + "loss": 0.2364, + "step": 59936 + }, + { + "epoch": 2.3, + "learning_rate": 1.1601051551679918e-05, + "loss": 0.3826, + "step": 59937 + }, + { + "epoch": 2.3, + "learning_rate": 1.1600410361631188e-05, + "loss": 0.3184, + "step": 59938 + }, + { + "epoch": 2.3, + "learning_rate": 1.1599769171582457e-05, + "loss": 0.1951, + "step": 59939 + }, + { + "epoch": 2.3, + "learning_rate": 1.1599127981533728e-05, + "loss": 0.2047, + "step": 59940 + }, + { + "epoch": 2.3, + "learning_rate": 1.1598486791484997e-05, + "loss": 0.2142, + "step": 59941 + }, + { + "epoch": 2.3, + "learning_rate": 1.1597845601436267e-05, + "loss": 0.3005, + "step": 59942 + }, + { + "epoch": 2.3, + "learning_rate": 1.1597204411387535e-05, + "loss": 0.23, + "step": 59943 + }, + { + "epoch": 2.3, + "learning_rate": 1.1596563221338805e-05, + "loss": 0.2841, + "step": 59944 + }, + { + "epoch": 2.3, + "learning_rate": 1.1595922031290075e-05, + "loss": 0.2355, + "step": 59945 + }, + { + "epoch": 2.3, + "learning_rate": 1.1595280841241345e-05, + "loss": 0.2089, + "step": 59946 + }, + { + "epoch": 2.3, + "learning_rate": 1.1594639651192613e-05, + "loss": 0.249, + "step": 59947 + }, + { + "epoch": 2.3, + "learning_rate": 1.1593998461143883e-05, + "loss": 0.3002, + "step": 59948 + }, + { + "epoch": 2.3, + "learning_rate": 1.1593357271095154e-05, + "loss": 0.2899, + "step": 59949 + }, + { + "epoch": 2.3, + "learning_rate": 1.1592716081046424e-05, + "loss": 0.3143, + "step": 59950 + }, + { + "epoch": 2.3, + "learning_rate": 1.1592074890997692e-05, + "loss": 0.2039, + "step": 59951 + }, + { + "epoch": 2.3, + "learning_rate": 1.1591433700948962e-05, + "loss": 0.2052, + "step": 59952 + }, + { + "epoch": 2.3, + "learning_rate": 1.159079251090023e-05, + "loss": 0.2113, + "step": 59953 + }, + { + "epoch": 2.31, + "learning_rate": 1.1590151320851502e-05, + "loss": 0.1005, + "step": 59954 + }, + { + "epoch": 2.31, + "learning_rate": 1.158951013080277e-05, + "loss": 0.2575, + "step": 59955 + }, + { + "epoch": 2.31, + "learning_rate": 1.158886894075404e-05, + "loss": 0.1329, + "step": 59956 + }, + { + "epoch": 2.31, + "learning_rate": 1.1588227750705309e-05, + "loss": 0.1794, + "step": 59957 + }, + { + "epoch": 2.31, + "learning_rate": 1.158758656065658e-05, + "loss": 0.1543, + "step": 59958 + }, + { + "epoch": 2.31, + "learning_rate": 1.1586945370607849e-05, + "loss": 0.2005, + "step": 59959 + }, + { + "epoch": 2.31, + "learning_rate": 1.1586304180559119e-05, + "loss": 0.2074, + "step": 59960 + }, + { + "epoch": 2.31, + "learning_rate": 1.1585662990510387e-05, + "loss": 0.3442, + "step": 59961 + }, + { + "epoch": 2.31, + "learning_rate": 1.1585021800461657e-05, + "loss": 0.2094, + "step": 59962 + }, + { + "epoch": 2.31, + "learning_rate": 1.1584380610412927e-05, + "loss": 0.2586, + "step": 59963 + }, + { + "epoch": 2.31, + "learning_rate": 1.1583739420364197e-05, + "loss": 0.1232, + "step": 59964 + }, + { + "epoch": 2.31, + "learning_rate": 1.1583098230315465e-05, + "loss": 0.2196, + "step": 59965 + }, + { + "epoch": 2.31, + "learning_rate": 1.1582457040266735e-05, + "loss": 0.3845, + "step": 59966 + }, + { + "epoch": 2.31, + "learning_rate": 1.1581815850218005e-05, + "loss": 0.216, + "step": 59967 + }, + { + "epoch": 2.31, + "learning_rate": 1.1581174660169275e-05, + "loss": 0.2154, + "step": 59968 + }, + { + "epoch": 2.31, + "learning_rate": 1.1580533470120544e-05, + "loss": 0.2307, + "step": 59969 + }, + { + "epoch": 2.31, + "learning_rate": 1.1579892280071814e-05, + "loss": 0.2389, + "step": 59970 + }, + { + "epoch": 2.31, + "learning_rate": 1.1579251090023082e-05, + "loss": 0.2828, + "step": 59971 + }, + { + "epoch": 2.31, + "learning_rate": 1.1578609899974354e-05, + "loss": 0.1656, + "step": 59972 + }, + { + "epoch": 2.31, + "learning_rate": 1.1577968709925622e-05, + "loss": 0.2448, + "step": 59973 + }, + { + "epoch": 2.31, + "learning_rate": 1.1577327519876892e-05, + "loss": 0.1788, + "step": 59974 + }, + { + "epoch": 2.31, + "learning_rate": 1.157668632982816e-05, + "loss": 0.2468, + "step": 59975 + }, + { + "epoch": 2.31, + "learning_rate": 1.1576045139779432e-05, + "loss": 0.2312, + "step": 59976 + }, + { + "epoch": 2.31, + "learning_rate": 1.15754039497307e-05, + "loss": 0.1385, + "step": 59977 + }, + { + "epoch": 2.31, + "learning_rate": 1.157476275968197e-05, + "loss": 0.271, + "step": 59978 + }, + { + "epoch": 2.31, + "learning_rate": 1.1574121569633239e-05, + "loss": 0.178, + "step": 59979 + }, + { + "epoch": 2.31, + "learning_rate": 1.157348037958451e-05, + "loss": 0.2706, + "step": 59980 + }, + { + "epoch": 2.31, + "learning_rate": 1.1572839189535779e-05, + "loss": 0.1695, + "step": 59981 + }, + { + "epoch": 2.31, + "learning_rate": 1.1572197999487049e-05, + "loss": 0.1532, + "step": 59982 + }, + { + "epoch": 2.31, + "learning_rate": 1.1571556809438317e-05, + "loss": 0.2441, + "step": 59983 + }, + { + "epoch": 2.31, + "learning_rate": 1.1570915619389587e-05, + "loss": 0.1074, + "step": 59984 + }, + { + "epoch": 2.31, + "learning_rate": 1.1570274429340857e-05, + "loss": 0.4412, + "step": 59985 + }, + { + "epoch": 2.31, + "learning_rate": 1.1569633239292127e-05, + "loss": 0.1919, + "step": 59986 + }, + { + "epoch": 2.31, + "learning_rate": 1.1568992049243396e-05, + "loss": 0.2124, + "step": 59987 + }, + { + "epoch": 2.31, + "learning_rate": 1.1568350859194666e-05, + "loss": 0.1537, + "step": 59988 + }, + { + "epoch": 2.31, + "learning_rate": 1.1567709669145936e-05, + "loss": 0.2255, + "step": 59989 + }, + { + "epoch": 2.31, + "learning_rate": 1.1567068479097206e-05, + "loss": 0.1607, + "step": 59990 + }, + { + "epoch": 2.31, + "learning_rate": 1.1566427289048474e-05, + "loss": 0.3327, + "step": 59991 + }, + { + "epoch": 2.31, + "learning_rate": 1.1565786098999744e-05, + "loss": 0.1463, + "step": 59992 + }, + { + "epoch": 2.31, + "learning_rate": 1.1565144908951012e-05, + "loss": 0.2877, + "step": 59993 + }, + { + "epoch": 2.31, + "learning_rate": 1.1564503718902284e-05, + "loss": 0.1254, + "step": 59994 + }, + { + "epoch": 2.31, + "learning_rate": 1.1563862528853553e-05, + "loss": 0.2697, + "step": 59995 + }, + { + "epoch": 2.31, + "learning_rate": 1.1563221338804823e-05, + "loss": 0.2936, + "step": 59996 + }, + { + "epoch": 2.31, + "learning_rate": 1.1562580148756091e-05, + "loss": 0.1789, + "step": 59997 + }, + { + "epoch": 2.31, + "learning_rate": 1.1561938958707363e-05, + "loss": 0.2948, + "step": 59998 + }, + { + "epoch": 2.31, + "learning_rate": 1.1561297768658631e-05, + "loss": 0.2597, + "step": 59999 + }, + { + "epoch": 2.31, + "learning_rate": 1.1560656578609901e-05, + "loss": 0.189, + "step": 60000 + }, + { + "epoch": 2.31, + "learning_rate": 1.156001538856117e-05, + "loss": 0.1691, + "step": 60001 + }, + { + "epoch": 2.31, + "learning_rate": 1.155937419851244e-05, + "loss": 0.1902, + "step": 60002 + }, + { + "epoch": 2.31, + "learning_rate": 1.155873300846371e-05, + "loss": 0.2971, + "step": 60003 + }, + { + "epoch": 2.31, + "learning_rate": 1.155809181841498e-05, + "loss": 0.1782, + "step": 60004 + }, + { + "epoch": 2.31, + "learning_rate": 1.1557450628366248e-05, + "loss": 0.2231, + "step": 60005 + }, + { + "epoch": 2.31, + "learning_rate": 1.1556809438317518e-05, + "loss": 0.1051, + "step": 60006 + }, + { + "epoch": 2.31, + "learning_rate": 1.1556168248268788e-05, + "loss": 0.1518, + "step": 60007 + }, + { + "epoch": 2.31, + "learning_rate": 1.1555527058220058e-05, + "loss": 0.3075, + "step": 60008 + }, + { + "epoch": 2.31, + "learning_rate": 1.1554885868171326e-05, + "loss": 0.1832, + "step": 60009 + }, + { + "epoch": 2.31, + "learning_rate": 1.1554244678122596e-05, + "loss": 0.2562, + "step": 60010 + }, + { + "epoch": 2.31, + "learning_rate": 1.1553603488073864e-05, + "loss": 0.2236, + "step": 60011 + }, + { + "epoch": 2.31, + "learning_rate": 1.1552962298025136e-05, + "loss": 0.2639, + "step": 60012 + }, + { + "epoch": 2.31, + "learning_rate": 1.1552321107976404e-05, + "loss": 0.2836, + "step": 60013 + }, + { + "epoch": 2.31, + "learning_rate": 1.1551679917927674e-05, + "loss": 0.3594, + "step": 60014 + }, + { + "epoch": 2.31, + "learning_rate": 1.1551038727878943e-05, + "loss": 0.2232, + "step": 60015 + }, + { + "epoch": 2.31, + "learning_rate": 1.1550397537830215e-05, + "loss": 0.4656, + "step": 60016 + }, + { + "epoch": 2.31, + "learning_rate": 1.1549756347781483e-05, + "loss": 0.2055, + "step": 60017 + }, + { + "epoch": 2.31, + "learning_rate": 1.1549115157732753e-05, + "loss": 0.3032, + "step": 60018 + }, + { + "epoch": 2.31, + "learning_rate": 1.1548473967684021e-05, + "loss": 0.2132, + "step": 60019 + }, + { + "epoch": 2.31, + "learning_rate": 1.1547832777635291e-05, + "loss": 0.3156, + "step": 60020 + }, + { + "epoch": 2.31, + "learning_rate": 1.1547191587586561e-05, + "loss": 0.2148, + "step": 60021 + }, + { + "epoch": 2.31, + "learning_rate": 1.1546550397537831e-05, + "loss": 0.1419, + "step": 60022 + }, + { + "epoch": 2.31, + "learning_rate": 1.15459092074891e-05, + "loss": 0.2512, + "step": 60023 + }, + { + "epoch": 2.31, + "learning_rate": 1.154526801744037e-05, + "loss": 0.3261, + "step": 60024 + }, + { + "epoch": 2.31, + "learning_rate": 1.154462682739164e-05, + "loss": 0.1869, + "step": 60025 + }, + { + "epoch": 2.31, + "learning_rate": 1.154398563734291e-05, + "loss": 0.2916, + "step": 60026 + }, + { + "epoch": 2.31, + "learning_rate": 1.1543344447294178e-05, + "loss": 0.3365, + "step": 60027 + }, + { + "epoch": 2.31, + "learning_rate": 1.1542703257245448e-05, + "loss": 0.2968, + "step": 60028 + }, + { + "epoch": 2.31, + "learning_rate": 1.1542062067196718e-05, + "loss": 0.2862, + "step": 60029 + }, + { + "epoch": 2.31, + "learning_rate": 1.1541420877147988e-05, + "loss": 0.2288, + "step": 60030 + }, + { + "epoch": 2.31, + "learning_rate": 1.1540779687099256e-05, + "loss": 0.1942, + "step": 60031 + }, + { + "epoch": 2.31, + "learning_rate": 1.1540138497050526e-05, + "loss": 0.2511, + "step": 60032 + }, + { + "epoch": 2.31, + "learning_rate": 1.1539497307001795e-05, + "loss": 0.2129, + "step": 60033 + }, + { + "epoch": 2.31, + "learning_rate": 1.1538856116953066e-05, + "loss": 0.2009, + "step": 60034 + }, + { + "epoch": 2.31, + "learning_rate": 1.1538214926904335e-05, + "loss": 0.1979, + "step": 60035 + }, + { + "epoch": 2.31, + "learning_rate": 1.1537573736855605e-05, + "loss": 0.1912, + "step": 60036 + }, + { + "epoch": 2.31, + "learning_rate": 1.1536932546806873e-05, + "loss": 0.27, + "step": 60037 + }, + { + "epoch": 2.31, + "learning_rate": 1.1536291356758145e-05, + "loss": 0.3462, + "step": 60038 + }, + { + "epoch": 2.31, + "learning_rate": 1.1535650166709413e-05, + "loss": 0.2713, + "step": 60039 + }, + { + "epoch": 2.31, + "learning_rate": 1.1535008976660683e-05, + "loss": 0.2069, + "step": 60040 + }, + { + "epoch": 2.31, + "learning_rate": 1.1534367786611952e-05, + "loss": 0.2381, + "step": 60041 + }, + { + "epoch": 2.31, + "learning_rate": 1.1533726596563222e-05, + "loss": 0.2571, + "step": 60042 + }, + { + "epoch": 2.31, + "learning_rate": 1.1533085406514492e-05, + "loss": 0.3277, + "step": 60043 + }, + { + "epoch": 2.31, + "learning_rate": 1.1532444216465762e-05, + "loss": 0.2813, + "step": 60044 + }, + { + "epoch": 2.31, + "learning_rate": 1.153180302641703e-05, + "loss": 0.2952, + "step": 60045 + }, + { + "epoch": 2.31, + "learning_rate": 1.15311618363683e-05, + "loss": 0.3393, + "step": 60046 + }, + { + "epoch": 2.31, + "learning_rate": 1.153052064631957e-05, + "loss": 0.269, + "step": 60047 + }, + { + "epoch": 2.31, + "learning_rate": 1.152987945627084e-05, + "loss": 0.1616, + "step": 60048 + }, + { + "epoch": 2.31, + "learning_rate": 1.1529238266222108e-05, + "loss": 0.1796, + "step": 60049 + }, + { + "epoch": 2.31, + "learning_rate": 1.1528597076173378e-05, + "loss": 0.1251, + "step": 60050 + }, + { + "epoch": 2.31, + "learning_rate": 1.1527955886124647e-05, + "loss": 0.2298, + "step": 60051 + }, + { + "epoch": 2.31, + "learning_rate": 1.1527314696075918e-05, + "loss": 0.3811, + "step": 60052 + }, + { + "epoch": 2.31, + "learning_rate": 1.1526673506027187e-05, + "loss": 0.2421, + "step": 60053 + }, + { + "epoch": 2.31, + "learning_rate": 1.1526032315978457e-05, + "loss": 0.2567, + "step": 60054 + }, + { + "epoch": 2.31, + "learning_rate": 1.1525391125929725e-05, + "loss": 0.2108, + "step": 60055 + }, + { + "epoch": 2.31, + "learning_rate": 1.1524749935880997e-05, + "loss": 0.2429, + "step": 60056 + }, + { + "epoch": 2.31, + "learning_rate": 1.1524108745832265e-05, + "loss": 0.2757, + "step": 60057 + }, + { + "epoch": 2.31, + "learning_rate": 1.1523467555783535e-05, + "loss": 0.1933, + "step": 60058 + }, + { + "epoch": 2.31, + "learning_rate": 1.1522826365734803e-05, + "loss": 0.2411, + "step": 60059 + }, + { + "epoch": 2.31, + "learning_rate": 1.1522185175686073e-05, + "loss": 0.2527, + "step": 60060 + }, + { + "epoch": 2.31, + "learning_rate": 1.1521543985637344e-05, + "loss": 0.2248, + "step": 60061 + }, + { + "epoch": 2.31, + "learning_rate": 1.1520902795588614e-05, + "loss": 0.1987, + "step": 60062 + }, + { + "epoch": 2.31, + "learning_rate": 1.1520261605539882e-05, + "loss": 0.259, + "step": 60063 + }, + { + "epoch": 2.31, + "learning_rate": 1.1519620415491152e-05, + "loss": 0.2012, + "step": 60064 + }, + { + "epoch": 2.31, + "learning_rate": 1.1518979225442422e-05, + "loss": 0.2377, + "step": 60065 + }, + { + "epoch": 2.31, + "learning_rate": 1.1518338035393692e-05, + "loss": 0.1296, + "step": 60066 + }, + { + "epoch": 2.31, + "learning_rate": 1.151769684534496e-05, + "loss": 0.2141, + "step": 60067 + }, + { + "epoch": 2.31, + "learning_rate": 1.151705565529623e-05, + "loss": 0.1688, + "step": 60068 + }, + { + "epoch": 2.31, + "learning_rate": 1.1516414465247499e-05, + "loss": 0.2064, + "step": 60069 + }, + { + "epoch": 2.31, + "learning_rate": 1.151577327519877e-05, + "loss": 0.3114, + "step": 60070 + }, + { + "epoch": 2.31, + "learning_rate": 1.1515132085150039e-05, + "loss": 0.167, + "step": 60071 + }, + { + "epoch": 2.31, + "learning_rate": 1.1514490895101309e-05, + "loss": 0.2195, + "step": 60072 + }, + { + "epoch": 2.31, + "learning_rate": 1.1513849705052577e-05, + "loss": 0.2335, + "step": 60073 + }, + { + "epoch": 2.31, + "learning_rate": 1.1513208515003849e-05, + "loss": 0.2678, + "step": 60074 + }, + { + "epoch": 2.31, + "learning_rate": 1.1512567324955117e-05, + "loss": 0.2308, + "step": 60075 + }, + { + "epoch": 2.31, + "learning_rate": 1.1511926134906387e-05, + "loss": 0.3321, + "step": 60076 + }, + { + "epoch": 2.31, + "learning_rate": 1.1511284944857655e-05, + "loss": 0.1193, + "step": 60077 + }, + { + "epoch": 2.31, + "learning_rate": 1.1510643754808927e-05, + "loss": 0.2174, + "step": 60078 + }, + { + "epoch": 2.31, + "learning_rate": 1.1510002564760195e-05, + "loss": 0.2036, + "step": 60079 + }, + { + "epoch": 2.31, + "learning_rate": 1.1509361374711465e-05, + "loss": 0.4135, + "step": 60080 + }, + { + "epoch": 2.31, + "learning_rate": 1.1508720184662734e-05, + "loss": 0.1782, + "step": 60081 + }, + { + "epoch": 2.31, + "learning_rate": 1.1508078994614004e-05, + "loss": 0.2463, + "step": 60082 + }, + { + "epoch": 2.31, + "learning_rate": 1.1507437804565274e-05, + "loss": 0.3341, + "step": 60083 + }, + { + "epoch": 2.31, + "learning_rate": 1.1506796614516544e-05, + "loss": 0.2837, + "step": 60084 + }, + { + "epoch": 2.31, + "learning_rate": 1.1506155424467812e-05, + "loss": 0.3392, + "step": 60085 + }, + { + "epoch": 2.31, + "learning_rate": 1.1505514234419082e-05, + "loss": 0.2876, + "step": 60086 + }, + { + "epoch": 2.31, + "learning_rate": 1.1504873044370352e-05, + "loss": 0.2872, + "step": 60087 + }, + { + "epoch": 2.31, + "learning_rate": 1.1504231854321622e-05, + "loss": 0.3228, + "step": 60088 + }, + { + "epoch": 2.31, + "learning_rate": 1.150359066427289e-05, + "loss": 0.1537, + "step": 60089 + }, + { + "epoch": 2.31, + "learning_rate": 1.150294947422416e-05, + "loss": 0.1329, + "step": 60090 + }, + { + "epoch": 2.31, + "learning_rate": 1.1502308284175429e-05, + "loss": 0.177, + "step": 60091 + }, + { + "epoch": 2.31, + "learning_rate": 1.15016670941267e-05, + "loss": 0.132, + "step": 60092 + }, + { + "epoch": 2.31, + "learning_rate": 1.1501025904077969e-05, + "loss": 0.3961, + "step": 60093 + }, + { + "epoch": 2.31, + "learning_rate": 1.1500384714029239e-05, + "loss": 0.3116, + "step": 60094 + }, + { + "epoch": 2.31, + "learning_rate": 1.1499743523980507e-05, + "loss": 0.2519, + "step": 60095 + }, + { + "epoch": 2.31, + "learning_rate": 1.1499102333931779e-05, + "loss": 0.1819, + "step": 60096 + }, + { + "epoch": 2.31, + "learning_rate": 1.1498461143883047e-05, + "loss": 0.1666, + "step": 60097 + }, + { + "epoch": 2.31, + "learning_rate": 1.1497819953834317e-05, + "loss": 0.2644, + "step": 60098 + }, + { + "epoch": 2.31, + "learning_rate": 1.1497178763785586e-05, + "loss": 0.2768, + "step": 60099 + }, + { + "epoch": 2.31, + "learning_rate": 1.1496537573736856e-05, + "loss": 0.178, + "step": 60100 + }, + { + "epoch": 2.31, + "learning_rate": 1.1495896383688126e-05, + "loss": 0.3512, + "step": 60101 + }, + { + "epoch": 2.31, + "learning_rate": 1.1495255193639396e-05, + "loss": 0.3881, + "step": 60102 + }, + { + "epoch": 2.31, + "learning_rate": 1.1494614003590664e-05, + "loss": 0.3405, + "step": 60103 + }, + { + "epoch": 2.31, + "learning_rate": 1.1493972813541934e-05, + "loss": 0.2447, + "step": 60104 + }, + { + "epoch": 2.31, + "learning_rate": 1.1493331623493204e-05, + "loss": 0.1707, + "step": 60105 + }, + { + "epoch": 2.31, + "learning_rate": 1.1492690433444474e-05, + "loss": 0.1346, + "step": 60106 + }, + { + "epoch": 2.31, + "learning_rate": 1.1492049243395743e-05, + "loss": 0.1728, + "step": 60107 + }, + { + "epoch": 2.31, + "learning_rate": 1.1491408053347013e-05, + "loss": 0.1943, + "step": 60108 + }, + { + "epoch": 2.31, + "learning_rate": 1.1490766863298281e-05, + "loss": 0.163, + "step": 60109 + }, + { + "epoch": 2.31, + "learning_rate": 1.1490125673249553e-05, + "loss": 0.2266, + "step": 60110 + }, + { + "epoch": 2.31, + "learning_rate": 1.1489484483200821e-05, + "loss": 0.2609, + "step": 60111 + }, + { + "epoch": 2.31, + "learning_rate": 1.1488843293152091e-05, + "loss": 0.136, + "step": 60112 + }, + { + "epoch": 2.31, + "learning_rate": 1.148820210310336e-05, + "loss": 0.2613, + "step": 60113 + }, + { + "epoch": 2.31, + "learning_rate": 1.1487560913054631e-05, + "loss": 0.1649, + "step": 60114 + }, + { + "epoch": 2.31, + "learning_rate": 1.14869197230059e-05, + "loss": 0.1568, + "step": 60115 + }, + { + "epoch": 2.31, + "learning_rate": 1.148627853295717e-05, + "loss": 0.301, + "step": 60116 + }, + { + "epoch": 2.31, + "learning_rate": 1.1485637342908438e-05, + "loss": 0.2412, + "step": 60117 + }, + { + "epoch": 2.31, + "learning_rate": 1.1484996152859708e-05, + "loss": 0.3049, + "step": 60118 + }, + { + "epoch": 2.31, + "learning_rate": 1.1484354962810978e-05, + "loss": 0.1658, + "step": 60119 + }, + { + "epoch": 2.31, + "learning_rate": 1.1483713772762248e-05, + "loss": 0.1834, + "step": 60120 + }, + { + "epoch": 2.31, + "learning_rate": 1.1483072582713516e-05, + "loss": 0.2032, + "step": 60121 + }, + { + "epoch": 2.31, + "learning_rate": 1.1482431392664786e-05, + "loss": 0.1332, + "step": 60122 + }, + { + "epoch": 2.31, + "learning_rate": 1.1481790202616056e-05, + "loss": 0.2725, + "step": 60123 + }, + { + "epoch": 2.31, + "learning_rate": 1.1481149012567326e-05, + "loss": 0.2372, + "step": 60124 + }, + { + "epoch": 2.31, + "learning_rate": 1.1480507822518594e-05, + "loss": 0.1911, + "step": 60125 + }, + { + "epoch": 2.31, + "learning_rate": 1.1479866632469864e-05, + "loss": 0.1873, + "step": 60126 + }, + { + "epoch": 2.31, + "learning_rate": 1.1479225442421135e-05, + "loss": 0.3916, + "step": 60127 + }, + { + "epoch": 2.31, + "learning_rate": 1.1478584252372405e-05, + "loss": 0.2896, + "step": 60128 + }, + { + "epoch": 2.31, + "learning_rate": 1.1477943062323673e-05, + "loss": 0.2355, + "step": 60129 + }, + { + "epoch": 2.31, + "learning_rate": 1.1477301872274943e-05, + "loss": 0.2537, + "step": 60130 + }, + { + "epoch": 2.31, + "learning_rate": 1.1476660682226211e-05, + "loss": 0.1982, + "step": 60131 + }, + { + "epoch": 2.31, + "learning_rate": 1.1476019492177483e-05, + "loss": 0.2609, + "step": 60132 + }, + { + "epoch": 2.31, + "learning_rate": 1.1475378302128751e-05, + "loss": 0.226, + "step": 60133 + }, + { + "epoch": 2.31, + "learning_rate": 1.1474737112080021e-05, + "loss": 0.3007, + "step": 60134 + }, + { + "epoch": 2.31, + "learning_rate": 1.147409592203129e-05, + "loss": 0.3097, + "step": 60135 + }, + { + "epoch": 2.31, + "learning_rate": 1.1473454731982561e-05, + "loss": 0.3481, + "step": 60136 + }, + { + "epoch": 2.31, + "learning_rate": 1.147281354193383e-05, + "loss": 0.2436, + "step": 60137 + }, + { + "epoch": 2.31, + "learning_rate": 1.14721723518851e-05, + "loss": 0.1854, + "step": 60138 + }, + { + "epoch": 2.31, + "learning_rate": 1.1471531161836368e-05, + "loss": 0.3669, + "step": 60139 + }, + { + "epoch": 2.31, + "learning_rate": 1.1470889971787638e-05, + "loss": 0.1284, + "step": 60140 + }, + { + "epoch": 2.31, + "learning_rate": 1.1470248781738908e-05, + "loss": 0.2316, + "step": 60141 + }, + { + "epoch": 2.31, + "learning_rate": 1.1469607591690178e-05, + "loss": 0.2465, + "step": 60142 + }, + { + "epoch": 2.31, + "learning_rate": 1.1468966401641446e-05, + "loss": 0.2301, + "step": 60143 + }, + { + "epoch": 2.31, + "learning_rate": 1.1468325211592716e-05, + "loss": 0.1786, + "step": 60144 + }, + { + "epoch": 2.31, + "learning_rate": 1.1467684021543986e-05, + "loss": 0.3105, + "step": 60145 + }, + { + "epoch": 2.31, + "learning_rate": 1.1467042831495256e-05, + "loss": 0.2211, + "step": 60146 + }, + { + "epoch": 2.31, + "learning_rate": 1.1466401641446525e-05, + "loss": 0.1605, + "step": 60147 + }, + { + "epoch": 2.31, + "learning_rate": 1.1465760451397795e-05, + "loss": 0.2425, + "step": 60148 + }, + { + "epoch": 2.31, + "learning_rate": 1.1465119261349063e-05, + "loss": 0.2355, + "step": 60149 + }, + { + "epoch": 2.31, + "learning_rate": 1.1464478071300335e-05, + "loss": 0.3154, + "step": 60150 + }, + { + "epoch": 2.31, + "learning_rate": 1.1463836881251603e-05, + "loss": 0.238, + "step": 60151 + }, + { + "epoch": 2.31, + "learning_rate": 1.1463195691202873e-05, + "loss": 0.3736, + "step": 60152 + }, + { + "epoch": 2.31, + "learning_rate": 1.1462554501154142e-05, + "loss": 0.3369, + "step": 60153 + }, + { + "epoch": 2.31, + "learning_rate": 1.1461913311105413e-05, + "loss": 0.1401, + "step": 60154 + }, + { + "epoch": 2.31, + "learning_rate": 1.1461272121056682e-05, + "loss": 0.2515, + "step": 60155 + }, + { + "epoch": 2.31, + "learning_rate": 1.1460630931007952e-05, + "loss": 0.157, + "step": 60156 + }, + { + "epoch": 2.31, + "learning_rate": 1.145998974095922e-05, + "loss": 0.2374, + "step": 60157 + }, + { + "epoch": 2.31, + "learning_rate": 1.145934855091049e-05, + "loss": 0.3421, + "step": 60158 + }, + { + "epoch": 2.31, + "learning_rate": 1.145870736086176e-05, + "loss": 0.1746, + "step": 60159 + }, + { + "epoch": 2.31, + "learning_rate": 1.145806617081303e-05, + "loss": 0.2698, + "step": 60160 + }, + { + "epoch": 2.31, + "learning_rate": 1.1457424980764298e-05, + "loss": 0.2218, + "step": 60161 + }, + { + "epoch": 2.31, + "learning_rate": 1.1456783790715568e-05, + "loss": 0.2888, + "step": 60162 + }, + { + "epoch": 2.31, + "learning_rate": 1.1456142600666838e-05, + "loss": 0.2544, + "step": 60163 + }, + { + "epoch": 2.31, + "learning_rate": 1.1455501410618108e-05, + "loss": 0.4133, + "step": 60164 + }, + { + "epoch": 2.31, + "learning_rate": 1.1454860220569377e-05, + "loss": 0.1334, + "step": 60165 + }, + { + "epoch": 2.31, + "learning_rate": 1.1454219030520647e-05, + "loss": 0.271, + "step": 60166 + }, + { + "epoch": 2.31, + "learning_rate": 1.1453577840471915e-05, + "loss": 0.1373, + "step": 60167 + }, + { + "epoch": 2.31, + "learning_rate": 1.1452936650423187e-05, + "loss": 0.3274, + "step": 60168 + }, + { + "epoch": 2.31, + "learning_rate": 1.1452295460374455e-05, + "loss": 0.2646, + "step": 60169 + }, + { + "epoch": 2.31, + "learning_rate": 1.1451654270325725e-05, + "loss": 0.1485, + "step": 60170 + }, + { + "epoch": 2.31, + "learning_rate": 1.1451013080276993e-05, + "loss": 0.3011, + "step": 60171 + }, + { + "epoch": 2.31, + "learning_rate": 1.1450371890228265e-05, + "loss": 0.2685, + "step": 60172 + }, + { + "epoch": 2.31, + "learning_rate": 1.1449730700179534e-05, + "loss": 0.162, + "step": 60173 + }, + { + "epoch": 2.31, + "learning_rate": 1.1449089510130804e-05, + "loss": 0.2596, + "step": 60174 + }, + { + "epoch": 2.31, + "learning_rate": 1.1448448320082072e-05, + "loss": 0.1648, + "step": 60175 + }, + { + "epoch": 2.31, + "learning_rate": 1.1447807130033344e-05, + "loss": 0.2546, + "step": 60176 + }, + { + "epoch": 2.31, + "learning_rate": 1.1447165939984612e-05, + "loss": 0.2438, + "step": 60177 + }, + { + "epoch": 2.31, + "learning_rate": 1.1446524749935882e-05, + "loss": 0.1998, + "step": 60178 + }, + { + "epoch": 2.31, + "learning_rate": 1.144588355988715e-05, + "loss": 0.2081, + "step": 60179 + }, + { + "epoch": 2.31, + "learning_rate": 1.144524236983842e-05, + "loss": 0.1986, + "step": 60180 + }, + { + "epoch": 2.31, + "learning_rate": 1.144460117978969e-05, + "loss": 0.3064, + "step": 60181 + }, + { + "epoch": 2.31, + "learning_rate": 1.144395998974096e-05, + "loss": 0.2202, + "step": 60182 + }, + { + "epoch": 2.31, + "learning_rate": 1.1443318799692229e-05, + "loss": 0.2648, + "step": 60183 + }, + { + "epoch": 2.31, + "learning_rate": 1.1442677609643499e-05, + "loss": 0.23, + "step": 60184 + }, + { + "epoch": 2.31, + "learning_rate": 1.1442036419594769e-05, + "loss": 0.1694, + "step": 60185 + }, + { + "epoch": 2.31, + "learning_rate": 1.1441395229546039e-05, + "loss": 0.2208, + "step": 60186 + }, + { + "epoch": 2.31, + "learning_rate": 1.1440754039497307e-05, + "loss": 0.2396, + "step": 60187 + }, + { + "epoch": 2.31, + "learning_rate": 1.1440112849448577e-05, + "loss": 0.2148, + "step": 60188 + }, + { + "epoch": 2.31, + "learning_rate": 1.1439471659399845e-05, + "loss": 0.1919, + "step": 60189 + }, + { + "epoch": 2.31, + "learning_rate": 1.1438830469351117e-05, + "loss": 0.1598, + "step": 60190 + }, + { + "epoch": 2.31, + "learning_rate": 1.1438189279302385e-05, + "loss": 0.2834, + "step": 60191 + }, + { + "epoch": 2.31, + "learning_rate": 1.1437548089253655e-05, + "loss": 0.1655, + "step": 60192 + }, + { + "epoch": 2.31, + "learning_rate": 1.1436906899204924e-05, + "loss": 0.125, + "step": 60193 + }, + { + "epoch": 2.31, + "learning_rate": 1.1436265709156196e-05, + "loss": 0.1726, + "step": 60194 + }, + { + "epoch": 2.31, + "learning_rate": 1.1435624519107464e-05, + "loss": 0.2873, + "step": 60195 + }, + { + "epoch": 2.31, + "learning_rate": 1.1434983329058734e-05, + "loss": 0.2624, + "step": 60196 + }, + { + "epoch": 2.31, + "learning_rate": 1.1434342139010002e-05, + "loss": 0.231, + "step": 60197 + }, + { + "epoch": 2.31, + "learning_rate": 1.1433700948961272e-05, + "loss": 0.267, + "step": 60198 + }, + { + "epoch": 2.31, + "learning_rate": 1.1433059758912542e-05, + "loss": 0.203, + "step": 60199 + }, + { + "epoch": 2.31, + "learning_rate": 1.1432418568863812e-05, + "loss": 0.2289, + "step": 60200 + }, + { + "epoch": 2.31, + "learning_rate": 1.143177737881508e-05, + "loss": 0.2842, + "step": 60201 + }, + { + "epoch": 2.31, + "learning_rate": 1.143113618876635e-05, + "loss": 0.2237, + "step": 60202 + }, + { + "epoch": 2.31, + "learning_rate": 1.143049499871762e-05, + "loss": 0.1835, + "step": 60203 + }, + { + "epoch": 2.31, + "learning_rate": 1.142985380866889e-05, + "loss": 0.1403, + "step": 60204 + }, + { + "epoch": 2.31, + "learning_rate": 1.1429212618620159e-05, + "loss": 0.2735, + "step": 60205 + }, + { + "epoch": 2.31, + "learning_rate": 1.1428571428571429e-05, + "loss": 0.2325, + "step": 60206 + }, + { + "epoch": 2.31, + "learning_rate": 1.1427930238522697e-05, + "loss": 0.2097, + "step": 60207 + }, + { + "epoch": 2.31, + "learning_rate": 1.1427289048473969e-05, + "loss": 0.1901, + "step": 60208 + }, + { + "epoch": 2.31, + "learning_rate": 1.1426647858425237e-05, + "loss": 0.2432, + "step": 60209 + }, + { + "epoch": 2.31, + "learning_rate": 1.1426006668376507e-05, + "loss": 0.2544, + "step": 60210 + }, + { + "epoch": 2.31, + "learning_rate": 1.1425365478327776e-05, + "loss": 0.3321, + "step": 60211 + }, + { + "epoch": 2.31, + "learning_rate": 1.1424724288279047e-05, + "loss": 0.2188, + "step": 60212 + }, + { + "epoch": 2.31, + "learning_rate": 1.1424083098230316e-05, + "loss": 0.37, + "step": 60213 + }, + { + "epoch": 2.32, + "learning_rate": 1.1423441908181586e-05, + "loss": 0.1678, + "step": 60214 + }, + { + "epoch": 2.32, + "learning_rate": 1.1422800718132854e-05, + "loss": 0.1743, + "step": 60215 + }, + { + "epoch": 2.32, + "learning_rate": 1.1422159528084126e-05, + "loss": 0.2809, + "step": 60216 + }, + { + "epoch": 2.32, + "learning_rate": 1.1421518338035394e-05, + "loss": 0.1999, + "step": 60217 + }, + { + "epoch": 2.32, + "learning_rate": 1.1420877147986664e-05, + "loss": 0.1502, + "step": 60218 + }, + { + "epoch": 2.32, + "learning_rate": 1.1420235957937933e-05, + "loss": 0.375, + "step": 60219 + }, + { + "epoch": 2.32, + "learning_rate": 1.1419594767889203e-05, + "loss": 0.2694, + "step": 60220 + }, + { + "epoch": 2.32, + "learning_rate": 1.1418953577840473e-05, + "loss": 0.2419, + "step": 60221 + }, + { + "epoch": 2.32, + "learning_rate": 1.1418312387791743e-05, + "loss": 0.1982, + "step": 60222 + }, + { + "epoch": 2.32, + "learning_rate": 1.1417671197743011e-05, + "loss": 0.2712, + "step": 60223 + }, + { + "epoch": 2.32, + "learning_rate": 1.1417030007694281e-05, + "loss": 0.1212, + "step": 60224 + }, + { + "epoch": 2.32, + "learning_rate": 1.1416388817645551e-05, + "loss": 0.3113, + "step": 60225 + }, + { + "epoch": 2.32, + "learning_rate": 1.1415747627596821e-05, + "loss": 0.1319, + "step": 60226 + }, + { + "epoch": 2.32, + "learning_rate": 1.141510643754809e-05, + "loss": 0.27, + "step": 60227 + }, + { + "epoch": 2.32, + "learning_rate": 1.141446524749936e-05, + "loss": 0.3844, + "step": 60228 + }, + { + "epoch": 2.32, + "learning_rate": 1.1413824057450628e-05, + "loss": 0.1704, + "step": 60229 + }, + { + "epoch": 2.32, + "learning_rate": 1.14131828674019e-05, + "loss": 0.1784, + "step": 60230 + }, + { + "epoch": 2.32, + "learning_rate": 1.1412541677353168e-05, + "loss": 0.4137, + "step": 60231 + }, + { + "epoch": 2.32, + "learning_rate": 1.1411900487304438e-05, + "loss": 0.2696, + "step": 60232 + }, + { + "epoch": 2.32, + "learning_rate": 1.1411259297255706e-05, + "loss": 0.1776, + "step": 60233 + }, + { + "epoch": 2.32, + "learning_rate": 1.1410618107206978e-05, + "loss": 0.2681, + "step": 60234 + }, + { + "epoch": 2.32, + "learning_rate": 1.1409976917158246e-05, + "loss": 0.2161, + "step": 60235 + }, + { + "epoch": 2.32, + "learning_rate": 1.1409335727109516e-05, + "loss": 0.2436, + "step": 60236 + }, + { + "epoch": 2.32, + "learning_rate": 1.1408694537060784e-05, + "loss": 0.2251, + "step": 60237 + }, + { + "epoch": 2.32, + "learning_rate": 1.1408053347012054e-05, + "loss": 0.17, + "step": 60238 + }, + { + "epoch": 2.32, + "learning_rate": 1.1407412156963325e-05, + "loss": 0.1711, + "step": 60239 + }, + { + "epoch": 2.32, + "learning_rate": 1.1406770966914595e-05, + "loss": 0.228, + "step": 60240 + }, + { + "epoch": 2.32, + "learning_rate": 1.1406129776865863e-05, + "loss": 0.1969, + "step": 60241 + }, + { + "epoch": 2.32, + "learning_rate": 1.1405488586817133e-05, + "loss": 0.2799, + "step": 60242 + }, + { + "epoch": 2.32, + "learning_rate": 1.1404847396768403e-05, + "loss": 0.1016, + "step": 60243 + }, + { + "epoch": 2.32, + "learning_rate": 1.1404206206719673e-05, + "loss": 0.1751, + "step": 60244 + }, + { + "epoch": 2.32, + "learning_rate": 1.1403565016670941e-05, + "loss": 0.2078, + "step": 60245 + }, + { + "epoch": 2.32, + "learning_rate": 1.1402923826622211e-05, + "loss": 0.2992, + "step": 60246 + }, + { + "epoch": 2.32, + "learning_rate": 1.140228263657348e-05, + "loss": 0.236, + "step": 60247 + }, + { + "epoch": 2.32, + "learning_rate": 1.1401641446524751e-05, + "loss": 0.2093, + "step": 60248 + }, + { + "epoch": 2.32, + "learning_rate": 1.140100025647602e-05, + "loss": 0.2701, + "step": 60249 + }, + { + "epoch": 2.32, + "learning_rate": 1.140035906642729e-05, + "loss": 0.3077, + "step": 60250 + }, + { + "epoch": 2.32, + "learning_rate": 1.1399717876378558e-05, + "loss": 0.3109, + "step": 60251 + }, + { + "epoch": 2.32, + "learning_rate": 1.139907668632983e-05, + "loss": 0.2768, + "step": 60252 + }, + { + "epoch": 2.32, + "learning_rate": 1.1398435496281098e-05, + "loss": 0.2396, + "step": 60253 + }, + { + "epoch": 2.32, + "learning_rate": 1.1397794306232368e-05, + "loss": 0.2179, + "step": 60254 + }, + { + "epoch": 2.32, + "learning_rate": 1.1397153116183636e-05, + "loss": 0.1816, + "step": 60255 + }, + { + "epoch": 2.32, + "learning_rate": 1.1396511926134906e-05, + "loss": 0.3362, + "step": 60256 + }, + { + "epoch": 2.32, + "learning_rate": 1.1395870736086176e-05, + "loss": 0.1965, + "step": 60257 + }, + { + "epoch": 2.32, + "learning_rate": 1.1395229546037446e-05, + "loss": 0.2684, + "step": 60258 + }, + { + "epoch": 2.32, + "learning_rate": 1.1394588355988715e-05, + "loss": 0.2106, + "step": 60259 + }, + { + "epoch": 2.32, + "learning_rate": 1.1393947165939985e-05, + "loss": 0.2493, + "step": 60260 + }, + { + "epoch": 2.32, + "learning_rate": 1.1393305975891255e-05, + "loss": 0.2692, + "step": 60261 + }, + { + "epoch": 2.32, + "learning_rate": 1.1392664785842525e-05, + "loss": 0.1552, + "step": 60262 + }, + { + "epoch": 2.32, + "learning_rate": 1.1392023595793793e-05, + "loss": 0.151, + "step": 60263 + }, + { + "epoch": 2.32, + "learning_rate": 1.1391382405745063e-05, + "loss": 0.1715, + "step": 60264 + }, + { + "epoch": 2.32, + "learning_rate": 1.1390741215696333e-05, + "loss": 0.2696, + "step": 60265 + }, + { + "epoch": 2.32, + "learning_rate": 1.1390100025647603e-05, + "loss": 0.1578, + "step": 60266 + }, + { + "epoch": 2.32, + "learning_rate": 1.1389458835598872e-05, + "loss": 0.2661, + "step": 60267 + }, + { + "epoch": 2.32, + "learning_rate": 1.1388817645550142e-05, + "loss": 0.1968, + "step": 60268 + }, + { + "epoch": 2.32, + "learning_rate": 1.138817645550141e-05, + "loss": 0.3196, + "step": 60269 + }, + { + "epoch": 2.32, + "learning_rate": 1.1387535265452682e-05, + "loss": 0.0963, + "step": 60270 + }, + { + "epoch": 2.32, + "learning_rate": 1.138689407540395e-05, + "loss": 0.1815, + "step": 60271 + }, + { + "epoch": 2.32, + "learning_rate": 1.138625288535522e-05, + "loss": 0.3259, + "step": 60272 + }, + { + "epoch": 2.32, + "learning_rate": 1.1385611695306488e-05, + "loss": 0.1914, + "step": 60273 + }, + { + "epoch": 2.32, + "learning_rate": 1.138497050525776e-05, + "loss": 0.265, + "step": 60274 + }, + { + "epoch": 2.32, + "learning_rate": 1.1384329315209028e-05, + "loss": 0.2026, + "step": 60275 + }, + { + "epoch": 2.32, + "learning_rate": 1.1383688125160298e-05, + "loss": 0.4158, + "step": 60276 + }, + { + "epoch": 2.32, + "learning_rate": 1.1383046935111567e-05, + "loss": 0.3357, + "step": 60277 + }, + { + "epoch": 2.32, + "learning_rate": 1.1382405745062837e-05, + "loss": 0.1862, + "step": 60278 + }, + { + "epoch": 2.32, + "learning_rate": 1.1381764555014107e-05, + "loss": 0.1931, + "step": 60279 + }, + { + "epoch": 2.32, + "learning_rate": 1.1381123364965377e-05, + "loss": 0.2062, + "step": 60280 + }, + { + "epoch": 2.32, + "learning_rate": 1.1380482174916645e-05, + "loss": 0.2552, + "step": 60281 + }, + { + "epoch": 2.32, + "learning_rate": 1.1379840984867915e-05, + "loss": 0.2485, + "step": 60282 + }, + { + "epoch": 2.32, + "learning_rate": 1.1379199794819185e-05, + "loss": 0.465, + "step": 60283 + }, + { + "epoch": 2.32, + "learning_rate": 1.1378558604770455e-05, + "loss": 0.0982, + "step": 60284 + }, + { + "epoch": 2.32, + "learning_rate": 1.1377917414721724e-05, + "loss": 0.1749, + "step": 60285 + }, + { + "epoch": 2.32, + "learning_rate": 1.1377276224672994e-05, + "loss": 0.2228, + "step": 60286 + }, + { + "epoch": 2.32, + "learning_rate": 1.1376635034624262e-05, + "loss": 0.0963, + "step": 60287 + }, + { + "epoch": 2.32, + "learning_rate": 1.1375993844575534e-05, + "loss": 0.239, + "step": 60288 + }, + { + "epoch": 2.32, + "learning_rate": 1.1375352654526802e-05, + "loss": 0.1493, + "step": 60289 + }, + { + "epoch": 2.32, + "learning_rate": 1.1374711464478072e-05, + "loss": 0.2816, + "step": 60290 + }, + { + "epoch": 2.32, + "learning_rate": 1.137407027442934e-05, + "loss": 0.1751, + "step": 60291 + }, + { + "epoch": 2.32, + "learning_rate": 1.1373429084380612e-05, + "loss": 0.2425, + "step": 60292 + }, + { + "epoch": 2.32, + "learning_rate": 1.137278789433188e-05, + "loss": 0.2174, + "step": 60293 + }, + { + "epoch": 2.32, + "learning_rate": 1.137214670428315e-05, + "loss": 0.1624, + "step": 60294 + }, + { + "epoch": 2.32, + "learning_rate": 1.1371505514234419e-05, + "loss": 0.215, + "step": 60295 + }, + { + "epoch": 2.32, + "learning_rate": 1.1370864324185689e-05, + "loss": 0.1489, + "step": 60296 + }, + { + "epoch": 2.32, + "learning_rate": 1.1370223134136959e-05, + "loss": 0.2459, + "step": 60297 + }, + { + "epoch": 2.32, + "learning_rate": 1.1369581944088229e-05, + "loss": 0.3505, + "step": 60298 + }, + { + "epoch": 2.32, + "learning_rate": 1.1368940754039497e-05, + "loss": 0.262, + "step": 60299 + }, + { + "epoch": 2.32, + "learning_rate": 1.1368299563990767e-05, + "loss": 0.2793, + "step": 60300 + }, + { + "epoch": 2.32, + "learning_rate": 1.1367658373942037e-05, + "loss": 0.2166, + "step": 60301 + }, + { + "epoch": 2.32, + "learning_rate": 1.1367017183893307e-05, + "loss": 0.2566, + "step": 60302 + }, + { + "epoch": 2.32, + "learning_rate": 1.1366375993844575e-05, + "loss": 0.15, + "step": 60303 + }, + { + "epoch": 2.32, + "learning_rate": 1.1365734803795845e-05, + "loss": 0.1716, + "step": 60304 + }, + { + "epoch": 2.32, + "learning_rate": 1.1365093613747114e-05, + "loss": 0.1833, + "step": 60305 + }, + { + "epoch": 2.32, + "learning_rate": 1.1364452423698386e-05, + "loss": 0.1894, + "step": 60306 + }, + { + "epoch": 2.32, + "learning_rate": 1.1363811233649654e-05, + "loss": 0.3149, + "step": 60307 + }, + { + "epoch": 2.32, + "learning_rate": 1.1363170043600924e-05, + "loss": 0.2496, + "step": 60308 + }, + { + "epoch": 2.32, + "learning_rate": 1.1362528853552192e-05, + "loss": 0.1181, + "step": 60309 + }, + { + "epoch": 2.32, + "learning_rate": 1.1361887663503464e-05, + "loss": 0.1677, + "step": 60310 + }, + { + "epoch": 2.32, + "learning_rate": 1.1361246473454732e-05, + "loss": 0.1778, + "step": 60311 + }, + { + "epoch": 2.32, + "learning_rate": 1.1360605283406002e-05, + "loss": 0.2807, + "step": 60312 + }, + { + "epoch": 2.32, + "learning_rate": 1.135996409335727e-05, + "loss": 0.3114, + "step": 60313 + }, + { + "epoch": 2.32, + "learning_rate": 1.1359322903308542e-05, + "loss": 0.2053, + "step": 60314 + }, + { + "epoch": 2.32, + "learning_rate": 1.135868171325981e-05, + "loss": 0.2226, + "step": 60315 + }, + { + "epoch": 2.32, + "learning_rate": 1.135804052321108e-05, + "loss": 0.2183, + "step": 60316 + }, + { + "epoch": 2.32, + "learning_rate": 1.1357399333162349e-05, + "loss": 0.2042, + "step": 60317 + }, + { + "epoch": 2.32, + "learning_rate": 1.1356758143113619e-05, + "loss": 0.3091, + "step": 60318 + }, + { + "epoch": 2.32, + "learning_rate": 1.1356116953064889e-05, + "loss": 0.2317, + "step": 60319 + }, + { + "epoch": 2.32, + "learning_rate": 1.1355475763016159e-05, + "loss": 0.1568, + "step": 60320 + }, + { + "epoch": 2.32, + "learning_rate": 1.1354834572967427e-05, + "loss": 0.2178, + "step": 60321 + }, + { + "epoch": 2.32, + "learning_rate": 1.1354193382918697e-05, + "loss": 0.1979, + "step": 60322 + }, + { + "epoch": 2.32, + "learning_rate": 1.1353552192869967e-05, + "loss": 0.257, + "step": 60323 + }, + { + "epoch": 2.32, + "learning_rate": 1.1352911002821237e-05, + "loss": 0.2799, + "step": 60324 + }, + { + "epoch": 2.32, + "learning_rate": 1.1352269812772506e-05, + "loss": 0.4063, + "step": 60325 + }, + { + "epoch": 2.32, + "learning_rate": 1.1351628622723776e-05, + "loss": 0.1629, + "step": 60326 + }, + { + "epoch": 2.32, + "learning_rate": 1.1350987432675044e-05, + "loss": 0.1318, + "step": 60327 + }, + { + "epoch": 2.32, + "learning_rate": 1.1350346242626316e-05, + "loss": 0.3435, + "step": 60328 + }, + { + "epoch": 2.32, + "learning_rate": 1.1349705052577584e-05, + "loss": 0.1822, + "step": 60329 + }, + { + "epoch": 2.32, + "learning_rate": 1.1349063862528854e-05, + "loss": 0.1956, + "step": 60330 + }, + { + "epoch": 2.32, + "learning_rate": 1.1348422672480123e-05, + "loss": 0.3524, + "step": 60331 + }, + { + "epoch": 2.32, + "learning_rate": 1.1347781482431394e-05, + "loss": 0.2766, + "step": 60332 + }, + { + "epoch": 2.32, + "learning_rate": 1.1347140292382663e-05, + "loss": 0.3267, + "step": 60333 + }, + { + "epoch": 2.32, + "learning_rate": 1.1346499102333933e-05, + "loss": 0.1823, + "step": 60334 + }, + { + "epoch": 2.32, + "learning_rate": 1.1345857912285201e-05, + "loss": 0.2135, + "step": 60335 + }, + { + "epoch": 2.32, + "learning_rate": 1.1345216722236471e-05, + "loss": 0.2037, + "step": 60336 + }, + { + "epoch": 2.32, + "learning_rate": 1.1344575532187741e-05, + "loss": 0.1655, + "step": 60337 + }, + { + "epoch": 2.32, + "learning_rate": 1.1343934342139011e-05, + "loss": 0.1784, + "step": 60338 + }, + { + "epoch": 2.32, + "learning_rate": 1.134329315209028e-05, + "loss": 0.2129, + "step": 60339 + }, + { + "epoch": 2.32, + "learning_rate": 1.134265196204155e-05, + "loss": 0.1757, + "step": 60340 + }, + { + "epoch": 2.32, + "learning_rate": 1.134201077199282e-05, + "loss": 0.4257, + "step": 60341 + }, + { + "epoch": 2.32, + "learning_rate": 1.134136958194409e-05, + "loss": 0.1897, + "step": 60342 + }, + { + "epoch": 2.32, + "learning_rate": 1.1340728391895358e-05, + "loss": 0.2008, + "step": 60343 + }, + { + "epoch": 2.32, + "learning_rate": 1.1340087201846628e-05, + "loss": 0.2589, + "step": 60344 + }, + { + "epoch": 2.32, + "learning_rate": 1.1339446011797896e-05, + "loss": 0.2006, + "step": 60345 + }, + { + "epoch": 2.32, + "learning_rate": 1.1338804821749168e-05, + "loss": 0.1422, + "step": 60346 + }, + { + "epoch": 2.32, + "learning_rate": 1.1338163631700436e-05, + "loss": 0.2113, + "step": 60347 + }, + { + "epoch": 2.32, + "learning_rate": 1.1337522441651706e-05, + "loss": 0.1733, + "step": 60348 + }, + { + "epoch": 2.32, + "learning_rate": 1.1336881251602974e-05, + "loss": 0.1891, + "step": 60349 + }, + { + "epoch": 2.32, + "learning_rate": 1.1336240061554246e-05, + "loss": 0.2679, + "step": 60350 + }, + { + "epoch": 2.32, + "learning_rate": 1.1335598871505515e-05, + "loss": 0.1889, + "step": 60351 + }, + { + "epoch": 2.32, + "learning_rate": 1.1334957681456785e-05, + "loss": 0.2912, + "step": 60352 + }, + { + "epoch": 2.32, + "learning_rate": 1.1334316491408053e-05, + "loss": 0.0945, + "step": 60353 + }, + { + "epoch": 2.32, + "learning_rate": 1.1333675301359323e-05, + "loss": 0.132, + "step": 60354 + }, + { + "epoch": 2.32, + "learning_rate": 1.1333034111310593e-05, + "loss": 0.2703, + "step": 60355 + }, + { + "epoch": 2.32, + "learning_rate": 1.1332392921261863e-05, + "loss": 0.3222, + "step": 60356 + }, + { + "epoch": 2.32, + "learning_rate": 1.1331751731213131e-05, + "loss": 0.1455, + "step": 60357 + }, + { + "epoch": 2.32, + "learning_rate": 1.1331110541164401e-05, + "loss": 0.4774, + "step": 60358 + }, + { + "epoch": 2.32, + "learning_rate": 1.1330469351115671e-05, + "loss": 0.4277, + "step": 60359 + }, + { + "epoch": 2.32, + "learning_rate": 1.1329828161066941e-05, + "loss": 0.1943, + "step": 60360 + }, + { + "epoch": 2.32, + "learning_rate": 1.132918697101821e-05, + "loss": 0.2529, + "step": 60361 + }, + { + "epoch": 2.32, + "learning_rate": 1.132854578096948e-05, + "loss": 0.2646, + "step": 60362 + }, + { + "epoch": 2.32, + "learning_rate": 1.132790459092075e-05, + "loss": 0.1702, + "step": 60363 + }, + { + "epoch": 2.32, + "learning_rate": 1.132726340087202e-05, + "loss": 0.1961, + "step": 60364 + }, + { + "epoch": 2.32, + "learning_rate": 1.1326622210823288e-05, + "loss": 0.26, + "step": 60365 + }, + { + "epoch": 2.32, + "learning_rate": 1.1325981020774558e-05, + "loss": 0.2228, + "step": 60366 + }, + { + "epoch": 2.32, + "learning_rate": 1.1325339830725826e-05, + "loss": 0.206, + "step": 60367 + }, + { + "epoch": 2.32, + "learning_rate": 1.1324698640677098e-05, + "loss": 0.2112, + "step": 60368 + }, + { + "epoch": 2.32, + "learning_rate": 1.1324057450628366e-05, + "loss": 0.1695, + "step": 60369 + }, + { + "epoch": 2.32, + "learning_rate": 1.1323416260579636e-05, + "loss": 0.1906, + "step": 60370 + }, + { + "epoch": 2.32, + "learning_rate": 1.1322775070530905e-05, + "loss": 0.1623, + "step": 60371 + }, + { + "epoch": 2.32, + "learning_rate": 1.1322133880482177e-05, + "loss": 0.14, + "step": 60372 + }, + { + "epoch": 2.32, + "learning_rate": 1.1321492690433445e-05, + "loss": 0.1929, + "step": 60373 + }, + { + "epoch": 2.32, + "learning_rate": 1.1320851500384715e-05, + "loss": 0.2084, + "step": 60374 + }, + { + "epoch": 2.32, + "learning_rate": 1.1320210310335983e-05, + "loss": 0.1875, + "step": 60375 + }, + { + "epoch": 2.32, + "learning_rate": 1.1319569120287253e-05, + "loss": 0.2307, + "step": 60376 + }, + { + "epoch": 2.32, + "learning_rate": 1.1318927930238523e-05, + "loss": 0.2632, + "step": 60377 + }, + { + "epoch": 2.32, + "learning_rate": 1.1318286740189793e-05, + "loss": 0.3247, + "step": 60378 + }, + { + "epoch": 2.32, + "learning_rate": 1.1317645550141062e-05, + "loss": 0.1732, + "step": 60379 + }, + { + "epoch": 2.32, + "learning_rate": 1.1317004360092332e-05, + "loss": 0.3016, + "step": 60380 + }, + { + "epoch": 2.32, + "learning_rate": 1.1316363170043602e-05, + "loss": 0.2102, + "step": 60381 + }, + { + "epoch": 2.32, + "learning_rate": 1.1315721979994872e-05, + "loss": 0.4072, + "step": 60382 + }, + { + "epoch": 2.32, + "learning_rate": 1.131508078994614e-05, + "loss": 0.2849, + "step": 60383 + }, + { + "epoch": 2.32, + "learning_rate": 1.131443959989741e-05, + "loss": 0.1867, + "step": 60384 + }, + { + "epoch": 2.32, + "learning_rate": 1.1313798409848678e-05, + "loss": 0.2355, + "step": 60385 + }, + { + "epoch": 2.32, + "learning_rate": 1.131315721979995e-05, + "loss": 0.1267, + "step": 60386 + }, + { + "epoch": 2.32, + "learning_rate": 1.1312516029751218e-05, + "loss": 0.4464, + "step": 60387 + }, + { + "epoch": 2.32, + "learning_rate": 1.1311874839702488e-05, + "loss": 0.3194, + "step": 60388 + }, + { + "epoch": 2.32, + "learning_rate": 1.1311233649653757e-05, + "loss": 0.2566, + "step": 60389 + }, + { + "epoch": 2.32, + "learning_rate": 1.1310592459605028e-05, + "loss": 0.2751, + "step": 60390 + }, + { + "epoch": 2.32, + "learning_rate": 1.1309951269556297e-05, + "loss": 0.2102, + "step": 60391 + }, + { + "epoch": 2.32, + "learning_rate": 1.1309310079507567e-05, + "loss": 0.2713, + "step": 60392 + }, + { + "epoch": 2.32, + "learning_rate": 1.1308668889458835e-05, + "loss": 0.3871, + "step": 60393 + }, + { + "epoch": 2.32, + "learning_rate": 1.1308027699410105e-05, + "loss": 0.3366, + "step": 60394 + }, + { + "epoch": 2.32, + "learning_rate": 1.1307386509361375e-05, + "loss": 0.2418, + "step": 60395 + }, + { + "epoch": 2.32, + "learning_rate": 1.1306745319312645e-05, + "loss": 0.2629, + "step": 60396 + }, + { + "epoch": 2.32, + "learning_rate": 1.1306104129263914e-05, + "loss": 0.2843, + "step": 60397 + }, + { + "epoch": 2.32, + "learning_rate": 1.1305462939215184e-05, + "loss": 0.309, + "step": 60398 + }, + { + "epoch": 2.32, + "learning_rate": 1.1304821749166454e-05, + "loss": 0.3052, + "step": 60399 + }, + { + "epoch": 2.32, + "learning_rate": 1.1304180559117724e-05, + "loss": 0.1859, + "step": 60400 + }, + { + "epoch": 2.32, + "learning_rate": 1.1303539369068992e-05, + "loss": 0.2422, + "step": 60401 + }, + { + "epoch": 2.32, + "learning_rate": 1.1302898179020262e-05, + "loss": 0.2295, + "step": 60402 + }, + { + "epoch": 2.32, + "learning_rate": 1.130225698897153e-05, + "loss": 0.2252, + "step": 60403 + }, + { + "epoch": 2.32, + "learning_rate": 1.1301615798922802e-05, + "loss": 0.2569, + "step": 60404 + }, + { + "epoch": 2.32, + "learning_rate": 1.130097460887407e-05, + "loss": 0.2583, + "step": 60405 + }, + { + "epoch": 2.32, + "learning_rate": 1.130033341882534e-05, + "loss": 0.2014, + "step": 60406 + }, + { + "epoch": 2.32, + "learning_rate": 1.1299692228776609e-05, + "loss": 0.3186, + "step": 60407 + }, + { + "epoch": 2.32, + "learning_rate": 1.129905103872788e-05, + "loss": 0.2127, + "step": 60408 + }, + { + "epoch": 2.32, + "learning_rate": 1.1298409848679149e-05, + "loss": 0.1454, + "step": 60409 + }, + { + "epoch": 2.32, + "learning_rate": 1.1297768658630419e-05, + "loss": 0.27, + "step": 60410 + }, + { + "epoch": 2.32, + "learning_rate": 1.1297127468581687e-05, + "loss": 0.2969, + "step": 60411 + }, + { + "epoch": 2.32, + "learning_rate": 1.1296486278532959e-05, + "loss": 0.2775, + "step": 60412 + }, + { + "epoch": 2.32, + "learning_rate": 1.1295845088484227e-05, + "loss": 0.3374, + "step": 60413 + }, + { + "epoch": 2.32, + "learning_rate": 1.1295203898435497e-05, + "loss": 0.144, + "step": 60414 + }, + { + "epoch": 2.32, + "learning_rate": 1.1294562708386765e-05, + "loss": 0.1984, + "step": 60415 + }, + { + "epoch": 2.32, + "learning_rate": 1.1293921518338036e-05, + "loss": 0.2661, + "step": 60416 + }, + { + "epoch": 2.32, + "learning_rate": 1.1293280328289306e-05, + "loss": 0.1045, + "step": 60417 + }, + { + "epoch": 2.32, + "learning_rate": 1.1292639138240576e-05, + "loss": 0.1847, + "step": 60418 + }, + { + "epoch": 2.32, + "learning_rate": 1.1291997948191844e-05, + "loss": 0.1947, + "step": 60419 + }, + { + "epoch": 2.32, + "learning_rate": 1.1291356758143114e-05, + "loss": 0.2647, + "step": 60420 + }, + { + "epoch": 2.32, + "learning_rate": 1.1290715568094384e-05, + "loss": 0.2785, + "step": 60421 + }, + { + "epoch": 2.32, + "learning_rate": 1.1290074378045654e-05, + "loss": 0.1769, + "step": 60422 + }, + { + "epoch": 2.32, + "learning_rate": 1.1289433187996922e-05, + "loss": 0.0948, + "step": 60423 + }, + { + "epoch": 2.32, + "learning_rate": 1.1288791997948192e-05, + "loss": 0.0841, + "step": 60424 + }, + { + "epoch": 2.32, + "learning_rate": 1.128815080789946e-05, + "loss": 0.3821, + "step": 60425 + }, + { + "epoch": 2.32, + "learning_rate": 1.1287509617850732e-05, + "loss": 0.1919, + "step": 60426 + }, + { + "epoch": 2.32, + "learning_rate": 1.1286868427802e-05, + "loss": 0.2171, + "step": 60427 + }, + { + "epoch": 2.32, + "learning_rate": 1.128622723775327e-05, + "loss": 0.2083, + "step": 60428 + }, + { + "epoch": 2.32, + "learning_rate": 1.1285586047704539e-05, + "loss": 0.3363, + "step": 60429 + }, + { + "epoch": 2.32, + "learning_rate": 1.128494485765581e-05, + "loss": 0.1914, + "step": 60430 + }, + { + "epoch": 2.32, + "learning_rate": 1.1284303667607079e-05, + "loss": 0.2441, + "step": 60431 + }, + { + "epoch": 2.32, + "learning_rate": 1.1283662477558349e-05, + "loss": 0.2434, + "step": 60432 + }, + { + "epoch": 2.32, + "learning_rate": 1.1283021287509617e-05, + "loss": 0.212, + "step": 60433 + }, + { + "epoch": 2.32, + "learning_rate": 1.1282380097460887e-05, + "loss": 0.1783, + "step": 60434 + }, + { + "epoch": 2.32, + "learning_rate": 1.1281738907412157e-05, + "loss": 0.2925, + "step": 60435 + }, + { + "epoch": 2.32, + "learning_rate": 1.1281097717363427e-05, + "loss": 0.2383, + "step": 60436 + }, + { + "epoch": 2.32, + "learning_rate": 1.1280456527314696e-05, + "loss": 0.186, + "step": 60437 + }, + { + "epoch": 2.32, + "learning_rate": 1.1279815337265966e-05, + "loss": 0.2522, + "step": 60438 + }, + { + "epoch": 2.32, + "learning_rate": 1.1279174147217236e-05, + "loss": 0.2092, + "step": 60439 + }, + { + "epoch": 2.32, + "learning_rate": 1.1278532957168506e-05, + "loss": 0.2687, + "step": 60440 + }, + { + "epoch": 2.32, + "learning_rate": 1.1277891767119774e-05, + "loss": 0.402, + "step": 60441 + }, + { + "epoch": 2.32, + "learning_rate": 1.1277250577071044e-05, + "loss": 0.1383, + "step": 60442 + }, + { + "epoch": 2.32, + "learning_rate": 1.1276609387022313e-05, + "loss": 0.1665, + "step": 60443 + }, + { + "epoch": 2.32, + "learning_rate": 1.1275968196973584e-05, + "loss": 0.2236, + "step": 60444 + }, + { + "epoch": 2.32, + "learning_rate": 1.1275327006924853e-05, + "loss": 0.3409, + "step": 60445 + }, + { + "epoch": 2.32, + "learning_rate": 1.1274685816876123e-05, + "loss": 0.1331, + "step": 60446 + }, + { + "epoch": 2.32, + "learning_rate": 1.1274044626827391e-05, + "loss": 0.1817, + "step": 60447 + }, + { + "epoch": 2.32, + "learning_rate": 1.1273403436778663e-05, + "loss": 0.2411, + "step": 60448 + }, + { + "epoch": 2.32, + "learning_rate": 1.1272762246729931e-05, + "loss": 0.3082, + "step": 60449 + }, + { + "epoch": 2.32, + "learning_rate": 1.1272121056681201e-05, + "loss": 0.1939, + "step": 60450 + }, + { + "epoch": 2.32, + "learning_rate": 1.127147986663247e-05, + "loss": 0.2071, + "step": 60451 + }, + { + "epoch": 2.32, + "learning_rate": 1.1270838676583741e-05, + "loss": 0.2236, + "step": 60452 + }, + { + "epoch": 2.32, + "learning_rate": 1.127019748653501e-05, + "loss": 0.3128, + "step": 60453 + }, + { + "epoch": 2.32, + "learning_rate": 1.126955629648628e-05, + "loss": 0.3719, + "step": 60454 + }, + { + "epoch": 2.32, + "learning_rate": 1.1268915106437548e-05, + "loss": 0.159, + "step": 60455 + }, + { + "epoch": 2.32, + "learning_rate": 1.1268273916388818e-05, + "loss": 0.2071, + "step": 60456 + }, + { + "epoch": 2.32, + "learning_rate": 1.1267632726340088e-05, + "loss": 0.1964, + "step": 60457 + }, + { + "epoch": 2.32, + "learning_rate": 1.1266991536291358e-05, + "loss": 0.1198, + "step": 60458 + }, + { + "epoch": 2.32, + "learning_rate": 1.1266350346242626e-05, + "loss": 0.2328, + "step": 60459 + }, + { + "epoch": 2.32, + "learning_rate": 1.1265709156193896e-05, + "loss": 0.2455, + "step": 60460 + }, + { + "epoch": 2.32, + "learning_rate": 1.1265067966145166e-05, + "loss": 0.1815, + "step": 60461 + }, + { + "epoch": 2.32, + "learning_rate": 1.1264426776096436e-05, + "loss": 0.2003, + "step": 60462 + }, + { + "epoch": 2.32, + "learning_rate": 1.1263785586047705e-05, + "loss": 0.1743, + "step": 60463 + }, + { + "epoch": 2.32, + "learning_rate": 1.1263144395998975e-05, + "loss": 0.2577, + "step": 60464 + }, + { + "epoch": 2.32, + "learning_rate": 1.1262503205950243e-05, + "loss": 0.2499, + "step": 60465 + }, + { + "epoch": 2.32, + "learning_rate": 1.1261862015901515e-05, + "loss": 0.3378, + "step": 60466 + }, + { + "epoch": 2.32, + "learning_rate": 1.1261220825852783e-05, + "loss": 0.1939, + "step": 60467 + }, + { + "epoch": 2.32, + "learning_rate": 1.1260579635804053e-05, + "loss": 0.1685, + "step": 60468 + }, + { + "epoch": 2.32, + "learning_rate": 1.1259938445755321e-05, + "loss": 0.0832, + "step": 60469 + }, + { + "epoch": 2.32, + "learning_rate": 1.1259297255706593e-05, + "loss": 0.1687, + "step": 60470 + }, + { + "epoch": 2.32, + "learning_rate": 1.1258656065657861e-05, + "loss": 0.165, + "step": 60471 + }, + { + "epoch": 2.32, + "learning_rate": 1.1258014875609131e-05, + "loss": 0.1736, + "step": 60472 + }, + { + "epoch": 2.32, + "learning_rate": 1.12573736855604e-05, + "loss": 0.3068, + "step": 60473 + }, + { + "epoch": 2.33, + "learning_rate": 1.125673249551167e-05, + "loss": 0.3015, + "step": 60474 + }, + { + "epoch": 2.33, + "learning_rate": 1.125609130546294e-05, + "loss": 0.2435, + "step": 60475 + }, + { + "epoch": 2.33, + "learning_rate": 1.125545011541421e-05, + "loss": 0.2358, + "step": 60476 + }, + { + "epoch": 2.33, + "learning_rate": 1.1254808925365478e-05, + "loss": 0.3015, + "step": 60477 + }, + { + "epoch": 2.33, + "learning_rate": 1.1254167735316748e-05, + "loss": 0.2192, + "step": 60478 + }, + { + "epoch": 2.33, + "learning_rate": 1.1253526545268018e-05, + "loss": 0.4567, + "step": 60479 + }, + { + "epoch": 2.33, + "learning_rate": 1.1252885355219288e-05, + "loss": 0.1697, + "step": 60480 + }, + { + "epoch": 2.33, + "learning_rate": 1.1252244165170556e-05, + "loss": 0.275, + "step": 60481 + }, + { + "epoch": 2.33, + "learning_rate": 1.1251602975121826e-05, + "loss": 0.196, + "step": 60482 + }, + { + "epoch": 2.33, + "learning_rate": 1.1250961785073095e-05, + "loss": 0.3303, + "step": 60483 + }, + { + "epoch": 2.33, + "learning_rate": 1.1250320595024367e-05, + "loss": 0.2825, + "step": 60484 + }, + { + "epoch": 2.33, + "learning_rate": 1.1249679404975635e-05, + "loss": 0.0968, + "step": 60485 + }, + { + "epoch": 2.33, + "learning_rate": 1.1249038214926905e-05, + "loss": 0.2799, + "step": 60486 + }, + { + "epoch": 2.33, + "learning_rate": 1.1248397024878173e-05, + "loss": 0.2166, + "step": 60487 + }, + { + "epoch": 2.33, + "learning_rate": 1.1247755834829445e-05, + "loss": 0.2512, + "step": 60488 + }, + { + "epoch": 2.33, + "learning_rate": 1.1247114644780713e-05, + "loss": 0.1964, + "step": 60489 + }, + { + "epoch": 2.33, + "learning_rate": 1.1246473454731983e-05, + "loss": 0.2754, + "step": 60490 + }, + { + "epoch": 2.33, + "learning_rate": 1.1245832264683252e-05, + "loss": 0.3413, + "step": 60491 + }, + { + "epoch": 2.33, + "learning_rate": 1.1245191074634522e-05, + "loss": 0.3274, + "step": 60492 + }, + { + "epoch": 2.33, + "learning_rate": 1.1244549884585792e-05, + "loss": 0.1854, + "step": 60493 + }, + { + "epoch": 2.33, + "learning_rate": 1.1243908694537062e-05, + "loss": 0.1317, + "step": 60494 + }, + { + "epoch": 2.33, + "learning_rate": 1.124326750448833e-05, + "loss": 0.2177, + "step": 60495 + }, + { + "epoch": 2.33, + "learning_rate": 1.12426263144396e-05, + "loss": 0.2987, + "step": 60496 + }, + { + "epoch": 2.33, + "learning_rate": 1.124198512439087e-05, + "loss": 0.2445, + "step": 60497 + }, + { + "epoch": 2.33, + "learning_rate": 1.124134393434214e-05, + "loss": 0.1922, + "step": 60498 + }, + { + "epoch": 2.33, + "learning_rate": 1.1240702744293408e-05, + "loss": 0.216, + "step": 60499 + }, + { + "epoch": 2.33, + "learning_rate": 1.1240061554244678e-05, + "loss": 0.2891, + "step": 60500 + }, + { + "epoch": 2.33, + "learning_rate": 1.1239420364195948e-05, + "loss": 0.203, + "step": 60501 + }, + { + "epoch": 2.33, + "learning_rate": 1.1238779174147218e-05, + "loss": 0.2458, + "step": 60502 + }, + { + "epoch": 2.33, + "learning_rate": 1.1238137984098487e-05, + "loss": 0.1884, + "step": 60503 + }, + { + "epoch": 2.33, + "learning_rate": 1.1237496794049757e-05, + "loss": 0.2398, + "step": 60504 + }, + { + "epoch": 2.33, + "learning_rate": 1.1236855604001025e-05, + "loss": 0.2249, + "step": 60505 + }, + { + "epoch": 2.33, + "learning_rate": 1.1236214413952297e-05, + "loss": 0.205, + "step": 60506 + }, + { + "epoch": 2.33, + "learning_rate": 1.1235573223903565e-05, + "loss": 0.2497, + "step": 60507 + }, + { + "epoch": 2.33, + "learning_rate": 1.1234932033854835e-05, + "loss": 0.1407, + "step": 60508 + }, + { + "epoch": 2.33, + "learning_rate": 1.1234290843806104e-05, + "loss": 0.2531, + "step": 60509 + }, + { + "epoch": 2.33, + "learning_rate": 1.1233649653757375e-05, + "loss": 0.1969, + "step": 60510 + }, + { + "epoch": 2.33, + "learning_rate": 1.1233008463708644e-05, + "loss": 0.2135, + "step": 60511 + }, + { + "epoch": 2.33, + "learning_rate": 1.1232367273659914e-05, + "loss": 0.1732, + "step": 60512 + }, + { + "epoch": 2.33, + "learning_rate": 1.1231726083611182e-05, + "loss": 0.1595, + "step": 60513 + }, + { + "epoch": 2.33, + "learning_rate": 1.1231084893562452e-05, + "loss": 0.2848, + "step": 60514 + }, + { + "epoch": 2.33, + "learning_rate": 1.1230443703513722e-05, + "loss": 0.1339, + "step": 60515 + }, + { + "epoch": 2.33, + "learning_rate": 1.1229802513464992e-05, + "loss": 0.2433, + "step": 60516 + }, + { + "epoch": 2.33, + "learning_rate": 1.122916132341626e-05, + "loss": 0.1771, + "step": 60517 + }, + { + "epoch": 2.33, + "learning_rate": 1.122852013336753e-05, + "loss": 0.2563, + "step": 60518 + }, + { + "epoch": 2.33, + "learning_rate": 1.12278789433188e-05, + "loss": 0.2253, + "step": 60519 + }, + { + "epoch": 2.33, + "learning_rate": 1.122723775327007e-05, + "loss": 0.2019, + "step": 60520 + }, + { + "epoch": 2.33, + "learning_rate": 1.1226596563221339e-05, + "loss": 0.137, + "step": 60521 + }, + { + "epoch": 2.33, + "learning_rate": 1.1225955373172609e-05, + "loss": 0.2606, + "step": 60522 + }, + { + "epoch": 2.33, + "learning_rate": 1.1225314183123877e-05, + "loss": 0.2194, + "step": 60523 + }, + { + "epoch": 2.33, + "learning_rate": 1.1224672993075149e-05, + "loss": 0.2441, + "step": 60524 + }, + { + "epoch": 2.33, + "learning_rate": 1.1224031803026417e-05, + "loss": 0.2302, + "step": 60525 + }, + { + "epoch": 2.33, + "learning_rate": 1.1223390612977687e-05, + "loss": 0.2026, + "step": 60526 + }, + { + "epoch": 2.33, + "learning_rate": 1.1222749422928955e-05, + "loss": 0.2878, + "step": 60527 + }, + { + "epoch": 2.33, + "learning_rate": 1.1222108232880227e-05, + "loss": 0.1816, + "step": 60528 + }, + { + "epoch": 2.33, + "learning_rate": 1.1221467042831496e-05, + "loss": 0.2006, + "step": 60529 + }, + { + "epoch": 2.33, + "learning_rate": 1.1220825852782766e-05, + "loss": 0.1256, + "step": 60530 + }, + { + "epoch": 2.33, + "learning_rate": 1.1220184662734034e-05, + "loss": 0.1918, + "step": 60531 + }, + { + "epoch": 2.33, + "learning_rate": 1.1219543472685304e-05, + "loss": 0.1898, + "step": 60532 + }, + { + "epoch": 2.33, + "learning_rate": 1.1218902282636574e-05, + "loss": 0.3482, + "step": 60533 + }, + { + "epoch": 2.33, + "learning_rate": 1.1218261092587844e-05, + "loss": 0.1991, + "step": 60534 + }, + { + "epoch": 2.33, + "learning_rate": 1.1217619902539112e-05, + "loss": 0.2111, + "step": 60535 + }, + { + "epoch": 2.33, + "learning_rate": 1.1216978712490382e-05, + "loss": 0.2517, + "step": 60536 + }, + { + "epoch": 2.33, + "learning_rate": 1.1216337522441652e-05, + "loss": 0.2088, + "step": 60537 + }, + { + "epoch": 2.33, + "learning_rate": 1.1215696332392922e-05, + "loss": 0.2775, + "step": 60538 + }, + { + "epoch": 2.33, + "learning_rate": 1.121505514234419e-05, + "loss": 0.5161, + "step": 60539 + }, + { + "epoch": 2.33, + "learning_rate": 1.121441395229546e-05, + "loss": 0.1255, + "step": 60540 + }, + { + "epoch": 2.33, + "learning_rate": 1.1213772762246729e-05, + "loss": 0.3009, + "step": 60541 + }, + { + "epoch": 2.33, + "learning_rate": 1.1213131572198e-05, + "loss": 0.3322, + "step": 60542 + }, + { + "epoch": 2.33, + "learning_rate": 1.1212490382149269e-05, + "loss": 0.2128, + "step": 60543 + }, + { + "epoch": 2.33, + "learning_rate": 1.1211849192100539e-05, + "loss": 0.1181, + "step": 60544 + }, + { + "epoch": 2.33, + "learning_rate": 1.1211208002051807e-05, + "loss": 0.1339, + "step": 60545 + }, + { + "epoch": 2.33, + "learning_rate": 1.1210566812003079e-05, + "loss": 0.3077, + "step": 60546 + }, + { + "epoch": 2.33, + "learning_rate": 1.1209925621954347e-05, + "loss": 0.2953, + "step": 60547 + }, + { + "epoch": 2.33, + "learning_rate": 1.1209284431905617e-05, + "loss": 0.2767, + "step": 60548 + }, + { + "epoch": 2.33, + "learning_rate": 1.1208643241856886e-05, + "loss": 0.2244, + "step": 60549 + }, + { + "epoch": 2.33, + "learning_rate": 1.1208002051808158e-05, + "loss": 0.2675, + "step": 60550 + }, + { + "epoch": 2.33, + "learning_rate": 1.1207360861759426e-05, + "loss": 0.2036, + "step": 60551 + }, + { + "epoch": 2.33, + "learning_rate": 1.1206719671710696e-05, + "loss": 0.2283, + "step": 60552 + }, + { + "epoch": 2.33, + "learning_rate": 1.1206078481661964e-05, + "loss": 0.2319, + "step": 60553 + }, + { + "epoch": 2.33, + "learning_rate": 1.1205437291613234e-05, + "loss": 0.2659, + "step": 60554 + }, + { + "epoch": 2.33, + "learning_rate": 1.1204796101564504e-05, + "loss": 0.3259, + "step": 60555 + }, + { + "epoch": 2.33, + "learning_rate": 1.1204154911515774e-05, + "loss": 0.1861, + "step": 60556 + }, + { + "epoch": 2.33, + "learning_rate": 1.1203513721467043e-05, + "loss": 0.1685, + "step": 60557 + }, + { + "epoch": 2.33, + "learning_rate": 1.1202872531418313e-05, + "loss": 0.2227, + "step": 60558 + }, + { + "epoch": 2.33, + "learning_rate": 1.1202231341369583e-05, + "loss": 0.1433, + "step": 60559 + }, + { + "epoch": 2.33, + "learning_rate": 1.1201590151320853e-05, + "loss": 0.0525, + "step": 60560 + }, + { + "epoch": 2.33, + "learning_rate": 1.1200948961272121e-05, + "loss": 0.1805, + "step": 60561 + }, + { + "epoch": 2.33, + "learning_rate": 1.1200307771223391e-05, + "loss": 0.234, + "step": 60562 + }, + { + "epoch": 2.33, + "learning_rate": 1.119966658117466e-05, + "loss": 0.234, + "step": 60563 + }, + { + "epoch": 2.33, + "learning_rate": 1.1199025391125931e-05, + "loss": 0.3199, + "step": 60564 + }, + { + "epoch": 2.33, + "learning_rate": 1.11983842010772e-05, + "loss": 0.2967, + "step": 60565 + }, + { + "epoch": 2.33, + "learning_rate": 1.119774301102847e-05, + "loss": 0.3016, + "step": 60566 + }, + { + "epoch": 2.33, + "learning_rate": 1.1197101820979738e-05, + "loss": 0.1395, + "step": 60567 + }, + { + "epoch": 2.33, + "learning_rate": 1.119646063093101e-05, + "loss": 0.188, + "step": 60568 + }, + { + "epoch": 2.33, + "learning_rate": 1.1195819440882278e-05, + "loss": 0.1858, + "step": 60569 + }, + { + "epoch": 2.33, + "learning_rate": 1.1195178250833548e-05, + "loss": 0.2583, + "step": 60570 + }, + { + "epoch": 2.33, + "learning_rate": 1.1194537060784816e-05, + "loss": 0.3103, + "step": 60571 + }, + { + "epoch": 2.33, + "learning_rate": 1.1193895870736086e-05, + "loss": 0.3334, + "step": 60572 + }, + { + "epoch": 2.33, + "learning_rate": 1.1193254680687356e-05, + "loss": 0.1848, + "step": 60573 + }, + { + "epoch": 2.33, + "learning_rate": 1.1192613490638626e-05, + "loss": 0.1907, + "step": 60574 + }, + { + "epoch": 2.33, + "learning_rate": 1.1191972300589895e-05, + "loss": 0.2442, + "step": 60575 + }, + { + "epoch": 2.33, + "learning_rate": 1.1191331110541165e-05, + "loss": 0.3068, + "step": 60576 + }, + { + "epoch": 2.33, + "learning_rate": 1.1190689920492435e-05, + "loss": 0.3426, + "step": 60577 + }, + { + "epoch": 2.33, + "learning_rate": 1.1190048730443705e-05, + "loss": 0.2522, + "step": 60578 + }, + { + "epoch": 2.33, + "learning_rate": 1.1189407540394973e-05, + "loss": 0.2971, + "step": 60579 + }, + { + "epoch": 2.33, + "learning_rate": 1.1188766350346243e-05, + "loss": 0.1314, + "step": 60580 + }, + { + "epoch": 2.33, + "learning_rate": 1.1188125160297511e-05, + "loss": 0.1966, + "step": 60581 + }, + { + "epoch": 2.33, + "learning_rate": 1.1187483970248783e-05, + "loss": 0.166, + "step": 60582 + }, + { + "epoch": 2.33, + "learning_rate": 1.1186842780200051e-05, + "loss": 0.2512, + "step": 60583 + }, + { + "epoch": 2.33, + "learning_rate": 1.1186201590151321e-05, + "loss": 0.2174, + "step": 60584 + }, + { + "epoch": 2.33, + "learning_rate": 1.118556040010259e-05, + "loss": 0.3054, + "step": 60585 + }, + { + "epoch": 2.33, + "learning_rate": 1.1184919210053861e-05, + "loss": 0.1712, + "step": 60586 + }, + { + "epoch": 2.33, + "learning_rate": 1.118427802000513e-05, + "loss": 0.2729, + "step": 60587 + }, + { + "epoch": 2.33, + "learning_rate": 1.11836368299564e-05, + "loss": 0.1616, + "step": 60588 + }, + { + "epoch": 2.33, + "learning_rate": 1.1182995639907668e-05, + "loss": 0.2271, + "step": 60589 + }, + { + "epoch": 2.33, + "learning_rate": 1.1182354449858938e-05, + "loss": 0.1484, + "step": 60590 + }, + { + "epoch": 2.33, + "learning_rate": 1.1181713259810208e-05, + "loss": 0.4181, + "step": 60591 + }, + { + "epoch": 2.33, + "learning_rate": 1.1181072069761478e-05, + "loss": 0.2275, + "step": 60592 + }, + { + "epoch": 2.33, + "learning_rate": 1.1180430879712746e-05, + "loss": 0.2392, + "step": 60593 + }, + { + "epoch": 2.33, + "learning_rate": 1.1179789689664017e-05, + "loss": 0.4138, + "step": 60594 + }, + { + "epoch": 2.33, + "learning_rate": 1.1179148499615287e-05, + "loss": 0.186, + "step": 60595 + }, + { + "epoch": 2.33, + "learning_rate": 1.1178507309566557e-05, + "loss": 0.299, + "step": 60596 + }, + { + "epoch": 2.33, + "learning_rate": 1.1177866119517825e-05, + "loss": 0.2427, + "step": 60597 + }, + { + "epoch": 2.33, + "learning_rate": 1.1177224929469095e-05, + "loss": 0.2679, + "step": 60598 + }, + { + "epoch": 2.33, + "learning_rate": 1.1176583739420365e-05, + "loss": 0.265, + "step": 60599 + }, + { + "epoch": 2.33, + "learning_rate": 1.1175942549371635e-05, + "loss": 0.1621, + "step": 60600 + }, + { + "epoch": 2.33, + "learning_rate": 1.1175301359322903e-05, + "loss": 0.2348, + "step": 60601 + }, + { + "epoch": 2.33, + "learning_rate": 1.1174660169274173e-05, + "loss": 0.2631, + "step": 60602 + }, + { + "epoch": 2.33, + "learning_rate": 1.1174018979225442e-05, + "loss": 0.1811, + "step": 60603 + }, + { + "epoch": 2.33, + "learning_rate": 1.1173377789176713e-05, + "loss": 0.17, + "step": 60604 + }, + { + "epoch": 2.33, + "learning_rate": 1.1172736599127982e-05, + "loss": 0.2384, + "step": 60605 + }, + { + "epoch": 2.33, + "learning_rate": 1.1172095409079252e-05, + "loss": 0.2409, + "step": 60606 + }, + { + "epoch": 2.33, + "learning_rate": 1.117145421903052e-05, + "loss": 0.2607, + "step": 60607 + }, + { + "epoch": 2.33, + "learning_rate": 1.1170813028981792e-05, + "loss": 0.2298, + "step": 60608 + }, + { + "epoch": 2.33, + "learning_rate": 1.117017183893306e-05, + "loss": 0.1587, + "step": 60609 + }, + { + "epoch": 2.33, + "learning_rate": 1.116953064888433e-05, + "loss": 0.1773, + "step": 60610 + }, + { + "epoch": 2.33, + "learning_rate": 1.1168889458835598e-05, + "loss": 0.2084, + "step": 60611 + }, + { + "epoch": 2.33, + "learning_rate": 1.1168248268786868e-05, + "loss": 0.3208, + "step": 60612 + }, + { + "epoch": 2.33, + "learning_rate": 1.1167607078738138e-05, + "loss": 0.1304, + "step": 60613 + }, + { + "epoch": 2.33, + "learning_rate": 1.1166965888689408e-05, + "loss": 0.2204, + "step": 60614 + }, + { + "epoch": 2.33, + "learning_rate": 1.1166324698640677e-05, + "loss": 0.2471, + "step": 60615 + }, + { + "epoch": 2.33, + "learning_rate": 1.1165683508591947e-05, + "loss": 0.1462, + "step": 60616 + }, + { + "epoch": 2.33, + "learning_rate": 1.1165042318543217e-05, + "loss": 0.1787, + "step": 60617 + }, + { + "epoch": 2.33, + "learning_rate": 1.1164401128494487e-05, + "loss": 0.111, + "step": 60618 + }, + { + "epoch": 2.33, + "learning_rate": 1.1163759938445755e-05, + "loss": 0.0709, + "step": 60619 + }, + { + "epoch": 2.33, + "learning_rate": 1.1163118748397025e-05, + "loss": 0.241, + "step": 60620 + }, + { + "epoch": 2.33, + "learning_rate": 1.1162477558348294e-05, + "loss": 0.2477, + "step": 60621 + }, + { + "epoch": 2.33, + "learning_rate": 1.1161836368299565e-05, + "loss": 0.237, + "step": 60622 + }, + { + "epoch": 2.33, + "learning_rate": 1.1161195178250834e-05, + "loss": 0.2306, + "step": 60623 + }, + { + "epoch": 2.33, + "learning_rate": 1.1160553988202104e-05, + "loss": 0.2494, + "step": 60624 + }, + { + "epoch": 2.33, + "learning_rate": 1.1159912798153372e-05, + "loss": 0.1544, + "step": 60625 + }, + { + "epoch": 2.33, + "learning_rate": 1.1159271608104644e-05, + "loss": 0.1384, + "step": 60626 + }, + { + "epoch": 2.33, + "learning_rate": 1.1158630418055912e-05, + "loss": 0.2051, + "step": 60627 + }, + { + "epoch": 2.33, + "learning_rate": 1.1157989228007182e-05, + "loss": 0.2314, + "step": 60628 + }, + { + "epoch": 2.33, + "learning_rate": 1.115734803795845e-05, + "loss": 0.2562, + "step": 60629 + }, + { + "epoch": 2.33, + "learning_rate": 1.115670684790972e-05, + "loss": 0.3673, + "step": 60630 + }, + { + "epoch": 2.33, + "learning_rate": 1.115606565786099e-05, + "loss": 0.1905, + "step": 60631 + }, + { + "epoch": 2.33, + "learning_rate": 1.115542446781226e-05, + "loss": 0.1551, + "step": 60632 + }, + { + "epoch": 2.33, + "learning_rate": 1.1154783277763529e-05, + "loss": 0.3248, + "step": 60633 + }, + { + "epoch": 2.33, + "learning_rate": 1.1154142087714799e-05, + "loss": 0.1445, + "step": 60634 + }, + { + "epoch": 2.33, + "learning_rate": 1.1153500897666069e-05, + "loss": 0.3647, + "step": 60635 + }, + { + "epoch": 2.33, + "learning_rate": 1.1152859707617339e-05, + "loss": 0.1975, + "step": 60636 + }, + { + "epoch": 2.33, + "learning_rate": 1.1152218517568607e-05, + "loss": 0.2168, + "step": 60637 + }, + { + "epoch": 2.33, + "learning_rate": 1.1151577327519877e-05, + "loss": 0.1896, + "step": 60638 + }, + { + "epoch": 2.33, + "learning_rate": 1.1150936137471146e-05, + "loss": 0.2064, + "step": 60639 + }, + { + "epoch": 2.33, + "learning_rate": 1.1150294947422417e-05, + "loss": 0.2439, + "step": 60640 + }, + { + "epoch": 2.33, + "learning_rate": 1.1149653757373686e-05, + "loss": 0.1395, + "step": 60641 + }, + { + "epoch": 2.33, + "learning_rate": 1.1149012567324956e-05, + "loss": 0.2121, + "step": 60642 + }, + { + "epoch": 2.33, + "learning_rate": 1.1148371377276224e-05, + "loss": 0.288, + "step": 60643 + }, + { + "epoch": 2.33, + "learning_rate": 1.1147730187227496e-05, + "loss": 0.3144, + "step": 60644 + }, + { + "epoch": 2.33, + "learning_rate": 1.1147088997178764e-05, + "loss": 0.2232, + "step": 60645 + }, + { + "epoch": 2.33, + "learning_rate": 1.1146447807130034e-05, + "loss": 0.3106, + "step": 60646 + }, + { + "epoch": 2.33, + "learning_rate": 1.1145806617081302e-05, + "loss": 0.1615, + "step": 60647 + }, + { + "epoch": 2.33, + "learning_rate": 1.1145165427032574e-05, + "loss": 0.3316, + "step": 60648 + }, + { + "epoch": 2.33, + "learning_rate": 1.1144524236983842e-05, + "loss": 0.2835, + "step": 60649 + }, + { + "epoch": 2.33, + "learning_rate": 1.1143883046935112e-05, + "loss": 0.3628, + "step": 60650 + }, + { + "epoch": 2.33, + "learning_rate": 1.114324185688638e-05, + "loss": 0.284, + "step": 60651 + }, + { + "epoch": 2.33, + "learning_rate": 1.114260066683765e-05, + "loss": 0.2367, + "step": 60652 + }, + { + "epoch": 2.33, + "learning_rate": 1.114195947678892e-05, + "loss": 0.1522, + "step": 60653 + }, + { + "epoch": 2.33, + "learning_rate": 1.114131828674019e-05, + "loss": 0.2247, + "step": 60654 + }, + { + "epoch": 2.33, + "learning_rate": 1.1140677096691459e-05, + "loss": 0.2263, + "step": 60655 + }, + { + "epoch": 2.33, + "learning_rate": 1.1140035906642729e-05, + "loss": 0.4539, + "step": 60656 + }, + { + "epoch": 2.33, + "learning_rate": 1.1139394716593999e-05, + "loss": 0.2035, + "step": 60657 + }, + { + "epoch": 2.33, + "learning_rate": 1.113875352654527e-05, + "loss": 0.3162, + "step": 60658 + }, + { + "epoch": 2.33, + "learning_rate": 1.1138112336496537e-05, + "loss": 0.2065, + "step": 60659 + }, + { + "epoch": 2.33, + "learning_rate": 1.1137471146447808e-05, + "loss": 0.2038, + "step": 60660 + }, + { + "epoch": 2.33, + "learning_rate": 1.1136829956399076e-05, + "loss": 0.1526, + "step": 60661 + }, + { + "epoch": 2.33, + "learning_rate": 1.1136188766350348e-05, + "loss": 0.1765, + "step": 60662 + }, + { + "epoch": 2.33, + "learning_rate": 1.1135547576301616e-05, + "loss": 0.2277, + "step": 60663 + }, + { + "epoch": 2.33, + "learning_rate": 1.1134906386252886e-05, + "loss": 0.1517, + "step": 60664 + }, + { + "epoch": 2.33, + "learning_rate": 1.1134265196204154e-05, + "loss": 0.3227, + "step": 60665 + }, + { + "epoch": 2.33, + "learning_rate": 1.1133624006155426e-05, + "loss": 0.2312, + "step": 60666 + }, + { + "epoch": 2.33, + "learning_rate": 1.1132982816106694e-05, + "loss": 0.2641, + "step": 60667 + }, + { + "epoch": 2.33, + "learning_rate": 1.1132341626057964e-05, + "loss": 0.2465, + "step": 60668 + }, + { + "epoch": 2.33, + "learning_rate": 1.1131700436009233e-05, + "loss": 0.3392, + "step": 60669 + }, + { + "epoch": 2.33, + "learning_rate": 1.1131059245960503e-05, + "loss": 0.246, + "step": 60670 + }, + { + "epoch": 2.33, + "learning_rate": 1.1130418055911773e-05, + "loss": 0.2398, + "step": 60671 + }, + { + "epoch": 2.33, + "learning_rate": 1.1129776865863043e-05, + "loss": 0.293, + "step": 60672 + }, + { + "epoch": 2.33, + "learning_rate": 1.1129135675814311e-05, + "loss": 0.281, + "step": 60673 + }, + { + "epoch": 2.33, + "learning_rate": 1.1128494485765581e-05, + "loss": 0.2087, + "step": 60674 + }, + { + "epoch": 2.33, + "learning_rate": 1.1127853295716851e-05, + "loss": 0.3362, + "step": 60675 + }, + { + "epoch": 2.33, + "learning_rate": 1.1127212105668121e-05, + "loss": 0.2306, + "step": 60676 + }, + { + "epoch": 2.33, + "learning_rate": 1.112657091561939e-05, + "loss": 0.2456, + "step": 60677 + }, + { + "epoch": 2.33, + "learning_rate": 1.112592972557066e-05, + "loss": 0.1776, + "step": 60678 + }, + { + "epoch": 2.33, + "learning_rate": 1.1125288535521928e-05, + "loss": 0.3151, + "step": 60679 + }, + { + "epoch": 2.33, + "learning_rate": 1.11246473454732e-05, + "loss": 0.2242, + "step": 60680 + }, + { + "epoch": 2.33, + "learning_rate": 1.1124006155424468e-05, + "loss": 0.2611, + "step": 60681 + }, + { + "epoch": 2.33, + "learning_rate": 1.1123364965375738e-05, + "loss": 0.1994, + "step": 60682 + }, + { + "epoch": 2.33, + "learning_rate": 1.1122723775327006e-05, + "loss": 0.1432, + "step": 60683 + }, + { + "epoch": 2.33, + "learning_rate": 1.1122082585278278e-05, + "loss": 0.1983, + "step": 60684 + }, + { + "epoch": 2.33, + "learning_rate": 1.1121441395229546e-05, + "loss": 0.2526, + "step": 60685 + }, + { + "epoch": 2.33, + "learning_rate": 1.1120800205180816e-05, + "loss": 0.2539, + "step": 60686 + }, + { + "epoch": 2.33, + "learning_rate": 1.1120159015132085e-05, + "loss": 0.208, + "step": 60687 + }, + { + "epoch": 2.33, + "learning_rate": 1.1119517825083356e-05, + "loss": 0.1806, + "step": 60688 + }, + { + "epoch": 2.33, + "learning_rate": 1.1118876635034625e-05, + "loss": 0.1798, + "step": 60689 + }, + { + "epoch": 2.33, + "learning_rate": 1.1118235444985895e-05, + "loss": 0.3035, + "step": 60690 + }, + { + "epoch": 2.33, + "learning_rate": 1.1117594254937163e-05, + "loss": 0.154, + "step": 60691 + }, + { + "epoch": 2.33, + "learning_rate": 1.1116953064888433e-05, + "loss": 0.1409, + "step": 60692 + }, + { + "epoch": 2.33, + "learning_rate": 1.1116311874839703e-05, + "loss": 0.2415, + "step": 60693 + }, + { + "epoch": 2.33, + "learning_rate": 1.1115670684790973e-05, + "loss": 0.2866, + "step": 60694 + }, + { + "epoch": 2.33, + "learning_rate": 1.1115029494742241e-05, + "loss": 0.3018, + "step": 60695 + }, + { + "epoch": 2.33, + "learning_rate": 1.1114388304693511e-05, + "loss": 0.1989, + "step": 60696 + }, + { + "epoch": 2.33, + "learning_rate": 1.1113747114644781e-05, + "loss": 0.3099, + "step": 60697 + }, + { + "epoch": 2.33, + "learning_rate": 1.1113105924596051e-05, + "loss": 0.2685, + "step": 60698 + }, + { + "epoch": 2.33, + "learning_rate": 1.111246473454732e-05, + "loss": 0.1294, + "step": 60699 + }, + { + "epoch": 2.33, + "learning_rate": 1.111182354449859e-05, + "loss": 0.2528, + "step": 60700 + }, + { + "epoch": 2.33, + "learning_rate": 1.1111182354449858e-05, + "loss": 0.2132, + "step": 60701 + }, + { + "epoch": 2.33, + "learning_rate": 1.111054116440113e-05, + "loss": 0.3033, + "step": 60702 + }, + { + "epoch": 2.33, + "learning_rate": 1.1109899974352398e-05, + "loss": 0.1934, + "step": 60703 + }, + { + "epoch": 2.33, + "learning_rate": 1.1109258784303668e-05, + "loss": 0.3067, + "step": 60704 + }, + { + "epoch": 2.33, + "learning_rate": 1.1108617594254936e-05, + "loss": 0.1702, + "step": 60705 + }, + { + "epoch": 2.33, + "learning_rate": 1.1107976404206208e-05, + "loss": 0.3332, + "step": 60706 + }, + { + "epoch": 2.33, + "learning_rate": 1.1107335214157477e-05, + "loss": 0.4793, + "step": 60707 + }, + { + "epoch": 2.33, + "learning_rate": 1.1106694024108747e-05, + "loss": 0.1838, + "step": 60708 + }, + { + "epoch": 2.33, + "learning_rate": 1.1106052834060015e-05, + "loss": 0.1573, + "step": 60709 + }, + { + "epoch": 2.33, + "learning_rate": 1.1105411644011285e-05, + "loss": 0.1728, + "step": 60710 + }, + { + "epoch": 2.33, + "learning_rate": 1.1104770453962555e-05, + "loss": 0.1723, + "step": 60711 + }, + { + "epoch": 2.33, + "learning_rate": 1.1104129263913825e-05, + "loss": 0.3642, + "step": 60712 + }, + { + "epoch": 2.33, + "learning_rate": 1.1103488073865093e-05, + "loss": 0.1915, + "step": 60713 + }, + { + "epoch": 2.33, + "learning_rate": 1.1102846883816363e-05, + "loss": 0.1901, + "step": 60714 + }, + { + "epoch": 2.33, + "learning_rate": 1.1102205693767633e-05, + "loss": 0.2649, + "step": 60715 + }, + { + "epoch": 2.33, + "learning_rate": 1.1101564503718903e-05, + "loss": 0.2747, + "step": 60716 + }, + { + "epoch": 2.33, + "learning_rate": 1.1100923313670172e-05, + "loss": 0.2157, + "step": 60717 + }, + { + "epoch": 2.33, + "learning_rate": 1.1100282123621442e-05, + "loss": 0.115, + "step": 60718 + }, + { + "epoch": 2.33, + "learning_rate": 1.109964093357271e-05, + "loss": 0.1538, + "step": 60719 + }, + { + "epoch": 2.33, + "learning_rate": 1.1098999743523982e-05, + "loss": 0.1958, + "step": 60720 + }, + { + "epoch": 2.33, + "learning_rate": 1.109835855347525e-05, + "loss": 0.1911, + "step": 60721 + }, + { + "epoch": 2.33, + "learning_rate": 1.109771736342652e-05, + "loss": 0.1525, + "step": 60722 + }, + { + "epoch": 2.33, + "learning_rate": 1.1097076173377788e-05, + "loss": 0.4118, + "step": 60723 + }, + { + "epoch": 2.33, + "learning_rate": 1.109643498332906e-05, + "loss": 0.2108, + "step": 60724 + }, + { + "epoch": 2.33, + "learning_rate": 1.1095793793280328e-05, + "loss": 0.1392, + "step": 60725 + }, + { + "epoch": 2.33, + "learning_rate": 1.1095152603231599e-05, + "loss": 0.1443, + "step": 60726 + }, + { + "epoch": 2.33, + "learning_rate": 1.1094511413182867e-05, + "loss": 0.3456, + "step": 60727 + }, + { + "epoch": 2.33, + "learning_rate": 1.1093870223134137e-05, + "loss": 0.2627, + "step": 60728 + }, + { + "epoch": 2.33, + "learning_rate": 1.1093229033085407e-05, + "loss": 0.1767, + "step": 60729 + }, + { + "epoch": 2.33, + "learning_rate": 1.1092587843036677e-05, + "loss": 0.2245, + "step": 60730 + }, + { + "epoch": 2.33, + "learning_rate": 1.1091946652987945e-05, + "loss": 0.1271, + "step": 60731 + }, + { + "epoch": 2.33, + "learning_rate": 1.1091305462939215e-05, + "loss": 0.1692, + "step": 60732 + }, + { + "epoch": 2.33, + "learning_rate": 1.1090664272890485e-05, + "loss": 0.2432, + "step": 60733 + }, + { + "epoch": 2.34, + "learning_rate": 1.1090023082841755e-05, + "loss": 0.1358, + "step": 60734 + }, + { + "epoch": 2.34, + "learning_rate": 1.1089381892793024e-05, + "loss": 0.2989, + "step": 60735 + }, + { + "epoch": 2.34, + "learning_rate": 1.1088740702744294e-05, + "loss": 0.3366, + "step": 60736 + }, + { + "epoch": 2.34, + "learning_rate": 1.1088099512695564e-05, + "loss": 0.2279, + "step": 60737 + }, + { + "epoch": 2.34, + "learning_rate": 1.1087458322646834e-05, + "loss": 0.2434, + "step": 60738 + }, + { + "epoch": 2.34, + "learning_rate": 1.1086817132598102e-05, + "loss": 0.2035, + "step": 60739 + }, + { + "epoch": 2.34, + "learning_rate": 1.1086175942549372e-05, + "loss": 0.1468, + "step": 60740 + }, + { + "epoch": 2.34, + "learning_rate": 1.108553475250064e-05, + "loss": 0.1633, + "step": 60741 + }, + { + "epoch": 2.34, + "learning_rate": 1.1084893562451912e-05, + "loss": 0.2571, + "step": 60742 + }, + { + "epoch": 2.34, + "learning_rate": 1.108425237240318e-05, + "loss": 0.2921, + "step": 60743 + }, + { + "epoch": 2.34, + "learning_rate": 1.108361118235445e-05, + "loss": 0.2477, + "step": 60744 + }, + { + "epoch": 2.34, + "learning_rate": 1.1082969992305719e-05, + "loss": 0.2357, + "step": 60745 + }, + { + "epoch": 2.34, + "learning_rate": 1.108232880225699e-05, + "loss": 0.2875, + "step": 60746 + }, + { + "epoch": 2.34, + "learning_rate": 1.1081687612208259e-05, + "loss": 0.1825, + "step": 60747 + }, + { + "epoch": 2.34, + "learning_rate": 1.1081046422159529e-05, + "loss": 0.1262, + "step": 60748 + }, + { + "epoch": 2.34, + "learning_rate": 1.1080405232110797e-05, + "loss": 0.1666, + "step": 60749 + }, + { + "epoch": 2.34, + "learning_rate": 1.1079764042062067e-05, + "loss": 0.2232, + "step": 60750 + }, + { + "epoch": 2.34, + "learning_rate": 1.1079122852013337e-05, + "loss": 0.1191, + "step": 60751 + }, + { + "epoch": 2.34, + "learning_rate": 1.1078481661964607e-05, + "loss": 0.2432, + "step": 60752 + }, + { + "epoch": 2.34, + "learning_rate": 1.1077840471915876e-05, + "loss": 0.1202, + "step": 60753 + }, + { + "epoch": 2.34, + "learning_rate": 1.1077199281867146e-05, + "loss": 0.2653, + "step": 60754 + }, + { + "epoch": 2.34, + "learning_rate": 1.1076558091818416e-05, + "loss": 0.2857, + "step": 60755 + }, + { + "epoch": 2.34, + "learning_rate": 1.1075916901769686e-05, + "loss": 0.3537, + "step": 60756 + }, + { + "epoch": 2.34, + "learning_rate": 1.1075275711720954e-05, + "loss": 0.115, + "step": 60757 + }, + { + "epoch": 2.34, + "learning_rate": 1.1074634521672224e-05, + "loss": 0.1002, + "step": 60758 + }, + { + "epoch": 2.34, + "learning_rate": 1.1073993331623492e-05, + "loss": 0.2785, + "step": 60759 + }, + { + "epoch": 2.34, + "learning_rate": 1.1073352141574764e-05, + "loss": 0.1996, + "step": 60760 + }, + { + "epoch": 2.34, + "learning_rate": 1.1072710951526032e-05, + "loss": 0.0846, + "step": 60761 + }, + { + "epoch": 2.34, + "learning_rate": 1.1072069761477302e-05, + "loss": 0.1877, + "step": 60762 + }, + { + "epoch": 2.34, + "learning_rate": 1.107142857142857e-05, + "loss": 0.2191, + "step": 60763 + }, + { + "epoch": 2.34, + "learning_rate": 1.1070787381379842e-05, + "loss": 0.1239, + "step": 60764 + }, + { + "epoch": 2.34, + "learning_rate": 1.107014619133111e-05, + "loss": 0.249, + "step": 60765 + }, + { + "epoch": 2.34, + "learning_rate": 1.106950500128238e-05, + "loss": 0.3114, + "step": 60766 + }, + { + "epoch": 2.34, + "learning_rate": 1.1068863811233649e-05, + "loss": 0.1856, + "step": 60767 + }, + { + "epoch": 2.34, + "learning_rate": 1.1068222621184919e-05, + "loss": 0.2385, + "step": 60768 + }, + { + "epoch": 2.34, + "learning_rate": 1.1067581431136189e-05, + "loss": 0.3822, + "step": 60769 + }, + { + "epoch": 2.34, + "learning_rate": 1.106694024108746e-05, + "loss": 0.2575, + "step": 60770 + }, + { + "epoch": 2.34, + "learning_rate": 1.1066299051038727e-05, + "loss": 0.204, + "step": 60771 + }, + { + "epoch": 2.34, + "learning_rate": 1.1065657860989998e-05, + "loss": 0.1937, + "step": 60772 + }, + { + "epoch": 2.34, + "learning_rate": 1.1065016670941268e-05, + "loss": 0.198, + "step": 60773 + }, + { + "epoch": 2.34, + "learning_rate": 1.1064375480892538e-05, + "loss": 0.1984, + "step": 60774 + }, + { + "epoch": 2.34, + "learning_rate": 1.1063734290843806e-05, + "loss": 0.2638, + "step": 60775 + }, + { + "epoch": 2.34, + "learning_rate": 1.1063093100795076e-05, + "loss": 0.1886, + "step": 60776 + }, + { + "epoch": 2.34, + "learning_rate": 1.1062451910746344e-05, + "loss": 0.1567, + "step": 60777 + }, + { + "epoch": 2.34, + "learning_rate": 1.1061810720697616e-05, + "loss": 0.1874, + "step": 60778 + }, + { + "epoch": 2.34, + "learning_rate": 1.1061169530648884e-05, + "loss": 0.1485, + "step": 60779 + }, + { + "epoch": 2.34, + "learning_rate": 1.1060528340600154e-05, + "loss": 0.236, + "step": 60780 + }, + { + "epoch": 2.34, + "learning_rate": 1.1059887150551423e-05, + "loss": 0.2814, + "step": 60781 + }, + { + "epoch": 2.34, + "learning_rate": 1.1059245960502694e-05, + "loss": 0.2575, + "step": 60782 + }, + { + "epoch": 2.34, + "learning_rate": 1.1058604770453963e-05, + "loss": 0.2049, + "step": 60783 + }, + { + "epoch": 2.34, + "learning_rate": 1.1057963580405233e-05, + "loss": 0.2259, + "step": 60784 + }, + { + "epoch": 2.34, + "learning_rate": 1.1057322390356501e-05, + "loss": 0.2606, + "step": 60785 + }, + { + "epoch": 2.34, + "learning_rate": 1.1056681200307773e-05, + "loss": 0.2059, + "step": 60786 + }, + { + "epoch": 2.34, + "learning_rate": 1.1056040010259041e-05, + "loss": 0.2166, + "step": 60787 + }, + { + "epoch": 2.34, + "learning_rate": 1.1055398820210311e-05, + "loss": 0.1275, + "step": 60788 + }, + { + "epoch": 2.34, + "learning_rate": 1.105475763016158e-05, + "loss": 0.2255, + "step": 60789 + }, + { + "epoch": 2.34, + "learning_rate": 1.105411644011285e-05, + "loss": 0.2831, + "step": 60790 + }, + { + "epoch": 2.34, + "learning_rate": 1.105347525006412e-05, + "loss": 0.2082, + "step": 60791 + }, + { + "epoch": 2.34, + "learning_rate": 1.105283406001539e-05, + "loss": 0.1982, + "step": 60792 + }, + { + "epoch": 2.34, + "learning_rate": 1.1052192869966658e-05, + "loss": 0.3105, + "step": 60793 + }, + { + "epoch": 2.34, + "learning_rate": 1.1051551679917928e-05, + "loss": 0.2229, + "step": 60794 + }, + { + "epoch": 2.34, + "learning_rate": 1.1050910489869198e-05, + "loss": 0.3781, + "step": 60795 + }, + { + "epoch": 2.34, + "learning_rate": 1.1050269299820468e-05, + "loss": 0.1331, + "step": 60796 + }, + { + "epoch": 2.34, + "learning_rate": 1.1049628109771736e-05, + "loss": 0.3229, + "step": 60797 + }, + { + "epoch": 2.34, + "learning_rate": 1.1048986919723006e-05, + "loss": 0.2445, + "step": 60798 + }, + { + "epoch": 2.34, + "learning_rate": 1.1048345729674275e-05, + "loss": 0.3075, + "step": 60799 + }, + { + "epoch": 2.34, + "learning_rate": 1.1047704539625546e-05, + "loss": 0.169, + "step": 60800 + }, + { + "epoch": 2.34, + "learning_rate": 1.1047063349576815e-05, + "loss": 0.2426, + "step": 60801 + }, + { + "epoch": 2.34, + "learning_rate": 1.1046422159528085e-05, + "loss": 0.1842, + "step": 60802 + }, + { + "epoch": 2.34, + "learning_rate": 1.1045780969479353e-05, + "loss": 0.3423, + "step": 60803 + }, + { + "epoch": 2.34, + "learning_rate": 1.1045139779430625e-05, + "loss": 0.3929, + "step": 60804 + }, + { + "epoch": 2.34, + "learning_rate": 1.1044498589381893e-05, + "loss": 0.1536, + "step": 60805 + }, + { + "epoch": 2.34, + "learning_rate": 1.1043857399333163e-05, + "loss": 0.1588, + "step": 60806 + }, + { + "epoch": 2.34, + "learning_rate": 1.1043216209284431e-05, + "loss": 0.1683, + "step": 60807 + }, + { + "epoch": 2.34, + "learning_rate": 1.1042575019235701e-05, + "loss": 0.2406, + "step": 60808 + }, + { + "epoch": 2.34, + "learning_rate": 1.1041933829186971e-05, + "loss": 0.1789, + "step": 60809 + }, + { + "epoch": 2.34, + "learning_rate": 1.1041292639138241e-05, + "loss": 0.2585, + "step": 60810 + }, + { + "epoch": 2.34, + "learning_rate": 1.104065144908951e-05, + "loss": 0.1815, + "step": 60811 + }, + { + "epoch": 2.34, + "learning_rate": 1.104001025904078e-05, + "loss": 0.2202, + "step": 60812 + }, + { + "epoch": 2.34, + "learning_rate": 1.103936906899205e-05, + "loss": 0.3557, + "step": 60813 + }, + { + "epoch": 2.34, + "learning_rate": 1.103872787894332e-05, + "loss": 0.1788, + "step": 60814 + }, + { + "epoch": 2.34, + "learning_rate": 1.1038086688894588e-05, + "loss": 0.3555, + "step": 60815 + }, + { + "epoch": 2.34, + "learning_rate": 1.1037445498845858e-05, + "loss": 0.4018, + "step": 60816 + }, + { + "epoch": 2.34, + "learning_rate": 1.1036804308797127e-05, + "loss": 0.4276, + "step": 60817 + }, + { + "epoch": 2.34, + "learning_rate": 1.1036163118748398e-05, + "loss": 0.2039, + "step": 60818 + }, + { + "epoch": 2.34, + "learning_rate": 1.1035521928699667e-05, + "loss": 0.2627, + "step": 60819 + }, + { + "epoch": 2.34, + "learning_rate": 1.1034880738650937e-05, + "loss": 0.1612, + "step": 60820 + }, + { + "epoch": 2.34, + "learning_rate": 1.1034239548602205e-05, + "loss": 0.3543, + "step": 60821 + }, + { + "epoch": 2.34, + "learning_rate": 1.1033598358553477e-05, + "loss": 0.2447, + "step": 60822 + }, + { + "epoch": 2.34, + "learning_rate": 1.1032957168504745e-05, + "loss": 0.2996, + "step": 60823 + }, + { + "epoch": 2.34, + "learning_rate": 1.1032315978456015e-05, + "loss": 0.2891, + "step": 60824 + }, + { + "epoch": 2.34, + "learning_rate": 1.1031674788407283e-05, + "loss": 0.1206, + "step": 60825 + }, + { + "epoch": 2.34, + "learning_rate": 1.1031033598358553e-05, + "loss": 0.2006, + "step": 60826 + }, + { + "epoch": 2.34, + "learning_rate": 1.1030392408309823e-05, + "loss": 0.3373, + "step": 60827 + }, + { + "epoch": 2.34, + "learning_rate": 1.1029751218261093e-05, + "loss": 0.134, + "step": 60828 + }, + { + "epoch": 2.34, + "learning_rate": 1.1029110028212362e-05, + "loss": 0.2544, + "step": 60829 + }, + { + "epoch": 2.34, + "learning_rate": 1.1028468838163632e-05, + "loss": 0.2374, + "step": 60830 + }, + { + "epoch": 2.34, + "learning_rate": 1.1027827648114902e-05, + "loss": 0.2109, + "step": 60831 + }, + { + "epoch": 2.34, + "learning_rate": 1.1027186458066172e-05, + "loss": 0.1842, + "step": 60832 + }, + { + "epoch": 2.34, + "learning_rate": 1.102654526801744e-05, + "loss": 0.1894, + "step": 60833 + }, + { + "epoch": 2.34, + "learning_rate": 1.102590407796871e-05, + "loss": 0.1896, + "step": 60834 + }, + { + "epoch": 2.34, + "learning_rate": 1.102526288791998e-05, + "loss": 0.1364, + "step": 60835 + }, + { + "epoch": 2.34, + "learning_rate": 1.102462169787125e-05, + "loss": 0.2057, + "step": 60836 + }, + { + "epoch": 2.34, + "learning_rate": 1.1023980507822518e-05, + "loss": 0.1404, + "step": 60837 + }, + { + "epoch": 2.34, + "learning_rate": 1.1023339317773789e-05, + "loss": 0.2602, + "step": 60838 + }, + { + "epoch": 2.34, + "learning_rate": 1.1022698127725057e-05, + "loss": 0.2954, + "step": 60839 + }, + { + "epoch": 2.34, + "learning_rate": 1.1022056937676329e-05, + "loss": 0.245, + "step": 60840 + }, + { + "epoch": 2.34, + "learning_rate": 1.1021415747627597e-05, + "loss": 0.2113, + "step": 60841 + }, + { + "epoch": 2.34, + "learning_rate": 1.1020774557578867e-05, + "loss": 0.2765, + "step": 60842 + }, + { + "epoch": 2.34, + "learning_rate": 1.1020133367530135e-05, + "loss": 0.2441, + "step": 60843 + }, + { + "epoch": 2.34, + "learning_rate": 1.1019492177481407e-05, + "loss": 0.2147, + "step": 60844 + }, + { + "epoch": 2.34, + "learning_rate": 1.1018850987432675e-05, + "loss": 0.2262, + "step": 60845 + }, + { + "epoch": 2.34, + "learning_rate": 1.1018209797383945e-05, + "loss": 0.1208, + "step": 60846 + }, + { + "epoch": 2.34, + "learning_rate": 1.1017568607335214e-05, + "loss": 0.2762, + "step": 60847 + }, + { + "epoch": 2.34, + "learning_rate": 1.1016927417286484e-05, + "loss": 0.1623, + "step": 60848 + }, + { + "epoch": 2.34, + "learning_rate": 1.1016286227237754e-05, + "loss": 0.1931, + "step": 60849 + }, + { + "epoch": 2.34, + "learning_rate": 1.1015645037189024e-05, + "loss": 0.2483, + "step": 60850 + }, + { + "epoch": 2.34, + "learning_rate": 1.1015003847140292e-05, + "loss": 0.4295, + "step": 60851 + }, + { + "epoch": 2.34, + "learning_rate": 1.1014362657091562e-05, + "loss": 0.1941, + "step": 60852 + }, + { + "epoch": 2.34, + "learning_rate": 1.1013721467042832e-05, + "loss": 0.1745, + "step": 60853 + }, + { + "epoch": 2.34, + "learning_rate": 1.1013080276994102e-05, + "loss": 0.2006, + "step": 60854 + }, + { + "epoch": 2.34, + "learning_rate": 1.101243908694537e-05, + "loss": 0.2185, + "step": 60855 + }, + { + "epoch": 2.34, + "learning_rate": 1.101179789689664e-05, + "loss": 0.1515, + "step": 60856 + }, + { + "epoch": 2.34, + "learning_rate": 1.1011156706847909e-05, + "loss": 0.2147, + "step": 60857 + }, + { + "epoch": 2.34, + "learning_rate": 1.101051551679918e-05, + "loss": 0.1999, + "step": 60858 + }, + { + "epoch": 2.34, + "learning_rate": 1.1009874326750449e-05, + "loss": 0.2957, + "step": 60859 + }, + { + "epoch": 2.34, + "learning_rate": 1.1009233136701719e-05, + "loss": 0.2405, + "step": 60860 + }, + { + "epoch": 2.34, + "learning_rate": 1.1008591946652987e-05, + "loss": 0.222, + "step": 60861 + }, + { + "epoch": 2.34, + "learning_rate": 1.1007950756604259e-05, + "loss": 0.0989, + "step": 60862 + }, + { + "epoch": 2.34, + "learning_rate": 1.1007309566555527e-05, + "loss": 0.252, + "step": 60863 + }, + { + "epoch": 2.34, + "learning_rate": 1.1006668376506797e-05, + "loss": 0.296, + "step": 60864 + }, + { + "epoch": 2.34, + "learning_rate": 1.1006027186458066e-05, + "loss": 0.1769, + "step": 60865 + }, + { + "epoch": 2.34, + "learning_rate": 1.1005385996409336e-05, + "loss": 0.2002, + "step": 60866 + }, + { + "epoch": 2.34, + "learning_rate": 1.1004744806360606e-05, + "loss": 0.1685, + "step": 60867 + }, + { + "epoch": 2.34, + "learning_rate": 1.1004103616311876e-05, + "loss": 0.4449, + "step": 60868 + }, + { + "epoch": 2.34, + "learning_rate": 1.1003462426263144e-05, + "loss": 0.1799, + "step": 60869 + }, + { + "epoch": 2.34, + "learning_rate": 1.1002821236214414e-05, + "loss": 0.2635, + "step": 60870 + }, + { + "epoch": 2.34, + "learning_rate": 1.1002180046165684e-05, + "loss": 0.3169, + "step": 60871 + }, + { + "epoch": 2.34, + "learning_rate": 1.1001538856116954e-05, + "loss": 0.2695, + "step": 60872 + }, + { + "epoch": 2.34, + "learning_rate": 1.1000897666068222e-05, + "loss": 0.2506, + "step": 60873 + }, + { + "epoch": 2.34, + "learning_rate": 1.1000256476019492e-05, + "loss": 0.1663, + "step": 60874 + }, + { + "epoch": 2.34, + "learning_rate": 1.0999615285970762e-05, + "loss": 0.2719, + "step": 60875 + }, + { + "epoch": 2.34, + "learning_rate": 1.0998974095922032e-05, + "loss": 0.2395, + "step": 60876 + }, + { + "epoch": 2.34, + "learning_rate": 1.09983329058733e-05, + "loss": 0.2286, + "step": 60877 + }, + { + "epoch": 2.34, + "learning_rate": 1.099769171582457e-05, + "loss": 0.2002, + "step": 60878 + }, + { + "epoch": 2.34, + "learning_rate": 1.0997050525775839e-05, + "loss": 0.2904, + "step": 60879 + }, + { + "epoch": 2.34, + "learning_rate": 1.099640933572711e-05, + "loss": 0.1722, + "step": 60880 + }, + { + "epoch": 2.34, + "learning_rate": 1.0995768145678379e-05, + "loss": 0.2164, + "step": 60881 + }, + { + "epoch": 2.34, + "learning_rate": 1.099512695562965e-05, + "loss": 0.2547, + "step": 60882 + }, + { + "epoch": 2.34, + "learning_rate": 1.0994485765580918e-05, + "loss": 0.2235, + "step": 60883 + }, + { + "epoch": 2.34, + "learning_rate": 1.099384457553219e-05, + "loss": 0.2963, + "step": 60884 + }, + { + "epoch": 2.34, + "learning_rate": 1.0993203385483458e-05, + "loss": 0.123, + "step": 60885 + }, + { + "epoch": 2.34, + "learning_rate": 1.0992562195434728e-05, + "loss": 0.1125, + "step": 60886 + }, + { + "epoch": 2.34, + "learning_rate": 1.0991921005385996e-05, + "loss": 0.3075, + "step": 60887 + }, + { + "epoch": 2.34, + "learning_rate": 1.0991279815337266e-05, + "loss": 0.1796, + "step": 60888 + }, + { + "epoch": 2.34, + "learning_rate": 1.0990638625288536e-05, + "loss": 0.2801, + "step": 60889 + }, + { + "epoch": 2.34, + "learning_rate": 1.0989997435239806e-05, + "loss": 0.2545, + "step": 60890 + }, + { + "epoch": 2.34, + "learning_rate": 1.0989356245191074e-05, + "loss": 0.1296, + "step": 60891 + }, + { + "epoch": 2.34, + "learning_rate": 1.0988715055142344e-05, + "loss": 0.3275, + "step": 60892 + }, + { + "epoch": 2.34, + "learning_rate": 1.0988073865093614e-05, + "loss": 0.2896, + "step": 60893 + }, + { + "epoch": 2.34, + "learning_rate": 1.0987432675044884e-05, + "loss": 0.2076, + "step": 60894 + }, + { + "epoch": 2.34, + "learning_rate": 1.0986791484996153e-05, + "loss": 0.1475, + "step": 60895 + }, + { + "epoch": 2.34, + "learning_rate": 1.0986150294947423e-05, + "loss": 0.2583, + "step": 60896 + }, + { + "epoch": 2.34, + "learning_rate": 1.0985509104898693e-05, + "loss": 0.1927, + "step": 60897 + }, + { + "epoch": 2.34, + "learning_rate": 1.0984867914849963e-05, + "loss": 0.201, + "step": 60898 + }, + { + "epoch": 2.34, + "learning_rate": 1.0984226724801231e-05, + "loss": 0.2845, + "step": 60899 + }, + { + "epoch": 2.34, + "learning_rate": 1.0983585534752501e-05, + "loss": 0.1676, + "step": 60900 + }, + { + "epoch": 2.34, + "learning_rate": 1.098294434470377e-05, + "loss": 0.1771, + "step": 60901 + }, + { + "epoch": 2.34, + "learning_rate": 1.0982303154655041e-05, + "loss": 0.2772, + "step": 60902 + }, + { + "epoch": 2.34, + "learning_rate": 1.098166196460631e-05, + "loss": 0.2577, + "step": 60903 + }, + { + "epoch": 2.34, + "learning_rate": 1.098102077455758e-05, + "loss": 0.2459, + "step": 60904 + }, + { + "epoch": 2.34, + "learning_rate": 1.0980379584508848e-05, + "loss": 0.2016, + "step": 60905 + }, + { + "epoch": 2.34, + "learning_rate": 1.0979738394460118e-05, + "loss": 0.1801, + "step": 60906 + }, + { + "epoch": 2.34, + "learning_rate": 1.0979097204411388e-05, + "loss": 0.3108, + "step": 60907 + }, + { + "epoch": 2.34, + "learning_rate": 1.0978456014362658e-05, + "loss": 0.2732, + "step": 60908 + }, + { + "epoch": 2.34, + "learning_rate": 1.0977814824313926e-05, + "loss": 0.2406, + "step": 60909 + }, + { + "epoch": 2.34, + "learning_rate": 1.0977173634265196e-05, + "loss": 0.2744, + "step": 60910 + }, + { + "epoch": 2.34, + "learning_rate": 1.0976532444216466e-05, + "loss": 0.2001, + "step": 60911 + }, + { + "epoch": 2.34, + "learning_rate": 1.0975891254167736e-05, + "loss": 0.2282, + "step": 60912 + }, + { + "epoch": 2.34, + "learning_rate": 1.0975250064119005e-05, + "loss": 0.2908, + "step": 60913 + }, + { + "epoch": 2.34, + "learning_rate": 1.0974608874070275e-05, + "loss": 0.2646, + "step": 60914 + }, + { + "epoch": 2.34, + "learning_rate": 1.0973967684021545e-05, + "loss": 0.138, + "step": 60915 + }, + { + "epoch": 2.34, + "learning_rate": 1.0973326493972815e-05, + "loss": 0.2713, + "step": 60916 + }, + { + "epoch": 2.34, + "learning_rate": 1.0972685303924083e-05, + "loss": 0.2432, + "step": 60917 + }, + { + "epoch": 2.34, + "learning_rate": 1.0972044113875353e-05, + "loss": 0.3607, + "step": 60918 + }, + { + "epoch": 2.34, + "learning_rate": 1.0971402923826621e-05, + "loss": 0.1767, + "step": 60919 + }, + { + "epoch": 2.34, + "learning_rate": 1.0970761733777893e-05, + "loss": 0.1854, + "step": 60920 + }, + { + "epoch": 2.34, + "learning_rate": 1.0970120543729161e-05, + "loss": 0.2747, + "step": 60921 + }, + { + "epoch": 2.34, + "learning_rate": 1.0969479353680431e-05, + "loss": 0.2236, + "step": 60922 + }, + { + "epoch": 2.34, + "learning_rate": 1.09688381636317e-05, + "loss": 0.1991, + "step": 60923 + }, + { + "epoch": 2.34, + "learning_rate": 1.096819697358297e-05, + "loss": 0.3596, + "step": 60924 + }, + { + "epoch": 2.34, + "learning_rate": 1.096755578353424e-05, + "loss": 0.2583, + "step": 60925 + }, + { + "epoch": 2.34, + "learning_rate": 1.096691459348551e-05, + "loss": 0.1755, + "step": 60926 + }, + { + "epoch": 2.34, + "learning_rate": 1.0966273403436778e-05, + "loss": 0.2299, + "step": 60927 + }, + { + "epoch": 2.34, + "learning_rate": 1.0965632213388048e-05, + "loss": 0.2166, + "step": 60928 + }, + { + "epoch": 2.34, + "learning_rate": 1.0964991023339318e-05, + "loss": 0.3429, + "step": 60929 + }, + { + "epoch": 2.34, + "learning_rate": 1.0964349833290588e-05, + "loss": 0.247, + "step": 60930 + }, + { + "epoch": 2.34, + "learning_rate": 1.0963708643241857e-05, + "loss": 0.1745, + "step": 60931 + }, + { + "epoch": 2.34, + "learning_rate": 1.0963067453193127e-05, + "loss": 0.1784, + "step": 60932 + }, + { + "epoch": 2.34, + "learning_rate": 1.0962426263144397e-05, + "loss": 0.204, + "step": 60933 + }, + { + "epoch": 2.34, + "learning_rate": 1.0961785073095667e-05, + "loss": 0.4208, + "step": 60934 + }, + { + "epoch": 2.34, + "learning_rate": 1.0961143883046935e-05, + "loss": 0.1188, + "step": 60935 + }, + { + "epoch": 2.34, + "learning_rate": 1.0960502692998205e-05, + "loss": 0.3042, + "step": 60936 + }, + { + "epoch": 2.34, + "learning_rate": 1.0959861502949475e-05, + "loss": 0.2183, + "step": 60937 + }, + { + "epoch": 2.34, + "learning_rate": 1.0959220312900745e-05, + "loss": 0.1753, + "step": 60938 + }, + { + "epoch": 2.34, + "learning_rate": 1.0958579122852013e-05, + "loss": 0.2443, + "step": 60939 + }, + { + "epoch": 2.34, + "learning_rate": 1.0957937932803283e-05, + "loss": 0.2235, + "step": 60940 + }, + { + "epoch": 2.34, + "learning_rate": 1.0957296742754552e-05, + "loss": 0.2873, + "step": 60941 + }, + { + "epoch": 2.34, + "learning_rate": 1.0956655552705823e-05, + "loss": 0.1763, + "step": 60942 + }, + { + "epoch": 2.34, + "learning_rate": 1.0956014362657092e-05, + "loss": 0.129, + "step": 60943 + }, + { + "epoch": 2.34, + "learning_rate": 1.0955373172608362e-05, + "loss": 0.2188, + "step": 60944 + }, + { + "epoch": 2.34, + "learning_rate": 1.095473198255963e-05, + "loss": 0.2052, + "step": 60945 + }, + { + "epoch": 2.34, + "learning_rate": 1.09540907925109e-05, + "loss": 0.2162, + "step": 60946 + }, + { + "epoch": 2.34, + "learning_rate": 1.095344960246217e-05, + "loss": 0.1996, + "step": 60947 + }, + { + "epoch": 2.34, + "learning_rate": 1.095280841241344e-05, + "loss": 0.2823, + "step": 60948 + }, + { + "epoch": 2.34, + "learning_rate": 1.0952167222364708e-05, + "loss": 0.2041, + "step": 60949 + }, + { + "epoch": 2.34, + "learning_rate": 1.0951526032315979e-05, + "loss": 0.1296, + "step": 60950 + }, + { + "epoch": 2.34, + "learning_rate": 1.0950884842267249e-05, + "loss": 0.3286, + "step": 60951 + }, + { + "epoch": 2.34, + "learning_rate": 1.0950243652218519e-05, + "loss": 0.1647, + "step": 60952 + }, + { + "epoch": 2.34, + "learning_rate": 1.0949602462169787e-05, + "loss": 0.2516, + "step": 60953 + }, + { + "epoch": 2.34, + "learning_rate": 1.0948961272121057e-05, + "loss": 0.2643, + "step": 60954 + }, + { + "epoch": 2.34, + "learning_rate": 1.0948320082072327e-05, + "loss": 0.2442, + "step": 60955 + }, + { + "epoch": 2.34, + "learning_rate": 1.0947678892023597e-05, + "loss": 0.2182, + "step": 60956 + }, + { + "epoch": 2.34, + "learning_rate": 1.0947037701974865e-05, + "loss": 0.242, + "step": 60957 + }, + { + "epoch": 2.34, + "learning_rate": 1.0946396511926135e-05, + "loss": 0.2702, + "step": 60958 + }, + { + "epoch": 2.34, + "learning_rate": 1.0945755321877405e-05, + "loss": 0.2666, + "step": 60959 + }, + { + "epoch": 2.34, + "learning_rate": 1.0945114131828675e-05, + "loss": 0.3458, + "step": 60960 + }, + { + "epoch": 2.34, + "learning_rate": 1.0944472941779944e-05, + "loss": 0.1994, + "step": 60961 + }, + { + "epoch": 2.34, + "learning_rate": 1.0943831751731214e-05, + "loss": 0.1781, + "step": 60962 + }, + { + "epoch": 2.34, + "learning_rate": 1.0943190561682482e-05, + "loss": 0.2359, + "step": 60963 + }, + { + "epoch": 2.34, + "learning_rate": 1.0942549371633752e-05, + "loss": 0.3457, + "step": 60964 + }, + { + "epoch": 2.34, + "learning_rate": 1.0941908181585022e-05, + "loss": 0.3133, + "step": 60965 + }, + { + "epoch": 2.34, + "learning_rate": 1.0941266991536292e-05, + "loss": 0.2552, + "step": 60966 + }, + { + "epoch": 2.34, + "learning_rate": 1.094062580148756e-05, + "loss": 0.2057, + "step": 60967 + }, + { + "epoch": 2.34, + "learning_rate": 1.093998461143883e-05, + "loss": 0.2685, + "step": 60968 + }, + { + "epoch": 2.34, + "learning_rate": 1.09393434213901e-05, + "loss": 0.207, + "step": 60969 + }, + { + "epoch": 2.34, + "learning_rate": 1.093870223134137e-05, + "loss": 0.1599, + "step": 60970 + }, + { + "epoch": 2.34, + "learning_rate": 1.0938061041292639e-05, + "loss": 0.171, + "step": 60971 + }, + { + "epoch": 2.34, + "learning_rate": 1.0937419851243909e-05, + "loss": 0.2376, + "step": 60972 + }, + { + "epoch": 2.34, + "learning_rate": 1.0936778661195179e-05, + "loss": 0.1252, + "step": 60973 + }, + { + "epoch": 2.34, + "learning_rate": 1.0936137471146449e-05, + "loss": 0.1782, + "step": 60974 + }, + { + "epoch": 2.34, + "learning_rate": 1.0935496281097717e-05, + "loss": 0.2714, + "step": 60975 + }, + { + "epoch": 2.34, + "learning_rate": 1.0934855091048987e-05, + "loss": 0.2235, + "step": 60976 + }, + { + "epoch": 2.34, + "learning_rate": 1.0934213901000257e-05, + "loss": 0.2025, + "step": 60977 + }, + { + "epoch": 2.34, + "learning_rate": 1.0933572710951527e-05, + "loss": 0.2123, + "step": 60978 + }, + { + "epoch": 2.34, + "learning_rate": 1.0932931520902796e-05, + "loss": 0.3139, + "step": 60979 + }, + { + "epoch": 2.34, + "learning_rate": 1.0932290330854066e-05, + "loss": 0.2821, + "step": 60980 + }, + { + "epoch": 2.34, + "learning_rate": 1.0931649140805334e-05, + "loss": 0.1798, + "step": 60981 + }, + { + "epoch": 2.34, + "learning_rate": 1.0931007950756606e-05, + "loss": 0.2911, + "step": 60982 + }, + { + "epoch": 2.34, + "learning_rate": 1.0930366760707874e-05, + "loss": 0.3, + "step": 60983 + }, + { + "epoch": 2.34, + "learning_rate": 1.0929725570659144e-05, + "loss": 0.0813, + "step": 60984 + }, + { + "epoch": 2.34, + "learning_rate": 1.0929084380610412e-05, + "loss": 0.0969, + "step": 60985 + }, + { + "epoch": 2.34, + "learning_rate": 1.0928443190561682e-05, + "loss": 0.2231, + "step": 60986 + }, + { + "epoch": 2.34, + "learning_rate": 1.0927802000512952e-05, + "loss": 0.1326, + "step": 60987 + }, + { + "epoch": 2.34, + "learning_rate": 1.0927160810464222e-05, + "loss": 0.2647, + "step": 60988 + }, + { + "epoch": 2.34, + "learning_rate": 1.092651962041549e-05, + "loss": 0.2996, + "step": 60989 + }, + { + "epoch": 2.34, + "learning_rate": 1.092587843036676e-05, + "loss": 0.1899, + "step": 60990 + }, + { + "epoch": 2.34, + "learning_rate": 1.092523724031803e-05, + "loss": 0.1546, + "step": 60991 + }, + { + "epoch": 2.34, + "learning_rate": 1.09245960502693e-05, + "loss": 0.3048, + "step": 60992 + }, + { + "epoch": 2.34, + "learning_rate": 1.092395486022057e-05, + "loss": 0.2726, + "step": 60993 + }, + { + "epoch": 2.35, + "learning_rate": 1.092331367017184e-05, + "loss": 0.2013, + "step": 60994 + }, + { + "epoch": 2.35, + "learning_rate": 1.092267248012311e-05, + "loss": 0.1675, + "step": 60995 + }, + { + "epoch": 2.35, + "learning_rate": 1.092203129007438e-05, + "loss": 0.2763, + "step": 60996 + }, + { + "epoch": 2.35, + "learning_rate": 1.0921390100025648e-05, + "loss": 0.1998, + "step": 60997 + }, + { + "epoch": 2.35, + "learning_rate": 1.0920748909976918e-05, + "loss": 0.3287, + "step": 60998 + }, + { + "epoch": 2.35, + "learning_rate": 1.0920107719928188e-05, + "loss": 0.2356, + "step": 60999 + }, + { + "epoch": 2.35, + "learning_rate": 1.0919466529879458e-05, + "loss": 0.2309, + "step": 61000 + }, + { + "epoch": 2.35, + "learning_rate": 1.0918825339830726e-05, + "loss": 0.2079, + "step": 61001 + }, + { + "epoch": 2.35, + "learning_rate": 1.0918184149781996e-05, + "loss": 0.2333, + "step": 61002 + }, + { + "epoch": 2.35, + "learning_rate": 1.0917542959733264e-05, + "loss": 0.2434, + "step": 61003 + }, + { + "epoch": 2.35, + "learning_rate": 1.0916901769684534e-05, + "loss": 0.2969, + "step": 61004 + }, + { + "epoch": 2.35, + "learning_rate": 1.0916260579635804e-05, + "loss": 0.2259, + "step": 61005 + }, + { + "epoch": 2.35, + "learning_rate": 1.0915619389587074e-05, + "loss": 0.1342, + "step": 61006 + }, + { + "epoch": 2.35, + "learning_rate": 1.0914978199538343e-05, + "loss": 0.179, + "step": 61007 + }, + { + "epoch": 2.35, + "learning_rate": 1.0914337009489613e-05, + "loss": 0.4136, + "step": 61008 + }, + { + "epoch": 2.35, + "learning_rate": 1.0913695819440883e-05, + "loss": 0.205, + "step": 61009 + }, + { + "epoch": 2.35, + "learning_rate": 1.0913054629392153e-05, + "loss": 0.1661, + "step": 61010 + }, + { + "epoch": 2.35, + "learning_rate": 1.0912413439343421e-05, + "loss": 0.2175, + "step": 61011 + }, + { + "epoch": 2.35, + "learning_rate": 1.0911772249294691e-05, + "loss": 0.2516, + "step": 61012 + }, + { + "epoch": 2.35, + "learning_rate": 1.0911131059245961e-05, + "loss": 0.1948, + "step": 61013 + }, + { + "epoch": 2.35, + "learning_rate": 1.0910489869197231e-05, + "loss": 0.1795, + "step": 61014 + }, + { + "epoch": 2.35, + "learning_rate": 1.09098486791485e-05, + "loss": 0.2967, + "step": 61015 + }, + { + "epoch": 2.35, + "learning_rate": 1.090920748909977e-05, + "loss": 0.2087, + "step": 61016 + }, + { + "epoch": 2.35, + "learning_rate": 1.090856629905104e-05, + "loss": 0.244, + "step": 61017 + }, + { + "epoch": 2.35, + "learning_rate": 1.090792510900231e-05, + "loss": 0.2088, + "step": 61018 + }, + { + "epoch": 2.35, + "learning_rate": 1.0907283918953578e-05, + "loss": 0.1389, + "step": 61019 + }, + { + "epoch": 2.35, + "learning_rate": 1.0906642728904848e-05, + "loss": 0.2236, + "step": 61020 + }, + { + "epoch": 2.35, + "learning_rate": 1.0906001538856116e-05, + "loss": 0.3361, + "step": 61021 + }, + { + "epoch": 2.35, + "learning_rate": 1.0905360348807388e-05, + "loss": 0.0851, + "step": 61022 + }, + { + "epoch": 2.35, + "learning_rate": 1.0904719158758656e-05, + "loss": 0.1675, + "step": 61023 + }, + { + "epoch": 2.35, + "learning_rate": 1.0904077968709926e-05, + "loss": 0.2278, + "step": 61024 + }, + { + "epoch": 2.35, + "learning_rate": 1.0903436778661195e-05, + "loss": 0.1895, + "step": 61025 + }, + { + "epoch": 2.35, + "learning_rate": 1.0902795588612465e-05, + "loss": 0.3905, + "step": 61026 + }, + { + "epoch": 2.35, + "learning_rate": 1.0902154398563735e-05, + "loss": 0.2046, + "step": 61027 + }, + { + "epoch": 2.35, + "learning_rate": 1.0901513208515005e-05, + "loss": 0.1345, + "step": 61028 + }, + { + "epoch": 2.35, + "learning_rate": 1.0900872018466273e-05, + "loss": 0.3038, + "step": 61029 + }, + { + "epoch": 2.35, + "learning_rate": 1.0900230828417543e-05, + "loss": 0.1958, + "step": 61030 + }, + { + "epoch": 2.35, + "learning_rate": 1.0899589638368813e-05, + "loss": 0.3507, + "step": 61031 + }, + { + "epoch": 2.35, + "learning_rate": 1.0898948448320083e-05, + "loss": 0.2487, + "step": 61032 + }, + { + "epoch": 2.35, + "learning_rate": 1.0898307258271351e-05, + "loss": 0.376, + "step": 61033 + }, + { + "epoch": 2.35, + "learning_rate": 1.0897666068222621e-05, + "loss": 0.2206, + "step": 61034 + }, + { + "epoch": 2.35, + "learning_rate": 1.0897024878173891e-05, + "loss": 0.3253, + "step": 61035 + }, + { + "epoch": 2.35, + "learning_rate": 1.0896383688125161e-05, + "loss": 0.3195, + "step": 61036 + }, + { + "epoch": 2.35, + "learning_rate": 1.089574249807643e-05, + "loss": 0.2284, + "step": 61037 + }, + { + "epoch": 2.35, + "learning_rate": 1.08951013080277e-05, + "loss": 0.1957, + "step": 61038 + }, + { + "epoch": 2.35, + "learning_rate": 1.089446011797897e-05, + "loss": 0.0808, + "step": 61039 + }, + { + "epoch": 2.35, + "learning_rate": 1.089381892793024e-05, + "loss": 0.2051, + "step": 61040 + }, + { + "epoch": 2.35, + "learning_rate": 1.0893177737881508e-05, + "loss": 0.1282, + "step": 61041 + }, + { + "epoch": 2.35, + "learning_rate": 1.0892536547832778e-05, + "loss": 0.1365, + "step": 61042 + }, + { + "epoch": 2.35, + "learning_rate": 1.0891895357784047e-05, + "loss": 0.1869, + "step": 61043 + }, + { + "epoch": 2.35, + "learning_rate": 1.0891254167735317e-05, + "loss": 0.22, + "step": 61044 + }, + { + "epoch": 2.35, + "learning_rate": 1.0890612977686587e-05, + "loss": 0.2626, + "step": 61045 + }, + { + "epoch": 2.35, + "learning_rate": 1.0889971787637857e-05, + "loss": 0.2207, + "step": 61046 + }, + { + "epoch": 2.35, + "learning_rate": 1.0889330597589125e-05, + "loss": 0.1531, + "step": 61047 + }, + { + "epoch": 2.35, + "learning_rate": 1.0888689407540395e-05, + "loss": 0.1736, + "step": 61048 + }, + { + "epoch": 2.35, + "learning_rate": 1.0888048217491665e-05, + "loss": 0.0693, + "step": 61049 + }, + { + "epoch": 2.35, + "learning_rate": 1.0887407027442935e-05, + "loss": 0.2089, + "step": 61050 + }, + { + "epoch": 2.35, + "learning_rate": 1.0886765837394203e-05, + "loss": 0.1605, + "step": 61051 + }, + { + "epoch": 2.35, + "learning_rate": 1.0886124647345473e-05, + "loss": 0.1687, + "step": 61052 + }, + { + "epoch": 2.35, + "learning_rate": 1.0885483457296743e-05, + "loss": 0.276, + "step": 61053 + }, + { + "epoch": 2.35, + "learning_rate": 1.0884842267248013e-05, + "loss": 0.2123, + "step": 61054 + }, + { + "epoch": 2.35, + "learning_rate": 1.0884201077199282e-05, + "loss": 0.1643, + "step": 61055 + }, + { + "epoch": 2.35, + "learning_rate": 1.0883559887150552e-05, + "loss": 0.267, + "step": 61056 + }, + { + "epoch": 2.35, + "learning_rate": 1.0882918697101822e-05, + "loss": 0.1419, + "step": 61057 + }, + { + "epoch": 2.35, + "learning_rate": 1.0882277507053092e-05, + "loss": 0.1508, + "step": 61058 + }, + { + "epoch": 2.35, + "learning_rate": 1.088163631700436e-05, + "loss": 0.2887, + "step": 61059 + }, + { + "epoch": 2.35, + "learning_rate": 1.088099512695563e-05, + "loss": 0.2302, + "step": 61060 + }, + { + "epoch": 2.35, + "learning_rate": 1.08803539369069e-05, + "loss": 0.2422, + "step": 61061 + }, + { + "epoch": 2.35, + "learning_rate": 1.0879712746858169e-05, + "loss": 0.2955, + "step": 61062 + }, + { + "epoch": 2.35, + "learning_rate": 1.0879071556809439e-05, + "loss": 0.2991, + "step": 61063 + }, + { + "epoch": 2.35, + "learning_rate": 1.0878430366760709e-05, + "loss": 0.2837, + "step": 61064 + }, + { + "epoch": 2.35, + "learning_rate": 1.0877789176711977e-05, + "loss": 0.2298, + "step": 61065 + }, + { + "epoch": 2.35, + "learning_rate": 1.0877147986663247e-05, + "loss": 0.1689, + "step": 61066 + }, + { + "epoch": 2.35, + "learning_rate": 1.0876506796614517e-05, + "loss": 0.1186, + "step": 61067 + }, + { + "epoch": 2.35, + "learning_rate": 1.0875865606565787e-05, + "loss": 0.2222, + "step": 61068 + }, + { + "epoch": 2.35, + "learning_rate": 1.0875224416517055e-05, + "loss": 0.3216, + "step": 61069 + }, + { + "epoch": 2.35, + "learning_rate": 1.0874583226468325e-05, + "loss": 0.1783, + "step": 61070 + }, + { + "epoch": 2.35, + "learning_rate": 1.0873942036419595e-05, + "loss": 0.0906, + "step": 61071 + }, + { + "epoch": 2.35, + "learning_rate": 1.0873300846370865e-05, + "loss": 0.2505, + "step": 61072 + }, + { + "epoch": 2.35, + "learning_rate": 1.0872659656322134e-05, + "loss": 0.181, + "step": 61073 + }, + { + "epoch": 2.35, + "learning_rate": 1.0872018466273404e-05, + "loss": 0.1906, + "step": 61074 + }, + { + "epoch": 2.35, + "learning_rate": 1.0871377276224674e-05, + "loss": 0.1947, + "step": 61075 + }, + { + "epoch": 2.35, + "learning_rate": 1.0870736086175944e-05, + "loss": 0.1505, + "step": 61076 + }, + { + "epoch": 2.35, + "learning_rate": 1.0870094896127212e-05, + "loss": 0.1981, + "step": 61077 + }, + { + "epoch": 2.35, + "learning_rate": 1.0869453706078482e-05, + "loss": 0.2519, + "step": 61078 + }, + { + "epoch": 2.35, + "learning_rate": 1.0868812516029752e-05, + "loss": 0.1615, + "step": 61079 + }, + { + "epoch": 2.35, + "learning_rate": 1.0868171325981022e-05, + "loss": 0.2079, + "step": 61080 + }, + { + "epoch": 2.35, + "learning_rate": 1.086753013593229e-05, + "loss": 0.097, + "step": 61081 + }, + { + "epoch": 2.35, + "learning_rate": 1.086688894588356e-05, + "loss": 0.1878, + "step": 61082 + }, + { + "epoch": 2.35, + "learning_rate": 1.0866247755834829e-05, + "loss": 0.2599, + "step": 61083 + }, + { + "epoch": 2.35, + "learning_rate": 1.0865606565786099e-05, + "loss": 0.2102, + "step": 61084 + }, + { + "epoch": 2.35, + "learning_rate": 1.0864965375737369e-05, + "loss": 0.2699, + "step": 61085 + }, + { + "epoch": 2.35, + "learning_rate": 1.0864324185688639e-05, + "loss": 0.1978, + "step": 61086 + }, + { + "epoch": 2.35, + "learning_rate": 1.0863682995639907e-05, + "loss": 0.2461, + "step": 61087 + }, + { + "epoch": 2.35, + "learning_rate": 1.0863041805591177e-05, + "loss": 0.1151, + "step": 61088 + }, + { + "epoch": 2.35, + "learning_rate": 1.0862400615542447e-05, + "loss": 0.2554, + "step": 61089 + }, + { + "epoch": 2.35, + "learning_rate": 1.0861759425493717e-05, + "loss": 0.2896, + "step": 61090 + }, + { + "epoch": 2.35, + "learning_rate": 1.0861118235444986e-05, + "loss": 0.1713, + "step": 61091 + }, + { + "epoch": 2.35, + "learning_rate": 1.0860477045396256e-05, + "loss": 0.2321, + "step": 61092 + }, + { + "epoch": 2.35, + "learning_rate": 1.0859835855347526e-05, + "loss": 0.117, + "step": 61093 + }, + { + "epoch": 2.35, + "learning_rate": 1.0859194665298796e-05, + "loss": 0.4242, + "step": 61094 + }, + { + "epoch": 2.35, + "learning_rate": 1.0858553475250064e-05, + "loss": 0.2063, + "step": 61095 + }, + { + "epoch": 2.35, + "learning_rate": 1.0857912285201334e-05, + "loss": 0.3579, + "step": 61096 + }, + { + "epoch": 2.35, + "learning_rate": 1.0857271095152604e-05, + "loss": 0.2099, + "step": 61097 + }, + { + "epoch": 2.35, + "learning_rate": 1.0856629905103874e-05, + "loss": 0.2403, + "step": 61098 + }, + { + "epoch": 2.35, + "learning_rate": 1.0855988715055142e-05, + "loss": 0.3833, + "step": 61099 + }, + { + "epoch": 2.35, + "learning_rate": 1.0855347525006412e-05, + "loss": 0.2497, + "step": 61100 + }, + { + "epoch": 2.35, + "learning_rate": 1.0854706334957682e-05, + "loss": 0.282, + "step": 61101 + }, + { + "epoch": 2.35, + "learning_rate": 1.085406514490895e-05, + "loss": 0.3328, + "step": 61102 + }, + { + "epoch": 2.35, + "learning_rate": 1.085342395486022e-05, + "loss": 0.2791, + "step": 61103 + }, + { + "epoch": 2.35, + "learning_rate": 1.085278276481149e-05, + "loss": 0.1839, + "step": 61104 + }, + { + "epoch": 2.35, + "learning_rate": 1.085214157476276e-05, + "loss": 0.2124, + "step": 61105 + }, + { + "epoch": 2.35, + "learning_rate": 1.085150038471403e-05, + "loss": 0.2987, + "step": 61106 + }, + { + "epoch": 2.35, + "learning_rate": 1.08508591946653e-05, + "loss": 0.1738, + "step": 61107 + }, + { + "epoch": 2.35, + "learning_rate": 1.085021800461657e-05, + "loss": 0.3464, + "step": 61108 + }, + { + "epoch": 2.35, + "learning_rate": 1.0849576814567838e-05, + "loss": 0.2997, + "step": 61109 + }, + { + "epoch": 2.35, + "learning_rate": 1.0848935624519108e-05, + "loss": 0.2382, + "step": 61110 + }, + { + "epoch": 2.35, + "learning_rate": 1.0848294434470378e-05, + "loss": 0.1873, + "step": 61111 + }, + { + "epoch": 2.35, + "learning_rate": 1.0847653244421648e-05, + "loss": 0.1536, + "step": 61112 + }, + { + "epoch": 2.35, + "learning_rate": 1.0847012054372916e-05, + "loss": 0.3424, + "step": 61113 + }, + { + "epoch": 2.35, + "learning_rate": 1.0846370864324186e-05, + "loss": 0.2475, + "step": 61114 + }, + { + "epoch": 2.35, + "learning_rate": 1.0845729674275456e-05, + "loss": 0.1417, + "step": 61115 + }, + { + "epoch": 2.35, + "learning_rate": 1.0845088484226726e-05, + "loss": 0.1835, + "step": 61116 + }, + { + "epoch": 2.35, + "learning_rate": 1.0844447294177994e-05, + "loss": 0.2859, + "step": 61117 + }, + { + "epoch": 2.35, + "learning_rate": 1.0843806104129264e-05, + "loss": 0.1625, + "step": 61118 + }, + { + "epoch": 2.35, + "learning_rate": 1.0843164914080534e-05, + "loss": 0.2327, + "step": 61119 + }, + { + "epoch": 2.35, + "learning_rate": 1.0842523724031804e-05, + "loss": 0.0962, + "step": 61120 + }, + { + "epoch": 2.35, + "learning_rate": 1.0841882533983073e-05, + "loss": 0.3327, + "step": 61121 + }, + { + "epoch": 2.35, + "learning_rate": 1.0841241343934343e-05, + "loss": 0.3057, + "step": 61122 + }, + { + "epoch": 2.35, + "learning_rate": 1.0840600153885611e-05, + "loss": 0.2532, + "step": 61123 + }, + { + "epoch": 2.35, + "learning_rate": 1.0839958963836881e-05, + "loss": 0.2198, + "step": 61124 + }, + { + "epoch": 2.35, + "learning_rate": 1.0839317773788151e-05, + "loss": 0.2303, + "step": 61125 + }, + { + "epoch": 2.35, + "learning_rate": 1.0838676583739421e-05, + "loss": 0.2819, + "step": 61126 + }, + { + "epoch": 2.35, + "learning_rate": 1.083803539369069e-05, + "loss": 0.2374, + "step": 61127 + }, + { + "epoch": 2.35, + "learning_rate": 1.083739420364196e-05, + "loss": 0.1791, + "step": 61128 + }, + { + "epoch": 2.35, + "learning_rate": 1.083675301359323e-05, + "loss": 0.2299, + "step": 61129 + }, + { + "epoch": 2.35, + "learning_rate": 1.08361118235445e-05, + "loss": 0.2386, + "step": 61130 + }, + { + "epoch": 2.35, + "learning_rate": 1.0835470633495768e-05, + "loss": 0.1791, + "step": 61131 + }, + { + "epoch": 2.35, + "learning_rate": 1.0834829443447038e-05, + "loss": 0.3298, + "step": 61132 + }, + { + "epoch": 2.35, + "learning_rate": 1.0834188253398308e-05, + "loss": 0.2283, + "step": 61133 + }, + { + "epoch": 2.35, + "learning_rate": 1.0833547063349578e-05, + "loss": 0.169, + "step": 61134 + }, + { + "epoch": 2.35, + "learning_rate": 1.0832905873300846e-05, + "loss": 0.2448, + "step": 61135 + }, + { + "epoch": 2.35, + "learning_rate": 1.0832264683252116e-05, + "loss": 0.5281, + "step": 61136 + }, + { + "epoch": 2.35, + "learning_rate": 1.0831623493203386e-05, + "loss": 0.1699, + "step": 61137 + }, + { + "epoch": 2.35, + "learning_rate": 1.0830982303154656e-05, + "loss": 0.1575, + "step": 61138 + }, + { + "epoch": 2.35, + "learning_rate": 1.0830341113105925e-05, + "loss": 0.1267, + "step": 61139 + }, + { + "epoch": 2.35, + "learning_rate": 1.0829699923057195e-05, + "loss": 0.1437, + "step": 61140 + }, + { + "epoch": 2.35, + "learning_rate": 1.0829058733008465e-05, + "loss": 0.2179, + "step": 61141 + }, + { + "epoch": 2.35, + "learning_rate": 1.0828417542959733e-05, + "loss": 0.2235, + "step": 61142 + }, + { + "epoch": 2.35, + "learning_rate": 1.0827776352911003e-05, + "loss": 0.2026, + "step": 61143 + }, + { + "epoch": 2.35, + "learning_rate": 1.0827135162862273e-05, + "loss": 0.3428, + "step": 61144 + }, + { + "epoch": 2.35, + "learning_rate": 1.0826493972813541e-05, + "loss": 0.2343, + "step": 61145 + }, + { + "epoch": 2.35, + "learning_rate": 1.0825852782764811e-05, + "loss": 0.1936, + "step": 61146 + }, + { + "epoch": 2.35, + "learning_rate": 1.0825211592716081e-05, + "loss": 0.2378, + "step": 61147 + }, + { + "epoch": 2.35, + "learning_rate": 1.0824570402667352e-05, + "loss": 0.3073, + "step": 61148 + }, + { + "epoch": 2.35, + "learning_rate": 1.082392921261862e-05, + "loss": 0.2996, + "step": 61149 + }, + { + "epoch": 2.35, + "learning_rate": 1.082328802256989e-05, + "loss": 0.2087, + "step": 61150 + }, + { + "epoch": 2.35, + "learning_rate": 1.082264683252116e-05, + "loss": 0.2659, + "step": 61151 + }, + { + "epoch": 2.35, + "learning_rate": 1.082200564247243e-05, + "loss": 0.2587, + "step": 61152 + }, + { + "epoch": 2.35, + "learning_rate": 1.0821364452423698e-05, + "loss": 0.2417, + "step": 61153 + }, + { + "epoch": 2.35, + "learning_rate": 1.0820723262374968e-05, + "loss": 0.3376, + "step": 61154 + }, + { + "epoch": 2.35, + "learning_rate": 1.0820082072326238e-05, + "loss": 0.1849, + "step": 61155 + }, + { + "epoch": 2.35, + "learning_rate": 1.0819440882277508e-05, + "loss": 0.2159, + "step": 61156 + }, + { + "epoch": 2.35, + "learning_rate": 1.0818799692228777e-05, + "loss": 0.2178, + "step": 61157 + }, + { + "epoch": 2.35, + "learning_rate": 1.0818158502180047e-05, + "loss": 0.2526, + "step": 61158 + }, + { + "epoch": 2.35, + "learning_rate": 1.0817517312131317e-05, + "loss": 0.2931, + "step": 61159 + }, + { + "epoch": 2.35, + "learning_rate": 1.0816876122082585e-05, + "loss": 0.2211, + "step": 61160 + }, + { + "epoch": 2.35, + "learning_rate": 1.0816234932033855e-05, + "loss": 0.3497, + "step": 61161 + }, + { + "epoch": 2.35, + "learning_rate": 1.0815593741985125e-05, + "loss": 0.231, + "step": 61162 + }, + { + "epoch": 2.35, + "learning_rate": 1.0814952551936395e-05, + "loss": 0.2592, + "step": 61163 + }, + { + "epoch": 2.35, + "learning_rate": 1.0814311361887663e-05, + "loss": 0.2229, + "step": 61164 + }, + { + "epoch": 2.35, + "learning_rate": 1.0813670171838933e-05, + "loss": 0.2607, + "step": 61165 + }, + { + "epoch": 2.35, + "learning_rate": 1.0813028981790203e-05, + "loss": 0.2119, + "step": 61166 + }, + { + "epoch": 2.35, + "learning_rate": 1.0812387791741472e-05, + "loss": 0.2154, + "step": 61167 + }, + { + "epoch": 2.35, + "learning_rate": 1.0811746601692742e-05, + "loss": 0.3785, + "step": 61168 + }, + { + "epoch": 2.35, + "learning_rate": 1.0811105411644012e-05, + "loss": 0.2589, + "step": 61169 + }, + { + "epoch": 2.35, + "learning_rate": 1.0810464221595282e-05, + "loss": 0.3401, + "step": 61170 + }, + { + "epoch": 2.35, + "learning_rate": 1.080982303154655e-05, + "loss": 0.2481, + "step": 61171 + }, + { + "epoch": 2.35, + "learning_rate": 1.080918184149782e-05, + "loss": 0.2616, + "step": 61172 + }, + { + "epoch": 2.35, + "learning_rate": 1.080854065144909e-05, + "loss": 0.2059, + "step": 61173 + }, + { + "epoch": 2.35, + "learning_rate": 1.080789946140036e-05, + "loss": 0.2826, + "step": 61174 + }, + { + "epoch": 2.35, + "learning_rate": 1.0807258271351629e-05, + "loss": 0.175, + "step": 61175 + }, + { + "epoch": 2.35, + "learning_rate": 1.0806617081302899e-05, + "loss": 0.2352, + "step": 61176 + }, + { + "epoch": 2.35, + "learning_rate": 1.0805975891254169e-05, + "loss": 0.1906, + "step": 61177 + }, + { + "epoch": 2.35, + "learning_rate": 1.0805334701205439e-05, + "loss": 0.3277, + "step": 61178 + }, + { + "epoch": 2.35, + "learning_rate": 1.0804693511156707e-05, + "loss": 0.2476, + "step": 61179 + }, + { + "epoch": 2.35, + "learning_rate": 1.0804052321107977e-05, + "loss": 0.1882, + "step": 61180 + }, + { + "epoch": 2.35, + "learning_rate": 1.0803411131059247e-05, + "loss": 0.1806, + "step": 61181 + }, + { + "epoch": 2.35, + "learning_rate": 1.0802769941010515e-05, + "loss": 0.2215, + "step": 61182 + }, + { + "epoch": 2.35, + "learning_rate": 1.0802128750961785e-05, + "loss": 0.2163, + "step": 61183 + }, + { + "epoch": 2.35, + "learning_rate": 1.0801487560913055e-05, + "loss": 0.1952, + "step": 61184 + }, + { + "epoch": 2.35, + "learning_rate": 1.0800846370864324e-05, + "loss": 0.272, + "step": 61185 + }, + { + "epoch": 2.35, + "learning_rate": 1.0800205180815594e-05, + "loss": 0.1272, + "step": 61186 + }, + { + "epoch": 2.35, + "learning_rate": 1.0799563990766864e-05, + "loss": 0.0915, + "step": 61187 + }, + { + "epoch": 2.35, + "learning_rate": 1.0798922800718134e-05, + "loss": 0.2973, + "step": 61188 + }, + { + "epoch": 2.35, + "learning_rate": 1.0798281610669402e-05, + "loss": 0.2034, + "step": 61189 + }, + { + "epoch": 2.35, + "learning_rate": 1.0797640420620672e-05, + "loss": 0.2451, + "step": 61190 + }, + { + "epoch": 2.35, + "learning_rate": 1.0796999230571942e-05, + "loss": 0.2559, + "step": 61191 + }, + { + "epoch": 2.35, + "learning_rate": 1.0796358040523212e-05, + "loss": 0.1582, + "step": 61192 + }, + { + "epoch": 2.35, + "learning_rate": 1.079571685047448e-05, + "loss": 0.2822, + "step": 61193 + }, + { + "epoch": 2.35, + "learning_rate": 1.079507566042575e-05, + "loss": 0.2731, + "step": 61194 + }, + { + "epoch": 2.35, + "learning_rate": 1.079443447037702e-05, + "loss": 0.2371, + "step": 61195 + }, + { + "epoch": 2.35, + "learning_rate": 1.079379328032829e-05, + "loss": 0.3759, + "step": 61196 + }, + { + "epoch": 2.35, + "learning_rate": 1.0793152090279559e-05, + "loss": 0.1977, + "step": 61197 + }, + { + "epoch": 2.35, + "learning_rate": 1.0792510900230829e-05, + "loss": 0.1681, + "step": 61198 + }, + { + "epoch": 2.35, + "learning_rate": 1.0791869710182099e-05, + "loss": 0.2985, + "step": 61199 + }, + { + "epoch": 2.35, + "learning_rate": 1.0791228520133367e-05, + "loss": 0.3606, + "step": 61200 + }, + { + "epoch": 2.35, + "learning_rate": 1.0790587330084637e-05, + "loss": 0.2189, + "step": 61201 + }, + { + "epoch": 2.35, + "learning_rate": 1.0789946140035907e-05, + "loss": 0.1988, + "step": 61202 + }, + { + "epoch": 2.35, + "learning_rate": 1.0789304949987177e-05, + "loss": 0.2717, + "step": 61203 + }, + { + "epoch": 2.35, + "learning_rate": 1.0788663759938446e-05, + "loss": 0.3499, + "step": 61204 + }, + { + "epoch": 2.35, + "learning_rate": 1.0788022569889716e-05, + "loss": 0.0892, + "step": 61205 + }, + { + "epoch": 2.35, + "learning_rate": 1.0787381379840986e-05, + "loss": 0.2894, + "step": 61206 + }, + { + "epoch": 2.35, + "learning_rate": 1.0786740189792254e-05, + "loss": 0.146, + "step": 61207 + }, + { + "epoch": 2.35, + "learning_rate": 1.0786098999743524e-05, + "loss": 0.3737, + "step": 61208 + }, + { + "epoch": 2.35, + "learning_rate": 1.0785457809694794e-05, + "loss": 0.1914, + "step": 61209 + }, + { + "epoch": 2.35, + "learning_rate": 1.0784816619646064e-05, + "loss": 0.3473, + "step": 61210 + }, + { + "epoch": 2.35, + "learning_rate": 1.0784175429597332e-05, + "loss": 0.2022, + "step": 61211 + }, + { + "epoch": 2.35, + "learning_rate": 1.0783534239548602e-05, + "loss": 0.2401, + "step": 61212 + }, + { + "epoch": 2.35, + "learning_rate": 1.0782893049499872e-05, + "loss": 0.4077, + "step": 61213 + }, + { + "epoch": 2.35, + "learning_rate": 1.0782251859451143e-05, + "loss": 0.1586, + "step": 61214 + }, + { + "epoch": 2.35, + "learning_rate": 1.078161066940241e-05, + "loss": 0.2477, + "step": 61215 + }, + { + "epoch": 2.35, + "learning_rate": 1.0780969479353681e-05, + "loss": 0.2972, + "step": 61216 + }, + { + "epoch": 2.35, + "learning_rate": 1.0780328289304951e-05, + "loss": 0.1921, + "step": 61217 + }, + { + "epoch": 2.35, + "learning_rate": 1.0779687099256221e-05, + "loss": 0.2111, + "step": 61218 + }, + { + "epoch": 2.35, + "learning_rate": 1.077904590920749e-05, + "loss": 0.4096, + "step": 61219 + }, + { + "epoch": 2.35, + "learning_rate": 1.077840471915876e-05, + "loss": 0.2227, + "step": 61220 + }, + { + "epoch": 2.35, + "learning_rate": 1.077776352911003e-05, + "loss": 0.1905, + "step": 61221 + }, + { + "epoch": 2.35, + "learning_rate": 1.0777122339061298e-05, + "loss": 0.2394, + "step": 61222 + }, + { + "epoch": 2.35, + "learning_rate": 1.0776481149012568e-05, + "loss": 0.2231, + "step": 61223 + }, + { + "epoch": 2.35, + "learning_rate": 1.0775839958963838e-05, + "loss": 0.2072, + "step": 61224 + }, + { + "epoch": 2.35, + "learning_rate": 1.0775198768915108e-05, + "loss": 0.1857, + "step": 61225 + }, + { + "epoch": 2.35, + "learning_rate": 1.0774557578866376e-05, + "loss": 0.249, + "step": 61226 + }, + { + "epoch": 2.35, + "learning_rate": 1.0773916388817646e-05, + "loss": 0.1498, + "step": 61227 + }, + { + "epoch": 2.35, + "learning_rate": 1.0773275198768916e-05, + "loss": 0.2186, + "step": 61228 + }, + { + "epoch": 2.35, + "learning_rate": 1.0772634008720184e-05, + "loss": 0.1677, + "step": 61229 + }, + { + "epoch": 2.35, + "learning_rate": 1.0771992818671454e-05, + "loss": 0.1746, + "step": 61230 + }, + { + "epoch": 2.35, + "learning_rate": 1.0771351628622724e-05, + "loss": 0.2395, + "step": 61231 + }, + { + "epoch": 2.35, + "learning_rate": 1.0770710438573994e-05, + "loss": 0.2153, + "step": 61232 + }, + { + "epoch": 2.35, + "learning_rate": 1.0770069248525263e-05, + "loss": 0.3817, + "step": 61233 + }, + { + "epoch": 2.35, + "learning_rate": 1.0769428058476533e-05, + "loss": 0.263, + "step": 61234 + }, + { + "epoch": 2.35, + "learning_rate": 1.0768786868427803e-05, + "loss": 0.3531, + "step": 61235 + }, + { + "epoch": 2.35, + "learning_rate": 1.0768145678379073e-05, + "loss": 0.2511, + "step": 61236 + }, + { + "epoch": 2.35, + "learning_rate": 1.0767504488330341e-05, + "loss": 0.2222, + "step": 61237 + }, + { + "epoch": 2.35, + "learning_rate": 1.0766863298281611e-05, + "loss": 0.278, + "step": 61238 + }, + { + "epoch": 2.35, + "learning_rate": 1.0766222108232881e-05, + "loss": 0.3067, + "step": 61239 + }, + { + "epoch": 2.35, + "learning_rate": 1.076558091818415e-05, + "loss": 0.1745, + "step": 61240 + }, + { + "epoch": 2.35, + "learning_rate": 1.076493972813542e-05, + "loss": 0.1654, + "step": 61241 + }, + { + "epoch": 2.35, + "learning_rate": 1.076429853808669e-05, + "loss": 0.2032, + "step": 61242 + }, + { + "epoch": 2.35, + "learning_rate": 1.076365734803796e-05, + "loss": 0.2482, + "step": 61243 + }, + { + "epoch": 2.35, + "learning_rate": 1.0763016157989228e-05, + "loss": 0.2352, + "step": 61244 + }, + { + "epoch": 2.35, + "learning_rate": 1.0762374967940498e-05, + "loss": 0.2402, + "step": 61245 + }, + { + "epoch": 2.35, + "learning_rate": 1.0761733777891768e-05, + "loss": 0.2066, + "step": 61246 + }, + { + "epoch": 2.35, + "learning_rate": 1.0761092587843036e-05, + "loss": 0.1467, + "step": 61247 + }, + { + "epoch": 2.35, + "learning_rate": 1.0760451397794306e-05, + "loss": 0.1714, + "step": 61248 + }, + { + "epoch": 2.35, + "learning_rate": 1.0759810207745576e-05, + "loss": 0.2135, + "step": 61249 + }, + { + "epoch": 2.35, + "learning_rate": 1.0759169017696846e-05, + "loss": 0.2393, + "step": 61250 + }, + { + "epoch": 2.35, + "learning_rate": 1.0758527827648115e-05, + "loss": 0.32, + "step": 61251 + }, + { + "epoch": 2.35, + "learning_rate": 1.0757886637599385e-05, + "loss": 0.191, + "step": 61252 + }, + { + "epoch": 2.35, + "learning_rate": 1.0757245447550655e-05, + "loss": 0.2082, + "step": 61253 + }, + { + "epoch": 2.36, + "learning_rate": 1.0756604257501925e-05, + "loss": 0.3569, + "step": 61254 + }, + { + "epoch": 2.36, + "learning_rate": 1.0755963067453193e-05, + "loss": 0.1992, + "step": 61255 + }, + { + "epoch": 2.36, + "learning_rate": 1.0755321877404463e-05, + "loss": 0.1258, + "step": 61256 + }, + { + "epoch": 2.36, + "learning_rate": 1.0754680687355733e-05, + "loss": 0.2254, + "step": 61257 + }, + { + "epoch": 2.36, + "learning_rate": 1.0754039497307003e-05, + "loss": 0.1769, + "step": 61258 + }, + { + "epoch": 2.36, + "learning_rate": 1.0753398307258271e-05, + "loss": 0.1426, + "step": 61259 + }, + { + "epoch": 2.36, + "learning_rate": 1.0752757117209542e-05, + "loss": 0.3069, + "step": 61260 + }, + { + "epoch": 2.36, + "learning_rate": 1.0752115927160812e-05, + "loss": 0.1752, + "step": 61261 + }, + { + "epoch": 2.36, + "learning_rate": 1.075147473711208e-05, + "loss": 0.213, + "step": 61262 + }, + { + "epoch": 2.36, + "learning_rate": 1.075083354706335e-05, + "loss": 0.2001, + "step": 61263 + }, + { + "epoch": 2.36, + "learning_rate": 1.075019235701462e-05, + "loss": 0.2434, + "step": 61264 + }, + { + "epoch": 2.36, + "learning_rate": 1.074955116696589e-05, + "loss": 0.2492, + "step": 61265 + }, + { + "epoch": 2.36, + "learning_rate": 1.0748909976917158e-05, + "loss": 0.2914, + "step": 61266 + }, + { + "epoch": 2.36, + "learning_rate": 1.0748268786868428e-05, + "loss": 0.1536, + "step": 61267 + }, + { + "epoch": 2.36, + "learning_rate": 1.0747627596819698e-05, + "loss": 0.2993, + "step": 61268 + }, + { + "epoch": 2.36, + "learning_rate": 1.0746986406770967e-05, + "loss": 0.2598, + "step": 61269 + }, + { + "epoch": 2.36, + "learning_rate": 1.0746345216722237e-05, + "loss": 0.1469, + "step": 61270 + }, + { + "epoch": 2.36, + "learning_rate": 1.0745704026673507e-05, + "loss": 0.1511, + "step": 61271 + }, + { + "epoch": 2.36, + "learning_rate": 1.0745062836624777e-05, + "loss": 0.2284, + "step": 61272 + }, + { + "epoch": 2.36, + "learning_rate": 1.0744421646576045e-05, + "loss": 0.1881, + "step": 61273 + }, + { + "epoch": 2.36, + "learning_rate": 1.0743780456527315e-05, + "loss": 0.4189, + "step": 61274 + }, + { + "epoch": 2.36, + "learning_rate": 1.0743139266478585e-05, + "loss": 0.1677, + "step": 61275 + }, + { + "epoch": 2.36, + "learning_rate": 1.0742498076429855e-05, + "loss": 0.2729, + "step": 61276 + }, + { + "epoch": 2.36, + "learning_rate": 1.0741856886381123e-05, + "loss": 0.1489, + "step": 61277 + }, + { + "epoch": 2.36, + "learning_rate": 1.0741215696332393e-05, + "loss": 0.2218, + "step": 61278 + }, + { + "epoch": 2.36, + "learning_rate": 1.0740574506283663e-05, + "loss": 0.139, + "step": 61279 + }, + { + "epoch": 2.36, + "learning_rate": 1.0739933316234932e-05, + "loss": 0.227, + "step": 61280 + }, + { + "epoch": 2.36, + "learning_rate": 1.0739292126186202e-05, + "loss": 0.2007, + "step": 61281 + }, + { + "epoch": 2.36, + "learning_rate": 1.0738650936137472e-05, + "loss": 0.2699, + "step": 61282 + }, + { + "epoch": 2.36, + "learning_rate": 1.0738009746088742e-05, + "loss": 0.3199, + "step": 61283 + }, + { + "epoch": 2.36, + "learning_rate": 1.073736855604001e-05, + "loss": 0.1855, + "step": 61284 + }, + { + "epoch": 2.36, + "learning_rate": 1.073672736599128e-05, + "loss": 0.2769, + "step": 61285 + }, + { + "epoch": 2.36, + "learning_rate": 1.073608617594255e-05, + "loss": 0.3814, + "step": 61286 + }, + { + "epoch": 2.36, + "learning_rate": 1.0735444985893819e-05, + "loss": 0.2102, + "step": 61287 + }, + { + "epoch": 2.36, + "learning_rate": 1.0734803795845089e-05, + "loss": 0.1852, + "step": 61288 + }, + { + "epoch": 2.36, + "learning_rate": 1.0734162605796359e-05, + "loss": 0.102, + "step": 61289 + }, + { + "epoch": 2.36, + "learning_rate": 1.0733521415747629e-05, + "loss": 0.1878, + "step": 61290 + }, + { + "epoch": 2.36, + "learning_rate": 1.0732880225698897e-05, + "loss": 0.2982, + "step": 61291 + }, + { + "epoch": 2.36, + "learning_rate": 1.0732239035650167e-05, + "loss": 0.1824, + "step": 61292 + }, + { + "epoch": 2.36, + "learning_rate": 1.0731597845601437e-05, + "loss": 0.1454, + "step": 61293 + }, + { + "epoch": 2.36, + "learning_rate": 1.0730956655552707e-05, + "loss": 0.2406, + "step": 61294 + }, + { + "epoch": 2.36, + "learning_rate": 1.0730315465503975e-05, + "loss": 0.2193, + "step": 61295 + }, + { + "epoch": 2.36, + "learning_rate": 1.0729674275455245e-05, + "loss": 0.2126, + "step": 61296 + }, + { + "epoch": 2.36, + "learning_rate": 1.0729033085406515e-05, + "loss": 0.2004, + "step": 61297 + }, + { + "epoch": 2.36, + "learning_rate": 1.0728391895357784e-05, + "loss": 0.1187, + "step": 61298 + }, + { + "epoch": 2.36, + "learning_rate": 1.0727750705309054e-05, + "loss": 0.3273, + "step": 61299 + }, + { + "epoch": 2.36, + "learning_rate": 1.0727109515260324e-05, + "loss": 0.1309, + "step": 61300 + }, + { + "epoch": 2.36, + "learning_rate": 1.0726468325211594e-05, + "loss": 0.1712, + "step": 61301 + }, + { + "epoch": 2.36, + "learning_rate": 1.0725827135162862e-05, + "loss": 0.2068, + "step": 61302 + }, + { + "epoch": 2.36, + "learning_rate": 1.0725185945114132e-05, + "loss": 0.2766, + "step": 61303 + }, + { + "epoch": 2.36, + "learning_rate": 1.0724544755065402e-05, + "loss": 0.1857, + "step": 61304 + }, + { + "epoch": 2.36, + "learning_rate": 1.0723903565016672e-05, + "loss": 0.1313, + "step": 61305 + }, + { + "epoch": 2.36, + "learning_rate": 1.072326237496794e-05, + "loss": 0.2685, + "step": 61306 + }, + { + "epoch": 2.36, + "learning_rate": 1.072262118491921e-05, + "loss": 0.089, + "step": 61307 + }, + { + "epoch": 2.36, + "learning_rate": 1.072197999487048e-05, + "loss": 0.2002, + "step": 61308 + }, + { + "epoch": 2.36, + "learning_rate": 1.0721338804821749e-05, + "loss": 0.2474, + "step": 61309 + }, + { + "epoch": 2.36, + "learning_rate": 1.0720697614773019e-05, + "loss": 0.1178, + "step": 61310 + }, + { + "epoch": 2.36, + "learning_rate": 1.0720056424724289e-05, + "loss": 0.2979, + "step": 61311 + }, + { + "epoch": 2.36, + "learning_rate": 1.0719415234675559e-05, + "loss": 0.2243, + "step": 61312 + }, + { + "epoch": 2.36, + "learning_rate": 1.0718774044626827e-05, + "loss": 0.32, + "step": 61313 + }, + { + "epoch": 2.36, + "learning_rate": 1.0718132854578097e-05, + "loss": 0.2312, + "step": 61314 + }, + { + "epoch": 2.36, + "learning_rate": 1.0717491664529367e-05, + "loss": 0.1902, + "step": 61315 + }, + { + "epoch": 2.36, + "learning_rate": 1.0716850474480637e-05, + "loss": 0.2617, + "step": 61316 + }, + { + "epoch": 2.36, + "learning_rate": 1.0716209284431906e-05, + "loss": 0.2372, + "step": 61317 + }, + { + "epoch": 2.36, + "learning_rate": 1.0715568094383176e-05, + "loss": 0.2107, + "step": 61318 + }, + { + "epoch": 2.36, + "learning_rate": 1.0714926904334446e-05, + "loss": 0.1791, + "step": 61319 + }, + { + "epoch": 2.36, + "learning_rate": 1.0714285714285714e-05, + "loss": 0.181, + "step": 61320 + }, + { + "epoch": 2.36, + "learning_rate": 1.0713644524236984e-05, + "loss": 0.2633, + "step": 61321 + }, + { + "epoch": 2.36, + "learning_rate": 1.0713003334188254e-05, + "loss": 0.1813, + "step": 61322 + }, + { + "epoch": 2.36, + "learning_rate": 1.0712362144139524e-05, + "loss": 0.2038, + "step": 61323 + }, + { + "epoch": 2.36, + "learning_rate": 1.0711720954090792e-05, + "loss": 0.2891, + "step": 61324 + }, + { + "epoch": 2.36, + "learning_rate": 1.0711079764042062e-05, + "loss": 0.3152, + "step": 61325 + }, + { + "epoch": 2.36, + "learning_rate": 1.0710438573993333e-05, + "loss": 0.2278, + "step": 61326 + }, + { + "epoch": 2.36, + "learning_rate": 1.0709797383944603e-05, + "loss": 0.1458, + "step": 61327 + }, + { + "epoch": 2.36, + "learning_rate": 1.0709156193895871e-05, + "loss": 0.2154, + "step": 61328 + }, + { + "epoch": 2.36, + "learning_rate": 1.0708515003847141e-05, + "loss": 0.1999, + "step": 61329 + }, + { + "epoch": 2.36, + "learning_rate": 1.0707873813798411e-05, + "loss": 0.1829, + "step": 61330 + }, + { + "epoch": 2.36, + "learning_rate": 1.070723262374968e-05, + "loss": 0.1603, + "step": 61331 + }, + { + "epoch": 2.36, + "learning_rate": 1.070659143370095e-05, + "loss": 0.39, + "step": 61332 + }, + { + "epoch": 2.36, + "learning_rate": 1.070595024365222e-05, + "loss": 0.2625, + "step": 61333 + }, + { + "epoch": 2.36, + "learning_rate": 1.070530905360349e-05, + "loss": 0.2938, + "step": 61334 + }, + { + "epoch": 2.36, + "learning_rate": 1.0704667863554758e-05, + "loss": 0.2344, + "step": 61335 + }, + { + "epoch": 2.36, + "learning_rate": 1.0704026673506028e-05, + "loss": 0.2966, + "step": 61336 + }, + { + "epoch": 2.36, + "learning_rate": 1.0703385483457298e-05, + "loss": 0.2498, + "step": 61337 + }, + { + "epoch": 2.36, + "learning_rate": 1.0702744293408566e-05, + "loss": 0.2978, + "step": 61338 + }, + { + "epoch": 2.36, + "learning_rate": 1.0702103103359836e-05, + "loss": 0.2148, + "step": 61339 + }, + { + "epoch": 2.36, + "learning_rate": 1.0701461913311106e-05, + "loss": 0.2608, + "step": 61340 + }, + { + "epoch": 2.36, + "learning_rate": 1.0700820723262376e-05, + "loss": 0.1859, + "step": 61341 + }, + { + "epoch": 2.36, + "learning_rate": 1.0700179533213644e-05, + "loss": 0.2232, + "step": 61342 + }, + { + "epoch": 2.36, + "learning_rate": 1.0699538343164914e-05, + "loss": 0.1426, + "step": 61343 + }, + { + "epoch": 2.36, + "learning_rate": 1.0698897153116184e-05, + "loss": 0.2578, + "step": 61344 + }, + { + "epoch": 2.36, + "learning_rate": 1.0698255963067454e-05, + "loss": 0.2642, + "step": 61345 + }, + { + "epoch": 2.36, + "learning_rate": 1.0697614773018723e-05, + "loss": 0.2908, + "step": 61346 + }, + { + "epoch": 2.36, + "learning_rate": 1.0696973582969993e-05, + "loss": 0.1717, + "step": 61347 + }, + { + "epoch": 2.36, + "learning_rate": 1.0696332392921263e-05, + "loss": 0.2186, + "step": 61348 + }, + { + "epoch": 2.36, + "learning_rate": 1.0695691202872531e-05, + "loss": 0.3057, + "step": 61349 + }, + { + "epoch": 2.36, + "learning_rate": 1.0695050012823801e-05, + "loss": 0.2486, + "step": 61350 + }, + { + "epoch": 2.36, + "learning_rate": 1.0694408822775071e-05, + "loss": 0.3122, + "step": 61351 + }, + { + "epoch": 2.36, + "learning_rate": 1.0693767632726341e-05, + "loss": 0.2273, + "step": 61352 + }, + { + "epoch": 2.36, + "learning_rate": 1.069312644267761e-05, + "loss": 0.2448, + "step": 61353 + }, + { + "epoch": 2.36, + "learning_rate": 1.069248525262888e-05, + "loss": 0.1304, + "step": 61354 + }, + { + "epoch": 2.36, + "learning_rate": 1.069184406258015e-05, + "loss": 0.129, + "step": 61355 + }, + { + "epoch": 2.36, + "learning_rate": 1.069120287253142e-05, + "loss": 0.1935, + "step": 61356 + }, + { + "epoch": 2.36, + "learning_rate": 1.0690561682482688e-05, + "loss": 0.2009, + "step": 61357 + }, + { + "epoch": 2.36, + "learning_rate": 1.0689920492433958e-05, + "loss": 0.2231, + "step": 61358 + }, + { + "epoch": 2.36, + "learning_rate": 1.0689279302385228e-05, + "loss": 0.2473, + "step": 61359 + }, + { + "epoch": 2.36, + "learning_rate": 1.0688638112336496e-05, + "loss": 0.2891, + "step": 61360 + }, + { + "epoch": 2.36, + "learning_rate": 1.0687996922287766e-05, + "loss": 0.317, + "step": 61361 + }, + { + "epoch": 2.36, + "learning_rate": 1.0687355732239036e-05, + "loss": 0.1868, + "step": 61362 + }, + { + "epoch": 2.36, + "learning_rate": 1.0686714542190306e-05, + "loss": 0.2589, + "step": 61363 + }, + { + "epoch": 2.36, + "learning_rate": 1.0686073352141575e-05, + "loss": 0.1628, + "step": 61364 + }, + { + "epoch": 2.36, + "learning_rate": 1.0685432162092845e-05, + "loss": 0.2656, + "step": 61365 + }, + { + "epoch": 2.36, + "learning_rate": 1.0684790972044115e-05, + "loss": 0.2335, + "step": 61366 + }, + { + "epoch": 2.36, + "learning_rate": 1.0684149781995385e-05, + "loss": 0.1357, + "step": 61367 + }, + { + "epoch": 2.36, + "learning_rate": 1.0683508591946653e-05, + "loss": 0.2518, + "step": 61368 + }, + { + "epoch": 2.36, + "learning_rate": 1.0682867401897923e-05, + "loss": 0.2044, + "step": 61369 + }, + { + "epoch": 2.36, + "learning_rate": 1.0682226211849193e-05, + "loss": 0.1606, + "step": 61370 + }, + { + "epoch": 2.36, + "learning_rate": 1.0681585021800462e-05, + "loss": 0.2098, + "step": 61371 + }, + { + "epoch": 2.36, + "learning_rate": 1.0680943831751732e-05, + "loss": 0.2856, + "step": 61372 + }, + { + "epoch": 2.36, + "learning_rate": 1.0680302641703002e-05, + "loss": 0.2264, + "step": 61373 + }, + { + "epoch": 2.36, + "learning_rate": 1.0679661451654272e-05, + "loss": 0.1668, + "step": 61374 + }, + { + "epoch": 2.36, + "learning_rate": 1.067902026160554e-05, + "loss": 0.1377, + "step": 61375 + }, + { + "epoch": 2.36, + "learning_rate": 1.067837907155681e-05, + "loss": 0.3119, + "step": 61376 + }, + { + "epoch": 2.36, + "learning_rate": 1.067773788150808e-05, + "loss": 0.1966, + "step": 61377 + }, + { + "epoch": 2.36, + "learning_rate": 1.0677096691459348e-05, + "loss": 0.1908, + "step": 61378 + }, + { + "epoch": 2.36, + "learning_rate": 1.0676455501410618e-05, + "loss": 0.3041, + "step": 61379 + }, + { + "epoch": 2.36, + "learning_rate": 1.0675814311361888e-05, + "loss": 0.225, + "step": 61380 + }, + { + "epoch": 2.36, + "learning_rate": 1.0675173121313158e-05, + "loss": 0.2318, + "step": 61381 + }, + { + "epoch": 2.36, + "learning_rate": 1.0674531931264427e-05, + "loss": 0.1772, + "step": 61382 + }, + { + "epoch": 2.36, + "learning_rate": 1.0673890741215697e-05, + "loss": 0.4646, + "step": 61383 + }, + { + "epoch": 2.36, + "learning_rate": 1.0673249551166967e-05, + "loss": 0.2041, + "step": 61384 + }, + { + "epoch": 2.36, + "learning_rate": 1.0672608361118237e-05, + "loss": 0.2687, + "step": 61385 + }, + { + "epoch": 2.36, + "learning_rate": 1.0671967171069505e-05, + "loss": 0.1878, + "step": 61386 + }, + { + "epoch": 2.36, + "learning_rate": 1.0671325981020775e-05, + "loss": 0.1953, + "step": 61387 + }, + { + "epoch": 2.36, + "learning_rate": 1.0670684790972045e-05, + "loss": 0.2197, + "step": 61388 + }, + { + "epoch": 2.36, + "learning_rate": 1.0670043600923313e-05, + "loss": 0.3028, + "step": 61389 + }, + { + "epoch": 2.36, + "learning_rate": 1.0669402410874583e-05, + "loss": 0.0906, + "step": 61390 + }, + { + "epoch": 2.36, + "learning_rate": 1.0668761220825853e-05, + "loss": 0.2557, + "step": 61391 + }, + { + "epoch": 2.36, + "learning_rate": 1.0668120030777124e-05, + "loss": 0.1462, + "step": 61392 + }, + { + "epoch": 2.36, + "learning_rate": 1.0667478840728392e-05, + "loss": 0.269, + "step": 61393 + }, + { + "epoch": 2.36, + "learning_rate": 1.0666837650679662e-05, + "loss": 0.2576, + "step": 61394 + }, + { + "epoch": 2.36, + "learning_rate": 1.0666196460630932e-05, + "loss": 0.2601, + "step": 61395 + }, + { + "epoch": 2.36, + "learning_rate": 1.06655552705822e-05, + "loss": 0.1988, + "step": 61396 + }, + { + "epoch": 2.36, + "learning_rate": 1.066491408053347e-05, + "loss": 0.3013, + "step": 61397 + }, + { + "epoch": 2.36, + "learning_rate": 1.066427289048474e-05, + "loss": 0.147, + "step": 61398 + }, + { + "epoch": 2.36, + "learning_rate": 1.066363170043601e-05, + "loss": 0.2271, + "step": 61399 + }, + { + "epoch": 2.36, + "learning_rate": 1.0662990510387279e-05, + "loss": 0.3104, + "step": 61400 + }, + { + "epoch": 2.36, + "learning_rate": 1.0662349320338549e-05, + "loss": 0.1678, + "step": 61401 + }, + { + "epoch": 2.36, + "learning_rate": 1.0661708130289819e-05, + "loss": 0.3656, + "step": 61402 + }, + { + "epoch": 2.36, + "learning_rate": 1.0661066940241089e-05, + "loss": 0.2412, + "step": 61403 + }, + { + "epoch": 2.36, + "learning_rate": 1.0660425750192357e-05, + "loss": 0.1383, + "step": 61404 + }, + { + "epoch": 2.36, + "learning_rate": 1.0659784560143627e-05, + "loss": 0.2236, + "step": 61405 + }, + { + "epoch": 2.36, + "learning_rate": 1.0659143370094897e-05, + "loss": 0.2808, + "step": 61406 + }, + { + "epoch": 2.36, + "learning_rate": 1.0658502180046167e-05, + "loss": 0.2149, + "step": 61407 + }, + { + "epoch": 2.36, + "learning_rate": 1.0657860989997435e-05, + "loss": 0.185, + "step": 61408 + }, + { + "epoch": 2.36, + "learning_rate": 1.0657219799948705e-05, + "loss": 0.1776, + "step": 61409 + }, + { + "epoch": 2.36, + "learning_rate": 1.0656578609899975e-05, + "loss": 0.1605, + "step": 61410 + }, + { + "epoch": 2.36, + "learning_rate": 1.0655937419851244e-05, + "loss": 0.1481, + "step": 61411 + }, + { + "epoch": 2.36, + "learning_rate": 1.0655296229802514e-05, + "loss": 0.2673, + "step": 61412 + }, + { + "epoch": 2.36, + "learning_rate": 1.0654655039753784e-05, + "loss": 0.322, + "step": 61413 + }, + { + "epoch": 2.36, + "learning_rate": 1.0654013849705054e-05, + "loss": 0.2675, + "step": 61414 + }, + { + "epoch": 2.36, + "learning_rate": 1.0653372659656322e-05, + "loss": 0.1868, + "step": 61415 + }, + { + "epoch": 2.36, + "learning_rate": 1.0652731469607592e-05, + "loss": 0.166, + "step": 61416 + }, + { + "epoch": 2.36, + "learning_rate": 1.0652090279558862e-05, + "loss": 0.1456, + "step": 61417 + }, + { + "epoch": 2.36, + "learning_rate": 1.065144908951013e-05, + "loss": 0.2291, + "step": 61418 + }, + { + "epoch": 2.36, + "learning_rate": 1.06508078994614e-05, + "loss": 0.1585, + "step": 61419 + }, + { + "epoch": 2.36, + "learning_rate": 1.065016670941267e-05, + "loss": 0.2709, + "step": 61420 + }, + { + "epoch": 2.36, + "learning_rate": 1.064952551936394e-05, + "loss": 0.1517, + "step": 61421 + }, + { + "epoch": 2.36, + "learning_rate": 1.0648884329315209e-05, + "loss": 0.2609, + "step": 61422 + }, + { + "epoch": 2.36, + "learning_rate": 1.0648243139266479e-05, + "loss": 0.1956, + "step": 61423 + }, + { + "epoch": 2.36, + "learning_rate": 1.0647601949217749e-05, + "loss": 0.2192, + "step": 61424 + }, + { + "epoch": 2.36, + "learning_rate": 1.0646960759169019e-05, + "loss": 0.1751, + "step": 61425 + }, + { + "epoch": 2.36, + "learning_rate": 1.0646319569120287e-05, + "loss": 0.2048, + "step": 61426 + }, + { + "epoch": 2.36, + "learning_rate": 1.0645678379071557e-05, + "loss": 0.3003, + "step": 61427 + }, + { + "epoch": 2.36, + "learning_rate": 1.0645037189022827e-05, + "loss": 0.2145, + "step": 61428 + }, + { + "epoch": 2.36, + "learning_rate": 1.0644395998974097e-05, + "loss": 0.1644, + "step": 61429 + }, + { + "epoch": 2.36, + "learning_rate": 1.0643754808925366e-05, + "loss": 0.2481, + "step": 61430 + }, + { + "epoch": 2.36, + "learning_rate": 1.0643113618876636e-05, + "loss": 0.1629, + "step": 61431 + }, + { + "epoch": 2.36, + "learning_rate": 1.0642472428827906e-05, + "loss": 0.177, + "step": 61432 + }, + { + "epoch": 2.36, + "learning_rate": 1.0641831238779174e-05, + "loss": 0.294, + "step": 61433 + }, + { + "epoch": 2.36, + "learning_rate": 1.0641190048730444e-05, + "loss": 0.3023, + "step": 61434 + }, + { + "epoch": 2.36, + "learning_rate": 1.0640548858681714e-05, + "loss": 0.2929, + "step": 61435 + }, + { + "epoch": 2.36, + "learning_rate": 1.0639907668632982e-05, + "loss": 0.2799, + "step": 61436 + }, + { + "epoch": 2.36, + "learning_rate": 1.0639266478584253e-05, + "loss": 0.2036, + "step": 61437 + }, + { + "epoch": 2.36, + "learning_rate": 1.0638625288535523e-05, + "loss": 0.3594, + "step": 61438 + }, + { + "epoch": 2.36, + "learning_rate": 1.0637984098486793e-05, + "loss": 0.1996, + "step": 61439 + }, + { + "epoch": 2.36, + "learning_rate": 1.0637342908438061e-05, + "loss": 0.3418, + "step": 61440 + }, + { + "epoch": 2.36, + "learning_rate": 1.0636701718389331e-05, + "loss": 0.2347, + "step": 61441 + }, + { + "epoch": 2.36, + "learning_rate": 1.0636060528340601e-05, + "loss": 0.3223, + "step": 61442 + }, + { + "epoch": 2.36, + "learning_rate": 1.0635419338291871e-05, + "loss": 0.2712, + "step": 61443 + }, + { + "epoch": 2.36, + "learning_rate": 1.063477814824314e-05, + "loss": 0.154, + "step": 61444 + }, + { + "epoch": 2.36, + "learning_rate": 1.063413695819441e-05, + "loss": 0.1242, + "step": 61445 + }, + { + "epoch": 2.36, + "learning_rate": 1.063349576814568e-05, + "loss": 0.2236, + "step": 61446 + }, + { + "epoch": 2.36, + "learning_rate": 1.063285457809695e-05, + "loss": 0.1861, + "step": 61447 + }, + { + "epoch": 2.36, + "learning_rate": 1.0632213388048218e-05, + "loss": 0.4689, + "step": 61448 + }, + { + "epoch": 2.36, + "learning_rate": 1.0631572197999488e-05, + "loss": 0.2419, + "step": 61449 + }, + { + "epoch": 2.36, + "learning_rate": 1.0630931007950758e-05, + "loss": 0.289, + "step": 61450 + }, + { + "epoch": 2.36, + "learning_rate": 1.0630289817902026e-05, + "loss": 0.1632, + "step": 61451 + }, + { + "epoch": 2.36, + "learning_rate": 1.0629648627853296e-05, + "loss": 0.2303, + "step": 61452 + }, + { + "epoch": 2.36, + "learning_rate": 1.0629007437804566e-05, + "loss": 0.1881, + "step": 61453 + }, + { + "epoch": 2.36, + "learning_rate": 1.0628366247755836e-05, + "loss": 0.1961, + "step": 61454 + }, + { + "epoch": 2.36, + "learning_rate": 1.0627725057707104e-05, + "loss": 0.1847, + "step": 61455 + }, + { + "epoch": 2.36, + "learning_rate": 1.0627083867658374e-05, + "loss": 0.1201, + "step": 61456 + }, + { + "epoch": 2.36, + "learning_rate": 1.0626442677609644e-05, + "loss": 0.2215, + "step": 61457 + }, + { + "epoch": 2.36, + "learning_rate": 1.0625801487560913e-05, + "loss": 0.1753, + "step": 61458 + }, + { + "epoch": 2.36, + "learning_rate": 1.0625160297512183e-05, + "loss": 0.1554, + "step": 61459 + }, + { + "epoch": 2.36, + "learning_rate": 1.0624519107463453e-05, + "loss": 0.3771, + "step": 61460 + }, + { + "epoch": 2.36, + "learning_rate": 1.0623877917414723e-05, + "loss": 0.2419, + "step": 61461 + }, + { + "epoch": 2.36, + "learning_rate": 1.0623236727365991e-05, + "loss": 0.3489, + "step": 61462 + }, + { + "epoch": 2.36, + "learning_rate": 1.0622595537317261e-05, + "loss": 0.2059, + "step": 61463 + }, + { + "epoch": 2.36, + "learning_rate": 1.0621954347268531e-05, + "loss": 0.1871, + "step": 61464 + }, + { + "epoch": 2.36, + "learning_rate": 1.0621313157219801e-05, + "loss": 0.1902, + "step": 61465 + }, + { + "epoch": 2.36, + "learning_rate": 1.062067196717107e-05, + "loss": 0.1713, + "step": 61466 + }, + { + "epoch": 2.36, + "learning_rate": 1.062003077712234e-05, + "loss": 0.2026, + "step": 61467 + }, + { + "epoch": 2.36, + "learning_rate": 1.061938958707361e-05, + "loss": 0.1868, + "step": 61468 + }, + { + "epoch": 2.36, + "learning_rate": 1.061874839702488e-05, + "loss": 0.2899, + "step": 61469 + }, + { + "epoch": 2.36, + "learning_rate": 1.0618107206976148e-05, + "loss": 0.1766, + "step": 61470 + }, + { + "epoch": 2.36, + "learning_rate": 1.0617466016927418e-05, + "loss": 0.3651, + "step": 61471 + }, + { + "epoch": 2.36, + "learning_rate": 1.0616824826878688e-05, + "loss": 0.2181, + "step": 61472 + }, + { + "epoch": 2.36, + "learning_rate": 1.0616183636829956e-05, + "loss": 0.2939, + "step": 61473 + }, + { + "epoch": 2.36, + "learning_rate": 1.0615542446781226e-05, + "loss": 0.273, + "step": 61474 + }, + { + "epoch": 2.36, + "learning_rate": 1.0614901256732496e-05, + "loss": 0.3384, + "step": 61475 + }, + { + "epoch": 2.36, + "learning_rate": 1.0614260066683765e-05, + "loss": 0.3519, + "step": 61476 + }, + { + "epoch": 2.36, + "learning_rate": 1.0613618876635035e-05, + "loss": 0.2675, + "step": 61477 + }, + { + "epoch": 2.36, + "learning_rate": 1.0612977686586305e-05, + "loss": 0.2165, + "step": 61478 + }, + { + "epoch": 2.36, + "learning_rate": 1.0612336496537575e-05, + "loss": 0.2445, + "step": 61479 + }, + { + "epoch": 2.36, + "learning_rate": 1.0611695306488843e-05, + "loss": 0.1267, + "step": 61480 + }, + { + "epoch": 2.36, + "learning_rate": 1.0611054116440113e-05, + "loss": 0.1481, + "step": 61481 + }, + { + "epoch": 2.36, + "learning_rate": 1.0610412926391383e-05, + "loss": 0.3526, + "step": 61482 + }, + { + "epoch": 2.36, + "learning_rate": 1.0609771736342653e-05, + "loss": 0.2152, + "step": 61483 + }, + { + "epoch": 2.36, + "learning_rate": 1.0609130546293922e-05, + "loss": 0.2785, + "step": 61484 + }, + { + "epoch": 2.36, + "learning_rate": 1.0608489356245192e-05, + "loss": 0.1918, + "step": 61485 + }, + { + "epoch": 2.36, + "learning_rate": 1.0607848166196462e-05, + "loss": 0.1714, + "step": 61486 + }, + { + "epoch": 2.36, + "learning_rate": 1.0607206976147732e-05, + "loss": 0.1802, + "step": 61487 + }, + { + "epoch": 2.36, + "learning_rate": 1.0606565786099e-05, + "loss": 0.2705, + "step": 61488 + }, + { + "epoch": 2.36, + "learning_rate": 1.060592459605027e-05, + "loss": 0.3193, + "step": 61489 + }, + { + "epoch": 2.36, + "learning_rate": 1.060528340600154e-05, + "loss": 0.218, + "step": 61490 + }, + { + "epoch": 2.36, + "learning_rate": 1.0604642215952808e-05, + "loss": 0.2115, + "step": 61491 + }, + { + "epoch": 2.36, + "learning_rate": 1.0604001025904078e-05, + "loss": 0.3022, + "step": 61492 + }, + { + "epoch": 2.36, + "learning_rate": 1.0603359835855348e-05, + "loss": 0.3551, + "step": 61493 + }, + { + "epoch": 2.36, + "learning_rate": 1.0602718645806618e-05, + "loss": 0.4016, + "step": 61494 + }, + { + "epoch": 2.36, + "learning_rate": 1.0602077455757887e-05, + "loss": 0.1996, + "step": 61495 + }, + { + "epoch": 2.36, + "learning_rate": 1.0601436265709157e-05, + "loss": 0.2119, + "step": 61496 + }, + { + "epoch": 2.36, + "learning_rate": 1.0600795075660427e-05, + "loss": 0.2211, + "step": 61497 + }, + { + "epoch": 2.36, + "learning_rate": 1.0600153885611695e-05, + "loss": 0.3496, + "step": 61498 + }, + { + "epoch": 2.36, + "learning_rate": 1.0599512695562965e-05, + "loss": 0.3174, + "step": 61499 + }, + { + "epoch": 2.36, + "learning_rate": 1.0598871505514235e-05, + "loss": 0.2665, + "step": 61500 + }, + { + "epoch": 2.36, + "learning_rate": 1.0598230315465505e-05, + "loss": 0.246, + "step": 61501 + }, + { + "epoch": 2.36, + "learning_rate": 1.0597589125416773e-05, + "loss": 0.2047, + "step": 61502 + }, + { + "epoch": 2.36, + "learning_rate": 1.0596947935368043e-05, + "loss": 0.2041, + "step": 61503 + }, + { + "epoch": 2.36, + "learning_rate": 1.0596306745319314e-05, + "loss": 0.2891, + "step": 61504 + }, + { + "epoch": 2.36, + "learning_rate": 1.0595665555270584e-05, + "loss": 0.2423, + "step": 61505 + }, + { + "epoch": 2.36, + "learning_rate": 1.0595024365221852e-05, + "loss": 0.2272, + "step": 61506 + }, + { + "epoch": 2.36, + "learning_rate": 1.0594383175173122e-05, + "loss": 0.1863, + "step": 61507 + }, + { + "epoch": 2.36, + "learning_rate": 1.0593741985124392e-05, + "loss": 0.2067, + "step": 61508 + }, + { + "epoch": 2.36, + "learning_rate": 1.0593100795075662e-05, + "loss": 0.1339, + "step": 61509 + }, + { + "epoch": 2.36, + "learning_rate": 1.059245960502693e-05, + "loss": 0.361, + "step": 61510 + }, + { + "epoch": 2.36, + "learning_rate": 1.05918184149782e-05, + "loss": 0.1717, + "step": 61511 + }, + { + "epoch": 2.36, + "learning_rate": 1.059117722492947e-05, + "loss": 0.2122, + "step": 61512 + }, + { + "epoch": 2.36, + "learning_rate": 1.0590536034880739e-05, + "loss": 0.3356, + "step": 61513 + }, + { + "epoch": 2.37, + "learning_rate": 1.0589894844832009e-05, + "loss": 0.1718, + "step": 61514 + }, + { + "epoch": 2.37, + "learning_rate": 1.0589253654783279e-05, + "loss": 0.3277, + "step": 61515 + }, + { + "epoch": 2.37, + "learning_rate": 1.0588612464734547e-05, + "loss": 0.1873, + "step": 61516 + }, + { + "epoch": 2.37, + "learning_rate": 1.0587971274685817e-05, + "loss": 0.1379, + "step": 61517 + }, + { + "epoch": 2.37, + "learning_rate": 1.0587330084637087e-05, + "loss": 0.2394, + "step": 61518 + }, + { + "epoch": 2.37, + "learning_rate": 1.0586688894588357e-05, + "loss": 0.1235, + "step": 61519 + }, + { + "epoch": 2.37, + "learning_rate": 1.0586047704539625e-05, + "loss": 0.3084, + "step": 61520 + }, + { + "epoch": 2.37, + "learning_rate": 1.0585406514490895e-05, + "loss": 0.1597, + "step": 61521 + }, + { + "epoch": 2.37, + "learning_rate": 1.0584765324442165e-05, + "loss": 0.1965, + "step": 61522 + }, + { + "epoch": 2.37, + "learning_rate": 1.0584124134393435e-05, + "loss": 0.1857, + "step": 61523 + }, + { + "epoch": 2.37, + "learning_rate": 1.0583482944344704e-05, + "loss": 0.2333, + "step": 61524 + }, + { + "epoch": 2.37, + "learning_rate": 1.0582841754295974e-05, + "loss": 0.21, + "step": 61525 + }, + { + "epoch": 2.37, + "learning_rate": 1.0582200564247244e-05, + "loss": 0.2006, + "step": 61526 + }, + { + "epoch": 2.37, + "learning_rate": 1.0581559374198514e-05, + "loss": 0.237, + "step": 61527 + }, + { + "epoch": 2.37, + "learning_rate": 1.0580918184149782e-05, + "loss": 0.2335, + "step": 61528 + }, + { + "epoch": 2.37, + "learning_rate": 1.0580276994101052e-05, + "loss": 0.2403, + "step": 61529 + }, + { + "epoch": 2.37, + "learning_rate": 1.0579635804052322e-05, + "loss": 0.154, + "step": 61530 + }, + { + "epoch": 2.37, + "learning_rate": 1.0578994614003592e-05, + "loss": 0.1276, + "step": 61531 + }, + { + "epoch": 2.37, + "learning_rate": 1.057835342395486e-05, + "loss": 0.1614, + "step": 61532 + }, + { + "epoch": 2.37, + "learning_rate": 1.057771223390613e-05, + "loss": 0.165, + "step": 61533 + }, + { + "epoch": 2.37, + "learning_rate": 1.0577071043857399e-05, + "loss": 0.1283, + "step": 61534 + }, + { + "epoch": 2.37, + "learning_rate": 1.0576429853808669e-05, + "loss": 0.2874, + "step": 61535 + }, + { + "epoch": 2.37, + "learning_rate": 1.0575788663759939e-05, + "loss": 0.1063, + "step": 61536 + }, + { + "epoch": 2.37, + "learning_rate": 1.0575147473711209e-05, + "loss": 0.2866, + "step": 61537 + }, + { + "epoch": 2.37, + "learning_rate": 1.0574506283662477e-05, + "loss": 0.2378, + "step": 61538 + }, + { + "epoch": 2.37, + "learning_rate": 1.0573865093613747e-05, + "loss": 0.2527, + "step": 61539 + }, + { + "epoch": 2.37, + "learning_rate": 1.0573223903565017e-05, + "loss": 0.2026, + "step": 61540 + }, + { + "epoch": 2.37, + "learning_rate": 1.0572582713516287e-05, + "loss": 0.2687, + "step": 61541 + }, + { + "epoch": 2.37, + "learning_rate": 1.0571941523467556e-05, + "loss": 0.3196, + "step": 61542 + }, + { + "epoch": 2.37, + "learning_rate": 1.0571300333418826e-05, + "loss": 0.19, + "step": 61543 + }, + { + "epoch": 2.37, + "learning_rate": 1.0570659143370096e-05, + "loss": 0.35, + "step": 61544 + }, + { + "epoch": 2.37, + "learning_rate": 1.0570017953321366e-05, + "loss": 0.1904, + "step": 61545 + }, + { + "epoch": 2.37, + "learning_rate": 1.0569376763272634e-05, + "loss": 0.4306, + "step": 61546 + }, + { + "epoch": 2.37, + "learning_rate": 1.0568735573223904e-05, + "loss": 0.2645, + "step": 61547 + }, + { + "epoch": 2.37, + "learning_rate": 1.0568094383175174e-05, + "loss": 0.1607, + "step": 61548 + }, + { + "epoch": 2.37, + "learning_rate": 1.0567453193126444e-05, + "loss": 0.2401, + "step": 61549 + }, + { + "epoch": 2.37, + "learning_rate": 1.0566812003077713e-05, + "loss": 0.2678, + "step": 61550 + }, + { + "epoch": 2.37, + "learning_rate": 1.0566170813028983e-05, + "loss": 0.1955, + "step": 61551 + }, + { + "epoch": 2.37, + "learning_rate": 1.0565529622980253e-05, + "loss": 0.2398, + "step": 61552 + }, + { + "epoch": 2.37, + "learning_rate": 1.0564888432931521e-05, + "loss": 0.3772, + "step": 61553 + }, + { + "epoch": 2.37, + "learning_rate": 1.0564247242882791e-05, + "loss": 0.2462, + "step": 61554 + }, + { + "epoch": 2.37, + "learning_rate": 1.0563606052834061e-05, + "loss": 0.2574, + "step": 61555 + }, + { + "epoch": 2.37, + "learning_rate": 1.056296486278533e-05, + "loss": 0.2505, + "step": 61556 + }, + { + "epoch": 2.37, + "learning_rate": 1.05623236727366e-05, + "loss": 0.2405, + "step": 61557 + }, + { + "epoch": 2.37, + "learning_rate": 1.056168248268787e-05, + "loss": 0.2843, + "step": 61558 + }, + { + "epoch": 2.37, + "learning_rate": 1.056104129263914e-05, + "loss": 0.2287, + "step": 61559 + }, + { + "epoch": 2.37, + "learning_rate": 1.0560400102590408e-05, + "loss": 0.2748, + "step": 61560 + }, + { + "epoch": 2.37, + "learning_rate": 1.0559758912541678e-05, + "loss": 0.2859, + "step": 61561 + }, + { + "epoch": 2.37, + "learning_rate": 1.0559117722492948e-05, + "loss": 0.28, + "step": 61562 + }, + { + "epoch": 2.37, + "learning_rate": 1.0558476532444218e-05, + "loss": 0.1835, + "step": 61563 + }, + { + "epoch": 2.37, + "learning_rate": 1.0557835342395486e-05, + "loss": 0.1617, + "step": 61564 + }, + { + "epoch": 2.37, + "learning_rate": 1.0557194152346756e-05, + "loss": 0.2684, + "step": 61565 + }, + { + "epoch": 2.37, + "learning_rate": 1.0556552962298026e-05, + "loss": 0.3314, + "step": 61566 + }, + { + "epoch": 2.37, + "learning_rate": 1.0555911772249296e-05, + "loss": 0.3111, + "step": 61567 + }, + { + "epoch": 2.37, + "learning_rate": 1.0555270582200564e-05, + "loss": 0.2211, + "step": 61568 + }, + { + "epoch": 2.37, + "learning_rate": 1.0554629392151834e-05, + "loss": 0.2001, + "step": 61569 + }, + { + "epoch": 2.37, + "learning_rate": 1.0553988202103105e-05, + "loss": 0.2447, + "step": 61570 + }, + { + "epoch": 2.37, + "learning_rate": 1.0553347012054375e-05, + "loss": 0.3, + "step": 61571 + }, + { + "epoch": 2.37, + "learning_rate": 1.0552705822005643e-05, + "loss": 0.1992, + "step": 61572 + }, + { + "epoch": 2.37, + "learning_rate": 1.0552064631956913e-05, + "loss": 0.2704, + "step": 61573 + }, + { + "epoch": 2.37, + "learning_rate": 1.0551423441908181e-05, + "loss": 0.2972, + "step": 61574 + }, + { + "epoch": 2.37, + "learning_rate": 1.0550782251859451e-05, + "loss": 0.1541, + "step": 61575 + }, + { + "epoch": 2.37, + "learning_rate": 1.0550141061810721e-05, + "loss": 0.2426, + "step": 61576 + }, + { + "epoch": 2.37, + "learning_rate": 1.0549499871761991e-05, + "loss": 0.2184, + "step": 61577 + }, + { + "epoch": 2.37, + "learning_rate": 1.054885868171326e-05, + "loss": 0.1509, + "step": 61578 + }, + { + "epoch": 2.37, + "learning_rate": 1.054821749166453e-05, + "loss": 0.1989, + "step": 61579 + }, + { + "epoch": 2.37, + "learning_rate": 1.05475763016158e-05, + "loss": 0.1622, + "step": 61580 + }, + { + "epoch": 2.37, + "learning_rate": 1.054693511156707e-05, + "loss": 0.2031, + "step": 61581 + }, + { + "epoch": 2.37, + "learning_rate": 1.0546293921518338e-05, + "loss": 0.1774, + "step": 61582 + }, + { + "epoch": 2.37, + "learning_rate": 1.0545652731469608e-05, + "loss": 0.1327, + "step": 61583 + }, + { + "epoch": 2.37, + "learning_rate": 1.0545011541420878e-05, + "loss": 0.2587, + "step": 61584 + }, + { + "epoch": 2.37, + "learning_rate": 1.0544370351372148e-05, + "loss": 0.1308, + "step": 61585 + }, + { + "epoch": 2.37, + "learning_rate": 1.0543729161323416e-05, + "loss": 0.1776, + "step": 61586 + }, + { + "epoch": 2.37, + "learning_rate": 1.0543087971274686e-05, + "loss": 0.221, + "step": 61587 + }, + { + "epoch": 2.37, + "learning_rate": 1.0542446781225956e-05, + "loss": 0.1569, + "step": 61588 + }, + { + "epoch": 2.37, + "learning_rate": 1.0541805591177226e-05, + "loss": 0.1161, + "step": 61589 + }, + { + "epoch": 2.37, + "learning_rate": 1.0541164401128495e-05, + "loss": 0.1237, + "step": 61590 + }, + { + "epoch": 2.37, + "learning_rate": 1.0540523211079765e-05, + "loss": 0.1822, + "step": 61591 + }, + { + "epoch": 2.37, + "learning_rate": 1.0539882021031035e-05, + "loss": 0.1601, + "step": 61592 + }, + { + "epoch": 2.37, + "learning_rate": 1.0539240830982305e-05, + "loss": 0.1983, + "step": 61593 + }, + { + "epoch": 2.37, + "learning_rate": 1.0538599640933573e-05, + "loss": 0.1468, + "step": 61594 + }, + { + "epoch": 2.37, + "learning_rate": 1.0537958450884843e-05, + "loss": 0.1611, + "step": 61595 + }, + { + "epoch": 2.37, + "learning_rate": 1.0537317260836112e-05, + "loss": 0.2014, + "step": 61596 + }, + { + "epoch": 2.37, + "learning_rate": 1.0536676070787382e-05, + "loss": 0.2948, + "step": 61597 + }, + { + "epoch": 2.37, + "learning_rate": 1.0536034880738652e-05, + "loss": 0.1226, + "step": 61598 + }, + { + "epoch": 2.37, + "learning_rate": 1.0535393690689922e-05, + "loss": 0.2692, + "step": 61599 + }, + { + "epoch": 2.37, + "learning_rate": 1.053475250064119e-05, + "loss": 0.3255, + "step": 61600 + }, + { + "epoch": 2.37, + "learning_rate": 1.053411131059246e-05, + "loss": 0.2404, + "step": 61601 + }, + { + "epoch": 2.37, + "learning_rate": 1.053347012054373e-05, + "loss": 0.1582, + "step": 61602 + }, + { + "epoch": 2.37, + "learning_rate": 1.0532828930495e-05, + "loss": 0.2841, + "step": 61603 + }, + { + "epoch": 2.37, + "learning_rate": 1.0532187740446268e-05, + "loss": 0.1751, + "step": 61604 + }, + { + "epoch": 2.37, + "learning_rate": 1.0531546550397538e-05, + "loss": 0.2175, + "step": 61605 + }, + { + "epoch": 2.37, + "learning_rate": 1.0530905360348808e-05, + "loss": 0.179, + "step": 61606 + }, + { + "epoch": 2.37, + "learning_rate": 1.0530264170300078e-05, + "loss": 0.1245, + "step": 61607 + }, + { + "epoch": 2.37, + "learning_rate": 1.0529622980251347e-05, + "loss": 0.1522, + "step": 61608 + }, + { + "epoch": 2.37, + "learning_rate": 1.0528981790202617e-05, + "loss": 0.1452, + "step": 61609 + }, + { + "epoch": 2.37, + "learning_rate": 1.0528340600153887e-05, + "loss": 0.4099, + "step": 61610 + }, + { + "epoch": 2.37, + "learning_rate": 1.0527699410105157e-05, + "loss": 0.2963, + "step": 61611 + }, + { + "epoch": 2.37, + "learning_rate": 1.0527058220056425e-05, + "loss": 0.3805, + "step": 61612 + }, + { + "epoch": 2.37, + "learning_rate": 1.0526417030007695e-05, + "loss": 0.2249, + "step": 61613 + }, + { + "epoch": 2.37, + "learning_rate": 1.0525775839958963e-05, + "loss": 0.2737, + "step": 61614 + }, + { + "epoch": 2.37, + "learning_rate": 1.0525134649910234e-05, + "loss": 0.3407, + "step": 61615 + }, + { + "epoch": 2.37, + "learning_rate": 1.0524493459861504e-05, + "loss": 0.1196, + "step": 61616 + }, + { + "epoch": 2.37, + "learning_rate": 1.0523852269812774e-05, + "loss": 0.2236, + "step": 61617 + }, + { + "epoch": 2.37, + "learning_rate": 1.0523211079764042e-05, + "loss": 0.1822, + "step": 61618 + }, + { + "epoch": 2.37, + "learning_rate": 1.0522569889715312e-05, + "loss": 0.3007, + "step": 61619 + }, + { + "epoch": 2.37, + "learning_rate": 1.0521928699666582e-05, + "loss": 0.227, + "step": 61620 + }, + { + "epoch": 2.37, + "learning_rate": 1.0521287509617852e-05, + "loss": 0.1753, + "step": 61621 + }, + { + "epoch": 2.37, + "learning_rate": 1.052064631956912e-05, + "loss": 0.2443, + "step": 61622 + }, + { + "epoch": 2.37, + "learning_rate": 1.052000512952039e-05, + "loss": 0.2059, + "step": 61623 + }, + { + "epoch": 2.37, + "learning_rate": 1.051936393947166e-05, + "loss": 0.1981, + "step": 61624 + }, + { + "epoch": 2.37, + "learning_rate": 1.051872274942293e-05, + "loss": 0.2591, + "step": 61625 + }, + { + "epoch": 2.37, + "learning_rate": 1.0518081559374199e-05, + "loss": 0.1755, + "step": 61626 + }, + { + "epoch": 2.37, + "learning_rate": 1.0517440369325469e-05, + "loss": 0.1758, + "step": 61627 + }, + { + "epoch": 2.37, + "learning_rate": 1.0516799179276739e-05, + "loss": 0.1786, + "step": 61628 + }, + { + "epoch": 2.37, + "learning_rate": 1.0516157989228009e-05, + "loss": 0.1282, + "step": 61629 + }, + { + "epoch": 2.37, + "learning_rate": 1.0515516799179277e-05, + "loss": 0.1193, + "step": 61630 + }, + { + "epoch": 2.37, + "learning_rate": 1.0514875609130547e-05, + "loss": 0.4042, + "step": 61631 + }, + { + "epoch": 2.37, + "learning_rate": 1.0514234419081815e-05, + "loss": 0.1538, + "step": 61632 + }, + { + "epoch": 2.37, + "learning_rate": 1.0513593229033087e-05, + "loss": 0.2311, + "step": 61633 + }, + { + "epoch": 2.37, + "learning_rate": 1.0512952038984355e-05, + "loss": 0.1894, + "step": 61634 + }, + { + "epoch": 2.37, + "learning_rate": 1.0512310848935625e-05, + "loss": 0.1751, + "step": 61635 + }, + { + "epoch": 2.37, + "learning_rate": 1.0511669658886894e-05, + "loss": 0.2281, + "step": 61636 + }, + { + "epoch": 2.37, + "learning_rate": 1.0511028468838164e-05, + "loss": 0.1409, + "step": 61637 + }, + { + "epoch": 2.37, + "learning_rate": 1.0510387278789434e-05, + "loss": 0.1538, + "step": 61638 + }, + { + "epoch": 2.37, + "learning_rate": 1.0509746088740704e-05, + "loss": 0.1922, + "step": 61639 + }, + { + "epoch": 2.37, + "learning_rate": 1.0509104898691972e-05, + "loss": 0.1433, + "step": 61640 + }, + { + "epoch": 2.37, + "learning_rate": 1.0508463708643242e-05, + "loss": 0.1715, + "step": 61641 + }, + { + "epoch": 2.37, + "learning_rate": 1.0507822518594512e-05, + "loss": 0.2319, + "step": 61642 + }, + { + "epoch": 2.37, + "learning_rate": 1.0507181328545782e-05, + "loss": 0.1852, + "step": 61643 + }, + { + "epoch": 2.37, + "learning_rate": 1.050654013849705e-05, + "loss": 0.2396, + "step": 61644 + }, + { + "epoch": 2.37, + "learning_rate": 1.050589894844832e-05, + "loss": 0.1368, + "step": 61645 + }, + { + "epoch": 2.37, + "learning_rate": 1.050525775839959e-05, + "loss": 0.397, + "step": 61646 + }, + { + "epoch": 2.37, + "learning_rate": 1.050461656835086e-05, + "loss": 0.3278, + "step": 61647 + }, + { + "epoch": 2.37, + "learning_rate": 1.0503975378302129e-05, + "loss": 0.1797, + "step": 61648 + }, + { + "epoch": 2.37, + "learning_rate": 1.0503334188253399e-05, + "loss": 0.2442, + "step": 61649 + }, + { + "epoch": 2.37, + "learning_rate": 1.0502692998204669e-05, + "loss": 0.1425, + "step": 61650 + }, + { + "epoch": 2.37, + "learning_rate": 1.0502051808155939e-05, + "loss": 0.237, + "step": 61651 + }, + { + "epoch": 2.37, + "learning_rate": 1.0501410618107207e-05, + "loss": 0.3067, + "step": 61652 + }, + { + "epoch": 2.37, + "learning_rate": 1.0500769428058477e-05, + "loss": 0.2166, + "step": 61653 + }, + { + "epoch": 2.37, + "learning_rate": 1.0500128238009746e-05, + "loss": 0.118, + "step": 61654 + }, + { + "epoch": 2.37, + "learning_rate": 1.0499487047961016e-05, + "loss": 0.2077, + "step": 61655 + }, + { + "epoch": 2.37, + "learning_rate": 1.0498845857912286e-05, + "loss": 0.2042, + "step": 61656 + }, + { + "epoch": 2.37, + "learning_rate": 1.0498204667863556e-05, + "loss": 0.2364, + "step": 61657 + }, + { + "epoch": 2.37, + "learning_rate": 1.0497563477814824e-05, + "loss": 0.1876, + "step": 61658 + }, + { + "epoch": 2.37, + "learning_rate": 1.0496922287766094e-05, + "loss": 0.2791, + "step": 61659 + }, + { + "epoch": 2.37, + "learning_rate": 1.0496281097717364e-05, + "loss": 0.2739, + "step": 61660 + }, + { + "epoch": 2.37, + "learning_rate": 1.0495639907668634e-05, + "loss": 0.1832, + "step": 61661 + }, + { + "epoch": 2.37, + "learning_rate": 1.0494998717619903e-05, + "loss": 0.2865, + "step": 61662 + }, + { + "epoch": 2.37, + "learning_rate": 1.0494357527571173e-05, + "loss": 0.2831, + "step": 61663 + }, + { + "epoch": 2.37, + "learning_rate": 1.0493716337522443e-05, + "loss": 0.332, + "step": 61664 + }, + { + "epoch": 2.37, + "learning_rate": 1.0493075147473713e-05, + "loss": 0.2369, + "step": 61665 + }, + { + "epoch": 2.37, + "learning_rate": 1.0492433957424981e-05, + "loss": 0.2505, + "step": 61666 + }, + { + "epoch": 2.37, + "learning_rate": 1.0491792767376251e-05, + "loss": 0.2009, + "step": 61667 + }, + { + "epoch": 2.37, + "learning_rate": 1.0491151577327521e-05, + "loss": 0.2851, + "step": 61668 + }, + { + "epoch": 2.37, + "learning_rate": 1.0490510387278791e-05, + "loss": 0.1922, + "step": 61669 + }, + { + "epoch": 2.37, + "learning_rate": 1.048986919723006e-05, + "loss": 0.2055, + "step": 61670 + }, + { + "epoch": 2.37, + "learning_rate": 1.048922800718133e-05, + "loss": 0.3595, + "step": 61671 + }, + { + "epoch": 2.37, + "learning_rate": 1.0488586817132598e-05, + "loss": 0.239, + "step": 61672 + }, + { + "epoch": 2.37, + "learning_rate": 1.048794562708387e-05, + "loss": 0.1894, + "step": 61673 + }, + { + "epoch": 2.37, + "learning_rate": 1.0487304437035138e-05, + "loss": 0.2792, + "step": 61674 + }, + { + "epoch": 2.37, + "learning_rate": 1.0486663246986408e-05, + "loss": 0.2807, + "step": 61675 + }, + { + "epoch": 2.37, + "learning_rate": 1.0486022056937676e-05, + "loss": 0.2619, + "step": 61676 + }, + { + "epoch": 2.37, + "learning_rate": 1.0485380866888946e-05, + "loss": 0.1959, + "step": 61677 + }, + { + "epoch": 2.37, + "learning_rate": 1.0484739676840216e-05, + "loss": 0.3319, + "step": 61678 + }, + { + "epoch": 2.37, + "learning_rate": 1.0484098486791486e-05, + "loss": 0.2761, + "step": 61679 + }, + { + "epoch": 2.37, + "learning_rate": 1.0483457296742754e-05, + "loss": 0.2498, + "step": 61680 + }, + { + "epoch": 2.37, + "learning_rate": 1.0482816106694025e-05, + "loss": 0.1591, + "step": 61681 + }, + { + "epoch": 2.37, + "learning_rate": 1.0482174916645295e-05, + "loss": 0.2053, + "step": 61682 + }, + { + "epoch": 2.37, + "learning_rate": 1.0481533726596565e-05, + "loss": 0.2491, + "step": 61683 + }, + { + "epoch": 2.37, + "learning_rate": 1.0480892536547833e-05, + "loss": 0.2404, + "step": 61684 + }, + { + "epoch": 2.37, + "learning_rate": 1.0480251346499103e-05, + "loss": 0.2337, + "step": 61685 + }, + { + "epoch": 2.37, + "learning_rate": 1.0479610156450373e-05, + "loss": 0.2537, + "step": 61686 + }, + { + "epoch": 2.37, + "learning_rate": 1.0478968966401643e-05, + "loss": 0.2123, + "step": 61687 + }, + { + "epoch": 2.37, + "learning_rate": 1.0478327776352911e-05, + "loss": 0.2243, + "step": 61688 + }, + { + "epoch": 2.37, + "learning_rate": 1.0477686586304181e-05, + "loss": 0.2983, + "step": 61689 + }, + { + "epoch": 2.37, + "learning_rate": 1.0477045396255451e-05, + "loss": 0.2336, + "step": 61690 + }, + { + "epoch": 2.37, + "learning_rate": 1.0476404206206721e-05, + "loss": 0.212, + "step": 61691 + }, + { + "epoch": 2.37, + "learning_rate": 1.047576301615799e-05, + "loss": 0.2491, + "step": 61692 + }, + { + "epoch": 2.37, + "learning_rate": 1.047512182610926e-05, + "loss": 0.1655, + "step": 61693 + }, + { + "epoch": 2.37, + "learning_rate": 1.0474480636060528e-05, + "loss": 0.1289, + "step": 61694 + }, + { + "epoch": 2.37, + "learning_rate": 1.04738394460118e-05, + "loss": 0.2777, + "step": 61695 + }, + { + "epoch": 2.37, + "learning_rate": 1.0473198255963068e-05, + "loss": 0.2215, + "step": 61696 + }, + { + "epoch": 2.37, + "learning_rate": 1.0472557065914338e-05, + "loss": 0.3554, + "step": 61697 + }, + { + "epoch": 2.37, + "learning_rate": 1.0471915875865606e-05, + "loss": 0.2511, + "step": 61698 + }, + { + "epoch": 2.37, + "learning_rate": 1.0471274685816876e-05, + "loss": 0.2516, + "step": 61699 + }, + { + "epoch": 2.37, + "learning_rate": 1.0470633495768146e-05, + "loss": 0.4145, + "step": 61700 + }, + { + "epoch": 2.37, + "learning_rate": 1.0469992305719416e-05, + "loss": 0.1506, + "step": 61701 + }, + { + "epoch": 2.37, + "learning_rate": 1.0469351115670685e-05, + "loss": 0.2144, + "step": 61702 + }, + { + "epoch": 2.37, + "learning_rate": 1.0468709925621955e-05, + "loss": 0.1273, + "step": 61703 + }, + { + "epoch": 2.37, + "learning_rate": 1.0468068735573225e-05, + "loss": 0.1423, + "step": 61704 + }, + { + "epoch": 2.37, + "learning_rate": 1.0467427545524495e-05, + "loss": 0.1257, + "step": 61705 + }, + { + "epoch": 2.37, + "learning_rate": 1.0466786355475763e-05, + "loss": 0.326, + "step": 61706 + }, + { + "epoch": 2.37, + "learning_rate": 1.0466145165427033e-05, + "loss": 0.351, + "step": 61707 + }, + { + "epoch": 2.37, + "learning_rate": 1.0465503975378303e-05, + "loss": 0.2619, + "step": 61708 + }, + { + "epoch": 2.37, + "learning_rate": 1.0464862785329573e-05, + "loss": 0.2281, + "step": 61709 + }, + { + "epoch": 2.37, + "learning_rate": 1.0464221595280842e-05, + "loss": 0.0883, + "step": 61710 + }, + { + "epoch": 2.37, + "learning_rate": 1.0463580405232112e-05, + "loss": 0.2639, + "step": 61711 + }, + { + "epoch": 2.37, + "learning_rate": 1.046293921518338e-05, + "loss": 0.1681, + "step": 61712 + }, + { + "epoch": 2.37, + "learning_rate": 1.0462298025134652e-05, + "loss": 0.0972, + "step": 61713 + }, + { + "epoch": 2.37, + "learning_rate": 1.046165683508592e-05, + "loss": 0.1839, + "step": 61714 + }, + { + "epoch": 2.37, + "learning_rate": 1.046101564503719e-05, + "loss": 0.1071, + "step": 61715 + }, + { + "epoch": 2.37, + "learning_rate": 1.0460374454988458e-05, + "loss": 0.225, + "step": 61716 + }, + { + "epoch": 2.37, + "learning_rate": 1.0459733264939728e-05, + "loss": 0.2458, + "step": 61717 + }, + { + "epoch": 2.37, + "learning_rate": 1.0459092074890998e-05, + "loss": 0.1443, + "step": 61718 + }, + { + "epoch": 2.37, + "learning_rate": 1.0458450884842268e-05, + "loss": 0.2856, + "step": 61719 + }, + { + "epoch": 2.37, + "learning_rate": 1.0457809694793537e-05, + "loss": 0.2096, + "step": 61720 + }, + { + "epoch": 2.37, + "learning_rate": 1.0457168504744807e-05, + "loss": 0.1868, + "step": 61721 + }, + { + "epoch": 2.37, + "learning_rate": 1.0456527314696077e-05, + "loss": 0.2816, + "step": 61722 + }, + { + "epoch": 2.37, + "learning_rate": 1.0455886124647347e-05, + "loss": 0.2555, + "step": 61723 + }, + { + "epoch": 2.37, + "learning_rate": 1.0455244934598615e-05, + "loss": 0.2249, + "step": 61724 + }, + { + "epoch": 2.37, + "learning_rate": 1.0454603744549885e-05, + "loss": 0.3203, + "step": 61725 + }, + { + "epoch": 2.37, + "learning_rate": 1.0453962554501155e-05, + "loss": 0.2051, + "step": 61726 + }, + { + "epoch": 2.37, + "learning_rate": 1.0453321364452425e-05, + "loss": 0.1348, + "step": 61727 + }, + { + "epoch": 2.37, + "learning_rate": 1.0452680174403694e-05, + "loss": 0.2716, + "step": 61728 + }, + { + "epoch": 2.37, + "learning_rate": 1.0452038984354964e-05, + "loss": 0.3662, + "step": 61729 + }, + { + "epoch": 2.37, + "learning_rate": 1.0451397794306234e-05, + "loss": 0.2026, + "step": 61730 + }, + { + "epoch": 2.37, + "learning_rate": 1.0450756604257504e-05, + "loss": 0.149, + "step": 61731 + }, + { + "epoch": 2.37, + "learning_rate": 1.0450115414208772e-05, + "loss": 0.164, + "step": 61732 + }, + { + "epoch": 2.37, + "learning_rate": 1.0449474224160042e-05, + "loss": 0.2898, + "step": 61733 + }, + { + "epoch": 2.37, + "learning_rate": 1.044883303411131e-05, + "loss": 0.3201, + "step": 61734 + }, + { + "epoch": 2.37, + "learning_rate": 1.0448191844062582e-05, + "loss": 0.1052, + "step": 61735 + }, + { + "epoch": 2.37, + "learning_rate": 1.044755065401385e-05, + "loss": 0.1674, + "step": 61736 + }, + { + "epoch": 2.37, + "learning_rate": 1.044690946396512e-05, + "loss": 0.1596, + "step": 61737 + }, + { + "epoch": 2.37, + "learning_rate": 1.0446268273916389e-05, + "loss": 0.1681, + "step": 61738 + }, + { + "epoch": 2.37, + "learning_rate": 1.0445627083867659e-05, + "loss": 0.1669, + "step": 61739 + }, + { + "epoch": 2.37, + "learning_rate": 1.0444985893818929e-05, + "loss": 0.21, + "step": 61740 + }, + { + "epoch": 2.37, + "learning_rate": 1.0444344703770199e-05, + "loss": 0.1733, + "step": 61741 + }, + { + "epoch": 2.37, + "learning_rate": 1.0443703513721467e-05, + "loss": 0.3887, + "step": 61742 + }, + { + "epoch": 2.37, + "learning_rate": 1.0443062323672737e-05, + "loss": 0.1195, + "step": 61743 + }, + { + "epoch": 2.37, + "learning_rate": 1.0442421133624007e-05, + "loss": 0.137, + "step": 61744 + }, + { + "epoch": 2.37, + "learning_rate": 1.0441779943575277e-05, + "loss": 0.1773, + "step": 61745 + }, + { + "epoch": 2.37, + "learning_rate": 1.0441138753526545e-05, + "loss": 0.1612, + "step": 61746 + }, + { + "epoch": 2.37, + "learning_rate": 1.0440497563477815e-05, + "loss": 0.2617, + "step": 61747 + }, + { + "epoch": 2.37, + "learning_rate": 1.0439856373429086e-05, + "loss": 0.353, + "step": 61748 + }, + { + "epoch": 2.37, + "learning_rate": 1.0439215183380356e-05, + "loss": 0.3344, + "step": 61749 + }, + { + "epoch": 2.37, + "learning_rate": 1.0438573993331624e-05, + "loss": 0.1256, + "step": 61750 + }, + { + "epoch": 2.37, + "learning_rate": 1.0437932803282894e-05, + "loss": 0.2571, + "step": 61751 + }, + { + "epoch": 2.37, + "learning_rate": 1.0437291613234162e-05, + "loss": 0.1313, + "step": 61752 + }, + { + "epoch": 2.37, + "learning_rate": 1.0436650423185434e-05, + "loss": 0.2135, + "step": 61753 + }, + { + "epoch": 2.37, + "learning_rate": 1.0436009233136702e-05, + "loss": 0.2545, + "step": 61754 + }, + { + "epoch": 2.37, + "learning_rate": 1.0435368043087972e-05, + "loss": 0.1514, + "step": 61755 + }, + { + "epoch": 2.37, + "learning_rate": 1.043472685303924e-05, + "loss": 0.0798, + "step": 61756 + }, + { + "epoch": 2.37, + "learning_rate": 1.043408566299051e-05, + "loss": 0.323, + "step": 61757 + }, + { + "epoch": 2.37, + "learning_rate": 1.043344447294178e-05, + "loss": 0.2269, + "step": 61758 + }, + { + "epoch": 2.37, + "learning_rate": 1.043280328289305e-05, + "loss": 0.1656, + "step": 61759 + }, + { + "epoch": 2.37, + "learning_rate": 1.0432162092844319e-05, + "loss": 0.1239, + "step": 61760 + }, + { + "epoch": 2.37, + "learning_rate": 1.0431520902795589e-05, + "loss": 0.3118, + "step": 61761 + }, + { + "epoch": 2.37, + "learning_rate": 1.0430879712746859e-05, + "loss": 0.2974, + "step": 61762 + }, + { + "epoch": 2.37, + "learning_rate": 1.0430238522698129e-05, + "loss": 0.2924, + "step": 61763 + }, + { + "epoch": 2.37, + "learning_rate": 1.0429597332649397e-05, + "loss": 0.3819, + "step": 61764 + }, + { + "epoch": 2.37, + "learning_rate": 1.0428956142600667e-05, + "loss": 0.2089, + "step": 61765 + }, + { + "epoch": 2.37, + "learning_rate": 1.0428314952551937e-05, + "loss": 0.3925, + "step": 61766 + }, + { + "epoch": 2.37, + "learning_rate": 1.0427673762503207e-05, + "loss": 0.208, + "step": 61767 + }, + { + "epoch": 2.37, + "learning_rate": 1.0427032572454476e-05, + "loss": 0.3177, + "step": 61768 + }, + { + "epoch": 2.37, + "learning_rate": 1.0426391382405746e-05, + "loss": 0.1792, + "step": 61769 + }, + { + "epoch": 2.37, + "learning_rate": 1.0425750192357014e-05, + "loss": 0.2841, + "step": 61770 + }, + { + "epoch": 2.37, + "learning_rate": 1.0425109002308286e-05, + "loss": 0.2404, + "step": 61771 + }, + { + "epoch": 2.37, + "learning_rate": 1.0424467812259554e-05, + "loss": 0.2424, + "step": 61772 + }, + { + "epoch": 2.37, + "learning_rate": 1.0423826622210824e-05, + "loss": 0.2654, + "step": 61773 + }, + { + "epoch": 2.38, + "learning_rate": 1.0423185432162093e-05, + "loss": 0.2481, + "step": 61774 + }, + { + "epoch": 2.38, + "learning_rate": 1.0422544242113364e-05, + "loss": 0.2148, + "step": 61775 + }, + { + "epoch": 2.38, + "learning_rate": 1.0421903052064633e-05, + "loss": 0.3455, + "step": 61776 + }, + { + "epoch": 2.38, + "learning_rate": 1.0421261862015903e-05, + "loss": 0.2875, + "step": 61777 + }, + { + "epoch": 2.38, + "learning_rate": 1.0420620671967171e-05, + "loss": 0.1494, + "step": 61778 + }, + { + "epoch": 2.38, + "learning_rate": 1.0419979481918441e-05, + "loss": 0.1881, + "step": 61779 + }, + { + "epoch": 2.38, + "learning_rate": 1.0419338291869711e-05, + "loss": 0.2351, + "step": 61780 + }, + { + "epoch": 2.38, + "learning_rate": 1.0418697101820981e-05, + "loss": 0.2276, + "step": 61781 + }, + { + "epoch": 2.38, + "learning_rate": 1.041805591177225e-05, + "loss": 0.1851, + "step": 61782 + }, + { + "epoch": 2.38, + "learning_rate": 1.041741472172352e-05, + "loss": 0.1733, + "step": 61783 + }, + { + "epoch": 2.38, + "learning_rate": 1.041677353167479e-05, + "loss": 0.1809, + "step": 61784 + }, + { + "epoch": 2.38, + "learning_rate": 1.041613234162606e-05, + "loss": 0.1842, + "step": 61785 + }, + { + "epoch": 2.38, + "learning_rate": 1.0415491151577328e-05, + "loss": 0.2085, + "step": 61786 + }, + { + "epoch": 2.38, + "learning_rate": 1.0414849961528598e-05, + "loss": 0.1495, + "step": 61787 + }, + { + "epoch": 2.38, + "learning_rate": 1.0414208771479868e-05, + "loss": 0.2711, + "step": 61788 + }, + { + "epoch": 2.38, + "learning_rate": 1.0413567581431138e-05, + "loss": 0.2338, + "step": 61789 + }, + { + "epoch": 2.38, + "learning_rate": 1.0412926391382406e-05, + "loss": 0.2053, + "step": 61790 + }, + { + "epoch": 2.38, + "learning_rate": 1.0412285201333676e-05, + "loss": 0.294, + "step": 61791 + }, + { + "epoch": 2.38, + "learning_rate": 1.0411644011284944e-05, + "loss": 0.2184, + "step": 61792 + }, + { + "epoch": 2.38, + "learning_rate": 1.0411002821236216e-05, + "loss": 0.1009, + "step": 61793 + }, + { + "epoch": 2.38, + "learning_rate": 1.0410361631187485e-05, + "loss": 0.229, + "step": 61794 + }, + { + "epoch": 2.38, + "learning_rate": 1.0409720441138755e-05, + "loss": 0.1865, + "step": 61795 + }, + { + "epoch": 2.38, + "learning_rate": 1.0409079251090023e-05, + "loss": 0.0989, + "step": 61796 + }, + { + "epoch": 2.38, + "learning_rate": 1.0408438061041295e-05, + "loss": 0.2553, + "step": 61797 + }, + { + "epoch": 2.38, + "learning_rate": 1.0407796870992563e-05, + "loss": 0.1508, + "step": 61798 + }, + { + "epoch": 2.38, + "learning_rate": 1.0407155680943833e-05, + "loss": 0.1348, + "step": 61799 + }, + { + "epoch": 2.38, + "learning_rate": 1.0406514490895101e-05, + "loss": 0.3646, + "step": 61800 + }, + { + "epoch": 2.38, + "learning_rate": 1.0405873300846371e-05, + "loss": 0.3102, + "step": 61801 + }, + { + "epoch": 2.38, + "learning_rate": 1.0405232110797641e-05, + "loss": 0.2515, + "step": 61802 + }, + { + "epoch": 2.38, + "learning_rate": 1.0404590920748911e-05, + "loss": 0.2688, + "step": 61803 + }, + { + "epoch": 2.38, + "learning_rate": 1.040394973070018e-05, + "loss": 0.2695, + "step": 61804 + }, + { + "epoch": 2.38, + "learning_rate": 1.040330854065145e-05, + "loss": 0.1932, + "step": 61805 + }, + { + "epoch": 2.38, + "learning_rate": 1.040266735060272e-05, + "loss": 0.2522, + "step": 61806 + }, + { + "epoch": 2.38, + "learning_rate": 1.040202616055399e-05, + "loss": 0.2199, + "step": 61807 + }, + { + "epoch": 2.38, + "learning_rate": 1.0401384970505258e-05, + "loss": 0.2129, + "step": 61808 + }, + { + "epoch": 2.38, + "learning_rate": 1.0400743780456528e-05, + "loss": 0.224, + "step": 61809 + }, + { + "epoch": 2.38, + "learning_rate": 1.0400102590407796e-05, + "loss": 0.2677, + "step": 61810 + }, + { + "epoch": 2.38, + "learning_rate": 1.0399461400359068e-05, + "loss": 0.0736, + "step": 61811 + }, + { + "epoch": 2.38, + "learning_rate": 1.0398820210310336e-05, + "loss": 0.2349, + "step": 61812 + }, + { + "epoch": 2.38, + "learning_rate": 1.0398179020261606e-05, + "loss": 0.1034, + "step": 61813 + }, + { + "epoch": 2.38, + "learning_rate": 1.0397537830212875e-05, + "loss": 0.2044, + "step": 61814 + }, + { + "epoch": 2.38, + "learning_rate": 1.0396896640164147e-05, + "loss": 0.2869, + "step": 61815 + }, + { + "epoch": 2.38, + "learning_rate": 1.0396255450115415e-05, + "loss": 0.2084, + "step": 61816 + }, + { + "epoch": 2.38, + "learning_rate": 1.0395614260066685e-05, + "loss": 0.2493, + "step": 61817 + }, + { + "epoch": 2.38, + "learning_rate": 1.0394973070017953e-05, + "loss": 0.1474, + "step": 61818 + }, + { + "epoch": 2.38, + "learning_rate": 1.0394331879969223e-05, + "loss": 0.1984, + "step": 61819 + }, + { + "epoch": 2.38, + "learning_rate": 1.0393690689920493e-05, + "loss": 0.2704, + "step": 61820 + }, + { + "epoch": 2.38, + "learning_rate": 1.0393049499871763e-05, + "loss": 0.2832, + "step": 61821 + }, + { + "epoch": 2.38, + "learning_rate": 1.0392408309823032e-05, + "loss": 0.1889, + "step": 61822 + }, + { + "epoch": 2.38, + "learning_rate": 1.0391767119774302e-05, + "loss": 0.1722, + "step": 61823 + }, + { + "epoch": 2.38, + "learning_rate": 1.0391125929725572e-05, + "loss": 0.171, + "step": 61824 + }, + { + "epoch": 2.38, + "learning_rate": 1.0390484739676842e-05, + "loss": 0.098, + "step": 61825 + }, + { + "epoch": 2.38, + "learning_rate": 1.038984354962811e-05, + "loss": 0.2503, + "step": 61826 + }, + { + "epoch": 2.38, + "learning_rate": 1.038920235957938e-05, + "loss": 0.2176, + "step": 61827 + }, + { + "epoch": 2.38, + "learning_rate": 1.038856116953065e-05, + "loss": 0.2325, + "step": 61828 + }, + { + "epoch": 2.38, + "learning_rate": 1.038791997948192e-05, + "loss": 0.1467, + "step": 61829 + }, + { + "epoch": 2.38, + "learning_rate": 1.0387278789433188e-05, + "loss": 0.108, + "step": 61830 + }, + { + "epoch": 2.38, + "learning_rate": 1.0386637599384458e-05, + "loss": 0.2284, + "step": 61831 + }, + { + "epoch": 2.38, + "learning_rate": 1.0385996409335727e-05, + "loss": 0.2552, + "step": 61832 + }, + { + "epoch": 2.38, + "learning_rate": 1.0385355219286998e-05, + "loss": 0.3977, + "step": 61833 + }, + { + "epoch": 2.38, + "learning_rate": 1.0384714029238267e-05, + "loss": 0.1567, + "step": 61834 + }, + { + "epoch": 2.38, + "learning_rate": 1.0384072839189537e-05, + "loss": 0.1513, + "step": 61835 + }, + { + "epoch": 2.38, + "learning_rate": 1.0383431649140805e-05, + "loss": 0.3561, + "step": 61836 + }, + { + "epoch": 2.38, + "learning_rate": 1.0382790459092077e-05, + "loss": 0.0888, + "step": 61837 + }, + { + "epoch": 2.38, + "learning_rate": 1.0382149269043345e-05, + "loss": 0.1566, + "step": 61838 + }, + { + "epoch": 2.38, + "learning_rate": 1.0381508078994615e-05, + "loss": 0.1624, + "step": 61839 + }, + { + "epoch": 2.38, + "learning_rate": 1.0380866888945884e-05, + "loss": 0.3305, + "step": 61840 + }, + { + "epoch": 2.38, + "learning_rate": 1.0380225698897154e-05, + "loss": 0.2029, + "step": 61841 + }, + { + "epoch": 2.38, + "learning_rate": 1.0379584508848424e-05, + "loss": 0.1954, + "step": 61842 + }, + { + "epoch": 2.38, + "learning_rate": 1.0378943318799694e-05, + "loss": 0.1875, + "step": 61843 + }, + { + "epoch": 2.38, + "learning_rate": 1.0378302128750962e-05, + "loss": 0.1719, + "step": 61844 + }, + { + "epoch": 2.38, + "learning_rate": 1.0377660938702232e-05, + "loss": 0.1237, + "step": 61845 + }, + { + "epoch": 2.38, + "learning_rate": 1.0377019748653502e-05, + "loss": 0.2223, + "step": 61846 + }, + { + "epoch": 2.38, + "learning_rate": 1.0376378558604772e-05, + "loss": 0.3127, + "step": 61847 + }, + { + "epoch": 2.38, + "learning_rate": 1.037573736855604e-05, + "loss": 0.2466, + "step": 61848 + }, + { + "epoch": 2.38, + "learning_rate": 1.037509617850731e-05, + "loss": 0.3172, + "step": 61849 + }, + { + "epoch": 2.38, + "learning_rate": 1.0374454988458579e-05, + "loss": 0.1634, + "step": 61850 + }, + { + "epoch": 2.38, + "learning_rate": 1.037381379840985e-05, + "loss": 0.2225, + "step": 61851 + }, + { + "epoch": 2.38, + "learning_rate": 1.0373172608361119e-05, + "loss": 0.3054, + "step": 61852 + }, + { + "epoch": 2.38, + "learning_rate": 1.0372531418312389e-05, + "loss": 0.1654, + "step": 61853 + }, + { + "epoch": 2.38, + "learning_rate": 1.0371890228263657e-05, + "loss": 0.2441, + "step": 61854 + }, + { + "epoch": 2.38, + "learning_rate": 1.0371249038214929e-05, + "loss": 0.163, + "step": 61855 + }, + { + "epoch": 2.38, + "learning_rate": 1.0370607848166197e-05, + "loss": 0.1533, + "step": 61856 + }, + { + "epoch": 2.38, + "learning_rate": 1.0369966658117467e-05, + "loss": 0.3774, + "step": 61857 + }, + { + "epoch": 2.38, + "learning_rate": 1.0369325468068735e-05, + "loss": 0.178, + "step": 61858 + }, + { + "epoch": 2.38, + "learning_rate": 1.0368684278020006e-05, + "loss": 0.2591, + "step": 61859 + }, + { + "epoch": 2.38, + "learning_rate": 1.0368043087971276e-05, + "loss": 0.1848, + "step": 61860 + }, + { + "epoch": 2.38, + "learning_rate": 1.0367401897922546e-05, + "loss": 0.1499, + "step": 61861 + }, + { + "epoch": 2.38, + "learning_rate": 1.0366760707873814e-05, + "loss": 0.2763, + "step": 61862 + }, + { + "epoch": 2.38, + "learning_rate": 1.0366119517825084e-05, + "loss": 0.1956, + "step": 61863 + }, + { + "epoch": 2.38, + "learning_rate": 1.0365478327776354e-05, + "loss": 0.1213, + "step": 61864 + }, + { + "epoch": 2.38, + "learning_rate": 1.0364837137727624e-05, + "loss": 0.1827, + "step": 61865 + }, + { + "epoch": 2.38, + "learning_rate": 1.0364195947678892e-05, + "loss": 0.3404, + "step": 61866 + }, + { + "epoch": 2.38, + "learning_rate": 1.0363554757630162e-05, + "loss": 0.3152, + "step": 61867 + }, + { + "epoch": 2.38, + "learning_rate": 1.036291356758143e-05, + "loss": 0.2693, + "step": 61868 + }, + { + "epoch": 2.38, + "learning_rate": 1.0362272377532702e-05, + "loss": 0.1954, + "step": 61869 + }, + { + "epoch": 2.38, + "learning_rate": 1.036163118748397e-05, + "loss": 0.2038, + "step": 61870 + }, + { + "epoch": 2.38, + "learning_rate": 1.036098999743524e-05, + "loss": 0.3363, + "step": 61871 + }, + { + "epoch": 2.38, + "learning_rate": 1.0360348807386509e-05, + "loss": 0.4007, + "step": 61872 + }, + { + "epoch": 2.38, + "learning_rate": 1.035970761733778e-05, + "loss": 0.3145, + "step": 61873 + }, + { + "epoch": 2.38, + "learning_rate": 1.0359066427289049e-05, + "loss": 0.2944, + "step": 61874 + }, + { + "epoch": 2.38, + "learning_rate": 1.0358425237240319e-05, + "loss": 0.1032, + "step": 61875 + }, + { + "epoch": 2.38, + "learning_rate": 1.0357784047191587e-05, + "loss": 0.1721, + "step": 61876 + }, + { + "epoch": 2.38, + "learning_rate": 1.0357142857142859e-05, + "loss": 0.3209, + "step": 61877 + }, + { + "epoch": 2.38, + "learning_rate": 1.0356501667094127e-05, + "loss": 0.2007, + "step": 61878 + }, + { + "epoch": 2.38, + "learning_rate": 1.0355860477045397e-05, + "loss": 0.1392, + "step": 61879 + }, + { + "epoch": 2.38, + "learning_rate": 1.0355219286996666e-05, + "loss": 0.2041, + "step": 61880 + }, + { + "epoch": 2.38, + "learning_rate": 1.0354578096947936e-05, + "loss": 0.3113, + "step": 61881 + }, + { + "epoch": 2.38, + "learning_rate": 1.0353936906899206e-05, + "loss": 0.1782, + "step": 61882 + }, + { + "epoch": 2.38, + "learning_rate": 1.0353295716850476e-05, + "loss": 0.2141, + "step": 61883 + }, + { + "epoch": 2.38, + "learning_rate": 1.0352654526801744e-05, + "loss": 0.1834, + "step": 61884 + }, + { + "epoch": 2.38, + "learning_rate": 1.0352013336753014e-05, + "loss": 0.2617, + "step": 61885 + }, + { + "epoch": 2.38, + "learning_rate": 1.0351372146704284e-05, + "loss": 0.3058, + "step": 61886 + }, + { + "epoch": 2.38, + "learning_rate": 1.0350730956655554e-05, + "loss": 0.241, + "step": 61887 + }, + { + "epoch": 2.38, + "learning_rate": 1.0350089766606823e-05, + "loss": 0.2705, + "step": 61888 + }, + { + "epoch": 2.38, + "learning_rate": 1.0349448576558093e-05, + "loss": 0.2942, + "step": 61889 + }, + { + "epoch": 2.38, + "learning_rate": 1.0348807386509361e-05, + "loss": 0.235, + "step": 61890 + }, + { + "epoch": 2.38, + "learning_rate": 1.0348166196460633e-05, + "loss": 0.3433, + "step": 61891 + }, + { + "epoch": 2.38, + "learning_rate": 1.0347525006411901e-05, + "loss": 0.1489, + "step": 61892 + }, + { + "epoch": 2.38, + "learning_rate": 1.0346883816363171e-05, + "loss": 0.3115, + "step": 61893 + }, + { + "epoch": 2.38, + "learning_rate": 1.034624262631444e-05, + "loss": 0.2608, + "step": 61894 + }, + { + "epoch": 2.38, + "learning_rate": 1.0345601436265711e-05, + "loss": 0.31, + "step": 61895 + }, + { + "epoch": 2.38, + "learning_rate": 1.034496024621698e-05, + "loss": 0.1644, + "step": 61896 + }, + { + "epoch": 2.38, + "learning_rate": 1.034431905616825e-05, + "loss": 0.1909, + "step": 61897 + }, + { + "epoch": 2.38, + "learning_rate": 1.0343677866119518e-05, + "loss": 0.34, + "step": 61898 + }, + { + "epoch": 2.38, + "learning_rate": 1.0343036676070788e-05, + "loss": 0.1972, + "step": 61899 + }, + { + "epoch": 2.38, + "learning_rate": 1.0342395486022058e-05, + "loss": 0.3228, + "step": 61900 + }, + { + "epoch": 2.38, + "learning_rate": 1.0341754295973328e-05, + "loss": 0.3906, + "step": 61901 + }, + { + "epoch": 2.38, + "learning_rate": 1.0341113105924596e-05, + "loss": 0.3509, + "step": 61902 + }, + { + "epoch": 2.38, + "learning_rate": 1.0340471915875866e-05, + "loss": 0.2183, + "step": 61903 + }, + { + "epoch": 2.38, + "learning_rate": 1.0339830725827136e-05, + "loss": 0.1715, + "step": 61904 + }, + { + "epoch": 2.38, + "learning_rate": 1.0339189535778406e-05, + "loss": 0.1957, + "step": 61905 + }, + { + "epoch": 2.38, + "learning_rate": 1.0338548345729675e-05, + "loss": 0.1618, + "step": 61906 + }, + { + "epoch": 2.38, + "learning_rate": 1.0337907155680945e-05, + "loss": 0.1101, + "step": 61907 + }, + { + "epoch": 2.38, + "learning_rate": 1.0337265965632213e-05, + "loss": 0.2214, + "step": 61908 + }, + { + "epoch": 2.38, + "learning_rate": 1.0336624775583485e-05, + "loss": 0.253, + "step": 61909 + }, + { + "epoch": 2.38, + "learning_rate": 1.0335983585534753e-05, + "loss": 0.1786, + "step": 61910 + }, + { + "epoch": 2.38, + "learning_rate": 1.0335342395486023e-05, + "loss": 0.1507, + "step": 61911 + }, + { + "epoch": 2.38, + "learning_rate": 1.0334701205437291e-05, + "loss": 0.1941, + "step": 61912 + }, + { + "epoch": 2.38, + "learning_rate": 1.0334060015388563e-05, + "loss": 0.1773, + "step": 61913 + }, + { + "epoch": 2.38, + "learning_rate": 1.0333418825339831e-05, + "loss": 0.1786, + "step": 61914 + }, + { + "epoch": 2.38, + "learning_rate": 1.0332777635291101e-05, + "loss": 0.1732, + "step": 61915 + }, + { + "epoch": 2.38, + "learning_rate": 1.033213644524237e-05, + "loss": 0.3007, + "step": 61916 + }, + { + "epoch": 2.38, + "learning_rate": 1.033149525519364e-05, + "loss": 0.253, + "step": 61917 + }, + { + "epoch": 2.38, + "learning_rate": 1.033085406514491e-05, + "loss": 0.2465, + "step": 61918 + }, + { + "epoch": 2.38, + "learning_rate": 1.033021287509618e-05, + "loss": 0.2358, + "step": 61919 + }, + { + "epoch": 2.38, + "learning_rate": 1.0329571685047448e-05, + "loss": 0.2089, + "step": 61920 + }, + { + "epoch": 2.38, + "learning_rate": 1.0328930494998718e-05, + "loss": 0.2291, + "step": 61921 + }, + { + "epoch": 2.38, + "learning_rate": 1.0328289304949988e-05, + "loss": 0.1325, + "step": 61922 + }, + { + "epoch": 2.38, + "learning_rate": 1.0327648114901258e-05, + "loss": 0.1558, + "step": 61923 + }, + { + "epoch": 2.38, + "learning_rate": 1.0327006924852526e-05, + "loss": 0.2624, + "step": 61924 + }, + { + "epoch": 2.38, + "learning_rate": 1.0326365734803797e-05, + "loss": 0.2048, + "step": 61925 + }, + { + "epoch": 2.38, + "learning_rate": 1.0325724544755067e-05, + "loss": 0.2456, + "step": 61926 + }, + { + "epoch": 2.38, + "learning_rate": 1.0325083354706337e-05, + "loss": 0.3188, + "step": 61927 + }, + { + "epoch": 2.38, + "learning_rate": 1.0324442164657605e-05, + "loss": 0.1977, + "step": 61928 + }, + { + "epoch": 2.38, + "learning_rate": 1.0323800974608875e-05, + "loss": 0.3108, + "step": 61929 + }, + { + "epoch": 2.38, + "learning_rate": 1.0323159784560143e-05, + "loss": 0.2311, + "step": 61930 + }, + { + "epoch": 2.38, + "learning_rate": 1.0322518594511415e-05, + "loss": 0.3596, + "step": 61931 + }, + { + "epoch": 2.38, + "learning_rate": 1.0321877404462683e-05, + "loss": 0.1391, + "step": 61932 + }, + { + "epoch": 2.38, + "learning_rate": 1.0321236214413953e-05, + "loss": 0.1499, + "step": 61933 + }, + { + "epoch": 2.38, + "learning_rate": 1.0320595024365222e-05, + "loss": 0.228, + "step": 61934 + }, + { + "epoch": 2.38, + "learning_rate": 1.0319953834316493e-05, + "loss": 0.2108, + "step": 61935 + }, + { + "epoch": 2.38, + "learning_rate": 1.0319312644267762e-05, + "loss": 0.3623, + "step": 61936 + }, + { + "epoch": 2.38, + "learning_rate": 1.0318671454219032e-05, + "loss": 0.1489, + "step": 61937 + }, + { + "epoch": 2.38, + "learning_rate": 1.03180302641703e-05, + "loss": 0.2258, + "step": 61938 + }, + { + "epoch": 2.38, + "learning_rate": 1.031738907412157e-05, + "loss": 0.2533, + "step": 61939 + }, + { + "epoch": 2.38, + "learning_rate": 1.031674788407284e-05, + "loss": 0.2964, + "step": 61940 + }, + { + "epoch": 2.38, + "learning_rate": 1.031610669402411e-05, + "loss": 0.3155, + "step": 61941 + }, + { + "epoch": 2.38, + "learning_rate": 1.0315465503975378e-05, + "loss": 0.2425, + "step": 61942 + }, + { + "epoch": 2.38, + "learning_rate": 1.0314824313926648e-05, + "loss": 0.1535, + "step": 61943 + }, + { + "epoch": 2.38, + "learning_rate": 1.0314183123877918e-05, + "loss": 0.2188, + "step": 61944 + }, + { + "epoch": 2.38, + "learning_rate": 1.0313541933829188e-05, + "loss": 0.1428, + "step": 61945 + }, + { + "epoch": 2.38, + "learning_rate": 1.0312900743780457e-05, + "loss": 0.3159, + "step": 61946 + }, + { + "epoch": 2.38, + "learning_rate": 1.0312259553731727e-05, + "loss": 0.1945, + "step": 61947 + }, + { + "epoch": 2.38, + "learning_rate": 1.0311618363682995e-05, + "loss": 0.1542, + "step": 61948 + }, + { + "epoch": 2.38, + "learning_rate": 1.0310977173634267e-05, + "loss": 0.1196, + "step": 61949 + }, + { + "epoch": 2.38, + "learning_rate": 1.0310335983585535e-05, + "loss": 0.1933, + "step": 61950 + }, + { + "epoch": 2.38, + "learning_rate": 1.0309694793536805e-05, + "loss": 0.3243, + "step": 61951 + }, + { + "epoch": 2.38, + "learning_rate": 1.0309053603488074e-05, + "loss": 0.1814, + "step": 61952 + }, + { + "epoch": 2.38, + "learning_rate": 1.0308412413439345e-05, + "loss": 0.1816, + "step": 61953 + }, + { + "epoch": 2.38, + "learning_rate": 1.0307771223390614e-05, + "loss": 0.1578, + "step": 61954 + }, + { + "epoch": 2.38, + "learning_rate": 1.0307130033341884e-05, + "loss": 0.116, + "step": 61955 + }, + { + "epoch": 2.38, + "learning_rate": 1.0306488843293152e-05, + "loss": 0.1614, + "step": 61956 + }, + { + "epoch": 2.38, + "learning_rate": 1.0305847653244422e-05, + "loss": 0.2466, + "step": 61957 + }, + { + "epoch": 2.38, + "learning_rate": 1.0305206463195692e-05, + "loss": 0.206, + "step": 61958 + }, + { + "epoch": 2.38, + "learning_rate": 1.0304565273146962e-05, + "loss": 0.2404, + "step": 61959 + }, + { + "epoch": 2.38, + "learning_rate": 1.030392408309823e-05, + "loss": 0.2399, + "step": 61960 + }, + { + "epoch": 2.38, + "learning_rate": 1.03032828930495e-05, + "loss": 0.2217, + "step": 61961 + }, + { + "epoch": 2.38, + "learning_rate": 1.030264170300077e-05, + "loss": 0.2034, + "step": 61962 + }, + { + "epoch": 2.38, + "learning_rate": 1.030200051295204e-05, + "loss": 0.1572, + "step": 61963 + }, + { + "epoch": 2.38, + "learning_rate": 1.0301359322903309e-05, + "loss": 0.222, + "step": 61964 + }, + { + "epoch": 2.38, + "learning_rate": 1.0300718132854579e-05, + "loss": 0.1824, + "step": 61965 + }, + { + "epoch": 2.38, + "learning_rate": 1.0300076942805849e-05, + "loss": 0.177, + "step": 61966 + }, + { + "epoch": 2.38, + "learning_rate": 1.0299435752757119e-05, + "loss": 0.2577, + "step": 61967 + }, + { + "epoch": 2.38, + "learning_rate": 1.0298794562708387e-05, + "loss": 0.2588, + "step": 61968 + }, + { + "epoch": 2.38, + "learning_rate": 1.0298153372659657e-05, + "loss": 0.236, + "step": 61969 + }, + { + "epoch": 2.38, + "learning_rate": 1.0297512182610925e-05, + "loss": 0.1814, + "step": 61970 + }, + { + "epoch": 2.38, + "learning_rate": 1.0296870992562197e-05, + "loss": 0.2111, + "step": 61971 + }, + { + "epoch": 2.38, + "learning_rate": 1.0296229802513466e-05, + "loss": 0.2024, + "step": 61972 + }, + { + "epoch": 2.38, + "learning_rate": 1.0295588612464736e-05, + "loss": 0.2082, + "step": 61973 + }, + { + "epoch": 2.38, + "learning_rate": 1.0294947422416004e-05, + "loss": 0.1992, + "step": 61974 + }, + { + "epoch": 2.38, + "learning_rate": 1.0294306232367276e-05, + "loss": 0.2944, + "step": 61975 + }, + { + "epoch": 2.38, + "learning_rate": 1.0293665042318544e-05, + "loss": 0.1329, + "step": 61976 + }, + { + "epoch": 2.38, + "learning_rate": 1.0293023852269814e-05, + "loss": 0.2974, + "step": 61977 + }, + { + "epoch": 2.38, + "learning_rate": 1.0292382662221082e-05, + "loss": 0.2434, + "step": 61978 + }, + { + "epoch": 2.38, + "learning_rate": 1.0291741472172352e-05, + "loss": 0.201, + "step": 61979 + }, + { + "epoch": 2.38, + "learning_rate": 1.0291100282123622e-05, + "loss": 0.1621, + "step": 61980 + }, + { + "epoch": 2.38, + "learning_rate": 1.0290459092074892e-05, + "loss": 0.2973, + "step": 61981 + }, + { + "epoch": 2.38, + "learning_rate": 1.028981790202616e-05, + "loss": 0.1684, + "step": 61982 + }, + { + "epoch": 2.38, + "learning_rate": 1.028917671197743e-05, + "loss": 0.2268, + "step": 61983 + }, + { + "epoch": 2.38, + "learning_rate": 1.02885355219287e-05, + "loss": 0.2507, + "step": 61984 + }, + { + "epoch": 2.38, + "learning_rate": 1.028789433187997e-05, + "loss": 0.2842, + "step": 61985 + }, + { + "epoch": 2.38, + "learning_rate": 1.0287253141831239e-05, + "loss": 0.1275, + "step": 61986 + }, + { + "epoch": 2.38, + "learning_rate": 1.0286611951782509e-05, + "loss": 0.2752, + "step": 61987 + }, + { + "epoch": 2.38, + "learning_rate": 1.0285970761733777e-05, + "loss": 0.252, + "step": 61988 + }, + { + "epoch": 2.38, + "learning_rate": 1.0285329571685049e-05, + "loss": 0.2311, + "step": 61989 + }, + { + "epoch": 2.38, + "learning_rate": 1.0284688381636317e-05, + "loss": 0.2392, + "step": 61990 + }, + { + "epoch": 2.38, + "learning_rate": 1.0284047191587588e-05, + "loss": 0.17, + "step": 61991 + }, + { + "epoch": 2.38, + "learning_rate": 1.0283406001538856e-05, + "loss": 0.2769, + "step": 61992 + }, + { + "epoch": 2.38, + "learning_rate": 1.0282764811490128e-05, + "loss": 0.2287, + "step": 61993 + }, + { + "epoch": 2.38, + "learning_rate": 1.0282123621441396e-05, + "loss": 0.3989, + "step": 61994 + }, + { + "epoch": 2.38, + "learning_rate": 1.0281482431392666e-05, + "loss": 0.1956, + "step": 61995 + }, + { + "epoch": 2.38, + "learning_rate": 1.0280841241343934e-05, + "loss": 0.2271, + "step": 61996 + }, + { + "epoch": 2.38, + "learning_rate": 1.0280200051295204e-05, + "loss": 0.267, + "step": 61997 + }, + { + "epoch": 2.38, + "learning_rate": 1.0279558861246474e-05, + "loss": 0.1468, + "step": 61998 + }, + { + "epoch": 2.38, + "learning_rate": 1.0278917671197744e-05, + "loss": 0.2123, + "step": 61999 + }, + { + "epoch": 2.38, + "learning_rate": 1.0278276481149013e-05, + "loss": 0.2406, + "step": 62000 + }, + { + "epoch": 2.38, + "learning_rate": 1.0277635291100283e-05, + "loss": 0.1487, + "step": 62001 + }, + { + "epoch": 2.38, + "learning_rate": 1.0276994101051553e-05, + "loss": 0.1293, + "step": 62002 + }, + { + "epoch": 2.38, + "learning_rate": 1.0276352911002823e-05, + "loss": 0.193, + "step": 62003 + }, + { + "epoch": 2.38, + "learning_rate": 1.0275711720954091e-05, + "loss": 0.2713, + "step": 62004 + }, + { + "epoch": 2.38, + "learning_rate": 1.0275070530905361e-05, + "loss": 0.1294, + "step": 62005 + }, + { + "epoch": 2.38, + "learning_rate": 1.027442934085663e-05, + "loss": 0.1966, + "step": 62006 + }, + { + "epoch": 2.38, + "learning_rate": 1.0273788150807901e-05, + "loss": 0.2192, + "step": 62007 + }, + { + "epoch": 2.38, + "learning_rate": 1.027314696075917e-05, + "loss": 0.2921, + "step": 62008 + }, + { + "epoch": 2.38, + "learning_rate": 1.027250577071044e-05, + "loss": 0.207, + "step": 62009 + }, + { + "epoch": 2.38, + "learning_rate": 1.0271864580661708e-05, + "loss": 0.2782, + "step": 62010 + }, + { + "epoch": 2.38, + "learning_rate": 1.027122339061298e-05, + "loss": 0.2576, + "step": 62011 + }, + { + "epoch": 2.38, + "learning_rate": 1.0270582200564248e-05, + "loss": 0.2644, + "step": 62012 + }, + { + "epoch": 2.38, + "learning_rate": 1.0269941010515518e-05, + "loss": 0.2944, + "step": 62013 + }, + { + "epoch": 2.38, + "learning_rate": 1.0269299820466786e-05, + "loss": 0.1759, + "step": 62014 + }, + { + "epoch": 2.38, + "learning_rate": 1.0268658630418058e-05, + "loss": 0.1364, + "step": 62015 + }, + { + "epoch": 2.38, + "learning_rate": 1.0268017440369326e-05, + "loss": 0.2429, + "step": 62016 + }, + { + "epoch": 2.38, + "learning_rate": 1.0267376250320596e-05, + "loss": 0.238, + "step": 62017 + }, + { + "epoch": 2.38, + "learning_rate": 1.0266735060271865e-05, + "loss": 0.2882, + "step": 62018 + }, + { + "epoch": 2.38, + "learning_rate": 1.0266093870223135e-05, + "loss": 0.2247, + "step": 62019 + }, + { + "epoch": 2.38, + "learning_rate": 1.0265452680174405e-05, + "loss": 0.1404, + "step": 62020 + }, + { + "epoch": 2.38, + "learning_rate": 1.0264811490125675e-05, + "loss": 0.2783, + "step": 62021 + }, + { + "epoch": 2.38, + "learning_rate": 1.0264170300076943e-05, + "loss": 0.2274, + "step": 62022 + }, + { + "epoch": 2.38, + "learning_rate": 1.0263529110028213e-05, + "loss": 0.1508, + "step": 62023 + }, + { + "epoch": 2.38, + "learning_rate": 1.0262887919979483e-05, + "loss": 0.2101, + "step": 62024 + }, + { + "epoch": 2.38, + "learning_rate": 1.0262246729930753e-05, + "loss": 0.3225, + "step": 62025 + }, + { + "epoch": 2.38, + "learning_rate": 1.0261605539882021e-05, + "loss": 0.2177, + "step": 62026 + }, + { + "epoch": 2.38, + "learning_rate": 1.0260964349833291e-05, + "loss": 0.2759, + "step": 62027 + }, + { + "epoch": 2.38, + "learning_rate": 1.026032315978456e-05, + "loss": 0.157, + "step": 62028 + }, + { + "epoch": 2.38, + "learning_rate": 1.0259681969735831e-05, + "loss": 0.1183, + "step": 62029 + }, + { + "epoch": 2.38, + "learning_rate": 1.02590407796871e-05, + "loss": 0.116, + "step": 62030 + }, + { + "epoch": 2.38, + "learning_rate": 1.025839958963837e-05, + "loss": 0.2023, + "step": 62031 + }, + { + "epoch": 2.38, + "learning_rate": 1.0257758399589638e-05, + "loss": 0.3513, + "step": 62032 + }, + { + "epoch": 2.38, + "learning_rate": 1.025711720954091e-05, + "loss": 0.252, + "step": 62033 + }, + { + "epoch": 2.39, + "learning_rate": 1.0256476019492178e-05, + "loss": 0.0968, + "step": 62034 + }, + { + "epoch": 2.39, + "learning_rate": 1.0255834829443448e-05, + "loss": 0.1813, + "step": 62035 + }, + { + "epoch": 2.39, + "learning_rate": 1.0255193639394716e-05, + "loss": 0.272, + "step": 62036 + }, + { + "epoch": 2.39, + "learning_rate": 1.0254552449345987e-05, + "loss": 0.2847, + "step": 62037 + }, + { + "epoch": 2.39, + "learning_rate": 1.0253911259297257e-05, + "loss": 0.2065, + "step": 62038 + }, + { + "epoch": 2.39, + "learning_rate": 1.0253270069248527e-05, + "loss": 0.2579, + "step": 62039 + }, + { + "epoch": 2.39, + "learning_rate": 1.0252628879199795e-05, + "loss": 0.222, + "step": 62040 + }, + { + "epoch": 2.39, + "learning_rate": 1.0251987689151065e-05, + "loss": 0.2546, + "step": 62041 + }, + { + "epoch": 2.39, + "learning_rate": 1.0251346499102335e-05, + "loss": 0.3739, + "step": 62042 + }, + { + "epoch": 2.39, + "learning_rate": 1.0250705309053605e-05, + "loss": 0.2336, + "step": 62043 + }, + { + "epoch": 2.39, + "learning_rate": 1.0250064119004873e-05, + "loss": 0.1719, + "step": 62044 + }, + { + "epoch": 2.39, + "learning_rate": 1.0249422928956143e-05, + "loss": 0.3602, + "step": 62045 + }, + { + "epoch": 2.39, + "learning_rate": 1.0248781738907412e-05, + "loss": 0.1579, + "step": 62046 + }, + { + "epoch": 2.39, + "learning_rate": 1.0248140548858683e-05, + "loss": 0.1688, + "step": 62047 + }, + { + "epoch": 2.39, + "learning_rate": 1.0247499358809952e-05, + "loss": 0.2574, + "step": 62048 + }, + { + "epoch": 2.39, + "learning_rate": 1.0246858168761222e-05, + "loss": 0.1984, + "step": 62049 + }, + { + "epoch": 2.39, + "learning_rate": 1.024621697871249e-05, + "loss": 0.2205, + "step": 62050 + }, + { + "epoch": 2.39, + "learning_rate": 1.0245575788663762e-05, + "loss": 0.1485, + "step": 62051 + }, + { + "epoch": 2.39, + "learning_rate": 1.024493459861503e-05, + "loss": 0.2169, + "step": 62052 + }, + { + "epoch": 2.39, + "learning_rate": 1.02442934085663e-05, + "loss": 0.2153, + "step": 62053 + }, + { + "epoch": 2.39, + "learning_rate": 1.0243652218517568e-05, + "loss": 0.2638, + "step": 62054 + }, + { + "epoch": 2.39, + "learning_rate": 1.0243011028468838e-05, + "loss": 0.2294, + "step": 62055 + }, + { + "epoch": 2.39, + "learning_rate": 1.0242369838420108e-05, + "loss": 0.1927, + "step": 62056 + }, + { + "epoch": 2.39, + "learning_rate": 1.0241728648371378e-05, + "loss": 0.2609, + "step": 62057 + }, + { + "epoch": 2.39, + "learning_rate": 1.0241087458322647e-05, + "loss": 0.151, + "step": 62058 + }, + { + "epoch": 2.39, + "learning_rate": 1.0240446268273917e-05, + "loss": 0.2681, + "step": 62059 + }, + { + "epoch": 2.39, + "learning_rate": 1.0239805078225187e-05, + "loss": 0.1955, + "step": 62060 + }, + { + "epoch": 2.39, + "learning_rate": 1.0239163888176457e-05, + "loss": 0.3194, + "step": 62061 + }, + { + "epoch": 2.39, + "learning_rate": 1.0238522698127725e-05, + "loss": 0.3487, + "step": 62062 + }, + { + "epoch": 2.39, + "learning_rate": 1.0237881508078995e-05, + "loss": 0.1272, + "step": 62063 + }, + { + "epoch": 2.39, + "learning_rate": 1.0237240318030265e-05, + "loss": 0.1836, + "step": 62064 + }, + { + "epoch": 2.39, + "learning_rate": 1.0236599127981535e-05, + "loss": 0.2215, + "step": 62065 + }, + { + "epoch": 2.39, + "learning_rate": 1.0235957937932804e-05, + "loss": 0.1732, + "step": 62066 + }, + { + "epoch": 2.39, + "learning_rate": 1.0235316747884074e-05, + "loss": 0.1997, + "step": 62067 + }, + { + "epoch": 2.39, + "learning_rate": 1.0234675557835342e-05, + "loss": 0.164, + "step": 62068 + }, + { + "epoch": 2.39, + "learning_rate": 1.0234034367786614e-05, + "loss": 0.2879, + "step": 62069 + }, + { + "epoch": 2.39, + "learning_rate": 1.0233393177737882e-05, + "loss": 0.2865, + "step": 62070 + }, + { + "epoch": 2.39, + "learning_rate": 1.0232751987689152e-05, + "loss": 0.1914, + "step": 62071 + }, + { + "epoch": 2.39, + "learning_rate": 1.023211079764042e-05, + "loss": 0.2066, + "step": 62072 + }, + { + "epoch": 2.39, + "learning_rate": 1.0231469607591692e-05, + "loss": 0.2454, + "step": 62073 + }, + { + "epoch": 2.39, + "learning_rate": 1.023082841754296e-05, + "loss": 0.1377, + "step": 62074 + }, + { + "epoch": 2.39, + "learning_rate": 1.023018722749423e-05, + "loss": 0.2426, + "step": 62075 + }, + { + "epoch": 2.39, + "learning_rate": 1.0229546037445499e-05, + "loss": 0.1861, + "step": 62076 + }, + { + "epoch": 2.39, + "learning_rate": 1.0228904847396769e-05, + "loss": 0.1991, + "step": 62077 + }, + { + "epoch": 2.39, + "learning_rate": 1.0228263657348039e-05, + "loss": 0.204, + "step": 62078 + }, + { + "epoch": 2.39, + "learning_rate": 1.0227622467299309e-05, + "loss": 0.1228, + "step": 62079 + }, + { + "epoch": 2.39, + "learning_rate": 1.0226981277250577e-05, + "loss": 0.2217, + "step": 62080 + }, + { + "epoch": 2.39, + "learning_rate": 1.0226340087201847e-05, + "loss": 0.2544, + "step": 62081 + }, + { + "epoch": 2.39, + "learning_rate": 1.0225698897153117e-05, + "loss": 0.1167, + "step": 62082 + }, + { + "epoch": 2.39, + "learning_rate": 1.0225057707104387e-05, + "loss": 0.1502, + "step": 62083 + }, + { + "epoch": 2.39, + "learning_rate": 1.0224416517055656e-05, + "loss": 0.1682, + "step": 62084 + }, + { + "epoch": 2.39, + "learning_rate": 1.0223775327006926e-05, + "loss": 0.3192, + "step": 62085 + }, + { + "epoch": 2.39, + "learning_rate": 1.0223134136958194e-05, + "loss": 0.2158, + "step": 62086 + }, + { + "epoch": 2.39, + "learning_rate": 1.0222492946909466e-05, + "loss": 0.2458, + "step": 62087 + }, + { + "epoch": 2.39, + "learning_rate": 1.0221851756860734e-05, + "loss": 0.1772, + "step": 62088 + }, + { + "epoch": 2.39, + "learning_rate": 1.0221210566812004e-05, + "loss": 0.1533, + "step": 62089 + }, + { + "epoch": 2.39, + "learning_rate": 1.0220569376763272e-05, + "loss": 0.2634, + "step": 62090 + }, + { + "epoch": 2.39, + "learning_rate": 1.0219928186714544e-05, + "loss": 0.1792, + "step": 62091 + }, + { + "epoch": 2.39, + "learning_rate": 1.0219286996665812e-05, + "loss": 0.1228, + "step": 62092 + }, + { + "epoch": 2.39, + "learning_rate": 1.0218645806617082e-05, + "loss": 0.2453, + "step": 62093 + }, + { + "epoch": 2.39, + "learning_rate": 1.021800461656835e-05, + "loss": 0.2884, + "step": 62094 + }, + { + "epoch": 2.39, + "learning_rate": 1.021736342651962e-05, + "loss": 0.2453, + "step": 62095 + }, + { + "epoch": 2.39, + "learning_rate": 1.021672223647089e-05, + "loss": 0.1642, + "step": 62096 + }, + { + "epoch": 2.39, + "learning_rate": 1.021608104642216e-05, + "loss": 0.2047, + "step": 62097 + }, + { + "epoch": 2.39, + "learning_rate": 1.0215439856373429e-05, + "loss": 0.2441, + "step": 62098 + }, + { + "epoch": 2.39, + "learning_rate": 1.0214798666324699e-05, + "loss": 0.2561, + "step": 62099 + }, + { + "epoch": 2.39, + "learning_rate": 1.0214157476275969e-05, + "loss": 0.2526, + "step": 62100 + }, + { + "epoch": 2.39, + "learning_rate": 1.021351628622724e-05, + "loss": 0.2926, + "step": 62101 + }, + { + "epoch": 2.39, + "learning_rate": 1.0212875096178507e-05, + "loss": 0.1789, + "step": 62102 + }, + { + "epoch": 2.39, + "learning_rate": 1.0212233906129778e-05, + "loss": 0.2255, + "step": 62103 + }, + { + "epoch": 2.39, + "learning_rate": 1.0211592716081046e-05, + "loss": 0.2249, + "step": 62104 + }, + { + "epoch": 2.39, + "learning_rate": 1.0210951526032318e-05, + "loss": 0.0652, + "step": 62105 + }, + { + "epoch": 2.39, + "learning_rate": 1.0210310335983586e-05, + "loss": 0.1929, + "step": 62106 + }, + { + "epoch": 2.39, + "learning_rate": 1.0209669145934856e-05, + "loss": 0.2512, + "step": 62107 + }, + { + "epoch": 2.39, + "learning_rate": 1.0209027955886124e-05, + "loss": 0.2429, + "step": 62108 + }, + { + "epoch": 2.39, + "learning_rate": 1.0208386765837396e-05, + "loss": 0.2872, + "step": 62109 + }, + { + "epoch": 2.39, + "learning_rate": 1.0207745575788664e-05, + "loss": 0.1845, + "step": 62110 + }, + { + "epoch": 2.39, + "learning_rate": 1.0207104385739934e-05, + "loss": 0.1872, + "step": 62111 + }, + { + "epoch": 2.39, + "learning_rate": 1.0206463195691203e-05, + "loss": 0.2881, + "step": 62112 + }, + { + "epoch": 2.39, + "learning_rate": 1.0205822005642474e-05, + "loss": 0.1174, + "step": 62113 + }, + { + "epoch": 2.39, + "learning_rate": 1.0205180815593743e-05, + "loss": 0.2352, + "step": 62114 + }, + { + "epoch": 2.39, + "learning_rate": 1.0204539625545013e-05, + "loss": 0.1315, + "step": 62115 + }, + { + "epoch": 2.39, + "learning_rate": 1.0203898435496281e-05, + "loss": 0.1719, + "step": 62116 + }, + { + "epoch": 2.39, + "learning_rate": 1.0203257245447551e-05, + "loss": 0.231, + "step": 62117 + }, + { + "epoch": 2.39, + "learning_rate": 1.0202616055398821e-05, + "loss": 0.164, + "step": 62118 + }, + { + "epoch": 2.39, + "learning_rate": 1.0201974865350091e-05, + "loss": 0.2572, + "step": 62119 + }, + { + "epoch": 2.39, + "learning_rate": 1.020133367530136e-05, + "loss": 0.1554, + "step": 62120 + }, + { + "epoch": 2.39, + "learning_rate": 1.020069248525263e-05, + "loss": 0.3354, + "step": 62121 + }, + { + "epoch": 2.39, + "learning_rate": 1.02000512952039e-05, + "loss": 0.3602, + "step": 62122 + }, + { + "epoch": 2.39, + "learning_rate": 1.019941010515517e-05, + "loss": 0.1508, + "step": 62123 + }, + { + "epoch": 2.39, + "learning_rate": 1.0198768915106438e-05, + "loss": 0.2013, + "step": 62124 + }, + { + "epoch": 2.39, + "learning_rate": 1.0198127725057708e-05, + "loss": 0.2235, + "step": 62125 + }, + { + "epoch": 2.39, + "learning_rate": 1.0197486535008976e-05, + "loss": 0.2565, + "step": 62126 + }, + { + "epoch": 2.39, + "learning_rate": 1.0196845344960248e-05, + "loss": 0.271, + "step": 62127 + }, + { + "epoch": 2.39, + "learning_rate": 1.0196204154911516e-05, + "loss": 0.2201, + "step": 62128 + }, + { + "epoch": 2.39, + "learning_rate": 1.0195562964862786e-05, + "loss": 0.2862, + "step": 62129 + }, + { + "epoch": 2.39, + "learning_rate": 1.0194921774814055e-05, + "loss": 0.1242, + "step": 62130 + }, + { + "epoch": 2.39, + "learning_rate": 1.0194280584765326e-05, + "loss": 0.2493, + "step": 62131 + }, + { + "epoch": 2.39, + "learning_rate": 1.0193639394716595e-05, + "loss": 0.1656, + "step": 62132 + }, + { + "epoch": 2.39, + "learning_rate": 1.0192998204667865e-05, + "loss": 0.2572, + "step": 62133 + }, + { + "epoch": 2.39, + "learning_rate": 1.0192357014619133e-05, + "loss": 0.2705, + "step": 62134 + }, + { + "epoch": 2.39, + "learning_rate": 1.0191715824570403e-05, + "loss": 0.2734, + "step": 62135 + }, + { + "epoch": 2.39, + "learning_rate": 1.0191074634521673e-05, + "loss": 0.189, + "step": 62136 + }, + { + "epoch": 2.39, + "learning_rate": 1.0190433444472943e-05, + "loss": 0.129, + "step": 62137 + }, + { + "epoch": 2.39, + "learning_rate": 1.0189792254424211e-05, + "loss": 0.2056, + "step": 62138 + }, + { + "epoch": 2.39, + "learning_rate": 1.0189151064375481e-05, + "loss": 0.2233, + "step": 62139 + }, + { + "epoch": 2.39, + "learning_rate": 1.0188509874326751e-05, + "loss": 0.1958, + "step": 62140 + }, + { + "epoch": 2.39, + "learning_rate": 1.0187868684278021e-05, + "loss": 0.2206, + "step": 62141 + }, + { + "epoch": 2.39, + "learning_rate": 1.018722749422929e-05, + "loss": 0.1946, + "step": 62142 + }, + { + "epoch": 2.39, + "learning_rate": 1.018658630418056e-05, + "loss": 0.2475, + "step": 62143 + }, + { + "epoch": 2.39, + "learning_rate": 1.0185945114131828e-05, + "loss": 0.1983, + "step": 62144 + }, + { + "epoch": 2.39, + "learning_rate": 1.01853039240831e-05, + "loss": 0.1695, + "step": 62145 + }, + { + "epoch": 2.39, + "learning_rate": 1.0184662734034368e-05, + "loss": 0.1907, + "step": 62146 + }, + { + "epoch": 2.39, + "learning_rate": 1.0184021543985638e-05, + "loss": 0.1438, + "step": 62147 + }, + { + "epoch": 2.39, + "learning_rate": 1.0183380353936907e-05, + "loss": 0.3827, + "step": 62148 + }, + { + "epoch": 2.39, + "learning_rate": 1.0182739163888178e-05, + "loss": 0.1874, + "step": 62149 + }, + { + "epoch": 2.39, + "learning_rate": 1.0182097973839447e-05, + "loss": 0.1661, + "step": 62150 + }, + { + "epoch": 2.39, + "learning_rate": 1.0181456783790717e-05, + "loss": 0.3137, + "step": 62151 + }, + { + "epoch": 2.39, + "learning_rate": 1.0180815593741985e-05, + "loss": 0.1378, + "step": 62152 + }, + { + "epoch": 2.39, + "learning_rate": 1.0180174403693255e-05, + "loss": 0.1942, + "step": 62153 + }, + { + "epoch": 2.39, + "learning_rate": 1.0179533213644525e-05, + "loss": 0.3588, + "step": 62154 + }, + { + "epoch": 2.39, + "learning_rate": 1.0178892023595795e-05, + "loss": 0.1965, + "step": 62155 + }, + { + "epoch": 2.39, + "learning_rate": 1.0178250833547063e-05, + "loss": 0.1791, + "step": 62156 + }, + { + "epoch": 2.39, + "learning_rate": 1.0177609643498333e-05, + "loss": 0.3386, + "step": 62157 + }, + { + "epoch": 2.39, + "learning_rate": 1.0176968453449603e-05, + "loss": 0.2294, + "step": 62158 + }, + { + "epoch": 2.39, + "learning_rate": 1.0176327263400873e-05, + "loss": 0.3139, + "step": 62159 + }, + { + "epoch": 2.39, + "learning_rate": 1.0175686073352142e-05, + "loss": 0.1471, + "step": 62160 + }, + { + "epoch": 2.39, + "learning_rate": 1.0175044883303412e-05, + "loss": 0.2091, + "step": 62161 + }, + { + "epoch": 2.39, + "learning_rate": 1.0174403693254682e-05, + "loss": 0.1901, + "step": 62162 + }, + { + "epoch": 2.39, + "learning_rate": 1.0173762503205952e-05, + "loss": 0.2697, + "step": 62163 + }, + { + "epoch": 2.39, + "learning_rate": 1.017312131315722e-05, + "loss": 0.2248, + "step": 62164 + }, + { + "epoch": 2.39, + "learning_rate": 1.017248012310849e-05, + "loss": 0.2676, + "step": 62165 + }, + { + "epoch": 2.39, + "learning_rate": 1.0171838933059758e-05, + "loss": 0.3622, + "step": 62166 + }, + { + "epoch": 2.39, + "learning_rate": 1.017119774301103e-05, + "loss": 0.2458, + "step": 62167 + }, + { + "epoch": 2.39, + "learning_rate": 1.0170556552962298e-05, + "loss": 0.2004, + "step": 62168 + }, + { + "epoch": 2.39, + "learning_rate": 1.0169915362913569e-05, + "loss": 0.1896, + "step": 62169 + }, + { + "epoch": 2.39, + "learning_rate": 1.0169274172864837e-05, + "loss": 0.3397, + "step": 62170 + }, + { + "epoch": 2.39, + "learning_rate": 1.0168632982816109e-05, + "loss": 0.1241, + "step": 62171 + }, + { + "epoch": 2.39, + "learning_rate": 1.0167991792767377e-05, + "loss": 0.1672, + "step": 62172 + }, + { + "epoch": 2.39, + "learning_rate": 1.0167350602718647e-05, + "loss": 0.2802, + "step": 62173 + }, + { + "epoch": 2.39, + "learning_rate": 1.0166709412669915e-05, + "loss": 0.1741, + "step": 62174 + }, + { + "epoch": 2.39, + "learning_rate": 1.0166068222621185e-05, + "loss": 0.3659, + "step": 62175 + }, + { + "epoch": 2.39, + "learning_rate": 1.0165427032572455e-05, + "loss": 0.2037, + "step": 62176 + }, + { + "epoch": 2.39, + "learning_rate": 1.0164785842523725e-05, + "loss": 0.2454, + "step": 62177 + }, + { + "epoch": 2.39, + "learning_rate": 1.0164144652474994e-05, + "loss": 0.1659, + "step": 62178 + }, + { + "epoch": 2.39, + "learning_rate": 1.0163503462426264e-05, + "loss": 0.249, + "step": 62179 + }, + { + "epoch": 2.39, + "learning_rate": 1.0162862272377534e-05, + "loss": 0.3103, + "step": 62180 + }, + { + "epoch": 2.39, + "learning_rate": 1.0162221082328804e-05, + "loss": 0.1945, + "step": 62181 + }, + { + "epoch": 2.39, + "learning_rate": 1.0161579892280072e-05, + "loss": 0.1937, + "step": 62182 + }, + { + "epoch": 2.39, + "learning_rate": 1.0160938702231342e-05, + "loss": 0.1569, + "step": 62183 + }, + { + "epoch": 2.39, + "learning_rate": 1.016029751218261e-05, + "loss": 0.2232, + "step": 62184 + }, + { + "epoch": 2.39, + "learning_rate": 1.0159656322133882e-05, + "loss": 0.3454, + "step": 62185 + }, + { + "epoch": 2.39, + "learning_rate": 1.015901513208515e-05, + "loss": 0.2469, + "step": 62186 + }, + { + "epoch": 2.39, + "learning_rate": 1.015837394203642e-05, + "loss": 0.2334, + "step": 62187 + }, + { + "epoch": 2.39, + "learning_rate": 1.0157732751987689e-05, + "loss": 0.1942, + "step": 62188 + }, + { + "epoch": 2.39, + "learning_rate": 1.015709156193896e-05, + "loss": 0.2413, + "step": 62189 + }, + { + "epoch": 2.39, + "learning_rate": 1.0156450371890229e-05, + "loss": 0.2155, + "step": 62190 + }, + { + "epoch": 2.39, + "learning_rate": 1.0155809181841499e-05, + "loss": 0.181, + "step": 62191 + }, + { + "epoch": 2.39, + "learning_rate": 1.0155167991792767e-05, + "loss": 0.3275, + "step": 62192 + }, + { + "epoch": 2.39, + "learning_rate": 1.0154526801744037e-05, + "loss": 0.2389, + "step": 62193 + }, + { + "epoch": 2.39, + "learning_rate": 1.0153885611695307e-05, + "loss": 0.1395, + "step": 62194 + }, + { + "epoch": 2.39, + "learning_rate": 1.0153244421646577e-05, + "loss": 0.2455, + "step": 62195 + }, + { + "epoch": 2.39, + "learning_rate": 1.0152603231597846e-05, + "loss": 0.2495, + "step": 62196 + }, + { + "epoch": 2.39, + "learning_rate": 1.0151962041549116e-05, + "loss": 0.2714, + "step": 62197 + }, + { + "epoch": 2.39, + "learning_rate": 1.0151320851500386e-05, + "loss": 0.1817, + "step": 62198 + }, + { + "epoch": 2.39, + "learning_rate": 1.0150679661451656e-05, + "loss": 0.277, + "step": 62199 + }, + { + "epoch": 2.39, + "learning_rate": 1.0150038471402924e-05, + "loss": 0.2317, + "step": 62200 + }, + { + "epoch": 2.39, + "learning_rate": 1.0149397281354194e-05, + "loss": 0.1759, + "step": 62201 + }, + { + "epoch": 2.39, + "learning_rate": 1.0148756091305462e-05, + "loss": 0.5099, + "step": 62202 + }, + { + "epoch": 2.39, + "learning_rate": 1.0148114901256734e-05, + "loss": 0.3053, + "step": 62203 + }, + { + "epoch": 2.39, + "learning_rate": 1.0147473711208002e-05, + "loss": 0.2113, + "step": 62204 + }, + { + "epoch": 2.39, + "learning_rate": 1.0146832521159272e-05, + "loss": 0.2402, + "step": 62205 + }, + { + "epoch": 2.39, + "learning_rate": 1.014619133111054e-05, + "loss": 0.1773, + "step": 62206 + }, + { + "epoch": 2.39, + "learning_rate": 1.0145550141061812e-05, + "loss": 0.1654, + "step": 62207 + }, + { + "epoch": 2.39, + "learning_rate": 1.014490895101308e-05, + "loss": 0.1927, + "step": 62208 + }, + { + "epoch": 2.39, + "learning_rate": 1.014426776096435e-05, + "loss": 0.3013, + "step": 62209 + }, + { + "epoch": 2.39, + "learning_rate": 1.0143626570915619e-05, + "loss": 0.1532, + "step": 62210 + }, + { + "epoch": 2.39, + "learning_rate": 1.014298538086689e-05, + "loss": 0.2681, + "step": 62211 + }, + { + "epoch": 2.39, + "learning_rate": 1.0142344190818159e-05, + "loss": 0.1356, + "step": 62212 + }, + { + "epoch": 2.39, + "learning_rate": 1.014170300076943e-05, + "loss": 0.2314, + "step": 62213 + }, + { + "epoch": 2.39, + "learning_rate": 1.0141061810720697e-05, + "loss": 0.1983, + "step": 62214 + }, + { + "epoch": 2.39, + "learning_rate": 1.0140420620671968e-05, + "loss": 0.2417, + "step": 62215 + }, + { + "epoch": 2.39, + "learning_rate": 1.0139779430623238e-05, + "loss": 0.191, + "step": 62216 + }, + { + "epoch": 2.39, + "learning_rate": 1.0139138240574508e-05, + "loss": 0.2455, + "step": 62217 + }, + { + "epoch": 2.39, + "learning_rate": 1.0138497050525776e-05, + "loss": 0.3522, + "step": 62218 + }, + { + "epoch": 2.39, + "learning_rate": 1.0137855860477046e-05, + "loss": 0.1226, + "step": 62219 + }, + { + "epoch": 2.39, + "learning_rate": 1.0137214670428316e-05, + "loss": 0.1591, + "step": 62220 + }, + { + "epoch": 2.39, + "learning_rate": 1.0136573480379586e-05, + "loss": 0.3324, + "step": 62221 + }, + { + "epoch": 2.39, + "learning_rate": 1.0135932290330854e-05, + "loss": 0.2965, + "step": 62222 + }, + { + "epoch": 2.39, + "learning_rate": 1.0135291100282124e-05, + "loss": 0.2442, + "step": 62223 + }, + { + "epoch": 2.39, + "learning_rate": 1.0134649910233393e-05, + "loss": 0.1833, + "step": 62224 + }, + { + "epoch": 2.39, + "learning_rate": 1.0134008720184664e-05, + "loss": 0.1893, + "step": 62225 + }, + { + "epoch": 2.39, + "learning_rate": 1.0133367530135933e-05, + "loss": 0.208, + "step": 62226 + }, + { + "epoch": 2.39, + "learning_rate": 1.0132726340087203e-05, + "loss": 0.3334, + "step": 62227 + }, + { + "epoch": 2.39, + "learning_rate": 1.0132085150038471e-05, + "loss": 0.1292, + "step": 62228 + }, + { + "epoch": 2.39, + "learning_rate": 1.0131443959989743e-05, + "loss": 0.1657, + "step": 62229 + }, + { + "epoch": 2.39, + "learning_rate": 1.0130802769941011e-05, + "loss": 0.3216, + "step": 62230 + }, + { + "epoch": 2.39, + "learning_rate": 1.0130161579892281e-05, + "loss": 0.1745, + "step": 62231 + }, + { + "epoch": 2.39, + "learning_rate": 1.012952038984355e-05, + "loss": 0.1837, + "step": 62232 + }, + { + "epoch": 2.39, + "learning_rate": 1.012887919979482e-05, + "loss": 0.1888, + "step": 62233 + }, + { + "epoch": 2.39, + "learning_rate": 1.012823800974609e-05, + "loss": 0.2088, + "step": 62234 + }, + { + "epoch": 2.39, + "learning_rate": 1.012759681969736e-05, + "loss": 0.2592, + "step": 62235 + }, + { + "epoch": 2.39, + "learning_rate": 1.0126955629648628e-05, + "loss": 0.2236, + "step": 62236 + }, + { + "epoch": 2.39, + "learning_rate": 1.0126314439599898e-05, + "loss": 0.2607, + "step": 62237 + }, + { + "epoch": 2.39, + "learning_rate": 1.0125673249551168e-05, + "loss": 0.3095, + "step": 62238 + }, + { + "epoch": 2.39, + "learning_rate": 1.0125032059502438e-05, + "loss": 0.2831, + "step": 62239 + }, + { + "epoch": 2.39, + "learning_rate": 1.0124390869453706e-05, + "loss": 0.2216, + "step": 62240 + }, + { + "epoch": 2.39, + "learning_rate": 1.0123749679404976e-05, + "loss": 0.2041, + "step": 62241 + }, + { + "epoch": 2.39, + "learning_rate": 1.0123108489356245e-05, + "loss": 0.1761, + "step": 62242 + }, + { + "epoch": 2.39, + "learning_rate": 1.0122467299307516e-05, + "loss": 0.1723, + "step": 62243 + }, + { + "epoch": 2.39, + "learning_rate": 1.0121826109258785e-05, + "loss": 0.1887, + "step": 62244 + }, + { + "epoch": 2.39, + "learning_rate": 1.0121184919210055e-05, + "loss": 0.3344, + "step": 62245 + }, + { + "epoch": 2.39, + "learning_rate": 1.0120543729161323e-05, + "loss": 0.2043, + "step": 62246 + }, + { + "epoch": 2.39, + "learning_rate": 1.0119902539112595e-05, + "loss": 0.1651, + "step": 62247 + }, + { + "epoch": 2.39, + "learning_rate": 1.0119261349063863e-05, + "loss": 0.2169, + "step": 62248 + }, + { + "epoch": 2.39, + "learning_rate": 1.0118620159015133e-05, + "loss": 0.2121, + "step": 62249 + }, + { + "epoch": 2.39, + "learning_rate": 1.0117978968966401e-05, + "loss": 0.2175, + "step": 62250 + }, + { + "epoch": 2.39, + "learning_rate": 1.0117337778917673e-05, + "loss": 0.1683, + "step": 62251 + }, + { + "epoch": 2.39, + "learning_rate": 1.0116696588868941e-05, + "loss": 0.2663, + "step": 62252 + }, + { + "epoch": 2.39, + "learning_rate": 1.0116055398820211e-05, + "loss": 0.1486, + "step": 62253 + }, + { + "epoch": 2.39, + "learning_rate": 1.011541420877148e-05, + "loss": 0.3152, + "step": 62254 + }, + { + "epoch": 2.39, + "learning_rate": 1.011477301872275e-05, + "loss": 0.183, + "step": 62255 + }, + { + "epoch": 2.39, + "learning_rate": 1.011413182867402e-05, + "loss": 0.1961, + "step": 62256 + }, + { + "epoch": 2.39, + "learning_rate": 1.011349063862529e-05, + "loss": 0.1581, + "step": 62257 + }, + { + "epoch": 2.39, + "learning_rate": 1.0112849448576558e-05, + "loss": 0.1837, + "step": 62258 + }, + { + "epoch": 2.39, + "learning_rate": 1.0112208258527828e-05, + "loss": 0.1069, + "step": 62259 + }, + { + "epoch": 2.39, + "learning_rate": 1.0111567068479098e-05, + "loss": 0.1987, + "step": 62260 + }, + { + "epoch": 2.39, + "learning_rate": 1.0110925878430368e-05, + "loss": 0.1681, + "step": 62261 + }, + { + "epoch": 2.39, + "learning_rate": 1.0110284688381637e-05, + "loss": 0.2226, + "step": 62262 + }, + { + "epoch": 2.39, + "learning_rate": 1.0109643498332907e-05, + "loss": 0.1754, + "step": 62263 + }, + { + "epoch": 2.39, + "learning_rate": 1.0109002308284175e-05, + "loss": 0.1516, + "step": 62264 + }, + { + "epoch": 2.39, + "learning_rate": 1.0108361118235447e-05, + "loss": 0.2819, + "step": 62265 + }, + { + "epoch": 2.39, + "learning_rate": 1.0107719928186715e-05, + "loss": 0.2, + "step": 62266 + }, + { + "epoch": 2.39, + "learning_rate": 1.0107078738137985e-05, + "loss": 0.1919, + "step": 62267 + }, + { + "epoch": 2.39, + "learning_rate": 1.0106437548089253e-05, + "loss": 0.2053, + "step": 62268 + }, + { + "epoch": 2.39, + "learning_rate": 1.0105796358040525e-05, + "loss": 0.2798, + "step": 62269 + }, + { + "epoch": 2.39, + "learning_rate": 1.0105155167991793e-05, + "loss": 0.3749, + "step": 62270 + }, + { + "epoch": 2.39, + "learning_rate": 1.0104513977943063e-05, + "loss": 0.1547, + "step": 62271 + }, + { + "epoch": 2.39, + "learning_rate": 1.0103872787894332e-05, + "loss": 0.2466, + "step": 62272 + }, + { + "epoch": 2.39, + "learning_rate": 1.0103231597845602e-05, + "loss": 0.2732, + "step": 62273 + }, + { + "epoch": 2.39, + "learning_rate": 1.0102590407796872e-05, + "loss": 0.1944, + "step": 62274 + }, + { + "epoch": 2.39, + "learning_rate": 1.0101949217748142e-05, + "loss": 0.2939, + "step": 62275 + }, + { + "epoch": 2.39, + "learning_rate": 1.010130802769941e-05, + "loss": 0.2526, + "step": 62276 + }, + { + "epoch": 2.39, + "learning_rate": 1.010066683765068e-05, + "loss": 0.1465, + "step": 62277 + }, + { + "epoch": 2.39, + "learning_rate": 1.010002564760195e-05, + "loss": 0.1484, + "step": 62278 + }, + { + "epoch": 2.39, + "learning_rate": 1.009938445755322e-05, + "loss": 0.2101, + "step": 62279 + }, + { + "epoch": 2.39, + "learning_rate": 1.0098743267504488e-05, + "loss": 0.2421, + "step": 62280 + }, + { + "epoch": 2.39, + "learning_rate": 1.0098102077455759e-05, + "loss": 0.1296, + "step": 62281 + }, + { + "epoch": 2.39, + "learning_rate": 1.0097460887407027e-05, + "loss": 0.21, + "step": 62282 + }, + { + "epoch": 2.39, + "learning_rate": 1.0096819697358299e-05, + "loss": 0.2623, + "step": 62283 + }, + { + "epoch": 2.39, + "learning_rate": 1.0096178507309567e-05, + "loss": 0.227, + "step": 62284 + }, + { + "epoch": 2.39, + "learning_rate": 1.0095537317260837e-05, + "loss": 0.2921, + "step": 62285 + }, + { + "epoch": 2.39, + "learning_rate": 1.0094896127212105e-05, + "loss": 0.2096, + "step": 62286 + }, + { + "epoch": 2.39, + "learning_rate": 1.0094254937163377e-05, + "loss": 0.2278, + "step": 62287 + }, + { + "epoch": 2.39, + "learning_rate": 1.0093613747114645e-05, + "loss": 0.2612, + "step": 62288 + }, + { + "epoch": 2.39, + "learning_rate": 1.0092972557065915e-05, + "loss": 0.3027, + "step": 62289 + }, + { + "epoch": 2.39, + "learning_rate": 1.0092331367017184e-05, + "loss": 0.225, + "step": 62290 + }, + { + "epoch": 2.39, + "learning_rate": 1.0091690176968454e-05, + "loss": 0.2669, + "step": 62291 + }, + { + "epoch": 2.39, + "learning_rate": 1.0091048986919724e-05, + "loss": 0.2008, + "step": 62292 + }, + { + "epoch": 2.39, + "learning_rate": 1.0090407796870994e-05, + "loss": 0.1194, + "step": 62293 + }, + { + "epoch": 2.4, + "learning_rate": 1.0089766606822262e-05, + "loss": 0.3125, + "step": 62294 + }, + { + "epoch": 2.4, + "learning_rate": 1.0089125416773532e-05, + "loss": 0.3174, + "step": 62295 + }, + { + "epoch": 2.4, + "learning_rate": 1.0088484226724802e-05, + "loss": 0.2829, + "step": 62296 + }, + { + "epoch": 2.4, + "learning_rate": 1.0087843036676072e-05, + "loss": 0.3362, + "step": 62297 + }, + { + "epoch": 2.4, + "learning_rate": 1.008720184662734e-05, + "loss": 0.1509, + "step": 62298 + }, + { + "epoch": 2.4, + "learning_rate": 1.008656065657861e-05, + "loss": 0.2629, + "step": 62299 + }, + { + "epoch": 2.4, + "learning_rate": 1.008591946652988e-05, + "loss": 0.1302, + "step": 62300 + }, + { + "epoch": 2.4, + "learning_rate": 1.008527827648115e-05, + "loss": 0.3887, + "step": 62301 + }, + { + "epoch": 2.4, + "learning_rate": 1.0084637086432419e-05, + "loss": 0.1188, + "step": 62302 + }, + { + "epoch": 2.4, + "learning_rate": 1.0083995896383689e-05, + "loss": 0.355, + "step": 62303 + }, + { + "epoch": 2.4, + "learning_rate": 1.0083354706334957e-05, + "loss": 0.2659, + "step": 62304 + }, + { + "epoch": 2.4, + "learning_rate": 1.0082713516286229e-05, + "loss": 0.3436, + "step": 62305 + }, + { + "epoch": 2.4, + "learning_rate": 1.0082072326237497e-05, + "loss": 0.1779, + "step": 62306 + }, + { + "epoch": 2.4, + "learning_rate": 1.0081431136188767e-05, + "loss": 0.2005, + "step": 62307 + }, + { + "epoch": 2.4, + "learning_rate": 1.0080789946140036e-05, + "loss": 0.1285, + "step": 62308 + }, + { + "epoch": 2.4, + "learning_rate": 1.0080148756091307e-05, + "loss": 0.2524, + "step": 62309 + }, + { + "epoch": 2.4, + "learning_rate": 1.0079507566042576e-05, + "loss": 0.2096, + "step": 62310 + }, + { + "epoch": 2.4, + "learning_rate": 1.0078866375993846e-05, + "loss": 0.2502, + "step": 62311 + }, + { + "epoch": 2.4, + "learning_rate": 1.0078225185945114e-05, + "loss": 0.1746, + "step": 62312 + }, + { + "epoch": 2.4, + "learning_rate": 1.0077583995896384e-05, + "loss": 0.1679, + "step": 62313 + }, + { + "epoch": 2.4, + "learning_rate": 1.0076942805847654e-05, + "loss": 0.1966, + "step": 62314 + }, + { + "epoch": 2.4, + "learning_rate": 1.0076301615798924e-05, + "loss": 0.1774, + "step": 62315 + }, + { + "epoch": 2.4, + "learning_rate": 1.0075660425750192e-05, + "loss": 0.158, + "step": 62316 + }, + { + "epoch": 2.4, + "learning_rate": 1.0075019235701462e-05, + "loss": 0.234, + "step": 62317 + }, + { + "epoch": 2.4, + "learning_rate": 1.0074378045652732e-05, + "loss": 0.3255, + "step": 62318 + }, + { + "epoch": 2.4, + "learning_rate": 1.0073736855604002e-05, + "loss": 0.2259, + "step": 62319 + }, + { + "epoch": 2.4, + "learning_rate": 1.007309566555527e-05, + "loss": 0.1409, + "step": 62320 + }, + { + "epoch": 2.4, + "learning_rate": 1.007245447550654e-05, + "loss": 0.1119, + "step": 62321 + }, + { + "epoch": 2.4, + "learning_rate": 1.0071813285457809e-05, + "loss": 0.1416, + "step": 62322 + }, + { + "epoch": 2.4, + "learning_rate": 1.007117209540908e-05, + "loss": 0.2171, + "step": 62323 + }, + { + "epoch": 2.4, + "learning_rate": 1.007053090536035e-05, + "loss": 0.1981, + "step": 62324 + }, + { + "epoch": 2.4, + "learning_rate": 1.006988971531162e-05, + "loss": 0.2791, + "step": 62325 + }, + { + "epoch": 2.4, + "learning_rate": 1.0069248525262888e-05, + "loss": 0.2077, + "step": 62326 + }, + { + "epoch": 2.4, + "learning_rate": 1.006860733521416e-05, + "loss": 0.2765, + "step": 62327 + }, + { + "epoch": 2.4, + "learning_rate": 1.0067966145165428e-05, + "loss": 0.2191, + "step": 62328 + }, + { + "epoch": 2.4, + "learning_rate": 1.0067324955116698e-05, + "loss": 0.1432, + "step": 62329 + }, + { + "epoch": 2.4, + "learning_rate": 1.0066683765067966e-05, + "loss": 0.1116, + "step": 62330 + }, + { + "epoch": 2.4, + "learning_rate": 1.0066042575019236e-05, + "loss": 0.1814, + "step": 62331 + }, + { + "epoch": 2.4, + "learning_rate": 1.0065401384970506e-05, + "loss": 0.232, + "step": 62332 + }, + { + "epoch": 2.4, + "learning_rate": 1.0064760194921776e-05, + "loss": 0.1306, + "step": 62333 + }, + { + "epoch": 2.4, + "learning_rate": 1.0064119004873044e-05, + "loss": 0.4416, + "step": 62334 + }, + { + "epoch": 2.4, + "learning_rate": 1.0063477814824314e-05, + "loss": 0.2924, + "step": 62335 + }, + { + "epoch": 2.4, + "learning_rate": 1.0062836624775584e-05, + "loss": 0.2679, + "step": 62336 + }, + { + "epoch": 2.4, + "learning_rate": 1.0062195434726854e-05, + "loss": 0.2404, + "step": 62337 + }, + { + "epoch": 2.4, + "learning_rate": 1.0061554244678123e-05, + "loss": 0.306, + "step": 62338 + }, + { + "epoch": 2.4, + "learning_rate": 1.0060913054629393e-05, + "loss": 0.1871, + "step": 62339 + }, + { + "epoch": 2.4, + "learning_rate": 1.0060271864580661e-05, + "loss": 0.1482, + "step": 62340 + }, + { + "epoch": 2.4, + "learning_rate": 1.0059630674531933e-05, + "loss": 0.2052, + "step": 62341 + }, + { + "epoch": 2.4, + "learning_rate": 1.0058989484483201e-05, + "loss": 0.1693, + "step": 62342 + }, + { + "epoch": 2.4, + "learning_rate": 1.0058348294434471e-05, + "loss": 0.1556, + "step": 62343 + }, + { + "epoch": 2.4, + "learning_rate": 1.005770710438574e-05, + "loss": 0.2289, + "step": 62344 + }, + { + "epoch": 2.4, + "learning_rate": 1.0057065914337011e-05, + "loss": 0.2189, + "step": 62345 + }, + { + "epoch": 2.4, + "learning_rate": 1.005642472428828e-05, + "loss": 0.1521, + "step": 62346 + }, + { + "epoch": 2.4, + "learning_rate": 1.005578353423955e-05, + "loss": 0.2864, + "step": 62347 + }, + { + "epoch": 2.4, + "learning_rate": 1.0055142344190818e-05, + "loss": 0.2026, + "step": 62348 + }, + { + "epoch": 2.4, + "learning_rate": 1.005450115414209e-05, + "loss": 0.1437, + "step": 62349 + }, + { + "epoch": 2.4, + "learning_rate": 1.0053859964093358e-05, + "loss": 0.1242, + "step": 62350 + }, + { + "epoch": 2.4, + "learning_rate": 1.0053218774044628e-05, + "loss": 0.2921, + "step": 62351 + }, + { + "epoch": 2.4, + "learning_rate": 1.0052577583995896e-05, + "loss": 0.2631, + "step": 62352 + }, + { + "epoch": 2.4, + "learning_rate": 1.0051936393947166e-05, + "loss": 0.2921, + "step": 62353 + }, + { + "epoch": 2.4, + "learning_rate": 1.0051295203898436e-05, + "loss": 0.2184, + "step": 62354 + }, + { + "epoch": 2.4, + "learning_rate": 1.0050654013849706e-05, + "loss": 0.1742, + "step": 62355 + }, + { + "epoch": 2.4, + "learning_rate": 1.0050012823800975e-05, + "loss": 0.2155, + "step": 62356 + }, + { + "epoch": 2.4, + "learning_rate": 1.0049371633752245e-05, + "loss": 0.2433, + "step": 62357 + }, + { + "epoch": 2.4, + "learning_rate": 1.0048730443703515e-05, + "loss": 0.1372, + "step": 62358 + }, + { + "epoch": 2.4, + "learning_rate": 1.0048089253654785e-05, + "loss": 0.2595, + "step": 62359 + }, + { + "epoch": 2.4, + "learning_rate": 1.0047448063606053e-05, + "loss": 0.2474, + "step": 62360 + }, + { + "epoch": 2.4, + "learning_rate": 1.0046806873557323e-05, + "loss": 0.3116, + "step": 62361 + }, + { + "epoch": 2.4, + "learning_rate": 1.0046165683508591e-05, + "loss": 0.3064, + "step": 62362 + }, + { + "epoch": 2.4, + "learning_rate": 1.0045524493459863e-05, + "loss": 0.1942, + "step": 62363 + }, + { + "epoch": 2.4, + "learning_rate": 1.0044883303411131e-05, + "loss": 0.1057, + "step": 62364 + }, + { + "epoch": 2.4, + "learning_rate": 1.0044242113362401e-05, + "loss": 0.2608, + "step": 62365 + }, + { + "epoch": 2.4, + "learning_rate": 1.004360092331367e-05, + "loss": 0.2035, + "step": 62366 + }, + { + "epoch": 2.4, + "learning_rate": 1.0042959733264941e-05, + "loss": 0.3513, + "step": 62367 + }, + { + "epoch": 2.4, + "learning_rate": 1.004231854321621e-05, + "loss": 0.2738, + "step": 62368 + }, + { + "epoch": 2.4, + "learning_rate": 1.004167735316748e-05, + "loss": 0.262, + "step": 62369 + }, + { + "epoch": 2.4, + "learning_rate": 1.0041036163118748e-05, + "loss": 0.2452, + "step": 62370 + }, + { + "epoch": 2.4, + "learning_rate": 1.0040394973070018e-05, + "loss": 0.2993, + "step": 62371 + }, + { + "epoch": 2.4, + "learning_rate": 1.0039753783021288e-05, + "loss": 0.1478, + "step": 62372 + }, + { + "epoch": 2.4, + "learning_rate": 1.0039112592972558e-05, + "loss": 0.1258, + "step": 62373 + }, + { + "epoch": 2.4, + "learning_rate": 1.0038471402923827e-05, + "loss": 0.2729, + "step": 62374 + }, + { + "epoch": 2.4, + "learning_rate": 1.0037830212875097e-05, + "loss": 0.1911, + "step": 62375 + }, + { + "epoch": 2.4, + "learning_rate": 1.0037189022826367e-05, + "loss": 0.2391, + "step": 62376 + }, + { + "epoch": 2.4, + "learning_rate": 1.0036547832777637e-05, + "loss": 0.3142, + "step": 62377 + }, + { + "epoch": 2.4, + "learning_rate": 1.0035906642728905e-05, + "loss": 0.269, + "step": 62378 + }, + { + "epoch": 2.4, + "learning_rate": 1.0035265452680175e-05, + "loss": 0.2413, + "step": 62379 + }, + { + "epoch": 2.4, + "learning_rate": 1.0034624262631443e-05, + "loss": 0.1089, + "step": 62380 + }, + { + "epoch": 2.4, + "learning_rate": 1.0033983072582715e-05, + "loss": 0.2284, + "step": 62381 + }, + { + "epoch": 2.4, + "learning_rate": 1.0033341882533983e-05, + "loss": 0.2445, + "step": 62382 + }, + { + "epoch": 2.4, + "learning_rate": 1.0032700692485253e-05, + "loss": 0.1767, + "step": 62383 + }, + { + "epoch": 2.4, + "learning_rate": 1.0032059502436522e-05, + "loss": 0.2708, + "step": 62384 + }, + { + "epoch": 2.4, + "learning_rate": 1.0031418312387793e-05, + "loss": 0.1815, + "step": 62385 + }, + { + "epoch": 2.4, + "learning_rate": 1.0030777122339062e-05, + "loss": 0.2611, + "step": 62386 + }, + { + "epoch": 2.4, + "learning_rate": 1.0030135932290332e-05, + "loss": 0.1958, + "step": 62387 + }, + { + "epoch": 2.4, + "learning_rate": 1.00294947422416e-05, + "loss": 0.1606, + "step": 62388 + }, + { + "epoch": 2.4, + "learning_rate": 1.002885355219287e-05, + "loss": 0.2911, + "step": 62389 + }, + { + "epoch": 2.4, + "learning_rate": 1.002821236214414e-05, + "loss": 0.1595, + "step": 62390 + }, + { + "epoch": 2.4, + "learning_rate": 1.002757117209541e-05, + "loss": 0.2675, + "step": 62391 + }, + { + "epoch": 2.4, + "learning_rate": 1.0026929982046679e-05, + "loss": 0.3005, + "step": 62392 + }, + { + "epoch": 2.4, + "learning_rate": 1.0026288791997949e-05, + "loss": 0.2737, + "step": 62393 + }, + { + "epoch": 2.4, + "learning_rate": 1.0025647601949219e-05, + "loss": 0.2529, + "step": 62394 + }, + { + "epoch": 2.4, + "learning_rate": 1.0025006411900489e-05, + "loss": 0.1766, + "step": 62395 + }, + { + "epoch": 2.4, + "learning_rate": 1.0024365221851757e-05, + "loss": 0.1932, + "step": 62396 + }, + { + "epoch": 2.4, + "learning_rate": 1.0023724031803027e-05, + "loss": 0.2875, + "step": 62397 + }, + { + "epoch": 2.4, + "learning_rate": 1.0023082841754297e-05, + "loss": 0.2015, + "step": 62398 + }, + { + "epoch": 2.4, + "learning_rate": 1.0022441651705567e-05, + "loss": 0.2536, + "step": 62399 + }, + { + "epoch": 2.4, + "learning_rate": 1.0021800461656835e-05, + "loss": 0.2292, + "step": 62400 + }, + { + "epoch": 2.4, + "learning_rate": 1.0021159271608105e-05, + "loss": 0.2222, + "step": 62401 + }, + { + "epoch": 2.4, + "learning_rate": 1.0020518081559374e-05, + "loss": 0.27, + "step": 62402 + }, + { + "epoch": 2.4, + "learning_rate": 1.0019876891510645e-05, + "loss": 0.1755, + "step": 62403 + }, + { + "epoch": 2.4, + "learning_rate": 1.0019235701461914e-05, + "loss": 0.1069, + "step": 62404 + }, + { + "epoch": 2.4, + "learning_rate": 1.0018594511413184e-05, + "loss": 0.1464, + "step": 62405 + }, + { + "epoch": 2.4, + "learning_rate": 1.0017953321364452e-05, + "loss": 0.1385, + "step": 62406 + }, + { + "epoch": 2.4, + "learning_rate": 1.0017312131315724e-05, + "loss": 0.2583, + "step": 62407 + }, + { + "epoch": 2.4, + "learning_rate": 1.0016670941266992e-05, + "loss": 0.2959, + "step": 62408 + }, + { + "epoch": 2.4, + "learning_rate": 1.0016029751218262e-05, + "loss": 0.1518, + "step": 62409 + }, + { + "epoch": 2.4, + "learning_rate": 1.001538856116953e-05, + "loss": 0.2686, + "step": 62410 + }, + { + "epoch": 2.4, + "learning_rate": 1.00147473711208e-05, + "loss": 0.0925, + "step": 62411 + }, + { + "epoch": 2.4, + "learning_rate": 1.001410618107207e-05, + "loss": 0.1823, + "step": 62412 + }, + { + "epoch": 2.4, + "learning_rate": 1.001346499102334e-05, + "loss": 0.1551, + "step": 62413 + }, + { + "epoch": 2.4, + "learning_rate": 1.0012823800974609e-05, + "loss": 0.183, + "step": 62414 + }, + { + "epoch": 2.4, + "learning_rate": 1.0012182610925879e-05, + "loss": 0.2277, + "step": 62415 + }, + { + "epoch": 2.4, + "learning_rate": 1.0011541420877149e-05, + "loss": 0.2793, + "step": 62416 + }, + { + "epoch": 2.4, + "learning_rate": 1.0010900230828419e-05, + "loss": 0.2273, + "step": 62417 + }, + { + "epoch": 2.4, + "learning_rate": 1.0010259040779687e-05, + "loss": 0.2411, + "step": 62418 + }, + { + "epoch": 2.4, + "learning_rate": 1.0009617850730957e-05, + "loss": 0.1931, + "step": 62419 + }, + { + "epoch": 2.4, + "learning_rate": 1.0008976660682226e-05, + "loss": 0.3106, + "step": 62420 + }, + { + "epoch": 2.4, + "learning_rate": 1.0008335470633497e-05, + "loss": 0.2588, + "step": 62421 + }, + { + "epoch": 2.4, + "learning_rate": 1.0007694280584766e-05, + "loss": 0.164, + "step": 62422 + }, + { + "epoch": 2.4, + "learning_rate": 1.0007053090536036e-05, + "loss": 0.1374, + "step": 62423 + }, + { + "epoch": 2.4, + "learning_rate": 1.0006411900487304e-05, + "loss": 0.2469, + "step": 62424 + }, + { + "epoch": 2.4, + "learning_rate": 1.0005770710438576e-05, + "loss": 0.3096, + "step": 62425 + }, + { + "epoch": 2.4, + "learning_rate": 1.0005129520389844e-05, + "loss": 0.4467, + "step": 62426 + }, + { + "epoch": 2.4, + "learning_rate": 1.0004488330341114e-05, + "loss": 0.175, + "step": 62427 + }, + { + "epoch": 2.4, + "learning_rate": 1.0003847140292382e-05, + "loss": 0.2766, + "step": 62428 + }, + { + "epoch": 2.4, + "learning_rate": 1.0003205950243652e-05, + "loss": 0.2221, + "step": 62429 + }, + { + "epoch": 2.4, + "learning_rate": 1.0002564760194922e-05, + "loss": 0.2095, + "step": 62430 + }, + { + "epoch": 2.4, + "learning_rate": 1.0001923570146192e-05, + "loss": 0.1995, + "step": 62431 + }, + { + "epoch": 2.4, + "learning_rate": 1.000128238009746e-05, + "loss": 0.2433, + "step": 62432 + }, + { + "epoch": 2.4, + "learning_rate": 1.000064119004873e-05, + "loss": 0.2607, + "step": 62433 + }, + { + "epoch": 2.4, + "learning_rate": 1e-05, + "loss": 0.1587, + "step": 62434 + }, + { + "epoch": 2.4, + "learning_rate": 9.99935880995127e-06, + "loss": 0.1428, + "step": 62435 + }, + { + "epoch": 2.4, + "learning_rate": 9.99871761990254e-06, + "loss": 0.2185, + "step": 62436 + }, + { + "epoch": 2.4, + "learning_rate": 9.99807642985381e-06, + "loss": 0.236, + "step": 62437 + }, + { + "epoch": 2.4, + "learning_rate": 9.997435239805078e-06, + "loss": 0.1728, + "step": 62438 + }, + { + "epoch": 2.4, + "learning_rate": 9.99679404975635e-06, + "loss": 0.3106, + "step": 62439 + }, + { + "epoch": 2.4, + "learning_rate": 9.996152859707618e-06, + "loss": 0.1645, + "step": 62440 + }, + { + "epoch": 2.4, + "learning_rate": 9.995511669658888e-06, + "loss": 0.2814, + "step": 62441 + }, + { + "epoch": 2.4, + "learning_rate": 9.994870479610156e-06, + "loss": 0.1638, + "step": 62442 + }, + { + "epoch": 2.4, + "learning_rate": 9.994229289561428e-06, + "loss": 0.2763, + "step": 62443 + }, + { + "epoch": 2.4, + "learning_rate": 9.993588099512696e-06, + "loss": 0.1938, + "step": 62444 + }, + { + "epoch": 2.4, + "learning_rate": 9.992946909463966e-06, + "loss": 0.1601, + "step": 62445 + }, + { + "epoch": 2.4, + "learning_rate": 9.992305719415234e-06, + "loss": 0.2243, + "step": 62446 + }, + { + "epoch": 2.4, + "learning_rate": 9.991664529366506e-06, + "loss": 0.2822, + "step": 62447 + }, + { + "epoch": 2.4, + "learning_rate": 9.991023339317774e-06, + "loss": 0.3371, + "step": 62448 + }, + { + "epoch": 2.4, + "learning_rate": 9.990382149269044e-06, + "loss": 0.1603, + "step": 62449 + }, + { + "epoch": 2.4, + "learning_rate": 9.989740959220313e-06, + "loss": 0.2607, + "step": 62450 + }, + { + "epoch": 2.4, + "learning_rate": 9.989099769171583e-06, + "loss": 0.2123, + "step": 62451 + }, + { + "epoch": 2.4, + "learning_rate": 9.988458579122853e-06, + "loss": 0.2052, + "step": 62452 + }, + { + "epoch": 2.4, + "learning_rate": 9.987817389074123e-06, + "loss": 0.3989, + "step": 62453 + }, + { + "epoch": 2.4, + "learning_rate": 9.987176199025391e-06, + "loss": 0.2645, + "step": 62454 + }, + { + "epoch": 2.4, + "learning_rate": 9.986535008976661e-06, + "loss": 0.2411, + "step": 62455 + }, + { + "epoch": 2.4, + "learning_rate": 9.985893818927931e-06, + "loss": 0.1594, + "step": 62456 + }, + { + "epoch": 2.4, + "learning_rate": 9.985252628879201e-06, + "loss": 0.1096, + "step": 62457 + }, + { + "epoch": 2.4, + "learning_rate": 9.98461143883047e-06, + "loss": 0.1702, + "step": 62458 + }, + { + "epoch": 2.4, + "learning_rate": 9.98397024878174e-06, + "loss": 0.1592, + "step": 62459 + }, + { + "epoch": 2.4, + "learning_rate": 9.983329058733008e-06, + "loss": 0.3077, + "step": 62460 + }, + { + "epoch": 2.4, + "learning_rate": 9.98268786868428e-06, + "loss": 0.3929, + "step": 62461 + }, + { + "epoch": 2.4, + "learning_rate": 9.982046678635548e-06, + "loss": 0.2338, + "step": 62462 + }, + { + "epoch": 2.4, + "learning_rate": 9.981405488586818e-06, + "loss": 0.1539, + "step": 62463 + }, + { + "epoch": 2.4, + "learning_rate": 9.980764298538086e-06, + "loss": 0.1107, + "step": 62464 + }, + { + "epoch": 2.4, + "learning_rate": 9.980123108489358e-06, + "loss": 0.2526, + "step": 62465 + }, + { + "epoch": 2.4, + "learning_rate": 9.979481918440626e-06, + "loss": 0.1237, + "step": 62466 + }, + { + "epoch": 2.4, + "learning_rate": 9.978840728391896e-06, + "loss": 0.251, + "step": 62467 + }, + { + "epoch": 2.4, + "learning_rate": 9.978199538343165e-06, + "loss": 0.3712, + "step": 62468 + }, + { + "epoch": 2.4, + "learning_rate": 9.977558348294435e-06, + "loss": 0.2668, + "step": 62469 + }, + { + "epoch": 2.4, + "learning_rate": 9.976917158245705e-06, + "loss": 0.2358, + "step": 62470 + }, + { + "epoch": 2.4, + "learning_rate": 9.976275968196975e-06, + "loss": 0.2612, + "step": 62471 + }, + { + "epoch": 2.4, + "learning_rate": 9.975634778148243e-06, + "loss": 0.2773, + "step": 62472 + }, + { + "epoch": 2.4, + "learning_rate": 9.974993588099513e-06, + "loss": 0.1408, + "step": 62473 + }, + { + "epoch": 2.4, + "learning_rate": 9.974352398050783e-06, + "loss": 0.2647, + "step": 62474 + }, + { + "epoch": 2.4, + "learning_rate": 9.973711208002053e-06, + "loss": 0.2432, + "step": 62475 + }, + { + "epoch": 2.4, + "learning_rate": 9.973070017953321e-06, + "loss": 0.2146, + "step": 62476 + }, + { + "epoch": 2.4, + "learning_rate": 9.972428827904591e-06, + "loss": 0.2269, + "step": 62477 + }, + { + "epoch": 2.4, + "learning_rate": 9.97178763785586e-06, + "loss": 0.1656, + "step": 62478 + }, + { + "epoch": 2.4, + "learning_rate": 9.971146447807132e-06, + "loss": 0.1956, + "step": 62479 + }, + { + "epoch": 2.4, + "learning_rate": 9.9705052577584e-06, + "loss": 0.3206, + "step": 62480 + }, + { + "epoch": 2.4, + "learning_rate": 9.96986406770967e-06, + "loss": 0.2526, + "step": 62481 + }, + { + "epoch": 2.4, + "learning_rate": 9.969222877660938e-06, + "loss": 0.2907, + "step": 62482 + }, + { + "epoch": 2.4, + "learning_rate": 9.96858168761221e-06, + "loss": 0.3312, + "step": 62483 + }, + { + "epoch": 2.4, + "learning_rate": 9.967940497563478e-06, + "loss": 0.3162, + "step": 62484 + }, + { + "epoch": 2.4, + "learning_rate": 9.967299307514748e-06, + "loss": 0.1592, + "step": 62485 + }, + { + "epoch": 2.4, + "learning_rate": 9.966658117466017e-06, + "loss": 0.2636, + "step": 62486 + }, + { + "epoch": 2.4, + "learning_rate": 9.966016927417288e-06, + "loss": 0.1626, + "step": 62487 + }, + { + "epoch": 2.4, + "learning_rate": 9.965375737368557e-06, + "loss": 0.3016, + "step": 62488 + }, + { + "epoch": 2.4, + "learning_rate": 9.964734547319827e-06, + "loss": 0.2105, + "step": 62489 + }, + { + "epoch": 2.4, + "learning_rate": 9.964093357271095e-06, + "loss": 0.1874, + "step": 62490 + }, + { + "epoch": 2.4, + "learning_rate": 9.963452167222365e-06, + "loss": 0.1799, + "step": 62491 + }, + { + "epoch": 2.4, + "learning_rate": 9.962810977173635e-06, + "loss": 0.2688, + "step": 62492 + }, + { + "epoch": 2.4, + "learning_rate": 9.962169787124905e-06, + "loss": 0.3807, + "step": 62493 + }, + { + "epoch": 2.4, + "learning_rate": 9.961528597076173e-06, + "loss": 0.2507, + "step": 62494 + }, + { + "epoch": 2.4, + "learning_rate": 9.960887407027443e-06, + "loss": 0.2023, + "step": 62495 + }, + { + "epoch": 2.4, + "learning_rate": 9.960246216978713e-06, + "loss": 0.224, + "step": 62496 + }, + { + "epoch": 2.4, + "learning_rate": 9.959605026929983e-06, + "loss": 0.1389, + "step": 62497 + }, + { + "epoch": 2.4, + "learning_rate": 9.958963836881252e-06, + "loss": 0.2655, + "step": 62498 + }, + { + "epoch": 2.4, + "learning_rate": 9.958322646832522e-06, + "loss": 0.2099, + "step": 62499 + }, + { + "epoch": 2.4, + "learning_rate": 9.95768145678379e-06, + "loss": 0.2593, + "step": 62500 + }, + { + "epoch": 2.4, + "learning_rate": 9.957040266735062e-06, + "loss": 0.2019, + "step": 62501 + }, + { + "epoch": 2.4, + "learning_rate": 9.95639907668633e-06, + "loss": 0.3325, + "step": 62502 + }, + { + "epoch": 2.4, + "learning_rate": 9.9557578866376e-06, + "loss": 0.2849, + "step": 62503 + }, + { + "epoch": 2.4, + "learning_rate": 9.955116696588869e-06, + "loss": 0.3271, + "step": 62504 + }, + { + "epoch": 2.4, + "learning_rate": 9.95447550654014e-06, + "loss": 0.2578, + "step": 62505 + }, + { + "epoch": 2.4, + "learning_rate": 9.953834316491409e-06, + "loss": 0.2809, + "step": 62506 + }, + { + "epoch": 2.4, + "learning_rate": 9.953193126442679e-06, + "loss": 0.291, + "step": 62507 + }, + { + "epoch": 2.4, + "learning_rate": 9.952551936393947e-06, + "loss": 0.3157, + "step": 62508 + }, + { + "epoch": 2.4, + "learning_rate": 9.951910746345217e-06, + "loss": 0.3507, + "step": 62509 + }, + { + "epoch": 2.4, + "learning_rate": 9.951269556296487e-06, + "loss": 0.1935, + "step": 62510 + }, + { + "epoch": 2.4, + "learning_rate": 9.950628366247757e-06, + "loss": 0.2763, + "step": 62511 + }, + { + "epoch": 2.4, + "learning_rate": 9.949987176199025e-06, + "loss": 0.1617, + "step": 62512 + }, + { + "epoch": 2.4, + "learning_rate": 9.949345986150295e-06, + "loss": 0.2661, + "step": 62513 + }, + { + "epoch": 2.4, + "learning_rate": 9.948704796101565e-06, + "loss": 0.2118, + "step": 62514 + }, + { + "epoch": 2.4, + "learning_rate": 9.948063606052835e-06, + "loss": 0.2827, + "step": 62515 + }, + { + "epoch": 2.4, + "learning_rate": 9.947422416004104e-06, + "loss": 0.2745, + "step": 62516 + }, + { + "epoch": 2.4, + "learning_rate": 9.946781225955374e-06, + "loss": 0.1652, + "step": 62517 + }, + { + "epoch": 2.4, + "learning_rate": 9.946140035906642e-06, + "loss": 0.2598, + "step": 62518 + }, + { + "epoch": 2.4, + "learning_rate": 9.945498845857914e-06, + "loss": 0.158, + "step": 62519 + }, + { + "epoch": 2.4, + "learning_rate": 9.944857655809182e-06, + "loss": 0.212, + "step": 62520 + }, + { + "epoch": 2.4, + "learning_rate": 9.944216465760452e-06, + "loss": 0.3725, + "step": 62521 + }, + { + "epoch": 2.4, + "learning_rate": 9.94357527571172e-06, + "loss": 0.1372, + "step": 62522 + }, + { + "epoch": 2.4, + "learning_rate": 9.942934085662992e-06, + "loss": 0.3011, + "step": 62523 + }, + { + "epoch": 2.4, + "learning_rate": 9.94229289561426e-06, + "loss": 0.1765, + "step": 62524 + }, + { + "epoch": 2.4, + "learning_rate": 9.94165170556553e-06, + "loss": 0.221, + "step": 62525 + }, + { + "epoch": 2.4, + "learning_rate": 9.941010515516799e-06, + "loss": 0.3269, + "step": 62526 + }, + { + "epoch": 2.4, + "learning_rate": 9.940369325468069e-06, + "loss": 0.2661, + "step": 62527 + }, + { + "epoch": 2.4, + "learning_rate": 9.939728135419339e-06, + "loss": 0.219, + "step": 62528 + }, + { + "epoch": 2.4, + "learning_rate": 9.939086945370609e-06, + "loss": 0.1941, + "step": 62529 + }, + { + "epoch": 2.4, + "learning_rate": 9.938445755321877e-06, + "loss": 0.1751, + "step": 62530 + }, + { + "epoch": 2.4, + "learning_rate": 9.937804565273147e-06, + "loss": 0.1777, + "step": 62531 + }, + { + "epoch": 2.4, + "learning_rate": 9.937163375224417e-06, + "loss": 0.2452, + "step": 62532 + }, + { + "epoch": 2.4, + "learning_rate": 9.936522185175687e-06, + "loss": 0.2021, + "step": 62533 + }, + { + "epoch": 2.4, + "learning_rate": 9.935880995126956e-06, + "loss": 0.3378, + "step": 62534 + }, + { + "epoch": 2.4, + "learning_rate": 9.935239805078226e-06, + "loss": 0.3422, + "step": 62535 + }, + { + "epoch": 2.4, + "learning_rate": 9.934598615029496e-06, + "loss": 0.2141, + "step": 62536 + }, + { + "epoch": 2.4, + "learning_rate": 9.933957424980766e-06, + "loss": 0.1791, + "step": 62537 + }, + { + "epoch": 2.4, + "learning_rate": 9.933316234932034e-06, + "loss": 0.1334, + "step": 62538 + }, + { + "epoch": 2.4, + "learning_rate": 9.932675044883304e-06, + "loss": 0.2265, + "step": 62539 + }, + { + "epoch": 2.4, + "learning_rate": 9.932033854834572e-06, + "loss": 0.1011, + "step": 62540 + }, + { + "epoch": 2.4, + "learning_rate": 9.931392664785844e-06, + "loss": 0.1644, + "step": 62541 + }, + { + "epoch": 2.4, + "learning_rate": 9.930751474737112e-06, + "loss": 0.3163, + "step": 62542 + }, + { + "epoch": 2.4, + "learning_rate": 9.930110284688382e-06, + "loss": 0.1676, + "step": 62543 + }, + { + "epoch": 2.4, + "learning_rate": 9.92946909463965e-06, + "loss": 0.3228, + "step": 62544 + }, + { + "epoch": 2.4, + "learning_rate": 9.928827904590922e-06, + "loss": 0.2771, + "step": 62545 + }, + { + "epoch": 2.4, + "learning_rate": 9.92818671454219e-06, + "loss": 0.2126, + "step": 62546 + }, + { + "epoch": 2.4, + "learning_rate": 9.927545524493461e-06, + "loss": 0.1185, + "step": 62547 + }, + { + "epoch": 2.4, + "learning_rate": 9.92690433444473e-06, + "loss": 0.2147, + "step": 62548 + }, + { + "epoch": 2.4, + "learning_rate": 9.926263144396e-06, + "loss": 0.2136, + "step": 62549 + }, + { + "epoch": 2.4, + "learning_rate": 9.92562195434727e-06, + "loss": 0.262, + "step": 62550 + }, + { + "epoch": 2.4, + "learning_rate": 9.92498076429854e-06, + "loss": 0.3053, + "step": 62551 + }, + { + "epoch": 2.4, + "learning_rate": 9.924339574249808e-06, + "loss": 0.2273, + "step": 62552 + }, + { + "epoch": 2.4, + "learning_rate": 9.923698384201078e-06, + "loss": 0.242, + "step": 62553 + }, + { + "epoch": 2.4, + "learning_rate": 9.923057194152348e-06, + "loss": 0.1661, + "step": 62554 + }, + { + "epoch": 2.41, + "learning_rate": 9.922416004103618e-06, + "loss": 0.2743, + "step": 62555 + }, + { + "epoch": 2.41, + "learning_rate": 9.921774814054886e-06, + "loss": 0.1992, + "step": 62556 + }, + { + "epoch": 2.41, + "learning_rate": 9.921133624006156e-06, + "loss": 0.2287, + "step": 62557 + }, + { + "epoch": 2.41, + "learning_rate": 9.920492433957424e-06, + "loss": 0.2033, + "step": 62558 + }, + { + "epoch": 2.41, + "learning_rate": 9.919851243908696e-06, + "loss": 0.1567, + "step": 62559 + }, + { + "epoch": 2.41, + "learning_rate": 9.919210053859964e-06, + "loss": 0.2434, + "step": 62560 + }, + { + "epoch": 2.41, + "learning_rate": 9.918568863811234e-06, + "loss": 0.2553, + "step": 62561 + }, + { + "epoch": 2.41, + "learning_rate": 9.917927673762503e-06, + "loss": 0.1889, + "step": 62562 + }, + { + "epoch": 2.41, + "learning_rate": 9.917286483713774e-06, + "loss": 0.2061, + "step": 62563 + }, + { + "epoch": 2.41, + "learning_rate": 9.916645293665043e-06, + "loss": 0.2522, + "step": 62564 + }, + { + "epoch": 2.41, + "learning_rate": 9.916004103616313e-06, + "loss": 0.1639, + "step": 62565 + }, + { + "epoch": 2.41, + "learning_rate": 9.915362913567581e-06, + "loss": 0.3187, + "step": 62566 + }, + { + "epoch": 2.41, + "learning_rate": 9.914721723518851e-06, + "loss": 0.2277, + "step": 62567 + }, + { + "epoch": 2.41, + "learning_rate": 9.914080533470121e-06, + "loss": 0.1412, + "step": 62568 + }, + { + "epoch": 2.41, + "learning_rate": 9.913439343421391e-06, + "loss": 0.176, + "step": 62569 + }, + { + "epoch": 2.41, + "learning_rate": 9.91279815337266e-06, + "loss": 0.1701, + "step": 62570 + }, + { + "epoch": 2.41, + "learning_rate": 9.91215696332393e-06, + "loss": 0.196, + "step": 62571 + }, + { + "epoch": 2.41, + "learning_rate": 9.9115157732752e-06, + "loss": 0.2668, + "step": 62572 + }, + { + "epoch": 2.41, + "learning_rate": 9.91087458322647e-06, + "loss": 0.1775, + "step": 62573 + }, + { + "epoch": 2.41, + "learning_rate": 9.910233393177738e-06, + "loss": 0.2725, + "step": 62574 + }, + { + "epoch": 2.41, + "learning_rate": 9.909592203129008e-06, + "loss": 0.1612, + "step": 62575 + }, + { + "epoch": 2.41, + "learning_rate": 9.908951013080276e-06, + "loss": 0.2978, + "step": 62576 + }, + { + "epoch": 2.41, + "learning_rate": 9.908309823031548e-06, + "loss": 0.1904, + "step": 62577 + }, + { + "epoch": 2.41, + "learning_rate": 9.907668632982816e-06, + "loss": 0.1313, + "step": 62578 + }, + { + "epoch": 2.41, + "learning_rate": 9.907027442934086e-06, + "loss": 0.3091, + "step": 62579 + }, + { + "epoch": 2.41, + "learning_rate": 9.906386252885355e-06, + "loss": 0.1829, + "step": 62580 + }, + { + "epoch": 2.41, + "learning_rate": 9.905745062836626e-06, + "loss": 0.279, + "step": 62581 + }, + { + "epoch": 2.41, + "learning_rate": 9.905103872787895e-06, + "loss": 0.1335, + "step": 62582 + }, + { + "epoch": 2.41, + "learning_rate": 9.904462682739165e-06, + "loss": 0.2636, + "step": 62583 + }, + { + "epoch": 2.41, + "learning_rate": 9.903821492690433e-06, + "loss": 0.1671, + "step": 62584 + }, + { + "epoch": 2.41, + "learning_rate": 9.903180302641705e-06, + "loss": 0.1591, + "step": 62585 + }, + { + "epoch": 2.41, + "learning_rate": 9.902539112592973e-06, + "loss": 0.3553, + "step": 62586 + }, + { + "epoch": 2.41, + "learning_rate": 9.901897922544243e-06, + "loss": 0.1463, + "step": 62587 + }, + { + "epoch": 2.41, + "learning_rate": 9.901256732495511e-06, + "loss": 0.3546, + "step": 62588 + }, + { + "epoch": 2.41, + "learning_rate": 9.900615542446781e-06, + "loss": 0.1856, + "step": 62589 + }, + { + "epoch": 2.41, + "learning_rate": 9.899974352398051e-06, + "loss": 0.2302, + "step": 62590 + }, + { + "epoch": 2.41, + "learning_rate": 9.899333162349322e-06, + "loss": 0.1894, + "step": 62591 + }, + { + "epoch": 2.41, + "learning_rate": 9.89869197230059e-06, + "loss": 0.1839, + "step": 62592 + }, + { + "epoch": 2.41, + "learning_rate": 9.89805078225186e-06, + "loss": 0.1943, + "step": 62593 + }, + { + "epoch": 2.41, + "learning_rate": 9.89740959220313e-06, + "loss": 0.2161, + "step": 62594 + }, + { + "epoch": 2.41, + "learning_rate": 9.8967684021544e-06, + "loss": 0.1838, + "step": 62595 + }, + { + "epoch": 2.41, + "learning_rate": 9.896127212105668e-06, + "loss": 0.2886, + "step": 62596 + }, + { + "epoch": 2.41, + "learning_rate": 9.895486022056938e-06, + "loss": 0.271, + "step": 62597 + }, + { + "epoch": 2.41, + "learning_rate": 9.894844832008207e-06, + "loss": 0.2531, + "step": 62598 + }, + { + "epoch": 2.41, + "learning_rate": 9.894203641959478e-06, + "loss": 0.1887, + "step": 62599 + }, + { + "epoch": 2.41, + "learning_rate": 9.893562451910747e-06, + "loss": 0.128, + "step": 62600 + }, + { + "epoch": 2.41, + "learning_rate": 9.892921261862017e-06, + "loss": 0.1699, + "step": 62601 + }, + { + "epoch": 2.41, + "learning_rate": 9.892280071813285e-06, + "loss": 0.1772, + "step": 62602 + }, + { + "epoch": 2.41, + "learning_rate": 9.891638881764557e-06, + "loss": 0.3111, + "step": 62603 + }, + { + "epoch": 2.41, + "learning_rate": 9.890997691715825e-06, + "loss": 0.191, + "step": 62604 + }, + { + "epoch": 2.41, + "learning_rate": 9.890356501667095e-06, + "loss": 0.1376, + "step": 62605 + }, + { + "epoch": 2.41, + "learning_rate": 9.889715311618363e-06, + "loss": 0.1834, + "step": 62606 + }, + { + "epoch": 2.41, + "learning_rate": 9.889074121569633e-06, + "loss": 0.2185, + "step": 62607 + }, + { + "epoch": 2.41, + "learning_rate": 9.888432931520903e-06, + "loss": 0.1438, + "step": 62608 + }, + { + "epoch": 2.41, + "learning_rate": 9.887791741472173e-06, + "loss": 0.327, + "step": 62609 + }, + { + "epoch": 2.41, + "learning_rate": 9.887150551423442e-06, + "loss": 0.2332, + "step": 62610 + }, + { + "epoch": 2.41, + "learning_rate": 9.886509361374712e-06, + "loss": 0.3294, + "step": 62611 + }, + { + "epoch": 2.41, + "learning_rate": 9.885868171325982e-06, + "loss": 0.1431, + "step": 62612 + }, + { + "epoch": 2.41, + "learning_rate": 9.885226981277252e-06, + "loss": 0.162, + "step": 62613 + }, + { + "epoch": 2.41, + "learning_rate": 9.88458579122852e-06, + "loss": 0.2734, + "step": 62614 + }, + { + "epoch": 2.41, + "learning_rate": 9.88394460117979e-06, + "loss": 0.2027, + "step": 62615 + }, + { + "epoch": 2.41, + "learning_rate": 9.883303411131059e-06, + "loss": 0.3241, + "step": 62616 + }, + { + "epoch": 2.41, + "learning_rate": 9.88266222108233e-06, + "loss": 0.2419, + "step": 62617 + }, + { + "epoch": 2.41, + "learning_rate": 9.882021031033599e-06, + "loss": 0.3191, + "step": 62618 + }, + { + "epoch": 2.41, + "learning_rate": 9.881379840984869e-06, + "loss": 0.3002, + "step": 62619 + }, + { + "epoch": 2.41, + "learning_rate": 9.880738650936137e-06, + "loss": 0.122, + "step": 62620 + }, + { + "epoch": 2.41, + "learning_rate": 9.880097460887409e-06, + "loss": 0.2453, + "step": 62621 + }, + { + "epoch": 2.41, + "learning_rate": 9.879456270838677e-06, + "loss": 0.1975, + "step": 62622 + }, + { + "epoch": 2.41, + "learning_rate": 9.878815080789947e-06, + "loss": 0.1367, + "step": 62623 + }, + { + "epoch": 2.41, + "learning_rate": 9.878173890741215e-06, + "loss": 0.2294, + "step": 62624 + }, + { + "epoch": 2.41, + "learning_rate": 9.877532700692485e-06, + "loss": 0.2516, + "step": 62625 + }, + { + "epoch": 2.41, + "learning_rate": 9.876891510643755e-06, + "loss": 0.2151, + "step": 62626 + }, + { + "epoch": 2.41, + "learning_rate": 9.876250320595025e-06, + "loss": 0.1998, + "step": 62627 + }, + { + "epoch": 2.41, + "learning_rate": 9.875609130546294e-06, + "loss": 0.2319, + "step": 62628 + }, + { + "epoch": 2.41, + "learning_rate": 9.874967940497564e-06, + "loss": 0.3147, + "step": 62629 + }, + { + "epoch": 2.41, + "learning_rate": 9.874326750448834e-06, + "loss": 0.2366, + "step": 62630 + }, + { + "epoch": 2.41, + "learning_rate": 9.873685560400104e-06, + "loss": 0.2507, + "step": 62631 + }, + { + "epoch": 2.41, + "learning_rate": 9.873044370351372e-06, + "loss": 0.1839, + "step": 62632 + }, + { + "epoch": 2.41, + "learning_rate": 9.872403180302642e-06, + "loss": 0.2765, + "step": 62633 + }, + { + "epoch": 2.41, + "learning_rate": 9.871761990253912e-06, + "loss": 0.1086, + "step": 62634 + }, + { + "epoch": 2.41, + "learning_rate": 9.871120800205182e-06, + "loss": 0.1621, + "step": 62635 + }, + { + "epoch": 2.41, + "learning_rate": 9.87047961015645e-06, + "loss": 0.234, + "step": 62636 + }, + { + "epoch": 2.41, + "learning_rate": 9.86983842010772e-06, + "loss": 0.2548, + "step": 62637 + }, + { + "epoch": 2.41, + "learning_rate": 9.869197230058989e-06, + "loss": 0.209, + "step": 62638 + }, + { + "epoch": 2.41, + "learning_rate": 9.86855604001026e-06, + "loss": 0.2981, + "step": 62639 + }, + { + "epoch": 2.41, + "learning_rate": 9.867914849961529e-06, + "loss": 0.5581, + "step": 62640 + }, + { + "epoch": 2.41, + "learning_rate": 9.867273659912799e-06, + "loss": 0.1862, + "step": 62641 + }, + { + "epoch": 2.41, + "learning_rate": 9.866632469864067e-06, + "loss": 0.2992, + "step": 62642 + }, + { + "epoch": 2.41, + "learning_rate": 9.865991279815339e-06, + "loss": 0.2268, + "step": 62643 + }, + { + "epoch": 2.41, + "learning_rate": 9.865350089766607e-06, + "loss": 0.3084, + "step": 62644 + }, + { + "epoch": 2.41, + "learning_rate": 9.864708899717877e-06, + "loss": 0.2189, + "step": 62645 + }, + { + "epoch": 2.41, + "learning_rate": 9.864067709669146e-06, + "loss": 0.1744, + "step": 62646 + }, + { + "epoch": 2.41, + "learning_rate": 9.863426519620416e-06, + "loss": 0.0968, + "step": 62647 + }, + { + "epoch": 2.41, + "learning_rate": 9.862785329571686e-06, + "loss": 0.1718, + "step": 62648 + }, + { + "epoch": 2.41, + "learning_rate": 9.862144139522956e-06, + "loss": 0.2682, + "step": 62649 + }, + { + "epoch": 2.41, + "learning_rate": 9.861502949474224e-06, + "loss": 0.1786, + "step": 62650 + }, + { + "epoch": 2.41, + "learning_rate": 9.860861759425494e-06, + "loss": 0.261, + "step": 62651 + }, + { + "epoch": 2.41, + "learning_rate": 9.860220569376764e-06, + "loss": 0.1954, + "step": 62652 + }, + { + "epoch": 2.41, + "learning_rate": 9.859579379328034e-06, + "loss": 0.3587, + "step": 62653 + }, + { + "epoch": 2.41, + "learning_rate": 9.858938189279302e-06, + "loss": 0.2591, + "step": 62654 + }, + { + "epoch": 2.41, + "learning_rate": 9.858296999230572e-06, + "loss": 0.1712, + "step": 62655 + }, + { + "epoch": 2.41, + "learning_rate": 9.85765580918184e-06, + "loss": 0.2557, + "step": 62656 + }, + { + "epoch": 2.41, + "learning_rate": 9.857014619133113e-06, + "loss": 0.2262, + "step": 62657 + }, + { + "epoch": 2.41, + "learning_rate": 9.85637342908438e-06, + "loss": 0.1542, + "step": 62658 + }, + { + "epoch": 2.41, + "learning_rate": 9.855732239035651e-06, + "loss": 0.3109, + "step": 62659 + }, + { + "epoch": 2.41, + "learning_rate": 9.85509104898692e-06, + "loss": 0.2381, + "step": 62660 + }, + { + "epoch": 2.41, + "learning_rate": 9.854449858938191e-06, + "loss": 0.1749, + "step": 62661 + }, + { + "epoch": 2.41, + "learning_rate": 9.85380866888946e-06, + "loss": 0.3164, + "step": 62662 + }, + { + "epoch": 2.41, + "learning_rate": 9.85316747884073e-06, + "loss": 0.4172, + "step": 62663 + }, + { + "epoch": 2.41, + "learning_rate": 9.852526288791998e-06, + "loss": 0.1699, + "step": 62664 + }, + { + "epoch": 2.41, + "learning_rate": 9.851885098743268e-06, + "loss": 0.3415, + "step": 62665 + }, + { + "epoch": 2.41, + "learning_rate": 9.851243908694538e-06, + "loss": 0.1249, + "step": 62666 + }, + { + "epoch": 2.41, + "learning_rate": 9.850602718645808e-06, + "loss": 0.2529, + "step": 62667 + }, + { + "epoch": 2.41, + "learning_rate": 9.849961528597076e-06, + "loss": 0.1722, + "step": 62668 + }, + { + "epoch": 2.41, + "learning_rate": 9.849320338548346e-06, + "loss": 0.1732, + "step": 62669 + }, + { + "epoch": 2.41, + "learning_rate": 9.848679148499616e-06, + "loss": 0.1414, + "step": 62670 + }, + { + "epoch": 2.41, + "learning_rate": 9.848037958450886e-06, + "loss": 0.2547, + "step": 62671 + }, + { + "epoch": 2.41, + "learning_rate": 9.847396768402154e-06, + "loss": 0.2791, + "step": 62672 + }, + { + "epoch": 2.41, + "learning_rate": 9.846755578353424e-06, + "loss": 0.2281, + "step": 62673 + }, + { + "epoch": 2.41, + "learning_rate": 9.846114388304693e-06, + "loss": 0.1976, + "step": 62674 + }, + { + "epoch": 2.41, + "learning_rate": 9.845473198255964e-06, + "loss": 0.2551, + "step": 62675 + }, + { + "epoch": 2.41, + "learning_rate": 9.844832008207233e-06, + "loss": 0.1828, + "step": 62676 + }, + { + "epoch": 2.41, + "learning_rate": 9.844190818158503e-06, + "loss": 0.26, + "step": 62677 + }, + { + "epoch": 2.41, + "learning_rate": 9.843549628109771e-06, + "loss": 0.1628, + "step": 62678 + }, + { + "epoch": 2.41, + "learning_rate": 9.842908438061043e-06, + "loss": 0.1492, + "step": 62679 + }, + { + "epoch": 2.41, + "learning_rate": 9.842267248012311e-06, + "loss": 0.1843, + "step": 62680 + }, + { + "epoch": 2.41, + "learning_rate": 9.841626057963581e-06, + "loss": 0.2406, + "step": 62681 + }, + { + "epoch": 2.41, + "learning_rate": 9.84098486791485e-06, + "loss": 0.3179, + "step": 62682 + }, + { + "epoch": 2.41, + "learning_rate": 9.840343677866121e-06, + "loss": 0.173, + "step": 62683 + }, + { + "epoch": 2.41, + "learning_rate": 9.83970248781739e-06, + "loss": 0.2612, + "step": 62684 + }, + { + "epoch": 2.41, + "learning_rate": 9.83906129776866e-06, + "loss": 0.2118, + "step": 62685 + }, + { + "epoch": 2.41, + "learning_rate": 9.838420107719928e-06, + "loss": 0.1742, + "step": 62686 + }, + { + "epoch": 2.41, + "learning_rate": 9.837778917671198e-06, + "loss": 0.2765, + "step": 62687 + }, + { + "epoch": 2.41, + "learning_rate": 9.837137727622468e-06, + "loss": 0.2491, + "step": 62688 + }, + { + "epoch": 2.41, + "learning_rate": 9.836496537573738e-06, + "loss": 0.1128, + "step": 62689 + }, + { + "epoch": 2.41, + "learning_rate": 9.835855347525006e-06, + "loss": 0.233, + "step": 62690 + }, + { + "epoch": 2.41, + "learning_rate": 9.835214157476276e-06, + "loss": 0.3408, + "step": 62691 + }, + { + "epoch": 2.41, + "learning_rate": 9.834572967427546e-06, + "loss": 0.3694, + "step": 62692 + }, + { + "epoch": 2.41, + "learning_rate": 9.833931777378816e-06, + "loss": 0.2006, + "step": 62693 + }, + { + "epoch": 2.41, + "learning_rate": 9.833290587330085e-06, + "loss": 0.1093, + "step": 62694 + }, + { + "epoch": 2.41, + "learning_rate": 9.832649397281355e-06, + "loss": 0.1291, + "step": 62695 + }, + { + "epoch": 2.41, + "learning_rate": 9.832008207232623e-06, + "loss": 0.2593, + "step": 62696 + }, + { + "epoch": 2.41, + "learning_rate": 9.831367017183895e-06, + "loss": 0.2613, + "step": 62697 + }, + { + "epoch": 2.41, + "learning_rate": 9.830725827135163e-06, + "loss": 0.4551, + "step": 62698 + }, + { + "epoch": 2.41, + "learning_rate": 9.830084637086433e-06, + "loss": 0.1656, + "step": 62699 + }, + { + "epoch": 2.41, + "learning_rate": 9.829443447037701e-06, + "loss": 0.2063, + "step": 62700 + }, + { + "epoch": 2.41, + "learning_rate": 9.828802256988973e-06, + "loss": 0.2507, + "step": 62701 + }, + { + "epoch": 2.41, + "learning_rate": 9.828161066940242e-06, + "loss": 0.1084, + "step": 62702 + }, + { + "epoch": 2.41, + "learning_rate": 9.827519876891512e-06, + "loss": 0.1753, + "step": 62703 + }, + { + "epoch": 2.41, + "learning_rate": 9.82687868684278e-06, + "loss": 0.1417, + "step": 62704 + }, + { + "epoch": 2.41, + "learning_rate": 9.82623749679405e-06, + "loss": 0.2528, + "step": 62705 + }, + { + "epoch": 2.41, + "learning_rate": 9.82559630674532e-06, + "loss": 0.1687, + "step": 62706 + }, + { + "epoch": 2.41, + "learning_rate": 9.82495511669659e-06, + "loss": 0.2238, + "step": 62707 + }, + { + "epoch": 2.41, + "learning_rate": 9.824313926647858e-06, + "loss": 0.2133, + "step": 62708 + }, + { + "epoch": 2.41, + "learning_rate": 9.823672736599128e-06, + "loss": 0.2449, + "step": 62709 + }, + { + "epoch": 2.41, + "learning_rate": 9.823031546550398e-06, + "loss": 0.3056, + "step": 62710 + }, + { + "epoch": 2.41, + "learning_rate": 9.822390356501668e-06, + "loss": 0.1604, + "step": 62711 + }, + { + "epoch": 2.41, + "learning_rate": 9.821749166452937e-06, + "loss": 0.1649, + "step": 62712 + }, + { + "epoch": 2.41, + "learning_rate": 9.821107976404207e-06, + "loss": 0.2536, + "step": 62713 + }, + { + "epoch": 2.41, + "learning_rate": 9.820466786355475e-06, + "loss": 0.3311, + "step": 62714 + }, + { + "epoch": 2.41, + "learning_rate": 9.819825596306747e-06, + "loss": 0.1875, + "step": 62715 + }, + { + "epoch": 2.41, + "learning_rate": 9.819184406258015e-06, + "loss": 0.1955, + "step": 62716 + }, + { + "epoch": 2.41, + "learning_rate": 9.818543216209285e-06, + "loss": 0.236, + "step": 62717 + }, + { + "epoch": 2.41, + "learning_rate": 9.817902026160553e-06, + "loss": 0.2375, + "step": 62718 + }, + { + "epoch": 2.41, + "learning_rate": 9.817260836111825e-06, + "loss": 0.2149, + "step": 62719 + }, + { + "epoch": 2.41, + "learning_rate": 9.816619646063093e-06, + "loss": 0.2712, + "step": 62720 + }, + { + "epoch": 2.41, + "learning_rate": 9.815978456014363e-06, + "loss": 0.2824, + "step": 62721 + }, + { + "epoch": 2.41, + "learning_rate": 9.815337265965632e-06, + "loss": 0.1317, + "step": 62722 + }, + { + "epoch": 2.41, + "learning_rate": 9.814696075916904e-06, + "loss": 0.1899, + "step": 62723 + }, + { + "epoch": 2.41, + "learning_rate": 9.814054885868172e-06, + "loss": 0.1953, + "step": 62724 + }, + { + "epoch": 2.41, + "learning_rate": 9.813413695819442e-06, + "loss": 0.2592, + "step": 62725 + }, + { + "epoch": 2.41, + "learning_rate": 9.81277250577071e-06, + "loss": 0.3137, + "step": 62726 + }, + { + "epoch": 2.41, + "learning_rate": 9.81213131572198e-06, + "loss": 0.1295, + "step": 62727 + }, + { + "epoch": 2.41, + "learning_rate": 9.81149012567325e-06, + "loss": 0.1923, + "step": 62728 + }, + { + "epoch": 2.41, + "learning_rate": 9.81084893562452e-06, + "loss": 0.1046, + "step": 62729 + }, + { + "epoch": 2.41, + "learning_rate": 9.810207745575789e-06, + "loss": 0.2746, + "step": 62730 + }, + { + "epoch": 2.41, + "learning_rate": 9.809566555527059e-06, + "loss": 0.2275, + "step": 62731 + }, + { + "epoch": 2.41, + "learning_rate": 9.808925365478329e-06, + "loss": 0.1143, + "step": 62732 + }, + { + "epoch": 2.41, + "learning_rate": 9.808284175429599e-06, + "loss": 0.1644, + "step": 62733 + }, + { + "epoch": 2.41, + "learning_rate": 9.807642985380867e-06, + "loss": 0.2439, + "step": 62734 + }, + { + "epoch": 2.41, + "learning_rate": 9.807001795332137e-06, + "loss": 0.1233, + "step": 62735 + }, + { + "epoch": 2.41, + "learning_rate": 9.806360605283405e-06, + "loss": 0.2357, + "step": 62736 + }, + { + "epoch": 2.41, + "learning_rate": 9.805719415234677e-06, + "loss": 0.1107, + "step": 62737 + }, + { + "epoch": 2.41, + "learning_rate": 9.805078225185945e-06, + "loss": 0.2187, + "step": 62738 + }, + { + "epoch": 2.41, + "learning_rate": 9.804437035137215e-06, + "loss": 0.1649, + "step": 62739 + }, + { + "epoch": 2.41, + "learning_rate": 9.803795845088484e-06, + "loss": 0.1545, + "step": 62740 + }, + { + "epoch": 2.41, + "learning_rate": 9.803154655039755e-06, + "loss": 0.1376, + "step": 62741 + }, + { + "epoch": 2.41, + "learning_rate": 9.802513464991024e-06, + "loss": 0.176, + "step": 62742 + }, + { + "epoch": 2.41, + "learning_rate": 9.801872274942294e-06, + "loss": 0.3051, + "step": 62743 + }, + { + "epoch": 2.41, + "learning_rate": 9.801231084893562e-06, + "loss": 0.3322, + "step": 62744 + }, + { + "epoch": 2.41, + "learning_rate": 9.800589894844832e-06, + "loss": 0.2045, + "step": 62745 + }, + { + "epoch": 2.41, + "learning_rate": 9.799948704796102e-06, + "loss": 0.208, + "step": 62746 + }, + { + "epoch": 2.41, + "learning_rate": 9.799307514747372e-06, + "loss": 0.209, + "step": 62747 + }, + { + "epoch": 2.41, + "learning_rate": 9.79866632469864e-06, + "loss": 0.2272, + "step": 62748 + }, + { + "epoch": 2.41, + "learning_rate": 9.79802513464991e-06, + "loss": 0.1576, + "step": 62749 + }, + { + "epoch": 2.41, + "learning_rate": 9.79738394460118e-06, + "loss": 0.3769, + "step": 62750 + }, + { + "epoch": 2.41, + "learning_rate": 9.79674275455245e-06, + "loss": 0.143, + "step": 62751 + }, + { + "epoch": 2.41, + "learning_rate": 9.796101564503719e-06, + "loss": 0.1886, + "step": 62752 + }, + { + "epoch": 2.41, + "learning_rate": 9.795460374454989e-06, + "loss": 0.2574, + "step": 62753 + }, + { + "epoch": 2.41, + "learning_rate": 9.794819184406257e-06, + "loss": 0.1545, + "step": 62754 + }, + { + "epoch": 2.41, + "learning_rate": 9.794177994357529e-06, + "loss": 0.2145, + "step": 62755 + }, + { + "epoch": 2.41, + "learning_rate": 9.793536804308797e-06, + "loss": 0.2248, + "step": 62756 + }, + { + "epoch": 2.41, + "learning_rate": 9.792895614260067e-06, + "loss": 0.2223, + "step": 62757 + }, + { + "epoch": 2.41, + "learning_rate": 9.792254424211336e-06, + "loss": 0.2079, + "step": 62758 + }, + { + "epoch": 2.41, + "learning_rate": 9.791613234162607e-06, + "loss": 0.2006, + "step": 62759 + }, + { + "epoch": 2.41, + "learning_rate": 9.790972044113876e-06, + "loss": 0.1162, + "step": 62760 + }, + { + "epoch": 2.41, + "learning_rate": 9.790330854065146e-06, + "loss": 0.292, + "step": 62761 + }, + { + "epoch": 2.41, + "learning_rate": 9.789689664016414e-06, + "loss": 0.2589, + "step": 62762 + }, + { + "epoch": 2.41, + "learning_rate": 9.789048473967684e-06, + "loss": 0.1817, + "step": 62763 + }, + { + "epoch": 2.41, + "learning_rate": 9.788407283918954e-06, + "loss": 0.1951, + "step": 62764 + }, + { + "epoch": 2.41, + "learning_rate": 9.787766093870224e-06, + "loss": 0.2169, + "step": 62765 + }, + { + "epoch": 2.41, + "learning_rate": 9.787124903821492e-06, + "loss": 0.2491, + "step": 62766 + }, + { + "epoch": 2.41, + "learning_rate": 9.786483713772762e-06, + "loss": 0.3584, + "step": 62767 + }, + { + "epoch": 2.41, + "learning_rate": 9.785842523724032e-06, + "loss": 0.2965, + "step": 62768 + }, + { + "epoch": 2.41, + "learning_rate": 9.785201333675303e-06, + "loss": 0.2272, + "step": 62769 + }, + { + "epoch": 2.41, + "learning_rate": 9.784560143626571e-06, + "loss": 0.1257, + "step": 62770 + }, + { + "epoch": 2.41, + "learning_rate": 9.783918953577841e-06, + "loss": 0.171, + "step": 62771 + }, + { + "epoch": 2.41, + "learning_rate": 9.783277763529111e-06, + "loss": 0.2088, + "step": 62772 + }, + { + "epoch": 2.41, + "learning_rate": 9.782636573480381e-06, + "loss": 0.2238, + "step": 62773 + }, + { + "epoch": 2.41, + "learning_rate": 9.78199538343165e-06, + "loss": 0.1336, + "step": 62774 + }, + { + "epoch": 2.41, + "learning_rate": 9.78135419338292e-06, + "loss": 0.1551, + "step": 62775 + }, + { + "epoch": 2.41, + "learning_rate": 9.780713003334188e-06, + "loss": 0.2385, + "step": 62776 + }, + { + "epoch": 2.41, + "learning_rate": 9.78007181328546e-06, + "loss": 0.2896, + "step": 62777 + }, + { + "epoch": 2.41, + "learning_rate": 9.779430623236728e-06, + "loss": 0.2499, + "step": 62778 + }, + { + "epoch": 2.41, + "learning_rate": 9.778789433187998e-06, + "loss": 0.1234, + "step": 62779 + }, + { + "epoch": 2.41, + "learning_rate": 9.778148243139266e-06, + "loss": 0.0674, + "step": 62780 + }, + { + "epoch": 2.41, + "learning_rate": 9.777507053090538e-06, + "loss": 0.31, + "step": 62781 + }, + { + "epoch": 2.41, + "learning_rate": 9.776865863041806e-06, + "loss": 0.2942, + "step": 62782 + }, + { + "epoch": 2.41, + "learning_rate": 9.776224672993076e-06, + "loss": 0.2085, + "step": 62783 + }, + { + "epoch": 2.41, + "learning_rate": 9.775583482944344e-06, + "loss": 0.1571, + "step": 62784 + }, + { + "epoch": 2.41, + "learning_rate": 9.774942292895614e-06, + "loss": 0.1268, + "step": 62785 + }, + { + "epoch": 2.41, + "learning_rate": 9.774301102846884e-06, + "loss": 0.3087, + "step": 62786 + }, + { + "epoch": 2.41, + "learning_rate": 9.773659912798154e-06, + "loss": 0.2534, + "step": 62787 + }, + { + "epoch": 2.41, + "learning_rate": 9.773018722749423e-06, + "loss": 0.1585, + "step": 62788 + }, + { + "epoch": 2.41, + "learning_rate": 9.772377532700693e-06, + "loss": 0.3269, + "step": 62789 + }, + { + "epoch": 2.41, + "learning_rate": 9.771736342651963e-06, + "loss": 0.1426, + "step": 62790 + }, + { + "epoch": 2.41, + "learning_rate": 9.771095152603233e-06, + "loss": 0.2283, + "step": 62791 + }, + { + "epoch": 2.41, + "learning_rate": 9.770453962554501e-06, + "loss": 0.1601, + "step": 62792 + }, + { + "epoch": 2.41, + "learning_rate": 9.769812772505771e-06, + "loss": 0.3659, + "step": 62793 + }, + { + "epoch": 2.41, + "learning_rate": 9.76917158245704e-06, + "loss": 0.2987, + "step": 62794 + }, + { + "epoch": 2.41, + "learning_rate": 9.768530392408311e-06, + "loss": 0.1902, + "step": 62795 + }, + { + "epoch": 2.41, + "learning_rate": 9.76788920235958e-06, + "loss": 0.3281, + "step": 62796 + }, + { + "epoch": 2.41, + "learning_rate": 9.76724801231085e-06, + "loss": 0.1404, + "step": 62797 + }, + { + "epoch": 2.41, + "learning_rate": 9.766606822262118e-06, + "loss": 0.1357, + "step": 62798 + }, + { + "epoch": 2.41, + "learning_rate": 9.76596563221339e-06, + "loss": 0.1453, + "step": 62799 + }, + { + "epoch": 2.41, + "learning_rate": 9.765324442164658e-06, + "loss": 0.337, + "step": 62800 + }, + { + "epoch": 2.41, + "learning_rate": 9.764683252115928e-06, + "loss": 0.2536, + "step": 62801 + }, + { + "epoch": 2.41, + "learning_rate": 9.764042062067196e-06, + "loss": 0.1455, + "step": 62802 + }, + { + "epoch": 2.41, + "learning_rate": 9.763400872018466e-06, + "loss": 0.3062, + "step": 62803 + }, + { + "epoch": 2.41, + "learning_rate": 9.762759681969736e-06, + "loss": 0.1823, + "step": 62804 + }, + { + "epoch": 2.41, + "learning_rate": 9.762118491921006e-06, + "loss": 0.2866, + "step": 62805 + }, + { + "epoch": 2.41, + "learning_rate": 9.761477301872275e-06, + "loss": 0.2299, + "step": 62806 + }, + { + "epoch": 2.41, + "learning_rate": 9.760836111823545e-06, + "loss": 0.1758, + "step": 62807 + }, + { + "epoch": 2.41, + "learning_rate": 9.760194921774815e-06, + "loss": 0.2655, + "step": 62808 + }, + { + "epoch": 2.41, + "learning_rate": 9.759553731726085e-06, + "loss": 0.172, + "step": 62809 + }, + { + "epoch": 2.41, + "learning_rate": 9.758912541677353e-06, + "loss": 0.2275, + "step": 62810 + }, + { + "epoch": 2.41, + "learning_rate": 9.758271351628623e-06, + "loss": 0.16, + "step": 62811 + }, + { + "epoch": 2.41, + "learning_rate": 9.757630161579891e-06, + "loss": 0.1976, + "step": 62812 + }, + { + "epoch": 2.41, + "learning_rate": 9.756988971531163e-06, + "loss": 0.131, + "step": 62813 + }, + { + "epoch": 2.41, + "learning_rate": 9.756347781482432e-06, + "loss": 0.1801, + "step": 62814 + }, + { + "epoch": 2.42, + "learning_rate": 9.755706591433702e-06, + "loss": 0.1801, + "step": 62815 + }, + { + "epoch": 2.42, + "learning_rate": 9.75506540138497e-06, + "loss": 0.235, + "step": 62816 + }, + { + "epoch": 2.42, + "learning_rate": 9.754424211336242e-06, + "loss": 0.1738, + "step": 62817 + }, + { + "epoch": 2.42, + "learning_rate": 9.75378302128751e-06, + "loss": 0.333, + "step": 62818 + }, + { + "epoch": 2.42, + "learning_rate": 9.75314183123878e-06, + "loss": 0.2263, + "step": 62819 + }, + { + "epoch": 2.42, + "learning_rate": 9.752500641190048e-06, + "loss": 0.2267, + "step": 62820 + }, + { + "epoch": 2.42, + "learning_rate": 9.75185945114132e-06, + "loss": 0.2473, + "step": 62821 + }, + { + "epoch": 2.42, + "learning_rate": 9.751218261092588e-06, + "loss": 0.3147, + "step": 62822 + }, + { + "epoch": 2.42, + "learning_rate": 9.750577071043858e-06, + "loss": 0.1693, + "step": 62823 + }, + { + "epoch": 2.42, + "learning_rate": 9.749935880995127e-06, + "loss": 0.1865, + "step": 62824 + }, + { + "epoch": 2.42, + "learning_rate": 9.749294690946397e-06, + "loss": 0.2699, + "step": 62825 + }, + { + "epoch": 2.42, + "learning_rate": 9.748653500897667e-06, + "loss": 0.3517, + "step": 62826 + }, + { + "epoch": 2.42, + "learning_rate": 9.748012310848937e-06, + "loss": 0.2311, + "step": 62827 + }, + { + "epoch": 2.42, + "learning_rate": 9.747371120800205e-06, + "loss": 0.1803, + "step": 62828 + }, + { + "epoch": 2.42, + "learning_rate": 9.746729930751475e-06, + "loss": 0.0988, + "step": 62829 + }, + { + "epoch": 2.42, + "learning_rate": 9.746088740702745e-06, + "loss": 0.1947, + "step": 62830 + }, + { + "epoch": 2.42, + "learning_rate": 9.745447550654015e-06, + "loss": 0.3718, + "step": 62831 + }, + { + "epoch": 2.42, + "learning_rate": 9.744806360605283e-06, + "loss": 0.1549, + "step": 62832 + }, + { + "epoch": 2.42, + "learning_rate": 9.744165170556553e-06, + "loss": 0.3686, + "step": 62833 + }, + { + "epoch": 2.42, + "learning_rate": 9.743523980507822e-06, + "loss": 0.2421, + "step": 62834 + }, + { + "epoch": 2.42, + "learning_rate": 9.742882790459094e-06, + "loss": 0.1404, + "step": 62835 + }, + { + "epoch": 2.42, + "learning_rate": 9.742241600410362e-06, + "loss": 0.17, + "step": 62836 + }, + { + "epoch": 2.42, + "learning_rate": 9.741600410361632e-06, + "loss": 0.3682, + "step": 62837 + }, + { + "epoch": 2.42, + "learning_rate": 9.7409592203129e-06, + "loss": 0.2191, + "step": 62838 + }, + { + "epoch": 2.42, + "learning_rate": 9.740318030264172e-06, + "loss": 0.2349, + "step": 62839 + }, + { + "epoch": 2.42, + "learning_rate": 9.73967684021544e-06, + "loss": 0.2928, + "step": 62840 + }, + { + "epoch": 2.42, + "learning_rate": 9.73903565016671e-06, + "loss": 0.2717, + "step": 62841 + }, + { + "epoch": 2.42, + "learning_rate": 9.738394460117979e-06, + "loss": 0.1703, + "step": 62842 + }, + { + "epoch": 2.42, + "learning_rate": 9.737753270069249e-06, + "loss": 0.2345, + "step": 62843 + }, + { + "epoch": 2.42, + "learning_rate": 9.737112080020519e-06, + "loss": 0.3091, + "step": 62844 + }, + { + "epoch": 2.42, + "learning_rate": 9.736470889971789e-06, + "loss": 0.1587, + "step": 62845 + }, + { + "epoch": 2.42, + "learning_rate": 9.735829699923057e-06, + "loss": 0.059, + "step": 62846 + }, + { + "epoch": 2.42, + "learning_rate": 9.735188509874327e-06, + "loss": 0.2507, + "step": 62847 + }, + { + "epoch": 2.42, + "learning_rate": 9.734547319825597e-06, + "loss": 0.1634, + "step": 62848 + }, + { + "epoch": 2.42, + "learning_rate": 9.733906129776867e-06, + "loss": 0.2098, + "step": 62849 + }, + { + "epoch": 2.42, + "learning_rate": 9.733264939728135e-06, + "loss": 0.3093, + "step": 62850 + }, + { + "epoch": 2.42, + "learning_rate": 9.732623749679405e-06, + "loss": 0.1784, + "step": 62851 + }, + { + "epoch": 2.42, + "learning_rate": 9.731982559630674e-06, + "loss": 0.2381, + "step": 62852 + }, + { + "epoch": 2.42, + "learning_rate": 9.731341369581945e-06, + "loss": 0.1891, + "step": 62853 + }, + { + "epoch": 2.42, + "learning_rate": 9.730700179533214e-06, + "loss": 0.224, + "step": 62854 + }, + { + "epoch": 2.42, + "learning_rate": 9.730058989484484e-06, + "loss": 0.1472, + "step": 62855 + }, + { + "epoch": 2.42, + "learning_rate": 9.729417799435752e-06, + "loss": 0.2975, + "step": 62856 + }, + { + "epoch": 2.42, + "learning_rate": 9.728776609387024e-06, + "loss": 0.2763, + "step": 62857 + }, + { + "epoch": 2.42, + "learning_rate": 9.728135419338292e-06, + "loss": 0.1935, + "step": 62858 + }, + { + "epoch": 2.42, + "learning_rate": 9.727494229289562e-06, + "loss": 0.215, + "step": 62859 + }, + { + "epoch": 2.42, + "learning_rate": 9.72685303924083e-06, + "loss": 0.2855, + "step": 62860 + }, + { + "epoch": 2.42, + "learning_rate": 9.7262118491921e-06, + "loss": 0.2162, + "step": 62861 + }, + { + "epoch": 2.42, + "learning_rate": 9.72557065914337e-06, + "loss": 0.2646, + "step": 62862 + }, + { + "epoch": 2.42, + "learning_rate": 9.72492946909464e-06, + "loss": 0.164, + "step": 62863 + }, + { + "epoch": 2.42, + "learning_rate": 9.724288279045909e-06, + "loss": 0.2349, + "step": 62864 + }, + { + "epoch": 2.42, + "learning_rate": 9.723647088997179e-06, + "loss": 0.1944, + "step": 62865 + }, + { + "epoch": 2.42, + "learning_rate": 9.723005898948449e-06, + "loss": 0.3876, + "step": 62866 + }, + { + "epoch": 2.42, + "learning_rate": 9.722364708899719e-06, + "loss": 0.2559, + "step": 62867 + }, + { + "epoch": 2.42, + "learning_rate": 9.721723518850987e-06, + "loss": 0.1558, + "step": 62868 + }, + { + "epoch": 2.42, + "learning_rate": 9.721082328802257e-06, + "loss": 0.3249, + "step": 62869 + }, + { + "epoch": 2.42, + "learning_rate": 9.720441138753527e-06, + "loss": 0.407, + "step": 62870 + }, + { + "epoch": 2.42, + "learning_rate": 9.719799948704797e-06, + "loss": 0.2489, + "step": 62871 + }, + { + "epoch": 2.42, + "learning_rate": 9.719158758656066e-06, + "loss": 0.1689, + "step": 62872 + }, + { + "epoch": 2.42, + "learning_rate": 9.718517568607336e-06, + "loss": 0.1559, + "step": 62873 + }, + { + "epoch": 2.42, + "learning_rate": 9.717876378558604e-06, + "loss": 0.0825, + "step": 62874 + }, + { + "epoch": 2.42, + "learning_rate": 9.717235188509876e-06, + "loss": 0.228, + "step": 62875 + }, + { + "epoch": 2.42, + "learning_rate": 9.716593998461144e-06, + "loss": 0.3455, + "step": 62876 + }, + { + "epoch": 2.42, + "learning_rate": 9.715952808412414e-06, + "loss": 0.1278, + "step": 62877 + }, + { + "epoch": 2.42, + "learning_rate": 9.715311618363682e-06, + "loss": 0.2165, + "step": 62878 + }, + { + "epoch": 2.42, + "learning_rate": 9.714670428314954e-06, + "loss": 0.1496, + "step": 62879 + }, + { + "epoch": 2.42, + "learning_rate": 9.714029238266223e-06, + "loss": 0.3105, + "step": 62880 + }, + { + "epoch": 2.42, + "learning_rate": 9.713388048217493e-06, + "loss": 0.1296, + "step": 62881 + }, + { + "epoch": 2.42, + "learning_rate": 9.712746858168761e-06, + "loss": 0.4155, + "step": 62882 + }, + { + "epoch": 2.42, + "learning_rate": 9.712105668120031e-06, + "loss": 0.2271, + "step": 62883 + }, + { + "epoch": 2.42, + "learning_rate": 9.711464478071301e-06, + "loss": 0.1021, + "step": 62884 + }, + { + "epoch": 2.42, + "learning_rate": 9.710823288022571e-06, + "loss": 0.3328, + "step": 62885 + }, + { + "epoch": 2.42, + "learning_rate": 9.71018209797384e-06, + "loss": 0.4443, + "step": 62886 + }, + { + "epoch": 2.42, + "learning_rate": 9.70954090792511e-06, + "loss": 0.1572, + "step": 62887 + }, + { + "epoch": 2.42, + "learning_rate": 9.70889971787638e-06, + "loss": 0.1552, + "step": 62888 + }, + { + "epoch": 2.42, + "learning_rate": 9.70825852782765e-06, + "loss": 0.3744, + "step": 62889 + }, + { + "epoch": 2.42, + "learning_rate": 9.707617337778918e-06, + "loss": 0.3137, + "step": 62890 + }, + { + "epoch": 2.42, + "learning_rate": 9.706976147730188e-06, + "loss": 0.2874, + "step": 62891 + }, + { + "epoch": 2.42, + "learning_rate": 9.706334957681456e-06, + "loss": 0.2413, + "step": 62892 + }, + { + "epoch": 2.42, + "learning_rate": 9.705693767632728e-06, + "loss": 0.1931, + "step": 62893 + }, + { + "epoch": 2.42, + "learning_rate": 9.705052577583996e-06, + "loss": 0.18, + "step": 62894 + }, + { + "epoch": 2.42, + "learning_rate": 9.704411387535266e-06, + "loss": 0.2042, + "step": 62895 + }, + { + "epoch": 2.42, + "learning_rate": 9.703770197486534e-06, + "loss": 0.2332, + "step": 62896 + }, + { + "epoch": 2.42, + "learning_rate": 9.703129007437806e-06, + "loss": 0.1934, + "step": 62897 + }, + { + "epoch": 2.42, + "learning_rate": 9.702487817389074e-06, + "loss": 0.3969, + "step": 62898 + }, + { + "epoch": 2.42, + "learning_rate": 9.701846627340344e-06, + "loss": 0.339, + "step": 62899 + }, + { + "epoch": 2.42, + "learning_rate": 9.701205437291613e-06, + "loss": 0.3195, + "step": 62900 + }, + { + "epoch": 2.42, + "learning_rate": 9.700564247242883e-06, + "loss": 0.2132, + "step": 62901 + }, + { + "epoch": 2.42, + "learning_rate": 9.699923057194153e-06, + "loss": 0.2576, + "step": 62902 + }, + { + "epoch": 2.42, + "learning_rate": 9.699281867145423e-06, + "loss": 0.2558, + "step": 62903 + }, + { + "epoch": 2.42, + "learning_rate": 9.698640677096691e-06, + "loss": 0.1946, + "step": 62904 + }, + { + "epoch": 2.42, + "learning_rate": 9.697999487047961e-06, + "loss": 0.2177, + "step": 62905 + }, + { + "epoch": 2.42, + "learning_rate": 9.697358296999231e-06, + "loss": 0.3926, + "step": 62906 + }, + { + "epoch": 2.42, + "learning_rate": 9.696717106950501e-06, + "loss": 0.2182, + "step": 62907 + }, + { + "epoch": 2.42, + "learning_rate": 9.69607591690177e-06, + "loss": 0.1427, + "step": 62908 + }, + { + "epoch": 2.42, + "learning_rate": 9.69543472685304e-06, + "loss": 0.2661, + "step": 62909 + }, + { + "epoch": 2.42, + "learning_rate": 9.694793536804308e-06, + "loss": 0.2274, + "step": 62910 + }, + { + "epoch": 2.42, + "learning_rate": 9.69415234675558e-06, + "loss": 0.143, + "step": 62911 + }, + { + "epoch": 2.42, + "learning_rate": 9.693511156706848e-06, + "loss": 0.1792, + "step": 62912 + }, + { + "epoch": 2.42, + "learning_rate": 9.692869966658118e-06, + "loss": 0.2597, + "step": 62913 + }, + { + "epoch": 2.42, + "learning_rate": 9.692228776609386e-06, + "loss": 0.2225, + "step": 62914 + }, + { + "epoch": 2.42, + "learning_rate": 9.691587586560658e-06, + "loss": 0.2251, + "step": 62915 + }, + { + "epoch": 2.42, + "learning_rate": 9.690946396511926e-06, + "loss": 0.2965, + "step": 62916 + }, + { + "epoch": 2.42, + "learning_rate": 9.690305206463196e-06, + "loss": 0.1846, + "step": 62917 + }, + { + "epoch": 2.42, + "learning_rate": 9.689664016414465e-06, + "loss": 0.1743, + "step": 62918 + }, + { + "epoch": 2.42, + "learning_rate": 9.689022826365736e-06, + "loss": 0.2577, + "step": 62919 + }, + { + "epoch": 2.42, + "learning_rate": 9.688381636317005e-06, + "loss": 0.1783, + "step": 62920 + }, + { + "epoch": 2.42, + "learning_rate": 9.687740446268275e-06, + "loss": 0.1419, + "step": 62921 + }, + { + "epoch": 2.42, + "learning_rate": 9.687099256219543e-06, + "loss": 0.21, + "step": 62922 + }, + { + "epoch": 2.42, + "learning_rate": 9.686458066170813e-06, + "loss": 0.2108, + "step": 62923 + }, + { + "epoch": 2.42, + "learning_rate": 9.685816876122083e-06, + "loss": 0.2484, + "step": 62924 + }, + { + "epoch": 2.42, + "learning_rate": 9.685175686073353e-06, + "loss": 0.1783, + "step": 62925 + }, + { + "epoch": 2.42, + "learning_rate": 9.684534496024622e-06, + "loss": 0.2427, + "step": 62926 + }, + { + "epoch": 2.42, + "learning_rate": 9.683893305975892e-06, + "loss": 0.1885, + "step": 62927 + }, + { + "epoch": 2.42, + "learning_rate": 9.683252115927162e-06, + "loss": 0.2784, + "step": 62928 + }, + { + "epoch": 2.42, + "learning_rate": 9.682610925878432e-06, + "loss": 0.1538, + "step": 62929 + }, + { + "epoch": 2.42, + "learning_rate": 9.6819697358297e-06, + "loss": 0.2307, + "step": 62930 + }, + { + "epoch": 2.42, + "learning_rate": 9.68132854578097e-06, + "loss": 0.228, + "step": 62931 + }, + { + "epoch": 2.42, + "learning_rate": 9.680687355732238e-06, + "loss": 0.2481, + "step": 62932 + }, + { + "epoch": 2.42, + "learning_rate": 9.68004616568351e-06, + "loss": 0.1951, + "step": 62933 + }, + { + "epoch": 2.42, + "learning_rate": 9.679404975634778e-06, + "loss": 0.2235, + "step": 62934 + }, + { + "epoch": 2.42, + "learning_rate": 9.678763785586048e-06, + "loss": 0.2046, + "step": 62935 + }, + { + "epoch": 2.42, + "learning_rate": 9.678122595537317e-06, + "loss": 0.1468, + "step": 62936 + }, + { + "epoch": 2.42, + "learning_rate": 9.677481405488588e-06, + "loss": 0.1428, + "step": 62937 + }, + { + "epoch": 2.42, + "learning_rate": 9.676840215439857e-06, + "loss": 0.288, + "step": 62938 + }, + { + "epoch": 2.42, + "learning_rate": 9.676199025391127e-06, + "loss": 0.1951, + "step": 62939 + }, + { + "epoch": 2.42, + "learning_rate": 9.675557835342395e-06, + "loss": 0.3686, + "step": 62940 + }, + { + "epoch": 2.42, + "learning_rate": 9.674916645293665e-06, + "loss": 0.2192, + "step": 62941 + }, + { + "epoch": 2.42, + "learning_rate": 9.674275455244935e-06, + "loss": 0.2177, + "step": 62942 + }, + { + "epoch": 2.42, + "learning_rate": 9.673634265196205e-06, + "loss": 0.1435, + "step": 62943 + }, + { + "epoch": 2.42, + "learning_rate": 9.672993075147473e-06, + "loss": 0.2288, + "step": 62944 + }, + { + "epoch": 2.42, + "learning_rate": 9.672351885098743e-06, + "loss": 0.1483, + "step": 62945 + }, + { + "epoch": 2.42, + "learning_rate": 9.671710695050014e-06, + "loss": 0.1519, + "step": 62946 + }, + { + "epoch": 2.42, + "learning_rate": 9.671069505001284e-06, + "loss": 0.1351, + "step": 62947 + }, + { + "epoch": 2.42, + "learning_rate": 9.670428314952552e-06, + "loss": 0.199, + "step": 62948 + }, + { + "epoch": 2.42, + "learning_rate": 9.669787124903822e-06, + "loss": 0.2562, + "step": 62949 + }, + { + "epoch": 2.42, + "learning_rate": 9.66914593485509e-06, + "loss": 0.298, + "step": 62950 + }, + { + "epoch": 2.42, + "learning_rate": 9.668504744806362e-06, + "loss": 0.0972, + "step": 62951 + }, + { + "epoch": 2.42, + "learning_rate": 9.66786355475763e-06, + "loss": 0.1762, + "step": 62952 + }, + { + "epoch": 2.42, + "learning_rate": 9.6672223647089e-06, + "loss": 0.3796, + "step": 62953 + }, + { + "epoch": 2.42, + "learning_rate": 9.666581174660169e-06, + "loss": 0.2185, + "step": 62954 + }, + { + "epoch": 2.42, + "learning_rate": 9.66593998461144e-06, + "loss": 0.2081, + "step": 62955 + }, + { + "epoch": 2.42, + "learning_rate": 9.665298794562709e-06, + "loss": 0.1561, + "step": 62956 + }, + { + "epoch": 2.42, + "learning_rate": 9.664657604513979e-06, + "loss": 0.1203, + "step": 62957 + }, + { + "epoch": 2.42, + "learning_rate": 9.664016414465247e-06, + "loss": 0.2957, + "step": 62958 + }, + { + "epoch": 2.42, + "learning_rate": 9.663375224416517e-06, + "loss": 0.3304, + "step": 62959 + }, + { + "epoch": 2.42, + "learning_rate": 9.662734034367787e-06, + "loss": 0.2716, + "step": 62960 + }, + { + "epoch": 2.42, + "learning_rate": 9.662092844319057e-06, + "loss": 0.2209, + "step": 62961 + }, + { + "epoch": 2.42, + "learning_rate": 9.661451654270325e-06, + "loss": 0.4901, + "step": 62962 + }, + { + "epoch": 2.42, + "learning_rate": 9.660810464221595e-06, + "loss": 0.2335, + "step": 62963 + }, + { + "epoch": 2.42, + "learning_rate": 9.660169274172865e-06, + "loss": 0.285, + "step": 62964 + }, + { + "epoch": 2.42, + "learning_rate": 9.659528084124135e-06, + "loss": 0.1606, + "step": 62965 + }, + { + "epoch": 2.42, + "learning_rate": 9.658886894075404e-06, + "loss": 0.2464, + "step": 62966 + }, + { + "epoch": 2.42, + "learning_rate": 9.658245704026674e-06, + "loss": 0.1377, + "step": 62967 + }, + { + "epoch": 2.42, + "learning_rate": 9.657604513977944e-06, + "loss": 0.1571, + "step": 62968 + }, + { + "epoch": 2.42, + "learning_rate": 9.656963323929214e-06, + "loss": 0.278, + "step": 62969 + }, + { + "epoch": 2.42, + "learning_rate": 9.656322133880482e-06, + "loss": 0.2058, + "step": 62970 + }, + { + "epoch": 2.42, + "learning_rate": 9.655680943831752e-06, + "loss": 0.1968, + "step": 62971 + }, + { + "epoch": 2.42, + "learning_rate": 9.65503975378302e-06, + "loss": 0.1678, + "step": 62972 + }, + { + "epoch": 2.42, + "learning_rate": 9.654398563734292e-06, + "loss": 0.1119, + "step": 62973 + }, + { + "epoch": 2.42, + "learning_rate": 9.65375737368556e-06, + "loss": 0.1293, + "step": 62974 + }, + { + "epoch": 2.42, + "learning_rate": 9.65311618363683e-06, + "loss": 0.1465, + "step": 62975 + }, + { + "epoch": 2.42, + "learning_rate": 9.652474993588099e-06, + "loss": 0.2393, + "step": 62976 + }, + { + "epoch": 2.42, + "learning_rate": 9.65183380353937e-06, + "loss": 0.4397, + "step": 62977 + }, + { + "epoch": 2.42, + "learning_rate": 9.651192613490639e-06, + "loss": 0.2289, + "step": 62978 + }, + { + "epoch": 2.42, + "learning_rate": 9.650551423441909e-06, + "loss": 0.2776, + "step": 62979 + }, + { + "epoch": 2.42, + "learning_rate": 9.649910233393177e-06, + "loss": 0.0862, + "step": 62980 + }, + { + "epoch": 2.42, + "learning_rate": 9.649269043344447e-06, + "loss": 0.2636, + "step": 62981 + }, + { + "epoch": 2.42, + "learning_rate": 9.648627853295717e-06, + "loss": 0.236, + "step": 62982 + }, + { + "epoch": 2.42, + "learning_rate": 9.647986663246987e-06, + "loss": 0.1545, + "step": 62983 + }, + { + "epoch": 2.42, + "learning_rate": 9.647345473198256e-06, + "loss": 0.2452, + "step": 62984 + }, + { + "epoch": 2.42, + "learning_rate": 9.646704283149526e-06, + "loss": 0.3023, + "step": 62985 + }, + { + "epoch": 2.42, + "learning_rate": 9.646063093100796e-06, + "loss": 0.2089, + "step": 62986 + }, + { + "epoch": 2.42, + "learning_rate": 9.645421903052066e-06, + "loss": 0.2613, + "step": 62987 + }, + { + "epoch": 2.42, + "learning_rate": 9.644780713003334e-06, + "loss": 0.1519, + "step": 62988 + }, + { + "epoch": 2.42, + "learning_rate": 9.644139522954604e-06, + "loss": 0.1946, + "step": 62989 + }, + { + "epoch": 2.42, + "learning_rate": 9.643498332905872e-06, + "loss": 0.241, + "step": 62990 + }, + { + "epoch": 2.42, + "learning_rate": 9.642857142857144e-06, + "loss": 0.1694, + "step": 62991 + }, + { + "epoch": 2.42, + "learning_rate": 9.642215952808413e-06, + "loss": 0.3408, + "step": 62992 + }, + { + "epoch": 2.42, + "learning_rate": 9.641574762759683e-06, + "loss": 0.2396, + "step": 62993 + }, + { + "epoch": 2.42, + "learning_rate": 9.640933572710951e-06, + "loss": 0.2626, + "step": 62994 + }, + { + "epoch": 2.42, + "learning_rate": 9.640292382662223e-06, + "loss": 0.2782, + "step": 62995 + }, + { + "epoch": 2.42, + "learning_rate": 9.639651192613491e-06, + "loss": 0.2588, + "step": 62996 + }, + { + "epoch": 2.42, + "learning_rate": 9.639010002564761e-06, + "loss": 0.1894, + "step": 62997 + }, + { + "epoch": 2.42, + "learning_rate": 9.63836881251603e-06, + "loss": 0.2814, + "step": 62998 + }, + { + "epoch": 2.42, + "learning_rate": 9.6377276224673e-06, + "loss": 0.1625, + "step": 62999 + }, + { + "epoch": 2.42, + "learning_rate": 9.63708643241857e-06, + "loss": 0.2192, + "step": 63000 + }, + { + "epoch": 2.42, + "learning_rate": 9.63644524236984e-06, + "loss": 0.2362, + "step": 63001 + }, + { + "epoch": 2.42, + "learning_rate": 9.635804052321108e-06, + "loss": 0.1438, + "step": 63002 + }, + { + "epoch": 2.42, + "learning_rate": 9.635162862272378e-06, + "loss": 0.2523, + "step": 63003 + }, + { + "epoch": 2.42, + "learning_rate": 9.634521672223648e-06, + "loss": 0.1778, + "step": 63004 + }, + { + "epoch": 2.42, + "learning_rate": 9.633880482174918e-06, + "loss": 0.1099, + "step": 63005 + }, + { + "epoch": 2.42, + "learning_rate": 9.633239292126186e-06, + "loss": 0.1829, + "step": 63006 + }, + { + "epoch": 2.42, + "learning_rate": 9.632598102077456e-06, + "loss": 0.3026, + "step": 63007 + }, + { + "epoch": 2.42, + "learning_rate": 9.631956912028726e-06, + "loss": 0.1608, + "step": 63008 + }, + { + "epoch": 2.42, + "learning_rate": 9.631315721979996e-06, + "loss": 0.2028, + "step": 63009 + }, + { + "epoch": 2.42, + "learning_rate": 9.630674531931264e-06, + "loss": 0.1393, + "step": 63010 + }, + { + "epoch": 2.42, + "learning_rate": 9.630033341882534e-06, + "loss": 0.2107, + "step": 63011 + }, + { + "epoch": 2.42, + "learning_rate": 9.629392151833803e-06, + "loss": 0.1317, + "step": 63012 + }, + { + "epoch": 2.42, + "learning_rate": 9.628750961785075e-06, + "loss": 0.3079, + "step": 63013 + }, + { + "epoch": 2.42, + "learning_rate": 9.628109771736343e-06, + "loss": 0.2427, + "step": 63014 + }, + { + "epoch": 2.42, + "learning_rate": 9.627468581687613e-06, + "loss": 0.1331, + "step": 63015 + }, + { + "epoch": 2.42, + "learning_rate": 9.626827391638881e-06, + "loss": 0.2052, + "step": 63016 + }, + { + "epoch": 2.42, + "learning_rate": 9.626186201590153e-06, + "loss": 0.2591, + "step": 63017 + }, + { + "epoch": 2.42, + "learning_rate": 9.625545011541421e-06, + "loss": 0.2862, + "step": 63018 + }, + { + "epoch": 2.42, + "learning_rate": 9.624903821492691e-06, + "loss": 0.2592, + "step": 63019 + }, + { + "epoch": 2.42, + "learning_rate": 9.62426263144396e-06, + "loss": 0.2041, + "step": 63020 + }, + { + "epoch": 2.42, + "learning_rate": 9.62362144139523e-06, + "loss": 0.2488, + "step": 63021 + }, + { + "epoch": 2.42, + "learning_rate": 9.6229802513465e-06, + "loss": 0.2117, + "step": 63022 + }, + { + "epoch": 2.42, + "learning_rate": 9.62233906129777e-06, + "loss": 0.2405, + "step": 63023 + }, + { + "epoch": 2.42, + "learning_rate": 9.621697871249038e-06, + "loss": 0.3649, + "step": 63024 + }, + { + "epoch": 2.42, + "learning_rate": 9.621056681200308e-06, + "loss": 0.286, + "step": 63025 + }, + { + "epoch": 2.42, + "learning_rate": 9.620415491151578e-06, + "loss": 0.1683, + "step": 63026 + }, + { + "epoch": 2.42, + "learning_rate": 9.619774301102848e-06, + "loss": 0.2965, + "step": 63027 + }, + { + "epoch": 2.42, + "learning_rate": 9.619133111054116e-06, + "loss": 0.1531, + "step": 63028 + }, + { + "epoch": 2.42, + "learning_rate": 9.618491921005386e-06, + "loss": 0.3373, + "step": 63029 + }, + { + "epoch": 2.42, + "learning_rate": 9.617850730956655e-06, + "loss": 0.1761, + "step": 63030 + }, + { + "epoch": 2.42, + "learning_rate": 9.617209540907926e-06, + "loss": 0.2053, + "step": 63031 + }, + { + "epoch": 2.42, + "learning_rate": 9.616568350859195e-06, + "loss": 0.3833, + "step": 63032 + }, + { + "epoch": 2.42, + "learning_rate": 9.615927160810465e-06, + "loss": 0.2338, + "step": 63033 + }, + { + "epoch": 2.42, + "learning_rate": 9.615285970761733e-06, + "loss": 0.1793, + "step": 63034 + }, + { + "epoch": 2.42, + "learning_rate": 9.614644780713005e-06, + "loss": 0.1304, + "step": 63035 + }, + { + "epoch": 2.42, + "learning_rate": 9.614003590664273e-06, + "loss": 0.2247, + "step": 63036 + }, + { + "epoch": 2.42, + "learning_rate": 9.613362400615543e-06, + "loss": 0.3154, + "step": 63037 + }, + { + "epoch": 2.42, + "learning_rate": 9.612721210566812e-06, + "loss": 0.1393, + "step": 63038 + }, + { + "epoch": 2.42, + "learning_rate": 9.612080020518082e-06, + "loss": 0.3218, + "step": 63039 + }, + { + "epoch": 2.42, + "learning_rate": 9.611438830469352e-06, + "loss": 0.2937, + "step": 63040 + }, + { + "epoch": 2.42, + "learning_rate": 9.610797640420622e-06, + "loss": 0.2616, + "step": 63041 + }, + { + "epoch": 2.42, + "learning_rate": 9.61015645037189e-06, + "loss": 0.1892, + "step": 63042 + }, + { + "epoch": 2.42, + "learning_rate": 9.60951526032316e-06, + "loss": 0.143, + "step": 63043 + }, + { + "epoch": 2.42, + "learning_rate": 9.60887407027443e-06, + "loss": 0.2343, + "step": 63044 + }, + { + "epoch": 2.42, + "learning_rate": 9.6082328802257e-06, + "loss": 0.1444, + "step": 63045 + }, + { + "epoch": 2.42, + "learning_rate": 9.607591690176968e-06, + "loss": 0.1722, + "step": 63046 + }, + { + "epoch": 2.42, + "learning_rate": 9.606950500128238e-06, + "loss": 0.3269, + "step": 63047 + }, + { + "epoch": 2.42, + "learning_rate": 9.606309310079507e-06, + "loss": 0.2579, + "step": 63048 + }, + { + "epoch": 2.42, + "learning_rate": 9.605668120030778e-06, + "loss": 0.2399, + "step": 63049 + }, + { + "epoch": 2.42, + "learning_rate": 9.605026929982047e-06, + "loss": 0.2071, + "step": 63050 + }, + { + "epoch": 2.42, + "learning_rate": 9.604385739933317e-06, + "loss": 0.2242, + "step": 63051 + }, + { + "epoch": 2.42, + "learning_rate": 9.603744549884585e-06, + "loss": 0.1508, + "step": 63052 + }, + { + "epoch": 2.42, + "learning_rate": 9.603103359835857e-06, + "loss": 0.2417, + "step": 63053 + }, + { + "epoch": 2.42, + "learning_rate": 9.602462169787125e-06, + "loss": 0.2114, + "step": 63054 + }, + { + "epoch": 2.42, + "learning_rate": 9.601820979738395e-06, + "loss": 0.1457, + "step": 63055 + }, + { + "epoch": 2.42, + "learning_rate": 9.601179789689663e-06, + "loss": 0.1149, + "step": 63056 + }, + { + "epoch": 2.42, + "learning_rate": 9.600538599640935e-06, + "loss": 0.2505, + "step": 63057 + }, + { + "epoch": 2.42, + "learning_rate": 9.599897409592204e-06, + "loss": 0.2337, + "step": 63058 + }, + { + "epoch": 2.42, + "learning_rate": 9.599256219543474e-06, + "loss": 0.2089, + "step": 63059 + }, + { + "epoch": 2.42, + "learning_rate": 9.598615029494742e-06, + "loss": 0.2207, + "step": 63060 + }, + { + "epoch": 2.42, + "learning_rate": 9.597973839446012e-06, + "loss": 0.2402, + "step": 63061 + }, + { + "epoch": 2.42, + "learning_rate": 9.597332649397282e-06, + "loss": 0.2854, + "step": 63062 + }, + { + "epoch": 2.42, + "learning_rate": 9.596691459348552e-06, + "loss": 0.2623, + "step": 63063 + }, + { + "epoch": 2.42, + "learning_rate": 9.59605026929982e-06, + "loss": 0.2361, + "step": 63064 + }, + { + "epoch": 2.42, + "learning_rate": 9.59540907925109e-06, + "loss": 0.2735, + "step": 63065 + }, + { + "epoch": 2.42, + "learning_rate": 9.59476788920236e-06, + "loss": 0.1953, + "step": 63066 + }, + { + "epoch": 2.42, + "learning_rate": 9.59412669915363e-06, + "loss": 0.2002, + "step": 63067 + }, + { + "epoch": 2.42, + "learning_rate": 9.593485509104899e-06, + "loss": 0.2277, + "step": 63068 + }, + { + "epoch": 2.42, + "learning_rate": 9.592844319056169e-06, + "loss": 0.2471, + "step": 63069 + }, + { + "epoch": 2.42, + "learning_rate": 9.592203129007437e-06, + "loss": 0.1632, + "step": 63070 + }, + { + "epoch": 2.42, + "learning_rate": 9.591561938958709e-06, + "loss": 0.3038, + "step": 63071 + }, + { + "epoch": 2.42, + "learning_rate": 9.590920748909977e-06, + "loss": 0.3068, + "step": 63072 + }, + { + "epoch": 2.42, + "learning_rate": 9.590279558861247e-06, + "loss": 0.1995, + "step": 63073 + }, + { + "epoch": 2.42, + "learning_rate": 9.589638368812515e-06, + "loss": 0.213, + "step": 63074 + }, + { + "epoch": 2.43, + "learning_rate": 9.588997178763787e-06, + "loss": 0.3072, + "step": 63075 + }, + { + "epoch": 2.43, + "learning_rate": 9.588355988715055e-06, + "loss": 0.214, + "step": 63076 + }, + { + "epoch": 2.43, + "learning_rate": 9.587714798666325e-06, + "loss": 0.1902, + "step": 63077 + }, + { + "epoch": 2.43, + "learning_rate": 9.587073608617594e-06, + "loss": 0.2882, + "step": 63078 + }, + { + "epoch": 2.43, + "learning_rate": 9.586432418568864e-06, + "loss": 0.3277, + "step": 63079 + }, + { + "epoch": 2.43, + "learning_rate": 9.585791228520134e-06, + "loss": 0.136, + "step": 63080 + }, + { + "epoch": 2.43, + "learning_rate": 9.585150038471404e-06, + "loss": 0.1411, + "step": 63081 + }, + { + "epoch": 2.43, + "learning_rate": 9.584508848422672e-06, + "loss": 0.2564, + "step": 63082 + }, + { + "epoch": 2.43, + "learning_rate": 9.583867658373942e-06, + "loss": 0.2734, + "step": 63083 + }, + { + "epoch": 2.43, + "learning_rate": 9.583226468325212e-06, + "loss": 0.1338, + "step": 63084 + }, + { + "epoch": 2.43, + "learning_rate": 9.582585278276482e-06, + "loss": 0.4113, + "step": 63085 + }, + { + "epoch": 2.43, + "learning_rate": 9.58194408822775e-06, + "loss": 0.2711, + "step": 63086 + }, + { + "epoch": 2.43, + "learning_rate": 9.58130289817902e-06, + "loss": 0.2087, + "step": 63087 + }, + { + "epoch": 2.43, + "learning_rate": 9.580661708130289e-06, + "loss": 0.2298, + "step": 63088 + }, + { + "epoch": 2.43, + "learning_rate": 9.58002051808156e-06, + "loss": 0.208, + "step": 63089 + }, + { + "epoch": 2.43, + "learning_rate": 9.579379328032829e-06, + "loss": 0.1528, + "step": 63090 + }, + { + "epoch": 2.43, + "learning_rate": 9.578738137984099e-06, + "loss": 0.1535, + "step": 63091 + }, + { + "epoch": 2.43, + "learning_rate": 9.578096947935367e-06, + "loss": 0.1421, + "step": 63092 + }, + { + "epoch": 2.43, + "learning_rate": 9.577455757886639e-06, + "loss": 0.2197, + "step": 63093 + }, + { + "epoch": 2.43, + "learning_rate": 9.576814567837907e-06, + "loss": 0.1554, + "step": 63094 + }, + { + "epoch": 2.43, + "learning_rate": 9.576173377789177e-06, + "loss": 0.3022, + "step": 63095 + }, + { + "epoch": 2.43, + "learning_rate": 9.575532187740446e-06, + "loss": 0.2011, + "step": 63096 + }, + { + "epoch": 2.43, + "learning_rate": 9.574890997691716e-06, + "loss": 0.2058, + "step": 63097 + }, + { + "epoch": 2.43, + "learning_rate": 9.574249807642986e-06, + "loss": 0.2457, + "step": 63098 + }, + { + "epoch": 2.43, + "learning_rate": 9.573608617594256e-06, + "loss": 0.279, + "step": 63099 + }, + { + "epoch": 2.43, + "learning_rate": 9.572967427545524e-06, + "loss": 0.2217, + "step": 63100 + }, + { + "epoch": 2.43, + "learning_rate": 9.572326237496794e-06, + "loss": 0.3325, + "step": 63101 + }, + { + "epoch": 2.43, + "learning_rate": 9.571685047448064e-06, + "loss": 0.2024, + "step": 63102 + }, + { + "epoch": 2.43, + "learning_rate": 9.571043857399334e-06, + "loss": 0.2116, + "step": 63103 + }, + { + "epoch": 2.43, + "learning_rate": 9.570402667350603e-06, + "loss": 0.2697, + "step": 63104 + }, + { + "epoch": 2.43, + "learning_rate": 9.569761477301873e-06, + "loss": 0.3177, + "step": 63105 + }, + { + "epoch": 2.43, + "learning_rate": 9.569120287253143e-06, + "loss": 0.1817, + "step": 63106 + }, + { + "epoch": 2.43, + "learning_rate": 9.568479097204413e-06, + "loss": 0.2278, + "step": 63107 + }, + { + "epoch": 2.43, + "learning_rate": 9.567837907155681e-06, + "loss": 0.241, + "step": 63108 + }, + { + "epoch": 2.43, + "learning_rate": 9.567196717106951e-06, + "loss": 0.4831, + "step": 63109 + }, + { + "epoch": 2.43, + "learning_rate": 9.56655552705822e-06, + "loss": 0.2884, + "step": 63110 + }, + { + "epoch": 2.43, + "learning_rate": 9.565914337009491e-06, + "loss": 0.2144, + "step": 63111 + }, + { + "epoch": 2.43, + "learning_rate": 9.56527314696076e-06, + "loss": 0.174, + "step": 63112 + }, + { + "epoch": 2.43, + "learning_rate": 9.56463195691203e-06, + "loss": 0.342, + "step": 63113 + }, + { + "epoch": 2.43, + "learning_rate": 9.563990766863298e-06, + "loss": 0.1602, + "step": 63114 + }, + { + "epoch": 2.43, + "learning_rate": 9.56334957681457e-06, + "loss": 0.1334, + "step": 63115 + }, + { + "epoch": 2.43, + "learning_rate": 9.562708386765838e-06, + "loss": 0.2887, + "step": 63116 + }, + { + "epoch": 2.43, + "learning_rate": 9.562067196717108e-06, + "loss": 0.2448, + "step": 63117 + }, + { + "epoch": 2.43, + "learning_rate": 9.561426006668376e-06, + "loss": 0.1769, + "step": 63118 + }, + { + "epoch": 2.43, + "learning_rate": 9.560784816619646e-06, + "loss": 0.215, + "step": 63119 + }, + { + "epoch": 2.43, + "learning_rate": 9.560143626570916e-06, + "loss": 0.3347, + "step": 63120 + }, + { + "epoch": 2.43, + "learning_rate": 9.559502436522186e-06, + "loss": 0.1828, + "step": 63121 + }, + { + "epoch": 2.43, + "learning_rate": 9.558861246473454e-06, + "loss": 0.1499, + "step": 63122 + }, + { + "epoch": 2.43, + "learning_rate": 9.558220056424724e-06, + "loss": 0.2299, + "step": 63123 + }, + { + "epoch": 2.43, + "learning_rate": 9.557578866375995e-06, + "loss": 0.131, + "step": 63124 + }, + { + "epoch": 2.43, + "learning_rate": 9.556937676327265e-06, + "loss": 0.2397, + "step": 63125 + }, + { + "epoch": 2.43, + "learning_rate": 9.556296486278533e-06, + "loss": 0.2164, + "step": 63126 + }, + { + "epoch": 2.43, + "learning_rate": 9.555655296229803e-06, + "loss": 0.3588, + "step": 63127 + }, + { + "epoch": 2.43, + "learning_rate": 9.555014106181071e-06, + "loss": 0.228, + "step": 63128 + }, + { + "epoch": 2.43, + "learning_rate": 9.554372916132343e-06, + "loss": 0.3704, + "step": 63129 + }, + { + "epoch": 2.43, + "learning_rate": 9.553731726083611e-06, + "loss": 0.2017, + "step": 63130 + }, + { + "epoch": 2.43, + "learning_rate": 9.553090536034881e-06, + "loss": 0.354, + "step": 63131 + }, + { + "epoch": 2.43, + "learning_rate": 9.55244934598615e-06, + "loss": 0.1576, + "step": 63132 + }, + { + "epoch": 2.43, + "learning_rate": 9.551808155937421e-06, + "loss": 0.2225, + "step": 63133 + }, + { + "epoch": 2.43, + "learning_rate": 9.55116696588869e-06, + "loss": 0.1976, + "step": 63134 + }, + { + "epoch": 2.43, + "learning_rate": 9.55052577583996e-06, + "loss": 0.2102, + "step": 63135 + }, + { + "epoch": 2.43, + "learning_rate": 9.549884585791228e-06, + "loss": 0.2198, + "step": 63136 + }, + { + "epoch": 2.43, + "learning_rate": 9.549243395742498e-06, + "loss": 0.1694, + "step": 63137 + }, + { + "epoch": 2.43, + "learning_rate": 9.548602205693768e-06, + "loss": 0.2878, + "step": 63138 + }, + { + "epoch": 2.43, + "learning_rate": 9.547961015645038e-06, + "loss": 0.1438, + "step": 63139 + }, + { + "epoch": 2.43, + "learning_rate": 9.547319825596306e-06, + "loss": 0.2395, + "step": 63140 + }, + { + "epoch": 2.43, + "learning_rate": 9.546678635547576e-06, + "loss": 0.2009, + "step": 63141 + }, + { + "epoch": 2.43, + "learning_rate": 9.546037445498846e-06, + "loss": 0.1731, + "step": 63142 + }, + { + "epoch": 2.43, + "learning_rate": 9.545396255450116e-06, + "loss": 0.1888, + "step": 63143 + }, + { + "epoch": 2.43, + "learning_rate": 9.544755065401385e-06, + "loss": 0.2933, + "step": 63144 + }, + { + "epoch": 2.43, + "learning_rate": 9.544113875352655e-06, + "loss": 0.2133, + "step": 63145 + }, + { + "epoch": 2.43, + "learning_rate": 9.543472685303923e-06, + "loss": 0.3515, + "step": 63146 + }, + { + "epoch": 2.43, + "learning_rate": 9.542831495255195e-06, + "loss": 0.1396, + "step": 63147 + }, + { + "epoch": 2.43, + "learning_rate": 9.542190305206463e-06, + "loss": 0.3119, + "step": 63148 + }, + { + "epoch": 2.43, + "learning_rate": 9.541549115157733e-06, + "loss": 0.3317, + "step": 63149 + }, + { + "epoch": 2.43, + "learning_rate": 9.540907925109002e-06, + "loss": 0.2125, + "step": 63150 + }, + { + "epoch": 2.43, + "learning_rate": 9.540266735060273e-06, + "loss": 0.229, + "step": 63151 + }, + { + "epoch": 2.43, + "learning_rate": 9.539625545011542e-06, + "loss": 0.2714, + "step": 63152 + }, + { + "epoch": 2.43, + "learning_rate": 9.538984354962812e-06, + "loss": 0.2058, + "step": 63153 + }, + { + "epoch": 2.43, + "learning_rate": 9.53834316491408e-06, + "loss": 0.2918, + "step": 63154 + }, + { + "epoch": 2.43, + "learning_rate": 9.537701974865352e-06, + "loss": 0.1538, + "step": 63155 + }, + { + "epoch": 2.43, + "learning_rate": 9.53706078481662e-06, + "loss": 0.2453, + "step": 63156 + }, + { + "epoch": 2.43, + "learning_rate": 9.53641959476789e-06, + "loss": 0.3104, + "step": 63157 + }, + { + "epoch": 2.43, + "learning_rate": 9.535778404719158e-06, + "loss": 0.1742, + "step": 63158 + }, + { + "epoch": 2.43, + "learning_rate": 9.535137214670428e-06, + "loss": 0.1945, + "step": 63159 + }, + { + "epoch": 2.43, + "learning_rate": 9.534496024621698e-06, + "loss": 0.2398, + "step": 63160 + }, + { + "epoch": 2.43, + "learning_rate": 9.533854834572968e-06, + "loss": 0.1627, + "step": 63161 + }, + { + "epoch": 2.43, + "learning_rate": 9.533213644524237e-06, + "loss": 0.2208, + "step": 63162 + }, + { + "epoch": 2.43, + "learning_rate": 9.532572454475507e-06, + "loss": 0.1635, + "step": 63163 + }, + { + "epoch": 2.43, + "learning_rate": 9.531931264426777e-06, + "loss": 0.1223, + "step": 63164 + }, + { + "epoch": 2.43, + "learning_rate": 9.531290074378047e-06, + "loss": 0.1786, + "step": 63165 + }, + { + "epoch": 2.43, + "learning_rate": 9.530648884329315e-06, + "loss": 0.1811, + "step": 63166 + }, + { + "epoch": 2.43, + "learning_rate": 9.530007694280585e-06, + "loss": 0.2149, + "step": 63167 + }, + { + "epoch": 2.43, + "learning_rate": 9.529366504231853e-06, + "loss": 0.2258, + "step": 63168 + }, + { + "epoch": 2.43, + "learning_rate": 9.528725314183125e-06, + "loss": 0.1207, + "step": 63169 + }, + { + "epoch": 2.43, + "learning_rate": 9.528084124134394e-06, + "loss": 0.1998, + "step": 63170 + }, + { + "epoch": 2.43, + "learning_rate": 9.527442934085664e-06, + "loss": 0.3091, + "step": 63171 + }, + { + "epoch": 2.43, + "learning_rate": 9.526801744036932e-06, + "loss": 0.2186, + "step": 63172 + }, + { + "epoch": 2.43, + "learning_rate": 9.526160553988204e-06, + "loss": 0.3432, + "step": 63173 + }, + { + "epoch": 2.43, + "learning_rate": 9.525519363939472e-06, + "loss": 0.2313, + "step": 63174 + }, + { + "epoch": 2.43, + "learning_rate": 9.524878173890742e-06, + "loss": 0.215, + "step": 63175 + }, + { + "epoch": 2.43, + "learning_rate": 9.52423698384201e-06, + "loss": 0.1999, + "step": 63176 + }, + { + "epoch": 2.43, + "learning_rate": 9.52359579379328e-06, + "loss": 0.2646, + "step": 63177 + }, + { + "epoch": 2.43, + "learning_rate": 9.52295460374455e-06, + "loss": 0.2395, + "step": 63178 + }, + { + "epoch": 2.43, + "learning_rate": 9.52231341369582e-06, + "loss": 0.1529, + "step": 63179 + }, + { + "epoch": 2.43, + "learning_rate": 9.521672223647089e-06, + "loss": 0.3305, + "step": 63180 + }, + { + "epoch": 2.43, + "learning_rate": 9.521031033598359e-06, + "loss": 0.1672, + "step": 63181 + }, + { + "epoch": 2.43, + "learning_rate": 9.520389843549629e-06, + "loss": 0.2165, + "step": 63182 + }, + { + "epoch": 2.43, + "learning_rate": 9.519748653500899e-06, + "loss": 0.178, + "step": 63183 + }, + { + "epoch": 2.43, + "learning_rate": 9.519107463452167e-06, + "loss": 0.2258, + "step": 63184 + }, + { + "epoch": 2.43, + "learning_rate": 9.518466273403437e-06, + "loss": 0.1829, + "step": 63185 + }, + { + "epoch": 2.43, + "learning_rate": 9.517825083354705e-06, + "loss": 0.3624, + "step": 63186 + }, + { + "epoch": 2.43, + "learning_rate": 9.517183893305977e-06, + "loss": 0.2595, + "step": 63187 + }, + { + "epoch": 2.43, + "learning_rate": 9.516542703257245e-06, + "loss": 0.2424, + "step": 63188 + }, + { + "epoch": 2.43, + "learning_rate": 9.515901513208515e-06, + "loss": 0.2153, + "step": 63189 + }, + { + "epoch": 2.43, + "learning_rate": 9.515260323159784e-06, + "loss": 0.1258, + "step": 63190 + }, + { + "epoch": 2.43, + "learning_rate": 9.514619133111056e-06, + "loss": 0.1994, + "step": 63191 + }, + { + "epoch": 2.43, + "learning_rate": 9.513977943062324e-06, + "loss": 0.2129, + "step": 63192 + }, + { + "epoch": 2.43, + "learning_rate": 9.513336753013594e-06, + "loss": 0.2415, + "step": 63193 + }, + { + "epoch": 2.43, + "learning_rate": 9.512695562964862e-06, + "loss": 0.174, + "step": 63194 + }, + { + "epoch": 2.43, + "learning_rate": 9.512054372916132e-06, + "loss": 0.1906, + "step": 63195 + }, + { + "epoch": 2.43, + "learning_rate": 9.511413182867402e-06, + "loss": 0.253, + "step": 63196 + }, + { + "epoch": 2.43, + "learning_rate": 9.510771992818672e-06, + "loss": 0.1533, + "step": 63197 + }, + { + "epoch": 2.43, + "learning_rate": 9.51013080276994e-06, + "loss": 0.2282, + "step": 63198 + }, + { + "epoch": 2.43, + "learning_rate": 9.50948961272121e-06, + "loss": 0.1311, + "step": 63199 + }, + { + "epoch": 2.43, + "learning_rate": 9.50884842267248e-06, + "loss": 0.2824, + "step": 63200 + }, + { + "epoch": 2.43, + "learning_rate": 9.50820723262375e-06, + "loss": 0.2433, + "step": 63201 + }, + { + "epoch": 2.43, + "learning_rate": 9.507566042575019e-06, + "loss": 0.2401, + "step": 63202 + }, + { + "epoch": 2.43, + "learning_rate": 9.506924852526289e-06, + "loss": 0.2693, + "step": 63203 + }, + { + "epoch": 2.43, + "learning_rate": 9.506283662477559e-06, + "loss": 0.1311, + "step": 63204 + }, + { + "epoch": 2.43, + "learning_rate": 9.505642472428829e-06, + "loss": 0.109, + "step": 63205 + }, + { + "epoch": 2.43, + "learning_rate": 9.505001282380097e-06, + "loss": 0.2106, + "step": 63206 + }, + { + "epoch": 2.43, + "learning_rate": 9.504360092331367e-06, + "loss": 0.2376, + "step": 63207 + }, + { + "epoch": 2.43, + "learning_rate": 9.503718902282636e-06, + "loss": 0.238, + "step": 63208 + }, + { + "epoch": 2.43, + "learning_rate": 9.503077712233907e-06, + "loss": 0.2321, + "step": 63209 + }, + { + "epoch": 2.43, + "learning_rate": 9.502436522185176e-06, + "loss": 0.3327, + "step": 63210 + }, + { + "epoch": 2.43, + "learning_rate": 9.501795332136446e-06, + "loss": 0.1362, + "step": 63211 + }, + { + "epoch": 2.43, + "learning_rate": 9.501154142087714e-06, + "loss": 0.2938, + "step": 63212 + }, + { + "epoch": 2.43, + "learning_rate": 9.500512952038986e-06, + "loss": 0.3101, + "step": 63213 + }, + { + "epoch": 2.43, + "learning_rate": 9.499871761990254e-06, + "loss": 0.2121, + "step": 63214 + }, + { + "epoch": 2.43, + "learning_rate": 9.499230571941524e-06, + "loss": 0.1934, + "step": 63215 + }, + { + "epoch": 2.43, + "learning_rate": 9.498589381892793e-06, + "loss": 0.3107, + "step": 63216 + }, + { + "epoch": 2.43, + "learning_rate": 9.497948191844063e-06, + "loss": 0.217, + "step": 63217 + }, + { + "epoch": 2.43, + "learning_rate": 9.497307001795333e-06, + "loss": 0.2336, + "step": 63218 + }, + { + "epoch": 2.43, + "learning_rate": 9.496665811746603e-06, + "loss": 0.1927, + "step": 63219 + }, + { + "epoch": 2.43, + "learning_rate": 9.496024621697871e-06, + "loss": 0.2588, + "step": 63220 + }, + { + "epoch": 2.43, + "learning_rate": 9.495383431649141e-06, + "loss": 0.1418, + "step": 63221 + }, + { + "epoch": 2.43, + "learning_rate": 9.494742241600411e-06, + "loss": 0.3477, + "step": 63222 + }, + { + "epoch": 2.43, + "learning_rate": 9.494101051551681e-06, + "loss": 0.2393, + "step": 63223 + }, + { + "epoch": 2.43, + "learning_rate": 9.49345986150295e-06, + "loss": 0.4425, + "step": 63224 + }, + { + "epoch": 2.43, + "learning_rate": 9.49281867145422e-06, + "loss": 0.2021, + "step": 63225 + }, + { + "epoch": 2.43, + "learning_rate": 9.492177481405488e-06, + "loss": 0.192, + "step": 63226 + }, + { + "epoch": 2.43, + "learning_rate": 9.49153629135676e-06, + "loss": 0.2405, + "step": 63227 + }, + { + "epoch": 2.43, + "learning_rate": 9.490895101308028e-06, + "loss": 0.3353, + "step": 63228 + }, + { + "epoch": 2.43, + "learning_rate": 9.490253911259298e-06, + "loss": 0.2925, + "step": 63229 + }, + { + "epoch": 2.43, + "learning_rate": 9.489612721210566e-06, + "loss": 0.2533, + "step": 63230 + }, + { + "epoch": 2.43, + "learning_rate": 9.488971531161838e-06, + "loss": 0.2355, + "step": 63231 + }, + { + "epoch": 2.43, + "learning_rate": 9.488330341113106e-06, + "loss": 0.2046, + "step": 63232 + }, + { + "epoch": 2.43, + "learning_rate": 9.487689151064376e-06, + "loss": 0.1921, + "step": 63233 + }, + { + "epoch": 2.43, + "learning_rate": 9.487047961015644e-06, + "loss": 0.2476, + "step": 63234 + }, + { + "epoch": 2.43, + "learning_rate": 9.486406770966914e-06, + "loss": 0.2595, + "step": 63235 + }, + { + "epoch": 2.43, + "learning_rate": 9.485765580918185e-06, + "loss": 0.2249, + "step": 63236 + }, + { + "epoch": 2.43, + "learning_rate": 9.485124390869455e-06, + "loss": 0.1351, + "step": 63237 + }, + { + "epoch": 2.43, + "learning_rate": 9.484483200820723e-06, + "loss": 0.2332, + "step": 63238 + }, + { + "epoch": 2.43, + "learning_rate": 9.483842010771993e-06, + "loss": 0.2005, + "step": 63239 + }, + { + "epoch": 2.43, + "learning_rate": 9.483200820723263e-06, + "loss": 0.3292, + "step": 63240 + }, + { + "epoch": 2.43, + "learning_rate": 9.482559630674533e-06, + "loss": 0.1834, + "step": 63241 + }, + { + "epoch": 2.43, + "learning_rate": 9.481918440625801e-06, + "loss": 0.2135, + "step": 63242 + }, + { + "epoch": 2.43, + "learning_rate": 9.481277250577071e-06, + "loss": 0.1799, + "step": 63243 + }, + { + "epoch": 2.43, + "learning_rate": 9.480636060528341e-06, + "loss": 0.3491, + "step": 63244 + }, + { + "epoch": 2.43, + "learning_rate": 9.479994870479611e-06, + "loss": 0.166, + "step": 63245 + }, + { + "epoch": 2.43, + "learning_rate": 9.47935368043088e-06, + "loss": 0.1472, + "step": 63246 + }, + { + "epoch": 2.43, + "learning_rate": 9.47871249038215e-06, + "loss": 0.1932, + "step": 63247 + }, + { + "epoch": 2.43, + "learning_rate": 9.478071300333418e-06, + "loss": 0.1343, + "step": 63248 + }, + { + "epoch": 2.43, + "learning_rate": 9.47743011028469e-06, + "loss": 0.3016, + "step": 63249 + }, + { + "epoch": 2.43, + "learning_rate": 9.476788920235958e-06, + "loss": 0.1898, + "step": 63250 + }, + { + "epoch": 2.43, + "learning_rate": 9.476147730187228e-06, + "loss": 0.1486, + "step": 63251 + }, + { + "epoch": 2.43, + "learning_rate": 9.475506540138496e-06, + "loss": 0.3284, + "step": 63252 + }, + { + "epoch": 2.43, + "learning_rate": 9.474865350089768e-06, + "loss": 0.1327, + "step": 63253 + }, + { + "epoch": 2.43, + "learning_rate": 9.474224160041036e-06, + "loss": 0.2342, + "step": 63254 + }, + { + "epoch": 2.43, + "learning_rate": 9.473582969992306e-06, + "loss": 0.2077, + "step": 63255 + }, + { + "epoch": 2.43, + "learning_rate": 9.472941779943575e-06, + "loss": 0.2057, + "step": 63256 + }, + { + "epoch": 2.43, + "learning_rate": 9.472300589894845e-06, + "loss": 0.1665, + "step": 63257 + }, + { + "epoch": 2.43, + "learning_rate": 9.471659399846115e-06, + "loss": 0.2521, + "step": 63258 + }, + { + "epoch": 2.43, + "learning_rate": 9.471018209797385e-06, + "loss": 0.2073, + "step": 63259 + }, + { + "epoch": 2.43, + "learning_rate": 9.470377019748653e-06, + "loss": 0.1015, + "step": 63260 + }, + { + "epoch": 2.43, + "learning_rate": 9.469735829699923e-06, + "loss": 0.3245, + "step": 63261 + }, + { + "epoch": 2.43, + "learning_rate": 9.469094639651193e-06, + "loss": 0.1645, + "step": 63262 + }, + { + "epoch": 2.43, + "learning_rate": 9.468453449602463e-06, + "loss": 0.341, + "step": 63263 + }, + { + "epoch": 2.43, + "learning_rate": 9.467812259553732e-06, + "loss": 0.157, + "step": 63264 + }, + { + "epoch": 2.43, + "learning_rate": 9.467171069505002e-06, + "loss": 0.2047, + "step": 63265 + }, + { + "epoch": 2.43, + "learning_rate": 9.46652987945627e-06, + "loss": 0.1791, + "step": 63266 + }, + { + "epoch": 2.43, + "learning_rate": 9.465888689407542e-06, + "loss": 0.302, + "step": 63267 + }, + { + "epoch": 2.43, + "learning_rate": 9.46524749935881e-06, + "loss": 0.2347, + "step": 63268 + }, + { + "epoch": 2.43, + "learning_rate": 9.46460630931008e-06, + "loss": 0.1844, + "step": 63269 + }, + { + "epoch": 2.43, + "learning_rate": 9.463965119261348e-06, + "loss": 0.1492, + "step": 63270 + }, + { + "epoch": 2.43, + "learning_rate": 9.46332392921262e-06, + "loss": 0.2245, + "step": 63271 + }, + { + "epoch": 2.43, + "learning_rate": 9.462682739163888e-06, + "loss": 0.222, + "step": 63272 + }, + { + "epoch": 2.43, + "learning_rate": 9.462041549115158e-06, + "loss": 0.1552, + "step": 63273 + }, + { + "epoch": 2.43, + "learning_rate": 9.461400359066427e-06, + "loss": 0.246, + "step": 63274 + }, + { + "epoch": 2.43, + "learning_rate": 9.460759169017697e-06, + "loss": 0.2215, + "step": 63275 + }, + { + "epoch": 2.43, + "learning_rate": 9.460117978968967e-06, + "loss": 0.315, + "step": 63276 + }, + { + "epoch": 2.43, + "learning_rate": 9.459476788920237e-06, + "loss": 0.2461, + "step": 63277 + }, + { + "epoch": 2.43, + "learning_rate": 9.458835598871505e-06, + "loss": 0.2765, + "step": 63278 + }, + { + "epoch": 2.43, + "learning_rate": 9.458194408822775e-06, + "loss": 0.2417, + "step": 63279 + }, + { + "epoch": 2.43, + "learning_rate": 9.457553218774045e-06, + "loss": 0.2175, + "step": 63280 + }, + { + "epoch": 2.43, + "learning_rate": 9.456912028725315e-06, + "loss": 0.1158, + "step": 63281 + }, + { + "epoch": 2.43, + "learning_rate": 9.456270838676584e-06, + "loss": 0.292, + "step": 63282 + }, + { + "epoch": 2.43, + "learning_rate": 9.455629648627854e-06, + "loss": 0.1674, + "step": 63283 + }, + { + "epoch": 2.43, + "learning_rate": 9.454988458579122e-06, + "loss": 0.2899, + "step": 63284 + }, + { + "epoch": 2.43, + "learning_rate": 9.454347268530394e-06, + "loss": 0.2173, + "step": 63285 + }, + { + "epoch": 2.43, + "learning_rate": 9.453706078481662e-06, + "loss": 0.2455, + "step": 63286 + }, + { + "epoch": 2.43, + "learning_rate": 9.453064888432932e-06, + "loss": 0.1846, + "step": 63287 + }, + { + "epoch": 2.43, + "learning_rate": 9.4524236983842e-06, + "loss": 0.153, + "step": 63288 + }, + { + "epoch": 2.43, + "learning_rate": 9.451782508335472e-06, + "loss": 0.3168, + "step": 63289 + }, + { + "epoch": 2.43, + "learning_rate": 9.45114131828674e-06, + "loss": 0.1575, + "step": 63290 + }, + { + "epoch": 2.43, + "learning_rate": 9.45050012823801e-06, + "loss": 0.2549, + "step": 63291 + }, + { + "epoch": 2.43, + "learning_rate": 9.449858938189279e-06, + "loss": 0.3775, + "step": 63292 + }, + { + "epoch": 2.43, + "learning_rate": 9.44921774814055e-06, + "loss": 0.255, + "step": 63293 + }, + { + "epoch": 2.43, + "learning_rate": 9.448576558091819e-06, + "loss": 0.2111, + "step": 63294 + }, + { + "epoch": 2.43, + "learning_rate": 9.447935368043089e-06, + "loss": 0.2706, + "step": 63295 + }, + { + "epoch": 2.43, + "learning_rate": 9.447294177994357e-06, + "loss": 0.3018, + "step": 63296 + }, + { + "epoch": 2.43, + "learning_rate": 9.446652987945627e-06, + "loss": 0.2248, + "step": 63297 + }, + { + "epoch": 2.43, + "learning_rate": 9.446011797896897e-06, + "loss": 0.1537, + "step": 63298 + }, + { + "epoch": 2.43, + "learning_rate": 9.445370607848167e-06, + "loss": 0.1288, + "step": 63299 + }, + { + "epoch": 2.43, + "learning_rate": 9.444729417799435e-06, + "loss": 0.2028, + "step": 63300 + }, + { + "epoch": 2.43, + "learning_rate": 9.444088227750705e-06, + "loss": 0.2043, + "step": 63301 + }, + { + "epoch": 2.43, + "learning_rate": 9.443447037701976e-06, + "loss": 0.2142, + "step": 63302 + }, + { + "epoch": 2.43, + "learning_rate": 9.442805847653246e-06, + "loss": 0.2836, + "step": 63303 + }, + { + "epoch": 2.43, + "learning_rate": 9.442164657604514e-06, + "loss": 0.157, + "step": 63304 + }, + { + "epoch": 2.43, + "learning_rate": 9.441523467555784e-06, + "loss": 0.1832, + "step": 63305 + }, + { + "epoch": 2.43, + "learning_rate": 9.440882277507052e-06, + "loss": 0.2941, + "step": 63306 + }, + { + "epoch": 2.43, + "learning_rate": 9.440241087458324e-06, + "loss": 0.2304, + "step": 63307 + }, + { + "epoch": 2.43, + "learning_rate": 9.439599897409592e-06, + "loss": 0.2785, + "step": 63308 + }, + { + "epoch": 2.43, + "learning_rate": 9.438958707360862e-06, + "loss": 0.2507, + "step": 63309 + }, + { + "epoch": 2.43, + "learning_rate": 9.43831751731213e-06, + "loss": 0.2681, + "step": 63310 + }, + { + "epoch": 2.43, + "learning_rate": 9.437676327263402e-06, + "loss": 0.2068, + "step": 63311 + }, + { + "epoch": 2.43, + "learning_rate": 9.43703513721467e-06, + "loss": 0.1812, + "step": 63312 + }, + { + "epoch": 2.43, + "learning_rate": 9.43639394716594e-06, + "loss": 0.1899, + "step": 63313 + }, + { + "epoch": 2.43, + "learning_rate": 9.435752757117209e-06, + "loss": 0.2333, + "step": 63314 + }, + { + "epoch": 2.43, + "learning_rate": 9.435111567068479e-06, + "loss": 0.2936, + "step": 63315 + }, + { + "epoch": 2.43, + "learning_rate": 9.434470377019749e-06, + "loss": 0.1533, + "step": 63316 + }, + { + "epoch": 2.43, + "learning_rate": 9.433829186971019e-06, + "loss": 0.1836, + "step": 63317 + }, + { + "epoch": 2.43, + "learning_rate": 9.433187996922287e-06, + "loss": 0.2255, + "step": 63318 + }, + { + "epoch": 2.43, + "learning_rate": 9.432546806873557e-06, + "loss": 0.1463, + "step": 63319 + }, + { + "epoch": 2.43, + "learning_rate": 9.431905616824827e-06, + "loss": 0.1503, + "step": 63320 + }, + { + "epoch": 2.43, + "learning_rate": 9.431264426776097e-06, + "loss": 0.3879, + "step": 63321 + }, + { + "epoch": 2.43, + "learning_rate": 9.430623236727366e-06, + "loss": 0.306, + "step": 63322 + }, + { + "epoch": 2.43, + "learning_rate": 9.429982046678636e-06, + "loss": 0.1314, + "step": 63323 + }, + { + "epoch": 2.43, + "learning_rate": 9.429340856629904e-06, + "loss": 0.1261, + "step": 63324 + }, + { + "epoch": 2.43, + "learning_rate": 9.428699666581176e-06, + "loss": 0.2553, + "step": 63325 + }, + { + "epoch": 2.43, + "learning_rate": 9.428058476532444e-06, + "loss": 0.0958, + "step": 63326 + }, + { + "epoch": 2.43, + "learning_rate": 9.427417286483714e-06, + "loss": 0.1451, + "step": 63327 + }, + { + "epoch": 2.43, + "learning_rate": 9.426776096434983e-06, + "loss": 0.1855, + "step": 63328 + }, + { + "epoch": 2.43, + "learning_rate": 9.426134906386254e-06, + "loss": 0.1942, + "step": 63329 + }, + { + "epoch": 2.43, + "learning_rate": 9.425493716337523e-06, + "loss": 0.2099, + "step": 63330 + }, + { + "epoch": 2.43, + "learning_rate": 9.424852526288793e-06, + "loss": 0.1986, + "step": 63331 + }, + { + "epoch": 2.43, + "learning_rate": 9.424211336240061e-06, + "loss": 0.2271, + "step": 63332 + }, + { + "epoch": 2.43, + "learning_rate": 9.423570146191331e-06, + "loss": 0.3037, + "step": 63333 + }, + { + "epoch": 2.43, + "learning_rate": 9.422928956142601e-06, + "loss": 0.3673, + "step": 63334 + }, + { + "epoch": 2.44, + "learning_rate": 9.422287766093871e-06, + "loss": 0.3405, + "step": 63335 + }, + { + "epoch": 2.44, + "learning_rate": 9.42164657604514e-06, + "loss": 0.2184, + "step": 63336 + }, + { + "epoch": 2.44, + "learning_rate": 9.42100538599641e-06, + "loss": 0.1399, + "step": 63337 + }, + { + "epoch": 2.44, + "learning_rate": 9.42036419594768e-06, + "loss": 0.2148, + "step": 63338 + }, + { + "epoch": 2.44, + "learning_rate": 9.41972300589895e-06, + "loss": 0.1424, + "step": 63339 + }, + { + "epoch": 2.44, + "learning_rate": 9.419081815850218e-06, + "loss": 0.2488, + "step": 63340 + }, + { + "epoch": 2.44, + "learning_rate": 9.418440625801488e-06, + "loss": 0.2693, + "step": 63341 + }, + { + "epoch": 2.44, + "learning_rate": 9.417799435752758e-06, + "loss": 0.3162, + "step": 63342 + }, + { + "epoch": 2.44, + "learning_rate": 9.417158245704028e-06, + "loss": 0.3657, + "step": 63343 + }, + { + "epoch": 2.44, + "learning_rate": 9.416517055655296e-06, + "loss": 0.201, + "step": 63344 + }, + { + "epoch": 2.44, + "learning_rate": 9.415875865606566e-06, + "loss": 0.1326, + "step": 63345 + }, + { + "epoch": 2.44, + "learning_rate": 9.415234675557834e-06, + "loss": 0.2188, + "step": 63346 + }, + { + "epoch": 2.44, + "learning_rate": 9.414593485509106e-06, + "loss": 0.1828, + "step": 63347 + }, + { + "epoch": 2.44, + "learning_rate": 9.413952295460375e-06, + "loss": 0.1451, + "step": 63348 + }, + { + "epoch": 2.44, + "learning_rate": 9.413311105411645e-06, + "loss": 0.1973, + "step": 63349 + }, + { + "epoch": 2.44, + "learning_rate": 9.412669915362913e-06, + "loss": 0.272, + "step": 63350 + }, + { + "epoch": 2.44, + "learning_rate": 9.412028725314185e-06, + "loss": 0.2528, + "step": 63351 + }, + { + "epoch": 2.44, + "learning_rate": 9.411387535265453e-06, + "loss": 0.2548, + "step": 63352 + }, + { + "epoch": 2.44, + "learning_rate": 9.410746345216723e-06, + "loss": 0.197, + "step": 63353 + }, + { + "epoch": 2.44, + "learning_rate": 9.410105155167991e-06, + "loss": 0.3005, + "step": 63354 + }, + { + "epoch": 2.44, + "learning_rate": 9.409463965119261e-06, + "loss": 0.1723, + "step": 63355 + }, + { + "epoch": 2.44, + "learning_rate": 9.408822775070531e-06, + "loss": 0.1409, + "step": 63356 + }, + { + "epoch": 2.44, + "learning_rate": 9.408181585021801e-06, + "loss": 0.3185, + "step": 63357 + }, + { + "epoch": 2.44, + "learning_rate": 9.40754039497307e-06, + "loss": 0.2645, + "step": 63358 + }, + { + "epoch": 2.44, + "learning_rate": 9.40689920492434e-06, + "loss": 0.2748, + "step": 63359 + }, + { + "epoch": 2.44, + "learning_rate": 9.40625801487561e-06, + "loss": 0.313, + "step": 63360 + }, + { + "epoch": 2.44, + "learning_rate": 9.40561682482688e-06, + "loss": 0.1895, + "step": 63361 + }, + { + "epoch": 2.44, + "learning_rate": 9.404975634778148e-06, + "loss": 0.157, + "step": 63362 + }, + { + "epoch": 2.44, + "learning_rate": 9.404334444729418e-06, + "loss": 0.228, + "step": 63363 + }, + { + "epoch": 2.44, + "learning_rate": 9.403693254680686e-06, + "loss": 0.2245, + "step": 63364 + }, + { + "epoch": 2.44, + "learning_rate": 9.403052064631958e-06, + "loss": 0.3033, + "step": 63365 + }, + { + "epoch": 2.44, + "learning_rate": 9.402410874583226e-06, + "loss": 0.2596, + "step": 63366 + }, + { + "epoch": 2.44, + "learning_rate": 9.401769684534496e-06, + "loss": 0.2922, + "step": 63367 + }, + { + "epoch": 2.44, + "learning_rate": 9.401128494485765e-06, + "loss": 0.3339, + "step": 63368 + }, + { + "epoch": 2.44, + "learning_rate": 9.400487304437037e-06, + "loss": 0.2184, + "step": 63369 + }, + { + "epoch": 2.44, + "learning_rate": 9.399846114388305e-06, + "loss": 0.2455, + "step": 63370 + }, + { + "epoch": 2.44, + "learning_rate": 9.399204924339575e-06, + "loss": 0.3385, + "step": 63371 + }, + { + "epoch": 2.44, + "learning_rate": 9.398563734290843e-06, + "loss": 0.2706, + "step": 63372 + }, + { + "epoch": 2.44, + "learning_rate": 9.397922544242113e-06, + "loss": 0.2559, + "step": 63373 + }, + { + "epoch": 2.44, + "learning_rate": 9.397281354193383e-06, + "loss": 0.211, + "step": 63374 + }, + { + "epoch": 2.44, + "learning_rate": 9.396640164144653e-06, + "loss": 0.3195, + "step": 63375 + }, + { + "epoch": 2.44, + "learning_rate": 9.395998974095922e-06, + "loss": 0.2234, + "step": 63376 + }, + { + "epoch": 2.44, + "learning_rate": 9.395357784047192e-06, + "loss": 0.2607, + "step": 63377 + }, + { + "epoch": 2.44, + "learning_rate": 9.394716593998462e-06, + "loss": 0.2117, + "step": 63378 + }, + { + "epoch": 2.44, + "learning_rate": 9.394075403949732e-06, + "loss": 0.301, + "step": 63379 + }, + { + "epoch": 2.44, + "learning_rate": 9.393434213901e-06, + "loss": 0.3053, + "step": 63380 + }, + { + "epoch": 2.44, + "learning_rate": 9.39279302385227e-06, + "loss": 0.2684, + "step": 63381 + }, + { + "epoch": 2.44, + "learning_rate": 9.392151833803538e-06, + "loss": 0.2883, + "step": 63382 + }, + { + "epoch": 2.44, + "learning_rate": 9.39151064375481e-06, + "loss": 0.2004, + "step": 63383 + }, + { + "epoch": 2.44, + "learning_rate": 9.390869453706078e-06, + "loss": 0.2563, + "step": 63384 + }, + { + "epoch": 2.44, + "learning_rate": 9.390228263657348e-06, + "loss": 0.2104, + "step": 63385 + }, + { + "epoch": 2.44, + "learning_rate": 9.389587073608617e-06, + "loss": 0.2056, + "step": 63386 + }, + { + "epoch": 2.44, + "learning_rate": 9.388945883559888e-06, + "loss": 0.3599, + "step": 63387 + }, + { + "epoch": 2.44, + "learning_rate": 9.388304693511157e-06, + "loss": 0.2696, + "step": 63388 + }, + { + "epoch": 2.44, + "learning_rate": 9.387663503462427e-06, + "loss": 0.198, + "step": 63389 + }, + { + "epoch": 2.44, + "learning_rate": 9.387022313413695e-06, + "loss": 0.1998, + "step": 63390 + }, + { + "epoch": 2.44, + "learning_rate": 9.386381123364967e-06, + "loss": 0.2497, + "step": 63391 + }, + { + "epoch": 2.44, + "learning_rate": 9.385739933316235e-06, + "loss": 0.2703, + "step": 63392 + }, + { + "epoch": 2.44, + "learning_rate": 9.385098743267505e-06, + "loss": 0.2584, + "step": 63393 + }, + { + "epoch": 2.44, + "learning_rate": 9.384457553218774e-06, + "loss": 0.235, + "step": 63394 + }, + { + "epoch": 2.44, + "learning_rate": 9.383816363170044e-06, + "loss": 0.1874, + "step": 63395 + }, + { + "epoch": 2.44, + "learning_rate": 9.383175173121314e-06, + "loss": 0.2303, + "step": 63396 + }, + { + "epoch": 2.44, + "learning_rate": 9.382533983072584e-06, + "loss": 0.2467, + "step": 63397 + }, + { + "epoch": 2.44, + "learning_rate": 9.381892793023852e-06, + "loss": 0.2442, + "step": 63398 + }, + { + "epoch": 2.44, + "learning_rate": 9.381251602975122e-06, + "loss": 0.2354, + "step": 63399 + }, + { + "epoch": 2.44, + "learning_rate": 9.380610412926392e-06, + "loss": 0.3112, + "step": 63400 + }, + { + "epoch": 2.44, + "learning_rate": 9.379969222877662e-06, + "loss": 0.1807, + "step": 63401 + }, + { + "epoch": 2.44, + "learning_rate": 9.37932803282893e-06, + "loss": 0.0873, + "step": 63402 + }, + { + "epoch": 2.44, + "learning_rate": 9.3786868427802e-06, + "loss": 0.2504, + "step": 63403 + }, + { + "epoch": 2.44, + "learning_rate": 9.378045652731469e-06, + "loss": 0.2119, + "step": 63404 + }, + { + "epoch": 2.44, + "learning_rate": 9.37740446268274e-06, + "loss": 0.2791, + "step": 63405 + }, + { + "epoch": 2.44, + "learning_rate": 9.376763272634009e-06, + "loss": 0.1547, + "step": 63406 + }, + { + "epoch": 2.44, + "learning_rate": 9.376122082585279e-06, + "loss": 0.2664, + "step": 63407 + }, + { + "epoch": 2.44, + "learning_rate": 9.375480892536547e-06, + "loss": 0.3, + "step": 63408 + }, + { + "epoch": 2.44, + "learning_rate": 9.374839702487819e-06, + "loss": 0.3212, + "step": 63409 + }, + { + "epoch": 2.44, + "learning_rate": 9.374198512439087e-06, + "loss": 0.2544, + "step": 63410 + }, + { + "epoch": 2.44, + "learning_rate": 9.373557322390357e-06, + "loss": 0.2295, + "step": 63411 + }, + { + "epoch": 2.44, + "learning_rate": 9.372916132341625e-06, + "loss": 0.2193, + "step": 63412 + }, + { + "epoch": 2.44, + "learning_rate": 9.372274942292896e-06, + "loss": 0.2426, + "step": 63413 + }, + { + "epoch": 2.44, + "learning_rate": 9.371633752244166e-06, + "loss": 0.1693, + "step": 63414 + }, + { + "epoch": 2.44, + "learning_rate": 9.370992562195436e-06, + "loss": 0.1827, + "step": 63415 + }, + { + "epoch": 2.44, + "learning_rate": 9.370351372146704e-06, + "loss": 0.4039, + "step": 63416 + }, + { + "epoch": 2.44, + "learning_rate": 9.369710182097974e-06, + "loss": 0.1789, + "step": 63417 + }, + { + "epoch": 2.44, + "learning_rate": 9.369068992049244e-06, + "loss": 0.2378, + "step": 63418 + }, + { + "epoch": 2.44, + "learning_rate": 9.368427802000514e-06, + "loss": 0.1666, + "step": 63419 + }, + { + "epoch": 2.44, + "learning_rate": 9.367786611951782e-06, + "loss": 0.1907, + "step": 63420 + }, + { + "epoch": 2.44, + "learning_rate": 9.367145421903052e-06, + "loss": 0.1497, + "step": 63421 + }, + { + "epoch": 2.44, + "learning_rate": 9.36650423185432e-06, + "loss": 0.141, + "step": 63422 + }, + { + "epoch": 2.44, + "learning_rate": 9.365863041805592e-06, + "loss": 0.1805, + "step": 63423 + }, + { + "epoch": 2.44, + "learning_rate": 9.36522185175686e-06, + "loss": 0.2195, + "step": 63424 + }, + { + "epoch": 2.44, + "learning_rate": 9.36458066170813e-06, + "loss": 0.2439, + "step": 63425 + }, + { + "epoch": 2.44, + "learning_rate": 9.363939471659399e-06, + "loss": 0.2521, + "step": 63426 + }, + { + "epoch": 2.44, + "learning_rate": 9.36329828161067e-06, + "loss": 0.2967, + "step": 63427 + }, + { + "epoch": 2.44, + "learning_rate": 9.362657091561939e-06, + "loss": 0.2108, + "step": 63428 + }, + { + "epoch": 2.44, + "learning_rate": 9.362015901513209e-06, + "loss": 0.1734, + "step": 63429 + }, + { + "epoch": 2.44, + "learning_rate": 9.361374711464477e-06, + "loss": 0.2675, + "step": 63430 + }, + { + "epoch": 2.44, + "learning_rate": 9.360733521415747e-06, + "loss": 0.1456, + "step": 63431 + }, + { + "epoch": 2.44, + "learning_rate": 9.360092331367017e-06, + "loss": 0.2476, + "step": 63432 + }, + { + "epoch": 2.44, + "learning_rate": 9.359451141318287e-06, + "loss": 0.2155, + "step": 63433 + }, + { + "epoch": 2.44, + "learning_rate": 9.358809951269556e-06, + "loss": 0.1781, + "step": 63434 + }, + { + "epoch": 2.44, + "learning_rate": 9.358168761220826e-06, + "loss": 0.1555, + "step": 63435 + }, + { + "epoch": 2.44, + "learning_rate": 9.357527571172096e-06, + "loss": 0.2103, + "step": 63436 + }, + { + "epoch": 2.44, + "learning_rate": 9.356886381123366e-06, + "loss": 0.1243, + "step": 63437 + }, + { + "epoch": 2.44, + "learning_rate": 9.356245191074634e-06, + "loss": 0.2389, + "step": 63438 + }, + { + "epoch": 2.44, + "learning_rate": 9.355604001025904e-06, + "loss": 0.2037, + "step": 63439 + }, + { + "epoch": 2.44, + "learning_rate": 9.354962810977174e-06, + "loss": 0.2967, + "step": 63440 + }, + { + "epoch": 2.44, + "learning_rate": 9.354321620928444e-06, + "loss": 0.3387, + "step": 63441 + }, + { + "epoch": 2.44, + "learning_rate": 9.353680430879713e-06, + "loss": 0.1569, + "step": 63442 + }, + { + "epoch": 2.44, + "learning_rate": 9.353039240830983e-06, + "loss": 0.1993, + "step": 63443 + }, + { + "epoch": 2.44, + "learning_rate": 9.352398050782251e-06, + "loss": 0.0884, + "step": 63444 + }, + { + "epoch": 2.44, + "learning_rate": 9.351756860733523e-06, + "loss": 0.2642, + "step": 63445 + }, + { + "epoch": 2.44, + "learning_rate": 9.351115670684791e-06, + "loss": 0.213, + "step": 63446 + }, + { + "epoch": 2.44, + "learning_rate": 9.350474480636061e-06, + "loss": 0.38, + "step": 63447 + }, + { + "epoch": 2.44, + "learning_rate": 9.34983329058733e-06, + "loss": 0.1828, + "step": 63448 + }, + { + "epoch": 2.44, + "learning_rate": 9.349192100538601e-06, + "loss": 0.23, + "step": 63449 + }, + { + "epoch": 2.44, + "learning_rate": 9.34855091048987e-06, + "loss": 0.1328, + "step": 63450 + }, + { + "epoch": 2.44, + "learning_rate": 9.34790972044114e-06, + "loss": 0.1578, + "step": 63451 + }, + { + "epoch": 2.44, + "learning_rate": 9.347268530392408e-06, + "loss": 0.1644, + "step": 63452 + }, + { + "epoch": 2.44, + "learning_rate": 9.346627340343678e-06, + "loss": 0.3237, + "step": 63453 + }, + { + "epoch": 2.44, + "learning_rate": 9.345986150294948e-06, + "loss": 0.2596, + "step": 63454 + }, + { + "epoch": 2.44, + "learning_rate": 9.345344960246218e-06, + "loss": 0.2279, + "step": 63455 + }, + { + "epoch": 2.44, + "learning_rate": 9.344703770197486e-06, + "loss": 0.2246, + "step": 63456 + }, + { + "epoch": 2.44, + "learning_rate": 9.344062580148756e-06, + "loss": 0.2204, + "step": 63457 + }, + { + "epoch": 2.44, + "learning_rate": 9.343421390100026e-06, + "loss": 0.1157, + "step": 63458 + }, + { + "epoch": 2.44, + "learning_rate": 9.342780200051296e-06, + "loss": 0.1446, + "step": 63459 + }, + { + "epoch": 2.44, + "learning_rate": 9.342139010002565e-06, + "loss": 0.404, + "step": 63460 + }, + { + "epoch": 2.44, + "learning_rate": 9.341497819953835e-06, + "loss": 0.2357, + "step": 63461 + }, + { + "epoch": 2.44, + "learning_rate": 9.340856629905103e-06, + "loss": 0.2364, + "step": 63462 + }, + { + "epoch": 2.44, + "learning_rate": 9.340215439856375e-06, + "loss": 0.1929, + "step": 63463 + }, + { + "epoch": 2.44, + "learning_rate": 9.339574249807643e-06, + "loss": 0.2568, + "step": 63464 + }, + { + "epoch": 2.44, + "learning_rate": 9.338933059758913e-06, + "loss": 0.149, + "step": 63465 + }, + { + "epoch": 2.44, + "learning_rate": 9.338291869710181e-06, + "loss": 0.312, + "step": 63466 + }, + { + "epoch": 2.44, + "learning_rate": 9.337650679661453e-06, + "loss": 0.2646, + "step": 63467 + }, + { + "epoch": 2.44, + "learning_rate": 9.337009489612721e-06, + "loss": 0.2448, + "step": 63468 + }, + { + "epoch": 2.44, + "learning_rate": 9.336368299563991e-06, + "loss": 0.192, + "step": 63469 + }, + { + "epoch": 2.44, + "learning_rate": 9.33572710951526e-06, + "loss": 0.1589, + "step": 63470 + }, + { + "epoch": 2.44, + "learning_rate": 9.33508591946653e-06, + "loss": 0.1831, + "step": 63471 + }, + { + "epoch": 2.44, + "learning_rate": 9.3344447294178e-06, + "loss": 0.1466, + "step": 63472 + }, + { + "epoch": 2.44, + "learning_rate": 9.33380353936907e-06, + "loss": 0.3293, + "step": 63473 + }, + { + "epoch": 2.44, + "learning_rate": 9.333162349320338e-06, + "loss": 0.2441, + "step": 63474 + }, + { + "epoch": 2.44, + "learning_rate": 9.332521159271608e-06, + "loss": 0.235, + "step": 63475 + }, + { + "epoch": 2.44, + "learning_rate": 9.331879969222878e-06, + "loss": 0.344, + "step": 63476 + }, + { + "epoch": 2.44, + "learning_rate": 9.331238779174148e-06, + "loss": 0.2403, + "step": 63477 + }, + { + "epoch": 2.44, + "learning_rate": 9.330597589125416e-06, + "loss": 0.2901, + "step": 63478 + }, + { + "epoch": 2.44, + "learning_rate": 9.329956399076686e-06, + "loss": 0.1786, + "step": 63479 + }, + { + "epoch": 2.44, + "learning_rate": 9.329315209027957e-06, + "loss": 0.1462, + "step": 63480 + }, + { + "epoch": 2.44, + "learning_rate": 9.328674018979227e-06, + "loss": 0.2804, + "step": 63481 + }, + { + "epoch": 2.44, + "learning_rate": 9.328032828930495e-06, + "loss": 0.263, + "step": 63482 + }, + { + "epoch": 2.44, + "learning_rate": 9.327391638881765e-06, + "loss": 0.1889, + "step": 63483 + }, + { + "epoch": 2.44, + "learning_rate": 9.326750448833033e-06, + "loss": 0.3268, + "step": 63484 + }, + { + "epoch": 2.44, + "learning_rate": 9.326109258784305e-06, + "loss": 0.2698, + "step": 63485 + }, + { + "epoch": 2.44, + "learning_rate": 9.325468068735573e-06, + "loss": 0.1687, + "step": 63486 + }, + { + "epoch": 2.44, + "learning_rate": 9.324826878686843e-06, + "loss": 0.2839, + "step": 63487 + }, + { + "epoch": 2.44, + "learning_rate": 9.324185688638112e-06, + "loss": 0.1477, + "step": 63488 + }, + { + "epoch": 2.44, + "learning_rate": 9.323544498589383e-06, + "loss": 0.3266, + "step": 63489 + }, + { + "epoch": 2.44, + "learning_rate": 9.322903308540652e-06, + "loss": 0.336, + "step": 63490 + }, + { + "epoch": 2.44, + "learning_rate": 9.322262118491922e-06, + "loss": 0.2106, + "step": 63491 + }, + { + "epoch": 2.44, + "learning_rate": 9.32162092844319e-06, + "loss": 0.1988, + "step": 63492 + }, + { + "epoch": 2.44, + "learning_rate": 9.32097973839446e-06, + "loss": 0.3244, + "step": 63493 + }, + { + "epoch": 2.44, + "learning_rate": 9.32033854834573e-06, + "loss": 0.2223, + "step": 63494 + }, + { + "epoch": 2.44, + "learning_rate": 9.319697358297e-06, + "loss": 0.2053, + "step": 63495 + }, + { + "epoch": 2.44, + "learning_rate": 9.319056168248268e-06, + "loss": 0.4744, + "step": 63496 + }, + { + "epoch": 2.44, + "learning_rate": 9.318414978199538e-06, + "loss": 0.3329, + "step": 63497 + }, + { + "epoch": 2.44, + "learning_rate": 9.317773788150808e-06, + "loss": 0.3817, + "step": 63498 + }, + { + "epoch": 2.44, + "learning_rate": 9.317132598102078e-06, + "loss": 0.2103, + "step": 63499 + }, + { + "epoch": 2.44, + "learning_rate": 9.316491408053347e-06, + "loss": 0.2399, + "step": 63500 + }, + { + "epoch": 2.44, + "learning_rate": 9.315850218004617e-06, + "loss": 0.1334, + "step": 63501 + }, + { + "epoch": 2.44, + "learning_rate": 9.315209027955885e-06, + "loss": 0.1002, + "step": 63502 + }, + { + "epoch": 2.44, + "learning_rate": 9.314567837907157e-06, + "loss": 0.1997, + "step": 63503 + }, + { + "epoch": 2.44, + "learning_rate": 9.313926647858425e-06, + "loss": 0.2887, + "step": 63504 + }, + { + "epoch": 2.44, + "learning_rate": 9.313285457809695e-06, + "loss": 0.3267, + "step": 63505 + }, + { + "epoch": 2.44, + "learning_rate": 9.312644267760964e-06, + "loss": 0.286, + "step": 63506 + }, + { + "epoch": 2.44, + "learning_rate": 9.312003077712235e-06, + "loss": 0.2309, + "step": 63507 + }, + { + "epoch": 2.44, + "learning_rate": 9.311361887663504e-06, + "loss": 0.1682, + "step": 63508 + }, + { + "epoch": 2.44, + "learning_rate": 9.310720697614774e-06, + "loss": 0.3401, + "step": 63509 + }, + { + "epoch": 2.44, + "learning_rate": 9.310079507566042e-06, + "loss": 0.3728, + "step": 63510 + }, + { + "epoch": 2.44, + "learning_rate": 9.309438317517312e-06, + "loss": 0.0978, + "step": 63511 + }, + { + "epoch": 2.44, + "learning_rate": 9.308797127468582e-06, + "loss": 0.3186, + "step": 63512 + }, + { + "epoch": 2.44, + "learning_rate": 9.308155937419852e-06, + "loss": 0.133, + "step": 63513 + }, + { + "epoch": 2.44, + "learning_rate": 9.30751474737112e-06, + "loss": 0.2256, + "step": 63514 + }, + { + "epoch": 2.44, + "learning_rate": 9.30687355732239e-06, + "loss": 0.1552, + "step": 63515 + }, + { + "epoch": 2.44, + "learning_rate": 9.30623236727366e-06, + "loss": 0.2709, + "step": 63516 + }, + { + "epoch": 2.44, + "learning_rate": 9.30559117722493e-06, + "loss": 0.1727, + "step": 63517 + }, + { + "epoch": 2.44, + "learning_rate": 9.304949987176199e-06, + "loss": 0.1227, + "step": 63518 + }, + { + "epoch": 2.44, + "learning_rate": 9.304308797127469e-06, + "loss": 0.211, + "step": 63519 + }, + { + "epoch": 2.44, + "learning_rate": 9.303667607078737e-06, + "loss": 0.2867, + "step": 63520 + }, + { + "epoch": 2.44, + "learning_rate": 9.303026417030009e-06, + "loss": 0.2474, + "step": 63521 + }, + { + "epoch": 2.44, + "learning_rate": 9.302385226981277e-06, + "loss": 0.2451, + "step": 63522 + }, + { + "epoch": 2.44, + "learning_rate": 9.301744036932547e-06, + "loss": 0.2924, + "step": 63523 + }, + { + "epoch": 2.44, + "learning_rate": 9.301102846883815e-06, + "loss": 0.2173, + "step": 63524 + }, + { + "epoch": 2.44, + "learning_rate": 9.300461656835087e-06, + "loss": 0.1362, + "step": 63525 + }, + { + "epoch": 2.44, + "learning_rate": 9.299820466786356e-06, + "loss": 0.1508, + "step": 63526 + }, + { + "epoch": 2.44, + "learning_rate": 9.299179276737626e-06, + "loss": 0.2555, + "step": 63527 + }, + { + "epoch": 2.44, + "learning_rate": 9.298538086688894e-06, + "loss": 0.2766, + "step": 63528 + }, + { + "epoch": 2.44, + "learning_rate": 9.297896896640166e-06, + "loss": 0.1555, + "step": 63529 + }, + { + "epoch": 2.44, + "learning_rate": 9.297255706591434e-06, + "loss": 0.098, + "step": 63530 + }, + { + "epoch": 2.44, + "learning_rate": 9.296614516542704e-06, + "loss": 0.2153, + "step": 63531 + }, + { + "epoch": 2.44, + "learning_rate": 9.295973326493972e-06, + "loss": 0.1561, + "step": 63532 + }, + { + "epoch": 2.44, + "learning_rate": 9.295332136445242e-06, + "loss": 0.2351, + "step": 63533 + }, + { + "epoch": 2.44, + "learning_rate": 9.294690946396512e-06, + "loss": 0.1367, + "step": 63534 + }, + { + "epoch": 2.44, + "learning_rate": 9.294049756347782e-06, + "loss": 0.2711, + "step": 63535 + }, + { + "epoch": 2.44, + "learning_rate": 9.29340856629905e-06, + "loss": 0.0972, + "step": 63536 + }, + { + "epoch": 2.44, + "learning_rate": 9.29276737625032e-06, + "loss": 0.2392, + "step": 63537 + }, + { + "epoch": 2.44, + "learning_rate": 9.29212618620159e-06, + "loss": 0.3995, + "step": 63538 + }, + { + "epoch": 2.44, + "learning_rate": 9.29148499615286e-06, + "loss": 0.2098, + "step": 63539 + }, + { + "epoch": 2.44, + "learning_rate": 9.290843806104129e-06, + "loss": 0.1586, + "step": 63540 + }, + { + "epoch": 2.44, + "learning_rate": 9.290202616055399e-06, + "loss": 0.1487, + "step": 63541 + }, + { + "epoch": 2.44, + "learning_rate": 9.289561426006667e-06, + "loss": 0.1449, + "step": 63542 + }, + { + "epoch": 2.44, + "learning_rate": 9.288920235957939e-06, + "loss": 0.191, + "step": 63543 + }, + { + "epoch": 2.44, + "learning_rate": 9.288279045909207e-06, + "loss": 0.2575, + "step": 63544 + }, + { + "epoch": 2.44, + "learning_rate": 9.287637855860477e-06, + "loss": 0.1552, + "step": 63545 + }, + { + "epoch": 2.44, + "learning_rate": 9.286996665811746e-06, + "loss": 0.2084, + "step": 63546 + }, + { + "epoch": 2.44, + "learning_rate": 9.286355475763018e-06, + "loss": 0.1782, + "step": 63547 + }, + { + "epoch": 2.44, + "learning_rate": 9.285714285714286e-06, + "loss": 0.192, + "step": 63548 + }, + { + "epoch": 2.44, + "learning_rate": 9.285073095665556e-06, + "loss": 0.2135, + "step": 63549 + }, + { + "epoch": 2.44, + "learning_rate": 9.284431905616824e-06, + "loss": 0.3798, + "step": 63550 + }, + { + "epoch": 2.44, + "learning_rate": 9.283790715568094e-06, + "loss": 0.191, + "step": 63551 + }, + { + "epoch": 2.44, + "learning_rate": 9.283149525519364e-06, + "loss": 0.1557, + "step": 63552 + }, + { + "epoch": 2.44, + "learning_rate": 9.282508335470634e-06, + "loss": 0.2641, + "step": 63553 + }, + { + "epoch": 2.44, + "learning_rate": 9.281867145421903e-06, + "loss": 0.2053, + "step": 63554 + }, + { + "epoch": 2.44, + "learning_rate": 9.281225955373173e-06, + "loss": 0.2451, + "step": 63555 + }, + { + "epoch": 2.44, + "learning_rate": 9.280584765324443e-06, + "loss": 0.2057, + "step": 63556 + }, + { + "epoch": 2.44, + "learning_rate": 9.279943575275713e-06, + "loss": 0.257, + "step": 63557 + }, + { + "epoch": 2.44, + "learning_rate": 9.279302385226981e-06, + "loss": 0.1864, + "step": 63558 + }, + { + "epoch": 2.44, + "learning_rate": 9.278661195178251e-06, + "loss": 0.2687, + "step": 63559 + }, + { + "epoch": 2.44, + "learning_rate": 9.27802000512952e-06, + "loss": 0.215, + "step": 63560 + }, + { + "epoch": 2.44, + "learning_rate": 9.277378815080791e-06, + "loss": 0.1472, + "step": 63561 + }, + { + "epoch": 2.44, + "learning_rate": 9.27673762503206e-06, + "loss": 0.2789, + "step": 63562 + }, + { + "epoch": 2.44, + "learning_rate": 9.27609643498333e-06, + "loss": 0.1271, + "step": 63563 + }, + { + "epoch": 2.44, + "learning_rate": 9.275455244934598e-06, + "loss": 0.204, + "step": 63564 + }, + { + "epoch": 2.44, + "learning_rate": 9.27481405488587e-06, + "loss": 0.127, + "step": 63565 + }, + { + "epoch": 2.44, + "learning_rate": 9.274172864837138e-06, + "loss": 0.3167, + "step": 63566 + }, + { + "epoch": 2.44, + "learning_rate": 9.273531674788408e-06, + "loss": 0.249, + "step": 63567 + }, + { + "epoch": 2.44, + "learning_rate": 9.272890484739676e-06, + "loss": 0.1678, + "step": 63568 + }, + { + "epoch": 2.44, + "learning_rate": 9.272249294690946e-06, + "loss": 0.1831, + "step": 63569 + }, + { + "epoch": 2.44, + "learning_rate": 9.271608104642216e-06, + "loss": 0.2219, + "step": 63570 + }, + { + "epoch": 2.44, + "learning_rate": 9.270966914593486e-06, + "loss": 0.2027, + "step": 63571 + }, + { + "epoch": 2.44, + "learning_rate": 9.270325724544755e-06, + "loss": 0.3239, + "step": 63572 + }, + { + "epoch": 2.44, + "learning_rate": 9.269684534496025e-06, + "loss": 0.1877, + "step": 63573 + }, + { + "epoch": 2.44, + "learning_rate": 9.269043344447295e-06, + "loss": 0.2503, + "step": 63574 + }, + { + "epoch": 2.44, + "learning_rate": 9.268402154398565e-06, + "loss": 0.1357, + "step": 63575 + }, + { + "epoch": 2.44, + "learning_rate": 9.267760964349833e-06, + "loss": 0.234, + "step": 63576 + }, + { + "epoch": 2.44, + "learning_rate": 9.267119774301103e-06, + "loss": 0.1483, + "step": 63577 + }, + { + "epoch": 2.44, + "learning_rate": 9.266478584252373e-06, + "loss": 0.1638, + "step": 63578 + }, + { + "epoch": 2.44, + "learning_rate": 9.265837394203643e-06, + "loss": 0.203, + "step": 63579 + }, + { + "epoch": 2.44, + "learning_rate": 9.265196204154911e-06, + "loss": 0.3083, + "step": 63580 + }, + { + "epoch": 2.44, + "learning_rate": 9.264555014106181e-06, + "loss": 0.2165, + "step": 63581 + }, + { + "epoch": 2.44, + "learning_rate": 9.26391382405745e-06, + "loss": 0.1666, + "step": 63582 + }, + { + "epoch": 2.44, + "learning_rate": 9.263272634008721e-06, + "loss": 0.1171, + "step": 63583 + }, + { + "epoch": 2.44, + "learning_rate": 9.26263144395999e-06, + "loss": 0.2105, + "step": 63584 + }, + { + "epoch": 2.44, + "learning_rate": 9.26199025391126e-06, + "loss": 0.1804, + "step": 63585 + }, + { + "epoch": 2.44, + "learning_rate": 9.261349063862528e-06, + "loss": 0.1613, + "step": 63586 + }, + { + "epoch": 2.44, + "learning_rate": 9.2607078738138e-06, + "loss": 0.1741, + "step": 63587 + }, + { + "epoch": 2.44, + "learning_rate": 9.260066683765068e-06, + "loss": 0.1117, + "step": 63588 + }, + { + "epoch": 2.44, + "learning_rate": 9.259425493716338e-06, + "loss": 0.1209, + "step": 63589 + }, + { + "epoch": 2.44, + "learning_rate": 9.258784303667606e-06, + "loss": 0.2226, + "step": 63590 + }, + { + "epoch": 2.44, + "learning_rate": 9.258143113618877e-06, + "loss": 0.2169, + "step": 63591 + }, + { + "epoch": 2.44, + "learning_rate": 9.257501923570147e-06, + "loss": 0.2507, + "step": 63592 + }, + { + "epoch": 2.44, + "learning_rate": 9.256860733521417e-06, + "loss": 0.3892, + "step": 63593 + }, + { + "epoch": 2.44, + "learning_rate": 9.256219543472685e-06, + "loss": 0.2526, + "step": 63594 + }, + { + "epoch": 2.45, + "learning_rate": 9.255578353423955e-06, + "loss": 0.1633, + "step": 63595 + }, + { + "epoch": 2.45, + "learning_rate": 9.254937163375225e-06, + "loss": 0.1884, + "step": 63596 + }, + { + "epoch": 2.45, + "learning_rate": 9.254295973326495e-06, + "loss": 0.3683, + "step": 63597 + }, + { + "epoch": 2.45, + "learning_rate": 9.253654783277763e-06, + "loss": 0.3124, + "step": 63598 + }, + { + "epoch": 2.45, + "learning_rate": 9.253013593229033e-06, + "loss": 0.2155, + "step": 63599 + }, + { + "epoch": 2.45, + "learning_rate": 9.252372403180302e-06, + "loss": 0.2442, + "step": 63600 + }, + { + "epoch": 2.45, + "learning_rate": 9.251731213131573e-06, + "loss": 0.2184, + "step": 63601 + }, + { + "epoch": 2.45, + "learning_rate": 9.251090023082842e-06, + "loss": 0.1963, + "step": 63602 + }, + { + "epoch": 2.45, + "learning_rate": 9.250448833034112e-06, + "loss": 0.2268, + "step": 63603 + }, + { + "epoch": 2.45, + "learning_rate": 9.24980764298538e-06, + "loss": 0.3493, + "step": 63604 + }, + { + "epoch": 2.45, + "learning_rate": 9.249166452936652e-06, + "loss": 0.0848, + "step": 63605 + }, + { + "epoch": 2.45, + "learning_rate": 9.24852526288792e-06, + "loss": 0.1766, + "step": 63606 + }, + { + "epoch": 2.45, + "learning_rate": 9.24788407283919e-06, + "loss": 0.1928, + "step": 63607 + }, + { + "epoch": 2.45, + "learning_rate": 9.247242882790458e-06, + "loss": 0.1679, + "step": 63608 + }, + { + "epoch": 2.45, + "learning_rate": 9.246601692741728e-06, + "loss": 0.277, + "step": 63609 + }, + { + "epoch": 2.45, + "learning_rate": 9.245960502692998e-06, + "loss": 0.2131, + "step": 63610 + }, + { + "epoch": 2.45, + "learning_rate": 9.245319312644268e-06, + "loss": 0.0804, + "step": 63611 + }, + { + "epoch": 2.45, + "learning_rate": 9.244678122595537e-06, + "loss": 0.1229, + "step": 63612 + }, + { + "epoch": 2.45, + "learning_rate": 9.244036932546807e-06, + "loss": 0.2305, + "step": 63613 + }, + { + "epoch": 2.45, + "learning_rate": 9.243395742498077e-06, + "loss": 0.2791, + "step": 63614 + }, + { + "epoch": 2.45, + "learning_rate": 9.242754552449347e-06, + "loss": 0.1283, + "step": 63615 + }, + { + "epoch": 2.45, + "learning_rate": 9.242113362400615e-06, + "loss": 0.281, + "step": 63616 + }, + { + "epoch": 2.45, + "learning_rate": 9.241472172351885e-06, + "loss": 0.2064, + "step": 63617 + }, + { + "epoch": 2.45, + "learning_rate": 9.240830982303155e-06, + "loss": 0.3113, + "step": 63618 + }, + { + "epoch": 2.45, + "learning_rate": 9.240189792254425e-06, + "loss": 0.1989, + "step": 63619 + }, + { + "epoch": 2.45, + "learning_rate": 9.239548602205694e-06, + "loss": 0.2282, + "step": 63620 + }, + { + "epoch": 2.45, + "learning_rate": 9.238907412156964e-06, + "loss": 0.2165, + "step": 63621 + }, + { + "epoch": 2.45, + "learning_rate": 9.238266222108232e-06, + "loss": 0.1638, + "step": 63622 + }, + { + "epoch": 2.45, + "learning_rate": 9.237625032059504e-06, + "loss": 0.207, + "step": 63623 + }, + { + "epoch": 2.45, + "learning_rate": 9.236983842010772e-06, + "loss": 0.2775, + "step": 63624 + }, + { + "epoch": 2.45, + "learning_rate": 9.236342651962042e-06, + "loss": 0.243, + "step": 63625 + }, + { + "epoch": 2.45, + "learning_rate": 9.23570146191331e-06, + "loss": 0.2556, + "step": 63626 + }, + { + "epoch": 2.45, + "learning_rate": 9.235060271864582e-06, + "loss": 0.2793, + "step": 63627 + }, + { + "epoch": 2.45, + "learning_rate": 9.23441908181585e-06, + "loss": 0.2638, + "step": 63628 + }, + { + "epoch": 2.45, + "learning_rate": 9.23377789176712e-06, + "loss": 0.1689, + "step": 63629 + }, + { + "epoch": 2.45, + "learning_rate": 9.233136701718389e-06, + "loss": 0.4343, + "step": 63630 + }, + { + "epoch": 2.45, + "learning_rate": 9.232495511669659e-06, + "loss": 0.3714, + "step": 63631 + }, + { + "epoch": 2.45, + "learning_rate": 9.231854321620929e-06, + "loss": 0.1492, + "step": 63632 + }, + { + "epoch": 2.45, + "learning_rate": 9.231213131572199e-06, + "loss": 0.2303, + "step": 63633 + }, + { + "epoch": 2.45, + "learning_rate": 9.230571941523467e-06, + "loss": 0.1366, + "step": 63634 + }, + { + "epoch": 2.45, + "learning_rate": 9.229930751474737e-06, + "loss": 0.1631, + "step": 63635 + }, + { + "epoch": 2.45, + "learning_rate": 9.229289561426007e-06, + "loss": 0.2098, + "step": 63636 + }, + { + "epoch": 2.45, + "learning_rate": 9.228648371377277e-06, + "loss": 0.1949, + "step": 63637 + }, + { + "epoch": 2.45, + "learning_rate": 9.228007181328546e-06, + "loss": 0.2413, + "step": 63638 + }, + { + "epoch": 2.45, + "learning_rate": 9.227365991279816e-06, + "loss": 0.1718, + "step": 63639 + }, + { + "epoch": 2.45, + "learning_rate": 9.226724801231084e-06, + "loss": 0.2129, + "step": 63640 + }, + { + "epoch": 2.45, + "learning_rate": 9.226083611182356e-06, + "loss": 0.2197, + "step": 63641 + }, + { + "epoch": 2.45, + "learning_rate": 9.225442421133624e-06, + "loss": 0.3736, + "step": 63642 + }, + { + "epoch": 2.45, + "learning_rate": 9.224801231084894e-06, + "loss": 0.1684, + "step": 63643 + }, + { + "epoch": 2.45, + "learning_rate": 9.224160041036162e-06, + "loss": 0.2571, + "step": 63644 + }, + { + "epoch": 2.45, + "learning_rate": 9.223518850987434e-06, + "loss": 0.2445, + "step": 63645 + }, + { + "epoch": 2.45, + "learning_rate": 9.222877660938702e-06, + "loss": 0.1706, + "step": 63646 + }, + { + "epoch": 2.45, + "learning_rate": 9.222236470889972e-06, + "loss": 0.2771, + "step": 63647 + }, + { + "epoch": 2.45, + "learning_rate": 9.22159528084124e-06, + "loss": 0.2504, + "step": 63648 + }, + { + "epoch": 2.45, + "learning_rate": 9.22095409079251e-06, + "loss": 0.1931, + "step": 63649 + }, + { + "epoch": 2.45, + "learning_rate": 9.22031290074378e-06, + "loss": 0.2516, + "step": 63650 + }, + { + "epoch": 2.45, + "learning_rate": 9.21967171069505e-06, + "loss": 0.2272, + "step": 63651 + }, + { + "epoch": 2.45, + "learning_rate": 9.219030520646319e-06, + "loss": 0.3162, + "step": 63652 + }, + { + "epoch": 2.45, + "learning_rate": 9.218389330597589e-06, + "loss": 0.2965, + "step": 63653 + }, + { + "epoch": 2.45, + "learning_rate": 9.217748140548859e-06, + "loss": 0.2042, + "step": 63654 + }, + { + "epoch": 2.45, + "learning_rate": 9.21710695050013e-06, + "loss": 0.3218, + "step": 63655 + }, + { + "epoch": 2.45, + "learning_rate": 9.216465760451397e-06, + "loss": 0.1048, + "step": 63656 + }, + { + "epoch": 2.45, + "learning_rate": 9.215824570402668e-06, + "loss": 0.1797, + "step": 63657 + }, + { + "epoch": 2.45, + "learning_rate": 9.215183380353938e-06, + "loss": 0.1654, + "step": 63658 + }, + { + "epoch": 2.45, + "learning_rate": 9.214542190305208e-06, + "loss": 0.1747, + "step": 63659 + }, + { + "epoch": 2.45, + "learning_rate": 9.213901000256476e-06, + "loss": 0.2559, + "step": 63660 + }, + { + "epoch": 2.45, + "learning_rate": 9.213259810207746e-06, + "loss": 0.1893, + "step": 63661 + }, + { + "epoch": 2.45, + "learning_rate": 9.212618620159014e-06, + "loss": 0.1338, + "step": 63662 + }, + { + "epoch": 2.45, + "learning_rate": 9.211977430110286e-06, + "loss": 0.2018, + "step": 63663 + }, + { + "epoch": 2.45, + "learning_rate": 9.211336240061554e-06, + "loss": 0.1442, + "step": 63664 + }, + { + "epoch": 2.45, + "learning_rate": 9.210695050012824e-06, + "loss": 0.2595, + "step": 63665 + }, + { + "epoch": 2.45, + "learning_rate": 9.210053859964093e-06, + "loss": 0.2797, + "step": 63666 + }, + { + "epoch": 2.45, + "learning_rate": 9.209412669915363e-06, + "loss": 0.3445, + "step": 63667 + }, + { + "epoch": 2.45, + "learning_rate": 9.208771479866633e-06, + "loss": 0.3392, + "step": 63668 + }, + { + "epoch": 2.45, + "learning_rate": 9.208130289817903e-06, + "loss": 0.2309, + "step": 63669 + }, + { + "epoch": 2.45, + "learning_rate": 9.207489099769171e-06, + "loss": 0.0911, + "step": 63670 + }, + { + "epoch": 2.45, + "learning_rate": 9.206847909720441e-06, + "loss": 0.2174, + "step": 63671 + }, + { + "epoch": 2.45, + "learning_rate": 9.206206719671711e-06, + "loss": 0.2209, + "step": 63672 + }, + { + "epoch": 2.45, + "learning_rate": 9.205565529622981e-06, + "loss": 0.3179, + "step": 63673 + }, + { + "epoch": 2.45, + "learning_rate": 9.20492433957425e-06, + "loss": 0.2655, + "step": 63674 + }, + { + "epoch": 2.45, + "learning_rate": 9.20428314952552e-06, + "loss": 0.2633, + "step": 63675 + }, + { + "epoch": 2.45, + "learning_rate": 9.20364195947679e-06, + "loss": 0.3126, + "step": 63676 + }, + { + "epoch": 2.45, + "learning_rate": 9.20300076942806e-06, + "loss": 0.2699, + "step": 63677 + }, + { + "epoch": 2.45, + "learning_rate": 9.202359579379328e-06, + "loss": 0.2297, + "step": 63678 + }, + { + "epoch": 2.45, + "learning_rate": 9.201718389330598e-06, + "loss": 0.1477, + "step": 63679 + }, + { + "epoch": 2.45, + "learning_rate": 9.201077199281866e-06, + "loss": 0.1905, + "step": 63680 + }, + { + "epoch": 2.45, + "learning_rate": 9.200436009233138e-06, + "loss": 0.1378, + "step": 63681 + }, + { + "epoch": 2.45, + "learning_rate": 9.199794819184406e-06, + "loss": 0.1276, + "step": 63682 + }, + { + "epoch": 2.45, + "learning_rate": 9.199153629135676e-06, + "loss": 0.2026, + "step": 63683 + }, + { + "epoch": 2.45, + "learning_rate": 9.198512439086945e-06, + "loss": 0.2481, + "step": 63684 + }, + { + "epoch": 2.45, + "learning_rate": 9.197871249038216e-06, + "loss": 0.1907, + "step": 63685 + }, + { + "epoch": 2.45, + "learning_rate": 9.197230058989485e-06, + "loss": 0.2088, + "step": 63686 + }, + { + "epoch": 2.45, + "learning_rate": 9.196588868940755e-06, + "loss": 0.2672, + "step": 63687 + }, + { + "epoch": 2.45, + "learning_rate": 9.195947678892023e-06, + "loss": 0.1607, + "step": 63688 + }, + { + "epoch": 2.45, + "learning_rate": 9.195306488843293e-06, + "loss": 0.2318, + "step": 63689 + }, + { + "epoch": 2.45, + "learning_rate": 9.194665298794563e-06, + "loss": 0.1901, + "step": 63690 + }, + { + "epoch": 2.45, + "learning_rate": 9.194024108745833e-06, + "loss": 0.2232, + "step": 63691 + }, + { + "epoch": 2.45, + "learning_rate": 9.193382918697101e-06, + "loss": 0.2651, + "step": 63692 + }, + { + "epoch": 2.45, + "learning_rate": 9.192741728648371e-06, + "loss": 0.29, + "step": 63693 + }, + { + "epoch": 2.45, + "learning_rate": 9.192100538599641e-06, + "loss": 0.2366, + "step": 63694 + }, + { + "epoch": 2.45, + "learning_rate": 9.191459348550911e-06, + "loss": 0.169, + "step": 63695 + }, + { + "epoch": 2.45, + "learning_rate": 9.19081815850218e-06, + "loss": 0.1516, + "step": 63696 + }, + { + "epoch": 2.45, + "learning_rate": 9.19017696845345e-06, + "loss": 0.3953, + "step": 63697 + }, + { + "epoch": 2.45, + "learning_rate": 9.18953577840472e-06, + "loss": 0.2383, + "step": 63698 + }, + { + "epoch": 2.45, + "learning_rate": 9.18889458835599e-06, + "loss": 0.2451, + "step": 63699 + }, + { + "epoch": 2.45, + "learning_rate": 9.188253398307258e-06, + "loss": 0.3162, + "step": 63700 + }, + { + "epoch": 2.45, + "learning_rate": 9.187612208258528e-06, + "loss": 0.1912, + "step": 63701 + }, + { + "epoch": 2.45, + "learning_rate": 9.186971018209796e-06, + "loss": 0.1895, + "step": 63702 + }, + { + "epoch": 2.45, + "learning_rate": 9.186329828161068e-06, + "loss": 0.131, + "step": 63703 + }, + { + "epoch": 2.45, + "learning_rate": 9.185688638112337e-06, + "loss": 0.1898, + "step": 63704 + }, + { + "epoch": 2.45, + "learning_rate": 9.185047448063607e-06, + "loss": 0.0991, + "step": 63705 + }, + { + "epoch": 2.45, + "learning_rate": 9.184406258014875e-06, + "loss": 0.1646, + "step": 63706 + }, + { + "epoch": 2.45, + "learning_rate": 9.183765067966145e-06, + "loss": 0.1611, + "step": 63707 + }, + { + "epoch": 2.45, + "learning_rate": 9.183123877917415e-06, + "loss": 0.144, + "step": 63708 + }, + { + "epoch": 2.45, + "learning_rate": 9.182482687868685e-06, + "loss": 0.1949, + "step": 63709 + }, + { + "epoch": 2.45, + "learning_rate": 9.181841497819953e-06, + "loss": 0.1599, + "step": 63710 + }, + { + "epoch": 2.45, + "learning_rate": 9.181200307771223e-06, + "loss": 0.2294, + "step": 63711 + }, + { + "epoch": 2.45, + "learning_rate": 9.180559117722493e-06, + "loss": 0.3077, + "step": 63712 + }, + { + "epoch": 2.45, + "learning_rate": 9.179917927673763e-06, + "loss": 0.1812, + "step": 63713 + }, + { + "epoch": 2.45, + "learning_rate": 9.179276737625032e-06, + "loss": 0.197, + "step": 63714 + }, + { + "epoch": 2.45, + "learning_rate": 9.178635547576302e-06, + "loss": 0.2281, + "step": 63715 + }, + { + "epoch": 2.45, + "learning_rate": 9.177994357527572e-06, + "loss": 0.1454, + "step": 63716 + }, + { + "epoch": 2.45, + "learning_rate": 9.177353167478842e-06, + "loss": 0.3191, + "step": 63717 + }, + { + "epoch": 2.45, + "learning_rate": 9.17671197743011e-06, + "loss": 0.1527, + "step": 63718 + }, + { + "epoch": 2.45, + "learning_rate": 9.17607078738138e-06, + "loss": 0.1648, + "step": 63719 + }, + { + "epoch": 2.45, + "learning_rate": 9.17542959733265e-06, + "loss": 0.4069, + "step": 63720 + }, + { + "epoch": 2.45, + "learning_rate": 9.17478840728392e-06, + "loss": 0.1867, + "step": 63721 + }, + { + "epoch": 2.45, + "learning_rate": 9.174147217235188e-06, + "loss": 0.2117, + "step": 63722 + }, + { + "epoch": 2.45, + "learning_rate": 9.173506027186459e-06, + "loss": 0.2027, + "step": 63723 + }, + { + "epoch": 2.45, + "learning_rate": 9.172864837137727e-06, + "loss": 0.2161, + "step": 63724 + }, + { + "epoch": 2.45, + "learning_rate": 9.172223647088999e-06, + "loss": 0.2052, + "step": 63725 + }, + { + "epoch": 2.45, + "learning_rate": 9.171582457040267e-06, + "loss": 0.123, + "step": 63726 + }, + { + "epoch": 2.45, + "learning_rate": 9.170941266991537e-06, + "loss": 0.321, + "step": 63727 + }, + { + "epoch": 2.45, + "learning_rate": 9.170300076942805e-06, + "loss": 0.316, + "step": 63728 + }, + { + "epoch": 2.45, + "learning_rate": 9.169658886894075e-06, + "loss": 0.1106, + "step": 63729 + }, + { + "epoch": 2.45, + "learning_rate": 9.169017696845345e-06, + "loss": 0.2335, + "step": 63730 + }, + { + "epoch": 2.45, + "learning_rate": 9.168376506796615e-06, + "loss": 0.1193, + "step": 63731 + }, + { + "epoch": 2.45, + "learning_rate": 9.167735316747884e-06, + "loss": 0.2, + "step": 63732 + }, + { + "epoch": 2.45, + "learning_rate": 9.167094126699154e-06, + "loss": 0.2412, + "step": 63733 + }, + { + "epoch": 2.45, + "learning_rate": 9.166452936650424e-06, + "loss": 0.1922, + "step": 63734 + }, + { + "epoch": 2.45, + "learning_rate": 9.165811746601694e-06, + "loss": 0.1893, + "step": 63735 + }, + { + "epoch": 2.45, + "learning_rate": 9.165170556552962e-06, + "loss": 0.1776, + "step": 63736 + }, + { + "epoch": 2.45, + "learning_rate": 9.164529366504232e-06, + "loss": 0.2333, + "step": 63737 + }, + { + "epoch": 2.45, + "learning_rate": 9.163888176455502e-06, + "loss": 0.4377, + "step": 63738 + }, + { + "epoch": 2.45, + "learning_rate": 9.163246986406772e-06, + "loss": 0.2338, + "step": 63739 + }, + { + "epoch": 2.45, + "learning_rate": 9.16260579635804e-06, + "loss": 0.2069, + "step": 63740 + }, + { + "epoch": 2.45, + "learning_rate": 9.16196460630931e-06, + "loss": 0.2126, + "step": 63741 + }, + { + "epoch": 2.45, + "learning_rate": 9.161323416260579e-06, + "loss": 0.1398, + "step": 63742 + }, + { + "epoch": 2.45, + "learning_rate": 9.16068222621185e-06, + "loss": 0.2093, + "step": 63743 + }, + { + "epoch": 2.45, + "learning_rate": 9.160041036163119e-06, + "loss": 0.2128, + "step": 63744 + }, + { + "epoch": 2.45, + "learning_rate": 9.159399846114389e-06, + "loss": 0.1722, + "step": 63745 + }, + { + "epoch": 2.45, + "learning_rate": 9.158758656065657e-06, + "loss": 0.2717, + "step": 63746 + }, + { + "epoch": 2.45, + "learning_rate": 9.158117466016927e-06, + "loss": 0.1659, + "step": 63747 + }, + { + "epoch": 2.45, + "learning_rate": 9.157476275968197e-06, + "loss": 0.2415, + "step": 63748 + }, + { + "epoch": 2.45, + "learning_rate": 9.156835085919467e-06, + "loss": 0.1453, + "step": 63749 + }, + { + "epoch": 2.45, + "learning_rate": 9.156193895870736e-06, + "loss": 0.1939, + "step": 63750 + }, + { + "epoch": 2.45, + "learning_rate": 9.155552705822006e-06, + "loss": 0.12, + "step": 63751 + }, + { + "epoch": 2.45, + "learning_rate": 9.154911515773276e-06, + "loss": 0.2426, + "step": 63752 + }, + { + "epoch": 2.45, + "learning_rate": 9.154270325724546e-06, + "loss": 0.4051, + "step": 63753 + }, + { + "epoch": 2.45, + "learning_rate": 9.153629135675814e-06, + "loss": 0.2018, + "step": 63754 + }, + { + "epoch": 2.45, + "learning_rate": 9.152987945627084e-06, + "loss": 0.3555, + "step": 63755 + }, + { + "epoch": 2.45, + "learning_rate": 9.152346755578354e-06, + "loss": 0.2019, + "step": 63756 + }, + { + "epoch": 2.45, + "learning_rate": 9.151705565529624e-06, + "loss": 0.1903, + "step": 63757 + }, + { + "epoch": 2.45, + "learning_rate": 9.151064375480892e-06, + "loss": 0.3056, + "step": 63758 + }, + { + "epoch": 2.45, + "learning_rate": 9.150423185432162e-06, + "loss": 0.1422, + "step": 63759 + }, + { + "epoch": 2.45, + "learning_rate": 9.149781995383432e-06, + "loss": 0.3066, + "step": 63760 + }, + { + "epoch": 2.45, + "learning_rate": 9.149140805334702e-06, + "loss": 0.1824, + "step": 63761 + }, + { + "epoch": 2.45, + "learning_rate": 9.14849961528597e-06, + "loss": 0.1278, + "step": 63762 + }, + { + "epoch": 2.45, + "learning_rate": 9.14785842523724e-06, + "loss": 0.1484, + "step": 63763 + }, + { + "epoch": 2.45, + "learning_rate": 9.147217235188509e-06, + "loss": 0.1362, + "step": 63764 + }, + { + "epoch": 2.45, + "learning_rate": 9.14657604513978e-06, + "loss": 0.2772, + "step": 63765 + }, + { + "epoch": 2.45, + "learning_rate": 9.145934855091049e-06, + "loss": 0.3812, + "step": 63766 + }, + { + "epoch": 2.45, + "learning_rate": 9.14529366504232e-06, + "loss": 0.2599, + "step": 63767 + }, + { + "epoch": 2.45, + "learning_rate": 9.144652474993587e-06, + "loss": 0.349, + "step": 63768 + }, + { + "epoch": 2.45, + "learning_rate": 9.144011284944858e-06, + "loss": 0.136, + "step": 63769 + }, + { + "epoch": 2.45, + "learning_rate": 9.143370094896128e-06, + "loss": 0.2272, + "step": 63770 + }, + { + "epoch": 2.45, + "learning_rate": 9.142728904847398e-06, + "loss": 0.2711, + "step": 63771 + }, + { + "epoch": 2.45, + "learning_rate": 9.142087714798666e-06, + "loss": 0.1631, + "step": 63772 + }, + { + "epoch": 2.45, + "learning_rate": 9.141446524749936e-06, + "loss": 0.265, + "step": 63773 + }, + { + "epoch": 2.45, + "learning_rate": 9.140805334701206e-06, + "loss": 0.1852, + "step": 63774 + }, + { + "epoch": 2.45, + "learning_rate": 9.140164144652476e-06, + "loss": 0.2133, + "step": 63775 + }, + { + "epoch": 2.45, + "learning_rate": 9.139522954603744e-06, + "loss": 0.2356, + "step": 63776 + }, + { + "epoch": 2.45, + "learning_rate": 9.138881764555014e-06, + "loss": 0.3502, + "step": 63777 + }, + { + "epoch": 2.45, + "learning_rate": 9.138240574506284e-06, + "loss": 0.3516, + "step": 63778 + }, + { + "epoch": 2.45, + "learning_rate": 9.137599384457554e-06, + "loss": 0.205, + "step": 63779 + }, + { + "epoch": 2.45, + "learning_rate": 9.136958194408823e-06, + "loss": 0.2246, + "step": 63780 + }, + { + "epoch": 2.45, + "learning_rate": 9.136317004360093e-06, + "loss": 0.234, + "step": 63781 + }, + { + "epoch": 2.45, + "learning_rate": 9.135675814311363e-06, + "loss": 0.14, + "step": 63782 + }, + { + "epoch": 2.45, + "learning_rate": 9.135034624262633e-06, + "loss": 0.2806, + "step": 63783 + }, + { + "epoch": 2.45, + "learning_rate": 9.134393434213901e-06, + "loss": 0.2531, + "step": 63784 + }, + { + "epoch": 2.45, + "learning_rate": 9.133752244165171e-06, + "loss": 0.2232, + "step": 63785 + }, + { + "epoch": 2.45, + "learning_rate": 9.13311105411644e-06, + "loss": 0.1611, + "step": 63786 + }, + { + "epoch": 2.45, + "learning_rate": 9.13246986406771e-06, + "loss": 0.1363, + "step": 63787 + }, + { + "epoch": 2.45, + "learning_rate": 9.13182867401898e-06, + "loss": 0.2795, + "step": 63788 + }, + { + "epoch": 2.45, + "learning_rate": 9.13118748397025e-06, + "loss": 0.2616, + "step": 63789 + }, + { + "epoch": 2.45, + "learning_rate": 9.130546293921518e-06, + "loss": 0.1871, + "step": 63790 + }, + { + "epoch": 2.45, + "learning_rate": 9.129905103872788e-06, + "loss": 0.3236, + "step": 63791 + }, + { + "epoch": 2.45, + "learning_rate": 9.129263913824058e-06, + "loss": 0.2324, + "step": 63792 + }, + { + "epoch": 2.45, + "learning_rate": 9.128622723775328e-06, + "loss": 0.3917, + "step": 63793 + }, + { + "epoch": 2.45, + "learning_rate": 9.127981533726596e-06, + "loss": 0.2809, + "step": 63794 + }, + { + "epoch": 2.45, + "learning_rate": 9.127340343677866e-06, + "loss": 0.1157, + "step": 63795 + }, + { + "epoch": 2.45, + "learning_rate": 9.126699153629136e-06, + "loss": 0.208, + "step": 63796 + }, + { + "epoch": 2.45, + "learning_rate": 9.126057963580406e-06, + "loss": 0.2485, + "step": 63797 + }, + { + "epoch": 2.45, + "learning_rate": 9.125416773531675e-06, + "loss": 0.2103, + "step": 63798 + }, + { + "epoch": 2.45, + "learning_rate": 9.124775583482945e-06, + "loss": 0.2331, + "step": 63799 + }, + { + "epoch": 2.45, + "learning_rate": 9.124134393434215e-06, + "loss": 0.2307, + "step": 63800 + }, + { + "epoch": 2.45, + "learning_rate": 9.123493203385485e-06, + "loss": 0.2184, + "step": 63801 + }, + { + "epoch": 2.45, + "learning_rate": 9.122852013336753e-06, + "loss": 0.2213, + "step": 63802 + }, + { + "epoch": 2.45, + "learning_rate": 9.122210823288023e-06, + "loss": 0.1509, + "step": 63803 + }, + { + "epoch": 2.45, + "learning_rate": 9.121569633239291e-06, + "loss": 0.3087, + "step": 63804 + }, + { + "epoch": 2.45, + "learning_rate": 9.120928443190561e-06, + "loss": 0.1941, + "step": 63805 + }, + { + "epoch": 2.45, + "learning_rate": 9.120287253141831e-06, + "loss": 0.3514, + "step": 63806 + }, + { + "epoch": 2.45, + "learning_rate": 9.119646063093101e-06, + "loss": 0.3253, + "step": 63807 + }, + { + "epoch": 2.45, + "learning_rate": 9.11900487304437e-06, + "loss": 0.1959, + "step": 63808 + }, + { + "epoch": 2.45, + "learning_rate": 9.11836368299564e-06, + "loss": 0.1677, + "step": 63809 + }, + { + "epoch": 2.45, + "learning_rate": 9.11772249294691e-06, + "loss": 0.2284, + "step": 63810 + }, + { + "epoch": 2.45, + "learning_rate": 9.11708130289818e-06, + "loss": 0.2775, + "step": 63811 + }, + { + "epoch": 2.45, + "learning_rate": 9.116440112849448e-06, + "loss": 0.2394, + "step": 63812 + }, + { + "epoch": 2.45, + "learning_rate": 9.115798922800718e-06, + "loss": 0.3125, + "step": 63813 + }, + { + "epoch": 2.45, + "learning_rate": 9.115157732751988e-06, + "loss": 0.1872, + "step": 63814 + }, + { + "epoch": 2.45, + "learning_rate": 9.114516542703258e-06, + "loss": 0.1727, + "step": 63815 + }, + { + "epoch": 2.45, + "learning_rate": 9.113875352654527e-06, + "loss": 0.2834, + "step": 63816 + }, + { + "epoch": 2.45, + "learning_rate": 9.113234162605797e-06, + "loss": 0.205, + "step": 63817 + }, + { + "epoch": 2.45, + "learning_rate": 9.112592972557067e-06, + "loss": 0.2403, + "step": 63818 + }, + { + "epoch": 2.45, + "learning_rate": 9.111951782508337e-06, + "loss": 0.1468, + "step": 63819 + }, + { + "epoch": 2.45, + "learning_rate": 9.111310592459605e-06, + "loss": 0.2347, + "step": 63820 + }, + { + "epoch": 2.45, + "learning_rate": 9.110669402410875e-06, + "loss": 0.1437, + "step": 63821 + }, + { + "epoch": 2.45, + "learning_rate": 9.110028212362145e-06, + "loss": 0.2114, + "step": 63822 + }, + { + "epoch": 2.45, + "learning_rate": 9.109387022313415e-06, + "loss": 0.1225, + "step": 63823 + }, + { + "epoch": 2.45, + "learning_rate": 9.108745832264683e-06, + "loss": 0.1869, + "step": 63824 + }, + { + "epoch": 2.45, + "learning_rate": 9.108104642215953e-06, + "loss": 0.2416, + "step": 63825 + }, + { + "epoch": 2.45, + "learning_rate": 9.107463452167222e-06, + "loss": 0.2334, + "step": 63826 + }, + { + "epoch": 2.45, + "learning_rate": 9.106822262118492e-06, + "loss": 0.1864, + "step": 63827 + }, + { + "epoch": 2.45, + "learning_rate": 9.106181072069762e-06, + "loss": 0.1993, + "step": 63828 + }, + { + "epoch": 2.45, + "learning_rate": 9.105539882021032e-06, + "loss": 0.3234, + "step": 63829 + }, + { + "epoch": 2.45, + "learning_rate": 9.1048986919723e-06, + "loss": 0.2577, + "step": 63830 + }, + { + "epoch": 2.45, + "learning_rate": 9.10425750192357e-06, + "loss": 0.3103, + "step": 63831 + }, + { + "epoch": 2.45, + "learning_rate": 9.10361631187484e-06, + "loss": 0.1946, + "step": 63832 + }, + { + "epoch": 2.45, + "learning_rate": 9.10297512182611e-06, + "loss": 0.1965, + "step": 63833 + }, + { + "epoch": 2.45, + "learning_rate": 9.102333931777378e-06, + "loss": 0.1546, + "step": 63834 + }, + { + "epoch": 2.45, + "learning_rate": 9.101692741728649e-06, + "loss": 0.1467, + "step": 63835 + }, + { + "epoch": 2.45, + "learning_rate": 9.101051551679919e-06, + "loss": 0.2921, + "step": 63836 + }, + { + "epoch": 2.45, + "learning_rate": 9.100410361631189e-06, + "loss": 0.2248, + "step": 63837 + }, + { + "epoch": 2.45, + "learning_rate": 9.099769171582457e-06, + "loss": 0.2981, + "step": 63838 + }, + { + "epoch": 2.45, + "learning_rate": 9.099127981533727e-06, + "loss": 0.1752, + "step": 63839 + }, + { + "epoch": 2.45, + "learning_rate": 9.098486791484997e-06, + "loss": 0.217, + "step": 63840 + }, + { + "epoch": 2.45, + "learning_rate": 9.097845601436267e-06, + "loss": 0.2038, + "step": 63841 + }, + { + "epoch": 2.45, + "learning_rate": 9.097204411387535e-06, + "loss": 0.2497, + "step": 63842 + }, + { + "epoch": 2.45, + "learning_rate": 9.096563221338805e-06, + "loss": 0.256, + "step": 63843 + }, + { + "epoch": 2.45, + "learning_rate": 9.095922031290074e-06, + "loss": 0.2604, + "step": 63844 + }, + { + "epoch": 2.45, + "learning_rate": 9.095280841241344e-06, + "loss": 0.1724, + "step": 63845 + }, + { + "epoch": 2.45, + "learning_rate": 9.094639651192614e-06, + "loss": 0.1382, + "step": 63846 + }, + { + "epoch": 2.45, + "learning_rate": 9.093998461143884e-06, + "loss": 0.1654, + "step": 63847 + }, + { + "epoch": 2.45, + "learning_rate": 9.093357271095152e-06, + "loss": 0.2131, + "step": 63848 + }, + { + "epoch": 2.45, + "learning_rate": 9.092716081046422e-06, + "loss": 0.1541, + "step": 63849 + }, + { + "epoch": 2.45, + "learning_rate": 9.092074890997692e-06, + "loss": 0.2991, + "step": 63850 + }, + { + "epoch": 2.45, + "learning_rate": 9.091433700948962e-06, + "loss": 0.1947, + "step": 63851 + }, + { + "epoch": 2.45, + "learning_rate": 9.09079251090023e-06, + "loss": 0.2569, + "step": 63852 + }, + { + "epoch": 2.45, + "learning_rate": 9.0901513208515e-06, + "loss": 0.172, + "step": 63853 + }, + { + "epoch": 2.45, + "learning_rate": 9.08951013080277e-06, + "loss": 0.1139, + "step": 63854 + }, + { + "epoch": 2.46, + "learning_rate": 9.08886894075404e-06, + "loss": 0.1526, + "step": 63855 + }, + { + "epoch": 2.46, + "learning_rate": 9.088227750705309e-06, + "loss": 0.1667, + "step": 63856 + }, + { + "epoch": 2.46, + "learning_rate": 9.087586560656579e-06, + "loss": 0.249, + "step": 63857 + }, + { + "epoch": 2.46, + "learning_rate": 9.086945370607849e-06, + "loss": 0.1364, + "step": 63858 + }, + { + "epoch": 2.46, + "learning_rate": 9.086304180559119e-06, + "loss": 0.181, + "step": 63859 + }, + { + "epoch": 2.46, + "learning_rate": 9.085662990510387e-06, + "loss": 0.205, + "step": 63860 + }, + { + "epoch": 2.46, + "learning_rate": 9.085021800461657e-06, + "loss": 0.0979, + "step": 63861 + }, + { + "epoch": 2.46, + "learning_rate": 9.084380610412927e-06, + "loss": 0.2278, + "step": 63862 + }, + { + "epoch": 2.46, + "learning_rate": 9.083739420364197e-06, + "loss": 0.1239, + "step": 63863 + }, + { + "epoch": 2.46, + "learning_rate": 9.083098230315466e-06, + "loss": 0.1769, + "step": 63864 + }, + { + "epoch": 2.46, + "learning_rate": 9.082457040266736e-06, + "loss": 0.2265, + "step": 63865 + }, + { + "epoch": 2.46, + "learning_rate": 9.081815850218004e-06, + "loss": 0.1786, + "step": 63866 + }, + { + "epoch": 2.46, + "learning_rate": 9.081174660169274e-06, + "loss": 0.2081, + "step": 63867 + }, + { + "epoch": 2.46, + "learning_rate": 9.080533470120544e-06, + "loss": 0.2476, + "step": 63868 + }, + { + "epoch": 2.46, + "learning_rate": 9.079892280071814e-06, + "loss": 0.0997, + "step": 63869 + }, + { + "epoch": 2.46, + "learning_rate": 9.079251090023082e-06, + "loss": 0.3583, + "step": 63870 + }, + { + "epoch": 2.46, + "learning_rate": 9.078609899974352e-06, + "loss": 0.1648, + "step": 63871 + }, + { + "epoch": 2.46, + "learning_rate": 9.077968709925622e-06, + "loss": 0.1886, + "step": 63872 + }, + { + "epoch": 2.46, + "learning_rate": 9.077327519876892e-06, + "loss": 0.1721, + "step": 63873 + }, + { + "epoch": 2.46, + "learning_rate": 9.07668632982816e-06, + "loss": 0.3248, + "step": 63874 + }, + { + "epoch": 2.46, + "learning_rate": 9.07604513977943e-06, + "loss": 0.1195, + "step": 63875 + }, + { + "epoch": 2.46, + "learning_rate": 9.0754039497307e-06, + "loss": 0.1642, + "step": 63876 + }, + { + "epoch": 2.46, + "learning_rate": 9.07476275968197e-06, + "loss": 0.1734, + "step": 63877 + }, + { + "epoch": 2.46, + "learning_rate": 9.074121569633239e-06, + "loss": 0.1422, + "step": 63878 + }, + { + "epoch": 2.46, + "learning_rate": 9.07348037958451e-06, + "loss": 0.2536, + "step": 63879 + }, + { + "epoch": 2.46, + "learning_rate": 9.07283918953578e-06, + "loss": 0.2197, + "step": 63880 + }, + { + "epoch": 2.46, + "learning_rate": 9.07219799948705e-06, + "loss": 0.3303, + "step": 63881 + }, + { + "epoch": 2.46, + "learning_rate": 9.071556809438318e-06, + "loss": 0.3134, + "step": 63882 + }, + { + "epoch": 2.46, + "learning_rate": 9.070915619389588e-06, + "loss": 0.1323, + "step": 63883 + }, + { + "epoch": 2.46, + "learning_rate": 9.070274429340858e-06, + "loss": 0.1704, + "step": 63884 + }, + { + "epoch": 2.46, + "learning_rate": 9.069633239292126e-06, + "loss": 0.1169, + "step": 63885 + }, + { + "epoch": 2.46, + "learning_rate": 9.068992049243396e-06, + "loss": 0.3035, + "step": 63886 + }, + { + "epoch": 2.46, + "learning_rate": 9.068350859194666e-06, + "loss": 0.2118, + "step": 63887 + }, + { + "epoch": 2.46, + "learning_rate": 9.067709669145934e-06, + "loss": 0.1896, + "step": 63888 + }, + { + "epoch": 2.46, + "learning_rate": 9.067068479097204e-06, + "loss": 0.221, + "step": 63889 + }, + { + "epoch": 2.46, + "learning_rate": 9.066427289048474e-06, + "loss": 0.2846, + "step": 63890 + }, + { + "epoch": 2.46, + "learning_rate": 9.065786098999744e-06, + "loss": 0.2038, + "step": 63891 + }, + { + "epoch": 2.46, + "learning_rate": 9.065144908951013e-06, + "loss": 0.0998, + "step": 63892 + }, + { + "epoch": 2.46, + "learning_rate": 9.064503718902283e-06, + "loss": 0.267, + "step": 63893 + }, + { + "epoch": 2.46, + "learning_rate": 9.063862528853553e-06, + "loss": 0.2003, + "step": 63894 + }, + { + "epoch": 2.46, + "learning_rate": 9.063221338804823e-06, + "loss": 0.2326, + "step": 63895 + }, + { + "epoch": 2.46, + "learning_rate": 9.062580148756091e-06, + "loss": 0.405, + "step": 63896 + }, + { + "epoch": 2.46, + "learning_rate": 9.061938958707361e-06, + "loss": 0.4409, + "step": 63897 + }, + { + "epoch": 2.46, + "learning_rate": 9.061297768658631e-06, + "loss": 0.1098, + "step": 63898 + }, + { + "epoch": 2.46, + "learning_rate": 9.060656578609901e-06, + "loss": 0.2451, + "step": 63899 + }, + { + "epoch": 2.46, + "learning_rate": 9.06001538856117e-06, + "loss": 0.2843, + "step": 63900 + }, + { + "epoch": 2.46, + "learning_rate": 9.05937419851244e-06, + "loss": 0.3137, + "step": 63901 + }, + { + "epoch": 2.46, + "learning_rate": 9.05873300846371e-06, + "loss": 0.1625, + "step": 63902 + }, + { + "epoch": 2.46, + "learning_rate": 9.058091818414978e-06, + "loss": 0.2847, + "step": 63903 + }, + { + "epoch": 2.46, + "learning_rate": 9.057450628366248e-06, + "loss": 0.4082, + "step": 63904 + }, + { + "epoch": 2.46, + "learning_rate": 9.056809438317518e-06, + "loss": 0.1563, + "step": 63905 + }, + { + "epoch": 2.46, + "learning_rate": 9.056168248268786e-06, + "loss": 0.1376, + "step": 63906 + }, + { + "epoch": 2.46, + "learning_rate": 9.055527058220056e-06, + "loss": 0.2631, + "step": 63907 + }, + { + "epoch": 2.46, + "learning_rate": 9.054885868171326e-06, + "loss": 0.2116, + "step": 63908 + }, + { + "epoch": 2.46, + "learning_rate": 9.054244678122596e-06, + "loss": 0.1465, + "step": 63909 + }, + { + "epoch": 2.46, + "learning_rate": 9.053603488073865e-06, + "loss": 0.2333, + "step": 63910 + }, + { + "epoch": 2.46, + "learning_rate": 9.052962298025135e-06, + "loss": 0.1383, + "step": 63911 + }, + { + "epoch": 2.46, + "learning_rate": 9.052321107976405e-06, + "loss": 0.2499, + "step": 63912 + }, + { + "epoch": 2.46, + "learning_rate": 9.051679917927675e-06, + "loss": 0.2485, + "step": 63913 + }, + { + "epoch": 2.46, + "learning_rate": 9.051038727878943e-06, + "loss": 0.3315, + "step": 63914 + }, + { + "epoch": 2.46, + "learning_rate": 9.050397537830213e-06, + "loss": 0.2895, + "step": 63915 + }, + { + "epoch": 2.46, + "learning_rate": 9.049756347781483e-06, + "loss": 0.2083, + "step": 63916 + }, + { + "epoch": 2.46, + "learning_rate": 9.049115157732753e-06, + "loss": 0.3061, + "step": 63917 + }, + { + "epoch": 2.46, + "learning_rate": 9.048473967684021e-06, + "loss": 0.1147, + "step": 63918 + }, + { + "epoch": 2.46, + "learning_rate": 9.047832777635291e-06, + "loss": 0.3483, + "step": 63919 + }, + { + "epoch": 2.46, + "learning_rate": 9.047191587586561e-06, + "loss": 0.257, + "step": 63920 + }, + { + "epoch": 2.46, + "learning_rate": 9.046550397537831e-06, + "loss": 0.0832, + "step": 63921 + }, + { + "epoch": 2.46, + "learning_rate": 9.0459092074891e-06, + "loss": 0.2231, + "step": 63922 + }, + { + "epoch": 2.46, + "learning_rate": 9.04526801744037e-06, + "loss": 0.1167, + "step": 63923 + }, + { + "epoch": 2.46, + "learning_rate": 9.04462682739164e-06, + "loss": 0.2561, + "step": 63924 + }, + { + "epoch": 2.46, + "learning_rate": 9.043985637342908e-06, + "loss": 0.1551, + "step": 63925 + }, + { + "epoch": 2.46, + "learning_rate": 9.043344447294178e-06, + "loss": 0.2344, + "step": 63926 + }, + { + "epoch": 2.46, + "learning_rate": 9.042703257245448e-06, + "loss": 0.2903, + "step": 63927 + }, + { + "epoch": 2.46, + "learning_rate": 9.042062067196717e-06, + "loss": 0.2015, + "step": 63928 + }, + { + "epoch": 2.46, + "learning_rate": 9.041420877147987e-06, + "loss": 0.21, + "step": 63929 + }, + { + "epoch": 2.46, + "learning_rate": 9.040779687099257e-06, + "loss": 0.152, + "step": 63930 + }, + { + "epoch": 2.46, + "learning_rate": 9.040138497050527e-06, + "loss": 0.1442, + "step": 63931 + }, + { + "epoch": 2.46, + "learning_rate": 9.039497307001795e-06, + "loss": 0.1768, + "step": 63932 + }, + { + "epoch": 2.46, + "learning_rate": 9.038856116953065e-06, + "loss": 0.2795, + "step": 63933 + }, + { + "epoch": 2.46, + "learning_rate": 9.038214926904335e-06, + "loss": 0.2402, + "step": 63934 + }, + { + "epoch": 2.46, + "learning_rate": 9.037573736855605e-06, + "loss": 0.2926, + "step": 63935 + }, + { + "epoch": 2.46, + "learning_rate": 9.036932546806873e-06, + "loss": 0.2264, + "step": 63936 + }, + { + "epoch": 2.46, + "learning_rate": 9.036291356758143e-06, + "loss": 0.2704, + "step": 63937 + }, + { + "epoch": 2.46, + "learning_rate": 9.035650166709413e-06, + "loss": 0.1788, + "step": 63938 + }, + { + "epoch": 2.46, + "learning_rate": 9.035008976660683e-06, + "loss": 0.1653, + "step": 63939 + }, + { + "epoch": 2.46, + "learning_rate": 9.034367786611952e-06, + "loss": 0.2563, + "step": 63940 + }, + { + "epoch": 2.46, + "learning_rate": 9.033726596563222e-06, + "loss": 0.1926, + "step": 63941 + }, + { + "epoch": 2.46, + "learning_rate": 9.033085406514492e-06, + "loss": 0.271, + "step": 63942 + }, + { + "epoch": 2.46, + "learning_rate": 9.03244421646576e-06, + "loss": 0.1945, + "step": 63943 + }, + { + "epoch": 2.46, + "learning_rate": 9.03180302641703e-06, + "loss": 0.3896, + "step": 63944 + }, + { + "epoch": 2.46, + "learning_rate": 9.0311618363683e-06, + "loss": 0.1425, + "step": 63945 + }, + { + "epoch": 2.46, + "learning_rate": 9.030520646319568e-06, + "loss": 0.2722, + "step": 63946 + }, + { + "epoch": 2.46, + "learning_rate": 9.029879456270839e-06, + "loss": 0.2112, + "step": 63947 + }, + { + "epoch": 2.46, + "learning_rate": 9.029238266222109e-06, + "loss": 0.1504, + "step": 63948 + }, + { + "epoch": 2.46, + "learning_rate": 9.028597076173379e-06, + "loss": 0.2186, + "step": 63949 + }, + { + "epoch": 2.46, + "learning_rate": 9.027955886124647e-06, + "loss": 0.2448, + "step": 63950 + }, + { + "epoch": 2.46, + "learning_rate": 9.027314696075917e-06, + "loss": 0.2078, + "step": 63951 + }, + { + "epoch": 2.46, + "learning_rate": 9.026673506027187e-06, + "loss": 0.245, + "step": 63952 + }, + { + "epoch": 2.46, + "learning_rate": 9.026032315978457e-06, + "loss": 0.2265, + "step": 63953 + }, + { + "epoch": 2.46, + "learning_rate": 9.025391125929725e-06, + "loss": 0.1696, + "step": 63954 + }, + { + "epoch": 2.46, + "learning_rate": 9.024749935880995e-06, + "loss": 0.1059, + "step": 63955 + }, + { + "epoch": 2.46, + "learning_rate": 9.024108745832265e-06, + "loss": 0.2931, + "step": 63956 + }, + { + "epoch": 2.46, + "learning_rate": 9.023467555783535e-06, + "loss": 0.2078, + "step": 63957 + }, + { + "epoch": 2.46, + "learning_rate": 9.022826365734804e-06, + "loss": 0.2671, + "step": 63958 + }, + { + "epoch": 2.46, + "learning_rate": 9.022185175686074e-06, + "loss": 0.145, + "step": 63959 + }, + { + "epoch": 2.46, + "learning_rate": 9.021543985637344e-06, + "loss": 0.388, + "step": 63960 + }, + { + "epoch": 2.46, + "learning_rate": 9.020902795588614e-06, + "loss": 0.1698, + "step": 63961 + }, + { + "epoch": 2.46, + "learning_rate": 9.020261605539882e-06, + "loss": 0.1679, + "step": 63962 + }, + { + "epoch": 2.46, + "learning_rate": 9.019620415491152e-06, + "loss": 0.1641, + "step": 63963 + }, + { + "epoch": 2.46, + "learning_rate": 9.018979225442422e-06, + "loss": 0.3074, + "step": 63964 + }, + { + "epoch": 2.46, + "learning_rate": 9.01833803539369e-06, + "loss": 0.2741, + "step": 63965 + }, + { + "epoch": 2.46, + "learning_rate": 9.01769684534496e-06, + "loss": 0.1538, + "step": 63966 + }, + { + "epoch": 2.46, + "learning_rate": 9.01705565529623e-06, + "loss": 0.1366, + "step": 63967 + }, + { + "epoch": 2.46, + "learning_rate": 9.016414465247499e-06, + "loss": 0.246, + "step": 63968 + }, + { + "epoch": 2.46, + "learning_rate": 9.015773275198769e-06, + "loss": 0.2144, + "step": 63969 + }, + { + "epoch": 2.46, + "learning_rate": 9.015132085150039e-06, + "loss": 0.3962, + "step": 63970 + }, + { + "epoch": 2.46, + "learning_rate": 9.014490895101309e-06, + "loss": 0.1963, + "step": 63971 + }, + { + "epoch": 2.46, + "learning_rate": 9.013849705052577e-06, + "loss": 0.1673, + "step": 63972 + }, + { + "epoch": 2.46, + "learning_rate": 9.013208515003847e-06, + "loss": 0.1326, + "step": 63973 + }, + { + "epoch": 2.46, + "learning_rate": 9.012567324955117e-06, + "loss": 0.2688, + "step": 63974 + }, + { + "epoch": 2.46, + "learning_rate": 9.011926134906387e-06, + "loss": 0.2146, + "step": 63975 + }, + { + "epoch": 2.46, + "learning_rate": 9.011284944857656e-06, + "loss": 0.162, + "step": 63976 + }, + { + "epoch": 2.46, + "learning_rate": 9.010643754808926e-06, + "loss": 0.318, + "step": 63977 + }, + { + "epoch": 2.46, + "learning_rate": 9.010002564760196e-06, + "loss": 0.1938, + "step": 63978 + }, + { + "epoch": 2.46, + "learning_rate": 9.009361374711466e-06, + "loss": 0.1627, + "step": 63979 + }, + { + "epoch": 2.46, + "learning_rate": 9.008720184662734e-06, + "loss": 0.2601, + "step": 63980 + }, + { + "epoch": 2.46, + "learning_rate": 9.008078994614004e-06, + "loss": 0.3061, + "step": 63981 + }, + { + "epoch": 2.46, + "learning_rate": 9.007437804565274e-06, + "loss": 0.2418, + "step": 63982 + }, + { + "epoch": 2.46, + "learning_rate": 9.006796614516542e-06, + "loss": 0.166, + "step": 63983 + }, + { + "epoch": 2.46, + "learning_rate": 9.006155424467812e-06, + "loss": 0.3214, + "step": 63984 + }, + { + "epoch": 2.46, + "learning_rate": 9.005514234419082e-06, + "loss": 0.1404, + "step": 63985 + }, + { + "epoch": 2.46, + "learning_rate": 9.004873044370352e-06, + "loss": 0.1521, + "step": 63986 + }, + { + "epoch": 2.46, + "learning_rate": 9.00423185432162e-06, + "loss": 0.2345, + "step": 63987 + }, + { + "epoch": 2.46, + "learning_rate": 9.00359066427289e-06, + "loss": 0.2522, + "step": 63988 + }, + { + "epoch": 2.46, + "learning_rate": 9.00294947422416e-06, + "loss": 0.2625, + "step": 63989 + }, + { + "epoch": 2.46, + "learning_rate": 9.00230828417543e-06, + "loss": 0.1522, + "step": 63990 + }, + { + "epoch": 2.46, + "learning_rate": 9.0016670941267e-06, + "loss": 0.1881, + "step": 63991 + }, + { + "epoch": 2.46, + "learning_rate": 9.00102590407797e-06, + "loss": 0.228, + "step": 63992 + }, + { + "epoch": 2.46, + "learning_rate": 9.00038471402924e-06, + "loss": 0.1499, + "step": 63993 + }, + { + "epoch": 2.46, + "learning_rate": 8.999743523980508e-06, + "loss": 0.2058, + "step": 63994 + }, + { + "epoch": 2.46, + "learning_rate": 8.999102333931778e-06, + "loss": 0.298, + "step": 63995 + }, + { + "epoch": 2.46, + "learning_rate": 8.998461143883048e-06, + "loss": 0.1428, + "step": 63996 + }, + { + "epoch": 2.46, + "learning_rate": 8.997819953834318e-06, + "loss": 0.1296, + "step": 63997 + }, + { + "epoch": 2.46, + "learning_rate": 8.997178763785586e-06, + "loss": 0.1557, + "step": 63998 + }, + { + "epoch": 2.46, + "learning_rate": 8.996537573736856e-06, + "loss": 0.1769, + "step": 63999 + }, + { + "epoch": 2.46, + "learning_rate": 8.995896383688126e-06, + "loss": 0.1444, + "step": 64000 + }, + { + "epoch": 2.46, + "learning_rate": 8.995255193639396e-06, + "loss": 0.2188, + "step": 64001 + }, + { + "epoch": 2.46, + "learning_rate": 8.994614003590664e-06, + "loss": 0.313, + "step": 64002 + }, + { + "epoch": 2.46, + "learning_rate": 8.993972813541934e-06, + "loss": 0.1676, + "step": 64003 + }, + { + "epoch": 2.46, + "learning_rate": 8.993331623493204e-06, + "loss": 0.3318, + "step": 64004 + }, + { + "epoch": 2.46, + "learning_rate": 8.992690433444473e-06, + "loss": 0.2076, + "step": 64005 + }, + { + "epoch": 2.46, + "learning_rate": 8.992049243395743e-06, + "loss": 0.2226, + "step": 64006 + }, + { + "epoch": 2.46, + "learning_rate": 8.991408053347013e-06, + "loss": 0.2966, + "step": 64007 + }, + { + "epoch": 2.46, + "learning_rate": 8.990766863298281e-06, + "loss": 0.2326, + "step": 64008 + }, + { + "epoch": 2.46, + "learning_rate": 8.990125673249551e-06, + "loss": 0.2882, + "step": 64009 + }, + { + "epoch": 2.46, + "learning_rate": 8.989484483200821e-06, + "loss": 0.3871, + "step": 64010 + }, + { + "epoch": 2.46, + "learning_rate": 8.988843293152091e-06, + "loss": 0.1834, + "step": 64011 + }, + { + "epoch": 2.46, + "learning_rate": 8.98820210310336e-06, + "loss": 0.114, + "step": 64012 + }, + { + "epoch": 2.46, + "learning_rate": 8.98756091305463e-06, + "loss": 0.2848, + "step": 64013 + }, + { + "epoch": 2.46, + "learning_rate": 8.9869197230059e-06, + "loss": 0.2589, + "step": 64014 + }, + { + "epoch": 2.46, + "learning_rate": 8.98627853295717e-06, + "loss": 0.4248, + "step": 64015 + }, + { + "epoch": 2.46, + "learning_rate": 8.985637342908438e-06, + "loss": 0.2039, + "step": 64016 + }, + { + "epoch": 2.46, + "learning_rate": 8.984996152859708e-06, + "loss": 0.2177, + "step": 64017 + }, + { + "epoch": 2.46, + "learning_rate": 8.984354962810978e-06, + "loss": 0.2002, + "step": 64018 + }, + { + "epoch": 2.46, + "learning_rate": 8.983713772762248e-06, + "loss": 0.2497, + "step": 64019 + }, + { + "epoch": 2.46, + "learning_rate": 8.983072582713516e-06, + "loss": 0.1667, + "step": 64020 + }, + { + "epoch": 2.46, + "learning_rate": 8.982431392664786e-06, + "loss": 0.3082, + "step": 64021 + }, + { + "epoch": 2.46, + "learning_rate": 8.981790202616056e-06, + "loss": 0.3265, + "step": 64022 + }, + { + "epoch": 2.46, + "learning_rate": 8.981149012567325e-06, + "loss": 0.1589, + "step": 64023 + }, + { + "epoch": 2.46, + "learning_rate": 8.980507822518595e-06, + "loss": 0.2217, + "step": 64024 + }, + { + "epoch": 2.46, + "learning_rate": 8.979866632469865e-06, + "loss": 0.3158, + "step": 64025 + }, + { + "epoch": 2.46, + "learning_rate": 8.979225442421135e-06, + "loss": 0.3536, + "step": 64026 + }, + { + "epoch": 2.46, + "learning_rate": 8.978584252372403e-06, + "loss": 0.2115, + "step": 64027 + }, + { + "epoch": 2.46, + "learning_rate": 8.977943062323673e-06, + "loss": 0.2432, + "step": 64028 + }, + { + "epoch": 2.46, + "learning_rate": 8.977301872274943e-06, + "loss": 0.1266, + "step": 64029 + }, + { + "epoch": 2.46, + "learning_rate": 8.976660682226211e-06, + "loss": 0.2193, + "step": 64030 + }, + { + "epoch": 2.46, + "learning_rate": 8.976019492177481e-06, + "loss": 0.1717, + "step": 64031 + }, + { + "epoch": 2.46, + "learning_rate": 8.975378302128751e-06, + "loss": 0.1997, + "step": 64032 + }, + { + "epoch": 2.46, + "learning_rate": 8.974737112080021e-06, + "loss": 0.2336, + "step": 64033 + }, + { + "epoch": 2.46, + "learning_rate": 8.97409592203129e-06, + "loss": 0.197, + "step": 64034 + }, + { + "epoch": 2.46, + "learning_rate": 8.97345473198256e-06, + "loss": 0.3204, + "step": 64035 + }, + { + "epoch": 2.46, + "learning_rate": 8.97281354193383e-06, + "loss": 0.1695, + "step": 64036 + }, + { + "epoch": 2.46, + "learning_rate": 8.9721723518851e-06, + "loss": 0.2208, + "step": 64037 + }, + { + "epoch": 2.46, + "learning_rate": 8.971531161836368e-06, + "loss": 0.2446, + "step": 64038 + }, + { + "epoch": 2.46, + "learning_rate": 8.970889971787638e-06, + "loss": 0.3631, + "step": 64039 + }, + { + "epoch": 2.46, + "learning_rate": 8.970248781738908e-06, + "loss": 0.2993, + "step": 64040 + }, + { + "epoch": 2.46, + "learning_rate": 8.969607591690177e-06, + "loss": 0.1921, + "step": 64041 + }, + { + "epoch": 2.46, + "learning_rate": 8.968966401641447e-06, + "loss": 0.245, + "step": 64042 + }, + { + "epoch": 2.46, + "learning_rate": 8.968325211592717e-06, + "loss": 0.1924, + "step": 64043 + }, + { + "epoch": 2.46, + "learning_rate": 8.967684021543987e-06, + "loss": 0.261, + "step": 64044 + }, + { + "epoch": 2.46, + "learning_rate": 8.967042831495255e-06, + "loss": 0.1633, + "step": 64045 + }, + { + "epoch": 2.46, + "learning_rate": 8.966401641446525e-06, + "loss": 0.1675, + "step": 64046 + }, + { + "epoch": 2.46, + "learning_rate": 8.965760451397795e-06, + "loss": 0.1936, + "step": 64047 + }, + { + "epoch": 2.46, + "learning_rate": 8.965119261349065e-06, + "loss": 0.2363, + "step": 64048 + }, + { + "epoch": 2.46, + "learning_rate": 8.964478071300333e-06, + "loss": 0.1937, + "step": 64049 + }, + { + "epoch": 2.46, + "learning_rate": 8.963836881251603e-06, + "loss": 0.2797, + "step": 64050 + }, + { + "epoch": 2.46, + "learning_rate": 8.963195691202873e-06, + "loss": 0.415, + "step": 64051 + }, + { + "epoch": 2.46, + "learning_rate": 8.962554501154142e-06, + "loss": 0.1895, + "step": 64052 + }, + { + "epoch": 2.46, + "learning_rate": 8.961913311105412e-06, + "loss": 0.2498, + "step": 64053 + }, + { + "epoch": 2.46, + "learning_rate": 8.961272121056682e-06, + "loss": 0.1455, + "step": 64054 + }, + { + "epoch": 2.46, + "learning_rate": 8.960630931007952e-06, + "loss": 0.245, + "step": 64055 + }, + { + "epoch": 2.46, + "learning_rate": 8.95998974095922e-06, + "loss": 0.1936, + "step": 64056 + }, + { + "epoch": 2.46, + "learning_rate": 8.95934855091049e-06, + "loss": 0.2728, + "step": 64057 + }, + { + "epoch": 2.46, + "learning_rate": 8.95870736086176e-06, + "loss": 0.268, + "step": 64058 + }, + { + "epoch": 2.46, + "learning_rate": 8.95806617081303e-06, + "loss": 0.1354, + "step": 64059 + }, + { + "epoch": 2.46, + "learning_rate": 8.957424980764299e-06, + "loss": 0.1228, + "step": 64060 + }, + { + "epoch": 2.46, + "learning_rate": 8.956783790715569e-06, + "loss": 0.2455, + "step": 64061 + }, + { + "epoch": 2.46, + "learning_rate": 8.956142600666839e-06, + "loss": 0.1732, + "step": 64062 + }, + { + "epoch": 2.46, + "learning_rate": 8.955501410618107e-06, + "loss": 0.2924, + "step": 64063 + }, + { + "epoch": 2.46, + "learning_rate": 8.954860220569377e-06, + "loss": 0.13, + "step": 64064 + }, + { + "epoch": 2.46, + "learning_rate": 8.954219030520647e-06, + "loss": 0.2118, + "step": 64065 + }, + { + "epoch": 2.46, + "learning_rate": 8.953577840471917e-06, + "loss": 0.2578, + "step": 64066 + }, + { + "epoch": 2.46, + "learning_rate": 8.952936650423185e-06, + "loss": 0.2064, + "step": 64067 + }, + { + "epoch": 2.46, + "learning_rate": 8.952295460374455e-06, + "loss": 0.2366, + "step": 64068 + }, + { + "epoch": 2.46, + "learning_rate": 8.951654270325725e-06, + "loss": 0.3187, + "step": 64069 + }, + { + "epoch": 2.46, + "learning_rate": 8.951013080276994e-06, + "loss": 0.1731, + "step": 64070 + }, + { + "epoch": 2.46, + "learning_rate": 8.950371890228264e-06, + "loss": 0.1722, + "step": 64071 + }, + { + "epoch": 2.46, + "learning_rate": 8.949730700179534e-06, + "loss": 0.163, + "step": 64072 + }, + { + "epoch": 2.46, + "learning_rate": 8.949089510130804e-06, + "loss": 0.2551, + "step": 64073 + }, + { + "epoch": 2.46, + "learning_rate": 8.948448320082072e-06, + "loss": 0.2223, + "step": 64074 + }, + { + "epoch": 2.46, + "learning_rate": 8.947807130033342e-06, + "loss": 0.2692, + "step": 64075 + }, + { + "epoch": 2.46, + "learning_rate": 8.947165939984612e-06, + "loss": 0.1094, + "step": 64076 + }, + { + "epoch": 2.46, + "learning_rate": 8.946524749935882e-06, + "loss": 0.4197, + "step": 64077 + }, + { + "epoch": 2.46, + "learning_rate": 8.94588355988715e-06, + "loss": 0.1884, + "step": 64078 + }, + { + "epoch": 2.46, + "learning_rate": 8.94524236983842e-06, + "loss": 0.2786, + "step": 64079 + }, + { + "epoch": 2.46, + "learning_rate": 8.94460117978969e-06, + "loss": 0.1955, + "step": 64080 + }, + { + "epoch": 2.46, + "learning_rate": 8.943959989740959e-06, + "loss": 0.1327, + "step": 64081 + }, + { + "epoch": 2.46, + "learning_rate": 8.943318799692229e-06, + "loss": 0.1502, + "step": 64082 + }, + { + "epoch": 2.46, + "learning_rate": 8.942677609643499e-06, + "loss": 0.2613, + "step": 64083 + }, + { + "epoch": 2.46, + "learning_rate": 8.942036419594769e-06, + "loss": 0.3361, + "step": 64084 + }, + { + "epoch": 2.46, + "learning_rate": 8.941395229546037e-06, + "loss": 0.2455, + "step": 64085 + }, + { + "epoch": 2.46, + "learning_rate": 8.940754039497307e-06, + "loss": 0.3087, + "step": 64086 + }, + { + "epoch": 2.46, + "learning_rate": 8.940112849448577e-06, + "loss": 0.2815, + "step": 64087 + }, + { + "epoch": 2.46, + "learning_rate": 8.939471659399847e-06, + "loss": 0.1888, + "step": 64088 + }, + { + "epoch": 2.46, + "learning_rate": 8.938830469351116e-06, + "loss": 0.2147, + "step": 64089 + }, + { + "epoch": 2.46, + "learning_rate": 8.938189279302386e-06, + "loss": 0.1465, + "step": 64090 + }, + { + "epoch": 2.46, + "learning_rate": 8.937548089253656e-06, + "loss": 0.2996, + "step": 64091 + }, + { + "epoch": 2.46, + "learning_rate": 8.936906899204924e-06, + "loss": 0.2085, + "step": 64092 + }, + { + "epoch": 2.46, + "learning_rate": 8.936265709156194e-06, + "loss": 0.1816, + "step": 64093 + }, + { + "epoch": 2.46, + "learning_rate": 8.935624519107464e-06, + "loss": 0.1993, + "step": 64094 + }, + { + "epoch": 2.46, + "learning_rate": 8.934983329058734e-06, + "loss": 0.2118, + "step": 64095 + }, + { + "epoch": 2.46, + "learning_rate": 8.934342139010002e-06, + "loss": 0.2586, + "step": 64096 + }, + { + "epoch": 2.46, + "learning_rate": 8.933700948961272e-06, + "loss": 0.2635, + "step": 64097 + }, + { + "epoch": 2.46, + "learning_rate": 8.933059758912542e-06, + "loss": 0.2041, + "step": 64098 + }, + { + "epoch": 2.46, + "learning_rate": 8.932418568863812e-06, + "loss": 0.1861, + "step": 64099 + }, + { + "epoch": 2.46, + "learning_rate": 8.93177737881508e-06, + "loss": 0.1487, + "step": 64100 + }, + { + "epoch": 2.46, + "learning_rate": 8.93113618876635e-06, + "loss": 0.2661, + "step": 64101 + }, + { + "epoch": 2.46, + "learning_rate": 8.930494998717621e-06, + "loss": 0.2106, + "step": 64102 + }, + { + "epoch": 2.46, + "learning_rate": 8.92985380866889e-06, + "loss": 0.1763, + "step": 64103 + }, + { + "epoch": 2.46, + "learning_rate": 8.92921261862016e-06, + "loss": 0.1779, + "step": 64104 + }, + { + "epoch": 2.46, + "learning_rate": 8.92857142857143e-06, + "loss": 0.1557, + "step": 64105 + }, + { + "epoch": 2.46, + "learning_rate": 8.9279302385227e-06, + "loss": 0.2104, + "step": 64106 + }, + { + "epoch": 2.46, + "learning_rate": 8.927289048473968e-06, + "loss": 0.2974, + "step": 64107 + }, + { + "epoch": 2.46, + "learning_rate": 8.926647858425238e-06, + "loss": 0.3788, + "step": 64108 + }, + { + "epoch": 2.46, + "learning_rate": 8.926006668376508e-06, + "loss": 0.2727, + "step": 64109 + }, + { + "epoch": 2.46, + "learning_rate": 8.925365478327776e-06, + "loss": 0.2571, + "step": 64110 + }, + { + "epoch": 2.46, + "learning_rate": 8.924724288279046e-06, + "loss": 0.2643, + "step": 64111 + }, + { + "epoch": 2.46, + "learning_rate": 8.924083098230316e-06, + "loss": 0.1539, + "step": 64112 + }, + { + "epoch": 2.46, + "learning_rate": 8.923441908181586e-06, + "loss": 0.361, + "step": 64113 + }, + { + "epoch": 2.46, + "learning_rate": 8.922800718132854e-06, + "loss": 0.2519, + "step": 64114 + }, + { + "epoch": 2.47, + "learning_rate": 8.922159528084124e-06, + "loss": 0.2552, + "step": 64115 + }, + { + "epoch": 2.47, + "learning_rate": 8.921518338035394e-06, + "loss": 0.2612, + "step": 64116 + }, + { + "epoch": 2.47, + "learning_rate": 8.920877147986664e-06, + "loss": 0.2657, + "step": 64117 + }, + { + "epoch": 2.47, + "learning_rate": 8.920235957937933e-06, + "loss": 0.3005, + "step": 64118 + }, + { + "epoch": 2.47, + "learning_rate": 8.919594767889203e-06, + "loss": 0.1509, + "step": 64119 + }, + { + "epoch": 2.47, + "learning_rate": 8.918953577840473e-06, + "loss": 0.2496, + "step": 64120 + }, + { + "epoch": 2.47, + "learning_rate": 8.918312387791741e-06, + "loss": 0.2321, + "step": 64121 + }, + { + "epoch": 2.47, + "learning_rate": 8.917671197743011e-06, + "loss": 0.2898, + "step": 64122 + }, + { + "epoch": 2.47, + "learning_rate": 8.917030007694281e-06, + "loss": 0.143, + "step": 64123 + }, + { + "epoch": 2.47, + "learning_rate": 8.916388817645551e-06, + "loss": 0.1753, + "step": 64124 + }, + { + "epoch": 2.47, + "learning_rate": 8.91574762759682e-06, + "loss": 0.1978, + "step": 64125 + }, + { + "epoch": 2.47, + "learning_rate": 8.91510643754809e-06, + "loss": 0.2947, + "step": 64126 + }, + { + "epoch": 2.47, + "learning_rate": 8.91446524749936e-06, + "loss": 0.2576, + "step": 64127 + }, + { + "epoch": 2.47, + "learning_rate": 8.91382405745063e-06, + "loss": 0.1987, + "step": 64128 + }, + { + "epoch": 2.47, + "learning_rate": 8.913182867401898e-06, + "loss": 0.1269, + "step": 64129 + }, + { + "epoch": 2.47, + "learning_rate": 8.912541677353168e-06, + "loss": 0.1668, + "step": 64130 + }, + { + "epoch": 2.47, + "learning_rate": 8.911900487304438e-06, + "loss": 0.053, + "step": 64131 + }, + { + "epoch": 2.47, + "learning_rate": 8.911259297255706e-06, + "loss": 0.1995, + "step": 64132 + }, + { + "epoch": 2.47, + "learning_rate": 8.910618107206976e-06, + "loss": 0.2309, + "step": 64133 + }, + { + "epoch": 2.47, + "learning_rate": 8.909976917158246e-06, + "loss": 0.2264, + "step": 64134 + }, + { + "epoch": 2.47, + "learning_rate": 8.909335727109516e-06, + "loss": 0.2617, + "step": 64135 + }, + { + "epoch": 2.47, + "learning_rate": 8.908694537060785e-06, + "loss": 0.3765, + "step": 64136 + }, + { + "epoch": 2.47, + "learning_rate": 8.908053347012055e-06, + "loss": 0.2507, + "step": 64137 + }, + { + "epoch": 2.47, + "learning_rate": 8.907412156963325e-06, + "loss": 0.184, + "step": 64138 + }, + { + "epoch": 2.47, + "learning_rate": 8.906770966914593e-06, + "loss": 0.1232, + "step": 64139 + }, + { + "epoch": 2.47, + "learning_rate": 8.906129776865863e-06, + "loss": 0.1485, + "step": 64140 + }, + { + "epoch": 2.47, + "learning_rate": 8.905488586817133e-06, + "loss": 0.2155, + "step": 64141 + }, + { + "epoch": 2.47, + "learning_rate": 8.904847396768403e-06, + "loss": 0.1476, + "step": 64142 + }, + { + "epoch": 2.47, + "learning_rate": 8.904206206719671e-06, + "loss": 0.2622, + "step": 64143 + }, + { + "epoch": 2.47, + "learning_rate": 8.903565016670941e-06, + "loss": 0.3362, + "step": 64144 + }, + { + "epoch": 2.47, + "learning_rate": 8.902923826622212e-06, + "loss": 0.1838, + "step": 64145 + }, + { + "epoch": 2.47, + "learning_rate": 8.902282636573482e-06, + "loss": 0.4986, + "step": 64146 + }, + { + "epoch": 2.47, + "learning_rate": 8.90164144652475e-06, + "loss": 0.2449, + "step": 64147 + }, + { + "epoch": 2.47, + "learning_rate": 8.90100025647602e-06, + "loss": 0.1779, + "step": 64148 + }, + { + "epoch": 2.47, + "learning_rate": 8.90035906642729e-06, + "loss": 0.1172, + "step": 64149 + }, + { + "epoch": 2.47, + "learning_rate": 8.89971787637856e-06, + "loss": 0.2685, + "step": 64150 + }, + { + "epoch": 2.47, + "learning_rate": 8.899076686329828e-06, + "loss": 0.3404, + "step": 64151 + }, + { + "epoch": 2.47, + "learning_rate": 8.898435496281098e-06, + "loss": 0.1616, + "step": 64152 + }, + { + "epoch": 2.47, + "learning_rate": 8.897794306232368e-06, + "loss": 0.1593, + "step": 64153 + }, + { + "epoch": 2.47, + "learning_rate": 8.897153116183637e-06, + "loss": 0.2875, + "step": 64154 + }, + { + "epoch": 2.47, + "learning_rate": 8.896511926134907e-06, + "loss": 0.1248, + "step": 64155 + }, + { + "epoch": 2.47, + "learning_rate": 8.895870736086177e-06, + "loss": 0.2667, + "step": 64156 + }, + { + "epoch": 2.47, + "learning_rate": 8.895229546037447e-06, + "loss": 0.1141, + "step": 64157 + }, + { + "epoch": 2.47, + "learning_rate": 8.894588355988715e-06, + "loss": 0.2032, + "step": 64158 + }, + { + "epoch": 2.47, + "learning_rate": 8.893947165939985e-06, + "loss": 0.235, + "step": 64159 + }, + { + "epoch": 2.47, + "learning_rate": 8.893305975891255e-06, + "loss": 0.2492, + "step": 64160 + }, + { + "epoch": 2.47, + "learning_rate": 8.892664785842523e-06, + "loss": 0.138, + "step": 64161 + }, + { + "epoch": 2.47, + "learning_rate": 8.892023595793793e-06, + "loss": 0.308, + "step": 64162 + }, + { + "epoch": 2.47, + "learning_rate": 8.891382405745063e-06, + "loss": 0.1614, + "step": 64163 + }, + { + "epoch": 2.47, + "learning_rate": 8.890741215696333e-06, + "loss": 0.1517, + "step": 64164 + }, + { + "epoch": 2.47, + "learning_rate": 8.890100025647602e-06, + "loss": 0.1184, + "step": 64165 + }, + { + "epoch": 2.47, + "learning_rate": 8.889458835598872e-06, + "loss": 0.142, + "step": 64166 + }, + { + "epoch": 2.47, + "learning_rate": 8.888817645550142e-06, + "loss": 0.2096, + "step": 64167 + }, + { + "epoch": 2.47, + "learning_rate": 8.888176455501412e-06, + "loss": 0.3996, + "step": 64168 + }, + { + "epoch": 2.47, + "learning_rate": 8.88753526545268e-06, + "loss": 0.2097, + "step": 64169 + }, + { + "epoch": 2.47, + "learning_rate": 8.88689407540395e-06, + "loss": 0.1752, + "step": 64170 + }, + { + "epoch": 2.47, + "learning_rate": 8.88625288535522e-06, + "loss": 0.1776, + "step": 64171 + }, + { + "epoch": 2.47, + "learning_rate": 8.885611695306489e-06, + "loss": 0.1526, + "step": 64172 + }, + { + "epoch": 2.47, + "learning_rate": 8.884970505257759e-06, + "loss": 0.2234, + "step": 64173 + }, + { + "epoch": 2.47, + "learning_rate": 8.884329315209029e-06, + "loss": 0.328, + "step": 64174 + }, + { + "epoch": 2.47, + "learning_rate": 8.883688125160299e-06, + "loss": 0.2134, + "step": 64175 + }, + { + "epoch": 2.47, + "learning_rate": 8.883046935111567e-06, + "loss": 0.2824, + "step": 64176 + }, + { + "epoch": 2.47, + "learning_rate": 8.882405745062837e-06, + "loss": 0.2303, + "step": 64177 + }, + { + "epoch": 2.47, + "learning_rate": 8.881764555014107e-06, + "loss": 0.2266, + "step": 64178 + }, + { + "epoch": 2.47, + "learning_rate": 8.881123364965375e-06, + "loss": 0.2087, + "step": 64179 + }, + { + "epoch": 2.47, + "learning_rate": 8.880482174916645e-06, + "loss": 0.2481, + "step": 64180 + }, + { + "epoch": 2.47, + "learning_rate": 8.879840984867915e-06, + "loss": 0.1717, + "step": 64181 + }, + { + "epoch": 2.47, + "learning_rate": 8.879199794819185e-06, + "loss": 0.2289, + "step": 64182 + }, + { + "epoch": 2.47, + "learning_rate": 8.878558604770454e-06, + "loss": 0.2315, + "step": 64183 + }, + { + "epoch": 2.47, + "learning_rate": 8.877917414721724e-06, + "loss": 0.2424, + "step": 64184 + }, + { + "epoch": 2.47, + "learning_rate": 8.877276224672994e-06, + "loss": 0.2445, + "step": 64185 + }, + { + "epoch": 2.47, + "learning_rate": 8.876635034624264e-06, + "loss": 0.1098, + "step": 64186 + }, + { + "epoch": 2.47, + "learning_rate": 8.875993844575532e-06, + "loss": 0.1677, + "step": 64187 + }, + { + "epoch": 2.47, + "learning_rate": 8.875352654526802e-06, + "loss": 0.1791, + "step": 64188 + }, + { + "epoch": 2.47, + "learning_rate": 8.874711464478072e-06, + "loss": 0.1456, + "step": 64189 + }, + { + "epoch": 2.47, + "learning_rate": 8.874070274429342e-06, + "loss": 0.2166, + "step": 64190 + }, + { + "epoch": 2.47, + "learning_rate": 8.87342908438061e-06, + "loss": 0.1981, + "step": 64191 + }, + { + "epoch": 2.47, + "learning_rate": 8.87278789433188e-06, + "loss": 0.3117, + "step": 64192 + }, + { + "epoch": 2.47, + "learning_rate": 8.87214670428315e-06, + "loss": 0.1461, + "step": 64193 + }, + { + "epoch": 2.47, + "learning_rate": 8.871505514234419e-06, + "loss": 0.2125, + "step": 64194 + }, + { + "epoch": 2.47, + "learning_rate": 8.870864324185689e-06, + "loss": 0.3275, + "step": 64195 + }, + { + "epoch": 2.47, + "learning_rate": 8.870223134136959e-06, + "loss": 0.198, + "step": 64196 + }, + { + "epoch": 2.47, + "learning_rate": 8.869581944088229e-06, + "loss": 0.2224, + "step": 64197 + }, + { + "epoch": 2.47, + "learning_rate": 8.868940754039497e-06, + "loss": 0.1904, + "step": 64198 + }, + { + "epoch": 2.47, + "learning_rate": 8.868299563990767e-06, + "loss": 0.1639, + "step": 64199 + }, + { + "epoch": 2.47, + "learning_rate": 8.867658373942037e-06, + "loss": 0.1974, + "step": 64200 + }, + { + "epoch": 2.47, + "learning_rate": 8.867017183893306e-06, + "loss": 0.1073, + "step": 64201 + }, + { + "epoch": 2.47, + "learning_rate": 8.866375993844576e-06, + "loss": 0.2569, + "step": 64202 + }, + { + "epoch": 2.47, + "learning_rate": 8.865734803795846e-06, + "loss": 0.1668, + "step": 64203 + }, + { + "epoch": 2.47, + "learning_rate": 8.865093613747116e-06, + "loss": 0.1993, + "step": 64204 + }, + { + "epoch": 2.47, + "learning_rate": 8.864452423698384e-06, + "loss": 0.2714, + "step": 64205 + }, + { + "epoch": 2.47, + "learning_rate": 8.863811233649654e-06, + "loss": 0.1991, + "step": 64206 + }, + { + "epoch": 2.47, + "learning_rate": 8.863170043600924e-06, + "loss": 0.2883, + "step": 64207 + }, + { + "epoch": 2.47, + "learning_rate": 8.862528853552194e-06, + "loss": 0.1718, + "step": 64208 + }, + { + "epoch": 2.47, + "learning_rate": 8.861887663503462e-06, + "loss": 0.1918, + "step": 64209 + }, + { + "epoch": 2.47, + "learning_rate": 8.861246473454732e-06, + "loss": 0.1823, + "step": 64210 + }, + { + "epoch": 2.47, + "learning_rate": 8.860605283406003e-06, + "loss": 0.2879, + "step": 64211 + }, + { + "epoch": 2.47, + "learning_rate": 8.85996409335727e-06, + "loss": 0.1995, + "step": 64212 + }, + { + "epoch": 2.47, + "learning_rate": 8.859322903308541e-06, + "loss": 0.133, + "step": 64213 + }, + { + "epoch": 2.47, + "learning_rate": 8.858681713259811e-06, + "loss": 0.3225, + "step": 64214 + }, + { + "epoch": 2.47, + "learning_rate": 8.858040523211081e-06, + "loss": 0.2599, + "step": 64215 + }, + { + "epoch": 2.47, + "learning_rate": 8.85739933316235e-06, + "loss": 0.255, + "step": 64216 + }, + { + "epoch": 2.47, + "learning_rate": 8.85675814311362e-06, + "loss": 0.2903, + "step": 64217 + }, + { + "epoch": 2.47, + "learning_rate": 8.85611695306489e-06, + "loss": 0.2597, + "step": 64218 + }, + { + "epoch": 2.47, + "learning_rate": 8.855475763016158e-06, + "loss": 0.2323, + "step": 64219 + }, + { + "epoch": 2.47, + "learning_rate": 8.854834572967428e-06, + "loss": 0.2199, + "step": 64220 + }, + { + "epoch": 2.47, + "learning_rate": 8.854193382918698e-06, + "loss": 0.3026, + "step": 64221 + }, + { + "epoch": 2.47, + "learning_rate": 8.853552192869968e-06, + "loss": 0.1408, + "step": 64222 + }, + { + "epoch": 2.47, + "learning_rate": 8.852911002821236e-06, + "loss": 0.2339, + "step": 64223 + }, + { + "epoch": 2.47, + "learning_rate": 8.852269812772506e-06, + "loss": 0.1357, + "step": 64224 + }, + { + "epoch": 2.47, + "learning_rate": 8.851628622723776e-06, + "loss": 0.1867, + "step": 64225 + }, + { + "epoch": 2.47, + "learning_rate": 8.850987432675046e-06, + "loss": 0.2567, + "step": 64226 + }, + { + "epoch": 2.47, + "learning_rate": 8.850346242626314e-06, + "loss": 0.2401, + "step": 64227 + }, + { + "epoch": 2.47, + "learning_rate": 8.849705052577584e-06, + "loss": 0.1419, + "step": 64228 + }, + { + "epoch": 2.47, + "learning_rate": 8.849063862528854e-06, + "loss": 0.0934, + "step": 64229 + }, + { + "epoch": 2.47, + "learning_rate": 8.848422672480124e-06, + "loss": 0.192, + "step": 64230 + }, + { + "epoch": 2.47, + "learning_rate": 8.847781482431393e-06, + "loss": 0.2077, + "step": 64231 + }, + { + "epoch": 2.47, + "learning_rate": 8.847140292382663e-06, + "loss": 0.2541, + "step": 64232 + }, + { + "epoch": 2.47, + "learning_rate": 8.846499102333933e-06, + "loss": 0.1949, + "step": 64233 + }, + { + "epoch": 2.47, + "learning_rate": 8.845857912285201e-06, + "loss": 0.1433, + "step": 64234 + }, + { + "epoch": 2.47, + "learning_rate": 8.845216722236471e-06, + "loss": 0.2976, + "step": 64235 + }, + { + "epoch": 2.47, + "learning_rate": 8.844575532187741e-06, + "loss": 0.186, + "step": 64236 + }, + { + "epoch": 2.47, + "learning_rate": 8.843934342139011e-06, + "loss": 0.3608, + "step": 64237 + }, + { + "epoch": 2.47, + "learning_rate": 8.84329315209028e-06, + "loss": 0.1647, + "step": 64238 + }, + { + "epoch": 2.47, + "learning_rate": 8.84265196204155e-06, + "loss": 0.1016, + "step": 64239 + }, + { + "epoch": 2.47, + "learning_rate": 8.84201077199282e-06, + "loss": 0.3158, + "step": 64240 + }, + { + "epoch": 2.47, + "learning_rate": 8.841369581944088e-06, + "loss": 0.2031, + "step": 64241 + }, + { + "epoch": 2.47, + "learning_rate": 8.840728391895358e-06, + "loss": 0.2415, + "step": 64242 + }, + { + "epoch": 2.47, + "learning_rate": 8.840087201846628e-06, + "loss": 0.3166, + "step": 64243 + }, + { + "epoch": 2.47, + "learning_rate": 8.839446011797898e-06, + "loss": 0.2388, + "step": 64244 + }, + { + "epoch": 2.47, + "learning_rate": 8.838804821749166e-06, + "loss": 0.1122, + "step": 64245 + }, + { + "epoch": 2.47, + "learning_rate": 8.838163631700436e-06, + "loss": 0.2539, + "step": 64246 + }, + { + "epoch": 2.47, + "learning_rate": 8.837522441651706e-06, + "loss": 0.3065, + "step": 64247 + }, + { + "epoch": 2.47, + "learning_rate": 8.836881251602976e-06, + "loss": 0.3285, + "step": 64248 + }, + { + "epoch": 2.47, + "learning_rate": 8.836240061554245e-06, + "loss": 0.1729, + "step": 64249 + }, + { + "epoch": 2.47, + "learning_rate": 8.835598871505515e-06, + "loss": 0.2253, + "step": 64250 + }, + { + "epoch": 2.47, + "learning_rate": 8.834957681456785e-06, + "loss": 0.2408, + "step": 64251 + }, + { + "epoch": 2.47, + "learning_rate": 8.834316491408055e-06, + "loss": 0.3372, + "step": 64252 + }, + { + "epoch": 2.47, + "learning_rate": 8.833675301359323e-06, + "loss": 0.1416, + "step": 64253 + }, + { + "epoch": 2.47, + "learning_rate": 8.833034111310593e-06, + "loss": 0.1363, + "step": 64254 + }, + { + "epoch": 2.47, + "learning_rate": 8.832392921261863e-06, + "loss": 0.1401, + "step": 64255 + }, + { + "epoch": 2.47, + "learning_rate": 8.831751731213131e-06, + "loss": 0.3171, + "step": 64256 + }, + { + "epoch": 2.47, + "learning_rate": 8.831110541164402e-06, + "loss": 0.1578, + "step": 64257 + }, + { + "epoch": 2.47, + "learning_rate": 8.830469351115672e-06, + "loss": 0.22, + "step": 64258 + }, + { + "epoch": 2.47, + "learning_rate": 8.82982816106694e-06, + "loss": 0.1932, + "step": 64259 + }, + { + "epoch": 2.47, + "learning_rate": 8.82918697101821e-06, + "loss": 0.3253, + "step": 64260 + }, + { + "epoch": 2.47, + "learning_rate": 8.82854578096948e-06, + "loss": 0.2622, + "step": 64261 + }, + { + "epoch": 2.47, + "learning_rate": 8.82790459092075e-06, + "loss": 0.155, + "step": 64262 + }, + { + "epoch": 2.47, + "learning_rate": 8.827263400872018e-06, + "loss": 0.2145, + "step": 64263 + }, + { + "epoch": 2.47, + "learning_rate": 8.826622210823288e-06, + "loss": 0.1818, + "step": 64264 + }, + { + "epoch": 2.47, + "learning_rate": 8.825981020774558e-06, + "loss": 0.3261, + "step": 64265 + }, + { + "epoch": 2.47, + "learning_rate": 8.825339830725828e-06, + "loss": 0.2754, + "step": 64266 + }, + { + "epoch": 2.47, + "learning_rate": 8.824698640677097e-06, + "loss": 0.1022, + "step": 64267 + }, + { + "epoch": 2.47, + "learning_rate": 8.824057450628367e-06, + "loss": 0.313, + "step": 64268 + }, + { + "epoch": 2.47, + "learning_rate": 8.823416260579637e-06, + "loss": 0.2478, + "step": 64269 + }, + { + "epoch": 2.47, + "learning_rate": 8.822775070530907e-06, + "loss": 0.1935, + "step": 64270 + }, + { + "epoch": 2.47, + "learning_rate": 8.822133880482175e-06, + "loss": 0.1915, + "step": 64271 + }, + { + "epoch": 2.47, + "learning_rate": 8.821492690433445e-06, + "loss": 0.175, + "step": 64272 + }, + { + "epoch": 2.47, + "learning_rate": 8.820851500384715e-06, + "loss": 0.3426, + "step": 64273 + }, + { + "epoch": 2.47, + "learning_rate": 8.820210310335983e-06, + "loss": 0.2041, + "step": 64274 + }, + { + "epoch": 2.47, + "learning_rate": 8.819569120287253e-06, + "loss": 0.2368, + "step": 64275 + }, + { + "epoch": 2.47, + "learning_rate": 8.818927930238523e-06, + "loss": 0.1365, + "step": 64276 + }, + { + "epoch": 2.47, + "learning_rate": 8.818286740189792e-06, + "loss": 0.3469, + "step": 64277 + }, + { + "epoch": 2.47, + "learning_rate": 8.817645550141062e-06, + "loss": 0.254, + "step": 64278 + }, + { + "epoch": 2.47, + "learning_rate": 8.817004360092332e-06, + "loss": 0.3118, + "step": 64279 + }, + { + "epoch": 2.47, + "learning_rate": 8.816363170043602e-06, + "loss": 0.2153, + "step": 64280 + }, + { + "epoch": 2.47, + "learning_rate": 8.81572197999487e-06, + "loss": 0.113, + "step": 64281 + }, + { + "epoch": 2.47, + "learning_rate": 8.81508078994614e-06, + "loss": 0.2275, + "step": 64282 + }, + { + "epoch": 2.47, + "learning_rate": 8.81443959989741e-06, + "loss": 0.3616, + "step": 64283 + }, + { + "epoch": 2.47, + "learning_rate": 8.81379840984868e-06, + "loss": 0.2448, + "step": 64284 + }, + { + "epoch": 2.47, + "learning_rate": 8.813157219799949e-06, + "loss": 0.1301, + "step": 64285 + }, + { + "epoch": 2.47, + "learning_rate": 8.812516029751219e-06, + "loss": 0.1653, + "step": 64286 + }, + { + "epoch": 2.47, + "learning_rate": 8.811874839702489e-06, + "loss": 0.1502, + "step": 64287 + }, + { + "epoch": 2.47, + "learning_rate": 8.811233649653759e-06, + "loss": 0.1443, + "step": 64288 + }, + { + "epoch": 2.47, + "learning_rate": 8.810592459605027e-06, + "loss": 0.2114, + "step": 64289 + }, + { + "epoch": 2.47, + "learning_rate": 8.809951269556297e-06, + "loss": 0.1249, + "step": 64290 + }, + { + "epoch": 2.47, + "learning_rate": 8.809310079507567e-06, + "loss": 0.1721, + "step": 64291 + }, + { + "epoch": 2.47, + "learning_rate": 8.808668889458837e-06, + "loss": 0.2389, + "step": 64292 + }, + { + "epoch": 2.47, + "learning_rate": 8.808027699410105e-06, + "loss": 0.1981, + "step": 64293 + }, + { + "epoch": 2.47, + "learning_rate": 8.807386509361375e-06, + "loss": 0.2869, + "step": 64294 + }, + { + "epoch": 2.47, + "learning_rate": 8.806745319312645e-06, + "loss": 0.205, + "step": 64295 + }, + { + "epoch": 2.47, + "learning_rate": 8.806104129263914e-06, + "loss": 0.1993, + "step": 64296 + }, + { + "epoch": 2.47, + "learning_rate": 8.805462939215184e-06, + "loss": 0.3126, + "step": 64297 + }, + { + "epoch": 2.47, + "learning_rate": 8.804821749166454e-06, + "loss": 0.2468, + "step": 64298 + }, + { + "epoch": 2.47, + "learning_rate": 8.804180559117722e-06, + "loss": 0.2262, + "step": 64299 + }, + { + "epoch": 2.47, + "learning_rate": 8.803539369068992e-06, + "loss": 0.0907, + "step": 64300 + }, + { + "epoch": 2.47, + "learning_rate": 8.802898179020262e-06, + "loss": 0.2514, + "step": 64301 + }, + { + "epoch": 2.47, + "learning_rate": 8.802256988971532e-06, + "loss": 0.1671, + "step": 64302 + }, + { + "epoch": 2.47, + "learning_rate": 8.8016157989228e-06, + "loss": 0.2916, + "step": 64303 + }, + { + "epoch": 2.47, + "learning_rate": 8.80097460887407e-06, + "loss": 0.2709, + "step": 64304 + }, + { + "epoch": 2.47, + "learning_rate": 8.80033341882534e-06, + "loss": 0.3154, + "step": 64305 + }, + { + "epoch": 2.47, + "learning_rate": 8.79969222877661e-06, + "loss": 0.2416, + "step": 64306 + }, + { + "epoch": 2.47, + "learning_rate": 8.799051038727879e-06, + "loss": 0.159, + "step": 64307 + }, + { + "epoch": 2.47, + "learning_rate": 8.798409848679149e-06, + "loss": 0.1931, + "step": 64308 + }, + { + "epoch": 2.47, + "learning_rate": 8.797768658630419e-06, + "loss": 0.2531, + "step": 64309 + }, + { + "epoch": 2.47, + "learning_rate": 8.797127468581689e-06, + "loss": 0.2188, + "step": 64310 + }, + { + "epoch": 2.47, + "learning_rate": 8.796486278532957e-06, + "loss": 0.2577, + "step": 64311 + }, + { + "epoch": 2.47, + "learning_rate": 8.795845088484227e-06, + "loss": 0.2693, + "step": 64312 + }, + { + "epoch": 2.47, + "learning_rate": 8.795203898435497e-06, + "loss": 0.2758, + "step": 64313 + }, + { + "epoch": 2.47, + "learning_rate": 8.794562708386766e-06, + "loss": 0.2163, + "step": 64314 + }, + { + "epoch": 2.47, + "learning_rate": 8.793921518338036e-06, + "loss": 0.1377, + "step": 64315 + }, + { + "epoch": 2.47, + "learning_rate": 8.793280328289306e-06, + "loss": 0.1385, + "step": 64316 + }, + { + "epoch": 2.47, + "learning_rate": 8.792639138240574e-06, + "loss": 0.3165, + "step": 64317 + }, + { + "epoch": 2.47, + "learning_rate": 8.791997948191844e-06, + "loss": 0.1716, + "step": 64318 + }, + { + "epoch": 2.47, + "learning_rate": 8.791356758143114e-06, + "loss": 0.3296, + "step": 64319 + }, + { + "epoch": 2.47, + "learning_rate": 8.790715568094384e-06, + "loss": 0.1886, + "step": 64320 + }, + { + "epoch": 2.47, + "learning_rate": 8.790074378045652e-06, + "loss": 0.219, + "step": 64321 + }, + { + "epoch": 2.47, + "learning_rate": 8.789433187996922e-06, + "loss": 0.1732, + "step": 64322 + }, + { + "epoch": 2.47, + "learning_rate": 8.788791997948193e-06, + "loss": 0.323, + "step": 64323 + }, + { + "epoch": 2.47, + "learning_rate": 8.788150807899463e-06, + "loss": 0.1845, + "step": 64324 + }, + { + "epoch": 2.47, + "learning_rate": 8.787509617850731e-06, + "loss": 0.2275, + "step": 64325 + }, + { + "epoch": 2.47, + "learning_rate": 8.786868427802001e-06, + "loss": 0.2331, + "step": 64326 + }, + { + "epoch": 2.47, + "learning_rate": 8.786227237753271e-06, + "loss": 0.2221, + "step": 64327 + }, + { + "epoch": 2.47, + "learning_rate": 8.785586047704541e-06, + "loss": 0.3028, + "step": 64328 + }, + { + "epoch": 2.47, + "learning_rate": 8.78494485765581e-06, + "loss": 0.1761, + "step": 64329 + }, + { + "epoch": 2.47, + "learning_rate": 8.78430366760708e-06, + "loss": 0.2114, + "step": 64330 + }, + { + "epoch": 2.47, + "learning_rate": 8.78366247755835e-06, + "loss": 0.1777, + "step": 64331 + }, + { + "epoch": 2.47, + "learning_rate": 8.78302128750962e-06, + "loss": 0.2901, + "step": 64332 + }, + { + "epoch": 2.47, + "learning_rate": 8.782380097460888e-06, + "loss": 0.1277, + "step": 64333 + }, + { + "epoch": 2.47, + "learning_rate": 8.781738907412158e-06, + "loss": 0.144, + "step": 64334 + }, + { + "epoch": 2.47, + "learning_rate": 8.781097717363428e-06, + "loss": 0.2298, + "step": 64335 + }, + { + "epoch": 2.47, + "learning_rate": 8.780456527314696e-06, + "loss": 0.1776, + "step": 64336 + }, + { + "epoch": 2.47, + "learning_rate": 8.779815337265966e-06, + "loss": 0.3496, + "step": 64337 + }, + { + "epoch": 2.47, + "learning_rate": 8.779174147217236e-06, + "loss": 0.3072, + "step": 64338 + }, + { + "epoch": 2.47, + "learning_rate": 8.778532957168504e-06, + "loss": 0.1965, + "step": 64339 + }, + { + "epoch": 2.47, + "learning_rate": 8.777891767119774e-06, + "loss": 0.1859, + "step": 64340 + }, + { + "epoch": 2.47, + "learning_rate": 8.777250577071044e-06, + "loss": 0.1672, + "step": 64341 + }, + { + "epoch": 2.47, + "learning_rate": 8.776609387022314e-06, + "loss": 0.2499, + "step": 64342 + }, + { + "epoch": 2.47, + "learning_rate": 8.775968196973583e-06, + "loss": 0.3244, + "step": 64343 + }, + { + "epoch": 2.47, + "learning_rate": 8.775327006924853e-06, + "loss": 0.1999, + "step": 64344 + }, + { + "epoch": 2.47, + "learning_rate": 8.774685816876123e-06, + "loss": 0.234, + "step": 64345 + }, + { + "epoch": 2.47, + "learning_rate": 8.774044626827393e-06, + "loss": 0.2991, + "step": 64346 + }, + { + "epoch": 2.47, + "learning_rate": 8.773403436778661e-06, + "loss": 0.3414, + "step": 64347 + }, + { + "epoch": 2.47, + "learning_rate": 8.772762246729931e-06, + "loss": 0.171, + "step": 64348 + }, + { + "epoch": 2.47, + "learning_rate": 8.772121056681201e-06, + "loss": 0.2295, + "step": 64349 + }, + { + "epoch": 2.47, + "learning_rate": 8.771479866632471e-06, + "loss": 0.2706, + "step": 64350 + }, + { + "epoch": 2.47, + "learning_rate": 8.77083867658374e-06, + "loss": 0.2588, + "step": 64351 + }, + { + "epoch": 2.47, + "learning_rate": 8.77019748653501e-06, + "loss": 0.161, + "step": 64352 + }, + { + "epoch": 2.47, + "learning_rate": 8.76955629648628e-06, + "loss": 0.2209, + "step": 64353 + }, + { + "epoch": 2.47, + "learning_rate": 8.76891510643755e-06, + "loss": 0.302, + "step": 64354 + }, + { + "epoch": 2.47, + "learning_rate": 8.768273916388818e-06, + "loss": 0.2654, + "step": 64355 + }, + { + "epoch": 2.47, + "learning_rate": 8.767632726340088e-06, + "loss": 0.1373, + "step": 64356 + }, + { + "epoch": 2.47, + "learning_rate": 8.766991536291356e-06, + "loss": 0.4266, + "step": 64357 + }, + { + "epoch": 2.47, + "learning_rate": 8.766350346242626e-06, + "loss": 0.2144, + "step": 64358 + }, + { + "epoch": 2.47, + "learning_rate": 8.765709156193896e-06, + "loss": 0.199, + "step": 64359 + }, + { + "epoch": 2.47, + "learning_rate": 8.765067966145166e-06, + "loss": 0.1819, + "step": 64360 + }, + { + "epoch": 2.47, + "learning_rate": 8.764426776096435e-06, + "loss": 0.3362, + "step": 64361 + }, + { + "epoch": 2.47, + "learning_rate": 8.763785586047705e-06, + "loss": 0.229, + "step": 64362 + }, + { + "epoch": 2.47, + "learning_rate": 8.763144395998975e-06, + "loss": 0.1683, + "step": 64363 + }, + { + "epoch": 2.47, + "learning_rate": 8.762503205950245e-06, + "loss": 0.2464, + "step": 64364 + }, + { + "epoch": 2.47, + "learning_rate": 8.761862015901513e-06, + "loss": 0.3877, + "step": 64365 + }, + { + "epoch": 2.47, + "learning_rate": 8.761220825852783e-06, + "loss": 0.1717, + "step": 64366 + }, + { + "epoch": 2.47, + "learning_rate": 8.760579635804053e-06, + "loss": 0.2395, + "step": 64367 + }, + { + "epoch": 2.47, + "learning_rate": 8.759938445755323e-06, + "loss": 0.1483, + "step": 64368 + }, + { + "epoch": 2.47, + "learning_rate": 8.759297255706592e-06, + "loss": 0.2761, + "step": 64369 + }, + { + "epoch": 2.47, + "learning_rate": 8.758656065657862e-06, + "loss": 0.2142, + "step": 64370 + }, + { + "epoch": 2.47, + "learning_rate": 8.758014875609132e-06, + "loss": 0.3019, + "step": 64371 + }, + { + "epoch": 2.47, + "learning_rate": 8.757373685560402e-06, + "loss": 0.1957, + "step": 64372 + }, + { + "epoch": 2.47, + "learning_rate": 8.75673249551167e-06, + "loss": 0.1467, + "step": 64373 + }, + { + "epoch": 2.47, + "learning_rate": 8.75609130546294e-06, + "loss": 0.2323, + "step": 64374 + }, + { + "epoch": 2.48, + "learning_rate": 8.755450115414208e-06, + "loss": 0.1282, + "step": 64375 + }, + { + "epoch": 2.48, + "learning_rate": 8.754808925365478e-06, + "loss": 0.209, + "step": 64376 + }, + { + "epoch": 2.48, + "learning_rate": 8.754167735316748e-06, + "loss": 0.1346, + "step": 64377 + }, + { + "epoch": 2.48, + "learning_rate": 8.753526545268018e-06, + "loss": 0.1774, + "step": 64378 + }, + { + "epoch": 2.48, + "learning_rate": 8.752885355219287e-06, + "loss": 0.2272, + "step": 64379 + }, + { + "epoch": 2.48, + "learning_rate": 8.752244165170557e-06, + "loss": 0.1776, + "step": 64380 + }, + { + "epoch": 2.48, + "learning_rate": 8.751602975121827e-06, + "loss": 0.0894, + "step": 64381 + }, + { + "epoch": 2.48, + "learning_rate": 8.750961785073097e-06, + "loss": 0.1898, + "step": 64382 + }, + { + "epoch": 2.48, + "learning_rate": 8.750320595024365e-06, + "loss": 0.2912, + "step": 64383 + }, + { + "epoch": 2.48, + "learning_rate": 8.749679404975635e-06, + "loss": 0.1839, + "step": 64384 + }, + { + "epoch": 2.48, + "learning_rate": 8.749038214926905e-06, + "loss": 0.1943, + "step": 64385 + }, + { + "epoch": 2.48, + "learning_rate": 8.748397024878175e-06, + "loss": 0.2965, + "step": 64386 + }, + { + "epoch": 2.48, + "learning_rate": 8.747755834829443e-06, + "loss": 0.1971, + "step": 64387 + }, + { + "epoch": 2.48, + "learning_rate": 8.747114644780713e-06, + "loss": 0.171, + "step": 64388 + }, + { + "epoch": 2.48, + "learning_rate": 8.746473454731984e-06, + "loss": 0.1416, + "step": 64389 + }, + { + "epoch": 2.48, + "learning_rate": 8.745832264683254e-06, + "loss": 0.212, + "step": 64390 + }, + { + "epoch": 2.48, + "learning_rate": 8.745191074634522e-06, + "loss": 0.1955, + "step": 64391 + }, + { + "epoch": 2.48, + "learning_rate": 8.744549884585792e-06, + "loss": 0.2185, + "step": 64392 + }, + { + "epoch": 2.48, + "learning_rate": 8.743908694537062e-06, + "loss": 0.1873, + "step": 64393 + }, + { + "epoch": 2.48, + "learning_rate": 8.743267504488332e-06, + "loss": 0.2537, + "step": 64394 + }, + { + "epoch": 2.48, + "learning_rate": 8.7426263144396e-06, + "loss": 0.2206, + "step": 64395 + }, + { + "epoch": 2.48, + "learning_rate": 8.74198512439087e-06, + "loss": 0.2248, + "step": 64396 + }, + { + "epoch": 2.48, + "learning_rate": 8.741343934342139e-06, + "loss": 0.175, + "step": 64397 + }, + { + "epoch": 2.48, + "learning_rate": 8.740702744293409e-06, + "loss": 0.1949, + "step": 64398 + }, + { + "epoch": 2.48, + "learning_rate": 8.740061554244679e-06, + "loss": 0.1991, + "step": 64399 + }, + { + "epoch": 2.48, + "learning_rate": 8.739420364195949e-06, + "loss": 0.1683, + "step": 64400 + }, + { + "epoch": 2.48, + "learning_rate": 8.738779174147217e-06, + "loss": 0.1319, + "step": 64401 + }, + { + "epoch": 2.48, + "learning_rate": 8.738137984098487e-06, + "loss": 0.1987, + "step": 64402 + }, + { + "epoch": 2.48, + "learning_rate": 8.737496794049757e-06, + "loss": 0.2063, + "step": 64403 + }, + { + "epoch": 2.48, + "learning_rate": 8.736855604001027e-06, + "loss": 0.2244, + "step": 64404 + }, + { + "epoch": 2.48, + "learning_rate": 8.736214413952295e-06, + "loss": 0.1791, + "step": 64405 + }, + { + "epoch": 2.48, + "learning_rate": 8.735573223903565e-06, + "loss": 0.3333, + "step": 64406 + }, + { + "epoch": 2.48, + "learning_rate": 8.734932033854835e-06, + "loss": 0.278, + "step": 64407 + }, + { + "epoch": 2.48, + "learning_rate": 8.734290843806105e-06, + "loss": 0.1963, + "step": 64408 + }, + { + "epoch": 2.48, + "learning_rate": 8.733649653757374e-06, + "loss": 0.1506, + "step": 64409 + }, + { + "epoch": 2.48, + "learning_rate": 8.733008463708644e-06, + "loss": 0.2927, + "step": 64410 + }, + { + "epoch": 2.48, + "learning_rate": 8.732367273659914e-06, + "loss": 0.1948, + "step": 64411 + }, + { + "epoch": 2.48, + "learning_rate": 8.731726083611184e-06, + "loss": 0.2082, + "step": 64412 + }, + { + "epoch": 2.48, + "learning_rate": 8.731084893562452e-06, + "loss": 0.1124, + "step": 64413 + }, + { + "epoch": 2.48, + "learning_rate": 8.730443703513722e-06, + "loss": 0.2302, + "step": 64414 + }, + { + "epoch": 2.48, + "learning_rate": 8.72980251346499e-06, + "loss": 0.1783, + "step": 64415 + }, + { + "epoch": 2.48, + "learning_rate": 8.729161323416262e-06, + "loss": 0.3071, + "step": 64416 + }, + { + "epoch": 2.48, + "learning_rate": 8.72852013336753e-06, + "loss": 0.2029, + "step": 64417 + }, + { + "epoch": 2.48, + "learning_rate": 8.7278789433188e-06, + "loss": 0.2218, + "step": 64418 + }, + { + "epoch": 2.48, + "learning_rate": 8.727237753270069e-06, + "loss": 0.2176, + "step": 64419 + }, + { + "epoch": 2.48, + "learning_rate": 8.726596563221339e-06, + "loss": 0.2636, + "step": 64420 + }, + { + "epoch": 2.48, + "learning_rate": 8.725955373172609e-06, + "loss": 0.2776, + "step": 64421 + }, + { + "epoch": 2.48, + "learning_rate": 8.725314183123879e-06, + "loss": 0.118, + "step": 64422 + }, + { + "epoch": 2.48, + "learning_rate": 8.724672993075147e-06, + "loss": 0.1328, + "step": 64423 + }, + { + "epoch": 2.48, + "learning_rate": 8.724031803026417e-06, + "loss": 0.2151, + "step": 64424 + }, + { + "epoch": 2.48, + "learning_rate": 8.723390612977687e-06, + "loss": 0.2383, + "step": 64425 + }, + { + "epoch": 2.48, + "learning_rate": 8.722749422928957e-06, + "loss": 0.2591, + "step": 64426 + }, + { + "epoch": 2.48, + "learning_rate": 8.722108232880226e-06, + "loss": 0.1355, + "step": 64427 + }, + { + "epoch": 2.48, + "learning_rate": 8.721467042831496e-06, + "loss": 0.2131, + "step": 64428 + }, + { + "epoch": 2.48, + "learning_rate": 8.720825852782766e-06, + "loss": 0.3572, + "step": 64429 + }, + { + "epoch": 2.48, + "learning_rate": 8.720184662734036e-06, + "loss": 0.1763, + "step": 64430 + }, + { + "epoch": 2.48, + "learning_rate": 8.719543472685304e-06, + "loss": 0.2254, + "step": 64431 + }, + { + "epoch": 2.48, + "learning_rate": 8.718902282636574e-06, + "loss": 0.2114, + "step": 64432 + }, + { + "epoch": 2.48, + "learning_rate": 8.718261092587844e-06, + "loss": 0.1887, + "step": 64433 + }, + { + "epoch": 2.48, + "learning_rate": 8.717619902539114e-06, + "loss": 0.3654, + "step": 64434 + }, + { + "epoch": 2.48, + "learning_rate": 8.716978712490383e-06, + "loss": 0.1586, + "step": 64435 + }, + { + "epoch": 2.48, + "learning_rate": 8.716337522441653e-06, + "loss": 0.1558, + "step": 64436 + }, + { + "epoch": 2.48, + "learning_rate": 8.715696332392921e-06, + "loss": 0.316, + "step": 64437 + }, + { + "epoch": 2.48, + "learning_rate": 8.715055142344191e-06, + "loss": 0.3497, + "step": 64438 + }, + { + "epoch": 2.48, + "learning_rate": 8.714413952295461e-06, + "loss": 0.3313, + "step": 64439 + }, + { + "epoch": 2.48, + "learning_rate": 8.713772762246731e-06, + "loss": 0.1627, + "step": 64440 + }, + { + "epoch": 2.48, + "learning_rate": 8.713131572198e-06, + "loss": 0.3163, + "step": 64441 + }, + { + "epoch": 2.48, + "learning_rate": 8.71249038214927e-06, + "loss": 0.1544, + "step": 64442 + }, + { + "epoch": 2.48, + "learning_rate": 8.71184919210054e-06, + "loss": 0.3892, + "step": 64443 + }, + { + "epoch": 2.48, + "learning_rate": 8.71120800205181e-06, + "loss": 0.2345, + "step": 64444 + }, + { + "epoch": 2.48, + "learning_rate": 8.710566812003078e-06, + "loss": 0.2211, + "step": 64445 + }, + { + "epoch": 2.48, + "learning_rate": 8.709925621954348e-06, + "loss": 0.2001, + "step": 64446 + }, + { + "epoch": 2.48, + "learning_rate": 8.709284431905618e-06, + "loss": 0.3095, + "step": 64447 + }, + { + "epoch": 2.48, + "learning_rate": 8.708643241856888e-06, + "loss": 0.1879, + "step": 64448 + }, + { + "epoch": 2.48, + "learning_rate": 8.708002051808156e-06, + "loss": 0.1933, + "step": 64449 + }, + { + "epoch": 2.48, + "learning_rate": 8.707360861759426e-06, + "loss": 0.1807, + "step": 64450 + }, + { + "epoch": 2.48, + "learning_rate": 8.706719671710696e-06, + "loss": 0.2551, + "step": 64451 + }, + { + "epoch": 2.48, + "learning_rate": 8.706078481661966e-06, + "loss": 0.3171, + "step": 64452 + }, + { + "epoch": 2.48, + "learning_rate": 8.705437291613234e-06, + "loss": 0.1887, + "step": 64453 + }, + { + "epoch": 2.48, + "learning_rate": 8.704796101564504e-06, + "loss": 0.2988, + "step": 64454 + }, + { + "epoch": 2.48, + "learning_rate": 8.704154911515773e-06, + "loss": 0.2528, + "step": 64455 + }, + { + "epoch": 2.48, + "learning_rate": 8.703513721467045e-06, + "loss": 0.2263, + "step": 64456 + }, + { + "epoch": 2.48, + "learning_rate": 8.702872531418313e-06, + "loss": 0.3113, + "step": 64457 + }, + { + "epoch": 2.48, + "learning_rate": 8.702231341369583e-06, + "loss": 0.2799, + "step": 64458 + }, + { + "epoch": 2.48, + "learning_rate": 8.701590151320851e-06, + "loss": 0.1207, + "step": 64459 + }, + { + "epoch": 2.48, + "learning_rate": 8.700948961272121e-06, + "loss": 0.3328, + "step": 64460 + }, + { + "epoch": 2.48, + "learning_rate": 8.700307771223391e-06, + "loss": 0.2779, + "step": 64461 + }, + { + "epoch": 2.48, + "learning_rate": 8.699666581174661e-06, + "loss": 0.2996, + "step": 64462 + }, + { + "epoch": 2.48, + "learning_rate": 8.69902539112593e-06, + "loss": 0.3283, + "step": 64463 + }, + { + "epoch": 2.48, + "learning_rate": 8.6983842010772e-06, + "loss": 0.1889, + "step": 64464 + }, + { + "epoch": 2.48, + "learning_rate": 8.69774301102847e-06, + "loss": 0.3048, + "step": 64465 + }, + { + "epoch": 2.48, + "learning_rate": 8.69710182097974e-06, + "loss": 0.187, + "step": 64466 + }, + { + "epoch": 2.48, + "learning_rate": 8.696460630931008e-06, + "loss": 0.1484, + "step": 64467 + }, + { + "epoch": 2.48, + "learning_rate": 8.695819440882278e-06, + "loss": 0.1165, + "step": 64468 + }, + { + "epoch": 2.48, + "learning_rate": 8.695178250833548e-06, + "loss": 0.2572, + "step": 64469 + }, + { + "epoch": 2.48, + "learning_rate": 8.694537060784818e-06, + "loss": 0.1685, + "step": 64470 + }, + { + "epoch": 2.48, + "learning_rate": 8.693895870736086e-06, + "loss": 0.2793, + "step": 64471 + }, + { + "epoch": 2.48, + "learning_rate": 8.693254680687356e-06, + "loss": 0.2603, + "step": 64472 + }, + { + "epoch": 2.48, + "learning_rate": 8.692613490638625e-06, + "loss": 0.1331, + "step": 64473 + }, + { + "epoch": 2.48, + "learning_rate": 8.691972300589896e-06, + "loss": 0.175, + "step": 64474 + }, + { + "epoch": 2.48, + "learning_rate": 8.691331110541165e-06, + "loss": 0.1668, + "step": 64475 + }, + { + "epoch": 2.48, + "learning_rate": 8.690689920492435e-06, + "loss": 0.2362, + "step": 64476 + }, + { + "epoch": 2.48, + "learning_rate": 8.690048730443703e-06, + "loss": 0.1701, + "step": 64477 + }, + { + "epoch": 2.48, + "learning_rate": 8.689407540394973e-06, + "loss": 0.1958, + "step": 64478 + }, + { + "epoch": 2.48, + "learning_rate": 8.688766350346243e-06, + "loss": 0.2071, + "step": 64479 + }, + { + "epoch": 2.48, + "learning_rate": 8.688125160297513e-06, + "loss": 0.2084, + "step": 64480 + }, + { + "epoch": 2.48, + "learning_rate": 8.687483970248782e-06, + "loss": 0.1408, + "step": 64481 + }, + { + "epoch": 2.48, + "learning_rate": 8.686842780200052e-06, + "loss": 0.1766, + "step": 64482 + }, + { + "epoch": 2.48, + "learning_rate": 8.686201590151322e-06, + "loss": 0.1006, + "step": 64483 + }, + { + "epoch": 2.48, + "learning_rate": 8.685560400102592e-06, + "loss": 0.2523, + "step": 64484 + }, + { + "epoch": 2.48, + "learning_rate": 8.68491921005386e-06, + "loss": 0.2665, + "step": 64485 + }, + { + "epoch": 2.48, + "learning_rate": 8.68427802000513e-06, + "loss": 0.1677, + "step": 64486 + }, + { + "epoch": 2.48, + "learning_rate": 8.6836368299564e-06, + "loss": 0.2967, + "step": 64487 + }, + { + "epoch": 2.48, + "learning_rate": 8.68299563990767e-06, + "loss": 0.1652, + "step": 64488 + }, + { + "epoch": 2.48, + "learning_rate": 8.682354449858938e-06, + "loss": 0.3122, + "step": 64489 + }, + { + "epoch": 2.48, + "learning_rate": 8.681713259810208e-06, + "loss": 0.1346, + "step": 64490 + }, + { + "epoch": 2.48, + "learning_rate": 8.681072069761478e-06, + "loss": 0.1819, + "step": 64491 + }, + { + "epoch": 2.48, + "learning_rate": 8.680430879712748e-06, + "loss": 0.1585, + "step": 64492 + }, + { + "epoch": 2.48, + "learning_rate": 8.679789689664017e-06, + "loss": 0.1208, + "step": 64493 + }, + { + "epoch": 2.48, + "learning_rate": 8.679148499615287e-06, + "loss": 0.2192, + "step": 64494 + }, + { + "epoch": 2.48, + "learning_rate": 8.678507309566555e-06, + "loss": 0.1459, + "step": 64495 + }, + { + "epoch": 2.48, + "learning_rate": 8.677866119517827e-06, + "loss": 0.224, + "step": 64496 + }, + { + "epoch": 2.48, + "learning_rate": 8.677224929469095e-06, + "loss": 0.2092, + "step": 64497 + }, + { + "epoch": 2.48, + "learning_rate": 8.676583739420365e-06, + "loss": 0.1801, + "step": 64498 + }, + { + "epoch": 2.48, + "learning_rate": 8.675942549371633e-06, + "loss": 0.1789, + "step": 64499 + }, + { + "epoch": 2.48, + "learning_rate": 8.675301359322903e-06, + "loss": 0.2404, + "step": 64500 + }, + { + "epoch": 2.48, + "learning_rate": 8.674660169274174e-06, + "loss": 0.2343, + "step": 64501 + }, + { + "epoch": 2.48, + "learning_rate": 8.674018979225444e-06, + "loss": 0.1955, + "step": 64502 + }, + { + "epoch": 2.48, + "learning_rate": 8.673377789176712e-06, + "loss": 0.2317, + "step": 64503 + }, + { + "epoch": 2.48, + "learning_rate": 8.672736599127982e-06, + "loss": 0.218, + "step": 64504 + }, + { + "epoch": 2.48, + "learning_rate": 8.672095409079252e-06, + "loss": 0.2113, + "step": 64505 + }, + { + "epoch": 2.48, + "learning_rate": 8.671454219030522e-06, + "loss": 0.1751, + "step": 64506 + }, + { + "epoch": 2.48, + "learning_rate": 8.67081302898179e-06, + "loss": 0.1418, + "step": 64507 + }, + { + "epoch": 2.48, + "learning_rate": 8.67017183893306e-06, + "loss": 0.273, + "step": 64508 + }, + { + "epoch": 2.48, + "learning_rate": 8.66953064888433e-06, + "loss": 0.1778, + "step": 64509 + }, + { + "epoch": 2.48, + "learning_rate": 8.6688894588356e-06, + "loss": 0.1773, + "step": 64510 + }, + { + "epoch": 2.48, + "learning_rate": 8.668248268786869e-06, + "loss": 0.1417, + "step": 64511 + }, + { + "epoch": 2.48, + "learning_rate": 8.667607078738139e-06, + "loss": 0.157, + "step": 64512 + }, + { + "epoch": 2.48, + "learning_rate": 8.666965888689407e-06, + "loss": 0.2267, + "step": 64513 + }, + { + "epoch": 2.48, + "learning_rate": 8.666324698640679e-06, + "loss": 0.14, + "step": 64514 + }, + { + "epoch": 2.48, + "learning_rate": 8.665683508591947e-06, + "loss": 0.3246, + "step": 64515 + }, + { + "epoch": 2.48, + "learning_rate": 8.665042318543217e-06, + "loss": 0.1435, + "step": 64516 + }, + { + "epoch": 2.48, + "learning_rate": 8.664401128494485e-06, + "loss": 0.341, + "step": 64517 + }, + { + "epoch": 2.48, + "learning_rate": 8.663759938445757e-06, + "loss": 0.3102, + "step": 64518 + }, + { + "epoch": 2.48, + "learning_rate": 8.663118748397025e-06, + "loss": 0.0905, + "step": 64519 + }, + { + "epoch": 2.48, + "learning_rate": 8.662477558348295e-06, + "loss": 0.4133, + "step": 64520 + }, + { + "epoch": 2.48, + "learning_rate": 8.661836368299564e-06, + "loss": 0.2148, + "step": 64521 + }, + { + "epoch": 2.48, + "learning_rate": 8.661195178250834e-06, + "loss": 0.0946, + "step": 64522 + }, + { + "epoch": 2.48, + "learning_rate": 8.660553988202104e-06, + "loss": 0.181, + "step": 64523 + }, + { + "epoch": 2.48, + "learning_rate": 8.659912798153374e-06, + "loss": 0.223, + "step": 64524 + }, + { + "epoch": 2.48, + "learning_rate": 8.659271608104642e-06, + "loss": 0.1585, + "step": 64525 + }, + { + "epoch": 2.48, + "learning_rate": 8.658630418055912e-06, + "loss": 0.251, + "step": 64526 + }, + { + "epoch": 2.48, + "learning_rate": 8.657989228007182e-06, + "loss": 0.3009, + "step": 64527 + }, + { + "epoch": 2.48, + "learning_rate": 8.657348037958452e-06, + "loss": 0.2517, + "step": 64528 + }, + { + "epoch": 2.48, + "learning_rate": 8.65670684790972e-06, + "loss": 0.1551, + "step": 64529 + }, + { + "epoch": 2.48, + "learning_rate": 8.65606565786099e-06, + "loss": 0.2527, + "step": 64530 + }, + { + "epoch": 2.48, + "learning_rate": 8.65542446781226e-06, + "loss": 0.0983, + "step": 64531 + }, + { + "epoch": 2.48, + "learning_rate": 8.65478327776353e-06, + "loss": 0.2966, + "step": 64532 + }, + { + "epoch": 2.48, + "learning_rate": 8.654142087714799e-06, + "loss": 0.2098, + "step": 64533 + }, + { + "epoch": 2.48, + "learning_rate": 8.653500897666069e-06, + "loss": 0.1835, + "step": 64534 + }, + { + "epoch": 2.48, + "learning_rate": 8.652859707617337e-06, + "loss": 0.2857, + "step": 64535 + }, + { + "epoch": 2.48, + "learning_rate": 8.652218517568609e-06, + "loss": 0.2214, + "step": 64536 + }, + { + "epoch": 2.48, + "learning_rate": 8.651577327519877e-06, + "loss": 0.2079, + "step": 64537 + }, + { + "epoch": 2.48, + "learning_rate": 8.650936137471147e-06, + "loss": 0.2101, + "step": 64538 + }, + { + "epoch": 2.48, + "learning_rate": 8.650294947422416e-06, + "loss": 0.0786, + "step": 64539 + }, + { + "epoch": 2.48, + "learning_rate": 8.649653757373686e-06, + "loss": 0.2157, + "step": 64540 + }, + { + "epoch": 2.48, + "learning_rate": 8.649012567324956e-06, + "loss": 0.2284, + "step": 64541 + }, + { + "epoch": 2.48, + "learning_rate": 8.648371377276226e-06, + "loss": 0.3308, + "step": 64542 + }, + { + "epoch": 2.48, + "learning_rate": 8.647730187227494e-06, + "loss": 0.166, + "step": 64543 + }, + { + "epoch": 2.48, + "learning_rate": 8.647088997178764e-06, + "loss": 0.1692, + "step": 64544 + }, + { + "epoch": 2.48, + "learning_rate": 8.646447807130034e-06, + "loss": 0.3498, + "step": 64545 + }, + { + "epoch": 2.48, + "learning_rate": 8.645806617081304e-06, + "loss": 0.1907, + "step": 64546 + }, + { + "epoch": 2.48, + "learning_rate": 8.645165427032573e-06, + "loss": 0.2036, + "step": 64547 + }, + { + "epoch": 2.48, + "learning_rate": 8.644524236983843e-06, + "loss": 0.3957, + "step": 64548 + }, + { + "epoch": 2.48, + "learning_rate": 8.643883046935113e-06, + "loss": 0.2765, + "step": 64549 + }, + { + "epoch": 2.48, + "learning_rate": 8.643241856886383e-06, + "loss": 0.2053, + "step": 64550 + }, + { + "epoch": 2.48, + "learning_rate": 8.642600666837651e-06, + "loss": 0.2347, + "step": 64551 + }, + { + "epoch": 2.48, + "learning_rate": 8.641959476788921e-06, + "loss": 0.1506, + "step": 64552 + }, + { + "epoch": 2.48, + "learning_rate": 8.64131828674019e-06, + "loss": 0.2277, + "step": 64553 + }, + { + "epoch": 2.48, + "learning_rate": 8.640677096691461e-06, + "loss": 0.2395, + "step": 64554 + }, + { + "epoch": 2.48, + "learning_rate": 8.64003590664273e-06, + "loss": 0.2017, + "step": 64555 + }, + { + "epoch": 2.48, + "learning_rate": 8.639394716594e-06, + "loss": 0.2732, + "step": 64556 + }, + { + "epoch": 2.48, + "learning_rate": 8.638753526545268e-06, + "loss": 0.1483, + "step": 64557 + }, + { + "epoch": 2.48, + "learning_rate": 8.63811233649654e-06, + "loss": 0.1754, + "step": 64558 + }, + { + "epoch": 2.48, + "learning_rate": 8.637471146447808e-06, + "loss": 0.179, + "step": 64559 + }, + { + "epoch": 2.48, + "learning_rate": 8.636829956399078e-06, + "loss": 0.1989, + "step": 64560 + }, + { + "epoch": 2.48, + "learning_rate": 8.636188766350346e-06, + "loss": 0.2276, + "step": 64561 + }, + { + "epoch": 2.48, + "learning_rate": 8.635547576301616e-06, + "loss": 0.2818, + "step": 64562 + }, + { + "epoch": 2.48, + "learning_rate": 8.634906386252886e-06, + "loss": 0.2241, + "step": 64563 + }, + { + "epoch": 2.48, + "learning_rate": 8.634265196204156e-06, + "loss": 0.1711, + "step": 64564 + }, + { + "epoch": 2.48, + "learning_rate": 8.633624006155424e-06, + "loss": 0.1681, + "step": 64565 + }, + { + "epoch": 2.48, + "learning_rate": 8.632982816106694e-06, + "loss": 0.1759, + "step": 64566 + }, + { + "epoch": 2.48, + "learning_rate": 8.632341626057965e-06, + "loss": 0.2509, + "step": 64567 + }, + { + "epoch": 2.48, + "learning_rate": 8.631700436009235e-06, + "loss": 0.3397, + "step": 64568 + }, + { + "epoch": 2.48, + "learning_rate": 8.631059245960503e-06, + "loss": 0.1272, + "step": 64569 + }, + { + "epoch": 2.48, + "learning_rate": 8.630418055911773e-06, + "loss": 0.1915, + "step": 64570 + }, + { + "epoch": 2.48, + "learning_rate": 8.629776865863043e-06, + "loss": 0.229, + "step": 64571 + }, + { + "epoch": 2.48, + "learning_rate": 8.629135675814313e-06, + "loss": 0.1774, + "step": 64572 + }, + { + "epoch": 2.48, + "learning_rate": 8.628494485765581e-06, + "loss": 0.2261, + "step": 64573 + }, + { + "epoch": 2.48, + "learning_rate": 8.627853295716851e-06, + "loss": 0.1832, + "step": 64574 + }, + { + "epoch": 2.48, + "learning_rate": 8.62721210566812e-06, + "loss": 0.2479, + "step": 64575 + }, + { + "epoch": 2.48, + "learning_rate": 8.626570915619391e-06, + "loss": 0.2312, + "step": 64576 + }, + { + "epoch": 2.48, + "learning_rate": 8.62592972557066e-06, + "loss": 0.3193, + "step": 64577 + }, + { + "epoch": 2.48, + "learning_rate": 8.62528853552193e-06, + "loss": 0.1598, + "step": 64578 + }, + { + "epoch": 2.48, + "learning_rate": 8.624647345473198e-06, + "loss": 0.302, + "step": 64579 + }, + { + "epoch": 2.48, + "learning_rate": 8.624006155424468e-06, + "loss": 0.1601, + "step": 64580 + }, + { + "epoch": 2.48, + "learning_rate": 8.623364965375738e-06, + "loss": 0.2121, + "step": 64581 + }, + { + "epoch": 2.48, + "learning_rate": 8.622723775327008e-06, + "loss": 0.1741, + "step": 64582 + }, + { + "epoch": 2.48, + "learning_rate": 8.622082585278276e-06, + "loss": 0.1656, + "step": 64583 + }, + { + "epoch": 2.48, + "learning_rate": 8.621441395229546e-06, + "loss": 0.1053, + "step": 64584 + }, + { + "epoch": 2.48, + "learning_rate": 8.620800205180816e-06, + "loss": 0.1555, + "step": 64585 + }, + { + "epoch": 2.48, + "learning_rate": 8.620159015132086e-06, + "loss": 0.114, + "step": 64586 + }, + { + "epoch": 2.48, + "learning_rate": 8.619517825083355e-06, + "loss": 0.2027, + "step": 64587 + }, + { + "epoch": 2.48, + "learning_rate": 8.618876635034625e-06, + "loss": 0.2116, + "step": 64588 + }, + { + "epoch": 2.48, + "learning_rate": 8.618235444985895e-06, + "loss": 0.2856, + "step": 64589 + }, + { + "epoch": 2.48, + "learning_rate": 8.617594254937165e-06, + "loss": 0.2278, + "step": 64590 + }, + { + "epoch": 2.48, + "learning_rate": 8.616953064888433e-06, + "loss": 0.1637, + "step": 64591 + }, + { + "epoch": 2.48, + "learning_rate": 8.616311874839703e-06, + "loss": 0.2717, + "step": 64592 + }, + { + "epoch": 2.48, + "learning_rate": 8.615670684790972e-06, + "loss": 0.2732, + "step": 64593 + }, + { + "epoch": 2.48, + "learning_rate": 8.615029494742243e-06, + "loss": 0.1207, + "step": 64594 + }, + { + "epoch": 2.48, + "learning_rate": 8.614388304693512e-06, + "loss": 0.1464, + "step": 64595 + }, + { + "epoch": 2.48, + "learning_rate": 8.613747114644782e-06, + "loss": 0.2638, + "step": 64596 + }, + { + "epoch": 2.48, + "learning_rate": 8.61310592459605e-06, + "loss": 0.1851, + "step": 64597 + }, + { + "epoch": 2.48, + "learning_rate": 8.612464734547322e-06, + "loss": 0.1098, + "step": 64598 + }, + { + "epoch": 2.48, + "learning_rate": 8.61182354449859e-06, + "loss": 0.2538, + "step": 64599 + }, + { + "epoch": 2.48, + "learning_rate": 8.61118235444986e-06, + "loss": 0.2956, + "step": 64600 + }, + { + "epoch": 2.48, + "learning_rate": 8.610541164401128e-06, + "loss": 0.2784, + "step": 64601 + }, + { + "epoch": 2.48, + "learning_rate": 8.609899974352398e-06, + "loss": 0.3041, + "step": 64602 + }, + { + "epoch": 2.48, + "learning_rate": 8.609258784303668e-06, + "loss": 0.1463, + "step": 64603 + }, + { + "epoch": 2.48, + "learning_rate": 8.608617594254938e-06, + "loss": 0.3426, + "step": 64604 + }, + { + "epoch": 2.48, + "learning_rate": 8.607976404206207e-06, + "loss": 0.2846, + "step": 64605 + }, + { + "epoch": 2.48, + "learning_rate": 8.607335214157477e-06, + "loss": 0.1729, + "step": 64606 + }, + { + "epoch": 2.48, + "learning_rate": 8.606694024108747e-06, + "loss": 0.154, + "step": 64607 + }, + { + "epoch": 2.48, + "learning_rate": 8.606052834060017e-06, + "loss": 0.3089, + "step": 64608 + }, + { + "epoch": 2.48, + "learning_rate": 8.605411644011285e-06, + "loss": 0.2466, + "step": 64609 + }, + { + "epoch": 2.48, + "learning_rate": 8.604770453962555e-06, + "loss": 0.2158, + "step": 64610 + }, + { + "epoch": 2.48, + "learning_rate": 8.604129263913823e-06, + "loss": 0.2375, + "step": 64611 + }, + { + "epoch": 2.48, + "learning_rate": 8.603488073865095e-06, + "loss": 0.2423, + "step": 64612 + }, + { + "epoch": 2.48, + "learning_rate": 8.602846883816364e-06, + "loss": 0.1735, + "step": 64613 + }, + { + "epoch": 2.48, + "learning_rate": 8.602205693767634e-06, + "loss": 0.3166, + "step": 64614 + }, + { + "epoch": 2.48, + "learning_rate": 8.601564503718902e-06, + "loss": 0.1711, + "step": 64615 + }, + { + "epoch": 2.48, + "learning_rate": 8.600923313670174e-06, + "loss": 0.3419, + "step": 64616 + }, + { + "epoch": 2.48, + "learning_rate": 8.600282123621442e-06, + "loss": 0.2022, + "step": 64617 + }, + { + "epoch": 2.48, + "learning_rate": 8.599640933572712e-06, + "loss": 0.3474, + "step": 64618 + }, + { + "epoch": 2.48, + "learning_rate": 8.59899974352398e-06, + "loss": 0.2129, + "step": 64619 + }, + { + "epoch": 2.48, + "learning_rate": 8.598358553475252e-06, + "loss": 0.2429, + "step": 64620 + }, + { + "epoch": 2.48, + "learning_rate": 8.59771736342652e-06, + "loss": 0.2569, + "step": 64621 + }, + { + "epoch": 2.48, + "learning_rate": 8.59707617337779e-06, + "loss": 0.0761, + "step": 64622 + }, + { + "epoch": 2.48, + "learning_rate": 8.596434983329059e-06, + "loss": 0.267, + "step": 64623 + }, + { + "epoch": 2.48, + "learning_rate": 8.595793793280329e-06, + "loss": 0.1312, + "step": 64624 + }, + { + "epoch": 2.48, + "learning_rate": 8.595152603231599e-06, + "loss": 0.3188, + "step": 64625 + }, + { + "epoch": 2.48, + "learning_rate": 8.594511413182869e-06, + "loss": 0.3048, + "step": 64626 + }, + { + "epoch": 2.48, + "learning_rate": 8.593870223134137e-06, + "loss": 0.306, + "step": 64627 + }, + { + "epoch": 2.48, + "learning_rate": 8.593229033085407e-06, + "loss": 0.1456, + "step": 64628 + }, + { + "epoch": 2.48, + "learning_rate": 8.592587843036677e-06, + "loss": 0.1447, + "step": 64629 + }, + { + "epoch": 2.48, + "learning_rate": 8.591946652987947e-06, + "loss": 0.1848, + "step": 64630 + }, + { + "epoch": 2.48, + "learning_rate": 8.591305462939215e-06, + "loss": 0.1263, + "step": 64631 + }, + { + "epoch": 2.48, + "learning_rate": 8.590664272890485e-06, + "loss": 0.2488, + "step": 64632 + }, + { + "epoch": 2.48, + "learning_rate": 8.590023082841754e-06, + "loss": 0.2029, + "step": 64633 + }, + { + "epoch": 2.48, + "learning_rate": 8.589381892793026e-06, + "loss": 0.3635, + "step": 64634 + }, + { + "epoch": 2.49, + "learning_rate": 8.588740702744294e-06, + "loss": 0.176, + "step": 64635 + }, + { + "epoch": 2.49, + "learning_rate": 8.588099512695564e-06, + "loss": 0.1175, + "step": 64636 + }, + { + "epoch": 2.49, + "learning_rate": 8.587458322646832e-06, + "loss": 0.2697, + "step": 64637 + }, + { + "epoch": 2.49, + "learning_rate": 8.586817132598104e-06, + "loss": 0.2152, + "step": 64638 + }, + { + "epoch": 2.49, + "learning_rate": 8.586175942549372e-06, + "loss": 0.1469, + "step": 64639 + }, + { + "epoch": 2.49, + "learning_rate": 8.585534752500642e-06, + "loss": 0.1372, + "step": 64640 + }, + { + "epoch": 2.49, + "learning_rate": 8.58489356245191e-06, + "loss": 0.2544, + "step": 64641 + }, + { + "epoch": 2.49, + "learning_rate": 8.58425237240318e-06, + "loss": 0.2381, + "step": 64642 + }, + { + "epoch": 2.49, + "learning_rate": 8.58361118235445e-06, + "loss": 0.1298, + "step": 64643 + }, + { + "epoch": 2.49, + "learning_rate": 8.58296999230572e-06, + "loss": 0.1219, + "step": 64644 + }, + { + "epoch": 2.49, + "learning_rate": 8.582328802256989e-06, + "loss": 0.1374, + "step": 64645 + }, + { + "epoch": 2.49, + "learning_rate": 8.581687612208259e-06, + "loss": 0.1291, + "step": 64646 + }, + { + "epoch": 2.49, + "learning_rate": 8.581046422159529e-06, + "loss": 0.2526, + "step": 64647 + }, + { + "epoch": 2.49, + "learning_rate": 8.580405232110799e-06, + "loss": 0.2874, + "step": 64648 + }, + { + "epoch": 2.49, + "learning_rate": 8.579764042062067e-06, + "loss": 0.2502, + "step": 64649 + }, + { + "epoch": 2.49, + "learning_rate": 8.579122852013337e-06, + "loss": 0.1473, + "step": 64650 + }, + { + "epoch": 2.49, + "learning_rate": 8.578481661964606e-06, + "loss": 0.2373, + "step": 64651 + }, + { + "epoch": 2.49, + "learning_rate": 8.577840471915877e-06, + "loss": 0.1705, + "step": 64652 + }, + { + "epoch": 2.49, + "learning_rate": 8.577199281867146e-06, + "loss": 0.138, + "step": 64653 + }, + { + "epoch": 2.49, + "learning_rate": 8.576558091818416e-06, + "loss": 0.1558, + "step": 64654 + }, + { + "epoch": 2.49, + "learning_rate": 8.575916901769684e-06, + "loss": 0.2354, + "step": 64655 + }, + { + "epoch": 2.49, + "learning_rate": 8.575275711720956e-06, + "loss": 0.3986, + "step": 64656 + }, + { + "epoch": 2.49, + "learning_rate": 8.574634521672224e-06, + "loss": 0.2461, + "step": 64657 + }, + { + "epoch": 2.49, + "learning_rate": 8.573993331623494e-06, + "loss": 0.2319, + "step": 64658 + }, + { + "epoch": 2.49, + "learning_rate": 8.573352141574763e-06, + "loss": 0.2463, + "step": 64659 + }, + { + "epoch": 2.49, + "learning_rate": 8.572710951526033e-06, + "loss": 0.3609, + "step": 64660 + }, + { + "epoch": 2.49, + "learning_rate": 8.572069761477303e-06, + "loss": 0.1936, + "step": 64661 + }, + { + "epoch": 2.49, + "learning_rate": 8.571428571428573e-06, + "loss": 0.4127, + "step": 64662 + }, + { + "epoch": 2.49, + "learning_rate": 8.570787381379841e-06, + "loss": 0.1923, + "step": 64663 + }, + { + "epoch": 2.49, + "learning_rate": 8.570146191331111e-06, + "loss": 0.2171, + "step": 64664 + }, + { + "epoch": 2.49, + "learning_rate": 8.569505001282381e-06, + "loss": 0.1436, + "step": 64665 + }, + { + "epoch": 2.49, + "learning_rate": 8.568863811233651e-06, + "loss": 0.1585, + "step": 64666 + }, + { + "epoch": 2.49, + "learning_rate": 8.56822262118492e-06, + "loss": 0.2886, + "step": 64667 + }, + { + "epoch": 2.49, + "learning_rate": 8.56758143113619e-06, + "loss": 0.1016, + "step": 64668 + }, + { + "epoch": 2.49, + "learning_rate": 8.56694024108746e-06, + "loss": 0.1697, + "step": 64669 + }, + { + "epoch": 2.49, + "learning_rate": 8.56629905103873e-06, + "loss": 0.0968, + "step": 64670 + }, + { + "epoch": 2.49, + "learning_rate": 8.565657860989998e-06, + "loss": 0.2788, + "step": 64671 + }, + { + "epoch": 2.49, + "learning_rate": 8.565016670941268e-06, + "loss": 0.2617, + "step": 64672 + }, + { + "epoch": 2.49, + "learning_rate": 8.564375480892536e-06, + "loss": 0.2421, + "step": 64673 + }, + { + "epoch": 2.49, + "learning_rate": 8.563734290843808e-06, + "loss": 0.2013, + "step": 64674 + }, + { + "epoch": 2.49, + "learning_rate": 8.563093100795076e-06, + "loss": 0.3092, + "step": 64675 + }, + { + "epoch": 2.49, + "learning_rate": 8.562451910746346e-06, + "loss": 0.2371, + "step": 64676 + }, + { + "epoch": 2.49, + "learning_rate": 8.561810720697614e-06, + "loss": 0.3233, + "step": 64677 + }, + { + "epoch": 2.49, + "learning_rate": 8.561169530648886e-06, + "loss": 0.2237, + "step": 64678 + }, + { + "epoch": 2.49, + "learning_rate": 8.560528340600155e-06, + "loss": 0.2083, + "step": 64679 + }, + { + "epoch": 2.49, + "learning_rate": 8.559887150551425e-06, + "loss": 0.0839, + "step": 64680 + }, + { + "epoch": 2.49, + "learning_rate": 8.559245960502693e-06, + "loss": 0.222, + "step": 64681 + }, + { + "epoch": 2.49, + "learning_rate": 8.558604770453963e-06, + "loss": 0.2139, + "step": 64682 + }, + { + "epoch": 2.49, + "learning_rate": 8.557963580405233e-06, + "loss": 0.1882, + "step": 64683 + }, + { + "epoch": 2.49, + "learning_rate": 8.557322390356503e-06, + "loss": 0.2098, + "step": 64684 + }, + { + "epoch": 2.49, + "learning_rate": 8.556681200307771e-06, + "loss": 0.1546, + "step": 64685 + }, + { + "epoch": 2.49, + "learning_rate": 8.556040010259041e-06, + "loss": 0.2324, + "step": 64686 + }, + { + "epoch": 2.49, + "learning_rate": 8.555398820210311e-06, + "loss": 0.2457, + "step": 64687 + }, + { + "epoch": 2.49, + "learning_rate": 8.554757630161581e-06, + "loss": 0.3004, + "step": 64688 + }, + { + "epoch": 2.49, + "learning_rate": 8.55411644011285e-06, + "loss": 0.3351, + "step": 64689 + }, + { + "epoch": 2.49, + "learning_rate": 8.55347525006412e-06, + "loss": 0.2369, + "step": 64690 + }, + { + "epoch": 2.49, + "learning_rate": 8.552834060015388e-06, + "loss": 0.3693, + "step": 64691 + }, + { + "epoch": 2.49, + "learning_rate": 8.55219286996666e-06, + "loss": 0.1905, + "step": 64692 + }, + { + "epoch": 2.49, + "learning_rate": 8.551551679917928e-06, + "loss": 0.1835, + "step": 64693 + }, + { + "epoch": 2.49, + "learning_rate": 8.550910489869198e-06, + "loss": 0.2975, + "step": 64694 + }, + { + "epoch": 2.49, + "learning_rate": 8.550269299820466e-06, + "loss": 0.1761, + "step": 64695 + }, + { + "epoch": 2.49, + "learning_rate": 8.549628109771738e-06, + "loss": 0.2571, + "step": 64696 + }, + { + "epoch": 2.49, + "learning_rate": 8.548986919723006e-06, + "loss": 0.3058, + "step": 64697 + }, + { + "epoch": 2.49, + "learning_rate": 8.548345729674276e-06, + "loss": 0.2829, + "step": 64698 + }, + { + "epoch": 2.49, + "learning_rate": 8.547704539625545e-06, + "loss": 0.1504, + "step": 64699 + }, + { + "epoch": 2.49, + "learning_rate": 8.547063349576815e-06, + "loss": 0.3691, + "step": 64700 + }, + { + "epoch": 2.49, + "learning_rate": 8.546422159528085e-06, + "loss": 0.2591, + "step": 64701 + }, + { + "epoch": 2.49, + "learning_rate": 8.545780969479355e-06, + "loss": 0.1691, + "step": 64702 + }, + { + "epoch": 2.49, + "learning_rate": 8.545139779430623e-06, + "loss": 0.2179, + "step": 64703 + }, + { + "epoch": 2.49, + "learning_rate": 8.544498589381893e-06, + "loss": 0.1394, + "step": 64704 + }, + { + "epoch": 2.49, + "learning_rate": 8.543857399333163e-06, + "loss": 0.2118, + "step": 64705 + }, + { + "epoch": 2.49, + "learning_rate": 8.543216209284433e-06, + "loss": 0.3523, + "step": 64706 + }, + { + "epoch": 2.49, + "learning_rate": 8.542575019235702e-06, + "loss": 0.1551, + "step": 64707 + }, + { + "epoch": 2.49, + "learning_rate": 8.541933829186972e-06, + "loss": 0.1389, + "step": 64708 + }, + { + "epoch": 2.49, + "learning_rate": 8.54129263913824e-06, + "loss": 0.2122, + "step": 64709 + }, + { + "epoch": 2.49, + "learning_rate": 8.540651449089512e-06, + "loss": 0.15, + "step": 64710 + }, + { + "epoch": 2.49, + "learning_rate": 8.54001025904078e-06, + "loss": 0.1618, + "step": 64711 + }, + { + "epoch": 2.49, + "learning_rate": 8.53936906899205e-06, + "loss": 0.202, + "step": 64712 + }, + { + "epoch": 2.49, + "learning_rate": 8.538727878943318e-06, + "loss": 0.2376, + "step": 64713 + }, + { + "epoch": 2.49, + "learning_rate": 8.53808668889459e-06, + "loss": 0.1341, + "step": 64714 + }, + { + "epoch": 2.49, + "learning_rate": 8.537445498845858e-06, + "loss": 0.1805, + "step": 64715 + }, + { + "epoch": 2.49, + "learning_rate": 8.536804308797128e-06, + "loss": 0.2696, + "step": 64716 + }, + { + "epoch": 2.49, + "learning_rate": 8.536163118748397e-06, + "loss": 0.1291, + "step": 64717 + }, + { + "epoch": 2.49, + "learning_rate": 8.535521928699668e-06, + "loss": 0.2179, + "step": 64718 + }, + { + "epoch": 2.49, + "learning_rate": 8.534880738650937e-06, + "loss": 0.2177, + "step": 64719 + }, + { + "epoch": 2.49, + "learning_rate": 8.534239548602207e-06, + "loss": 0.2182, + "step": 64720 + }, + { + "epoch": 2.49, + "learning_rate": 8.533598358553475e-06, + "loss": 0.2357, + "step": 64721 + }, + { + "epoch": 2.49, + "learning_rate": 8.532957168504745e-06, + "loss": 0.2562, + "step": 64722 + }, + { + "epoch": 2.49, + "learning_rate": 8.532315978456015e-06, + "loss": 0.1279, + "step": 64723 + }, + { + "epoch": 2.49, + "learning_rate": 8.531674788407285e-06, + "loss": 0.2381, + "step": 64724 + }, + { + "epoch": 2.49, + "learning_rate": 8.531033598358554e-06, + "loss": 0.1631, + "step": 64725 + }, + { + "epoch": 2.49, + "learning_rate": 8.530392408309824e-06, + "loss": 0.1293, + "step": 64726 + }, + { + "epoch": 2.49, + "learning_rate": 8.529751218261094e-06, + "loss": 0.2688, + "step": 64727 + }, + { + "epoch": 2.49, + "learning_rate": 8.529110028212364e-06, + "loss": 0.2352, + "step": 64728 + }, + { + "epoch": 2.49, + "learning_rate": 8.528468838163632e-06, + "loss": 0.1681, + "step": 64729 + }, + { + "epoch": 2.49, + "learning_rate": 8.527827648114902e-06, + "loss": 0.2246, + "step": 64730 + }, + { + "epoch": 2.49, + "learning_rate": 8.52718645806617e-06, + "loss": 0.2769, + "step": 64731 + }, + { + "epoch": 2.49, + "learning_rate": 8.526545268017442e-06, + "loss": 0.193, + "step": 64732 + }, + { + "epoch": 2.49, + "learning_rate": 8.52590407796871e-06, + "loss": 0.2097, + "step": 64733 + }, + { + "epoch": 2.49, + "learning_rate": 8.52526288791998e-06, + "loss": 0.1515, + "step": 64734 + }, + { + "epoch": 2.49, + "learning_rate": 8.524621697871249e-06, + "loss": 0.1894, + "step": 64735 + }, + { + "epoch": 2.49, + "learning_rate": 8.52398050782252e-06, + "loss": 0.2322, + "step": 64736 + }, + { + "epoch": 2.49, + "learning_rate": 8.523339317773789e-06, + "loss": 0.1415, + "step": 64737 + }, + { + "epoch": 2.49, + "learning_rate": 8.522698127725059e-06, + "loss": 0.174, + "step": 64738 + }, + { + "epoch": 2.49, + "learning_rate": 8.522056937676327e-06, + "loss": 0.1247, + "step": 64739 + }, + { + "epoch": 2.49, + "learning_rate": 8.521415747627597e-06, + "loss": 0.2063, + "step": 64740 + }, + { + "epoch": 2.49, + "learning_rate": 8.520774557578867e-06, + "loss": 0.3501, + "step": 64741 + }, + { + "epoch": 2.49, + "learning_rate": 8.520133367530137e-06, + "loss": 0.1182, + "step": 64742 + }, + { + "epoch": 2.49, + "learning_rate": 8.519492177481405e-06, + "loss": 0.3744, + "step": 64743 + }, + { + "epoch": 2.49, + "learning_rate": 8.518850987432675e-06, + "loss": 0.1352, + "step": 64744 + }, + { + "epoch": 2.49, + "learning_rate": 8.518209797383946e-06, + "loss": 0.2224, + "step": 64745 + }, + { + "epoch": 2.49, + "learning_rate": 8.517568607335216e-06, + "loss": 0.1535, + "step": 64746 + }, + { + "epoch": 2.49, + "learning_rate": 8.516927417286484e-06, + "loss": 0.1933, + "step": 64747 + }, + { + "epoch": 2.49, + "learning_rate": 8.516286227237754e-06, + "loss": 0.2757, + "step": 64748 + }, + { + "epoch": 2.49, + "learning_rate": 8.515645037189022e-06, + "loss": 0.2311, + "step": 64749 + }, + { + "epoch": 2.49, + "learning_rate": 8.515003847140294e-06, + "loss": 0.1313, + "step": 64750 + }, + { + "epoch": 2.49, + "learning_rate": 8.514362657091562e-06, + "loss": 0.2153, + "step": 64751 + }, + { + "epoch": 2.49, + "learning_rate": 8.513721467042832e-06, + "loss": 0.2003, + "step": 64752 + }, + { + "epoch": 2.49, + "learning_rate": 8.5130802769941e-06, + "loss": 0.149, + "step": 64753 + }, + { + "epoch": 2.49, + "learning_rate": 8.512439086945372e-06, + "loss": 0.2298, + "step": 64754 + }, + { + "epoch": 2.49, + "learning_rate": 8.51179789689664e-06, + "loss": 0.2521, + "step": 64755 + }, + { + "epoch": 2.49, + "learning_rate": 8.51115670684791e-06, + "loss": 0.2382, + "step": 64756 + }, + { + "epoch": 2.49, + "learning_rate": 8.510515516799179e-06, + "loss": 0.2473, + "step": 64757 + }, + { + "epoch": 2.49, + "learning_rate": 8.50987432675045e-06, + "loss": 0.1081, + "step": 64758 + }, + { + "epoch": 2.49, + "learning_rate": 8.509233136701719e-06, + "loss": 0.1649, + "step": 64759 + }, + { + "epoch": 2.49, + "learning_rate": 8.508591946652989e-06, + "loss": 0.203, + "step": 64760 + }, + { + "epoch": 2.49, + "learning_rate": 8.507950756604257e-06, + "loss": 0.3247, + "step": 64761 + }, + { + "epoch": 2.49, + "learning_rate": 8.507309566555527e-06, + "loss": 0.1912, + "step": 64762 + }, + { + "epoch": 2.49, + "learning_rate": 8.506668376506797e-06, + "loss": 0.1414, + "step": 64763 + }, + { + "epoch": 2.49, + "learning_rate": 8.506027186458067e-06, + "loss": 0.2837, + "step": 64764 + }, + { + "epoch": 2.49, + "learning_rate": 8.505385996409336e-06, + "loss": 0.2091, + "step": 64765 + }, + { + "epoch": 2.49, + "learning_rate": 8.504744806360606e-06, + "loss": 0.218, + "step": 64766 + }, + { + "epoch": 2.49, + "learning_rate": 8.504103616311876e-06, + "loss": 0.3352, + "step": 64767 + }, + { + "epoch": 2.49, + "learning_rate": 8.503462426263146e-06, + "loss": 0.2919, + "step": 64768 + }, + { + "epoch": 2.49, + "learning_rate": 8.502821236214414e-06, + "loss": 0.2779, + "step": 64769 + }, + { + "epoch": 2.49, + "learning_rate": 8.502180046165684e-06, + "loss": 0.2428, + "step": 64770 + }, + { + "epoch": 2.49, + "learning_rate": 8.501538856116953e-06, + "loss": 0.3384, + "step": 64771 + }, + { + "epoch": 2.49, + "learning_rate": 8.500897666068224e-06, + "loss": 0.1396, + "step": 64772 + }, + { + "epoch": 2.49, + "learning_rate": 8.500256476019493e-06, + "loss": 0.2225, + "step": 64773 + }, + { + "epoch": 2.49, + "learning_rate": 8.499615285970763e-06, + "loss": 0.2037, + "step": 64774 + }, + { + "epoch": 2.49, + "learning_rate": 8.498974095922031e-06, + "loss": 0.1594, + "step": 64775 + }, + { + "epoch": 2.49, + "learning_rate": 8.498332905873303e-06, + "loss": 0.2004, + "step": 64776 + }, + { + "epoch": 2.49, + "learning_rate": 8.497691715824571e-06, + "loss": 0.1784, + "step": 64777 + }, + { + "epoch": 2.49, + "learning_rate": 8.497050525775841e-06, + "loss": 0.3149, + "step": 64778 + }, + { + "epoch": 2.49, + "learning_rate": 8.49640933572711e-06, + "loss": 0.2449, + "step": 64779 + }, + { + "epoch": 2.49, + "learning_rate": 8.49576814567838e-06, + "loss": 0.1457, + "step": 64780 + }, + { + "epoch": 2.49, + "learning_rate": 8.49512695562965e-06, + "loss": 0.2515, + "step": 64781 + }, + { + "epoch": 2.49, + "learning_rate": 8.49448576558092e-06, + "loss": 0.2407, + "step": 64782 + }, + { + "epoch": 2.49, + "learning_rate": 8.493844575532188e-06, + "loss": 0.2051, + "step": 64783 + }, + { + "epoch": 2.49, + "learning_rate": 8.493203385483458e-06, + "loss": 0.1382, + "step": 64784 + }, + { + "epoch": 2.49, + "learning_rate": 8.492562195434728e-06, + "loss": 0.1561, + "step": 64785 + }, + { + "epoch": 2.49, + "learning_rate": 8.491921005385998e-06, + "loss": 0.2378, + "step": 64786 + }, + { + "epoch": 2.49, + "learning_rate": 8.491279815337266e-06, + "loss": 0.2551, + "step": 64787 + }, + { + "epoch": 2.49, + "learning_rate": 8.490638625288536e-06, + "loss": 0.1888, + "step": 64788 + }, + { + "epoch": 2.49, + "learning_rate": 8.489997435239804e-06, + "loss": 0.2475, + "step": 64789 + }, + { + "epoch": 2.49, + "learning_rate": 8.489356245191076e-06, + "loss": 0.168, + "step": 64790 + }, + { + "epoch": 2.49, + "learning_rate": 8.488715055142345e-06, + "loss": 0.177, + "step": 64791 + }, + { + "epoch": 2.49, + "learning_rate": 8.488073865093615e-06, + "loss": 0.4421, + "step": 64792 + }, + { + "epoch": 2.49, + "learning_rate": 8.487432675044883e-06, + "loss": 0.1632, + "step": 64793 + }, + { + "epoch": 2.49, + "learning_rate": 8.486791484996155e-06, + "loss": 0.2174, + "step": 64794 + }, + { + "epoch": 2.49, + "learning_rate": 8.486150294947423e-06, + "loss": 0.3195, + "step": 64795 + }, + { + "epoch": 2.49, + "learning_rate": 8.485509104898693e-06, + "loss": 0.1847, + "step": 64796 + }, + { + "epoch": 2.49, + "learning_rate": 8.484867914849961e-06, + "loss": 0.2097, + "step": 64797 + }, + { + "epoch": 2.49, + "learning_rate": 8.484226724801231e-06, + "loss": 0.2906, + "step": 64798 + }, + { + "epoch": 2.49, + "learning_rate": 8.483585534752501e-06, + "loss": 0.2579, + "step": 64799 + }, + { + "epoch": 2.49, + "learning_rate": 8.482944344703771e-06, + "loss": 0.2807, + "step": 64800 + }, + { + "epoch": 2.49, + "learning_rate": 8.48230315465504e-06, + "loss": 0.2432, + "step": 64801 + }, + { + "epoch": 2.49, + "learning_rate": 8.48166196460631e-06, + "loss": 0.1069, + "step": 64802 + }, + { + "epoch": 2.49, + "learning_rate": 8.48102077455758e-06, + "loss": 0.1608, + "step": 64803 + }, + { + "epoch": 2.49, + "learning_rate": 8.48037958450885e-06, + "loss": 0.2554, + "step": 64804 + }, + { + "epoch": 2.49, + "learning_rate": 8.479738394460118e-06, + "loss": 0.188, + "step": 64805 + }, + { + "epoch": 2.49, + "learning_rate": 8.479097204411388e-06, + "loss": 0.2704, + "step": 64806 + }, + { + "epoch": 2.49, + "learning_rate": 8.478456014362658e-06, + "loss": 0.2014, + "step": 64807 + }, + { + "epoch": 2.49, + "learning_rate": 8.477814824313928e-06, + "loss": 0.148, + "step": 64808 + }, + { + "epoch": 2.49, + "learning_rate": 8.477173634265196e-06, + "loss": 0.2481, + "step": 64809 + }, + { + "epoch": 2.49, + "learning_rate": 8.476532444216466e-06, + "loss": 0.1614, + "step": 64810 + }, + { + "epoch": 2.49, + "learning_rate": 8.475891254167735e-06, + "loss": 0.2918, + "step": 64811 + }, + { + "epoch": 2.49, + "learning_rate": 8.475250064119007e-06, + "loss": 0.1747, + "step": 64812 + }, + { + "epoch": 2.49, + "learning_rate": 8.474608874070275e-06, + "loss": 0.1115, + "step": 64813 + }, + { + "epoch": 2.49, + "learning_rate": 8.473967684021545e-06, + "loss": 0.2214, + "step": 64814 + }, + { + "epoch": 2.49, + "learning_rate": 8.473326493972813e-06, + "loss": 0.2223, + "step": 64815 + }, + { + "epoch": 2.49, + "learning_rate": 8.472685303924085e-06, + "loss": 0.1277, + "step": 64816 + }, + { + "epoch": 2.49, + "learning_rate": 8.472044113875353e-06, + "loss": 0.2824, + "step": 64817 + }, + { + "epoch": 2.49, + "learning_rate": 8.471402923826623e-06, + "loss": 0.1906, + "step": 64818 + }, + { + "epoch": 2.49, + "learning_rate": 8.470761733777892e-06, + "loss": 0.137, + "step": 64819 + }, + { + "epoch": 2.49, + "learning_rate": 8.470120543729162e-06, + "loss": 0.2059, + "step": 64820 + }, + { + "epoch": 2.49, + "learning_rate": 8.469479353680432e-06, + "loss": 0.1713, + "step": 64821 + }, + { + "epoch": 2.49, + "learning_rate": 8.468838163631702e-06, + "loss": 0.198, + "step": 64822 + }, + { + "epoch": 2.49, + "learning_rate": 8.46819697358297e-06, + "loss": 0.1596, + "step": 64823 + }, + { + "epoch": 2.49, + "learning_rate": 8.46755578353424e-06, + "loss": 0.2813, + "step": 64824 + }, + { + "epoch": 2.49, + "learning_rate": 8.46691459348551e-06, + "loss": 0.261, + "step": 64825 + }, + { + "epoch": 2.49, + "learning_rate": 8.46627340343678e-06, + "loss": 0.3699, + "step": 64826 + }, + { + "epoch": 2.49, + "learning_rate": 8.465632213388048e-06, + "loss": 0.2922, + "step": 64827 + }, + { + "epoch": 2.49, + "learning_rate": 8.464991023339318e-06, + "loss": 0.1434, + "step": 64828 + }, + { + "epoch": 2.49, + "learning_rate": 8.464349833290587e-06, + "loss": 0.3546, + "step": 64829 + }, + { + "epoch": 2.49, + "learning_rate": 8.463708643241858e-06, + "loss": 0.1277, + "step": 64830 + }, + { + "epoch": 2.49, + "learning_rate": 8.463067453193127e-06, + "loss": 0.2248, + "step": 64831 + }, + { + "epoch": 2.49, + "learning_rate": 8.462426263144397e-06, + "loss": 0.1698, + "step": 64832 + }, + { + "epoch": 2.49, + "learning_rate": 8.461785073095665e-06, + "loss": 0.1849, + "step": 64833 + }, + { + "epoch": 2.49, + "learning_rate": 8.461143883046937e-06, + "loss": 0.2848, + "step": 64834 + }, + { + "epoch": 2.49, + "learning_rate": 8.460502692998205e-06, + "loss": 0.2741, + "step": 64835 + }, + { + "epoch": 2.49, + "learning_rate": 8.459861502949475e-06, + "loss": 0.2667, + "step": 64836 + }, + { + "epoch": 2.49, + "learning_rate": 8.459220312900744e-06, + "loss": 0.2501, + "step": 64837 + }, + { + "epoch": 2.49, + "learning_rate": 8.458579122852014e-06, + "loss": 0.3246, + "step": 64838 + }, + { + "epoch": 2.49, + "learning_rate": 8.457937932803284e-06, + "loss": 0.1725, + "step": 64839 + }, + { + "epoch": 2.49, + "learning_rate": 8.457296742754554e-06, + "loss": 0.1855, + "step": 64840 + }, + { + "epoch": 2.49, + "learning_rate": 8.456655552705822e-06, + "loss": 0.2065, + "step": 64841 + }, + { + "epoch": 2.49, + "learning_rate": 8.456014362657092e-06, + "loss": 0.2984, + "step": 64842 + }, + { + "epoch": 2.49, + "learning_rate": 8.455373172608362e-06, + "loss": 0.1766, + "step": 64843 + }, + { + "epoch": 2.49, + "learning_rate": 8.454731982559632e-06, + "loss": 0.178, + "step": 64844 + }, + { + "epoch": 2.49, + "learning_rate": 8.4540907925109e-06, + "loss": 0.2782, + "step": 64845 + }, + { + "epoch": 2.49, + "learning_rate": 8.45344960246217e-06, + "loss": 0.2003, + "step": 64846 + }, + { + "epoch": 2.49, + "learning_rate": 8.452808412413439e-06, + "loss": 0.2618, + "step": 64847 + }, + { + "epoch": 2.49, + "learning_rate": 8.45216722236471e-06, + "loss": 0.2173, + "step": 64848 + }, + { + "epoch": 2.49, + "learning_rate": 8.451526032315979e-06, + "loss": 0.3033, + "step": 64849 + }, + { + "epoch": 2.49, + "learning_rate": 8.450884842267249e-06, + "loss": 0.2071, + "step": 64850 + }, + { + "epoch": 2.49, + "learning_rate": 8.450243652218517e-06, + "loss": 0.2307, + "step": 64851 + }, + { + "epoch": 2.49, + "learning_rate": 8.449602462169789e-06, + "loss": 0.1186, + "step": 64852 + }, + { + "epoch": 2.49, + "learning_rate": 8.448961272121057e-06, + "loss": 0.2825, + "step": 64853 + }, + { + "epoch": 2.49, + "learning_rate": 8.448320082072327e-06, + "loss": 0.1282, + "step": 64854 + }, + { + "epoch": 2.49, + "learning_rate": 8.447678892023595e-06, + "loss": 0.187, + "step": 64855 + }, + { + "epoch": 2.49, + "learning_rate": 8.447037701974867e-06, + "loss": 0.174, + "step": 64856 + }, + { + "epoch": 2.49, + "learning_rate": 8.446396511926136e-06, + "loss": 0.2524, + "step": 64857 + }, + { + "epoch": 2.49, + "learning_rate": 8.445755321877406e-06, + "loss": 0.1642, + "step": 64858 + }, + { + "epoch": 2.49, + "learning_rate": 8.445114131828674e-06, + "loss": 0.2575, + "step": 64859 + }, + { + "epoch": 2.49, + "learning_rate": 8.444472941779944e-06, + "loss": 0.1902, + "step": 64860 + }, + { + "epoch": 2.49, + "learning_rate": 8.443831751731214e-06, + "loss": 0.2359, + "step": 64861 + }, + { + "epoch": 2.49, + "learning_rate": 8.443190561682484e-06, + "loss": 0.2622, + "step": 64862 + }, + { + "epoch": 2.49, + "learning_rate": 8.442549371633752e-06, + "loss": 0.1243, + "step": 64863 + }, + { + "epoch": 2.49, + "learning_rate": 8.441908181585022e-06, + "loss": 0.2264, + "step": 64864 + }, + { + "epoch": 2.49, + "learning_rate": 8.441266991536292e-06, + "loss": 0.1847, + "step": 64865 + }, + { + "epoch": 2.49, + "learning_rate": 8.440625801487562e-06, + "loss": 0.2706, + "step": 64866 + }, + { + "epoch": 2.49, + "learning_rate": 8.43998461143883e-06, + "loss": 0.3181, + "step": 64867 + }, + { + "epoch": 2.49, + "learning_rate": 8.4393434213901e-06, + "loss": 0.2193, + "step": 64868 + }, + { + "epoch": 2.49, + "learning_rate": 8.438702231341369e-06, + "loss": 0.2654, + "step": 64869 + }, + { + "epoch": 2.49, + "learning_rate": 8.43806104129264e-06, + "loss": 0.1994, + "step": 64870 + }, + { + "epoch": 2.49, + "learning_rate": 8.437419851243909e-06, + "loss": 0.2799, + "step": 64871 + }, + { + "epoch": 2.49, + "learning_rate": 8.436778661195179e-06, + "loss": 0.2443, + "step": 64872 + }, + { + "epoch": 2.49, + "learning_rate": 8.436137471146447e-06, + "loss": 0.2532, + "step": 64873 + }, + { + "epoch": 2.49, + "learning_rate": 8.435496281097719e-06, + "loss": 0.2688, + "step": 64874 + }, + { + "epoch": 2.49, + "learning_rate": 8.434855091048987e-06, + "loss": 0.1557, + "step": 64875 + }, + { + "epoch": 2.49, + "learning_rate": 8.434213901000257e-06, + "loss": 0.1516, + "step": 64876 + }, + { + "epoch": 2.49, + "learning_rate": 8.433572710951526e-06, + "loss": 0.1498, + "step": 64877 + }, + { + "epoch": 2.49, + "learning_rate": 8.432931520902796e-06, + "loss": 0.2211, + "step": 64878 + }, + { + "epoch": 2.49, + "learning_rate": 8.432290330854066e-06, + "loss": 0.1107, + "step": 64879 + }, + { + "epoch": 2.49, + "learning_rate": 8.431649140805336e-06, + "loss": 0.177, + "step": 64880 + }, + { + "epoch": 2.49, + "learning_rate": 8.431007950756604e-06, + "loss": 0.1471, + "step": 64881 + }, + { + "epoch": 2.49, + "learning_rate": 8.430366760707874e-06, + "loss": 0.2567, + "step": 64882 + }, + { + "epoch": 2.49, + "learning_rate": 8.429725570659144e-06, + "loss": 0.2029, + "step": 64883 + }, + { + "epoch": 2.49, + "learning_rate": 8.429084380610414e-06, + "loss": 0.2589, + "step": 64884 + }, + { + "epoch": 2.49, + "learning_rate": 8.428443190561683e-06, + "loss": 0.1427, + "step": 64885 + }, + { + "epoch": 2.49, + "learning_rate": 8.427802000512953e-06, + "loss": 0.1259, + "step": 64886 + }, + { + "epoch": 2.49, + "learning_rate": 8.427160810464221e-06, + "loss": 0.1121, + "step": 64887 + }, + { + "epoch": 2.49, + "learning_rate": 8.426519620415493e-06, + "loss": 0.113, + "step": 64888 + }, + { + "epoch": 2.49, + "learning_rate": 8.425878430366761e-06, + "loss": 0.2065, + "step": 64889 + }, + { + "epoch": 2.49, + "learning_rate": 8.425237240318031e-06, + "loss": 0.1991, + "step": 64890 + }, + { + "epoch": 2.49, + "learning_rate": 8.4245960502693e-06, + "loss": 0.163, + "step": 64891 + }, + { + "epoch": 2.49, + "learning_rate": 8.423954860220571e-06, + "loss": 0.258, + "step": 64892 + }, + { + "epoch": 2.49, + "learning_rate": 8.42331367017184e-06, + "loss": 0.1309, + "step": 64893 + }, + { + "epoch": 2.49, + "learning_rate": 8.42267248012311e-06, + "loss": 0.2082, + "step": 64894 + }, + { + "epoch": 2.5, + "learning_rate": 8.422031290074378e-06, + "loss": 0.255, + "step": 64895 + }, + { + "epoch": 2.5, + "learning_rate": 8.421390100025648e-06, + "loss": 0.2339, + "step": 64896 + }, + { + "epoch": 2.5, + "learning_rate": 8.420748909976918e-06, + "loss": 0.3876, + "step": 64897 + }, + { + "epoch": 2.5, + "learning_rate": 8.420107719928188e-06, + "loss": 0.1519, + "step": 64898 + }, + { + "epoch": 2.5, + "learning_rate": 8.419466529879456e-06, + "loss": 0.2775, + "step": 64899 + }, + { + "epoch": 2.5, + "learning_rate": 8.418825339830726e-06, + "loss": 0.2861, + "step": 64900 + }, + { + "epoch": 2.5, + "learning_rate": 8.418184149781996e-06, + "loss": 0.2325, + "step": 64901 + }, + { + "epoch": 2.5, + "learning_rate": 8.417542959733266e-06, + "loss": 0.3052, + "step": 64902 + }, + { + "epoch": 2.5, + "learning_rate": 8.416901769684535e-06, + "loss": 0.2318, + "step": 64903 + }, + { + "epoch": 2.5, + "learning_rate": 8.416260579635805e-06, + "loss": 0.2338, + "step": 64904 + }, + { + "epoch": 2.5, + "learning_rate": 8.415619389587075e-06, + "loss": 0.1127, + "step": 64905 + }, + { + "epoch": 2.5, + "learning_rate": 8.414978199538345e-06, + "loss": 0.3109, + "step": 64906 + }, + { + "epoch": 2.5, + "learning_rate": 8.414337009489613e-06, + "loss": 0.2903, + "step": 64907 + }, + { + "epoch": 2.5, + "learning_rate": 8.413695819440883e-06, + "loss": 0.1539, + "step": 64908 + }, + { + "epoch": 2.5, + "learning_rate": 8.413054629392151e-06, + "loss": 0.26, + "step": 64909 + }, + { + "epoch": 2.5, + "learning_rate": 8.412413439343423e-06, + "loss": 0.1914, + "step": 64910 + }, + { + "epoch": 2.5, + "learning_rate": 8.411772249294691e-06, + "loss": 0.2733, + "step": 64911 + }, + { + "epoch": 2.5, + "learning_rate": 8.411131059245961e-06, + "loss": 0.1858, + "step": 64912 + }, + { + "epoch": 2.5, + "learning_rate": 8.41048986919723e-06, + "loss": 0.3035, + "step": 64913 + }, + { + "epoch": 2.5, + "learning_rate": 8.409848679148501e-06, + "loss": 0.3005, + "step": 64914 + }, + { + "epoch": 2.5, + "learning_rate": 8.40920748909977e-06, + "loss": 0.4899, + "step": 64915 + }, + { + "epoch": 2.5, + "learning_rate": 8.40856629905104e-06, + "loss": 0.1771, + "step": 64916 + }, + { + "epoch": 2.5, + "learning_rate": 8.407925109002308e-06, + "loss": 0.1091, + "step": 64917 + }, + { + "epoch": 2.5, + "learning_rate": 8.407283918953578e-06, + "loss": 0.1918, + "step": 64918 + }, + { + "epoch": 2.5, + "learning_rate": 8.406642728904848e-06, + "loss": 0.2532, + "step": 64919 + }, + { + "epoch": 2.5, + "learning_rate": 8.406001538856118e-06, + "loss": 0.1079, + "step": 64920 + }, + { + "epoch": 2.5, + "learning_rate": 8.405360348807386e-06, + "loss": 0.1545, + "step": 64921 + }, + { + "epoch": 2.5, + "learning_rate": 8.404719158758657e-06, + "loss": 0.4358, + "step": 64922 + }, + { + "epoch": 2.5, + "learning_rate": 8.404077968709927e-06, + "loss": 0.3114, + "step": 64923 + }, + { + "epoch": 2.5, + "learning_rate": 8.403436778661197e-06, + "loss": 0.1539, + "step": 64924 + }, + { + "epoch": 2.5, + "learning_rate": 8.402795588612465e-06, + "loss": 0.2506, + "step": 64925 + }, + { + "epoch": 2.5, + "learning_rate": 8.402154398563735e-06, + "loss": 0.1288, + "step": 64926 + }, + { + "epoch": 2.5, + "learning_rate": 8.401513208515003e-06, + "loss": 0.1846, + "step": 64927 + }, + { + "epoch": 2.5, + "learning_rate": 8.400872018466275e-06, + "loss": 0.2351, + "step": 64928 + }, + { + "epoch": 2.5, + "learning_rate": 8.400230828417543e-06, + "loss": 0.3575, + "step": 64929 + }, + { + "epoch": 2.5, + "learning_rate": 8.399589638368813e-06, + "loss": 0.3729, + "step": 64930 + }, + { + "epoch": 2.5, + "learning_rate": 8.398948448320082e-06, + "loss": 0.3077, + "step": 64931 + }, + { + "epoch": 2.5, + "learning_rate": 8.398307258271353e-06, + "loss": 0.3058, + "step": 64932 + }, + { + "epoch": 2.5, + "learning_rate": 8.397666068222622e-06, + "loss": 0.2946, + "step": 64933 + }, + { + "epoch": 2.5, + "learning_rate": 8.397024878173892e-06, + "loss": 0.1916, + "step": 64934 + }, + { + "epoch": 2.5, + "learning_rate": 8.39638368812516e-06, + "loss": 0.2557, + "step": 64935 + }, + { + "epoch": 2.5, + "learning_rate": 8.39574249807643e-06, + "loss": 0.2218, + "step": 64936 + }, + { + "epoch": 2.5, + "learning_rate": 8.3951013080277e-06, + "loss": 0.1928, + "step": 64937 + }, + { + "epoch": 2.5, + "learning_rate": 8.39446011797897e-06, + "loss": 0.2765, + "step": 64938 + }, + { + "epoch": 2.5, + "learning_rate": 8.393818927930238e-06, + "loss": 0.1117, + "step": 64939 + }, + { + "epoch": 2.5, + "learning_rate": 8.393177737881508e-06, + "loss": 0.1591, + "step": 64940 + }, + { + "epoch": 2.5, + "learning_rate": 8.392536547832778e-06, + "loss": 0.2888, + "step": 64941 + }, + { + "epoch": 2.5, + "learning_rate": 8.391895357784048e-06, + "loss": 0.3386, + "step": 64942 + }, + { + "epoch": 2.5, + "learning_rate": 8.391254167735317e-06, + "loss": 0.241, + "step": 64943 + }, + { + "epoch": 2.5, + "learning_rate": 8.390612977686587e-06, + "loss": 0.2195, + "step": 64944 + }, + { + "epoch": 2.5, + "learning_rate": 8.389971787637855e-06, + "loss": 0.2916, + "step": 64945 + }, + { + "epoch": 2.5, + "learning_rate": 8.389330597589127e-06, + "loss": 0.1185, + "step": 64946 + }, + { + "epoch": 2.5, + "learning_rate": 8.388689407540395e-06, + "loss": 0.2782, + "step": 64947 + }, + { + "epoch": 2.5, + "learning_rate": 8.388048217491665e-06, + "loss": 0.2748, + "step": 64948 + }, + { + "epoch": 2.5, + "learning_rate": 8.387407027442934e-06, + "loss": 0.13, + "step": 64949 + }, + { + "epoch": 2.5, + "learning_rate": 8.386765837394205e-06, + "loss": 0.179, + "step": 64950 + }, + { + "epoch": 2.5, + "learning_rate": 8.386124647345474e-06, + "loss": 0.2682, + "step": 64951 + }, + { + "epoch": 2.5, + "learning_rate": 8.385483457296744e-06, + "loss": 0.1465, + "step": 64952 + }, + { + "epoch": 2.5, + "learning_rate": 8.384842267248012e-06, + "loss": 0.2198, + "step": 64953 + }, + { + "epoch": 2.5, + "learning_rate": 8.384201077199284e-06, + "loss": 0.2266, + "step": 64954 + }, + { + "epoch": 2.5, + "learning_rate": 8.383559887150552e-06, + "loss": 0.1669, + "step": 64955 + }, + { + "epoch": 2.5, + "learning_rate": 8.382918697101822e-06, + "loss": 0.162, + "step": 64956 + }, + { + "epoch": 2.5, + "learning_rate": 8.38227750705309e-06, + "loss": 0.1776, + "step": 64957 + }, + { + "epoch": 2.5, + "learning_rate": 8.38163631700436e-06, + "loss": 0.2926, + "step": 64958 + }, + { + "epoch": 2.5, + "learning_rate": 8.38099512695563e-06, + "loss": 0.1294, + "step": 64959 + }, + { + "epoch": 2.5, + "learning_rate": 8.3803539369069e-06, + "loss": 0.2366, + "step": 64960 + }, + { + "epoch": 2.5, + "learning_rate": 8.379712746858169e-06, + "loss": 0.2281, + "step": 64961 + }, + { + "epoch": 2.5, + "learning_rate": 8.379071556809439e-06, + "loss": 0.0871, + "step": 64962 + }, + { + "epoch": 2.5, + "learning_rate": 8.378430366760709e-06, + "loss": 0.2089, + "step": 64963 + }, + { + "epoch": 2.5, + "learning_rate": 8.377789176711979e-06, + "loss": 0.1538, + "step": 64964 + }, + { + "epoch": 2.5, + "learning_rate": 8.377147986663247e-06, + "loss": 0.2358, + "step": 64965 + }, + { + "epoch": 2.5, + "learning_rate": 8.376506796614517e-06, + "loss": 0.1379, + "step": 64966 + }, + { + "epoch": 2.5, + "learning_rate": 8.375865606565785e-06, + "loss": 0.1985, + "step": 64967 + }, + { + "epoch": 2.5, + "learning_rate": 8.375224416517057e-06, + "loss": 0.133, + "step": 64968 + }, + { + "epoch": 2.5, + "learning_rate": 8.374583226468326e-06, + "loss": 0.2307, + "step": 64969 + }, + { + "epoch": 2.5, + "learning_rate": 8.373942036419596e-06, + "loss": 0.2103, + "step": 64970 + }, + { + "epoch": 2.5, + "learning_rate": 8.373300846370864e-06, + "loss": 0.2477, + "step": 64971 + }, + { + "epoch": 2.5, + "learning_rate": 8.372659656322136e-06, + "loss": 0.2281, + "step": 64972 + }, + { + "epoch": 2.5, + "learning_rate": 8.372018466273404e-06, + "loss": 0.3216, + "step": 64973 + }, + { + "epoch": 2.5, + "learning_rate": 8.371377276224674e-06, + "loss": 0.3685, + "step": 64974 + }, + { + "epoch": 2.5, + "learning_rate": 8.370736086175942e-06, + "loss": 0.1357, + "step": 64975 + }, + { + "epoch": 2.5, + "learning_rate": 8.370094896127212e-06, + "loss": 0.2067, + "step": 64976 + }, + { + "epoch": 2.5, + "learning_rate": 8.369453706078482e-06, + "loss": 0.1355, + "step": 64977 + }, + { + "epoch": 2.5, + "learning_rate": 8.368812516029752e-06, + "loss": 0.2139, + "step": 64978 + }, + { + "epoch": 2.5, + "learning_rate": 8.36817132598102e-06, + "loss": 0.1373, + "step": 64979 + }, + { + "epoch": 2.5, + "learning_rate": 8.36753013593229e-06, + "loss": 0.2421, + "step": 64980 + }, + { + "epoch": 2.5, + "learning_rate": 8.36688894588356e-06, + "loss": 0.3412, + "step": 64981 + }, + { + "epoch": 2.5, + "learning_rate": 8.36624775583483e-06, + "loss": 0.1004, + "step": 64982 + }, + { + "epoch": 2.5, + "learning_rate": 8.365606565786099e-06, + "loss": 0.2926, + "step": 64983 + }, + { + "epoch": 2.5, + "learning_rate": 8.364965375737369e-06, + "loss": 0.1564, + "step": 64984 + }, + { + "epoch": 2.5, + "learning_rate": 8.364324185688637e-06, + "loss": 0.1056, + "step": 64985 + }, + { + "epoch": 2.5, + "learning_rate": 8.363682995639909e-06, + "loss": 0.1358, + "step": 64986 + }, + { + "epoch": 2.5, + "learning_rate": 8.363041805591177e-06, + "loss": 0.2381, + "step": 64987 + }, + { + "epoch": 2.5, + "learning_rate": 8.362400615542448e-06, + "loss": 0.1459, + "step": 64988 + }, + { + "epoch": 2.5, + "learning_rate": 8.361759425493716e-06, + "loss": 0.2089, + "step": 64989 + }, + { + "epoch": 2.5, + "learning_rate": 8.361118235444988e-06, + "loss": 0.21, + "step": 64990 + }, + { + "epoch": 2.5, + "learning_rate": 8.360477045396256e-06, + "loss": 0.1709, + "step": 64991 + }, + { + "epoch": 2.5, + "learning_rate": 8.359835855347526e-06, + "loss": 0.1268, + "step": 64992 + }, + { + "epoch": 2.5, + "learning_rate": 8.359194665298794e-06, + "loss": 0.2735, + "step": 64993 + }, + { + "epoch": 2.5, + "learning_rate": 8.358553475250066e-06, + "loss": 0.1368, + "step": 64994 + }, + { + "epoch": 2.5, + "learning_rate": 8.357912285201334e-06, + "loss": 0.2097, + "step": 64995 + }, + { + "epoch": 2.5, + "learning_rate": 8.357271095152604e-06, + "loss": 0.1561, + "step": 64996 + }, + { + "epoch": 2.5, + "learning_rate": 8.356629905103873e-06, + "loss": 0.0986, + "step": 64997 + }, + { + "epoch": 2.5, + "learning_rate": 8.355988715055143e-06, + "loss": 0.2756, + "step": 64998 + }, + { + "epoch": 2.5, + "learning_rate": 8.355347525006413e-06, + "loss": 0.358, + "step": 64999 + }, + { + "epoch": 2.5, + "learning_rate": 8.354706334957683e-06, + "loss": 0.1115, + "step": 65000 + }, + { + "epoch": 2.5, + "learning_rate": 8.354065144908951e-06, + "loss": 0.1873, + "step": 65001 + }, + { + "epoch": 2.5, + "learning_rate": 8.353423954860221e-06, + "loss": 0.1856, + "step": 65002 + }, + { + "epoch": 2.5, + "learning_rate": 8.352782764811491e-06, + "loss": 0.1087, + "step": 65003 + }, + { + "epoch": 2.5, + "learning_rate": 8.352141574762761e-06, + "loss": 0.2217, + "step": 65004 + }, + { + "epoch": 2.5, + "learning_rate": 8.35150038471403e-06, + "loss": 0.2249, + "step": 65005 + }, + { + "epoch": 2.5, + "learning_rate": 8.3508591946653e-06, + "loss": 0.2734, + "step": 65006 + }, + { + "epoch": 2.5, + "learning_rate": 8.350218004616568e-06, + "loss": 0.1753, + "step": 65007 + }, + { + "epoch": 2.5, + "learning_rate": 8.34957681456784e-06, + "loss": 0.1338, + "step": 65008 + }, + { + "epoch": 2.5, + "learning_rate": 8.348935624519108e-06, + "loss": 0.3013, + "step": 65009 + }, + { + "epoch": 2.5, + "learning_rate": 8.348294434470378e-06, + "loss": 0.1911, + "step": 65010 + }, + { + "epoch": 2.5, + "learning_rate": 8.347653244421646e-06, + "loss": 0.1315, + "step": 65011 + }, + { + "epoch": 2.5, + "learning_rate": 8.347012054372918e-06, + "loss": 0.15, + "step": 65012 + }, + { + "epoch": 2.5, + "learning_rate": 8.346370864324186e-06, + "loss": 0.254, + "step": 65013 + }, + { + "epoch": 2.5, + "learning_rate": 8.345729674275456e-06, + "loss": 0.2606, + "step": 65014 + }, + { + "epoch": 2.5, + "learning_rate": 8.345088484226725e-06, + "loss": 0.3082, + "step": 65015 + }, + { + "epoch": 2.5, + "learning_rate": 8.344447294177995e-06, + "loss": 0.2765, + "step": 65016 + }, + { + "epoch": 2.5, + "learning_rate": 8.343806104129265e-06, + "loss": 0.1906, + "step": 65017 + }, + { + "epoch": 2.5, + "learning_rate": 8.343164914080535e-06, + "loss": 0.215, + "step": 65018 + }, + { + "epoch": 2.5, + "learning_rate": 8.342523724031803e-06, + "loss": 0.3546, + "step": 65019 + }, + { + "epoch": 2.5, + "learning_rate": 8.341882533983073e-06, + "loss": 0.2423, + "step": 65020 + }, + { + "epoch": 2.5, + "learning_rate": 8.341241343934343e-06, + "loss": 0.1626, + "step": 65021 + }, + { + "epoch": 2.5, + "learning_rate": 8.340600153885613e-06, + "loss": 0.2038, + "step": 65022 + }, + { + "epoch": 2.5, + "learning_rate": 8.339958963836881e-06, + "loss": 0.2235, + "step": 65023 + }, + { + "epoch": 2.5, + "learning_rate": 8.339317773788151e-06, + "loss": 0.1139, + "step": 65024 + }, + { + "epoch": 2.5, + "learning_rate": 8.33867658373942e-06, + "loss": 0.0936, + "step": 65025 + }, + { + "epoch": 2.5, + "learning_rate": 8.338035393690691e-06, + "loss": 0.4226, + "step": 65026 + }, + { + "epoch": 2.5, + "learning_rate": 8.33739420364196e-06, + "loss": 0.1538, + "step": 65027 + }, + { + "epoch": 2.5, + "learning_rate": 8.33675301359323e-06, + "loss": 0.0518, + "step": 65028 + }, + { + "epoch": 2.5, + "learning_rate": 8.336111823544498e-06, + "loss": 0.2605, + "step": 65029 + }, + { + "epoch": 2.5, + "learning_rate": 8.33547063349577e-06, + "loss": 0.2338, + "step": 65030 + }, + { + "epoch": 2.5, + "learning_rate": 8.334829443447038e-06, + "loss": 0.2012, + "step": 65031 + }, + { + "epoch": 2.5, + "learning_rate": 8.334188253398308e-06, + "loss": 0.346, + "step": 65032 + }, + { + "epoch": 2.5, + "learning_rate": 8.333547063349576e-06, + "loss": 0.2563, + "step": 65033 + }, + { + "epoch": 2.5, + "learning_rate": 8.332905873300847e-06, + "loss": 0.4135, + "step": 65034 + }, + { + "epoch": 2.5, + "learning_rate": 8.332264683252117e-06, + "loss": 0.2235, + "step": 65035 + }, + { + "epoch": 2.5, + "learning_rate": 8.331623493203387e-06, + "loss": 0.3426, + "step": 65036 + }, + { + "epoch": 2.5, + "learning_rate": 8.330982303154655e-06, + "loss": 0.1796, + "step": 65037 + }, + { + "epoch": 2.5, + "learning_rate": 8.330341113105925e-06, + "loss": 0.2815, + "step": 65038 + }, + { + "epoch": 2.5, + "learning_rate": 8.329699923057195e-06, + "loss": 0.2034, + "step": 65039 + }, + { + "epoch": 2.5, + "learning_rate": 8.329058733008465e-06, + "loss": 0.2332, + "step": 65040 + }, + { + "epoch": 2.5, + "learning_rate": 8.328417542959733e-06, + "loss": 0.194, + "step": 65041 + }, + { + "epoch": 2.5, + "learning_rate": 8.327776352911003e-06, + "loss": 0.1365, + "step": 65042 + }, + { + "epoch": 2.5, + "learning_rate": 8.327135162862273e-06, + "loss": 0.2451, + "step": 65043 + }, + { + "epoch": 2.5, + "learning_rate": 8.326493972813543e-06, + "loss": 0.2594, + "step": 65044 + }, + { + "epoch": 2.5, + "learning_rate": 8.325852782764812e-06, + "loss": 0.1728, + "step": 65045 + }, + { + "epoch": 2.5, + "learning_rate": 8.325211592716082e-06, + "loss": 0.1504, + "step": 65046 + }, + { + "epoch": 2.5, + "learning_rate": 8.32457040266735e-06, + "loss": 0.3084, + "step": 65047 + }, + { + "epoch": 2.5, + "learning_rate": 8.323929212618622e-06, + "loss": 0.1904, + "step": 65048 + }, + { + "epoch": 2.5, + "learning_rate": 8.32328802256989e-06, + "loss": 0.2685, + "step": 65049 + }, + { + "epoch": 2.5, + "learning_rate": 8.32264683252116e-06, + "loss": 0.1196, + "step": 65050 + }, + { + "epoch": 2.5, + "learning_rate": 8.322005642472428e-06, + "loss": 0.1757, + "step": 65051 + }, + { + "epoch": 2.5, + "learning_rate": 8.3213644524237e-06, + "loss": 0.1994, + "step": 65052 + }, + { + "epoch": 2.5, + "learning_rate": 8.320723262374968e-06, + "loss": 0.3289, + "step": 65053 + }, + { + "epoch": 2.5, + "learning_rate": 8.320082072326238e-06, + "loss": 0.2431, + "step": 65054 + }, + { + "epoch": 2.5, + "learning_rate": 8.319440882277507e-06, + "loss": 0.2064, + "step": 65055 + }, + { + "epoch": 2.5, + "learning_rate": 8.318799692228777e-06, + "loss": 0.2222, + "step": 65056 + }, + { + "epoch": 2.5, + "learning_rate": 8.318158502180047e-06, + "loss": 0.1938, + "step": 65057 + }, + { + "epoch": 2.5, + "learning_rate": 8.317517312131317e-06, + "loss": 0.1929, + "step": 65058 + }, + { + "epoch": 2.5, + "learning_rate": 8.316876122082585e-06, + "loss": 0.2997, + "step": 65059 + }, + { + "epoch": 2.5, + "learning_rate": 8.316234932033855e-06, + "loss": 0.2425, + "step": 65060 + }, + { + "epoch": 2.5, + "learning_rate": 8.315593741985125e-06, + "loss": 0.435, + "step": 65061 + }, + { + "epoch": 2.5, + "learning_rate": 8.314952551936395e-06, + "loss": 0.1284, + "step": 65062 + }, + { + "epoch": 2.5, + "learning_rate": 8.314311361887664e-06, + "loss": 0.2979, + "step": 65063 + }, + { + "epoch": 2.5, + "learning_rate": 8.313670171838934e-06, + "loss": 0.2187, + "step": 65064 + }, + { + "epoch": 2.5, + "learning_rate": 8.313028981790202e-06, + "loss": 0.1579, + "step": 65065 + }, + { + "epoch": 2.5, + "learning_rate": 8.312387791741474e-06, + "loss": 0.3071, + "step": 65066 + }, + { + "epoch": 2.5, + "learning_rate": 8.311746601692742e-06, + "loss": 0.1322, + "step": 65067 + }, + { + "epoch": 2.5, + "learning_rate": 8.311105411644012e-06, + "loss": 0.1384, + "step": 65068 + }, + { + "epoch": 2.5, + "learning_rate": 8.31046422159528e-06, + "loss": 0.2863, + "step": 65069 + }, + { + "epoch": 2.5, + "learning_rate": 8.309823031546552e-06, + "loss": 0.093, + "step": 65070 + }, + { + "epoch": 2.5, + "learning_rate": 8.30918184149782e-06, + "loss": 0.3815, + "step": 65071 + }, + { + "epoch": 2.5, + "learning_rate": 8.30854065144909e-06, + "loss": 0.1548, + "step": 65072 + }, + { + "epoch": 2.5, + "learning_rate": 8.307899461400359e-06, + "loss": 0.1261, + "step": 65073 + }, + { + "epoch": 2.5, + "learning_rate": 8.307258271351629e-06, + "loss": 0.1436, + "step": 65074 + }, + { + "epoch": 2.5, + "learning_rate": 8.306617081302899e-06, + "loss": 0.3797, + "step": 65075 + }, + { + "epoch": 2.5, + "learning_rate": 8.305975891254169e-06, + "loss": 0.2596, + "step": 65076 + }, + { + "epoch": 2.5, + "learning_rate": 8.305334701205437e-06, + "loss": 0.1558, + "step": 65077 + }, + { + "epoch": 2.5, + "learning_rate": 8.304693511156707e-06, + "loss": 0.1686, + "step": 65078 + }, + { + "epoch": 2.5, + "learning_rate": 8.304052321107977e-06, + "loss": 0.1688, + "step": 65079 + }, + { + "epoch": 2.5, + "learning_rate": 8.303411131059247e-06, + "loss": 0.2933, + "step": 65080 + }, + { + "epoch": 2.5, + "learning_rate": 8.302769941010516e-06, + "loss": 0.2848, + "step": 65081 + }, + { + "epoch": 2.5, + "learning_rate": 8.302128750961786e-06, + "loss": 0.2288, + "step": 65082 + }, + { + "epoch": 2.5, + "learning_rate": 8.301487560913054e-06, + "loss": 0.2819, + "step": 65083 + }, + { + "epoch": 2.5, + "learning_rate": 8.300846370864326e-06, + "loss": 0.1662, + "step": 65084 + }, + { + "epoch": 2.5, + "learning_rate": 8.300205180815594e-06, + "loss": 0.1591, + "step": 65085 + }, + { + "epoch": 2.5, + "learning_rate": 8.299563990766864e-06, + "loss": 0.1713, + "step": 65086 + }, + { + "epoch": 2.5, + "learning_rate": 8.298922800718132e-06, + "loss": 0.2574, + "step": 65087 + }, + { + "epoch": 2.5, + "learning_rate": 8.298281610669404e-06, + "loss": 0.2336, + "step": 65088 + }, + { + "epoch": 2.5, + "learning_rate": 8.297640420620672e-06, + "loss": 0.1737, + "step": 65089 + }, + { + "epoch": 2.5, + "learning_rate": 8.296999230571942e-06, + "loss": 0.1828, + "step": 65090 + }, + { + "epoch": 2.5, + "learning_rate": 8.29635804052321e-06, + "loss": 0.1182, + "step": 65091 + }, + { + "epoch": 2.5, + "learning_rate": 8.295716850474482e-06, + "loss": 0.2562, + "step": 65092 + }, + { + "epoch": 2.5, + "learning_rate": 8.29507566042575e-06, + "loss": 0.279, + "step": 65093 + }, + { + "epoch": 2.5, + "learning_rate": 8.29443447037702e-06, + "loss": 0.1611, + "step": 65094 + }, + { + "epoch": 2.5, + "learning_rate": 8.293793280328289e-06, + "loss": 0.2052, + "step": 65095 + }, + { + "epoch": 2.5, + "learning_rate": 8.293152090279559e-06, + "loss": 0.2384, + "step": 65096 + }, + { + "epoch": 2.5, + "learning_rate": 8.292510900230829e-06, + "loss": 0.3331, + "step": 65097 + }, + { + "epoch": 2.5, + "learning_rate": 8.2918697101821e-06, + "loss": 0.2366, + "step": 65098 + }, + { + "epoch": 2.5, + "learning_rate": 8.291228520133367e-06, + "loss": 0.1752, + "step": 65099 + }, + { + "epoch": 2.5, + "learning_rate": 8.290587330084638e-06, + "loss": 0.1609, + "step": 65100 + }, + { + "epoch": 2.5, + "learning_rate": 8.289946140035908e-06, + "loss": 0.2292, + "step": 65101 + }, + { + "epoch": 2.5, + "learning_rate": 8.289304949987178e-06, + "loss": 0.2476, + "step": 65102 + }, + { + "epoch": 2.5, + "learning_rate": 8.288663759938446e-06, + "loss": 0.3087, + "step": 65103 + }, + { + "epoch": 2.5, + "learning_rate": 8.288022569889716e-06, + "loss": 0.0913, + "step": 65104 + }, + { + "epoch": 2.5, + "learning_rate": 8.287381379840984e-06, + "loss": 0.2192, + "step": 65105 + }, + { + "epoch": 2.5, + "learning_rate": 8.286740189792256e-06, + "loss": 0.163, + "step": 65106 + }, + { + "epoch": 2.5, + "learning_rate": 8.286098999743524e-06, + "loss": 0.1476, + "step": 65107 + }, + { + "epoch": 2.5, + "learning_rate": 8.285457809694794e-06, + "loss": 0.3085, + "step": 65108 + }, + { + "epoch": 2.5, + "learning_rate": 8.284816619646063e-06, + "loss": 0.2987, + "step": 65109 + }, + { + "epoch": 2.5, + "learning_rate": 8.284175429597334e-06, + "loss": 0.2511, + "step": 65110 + }, + { + "epoch": 2.5, + "learning_rate": 8.283534239548603e-06, + "loss": 0.2079, + "step": 65111 + }, + { + "epoch": 2.5, + "learning_rate": 8.282893049499873e-06, + "loss": 0.2883, + "step": 65112 + }, + { + "epoch": 2.5, + "learning_rate": 8.282251859451141e-06, + "loss": 0.2809, + "step": 65113 + }, + { + "epoch": 2.5, + "learning_rate": 8.281610669402411e-06, + "loss": 0.1869, + "step": 65114 + }, + { + "epoch": 2.5, + "learning_rate": 8.280969479353681e-06, + "loss": 0.1569, + "step": 65115 + }, + { + "epoch": 2.5, + "learning_rate": 8.280328289304951e-06, + "loss": 0.2184, + "step": 65116 + }, + { + "epoch": 2.5, + "learning_rate": 8.27968709925622e-06, + "loss": 0.183, + "step": 65117 + }, + { + "epoch": 2.5, + "learning_rate": 8.27904590920749e-06, + "loss": 0.2665, + "step": 65118 + }, + { + "epoch": 2.5, + "learning_rate": 8.27840471915876e-06, + "loss": 0.1686, + "step": 65119 + }, + { + "epoch": 2.5, + "learning_rate": 8.27776352911003e-06, + "loss": 0.1706, + "step": 65120 + }, + { + "epoch": 2.5, + "learning_rate": 8.277122339061298e-06, + "loss": 0.2889, + "step": 65121 + }, + { + "epoch": 2.5, + "learning_rate": 8.276481149012568e-06, + "loss": 0.1546, + "step": 65122 + }, + { + "epoch": 2.5, + "learning_rate": 8.275839958963836e-06, + "loss": 0.2137, + "step": 65123 + }, + { + "epoch": 2.5, + "learning_rate": 8.275198768915108e-06, + "loss": 0.2915, + "step": 65124 + }, + { + "epoch": 2.5, + "learning_rate": 8.274557578866376e-06, + "loss": 0.2273, + "step": 65125 + }, + { + "epoch": 2.5, + "learning_rate": 8.273916388817646e-06, + "loss": 0.2523, + "step": 65126 + }, + { + "epoch": 2.5, + "learning_rate": 8.273275198768915e-06, + "loss": 0.2567, + "step": 65127 + }, + { + "epoch": 2.5, + "learning_rate": 8.272634008720186e-06, + "loss": 0.3427, + "step": 65128 + }, + { + "epoch": 2.5, + "learning_rate": 8.271992818671455e-06, + "loss": 0.244, + "step": 65129 + }, + { + "epoch": 2.5, + "learning_rate": 8.271351628622725e-06, + "loss": 0.2333, + "step": 65130 + }, + { + "epoch": 2.5, + "learning_rate": 8.270710438573993e-06, + "loss": 0.2393, + "step": 65131 + }, + { + "epoch": 2.5, + "learning_rate": 8.270069248525263e-06, + "loss": 0.1528, + "step": 65132 + }, + { + "epoch": 2.5, + "learning_rate": 8.269428058476533e-06, + "loss": 0.381, + "step": 65133 + }, + { + "epoch": 2.5, + "learning_rate": 8.268786868427803e-06, + "loss": 0.3064, + "step": 65134 + }, + { + "epoch": 2.5, + "learning_rate": 8.268145678379071e-06, + "loss": 0.1102, + "step": 65135 + }, + { + "epoch": 2.5, + "learning_rate": 8.267504488330341e-06, + "loss": 0.3509, + "step": 65136 + }, + { + "epoch": 2.5, + "learning_rate": 8.266863298281611e-06, + "loss": 0.1004, + "step": 65137 + }, + { + "epoch": 2.5, + "learning_rate": 8.266222108232881e-06, + "loss": 0.1473, + "step": 65138 + }, + { + "epoch": 2.5, + "learning_rate": 8.26558091818415e-06, + "loss": 0.1285, + "step": 65139 + }, + { + "epoch": 2.5, + "learning_rate": 8.26493972813542e-06, + "loss": 0.2184, + "step": 65140 + }, + { + "epoch": 2.5, + "learning_rate": 8.26429853808669e-06, + "loss": 0.2325, + "step": 65141 + }, + { + "epoch": 2.5, + "learning_rate": 8.26365734803796e-06, + "loss": 0.2795, + "step": 65142 + }, + { + "epoch": 2.5, + "learning_rate": 8.263016157989228e-06, + "loss": 0.3628, + "step": 65143 + }, + { + "epoch": 2.5, + "learning_rate": 8.262374967940498e-06, + "loss": 0.1841, + "step": 65144 + }, + { + "epoch": 2.5, + "learning_rate": 8.261733777891767e-06, + "loss": 0.2025, + "step": 65145 + }, + { + "epoch": 2.5, + "learning_rate": 8.261092587843038e-06, + "loss": 0.2394, + "step": 65146 + }, + { + "epoch": 2.5, + "learning_rate": 8.260451397794307e-06, + "loss": 0.2461, + "step": 65147 + }, + { + "epoch": 2.5, + "learning_rate": 8.259810207745577e-06, + "loss": 0.2598, + "step": 65148 + }, + { + "epoch": 2.5, + "learning_rate": 8.259169017696845e-06, + "loss": 0.2518, + "step": 65149 + }, + { + "epoch": 2.5, + "learning_rate": 8.258527827648117e-06, + "loss": 0.1393, + "step": 65150 + }, + { + "epoch": 2.5, + "learning_rate": 8.257886637599385e-06, + "loss": 0.2217, + "step": 65151 + }, + { + "epoch": 2.5, + "learning_rate": 8.257245447550655e-06, + "loss": 0.2141, + "step": 65152 + }, + { + "epoch": 2.5, + "learning_rate": 8.256604257501923e-06, + "loss": 0.3594, + "step": 65153 + }, + { + "epoch": 2.5, + "learning_rate": 8.255963067453193e-06, + "loss": 0.1133, + "step": 65154 + }, + { + "epoch": 2.5, + "learning_rate": 8.255321877404463e-06, + "loss": 0.2656, + "step": 65155 + }, + { + "epoch": 2.51, + "learning_rate": 8.254680687355733e-06, + "loss": 0.2759, + "step": 65156 + }, + { + "epoch": 2.51, + "learning_rate": 8.254039497307002e-06, + "loss": 0.1889, + "step": 65157 + }, + { + "epoch": 2.51, + "learning_rate": 8.253398307258272e-06, + "loss": 0.2306, + "step": 65158 + }, + { + "epoch": 2.51, + "learning_rate": 8.252757117209542e-06, + "loss": 0.268, + "step": 65159 + }, + { + "epoch": 2.51, + "learning_rate": 8.252115927160812e-06, + "loss": 0.4014, + "step": 65160 + }, + { + "epoch": 2.51, + "learning_rate": 8.25147473711208e-06, + "loss": 0.2473, + "step": 65161 + }, + { + "epoch": 2.51, + "learning_rate": 8.25083354706335e-06, + "loss": 0.1507, + "step": 65162 + }, + { + "epoch": 2.51, + "learning_rate": 8.250192357014618e-06, + "loss": 0.2386, + "step": 65163 + }, + { + "epoch": 2.51, + "learning_rate": 8.24955116696589e-06, + "loss": 0.2027, + "step": 65164 + }, + { + "epoch": 2.51, + "learning_rate": 8.248909976917158e-06, + "loss": 0.1551, + "step": 65165 + }, + { + "epoch": 2.51, + "learning_rate": 8.248268786868429e-06, + "loss": 0.1117, + "step": 65166 + }, + { + "epoch": 2.51, + "learning_rate": 8.247627596819697e-06, + "loss": 0.3872, + "step": 65167 + }, + { + "epoch": 2.51, + "learning_rate": 8.246986406770969e-06, + "loss": 0.3896, + "step": 65168 + }, + { + "epoch": 2.51, + "learning_rate": 8.246345216722237e-06, + "loss": 0.1645, + "step": 65169 + }, + { + "epoch": 2.51, + "learning_rate": 8.245704026673507e-06, + "loss": 0.2921, + "step": 65170 + }, + { + "epoch": 2.51, + "learning_rate": 8.245062836624775e-06, + "loss": 0.1997, + "step": 65171 + }, + { + "epoch": 2.51, + "learning_rate": 8.244421646576045e-06, + "loss": 0.2155, + "step": 65172 + }, + { + "epoch": 2.51, + "learning_rate": 8.243780456527315e-06, + "loss": 0.2201, + "step": 65173 + }, + { + "epoch": 2.51, + "learning_rate": 8.243139266478585e-06, + "loss": 0.2933, + "step": 65174 + }, + { + "epoch": 2.51, + "learning_rate": 8.242498076429854e-06, + "loss": 0.2316, + "step": 65175 + }, + { + "epoch": 2.51, + "learning_rate": 8.241856886381124e-06, + "loss": 0.2193, + "step": 65176 + }, + { + "epoch": 2.51, + "learning_rate": 8.241215696332394e-06, + "loss": 0.1837, + "step": 65177 + }, + { + "epoch": 2.51, + "learning_rate": 8.240574506283664e-06, + "loss": 0.2591, + "step": 65178 + }, + { + "epoch": 2.51, + "learning_rate": 8.239933316234932e-06, + "loss": 0.1447, + "step": 65179 + }, + { + "epoch": 2.51, + "learning_rate": 8.239292126186202e-06, + "loss": 0.233, + "step": 65180 + }, + { + "epoch": 2.51, + "learning_rate": 8.23865093613747e-06, + "loss": 0.2327, + "step": 65181 + }, + { + "epoch": 2.51, + "learning_rate": 8.238009746088742e-06, + "loss": 0.2193, + "step": 65182 + }, + { + "epoch": 2.51, + "learning_rate": 8.23736855604001e-06, + "loss": 0.1855, + "step": 65183 + }, + { + "epoch": 2.51, + "learning_rate": 8.23672736599128e-06, + "loss": 0.172, + "step": 65184 + }, + { + "epoch": 2.51, + "learning_rate": 8.236086175942549e-06, + "loss": 0.2444, + "step": 65185 + }, + { + "epoch": 2.51, + "learning_rate": 8.23544498589382e-06, + "loss": 0.139, + "step": 65186 + }, + { + "epoch": 2.51, + "learning_rate": 8.234803795845089e-06, + "loss": 0.2314, + "step": 65187 + }, + { + "epoch": 2.51, + "learning_rate": 8.234162605796359e-06, + "loss": 0.1934, + "step": 65188 + }, + { + "epoch": 2.51, + "learning_rate": 8.233521415747627e-06, + "loss": 0.3264, + "step": 65189 + }, + { + "epoch": 2.51, + "learning_rate": 8.232880225698899e-06, + "loss": 0.1873, + "step": 65190 + }, + { + "epoch": 2.51, + "learning_rate": 8.232239035650167e-06, + "loss": 0.1805, + "step": 65191 + }, + { + "epoch": 2.51, + "learning_rate": 8.231597845601437e-06, + "loss": 0.1106, + "step": 65192 + }, + { + "epoch": 2.51, + "learning_rate": 8.230956655552706e-06, + "loss": 0.1811, + "step": 65193 + }, + { + "epoch": 2.51, + "learning_rate": 8.230315465503976e-06, + "loss": 0.1864, + "step": 65194 + }, + { + "epoch": 2.51, + "learning_rate": 8.229674275455246e-06, + "loss": 0.2107, + "step": 65195 + }, + { + "epoch": 2.51, + "learning_rate": 8.229033085406516e-06, + "loss": 0.362, + "step": 65196 + }, + { + "epoch": 2.51, + "learning_rate": 8.228391895357784e-06, + "loss": 0.1666, + "step": 65197 + }, + { + "epoch": 2.51, + "learning_rate": 8.227750705309054e-06, + "loss": 0.1609, + "step": 65198 + }, + { + "epoch": 2.51, + "learning_rate": 8.227109515260324e-06, + "loss": 0.1926, + "step": 65199 + }, + { + "epoch": 2.51, + "learning_rate": 8.226468325211594e-06, + "loss": 0.3873, + "step": 65200 + }, + { + "epoch": 2.51, + "learning_rate": 8.225827135162862e-06, + "loss": 0.2527, + "step": 65201 + }, + { + "epoch": 2.51, + "learning_rate": 8.225185945114132e-06, + "loss": 0.3779, + "step": 65202 + }, + { + "epoch": 2.51, + "learning_rate": 8.2245447550654e-06, + "loss": 0.233, + "step": 65203 + }, + { + "epoch": 2.51, + "learning_rate": 8.223903565016672e-06, + "loss": 0.2812, + "step": 65204 + }, + { + "epoch": 2.51, + "learning_rate": 8.22326237496794e-06, + "loss": 0.181, + "step": 65205 + }, + { + "epoch": 2.51, + "learning_rate": 8.22262118491921e-06, + "loss": 0.1708, + "step": 65206 + }, + { + "epoch": 2.51, + "learning_rate": 8.221979994870479e-06, + "loss": 0.371, + "step": 65207 + }, + { + "epoch": 2.51, + "learning_rate": 8.22133880482175e-06, + "loss": 0.1503, + "step": 65208 + }, + { + "epoch": 2.51, + "learning_rate": 8.220697614773019e-06, + "loss": 0.2336, + "step": 65209 + }, + { + "epoch": 2.51, + "learning_rate": 8.22005642472429e-06, + "loss": 0.1813, + "step": 65210 + }, + { + "epoch": 2.51, + "learning_rate": 8.219415234675557e-06, + "loss": 0.2211, + "step": 65211 + }, + { + "epoch": 2.51, + "learning_rate": 8.218774044626828e-06, + "loss": 0.3329, + "step": 65212 + }, + { + "epoch": 2.51, + "learning_rate": 8.218132854578098e-06, + "loss": 0.2148, + "step": 65213 + }, + { + "epoch": 2.51, + "learning_rate": 8.217491664529368e-06, + "loss": 0.1943, + "step": 65214 + }, + { + "epoch": 2.51, + "learning_rate": 8.216850474480636e-06, + "loss": 0.1725, + "step": 65215 + }, + { + "epoch": 2.51, + "learning_rate": 8.216209284431906e-06, + "loss": 0.1712, + "step": 65216 + }, + { + "epoch": 2.51, + "learning_rate": 8.215568094383176e-06, + "loss": 0.1526, + "step": 65217 + }, + { + "epoch": 2.51, + "learning_rate": 8.214926904334446e-06, + "loss": 0.2573, + "step": 65218 + }, + { + "epoch": 2.51, + "learning_rate": 8.214285714285714e-06, + "loss": 0.342, + "step": 65219 + }, + { + "epoch": 2.51, + "learning_rate": 8.213644524236984e-06, + "loss": 0.3174, + "step": 65220 + }, + { + "epoch": 2.51, + "learning_rate": 8.213003334188253e-06, + "loss": 0.4321, + "step": 65221 + }, + { + "epoch": 2.51, + "learning_rate": 8.212362144139524e-06, + "loss": 0.1908, + "step": 65222 + }, + { + "epoch": 2.51, + "learning_rate": 8.211720954090793e-06, + "loss": 0.1682, + "step": 65223 + }, + { + "epoch": 2.51, + "learning_rate": 8.211079764042063e-06, + "loss": 0.2214, + "step": 65224 + }, + { + "epoch": 2.51, + "learning_rate": 8.210438573993331e-06, + "loss": 0.2919, + "step": 65225 + }, + { + "epoch": 2.51, + "learning_rate": 8.209797383944603e-06, + "loss": 0.2307, + "step": 65226 + }, + { + "epoch": 2.51, + "learning_rate": 8.209156193895871e-06, + "loss": 0.2535, + "step": 65227 + }, + { + "epoch": 2.51, + "learning_rate": 8.208515003847141e-06, + "loss": 0.2289, + "step": 65228 + }, + { + "epoch": 2.51, + "learning_rate": 8.20787381379841e-06, + "loss": 0.1891, + "step": 65229 + }, + { + "epoch": 2.51, + "learning_rate": 8.20723262374968e-06, + "loss": 0.2722, + "step": 65230 + }, + { + "epoch": 2.51, + "learning_rate": 8.20659143370095e-06, + "loss": 0.2251, + "step": 65231 + }, + { + "epoch": 2.51, + "learning_rate": 8.20595024365222e-06, + "loss": 0.2069, + "step": 65232 + }, + { + "epoch": 2.51, + "learning_rate": 8.205309053603488e-06, + "loss": 0.277, + "step": 65233 + }, + { + "epoch": 2.51, + "learning_rate": 8.204667863554758e-06, + "loss": 0.3036, + "step": 65234 + }, + { + "epoch": 2.51, + "learning_rate": 8.204026673506028e-06, + "loss": 0.302, + "step": 65235 + }, + { + "epoch": 2.51, + "learning_rate": 8.203385483457298e-06, + "loss": 0.2946, + "step": 65236 + }, + { + "epoch": 2.51, + "learning_rate": 8.202744293408566e-06, + "loss": 0.2846, + "step": 65237 + }, + { + "epoch": 2.51, + "learning_rate": 8.202103103359836e-06, + "loss": 0.1754, + "step": 65238 + }, + { + "epoch": 2.51, + "learning_rate": 8.201461913311106e-06, + "loss": 0.2826, + "step": 65239 + }, + { + "epoch": 2.51, + "learning_rate": 8.200820723262376e-06, + "loss": 0.1413, + "step": 65240 + }, + { + "epoch": 2.51, + "learning_rate": 8.200179533213645e-06, + "loss": 0.1442, + "step": 65241 + }, + { + "epoch": 2.51, + "learning_rate": 8.199538343164915e-06, + "loss": 0.3223, + "step": 65242 + }, + { + "epoch": 2.51, + "learning_rate": 8.198897153116183e-06, + "loss": 0.2363, + "step": 65243 + }, + { + "epoch": 2.51, + "learning_rate": 8.198255963067455e-06, + "loss": 0.2906, + "step": 65244 + }, + { + "epoch": 2.51, + "learning_rate": 8.197614773018723e-06, + "loss": 0.1902, + "step": 65245 + }, + { + "epoch": 2.51, + "learning_rate": 8.196973582969993e-06, + "loss": 0.2314, + "step": 65246 + }, + { + "epoch": 2.51, + "learning_rate": 8.196332392921261e-06, + "loss": 0.1425, + "step": 65247 + }, + { + "epoch": 2.51, + "learning_rate": 8.195691202872533e-06, + "loss": 0.2403, + "step": 65248 + }, + { + "epoch": 2.51, + "learning_rate": 8.195050012823801e-06, + "loss": 0.2191, + "step": 65249 + }, + { + "epoch": 2.51, + "learning_rate": 8.194408822775071e-06, + "loss": 0.3939, + "step": 65250 + }, + { + "epoch": 2.51, + "learning_rate": 8.19376763272634e-06, + "loss": 0.2316, + "step": 65251 + }, + { + "epoch": 2.51, + "learning_rate": 8.19312644267761e-06, + "loss": 0.1624, + "step": 65252 + }, + { + "epoch": 2.51, + "learning_rate": 8.19248525262888e-06, + "loss": 0.199, + "step": 65253 + }, + { + "epoch": 2.51, + "learning_rate": 8.19184406258015e-06, + "loss": 0.234, + "step": 65254 + }, + { + "epoch": 2.51, + "learning_rate": 8.191202872531418e-06, + "loss": 0.1858, + "step": 65255 + }, + { + "epoch": 2.51, + "learning_rate": 8.190561682482688e-06, + "loss": 0.1736, + "step": 65256 + }, + { + "epoch": 2.51, + "learning_rate": 8.189920492433958e-06, + "loss": 0.1578, + "step": 65257 + }, + { + "epoch": 2.51, + "learning_rate": 8.189279302385228e-06, + "loss": 0.2914, + "step": 65258 + }, + { + "epoch": 2.51, + "learning_rate": 8.188638112336497e-06, + "loss": 0.2905, + "step": 65259 + }, + { + "epoch": 2.51, + "learning_rate": 8.187996922287767e-06, + "loss": 0.1789, + "step": 65260 + }, + { + "epoch": 2.51, + "learning_rate": 8.187355732239035e-06, + "loss": 0.1873, + "step": 65261 + }, + { + "epoch": 2.51, + "learning_rate": 8.186714542190307e-06, + "loss": 0.3577, + "step": 65262 + }, + { + "epoch": 2.51, + "learning_rate": 8.186073352141575e-06, + "loss": 0.2236, + "step": 65263 + }, + { + "epoch": 2.51, + "learning_rate": 8.185432162092845e-06, + "loss": 0.1253, + "step": 65264 + }, + { + "epoch": 2.51, + "learning_rate": 8.184790972044113e-06, + "loss": 0.1649, + "step": 65265 + }, + { + "epoch": 2.51, + "learning_rate": 8.184149781995385e-06, + "loss": 0.2185, + "step": 65266 + }, + { + "epoch": 2.51, + "learning_rate": 8.183508591946653e-06, + "loss": 0.1747, + "step": 65267 + }, + { + "epoch": 2.51, + "learning_rate": 8.182867401897923e-06, + "loss": 0.2431, + "step": 65268 + }, + { + "epoch": 2.51, + "learning_rate": 8.182226211849192e-06, + "loss": 0.1616, + "step": 65269 + }, + { + "epoch": 2.51, + "learning_rate": 8.181585021800462e-06, + "loss": 0.1835, + "step": 65270 + }, + { + "epoch": 2.51, + "learning_rate": 8.180943831751732e-06, + "loss": 0.149, + "step": 65271 + }, + { + "epoch": 2.51, + "learning_rate": 8.180302641703002e-06, + "loss": 0.1588, + "step": 65272 + }, + { + "epoch": 2.51, + "learning_rate": 8.17966145165427e-06, + "loss": 0.2314, + "step": 65273 + }, + { + "epoch": 2.51, + "learning_rate": 8.17902026160554e-06, + "loss": 0.1971, + "step": 65274 + }, + { + "epoch": 2.51, + "learning_rate": 8.17837907155681e-06, + "loss": 0.1672, + "step": 65275 + }, + { + "epoch": 2.51, + "learning_rate": 8.17773788150808e-06, + "loss": 0.142, + "step": 65276 + }, + { + "epoch": 2.51, + "learning_rate": 8.177096691459348e-06, + "loss": 0.1911, + "step": 65277 + }, + { + "epoch": 2.51, + "learning_rate": 8.176455501410619e-06, + "loss": 0.348, + "step": 65278 + }, + { + "epoch": 2.51, + "learning_rate": 8.175814311361889e-06, + "loss": 0.2307, + "step": 65279 + }, + { + "epoch": 2.51, + "learning_rate": 8.175173121313159e-06, + "loss": 0.3087, + "step": 65280 + }, + { + "epoch": 2.51, + "learning_rate": 8.174531931264427e-06, + "loss": 0.1566, + "step": 65281 + }, + { + "epoch": 2.51, + "learning_rate": 8.173890741215697e-06, + "loss": 0.2479, + "step": 65282 + }, + { + "epoch": 2.51, + "learning_rate": 8.173249551166965e-06, + "loss": 0.2123, + "step": 65283 + }, + { + "epoch": 2.51, + "learning_rate": 8.172608361118237e-06, + "loss": 0.3111, + "step": 65284 + }, + { + "epoch": 2.51, + "learning_rate": 8.171967171069505e-06, + "loss": 0.3094, + "step": 65285 + }, + { + "epoch": 2.51, + "learning_rate": 8.171325981020775e-06, + "loss": 0.2607, + "step": 65286 + }, + { + "epoch": 2.51, + "learning_rate": 8.170684790972044e-06, + "loss": 0.2835, + "step": 65287 + }, + { + "epoch": 2.51, + "learning_rate": 8.170043600923315e-06, + "loss": 0.2022, + "step": 65288 + }, + { + "epoch": 2.51, + "learning_rate": 8.169402410874584e-06, + "loss": 0.1359, + "step": 65289 + }, + { + "epoch": 2.51, + "learning_rate": 8.168761220825854e-06, + "loss": 0.1602, + "step": 65290 + }, + { + "epoch": 2.51, + "learning_rate": 8.168120030777122e-06, + "loss": 0.2868, + "step": 65291 + }, + { + "epoch": 2.51, + "learning_rate": 8.167478840728392e-06, + "loss": 0.1849, + "step": 65292 + }, + { + "epoch": 2.51, + "learning_rate": 8.166837650679662e-06, + "loss": 0.2976, + "step": 65293 + }, + { + "epoch": 2.51, + "learning_rate": 8.166196460630932e-06, + "loss": 0.201, + "step": 65294 + }, + { + "epoch": 2.51, + "learning_rate": 8.1655552705822e-06, + "loss": 0.2033, + "step": 65295 + }, + { + "epoch": 2.51, + "learning_rate": 8.16491408053347e-06, + "loss": 0.2768, + "step": 65296 + }, + { + "epoch": 2.51, + "learning_rate": 8.16427289048474e-06, + "loss": 0.3769, + "step": 65297 + }, + { + "epoch": 2.51, + "learning_rate": 8.16363170043601e-06, + "loss": 0.2993, + "step": 65298 + }, + { + "epoch": 2.51, + "learning_rate": 8.162990510387279e-06, + "loss": 0.1782, + "step": 65299 + }, + { + "epoch": 2.51, + "learning_rate": 8.162349320338549e-06, + "loss": 0.4322, + "step": 65300 + }, + { + "epoch": 2.51, + "learning_rate": 8.161708130289817e-06, + "loss": 0.2319, + "step": 65301 + }, + { + "epoch": 2.51, + "learning_rate": 8.161066940241089e-06, + "loss": 0.1897, + "step": 65302 + }, + { + "epoch": 2.51, + "learning_rate": 8.160425750192357e-06, + "loss": 0.294, + "step": 65303 + }, + { + "epoch": 2.51, + "learning_rate": 8.159784560143627e-06, + "loss": 0.2592, + "step": 65304 + }, + { + "epoch": 2.51, + "learning_rate": 8.159143370094896e-06, + "loss": 0.2327, + "step": 65305 + }, + { + "epoch": 2.51, + "learning_rate": 8.158502180046167e-06, + "loss": 0.2484, + "step": 65306 + }, + { + "epoch": 2.51, + "learning_rate": 8.157860989997436e-06, + "loss": 0.2132, + "step": 65307 + }, + { + "epoch": 2.51, + "learning_rate": 8.157219799948706e-06, + "loss": 0.1489, + "step": 65308 + }, + { + "epoch": 2.51, + "learning_rate": 8.156578609899974e-06, + "loss": 0.1468, + "step": 65309 + }, + { + "epoch": 2.51, + "learning_rate": 8.155937419851244e-06, + "loss": 0.3275, + "step": 65310 + }, + { + "epoch": 2.51, + "learning_rate": 8.155296229802514e-06, + "loss": 0.1557, + "step": 65311 + }, + { + "epoch": 2.51, + "learning_rate": 8.154655039753784e-06, + "loss": 0.1365, + "step": 65312 + }, + { + "epoch": 2.51, + "learning_rate": 8.154013849705052e-06, + "loss": 0.2249, + "step": 65313 + }, + { + "epoch": 2.51, + "learning_rate": 8.153372659656322e-06, + "loss": 0.256, + "step": 65314 + }, + { + "epoch": 2.51, + "learning_rate": 8.152731469607592e-06, + "loss": 0.1688, + "step": 65315 + }, + { + "epoch": 2.51, + "learning_rate": 8.152090279558862e-06, + "loss": 0.2751, + "step": 65316 + }, + { + "epoch": 2.51, + "learning_rate": 8.15144908951013e-06, + "loss": 0.1486, + "step": 65317 + }, + { + "epoch": 2.51, + "learning_rate": 8.1508078994614e-06, + "loss": 0.1928, + "step": 65318 + }, + { + "epoch": 2.51, + "learning_rate": 8.150166709412669e-06, + "loss": 0.247, + "step": 65319 + }, + { + "epoch": 2.51, + "learning_rate": 8.14952551936394e-06, + "loss": 0.1463, + "step": 65320 + }, + { + "epoch": 2.51, + "learning_rate": 8.14888432931521e-06, + "loss": 0.3307, + "step": 65321 + }, + { + "epoch": 2.51, + "learning_rate": 8.14824313926648e-06, + "loss": 0.2114, + "step": 65322 + }, + { + "epoch": 2.51, + "learning_rate": 8.147601949217748e-06, + "loss": 0.2035, + "step": 65323 + }, + { + "epoch": 2.51, + "learning_rate": 8.14696075916902e-06, + "loss": 0.2496, + "step": 65324 + }, + { + "epoch": 2.51, + "learning_rate": 8.146319569120288e-06, + "loss": 0.2142, + "step": 65325 + }, + { + "epoch": 2.51, + "learning_rate": 8.145678379071558e-06, + "loss": 0.2041, + "step": 65326 + }, + { + "epoch": 2.51, + "learning_rate": 8.145037189022826e-06, + "loss": 0.28, + "step": 65327 + }, + { + "epoch": 2.51, + "learning_rate": 8.144395998974098e-06, + "loss": 0.2969, + "step": 65328 + }, + { + "epoch": 2.51, + "learning_rate": 8.143754808925366e-06, + "loss": 0.2388, + "step": 65329 + }, + { + "epoch": 2.51, + "learning_rate": 8.143113618876636e-06, + "loss": 0.1473, + "step": 65330 + }, + { + "epoch": 2.51, + "learning_rate": 8.142472428827904e-06, + "loss": 0.2543, + "step": 65331 + }, + { + "epoch": 2.51, + "learning_rate": 8.141831238779174e-06, + "loss": 0.2322, + "step": 65332 + }, + { + "epoch": 2.51, + "learning_rate": 8.141190048730444e-06, + "loss": 0.1981, + "step": 65333 + }, + { + "epoch": 2.51, + "learning_rate": 8.140548858681714e-06, + "loss": 0.1045, + "step": 65334 + }, + { + "epoch": 2.51, + "learning_rate": 8.139907668632983e-06, + "loss": 0.2324, + "step": 65335 + }, + { + "epoch": 2.51, + "learning_rate": 8.139266478584253e-06, + "loss": 0.1811, + "step": 65336 + }, + { + "epoch": 2.51, + "learning_rate": 8.138625288535523e-06, + "loss": 0.1638, + "step": 65337 + }, + { + "epoch": 2.51, + "learning_rate": 8.137984098486793e-06, + "loss": 0.1232, + "step": 65338 + }, + { + "epoch": 2.51, + "learning_rate": 8.137342908438061e-06, + "loss": 0.2495, + "step": 65339 + }, + { + "epoch": 2.51, + "learning_rate": 8.136701718389331e-06, + "loss": 0.2121, + "step": 65340 + }, + { + "epoch": 2.51, + "learning_rate": 8.1360605283406e-06, + "loss": 0.3602, + "step": 65341 + }, + { + "epoch": 2.51, + "learning_rate": 8.135419338291871e-06, + "loss": 0.209, + "step": 65342 + }, + { + "epoch": 2.51, + "learning_rate": 8.13477814824314e-06, + "loss": 0.2397, + "step": 65343 + }, + { + "epoch": 2.51, + "learning_rate": 8.13413695819441e-06, + "loss": 0.3451, + "step": 65344 + }, + { + "epoch": 2.51, + "learning_rate": 8.133495768145678e-06, + "loss": 0.2691, + "step": 65345 + }, + { + "epoch": 2.51, + "learning_rate": 8.13285457809695e-06, + "loss": 0.1692, + "step": 65346 + }, + { + "epoch": 2.51, + "learning_rate": 8.132213388048218e-06, + "loss": 0.0876, + "step": 65347 + }, + { + "epoch": 2.51, + "learning_rate": 8.131572197999488e-06, + "loss": 0.1348, + "step": 65348 + }, + { + "epoch": 2.51, + "learning_rate": 8.130931007950756e-06, + "loss": 0.3601, + "step": 65349 + }, + { + "epoch": 2.51, + "learning_rate": 8.130289817902026e-06, + "loss": 0.1272, + "step": 65350 + }, + { + "epoch": 2.51, + "learning_rate": 8.129648627853296e-06, + "loss": 0.2309, + "step": 65351 + }, + { + "epoch": 2.51, + "learning_rate": 8.129007437804566e-06, + "loss": 0.1778, + "step": 65352 + }, + { + "epoch": 2.51, + "learning_rate": 8.128366247755835e-06, + "loss": 0.2543, + "step": 65353 + }, + { + "epoch": 2.51, + "learning_rate": 8.127725057707105e-06, + "loss": 0.1568, + "step": 65354 + }, + { + "epoch": 2.51, + "learning_rate": 8.127083867658375e-06, + "loss": 0.1675, + "step": 65355 + }, + { + "epoch": 2.51, + "learning_rate": 8.126442677609645e-06, + "loss": 0.1443, + "step": 65356 + }, + { + "epoch": 2.51, + "learning_rate": 8.125801487560913e-06, + "loss": 0.2999, + "step": 65357 + }, + { + "epoch": 2.51, + "learning_rate": 8.125160297512183e-06, + "loss": 0.2162, + "step": 65358 + }, + { + "epoch": 2.51, + "learning_rate": 8.124519107463451e-06, + "loss": 0.1819, + "step": 65359 + }, + { + "epoch": 2.51, + "learning_rate": 8.123877917414723e-06, + "loss": 0.1643, + "step": 65360 + }, + { + "epoch": 2.51, + "learning_rate": 8.123236727365991e-06, + "loss": 0.1569, + "step": 65361 + }, + { + "epoch": 2.51, + "learning_rate": 8.122595537317261e-06, + "loss": 0.1953, + "step": 65362 + }, + { + "epoch": 2.51, + "learning_rate": 8.12195434726853e-06, + "loss": 0.1329, + "step": 65363 + }, + { + "epoch": 2.51, + "learning_rate": 8.121313157219801e-06, + "loss": 0.1412, + "step": 65364 + }, + { + "epoch": 2.51, + "learning_rate": 8.12067196717107e-06, + "loss": 0.1587, + "step": 65365 + }, + { + "epoch": 2.51, + "learning_rate": 8.12003077712234e-06, + "loss": 0.1842, + "step": 65366 + }, + { + "epoch": 2.51, + "learning_rate": 8.119389587073608e-06, + "loss": 0.1171, + "step": 65367 + }, + { + "epoch": 2.51, + "learning_rate": 8.118748397024878e-06, + "loss": 0.1587, + "step": 65368 + }, + { + "epoch": 2.51, + "learning_rate": 8.118107206976148e-06, + "loss": 0.1938, + "step": 65369 + }, + { + "epoch": 2.51, + "learning_rate": 8.117466016927418e-06, + "loss": 0.1974, + "step": 65370 + }, + { + "epoch": 2.51, + "learning_rate": 8.116824826878687e-06, + "loss": 0.1252, + "step": 65371 + }, + { + "epoch": 2.51, + "learning_rate": 8.116183636829957e-06, + "loss": 0.1542, + "step": 65372 + }, + { + "epoch": 2.51, + "learning_rate": 8.115542446781227e-06, + "loss": 0.2901, + "step": 65373 + }, + { + "epoch": 2.51, + "learning_rate": 8.114901256732497e-06, + "loss": 0.1169, + "step": 65374 + }, + { + "epoch": 2.51, + "learning_rate": 8.114260066683765e-06, + "loss": 0.3082, + "step": 65375 + }, + { + "epoch": 2.51, + "learning_rate": 8.113618876635035e-06, + "loss": 0.2495, + "step": 65376 + }, + { + "epoch": 2.51, + "learning_rate": 8.112977686586305e-06, + "loss": 0.1595, + "step": 65377 + }, + { + "epoch": 2.51, + "learning_rate": 8.112336496537575e-06, + "loss": 0.3396, + "step": 65378 + }, + { + "epoch": 2.51, + "learning_rate": 8.111695306488843e-06, + "loss": 0.194, + "step": 65379 + }, + { + "epoch": 2.51, + "learning_rate": 8.111054116440113e-06, + "loss": 0.1886, + "step": 65380 + }, + { + "epoch": 2.51, + "learning_rate": 8.110412926391382e-06, + "loss": 0.1404, + "step": 65381 + }, + { + "epoch": 2.51, + "learning_rate": 8.109771736342653e-06, + "loss": 0.1257, + "step": 65382 + }, + { + "epoch": 2.51, + "learning_rate": 8.109130546293922e-06, + "loss": 0.2193, + "step": 65383 + }, + { + "epoch": 2.51, + "learning_rate": 8.108489356245192e-06, + "loss": 0.1646, + "step": 65384 + }, + { + "epoch": 2.51, + "learning_rate": 8.10784816619646e-06, + "loss": 0.32, + "step": 65385 + }, + { + "epoch": 2.51, + "learning_rate": 8.107206976147732e-06, + "loss": 0.3548, + "step": 65386 + }, + { + "epoch": 2.51, + "learning_rate": 8.106565786099e-06, + "loss": 0.235, + "step": 65387 + }, + { + "epoch": 2.51, + "learning_rate": 8.10592459605027e-06, + "loss": 0.1861, + "step": 65388 + }, + { + "epoch": 2.51, + "learning_rate": 8.105283406001539e-06, + "loss": 0.1468, + "step": 65389 + }, + { + "epoch": 2.51, + "learning_rate": 8.104642215952809e-06, + "loss": 0.2687, + "step": 65390 + }, + { + "epoch": 2.51, + "learning_rate": 8.104001025904079e-06, + "loss": 0.3889, + "step": 65391 + }, + { + "epoch": 2.51, + "learning_rate": 8.103359835855349e-06, + "loss": 0.2788, + "step": 65392 + }, + { + "epoch": 2.51, + "learning_rate": 8.102718645806617e-06, + "loss": 0.1304, + "step": 65393 + }, + { + "epoch": 2.51, + "learning_rate": 8.102077455757887e-06, + "loss": 0.3183, + "step": 65394 + }, + { + "epoch": 2.51, + "learning_rate": 8.101436265709157e-06, + "loss": 0.2504, + "step": 65395 + }, + { + "epoch": 2.51, + "learning_rate": 8.100795075660427e-06, + "loss": 0.3399, + "step": 65396 + }, + { + "epoch": 2.51, + "learning_rate": 8.100153885611695e-06, + "loss": 0.1752, + "step": 65397 + }, + { + "epoch": 2.51, + "learning_rate": 8.099512695562965e-06, + "loss": 0.1133, + "step": 65398 + }, + { + "epoch": 2.51, + "learning_rate": 8.098871505514234e-06, + "loss": 0.1778, + "step": 65399 + }, + { + "epoch": 2.51, + "learning_rate": 8.098230315465505e-06, + "loss": 0.261, + "step": 65400 + }, + { + "epoch": 2.51, + "learning_rate": 8.097589125416774e-06, + "loss": 0.1426, + "step": 65401 + }, + { + "epoch": 2.51, + "learning_rate": 8.096947935368044e-06, + "loss": 0.1457, + "step": 65402 + }, + { + "epoch": 2.51, + "learning_rate": 8.096306745319312e-06, + "loss": 0.427, + "step": 65403 + }, + { + "epoch": 2.51, + "learning_rate": 8.095665555270584e-06, + "loss": 0.112, + "step": 65404 + }, + { + "epoch": 2.51, + "learning_rate": 8.095024365221852e-06, + "loss": 0.0817, + "step": 65405 + }, + { + "epoch": 2.51, + "learning_rate": 8.094383175173122e-06, + "loss": 0.3268, + "step": 65406 + }, + { + "epoch": 2.51, + "learning_rate": 8.09374198512439e-06, + "loss": 0.1993, + "step": 65407 + }, + { + "epoch": 2.51, + "learning_rate": 8.09310079507566e-06, + "loss": 0.2495, + "step": 65408 + }, + { + "epoch": 2.51, + "learning_rate": 8.09245960502693e-06, + "loss": 0.1747, + "step": 65409 + }, + { + "epoch": 2.51, + "learning_rate": 8.0918184149782e-06, + "loss": 0.1965, + "step": 65410 + }, + { + "epoch": 2.51, + "learning_rate": 8.091177224929469e-06, + "loss": 0.1605, + "step": 65411 + }, + { + "epoch": 2.51, + "learning_rate": 8.090536034880739e-06, + "loss": 0.2342, + "step": 65412 + }, + { + "epoch": 2.51, + "learning_rate": 8.089894844832009e-06, + "loss": 0.197, + "step": 65413 + }, + { + "epoch": 2.51, + "learning_rate": 8.089253654783279e-06, + "loss": 0.327, + "step": 65414 + }, + { + "epoch": 2.51, + "learning_rate": 8.088612464734547e-06, + "loss": 0.1631, + "step": 65415 + }, + { + "epoch": 2.52, + "learning_rate": 8.087971274685817e-06, + "loss": 0.1527, + "step": 65416 + }, + { + "epoch": 2.52, + "learning_rate": 8.087330084637086e-06, + "loss": 0.1884, + "step": 65417 + }, + { + "epoch": 2.52, + "learning_rate": 8.086688894588357e-06, + "loss": 0.2397, + "step": 65418 + }, + { + "epoch": 2.52, + "learning_rate": 8.086047704539626e-06, + "loss": 0.2558, + "step": 65419 + }, + { + "epoch": 2.52, + "learning_rate": 8.085406514490896e-06, + "loss": 0.1813, + "step": 65420 + }, + { + "epoch": 2.52, + "learning_rate": 8.084765324442164e-06, + "loss": 0.3333, + "step": 65421 + }, + { + "epoch": 2.52, + "learning_rate": 8.084124134393436e-06, + "loss": 0.1784, + "step": 65422 + }, + { + "epoch": 2.52, + "learning_rate": 8.083482944344704e-06, + "loss": 0.18, + "step": 65423 + }, + { + "epoch": 2.52, + "learning_rate": 8.082841754295974e-06, + "loss": 0.3851, + "step": 65424 + }, + { + "epoch": 2.52, + "learning_rate": 8.082200564247242e-06, + "loss": 0.2038, + "step": 65425 + }, + { + "epoch": 2.52, + "learning_rate": 8.081559374198514e-06, + "loss": 0.1599, + "step": 65426 + }, + { + "epoch": 2.52, + "learning_rate": 8.080918184149782e-06, + "loss": 0.2025, + "step": 65427 + }, + { + "epoch": 2.52, + "learning_rate": 8.080276994101052e-06, + "loss": 0.3456, + "step": 65428 + }, + { + "epoch": 2.52, + "learning_rate": 8.07963580405232e-06, + "loss": 0.3645, + "step": 65429 + }, + { + "epoch": 2.52, + "learning_rate": 8.07899461400359e-06, + "loss": 0.2328, + "step": 65430 + }, + { + "epoch": 2.52, + "learning_rate": 8.07835342395486e-06, + "loss": 0.2967, + "step": 65431 + }, + { + "epoch": 2.52, + "learning_rate": 8.07771223390613e-06, + "loss": 0.141, + "step": 65432 + }, + { + "epoch": 2.52, + "learning_rate": 8.0770710438574e-06, + "loss": 0.271, + "step": 65433 + }, + { + "epoch": 2.52, + "learning_rate": 8.07642985380867e-06, + "loss": 0.1109, + "step": 65434 + }, + { + "epoch": 2.52, + "learning_rate": 8.07578866375994e-06, + "loss": 0.1744, + "step": 65435 + }, + { + "epoch": 2.52, + "learning_rate": 8.07514747371121e-06, + "loss": 0.3172, + "step": 65436 + }, + { + "epoch": 2.52, + "learning_rate": 8.074506283662478e-06, + "loss": 0.1574, + "step": 65437 + }, + { + "epoch": 2.52, + "learning_rate": 8.073865093613748e-06, + "loss": 0.1881, + "step": 65438 + }, + { + "epoch": 2.52, + "learning_rate": 8.073223903565016e-06, + "loss": 0.1791, + "step": 65439 + }, + { + "epoch": 2.52, + "learning_rate": 8.072582713516288e-06, + "loss": 0.1467, + "step": 65440 + }, + { + "epoch": 2.52, + "learning_rate": 8.071941523467556e-06, + "loss": 0.1555, + "step": 65441 + }, + { + "epoch": 2.52, + "learning_rate": 8.071300333418826e-06, + "loss": 0.2241, + "step": 65442 + }, + { + "epoch": 2.52, + "learning_rate": 8.070659143370094e-06, + "loss": 0.2501, + "step": 65443 + }, + { + "epoch": 2.52, + "learning_rate": 8.070017953321366e-06, + "loss": 0.238, + "step": 65444 + }, + { + "epoch": 2.52, + "learning_rate": 8.069376763272634e-06, + "loss": 0.2768, + "step": 65445 + }, + { + "epoch": 2.52, + "learning_rate": 8.068735573223904e-06, + "loss": 0.2154, + "step": 65446 + }, + { + "epoch": 2.52, + "learning_rate": 8.068094383175173e-06, + "loss": 0.1792, + "step": 65447 + }, + { + "epoch": 2.52, + "learning_rate": 8.067453193126443e-06, + "loss": 0.1782, + "step": 65448 + }, + { + "epoch": 2.52, + "learning_rate": 8.066812003077713e-06, + "loss": 0.2324, + "step": 65449 + }, + { + "epoch": 2.52, + "learning_rate": 8.066170813028983e-06, + "loss": 0.2067, + "step": 65450 + }, + { + "epoch": 2.52, + "learning_rate": 8.065529622980251e-06, + "loss": 0.1674, + "step": 65451 + }, + { + "epoch": 2.52, + "learning_rate": 8.064888432931521e-06, + "loss": 0.2272, + "step": 65452 + }, + { + "epoch": 2.52, + "learning_rate": 8.064247242882791e-06, + "loss": 0.1832, + "step": 65453 + }, + { + "epoch": 2.52, + "learning_rate": 8.063606052834061e-06, + "loss": 0.3085, + "step": 65454 + }, + { + "epoch": 2.52, + "learning_rate": 8.06296486278533e-06, + "loss": 0.1841, + "step": 65455 + }, + { + "epoch": 2.52, + "learning_rate": 8.0623236727366e-06, + "loss": 0.1122, + "step": 65456 + }, + { + "epoch": 2.52, + "learning_rate": 8.061682482687868e-06, + "loss": 0.1818, + "step": 65457 + }, + { + "epoch": 2.52, + "learning_rate": 8.06104129263914e-06, + "loss": 0.099, + "step": 65458 + }, + { + "epoch": 2.52, + "learning_rate": 8.060400102590408e-06, + "loss": 0.1864, + "step": 65459 + }, + { + "epoch": 2.52, + "learning_rate": 8.059758912541678e-06, + "loss": 0.2964, + "step": 65460 + }, + { + "epoch": 2.52, + "learning_rate": 8.059117722492946e-06, + "loss": 0.211, + "step": 65461 + }, + { + "epoch": 2.52, + "learning_rate": 8.058476532444218e-06, + "loss": 0.1639, + "step": 65462 + }, + { + "epoch": 2.52, + "learning_rate": 8.057835342395486e-06, + "loss": 0.2317, + "step": 65463 + }, + { + "epoch": 2.52, + "learning_rate": 8.057194152346756e-06, + "loss": 0.2976, + "step": 65464 + }, + { + "epoch": 2.52, + "learning_rate": 8.056552962298025e-06, + "loss": 0.2451, + "step": 65465 + }, + { + "epoch": 2.52, + "learning_rate": 8.055911772249295e-06, + "loss": 0.1207, + "step": 65466 + }, + { + "epoch": 2.52, + "learning_rate": 8.055270582200565e-06, + "loss": 0.2839, + "step": 65467 + }, + { + "epoch": 2.52, + "learning_rate": 8.054629392151835e-06, + "loss": 0.1196, + "step": 65468 + }, + { + "epoch": 2.52, + "learning_rate": 8.053988202103103e-06, + "loss": 0.1562, + "step": 65469 + }, + { + "epoch": 2.52, + "learning_rate": 8.053347012054373e-06, + "loss": 0.2441, + "step": 65470 + }, + { + "epoch": 2.52, + "learning_rate": 8.052705822005643e-06, + "loss": 0.1133, + "step": 65471 + }, + { + "epoch": 2.52, + "learning_rate": 8.052064631956913e-06, + "loss": 0.2223, + "step": 65472 + }, + { + "epoch": 2.52, + "learning_rate": 8.051423441908181e-06, + "loss": 0.268, + "step": 65473 + }, + { + "epoch": 2.52, + "learning_rate": 8.050782251859451e-06, + "loss": 0.1224, + "step": 65474 + }, + { + "epoch": 2.52, + "learning_rate": 8.050141061810721e-06, + "loss": 0.3221, + "step": 65475 + }, + { + "epoch": 2.52, + "learning_rate": 8.049499871761992e-06, + "loss": 0.2656, + "step": 65476 + }, + { + "epoch": 2.52, + "learning_rate": 8.04885868171326e-06, + "loss": 0.0583, + "step": 65477 + }, + { + "epoch": 2.52, + "learning_rate": 8.04821749166453e-06, + "loss": 0.2394, + "step": 65478 + }, + { + "epoch": 2.52, + "learning_rate": 8.047576301615798e-06, + "loss": 0.3899, + "step": 65479 + }, + { + "epoch": 2.52, + "learning_rate": 8.04693511156707e-06, + "loss": 0.1416, + "step": 65480 + }, + { + "epoch": 2.52, + "learning_rate": 8.046293921518338e-06, + "loss": 0.1646, + "step": 65481 + }, + { + "epoch": 2.52, + "learning_rate": 8.045652731469608e-06, + "loss": 0.3342, + "step": 65482 + }, + { + "epoch": 2.52, + "learning_rate": 8.045011541420877e-06, + "loss": 0.2618, + "step": 65483 + }, + { + "epoch": 2.52, + "learning_rate": 8.044370351372148e-06, + "loss": 0.3178, + "step": 65484 + }, + { + "epoch": 2.52, + "learning_rate": 8.043729161323417e-06, + "loss": 0.1623, + "step": 65485 + }, + { + "epoch": 2.52, + "learning_rate": 8.043087971274687e-06, + "loss": 0.1224, + "step": 65486 + }, + { + "epoch": 2.52, + "learning_rate": 8.042446781225955e-06, + "loss": 0.0931, + "step": 65487 + }, + { + "epoch": 2.52, + "learning_rate": 8.041805591177225e-06, + "loss": 0.1215, + "step": 65488 + }, + { + "epoch": 2.52, + "learning_rate": 8.041164401128495e-06, + "loss": 0.2428, + "step": 65489 + }, + { + "epoch": 2.52, + "learning_rate": 8.040523211079765e-06, + "loss": 0.2411, + "step": 65490 + }, + { + "epoch": 2.52, + "learning_rate": 8.039882021031033e-06, + "loss": 0.3526, + "step": 65491 + }, + { + "epoch": 2.52, + "learning_rate": 8.039240830982303e-06, + "loss": 0.1964, + "step": 65492 + }, + { + "epoch": 2.52, + "learning_rate": 8.038599640933573e-06, + "loss": 0.2479, + "step": 65493 + }, + { + "epoch": 2.52, + "learning_rate": 8.037958450884843e-06, + "loss": 0.3216, + "step": 65494 + }, + { + "epoch": 2.52, + "learning_rate": 8.037317260836112e-06, + "loss": 0.2678, + "step": 65495 + }, + { + "epoch": 2.52, + "learning_rate": 8.036676070787382e-06, + "loss": 0.2293, + "step": 65496 + }, + { + "epoch": 2.52, + "learning_rate": 8.03603488073865e-06, + "loss": 0.2236, + "step": 65497 + }, + { + "epoch": 2.52, + "learning_rate": 8.035393690689922e-06, + "loss": 0.1267, + "step": 65498 + }, + { + "epoch": 2.52, + "learning_rate": 8.03475250064119e-06, + "loss": 0.2458, + "step": 65499 + }, + { + "epoch": 2.52, + "learning_rate": 8.03411131059246e-06, + "loss": 0.1994, + "step": 65500 + }, + { + "epoch": 2.52, + "learning_rate": 8.033470120543729e-06, + "loss": 0.2456, + "step": 65501 + }, + { + "epoch": 2.52, + "learning_rate": 8.032828930495e-06, + "loss": 0.1517, + "step": 65502 + }, + { + "epoch": 2.52, + "learning_rate": 8.032187740446269e-06, + "loss": 0.3262, + "step": 65503 + }, + { + "epoch": 2.52, + "learning_rate": 8.031546550397539e-06, + "loss": 0.29, + "step": 65504 + }, + { + "epoch": 2.52, + "learning_rate": 8.030905360348807e-06, + "loss": 0.1886, + "step": 65505 + }, + { + "epoch": 2.52, + "learning_rate": 8.030264170300077e-06, + "loss": 0.1168, + "step": 65506 + }, + { + "epoch": 2.52, + "learning_rate": 8.029622980251347e-06, + "loss": 0.2025, + "step": 65507 + }, + { + "epoch": 2.52, + "learning_rate": 8.028981790202617e-06, + "loss": 0.2238, + "step": 65508 + }, + { + "epoch": 2.52, + "learning_rate": 8.028340600153885e-06, + "loss": 0.1695, + "step": 65509 + }, + { + "epoch": 2.52, + "learning_rate": 8.027699410105155e-06, + "loss": 0.1898, + "step": 65510 + }, + { + "epoch": 2.52, + "learning_rate": 8.027058220056425e-06, + "loss": 0.2344, + "step": 65511 + }, + { + "epoch": 2.52, + "learning_rate": 8.026417030007695e-06, + "loss": 0.2971, + "step": 65512 + }, + { + "epoch": 2.52, + "learning_rate": 8.025775839958964e-06, + "loss": 0.1971, + "step": 65513 + }, + { + "epoch": 2.52, + "learning_rate": 8.025134649910234e-06, + "loss": 0.1919, + "step": 65514 + }, + { + "epoch": 2.52, + "learning_rate": 8.024493459861504e-06, + "loss": 0.1468, + "step": 65515 + }, + { + "epoch": 2.52, + "learning_rate": 8.023852269812774e-06, + "loss": 0.1318, + "step": 65516 + }, + { + "epoch": 2.52, + "learning_rate": 8.023211079764042e-06, + "loss": 0.1843, + "step": 65517 + }, + { + "epoch": 2.52, + "learning_rate": 8.022569889715312e-06, + "loss": 0.2769, + "step": 65518 + }, + { + "epoch": 2.52, + "learning_rate": 8.02192869966658e-06, + "loss": 0.291, + "step": 65519 + }, + { + "epoch": 2.52, + "learning_rate": 8.021287509617852e-06, + "loss": 0.22, + "step": 65520 + }, + { + "epoch": 2.52, + "learning_rate": 8.02064631956912e-06, + "loss": 0.3472, + "step": 65521 + }, + { + "epoch": 2.52, + "learning_rate": 8.02000512952039e-06, + "loss": 0.2774, + "step": 65522 + }, + { + "epoch": 2.52, + "learning_rate": 8.019363939471659e-06, + "loss": 0.1961, + "step": 65523 + }, + { + "epoch": 2.52, + "learning_rate": 8.01872274942293e-06, + "loss": 0.1847, + "step": 65524 + }, + { + "epoch": 2.52, + "learning_rate": 8.018081559374199e-06, + "loss": 0.409, + "step": 65525 + }, + { + "epoch": 2.52, + "learning_rate": 8.017440369325469e-06, + "loss": 0.172, + "step": 65526 + }, + { + "epoch": 2.52, + "learning_rate": 8.016799179276737e-06, + "loss": 0.2046, + "step": 65527 + }, + { + "epoch": 2.52, + "learning_rate": 8.016157989228007e-06, + "loss": 0.1387, + "step": 65528 + }, + { + "epoch": 2.52, + "learning_rate": 8.015516799179277e-06, + "loss": 0.2182, + "step": 65529 + }, + { + "epoch": 2.52, + "learning_rate": 8.014875609130547e-06, + "loss": 0.1937, + "step": 65530 + }, + { + "epoch": 2.52, + "learning_rate": 8.014234419081816e-06, + "loss": 0.1489, + "step": 65531 + }, + { + "epoch": 2.52, + "learning_rate": 8.013593229033086e-06, + "loss": 0.3087, + "step": 65532 + }, + { + "epoch": 2.52, + "learning_rate": 8.012952038984356e-06, + "loss": 0.1834, + "step": 65533 + }, + { + "epoch": 2.52, + "learning_rate": 8.012310848935626e-06, + "loss": 0.1751, + "step": 65534 + }, + { + "epoch": 2.52, + "learning_rate": 8.011669658886894e-06, + "loss": 0.1641, + "step": 65535 + }, + { + "epoch": 2.52, + "learning_rate": 8.011028468838164e-06, + "loss": 0.1906, + "step": 65536 + }, + { + "epoch": 2.52, + "learning_rate": 8.010387278789432e-06, + "loss": 0.1786, + "step": 65537 + }, + { + "epoch": 2.52, + "learning_rate": 8.009746088740704e-06, + "loss": 0.173, + "step": 65538 + }, + { + "epoch": 2.52, + "learning_rate": 8.009104898691972e-06, + "loss": 0.1571, + "step": 65539 + }, + { + "epoch": 2.52, + "learning_rate": 8.008463708643242e-06, + "loss": 0.2369, + "step": 65540 + }, + { + "epoch": 2.52, + "learning_rate": 8.00782251859451e-06, + "loss": 0.2242, + "step": 65541 + }, + { + "epoch": 2.52, + "learning_rate": 8.007181328545783e-06, + "loss": 0.1303, + "step": 65542 + }, + { + "epoch": 2.52, + "learning_rate": 8.00654013849705e-06, + "loss": 0.1827, + "step": 65543 + }, + { + "epoch": 2.52, + "learning_rate": 8.005898948448321e-06, + "loss": 0.3077, + "step": 65544 + }, + { + "epoch": 2.52, + "learning_rate": 8.00525775839959e-06, + "loss": 0.2126, + "step": 65545 + }, + { + "epoch": 2.52, + "learning_rate": 8.00461656835086e-06, + "loss": 0.2481, + "step": 65546 + }, + { + "epoch": 2.52, + "learning_rate": 8.00397537830213e-06, + "loss": 0.1645, + "step": 65547 + }, + { + "epoch": 2.52, + "learning_rate": 8.0033341882534e-06, + "loss": 0.2769, + "step": 65548 + }, + { + "epoch": 2.52, + "learning_rate": 8.002692998204668e-06, + "loss": 0.1157, + "step": 65549 + }, + { + "epoch": 2.52, + "learning_rate": 8.002051808155938e-06, + "loss": 0.1738, + "step": 65550 + }, + { + "epoch": 2.52, + "learning_rate": 8.001410618107208e-06, + "loss": 0.3186, + "step": 65551 + }, + { + "epoch": 2.52, + "learning_rate": 8.000769428058478e-06, + "loss": 0.2678, + "step": 65552 + }, + { + "epoch": 2.52, + "learning_rate": 8.000128238009746e-06, + "loss": 0.1252, + "step": 65553 + }, + { + "epoch": 2.52, + "learning_rate": 7.999487047961016e-06, + "loss": 0.2772, + "step": 65554 + }, + { + "epoch": 2.52, + "learning_rate": 7.998845857912284e-06, + "loss": 0.284, + "step": 65555 + }, + { + "epoch": 2.52, + "learning_rate": 7.998204667863556e-06, + "loss": 0.4352, + "step": 65556 + }, + { + "epoch": 2.52, + "learning_rate": 7.997563477814824e-06, + "loss": 0.1624, + "step": 65557 + }, + { + "epoch": 2.52, + "learning_rate": 7.996922287766094e-06, + "loss": 0.1954, + "step": 65558 + }, + { + "epoch": 2.52, + "learning_rate": 7.996281097717363e-06, + "loss": 0.2739, + "step": 65559 + }, + { + "epoch": 2.52, + "learning_rate": 7.995639907668634e-06, + "loss": 0.2095, + "step": 65560 + }, + { + "epoch": 2.52, + "learning_rate": 7.994998717619903e-06, + "loss": 0.1929, + "step": 65561 + }, + { + "epoch": 2.52, + "learning_rate": 7.994357527571173e-06, + "loss": 0.2141, + "step": 65562 + }, + { + "epoch": 2.52, + "learning_rate": 7.993716337522441e-06, + "loss": 0.1812, + "step": 65563 + }, + { + "epoch": 2.52, + "learning_rate": 7.993075147473713e-06, + "loss": 0.2717, + "step": 65564 + }, + { + "epoch": 2.52, + "learning_rate": 7.992433957424981e-06, + "loss": 0.2384, + "step": 65565 + }, + { + "epoch": 2.52, + "learning_rate": 7.991792767376251e-06, + "loss": 0.251, + "step": 65566 + }, + { + "epoch": 2.52, + "learning_rate": 7.99115157732752e-06, + "loss": 0.304, + "step": 65567 + }, + { + "epoch": 2.52, + "learning_rate": 7.99051038727879e-06, + "loss": 0.1377, + "step": 65568 + }, + { + "epoch": 2.52, + "learning_rate": 7.98986919723006e-06, + "loss": 0.3499, + "step": 65569 + }, + { + "epoch": 2.52, + "learning_rate": 7.98922800718133e-06, + "loss": 0.267, + "step": 65570 + }, + { + "epoch": 2.52, + "learning_rate": 7.988586817132598e-06, + "loss": 0.2326, + "step": 65571 + }, + { + "epoch": 2.52, + "learning_rate": 7.987945627083868e-06, + "loss": 0.2203, + "step": 65572 + }, + { + "epoch": 2.52, + "learning_rate": 7.987304437035138e-06, + "loss": 0.1864, + "step": 65573 + }, + { + "epoch": 2.52, + "learning_rate": 7.986663246986408e-06, + "loss": 0.3436, + "step": 65574 + }, + { + "epoch": 2.52, + "learning_rate": 7.986022056937676e-06, + "loss": 0.2024, + "step": 65575 + }, + { + "epoch": 2.52, + "learning_rate": 7.985380866888946e-06, + "loss": 0.1969, + "step": 65576 + }, + { + "epoch": 2.52, + "learning_rate": 7.984739676840215e-06, + "loss": 0.175, + "step": 65577 + }, + { + "epoch": 2.52, + "learning_rate": 7.984098486791486e-06, + "loss": 0.1392, + "step": 65578 + }, + { + "epoch": 2.52, + "learning_rate": 7.983457296742755e-06, + "loss": 0.1564, + "step": 65579 + }, + { + "epoch": 2.52, + "learning_rate": 7.982816106694025e-06, + "loss": 0.188, + "step": 65580 + }, + { + "epoch": 2.52, + "learning_rate": 7.982174916645293e-06, + "loss": 0.224, + "step": 65581 + }, + { + "epoch": 2.52, + "learning_rate": 7.981533726596565e-06, + "loss": 0.1745, + "step": 65582 + }, + { + "epoch": 2.52, + "learning_rate": 7.980892536547833e-06, + "loss": 0.2285, + "step": 65583 + }, + { + "epoch": 2.52, + "learning_rate": 7.980251346499103e-06, + "loss": 0.1755, + "step": 65584 + }, + { + "epoch": 2.52, + "learning_rate": 7.979610156450371e-06, + "loss": 0.1926, + "step": 65585 + }, + { + "epoch": 2.52, + "learning_rate": 7.978968966401641e-06, + "loss": 0.2133, + "step": 65586 + }, + { + "epoch": 2.52, + "learning_rate": 7.978327776352911e-06, + "loss": 0.2528, + "step": 65587 + }, + { + "epoch": 2.52, + "learning_rate": 7.977686586304182e-06, + "loss": 0.2533, + "step": 65588 + }, + { + "epoch": 2.52, + "learning_rate": 7.97704539625545e-06, + "loss": 0.2323, + "step": 65589 + }, + { + "epoch": 2.52, + "learning_rate": 7.97640420620672e-06, + "loss": 0.165, + "step": 65590 + }, + { + "epoch": 2.52, + "learning_rate": 7.97576301615799e-06, + "loss": 0.3188, + "step": 65591 + }, + { + "epoch": 2.52, + "learning_rate": 7.97512182610926e-06, + "loss": 0.1974, + "step": 65592 + }, + { + "epoch": 2.52, + "learning_rate": 7.974480636060528e-06, + "loss": 0.1911, + "step": 65593 + }, + { + "epoch": 2.52, + "learning_rate": 7.973839446011798e-06, + "loss": 0.1823, + "step": 65594 + }, + { + "epoch": 2.52, + "learning_rate": 7.973198255963067e-06, + "loss": 0.3001, + "step": 65595 + }, + { + "epoch": 2.52, + "learning_rate": 7.972557065914338e-06, + "loss": 0.3786, + "step": 65596 + }, + { + "epoch": 2.52, + "learning_rate": 7.971915875865607e-06, + "loss": 0.2885, + "step": 65597 + }, + { + "epoch": 2.52, + "learning_rate": 7.971274685816877e-06, + "loss": 0.3199, + "step": 65598 + }, + { + "epoch": 2.52, + "learning_rate": 7.970633495768145e-06, + "loss": 0.2767, + "step": 65599 + }, + { + "epoch": 2.52, + "learning_rate": 7.969992305719417e-06, + "loss": 0.1197, + "step": 65600 + }, + { + "epoch": 2.52, + "learning_rate": 7.969351115670685e-06, + "loss": 0.1879, + "step": 65601 + }, + { + "epoch": 2.52, + "learning_rate": 7.968709925621955e-06, + "loss": 0.2146, + "step": 65602 + }, + { + "epoch": 2.52, + "learning_rate": 7.968068735573223e-06, + "loss": 0.1748, + "step": 65603 + }, + { + "epoch": 2.52, + "learning_rate": 7.967427545524493e-06, + "loss": 0.1851, + "step": 65604 + }, + { + "epoch": 2.52, + "learning_rate": 7.966786355475763e-06, + "loss": 0.2844, + "step": 65605 + }, + { + "epoch": 2.52, + "learning_rate": 7.966145165427033e-06, + "loss": 0.1052, + "step": 65606 + }, + { + "epoch": 2.52, + "learning_rate": 7.965503975378302e-06, + "loss": 0.2907, + "step": 65607 + }, + { + "epoch": 2.52, + "learning_rate": 7.964862785329572e-06, + "loss": 0.2299, + "step": 65608 + }, + { + "epoch": 2.52, + "learning_rate": 7.964221595280842e-06, + "loss": 0.1723, + "step": 65609 + }, + { + "epoch": 2.52, + "learning_rate": 7.963580405232112e-06, + "loss": 0.1638, + "step": 65610 + }, + { + "epoch": 2.52, + "learning_rate": 7.96293921518338e-06, + "loss": 0.1689, + "step": 65611 + }, + { + "epoch": 2.52, + "learning_rate": 7.96229802513465e-06, + "loss": 0.106, + "step": 65612 + }, + { + "epoch": 2.52, + "learning_rate": 7.96165683508592e-06, + "loss": 0.2846, + "step": 65613 + }, + { + "epoch": 2.52, + "learning_rate": 7.96101564503719e-06, + "loss": 0.1917, + "step": 65614 + }, + { + "epoch": 2.52, + "learning_rate": 7.960374454988459e-06, + "loss": 0.1519, + "step": 65615 + }, + { + "epoch": 2.52, + "learning_rate": 7.959733264939729e-06, + "loss": 0.2135, + "step": 65616 + }, + { + "epoch": 2.52, + "learning_rate": 7.959092074890997e-06, + "loss": 0.3018, + "step": 65617 + }, + { + "epoch": 2.52, + "learning_rate": 7.958450884842269e-06, + "loss": 0.1757, + "step": 65618 + }, + { + "epoch": 2.52, + "learning_rate": 7.957809694793537e-06, + "loss": 0.1988, + "step": 65619 + }, + { + "epoch": 2.52, + "learning_rate": 7.957168504744807e-06, + "loss": 0.2168, + "step": 65620 + }, + { + "epoch": 2.52, + "learning_rate": 7.956527314696075e-06, + "loss": 0.2748, + "step": 65621 + }, + { + "epoch": 2.52, + "learning_rate": 7.955886124647347e-06, + "loss": 0.2031, + "step": 65622 + }, + { + "epoch": 2.52, + "learning_rate": 7.955244934598615e-06, + "loss": 0.1963, + "step": 65623 + }, + { + "epoch": 2.52, + "learning_rate": 7.954603744549885e-06, + "loss": 0.2396, + "step": 65624 + }, + { + "epoch": 2.52, + "learning_rate": 7.953962554501154e-06, + "loss": 0.1777, + "step": 65625 + }, + { + "epoch": 2.52, + "learning_rate": 7.953321364452424e-06, + "loss": 0.291, + "step": 65626 + }, + { + "epoch": 2.52, + "learning_rate": 7.952680174403694e-06, + "loss": 0.2415, + "step": 65627 + }, + { + "epoch": 2.52, + "learning_rate": 7.952038984354964e-06, + "loss": 0.2229, + "step": 65628 + }, + { + "epoch": 2.52, + "learning_rate": 7.951397794306232e-06, + "loss": 0.1849, + "step": 65629 + }, + { + "epoch": 2.52, + "learning_rate": 7.950756604257502e-06, + "loss": 0.3055, + "step": 65630 + }, + { + "epoch": 2.52, + "learning_rate": 7.950115414208772e-06, + "loss": 0.2706, + "step": 65631 + }, + { + "epoch": 2.52, + "learning_rate": 7.949474224160042e-06, + "loss": 0.3019, + "step": 65632 + }, + { + "epoch": 2.52, + "learning_rate": 7.94883303411131e-06, + "loss": 0.1844, + "step": 65633 + }, + { + "epoch": 2.52, + "learning_rate": 7.94819184406258e-06, + "loss": 0.2281, + "step": 65634 + }, + { + "epoch": 2.52, + "learning_rate": 7.947550654013849e-06, + "loss": 0.158, + "step": 65635 + }, + { + "epoch": 2.52, + "learning_rate": 7.94690946396512e-06, + "loss": 0.2399, + "step": 65636 + }, + { + "epoch": 2.52, + "learning_rate": 7.946268273916389e-06, + "loss": 0.2413, + "step": 65637 + }, + { + "epoch": 2.52, + "learning_rate": 7.945627083867659e-06, + "loss": 0.2511, + "step": 65638 + }, + { + "epoch": 2.52, + "learning_rate": 7.944985893818927e-06, + "loss": 0.2037, + "step": 65639 + }, + { + "epoch": 2.52, + "learning_rate": 7.944344703770199e-06, + "loss": 0.1849, + "step": 65640 + }, + { + "epoch": 2.52, + "learning_rate": 7.943703513721467e-06, + "loss": 0.2862, + "step": 65641 + }, + { + "epoch": 2.52, + "learning_rate": 7.943062323672737e-06, + "loss": 0.236, + "step": 65642 + }, + { + "epoch": 2.52, + "learning_rate": 7.942421133624006e-06, + "loss": 0.1401, + "step": 65643 + }, + { + "epoch": 2.52, + "learning_rate": 7.941779943575276e-06, + "loss": 0.2634, + "step": 65644 + }, + { + "epoch": 2.52, + "learning_rate": 7.941138753526546e-06, + "loss": 0.2107, + "step": 65645 + }, + { + "epoch": 2.52, + "learning_rate": 7.940497563477816e-06, + "loss": 0.3625, + "step": 65646 + }, + { + "epoch": 2.52, + "learning_rate": 7.939856373429084e-06, + "loss": 0.2198, + "step": 65647 + }, + { + "epoch": 2.52, + "learning_rate": 7.939215183380354e-06, + "loss": 0.1304, + "step": 65648 + }, + { + "epoch": 2.52, + "learning_rate": 7.938573993331624e-06, + "loss": 0.1629, + "step": 65649 + }, + { + "epoch": 2.52, + "learning_rate": 7.937932803282894e-06, + "loss": 0.2548, + "step": 65650 + }, + { + "epoch": 2.52, + "learning_rate": 7.937291613234162e-06, + "loss": 0.2275, + "step": 65651 + }, + { + "epoch": 2.52, + "learning_rate": 7.936650423185432e-06, + "loss": 0.1855, + "step": 65652 + }, + { + "epoch": 2.52, + "learning_rate": 7.9360092331367e-06, + "loss": 0.2276, + "step": 65653 + }, + { + "epoch": 2.52, + "learning_rate": 7.935368043087973e-06, + "loss": 0.2934, + "step": 65654 + }, + { + "epoch": 2.52, + "learning_rate": 7.93472685303924e-06, + "loss": 0.2919, + "step": 65655 + }, + { + "epoch": 2.52, + "learning_rate": 7.934085662990511e-06, + "loss": 0.2255, + "step": 65656 + }, + { + "epoch": 2.52, + "learning_rate": 7.93344447294178e-06, + "loss": 0.3112, + "step": 65657 + }, + { + "epoch": 2.52, + "learning_rate": 7.932803282893051e-06, + "loss": 0.1852, + "step": 65658 + }, + { + "epoch": 2.52, + "learning_rate": 7.93216209284432e-06, + "loss": 0.2149, + "step": 65659 + }, + { + "epoch": 2.52, + "learning_rate": 7.93152090279559e-06, + "loss": 0.2894, + "step": 65660 + }, + { + "epoch": 2.52, + "learning_rate": 7.930879712746858e-06, + "loss": 0.1276, + "step": 65661 + }, + { + "epoch": 2.52, + "learning_rate": 7.93023852269813e-06, + "loss": 0.1475, + "step": 65662 + }, + { + "epoch": 2.52, + "learning_rate": 7.929597332649398e-06, + "loss": 0.2756, + "step": 65663 + }, + { + "epoch": 2.52, + "learning_rate": 7.928956142600668e-06, + "loss": 0.2786, + "step": 65664 + }, + { + "epoch": 2.52, + "learning_rate": 7.928314952551936e-06, + "loss": 0.1699, + "step": 65665 + }, + { + "epoch": 2.52, + "learning_rate": 7.927673762503206e-06, + "loss": 0.1728, + "step": 65666 + }, + { + "epoch": 2.52, + "learning_rate": 7.927032572454476e-06, + "loss": 0.1712, + "step": 65667 + }, + { + "epoch": 2.52, + "learning_rate": 7.926391382405746e-06, + "loss": 0.3196, + "step": 65668 + }, + { + "epoch": 2.52, + "learning_rate": 7.925750192357014e-06, + "loss": 0.1098, + "step": 65669 + }, + { + "epoch": 2.52, + "learning_rate": 7.925109002308284e-06, + "loss": 0.1595, + "step": 65670 + }, + { + "epoch": 2.52, + "learning_rate": 7.924467812259554e-06, + "loss": 0.1248, + "step": 65671 + }, + { + "epoch": 2.52, + "learning_rate": 7.923826622210824e-06, + "loss": 0.1338, + "step": 65672 + }, + { + "epoch": 2.52, + "learning_rate": 7.923185432162093e-06, + "loss": 0.218, + "step": 65673 + }, + { + "epoch": 2.52, + "learning_rate": 7.922544242113363e-06, + "loss": 0.1722, + "step": 65674 + }, + { + "epoch": 2.52, + "learning_rate": 7.921903052064631e-06, + "loss": 0.1597, + "step": 65675 + }, + { + "epoch": 2.53, + "learning_rate": 7.921261862015903e-06, + "loss": 0.3172, + "step": 65676 + }, + { + "epoch": 2.53, + "learning_rate": 7.920620671967171e-06, + "loss": 0.1597, + "step": 65677 + }, + { + "epoch": 2.53, + "learning_rate": 7.919979481918441e-06, + "loss": 0.3635, + "step": 65678 + }, + { + "epoch": 2.53, + "learning_rate": 7.91933829186971e-06, + "loss": 0.2993, + "step": 65679 + }, + { + "epoch": 2.53, + "learning_rate": 7.918697101820981e-06, + "loss": 0.4209, + "step": 65680 + }, + { + "epoch": 2.53, + "learning_rate": 7.91805591177225e-06, + "loss": 0.3209, + "step": 65681 + }, + { + "epoch": 2.53, + "learning_rate": 7.91741472172352e-06, + "loss": 0.3081, + "step": 65682 + }, + { + "epoch": 2.53, + "learning_rate": 7.916773531674788e-06, + "loss": 0.1616, + "step": 65683 + }, + { + "epoch": 2.53, + "learning_rate": 7.916132341626058e-06, + "loss": 0.1944, + "step": 65684 + }, + { + "epoch": 2.53, + "learning_rate": 7.915491151577328e-06, + "loss": 0.1413, + "step": 65685 + }, + { + "epoch": 2.53, + "learning_rate": 7.914849961528598e-06, + "loss": 0.2106, + "step": 65686 + }, + { + "epoch": 2.53, + "learning_rate": 7.914208771479866e-06, + "loss": 0.1925, + "step": 65687 + }, + { + "epoch": 2.53, + "learning_rate": 7.913567581431136e-06, + "loss": 0.1509, + "step": 65688 + }, + { + "epoch": 2.53, + "learning_rate": 7.912926391382406e-06, + "loss": 0.1794, + "step": 65689 + }, + { + "epoch": 2.53, + "learning_rate": 7.912285201333676e-06, + "loss": 0.2209, + "step": 65690 + }, + { + "epoch": 2.53, + "learning_rate": 7.911644011284945e-06, + "loss": 0.24, + "step": 65691 + }, + { + "epoch": 2.53, + "learning_rate": 7.911002821236215e-06, + "loss": 0.2258, + "step": 65692 + }, + { + "epoch": 2.53, + "learning_rate": 7.910361631187483e-06, + "loss": 0.1958, + "step": 65693 + }, + { + "epoch": 2.53, + "learning_rate": 7.909720441138755e-06, + "loss": 0.1925, + "step": 65694 + }, + { + "epoch": 2.53, + "learning_rate": 7.909079251090023e-06, + "loss": 0.1325, + "step": 65695 + }, + { + "epoch": 2.53, + "learning_rate": 7.908438061041293e-06, + "loss": 0.2164, + "step": 65696 + }, + { + "epoch": 2.53, + "learning_rate": 7.907796870992561e-06, + "loss": 0.269, + "step": 65697 + }, + { + "epoch": 2.53, + "learning_rate": 7.907155680943833e-06, + "loss": 0.1219, + "step": 65698 + }, + { + "epoch": 2.53, + "learning_rate": 7.906514490895102e-06, + "loss": 0.1873, + "step": 65699 + }, + { + "epoch": 2.53, + "learning_rate": 7.905873300846372e-06, + "loss": 0.3011, + "step": 65700 + }, + { + "epoch": 2.53, + "learning_rate": 7.90523211079764e-06, + "loss": 0.267, + "step": 65701 + }, + { + "epoch": 2.53, + "learning_rate": 7.90459092074891e-06, + "loss": 0.2091, + "step": 65702 + }, + { + "epoch": 2.53, + "learning_rate": 7.90394973070018e-06, + "loss": 0.1867, + "step": 65703 + }, + { + "epoch": 2.53, + "learning_rate": 7.90330854065145e-06, + "loss": 0.1864, + "step": 65704 + }, + { + "epoch": 2.53, + "learning_rate": 7.902667350602718e-06, + "loss": 0.2318, + "step": 65705 + }, + { + "epoch": 2.53, + "learning_rate": 7.902026160553988e-06, + "loss": 0.22, + "step": 65706 + }, + { + "epoch": 2.53, + "learning_rate": 7.901384970505258e-06, + "loss": 0.2311, + "step": 65707 + }, + { + "epoch": 2.53, + "learning_rate": 7.900743780456528e-06, + "loss": 0.1631, + "step": 65708 + }, + { + "epoch": 2.53, + "learning_rate": 7.900102590407797e-06, + "loss": 0.1161, + "step": 65709 + }, + { + "epoch": 2.53, + "learning_rate": 7.899461400359067e-06, + "loss": 0.192, + "step": 65710 + }, + { + "epoch": 2.53, + "learning_rate": 7.898820210310337e-06, + "loss": 0.143, + "step": 65711 + }, + { + "epoch": 2.53, + "learning_rate": 7.898179020261607e-06, + "loss": 0.2629, + "step": 65712 + }, + { + "epoch": 2.53, + "learning_rate": 7.897537830212875e-06, + "loss": 0.2372, + "step": 65713 + }, + { + "epoch": 2.53, + "learning_rate": 7.896896640164145e-06, + "loss": 0.2563, + "step": 65714 + }, + { + "epoch": 2.53, + "learning_rate": 7.896255450115413e-06, + "loss": 0.3053, + "step": 65715 + }, + { + "epoch": 2.53, + "learning_rate": 7.895614260066685e-06, + "loss": 0.1825, + "step": 65716 + }, + { + "epoch": 2.53, + "learning_rate": 7.894973070017953e-06, + "loss": 0.112, + "step": 65717 + }, + { + "epoch": 2.53, + "learning_rate": 7.894331879969223e-06, + "loss": 0.212, + "step": 65718 + }, + { + "epoch": 2.53, + "learning_rate": 7.893690689920492e-06, + "loss": 0.2171, + "step": 65719 + }, + { + "epoch": 2.53, + "learning_rate": 7.893049499871764e-06, + "loss": 0.1524, + "step": 65720 + }, + { + "epoch": 2.53, + "learning_rate": 7.892408309823032e-06, + "loss": 0.1773, + "step": 65721 + }, + { + "epoch": 2.53, + "learning_rate": 7.891767119774302e-06, + "loss": 0.3324, + "step": 65722 + }, + { + "epoch": 2.53, + "learning_rate": 7.89112592972557e-06, + "loss": 0.1667, + "step": 65723 + }, + { + "epoch": 2.53, + "learning_rate": 7.89048473967684e-06, + "loss": 0.1241, + "step": 65724 + }, + { + "epoch": 2.53, + "learning_rate": 7.88984354962811e-06, + "loss": 0.2238, + "step": 65725 + }, + { + "epoch": 2.53, + "learning_rate": 7.88920235957938e-06, + "loss": 0.147, + "step": 65726 + }, + { + "epoch": 2.53, + "learning_rate": 7.888561169530649e-06, + "loss": 0.2562, + "step": 65727 + }, + { + "epoch": 2.53, + "learning_rate": 7.887919979481919e-06, + "loss": 0.344, + "step": 65728 + }, + { + "epoch": 2.53, + "learning_rate": 7.887278789433189e-06, + "loss": 0.3486, + "step": 65729 + }, + { + "epoch": 2.53, + "learning_rate": 7.886637599384459e-06, + "loss": 0.34, + "step": 65730 + }, + { + "epoch": 2.53, + "learning_rate": 7.885996409335727e-06, + "loss": 0.2292, + "step": 65731 + }, + { + "epoch": 2.53, + "learning_rate": 7.885355219286997e-06, + "loss": 0.2751, + "step": 65732 + }, + { + "epoch": 2.53, + "learning_rate": 7.884714029238265e-06, + "loss": 0.1832, + "step": 65733 + }, + { + "epoch": 2.53, + "learning_rate": 7.884072839189537e-06, + "loss": 0.249, + "step": 65734 + }, + { + "epoch": 2.53, + "learning_rate": 7.883431649140805e-06, + "loss": 0.3025, + "step": 65735 + }, + { + "epoch": 2.53, + "learning_rate": 7.882790459092075e-06, + "loss": 0.1603, + "step": 65736 + }, + { + "epoch": 2.53, + "learning_rate": 7.882149269043344e-06, + "loss": 0.1883, + "step": 65737 + }, + { + "epoch": 2.53, + "learning_rate": 7.881508078994615e-06, + "loss": 0.2264, + "step": 65738 + }, + { + "epoch": 2.53, + "learning_rate": 7.880866888945884e-06, + "loss": 0.3854, + "step": 65739 + }, + { + "epoch": 2.53, + "learning_rate": 7.880225698897154e-06, + "loss": 0.2022, + "step": 65740 + }, + { + "epoch": 2.53, + "learning_rate": 7.879584508848422e-06, + "loss": 0.141, + "step": 65741 + }, + { + "epoch": 2.53, + "learning_rate": 7.878943318799692e-06, + "loss": 0.3115, + "step": 65742 + }, + { + "epoch": 2.53, + "learning_rate": 7.878302128750962e-06, + "loss": 0.2264, + "step": 65743 + }, + { + "epoch": 2.53, + "learning_rate": 7.877660938702232e-06, + "loss": 0.1961, + "step": 65744 + }, + { + "epoch": 2.53, + "learning_rate": 7.8770197486535e-06, + "loss": 0.2307, + "step": 65745 + }, + { + "epoch": 2.53, + "learning_rate": 7.87637855860477e-06, + "loss": 0.2796, + "step": 65746 + }, + { + "epoch": 2.53, + "learning_rate": 7.87573736855604e-06, + "loss": 0.2271, + "step": 65747 + }, + { + "epoch": 2.53, + "learning_rate": 7.87509617850731e-06, + "loss": 0.1958, + "step": 65748 + }, + { + "epoch": 2.53, + "learning_rate": 7.874454988458579e-06, + "loss": 0.2476, + "step": 65749 + }, + { + "epoch": 2.53, + "learning_rate": 7.873813798409849e-06, + "loss": 0.3671, + "step": 65750 + }, + { + "epoch": 2.53, + "learning_rate": 7.873172608361119e-06, + "loss": 0.1995, + "step": 65751 + }, + { + "epoch": 2.53, + "learning_rate": 7.872531418312389e-06, + "loss": 0.2761, + "step": 65752 + }, + { + "epoch": 2.53, + "learning_rate": 7.871890228263657e-06, + "loss": 0.0835, + "step": 65753 + }, + { + "epoch": 2.53, + "learning_rate": 7.871249038214927e-06, + "loss": 0.2267, + "step": 65754 + }, + { + "epoch": 2.53, + "learning_rate": 7.870607848166196e-06, + "loss": 0.1244, + "step": 65755 + }, + { + "epoch": 2.53, + "learning_rate": 7.869966658117467e-06, + "loss": 0.3359, + "step": 65756 + }, + { + "epoch": 2.53, + "learning_rate": 7.869325468068736e-06, + "loss": 0.3688, + "step": 65757 + }, + { + "epoch": 2.53, + "learning_rate": 7.868684278020006e-06, + "loss": 0.192, + "step": 65758 + }, + { + "epoch": 2.53, + "learning_rate": 7.868043087971274e-06, + "loss": 0.226, + "step": 65759 + }, + { + "epoch": 2.53, + "learning_rate": 7.867401897922546e-06, + "loss": 0.236, + "step": 65760 + }, + { + "epoch": 2.53, + "learning_rate": 7.866760707873814e-06, + "loss": 0.2597, + "step": 65761 + }, + { + "epoch": 2.53, + "learning_rate": 7.866119517825084e-06, + "loss": 0.2437, + "step": 65762 + }, + { + "epoch": 2.53, + "learning_rate": 7.865478327776352e-06, + "loss": 0.2093, + "step": 65763 + }, + { + "epoch": 2.53, + "learning_rate": 7.864837137727622e-06, + "loss": 0.2568, + "step": 65764 + }, + { + "epoch": 2.53, + "learning_rate": 7.864195947678892e-06, + "loss": 0.2512, + "step": 65765 + }, + { + "epoch": 2.53, + "learning_rate": 7.863554757630163e-06, + "loss": 0.1105, + "step": 65766 + }, + { + "epoch": 2.53, + "learning_rate": 7.862913567581431e-06, + "loss": 0.2226, + "step": 65767 + }, + { + "epoch": 2.53, + "learning_rate": 7.862272377532701e-06, + "loss": 0.3377, + "step": 65768 + }, + { + "epoch": 2.53, + "learning_rate": 7.861631187483971e-06, + "loss": 0.2895, + "step": 65769 + }, + { + "epoch": 2.53, + "learning_rate": 7.860989997435241e-06, + "loss": 0.1802, + "step": 65770 + }, + { + "epoch": 2.53, + "learning_rate": 7.86034880738651e-06, + "loss": 0.2078, + "step": 65771 + }, + { + "epoch": 2.53, + "learning_rate": 7.85970761733778e-06, + "loss": 0.198, + "step": 65772 + }, + { + "epoch": 2.53, + "learning_rate": 7.859066427289048e-06, + "loss": 0.2186, + "step": 65773 + }, + { + "epoch": 2.53, + "learning_rate": 7.85842523724032e-06, + "loss": 0.2905, + "step": 65774 + }, + { + "epoch": 2.53, + "learning_rate": 7.857784047191588e-06, + "loss": 0.2175, + "step": 65775 + }, + { + "epoch": 2.53, + "learning_rate": 7.857142857142858e-06, + "loss": 0.1591, + "step": 65776 + }, + { + "epoch": 2.53, + "learning_rate": 7.856501667094126e-06, + "loss": 0.2076, + "step": 65777 + }, + { + "epoch": 2.53, + "learning_rate": 7.855860477045398e-06, + "loss": 0.22, + "step": 65778 + }, + { + "epoch": 2.53, + "learning_rate": 7.855219286996666e-06, + "loss": 0.1755, + "step": 65779 + }, + { + "epoch": 2.53, + "learning_rate": 7.854578096947936e-06, + "loss": 0.1092, + "step": 65780 + }, + { + "epoch": 2.53, + "learning_rate": 7.853936906899204e-06, + "loss": 0.345, + "step": 65781 + }, + { + "epoch": 2.53, + "learning_rate": 7.853295716850474e-06, + "loss": 0.2766, + "step": 65782 + }, + { + "epoch": 2.53, + "learning_rate": 7.852654526801744e-06, + "loss": 0.2446, + "step": 65783 + }, + { + "epoch": 2.53, + "learning_rate": 7.852013336753014e-06, + "loss": 0.2087, + "step": 65784 + }, + { + "epoch": 2.53, + "learning_rate": 7.851372146704283e-06, + "loss": 0.2015, + "step": 65785 + }, + { + "epoch": 2.53, + "learning_rate": 7.850730956655553e-06, + "loss": 0.1535, + "step": 65786 + }, + { + "epoch": 2.53, + "learning_rate": 7.850089766606823e-06, + "loss": 0.1869, + "step": 65787 + }, + { + "epoch": 2.53, + "learning_rate": 7.849448576558093e-06, + "loss": 0.152, + "step": 65788 + }, + { + "epoch": 2.53, + "learning_rate": 7.848807386509361e-06, + "loss": 0.2311, + "step": 65789 + }, + { + "epoch": 2.53, + "learning_rate": 7.848166196460631e-06, + "loss": 0.2074, + "step": 65790 + }, + { + "epoch": 2.53, + "learning_rate": 7.8475250064119e-06, + "loss": 0.2023, + "step": 65791 + }, + { + "epoch": 2.53, + "learning_rate": 7.846883816363171e-06, + "loss": 0.1596, + "step": 65792 + }, + { + "epoch": 2.53, + "learning_rate": 7.84624262631444e-06, + "loss": 0.2521, + "step": 65793 + }, + { + "epoch": 2.53, + "learning_rate": 7.84560143626571e-06, + "loss": 0.4352, + "step": 65794 + }, + { + "epoch": 2.53, + "learning_rate": 7.844960246216978e-06, + "loss": 0.3084, + "step": 65795 + }, + { + "epoch": 2.53, + "learning_rate": 7.84431905616825e-06, + "loss": 0.4222, + "step": 65796 + }, + { + "epoch": 2.53, + "learning_rate": 7.843677866119518e-06, + "loss": 0.1387, + "step": 65797 + }, + { + "epoch": 2.53, + "learning_rate": 7.843036676070788e-06, + "loss": 0.1947, + "step": 65798 + }, + { + "epoch": 2.53, + "learning_rate": 7.842395486022056e-06, + "loss": 0.2797, + "step": 65799 + }, + { + "epoch": 2.53, + "learning_rate": 7.841754295973328e-06, + "loss": 0.2039, + "step": 65800 + }, + { + "epoch": 2.53, + "learning_rate": 7.841113105924596e-06, + "loss": 0.2569, + "step": 65801 + }, + { + "epoch": 2.53, + "learning_rate": 7.840471915875866e-06, + "loss": 0.0969, + "step": 65802 + }, + { + "epoch": 2.53, + "learning_rate": 7.839830725827135e-06, + "loss": 0.1394, + "step": 65803 + }, + { + "epoch": 2.53, + "learning_rate": 7.839189535778405e-06, + "loss": 0.1682, + "step": 65804 + }, + { + "epoch": 2.53, + "learning_rate": 7.838548345729675e-06, + "loss": 0.1861, + "step": 65805 + }, + { + "epoch": 2.53, + "learning_rate": 7.837907155680945e-06, + "loss": 0.1583, + "step": 65806 + }, + { + "epoch": 2.53, + "learning_rate": 7.837265965632213e-06, + "loss": 0.2123, + "step": 65807 + }, + { + "epoch": 2.53, + "learning_rate": 7.836624775583483e-06, + "loss": 0.2598, + "step": 65808 + }, + { + "epoch": 2.53, + "learning_rate": 7.835983585534753e-06, + "loss": 0.1744, + "step": 65809 + }, + { + "epoch": 2.53, + "learning_rate": 7.835342395486023e-06, + "loss": 0.2804, + "step": 65810 + }, + { + "epoch": 2.53, + "learning_rate": 7.834701205437292e-06, + "loss": 0.2248, + "step": 65811 + }, + { + "epoch": 2.53, + "learning_rate": 7.834060015388562e-06, + "loss": 0.3645, + "step": 65812 + }, + { + "epoch": 2.53, + "learning_rate": 7.83341882533983e-06, + "loss": 0.1583, + "step": 65813 + }, + { + "epoch": 2.53, + "learning_rate": 7.832777635291102e-06, + "loss": 0.1844, + "step": 65814 + }, + { + "epoch": 2.53, + "learning_rate": 7.83213644524237e-06, + "loss": 0.2345, + "step": 65815 + }, + { + "epoch": 2.53, + "learning_rate": 7.83149525519364e-06, + "loss": 0.2326, + "step": 65816 + }, + { + "epoch": 2.53, + "learning_rate": 7.830854065144908e-06, + "loss": 0.1418, + "step": 65817 + }, + { + "epoch": 2.53, + "learning_rate": 7.83021287509618e-06, + "loss": 0.21, + "step": 65818 + }, + { + "epoch": 2.53, + "learning_rate": 7.829571685047448e-06, + "loss": 0.3164, + "step": 65819 + }, + { + "epoch": 2.53, + "learning_rate": 7.828930494998718e-06, + "loss": 0.2332, + "step": 65820 + }, + { + "epoch": 2.53, + "learning_rate": 7.828289304949987e-06, + "loss": 0.18, + "step": 65821 + }, + { + "epoch": 2.53, + "learning_rate": 7.827648114901257e-06, + "loss": 0.2193, + "step": 65822 + }, + { + "epoch": 2.53, + "learning_rate": 7.827006924852527e-06, + "loss": 0.1347, + "step": 65823 + }, + { + "epoch": 2.53, + "learning_rate": 7.826365734803797e-06, + "loss": 0.2736, + "step": 65824 + }, + { + "epoch": 2.53, + "learning_rate": 7.825724544755065e-06, + "loss": 0.1897, + "step": 65825 + }, + { + "epoch": 2.53, + "learning_rate": 7.825083354706335e-06, + "loss": 0.206, + "step": 65826 + }, + { + "epoch": 2.53, + "learning_rate": 7.824442164657605e-06, + "loss": 0.2851, + "step": 65827 + }, + { + "epoch": 2.53, + "learning_rate": 7.823800974608875e-06, + "loss": 0.3623, + "step": 65828 + }, + { + "epoch": 2.53, + "learning_rate": 7.823159784560143e-06, + "loss": 0.1882, + "step": 65829 + }, + { + "epoch": 2.53, + "learning_rate": 7.822518594511413e-06, + "loss": 0.1312, + "step": 65830 + }, + { + "epoch": 2.53, + "learning_rate": 7.821877404462682e-06, + "loss": 0.1936, + "step": 65831 + }, + { + "epoch": 2.53, + "learning_rate": 7.821236214413954e-06, + "loss": 0.213, + "step": 65832 + }, + { + "epoch": 2.53, + "learning_rate": 7.820595024365222e-06, + "loss": 0.1794, + "step": 65833 + }, + { + "epoch": 2.53, + "learning_rate": 7.819953834316492e-06, + "loss": 0.3662, + "step": 65834 + }, + { + "epoch": 2.53, + "learning_rate": 7.81931264426776e-06, + "loss": 0.1607, + "step": 65835 + }, + { + "epoch": 2.53, + "learning_rate": 7.818671454219032e-06, + "loss": 0.2173, + "step": 65836 + }, + { + "epoch": 2.53, + "learning_rate": 7.8180302641703e-06, + "loss": 0.1352, + "step": 65837 + }, + { + "epoch": 2.53, + "learning_rate": 7.81738907412157e-06, + "loss": 0.1998, + "step": 65838 + }, + { + "epoch": 2.53, + "learning_rate": 7.816747884072839e-06, + "loss": 0.1511, + "step": 65839 + }, + { + "epoch": 2.53, + "learning_rate": 7.816106694024109e-06, + "loss": 0.261, + "step": 65840 + }, + { + "epoch": 2.53, + "learning_rate": 7.815465503975379e-06, + "loss": 0.243, + "step": 65841 + }, + { + "epoch": 2.53, + "learning_rate": 7.814824313926649e-06, + "loss": 0.1863, + "step": 65842 + }, + { + "epoch": 2.53, + "learning_rate": 7.814183123877917e-06, + "loss": 0.2506, + "step": 65843 + }, + { + "epoch": 2.53, + "learning_rate": 7.813541933829187e-06, + "loss": 0.2915, + "step": 65844 + }, + { + "epoch": 2.53, + "learning_rate": 7.812900743780457e-06, + "loss": 0.2717, + "step": 65845 + }, + { + "epoch": 2.53, + "learning_rate": 7.812259553731727e-06, + "loss": 0.2902, + "step": 65846 + }, + { + "epoch": 2.53, + "learning_rate": 7.811618363682995e-06, + "loss": 0.3886, + "step": 65847 + }, + { + "epoch": 2.53, + "learning_rate": 7.810977173634265e-06, + "loss": 0.2561, + "step": 65848 + }, + { + "epoch": 2.53, + "learning_rate": 7.810335983585535e-06, + "loss": 0.2081, + "step": 65849 + }, + { + "epoch": 2.53, + "learning_rate": 7.809694793536805e-06, + "loss": 0.2965, + "step": 65850 + }, + { + "epoch": 2.53, + "learning_rate": 7.809053603488074e-06, + "loss": 0.331, + "step": 65851 + }, + { + "epoch": 2.53, + "learning_rate": 7.808412413439344e-06, + "loss": 0.1817, + "step": 65852 + }, + { + "epoch": 2.53, + "learning_rate": 7.807771223390612e-06, + "loss": 0.1381, + "step": 65853 + }, + { + "epoch": 2.53, + "learning_rate": 7.807130033341884e-06, + "loss": 0.361, + "step": 65854 + }, + { + "epoch": 2.53, + "learning_rate": 7.806488843293152e-06, + "loss": 0.1359, + "step": 65855 + }, + { + "epoch": 2.53, + "learning_rate": 7.805847653244422e-06, + "loss": 0.1461, + "step": 65856 + }, + { + "epoch": 2.53, + "learning_rate": 7.80520646319569e-06, + "loss": 0.2635, + "step": 65857 + }, + { + "epoch": 2.53, + "learning_rate": 7.804565273146962e-06, + "loss": 0.3373, + "step": 65858 + }, + { + "epoch": 2.53, + "learning_rate": 7.80392408309823e-06, + "loss": 0.2902, + "step": 65859 + }, + { + "epoch": 2.53, + "learning_rate": 7.8032828930495e-06, + "loss": 0.2496, + "step": 65860 + }, + { + "epoch": 2.53, + "learning_rate": 7.802641703000769e-06, + "loss": 0.1855, + "step": 65861 + }, + { + "epoch": 2.53, + "learning_rate": 7.802000512952039e-06, + "loss": 0.1034, + "step": 65862 + }, + { + "epoch": 2.53, + "learning_rate": 7.801359322903309e-06, + "loss": 0.185, + "step": 65863 + }, + { + "epoch": 2.53, + "learning_rate": 7.800718132854579e-06, + "loss": 0.2579, + "step": 65864 + }, + { + "epoch": 2.53, + "learning_rate": 7.800076942805847e-06, + "loss": 0.198, + "step": 65865 + }, + { + "epoch": 2.53, + "learning_rate": 7.799435752757117e-06, + "loss": 0.2198, + "step": 65866 + }, + { + "epoch": 2.53, + "learning_rate": 7.798794562708387e-06, + "loss": 0.1111, + "step": 65867 + }, + { + "epoch": 2.53, + "learning_rate": 7.798153372659657e-06, + "loss": 0.2082, + "step": 65868 + }, + { + "epoch": 2.53, + "learning_rate": 7.797512182610926e-06, + "loss": 0.1659, + "step": 65869 + }, + { + "epoch": 2.53, + "learning_rate": 7.796870992562196e-06, + "loss": 0.2129, + "step": 65870 + }, + { + "epoch": 2.53, + "learning_rate": 7.796229802513464e-06, + "loss": 0.3273, + "step": 65871 + }, + { + "epoch": 2.53, + "learning_rate": 7.795588612464736e-06, + "loss": 0.2263, + "step": 65872 + }, + { + "epoch": 2.53, + "learning_rate": 7.794947422416004e-06, + "loss": 0.2032, + "step": 65873 + }, + { + "epoch": 2.53, + "learning_rate": 7.794306232367274e-06, + "loss": 0.1268, + "step": 65874 + }, + { + "epoch": 2.53, + "learning_rate": 7.793665042318542e-06, + "loss": 0.2863, + "step": 65875 + }, + { + "epoch": 2.53, + "learning_rate": 7.793023852269814e-06, + "loss": 0.2384, + "step": 65876 + }, + { + "epoch": 2.53, + "learning_rate": 7.792382662221083e-06, + "loss": 0.158, + "step": 65877 + }, + { + "epoch": 2.53, + "learning_rate": 7.791741472172353e-06, + "loss": 0.0947, + "step": 65878 + }, + { + "epoch": 2.53, + "learning_rate": 7.791100282123621e-06, + "loss": 0.2299, + "step": 65879 + }, + { + "epoch": 2.53, + "learning_rate": 7.790459092074891e-06, + "loss": 0.1234, + "step": 65880 + }, + { + "epoch": 2.53, + "learning_rate": 7.789817902026161e-06, + "loss": 0.3913, + "step": 65881 + }, + { + "epoch": 2.53, + "learning_rate": 7.789176711977431e-06, + "loss": 0.264, + "step": 65882 + }, + { + "epoch": 2.53, + "learning_rate": 7.7885355219287e-06, + "loss": 0.3045, + "step": 65883 + }, + { + "epoch": 2.53, + "learning_rate": 7.78789433187997e-06, + "loss": 0.2294, + "step": 65884 + }, + { + "epoch": 2.53, + "learning_rate": 7.78725314183124e-06, + "loss": 0.4221, + "step": 65885 + }, + { + "epoch": 2.53, + "learning_rate": 7.78661195178251e-06, + "loss": 0.19, + "step": 65886 + }, + { + "epoch": 2.53, + "learning_rate": 7.785970761733778e-06, + "loss": 0.2179, + "step": 65887 + }, + { + "epoch": 2.53, + "learning_rate": 7.785329571685048e-06, + "loss": 0.2305, + "step": 65888 + }, + { + "epoch": 2.53, + "learning_rate": 7.784688381636316e-06, + "loss": 0.2597, + "step": 65889 + }, + { + "epoch": 2.53, + "learning_rate": 7.784047191587588e-06, + "loss": 0.2535, + "step": 65890 + }, + { + "epoch": 2.53, + "learning_rate": 7.783406001538856e-06, + "loss": 0.172, + "step": 65891 + }, + { + "epoch": 2.53, + "learning_rate": 7.782764811490126e-06, + "loss": 0.1462, + "step": 65892 + }, + { + "epoch": 2.53, + "learning_rate": 7.782123621441394e-06, + "loss": 0.2765, + "step": 65893 + }, + { + "epoch": 2.53, + "learning_rate": 7.781482431392666e-06, + "loss": 0.2175, + "step": 65894 + }, + { + "epoch": 2.53, + "learning_rate": 7.780841241343934e-06, + "loss": 0.1337, + "step": 65895 + }, + { + "epoch": 2.53, + "learning_rate": 7.780200051295204e-06, + "loss": 0.1447, + "step": 65896 + }, + { + "epoch": 2.53, + "learning_rate": 7.779558861246473e-06, + "loss": 0.2717, + "step": 65897 + }, + { + "epoch": 2.53, + "learning_rate": 7.778917671197745e-06, + "loss": 0.2426, + "step": 65898 + }, + { + "epoch": 2.53, + "learning_rate": 7.778276481149013e-06, + "loss": 0.1591, + "step": 65899 + }, + { + "epoch": 2.53, + "learning_rate": 7.777635291100283e-06, + "loss": 0.2612, + "step": 65900 + }, + { + "epoch": 2.53, + "learning_rate": 7.776994101051551e-06, + "loss": 0.1712, + "step": 65901 + }, + { + "epoch": 2.53, + "learning_rate": 7.776352911002821e-06, + "loss": 0.1272, + "step": 65902 + }, + { + "epoch": 2.53, + "learning_rate": 7.775711720954091e-06, + "loss": 0.1932, + "step": 65903 + }, + { + "epoch": 2.53, + "learning_rate": 7.775070530905361e-06, + "loss": 0.2582, + "step": 65904 + }, + { + "epoch": 2.53, + "learning_rate": 7.77442934085663e-06, + "loss": 0.1782, + "step": 65905 + }, + { + "epoch": 2.53, + "learning_rate": 7.7737881508079e-06, + "loss": 0.2104, + "step": 65906 + }, + { + "epoch": 2.53, + "learning_rate": 7.77314696075917e-06, + "loss": 0.1615, + "step": 65907 + }, + { + "epoch": 2.53, + "learning_rate": 7.77250577071044e-06, + "loss": 0.3506, + "step": 65908 + }, + { + "epoch": 2.53, + "learning_rate": 7.771864580661708e-06, + "loss": 0.253, + "step": 65909 + }, + { + "epoch": 2.53, + "learning_rate": 7.771223390612978e-06, + "loss": 0.1621, + "step": 65910 + }, + { + "epoch": 2.53, + "learning_rate": 7.770582200564246e-06, + "loss": 0.1205, + "step": 65911 + }, + { + "epoch": 2.53, + "learning_rate": 7.769941010515518e-06, + "loss": 0.2597, + "step": 65912 + }, + { + "epoch": 2.53, + "learning_rate": 7.769299820466786e-06, + "loss": 0.2199, + "step": 65913 + }, + { + "epoch": 2.53, + "learning_rate": 7.768658630418056e-06, + "loss": 0.3331, + "step": 65914 + }, + { + "epoch": 2.53, + "learning_rate": 7.768017440369325e-06, + "loss": 0.1829, + "step": 65915 + }, + { + "epoch": 2.53, + "learning_rate": 7.767376250320596e-06, + "loss": 0.363, + "step": 65916 + }, + { + "epoch": 2.53, + "learning_rate": 7.766735060271865e-06, + "loss": 0.157, + "step": 65917 + }, + { + "epoch": 2.53, + "learning_rate": 7.766093870223135e-06, + "loss": 0.2599, + "step": 65918 + }, + { + "epoch": 2.53, + "learning_rate": 7.765452680174403e-06, + "loss": 0.167, + "step": 65919 + }, + { + "epoch": 2.53, + "learning_rate": 7.764811490125673e-06, + "loss": 0.2722, + "step": 65920 + }, + { + "epoch": 2.53, + "learning_rate": 7.764170300076943e-06, + "loss": 0.2927, + "step": 65921 + }, + { + "epoch": 2.53, + "learning_rate": 7.763529110028213e-06, + "loss": 0.1489, + "step": 65922 + }, + { + "epoch": 2.53, + "learning_rate": 7.762887919979482e-06, + "loss": 0.1952, + "step": 65923 + }, + { + "epoch": 2.53, + "learning_rate": 7.762246729930752e-06, + "loss": 0.3008, + "step": 65924 + }, + { + "epoch": 2.53, + "learning_rate": 7.761605539882022e-06, + "loss": 0.3088, + "step": 65925 + }, + { + "epoch": 2.53, + "learning_rate": 7.760964349833292e-06, + "loss": 0.2341, + "step": 65926 + }, + { + "epoch": 2.53, + "learning_rate": 7.76032315978456e-06, + "loss": 0.3075, + "step": 65927 + }, + { + "epoch": 2.53, + "learning_rate": 7.75968196973583e-06, + "loss": 0.1144, + "step": 65928 + }, + { + "epoch": 2.53, + "learning_rate": 7.759040779687098e-06, + "loss": 0.3193, + "step": 65929 + }, + { + "epoch": 2.53, + "learning_rate": 7.75839958963837e-06, + "loss": 0.206, + "step": 65930 + }, + { + "epoch": 2.53, + "learning_rate": 7.757758399589638e-06, + "loss": 0.1681, + "step": 65931 + }, + { + "epoch": 2.53, + "learning_rate": 7.757117209540908e-06, + "loss": 0.1091, + "step": 65932 + }, + { + "epoch": 2.53, + "learning_rate": 7.756476019492177e-06, + "loss": 0.1128, + "step": 65933 + }, + { + "epoch": 2.53, + "learning_rate": 7.755834829443448e-06, + "loss": 0.1523, + "step": 65934 + }, + { + "epoch": 2.53, + "learning_rate": 7.755193639394717e-06, + "loss": 0.1695, + "step": 65935 + }, + { + "epoch": 2.54, + "learning_rate": 7.754552449345987e-06, + "loss": 0.3341, + "step": 65936 + }, + { + "epoch": 2.54, + "learning_rate": 7.753911259297255e-06, + "loss": 0.3187, + "step": 65937 + }, + { + "epoch": 2.54, + "learning_rate": 7.753270069248525e-06, + "loss": 0.145, + "step": 65938 + }, + { + "epoch": 2.54, + "learning_rate": 7.752628879199795e-06, + "loss": 0.2488, + "step": 65939 + }, + { + "epoch": 2.54, + "learning_rate": 7.751987689151065e-06, + "loss": 0.2028, + "step": 65940 + }, + { + "epoch": 2.54, + "learning_rate": 7.751346499102333e-06, + "loss": 0.3041, + "step": 65941 + }, + { + "epoch": 2.54, + "learning_rate": 7.750705309053603e-06, + "loss": 0.1919, + "step": 65942 + }, + { + "epoch": 2.54, + "learning_rate": 7.750064119004874e-06, + "loss": 0.2534, + "step": 65943 + }, + { + "epoch": 2.54, + "learning_rate": 7.749422928956144e-06, + "loss": 0.3182, + "step": 65944 + }, + { + "epoch": 2.54, + "learning_rate": 7.748781738907412e-06, + "loss": 0.1471, + "step": 65945 + }, + { + "epoch": 2.54, + "learning_rate": 7.748140548858682e-06, + "loss": 0.2254, + "step": 65946 + }, + { + "epoch": 2.54, + "learning_rate": 7.747499358809952e-06, + "loss": 0.1842, + "step": 65947 + }, + { + "epoch": 2.54, + "learning_rate": 7.746858168761222e-06, + "loss": 0.1716, + "step": 65948 + }, + { + "epoch": 2.54, + "learning_rate": 7.74621697871249e-06, + "loss": 0.2224, + "step": 65949 + }, + { + "epoch": 2.54, + "learning_rate": 7.74557578866376e-06, + "loss": 0.1699, + "step": 65950 + }, + { + "epoch": 2.54, + "learning_rate": 7.744934598615029e-06, + "loss": 0.2931, + "step": 65951 + }, + { + "epoch": 2.54, + "learning_rate": 7.7442934085663e-06, + "loss": 0.1293, + "step": 65952 + }, + { + "epoch": 2.54, + "learning_rate": 7.743652218517569e-06, + "loss": 0.1507, + "step": 65953 + }, + { + "epoch": 2.54, + "learning_rate": 7.743011028468839e-06, + "loss": 0.1613, + "step": 65954 + }, + { + "epoch": 2.54, + "learning_rate": 7.742369838420107e-06, + "loss": 0.2408, + "step": 65955 + }, + { + "epoch": 2.54, + "learning_rate": 7.741728648371379e-06, + "loss": 0.2561, + "step": 65956 + }, + { + "epoch": 2.54, + "learning_rate": 7.741087458322647e-06, + "loss": 0.3269, + "step": 65957 + }, + { + "epoch": 2.54, + "learning_rate": 7.740446268273917e-06, + "loss": 0.343, + "step": 65958 + }, + { + "epoch": 2.54, + "learning_rate": 7.739805078225185e-06, + "loss": 0.2942, + "step": 65959 + }, + { + "epoch": 2.54, + "learning_rate": 7.739163888176455e-06, + "loss": 0.2905, + "step": 65960 + }, + { + "epoch": 2.54, + "learning_rate": 7.738522698127725e-06, + "loss": 0.1559, + "step": 65961 + }, + { + "epoch": 2.54, + "learning_rate": 7.737881508078995e-06, + "loss": 0.2189, + "step": 65962 + }, + { + "epoch": 2.54, + "learning_rate": 7.737240318030264e-06, + "loss": 0.3281, + "step": 65963 + }, + { + "epoch": 2.54, + "learning_rate": 7.736599127981534e-06, + "loss": 0.3914, + "step": 65964 + }, + { + "epoch": 2.54, + "learning_rate": 7.735957937932804e-06, + "loss": 0.2587, + "step": 65965 + }, + { + "epoch": 2.54, + "learning_rate": 7.735316747884074e-06, + "loss": 0.2716, + "step": 65966 + }, + { + "epoch": 2.54, + "learning_rate": 7.734675557835342e-06, + "loss": 0.3071, + "step": 65967 + }, + { + "epoch": 2.54, + "learning_rate": 7.734034367786612e-06, + "loss": 0.1017, + "step": 65968 + }, + { + "epoch": 2.54, + "learning_rate": 7.73339317773788e-06, + "loss": 0.2013, + "step": 65969 + }, + { + "epoch": 2.54, + "learning_rate": 7.732751987689152e-06, + "loss": 0.1467, + "step": 65970 + }, + { + "epoch": 2.54, + "learning_rate": 7.73211079764042e-06, + "loss": 0.2362, + "step": 65971 + }, + { + "epoch": 2.54, + "learning_rate": 7.73146960759169e-06, + "loss": 0.1809, + "step": 65972 + }, + { + "epoch": 2.54, + "learning_rate": 7.730828417542959e-06, + "loss": 0.1811, + "step": 65973 + }, + { + "epoch": 2.54, + "learning_rate": 7.73018722749423e-06, + "loss": 0.2912, + "step": 65974 + }, + { + "epoch": 2.54, + "learning_rate": 7.729546037445499e-06, + "loss": 0.2285, + "step": 65975 + }, + { + "epoch": 2.54, + "learning_rate": 7.728904847396769e-06, + "loss": 0.1974, + "step": 65976 + }, + { + "epoch": 2.54, + "learning_rate": 7.728263657348037e-06, + "loss": 0.1582, + "step": 65977 + }, + { + "epoch": 2.54, + "learning_rate": 7.727622467299307e-06, + "loss": 0.2056, + "step": 65978 + }, + { + "epoch": 2.54, + "learning_rate": 7.726981277250577e-06, + "loss": 0.2332, + "step": 65979 + }, + { + "epoch": 2.54, + "learning_rate": 7.726340087201847e-06, + "loss": 0.3422, + "step": 65980 + }, + { + "epoch": 2.54, + "learning_rate": 7.725698897153116e-06, + "loss": 0.1645, + "step": 65981 + }, + { + "epoch": 2.54, + "learning_rate": 7.725057707104386e-06, + "loss": 0.3133, + "step": 65982 + }, + { + "epoch": 2.54, + "learning_rate": 7.724416517055656e-06, + "loss": 0.3007, + "step": 65983 + }, + { + "epoch": 2.54, + "learning_rate": 7.723775327006926e-06, + "loss": 0.2291, + "step": 65984 + }, + { + "epoch": 2.54, + "learning_rate": 7.723134136958194e-06, + "loss": 0.1473, + "step": 65985 + }, + { + "epoch": 2.54, + "learning_rate": 7.722492946909464e-06, + "loss": 0.2816, + "step": 65986 + }, + { + "epoch": 2.54, + "learning_rate": 7.721851756860732e-06, + "loss": 0.1638, + "step": 65987 + }, + { + "epoch": 2.54, + "learning_rate": 7.721210566812004e-06, + "loss": 0.1902, + "step": 65988 + }, + { + "epoch": 2.54, + "learning_rate": 7.720569376763273e-06, + "loss": 0.2309, + "step": 65989 + }, + { + "epoch": 2.54, + "learning_rate": 7.719928186714543e-06, + "loss": 0.1473, + "step": 65990 + }, + { + "epoch": 2.54, + "learning_rate": 7.719286996665811e-06, + "loss": 0.2536, + "step": 65991 + }, + { + "epoch": 2.54, + "learning_rate": 7.718645806617083e-06, + "loss": 0.2753, + "step": 65992 + }, + { + "epoch": 2.54, + "learning_rate": 7.718004616568351e-06, + "loss": 0.1857, + "step": 65993 + }, + { + "epoch": 2.54, + "learning_rate": 7.717363426519621e-06, + "loss": 0.181, + "step": 65994 + }, + { + "epoch": 2.54, + "learning_rate": 7.71672223647089e-06, + "loss": 0.1857, + "step": 65995 + }, + { + "epoch": 2.54, + "learning_rate": 7.716081046422161e-06, + "loss": 0.3967, + "step": 65996 + }, + { + "epoch": 2.54, + "learning_rate": 7.71543985637343e-06, + "loss": 0.2219, + "step": 65997 + }, + { + "epoch": 2.54, + "learning_rate": 7.7147986663247e-06, + "loss": 0.1015, + "step": 65998 + }, + { + "epoch": 2.54, + "learning_rate": 7.714157476275968e-06, + "loss": 0.104, + "step": 65999 + }, + { + "epoch": 2.54, + "learning_rate": 7.713516286227238e-06, + "loss": 0.2589, + "step": 66000 + }, + { + "epoch": 2.54, + "learning_rate": 7.712875096178508e-06, + "loss": 0.2215, + "step": 66001 + }, + { + "epoch": 2.54, + "learning_rate": 7.712233906129778e-06, + "loss": 0.1998, + "step": 66002 + }, + { + "epoch": 2.54, + "learning_rate": 7.711592716081046e-06, + "loss": 0.2755, + "step": 66003 + }, + { + "epoch": 2.54, + "learning_rate": 7.710951526032316e-06, + "loss": 0.2477, + "step": 66004 + }, + { + "epoch": 2.54, + "learning_rate": 7.710310335983586e-06, + "loss": 0.187, + "step": 66005 + }, + { + "epoch": 2.54, + "learning_rate": 7.709669145934856e-06, + "loss": 0.1, + "step": 66006 + }, + { + "epoch": 2.54, + "learning_rate": 7.709027955886124e-06, + "loss": 0.2211, + "step": 66007 + }, + { + "epoch": 2.54, + "learning_rate": 7.708386765837394e-06, + "loss": 0.2937, + "step": 66008 + }, + { + "epoch": 2.54, + "learning_rate": 7.707745575788663e-06, + "loss": 0.2528, + "step": 66009 + }, + { + "epoch": 2.54, + "learning_rate": 7.707104385739935e-06, + "loss": 0.3029, + "step": 66010 + }, + { + "epoch": 2.54, + "learning_rate": 7.706463195691203e-06, + "loss": 0.2494, + "step": 66011 + }, + { + "epoch": 2.54, + "learning_rate": 7.705822005642473e-06, + "loss": 0.2588, + "step": 66012 + }, + { + "epoch": 2.54, + "learning_rate": 7.705180815593741e-06, + "loss": 0.3506, + "step": 66013 + }, + { + "epoch": 2.54, + "learning_rate": 7.704539625545013e-06, + "loss": 0.0962, + "step": 66014 + }, + { + "epoch": 2.54, + "learning_rate": 7.703898435496281e-06, + "loss": 0.1874, + "step": 66015 + }, + { + "epoch": 2.54, + "learning_rate": 7.703257245447551e-06, + "loss": 0.2422, + "step": 66016 + }, + { + "epoch": 2.54, + "learning_rate": 7.70261605539882e-06, + "loss": 0.2558, + "step": 66017 + }, + { + "epoch": 2.54, + "learning_rate": 7.70197486535009e-06, + "loss": 0.2219, + "step": 66018 + }, + { + "epoch": 2.54, + "learning_rate": 7.70133367530136e-06, + "loss": 0.2797, + "step": 66019 + }, + { + "epoch": 2.54, + "learning_rate": 7.70069248525263e-06, + "loss": 0.1834, + "step": 66020 + }, + { + "epoch": 2.54, + "learning_rate": 7.700051295203898e-06, + "loss": 0.28, + "step": 66021 + }, + { + "epoch": 2.54, + "learning_rate": 7.699410105155168e-06, + "loss": 0.3026, + "step": 66022 + }, + { + "epoch": 2.54, + "learning_rate": 7.698768915106438e-06, + "loss": 0.2694, + "step": 66023 + }, + { + "epoch": 2.54, + "learning_rate": 7.698127725057708e-06, + "loss": 0.192, + "step": 66024 + }, + { + "epoch": 2.54, + "learning_rate": 7.697486535008976e-06, + "loss": 0.2887, + "step": 66025 + }, + { + "epoch": 2.54, + "learning_rate": 7.696845344960246e-06, + "loss": 0.1348, + "step": 66026 + }, + { + "epoch": 2.54, + "learning_rate": 7.696204154911515e-06, + "loss": 0.1705, + "step": 66027 + }, + { + "epoch": 2.54, + "learning_rate": 7.695562964862786e-06, + "loss": 0.1933, + "step": 66028 + }, + { + "epoch": 2.54, + "learning_rate": 7.694921774814055e-06, + "loss": 0.3042, + "step": 66029 + }, + { + "epoch": 2.54, + "learning_rate": 7.694280584765325e-06, + "loss": 0.2023, + "step": 66030 + }, + { + "epoch": 2.54, + "learning_rate": 7.693639394716593e-06, + "loss": 0.2162, + "step": 66031 + }, + { + "epoch": 2.54, + "learning_rate": 7.692998204667865e-06, + "loss": 0.2163, + "step": 66032 + }, + { + "epoch": 2.54, + "learning_rate": 7.692357014619133e-06, + "loss": 0.1919, + "step": 66033 + }, + { + "epoch": 2.54, + "learning_rate": 7.691715824570403e-06, + "loss": 0.1913, + "step": 66034 + }, + { + "epoch": 2.54, + "learning_rate": 7.691074634521672e-06, + "loss": 0.1667, + "step": 66035 + }, + { + "epoch": 2.54, + "learning_rate": 7.690433444472943e-06, + "loss": 0.2509, + "step": 66036 + }, + { + "epoch": 2.54, + "learning_rate": 7.689792254424212e-06, + "loss": 0.1779, + "step": 66037 + }, + { + "epoch": 2.54, + "learning_rate": 7.689151064375482e-06, + "loss": 0.1786, + "step": 66038 + }, + { + "epoch": 2.54, + "learning_rate": 7.68850987432675e-06, + "loss": 0.189, + "step": 66039 + }, + { + "epoch": 2.54, + "learning_rate": 7.68786868427802e-06, + "loss": 0.2526, + "step": 66040 + }, + { + "epoch": 2.54, + "learning_rate": 7.68722749422929e-06, + "loss": 0.2171, + "step": 66041 + }, + { + "epoch": 2.54, + "learning_rate": 7.68658630418056e-06, + "loss": 0.2162, + "step": 66042 + }, + { + "epoch": 2.54, + "learning_rate": 7.685945114131828e-06, + "loss": 0.1909, + "step": 66043 + }, + { + "epoch": 2.54, + "learning_rate": 7.685303924083098e-06, + "loss": 0.1797, + "step": 66044 + }, + { + "epoch": 2.54, + "learning_rate": 7.684662734034368e-06, + "loss": 0.1914, + "step": 66045 + }, + { + "epoch": 2.54, + "learning_rate": 7.684021543985638e-06, + "loss": 0.1259, + "step": 66046 + }, + { + "epoch": 2.54, + "learning_rate": 7.683380353936907e-06, + "loss": 0.1523, + "step": 66047 + }, + { + "epoch": 2.54, + "learning_rate": 7.682739163888177e-06, + "loss": 0.2439, + "step": 66048 + }, + { + "epoch": 2.54, + "learning_rate": 7.682097973839445e-06, + "loss": 0.2178, + "step": 66049 + }, + { + "epoch": 2.54, + "learning_rate": 7.681456783790717e-06, + "loss": 0.1749, + "step": 66050 + }, + { + "epoch": 2.54, + "learning_rate": 7.680815593741985e-06, + "loss": 0.2259, + "step": 66051 + }, + { + "epoch": 2.54, + "learning_rate": 7.680174403693255e-06, + "loss": 0.1629, + "step": 66052 + }, + { + "epoch": 2.54, + "learning_rate": 7.679533213644523e-06, + "loss": 0.1387, + "step": 66053 + }, + { + "epoch": 2.54, + "learning_rate": 7.678892023595795e-06, + "loss": 0.2883, + "step": 66054 + }, + { + "epoch": 2.54, + "learning_rate": 7.678250833547064e-06, + "loss": 0.3084, + "step": 66055 + }, + { + "epoch": 2.54, + "learning_rate": 7.677609643498334e-06, + "loss": 0.2438, + "step": 66056 + }, + { + "epoch": 2.54, + "learning_rate": 7.676968453449602e-06, + "loss": 0.3194, + "step": 66057 + }, + { + "epoch": 2.54, + "learning_rate": 7.676327263400872e-06, + "loss": 0.1693, + "step": 66058 + }, + { + "epoch": 2.54, + "learning_rate": 7.675686073352142e-06, + "loss": 0.2125, + "step": 66059 + }, + { + "epoch": 2.54, + "learning_rate": 7.675044883303412e-06, + "loss": 0.2531, + "step": 66060 + }, + { + "epoch": 2.54, + "learning_rate": 7.67440369325468e-06, + "loss": 0.1745, + "step": 66061 + }, + { + "epoch": 2.54, + "learning_rate": 7.67376250320595e-06, + "loss": 0.156, + "step": 66062 + }, + { + "epoch": 2.54, + "learning_rate": 7.67312131315722e-06, + "loss": 0.2408, + "step": 66063 + }, + { + "epoch": 2.54, + "learning_rate": 7.67248012310849e-06, + "loss": 0.1278, + "step": 66064 + }, + { + "epoch": 2.54, + "learning_rate": 7.671838933059759e-06, + "loss": 0.1949, + "step": 66065 + }, + { + "epoch": 2.54, + "learning_rate": 7.671197743011029e-06, + "loss": 0.1269, + "step": 66066 + }, + { + "epoch": 2.54, + "learning_rate": 7.670556552962297e-06, + "loss": 0.3196, + "step": 66067 + }, + { + "epoch": 2.54, + "learning_rate": 7.669915362913569e-06, + "loss": 0.1895, + "step": 66068 + }, + { + "epoch": 2.54, + "learning_rate": 7.669274172864837e-06, + "loss": 0.3314, + "step": 66069 + }, + { + "epoch": 2.54, + "learning_rate": 7.668632982816107e-06, + "loss": 0.3471, + "step": 66070 + }, + { + "epoch": 2.54, + "learning_rate": 7.667991792767375e-06, + "loss": 0.2324, + "step": 66071 + }, + { + "epoch": 2.54, + "learning_rate": 7.667350602718647e-06, + "loss": 0.2419, + "step": 66072 + }, + { + "epoch": 2.54, + "learning_rate": 7.666709412669915e-06, + "loss": 0.1223, + "step": 66073 + }, + { + "epoch": 2.54, + "learning_rate": 7.666068222621185e-06, + "loss": 0.3092, + "step": 66074 + }, + { + "epoch": 2.54, + "learning_rate": 7.665427032572454e-06, + "loss": 0.1545, + "step": 66075 + }, + { + "epoch": 2.54, + "learning_rate": 7.664785842523724e-06, + "loss": 0.2431, + "step": 66076 + }, + { + "epoch": 2.54, + "learning_rate": 7.664144652474994e-06, + "loss": 0.23, + "step": 66077 + }, + { + "epoch": 2.54, + "learning_rate": 7.663503462426264e-06, + "loss": 0.1562, + "step": 66078 + }, + { + "epoch": 2.54, + "learning_rate": 7.662862272377532e-06, + "loss": 0.1563, + "step": 66079 + }, + { + "epoch": 2.54, + "learning_rate": 7.662221082328802e-06, + "loss": 0.2524, + "step": 66080 + }, + { + "epoch": 2.54, + "learning_rate": 7.661579892280072e-06, + "loss": 0.1776, + "step": 66081 + }, + { + "epoch": 2.54, + "learning_rate": 7.660938702231342e-06, + "loss": 0.2877, + "step": 66082 + }, + { + "epoch": 2.54, + "learning_rate": 7.66029751218261e-06, + "loss": 0.1953, + "step": 66083 + }, + { + "epoch": 2.54, + "learning_rate": 7.65965632213388e-06, + "loss": 0.1849, + "step": 66084 + }, + { + "epoch": 2.54, + "learning_rate": 7.65901513208515e-06, + "loss": 0.1417, + "step": 66085 + }, + { + "epoch": 2.54, + "learning_rate": 7.65837394203642e-06, + "loss": 0.1693, + "step": 66086 + }, + { + "epoch": 2.54, + "learning_rate": 7.657732751987689e-06, + "loss": 0.3662, + "step": 66087 + }, + { + "epoch": 2.54, + "learning_rate": 7.657091561938959e-06, + "loss": 0.1332, + "step": 66088 + }, + { + "epoch": 2.54, + "learning_rate": 7.656450371890227e-06, + "loss": 0.2599, + "step": 66089 + }, + { + "epoch": 2.54, + "learning_rate": 7.655809181841499e-06, + "loss": 0.2577, + "step": 66090 + }, + { + "epoch": 2.54, + "learning_rate": 7.655167991792767e-06, + "loss": 0.2111, + "step": 66091 + }, + { + "epoch": 2.54, + "learning_rate": 7.654526801744037e-06, + "loss": 0.2485, + "step": 66092 + }, + { + "epoch": 2.54, + "learning_rate": 7.653885611695306e-06, + "loss": 0.2327, + "step": 66093 + }, + { + "epoch": 2.54, + "learning_rate": 7.653244421646577e-06, + "loss": 0.1992, + "step": 66094 + }, + { + "epoch": 2.54, + "learning_rate": 7.652603231597846e-06, + "loss": 0.317, + "step": 66095 + }, + { + "epoch": 2.54, + "learning_rate": 7.651962041549116e-06, + "loss": 0.175, + "step": 66096 + }, + { + "epoch": 2.54, + "learning_rate": 7.651320851500384e-06, + "loss": 0.3413, + "step": 66097 + }, + { + "epoch": 2.54, + "learning_rate": 7.650679661451654e-06, + "loss": 0.145, + "step": 66098 + }, + { + "epoch": 2.54, + "learning_rate": 7.650038471402924e-06, + "loss": 0.2192, + "step": 66099 + }, + { + "epoch": 2.54, + "learning_rate": 7.649397281354194e-06, + "loss": 0.2608, + "step": 66100 + }, + { + "epoch": 2.54, + "learning_rate": 7.648756091305463e-06, + "loss": 0.1481, + "step": 66101 + }, + { + "epoch": 2.54, + "learning_rate": 7.648114901256733e-06, + "loss": 0.3906, + "step": 66102 + }, + { + "epoch": 2.54, + "learning_rate": 7.647473711208003e-06, + "loss": 0.2194, + "step": 66103 + }, + { + "epoch": 2.54, + "learning_rate": 7.646832521159273e-06, + "loss": 0.3039, + "step": 66104 + }, + { + "epoch": 2.54, + "learning_rate": 7.646191331110541e-06, + "loss": 0.2245, + "step": 66105 + }, + { + "epoch": 2.54, + "learning_rate": 7.645550141061811e-06, + "loss": 0.2082, + "step": 66106 + }, + { + "epoch": 2.54, + "learning_rate": 7.64490895101308e-06, + "loss": 0.242, + "step": 66107 + }, + { + "epoch": 2.54, + "learning_rate": 7.644267760964351e-06, + "loss": 0.227, + "step": 66108 + }, + { + "epoch": 2.54, + "learning_rate": 7.64362657091562e-06, + "loss": 0.172, + "step": 66109 + }, + { + "epoch": 2.54, + "learning_rate": 7.64298538086689e-06, + "loss": 0.2407, + "step": 66110 + }, + { + "epoch": 2.54, + "learning_rate": 7.642344190818158e-06, + "loss": 0.2361, + "step": 66111 + }, + { + "epoch": 2.54, + "learning_rate": 7.64170300076943e-06, + "loss": 0.2486, + "step": 66112 + }, + { + "epoch": 2.54, + "learning_rate": 7.641061810720698e-06, + "loss": 0.14, + "step": 66113 + }, + { + "epoch": 2.54, + "learning_rate": 7.640420620671968e-06, + "loss": 0.1571, + "step": 66114 + }, + { + "epoch": 2.54, + "learning_rate": 7.639779430623236e-06, + "loss": 0.1752, + "step": 66115 + }, + { + "epoch": 2.54, + "learning_rate": 7.639138240574506e-06, + "loss": 0.3063, + "step": 66116 + }, + { + "epoch": 2.54, + "learning_rate": 7.638497050525776e-06, + "loss": 0.2677, + "step": 66117 + }, + { + "epoch": 2.54, + "learning_rate": 7.637855860477046e-06, + "loss": 0.1895, + "step": 66118 + }, + { + "epoch": 2.54, + "learning_rate": 7.637214670428314e-06, + "loss": 0.131, + "step": 66119 + }, + { + "epoch": 2.54, + "learning_rate": 7.636573480379584e-06, + "loss": 0.1652, + "step": 66120 + }, + { + "epoch": 2.54, + "learning_rate": 7.635932290330855e-06, + "loss": 0.1085, + "step": 66121 + }, + { + "epoch": 2.54, + "learning_rate": 7.635291100282125e-06, + "loss": 0.2409, + "step": 66122 + }, + { + "epoch": 2.54, + "learning_rate": 7.634649910233393e-06, + "loss": 0.2162, + "step": 66123 + }, + { + "epoch": 2.54, + "learning_rate": 7.634008720184663e-06, + "loss": 0.119, + "step": 66124 + }, + { + "epoch": 2.54, + "learning_rate": 7.633367530135931e-06, + "loss": 0.1319, + "step": 66125 + }, + { + "epoch": 2.54, + "learning_rate": 7.632726340087203e-06, + "loss": 0.249, + "step": 66126 + }, + { + "epoch": 2.54, + "learning_rate": 7.632085150038471e-06, + "loss": 0.2681, + "step": 66127 + }, + { + "epoch": 2.54, + "learning_rate": 7.631443959989741e-06, + "loss": 0.2062, + "step": 66128 + }, + { + "epoch": 2.54, + "learning_rate": 7.63080276994101e-06, + "loss": 0.1766, + "step": 66129 + }, + { + "epoch": 2.54, + "learning_rate": 7.630161579892281e-06, + "loss": 0.2708, + "step": 66130 + }, + { + "epoch": 2.54, + "learning_rate": 7.62952038984355e-06, + "loss": 0.1006, + "step": 66131 + }, + { + "epoch": 2.54, + "learning_rate": 7.62887919979482e-06, + "loss": 0.1877, + "step": 66132 + }, + { + "epoch": 2.54, + "learning_rate": 7.628238009746089e-06, + "loss": 0.2134, + "step": 66133 + }, + { + "epoch": 2.54, + "learning_rate": 7.62759681969736e-06, + "loss": 0.1784, + "step": 66134 + }, + { + "epoch": 2.54, + "learning_rate": 7.626955629648629e-06, + "loss": 0.3459, + "step": 66135 + }, + { + "epoch": 2.54, + "learning_rate": 7.626314439599898e-06, + "loss": 0.1985, + "step": 66136 + }, + { + "epoch": 2.54, + "learning_rate": 7.625673249551167e-06, + "loss": 0.1765, + "step": 66137 + }, + { + "epoch": 2.54, + "learning_rate": 7.6250320595024364e-06, + "loss": 0.3089, + "step": 66138 + }, + { + "epoch": 2.54, + "learning_rate": 7.624390869453707e-06, + "loss": 0.1114, + "step": 66139 + }, + { + "epoch": 2.54, + "learning_rate": 7.6237496794049765e-06, + "loss": 0.242, + "step": 66140 + }, + { + "epoch": 2.54, + "learning_rate": 7.623108489356246e-06, + "loss": 0.1766, + "step": 66141 + }, + { + "epoch": 2.54, + "learning_rate": 7.622467299307515e-06, + "loss": 0.1937, + "step": 66142 + }, + { + "epoch": 2.54, + "learning_rate": 7.621826109258786e-06, + "loss": 0.3088, + "step": 66143 + }, + { + "epoch": 2.54, + "learning_rate": 7.621184919210055e-06, + "loss": 0.222, + "step": 66144 + }, + { + "epoch": 2.54, + "learning_rate": 7.620543729161324e-06, + "loss": 0.2983, + "step": 66145 + }, + { + "epoch": 2.54, + "learning_rate": 7.619902539112593e-06, + "loss": 0.1837, + "step": 66146 + }, + { + "epoch": 2.54, + "learning_rate": 7.619261349063862e-06, + "loss": 0.1829, + "step": 66147 + }, + { + "epoch": 2.54, + "learning_rate": 7.618620159015133e-06, + "loss": 0.1677, + "step": 66148 + }, + { + "epoch": 2.54, + "learning_rate": 7.6179789689664024e-06, + "loss": 0.2495, + "step": 66149 + }, + { + "epoch": 2.54, + "learning_rate": 7.617337778917672e-06, + "loss": 0.1997, + "step": 66150 + }, + { + "epoch": 2.54, + "learning_rate": 7.616696588868941e-06, + "loss": 0.2475, + "step": 66151 + }, + { + "epoch": 2.54, + "learning_rate": 7.616055398820212e-06, + "loss": 0.1712, + "step": 66152 + }, + { + "epoch": 2.54, + "learning_rate": 7.615414208771481e-06, + "loss": 0.3479, + "step": 66153 + }, + { + "epoch": 2.54, + "learning_rate": 7.61477301872275e-06, + "loss": 0.185, + "step": 66154 + }, + { + "epoch": 2.54, + "learning_rate": 7.614131828674019e-06, + "loss": 0.2205, + "step": 66155 + }, + { + "epoch": 2.54, + "learning_rate": 7.613490638625288e-06, + "loss": 0.1289, + "step": 66156 + }, + { + "epoch": 2.54, + "learning_rate": 7.612849448576559e-06, + "loss": 0.1569, + "step": 66157 + }, + { + "epoch": 2.54, + "learning_rate": 7.612208258527828e-06, + "loss": 0.308, + "step": 66158 + }, + { + "epoch": 2.54, + "learning_rate": 7.611567068479098e-06, + "loss": 0.2769, + "step": 66159 + }, + { + "epoch": 2.54, + "learning_rate": 7.610925878430367e-06, + "loss": 0.2067, + "step": 66160 + }, + { + "epoch": 2.54, + "learning_rate": 7.610284688381638e-06, + "loss": 0.2178, + "step": 66161 + }, + { + "epoch": 2.54, + "learning_rate": 7.609643498332907e-06, + "loss": 0.1441, + "step": 66162 + }, + { + "epoch": 2.54, + "learning_rate": 7.609002308284176e-06, + "loss": 0.1998, + "step": 66163 + }, + { + "epoch": 2.54, + "learning_rate": 7.608361118235445e-06, + "loss": 0.2312, + "step": 66164 + }, + { + "epoch": 2.54, + "learning_rate": 7.607719928186714e-06, + "loss": 0.1661, + "step": 66165 + }, + { + "epoch": 2.54, + "learning_rate": 7.607078738137985e-06, + "loss": 0.1728, + "step": 66166 + }, + { + "epoch": 2.54, + "learning_rate": 7.606437548089254e-06, + "loss": 0.218, + "step": 66167 + }, + { + "epoch": 2.54, + "learning_rate": 7.6057963580405235e-06, + "loss": 0.212, + "step": 66168 + }, + { + "epoch": 2.54, + "learning_rate": 7.605155167991793e-06, + "loss": 0.18, + "step": 66169 + }, + { + "epoch": 2.54, + "learning_rate": 7.604513977943064e-06, + "loss": 0.162, + "step": 66170 + }, + { + "epoch": 2.54, + "learning_rate": 7.603872787894333e-06, + "loss": 0.2763, + "step": 66171 + }, + { + "epoch": 2.54, + "learning_rate": 7.603231597845602e-06, + "loss": 0.2202, + "step": 66172 + }, + { + "epoch": 2.54, + "learning_rate": 7.602590407796871e-06, + "loss": 0.2116, + "step": 66173 + }, + { + "epoch": 2.54, + "learning_rate": 7.60194921774814e-06, + "loss": 0.208, + "step": 66174 + }, + { + "epoch": 2.54, + "learning_rate": 7.601308027699411e-06, + "loss": 0.1118, + "step": 66175 + }, + { + "epoch": 2.54, + "learning_rate": 7.60066683765068e-06, + "loss": 0.2435, + "step": 66176 + }, + { + "epoch": 2.54, + "learning_rate": 7.6000256476019495e-06, + "loss": 0.1983, + "step": 66177 + }, + { + "epoch": 2.54, + "learning_rate": 7.599384457553219e-06, + "loss": 0.1878, + "step": 66178 + }, + { + "epoch": 2.54, + "learning_rate": 7.5987432675044896e-06, + "loss": 0.1887, + "step": 66179 + }, + { + "epoch": 2.54, + "learning_rate": 7.598102077455759e-06, + "loss": 0.2829, + "step": 66180 + }, + { + "epoch": 2.54, + "learning_rate": 7.597460887407028e-06, + "loss": 0.1826, + "step": 66181 + }, + { + "epoch": 2.54, + "learning_rate": 7.596819697358297e-06, + "loss": 0.2037, + "step": 66182 + }, + { + "epoch": 2.54, + "learning_rate": 7.596178507309568e-06, + "loss": 0.2612, + "step": 66183 + }, + { + "epoch": 2.54, + "learning_rate": 7.595537317260837e-06, + "loss": 0.2081, + "step": 66184 + }, + { + "epoch": 2.54, + "learning_rate": 7.594896127212106e-06, + "loss": 0.2155, + "step": 66185 + }, + { + "epoch": 2.54, + "learning_rate": 7.5942549371633755e-06, + "loss": 0.2005, + "step": 66186 + }, + { + "epoch": 2.54, + "learning_rate": 7.593613747114645e-06, + "loss": 0.195, + "step": 66187 + }, + { + "epoch": 2.54, + "learning_rate": 7.5929725570659155e-06, + "loss": 0.2519, + "step": 66188 + }, + { + "epoch": 2.54, + "learning_rate": 7.592331367017185e-06, + "loss": 0.1807, + "step": 66189 + }, + { + "epoch": 2.54, + "learning_rate": 7.591690176968454e-06, + "loss": 0.1703, + "step": 66190 + }, + { + "epoch": 2.54, + "learning_rate": 7.591048986919723e-06, + "loss": 0.1522, + "step": 66191 + }, + { + "epoch": 2.54, + "learning_rate": 7.590407796870994e-06, + "loss": 0.1996, + "step": 66192 + }, + { + "epoch": 2.54, + "learning_rate": 7.589766606822263e-06, + "loss": 0.2132, + "step": 66193 + }, + { + "epoch": 2.54, + "learning_rate": 7.589125416773532e-06, + "loss": 0.3028, + "step": 66194 + }, + { + "epoch": 2.54, + "learning_rate": 7.5884842267248015e-06, + "loss": 0.2688, + "step": 66195 + }, + { + "epoch": 2.55, + "learning_rate": 7.587843036676071e-06, + "loss": 0.1544, + "step": 66196 + }, + { + "epoch": 2.55, + "learning_rate": 7.5872018466273415e-06, + "loss": 0.1458, + "step": 66197 + }, + { + "epoch": 2.55, + "learning_rate": 7.586560656578611e-06, + "loss": 0.1028, + "step": 66198 + }, + { + "epoch": 2.55, + "learning_rate": 7.58591946652988e-06, + "loss": 0.1786, + "step": 66199 + }, + { + "epoch": 2.55, + "learning_rate": 7.585278276481149e-06, + "loss": 0.1699, + "step": 66200 + }, + { + "epoch": 2.55, + "learning_rate": 7.58463708643242e-06, + "loss": 0.1824, + "step": 66201 + }, + { + "epoch": 2.55, + "learning_rate": 7.583995896383689e-06, + "loss": 0.2441, + "step": 66202 + }, + { + "epoch": 2.55, + "learning_rate": 7.583354706334958e-06, + "loss": 0.4089, + "step": 66203 + }, + { + "epoch": 2.55, + "learning_rate": 7.582713516286227e-06, + "loss": 0.1815, + "step": 66204 + }, + { + "epoch": 2.55, + "learning_rate": 7.582072326237497e-06, + "loss": 0.1609, + "step": 66205 + }, + { + "epoch": 2.55, + "learning_rate": 7.5814311361887675e-06, + "loss": 0.165, + "step": 66206 + }, + { + "epoch": 2.55, + "learning_rate": 7.580789946140037e-06, + "loss": 0.3646, + "step": 66207 + }, + { + "epoch": 2.55, + "learning_rate": 7.580148756091306e-06, + "loss": 0.1059, + "step": 66208 + }, + { + "epoch": 2.55, + "learning_rate": 7.579507566042575e-06, + "loss": 0.1213, + "step": 66209 + }, + { + "epoch": 2.55, + "learning_rate": 7.578866375993846e-06, + "loss": 0.191, + "step": 66210 + }, + { + "epoch": 2.55, + "learning_rate": 7.578225185945115e-06, + "loss": 0.2312, + "step": 66211 + }, + { + "epoch": 2.55, + "learning_rate": 7.577583995896384e-06, + "loss": 0.2145, + "step": 66212 + }, + { + "epoch": 2.55, + "learning_rate": 7.576942805847653e-06, + "loss": 0.1665, + "step": 66213 + }, + { + "epoch": 2.55, + "learning_rate": 7.5763016157989226e-06, + "loss": 0.2173, + "step": 66214 + }, + { + "epoch": 2.55, + "learning_rate": 7.5756604257501934e-06, + "loss": 0.2275, + "step": 66215 + }, + { + "epoch": 2.55, + "learning_rate": 7.575019235701463e-06, + "loss": 0.2973, + "step": 66216 + }, + { + "epoch": 2.55, + "learning_rate": 7.574378045652732e-06, + "loss": 0.1475, + "step": 66217 + }, + { + "epoch": 2.55, + "learning_rate": 7.573736855604001e-06, + "loss": 0.1458, + "step": 66218 + }, + { + "epoch": 2.55, + "learning_rate": 7.573095665555272e-06, + "loss": 0.2185, + "step": 66219 + }, + { + "epoch": 2.55, + "learning_rate": 7.572454475506541e-06, + "loss": 0.1167, + "step": 66220 + }, + { + "epoch": 2.55, + "learning_rate": 7.57181328545781e-06, + "loss": 0.2953, + "step": 66221 + }, + { + "epoch": 2.55, + "learning_rate": 7.571172095409079e-06, + "loss": 0.1917, + "step": 66222 + }, + { + "epoch": 2.55, + "learning_rate": 7.5705309053603485e-06, + "loss": 0.4166, + "step": 66223 + }, + { + "epoch": 2.55, + "learning_rate": 7.569889715311619e-06, + "loss": 0.1414, + "step": 66224 + }, + { + "epoch": 2.55, + "learning_rate": 7.5692485252628886e-06, + "loss": 0.1749, + "step": 66225 + }, + { + "epoch": 2.55, + "learning_rate": 7.568607335214158e-06, + "loss": 0.2641, + "step": 66226 + }, + { + "epoch": 2.55, + "learning_rate": 7.567966145165427e-06, + "loss": 0.2186, + "step": 66227 + }, + { + "epoch": 2.55, + "learning_rate": 7.567324955116698e-06, + "loss": 0.1021, + "step": 66228 + }, + { + "epoch": 2.55, + "learning_rate": 7.566683765067967e-06, + "loss": 0.1483, + "step": 66229 + }, + { + "epoch": 2.55, + "learning_rate": 7.566042575019236e-06, + "loss": 0.2029, + "step": 66230 + }, + { + "epoch": 2.55, + "learning_rate": 7.565401384970505e-06, + "loss": 0.3346, + "step": 66231 + }, + { + "epoch": 2.55, + "learning_rate": 7.564760194921776e-06, + "loss": 0.1433, + "step": 66232 + }, + { + "epoch": 2.55, + "learning_rate": 7.564119004873045e-06, + "loss": 0.1353, + "step": 66233 + }, + { + "epoch": 2.55, + "learning_rate": 7.5634778148243145e-06, + "loss": 0.1384, + "step": 66234 + }, + { + "epoch": 2.55, + "learning_rate": 7.562836624775584e-06, + "loss": 0.209, + "step": 66235 + }, + { + "epoch": 2.55, + "learning_rate": 7.562195434726853e-06, + "loss": 0.1649, + "step": 66236 + }, + { + "epoch": 2.55, + "learning_rate": 7.561554244678124e-06, + "loss": 0.1095, + "step": 66237 + }, + { + "epoch": 2.55, + "learning_rate": 7.560913054629393e-06, + "loss": 0.2506, + "step": 66238 + }, + { + "epoch": 2.55, + "learning_rate": 7.560271864580662e-06, + "loss": 0.2241, + "step": 66239 + }, + { + "epoch": 2.55, + "learning_rate": 7.559630674531931e-06, + "loss": 0.3265, + "step": 66240 + }, + { + "epoch": 2.55, + "learning_rate": 7.558989484483202e-06, + "loss": 0.1241, + "step": 66241 + }, + { + "epoch": 2.55, + "learning_rate": 7.558348294434471e-06, + "loss": 0.279, + "step": 66242 + }, + { + "epoch": 2.55, + "learning_rate": 7.5577071043857405e-06, + "loss": 0.2622, + "step": 66243 + }, + { + "epoch": 2.55, + "learning_rate": 7.55706591433701e-06, + "loss": 0.2651, + "step": 66244 + }, + { + "epoch": 2.55, + "learning_rate": 7.556424724288279e-06, + "loss": 0.1571, + "step": 66245 + }, + { + "epoch": 2.55, + "learning_rate": 7.55578353423955e-06, + "loss": 0.1964, + "step": 66246 + }, + { + "epoch": 2.55, + "learning_rate": 7.555142344190819e-06, + "loss": 0.2934, + "step": 66247 + }, + { + "epoch": 2.55, + "learning_rate": 7.554501154142088e-06, + "loss": 0.2075, + "step": 66248 + }, + { + "epoch": 2.55, + "learning_rate": 7.553859964093357e-06, + "loss": 0.2658, + "step": 66249 + }, + { + "epoch": 2.55, + "learning_rate": 7.553218774044628e-06, + "loss": 0.234, + "step": 66250 + }, + { + "epoch": 2.55, + "learning_rate": 7.552577583995897e-06, + "loss": 0.3537, + "step": 66251 + }, + { + "epoch": 2.55, + "learning_rate": 7.5519363939471665e-06, + "loss": 0.3317, + "step": 66252 + }, + { + "epoch": 2.55, + "learning_rate": 7.551295203898436e-06, + "loss": 0.2153, + "step": 66253 + }, + { + "epoch": 2.55, + "learning_rate": 7.550654013849705e-06, + "loss": 0.3822, + "step": 66254 + }, + { + "epoch": 2.55, + "learning_rate": 7.550012823800976e-06, + "loss": 0.288, + "step": 66255 + }, + { + "epoch": 2.55, + "learning_rate": 7.549371633752245e-06, + "loss": 0.1745, + "step": 66256 + }, + { + "epoch": 2.55, + "learning_rate": 7.548730443703514e-06, + "loss": 0.285, + "step": 66257 + }, + { + "epoch": 2.55, + "learning_rate": 7.548089253654783e-06, + "loss": 0.3398, + "step": 66258 + }, + { + "epoch": 2.55, + "learning_rate": 7.547448063606054e-06, + "loss": 0.235, + "step": 66259 + }, + { + "epoch": 2.55, + "learning_rate": 7.546806873557323e-06, + "loss": 0.2378, + "step": 66260 + }, + { + "epoch": 2.55, + "learning_rate": 7.5461656835085924e-06, + "loss": 0.217, + "step": 66261 + }, + { + "epoch": 2.55, + "learning_rate": 7.545524493459862e-06, + "loss": 0.1145, + "step": 66262 + }, + { + "epoch": 2.55, + "learning_rate": 7.544883303411131e-06, + "loss": 0.2182, + "step": 66263 + }, + { + "epoch": 2.55, + "learning_rate": 7.544242113362402e-06, + "loss": 0.3236, + "step": 66264 + }, + { + "epoch": 2.55, + "learning_rate": 7.543600923313671e-06, + "loss": 0.2856, + "step": 66265 + }, + { + "epoch": 2.55, + "learning_rate": 7.54295973326494e-06, + "loss": 0.3528, + "step": 66266 + }, + { + "epoch": 2.55, + "learning_rate": 7.542318543216209e-06, + "loss": 0.2144, + "step": 66267 + }, + { + "epoch": 2.55, + "learning_rate": 7.54167735316748e-06, + "loss": 0.2599, + "step": 66268 + }, + { + "epoch": 2.55, + "learning_rate": 7.541036163118749e-06, + "loss": 0.1333, + "step": 66269 + }, + { + "epoch": 2.55, + "learning_rate": 7.540394973070018e-06, + "loss": 0.2398, + "step": 66270 + }, + { + "epoch": 2.55, + "learning_rate": 7.539753783021288e-06, + "loss": 0.1957, + "step": 66271 + }, + { + "epoch": 2.55, + "learning_rate": 7.5391125929725585e-06, + "loss": 0.2819, + "step": 66272 + }, + { + "epoch": 2.55, + "learning_rate": 7.538471402923828e-06, + "loss": 0.1849, + "step": 66273 + }, + { + "epoch": 2.55, + "learning_rate": 7.537830212875097e-06, + "loss": 0.1963, + "step": 66274 + }, + { + "epoch": 2.55, + "learning_rate": 7.537189022826366e-06, + "loss": 0.3405, + "step": 66275 + }, + { + "epoch": 2.55, + "learning_rate": 7.536547832777635e-06, + "loss": 0.2977, + "step": 66276 + }, + { + "epoch": 2.55, + "learning_rate": 7.535906642728906e-06, + "loss": 0.2882, + "step": 66277 + }, + { + "epoch": 2.55, + "learning_rate": 7.535265452680175e-06, + "loss": 0.1778, + "step": 66278 + }, + { + "epoch": 2.55, + "learning_rate": 7.534624262631444e-06, + "loss": 0.1873, + "step": 66279 + }, + { + "epoch": 2.55, + "learning_rate": 7.5339830725827136e-06, + "loss": 0.2653, + "step": 66280 + }, + { + "epoch": 2.55, + "learning_rate": 7.5333418825339844e-06, + "loss": 0.1872, + "step": 66281 + }, + { + "epoch": 2.55, + "learning_rate": 7.532700692485254e-06, + "loss": 0.1457, + "step": 66282 + }, + { + "epoch": 2.55, + "learning_rate": 7.532059502436523e-06, + "loss": 0.3124, + "step": 66283 + }, + { + "epoch": 2.55, + "learning_rate": 7.531418312387792e-06, + "loss": 0.1525, + "step": 66284 + }, + { + "epoch": 2.55, + "learning_rate": 7.530777122339061e-06, + "loss": 0.1278, + "step": 66285 + }, + { + "epoch": 2.55, + "learning_rate": 7.530135932290332e-06, + "loss": 0.1511, + "step": 66286 + }, + { + "epoch": 2.55, + "learning_rate": 7.529494742241601e-06, + "loss": 0.1292, + "step": 66287 + }, + { + "epoch": 2.55, + "learning_rate": 7.52885355219287e-06, + "loss": 0.1768, + "step": 66288 + }, + { + "epoch": 2.55, + "learning_rate": 7.5282123621441395e-06, + "loss": 0.252, + "step": 66289 + }, + { + "epoch": 2.55, + "learning_rate": 7.52757117209541e-06, + "loss": 0.2355, + "step": 66290 + }, + { + "epoch": 2.55, + "learning_rate": 7.5269299820466796e-06, + "loss": 0.2879, + "step": 66291 + }, + { + "epoch": 2.55, + "learning_rate": 7.526288791997949e-06, + "loss": 0.1904, + "step": 66292 + }, + { + "epoch": 2.55, + "learning_rate": 7.525647601949218e-06, + "loss": 0.1977, + "step": 66293 + }, + { + "epoch": 2.55, + "learning_rate": 7.525006411900487e-06, + "loss": 0.1868, + "step": 66294 + }, + { + "epoch": 2.55, + "learning_rate": 7.524365221851758e-06, + "loss": 0.2924, + "step": 66295 + }, + { + "epoch": 2.55, + "learning_rate": 7.523724031803027e-06, + "loss": 0.1791, + "step": 66296 + }, + { + "epoch": 2.55, + "learning_rate": 7.523082841754296e-06, + "loss": 0.1735, + "step": 66297 + }, + { + "epoch": 2.55, + "learning_rate": 7.5224416517055655e-06, + "loss": 0.208, + "step": 66298 + }, + { + "epoch": 2.55, + "learning_rate": 7.521800461656836e-06, + "loss": 0.1368, + "step": 66299 + }, + { + "epoch": 2.55, + "learning_rate": 7.5211592716081055e-06, + "loss": 0.2439, + "step": 66300 + }, + { + "epoch": 2.55, + "learning_rate": 7.520518081559375e-06, + "loss": 0.3216, + "step": 66301 + }, + { + "epoch": 2.55, + "learning_rate": 7.519876891510644e-06, + "loss": 0.2437, + "step": 66302 + }, + { + "epoch": 2.55, + "learning_rate": 7.519235701461913e-06, + "loss": 0.2927, + "step": 66303 + }, + { + "epoch": 2.55, + "learning_rate": 7.518594511413184e-06, + "loss": 0.1616, + "step": 66304 + }, + { + "epoch": 2.55, + "learning_rate": 7.517953321364453e-06, + "loss": 0.2613, + "step": 66305 + }, + { + "epoch": 2.55, + "learning_rate": 7.517312131315722e-06, + "loss": 0.24, + "step": 66306 + }, + { + "epoch": 2.55, + "learning_rate": 7.5166709412669915e-06, + "loss": 0.1416, + "step": 66307 + }, + { + "epoch": 2.55, + "learning_rate": 7.516029751218262e-06, + "loss": 0.2309, + "step": 66308 + }, + { + "epoch": 2.55, + "learning_rate": 7.5153885611695315e-06, + "loss": 0.2017, + "step": 66309 + }, + { + "epoch": 2.55, + "learning_rate": 7.514747371120801e-06, + "loss": 0.2528, + "step": 66310 + }, + { + "epoch": 2.55, + "learning_rate": 7.51410618107207e-06, + "loss": 0.2244, + "step": 66311 + }, + { + "epoch": 2.55, + "learning_rate": 7.513464991023339e-06, + "loss": 0.2343, + "step": 66312 + }, + { + "epoch": 2.55, + "learning_rate": 7.51282380097461e-06, + "loss": 0.1528, + "step": 66313 + }, + { + "epoch": 2.55, + "learning_rate": 7.512182610925879e-06, + "loss": 0.1693, + "step": 66314 + }, + { + "epoch": 2.55, + "learning_rate": 7.511541420877148e-06, + "loss": 0.1143, + "step": 66315 + }, + { + "epoch": 2.55, + "learning_rate": 7.5109002308284174e-06, + "loss": 0.2545, + "step": 66316 + }, + { + "epoch": 2.55, + "learning_rate": 7.510259040779688e-06, + "loss": 0.2174, + "step": 66317 + }, + { + "epoch": 2.55, + "learning_rate": 7.5096178507309575e-06, + "loss": 0.3517, + "step": 66318 + }, + { + "epoch": 2.55, + "learning_rate": 7.508976660682227e-06, + "loss": 0.3048, + "step": 66319 + }, + { + "epoch": 2.55, + "learning_rate": 7.508335470633496e-06, + "loss": 0.3, + "step": 66320 + }, + { + "epoch": 2.55, + "learning_rate": 7.507694280584767e-06, + "loss": 0.1283, + "step": 66321 + }, + { + "epoch": 2.55, + "learning_rate": 7.507053090536036e-06, + "loss": 0.2299, + "step": 66322 + }, + { + "epoch": 2.55, + "learning_rate": 7.506411900487305e-06, + "loss": 0.2574, + "step": 66323 + }, + { + "epoch": 2.55, + "learning_rate": 7.505770710438574e-06, + "loss": 0.2959, + "step": 66324 + }, + { + "epoch": 2.55, + "learning_rate": 7.505129520389843e-06, + "loss": 0.2901, + "step": 66325 + }, + { + "epoch": 2.55, + "learning_rate": 7.504488330341114e-06, + "loss": 0.1807, + "step": 66326 + }, + { + "epoch": 2.55, + "learning_rate": 7.5038471402923834e-06, + "loss": 0.305, + "step": 66327 + }, + { + "epoch": 2.55, + "learning_rate": 7.503205950243653e-06, + "loss": 0.2056, + "step": 66328 + }, + { + "epoch": 2.55, + "learning_rate": 7.502564760194922e-06, + "loss": 0.2005, + "step": 66329 + }, + { + "epoch": 2.55, + "learning_rate": 7.501923570146193e-06, + "loss": 0.1972, + "step": 66330 + }, + { + "epoch": 2.55, + "learning_rate": 7.501282380097462e-06, + "loss": 0.2265, + "step": 66331 + }, + { + "epoch": 2.55, + "learning_rate": 7.500641190048731e-06, + "loss": 0.1232, + "step": 66332 + }, + { + "epoch": 2.55, + "learning_rate": 7.5e-06, + "loss": 0.1722, + "step": 66333 + }, + { + "epoch": 2.55, + "learning_rate": 7.499358809951269e-06, + "loss": 0.2129, + "step": 66334 + }, + { + "epoch": 2.55, + "learning_rate": 7.49871761990254e-06, + "loss": 0.2988, + "step": 66335 + }, + { + "epoch": 2.55, + "learning_rate": 7.498076429853809e-06, + "loss": 0.3207, + "step": 66336 + }, + { + "epoch": 2.55, + "learning_rate": 7.497435239805079e-06, + "loss": 0.2397, + "step": 66337 + }, + { + "epoch": 2.55, + "learning_rate": 7.496794049756348e-06, + "loss": 0.1094, + "step": 66338 + }, + { + "epoch": 2.55, + "learning_rate": 7.496152859707619e-06, + "loss": 0.2817, + "step": 66339 + }, + { + "epoch": 2.55, + "learning_rate": 7.495511669658888e-06, + "loss": 0.1561, + "step": 66340 + }, + { + "epoch": 2.55, + "learning_rate": 7.494870479610157e-06, + "loss": 0.1942, + "step": 66341 + }, + { + "epoch": 2.55, + "learning_rate": 7.494229289561426e-06, + "loss": 0.1502, + "step": 66342 + }, + { + "epoch": 2.55, + "learning_rate": 7.493588099512695e-06, + "loss": 0.2648, + "step": 66343 + }, + { + "epoch": 2.55, + "learning_rate": 7.492946909463966e-06, + "loss": 0.1819, + "step": 66344 + }, + { + "epoch": 2.55, + "learning_rate": 7.492305719415235e-06, + "loss": 0.2402, + "step": 66345 + }, + { + "epoch": 2.55, + "learning_rate": 7.4916645293665046e-06, + "loss": 0.2143, + "step": 66346 + }, + { + "epoch": 2.55, + "learning_rate": 7.491023339317774e-06, + "loss": 0.2697, + "step": 66347 + }, + { + "epoch": 2.55, + "learning_rate": 7.490382149269045e-06, + "loss": 0.3063, + "step": 66348 + }, + { + "epoch": 2.55, + "learning_rate": 7.489740959220314e-06, + "loss": 0.4034, + "step": 66349 + }, + { + "epoch": 2.55, + "learning_rate": 7.489099769171583e-06, + "loss": 0.2553, + "step": 66350 + }, + { + "epoch": 2.55, + "learning_rate": 7.488458579122852e-06, + "loss": 0.2038, + "step": 66351 + }, + { + "epoch": 2.55, + "learning_rate": 7.487817389074121e-06, + "loss": 0.3015, + "step": 66352 + }, + { + "epoch": 2.55, + "learning_rate": 7.487176199025392e-06, + "loss": 0.1071, + "step": 66353 + }, + { + "epoch": 2.55, + "learning_rate": 7.486535008976661e-06, + "loss": 0.1899, + "step": 66354 + }, + { + "epoch": 2.55, + "learning_rate": 7.4858938189279305e-06, + "loss": 0.3583, + "step": 66355 + }, + { + "epoch": 2.55, + "learning_rate": 7.4852526288792e-06, + "loss": 0.2191, + "step": 66356 + }, + { + "epoch": 2.55, + "learning_rate": 7.4846114388304706e-06, + "loss": 0.189, + "step": 66357 + }, + { + "epoch": 2.55, + "learning_rate": 7.48397024878174e-06, + "loss": 0.269, + "step": 66358 + }, + { + "epoch": 2.55, + "learning_rate": 7.483329058733009e-06, + "loss": 0.262, + "step": 66359 + }, + { + "epoch": 2.55, + "learning_rate": 7.482687868684278e-06, + "loss": 0.1958, + "step": 66360 + }, + { + "epoch": 2.55, + "learning_rate": 7.482046678635547e-06, + "loss": 0.1399, + "step": 66361 + }, + { + "epoch": 2.55, + "learning_rate": 7.481405488586818e-06, + "loss": 0.1893, + "step": 66362 + }, + { + "epoch": 2.55, + "learning_rate": 7.480764298538087e-06, + "loss": 0.3046, + "step": 66363 + }, + { + "epoch": 2.55, + "learning_rate": 7.4801231084893565e-06, + "loss": 0.1747, + "step": 66364 + }, + { + "epoch": 2.55, + "learning_rate": 7.479481918440626e-06, + "loss": 0.187, + "step": 66365 + }, + { + "epoch": 2.55, + "learning_rate": 7.4788407283918965e-06, + "loss": 0.3257, + "step": 66366 + }, + { + "epoch": 2.55, + "learning_rate": 7.478199538343166e-06, + "loss": 0.1217, + "step": 66367 + }, + { + "epoch": 2.55, + "learning_rate": 7.477558348294435e-06, + "loss": 0.2902, + "step": 66368 + }, + { + "epoch": 2.55, + "learning_rate": 7.476917158245704e-06, + "loss": 0.1904, + "step": 66369 + }, + { + "epoch": 2.55, + "learning_rate": 7.476275968196975e-06, + "loss": 0.1383, + "step": 66370 + }, + { + "epoch": 2.55, + "learning_rate": 7.475634778148244e-06, + "loss": 0.1561, + "step": 66371 + }, + { + "epoch": 2.55, + "learning_rate": 7.474993588099513e-06, + "loss": 0.1811, + "step": 66372 + }, + { + "epoch": 2.55, + "learning_rate": 7.4743523980507825e-06, + "loss": 0.4018, + "step": 66373 + }, + { + "epoch": 2.55, + "learning_rate": 7.473711208002052e-06, + "loss": 0.3625, + "step": 66374 + }, + { + "epoch": 2.55, + "learning_rate": 7.4730700179533225e-06, + "loss": 0.2982, + "step": 66375 + }, + { + "epoch": 2.55, + "learning_rate": 7.472428827904592e-06, + "loss": 0.1022, + "step": 66376 + }, + { + "epoch": 2.55, + "learning_rate": 7.471787637855861e-06, + "loss": 0.2191, + "step": 66377 + }, + { + "epoch": 2.55, + "learning_rate": 7.47114644780713e-06, + "loss": 0.2074, + "step": 66378 + }, + { + "epoch": 2.55, + "learning_rate": 7.470505257758401e-06, + "loss": 0.1321, + "step": 66379 + }, + { + "epoch": 2.55, + "learning_rate": 7.46986406770967e-06, + "loss": 0.1528, + "step": 66380 + }, + { + "epoch": 2.55, + "learning_rate": 7.469222877660939e-06, + "loss": 0.2253, + "step": 66381 + }, + { + "epoch": 2.55, + "learning_rate": 7.4685816876122084e-06, + "loss": 0.3312, + "step": 66382 + }, + { + "epoch": 2.55, + "learning_rate": 7.467940497563478e-06, + "loss": 0.1131, + "step": 66383 + }, + { + "epoch": 2.55, + "learning_rate": 7.4672993075147485e-06, + "loss": 0.1259, + "step": 66384 + }, + { + "epoch": 2.55, + "learning_rate": 7.466658117466018e-06, + "loss": 0.2322, + "step": 66385 + }, + { + "epoch": 2.55, + "learning_rate": 7.466016927417287e-06, + "loss": 0.2949, + "step": 66386 + }, + { + "epoch": 2.55, + "learning_rate": 7.465375737368556e-06, + "loss": 0.2321, + "step": 66387 + }, + { + "epoch": 2.55, + "learning_rate": 7.464734547319827e-06, + "loss": 0.2336, + "step": 66388 + }, + { + "epoch": 2.55, + "learning_rate": 7.464093357271096e-06, + "loss": 0.2961, + "step": 66389 + }, + { + "epoch": 2.55, + "learning_rate": 7.463452167222365e-06, + "loss": 0.2116, + "step": 66390 + }, + { + "epoch": 2.55, + "learning_rate": 7.462810977173634e-06, + "loss": 0.2411, + "step": 66391 + }, + { + "epoch": 2.55, + "learning_rate": 7.4621697871249036e-06, + "loss": 0.1497, + "step": 66392 + }, + { + "epoch": 2.55, + "learning_rate": 7.4615285970761744e-06, + "loss": 0.0944, + "step": 66393 + }, + { + "epoch": 2.55, + "learning_rate": 7.460887407027444e-06, + "loss": 0.2018, + "step": 66394 + }, + { + "epoch": 2.55, + "learning_rate": 7.460246216978713e-06, + "loss": 0.2039, + "step": 66395 + }, + { + "epoch": 2.55, + "learning_rate": 7.459605026929982e-06, + "loss": 0.1219, + "step": 66396 + }, + { + "epoch": 2.55, + "learning_rate": 7.458963836881253e-06, + "loss": 0.2144, + "step": 66397 + }, + { + "epoch": 2.55, + "learning_rate": 7.458322646832522e-06, + "loss": 0.2422, + "step": 66398 + }, + { + "epoch": 2.55, + "learning_rate": 7.457681456783791e-06, + "loss": 0.1429, + "step": 66399 + }, + { + "epoch": 2.55, + "learning_rate": 7.45704026673506e-06, + "loss": 0.2356, + "step": 66400 + }, + { + "epoch": 2.55, + "learning_rate": 7.4563990766863295e-06, + "loss": 0.1245, + "step": 66401 + }, + { + "epoch": 2.55, + "learning_rate": 7.4557578866376e-06, + "loss": 0.2181, + "step": 66402 + }, + { + "epoch": 2.55, + "learning_rate": 7.45511669658887e-06, + "loss": 0.247, + "step": 66403 + }, + { + "epoch": 2.55, + "learning_rate": 7.454475506540139e-06, + "loss": 0.217, + "step": 66404 + }, + { + "epoch": 2.55, + "learning_rate": 7.453834316491408e-06, + "loss": 0.2698, + "step": 66405 + }, + { + "epoch": 2.55, + "learning_rate": 7.453193126442679e-06, + "loss": 0.1403, + "step": 66406 + }, + { + "epoch": 2.55, + "learning_rate": 7.452551936393948e-06, + "loss": 0.2346, + "step": 66407 + }, + { + "epoch": 2.55, + "learning_rate": 7.451910746345217e-06, + "loss": 0.2758, + "step": 66408 + }, + { + "epoch": 2.55, + "learning_rate": 7.451269556296486e-06, + "loss": 0.2551, + "step": 66409 + }, + { + "epoch": 2.55, + "learning_rate": 7.4506283662477555e-06, + "loss": 0.1961, + "step": 66410 + }, + { + "epoch": 2.55, + "learning_rate": 7.449987176199026e-06, + "loss": 0.2355, + "step": 66411 + }, + { + "epoch": 2.55, + "learning_rate": 7.4493459861502955e-06, + "loss": 0.1879, + "step": 66412 + }, + { + "epoch": 2.55, + "learning_rate": 7.448704796101565e-06, + "loss": 0.233, + "step": 66413 + }, + { + "epoch": 2.55, + "learning_rate": 7.448063606052834e-06, + "loss": 0.303, + "step": 66414 + }, + { + "epoch": 2.55, + "learning_rate": 7.447422416004105e-06, + "loss": 0.1157, + "step": 66415 + }, + { + "epoch": 2.55, + "learning_rate": 7.446781225955374e-06, + "loss": 0.2345, + "step": 66416 + }, + { + "epoch": 2.55, + "learning_rate": 7.446140035906643e-06, + "loss": 0.3514, + "step": 66417 + }, + { + "epoch": 2.55, + "learning_rate": 7.445498845857912e-06, + "loss": 0.2799, + "step": 66418 + }, + { + "epoch": 2.55, + "learning_rate": 7.444857655809183e-06, + "loss": 0.3843, + "step": 66419 + }, + { + "epoch": 2.55, + "learning_rate": 7.444216465760452e-06, + "loss": 0.23, + "step": 66420 + }, + { + "epoch": 2.55, + "learning_rate": 7.4435752757117215e-06, + "loss": 0.28, + "step": 66421 + }, + { + "epoch": 2.55, + "learning_rate": 7.442934085662991e-06, + "loss": 0.3048, + "step": 66422 + }, + { + "epoch": 2.55, + "learning_rate": 7.44229289561426e-06, + "loss": 0.2321, + "step": 66423 + }, + { + "epoch": 2.55, + "learning_rate": 7.441651705565531e-06, + "loss": 0.3143, + "step": 66424 + }, + { + "epoch": 2.55, + "learning_rate": 7.4410105155168e-06, + "loss": 0.2558, + "step": 66425 + }, + { + "epoch": 2.55, + "learning_rate": 7.440369325468069e-06, + "loss": 0.2217, + "step": 66426 + }, + { + "epoch": 2.55, + "learning_rate": 7.439728135419338e-06, + "loss": 0.3274, + "step": 66427 + }, + { + "epoch": 2.55, + "learning_rate": 7.439086945370609e-06, + "loss": 0.2512, + "step": 66428 + }, + { + "epoch": 2.55, + "learning_rate": 7.438445755321878e-06, + "loss": 0.2921, + "step": 66429 + }, + { + "epoch": 2.55, + "learning_rate": 7.4378045652731475e-06, + "loss": 0.2888, + "step": 66430 + }, + { + "epoch": 2.55, + "learning_rate": 7.437163375224417e-06, + "loss": 0.2024, + "step": 66431 + }, + { + "epoch": 2.55, + "learning_rate": 7.436522185175686e-06, + "loss": 0.1409, + "step": 66432 + }, + { + "epoch": 2.55, + "learning_rate": 7.435880995126957e-06, + "loss": 0.3095, + "step": 66433 + }, + { + "epoch": 2.55, + "learning_rate": 7.435239805078226e-06, + "loss": 0.2072, + "step": 66434 + }, + { + "epoch": 2.55, + "learning_rate": 7.434598615029495e-06, + "loss": 0.2737, + "step": 66435 + }, + { + "epoch": 2.55, + "learning_rate": 7.433957424980764e-06, + "loss": 0.2113, + "step": 66436 + }, + { + "epoch": 2.55, + "learning_rate": 7.433316234932035e-06, + "loss": 0.1315, + "step": 66437 + }, + { + "epoch": 2.55, + "learning_rate": 7.432675044883304e-06, + "loss": 0.2408, + "step": 66438 + }, + { + "epoch": 2.55, + "learning_rate": 7.4320338548345735e-06, + "loss": 0.1954, + "step": 66439 + }, + { + "epoch": 2.55, + "learning_rate": 7.431392664785843e-06, + "loss": 0.1879, + "step": 66440 + }, + { + "epoch": 2.55, + "learning_rate": 7.430751474737112e-06, + "loss": 0.3234, + "step": 66441 + }, + { + "epoch": 2.55, + "learning_rate": 7.430110284688383e-06, + "loss": 0.1833, + "step": 66442 + }, + { + "epoch": 2.55, + "learning_rate": 7.429469094639652e-06, + "loss": 0.3013, + "step": 66443 + }, + { + "epoch": 2.55, + "learning_rate": 7.428827904590921e-06, + "loss": 0.2457, + "step": 66444 + }, + { + "epoch": 2.55, + "learning_rate": 7.42818671454219e-06, + "loss": 0.1555, + "step": 66445 + }, + { + "epoch": 2.55, + "learning_rate": 7.427545524493461e-06, + "loss": 0.3213, + "step": 66446 + }, + { + "epoch": 2.55, + "learning_rate": 7.42690433444473e-06, + "loss": 0.2809, + "step": 66447 + }, + { + "epoch": 2.55, + "learning_rate": 7.426263144395999e-06, + "loss": 0.1601, + "step": 66448 + }, + { + "epoch": 2.55, + "learning_rate": 7.425621954347269e-06, + "loss": 0.2596, + "step": 66449 + }, + { + "epoch": 2.55, + "learning_rate": 7.424980764298538e-06, + "loss": 0.2396, + "step": 66450 + }, + { + "epoch": 2.55, + "learning_rate": 7.424339574249809e-06, + "loss": 0.1534, + "step": 66451 + }, + { + "epoch": 2.55, + "learning_rate": 7.423698384201078e-06, + "loss": 0.2835, + "step": 66452 + }, + { + "epoch": 2.55, + "learning_rate": 7.423057194152347e-06, + "loss": 0.1712, + "step": 66453 + }, + { + "epoch": 2.55, + "learning_rate": 7.422416004103616e-06, + "loss": 0.2074, + "step": 66454 + }, + { + "epoch": 2.55, + "learning_rate": 7.421774814054887e-06, + "loss": 0.2537, + "step": 66455 + }, + { + "epoch": 2.56, + "learning_rate": 7.421133624006156e-06, + "loss": 0.1429, + "step": 66456 + }, + { + "epoch": 2.56, + "learning_rate": 7.420492433957425e-06, + "loss": 0.1184, + "step": 66457 + }, + { + "epoch": 2.56, + "learning_rate": 7.4198512439086946e-06, + "loss": 0.1501, + "step": 66458 + }, + { + "epoch": 2.56, + "learning_rate": 7.419210053859964e-06, + "loss": 0.206, + "step": 66459 + }, + { + "epoch": 2.56, + "learning_rate": 7.418568863811235e-06, + "loss": 0.1901, + "step": 66460 + }, + { + "epoch": 2.56, + "learning_rate": 7.417927673762504e-06, + "loss": 0.1409, + "step": 66461 + }, + { + "epoch": 2.56, + "learning_rate": 7.417286483713773e-06, + "loss": 0.2965, + "step": 66462 + }, + { + "epoch": 2.56, + "learning_rate": 7.416645293665042e-06, + "loss": 0.1934, + "step": 66463 + }, + { + "epoch": 2.56, + "learning_rate": 7.416004103616313e-06, + "loss": 0.1291, + "step": 66464 + }, + { + "epoch": 2.56, + "learning_rate": 7.415362913567582e-06, + "loss": 0.3403, + "step": 66465 + }, + { + "epoch": 2.56, + "learning_rate": 7.414721723518851e-06, + "loss": 0.2837, + "step": 66466 + }, + { + "epoch": 2.56, + "learning_rate": 7.4140805334701205e-06, + "loss": 0.2825, + "step": 66467 + }, + { + "epoch": 2.56, + "learning_rate": 7.413439343421391e-06, + "loss": 0.2072, + "step": 66468 + }, + { + "epoch": 2.56, + "learning_rate": 7.412798153372661e-06, + "loss": 0.1795, + "step": 66469 + }, + { + "epoch": 2.56, + "learning_rate": 7.41215696332393e-06, + "loss": 0.2954, + "step": 66470 + }, + { + "epoch": 2.56, + "learning_rate": 7.411515773275199e-06, + "loss": 0.2279, + "step": 66471 + }, + { + "epoch": 2.56, + "learning_rate": 7.410874583226468e-06, + "loss": 0.3129, + "step": 66472 + }, + { + "epoch": 2.56, + "learning_rate": 7.410233393177739e-06, + "loss": 0.3323, + "step": 66473 + }, + { + "epoch": 2.56, + "learning_rate": 7.409592203129008e-06, + "loss": 0.2199, + "step": 66474 + }, + { + "epoch": 2.56, + "learning_rate": 7.408951013080277e-06, + "loss": 0.2353, + "step": 66475 + }, + { + "epoch": 2.56, + "learning_rate": 7.4083098230315465e-06, + "loss": 0.1978, + "step": 66476 + }, + { + "epoch": 2.56, + "learning_rate": 7.407668632982817e-06, + "loss": 0.2526, + "step": 66477 + }, + { + "epoch": 2.56, + "learning_rate": 7.4070274429340865e-06, + "loss": 0.1747, + "step": 66478 + }, + { + "epoch": 2.56, + "learning_rate": 7.406386252885356e-06, + "loss": 0.1464, + "step": 66479 + }, + { + "epoch": 2.56, + "learning_rate": 7.405745062836625e-06, + "loss": 0.1606, + "step": 66480 + }, + { + "epoch": 2.56, + "learning_rate": 7.405103872787894e-06, + "loss": 0.1068, + "step": 66481 + }, + { + "epoch": 2.56, + "learning_rate": 7.404462682739165e-06, + "loss": 0.1825, + "step": 66482 + }, + { + "epoch": 2.56, + "learning_rate": 7.403821492690434e-06, + "loss": 0.2004, + "step": 66483 + }, + { + "epoch": 2.56, + "learning_rate": 7.403180302641703e-06, + "loss": 0.249, + "step": 66484 + }, + { + "epoch": 2.56, + "learning_rate": 7.4025391125929725e-06, + "loss": 0.1836, + "step": 66485 + }, + { + "epoch": 2.56, + "learning_rate": 7.401897922544243e-06, + "loss": 0.2181, + "step": 66486 + }, + { + "epoch": 2.56, + "learning_rate": 7.4012567324955125e-06, + "loss": 0.2047, + "step": 66487 + }, + { + "epoch": 2.56, + "learning_rate": 7.400615542446782e-06, + "loss": 0.1324, + "step": 66488 + }, + { + "epoch": 2.56, + "learning_rate": 7.399974352398051e-06, + "loss": 0.3072, + "step": 66489 + }, + { + "epoch": 2.56, + "learning_rate": 7.39933316234932e-06, + "loss": 0.1563, + "step": 66490 + }, + { + "epoch": 2.56, + "learning_rate": 7.398691972300591e-06, + "loss": 0.1589, + "step": 66491 + }, + { + "epoch": 2.56, + "learning_rate": 7.39805078225186e-06, + "loss": 0.236, + "step": 66492 + }, + { + "epoch": 2.56, + "learning_rate": 7.397409592203129e-06, + "loss": 0.1855, + "step": 66493 + }, + { + "epoch": 2.56, + "learning_rate": 7.3967684021543984e-06, + "loss": 0.3025, + "step": 66494 + }, + { + "epoch": 2.56, + "learning_rate": 7.396127212105669e-06, + "loss": 0.1758, + "step": 66495 + }, + { + "epoch": 2.56, + "learning_rate": 7.3954860220569385e-06, + "loss": 0.2247, + "step": 66496 + }, + { + "epoch": 2.56, + "learning_rate": 7.394844832008208e-06, + "loss": 0.2113, + "step": 66497 + }, + { + "epoch": 2.56, + "learning_rate": 7.394203641959477e-06, + "loss": 0.3265, + "step": 66498 + }, + { + "epoch": 2.56, + "learning_rate": 7.393562451910746e-06, + "loss": 0.1417, + "step": 66499 + }, + { + "epoch": 2.56, + "learning_rate": 7.392921261862017e-06, + "loss": 0.2798, + "step": 66500 + }, + { + "epoch": 2.56, + "learning_rate": 7.392280071813286e-06, + "loss": 0.238, + "step": 66501 + }, + { + "epoch": 2.56, + "learning_rate": 7.391638881764555e-06, + "loss": 0.1833, + "step": 66502 + }, + { + "epoch": 2.56, + "learning_rate": 7.390997691715824e-06, + "loss": 0.2222, + "step": 66503 + }, + { + "epoch": 2.56, + "learning_rate": 7.390356501667095e-06, + "loss": 0.205, + "step": 66504 + }, + { + "epoch": 2.56, + "learning_rate": 7.3897153116183644e-06, + "loss": 0.0932, + "step": 66505 + }, + { + "epoch": 2.56, + "learning_rate": 7.389074121569634e-06, + "loss": 0.3599, + "step": 66506 + }, + { + "epoch": 2.56, + "learning_rate": 7.388432931520903e-06, + "loss": 0.2728, + "step": 66507 + }, + { + "epoch": 2.56, + "learning_rate": 7.387791741472172e-06, + "loss": 0.184, + "step": 66508 + }, + { + "epoch": 2.56, + "learning_rate": 7.387150551423443e-06, + "loss": 0.2599, + "step": 66509 + }, + { + "epoch": 2.56, + "learning_rate": 7.386509361374712e-06, + "loss": 0.3647, + "step": 66510 + }, + { + "epoch": 2.56, + "learning_rate": 7.385868171325981e-06, + "loss": 0.2247, + "step": 66511 + }, + { + "epoch": 2.56, + "learning_rate": 7.38522698127725e-06, + "loss": 0.0968, + "step": 66512 + }, + { + "epoch": 2.56, + "learning_rate": 7.384585791228521e-06, + "loss": 0.2222, + "step": 66513 + }, + { + "epoch": 2.56, + "learning_rate": 7.38394460117979e-06, + "loss": 0.1874, + "step": 66514 + }, + { + "epoch": 2.56, + "learning_rate": 7.38330341113106e-06, + "loss": 0.1003, + "step": 66515 + }, + { + "epoch": 2.56, + "learning_rate": 7.382662221082329e-06, + "loss": 0.1897, + "step": 66516 + }, + { + "epoch": 2.56, + "learning_rate": 7.3820210310336e-06, + "loss": 0.1351, + "step": 66517 + }, + { + "epoch": 2.56, + "learning_rate": 7.381379840984869e-06, + "loss": 0.1983, + "step": 66518 + }, + { + "epoch": 2.56, + "learning_rate": 7.380738650936138e-06, + "loss": 0.2188, + "step": 66519 + }, + { + "epoch": 2.56, + "learning_rate": 7.380097460887407e-06, + "loss": 0.1045, + "step": 66520 + }, + { + "epoch": 2.56, + "learning_rate": 7.379456270838676e-06, + "loss": 0.2606, + "step": 66521 + }, + { + "epoch": 2.56, + "learning_rate": 7.378815080789947e-06, + "loss": 0.173, + "step": 66522 + }, + { + "epoch": 2.56, + "learning_rate": 7.378173890741216e-06, + "loss": 0.2565, + "step": 66523 + }, + { + "epoch": 2.56, + "learning_rate": 7.3775327006924856e-06, + "loss": 0.2414, + "step": 66524 + }, + { + "epoch": 2.56, + "learning_rate": 7.376891510643755e-06, + "loss": 0.2445, + "step": 66525 + }, + { + "epoch": 2.56, + "learning_rate": 7.376250320595026e-06, + "loss": 0.1826, + "step": 66526 + }, + { + "epoch": 2.56, + "learning_rate": 7.375609130546295e-06, + "loss": 0.3669, + "step": 66527 + }, + { + "epoch": 2.56, + "learning_rate": 7.374967940497564e-06, + "loss": 0.2714, + "step": 66528 + }, + { + "epoch": 2.56, + "learning_rate": 7.374326750448833e-06, + "loss": 0.2056, + "step": 66529 + }, + { + "epoch": 2.56, + "learning_rate": 7.373685560400102e-06, + "loss": 0.1124, + "step": 66530 + }, + { + "epoch": 2.56, + "learning_rate": 7.373044370351373e-06, + "loss": 0.1422, + "step": 66531 + }, + { + "epoch": 2.56, + "learning_rate": 7.372403180302642e-06, + "loss": 0.1109, + "step": 66532 + }, + { + "epoch": 2.56, + "learning_rate": 7.3717619902539115e-06, + "loss": 0.2203, + "step": 66533 + }, + { + "epoch": 2.56, + "learning_rate": 7.371120800205181e-06, + "loss": 0.407, + "step": 66534 + }, + { + "epoch": 2.56, + "learning_rate": 7.3704796101564516e-06, + "loss": 0.2218, + "step": 66535 + }, + { + "epoch": 2.56, + "learning_rate": 7.369838420107721e-06, + "loss": 0.169, + "step": 66536 + }, + { + "epoch": 2.56, + "learning_rate": 7.36919723005899e-06, + "loss": 0.2947, + "step": 66537 + }, + { + "epoch": 2.56, + "learning_rate": 7.368556040010259e-06, + "loss": 0.1857, + "step": 66538 + }, + { + "epoch": 2.56, + "learning_rate": 7.367914849961528e-06, + "loss": 0.1442, + "step": 66539 + }, + { + "epoch": 2.56, + "learning_rate": 7.367273659912799e-06, + "loss": 0.1592, + "step": 66540 + }, + { + "epoch": 2.56, + "learning_rate": 7.366632469864068e-06, + "loss": 0.2889, + "step": 66541 + }, + { + "epoch": 2.56, + "learning_rate": 7.3659912798153375e-06, + "loss": 0.2678, + "step": 66542 + }, + { + "epoch": 2.56, + "learning_rate": 7.365350089766607e-06, + "loss": 0.3133, + "step": 66543 + }, + { + "epoch": 2.56, + "learning_rate": 7.3647088997178775e-06, + "loss": 0.1003, + "step": 66544 + }, + { + "epoch": 2.56, + "learning_rate": 7.364067709669147e-06, + "loss": 0.2129, + "step": 66545 + }, + { + "epoch": 2.56, + "learning_rate": 7.363426519620416e-06, + "loss": 0.1513, + "step": 66546 + }, + { + "epoch": 2.56, + "learning_rate": 7.362785329571685e-06, + "loss": 0.1722, + "step": 66547 + }, + { + "epoch": 2.56, + "learning_rate": 7.362144139522954e-06, + "loss": 0.2843, + "step": 66548 + }, + { + "epoch": 2.56, + "learning_rate": 7.361502949474225e-06, + "loss": 0.1159, + "step": 66549 + }, + { + "epoch": 2.56, + "learning_rate": 7.360861759425494e-06, + "loss": 0.2314, + "step": 66550 + }, + { + "epoch": 2.56, + "learning_rate": 7.3602205693767635e-06, + "loss": 0.22, + "step": 66551 + }, + { + "epoch": 2.56, + "learning_rate": 7.359579379328033e-06, + "loss": 0.1583, + "step": 66552 + }, + { + "epoch": 2.56, + "learning_rate": 7.3589381892793035e-06, + "loss": 0.2363, + "step": 66553 + }, + { + "epoch": 2.56, + "learning_rate": 7.358296999230573e-06, + "loss": 0.238, + "step": 66554 + }, + { + "epoch": 2.56, + "learning_rate": 7.357655809181842e-06, + "loss": 0.2991, + "step": 66555 + }, + { + "epoch": 2.56, + "learning_rate": 7.357014619133111e-06, + "loss": 0.2311, + "step": 66556 + }, + { + "epoch": 2.56, + "learning_rate": 7.356373429084382e-06, + "loss": 0.1945, + "step": 66557 + }, + { + "epoch": 2.56, + "learning_rate": 7.355732239035651e-06, + "loss": 0.3004, + "step": 66558 + }, + { + "epoch": 2.56, + "learning_rate": 7.35509104898692e-06, + "loss": 0.2589, + "step": 66559 + }, + { + "epoch": 2.56, + "learning_rate": 7.3544498589381894e-06, + "loss": 0.1864, + "step": 66560 + }, + { + "epoch": 2.56, + "learning_rate": 7.353808668889459e-06, + "loss": 0.2392, + "step": 66561 + }, + { + "epoch": 2.56, + "learning_rate": 7.3531674788407295e-06, + "loss": 0.1675, + "step": 66562 + }, + { + "epoch": 2.56, + "learning_rate": 7.352526288791999e-06, + "loss": 0.2329, + "step": 66563 + }, + { + "epoch": 2.56, + "learning_rate": 7.351885098743268e-06, + "loss": 0.2106, + "step": 66564 + }, + { + "epoch": 2.56, + "learning_rate": 7.351243908694537e-06, + "loss": 0.1959, + "step": 66565 + }, + { + "epoch": 2.56, + "learning_rate": 7.350602718645808e-06, + "loss": 0.2198, + "step": 66566 + }, + { + "epoch": 2.56, + "learning_rate": 7.349961528597077e-06, + "loss": 0.1907, + "step": 66567 + }, + { + "epoch": 2.56, + "learning_rate": 7.349320338548346e-06, + "loss": 0.212, + "step": 66568 + }, + { + "epoch": 2.56, + "learning_rate": 7.348679148499615e-06, + "loss": 0.1429, + "step": 66569 + }, + { + "epoch": 2.56, + "learning_rate": 7.3480379584508846e-06, + "loss": 0.1472, + "step": 66570 + }, + { + "epoch": 2.56, + "learning_rate": 7.3473967684021554e-06, + "loss": 0.2294, + "step": 66571 + }, + { + "epoch": 2.56, + "learning_rate": 7.346755578353425e-06, + "loss": 0.1855, + "step": 66572 + }, + { + "epoch": 2.56, + "learning_rate": 7.346114388304694e-06, + "loss": 0.1328, + "step": 66573 + }, + { + "epoch": 2.56, + "learning_rate": 7.345473198255963e-06, + "loss": 0.2231, + "step": 66574 + }, + { + "epoch": 2.56, + "learning_rate": 7.344832008207234e-06, + "loss": 0.3594, + "step": 66575 + }, + { + "epoch": 2.56, + "learning_rate": 7.344190818158503e-06, + "loss": 0.2771, + "step": 66576 + }, + { + "epoch": 2.56, + "learning_rate": 7.343549628109772e-06, + "loss": 0.1173, + "step": 66577 + }, + { + "epoch": 2.56, + "learning_rate": 7.342908438061041e-06, + "loss": 0.321, + "step": 66578 + }, + { + "epoch": 2.56, + "learning_rate": 7.3422672480123105e-06, + "loss": 0.1289, + "step": 66579 + }, + { + "epoch": 2.56, + "learning_rate": 7.341626057963581e-06, + "loss": 0.2351, + "step": 66580 + }, + { + "epoch": 2.56, + "learning_rate": 7.340984867914851e-06, + "loss": 0.3023, + "step": 66581 + }, + { + "epoch": 2.56, + "learning_rate": 7.34034367786612e-06, + "loss": 0.2833, + "step": 66582 + }, + { + "epoch": 2.56, + "learning_rate": 7.339702487817389e-06, + "loss": 0.2225, + "step": 66583 + }, + { + "epoch": 2.56, + "learning_rate": 7.33906129776866e-06, + "loss": 0.1543, + "step": 66584 + }, + { + "epoch": 2.56, + "learning_rate": 7.338420107719929e-06, + "loss": 0.1494, + "step": 66585 + }, + { + "epoch": 2.56, + "learning_rate": 7.337778917671198e-06, + "loss": 0.2754, + "step": 66586 + }, + { + "epoch": 2.56, + "learning_rate": 7.337137727622467e-06, + "loss": 0.1601, + "step": 66587 + }, + { + "epoch": 2.56, + "learning_rate": 7.3364965375737365e-06, + "loss": 0.1085, + "step": 66588 + }, + { + "epoch": 2.56, + "learning_rate": 7.335855347525007e-06, + "loss": 0.0857, + "step": 66589 + }, + { + "epoch": 2.56, + "learning_rate": 7.3352141574762766e-06, + "loss": 0.1529, + "step": 66590 + }, + { + "epoch": 2.56, + "learning_rate": 7.334572967427546e-06, + "loss": 0.2163, + "step": 66591 + }, + { + "epoch": 2.56, + "learning_rate": 7.333931777378815e-06, + "loss": 0.1887, + "step": 66592 + }, + { + "epoch": 2.56, + "learning_rate": 7.333290587330086e-06, + "loss": 0.1915, + "step": 66593 + }, + { + "epoch": 2.56, + "learning_rate": 7.332649397281355e-06, + "loss": 0.3532, + "step": 66594 + }, + { + "epoch": 2.56, + "learning_rate": 7.332008207232624e-06, + "loss": 0.1295, + "step": 66595 + }, + { + "epoch": 2.56, + "learning_rate": 7.331367017183893e-06, + "loss": 0.1612, + "step": 66596 + }, + { + "epoch": 2.56, + "learning_rate": 7.3307258271351625e-06, + "loss": 0.1406, + "step": 66597 + }, + { + "epoch": 2.56, + "learning_rate": 7.330084637086433e-06, + "loss": 0.2443, + "step": 66598 + }, + { + "epoch": 2.56, + "learning_rate": 7.3294434470377025e-06, + "loss": 0.187, + "step": 66599 + }, + { + "epoch": 2.56, + "learning_rate": 7.328802256988972e-06, + "loss": 0.1296, + "step": 66600 + }, + { + "epoch": 2.56, + "learning_rate": 7.328161066940241e-06, + "loss": 0.0977, + "step": 66601 + }, + { + "epoch": 2.56, + "learning_rate": 7.327519876891512e-06, + "loss": 0.2834, + "step": 66602 + }, + { + "epoch": 2.56, + "learning_rate": 7.326878686842781e-06, + "loss": 0.1319, + "step": 66603 + }, + { + "epoch": 2.56, + "learning_rate": 7.32623749679405e-06, + "loss": 0.2216, + "step": 66604 + }, + { + "epoch": 2.56, + "learning_rate": 7.325596306745319e-06, + "loss": 0.2684, + "step": 66605 + }, + { + "epoch": 2.56, + "learning_rate": 7.32495511669659e-06, + "loss": 0.2662, + "step": 66606 + }, + { + "epoch": 2.56, + "learning_rate": 7.324313926647859e-06, + "loss": 0.3, + "step": 66607 + }, + { + "epoch": 2.56, + "learning_rate": 7.3236727365991285e-06, + "loss": 0.2485, + "step": 66608 + }, + { + "epoch": 2.56, + "learning_rate": 7.323031546550398e-06, + "loss": 0.3874, + "step": 66609 + }, + { + "epoch": 2.56, + "learning_rate": 7.322390356501667e-06, + "loss": 0.1467, + "step": 66610 + }, + { + "epoch": 2.56, + "learning_rate": 7.321749166452938e-06, + "loss": 0.1984, + "step": 66611 + }, + { + "epoch": 2.56, + "learning_rate": 7.321107976404207e-06, + "loss": 0.1994, + "step": 66612 + }, + { + "epoch": 2.56, + "learning_rate": 7.320466786355476e-06, + "loss": 0.2647, + "step": 66613 + }, + { + "epoch": 2.56, + "learning_rate": 7.319825596306745e-06, + "loss": 0.2062, + "step": 66614 + }, + { + "epoch": 2.56, + "learning_rate": 7.319184406258016e-06, + "loss": 0.2182, + "step": 66615 + }, + { + "epoch": 2.56, + "learning_rate": 7.318543216209285e-06, + "loss": 0.1563, + "step": 66616 + }, + { + "epoch": 2.56, + "learning_rate": 7.3179020261605545e-06, + "loss": 0.1752, + "step": 66617 + }, + { + "epoch": 2.56, + "learning_rate": 7.317260836111824e-06, + "loss": 0.1953, + "step": 66618 + }, + { + "epoch": 2.56, + "learning_rate": 7.316619646063093e-06, + "loss": 0.107, + "step": 66619 + }, + { + "epoch": 2.56, + "learning_rate": 7.315978456014364e-06, + "loss": 0.096, + "step": 66620 + }, + { + "epoch": 2.56, + "learning_rate": 7.315337265965633e-06, + "loss": 0.1831, + "step": 66621 + }, + { + "epoch": 2.56, + "learning_rate": 7.314696075916902e-06, + "loss": 0.1528, + "step": 66622 + }, + { + "epoch": 2.56, + "learning_rate": 7.314054885868171e-06, + "loss": 0.2429, + "step": 66623 + }, + { + "epoch": 2.56, + "learning_rate": 7.313413695819442e-06, + "loss": 0.2293, + "step": 66624 + }, + { + "epoch": 2.56, + "learning_rate": 7.312772505770711e-06, + "loss": 0.2438, + "step": 66625 + }, + { + "epoch": 2.56, + "learning_rate": 7.3121313157219804e-06, + "loss": 0.2718, + "step": 66626 + }, + { + "epoch": 2.56, + "learning_rate": 7.31149012567325e-06, + "loss": 0.111, + "step": 66627 + }, + { + "epoch": 2.56, + "learning_rate": 7.310848935624519e-06, + "loss": 0.2606, + "step": 66628 + }, + { + "epoch": 2.56, + "learning_rate": 7.31020774557579e-06, + "loss": 0.2103, + "step": 66629 + }, + { + "epoch": 2.56, + "learning_rate": 7.309566555527059e-06, + "loss": 0.1399, + "step": 66630 + }, + { + "epoch": 2.56, + "learning_rate": 7.308925365478328e-06, + "loss": 0.2078, + "step": 66631 + }, + { + "epoch": 2.56, + "learning_rate": 7.308284175429597e-06, + "loss": 0.1722, + "step": 66632 + }, + { + "epoch": 2.56, + "learning_rate": 7.307642985380868e-06, + "loss": 0.2256, + "step": 66633 + }, + { + "epoch": 2.56, + "learning_rate": 7.307001795332137e-06, + "loss": 0.2237, + "step": 66634 + }, + { + "epoch": 2.56, + "learning_rate": 7.306360605283406e-06, + "loss": 0.1602, + "step": 66635 + }, + { + "epoch": 2.56, + "learning_rate": 7.3057194152346756e-06, + "loss": 0.2657, + "step": 66636 + }, + { + "epoch": 2.56, + "learning_rate": 7.305078225185945e-06, + "loss": 0.3015, + "step": 66637 + }, + { + "epoch": 2.56, + "learning_rate": 7.304437035137216e-06, + "loss": 0.3219, + "step": 66638 + }, + { + "epoch": 2.56, + "learning_rate": 7.303795845088485e-06, + "loss": 0.3435, + "step": 66639 + }, + { + "epoch": 2.56, + "learning_rate": 7.303154655039754e-06, + "loss": 0.3105, + "step": 66640 + }, + { + "epoch": 2.56, + "learning_rate": 7.302513464991023e-06, + "loss": 0.2665, + "step": 66641 + }, + { + "epoch": 2.56, + "learning_rate": 7.301872274942294e-06, + "loss": 0.1939, + "step": 66642 + }, + { + "epoch": 2.56, + "learning_rate": 7.301231084893563e-06, + "loss": 0.167, + "step": 66643 + }, + { + "epoch": 2.56, + "learning_rate": 7.300589894844832e-06, + "loss": 0.1735, + "step": 66644 + }, + { + "epoch": 2.56, + "learning_rate": 7.2999487047961015e-06, + "loss": 0.3741, + "step": 66645 + }, + { + "epoch": 2.56, + "learning_rate": 7.299307514747371e-06, + "loss": 0.1686, + "step": 66646 + }, + { + "epoch": 2.56, + "learning_rate": 7.298666324698642e-06, + "loss": 0.221, + "step": 66647 + }, + { + "epoch": 2.56, + "learning_rate": 7.298025134649911e-06, + "loss": 0.2327, + "step": 66648 + }, + { + "epoch": 2.56, + "learning_rate": 7.29738394460118e-06, + "loss": 0.2321, + "step": 66649 + }, + { + "epoch": 2.56, + "learning_rate": 7.296742754552449e-06, + "loss": 0.2273, + "step": 66650 + }, + { + "epoch": 2.56, + "learning_rate": 7.29610156450372e-06, + "loss": 0.1984, + "step": 66651 + }, + { + "epoch": 2.56, + "learning_rate": 7.295460374454989e-06, + "loss": 0.1812, + "step": 66652 + }, + { + "epoch": 2.56, + "learning_rate": 7.294819184406258e-06, + "loss": 0.207, + "step": 66653 + }, + { + "epoch": 2.56, + "learning_rate": 7.2941779943575275e-06, + "loss": 0.2138, + "step": 66654 + }, + { + "epoch": 2.56, + "learning_rate": 7.293536804308798e-06, + "loss": 0.164, + "step": 66655 + }, + { + "epoch": 2.56, + "learning_rate": 7.2928956142600676e-06, + "loss": 0.2948, + "step": 66656 + }, + { + "epoch": 2.56, + "learning_rate": 7.292254424211337e-06, + "loss": 0.1268, + "step": 66657 + }, + { + "epoch": 2.56, + "learning_rate": 7.291613234162606e-06, + "loss": 0.1688, + "step": 66658 + }, + { + "epoch": 2.56, + "learning_rate": 7.290972044113875e-06, + "loss": 0.1939, + "step": 66659 + }, + { + "epoch": 2.56, + "learning_rate": 7.290330854065146e-06, + "loss": 0.1157, + "step": 66660 + }, + { + "epoch": 2.56, + "learning_rate": 7.289689664016415e-06, + "loss": 0.2459, + "step": 66661 + }, + { + "epoch": 2.56, + "learning_rate": 7.289048473967684e-06, + "loss": 0.1606, + "step": 66662 + }, + { + "epoch": 2.56, + "learning_rate": 7.2884072839189535e-06, + "loss": 0.2831, + "step": 66663 + }, + { + "epoch": 2.56, + "learning_rate": 7.287766093870224e-06, + "loss": 0.1936, + "step": 66664 + }, + { + "epoch": 2.56, + "learning_rate": 7.2871249038214935e-06, + "loss": 0.1891, + "step": 66665 + }, + { + "epoch": 2.56, + "learning_rate": 7.286483713772763e-06, + "loss": 0.216, + "step": 66666 + }, + { + "epoch": 2.56, + "learning_rate": 7.285842523724032e-06, + "loss": 0.174, + "step": 66667 + }, + { + "epoch": 2.56, + "learning_rate": 7.285201333675301e-06, + "loss": 0.2102, + "step": 66668 + }, + { + "epoch": 2.56, + "learning_rate": 7.284560143626572e-06, + "loss": 0.1963, + "step": 66669 + }, + { + "epoch": 2.56, + "learning_rate": 7.283918953577841e-06, + "loss": 0.1161, + "step": 66670 + }, + { + "epoch": 2.56, + "learning_rate": 7.28327776352911e-06, + "loss": 0.1421, + "step": 66671 + }, + { + "epoch": 2.56, + "learning_rate": 7.2826365734803794e-06, + "loss": 0.2038, + "step": 66672 + }, + { + "epoch": 2.56, + "learning_rate": 7.28199538343165e-06, + "loss": 0.2079, + "step": 66673 + }, + { + "epoch": 2.56, + "learning_rate": 7.2813541933829195e-06, + "loss": 0.1665, + "step": 66674 + }, + { + "epoch": 2.56, + "learning_rate": 7.280713003334189e-06, + "loss": 0.3535, + "step": 66675 + }, + { + "epoch": 2.56, + "learning_rate": 7.280071813285458e-06, + "loss": 0.2414, + "step": 66676 + }, + { + "epoch": 2.56, + "learning_rate": 7.279430623236727e-06, + "loss": 0.2129, + "step": 66677 + }, + { + "epoch": 2.56, + "learning_rate": 7.278789433187998e-06, + "loss": 0.2411, + "step": 66678 + }, + { + "epoch": 2.56, + "learning_rate": 7.278148243139267e-06, + "loss": 0.3171, + "step": 66679 + }, + { + "epoch": 2.56, + "learning_rate": 7.277507053090536e-06, + "loss": 0.1684, + "step": 66680 + }, + { + "epoch": 2.56, + "learning_rate": 7.276865863041805e-06, + "loss": 0.186, + "step": 66681 + }, + { + "epoch": 2.56, + "learning_rate": 7.276224672993076e-06, + "loss": 0.1251, + "step": 66682 + }, + { + "epoch": 2.56, + "learning_rate": 7.2755834829443455e-06, + "loss": 0.1068, + "step": 66683 + }, + { + "epoch": 2.56, + "learning_rate": 7.274942292895615e-06, + "loss": 0.1961, + "step": 66684 + }, + { + "epoch": 2.56, + "learning_rate": 7.274301102846884e-06, + "loss": 0.1989, + "step": 66685 + }, + { + "epoch": 2.56, + "learning_rate": 7.273659912798153e-06, + "loss": 0.1062, + "step": 66686 + }, + { + "epoch": 2.56, + "learning_rate": 7.273018722749424e-06, + "loss": 0.1103, + "step": 66687 + }, + { + "epoch": 2.56, + "learning_rate": 7.272377532700693e-06, + "loss": 0.3345, + "step": 66688 + }, + { + "epoch": 2.56, + "learning_rate": 7.271736342651962e-06, + "loss": 0.1357, + "step": 66689 + }, + { + "epoch": 2.56, + "learning_rate": 7.271095152603231e-06, + "loss": 0.2476, + "step": 66690 + }, + { + "epoch": 2.56, + "learning_rate": 7.270453962554502e-06, + "loss": 0.2174, + "step": 66691 + }, + { + "epoch": 2.56, + "learning_rate": 7.269812772505771e-06, + "loss": 0.2599, + "step": 66692 + }, + { + "epoch": 2.56, + "learning_rate": 7.269171582457041e-06, + "loss": 0.1516, + "step": 66693 + }, + { + "epoch": 2.56, + "learning_rate": 7.26853039240831e-06, + "loss": 0.2142, + "step": 66694 + }, + { + "epoch": 2.56, + "learning_rate": 7.267889202359579e-06, + "loss": 0.2711, + "step": 66695 + }, + { + "epoch": 2.56, + "learning_rate": 7.26724801231085e-06, + "loss": 0.205, + "step": 66696 + }, + { + "epoch": 2.56, + "learning_rate": 7.266606822262119e-06, + "loss": 0.218, + "step": 66697 + }, + { + "epoch": 2.56, + "learning_rate": 7.265965632213388e-06, + "loss": 0.2235, + "step": 66698 + }, + { + "epoch": 2.56, + "learning_rate": 7.265324442164657e-06, + "loss": 0.2748, + "step": 66699 + }, + { + "epoch": 2.56, + "learning_rate": 7.264683252115928e-06, + "loss": 0.1722, + "step": 66700 + }, + { + "epoch": 2.56, + "learning_rate": 7.264042062067197e-06, + "loss": 0.2325, + "step": 66701 + }, + { + "epoch": 2.56, + "learning_rate": 7.2634008720184666e-06, + "loss": 0.2829, + "step": 66702 + }, + { + "epoch": 2.56, + "learning_rate": 7.262759681969736e-06, + "loss": 0.41, + "step": 66703 + }, + { + "epoch": 2.56, + "learning_rate": 7.262118491921007e-06, + "loss": 0.1434, + "step": 66704 + }, + { + "epoch": 2.56, + "learning_rate": 7.261477301872276e-06, + "loss": 0.1667, + "step": 66705 + }, + { + "epoch": 2.56, + "learning_rate": 7.260836111823545e-06, + "loss": 0.109, + "step": 66706 + }, + { + "epoch": 2.56, + "learning_rate": 7.260194921774814e-06, + "loss": 0.184, + "step": 66707 + }, + { + "epoch": 2.56, + "learning_rate": 7.259553731726083e-06, + "loss": 0.1608, + "step": 66708 + }, + { + "epoch": 2.56, + "learning_rate": 7.258912541677354e-06, + "loss": 0.1799, + "step": 66709 + }, + { + "epoch": 2.56, + "learning_rate": 7.258271351628623e-06, + "loss": 0.1808, + "step": 66710 + }, + { + "epoch": 2.56, + "learning_rate": 7.2576301615798925e-06, + "loss": 0.1514, + "step": 66711 + }, + { + "epoch": 2.56, + "learning_rate": 7.256988971531162e-06, + "loss": 0.2748, + "step": 66712 + }, + { + "epoch": 2.56, + "learning_rate": 7.256347781482433e-06, + "loss": 0.2686, + "step": 66713 + }, + { + "epoch": 2.56, + "learning_rate": 7.255706591433702e-06, + "loss": 0.2033, + "step": 66714 + }, + { + "epoch": 2.56, + "learning_rate": 7.255065401384971e-06, + "loss": 0.1415, + "step": 66715 + }, + { + "epoch": 2.57, + "learning_rate": 7.25442421133624e-06, + "loss": 0.172, + "step": 66716 + }, + { + "epoch": 2.57, + "learning_rate": 7.253783021287509e-06, + "loss": 0.2746, + "step": 66717 + }, + { + "epoch": 2.57, + "learning_rate": 7.25314183123878e-06, + "loss": 0.2201, + "step": 66718 + }, + { + "epoch": 2.57, + "learning_rate": 7.252500641190049e-06, + "loss": 0.3729, + "step": 66719 + }, + { + "epoch": 2.57, + "learning_rate": 7.2518594511413185e-06, + "loss": 0.1216, + "step": 66720 + }, + { + "epoch": 2.57, + "learning_rate": 7.251218261092588e-06, + "loss": 0.1692, + "step": 66721 + }, + { + "epoch": 2.57, + "learning_rate": 7.2505770710438585e-06, + "loss": 0.2034, + "step": 66722 + }, + { + "epoch": 2.57, + "learning_rate": 7.249935880995128e-06, + "loss": 0.1902, + "step": 66723 + }, + { + "epoch": 2.57, + "learning_rate": 7.249294690946397e-06, + "loss": 0.2784, + "step": 66724 + }, + { + "epoch": 2.57, + "learning_rate": 7.248653500897666e-06, + "loss": 0.1784, + "step": 66725 + }, + { + "epoch": 2.57, + "learning_rate": 7.248012310848935e-06, + "loss": 0.1148, + "step": 66726 + }, + { + "epoch": 2.57, + "learning_rate": 7.247371120800206e-06, + "loss": 0.1338, + "step": 66727 + }, + { + "epoch": 2.57, + "learning_rate": 7.246729930751475e-06, + "loss": 0.1997, + "step": 66728 + }, + { + "epoch": 2.57, + "learning_rate": 7.2460887407027445e-06, + "loss": 0.3295, + "step": 66729 + }, + { + "epoch": 2.57, + "learning_rate": 7.245447550654014e-06, + "loss": 0.2381, + "step": 66730 + }, + { + "epoch": 2.57, + "learning_rate": 7.2448063606052845e-06, + "loss": 0.1904, + "step": 66731 + }, + { + "epoch": 2.57, + "learning_rate": 7.244165170556554e-06, + "loss": 0.2844, + "step": 66732 + }, + { + "epoch": 2.57, + "learning_rate": 7.243523980507823e-06, + "loss": 0.2424, + "step": 66733 + }, + { + "epoch": 2.57, + "learning_rate": 7.242882790459092e-06, + "loss": 0.264, + "step": 66734 + }, + { + "epoch": 2.57, + "learning_rate": 7.242241600410361e-06, + "loss": 0.3119, + "step": 66735 + }, + { + "epoch": 2.57, + "learning_rate": 7.241600410361632e-06, + "loss": 0.243, + "step": 66736 + }, + { + "epoch": 2.57, + "learning_rate": 7.240959220312901e-06, + "loss": 0.2607, + "step": 66737 + }, + { + "epoch": 2.57, + "learning_rate": 7.2403180302641704e-06, + "loss": 0.1659, + "step": 66738 + }, + { + "epoch": 2.57, + "learning_rate": 7.23967684021544e-06, + "loss": 0.1779, + "step": 66739 + }, + { + "epoch": 2.57, + "learning_rate": 7.2390356501667105e-06, + "loss": 0.1136, + "step": 66740 + }, + { + "epoch": 2.57, + "learning_rate": 7.23839446011798e-06, + "loss": 0.3556, + "step": 66741 + }, + { + "epoch": 2.57, + "learning_rate": 7.237753270069249e-06, + "loss": 0.3026, + "step": 66742 + }, + { + "epoch": 2.57, + "learning_rate": 7.237112080020518e-06, + "loss": 0.1829, + "step": 66743 + }, + { + "epoch": 2.57, + "learning_rate": 7.236470889971787e-06, + "loss": 0.1812, + "step": 66744 + }, + { + "epoch": 2.57, + "learning_rate": 7.235829699923058e-06, + "loss": 0.1786, + "step": 66745 + }, + { + "epoch": 2.57, + "learning_rate": 7.235188509874327e-06, + "loss": 0.1782, + "step": 66746 + }, + { + "epoch": 2.57, + "learning_rate": 7.234547319825596e-06, + "loss": 0.1966, + "step": 66747 + }, + { + "epoch": 2.57, + "learning_rate": 7.233906129776866e-06, + "loss": 0.2071, + "step": 66748 + }, + { + "epoch": 2.57, + "learning_rate": 7.2332649397281365e-06, + "loss": 0.1722, + "step": 66749 + }, + { + "epoch": 2.57, + "learning_rate": 7.232623749679406e-06, + "loss": 0.2462, + "step": 66750 + }, + { + "epoch": 2.57, + "learning_rate": 7.231982559630675e-06, + "loss": 0.2188, + "step": 66751 + }, + { + "epoch": 2.57, + "learning_rate": 7.231341369581944e-06, + "loss": 0.2054, + "step": 66752 + }, + { + "epoch": 2.57, + "learning_rate": 7.230700179533215e-06, + "loss": 0.1618, + "step": 66753 + }, + { + "epoch": 2.57, + "learning_rate": 7.230058989484484e-06, + "loss": 0.2196, + "step": 66754 + }, + { + "epoch": 2.57, + "learning_rate": 7.229417799435753e-06, + "loss": 0.1607, + "step": 66755 + }, + { + "epoch": 2.57, + "learning_rate": 7.228776609387022e-06, + "loss": 0.1794, + "step": 66756 + }, + { + "epoch": 2.57, + "learning_rate": 7.2281354193382915e-06, + "loss": 0.1707, + "step": 66757 + }, + { + "epoch": 2.57, + "learning_rate": 7.227494229289562e-06, + "loss": 0.1891, + "step": 66758 + }, + { + "epoch": 2.57, + "learning_rate": 7.226853039240832e-06, + "loss": 0.247, + "step": 66759 + }, + { + "epoch": 2.57, + "learning_rate": 7.226211849192101e-06, + "loss": 0.19, + "step": 66760 + }, + { + "epoch": 2.57, + "learning_rate": 7.22557065914337e-06, + "loss": 0.2425, + "step": 66761 + }, + { + "epoch": 2.57, + "learning_rate": 7.224929469094641e-06, + "loss": 0.1588, + "step": 66762 + }, + { + "epoch": 2.57, + "learning_rate": 7.22428827904591e-06, + "loss": 0.4062, + "step": 66763 + }, + { + "epoch": 2.57, + "learning_rate": 7.223647088997179e-06, + "loss": 0.2621, + "step": 66764 + }, + { + "epoch": 2.57, + "learning_rate": 7.223005898948448e-06, + "loss": 0.2263, + "step": 66765 + }, + { + "epoch": 2.57, + "learning_rate": 7.2223647088997175e-06, + "loss": 0.2247, + "step": 66766 + }, + { + "epoch": 2.57, + "learning_rate": 7.221723518850988e-06, + "loss": 0.2461, + "step": 66767 + }, + { + "epoch": 2.57, + "learning_rate": 7.2210823288022576e-06, + "loss": 0.1126, + "step": 66768 + }, + { + "epoch": 2.57, + "learning_rate": 7.220441138753527e-06, + "loss": 0.2536, + "step": 66769 + }, + { + "epoch": 2.57, + "learning_rate": 7.219799948704796e-06, + "loss": 0.1942, + "step": 66770 + }, + { + "epoch": 2.57, + "learning_rate": 7.219158758656067e-06, + "loss": 0.283, + "step": 66771 + }, + { + "epoch": 2.57, + "learning_rate": 7.218517568607336e-06, + "loss": 0.1059, + "step": 66772 + }, + { + "epoch": 2.57, + "learning_rate": 7.217876378558605e-06, + "loss": 0.2708, + "step": 66773 + }, + { + "epoch": 2.57, + "learning_rate": 7.217235188509874e-06, + "loss": 0.2736, + "step": 66774 + }, + { + "epoch": 2.57, + "learning_rate": 7.2165939984611435e-06, + "loss": 0.2342, + "step": 66775 + }, + { + "epoch": 2.57, + "learning_rate": 7.215952808412414e-06, + "loss": 0.368, + "step": 66776 + }, + { + "epoch": 2.57, + "learning_rate": 7.2153116183636835e-06, + "loss": 0.2023, + "step": 66777 + }, + { + "epoch": 2.57, + "learning_rate": 7.214670428314953e-06, + "loss": 0.2754, + "step": 66778 + }, + { + "epoch": 2.57, + "learning_rate": 7.214029238266222e-06, + "loss": 0.1286, + "step": 66779 + }, + { + "epoch": 2.57, + "learning_rate": 7.213388048217493e-06, + "loss": 0.2474, + "step": 66780 + }, + { + "epoch": 2.57, + "learning_rate": 7.212746858168762e-06, + "loss": 0.2176, + "step": 66781 + }, + { + "epoch": 2.57, + "learning_rate": 7.212105668120031e-06, + "loss": 0.0927, + "step": 66782 + }, + { + "epoch": 2.57, + "learning_rate": 7.2114644780713e-06, + "loss": 0.1373, + "step": 66783 + }, + { + "epoch": 2.57, + "learning_rate": 7.2108232880225695e-06, + "loss": 0.191, + "step": 66784 + }, + { + "epoch": 2.57, + "learning_rate": 7.21018209797384e-06, + "loss": 0.2504, + "step": 66785 + }, + { + "epoch": 2.57, + "learning_rate": 7.2095409079251095e-06, + "loss": 0.2063, + "step": 66786 + }, + { + "epoch": 2.57, + "learning_rate": 7.208899717876379e-06, + "loss": 0.1774, + "step": 66787 + }, + { + "epoch": 2.57, + "learning_rate": 7.208258527827648e-06, + "loss": 0.1632, + "step": 66788 + }, + { + "epoch": 2.57, + "learning_rate": 7.207617337778919e-06, + "loss": 0.2379, + "step": 66789 + }, + { + "epoch": 2.57, + "learning_rate": 7.206976147730188e-06, + "loss": 0.1806, + "step": 66790 + }, + { + "epoch": 2.57, + "learning_rate": 7.206334957681457e-06, + "loss": 0.2171, + "step": 66791 + }, + { + "epoch": 2.57, + "learning_rate": 7.205693767632726e-06, + "loss": 0.2176, + "step": 66792 + }, + { + "epoch": 2.57, + "learning_rate": 7.205052577583997e-06, + "loss": 0.1539, + "step": 66793 + }, + { + "epoch": 2.57, + "learning_rate": 7.204411387535266e-06, + "loss": 0.1671, + "step": 66794 + }, + { + "epoch": 2.57, + "learning_rate": 7.2037701974865355e-06, + "loss": 0.1756, + "step": 66795 + }, + { + "epoch": 2.57, + "learning_rate": 7.203129007437805e-06, + "loss": 0.2855, + "step": 66796 + }, + { + "epoch": 2.57, + "learning_rate": 7.202487817389074e-06, + "loss": 0.2745, + "step": 66797 + }, + { + "epoch": 2.57, + "learning_rate": 7.201846627340345e-06, + "loss": 0.1726, + "step": 66798 + }, + { + "epoch": 2.57, + "learning_rate": 7.201205437291614e-06, + "loss": 0.2459, + "step": 66799 + }, + { + "epoch": 2.57, + "learning_rate": 7.200564247242883e-06, + "loss": 0.4328, + "step": 66800 + }, + { + "epoch": 2.57, + "learning_rate": 7.199923057194152e-06, + "loss": 0.3027, + "step": 66801 + }, + { + "epoch": 2.57, + "learning_rate": 7.199281867145423e-06, + "loss": 0.2791, + "step": 66802 + }, + { + "epoch": 2.57, + "learning_rate": 7.198640677096692e-06, + "loss": 0.4792, + "step": 66803 + }, + { + "epoch": 2.57, + "learning_rate": 7.1979994870479614e-06, + "loss": 0.2341, + "step": 66804 + }, + { + "epoch": 2.57, + "learning_rate": 7.197358296999231e-06, + "loss": 0.3102, + "step": 66805 + }, + { + "epoch": 2.57, + "learning_rate": 7.1967171069505e-06, + "loss": 0.2856, + "step": 66806 + }, + { + "epoch": 2.57, + "learning_rate": 7.196075916901771e-06, + "loss": 0.2533, + "step": 66807 + }, + { + "epoch": 2.57, + "learning_rate": 7.19543472685304e-06, + "loss": 0.1939, + "step": 66808 + }, + { + "epoch": 2.57, + "learning_rate": 7.194793536804309e-06, + "loss": 0.2321, + "step": 66809 + }, + { + "epoch": 2.57, + "learning_rate": 7.194152346755578e-06, + "loss": 0.1146, + "step": 66810 + }, + { + "epoch": 2.57, + "learning_rate": 7.193511156706849e-06, + "loss": 0.1894, + "step": 66811 + }, + { + "epoch": 2.57, + "learning_rate": 7.192869966658118e-06, + "loss": 0.1889, + "step": 66812 + }, + { + "epoch": 2.57, + "learning_rate": 7.192228776609387e-06, + "loss": 0.1567, + "step": 66813 + }, + { + "epoch": 2.57, + "learning_rate": 7.1915875865606566e-06, + "loss": 0.2459, + "step": 66814 + }, + { + "epoch": 2.57, + "learning_rate": 7.190946396511926e-06, + "loss": 0.1176, + "step": 66815 + }, + { + "epoch": 2.57, + "learning_rate": 7.190305206463197e-06, + "loss": 0.2447, + "step": 66816 + }, + { + "epoch": 2.57, + "learning_rate": 7.189664016414466e-06, + "loss": 0.232, + "step": 66817 + }, + { + "epoch": 2.57, + "learning_rate": 7.189022826365735e-06, + "loss": 0.1525, + "step": 66818 + }, + { + "epoch": 2.57, + "learning_rate": 7.188381636317004e-06, + "loss": 0.1575, + "step": 66819 + }, + { + "epoch": 2.57, + "learning_rate": 7.187740446268275e-06, + "loss": 0.2704, + "step": 66820 + }, + { + "epoch": 2.57, + "learning_rate": 7.187099256219544e-06, + "loss": 0.251, + "step": 66821 + }, + { + "epoch": 2.57, + "learning_rate": 7.186458066170813e-06, + "loss": 0.2418, + "step": 66822 + }, + { + "epoch": 2.57, + "learning_rate": 7.1858168761220825e-06, + "loss": 0.249, + "step": 66823 + }, + { + "epoch": 2.57, + "learning_rate": 7.185175686073352e-06, + "loss": 0.2325, + "step": 66824 + }, + { + "epoch": 2.57, + "learning_rate": 7.184534496024623e-06, + "loss": 0.1656, + "step": 66825 + }, + { + "epoch": 2.57, + "learning_rate": 7.183893305975892e-06, + "loss": 0.217, + "step": 66826 + }, + { + "epoch": 2.57, + "learning_rate": 7.183252115927161e-06, + "loss": 0.3052, + "step": 66827 + }, + { + "epoch": 2.57, + "learning_rate": 7.18261092587843e-06, + "loss": 0.2005, + "step": 66828 + }, + { + "epoch": 2.57, + "learning_rate": 7.181969735829701e-06, + "loss": 0.2357, + "step": 66829 + }, + { + "epoch": 2.57, + "learning_rate": 7.18132854578097e-06, + "loss": 0.1658, + "step": 66830 + }, + { + "epoch": 2.57, + "learning_rate": 7.180687355732239e-06, + "loss": 0.1977, + "step": 66831 + }, + { + "epoch": 2.57, + "learning_rate": 7.1800461656835085e-06, + "loss": 0.2554, + "step": 66832 + }, + { + "epoch": 2.57, + "learning_rate": 7.179404975634778e-06, + "loss": 0.3974, + "step": 66833 + }, + { + "epoch": 2.57, + "learning_rate": 7.1787637855860486e-06, + "loss": 0.1776, + "step": 66834 + }, + { + "epoch": 2.57, + "learning_rate": 7.178122595537318e-06, + "loss": 0.204, + "step": 66835 + }, + { + "epoch": 2.57, + "learning_rate": 7.177481405488587e-06, + "loss": 0.1838, + "step": 66836 + }, + { + "epoch": 2.57, + "learning_rate": 7.176840215439856e-06, + "loss": 0.1652, + "step": 66837 + }, + { + "epoch": 2.57, + "learning_rate": 7.176199025391127e-06, + "loss": 0.2446, + "step": 66838 + }, + { + "epoch": 2.57, + "learning_rate": 7.175557835342396e-06, + "loss": 0.2404, + "step": 66839 + }, + { + "epoch": 2.57, + "learning_rate": 7.174916645293665e-06, + "loss": 0.2478, + "step": 66840 + }, + { + "epoch": 2.57, + "learning_rate": 7.1742754552449345e-06, + "loss": 0.3016, + "step": 66841 + }, + { + "epoch": 2.57, + "learning_rate": 7.173634265196205e-06, + "loss": 0.1389, + "step": 66842 + }, + { + "epoch": 2.57, + "learning_rate": 7.1729930751474745e-06, + "loss": 0.1859, + "step": 66843 + }, + { + "epoch": 2.57, + "learning_rate": 7.172351885098744e-06, + "loss": 0.1972, + "step": 66844 + }, + { + "epoch": 2.57, + "learning_rate": 7.171710695050013e-06, + "loss": 0.2874, + "step": 66845 + }, + { + "epoch": 2.57, + "learning_rate": 7.171069505001282e-06, + "loss": 0.2718, + "step": 66846 + }, + { + "epoch": 2.57, + "learning_rate": 7.170428314952553e-06, + "loss": 0.2513, + "step": 66847 + }, + { + "epoch": 2.57, + "learning_rate": 7.169787124903822e-06, + "loss": 0.1419, + "step": 66848 + }, + { + "epoch": 2.57, + "learning_rate": 7.169145934855091e-06, + "loss": 0.3379, + "step": 66849 + }, + { + "epoch": 2.57, + "learning_rate": 7.1685047448063604e-06, + "loss": 0.2304, + "step": 66850 + }, + { + "epoch": 2.57, + "learning_rate": 7.167863554757631e-06, + "loss": 0.3103, + "step": 66851 + }, + { + "epoch": 2.57, + "learning_rate": 7.1672223647089005e-06, + "loss": 0.1811, + "step": 66852 + }, + { + "epoch": 2.57, + "learning_rate": 7.16658117466017e-06, + "loss": 0.2449, + "step": 66853 + }, + { + "epoch": 2.57, + "learning_rate": 7.165939984611439e-06, + "loss": 0.279, + "step": 66854 + }, + { + "epoch": 2.57, + "learning_rate": 7.165298794562708e-06, + "loss": 0.3412, + "step": 66855 + }, + { + "epoch": 2.57, + "learning_rate": 7.164657604513979e-06, + "loss": 0.2147, + "step": 66856 + }, + { + "epoch": 2.57, + "learning_rate": 7.164016414465248e-06, + "loss": 0.1573, + "step": 66857 + }, + { + "epoch": 2.57, + "learning_rate": 7.163375224416517e-06, + "loss": 0.1602, + "step": 66858 + }, + { + "epoch": 2.57, + "learning_rate": 7.162734034367786e-06, + "loss": 0.1394, + "step": 66859 + }, + { + "epoch": 2.57, + "learning_rate": 7.162092844319057e-06, + "loss": 0.2133, + "step": 66860 + }, + { + "epoch": 2.57, + "learning_rate": 7.1614516542703265e-06, + "loss": 0.2103, + "step": 66861 + }, + { + "epoch": 2.57, + "learning_rate": 7.160810464221596e-06, + "loss": 0.1855, + "step": 66862 + }, + { + "epoch": 2.57, + "learning_rate": 7.160169274172865e-06, + "loss": 0.2297, + "step": 66863 + }, + { + "epoch": 2.57, + "learning_rate": 7.159528084124134e-06, + "loss": 0.241, + "step": 66864 + }, + { + "epoch": 2.57, + "learning_rate": 7.158886894075405e-06, + "loss": 0.2009, + "step": 66865 + }, + { + "epoch": 2.57, + "learning_rate": 7.158245704026674e-06, + "loss": 0.2589, + "step": 66866 + }, + { + "epoch": 2.57, + "learning_rate": 7.157604513977943e-06, + "loss": 0.2562, + "step": 66867 + }, + { + "epoch": 2.57, + "learning_rate": 7.156963323929212e-06, + "loss": 0.3824, + "step": 66868 + }, + { + "epoch": 2.57, + "learning_rate": 7.156322133880483e-06, + "loss": 0.2078, + "step": 66869 + }, + { + "epoch": 2.57, + "learning_rate": 7.1556809438317524e-06, + "loss": 0.3612, + "step": 66870 + }, + { + "epoch": 2.57, + "learning_rate": 7.155039753783022e-06, + "loss": 0.2105, + "step": 66871 + }, + { + "epoch": 2.57, + "learning_rate": 7.154398563734291e-06, + "loss": 0.2627, + "step": 66872 + }, + { + "epoch": 2.57, + "learning_rate": 7.15375737368556e-06, + "loss": 0.3122, + "step": 66873 + }, + { + "epoch": 2.57, + "learning_rate": 7.153116183636831e-06, + "loss": 0.2093, + "step": 66874 + }, + { + "epoch": 2.57, + "learning_rate": 7.1524749935881e-06, + "loss": 0.1898, + "step": 66875 + }, + { + "epoch": 2.57, + "learning_rate": 7.151833803539369e-06, + "loss": 0.2122, + "step": 66876 + }, + { + "epoch": 2.57, + "learning_rate": 7.151192613490638e-06, + "loss": 0.1736, + "step": 66877 + }, + { + "epoch": 2.57, + "learning_rate": 7.150551423441909e-06, + "loss": 0.2666, + "step": 66878 + }, + { + "epoch": 2.57, + "learning_rate": 7.149910233393178e-06, + "loss": 0.2233, + "step": 66879 + }, + { + "epoch": 2.57, + "learning_rate": 7.1492690433444476e-06, + "loss": 0.1771, + "step": 66880 + }, + { + "epoch": 2.57, + "learning_rate": 7.148627853295717e-06, + "loss": 0.2983, + "step": 66881 + }, + { + "epoch": 2.57, + "learning_rate": 7.147986663246986e-06, + "loss": 0.1591, + "step": 66882 + }, + { + "epoch": 2.57, + "learning_rate": 7.147345473198257e-06, + "loss": 0.1699, + "step": 66883 + }, + { + "epoch": 2.57, + "learning_rate": 7.146704283149526e-06, + "loss": 0.2627, + "step": 66884 + }, + { + "epoch": 2.57, + "learning_rate": 7.146063093100795e-06, + "loss": 0.2121, + "step": 66885 + }, + { + "epoch": 2.57, + "learning_rate": 7.145421903052064e-06, + "loss": 0.1643, + "step": 66886 + }, + { + "epoch": 2.57, + "learning_rate": 7.144780713003335e-06, + "loss": 0.3309, + "step": 66887 + }, + { + "epoch": 2.57, + "learning_rate": 7.144139522954604e-06, + "loss": 0.2303, + "step": 66888 + }, + { + "epoch": 2.57, + "learning_rate": 7.1434983329058735e-06, + "loss": 0.2422, + "step": 66889 + }, + { + "epoch": 2.57, + "learning_rate": 7.142857142857143e-06, + "loss": 0.3588, + "step": 66890 + }, + { + "epoch": 2.57, + "learning_rate": 7.142215952808414e-06, + "loss": 0.2115, + "step": 66891 + }, + { + "epoch": 2.57, + "learning_rate": 7.141574762759683e-06, + "loss": 0.1538, + "step": 66892 + }, + { + "epoch": 2.57, + "learning_rate": 7.140933572710952e-06, + "loss": 0.1863, + "step": 66893 + }, + { + "epoch": 2.57, + "learning_rate": 7.140292382662221e-06, + "loss": 0.1122, + "step": 66894 + }, + { + "epoch": 2.57, + "learning_rate": 7.13965119261349e-06, + "loss": 0.2449, + "step": 66895 + }, + { + "epoch": 2.57, + "learning_rate": 7.139010002564761e-06, + "loss": 0.1902, + "step": 66896 + }, + { + "epoch": 2.57, + "learning_rate": 7.13836881251603e-06, + "loss": 0.253, + "step": 66897 + }, + { + "epoch": 2.57, + "learning_rate": 7.1377276224672995e-06, + "loss": 0.2446, + "step": 66898 + }, + { + "epoch": 2.57, + "learning_rate": 7.137086432418569e-06, + "loss": 0.1497, + "step": 66899 + }, + { + "epoch": 2.57, + "learning_rate": 7.1364452423698396e-06, + "loss": 0.2464, + "step": 66900 + }, + { + "epoch": 2.57, + "learning_rate": 7.135804052321109e-06, + "loss": 0.1734, + "step": 66901 + }, + { + "epoch": 2.57, + "learning_rate": 7.135162862272378e-06, + "loss": 0.3146, + "step": 66902 + }, + { + "epoch": 2.57, + "learning_rate": 7.134521672223647e-06, + "loss": 0.2893, + "step": 66903 + }, + { + "epoch": 2.57, + "learning_rate": 7.133880482174916e-06, + "loss": 0.198, + "step": 66904 + }, + { + "epoch": 2.57, + "learning_rate": 7.133239292126187e-06, + "loss": 0.1855, + "step": 66905 + }, + { + "epoch": 2.57, + "learning_rate": 7.132598102077456e-06, + "loss": 0.2504, + "step": 66906 + }, + { + "epoch": 2.57, + "learning_rate": 7.1319569120287255e-06, + "loss": 0.2849, + "step": 66907 + }, + { + "epoch": 2.57, + "learning_rate": 7.131315721979995e-06, + "loss": 0.205, + "step": 66908 + }, + { + "epoch": 2.57, + "learning_rate": 7.1306745319312655e-06, + "loss": 0.2745, + "step": 66909 + }, + { + "epoch": 2.57, + "learning_rate": 7.130033341882535e-06, + "loss": 0.2374, + "step": 66910 + }, + { + "epoch": 2.57, + "learning_rate": 7.129392151833804e-06, + "loss": 0.1501, + "step": 66911 + }, + { + "epoch": 2.57, + "learning_rate": 7.128750961785073e-06, + "loss": 0.2425, + "step": 66912 + }, + { + "epoch": 2.57, + "learning_rate": 7.128109771736342e-06, + "loss": 0.2249, + "step": 66913 + }, + { + "epoch": 2.57, + "learning_rate": 7.127468581687613e-06, + "loss": 0.2556, + "step": 66914 + }, + { + "epoch": 2.57, + "learning_rate": 7.126827391638882e-06, + "loss": 0.3036, + "step": 66915 + }, + { + "epoch": 2.57, + "learning_rate": 7.1261862015901514e-06, + "loss": 0.2519, + "step": 66916 + }, + { + "epoch": 2.57, + "learning_rate": 7.125545011541421e-06, + "loss": 0.1522, + "step": 66917 + }, + { + "epoch": 2.57, + "learning_rate": 7.1249038214926915e-06, + "loss": 0.358, + "step": 66918 + }, + { + "epoch": 2.57, + "learning_rate": 7.124262631443961e-06, + "loss": 0.1539, + "step": 66919 + }, + { + "epoch": 2.57, + "learning_rate": 7.12362144139523e-06, + "loss": 0.2328, + "step": 66920 + }, + { + "epoch": 2.57, + "learning_rate": 7.122980251346499e-06, + "loss": 0.1853, + "step": 66921 + }, + { + "epoch": 2.57, + "learning_rate": 7.122339061297768e-06, + "loss": 0.1774, + "step": 66922 + }, + { + "epoch": 2.57, + "learning_rate": 7.121697871249039e-06, + "loss": 0.2146, + "step": 66923 + }, + { + "epoch": 2.57, + "learning_rate": 7.121056681200308e-06, + "loss": 0.3064, + "step": 66924 + }, + { + "epoch": 2.57, + "learning_rate": 7.120415491151577e-06, + "loss": 0.2427, + "step": 66925 + }, + { + "epoch": 2.57, + "learning_rate": 7.119774301102847e-06, + "loss": 0.1512, + "step": 66926 + }, + { + "epoch": 2.57, + "learning_rate": 7.1191331110541175e-06, + "loss": 0.1851, + "step": 66927 + }, + { + "epoch": 2.57, + "learning_rate": 7.118491921005387e-06, + "loss": 0.0856, + "step": 66928 + }, + { + "epoch": 2.57, + "learning_rate": 7.117850730956656e-06, + "loss": 0.2535, + "step": 66929 + }, + { + "epoch": 2.57, + "learning_rate": 7.117209540907925e-06, + "loss": 0.1193, + "step": 66930 + }, + { + "epoch": 2.57, + "learning_rate": 7.116568350859194e-06, + "loss": 0.2273, + "step": 66931 + }, + { + "epoch": 2.57, + "learning_rate": 7.115927160810465e-06, + "loss": 0.2111, + "step": 66932 + }, + { + "epoch": 2.57, + "learning_rate": 7.115285970761734e-06, + "loss": 0.1302, + "step": 66933 + }, + { + "epoch": 2.57, + "learning_rate": 7.114644780713003e-06, + "loss": 0.2095, + "step": 66934 + }, + { + "epoch": 2.57, + "learning_rate": 7.1140035906642726e-06, + "loss": 0.2304, + "step": 66935 + }, + { + "epoch": 2.57, + "learning_rate": 7.1133624006155434e-06, + "loss": 0.1631, + "step": 66936 + }, + { + "epoch": 2.57, + "learning_rate": 7.112721210566813e-06, + "loss": 0.1981, + "step": 66937 + }, + { + "epoch": 2.57, + "learning_rate": 7.112080020518082e-06, + "loss": 0.1919, + "step": 66938 + }, + { + "epoch": 2.57, + "learning_rate": 7.111438830469351e-06, + "loss": 0.2289, + "step": 66939 + }, + { + "epoch": 2.57, + "learning_rate": 7.110797640420622e-06, + "loss": 0.3836, + "step": 66940 + }, + { + "epoch": 2.57, + "learning_rate": 7.110156450371891e-06, + "loss": 0.2341, + "step": 66941 + }, + { + "epoch": 2.57, + "learning_rate": 7.10951526032316e-06, + "loss": 0.1944, + "step": 66942 + }, + { + "epoch": 2.57, + "learning_rate": 7.108874070274429e-06, + "loss": 0.416, + "step": 66943 + }, + { + "epoch": 2.57, + "learning_rate": 7.1082328802256985e-06, + "loss": 0.2884, + "step": 66944 + }, + { + "epoch": 2.57, + "learning_rate": 7.107591690176969e-06, + "loss": 0.1331, + "step": 66945 + }, + { + "epoch": 2.57, + "learning_rate": 7.1069505001282386e-06, + "loss": 0.0956, + "step": 66946 + }, + { + "epoch": 2.57, + "learning_rate": 7.106309310079508e-06, + "loss": 0.1557, + "step": 66947 + }, + { + "epoch": 2.57, + "learning_rate": 7.105668120030777e-06, + "loss": 0.3231, + "step": 66948 + }, + { + "epoch": 2.57, + "learning_rate": 7.105026929982048e-06, + "loss": 0.1535, + "step": 66949 + }, + { + "epoch": 2.57, + "learning_rate": 7.104385739933317e-06, + "loss": 0.1812, + "step": 66950 + }, + { + "epoch": 2.57, + "learning_rate": 7.103744549884586e-06, + "loss": 0.1705, + "step": 66951 + }, + { + "epoch": 2.57, + "learning_rate": 7.103103359835855e-06, + "loss": 0.1379, + "step": 66952 + }, + { + "epoch": 2.57, + "learning_rate": 7.1024621697871245e-06, + "loss": 0.1755, + "step": 66953 + }, + { + "epoch": 2.57, + "learning_rate": 7.101820979738395e-06, + "loss": 0.149, + "step": 66954 + }, + { + "epoch": 2.57, + "learning_rate": 7.1011797896896645e-06, + "loss": 0.2951, + "step": 66955 + }, + { + "epoch": 2.57, + "learning_rate": 7.100538599640934e-06, + "loss": 0.1813, + "step": 66956 + }, + { + "epoch": 2.57, + "learning_rate": 7.099897409592203e-06, + "loss": 0.242, + "step": 66957 + }, + { + "epoch": 2.57, + "learning_rate": 7.099256219543474e-06, + "loss": 0.1373, + "step": 66958 + }, + { + "epoch": 2.57, + "learning_rate": 7.098615029494743e-06, + "loss": 0.1292, + "step": 66959 + }, + { + "epoch": 2.57, + "learning_rate": 7.097973839446012e-06, + "loss": 0.239, + "step": 66960 + }, + { + "epoch": 2.57, + "learning_rate": 7.097332649397281e-06, + "loss": 0.2282, + "step": 66961 + }, + { + "epoch": 2.57, + "learning_rate": 7.0966914593485505e-06, + "loss": 0.2513, + "step": 66962 + }, + { + "epoch": 2.57, + "learning_rate": 7.096050269299821e-06, + "loss": 0.1934, + "step": 66963 + }, + { + "epoch": 2.57, + "learning_rate": 7.0954090792510905e-06, + "loss": 0.1654, + "step": 66964 + }, + { + "epoch": 2.57, + "learning_rate": 7.09476788920236e-06, + "loss": 0.0852, + "step": 66965 + }, + { + "epoch": 2.57, + "learning_rate": 7.094126699153629e-06, + "loss": 0.1506, + "step": 66966 + }, + { + "epoch": 2.57, + "learning_rate": 7.0934855091049e-06, + "loss": 0.1192, + "step": 66967 + }, + { + "epoch": 2.57, + "learning_rate": 7.092844319056169e-06, + "loss": 0.1163, + "step": 66968 + }, + { + "epoch": 2.57, + "learning_rate": 7.092203129007438e-06, + "loss": 0.2344, + "step": 66969 + }, + { + "epoch": 2.57, + "learning_rate": 7.091561938958707e-06, + "loss": 0.1437, + "step": 66970 + }, + { + "epoch": 2.57, + "learning_rate": 7.0909207489099764e-06, + "loss": 0.1328, + "step": 66971 + }, + { + "epoch": 2.57, + "learning_rate": 7.090279558861247e-06, + "loss": 0.2097, + "step": 66972 + }, + { + "epoch": 2.57, + "learning_rate": 7.0896383688125165e-06, + "loss": 0.1926, + "step": 66973 + }, + { + "epoch": 2.57, + "learning_rate": 7.088997178763786e-06, + "loss": 0.2455, + "step": 66974 + }, + { + "epoch": 2.57, + "learning_rate": 7.088355988715055e-06, + "loss": 0.2742, + "step": 66975 + }, + { + "epoch": 2.58, + "learning_rate": 7.087714798666326e-06, + "loss": 0.249, + "step": 66976 + }, + { + "epoch": 2.58, + "learning_rate": 7.087073608617595e-06, + "loss": 0.2059, + "step": 66977 + }, + { + "epoch": 2.58, + "learning_rate": 7.086432418568864e-06, + "loss": 0.2178, + "step": 66978 + }, + { + "epoch": 2.58, + "learning_rate": 7.085791228520133e-06, + "loss": 0.331, + "step": 66979 + }, + { + "epoch": 2.58, + "learning_rate": 7.085150038471402e-06, + "loss": 0.2516, + "step": 66980 + }, + { + "epoch": 2.58, + "learning_rate": 7.084508848422673e-06, + "loss": 0.1986, + "step": 66981 + }, + { + "epoch": 2.58, + "learning_rate": 7.0838676583739424e-06, + "loss": 0.3026, + "step": 66982 + }, + { + "epoch": 2.58, + "learning_rate": 7.083226468325212e-06, + "loss": 0.1739, + "step": 66983 + }, + { + "epoch": 2.58, + "learning_rate": 7.082585278276481e-06, + "loss": 0.1638, + "step": 66984 + }, + { + "epoch": 2.58, + "learning_rate": 7.081944088227752e-06, + "loss": 0.2738, + "step": 66985 + }, + { + "epoch": 2.58, + "learning_rate": 7.081302898179021e-06, + "loss": 0.2429, + "step": 66986 + }, + { + "epoch": 2.58, + "learning_rate": 7.08066170813029e-06, + "loss": 0.2843, + "step": 66987 + }, + { + "epoch": 2.58, + "learning_rate": 7.080020518081559e-06, + "loss": 0.2888, + "step": 66988 + }, + { + "epoch": 2.58, + "learning_rate": 7.07937932803283e-06, + "loss": 0.28, + "step": 66989 + }, + { + "epoch": 2.58, + "learning_rate": 7.078738137984099e-06, + "loss": 0.2006, + "step": 66990 + }, + { + "epoch": 2.58, + "learning_rate": 7.078096947935368e-06, + "loss": 0.129, + "step": 66991 + }, + { + "epoch": 2.58, + "learning_rate": 7.077455757886638e-06, + "loss": 0.4079, + "step": 66992 + }, + { + "epoch": 2.58, + "learning_rate": 7.076814567837907e-06, + "loss": 0.1217, + "step": 66993 + }, + { + "epoch": 2.58, + "learning_rate": 7.076173377789178e-06, + "loss": 0.1679, + "step": 66994 + }, + { + "epoch": 2.58, + "learning_rate": 7.075532187740447e-06, + "loss": 0.169, + "step": 66995 + }, + { + "epoch": 2.58, + "learning_rate": 7.074890997691716e-06, + "loss": 0.1633, + "step": 66996 + }, + { + "epoch": 2.58, + "learning_rate": 7.074249807642985e-06, + "loss": 0.1157, + "step": 66997 + }, + { + "epoch": 2.58, + "learning_rate": 7.073608617594256e-06, + "loss": 0.177, + "step": 66998 + }, + { + "epoch": 2.58, + "learning_rate": 7.072967427545525e-06, + "loss": 0.2004, + "step": 66999 + }, + { + "epoch": 2.58, + "learning_rate": 7.072326237496794e-06, + "loss": 0.2298, + "step": 67000 + }, + { + "epoch": 2.58, + "learning_rate": 7.0716850474480636e-06, + "loss": 0.2147, + "step": 67001 + }, + { + "epoch": 2.58, + "learning_rate": 7.071043857399333e-06, + "loss": 0.1834, + "step": 67002 + }, + { + "epoch": 2.58, + "learning_rate": 7.070402667350604e-06, + "loss": 0.234, + "step": 67003 + }, + { + "epoch": 2.58, + "learning_rate": 7.069761477301873e-06, + "loss": 0.2134, + "step": 67004 + }, + { + "epoch": 2.58, + "learning_rate": 7.069120287253142e-06, + "loss": 0.234, + "step": 67005 + }, + { + "epoch": 2.58, + "learning_rate": 7.068479097204411e-06, + "loss": 0.2169, + "step": 67006 + }, + { + "epoch": 2.58, + "learning_rate": 7.067837907155682e-06, + "loss": 0.1827, + "step": 67007 + }, + { + "epoch": 2.58, + "learning_rate": 7.067196717106951e-06, + "loss": 0.1209, + "step": 67008 + }, + { + "epoch": 2.58, + "learning_rate": 7.06655552705822e-06, + "loss": 0.3132, + "step": 67009 + }, + { + "epoch": 2.58, + "learning_rate": 7.0659143370094895e-06, + "loss": 0.2425, + "step": 67010 + }, + { + "epoch": 2.58, + "learning_rate": 7.065273146960759e-06, + "loss": 0.2769, + "step": 67011 + }, + { + "epoch": 2.58, + "learning_rate": 7.0646319569120296e-06, + "loss": 0.1559, + "step": 67012 + }, + { + "epoch": 2.58, + "learning_rate": 7.063990766863299e-06, + "loss": 0.2997, + "step": 67013 + }, + { + "epoch": 2.58, + "learning_rate": 7.063349576814568e-06, + "loss": 0.2638, + "step": 67014 + }, + { + "epoch": 2.58, + "learning_rate": 7.062708386765837e-06, + "loss": 0.2644, + "step": 67015 + }, + { + "epoch": 2.58, + "learning_rate": 7.062067196717108e-06, + "loss": 0.2672, + "step": 67016 + }, + { + "epoch": 2.58, + "learning_rate": 7.061426006668377e-06, + "loss": 0.1937, + "step": 67017 + }, + { + "epoch": 2.58, + "learning_rate": 7.060784816619646e-06, + "loss": 0.1554, + "step": 67018 + }, + { + "epoch": 2.58, + "learning_rate": 7.0601436265709155e-06, + "loss": 0.3131, + "step": 67019 + }, + { + "epoch": 2.58, + "learning_rate": 7.059502436522185e-06, + "loss": 0.1558, + "step": 67020 + }, + { + "epoch": 2.58, + "learning_rate": 7.0588612464734555e-06, + "loss": 0.2624, + "step": 67021 + }, + { + "epoch": 2.58, + "learning_rate": 7.058220056424725e-06, + "loss": 0.2707, + "step": 67022 + }, + { + "epoch": 2.58, + "learning_rate": 7.057578866375994e-06, + "loss": 0.4, + "step": 67023 + }, + { + "epoch": 2.58, + "learning_rate": 7.056937676327263e-06, + "loss": 0.2154, + "step": 67024 + }, + { + "epoch": 2.58, + "learning_rate": 7.056296486278534e-06, + "loss": 0.2068, + "step": 67025 + }, + { + "epoch": 2.58, + "learning_rate": 7.055655296229803e-06, + "loss": 0.1124, + "step": 67026 + }, + { + "epoch": 2.58, + "learning_rate": 7.055014106181072e-06, + "loss": 0.2088, + "step": 67027 + }, + { + "epoch": 2.58, + "learning_rate": 7.0543729161323415e-06, + "loss": 0.2613, + "step": 67028 + }, + { + "epoch": 2.58, + "learning_rate": 7.053731726083612e-06, + "loss": 0.1146, + "step": 67029 + }, + { + "epoch": 2.58, + "learning_rate": 7.0530905360348815e-06, + "loss": 0.2674, + "step": 67030 + }, + { + "epoch": 2.58, + "learning_rate": 7.052449345986151e-06, + "loss": 0.1781, + "step": 67031 + }, + { + "epoch": 2.58, + "learning_rate": 7.05180815593742e-06, + "loss": 0.1634, + "step": 67032 + }, + { + "epoch": 2.58, + "learning_rate": 7.051166965888689e-06, + "loss": 0.3836, + "step": 67033 + }, + { + "epoch": 2.58, + "learning_rate": 7.05052577583996e-06, + "loss": 0.155, + "step": 67034 + }, + { + "epoch": 2.58, + "learning_rate": 7.049884585791229e-06, + "loss": 0.2581, + "step": 67035 + }, + { + "epoch": 2.58, + "learning_rate": 7.049243395742498e-06, + "loss": 0.1759, + "step": 67036 + }, + { + "epoch": 2.58, + "learning_rate": 7.048602205693767e-06, + "loss": 0.2322, + "step": 67037 + }, + { + "epoch": 2.58, + "learning_rate": 7.047961015645038e-06, + "loss": 0.3252, + "step": 67038 + }, + { + "epoch": 2.58, + "learning_rate": 7.0473198255963075e-06, + "loss": 0.1777, + "step": 67039 + }, + { + "epoch": 2.58, + "learning_rate": 7.046678635547577e-06, + "loss": 0.1759, + "step": 67040 + }, + { + "epoch": 2.58, + "learning_rate": 7.046037445498846e-06, + "loss": 0.1898, + "step": 67041 + }, + { + "epoch": 2.58, + "learning_rate": 7.045396255450115e-06, + "loss": 0.3692, + "step": 67042 + }, + { + "epoch": 2.58, + "learning_rate": 7.044755065401386e-06, + "loss": 0.1526, + "step": 67043 + }, + { + "epoch": 2.58, + "learning_rate": 7.044113875352655e-06, + "loss": 0.1, + "step": 67044 + }, + { + "epoch": 2.58, + "learning_rate": 7.043472685303924e-06, + "loss": 0.2352, + "step": 67045 + }, + { + "epoch": 2.58, + "learning_rate": 7.042831495255193e-06, + "loss": 0.248, + "step": 67046 + }, + { + "epoch": 2.58, + "learning_rate": 7.042190305206464e-06, + "loss": 0.1979, + "step": 67047 + }, + { + "epoch": 2.58, + "learning_rate": 7.0415491151577334e-06, + "loss": 0.2734, + "step": 67048 + }, + { + "epoch": 2.58, + "learning_rate": 7.040907925109003e-06, + "loss": 0.1801, + "step": 67049 + }, + { + "epoch": 2.58, + "learning_rate": 7.040266735060272e-06, + "loss": 0.2957, + "step": 67050 + }, + { + "epoch": 2.58, + "learning_rate": 7.039625545011541e-06, + "loss": 0.2389, + "step": 67051 + }, + { + "epoch": 2.58, + "learning_rate": 7.038984354962812e-06, + "loss": 0.1649, + "step": 67052 + }, + { + "epoch": 2.58, + "learning_rate": 7.038343164914081e-06, + "loss": 0.1812, + "step": 67053 + }, + { + "epoch": 2.58, + "learning_rate": 7.03770197486535e-06, + "loss": 0.1187, + "step": 67054 + }, + { + "epoch": 2.58, + "learning_rate": 7.037060784816619e-06, + "loss": 0.2549, + "step": 67055 + }, + { + "epoch": 2.58, + "learning_rate": 7.03641959476789e-06, + "loss": 0.2418, + "step": 67056 + }, + { + "epoch": 2.58, + "learning_rate": 7.035778404719159e-06, + "loss": 0.2904, + "step": 67057 + }, + { + "epoch": 2.58, + "learning_rate": 7.035137214670429e-06, + "loss": 0.2714, + "step": 67058 + }, + { + "epoch": 2.58, + "learning_rate": 7.034496024621698e-06, + "loss": 0.1707, + "step": 67059 + }, + { + "epoch": 2.58, + "learning_rate": 7.033854834572967e-06, + "loss": 0.2191, + "step": 67060 + }, + { + "epoch": 2.58, + "learning_rate": 7.033213644524238e-06, + "loss": 0.1602, + "step": 67061 + }, + { + "epoch": 2.58, + "learning_rate": 7.032572454475507e-06, + "loss": 0.2071, + "step": 67062 + }, + { + "epoch": 2.58, + "learning_rate": 7.031931264426776e-06, + "loss": 0.2019, + "step": 67063 + }, + { + "epoch": 2.58, + "learning_rate": 7.031290074378045e-06, + "loss": 0.1681, + "step": 67064 + }, + { + "epoch": 2.58, + "learning_rate": 7.030648884329316e-06, + "loss": 0.2348, + "step": 67065 + }, + { + "epoch": 2.58, + "learning_rate": 7.030007694280585e-06, + "loss": 0.2754, + "step": 67066 + }, + { + "epoch": 2.58, + "learning_rate": 7.0293665042318545e-06, + "loss": 0.2493, + "step": 67067 + }, + { + "epoch": 2.58, + "learning_rate": 7.028725314183124e-06, + "loss": 0.2508, + "step": 67068 + }, + { + "epoch": 2.58, + "learning_rate": 7.028084124134393e-06, + "loss": 0.199, + "step": 67069 + }, + { + "epoch": 2.58, + "learning_rate": 7.027442934085664e-06, + "loss": 0.2803, + "step": 67070 + }, + { + "epoch": 2.58, + "learning_rate": 7.026801744036933e-06, + "loss": 0.1512, + "step": 67071 + }, + { + "epoch": 2.58, + "learning_rate": 7.026160553988202e-06, + "loss": 0.1162, + "step": 67072 + }, + { + "epoch": 2.58, + "learning_rate": 7.025519363939471e-06, + "loss": 0.2774, + "step": 67073 + }, + { + "epoch": 2.58, + "learning_rate": 7.024878173890742e-06, + "loss": 0.1728, + "step": 67074 + }, + { + "epoch": 2.58, + "learning_rate": 7.024236983842011e-06, + "loss": 0.1189, + "step": 67075 + }, + { + "epoch": 2.58, + "learning_rate": 7.0235957937932805e-06, + "loss": 0.2215, + "step": 67076 + }, + { + "epoch": 2.58, + "learning_rate": 7.02295460374455e-06, + "loss": 0.3061, + "step": 67077 + }, + { + "epoch": 2.58, + "learning_rate": 7.0223134136958206e-06, + "loss": 0.1597, + "step": 67078 + }, + { + "epoch": 2.58, + "learning_rate": 7.02167222364709e-06, + "loss": 0.2393, + "step": 67079 + }, + { + "epoch": 2.58, + "learning_rate": 7.021031033598359e-06, + "loss": 0.3003, + "step": 67080 + }, + { + "epoch": 2.58, + "learning_rate": 7.020389843549628e-06, + "loss": 0.2522, + "step": 67081 + }, + { + "epoch": 2.58, + "learning_rate": 7.019748653500897e-06, + "loss": 0.2158, + "step": 67082 + }, + { + "epoch": 2.58, + "learning_rate": 7.019107463452168e-06, + "loss": 0.2538, + "step": 67083 + }, + { + "epoch": 2.58, + "learning_rate": 7.018466273403437e-06, + "loss": 0.2207, + "step": 67084 + }, + { + "epoch": 2.58, + "learning_rate": 7.0178250833547065e-06, + "loss": 0.1992, + "step": 67085 + }, + { + "epoch": 2.58, + "learning_rate": 7.017183893305976e-06, + "loss": 0.257, + "step": 67086 + }, + { + "epoch": 2.58, + "learning_rate": 7.0165427032572465e-06, + "loss": 0.1399, + "step": 67087 + }, + { + "epoch": 2.58, + "learning_rate": 7.015901513208516e-06, + "loss": 0.1506, + "step": 67088 + }, + { + "epoch": 2.58, + "learning_rate": 7.015260323159785e-06, + "loss": 0.2308, + "step": 67089 + }, + { + "epoch": 2.58, + "learning_rate": 7.014619133111054e-06, + "loss": 0.4253, + "step": 67090 + }, + { + "epoch": 2.58, + "learning_rate": 7.013977943062323e-06, + "loss": 0.1859, + "step": 67091 + }, + { + "epoch": 2.58, + "learning_rate": 7.013336753013594e-06, + "loss": 0.2665, + "step": 67092 + }, + { + "epoch": 2.58, + "learning_rate": 7.012695562964863e-06, + "loss": 0.2656, + "step": 67093 + }, + { + "epoch": 2.58, + "learning_rate": 7.0120543729161324e-06, + "loss": 0.2679, + "step": 67094 + }, + { + "epoch": 2.58, + "learning_rate": 7.011413182867402e-06, + "loss": 0.2552, + "step": 67095 + }, + { + "epoch": 2.58, + "learning_rate": 7.0107719928186725e-06, + "loss": 0.2285, + "step": 67096 + }, + { + "epoch": 2.58, + "learning_rate": 7.010130802769942e-06, + "loss": 0.1998, + "step": 67097 + }, + { + "epoch": 2.58, + "learning_rate": 7.009489612721211e-06, + "loss": 0.2658, + "step": 67098 + }, + { + "epoch": 2.58, + "learning_rate": 7.00884842267248e-06, + "loss": 0.237, + "step": 67099 + }, + { + "epoch": 2.58, + "learning_rate": 7.008207232623749e-06, + "loss": 0.2073, + "step": 67100 + }, + { + "epoch": 2.58, + "learning_rate": 7.00756604257502e-06, + "loss": 0.1812, + "step": 67101 + }, + { + "epoch": 2.58, + "learning_rate": 7.006924852526289e-06, + "loss": 0.2792, + "step": 67102 + }, + { + "epoch": 2.58, + "learning_rate": 7.006283662477558e-06, + "loss": 0.2145, + "step": 67103 + }, + { + "epoch": 2.58, + "learning_rate": 7.005642472428828e-06, + "loss": 0.1777, + "step": 67104 + }, + { + "epoch": 2.58, + "learning_rate": 7.0050012823800985e-06, + "loss": 0.377, + "step": 67105 + }, + { + "epoch": 2.58, + "learning_rate": 7.004360092331368e-06, + "loss": 0.2194, + "step": 67106 + }, + { + "epoch": 2.58, + "learning_rate": 7.003718902282637e-06, + "loss": 0.2706, + "step": 67107 + }, + { + "epoch": 2.58, + "learning_rate": 7.003077712233906e-06, + "loss": 0.2178, + "step": 67108 + }, + { + "epoch": 2.58, + "learning_rate": 7.002436522185175e-06, + "loss": 0.1429, + "step": 67109 + }, + { + "epoch": 2.58, + "learning_rate": 7.001795332136446e-06, + "loss": 0.2125, + "step": 67110 + }, + { + "epoch": 2.58, + "learning_rate": 7.001154142087715e-06, + "loss": 0.1888, + "step": 67111 + }, + { + "epoch": 2.58, + "learning_rate": 7.000512952038984e-06, + "loss": 0.1926, + "step": 67112 + }, + { + "epoch": 2.58, + "learning_rate": 6.9998717619902536e-06, + "loss": 0.2207, + "step": 67113 + }, + { + "epoch": 2.58, + "learning_rate": 6.9992305719415244e-06, + "loss": 0.2003, + "step": 67114 + }, + { + "epoch": 2.58, + "learning_rate": 6.998589381892794e-06, + "loss": 0.2576, + "step": 67115 + }, + { + "epoch": 2.58, + "learning_rate": 6.997948191844063e-06, + "loss": 0.2141, + "step": 67116 + }, + { + "epoch": 2.58, + "learning_rate": 6.997307001795332e-06, + "loss": 0.1996, + "step": 67117 + }, + { + "epoch": 2.58, + "learning_rate": 6.996665811746601e-06, + "loss": 0.223, + "step": 67118 + }, + { + "epoch": 2.58, + "learning_rate": 6.996024621697872e-06, + "loss": 0.2807, + "step": 67119 + }, + { + "epoch": 2.58, + "learning_rate": 6.995383431649141e-06, + "loss": 0.3023, + "step": 67120 + }, + { + "epoch": 2.58, + "learning_rate": 6.99474224160041e-06, + "loss": 0.2405, + "step": 67121 + }, + { + "epoch": 2.58, + "learning_rate": 6.9941010515516795e-06, + "loss": 0.1493, + "step": 67122 + }, + { + "epoch": 2.58, + "learning_rate": 6.99345986150295e-06, + "loss": 0.1778, + "step": 67123 + }, + { + "epoch": 2.58, + "learning_rate": 6.9928186714542196e-06, + "loss": 0.0821, + "step": 67124 + }, + { + "epoch": 2.58, + "learning_rate": 6.992177481405489e-06, + "loss": 0.2578, + "step": 67125 + }, + { + "epoch": 2.58, + "learning_rate": 6.991536291356758e-06, + "loss": 0.1967, + "step": 67126 + }, + { + "epoch": 2.58, + "learning_rate": 6.990895101308029e-06, + "loss": 0.3252, + "step": 67127 + }, + { + "epoch": 2.58, + "learning_rate": 6.990253911259298e-06, + "loss": 0.4273, + "step": 67128 + }, + { + "epoch": 2.58, + "learning_rate": 6.989612721210567e-06, + "loss": 0.2054, + "step": 67129 + }, + { + "epoch": 2.58, + "learning_rate": 6.988971531161836e-06, + "loss": 0.1556, + "step": 67130 + }, + { + "epoch": 2.58, + "learning_rate": 6.9883303411131055e-06, + "loss": 0.2522, + "step": 67131 + }, + { + "epoch": 2.58, + "learning_rate": 6.987689151064376e-06, + "loss": 0.1815, + "step": 67132 + }, + { + "epoch": 2.58, + "learning_rate": 6.9870479610156455e-06, + "loss": 0.2154, + "step": 67133 + }, + { + "epoch": 2.58, + "learning_rate": 6.986406770966915e-06, + "loss": 0.157, + "step": 67134 + }, + { + "epoch": 2.58, + "learning_rate": 6.985765580918184e-06, + "loss": 0.1237, + "step": 67135 + }, + { + "epoch": 2.58, + "learning_rate": 6.985124390869455e-06, + "loss": 0.2856, + "step": 67136 + }, + { + "epoch": 2.58, + "learning_rate": 6.984483200820724e-06, + "loss": 0.169, + "step": 67137 + }, + { + "epoch": 2.58, + "learning_rate": 6.983842010771993e-06, + "loss": 0.2167, + "step": 67138 + }, + { + "epoch": 2.58, + "learning_rate": 6.983200820723262e-06, + "loss": 0.3547, + "step": 67139 + }, + { + "epoch": 2.58, + "learning_rate": 6.9825596306745315e-06, + "loss": 0.3143, + "step": 67140 + }, + { + "epoch": 2.58, + "learning_rate": 6.981918440625802e-06, + "loss": 0.2315, + "step": 67141 + }, + { + "epoch": 2.58, + "learning_rate": 6.9812772505770715e-06, + "loss": 0.1613, + "step": 67142 + }, + { + "epoch": 2.58, + "learning_rate": 6.980636060528341e-06, + "loss": 0.1143, + "step": 67143 + }, + { + "epoch": 2.58, + "learning_rate": 6.97999487047961e-06, + "loss": 0.2101, + "step": 67144 + }, + { + "epoch": 2.58, + "learning_rate": 6.979353680430881e-06, + "loss": 0.2137, + "step": 67145 + }, + { + "epoch": 2.58, + "learning_rate": 6.97871249038215e-06, + "loss": 0.2638, + "step": 67146 + }, + { + "epoch": 2.58, + "learning_rate": 6.978071300333419e-06, + "loss": 0.29, + "step": 67147 + }, + { + "epoch": 2.58, + "learning_rate": 6.977430110284688e-06, + "loss": 0.3122, + "step": 67148 + }, + { + "epoch": 2.58, + "learning_rate": 6.9767889202359574e-06, + "loss": 0.1214, + "step": 67149 + }, + { + "epoch": 2.58, + "learning_rate": 6.976147730187228e-06, + "loss": 0.1808, + "step": 67150 + }, + { + "epoch": 2.58, + "learning_rate": 6.9755065401384975e-06, + "loss": 0.1884, + "step": 67151 + }, + { + "epoch": 2.58, + "learning_rate": 6.974865350089767e-06, + "loss": 0.2429, + "step": 67152 + }, + { + "epoch": 2.58, + "learning_rate": 6.974224160041036e-06, + "loss": 0.2138, + "step": 67153 + }, + { + "epoch": 2.58, + "learning_rate": 6.973582969992307e-06, + "loss": 0.2734, + "step": 67154 + }, + { + "epoch": 2.58, + "learning_rate": 6.972941779943576e-06, + "loss": 0.288, + "step": 67155 + }, + { + "epoch": 2.58, + "learning_rate": 6.972300589894845e-06, + "loss": 0.1581, + "step": 67156 + }, + { + "epoch": 2.58, + "learning_rate": 6.971659399846114e-06, + "loss": 0.3626, + "step": 67157 + }, + { + "epoch": 2.58, + "learning_rate": 6.971018209797383e-06, + "loss": 0.3381, + "step": 67158 + }, + { + "epoch": 2.58, + "learning_rate": 6.970377019748654e-06, + "loss": 0.233, + "step": 67159 + }, + { + "epoch": 2.58, + "learning_rate": 6.9697358296999234e-06, + "loss": 0.1268, + "step": 67160 + }, + { + "epoch": 2.58, + "learning_rate": 6.969094639651193e-06, + "loss": 0.1851, + "step": 67161 + }, + { + "epoch": 2.58, + "learning_rate": 6.968453449602462e-06, + "loss": 0.2093, + "step": 67162 + }, + { + "epoch": 2.58, + "learning_rate": 6.967812259553733e-06, + "loss": 0.2502, + "step": 67163 + }, + { + "epoch": 2.58, + "learning_rate": 6.967171069505002e-06, + "loss": 0.2412, + "step": 67164 + }, + { + "epoch": 2.58, + "learning_rate": 6.966529879456271e-06, + "loss": 0.2897, + "step": 67165 + }, + { + "epoch": 2.58, + "learning_rate": 6.96588868940754e-06, + "loss": 0.1778, + "step": 67166 + }, + { + "epoch": 2.58, + "learning_rate": 6.965247499358809e-06, + "loss": 0.1475, + "step": 67167 + }, + { + "epoch": 2.58, + "learning_rate": 6.96460630931008e-06, + "loss": 0.2456, + "step": 67168 + }, + { + "epoch": 2.58, + "learning_rate": 6.963965119261349e-06, + "loss": 0.3005, + "step": 67169 + }, + { + "epoch": 2.58, + "learning_rate": 6.963323929212619e-06, + "loss": 0.1861, + "step": 67170 + }, + { + "epoch": 2.58, + "learning_rate": 6.962682739163888e-06, + "loss": 0.3358, + "step": 67171 + }, + { + "epoch": 2.58, + "learning_rate": 6.962041549115159e-06, + "loss": 0.143, + "step": 67172 + }, + { + "epoch": 2.58, + "learning_rate": 6.961400359066428e-06, + "loss": 0.1325, + "step": 67173 + }, + { + "epoch": 2.58, + "learning_rate": 6.960759169017697e-06, + "loss": 0.3253, + "step": 67174 + }, + { + "epoch": 2.58, + "learning_rate": 6.960117978968966e-06, + "loss": 0.1885, + "step": 67175 + }, + { + "epoch": 2.58, + "learning_rate": 6.959476788920237e-06, + "loss": 0.2629, + "step": 67176 + }, + { + "epoch": 2.58, + "learning_rate": 6.958835598871506e-06, + "loss": 0.2068, + "step": 67177 + }, + { + "epoch": 2.58, + "learning_rate": 6.958194408822775e-06, + "loss": 0.2014, + "step": 67178 + }, + { + "epoch": 2.58, + "learning_rate": 6.9575532187740446e-06, + "loss": 0.3319, + "step": 67179 + }, + { + "epoch": 2.58, + "learning_rate": 6.956912028725314e-06, + "loss": 0.2784, + "step": 67180 + }, + { + "epoch": 2.58, + "learning_rate": 6.956270838676585e-06, + "loss": 0.1726, + "step": 67181 + }, + { + "epoch": 2.58, + "learning_rate": 6.955629648627854e-06, + "loss": 0.1538, + "step": 67182 + }, + { + "epoch": 2.58, + "learning_rate": 6.954988458579123e-06, + "loss": 0.1264, + "step": 67183 + }, + { + "epoch": 2.58, + "learning_rate": 6.954347268530392e-06, + "loss": 0.1671, + "step": 67184 + }, + { + "epoch": 2.58, + "learning_rate": 6.953706078481663e-06, + "loss": 0.1809, + "step": 67185 + }, + { + "epoch": 2.58, + "learning_rate": 6.953064888432932e-06, + "loss": 0.1807, + "step": 67186 + }, + { + "epoch": 2.58, + "learning_rate": 6.952423698384201e-06, + "loss": 0.2702, + "step": 67187 + }, + { + "epoch": 2.58, + "learning_rate": 6.9517825083354705e-06, + "loss": 0.14, + "step": 67188 + }, + { + "epoch": 2.58, + "learning_rate": 6.95114131828674e-06, + "loss": 0.1673, + "step": 67189 + }, + { + "epoch": 2.58, + "learning_rate": 6.9505001282380106e-06, + "loss": 0.1287, + "step": 67190 + }, + { + "epoch": 2.58, + "learning_rate": 6.94985893818928e-06, + "loss": 0.1769, + "step": 67191 + }, + { + "epoch": 2.58, + "learning_rate": 6.949217748140549e-06, + "loss": 0.2422, + "step": 67192 + }, + { + "epoch": 2.58, + "learning_rate": 6.948576558091818e-06, + "loss": 0.1265, + "step": 67193 + }, + { + "epoch": 2.58, + "learning_rate": 6.947935368043089e-06, + "loss": 0.3879, + "step": 67194 + }, + { + "epoch": 2.58, + "learning_rate": 6.947294177994358e-06, + "loss": 0.2271, + "step": 67195 + }, + { + "epoch": 2.58, + "learning_rate": 6.946652987945627e-06, + "loss": 0.1334, + "step": 67196 + }, + { + "epoch": 2.58, + "learning_rate": 6.9460117978968965e-06, + "loss": 0.0586, + "step": 67197 + }, + { + "epoch": 2.58, + "learning_rate": 6.945370607848166e-06, + "loss": 0.1814, + "step": 67198 + }, + { + "epoch": 2.58, + "learning_rate": 6.9447294177994365e-06, + "loss": 0.195, + "step": 67199 + }, + { + "epoch": 2.58, + "learning_rate": 6.944088227750706e-06, + "loss": 0.5049, + "step": 67200 + }, + { + "epoch": 2.58, + "learning_rate": 6.943447037701975e-06, + "loss": 0.2045, + "step": 67201 + }, + { + "epoch": 2.58, + "learning_rate": 6.942805847653244e-06, + "loss": 0.1311, + "step": 67202 + }, + { + "epoch": 2.58, + "learning_rate": 6.942164657604515e-06, + "loss": 0.1421, + "step": 67203 + }, + { + "epoch": 2.58, + "learning_rate": 6.941523467555784e-06, + "loss": 0.1748, + "step": 67204 + }, + { + "epoch": 2.58, + "learning_rate": 6.940882277507053e-06, + "loss": 0.1894, + "step": 67205 + }, + { + "epoch": 2.58, + "learning_rate": 6.9402410874583225e-06, + "loss": 0.2144, + "step": 67206 + }, + { + "epoch": 2.58, + "learning_rate": 6.939599897409592e-06, + "loss": 0.2343, + "step": 67207 + }, + { + "epoch": 2.58, + "learning_rate": 6.9389587073608625e-06, + "loss": 0.24, + "step": 67208 + }, + { + "epoch": 2.58, + "learning_rate": 6.938317517312132e-06, + "loss": 0.1823, + "step": 67209 + }, + { + "epoch": 2.58, + "learning_rate": 6.937676327263401e-06, + "loss": 0.2691, + "step": 67210 + }, + { + "epoch": 2.58, + "learning_rate": 6.93703513721467e-06, + "loss": 0.2162, + "step": 67211 + }, + { + "epoch": 2.58, + "learning_rate": 6.936393947165941e-06, + "loss": 0.2357, + "step": 67212 + }, + { + "epoch": 2.58, + "learning_rate": 6.93575275711721e-06, + "loss": 0.3498, + "step": 67213 + }, + { + "epoch": 2.58, + "learning_rate": 6.935111567068479e-06, + "loss": 0.1324, + "step": 67214 + }, + { + "epoch": 2.58, + "learning_rate": 6.9344703770197484e-06, + "loss": 0.2669, + "step": 67215 + }, + { + "epoch": 2.58, + "learning_rate": 6.933829186971018e-06, + "loss": 0.2044, + "step": 67216 + }, + { + "epoch": 2.58, + "learning_rate": 6.9331879969222885e-06, + "loss": 0.2121, + "step": 67217 + }, + { + "epoch": 2.58, + "learning_rate": 6.932546806873558e-06, + "loss": 0.0797, + "step": 67218 + }, + { + "epoch": 2.58, + "learning_rate": 6.931905616824827e-06, + "loss": 0.233, + "step": 67219 + }, + { + "epoch": 2.58, + "learning_rate": 6.931264426776096e-06, + "loss": 0.1965, + "step": 67220 + }, + { + "epoch": 2.58, + "learning_rate": 6.930623236727367e-06, + "loss": 0.2932, + "step": 67221 + }, + { + "epoch": 2.58, + "learning_rate": 6.929982046678636e-06, + "loss": 0.1362, + "step": 67222 + }, + { + "epoch": 2.58, + "learning_rate": 6.929340856629905e-06, + "loss": 0.1738, + "step": 67223 + }, + { + "epoch": 2.58, + "learning_rate": 6.928699666581174e-06, + "loss": 0.2036, + "step": 67224 + }, + { + "epoch": 2.58, + "learning_rate": 6.928058476532445e-06, + "loss": 0.1569, + "step": 67225 + }, + { + "epoch": 2.58, + "learning_rate": 6.9274172864837144e-06, + "loss": 0.2544, + "step": 67226 + }, + { + "epoch": 2.58, + "learning_rate": 6.926776096434984e-06, + "loss": 0.1988, + "step": 67227 + }, + { + "epoch": 2.58, + "learning_rate": 6.926134906386253e-06, + "loss": 0.1353, + "step": 67228 + }, + { + "epoch": 2.58, + "learning_rate": 6.925493716337522e-06, + "loss": 0.3745, + "step": 67229 + }, + { + "epoch": 2.58, + "learning_rate": 6.924852526288793e-06, + "loss": 0.2017, + "step": 67230 + }, + { + "epoch": 2.58, + "learning_rate": 6.924211336240062e-06, + "loss": 0.1828, + "step": 67231 + }, + { + "epoch": 2.58, + "learning_rate": 6.923570146191331e-06, + "loss": 0.2348, + "step": 67232 + }, + { + "epoch": 2.58, + "learning_rate": 6.9229289561426e-06, + "loss": 0.1901, + "step": 67233 + }, + { + "epoch": 2.58, + "learning_rate": 6.922287766093871e-06, + "loss": 0.29, + "step": 67234 + }, + { + "epoch": 2.58, + "learning_rate": 6.92164657604514e-06, + "loss": 0.2172, + "step": 67235 + }, + { + "epoch": 2.59, + "learning_rate": 6.92100538599641e-06, + "loss": 0.33, + "step": 67236 + }, + { + "epoch": 2.59, + "learning_rate": 6.920364195947679e-06, + "loss": 0.191, + "step": 67237 + }, + { + "epoch": 2.59, + "learning_rate": 6.919723005898948e-06, + "loss": 0.1099, + "step": 67238 + }, + { + "epoch": 2.59, + "learning_rate": 6.919081815850219e-06, + "loss": 0.1887, + "step": 67239 + }, + { + "epoch": 2.59, + "learning_rate": 6.918440625801488e-06, + "loss": 0.0851, + "step": 67240 + }, + { + "epoch": 2.59, + "learning_rate": 6.917799435752757e-06, + "loss": 0.1605, + "step": 67241 + }, + { + "epoch": 2.59, + "learning_rate": 6.917158245704026e-06, + "loss": 0.1013, + "step": 67242 + }, + { + "epoch": 2.59, + "learning_rate": 6.916517055655297e-06, + "loss": 0.2617, + "step": 67243 + }, + { + "epoch": 2.59, + "learning_rate": 6.915875865606566e-06, + "loss": 0.1877, + "step": 67244 + }, + { + "epoch": 2.59, + "learning_rate": 6.9152346755578356e-06, + "loss": 0.1909, + "step": 67245 + }, + { + "epoch": 2.59, + "learning_rate": 6.914593485509105e-06, + "loss": 0.1565, + "step": 67246 + }, + { + "epoch": 2.59, + "learning_rate": 6.913952295460374e-06, + "loss": 0.2264, + "step": 67247 + }, + { + "epoch": 2.59, + "learning_rate": 6.913311105411645e-06, + "loss": 0.1746, + "step": 67248 + }, + { + "epoch": 2.59, + "learning_rate": 6.912669915362914e-06, + "loss": 0.1129, + "step": 67249 + }, + { + "epoch": 2.59, + "learning_rate": 6.912028725314183e-06, + "loss": 0.2537, + "step": 67250 + }, + { + "epoch": 2.59, + "learning_rate": 6.911387535265452e-06, + "loss": 0.3121, + "step": 67251 + }, + { + "epoch": 2.59, + "learning_rate": 6.910746345216723e-06, + "loss": 0.2196, + "step": 67252 + }, + { + "epoch": 2.59, + "learning_rate": 6.910105155167992e-06, + "loss": 0.3829, + "step": 67253 + }, + { + "epoch": 2.59, + "learning_rate": 6.9094639651192615e-06, + "loss": 0.0597, + "step": 67254 + }, + { + "epoch": 2.59, + "learning_rate": 6.908822775070531e-06, + "loss": 0.231, + "step": 67255 + }, + { + "epoch": 2.59, + "learning_rate": 6.9081815850218e-06, + "loss": 0.1901, + "step": 67256 + }, + { + "epoch": 2.59, + "learning_rate": 6.907540394973071e-06, + "loss": 0.2009, + "step": 67257 + }, + { + "epoch": 2.59, + "learning_rate": 6.90689920492434e-06, + "loss": 0.257, + "step": 67258 + }, + { + "epoch": 2.59, + "learning_rate": 6.906258014875609e-06, + "loss": 0.2688, + "step": 67259 + }, + { + "epoch": 2.59, + "learning_rate": 6.905616824826878e-06, + "loss": 0.1496, + "step": 67260 + }, + { + "epoch": 2.59, + "learning_rate": 6.904975634778149e-06, + "loss": 0.2084, + "step": 67261 + }, + { + "epoch": 2.59, + "learning_rate": 6.904334444729418e-06, + "loss": 0.2786, + "step": 67262 + }, + { + "epoch": 2.59, + "learning_rate": 6.9036932546806875e-06, + "loss": 0.2871, + "step": 67263 + }, + { + "epoch": 2.59, + "learning_rate": 6.903052064631957e-06, + "loss": 0.2597, + "step": 67264 + }, + { + "epoch": 2.59, + "learning_rate": 6.902410874583226e-06, + "loss": 0.2421, + "step": 67265 + }, + { + "epoch": 2.59, + "learning_rate": 6.901769684534497e-06, + "loss": 0.2147, + "step": 67266 + }, + { + "epoch": 2.59, + "learning_rate": 6.901128494485766e-06, + "loss": 0.2622, + "step": 67267 + }, + { + "epoch": 2.59, + "learning_rate": 6.900487304437035e-06, + "loss": 0.2966, + "step": 67268 + }, + { + "epoch": 2.59, + "learning_rate": 6.899846114388304e-06, + "loss": 0.2077, + "step": 67269 + }, + { + "epoch": 2.59, + "learning_rate": 6.899204924339575e-06, + "loss": 0.1022, + "step": 67270 + }, + { + "epoch": 2.59, + "learning_rate": 6.898563734290844e-06, + "loss": 0.1729, + "step": 67271 + }, + { + "epoch": 2.59, + "learning_rate": 6.8979225442421135e-06, + "loss": 0.2615, + "step": 67272 + }, + { + "epoch": 2.59, + "learning_rate": 6.897281354193383e-06, + "loss": 0.1608, + "step": 67273 + }, + { + "epoch": 2.59, + "learning_rate": 6.8966401641446535e-06, + "loss": 0.0994, + "step": 67274 + }, + { + "epoch": 2.59, + "learning_rate": 6.895998974095923e-06, + "loss": 0.244, + "step": 67275 + }, + { + "epoch": 2.59, + "learning_rate": 6.895357784047192e-06, + "loss": 0.2576, + "step": 67276 + }, + { + "epoch": 2.59, + "learning_rate": 6.894716593998461e-06, + "loss": 0.1864, + "step": 67277 + }, + { + "epoch": 2.59, + "learning_rate": 6.89407540394973e-06, + "loss": 0.2982, + "step": 67278 + }, + { + "epoch": 2.59, + "learning_rate": 6.893434213901001e-06, + "loss": 0.1886, + "step": 67279 + }, + { + "epoch": 2.59, + "learning_rate": 6.89279302385227e-06, + "loss": 0.1799, + "step": 67280 + }, + { + "epoch": 2.59, + "learning_rate": 6.892151833803539e-06, + "loss": 0.0925, + "step": 67281 + }, + { + "epoch": 2.59, + "learning_rate": 6.891510643754809e-06, + "loss": 0.2744, + "step": 67282 + }, + { + "epoch": 2.59, + "learning_rate": 6.8908694537060795e-06, + "loss": 0.3041, + "step": 67283 + }, + { + "epoch": 2.59, + "learning_rate": 6.890228263657349e-06, + "loss": 0.1394, + "step": 67284 + }, + { + "epoch": 2.59, + "learning_rate": 6.889587073608618e-06, + "loss": 0.1742, + "step": 67285 + }, + { + "epoch": 2.59, + "learning_rate": 6.888945883559887e-06, + "loss": 0.1795, + "step": 67286 + }, + { + "epoch": 2.59, + "learning_rate": 6.888304693511156e-06, + "loss": 0.2344, + "step": 67287 + }, + { + "epoch": 2.59, + "learning_rate": 6.887663503462427e-06, + "loss": 0.2479, + "step": 67288 + }, + { + "epoch": 2.59, + "learning_rate": 6.887022313413696e-06, + "loss": 0.1194, + "step": 67289 + }, + { + "epoch": 2.59, + "learning_rate": 6.886381123364965e-06, + "loss": 0.3371, + "step": 67290 + }, + { + "epoch": 2.59, + "learning_rate": 6.8857399333162346e-06, + "loss": 0.0877, + "step": 67291 + }, + { + "epoch": 2.59, + "learning_rate": 6.8850987432675054e-06, + "loss": 0.3363, + "step": 67292 + }, + { + "epoch": 2.59, + "learning_rate": 6.884457553218775e-06, + "loss": 0.1982, + "step": 67293 + }, + { + "epoch": 2.59, + "learning_rate": 6.883816363170044e-06, + "loss": 0.1654, + "step": 67294 + }, + { + "epoch": 2.59, + "learning_rate": 6.883175173121313e-06, + "loss": 0.1173, + "step": 67295 + }, + { + "epoch": 2.59, + "learning_rate": 6.882533983072582e-06, + "loss": 0.2432, + "step": 67296 + }, + { + "epoch": 2.59, + "learning_rate": 6.881892793023853e-06, + "loss": 0.1634, + "step": 67297 + }, + { + "epoch": 2.59, + "learning_rate": 6.881251602975122e-06, + "loss": 0.2667, + "step": 67298 + }, + { + "epoch": 2.59, + "learning_rate": 6.880610412926391e-06, + "loss": 0.4859, + "step": 67299 + }, + { + "epoch": 2.59, + "learning_rate": 6.8799692228776605e-06, + "loss": 0.1888, + "step": 67300 + }, + { + "epoch": 2.59, + "learning_rate": 6.879328032828931e-06, + "loss": 0.2023, + "step": 67301 + }, + { + "epoch": 2.59, + "learning_rate": 6.878686842780201e-06, + "loss": 0.1455, + "step": 67302 + }, + { + "epoch": 2.59, + "learning_rate": 6.87804565273147e-06, + "loss": 0.1438, + "step": 67303 + }, + { + "epoch": 2.59, + "learning_rate": 6.877404462682739e-06, + "loss": 0.2659, + "step": 67304 + }, + { + "epoch": 2.59, + "learning_rate": 6.876763272634008e-06, + "loss": 0.3019, + "step": 67305 + }, + { + "epoch": 2.59, + "learning_rate": 6.876122082585279e-06, + "loss": 0.1705, + "step": 67306 + }, + { + "epoch": 2.59, + "learning_rate": 6.875480892536548e-06, + "loss": 0.2115, + "step": 67307 + }, + { + "epoch": 2.59, + "learning_rate": 6.874839702487817e-06, + "loss": 0.2875, + "step": 67308 + }, + { + "epoch": 2.59, + "learning_rate": 6.8741985124390865e-06, + "loss": 0.2519, + "step": 67309 + }, + { + "epoch": 2.59, + "learning_rate": 6.873557322390357e-06, + "loss": 0.2907, + "step": 67310 + }, + { + "epoch": 2.59, + "learning_rate": 6.8729161323416265e-06, + "loss": 0.138, + "step": 67311 + }, + { + "epoch": 2.59, + "learning_rate": 6.872274942292896e-06, + "loss": 0.1979, + "step": 67312 + }, + { + "epoch": 2.59, + "learning_rate": 6.871633752244165e-06, + "loss": 0.3156, + "step": 67313 + }, + { + "epoch": 2.59, + "learning_rate": 6.870992562195436e-06, + "loss": 0.2877, + "step": 67314 + }, + { + "epoch": 2.59, + "learning_rate": 6.870351372146705e-06, + "loss": 0.2259, + "step": 67315 + }, + { + "epoch": 2.59, + "learning_rate": 6.869710182097974e-06, + "loss": 0.1222, + "step": 67316 + }, + { + "epoch": 2.59, + "learning_rate": 6.869068992049243e-06, + "loss": 0.1429, + "step": 67317 + }, + { + "epoch": 2.59, + "learning_rate": 6.8684278020005125e-06, + "loss": 0.1946, + "step": 67318 + }, + { + "epoch": 2.59, + "learning_rate": 6.867786611951783e-06, + "loss": 0.2239, + "step": 67319 + }, + { + "epoch": 2.59, + "learning_rate": 6.8671454219030525e-06, + "loss": 0.1834, + "step": 67320 + }, + { + "epoch": 2.59, + "learning_rate": 6.866504231854322e-06, + "loss": 0.1314, + "step": 67321 + }, + { + "epoch": 2.59, + "learning_rate": 6.865863041805591e-06, + "loss": 0.2514, + "step": 67322 + }, + { + "epoch": 2.59, + "learning_rate": 6.865221851756862e-06, + "loss": 0.1866, + "step": 67323 + }, + { + "epoch": 2.59, + "learning_rate": 6.864580661708131e-06, + "loss": 0.1661, + "step": 67324 + }, + { + "epoch": 2.59, + "learning_rate": 6.8639394716594e-06, + "loss": 0.3054, + "step": 67325 + }, + { + "epoch": 2.59, + "learning_rate": 6.863298281610669e-06, + "loss": 0.2074, + "step": 67326 + }, + { + "epoch": 2.59, + "learning_rate": 6.8626570915619384e-06, + "loss": 0.1591, + "step": 67327 + }, + { + "epoch": 2.59, + "learning_rate": 6.862015901513209e-06, + "loss": 0.2224, + "step": 67328 + }, + { + "epoch": 2.59, + "learning_rate": 6.8613747114644785e-06, + "loss": 0.1416, + "step": 67329 + }, + { + "epoch": 2.59, + "learning_rate": 6.860733521415748e-06, + "loss": 0.2516, + "step": 67330 + }, + { + "epoch": 2.59, + "learning_rate": 6.860092331367017e-06, + "loss": 0.1834, + "step": 67331 + }, + { + "epoch": 2.59, + "learning_rate": 6.859451141318288e-06, + "loss": 0.1418, + "step": 67332 + }, + { + "epoch": 2.59, + "learning_rate": 6.858809951269557e-06, + "loss": 0.2077, + "step": 67333 + }, + { + "epoch": 2.59, + "learning_rate": 6.858168761220826e-06, + "loss": 0.1678, + "step": 67334 + }, + { + "epoch": 2.59, + "learning_rate": 6.857527571172095e-06, + "loss": 0.427, + "step": 67335 + }, + { + "epoch": 2.59, + "learning_rate": 6.856886381123364e-06, + "loss": 0.19, + "step": 67336 + }, + { + "epoch": 2.59, + "learning_rate": 6.856245191074635e-06, + "loss": 0.37, + "step": 67337 + }, + { + "epoch": 2.59, + "learning_rate": 6.8556040010259045e-06, + "loss": 0.2556, + "step": 67338 + }, + { + "epoch": 2.59, + "learning_rate": 6.854962810977174e-06, + "loss": 0.1847, + "step": 67339 + }, + { + "epoch": 2.59, + "learning_rate": 6.854321620928443e-06, + "loss": 0.3728, + "step": 67340 + }, + { + "epoch": 2.59, + "learning_rate": 6.853680430879714e-06, + "loss": 0.2772, + "step": 67341 + }, + { + "epoch": 2.59, + "learning_rate": 6.853039240830983e-06, + "loss": 0.3154, + "step": 67342 + }, + { + "epoch": 2.59, + "learning_rate": 6.852398050782252e-06, + "loss": 0.2051, + "step": 67343 + }, + { + "epoch": 2.59, + "learning_rate": 6.851756860733521e-06, + "loss": 0.289, + "step": 67344 + }, + { + "epoch": 2.59, + "learning_rate": 6.85111567068479e-06, + "loss": 0.1461, + "step": 67345 + }, + { + "epoch": 2.59, + "learning_rate": 6.850474480636061e-06, + "loss": 0.1865, + "step": 67346 + }, + { + "epoch": 2.59, + "learning_rate": 6.84983329058733e-06, + "loss": 0.2082, + "step": 67347 + }, + { + "epoch": 2.59, + "learning_rate": 6.8491921005386e-06, + "loss": 0.3111, + "step": 67348 + }, + { + "epoch": 2.59, + "learning_rate": 6.848550910489869e-06, + "loss": 0.159, + "step": 67349 + }, + { + "epoch": 2.59, + "learning_rate": 6.84790972044114e-06, + "loss": 0.2115, + "step": 67350 + }, + { + "epoch": 2.59, + "learning_rate": 6.847268530392409e-06, + "loss": 0.138, + "step": 67351 + }, + { + "epoch": 2.59, + "learning_rate": 6.846627340343678e-06, + "loss": 0.1469, + "step": 67352 + }, + { + "epoch": 2.59, + "learning_rate": 6.845986150294947e-06, + "loss": 0.1666, + "step": 67353 + }, + { + "epoch": 2.59, + "learning_rate": 6.845344960246216e-06, + "loss": 0.1864, + "step": 67354 + }, + { + "epoch": 2.59, + "learning_rate": 6.844703770197487e-06, + "loss": 0.2447, + "step": 67355 + }, + { + "epoch": 2.59, + "learning_rate": 6.844062580148756e-06, + "loss": 0.1501, + "step": 67356 + }, + { + "epoch": 2.59, + "learning_rate": 6.8434213901000256e-06, + "loss": 0.1554, + "step": 67357 + }, + { + "epoch": 2.59, + "learning_rate": 6.842780200051295e-06, + "loss": 0.254, + "step": 67358 + }, + { + "epoch": 2.59, + "learning_rate": 6.842139010002566e-06, + "loss": 0.3629, + "step": 67359 + }, + { + "epoch": 2.59, + "learning_rate": 6.841497819953835e-06, + "loss": 0.1356, + "step": 67360 + }, + { + "epoch": 2.59, + "learning_rate": 6.840856629905104e-06, + "loss": 0.1185, + "step": 67361 + }, + { + "epoch": 2.59, + "learning_rate": 6.840215439856373e-06, + "loss": 0.1844, + "step": 67362 + }, + { + "epoch": 2.59, + "learning_rate": 6.839574249807644e-06, + "loss": 0.073, + "step": 67363 + }, + { + "epoch": 2.59, + "learning_rate": 6.838933059758913e-06, + "loss": 0.2337, + "step": 67364 + }, + { + "epoch": 2.59, + "learning_rate": 6.838291869710182e-06, + "loss": 0.1673, + "step": 67365 + }, + { + "epoch": 2.59, + "learning_rate": 6.8376506796614515e-06, + "loss": 0.3399, + "step": 67366 + }, + { + "epoch": 2.59, + "learning_rate": 6.837009489612721e-06, + "loss": 0.1934, + "step": 67367 + }, + { + "epoch": 2.59, + "learning_rate": 6.8363682995639916e-06, + "loss": 0.2793, + "step": 67368 + }, + { + "epoch": 2.59, + "learning_rate": 6.835727109515261e-06, + "loss": 0.2753, + "step": 67369 + }, + { + "epoch": 2.59, + "learning_rate": 6.83508591946653e-06, + "loss": 0.1852, + "step": 67370 + }, + { + "epoch": 2.59, + "learning_rate": 6.834444729417799e-06, + "loss": 0.2037, + "step": 67371 + }, + { + "epoch": 2.59, + "learning_rate": 6.83380353936907e-06, + "loss": 0.1959, + "step": 67372 + }, + { + "epoch": 2.59, + "learning_rate": 6.833162349320339e-06, + "loss": 0.127, + "step": 67373 + }, + { + "epoch": 2.59, + "learning_rate": 6.832521159271608e-06, + "loss": 0.1345, + "step": 67374 + }, + { + "epoch": 2.59, + "learning_rate": 6.8318799692228775e-06, + "loss": 0.1582, + "step": 67375 + }, + { + "epoch": 2.59, + "learning_rate": 6.831238779174147e-06, + "loss": 0.3124, + "step": 67376 + }, + { + "epoch": 2.59, + "learning_rate": 6.8305975891254175e-06, + "loss": 0.186, + "step": 67377 + }, + { + "epoch": 2.59, + "learning_rate": 6.829956399076687e-06, + "loss": 0.2133, + "step": 67378 + }, + { + "epoch": 2.59, + "learning_rate": 6.829315209027956e-06, + "loss": 0.2978, + "step": 67379 + }, + { + "epoch": 2.59, + "learning_rate": 6.828674018979225e-06, + "loss": 0.2527, + "step": 67380 + }, + { + "epoch": 2.59, + "learning_rate": 6.828032828930496e-06, + "loss": 0.165, + "step": 67381 + }, + { + "epoch": 2.59, + "learning_rate": 6.827391638881765e-06, + "loss": 0.2845, + "step": 67382 + }, + { + "epoch": 2.59, + "learning_rate": 6.826750448833034e-06, + "loss": 0.2109, + "step": 67383 + }, + { + "epoch": 2.59, + "learning_rate": 6.8261092587843035e-06, + "loss": 0.2477, + "step": 67384 + }, + { + "epoch": 2.59, + "learning_rate": 6.825468068735573e-06, + "loss": 0.2122, + "step": 67385 + }, + { + "epoch": 2.59, + "learning_rate": 6.8248268786868435e-06, + "loss": 0.2019, + "step": 67386 + }, + { + "epoch": 2.59, + "learning_rate": 6.824185688638113e-06, + "loss": 0.2479, + "step": 67387 + }, + { + "epoch": 2.59, + "learning_rate": 6.823544498589382e-06, + "loss": 0.2598, + "step": 67388 + }, + { + "epoch": 2.59, + "learning_rate": 6.822903308540651e-06, + "loss": 0.1889, + "step": 67389 + }, + { + "epoch": 2.59, + "learning_rate": 6.822262118491922e-06, + "loss": 0.2784, + "step": 67390 + }, + { + "epoch": 2.59, + "learning_rate": 6.821620928443191e-06, + "loss": 0.2554, + "step": 67391 + }, + { + "epoch": 2.59, + "learning_rate": 6.82097973839446e-06, + "loss": 0.1311, + "step": 67392 + }, + { + "epoch": 2.59, + "learning_rate": 6.8203385483457294e-06, + "loss": 0.4208, + "step": 67393 + }, + { + "epoch": 2.59, + "learning_rate": 6.819697358296999e-06, + "loss": 0.2225, + "step": 67394 + }, + { + "epoch": 2.59, + "learning_rate": 6.8190561682482695e-06, + "loss": 0.207, + "step": 67395 + }, + { + "epoch": 2.59, + "learning_rate": 6.818414978199539e-06, + "loss": 0.2564, + "step": 67396 + }, + { + "epoch": 2.59, + "learning_rate": 6.817773788150808e-06, + "loss": 0.2204, + "step": 67397 + }, + { + "epoch": 2.59, + "learning_rate": 6.817132598102077e-06, + "loss": 0.1837, + "step": 67398 + }, + { + "epoch": 2.59, + "learning_rate": 6.816491408053348e-06, + "loss": 0.2099, + "step": 67399 + }, + { + "epoch": 2.59, + "learning_rate": 6.815850218004617e-06, + "loss": 0.2108, + "step": 67400 + }, + { + "epoch": 2.59, + "learning_rate": 6.815209027955886e-06, + "loss": 0.3887, + "step": 67401 + }, + { + "epoch": 2.59, + "learning_rate": 6.814567837907155e-06, + "loss": 0.1075, + "step": 67402 + }, + { + "epoch": 2.59, + "learning_rate": 6.813926647858425e-06, + "loss": 0.2048, + "step": 67403 + }, + { + "epoch": 2.59, + "learning_rate": 6.8132854578096954e-06, + "loss": 0.1727, + "step": 67404 + }, + { + "epoch": 2.59, + "learning_rate": 6.812644267760965e-06, + "loss": 0.2004, + "step": 67405 + }, + { + "epoch": 2.59, + "learning_rate": 6.812003077712234e-06, + "loss": 0.2891, + "step": 67406 + }, + { + "epoch": 2.59, + "learning_rate": 6.811361887663503e-06, + "loss": 0.3293, + "step": 67407 + }, + { + "epoch": 2.59, + "learning_rate": 6.810720697614774e-06, + "loss": 0.1497, + "step": 67408 + }, + { + "epoch": 2.59, + "learning_rate": 6.810079507566043e-06, + "loss": 0.1057, + "step": 67409 + }, + { + "epoch": 2.59, + "learning_rate": 6.809438317517312e-06, + "loss": 0.1557, + "step": 67410 + }, + { + "epoch": 2.59, + "learning_rate": 6.808797127468581e-06, + "loss": 0.2037, + "step": 67411 + }, + { + "epoch": 2.59, + "learning_rate": 6.808155937419852e-06, + "loss": 0.2274, + "step": 67412 + }, + { + "epoch": 2.59, + "learning_rate": 6.807514747371121e-06, + "loss": 0.2112, + "step": 67413 + }, + { + "epoch": 2.59, + "learning_rate": 6.806873557322391e-06, + "loss": 0.3393, + "step": 67414 + }, + { + "epoch": 2.59, + "learning_rate": 6.80623236727366e-06, + "loss": 0.1352, + "step": 67415 + }, + { + "epoch": 2.59, + "learning_rate": 6.805591177224929e-06, + "loss": 0.2363, + "step": 67416 + }, + { + "epoch": 2.59, + "learning_rate": 6.8049499871762e-06, + "loss": 0.4377, + "step": 67417 + }, + { + "epoch": 2.59, + "learning_rate": 6.804308797127469e-06, + "loss": 0.2349, + "step": 67418 + }, + { + "epoch": 2.59, + "learning_rate": 6.803667607078738e-06, + "loss": 0.2314, + "step": 67419 + }, + { + "epoch": 2.59, + "learning_rate": 6.803026417030007e-06, + "loss": 0.2348, + "step": 67420 + }, + { + "epoch": 2.59, + "learning_rate": 6.802385226981278e-06, + "loss": 0.2801, + "step": 67421 + }, + { + "epoch": 2.59, + "learning_rate": 6.801744036932547e-06, + "loss": 0.21, + "step": 67422 + }, + { + "epoch": 2.59, + "learning_rate": 6.8011028468838166e-06, + "loss": 0.2744, + "step": 67423 + }, + { + "epoch": 2.59, + "learning_rate": 6.800461656835086e-06, + "loss": 0.14, + "step": 67424 + }, + { + "epoch": 2.59, + "learning_rate": 6.799820466786355e-06, + "loss": 0.2831, + "step": 67425 + }, + { + "epoch": 2.59, + "learning_rate": 6.799179276737626e-06, + "loss": 0.2242, + "step": 67426 + }, + { + "epoch": 2.59, + "learning_rate": 6.798538086688895e-06, + "loss": 0.2274, + "step": 67427 + }, + { + "epoch": 2.59, + "learning_rate": 6.797896896640164e-06, + "loss": 0.1682, + "step": 67428 + }, + { + "epoch": 2.59, + "learning_rate": 6.797255706591433e-06, + "loss": 0.3901, + "step": 67429 + }, + { + "epoch": 2.59, + "learning_rate": 6.796614516542704e-06, + "loss": 0.1374, + "step": 67430 + }, + { + "epoch": 2.59, + "learning_rate": 6.795973326493973e-06, + "loss": 0.2111, + "step": 67431 + }, + { + "epoch": 2.59, + "learning_rate": 6.7953321364452425e-06, + "loss": 0.2173, + "step": 67432 + }, + { + "epoch": 2.59, + "learning_rate": 6.794690946396512e-06, + "loss": 0.219, + "step": 67433 + }, + { + "epoch": 2.59, + "learning_rate": 6.794049756347781e-06, + "loss": 0.2901, + "step": 67434 + }, + { + "epoch": 2.59, + "learning_rate": 6.793408566299052e-06, + "loss": 0.1534, + "step": 67435 + }, + { + "epoch": 2.59, + "learning_rate": 6.792767376250321e-06, + "loss": 0.2374, + "step": 67436 + }, + { + "epoch": 2.59, + "learning_rate": 6.79212618620159e-06, + "loss": 0.2123, + "step": 67437 + }, + { + "epoch": 2.59, + "learning_rate": 6.791484996152859e-06, + "loss": 0.168, + "step": 67438 + }, + { + "epoch": 2.59, + "learning_rate": 6.79084380610413e-06, + "loss": 0.2814, + "step": 67439 + }, + { + "epoch": 2.59, + "learning_rate": 6.790202616055399e-06, + "loss": 0.154, + "step": 67440 + }, + { + "epoch": 2.59, + "learning_rate": 6.7895614260066685e-06, + "loss": 0.2791, + "step": 67441 + }, + { + "epoch": 2.59, + "learning_rate": 6.788920235957938e-06, + "loss": 0.1642, + "step": 67442 + }, + { + "epoch": 2.59, + "learning_rate": 6.788279045909207e-06, + "loss": 0.2618, + "step": 67443 + }, + { + "epoch": 2.59, + "learning_rate": 6.787637855860478e-06, + "loss": 0.2173, + "step": 67444 + }, + { + "epoch": 2.59, + "learning_rate": 6.786996665811747e-06, + "loss": 0.2123, + "step": 67445 + }, + { + "epoch": 2.59, + "learning_rate": 6.786355475763016e-06, + "loss": 0.1501, + "step": 67446 + }, + { + "epoch": 2.59, + "learning_rate": 6.785714285714285e-06, + "loss": 0.1712, + "step": 67447 + }, + { + "epoch": 2.59, + "learning_rate": 6.785073095665556e-06, + "loss": 0.2293, + "step": 67448 + }, + { + "epoch": 2.59, + "learning_rate": 6.784431905616825e-06, + "loss": 0.1999, + "step": 67449 + }, + { + "epoch": 2.59, + "learning_rate": 6.7837907155680945e-06, + "loss": 0.2236, + "step": 67450 + }, + { + "epoch": 2.59, + "learning_rate": 6.783149525519364e-06, + "loss": 0.1487, + "step": 67451 + }, + { + "epoch": 2.59, + "learning_rate": 6.782508335470633e-06, + "loss": 0.2275, + "step": 67452 + }, + { + "epoch": 2.59, + "learning_rate": 6.781867145421904e-06, + "loss": 0.2095, + "step": 67453 + }, + { + "epoch": 2.59, + "learning_rate": 6.781225955373173e-06, + "loss": 0.253, + "step": 67454 + }, + { + "epoch": 2.59, + "learning_rate": 6.780584765324442e-06, + "loss": 0.1822, + "step": 67455 + }, + { + "epoch": 2.59, + "learning_rate": 6.779943575275711e-06, + "loss": 0.3138, + "step": 67456 + }, + { + "epoch": 2.59, + "learning_rate": 6.779302385226982e-06, + "loss": 0.2504, + "step": 67457 + }, + { + "epoch": 2.59, + "learning_rate": 6.778661195178251e-06, + "loss": 0.3132, + "step": 67458 + }, + { + "epoch": 2.59, + "learning_rate": 6.7780200051295204e-06, + "loss": 0.2353, + "step": 67459 + }, + { + "epoch": 2.59, + "learning_rate": 6.77737881508079e-06, + "loss": 0.1384, + "step": 67460 + }, + { + "epoch": 2.59, + "learning_rate": 6.7767376250320605e-06, + "loss": 0.1738, + "step": 67461 + }, + { + "epoch": 2.59, + "learning_rate": 6.77609643498333e-06, + "loss": 0.2034, + "step": 67462 + }, + { + "epoch": 2.59, + "learning_rate": 6.775455244934599e-06, + "loss": 0.1081, + "step": 67463 + }, + { + "epoch": 2.59, + "learning_rate": 6.774814054885868e-06, + "loss": 0.1812, + "step": 67464 + }, + { + "epoch": 2.59, + "learning_rate": 6.774172864837137e-06, + "loss": 0.174, + "step": 67465 + }, + { + "epoch": 2.59, + "learning_rate": 6.773531674788408e-06, + "loss": 0.2789, + "step": 67466 + }, + { + "epoch": 2.59, + "learning_rate": 6.772890484739677e-06, + "loss": 0.2044, + "step": 67467 + }, + { + "epoch": 2.59, + "learning_rate": 6.772249294690946e-06, + "loss": 0.1419, + "step": 67468 + }, + { + "epoch": 2.59, + "learning_rate": 6.7716081046422156e-06, + "loss": 0.2674, + "step": 67469 + }, + { + "epoch": 2.59, + "learning_rate": 6.7709669145934864e-06, + "loss": 0.1527, + "step": 67470 + }, + { + "epoch": 2.59, + "learning_rate": 6.770325724544756e-06, + "loss": 0.2752, + "step": 67471 + }, + { + "epoch": 2.59, + "learning_rate": 6.769684534496025e-06, + "loss": 0.2622, + "step": 67472 + }, + { + "epoch": 2.59, + "learning_rate": 6.769043344447294e-06, + "loss": 0.2883, + "step": 67473 + }, + { + "epoch": 2.59, + "learning_rate": 6.768402154398563e-06, + "loss": 0.1473, + "step": 67474 + }, + { + "epoch": 2.59, + "learning_rate": 6.767760964349834e-06, + "loss": 0.2933, + "step": 67475 + }, + { + "epoch": 2.59, + "learning_rate": 6.767119774301103e-06, + "loss": 0.1882, + "step": 67476 + }, + { + "epoch": 2.59, + "learning_rate": 6.766478584252372e-06, + "loss": 0.26, + "step": 67477 + }, + { + "epoch": 2.59, + "learning_rate": 6.7658373942036415e-06, + "loss": 0.2737, + "step": 67478 + }, + { + "epoch": 2.59, + "learning_rate": 6.765196204154912e-06, + "loss": 0.2342, + "step": 67479 + }, + { + "epoch": 2.59, + "learning_rate": 6.764555014106182e-06, + "loss": 0.2299, + "step": 67480 + }, + { + "epoch": 2.59, + "learning_rate": 6.763913824057451e-06, + "loss": 0.228, + "step": 67481 + }, + { + "epoch": 2.59, + "learning_rate": 6.76327263400872e-06, + "loss": 0.1704, + "step": 67482 + }, + { + "epoch": 2.59, + "learning_rate": 6.762631443959989e-06, + "loss": 0.2427, + "step": 67483 + }, + { + "epoch": 2.59, + "learning_rate": 6.76199025391126e-06, + "loss": 0.2259, + "step": 67484 + }, + { + "epoch": 2.59, + "learning_rate": 6.761349063862529e-06, + "loss": 0.2143, + "step": 67485 + }, + { + "epoch": 2.59, + "learning_rate": 6.760707873813798e-06, + "loss": 0.2376, + "step": 67486 + }, + { + "epoch": 2.59, + "learning_rate": 6.7600666837650675e-06, + "loss": 0.2738, + "step": 67487 + }, + { + "epoch": 2.59, + "learning_rate": 6.759425493716338e-06, + "loss": 0.2937, + "step": 67488 + }, + { + "epoch": 2.59, + "learning_rate": 6.7587843036676076e-06, + "loss": 0.2485, + "step": 67489 + }, + { + "epoch": 2.59, + "learning_rate": 6.758143113618877e-06, + "loss": 0.2572, + "step": 67490 + }, + { + "epoch": 2.59, + "learning_rate": 6.757501923570146e-06, + "loss": 0.1416, + "step": 67491 + }, + { + "epoch": 2.59, + "learning_rate": 6.756860733521415e-06, + "loss": 0.2108, + "step": 67492 + }, + { + "epoch": 2.59, + "learning_rate": 6.756219543472686e-06, + "loss": 0.2275, + "step": 67493 + }, + { + "epoch": 2.59, + "learning_rate": 6.755578353423955e-06, + "loss": 0.165, + "step": 67494 + }, + { + "epoch": 2.59, + "learning_rate": 6.754937163375224e-06, + "loss": 0.1948, + "step": 67495 + }, + { + "epoch": 2.6, + "learning_rate": 6.7542959733264935e-06, + "loss": 0.0811, + "step": 67496 + }, + { + "epoch": 2.6, + "learning_rate": 6.753654783277764e-06, + "loss": 0.181, + "step": 67497 + }, + { + "epoch": 2.6, + "learning_rate": 6.7530135932290335e-06, + "loss": 0.2792, + "step": 67498 + }, + { + "epoch": 2.6, + "learning_rate": 6.752372403180303e-06, + "loss": 0.1616, + "step": 67499 + }, + { + "epoch": 2.6, + "learning_rate": 6.751731213131572e-06, + "loss": 0.2008, + "step": 67500 + }, + { + "epoch": 2.6, + "learning_rate": 6.751090023082841e-06, + "loss": 0.2254, + "step": 67501 + }, + { + "epoch": 2.6, + "learning_rate": 6.750448833034112e-06, + "loss": 0.1909, + "step": 67502 + }, + { + "epoch": 2.6, + "learning_rate": 6.749807642985381e-06, + "loss": 0.2474, + "step": 67503 + }, + { + "epoch": 2.6, + "learning_rate": 6.74916645293665e-06, + "loss": 0.1117, + "step": 67504 + }, + { + "epoch": 2.6, + "learning_rate": 6.7485252628879194e-06, + "loss": 0.2698, + "step": 67505 + }, + { + "epoch": 2.6, + "learning_rate": 6.74788407283919e-06, + "loss": 0.2068, + "step": 67506 + }, + { + "epoch": 2.6, + "learning_rate": 6.7472428827904595e-06, + "loss": 0.1924, + "step": 67507 + }, + { + "epoch": 2.6, + "learning_rate": 6.746601692741729e-06, + "loss": 0.2947, + "step": 67508 + }, + { + "epoch": 2.6, + "learning_rate": 6.745960502692998e-06, + "loss": 0.1258, + "step": 67509 + }, + { + "epoch": 2.6, + "learning_rate": 6.745319312644269e-06, + "loss": 0.1995, + "step": 67510 + }, + { + "epoch": 2.6, + "learning_rate": 6.744678122595538e-06, + "loss": 0.3138, + "step": 67511 + }, + { + "epoch": 2.6, + "learning_rate": 6.744036932546807e-06, + "loss": 0.1109, + "step": 67512 + }, + { + "epoch": 2.6, + "learning_rate": 6.743395742498076e-06, + "loss": 0.2264, + "step": 67513 + }, + { + "epoch": 2.6, + "learning_rate": 6.742754552449345e-06, + "loss": 0.1962, + "step": 67514 + }, + { + "epoch": 2.6, + "learning_rate": 6.742113362400616e-06, + "loss": 0.215, + "step": 67515 + }, + { + "epoch": 2.6, + "learning_rate": 6.7414721723518855e-06, + "loss": 0.2786, + "step": 67516 + }, + { + "epoch": 2.6, + "learning_rate": 6.740830982303155e-06, + "loss": 0.2796, + "step": 67517 + }, + { + "epoch": 2.6, + "learning_rate": 6.740189792254424e-06, + "loss": 0.1583, + "step": 67518 + }, + { + "epoch": 2.6, + "learning_rate": 6.739548602205695e-06, + "loss": 0.2307, + "step": 67519 + }, + { + "epoch": 2.6, + "learning_rate": 6.738907412156964e-06, + "loss": 0.2872, + "step": 67520 + }, + { + "epoch": 2.6, + "learning_rate": 6.738266222108233e-06, + "loss": 0.1796, + "step": 67521 + }, + { + "epoch": 2.6, + "learning_rate": 6.737625032059502e-06, + "loss": 0.1673, + "step": 67522 + }, + { + "epoch": 2.6, + "learning_rate": 6.736983842010771e-06, + "loss": 0.1693, + "step": 67523 + }, + { + "epoch": 2.6, + "learning_rate": 6.736342651962042e-06, + "loss": 0.1893, + "step": 67524 + }, + { + "epoch": 2.6, + "learning_rate": 6.7357014619133114e-06, + "loss": 0.1289, + "step": 67525 + }, + { + "epoch": 2.6, + "learning_rate": 6.735060271864581e-06, + "loss": 0.2702, + "step": 67526 + }, + { + "epoch": 2.6, + "learning_rate": 6.73441908181585e-06, + "loss": 0.2251, + "step": 67527 + }, + { + "epoch": 2.6, + "learning_rate": 6.733777891767121e-06, + "loss": 0.3664, + "step": 67528 + }, + { + "epoch": 2.6, + "learning_rate": 6.73313670171839e-06, + "loss": 0.2869, + "step": 67529 + }, + { + "epoch": 2.6, + "learning_rate": 6.732495511669659e-06, + "loss": 0.2495, + "step": 67530 + }, + { + "epoch": 2.6, + "learning_rate": 6.731854321620928e-06, + "loss": 0.0757, + "step": 67531 + }, + { + "epoch": 2.6, + "learning_rate": 6.731213131572197e-06, + "loss": 0.3021, + "step": 67532 + }, + { + "epoch": 2.6, + "learning_rate": 6.730571941523468e-06, + "loss": 0.231, + "step": 67533 + }, + { + "epoch": 2.6, + "learning_rate": 6.729930751474737e-06, + "loss": 0.2499, + "step": 67534 + }, + { + "epoch": 2.6, + "learning_rate": 6.7292895614260066e-06, + "loss": 0.2434, + "step": 67535 + }, + { + "epoch": 2.6, + "learning_rate": 6.728648371377276e-06, + "loss": 0.1348, + "step": 67536 + }, + { + "epoch": 2.6, + "learning_rate": 6.728007181328547e-06, + "loss": 0.1573, + "step": 67537 + }, + { + "epoch": 2.6, + "learning_rate": 6.727365991279816e-06, + "loss": 0.0828, + "step": 67538 + }, + { + "epoch": 2.6, + "learning_rate": 6.726724801231085e-06, + "loss": 0.1915, + "step": 67539 + }, + { + "epoch": 2.6, + "learning_rate": 6.726083611182354e-06, + "loss": 0.1816, + "step": 67540 + }, + { + "epoch": 2.6, + "learning_rate": 6.725442421133623e-06, + "loss": 0.3347, + "step": 67541 + }, + { + "epoch": 2.6, + "learning_rate": 6.724801231084894e-06, + "loss": 0.1717, + "step": 67542 + }, + { + "epoch": 2.6, + "learning_rate": 6.724160041036163e-06, + "loss": 0.2339, + "step": 67543 + }, + { + "epoch": 2.6, + "learning_rate": 6.7235188509874325e-06, + "loss": 0.1576, + "step": 67544 + }, + { + "epoch": 2.6, + "learning_rate": 6.722877660938702e-06, + "loss": 0.1834, + "step": 67545 + }, + { + "epoch": 2.6, + "learning_rate": 6.722236470889973e-06, + "loss": 0.2237, + "step": 67546 + }, + { + "epoch": 2.6, + "learning_rate": 6.721595280841242e-06, + "loss": 0.2436, + "step": 67547 + }, + { + "epoch": 2.6, + "learning_rate": 6.720954090792511e-06, + "loss": 0.1395, + "step": 67548 + }, + { + "epoch": 2.6, + "learning_rate": 6.72031290074378e-06, + "loss": 0.1404, + "step": 67549 + }, + { + "epoch": 2.6, + "learning_rate": 6.719671710695051e-06, + "loss": 0.1444, + "step": 67550 + }, + { + "epoch": 2.6, + "learning_rate": 6.71903052064632e-06, + "loss": 0.1637, + "step": 67551 + }, + { + "epoch": 2.6, + "learning_rate": 6.718389330597589e-06, + "loss": 0.2468, + "step": 67552 + }, + { + "epoch": 2.6, + "learning_rate": 6.7177481405488585e-06, + "loss": 0.1503, + "step": 67553 + }, + { + "epoch": 2.6, + "learning_rate": 6.717106950500128e-06, + "loss": 0.2001, + "step": 67554 + }, + { + "epoch": 2.6, + "learning_rate": 6.7164657604513985e-06, + "loss": 0.2528, + "step": 67555 + }, + { + "epoch": 2.6, + "learning_rate": 6.715824570402668e-06, + "loss": 0.2276, + "step": 67556 + }, + { + "epoch": 2.6, + "learning_rate": 6.715183380353937e-06, + "loss": 0.321, + "step": 67557 + }, + { + "epoch": 2.6, + "learning_rate": 6.714542190305206e-06, + "loss": 0.3492, + "step": 67558 + }, + { + "epoch": 2.6, + "learning_rate": 6.713901000256477e-06, + "loss": 0.1625, + "step": 67559 + }, + { + "epoch": 2.6, + "learning_rate": 6.713259810207746e-06, + "loss": 0.1873, + "step": 67560 + }, + { + "epoch": 2.6, + "learning_rate": 6.712618620159015e-06, + "loss": 0.2757, + "step": 67561 + }, + { + "epoch": 2.6, + "learning_rate": 6.7119774301102845e-06, + "loss": 0.1123, + "step": 67562 + }, + { + "epoch": 2.6, + "learning_rate": 6.711336240061554e-06, + "loss": 0.243, + "step": 67563 + }, + { + "epoch": 2.6, + "learning_rate": 6.7106950500128245e-06, + "loss": 0.1655, + "step": 67564 + }, + { + "epoch": 2.6, + "learning_rate": 6.710053859964094e-06, + "loss": 0.2255, + "step": 67565 + }, + { + "epoch": 2.6, + "learning_rate": 6.709412669915363e-06, + "loss": 0.2404, + "step": 67566 + }, + { + "epoch": 2.6, + "learning_rate": 6.708771479866632e-06, + "loss": 0.1908, + "step": 67567 + }, + { + "epoch": 2.6, + "learning_rate": 6.708130289817903e-06, + "loss": 0.2272, + "step": 67568 + }, + { + "epoch": 2.6, + "learning_rate": 6.707489099769172e-06, + "loss": 0.1335, + "step": 67569 + }, + { + "epoch": 2.6, + "learning_rate": 6.706847909720441e-06, + "loss": 0.1324, + "step": 67570 + }, + { + "epoch": 2.6, + "learning_rate": 6.7062067196717104e-06, + "loss": 0.2332, + "step": 67571 + }, + { + "epoch": 2.6, + "learning_rate": 6.70556552962298e-06, + "loss": 0.2126, + "step": 67572 + }, + { + "epoch": 2.6, + "learning_rate": 6.7049243395742505e-06, + "loss": 0.2905, + "step": 67573 + }, + { + "epoch": 2.6, + "learning_rate": 6.70428314952552e-06, + "loss": 0.2235, + "step": 67574 + }, + { + "epoch": 2.6, + "learning_rate": 6.703641959476789e-06, + "loss": 0.3261, + "step": 67575 + }, + { + "epoch": 2.6, + "learning_rate": 6.703000769428058e-06, + "loss": 0.1422, + "step": 67576 + }, + { + "epoch": 2.6, + "learning_rate": 6.702359579379329e-06, + "loss": 0.1782, + "step": 67577 + }, + { + "epoch": 2.6, + "learning_rate": 6.701718389330598e-06, + "loss": 0.2368, + "step": 67578 + }, + { + "epoch": 2.6, + "learning_rate": 6.701077199281867e-06, + "loss": 0.1665, + "step": 67579 + }, + { + "epoch": 2.6, + "learning_rate": 6.700436009233136e-06, + "loss": 0.2431, + "step": 67580 + }, + { + "epoch": 2.6, + "learning_rate": 6.699794819184406e-06, + "loss": 0.2268, + "step": 67581 + }, + { + "epoch": 2.6, + "learning_rate": 6.6991536291356765e-06, + "loss": 0.2129, + "step": 67582 + }, + { + "epoch": 2.6, + "learning_rate": 6.698512439086946e-06, + "loss": 0.2027, + "step": 67583 + }, + { + "epoch": 2.6, + "learning_rate": 6.697871249038215e-06, + "loss": 0.2311, + "step": 67584 + }, + { + "epoch": 2.6, + "learning_rate": 6.697230058989484e-06, + "loss": 0.2526, + "step": 67585 + }, + { + "epoch": 2.6, + "learning_rate": 6.696588868940755e-06, + "loss": 0.1639, + "step": 67586 + }, + { + "epoch": 2.6, + "learning_rate": 6.695947678892024e-06, + "loss": 0.1277, + "step": 67587 + }, + { + "epoch": 2.6, + "learning_rate": 6.695306488843293e-06, + "loss": 0.1855, + "step": 67588 + }, + { + "epoch": 2.6, + "learning_rate": 6.694665298794562e-06, + "loss": 0.1528, + "step": 67589 + }, + { + "epoch": 2.6, + "learning_rate": 6.6940241087458316e-06, + "loss": 0.249, + "step": 67590 + }, + { + "epoch": 2.6, + "learning_rate": 6.693382918697102e-06, + "loss": 0.11, + "step": 67591 + }, + { + "epoch": 2.6, + "learning_rate": 6.692741728648372e-06, + "loss": 0.2442, + "step": 67592 + }, + { + "epoch": 2.6, + "learning_rate": 6.692100538599641e-06, + "loss": 0.2374, + "step": 67593 + }, + { + "epoch": 2.6, + "learning_rate": 6.69145934855091e-06, + "loss": 0.1982, + "step": 67594 + }, + { + "epoch": 2.6, + "learning_rate": 6.690818158502181e-06, + "loss": 0.2142, + "step": 67595 + }, + { + "epoch": 2.6, + "learning_rate": 6.69017696845345e-06, + "loss": 0.2468, + "step": 67596 + }, + { + "epoch": 2.6, + "learning_rate": 6.689535778404719e-06, + "loss": 0.1562, + "step": 67597 + }, + { + "epoch": 2.6, + "learning_rate": 6.688894588355988e-06, + "loss": 0.182, + "step": 67598 + }, + { + "epoch": 2.6, + "learning_rate": 6.688253398307259e-06, + "loss": 0.2092, + "step": 67599 + }, + { + "epoch": 2.6, + "learning_rate": 6.687612208258528e-06, + "loss": 0.2193, + "step": 67600 + }, + { + "epoch": 2.6, + "learning_rate": 6.6869710182097976e-06, + "loss": 0.3675, + "step": 67601 + }, + { + "epoch": 2.6, + "learning_rate": 6.686329828161067e-06, + "loss": 0.2228, + "step": 67602 + }, + { + "epoch": 2.6, + "learning_rate": 6.685688638112336e-06, + "loss": 0.2553, + "step": 67603 + }, + { + "epoch": 2.6, + "learning_rate": 6.685047448063607e-06, + "loss": 0.162, + "step": 67604 + }, + { + "epoch": 2.6, + "learning_rate": 6.684406258014876e-06, + "loss": 0.1155, + "step": 67605 + }, + { + "epoch": 2.6, + "learning_rate": 6.683765067966145e-06, + "loss": 0.1688, + "step": 67606 + }, + { + "epoch": 2.6, + "learning_rate": 6.683123877917414e-06, + "loss": 0.2654, + "step": 67607 + }, + { + "epoch": 2.6, + "learning_rate": 6.682482687868685e-06, + "loss": 0.2392, + "step": 67608 + }, + { + "epoch": 2.6, + "learning_rate": 6.681841497819954e-06, + "loss": 0.2042, + "step": 67609 + }, + { + "epoch": 2.6, + "learning_rate": 6.6812003077712235e-06, + "loss": 0.2524, + "step": 67610 + }, + { + "epoch": 2.6, + "learning_rate": 6.680559117722493e-06, + "loss": 0.2691, + "step": 67611 + }, + { + "epoch": 2.6, + "learning_rate": 6.679917927673762e-06, + "loss": 0.2172, + "step": 67612 + }, + { + "epoch": 2.6, + "learning_rate": 6.679276737625033e-06, + "loss": 0.2133, + "step": 67613 + }, + { + "epoch": 2.6, + "learning_rate": 6.678635547576302e-06, + "loss": 0.2045, + "step": 67614 + }, + { + "epoch": 2.6, + "learning_rate": 6.677994357527571e-06, + "loss": 0.3081, + "step": 67615 + }, + { + "epoch": 2.6, + "learning_rate": 6.67735316747884e-06, + "loss": 0.1757, + "step": 67616 + }, + { + "epoch": 2.6, + "learning_rate": 6.676711977430111e-06, + "loss": 0.3376, + "step": 67617 + }, + { + "epoch": 2.6, + "learning_rate": 6.67607078738138e-06, + "loss": 0.1486, + "step": 67618 + }, + { + "epoch": 2.6, + "learning_rate": 6.6754295973326495e-06, + "loss": 0.1951, + "step": 67619 + }, + { + "epoch": 2.6, + "learning_rate": 6.674788407283919e-06, + "loss": 0.2278, + "step": 67620 + }, + { + "epoch": 2.6, + "learning_rate": 6.674147217235188e-06, + "loss": 0.1867, + "step": 67621 + }, + { + "epoch": 2.6, + "learning_rate": 6.673506027186459e-06, + "loss": 0.2598, + "step": 67622 + }, + { + "epoch": 2.6, + "learning_rate": 6.672864837137728e-06, + "loss": 0.1266, + "step": 67623 + }, + { + "epoch": 2.6, + "learning_rate": 6.672223647088997e-06, + "loss": 0.1388, + "step": 67624 + }, + { + "epoch": 2.6, + "learning_rate": 6.671582457040266e-06, + "loss": 0.3705, + "step": 67625 + }, + { + "epoch": 2.6, + "learning_rate": 6.670941266991537e-06, + "loss": 0.2102, + "step": 67626 + }, + { + "epoch": 2.6, + "learning_rate": 6.670300076942806e-06, + "loss": 0.1566, + "step": 67627 + }, + { + "epoch": 2.6, + "learning_rate": 6.6696588868940755e-06, + "loss": 0.2302, + "step": 67628 + }, + { + "epoch": 2.6, + "learning_rate": 6.669017696845345e-06, + "loss": 0.2358, + "step": 67629 + }, + { + "epoch": 2.6, + "learning_rate": 6.668376506796614e-06, + "loss": 0.274, + "step": 67630 + }, + { + "epoch": 2.6, + "learning_rate": 6.667735316747885e-06, + "loss": 0.2467, + "step": 67631 + }, + { + "epoch": 2.6, + "learning_rate": 6.667094126699154e-06, + "loss": 0.2935, + "step": 67632 + }, + { + "epoch": 2.6, + "learning_rate": 6.666452936650423e-06, + "loss": 0.1529, + "step": 67633 + }, + { + "epoch": 2.6, + "learning_rate": 6.665811746601692e-06, + "loss": 0.3391, + "step": 67634 + }, + { + "epoch": 2.6, + "learning_rate": 6.665170556552963e-06, + "loss": 0.2628, + "step": 67635 + }, + { + "epoch": 2.6, + "learning_rate": 6.664529366504232e-06, + "loss": 0.1731, + "step": 67636 + }, + { + "epoch": 2.6, + "learning_rate": 6.6638881764555014e-06, + "loss": 0.1322, + "step": 67637 + }, + { + "epoch": 2.6, + "learning_rate": 6.663246986406771e-06, + "loss": 0.215, + "step": 67638 + }, + { + "epoch": 2.6, + "learning_rate": 6.66260579635804e-06, + "loss": 0.1999, + "step": 67639 + }, + { + "epoch": 2.6, + "learning_rate": 6.661964606309311e-06, + "loss": 0.1939, + "step": 67640 + }, + { + "epoch": 2.6, + "learning_rate": 6.66132341626058e-06, + "loss": 0.2269, + "step": 67641 + }, + { + "epoch": 2.6, + "learning_rate": 6.660682226211849e-06, + "loss": 0.1629, + "step": 67642 + }, + { + "epoch": 2.6, + "learning_rate": 6.660041036163118e-06, + "loss": 0.1328, + "step": 67643 + }, + { + "epoch": 2.6, + "learning_rate": 6.659399846114389e-06, + "loss": 0.1961, + "step": 67644 + }, + { + "epoch": 2.6, + "learning_rate": 6.658758656065658e-06, + "loss": 0.1824, + "step": 67645 + }, + { + "epoch": 2.6, + "learning_rate": 6.658117466016927e-06, + "loss": 0.204, + "step": 67646 + }, + { + "epoch": 2.6, + "learning_rate": 6.657476275968197e-06, + "loss": 0.2364, + "step": 67647 + }, + { + "epoch": 2.6, + "learning_rate": 6.6568350859194674e-06, + "loss": 0.1309, + "step": 67648 + }, + { + "epoch": 2.6, + "learning_rate": 6.656193895870737e-06, + "loss": 0.1887, + "step": 67649 + }, + { + "epoch": 2.6, + "learning_rate": 6.655552705822006e-06, + "loss": 0.3559, + "step": 67650 + }, + { + "epoch": 2.6, + "learning_rate": 6.654911515773275e-06, + "loss": 0.311, + "step": 67651 + }, + { + "epoch": 2.6, + "learning_rate": 6.654270325724544e-06, + "loss": 0.1156, + "step": 67652 + }, + { + "epoch": 2.6, + "learning_rate": 6.653629135675815e-06, + "loss": 0.3015, + "step": 67653 + }, + { + "epoch": 2.6, + "learning_rate": 6.652987945627084e-06, + "loss": 0.2317, + "step": 67654 + }, + { + "epoch": 2.6, + "learning_rate": 6.652346755578353e-06, + "loss": 0.1629, + "step": 67655 + }, + { + "epoch": 2.6, + "learning_rate": 6.6517055655296225e-06, + "loss": 0.1916, + "step": 67656 + }, + { + "epoch": 2.6, + "learning_rate": 6.651064375480893e-06, + "loss": 0.1521, + "step": 67657 + }, + { + "epoch": 2.6, + "learning_rate": 6.650423185432163e-06, + "loss": 0.1582, + "step": 67658 + }, + { + "epoch": 2.6, + "learning_rate": 6.649781995383432e-06, + "loss": 0.1891, + "step": 67659 + }, + { + "epoch": 2.6, + "learning_rate": 6.649140805334701e-06, + "loss": 0.2777, + "step": 67660 + }, + { + "epoch": 2.6, + "learning_rate": 6.64849961528597e-06, + "loss": 0.2938, + "step": 67661 + }, + { + "epoch": 2.6, + "learning_rate": 6.647858425237241e-06, + "loss": 0.2047, + "step": 67662 + }, + { + "epoch": 2.6, + "learning_rate": 6.64721723518851e-06, + "loss": 0.1564, + "step": 67663 + }, + { + "epoch": 2.6, + "learning_rate": 6.646576045139779e-06, + "loss": 0.1301, + "step": 67664 + }, + { + "epoch": 2.6, + "learning_rate": 6.6459348550910485e-06, + "loss": 0.2089, + "step": 67665 + }, + { + "epoch": 2.6, + "learning_rate": 6.645293665042319e-06, + "loss": 0.2072, + "step": 67666 + }, + { + "epoch": 2.6, + "learning_rate": 6.6446524749935886e-06, + "loss": 0.281, + "step": 67667 + }, + { + "epoch": 2.6, + "learning_rate": 6.644011284944858e-06, + "loss": 0.1092, + "step": 67668 + }, + { + "epoch": 2.6, + "learning_rate": 6.643370094896127e-06, + "loss": 0.2233, + "step": 67669 + }, + { + "epoch": 2.6, + "learning_rate": 6.642728904847396e-06, + "loss": 0.2645, + "step": 67670 + }, + { + "epoch": 2.6, + "learning_rate": 6.642087714798667e-06, + "loss": 0.2397, + "step": 67671 + }, + { + "epoch": 2.6, + "learning_rate": 6.641446524749936e-06, + "loss": 0.0968, + "step": 67672 + }, + { + "epoch": 2.6, + "learning_rate": 6.640805334701205e-06, + "loss": 0.1819, + "step": 67673 + }, + { + "epoch": 2.6, + "learning_rate": 6.6401641446524745e-06, + "loss": 0.2771, + "step": 67674 + }, + { + "epoch": 2.6, + "learning_rate": 6.639522954603745e-06, + "loss": 0.1815, + "step": 67675 + }, + { + "epoch": 2.6, + "learning_rate": 6.6388817645550145e-06, + "loss": 0.2248, + "step": 67676 + }, + { + "epoch": 2.6, + "learning_rate": 6.638240574506284e-06, + "loss": 0.3354, + "step": 67677 + }, + { + "epoch": 2.6, + "learning_rate": 6.637599384457553e-06, + "loss": 0.2293, + "step": 67678 + }, + { + "epoch": 2.6, + "learning_rate": 6.636958194408822e-06, + "loss": 0.1967, + "step": 67679 + }, + { + "epoch": 2.6, + "learning_rate": 6.636317004360093e-06, + "loss": 0.2753, + "step": 67680 + }, + { + "epoch": 2.6, + "learning_rate": 6.635675814311362e-06, + "loss": 0.2174, + "step": 67681 + }, + { + "epoch": 2.6, + "learning_rate": 6.635034624262631e-06, + "loss": 0.2753, + "step": 67682 + }, + { + "epoch": 2.6, + "learning_rate": 6.6343934342139005e-06, + "loss": 0.324, + "step": 67683 + }, + { + "epoch": 2.6, + "learning_rate": 6.633752244165171e-06, + "loss": 0.1948, + "step": 67684 + }, + { + "epoch": 2.6, + "learning_rate": 6.6331110541164405e-06, + "loss": 0.2054, + "step": 67685 + }, + { + "epoch": 2.6, + "learning_rate": 6.63246986406771e-06, + "loss": 0.2661, + "step": 67686 + }, + { + "epoch": 2.6, + "learning_rate": 6.631828674018979e-06, + "loss": 0.2461, + "step": 67687 + }, + { + "epoch": 2.6, + "learning_rate": 6.631187483970248e-06, + "loss": 0.2641, + "step": 67688 + }, + { + "epoch": 2.6, + "learning_rate": 6.630546293921519e-06, + "loss": 0.2748, + "step": 67689 + }, + { + "epoch": 2.6, + "learning_rate": 6.629905103872788e-06, + "loss": 0.2628, + "step": 67690 + }, + { + "epoch": 2.6, + "learning_rate": 6.629263913824057e-06, + "loss": 0.2013, + "step": 67691 + }, + { + "epoch": 2.6, + "learning_rate": 6.628622723775326e-06, + "loss": 0.2267, + "step": 67692 + }, + { + "epoch": 2.6, + "learning_rate": 6.627981533726597e-06, + "loss": 0.1663, + "step": 67693 + }, + { + "epoch": 2.6, + "learning_rate": 6.6273403436778665e-06, + "loss": 0.2522, + "step": 67694 + }, + { + "epoch": 2.6, + "learning_rate": 6.626699153629136e-06, + "loss": 0.2963, + "step": 67695 + }, + { + "epoch": 2.6, + "learning_rate": 6.626057963580405e-06, + "loss": 0.1189, + "step": 67696 + }, + { + "epoch": 2.6, + "learning_rate": 6.625416773531676e-06, + "loss": 0.1865, + "step": 67697 + }, + { + "epoch": 2.6, + "learning_rate": 6.624775583482945e-06, + "loss": 0.2502, + "step": 67698 + }, + { + "epoch": 2.6, + "learning_rate": 6.624134393434214e-06, + "loss": 0.2232, + "step": 67699 + }, + { + "epoch": 2.6, + "learning_rate": 6.623493203385483e-06, + "loss": 0.2438, + "step": 67700 + }, + { + "epoch": 2.6, + "learning_rate": 6.622852013336752e-06, + "loss": 0.2316, + "step": 67701 + }, + { + "epoch": 2.6, + "learning_rate": 6.622210823288023e-06, + "loss": 0.1916, + "step": 67702 + }, + { + "epoch": 2.6, + "learning_rate": 6.6215696332392924e-06, + "loss": 0.1535, + "step": 67703 + }, + { + "epoch": 2.6, + "learning_rate": 6.620928443190562e-06, + "loss": 0.1749, + "step": 67704 + }, + { + "epoch": 2.6, + "learning_rate": 6.620287253141831e-06, + "loss": 0.2431, + "step": 67705 + }, + { + "epoch": 2.6, + "learning_rate": 6.619646063093102e-06, + "loss": 0.305, + "step": 67706 + }, + { + "epoch": 2.6, + "learning_rate": 6.619004873044371e-06, + "loss": 0.2291, + "step": 67707 + }, + { + "epoch": 2.6, + "learning_rate": 6.61836368299564e-06, + "loss": 0.0918, + "step": 67708 + }, + { + "epoch": 2.6, + "learning_rate": 6.617722492946909e-06, + "loss": 0.187, + "step": 67709 + }, + { + "epoch": 2.6, + "learning_rate": 6.617081302898178e-06, + "loss": 0.2116, + "step": 67710 + }, + { + "epoch": 2.6, + "learning_rate": 6.616440112849449e-06, + "loss": 0.196, + "step": 67711 + }, + { + "epoch": 2.6, + "learning_rate": 6.615798922800718e-06, + "loss": 0.1814, + "step": 67712 + }, + { + "epoch": 2.6, + "learning_rate": 6.6151577327519876e-06, + "loss": 0.3479, + "step": 67713 + }, + { + "epoch": 2.6, + "learning_rate": 6.614516542703257e-06, + "loss": 0.1744, + "step": 67714 + }, + { + "epoch": 2.6, + "learning_rate": 6.613875352654528e-06, + "loss": 0.2587, + "step": 67715 + }, + { + "epoch": 2.6, + "learning_rate": 6.613234162605797e-06, + "loss": 0.2757, + "step": 67716 + }, + { + "epoch": 2.6, + "learning_rate": 6.612592972557066e-06, + "loss": 0.2699, + "step": 67717 + }, + { + "epoch": 2.6, + "learning_rate": 6.611951782508335e-06, + "loss": 0.2123, + "step": 67718 + }, + { + "epoch": 2.6, + "learning_rate": 6.611310592459604e-06, + "loss": 0.2157, + "step": 67719 + }, + { + "epoch": 2.6, + "learning_rate": 6.610669402410875e-06, + "loss": 0.1015, + "step": 67720 + }, + { + "epoch": 2.6, + "learning_rate": 6.610028212362144e-06, + "loss": 0.2549, + "step": 67721 + }, + { + "epoch": 2.6, + "learning_rate": 6.6093870223134135e-06, + "loss": 0.2799, + "step": 67722 + }, + { + "epoch": 2.6, + "learning_rate": 6.608745832264683e-06, + "loss": 0.4073, + "step": 67723 + }, + { + "epoch": 2.6, + "learning_rate": 6.608104642215954e-06, + "loss": 0.3427, + "step": 67724 + }, + { + "epoch": 2.6, + "learning_rate": 6.607463452167223e-06, + "loss": 0.1121, + "step": 67725 + }, + { + "epoch": 2.6, + "learning_rate": 6.606822262118492e-06, + "loss": 0.1997, + "step": 67726 + }, + { + "epoch": 2.6, + "learning_rate": 6.606181072069761e-06, + "loss": 0.0953, + "step": 67727 + }, + { + "epoch": 2.6, + "learning_rate": 6.60553988202103e-06, + "loss": 0.1587, + "step": 67728 + }, + { + "epoch": 2.6, + "learning_rate": 6.604898691972301e-06, + "loss": 0.1518, + "step": 67729 + }, + { + "epoch": 2.6, + "learning_rate": 6.60425750192357e-06, + "loss": 0.1118, + "step": 67730 + }, + { + "epoch": 2.6, + "learning_rate": 6.6036163118748395e-06, + "loss": 0.1349, + "step": 67731 + }, + { + "epoch": 2.6, + "learning_rate": 6.602975121826109e-06, + "loss": 0.2229, + "step": 67732 + }, + { + "epoch": 2.6, + "learning_rate": 6.6023339317773796e-06, + "loss": 0.2965, + "step": 67733 + }, + { + "epoch": 2.6, + "learning_rate": 6.601692741728649e-06, + "loss": 0.1734, + "step": 67734 + }, + { + "epoch": 2.6, + "learning_rate": 6.601051551679918e-06, + "loss": 0.1984, + "step": 67735 + }, + { + "epoch": 2.6, + "learning_rate": 6.600410361631187e-06, + "loss": 0.1914, + "step": 67736 + }, + { + "epoch": 2.6, + "learning_rate": 6.599769171582456e-06, + "loss": 0.2029, + "step": 67737 + }, + { + "epoch": 2.6, + "learning_rate": 6.599127981533727e-06, + "loss": 0.273, + "step": 67738 + }, + { + "epoch": 2.6, + "learning_rate": 6.598486791484996e-06, + "loss": 0.1192, + "step": 67739 + }, + { + "epoch": 2.6, + "learning_rate": 6.5978456014362655e-06, + "loss": 0.3566, + "step": 67740 + }, + { + "epoch": 2.6, + "learning_rate": 6.597204411387535e-06, + "loss": 0.1514, + "step": 67741 + }, + { + "epoch": 2.6, + "learning_rate": 6.5965632213388055e-06, + "loss": 0.267, + "step": 67742 + }, + { + "epoch": 2.6, + "learning_rate": 6.595922031290075e-06, + "loss": 0.3091, + "step": 67743 + }, + { + "epoch": 2.6, + "learning_rate": 6.595280841241344e-06, + "loss": 0.2519, + "step": 67744 + }, + { + "epoch": 2.6, + "learning_rate": 6.594639651192613e-06, + "loss": 0.1779, + "step": 67745 + }, + { + "epoch": 2.6, + "learning_rate": 6.593998461143884e-06, + "loss": 0.1727, + "step": 67746 + }, + { + "epoch": 2.6, + "learning_rate": 6.593357271095153e-06, + "loss": 0.1864, + "step": 67747 + }, + { + "epoch": 2.6, + "learning_rate": 6.592716081046422e-06, + "loss": 0.2178, + "step": 67748 + }, + { + "epoch": 2.6, + "learning_rate": 6.5920748909976914e-06, + "loss": 0.2318, + "step": 67749 + }, + { + "epoch": 2.6, + "learning_rate": 6.591433700948961e-06, + "loss": 0.1126, + "step": 67750 + }, + { + "epoch": 2.6, + "learning_rate": 6.5907925109002315e-06, + "loss": 0.2989, + "step": 67751 + }, + { + "epoch": 2.6, + "learning_rate": 6.590151320851501e-06, + "loss": 0.1706, + "step": 67752 + }, + { + "epoch": 2.6, + "learning_rate": 6.58951013080277e-06, + "loss": 0.2273, + "step": 67753 + }, + { + "epoch": 2.6, + "learning_rate": 6.588868940754039e-06, + "loss": 0.2312, + "step": 67754 + }, + { + "epoch": 2.6, + "learning_rate": 6.58822775070531e-06, + "loss": 0.1785, + "step": 67755 + }, + { + "epoch": 2.6, + "learning_rate": 6.587586560656579e-06, + "loss": 0.203, + "step": 67756 + }, + { + "epoch": 2.61, + "learning_rate": 6.586945370607848e-06, + "loss": 0.1597, + "step": 67757 + }, + { + "epoch": 2.61, + "learning_rate": 6.586304180559117e-06, + "loss": 0.1731, + "step": 67758 + }, + { + "epoch": 2.61, + "learning_rate": 6.585662990510387e-06, + "loss": 0.1819, + "step": 67759 + }, + { + "epoch": 2.61, + "learning_rate": 6.5850218004616575e-06, + "loss": 0.1336, + "step": 67760 + }, + { + "epoch": 2.61, + "learning_rate": 6.584380610412927e-06, + "loss": 0.3058, + "step": 67761 + }, + { + "epoch": 2.61, + "learning_rate": 6.583739420364196e-06, + "loss": 0.1794, + "step": 67762 + }, + { + "epoch": 2.61, + "learning_rate": 6.583098230315465e-06, + "loss": 0.3235, + "step": 67763 + }, + { + "epoch": 2.61, + "learning_rate": 6.582457040266736e-06, + "loss": 0.2073, + "step": 67764 + }, + { + "epoch": 2.61, + "learning_rate": 6.581815850218005e-06, + "loss": 0.4063, + "step": 67765 + }, + { + "epoch": 2.61, + "learning_rate": 6.581174660169274e-06, + "loss": 0.1312, + "step": 67766 + }, + { + "epoch": 2.61, + "learning_rate": 6.580533470120543e-06, + "loss": 0.1428, + "step": 67767 + }, + { + "epoch": 2.61, + "learning_rate": 6.5798922800718126e-06, + "loss": 0.249, + "step": 67768 + }, + { + "epoch": 2.61, + "learning_rate": 6.5792510900230834e-06, + "loss": 0.2674, + "step": 67769 + }, + { + "epoch": 2.61, + "learning_rate": 6.578609899974353e-06, + "loss": 0.1211, + "step": 67770 + }, + { + "epoch": 2.61, + "learning_rate": 6.577968709925622e-06, + "loss": 0.0677, + "step": 67771 + }, + { + "epoch": 2.61, + "learning_rate": 6.577327519876891e-06, + "loss": 0.1839, + "step": 67772 + }, + { + "epoch": 2.61, + "learning_rate": 6.576686329828162e-06, + "loss": 0.1826, + "step": 67773 + }, + { + "epoch": 2.61, + "learning_rate": 6.576045139779431e-06, + "loss": 0.1612, + "step": 67774 + }, + { + "epoch": 2.61, + "learning_rate": 6.5754039497307e-06, + "loss": 0.2326, + "step": 67775 + }, + { + "epoch": 2.61, + "learning_rate": 6.574762759681969e-06, + "loss": 0.1824, + "step": 67776 + }, + { + "epoch": 2.61, + "learning_rate": 6.5741215696332385e-06, + "loss": 0.4279, + "step": 67777 + }, + { + "epoch": 2.61, + "learning_rate": 6.573480379584509e-06, + "loss": 0.198, + "step": 67778 + }, + { + "epoch": 2.61, + "learning_rate": 6.5728391895357786e-06, + "loss": 0.1553, + "step": 67779 + }, + { + "epoch": 2.61, + "learning_rate": 6.572197999487048e-06, + "loss": 0.1716, + "step": 67780 + }, + { + "epoch": 2.61, + "learning_rate": 6.571556809438317e-06, + "loss": 0.2055, + "step": 67781 + }, + { + "epoch": 2.61, + "learning_rate": 6.570915619389588e-06, + "loss": 0.1932, + "step": 67782 + }, + { + "epoch": 2.61, + "learning_rate": 6.570274429340857e-06, + "loss": 0.1735, + "step": 67783 + }, + { + "epoch": 2.61, + "learning_rate": 6.569633239292126e-06, + "loss": 0.1682, + "step": 67784 + }, + { + "epoch": 2.61, + "learning_rate": 6.568992049243395e-06, + "loss": 0.2936, + "step": 67785 + }, + { + "epoch": 2.61, + "learning_rate": 6.568350859194666e-06, + "loss": 0.1176, + "step": 67786 + }, + { + "epoch": 2.61, + "learning_rate": 6.567709669145935e-06, + "loss": 0.2365, + "step": 67787 + }, + { + "epoch": 2.61, + "learning_rate": 6.5670684790972045e-06, + "loss": 0.3904, + "step": 67788 + }, + { + "epoch": 2.61, + "learning_rate": 6.566427289048474e-06, + "loss": 0.1691, + "step": 67789 + }, + { + "epoch": 2.61, + "learning_rate": 6.565786098999743e-06, + "loss": 0.2243, + "step": 67790 + }, + { + "epoch": 2.61, + "learning_rate": 6.565144908951014e-06, + "loss": 0.3221, + "step": 67791 + }, + { + "epoch": 2.61, + "learning_rate": 6.564503718902283e-06, + "loss": 0.1993, + "step": 67792 + }, + { + "epoch": 2.61, + "learning_rate": 6.563862528853552e-06, + "loss": 0.2209, + "step": 67793 + }, + { + "epoch": 2.61, + "learning_rate": 6.563221338804821e-06, + "loss": 0.2048, + "step": 67794 + }, + { + "epoch": 2.61, + "learning_rate": 6.562580148756092e-06, + "loss": 0.2175, + "step": 67795 + }, + { + "epoch": 2.61, + "learning_rate": 6.561938958707361e-06, + "loss": 0.3114, + "step": 67796 + }, + { + "epoch": 2.61, + "learning_rate": 6.5612977686586305e-06, + "loss": 0.2824, + "step": 67797 + }, + { + "epoch": 2.61, + "learning_rate": 6.5606565786099e-06, + "loss": 0.192, + "step": 67798 + }, + { + "epoch": 2.61, + "learning_rate": 6.560015388561169e-06, + "loss": 0.146, + "step": 67799 + }, + { + "epoch": 2.61, + "learning_rate": 6.55937419851244e-06, + "loss": 0.2408, + "step": 67800 + }, + { + "epoch": 2.61, + "learning_rate": 6.558733008463709e-06, + "loss": 0.1739, + "step": 67801 + }, + { + "epoch": 2.61, + "learning_rate": 6.558091818414978e-06, + "loss": 0.1906, + "step": 67802 + }, + { + "epoch": 2.61, + "learning_rate": 6.557450628366247e-06, + "loss": 0.1282, + "step": 67803 + }, + { + "epoch": 2.61, + "learning_rate": 6.556809438317518e-06, + "loss": 0.1841, + "step": 67804 + }, + { + "epoch": 2.61, + "learning_rate": 6.556168248268787e-06, + "loss": 0.2616, + "step": 67805 + }, + { + "epoch": 2.61, + "learning_rate": 6.5555270582200565e-06, + "loss": 0.1232, + "step": 67806 + }, + { + "epoch": 2.61, + "learning_rate": 6.554885868171326e-06, + "loss": 0.3141, + "step": 67807 + }, + { + "epoch": 2.61, + "learning_rate": 6.554244678122595e-06, + "loss": 0.1807, + "step": 67808 + }, + { + "epoch": 2.61, + "learning_rate": 6.553603488073866e-06, + "loss": 0.2329, + "step": 67809 + }, + { + "epoch": 2.61, + "learning_rate": 6.552962298025135e-06, + "loss": 0.225, + "step": 67810 + }, + { + "epoch": 2.61, + "learning_rate": 6.552321107976404e-06, + "loss": 0.1835, + "step": 67811 + }, + { + "epoch": 2.61, + "learning_rate": 6.551679917927673e-06, + "loss": 0.2531, + "step": 67812 + }, + { + "epoch": 2.61, + "learning_rate": 6.551038727878944e-06, + "loss": 0.1771, + "step": 67813 + }, + { + "epoch": 2.61, + "learning_rate": 6.550397537830213e-06, + "loss": 0.2273, + "step": 67814 + }, + { + "epoch": 2.61, + "learning_rate": 6.5497563477814824e-06, + "loss": 0.2346, + "step": 67815 + }, + { + "epoch": 2.61, + "learning_rate": 6.549115157732752e-06, + "loss": 0.3001, + "step": 67816 + }, + { + "epoch": 2.61, + "learning_rate": 6.548473967684022e-06, + "loss": 0.2072, + "step": 67817 + }, + { + "epoch": 2.61, + "learning_rate": 6.547832777635292e-06, + "loss": 0.1665, + "step": 67818 + }, + { + "epoch": 2.61, + "learning_rate": 6.547191587586561e-06, + "loss": 0.1899, + "step": 67819 + }, + { + "epoch": 2.61, + "learning_rate": 6.54655039753783e-06, + "loss": 0.3089, + "step": 67820 + }, + { + "epoch": 2.61, + "learning_rate": 6.545909207489099e-06, + "loss": 0.1211, + "step": 67821 + }, + { + "epoch": 2.61, + "learning_rate": 6.54526801744037e-06, + "loss": 0.2302, + "step": 67822 + }, + { + "epoch": 2.61, + "learning_rate": 6.544626827391639e-06, + "loss": 0.2615, + "step": 67823 + }, + { + "epoch": 2.61, + "learning_rate": 6.543985637342908e-06, + "loss": 0.2319, + "step": 67824 + }, + { + "epoch": 2.61, + "learning_rate": 6.543344447294178e-06, + "loss": 0.3087, + "step": 67825 + }, + { + "epoch": 2.61, + "learning_rate": 6.542703257245448e-06, + "loss": 0.176, + "step": 67826 + }, + { + "epoch": 2.61, + "learning_rate": 6.542062067196718e-06, + "loss": 0.1567, + "step": 67827 + }, + { + "epoch": 2.61, + "learning_rate": 6.541420877147987e-06, + "loss": 0.2906, + "step": 67828 + }, + { + "epoch": 2.61, + "learning_rate": 6.540779687099256e-06, + "loss": 0.3197, + "step": 67829 + }, + { + "epoch": 2.61, + "learning_rate": 6.540138497050525e-06, + "loss": 0.1722, + "step": 67830 + }, + { + "epoch": 2.61, + "learning_rate": 6.539497307001796e-06, + "loss": 0.1456, + "step": 67831 + }, + { + "epoch": 2.61, + "learning_rate": 6.538856116953065e-06, + "loss": 0.2604, + "step": 67832 + }, + { + "epoch": 2.61, + "learning_rate": 6.538214926904334e-06, + "loss": 0.2424, + "step": 67833 + }, + { + "epoch": 2.61, + "learning_rate": 6.5375737368556036e-06, + "loss": 0.2004, + "step": 67834 + }, + { + "epoch": 2.61, + "learning_rate": 6.536932546806874e-06, + "loss": 0.1063, + "step": 67835 + }, + { + "epoch": 2.61, + "learning_rate": 6.536291356758144e-06, + "loss": 0.2403, + "step": 67836 + }, + { + "epoch": 2.61, + "learning_rate": 6.535650166709413e-06, + "loss": 0.126, + "step": 67837 + }, + { + "epoch": 2.61, + "learning_rate": 6.535008976660682e-06, + "loss": 0.1827, + "step": 67838 + }, + { + "epoch": 2.61, + "learning_rate": 6.534367786611951e-06, + "loss": 0.2117, + "step": 67839 + }, + { + "epoch": 2.61, + "learning_rate": 6.533726596563222e-06, + "loss": 0.1832, + "step": 67840 + }, + { + "epoch": 2.61, + "learning_rate": 6.533085406514491e-06, + "loss": 0.1886, + "step": 67841 + }, + { + "epoch": 2.61, + "learning_rate": 6.53244421646576e-06, + "loss": 0.1249, + "step": 67842 + }, + { + "epoch": 2.61, + "learning_rate": 6.5318030264170295e-06, + "loss": 0.3273, + "step": 67843 + }, + { + "epoch": 2.61, + "learning_rate": 6.5311618363683e-06, + "loss": 0.1583, + "step": 67844 + }, + { + "epoch": 2.61, + "learning_rate": 6.5305206463195696e-06, + "loss": 0.1696, + "step": 67845 + }, + { + "epoch": 2.61, + "learning_rate": 6.529879456270839e-06, + "loss": 0.3602, + "step": 67846 + }, + { + "epoch": 2.61, + "learning_rate": 6.529238266222108e-06, + "loss": 0.2941, + "step": 67847 + }, + { + "epoch": 2.61, + "learning_rate": 6.528597076173377e-06, + "loss": 0.2145, + "step": 67848 + }, + { + "epoch": 2.61, + "learning_rate": 6.527955886124648e-06, + "loss": 0.1668, + "step": 67849 + }, + { + "epoch": 2.61, + "learning_rate": 6.527314696075917e-06, + "loss": 0.0543, + "step": 67850 + }, + { + "epoch": 2.61, + "learning_rate": 6.526673506027186e-06, + "loss": 0.2839, + "step": 67851 + }, + { + "epoch": 2.61, + "learning_rate": 6.5260323159784555e-06, + "loss": 0.3621, + "step": 67852 + }, + { + "epoch": 2.61, + "learning_rate": 6.525391125929726e-06, + "loss": 0.1212, + "step": 67853 + }, + { + "epoch": 2.61, + "learning_rate": 6.5247499358809955e-06, + "loss": 0.2351, + "step": 67854 + }, + { + "epoch": 2.61, + "learning_rate": 6.524108745832265e-06, + "loss": 0.2663, + "step": 67855 + }, + { + "epoch": 2.61, + "learning_rate": 6.523467555783534e-06, + "loss": 0.2647, + "step": 67856 + }, + { + "epoch": 2.61, + "learning_rate": 6.522826365734804e-06, + "loss": 0.2674, + "step": 67857 + }, + { + "epoch": 2.61, + "learning_rate": 6.522185175686074e-06, + "loss": 0.2741, + "step": 67858 + }, + { + "epoch": 2.61, + "learning_rate": 6.521543985637343e-06, + "loss": 0.2949, + "step": 67859 + }, + { + "epoch": 2.61, + "learning_rate": 6.520902795588612e-06, + "loss": 0.2942, + "step": 67860 + }, + { + "epoch": 2.61, + "learning_rate": 6.5202616055398815e-06, + "loss": 0.2073, + "step": 67861 + }, + { + "epoch": 2.61, + "learning_rate": 6.519620415491152e-06, + "loss": 0.2198, + "step": 67862 + }, + { + "epoch": 2.61, + "learning_rate": 6.5189792254424215e-06, + "loss": 0.1713, + "step": 67863 + }, + { + "epoch": 2.61, + "learning_rate": 6.518338035393691e-06, + "loss": 0.1468, + "step": 67864 + }, + { + "epoch": 2.61, + "learning_rate": 6.51769684534496e-06, + "loss": 0.291, + "step": 67865 + }, + { + "epoch": 2.61, + "learning_rate": 6.51705565529623e-06, + "loss": 0.1661, + "step": 67866 + }, + { + "epoch": 2.61, + "learning_rate": 6.5164144652475e-06, + "loss": 0.2365, + "step": 67867 + }, + { + "epoch": 2.61, + "learning_rate": 6.515773275198769e-06, + "loss": 0.1974, + "step": 67868 + }, + { + "epoch": 2.61, + "learning_rate": 6.515132085150038e-06, + "loss": 0.3383, + "step": 67869 + }, + { + "epoch": 2.61, + "learning_rate": 6.5144908951013074e-06, + "loss": 0.2532, + "step": 67870 + }, + { + "epoch": 2.61, + "learning_rate": 6.513849705052578e-06, + "loss": 0.1433, + "step": 67871 + }, + { + "epoch": 2.61, + "learning_rate": 6.5132085150038475e-06, + "loss": 0.2066, + "step": 67872 + }, + { + "epoch": 2.61, + "learning_rate": 6.512567324955117e-06, + "loss": 0.237, + "step": 67873 + }, + { + "epoch": 2.61, + "learning_rate": 6.511926134906386e-06, + "loss": 0.2797, + "step": 67874 + }, + { + "epoch": 2.61, + "learning_rate": 6.511284944857656e-06, + "loss": 0.2519, + "step": 67875 + }, + { + "epoch": 2.61, + "learning_rate": 6.510643754808926e-06, + "loss": 0.2647, + "step": 67876 + }, + { + "epoch": 2.61, + "learning_rate": 6.510002564760195e-06, + "loss": 0.3053, + "step": 67877 + }, + { + "epoch": 2.61, + "learning_rate": 6.509361374711464e-06, + "loss": 0.2843, + "step": 67878 + }, + { + "epoch": 2.61, + "learning_rate": 6.508720184662733e-06, + "loss": 0.2616, + "step": 67879 + }, + { + "epoch": 2.61, + "learning_rate": 6.508078994614004e-06, + "loss": 0.1795, + "step": 67880 + }, + { + "epoch": 2.61, + "learning_rate": 6.5074378045652734e-06, + "loss": 0.1578, + "step": 67881 + }, + { + "epoch": 2.61, + "learning_rate": 6.506796614516543e-06, + "loss": 0.1962, + "step": 67882 + }, + { + "epoch": 2.61, + "learning_rate": 6.506155424467812e-06, + "loss": 0.2313, + "step": 67883 + }, + { + "epoch": 2.61, + "learning_rate": 6.505514234419083e-06, + "loss": 0.1854, + "step": 67884 + }, + { + "epoch": 2.61, + "learning_rate": 6.504873044370352e-06, + "loss": 0.1711, + "step": 67885 + }, + { + "epoch": 2.61, + "learning_rate": 6.504231854321621e-06, + "loss": 0.154, + "step": 67886 + }, + { + "epoch": 2.61, + "learning_rate": 6.50359066427289e-06, + "loss": 0.1722, + "step": 67887 + }, + { + "epoch": 2.61, + "learning_rate": 6.50294947422416e-06, + "loss": 0.1668, + "step": 67888 + }, + { + "epoch": 2.61, + "learning_rate": 6.50230828417543e-06, + "loss": 0.1795, + "step": 67889 + }, + { + "epoch": 2.61, + "learning_rate": 6.501667094126699e-06, + "loss": 0.1679, + "step": 67890 + }, + { + "epoch": 2.61, + "learning_rate": 6.501025904077969e-06, + "loss": 0.2887, + "step": 67891 + }, + { + "epoch": 2.61, + "learning_rate": 6.500384714029238e-06, + "loss": 0.2226, + "step": 67892 + }, + { + "epoch": 2.61, + "learning_rate": 6.499743523980509e-06, + "loss": 0.1381, + "step": 67893 + }, + { + "epoch": 2.61, + "learning_rate": 6.499102333931778e-06, + "loss": 0.2949, + "step": 67894 + }, + { + "epoch": 2.61, + "learning_rate": 6.498461143883047e-06, + "loss": 0.1856, + "step": 67895 + }, + { + "epoch": 2.61, + "learning_rate": 6.497819953834316e-06, + "loss": 0.2041, + "step": 67896 + }, + { + "epoch": 2.61, + "learning_rate": 6.497178763785586e-06, + "loss": 0.2634, + "step": 67897 + }, + { + "epoch": 2.61, + "learning_rate": 6.496537573736856e-06, + "loss": 0.2644, + "step": 67898 + }, + { + "epoch": 2.61, + "learning_rate": 6.495896383688125e-06, + "loss": 0.2472, + "step": 67899 + }, + { + "epoch": 2.61, + "learning_rate": 6.4952551936393945e-06, + "loss": 0.2563, + "step": 67900 + }, + { + "epoch": 2.61, + "learning_rate": 6.494614003590664e-06, + "loss": 0.2552, + "step": 67901 + }, + { + "epoch": 2.61, + "learning_rate": 6.493972813541935e-06, + "loss": 0.2883, + "step": 67902 + }, + { + "epoch": 2.61, + "learning_rate": 6.493331623493204e-06, + "loss": 0.2484, + "step": 67903 + }, + { + "epoch": 2.61, + "learning_rate": 6.492690433444473e-06, + "loss": 0.1604, + "step": 67904 + }, + { + "epoch": 2.61, + "learning_rate": 6.492049243395742e-06, + "loss": 0.0735, + "step": 67905 + }, + { + "epoch": 2.61, + "learning_rate": 6.491408053347012e-06, + "loss": 0.2518, + "step": 67906 + }, + { + "epoch": 2.61, + "learning_rate": 6.490766863298282e-06, + "loss": 0.1737, + "step": 67907 + }, + { + "epoch": 2.61, + "learning_rate": 6.490125673249551e-06, + "loss": 0.2802, + "step": 67908 + }, + { + "epoch": 2.61, + "learning_rate": 6.4894844832008205e-06, + "loss": 0.343, + "step": 67909 + }, + { + "epoch": 2.61, + "learning_rate": 6.48884329315209e-06, + "loss": 0.2677, + "step": 67910 + }, + { + "epoch": 2.61, + "learning_rate": 6.4882021031033606e-06, + "loss": 0.1229, + "step": 67911 + }, + { + "epoch": 2.61, + "learning_rate": 6.48756091305463e-06, + "loss": 0.1984, + "step": 67912 + }, + { + "epoch": 2.61, + "learning_rate": 6.486919723005899e-06, + "loss": 0.2758, + "step": 67913 + }, + { + "epoch": 2.61, + "learning_rate": 6.486278532957168e-06, + "loss": 0.2477, + "step": 67914 + }, + { + "epoch": 2.61, + "learning_rate": 6.485637342908438e-06, + "loss": 0.2209, + "step": 67915 + }, + { + "epoch": 2.61, + "learning_rate": 6.484996152859708e-06, + "loss": 0.1047, + "step": 67916 + }, + { + "epoch": 2.61, + "learning_rate": 6.484354962810977e-06, + "loss": 0.1619, + "step": 67917 + }, + { + "epoch": 2.61, + "learning_rate": 6.4837137727622465e-06, + "loss": 0.1077, + "step": 67918 + }, + { + "epoch": 2.61, + "learning_rate": 6.4830725827135165e-06, + "loss": 0.2212, + "step": 67919 + }, + { + "epoch": 2.61, + "learning_rate": 6.4824313926647865e-06, + "loss": 0.1841, + "step": 67920 + }, + { + "epoch": 2.61, + "learning_rate": 6.481790202616056e-06, + "loss": 0.2265, + "step": 67921 + }, + { + "epoch": 2.61, + "learning_rate": 6.481149012567325e-06, + "loss": 0.2366, + "step": 67922 + }, + { + "epoch": 2.61, + "learning_rate": 6.480507822518594e-06, + "loss": 0.1091, + "step": 67923 + }, + { + "epoch": 2.61, + "learning_rate": 6.479866632469864e-06, + "loss": 0.2039, + "step": 67924 + }, + { + "epoch": 2.61, + "learning_rate": 6.479225442421134e-06, + "loss": 0.2901, + "step": 67925 + }, + { + "epoch": 2.61, + "learning_rate": 6.478584252372403e-06, + "loss": 0.1324, + "step": 67926 + }, + { + "epoch": 2.61, + "learning_rate": 6.4779430623236725e-06, + "loss": 0.1693, + "step": 67927 + }, + { + "epoch": 2.61, + "learning_rate": 6.4773018722749425e-06, + "loss": 0.1116, + "step": 67928 + }, + { + "epoch": 2.61, + "learning_rate": 6.4766606822262125e-06, + "loss": 0.2955, + "step": 67929 + }, + { + "epoch": 2.61, + "learning_rate": 6.476019492177482e-06, + "loss": 0.2191, + "step": 67930 + }, + { + "epoch": 2.61, + "learning_rate": 6.475378302128751e-06, + "loss": 0.1727, + "step": 67931 + }, + { + "epoch": 2.61, + "learning_rate": 6.47473711208002e-06, + "loss": 0.2169, + "step": 67932 + }, + { + "epoch": 2.61, + "learning_rate": 6.474095922031291e-06, + "loss": 0.1934, + "step": 67933 + }, + { + "epoch": 2.61, + "learning_rate": 6.47345473198256e-06, + "loss": 0.1681, + "step": 67934 + }, + { + "epoch": 2.61, + "learning_rate": 6.472813541933829e-06, + "loss": 0.2186, + "step": 67935 + }, + { + "epoch": 2.61, + "learning_rate": 6.472172351885098e-06, + "loss": 0.2308, + "step": 67936 + }, + { + "epoch": 2.61, + "learning_rate": 6.4715311618363684e-06, + "loss": 0.1761, + "step": 67937 + }, + { + "epoch": 2.61, + "learning_rate": 6.4708899717876385e-06, + "loss": 0.2941, + "step": 67938 + }, + { + "epoch": 2.61, + "learning_rate": 6.470248781738908e-06, + "loss": 0.1369, + "step": 67939 + }, + { + "epoch": 2.61, + "learning_rate": 6.469607591690177e-06, + "loss": 0.2488, + "step": 67940 + }, + { + "epoch": 2.61, + "learning_rate": 6.468966401641446e-06, + "loss": 0.1705, + "step": 67941 + }, + { + "epoch": 2.61, + "learning_rate": 6.468325211592717e-06, + "loss": 0.2602, + "step": 67942 + }, + { + "epoch": 2.61, + "learning_rate": 6.467684021543986e-06, + "loss": 0.3103, + "step": 67943 + }, + { + "epoch": 2.61, + "learning_rate": 6.467042831495255e-06, + "loss": 0.1009, + "step": 67944 + }, + { + "epoch": 2.61, + "learning_rate": 6.466401641446524e-06, + "loss": 0.2048, + "step": 67945 + }, + { + "epoch": 2.61, + "learning_rate": 6.465760451397794e-06, + "loss": 0.1591, + "step": 67946 + }, + { + "epoch": 2.61, + "learning_rate": 6.4651192613490644e-06, + "loss": 0.3842, + "step": 67947 + }, + { + "epoch": 2.61, + "learning_rate": 6.464478071300334e-06, + "loss": 0.2126, + "step": 67948 + }, + { + "epoch": 2.61, + "learning_rate": 6.463836881251603e-06, + "loss": 0.1454, + "step": 67949 + }, + { + "epoch": 2.61, + "learning_rate": 6.463195691202873e-06, + "loss": 0.2205, + "step": 67950 + }, + { + "epoch": 2.61, + "learning_rate": 6.462554501154143e-06, + "loss": 0.2709, + "step": 67951 + }, + { + "epoch": 2.61, + "learning_rate": 6.461913311105412e-06, + "loss": 0.3774, + "step": 67952 + }, + { + "epoch": 2.61, + "learning_rate": 6.461272121056681e-06, + "loss": 0.205, + "step": 67953 + }, + { + "epoch": 2.61, + "learning_rate": 6.46063093100795e-06, + "loss": 0.1644, + "step": 67954 + }, + { + "epoch": 2.61, + "learning_rate": 6.45998974095922e-06, + "loss": 0.1334, + "step": 67955 + }, + { + "epoch": 2.61, + "learning_rate": 6.45934855091049e-06, + "loss": 0.1359, + "step": 67956 + }, + { + "epoch": 2.61, + "learning_rate": 6.4587073608617596e-06, + "loss": 0.1023, + "step": 67957 + }, + { + "epoch": 2.61, + "learning_rate": 6.458066170813029e-06, + "loss": 0.3302, + "step": 67958 + }, + { + "epoch": 2.61, + "learning_rate": 6.457424980764299e-06, + "loss": 0.1645, + "step": 67959 + }, + { + "epoch": 2.61, + "learning_rate": 6.456783790715569e-06, + "loss": 0.2199, + "step": 67960 + }, + { + "epoch": 2.61, + "learning_rate": 6.456142600666838e-06, + "loss": 0.1011, + "step": 67961 + }, + { + "epoch": 2.61, + "learning_rate": 6.455501410618107e-06, + "loss": 0.1029, + "step": 67962 + }, + { + "epoch": 2.61, + "learning_rate": 6.454860220569376e-06, + "loss": 0.1991, + "step": 67963 + }, + { + "epoch": 2.61, + "learning_rate": 6.454219030520646e-06, + "loss": 0.1801, + "step": 67964 + }, + { + "epoch": 2.61, + "learning_rate": 6.453577840471916e-06, + "loss": 0.2721, + "step": 67965 + }, + { + "epoch": 2.61, + "learning_rate": 6.4529366504231855e-06, + "loss": 0.2466, + "step": 67966 + }, + { + "epoch": 2.61, + "learning_rate": 6.452295460374455e-06, + "loss": 0.2518, + "step": 67967 + }, + { + "epoch": 2.61, + "learning_rate": 6.451654270325725e-06, + "loss": 0.1646, + "step": 67968 + }, + { + "epoch": 2.61, + "learning_rate": 6.451013080276995e-06, + "loss": 0.1958, + "step": 67969 + }, + { + "epoch": 2.61, + "learning_rate": 6.450371890228264e-06, + "loss": 0.2202, + "step": 67970 + }, + { + "epoch": 2.61, + "learning_rate": 6.449730700179533e-06, + "loss": 0.1838, + "step": 67971 + }, + { + "epoch": 2.61, + "learning_rate": 6.449089510130802e-06, + "loss": 0.1358, + "step": 67972 + }, + { + "epoch": 2.61, + "learning_rate": 6.448448320082072e-06, + "loss": 0.1987, + "step": 67973 + }, + { + "epoch": 2.61, + "learning_rate": 6.447807130033342e-06, + "loss": 0.1608, + "step": 67974 + }, + { + "epoch": 2.61, + "learning_rate": 6.4471659399846115e-06, + "loss": 0.1855, + "step": 67975 + }, + { + "epoch": 2.61, + "learning_rate": 6.446524749935881e-06, + "loss": 0.3144, + "step": 67976 + }, + { + "epoch": 2.61, + "learning_rate": 6.445883559887151e-06, + "loss": 0.3815, + "step": 67977 + }, + { + "epoch": 2.61, + "learning_rate": 6.445242369838421e-06, + "loss": 0.3713, + "step": 67978 + }, + { + "epoch": 2.61, + "learning_rate": 6.44460117978969e-06, + "loss": 0.2572, + "step": 67979 + }, + { + "epoch": 2.61, + "learning_rate": 6.443959989740959e-06, + "loss": 0.2025, + "step": 67980 + }, + { + "epoch": 2.61, + "learning_rate": 6.443318799692228e-06, + "loss": 0.1314, + "step": 67981 + }, + { + "epoch": 2.61, + "learning_rate": 6.442677609643499e-06, + "loss": 0.227, + "step": 67982 + }, + { + "epoch": 2.61, + "learning_rate": 6.442036419594768e-06, + "loss": 0.1185, + "step": 67983 + }, + { + "epoch": 2.61, + "learning_rate": 6.4413952295460375e-06, + "loss": 0.236, + "step": 67984 + }, + { + "epoch": 2.61, + "learning_rate": 6.440754039497307e-06, + "loss": 0.1337, + "step": 67985 + }, + { + "epoch": 2.61, + "learning_rate": 6.440112849448577e-06, + "loss": 0.1967, + "step": 67986 + }, + { + "epoch": 2.61, + "learning_rate": 6.439471659399847e-06, + "loss": 0.1927, + "step": 67987 + }, + { + "epoch": 2.61, + "learning_rate": 6.438830469351116e-06, + "loss": 0.2052, + "step": 67988 + }, + { + "epoch": 2.61, + "learning_rate": 6.438189279302385e-06, + "loss": 0.0867, + "step": 67989 + }, + { + "epoch": 2.61, + "learning_rate": 6.437548089253655e-06, + "loss": 0.1119, + "step": 67990 + }, + { + "epoch": 2.61, + "learning_rate": 6.436906899204925e-06, + "loss": 0.2321, + "step": 67991 + }, + { + "epoch": 2.61, + "learning_rate": 6.436265709156194e-06, + "loss": 0.2169, + "step": 67992 + }, + { + "epoch": 2.61, + "learning_rate": 6.4356245191074634e-06, + "loss": 0.1685, + "step": 67993 + }, + { + "epoch": 2.61, + "learning_rate": 6.434983329058733e-06, + "loss": 0.2731, + "step": 67994 + }, + { + "epoch": 2.61, + "learning_rate": 6.434342139010003e-06, + "loss": 0.2519, + "step": 67995 + }, + { + "epoch": 2.61, + "learning_rate": 6.433700948961273e-06, + "loss": 0.4607, + "step": 67996 + }, + { + "epoch": 2.61, + "learning_rate": 6.433059758912542e-06, + "loss": 0.316, + "step": 67997 + }, + { + "epoch": 2.61, + "learning_rate": 6.432418568863811e-06, + "loss": 0.2202, + "step": 67998 + }, + { + "epoch": 2.61, + "learning_rate": 6.431777378815081e-06, + "loss": 0.1161, + "step": 67999 + }, + { + "epoch": 2.61, + "learning_rate": 6.431136188766351e-06, + "loss": 0.1667, + "step": 68000 + }, + { + "epoch": 2.61, + "learning_rate": 6.43049499871762e-06, + "loss": 0.1942, + "step": 68001 + }, + { + "epoch": 2.61, + "learning_rate": 6.429853808668889e-06, + "loss": 0.1445, + "step": 68002 + }, + { + "epoch": 2.61, + "learning_rate": 6.429212618620159e-06, + "loss": 0.2819, + "step": 68003 + }, + { + "epoch": 2.61, + "learning_rate": 6.428571428571429e-06, + "loss": 0.313, + "step": 68004 + }, + { + "epoch": 2.61, + "learning_rate": 6.427930238522699e-06, + "loss": 0.2287, + "step": 68005 + }, + { + "epoch": 2.61, + "learning_rate": 6.427289048473968e-06, + "loss": 0.2389, + "step": 68006 + }, + { + "epoch": 2.61, + "learning_rate": 6.426647858425237e-06, + "loss": 0.1846, + "step": 68007 + }, + { + "epoch": 2.61, + "learning_rate": 6.426006668376507e-06, + "loss": 0.1374, + "step": 68008 + }, + { + "epoch": 2.61, + "learning_rate": 6.425365478327777e-06, + "loss": 0.1787, + "step": 68009 + }, + { + "epoch": 2.61, + "learning_rate": 6.424724288279046e-06, + "loss": 0.347, + "step": 68010 + }, + { + "epoch": 2.61, + "learning_rate": 6.424083098230315e-06, + "loss": 0.2149, + "step": 68011 + }, + { + "epoch": 2.61, + "learning_rate": 6.4234419081815846e-06, + "loss": 0.2307, + "step": 68012 + }, + { + "epoch": 2.61, + "learning_rate": 6.422800718132855e-06, + "loss": 0.238, + "step": 68013 + }, + { + "epoch": 2.61, + "learning_rate": 6.422159528084125e-06, + "loss": 0.1799, + "step": 68014 + }, + { + "epoch": 2.61, + "learning_rate": 6.421518338035394e-06, + "loss": 0.1207, + "step": 68015 + }, + { + "epoch": 2.61, + "learning_rate": 6.420877147986663e-06, + "loss": 0.2358, + "step": 68016 + }, + { + "epoch": 2.62, + "learning_rate": 6.420235957937933e-06, + "loss": 0.1671, + "step": 68017 + }, + { + "epoch": 2.62, + "learning_rate": 6.419594767889203e-06, + "loss": 0.148, + "step": 68018 + }, + { + "epoch": 2.62, + "learning_rate": 6.418953577840472e-06, + "loss": 0.1766, + "step": 68019 + }, + { + "epoch": 2.62, + "learning_rate": 6.418312387791741e-06, + "loss": 0.1816, + "step": 68020 + }, + { + "epoch": 2.62, + "learning_rate": 6.417671197743011e-06, + "loss": 0.0588, + "step": 68021 + }, + { + "epoch": 2.62, + "learning_rate": 6.4170300076942805e-06, + "loss": 0.2353, + "step": 68022 + }, + { + "epoch": 2.62, + "learning_rate": 6.4163888176455506e-06, + "loss": 0.1406, + "step": 68023 + }, + { + "epoch": 2.62, + "learning_rate": 6.41574762759682e-06, + "loss": 0.2099, + "step": 68024 + }, + { + "epoch": 2.62, + "learning_rate": 6.415106437548089e-06, + "loss": 0.2533, + "step": 68025 + }, + { + "epoch": 2.62, + "learning_rate": 6.414465247499359e-06, + "loss": 0.1852, + "step": 68026 + }, + { + "epoch": 2.62, + "learning_rate": 6.413824057450629e-06, + "loss": 0.2904, + "step": 68027 + }, + { + "epoch": 2.62, + "learning_rate": 6.413182867401898e-06, + "loss": 0.3149, + "step": 68028 + }, + { + "epoch": 2.62, + "learning_rate": 6.412541677353167e-06, + "loss": 0.3699, + "step": 68029 + }, + { + "epoch": 2.62, + "learning_rate": 6.411900487304437e-06, + "loss": 0.1033, + "step": 68030 + }, + { + "epoch": 2.62, + "learning_rate": 6.411259297255707e-06, + "loss": 0.2695, + "step": 68031 + }, + { + "epoch": 2.62, + "learning_rate": 6.4106181072069765e-06, + "loss": 0.112, + "step": 68032 + }, + { + "epoch": 2.62, + "learning_rate": 6.409976917158246e-06, + "loss": 0.26, + "step": 68033 + }, + { + "epoch": 2.62, + "learning_rate": 6.409335727109515e-06, + "loss": 0.0637, + "step": 68034 + }, + { + "epoch": 2.62, + "learning_rate": 6.408694537060785e-06, + "loss": 0.2228, + "step": 68035 + }, + { + "epoch": 2.62, + "learning_rate": 6.408053347012055e-06, + "loss": 0.1386, + "step": 68036 + }, + { + "epoch": 2.62, + "learning_rate": 6.407412156963324e-06, + "loss": 0.1567, + "step": 68037 + }, + { + "epoch": 2.62, + "learning_rate": 6.406770966914593e-06, + "loss": 0.4492, + "step": 68038 + }, + { + "epoch": 2.62, + "learning_rate": 6.406129776865863e-06, + "loss": 0.1359, + "step": 68039 + }, + { + "epoch": 2.62, + "learning_rate": 6.405488586817133e-06, + "loss": 0.185, + "step": 68040 + }, + { + "epoch": 2.62, + "learning_rate": 6.4048473967684025e-06, + "loss": 0.2905, + "step": 68041 + }, + { + "epoch": 2.62, + "learning_rate": 6.404206206719672e-06, + "loss": 0.1593, + "step": 68042 + }, + { + "epoch": 2.62, + "learning_rate": 6.403565016670941e-06, + "loss": 0.2461, + "step": 68043 + }, + { + "epoch": 2.62, + "learning_rate": 6.402923826622211e-06, + "loss": 0.2013, + "step": 68044 + }, + { + "epoch": 2.62, + "learning_rate": 6.402282636573481e-06, + "loss": 0.3356, + "step": 68045 + }, + { + "epoch": 2.62, + "learning_rate": 6.40164144652475e-06, + "loss": 0.2295, + "step": 68046 + }, + { + "epoch": 2.62, + "learning_rate": 6.401000256476019e-06, + "loss": 0.2219, + "step": 68047 + }, + { + "epoch": 2.62, + "learning_rate": 6.400359066427289e-06, + "loss": 0.2629, + "step": 68048 + }, + { + "epoch": 2.62, + "learning_rate": 6.399717876378559e-06, + "loss": 0.23, + "step": 68049 + }, + { + "epoch": 2.62, + "learning_rate": 6.3990766863298285e-06, + "loss": 0.1682, + "step": 68050 + }, + { + "epoch": 2.62, + "learning_rate": 6.398435496281098e-06, + "loss": 0.2132, + "step": 68051 + }, + { + "epoch": 2.62, + "learning_rate": 6.397794306232368e-06, + "loss": 0.3543, + "step": 68052 + }, + { + "epoch": 2.62, + "learning_rate": 6.397153116183637e-06, + "loss": 0.2226, + "step": 68053 + }, + { + "epoch": 2.62, + "learning_rate": 6.396511926134907e-06, + "loss": 0.1922, + "step": 68054 + }, + { + "epoch": 2.62, + "learning_rate": 6.395870736086176e-06, + "loss": 0.2207, + "step": 68055 + }, + { + "epoch": 2.62, + "learning_rate": 6.395229546037445e-06, + "loss": 0.2785, + "step": 68056 + }, + { + "epoch": 2.62, + "learning_rate": 6.394588355988715e-06, + "loss": 0.2119, + "step": 68057 + }, + { + "epoch": 2.62, + "learning_rate": 6.393947165939985e-06, + "loss": 0.3026, + "step": 68058 + }, + { + "epoch": 2.62, + "learning_rate": 6.3933059758912544e-06, + "loss": 0.3382, + "step": 68059 + }, + { + "epoch": 2.62, + "learning_rate": 6.392664785842524e-06, + "loss": 0.1712, + "step": 68060 + }, + { + "epoch": 2.62, + "learning_rate": 6.392023595793794e-06, + "loss": 0.1443, + "step": 68061 + }, + { + "epoch": 2.62, + "learning_rate": 6.391382405745063e-06, + "loss": 0.1973, + "step": 68062 + }, + { + "epoch": 2.62, + "learning_rate": 6.390741215696333e-06, + "loss": 0.2056, + "step": 68063 + }, + { + "epoch": 2.62, + "learning_rate": 6.390100025647602e-06, + "loss": 0.2824, + "step": 68064 + }, + { + "epoch": 2.62, + "learning_rate": 6.389458835598871e-06, + "loss": 0.21, + "step": 68065 + }, + { + "epoch": 2.62, + "learning_rate": 6.388817645550141e-06, + "loss": 0.2091, + "step": 68066 + }, + { + "epoch": 2.62, + "learning_rate": 6.388176455501411e-06, + "loss": 0.2097, + "step": 68067 + }, + { + "epoch": 2.62, + "learning_rate": 6.38753526545268e-06, + "loss": 0.294, + "step": 68068 + }, + { + "epoch": 2.62, + "learning_rate": 6.38689407540395e-06, + "loss": 0.2899, + "step": 68069 + }, + { + "epoch": 2.62, + "learning_rate": 6.38625288535522e-06, + "loss": 0.1479, + "step": 68070 + }, + { + "epoch": 2.62, + "learning_rate": 6.38561169530649e-06, + "loss": 0.2934, + "step": 68071 + }, + { + "epoch": 2.62, + "learning_rate": 6.384970505257759e-06, + "loss": 0.2915, + "step": 68072 + }, + { + "epoch": 2.62, + "learning_rate": 6.384329315209028e-06, + "loss": 0.2453, + "step": 68073 + }, + { + "epoch": 2.62, + "learning_rate": 6.383688125160297e-06, + "loss": 0.2645, + "step": 68074 + }, + { + "epoch": 2.62, + "learning_rate": 6.383046935111567e-06, + "loss": 0.1187, + "step": 68075 + }, + { + "epoch": 2.62, + "learning_rate": 6.382405745062837e-06, + "loss": 0.3479, + "step": 68076 + }, + { + "epoch": 2.62, + "learning_rate": 6.381764555014106e-06, + "loss": 0.299, + "step": 68077 + }, + { + "epoch": 2.62, + "learning_rate": 6.3811233649653756e-06, + "loss": 0.1542, + "step": 68078 + }, + { + "epoch": 2.62, + "learning_rate": 6.380482174916646e-06, + "loss": 0.2448, + "step": 68079 + }, + { + "epoch": 2.62, + "learning_rate": 6.379840984867916e-06, + "loss": 0.212, + "step": 68080 + }, + { + "epoch": 2.62, + "learning_rate": 6.379199794819185e-06, + "loss": 0.2277, + "step": 68081 + }, + { + "epoch": 2.62, + "learning_rate": 6.378558604770454e-06, + "loss": 0.2292, + "step": 68082 + }, + { + "epoch": 2.62, + "learning_rate": 6.377917414721724e-06, + "loss": 0.286, + "step": 68083 + }, + { + "epoch": 2.62, + "learning_rate": 6.377276224672993e-06, + "loss": 0.3295, + "step": 68084 + }, + { + "epoch": 2.62, + "learning_rate": 6.376635034624263e-06, + "loss": 0.1384, + "step": 68085 + }, + { + "epoch": 2.62, + "learning_rate": 6.375993844575532e-06, + "loss": 0.2783, + "step": 68086 + }, + { + "epoch": 2.62, + "learning_rate": 6.3753526545268015e-06, + "loss": 0.2949, + "step": 68087 + }, + { + "epoch": 2.62, + "learning_rate": 6.3747114644780715e-06, + "loss": 0.1502, + "step": 68088 + }, + { + "epoch": 2.62, + "learning_rate": 6.3740702744293416e-06, + "loss": 0.2024, + "step": 68089 + }, + { + "epoch": 2.62, + "learning_rate": 6.373429084380611e-06, + "loss": 0.0787, + "step": 68090 + }, + { + "epoch": 2.62, + "learning_rate": 6.37278789433188e-06, + "loss": 0.1919, + "step": 68091 + }, + { + "epoch": 2.62, + "learning_rate": 6.37214670428315e-06, + "loss": 0.2782, + "step": 68092 + }, + { + "epoch": 2.62, + "learning_rate": 6.371505514234419e-06, + "loss": 0.2339, + "step": 68093 + }, + { + "epoch": 2.62, + "learning_rate": 6.370864324185689e-06, + "loss": 0.3018, + "step": 68094 + }, + { + "epoch": 2.62, + "learning_rate": 6.370223134136958e-06, + "loss": 0.3156, + "step": 68095 + }, + { + "epoch": 2.62, + "learning_rate": 6.3695819440882275e-06, + "loss": 0.1234, + "step": 68096 + }, + { + "epoch": 2.62, + "learning_rate": 6.3689407540394975e-06, + "loss": 0.1474, + "step": 68097 + }, + { + "epoch": 2.62, + "learning_rate": 6.3682995639907675e-06, + "loss": 0.1744, + "step": 68098 + }, + { + "epoch": 2.62, + "learning_rate": 6.367658373942037e-06, + "loss": 0.1983, + "step": 68099 + }, + { + "epoch": 2.62, + "learning_rate": 6.367017183893306e-06, + "loss": 0.2237, + "step": 68100 + }, + { + "epoch": 2.62, + "learning_rate": 6.366375993844576e-06, + "loss": 0.4579, + "step": 68101 + }, + { + "epoch": 2.62, + "learning_rate": 6.365734803795845e-06, + "loss": 0.2383, + "step": 68102 + }, + { + "epoch": 2.62, + "learning_rate": 6.365093613747115e-06, + "loss": 0.147, + "step": 68103 + }, + { + "epoch": 2.62, + "learning_rate": 6.364452423698384e-06, + "loss": 0.2538, + "step": 68104 + }, + { + "epoch": 2.62, + "learning_rate": 6.3638112336496535e-06, + "loss": 0.4413, + "step": 68105 + }, + { + "epoch": 2.62, + "learning_rate": 6.3631700436009235e-06, + "loss": 0.2316, + "step": 68106 + }, + { + "epoch": 2.62, + "learning_rate": 6.3625288535521935e-06, + "loss": 0.2555, + "step": 68107 + }, + { + "epoch": 2.62, + "learning_rate": 6.361887663503463e-06, + "loss": 0.2657, + "step": 68108 + }, + { + "epoch": 2.62, + "learning_rate": 6.361246473454732e-06, + "loss": 0.1887, + "step": 68109 + }, + { + "epoch": 2.62, + "learning_rate": 6.360605283406002e-06, + "loss": 0.2317, + "step": 68110 + }, + { + "epoch": 2.62, + "learning_rate": 6.359964093357271e-06, + "loss": 0.1582, + "step": 68111 + }, + { + "epoch": 2.62, + "learning_rate": 6.359322903308541e-06, + "loss": 0.249, + "step": 68112 + }, + { + "epoch": 2.62, + "learning_rate": 6.35868171325981e-06, + "loss": 0.3388, + "step": 68113 + }, + { + "epoch": 2.62, + "learning_rate": 6.3580405232110794e-06, + "loss": 0.175, + "step": 68114 + }, + { + "epoch": 2.62, + "learning_rate": 6.3573993331623494e-06, + "loss": 0.1996, + "step": 68115 + }, + { + "epoch": 2.62, + "learning_rate": 6.3567581431136195e-06, + "loss": 0.285, + "step": 68116 + }, + { + "epoch": 2.62, + "learning_rate": 6.356116953064889e-06, + "loss": 0.1988, + "step": 68117 + }, + { + "epoch": 2.62, + "learning_rate": 6.355475763016158e-06, + "loss": 0.192, + "step": 68118 + }, + { + "epoch": 2.62, + "learning_rate": 6.354834572967428e-06, + "loss": 0.2318, + "step": 68119 + }, + { + "epoch": 2.62, + "learning_rate": 6.354193382918698e-06, + "loss": 0.2898, + "step": 68120 + }, + { + "epoch": 2.62, + "learning_rate": 6.353552192869967e-06, + "loss": 0.2337, + "step": 68121 + }, + { + "epoch": 2.62, + "learning_rate": 6.352911002821236e-06, + "loss": 0.2871, + "step": 68122 + }, + { + "epoch": 2.62, + "learning_rate": 6.352269812772506e-06, + "loss": 0.1834, + "step": 68123 + }, + { + "epoch": 2.62, + "learning_rate": 6.351628622723775e-06, + "loss": 0.1402, + "step": 68124 + }, + { + "epoch": 2.62, + "learning_rate": 6.3509874326750454e-06, + "loss": 0.1956, + "step": 68125 + }, + { + "epoch": 2.62, + "learning_rate": 6.350346242626315e-06, + "loss": 0.175, + "step": 68126 + }, + { + "epoch": 2.62, + "learning_rate": 6.349705052577584e-06, + "loss": 0.1374, + "step": 68127 + }, + { + "epoch": 2.62, + "learning_rate": 6.349063862528854e-06, + "loss": 0.0988, + "step": 68128 + }, + { + "epoch": 2.62, + "learning_rate": 6.348422672480124e-06, + "loss": 0.2481, + "step": 68129 + }, + { + "epoch": 2.62, + "learning_rate": 6.347781482431393e-06, + "loss": 0.1962, + "step": 68130 + }, + { + "epoch": 2.62, + "learning_rate": 6.347140292382662e-06, + "loss": 0.1735, + "step": 68131 + }, + { + "epoch": 2.62, + "learning_rate": 6.346499102333932e-06, + "loss": 0.2471, + "step": 68132 + }, + { + "epoch": 2.62, + "learning_rate": 6.345857912285201e-06, + "loss": 0.1788, + "step": 68133 + }, + { + "epoch": 2.62, + "learning_rate": 6.345216722236471e-06, + "loss": 0.1209, + "step": 68134 + }, + { + "epoch": 2.62, + "learning_rate": 6.344575532187741e-06, + "loss": 0.209, + "step": 68135 + }, + { + "epoch": 2.62, + "learning_rate": 6.34393434213901e-06, + "loss": 0.218, + "step": 68136 + }, + { + "epoch": 2.62, + "learning_rate": 6.34329315209028e-06, + "loss": 0.2787, + "step": 68137 + }, + { + "epoch": 2.62, + "learning_rate": 6.34265196204155e-06, + "loss": 0.2419, + "step": 68138 + }, + { + "epoch": 2.62, + "learning_rate": 6.342010771992819e-06, + "loss": 0.2117, + "step": 68139 + }, + { + "epoch": 2.62, + "learning_rate": 6.341369581944088e-06, + "loss": 0.2503, + "step": 68140 + }, + { + "epoch": 2.62, + "learning_rate": 6.340728391895358e-06, + "loss": 0.1116, + "step": 68141 + }, + { + "epoch": 2.62, + "learning_rate": 6.340087201846627e-06, + "loss": 0.1532, + "step": 68142 + }, + { + "epoch": 2.62, + "learning_rate": 6.339446011797897e-06, + "loss": 0.1728, + "step": 68143 + }, + { + "epoch": 2.62, + "learning_rate": 6.3388048217491665e-06, + "loss": 0.197, + "step": 68144 + }, + { + "epoch": 2.62, + "learning_rate": 6.338163631700436e-06, + "loss": 0.1325, + "step": 68145 + }, + { + "epoch": 2.62, + "learning_rate": 6.337522441651706e-06, + "loss": 0.0937, + "step": 68146 + }, + { + "epoch": 2.62, + "learning_rate": 6.336881251602976e-06, + "loss": 0.156, + "step": 68147 + }, + { + "epoch": 2.62, + "learning_rate": 6.336240061554245e-06, + "loss": 0.3321, + "step": 68148 + }, + { + "epoch": 2.62, + "learning_rate": 6.335598871505514e-06, + "loss": 0.2521, + "step": 68149 + }, + { + "epoch": 2.62, + "learning_rate": 6.334957681456784e-06, + "loss": 0.1863, + "step": 68150 + }, + { + "epoch": 2.62, + "learning_rate": 6.334316491408053e-06, + "loss": 0.181, + "step": 68151 + }, + { + "epoch": 2.62, + "learning_rate": 6.333675301359323e-06, + "loss": 0.282, + "step": 68152 + }, + { + "epoch": 2.62, + "learning_rate": 6.3330341113105925e-06, + "loss": 0.3174, + "step": 68153 + }, + { + "epoch": 2.62, + "learning_rate": 6.3323929212618625e-06, + "loss": 0.2077, + "step": 68154 + }, + { + "epoch": 2.62, + "learning_rate": 6.331751731213132e-06, + "loss": 0.1532, + "step": 68155 + }, + { + "epoch": 2.62, + "learning_rate": 6.331110541164402e-06, + "loss": 0.3715, + "step": 68156 + }, + { + "epoch": 2.62, + "learning_rate": 6.330469351115671e-06, + "loss": 0.222, + "step": 68157 + }, + { + "epoch": 2.62, + "learning_rate": 6.32982816106694e-06, + "loss": 0.1624, + "step": 68158 + }, + { + "epoch": 2.62, + "learning_rate": 6.32918697101821e-06, + "loss": 0.1883, + "step": 68159 + }, + { + "epoch": 2.62, + "learning_rate": 6.328545780969479e-06, + "loss": 0.3021, + "step": 68160 + }, + { + "epoch": 2.62, + "learning_rate": 6.327904590920749e-06, + "loss": 0.2311, + "step": 68161 + }, + { + "epoch": 2.62, + "learning_rate": 6.3272634008720185e-06, + "loss": 0.1878, + "step": 68162 + }, + { + "epoch": 2.62, + "learning_rate": 6.3266222108232885e-06, + "loss": 0.2725, + "step": 68163 + }, + { + "epoch": 2.62, + "learning_rate": 6.325981020774558e-06, + "loss": 0.1029, + "step": 68164 + }, + { + "epoch": 2.62, + "learning_rate": 6.325339830725828e-06, + "loss": 0.3763, + "step": 68165 + }, + { + "epoch": 2.62, + "learning_rate": 6.324698640677097e-06, + "loss": 0.2925, + "step": 68166 + }, + { + "epoch": 2.62, + "learning_rate": 6.324057450628366e-06, + "loss": 0.2272, + "step": 68167 + }, + { + "epoch": 2.62, + "learning_rate": 6.323416260579636e-06, + "loss": 0.3673, + "step": 68168 + }, + { + "epoch": 2.62, + "learning_rate": 6.322775070530906e-06, + "loss": 0.1911, + "step": 68169 + }, + { + "epoch": 2.62, + "learning_rate": 6.322133880482175e-06, + "loss": 0.271, + "step": 68170 + }, + { + "epoch": 2.62, + "learning_rate": 6.3214926904334445e-06, + "loss": 0.2249, + "step": 68171 + }, + { + "epoch": 2.62, + "learning_rate": 6.3208515003847145e-06, + "loss": 0.2025, + "step": 68172 + }, + { + "epoch": 2.62, + "learning_rate": 6.320210310335984e-06, + "loss": 0.1415, + "step": 68173 + }, + { + "epoch": 2.62, + "learning_rate": 6.319569120287254e-06, + "loss": 0.236, + "step": 68174 + }, + { + "epoch": 2.62, + "learning_rate": 6.318927930238523e-06, + "loss": 0.2132, + "step": 68175 + }, + { + "epoch": 2.62, + "learning_rate": 6.318286740189792e-06, + "loss": 0.1295, + "step": 68176 + }, + { + "epoch": 2.62, + "learning_rate": 6.317645550141062e-06, + "loss": 0.1623, + "step": 68177 + }, + { + "epoch": 2.62, + "learning_rate": 6.317004360092332e-06, + "loss": 0.1218, + "step": 68178 + }, + { + "epoch": 2.62, + "learning_rate": 6.316363170043601e-06, + "loss": 0.1864, + "step": 68179 + }, + { + "epoch": 2.62, + "learning_rate": 6.31572197999487e-06, + "loss": 0.106, + "step": 68180 + }, + { + "epoch": 2.62, + "learning_rate": 6.3150807899461404e-06, + "loss": 0.1262, + "step": 68181 + }, + { + "epoch": 2.62, + "learning_rate": 6.31443959989741e-06, + "loss": 0.1077, + "step": 68182 + }, + { + "epoch": 2.62, + "learning_rate": 6.31379840984868e-06, + "loss": 0.151, + "step": 68183 + }, + { + "epoch": 2.62, + "learning_rate": 6.313157219799949e-06, + "loss": 0.1912, + "step": 68184 + }, + { + "epoch": 2.62, + "learning_rate": 6.312516029751219e-06, + "loss": 0.2627, + "step": 68185 + }, + { + "epoch": 2.62, + "learning_rate": 6.311874839702488e-06, + "loss": 0.3215, + "step": 68186 + }, + { + "epoch": 2.62, + "learning_rate": 6.311233649653758e-06, + "loss": 0.0907, + "step": 68187 + }, + { + "epoch": 2.62, + "learning_rate": 6.310592459605027e-06, + "loss": 0.2238, + "step": 68188 + }, + { + "epoch": 2.62, + "learning_rate": 6.309951269556296e-06, + "loss": 0.1751, + "step": 68189 + }, + { + "epoch": 2.62, + "learning_rate": 6.309310079507566e-06, + "loss": 0.344, + "step": 68190 + }, + { + "epoch": 2.62, + "learning_rate": 6.308668889458836e-06, + "loss": 0.2068, + "step": 68191 + }, + { + "epoch": 2.62, + "learning_rate": 6.308027699410106e-06, + "loss": 0.1897, + "step": 68192 + }, + { + "epoch": 2.62, + "learning_rate": 6.307386509361375e-06, + "loss": 0.2098, + "step": 68193 + }, + { + "epoch": 2.62, + "learning_rate": 6.306745319312645e-06, + "loss": 0.2308, + "step": 68194 + }, + { + "epoch": 2.62, + "learning_rate": 6.306104129263914e-06, + "loss": 0.1197, + "step": 68195 + }, + { + "epoch": 2.62, + "learning_rate": 6.305462939215184e-06, + "loss": 0.274, + "step": 68196 + }, + { + "epoch": 2.62, + "learning_rate": 6.304821749166453e-06, + "loss": 0.0981, + "step": 68197 + }, + { + "epoch": 2.62, + "learning_rate": 6.304180559117722e-06, + "loss": 0.2199, + "step": 68198 + }, + { + "epoch": 2.62, + "learning_rate": 6.303539369068992e-06, + "loss": 0.2109, + "step": 68199 + }, + { + "epoch": 2.62, + "learning_rate": 6.3028981790202616e-06, + "loss": 0.1211, + "step": 68200 + }, + { + "epoch": 2.62, + "learning_rate": 6.302256988971532e-06, + "loss": 0.2233, + "step": 68201 + }, + { + "epoch": 2.62, + "learning_rate": 6.301615798922801e-06, + "loss": 0.28, + "step": 68202 + }, + { + "epoch": 2.62, + "learning_rate": 6.300974608874071e-06, + "loss": 0.1691, + "step": 68203 + }, + { + "epoch": 2.62, + "learning_rate": 6.30033341882534e-06, + "loss": 0.2513, + "step": 68204 + }, + { + "epoch": 2.62, + "learning_rate": 6.29969222877661e-06, + "loss": 0.2572, + "step": 68205 + }, + { + "epoch": 2.62, + "learning_rate": 6.299051038727879e-06, + "loss": 0.2169, + "step": 68206 + }, + { + "epoch": 2.62, + "learning_rate": 6.298409848679148e-06, + "loss": 0.1476, + "step": 68207 + }, + { + "epoch": 2.62, + "learning_rate": 6.297768658630418e-06, + "loss": 0.2417, + "step": 68208 + }, + { + "epoch": 2.62, + "learning_rate": 6.2971274685816875e-06, + "loss": 0.1138, + "step": 68209 + }, + { + "epoch": 2.62, + "learning_rate": 6.2964862785329575e-06, + "loss": 0.3007, + "step": 68210 + }, + { + "epoch": 2.62, + "learning_rate": 6.295845088484227e-06, + "loss": 0.1935, + "step": 68211 + }, + { + "epoch": 2.62, + "learning_rate": 6.295203898435497e-06, + "loss": 0.157, + "step": 68212 + }, + { + "epoch": 2.62, + "learning_rate": 6.294562708386766e-06, + "loss": 0.1505, + "step": 68213 + }, + { + "epoch": 2.62, + "learning_rate": 6.293921518338036e-06, + "loss": 0.5211, + "step": 68214 + }, + { + "epoch": 2.62, + "learning_rate": 6.293280328289305e-06, + "loss": 0.2278, + "step": 68215 + }, + { + "epoch": 2.62, + "learning_rate": 6.292639138240575e-06, + "loss": 0.1561, + "step": 68216 + }, + { + "epoch": 2.62, + "learning_rate": 6.291997948191844e-06, + "loss": 0.2581, + "step": 68217 + }, + { + "epoch": 2.62, + "learning_rate": 6.291356758143114e-06, + "loss": 0.1269, + "step": 68218 + }, + { + "epoch": 2.62, + "learning_rate": 6.2907155680943835e-06, + "loss": 0.3081, + "step": 68219 + }, + { + "epoch": 2.62, + "learning_rate": 6.290074378045653e-06, + "loss": 0.2168, + "step": 68220 + }, + { + "epoch": 2.62, + "learning_rate": 6.289433187996923e-06, + "loss": 0.2126, + "step": 68221 + }, + { + "epoch": 2.62, + "learning_rate": 6.288791997948192e-06, + "loss": 0.2132, + "step": 68222 + }, + { + "epoch": 2.62, + "learning_rate": 6.288150807899462e-06, + "loss": 0.1815, + "step": 68223 + }, + { + "epoch": 2.62, + "learning_rate": 6.287509617850731e-06, + "loss": 0.2, + "step": 68224 + }, + { + "epoch": 2.62, + "learning_rate": 6.286868427802001e-06, + "loss": 0.3078, + "step": 68225 + }, + { + "epoch": 2.62, + "learning_rate": 6.28622723775327e-06, + "loss": 0.1628, + "step": 68226 + }, + { + "epoch": 2.62, + "learning_rate": 6.28558604770454e-06, + "loss": 0.1425, + "step": 68227 + }, + { + "epoch": 2.62, + "learning_rate": 6.2849448576558095e-06, + "loss": 0.1691, + "step": 68228 + }, + { + "epoch": 2.62, + "learning_rate": 6.284303667607079e-06, + "loss": 0.1512, + "step": 68229 + }, + { + "epoch": 2.62, + "learning_rate": 6.283662477558349e-06, + "loss": 0.1724, + "step": 68230 + }, + { + "epoch": 2.62, + "learning_rate": 6.283021287509618e-06, + "loss": 0.3454, + "step": 68231 + }, + { + "epoch": 2.62, + "learning_rate": 6.282380097460888e-06, + "loss": 0.3387, + "step": 68232 + }, + { + "epoch": 2.62, + "learning_rate": 6.281738907412157e-06, + "loss": 0.1856, + "step": 68233 + }, + { + "epoch": 2.62, + "learning_rate": 6.281097717363427e-06, + "loss": 0.1095, + "step": 68234 + }, + { + "epoch": 2.62, + "learning_rate": 6.280456527314696e-06, + "loss": 0.2278, + "step": 68235 + }, + { + "epoch": 2.62, + "learning_rate": 6.279815337265966e-06, + "loss": 0.3159, + "step": 68236 + }, + { + "epoch": 2.62, + "learning_rate": 6.2791741472172354e-06, + "loss": 0.4027, + "step": 68237 + }, + { + "epoch": 2.62, + "learning_rate": 6.278532957168505e-06, + "loss": 0.2652, + "step": 68238 + }, + { + "epoch": 2.62, + "learning_rate": 6.277891767119775e-06, + "loss": 0.2575, + "step": 68239 + }, + { + "epoch": 2.62, + "learning_rate": 6.277250577071044e-06, + "loss": 0.1834, + "step": 68240 + }, + { + "epoch": 2.62, + "learning_rate": 6.276609387022314e-06, + "loss": 0.1435, + "step": 68241 + }, + { + "epoch": 2.62, + "learning_rate": 6.275968196973583e-06, + "loss": 0.0865, + "step": 68242 + }, + { + "epoch": 2.62, + "learning_rate": 6.275327006924853e-06, + "loss": 0.2127, + "step": 68243 + }, + { + "epoch": 2.62, + "learning_rate": 6.274685816876122e-06, + "loss": 0.2338, + "step": 68244 + }, + { + "epoch": 2.62, + "learning_rate": 6.274044626827392e-06, + "loss": 0.2442, + "step": 68245 + }, + { + "epoch": 2.62, + "learning_rate": 6.273403436778661e-06, + "loss": 0.1414, + "step": 68246 + }, + { + "epoch": 2.62, + "learning_rate": 6.272762246729931e-06, + "loss": 0.2253, + "step": 68247 + }, + { + "epoch": 2.62, + "learning_rate": 6.272121056681201e-06, + "loss": 0.2629, + "step": 68248 + }, + { + "epoch": 2.62, + "learning_rate": 6.27147986663247e-06, + "loss": 0.1705, + "step": 68249 + }, + { + "epoch": 2.62, + "learning_rate": 6.27083867658374e-06, + "loss": 0.1761, + "step": 68250 + }, + { + "epoch": 2.62, + "learning_rate": 6.270197486535009e-06, + "loss": 0.233, + "step": 68251 + }, + { + "epoch": 2.62, + "learning_rate": 6.269556296486279e-06, + "loss": 0.1558, + "step": 68252 + }, + { + "epoch": 2.62, + "learning_rate": 6.268915106437548e-06, + "loss": 0.1944, + "step": 68253 + }, + { + "epoch": 2.62, + "learning_rate": 6.268273916388818e-06, + "loss": 0.168, + "step": 68254 + }, + { + "epoch": 2.62, + "learning_rate": 6.267632726340087e-06, + "loss": 0.163, + "step": 68255 + }, + { + "epoch": 2.62, + "learning_rate": 6.266991536291357e-06, + "loss": 0.0873, + "step": 68256 + }, + { + "epoch": 2.62, + "learning_rate": 6.266350346242627e-06, + "loss": 0.2559, + "step": 68257 + }, + { + "epoch": 2.62, + "learning_rate": 6.265709156193896e-06, + "loss": 0.1805, + "step": 68258 + }, + { + "epoch": 2.62, + "learning_rate": 6.265067966145166e-06, + "loss": 0.1845, + "step": 68259 + }, + { + "epoch": 2.62, + "learning_rate": 6.264426776096435e-06, + "loss": 0.3513, + "step": 68260 + }, + { + "epoch": 2.62, + "learning_rate": 6.263785586047705e-06, + "loss": 0.2775, + "step": 68261 + }, + { + "epoch": 2.62, + "learning_rate": 6.263144395998974e-06, + "loss": 0.1505, + "step": 68262 + }, + { + "epoch": 2.62, + "learning_rate": 6.262503205950244e-06, + "loss": 0.269, + "step": 68263 + }, + { + "epoch": 2.62, + "learning_rate": 6.261862015901513e-06, + "loss": 0.1526, + "step": 68264 + }, + { + "epoch": 2.62, + "learning_rate": 6.261220825852783e-06, + "loss": 0.1885, + "step": 68265 + }, + { + "epoch": 2.62, + "learning_rate": 6.2605796358040526e-06, + "loss": 0.2414, + "step": 68266 + }, + { + "epoch": 2.62, + "learning_rate": 6.2599384457553226e-06, + "loss": 0.1958, + "step": 68267 + }, + { + "epoch": 2.62, + "learning_rate": 6.259297255706592e-06, + "loss": 0.3192, + "step": 68268 + }, + { + "epoch": 2.62, + "learning_rate": 6.258656065657861e-06, + "loss": 0.1583, + "step": 68269 + }, + { + "epoch": 2.62, + "learning_rate": 6.258014875609131e-06, + "loss": 0.2125, + "step": 68270 + }, + { + "epoch": 2.62, + "learning_rate": 6.2573736855604e-06, + "loss": 0.2806, + "step": 68271 + }, + { + "epoch": 2.62, + "learning_rate": 6.25673249551167e-06, + "loss": 0.1948, + "step": 68272 + }, + { + "epoch": 2.62, + "learning_rate": 6.256091305462939e-06, + "loss": 0.2275, + "step": 68273 + }, + { + "epoch": 2.62, + "learning_rate": 6.255450115414209e-06, + "loss": 0.2717, + "step": 68274 + }, + { + "epoch": 2.62, + "learning_rate": 6.2548089253654785e-06, + "loss": 0.1864, + "step": 68275 + }, + { + "epoch": 2.62, + "learning_rate": 6.2541677353167485e-06, + "loss": 0.2699, + "step": 68276 + }, + { + "epoch": 2.63, + "learning_rate": 6.253526545268018e-06, + "loss": 0.2578, + "step": 68277 + }, + { + "epoch": 2.63, + "learning_rate": 6.252885355219287e-06, + "loss": 0.0946, + "step": 68278 + }, + { + "epoch": 2.63, + "learning_rate": 6.252244165170557e-06, + "loss": 0.1501, + "step": 68279 + }, + { + "epoch": 2.63, + "learning_rate": 6.251602975121826e-06, + "loss": 0.1487, + "step": 68280 + }, + { + "epoch": 2.63, + "learning_rate": 6.250961785073096e-06, + "loss": 0.2704, + "step": 68281 + }, + { + "epoch": 2.63, + "learning_rate": 6.250320595024365e-06, + "loss": 0.2027, + "step": 68282 + }, + { + "epoch": 2.63, + "learning_rate": 6.249679404975635e-06, + "loss": 0.1612, + "step": 68283 + }, + { + "epoch": 2.63, + "learning_rate": 6.2490382149269045e-06, + "loss": 0.1493, + "step": 68284 + }, + { + "epoch": 2.63, + "learning_rate": 6.2483970248781745e-06, + "loss": 0.1312, + "step": 68285 + }, + { + "epoch": 2.63, + "learning_rate": 6.247755834829444e-06, + "loss": 0.2217, + "step": 68286 + }, + { + "epoch": 2.63, + "learning_rate": 6.247114644780714e-06, + "loss": 0.3387, + "step": 68287 + }, + { + "epoch": 2.63, + "learning_rate": 6.246473454731983e-06, + "loss": 0.1479, + "step": 68288 + }, + { + "epoch": 2.63, + "learning_rate": 6.245832264683252e-06, + "loss": 0.1659, + "step": 68289 + }, + { + "epoch": 2.63, + "learning_rate": 6.245191074634522e-06, + "loss": 0.133, + "step": 68290 + }, + { + "epoch": 2.63, + "learning_rate": 6.244549884585791e-06, + "loss": 0.2098, + "step": 68291 + }, + { + "epoch": 2.63, + "learning_rate": 6.243908694537061e-06, + "loss": 0.1319, + "step": 68292 + }, + { + "epoch": 2.63, + "learning_rate": 6.2432675044883305e-06, + "loss": 0.2141, + "step": 68293 + }, + { + "epoch": 2.63, + "learning_rate": 6.2426263144396005e-06, + "loss": 0.2113, + "step": 68294 + }, + { + "epoch": 2.63, + "learning_rate": 6.24198512439087e-06, + "loss": 0.1842, + "step": 68295 + }, + { + "epoch": 2.63, + "learning_rate": 6.24134393434214e-06, + "loss": 0.3591, + "step": 68296 + }, + { + "epoch": 2.63, + "learning_rate": 6.240702744293409e-06, + "loss": 0.2251, + "step": 68297 + }, + { + "epoch": 2.63, + "learning_rate": 6.240061554244678e-06, + "loss": 0.2565, + "step": 68298 + }, + { + "epoch": 2.63, + "learning_rate": 6.239420364195948e-06, + "loss": 0.3231, + "step": 68299 + }, + { + "epoch": 2.63, + "learning_rate": 6.238779174147217e-06, + "loss": 0.2735, + "step": 68300 + }, + { + "epoch": 2.63, + "learning_rate": 6.238137984098487e-06, + "loss": 0.2291, + "step": 68301 + }, + { + "epoch": 2.63, + "learning_rate": 6.237496794049756e-06, + "loss": 0.3657, + "step": 68302 + }, + { + "epoch": 2.63, + "learning_rate": 6.2368556040010264e-06, + "loss": 0.2277, + "step": 68303 + }, + { + "epoch": 2.63, + "learning_rate": 6.236214413952296e-06, + "loss": 0.1296, + "step": 68304 + }, + { + "epoch": 2.63, + "learning_rate": 6.235573223903566e-06, + "loss": 0.3142, + "step": 68305 + }, + { + "epoch": 2.63, + "learning_rate": 6.234932033854835e-06, + "loss": 0.2064, + "step": 68306 + }, + { + "epoch": 2.63, + "learning_rate": 6.234290843806105e-06, + "loss": 0.2635, + "step": 68307 + }, + { + "epoch": 2.63, + "learning_rate": 6.233649653757374e-06, + "loss": 0.1865, + "step": 68308 + }, + { + "epoch": 2.63, + "learning_rate": 6.233008463708643e-06, + "loss": 0.2415, + "step": 68309 + }, + { + "epoch": 2.63, + "learning_rate": 6.232367273659913e-06, + "loss": 0.1571, + "step": 68310 + }, + { + "epoch": 2.63, + "learning_rate": 6.231726083611182e-06, + "loss": 0.2618, + "step": 68311 + }, + { + "epoch": 2.63, + "learning_rate": 6.231084893562452e-06, + "loss": 0.1805, + "step": 68312 + }, + { + "epoch": 2.63, + "learning_rate": 6.230443703513722e-06, + "loss": 0.2963, + "step": 68313 + }, + { + "epoch": 2.63, + "learning_rate": 6.229802513464992e-06, + "loss": 0.1346, + "step": 68314 + }, + { + "epoch": 2.63, + "learning_rate": 6.229161323416261e-06, + "loss": 0.2015, + "step": 68315 + }, + { + "epoch": 2.63, + "learning_rate": 6.228520133367531e-06, + "loss": 0.2041, + "step": 68316 + }, + { + "epoch": 2.63, + "learning_rate": 6.2278789433188e-06, + "loss": 0.1766, + "step": 68317 + }, + { + "epoch": 2.63, + "learning_rate": 6.22723775327007e-06, + "loss": 0.1613, + "step": 68318 + }, + { + "epoch": 2.63, + "learning_rate": 6.226596563221339e-06, + "loss": 0.2716, + "step": 68319 + }, + { + "epoch": 2.63, + "learning_rate": 6.225955373172608e-06, + "loss": 0.2958, + "step": 68320 + }, + { + "epoch": 2.63, + "learning_rate": 6.225314183123878e-06, + "loss": 0.3453, + "step": 68321 + }, + { + "epoch": 2.63, + "learning_rate": 6.2246729930751476e-06, + "loss": 0.1636, + "step": 68322 + }, + { + "epoch": 2.63, + "learning_rate": 6.224031803026418e-06, + "loss": 0.2854, + "step": 68323 + }, + { + "epoch": 2.63, + "learning_rate": 6.223390612977687e-06, + "loss": 0.2373, + "step": 68324 + }, + { + "epoch": 2.63, + "learning_rate": 6.222749422928957e-06, + "loss": 0.186, + "step": 68325 + }, + { + "epoch": 2.63, + "learning_rate": 6.222108232880226e-06, + "loss": 0.1121, + "step": 68326 + }, + { + "epoch": 2.63, + "learning_rate": 6.221467042831496e-06, + "loss": 0.1876, + "step": 68327 + }, + { + "epoch": 2.63, + "learning_rate": 6.220825852782765e-06, + "loss": 0.1816, + "step": 68328 + }, + { + "epoch": 2.63, + "learning_rate": 6.220184662734034e-06, + "loss": 0.1676, + "step": 68329 + }, + { + "epoch": 2.63, + "learning_rate": 6.219543472685304e-06, + "loss": 0.1392, + "step": 68330 + }, + { + "epoch": 2.63, + "learning_rate": 6.2189022826365735e-06, + "loss": 0.2953, + "step": 68331 + }, + { + "epoch": 2.63, + "learning_rate": 6.2182610925878435e-06, + "loss": 0.1578, + "step": 68332 + }, + { + "epoch": 2.63, + "learning_rate": 6.217619902539113e-06, + "loss": 0.228, + "step": 68333 + }, + { + "epoch": 2.63, + "learning_rate": 6.216978712490383e-06, + "loss": 0.131, + "step": 68334 + }, + { + "epoch": 2.63, + "learning_rate": 6.216337522441652e-06, + "loss": 0.1588, + "step": 68335 + }, + { + "epoch": 2.63, + "learning_rate": 6.215696332392922e-06, + "loss": 0.0918, + "step": 68336 + }, + { + "epoch": 2.63, + "learning_rate": 6.215055142344191e-06, + "loss": 0.1398, + "step": 68337 + }, + { + "epoch": 2.63, + "learning_rate": 6.214413952295461e-06, + "loss": 0.1909, + "step": 68338 + }, + { + "epoch": 2.63, + "learning_rate": 6.21377276224673e-06, + "loss": 0.23, + "step": 68339 + }, + { + "epoch": 2.63, + "learning_rate": 6.2131315721979995e-06, + "loss": 0.257, + "step": 68340 + }, + { + "epoch": 2.63, + "learning_rate": 6.2124903821492695e-06, + "loss": 0.2079, + "step": 68341 + }, + { + "epoch": 2.63, + "learning_rate": 6.211849192100539e-06, + "loss": 0.1027, + "step": 68342 + }, + { + "epoch": 2.63, + "learning_rate": 6.211208002051809e-06, + "loss": 0.2031, + "step": 68343 + }, + { + "epoch": 2.63, + "learning_rate": 6.210566812003078e-06, + "loss": 0.1983, + "step": 68344 + }, + { + "epoch": 2.63, + "learning_rate": 6.209925621954348e-06, + "loss": 0.2369, + "step": 68345 + }, + { + "epoch": 2.63, + "learning_rate": 6.209284431905617e-06, + "loss": 0.2752, + "step": 68346 + }, + { + "epoch": 2.63, + "learning_rate": 6.208643241856887e-06, + "loss": 0.197, + "step": 68347 + }, + { + "epoch": 2.63, + "learning_rate": 6.208002051808156e-06, + "loss": 0.2092, + "step": 68348 + }, + { + "epoch": 2.63, + "learning_rate": 6.2073608617594255e-06, + "loss": 0.1502, + "step": 68349 + }, + { + "epoch": 2.63, + "learning_rate": 6.2067196717106955e-06, + "loss": 0.2631, + "step": 68350 + }, + { + "epoch": 2.63, + "learning_rate": 6.206078481661965e-06, + "loss": 0.2672, + "step": 68351 + }, + { + "epoch": 2.63, + "learning_rate": 6.205437291613235e-06, + "loss": 0.213, + "step": 68352 + }, + { + "epoch": 2.63, + "learning_rate": 6.204796101564504e-06, + "loss": 0.1972, + "step": 68353 + }, + { + "epoch": 2.63, + "learning_rate": 6.204154911515774e-06, + "loss": 0.1576, + "step": 68354 + }, + { + "epoch": 2.63, + "learning_rate": 6.203513721467043e-06, + "loss": 0.1129, + "step": 68355 + }, + { + "epoch": 2.63, + "learning_rate": 6.202872531418313e-06, + "loss": 0.1833, + "step": 68356 + }, + { + "epoch": 2.63, + "learning_rate": 6.202231341369582e-06, + "loss": 0.1108, + "step": 68357 + }, + { + "epoch": 2.63, + "learning_rate": 6.201590151320852e-06, + "loss": 0.1264, + "step": 68358 + }, + { + "epoch": 2.63, + "learning_rate": 6.2009489612721214e-06, + "loss": 0.3058, + "step": 68359 + }, + { + "epoch": 2.63, + "learning_rate": 6.200307771223391e-06, + "loss": 0.3677, + "step": 68360 + }, + { + "epoch": 2.63, + "learning_rate": 6.199666581174661e-06, + "loss": 0.1968, + "step": 68361 + }, + { + "epoch": 2.63, + "learning_rate": 6.19902539112593e-06, + "loss": 0.2588, + "step": 68362 + }, + { + "epoch": 2.63, + "learning_rate": 6.1983842010772e-06, + "loss": 0.3117, + "step": 68363 + }, + { + "epoch": 2.63, + "learning_rate": 6.197743011028469e-06, + "loss": 0.1571, + "step": 68364 + }, + { + "epoch": 2.63, + "learning_rate": 6.197101820979739e-06, + "loss": 0.3261, + "step": 68365 + }, + { + "epoch": 2.63, + "learning_rate": 6.196460630931008e-06, + "loss": 0.2868, + "step": 68366 + }, + { + "epoch": 2.63, + "learning_rate": 6.195819440882278e-06, + "loss": 0.2086, + "step": 68367 + }, + { + "epoch": 2.63, + "learning_rate": 6.195178250833547e-06, + "loss": 0.1697, + "step": 68368 + }, + { + "epoch": 2.63, + "learning_rate": 6.1945370607848174e-06, + "loss": 0.1594, + "step": 68369 + }, + { + "epoch": 2.63, + "learning_rate": 6.193895870736087e-06, + "loss": 0.202, + "step": 68370 + }, + { + "epoch": 2.63, + "learning_rate": 6.193254680687356e-06, + "loss": 0.1194, + "step": 68371 + }, + { + "epoch": 2.63, + "learning_rate": 6.192613490638626e-06, + "loss": 0.1933, + "step": 68372 + }, + { + "epoch": 2.63, + "learning_rate": 6.191972300589895e-06, + "loss": 0.0863, + "step": 68373 + }, + { + "epoch": 2.63, + "learning_rate": 6.191331110541165e-06, + "loss": 0.229, + "step": 68374 + }, + { + "epoch": 2.63, + "learning_rate": 6.190689920492434e-06, + "loss": 0.2352, + "step": 68375 + }, + { + "epoch": 2.63, + "learning_rate": 6.190048730443704e-06, + "loss": 0.1915, + "step": 68376 + }, + { + "epoch": 2.63, + "learning_rate": 6.189407540394973e-06, + "loss": 0.188, + "step": 68377 + }, + { + "epoch": 2.63, + "learning_rate": 6.188766350346243e-06, + "loss": 0.1619, + "step": 68378 + }, + { + "epoch": 2.63, + "learning_rate": 6.188125160297513e-06, + "loss": 0.2267, + "step": 68379 + }, + { + "epoch": 2.63, + "learning_rate": 6.187483970248782e-06, + "loss": 0.2209, + "step": 68380 + }, + { + "epoch": 2.63, + "learning_rate": 6.186842780200052e-06, + "loss": 0.1034, + "step": 68381 + }, + { + "epoch": 2.63, + "learning_rate": 6.186201590151321e-06, + "loss": 0.245, + "step": 68382 + }, + { + "epoch": 2.63, + "learning_rate": 6.185560400102591e-06, + "loss": 0.1842, + "step": 68383 + }, + { + "epoch": 2.63, + "learning_rate": 6.18491921005386e-06, + "loss": 0.2667, + "step": 68384 + }, + { + "epoch": 2.63, + "learning_rate": 6.18427802000513e-06, + "loss": 0.2616, + "step": 68385 + }, + { + "epoch": 2.63, + "learning_rate": 6.183636829956399e-06, + "loss": 0.237, + "step": 68386 + }, + { + "epoch": 2.63, + "learning_rate": 6.182995639907669e-06, + "loss": 0.1215, + "step": 68387 + }, + { + "epoch": 2.63, + "learning_rate": 6.1823544498589386e-06, + "loss": 0.1661, + "step": 68388 + }, + { + "epoch": 2.63, + "learning_rate": 6.1817132598102086e-06, + "loss": 0.2522, + "step": 68389 + }, + { + "epoch": 2.63, + "learning_rate": 6.181072069761478e-06, + "loss": 0.3139, + "step": 68390 + }, + { + "epoch": 2.63, + "learning_rate": 6.180430879712747e-06, + "loss": 0.1725, + "step": 68391 + }, + { + "epoch": 2.63, + "learning_rate": 6.179789689664017e-06, + "loss": 0.1531, + "step": 68392 + }, + { + "epoch": 2.63, + "learning_rate": 6.179148499615286e-06, + "loss": 0.0586, + "step": 68393 + }, + { + "epoch": 2.63, + "learning_rate": 6.178507309566556e-06, + "loss": 0.2698, + "step": 68394 + }, + { + "epoch": 2.63, + "learning_rate": 6.177866119517825e-06, + "loss": 0.2061, + "step": 68395 + }, + { + "epoch": 2.63, + "learning_rate": 6.177224929469095e-06, + "loss": 0.2624, + "step": 68396 + }, + { + "epoch": 2.63, + "learning_rate": 6.1765837394203645e-06, + "loss": 0.2448, + "step": 68397 + }, + { + "epoch": 2.63, + "learning_rate": 6.1759425493716345e-06, + "loss": 0.2976, + "step": 68398 + }, + { + "epoch": 2.63, + "learning_rate": 6.175301359322904e-06, + "loss": 0.3109, + "step": 68399 + }, + { + "epoch": 2.63, + "learning_rate": 6.174660169274174e-06, + "loss": 0.1653, + "step": 68400 + }, + { + "epoch": 2.63, + "learning_rate": 6.174018979225443e-06, + "loss": 0.1594, + "step": 68401 + }, + { + "epoch": 2.63, + "learning_rate": 6.173377789176712e-06, + "loss": 0.0829, + "step": 68402 + }, + { + "epoch": 2.63, + "learning_rate": 6.172736599127982e-06, + "loss": 0.2154, + "step": 68403 + }, + { + "epoch": 2.63, + "learning_rate": 6.172095409079251e-06, + "loss": 0.3113, + "step": 68404 + }, + { + "epoch": 2.63, + "learning_rate": 6.171454219030521e-06, + "loss": 0.2143, + "step": 68405 + }, + { + "epoch": 2.63, + "learning_rate": 6.1708130289817905e-06, + "loss": 0.3285, + "step": 68406 + }, + { + "epoch": 2.63, + "learning_rate": 6.1701718389330605e-06, + "loss": 0.29, + "step": 68407 + }, + { + "epoch": 2.63, + "learning_rate": 6.16953064888433e-06, + "loss": 0.1707, + "step": 68408 + }, + { + "epoch": 2.63, + "learning_rate": 6.1688894588356e-06, + "loss": 0.1558, + "step": 68409 + }, + { + "epoch": 2.63, + "learning_rate": 6.168248268786869e-06, + "loss": 0.3397, + "step": 68410 + }, + { + "epoch": 2.63, + "learning_rate": 6.167607078738138e-06, + "loss": 0.2083, + "step": 68411 + }, + { + "epoch": 2.63, + "learning_rate": 6.166965888689408e-06, + "loss": 0.2054, + "step": 68412 + }, + { + "epoch": 2.63, + "learning_rate": 6.166324698640677e-06, + "loss": 0.3178, + "step": 68413 + }, + { + "epoch": 2.63, + "learning_rate": 6.165683508591947e-06, + "loss": 0.0627, + "step": 68414 + }, + { + "epoch": 2.63, + "learning_rate": 6.1650423185432165e-06, + "loss": 0.2513, + "step": 68415 + }, + { + "epoch": 2.63, + "learning_rate": 6.1644011284944865e-06, + "loss": 0.2662, + "step": 68416 + }, + { + "epoch": 2.63, + "learning_rate": 6.163759938445756e-06, + "loss": 0.2308, + "step": 68417 + }, + { + "epoch": 2.63, + "learning_rate": 6.163118748397026e-06, + "loss": 0.3774, + "step": 68418 + }, + { + "epoch": 2.63, + "learning_rate": 6.162477558348295e-06, + "loss": 0.2714, + "step": 68419 + }, + { + "epoch": 2.63, + "learning_rate": 6.161836368299565e-06, + "loss": 0.2444, + "step": 68420 + }, + { + "epoch": 2.63, + "learning_rate": 6.161195178250834e-06, + "loss": 0.1793, + "step": 68421 + }, + { + "epoch": 2.63, + "learning_rate": 6.160553988202103e-06, + "loss": 0.2391, + "step": 68422 + }, + { + "epoch": 2.63, + "learning_rate": 6.159912798153373e-06, + "loss": 0.1629, + "step": 68423 + }, + { + "epoch": 2.63, + "learning_rate": 6.159271608104642e-06, + "loss": 0.2159, + "step": 68424 + }, + { + "epoch": 2.63, + "learning_rate": 6.1586304180559124e-06, + "loss": 0.1616, + "step": 68425 + }, + { + "epoch": 2.63, + "learning_rate": 6.157989228007182e-06, + "loss": 0.2687, + "step": 68426 + }, + { + "epoch": 2.63, + "learning_rate": 6.157348037958452e-06, + "loss": 0.1619, + "step": 68427 + }, + { + "epoch": 2.63, + "learning_rate": 6.156706847909721e-06, + "loss": 0.2722, + "step": 68428 + }, + { + "epoch": 2.63, + "learning_rate": 6.156065657860991e-06, + "loss": 0.1759, + "step": 68429 + }, + { + "epoch": 2.63, + "learning_rate": 6.15542446781226e-06, + "loss": 0.1289, + "step": 68430 + }, + { + "epoch": 2.63, + "learning_rate": 6.154783277763529e-06, + "loss": 0.1635, + "step": 68431 + }, + { + "epoch": 2.63, + "learning_rate": 6.154142087714799e-06, + "loss": 0.1084, + "step": 68432 + }, + { + "epoch": 2.63, + "learning_rate": 6.153500897666068e-06, + "loss": 0.1994, + "step": 68433 + }, + { + "epoch": 2.63, + "learning_rate": 6.152859707617338e-06, + "loss": 0.1208, + "step": 68434 + }, + { + "epoch": 2.63, + "learning_rate": 6.152218517568608e-06, + "loss": 0.1557, + "step": 68435 + }, + { + "epoch": 2.63, + "learning_rate": 6.151577327519878e-06, + "loss": 0.225, + "step": 68436 + }, + { + "epoch": 2.63, + "learning_rate": 6.150936137471147e-06, + "loss": 0.1804, + "step": 68437 + }, + { + "epoch": 2.63, + "learning_rate": 6.150294947422417e-06, + "loss": 0.2442, + "step": 68438 + }, + { + "epoch": 2.63, + "learning_rate": 6.149653757373686e-06, + "loss": 0.2539, + "step": 68439 + }, + { + "epoch": 2.63, + "learning_rate": 6.149012567324956e-06, + "loss": 0.2342, + "step": 68440 + }, + { + "epoch": 2.63, + "learning_rate": 6.148371377276225e-06, + "loss": 0.2032, + "step": 68441 + }, + { + "epoch": 2.63, + "learning_rate": 6.147730187227494e-06, + "loss": 0.159, + "step": 68442 + }, + { + "epoch": 2.63, + "learning_rate": 6.147088997178764e-06, + "loss": 0.252, + "step": 68443 + }, + { + "epoch": 2.63, + "learning_rate": 6.1464478071300336e-06, + "loss": 0.2056, + "step": 68444 + }, + { + "epoch": 2.63, + "learning_rate": 6.145806617081304e-06, + "loss": 0.2319, + "step": 68445 + }, + { + "epoch": 2.63, + "learning_rate": 6.145165427032573e-06, + "loss": 0.2343, + "step": 68446 + }, + { + "epoch": 2.63, + "learning_rate": 6.144524236983843e-06, + "loss": 0.3562, + "step": 68447 + }, + { + "epoch": 2.63, + "learning_rate": 6.143883046935112e-06, + "loss": 0.065, + "step": 68448 + }, + { + "epoch": 2.63, + "learning_rate": 6.143241856886382e-06, + "loss": 0.1441, + "step": 68449 + }, + { + "epoch": 2.63, + "learning_rate": 6.142600666837651e-06, + "loss": 0.2133, + "step": 68450 + }, + { + "epoch": 2.63, + "learning_rate": 6.141959476788921e-06, + "loss": 0.088, + "step": 68451 + }, + { + "epoch": 2.63, + "learning_rate": 6.14131828674019e-06, + "loss": 0.1502, + "step": 68452 + }, + { + "epoch": 2.63, + "learning_rate": 6.1406770966914595e-06, + "loss": 0.2956, + "step": 68453 + }, + { + "epoch": 2.63, + "learning_rate": 6.1400359066427295e-06, + "loss": 0.177, + "step": 68454 + }, + { + "epoch": 2.63, + "learning_rate": 6.139394716593999e-06, + "loss": 0.2613, + "step": 68455 + }, + { + "epoch": 2.63, + "learning_rate": 6.138753526545269e-06, + "loss": 0.2085, + "step": 68456 + }, + { + "epoch": 2.63, + "learning_rate": 6.138112336496538e-06, + "loss": 0.2401, + "step": 68457 + }, + { + "epoch": 2.63, + "learning_rate": 6.137471146447808e-06, + "loss": 0.287, + "step": 68458 + }, + { + "epoch": 2.63, + "learning_rate": 6.136829956399077e-06, + "loss": 0.1618, + "step": 68459 + }, + { + "epoch": 2.63, + "learning_rate": 6.136188766350347e-06, + "loss": 0.3687, + "step": 68460 + }, + { + "epoch": 2.63, + "learning_rate": 6.135547576301616e-06, + "loss": 0.3145, + "step": 68461 + }, + { + "epoch": 2.63, + "learning_rate": 6.1349063862528855e-06, + "loss": 0.2389, + "step": 68462 + }, + { + "epoch": 2.63, + "learning_rate": 6.1342651962041555e-06, + "loss": 0.2352, + "step": 68463 + }, + { + "epoch": 2.63, + "learning_rate": 6.133624006155425e-06, + "loss": 0.1865, + "step": 68464 + }, + { + "epoch": 2.63, + "learning_rate": 6.132982816106695e-06, + "loss": 0.2053, + "step": 68465 + }, + { + "epoch": 2.63, + "learning_rate": 6.132341626057964e-06, + "loss": 0.2738, + "step": 68466 + }, + { + "epoch": 2.63, + "learning_rate": 6.131700436009234e-06, + "loss": 0.1631, + "step": 68467 + }, + { + "epoch": 2.63, + "learning_rate": 6.131059245960503e-06, + "loss": 0.1733, + "step": 68468 + }, + { + "epoch": 2.63, + "learning_rate": 6.130418055911773e-06, + "loss": 0.3576, + "step": 68469 + }, + { + "epoch": 2.63, + "learning_rate": 6.129776865863042e-06, + "loss": 0.194, + "step": 68470 + }, + { + "epoch": 2.63, + "learning_rate": 6.129135675814312e-06, + "loss": 0.1522, + "step": 68471 + }, + { + "epoch": 2.63, + "learning_rate": 6.1284944857655815e-06, + "loss": 0.1594, + "step": 68472 + }, + { + "epoch": 2.63, + "learning_rate": 6.127853295716851e-06, + "loss": 0.2865, + "step": 68473 + }, + { + "epoch": 2.63, + "learning_rate": 6.127212105668121e-06, + "loss": 0.1464, + "step": 68474 + }, + { + "epoch": 2.63, + "learning_rate": 6.12657091561939e-06, + "loss": 0.1569, + "step": 68475 + }, + { + "epoch": 2.63, + "learning_rate": 6.12592972557066e-06, + "loss": 0.2195, + "step": 68476 + }, + { + "epoch": 2.63, + "learning_rate": 6.125288535521929e-06, + "loss": 0.1557, + "step": 68477 + }, + { + "epoch": 2.63, + "learning_rate": 6.124647345473199e-06, + "loss": 0.198, + "step": 68478 + }, + { + "epoch": 2.63, + "learning_rate": 6.124006155424468e-06, + "loss": 0.1389, + "step": 68479 + }, + { + "epoch": 2.63, + "learning_rate": 6.123364965375738e-06, + "loss": 0.1433, + "step": 68480 + }, + { + "epoch": 2.63, + "learning_rate": 6.1227237753270075e-06, + "loss": 0.2837, + "step": 68481 + }, + { + "epoch": 2.63, + "learning_rate": 6.122082585278277e-06, + "loss": 0.1539, + "step": 68482 + }, + { + "epoch": 2.63, + "learning_rate": 6.121441395229547e-06, + "loss": 0.1159, + "step": 68483 + }, + { + "epoch": 2.63, + "learning_rate": 6.120800205180816e-06, + "loss": 0.388, + "step": 68484 + }, + { + "epoch": 2.63, + "learning_rate": 6.120159015132086e-06, + "loss": 0.1968, + "step": 68485 + }, + { + "epoch": 2.63, + "learning_rate": 6.119517825083355e-06, + "loss": 0.2804, + "step": 68486 + }, + { + "epoch": 2.63, + "learning_rate": 6.118876635034625e-06, + "loss": 0.0878, + "step": 68487 + }, + { + "epoch": 2.63, + "learning_rate": 6.118235444985894e-06, + "loss": 0.174, + "step": 68488 + }, + { + "epoch": 2.63, + "learning_rate": 6.117594254937164e-06, + "loss": 0.1426, + "step": 68489 + }, + { + "epoch": 2.63, + "learning_rate": 6.116953064888433e-06, + "loss": 0.2196, + "step": 68490 + }, + { + "epoch": 2.63, + "learning_rate": 6.1163118748397034e-06, + "loss": 0.1984, + "step": 68491 + }, + { + "epoch": 2.63, + "learning_rate": 6.115670684790973e-06, + "loss": 0.4045, + "step": 68492 + }, + { + "epoch": 2.63, + "learning_rate": 6.115029494742242e-06, + "loss": 0.2997, + "step": 68493 + }, + { + "epoch": 2.63, + "learning_rate": 6.114388304693512e-06, + "loss": 0.1885, + "step": 68494 + }, + { + "epoch": 2.63, + "learning_rate": 6.113747114644781e-06, + "loss": 0.2597, + "step": 68495 + }, + { + "epoch": 2.63, + "learning_rate": 6.113105924596051e-06, + "loss": 0.2145, + "step": 68496 + }, + { + "epoch": 2.63, + "learning_rate": 6.11246473454732e-06, + "loss": 0.2044, + "step": 68497 + }, + { + "epoch": 2.63, + "learning_rate": 6.11182354449859e-06, + "loss": 0.2153, + "step": 68498 + }, + { + "epoch": 2.63, + "learning_rate": 6.111182354449859e-06, + "loss": 0.2504, + "step": 68499 + }, + { + "epoch": 2.63, + "learning_rate": 6.110541164401129e-06, + "loss": 0.1957, + "step": 68500 + }, + { + "epoch": 2.63, + "learning_rate": 6.109899974352399e-06, + "loss": 0.1928, + "step": 68501 + }, + { + "epoch": 2.63, + "learning_rate": 6.109258784303669e-06, + "loss": 0.1841, + "step": 68502 + }, + { + "epoch": 2.63, + "learning_rate": 6.108617594254938e-06, + "loss": 0.2912, + "step": 68503 + }, + { + "epoch": 2.63, + "learning_rate": 6.107976404206207e-06, + "loss": 0.21, + "step": 68504 + }, + { + "epoch": 2.63, + "learning_rate": 6.107335214157477e-06, + "loss": 0.2278, + "step": 68505 + }, + { + "epoch": 2.63, + "learning_rate": 6.106694024108746e-06, + "loss": 0.3021, + "step": 68506 + }, + { + "epoch": 2.63, + "learning_rate": 6.106052834060016e-06, + "loss": 0.1975, + "step": 68507 + }, + { + "epoch": 2.63, + "learning_rate": 6.105411644011285e-06, + "loss": 0.2504, + "step": 68508 + }, + { + "epoch": 2.63, + "learning_rate": 6.104770453962555e-06, + "loss": 0.1326, + "step": 68509 + }, + { + "epoch": 2.63, + "learning_rate": 6.1041292639138246e-06, + "loss": 0.1785, + "step": 68510 + }, + { + "epoch": 2.63, + "learning_rate": 6.1034880738650946e-06, + "loss": 0.2028, + "step": 68511 + }, + { + "epoch": 2.63, + "learning_rate": 6.102846883816364e-06, + "loss": 0.2224, + "step": 68512 + }, + { + "epoch": 2.63, + "learning_rate": 6.102205693767633e-06, + "loss": 0.2166, + "step": 68513 + }, + { + "epoch": 2.63, + "learning_rate": 6.101564503718903e-06, + "loss": 0.2171, + "step": 68514 + }, + { + "epoch": 2.63, + "learning_rate": 6.100923313670172e-06, + "loss": 0.1987, + "step": 68515 + }, + { + "epoch": 2.63, + "learning_rate": 6.100282123621442e-06, + "loss": 0.1967, + "step": 68516 + }, + { + "epoch": 2.63, + "learning_rate": 6.099640933572711e-06, + "loss": 0.1247, + "step": 68517 + }, + { + "epoch": 2.63, + "learning_rate": 6.098999743523981e-06, + "loss": 0.2267, + "step": 68518 + }, + { + "epoch": 2.63, + "learning_rate": 6.0983585534752505e-06, + "loss": 0.1535, + "step": 68519 + }, + { + "epoch": 2.63, + "learning_rate": 6.0977173634265205e-06, + "loss": 0.23, + "step": 68520 + }, + { + "epoch": 2.63, + "learning_rate": 6.09707617337779e-06, + "loss": 0.2171, + "step": 68521 + }, + { + "epoch": 2.63, + "learning_rate": 6.09643498332906e-06, + "loss": 0.1312, + "step": 68522 + }, + { + "epoch": 2.63, + "learning_rate": 6.095793793280329e-06, + "loss": 0.1733, + "step": 68523 + }, + { + "epoch": 2.63, + "learning_rate": 6.095152603231598e-06, + "loss": 0.2981, + "step": 68524 + }, + { + "epoch": 2.63, + "learning_rate": 6.094511413182868e-06, + "loss": 0.2411, + "step": 68525 + }, + { + "epoch": 2.63, + "learning_rate": 6.093870223134137e-06, + "loss": 0.1699, + "step": 68526 + }, + { + "epoch": 2.63, + "learning_rate": 6.093229033085407e-06, + "loss": 0.1493, + "step": 68527 + }, + { + "epoch": 2.63, + "learning_rate": 6.0925878430366765e-06, + "loss": 0.175, + "step": 68528 + }, + { + "epoch": 2.63, + "learning_rate": 6.0919466529879465e-06, + "loss": 0.1513, + "step": 68529 + }, + { + "epoch": 2.63, + "learning_rate": 6.091305462939216e-06, + "loss": 0.2702, + "step": 68530 + }, + { + "epoch": 2.63, + "learning_rate": 6.090664272890486e-06, + "loss": 0.1984, + "step": 68531 + }, + { + "epoch": 2.63, + "learning_rate": 6.090023082841755e-06, + "loss": 0.1779, + "step": 68532 + }, + { + "epoch": 2.63, + "learning_rate": 6.089381892793025e-06, + "loss": 0.1254, + "step": 68533 + }, + { + "epoch": 2.63, + "learning_rate": 6.088740702744294e-06, + "loss": 0.2893, + "step": 68534 + }, + { + "epoch": 2.63, + "learning_rate": 6.088099512695563e-06, + "loss": 0.2224, + "step": 68535 + }, + { + "epoch": 2.63, + "learning_rate": 6.087458322646833e-06, + "loss": 0.2273, + "step": 68536 + }, + { + "epoch": 2.64, + "learning_rate": 6.0868171325981025e-06, + "loss": 0.2139, + "step": 68537 + }, + { + "epoch": 2.64, + "learning_rate": 6.0861759425493725e-06, + "loss": 0.309, + "step": 68538 + }, + { + "epoch": 2.64, + "learning_rate": 6.085534752500642e-06, + "loss": 0.223, + "step": 68539 + }, + { + "epoch": 2.64, + "learning_rate": 6.084893562451912e-06, + "loss": 0.1802, + "step": 68540 + }, + { + "epoch": 2.64, + "learning_rate": 6.084252372403181e-06, + "loss": 0.1125, + "step": 68541 + }, + { + "epoch": 2.64, + "learning_rate": 6.083611182354451e-06, + "loss": 0.1964, + "step": 68542 + }, + { + "epoch": 2.64, + "learning_rate": 6.08296999230572e-06, + "loss": 0.168, + "step": 68543 + }, + { + "epoch": 2.64, + "learning_rate": 6.082328802256989e-06, + "loss": 0.1786, + "step": 68544 + }, + { + "epoch": 2.64, + "learning_rate": 6.081687612208259e-06, + "loss": 0.1764, + "step": 68545 + }, + { + "epoch": 2.64, + "learning_rate": 6.0810464221595284e-06, + "loss": 0.2332, + "step": 68546 + }, + { + "epoch": 2.64, + "learning_rate": 6.0804052321107984e-06, + "loss": 0.1752, + "step": 68547 + }, + { + "epoch": 2.64, + "learning_rate": 6.079764042062068e-06, + "loss": 0.2139, + "step": 68548 + }, + { + "epoch": 2.64, + "learning_rate": 6.079122852013338e-06, + "loss": 0.1958, + "step": 68549 + }, + { + "epoch": 2.64, + "learning_rate": 6.078481661964607e-06, + "loss": 0.1959, + "step": 68550 + }, + { + "epoch": 2.64, + "learning_rate": 6.077840471915877e-06, + "loss": 0.2558, + "step": 68551 + }, + { + "epoch": 2.64, + "learning_rate": 6.077199281867146e-06, + "loss": 0.3135, + "step": 68552 + }, + { + "epoch": 2.64, + "learning_rate": 6.076558091818415e-06, + "loss": 0.2074, + "step": 68553 + }, + { + "epoch": 2.64, + "learning_rate": 6.075916901769685e-06, + "loss": 0.2832, + "step": 68554 + }, + { + "epoch": 2.64, + "learning_rate": 6.075275711720954e-06, + "loss": 0.1615, + "step": 68555 + }, + { + "epoch": 2.64, + "learning_rate": 6.074634521672224e-06, + "loss": 0.2139, + "step": 68556 + }, + { + "epoch": 2.64, + "learning_rate": 6.073993331623494e-06, + "loss": 0.2011, + "step": 68557 + }, + { + "epoch": 2.64, + "learning_rate": 6.073352141574764e-06, + "loss": 0.2806, + "step": 68558 + }, + { + "epoch": 2.64, + "learning_rate": 6.072710951526033e-06, + "loss": 0.1579, + "step": 68559 + }, + { + "epoch": 2.64, + "learning_rate": 6.072069761477303e-06, + "loss": 0.1747, + "step": 68560 + }, + { + "epoch": 2.64, + "learning_rate": 6.071428571428572e-06, + "loss": 0.1524, + "step": 68561 + }, + { + "epoch": 2.64, + "learning_rate": 6.070787381379842e-06, + "loss": 0.2527, + "step": 68562 + }, + { + "epoch": 2.64, + "learning_rate": 6.070146191331111e-06, + "loss": 0.2042, + "step": 68563 + }, + { + "epoch": 2.64, + "learning_rate": 6.06950500128238e-06, + "loss": 0.3632, + "step": 68564 + }, + { + "epoch": 2.64, + "learning_rate": 6.06886381123365e-06, + "loss": 0.2017, + "step": 68565 + }, + { + "epoch": 2.64, + "learning_rate": 6.0682226211849196e-06, + "loss": 0.1784, + "step": 68566 + }, + { + "epoch": 2.64, + "learning_rate": 6.06758143113619e-06, + "loss": 0.3701, + "step": 68567 + }, + { + "epoch": 2.64, + "learning_rate": 6.066940241087459e-06, + "loss": 0.3138, + "step": 68568 + }, + { + "epoch": 2.64, + "learning_rate": 6.066299051038729e-06, + "loss": 0.1371, + "step": 68569 + }, + { + "epoch": 2.64, + "learning_rate": 6.065657860989998e-06, + "loss": 0.1559, + "step": 68570 + }, + { + "epoch": 2.64, + "learning_rate": 6.065016670941268e-06, + "loss": 0.2165, + "step": 68571 + }, + { + "epoch": 2.64, + "learning_rate": 6.064375480892537e-06, + "loss": 0.1435, + "step": 68572 + }, + { + "epoch": 2.64, + "learning_rate": 6.063734290843806e-06, + "loss": 0.2222, + "step": 68573 + }, + { + "epoch": 2.64, + "learning_rate": 6.063093100795076e-06, + "loss": 0.1558, + "step": 68574 + }, + { + "epoch": 2.64, + "learning_rate": 6.0624519107463455e-06, + "loss": 0.2743, + "step": 68575 + }, + { + "epoch": 2.64, + "learning_rate": 6.0618107206976155e-06, + "loss": 0.1216, + "step": 68576 + }, + { + "epoch": 2.64, + "learning_rate": 6.061169530648885e-06, + "loss": 0.0636, + "step": 68577 + }, + { + "epoch": 2.64, + "learning_rate": 6.060528340600155e-06, + "loss": 0.1548, + "step": 68578 + }, + { + "epoch": 2.64, + "learning_rate": 6.059887150551424e-06, + "loss": 0.1169, + "step": 68579 + }, + { + "epoch": 2.64, + "learning_rate": 6.059245960502694e-06, + "loss": 0.2359, + "step": 68580 + }, + { + "epoch": 2.64, + "learning_rate": 6.058604770453963e-06, + "loss": 0.2791, + "step": 68581 + }, + { + "epoch": 2.64, + "learning_rate": 6.057963580405233e-06, + "loss": 0.1182, + "step": 68582 + }, + { + "epoch": 2.64, + "learning_rate": 6.057322390356502e-06, + "loss": 0.1689, + "step": 68583 + }, + { + "epoch": 2.64, + "learning_rate": 6.0566812003077715e-06, + "loss": 0.3686, + "step": 68584 + }, + { + "epoch": 2.64, + "learning_rate": 6.0560400102590415e-06, + "loss": 0.1929, + "step": 68585 + }, + { + "epoch": 2.64, + "learning_rate": 6.055398820210311e-06, + "loss": 0.1471, + "step": 68586 + }, + { + "epoch": 2.64, + "learning_rate": 6.054757630161581e-06, + "loss": 0.1878, + "step": 68587 + }, + { + "epoch": 2.64, + "learning_rate": 6.05411644011285e-06, + "loss": 0.3177, + "step": 68588 + }, + { + "epoch": 2.64, + "learning_rate": 6.05347525006412e-06, + "loss": 0.1876, + "step": 68589 + }, + { + "epoch": 2.64, + "learning_rate": 6.052834060015389e-06, + "loss": 0.2397, + "step": 68590 + }, + { + "epoch": 2.64, + "learning_rate": 6.052192869966659e-06, + "loss": 0.1344, + "step": 68591 + }, + { + "epoch": 2.64, + "learning_rate": 6.051551679917928e-06, + "loss": 0.2584, + "step": 68592 + }, + { + "epoch": 2.64, + "learning_rate": 6.0509104898691975e-06, + "loss": 0.2367, + "step": 68593 + }, + { + "epoch": 2.64, + "learning_rate": 6.0502692998204675e-06, + "loss": 0.2289, + "step": 68594 + }, + { + "epoch": 2.64, + "learning_rate": 6.049628109771737e-06, + "loss": 0.2615, + "step": 68595 + }, + { + "epoch": 2.64, + "learning_rate": 6.048986919723007e-06, + "loss": 0.3633, + "step": 68596 + }, + { + "epoch": 2.64, + "learning_rate": 6.048345729674276e-06, + "loss": 0.1615, + "step": 68597 + }, + { + "epoch": 2.64, + "learning_rate": 6.047704539625546e-06, + "loss": 0.1832, + "step": 68598 + }, + { + "epoch": 2.64, + "learning_rate": 6.047063349576815e-06, + "loss": 0.3012, + "step": 68599 + }, + { + "epoch": 2.64, + "learning_rate": 6.046422159528085e-06, + "loss": 0.2049, + "step": 68600 + }, + { + "epoch": 2.64, + "learning_rate": 6.045780969479354e-06, + "loss": 0.267, + "step": 68601 + }, + { + "epoch": 2.64, + "learning_rate": 6.045139779430624e-06, + "loss": 0.1918, + "step": 68602 + }, + { + "epoch": 2.64, + "learning_rate": 6.0444985893818935e-06, + "loss": 0.2364, + "step": 68603 + }, + { + "epoch": 2.64, + "learning_rate": 6.043857399333163e-06, + "loss": 0.325, + "step": 68604 + }, + { + "epoch": 2.64, + "learning_rate": 6.043216209284433e-06, + "loss": 0.1743, + "step": 68605 + }, + { + "epoch": 2.64, + "learning_rate": 6.042575019235702e-06, + "loss": 0.1842, + "step": 68606 + }, + { + "epoch": 2.64, + "learning_rate": 6.041933829186972e-06, + "loss": 0.2313, + "step": 68607 + }, + { + "epoch": 2.64, + "learning_rate": 6.041292639138241e-06, + "loss": 0.1903, + "step": 68608 + }, + { + "epoch": 2.64, + "learning_rate": 6.040651449089511e-06, + "loss": 0.213, + "step": 68609 + }, + { + "epoch": 2.64, + "learning_rate": 6.04001025904078e-06, + "loss": 0.2289, + "step": 68610 + }, + { + "epoch": 2.64, + "learning_rate": 6.03936906899205e-06, + "loss": 0.1681, + "step": 68611 + }, + { + "epoch": 2.64, + "learning_rate": 6.038727878943319e-06, + "loss": 0.0925, + "step": 68612 + }, + { + "epoch": 2.64, + "learning_rate": 6.038086688894589e-06, + "loss": 0.1492, + "step": 68613 + }, + { + "epoch": 2.64, + "learning_rate": 6.037445498845859e-06, + "loss": 0.1688, + "step": 68614 + }, + { + "epoch": 2.64, + "learning_rate": 6.036804308797128e-06, + "loss": 0.1904, + "step": 68615 + }, + { + "epoch": 2.64, + "learning_rate": 6.036163118748398e-06, + "loss": 0.1298, + "step": 68616 + }, + { + "epoch": 2.64, + "learning_rate": 6.035521928699667e-06, + "loss": 0.1753, + "step": 68617 + }, + { + "epoch": 2.64, + "learning_rate": 6.034880738650937e-06, + "loss": 0.1345, + "step": 68618 + }, + { + "epoch": 2.64, + "learning_rate": 6.034239548602206e-06, + "loss": 0.2084, + "step": 68619 + }, + { + "epoch": 2.64, + "learning_rate": 6.033598358553476e-06, + "loss": 0.1796, + "step": 68620 + }, + { + "epoch": 2.64, + "learning_rate": 6.032957168504745e-06, + "loss": 0.1775, + "step": 68621 + }, + { + "epoch": 2.64, + "learning_rate": 6.0323159784560146e-06, + "loss": 0.3765, + "step": 68622 + }, + { + "epoch": 2.64, + "learning_rate": 6.031674788407285e-06, + "loss": 0.3537, + "step": 68623 + }, + { + "epoch": 2.64, + "learning_rate": 6.031033598358554e-06, + "loss": 0.1747, + "step": 68624 + }, + { + "epoch": 2.64, + "learning_rate": 6.030392408309824e-06, + "loss": 0.2708, + "step": 68625 + }, + { + "epoch": 2.64, + "learning_rate": 6.029751218261093e-06, + "loss": 0.2909, + "step": 68626 + }, + { + "epoch": 2.64, + "learning_rate": 6.029110028212363e-06, + "loss": 0.3044, + "step": 68627 + }, + { + "epoch": 2.64, + "learning_rate": 6.028468838163632e-06, + "loss": 0.2849, + "step": 68628 + }, + { + "epoch": 2.64, + "learning_rate": 6.027827648114902e-06, + "loss": 0.1364, + "step": 68629 + }, + { + "epoch": 2.64, + "learning_rate": 6.027186458066171e-06, + "loss": 0.2087, + "step": 68630 + }, + { + "epoch": 2.64, + "learning_rate": 6.026545268017441e-06, + "loss": 0.1666, + "step": 68631 + }, + { + "epoch": 2.64, + "learning_rate": 6.0259040779687106e-06, + "loss": 0.2035, + "step": 68632 + }, + { + "epoch": 2.64, + "learning_rate": 6.02526288791998e-06, + "loss": 0.2486, + "step": 68633 + }, + { + "epoch": 2.64, + "learning_rate": 6.02462169787125e-06, + "loss": 0.1343, + "step": 68634 + }, + { + "epoch": 2.64, + "learning_rate": 6.023980507822519e-06, + "loss": 0.4237, + "step": 68635 + }, + { + "epoch": 2.64, + "learning_rate": 6.023339317773789e-06, + "loss": 0.3685, + "step": 68636 + }, + { + "epoch": 2.64, + "learning_rate": 6.022698127725058e-06, + "loss": 0.198, + "step": 68637 + }, + { + "epoch": 2.64, + "learning_rate": 6.022056937676328e-06, + "loss": 0.339, + "step": 68638 + }, + { + "epoch": 2.64, + "learning_rate": 6.021415747627597e-06, + "loss": 0.2829, + "step": 68639 + }, + { + "epoch": 2.64, + "learning_rate": 6.020774557578867e-06, + "loss": 0.1317, + "step": 68640 + }, + { + "epoch": 2.64, + "learning_rate": 6.0201333675301365e-06, + "loss": 0.159, + "step": 68641 + }, + { + "epoch": 2.64, + "learning_rate": 6.019492177481406e-06, + "loss": 0.1683, + "step": 68642 + }, + { + "epoch": 2.64, + "learning_rate": 6.018850987432676e-06, + "loss": 0.2765, + "step": 68643 + }, + { + "epoch": 2.64, + "learning_rate": 6.018209797383945e-06, + "loss": 0.2021, + "step": 68644 + }, + { + "epoch": 2.64, + "learning_rate": 6.017568607335215e-06, + "loss": 0.3675, + "step": 68645 + }, + { + "epoch": 2.64, + "learning_rate": 6.016927417286484e-06, + "loss": 0.2679, + "step": 68646 + }, + { + "epoch": 2.64, + "learning_rate": 6.016286227237754e-06, + "loss": 0.1672, + "step": 68647 + }, + { + "epoch": 2.64, + "learning_rate": 6.015645037189023e-06, + "loss": 0.1639, + "step": 68648 + }, + { + "epoch": 2.64, + "learning_rate": 6.015003847140293e-06, + "loss": 0.1147, + "step": 68649 + }, + { + "epoch": 2.64, + "learning_rate": 6.0143626570915625e-06, + "loss": 0.1497, + "step": 68650 + }, + { + "epoch": 2.64, + "learning_rate": 6.0137214670428325e-06, + "loss": 0.1415, + "step": 68651 + }, + { + "epoch": 2.64, + "learning_rate": 6.013080276994102e-06, + "loss": 0.136, + "step": 68652 + }, + { + "epoch": 2.64, + "learning_rate": 6.012439086945371e-06, + "loss": 0.411, + "step": 68653 + }, + { + "epoch": 2.64, + "learning_rate": 6.011797896896641e-06, + "loss": 0.1911, + "step": 68654 + }, + { + "epoch": 2.64, + "learning_rate": 6.01115670684791e-06, + "loss": 0.3438, + "step": 68655 + }, + { + "epoch": 2.64, + "learning_rate": 6.01051551679918e-06, + "loss": 0.2505, + "step": 68656 + }, + { + "epoch": 2.64, + "learning_rate": 6.009874326750449e-06, + "loss": 0.1053, + "step": 68657 + }, + { + "epoch": 2.64, + "learning_rate": 6.009233136701719e-06, + "loss": 0.1745, + "step": 68658 + }, + { + "epoch": 2.64, + "learning_rate": 6.0085919466529885e-06, + "loss": 0.1345, + "step": 68659 + }, + { + "epoch": 2.64, + "learning_rate": 6.0079507566042585e-06, + "loss": 0.2135, + "step": 68660 + }, + { + "epoch": 2.64, + "learning_rate": 6.007309566555528e-06, + "loss": 0.1502, + "step": 68661 + }, + { + "epoch": 2.64, + "learning_rate": 6.006668376506797e-06, + "loss": 0.2525, + "step": 68662 + }, + { + "epoch": 2.64, + "learning_rate": 6.006027186458067e-06, + "loss": 0.1591, + "step": 68663 + }, + { + "epoch": 2.64, + "learning_rate": 6.005385996409336e-06, + "loss": 0.3112, + "step": 68664 + }, + { + "epoch": 2.64, + "learning_rate": 6.004744806360606e-06, + "loss": 0.2925, + "step": 68665 + }, + { + "epoch": 2.64, + "learning_rate": 6.004103616311875e-06, + "loss": 0.1478, + "step": 68666 + }, + { + "epoch": 2.64, + "learning_rate": 6.003462426263145e-06, + "loss": 0.217, + "step": 68667 + }, + { + "epoch": 2.64, + "learning_rate": 6.0028212362144144e-06, + "loss": 0.2912, + "step": 68668 + }, + { + "epoch": 2.64, + "learning_rate": 6.0021800461656844e-06, + "loss": 0.1722, + "step": 68669 + }, + { + "epoch": 2.64, + "learning_rate": 6.001538856116954e-06, + "loss": 0.2497, + "step": 68670 + }, + { + "epoch": 2.64, + "learning_rate": 6.000897666068223e-06, + "loss": 0.249, + "step": 68671 + }, + { + "epoch": 2.64, + "learning_rate": 6.000256476019493e-06, + "loss": 0.1844, + "step": 68672 + }, + { + "epoch": 2.64, + "learning_rate": 5.999615285970762e-06, + "loss": 0.23, + "step": 68673 + }, + { + "epoch": 2.64, + "learning_rate": 5.998974095922032e-06, + "loss": 0.1773, + "step": 68674 + }, + { + "epoch": 2.64, + "learning_rate": 5.998332905873301e-06, + "loss": 0.1194, + "step": 68675 + }, + { + "epoch": 2.64, + "learning_rate": 5.997691715824571e-06, + "loss": 0.3574, + "step": 68676 + }, + { + "epoch": 2.64, + "learning_rate": 5.99705052577584e-06, + "loss": 0.284, + "step": 68677 + }, + { + "epoch": 2.64, + "learning_rate": 5.99640933572711e-06, + "loss": 0.1237, + "step": 68678 + }, + { + "epoch": 2.64, + "learning_rate": 5.99576814567838e-06, + "loss": 0.1511, + "step": 68679 + }, + { + "epoch": 2.64, + "learning_rate": 5.99512695562965e-06, + "loss": 0.3049, + "step": 68680 + }, + { + "epoch": 2.64, + "learning_rate": 5.994485765580919e-06, + "loss": 0.1144, + "step": 68681 + }, + { + "epoch": 2.64, + "learning_rate": 5.993844575532188e-06, + "loss": 0.0887, + "step": 68682 + }, + { + "epoch": 2.64, + "learning_rate": 5.993203385483458e-06, + "loss": 0.2625, + "step": 68683 + }, + { + "epoch": 2.64, + "learning_rate": 5.992562195434727e-06, + "loss": 0.1614, + "step": 68684 + }, + { + "epoch": 2.64, + "learning_rate": 5.991921005385997e-06, + "loss": 0.2712, + "step": 68685 + }, + { + "epoch": 2.64, + "learning_rate": 5.991279815337266e-06, + "loss": 0.2603, + "step": 68686 + }, + { + "epoch": 2.64, + "learning_rate": 5.990638625288536e-06, + "loss": 0.3739, + "step": 68687 + }, + { + "epoch": 2.64, + "learning_rate": 5.9899974352398056e-06, + "loss": 0.2353, + "step": 68688 + }, + { + "epoch": 2.64, + "learning_rate": 5.989356245191076e-06, + "loss": 0.2044, + "step": 68689 + }, + { + "epoch": 2.64, + "learning_rate": 5.988715055142345e-06, + "loss": 0.1704, + "step": 68690 + }, + { + "epoch": 2.64, + "learning_rate": 5.988073865093614e-06, + "loss": 0.147, + "step": 68691 + }, + { + "epoch": 2.64, + "learning_rate": 5.987432675044884e-06, + "loss": 0.2368, + "step": 68692 + }, + { + "epoch": 2.64, + "learning_rate": 5.986791484996153e-06, + "loss": 0.1577, + "step": 68693 + }, + { + "epoch": 2.64, + "learning_rate": 5.986150294947423e-06, + "loss": 0.3322, + "step": 68694 + }, + { + "epoch": 2.64, + "learning_rate": 5.985509104898692e-06, + "loss": 0.2641, + "step": 68695 + }, + { + "epoch": 2.64, + "learning_rate": 5.984867914849962e-06, + "loss": 0.1671, + "step": 68696 + }, + { + "epoch": 2.64, + "learning_rate": 5.9842267248012315e-06, + "loss": 0.1508, + "step": 68697 + }, + { + "epoch": 2.64, + "learning_rate": 5.9835855347525015e-06, + "loss": 0.1033, + "step": 68698 + }, + { + "epoch": 2.64, + "learning_rate": 5.982944344703771e-06, + "loss": 0.1407, + "step": 68699 + }, + { + "epoch": 2.64, + "learning_rate": 5.982303154655041e-06, + "loss": 0.1829, + "step": 68700 + }, + { + "epoch": 2.64, + "learning_rate": 5.98166196460631e-06, + "loss": 0.3168, + "step": 68701 + }, + { + "epoch": 2.64, + "learning_rate": 5.981020774557579e-06, + "loss": 0.1715, + "step": 68702 + }, + { + "epoch": 2.64, + "learning_rate": 5.980379584508849e-06, + "loss": 0.1893, + "step": 68703 + }, + { + "epoch": 2.64, + "learning_rate": 5.979738394460118e-06, + "loss": 0.2245, + "step": 68704 + }, + { + "epoch": 2.64, + "learning_rate": 5.979097204411388e-06, + "loss": 0.2505, + "step": 68705 + }, + { + "epoch": 2.64, + "learning_rate": 5.9784560143626575e-06, + "loss": 0.176, + "step": 68706 + }, + { + "epoch": 2.64, + "learning_rate": 5.9778148243139275e-06, + "loss": 0.2256, + "step": 68707 + }, + { + "epoch": 2.64, + "learning_rate": 5.977173634265197e-06, + "loss": 0.1319, + "step": 68708 + }, + { + "epoch": 2.64, + "learning_rate": 5.976532444216467e-06, + "loss": 0.2519, + "step": 68709 + }, + { + "epoch": 2.64, + "learning_rate": 5.975891254167736e-06, + "loss": 0.4979, + "step": 68710 + }, + { + "epoch": 2.64, + "learning_rate": 5.975250064119005e-06, + "loss": 0.211, + "step": 68711 + }, + { + "epoch": 2.64, + "learning_rate": 5.974608874070275e-06, + "loss": 0.3542, + "step": 68712 + }, + { + "epoch": 2.64, + "learning_rate": 5.973967684021544e-06, + "loss": 0.1306, + "step": 68713 + }, + { + "epoch": 2.64, + "learning_rate": 5.973326493972814e-06, + "loss": 0.288, + "step": 68714 + }, + { + "epoch": 2.64, + "learning_rate": 5.9726853039240835e-06, + "loss": 0.2377, + "step": 68715 + }, + { + "epoch": 2.64, + "learning_rate": 5.9720441138753535e-06, + "loss": 0.1554, + "step": 68716 + }, + { + "epoch": 2.64, + "learning_rate": 5.971402923826623e-06, + "loss": 0.2942, + "step": 68717 + }, + { + "epoch": 2.64, + "learning_rate": 5.970761733777893e-06, + "loss": 0.409, + "step": 68718 + }, + { + "epoch": 2.64, + "learning_rate": 5.970120543729162e-06, + "loss": 0.2614, + "step": 68719 + }, + { + "epoch": 2.64, + "learning_rate": 5.969479353680431e-06, + "loss": 0.1864, + "step": 68720 + }, + { + "epoch": 2.64, + "learning_rate": 5.968838163631701e-06, + "loss": 0.2823, + "step": 68721 + }, + { + "epoch": 2.64, + "learning_rate": 5.96819697358297e-06, + "loss": 0.3833, + "step": 68722 + }, + { + "epoch": 2.64, + "learning_rate": 5.96755578353424e-06, + "loss": 0.1249, + "step": 68723 + }, + { + "epoch": 2.64, + "learning_rate": 5.9669145934855094e-06, + "loss": 0.2552, + "step": 68724 + }, + { + "epoch": 2.64, + "learning_rate": 5.9662734034367795e-06, + "loss": 0.1754, + "step": 68725 + }, + { + "epoch": 2.64, + "learning_rate": 5.965632213388049e-06, + "loss": 0.18, + "step": 68726 + }, + { + "epoch": 2.64, + "learning_rate": 5.964991023339319e-06, + "loss": 0.2141, + "step": 68727 + }, + { + "epoch": 2.64, + "learning_rate": 5.964349833290588e-06, + "loss": 0.3189, + "step": 68728 + }, + { + "epoch": 2.64, + "learning_rate": 5.963708643241858e-06, + "loss": 0.1126, + "step": 68729 + }, + { + "epoch": 2.64, + "learning_rate": 5.963067453193127e-06, + "loss": 0.337, + "step": 68730 + }, + { + "epoch": 2.64, + "learning_rate": 5.962426263144396e-06, + "loss": 0.4366, + "step": 68731 + }, + { + "epoch": 2.64, + "learning_rate": 5.961785073095666e-06, + "loss": 0.2512, + "step": 68732 + }, + { + "epoch": 2.64, + "learning_rate": 5.961143883046935e-06, + "loss": 0.1394, + "step": 68733 + }, + { + "epoch": 2.64, + "learning_rate": 5.960502692998205e-06, + "loss": 0.2092, + "step": 68734 + }, + { + "epoch": 2.64, + "learning_rate": 5.959861502949475e-06, + "loss": 0.228, + "step": 68735 + }, + { + "epoch": 2.64, + "learning_rate": 5.959220312900745e-06, + "loss": 0.1046, + "step": 68736 + }, + { + "epoch": 2.64, + "learning_rate": 5.958579122852014e-06, + "loss": 0.226, + "step": 68737 + }, + { + "epoch": 2.64, + "learning_rate": 5.957937932803284e-06, + "loss": 0.2828, + "step": 68738 + }, + { + "epoch": 2.64, + "learning_rate": 5.957296742754553e-06, + "loss": 0.1442, + "step": 68739 + }, + { + "epoch": 2.64, + "learning_rate": 5.956655552705822e-06, + "loss": 0.2739, + "step": 68740 + }, + { + "epoch": 2.64, + "learning_rate": 5.956014362657092e-06, + "loss": 0.2335, + "step": 68741 + }, + { + "epoch": 2.64, + "learning_rate": 5.955373172608361e-06, + "loss": 0.1716, + "step": 68742 + }, + { + "epoch": 2.64, + "learning_rate": 5.954731982559631e-06, + "loss": 0.2404, + "step": 68743 + }, + { + "epoch": 2.64, + "learning_rate": 5.9540907925109006e-06, + "loss": 0.2165, + "step": 68744 + }, + { + "epoch": 2.64, + "learning_rate": 5.953449602462171e-06, + "loss": 0.3059, + "step": 68745 + }, + { + "epoch": 2.64, + "learning_rate": 5.95280841241344e-06, + "loss": 0.2038, + "step": 68746 + }, + { + "epoch": 2.64, + "learning_rate": 5.95216722236471e-06, + "loss": 0.2616, + "step": 68747 + }, + { + "epoch": 2.64, + "learning_rate": 5.951526032315979e-06, + "loss": 0.2904, + "step": 68748 + }, + { + "epoch": 2.64, + "learning_rate": 5.950884842267249e-06, + "loss": 0.2441, + "step": 68749 + }, + { + "epoch": 2.64, + "learning_rate": 5.950243652218518e-06, + "loss": 0.2347, + "step": 68750 + }, + { + "epoch": 2.64, + "learning_rate": 5.949602462169787e-06, + "loss": 0.16, + "step": 68751 + }, + { + "epoch": 2.64, + "learning_rate": 5.948961272121057e-06, + "loss": 0.1803, + "step": 68752 + }, + { + "epoch": 2.64, + "learning_rate": 5.9483200820723265e-06, + "loss": 0.2379, + "step": 68753 + }, + { + "epoch": 2.64, + "learning_rate": 5.9476788920235966e-06, + "loss": 0.2312, + "step": 68754 + }, + { + "epoch": 2.64, + "learning_rate": 5.947037701974866e-06, + "loss": 0.2825, + "step": 68755 + }, + { + "epoch": 2.64, + "learning_rate": 5.946396511926136e-06, + "loss": 0.2097, + "step": 68756 + }, + { + "epoch": 2.64, + "learning_rate": 5.945755321877405e-06, + "loss": 0.1165, + "step": 68757 + }, + { + "epoch": 2.64, + "learning_rate": 5.945114131828675e-06, + "loss": 0.1872, + "step": 68758 + }, + { + "epoch": 2.64, + "learning_rate": 5.944472941779944e-06, + "loss": 0.2901, + "step": 68759 + }, + { + "epoch": 2.64, + "learning_rate": 5.943831751731213e-06, + "loss": 0.1374, + "step": 68760 + }, + { + "epoch": 2.64, + "learning_rate": 5.943190561682483e-06, + "loss": 0.1341, + "step": 68761 + }, + { + "epoch": 2.64, + "learning_rate": 5.9425493716337525e-06, + "loss": 0.1098, + "step": 68762 + }, + { + "epoch": 2.64, + "learning_rate": 5.9419081815850225e-06, + "loss": 0.161, + "step": 68763 + }, + { + "epoch": 2.64, + "learning_rate": 5.941266991536292e-06, + "loss": 0.2731, + "step": 68764 + }, + { + "epoch": 2.64, + "learning_rate": 5.940625801487562e-06, + "loss": 0.2138, + "step": 68765 + }, + { + "epoch": 2.64, + "learning_rate": 5.939984611438831e-06, + "loss": 0.2554, + "step": 68766 + }, + { + "epoch": 2.64, + "learning_rate": 5.939343421390101e-06, + "loss": 0.2134, + "step": 68767 + }, + { + "epoch": 2.64, + "learning_rate": 5.93870223134137e-06, + "loss": 0.3143, + "step": 68768 + }, + { + "epoch": 2.64, + "learning_rate": 5.93806104129264e-06, + "loss": 0.2004, + "step": 68769 + }, + { + "epoch": 2.64, + "learning_rate": 5.937419851243909e-06, + "loss": 0.2719, + "step": 68770 + }, + { + "epoch": 2.64, + "learning_rate": 5.9367786611951785e-06, + "loss": 0.2397, + "step": 68771 + }, + { + "epoch": 2.64, + "learning_rate": 5.9361374711464485e-06, + "loss": 0.1883, + "step": 68772 + }, + { + "epoch": 2.64, + "learning_rate": 5.935496281097718e-06, + "loss": 0.2265, + "step": 68773 + }, + { + "epoch": 2.64, + "learning_rate": 5.934855091048988e-06, + "loss": 0.139, + "step": 68774 + }, + { + "epoch": 2.64, + "learning_rate": 5.934213901000257e-06, + "loss": 0.2016, + "step": 68775 + }, + { + "epoch": 2.64, + "learning_rate": 5.933572710951527e-06, + "loss": 0.1833, + "step": 68776 + }, + { + "epoch": 2.64, + "learning_rate": 5.932931520902796e-06, + "loss": 0.1177, + "step": 68777 + }, + { + "epoch": 2.64, + "learning_rate": 5.932290330854066e-06, + "loss": 0.3065, + "step": 68778 + }, + { + "epoch": 2.64, + "learning_rate": 5.931649140805335e-06, + "loss": 0.229, + "step": 68779 + }, + { + "epoch": 2.64, + "learning_rate": 5.9310079507566044e-06, + "loss": 0.2047, + "step": 68780 + }, + { + "epoch": 2.64, + "learning_rate": 5.9303667607078745e-06, + "loss": 0.1224, + "step": 68781 + }, + { + "epoch": 2.64, + "learning_rate": 5.929725570659144e-06, + "loss": 0.2611, + "step": 68782 + }, + { + "epoch": 2.64, + "learning_rate": 5.929084380610414e-06, + "loss": 0.1755, + "step": 68783 + }, + { + "epoch": 2.64, + "learning_rate": 5.928443190561683e-06, + "loss": 0.2352, + "step": 68784 + }, + { + "epoch": 2.64, + "learning_rate": 5.927802000512953e-06, + "loss": 0.1452, + "step": 68785 + }, + { + "epoch": 2.64, + "learning_rate": 5.927160810464222e-06, + "loss": 0.1578, + "step": 68786 + }, + { + "epoch": 2.64, + "learning_rate": 5.926519620415492e-06, + "loss": 0.22, + "step": 68787 + }, + { + "epoch": 2.64, + "learning_rate": 5.925878430366761e-06, + "loss": 0.2123, + "step": 68788 + }, + { + "epoch": 2.64, + "learning_rate": 5.92523724031803e-06, + "loss": 0.157, + "step": 68789 + }, + { + "epoch": 2.64, + "learning_rate": 5.9245960502693004e-06, + "loss": 0.1755, + "step": 68790 + }, + { + "epoch": 2.64, + "learning_rate": 5.92395486022057e-06, + "loss": 0.1645, + "step": 68791 + }, + { + "epoch": 2.64, + "learning_rate": 5.92331367017184e-06, + "loss": 0.2019, + "step": 68792 + }, + { + "epoch": 2.64, + "learning_rate": 5.922672480123109e-06, + "loss": 0.1896, + "step": 68793 + }, + { + "epoch": 2.64, + "learning_rate": 5.922031290074379e-06, + "loss": 0.1913, + "step": 68794 + }, + { + "epoch": 2.64, + "learning_rate": 5.921390100025648e-06, + "loss": 0.1673, + "step": 68795 + }, + { + "epoch": 2.64, + "learning_rate": 5.920748909976918e-06, + "loss": 0.154, + "step": 68796 + }, + { + "epoch": 2.65, + "learning_rate": 5.920107719928187e-06, + "loss": 0.1843, + "step": 68797 + }, + { + "epoch": 2.65, + "learning_rate": 5.919466529879457e-06, + "loss": 0.2265, + "step": 68798 + }, + { + "epoch": 2.65, + "learning_rate": 5.918825339830726e-06, + "loss": 0.4188, + "step": 68799 + }, + { + "epoch": 2.65, + "learning_rate": 5.9181841497819956e-06, + "loss": 0.2156, + "step": 68800 + }, + { + "epoch": 2.65, + "learning_rate": 5.917542959733266e-06, + "loss": 0.1755, + "step": 68801 + }, + { + "epoch": 2.65, + "learning_rate": 5.916901769684535e-06, + "loss": 0.2681, + "step": 68802 + }, + { + "epoch": 2.65, + "learning_rate": 5.916260579635805e-06, + "loss": 0.1374, + "step": 68803 + }, + { + "epoch": 2.65, + "learning_rate": 5.915619389587074e-06, + "loss": 0.1591, + "step": 68804 + }, + { + "epoch": 2.65, + "learning_rate": 5.914978199538344e-06, + "loss": 0.2179, + "step": 68805 + }, + { + "epoch": 2.65, + "learning_rate": 5.914337009489613e-06, + "loss": 0.1461, + "step": 68806 + }, + { + "epoch": 2.65, + "learning_rate": 5.913695819440883e-06, + "loss": 0.1663, + "step": 68807 + }, + { + "epoch": 2.65, + "learning_rate": 5.913054629392152e-06, + "loss": 0.2388, + "step": 68808 + }, + { + "epoch": 2.65, + "learning_rate": 5.9124134393434215e-06, + "loss": 0.159, + "step": 68809 + }, + { + "epoch": 2.65, + "learning_rate": 5.9117722492946916e-06, + "loss": 0.0859, + "step": 68810 + }, + { + "epoch": 2.65, + "learning_rate": 5.911131059245961e-06, + "loss": 0.1371, + "step": 68811 + }, + { + "epoch": 2.65, + "learning_rate": 5.910489869197231e-06, + "loss": 0.1267, + "step": 68812 + }, + { + "epoch": 2.65, + "learning_rate": 5.9098486791485e-06, + "loss": 0.3021, + "step": 68813 + }, + { + "epoch": 2.65, + "learning_rate": 5.90920748909977e-06, + "loss": 0.2116, + "step": 68814 + }, + { + "epoch": 2.65, + "learning_rate": 5.908566299051039e-06, + "loss": 0.0973, + "step": 68815 + }, + { + "epoch": 2.65, + "learning_rate": 5.907925109002309e-06, + "loss": 0.145, + "step": 68816 + }, + { + "epoch": 2.65, + "learning_rate": 5.907283918953578e-06, + "loss": 0.1516, + "step": 68817 + }, + { + "epoch": 2.65, + "learning_rate": 5.906642728904848e-06, + "loss": 0.3569, + "step": 68818 + }, + { + "epoch": 2.65, + "learning_rate": 5.9060015388561175e-06, + "loss": 0.3046, + "step": 68819 + }, + { + "epoch": 2.65, + "learning_rate": 5.905360348807387e-06, + "loss": 0.2733, + "step": 68820 + }, + { + "epoch": 2.65, + "learning_rate": 5.904719158758657e-06, + "loss": 0.2191, + "step": 68821 + }, + { + "epoch": 2.65, + "learning_rate": 5.904077968709926e-06, + "loss": 0.1141, + "step": 68822 + }, + { + "epoch": 2.65, + "learning_rate": 5.903436778661196e-06, + "loss": 0.2166, + "step": 68823 + }, + { + "epoch": 2.65, + "learning_rate": 5.902795588612465e-06, + "loss": 0.2063, + "step": 68824 + }, + { + "epoch": 2.65, + "learning_rate": 5.902154398563735e-06, + "loss": 0.1625, + "step": 68825 + }, + { + "epoch": 2.65, + "learning_rate": 5.901513208515004e-06, + "loss": 0.2792, + "step": 68826 + }, + { + "epoch": 2.65, + "learning_rate": 5.900872018466274e-06, + "loss": 0.1119, + "step": 68827 + }, + { + "epoch": 2.65, + "learning_rate": 5.9002308284175435e-06, + "loss": 0.2623, + "step": 68828 + }, + { + "epoch": 2.65, + "learning_rate": 5.899589638368813e-06, + "loss": 0.2655, + "step": 68829 + }, + { + "epoch": 2.65, + "learning_rate": 5.898948448320083e-06, + "loss": 0.1255, + "step": 68830 + }, + { + "epoch": 2.65, + "learning_rate": 5.898307258271352e-06, + "loss": 0.0864, + "step": 68831 + }, + { + "epoch": 2.65, + "learning_rate": 5.897666068222622e-06, + "loss": 0.173, + "step": 68832 + }, + { + "epoch": 2.65, + "learning_rate": 5.897024878173891e-06, + "loss": 0.1848, + "step": 68833 + }, + { + "epoch": 2.65, + "learning_rate": 5.896383688125161e-06, + "loss": 0.0754, + "step": 68834 + }, + { + "epoch": 2.65, + "learning_rate": 5.89574249807643e-06, + "loss": 0.242, + "step": 68835 + }, + { + "epoch": 2.65, + "learning_rate": 5.8951013080277e-06, + "loss": 0.3707, + "step": 68836 + }, + { + "epoch": 2.65, + "learning_rate": 5.8944601179789695e-06, + "loss": 0.258, + "step": 68837 + }, + { + "epoch": 2.65, + "learning_rate": 5.893818927930239e-06, + "loss": 0.2107, + "step": 68838 + }, + { + "epoch": 2.65, + "learning_rate": 5.893177737881509e-06, + "loss": 0.1884, + "step": 68839 + }, + { + "epoch": 2.65, + "learning_rate": 5.892536547832778e-06, + "loss": 0.0706, + "step": 68840 + }, + { + "epoch": 2.65, + "learning_rate": 5.891895357784048e-06, + "loss": 0.1446, + "step": 68841 + }, + { + "epoch": 2.65, + "learning_rate": 5.891254167735317e-06, + "loss": 0.1714, + "step": 68842 + }, + { + "epoch": 2.65, + "learning_rate": 5.890612977686587e-06, + "loss": 0.1969, + "step": 68843 + }, + { + "epoch": 2.65, + "learning_rate": 5.889971787637856e-06, + "loss": 0.2208, + "step": 68844 + }, + { + "epoch": 2.65, + "learning_rate": 5.889330597589126e-06, + "loss": 0.2217, + "step": 68845 + }, + { + "epoch": 2.65, + "learning_rate": 5.8886894075403954e-06, + "loss": 0.1655, + "step": 68846 + }, + { + "epoch": 2.65, + "learning_rate": 5.8880482174916655e-06, + "loss": 0.1864, + "step": 68847 + }, + { + "epoch": 2.65, + "learning_rate": 5.887407027442935e-06, + "loss": 0.0813, + "step": 68848 + }, + { + "epoch": 2.65, + "learning_rate": 5.886765837394204e-06, + "loss": 0.2687, + "step": 68849 + }, + { + "epoch": 2.65, + "learning_rate": 5.886124647345474e-06, + "loss": 0.2061, + "step": 68850 + }, + { + "epoch": 2.65, + "learning_rate": 5.885483457296743e-06, + "loss": 0.2283, + "step": 68851 + }, + { + "epoch": 2.65, + "learning_rate": 5.884842267248013e-06, + "loss": 0.2068, + "step": 68852 + }, + { + "epoch": 2.65, + "learning_rate": 5.884201077199282e-06, + "loss": 0.2319, + "step": 68853 + }, + { + "epoch": 2.65, + "learning_rate": 5.883559887150552e-06, + "loss": 0.1675, + "step": 68854 + }, + { + "epoch": 2.65, + "learning_rate": 5.882918697101821e-06, + "loss": 0.2006, + "step": 68855 + }, + { + "epoch": 2.65, + "learning_rate": 5.882277507053091e-06, + "loss": 0.3281, + "step": 68856 + }, + { + "epoch": 2.65, + "learning_rate": 5.881636317004361e-06, + "loss": 0.2324, + "step": 68857 + }, + { + "epoch": 2.65, + "learning_rate": 5.88099512695563e-06, + "loss": 0.2401, + "step": 68858 + }, + { + "epoch": 2.65, + "learning_rate": 5.8803539369069e-06, + "loss": 0.369, + "step": 68859 + }, + { + "epoch": 2.65, + "learning_rate": 5.879712746858169e-06, + "loss": 0.1792, + "step": 68860 + }, + { + "epoch": 2.65, + "learning_rate": 5.879071556809439e-06, + "loss": 0.1498, + "step": 68861 + }, + { + "epoch": 2.65, + "learning_rate": 5.878430366760708e-06, + "loss": 0.2313, + "step": 68862 + }, + { + "epoch": 2.65, + "learning_rate": 5.877789176711978e-06, + "loss": 0.2129, + "step": 68863 + }, + { + "epoch": 2.65, + "learning_rate": 5.877147986663247e-06, + "loss": 0.1274, + "step": 68864 + }, + { + "epoch": 2.65, + "learning_rate": 5.876506796614517e-06, + "loss": 0.1339, + "step": 68865 + }, + { + "epoch": 2.65, + "learning_rate": 5.8758656065657866e-06, + "loss": 0.2318, + "step": 68866 + }, + { + "epoch": 2.65, + "learning_rate": 5.875224416517057e-06, + "loss": 0.3209, + "step": 68867 + }, + { + "epoch": 2.65, + "learning_rate": 5.874583226468326e-06, + "loss": 0.2119, + "step": 68868 + }, + { + "epoch": 2.65, + "learning_rate": 5.873942036419595e-06, + "loss": 0.1429, + "step": 68869 + }, + { + "epoch": 2.65, + "learning_rate": 5.873300846370865e-06, + "loss": 0.2189, + "step": 68870 + }, + { + "epoch": 2.65, + "learning_rate": 5.872659656322134e-06, + "loss": 0.2344, + "step": 68871 + }, + { + "epoch": 2.65, + "learning_rate": 5.872018466273404e-06, + "loss": 0.131, + "step": 68872 + }, + { + "epoch": 2.65, + "learning_rate": 5.871377276224673e-06, + "loss": 0.2068, + "step": 68873 + }, + { + "epoch": 2.65, + "learning_rate": 5.870736086175943e-06, + "loss": 0.1992, + "step": 68874 + }, + { + "epoch": 2.65, + "learning_rate": 5.8700948961272125e-06, + "loss": 0.1724, + "step": 68875 + }, + { + "epoch": 2.65, + "learning_rate": 5.8694537060784826e-06, + "loss": 0.1361, + "step": 68876 + }, + { + "epoch": 2.65, + "learning_rate": 5.868812516029752e-06, + "loss": 0.2037, + "step": 68877 + }, + { + "epoch": 2.65, + "learning_rate": 5.868171325981021e-06, + "loss": 0.1907, + "step": 68878 + }, + { + "epoch": 2.65, + "learning_rate": 5.867530135932291e-06, + "loss": 0.2711, + "step": 68879 + }, + { + "epoch": 2.65, + "learning_rate": 5.86688894588356e-06, + "loss": 0.2799, + "step": 68880 + }, + { + "epoch": 2.65, + "learning_rate": 5.86624775583483e-06, + "loss": 0.3307, + "step": 68881 + }, + { + "epoch": 2.65, + "learning_rate": 5.865606565786099e-06, + "loss": 0.1715, + "step": 68882 + }, + { + "epoch": 2.65, + "learning_rate": 5.864965375737369e-06, + "loss": 0.2196, + "step": 68883 + }, + { + "epoch": 2.65, + "learning_rate": 5.8643241856886385e-06, + "loss": 0.271, + "step": 68884 + }, + { + "epoch": 2.65, + "learning_rate": 5.8636829956399085e-06, + "loss": 0.1662, + "step": 68885 + }, + { + "epoch": 2.65, + "learning_rate": 5.863041805591178e-06, + "loss": 0.2095, + "step": 68886 + }, + { + "epoch": 2.65, + "learning_rate": 5.862400615542448e-06, + "loss": 0.2211, + "step": 68887 + }, + { + "epoch": 2.65, + "learning_rate": 5.861759425493717e-06, + "loss": 0.2374, + "step": 68888 + }, + { + "epoch": 2.65, + "learning_rate": 5.861118235444986e-06, + "loss": 0.1616, + "step": 68889 + }, + { + "epoch": 2.65, + "learning_rate": 5.860477045396256e-06, + "loss": 0.3757, + "step": 68890 + }, + { + "epoch": 2.65, + "learning_rate": 5.859835855347525e-06, + "loss": 0.1949, + "step": 68891 + }, + { + "epoch": 2.65, + "learning_rate": 5.859194665298795e-06, + "loss": 0.2276, + "step": 68892 + }, + { + "epoch": 2.65, + "learning_rate": 5.8585534752500645e-06, + "loss": 0.196, + "step": 68893 + }, + { + "epoch": 2.65, + "learning_rate": 5.8579122852013345e-06, + "loss": 0.1958, + "step": 68894 + }, + { + "epoch": 2.65, + "learning_rate": 5.857271095152604e-06, + "loss": 0.1547, + "step": 68895 + }, + { + "epoch": 2.65, + "learning_rate": 5.856629905103874e-06, + "loss": 0.3206, + "step": 68896 + }, + { + "epoch": 2.65, + "learning_rate": 5.855988715055143e-06, + "loss": 0.26, + "step": 68897 + }, + { + "epoch": 2.65, + "learning_rate": 5.855347525006412e-06, + "loss": 0.149, + "step": 68898 + }, + { + "epoch": 2.65, + "learning_rate": 5.854706334957682e-06, + "loss": 0.2517, + "step": 68899 + }, + { + "epoch": 2.65, + "learning_rate": 5.854065144908951e-06, + "loss": 0.2586, + "step": 68900 + }, + { + "epoch": 2.65, + "learning_rate": 5.853423954860221e-06, + "loss": 0.2492, + "step": 68901 + }, + { + "epoch": 2.65, + "learning_rate": 5.8527827648114904e-06, + "loss": 0.1951, + "step": 68902 + }, + { + "epoch": 2.65, + "learning_rate": 5.8521415747627605e-06, + "loss": 0.1528, + "step": 68903 + }, + { + "epoch": 2.65, + "learning_rate": 5.85150038471403e-06, + "loss": 0.1733, + "step": 68904 + }, + { + "epoch": 2.65, + "learning_rate": 5.8508591946653e-06, + "loss": 0.238, + "step": 68905 + }, + { + "epoch": 2.65, + "learning_rate": 5.850218004616569e-06, + "loss": 0.168, + "step": 68906 + }, + { + "epoch": 2.65, + "learning_rate": 5.849576814567838e-06, + "loss": 0.2286, + "step": 68907 + }, + { + "epoch": 2.65, + "learning_rate": 5.848935624519108e-06, + "loss": 0.2074, + "step": 68908 + }, + { + "epoch": 2.65, + "learning_rate": 5.848294434470377e-06, + "loss": 0.2441, + "step": 68909 + }, + { + "epoch": 2.65, + "learning_rate": 5.847653244421647e-06, + "loss": 0.3138, + "step": 68910 + }, + { + "epoch": 2.65, + "learning_rate": 5.847012054372916e-06, + "loss": 0.2309, + "step": 68911 + }, + { + "epoch": 2.65, + "learning_rate": 5.8463708643241864e-06, + "loss": 0.3781, + "step": 68912 + }, + { + "epoch": 2.65, + "learning_rate": 5.845729674275456e-06, + "loss": 0.1463, + "step": 68913 + }, + { + "epoch": 2.65, + "learning_rate": 5.845088484226726e-06, + "loss": 0.3269, + "step": 68914 + }, + { + "epoch": 2.65, + "learning_rate": 5.844447294177995e-06, + "loss": 0.1529, + "step": 68915 + }, + { + "epoch": 2.65, + "learning_rate": 5.843806104129265e-06, + "loss": 0.1869, + "step": 68916 + }, + { + "epoch": 2.65, + "learning_rate": 5.843164914080534e-06, + "loss": 0.3438, + "step": 68917 + }, + { + "epoch": 2.65, + "learning_rate": 5.842523724031803e-06, + "loss": 0.3668, + "step": 68918 + }, + { + "epoch": 2.65, + "learning_rate": 5.841882533983073e-06, + "loss": 0.2516, + "step": 68919 + }, + { + "epoch": 2.65, + "learning_rate": 5.841241343934342e-06, + "loss": 0.2697, + "step": 68920 + }, + { + "epoch": 2.65, + "learning_rate": 5.840600153885612e-06, + "loss": 0.1874, + "step": 68921 + }, + { + "epoch": 2.65, + "learning_rate": 5.8399589638368816e-06, + "loss": 0.229, + "step": 68922 + }, + { + "epoch": 2.65, + "learning_rate": 5.839317773788152e-06, + "loss": 0.2612, + "step": 68923 + }, + { + "epoch": 2.65, + "learning_rate": 5.838676583739421e-06, + "loss": 0.1809, + "step": 68924 + }, + { + "epoch": 2.65, + "learning_rate": 5.838035393690691e-06, + "loss": 0.3581, + "step": 68925 + }, + { + "epoch": 2.65, + "learning_rate": 5.83739420364196e-06, + "loss": 0.3117, + "step": 68926 + }, + { + "epoch": 2.65, + "learning_rate": 5.836753013593229e-06, + "loss": 0.1025, + "step": 68927 + }, + { + "epoch": 2.65, + "learning_rate": 5.836111823544499e-06, + "loss": 0.2732, + "step": 68928 + }, + { + "epoch": 2.65, + "learning_rate": 5.835470633495768e-06, + "loss": 0.1897, + "step": 68929 + }, + { + "epoch": 2.65, + "learning_rate": 5.834829443447038e-06, + "loss": 0.2512, + "step": 68930 + }, + { + "epoch": 2.65, + "learning_rate": 5.8341882533983075e-06, + "loss": 0.1255, + "step": 68931 + }, + { + "epoch": 2.65, + "learning_rate": 5.8335470633495776e-06, + "loss": 0.1798, + "step": 68932 + }, + { + "epoch": 2.65, + "learning_rate": 5.832905873300847e-06, + "loss": 0.1139, + "step": 68933 + }, + { + "epoch": 2.65, + "learning_rate": 5.832264683252117e-06, + "loss": 0.2138, + "step": 68934 + }, + { + "epoch": 2.65, + "learning_rate": 5.831623493203386e-06, + "loss": 0.1683, + "step": 68935 + }, + { + "epoch": 2.65, + "learning_rate": 5.830982303154656e-06, + "loss": 0.1891, + "step": 68936 + }, + { + "epoch": 2.65, + "learning_rate": 5.830341113105925e-06, + "loss": 0.207, + "step": 68937 + }, + { + "epoch": 2.65, + "learning_rate": 5.829699923057194e-06, + "loss": 0.1259, + "step": 68938 + }, + { + "epoch": 2.65, + "learning_rate": 5.829058733008464e-06, + "loss": 0.1913, + "step": 68939 + }, + { + "epoch": 2.65, + "learning_rate": 5.8284175429597335e-06, + "loss": 0.22, + "step": 68940 + }, + { + "epoch": 2.65, + "learning_rate": 5.8277763529110035e-06, + "loss": 0.2196, + "step": 68941 + }, + { + "epoch": 2.65, + "learning_rate": 5.827135162862273e-06, + "loss": 0.1689, + "step": 68942 + }, + { + "epoch": 2.65, + "learning_rate": 5.826493972813543e-06, + "loss": 0.276, + "step": 68943 + }, + { + "epoch": 2.65, + "learning_rate": 5.825852782764812e-06, + "loss": 0.2137, + "step": 68944 + }, + { + "epoch": 2.65, + "learning_rate": 5.825211592716082e-06, + "loss": 0.3084, + "step": 68945 + }, + { + "epoch": 2.65, + "learning_rate": 5.824570402667351e-06, + "loss": 0.0825, + "step": 68946 + }, + { + "epoch": 2.65, + "learning_rate": 5.82392921261862e-06, + "loss": 0.2594, + "step": 68947 + }, + { + "epoch": 2.65, + "learning_rate": 5.82328802256989e-06, + "loss": 0.378, + "step": 68948 + }, + { + "epoch": 2.65, + "learning_rate": 5.8226468325211595e-06, + "loss": 0.2947, + "step": 68949 + }, + { + "epoch": 2.65, + "learning_rate": 5.8220056424724295e-06, + "loss": 0.2093, + "step": 68950 + }, + { + "epoch": 2.65, + "learning_rate": 5.821364452423699e-06, + "loss": 0.2111, + "step": 68951 + }, + { + "epoch": 2.65, + "learning_rate": 5.820723262374969e-06, + "loss": 0.2082, + "step": 68952 + }, + { + "epoch": 2.65, + "learning_rate": 5.820082072326238e-06, + "loss": 0.2051, + "step": 68953 + }, + { + "epoch": 2.65, + "learning_rate": 5.819440882277508e-06, + "loss": 0.324, + "step": 68954 + }, + { + "epoch": 2.65, + "learning_rate": 5.818799692228777e-06, + "loss": 0.2039, + "step": 68955 + }, + { + "epoch": 2.65, + "learning_rate": 5.818158502180046e-06, + "loss": 0.151, + "step": 68956 + }, + { + "epoch": 2.65, + "learning_rate": 5.817517312131316e-06, + "loss": 0.2732, + "step": 68957 + }, + { + "epoch": 2.65, + "learning_rate": 5.8168761220825854e-06, + "loss": 0.2772, + "step": 68958 + }, + { + "epoch": 2.65, + "learning_rate": 5.8162349320338555e-06, + "loss": 0.0961, + "step": 68959 + }, + { + "epoch": 2.65, + "learning_rate": 5.815593741985125e-06, + "loss": 0.2686, + "step": 68960 + }, + { + "epoch": 2.65, + "learning_rate": 5.814952551936395e-06, + "loss": 0.2158, + "step": 68961 + }, + { + "epoch": 2.65, + "learning_rate": 5.814311361887664e-06, + "loss": 0.1689, + "step": 68962 + }, + { + "epoch": 2.65, + "learning_rate": 5.813670171838934e-06, + "loss": 0.1909, + "step": 68963 + }, + { + "epoch": 2.65, + "learning_rate": 5.813028981790203e-06, + "loss": 0.2076, + "step": 68964 + }, + { + "epoch": 2.65, + "learning_rate": 5.812387791741473e-06, + "loss": 0.255, + "step": 68965 + }, + { + "epoch": 2.65, + "learning_rate": 5.811746601692742e-06, + "loss": 0.1796, + "step": 68966 + }, + { + "epoch": 2.65, + "learning_rate": 5.811105411644011e-06, + "loss": 0.2043, + "step": 68967 + }, + { + "epoch": 2.65, + "learning_rate": 5.8104642215952814e-06, + "loss": 0.12, + "step": 68968 + }, + { + "epoch": 2.65, + "learning_rate": 5.809823031546551e-06, + "loss": 0.2095, + "step": 68969 + }, + { + "epoch": 2.65, + "learning_rate": 5.809181841497821e-06, + "loss": 0.2059, + "step": 68970 + }, + { + "epoch": 2.65, + "learning_rate": 5.80854065144909e-06, + "loss": 0.1391, + "step": 68971 + }, + { + "epoch": 2.65, + "learning_rate": 5.80789946140036e-06, + "loss": 0.2308, + "step": 68972 + }, + { + "epoch": 2.65, + "learning_rate": 5.807258271351629e-06, + "loss": 0.1506, + "step": 68973 + }, + { + "epoch": 2.65, + "learning_rate": 5.806617081302899e-06, + "loss": 0.1874, + "step": 68974 + }, + { + "epoch": 2.65, + "learning_rate": 5.805975891254168e-06, + "loss": 0.1811, + "step": 68975 + }, + { + "epoch": 2.65, + "learning_rate": 5.805334701205437e-06, + "loss": 0.3245, + "step": 68976 + }, + { + "epoch": 2.65, + "learning_rate": 5.804693511156707e-06, + "loss": 0.3017, + "step": 68977 + }, + { + "epoch": 2.65, + "learning_rate": 5.8040523211079766e-06, + "loss": 0.1788, + "step": 68978 + }, + { + "epoch": 2.65, + "learning_rate": 5.803411131059247e-06, + "loss": 0.1499, + "step": 68979 + }, + { + "epoch": 2.65, + "learning_rate": 5.802769941010516e-06, + "loss": 0.1755, + "step": 68980 + }, + { + "epoch": 2.65, + "learning_rate": 5.802128750961786e-06, + "loss": 0.1654, + "step": 68981 + }, + { + "epoch": 2.65, + "learning_rate": 5.801487560913055e-06, + "loss": 0.2099, + "step": 68982 + }, + { + "epoch": 2.65, + "learning_rate": 5.800846370864325e-06, + "loss": 0.1745, + "step": 68983 + }, + { + "epoch": 2.65, + "learning_rate": 5.800205180815594e-06, + "loss": 0.1546, + "step": 68984 + }, + { + "epoch": 2.65, + "learning_rate": 5.799563990766864e-06, + "loss": 0.2331, + "step": 68985 + }, + { + "epoch": 2.65, + "learning_rate": 5.798922800718133e-06, + "loss": 0.1799, + "step": 68986 + }, + { + "epoch": 2.65, + "learning_rate": 5.7982816106694025e-06, + "loss": 0.2392, + "step": 68987 + }, + { + "epoch": 2.65, + "learning_rate": 5.7976404206206726e-06, + "loss": 0.3351, + "step": 68988 + }, + { + "epoch": 2.65, + "learning_rate": 5.796999230571942e-06, + "loss": 0.1824, + "step": 68989 + }, + { + "epoch": 2.65, + "learning_rate": 5.796358040523212e-06, + "loss": 0.1533, + "step": 68990 + }, + { + "epoch": 2.65, + "learning_rate": 5.795716850474481e-06, + "loss": 0.2459, + "step": 68991 + }, + { + "epoch": 2.65, + "learning_rate": 5.795075660425751e-06, + "loss": 0.1471, + "step": 68992 + }, + { + "epoch": 2.65, + "learning_rate": 5.79443447037702e-06, + "loss": 0.1949, + "step": 68993 + }, + { + "epoch": 2.65, + "learning_rate": 5.79379328032829e-06, + "loss": 0.2057, + "step": 68994 + }, + { + "epoch": 2.65, + "learning_rate": 5.793152090279559e-06, + "loss": 0.2904, + "step": 68995 + }, + { + "epoch": 2.65, + "learning_rate": 5.7925109002308285e-06, + "loss": 0.175, + "step": 68996 + }, + { + "epoch": 2.65, + "learning_rate": 5.7918697101820985e-06, + "loss": 0.3607, + "step": 68997 + }, + { + "epoch": 2.65, + "learning_rate": 5.791228520133368e-06, + "loss": 0.145, + "step": 68998 + }, + { + "epoch": 2.65, + "learning_rate": 5.790587330084638e-06, + "loss": 0.2207, + "step": 68999 + }, + { + "epoch": 2.65, + "learning_rate": 5.789946140035907e-06, + "loss": 0.1586, + "step": 69000 + }, + { + "epoch": 2.65, + "learning_rate": 5.789304949987177e-06, + "loss": 0.2957, + "step": 69001 + }, + { + "epoch": 2.65, + "learning_rate": 5.788663759938446e-06, + "loss": 0.135, + "step": 69002 + }, + { + "epoch": 2.65, + "learning_rate": 5.788022569889716e-06, + "loss": 0.2464, + "step": 69003 + }, + { + "epoch": 2.65, + "learning_rate": 5.787381379840985e-06, + "loss": 0.2159, + "step": 69004 + }, + { + "epoch": 2.65, + "learning_rate": 5.786740189792255e-06, + "loss": 0.141, + "step": 69005 + }, + { + "epoch": 2.65, + "learning_rate": 5.7860989997435245e-06, + "loss": 0.2188, + "step": 69006 + }, + { + "epoch": 2.65, + "learning_rate": 5.785457809694794e-06, + "loss": 0.2785, + "step": 69007 + }, + { + "epoch": 2.65, + "learning_rate": 5.784816619646064e-06, + "loss": 0.2555, + "step": 69008 + }, + { + "epoch": 2.65, + "learning_rate": 5.784175429597333e-06, + "loss": 0.2802, + "step": 69009 + }, + { + "epoch": 2.65, + "learning_rate": 5.783534239548603e-06, + "loss": 0.148, + "step": 69010 + }, + { + "epoch": 2.65, + "learning_rate": 5.782893049499872e-06, + "loss": 0.3272, + "step": 69011 + }, + { + "epoch": 2.65, + "learning_rate": 5.782251859451142e-06, + "loss": 0.2584, + "step": 69012 + }, + { + "epoch": 2.65, + "learning_rate": 5.781610669402411e-06, + "loss": 0.3168, + "step": 69013 + }, + { + "epoch": 2.65, + "learning_rate": 5.780969479353681e-06, + "loss": 0.228, + "step": 69014 + }, + { + "epoch": 2.65, + "learning_rate": 5.7803282893049505e-06, + "loss": 0.1187, + "step": 69015 + }, + { + "epoch": 2.65, + "learning_rate": 5.77968709925622e-06, + "loss": 0.3001, + "step": 69016 + }, + { + "epoch": 2.65, + "learning_rate": 5.77904590920749e-06, + "loss": 0.1918, + "step": 69017 + }, + { + "epoch": 2.65, + "learning_rate": 5.778404719158759e-06, + "loss": 0.3109, + "step": 69018 + }, + { + "epoch": 2.65, + "learning_rate": 5.777763529110029e-06, + "loss": 0.2002, + "step": 69019 + }, + { + "epoch": 2.65, + "learning_rate": 5.777122339061298e-06, + "loss": 0.2957, + "step": 69020 + }, + { + "epoch": 2.65, + "learning_rate": 5.776481149012568e-06, + "loss": 0.2411, + "step": 69021 + }, + { + "epoch": 2.65, + "learning_rate": 5.775839958963837e-06, + "loss": 0.2664, + "step": 69022 + }, + { + "epoch": 2.65, + "learning_rate": 5.775198768915107e-06, + "loss": 0.12, + "step": 69023 + }, + { + "epoch": 2.65, + "learning_rate": 5.7745575788663764e-06, + "loss": 0.1465, + "step": 69024 + }, + { + "epoch": 2.65, + "learning_rate": 5.773916388817646e-06, + "loss": 0.1139, + "step": 69025 + }, + { + "epoch": 2.65, + "learning_rate": 5.773275198768916e-06, + "loss": 0.192, + "step": 69026 + }, + { + "epoch": 2.65, + "learning_rate": 5.772634008720185e-06, + "loss": 0.2501, + "step": 69027 + }, + { + "epoch": 2.65, + "learning_rate": 5.771992818671455e-06, + "loss": 0.1601, + "step": 69028 + }, + { + "epoch": 2.65, + "learning_rate": 5.771351628622724e-06, + "loss": 0.2249, + "step": 69029 + }, + { + "epoch": 2.65, + "learning_rate": 5.770710438573994e-06, + "loss": 0.247, + "step": 69030 + }, + { + "epoch": 2.65, + "learning_rate": 5.770069248525263e-06, + "loss": 0.2861, + "step": 69031 + }, + { + "epoch": 2.65, + "learning_rate": 5.769428058476533e-06, + "loss": 0.3407, + "step": 69032 + }, + { + "epoch": 2.65, + "learning_rate": 5.768786868427802e-06, + "loss": 0.1595, + "step": 69033 + }, + { + "epoch": 2.65, + "learning_rate": 5.7681456783790724e-06, + "loss": 0.2465, + "step": 69034 + }, + { + "epoch": 2.65, + "learning_rate": 5.767504488330342e-06, + "loss": 0.2994, + "step": 69035 + }, + { + "epoch": 2.65, + "learning_rate": 5.766863298281611e-06, + "loss": 0.2505, + "step": 69036 + }, + { + "epoch": 2.65, + "learning_rate": 5.766222108232881e-06, + "loss": 0.1553, + "step": 69037 + }, + { + "epoch": 2.65, + "learning_rate": 5.76558091818415e-06, + "loss": 0.1767, + "step": 69038 + }, + { + "epoch": 2.65, + "learning_rate": 5.76493972813542e-06, + "loss": 0.1681, + "step": 69039 + }, + { + "epoch": 2.65, + "learning_rate": 5.764298538086689e-06, + "loss": 0.2147, + "step": 69040 + }, + { + "epoch": 2.65, + "learning_rate": 5.763657348037959e-06, + "loss": 0.1698, + "step": 69041 + }, + { + "epoch": 2.65, + "learning_rate": 5.763016157989228e-06, + "loss": 0.2012, + "step": 69042 + }, + { + "epoch": 2.65, + "learning_rate": 5.762374967940498e-06, + "loss": 0.1537, + "step": 69043 + }, + { + "epoch": 2.65, + "learning_rate": 5.7617337778917676e-06, + "loss": 0.2055, + "step": 69044 + }, + { + "epoch": 2.65, + "learning_rate": 5.761092587843037e-06, + "loss": 0.2166, + "step": 69045 + }, + { + "epoch": 2.65, + "learning_rate": 5.760451397794307e-06, + "loss": 0.2477, + "step": 69046 + }, + { + "epoch": 2.65, + "learning_rate": 5.759810207745576e-06, + "loss": 0.2263, + "step": 69047 + }, + { + "epoch": 2.65, + "learning_rate": 5.759169017696846e-06, + "loss": 0.2097, + "step": 69048 + }, + { + "epoch": 2.65, + "learning_rate": 5.758527827648115e-06, + "loss": 0.3708, + "step": 69049 + }, + { + "epoch": 2.65, + "learning_rate": 5.757886637599385e-06, + "loss": 0.1389, + "step": 69050 + }, + { + "epoch": 2.65, + "learning_rate": 5.757245447550654e-06, + "loss": 0.1363, + "step": 69051 + }, + { + "epoch": 2.65, + "learning_rate": 5.756604257501924e-06, + "loss": 0.1719, + "step": 69052 + }, + { + "epoch": 2.65, + "learning_rate": 5.7559630674531935e-06, + "loss": 0.2544, + "step": 69053 + }, + { + "epoch": 2.65, + "learning_rate": 5.7553218774044636e-06, + "loss": 0.3078, + "step": 69054 + }, + { + "epoch": 2.65, + "learning_rate": 5.754680687355733e-06, + "loss": 0.284, + "step": 69055 + }, + { + "epoch": 2.65, + "learning_rate": 5.754039497307002e-06, + "loss": 0.2459, + "step": 69056 + }, + { + "epoch": 2.66, + "learning_rate": 5.753398307258272e-06, + "loss": 0.1312, + "step": 69057 + }, + { + "epoch": 2.66, + "learning_rate": 5.752757117209541e-06, + "loss": 0.1424, + "step": 69058 + }, + { + "epoch": 2.66, + "learning_rate": 5.752115927160811e-06, + "loss": 0.2879, + "step": 69059 + }, + { + "epoch": 2.66, + "learning_rate": 5.75147473711208e-06, + "loss": 0.1342, + "step": 69060 + }, + { + "epoch": 2.66, + "learning_rate": 5.75083354706335e-06, + "loss": 0.1255, + "step": 69061 + }, + { + "epoch": 2.66, + "learning_rate": 5.7501923570146195e-06, + "loss": 0.1298, + "step": 69062 + }, + { + "epoch": 2.66, + "learning_rate": 5.7495511669658895e-06, + "loss": 0.238, + "step": 69063 + }, + { + "epoch": 2.66, + "learning_rate": 5.748909976917159e-06, + "loss": 0.1069, + "step": 69064 + }, + { + "epoch": 2.66, + "learning_rate": 5.748268786868428e-06, + "loss": 0.2274, + "step": 69065 + }, + { + "epoch": 2.66, + "learning_rate": 5.747627596819698e-06, + "loss": 0.2856, + "step": 69066 + }, + { + "epoch": 2.66, + "learning_rate": 5.746986406770967e-06, + "loss": 0.1952, + "step": 69067 + }, + { + "epoch": 2.66, + "learning_rate": 5.746345216722237e-06, + "loss": 0.2208, + "step": 69068 + }, + { + "epoch": 2.66, + "learning_rate": 5.745704026673506e-06, + "loss": 0.2532, + "step": 69069 + }, + { + "epoch": 2.66, + "learning_rate": 5.745062836624776e-06, + "loss": 0.2056, + "step": 69070 + }, + { + "epoch": 2.66, + "learning_rate": 5.7444216465760455e-06, + "loss": 0.1595, + "step": 69071 + }, + { + "epoch": 2.66, + "learning_rate": 5.7437804565273155e-06, + "loss": 0.1642, + "step": 69072 + }, + { + "epoch": 2.66, + "learning_rate": 5.743139266478585e-06, + "loss": 0.2497, + "step": 69073 + }, + { + "epoch": 2.66, + "learning_rate": 5.742498076429854e-06, + "loss": 0.2237, + "step": 69074 + }, + { + "epoch": 2.66, + "learning_rate": 5.741856886381124e-06, + "loss": 0.1872, + "step": 69075 + }, + { + "epoch": 2.66, + "learning_rate": 5.741215696332393e-06, + "loss": 0.243, + "step": 69076 + }, + { + "epoch": 2.66, + "learning_rate": 5.740574506283663e-06, + "loss": 0.2082, + "step": 69077 + }, + { + "epoch": 2.66, + "learning_rate": 5.739933316234932e-06, + "loss": 0.1203, + "step": 69078 + }, + { + "epoch": 2.66, + "learning_rate": 5.739292126186202e-06, + "loss": 0.1665, + "step": 69079 + }, + { + "epoch": 2.66, + "learning_rate": 5.7386509361374714e-06, + "loss": 0.0793, + "step": 69080 + }, + { + "epoch": 2.66, + "learning_rate": 5.7380097460887415e-06, + "loss": 0.2539, + "step": 69081 + }, + { + "epoch": 2.66, + "learning_rate": 5.737368556040011e-06, + "loss": 0.1652, + "step": 69082 + }, + { + "epoch": 2.66, + "learning_rate": 5.736727365991281e-06, + "loss": 0.2482, + "step": 69083 + }, + { + "epoch": 2.66, + "learning_rate": 5.73608617594255e-06, + "loss": 0.3086, + "step": 69084 + }, + { + "epoch": 2.66, + "learning_rate": 5.735444985893819e-06, + "loss": 0.0706, + "step": 69085 + }, + { + "epoch": 2.66, + "learning_rate": 5.734803795845089e-06, + "loss": 0.3062, + "step": 69086 + }, + { + "epoch": 2.66, + "learning_rate": 5.734162605796358e-06, + "loss": 0.1122, + "step": 69087 + }, + { + "epoch": 2.66, + "learning_rate": 5.733521415747628e-06, + "loss": 0.0969, + "step": 69088 + }, + { + "epoch": 2.66, + "learning_rate": 5.732880225698897e-06, + "loss": 0.1903, + "step": 69089 + }, + { + "epoch": 2.66, + "learning_rate": 5.7322390356501674e-06, + "loss": 0.4765, + "step": 69090 + }, + { + "epoch": 2.66, + "learning_rate": 5.731597845601437e-06, + "loss": 0.0932, + "step": 69091 + }, + { + "epoch": 2.66, + "learning_rate": 5.730956655552707e-06, + "loss": 0.1714, + "step": 69092 + }, + { + "epoch": 2.66, + "learning_rate": 5.730315465503976e-06, + "loss": 0.2732, + "step": 69093 + }, + { + "epoch": 2.66, + "learning_rate": 5.729674275455245e-06, + "loss": 0.1789, + "step": 69094 + }, + { + "epoch": 2.66, + "learning_rate": 5.729033085406515e-06, + "loss": 0.1199, + "step": 69095 + }, + { + "epoch": 2.66, + "learning_rate": 5.728391895357784e-06, + "loss": 0.1299, + "step": 69096 + }, + { + "epoch": 2.66, + "learning_rate": 5.727750705309054e-06, + "loss": 0.2267, + "step": 69097 + }, + { + "epoch": 2.66, + "learning_rate": 5.727109515260323e-06, + "loss": 0.2001, + "step": 69098 + }, + { + "epoch": 2.66, + "learning_rate": 5.726468325211593e-06, + "loss": 0.2798, + "step": 69099 + }, + { + "epoch": 2.66, + "learning_rate": 5.7258271351628626e-06, + "loss": 0.1481, + "step": 69100 + }, + { + "epoch": 2.66, + "learning_rate": 5.725185945114133e-06, + "loss": 0.2425, + "step": 69101 + }, + { + "epoch": 2.66, + "learning_rate": 5.724544755065402e-06, + "loss": 0.2533, + "step": 69102 + }, + { + "epoch": 2.66, + "learning_rate": 5.723903565016672e-06, + "loss": 0.2595, + "step": 69103 + }, + { + "epoch": 2.66, + "learning_rate": 5.723262374967941e-06, + "loss": 0.2957, + "step": 69104 + }, + { + "epoch": 2.66, + "learning_rate": 5.72262118491921e-06, + "loss": 0.2665, + "step": 69105 + }, + { + "epoch": 2.66, + "learning_rate": 5.72197999487048e-06, + "loss": 0.1733, + "step": 69106 + }, + { + "epoch": 2.66, + "learning_rate": 5.721338804821749e-06, + "loss": 0.2843, + "step": 69107 + }, + { + "epoch": 2.66, + "learning_rate": 5.720697614773019e-06, + "loss": 0.2181, + "step": 69108 + }, + { + "epoch": 2.66, + "learning_rate": 5.7200564247242885e-06, + "loss": 0.1652, + "step": 69109 + }, + { + "epoch": 2.66, + "learning_rate": 5.7194152346755586e-06, + "loss": 0.1462, + "step": 69110 + }, + { + "epoch": 2.66, + "learning_rate": 5.718774044626828e-06, + "loss": 0.2881, + "step": 69111 + }, + { + "epoch": 2.66, + "learning_rate": 5.718132854578098e-06, + "loss": 0.1584, + "step": 69112 + }, + { + "epoch": 2.66, + "learning_rate": 5.717491664529367e-06, + "loss": 0.2489, + "step": 69113 + }, + { + "epoch": 2.66, + "learning_rate": 5.716850474480636e-06, + "loss": 0.1804, + "step": 69114 + }, + { + "epoch": 2.66, + "learning_rate": 5.716209284431906e-06, + "loss": 0.2648, + "step": 69115 + }, + { + "epoch": 2.66, + "learning_rate": 5.715568094383175e-06, + "loss": 0.2275, + "step": 69116 + }, + { + "epoch": 2.66, + "learning_rate": 5.714926904334445e-06, + "loss": 0.1869, + "step": 69117 + }, + { + "epoch": 2.66, + "learning_rate": 5.7142857142857145e-06, + "loss": 0.1706, + "step": 69118 + }, + { + "epoch": 2.66, + "learning_rate": 5.7136445242369845e-06, + "loss": 0.1571, + "step": 69119 + }, + { + "epoch": 2.66, + "learning_rate": 5.713003334188254e-06, + "loss": 0.3694, + "step": 69120 + }, + { + "epoch": 2.66, + "learning_rate": 5.712362144139524e-06, + "loss": 0.1804, + "step": 69121 + }, + { + "epoch": 2.66, + "learning_rate": 5.711720954090793e-06, + "loss": 0.2951, + "step": 69122 + }, + { + "epoch": 2.66, + "learning_rate": 5.711079764042063e-06, + "loss": 0.1549, + "step": 69123 + }, + { + "epoch": 2.66, + "learning_rate": 5.710438573993332e-06, + "loss": 0.2057, + "step": 69124 + }, + { + "epoch": 2.66, + "learning_rate": 5.709797383944601e-06, + "loss": 0.2676, + "step": 69125 + }, + { + "epoch": 2.66, + "learning_rate": 5.709156193895871e-06, + "loss": 0.1937, + "step": 69126 + }, + { + "epoch": 2.66, + "learning_rate": 5.7085150038471405e-06, + "loss": 0.2499, + "step": 69127 + }, + { + "epoch": 2.66, + "learning_rate": 5.7078738137984105e-06, + "loss": 0.1849, + "step": 69128 + }, + { + "epoch": 2.66, + "learning_rate": 5.70723262374968e-06, + "loss": 0.1931, + "step": 69129 + }, + { + "epoch": 2.66, + "learning_rate": 5.70659143370095e-06, + "loss": 0.1731, + "step": 69130 + }, + { + "epoch": 2.66, + "learning_rate": 5.705950243652219e-06, + "loss": 0.1696, + "step": 69131 + }, + { + "epoch": 2.66, + "learning_rate": 5.705309053603489e-06, + "loss": 0.2173, + "step": 69132 + }, + { + "epoch": 2.66, + "learning_rate": 5.704667863554758e-06, + "loss": 0.3196, + "step": 69133 + }, + { + "epoch": 2.66, + "learning_rate": 5.704026673506027e-06, + "loss": 0.2278, + "step": 69134 + }, + { + "epoch": 2.66, + "learning_rate": 5.703385483457297e-06, + "loss": 0.2442, + "step": 69135 + }, + { + "epoch": 2.66, + "learning_rate": 5.7027442934085664e-06, + "loss": 0.1276, + "step": 69136 + }, + { + "epoch": 2.66, + "learning_rate": 5.7021031033598365e-06, + "loss": 0.264, + "step": 69137 + }, + { + "epoch": 2.66, + "learning_rate": 5.701461913311106e-06, + "loss": 0.1875, + "step": 69138 + }, + { + "epoch": 2.66, + "learning_rate": 5.700820723262376e-06, + "loss": 0.2667, + "step": 69139 + }, + { + "epoch": 2.66, + "learning_rate": 5.700179533213645e-06, + "loss": 0.2862, + "step": 69140 + }, + { + "epoch": 2.66, + "learning_rate": 5.699538343164915e-06, + "loss": 0.2324, + "step": 69141 + }, + { + "epoch": 2.66, + "learning_rate": 5.698897153116184e-06, + "loss": 0.1815, + "step": 69142 + }, + { + "epoch": 2.66, + "learning_rate": 5.698255963067453e-06, + "loss": 0.1485, + "step": 69143 + }, + { + "epoch": 2.66, + "learning_rate": 5.697614773018723e-06, + "loss": 0.1742, + "step": 69144 + }, + { + "epoch": 2.66, + "learning_rate": 5.696973582969992e-06, + "loss": 0.1604, + "step": 69145 + }, + { + "epoch": 2.66, + "learning_rate": 5.6963323929212624e-06, + "loss": 0.1661, + "step": 69146 + }, + { + "epoch": 2.66, + "learning_rate": 5.695691202872532e-06, + "loss": 0.3343, + "step": 69147 + }, + { + "epoch": 2.66, + "learning_rate": 5.695050012823802e-06, + "loss": 0.3783, + "step": 69148 + }, + { + "epoch": 2.66, + "learning_rate": 5.694408822775071e-06, + "loss": 0.1409, + "step": 69149 + }, + { + "epoch": 2.66, + "learning_rate": 5.693767632726341e-06, + "loss": 0.2064, + "step": 69150 + }, + { + "epoch": 2.66, + "learning_rate": 5.69312644267761e-06, + "loss": 0.2099, + "step": 69151 + }, + { + "epoch": 2.66, + "learning_rate": 5.69248525262888e-06, + "loss": 0.2011, + "step": 69152 + }, + { + "epoch": 2.66, + "learning_rate": 5.691844062580149e-06, + "loss": 0.11, + "step": 69153 + }, + { + "epoch": 2.66, + "learning_rate": 5.691202872531418e-06, + "loss": 0.123, + "step": 69154 + }, + { + "epoch": 2.66, + "learning_rate": 5.690561682482688e-06, + "loss": 0.2013, + "step": 69155 + }, + { + "epoch": 2.66, + "learning_rate": 5.689920492433958e-06, + "loss": 0.197, + "step": 69156 + }, + { + "epoch": 2.66, + "learning_rate": 5.689279302385228e-06, + "loss": 0.2643, + "step": 69157 + }, + { + "epoch": 2.66, + "learning_rate": 5.688638112336497e-06, + "loss": 0.1843, + "step": 69158 + }, + { + "epoch": 2.66, + "learning_rate": 5.687996922287767e-06, + "loss": 0.1548, + "step": 69159 + }, + { + "epoch": 2.66, + "learning_rate": 5.687355732239036e-06, + "loss": 0.2835, + "step": 69160 + }, + { + "epoch": 2.66, + "learning_rate": 5.686714542190306e-06, + "loss": 0.2358, + "step": 69161 + }, + { + "epoch": 2.66, + "learning_rate": 5.686073352141575e-06, + "loss": 0.3402, + "step": 69162 + }, + { + "epoch": 2.66, + "learning_rate": 5.685432162092844e-06, + "loss": 0.2377, + "step": 69163 + }, + { + "epoch": 2.66, + "learning_rate": 5.684790972044114e-06, + "loss": 0.2027, + "step": 69164 + }, + { + "epoch": 2.66, + "learning_rate": 5.6841497819953835e-06, + "loss": 0.2616, + "step": 69165 + }, + { + "epoch": 2.66, + "learning_rate": 5.6835085919466536e-06, + "loss": 0.3567, + "step": 69166 + }, + { + "epoch": 2.66, + "learning_rate": 5.682867401897923e-06, + "loss": 0.2363, + "step": 69167 + }, + { + "epoch": 2.66, + "learning_rate": 5.682226211849193e-06, + "loss": 0.1896, + "step": 69168 + }, + { + "epoch": 2.66, + "learning_rate": 5.681585021800462e-06, + "loss": 0.2072, + "step": 69169 + }, + { + "epoch": 2.66, + "learning_rate": 5.680943831751732e-06, + "loss": 0.4132, + "step": 69170 + }, + { + "epoch": 2.66, + "learning_rate": 5.680302641703001e-06, + "loss": 0.3522, + "step": 69171 + }, + { + "epoch": 2.66, + "learning_rate": 5.679661451654271e-06, + "loss": 0.0877, + "step": 69172 + }, + { + "epoch": 2.66, + "learning_rate": 5.67902026160554e-06, + "loss": 0.2101, + "step": 69173 + }, + { + "epoch": 2.66, + "learning_rate": 5.6783790715568095e-06, + "loss": 0.3121, + "step": 69174 + }, + { + "epoch": 2.66, + "learning_rate": 5.6777378815080795e-06, + "loss": 0.1651, + "step": 69175 + }, + { + "epoch": 2.66, + "learning_rate": 5.677096691459349e-06, + "loss": 0.2184, + "step": 69176 + }, + { + "epoch": 2.66, + "learning_rate": 5.676455501410619e-06, + "loss": 0.1493, + "step": 69177 + }, + { + "epoch": 2.66, + "learning_rate": 5.675814311361888e-06, + "loss": 0.359, + "step": 69178 + }, + { + "epoch": 2.66, + "learning_rate": 5.675173121313158e-06, + "loss": 0.2726, + "step": 69179 + }, + { + "epoch": 2.66, + "learning_rate": 5.674531931264427e-06, + "loss": 0.2713, + "step": 69180 + }, + { + "epoch": 2.66, + "learning_rate": 5.673890741215697e-06, + "loss": 0.3605, + "step": 69181 + }, + { + "epoch": 2.66, + "learning_rate": 5.673249551166966e-06, + "loss": 0.0596, + "step": 69182 + }, + { + "epoch": 2.66, + "learning_rate": 5.6726083611182355e-06, + "loss": 0.2653, + "step": 69183 + }, + { + "epoch": 2.66, + "learning_rate": 5.6719671710695055e-06, + "loss": 0.3976, + "step": 69184 + }, + { + "epoch": 2.66, + "learning_rate": 5.671325981020775e-06, + "loss": 0.1634, + "step": 69185 + }, + { + "epoch": 2.66, + "learning_rate": 5.670684790972045e-06, + "loss": 0.1946, + "step": 69186 + }, + { + "epoch": 2.66, + "learning_rate": 5.670043600923314e-06, + "loss": 0.1499, + "step": 69187 + }, + { + "epoch": 2.66, + "learning_rate": 5.669402410874584e-06, + "loss": 0.2211, + "step": 69188 + }, + { + "epoch": 2.66, + "learning_rate": 5.668761220825853e-06, + "loss": 0.3277, + "step": 69189 + }, + { + "epoch": 2.66, + "learning_rate": 5.668120030777123e-06, + "loss": 0.2312, + "step": 69190 + }, + { + "epoch": 2.66, + "learning_rate": 5.667478840728392e-06, + "loss": 0.1355, + "step": 69191 + }, + { + "epoch": 2.66, + "learning_rate": 5.6668376506796615e-06, + "loss": 0.2485, + "step": 69192 + }, + { + "epoch": 2.66, + "learning_rate": 5.6661964606309315e-06, + "loss": 0.2207, + "step": 69193 + }, + { + "epoch": 2.66, + "learning_rate": 5.665555270582201e-06, + "loss": 0.1771, + "step": 69194 + }, + { + "epoch": 2.66, + "learning_rate": 5.664914080533471e-06, + "loss": 0.2475, + "step": 69195 + }, + { + "epoch": 2.66, + "learning_rate": 5.66427289048474e-06, + "loss": 0.2339, + "step": 69196 + }, + { + "epoch": 2.66, + "learning_rate": 5.66363170043601e-06, + "loss": 0.1593, + "step": 69197 + }, + { + "epoch": 2.66, + "learning_rate": 5.662990510387279e-06, + "loss": 0.1524, + "step": 69198 + }, + { + "epoch": 2.66, + "learning_rate": 5.662349320338549e-06, + "loss": 0.2215, + "step": 69199 + }, + { + "epoch": 2.66, + "learning_rate": 5.661708130289818e-06, + "loss": 0.1983, + "step": 69200 + }, + { + "epoch": 2.66, + "learning_rate": 5.661066940241088e-06, + "loss": 0.1192, + "step": 69201 + }, + { + "epoch": 2.66, + "learning_rate": 5.6604257501923574e-06, + "loss": 0.3812, + "step": 69202 + }, + { + "epoch": 2.66, + "learning_rate": 5.659784560143627e-06, + "loss": 0.2525, + "step": 69203 + }, + { + "epoch": 2.66, + "learning_rate": 5.659143370094897e-06, + "loss": 0.1446, + "step": 69204 + }, + { + "epoch": 2.66, + "learning_rate": 5.658502180046166e-06, + "loss": 0.2125, + "step": 69205 + }, + { + "epoch": 2.66, + "learning_rate": 5.657860989997436e-06, + "loss": 0.1638, + "step": 69206 + }, + { + "epoch": 2.66, + "learning_rate": 5.657219799948705e-06, + "loss": 0.1675, + "step": 69207 + }, + { + "epoch": 2.66, + "learning_rate": 5.656578609899975e-06, + "loss": 0.1835, + "step": 69208 + }, + { + "epoch": 2.66, + "learning_rate": 5.655937419851244e-06, + "loss": 0.1158, + "step": 69209 + }, + { + "epoch": 2.66, + "learning_rate": 5.655296229802514e-06, + "loss": 0.3281, + "step": 69210 + }, + { + "epoch": 2.66, + "learning_rate": 5.654655039753783e-06, + "loss": 0.1875, + "step": 69211 + }, + { + "epoch": 2.66, + "learning_rate": 5.654013849705053e-06, + "loss": 0.1393, + "step": 69212 + }, + { + "epoch": 2.66, + "learning_rate": 5.653372659656323e-06, + "loss": 0.2446, + "step": 69213 + }, + { + "epoch": 2.66, + "learning_rate": 5.652731469607592e-06, + "loss": 0.1481, + "step": 69214 + }, + { + "epoch": 2.66, + "learning_rate": 5.652090279558862e-06, + "loss": 0.1576, + "step": 69215 + }, + { + "epoch": 2.66, + "learning_rate": 5.651449089510131e-06, + "loss": 0.1745, + "step": 69216 + }, + { + "epoch": 2.66, + "learning_rate": 5.650807899461401e-06, + "loss": 0.1623, + "step": 69217 + }, + { + "epoch": 2.66, + "learning_rate": 5.65016670941267e-06, + "loss": 0.2658, + "step": 69218 + }, + { + "epoch": 2.66, + "learning_rate": 5.64952551936394e-06, + "loss": 0.3903, + "step": 69219 + }, + { + "epoch": 2.66, + "learning_rate": 5.648884329315209e-06, + "loss": 0.076, + "step": 69220 + }, + { + "epoch": 2.66, + "learning_rate": 5.648243139266479e-06, + "loss": 0.2749, + "step": 69221 + }, + { + "epoch": 2.66, + "learning_rate": 5.6476019492177486e-06, + "loss": 0.1529, + "step": 69222 + }, + { + "epoch": 2.66, + "learning_rate": 5.646960759169018e-06, + "loss": 0.4004, + "step": 69223 + }, + { + "epoch": 2.66, + "learning_rate": 5.646319569120288e-06, + "loss": 0.1654, + "step": 69224 + }, + { + "epoch": 2.66, + "learning_rate": 5.645678379071557e-06, + "loss": 0.2592, + "step": 69225 + }, + { + "epoch": 2.66, + "learning_rate": 5.645037189022827e-06, + "loss": 0.2552, + "step": 69226 + }, + { + "epoch": 2.66, + "learning_rate": 5.644395998974096e-06, + "loss": 0.0678, + "step": 69227 + }, + { + "epoch": 2.66, + "learning_rate": 5.643754808925366e-06, + "loss": 0.2783, + "step": 69228 + }, + { + "epoch": 2.66, + "learning_rate": 5.643113618876635e-06, + "loss": 0.3085, + "step": 69229 + }, + { + "epoch": 2.66, + "learning_rate": 5.642472428827905e-06, + "loss": 0.1861, + "step": 69230 + }, + { + "epoch": 2.66, + "learning_rate": 5.6418312387791745e-06, + "loss": 0.2233, + "step": 69231 + }, + { + "epoch": 2.66, + "learning_rate": 5.641190048730444e-06, + "loss": 0.0552, + "step": 69232 + }, + { + "epoch": 2.66, + "learning_rate": 5.640548858681714e-06, + "loss": 0.1957, + "step": 69233 + }, + { + "epoch": 2.66, + "learning_rate": 5.639907668632983e-06, + "loss": 0.2221, + "step": 69234 + }, + { + "epoch": 2.66, + "learning_rate": 5.639266478584253e-06, + "loss": 0.262, + "step": 69235 + }, + { + "epoch": 2.66, + "learning_rate": 5.638625288535522e-06, + "loss": 0.3355, + "step": 69236 + }, + { + "epoch": 2.66, + "learning_rate": 5.637984098486792e-06, + "loss": 0.1862, + "step": 69237 + }, + { + "epoch": 2.66, + "learning_rate": 5.637342908438061e-06, + "loss": 0.2138, + "step": 69238 + }, + { + "epoch": 2.66, + "learning_rate": 5.636701718389331e-06, + "loss": 0.1539, + "step": 69239 + }, + { + "epoch": 2.66, + "learning_rate": 5.6360605283406005e-06, + "loss": 0.2565, + "step": 69240 + }, + { + "epoch": 2.66, + "learning_rate": 5.6354193382918705e-06, + "loss": 0.2359, + "step": 69241 + }, + { + "epoch": 2.66, + "learning_rate": 5.63477814824314e-06, + "loss": 0.194, + "step": 69242 + }, + { + "epoch": 2.66, + "learning_rate": 5.634136958194409e-06, + "loss": 0.2498, + "step": 69243 + }, + { + "epoch": 2.66, + "learning_rate": 5.633495768145679e-06, + "loss": 0.234, + "step": 69244 + }, + { + "epoch": 2.66, + "learning_rate": 5.632854578096948e-06, + "loss": 0.1708, + "step": 69245 + }, + { + "epoch": 2.66, + "learning_rate": 5.632213388048218e-06, + "loss": 0.2312, + "step": 69246 + }, + { + "epoch": 2.66, + "learning_rate": 5.631572197999487e-06, + "loss": 0.2314, + "step": 69247 + }, + { + "epoch": 2.66, + "learning_rate": 5.630931007950757e-06, + "loss": 0.1998, + "step": 69248 + }, + { + "epoch": 2.66, + "learning_rate": 5.6302898179020265e-06, + "loss": 0.1288, + "step": 69249 + }, + { + "epoch": 2.66, + "learning_rate": 5.6296486278532965e-06, + "loss": 0.2326, + "step": 69250 + }, + { + "epoch": 2.66, + "learning_rate": 5.629007437804566e-06, + "loss": 0.3359, + "step": 69251 + }, + { + "epoch": 2.66, + "learning_rate": 5.628366247755835e-06, + "loss": 0.2664, + "step": 69252 + }, + { + "epoch": 2.66, + "learning_rate": 5.627725057707105e-06, + "loss": 0.2543, + "step": 69253 + }, + { + "epoch": 2.66, + "learning_rate": 5.627083867658374e-06, + "loss": 0.1516, + "step": 69254 + }, + { + "epoch": 2.66, + "learning_rate": 5.626442677609644e-06, + "loss": 0.2037, + "step": 69255 + }, + { + "epoch": 2.66, + "learning_rate": 5.625801487560913e-06, + "loss": 0.3306, + "step": 69256 + }, + { + "epoch": 2.66, + "learning_rate": 5.625160297512183e-06, + "loss": 0.427, + "step": 69257 + }, + { + "epoch": 2.66, + "learning_rate": 5.6245191074634524e-06, + "loss": 0.1083, + "step": 69258 + }, + { + "epoch": 2.66, + "learning_rate": 5.6238779174147225e-06, + "loss": 0.2454, + "step": 69259 + }, + { + "epoch": 2.66, + "learning_rate": 5.623236727365992e-06, + "loss": 0.2341, + "step": 69260 + }, + { + "epoch": 2.66, + "learning_rate": 5.622595537317261e-06, + "loss": 0.257, + "step": 69261 + }, + { + "epoch": 2.66, + "learning_rate": 5.621954347268531e-06, + "loss": 0.2626, + "step": 69262 + }, + { + "epoch": 2.66, + "learning_rate": 5.6213131572198e-06, + "loss": 0.2077, + "step": 69263 + }, + { + "epoch": 2.66, + "learning_rate": 5.62067196717107e-06, + "loss": 0.1549, + "step": 69264 + }, + { + "epoch": 2.66, + "learning_rate": 5.620030777122339e-06, + "loss": 0.1798, + "step": 69265 + }, + { + "epoch": 2.66, + "learning_rate": 5.619389587073609e-06, + "loss": 0.1971, + "step": 69266 + }, + { + "epoch": 2.66, + "learning_rate": 5.618748397024878e-06, + "loss": 0.2524, + "step": 69267 + }, + { + "epoch": 2.66, + "learning_rate": 5.6181072069761484e-06, + "loss": 0.1919, + "step": 69268 + }, + { + "epoch": 2.66, + "learning_rate": 5.617466016927418e-06, + "loss": 0.252, + "step": 69269 + }, + { + "epoch": 2.66, + "learning_rate": 5.616824826878688e-06, + "loss": 0.2638, + "step": 69270 + }, + { + "epoch": 2.66, + "learning_rate": 5.616183636829957e-06, + "loss": 0.2684, + "step": 69271 + }, + { + "epoch": 2.66, + "learning_rate": 5.615542446781226e-06, + "loss": 0.1979, + "step": 69272 + }, + { + "epoch": 2.66, + "learning_rate": 5.614901256732496e-06, + "loss": 0.2861, + "step": 69273 + }, + { + "epoch": 2.66, + "learning_rate": 5.614260066683765e-06, + "loss": 0.2066, + "step": 69274 + }, + { + "epoch": 2.66, + "learning_rate": 5.613618876635035e-06, + "loss": 0.1139, + "step": 69275 + }, + { + "epoch": 2.66, + "learning_rate": 5.612977686586304e-06, + "loss": 0.1525, + "step": 69276 + }, + { + "epoch": 2.66, + "learning_rate": 5.612336496537574e-06, + "loss": 0.2338, + "step": 69277 + }, + { + "epoch": 2.66, + "learning_rate": 5.611695306488844e-06, + "loss": 0.2122, + "step": 69278 + }, + { + "epoch": 2.66, + "learning_rate": 5.611054116440114e-06, + "loss": 0.2431, + "step": 69279 + }, + { + "epoch": 2.66, + "learning_rate": 5.610412926391383e-06, + "loss": 0.3037, + "step": 69280 + }, + { + "epoch": 2.66, + "learning_rate": 5.609771736342652e-06, + "loss": 0.1808, + "step": 69281 + }, + { + "epoch": 2.66, + "learning_rate": 5.609130546293922e-06, + "loss": 0.3143, + "step": 69282 + }, + { + "epoch": 2.66, + "learning_rate": 5.608489356245191e-06, + "loss": 0.2116, + "step": 69283 + }, + { + "epoch": 2.66, + "learning_rate": 5.607848166196461e-06, + "loss": 0.1485, + "step": 69284 + }, + { + "epoch": 2.66, + "learning_rate": 5.60720697614773e-06, + "loss": 0.1675, + "step": 69285 + }, + { + "epoch": 2.66, + "learning_rate": 5.606565786099e-06, + "loss": 0.2208, + "step": 69286 + }, + { + "epoch": 2.66, + "learning_rate": 5.6059245960502695e-06, + "loss": 0.2873, + "step": 69287 + }, + { + "epoch": 2.66, + "learning_rate": 5.6052834060015396e-06, + "loss": 0.3351, + "step": 69288 + }, + { + "epoch": 2.66, + "learning_rate": 5.604642215952809e-06, + "loss": 0.1774, + "step": 69289 + }, + { + "epoch": 2.66, + "learning_rate": 5.604001025904079e-06, + "loss": 0.2021, + "step": 69290 + }, + { + "epoch": 2.66, + "learning_rate": 5.603359835855348e-06, + "loss": 0.1794, + "step": 69291 + }, + { + "epoch": 2.66, + "learning_rate": 5.602718645806617e-06, + "loss": 0.1596, + "step": 69292 + }, + { + "epoch": 2.66, + "learning_rate": 5.602077455757887e-06, + "loss": 0.3841, + "step": 69293 + }, + { + "epoch": 2.66, + "learning_rate": 5.601436265709156e-06, + "loss": 0.2691, + "step": 69294 + }, + { + "epoch": 2.66, + "learning_rate": 5.600795075660426e-06, + "loss": 0.1295, + "step": 69295 + }, + { + "epoch": 2.66, + "learning_rate": 5.6001538856116955e-06, + "loss": 0.2066, + "step": 69296 + }, + { + "epoch": 2.66, + "learning_rate": 5.5995126955629655e-06, + "loss": 0.1577, + "step": 69297 + }, + { + "epoch": 2.66, + "learning_rate": 5.598871505514235e-06, + "loss": 0.2248, + "step": 69298 + }, + { + "epoch": 2.66, + "learning_rate": 5.598230315465505e-06, + "loss": 0.2335, + "step": 69299 + }, + { + "epoch": 2.66, + "learning_rate": 5.597589125416774e-06, + "loss": 0.4284, + "step": 69300 + }, + { + "epoch": 2.66, + "learning_rate": 5.596947935368043e-06, + "loss": 0.2587, + "step": 69301 + }, + { + "epoch": 2.66, + "learning_rate": 5.596306745319313e-06, + "loss": 0.1561, + "step": 69302 + }, + { + "epoch": 2.66, + "learning_rate": 5.595665555270582e-06, + "loss": 0.1846, + "step": 69303 + }, + { + "epoch": 2.66, + "learning_rate": 5.595024365221852e-06, + "loss": 0.1988, + "step": 69304 + }, + { + "epoch": 2.66, + "learning_rate": 5.5943831751731215e-06, + "loss": 0.3659, + "step": 69305 + }, + { + "epoch": 2.66, + "learning_rate": 5.5937419851243915e-06, + "loss": 0.1985, + "step": 69306 + }, + { + "epoch": 2.66, + "learning_rate": 5.593100795075661e-06, + "loss": 0.1419, + "step": 69307 + }, + { + "epoch": 2.66, + "learning_rate": 5.592459605026931e-06, + "loss": 0.1576, + "step": 69308 + }, + { + "epoch": 2.66, + "learning_rate": 5.5918184149782e-06, + "loss": 0.1529, + "step": 69309 + }, + { + "epoch": 2.66, + "learning_rate": 5.591177224929469e-06, + "loss": 0.197, + "step": 69310 + }, + { + "epoch": 2.66, + "learning_rate": 5.590536034880739e-06, + "loss": 0.2453, + "step": 69311 + }, + { + "epoch": 2.66, + "learning_rate": 5.589894844832008e-06, + "loss": 0.1373, + "step": 69312 + }, + { + "epoch": 2.66, + "learning_rate": 5.589253654783278e-06, + "loss": 0.2344, + "step": 69313 + }, + { + "epoch": 2.66, + "learning_rate": 5.5886124647345475e-06, + "loss": 0.1779, + "step": 69314 + }, + { + "epoch": 2.66, + "learning_rate": 5.5879712746858175e-06, + "loss": 0.2058, + "step": 69315 + }, + { + "epoch": 2.66, + "learning_rate": 5.587330084637087e-06, + "loss": 0.1802, + "step": 69316 + }, + { + "epoch": 2.67, + "learning_rate": 5.586688894588357e-06, + "loss": 0.1965, + "step": 69317 + }, + { + "epoch": 2.67, + "learning_rate": 5.586047704539626e-06, + "loss": 0.2406, + "step": 69318 + }, + { + "epoch": 2.67, + "learning_rate": 5.585406514490896e-06, + "loss": 0.4339, + "step": 69319 + }, + { + "epoch": 2.67, + "learning_rate": 5.584765324442165e-06, + "loss": 0.2681, + "step": 69320 + }, + { + "epoch": 2.67, + "learning_rate": 5.584124134393434e-06, + "loss": 0.1976, + "step": 69321 + }, + { + "epoch": 2.67, + "learning_rate": 5.583482944344704e-06, + "loss": 0.1434, + "step": 69322 + }, + { + "epoch": 2.67, + "learning_rate": 5.582841754295973e-06, + "loss": 0.2744, + "step": 69323 + }, + { + "epoch": 2.67, + "learning_rate": 5.5822005642472434e-06, + "loss": 0.2743, + "step": 69324 + }, + { + "epoch": 2.67, + "learning_rate": 5.581559374198513e-06, + "loss": 0.1904, + "step": 69325 + }, + { + "epoch": 2.67, + "learning_rate": 5.580918184149783e-06, + "loss": 0.3297, + "step": 69326 + }, + { + "epoch": 2.67, + "learning_rate": 5.580276994101052e-06, + "loss": 0.2305, + "step": 69327 + }, + { + "epoch": 2.67, + "learning_rate": 5.579635804052322e-06, + "loss": 0.259, + "step": 69328 + }, + { + "epoch": 2.67, + "learning_rate": 5.578994614003591e-06, + "loss": 0.1644, + "step": 69329 + }, + { + "epoch": 2.67, + "learning_rate": 5.57835342395486e-06, + "loss": 0.3054, + "step": 69330 + }, + { + "epoch": 2.67, + "learning_rate": 5.57771223390613e-06, + "loss": 0.3252, + "step": 69331 + }, + { + "epoch": 2.67, + "learning_rate": 5.577071043857399e-06, + "loss": 0.2691, + "step": 69332 + }, + { + "epoch": 2.67, + "learning_rate": 5.576429853808669e-06, + "loss": 0.1856, + "step": 69333 + }, + { + "epoch": 2.67, + "learning_rate": 5.575788663759939e-06, + "loss": 0.1572, + "step": 69334 + }, + { + "epoch": 2.67, + "learning_rate": 5.575147473711209e-06, + "loss": 0.169, + "step": 69335 + }, + { + "epoch": 2.67, + "learning_rate": 5.574506283662478e-06, + "loss": 0.2164, + "step": 69336 + }, + { + "epoch": 2.67, + "learning_rate": 5.573865093613748e-06, + "loss": 0.2165, + "step": 69337 + }, + { + "epoch": 2.67, + "learning_rate": 5.573223903565017e-06, + "loss": 0.4191, + "step": 69338 + }, + { + "epoch": 2.67, + "learning_rate": 5.572582713516287e-06, + "loss": 0.2157, + "step": 69339 + }, + { + "epoch": 2.67, + "learning_rate": 5.571941523467556e-06, + "loss": 0.1558, + "step": 69340 + }, + { + "epoch": 2.67, + "learning_rate": 5.571300333418825e-06, + "loss": 0.1804, + "step": 69341 + }, + { + "epoch": 2.67, + "learning_rate": 5.570659143370095e-06, + "loss": 0.2333, + "step": 69342 + }, + { + "epoch": 2.67, + "learning_rate": 5.5700179533213646e-06, + "loss": 0.2446, + "step": 69343 + }, + { + "epoch": 2.67, + "learning_rate": 5.569376763272635e-06, + "loss": 0.2967, + "step": 69344 + }, + { + "epoch": 2.67, + "learning_rate": 5.568735573223904e-06, + "loss": 0.2177, + "step": 69345 + }, + { + "epoch": 2.67, + "learning_rate": 5.568094383175174e-06, + "loss": 0.1324, + "step": 69346 + }, + { + "epoch": 2.67, + "learning_rate": 5.567453193126443e-06, + "loss": 0.0707, + "step": 69347 + }, + { + "epoch": 2.67, + "learning_rate": 5.566812003077713e-06, + "loss": 0.2672, + "step": 69348 + }, + { + "epoch": 2.67, + "learning_rate": 5.566170813028982e-06, + "loss": 0.2164, + "step": 69349 + }, + { + "epoch": 2.67, + "learning_rate": 5.565529622980251e-06, + "loss": 0.3434, + "step": 69350 + }, + { + "epoch": 2.67, + "learning_rate": 5.564888432931521e-06, + "loss": 0.0826, + "step": 69351 + }, + { + "epoch": 2.67, + "learning_rate": 5.5642472428827905e-06, + "loss": 0.2058, + "step": 69352 + }, + { + "epoch": 2.67, + "learning_rate": 5.5636060528340605e-06, + "loss": 0.1371, + "step": 69353 + }, + { + "epoch": 2.67, + "learning_rate": 5.56296486278533e-06, + "loss": 0.1128, + "step": 69354 + }, + { + "epoch": 2.67, + "learning_rate": 5.5623236727366e-06, + "loss": 0.1664, + "step": 69355 + }, + { + "epoch": 2.67, + "learning_rate": 5.561682482687869e-06, + "loss": 0.2251, + "step": 69356 + }, + { + "epoch": 2.67, + "learning_rate": 5.561041292639139e-06, + "loss": 0.2405, + "step": 69357 + }, + { + "epoch": 2.67, + "learning_rate": 5.560400102590408e-06, + "loss": 0.2081, + "step": 69358 + }, + { + "epoch": 2.67, + "learning_rate": 5.559758912541678e-06, + "loss": 0.2618, + "step": 69359 + }, + { + "epoch": 2.67, + "learning_rate": 5.559117722492947e-06, + "loss": 0.226, + "step": 69360 + }, + { + "epoch": 2.67, + "learning_rate": 5.5584765324442165e-06, + "loss": 0.1336, + "step": 69361 + }, + { + "epoch": 2.67, + "learning_rate": 5.5578353423954865e-06, + "loss": 0.1213, + "step": 69362 + }, + { + "epoch": 2.67, + "learning_rate": 5.557194152346756e-06, + "loss": 0.2128, + "step": 69363 + }, + { + "epoch": 2.67, + "learning_rate": 5.556552962298026e-06, + "loss": 0.1983, + "step": 69364 + }, + { + "epoch": 2.67, + "learning_rate": 5.555911772249295e-06, + "loss": 0.2461, + "step": 69365 + }, + { + "epoch": 2.67, + "learning_rate": 5.555270582200565e-06, + "loss": 0.1872, + "step": 69366 + }, + { + "epoch": 2.67, + "learning_rate": 5.554629392151834e-06, + "loss": 0.1524, + "step": 69367 + }, + { + "epoch": 2.67, + "learning_rate": 5.553988202103104e-06, + "loss": 0.1725, + "step": 69368 + }, + { + "epoch": 2.67, + "learning_rate": 5.553347012054373e-06, + "loss": 0.2474, + "step": 69369 + }, + { + "epoch": 2.67, + "learning_rate": 5.5527058220056425e-06, + "loss": 0.2947, + "step": 69370 + }, + { + "epoch": 2.67, + "learning_rate": 5.5520646319569125e-06, + "loss": 0.2778, + "step": 69371 + }, + { + "epoch": 2.67, + "learning_rate": 5.551423441908182e-06, + "loss": 0.2348, + "step": 69372 + }, + { + "epoch": 2.67, + "learning_rate": 5.550782251859452e-06, + "loss": 0.2007, + "step": 69373 + }, + { + "epoch": 2.67, + "learning_rate": 5.550141061810721e-06, + "loss": 0.3211, + "step": 69374 + }, + { + "epoch": 2.67, + "learning_rate": 5.549499871761991e-06, + "loss": 0.2541, + "step": 69375 + }, + { + "epoch": 2.67, + "learning_rate": 5.54885868171326e-06, + "loss": 0.1761, + "step": 69376 + }, + { + "epoch": 2.67, + "learning_rate": 5.54821749166453e-06, + "loss": 0.4452, + "step": 69377 + }, + { + "epoch": 2.67, + "learning_rate": 5.547576301615799e-06, + "loss": 0.1611, + "step": 69378 + }, + { + "epoch": 2.67, + "learning_rate": 5.5469351115670684e-06, + "loss": 0.1395, + "step": 69379 + }, + { + "epoch": 2.67, + "learning_rate": 5.5462939215183384e-06, + "loss": 0.1477, + "step": 69380 + }, + { + "epoch": 2.67, + "learning_rate": 5.545652731469608e-06, + "loss": 0.1895, + "step": 69381 + }, + { + "epoch": 2.67, + "learning_rate": 5.545011541420878e-06, + "loss": 0.1387, + "step": 69382 + }, + { + "epoch": 2.67, + "learning_rate": 5.544370351372147e-06, + "loss": 0.1815, + "step": 69383 + }, + { + "epoch": 2.67, + "learning_rate": 5.543729161323417e-06, + "loss": 0.2962, + "step": 69384 + }, + { + "epoch": 2.67, + "learning_rate": 5.543087971274686e-06, + "loss": 0.1397, + "step": 69385 + }, + { + "epoch": 2.67, + "learning_rate": 5.542446781225956e-06, + "loss": 0.1276, + "step": 69386 + }, + { + "epoch": 2.67, + "learning_rate": 5.541805591177225e-06, + "loss": 0.0668, + "step": 69387 + }, + { + "epoch": 2.67, + "learning_rate": 5.541164401128495e-06, + "loss": 0.1375, + "step": 69388 + }, + { + "epoch": 2.67, + "learning_rate": 5.540523211079764e-06, + "loss": 0.2228, + "step": 69389 + }, + { + "epoch": 2.67, + "learning_rate": 5.539882021031034e-06, + "loss": 0.1789, + "step": 69390 + }, + { + "epoch": 2.67, + "learning_rate": 5.539240830982304e-06, + "loss": 0.1341, + "step": 69391 + }, + { + "epoch": 2.67, + "learning_rate": 5.538599640933573e-06, + "loss": 0.1412, + "step": 69392 + }, + { + "epoch": 2.67, + "learning_rate": 5.537958450884843e-06, + "loss": 0.2194, + "step": 69393 + }, + { + "epoch": 2.67, + "learning_rate": 5.537317260836112e-06, + "loss": 0.1924, + "step": 69394 + }, + { + "epoch": 2.67, + "learning_rate": 5.536676070787382e-06, + "loss": 0.3169, + "step": 69395 + }, + { + "epoch": 2.67, + "learning_rate": 5.536034880738651e-06, + "loss": 0.1704, + "step": 69396 + }, + { + "epoch": 2.67, + "learning_rate": 5.535393690689921e-06, + "loss": 0.1292, + "step": 69397 + }, + { + "epoch": 2.67, + "learning_rate": 5.53475250064119e-06, + "loss": 0.2234, + "step": 69398 + }, + { + "epoch": 2.67, + "learning_rate": 5.5341113105924596e-06, + "loss": 0.113, + "step": 69399 + }, + { + "epoch": 2.67, + "learning_rate": 5.53347012054373e-06, + "loss": 0.1332, + "step": 69400 + }, + { + "epoch": 2.67, + "learning_rate": 5.532828930494999e-06, + "loss": 0.1079, + "step": 69401 + }, + { + "epoch": 2.67, + "learning_rate": 5.532187740446269e-06, + "loss": 0.1767, + "step": 69402 + }, + { + "epoch": 2.67, + "learning_rate": 5.531546550397538e-06, + "loss": 0.1584, + "step": 69403 + }, + { + "epoch": 2.67, + "learning_rate": 5.530905360348808e-06, + "loss": 0.2393, + "step": 69404 + }, + { + "epoch": 2.67, + "learning_rate": 5.530264170300077e-06, + "loss": 0.1589, + "step": 69405 + }, + { + "epoch": 2.67, + "learning_rate": 5.529622980251347e-06, + "loss": 0.2641, + "step": 69406 + }, + { + "epoch": 2.67, + "learning_rate": 5.528981790202616e-06, + "loss": 0.1434, + "step": 69407 + }, + { + "epoch": 2.67, + "learning_rate": 5.528340600153886e-06, + "loss": 0.1024, + "step": 69408 + }, + { + "epoch": 2.67, + "learning_rate": 5.5276994101051556e-06, + "loss": 0.3292, + "step": 69409 + }, + { + "epoch": 2.67, + "learning_rate": 5.527058220056425e-06, + "loss": 0.1977, + "step": 69410 + }, + { + "epoch": 2.67, + "learning_rate": 5.526417030007695e-06, + "loss": 0.1861, + "step": 69411 + }, + { + "epoch": 2.67, + "learning_rate": 5.525775839958964e-06, + "loss": 0.298, + "step": 69412 + }, + { + "epoch": 2.67, + "learning_rate": 5.525134649910234e-06, + "loss": 0.1793, + "step": 69413 + }, + { + "epoch": 2.67, + "learning_rate": 5.524493459861503e-06, + "loss": 0.2583, + "step": 69414 + }, + { + "epoch": 2.67, + "learning_rate": 5.523852269812773e-06, + "loss": 0.1085, + "step": 69415 + }, + { + "epoch": 2.67, + "learning_rate": 5.523211079764042e-06, + "loss": 0.102, + "step": 69416 + }, + { + "epoch": 2.67, + "learning_rate": 5.522569889715312e-06, + "loss": 0.2481, + "step": 69417 + }, + { + "epoch": 2.67, + "learning_rate": 5.5219286996665815e-06, + "loss": 0.2189, + "step": 69418 + }, + { + "epoch": 2.67, + "learning_rate": 5.521287509617851e-06, + "loss": 0.2805, + "step": 69419 + }, + { + "epoch": 2.67, + "learning_rate": 5.520646319569121e-06, + "loss": 0.2699, + "step": 69420 + }, + { + "epoch": 2.67, + "learning_rate": 5.52000512952039e-06, + "loss": 0.1666, + "step": 69421 + }, + { + "epoch": 2.67, + "learning_rate": 5.51936393947166e-06, + "loss": 0.2221, + "step": 69422 + }, + { + "epoch": 2.67, + "learning_rate": 5.518722749422929e-06, + "loss": 0.1768, + "step": 69423 + }, + { + "epoch": 2.67, + "learning_rate": 5.518081559374199e-06, + "loss": 0.1403, + "step": 69424 + }, + { + "epoch": 2.67, + "learning_rate": 5.517440369325468e-06, + "loss": 0.1529, + "step": 69425 + }, + { + "epoch": 2.67, + "learning_rate": 5.516799179276738e-06, + "loss": 0.1096, + "step": 69426 + }, + { + "epoch": 2.67, + "learning_rate": 5.5161579892280075e-06, + "loss": 0.2969, + "step": 69427 + }, + { + "epoch": 2.67, + "learning_rate": 5.515516799179277e-06, + "loss": 0.1338, + "step": 69428 + }, + { + "epoch": 2.67, + "learning_rate": 5.514875609130547e-06, + "loss": 0.1387, + "step": 69429 + }, + { + "epoch": 2.67, + "learning_rate": 5.514234419081816e-06, + "loss": 0.2897, + "step": 69430 + }, + { + "epoch": 2.67, + "learning_rate": 5.513593229033086e-06, + "loss": 0.4237, + "step": 69431 + }, + { + "epoch": 2.67, + "learning_rate": 5.512952038984355e-06, + "loss": 0.2756, + "step": 69432 + }, + { + "epoch": 2.67, + "learning_rate": 5.512310848935625e-06, + "loss": 0.2922, + "step": 69433 + }, + { + "epoch": 2.67, + "learning_rate": 5.511669658886894e-06, + "loss": 0.1687, + "step": 69434 + }, + { + "epoch": 2.67, + "learning_rate": 5.511028468838164e-06, + "loss": 0.1369, + "step": 69435 + }, + { + "epoch": 2.67, + "learning_rate": 5.5103872787894335e-06, + "loss": 0.2571, + "step": 69436 + }, + { + "epoch": 2.67, + "learning_rate": 5.5097460887407035e-06, + "loss": 0.1961, + "step": 69437 + }, + { + "epoch": 2.67, + "learning_rate": 5.509104898691973e-06, + "loss": 0.4084, + "step": 69438 + }, + { + "epoch": 2.67, + "learning_rate": 5.508463708643242e-06, + "loss": 0.2919, + "step": 69439 + }, + { + "epoch": 2.67, + "learning_rate": 5.507822518594512e-06, + "loss": 0.2211, + "step": 69440 + }, + { + "epoch": 2.67, + "learning_rate": 5.507181328545781e-06, + "loss": 0.2363, + "step": 69441 + }, + { + "epoch": 2.67, + "learning_rate": 5.506540138497051e-06, + "loss": 0.164, + "step": 69442 + }, + { + "epoch": 2.67, + "learning_rate": 5.50589894844832e-06, + "loss": 0.3506, + "step": 69443 + }, + { + "epoch": 2.67, + "learning_rate": 5.50525775839959e-06, + "loss": 0.1177, + "step": 69444 + }, + { + "epoch": 2.67, + "learning_rate": 5.504616568350859e-06, + "loss": 0.1894, + "step": 69445 + }, + { + "epoch": 2.67, + "learning_rate": 5.5039753783021294e-06, + "loss": 0.294, + "step": 69446 + }, + { + "epoch": 2.67, + "learning_rate": 5.503334188253399e-06, + "loss": 0.2557, + "step": 69447 + }, + { + "epoch": 2.67, + "learning_rate": 5.502692998204668e-06, + "loss": 0.2632, + "step": 69448 + }, + { + "epoch": 2.67, + "learning_rate": 5.502051808155938e-06, + "loss": 0.3884, + "step": 69449 + }, + { + "epoch": 2.67, + "learning_rate": 5.501410618107207e-06, + "loss": 0.2088, + "step": 69450 + }, + { + "epoch": 2.67, + "learning_rate": 5.500769428058477e-06, + "loss": 0.1705, + "step": 69451 + }, + { + "epoch": 2.67, + "learning_rate": 5.500128238009746e-06, + "loss": 0.231, + "step": 69452 + }, + { + "epoch": 2.67, + "learning_rate": 5.499487047961016e-06, + "loss": 0.2189, + "step": 69453 + }, + { + "epoch": 2.67, + "learning_rate": 5.498845857912285e-06, + "loss": 0.1145, + "step": 69454 + }, + { + "epoch": 2.67, + "learning_rate": 5.498204667863555e-06, + "loss": 0.181, + "step": 69455 + }, + { + "epoch": 2.67, + "learning_rate": 5.497563477814825e-06, + "loss": 0.1827, + "step": 69456 + }, + { + "epoch": 2.67, + "learning_rate": 5.496922287766095e-06, + "loss": 0.1553, + "step": 69457 + }, + { + "epoch": 2.67, + "learning_rate": 5.496281097717364e-06, + "loss": 0.3507, + "step": 69458 + }, + { + "epoch": 2.67, + "learning_rate": 5.495639907668633e-06, + "loss": 0.2261, + "step": 69459 + }, + { + "epoch": 2.67, + "learning_rate": 5.494998717619903e-06, + "loss": 0.1698, + "step": 69460 + }, + { + "epoch": 2.67, + "learning_rate": 5.494357527571172e-06, + "loss": 0.1509, + "step": 69461 + }, + { + "epoch": 2.67, + "learning_rate": 5.493716337522442e-06, + "loss": 0.1965, + "step": 69462 + }, + { + "epoch": 2.67, + "learning_rate": 5.493075147473711e-06, + "loss": 0.2851, + "step": 69463 + }, + { + "epoch": 2.67, + "learning_rate": 5.492433957424981e-06, + "loss": 0.2062, + "step": 69464 + }, + { + "epoch": 2.67, + "learning_rate": 5.4917927673762506e-06, + "loss": 0.2606, + "step": 69465 + }, + { + "epoch": 2.67, + "learning_rate": 5.491151577327521e-06, + "loss": 0.3564, + "step": 69466 + }, + { + "epoch": 2.67, + "learning_rate": 5.49051038727879e-06, + "loss": 0.2193, + "step": 69467 + }, + { + "epoch": 2.67, + "learning_rate": 5.489869197230059e-06, + "loss": 0.1604, + "step": 69468 + }, + { + "epoch": 2.67, + "learning_rate": 5.489228007181329e-06, + "loss": 0.214, + "step": 69469 + }, + { + "epoch": 2.67, + "learning_rate": 5.488586817132598e-06, + "loss": 0.2684, + "step": 69470 + }, + { + "epoch": 2.67, + "learning_rate": 5.487945627083868e-06, + "loss": 0.2418, + "step": 69471 + }, + { + "epoch": 2.67, + "learning_rate": 5.487304437035137e-06, + "loss": 0.1708, + "step": 69472 + }, + { + "epoch": 2.67, + "learning_rate": 5.486663246986407e-06, + "loss": 0.2802, + "step": 69473 + }, + { + "epoch": 2.67, + "learning_rate": 5.4860220569376765e-06, + "loss": 0.2115, + "step": 69474 + }, + { + "epoch": 2.67, + "learning_rate": 5.4853808668889465e-06, + "loss": 0.4065, + "step": 69475 + }, + { + "epoch": 2.67, + "learning_rate": 5.484739676840216e-06, + "loss": 0.1092, + "step": 69476 + }, + { + "epoch": 2.67, + "learning_rate": 5.484098486791485e-06, + "loss": 0.1009, + "step": 69477 + }, + { + "epoch": 2.67, + "learning_rate": 5.483457296742755e-06, + "loss": 0.2457, + "step": 69478 + }, + { + "epoch": 2.67, + "learning_rate": 5.482816106694024e-06, + "loss": 0.2714, + "step": 69479 + }, + { + "epoch": 2.67, + "learning_rate": 5.482174916645294e-06, + "loss": 0.1385, + "step": 69480 + }, + { + "epoch": 2.67, + "learning_rate": 5.481533726596563e-06, + "loss": 0.2216, + "step": 69481 + }, + { + "epoch": 2.67, + "learning_rate": 5.480892536547833e-06, + "loss": 0.2187, + "step": 69482 + }, + { + "epoch": 2.67, + "learning_rate": 5.4802513464991025e-06, + "loss": 0.2049, + "step": 69483 + }, + { + "epoch": 2.67, + "learning_rate": 5.4796101564503725e-06, + "loss": 0.2926, + "step": 69484 + }, + { + "epoch": 2.67, + "learning_rate": 5.478968966401642e-06, + "loss": 0.3562, + "step": 69485 + }, + { + "epoch": 2.67, + "learning_rate": 5.478327776352912e-06, + "loss": 0.216, + "step": 69486 + }, + { + "epoch": 2.67, + "learning_rate": 5.477686586304181e-06, + "loss": 0.1126, + "step": 69487 + }, + { + "epoch": 2.67, + "learning_rate": 5.47704539625545e-06, + "loss": 0.2664, + "step": 69488 + }, + { + "epoch": 2.67, + "learning_rate": 5.47640420620672e-06, + "loss": 0.186, + "step": 69489 + }, + { + "epoch": 2.67, + "learning_rate": 5.475763016157989e-06, + "loss": 0.135, + "step": 69490 + }, + { + "epoch": 2.67, + "learning_rate": 5.475121826109259e-06, + "loss": 0.1833, + "step": 69491 + }, + { + "epoch": 2.67, + "learning_rate": 5.4744806360605285e-06, + "loss": 0.1305, + "step": 69492 + }, + { + "epoch": 2.67, + "learning_rate": 5.4738394460117985e-06, + "loss": 0.2254, + "step": 69493 + }, + { + "epoch": 2.67, + "learning_rate": 5.473198255963068e-06, + "loss": 0.1589, + "step": 69494 + }, + { + "epoch": 2.67, + "learning_rate": 5.472557065914338e-06, + "loss": 0.1857, + "step": 69495 + }, + { + "epoch": 2.67, + "learning_rate": 5.471915875865607e-06, + "loss": 0.154, + "step": 69496 + }, + { + "epoch": 2.67, + "learning_rate": 5.471274685816876e-06, + "loss": 0.1884, + "step": 69497 + }, + { + "epoch": 2.67, + "learning_rate": 5.470633495768146e-06, + "loss": 0.0856, + "step": 69498 + }, + { + "epoch": 2.67, + "learning_rate": 5.469992305719415e-06, + "loss": 0.2281, + "step": 69499 + }, + { + "epoch": 2.67, + "learning_rate": 5.469351115670685e-06, + "loss": 0.243, + "step": 69500 + }, + { + "epoch": 2.67, + "learning_rate": 5.4687099256219544e-06, + "loss": 0.2633, + "step": 69501 + }, + { + "epoch": 2.67, + "learning_rate": 5.4680687355732244e-06, + "loss": 0.2378, + "step": 69502 + }, + { + "epoch": 2.67, + "learning_rate": 5.467427545524494e-06, + "loss": 0.2016, + "step": 69503 + }, + { + "epoch": 2.67, + "learning_rate": 5.466786355475764e-06, + "loss": 0.1854, + "step": 69504 + }, + { + "epoch": 2.67, + "learning_rate": 5.466145165427033e-06, + "loss": 0.1979, + "step": 69505 + }, + { + "epoch": 2.67, + "learning_rate": 5.465503975378303e-06, + "loss": 0.2147, + "step": 69506 + }, + { + "epoch": 2.67, + "learning_rate": 5.464862785329572e-06, + "loss": 0.1107, + "step": 69507 + }, + { + "epoch": 2.67, + "learning_rate": 5.464221595280841e-06, + "loss": 0.2009, + "step": 69508 + }, + { + "epoch": 2.67, + "learning_rate": 5.463580405232111e-06, + "loss": 0.1298, + "step": 69509 + }, + { + "epoch": 2.67, + "learning_rate": 5.46293921518338e-06, + "loss": 0.1848, + "step": 69510 + }, + { + "epoch": 2.67, + "learning_rate": 5.46229802513465e-06, + "loss": 0.2616, + "step": 69511 + }, + { + "epoch": 2.67, + "learning_rate": 5.46165683508592e-06, + "loss": 0.1774, + "step": 69512 + }, + { + "epoch": 2.67, + "learning_rate": 5.46101564503719e-06, + "loss": 0.1992, + "step": 69513 + }, + { + "epoch": 2.67, + "learning_rate": 5.460374454988459e-06, + "loss": 0.3045, + "step": 69514 + }, + { + "epoch": 2.67, + "learning_rate": 5.459733264939729e-06, + "loss": 0.2148, + "step": 69515 + }, + { + "epoch": 2.67, + "learning_rate": 5.459092074890998e-06, + "loss": 0.2324, + "step": 69516 + }, + { + "epoch": 2.67, + "learning_rate": 5.458450884842267e-06, + "loss": 0.1931, + "step": 69517 + }, + { + "epoch": 2.67, + "learning_rate": 5.457809694793537e-06, + "loss": 0.0991, + "step": 69518 + }, + { + "epoch": 2.67, + "learning_rate": 5.457168504744806e-06, + "loss": 0.2026, + "step": 69519 + }, + { + "epoch": 2.67, + "learning_rate": 5.456527314696076e-06, + "loss": 0.1808, + "step": 69520 + }, + { + "epoch": 2.67, + "learning_rate": 5.4558861246473456e-06, + "loss": 0.2223, + "step": 69521 + }, + { + "epoch": 2.67, + "learning_rate": 5.455244934598616e-06, + "loss": 0.2485, + "step": 69522 + }, + { + "epoch": 2.67, + "learning_rate": 5.454603744549885e-06, + "loss": 0.2117, + "step": 69523 + }, + { + "epoch": 2.67, + "learning_rate": 5.453962554501155e-06, + "loss": 0.1962, + "step": 69524 + }, + { + "epoch": 2.67, + "learning_rate": 5.453321364452424e-06, + "loss": 0.3075, + "step": 69525 + }, + { + "epoch": 2.67, + "learning_rate": 5.452680174403694e-06, + "loss": 0.1192, + "step": 69526 + }, + { + "epoch": 2.67, + "learning_rate": 5.452038984354963e-06, + "loss": 0.204, + "step": 69527 + }, + { + "epoch": 2.67, + "learning_rate": 5.451397794306232e-06, + "loss": 0.132, + "step": 69528 + }, + { + "epoch": 2.67, + "learning_rate": 5.450756604257502e-06, + "loss": 0.2336, + "step": 69529 + }, + { + "epoch": 2.67, + "learning_rate": 5.4501154142087715e-06, + "loss": 0.2522, + "step": 69530 + }, + { + "epoch": 2.67, + "learning_rate": 5.4494742241600416e-06, + "loss": 0.3672, + "step": 69531 + }, + { + "epoch": 2.67, + "learning_rate": 5.448833034111311e-06, + "loss": 0.2012, + "step": 69532 + }, + { + "epoch": 2.67, + "learning_rate": 5.448191844062581e-06, + "loss": 0.2041, + "step": 69533 + }, + { + "epoch": 2.67, + "learning_rate": 5.44755065401385e-06, + "loss": 0.2409, + "step": 69534 + }, + { + "epoch": 2.67, + "learning_rate": 5.44690946396512e-06, + "loss": 0.1801, + "step": 69535 + }, + { + "epoch": 2.67, + "learning_rate": 5.446268273916389e-06, + "loss": 0.2615, + "step": 69536 + }, + { + "epoch": 2.67, + "learning_rate": 5.445627083867658e-06, + "loss": 0.1834, + "step": 69537 + }, + { + "epoch": 2.67, + "learning_rate": 5.444985893818928e-06, + "loss": 0.1002, + "step": 69538 + }, + { + "epoch": 2.67, + "learning_rate": 5.4443447037701975e-06, + "loss": 0.0749, + "step": 69539 + }, + { + "epoch": 2.67, + "learning_rate": 5.4437035137214675e-06, + "loss": 0.2074, + "step": 69540 + }, + { + "epoch": 2.67, + "learning_rate": 5.443062323672737e-06, + "loss": 0.149, + "step": 69541 + }, + { + "epoch": 2.67, + "learning_rate": 5.442421133624007e-06, + "loss": 0.1783, + "step": 69542 + }, + { + "epoch": 2.67, + "learning_rate": 5.441779943575276e-06, + "loss": 0.2497, + "step": 69543 + }, + { + "epoch": 2.67, + "learning_rate": 5.441138753526546e-06, + "loss": 0.1695, + "step": 69544 + }, + { + "epoch": 2.67, + "learning_rate": 5.440497563477815e-06, + "loss": 0.281, + "step": 69545 + }, + { + "epoch": 2.67, + "learning_rate": 5.439856373429084e-06, + "loss": 0.2446, + "step": 69546 + }, + { + "epoch": 2.67, + "learning_rate": 5.439215183380354e-06, + "loss": 0.2404, + "step": 69547 + }, + { + "epoch": 2.67, + "learning_rate": 5.4385739933316235e-06, + "loss": 0.3466, + "step": 69548 + }, + { + "epoch": 2.67, + "learning_rate": 5.4379328032828935e-06, + "loss": 0.2535, + "step": 69549 + }, + { + "epoch": 2.67, + "learning_rate": 5.437291613234163e-06, + "loss": 0.1455, + "step": 69550 + }, + { + "epoch": 2.67, + "learning_rate": 5.436650423185433e-06, + "loss": 0.2529, + "step": 69551 + }, + { + "epoch": 2.67, + "learning_rate": 5.436009233136702e-06, + "loss": 0.2717, + "step": 69552 + }, + { + "epoch": 2.67, + "learning_rate": 5.435368043087972e-06, + "loss": 0.2506, + "step": 69553 + }, + { + "epoch": 2.67, + "learning_rate": 5.434726853039241e-06, + "loss": 0.1978, + "step": 69554 + }, + { + "epoch": 2.67, + "learning_rate": 5.434085662990511e-06, + "loss": 0.1431, + "step": 69555 + }, + { + "epoch": 2.67, + "learning_rate": 5.43344447294178e-06, + "loss": 0.4683, + "step": 69556 + }, + { + "epoch": 2.67, + "learning_rate": 5.4328032828930494e-06, + "loss": 0.164, + "step": 69557 + }, + { + "epoch": 2.67, + "learning_rate": 5.4321620928443195e-06, + "loss": 0.1621, + "step": 69558 + }, + { + "epoch": 2.67, + "learning_rate": 5.431520902795589e-06, + "loss": 0.2999, + "step": 69559 + }, + { + "epoch": 2.67, + "learning_rate": 5.430879712746859e-06, + "loss": 0.1462, + "step": 69560 + }, + { + "epoch": 2.67, + "learning_rate": 5.430238522698128e-06, + "loss": 0.1946, + "step": 69561 + }, + { + "epoch": 2.67, + "learning_rate": 5.429597332649398e-06, + "loss": 0.2475, + "step": 69562 + }, + { + "epoch": 2.67, + "learning_rate": 5.428956142600667e-06, + "loss": 0.1209, + "step": 69563 + }, + { + "epoch": 2.67, + "learning_rate": 5.428314952551937e-06, + "loss": 0.2626, + "step": 69564 + }, + { + "epoch": 2.67, + "learning_rate": 5.427673762503206e-06, + "loss": 0.2748, + "step": 69565 + }, + { + "epoch": 2.67, + "learning_rate": 5.427032572454475e-06, + "loss": 0.1076, + "step": 69566 + }, + { + "epoch": 2.67, + "learning_rate": 5.426391382405745e-06, + "loss": 0.2956, + "step": 69567 + }, + { + "epoch": 2.67, + "learning_rate": 5.425750192357015e-06, + "loss": 0.094, + "step": 69568 + }, + { + "epoch": 2.67, + "learning_rate": 5.425109002308285e-06, + "loss": 0.3404, + "step": 69569 + }, + { + "epoch": 2.67, + "learning_rate": 5.424467812259554e-06, + "loss": 0.2989, + "step": 69570 + }, + { + "epoch": 2.67, + "learning_rate": 5.423826622210824e-06, + "loss": 0.0884, + "step": 69571 + }, + { + "epoch": 2.67, + "learning_rate": 5.423185432162093e-06, + "loss": 0.2306, + "step": 69572 + }, + { + "epoch": 2.67, + "learning_rate": 5.422544242113363e-06, + "loss": 0.1656, + "step": 69573 + }, + { + "epoch": 2.67, + "learning_rate": 5.421903052064632e-06, + "loss": 0.2287, + "step": 69574 + }, + { + "epoch": 2.67, + "learning_rate": 5.421261862015902e-06, + "loss": 0.1246, + "step": 69575 + }, + { + "epoch": 2.67, + "learning_rate": 5.420620671967171e-06, + "loss": 0.1851, + "step": 69576 + }, + { + "epoch": 2.68, + "learning_rate": 5.4199794819184406e-06, + "loss": 0.2322, + "step": 69577 + }, + { + "epoch": 2.68, + "learning_rate": 5.419338291869711e-06, + "loss": 0.2204, + "step": 69578 + }, + { + "epoch": 2.68, + "learning_rate": 5.41869710182098e-06, + "loss": 0.196, + "step": 69579 + }, + { + "epoch": 2.68, + "learning_rate": 5.41805591177225e-06, + "loss": 0.3393, + "step": 69580 + }, + { + "epoch": 2.68, + "learning_rate": 5.417414721723519e-06, + "loss": 0.2131, + "step": 69581 + }, + { + "epoch": 2.68, + "learning_rate": 5.416773531674789e-06, + "loss": 0.232, + "step": 69582 + }, + { + "epoch": 2.68, + "learning_rate": 5.416132341626058e-06, + "loss": 0.2643, + "step": 69583 + }, + { + "epoch": 2.68, + "learning_rate": 5.415491151577328e-06, + "loss": 0.2453, + "step": 69584 + }, + { + "epoch": 2.68, + "learning_rate": 5.414849961528597e-06, + "loss": 0.2959, + "step": 69585 + }, + { + "epoch": 2.68, + "learning_rate": 5.4142087714798665e-06, + "loss": 0.3615, + "step": 69586 + }, + { + "epoch": 2.68, + "learning_rate": 5.4135675814311366e-06, + "loss": 0.223, + "step": 69587 + }, + { + "epoch": 2.68, + "learning_rate": 5.412926391382406e-06, + "loss": 0.3874, + "step": 69588 + }, + { + "epoch": 2.68, + "learning_rate": 5.412285201333676e-06, + "loss": 0.3949, + "step": 69589 + }, + { + "epoch": 2.68, + "learning_rate": 5.411644011284945e-06, + "loss": 0.168, + "step": 69590 + }, + { + "epoch": 2.68, + "learning_rate": 5.411002821236215e-06, + "loss": 0.1359, + "step": 69591 + }, + { + "epoch": 2.68, + "learning_rate": 5.410361631187484e-06, + "loss": 0.2499, + "step": 69592 + }, + { + "epoch": 2.68, + "learning_rate": 5.409720441138754e-06, + "loss": 0.3043, + "step": 69593 + }, + { + "epoch": 2.68, + "learning_rate": 5.409079251090023e-06, + "loss": 0.1299, + "step": 69594 + }, + { + "epoch": 2.68, + "learning_rate": 5.4084380610412925e-06, + "loss": 0.2623, + "step": 69595 + }, + { + "epoch": 2.68, + "learning_rate": 5.4077968709925625e-06, + "loss": 0.192, + "step": 69596 + }, + { + "epoch": 2.68, + "learning_rate": 5.407155680943832e-06, + "loss": 0.2274, + "step": 69597 + }, + { + "epoch": 2.68, + "learning_rate": 5.406514490895102e-06, + "loss": 0.194, + "step": 69598 + }, + { + "epoch": 2.68, + "learning_rate": 5.405873300846371e-06, + "loss": 0.2441, + "step": 69599 + }, + { + "epoch": 2.68, + "learning_rate": 5.405232110797641e-06, + "loss": 0.3374, + "step": 69600 + }, + { + "epoch": 2.68, + "learning_rate": 5.40459092074891e-06, + "loss": 0.3577, + "step": 69601 + }, + { + "epoch": 2.68, + "learning_rate": 5.40394973070018e-06, + "loss": 0.1179, + "step": 69602 + }, + { + "epoch": 2.68, + "learning_rate": 5.403308540651449e-06, + "loss": 0.1961, + "step": 69603 + }, + { + "epoch": 2.68, + "learning_rate": 5.402667350602719e-06, + "loss": 0.1677, + "step": 69604 + }, + { + "epoch": 2.68, + "learning_rate": 5.4020261605539885e-06, + "loss": 0.1028, + "step": 69605 + }, + { + "epoch": 2.68, + "learning_rate": 5.401384970505258e-06, + "loss": 0.0889, + "step": 69606 + }, + { + "epoch": 2.68, + "learning_rate": 5.400743780456528e-06, + "loss": 0.1477, + "step": 69607 + }, + { + "epoch": 2.68, + "learning_rate": 5.400102590407797e-06, + "loss": 0.1296, + "step": 69608 + }, + { + "epoch": 2.68, + "learning_rate": 5.399461400359067e-06, + "loss": 0.183, + "step": 69609 + }, + { + "epoch": 2.68, + "learning_rate": 5.398820210310336e-06, + "loss": 0.2163, + "step": 69610 + }, + { + "epoch": 2.68, + "learning_rate": 5.398179020261606e-06, + "loss": 0.2185, + "step": 69611 + }, + { + "epoch": 2.68, + "learning_rate": 5.397537830212875e-06, + "loss": 0.2538, + "step": 69612 + }, + { + "epoch": 2.68, + "learning_rate": 5.396896640164145e-06, + "loss": 0.365, + "step": 69613 + }, + { + "epoch": 2.68, + "learning_rate": 5.3962554501154145e-06, + "loss": 0.185, + "step": 69614 + }, + { + "epoch": 2.68, + "learning_rate": 5.395614260066684e-06, + "loss": 0.1289, + "step": 69615 + }, + { + "epoch": 2.68, + "learning_rate": 5.394973070017954e-06, + "loss": 0.3536, + "step": 69616 + }, + { + "epoch": 2.68, + "learning_rate": 5.394331879969223e-06, + "loss": 0.4731, + "step": 69617 + }, + { + "epoch": 2.68, + "learning_rate": 5.393690689920493e-06, + "loss": 0.1751, + "step": 69618 + }, + { + "epoch": 2.68, + "learning_rate": 5.393049499871762e-06, + "loss": 0.3453, + "step": 69619 + }, + { + "epoch": 2.68, + "learning_rate": 5.392408309823032e-06, + "loss": 0.1869, + "step": 69620 + }, + { + "epoch": 2.68, + "learning_rate": 5.391767119774301e-06, + "loss": 0.1724, + "step": 69621 + }, + { + "epoch": 2.68, + "learning_rate": 5.391125929725571e-06, + "loss": 0.2875, + "step": 69622 + }, + { + "epoch": 2.68, + "learning_rate": 5.3904847396768404e-06, + "loss": 0.218, + "step": 69623 + }, + { + "epoch": 2.68, + "learning_rate": 5.3898435496281105e-06, + "loss": 0.2124, + "step": 69624 + }, + { + "epoch": 2.68, + "learning_rate": 5.38920235957938e-06, + "loss": 0.1875, + "step": 69625 + }, + { + "epoch": 2.68, + "learning_rate": 5.388561169530649e-06, + "loss": 0.1755, + "step": 69626 + }, + { + "epoch": 2.68, + "learning_rate": 5.387919979481919e-06, + "loss": 0.2195, + "step": 69627 + }, + { + "epoch": 2.68, + "learning_rate": 5.387278789433188e-06, + "loss": 0.1784, + "step": 69628 + }, + { + "epoch": 2.68, + "learning_rate": 5.386637599384458e-06, + "loss": 0.2331, + "step": 69629 + }, + { + "epoch": 2.68, + "learning_rate": 5.385996409335727e-06, + "loss": 0.1341, + "step": 69630 + }, + { + "epoch": 2.68, + "learning_rate": 5.385355219286997e-06, + "loss": 0.1231, + "step": 69631 + }, + { + "epoch": 2.68, + "learning_rate": 5.384714029238266e-06, + "loss": 0.2235, + "step": 69632 + }, + { + "epoch": 2.68, + "learning_rate": 5.384072839189536e-06, + "loss": 0.2478, + "step": 69633 + }, + { + "epoch": 2.68, + "learning_rate": 5.383431649140806e-06, + "loss": 0.2571, + "step": 69634 + }, + { + "epoch": 2.68, + "learning_rate": 5.382790459092075e-06, + "loss": 0.3152, + "step": 69635 + }, + { + "epoch": 2.68, + "learning_rate": 5.382149269043345e-06, + "loss": 0.2987, + "step": 69636 + }, + { + "epoch": 2.68, + "learning_rate": 5.381508078994614e-06, + "loss": 0.206, + "step": 69637 + }, + { + "epoch": 2.68, + "learning_rate": 5.380866888945884e-06, + "loss": 0.1517, + "step": 69638 + }, + { + "epoch": 2.68, + "learning_rate": 5.380225698897153e-06, + "loss": 0.3012, + "step": 69639 + }, + { + "epoch": 2.68, + "learning_rate": 5.379584508848423e-06, + "loss": 0.1627, + "step": 69640 + }, + { + "epoch": 2.68, + "learning_rate": 5.378943318799692e-06, + "loss": 0.3992, + "step": 69641 + }, + { + "epoch": 2.68, + "learning_rate": 5.378302128750962e-06, + "loss": 0.1612, + "step": 69642 + }, + { + "epoch": 2.68, + "learning_rate": 5.3776609387022316e-06, + "loss": 0.1474, + "step": 69643 + }, + { + "epoch": 2.68, + "learning_rate": 5.377019748653502e-06, + "loss": 0.3207, + "step": 69644 + }, + { + "epoch": 2.68, + "learning_rate": 5.376378558604771e-06, + "loss": 0.1605, + "step": 69645 + }, + { + "epoch": 2.68, + "learning_rate": 5.37573736855604e-06, + "loss": 0.1561, + "step": 69646 + }, + { + "epoch": 2.68, + "learning_rate": 5.37509617850731e-06, + "loss": 0.2534, + "step": 69647 + }, + { + "epoch": 2.68, + "learning_rate": 5.374454988458579e-06, + "loss": 0.2188, + "step": 69648 + }, + { + "epoch": 2.68, + "learning_rate": 5.373813798409849e-06, + "loss": 0.2234, + "step": 69649 + }, + { + "epoch": 2.68, + "learning_rate": 5.373172608361118e-06, + "loss": 0.2075, + "step": 69650 + }, + { + "epoch": 2.68, + "learning_rate": 5.372531418312388e-06, + "loss": 0.2783, + "step": 69651 + }, + { + "epoch": 2.68, + "learning_rate": 5.3718902282636575e-06, + "loss": 0.1818, + "step": 69652 + }, + { + "epoch": 2.68, + "learning_rate": 5.3712490382149276e-06, + "loss": 0.2541, + "step": 69653 + }, + { + "epoch": 2.68, + "learning_rate": 5.370607848166197e-06, + "loss": 0.2238, + "step": 69654 + }, + { + "epoch": 2.68, + "learning_rate": 5.369966658117466e-06, + "loss": 0.1644, + "step": 69655 + }, + { + "epoch": 2.68, + "learning_rate": 5.369325468068736e-06, + "loss": 0.1733, + "step": 69656 + }, + { + "epoch": 2.68, + "learning_rate": 5.368684278020005e-06, + "loss": 0.1602, + "step": 69657 + }, + { + "epoch": 2.68, + "learning_rate": 5.368043087971275e-06, + "loss": 0.2058, + "step": 69658 + }, + { + "epoch": 2.68, + "learning_rate": 5.367401897922544e-06, + "loss": 0.1695, + "step": 69659 + }, + { + "epoch": 2.68, + "learning_rate": 5.366760707873814e-06, + "loss": 0.2511, + "step": 69660 + }, + { + "epoch": 2.68, + "learning_rate": 5.3661195178250835e-06, + "loss": 0.1284, + "step": 69661 + }, + { + "epoch": 2.68, + "learning_rate": 5.3654783277763535e-06, + "loss": 0.2476, + "step": 69662 + }, + { + "epoch": 2.68, + "learning_rate": 5.364837137727623e-06, + "loss": 0.2695, + "step": 69663 + }, + { + "epoch": 2.68, + "learning_rate": 5.364195947678892e-06, + "loss": 0.2827, + "step": 69664 + }, + { + "epoch": 2.68, + "learning_rate": 5.363554757630162e-06, + "loss": 0.1901, + "step": 69665 + }, + { + "epoch": 2.68, + "learning_rate": 5.362913567581431e-06, + "loss": 0.2305, + "step": 69666 + }, + { + "epoch": 2.68, + "learning_rate": 5.362272377532701e-06, + "loss": 0.2191, + "step": 69667 + }, + { + "epoch": 2.68, + "learning_rate": 5.36163118748397e-06, + "loss": 0.1675, + "step": 69668 + }, + { + "epoch": 2.68, + "learning_rate": 5.36098999743524e-06, + "loss": 0.2559, + "step": 69669 + }, + { + "epoch": 2.68, + "learning_rate": 5.3603488073865095e-06, + "loss": 0.1122, + "step": 69670 + }, + { + "epoch": 2.68, + "learning_rate": 5.3597076173377795e-06, + "loss": 0.2164, + "step": 69671 + }, + { + "epoch": 2.68, + "learning_rate": 5.359066427289049e-06, + "loss": 0.1536, + "step": 69672 + }, + { + "epoch": 2.68, + "learning_rate": 5.358425237240319e-06, + "loss": 0.175, + "step": 69673 + }, + { + "epoch": 2.68, + "learning_rate": 5.357784047191588e-06, + "loss": 0.1857, + "step": 69674 + }, + { + "epoch": 2.68, + "learning_rate": 5.357142857142857e-06, + "loss": 0.3016, + "step": 69675 + }, + { + "epoch": 2.68, + "learning_rate": 5.356501667094127e-06, + "loss": 0.2198, + "step": 69676 + }, + { + "epoch": 2.68, + "learning_rate": 5.355860477045396e-06, + "loss": 0.1111, + "step": 69677 + }, + { + "epoch": 2.68, + "learning_rate": 5.355219286996666e-06, + "loss": 0.2725, + "step": 69678 + }, + { + "epoch": 2.68, + "learning_rate": 5.3545780969479354e-06, + "loss": 0.1207, + "step": 69679 + }, + { + "epoch": 2.68, + "learning_rate": 5.3539369068992055e-06, + "loss": 0.2309, + "step": 69680 + }, + { + "epoch": 2.68, + "learning_rate": 5.353295716850475e-06, + "loss": 0.1816, + "step": 69681 + }, + { + "epoch": 2.68, + "learning_rate": 5.352654526801745e-06, + "loss": 0.2297, + "step": 69682 + }, + { + "epoch": 2.68, + "learning_rate": 5.352013336753014e-06, + "loss": 0.2564, + "step": 69683 + }, + { + "epoch": 2.68, + "learning_rate": 5.351372146704283e-06, + "loss": 0.3541, + "step": 69684 + }, + { + "epoch": 2.68, + "learning_rate": 5.350730956655553e-06, + "loss": 0.1358, + "step": 69685 + }, + { + "epoch": 2.68, + "learning_rate": 5.350089766606822e-06, + "loss": 0.2056, + "step": 69686 + }, + { + "epoch": 2.68, + "learning_rate": 5.349448576558092e-06, + "loss": 0.2939, + "step": 69687 + }, + { + "epoch": 2.68, + "learning_rate": 5.348807386509361e-06, + "loss": 0.2448, + "step": 69688 + }, + { + "epoch": 2.68, + "learning_rate": 5.348166196460631e-06, + "loss": 0.0861, + "step": 69689 + }, + { + "epoch": 2.68, + "learning_rate": 5.347525006411901e-06, + "loss": 0.2189, + "step": 69690 + }, + { + "epoch": 2.68, + "learning_rate": 5.346883816363171e-06, + "loss": 0.1778, + "step": 69691 + }, + { + "epoch": 2.68, + "learning_rate": 5.34624262631444e-06, + "loss": 0.2284, + "step": 69692 + }, + { + "epoch": 2.68, + "learning_rate": 5.34560143626571e-06, + "loss": 0.1486, + "step": 69693 + }, + { + "epoch": 2.68, + "learning_rate": 5.344960246216979e-06, + "loss": 0.1742, + "step": 69694 + }, + { + "epoch": 2.68, + "learning_rate": 5.344319056168248e-06, + "loss": 0.2365, + "step": 69695 + }, + { + "epoch": 2.68, + "learning_rate": 5.343677866119518e-06, + "loss": 0.2766, + "step": 69696 + }, + { + "epoch": 2.68, + "learning_rate": 5.343036676070787e-06, + "loss": 0.1625, + "step": 69697 + }, + { + "epoch": 2.68, + "learning_rate": 5.342395486022057e-06, + "loss": 0.2692, + "step": 69698 + }, + { + "epoch": 2.68, + "learning_rate": 5.3417542959733266e-06, + "loss": 0.1362, + "step": 69699 + }, + { + "epoch": 2.68, + "learning_rate": 5.341113105924597e-06, + "loss": 0.2198, + "step": 69700 + }, + { + "epoch": 2.68, + "learning_rate": 5.340471915875866e-06, + "loss": 0.1686, + "step": 69701 + }, + { + "epoch": 2.68, + "learning_rate": 5.339830725827136e-06, + "loss": 0.2219, + "step": 69702 + }, + { + "epoch": 2.68, + "learning_rate": 5.339189535778405e-06, + "loss": 0.2817, + "step": 69703 + }, + { + "epoch": 2.68, + "learning_rate": 5.338548345729674e-06, + "loss": 0.2925, + "step": 69704 + }, + { + "epoch": 2.68, + "learning_rate": 5.337907155680944e-06, + "loss": 0.2612, + "step": 69705 + }, + { + "epoch": 2.68, + "learning_rate": 5.337265965632213e-06, + "loss": 0.146, + "step": 69706 + }, + { + "epoch": 2.68, + "learning_rate": 5.336624775583483e-06, + "loss": 0.2659, + "step": 69707 + }, + { + "epoch": 2.68, + "learning_rate": 5.3359835855347525e-06, + "loss": 0.2048, + "step": 69708 + }, + { + "epoch": 2.68, + "learning_rate": 5.3353423954860226e-06, + "loss": 0.1532, + "step": 69709 + }, + { + "epoch": 2.68, + "learning_rate": 5.334701205437292e-06, + "loss": 0.3004, + "step": 69710 + }, + { + "epoch": 2.68, + "learning_rate": 5.334060015388562e-06, + "loss": 0.2501, + "step": 69711 + }, + { + "epoch": 2.68, + "learning_rate": 5.333418825339831e-06, + "loss": 0.3084, + "step": 69712 + }, + { + "epoch": 2.68, + "learning_rate": 5.3327776352911e-06, + "loss": 0.1608, + "step": 69713 + }, + { + "epoch": 2.68, + "learning_rate": 5.33213644524237e-06, + "loss": 0.2656, + "step": 69714 + }, + { + "epoch": 2.68, + "learning_rate": 5.331495255193639e-06, + "loss": 0.1636, + "step": 69715 + }, + { + "epoch": 2.68, + "learning_rate": 5.330854065144909e-06, + "loss": 0.2547, + "step": 69716 + }, + { + "epoch": 2.68, + "learning_rate": 5.3302128750961785e-06, + "loss": 0.1982, + "step": 69717 + }, + { + "epoch": 2.68, + "learning_rate": 5.3295716850474485e-06, + "loss": 0.1959, + "step": 69718 + }, + { + "epoch": 2.68, + "learning_rate": 5.328930494998718e-06, + "loss": 0.2541, + "step": 69719 + }, + { + "epoch": 2.68, + "learning_rate": 5.328289304949988e-06, + "loss": 0.167, + "step": 69720 + }, + { + "epoch": 2.68, + "learning_rate": 5.327648114901257e-06, + "loss": 0.1339, + "step": 69721 + }, + { + "epoch": 2.68, + "learning_rate": 5.327006924852527e-06, + "loss": 0.2738, + "step": 69722 + }, + { + "epoch": 2.68, + "learning_rate": 5.326365734803796e-06, + "loss": 0.2097, + "step": 69723 + }, + { + "epoch": 2.68, + "learning_rate": 5.325724544755065e-06, + "loss": 0.308, + "step": 69724 + }, + { + "epoch": 2.68, + "learning_rate": 5.325083354706335e-06, + "loss": 0.243, + "step": 69725 + }, + { + "epoch": 2.68, + "learning_rate": 5.3244421646576045e-06, + "loss": 0.3156, + "step": 69726 + }, + { + "epoch": 2.68, + "learning_rate": 5.3238009746088745e-06, + "loss": 0.1716, + "step": 69727 + }, + { + "epoch": 2.68, + "learning_rate": 5.323159784560144e-06, + "loss": 0.2172, + "step": 69728 + }, + { + "epoch": 2.68, + "learning_rate": 5.322518594511414e-06, + "loss": 0.1359, + "step": 69729 + }, + { + "epoch": 2.68, + "learning_rate": 5.321877404462683e-06, + "loss": 0.1338, + "step": 69730 + }, + { + "epoch": 2.68, + "learning_rate": 5.321236214413953e-06, + "loss": 0.1428, + "step": 69731 + }, + { + "epoch": 2.68, + "learning_rate": 5.320595024365222e-06, + "loss": 0.2341, + "step": 69732 + }, + { + "epoch": 2.68, + "learning_rate": 5.319953834316491e-06, + "loss": 0.3083, + "step": 69733 + }, + { + "epoch": 2.68, + "learning_rate": 5.319312644267761e-06, + "loss": 0.2219, + "step": 69734 + }, + { + "epoch": 2.68, + "learning_rate": 5.3186714542190304e-06, + "loss": 0.2403, + "step": 69735 + }, + { + "epoch": 2.68, + "learning_rate": 5.3180302641703005e-06, + "loss": 0.2053, + "step": 69736 + }, + { + "epoch": 2.68, + "learning_rate": 5.31738907412157e-06, + "loss": 0.2662, + "step": 69737 + }, + { + "epoch": 2.68, + "learning_rate": 5.31674788407284e-06, + "loss": 0.0682, + "step": 69738 + }, + { + "epoch": 2.68, + "learning_rate": 5.316106694024109e-06, + "loss": 0.1809, + "step": 69739 + }, + { + "epoch": 2.68, + "learning_rate": 5.315465503975379e-06, + "loss": 0.2007, + "step": 69740 + }, + { + "epoch": 2.68, + "learning_rate": 5.314824313926648e-06, + "loss": 0.2209, + "step": 69741 + }, + { + "epoch": 2.68, + "learning_rate": 5.314183123877918e-06, + "loss": 0.1455, + "step": 69742 + }, + { + "epoch": 2.68, + "learning_rate": 5.313541933829187e-06, + "loss": 0.2011, + "step": 69743 + }, + { + "epoch": 2.68, + "learning_rate": 5.312900743780456e-06, + "loss": 0.2564, + "step": 69744 + }, + { + "epoch": 2.68, + "learning_rate": 5.3122595537317264e-06, + "loss": 0.2234, + "step": 69745 + }, + { + "epoch": 2.68, + "learning_rate": 5.311618363682996e-06, + "loss": 0.1904, + "step": 69746 + }, + { + "epoch": 2.68, + "learning_rate": 5.310977173634266e-06, + "loss": 0.1789, + "step": 69747 + }, + { + "epoch": 2.68, + "learning_rate": 5.310335983585535e-06, + "loss": 0.4513, + "step": 69748 + }, + { + "epoch": 2.68, + "learning_rate": 5.309694793536805e-06, + "loss": 0.2488, + "step": 69749 + }, + { + "epoch": 2.68, + "learning_rate": 5.309053603488074e-06, + "loss": 0.1688, + "step": 69750 + }, + { + "epoch": 2.68, + "learning_rate": 5.308412413439344e-06, + "loss": 0.1799, + "step": 69751 + }, + { + "epoch": 2.68, + "learning_rate": 5.307771223390613e-06, + "loss": 0.2399, + "step": 69752 + }, + { + "epoch": 2.68, + "learning_rate": 5.307130033341882e-06, + "loss": 0.2687, + "step": 69753 + }, + { + "epoch": 2.68, + "learning_rate": 5.306488843293152e-06, + "loss": 0.2736, + "step": 69754 + }, + { + "epoch": 2.68, + "learning_rate": 5.3058476532444216e-06, + "loss": 0.1977, + "step": 69755 + }, + { + "epoch": 2.68, + "learning_rate": 5.305206463195692e-06, + "loss": 0.3195, + "step": 69756 + }, + { + "epoch": 2.68, + "learning_rate": 5.304565273146961e-06, + "loss": 0.187, + "step": 69757 + }, + { + "epoch": 2.68, + "learning_rate": 5.303924083098231e-06, + "loss": 0.3309, + "step": 69758 + }, + { + "epoch": 2.68, + "learning_rate": 5.3032828930495e-06, + "loss": 0.1873, + "step": 69759 + }, + { + "epoch": 2.68, + "learning_rate": 5.30264170300077e-06, + "loss": 0.1371, + "step": 69760 + }, + { + "epoch": 2.68, + "learning_rate": 5.302000512952039e-06, + "loss": 0.1347, + "step": 69761 + }, + { + "epoch": 2.68, + "learning_rate": 5.301359322903309e-06, + "loss": 0.1597, + "step": 69762 + }, + { + "epoch": 2.68, + "learning_rate": 5.300718132854578e-06, + "loss": 0.3134, + "step": 69763 + }, + { + "epoch": 2.68, + "learning_rate": 5.3000769428058475e-06, + "loss": 0.2264, + "step": 69764 + }, + { + "epoch": 2.68, + "learning_rate": 5.2994357527571176e-06, + "loss": 0.154, + "step": 69765 + }, + { + "epoch": 2.68, + "learning_rate": 5.298794562708387e-06, + "loss": 0.2042, + "step": 69766 + }, + { + "epoch": 2.68, + "learning_rate": 5.298153372659657e-06, + "loss": 0.3136, + "step": 69767 + }, + { + "epoch": 2.68, + "learning_rate": 5.297512182610926e-06, + "loss": 0.1825, + "step": 69768 + }, + { + "epoch": 2.68, + "learning_rate": 5.296870992562196e-06, + "loss": 0.1573, + "step": 69769 + }, + { + "epoch": 2.68, + "learning_rate": 5.296229802513465e-06, + "loss": 0.2496, + "step": 69770 + }, + { + "epoch": 2.68, + "learning_rate": 5.295588612464735e-06, + "loss": 0.4561, + "step": 69771 + }, + { + "epoch": 2.68, + "learning_rate": 5.294947422416004e-06, + "loss": 0.2791, + "step": 69772 + }, + { + "epoch": 2.68, + "learning_rate": 5.2943062323672735e-06, + "loss": 0.2553, + "step": 69773 + }, + { + "epoch": 2.68, + "learning_rate": 5.2936650423185435e-06, + "loss": 0.2817, + "step": 69774 + }, + { + "epoch": 2.68, + "learning_rate": 5.293023852269813e-06, + "loss": 0.2354, + "step": 69775 + }, + { + "epoch": 2.68, + "learning_rate": 5.292382662221083e-06, + "loss": 0.245, + "step": 69776 + }, + { + "epoch": 2.68, + "learning_rate": 5.291741472172352e-06, + "loss": 0.3151, + "step": 69777 + }, + { + "epoch": 2.68, + "learning_rate": 5.291100282123622e-06, + "loss": 0.1154, + "step": 69778 + }, + { + "epoch": 2.68, + "learning_rate": 5.290459092074891e-06, + "loss": 0.187, + "step": 69779 + }, + { + "epoch": 2.68, + "learning_rate": 5.289817902026161e-06, + "loss": 0.2031, + "step": 69780 + }, + { + "epoch": 2.68, + "learning_rate": 5.28917671197743e-06, + "loss": 0.1278, + "step": 69781 + }, + { + "epoch": 2.68, + "learning_rate": 5.2885355219286995e-06, + "loss": 0.2236, + "step": 69782 + }, + { + "epoch": 2.68, + "learning_rate": 5.2878943318799695e-06, + "loss": 0.3412, + "step": 69783 + }, + { + "epoch": 2.68, + "learning_rate": 5.287253141831239e-06, + "loss": 0.2322, + "step": 69784 + }, + { + "epoch": 2.68, + "learning_rate": 5.286611951782509e-06, + "loss": 0.2352, + "step": 69785 + }, + { + "epoch": 2.68, + "learning_rate": 5.285970761733778e-06, + "loss": 0.1705, + "step": 69786 + }, + { + "epoch": 2.68, + "learning_rate": 5.285329571685048e-06, + "loss": 0.0873, + "step": 69787 + }, + { + "epoch": 2.68, + "learning_rate": 5.284688381636317e-06, + "loss": 0.2538, + "step": 69788 + }, + { + "epoch": 2.68, + "learning_rate": 5.284047191587587e-06, + "loss": 0.1845, + "step": 69789 + }, + { + "epoch": 2.68, + "learning_rate": 5.283406001538856e-06, + "loss": 0.2668, + "step": 69790 + }, + { + "epoch": 2.68, + "learning_rate": 5.282764811490126e-06, + "loss": 0.1809, + "step": 69791 + }, + { + "epoch": 2.68, + "learning_rate": 5.2821236214413955e-06, + "loss": 0.2794, + "step": 69792 + }, + { + "epoch": 2.68, + "learning_rate": 5.281482431392665e-06, + "loss": 0.2368, + "step": 69793 + }, + { + "epoch": 2.68, + "learning_rate": 5.280841241343935e-06, + "loss": 0.215, + "step": 69794 + }, + { + "epoch": 2.68, + "learning_rate": 5.280200051295204e-06, + "loss": 0.1532, + "step": 69795 + }, + { + "epoch": 2.68, + "learning_rate": 5.279558861246474e-06, + "loss": 0.1869, + "step": 69796 + }, + { + "epoch": 2.68, + "learning_rate": 5.278917671197743e-06, + "loss": 0.1777, + "step": 69797 + }, + { + "epoch": 2.68, + "learning_rate": 5.278276481149013e-06, + "loss": 0.1899, + "step": 69798 + }, + { + "epoch": 2.68, + "learning_rate": 5.277635291100282e-06, + "loss": 0.185, + "step": 69799 + }, + { + "epoch": 2.68, + "learning_rate": 5.276994101051552e-06, + "loss": 0.2418, + "step": 69800 + }, + { + "epoch": 2.68, + "learning_rate": 5.2763529110028214e-06, + "loss": 0.1373, + "step": 69801 + }, + { + "epoch": 2.68, + "learning_rate": 5.275711720954091e-06, + "loss": 0.2729, + "step": 69802 + }, + { + "epoch": 2.68, + "learning_rate": 5.275070530905361e-06, + "loss": 0.2239, + "step": 69803 + }, + { + "epoch": 2.68, + "learning_rate": 5.27442934085663e-06, + "loss": 0.1278, + "step": 69804 + }, + { + "epoch": 2.68, + "learning_rate": 5.2737881508079e-06, + "loss": 0.2321, + "step": 69805 + }, + { + "epoch": 2.68, + "learning_rate": 5.273146960759169e-06, + "loss": 0.1341, + "step": 69806 + }, + { + "epoch": 2.68, + "learning_rate": 5.272505770710439e-06, + "loss": 0.113, + "step": 69807 + }, + { + "epoch": 2.68, + "learning_rate": 5.271864580661708e-06, + "loss": 0.211, + "step": 69808 + }, + { + "epoch": 2.68, + "learning_rate": 5.271223390612978e-06, + "loss": 0.1594, + "step": 69809 + }, + { + "epoch": 2.68, + "learning_rate": 5.270582200564247e-06, + "loss": 0.3219, + "step": 69810 + }, + { + "epoch": 2.68, + "learning_rate": 5.2699410105155174e-06, + "loss": 0.1692, + "step": 69811 + }, + { + "epoch": 2.68, + "learning_rate": 5.269299820466787e-06, + "loss": 0.2272, + "step": 69812 + }, + { + "epoch": 2.68, + "learning_rate": 5.268658630418056e-06, + "loss": 0.3216, + "step": 69813 + }, + { + "epoch": 2.68, + "learning_rate": 5.268017440369326e-06, + "loss": 0.1295, + "step": 69814 + }, + { + "epoch": 2.68, + "learning_rate": 5.267376250320595e-06, + "loss": 0.1901, + "step": 69815 + }, + { + "epoch": 2.68, + "learning_rate": 5.266735060271865e-06, + "loss": 0.1575, + "step": 69816 + }, + { + "epoch": 2.68, + "learning_rate": 5.266093870223134e-06, + "loss": 0.1733, + "step": 69817 + }, + { + "epoch": 2.68, + "learning_rate": 5.265452680174404e-06, + "loss": 0.2629, + "step": 69818 + }, + { + "epoch": 2.68, + "learning_rate": 5.264811490125673e-06, + "loss": 0.3436, + "step": 69819 + }, + { + "epoch": 2.68, + "learning_rate": 5.264170300076943e-06, + "loss": 0.2154, + "step": 69820 + }, + { + "epoch": 2.68, + "learning_rate": 5.2635291100282126e-06, + "loss": 0.1894, + "step": 69821 + }, + { + "epoch": 2.68, + "learning_rate": 5.262887919979482e-06, + "loss": 0.2274, + "step": 69822 + }, + { + "epoch": 2.68, + "learning_rate": 5.262246729930752e-06, + "loss": 0.1613, + "step": 69823 + }, + { + "epoch": 2.68, + "learning_rate": 5.261605539882021e-06, + "loss": 0.3111, + "step": 69824 + }, + { + "epoch": 2.68, + "learning_rate": 5.260964349833291e-06, + "loss": 0.2442, + "step": 69825 + }, + { + "epoch": 2.68, + "learning_rate": 5.26032315978456e-06, + "loss": 0.1406, + "step": 69826 + }, + { + "epoch": 2.68, + "learning_rate": 5.25968196973583e-06, + "loss": 0.2616, + "step": 69827 + }, + { + "epoch": 2.68, + "learning_rate": 5.259040779687099e-06, + "loss": 0.2715, + "step": 69828 + }, + { + "epoch": 2.68, + "learning_rate": 5.258399589638369e-06, + "loss": 0.1601, + "step": 69829 + }, + { + "epoch": 2.68, + "learning_rate": 5.2577583995896385e-06, + "loss": 0.3898, + "step": 69830 + }, + { + "epoch": 2.68, + "learning_rate": 5.257117209540908e-06, + "loss": 0.2522, + "step": 69831 + }, + { + "epoch": 2.68, + "learning_rate": 5.256476019492178e-06, + "loss": 0.2084, + "step": 69832 + }, + { + "epoch": 2.68, + "learning_rate": 5.255834829443447e-06, + "loss": 0.2858, + "step": 69833 + }, + { + "epoch": 2.68, + "learning_rate": 5.255193639394717e-06, + "loss": 0.1312, + "step": 69834 + }, + { + "epoch": 2.68, + "learning_rate": 5.254552449345986e-06, + "loss": 0.2397, + "step": 69835 + }, + { + "epoch": 2.68, + "learning_rate": 5.253911259297256e-06, + "loss": 0.2819, + "step": 69836 + }, + { + "epoch": 2.69, + "learning_rate": 5.253270069248525e-06, + "loss": 0.1293, + "step": 69837 + }, + { + "epoch": 2.69, + "learning_rate": 5.252628879199795e-06, + "loss": 0.1671, + "step": 69838 + }, + { + "epoch": 2.69, + "learning_rate": 5.2519876891510645e-06, + "loss": 0.2357, + "step": 69839 + }, + { + "epoch": 2.69, + "learning_rate": 5.2513464991023345e-06, + "loss": 0.1915, + "step": 69840 + }, + { + "epoch": 2.69, + "learning_rate": 5.250705309053604e-06, + "loss": 0.1865, + "step": 69841 + }, + { + "epoch": 2.69, + "learning_rate": 5.250064119004873e-06, + "loss": 0.1637, + "step": 69842 + }, + { + "epoch": 2.69, + "learning_rate": 5.249422928956143e-06, + "loss": 0.1548, + "step": 69843 + }, + { + "epoch": 2.69, + "learning_rate": 5.248781738907412e-06, + "loss": 0.1407, + "step": 69844 + }, + { + "epoch": 2.69, + "learning_rate": 5.248140548858682e-06, + "loss": 0.2634, + "step": 69845 + }, + { + "epoch": 2.69, + "learning_rate": 5.247499358809951e-06, + "loss": 0.142, + "step": 69846 + }, + { + "epoch": 2.69, + "learning_rate": 5.246858168761221e-06, + "loss": 0.242, + "step": 69847 + }, + { + "epoch": 2.69, + "learning_rate": 5.2462169787124905e-06, + "loss": 0.1621, + "step": 69848 + }, + { + "epoch": 2.69, + "learning_rate": 5.2455757886637605e-06, + "loss": 0.3109, + "step": 69849 + }, + { + "epoch": 2.69, + "learning_rate": 5.24493459861503e-06, + "loss": 0.3733, + "step": 69850 + }, + { + "epoch": 2.69, + "learning_rate": 5.244293408566299e-06, + "loss": 0.2017, + "step": 69851 + }, + { + "epoch": 2.69, + "learning_rate": 5.243652218517569e-06, + "loss": 0.3076, + "step": 69852 + }, + { + "epoch": 2.69, + "learning_rate": 5.243011028468838e-06, + "loss": 0.1519, + "step": 69853 + }, + { + "epoch": 2.69, + "learning_rate": 5.242369838420108e-06, + "loss": 0.2491, + "step": 69854 + }, + { + "epoch": 2.69, + "learning_rate": 5.241728648371377e-06, + "loss": 0.1971, + "step": 69855 + }, + { + "epoch": 2.69, + "learning_rate": 5.241087458322647e-06, + "loss": 0.208, + "step": 69856 + }, + { + "epoch": 2.69, + "learning_rate": 5.2404462682739164e-06, + "loss": 0.2827, + "step": 69857 + }, + { + "epoch": 2.69, + "learning_rate": 5.2398050782251865e-06, + "loss": 0.1529, + "step": 69858 + }, + { + "epoch": 2.69, + "learning_rate": 5.239163888176456e-06, + "loss": 0.1281, + "step": 69859 + }, + { + "epoch": 2.69, + "learning_rate": 5.238522698127726e-06, + "loss": 0.3344, + "step": 69860 + }, + { + "epoch": 2.69, + "learning_rate": 5.237881508078995e-06, + "loss": 0.2057, + "step": 69861 + }, + { + "epoch": 2.69, + "learning_rate": 5.237240318030264e-06, + "loss": 0.1351, + "step": 69862 + }, + { + "epoch": 2.69, + "learning_rate": 5.236599127981534e-06, + "loss": 0.1537, + "step": 69863 + }, + { + "epoch": 2.69, + "learning_rate": 5.235957937932803e-06, + "loss": 0.1788, + "step": 69864 + }, + { + "epoch": 2.69, + "learning_rate": 5.235316747884073e-06, + "loss": 0.202, + "step": 69865 + }, + { + "epoch": 2.69, + "learning_rate": 5.234675557835342e-06, + "loss": 0.3211, + "step": 69866 + }, + { + "epoch": 2.69, + "learning_rate": 5.2340343677866124e-06, + "loss": 0.1587, + "step": 69867 + }, + { + "epoch": 2.69, + "learning_rate": 5.233393177737882e-06, + "loss": 0.2223, + "step": 69868 + }, + { + "epoch": 2.69, + "learning_rate": 5.232751987689152e-06, + "loss": 0.2531, + "step": 69869 + }, + { + "epoch": 2.69, + "learning_rate": 5.232110797640421e-06, + "loss": 0.2968, + "step": 69870 + }, + { + "epoch": 2.69, + "learning_rate": 5.23146960759169e-06, + "loss": 0.2404, + "step": 69871 + }, + { + "epoch": 2.69, + "learning_rate": 5.23082841754296e-06, + "loss": 0.2157, + "step": 69872 + }, + { + "epoch": 2.69, + "learning_rate": 5.230187227494229e-06, + "loss": 0.2598, + "step": 69873 + }, + { + "epoch": 2.69, + "learning_rate": 5.229546037445499e-06, + "loss": 0.1974, + "step": 69874 + }, + { + "epoch": 2.69, + "learning_rate": 5.228904847396768e-06, + "loss": 0.2708, + "step": 69875 + }, + { + "epoch": 2.69, + "learning_rate": 5.228263657348038e-06, + "loss": 0.2, + "step": 69876 + }, + { + "epoch": 2.69, + "learning_rate": 5.2276224672993076e-06, + "loss": 0.253, + "step": 69877 + }, + { + "epoch": 2.69, + "learning_rate": 5.226981277250578e-06, + "loss": 0.1805, + "step": 69878 + }, + { + "epoch": 2.69, + "learning_rate": 5.226340087201847e-06, + "loss": 0.1224, + "step": 69879 + }, + { + "epoch": 2.69, + "learning_rate": 5.225698897153117e-06, + "loss": 0.5079, + "step": 69880 + }, + { + "epoch": 2.69, + "learning_rate": 5.225057707104386e-06, + "loss": 0.1517, + "step": 69881 + }, + { + "epoch": 2.69, + "learning_rate": 5.224416517055655e-06, + "loss": 0.1322, + "step": 69882 + }, + { + "epoch": 2.69, + "learning_rate": 5.223775327006925e-06, + "loss": 0.1794, + "step": 69883 + }, + { + "epoch": 2.69, + "learning_rate": 5.223134136958194e-06, + "loss": 0.1986, + "step": 69884 + }, + { + "epoch": 2.69, + "learning_rate": 5.222492946909464e-06, + "loss": 0.1033, + "step": 69885 + }, + { + "epoch": 2.69, + "learning_rate": 5.2218517568607335e-06, + "loss": 0.3247, + "step": 69886 + }, + { + "epoch": 2.69, + "learning_rate": 5.2212105668120036e-06, + "loss": 0.2784, + "step": 69887 + }, + { + "epoch": 2.69, + "learning_rate": 5.220569376763273e-06, + "loss": 0.1209, + "step": 69888 + }, + { + "epoch": 2.69, + "learning_rate": 5.219928186714543e-06, + "loss": 0.1839, + "step": 69889 + }, + { + "epoch": 2.69, + "learning_rate": 5.219286996665812e-06, + "loss": 0.1219, + "step": 69890 + }, + { + "epoch": 2.69, + "learning_rate": 5.218645806617081e-06, + "loss": 0.2789, + "step": 69891 + }, + { + "epoch": 2.69, + "learning_rate": 5.218004616568351e-06, + "loss": 0.2369, + "step": 69892 + }, + { + "epoch": 2.69, + "learning_rate": 5.21736342651962e-06, + "loss": 0.1463, + "step": 69893 + }, + { + "epoch": 2.69, + "learning_rate": 5.21672223647089e-06, + "loss": 0.3109, + "step": 69894 + }, + { + "epoch": 2.69, + "learning_rate": 5.2160810464221595e-06, + "loss": 0.1758, + "step": 69895 + }, + { + "epoch": 2.69, + "learning_rate": 5.2154398563734295e-06, + "loss": 0.2515, + "step": 69896 + }, + { + "epoch": 2.69, + "learning_rate": 5.214798666324699e-06, + "loss": 0.2676, + "step": 69897 + }, + { + "epoch": 2.69, + "learning_rate": 5.214157476275969e-06, + "loss": 0.3382, + "step": 69898 + }, + { + "epoch": 2.69, + "learning_rate": 5.213516286227238e-06, + "loss": 0.1496, + "step": 69899 + }, + { + "epoch": 2.69, + "learning_rate": 5.212875096178507e-06, + "loss": 0.2332, + "step": 69900 + }, + { + "epoch": 2.69, + "learning_rate": 5.212233906129777e-06, + "loss": 0.3338, + "step": 69901 + }, + { + "epoch": 2.69, + "learning_rate": 5.211592716081046e-06, + "loss": 0.1915, + "step": 69902 + }, + { + "epoch": 2.69, + "learning_rate": 5.210951526032316e-06, + "loss": 0.0673, + "step": 69903 + }, + { + "epoch": 2.69, + "learning_rate": 5.2103103359835855e-06, + "loss": 0.3109, + "step": 69904 + }, + { + "epoch": 2.69, + "learning_rate": 5.2096691459348555e-06, + "loss": 0.1899, + "step": 69905 + }, + { + "epoch": 2.69, + "learning_rate": 5.209027955886125e-06, + "loss": 0.1596, + "step": 69906 + }, + { + "epoch": 2.69, + "learning_rate": 5.208386765837395e-06, + "loss": 0.259, + "step": 69907 + }, + { + "epoch": 2.69, + "learning_rate": 5.207745575788664e-06, + "loss": 0.2155, + "step": 69908 + }, + { + "epoch": 2.69, + "learning_rate": 5.207104385739934e-06, + "loss": 0.2384, + "step": 69909 + }, + { + "epoch": 2.69, + "learning_rate": 5.206463195691203e-06, + "loss": 0.1701, + "step": 69910 + }, + { + "epoch": 2.69, + "learning_rate": 5.205822005642472e-06, + "loss": 0.2145, + "step": 69911 + }, + { + "epoch": 2.69, + "learning_rate": 5.205180815593742e-06, + "loss": 0.2783, + "step": 69912 + }, + { + "epoch": 2.69, + "learning_rate": 5.2045396255450114e-06, + "loss": 0.1829, + "step": 69913 + }, + { + "epoch": 2.69, + "learning_rate": 5.2038984354962815e-06, + "loss": 0.278, + "step": 69914 + }, + { + "epoch": 2.69, + "learning_rate": 5.203257245447551e-06, + "loss": 0.296, + "step": 69915 + }, + { + "epoch": 2.69, + "learning_rate": 5.202616055398821e-06, + "loss": 0.1913, + "step": 69916 + }, + { + "epoch": 2.69, + "learning_rate": 5.20197486535009e-06, + "loss": 0.1871, + "step": 69917 + }, + { + "epoch": 2.69, + "learning_rate": 5.20133367530136e-06, + "loss": 0.1093, + "step": 69918 + }, + { + "epoch": 2.69, + "learning_rate": 5.200692485252629e-06, + "loss": 0.2431, + "step": 69919 + }, + { + "epoch": 2.69, + "learning_rate": 5.200051295203898e-06, + "loss": 0.4988, + "step": 69920 + }, + { + "epoch": 2.69, + "learning_rate": 5.199410105155168e-06, + "loss": 0.1857, + "step": 69921 + }, + { + "epoch": 2.69, + "learning_rate": 5.198768915106437e-06, + "loss": 0.1812, + "step": 69922 + }, + { + "epoch": 2.69, + "learning_rate": 5.1981277250577074e-06, + "loss": 0.167, + "step": 69923 + }, + { + "epoch": 2.69, + "learning_rate": 5.197486535008977e-06, + "loss": 0.2338, + "step": 69924 + }, + { + "epoch": 2.69, + "learning_rate": 5.196845344960247e-06, + "loss": 0.1236, + "step": 69925 + }, + { + "epoch": 2.69, + "learning_rate": 5.196204154911516e-06, + "loss": 0.1837, + "step": 69926 + }, + { + "epoch": 2.69, + "learning_rate": 5.195562964862786e-06, + "loss": 0.2058, + "step": 69927 + }, + { + "epoch": 2.69, + "learning_rate": 5.194921774814055e-06, + "loss": 0.2983, + "step": 69928 + }, + { + "epoch": 2.69, + "learning_rate": 5.194280584765325e-06, + "loss": 0.2397, + "step": 69929 + }, + { + "epoch": 2.69, + "learning_rate": 5.193639394716594e-06, + "loss": 0.1493, + "step": 69930 + }, + { + "epoch": 2.69, + "learning_rate": 5.192998204667863e-06, + "loss": 0.1316, + "step": 69931 + }, + { + "epoch": 2.69, + "learning_rate": 5.192357014619133e-06, + "loss": 0.2058, + "step": 69932 + }, + { + "epoch": 2.69, + "learning_rate": 5.191715824570403e-06, + "loss": 0.2089, + "step": 69933 + }, + { + "epoch": 2.69, + "learning_rate": 5.191074634521673e-06, + "loss": 0.1165, + "step": 69934 + }, + { + "epoch": 2.69, + "learning_rate": 5.190433444472942e-06, + "loss": 0.2946, + "step": 69935 + }, + { + "epoch": 2.69, + "learning_rate": 5.189792254424212e-06, + "loss": 0.141, + "step": 69936 + }, + { + "epoch": 2.69, + "learning_rate": 5.189151064375481e-06, + "loss": 0.298, + "step": 69937 + }, + { + "epoch": 2.69, + "learning_rate": 5.188509874326751e-06, + "loss": 0.1997, + "step": 69938 + }, + { + "epoch": 2.69, + "learning_rate": 5.18786868427802e-06, + "loss": 0.1442, + "step": 69939 + }, + { + "epoch": 2.69, + "learning_rate": 5.187227494229289e-06, + "loss": 0.1186, + "step": 69940 + }, + { + "epoch": 2.69, + "learning_rate": 5.186586304180559e-06, + "loss": 0.1957, + "step": 69941 + }, + { + "epoch": 2.69, + "learning_rate": 5.1859451141318285e-06, + "loss": 0.1142, + "step": 69942 + }, + { + "epoch": 2.69, + "learning_rate": 5.1853039240830986e-06, + "loss": 0.2406, + "step": 69943 + }, + { + "epoch": 2.69, + "learning_rate": 5.184662734034368e-06, + "loss": 0.1862, + "step": 69944 + }, + { + "epoch": 2.69, + "learning_rate": 5.184021543985638e-06, + "loss": 0.1361, + "step": 69945 + }, + { + "epoch": 2.69, + "learning_rate": 5.183380353936907e-06, + "loss": 0.1142, + "step": 69946 + }, + { + "epoch": 2.69, + "learning_rate": 5.182739163888177e-06, + "loss": 0.0913, + "step": 69947 + }, + { + "epoch": 2.69, + "learning_rate": 5.182097973839446e-06, + "loss": 0.1966, + "step": 69948 + }, + { + "epoch": 2.69, + "learning_rate": 5.181456783790715e-06, + "loss": 0.1932, + "step": 69949 + }, + { + "epoch": 2.69, + "learning_rate": 5.180815593741985e-06, + "loss": 0.2585, + "step": 69950 + }, + { + "epoch": 2.69, + "learning_rate": 5.1801744036932545e-06, + "loss": 0.1852, + "step": 69951 + }, + { + "epoch": 2.69, + "learning_rate": 5.1795332136445245e-06, + "loss": 0.1807, + "step": 69952 + }, + { + "epoch": 2.69, + "learning_rate": 5.178892023595794e-06, + "loss": 0.3561, + "step": 69953 + }, + { + "epoch": 2.69, + "learning_rate": 5.178250833547064e-06, + "loss": 0.1833, + "step": 69954 + }, + { + "epoch": 2.69, + "learning_rate": 5.177609643498333e-06, + "loss": 0.142, + "step": 69955 + }, + { + "epoch": 2.69, + "learning_rate": 5.176968453449603e-06, + "loss": 0.3404, + "step": 69956 + }, + { + "epoch": 2.69, + "learning_rate": 5.176327263400872e-06, + "loss": 0.119, + "step": 69957 + }, + { + "epoch": 2.69, + "learning_rate": 5.175686073352142e-06, + "loss": 0.3316, + "step": 69958 + }, + { + "epoch": 2.69, + "learning_rate": 5.175044883303411e-06, + "loss": 0.188, + "step": 69959 + }, + { + "epoch": 2.69, + "learning_rate": 5.1744036932546805e-06, + "loss": 0.2519, + "step": 69960 + }, + { + "epoch": 2.69, + "learning_rate": 5.1737625032059505e-06, + "loss": 0.1368, + "step": 69961 + }, + { + "epoch": 2.69, + "learning_rate": 5.17312131315722e-06, + "loss": 0.2817, + "step": 69962 + }, + { + "epoch": 2.69, + "learning_rate": 5.17248012310849e-06, + "loss": 0.2167, + "step": 69963 + }, + { + "epoch": 2.69, + "learning_rate": 5.171838933059759e-06, + "loss": 0.2036, + "step": 69964 + }, + { + "epoch": 2.69, + "learning_rate": 5.171197743011029e-06, + "loss": 0.2198, + "step": 69965 + }, + { + "epoch": 2.69, + "learning_rate": 5.170556552962298e-06, + "loss": 0.2508, + "step": 69966 + }, + { + "epoch": 2.69, + "learning_rate": 5.169915362913568e-06, + "loss": 0.2554, + "step": 69967 + }, + { + "epoch": 2.69, + "learning_rate": 5.169274172864837e-06, + "loss": 0.2154, + "step": 69968 + }, + { + "epoch": 2.69, + "learning_rate": 5.1686329828161064e-06, + "loss": 0.1432, + "step": 69969 + }, + { + "epoch": 2.69, + "learning_rate": 5.1679917927673765e-06, + "loss": 0.2477, + "step": 69970 + }, + { + "epoch": 2.69, + "learning_rate": 5.167350602718646e-06, + "loss": 0.1841, + "step": 69971 + }, + { + "epoch": 2.69, + "learning_rate": 5.166709412669916e-06, + "loss": 0.2953, + "step": 69972 + }, + { + "epoch": 2.69, + "learning_rate": 5.166068222621185e-06, + "loss": 0.1599, + "step": 69973 + }, + { + "epoch": 2.69, + "learning_rate": 5.165427032572455e-06, + "loss": 0.1813, + "step": 69974 + }, + { + "epoch": 2.69, + "learning_rate": 5.164785842523724e-06, + "loss": 0.1401, + "step": 69975 + }, + { + "epoch": 2.69, + "learning_rate": 5.164144652474994e-06, + "loss": 0.1269, + "step": 69976 + }, + { + "epoch": 2.69, + "learning_rate": 5.163503462426263e-06, + "loss": 0.2983, + "step": 69977 + }, + { + "epoch": 2.69, + "learning_rate": 5.162862272377533e-06, + "loss": 0.1458, + "step": 69978 + }, + { + "epoch": 2.69, + "learning_rate": 5.1622210823288024e-06, + "loss": 0.2144, + "step": 69979 + }, + { + "epoch": 2.69, + "learning_rate": 5.161579892280072e-06, + "loss": 0.2855, + "step": 69980 + }, + { + "epoch": 2.69, + "learning_rate": 5.160938702231342e-06, + "loss": 0.1094, + "step": 69981 + }, + { + "epoch": 2.69, + "learning_rate": 5.160297512182611e-06, + "loss": 0.2714, + "step": 69982 + }, + { + "epoch": 2.69, + "learning_rate": 5.159656322133881e-06, + "loss": 0.1434, + "step": 69983 + }, + { + "epoch": 2.69, + "learning_rate": 5.15901513208515e-06, + "loss": 0.1652, + "step": 69984 + }, + { + "epoch": 2.69, + "learning_rate": 5.15837394203642e-06, + "loss": 0.2614, + "step": 69985 + }, + { + "epoch": 2.69, + "learning_rate": 5.157732751987689e-06, + "loss": 0.2749, + "step": 69986 + }, + { + "epoch": 2.69, + "learning_rate": 5.157091561938959e-06, + "loss": 0.1813, + "step": 69987 + }, + { + "epoch": 2.69, + "learning_rate": 5.156450371890228e-06, + "loss": 0.3397, + "step": 69988 + }, + { + "epoch": 2.69, + "learning_rate": 5.155809181841498e-06, + "loss": 0.1719, + "step": 69989 + }, + { + "epoch": 2.69, + "learning_rate": 5.155167991792768e-06, + "loss": 0.2786, + "step": 69990 + }, + { + "epoch": 2.69, + "learning_rate": 5.154526801744037e-06, + "loss": 0.0813, + "step": 69991 + }, + { + "epoch": 2.69, + "learning_rate": 5.153885611695307e-06, + "loss": 0.2029, + "step": 69992 + }, + { + "epoch": 2.69, + "learning_rate": 5.153244421646576e-06, + "loss": 0.2808, + "step": 69993 + }, + { + "epoch": 2.69, + "learning_rate": 5.152603231597846e-06, + "loss": 0.1234, + "step": 69994 + }, + { + "epoch": 2.69, + "learning_rate": 5.151962041549115e-06, + "loss": 0.1959, + "step": 69995 + }, + { + "epoch": 2.69, + "learning_rate": 5.151320851500385e-06, + "loss": 0.2494, + "step": 69996 + }, + { + "epoch": 2.69, + "learning_rate": 5.150679661451654e-06, + "loss": 0.1656, + "step": 69997 + }, + { + "epoch": 2.69, + "learning_rate": 5.150038471402924e-06, + "loss": 0.1975, + "step": 69998 + }, + { + "epoch": 2.69, + "learning_rate": 5.1493972813541936e-06, + "loss": 0.2005, + "step": 69999 + }, + { + "epoch": 2.69, + "learning_rate": 5.148756091305463e-06, + "loss": 0.1766, + "step": 70000 + }, + { + "epoch": 2.69, + "learning_rate": 5.148114901256733e-06, + "loss": 0.1367, + "step": 70001 + }, + { + "epoch": 2.69, + "learning_rate": 5.147473711208002e-06, + "loss": 0.269, + "step": 70002 + }, + { + "epoch": 2.69, + "learning_rate": 5.146832521159272e-06, + "loss": 0.2967, + "step": 70003 + }, + { + "epoch": 2.69, + "learning_rate": 5.146191331110541e-06, + "loss": 0.2892, + "step": 70004 + }, + { + "epoch": 2.69, + "learning_rate": 5.145550141061811e-06, + "loss": 0.2113, + "step": 70005 + }, + { + "epoch": 2.69, + "learning_rate": 5.14490895101308e-06, + "loss": 0.318, + "step": 70006 + }, + { + "epoch": 2.69, + "learning_rate": 5.14426776096435e-06, + "loss": 0.0998, + "step": 70007 + }, + { + "epoch": 2.69, + "learning_rate": 5.1436265709156195e-06, + "loss": 0.139, + "step": 70008 + }, + { + "epoch": 2.69, + "learning_rate": 5.142985380866889e-06, + "loss": 0.3044, + "step": 70009 + }, + { + "epoch": 2.69, + "learning_rate": 5.142344190818159e-06, + "loss": 0.2445, + "step": 70010 + }, + { + "epoch": 2.69, + "learning_rate": 5.141703000769428e-06, + "loss": 0.2133, + "step": 70011 + }, + { + "epoch": 2.69, + "learning_rate": 5.141061810720698e-06, + "loss": 0.1825, + "step": 70012 + }, + { + "epoch": 2.69, + "learning_rate": 5.140420620671967e-06, + "loss": 0.1706, + "step": 70013 + }, + { + "epoch": 2.69, + "learning_rate": 5.139779430623237e-06, + "loss": 0.1635, + "step": 70014 + }, + { + "epoch": 2.69, + "learning_rate": 5.139138240574506e-06, + "loss": 0.1811, + "step": 70015 + }, + { + "epoch": 2.69, + "learning_rate": 5.138497050525776e-06, + "loss": 0.3179, + "step": 70016 + }, + { + "epoch": 2.69, + "learning_rate": 5.1378558604770455e-06, + "loss": 0.0613, + "step": 70017 + }, + { + "epoch": 2.69, + "learning_rate": 5.137214670428315e-06, + "loss": 0.1863, + "step": 70018 + }, + { + "epoch": 2.69, + "learning_rate": 5.136573480379585e-06, + "loss": 0.2293, + "step": 70019 + }, + { + "epoch": 2.69, + "learning_rate": 5.135932290330854e-06, + "loss": 0.2316, + "step": 70020 + }, + { + "epoch": 2.69, + "learning_rate": 5.135291100282124e-06, + "loss": 0.2767, + "step": 70021 + }, + { + "epoch": 2.69, + "learning_rate": 5.134649910233393e-06, + "loss": 0.329, + "step": 70022 + }, + { + "epoch": 2.69, + "learning_rate": 5.134008720184663e-06, + "loss": 0.1849, + "step": 70023 + }, + { + "epoch": 2.69, + "learning_rate": 5.133367530135932e-06, + "loss": 0.1913, + "step": 70024 + }, + { + "epoch": 2.69, + "learning_rate": 5.132726340087202e-06, + "loss": 0.2365, + "step": 70025 + }, + { + "epoch": 2.69, + "learning_rate": 5.1320851500384715e-06, + "loss": 0.2329, + "step": 70026 + }, + { + "epoch": 2.69, + "learning_rate": 5.1314439599897415e-06, + "loss": 0.1748, + "step": 70027 + }, + { + "epoch": 2.69, + "learning_rate": 5.130802769941011e-06, + "loss": 0.2061, + "step": 70028 + }, + { + "epoch": 2.69, + "learning_rate": 5.13016157989228e-06, + "loss": 0.1852, + "step": 70029 + }, + { + "epoch": 2.69, + "learning_rate": 5.12952038984355e-06, + "loss": 0.2655, + "step": 70030 + }, + { + "epoch": 2.69, + "learning_rate": 5.128879199794819e-06, + "loss": 0.1771, + "step": 70031 + }, + { + "epoch": 2.69, + "learning_rate": 5.128238009746089e-06, + "loss": 0.2515, + "step": 70032 + }, + { + "epoch": 2.69, + "learning_rate": 5.127596819697358e-06, + "loss": 0.1134, + "step": 70033 + }, + { + "epoch": 2.69, + "learning_rate": 5.126955629648628e-06, + "loss": 0.1937, + "step": 70034 + }, + { + "epoch": 2.69, + "learning_rate": 5.1263144395998974e-06, + "loss": 0.3247, + "step": 70035 + }, + { + "epoch": 2.69, + "learning_rate": 5.1256732495511675e-06, + "loss": 0.1923, + "step": 70036 + }, + { + "epoch": 2.69, + "learning_rate": 5.125032059502437e-06, + "loss": 0.3153, + "step": 70037 + }, + { + "epoch": 2.69, + "learning_rate": 5.124390869453706e-06, + "loss": 0.1982, + "step": 70038 + }, + { + "epoch": 2.69, + "learning_rate": 5.123749679404976e-06, + "loss": 0.2191, + "step": 70039 + }, + { + "epoch": 2.69, + "learning_rate": 5.123108489356245e-06, + "loss": 0.1631, + "step": 70040 + }, + { + "epoch": 2.69, + "learning_rate": 5.122467299307515e-06, + "loss": 0.1714, + "step": 70041 + }, + { + "epoch": 2.69, + "learning_rate": 5.121826109258784e-06, + "loss": 0.0861, + "step": 70042 + }, + { + "epoch": 2.69, + "learning_rate": 5.121184919210054e-06, + "loss": 0.1801, + "step": 70043 + }, + { + "epoch": 2.69, + "learning_rate": 5.120543729161323e-06, + "loss": 0.075, + "step": 70044 + }, + { + "epoch": 2.69, + "learning_rate": 5.1199025391125934e-06, + "loss": 0.2932, + "step": 70045 + }, + { + "epoch": 2.69, + "learning_rate": 5.119261349063863e-06, + "loss": 0.1162, + "step": 70046 + }, + { + "epoch": 2.69, + "learning_rate": 5.118620159015133e-06, + "loss": 0.1306, + "step": 70047 + }, + { + "epoch": 2.69, + "learning_rate": 5.117978968966402e-06, + "loss": 0.2789, + "step": 70048 + }, + { + "epoch": 2.69, + "learning_rate": 5.117337778917671e-06, + "loss": 0.2233, + "step": 70049 + }, + { + "epoch": 2.69, + "learning_rate": 5.116696588868941e-06, + "loss": 0.281, + "step": 70050 + }, + { + "epoch": 2.69, + "learning_rate": 5.11605539882021e-06, + "loss": 0.2167, + "step": 70051 + }, + { + "epoch": 2.69, + "learning_rate": 5.11541420877148e-06, + "loss": 0.2066, + "step": 70052 + }, + { + "epoch": 2.69, + "learning_rate": 5.114773018722749e-06, + "loss": 0.1605, + "step": 70053 + }, + { + "epoch": 2.69, + "learning_rate": 5.114131828674019e-06, + "loss": 0.18, + "step": 70054 + }, + { + "epoch": 2.69, + "learning_rate": 5.113490638625289e-06, + "loss": 0.1037, + "step": 70055 + }, + { + "epoch": 2.69, + "learning_rate": 5.112849448576559e-06, + "loss": 0.2648, + "step": 70056 + }, + { + "epoch": 2.69, + "learning_rate": 5.112208258527828e-06, + "loss": 0.1611, + "step": 70057 + }, + { + "epoch": 2.69, + "learning_rate": 5.111567068479097e-06, + "loss": 0.1307, + "step": 70058 + }, + { + "epoch": 2.69, + "learning_rate": 5.110925878430367e-06, + "loss": 0.2441, + "step": 70059 + }, + { + "epoch": 2.69, + "learning_rate": 5.110284688381636e-06, + "loss": 0.3692, + "step": 70060 + }, + { + "epoch": 2.69, + "learning_rate": 5.109643498332906e-06, + "loss": 0.2429, + "step": 70061 + }, + { + "epoch": 2.69, + "learning_rate": 5.109002308284175e-06, + "loss": 0.2781, + "step": 70062 + }, + { + "epoch": 2.69, + "learning_rate": 5.108361118235445e-06, + "loss": 0.3055, + "step": 70063 + }, + { + "epoch": 2.69, + "learning_rate": 5.1077199281867145e-06, + "loss": 0.1852, + "step": 70064 + }, + { + "epoch": 2.69, + "learning_rate": 5.1070787381379846e-06, + "loss": 0.2188, + "step": 70065 + }, + { + "epoch": 2.69, + "learning_rate": 5.106437548089254e-06, + "loss": 0.2092, + "step": 70066 + }, + { + "epoch": 2.69, + "learning_rate": 5.105796358040523e-06, + "loss": 0.2214, + "step": 70067 + }, + { + "epoch": 2.69, + "learning_rate": 5.105155167991793e-06, + "loss": 0.2267, + "step": 70068 + }, + { + "epoch": 2.69, + "learning_rate": 5.104513977943062e-06, + "loss": 0.2698, + "step": 70069 + }, + { + "epoch": 2.69, + "learning_rate": 5.103872787894332e-06, + "loss": 0.2463, + "step": 70070 + }, + { + "epoch": 2.69, + "learning_rate": 5.103231597845601e-06, + "loss": 0.1401, + "step": 70071 + }, + { + "epoch": 2.69, + "learning_rate": 5.102590407796871e-06, + "loss": 0.1458, + "step": 70072 + }, + { + "epoch": 2.69, + "learning_rate": 5.1019492177481405e-06, + "loss": 0.2336, + "step": 70073 + }, + { + "epoch": 2.69, + "learning_rate": 5.1013080276994105e-06, + "loss": 0.1763, + "step": 70074 + }, + { + "epoch": 2.69, + "learning_rate": 5.10066683765068e-06, + "loss": 0.2111, + "step": 70075 + }, + { + "epoch": 2.69, + "learning_rate": 5.10002564760195e-06, + "loss": 0.2663, + "step": 70076 + }, + { + "epoch": 2.69, + "learning_rate": 5.099384457553219e-06, + "loss": 0.1928, + "step": 70077 + }, + { + "epoch": 2.69, + "learning_rate": 5.098743267504488e-06, + "loss": 0.2514, + "step": 70078 + }, + { + "epoch": 2.69, + "learning_rate": 5.098102077455758e-06, + "loss": 0.2045, + "step": 70079 + }, + { + "epoch": 2.69, + "learning_rate": 5.097460887407027e-06, + "loss": 0.167, + "step": 70080 + }, + { + "epoch": 2.69, + "learning_rate": 5.096819697358297e-06, + "loss": 0.1912, + "step": 70081 + }, + { + "epoch": 2.69, + "learning_rate": 5.0961785073095665e-06, + "loss": 0.1652, + "step": 70082 + }, + { + "epoch": 2.69, + "learning_rate": 5.0955373172608365e-06, + "loss": 0.232, + "step": 70083 + }, + { + "epoch": 2.69, + "learning_rate": 5.094896127212106e-06, + "loss": 0.2922, + "step": 70084 + }, + { + "epoch": 2.69, + "learning_rate": 5.094254937163376e-06, + "loss": 0.228, + "step": 70085 + }, + { + "epoch": 2.69, + "learning_rate": 5.093613747114645e-06, + "loss": 0.185, + "step": 70086 + }, + { + "epoch": 2.69, + "learning_rate": 5.092972557065914e-06, + "loss": 0.0654, + "step": 70087 + }, + { + "epoch": 2.69, + "learning_rate": 5.092331367017184e-06, + "loss": 0.2181, + "step": 70088 + }, + { + "epoch": 2.69, + "learning_rate": 5.091690176968453e-06, + "loss": 0.1926, + "step": 70089 + }, + { + "epoch": 2.69, + "learning_rate": 5.091048986919723e-06, + "loss": 0.1014, + "step": 70090 + }, + { + "epoch": 2.69, + "learning_rate": 5.0904077968709924e-06, + "loss": 0.2912, + "step": 70091 + }, + { + "epoch": 2.69, + "learning_rate": 5.0897666068222625e-06, + "loss": 0.2845, + "step": 70092 + }, + { + "epoch": 2.69, + "learning_rate": 5.089125416773532e-06, + "loss": 0.1619, + "step": 70093 + }, + { + "epoch": 2.69, + "learning_rate": 5.088484226724802e-06, + "loss": 0.2717, + "step": 70094 + }, + { + "epoch": 2.69, + "learning_rate": 5.087843036676071e-06, + "loss": 0.2839, + "step": 70095 + }, + { + "epoch": 2.69, + "learning_rate": 5.087201846627341e-06, + "loss": 0.244, + "step": 70096 + }, + { + "epoch": 2.7, + "learning_rate": 5.08656065657861e-06, + "loss": 0.1267, + "step": 70097 + }, + { + "epoch": 2.7, + "learning_rate": 5.085919466529879e-06, + "loss": 0.2306, + "step": 70098 + }, + { + "epoch": 2.7, + "learning_rate": 5.085278276481149e-06, + "loss": 0.2927, + "step": 70099 + }, + { + "epoch": 2.7, + "learning_rate": 5.084637086432418e-06, + "loss": 0.2208, + "step": 70100 + }, + { + "epoch": 2.7, + "learning_rate": 5.0839958963836884e-06, + "loss": 0.3482, + "step": 70101 + }, + { + "epoch": 2.7, + "learning_rate": 5.083354706334958e-06, + "loss": 0.21, + "step": 70102 + }, + { + "epoch": 2.7, + "learning_rate": 5.082713516286228e-06, + "loss": 0.1861, + "step": 70103 + }, + { + "epoch": 2.7, + "learning_rate": 5.082072326237497e-06, + "loss": 0.2369, + "step": 70104 + }, + { + "epoch": 2.7, + "learning_rate": 5.081431136188767e-06, + "loss": 0.2493, + "step": 70105 + }, + { + "epoch": 2.7, + "learning_rate": 5.080789946140036e-06, + "loss": 0.1368, + "step": 70106 + }, + { + "epoch": 2.7, + "learning_rate": 5.080148756091305e-06, + "loss": 0.2884, + "step": 70107 + }, + { + "epoch": 2.7, + "learning_rate": 5.079507566042575e-06, + "loss": 0.2384, + "step": 70108 + }, + { + "epoch": 2.7, + "learning_rate": 5.078866375993844e-06, + "loss": 0.2609, + "step": 70109 + }, + { + "epoch": 2.7, + "learning_rate": 5.078225185945114e-06, + "loss": 0.1692, + "step": 70110 + }, + { + "epoch": 2.7, + "learning_rate": 5.077583995896384e-06, + "loss": 0.3561, + "step": 70111 + }, + { + "epoch": 2.7, + "learning_rate": 5.076942805847654e-06, + "loss": 0.0854, + "step": 70112 + }, + { + "epoch": 2.7, + "learning_rate": 5.076301615798923e-06, + "loss": 0.4025, + "step": 70113 + }, + { + "epoch": 2.7, + "learning_rate": 5.075660425750193e-06, + "loss": 0.2933, + "step": 70114 + }, + { + "epoch": 2.7, + "learning_rate": 5.075019235701462e-06, + "loss": 0.1619, + "step": 70115 + }, + { + "epoch": 2.7, + "learning_rate": 5.074378045652731e-06, + "loss": 0.1003, + "step": 70116 + }, + { + "epoch": 2.7, + "learning_rate": 5.073736855604001e-06, + "loss": 0.1774, + "step": 70117 + }, + { + "epoch": 2.7, + "learning_rate": 5.07309566555527e-06, + "loss": 0.2507, + "step": 70118 + }, + { + "epoch": 2.7, + "learning_rate": 5.07245447550654e-06, + "loss": 0.1746, + "step": 70119 + }, + { + "epoch": 2.7, + "learning_rate": 5.0718132854578096e-06, + "loss": 0.1966, + "step": 70120 + }, + { + "epoch": 2.7, + "learning_rate": 5.0711720954090796e-06, + "loss": 0.2031, + "step": 70121 + }, + { + "epoch": 2.7, + "learning_rate": 5.070530905360349e-06, + "loss": 0.1632, + "step": 70122 + }, + { + "epoch": 2.7, + "learning_rate": 5.069889715311619e-06, + "loss": 0.3202, + "step": 70123 + }, + { + "epoch": 2.7, + "learning_rate": 5.069248525262888e-06, + "loss": 0.2033, + "step": 70124 + }, + { + "epoch": 2.7, + "learning_rate": 5.068607335214158e-06, + "loss": 0.2728, + "step": 70125 + }, + { + "epoch": 2.7, + "learning_rate": 5.067966145165427e-06, + "loss": 0.3683, + "step": 70126 + }, + { + "epoch": 2.7, + "learning_rate": 5.067324955116696e-06, + "loss": 0.3386, + "step": 70127 + }, + { + "epoch": 2.7, + "learning_rate": 5.066683765067966e-06, + "loss": 0.2153, + "step": 70128 + }, + { + "epoch": 2.7, + "learning_rate": 5.0660425750192355e-06, + "loss": 0.1704, + "step": 70129 + }, + { + "epoch": 2.7, + "learning_rate": 5.0654013849705055e-06, + "loss": 0.243, + "step": 70130 + }, + { + "epoch": 2.7, + "learning_rate": 5.064760194921775e-06, + "loss": 0.2462, + "step": 70131 + }, + { + "epoch": 2.7, + "learning_rate": 5.064119004873045e-06, + "loss": 0.2681, + "step": 70132 + }, + { + "epoch": 2.7, + "learning_rate": 5.063477814824314e-06, + "loss": 0.1783, + "step": 70133 + }, + { + "epoch": 2.7, + "learning_rate": 5.062836624775584e-06, + "loss": 0.2313, + "step": 70134 + }, + { + "epoch": 2.7, + "learning_rate": 5.062195434726853e-06, + "loss": 0.2244, + "step": 70135 + }, + { + "epoch": 2.7, + "learning_rate": 5.061554244678122e-06, + "loss": 0.2737, + "step": 70136 + }, + { + "epoch": 2.7, + "learning_rate": 5.060913054629392e-06, + "loss": 0.2362, + "step": 70137 + }, + { + "epoch": 2.7, + "learning_rate": 5.0602718645806615e-06, + "loss": 0.2521, + "step": 70138 + }, + { + "epoch": 2.7, + "learning_rate": 5.0596306745319315e-06, + "loss": 0.247, + "step": 70139 + }, + { + "epoch": 2.7, + "learning_rate": 5.058989484483201e-06, + "loss": 0.2196, + "step": 70140 + }, + { + "epoch": 2.7, + "learning_rate": 5.058348294434471e-06, + "loss": 0.0837, + "step": 70141 + }, + { + "epoch": 2.7, + "learning_rate": 5.05770710438574e-06, + "loss": 0.1303, + "step": 70142 + }, + { + "epoch": 2.7, + "learning_rate": 5.05706591433701e-06, + "loss": 0.2526, + "step": 70143 + }, + { + "epoch": 2.7, + "learning_rate": 5.056424724288279e-06, + "loss": 0.1885, + "step": 70144 + }, + { + "epoch": 2.7, + "learning_rate": 5.055783534239549e-06, + "loss": 0.2041, + "step": 70145 + }, + { + "epoch": 2.7, + "learning_rate": 5.055142344190818e-06, + "loss": 0.1458, + "step": 70146 + }, + { + "epoch": 2.7, + "learning_rate": 5.0545011541420875e-06, + "loss": 0.3277, + "step": 70147 + }, + { + "epoch": 2.7, + "learning_rate": 5.0538599640933575e-06, + "loss": 0.2012, + "step": 70148 + }, + { + "epoch": 2.7, + "learning_rate": 5.053218774044627e-06, + "loss": 0.1245, + "step": 70149 + }, + { + "epoch": 2.7, + "learning_rate": 5.052577583995897e-06, + "loss": 0.1743, + "step": 70150 + }, + { + "epoch": 2.7, + "learning_rate": 5.051936393947166e-06, + "loss": 0.1602, + "step": 70151 + }, + { + "epoch": 2.7, + "learning_rate": 5.051295203898436e-06, + "loss": 0.1453, + "step": 70152 + }, + { + "epoch": 2.7, + "learning_rate": 5.050654013849705e-06, + "loss": 0.2592, + "step": 70153 + }, + { + "epoch": 2.7, + "learning_rate": 5.050012823800975e-06, + "loss": 0.2255, + "step": 70154 + }, + { + "epoch": 2.7, + "learning_rate": 5.049371633752244e-06, + "loss": 0.1703, + "step": 70155 + }, + { + "epoch": 2.7, + "learning_rate": 5.048730443703513e-06, + "loss": 0.1573, + "step": 70156 + }, + { + "epoch": 2.7, + "learning_rate": 5.0480892536547834e-06, + "loss": 0.188, + "step": 70157 + }, + { + "epoch": 2.7, + "learning_rate": 5.047448063606053e-06, + "loss": 0.2428, + "step": 70158 + }, + { + "epoch": 2.7, + "learning_rate": 5.046806873557323e-06, + "loss": 0.1987, + "step": 70159 + }, + { + "epoch": 2.7, + "learning_rate": 5.046165683508592e-06, + "loss": 0.2526, + "step": 70160 + }, + { + "epoch": 2.7, + "learning_rate": 5.045524493459862e-06, + "loss": 0.0718, + "step": 70161 + }, + { + "epoch": 2.7, + "learning_rate": 5.044883303411131e-06, + "loss": 0.1599, + "step": 70162 + }, + { + "epoch": 2.7, + "learning_rate": 5.044242113362401e-06, + "loss": 0.3408, + "step": 70163 + }, + { + "epoch": 2.7, + "learning_rate": 5.04360092331367e-06, + "loss": 0.2143, + "step": 70164 + }, + { + "epoch": 2.7, + "learning_rate": 5.04295973326494e-06, + "loss": 0.3011, + "step": 70165 + }, + { + "epoch": 2.7, + "learning_rate": 5.042318543216209e-06, + "loss": 0.1211, + "step": 70166 + }, + { + "epoch": 2.7, + "learning_rate": 5.041677353167479e-06, + "loss": 0.1462, + "step": 70167 + }, + { + "epoch": 2.7, + "learning_rate": 5.041036163118749e-06, + "loss": 0.1426, + "step": 70168 + }, + { + "epoch": 2.7, + "learning_rate": 5.040394973070018e-06, + "loss": 0.1527, + "step": 70169 + }, + { + "epoch": 2.7, + "learning_rate": 5.039753783021288e-06, + "loss": 0.2224, + "step": 70170 + }, + { + "epoch": 2.7, + "learning_rate": 5.039112592972557e-06, + "loss": 0.1739, + "step": 70171 + }, + { + "epoch": 2.7, + "learning_rate": 5.038471402923827e-06, + "loss": 0.2178, + "step": 70172 + }, + { + "epoch": 2.7, + "learning_rate": 5.037830212875096e-06, + "loss": 0.2152, + "step": 70173 + }, + { + "epoch": 2.7, + "learning_rate": 5.037189022826366e-06, + "loss": 0.1452, + "step": 70174 + }, + { + "epoch": 2.7, + "learning_rate": 5.036547832777635e-06, + "loss": 0.3365, + "step": 70175 + }, + { + "epoch": 2.7, + "learning_rate": 5.0359066427289046e-06, + "loss": 0.1926, + "step": 70176 + }, + { + "epoch": 2.7, + "learning_rate": 5.035265452680175e-06, + "loss": 0.2041, + "step": 70177 + }, + { + "epoch": 2.7, + "learning_rate": 5.034624262631444e-06, + "loss": 0.4118, + "step": 70178 + }, + { + "epoch": 2.7, + "learning_rate": 5.033983072582714e-06, + "loss": 0.1848, + "step": 70179 + }, + { + "epoch": 2.7, + "learning_rate": 5.033341882533983e-06, + "loss": 0.2444, + "step": 70180 + }, + { + "epoch": 2.7, + "learning_rate": 5.032700692485253e-06, + "loss": 0.1942, + "step": 70181 + }, + { + "epoch": 2.7, + "learning_rate": 5.032059502436522e-06, + "loss": 0.1113, + "step": 70182 + }, + { + "epoch": 2.7, + "learning_rate": 5.031418312387792e-06, + "loss": 0.1269, + "step": 70183 + }, + { + "epoch": 2.7, + "learning_rate": 5.030777122339061e-06, + "loss": 0.1548, + "step": 70184 + }, + { + "epoch": 2.7, + "learning_rate": 5.0301359322903305e-06, + "loss": 0.136, + "step": 70185 + }, + { + "epoch": 2.7, + "learning_rate": 5.0294947422416005e-06, + "loss": 0.2242, + "step": 70186 + }, + { + "epoch": 2.7, + "learning_rate": 5.02885355219287e-06, + "loss": 0.1015, + "step": 70187 + }, + { + "epoch": 2.7, + "learning_rate": 5.02821236214414e-06, + "loss": 0.1902, + "step": 70188 + }, + { + "epoch": 2.7, + "learning_rate": 5.027571172095409e-06, + "loss": 0.2878, + "step": 70189 + }, + { + "epoch": 2.7, + "learning_rate": 5.026929982046679e-06, + "loss": 0.18, + "step": 70190 + }, + { + "epoch": 2.7, + "learning_rate": 5.026288791997948e-06, + "loss": 0.2296, + "step": 70191 + }, + { + "epoch": 2.7, + "learning_rate": 5.025647601949218e-06, + "loss": 0.2021, + "step": 70192 + }, + { + "epoch": 2.7, + "learning_rate": 5.025006411900487e-06, + "loss": 0.2032, + "step": 70193 + }, + { + "epoch": 2.7, + "learning_rate": 5.024365221851757e-06, + "loss": 0.1086, + "step": 70194 + }, + { + "epoch": 2.7, + "learning_rate": 5.0237240318030265e-06, + "loss": 0.3047, + "step": 70195 + }, + { + "epoch": 2.7, + "learning_rate": 5.023082841754296e-06, + "loss": 0.1345, + "step": 70196 + }, + { + "epoch": 2.7, + "learning_rate": 5.022441651705566e-06, + "loss": 0.3431, + "step": 70197 + }, + { + "epoch": 2.7, + "learning_rate": 5.021800461656835e-06, + "loss": 0.2163, + "step": 70198 + }, + { + "epoch": 2.7, + "learning_rate": 5.021159271608105e-06, + "loss": 0.3472, + "step": 70199 + }, + { + "epoch": 2.7, + "learning_rate": 5.020518081559374e-06, + "loss": 0.5429, + "step": 70200 + }, + { + "epoch": 2.7, + "learning_rate": 5.019876891510644e-06, + "loss": 0.2572, + "step": 70201 + }, + { + "epoch": 2.7, + "learning_rate": 5.019235701461913e-06, + "loss": 0.2521, + "step": 70202 + }, + { + "epoch": 2.7, + "learning_rate": 5.018594511413183e-06, + "loss": 0.2734, + "step": 70203 + }, + { + "epoch": 2.7, + "learning_rate": 5.0179533213644525e-06, + "loss": 0.1962, + "step": 70204 + }, + { + "epoch": 2.7, + "learning_rate": 5.017312131315722e-06, + "loss": 0.2139, + "step": 70205 + }, + { + "epoch": 2.7, + "learning_rate": 5.016670941266992e-06, + "loss": 0.0713, + "step": 70206 + }, + { + "epoch": 2.7, + "learning_rate": 5.016029751218261e-06, + "loss": 0.2198, + "step": 70207 + }, + { + "epoch": 2.7, + "learning_rate": 5.015388561169531e-06, + "loss": 0.2412, + "step": 70208 + }, + { + "epoch": 2.7, + "learning_rate": 5.0147473711208e-06, + "loss": 0.2324, + "step": 70209 + }, + { + "epoch": 2.7, + "learning_rate": 5.01410618107207e-06, + "loss": 0.1658, + "step": 70210 + }, + { + "epoch": 2.7, + "learning_rate": 5.013464991023339e-06, + "loss": 0.2018, + "step": 70211 + }, + { + "epoch": 2.7, + "learning_rate": 5.012823800974609e-06, + "loss": 0.2543, + "step": 70212 + }, + { + "epoch": 2.7, + "learning_rate": 5.0121826109258785e-06, + "loss": 0.3114, + "step": 70213 + }, + { + "epoch": 2.7, + "learning_rate": 5.0115414208771485e-06, + "loss": 0.132, + "step": 70214 + }, + { + "epoch": 2.7, + "learning_rate": 5.010900230828418e-06, + "loss": 0.2423, + "step": 70215 + }, + { + "epoch": 2.7, + "learning_rate": 5.010259040779687e-06, + "loss": 0.2992, + "step": 70216 + }, + { + "epoch": 2.7, + "learning_rate": 5.009617850730957e-06, + "loss": 0.1726, + "step": 70217 + }, + { + "epoch": 2.7, + "learning_rate": 5.008976660682226e-06, + "loss": 0.1646, + "step": 70218 + }, + { + "epoch": 2.7, + "learning_rate": 5.008335470633496e-06, + "loss": 0.2404, + "step": 70219 + }, + { + "epoch": 2.7, + "learning_rate": 5.007694280584765e-06, + "loss": 0.2967, + "step": 70220 + }, + { + "epoch": 2.7, + "learning_rate": 5.007053090536035e-06, + "loss": 0.1631, + "step": 70221 + }, + { + "epoch": 2.7, + "learning_rate": 5.006411900487304e-06, + "loss": 0.1684, + "step": 70222 + }, + { + "epoch": 2.7, + "learning_rate": 5.0057707104385744e-06, + "loss": 0.1556, + "step": 70223 + }, + { + "epoch": 2.7, + "learning_rate": 5.005129520389844e-06, + "loss": 0.2051, + "step": 70224 + }, + { + "epoch": 2.7, + "learning_rate": 5.004488330341113e-06, + "loss": 0.1266, + "step": 70225 + }, + { + "epoch": 2.7, + "learning_rate": 5.003847140292383e-06, + "loss": 0.2124, + "step": 70226 + }, + { + "epoch": 2.7, + "learning_rate": 5.003205950243652e-06, + "loss": 0.1446, + "step": 70227 + }, + { + "epoch": 2.7, + "learning_rate": 5.002564760194922e-06, + "loss": 0.1119, + "step": 70228 + }, + { + "epoch": 2.7, + "learning_rate": 5.001923570146191e-06, + "loss": 0.197, + "step": 70229 + }, + { + "epoch": 2.7, + "learning_rate": 5.001282380097461e-06, + "loss": 0.2509, + "step": 70230 + }, + { + "epoch": 2.7, + "learning_rate": 5.00064119004873e-06, + "loss": 0.1929, + "step": 70231 + }, + { + "epoch": 2.7, + "learning_rate": 5e-06, + "loss": 0.223, + "step": 70232 + }, + { + "epoch": 2.7, + "learning_rate": 4.99935880995127e-06, + "loss": 0.2718, + "step": 70233 + }, + { + "epoch": 2.7, + "learning_rate": 4.998717619902539e-06, + "loss": 0.2028, + "step": 70234 + }, + { + "epoch": 2.7, + "learning_rate": 4.998076429853809e-06, + "loss": 0.3025, + "step": 70235 + }, + { + "epoch": 2.7, + "learning_rate": 4.997435239805078e-06, + "loss": 0.1811, + "step": 70236 + }, + { + "epoch": 2.7, + "learning_rate": 4.996794049756348e-06, + "loss": 0.077, + "step": 70237 + }, + { + "epoch": 2.7, + "learning_rate": 4.996152859707617e-06, + "loss": 0.1755, + "step": 70238 + }, + { + "epoch": 2.7, + "learning_rate": 4.995511669658887e-06, + "loss": 0.2141, + "step": 70239 + }, + { + "epoch": 2.7, + "learning_rate": 4.994870479610156e-06, + "loss": 0.2775, + "step": 70240 + }, + { + "epoch": 2.7, + "learning_rate": 4.994229289561426e-06, + "loss": 0.2134, + "step": 70241 + }, + { + "epoch": 2.7, + "learning_rate": 4.9935880995126956e-06, + "loss": 0.1671, + "step": 70242 + }, + { + "epoch": 2.7, + "learning_rate": 4.9929469094639656e-06, + "loss": 0.2631, + "step": 70243 + }, + { + "epoch": 2.7, + "learning_rate": 4.992305719415235e-06, + "loss": 0.2463, + "step": 70244 + }, + { + "epoch": 2.7, + "learning_rate": 4.991664529366504e-06, + "loss": 0.152, + "step": 70245 + }, + { + "epoch": 2.7, + "learning_rate": 4.991023339317774e-06, + "loss": 0.1093, + "step": 70246 + }, + { + "epoch": 2.7, + "learning_rate": 4.990382149269043e-06, + "loss": 0.2263, + "step": 70247 + }, + { + "epoch": 2.7, + "learning_rate": 4.989740959220313e-06, + "loss": 0.2612, + "step": 70248 + }, + { + "epoch": 2.7, + "learning_rate": 4.989099769171582e-06, + "loss": 0.309, + "step": 70249 + }, + { + "epoch": 2.7, + "learning_rate": 4.988458579122852e-06, + "loss": 0.3404, + "step": 70250 + }, + { + "epoch": 2.7, + "learning_rate": 4.9878173890741215e-06, + "loss": 0.2026, + "step": 70251 + }, + { + "epoch": 2.7, + "learning_rate": 4.9871761990253915e-06, + "loss": 0.0612, + "step": 70252 + }, + { + "epoch": 2.7, + "learning_rate": 4.986535008976661e-06, + "loss": 0.1752, + "step": 70253 + }, + { + "epoch": 2.7, + "learning_rate": 4.98589381892793e-06, + "loss": 0.2073, + "step": 70254 + }, + { + "epoch": 2.7, + "learning_rate": 4.9852526288792e-06, + "loss": 0.1248, + "step": 70255 + }, + { + "epoch": 2.7, + "learning_rate": 4.984611438830469e-06, + "loss": 0.2615, + "step": 70256 + }, + { + "epoch": 2.7, + "learning_rate": 4.983970248781739e-06, + "loss": 0.2558, + "step": 70257 + }, + { + "epoch": 2.7, + "learning_rate": 4.983329058733008e-06, + "loss": 0.2667, + "step": 70258 + }, + { + "epoch": 2.7, + "learning_rate": 4.982687868684278e-06, + "loss": 0.1987, + "step": 70259 + }, + { + "epoch": 2.7, + "learning_rate": 4.9820466786355475e-06, + "loss": 0.2329, + "step": 70260 + }, + { + "epoch": 2.7, + "learning_rate": 4.9814054885868175e-06, + "loss": 0.1146, + "step": 70261 + }, + { + "epoch": 2.7, + "learning_rate": 4.980764298538087e-06, + "loss": 0.2372, + "step": 70262 + }, + { + "epoch": 2.7, + "learning_rate": 4.980123108489357e-06, + "loss": 0.1964, + "step": 70263 + }, + { + "epoch": 2.7, + "learning_rate": 4.979481918440626e-06, + "loss": 0.1992, + "step": 70264 + }, + { + "epoch": 2.7, + "learning_rate": 4.978840728391895e-06, + "loss": 0.1891, + "step": 70265 + }, + { + "epoch": 2.7, + "learning_rate": 4.978199538343165e-06, + "loss": 0.1619, + "step": 70266 + }, + { + "epoch": 2.7, + "learning_rate": 4.977558348294434e-06, + "loss": 0.3166, + "step": 70267 + }, + { + "epoch": 2.7, + "learning_rate": 4.976917158245704e-06, + "loss": 0.2358, + "step": 70268 + }, + { + "epoch": 2.7, + "learning_rate": 4.9762759681969735e-06, + "loss": 0.2183, + "step": 70269 + }, + { + "epoch": 2.7, + "learning_rate": 4.9756347781482435e-06, + "loss": 0.3535, + "step": 70270 + }, + { + "epoch": 2.7, + "learning_rate": 4.974993588099513e-06, + "loss": 0.0983, + "step": 70271 + }, + { + "epoch": 2.7, + "learning_rate": 4.974352398050783e-06, + "loss": 0.2304, + "step": 70272 + }, + { + "epoch": 2.7, + "learning_rate": 4.973711208002052e-06, + "loss": 0.1967, + "step": 70273 + }, + { + "epoch": 2.7, + "learning_rate": 4.973070017953321e-06, + "loss": 0.2138, + "step": 70274 + }, + { + "epoch": 2.7, + "learning_rate": 4.972428827904591e-06, + "loss": 0.1986, + "step": 70275 + }, + { + "epoch": 2.7, + "learning_rate": 4.97178763785586e-06, + "loss": 0.2052, + "step": 70276 + }, + { + "epoch": 2.7, + "learning_rate": 4.97114644780713e-06, + "loss": 0.3202, + "step": 70277 + }, + { + "epoch": 2.7, + "learning_rate": 4.9705052577583994e-06, + "loss": 0.1882, + "step": 70278 + }, + { + "epoch": 2.7, + "learning_rate": 4.9698640677096694e-06, + "loss": 0.1255, + "step": 70279 + }, + { + "epoch": 2.7, + "learning_rate": 4.969222877660939e-06, + "loss": 0.2131, + "step": 70280 + }, + { + "epoch": 2.7, + "learning_rate": 4.968581687612209e-06, + "loss": 0.2468, + "step": 70281 + }, + { + "epoch": 2.7, + "learning_rate": 4.967940497563478e-06, + "loss": 0.1117, + "step": 70282 + }, + { + "epoch": 2.7, + "learning_rate": 4.967299307514748e-06, + "loss": 0.2036, + "step": 70283 + }, + { + "epoch": 2.7, + "learning_rate": 4.966658117466017e-06, + "loss": 0.2452, + "step": 70284 + }, + { + "epoch": 2.7, + "learning_rate": 4.966016927417286e-06, + "loss": 0.114, + "step": 70285 + }, + { + "epoch": 2.7, + "learning_rate": 4.965375737368556e-06, + "loss": 0.1326, + "step": 70286 + }, + { + "epoch": 2.7, + "learning_rate": 4.964734547319825e-06, + "loss": 0.1583, + "step": 70287 + }, + { + "epoch": 2.7, + "learning_rate": 4.964093357271095e-06, + "loss": 0.2675, + "step": 70288 + }, + { + "epoch": 2.7, + "learning_rate": 4.963452167222365e-06, + "loss": 0.3476, + "step": 70289 + }, + { + "epoch": 2.7, + "learning_rate": 4.962810977173635e-06, + "loss": 0.1313, + "step": 70290 + }, + { + "epoch": 2.7, + "learning_rate": 4.962169787124904e-06, + "loss": 0.1798, + "step": 70291 + }, + { + "epoch": 2.7, + "learning_rate": 4.961528597076174e-06, + "loss": 0.1248, + "step": 70292 + }, + { + "epoch": 2.7, + "learning_rate": 4.960887407027443e-06, + "loss": 0.1861, + "step": 70293 + }, + { + "epoch": 2.7, + "learning_rate": 4.960246216978712e-06, + "loss": 0.144, + "step": 70294 + }, + { + "epoch": 2.7, + "learning_rate": 4.959605026929982e-06, + "loss": 0.2202, + "step": 70295 + }, + { + "epoch": 2.7, + "learning_rate": 4.958963836881251e-06, + "loss": 0.162, + "step": 70296 + }, + { + "epoch": 2.7, + "learning_rate": 4.958322646832521e-06, + "loss": 0.2683, + "step": 70297 + }, + { + "epoch": 2.7, + "learning_rate": 4.9576814567837906e-06, + "loss": 0.1122, + "step": 70298 + }, + { + "epoch": 2.7, + "learning_rate": 4.957040266735061e-06, + "loss": 0.2733, + "step": 70299 + }, + { + "epoch": 2.7, + "learning_rate": 4.95639907668633e-06, + "loss": 0.3291, + "step": 70300 + }, + { + "epoch": 2.7, + "learning_rate": 4.9557578866376e-06, + "loss": 0.2518, + "step": 70301 + }, + { + "epoch": 2.7, + "learning_rate": 4.955116696588869e-06, + "loss": 0.213, + "step": 70302 + }, + { + "epoch": 2.7, + "learning_rate": 4.954475506540138e-06, + "loss": 0.2455, + "step": 70303 + }, + { + "epoch": 2.7, + "learning_rate": 4.953834316491408e-06, + "loss": 0.2685, + "step": 70304 + }, + { + "epoch": 2.7, + "learning_rate": 4.953193126442677e-06, + "loss": 0.2079, + "step": 70305 + }, + { + "epoch": 2.7, + "learning_rate": 4.952551936393947e-06, + "loss": 0.1438, + "step": 70306 + }, + { + "epoch": 2.7, + "learning_rate": 4.9519107463452165e-06, + "loss": 0.1982, + "step": 70307 + }, + { + "epoch": 2.7, + "learning_rate": 4.9512695562964865e-06, + "loss": 0.1463, + "step": 70308 + }, + { + "epoch": 2.7, + "learning_rate": 4.950628366247756e-06, + "loss": 0.2429, + "step": 70309 + }, + { + "epoch": 2.7, + "learning_rate": 4.949987176199026e-06, + "loss": 0.151, + "step": 70310 + }, + { + "epoch": 2.7, + "learning_rate": 4.949345986150295e-06, + "loss": 0.2205, + "step": 70311 + }, + { + "epoch": 2.7, + "learning_rate": 4.948704796101565e-06, + "loss": 0.3095, + "step": 70312 + }, + { + "epoch": 2.7, + "learning_rate": 4.948063606052834e-06, + "loss": 0.1639, + "step": 70313 + }, + { + "epoch": 2.7, + "learning_rate": 4.947422416004103e-06, + "loss": 0.3173, + "step": 70314 + }, + { + "epoch": 2.7, + "learning_rate": 4.946781225955373e-06, + "loss": 0.1605, + "step": 70315 + }, + { + "epoch": 2.7, + "learning_rate": 4.9461400359066425e-06, + "loss": 0.244, + "step": 70316 + }, + { + "epoch": 2.7, + "learning_rate": 4.9454988458579125e-06, + "loss": 0.1432, + "step": 70317 + }, + { + "epoch": 2.7, + "learning_rate": 4.944857655809182e-06, + "loss": 0.1688, + "step": 70318 + }, + { + "epoch": 2.7, + "learning_rate": 4.944216465760452e-06, + "loss": 0.148, + "step": 70319 + }, + { + "epoch": 2.7, + "learning_rate": 4.943575275711721e-06, + "loss": 0.0677, + "step": 70320 + }, + { + "epoch": 2.7, + "learning_rate": 4.942934085662991e-06, + "loss": 0.1231, + "step": 70321 + }, + { + "epoch": 2.7, + "learning_rate": 4.94229289561426e-06, + "loss": 0.178, + "step": 70322 + }, + { + "epoch": 2.7, + "learning_rate": 4.941651705565529e-06, + "loss": 0.3856, + "step": 70323 + }, + { + "epoch": 2.7, + "learning_rate": 4.941010515516799e-06, + "loss": 0.1757, + "step": 70324 + }, + { + "epoch": 2.7, + "learning_rate": 4.9403693254680685e-06, + "loss": 0.0861, + "step": 70325 + }, + { + "epoch": 2.7, + "learning_rate": 4.9397281354193385e-06, + "loss": 0.3427, + "step": 70326 + }, + { + "epoch": 2.7, + "learning_rate": 4.939086945370608e-06, + "loss": 0.1119, + "step": 70327 + }, + { + "epoch": 2.7, + "learning_rate": 4.938445755321878e-06, + "loss": 0.3116, + "step": 70328 + }, + { + "epoch": 2.7, + "learning_rate": 4.937804565273147e-06, + "loss": 0.2362, + "step": 70329 + }, + { + "epoch": 2.7, + "learning_rate": 4.937163375224417e-06, + "loss": 0.2448, + "step": 70330 + }, + { + "epoch": 2.7, + "learning_rate": 4.936522185175686e-06, + "loss": 0.2335, + "step": 70331 + }, + { + "epoch": 2.7, + "learning_rate": 4.935880995126956e-06, + "loss": 0.2671, + "step": 70332 + }, + { + "epoch": 2.7, + "learning_rate": 4.935239805078225e-06, + "loss": 0.2865, + "step": 70333 + }, + { + "epoch": 2.7, + "learning_rate": 4.9345986150294944e-06, + "loss": 0.2263, + "step": 70334 + }, + { + "epoch": 2.7, + "learning_rate": 4.9339574249807645e-06, + "loss": 0.1361, + "step": 70335 + }, + { + "epoch": 2.7, + "learning_rate": 4.933316234932034e-06, + "loss": 0.2335, + "step": 70336 + }, + { + "epoch": 2.7, + "learning_rate": 4.932675044883304e-06, + "loss": 0.4016, + "step": 70337 + }, + { + "epoch": 2.7, + "learning_rate": 4.932033854834573e-06, + "loss": 0.1605, + "step": 70338 + }, + { + "epoch": 2.7, + "learning_rate": 4.931392664785843e-06, + "loss": 0.2862, + "step": 70339 + }, + { + "epoch": 2.7, + "learning_rate": 4.930751474737112e-06, + "loss": 0.2205, + "step": 70340 + }, + { + "epoch": 2.7, + "learning_rate": 4.930110284688382e-06, + "loss": 0.2731, + "step": 70341 + }, + { + "epoch": 2.7, + "learning_rate": 4.929469094639651e-06, + "loss": 0.3463, + "step": 70342 + }, + { + "epoch": 2.7, + "learning_rate": 4.92882790459092e-06, + "loss": 0.1666, + "step": 70343 + }, + { + "epoch": 2.7, + "learning_rate": 4.92818671454219e-06, + "loss": 0.1396, + "step": 70344 + }, + { + "epoch": 2.7, + "learning_rate": 4.92754552449346e-06, + "loss": 0.2184, + "step": 70345 + }, + { + "epoch": 2.7, + "learning_rate": 4.92690433444473e-06, + "loss": 0.1877, + "step": 70346 + }, + { + "epoch": 2.7, + "learning_rate": 4.926263144395999e-06, + "loss": 0.3656, + "step": 70347 + }, + { + "epoch": 2.7, + "learning_rate": 4.925621954347269e-06, + "loss": 0.1533, + "step": 70348 + }, + { + "epoch": 2.7, + "learning_rate": 4.924980764298538e-06, + "loss": 0.2954, + "step": 70349 + }, + { + "epoch": 2.7, + "learning_rate": 4.924339574249808e-06, + "loss": 0.3167, + "step": 70350 + }, + { + "epoch": 2.7, + "learning_rate": 4.923698384201077e-06, + "loss": 0.1976, + "step": 70351 + }, + { + "epoch": 2.7, + "learning_rate": 4.923057194152346e-06, + "loss": 0.3805, + "step": 70352 + }, + { + "epoch": 2.7, + "learning_rate": 4.922416004103616e-06, + "loss": 0.1363, + "step": 70353 + }, + { + "epoch": 2.7, + "learning_rate": 4.9217748140548856e-06, + "loss": 0.1892, + "step": 70354 + }, + { + "epoch": 2.7, + "learning_rate": 4.921133624006156e-06, + "loss": 0.2772, + "step": 70355 + }, + { + "epoch": 2.7, + "learning_rate": 4.920492433957425e-06, + "loss": 0.1733, + "step": 70356 + }, + { + "epoch": 2.7, + "learning_rate": 4.919851243908695e-06, + "loss": 0.1894, + "step": 70357 + }, + { + "epoch": 2.71, + "learning_rate": 4.919210053859964e-06, + "loss": 0.1278, + "step": 70358 + }, + { + "epoch": 2.71, + "learning_rate": 4.918568863811234e-06, + "loss": 0.2592, + "step": 70359 + }, + { + "epoch": 2.71, + "learning_rate": 4.917927673762503e-06, + "loss": 0.2194, + "step": 70360 + }, + { + "epoch": 2.71, + "learning_rate": 4.917286483713773e-06, + "loss": 0.1531, + "step": 70361 + }, + { + "epoch": 2.71, + "learning_rate": 4.916645293665042e-06, + "loss": 0.1367, + "step": 70362 + }, + { + "epoch": 2.71, + "learning_rate": 4.9160041036163115e-06, + "loss": 0.1622, + "step": 70363 + }, + { + "epoch": 2.71, + "learning_rate": 4.9153629135675816e-06, + "loss": 0.1853, + "step": 70364 + }, + { + "epoch": 2.71, + "learning_rate": 4.914721723518851e-06, + "loss": 0.2571, + "step": 70365 + }, + { + "epoch": 2.71, + "learning_rate": 4.914080533470121e-06, + "loss": 0.2104, + "step": 70366 + }, + { + "epoch": 2.71, + "learning_rate": 4.91343934342139e-06, + "loss": 0.0893, + "step": 70367 + }, + { + "epoch": 2.71, + "learning_rate": 4.91279815337266e-06, + "loss": 0.2319, + "step": 70368 + }, + { + "epoch": 2.71, + "learning_rate": 4.912156963323929e-06, + "loss": 0.1993, + "step": 70369 + }, + { + "epoch": 2.71, + "learning_rate": 4.911515773275199e-06, + "loss": 0.2249, + "step": 70370 + }, + { + "epoch": 2.71, + "learning_rate": 4.910874583226468e-06, + "loss": 0.2272, + "step": 70371 + }, + { + "epoch": 2.71, + "learning_rate": 4.9102333931777375e-06, + "loss": 0.154, + "step": 70372 + }, + { + "epoch": 2.71, + "learning_rate": 4.9095922031290075e-06, + "loss": 0.2342, + "step": 70373 + }, + { + "epoch": 2.71, + "learning_rate": 4.908951013080277e-06, + "loss": 0.1892, + "step": 70374 + }, + { + "epoch": 2.71, + "learning_rate": 4.908309823031547e-06, + "loss": 0.1075, + "step": 70375 + }, + { + "epoch": 2.71, + "learning_rate": 4.907668632982816e-06, + "loss": 0.1132, + "step": 70376 + }, + { + "epoch": 2.71, + "learning_rate": 4.907027442934086e-06, + "loss": 0.1184, + "step": 70377 + }, + { + "epoch": 2.71, + "learning_rate": 4.906386252885355e-06, + "loss": 0.2745, + "step": 70378 + }, + { + "epoch": 2.71, + "learning_rate": 4.905745062836625e-06, + "loss": 0.3462, + "step": 70379 + }, + { + "epoch": 2.71, + "learning_rate": 4.905103872787894e-06, + "loss": 0.2487, + "step": 70380 + }, + { + "epoch": 2.71, + "learning_rate": 4.904462682739164e-06, + "loss": 0.1698, + "step": 70381 + }, + { + "epoch": 2.71, + "learning_rate": 4.9038214926904335e-06, + "loss": 0.1142, + "step": 70382 + }, + { + "epoch": 2.71, + "learning_rate": 4.903180302641703e-06, + "loss": 0.1988, + "step": 70383 + }, + { + "epoch": 2.71, + "learning_rate": 4.902539112592973e-06, + "loss": 0.2315, + "step": 70384 + }, + { + "epoch": 2.71, + "learning_rate": 4.901897922544242e-06, + "loss": 0.15, + "step": 70385 + }, + { + "epoch": 2.71, + "learning_rate": 4.901256732495512e-06, + "loss": 0.1336, + "step": 70386 + }, + { + "epoch": 2.71, + "learning_rate": 4.900615542446781e-06, + "loss": 0.1646, + "step": 70387 + }, + { + "epoch": 2.71, + "learning_rate": 4.899974352398051e-06, + "loss": 0.1462, + "step": 70388 + }, + { + "epoch": 2.71, + "learning_rate": 4.89933316234932e-06, + "loss": 0.2844, + "step": 70389 + }, + { + "epoch": 2.71, + "learning_rate": 4.89869197230059e-06, + "loss": 0.1378, + "step": 70390 + }, + { + "epoch": 2.71, + "learning_rate": 4.8980507822518595e-06, + "loss": 0.2313, + "step": 70391 + }, + { + "epoch": 2.71, + "learning_rate": 4.897409592203129e-06, + "loss": 0.1957, + "step": 70392 + }, + { + "epoch": 2.71, + "learning_rate": 4.896768402154399e-06, + "loss": 0.1715, + "step": 70393 + }, + { + "epoch": 2.71, + "learning_rate": 4.896127212105668e-06, + "loss": 0.2244, + "step": 70394 + }, + { + "epoch": 2.71, + "learning_rate": 4.895486022056938e-06, + "loss": 0.2617, + "step": 70395 + }, + { + "epoch": 2.71, + "learning_rate": 4.894844832008207e-06, + "loss": 0.2015, + "step": 70396 + }, + { + "epoch": 2.71, + "learning_rate": 4.894203641959477e-06, + "loss": 0.113, + "step": 70397 + }, + { + "epoch": 2.71, + "learning_rate": 4.893562451910746e-06, + "loss": 0.3548, + "step": 70398 + }, + { + "epoch": 2.71, + "learning_rate": 4.892921261862016e-06, + "loss": 0.1001, + "step": 70399 + }, + { + "epoch": 2.71, + "learning_rate": 4.8922800718132854e-06, + "loss": 0.2476, + "step": 70400 + }, + { + "epoch": 2.71, + "learning_rate": 4.8916388817645554e-06, + "loss": 0.1486, + "step": 70401 + }, + { + "epoch": 2.71, + "learning_rate": 4.890997691715825e-06, + "loss": 0.2444, + "step": 70402 + }, + { + "epoch": 2.71, + "learning_rate": 4.890356501667094e-06, + "loss": 0.2242, + "step": 70403 + }, + { + "epoch": 2.71, + "learning_rate": 4.889715311618364e-06, + "loss": 0.1966, + "step": 70404 + }, + { + "epoch": 2.71, + "learning_rate": 4.889074121569633e-06, + "loss": 0.1167, + "step": 70405 + }, + { + "epoch": 2.71, + "learning_rate": 4.888432931520903e-06, + "loss": 0.1171, + "step": 70406 + }, + { + "epoch": 2.71, + "learning_rate": 4.887791741472172e-06, + "loss": 0.322, + "step": 70407 + }, + { + "epoch": 2.71, + "learning_rate": 4.887150551423442e-06, + "loss": 0.0841, + "step": 70408 + }, + { + "epoch": 2.71, + "learning_rate": 4.886509361374711e-06, + "loss": 0.1283, + "step": 70409 + }, + { + "epoch": 2.71, + "learning_rate": 4.885868171325981e-06, + "loss": 0.2772, + "step": 70410 + }, + { + "epoch": 2.71, + "learning_rate": 4.885226981277251e-06, + "loss": 0.1906, + "step": 70411 + }, + { + "epoch": 2.71, + "learning_rate": 4.88458579122852e-06, + "loss": 0.198, + "step": 70412 + }, + { + "epoch": 2.71, + "learning_rate": 4.88394460117979e-06, + "loss": 0.3038, + "step": 70413 + }, + { + "epoch": 2.71, + "learning_rate": 4.883303411131059e-06, + "loss": 0.1741, + "step": 70414 + }, + { + "epoch": 2.71, + "learning_rate": 4.882662221082329e-06, + "loss": 0.1077, + "step": 70415 + }, + { + "epoch": 2.71, + "learning_rate": 4.882021031033598e-06, + "loss": 0.1342, + "step": 70416 + }, + { + "epoch": 2.71, + "learning_rate": 4.881379840984868e-06, + "loss": 0.2533, + "step": 70417 + }, + { + "epoch": 2.71, + "learning_rate": 4.880738650936137e-06, + "loss": 0.1948, + "step": 70418 + }, + { + "epoch": 2.71, + "learning_rate": 4.880097460887407e-06, + "loss": 0.1141, + "step": 70419 + }, + { + "epoch": 2.71, + "learning_rate": 4.8794562708386766e-06, + "loss": 0.1663, + "step": 70420 + }, + { + "epoch": 2.71, + "learning_rate": 4.878815080789946e-06, + "loss": 0.1071, + "step": 70421 + }, + { + "epoch": 2.71, + "learning_rate": 4.878173890741216e-06, + "loss": 0.2031, + "step": 70422 + }, + { + "epoch": 2.71, + "learning_rate": 4.877532700692485e-06, + "loss": 0.3413, + "step": 70423 + }, + { + "epoch": 2.71, + "learning_rate": 4.876891510643755e-06, + "loss": 0.1925, + "step": 70424 + }, + { + "epoch": 2.71, + "learning_rate": 4.876250320595024e-06, + "loss": 0.2377, + "step": 70425 + }, + { + "epoch": 2.71, + "learning_rate": 4.875609130546294e-06, + "loss": 0.3493, + "step": 70426 + }, + { + "epoch": 2.71, + "learning_rate": 4.874967940497563e-06, + "loss": 0.3767, + "step": 70427 + }, + { + "epoch": 2.71, + "learning_rate": 4.874326750448833e-06, + "loss": 0.1225, + "step": 70428 + }, + { + "epoch": 2.71, + "learning_rate": 4.8736855604001025e-06, + "loss": 0.0873, + "step": 70429 + }, + { + "epoch": 2.71, + "learning_rate": 4.8730443703513725e-06, + "loss": 0.1716, + "step": 70430 + }, + { + "epoch": 2.71, + "learning_rate": 4.872403180302642e-06, + "loss": 0.2766, + "step": 70431 + }, + { + "epoch": 2.71, + "learning_rate": 4.871761990253911e-06, + "loss": 0.3411, + "step": 70432 + }, + { + "epoch": 2.71, + "learning_rate": 4.871120800205181e-06, + "loss": 0.167, + "step": 70433 + }, + { + "epoch": 2.71, + "learning_rate": 4.87047961015645e-06, + "loss": 0.151, + "step": 70434 + }, + { + "epoch": 2.71, + "learning_rate": 4.86983842010772e-06, + "loss": 0.353, + "step": 70435 + }, + { + "epoch": 2.71, + "learning_rate": 4.869197230058989e-06, + "loss": 0.4012, + "step": 70436 + }, + { + "epoch": 2.71, + "learning_rate": 4.868556040010259e-06, + "loss": 0.2452, + "step": 70437 + }, + { + "epoch": 2.71, + "learning_rate": 4.8679148499615285e-06, + "loss": 0.2119, + "step": 70438 + }, + { + "epoch": 2.71, + "learning_rate": 4.8672736599127985e-06, + "loss": 0.1554, + "step": 70439 + }, + { + "epoch": 2.71, + "learning_rate": 4.866632469864068e-06, + "loss": 0.2103, + "step": 70440 + }, + { + "epoch": 2.71, + "learning_rate": 4.865991279815337e-06, + "loss": 0.1696, + "step": 70441 + }, + { + "epoch": 2.71, + "learning_rate": 4.865350089766607e-06, + "loss": 0.1617, + "step": 70442 + }, + { + "epoch": 2.71, + "learning_rate": 4.864708899717876e-06, + "loss": 0.2769, + "step": 70443 + }, + { + "epoch": 2.71, + "learning_rate": 4.864067709669146e-06, + "loss": 0.112, + "step": 70444 + }, + { + "epoch": 2.71, + "learning_rate": 4.863426519620415e-06, + "loss": 0.3369, + "step": 70445 + }, + { + "epoch": 2.71, + "learning_rate": 4.862785329571685e-06, + "loss": 0.2113, + "step": 70446 + }, + { + "epoch": 2.71, + "learning_rate": 4.8621441395229545e-06, + "loss": 0.2217, + "step": 70447 + }, + { + "epoch": 2.71, + "learning_rate": 4.8615029494742245e-06, + "loss": 0.1964, + "step": 70448 + }, + { + "epoch": 2.71, + "learning_rate": 4.860861759425494e-06, + "loss": 0.2192, + "step": 70449 + }, + { + "epoch": 2.71, + "learning_rate": 4.860220569376764e-06, + "loss": 0.2727, + "step": 70450 + }, + { + "epoch": 2.71, + "learning_rate": 4.859579379328033e-06, + "loss": 0.2325, + "step": 70451 + }, + { + "epoch": 2.71, + "learning_rate": 4.858938189279302e-06, + "loss": 0.271, + "step": 70452 + }, + { + "epoch": 2.71, + "learning_rate": 4.858296999230572e-06, + "loss": 0.151, + "step": 70453 + }, + { + "epoch": 2.71, + "learning_rate": 4.857655809181841e-06, + "loss": 0.1507, + "step": 70454 + }, + { + "epoch": 2.71, + "learning_rate": 4.857014619133111e-06, + "loss": 0.1624, + "step": 70455 + }, + { + "epoch": 2.71, + "learning_rate": 4.8563734290843804e-06, + "loss": 0.1912, + "step": 70456 + }, + { + "epoch": 2.71, + "learning_rate": 4.8557322390356505e-06, + "loss": 0.3892, + "step": 70457 + }, + { + "epoch": 2.71, + "learning_rate": 4.85509104898692e-06, + "loss": 0.1523, + "step": 70458 + }, + { + "epoch": 2.71, + "learning_rate": 4.85444985893819e-06, + "loss": 0.2755, + "step": 70459 + }, + { + "epoch": 2.71, + "learning_rate": 4.853808668889459e-06, + "loss": 0.2149, + "step": 70460 + }, + { + "epoch": 2.71, + "learning_rate": 4.853167478840728e-06, + "loss": 0.4942, + "step": 70461 + }, + { + "epoch": 2.71, + "learning_rate": 4.852526288791998e-06, + "loss": 0.2347, + "step": 70462 + }, + { + "epoch": 2.71, + "learning_rate": 4.851885098743267e-06, + "loss": 0.1755, + "step": 70463 + }, + { + "epoch": 2.71, + "learning_rate": 4.851243908694537e-06, + "loss": 0.1847, + "step": 70464 + }, + { + "epoch": 2.71, + "learning_rate": 4.850602718645806e-06, + "loss": 0.2367, + "step": 70465 + }, + { + "epoch": 2.71, + "learning_rate": 4.849961528597076e-06, + "loss": 0.3625, + "step": 70466 + }, + { + "epoch": 2.71, + "learning_rate": 4.849320338548346e-06, + "loss": 0.1338, + "step": 70467 + }, + { + "epoch": 2.71, + "learning_rate": 4.848679148499616e-06, + "loss": 0.1783, + "step": 70468 + }, + { + "epoch": 2.71, + "learning_rate": 4.848037958450885e-06, + "loss": 0.2192, + "step": 70469 + }, + { + "epoch": 2.71, + "learning_rate": 4.847396768402154e-06, + "loss": 0.1368, + "step": 70470 + }, + { + "epoch": 2.71, + "learning_rate": 4.846755578353424e-06, + "loss": 0.203, + "step": 70471 + }, + { + "epoch": 2.71, + "learning_rate": 4.846114388304693e-06, + "loss": 0.2147, + "step": 70472 + }, + { + "epoch": 2.71, + "learning_rate": 4.845473198255963e-06, + "loss": 0.1708, + "step": 70473 + }, + { + "epoch": 2.71, + "learning_rate": 4.844832008207232e-06, + "loss": 0.2326, + "step": 70474 + }, + { + "epoch": 2.71, + "learning_rate": 4.844190818158502e-06, + "loss": 0.2629, + "step": 70475 + }, + { + "epoch": 2.71, + "learning_rate": 4.8435496281097716e-06, + "loss": 0.2848, + "step": 70476 + }, + { + "epoch": 2.71, + "learning_rate": 4.842908438061042e-06, + "loss": 0.279, + "step": 70477 + }, + { + "epoch": 2.71, + "learning_rate": 4.842267248012311e-06, + "loss": 0.1957, + "step": 70478 + }, + { + "epoch": 2.71, + "learning_rate": 4.841626057963581e-06, + "loss": 0.3872, + "step": 70479 + }, + { + "epoch": 2.71, + "learning_rate": 4.84098486791485e-06, + "loss": 0.2926, + "step": 70480 + }, + { + "epoch": 2.71, + "learning_rate": 4.840343677866119e-06, + "loss": 0.3018, + "step": 70481 + }, + { + "epoch": 2.71, + "learning_rate": 4.839702487817389e-06, + "loss": 0.175, + "step": 70482 + }, + { + "epoch": 2.71, + "learning_rate": 4.839061297768658e-06, + "loss": 0.2475, + "step": 70483 + }, + { + "epoch": 2.71, + "learning_rate": 4.838420107719928e-06, + "loss": 0.2181, + "step": 70484 + }, + { + "epoch": 2.71, + "learning_rate": 4.8377789176711975e-06, + "loss": 0.2568, + "step": 70485 + }, + { + "epoch": 2.71, + "learning_rate": 4.8371377276224676e-06, + "loss": 0.3011, + "step": 70486 + }, + { + "epoch": 2.71, + "learning_rate": 4.836496537573737e-06, + "loss": 0.1468, + "step": 70487 + }, + { + "epoch": 2.71, + "learning_rate": 4.835855347525007e-06, + "loss": 0.3631, + "step": 70488 + }, + { + "epoch": 2.71, + "learning_rate": 4.835214157476276e-06, + "loss": 0.2592, + "step": 70489 + }, + { + "epoch": 2.71, + "learning_rate": 4.834572967427545e-06, + "loss": 0.1786, + "step": 70490 + }, + { + "epoch": 2.71, + "learning_rate": 4.833931777378815e-06, + "loss": 0.1169, + "step": 70491 + }, + { + "epoch": 2.71, + "learning_rate": 4.833290587330084e-06, + "loss": 0.1648, + "step": 70492 + }, + { + "epoch": 2.71, + "learning_rate": 4.832649397281354e-06, + "loss": 0.218, + "step": 70493 + }, + { + "epoch": 2.71, + "learning_rate": 4.8320082072326235e-06, + "loss": 0.1411, + "step": 70494 + }, + { + "epoch": 2.71, + "learning_rate": 4.8313670171838935e-06, + "loss": 0.17, + "step": 70495 + }, + { + "epoch": 2.71, + "learning_rate": 4.830725827135163e-06, + "loss": 0.1532, + "step": 70496 + }, + { + "epoch": 2.71, + "learning_rate": 4.830084637086433e-06, + "loss": 0.1817, + "step": 70497 + }, + { + "epoch": 2.71, + "learning_rate": 4.829443447037702e-06, + "loss": 0.1782, + "step": 70498 + }, + { + "epoch": 2.71, + "learning_rate": 4.828802256988972e-06, + "loss": 0.2759, + "step": 70499 + }, + { + "epoch": 2.71, + "learning_rate": 4.828161066940241e-06, + "loss": 0.2587, + "step": 70500 + }, + { + "epoch": 2.71, + "learning_rate": 4.82751987689151e-06, + "loss": 0.2296, + "step": 70501 + }, + { + "epoch": 2.71, + "learning_rate": 4.82687868684278e-06, + "loss": 0.235, + "step": 70502 + }, + { + "epoch": 2.71, + "learning_rate": 4.8262374967940495e-06, + "loss": 0.1876, + "step": 70503 + }, + { + "epoch": 2.71, + "learning_rate": 4.8255963067453195e-06, + "loss": 0.0887, + "step": 70504 + }, + { + "epoch": 2.71, + "learning_rate": 4.824955116696589e-06, + "loss": 0.2126, + "step": 70505 + }, + { + "epoch": 2.71, + "learning_rate": 4.824313926647859e-06, + "loss": 0.1956, + "step": 70506 + }, + { + "epoch": 2.71, + "learning_rate": 4.823672736599128e-06, + "loss": 0.2911, + "step": 70507 + }, + { + "epoch": 2.71, + "learning_rate": 4.823031546550398e-06, + "loss": 0.2704, + "step": 70508 + }, + { + "epoch": 2.71, + "learning_rate": 4.822390356501667e-06, + "loss": 0.2522, + "step": 70509 + }, + { + "epoch": 2.71, + "learning_rate": 4.821749166452936e-06, + "loss": 0.1629, + "step": 70510 + }, + { + "epoch": 2.71, + "learning_rate": 4.821107976404206e-06, + "loss": 0.1879, + "step": 70511 + }, + { + "epoch": 2.71, + "learning_rate": 4.8204667863554754e-06, + "loss": 0.1439, + "step": 70512 + }, + { + "epoch": 2.71, + "learning_rate": 4.8198255963067455e-06, + "loss": 0.1908, + "step": 70513 + }, + { + "epoch": 2.71, + "learning_rate": 4.819184406258015e-06, + "loss": 0.1525, + "step": 70514 + }, + { + "epoch": 2.71, + "learning_rate": 4.818543216209285e-06, + "loss": 0.3163, + "step": 70515 + }, + { + "epoch": 2.71, + "learning_rate": 4.817902026160554e-06, + "loss": 0.1733, + "step": 70516 + }, + { + "epoch": 2.71, + "learning_rate": 4.817260836111824e-06, + "loss": 0.1803, + "step": 70517 + }, + { + "epoch": 2.71, + "learning_rate": 4.816619646063093e-06, + "loss": 0.1772, + "step": 70518 + }, + { + "epoch": 2.71, + "learning_rate": 4.815978456014363e-06, + "loss": 0.329, + "step": 70519 + }, + { + "epoch": 2.71, + "learning_rate": 4.815337265965632e-06, + "loss": 0.1451, + "step": 70520 + }, + { + "epoch": 2.71, + "learning_rate": 4.814696075916901e-06, + "loss": 0.166, + "step": 70521 + }, + { + "epoch": 2.71, + "learning_rate": 4.8140548858681714e-06, + "loss": 0.2502, + "step": 70522 + }, + { + "epoch": 2.71, + "learning_rate": 4.813413695819441e-06, + "loss": 0.2158, + "step": 70523 + }, + { + "epoch": 2.71, + "learning_rate": 4.812772505770711e-06, + "loss": 0.1327, + "step": 70524 + }, + { + "epoch": 2.71, + "learning_rate": 4.81213131572198e-06, + "loss": 0.0857, + "step": 70525 + }, + { + "epoch": 2.71, + "learning_rate": 4.81149012567325e-06, + "loss": 0.1249, + "step": 70526 + }, + { + "epoch": 2.71, + "learning_rate": 4.810848935624519e-06, + "loss": 0.1833, + "step": 70527 + }, + { + "epoch": 2.71, + "learning_rate": 4.810207745575789e-06, + "loss": 0.2104, + "step": 70528 + }, + { + "epoch": 2.71, + "learning_rate": 4.809566555527058e-06, + "loss": 0.0947, + "step": 70529 + }, + { + "epoch": 2.71, + "learning_rate": 4.808925365478327e-06, + "loss": 0.1281, + "step": 70530 + }, + { + "epoch": 2.71, + "learning_rate": 4.808284175429597e-06, + "loss": 0.1424, + "step": 70531 + }, + { + "epoch": 2.71, + "learning_rate": 4.8076429853808666e-06, + "loss": 0.1116, + "step": 70532 + }, + { + "epoch": 2.71, + "learning_rate": 4.807001795332137e-06, + "loss": 0.1627, + "step": 70533 + }, + { + "epoch": 2.71, + "learning_rate": 4.806360605283406e-06, + "loss": 0.1867, + "step": 70534 + }, + { + "epoch": 2.71, + "learning_rate": 4.805719415234676e-06, + "loss": 0.2882, + "step": 70535 + }, + { + "epoch": 2.71, + "learning_rate": 4.805078225185945e-06, + "loss": 0.2801, + "step": 70536 + }, + { + "epoch": 2.71, + "learning_rate": 4.804437035137215e-06, + "loss": 0.2038, + "step": 70537 + }, + { + "epoch": 2.71, + "learning_rate": 4.803795845088484e-06, + "loss": 0.1534, + "step": 70538 + }, + { + "epoch": 2.71, + "learning_rate": 4.803154655039753e-06, + "loss": 0.2396, + "step": 70539 + }, + { + "epoch": 2.71, + "learning_rate": 4.802513464991023e-06, + "loss": 0.1309, + "step": 70540 + }, + { + "epoch": 2.71, + "learning_rate": 4.8018722749422925e-06, + "loss": 0.253, + "step": 70541 + }, + { + "epoch": 2.71, + "learning_rate": 4.8012310848935626e-06, + "loss": 0.1222, + "step": 70542 + }, + { + "epoch": 2.71, + "learning_rate": 4.800589894844832e-06, + "loss": 0.222, + "step": 70543 + }, + { + "epoch": 2.71, + "learning_rate": 4.799948704796102e-06, + "loss": 0.3185, + "step": 70544 + }, + { + "epoch": 2.71, + "learning_rate": 4.799307514747371e-06, + "loss": 0.2487, + "step": 70545 + }, + { + "epoch": 2.71, + "learning_rate": 4.798666324698641e-06, + "loss": 0.203, + "step": 70546 + }, + { + "epoch": 2.71, + "learning_rate": 4.79802513464991e-06, + "loss": 0.2013, + "step": 70547 + }, + { + "epoch": 2.71, + "learning_rate": 4.79738394460118e-06, + "loss": 0.1711, + "step": 70548 + }, + { + "epoch": 2.71, + "learning_rate": 4.796742754552449e-06, + "loss": 0.1508, + "step": 70549 + }, + { + "epoch": 2.71, + "learning_rate": 4.7961015645037185e-06, + "loss": 0.2918, + "step": 70550 + }, + { + "epoch": 2.71, + "learning_rate": 4.7954603744549885e-06, + "loss": 0.3872, + "step": 70551 + }, + { + "epoch": 2.71, + "learning_rate": 4.794819184406258e-06, + "loss": 0.247, + "step": 70552 + }, + { + "epoch": 2.71, + "learning_rate": 4.794177994357528e-06, + "loss": 0.2373, + "step": 70553 + }, + { + "epoch": 2.71, + "learning_rate": 4.793536804308797e-06, + "loss": 0.2737, + "step": 70554 + }, + { + "epoch": 2.71, + "learning_rate": 4.792895614260067e-06, + "loss": 0.2302, + "step": 70555 + }, + { + "epoch": 2.71, + "learning_rate": 4.792254424211336e-06, + "loss": 0.2407, + "step": 70556 + }, + { + "epoch": 2.71, + "learning_rate": 4.791613234162606e-06, + "loss": 0.2071, + "step": 70557 + }, + { + "epoch": 2.71, + "learning_rate": 4.790972044113875e-06, + "loss": 0.106, + "step": 70558 + }, + { + "epoch": 2.71, + "learning_rate": 4.7903308540651445e-06, + "loss": 0.3229, + "step": 70559 + }, + { + "epoch": 2.71, + "learning_rate": 4.7896896640164145e-06, + "loss": 0.246, + "step": 70560 + }, + { + "epoch": 2.71, + "learning_rate": 4.789048473967684e-06, + "loss": 0.2299, + "step": 70561 + }, + { + "epoch": 2.71, + "learning_rate": 4.788407283918954e-06, + "loss": 0.0935, + "step": 70562 + }, + { + "epoch": 2.71, + "learning_rate": 4.787766093870223e-06, + "loss": 0.3328, + "step": 70563 + }, + { + "epoch": 2.71, + "learning_rate": 4.787124903821493e-06, + "loss": 0.2827, + "step": 70564 + }, + { + "epoch": 2.71, + "learning_rate": 4.786483713772762e-06, + "loss": 0.153, + "step": 70565 + }, + { + "epoch": 2.71, + "learning_rate": 4.785842523724032e-06, + "loss": 0.203, + "step": 70566 + }, + { + "epoch": 2.71, + "learning_rate": 4.785201333675301e-06, + "loss": 0.2952, + "step": 70567 + }, + { + "epoch": 2.71, + "learning_rate": 4.784560143626571e-06, + "loss": 0.1316, + "step": 70568 + }, + { + "epoch": 2.71, + "learning_rate": 4.7839189535778405e-06, + "loss": 0.3274, + "step": 70569 + }, + { + "epoch": 2.71, + "learning_rate": 4.78327776352911e-06, + "loss": 0.1785, + "step": 70570 + }, + { + "epoch": 2.71, + "learning_rate": 4.78263657348038e-06, + "loss": 0.1573, + "step": 70571 + }, + { + "epoch": 2.71, + "learning_rate": 4.781995383431649e-06, + "loss": 0.2295, + "step": 70572 + }, + { + "epoch": 2.71, + "learning_rate": 4.781354193382919e-06, + "loss": 0.1127, + "step": 70573 + }, + { + "epoch": 2.71, + "learning_rate": 4.780713003334188e-06, + "loss": 0.2581, + "step": 70574 + }, + { + "epoch": 2.71, + "learning_rate": 4.780071813285458e-06, + "loss": 0.2696, + "step": 70575 + }, + { + "epoch": 2.71, + "learning_rate": 4.779430623236727e-06, + "loss": 0.2058, + "step": 70576 + }, + { + "epoch": 2.71, + "learning_rate": 4.778789433187997e-06, + "loss": 0.2239, + "step": 70577 + }, + { + "epoch": 2.71, + "learning_rate": 4.7781482431392664e-06, + "loss": 0.2734, + "step": 70578 + }, + { + "epoch": 2.71, + "learning_rate": 4.777507053090536e-06, + "loss": 0.1985, + "step": 70579 + }, + { + "epoch": 2.71, + "learning_rate": 4.776865863041806e-06, + "loss": 0.1668, + "step": 70580 + }, + { + "epoch": 2.71, + "learning_rate": 4.776224672993075e-06, + "loss": 0.3853, + "step": 70581 + }, + { + "epoch": 2.71, + "learning_rate": 4.775583482944345e-06, + "loss": 0.3202, + "step": 70582 + }, + { + "epoch": 2.71, + "learning_rate": 4.774942292895614e-06, + "loss": 0.1507, + "step": 70583 + }, + { + "epoch": 2.71, + "learning_rate": 4.774301102846884e-06, + "loss": 0.143, + "step": 70584 + }, + { + "epoch": 2.71, + "learning_rate": 4.773659912798153e-06, + "loss": 0.1375, + "step": 70585 + }, + { + "epoch": 2.71, + "learning_rate": 4.773018722749423e-06, + "loss": 0.1134, + "step": 70586 + }, + { + "epoch": 2.71, + "learning_rate": 4.772377532700692e-06, + "loss": 0.182, + "step": 70587 + }, + { + "epoch": 2.71, + "learning_rate": 4.7717363426519616e-06, + "loss": 0.2275, + "step": 70588 + }, + { + "epoch": 2.71, + "learning_rate": 4.771095152603232e-06, + "loss": 0.125, + "step": 70589 + }, + { + "epoch": 2.71, + "learning_rate": 4.770453962554501e-06, + "loss": 0.2029, + "step": 70590 + }, + { + "epoch": 2.71, + "learning_rate": 4.769812772505771e-06, + "loss": 0.3062, + "step": 70591 + }, + { + "epoch": 2.71, + "learning_rate": 4.76917158245704e-06, + "loss": 0.1619, + "step": 70592 + }, + { + "epoch": 2.71, + "learning_rate": 4.76853039240831e-06, + "loss": 0.2057, + "step": 70593 + }, + { + "epoch": 2.71, + "learning_rate": 4.767889202359579e-06, + "loss": 0.1837, + "step": 70594 + }, + { + "epoch": 2.71, + "learning_rate": 4.767248012310849e-06, + "loss": 0.2179, + "step": 70595 + }, + { + "epoch": 2.71, + "learning_rate": 4.766606822262118e-06, + "loss": 0.2306, + "step": 70596 + }, + { + "epoch": 2.71, + "learning_rate": 4.765965632213388e-06, + "loss": 0.3431, + "step": 70597 + }, + { + "epoch": 2.71, + "learning_rate": 4.7653244421646576e-06, + "loss": 0.1063, + "step": 70598 + }, + { + "epoch": 2.71, + "learning_rate": 4.764683252115927e-06, + "loss": 0.223, + "step": 70599 + }, + { + "epoch": 2.71, + "learning_rate": 4.764042062067197e-06, + "loss": 0.1866, + "step": 70600 + }, + { + "epoch": 2.71, + "learning_rate": 4.763400872018466e-06, + "loss": 0.2539, + "step": 70601 + }, + { + "epoch": 2.71, + "learning_rate": 4.762759681969736e-06, + "loss": 0.197, + "step": 70602 + }, + { + "epoch": 2.71, + "learning_rate": 4.762118491921005e-06, + "loss": 0.1188, + "step": 70603 + }, + { + "epoch": 2.71, + "learning_rate": 4.761477301872275e-06, + "loss": 0.287, + "step": 70604 + }, + { + "epoch": 2.71, + "learning_rate": 4.760836111823544e-06, + "loss": 0.1745, + "step": 70605 + }, + { + "epoch": 2.71, + "learning_rate": 4.760194921774814e-06, + "loss": 0.1978, + "step": 70606 + }, + { + "epoch": 2.71, + "learning_rate": 4.7595537317260835e-06, + "loss": 0.1049, + "step": 70607 + }, + { + "epoch": 2.71, + "learning_rate": 4.758912541677353e-06, + "loss": 0.2105, + "step": 70608 + }, + { + "epoch": 2.71, + "learning_rate": 4.758271351628623e-06, + "loss": 0.2833, + "step": 70609 + }, + { + "epoch": 2.71, + "learning_rate": 4.757630161579892e-06, + "loss": 0.3281, + "step": 70610 + }, + { + "epoch": 2.71, + "learning_rate": 4.756988971531162e-06, + "loss": 0.1664, + "step": 70611 + }, + { + "epoch": 2.71, + "learning_rate": 4.756347781482431e-06, + "loss": 0.167, + "step": 70612 + }, + { + "epoch": 2.71, + "learning_rate": 4.755706591433701e-06, + "loss": 0.1414, + "step": 70613 + }, + { + "epoch": 2.71, + "learning_rate": 4.75506540138497e-06, + "loss": 0.1433, + "step": 70614 + }, + { + "epoch": 2.71, + "learning_rate": 4.75442421133624e-06, + "loss": 0.2135, + "step": 70615 + }, + { + "epoch": 2.71, + "learning_rate": 4.7537830212875095e-06, + "loss": 0.2498, + "step": 70616 + }, + { + "epoch": 2.71, + "learning_rate": 4.7531418312387795e-06, + "loss": 0.216, + "step": 70617 + }, + { + "epoch": 2.72, + "learning_rate": 4.752500641190049e-06, + "loss": 0.1524, + "step": 70618 + }, + { + "epoch": 2.72, + "learning_rate": 4.751859451141318e-06, + "loss": 0.0827, + "step": 70619 + }, + { + "epoch": 2.72, + "learning_rate": 4.751218261092588e-06, + "loss": 0.1577, + "step": 70620 + }, + { + "epoch": 2.72, + "learning_rate": 4.750577071043857e-06, + "loss": 0.2369, + "step": 70621 + }, + { + "epoch": 2.72, + "learning_rate": 4.749935880995127e-06, + "loss": 0.3082, + "step": 70622 + }, + { + "epoch": 2.72, + "learning_rate": 4.749294690946396e-06, + "loss": 0.1628, + "step": 70623 + }, + { + "epoch": 2.72, + "learning_rate": 4.748653500897666e-06, + "loss": 0.1855, + "step": 70624 + }, + { + "epoch": 2.72, + "learning_rate": 4.7480123108489355e-06, + "loss": 0.219, + "step": 70625 + }, + { + "epoch": 2.72, + "learning_rate": 4.7473711208002055e-06, + "loss": 0.2181, + "step": 70626 + }, + { + "epoch": 2.72, + "learning_rate": 4.746729930751475e-06, + "loss": 0.1888, + "step": 70627 + }, + { + "epoch": 2.72, + "learning_rate": 4.746088740702744e-06, + "loss": 0.2333, + "step": 70628 + }, + { + "epoch": 2.72, + "learning_rate": 4.745447550654014e-06, + "loss": 0.1462, + "step": 70629 + }, + { + "epoch": 2.72, + "learning_rate": 4.744806360605283e-06, + "loss": 0.1525, + "step": 70630 + }, + { + "epoch": 2.72, + "learning_rate": 4.744165170556553e-06, + "loss": 0.1942, + "step": 70631 + }, + { + "epoch": 2.72, + "learning_rate": 4.743523980507822e-06, + "loss": 0.1022, + "step": 70632 + }, + { + "epoch": 2.72, + "learning_rate": 4.742882790459092e-06, + "loss": 0.2138, + "step": 70633 + }, + { + "epoch": 2.72, + "learning_rate": 4.7422416004103614e-06, + "loss": 0.2448, + "step": 70634 + }, + { + "epoch": 2.72, + "learning_rate": 4.7416004103616315e-06, + "loss": 0.1497, + "step": 70635 + }, + { + "epoch": 2.72, + "learning_rate": 4.740959220312901e-06, + "loss": 0.1087, + "step": 70636 + }, + { + "epoch": 2.72, + "learning_rate": 4.740318030264171e-06, + "loss": 0.3063, + "step": 70637 + }, + { + "epoch": 2.72, + "learning_rate": 4.73967684021544e-06, + "loss": 0.2051, + "step": 70638 + }, + { + "epoch": 2.72, + "learning_rate": 4.739035650166709e-06, + "loss": 0.1708, + "step": 70639 + }, + { + "epoch": 2.72, + "learning_rate": 4.738394460117979e-06, + "loss": 0.2106, + "step": 70640 + }, + { + "epoch": 2.72, + "learning_rate": 4.737753270069248e-06, + "loss": 0.1761, + "step": 70641 + }, + { + "epoch": 2.72, + "learning_rate": 4.737112080020518e-06, + "loss": 0.1924, + "step": 70642 + }, + { + "epoch": 2.72, + "learning_rate": 4.736470889971787e-06, + "loss": 0.3819, + "step": 70643 + }, + { + "epoch": 2.72, + "learning_rate": 4.7358296999230574e-06, + "loss": 0.2469, + "step": 70644 + }, + { + "epoch": 2.72, + "learning_rate": 4.735188509874327e-06, + "loss": 0.1459, + "step": 70645 + }, + { + "epoch": 2.72, + "learning_rate": 4.734547319825597e-06, + "loss": 0.137, + "step": 70646 + }, + { + "epoch": 2.72, + "learning_rate": 4.733906129776866e-06, + "loss": 0.3082, + "step": 70647 + }, + { + "epoch": 2.72, + "learning_rate": 4.733264939728135e-06, + "loss": 0.334, + "step": 70648 + }, + { + "epoch": 2.72, + "learning_rate": 4.732623749679405e-06, + "loss": 0.1836, + "step": 70649 + }, + { + "epoch": 2.72, + "learning_rate": 4.731982559630674e-06, + "loss": 0.1946, + "step": 70650 + }, + { + "epoch": 2.72, + "learning_rate": 4.731341369581944e-06, + "loss": 0.194, + "step": 70651 + }, + { + "epoch": 2.72, + "learning_rate": 4.730700179533213e-06, + "loss": 0.2065, + "step": 70652 + }, + { + "epoch": 2.72, + "learning_rate": 4.730058989484483e-06, + "loss": 0.1826, + "step": 70653 + }, + { + "epoch": 2.72, + "learning_rate": 4.7294177994357526e-06, + "loss": 0.2159, + "step": 70654 + }, + { + "epoch": 2.72, + "learning_rate": 4.728776609387023e-06, + "loss": 0.1627, + "step": 70655 + }, + { + "epoch": 2.72, + "learning_rate": 4.728135419338292e-06, + "loss": 0.2554, + "step": 70656 + }, + { + "epoch": 2.72, + "learning_rate": 4.727494229289561e-06, + "loss": 0.168, + "step": 70657 + }, + { + "epoch": 2.72, + "learning_rate": 4.726853039240831e-06, + "loss": 0.1762, + "step": 70658 + }, + { + "epoch": 2.72, + "learning_rate": 4.7262118491921e-06, + "loss": 0.2634, + "step": 70659 + }, + { + "epoch": 2.72, + "learning_rate": 4.72557065914337e-06, + "loss": 0.1427, + "step": 70660 + }, + { + "epoch": 2.72, + "learning_rate": 4.724929469094639e-06, + "loss": 0.3007, + "step": 70661 + }, + { + "epoch": 2.72, + "learning_rate": 4.724288279045909e-06, + "loss": 0.0864, + "step": 70662 + }, + { + "epoch": 2.72, + "learning_rate": 4.7236470889971785e-06, + "loss": 0.3541, + "step": 70663 + }, + { + "epoch": 2.72, + "learning_rate": 4.7230058989484486e-06, + "loss": 0.2838, + "step": 70664 + }, + { + "epoch": 2.72, + "learning_rate": 4.722364708899718e-06, + "loss": 0.2911, + "step": 70665 + }, + { + "epoch": 2.72, + "learning_rate": 4.721723518850988e-06, + "loss": 0.2913, + "step": 70666 + }, + { + "epoch": 2.72, + "learning_rate": 4.721082328802257e-06, + "loss": 0.3064, + "step": 70667 + }, + { + "epoch": 2.72, + "learning_rate": 4.720441138753526e-06, + "loss": 0.3133, + "step": 70668 + }, + { + "epoch": 2.72, + "learning_rate": 4.719799948704796e-06, + "loss": 0.1952, + "step": 70669 + }, + { + "epoch": 2.72, + "learning_rate": 4.719158758656065e-06, + "loss": 0.2893, + "step": 70670 + }, + { + "epoch": 2.72, + "learning_rate": 4.718517568607335e-06, + "loss": 0.2253, + "step": 70671 + }, + { + "epoch": 2.72, + "learning_rate": 4.7178763785586045e-06, + "loss": 0.2192, + "step": 70672 + }, + { + "epoch": 2.72, + "learning_rate": 4.7172351885098745e-06, + "loss": 0.3701, + "step": 70673 + }, + { + "epoch": 2.72, + "learning_rate": 4.716593998461144e-06, + "loss": 0.2757, + "step": 70674 + }, + { + "epoch": 2.72, + "learning_rate": 4.715952808412414e-06, + "loss": 0.184, + "step": 70675 + }, + { + "epoch": 2.72, + "learning_rate": 4.715311618363683e-06, + "loss": 0.1659, + "step": 70676 + }, + { + "epoch": 2.72, + "learning_rate": 4.714670428314952e-06, + "loss": 0.1708, + "step": 70677 + }, + { + "epoch": 2.72, + "learning_rate": 4.714029238266222e-06, + "loss": 0.1737, + "step": 70678 + }, + { + "epoch": 2.72, + "learning_rate": 4.713388048217491e-06, + "loss": 0.2502, + "step": 70679 + }, + { + "epoch": 2.72, + "learning_rate": 4.712746858168761e-06, + "loss": 0.1564, + "step": 70680 + }, + { + "epoch": 2.72, + "learning_rate": 4.7121056681200305e-06, + "loss": 0.23, + "step": 70681 + }, + { + "epoch": 2.72, + "learning_rate": 4.7114644780713005e-06, + "loss": 0.307, + "step": 70682 + }, + { + "epoch": 2.72, + "learning_rate": 4.71082328802257e-06, + "loss": 0.2667, + "step": 70683 + }, + { + "epoch": 2.72, + "learning_rate": 4.71018209797384e-06, + "loss": 0.2359, + "step": 70684 + }, + { + "epoch": 2.72, + "learning_rate": 4.709540907925109e-06, + "loss": 0.12, + "step": 70685 + }, + { + "epoch": 2.72, + "learning_rate": 4.708899717876379e-06, + "loss": 0.1536, + "step": 70686 + }, + { + "epoch": 2.72, + "learning_rate": 4.708258527827648e-06, + "loss": 0.23, + "step": 70687 + }, + { + "epoch": 2.72, + "learning_rate": 4.707617337778917e-06, + "loss": 0.2355, + "step": 70688 + }, + { + "epoch": 2.72, + "learning_rate": 4.706976147730187e-06, + "loss": 0.2422, + "step": 70689 + }, + { + "epoch": 2.72, + "learning_rate": 4.7063349576814564e-06, + "loss": 0.1633, + "step": 70690 + }, + { + "epoch": 2.72, + "learning_rate": 4.7056937676327265e-06, + "loss": 0.3007, + "step": 70691 + }, + { + "epoch": 2.72, + "learning_rate": 4.705052577583996e-06, + "loss": 0.352, + "step": 70692 + }, + { + "epoch": 2.72, + "learning_rate": 4.704411387535266e-06, + "loss": 0.1163, + "step": 70693 + }, + { + "epoch": 2.72, + "learning_rate": 4.703770197486535e-06, + "loss": 0.1979, + "step": 70694 + }, + { + "epoch": 2.72, + "learning_rate": 4.703129007437805e-06, + "loss": 0.3138, + "step": 70695 + }, + { + "epoch": 2.72, + "learning_rate": 4.702487817389074e-06, + "loss": 0.2025, + "step": 70696 + }, + { + "epoch": 2.72, + "learning_rate": 4.701846627340343e-06, + "loss": 0.2358, + "step": 70697 + }, + { + "epoch": 2.72, + "learning_rate": 4.701205437291613e-06, + "loss": 0.1872, + "step": 70698 + }, + { + "epoch": 2.72, + "learning_rate": 4.700564247242882e-06, + "loss": 0.2904, + "step": 70699 + }, + { + "epoch": 2.72, + "learning_rate": 4.6999230571941524e-06, + "loss": 0.1892, + "step": 70700 + }, + { + "epoch": 2.72, + "learning_rate": 4.699281867145422e-06, + "loss": 0.1353, + "step": 70701 + }, + { + "epoch": 2.72, + "learning_rate": 4.698640677096692e-06, + "loss": 0.2066, + "step": 70702 + }, + { + "epoch": 2.72, + "learning_rate": 4.697999487047961e-06, + "loss": 0.1553, + "step": 70703 + }, + { + "epoch": 2.72, + "learning_rate": 4.697358296999231e-06, + "loss": 0.1367, + "step": 70704 + }, + { + "epoch": 2.72, + "learning_rate": 4.6967171069505e-06, + "loss": 0.2898, + "step": 70705 + }, + { + "epoch": 2.72, + "learning_rate": 4.696075916901769e-06, + "loss": 0.1631, + "step": 70706 + }, + { + "epoch": 2.72, + "learning_rate": 4.695434726853039e-06, + "loss": 0.1192, + "step": 70707 + }, + { + "epoch": 2.72, + "learning_rate": 4.694793536804308e-06, + "loss": 0.1977, + "step": 70708 + }, + { + "epoch": 2.72, + "learning_rate": 4.694152346755578e-06, + "loss": 0.2064, + "step": 70709 + }, + { + "epoch": 2.72, + "learning_rate": 4.6935111567068476e-06, + "loss": 0.1583, + "step": 70710 + }, + { + "epoch": 2.72, + "learning_rate": 4.692869966658118e-06, + "loss": 0.2718, + "step": 70711 + }, + { + "epoch": 2.72, + "learning_rate": 4.692228776609387e-06, + "loss": 0.3333, + "step": 70712 + }, + { + "epoch": 2.72, + "learning_rate": 4.691587586560657e-06, + "loss": 0.1312, + "step": 70713 + }, + { + "epoch": 2.72, + "learning_rate": 4.690946396511926e-06, + "loss": 0.2712, + "step": 70714 + }, + { + "epoch": 2.72, + "learning_rate": 4.690305206463196e-06, + "loss": 0.2128, + "step": 70715 + }, + { + "epoch": 2.72, + "learning_rate": 4.689664016414465e-06, + "loss": 0.1728, + "step": 70716 + }, + { + "epoch": 2.72, + "learning_rate": 4.689022826365734e-06, + "loss": 0.3381, + "step": 70717 + }, + { + "epoch": 2.72, + "learning_rate": 4.688381636317004e-06, + "loss": 0.1727, + "step": 70718 + }, + { + "epoch": 2.72, + "learning_rate": 4.6877404462682735e-06, + "loss": 0.1577, + "step": 70719 + }, + { + "epoch": 2.72, + "learning_rate": 4.6870992562195436e-06, + "loss": 0.2132, + "step": 70720 + }, + { + "epoch": 2.72, + "learning_rate": 4.686458066170813e-06, + "loss": 0.2449, + "step": 70721 + }, + { + "epoch": 2.72, + "learning_rate": 4.685816876122083e-06, + "loss": 0.2184, + "step": 70722 + }, + { + "epoch": 2.72, + "learning_rate": 4.685175686073352e-06, + "loss": 0.3564, + "step": 70723 + }, + { + "epoch": 2.72, + "learning_rate": 4.684534496024622e-06, + "loss": 0.2858, + "step": 70724 + }, + { + "epoch": 2.72, + "learning_rate": 4.683893305975891e-06, + "loss": 0.1994, + "step": 70725 + }, + { + "epoch": 2.72, + "learning_rate": 4.68325211592716e-06, + "loss": 0.452, + "step": 70726 + }, + { + "epoch": 2.72, + "learning_rate": 4.68261092587843e-06, + "loss": 0.137, + "step": 70727 + }, + { + "epoch": 2.72, + "learning_rate": 4.6819697358296995e-06, + "loss": 0.1709, + "step": 70728 + }, + { + "epoch": 2.72, + "learning_rate": 4.6813285457809695e-06, + "loss": 0.1989, + "step": 70729 + }, + { + "epoch": 2.72, + "learning_rate": 4.680687355732239e-06, + "loss": 0.1687, + "step": 70730 + }, + { + "epoch": 2.72, + "learning_rate": 4.680046165683509e-06, + "loss": 0.2304, + "step": 70731 + }, + { + "epoch": 2.72, + "learning_rate": 4.679404975634778e-06, + "loss": 0.1852, + "step": 70732 + }, + { + "epoch": 2.72, + "learning_rate": 4.678763785586048e-06, + "loss": 0.2183, + "step": 70733 + }, + { + "epoch": 2.72, + "learning_rate": 4.678122595537317e-06, + "loss": 0.238, + "step": 70734 + }, + { + "epoch": 2.72, + "learning_rate": 4.677481405488587e-06, + "loss": 0.1481, + "step": 70735 + }, + { + "epoch": 2.72, + "learning_rate": 4.676840215439856e-06, + "loss": 0.1574, + "step": 70736 + }, + { + "epoch": 2.72, + "learning_rate": 4.6761990253911255e-06, + "loss": 0.2523, + "step": 70737 + }, + { + "epoch": 2.72, + "learning_rate": 4.6755578353423955e-06, + "loss": 0.2752, + "step": 70738 + }, + { + "epoch": 2.72, + "learning_rate": 4.674916645293665e-06, + "loss": 0.2828, + "step": 70739 + }, + { + "epoch": 2.72, + "learning_rate": 4.674275455244935e-06, + "loss": 0.2367, + "step": 70740 + }, + { + "epoch": 2.72, + "learning_rate": 4.673634265196204e-06, + "loss": 0.1974, + "step": 70741 + }, + { + "epoch": 2.72, + "learning_rate": 4.672993075147474e-06, + "loss": 0.1601, + "step": 70742 + }, + { + "epoch": 2.72, + "learning_rate": 4.672351885098743e-06, + "loss": 0.2148, + "step": 70743 + }, + { + "epoch": 2.72, + "learning_rate": 4.671710695050013e-06, + "loss": 0.2484, + "step": 70744 + }, + { + "epoch": 2.72, + "learning_rate": 4.671069505001282e-06, + "loss": 0.325, + "step": 70745 + }, + { + "epoch": 2.72, + "learning_rate": 4.6704283149525514e-06, + "loss": 0.2621, + "step": 70746 + }, + { + "epoch": 2.72, + "learning_rate": 4.6697871249038215e-06, + "loss": 0.3621, + "step": 70747 + }, + { + "epoch": 2.72, + "learning_rate": 4.669145934855091e-06, + "loss": 0.2791, + "step": 70748 + }, + { + "epoch": 2.72, + "learning_rate": 4.668504744806361e-06, + "loss": 0.1291, + "step": 70749 + }, + { + "epoch": 2.72, + "learning_rate": 4.66786355475763e-06, + "loss": 0.157, + "step": 70750 + }, + { + "epoch": 2.72, + "learning_rate": 4.6672223647089e-06, + "loss": 0.2708, + "step": 70751 + }, + { + "epoch": 2.72, + "learning_rate": 4.666581174660169e-06, + "loss": 0.1714, + "step": 70752 + }, + { + "epoch": 2.72, + "learning_rate": 4.665939984611439e-06, + "loss": 0.2182, + "step": 70753 + }, + { + "epoch": 2.72, + "learning_rate": 4.665298794562708e-06, + "loss": 0.1775, + "step": 70754 + }, + { + "epoch": 2.72, + "learning_rate": 4.664657604513978e-06, + "loss": 0.2493, + "step": 70755 + }, + { + "epoch": 2.72, + "learning_rate": 4.6640164144652474e-06, + "loss": 0.2121, + "step": 70756 + }, + { + "epoch": 2.72, + "learning_rate": 4.663375224416517e-06, + "loss": 0.1734, + "step": 70757 + }, + { + "epoch": 2.72, + "learning_rate": 4.662734034367787e-06, + "loss": 0.1561, + "step": 70758 + }, + { + "epoch": 2.72, + "learning_rate": 4.662092844319056e-06, + "loss": 0.1648, + "step": 70759 + }, + { + "epoch": 2.72, + "learning_rate": 4.661451654270326e-06, + "loss": 0.1585, + "step": 70760 + }, + { + "epoch": 2.72, + "learning_rate": 4.660810464221595e-06, + "loss": 0.2259, + "step": 70761 + }, + { + "epoch": 2.72, + "learning_rate": 4.660169274172865e-06, + "loss": 0.1713, + "step": 70762 + }, + { + "epoch": 2.72, + "learning_rate": 4.659528084124134e-06, + "loss": 0.1025, + "step": 70763 + }, + { + "epoch": 2.72, + "learning_rate": 4.658886894075404e-06, + "loss": 0.1228, + "step": 70764 + }, + { + "epoch": 2.72, + "learning_rate": 4.658245704026673e-06, + "loss": 0.1409, + "step": 70765 + }, + { + "epoch": 2.72, + "learning_rate": 4.657604513977943e-06, + "loss": 0.2054, + "step": 70766 + }, + { + "epoch": 2.72, + "learning_rate": 4.656963323929213e-06, + "loss": 0.2163, + "step": 70767 + }, + { + "epoch": 2.72, + "learning_rate": 4.656322133880482e-06, + "loss": 0.27, + "step": 70768 + }, + { + "epoch": 2.72, + "learning_rate": 4.655680943831752e-06, + "loss": 0.1345, + "step": 70769 + }, + { + "epoch": 2.72, + "learning_rate": 4.655039753783021e-06, + "loss": 0.1996, + "step": 70770 + }, + { + "epoch": 2.72, + "learning_rate": 4.654398563734291e-06, + "loss": 0.1357, + "step": 70771 + }, + { + "epoch": 2.72, + "learning_rate": 4.65375737368556e-06, + "loss": 0.2035, + "step": 70772 + }, + { + "epoch": 2.72, + "learning_rate": 4.65311618363683e-06, + "loss": 0.0972, + "step": 70773 + }, + { + "epoch": 2.72, + "learning_rate": 4.652474993588099e-06, + "loss": 0.2354, + "step": 70774 + }, + { + "epoch": 2.72, + "learning_rate": 4.6518338035393685e-06, + "loss": 0.2368, + "step": 70775 + }, + { + "epoch": 2.72, + "learning_rate": 4.6511926134906386e-06, + "loss": 0.2604, + "step": 70776 + }, + { + "epoch": 2.72, + "learning_rate": 4.650551423441908e-06, + "loss": 0.1558, + "step": 70777 + }, + { + "epoch": 2.72, + "learning_rate": 4.649910233393178e-06, + "loss": 0.1261, + "step": 70778 + }, + { + "epoch": 2.72, + "learning_rate": 4.649269043344447e-06, + "loss": 0.1588, + "step": 70779 + }, + { + "epoch": 2.72, + "learning_rate": 4.648627853295717e-06, + "loss": 0.1056, + "step": 70780 + }, + { + "epoch": 2.72, + "learning_rate": 4.647986663246986e-06, + "loss": 0.2869, + "step": 70781 + }, + { + "epoch": 2.72, + "learning_rate": 4.647345473198256e-06, + "loss": 0.2132, + "step": 70782 + }, + { + "epoch": 2.72, + "learning_rate": 4.646704283149525e-06, + "loss": 0.1921, + "step": 70783 + }, + { + "epoch": 2.72, + "learning_rate": 4.646063093100795e-06, + "loss": 0.2997, + "step": 70784 + }, + { + "epoch": 2.72, + "learning_rate": 4.6454219030520645e-06, + "loss": 0.2414, + "step": 70785 + }, + { + "epoch": 2.72, + "learning_rate": 4.644780713003334e-06, + "loss": 0.2668, + "step": 70786 + }, + { + "epoch": 2.72, + "learning_rate": 4.644139522954604e-06, + "loss": 0.1111, + "step": 70787 + }, + { + "epoch": 2.72, + "learning_rate": 4.643498332905873e-06, + "loss": 0.2786, + "step": 70788 + }, + { + "epoch": 2.72, + "learning_rate": 4.642857142857143e-06, + "loss": 0.3102, + "step": 70789 + }, + { + "epoch": 2.72, + "learning_rate": 4.642215952808412e-06, + "loss": 0.2763, + "step": 70790 + }, + { + "epoch": 2.72, + "learning_rate": 4.641574762759682e-06, + "loss": 0.118, + "step": 70791 + }, + { + "epoch": 2.72, + "learning_rate": 4.640933572710951e-06, + "loss": 0.2414, + "step": 70792 + }, + { + "epoch": 2.72, + "learning_rate": 4.640292382662221e-06, + "loss": 0.2859, + "step": 70793 + }, + { + "epoch": 2.72, + "learning_rate": 4.6396511926134905e-06, + "loss": 0.208, + "step": 70794 + }, + { + "epoch": 2.72, + "learning_rate": 4.63901000256476e-06, + "loss": 0.1913, + "step": 70795 + }, + { + "epoch": 2.72, + "learning_rate": 4.63836881251603e-06, + "loss": 0.1825, + "step": 70796 + }, + { + "epoch": 2.72, + "learning_rate": 4.637727622467299e-06, + "loss": 0.1592, + "step": 70797 + }, + { + "epoch": 2.72, + "learning_rate": 4.637086432418569e-06, + "loss": 0.1964, + "step": 70798 + }, + { + "epoch": 2.72, + "learning_rate": 4.636445242369838e-06, + "loss": 0.1897, + "step": 70799 + }, + { + "epoch": 2.72, + "learning_rate": 4.635804052321108e-06, + "loss": 0.2184, + "step": 70800 + }, + { + "epoch": 2.72, + "learning_rate": 4.635162862272377e-06, + "loss": 0.25, + "step": 70801 + }, + { + "epoch": 2.72, + "learning_rate": 4.634521672223647e-06, + "loss": 0.2406, + "step": 70802 + }, + { + "epoch": 2.72, + "learning_rate": 4.6338804821749165e-06, + "loss": 0.2102, + "step": 70803 + }, + { + "epoch": 2.72, + "learning_rate": 4.6332392921261865e-06, + "loss": 0.2108, + "step": 70804 + }, + { + "epoch": 2.72, + "learning_rate": 4.632598102077456e-06, + "loss": 0.1436, + "step": 70805 + }, + { + "epoch": 2.72, + "learning_rate": 4.631956912028725e-06, + "loss": 0.2643, + "step": 70806 + }, + { + "epoch": 2.72, + "learning_rate": 4.631315721979995e-06, + "loss": 0.2253, + "step": 70807 + }, + { + "epoch": 2.72, + "learning_rate": 4.630674531931264e-06, + "loss": 0.25, + "step": 70808 + }, + { + "epoch": 2.72, + "learning_rate": 4.630033341882534e-06, + "loss": 0.2644, + "step": 70809 + }, + { + "epoch": 2.72, + "learning_rate": 4.629392151833803e-06, + "loss": 0.2986, + "step": 70810 + }, + { + "epoch": 2.72, + "learning_rate": 4.628750961785073e-06, + "loss": 0.2041, + "step": 70811 + }, + { + "epoch": 2.72, + "learning_rate": 4.6281097717363424e-06, + "loss": 0.2703, + "step": 70812 + }, + { + "epoch": 2.72, + "learning_rate": 4.6274685816876125e-06, + "loss": 0.1986, + "step": 70813 + }, + { + "epoch": 2.72, + "learning_rate": 4.626827391638882e-06, + "loss": 0.2327, + "step": 70814 + }, + { + "epoch": 2.72, + "learning_rate": 4.626186201590151e-06, + "loss": 0.2772, + "step": 70815 + }, + { + "epoch": 2.72, + "learning_rate": 4.625545011541421e-06, + "loss": 0.1637, + "step": 70816 + }, + { + "epoch": 2.72, + "learning_rate": 4.62490382149269e-06, + "loss": 0.2471, + "step": 70817 + }, + { + "epoch": 2.72, + "learning_rate": 4.62426263144396e-06, + "loss": 0.1873, + "step": 70818 + }, + { + "epoch": 2.72, + "learning_rate": 4.623621441395229e-06, + "loss": 0.2481, + "step": 70819 + }, + { + "epoch": 2.72, + "learning_rate": 4.622980251346499e-06, + "loss": 0.1656, + "step": 70820 + }, + { + "epoch": 2.72, + "learning_rate": 4.622339061297768e-06, + "loss": 0.1546, + "step": 70821 + }, + { + "epoch": 2.72, + "learning_rate": 4.6216978712490384e-06, + "loss": 0.1304, + "step": 70822 + }, + { + "epoch": 2.72, + "learning_rate": 4.621056681200308e-06, + "loss": 0.3468, + "step": 70823 + }, + { + "epoch": 2.72, + "learning_rate": 4.620415491151578e-06, + "loss": 0.2661, + "step": 70824 + }, + { + "epoch": 2.72, + "learning_rate": 4.619774301102847e-06, + "loss": 0.2234, + "step": 70825 + }, + { + "epoch": 2.72, + "learning_rate": 4.619133111054116e-06, + "loss": 0.2262, + "step": 70826 + }, + { + "epoch": 2.72, + "learning_rate": 4.618491921005386e-06, + "loss": 0.1282, + "step": 70827 + }, + { + "epoch": 2.72, + "learning_rate": 4.617850730956655e-06, + "loss": 0.2211, + "step": 70828 + }, + { + "epoch": 2.72, + "learning_rate": 4.617209540907925e-06, + "loss": 0.1882, + "step": 70829 + }, + { + "epoch": 2.72, + "learning_rate": 4.616568350859194e-06, + "loss": 0.1636, + "step": 70830 + }, + { + "epoch": 2.72, + "learning_rate": 4.615927160810464e-06, + "loss": 0.1346, + "step": 70831 + }, + { + "epoch": 2.72, + "learning_rate": 4.6152859707617336e-06, + "loss": 0.2922, + "step": 70832 + }, + { + "epoch": 2.72, + "learning_rate": 4.614644780713004e-06, + "loss": 0.2443, + "step": 70833 + }, + { + "epoch": 2.72, + "learning_rate": 4.614003590664273e-06, + "loss": 0.3324, + "step": 70834 + }, + { + "epoch": 2.72, + "learning_rate": 4.613362400615542e-06, + "loss": 0.3832, + "step": 70835 + }, + { + "epoch": 2.72, + "learning_rate": 4.612721210566812e-06, + "loss": 0.2192, + "step": 70836 + }, + { + "epoch": 2.72, + "learning_rate": 4.612080020518081e-06, + "loss": 0.2314, + "step": 70837 + }, + { + "epoch": 2.72, + "learning_rate": 4.611438830469351e-06, + "loss": 0.1174, + "step": 70838 + }, + { + "epoch": 2.72, + "learning_rate": 4.61079764042062e-06, + "loss": 0.1598, + "step": 70839 + }, + { + "epoch": 2.72, + "learning_rate": 4.61015645037189e-06, + "loss": 0.1276, + "step": 70840 + }, + { + "epoch": 2.72, + "learning_rate": 4.6095152603231595e-06, + "loss": 0.1131, + "step": 70841 + }, + { + "epoch": 2.72, + "learning_rate": 4.6088740702744296e-06, + "loss": 0.2187, + "step": 70842 + }, + { + "epoch": 2.72, + "learning_rate": 4.608232880225699e-06, + "loss": 0.1633, + "step": 70843 + }, + { + "epoch": 2.72, + "learning_rate": 4.607591690176969e-06, + "loss": 0.2169, + "step": 70844 + }, + { + "epoch": 2.72, + "learning_rate": 4.606950500128238e-06, + "loss": 0.164, + "step": 70845 + }, + { + "epoch": 2.72, + "learning_rate": 4.606309310079507e-06, + "loss": 0.1876, + "step": 70846 + }, + { + "epoch": 2.72, + "learning_rate": 4.605668120030777e-06, + "loss": 0.1471, + "step": 70847 + }, + { + "epoch": 2.72, + "learning_rate": 4.605026929982046e-06, + "loss": 0.2517, + "step": 70848 + }, + { + "epoch": 2.72, + "learning_rate": 4.604385739933316e-06, + "loss": 0.2197, + "step": 70849 + }, + { + "epoch": 2.72, + "learning_rate": 4.6037445498845855e-06, + "loss": 0.1181, + "step": 70850 + }, + { + "epoch": 2.72, + "learning_rate": 4.6031033598358555e-06, + "loss": 0.1134, + "step": 70851 + }, + { + "epoch": 2.72, + "learning_rate": 4.602462169787125e-06, + "loss": 0.2344, + "step": 70852 + }, + { + "epoch": 2.72, + "learning_rate": 4.601820979738395e-06, + "loss": 0.2378, + "step": 70853 + }, + { + "epoch": 2.72, + "learning_rate": 4.601179789689664e-06, + "loss": 0.236, + "step": 70854 + }, + { + "epoch": 2.72, + "learning_rate": 4.600538599640933e-06, + "loss": 0.198, + "step": 70855 + }, + { + "epoch": 2.72, + "learning_rate": 4.599897409592203e-06, + "loss": 0.2233, + "step": 70856 + }, + { + "epoch": 2.72, + "learning_rate": 4.599256219543472e-06, + "loss": 0.1283, + "step": 70857 + }, + { + "epoch": 2.72, + "learning_rate": 4.598615029494742e-06, + "loss": 0.1494, + "step": 70858 + }, + { + "epoch": 2.72, + "learning_rate": 4.5979738394460115e-06, + "loss": 0.2206, + "step": 70859 + }, + { + "epoch": 2.72, + "learning_rate": 4.5973326493972815e-06, + "loss": 0.1449, + "step": 70860 + }, + { + "epoch": 2.72, + "learning_rate": 4.596691459348551e-06, + "loss": 0.1892, + "step": 70861 + }, + { + "epoch": 2.72, + "learning_rate": 4.596050269299821e-06, + "loss": 0.1885, + "step": 70862 + }, + { + "epoch": 2.72, + "learning_rate": 4.59540907925109e-06, + "loss": 0.2073, + "step": 70863 + }, + { + "epoch": 2.72, + "learning_rate": 4.59476788920236e-06, + "loss": 0.2296, + "step": 70864 + }, + { + "epoch": 2.72, + "learning_rate": 4.594126699153629e-06, + "loss": 0.193, + "step": 70865 + }, + { + "epoch": 2.72, + "learning_rate": 4.593485509104898e-06, + "loss": 0.1356, + "step": 70866 + }, + { + "epoch": 2.72, + "learning_rate": 4.592844319056168e-06, + "loss": 0.209, + "step": 70867 + }, + { + "epoch": 2.72, + "learning_rate": 4.5922031290074374e-06, + "loss": 0.3122, + "step": 70868 + }, + { + "epoch": 2.72, + "learning_rate": 4.5915619389587075e-06, + "loss": 0.2088, + "step": 70869 + }, + { + "epoch": 2.72, + "learning_rate": 4.590920748909977e-06, + "loss": 0.1978, + "step": 70870 + }, + { + "epoch": 2.72, + "learning_rate": 4.590279558861247e-06, + "loss": 0.1465, + "step": 70871 + }, + { + "epoch": 2.72, + "learning_rate": 4.589638368812516e-06, + "loss": 0.1933, + "step": 70872 + }, + { + "epoch": 2.72, + "learning_rate": 4.588997178763786e-06, + "loss": 0.2225, + "step": 70873 + }, + { + "epoch": 2.72, + "learning_rate": 4.588355988715055e-06, + "loss": 0.0887, + "step": 70874 + }, + { + "epoch": 2.72, + "learning_rate": 4.587714798666325e-06, + "loss": 0.2087, + "step": 70875 + }, + { + "epoch": 2.72, + "learning_rate": 4.587073608617594e-06, + "loss": 0.1483, + "step": 70876 + }, + { + "epoch": 2.72, + "learning_rate": 4.586432418568863e-06, + "loss": 0.0997, + "step": 70877 + }, + { + "epoch": 2.73, + "learning_rate": 4.5857912285201334e-06, + "loss": 0.1824, + "step": 70878 + }, + { + "epoch": 2.73, + "learning_rate": 4.585150038471403e-06, + "loss": 0.144, + "step": 70879 + }, + { + "epoch": 2.73, + "learning_rate": 4.584508848422673e-06, + "loss": 0.1, + "step": 70880 + }, + { + "epoch": 2.73, + "learning_rate": 4.583867658373942e-06, + "loss": 0.2254, + "step": 70881 + }, + { + "epoch": 2.73, + "learning_rate": 4.583226468325212e-06, + "loss": 0.1899, + "step": 70882 + }, + { + "epoch": 2.73, + "learning_rate": 4.582585278276481e-06, + "loss": 0.2725, + "step": 70883 + }, + { + "epoch": 2.73, + "learning_rate": 4.581944088227751e-06, + "loss": 0.1948, + "step": 70884 + }, + { + "epoch": 2.73, + "learning_rate": 4.58130289817902e-06, + "loss": 0.2434, + "step": 70885 + }, + { + "epoch": 2.73, + "learning_rate": 4.580661708130289e-06, + "loss": 0.268, + "step": 70886 + }, + { + "epoch": 2.73, + "learning_rate": 4.580020518081559e-06, + "loss": 0.1263, + "step": 70887 + }, + { + "epoch": 2.73, + "learning_rate": 4.579379328032829e-06, + "loss": 0.1551, + "step": 70888 + }, + { + "epoch": 2.73, + "learning_rate": 4.578738137984099e-06, + "loss": 0.1649, + "step": 70889 + }, + { + "epoch": 2.73, + "learning_rate": 4.578096947935368e-06, + "loss": 0.224, + "step": 70890 + }, + { + "epoch": 2.73, + "learning_rate": 4.577455757886638e-06, + "loss": 0.1877, + "step": 70891 + }, + { + "epoch": 2.73, + "learning_rate": 4.576814567837907e-06, + "loss": 0.2222, + "step": 70892 + }, + { + "epoch": 2.73, + "learning_rate": 4.576173377789177e-06, + "loss": 0.2735, + "step": 70893 + }, + { + "epoch": 2.73, + "learning_rate": 4.575532187740446e-06, + "loss": 0.2107, + "step": 70894 + }, + { + "epoch": 2.73, + "learning_rate": 4.574890997691716e-06, + "loss": 0.1836, + "step": 70895 + }, + { + "epoch": 2.73, + "learning_rate": 4.574249807642985e-06, + "loss": 0.1371, + "step": 70896 + }, + { + "epoch": 2.73, + "learning_rate": 4.5736086175942545e-06, + "loss": 0.1915, + "step": 70897 + }, + { + "epoch": 2.73, + "learning_rate": 4.5729674275455246e-06, + "loss": 0.2409, + "step": 70898 + }, + { + "epoch": 2.73, + "learning_rate": 4.572326237496794e-06, + "loss": 0.2697, + "step": 70899 + }, + { + "epoch": 2.73, + "learning_rate": 4.571685047448064e-06, + "loss": 0.2213, + "step": 70900 + }, + { + "epoch": 2.73, + "learning_rate": 4.571043857399333e-06, + "loss": 0.1299, + "step": 70901 + }, + { + "epoch": 2.73, + "learning_rate": 4.570402667350603e-06, + "loss": 0.2706, + "step": 70902 + }, + { + "epoch": 2.73, + "learning_rate": 4.569761477301872e-06, + "loss": 0.2207, + "step": 70903 + }, + { + "epoch": 2.73, + "learning_rate": 4.569120287253142e-06, + "loss": 0.3437, + "step": 70904 + }, + { + "epoch": 2.73, + "learning_rate": 4.568479097204411e-06, + "loss": 0.246, + "step": 70905 + }, + { + "epoch": 2.73, + "learning_rate": 4.567837907155681e-06, + "loss": 0.1923, + "step": 70906 + }, + { + "epoch": 2.73, + "learning_rate": 4.5671967171069505e-06, + "loss": 0.1856, + "step": 70907 + }, + { + "epoch": 2.73, + "learning_rate": 4.56655552705822e-06, + "loss": 0.1833, + "step": 70908 + }, + { + "epoch": 2.73, + "learning_rate": 4.56591433700949e-06, + "loss": 0.2346, + "step": 70909 + }, + { + "epoch": 2.73, + "learning_rate": 4.565273146960759e-06, + "loss": 0.1905, + "step": 70910 + }, + { + "epoch": 2.73, + "learning_rate": 4.564631956912029e-06, + "loss": 0.2109, + "step": 70911 + }, + { + "epoch": 2.73, + "learning_rate": 4.563990766863298e-06, + "loss": 0.1873, + "step": 70912 + }, + { + "epoch": 2.73, + "learning_rate": 4.563349576814568e-06, + "loss": 0.1727, + "step": 70913 + }, + { + "epoch": 2.73, + "learning_rate": 4.562708386765837e-06, + "loss": 0.3621, + "step": 70914 + }, + { + "epoch": 2.73, + "learning_rate": 4.562067196717107e-06, + "loss": 0.2282, + "step": 70915 + }, + { + "epoch": 2.73, + "learning_rate": 4.5614260066683765e-06, + "loss": 0.2185, + "step": 70916 + }, + { + "epoch": 2.73, + "learning_rate": 4.560784816619646e-06, + "loss": 0.1687, + "step": 70917 + }, + { + "epoch": 2.73, + "learning_rate": 4.560143626570916e-06, + "loss": 0.1621, + "step": 70918 + }, + { + "epoch": 2.73, + "learning_rate": 4.559502436522185e-06, + "loss": 0.1671, + "step": 70919 + }, + { + "epoch": 2.73, + "learning_rate": 4.558861246473455e-06, + "loss": 0.1375, + "step": 70920 + }, + { + "epoch": 2.73, + "learning_rate": 4.558220056424724e-06, + "loss": 0.1406, + "step": 70921 + }, + { + "epoch": 2.73, + "learning_rate": 4.557578866375994e-06, + "loss": 0.129, + "step": 70922 + }, + { + "epoch": 2.73, + "learning_rate": 4.556937676327263e-06, + "loss": 0.1741, + "step": 70923 + }, + { + "epoch": 2.73, + "learning_rate": 4.556296486278533e-06, + "loss": 0.2208, + "step": 70924 + }, + { + "epoch": 2.73, + "learning_rate": 4.5556552962298025e-06, + "loss": 0.1315, + "step": 70925 + }, + { + "epoch": 2.73, + "learning_rate": 4.5550141061810725e-06, + "loss": 0.2304, + "step": 70926 + }, + { + "epoch": 2.73, + "learning_rate": 4.554372916132342e-06, + "loss": 0.3422, + "step": 70927 + }, + { + "epoch": 2.73, + "learning_rate": 4.553731726083611e-06, + "loss": 0.2381, + "step": 70928 + }, + { + "epoch": 2.73, + "learning_rate": 4.553090536034881e-06, + "loss": 0.1526, + "step": 70929 + }, + { + "epoch": 2.73, + "learning_rate": 4.55244934598615e-06, + "loss": 0.1798, + "step": 70930 + }, + { + "epoch": 2.73, + "learning_rate": 4.55180815593742e-06, + "loss": 0.2342, + "step": 70931 + }, + { + "epoch": 2.73, + "learning_rate": 4.551166965888689e-06, + "loss": 0.2978, + "step": 70932 + }, + { + "epoch": 2.73, + "learning_rate": 4.550525775839959e-06, + "loss": 0.1881, + "step": 70933 + }, + { + "epoch": 2.73, + "learning_rate": 4.5498845857912284e-06, + "loss": 0.2912, + "step": 70934 + }, + { + "epoch": 2.73, + "learning_rate": 4.5492433957424985e-06, + "loss": 0.1202, + "step": 70935 + }, + { + "epoch": 2.73, + "learning_rate": 4.548602205693768e-06, + "loss": 0.2101, + "step": 70936 + }, + { + "epoch": 2.73, + "learning_rate": 4.547961015645037e-06, + "loss": 0.3155, + "step": 70937 + }, + { + "epoch": 2.73, + "learning_rate": 4.547319825596307e-06, + "loss": 0.2496, + "step": 70938 + }, + { + "epoch": 2.73, + "learning_rate": 4.546678635547576e-06, + "loss": 0.1231, + "step": 70939 + }, + { + "epoch": 2.73, + "learning_rate": 4.546037445498846e-06, + "loss": 0.2336, + "step": 70940 + }, + { + "epoch": 2.73, + "learning_rate": 4.545396255450115e-06, + "loss": 0.1705, + "step": 70941 + }, + { + "epoch": 2.73, + "learning_rate": 4.544755065401385e-06, + "loss": 0.1933, + "step": 70942 + }, + { + "epoch": 2.73, + "learning_rate": 4.544113875352654e-06, + "loss": 0.1712, + "step": 70943 + }, + { + "epoch": 2.73, + "learning_rate": 4.5434726853039244e-06, + "loss": 0.1851, + "step": 70944 + }, + { + "epoch": 2.73, + "learning_rate": 4.542831495255194e-06, + "loss": 0.3274, + "step": 70945 + }, + { + "epoch": 2.73, + "learning_rate": 4.542190305206464e-06, + "loss": 0.2262, + "step": 70946 + }, + { + "epoch": 2.73, + "learning_rate": 4.541549115157733e-06, + "loss": 0.1755, + "step": 70947 + }, + { + "epoch": 2.73, + "learning_rate": 4.540907925109002e-06, + "loss": 0.1382, + "step": 70948 + }, + { + "epoch": 2.73, + "learning_rate": 4.540266735060272e-06, + "loss": 0.1567, + "step": 70949 + }, + { + "epoch": 2.73, + "learning_rate": 4.539625545011541e-06, + "loss": 0.1968, + "step": 70950 + }, + { + "epoch": 2.73, + "learning_rate": 4.538984354962811e-06, + "loss": 0.2383, + "step": 70951 + }, + { + "epoch": 2.73, + "learning_rate": 4.53834316491408e-06, + "loss": 0.2126, + "step": 70952 + }, + { + "epoch": 2.73, + "learning_rate": 4.53770197486535e-06, + "loss": 0.1795, + "step": 70953 + }, + { + "epoch": 2.73, + "learning_rate": 4.5370607848166196e-06, + "loss": 0.221, + "step": 70954 + }, + { + "epoch": 2.73, + "learning_rate": 4.53641959476789e-06, + "loss": 0.1161, + "step": 70955 + }, + { + "epoch": 2.73, + "learning_rate": 4.535778404719159e-06, + "loss": 0.2797, + "step": 70956 + }, + { + "epoch": 2.73, + "learning_rate": 4.535137214670429e-06, + "loss": 0.2121, + "step": 70957 + }, + { + "epoch": 2.73, + "learning_rate": 4.534496024621698e-06, + "loss": 0.2853, + "step": 70958 + }, + { + "epoch": 2.73, + "learning_rate": 4.533854834572967e-06, + "loss": 0.2457, + "step": 70959 + }, + { + "epoch": 2.73, + "learning_rate": 4.533213644524237e-06, + "loss": 0.1606, + "step": 70960 + }, + { + "epoch": 2.73, + "learning_rate": 4.532572454475506e-06, + "loss": 0.1793, + "step": 70961 + }, + { + "epoch": 2.73, + "learning_rate": 4.531931264426776e-06, + "loss": 0.1856, + "step": 70962 + }, + { + "epoch": 2.73, + "learning_rate": 4.5312900743780455e-06, + "loss": 0.2431, + "step": 70963 + }, + { + "epoch": 2.73, + "learning_rate": 4.5306488843293156e-06, + "loss": 0.2145, + "step": 70964 + }, + { + "epoch": 2.73, + "learning_rate": 4.530007694280585e-06, + "loss": 0.1553, + "step": 70965 + }, + { + "epoch": 2.73, + "learning_rate": 4.529366504231855e-06, + "loss": 0.2284, + "step": 70966 + }, + { + "epoch": 2.73, + "learning_rate": 4.528725314183124e-06, + "loss": 0.1933, + "step": 70967 + }, + { + "epoch": 2.73, + "learning_rate": 4.528084124134393e-06, + "loss": 0.1984, + "step": 70968 + }, + { + "epoch": 2.73, + "learning_rate": 4.527442934085663e-06, + "loss": 0.1509, + "step": 70969 + }, + { + "epoch": 2.73, + "learning_rate": 4.526801744036932e-06, + "loss": 0.2349, + "step": 70970 + }, + { + "epoch": 2.73, + "learning_rate": 4.526160553988202e-06, + "loss": 0.2751, + "step": 70971 + }, + { + "epoch": 2.73, + "learning_rate": 4.5255193639394715e-06, + "loss": 0.2254, + "step": 70972 + }, + { + "epoch": 2.73, + "learning_rate": 4.5248781738907415e-06, + "loss": 0.1213, + "step": 70973 + }, + { + "epoch": 2.73, + "learning_rate": 4.524236983842011e-06, + "loss": 0.2836, + "step": 70974 + }, + { + "epoch": 2.73, + "learning_rate": 4.523595793793281e-06, + "loss": 0.2641, + "step": 70975 + }, + { + "epoch": 2.73, + "learning_rate": 4.52295460374455e-06, + "loss": 0.1729, + "step": 70976 + }, + { + "epoch": 2.73, + "learning_rate": 4.52231341369582e-06, + "loss": 0.1064, + "step": 70977 + }, + { + "epoch": 2.73, + "learning_rate": 4.521672223647089e-06, + "loss": 0.1977, + "step": 70978 + }, + { + "epoch": 2.73, + "learning_rate": 4.521031033598358e-06, + "loss": 0.224, + "step": 70979 + }, + { + "epoch": 2.73, + "learning_rate": 4.520389843549628e-06, + "loss": 0.332, + "step": 70980 + }, + { + "epoch": 2.73, + "learning_rate": 4.5197486535008975e-06, + "loss": 0.0668, + "step": 70981 + }, + { + "epoch": 2.73, + "learning_rate": 4.5191074634521675e-06, + "loss": 0.2823, + "step": 70982 + }, + { + "epoch": 2.73, + "learning_rate": 4.518466273403437e-06, + "loss": 0.2364, + "step": 70983 + }, + { + "epoch": 2.73, + "learning_rate": 4.517825083354707e-06, + "loss": 0.114, + "step": 70984 + }, + { + "epoch": 2.73, + "learning_rate": 4.517183893305976e-06, + "loss": 0.2628, + "step": 70985 + }, + { + "epoch": 2.73, + "learning_rate": 4.516542703257246e-06, + "loss": 0.1821, + "step": 70986 + }, + { + "epoch": 2.73, + "learning_rate": 4.515901513208515e-06, + "loss": 0.1205, + "step": 70987 + }, + { + "epoch": 2.73, + "learning_rate": 4.515260323159784e-06, + "loss": 0.2168, + "step": 70988 + }, + { + "epoch": 2.73, + "learning_rate": 4.514619133111054e-06, + "loss": 0.2521, + "step": 70989 + }, + { + "epoch": 2.73, + "learning_rate": 4.5139779430623234e-06, + "loss": 0.2134, + "step": 70990 + }, + { + "epoch": 2.73, + "learning_rate": 4.5133367530135935e-06, + "loss": 0.1621, + "step": 70991 + }, + { + "epoch": 2.73, + "learning_rate": 4.512695562964863e-06, + "loss": 0.186, + "step": 70992 + }, + { + "epoch": 2.73, + "learning_rate": 4.512054372916133e-06, + "loss": 0.3037, + "step": 70993 + }, + { + "epoch": 2.73, + "learning_rate": 4.511413182867402e-06, + "loss": 0.1636, + "step": 70994 + }, + { + "epoch": 2.73, + "learning_rate": 4.510771992818672e-06, + "loss": 0.1762, + "step": 70995 + }, + { + "epoch": 2.73, + "learning_rate": 4.510130802769941e-06, + "loss": 0.2312, + "step": 70996 + }, + { + "epoch": 2.73, + "learning_rate": 4.509489612721211e-06, + "loss": 0.2588, + "step": 70997 + }, + { + "epoch": 2.73, + "learning_rate": 4.50884842267248e-06, + "loss": 0.2484, + "step": 70998 + }, + { + "epoch": 2.73, + "learning_rate": 4.508207232623749e-06, + "loss": 0.1462, + "step": 70999 + }, + { + "epoch": 2.73, + "learning_rate": 4.5075660425750194e-06, + "loss": 0.2284, + "step": 71000 + }, + { + "epoch": 2.73, + "learning_rate": 4.506924852526289e-06, + "loss": 0.1255, + "step": 71001 + }, + { + "epoch": 2.73, + "learning_rate": 4.506283662477559e-06, + "loss": 0.1417, + "step": 71002 + }, + { + "epoch": 2.73, + "learning_rate": 4.505642472428828e-06, + "loss": 0.3597, + "step": 71003 + }, + { + "epoch": 2.73, + "learning_rate": 4.505001282380098e-06, + "loss": 0.33, + "step": 71004 + }, + { + "epoch": 2.73, + "learning_rate": 4.504360092331367e-06, + "loss": 0.1682, + "step": 71005 + }, + { + "epoch": 2.73, + "learning_rate": 4.503718902282637e-06, + "loss": 0.2713, + "step": 71006 + }, + { + "epoch": 2.73, + "learning_rate": 4.503077712233906e-06, + "loss": 0.4102, + "step": 71007 + }, + { + "epoch": 2.73, + "learning_rate": 4.502436522185176e-06, + "loss": 0.206, + "step": 71008 + }, + { + "epoch": 2.73, + "learning_rate": 4.501795332136445e-06, + "loss": 0.3099, + "step": 71009 + }, + { + "epoch": 2.73, + "learning_rate": 4.501154142087715e-06, + "loss": 0.291, + "step": 71010 + }, + { + "epoch": 2.73, + "learning_rate": 4.500512952038985e-06, + "loss": 0.2676, + "step": 71011 + }, + { + "epoch": 2.73, + "learning_rate": 4.499871761990254e-06, + "loss": 0.3173, + "step": 71012 + }, + { + "epoch": 2.73, + "learning_rate": 4.499230571941524e-06, + "loss": 0.3058, + "step": 71013 + }, + { + "epoch": 2.73, + "learning_rate": 4.498589381892793e-06, + "loss": 0.118, + "step": 71014 + }, + { + "epoch": 2.73, + "learning_rate": 4.497948191844063e-06, + "loss": 0.3064, + "step": 71015 + }, + { + "epoch": 2.73, + "learning_rate": 4.497307001795332e-06, + "loss": 0.3963, + "step": 71016 + }, + { + "epoch": 2.73, + "learning_rate": 4.496665811746602e-06, + "loss": 0.1646, + "step": 71017 + }, + { + "epoch": 2.73, + "learning_rate": 4.496024621697871e-06, + "loss": 0.2919, + "step": 71018 + }, + { + "epoch": 2.73, + "learning_rate": 4.4953834316491405e-06, + "loss": 0.2964, + "step": 71019 + }, + { + "epoch": 2.73, + "learning_rate": 4.4947422416004106e-06, + "loss": 0.1278, + "step": 71020 + }, + { + "epoch": 2.73, + "learning_rate": 4.49410105155168e-06, + "loss": 0.1957, + "step": 71021 + }, + { + "epoch": 2.73, + "learning_rate": 4.49345986150295e-06, + "loss": 0.2067, + "step": 71022 + }, + { + "epoch": 2.73, + "learning_rate": 4.492818671454219e-06, + "loss": 0.1672, + "step": 71023 + }, + { + "epoch": 2.73, + "learning_rate": 4.492177481405489e-06, + "loss": 0.1407, + "step": 71024 + }, + { + "epoch": 2.73, + "learning_rate": 4.491536291356758e-06, + "loss": 0.2172, + "step": 71025 + }, + { + "epoch": 2.73, + "learning_rate": 4.490895101308028e-06, + "loss": 0.2114, + "step": 71026 + }, + { + "epoch": 2.73, + "learning_rate": 4.490253911259297e-06, + "loss": 0.2634, + "step": 71027 + }, + { + "epoch": 2.73, + "learning_rate": 4.489612721210567e-06, + "loss": 0.0824, + "step": 71028 + }, + { + "epoch": 2.73, + "learning_rate": 4.4889715311618365e-06, + "loss": 0.1615, + "step": 71029 + }, + { + "epoch": 2.73, + "learning_rate": 4.488330341113106e-06, + "loss": 0.1823, + "step": 71030 + }, + { + "epoch": 2.73, + "learning_rate": 4.487689151064376e-06, + "loss": 0.2114, + "step": 71031 + }, + { + "epoch": 2.73, + "learning_rate": 4.487047961015645e-06, + "loss": 0.1902, + "step": 71032 + }, + { + "epoch": 2.73, + "learning_rate": 4.486406770966915e-06, + "loss": 0.1945, + "step": 71033 + }, + { + "epoch": 2.73, + "learning_rate": 4.485765580918184e-06, + "loss": 0.3947, + "step": 71034 + }, + { + "epoch": 2.73, + "learning_rate": 4.485124390869454e-06, + "loss": 0.2676, + "step": 71035 + }, + { + "epoch": 2.73, + "learning_rate": 4.484483200820723e-06, + "loss": 0.132, + "step": 71036 + }, + { + "epoch": 2.73, + "learning_rate": 4.483842010771993e-06, + "loss": 0.1356, + "step": 71037 + }, + { + "epoch": 2.73, + "learning_rate": 4.4832008207232625e-06, + "loss": 0.2052, + "step": 71038 + }, + { + "epoch": 2.73, + "learning_rate": 4.4825596306745325e-06, + "loss": 0.199, + "step": 71039 + }, + { + "epoch": 2.73, + "learning_rate": 4.481918440625802e-06, + "loss": 0.3785, + "step": 71040 + }, + { + "epoch": 2.73, + "learning_rate": 4.481277250577071e-06, + "loss": 0.3334, + "step": 71041 + }, + { + "epoch": 2.73, + "learning_rate": 4.480636060528341e-06, + "loss": 0.2367, + "step": 71042 + }, + { + "epoch": 2.73, + "learning_rate": 4.47999487047961e-06, + "loss": 0.2726, + "step": 71043 + }, + { + "epoch": 2.73, + "learning_rate": 4.47935368043088e-06, + "loss": 0.2075, + "step": 71044 + }, + { + "epoch": 2.73, + "learning_rate": 4.478712490382149e-06, + "loss": 0.1493, + "step": 71045 + }, + { + "epoch": 2.73, + "learning_rate": 4.478071300333419e-06, + "loss": 0.2157, + "step": 71046 + }, + { + "epoch": 2.73, + "learning_rate": 4.4774301102846885e-06, + "loss": 0.2539, + "step": 71047 + }, + { + "epoch": 2.73, + "learning_rate": 4.4767889202359585e-06, + "loss": 0.2225, + "step": 71048 + }, + { + "epoch": 2.73, + "learning_rate": 4.476147730187228e-06, + "loss": 0.2396, + "step": 71049 + }, + { + "epoch": 2.73, + "learning_rate": 4.475506540138497e-06, + "loss": 0.1478, + "step": 71050 + }, + { + "epoch": 2.73, + "learning_rate": 4.474865350089767e-06, + "loss": 0.1784, + "step": 71051 + }, + { + "epoch": 2.73, + "learning_rate": 4.474224160041036e-06, + "loss": 0.264, + "step": 71052 + }, + { + "epoch": 2.73, + "learning_rate": 4.473582969992306e-06, + "loss": 0.2729, + "step": 71053 + }, + { + "epoch": 2.73, + "learning_rate": 4.472941779943575e-06, + "loss": 0.2373, + "step": 71054 + }, + { + "epoch": 2.73, + "learning_rate": 4.472300589894845e-06, + "loss": 0.0851, + "step": 71055 + }, + { + "epoch": 2.73, + "learning_rate": 4.4716593998461144e-06, + "loss": 0.2712, + "step": 71056 + }, + { + "epoch": 2.73, + "learning_rate": 4.4710182097973845e-06, + "loss": 0.19, + "step": 71057 + }, + { + "epoch": 2.73, + "learning_rate": 4.470377019748654e-06, + "loss": 0.1935, + "step": 71058 + }, + { + "epoch": 2.73, + "learning_rate": 4.469735829699924e-06, + "loss": 0.1887, + "step": 71059 + }, + { + "epoch": 2.73, + "learning_rate": 4.469094639651193e-06, + "loss": 0.1458, + "step": 71060 + }, + { + "epoch": 2.73, + "learning_rate": 4.468453449602462e-06, + "loss": 0.167, + "step": 71061 + }, + { + "epoch": 2.73, + "learning_rate": 4.467812259553732e-06, + "loss": 0.1202, + "step": 71062 + }, + { + "epoch": 2.73, + "learning_rate": 4.467171069505001e-06, + "loss": 0.2165, + "step": 71063 + }, + { + "epoch": 2.73, + "learning_rate": 4.466529879456271e-06, + "loss": 0.2591, + "step": 71064 + }, + { + "epoch": 2.73, + "learning_rate": 4.46588868940754e-06, + "loss": 0.2104, + "step": 71065 + }, + { + "epoch": 2.73, + "learning_rate": 4.4652474993588104e-06, + "loss": 0.1854, + "step": 71066 + }, + { + "epoch": 2.73, + "learning_rate": 4.46460630931008e-06, + "loss": 0.2527, + "step": 71067 + }, + { + "epoch": 2.73, + "learning_rate": 4.46396511926135e-06, + "loss": 0.2238, + "step": 71068 + }, + { + "epoch": 2.73, + "learning_rate": 4.463323929212619e-06, + "loss": 0.2126, + "step": 71069 + }, + { + "epoch": 2.73, + "learning_rate": 4.462682739163888e-06, + "loss": 0.1689, + "step": 71070 + }, + { + "epoch": 2.73, + "learning_rate": 4.462041549115158e-06, + "loss": 0.3572, + "step": 71071 + }, + { + "epoch": 2.73, + "learning_rate": 4.461400359066427e-06, + "loss": 0.2332, + "step": 71072 + }, + { + "epoch": 2.73, + "learning_rate": 4.460759169017697e-06, + "loss": 0.2154, + "step": 71073 + }, + { + "epoch": 2.73, + "learning_rate": 4.460117978968966e-06, + "loss": 0.2604, + "step": 71074 + }, + { + "epoch": 2.73, + "learning_rate": 4.459476788920236e-06, + "loss": 0.1774, + "step": 71075 + }, + { + "epoch": 2.73, + "learning_rate": 4.458835598871506e-06, + "loss": 0.2151, + "step": 71076 + }, + { + "epoch": 2.73, + "learning_rate": 4.458194408822776e-06, + "loss": 0.3201, + "step": 71077 + }, + { + "epoch": 2.73, + "learning_rate": 4.457553218774045e-06, + "loss": 0.1528, + "step": 71078 + }, + { + "epoch": 2.73, + "learning_rate": 4.456912028725315e-06, + "loss": 0.2031, + "step": 71079 + }, + { + "epoch": 2.73, + "learning_rate": 4.456270838676584e-06, + "loss": 0.2448, + "step": 71080 + }, + { + "epoch": 2.73, + "learning_rate": 4.455629648627853e-06, + "loss": 0.2794, + "step": 71081 + }, + { + "epoch": 2.73, + "learning_rate": 4.454988458579123e-06, + "loss": 0.2941, + "step": 71082 + }, + { + "epoch": 2.73, + "learning_rate": 4.454347268530392e-06, + "loss": 0.2368, + "step": 71083 + }, + { + "epoch": 2.73, + "learning_rate": 4.453706078481662e-06, + "loss": 0.3933, + "step": 71084 + }, + { + "epoch": 2.73, + "learning_rate": 4.4530648884329315e-06, + "loss": 0.1673, + "step": 71085 + }, + { + "epoch": 2.73, + "learning_rate": 4.4524236983842016e-06, + "loss": 0.2029, + "step": 71086 + }, + { + "epoch": 2.73, + "learning_rate": 4.451782508335471e-06, + "loss": 0.3074, + "step": 71087 + }, + { + "epoch": 2.73, + "learning_rate": 4.451141318286741e-06, + "loss": 0.1713, + "step": 71088 + }, + { + "epoch": 2.73, + "learning_rate": 4.45050012823801e-06, + "loss": 0.1826, + "step": 71089 + }, + { + "epoch": 2.73, + "learning_rate": 4.44985893818928e-06, + "loss": 0.3354, + "step": 71090 + }, + { + "epoch": 2.73, + "learning_rate": 4.449217748140549e-06, + "loss": 0.1442, + "step": 71091 + }, + { + "epoch": 2.73, + "learning_rate": 4.448576558091818e-06, + "loss": 0.2218, + "step": 71092 + }, + { + "epoch": 2.73, + "learning_rate": 4.447935368043088e-06, + "loss": 0.1911, + "step": 71093 + }, + { + "epoch": 2.73, + "learning_rate": 4.4472941779943575e-06, + "loss": 0.182, + "step": 71094 + }, + { + "epoch": 2.73, + "learning_rate": 4.4466529879456275e-06, + "loss": 0.2212, + "step": 71095 + }, + { + "epoch": 2.73, + "learning_rate": 4.446011797896897e-06, + "loss": 0.1958, + "step": 71096 + }, + { + "epoch": 2.73, + "learning_rate": 4.445370607848167e-06, + "loss": 0.2131, + "step": 71097 + }, + { + "epoch": 2.73, + "learning_rate": 4.444729417799436e-06, + "loss": 0.1633, + "step": 71098 + }, + { + "epoch": 2.73, + "learning_rate": 4.444088227750706e-06, + "loss": 0.2013, + "step": 71099 + }, + { + "epoch": 2.73, + "learning_rate": 4.443447037701975e-06, + "loss": 0.1465, + "step": 71100 + }, + { + "epoch": 2.73, + "learning_rate": 4.442805847653244e-06, + "loss": 0.1668, + "step": 71101 + }, + { + "epoch": 2.73, + "learning_rate": 4.442164657604514e-06, + "loss": 0.1635, + "step": 71102 + }, + { + "epoch": 2.73, + "learning_rate": 4.4415234675557835e-06, + "loss": 0.2977, + "step": 71103 + }, + { + "epoch": 2.73, + "learning_rate": 4.4408822775070535e-06, + "loss": 0.2112, + "step": 71104 + }, + { + "epoch": 2.73, + "learning_rate": 4.440241087458323e-06, + "loss": 0.3176, + "step": 71105 + }, + { + "epoch": 2.73, + "learning_rate": 4.439599897409593e-06, + "loss": 0.2655, + "step": 71106 + }, + { + "epoch": 2.73, + "learning_rate": 4.438958707360862e-06, + "loss": 0.1098, + "step": 71107 + }, + { + "epoch": 2.73, + "learning_rate": 4.438317517312132e-06, + "loss": 0.2161, + "step": 71108 + }, + { + "epoch": 2.73, + "learning_rate": 4.437676327263401e-06, + "loss": 0.143, + "step": 71109 + }, + { + "epoch": 2.73, + "learning_rate": 4.437035137214671e-06, + "loss": 0.2878, + "step": 71110 + }, + { + "epoch": 2.73, + "learning_rate": 4.43639394716594e-06, + "loss": 0.1451, + "step": 71111 + }, + { + "epoch": 2.73, + "learning_rate": 4.4357527571172094e-06, + "loss": 0.2036, + "step": 71112 + }, + { + "epoch": 2.73, + "learning_rate": 4.4351115670684795e-06, + "loss": 0.1508, + "step": 71113 + }, + { + "epoch": 2.73, + "learning_rate": 4.434470377019749e-06, + "loss": 0.2376, + "step": 71114 + }, + { + "epoch": 2.73, + "learning_rate": 4.433829186971019e-06, + "loss": 0.3028, + "step": 71115 + }, + { + "epoch": 2.73, + "learning_rate": 4.433187996922288e-06, + "loss": 0.1529, + "step": 71116 + }, + { + "epoch": 2.73, + "learning_rate": 4.432546806873558e-06, + "loss": 0.3306, + "step": 71117 + }, + { + "epoch": 2.73, + "learning_rate": 4.431905616824827e-06, + "loss": 0.2741, + "step": 71118 + }, + { + "epoch": 2.73, + "learning_rate": 4.431264426776097e-06, + "loss": 0.1763, + "step": 71119 + }, + { + "epoch": 2.73, + "learning_rate": 4.430623236727366e-06, + "loss": 0.2994, + "step": 71120 + }, + { + "epoch": 2.73, + "learning_rate": 4.429982046678635e-06, + "loss": 0.1741, + "step": 71121 + }, + { + "epoch": 2.73, + "learning_rate": 4.4293408566299054e-06, + "loss": 0.3267, + "step": 71122 + }, + { + "epoch": 2.73, + "learning_rate": 4.428699666581175e-06, + "loss": 0.3167, + "step": 71123 + }, + { + "epoch": 2.73, + "learning_rate": 4.428058476532445e-06, + "loss": 0.153, + "step": 71124 + }, + { + "epoch": 2.73, + "learning_rate": 4.427417286483714e-06, + "loss": 0.2217, + "step": 71125 + }, + { + "epoch": 2.73, + "learning_rate": 4.426776096434984e-06, + "loss": 0.1929, + "step": 71126 + }, + { + "epoch": 2.73, + "learning_rate": 4.426134906386253e-06, + "loss": 0.1684, + "step": 71127 + }, + { + "epoch": 2.73, + "learning_rate": 4.425493716337523e-06, + "loss": 0.258, + "step": 71128 + }, + { + "epoch": 2.73, + "learning_rate": 4.424852526288792e-06, + "loss": 0.2359, + "step": 71129 + }, + { + "epoch": 2.73, + "learning_rate": 4.424211336240062e-06, + "loss": 0.1189, + "step": 71130 + }, + { + "epoch": 2.73, + "learning_rate": 4.423570146191331e-06, + "loss": 0.1868, + "step": 71131 + }, + { + "epoch": 2.73, + "learning_rate": 4.422928956142601e-06, + "loss": 0.2866, + "step": 71132 + }, + { + "epoch": 2.73, + "learning_rate": 4.422287766093871e-06, + "loss": 0.2138, + "step": 71133 + }, + { + "epoch": 2.73, + "learning_rate": 4.42164657604514e-06, + "loss": 0.1151, + "step": 71134 + }, + { + "epoch": 2.73, + "learning_rate": 4.42100538599641e-06, + "loss": 0.2746, + "step": 71135 + }, + { + "epoch": 2.73, + "learning_rate": 4.420364195947679e-06, + "loss": 0.2482, + "step": 71136 + }, + { + "epoch": 2.73, + "learning_rate": 4.419723005898949e-06, + "loss": 0.2109, + "step": 71137 + }, + { + "epoch": 2.74, + "learning_rate": 4.419081815850218e-06, + "loss": 0.3174, + "step": 71138 + }, + { + "epoch": 2.74, + "learning_rate": 4.418440625801488e-06, + "loss": 0.2057, + "step": 71139 + }, + { + "epoch": 2.74, + "learning_rate": 4.417799435752757e-06, + "loss": 0.1507, + "step": 71140 + }, + { + "epoch": 2.74, + "learning_rate": 4.417158245704027e-06, + "loss": 0.1576, + "step": 71141 + }, + { + "epoch": 2.74, + "learning_rate": 4.4165170556552966e-06, + "loss": 0.2521, + "step": 71142 + }, + { + "epoch": 2.74, + "learning_rate": 4.415875865606566e-06, + "loss": 0.2619, + "step": 71143 + }, + { + "epoch": 2.74, + "learning_rate": 4.415234675557836e-06, + "loss": 0.2709, + "step": 71144 + }, + { + "epoch": 2.74, + "learning_rate": 4.414593485509105e-06, + "loss": 0.3805, + "step": 71145 + }, + { + "epoch": 2.74, + "learning_rate": 4.413952295460375e-06, + "loss": 0.1556, + "step": 71146 + }, + { + "epoch": 2.74, + "learning_rate": 4.413311105411644e-06, + "loss": 0.1069, + "step": 71147 + }, + { + "epoch": 2.74, + "learning_rate": 4.412669915362914e-06, + "loss": 0.1784, + "step": 71148 + }, + { + "epoch": 2.74, + "learning_rate": 4.412028725314183e-06, + "loss": 0.1869, + "step": 71149 + }, + { + "epoch": 2.74, + "learning_rate": 4.411387535265453e-06, + "loss": 0.1797, + "step": 71150 + }, + { + "epoch": 2.74, + "learning_rate": 4.4107463452167225e-06, + "loss": 0.2539, + "step": 71151 + }, + { + "epoch": 2.74, + "learning_rate": 4.410105155167992e-06, + "loss": 0.2232, + "step": 71152 + }, + { + "epoch": 2.74, + "learning_rate": 4.409463965119262e-06, + "loss": 0.264, + "step": 71153 + }, + { + "epoch": 2.74, + "learning_rate": 4.408822775070531e-06, + "loss": 0.1853, + "step": 71154 + }, + { + "epoch": 2.74, + "learning_rate": 4.408181585021801e-06, + "loss": 0.2193, + "step": 71155 + }, + { + "epoch": 2.74, + "learning_rate": 4.40754039497307e-06, + "loss": 0.1252, + "step": 71156 + }, + { + "epoch": 2.74, + "learning_rate": 4.40689920492434e-06, + "loss": 0.133, + "step": 71157 + }, + { + "epoch": 2.74, + "learning_rate": 4.406258014875609e-06, + "loss": 0.4206, + "step": 71158 + }, + { + "epoch": 2.74, + "learning_rate": 4.405616824826879e-06, + "loss": 0.1805, + "step": 71159 + }, + { + "epoch": 2.74, + "learning_rate": 4.4049756347781485e-06, + "loss": 0.128, + "step": 71160 + }, + { + "epoch": 2.74, + "learning_rate": 4.4043344447294185e-06, + "loss": 0.1764, + "step": 71161 + }, + { + "epoch": 2.74, + "learning_rate": 4.403693254680688e-06, + "loss": 0.145, + "step": 71162 + }, + { + "epoch": 2.74, + "learning_rate": 4.403052064631957e-06, + "loss": 0.2029, + "step": 71163 + }, + { + "epoch": 2.74, + "learning_rate": 4.402410874583227e-06, + "loss": 0.1732, + "step": 71164 + }, + { + "epoch": 2.74, + "learning_rate": 4.401769684534496e-06, + "loss": 0.0674, + "step": 71165 + }, + { + "epoch": 2.74, + "learning_rate": 4.401128494485766e-06, + "loss": 0.4213, + "step": 71166 + }, + { + "epoch": 2.74, + "learning_rate": 4.400487304437035e-06, + "loss": 0.1629, + "step": 71167 + }, + { + "epoch": 2.74, + "learning_rate": 4.399846114388305e-06, + "loss": 0.1279, + "step": 71168 + }, + { + "epoch": 2.74, + "learning_rate": 4.3992049243395745e-06, + "loss": 0.2098, + "step": 71169 + }, + { + "epoch": 2.74, + "learning_rate": 4.3985637342908445e-06, + "loss": 0.1696, + "step": 71170 + }, + { + "epoch": 2.74, + "learning_rate": 4.397922544242114e-06, + "loss": 0.1413, + "step": 71171 + }, + { + "epoch": 2.74, + "learning_rate": 4.397281354193383e-06, + "loss": 0.2186, + "step": 71172 + }, + { + "epoch": 2.74, + "learning_rate": 4.396640164144653e-06, + "loss": 0.2402, + "step": 71173 + }, + { + "epoch": 2.74, + "learning_rate": 4.395998974095922e-06, + "loss": 0.177, + "step": 71174 + }, + { + "epoch": 2.74, + "learning_rate": 4.395357784047192e-06, + "loss": 0.1256, + "step": 71175 + }, + { + "epoch": 2.74, + "learning_rate": 4.394716593998461e-06, + "loss": 0.1989, + "step": 71176 + }, + { + "epoch": 2.74, + "learning_rate": 4.394075403949731e-06, + "loss": 0.141, + "step": 71177 + }, + { + "epoch": 2.74, + "learning_rate": 4.3934342139010004e-06, + "loss": 0.4152, + "step": 71178 + }, + { + "epoch": 2.74, + "learning_rate": 4.3927930238522705e-06, + "loss": 0.2335, + "step": 71179 + }, + { + "epoch": 2.74, + "learning_rate": 4.39215183380354e-06, + "loss": 0.3321, + "step": 71180 + }, + { + "epoch": 2.74, + "learning_rate": 4.39151064375481e-06, + "loss": 0.1006, + "step": 71181 + }, + { + "epoch": 2.74, + "learning_rate": 4.390869453706079e-06, + "loss": 0.2003, + "step": 71182 + }, + { + "epoch": 2.74, + "learning_rate": 4.390228263657348e-06, + "loss": 0.2445, + "step": 71183 + }, + { + "epoch": 2.74, + "learning_rate": 4.389587073608618e-06, + "loss": 0.3335, + "step": 71184 + }, + { + "epoch": 2.74, + "learning_rate": 4.388945883559887e-06, + "loss": 0.2721, + "step": 71185 + }, + { + "epoch": 2.74, + "learning_rate": 4.388304693511157e-06, + "loss": 0.198, + "step": 71186 + }, + { + "epoch": 2.74, + "learning_rate": 4.387663503462426e-06, + "loss": 0.1529, + "step": 71187 + }, + { + "epoch": 2.74, + "learning_rate": 4.3870223134136964e-06, + "loss": 0.2731, + "step": 71188 + }, + { + "epoch": 2.74, + "learning_rate": 4.386381123364966e-06, + "loss": 0.1572, + "step": 71189 + }, + { + "epoch": 2.74, + "learning_rate": 4.385739933316236e-06, + "loss": 0.231, + "step": 71190 + }, + { + "epoch": 2.74, + "learning_rate": 4.385098743267505e-06, + "loss": 0.1357, + "step": 71191 + }, + { + "epoch": 2.74, + "learning_rate": 4.384457553218775e-06, + "loss": 0.204, + "step": 71192 + }, + { + "epoch": 2.74, + "learning_rate": 4.383816363170044e-06, + "loss": 0.1161, + "step": 71193 + }, + { + "epoch": 2.74, + "learning_rate": 4.383175173121313e-06, + "loss": 0.2641, + "step": 71194 + }, + { + "epoch": 2.74, + "learning_rate": 4.382533983072583e-06, + "loss": 0.335, + "step": 71195 + }, + { + "epoch": 2.74, + "learning_rate": 4.381892793023852e-06, + "loss": 0.1211, + "step": 71196 + }, + { + "epoch": 2.74, + "learning_rate": 4.381251602975122e-06, + "loss": 0.2927, + "step": 71197 + }, + { + "epoch": 2.74, + "learning_rate": 4.380610412926392e-06, + "loss": 0.1212, + "step": 71198 + }, + { + "epoch": 2.74, + "learning_rate": 4.379969222877662e-06, + "loss": 0.2639, + "step": 71199 + }, + { + "epoch": 2.74, + "learning_rate": 4.379328032828931e-06, + "loss": 0.166, + "step": 71200 + }, + { + "epoch": 2.74, + "learning_rate": 4.378686842780201e-06, + "loss": 0.1421, + "step": 71201 + }, + { + "epoch": 2.74, + "learning_rate": 4.37804565273147e-06, + "loss": 0.3495, + "step": 71202 + }, + { + "epoch": 2.74, + "learning_rate": 4.377404462682739e-06, + "loss": 0.2653, + "step": 71203 + }, + { + "epoch": 2.74, + "learning_rate": 4.376763272634009e-06, + "loss": 0.1472, + "step": 71204 + }, + { + "epoch": 2.74, + "learning_rate": 4.376122082585278e-06, + "loss": 0.1034, + "step": 71205 + }, + { + "epoch": 2.74, + "learning_rate": 4.375480892536548e-06, + "loss": 0.3783, + "step": 71206 + }, + { + "epoch": 2.74, + "learning_rate": 4.3748397024878175e-06, + "loss": 0.151, + "step": 71207 + }, + { + "epoch": 2.74, + "learning_rate": 4.3741985124390876e-06, + "loss": 0.2841, + "step": 71208 + }, + { + "epoch": 2.74, + "learning_rate": 4.373557322390357e-06, + "loss": 0.2915, + "step": 71209 + }, + { + "epoch": 2.74, + "learning_rate": 4.372916132341627e-06, + "loss": 0.1667, + "step": 71210 + }, + { + "epoch": 2.74, + "learning_rate": 4.372274942292896e-06, + "loss": 0.1674, + "step": 71211 + }, + { + "epoch": 2.74, + "learning_rate": 4.371633752244166e-06, + "loss": 0.1431, + "step": 71212 + }, + { + "epoch": 2.74, + "learning_rate": 4.370992562195435e-06, + "loss": 0.1453, + "step": 71213 + }, + { + "epoch": 2.74, + "learning_rate": 4.370351372146704e-06, + "loss": 0.1857, + "step": 71214 + }, + { + "epoch": 2.74, + "learning_rate": 4.369710182097974e-06, + "loss": 0.1244, + "step": 71215 + }, + { + "epoch": 2.74, + "learning_rate": 4.3690689920492435e-06, + "loss": 0.262, + "step": 71216 + }, + { + "epoch": 2.74, + "learning_rate": 4.3684278020005135e-06, + "loss": 0.216, + "step": 71217 + }, + { + "epoch": 2.74, + "learning_rate": 4.367786611951783e-06, + "loss": 0.1712, + "step": 71218 + }, + { + "epoch": 2.74, + "learning_rate": 4.367145421903053e-06, + "loss": 0.4647, + "step": 71219 + }, + { + "epoch": 2.74, + "learning_rate": 4.366504231854322e-06, + "loss": 0.3484, + "step": 71220 + }, + { + "epoch": 2.74, + "learning_rate": 4.365863041805592e-06, + "loss": 0.2747, + "step": 71221 + }, + { + "epoch": 2.74, + "learning_rate": 4.365221851756861e-06, + "loss": 0.2801, + "step": 71222 + }, + { + "epoch": 2.74, + "learning_rate": 4.364580661708131e-06, + "loss": 0.1684, + "step": 71223 + }, + { + "epoch": 2.74, + "learning_rate": 4.3639394716594e-06, + "loss": 0.1624, + "step": 71224 + }, + { + "epoch": 2.74, + "learning_rate": 4.3632982816106695e-06, + "loss": 0.1927, + "step": 71225 + }, + { + "epoch": 2.74, + "learning_rate": 4.3626570915619395e-06, + "loss": 0.14, + "step": 71226 + }, + { + "epoch": 2.74, + "learning_rate": 4.362015901513209e-06, + "loss": 0.2207, + "step": 71227 + }, + { + "epoch": 2.74, + "learning_rate": 4.361374711464479e-06, + "loss": 0.2911, + "step": 71228 + }, + { + "epoch": 2.74, + "learning_rate": 4.360733521415748e-06, + "loss": 0.3244, + "step": 71229 + }, + { + "epoch": 2.74, + "learning_rate": 4.360092331367018e-06, + "loss": 0.265, + "step": 71230 + }, + { + "epoch": 2.74, + "learning_rate": 4.359451141318287e-06, + "loss": 0.1035, + "step": 71231 + }, + { + "epoch": 2.74, + "learning_rate": 4.358809951269557e-06, + "loss": 0.1247, + "step": 71232 + }, + { + "epoch": 2.74, + "learning_rate": 4.358168761220826e-06, + "loss": 0.3675, + "step": 71233 + }, + { + "epoch": 2.74, + "learning_rate": 4.3575275711720954e-06, + "loss": 0.2437, + "step": 71234 + }, + { + "epoch": 2.74, + "learning_rate": 4.3568863811233655e-06, + "loss": 0.2026, + "step": 71235 + }, + { + "epoch": 2.74, + "learning_rate": 4.356245191074635e-06, + "loss": 0.4146, + "step": 71236 + }, + { + "epoch": 2.74, + "learning_rate": 4.355604001025905e-06, + "loss": 0.2778, + "step": 71237 + }, + { + "epoch": 2.74, + "learning_rate": 4.354962810977174e-06, + "loss": 0.2068, + "step": 71238 + }, + { + "epoch": 2.74, + "learning_rate": 4.354321620928444e-06, + "loss": 0.215, + "step": 71239 + }, + { + "epoch": 2.74, + "learning_rate": 4.353680430879713e-06, + "loss": 0.2461, + "step": 71240 + }, + { + "epoch": 2.74, + "learning_rate": 4.353039240830983e-06, + "loss": 0.1943, + "step": 71241 + }, + { + "epoch": 2.74, + "learning_rate": 4.352398050782252e-06, + "loss": 0.1519, + "step": 71242 + }, + { + "epoch": 2.74, + "learning_rate": 4.351756860733522e-06, + "loss": 0.1774, + "step": 71243 + }, + { + "epoch": 2.74, + "learning_rate": 4.3511156706847914e-06, + "loss": 0.1259, + "step": 71244 + }, + { + "epoch": 2.74, + "learning_rate": 4.350474480636061e-06, + "loss": 0.3278, + "step": 71245 + }, + { + "epoch": 2.74, + "learning_rate": 4.349833290587331e-06, + "loss": 0.2193, + "step": 71246 + }, + { + "epoch": 2.74, + "learning_rate": 4.3491921005386e-06, + "loss": 0.2266, + "step": 71247 + }, + { + "epoch": 2.74, + "learning_rate": 4.34855091048987e-06, + "loss": 0.246, + "step": 71248 + }, + { + "epoch": 2.74, + "learning_rate": 4.347909720441139e-06, + "loss": 0.2846, + "step": 71249 + }, + { + "epoch": 2.74, + "learning_rate": 4.347268530392409e-06, + "loss": 0.2138, + "step": 71250 + }, + { + "epoch": 2.74, + "learning_rate": 4.346627340343678e-06, + "loss": 0.2443, + "step": 71251 + }, + { + "epoch": 2.74, + "learning_rate": 4.345986150294948e-06, + "loss": 0.1632, + "step": 71252 + }, + { + "epoch": 2.74, + "learning_rate": 4.345344960246217e-06, + "loss": 0.1522, + "step": 71253 + }, + { + "epoch": 2.74, + "learning_rate": 4.344703770197487e-06, + "loss": 0.2525, + "step": 71254 + }, + { + "epoch": 2.74, + "learning_rate": 4.344062580148757e-06, + "loss": 0.2112, + "step": 71255 + }, + { + "epoch": 2.74, + "learning_rate": 4.343421390100026e-06, + "loss": 0.1906, + "step": 71256 + }, + { + "epoch": 2.74, + "learning_rate": 4.342780200051296e-06, + "loss": 0.1392, + "step": 71257 + }, + { + "epoch": 2.74, + "learning_rate": 4.342139010002565e-06, + "loss": 0.2644, + "step": 71258 + }, + { + "epoch": 2.74, + "learning_rate": 4.341497819953835e-06, + "loss": 0.2703, + "step": 71259 + }, + { + "epoch": 2.74, + "learning_rate": 4.340856629905104e-06, + "loss": 0.3516, + "step": 71260 + }, + { + "epoch": 2.74, + "learning_rate": 4.340215439856374e-06, + "loss": 0.1556, + "step": 71261 + }, + { + "epoch": 2.74, + "learning_rate": 4.339574249807643e-06, + "loss": 0.1958, + "step": 71262 + }, + { + "epoch": 2.74, + "learning_rate": 4.338933059758913e-06, + "loss": 0.2337, + "step": 71263 + }, + { + "epoch": 2.74, + "learning_rate": 4.3382918697101826e-06, + "loss": 0.1281, + "step": 71264 + }, + { + "epoch": 2.74, + "learning_rate": 4.337650679661452e-06, + "loss": 0.0862, + "step": 71265 + }, + { + "epoch": 2.74, + "learning_rate": 4.337009489612722e-06, + "loss": 0.0851, + "step": 71266 + }, + { + "epoch": 2.74, + "learning_rate": 4.336368299563991e-06, + "loss": 0.2075, + "step": 71267 + }, + { + "epoch": 2.74, + "learning_rate": 4.335727109515261e-06, + "loss": 0.1554, + "step": 71268 + }, + { + "epoch": 2.74, + "learning_rate": 4.33508591946653e-06, + "loss": 0.2648, + "step": 71269 + }, + { + "epoch": 2.74, + "learning_rate": 4.3344447294178e-06, + "loss": 0.2628, + "step": 71270 + }, + { + "epoch": 2.74, + "learning_rate": 4.333803539369069e-06, + "loss": 0.2378, + "step": 71271 + }, + { + "epoch": 2.74, + "learning_rate": 4.333162349320339e-06, + "loss": 0.1747, + "step": 71272 + }, + { + "epoch": 2.74, + "learning_rate": 4.3325211592716085e-06, + "loss": 0.2363, + "step": 71273 + }, + { + "epoch": 2.74, + "learning_rate": 4.3318799692228786e-06, + "loss": 0.1176, + "step": 71274 + }, + { + "epoch": 2.74, + "learning_rate": 4.331238779174148e-06, + "loss": 0.169, + "step": 71275 + }, + { + "epoch": 2.74, + "learning_rate": 4.330597589125417e-06, + "loss": 0.1928, + "step": 71276 + }, + { + "epoch": 2.74, + "learning_rate": 4.329956399076687e-06, + "loss": 0.2231, + "step": 71277 + }, + { + "epoch": 2.74, + "learning_rate": 4.329315209027956e-06, + "loss": 0.258, + "step": 71278 + }, + { + "epoch": 2.74, + "learning_rate": 4.328674018979226e-06, + "loss": 0.2641, + "step": 71279 + }, + { + "epoch": 2.74, + "learning_rate": 4.328032828930495e-06, + "loss": 0.2025, + "step": 71280 + }, + { + "epoch": 2.74, + "learning_rate": 4.327391638881765e-06, + "loss": 0.224, + "step": 71281 + }, + { + "epoch": 2.74, + "learning_rate": 4.3267504488330345e-06, + "loss": 0.1939, + "step": 71282 + }, + { + "epoch": 2.74, + "learning_rate": 4.3261092587843045e-06, + "loss": 0.0937, + "step": 71283 + }, + { + "epoch": 2.74, + "learning_rate": 4.325468068735574e-06, + "loss": 0.1156, + "step": 71284 + }, + { + "epoch": 2.74, + "learning_rate": 4.324826878686843e-06, + "loss": 0.2994, + "step": 71285 + }, + { + "epoch": 2.74, + "learning_rate": 4.324185688638113e-06, + "loss": 0.4453, + "step": 71286 + }, + { + "epoch": 2.74, + "learning_rate": 4.323544498589382e-06, + "loss": 0.2888, + "step": 71287 + }, + { + "epoch": 2.74, + "learning_rate": 4.322903308540652e-06, + "loss": 0.1904, + "step": 71288 + }, + { + "epoch": 2.74, + "learning_rate": 4.322262118491921e-06, + "loss": 0.3356, + "step": 71289 + }, + { + "epoch": 2.74, + "learning_rate": 4.321620928443191e-06, + "loss": 0.11, + "step": 71290 + }, + { + "epoch": 2.74, + "learning_rate": 4.3209797383944605e-06, + "loss": 0.2135, + "step": 71291 + }, + { + "epoch": 2.74, + "learning_rate": 4.3203385483457305e-06, + "loss": 0.1754, + "step": 71292 + }, + { + "epoch": 2.74, + "learning_rate": 4.319697358297e-06, + "loss": 0.221, + "step": 71293 + }, + { + "epoch": 2.74, + "learning_rate": 4.31905616824827e-06, + "loss": 0.2474, + "step": 71294 + }, + { + "epoch": 2.74, + "learning_rate": 4.318414978199539e-06, + "loss": 0.2744, + "step": 71295 + }, + { + "epoch": 2.74, + "learning_rate": 4.317773788150808e-06, + "loss": 0.2734, + "step": 71296 + }, + { + "epoch": 2.74, + "learning_rate": 4.317132598102078e-06, + "loss": 0.1398, + "step": 71297 + }, + { + "epoch": 2.74, + "learning_rate": 4.316491408053347e-06, + "loss": 0.3499, + "step": 71298 + }, + { + "epoch": 2.74, + "learning_rate": 4.315850218004617e-06, + "loss": 0.1112, + "step": 71299 + }, + { + "epoch": 2.74, + "learning_rate": 4.3152090279558864e-06, + "loss": 0.1658, + "step": 71300 + }, + { + "epoch": 2.74, + "learning_rate": 4.3145678379071565e-06, + "loss": 0.1712, + "step": 71301 + }, + { + "epoch": 2.74, + "learning_rate": 4.313926647858426e-06, + "loss": 0.2079, + "step": 71302 + }, + { + "epoch": 2.74, + "learning_rate": 4.313285457809696e-06, + "loss": 0.1782, + "step": 71303 + }, + { + "epoch": 2.74, + "learning_rate": 4.312644267760965e-06, + "loss": 0.0898, + "step": 71304 + }, + { + "epoch": 2.74, + "learning_rate": 4.312003077712234e-06, + "loss": 0.377, + "step": 71305 + }, + { + "epoch": 2.74, + "learning_rate": 4.311361887663504e-06, + "loss": 0.3452, + "step": 71306 + }, + { + "epoch": 2.74, + "learning_rate": 4.310720697614773e-06, + "loss": 0.1104, + "step": 71307 + }, + { + "epoch": 2.74, + "learning_rate": 4.310079507566043e-06, + "loss": 0.1778, + "step": 71308 + }, + { + "epoch": 2.74, + "learning_rate": 4.309438317517312e-06, + "loss": 0.4003, + "step": 71309 + }, + { + "epoch": 2.74, + "learning_rate": 4.3087971274685824e-06, + "loss": 0.157, + "step": 71310 + }, + { + "epoch": 2.74, + "learning_rate": 4.308155937419852e-06, + "loss": 0.2225, + "step": 71311 + }, + { + "epoch": 2.74, + "learning_rate": 4.307514747371122e-06, + "loss": 0.1406, + "step": 71312 + }, + { + "epoch": 2.74, + "learning_rate": 4.306873557322391e-06, + "loss": 0.1869, + "step": 71313 + }, + { + "epoch": 2.74, + "learning_rate": 4.306232367273661e-06, + "loss": 0.1842, + "step": 71314 + }, + { + "epoch": 2.74, + "learning_rate": 4.30559117722493e-06, + "loss": 0.1946, + "step": 71315 + }, + { + "epoch": 2.74, + "learning_rate": 4.304949987176199e-06, + "loss": 0.225, + "step": 71316 + }, + { + "epoch": 2.74, + "learning_rate": 4.304308797127469e-06, + "loss": 0.2782, + "step": 71317 + }, + { + "epoch": 2.74, + "learning_rate": 4.303667607078738e-06, + "loss": 0.2788, + "step": 71318 + }, + { + "epoch": 2.74, + "learning_rate": 4.303026417030008e-06, + "loss": 0.2666, + "step": 71319 + }, + { + "epoch": 2.74, + "learning_rate": 4.302385226981278e-06, + "loss": 0.2157, + "step": 71320 + }, + { + "epoch": 2.74, + "learning_rate": 4.301744036932548e-06, + "loss": 0.1953, + "step": 71321 + }, + { + "epoch": 2.74, + "learning_rate": 4.301102846883817e-06, + "loss": 0.1439, + "step": 71322 + }, + { + "epoch": 2.74, + "learning_rate": 4.300461656835087e-06, + "loss": 0.2195, + "step": 71323 + }, + { + "epoch": 2.74, + "learning_rate": 4.299820466786356e-06, + "loss": 0.1519, + "step": 71324 + }, + { + "epoch": 2.74, + "learning_rate": 4.299179276737626e-06, + "loss": 0.2247, + "step": 71325 + }, + { + "epoch": 2.74, + "learning_rate": 4.298538086688895e-06, + "loss": 0.2236, + "step": 71326 + }, + { + "epoch": 2.74, + "learning_rate": 4.297896896640164e-06, + "loss": 0.24, + "step": 71327 + }, + { + "epoch": 2.74, + "learning_rate": 4.297255706591434e-06, + "loss": 0.2008, + "step": 71328 + }, + { + "epoch": 2.74, + "learning_rate": 4.2966145165427035e-06, + "loss": 0.3712, + "step": 71329 + }, + { + "epoch": 2.74, + "learning_rate": 4.2959733264939736e-06, + "loss": 0.2586, + "step": 71330 + }, + { + "epoch": 2.74, + "learning_rate": 4.295332136445243e-06, + "loss": 0.1828, + "step": 71331 + }, + { + "epoch": 2.74, + "learning_rate": 4.294690946396513e-06, + "loss": 0.3064, + "step": 71332 + }, + { + "epoch": 2.74, + "learning_rate": 4.294049756347782e-06, + "loss": 0.1196, + "step": 71333 + }, + { + "epoch": 2.74, + "learning_rate": 4.293408566299052e-06, + "loss": 0.3345, + "step": 71334 + }, + { + "epoch": 2.74, + "learning_rate": 4.292767376250321e-06, + "loss": 0.3163, + "step": 71335 + }, + { + "epoch": 2.74, + "learning_rate": 4.29212618620159e-06, + "loss": 0.2788, + "step": 71336 + }, + { + "epoch": 2.74, + "learning_rate": 4.29148499615286e-06, + "loss": 0.1635, + "step": 71337 + }, + { + "epoch": 2.74, + "learning_rate": 4.2908438061041295e-06, + "loss": 0.2334, + "step": 71338 + }, + { + "epoch": 2.74, + "learning_rate": 4.2902026160553995e-06, + "loss": 0.1465, + "step": 71339 + }, + { + "epoch": 2.74, + "learning_rate": 4.289561426006669e-06, + "loss": 0.2808, + "step": 71340 + }, + { + "epoch": 2.74, + "learning_rate": 4.288920235957939e-06, + "loss": 0.2024, + "step": 71341 + }, + { + "epoch": 2.74, + "learning_rate": 4.288279045909208e-06, + "loss": 0.1256, + "step": 71342 + }, + { + "epoch": 2.74, + "learning_rate": 4.287637855860478e-06, + "loss": 0.1593, + "step": 71343 + }, + { + "epoch": 2.74, + "learning_rate": 4.286996665811747e-06, + "loss": 0.2342, + "step": 71344 + }, + { + "epoch": 2.74, + "learning_rate": 4.286355475763016e-06, + "loss": 0.2465, + "step": 71345 + }, + { + "epoch": 2.74, + "learning_rate": 4.285714285714286e-06, + "loss": 0.331, + "step": 71346 + }, + { + "epoch": 2.74, + "learning_rate": 4.2850730956655555e-06, + "loss": 0.1146, + "step": 71347 + }, + { + "epoch": 2.74, + "learning_rate": 4.2844319056168255e-06, + "loss": 0.153, + "step": 71348 + }, + { + "epoch": 2.74, + "learning_rate": 4.283790715568095e-06, + "loss": 0.1192, + "step": 71349 + }, + { + "epoch": 2.74, + "learning_rate": 4.283149525519365e-06, + "loss": 0.1333, + "step": 71350 + }, + { + "epoch": 2.74, + "learning_rate": 4.282508335470634e-06, + "loss": 0.2724, + "step": 71351 + }, + { + "epoch": 2.74, + "learning_rate": 4.281867145421904e-06, + "loss": 0.1342, + "step": 71352 + }, + { + "epoch": 2.74, + "learning_rate": 4.281225955373173e-06, + "loss": 0.1369, + "step": 71353 + }, + { + "epoch": 2.74, + "learning_rate": 4.280584765324443e-06, + "loss": 0.1524, + "step": 71354 + }, + { + "epoch": 2.74, + "learning_rate": 4.279943575275712e-06, + "loss": 0.1242, + "step": 71355 + }, + { + "epoch": 2.74, + "learning_rate": 4.2793023852269815e-06, + "loss": 0.2106, + "step": 71356 + }, + { + "epoch": 2.74, + "learning_rate": 4.2786611951782515e-06, + "loss": 0.1869, + "step": 71357 + }, + { + "epoch": 2.74, + "learning_rate": 4.278020005129521e-06, + "loss": 0.3453, + "step": 71358 + }, + { + "epoch": 2.74, + "learning_rate": 4.277378815080791e-06, + "loss": 0.2237, + "step": 71359 + }, + { + "epoch": 2.74, + "learning_rate": 4.27673762503206e-06, + "loss": 0.2566, + "step": 71360 + }, + { + "epoch": 2.74, + "learning_rate": 4.27609643498333e-06, + "loss": 0.2231, + "step": 71361 + }, + { + "epoch": 2.74, + "learning_rate": 4.275455244934599e-06, + "loss": 0.2685, + "step": 71362 + }, + { + "epoch": 2.74, + "learning_rate": 4.274814054885869e-06, + "loss": 0.3649, + "step": 71363 + }, + { + "epoch": 2.74, + "learning_rate": 4.274172864837138e-06, + "loss": 0.0918, + "step": 71364 + }, + { + "epoch": 2.74, + "learning_rate": 4.273531674788407e-06, + "loss": 0.1729, + "step": 71365 + }, + { + "epoch": 2.74, + "learning_rate": 4.2728904847396774e-06, + "loss": 0.2186, + "step": 71366 + }, + { + "epoch": 2.74, + "learning_rate": 4.272249294690947e-06, + "loss": 0.2897, + "step": 71367 + }, + { + "epoch": 2.74, + "learning_rate": 4.271608104642217e-06, + "loss": 0.2652, + "step": 71368 + }, + { + "epoch": 2.74, + "learning_rate": 4.270966914593486e-06, + "loss": 0.2069, + "step": 71369 + }, + { + "epoch": 2.74, + "learning_rate": 4.270325724544756e-06, + "loss": 0.3256, + "step": 71370 + }, + { + "epoch": 2.74, + "learning_rate": 4.269684534496025e-06, + "loss": 0.1973, + "step": 71371 + }, + { + "epoch": 2.74, + "learning_rate": 4.269043344447295e-06, + "loss": 0.2249, + "step": 71372 + }, + { + "epoch": 2.74, + "learning_rate": 4.268402154398564e-06, + "loss": 0.3101, + "step": 71373 + }, + { + "epoch": 2.74, + "learning_rate": 4.267760964349834e-06, + "loss": 0.1547, + "step": 71374 + }, + { + "epoch": 2.74, + "learning_rate": 4.267119774301103e-06, + "loss": 0.2035, + "step": 71375 + }, + { + "epoch": 2.74, + "learning_rate": 4.266478584252373e-06, + "loss": 0.1841, + "step": 71376 + }, + { + "epoch": 2.74, + "learning_rate": 4.265837394203643e-06, + "loss": 0.2116, + "step": 71377 + }, + { + "epoch": 2.74, + "learning_rate": 4.265196204154912e-06, + "loss": 0.1524, + "step": 71378 + }, + { + "epoch": 2.74, + "learning_rate": 4.264555014106182e-06, + "loss": 0.1562, + "step": 71379 + }, + { + "epoch": 2.74, + "learning_rate": 4.263913824057451e-06, + "loss": 0.2228, + "step": 71380 + }, + { + "epoch": 2.74, + "learning_rate": 4.263272634008721e-06, + "loss": 0.2004, + "step": 71381 + }, + { + "epoch": 2.74, + "learning_rate": 4.26263144395999e-06, + "loss": 0.1335, + "step": 71382 + }, + { + "epoch": 2.74, + "learning_rate": 4.26199025391126e-06, + "loss": 0.1452, + "step": 71383 + }, + { + "epoch": 2.74, + "learning_rate": 4.261349063862529e-06, + "loss": 0.3192, + "step": 71384 + }, + { + "epoch": 2.74, + "learning_rate": 4.2607078738137986e-06, + "loss": 0.2142, + "step": 71385 + }, + { + "epoch": 2.74, + "learning_rate": 4.2600666837650686e-06, + "loss": 0.3288, + "step": 71386 + }, + { + "epoch": 2.74, + "learning_rate": 4.259425493716338e-06, + "loss": 0.1334, + "step": 71387 + }, + { + "epoch": 2.74, + "learning_rate": 4.258784303667608e-06, + "loss": 0.2923, + "step": 71388 + }, + { + "epoch": 2.74, + "learning_rate": 4.258143113618877e-06, + "loss": 0.1791, + "step": 71389 + }, + { + "epoch": 2.74, + "learning_rate": 4.257501923570147e-06, + "loss": 0.2389, + "step": 71390 + }, + { + "epoch": 2.74, + "learning_rate": 4.256860733521416e-06, + "loss": 0.1757, + "step": 71391 + }, + { + "epoch": 2.74, + "learning_rate": 4.256219543472686e-06, + "loss": 0.3335, + "step": 71392 + }, + { + "epoch": 2.74, + "learning_rate": 4.255578353423955e-06, + "loss": 0.2148, + "step": 71393 + }, + { + "epoch": 2.74, + "learning_rate": 4.254937163375225e-06, + "loss": 0.1733, + "step": 71394 + }, + { + "epoch": 2.74, + "learning_rate": 4.2542959733264945e-06, + "loss": 0.2992, + "step": 71395 + }, + { + "epoch": 2.74, + "learning_rate": 4.253654783277764e-06, + "loss": 0.2528, + "step": 71396 + }, + { + "epoch": 2.74, + "learning_rate": 4.253013593229034e-06, + "loss": 0.1907, + "step": 71397 + }, + { + "epoch": 2.75, + "learning_rate": 4.252372403180303e-06, + "loss": 0.2563, + "step": 71398 + }, + { + "epoch": 2.75, + "learning_rate": 4.251731213131573e-06, + "loss": 0.3311, + "step": 71399 + }, + { + "epoch": 2.75, + "learning_rate": 4.251090023082842e-06, + "loss": 0.2035, + "step": 71400 + }, + { + "epoch": 2.75, + "learning_rate": 4.250448833034112e-06, + "loss": 0.2324, + "step": 71401 + }, + { + "epoch": 2.75, + "learning_rate": 4.249807642985381e-06, + "loss": 0.2347, + "step": 71402 + }, + { + "epoch": 2.75, + "learning_rate": 4.249166452936651e-06, + "loss": 0.1248, + "step": 71403 + }, + { + "epoch": 2.75, + "learning_rate": 4.2485252628879205e-06, + "loss": 0.1147, + "step": 71404 + }, + { + "epoch": 2.75, + "learning_rate": 4.24788407283919e-06, + "loss": 0.1669, + "step": 71405 + }, + { + "epoch": 2.75, + "learning_rate": 4.24724288279046e-06, + "loss": 0.4087, + "step": 71406 + }, + { + "epoch": 2.75, + "learning_rate": 4.246601692741729e-06, + "loss": 0.2285, + "step": 71407 + }, + { + "epoch": 2.75, + "learning_rate": 4.245960502692999e-06, + "loss": 0.1884, + "step": 71408 + }, + { + "epoch": 2.75, + "learning_rate": 4.245319312644268e-06, + "loss": 0.3033, + "step": 71409 + }, + { + "epoch": 2.75, + "learning_rate": 4.244678122595538e-06, + "loss": 0.1935, + "step": 71410 + }, + { + "epoch": 2.75, + "learning_rate": 4.244036932546807e-06, + "loss": 0.3844, + "step": 71411 + }, + { + "epoch": 2.75, + "learning_rate": 4.243395742498077e-06, + "loss": 0.1835, + "step": 71412 + }, + { + "epoch": 2.75, + "learning_rate": 4.2427545524493465e-06, + "loss": 0.2223, + "step": 71413 + }, + { + "epoch": 2.75, + "learning_rate": 4.242113362400616e-06, + "loss": 0.1832, + "step": 71414 + }, + { + "epoch": 2.75, + "learning_rate": 4.241472172351886e-06, + "loss": 0.1104, + "step": 71415 + }, + { + "epoch": 2.75, + "learning_rate": 4.240830982303155e-06, + "loss": 0.2082, + "step": 71416 + }, + { + "epoch": 2.75, + "learning_rate": 4.240189792254425e-06, + "loss": 0.181, + "step": 71417 + }, + { + "epoch": 2.75, + "learning_rate": 4.239548602205694e-06, + "loss": 0.1748, + "step": 71418 + }, + { + "epoch": 2.75, + "learning_rate": 4.238907412156964e-06, + "loss": 0.2271, + "step": 71419 + }, + { + "epoch": 2.75, + "learning_rate": 4.238266222108233e-06, + "loss": 0.1465, + "step": 71420 + }, + { + "epoch": 2.75, + "learning_rate": 4.237625032059503e-06, + "loss": 0.302, + "step": 71421 + }, + { + "epoch": 2.75, + "learning_rate": 4.2369838420107724e-06, + "loss": 0.1914, + "step": 71422 + }, + { + "epoch": 2.75, + "learning_rate": 4.2363426519620425e-06, + "loss": 0.2, + "step": 71423 + }, + { + "epoch": 2.75, + "learning_rate": 4.235701461913312e-06, + "loss": 0.3037, + "step": 71424 + }, + { + "epoch": 2.75, + "learning_rate": 4.235060271864581e-06, + "loss": 0.2911, + "step": 71425 + }, + { + "epoch": 2.75, + "learning_rate": 4.234419081815851e-06, + "loss": 0.2465, + "step": 71426 + }, + { + "epoch": 2.75, + "learning_rate": 4.23377789176712e-06, + "loss": 0.301, + "step": 71427 + }, + { + "epoch": 2.75, + "learning_rate": 4.23313670171839e-06, + "loss": 0.0481, + "step": 71428 + }, + { + "epoch": 2.75, + "learning_rate": 4.232495511669659e-06, + "loss": 0.2372, + "step": 71429 + }, + { + "epoch": 2.75, + "learning_rate": 4.231854321620929e-06, + "loss": 0.2218, + "step": 71430 + }, + { + "epoch": 2.75, + "learning_rate": 4.231213131572198e-06, + "loss": 0.2956, + "step": 71431 + }, + { + "epoch": 2.75, + "learning_rate": 4.2305719415234684e-06, + "loss": 0.2177, + "step": 71432 + }, + { + "epoch": 2.75, + "learning_rate": 4.229930751474738e-06, + "loss": 0.1827, + "step": 71433 + }, + { + "epoch": 2.75, + "learning_rate": 4.229289561426007e-06, + "loss": 0.1807, + "step": 71434 + }, + { + "epoch": 2.75, + "learning_rate": 4.228648371377277e-06, + "loss": 0.2489, + "step": 71435 + }, + { + "epoch": 2.75, + "learning_rate": 4.228007181328546e-06, + "loss": 0.1905, + "step": 71436 + }, + { + "epoch": 2.75, + "learning_rate": 4.227365991279816e-06, + "loss": 0.104, + "step": 71437 + }, + { + "epoch": 2.75, + "learning_rate": 4.226724801231085e-06, + "loss": 0.2373, + "step": 71438 + }, + { + "epoch": 2.75, + "learning_rate": 4.226083611182355e-06, + "loss": 0.1766, + "step": 71439 + }, + { + "epoch": 2.75, + "learning_rate": 4.225442421133624e-06, + "loss": 0.1399, + "step": 71440 + }, + { + "epoch": 2.75, + "learning_rate": 4.224801231084894e-06, + "loss": 0.3537, + "step": 71441 + }, + { + "epoch": 2.75, + "learning_rate": 4.224160041036164e-06, + "loss": 0.1827, + "step": 71442 + }, + { + "epoch": 2.75, + "learning_rate": 4.223518850987434e-06, + "loss": 0.1278, + "step": 71443 + }, + { + "epoch": 2.75, + "learning_rate": 4.222877660938703e-06, + "loss": 0.2028, + "step": 71444 + }, + { + "epoch": 2.75, + "learning_rate": 4.222236470889972e-06, + "loss": 0.3361, + "step": 71445 + }, + { + "epoch": 2.75, + "learning_rate": 4.221595280841242e-06, + "loss": 0.1912, + "step": 71446 + }, + { + "epoch": 2.75, + "learning_rate": 4.220954090792511e-06, + "loss": 0.1409, + "step": 71447 + }, + { + "epoch": 2.75, + "learning_rate": 4.220312900743781e-06, + "loss": 0.1729, + "step": 71448 + }, + { + "epoch": 2.75, + "learning_rate": 4.21967171069505e-06, + "loss": 0.1246, + "step": 71449 + }, + { + "epoch": 2.75, + "learning_rate": 4.21903052064632e-06, + "loss": 0.3482, + "step": 71450 + }, + { + "epoch": 2.75, + "learning_rate": 4.2183893305975895e-06, + "loss": 0.2619, + "step": 71451 + }, + { + "epoch": 2.75, + "learning_rate": 4.2177481405488596e-06, + "loss": 0.1914, + "step": 71452 + }, + { + "epoch": 2.75, + "learning_rate": 4.217106950500129e-06, + "loss": 0.195, + "step": 71453 + }, + { + "epoch": 2.75, + "learning_rate": 4.216465760451398e-06, + "loss": 0.2586, + "step": 71454 + }, + { + "epoch": 2.75, + "learning_rate": 4.215824570402668e-06, + "loss": 0.1396, + "step": 71455 + }, + { + "epoch": 2.75, + "learning_rate": 4.215183380353937e-06, + "loss": 0.2931, + "step": 71456 + }, + { + "epoch": 2.75, + "learning_rate": 4.214542190305207e-06, + "loss": 0.2134, + "step": 71457 + }, + { + "epoch": 2.75, + "learning_rate": 4.213901000256476e-06, + "loss": 0.1665, + "step": 71458 + }, + { + "epoch": 2.75, + "learning_rate": 4.213259810207746e-06, + "loss": 0.1756, + "step": 71459 + }, + { + "epoch": 2.75, + "learning_rate": 4.2126186201590155e-06, + "loss": 0.1502, + "step": 71460 + }, + { + "epoch": 2.75, + "learning_rate": 4.2119774301102855e-06, + "loss": 0.2752, + "step": 71461 + }, + { + "epoch": 2.75, + "learning_rate": 4.211336240061555e-06, + "loss": 0.1769, + "step": 71462 + }, + { + "epoch": 2.75, + "learning_rate": 4.210695050012824e-06, + "loss": 0.2578, + "step": 71463 + }, + { + "epoch": 2.75, + "learning_rate": 4.210053859964094e-06, + "loss": 0.2157, + "step": 71464 + }, + { + "epoch": 2.75, + "learning_rate": 4.209412669915363e-06, + "loss": 0.1767, + "step": 71465 + }, + { + "epoch": 2.75, + "learning_rate": 4.208771479866633e-06, + "loss": 0.1141, + "step": 71466 + }, + { + "epoch": 2.75, + "learning_rate": 4.208130289817902e-06, + "loss": 0.1975, + "step": 71467 + }, + { + "epoch": 2.75, + "learning_rate": 4.207489099769172e-06, + "loss": 0.1831, + "step": 71468 + }, + { + "epoch": 2.75, + "learning_rate": 4.2068479097204415e-06, + "loss": 0.2248, + "step": 71469 + }, + { + "epoch": 2.75, + "learning_rate": 4.2062067196717115e-06, + "loss": 0.268, + "step": 71470 + }, + { + "epoch": 2.75, + "learning_rate": 4.205565529622981e-06, + "loss": 0.118, + "step": 71471 + }, + { + "epoch": 2.75, + "learning_rate": 4.204924339574251e-06, + "loss": 0.245, + "step": 71472 + }, + { + "epoch": 2.75, + "learning_rate": 4.20428314952552e-06, + "loss": 0.2349, + "step": 71473 + }, + { + "epoch": 2.75, + "learning_rate": 4.203641959476789e-06, + "loss": 0.1579, + "step": 71474 + }, + { + "epoch": 2.75, + "learning_rate": 4.203000769428059e-06, + "loss": 0.1585, + "step": 71475 + }, + { + "epoch": 2.75, + "learning_rate": 4.202359579379328e-06, + "loss": 0.3049, + "step": 71476 + }, + { + "epoch": 2.75, + "learning_rate": 4.201718389330598e-06, + "loss": 0.2028, + "step": 71477 + }, + { + "epoch": 2.75, + "learning_rate": 4.2010771992818675e-06, + "loss": 0.3596, + "step": 71478 + }, + { + "epoch": 2.75, + "learning_rate": 4.2004360092331375e-06, + "loss": 0.2901, + "step": 71479 + }, + { + "epoch": 2.75, + "learning_rate": 4.199794819184407e-06, + "loss": 0.2425, + "step": 71480 + }, + { + "epoch": 2.75, + "learning_rate": 4.199153629135677e-06, + "loss": 0.1593, + "step": 71481 + }, + { + "epoch": 2.75, + "learning_rate": 4.198512439086946e-06, + "loss": 0.1833, + "step": 71482 + }, + { + "epoch": 2.75, + "learning_rate": 4.197871249038215e-06, + "loss": 0.2867, + "step": 71483 + }, + { + "epoch": 2.75, + "learning_rate": 4.197230058989485e-06, + "loss": 0.2092, + "step": 71484 + }, + { + "epoch": 2.75, + "learning_rate": 4.196588868940754e-06, + "loss": 0.2963, + "step": 71485 + }, + { + "epoch": 2.75, + "learning_rate": 4.195947678892024e-06, + "loss": 0.1484, + "step": 71486 + }, + { + "epoch": 2.75, + "learning_rate": 4.195306488843293e-06, + "loss": 0.1626, + "step": 71487 + }, + { + "epoch": 2.75, + "learning_rate": 4.1946652987945634e-06, + "loss": 0.2185, + "step": 71488 + }, + { + "epoch": 2.75, + "learning_rate": 4.194024108745833e-06, + "loss": 0.175, + "step": 71489 + }, + { + "epoch": 2.75, + "learning_rate": 4.193382918697103e-06, + "loss": 0.2816, + "step": 71490 + }, + { + "epoch": 2.75, + "learning_rate": 4.192741728648372e-06, + "loss": 0.2931, + "step": 71491 + }, + { + "epoch": 2.75, + "learning_rate": 4.192100538599642e-06, + "loss": 0.1514, + "step": 71492 + }, + { + "epoch": 2.75, + "learning_rate": 4.191459348550911e-06, + "loss": 0.2305, + "step": 71493 + }, + { + "epoch": 2.75, + "learning_rate": 4.19081815850218e-06, + "loss": 0.1809, + "step": 71494 + }, + { + "epoch": 2.75, + "learning_rate": 4.19017696845345e-06, + "loss": 0.2245, + "step": 71495 + }, + { + "epoch": 2.75, + "learning_rate": 4.189535778404719e-06, + "loss": 0.1216, + "step": 71496 + }, + { + "epoch": 2.75, + "learning_rate": 4.188894588355989e-06, + "loss": 0.2471, + "step": 71497 + }, + { + "epoch": 2.75, + "learning_rate": 4.188253398307259e-06, + "loss": 0.1636, + "step": 71498 + }, + { + "epoch": 2.75, + "learning_rate": 4.187612208258529e-06, + "loss": 0.2772, + "step": 71499 + }, + { + "epoch": 2.75, + "learning_rate": 4.186971018209798e-06, + "loss": 0.2116, + "step": 71500 + }, + { + "epoch": 2.75, + "learning_rate": 4.186329828161068e-06, + "loss": 0.2555, + "step": 71501 + }, + { + "epoch": 2.75, + "learning_rate": 4.185688638112337e-06, + "loss": 0.1729, + "step": 71502 + }, + { + "epoch": 2.75, + "learning_rate": 4.185047448063606e-06, + "loss": 0.2348, + "step": 71503 + }, + { + "epoch": 2.75, + "learning_rate": 4.184406258014876e-06, + "loss": 0.1374, + "step": 71504 + }, + { + "epoch": 2.75, + "learning_rate": 4.183765067966145e-06, + "loss": 0.2273, + "step": 71505 + }, + { + "epoch": 2.75, + "learning_rate": 4.183123877917415e-06, + "loss": 0.3024, + "step": 71506 + }, + { + "epoch": 2.75, + "learning_rate": 4.1824826878686846e-06, + "loss": 0.3057, + "step": 71507 + }, + { + "epoch": 2.75, + "learning_rate": 4.1818414978199546e-06, + "loss": 0.2629, + "step": 71508 + }, + { + "epoch": 2.75, + "learning_rate": 4.181200307771224e-06, + "loss": 0.2308, + "step": 71509 + }, + { + "epoch": 2.75, + "learning_rate": 4.180559117722494e-06, + "loss": 0.2092, + "step": 71510 + }, + { + "epoch": 2.75, + "learning_rate": 4.179917927673763e-06, + "loss": 0.175, + "step": 71511 + }, + { + "epoch": 2.75, + "learning_rate": 4.179276737625033e-06, + "loss": 0.1284, + "step": 71512 + }, + { + "epoch": 2.75, + "learning_rate": 4.178635547576302e-06, + "loss": 0.1376, + "step": 71513 + }, + { + "epoch": 2.75, + "learning_rate": 4.177994357527571e-06, + "loss": 0.1973, + "step": 71514 + }, + { + "epoch": 2.75, + "learning_rate": 4.177353167478841e-06, + "loss": 0.1138, + "step": 71515 + }, + { + "epoch": 2.75, + "learning_rate": 4.1767119774301105e-06, + "loss": 0.1517, + "step": 71516 + }, + { + "epoch": 2.75, + "learning_rate": 4.1760707873813805e-06, + "loss": 0.1598, + "step": 71517 + }, + { + "epoch": 2.75, + "learning_rate": 4.17542959733265e-06, + "loss": 0.2825, + "step": 71518 + }, + { + "epoch": 2.75, + "learning_rate": 4.17478840728392e-06, + "loss": 0.1239, + "step": 71519 + }, + { + "epoch": 2.75, + "learning_rate": 4.174147217235189e-06, + "loss": 0.1282, + "step": 71520 + }, + { + "epoch": 2.75, + "learning_rate": 4.173506027186459e-06, + "loss": 0.2625, + "step": 71521 + }, + { + "epoch": 2.75, + "learning_rate": 4.172864837137728e-06, + "loss": 0.182, + "step": 71522 + }, + { + "epoch": 2.75, + "learning_rate": 4.172223647088997e-06, + "loss": 0.1264, + "step": 71523 + }, + { + "epoch": 2.75, + "learning_rate": 4.171582457040267e-06, + "loss": 0.1747, + "step": 71524 + }, + { + "epoch": 2.75, + "learning_rate": 4.1709412669915365e-06, + "loss": 0.2469, + "step": 71525 + }, + { + "epoch": 2.75, + "learning_rate": 4.1703000769428065e-06, + "loss": 0.2574, + "step": 71526 + }, + { + "epoch": 2.75, + "learning_rate": 4.169658886894076e-06, + "loss": 0.2488, + "step": 71527 + }, + { + "epoch": 2.75, + "learning_rate": 4.169017696845346e-06, + "loss": 0.1379, + "step": 71528 + }, + { + "epoch": 2.75, + "learning_rate": 4.168376506796615e-06, + "loss": 0.2229, + "step": 71529 + }, + { + "epoch": 2.75, + "learning_rate": 4.167735316747885e-06, + "loss": 0.2016, + "step": 71530 + }, + { + "epoch": 2.75, + "learning_rate": 4.167094126699154e-06, + "loss": 0.1744, + "step": 71531 + }, + { + "epoch": 2.75, + "learning_rate": 4.166452936650423e-06, + "loss": 0.1929, + "step": 71532 + }, + { + "epoch": 2.75, + "learning_rate": 4.165811746601693e-06, + "loss": 0.2862, + "step": 71533 + }, + { + "epoch": 2.75, + "learning_rate": 4.1651705565529625e-06, + "loss": 0.1625, + "step": 71534 + }, + { + "epoch": 2.75, + "learning_rate": 4.1645293665042325e-06, + "loss": 0.178, + "step": 71535 + }, + { + "epoch": 2.75, + "learning_rate": 4.163888176455502e-06, + "loss": 0.1077, + "step": 71536 + }, + { + "epoch": 2.75, + "learning_rate": 4.163246986406772e-06, + "loss": 0.1421, + "step": 71537 + }, + { + "epoch": 2.75, + "learning_rate": 4.162605796358041e-06, + "loss": 0.2236, + "step": 71538 + }, + { + "epoch": 2.75, + "learning_rate": 4.161964606309311e-06, + "loss": 0.21, + "step": 71539 + }, + { + "epoch": 2.75, + "learning_rate": 4.16132341626058e-06, + "loss": 0.1402, + "step": 71540 + }, + { + "epoch": 2.75, + "learning_rate": 4.16068222621185e-06, + "loss": 0.1598, + "step": 71541 + }, + { + "epoch": 2.75, + "learning_rate": 4.160041036163119e-06, + "loss": 0.2251, + "step": 71542 + }, + { + "epoch": 2.75, + "learning_rate": 4.1593998461143884e-06, + "loss": 0.3011, + "step": 71543 + }, + { + "epoch": 2.75, + "learning_rate": 4.1587586560656584e-06, + "loss": 0.1993, + "step": 71544 + }, + { + "epoch": 2.75, + "learning_rate": 4.158117466016928e-06, + "loss": 0.1382, + "step": 71545 + }, + { + "epoch": 2.75, + "learning_rate": 4.157476275968198e-06, + "loss": 0.1787, + "step": 71546 + }, + { + "epoch": 2.75, + "learning_rate": 4.156835085919467e-06, + "loss": 0.1459, + "step": 71547 + }, + { + "epoch": 2.75, + "learning_rate": 4.156193895870737e-06, + "loss": 0.1401, + "step": 71548 + }, + { + "epoch": 2.75, + "learning_rate": 4.155552705822006e-06, + "loss": 0.0792, + "step": 71549 + }, + { + "epoch": 2.75, + "learning_rate": 4.154911515773276e-06, + "loss": 0.2063, + "step": 71550 + }, + { + "epoch": 2.75, + "learning_rate": 4.154270325724545e-06, + "loss": 0.1289, + "step": 71551 + }, + { + "epoch": 2.75, + "learning_rate": 4.153629135675814e-06, + "loss": 0.242, + "step": 71552 + }, + { + "epoch": 2.75, + "learning_rate": 4.152987945627084e-06, + "loss": 0.1964, + "step": 71553 + }, + { + "epoch": 2.75, + "learning_rate": 4.152346755578354e-06, + "loss": 0.317, + "step": 71554 + }, + { + "epoch": 2.75, + "learning_rate": 4.151705565529624e-06, + "loss": 0.2657, + "step": 71555 + }, + { + "epoch": 2.75, + "learning_rate": 4.151064375480893e-06, + "loss": 0.1727, + "step": 71556 + }, + { + "epoch": 2.75, + "learning_rate": 4.150423185432163e-06, + "loss": 0.1748, + "step": 71557 + }, + { + "epoch": 2.75, + "learning_rate": 4.149781995383432e-06, + "loss": 0.1599, + "step": 71558 + }, + { + "epoch": 2.75, + "learning_rate": 4.149140805334702e-06, + "loss": 0.2986, + "step": 71559 + }, + { + "epoch": 2.75, + "learning_rate": 4.148499615285971e-06, + "loss": 0.1961, + "step": 71560 + }, + { + "epoch": 2.75, + "learning_rate": 4.147858425237241e-06, + "loss": 0.2801, + "step": 71561 + }, + { + "epoch": 2.75, + "learning_rate": 4.14721723518851e-06, + "loss": 0.1838, + "step": 71562 + }, + { + "epoch": 2.75, + "learning_rate": 4.1465760451397796e-06, + "loss": 0.2379, + "step": 71563 + }, + { + "epoch": 2.75, + "learning_rate": 4.14593485509105e-06, + "loss": 0.2978, + "step": 71564 + }, + { + "epoch": 2.75, + "learning_rate": 4.145293665042319e-06, + "loss": 0.2994, + "step": 71565 + }, + { + "epoch": 2.75, + "learning_rate": 4.144652474993589e-06, + "loss": 0.2201, + "step": 71566 + }, + { + "epoch": 2.75, + "learning_rate": 4.144011284944858e-06, + "loss": 0.2111, + "step": 71567 + }, + { + "epoch": 2.75, + "learning_rate": 4.143370094896128e-06, + "loss": 0.2341, + "step": 71568 + }, + { + "epoch": 2.75, + "learning_rate": 4.142728904847397e-06, + "loss": 0.1251, + "step": 71569 + }, + { + "epoch": 2.75, + "learning_rate": 4.142087714798667e-06, + "loss": 0.1687, + "step": 71570 + }, + { + "epoch": 2.75, + "learning_rate": 4.141446524749936e-06, + "loss": 0.1689, + "step": 71571 + }, + { + "epoch": 2.75, + "learning_rate": 4.1408053347012055e-06, + "loss": 0.1912, + "step": 71572 + }, + { + "epoch": 2.75, + "learning_rate": 4.1401641446524755e-06, + "loss": 0.2537, + "step": 71573 + }, + { + "epoch": 2.75, + "learning_rate": 4.139522954603745e-06, + "loss": 0.2731, + "step": 71574 + }, + { + "epoch": 2.75, + "learning_rate": 4.138881764555015e-06, + "loss": 0.3108, + "step": 71575 + }, + { + "epoch": 2.75, + "learning_rate": 4.138240574506284e-06, + "loss": 0.2116, + "step": 71576 + }, + { + "epoch": 2.75, + "learning_rate": 4.137599384457554e-06, + "loss": 0.176, + "step": 71577 + }, + { + "epoch": 2.75, + "learning_rate": 4.136958194408823e-06, + "loss": 0.1389, + "step": 71578 + }, + { + "epoch": 2.75, + "learning_rate": 4.136317004360093e-06, + "loss": 0.1642, + "step": 71579 + }, + { + "epoch": 2.75, + "learning_rate": 4.135675814311362e-06, + "loss": 0.3636, + "step": 71580 + }, + { + "epoch": 2.75, + "learning_rate": 4.1350346242626315e-06, + "loss": 0.1336, + "step": 71581 + }, + { + "epoch": 2.75, + "learning_rate": 4.1343934342139015e-06, + "loss": 0.2082, + "step": 71582 + }, + { + "epoch": 2.75, + "learning_rate": 4.133752244165171e-06, + "loss": 0.1275, + "step": 71583 + }, + { + "epoch": 2.75, + "learning_rate": 4.133111054116441e-06, + "loss": 0.379, + "step": 71584 + }, + { + "epoch": 2.75, + "learning_rate": 4.13246986406771e-06, + "loss": 0.1314, + "step": 71585 + }, + { + "epoch": 2.75, + "learning_rate": 4.13182867401898e-06, + "loss": 0.1304, + "step": 71586 + }, + { + "epoch": 2.75, + "learning_rate": 4.131187483970249e-06, + "loss": 0.2609, + "step": 71587 + }, + { + "epoch": 2.75, + "learning_rate": 4.130546293921519e-06, + "loss": 0.2635, + "step": 71588 + }, + { + "epoch": 2.75, + "learning_rate": 4.129905103872788e-06, + "loss": 0.2165, + "step": 71589 + }, + { + "epoch": 2.75, + "learning_rate": 4.129263913824058e-06, + "loss": 0.2256, + "step": 71590 + }, + { + "epoch": 2.75, + "learning_rate": 4.1286227237753275e-06, + "loss": 0.241, + "step": 71591 + }, + { + "epoch": 2.75, + "learning_rate": 4.127981533726597e-06, + "loss": 0.2246, + "step": 71592 + }, + { + "epoch": 2.75, + "learning_rate": 4.127340343677867e-06, + "loss": 0.2505, + "step": 71593 + }, + { + "epoch": 2.75, + "learning_rate": 4.126699153629136e-06, + "loss": 0.1817, + "step": 71594 + }, + { + "epoch": 2.75, + "learning_rate": 4.126057963580406e-06, + "loss": 0.1595, + "step": 71595 + }, + { + "epoch": 2.75, + "learning_rate": 4.125416773531675e-06, + "loss": 0.2922, + "step": 71596 + }, + { + "epoch": 2.75, + "learning_rate": 4.124775583482945e-06, + "loss": 0.1427, + "step": 71597 + }, + { + "epoch": 2.75, + "learning_rate": 4.124134393434214e-06, + "loss": 0.2585, + "step": 71598 + }, + { + "epoch": 2.75, + "learning_rate": 4.123493203385484e-06, + "loss": 0.3206, + "step": 71599 + }, + { + "epoch": 2.75, + "learning_rate": 4.1228520133367535e-06, + "loss": 0.3578, + "step": 71600 + }, + { + "epoch": 2.75, + "learning_rate": 4.122210823288023e-06, + "loss": 0.1901, + "step": 71601 + }, + { + "epoch": 2.75, + "learning_rate": 4.121569633239293e-06, + "loss": 0.174, + "step": 71602 + }, + { + "epoch": 2.75, + "learning_rate": 4.120928443190562e-06, + "loss": 0.3067, + "step": 71603 + }, + { + "epoch": 2.75, + "learning_rate": 4.120287253141832e-06, + "loss": 0.1874, + "step": 71604 + }, + { + "epoch": 2.75, + "learning_rate": 4.119646063093101e-06, + "loss": 0.1848, + "step": 71605 + }, + { + "epoch": 2.75, + "learning_rate": 4.119004873044371e-06, + "loss": 0.3644, + "step": 71606 + }, + { + "epoch": 2.75, + "learning_rate": 4.11836368299564e-06, + "loss": 0.1468, + "step": 71607 + }, + { + "epoch": 2.75, + "learning_rate": 4.11772249294691e-06, + "loss": 0.2323, + "step": 71608 + }, + { + "epoch": 2.75, + "learning_rate": 4.117081302898179e-06, + "loss": 0.1367, + "step": 71609 + }, + { + "epoch": 2.75, + "learning_rate": 4.1164401128494494e-06, + "loss": 0.1697, + "step": 71610 + }, + { + "epoch": 2.75, + "learning_rate": 4.115798922800719e-06, + "loss": 0.3067, + "step": 71611 + }, + { + "epoch": 2.75, + "learning_rate": 4.115157732751988e-06, + "loss": 0.1801, + "step": 71612 + }, + { + "epoch": 2.75, + "learning_rate": 4.114516542703258e-06, + "loss": 0.3157, + "step": 71613 + }, + { + "epoch": 2.75, + "learning_rate": 4.113875352654527e-06, + "loss": 0.255, + "step": 71614 + }, + { + "epoch": 2.75, + "learning_rate": 4.113234162605797e-06, + "loss": 0.3732, + "step": 71615 + }, + { + "epoch": 2.75, + "learning_rate": 4.112592972557066e-06, + "loss": 0.2929, + "step": 71616 + }, + { + "epoch": 2.75, + "learning_rate": 4.111951782508336e-06, + "loss": 0.29, + "step": 71617 + }, + { + "epoch": 2.75, + "learning_rate": 4.111310592459605e-06, + "loss": 0.2562, + "step": 71618 + }, + { + "epoch": 2.75, + "learning_rate": 4.110669402410875e-06, + "loss": 0.2206, + "step": 71619 + }, + { + "epoch": 2.75, + "learning_rate": 4.110028212362145e-06, + "loss": 0.1855, + "step": 71620 + }, + { + "epoch": 2.75, + "learning_rate": 4.109387022313414e-06, + "loss": 0.2353, + "step": 71621 + }, + { + "epoch": 2.75, + "learning_rate": 4.108745832264684e-06, + "loss": 0.275, + "step": 71622 + }, + { + "epoch": 2.75, + "learning_rate": 4.108104642215953e-06, + "loss": 0.2519, + "step": 71623 + }, + { + "epoch": 2.75, + "learning_rate": 4.107463452167223e-06, + "loss": 0.2426, + "step": 71624 + }, + { + "epoch": 2.75, + "learning_rate": 4.106822262118492e-06, + "loss": 0.1433, + "step": 71625 + }, + { + "epoch": 2.75, + "learning_rate": 4.106181072069762e-06, + "loss": 0.1901, + "step": 71626 + }, + { + "epoch": 2.75, + "learning_rate": 4.105539882021031e-06, + "loss": 0.185, + "step": 71627 + }, + { + "epoch": 2.75, + "learning_rate": 4.104898691972301e-06, + "loss": 0.2936, + "step": 71628 + }, + { + "epoch": 2.75, + "learning_rate": 4.1042575019235706e-06, + "loss": 0.1389, + "step": 71629 + }, + { + "epoch": 2.75, + "learning_rate": 4.10361631187484e-06, + "loss": 0.2068, + "step": 71630 + }, + { + "epoch": 2.75, + "learning_rate": 4.10297512182611e-06, + "loss": 0.1343, + "step": 71631 + }, + { + "epoch": 2.75, + "learning_rate": 4.102333931777379e-06, + "loss": 0.2117, + "step": 71632 + }, + { + "epoch": 2.75, + "learning_rate": 4.101692741728649e-06, + "loss": 0.3404, + "step": 71633 + }, + { + "epoch": 2.75, + "learning_rate": 4.101051551679918e-06, + "loss": 0.2333, + "step": 71634 + }, + { + "epoch": 2.75, + "learning_rate": 4.100410361631188e-06, + "loss": 0.1406, + "step": 71635 + }, + { + "epoch": 2.75, + "learning_rate": 4.099769171582457e-06, + "loss": 0.262, + "step": 71636 + }, + { + "epoch": 2.75, + "learning_rate": 4.099127981533727e-06, + "loss": 0.189, + "step": 71637 + }, + { + "epoch": 2.75, + "learning_rate": 4.0984867914849965e-06, + "loss": 0.3463, + "step": 71638 + }, + { + "epoch": 2.75, + "learning_rate": 4.0978456014362665e-06, + "loss": 0.2287, + "step": 71639 + }, + { + "epoch": 2.75, + "learning_rate": 4.097204411387536e-06, + "loss": 0.1551, + "step": 71640 + }, + { + "epoch": 2.75, + "learning_rate": 4.096563221338805e-06, + "loss": 0.4531, + "step": 71641 + }, + { + "epoch": 2.75, + "learning_rate": 4.095922031290075e-06, + "loss": 0.1875, + "step": 71642 + }, + { + "epoch": 2.75, + "learning_rate": 4.095280841241344e-06, + "loss": 0.2317, + "step": 71643 + }, + { + "epoch": 2.75, + "learning_rate": 4.094639651192614e-06, + "loss": 0.2152, + "step": 71644 + }, + { + "epoch": 2.75, + "learning_rate": 4.093998461143883e-06, + "loss": 0.1282, + "step": 71645 + }, + { + "epoch": 2.75, + "learning_rate": 4.093357271095153e-06, + "loss": 0.1785, + "step": 71646 + }, + { + "epoch": 2.75, + "learning_rate": 4.0927160810464225e-06, + "loss": 0.2972, + "step": 71647 + }, + { + "epoch": 2.75, + "learning_rate": 4.0920748909976925e-06, + "loss": 0.2633, + "step": 71648 + }, + { + "epoch": 2.75, + "learning_rate": 4.091433700948962e-06, + "loss": 0.2238, + "step": 71649 + }, + { + "epoch": 2.75, + "learning_rate": 4.090792510900231e-06, + "loss": 0.2189, + "step": 71650 + }, + { + "epoch": 2.75, + "learning_rate": 4.090151320851501e-06, + "loss": 0.2569, + "step": 71651 + }, + { + "epoch": 2.75, + "learning_rate": 4.08951013080277e-06, + "loss": 0.2734, + "step": 71652 + }, + { + "epoch": 2.75, + "learning_rate": 4.08886894075404e-06, + "loss": 0.2252, + "step": 71653 + }, + { + "epoch": 2.75, + "learning_rate": 4.088227750705309e-06, + "loss": 0.2367, + "step": 71654 + }, + { + "epoch": 2.75, + "learning_rate": 4.087586560656579e-06, + "loss": 0.2573, + "step": 71655 + }, + { + "epoch": 2.75, + "learning_rate": 4.0869453706078485e-06, + "loss": 0.323, + "step": 71656 + }, + { + "epoch": 2.75, + "learning_rate": 4.0863041805591185e-06, + "loss": 0.2633, + "step": 71657 + }, + { + "epoch": 2.76, + "learning_rate": 4.085662990510388e-06, + "loss": 0.1931, + "step": 71658 + }, + { + "epoch": 2.76, + "learning_rate": 4.085021800461658e-06, + "loss": 0.305, + "step": 71659 + }, + { + "epoch": 2.76, + "learning_rate": 4.084380610412927e-06, + "loss": 0.2731, + "step": 71660 + }, + { + "epoch": 2.76, + "learning_rate": 4.083739420364196e-06, + "loss": 0.2024, + "step": 71661 + }, + { + "epoch": 2.76, + "learning_rate": 4.083098230315466e-06, + "loss": 0.3933, + "step": 71662 + }, + { + "epoch": 2.76, + "learning_rate": 4.082457040266735e-06, + "loss": 0.0967, + "step": 71663 + }, + { + "epoch": 2.76, + "learning_rate": 4.081815850218005e-06, + "loss": 0.2121, + "step": 71664 + }, + { + "epoch": 2.76, + "learning_rate": 4.0811746601692744e-06, + "loss": 0.2937, + "step": 71665 + }, + { + "epoch": 2.76, + "learning_rate": 4.0805334701205444e-06, + "loss": 0.1774, + "step": 71666 + }, + { + "epoch": 2.76, + "learning_rate": 4.079892280071814e-06, + "loss": 0.3508, + "step": 71667 + }, + { + "epoch": 2.76, + "learning_rate": 4.079251090023084e-06, + "loss": 0.178, + "step": 71668 + }, + { + "epoch": 2.76, + "learning_rate": 4.078609899974353e-06, + "loss": 0.1399, + "step": 71669 + }, + { + "epoch": 2.76, + "learning_rate": 4.077968709925622e-06, + "loss": 0.3019, + "step": 71670 + }, + { + "epoch": 2.76, + "learning_rate": 4.077327519876892e-06, + "loss": 0.2841, + "step": 71671 + }, + { + "epoch": 2.76, + "learning_rate": 4.076686329828161e-06, + "loss": 0.133, + "step": 71672 + }, + { + "epoch": 2.76, + "learning_rate": 4.076045139779431e-06, + "loss": 0.2322, + "step": 71673 + }, + { + "epoch": 2.76, + "learning_rate": 4.0754039497307e-06, + "loss": 0.1252, + "step": 71674 + }, + { + "epoch": 2.76, + "learning_rate": 4.07476275968197e-06, + "loss": 0.2845, + "step": 71675 + }, + { + "epoch": 2.76, + "learning_rate": 4.07412156963324e-06, + "loss": 0.1559, + "step": 71676 + }, + { + "epoch": 2.76, + "learning_rate": 4.07348037958451e-06, + "loss": 0.1275, + "step": 71677 + }, + { + "epoch": 2.76, + "learning_rate": 4.072839189535779e-06, + "loss": 0.1727, + "step": 71678 + }, + { + "epoch": 2.76, + "learning_rate": 4.072197999487049e-06, + "loss": 0.2363, + "step": 71679 + }, + { + "epoch": 2.76, + "learning_rate": 4.071556809438318e-06, + "loss": 0.1829, + "step": 71680 + }, + { + "epoch": 2.76, + "learning_rate": 4.070915619389587e-06, + "loss": 0.1513, + "step": 71681 + }, + { + "epoch": 2.76, + "learning_rate": 4.070274429340857e-06, + "loss": 0.321, + "step": 71682 + }, + { + "epoch": 2.76, + "learning_rate": 4.069633239292126e-06, + "loss": 0.1381, + "step": 71683 + }, + { + "epoch": 2.76, + "learning_rate": 4.068992049243396e-06, + "loss": 0.2038, + "step": 71684 + }, + { + "epoch": 2.76, + "learning_rate": 4.0683508591946656e-06, + "loss": 0.191, + "step": 71685 + }, + { + "epoch": 2.76, + "learning_rate": 4.067709669145936e-06, + "loss": 0.1405, + "step": 71686 + }, + { + "epoch": 2.76, + "learning_rate": 4.067068479097205e-06, + "loss": 0.3208, + "step": 71687 + }, + { + "epoch": 2.76, + "learning_rate": 4.066427289048475e-06, + "loss": 0.2479, + "step": 71688 + }, + { + "epoch": 2.76, + "learning_rate": 4.065786098999744e-06, + "loss": 0.2109, + "step": 71689 + }, + { + "epoch": 2.76, + "learning_rate": 4.065144908951013e-06, + "loss": 0.2203, + "step": 71690 + }, + { + "epoch": 2.76, + "learning_rate": 4.064503718902283e-06, + "loss": 0.1841, + "step": 71691 + }, + { + "epoch": 2.76, + "learning_rate": 4.063862528853552e-06, + "loss": 0.219, + "step": 71692 + }, + { + "epoch": 2.76, + "learning_rate": 4.063221338804822e-06, + "loss": 0.2786, + "step": 71693 + }, + { + "epoch": 2.76, + "learning_rate": 4.0625801487560915e-06, + "loss": 0.2861, + "step": 71694 + }, + { + "epoch": 2.76, + "learning_rate": 4.0619389587073615e-06, + "loss": 0.2054, + "step": 71695 + }, + { + "epoch": 2.76, + "learning_rate": 4.061297768658631e-06, + "loss": 0.2682, + "step": 71696 + }, + { + "epoch": 2.76, + "learning_rate": 4.060656578609901e-06, + "loss": 0.2065, + "step": 71697 + }, + { + "epoch": 2.76, + "learning_rate": 4.06001538856117e-06, + "loss": 0.1813, + "step": 71698 + }, + { + "epoch": 2.76, + "learning_rate": 4.059374198512439e-06, + "loss": 0.1809, + "step": 71699 + }, + { + "epoch": 2.76, + "learning_rate": 4.058733008463709e-06, + "loss": 0.2234, + "step": 71700 + }, + { + "epoch": 2.76, + "learning_rate": 4.058091818414978e-06, + "loss": 0.2816, + "step": 71701 + }, + { + "epoch": 2.76, + "learning_rate": 4.057450628366248e-06, + "loss": 0.0891, + "step": 71702 + }, + { + "epoch": 2.76, + "learning_rate": 4.0568094383175175e-06, + "loss": 0.1887, + "step": 71703 + }, + { + "epoch": 2.76, + "learning_rate": 4.0561682482687875e-06, + "loss": 0.2028, + "step": 71704 + }, + { + "epoch": 2.76, + "learning_rate": 4.055527058220057e-06, + "loss": 0.131, + "step": 71705 + }, + { + "epoch": 2.76, + "learning_rate": 4.054885868171327e-06, + "loss": 0.4456, + "step": 71706 + }, + { + "epoch": 2.76, + "learning_rate": 4.054244678122596e-06, + "loss": 0.1333, + "step": 71707 + }, + { + "epoch": 2.76, + "learning_rate": 4.053603488073866e-06, + "loss": 0.185, + "step": 71708 + }, + { + "epoch": 2.76, + "learning_rate": 4.052962298025135e-06, + "loss": 0.2115, + "step": 71709 + }, + { + "epoch": 2.76, + "learning_rate": 4.052321107976404e-06, + "loss": 0.2976, + "step": 71710 + }, + { + "epoch": 2.76, + "learning_rate": 4.051679917927674e-06, + "loss": 0.1744, + "step": 71711 + }, + { + "epoch": 2.76, + "learning_rate": 4.0510387278789435e-06, + "loss": 0.2496, + "step": 71712 + }, + { + "epoch": 2.76, + "learning_rate": 4.0503975378302135e-06, + "loss": 0.1255, + "step": 71713 + }, + { + "epoch": 2.76, + "learning_rate": 4.049756347781483e-06, + "loss": 0.2283, + "step": 71714 + }, + { + "epoch": 2.76, + "learning_rate": 4.049115157732753e-06, + "loss": 0.2747, + "step": 71715 + }, + { + "epoch": 2.76, + "learning_rate": 4.048473967684022e-06, + "loss": 0.2854, + "step": 71716 + }, + { + "epoch": 2.76, + "learning_rate": 4.047832777635292e-06, + "loss": 0.1027, + "step": 71717 + }, + { + "epoch": 2.76, + "learning_rate": 4.047191587586561e-06, + "loss": 0.2129, + "step": 71718 + }, + { + "epoch": 2.76, + "learning_rate": 4.04655039753783e-06, + "loss": 0.3078, + "step": 71719 + }, + { + "epoch": 2.76, + "learning_rate": 4.0459092074891e-06, + "loss": 0.2546, + "step": 71720 + }, + { + "epoch": 2.76, + "learning_rate": 4.0452680174403694e-06, + "loss": 0.197, + "step": 71721 + }, + { + "epoch": 2.76, + "learning_rate": 4.0446268273916395e-06, + "loss": 0.1278, + "step": 71722 + }, + { + "epoch": 2.76, + "learning_rate": 4.043985637342909e-06, + "loss": 0.2569, + "step": 71723 + }, + { + "epoch": 2.76, + "learning_rate": 4.043344447294179e-06, + "loss": 0.0963, + "step": 71724 + }, + { + "epoch": 2.76, + "learning_rate": 4.042703257245448e-06, + "loss": 0.2191, + "step": 71725 + }, + { + "epoch": 2.76, + "learning_rate": 4.042062067196718e-06, + "loss": 0.1902, + "step": 71726 + }, + { + "epoch": 2.76, + "learning_rate": 4.041420877147987e-06, + "loss": 0.2933, + "step": 71727 + }, + { + "epoch": 2.76, + "learning_rate": 4.040779687099257e-06, + "loss": 0.3195, + "step": 71728 + }, + { + "epoch": 2.76, + "learning_rate": 4.040138497050526e-06, + "loss": 0.1523, + "step": 71729 + }, + { + "epoch": 2.76, + "learning_rate": 4.039497307001795e-06, + "loss": 0.22, + "step": 71730 + }, + { + "epoch": 2.76, + "learning_rate": 4.038856116953065e-06, + "loss": 0.1545, + "step": 71731 + }, + { + "epoch": 2.76, + "learning_rate": 4.038214926904335e-06, + "loss": 0.1959, + "step": 71732 + }, + { + "epoch": 2.76, + "learning_rate": 4.037573736855605e-06, + "loss": 0.1603, + "step": 71733 + }, + { + "epoch": 2.76, + "learning_rate": 4.036932546806874e-06, + "loss": 0.1028, + "step": 71734 + }, + { + "epoch": 2.76, + "learning_rate": 4.036291356758144e-06, + "loss": 0.3004, + "step": 71735 + }, + { + "epoch": 2.76, + "learning_rate": 4.035650166709413e-06, + "loss": 0.1532, + "step": 71736 + }, + { + "epoch": 2.76, + "learning_rate": 4.035008976660683e-06, + "loss": 0.2621, + "step": 71737 + }, + { + "epoch": 2.76, + "learning_rate": 4.034367786611952e-06, + "loss": 0.2011, + "step": 71738 + }, + { + "epoch": 2.76, + "learning_rate": 4.033726596563221e-06, + "loss": 0.1476, + "step": 71739 + }, + { + "epoch": 2.76, + "learning_rate": 4.033085406514491e-06, + "loss": 0.1922, + "step": 71740 + }, + { + "epoch": 2.76, + "learning_rate": 4.0324442164657606e-06, + "loss": 0.2749, + "step": 71741 + }, + { + "epoch": 2.76, + "learning_rate": 4.031803026417031e-06, + "loss": 0.1643, + "step": 71742 + }, + { + "epoch": 2.76, + "learning_rate": 4.0311618363683e-06, + "loss": 0.1869, + "step": 71743 + }, + { + "epoch": 2.76, + "learning_rate": 4.03052064631957e-06, + "loss": 0.245, + "step": 71744 + }, + { + "epoch": 2.76, + "learning_rate": 4.029879456270839e-06, + "loss": 0.0737, + "step": 71745 + }, + { + "epoch": 2.76, + "learning_rate": 4.029238266222109e-06, + "loss": 0.0886, + "step": 71746 + }, + { + "epoch": 2.76, + "learning_rate": 4.028597076173378e-06, + "loss": 0.4046, + "step": 71747 + }, + { + "epoch": 2.76, + "learning_rate": 4.027955886124647e-06, + "loss": 0.2308, + "step": 71748 + }, + { + "epoch": 2.76, + "learning_rate": 4.027314696075917e-06, + "loss": 0.1897, + "step": 71749 + }, + { + "epoch": 2.76, + "learning_rate": 4.0266735060271865e-06, + "loss": 0.1587, + "step": 71750 + }, + { + "epoch": 2.76, + "learning_rate": 4.0260323159784566e-06, + "loss": 0.1755, + "step": 71751 + }, + { + "epoch": 2.76, + "learning_rate": 4.025391125929726e-06, + "loss": 0.1107, + "step": 71752 + }, + { + "epoch": 2.76, + "learning_rate": 4.024749935880996e-06, + "loss": 0.1868, + "step": 71753 + }, + { + "epoch": 2.76, + "learning_rate": 4.024108745832265e-06, + "loss": 0.2915, + "step": 71754 + }, + { + "epoch": 2.76, + "learning_rate": 4.023467555783535e-06, + "loss": 0.2761, + "step": 71755 + }, + { + "epoch": 2.76, + "learning_rate": 4.022826365734804e-06, + "loss": 0.0985, + "step": 71756 + }, + { + "epoch": 2.76, + "learning_rate": 4.022185175686074e-06, + "loss": 0.1959, + "step": 71757 + }, + { + "epoch": 2.76, + "learning_rate": 4.021543985637343e-06, + "loss": 0.1879, + "step": 71758 + }, + { + "epoch": 2.76, + "learning_rate": 4.0209027955886125e-06, + "loss": 0.1936, + "step": 71759 + }, + { + "epoch": 2.76, + "learning_rate": 4.0202616055398825e-06, + "loss": 0.1069, + "step": 71760 + }, + { + "epoch": 2.76, + "learning_rate": 4.019620415491152e-06, + "loss": 0.1573, + "step": 71761 + }, + { + "epoch": 2.76, + "learning_rate": 4.018979225442422e-06, + "loss": 0.422, + "step": 71762 + }, + { + "epoch": 2.76, + "learning_rate": 4.018338035393691e-06, + "loss": 0.1022, + "step": 71763 + }, + { + "epoch": 2.76, + "learning_rate": 4.017696845344961e-06, + "loss": 0.1821, + "step": 71764 + }, + { + "epoch": 2.76, + "learning_rate": 4.01705565529623e-06, + "loss": 0.1336, + "step": 71765 + }, + { + "epoch": 2.76, + "learning_rate": 4.0164144652475e-06, + "loss": 0.2562, + "step": 71766 + }, + { + "epoch": 2.76, + "learning_rate": 4.015773275198769e-06, + "loss": 0.2453, + "step": 71767 + }, + { + "epoch": 2.76, + "learning_rate": 4.0151320851500385e-06, + "loss": 0.1579, + "step": 71768 + }, + { + "epoch": 2.76, + "learning_rate": 4.0144908951013085e-06, + "loss": 0.1362, + "step": 71769 + }, + { + "epoch": 2.76, + "learning_rate": 4.013849705052578e-06, + "loss": 0.1139, + "step": 71770 + }, + { + "epoch": 2.76, + "learning_rate": 4.013208515003848e-06, + "loss": 0.1493, + "step": 71771 + }, + { + "epoch": 2.76, + "learning_rate": 4.012567324955117e-06, + "loss": 0.2442, + "step": 71772 + }, + { + "epoch": 2.76, + "learning_rate": 4.011926134906387e-06, + "loss": 0.189, + "step": 71773 + }, + { + "epoch": 2.76, + "learning_rate": 4.011284944857656e-06, + "loss": 0.1491, + "step": 71774 + }, + { + "epoch": 2.76, + "learning_rate": 4.010643754808926e-06, + "loss": 0.0832, + "step": 71775 + }, + { + "epoch": 2.76, + "learning_rate": 4.010002564760195e-06, + "loss": 0.2785, + "step": 71776 + }, + { + "epoch": 2.76, + "learning_rate": 4.009361374711465e-06, + "loss": 0.1799, + "step": 71777 + }, + { + "epoch": 2.76, + "learning_rate": 4.0087201846627345e-06, + "loss": 0.1932, + "step": 71778 + }, + { + "epoch": 2.76, + "learning_rate": 4.008078994614004e-06, + "loss": 0.1867, + "step": 71779 + }, + { + "epoch": 2.76, + "learning_rate": 4.007437804565274e-06, + "loss": 0.0983, + "step": 71780 + }, + { + "epoch": 2.76, + "learning_rate": 4.006796614516543e-06, + "loss": 0.2798, + "step": 71781 + }, + { + "epoch": 2.76, + "learning_rate": 4.006155424467813e-06, + "loss": 0.2106, + "step": 71782 + }, + { + "epoch": 2.76, + "learning_rate": 4.005514234419082e-06, + "loss": 0.246, + "step": 71783 + }, + { + "epoch": 2.76, + "learning_rate": 4.004873044370352e-06, + "loss": 0.2926, + "step": 71784 + }, + { + "epoch": 2.76, + "learning_rate": 4.004231854321621e-06, + "loss": 0.2101, + "step": 71785 + }, + { + "epoch": 2.76, + "learning_rate": 4.003590664272891e-06, + "loss": 0.304, + "step": 71786 + }, + { + "epoch": 2.76, + "learning_rate": 4.0029494742241604e-06, + "loss": 0.1371, + "step": 71787 + }, + { + "epoch": 2.76, + "learning_rate": 4.00230828417543e-06, + "loss": 0.2735, + "step": 71788 + }, + { + "epoch": 2.76, + "learning_rate": 4.0016670941267e-06, + "loss": 0.1842, + "step": 71789 + }, + { + "epoch": 2.76, + "learning_rate": 4.001025904077969e-06, + "loss": 0.1875, + "step": 71790 + }, + { + "epoch": 2.76, + "learning_rate": 4.000384714029239e-06, + "loss": 0.2011, + "step": 71791 + }, + { + "epoch": 2.76, + "learning_rate": 3.999743523980508e-06, + "loss": 0.2644, + "step": 71792 + }, + { + "epoch": 2.76, + "learning_rate": 3.999102333931778e-06, + "loss": 0.1927, + "step": 71793 + }, + { + "epoch": 2.76, + "learning_rate": 3.998461143883047e-06, + "loss": 0.1253, + "step": 71794 + }, + { + "epoch": 2.76, + "learning_rate": 3.997819953834317e-06, + "loss": 0.2254, + "step": 71795 + }, + { + "epoch": 2.76, + "learning_rate": 3.997178763785586e-06, + "loss": 0.175, + "step": 71796 + }, + { + "epoch": 2.76, + "learning_rate": 3.996537573736856e-06, + "loss": 0.2049, + "step": 71797 + }, + { + "epoch": 2.76, + "learning_rate": 3.995896383688126e-06, + "loss": 0.2947, + "step": 71798 + }, + { + "epoch": 2.76, + "learning_rate": 3.995255193639395e-06, + "loss": 0.2319, + "step": 71799 + }, + { + "epoch": 2.76, + "learning_rate": 3.994614003590665e-06, + "loss": 0.287, + "step": 71800 + }, + { + "epoch": 2.76, + "learning_rate": 3.993972813541934e-06, + "loss": 0.2439, + "step": 71801 + }, + { + "epoch": 2.76, + "learning_rate": 3.993331623493204e-06, + "loss": 0.1904, + "step": 71802 + }, + { + "epoch": 2.76, + "learning_rate": 3.992690433444473e-06, + "loss": 0.104, + "step": 71803 + }, + { + "epoch": 2.76, + "learning_rate": 3.992049243395743e-06, + "loss": 0.1818, + "step": 71804 + }, + { + "epoch": 2.76, + "learning_rate": 3.991408053347012e-06, + "loss": 0.1373, + "step": 71805 + }, + { + "epoch": 2.76, + "learning_rate": 3.990766863298282e-06, + "loss": 0.22, + "step": 71806 + }, + { + "epoch": 2.76, + "learning_rate": 3.9901256732495516e-06, + "loss": 0.1739, + "step": 71807 + }, + { + "epoch": 2.76, + "learning_rate": 3.989484483200821e-06, + "loss": 0.2383, + "step": 71808 + }, + { + "epoch": 2.76, + "learning_rate": 3.988843293152091e-06, + "loss": 0.1332, + "step": 71809 + }, + { + "epoch": 2.76, + "learning_rate": 3.98820210310336e-06, + "loss": 0.2083, + "step": 71810 + }, + { + "epoch": 2.76, + "learning_rate": 3.98756091305463e-06, + "loss": 0.2446, + "step": 71811 + }, + { + "epoch": 2.76, + "learning_rate": 3.986919723005899e-06, + "loss": 0.0921, + "step": 71812 + }, + { + "epoch": 2.76, + "learning_rate": 3.986278532957169e-06, + "loss": 0.1152, + "step": 71813 + }, + { + "epoch": 2.76, + "learning_rate": 3.985637342908438e-06, + "loss": 0.2309, + "step": 71814 + }, + { + "epoch": 2.76, + "learning_rate": 3.984996152859708e-06, + "loss": 0.2441, + "step": 71815 + }, + { + "epoch": 2.76, + "learning_rate": 3.9843549628109775e-06, + "loss": 0.3357, + "step": 71816 + }, + { + "epoch": 2.76, + "learning_rate": 3.983713772762247e-06, + "loss": 0.2055, + "step": 71817 + }, + { + "epoch": 2.76, + "learning_rate": 3.983072582713517e-06, + "loss": 0.2338, + "step": 71818 + }, + { + "epoch": 2.76, + "learning_rate": 3.982431392664786e-06, + "loss": 0.2881, + "step": 71819 + }, + { + "epoch": 2.76, + "learning_rate": 3.981790202616056e-06, + "loss": 0.3715, + "step": 71820 + }, + { + "epoch": 2.76, + "learning_rate": 3.981149012567325e-06, + "loss": 0.1808, + "step": 71821 + }, + { + "epoch": 2.76, + "learning_rate": 3.980507822518595e-06, + "loss": 0.2122, + "step": 71822 + }, + { + "epoch": 2.76, + "learning_rate": 3.979866632469864e-06, + "loss": 0.1499, + "step": 71823 + }, + { + "epoch": 2.76, + "learning_rate": 3.979225442421134e-06, + "loss": 0.2069, + "step": 71824 + }, + { + "epoch": 2.76, + "learning_rate": 3.9785842523724035e-06, + "loss": 0.2167, + "step": 71825 + }, + { + "epoch": 2.76, + "learning_rate": 3.9779430623236735e-06, + "loss": 0.253, + "step": 71826 + }, + { + "epoch": 2.76, + "learning_rate": 3.977301872274943e-06, + "loss": 0.2729, + "step": 71827 + }, + { + "epoch": 2.76, + "learning_rate": 3.976660682226212e-06, + "loss": 0.0995, + "step": 71828 + }, + { + "epoch": 2.76, + "learning_rate": 3.976019492177482e-06, + "loss": 0.0854, + "step": 71829 + }, + { + "epoch": 2.76, + "learning_rate": 3.975378302128751e-06, + "loss": 0.1544, + "step": 71830 + }, + { + "epoch": 2.76, + "learning_rate": 3.974737112080021e-06, + "loss": 0.3464, + "step": 71831 + }, + { + "epoch": 2.76, + "learning_rate": 3.97409592203129e-06, + "loss": 0.1504, + "step": 71832 + }, + { + "epoch": 2.76, + "learning_rate": 3.97345473198256e-06, + "loss": 0.1549, + "step": 71833 + }, + { + "epoch": 2.76, + "learning_rate": 3.9728135419338295e-06, + "loss": 0.1645, + "step": 71834 + }, + { + "epoch": 2.76, + "learning_rate": 3.9721723518850995e-06, + "loss": 0.2131, + "step": 71835 + }, + { + "epoch": 2.76, + "learning_rate": 3.971531161836369e-06, + "loss": 0.3133, + "step": 71836 + }, + { + "epoch": 2.76, + "learning_rate": 3.970889971787638e-06, + "loss": 0.4522, + "step": 71837 + }, + { + "epoch": 2.76, + "learning_rate": 3.970248781738908e-06, + "loss": 0.1981, + "step": 71838 + }, + { + "epoch": 2.76, + "learning_rate": 3.969607591690177e-06, + "loss": 0.083, + "step": 71839 + }, + { + "epoch": 2.76, + "learning_rate": 3.968966401641447e-06, + "loss": 0.1923, + "step": 71840 + }, + { + "epoch": 2.76, + "learning_rate": 3.968325211592716e-06, + "loss": 0.2027, + "step": 71841 + }, + { + "epoch": 2.76, + "learning_rate": 3.967684021543986e-06, + "loss": 0.1041, + "step": 71842 + }, + { + "epoch": 2.76, + "learning_rate": 3.9670428314952554e-06, + "loss": 0.1223, + "step": 71843 + }, + { + "epoch": 2.76, + "learning_rate": 3.9664016414465255e-06, + "loss": 0.2586, + "step": 71844 + }, + { + "epoch": 2.76, + "learning_rate": 3.965760451397795e-06, + "loss": 0.2122, + "step": 71845 + }, + { + "epoch": 2.76, + "learning_rate": 3.965119261349065e-06, + "loss": 0.2576, + "step": 71846 + }, + { + "epoch": 2.76, + "learning_rate": 3.964478071300334e-06, + "loss": 0.2379, + "step": 71847 + }, + { + "epoch": 2.76, + "learning_rate": 3.963836881251603e-06, + "loss": 0.1648, + "step": 71848 + }, + { + "epoch": 2.76, + "learning_rate": 3.963195691202873e-06, + "loss": 0.1917, + "step": 71849 + }, + { + "epoch": 2.76, + "learning_rate": 3.962554501154142e-06, + "loss": 0.1311, + "step": 71850 + }, + { + "epoch": 2.76, + "learning_rate": 3.961913311105412e-06, + "loss": 0.2863, + "step": 71851 + }, + { + "epoch": 2.76, + "learning_rate": 3.961272121056681e-06, + "loss": 0.1441, + "step": 71852 + }, + { + "epoch": 2.76, + "learning_rate": 3.960630931007951e-06, + "loss": 0.1111, + "step": 71853 + }, + { + "epoch": 2.76, + "learning_rate": 3.959989740959221e-06, + "loss": 0.1325, + "step": 71854 + }, + { + "epoch": 2.76, + "learning_rate": 3.959348550910491e-06, + "loss": 0.3044, + "step": 71855 + }, + { + "epoch": 2.76, + "learning_rate": 3.95870736086176e-06, + "loss": 0.1881, + "step": 71856 + }, + { + "epoch": 2.76, + "learning_rate": 3.958066170813029e-06, + "loss": 0.345, + "step": 71857 + }, + { + "epoch": 2.76, + "learning_rate": 3.957424980764299e-06, + "loss": 0.3049, + "step": 71858 + }, + { + "epoch": 2.76, + "learning_rate": 3.956783790715568e-06, + "loss": 0.093, + "step": 71859 + }, + { + "epoch": 2.76, + "learning_rate": 3.956142600666838e-06, + "loss": 0.1596, + "step": 71860 + }, + { + "epoch": 2.76, + "learning_rate": 3.955501410618107e-06, + "loss": 0.1891, + "step": 71861 + }, + { + "epoch": 2.76, + "learning_rate": 3.954860220569377e-06, + "loss": 0.2646, + "step": 71862 + }, + { + "epoch": 2.76, + "learning_rate": 3.9542190305206466e-06, + "loss": 0.1973, + "step": 71863 + }, + { + "epoch": 2.76, + "learning_rate": 3.953577840471917e-06, + "loss": 0.1739, + "step": 71864 + }, + { + "epoch": 2.76, + "learning_rate": 3.952936650423186e-06, + "loss": 0.1412, + "step": 71865 + }, + { + "epoch": 2.76, + "learning_rate": 3.952295460374455e-06, + "loss": 0.1619, + "step": 71866 + }, + { + "epoch": 2.76, + "learning_rate": 3.951654270325725e-06, + "loss": 0.2725, + "step": 71867 + }, + { + "epoch": 2.76, + "learning_rate": 3.951013080276994e-06, + "loss": 0.2187, + "step": 71868 + }, + { + "epoch": 2.76, + "learning_rate": 3.950371890228264e-06, + "loss": 0.1812, + "step": 71869 + }, + { + "epoch": 2.76, + "learning_rate": 3.949730700179533e-06, + "loss": 0.2273, + "step": 71870 + }, + { + "epoch": 2.76, + "learning_rate": 3.949089510130803e-06, + "loss": 0.3362, + "step": 71871 + }, + { + "epoch": 2.76, + "learning_rate": 3.9484483200820725e-06, + "loss": 0.2318, + "step": 71872 + }, + { + "epoch": 2.76, + "learning_rate": 3.9478071300333426e-06, + "loss": 0.12, + "step": 71873 + }, + { + "epoch": 2.76, + "learning_rate": 3.947165939984612e-06, + "loss": 0.2253, + "step": 71874 + }, + { + "epoch": 2.76, + "learning_rate": 3.946524749935882e-06, + "loss": 0.2258, + "step": 71875 + }, + { + "epoch": 2.76, + "learning_rate": 3.945883559887151e-06, + "loss": 0.2617, + "step": 71876 + }, + { + "epoch": 2.76, + "learning_rate": 3.94524236983842e-06, + "loss": 0.2492, + "step": 71877 + }, + { + "epoch": 2.76, + "learning_rate": 3.94460117978969e-06, + "loss": 0.2704, + "step": 71878 + }, + { + "epoch": 2.76, + "learning_rate": 3.943959989740959e-06, + "loss": 0.4237, + "step": 71879 + }, + { + "epoch": 2.76, + "learning_rate": 3.943318799692229e-06, + "loss": 0.3265, + "step": 71880 + }, + { + "epoch": 2.76, + "learning_rate": 3.9426776096434985e-06, + "loss": 0.1707, + "step": 71881 + }, + { + "epoch": 2.76, + "learning_rate": 3.9420364195947685e-06, + "loss": 0.2559, + "step": 71882 + }, + { + "epoch": 2.76, + "learning_rate": 3.941395229546038e-06, + "loss": 0.0453, + "step": 71883 + }, + { + "epoch": 2.76, + "learning_rate": 3.940754039497308e-06, + "loss": 0.2565, + "step": 71884 + }, + { + "epoch": 2.76, + "learning_rate": 3.940112849448577e-06, + "loss": 0.2355, + "step": 71885 + }, + { + "epoch": 2.76, + "learning_rate": 3.939471659399846e-06, + "loss": 0.1995, + "step": 71886 + }, + { + "epoch": 2.76, + "learning_rate": 3.938830469351116e-06, + "loss": 0.1694, + "step": 71887 + }, + { + "epoch": 2.76, + "learning_rate": 3.938189279302385e-06, + "loss": 0.2807, + "step": 71888 + }, + { + "epoch": 2.76, + "learning_rate": 3.937548089253655e-06, + "loss": 0.1946, + "step": 71889 + }, + { + "epoch": 2.76, + "learning_rate": 3.9369068992049245e-06, + "loss": 0.4054, + "step": 71890 + }, + { + "epoch": 2.76, + "learning_rate": 3.9362657091561945e-06, + "loss": 0.1205, + "step": 71891 + }, + { + "epoch": 2.76, + "learning_rate": 3.935624519107464e-06, + "loss": 0.1207, + "step": 71892 + }, + { + "epoch": 2.76, + "learning_rate": 3.934983329058734e-06, + "loss": 0.1131, + "step": 71893 + }, + { + "epoch": 2.76, + "learning_rate": 3.934342139010003e-06, + "loss": 0.2673, + "step": 71894 + }, + { + "epoch": 2.76, + "learning_rate": 3.933700948961273e-06, + "loss": 0.1937, + "step": 71895 + }, + { + "epoch": 2.76, + "learning_rate": 3.933059758912542e-06, + "loss": 0.3225, + "step": 71896 + }, + { + "epoch": 2.76, + "learning_rate": 3.932418568863811e-06, + "loss": 0.1515, + "step": 71897 + }, + { + "epoch": 2.76, + "learning_rate": 3.931777378815081e-06, + "loss": 0.1805, + "step": 71898 + }, + { + "epoch": 2.76, + "learning_rate": 3.9311361887663504e-06, + "loss": 0.367, + "step": 71899 + }, + { + "epoch": 2.76, + "learning_rate": 3.9304949987176205e-06, + "loss": 0.3059, + "step": 71900 + }, + { + "epoch": 2.76, + "learning_rate": 3.92985380866889e-06, + "loss": 0.3392, + "step": 71901 + }, + { + "epoch": 2.76, + "learning_rate": 3.92921261862016e-06, + "loss": 0.0843, + "step": 71902 + }, + { + "epoch": 2.76, + "learning_rate": 3.928571428571429e-06, + "loss": 0.2111, + "step": 71903 + }, + { + "epoch": 2.76, + "learning_rate": 3.927930238522699e-06, + "loss": 0.1706, + "step": 71904 + }, + { + "epoch": 2.76, + "learning_rate": 3.927289048473968e-06, + "loss": 0.2901, + "step": 71905 + }, + { + "epoch": 2.76, + "learning_rate": 3.926647858425237e-06, + "loss": 0.2018, + "step": 71906 + }, + { + "epoch": 2.76, + "learning_rate": 3.926006668376507e-06, + "loss": 0.2522, + "step": 71907 + }, + { + "epoch": 2.76, + "learning_rate": 3.925365478327776e-06, + "loss": 0.1718, + "step": 71908 + }, + { + "epoch": 2.76, + "learning_rate": 3.9247242882790464e-06, + "loss": 0.1483, + "step": 71909 + }, + { + "epoch": 2.76, + "learning_rate": 3.924083098230316e-06, + "loss": 0.2849, + "step": 71910 + }, + { + "epoch": 2.76, + "learning_rate": 3.923441908181586e-06, + "loss": 0.2562, + "step": 71911 + }, + { + "epoch": 2.76, + "learning_rate": 3.922800718132855e-06, + "loss": 0.3068, + "step": 71912 + }, + { + "epoch": 2.76, + "learning_rate": 3.922159528084125e-06, + "loss": 0.2262, + "step": 71913 + }, + { + "epoch": 2.76, + "learning_rate": 3.921518338035394e-06, + "loss": 0.164, + "step": 71914 + }, + { + "epoch": 2.76, + "learning_rate": 3.920877147986664e-06, + "loss": 0.1481, + "step": 71915 + }, + { + "epoch": 2.76, + "learning_rate": 3.920235957937933e-06, + "loss": 0.1151, + "step": 71916 + }, + { + "epoch": 2.76, + "learning_rate": 3.919594767889202e-06, + "loss": 0.1483, + "step": 71917 + }, + { + "epoch": 2.77, + "learning_rate": 3.918953577840472e-06, + "loss": 0.1864, + "step": 71918 + }, + { + "epoch": 2.77, + "learning_rate": 3.9183123877917416e-06, + "loss": 0.1763, + "step": 71919 + }, + { + "epoch": 2.77, + "learning_rate": 3.917671197743012e-06, + "loss": 0.2251, + "step": 71920 + }, + { + "epoch": 2.77, + "learning_rate": 3.917030007694281e-06, + "loss": 0.1962, + "step": 71921 + }, + { + "epoch": 2.77, + "learning_rate": 3.916388817645551e-06, + "loss": 0.1898, + "step": 71922 + }, + { + "epoch": 2.77, + "learning_rate": 3.91574762759682e-06, + "loss": 0.1097, + "step": 71923 + }, + { + "epoch": 2.77, + "learning_rate": 3.91510643754809e-06, + "loss": 0.2915, + "step": 71924 + }, + { + "epoch": 2.77, + "learning_rate": 3.914465247499359e-06, + "loss": 0.2011, + "step": 71925 + }, + { + "epoch": 2.77, + "learning_rate": 3.913824057450628e-06, + "loss": 0.1929, + "step": 71926 + }, + { + "epoch": 2.77, + "learning_rate": 3.913182867401898e-06, + "loss": 0.2748, + "step": 71927 + }, + { + "epoch": 2.77, + "learning_rate": 3.9125416773531675e-06, + "loss": 0.1218, + "step": 71928 + }, + { + "epoch": 2.77, + "learning_rate": 3.9119004873044376e-06, + "loss": 0.214, + "step": 71929 + }, + { + "epoch": 2.77, + "learning_rate": 3.911259297255707e-06, + "loss": 0.2474, + "step": 71930 + }, + { + "epoch": 2.77, + "learning_rate": 3.910618107206977e-06, + "loss": 0.1532, + "step": 71931 + }, + { + "epoch": 2.77, + "learning_rate": 3.909976917158246e-06, + "loss": 0.201, + "step": 71932 + }, + { + "epoch": 2.77, + "learning_rate": 3.909335727109516e-06, + "loss": 0.2282, + "step": 71933 + }, + { + "epoch": 2.77, + "learning_rate": 3.908694537060785e-06, + "loss": 0.1559, + "step": 71934 + }, + { + "epoch": 2.77, + "learning_rate": 3.908053347012054e-06, + "loss": 0.2788, + "step": 71935 + }, + { + "epoch": 2.77, + "learning_rate": 3.907412156963324e-06, + "loss": 0.1148, + "step": 71936 + }, + { + "epoch": 2.77, + "learning_rate": 3.9067709669145935e-06, + "loss": 0.2794, + "step": 71937 + }, + { + "epoch": 2.77, + "learning_rate": 3.9061297768658635e-06, + "loss": 0.1626, + "step": 71938 + }, + { + "epoch": 2.77, + "learning_rate": 3.905488586817133e-06, + "loss": 0.205, + "step": 71939 + }, + { + "epoch": 2.77, + "learning_rate": 3.904847396768403e-06, + "loss": 0.2218, + "step": 71940 + }, + { + "epoch": 2.77, + "learning_rate": 3.904206206719672e-06, + "loss": 0.2315, + "step": 71941 + }, + { + "epoch": 2.77, + "learning_rate": 3.903565016670942e-06, + "loss": 0.1698, + "step": 71942 + }, + { + "epoch": 2.77, + "learning_rate": 3.902923826622211e-06, + "loss": 0.2241, + "step": 71943 + }, + { + "epoch": 2.77, + "learning_rate": 3.902282636573481e-06, + "loss": 0.1806, + "step": 71944 + }, + { + "epoch": 2.77, + "learning_rate": 3.90164144652475e-06, + "loss": 0.0927, + "step": 71945 + }, + { + "epoch": 2.77, + "learning_rate": 3.9010002564760195e-06, + "loss": 0.2617, + "step": 71946 + }, + { + "epoch": 2.77, + "learning_rate": 3.9003590664272895e-06, + "loss": 0.191, + "step": 71947 + }, + { + "epoch": 2.77, + "learning_rate": 3.899717876378559e-06, + "loss": 0.1192, + "step": 71948 + }, + { + "epoch": 2.77, + "learning_rate": 3.899076686329829e-06, + "loss": 0.1373, + "step": 71949 + }, + { + "epoch": 2.77, + "learning_rate": 3.898435496281098e-06, + "loss": 0.1654, + "step": 71950 + }, + { + "epoch": 2.77, + "learning_rate": 3.897794306232368e-06, + "loss": 0.1276, + "step": 71951 + }, + { + "epoch": 2.77, + "learning_rate": 3.897153116183637e-06, + "loss": 0.2361, + "step": 71952 + }, + { + "epoch": 2.77, + "learning_rate": 3.896511926134907e-06, + "loss": 0.081, + "step": 71953 + }, + { + "epoch": 2.77, + "learning_rate": 3.895870736086176e-06, + "loss": 0.2072, + "step": 71954 + }, + { + "epoch": 2.77, + "learning_rate": 3.8952295460374454e-06, + "loss": 0.1953, + "step": 71955 + }, + { + "epoch": 2.77, + "learning_rate": 3.8945883559887155e-06, + "loss": 0.2241, + "step": 71956 + }, + { + "epoch": 2.77, + "learning_rate": 3.893947165939985e-06, + "loss": 0.2623, + "step": 71957 + }, + { + "epoch": 2.77, + "learning_rate": 3.893305975891255e-06, + "loss": 0.1689, + "step": 71958 + }, + { + "epoch": 2.77, + "learning_rate": 3.892664785842524e-06, + "loss": 0.1087, + "step": 71959 + }, + { + "epoch": 2.77, + "learning_rate": 3.892023595793794e-06, + "loss": 0.2651, + "step": 71960 + }, + { + "epoch": 2.77, + "learning_rate": 3.891382405745063e-06, + "loss": 0.2031, + "step": 71961 + }, + { + "epoch": 2.77, + "learning_rate": 3.890741215696333e-06, + "loss": 0.2547, + "step": 71962 + }, + { + "epoch": 2.77, + "learning_rate": 3.890100025647602e-06, + "loss": 0.1485, + "step": 71963 + }, + { + "epoch": 2.77, + "learning_rate": 3.889458835598872e-06, + "loss": 0.1979, + "step": 71964 + }, + { + "epoch": 2.77, + "learning_rate": 3.8888176455501414e-06, + "loss": 0.2414, + "step": 71965 + }, + { + "epoch": 2.77, + "learning_rate": 3.888176455501411e-06, + "loss": 0.2681, + "step": 71966 + }, + { + "epoch": 2.77, + "learning_rate": 3.887535265452681e-06, + "loss": 0.1478, + "step": 71967 + }, + { + "epoch": 2.77, + "learning_rate": 3.88689407540395e-06, + "loss": 0.2194, + "step": 71968 + }, + { + "epoch": 2.77, + "learning_rate": 3.88625288535522e-06, + "loss": 0.2066, + "step": 71969 + }, + { + "epoch": 2.77, + "learning_rate": 3.885611695306489e-06, + "loss": 0.1946, + "step": 71970 + }, + { + "epoch": 2.77, + "learning_rate": 3.884970505257759e-06, + "loss": 0.2274, + "step": 71971 + }, + { + "epoch": 2.77, + "learning_rate": 3.884329315209028e-06, + "loss": 0.171, + "step": 71972 + }, + { + "epoch": 2.77, + "learning_rate": 3.883688125160298e-06, + "loss": 0.0633, + "step": 71973 + }, + { + "epoch": 2.77, + "learning_rate": 3.883046935111567e-06, + "loss": 0.2636, + "step": 71974 + }, + { + "epoch": 2.77, + "learning_rate": 3.8824057450628366e-06, + "loss": 0.255, + "step": 71975 + }, + { + "epoch": 2.77, + "learning_rate": 3.881764555014107e-06, + "loss": 0.122, + "step": 71976 + }, + { + "epoch": 2.77, + "learning_rate": 3.881123364965376e-06, + "loss": 0.0952, + "step": 71977 + }, + { + "epoch": 2.77, + "learning_rate": 3.880482174916646e-06, + "loss": 0.2753, + "step": 71978 + }, + { + "epoch": 2.77, + "learning_rate": 3.879840984867915e-06, + "loss": 0.2313, + "step": 71979 + }, + { + "epoch": 2.77, + "learning_rate": 3.879199794819185e-06, + "loss": 0.1468, + "step": 71980 + }, + { + "epoch": 2.77, + "learning_rate": 3.878558604770454e-06, + "loss": 0.3069, + "step": 71981 + }, + { + "epoch": 2.77, + "learning_rate": 3.877917414721724e-06, + "loss": 0.2713, + "step": 71982 + }, + { + "epoch": 2.77, + "learning_rate": 3.877276224672993e-06, + "loss": 0.16, + "step": 71983 + }, + { + "epoch": 2.77, + "learning_rate": 3.8766350346242625e-06, + "loss": 0.2824, + "step": 71984 + }, + { + "epoch": 2.77, + "learning_rate": 3.8759938445755326e-06, + "loss": 0.3469, + "step": 71985 + }, + { + "epoch": 2.77, + "learning_rate": 3.875352654526802e-06, + "loss": 0.2117, + "step": 71986 + }, + { + "epoch": 2.77, + "learning_rate": 3.874711464478072e-06, + "loss": 0.1955, + "step": 71987 + }, + { + "epoch": 2.77, + "learning_rate": 3.874070274429341e-06, + "loss": 0.1088, + "step": 71988 + }, + { + "epoch": 2.77, + "learning_rate": 3.873429084380611e-06, + "loss": 0.1673, + "step": 71989 + }, + { + "epoch": 2.77, + "learning_rate": 3.87278789433188e-06, + "loss": 0.1971, + "step": 71990 + }, + { + "epoch": 2.77, + "learning_rate": 3.87214670428315e-06, + "loss": 0.1869, + "step": 71991 + }, + { + "epoch": 2.77, + "learning_rate": 3.871505514234419e-06, + "loss": 0.3253, + "step": 71992 + }, + { + "epoch": 2.77, + "learning_rate": 3.870864324185689e-06, + "loss": 0.2084, + "step": 71993 + }, + { + "epoch": 2.77, + "learning_rate": 3.8702231341369585e-06, + "loss": 0.1556, + "step": 71994 + }, + { + "epoch": 2.77, + "learning_rate": 3.869581944088228e-06, + "loss": 0.321, + "step": 71995 + }, + { + "epoch": 2.77, + "learning_rate": 3.868940754039498e-06, + "loss": 0.2297, + "step": 71996 + }, + { + "epoch": 2.77, + "learning_rate": 3.868299563990767e-06, + "loss": 0.1415, + "step": 71997 + }, + { + "epoch": 2.77, + "learning_rate": 3.867658373942037e-06, + "loss": 0.1832, + "step": 71998 + }, + { + "epoch": 2.77, + "learning_rate": 3.867017183893306e-06, + "loss": 0.3556, + "step": 71999 + }, + { + "epoch": 2.77, + "learning_rate": 3.866375993844576e-06, + "loss": 0.3058, + "step": 72000 + }, + { + "epoch": 2.77, + "learning_rate": 3.865734803795845e-06, + "loss": 0.3286, + "step": 72001 + }, + { + "epoch": 2.77, + "learning_rate": 3.865093613747115e-06, + "loss": 0.1029, + "step": 72002 + }, + { + "epoch": 2.77, + "learning_rate": 3.8644524236983845e-06, + "loss": 0.179, + "step": 72003 + }, + { + "epoch": 2.77, + "learning_rate": 3.863811233649654e-06, + "loss": 0.1686, + "step": 72004 + }, + { + "epoch": 2.77, + "learning_rate": 3.863170043600924e-06, + "loss": 0.2486, + "step": 72005 + }, + { + "epoch": 2.77, + "learning_rate": 3.862528853552193e-06, + "loss": 0.1475, + "step": 72006 + }, + { + "epoch": 2.77, + "learning_rate": 3.861887663503463e-06, + "loss": 0.3157, + "step": 72007 + }, + { + "epoch": 2.77, + "learning_rate": 3.861246473454732e-06, + "loss": 0.2084, + "step": 72008 + }, + { + "epoch": 2.77, + "learning_rate": 3.860605283406002e-06, + "loss": 0.2587, + "step": 72009 + }, + { + "epoch": 2.77, + "learning_rate": 3.859964093357271e-06, + "loss": 0.2576, + "step": 72010 + }, + { + "epoch": 2.77, + "learning_rate": 3.859322903308541e-06, + "loss": 0.152, + "step": 72011 + }, + { + "epoch": 2.77, + "learning_rate": 3.8586817132598105e-06, + "loss": 0.3916, + "step": 72012 + }, + { + "epoch": 2.77, + "learning_rate": 3.8580405232110805e-06, + "loss": 0.1618, + "step": 72013 + }, + { + "epoch": 2.77, + "learning_rate": 3.85739933316235e-06, + "loss": 0.2619, + "step": 72014 + }, + { + "epoch": 2.77, + "learning_rate": 3.856758143113619e-06, + "loss": 0.1343, + "step": 72015 + }, + { + "epoch": 2.77, + "learning_rate": 3.856116953064889e-06, + "loss": 0.2311, + "step": 72016 + }, + { + "epoch": 2.77, + "learning_rate": 3.855475763016158e-06, + "loss": 0.1295, + "step": 72017 + }, + { + "epoch": 2.77, + "learning_rate": 3.854834572967428e-06, + "loss": 0.1323, + "step": 72018 + }, + { + "epoch": 2.77, + "learning_rate": 3.854193382918697e-06, + "loss": 0.2259, + "step": 72019 + }, + { + "epoch": 2.77, + "learning_rate": 3.853552192869967e-06, + "loss": 0.1793, + "step": 72020 + }, + { + "epoch": 2.77, + "learning_rate": 3.8529110028212364e-06, + "loss": 0.1518, + "step": 72021 + }, + { + "epoch": 2.77, + "learning_rate": 3.8522698127725065e-06, + "loss": 0.2425, + "step": 72022 + }, + { + "epoch": 2.77, + "learning_rate": 3.851628622723776e-06, + "loss": 0.2157, + "step": 72023 + }, + { + "epoch": 2.77, + "learning_rate": 3.850987432675045e-06, + "loss": 0.2929, + "step": 72024 + }, + { + "epoch": 2.77, + "learning_rate": 3.850346242626315e-06, + "loss": 0.2718, + "step": 72025 + }, + { + "epoch": 2.77, + "learning_rate": 3.849705052577584e-06, + "loss": 0.2406, + "step": 72026 + }, + { + "epoch": 2.77, + "learning_rate": 3.849063862528854e-06, + "loss": 0.1931, + "step": 72027 + }, + { + "epoch": 2.77, + "learning_rate": 3.848422672480123e-06, + "loss": 0.2431, + "step": 72028 + }, + { + "epoch": 2.77, + "learning_rate": 3.847781482431393e-06, + "loss": 0.1053, + "step": 72029 + }, + { + "epoch": 2.77, + "learning_rate": 3.847140292382662e-06, + "loss": 0.2016, + "step": 72030 + }, + { + "epoch": 2.77, + "learning_rate": 3.8464991023339324e-06, + "loss": 0.2187, + "step": 72031 + }, + { + "epoch": 2.77, + "learning_rate": 3.845857912285202e-06, + "loss": 0.2654, + "step": 72032 + }, + { + "epoch": 2.77, + "learning_rate": 3.845216722236472e-06, + "loss": 0.2355, + "step": 72033 + }, + { + "epoch": 2.77, + "learning_rate": 3.844575532187741e-06, + "loss": 0.2384, + "step": 72034 + }, + { + "epoch": 2.77, + "learning_rate": 3.84393434213901e-06, + "loss": 0.2127, + "step": 72035 + }, + { + "epoch": 2.77, + "learning_rate": 3.84329315209028e-06, + "loss": 0.2104, + "step": 72036 + }, + { + "epoch": 2.77, + "learning_rate": 3.842651962041549e-06, + "loss": 0.0971, + "step": 72037 + }, + { + "epoch": 2.77, + "learning_rate": 3.842010771992819e-06, + "loss": 0.225, + "step": 72038 + }, + { + "epoch": 2.77, + "learning_rate": 3.841369581944088e-06, + "loss": 0.2146, + "step": 72039 + }, + { + "epoch": 2.77, + "learning_rate": 3.840728391895358e-06, + "loss": 0.2293, + "step": 72040 + }, + { + "epoch": 2.77, + "learning_rate": 3.8400872018466276e-06, + "loss": 0.1456, + "step": 72041 + }, + { + "epoch": 2.77, + "learning_rate": 3.839446011797898e-06, + "loss": 0.1045, + "step": 72042 + }, + { + "epoch": 2.77, + "learning_rate": 3.838804821749167e-06, + "loss": 0.2702, + "step": 72043 + }, + { + "epoch": 2.77, + "learning_rate": 3.838163631700436e-06, + "loss": 0.1454, + "step": 72044 + }, + { + "epoch": 2.77, + "learning_rate": 3.837522441651706e-06, + "loss": 0.2409, + "step": 72045 + }, + { + "epoch": 2.77, + "learning_rate": 3.836881251602975e-06, + "loss": 0.2094, + "step": 72046 + }, + { + "epoch": 2.77, + "learning_rate": 3.836240061554245e-06, + "loss": 0.1068, + "step": 72047 + }, + { + "epoch": 2.77, + "learning_rate": 3.835598871505514e-06, + "loss": 0.1535, + "step": 72048 + }, + { + "epoch": 2.77, + "learning_rate": 3.834957681456784e-06, + "loss": 0.2425, + "step": 72049 + }, + { + "epoch": 2.77, + "learning_rate": 3.8343164914080535e-06, + "loss": 0.1809, + "step": 72050 + }, + { + "epoch": 2.77, + "learning_rate": 3.8336753013593236e-06, + "loss": 0.3062, + "step": 72051 + }, + { + "epoch": 2.77, + "learning_rate": 3.833034111310593e-06, + "loss": 0.2473, + "step": 72052 + }, + { + "epoch": 2.77, + "learning_rate": 3.832392921261862e-06, + "loss": 0.2108, + "step": 72053 + }, + { + "epoch": 2.77, + "learning_rate": 3.831751731213132e-06, + "loss": 0.2025, + "step": 72054 + }, + { + "epoch": 2.77, + "learning_rate": 3.831110541164401e-06, + "loss": 0.3168, + "step": 72055 + }, + { + "epoch": 2.77, + "learning_rate": 3.830469351115671e-06, + "loss": 0.2309, + "step": 72056 + }, + { + "epoch": 2.77, + "learning_rate": 3.82982816106694e-06, + "loss": 0.2896, + "step": 72057 + }, + { + "epoch": 2.77, + "learning_rate": 3.82918697101821e-06, + "loss": 0.341, + "step": 72058 + }, + { + "epoch": 2.77, + "learning_rate": 3.8285457809694795e-06, + "loss": 0.2776, + "step": 72059 + }, + { + "epoch": 2.77, + "learning_rate": 3.8279045909207495e-06, + "loss": 0.1204, + "step": 72060 + }, + { + "epoch": 2.77, + "learning_rate": 3.827263400872019e-06, + "loss": 0.216, + "step": 72061 + }, + { + "epoch": 2.77, + "learning_rate": 3.826622210823289e-06, + "loss": 0.1617, + "step": 72062 + }, + { + "epoch": 2.77, + "learning_rate": 3.825981020774558e-06, + "loss": 0.1838, + "step": 72063 + }, + { + "epoch": 2.77, + "learning_rate": 3.825339830725827e-06, + "loss": 0.2069, + "step": 72064 + }, + { + "epoch": 2.77, + "learning_rate": 3.824698640677097e-06, + "loss": 0.1973, + "step": 72065 + }, + { + "epoch": 2.77, + "learning_rate": 3.824057450628366e-06, + "loss": 0.1665, + "step": 72066 + }, + { + "epoch": 2.77, + "learning_rate": 3.823416260579636e-06, + "loss": 0.3261, + "step": 72067 + }, + { + "epoch": 2.77, + "learning_rate": 3.8227750705309055e-06, + "loss": 0.2448, + "step": 72068 + }, + { + "epoch": 2.77, + "learning_rate": 3.8221338804821755e-06, + "loss": 0.16, + "step": 72069 + }, + { + "epoch": 2.77, + "learning_rate": 3.821492690433445e-06, + "loss": 0.2175, + "step": 72070 + }, + { + "epoch": 2.77, + "learning_rate": 3.820851500384715e-06, + "loss": 0.1918, + "step": 72071 + }, + { + "epoch": 2.77, + "learning_rate": 3.820210310335984e-06, + "loss": 0.1877, + "step": 72072 + }, + { + "epoch": 2.77, + "learning_rate": 3.819569120287253e-06, + "loss": 0.2136, + "step": 72073 + }, + { + "epoch": 2.77, + "learning_rate": 3.818927930238523e-06, + "loss": 0.1892, + "step": 72074 + }, + { + "epoch": 2.77, + "learning_rate": 3.818286740189792e-06, + "loss": 0.1953, + "step": 72075 + }, + { + "epoch": 2.77, + "learning_rate": 3.817645550141062e-06, + "loss": 0.2657, + "step": 72076 + }, + { + "epoch": 2.77, + "learning_rate": 3.8170043600923314e-06, + "loss": 0.2461, + "step": 72077 + }, + { + "epoch": 2.77, + "learning_rate": 3.8163631700436015e-06, + "loss": 0.2038, + "step": 72078 + }, + { + "epoch": 2.77, + "learning_rate": 3.815721979994871e-06, + "loss": 0.2314, + "step": 72079 + }, + { + "epoch": 2.77, + "learning_rate": 3.815080789946141e-06, + "loss": 0.1715, + "step": 72080 + }, + { + "epoch": 2.77, + "learning_rate": 3.81443959989741e-06, + "loss": 0.2343, + "step": 72081 + }, + { + "epoch": 2.77, + "learning_rate": 3.81379840984868e-06, + "loss": 0.2006, + "step": 72082 + }, + { + "epoch": 2.77, + "learning_rate": 3.813157219799949e-06, + "loss": 0.1672, + "step": 72083 + }, + { + "epoch": 2.77, + "learning_rate": 3.8125160297512182e-06, + "loss": 0.1941, + "step": 72084 + }, + { + "epoch": 2.77, + "learning_rate": 3.8118748397024882e-06, + "loss": 0.2377, + "step": 72085 + }, + { + "epoch": 2.77, + "learning_rate": 3.8112336496537574e-06, + "loss": 0.1558, + "step": 72086 + }, + { + "epoch": 2.77, + "learning_rate": 3.8105924596050274e-06, + "loss": 0.1583, + "step": 72087 + }, + { + "epoch": 2.77, + "learning_rate": 3.8099512695562966e-06, + "loss": 0.305, + "step": 72088 + }, + { + "epoch": 2.77, + "learning_rate": 3.8093100795075666e-06, + "loss": 0.246, + "step": 72089 + }, + { + "epoch": 2.77, + "learning_rate": 3.808668889458836e-06, + "loss": 0.2002, + "step": 72090 + }, + { + "epoch": 2.77, + "learning_rate": 3.808027699410106e-06, + "loss": 0.1315, + "step": 72091 + }, + { + "epoch": 2.77, + "learning_rate": 3.807386509361375e-06, + "loss": 0.1084, + "step": 72092 + }, + { + "epoch": 2.77, + "learning_rate": 3.806745319312644e-06, + "loss": 0.2796, + "step": 72093 + }, + { + "epoch": 2.77, + "learning_rate": 3.806104129263914e-06, + "loss": 0.1997, + "step": 72094 + }, + { + "epoch": 2.77, + "learning_rate": 3.8054629392151834e-06, + "loss": 0.1408, + "step": 72095 + }, + { + "epoch": 2.77, + "learning_rate": 3.8048217491664534e-06, + "loss": 0.1945, + "step": 72096 + }, + { + "epoch": 2.77, + "learning_rate": 3.8041805591177226e-06, + "loss": 0.2296, + "step": 72097 + }, + { + "epoch": 2.77, + "learning_rate": 3.8035393690689926e-06, + "loss": 0.1588, + "step": 72098 + }, + { + "epoch": 2.77, + "learning_rate": 3.8028981790202618e-06, + "loss": 0.1776, + "step": 72099 + }, + { + "epoch": 2.77, + "learning_rate": 3.802256988971532e-06, + "loss": 0.3159, + "step": 72100 + }, + { + "epoch": 2.77, + "learning_rate": 3.801615798922801e-06, + "loss": 0.2256, + "step": 72101 + }, + { + "epoch": 2.77, + "learning_rate": 3.80097460887407e-06, + "loss": 0.1316, + "step": 72102 + }, + { + "epoch": 2.77, + "learning_rate": 3.80033341882534e-06, + "loss": 0.2739, + "step": 72103 + }, + { + "epoch": 2.77, + "learning_rate": 3.7996922287766093e-06, + "loss": 0.1461, + "step": 72104 + }, + { + "epoch": 2.77, + "learning_rate": 3.7990510387278794e-06, + "loss": 0.315, + "step": 72105 + }, + { + "epoch": 2.77, + "learning_rate": 3.7984098486791485e-06, + "loss": 0.2316, + "step": 72106 + }, + { + "epoch": 2.77, + "learning_rate": 3.7977686586304186e-06, + "loss": 0.3178, + "step": 72107 + }, + { + "epoch": 2.77, + "learning_rate": 3.7971274685816877e-06, + "loss": 0.1361, + "step": 72108 + }, + { + "epoch": 2.77, + "learning_rate": 3.7964862785329578e-06, + "loss": 0.1703, + "step": 72109 + }, + { + "epoch": 2.77, + "learning_rate": 3.795845088484227e-06, + "loss": 0.1979, + "step": 72110 + }, + { + "epoch": 2.77, + "learning_rate": 3.795203898435497e-06, + "loss": 0.1891, + "step": 72111 + }, + { + "epoch": 2.77, + "learning_rate": 3.794562708386766e-06, + "loss": 0.1807, + "step": 72112 + }, + { + "epoch": 2.77, + "learning_rate": 3.7939215183380353e-06, + "loss": 0.1823, + "step": 72113 + }, + { + "epoch": 2.77, + "learning_rate": 3.7932803282893053e-06, + "loss": 0.1957, + "step": 72114 + }, + { + "epoch": 2.77, + "learning_rate": 3.7926391382405745e-06, + "loss": 0.1477, + "step": 72115 + }, + { + "epoch": 2.77, + "learning_rate": 3.7919979481918445e-06, + "loss": 0.1183, + "step": 72116 + }, + { + "epoch": 2.77, + "learning_rate": 3.7913567581431137e-06, + "loss": 0.1964, + "step": 72117 + }, + { + "epoch": 2.77, + "learning_rate": 3.7907155680943837e-06, + "loss": 0.2262, + "step": 72118 + }, + { + "epoch": 2.77, + "learning_rate": 3.790074378045653e-06, + "loss": 0.1502, + "step": 72119 + }, + { + "epoch": 2.77, + "learning_rate": 3.789433187996923e-06, + "loss": 0.1995, + "step": 72120 + }, + { + "epoch": 2.77, + "learning_rate": 3.788791997948192e-06, + "loss": 0.158, + "step": 72121 + }, + { + "epoch": 2.77, + "learning_rate": 3.7881508078994613e-06, + "loss": 0.1895, + "step": 72122 + }, + { + "epoch": 2.77, + "learning_rate": 3.7875096178507313e-06, + "loss": 0.1609, + "step": 72123 + }, + { + "epoch": 2.77, + "learning_rate": 3.7868684278020005e-06, + "loss": 0.2579, + "step": 72124 + }, + { + "epoch": 2.77, + "learning_rate": 3.7862272377532705e-06, + "loss": 0.2766, + "step": 72125 + }, + { + "epoch": 2.77, + "learning_rate": 3.7855860477045397e-06, + "loss": 0.1975, + "step": 72126 + }, + { + "epoch": 2.77, + "learning_rate": 3.7849448576558097e-06, + "loss": 0.1538, + "step": 72127 + }, + { + "epoch": 2.77, + "learning_rate": 3.784303667607079e-06, + "loss": 0.2564, + "step": 72128 + }, + { + "epoch": 2.77, + "learning_rate": 3.783662477558349e-06, + "loss": 0.2668, + "step": 72129 + }, + { + "epoch": 2.77, + "learning_rate": 3.783021287509618e-06, + "loss": 0.1126, + "step": 72130 + }, + { + "epoch": 2.77, + "learning_rate": 3.782380097460888e-06, + "loss": 0.1023, + "step": 72131 + }, + { + "epoch": 2.77, + "learning_rate": 3.7817389074121573e-06, + "loss": 0.0777, + "step": 72132 + }, + { + "epoch": 2.77, + "learning_rate": 3.7810977173634264e-06, + "loss": 0.3056, + "step": 72133 + }, + { + "epoch": 2.77, + "learning_rate": 3.7804565273146965e-06, + "loss": 0.2033, + "step": 72134 + }, + { + "epoch": 2.77, + "learning_rate": 3.7798153372659656e-06, + "loss": 0.2304, + "step": 72135 + }, + { + "epoch": 2.77, + "learning_rate": 3.7791741472172357e-06, + "loss": 0.1143, + "step": 72136 + }, + { + "epoch": 2.77, + "learning_rate": 3.778532957168505e-06, + "loss": 0.1749, + "step": 72137 + }, + { + "epoch": 2.77, + "learning_rate": 3.777891767119775e-06, + "loss": 0.1589, + "step": 72138 + }, + { + "epoch": 2.77, + "learning_rate": 3.777250577071044e-06, + "loss": 0.3134, + "step": 72139 + }, + { + "epoch": 2.77, + "learning_rate": 3.776609387022314e-06, + "loss": 0.15, + "step": 72140 + }, + { + "epoch": 2.77, + "learning_rate": 3.7759681969735832e-06, + "loss": 0.2032, + "step": 72141 + }, + { + "epoch": 2.77, + "learning_rate": 3.7753270069248524e-06, + "loss": 0.1347, + "step": 72142 + }, + { + "epoch": 2.77, + "learning_rate": 3.7746858168761224e-06, + "loss": 0.168, + "step": 72143 + }, + { + "epoch": 2.77, + "learning_rate": 3.7740446268273916e-06, + "loss": 0.1573, + "step": 72144 + }, + { + "epoch": 2.77, + "learning_rate": 3.7734034367786616e-06, + "loss": 0.2461, + "step": 72145 + }, + { + "epoch": 2.77, + "learning_rate": 3.772762246729931e-06, + "loss": 0.168, + "step": 72146 + }, + { + "epoch": 2.77, + "learning_rate": 3.772121056681201e-06, + "loss": 0.2374, + "step": 72147 + }, + { + "epoch": 2.77, + "learning_rate": 3.77147986663247e-06, + "loss": 0.1193, + "step": 72148 + }, + { + "epoch": 2.77, + "learning_rate": 3.77083867658374e-06, + "loss": 0.2877, + "step": 72149 + }, + { + "epoch": 2.77, + "learning_rate": 3.770197486535009e-06, + "loss": 0.2434, + "step": 72150 + }, + { + "epoch": 2.77, + "learning_rate": 3.7695562964862792e-06, + "loss": 0.1509, + "step": 72151 + }, + { + "epoch": 2.77, + "learning_rate": 3.7689151064375484e-06, + "loss": 0.1851, + "step": 72152 + }, + { + "epoch": 2.77, + "learning_rate": 3.7682739163888176e-06, + "loss": 0.1643, + "step": 72153 + }, + { + "epoch": 2.77, + "learning_rate": 3.7676327263400876e-06, + "loss": 0.1945, + "step": 72154 + }, + { + "epoch": 2.77, + "learning_rate": 3.7669915362913568e-06, + "loss": 0.1511, + "step": 72155 + }, + { + "epoch": 2.77, + "learning_rate": 3.766350346242627e-06, + "loss": 0.1366, + "step": 72156 + }, + { + "epoch": 2.77, + "learning_rate": 3.765709156193896e-06, + "loss": 0.1659, + "step": 72157 + }, + { + "epoch": 2.77, + "learning_rate": 3.765067966145166e-06, + "loss": 0.1995, + "step": 72158 + }, + { + "epoch": 2.77, + "learning_rate": 3.764426776096435e-06, + "loss": 0.1119, + "step": 72159 + }, + { + "epoch": 2.77, + "learning_rate": 3.763785586047705e-06, + "loss": 0.3484, + "step": 72160 + }, + { + "epoch": 2.77, + "learning_rate": 3.7631443959989744e-06, + "loss": 0.2198, + "step": 72161 + }, + { + "epoch": 2.77, + "learning_rate": 3.7625032059502435e-06, + "loss": 0.1463, + "step": 72162 + }, + { + "epoch": 2.77, + "learning_rate": 3.7618620159015136e-06, + "loss": 0.1891, + "step": 72163 + }, + { + "epoch": 2.77, + "learning_rate": 3.7612208258527827e-06, + "loss": 0.1022, + "step": 72164 + }, + { + "epoch": 2.77, + "learning_rate": 3.7605796358040528e-06, + "loss": 0.306, + "step": 72165 + }, + { + "epoch": 2.77, + "learning_rate": 3.759938445755322e-06, + "loss": 0.1035, + "step": 72166 + }, + { + "epoch": 2.77, + "learning_rate": 3.759297255706592e-06, + "loss": 0.2732, + "step": 72167 + }, + { + "epoch": 2.77, + "learning_rate": 3.758656065657861e-06, + "loss": 0.076, + "step": 72168 + }, + { + "epoch": 2.77, + "learning_rate": 3.758014875609131e-06, + "loss": 0.2961, + "step": 72169 + }, + { + "epoch": 2.77, + "learning_rate": 3.7573736855604003e-06, + "loss": 0.2945, + "step": 72170 + }, + { + "epoch": 2.77, + "learning_rate": 3.7567324955116695e-06, + "loss": 0.1648, + "step": 72171 + }, + { + "epoch": 2.77, + "learning_rate": 3.7560913054629395e-06, + "loss": 0.1996, + "step": 72172 + }, + { + "epoch": 2.77, + "learning_rate": 3.7554501154142087e-06, + "loss": 0.1324, + "step": 72173 + }, + { + "epoch": 2.77, + "learning_rate": 3.7548089253654787e-06, + "loss": 0.1484, + "step": 72174 + }, + { + "epoch": 2.77, + "learning_rate": 3.754167735316748e-06, + "loss": 0.2286, + "step": 72175 + }, + { + "epoch": 2.77, + "learning_rate": 3.753526545268018e-06, + "loss": 0.2505, + "step": 72176 + }, + { + "epoch": 2.77, + "learning_rate": 3.752885355219287e-06, + "loss": 0.1556, + "step": 72177 + }, + { + "epoch": 2.78, + "learning_rate": 3.752244165170557e-06, + "loss": 0.1916, + "step": 72178 + }, + { + "epoch": 2.78, + "learning_rate": 3.7516029751218263e-06, + "loss": 0.2074, + "step": 72179 + }, + { + "epoch": 2.78, + "learning_rate": 3.7509617850730963e-06, + "loss": 0.1611, + "step": 72180 + }, + { + "epoch": 2.78, + "learning_rate": 3.7503205950243655e-06, + "loss": 0.2653, + "step": 72181 + }, + { + "epoch": 2.78, + "learning_rate": 3.7496794049756347e-06, + "loss": 0.1817, + "step": 72182 + }, + { + "epoch": 2.78, + "learning_rate": 3.7490382149269047e-06, + "loss": 0.157, + "step": 72183 + }, + { + "epoch": 2.78, + "learning_rate": 3.748397024878174e-06, + "loss": 0.2816, + "step": 72184 + }, + { + "epoch": 2.78, + "learning_rate": 3.747755834829444e-06, + "loss": 0.1934, + "step": 72185 + }, + { + "epoch": 2.78, + "learning_rate": 3.747114644780713e-06, + "loss": 0.2287, + "step": 72186 + }, + { + "epoch": 2.78, + "learning_rate": 3.746473454731983e-06, + "loss": 0.2596, + "step": 72187 + }, + { + "epoch": 2.78, + "learning_rate": 3.7458322646832523e-06, + "loss": 0.1551, + "step": 72188 + }, + { + "epoch": 2.78, + "learning_rate": 3.7451910746345223e-06, + "loss": 0.2385, + "step": 72189 + }, + { + "epoch": 2.78, + "learning_rate": 3.7445498845857915e-06, + "loss": 0.1481, + "step": 72190 + }, + { + "epoch": 2.78, + "learning_rate": 3.7439086945370607e-06, + "loss": 0.217, + "step": 72191 + }, + { + "epoch": 2.78, + "learning_rate": 3.7432675044883307e-06, + "loss": 0.1321, + "step": 72192 + }, + { + "epoch": 2.78, + "learning_rate": 3.7426263144396e-06, + "loss": 0.2515, + "step": 72193 + }, + { + "epoch": 2.78, + "learning_rate": 3.74198512439087e-06, + "loss": 0.1855, + "step": 72194 + }, + { + "epoch": 2.78, + "learning_rate": 3.741343934342139e-06, + "loss": 0.2206, + "step": 72195 + }, + { + "epoch": 2.78, + "learning_rate": 3.740702744293409e-06, + "loss": 0.1993, + "step": 72196 + }, + { + "epoch": 2.78, + "learning_rate": 3.7400615542446782e-06, + "loss": 0.1292, + "step": 72197 + }, + { + "epoch": 2.78, + "learning_rate": 3.7394203641959483e-06, + "loss": 0.0605, + "step": 72198 + }, + { + "epoch": 2.78, + "learning_rate": 3.7387791741472174e-06, + "loss": 0.1589, + "step": 72199 + }, + { + "epoch": 2.78, + "learning_rate": 3.7381379840984875e-06, + "loss": 0.2212, + "step": 72200 + }, + { + "epoch": 2.78, + "learning_rate": 3.7374967940497566e-06, + "loss": 0.2268, + "step": 72201 + }, + { + "epoch": 2.78, + "learning_rate": 3.736855604001026e-06, + "loss": 0.1799, + "step": 72202 + }, + { + "epoch": 2.78, + "learning_rate": 3.736214413952296e-06, + "loss": 0.3304, + "step": 72203 + }, + { + "epoch": 2.78, + "learning_rate": 3.735573223903565e-06, + "loss": 0.1388, + "step": 72204 + }, + { + "epoch": 2.78, + "learning_rate": 3.734932033854835e-06, + "loss": 0.1896, + "step": 72205 + }, + { + "epoch": 2.78, + "learning_rate": 3.7342908438061042e-06, + "loss": 0.1544, + "step": 72206 + }, + { + "epoch": 2.78, + "learning_rate": 3.7336496537573742e-06, + "loss": 0.2011, + "step": 72207 + }, + { + "epoch": 2.78, + "learning_rate": 3.7330084637086434e-06, + "loss": 0.2402, + "step": 72208 + }, + { + "epoch": 2.78, + "learning_rate": 3.7323672736599134e-06, + "loss": 0.2952, + "step": 72209 + }, + { + "epoch": 2.78, + "learning_rate": 3.7317260836111826e-06, + "loss": 0.1503, + "step": 72210 + }, + { + "epoch": 2.78, + "learning_rate": 3.7310848935624518e-06, + "loss": 0.1528, + "step": 72211 + }, + { + "epoch": 2.78, + "learning_rate": 3.730443703513722e-06, + "loss": 0.1664, + "step": 72212 + }, + { + "epoch": 2.78, + "learning_rate": 3.729802513464991e-06, + "loss": 0.2916, + "step": 72213 + }, + { + "epoch": 2.78, + "learning_rate": 3.729161323416261e-06, + "loss": 0.184, + "step": 72214 + }, + { + "epoch": 2.78, + "learning_rate": 3.72852013336753e-06, + "loss": 0.1895, + "step": 72215 + }, + { + "epoch": 2.78, + "learning_rate": 3.7278789433188e-06, + "loss": 0.1469, + "step": 72216 + }, + { + "epoch": 2.78, + "learning_rate": 3.7272377532700694e-06, + "loss": 0.2347, + "step": 72217 + }, + { + "epoch": 2.78, + "learning_rate": 3.7265965632213394e-06, + "loss": 0.2512, + "step": 72218 + }, + { + "epoch": 2.78, + "learning_rate": 3.7259553731726086e-06, + "loss": 0.2308, + "step": 72219 + }, + { + "epoch": 2.78, + "learning_rate": 3.7253141831238778e-06, + "loss": 0.1763, + "step": 72220 + }, + { + "epoch": 2.78, + "learning_rate": 3.7246729930751478e-06, + "loss": 0.2244, + "step": 72221 + }, + { + "epoch": 2.78, + "learning_rate": 3.724031803026417e-06, + "loss": 0.1272, + "step": 72222 + }, + { + "epoch": 2.78, + "learning_rate": 3.723390612977687e-06, + "loss": 0.046, + "step": 72223 + }, + { + "epoch": 2.78, + "learning_rate": 3.722749422928956e-06, + "loss": 0.1674, + "step": 72224 + }, + { + "epoch": 2.78, + "learning_rate": 3.722108232880226e-06, + "loss": 0.1977, + "step": 72225 + }, + { + "epoch": 2.78, + "learning_rate": 3.7214670428314953e-06, + "loss": 0.1919, + "step": 72226 + }, + { + "epoch": 2.78, + "learning_rate": 3.7208258527827654e-06, + "loss": 0.2861, + "step": 72227 + }, + { + "epoch": 2.78, + "learning_rate": 3.7201846627340345e-06, + "loss": 0.1245, + "step": 72228 + }, + { + "epoch": 2.78, + "learning_rate": 3.7195434726853046e-06, + "loss": 0.2717, + "step": 72229 + }, + { + "epoch": 2.78, + "learning_rate": 3.7189022826365737e-06, + "loss": 0.1446, + "step": 72230 + }, + { + "epoch": 2.78, + "learning_rate": 3.718261092587843e-06, + "loss": 0.1973, + "step": 72231 + }, + { + "epoch": 2.78, + "learning_rate": 3.717619902539113e-06, + "loss": 0.1676, + "step": 72232 + }, + { + "epoch": 2.78, + "learning_rate": 3.716978712490382e-06, + "loss": 0.2223, + "step": 72233 + }, + { + "epoch": 2.78, + "learning_rate": 3.716337522441652e-06, + "loss": 0.1511, + "step": 72234 + }, + { + "epoch": 2.78, + "learning_rate": 3.7156963323929213e-06, + "loss": 0.1893, + "step": 72235 + }, + { + "epoch": 2.78, + "learning_rate": 3.7150551423441913e-06, + "loss": 0.2789, + "step": 72236 + }, + { + "epoch": 2.78, + "learning_rate": 3.7144139522954605e-06, + "loss": 0.2059, + "step": 72237 + }, + { + "epoch": 2.78, + "learning_rate": 3.7137727622467305e-06, + "loss": 0.313, + "step": 72238 + }, + { + "epoch": 2.78, + "learning_rate": 3.7131315721979997e-06, + "loss": 0.2632, + "step": 72239 + }, + { + "epoch": 2.78, + "learning_rate": 3.712490382149269e-06, + "loss": 0.2221, + "step": 72240 + }, + { + "epoch": 2.78, + "learning_rate": 3.711849192100539e-06, + "loss": 0.2059, + "step": 72241 + }, + { + "epoch": 2.78, + "learning_rate": 3.711208002051808e-06, + "loss": 0.2261, + "step": 72242 + }, + { + "epoch": 2.78, + "learning_rate": 3.710566812003078e-06, + "loss": 0.3414, + "step": 72243 + }, + { + "epoch": 2.78, + "learning_rate": 3.7099256219543473e-06, + "loss": 0.1811, + "step": 72244 + }, + { + "epoch": 2.78, + "learning_rate": 3.7092844319056173e-06, + "loss": 0.3224, + "step": 72245 + }, + { + "epoch": 2.78, + "learning_rate": 3.7086432418568865e-06, + "loss": 0.1342, + "step": 72246 + }, + { + "epoch": 2.78, + "learning_rate": 3.7080020518081565e-06, + "loss": 0.2778, + "step": 72247 + }, + { + "epoch": 2.78, + "learning_rate": 3.7073608617594257e-06, + "loss": 0.2552, + "step": 72248 + }, + { + "epoch": 2.78, + "learning_rate": 3.7067196717106957e-06, + "loss": 0.1834, + "step": 72249 + }, + { + "epoch": 2.78, + "learning_rate": 3.706078481661965e-06, + "loss": 0.1709, + "step": 72250 + }, + { + "epoch": 2.78, + "learning_rate": 3.705437291613234e-06, + "loss": 0.2218, + "step": 72251 + }, + { + "epoch": 2.78, + "learning_rate": 3.704796101564504e-06, + "loss": 0.1648, + "step": 72252 + }, + { + "epoch": 2.78, + "learning_rate": 3.7041549115157733e-06, + "loss": 0.2297, + "step": 72253 + }, + { + "epoch": 2.78, + "learning_rate": 3.7035137214670433e-06, + "loss": 0.1596, + "step": 72254 + }, + { + "epoch": 2.78, + "learning_rate": 3.7028725314183124e-06, + "loss": 0.1733, + "step": 72255 + }, + { + "epoch": 2.78, + "learning_rate": 3.7022313413695825e-06, + "loss": 0.2404, + "step": 72256 + }, + { + "epoch": 2.78, + "learning_rate": 3.7015901513208516e-06, + "loss": 0.116, + "step": 72257 + }, + { + "epoch": 2.78, + "learning_rate": 3.7009489612721217e-06, + "loss": 0.1781, + "step": 72258 + }, + { + "epoch": 2.78, + "learning_rate": 3.700307771223391e-06, + "loss": 0.1872, + "step": 72259 + }, + { + "epoch": 2.78, + "learning_rate": 3.69966658117466e-06, + "loss": 0.3012, + "step": 72260 + }, + { + "epoch": 2.78, + "learning_rate": 3.69902539112593e-06, + "loss": 0.236, + "step": 72261 + }, + { + "epoch": 2.78, + "learning_rate": 3.6983842010771992e-06, + "loss": 0.205, + "step": 72262 + }, + { + "epoch": 2.78, + "learning_rate": 3.6977430110284692e-06, + "loss": 0.2205, + "step": 72263 + }, + { + "epoch": 2.78, + "learning_rate": 3.6971018209797384e-06, + "loss": 0.2298, + "step": 72264 + }, + { + "epoch": 2.78, + "learning_rate": 3.6964606309310084e-06, + "loss": 0.163, + "step": 72265 + }, + { + "epoch": 2.78, + "learning_rate": 3.6958194408822776e-06, + "loss": 0.1397, + "step": 72266 + }, + { + "epoch": 2.78, + "learning_rate": 3.6951782508335476e-06, + "loss": 0.1624, + "step": 72267 + }, + { + "epoch": 2.78, + "learning_rate": 3.694537060784817e-06, + "loss": 0.2504, + "step": 72268 + }, + { + "epoch": 2.78, + "learning_rate": 3.693895870736086e-06, + "loss": 0.2261, + "step": 72269 + }, + { + "epoch": 2.78, + "learning_rate": 3.693254680687356e-06, + "loss": 0.1146, + "step": 72270 + }, + { + "epoch": 2.78, + "learning_rate": 3.692613490638625e-06, + "loss": 0.1179, + "step": 72271 + }, + { + "epoch": 2.78, + "learning_rate": 3.691972300589895e-06, + "loss": 0.2061, + "step": 72272 + }, + { + "epoch": 2.78, + "learning_rate": 3.6913311105411644e-06, + "loss": 0.1012, + "step": 72273 + }, + { + "epoch": 2.78, + "learning_rate": 3.6906899204924344e-06, + "loss": 0.2934, + "step": 72274 + }, + { + "epoch": 2.78, + "learning_rate": 3.6900487304437036e-06, + "loss": 0.2337, + "step": 72275 + }, + { + "epoch": 2.78, + "learning_rate": 3.6894075403949736e-06, + "loss": 0.2168, + "step": 72276 + }, + { + "epoch": 2.78, + "learning_rate": 3.6887663503462428e-06, + "loss": 0.2187, + "step": 72277 + }, + { + "epoch": 2.78, + "learning_rate": 3.688125160297513e-06, + "loss": 0.1746, + "step": 72278 + }, + { + "epoch": 2.78, + "learning_rate": 3.687483970248782e-06, + "loss": 0.1602, + "step": 72279 + }, + { + "epoch": 2.78, + "learning_rate": 3.686842780200051e-06, + "loss": 0.1514, + "step": 72280 + }, + { + "epoch": 2.78, + "learning_rate": 3.686201590151321e-06, + "loss": 0.1379, + "step": 72281 + }, + { + "epoch": 2.78, + "learning_rate": 3.6855604001025904e-06, + "loss": 0.3145, + "step": 72282 + }, + { + "epoch": 2.78, + "learning_rate": 3.6849192100538604e-06, + "loss": 0.2117, + "step": 72283 + }, + { + "epoch": 2.78, + "learning_rate": 3.6842780200051295e-06, + "loss": 0.2508, + "step": 72284 + }, + { + "epoch": 2.78, + "learning_rate": 3.6836368299563996e-06, + "loss": 0.1759, + "step": 72285 + }, + { + "epoch": 2.78, + "learning_rate": 3.6829956399076687e-06, + "loss": 0.1527, + "step": 72286 + }, + { + "epoch": 2.78, + "learning_rate": 3.6823544498589388e-06, + "loss": 0.1396, + "step": 72287 + }, + { + "epoch": 2.78, + "learning_rate": 3.681713259810208e-06, + "loss": 0.162, + "step": 72288 + }, + { + "epoch": 2.78, + "learning_rate": 3.681072069761477e-06, + "loss": 0.1767, + "step": 72289 + }, + { + "epoch": 2.78, + "learning_rate": 3.680430879712747e-06, + "loss": 0.1785, + "step": 72290 + }, + { + "epoch": 2.78, + "learning_rate": 3.6797896896640163e-06, + "loss": 0.1642, + "step": 72291 + }, + { + "epoch": 2.78, + "learning_rate": 3.6791484996152863e-06, + "loss": 0.1193, + "step": 72292 + }, + { + "epoch": 2.78, + "learning_rate": 3.6785073095665555e-06, + "loss": 0.3671, + "step": 72293 + }, + { + "epoch": 2.78, + "learning_rate": 3.6778661195178255e-06, + "loss": 0.143, + "step": 72294 + }, + { + "epoch": 2.78, + "learning_rate": 3.6772249294690947e-06, + "loss": 0.2532, + "step": 72295 + }, + { + "epoch": 2.78, + "learning_rate": 3.6765837394203647e-06, + "loss": 0.2351, + "step": 72296 + }, + { + "epoch": 2.78, + "learning_rate": 3.675942549371634e-06, + "loss": 0.2338, + "step": 72297 + }, + { + "epoch": 2.78, + "learning_rate": 3.675301359322904e-06, + "loss": 0.2307, + "step": 72298 + }, + { + "epoch": 2.78, + "learning_rate": 3.674660169274173e-06, + "loss": 0.3436, + "step": 72299 + }, + { + "epoch": 2.78, + "learning_rate": 3.6740189792254423e-06, + "loss": 0.2224, + "step": 72300 + }, + { + "epoch": 2.78, + "learning_rate": 3.6733777891767123e-06, + "loss": 0.1588, + "step": 72301 + }, + { + "epoch": 2.78, + "learning_rate": 3.6727365991279815e-06, + "loss": 0.2546, + "step": 72302 + }, + { + "epoch": 2.78, + "learning_rate": 3.6720954090792515e-06, + "loss": 0.2454, + "step": 72303 + }, + { + "epoch": 2.78, + "learning_rate": 3.6714542190305207e-06, + "loss": 0.325, + "step": 72304 + }, + { + "epoch": 2.78, + "learning_rate": 3.6708130289817907e-06, + "loss": 0.2065, + "step": 72305 + }, + { + "epoch": 2.78, + "learning_rate": 3.67017183893306e-06, + "loss": 0.1757, + "step": 72306 + }, + { + "epoch": 2.78, + "learning_rate": 3.66953064888433e-06, + "loss": 0.2365, + "step": 72307 + }, + { + "epoch": 2.78, + "learning_rate": 3.668889458835599e-06, + "loss": 0.2447, + "step": 72308 + }, + { + "epoch": 2.78, + "learning_rate": 3.6682482687868683e-06, + "loss": 0.1332, + "step": 72309 + }, + { + "epoch": 2.78, + "learning_rate": 3.6676070787381383e-06, + "loss": 0.183, + "step": 72310 + }, + { + "epoch": 2.78, + "learning_rate": 3.6669658886894075e-06, + "loss": 0.236, + "step": 72311 + }, + { + "epoch": 2.78, + "learning_rate": 3.6663246986406775e-06, + "loss": 0.2085, + "step": 72312 + }, + { + "epoch": 2.78, + "learning_rate": 3.6656835085919467e-06, + "loss": 0.2556, + "step": 72313 + }, + { + "epoch": 2.78, + "learning_rate": 3.6650423185432167e-06, + "loss": 0.2339, + "step": 72314 + }, + { + "epoch": 2.78, + "learning_rate": 3.664401128494486e-06, + "loss": 0.3595, + "step": 72315 + }, + { + "epoch": 2.78, + "learning_rate": 3.663759938445756e-06, + "loss": 0.2355, + "step": 72316 + }, + { + "epoch": 2.78, + "learning_rate": 3.663118748397025e-06, + "loss": 0.23, + "step": 72317 + }, + { + "epoch": 2.78, + "learning_rate": 3.662477558348295e-06, + "loss": 0.0913, + "step": 72318 + }, + { + "epoch": 2.78, + "learning_rate": 3.6618363682995642e-06, + "loss": 0.2538, + "step": 72319 + }, + { + "epoch": 2.78, + "learning_rate": 3.6611951782508334e-06, + "loss": 0.1066, + "step": 72320 + }, + { + "epoch": 2.78, + "learning_rate": 3.6605539882021034e-06, + "loss": 0.3487, + "step": 72321 + }, + { + "epoch": 2.78, + "learning_rate": 3.6599127981533726e-06, + "loss": 0.1903, + "step": 72322 + }, + { + "epoch": 2.78, + "learning_rate": 3.6592716081046426e-06, + "loss": 0.1318, + "step": 72323 + }, + { + "epoch": 2.78, + "learning_rate": 3.658630418055912e-06, + "loss": 0.1863, + "step": 72324 + }, + { + "epoch": 2.78, + "learning_rate": 3.657989228007182e-06, + "loss": 0.1752, + "step": 72325 + }, + { + "epoch": 2.78, + "learning_rate": 3.657348037958451e-06, + "loss": 0.1642, + "step": 72326 + }, + { + "epoch": 2.78, + "learning_rate": 3.656706847909721e-06, + "loss": 0.1962, + "step": 72327 + }, + { + "epoch": 2.78, + "learning_rate": 3.6560656578609902e-06, + "loss": 0.2252, + "step": 72328 + }, + { + "epoch": 2.78, + "learning_rate": 3.6554244678122594e-06, + "loss": 0.2042, + "step": 72329 + }, + { + "epoch": 2.78, + "learning_rate": 3.6547832777635294e-06, + "loss": 0.214, + "step": 72330 + }, + { + "epoch": 2.78, + "learning_rate": 3.6541420877147986e-06, + "loss": 0.3095, + "step": 72331 + }, + { + "epoch": 2.78, + "learning_rate": 3.6535008976660686e-06, + "loss": 0.2038, + "step": 72332 + }, + { + "epoch": 2.78, + "learning_rate": 3.6528597076173378e-06, + "loss": 0.3343, + "step": 72333 + }, + { + "epoch": 2.78, + "learning_rate": 3.652218517568608e-06, + "loss": 0.1667, + "step": 72334 + }, + { + "epoch": 2.78, + "learning_rate": 3.651577327519877e-06, + "loss": 0.1876, + "step": 72335 + }, + { + "epoch": 2.78, + "learning_rate": 3.650936137471147e-06, + "loss": 0.3805, + "step": 72336 + }, + { + "epoch": 2.78, + "learning_rate": 3.650294947422416e-06, + "loss": 0.237, + "step": 72337 + }, + { + "epoch": 2.78, + "learning_rate": 3.6496537573736854e-06, + "loss": 0.2128, + "step": 72338 + }, + { + "epoch": 2.78, + "learning_rate": 3.6490125673249554e-06, + "loss": 0.1585, + "step": 72339 + }, + { + "epoch": 2.78, + "learning_rate": 3.6483713772762246e-06, + "loss": 0.2058, + "step": 72340 + }, + { + "epoch": 2.78, + "learning_rate": 3.6477301872274946e-06, + "loss": 0.1858, + "step": 72341 + }, + { + "epoch": 2.78, + "learning_rate": 3.6470889971787638e-06, + "loss": 0.2239, + "step": 72342 + }, + { + "epoch": 2.78, + "learning_rate": 3.6464478071300338e-06, + "loss": 0.2331, + "step": 72343 + }, + { + "epoch": 2.78, + "learning_rate": 3.645806617081303e-06, + "loss": 0.3176, + "step": 72344 + }, + { + "epoch": 2.78, + "learning_rate": 3.645165427032573e-06, + "loss": 0.2207, + "step": 72345 + }, + { + "epoch": 2.78, + "learning_rate": 3.644524236983842e-06, + "loss": 0.2789, + "step": 72346 + }, + { + "epoch": 2.78, + "learning_rate": 3.643883046935112e-06, + "loss": 0.1689, + "step": 72347 + }, + { + "epoch": 2.78, + "learning_rate": 3.6432418568863813e-06, + "loss": 0.0782, + "step": 72348 + }, + { + "epoch": 2.78, + "learning_rate": 3.6426006668376505e-06, + "loss": 0.0979, + "step": 72349 + }, + { + "epoch": 2.78, + "learning_rate": 3.6419594767889205e-06, + "loss": 0.3292, + "step": 72350 + }, + { + "epoch": 2.78, + "learning_rate": 3.6413182867401897e-06, + "loss": 0.2216, + "step": 72351 + }, + { + "epoch": 2.78, + "learning_rate": 3.6406770966914597e-06, + "loss": 0.2703, + "step": 72352 + }, + { + "epoch": 2.78, + "learning_rate": 3.640035906642729e-06, + "loss": 0.399, + "step": 72353 + }, + { + "epoch": 2.78, + "learning_rate": 3.639394716593999e-06, + "loss": 0.3179, + "step": 72354 + }, + { + "epoch": 2.78, + "learning_rate": 3.638753526545268e-06, + "loss": 0.1633, + "step": 72355 + }, + { + "epoch": 2.78, + "learning_rate": 3.638112336496538e-06, + "loss": 0.0485, + "step": 72356 + }, + { + "epoch": 2.78, + "learning_rate": 3.6374711464478073e-06, + "loss": 0.1634, + "step": 72357 + }, + { + "epoch": 2.78, + "learning_rate": 3.6368299563990765e-06, + "loss": 0.1657, + "step": 72358 + }, + { + "epoch": 2.78, + "learning_rate": 3.6361887663503465e-06, + "loss": 0.09, + "step": 72359 + }, + { + "epoch": 2.78, + "learning_rate": 3.6355475763016157e-06, + "loss": 0.1645, + "step": 72360 + }, + { + "epoch": 2.78, + "learning_rate": 3.6349063862528857e-06, + "loss": 0.1869, + "step": 72361 + }, + { + "epoch": 2.78, + "learning_rate": 3.634265196204155e-06, + "loss": 0.1929, + "step": 72362 + }, + { + "epoch": 2.78, + "learning_rate": 3.633624006155425e-06, + "loss": 0.1259, + "step": 72363 + }, + { + "epoch": 2.78, + "learning_rate": 3.632982816106694e-06, + "loss": 0.1653, + "step": 72364 + }, + { + "epoch": 2.78, + "learning_rate": 3.632341626057964e-06, + "loss": 0.1049, + "step": 72365 + }, + { + "epoch": 2.78, + "learning_rate": 3.6317004360092333e-06, + "loss": 0.2581, + "step": 72366 + }, + { + "epoch": 2.78, + "learning_rate": 3.6310592459605033e-06, + "loss": 0.3111, + "step": 72367 + }, + { + "epoch": 2.78, + "learning_rate": 3.6304180559117725e-06, + "loss": 0.1844, + "step": 72368 + }, + { + "epoch": 2.78, + "learning_rate": 3.6297768658630417e-06, + "loss": 0.1087, + "step": 72369 + }, + { + "epoch": 2.78, + "learning_rate": 3.6291356758143117e-06, + "loss": 0.262, + "step": 72370 + }, + { + "epoch": 2.78, + "learning_rate": 3.628494485765581e-06, + "loss": 0.2515, + "step": 72371 + }, + { + "epoch": 2.78, + "learning_rate": 3.627853295716851e-06, + "loss": 0.2388, + "step": 72372 + }, + { + "epoch": 2.78, + "learning_rate": 3.62721210566812e-06, + "loss": 0.2716, + "step": 72373 + }, + { + "epoch": 2.78, + "learning_rate": 3.62657091561939e-06, + "loss": 0.1181, + "step": 72374 + }, + { + "epoch": 2.78, + "learning_rate": 3.6259297255706593e-06, + "loss": 0.1754, + "step": 72375 + }, + { + "epoch": 2.78, + "learning_rate": 3.6252885355219293e-06, + "loss": 0.1702, + "step": 72376 + }, + { + "epoch": 2.78, + "learning_rate": 3.6246473454731984e-06, + "loss": 0.1938, + "step": 72377 + }, + { + "epoch": 2.78, + "learning_rate": 3.6240061554244676e-06, + "loss": 0.3122, + "step": 72378 + }, + { + "epoch": 2.78, + "learning_rate": 3.6233649653757376e-06, + "loss": 0.1667, + "step": 72379 + }, + { + "epoch": 2.78, + "learning_rate": 3.622723775327007e-06, + "loss": 0.167, + "step": 72380 + }, + { + "epoch": 2.78, + "learning_rate": 3.622082585278277e-06, + "loss": 0.2448, + "step": 72381 + }, + { + "epoch": 2.78, + "learning_rate": 3.621441395229546e-06, + "loss": 0.2057, + "step": 72382 + }, + { + "epoch": 2.78, + "learning_rate": 3.620800205180816e-06, + "loss": 0.1998, + "step": 72383 + }, + { + "epoch": 2.78, + "learning_rate": 3.6201590151320852e-06, + "loss": 0.2401, + "step": 72384 + }, + { + "epoch": 2.78, + "learning_rate": 3.6195178250833552e-06, + "loss": 0.174, + "step": 72385 + }, + { + "epoch": 2.78, + "learning_rate": 3.6188766350346244e-06, + "loss": 0.3203, + "step": 72386 + }, + { + "epoch": 2.78, + "learning_rate": 3.6182354449858936e-06, + "loss": 0.1412, + "step": 72387 + }, + { + "epoch": 2.78, + "learning_rate": 3.6175942549371636e-06, + "loss": 0.2732, + "step": 72388 + }, + { + "epoch": 2.78, + "learning_rate": 3.616953064888433e-06, + "loss": 0.1446, + "step": 72389 + }, + { + "epoch": 2.78, + "learning_rate": 3.616311874839703e-06, + "loss": 0.2359, + "step": 72390 + }, + { + "epoch": 2.78, + "learning_rate": 3.615670684790972e-06, + "loss": 0.1362, + "step": 72391 + }, + { + "epoch": 2.78, + "learning_rate": 3.615029494742242e-06, + "loss": 0.2456, + "step": 72392 + }, + { + "epoch": 2.78, + "learning_rate": 3.614388304693511e-06, + "loss": 0.1921, + "step": 72393 + }, + { + "epoch": 2.78, + "learning_rate": 3.613747114644781e-06, + "loss": 0.2052, + "step": 72394 + }, + { + "epoch": 2.78, + "learning_rate": 3.6131059245960504e-06, + "loss": 0.3363, + "step": 72395 + }, + { + "epoch": 2.78, + "learning_rate": 3.6124647345473204e-06, + "loss": 0.2808, + "step": 72396 + }, + { + "epoch": 2.78, + "learning_rate": 3.6118235444985896e-06, + "loss": 0.1444, + "step": 72397 + }, + { + "epoch": 2.78, + "learning_rate": 3.6111823544498588e-06, + "loss": 0.2662, + "step": 72398 + }, + { + "epoch": 2.78, + "learning_rate": 3.6105411644011288e-06, + "loss": 0.1973, + "step": 72399 + }, + { + "epoch": 2.78, + "learning_rate": 3.609899974352398e-06, + "loss": 0.1556, + "step": 72400 + }, + { + "epoch": 2.78, + "learning_rate": 3.609258784303668e-06, + "loss": 0.2277, + "step": 72401 + }, + { + "epoch": 2.78, + "learning_rate": 3.608617594254937e-06, + "loss": 0.2725, + "step": 72402 + }, + { + "epoch": 2.78, + "learning_rate": 3.607976404206207e-06, + "loss": 0.0615, + "step": 72403 + }, + { + "epoch": 2.78, + "learning_rate": 3.6073352141574764e-06, + "loss": 0.2771, + "step": 72404 + }, + { + "epoch": 2.78, + "learning_rate": 3.6066940241087464e-06, + "loss": 0.2599, + "step": 72405 + }, + { + "epoch": 2.78, + "learning_rate": 3.6060528340600156e-06, + "loss": 0.1732, + "step": 72406 + }, + { + "epoch": 2.78, + "learning_rate": 3.6054116440112847e-06, + "loss": 0.1388, + "step": 72407 + }, + { + "epoch": 2.78, + "learning_rate": 3.6047704539625547e-06, + "loss": 0.146, + "step": 72408 + }, + { + "epoch": 2.78, + "learning_rate": 3.604129263913824e-06, + "loss": 0.1662, + "step": 72409 + }, + { + "epoch": 2.78, + "learning_rate": 3.603488073865094e-06, + "loss": 0.2832, + "step": 72410 + }, + { + "epoch": 2.78, + "learning_rate": 3.602846883816363e-06, + "loss": 0.2007, + "step": 72411 + }, + { + "epoch": 2.78, + "learning_rate": 3.602205693767633e-06, + "loss": 0.1409, + "step": 72412 + }, + { + "epoch": 2.78, + "learning_rate": 3.6015645037189023e-06, + "loss": 0.3217, + "step": 72413 + }, + { + "epoch": 2.78, + "learning_rate": 3.6009233136701723e-06, + "loss": 0.2425, + "step": 72414 + }, + { + "epoch": 2.78, + "learning_rate": 3.6002821236214415e-06, + "loss": 0.2245, + "step": 72415 + }, + { + "epoch": 2.78, + "learning_rate": 3.5996409335727115e-06, + "loss": 0.2895, + "step": 72416 + }, + { + "epoch": 2.78, + "learning_rate": 3.5989997435239807e-06, + "loss": 0.0949, + "step": 72417 + }, + { + "epoch": 2.78, + "learning_rate": 3.59835855347525e-06, + "loss": 0.1523, + "step": 72418 + }, + { + "epoch": 2.78, + "learning_rate": 3.59771736342652e-06, + "loss": 0.3311, + "step": 72419 + }, + { + "epoch": 2.78, + "learning_rate": 3.597076173377789e-06, + "loss": 0.2408, + "step": 72420 + }, + { + "epoch": 2.78, + "learning_rate": 3.596434983329059e-06, + "loss": 0.2199, + "step": 72421 + }, + { + "epoch": 2.78, + "learning_rate": 3.5957937932803283e-06, + "loss": 0.1391, + "step": 72422 + }, + { + "epoch": 2.78, + "learning_rate": 3.5951526032315983e-06, + "loss": 0.141, + "step": 72423 + }, + { + "epoch": 2.78, + "learning_rate": 3.5945114131828675e-06, + "loss": 0.1445, + "step": 72424 + }, + { + "epoch": 2.78, + "learning_rate": 3.5938702231341375e-06, + "loss": 0.1095, + "step": 72425 + }, + { + "epoch": 2.78, + "learning_rate": 3.5932290330854067e-06, + "loss": 0.2199, + "step": 72426 + }, + { + "epoch": 2.78, + "learning_rate": 3.592587843036676e-06, + "loss": 0.3524, + "step": 72427 + }, + { + "epoch": 2.78, + "learning_rate": 3.591946652987946e-06, + "loss": 0.2865, + "step": 72428 + }, + { + "epoch": 2.78, + "learning_rate": 3.591305462939215e-06, + "loss": 0.2458, + "step": 72429 + }, + { + "epoch": 2.78, + "learning_rate": 3.590664272890485e-06, + "loss": 0.2158, + "step": 72430 + }, + { + "epoch": 2.78, + "learning_rate": 3.5900230828417543e-06, + "loss": 0.1922, + "step": 72431 + }, + { + "epoch": 2.78, + "learning_rate": 3.5893818927930243e-06, + "loss": 0.1424, + "step": 72432 + }, + { + "epoch": 2.78, + "learning_rate": 3.5887407027442935e-06, + "loss": 0.2119, + "step": 72433 + }, + { + "epoch": 2.78, + "learning_rate": 3.5880995126955635e-06, + "loss": 0.1581, + "step": 72434 + }, + { + "epoch": 2.78, + "learning_rate": 3.5874583226468327e-06, + "loss": 0.2637, + "step": 72435 + }, + { + "epoch": 2.78, + "learning_rate": 3.5868171325981027e-06, + "loss": 0.1421, + "step": 72436 + }, + { + "epoch": 2.78, + "learning_rate": 3.586175942549372e-06, + "loss": 0.1763, + "step": 72437 + }, + { + "epoch": 2.79, + "learning_rate": 3.585534752500641e-06, + "loss": 0.173, + "step": 72438 + }, + { + "epoch": 2.79, + "learning_rate": 3.584893562451911e-06, + "loss": 0.15, + "step": 72439 + }, + { + "epoch": 2.79, + "learning_rate": 3.5842523724031802e-06, + "loss": 0.3748, + "step": 72440 + }, + { + "epoch": 2.79, + "learning_rate": 3.5836111823544502e-06, + "loss": 0.2243, + "step": 72441 + }, + { + "epoch": 2.79, + "learning_rate": 3.5829699923057194e-06, + "loss": 0.2491, + "step": 72442 + }, + { + "epoch": 2.79, + "learning_rate": 3.5823288022569894e-06, + "loss": 0.2072, + "step": 72443 + }, + { + "epoch": 2.79, + "learning_rate": 3.5816876122082586e-06, + "loss": 0.139, + "step": 72444 + }, + { + "epoch": 2.79, + "learning_rate": 3.5810464221595286e-06, + "loss": 0.2964, + "step": 72445 + }, + { + "epoch": 2.79, + "learning_rate": 3.580405232110798e-06, + "loss": 0.2397, + "step": 72446 + }, + { + "epoch": 2.79, + "learning_rate": 3.579764042062067e-06, + "loss": 0.2082, + "step": 72447 + }, + { + "epoch": 2.79, + "learning_rate": 3.579122852013337e-06, + "loss": 0.184, + "step": 72448 + }, + { + "epoch": 2.79, + "learning_rate": 3.578481661964606e-06, + "loss": 0.2579, + "step": 72449 + }, + { + "epoch": 2.79, + "learning_rate": 3.5778404719158762e-06, + "loss": 0.3065, + "step": 72450 + }, + { + "epoch": 2.79, + "learning_rate": 3.5771992818671454e-06, + "loss": 0.0838, + "step": 72451 + }, + { + "epoch": 2.79, + "learning_rate": 3.5765580918184154e-06, + "loss": 0.1994, + "step": 72452 + }, + { + "epoch": 2.79, + "learning_rate": 3.5759169017696846e-06, + "loss": 0.2153, + "step": 72453 + }, + { + "epoch": 2.79, + "learning_rate": 3.5752757117209546e-06, + "loss": 0.1392, + "step": 72454 + }, + { + "epoch": 2.79, + "learning_rate": 3.5746345216722238e-06, + "loss": 0.1796, + "step": 72455 + }, + { + "epoch": 2.79, + "learning_rate": 3.573993331623493e-06, + "loss": 0.288, + "step": 72456 + }, + { + "epoch": 2.79, + "learning_rate": 3.573352141574763e-06, + "loss": 0.2442, + "step": 72457 + }, + { + "epoch": 2.79, + "learning_rate": 3.572710951526032e-06, + "loss": 0.1373, + "step": 72458 + }, + { + "epoch": 2.79, + "learning_rate": 3.572069761477302e-06, + "loss": 0.2039, + "step": 72459 + }, + { + "epoch": 2.79, + "learning_rate": 3.5714285714285714e-06, + "loss": 0.0973, + "step": 72460 + }, + { + "epoch": 2.79, + "learning_rate": 3.5707873813798414e-06, + "loss": 0.1405, + "step": 72461 + }, + { + "epoch": 2.79, + "learning_rate": 3.5701461913311106e-06, + "loss": 0.1558, + "step": 72462 + }, + { + "epoch": 2.79, + "learning_rate": 3.5695050012823806e-06, + "loss": 0.2116, + "step": 72463 + }, + { + "epoch": 2.79, + "learning_rate": 3.5688638112336498e-06, + "loss": 0.3592, + "step": 72464 + }, + { + "epoch": 2.79, + "learning_rate": 3.5682226211849198e-06, + "loss": 0.1554, + "step": 72465 + }, + { + "epoch": 2.79, + "learning_rate": 3.567581431136189e-06, + "loss": 0.2632, + "step": 72466 + }, + { + "epoch": 2.79, + "learning_rate": 3.566940241087458e-06, + "loss": 0.3464, + "step": 72467 + }, + { + "epoch": 2.79, + "learning_rate": 3.566299051038728e-06, + "loss": 0.1134, + "step": 72468 + }, + { + "epoch": 2.79, + "learning_rate": 3.5656578609899973e-06, + "loss": 0.131, + "step": 72469 + }, + { + "epoch": 2.79, + "learning_rate": 3.5650166709412673e-06, + "loss": 0.2693, + "step": 72470 + }, + { + "epoch": 2.79, + "learning_rate": 3.5643754808925365e-06, + "loss": 0.1406, + "step": 72471 + }, + { + "epoch": 2.79, + "learning_rate": 3.5637342908438065e-06, + "loss": 0.1368, + "step": 72472 + }, + { + "epoch": 2.79, + "learning_rate": 3.5630931007950757e-06, + "loss": 0.2017, + "step": 72473 + }, + { + "epoch": 2.79, + "learning_rate": 3.5624519107463457e-06, + "loss": 0.1262, + "step": 72474 + }, + { + "epoch": 2.79, + "learning_rate": 3.561810720697615e-06, + "loss": 0.22, + "step": 72475 + }, + { + "epoch": 2.79, + "learning_rate": 3.561169530648884e-06, + "loss": 0.3435, + "step": 72476 + }, + { + "epoch": 2.79, + "learning_rate": 3.560528340600154e-06, + "loss": 0.1712, + "step": 72477 + }, + { + "epoch": 2.79, + "learning_rate": 3.5598871505514233e-06, + "loss": 0.1691, + "step": 72478 + }, + { + "epoch": 2.79, + "learning_rate": 3.5592459605026933e-06, + "loss": 0.209, + "step": 72479 + }, + { + "epoch": 2.79, + "learning_rate": 3.5586047704539625e-06, + "loss": 0.214, + "step": 72480 + }, + { + "epoch": 2.79, + "learning_rate": 3.5579635804052325e-06, + "loss": 0.2423, + "step": 72481 + }, + { + "epoch": 2.79, + "learning_rate": 3.5573223903565017e-06, + "loss": 0.2004, + "step": 72482 + }, + { + "epoch": 2.79, + "learning_rate": 3.5566812003077717e-06, + "loss": 0.3406, + "step": 72483 + }, + { + "epoch": 2.79, + "learning_rate": 3.556040010259041e-06, + "loss": 0.1307, + "step": 72484 + }, + { + "epoch": 2.79, + "learning_rate": 3.555398820210311e-06, + "loss": 0.2054, + "step": 72485 + }, + { + "epoch": 2.79, + "learning_rate": 3.55475763016158e-06, + "loss": 0.3178, + "step": 72486 + }, + { + "epoch": 2.79, + "learning_rate": 3.5541164401128493e-06, + "loss": 0.3225, + "step": 72487 + }, + { + "epoch": 2.79, + "learning_rate": 3.5534752500641193e-06, + "loss": 0.2368, + "step": 72488 + }, + { + "epoch": 2.79, + "learning_rate": 3.5528340600153885e-06, + "loss": 0.2315, + "step": 72489 + }, + { + "epoch": 2.79, + "learning_rate": 3.5521928699666585e-06, + "loss": 0.1176, + "step": 72490 + }, + { + "epoch": 2.79, + "learning_rate": 3.5515516799179277e-06, + "loss": 0.3066, + "step": 72491 + }, + { + "epoch": 2.79, + "learning_rate": 3.5509104898691977e-06, + "loss": 0.2328, + "step": 72492 + }, + { + "epoch": 2.79, + "learning_rate": 3.550269299820467e-06, + "loss": 0.0767, + "step": 72493 + }, + { + "epoch": 2.79, + "learning_rate": 3.549628109771737e-06, + "loss": 0.1822, + "step": 72494 + }, + { + "epoch": 2.79, + "learning_rate": 3.548986919723006e-06, + "loss": 0.1942, + "step": 72495 + }, + { + "epoch": 2.79, + "learning_rate": 3.5483457296742752e-06, + "loss": 0.3122, + "step": 72496 + }, + { + "epoch": 2.79, + "learning_rate": 3.5477045396255453e-06, + "loss": 0.2599, + "step": 72497 + }, + { + "epoch": 2.79, + "learning_rate": 3.5470633495768144e-06, + "loss": 0.3632, + "step": 72498 + }, + { + "epoch": 2.79, + "learning_rate": 3.5464221595280844e-06, + "loss": 0.2939, + "step": 72499 + }, + { + "epoch": 2.79, + "learning_rate": 3.5457809694793536e-06, + "loss": 0.2525, + "step": 72500 + }, + { + "epoch": 2.79, + "learning_rate": 3.5451397794306236e-06, + "loss": 0.1935, + "step": 72501 + }, + { + "epoch": 2.79, + "learning_rate": 3.544498589381893e-06, + "loss": 0.2438, + "step": 72502 + }, + { + "epoch": 2.79, + "learning_rate": 3.543857399333163e-06, + "loss": 0.2701, + "step": 72503 + }, + { + "epoch": 2.79, + "learning_rate": 3.543216209284432e-06, + "loss": 0.1327, + "step": 72504 + }, + { + "epoch": 2.79, + "learning_rate": 3.542575019235701e-06, + "loss": 0.1466, + "step": 72505 + }, + { + "epoch": 2.79, + "learning_rate": 3.5419338291869712e-06, + "loss": 0.3092, + "step": 72506 + }, + { + "epoch": 2.79, + "learning_rate": 3.5412926391382404e-06, + "loss": 0.2385, + "step": 72507 + }, + { + "epoch": 2.79, + "learning_rate": 3.5406514490895104e-06, + "loss": 0.2584, + "step": 72508 + }, + { + "epoch": 2.79, + "learning_rate": 3.5400102590407796e-06, + "loss": 0.1872, + "step": 72509 + }, + { + "epoch": 2.79, + "learning_rate": 3.5393690689920496e-06, + "loss": 0.1084, + "step": 72510 + }, + { + "epoch": 2.79, + "learning_rate": 3.538727878943319e-06, + "loss": 0.1379, + "step": 72511 + }, + { + "epoch": 2.79, + "learning_rate": 3.538086688894589e-06, + "loss": 0.2694, + "step": 72512 + }, + { + "epoch": 2.79, + "learning_rate": 3.537445498845858e-06, + "loss": 0.185, + "step": 72513 + }, + { + "epoch": 2.79, + "learning_rate": 3.536804308797128e-06, + "loss": 0.1666, + "step": 72514 + }, + { + "epoch": 2.79, + "learning_rate": 3.536163118748397e-06, + "loss": 0.1818, + "step": 72515 + }, + { + "epoch": 2.79, + "learning_rate": 3.5355219286996664e-06, + "loss": 0.2518, + "step": 72516 + }, + { + "epoch": 2.79, + "learning_rate": 3.5348807386509364e-06, + "loss": 0.174, + "step": 72517 + }, + { + "epoch": 2.79, + "learning_rate": 3.5342395486022056e-06, + "loss": 0.1249, + "step": 72518 + }, + { + "epoch": 2.79, + "learning_rate": 3.5335983585534756e-06, + "loss": 0.2342, + "step": 72519 + }, + { + "epoch": 2.79, + "learning_rate": 3.5329571685047448e-06, + "loss": 0.158, + "step": 72520 + }, + { + "epoch": 2.79, + "learning_rate": 3.5323159784560148e-06, + "loss": 0.2131, + "step": 72521 + }, + { + "epoch": 2.79, + "learning_rate": 3.531674788407284e-06, + "loss": 0.1881, + "step": 72522 + }, + { + "epoch": 2.79, + "learning_rate": 3.531033598358554e-06, + "loss": 0.1648, + "step": 72523 + }, + { + "epoch": 2.79, + "learning_rate": 3.530392408309823e-06, + "loss": 0.3389, + "step": 72524 + }, + { + "epoch": 2.79, + "learning_rate": 3.5297512182610923e-06, + "loss": 0.2974, + "step": 72525 + }, + { + "epoch": 2.79, + "learning_rate": 3.5291100282123624e-06, + "loss": 0.2191, + "step": 72526 + }, + { + "epoch": 2.79, + "learning_rate": 3.5284688381636315e-06, + "loss": 0.3813, + "step": 72527 + }, + { + "epoch": 2.79, + "learning_rate": 3.5278276481149016e-06, + "loss": 0.2172, + "step": 72528 + }, + { + "epoch": 2.79, + "learning_rate": 3.5271864580661707e-06, + "loss": 0.206, + "step": 72529 + }, + { + "epoch": 2.79, + "learning_rate": 3.5265452680174407e-06, + "loss": 0.228, + "step": 72530 + }, + { + "epoch": 2.79, + "learning_rate": 3.52590407796871e-06, + "loss": 0.1574, + "step": 72531 + }, + { + "epoch": 2.79, + "learning_rate": 3.52526288791998e-06, + "loss": 0.1699, + "step": 72532 + }, + { + "epoch": 2.79, + "learning_rate": 3.524621697871249e-06, + "loss": 0.1448, + "step": 72533 + }, + { + "epoch": 2.79, + "learning_rate": 3.523980507822519e-06, + "loss": 0.2556, + "step": 72534 + }, + { + "epoch": 2.79, + "learning_rate": 3.5233393177737883e-06, + "loss": 0.3048, + "step": 72535 + }, + { + "epoch": 2.79, + "learning_rate": 3.5226981277250575e-06, + "loss": 0.1105, + "step": 72536 + }, + { + "epoch": 2.79, + "learning_rate": 3.5220569376763275e-06, + "loss": 0.1675, + "step": 72537 + }, + { + "epoch": 2.79, + "learning_rate": 3.5214157476275967e-06, + "loss": 0.2555, + "step": 72538 + }, + { + "epoch": 2.79, + "learning_rate": 3.5207745575788667e-06, + "loss": 0.1798, + "step": 72539 + }, + { + "epoch": 2.79, + "learning_rate": 3.520133367530136e-06, + "loss": 0.2315, + "step": 72540 + }, + { + "epoch": 2.79, + "learning_rate": 3.519492177481406e-06, + "loss": 0.1491, + "step": 72541 + }, + { + "epoch": 2.79, + "learning_rate": 3.518850987432675e-06, + "loss": 0.1817, + "step": 72542 + }, + { + "epoch": 2.79, + "learning_rate": 3.518209797383945e-06, + "loss": 0.3762, + "step": 72543 + }, + { + "epoch": 2.79, + "learning_rate": 3.5175686073352143e-06, + "loss": 0.2414, + "step": 72544 + }, + { + "epoch": 2.79, + "learning_rate": 3.5169274172864835e-06, + "loss": 0.149, + "step": 72545 + }, + { + "epoch": 2.79, + "learning_rate": 3.5162862272377535e-06, + "loss": 0.1646, + "step": 72546 + }, + { + "epoch": 2.79, + "learning_rate": 3.5156450371890227e-06, + "loss": 0.2718, + "step": 72547 + }, + { + "epoch": 2.79, + "learning_rate": 3.5150038471402927e-06, + "loss": 0.1729, + "step": 72548 + }, + { + "epoch": 2.79, + "learning_rate": 3.514362657091562e-06, + "loss": 0.1267, + "step": 72549 + }, + { + "epoch": 2.79, + "learning_rate": 3.513721467042832e-06, + "loss": 0.1515, + "step": 72550 + }, + { + "epoch": 2.79, + "learning_rate": 3.513080276994101e-06, + "loss": 0.1992, + "step": 72551 + }, + { + "epoch": 2.79, + "learning_rate": 3.512439086945371e-06, + "loss": 0.3016, + "step": 72552 + }, + { + "epoch": 2.79, + "learning_rate": 3.5117978968966403e-06, + "loss": 0.2468, + "step": 72553 + }, + { + "epoch": 2.79, + "learning_rate": 3.5111567068479103e-06, + "loss": 0.3343, + "step": 72554 + }, + { + "epoch": 2.79, + "learning_rate": 3.5105155167991795e-06, + "loss": 0.1709, + "step": 72555 + }, + { + "epoch": 2.79, + "learning_rate": 3.5098743267504486e-06, + "loss": 0.304, + "step": 72556 + }, + { + "epoch": 2.79, + "learning_rate": 3.5092331367017187e-06, + "loss": 0.2426, + "step": 72557 + }, + { + "epoch": 2.79, + "learning_rate": 3.508591946652988e-06, + "loss": 0.2325, + "step": 72558 + }, + { + "epoch": 2.79, + "learning_rate": 3.507950756604258e-06, + "loss": 0.2086, + "step": 72559 + }, + { + "epoch": 2.79, + "learning_rate": 3.507309566555527e-06, + "loss": 0.2495, + "step": 72560 + }, + { + "epoch": 2.79, + "learning_rate": 3.506668376506797e-06, + "loss": 0.2507, + "step": 72561 + }, + { + "epoch": 2.79, + "learning_rate": 3.5060271864580662e-06, + "loss": 0.1218, + "step": 72562 + }, + { + "epoch": 2.79, + "learning_rate": 3.5053859964093362e-06, + "loss": 0.3987, + "step": 72563 + }, + { + "epoch": 2.79, + "learning_rate": 3.5047448063606054e-06, + "loss": 0.2599, + "step": 72564 + }, + { + "epoch": 2.79, + "learning_rate": 3.5041036163118746e-06, + "loss": 0.2294, + "step": 72565 + }, + { + "epoch": 2.79, + "learning_rate": 3.5034624262631446e-06, + "loss": 0.375, + "step": 72566 + }, + { + "epoch": 2.79, + "learning_rate": 3.502821236214414e-06, + "loss": 0.1983, + "step": 72567 + }, + { + "epoch": 2.79, + "learning_rate": 3.502180046165684e-06, + "loss": 0.2472, + "step": 72568 + }, + { + "epoch": 2.79, + "learning_rate": 3.501538856116953e-06, + "loss": 0.2646, + "step": 72569 + }, + { + "epoch": 2.79, + "learning_rate": 3.500897666068223e-06, + "loss": 0.2554, + "step": 72570 + }, + { + "epoch": 2.79, + "learning_rate": 3.500256476019492e-06, + "loss": 0.1645, + "step": 72571 + }, + { + "epoch": 2.79, + "learning_rate": 3.4996152859707622e-06, + "loss": 0.1719, + "step": 72572 + }, + { + "epoch": 2.79, + "learning_rate": 3.4989740959220314e-06, + "loss": 0.149, + "step": 72573 + }, + { + "epoch": 2.79, + "learning_rate": 3.4983329058733006e-06, + "loss": 0.1088, + "step": 72574 + }, + { + "epoch": 2.79, + "learning_rate": 3.4976917158245706e-06, + "loss": 0.2493, + "step": 72575 + }, + { + "epoch": 2.79, + "learning_rate": 3.4970505257758398e-06, + "loss": 0.27, + "step": 72576 + }, + { + "epoch": 2.79, + "learning_rate": 3.4964093357271098e-06, + "loss": 0.2913, + "step": 72577 + }, + { + "epoch": 2.79, + "learning_rate": 3.495768145678379e-06, + "loss": 0.1986, + "step": 72578 + }, + { + "epoch": 2.79, + "learning_rate": 3.495126955629649e-06, + "loss": 0.1107, + "step": 72579 + }, + { + "epoch": 2.79, + "learning_rate": 3.494485765580918e-06, + "loss": 0.2198, + "step": 72580 + }, + { + "epoch": 2.79, + "learning_rate": 3.493844575532188e-06, + "loss": 0.2036, + "step": 72581 + }, + { + "epoch": 2.79, + "learning_rate": 3.4932033854834574e-06, + "loss": 0.1706, + "step": 72582 + }, + { + "epoch": 2.79, + "learning_rate": 3.4925621954347274e-06, + "loss": 0.2796, + "step": 72583 + }, + { + "epoch": 2.79, + "learning_rate": 3.4919210053859966e-06, + "loss": 0.1375, + "step": 72584 + }, + { + "epoch": 2.79, + "learning_rate": 3.4912798153372657e-06, + "loss": 0.2135, + "step": 72585 + }, + { + "epoch": 2.79, + "learning_rate": 3.4906386252885358e-06, + "loss": 0.2338, + "step": 72586 + }, + { + "epoch": 2.79, + "learning_rate": 3.489997435239805e-06, + "loss": 0.0844, + "step": 72587 + }, + { + "epoch": 2.79, + "learning_rate": 3.489356245191075e-06, + "loss": 0.2448, + "step": 72588 + }, + { + "epoch": 2.79, + "learning_rate": 3.488715055142344e-06, + "loss": 0.2267, + "step": 72589 + }, + { + "epoch": 2.79, + "learning_rate": 3.488073865093614e-06, + "loss": 0.1878, + "step": 72590 + }, + { + "epoch": 2.79, + "learning_rate": 3.4874326750448833e-06, + "loss": 0.145, + "step": 72591 + }, + { + "epoch": 2.79, + "learning_rate": 3.4867914849961533e-06, + "loss": 0.1282, + "step": 72592 + }, + { + "epoch": 2.79, + "learning_rate": 3.4861502949474225e-06, + "loss": 0.2417, + "step": 72593 + }, + { + "epoch": 2.79, + "learning_rate": 3.4855091048986917e-06, + "loss": 0.1563, + "step": 72594 + }, + { + "epoch": 2.79, + "learning_rate": 3.4848679148499617e-06, + "loss": 0.1106, + "step": 72595 + }, + { + "epoch": 2.79, + "learning_rate": 3.484226724801231e-06, + "loss": 0.2148, + "step": 72596 + }, + { + "epoch": 2.79, + "learning_rate": 3.483585534752501e-06, + "loss": 0.1388, + "step": 72597 + }, + { + "epoch": 2.79, + "learning_rate": 3.48294434470377e-06, + "loss": 0.1207, + "step": 72598 + }, + { + "epoch": 2.79, + "learning_rate": 3.48230315465504e-06, + "loss": 0.1926, + "step": 72599 + }, + { + "epoch": 2.79, + "learning_rate": 3.4816619646063093e-06, + "loss": 0.1238, + "step": 72600 + }, + { + "epoch": 2.79, + "learning_rate": 3.4810207745575793e-06, + "loss": 0.2173, + "step": 72601 + }, + { + "epoch": 2.79, + "learning_rate": 3.4803795845088485e-06, + "loss": 0.1599, + "step": 72602 + }, + { + "epoch": 2.79, + "learning_rate": 3.4797383944601185e-06, + "loss": 0.2468, + "step": 72603 + }, + { + "epoch": 2.79, + "learning_rate": 3.4790972044113877e-06, + "loss": 0.175, + "step": 72604 + }, + { + "epoch": 2.79, + "learning_rate": 3.478456014362657e-06, + "loss": 0.2851, + "step": 72605 + }, + { + "epoch": 2.79, + "learning_rate": 3.477814824313927e-06, + "loss": 0.2791, + "step": 72606 + }, + { + "epoch": 2.79, + "learning_rate": 3.477173634265196e-06, + "loss": 0.1971, + "step": 72607 + }, + { + "epoch": 2.79, + "learning_rate": 3.476532444216466e-06, + "loss": 0.1436, + "step": 72608 + }, + { + "epoch": 2.79, + "learning_rate": 3.4758912541677353e-06, + "loss": 0.2362, + "step": 72609 + }, + { + "epoch": 2.79, + "learning_rate": 3.4752500641190053e-06, + "loss": 0.1005, + "step": 72610 + }, + { + "epoch": 2.79, + "learning_rate": 3.4746088740702745e-06, + "loss": 0.1397, + "step": 72611 + }, + { + "epoch": 2.79, + "learning_rate": 3.4739676840215445e-06, + "loss": 0.2547, + "step": 72612 + }, + { + "epoch": 2.79, + "learning_rate": 3.4733264939728137e-06, + "loss": 0.1927, + "step": 72613 + }, + { + "epoch": 2.79, + "learning_rate": 3.472685303924083e-06, + "loss": 0.2292, + "step": 72614 + }, + { + "epoch": 2.79, + "learning_rate": 3.472044113875353e-06, + "loss": 0.0863, + "step": 72615 + }, + { + "epoch": 2.79, + "learning_rate": 3.471402923826622e-06, + "loss": 0.2387, + "step": 72616 + }, + { + "epoch": 2.79, + "learning_rate": 3.470761733777892e-06, + "loss": 0.1941, + "step": 72617 + }, + { + "epoch": 2.79, + "learning_rate": 3.4701205437291612e-06, + "loss": 0.1059, + "step": 72618 + }, + { + "epoch": 2.79, + "learning_rate": 3.4694793536804313e-06, + "loss": 0.2752, + "step": 72619 + }, + { + "epoch": 2.79, + "learning_rate": 3.4688381636317004e-06, + "loss": 0.2227, + "step": 72620 + }, + { + "epoch": 2.79, + "learning_rate": 3.4681969735829705e-06, + "loss": 0.1837, + "step": 72621 + }, + { + "epoch": 2.79, + "learning_rate": 3.4675557835342396e-06, + "loss": 0.1207, + "step": 72622 + }, + { + "epoch": 2.79, + "learning_rate": 3.466914593485509e-06, + "loss": 0.1878, + "step": 72623 + }, + { + "epoch": 2.79, + "learning_rate": 3.466273403436779e-06, + "loss": 0.2586, + "step": 72624 + }, + { + "epoch": 2.79, + "learning_rate": 3.465632213388048e-06, + "loss": 0.1189, + "step": 72625 + }, + { + "epoch": 2.79, + "learning_rate": 3.464991023339318e-06, + "loss": 0.3595, + "step": 72626 + }, + { + "epoch": 2.79, + "learning_rate": 3.464349833290587e-06, + "loss": 0.167, + "step": 72627 + }, + { + "epoch": 2.79, + "learning_rate": 3.4637086432418572e-06, + "loss": 0.1792, + "step": 72628 + }, + { + "epoch": 2.79, + "learning_rate": 3.4630674531931264e-06, + "loss": 0.2118, + "step": 72629 + }, + { + "epoch": 2.79, + "learning_rate": 3.4624262631443964e-06, + "loss": 0.1964, + "step": 72630 + }, + { + "epoch": 2.79, + "learning_rate": 3.4617850730956656e-06, + "loss": 0.1317, + "step": 72631 + }, + { + "epoch": 2.79, + "learning_rate": 3.4611438830469356e-06, + "loss": 0.2401, + "step": 72632 + }, + { + "epoch": 2.79, + "learning_rate": 3.460502692998205e-06, + "loss": 0.1371, + "step": 72633 + }, + { + "epoch": 2.79, + "learning_rate": 3.459861502949474e-06, + "loss": 0.2833, + "step": 72634 + }, + { + "epoch": 2.79, + "learning_rate": 3.459220312900744e-06, + "loss": 0.1796, + "step": 72635 + }, + { + "epoch": 2.79, + "learning_rate": 3.458579122852013e-06, + "loss": 0.1526, + "step": 72636 + }, + { + "epoch": 2.79, + "learning_rate": 3.457937932803283e-06, + "loss": 0.249, + "step": 72637 + }, + { + "epoch": 2.79, + "learning_rate": 3.4572967427545524e-06, + "loss": 0.204, + "step": 72638 + }, + { + "epoch": 2.79, + "learning_rate": 3.4566555527058224e-06, + "loss": 0.1572, + "step": 72639 + }, + { + "epoch": 2.79, + "learning_rate": 3.4560143626570916e-06, + "loss": 0.2058, + "step": 72640 + }, + { + "epoch": 2.79, + "learning_rate": 3.4553731726083616e-06, + "loss": 0.1722, + "step": 72641 + }, + { + "epoch": 2.79, + "learning_rate": 3.4547319825596308e-06, + "loss": 0.1158, + "step": 72642 + }, + { + "epoch": 2.79, + "learning_rate": 3.4540907925109e-06, + "loss": 0.3724, + "step": 72643 + }, + { + "epoch": 2.79, + "learning_rate": 3.45344960246217e-06, + "loss": 0.1954, + "step": 72644 + }, + { + "epoch": 2.79, + "learning_rate": 3.452808412413439e-06, + "loss": 0.1647, + "step": 72645 + }, + { + "epoch": 2.79, + "learning_rate": 3.452167222364709e-06, + "loss": 0.235, + "step": 72646 + }, + { + "epoch": 2.79, + "learning_rate": 3.4515260323159783e-06, + "loss": 0.172, + "step": 72647 + }, + { + "epoch": 2.79, + "learning_rate": 3.4508848422672484e-06, + "loss": 0.2396, + "step": 72648 + }, + { + "epoch": 2.79, + "learning_rate": 3.4502436522185175e-06, + "loss": 0.3165, + "step": 72649 + }, + { + "epoch": 2.79, + "learning_rate": 3.4496024621697876e-06, + "loss": 0.117, + "step": 72650 + }, + { + "epoch": 2.79, + "learning_rate": 3.4489612721210567e-06, + "loss": 0.1423, + "step": 72651 + }, + { + "epoch": 2.79, + "learning_rate": 3.4483200820723268e-06, + "loss": 0.3079, + "step": 72652 + }, + { + "epoch": 2.79, + "learning_rate": 3.447678892023596e-06, + "loss": 0.2532, + "step": 72653 + }, + { + "epoch": 2.79, + "learning_rate": 3.447037701974865e-06, + "loss": 0.2038, + "step": 72654 + }, + { + "epoch": 2.79, + "learning_rate": 3.446396511926135e-06, + "loss": 0.2319, + "step": 72655 + }, + { + "epoch": 2.79, + "learning_rate": 3.4457553218774043e-06, + "loss": 0.199, + "step": 72656 + }, + { + "epoch": 2.79, + "learning_rate": 3.4451141318286743e-06, + "loss": 0.283, + "step": 72657 + }, + { + "epoch": 2.79, + "learning_rate": 3.4444729417799435e-06, + "loss": 0.1543, + "step": 72658 + }, + { + "epoch": 2.79, + "learning_rate": 3.4438317517312135e-06, + "loss": 0.3257, + "step": 72659 + }, + { + "epoch": 2.79, + "learning_rate": 3.4431905616824827e-06, + "loss": 0.4895, + "step": 72660 + }, + { + "epoch": 2.79, + "learning_rate": 3.4425493716337527e-06, + "loss": 0.1624, + "step": 72661 + }, + { + "epoch": 2.79, + "learning_rate": 3.441908181585022e-06, + "loss": 0.0898, + "step": 72662 + }, + { + "epoch": 2.79, + "learning_rate": 3.441266991536291e-06, + "loss": 0.1515, + "step": 72663 + }, + { + "epoch": 2.79, + "learning_rate": 3.440625801487561e-06, + "loss": 0.1978, + "step": 72664 + }, + { + "epoch": 2.79, + "learning_rate": 3.4399846114388303e-06, + "loss": 0.2494, + "step": 72665 + }, + { + "epoch": 2.79, + "learning_rate": 3.4393434213901003e-06, + "loss": 0.1064, + "step": 72666 + }, + { + "epoch": 2.79, + "learning_rate": 3.4387022313413695e-06, + "loss": 0.1704, + "step": 72667 + }, + { + "epoch": 2.79, + "learning_rate": 3.4380610412926395e-06, + "loss": 0.2802, + "step": 72668 + }, + { + "epoch": 2.79, + "learning_rate": 3.4374198512439087e-06, + "loss": 0.345, + "step": 72669 + }, + { + "epoch": 2.79, + "learning_rate": 3.4367786611951787e-06, + "loss": 0.1885, + "step": 72670 + }, + { + "epoch": 2.79, + "learning_rate": 3.436137471146448e-06, + "loss": 0.111, + "step": 72671 + }, + { + "epoch": 2.79, + "learning_rate": 3.435496281097718e-06, + "loss": 0.1625, + "step": 72672 + }, + { + "epoch": 2.79, + "learning_rate": 3.434855091048987e-06, + "loss": 0.1639, + "step": 72673 + }, + { + "epoch": 2.79, + "learning_rate": 3.4342139010002562e-06, + "loss": 0.1644, + "step": 72674 + }, + { + "epoch": 2.79, + "learning_rate": 3.4335727109515263e-06, + "loss": 0.1079, + "step": 72675 + }, + { + "epoch": 2.79, + "learning_rate": 3.4329315209027954e-06, + "loss": 0.1697, + "step": 72676 + }, + { + "epoch": 2.79, + "learning_rate": 3.4322903308540655e-06, + "loss": 0.178, + "step": 72677 + }, + { + "epoch": 2.79, + "learning_rate": 3.4316491408053346e-06, + "loss": 0.2831, + "step": 72678 + }, + { + "epoch": 2.79, + "learning_rate": 3.4310079507566047e-06, + "loss": 0.199, + "step": 72679 + }, + { + "epoch": 2.79, + "learning_rate": 3.430366760707874e-06, + "loss": 0.2144, + "step": 72680 + }, + { + "epoch": 2.79, + "learning_rate": 3.429725570659144e-06, + "loss": 0.2637, + "step": 72681 + }, + { + "epoch": 2.79, + "learning_rate": 3.429084380610413e-06, + "loss": 0.2061, + "step": 72682 + }, + { + "epoch": 2.79, + "learning_rate": 3.428443190561682e-06, + "loss": 0.2589, + "step": 72683 + }, + { + "epoch": 2.79, + "learning_rate": 3.4278020005129522e-06, + "loss": 0.2168, + "step": 72684 + }, + { + "epoch": 2.79, + "learning_rate": 3.4271608104642214e-06, + "loss": 0.2875, + "step": 72685 + }, + { + "epoch": 2.79, + "learning_rate": 3.4265196204154914e-06, + "loss": 0.2361, + "step": 72686 + }, + { + "epoch": 2.79, + "learning_rate": 3.4258784303667606e-06, + "loss": 0.3852, + "step": 72687 + }, + { + "epoch": 2.79, + "learning_rate": 3.4252372403180306e-06, + "loss": 0.1532, + "step": 72688 + }, + { + "epoch": 2.79, + "learning_rate": 3.4245960502693e-06, + "loss": 0.1853, + "step": 72689 + }, + { + "epoch": 2.79, + "learning_rate": 3.42395486022057e-06, + "loss": 0.1523, + "step": 72690 + }, + { + "epoch": 2.79, + "learning_rate": 3.423313670171839e-06, + "loss": 0.3032, + "step": 72691 + }, + { + "epoch": 2.79, + "learning_rate": 3.422672480123108e-06, + "loss": 0.1754, + "step": 72692 + }, + { + "epoch": 2.79, + "learning_rate": 3.422031290074378e-06, + "loss": 0.3067, + "step": 72693 + }, + { + "epoch": 2.79, + "learning_rate": 3.4213901000256474e-06, + "loss": 0.2699, + "step": 72694 + }, + { + "epoch": 2.79, + "learning_rate": 3.4207489099769174e-06, + "loss": 0.2748, + "step": 72695 + }, + { + "epoch": 2.79, + "learning_rate": 3.4201077199281866e-06, + "loss": 0.1935, + "step": 72696 + }, + { + "epoch": 2.79, + "learning_rate": 3.4194665298794566e-06, + "loss": 0.2485, + "step": 72697 + }, + { + "epoch": 2.8, + "learning_rate": 3.4188253398307258e-06, + "loss": 0.2153, + "step": 72698 + }, + { + "epoch": 2.8, + "learning_rate": 3.4181841497819958e-06, + "loss": 0.1882, + "step": 72699 + }, + { + "epoch": 2.8, + "learning_rate": 3.417542959733265e-06, + "loss": 0.315, + "step": 72700 + }, + { + "epoch": 2.8, + "learning_rate": 3.416901769684535e-06, + "loss": 0.2798, + "step": 72701 + }, + { + "epoch": 2.8, + "learning_rate": 3.416260579635804e-06, + "loss": 0.2356, + "step": 72702 + }, + { + "epoch": 2.8, + "learning_rate": 3.4156193895870733e-06, + "loss": 0.2102, + "step": 72703 + }, + { + "epoch": 2.8, + "learning_rate": 3.4149781995383434e-06, + "loss": 0.1573, + "step": 72704 + }, + { + "epoch": 2.8, + "learning_rate": 3.4143370094896125e-06, + "loss": 0.2464, + "step": 72705 + }, + { + "epoch": 2.8, + "learning_rate": 3.4136958194408826e-06, + "loss": 0.2153, + "step": 72706 + }, + { + "epoch": 2.8, + "learning_rate": 3.4130546293921517e-06, + "loss": 0.2587, + "step": 72707 + }, + { + "epoch": 2.8, + "learning_rate": 3.4124134393434218e-06, + "loss": 0.1218, + "step": 72708 + }, + { + "epoch": 2.8, + "learning_rate": 3.411772249294691e-06, + "loss": 0.2144, + "step": 72709 + }, + { + "epoch": 2.8, + "learning_rate": 3.411131059245961e-06, + "loss": 0.1539, + "step": 72710 + }, + { + "epoch": 2.8, + "learning_rate": 3.41048986919723e-06, + "loss": 0.3644, + "step": 72711 + }, + { + "epoch": 2.8, + "learning_rate": 3.4098486791484993e-06, + "loss": 0.2271, + "step": 72712 + }, + { + "epoch": 2.8, + "learning_rate": 3.4092074890997693e-06, + "loss": 0.1082, + "step": 72713 + }, + { + "epoch": 2.8, + "learning_rate": 3.4085662990510385e-06, + "loss": 0.2143, + "step": 72714 + }, + { + "epoch": 2.8, + "learning_rate": 3.4079251090023085e-06, + "loss": 0.1069, + "step": 72715 + }, + { + "epoch": 2.8, + "learning_rate": 3.4072839189535777e-06, + "loss": 0.1405, + "step": 72716 + }, + { + "epoch": 2.8, + "learning_rate": 3.4066427289048477e-06, + "loss": 0.166, + "step": 72717 + }, + { + "epoch": 2.8, + "learning_rate": 3.406001538856117e-06, + "loss": 0.1676, + "step": 72718 + }, + { + "epoch": 2.8, + "learning_rate": 3.405360348807387e-06, + "loss": 0.0771, + "step": 72719 + }, + { + "epoch": 2.8, + "learning_rate": 3.404719158758656e-06, + "loss": 0.2678, + "step": 72720 + }, + { + "epoch": 2.8, + "learning_rate": 3.404077968709926e-06, + "loss": 0.1443, + "step": 72721 + }, + { + "epoch": 2.8, + "learning_rate": 3.4034367786611953e-06, + "loss": 0.1849, + "step": 72722 + }, + { + "epoch": 2.8, + "learning_rate": 3.4027955886124645e-06, + "loss": 0.2519, + "step": 72723 + }, + { + "epoch": 2.8, + "learning_rate": 3.4021543985637345e-06, + "loss": 0.2204, + "step": 72724 + }, + { + "epoch": 2.8, + "learning_rate": 3.4015132085150037e-06, + "loss": 0.1817, + "step": 72725 + }, + { + "epoch": 2.8, + "learning_rate": 3.4008720184662737e-06, + "loss": 0.2345, + "step": 72726 + }, + { + "epoch": 2.8, + "learning_rate": 3.400230828417543e-06, + "loss": 0.1718, + "step": 72727 + }, + { + "epoch": 2.8, + "learning_rate": 3.399589638368813e-06, + "loss": 0.3712, + "step": 72728 + }, + { + "epoch": 2.8, + "learning_rate": 3.398948448320082e-06, + "loss": 0.1209, + "step": 72729 + }, + { + "epoch": 2.8, + "learning_rate": 3.398307258271352e-06, + "loss": 0.1387, + "step": 72730 + }, + { + "epoch": 2.8, + "learning_rate": 3.3976660682226213e-06, + "loss": 0.2143, + "step": 72731 + }, + { + "epoch": 2.8, + "learning_rate": 3.3970248781738904e-06, + "loss": 0.1195, + "step": 72732 + }, + { + "epoch": 2.8, + "learning_rate": 3.3963836881251605e-06, + "loss": 0.1936, + "step": 72733 + }, + { + "epoch": 2.8, + "learning_rate": 3.3957424980764296e-06, + "loss": 0.3141, + "step": 72734 + }, + { + "epoch": 2.8, + "learning_rate": 3.3951013080276997e-06, + "loss": 0.1706, + "step": 72735 + }, + { + "epoch": 2.8, + "learning_rate": 3.394460117978969e-06, + "loss": 0.1623, + "step": 72736 + }, + { + "epoch": 2.8, + "learning_rate": 3.393818927930239e-06, + "loss": 0.3119, + "step": 72737 + }, + { + "epoch": 2.8, + "learning_rate": 3.393177737881508e-06, + "loss": 0.1945, + "step": 72738 + }, + { + "epoch": 2.8, + "learning_rate": 3.392536547832778e-06, + "loss": 0.1736, + "step": 72739 + }, + { + "epoch": 2.8, + "learning_rate": 3.3918953577840472e-06, + "loss": 0.2588, + "step": 72740 + }, + { + "epoch": 2.8, + "learning_rate": 3.3912541677353164e-06, + "loss": 0.1479, + "step": 72741 + }, + { + "epoch": 2.8, + "learning_rate": 3.3906129776865864e-06, + "loss": 0.267, + "step": 72742 + }, + { + "epoch": 2.8, + "learning_rate": 3.3899717876378556e-06, + "loss": 0.205, + "step": 72743 + }, + { + "epoch": 2.8, + "learning_rate": 3.3893305975891256e-06, + "loss": 0.2791, + "step": 72744 + }, + { + "epoch": 2.8, + "learning_rate": 3.388689407540395e-06, + "loss": 0.1552, + "step": 72745 + }, + { + "epoch": 2.8, + "learning_rate": 3.388048217491665e-06, + "loss": 0.2851, + "step": 72746 + }, + { + "epoch": 2.8, + "learning_rate": 3.387407027442934e-06, + "loss": 0.2325, + "step": 72747 + }, + { + "epoch": 2.8, + "learning_rate": 3.386765837394204e-06, + "loss": 0.1709, + "step": 72748 + }, + { + "epoch": 2.8, + "learning_rate": 3.386124647345473e-06, + "loss": 0.2196, + "step": 72749 + }, + { + "epoch": 2.8, + "learning_rate": 3.3854834572967432e-06, + "loss": 0.3676, + "step": 72750 + }, + { + "epoch": 2.8, + "learning_rate": 3.3848422672480124e-06, + "loss": 0.336, + "step": 72751 + }, + { + "epoch": 2.8, + "learning_rate": 3.3842010771992816e-06, + "loss": 0.1347, + "step": 72752 + }, + { + "epoch": 2.8, + "learning_rate": 3.3835598871505516e-06, + "loss": 0.2357, + "step": 72753 + }, + { + "epoch": 2.8, + "learning_rate": 3.3829186971018208e-06, + "loss": 0.131, + "step": 72754 + }, + { + "epoch": 2.8, + "learning_rate": 3.382277507053091e-06, + "loss": 0.2514, + "step": 72755 + }, + { + "epoch": 2.8, + "learning_rate": 3.38163631700436e-06, + "loss": 0.1838, + "step": 72756 + }, + { + "epoch": 2.8, + "learning_rate": 3.38099512695563e-06, + "loss": 0.2161, + "step": 72757 + }, + { + "epoch": 2.8, + "learning_rate": 3.380353936906899e-06, + "loss": 0.2843, + "step": 72758 + }, + { + "epoch": 2.8, + "learning_rate": 3.379712746858169e-06, + "loss": 0.1666, + "step": 72759 + }, + { + "epoch": 2.8, + "learning_rate": 3.3790715568094384e-06, + "loss": 0.2345, + "step": 72760 + }, + { + "epoch": 2.8, + "learning_rate": 3.3784303667607075e-06, + "loss": 0.1903, + "step": 72761 + }, + { + "epoch": 2.8, + "learning_rate": 3.3777891767119776e-06, + "loss": 0.2022, + "step": 72762 + }, + { + "epoch": 2.8, + "learning_rate": 3.3771479866632467e-06, + "loss": 0.2706, + "step": 72763 + }, + { + "epoch": 2.8, + "learning_rate": 3.3765067966145168e-06, + "loss": 0.1727, + "step": 72764 + }, + { + "epoch": 2.8, + "learning_rate": 3.375865606565786e-06, + "loss": 0.1142, + "step": 72765 + }, + { + "epoch": 2.8, + "learning_rate": 3.375224416517056e-06, + "loss": 0.2075, + "step": 72766 + }, + { + "epoch": 2.8, + "learning_rate": 3.374583226468325e-06, + "loss": 0.2859, + "step": 72767 + }, + { + "epoch": 2.8, + "learning_rate": 3.373942036419595e-06, + "loss": 0.1799, + "step": 72768 + }, + { + "epoch": 2.8, + "learning_rate": 3.3733008463708643e-06, + "loss": 0.1317, + "step": 72769 + }, + { + "epoch": 2.8, + "learning_rate": 3.3726596563221344e-06, + "loss": 0.2891, + "step": 72770 + }, + { + "epoch": 2.8, + "learning_rate": 3.3720184662734035e-06, + "loss": 0.2149, + "step": 72771 + }, + { + "epoch": 2.8, + "learning_rate": 3.3713772762246727e-06, + "loss": 0.2415, + "step": 72772 + }, + { + "epoch": 2.8, + "learning_rate": 3.3707360861759427e-06, + "loss": 0.1159, + "step": 72773 + }, + { + "epoch": 2.8, + "learning_rate": 3.370094896127212e-06, + "loss": 0.127, + "step": 72774 + }, + { + "epoch": 2.8, + "learning_rate": 3.369453706078482e-06, + "loss": 0.2397, + "step": 72775 + }, + { + "epoch": 2.8, + "learning_rate": 3.368812516029751e-06, + "loss": 0.2108, + "step": 72776 + }, + { + "epoch": 2.8, + "learning_rate": 3.368171325981021e-06, + "loss": 0.1913, + "step": 72777 + }, + { + "epoch": 2.8, + "learning_rate": 3.3675301359322903e-06, + "loss": 0.2994, + "step": 72778 + }, + { + "epoch": 2.8, + "learning_rate": 3.3668889458835603e-06, + "loss": 0.4022, + "step": 72779 + }, + { + "epoch": 2.8, + "learning_rate": 3.3662477558348295e-06, + "loss": 0.3003, + "step": 72780 + }, + { + "epoch": 2.8, + "learning_rate": 3.3656065657860987e-06, + "loss": 0.1421, + "step": 72781 + }, + { + "epoch": 2.8, + "learning_rate": 3.3649653757373687e-06, + "loss": 0.2211, + "step": 72782 + }, + { + "epoch": 2.8, + "learning_rate": 3.364324185688638e-06, + "loss": 0.1821, + "step": 72783 + }, + { + "epoch": 2.8, + "learning_rate": 3.363682995639908e-06, + "loss": 0.1379, + "step": 72784 + }, + { + "epoch": 2.8, + "learning_rate": 3.363041805591177e-06, + "loss": 0.1961, + "step": 72785 + }, + { + "epoch": 2.8, + "learning_rate": 3.362400615542447e-06, + "loss": 0.428, + "step": 72786 + }, + { + "epoch": 2.8, + "learning_rate": 3.3617594254937163e-06, + "loss": 0.2366, + "step": 72787 + }, + { + "epoch": 2.8, + "learning_rate": 3.3611182354449863e-06, + "loss": 0.177, + "step": 72788 + }, + { + "epoch": 2.8, + "learning_rate": 3.3604770453962555e-06, + "loss": 0.2546, + "step": 72789 + }, + { + "epoch": 2.8, + "learning_rate": 3.3598358553475255e-06, + "loss": 0.1376, + "step": 72790 + }, + { + "epoch": 2.8, + "learning_rate": 3.3591946652987947e-06, + "loss": 0.1047, + "step": 72791 + }, + { + "epoch": 2.8, + "learning_rate": 3.358553475250064e-06, + "loss": 0.2513, + "step": 72792 + }, + { + "epoch": 2.8, + "learning_rate": 3.357912285201334e-06, + "loss": 0.209, + "step": 72793 + }, + { + "epoch": 2.8, + "learning_rate": 3.357271095152603e-06, + "loss": 0.1544, + "step": 72794 + }, + { + "epoch": 2.8, + "learning_rate": 3.356629905103873e-06, + "loss": 0.1986, + "step": 72795 + }, + { + "epoch": 2.8, + "learning_rate": 3.3559887150551422e-06, + "loss": 0.1202, + "step": 72796 + }, + { + "epoch": 2.8, + "learning_rate": 3.3553475250064123e-06, + "loss": 0.2462, + "step": 72797 + }, + { + "epoch": 2.8, + "learning_rate": 3.3547063349576814e-06, + "loss": 0.4153, + "step": 72798 + }, + { + "epoch": 2.8, + "learning_rate": 3.3540651449089515e-06, + "loss": 0.371, + "step": 72799 + }, + { + "epoch": 2.8, + "learning_rate": 3.3534239548602206e-06, + "loss": 0.1485, + "step": 72800 + }, + { + "epoch": 2.8, + "learning_rate": 3.35278276481149e-06, + "loss": 0.2575, + "step": 72801 + }, + { + "epoch": 2.8, + "learning_rate": 3.35214157476276e-06, + "loss": 0.1894, + "step": 72802 + }, + { + "epoch": 2.8, + "learning_rate": 3.351500384714029e-06, + "loss": 0.1405, + "step": 72803 + }, + { + "epoch": 2.8, + "learning_rate": 3.350859194665299e-06, + "loss": 0.1595, + "step": 72804 + }, + { + "epoch": 2.8, + "learning_rate": 3.350218004616568e-06, + "loss": 0.1805, + "step": 72805 + }, + { + "epoch": 2.8, + "learning_rate": 3.3495768145678382e-06, + "loss": 0.1374, + "step": 72806 + }, + { + "epoch": 2.8, + "learning_rate": 3.3489356245191074e-06, + "loss": 0.2358, + "step": 72807 + }, + { + "epoch": 2.8, + "learning_rate": 3.3482944344703774e-06, + "loss": 0.1989, + "step": 72808 + }, + { + "epoch": 2.8, + "learning_rate": 3.3476532444216466e-06, + "loss": 0.113, + "step": 72809 + }, + { + "epoch": 2.8, + "learning_rate": 3.3470120543729158e-06, + "loss": 0.2568, + "step": 72810 + }, + { + "epoch": 2.8, + "learning_rate": 3.346370864324186e-06, + "loss": 0.1869, + "step": 72811 + }, + { + "epoch": 2.8, + "learning_rate": 3.345729674275455e-06, + "loss": 0.1534, + "step": 72812 + }, + { + "epoch": 2.8, + "learning_rate": 3.345088484226725e-06, + "loss": 0.2532, + "step": 72813 + }, + { + "epoch": 2.8, + "learning_rate": 3.344447294177994e-06, + "loss": 0.1526, + "step": 72814 + }, + { + "epoch": 2.8, + "learning_rate": 3.343806104129264e-06, + "loss": 0.1675, + "step": 72815 + }, + { + "epoch": 2.8, + "learning_rate": 3.3431649140805334e-06, + "loss": 0.1533, + "step": 72816 + }, + { + "epoch": 2.8, + "learning_rate": 3.3425237240318034e-06, + "loss": 0.1841, + "step": 72817 + }, + { + "epoch": 2.8, + "learning_rate": 3.3418825339830726e-06, + "loss": 0.2591, + "step": 72818 + }, + { + "epoch": 2.8, + "learning_rate": 3.3412413439343426e-06, + "loss": 0.1123, + "step": 72819 + }, + { + "epoch": 2.8, + "learning_rate": 3.3406001538856118e-06, + "loss": 0.0889, + "step": 72820 + }, + { + "epoch": 2.8, + "learning_rate": 3.339958963836881e-06, + "loss": 0.2576, + "step": 72821 + }, + { + "epoch": 2.8, + "learning_rate": 3.339317773788151e-06, + "loss": 0.1583, + "step": 72822 + }, + { + "epoch": 2.8, + "learning_rate": 3.33867658373942e-06, + "loss": 0.2593, + "step": 72823 + }, + { + "epoch": 2.8, + "learning_rate": 3.33803539369069e-06, + "loss": 0.2202, + "step": 72824 + }, + { + "epoch": 2.8, + "learning_rate": 3.3373942036419593e-06, + "loss": 0.2102, + "step": 72825 + }, + { + "epoch": 2.8, + "learning_rate": 3.3367530135932294e-06, + "loss": 0.1935, + "step": 72826 + }, + { + "epoch": 2.8, + "learning_rate": 3.3361118235444985e-06, + "loss": 0.2951, + "step": 72827 + }, + { + "epoch": 2.8, + "learning_rate": 3.3354706334957686e-06, + "loss": 0.1765, + "step": 72828 + }, + { + "epoch": 2.8, + "learning_rate": 3.3348294434470377e-06, + "loss": 0.2072, + "step": 72829 + }, + { + "epoch": 2.8, + "learning_rate": 3.334188253398307e-06, + "loss": 0.2547, + "step": 72830 + }, + { + "epoch": 2.8, + "learning_rate": 3.333547063349577e-06, + "loss": 0.2252, + "step": 72831 + }, + { + "epoch": 2.8, + "learning_rate": 3.332905873300846e-06, + "loss": 0.3686, + "step": 72832 + }, + { + "epoch": 2.8, + "learning_rate": 3.332264683252116e-06, + "loss": 0.2903, + "step": 72833 + }, + { + "epoch": 2.8, + "learning_rate": 3.3316234932033853e-06, + "loss": 0.2133, + "step": 72834 + }, + { + "epoch": 2.8, + "learning_rate": 3.3309823031546553e-06, + "loss": 0.1721, + "step": 72835 + }, + { + "epoch": 2.8, + "learning_rate": 3.3303411131059245e-06, + "loss": 0.1545, + "step": 72836 + }, + { + "epoch": 2.8, + "learning_rate": 3.3296999230571945e-06, + "loss": 0.2424, + "step": 72837 + }, + { + "epoch": 2.8, + "learning_rate": 3.3290587330084637e-06, + "loss": 0.1538, + "step": 72838 + }, + { + "epoch": 2.8, + "learning_rate": 3.3284175429597337e-06, + "loss": 0.307, + "step": 72839 + }, + { + "epoch": 2.8, + "learning_rate": 3.327776352911003e-06, + "loss": 0.1519, + "step": 72840 + }, + { + "epoch": 2.8, + "learning_rate": 3.327135162862272e-06, + "loss": 0.2447, + "step": 72841 + }, + { + "epoch": 2.8, + "learning_rate": 3.326493972813542e-06, + "loss": 0.2126, + "step": 72842 + }, + { + "epoch": 2.8, + "learning_rate": 3.3258527827648113e-06, + "loss": 0.183, + "step": 72843 + }, + { + "epoch": 2.8, + "learning_rate": 3.3252115927160813e-06, + "loss": 0.276, + "step": 72844 + }, + { + "epoch": 2.8, + "learning_rate": 3.3245704026673505e-06, + "loss": 0.1959, + "step": 72845 + }, + { + "epoch": 2.8, + "learning_rate": 3.3239292126186205e-06, + "loss": 0.0997, + "step": 72846 + }, + { + "epoch": 2.8, + "learning_rate": 3.3232880225698897e-06, + "loss": 0.311, + "step": 72847 + }, + { + "epoch": 2.8, + "learning_rate": 3.3226468325211597e-06, + "loss": 0.1902, + "step": 72848 + }, + { + "epoch": 2.8, + "learning_rate": 3.322005642472429e-06, + "loss": 0.1964, + "step": 72849 + }, + { + "epoch": 2.8, + "learning_rate": 3.321364452423698e-06, + "loss": 0.193, + "step": 72850 + }, + { + "epoch": 2.8, + "learning_rate": 3.320723262374968e-06, + "loss": 0.188, + "step": 72851 + }, + { + "epoch": 2.8, + "learning_rate": 3.3200820723262372e-06, + "loss": 0.1805, + "step": 72852 + }, + { + "epoch": 2.8, + "learning_rate": 3.3194408822775073e-06, + "loss": 0.1916, + "step": 72853 + }, + { + "epoch": 2.8, + "learning_rate": 3.3187996922287764e-06, + "loss": 0.2524, + "step": 72854 + }, + { + "epoch": 2.8, + "learning_rate": 3.3181585021800465e-06, + "loss": 0.248, + "step": 72855 + }, + { + "epoch": 2.8, + "learning_rate": 3.3175173121313156e-06, + "loss": 0.1666, + "step": 72856 + }, + { + "epoch": 2.8, + "learning_rate": 3.3168761220825857e-06, + "loss": 0.2159, + "step": 72857 + }, + { + "epoch": 2.8, + "learning_rate": 3.316234932033855e-06, + "loss": 0.2449, + "step": 72858 + }, + { + "epoch": 2.8, + "learning_rate": 3.315593741985124e-06, + "loss": 0.1882, + "step": 72859 + }, + { + "epoch": 2.8, + "learning_rate": 3.314952551936394e-06, + "loss": 0.16, + "step": 72860 + }, + { + "epoch": 2.8, + "learning_rate": 3.314311361887663e-06, + "loss": 0.2619, + "step": 72861 + }, + { + "epoch": 2.8, + "learning_rate": 3.3136701718389332e-06, + "loss": 0.2337, + "step": 72862 + }, + { + "epoch": 2.8, + "learning_rate": 3.3130289817902024e-06, + "loss": 0.1849, + "step": 72863 + }, + { + "epoch": 2.8, + "learning_rate": 3.3123877917414724e-06, + "loss": 0.3018, + "step": 72864 + }, + { + "epoch": 2.8, + "learning_rate": 3.3117466016927416e-06, + "loss": 0.1704, + "step": 72865 + }, + { + "epoch": 2.8, + "learning_rate": 3.3111054116440116e-06, + "loss": 0.1197, + "step": 72866 + }, + { + "epoch": 2.8, + "learning_rate": 3.310464221595281e-06, + "loss": 0.2767, + "step": 72867 + }, + { + "epoch": 2.8, + "learning_rate": 3.309823031546551e-06, + "loss": 0.2035, + "step": 72868 + }, + { + "epoch": 2.8, + "learning_rate": 3.30918184149782e-06, + "loss": 0.0935, + "step": 72869 + }, + { + "epoch": 2.8, + "learning_rate": 3.308540651449089e-06, + "loss": 0.2946, + "step": 72870 + }, + { + "epoch": 2.8, + "learning_rate": 3.307899461400359e-06, + "loss": 0.2488, + "step": 72871 + }, + { + "epoch": 2.8, + "learning_rate": 3.3072582713516284e-06, + "loss": 0.1273, + "step": 72872 + }, + { + "epoch": 2.8, + "learning_rate": 3.3066170813028984e-06, + "loss": 0.3233, + "step": 72873 + }, + { + "epoch": 2.8, + "learning_rate": 3.3059758912541676e-06, + "loss": 0.2063, + "step": 72874 + }, + { + "epoch": 2.8, + "learning_rate": 3.3053347012054376e-06, + "loss": 0.2779, + "step": 72875 + }, + { + "epoch": 2.8, + "learning_rate": 3.3046935111567068e-06, + "loss": 0.22, + "step": 72876 + }, + { + "epoch": 2.8, + "learning_rate": 3.304052321107977e-06, + "loss": 0.1397, + "step": 72877 + }, + { + "epoch": 2.8, + "learning_rate": 3.303411131059246e-06, + "loss": 0.1508, + "step": 72878 + }, + { + "epoch": 2.8, + "learning_rate": 3.302769941010515e-06, + "loss": 0.2753, + "step": 72879 + }, + { + "epoch": 2.8, + "learning_rate": 3.302128750961785e-06, + "loss": 0.187, + "step": 72880 + }, + { + "epoch": 2.8, + "learning_rate": 3.3014875609130543e-06, + "loss": 0.3016, + "step": 72881 + }, + { + "epoch": 2.8, + "learning_rate": 3.3008463708643244e-06, + "loss": 0.1767, + "step": 72882 + }, + { + "epoch": 2.8, + "learning_rate": 3.3002051808155935e-06, + "loss": 0.1717, + "step": 72883 + }, + { + "epoch": 2.8, + "learning_rate": 3.2995639907668636e-06, + "loss": 0.1136, + "step": 72884 + }, + { + "epoch": 2.8, + "learning_rate": 3.2989228007181327e-06, + "loss": 0.3132, + "step": 72885 + }, + { + "epoch": 2.8, + "learning_rate": 3.2982816106694028e-06, + "loss": 0.1767, + "step": 72886 + }, + { + "epoch": 2.8, + "learning_rate": 3.297640420620672e-06, + "loss": 0.324, + "step": 72887 + }, + { + "epoch": 2.8, + "learning_rate": 3.296999230571942e-06, + "loss": 0.2619, + "step": 72888 + }, + { + "epoch": 2.8, + "learning_rate": 3.296358040523211e-06, + "loss": 0.217, + "step": 72889 + }, + { + "epoch": 2.8, + "learning_rate": 3.2957168504744803e-06, + "loss": 0.2414, + "step": 72890 + }, + { + "epoch": 2.8, + "learning_rate": 3.2950756604257503e-06, + "loss": 0.1336, + "step": 72891 + }, + { + "epoch": 2.8, + "learning_rate": 3.2944344703770195e-06, + "loss": 0.1575, + "step": 72892 + }, + { + "epoch": 2.8, + "learning_rate": 3.2937932803282895e-06, + "loss": 0.1778, + "step": 72893 + }, + { + "epoch": 2.8, + "learning_rate": 3.2931520902795587e-06, + "loss": 0.1498, + "step": 72894 + }, + { + "epoch": 2.8, + "learning_rate": 3.2925109002308287e-06, + "loss": 0.2961, + "step": 72895 + }, + { + "epoch": 2.8, + "learning_rate": 3.291869710182098e-06, + "loss": 0.1576, + "step": 72896 + }, + { + "epoch": 2.8, + "learning_rate": 3.291228520133368e-06, + "loss": 0.3842, + "step": 72897 + }, + { + "epoch": 2.8, + "learning_rate": 3.290587330084637e-06, + "loss": 0.1575, + "step": 72898 + }, + { + "epoch": 2.8, + "learning_rate": 3.2899461400359063e-06, + "loss": 0.1608, + "step": 72899 + }, + { + "epoch": 2.8, + "learning_rate": 3.2893049499871763e-06, + "loss": 0.1158, + "step": 72900 + }, + { + "epoch": 2.8, + "learning_rate": 3.2886637599384455e-06, + "loss": 0.0968, + "step": 72901 + }, + { + "epoch": 2.8, + "learning_rate": 3.2880225698897155e-06, + "loss": 0.185, + "step": 72902 + }, + { + "epoch": 2.8, + "learning_rate": 3.2873813798409847e-06, + "loss": 0.1759, + "step": 72903 + }, + { + "epoch": 2.8, + "learning_rate": 3.2867401897922547e-06, + "loss": 0.1881, + "step": 72904 + }, + { + "epoch": 2.8, + "learning_rate": 3.286098999743524e-06, + "loss": 0.1761, + "step": 72905 + }, + { + "epoch": 2.8, + "learning_rate": 3.285457809694794e-06, + "loss": 0.1734, + "step": 72906 + }, + { + "epoch": 2.8, + "learning_rate": 3.284816619646063e-06, + "loss": 0.2166, + "step": 72907 + }, + { + "epoch": 2.8, + "learning_rate": 3.284175429597333e-06, + "loss": 0.275, + "step": 72908 + }, + { + "epoch": 2.8, + "learning_rate": 3.2835342395486023e-06, + "loss": 0.2468, + "step": 72909 + }, + { + "epoch": 2.8, + "learning_rate": 3.2828930494998714e-06, + "loss": 0.2341, + "step": 72910 + }, + { + "epoch": 2.8, + "learning_rate": 3.2822518594511415e-06, + "loss": 0.1737, + "step": 72911 + }, + { + "epoch": 2.8, + "learning_rate": 3.2816106694024106e-06, + "loss": 0.2874, + "step": 72912 + }, + { + "epoch": 2.8, + "learning_rate": 3.2809694793536807e-06, + "loss": 0.0718, + "step": 72913 + }, + { + "epoch": 2.8, + "learning_rate": 3.28032828930495e-06, + "loss": 0.2341, + "step": 72914 + }, + { + "epoch": 2.8, + "learning_rate": 3.27968709925622e-06, + "loss": 0.1393, + "step": 72915 + }, + { + "epoch": 2.8, + "learning_rate": 3.279045909207489e-06, + "loss": 0.2458, + "step": 72916 + }, + { + "epoch": 2.8, + "learning_rate": 3.278404719158759e-06, + "loss": 0.252, + "step": 72917 + }, + { + "epoch": 2.8, + "learning_rate": 3.2777635291100282e-06, + "loss": 0.2722, + "step": 72918 + }, + { + "epoch": 2.8, + "learning_rate": 3.2771223390612974e-06, + "loss": 0.257, + "step": 72919 + }, + { + "epoch": 2.8, + "learning_rate": 3.2764811490125674e-06, + "loss": 0.1277, + "step": 72920 + }, + { + "epoch": 2.8, + "learning_rate": 3.2758399589638366e-06, + "loss": 0.2118, + "step": 72921 + }, + { + "epoch": 2.8, + "learning_rate": 3.2751987689151066e-06, + "loss": 0.2491, + "step": 72922 + }, + { + "epoch": 2.8, + "learning_rate": 3.274557578866376e-06, + "loss": 0.1526, + "step": 72923 + }, + { + "epoch": 2.8, + "learning_rate": 3.273916388817646e-06, + "loss": 0.2793, + "step": 72924 + }, + { + "epoch": 2.8, + "learning_rate": 3.273275198768915e-06, + "loss": 0.2279, + "step": 72925 + }, + { + "epoch": 2.8, + "learning_rate": 3.272634008720185e-06, + "loss": 0.198, + "step": 72926 + }, + { + "epoch": 2.8, + "learning_rate": 3.271992818671454e-06, + "loss": 0.2727, + "step": 72927 + }, + { + "epoch": 2.8, + "learning_rate": 3.271351628622724e-06, + "loss": 0.2238, + "step": 72928 + }, + { + "epoch": 2.8, + "learning_rate": 3.2707104385739934e-06, + "loss": 0.2733, + "step": 72929 + }, + { + "epoch": 2.8, + "learning_rate": 3.2700692485252626e-06, + "loss": 0.2272, + "step": 72930 + }, + { + "epoch": 2.8, + "learning_rate": 3.2694280584765326e-06, + "loss": 0.3047, + "step": 72931 + }, + { + "epoch": 2.8, + "learning_rate": 3.2687868684278018e-06, + "loss": 0.211, + "step": 72932 + }, + { + "epoch": 2.8, + "learning_rate": 3.268145678379072e-06, + "loss": 0.2552, + "step": 72933 + }, + { + "epoch": 2.8, + "learning_rate": 3.267504488330341e-06, + "loss": 0.1231, + "step": 72934 + }, + { + "epoch": 2.8, + "learning_rate": 3.266863298281611e-06, + "loss": 0.1993, + "step": 72935 + }, + { + "epoch": 2.8, + "learning_rate": 3.26622210823288e-06, + "loss": 0.1748, + "step": 72936 + }, + { + "epoch": 2.8, + "learning_rate": 3.26558091818415e-06, + "loss": 0.1527, + "step": 72937 + }, + { + "epoch": 2.8, + "learning_rate": 3.2649397281354194e-06, + "loss": 0.1339, + "step": 72938 + }, + { + "epoch": 2.8, + "learning_rate": 3.2642985380866885e-06, + "loss": 0.2714, + "step": 72939 + }, + { + "epoch": 2.8, + "learning_rate": 3.2636573480379586e-06, + "loss": 0.1741, + "step": 72940 + }, + { + "epoch": 2.8, + "learning_rate": 3.2630161579892277e-06, + "loss": 0.2564, + "step": 72941 + }, + { + "epoch": 2.8, + "learning_rate": 3.2623749679404978e-06, + "loss": 0.1727, + "step": 72942 + }, + { + "epoch": 2.8, + "learning_rate": 3.261733777891767e-06, + "loss": 0.2341, + "step": 72943 + }, + { + "epoch": 2.8, + "learning_rate": 3.261092587843037e-06, + "loss": 0.1599, + "step": 72944 + }, + { + "epoch": 2.8, + "learning_rate": 3.260451397794306e-06, + "loss": 0.2218, + "step": 72945 + }, + { + "epoch": 2.8, + "learning_rate": 3.259810207745576e-06, + "loss": 0.2098, + "step": 72946 + }, + { + "epoch": 2.8, + "learning_rate": 3.2591690176968453e-06, + "loss": 0.1613, + "step": 72947 + }, + { + "epoch": 2.8, + "learning_rate": 3.258527827648115e-06, + "loss": 0.1268, + "step": 72948 + }, + { + "epoch": 2.8, + "learning_rate": 3.2578866375993845e-06, + "loss": 0.1, + "step": 72949 + }, + { + "epoch": 2.8, + "learning_rate": 3.2572454475506537e-06, + "loss": 0.3298, + "step": 72950 + }, + { + "epoch": 2.8, + "learning_rate": 3.2566042575019237e-06, + "loss": 0.1263, + "step": 72951 + }, + { + "epoch": 2.8, + "learning_rate": 3.255963067453193e-06, + "loss": 0.1716, + "step": 72952 + }, + { + "epoch": 2.8, + "learning_rate": 3.255321877404463e-06, + "loss": 0.3257, + "step": 72953 + }, + { + "epoch": 2.8, + "learning_rate": 3.254680687355732e-06, + "loss": 0.1704, + "step": 72954 + }, + { + "epoch": 2.8, + "learning_rate": 3.254039497307002e-06, + "loss": 0.1993, + "step": 72955 + }, + { + "epoch": 2.8, + "learning_rate": 3.2533983072582713e-06, + "loss": 0.2085, + "step": 72956 + }, + { + "epoch": 2.8, + "learning_rate": 3.2527571172095413e-06, + "loss": 0.1167, + "step": 72957 + }, + { + "epoch": 2.8, + "learning_rate": 3.2521159271608105e-06, + "loss": 0.2634, + "step": 72958 + }, + { + "epoch": 2.81, + "learning_rate": 3.25147473711208e-06, + "loss": 0.1451, + "step": 72959 + }, + { + "epoch": 2.81, + "learning_rate": 3.2508335470633497e-06, + "loss": 0.26, + "step": 72960 + }, + { + "epoch": 2.81, + "learning_rate": 3.250192357014619e-06, + "loss": 0.1841, + "step": 72961 + }, + { + "epoch": 2.81, + "learning_rate": 3.249551166965889e-06, + "loss": 0.2053, + "step": 72962 + }, + { + "epoch": 2.81, + "learning_rate": 3.248909976917158e-06, + "loss": 0.1989, + "step": 72963 + }, + { + "epoch": 2.81, + "learning_rate": 3.248268786868428e-06, + "loss": 0.2483, + "step": 72964 + }, + { + "epoch": 2.81, + "learning_rate": 3.2476275968196973e-06, + "loss": 0.1602, + "step": 72965 + }, + { + "epoch": 2.81, + "learning_rate": 3.2469864067709673e-06, + "loss": 0.2659, + "step": 72966 + }, + { + "epoch": 2.81, + "learning_rate": 3.2463452167222365e-06, + "loss": 0.3027, + "step": 72967 + }, + { + "epoch": 2.81, + "learning_rate": 3.245704026673506e-06, + "loss": 0.2538, + "step": 72968 + }, + { + "epoch": 2.81, + "learning_rate": 3.2450628366247757e-06, + "loss": 0.1757, + "step": 72969 + }, + { + "epoch": 2.81, + "learning_rate": 3.244421646576045e-06, + "loss": 0.1477, + "step": 72970 + }, + { + "epoch": 2.81, + "learning_rate": 3.243780456527315e-06, + "loss": 0.1834, + "step": 72971 + }, + { + "epoch": 2.81, + "learning_rate": 3.243139266478584e-06, + "loss": 0.2271, + "step": 72972 + }, + { + "epoch": 2.81, + "learning_rate": 3.242498076429854e-06, + "loss": 0.1651, + "step": 72973 + }, + { + "epoch": 2.81, + "learning_rate": 3.2418568863811232e-06, + "loss": 0.3126, + "step": 72974 + }, + { + "epoch": 2.81, + "learning_rate": 3.2412156963323933e-06, + "loss": 0.1658, + "step": 72975 + }, + { + "epoch": 2.81, + "learning_rate": 3.2405745062836624e-06, + "loss": 0.2586, + "step": 72976 + }, + { + "epoch": 2.81, + "learning_rate": 3.239933316234932e-06, + "loss": 0.1495, + "step": 72977 + }, + { + "epoch": 2.81, + "learning_rate": 3.2392921261862016e-06, + "loss": 0.2468, + "step": 72978 + }, + { + "epoch": 2.81, + "learning_rate": 3.2386509361374712e-06, + "loss": 0.1186, + "step": 72979 + }, + { + "epoch": 2.81, + "learning_rate": 3.238009746088741e-06, + "loss": 0.1507, + "step": 72980 + }, + { + "epoch": 2.81, + "learning_rate": 3.23736855604001e-06, + "loss": 0.1465, + "step": 72981 + }, + { + "epoch": 2.81, + "learning_rate": 3.23672736599128e-06, + "loss": 0.1211, + "step": 72982 + }, + { + "epoch": 2.81, + "learning_rate": 3.236086175942549e-06, + "loss": 0.2448, + "step": 72983 + }, + { + "epoch": 2.81, + "learning_rate": 3.2354449858938192e-06, + "loss": 0.1985, + "step": 72984 + }, + { + "epoch": 2.81, + "learning_rate": 3.2348037958450884e-06, + "loss": 0.1787, + "step": 72985 + }, + { + "epoch": 2.81, + "learning_rate": 3.2341626057963584e-06, + "loss": 0.2351, + "step": 72986 + }, + { + "epoch": 2.81, + "learning_rate": 3.2335214157476276e-06, + "loss": 0.1466, + "step": 72987 + }, + { + "epoch": 2.81, + "learning_rate": 3.232880225698897e-06, + "loss": 0.2316, + "step": 72988 + }, + { + "epoch": 2.81, + "learning_rate": 3.232239035650167e-06, + "loss": 0.2138, + "step": 72989 + }, + { + "epoch": 2.81, + "learning_rate": 3.2315978456014364e-06, + "loss": 0.1577, + "step": 72990 + }, + { + "epoch": 2.81, + "learning_rate": 3.230956655552706e-06, + "loss": 0.3016, + "step": 72991 + }, + { + "epoch": 2.81, + "learning_rate": 3.230315465503975e-06, + "loss": 0.1601, + "step": 72992 + }, + { + "epoch": 2.81, + "learning_rate": 3.229674275455245e-06, + "loss": 0.2994, + "step": 72993 + }, + { + "epoch": 2.81, + "learning_rate": 3.2290330854065144e-06, + "loss": 0.1265, + "step": 72994 + }, + { + "epoch": 2.81, + "learning_rate": 3.2283918953577844e-06, + "loss": 0.2378, + "step": 72995 + }, + { + "epoch": 2.81, + "learning_rate": 3.2277507053090536e-06, + "loss": 0.1819, + "step": 72996 + }, + { + "epoch": 2.81, + "learning_rate": 3.227109515260323e-06, + "loss": 0.2344, + "step": 72997 + }, + { + "epoch": 2.81, + "learning_rate": 3.2264683252115928e-06, + "loss": 0.2359, + "step": 72998 + }, + { + "epoch": 2.81, + "learning_rate": 3.2258271351628624e-06, + "loss": 0.2351, + "step": 72999 + }, + { + "epoch": 2.81, + "learning_rate": 3.225185945114132e-06, + "loss": 0.1884, + "step": 73000 + }, + { + "epoch": 2.81, + "learning_rate": 3.224544755065401e-06, + "loss": 0.2596, + "step": 73001 + }, + { + "epoch": 2.81, + "learning_rate": 3.223903565016671e-06, + "loss": 0.2281, + "step": 73002 + }, + { + "epoch": 2.81, + "learning_rate": 3.2232623749679403e-06, + "loss": 0.2497, + "step": 73003 + }, + { + "epoch": 2.81, + "learning_rate": 3.2226211849192104e-06, + "loss": 0.3023, + "step": 73004 + }, + { + "epoch": 2.81, + "learning_rate": 3.2219799948704795e-06, + "loss": 0.2703, + "step": 73005 + }, + { + "epoch": 2.81, + "learning_rate": 3.2213388048217496e-06, + "loss": 0.2281, + "step": 73006 + }, + { + "epoch": 2.81, + "learning_rate": 3.2206976147730187e-06, + "loss": 0.1859, + "step": 73007 + }, + { + "epoch": 2.81, + "learning_rate": 3.2200564247242883e-06, + "loss": 0.1619, + "step": 73008 + }, + { + "epoch": 2.81, + "learning_rate": 3.219415234675558e-06, + "loss": 0.193, + "step": 73009 + }, + { + "epoch": 2.81, + "learning_rate": 3.2187740446268275e-06, + "loss": 0.2778, + "step": 73010 + }, + { + "epoch": 2.81, + "learning_rate": 3.218132854578097e-06, + "loss": 0.3205, + "step": 73011 + }, + { + "epoch": 2.81, + "learning_rate": 3.2174916645293663e-06, + "loss": 0.1775, + "step": 73012 + }, + { + "epoch": 2.81, + "learning_rate": 3.2168504744806363e-06, + "loss": 0.3324, + "step": 73013 + }, + { + "epoch": 2.81, + "learning_rate": 3.2162092844319055e-06, + "loss": 0.2379, + "step": 73014 + }, + { + "epoch": 2.81, + "learning_rate": 3.2155680943831755e-06, + "loss": 0.2074, + "step": 73015 + }, + { + "epoch": 2.81, + "learning_rate": 3.2149269043344447e-06, + "loss": 0.2492, + "step": 73016 + }, + { + "epoch": 2.81, + "learning_rate": 3.2142857142857143e-06, + "loss": 0.1944, + "step": 73017 + }, + { + "epoch": 2.81, + "learning_rate": 3.213644524236984e-06, + "loss": 0.2042, + "step": 73018 + }, + { + "epoch": 2.81, + "learning_rate": 3.2130033341882535e-06, + "loss": 0.1052, + "step": 73019 + }, + { + "epoch": 2.81, + "learning_rate": 3.212362144139523e-06, + "loss": 0.5067, + "step": 73020 + }, + { + "epoch": 2.81, + "learning_rate": 3.2117209540907923e-06, + "loss": 0.2661, + "step": 73021 + }, + { + "epoch": 2.81, + "learning_rate": 3.2110797640420623e-06, + "loss": 0.0936, + "step": 73022 + }, + { + "epoch": 2.81, + "learning_rate": 3.2104385739933315e-06, + "loss": 0.1755, + "step": 73023 + }, + { + "epoch": 2.81, + "learning_rate": 3.2097973839446015e-06, + "loss": 0.2064, + "step": 73024 + }, + { + "epoch": 2.81, + "learning_rate": 3.2091561938958707e-06, + "loss": 0.2301, + "step": 73025 + }, + { + "epoch": 2.81, + "learning_rate": 3.2085150038471403e-06, + "loss": 0.2468, + "step": 73026 + }, + { + "epoch": 2.81, + "learning_rate": 3.20787381379841e-06, + "loss": 0.209, + "step": 73027 + }, + { + "epoch": 2.81, + "learning_rate": 3.2072326237496795e-06, + "loss": 0.1466, + "step": 73028 + }, + { + "epoch": 2.81, + "learning_rate": 3.206591433700949e-06, + "loss": 0.1487, + "step": 73029 + }, + { + "epoch": 2.81, + "learning_rate": 3.2059502436522187e-06, + "loss": 0.3141, + "step": 73030 + }, + { + "epoch": 2.81, + "learning_rate": 3.2053090536034883e-06, + "loss": 0.2793, + "step": 73031 + }, + { + "epoch": 2.81, + "learning_rate": 3.2046678635547574e-06, + "loss": 0.1629, + "step": 73032 + }, + { + "epoch": 2.81, + "learning_rate": 3.2040266735060275e-06, + "loss": 0.2973, + "step": 73033 + }, + { + "epoch": 2.81, + "learning_rate": 3.2033854834572966e-06, + "loss": 0.1527, + "step": 73034 + }, + { + "epoch": 2.81, + "learning_rate": 3.2027442934085667e-06, + "loss": 0.2785, + "step": 73035 + }, + { + "epoch": 2.81, + "learning_rate": 3.202103103359836e-06, + "loss": 0.2558, + "step": 73036 + }, + { + "epoch": 2.81, + "learning_rate": 3.2014619133111054e-06, + "loss": 0.2928, + "step": 73037 + }, + { + "epoch": 2.81, + "learning_rate": 3.200820723262375e-06, + "loss": 0.1843, + "step": 73038 + }, + { + "epoch": 2.81, + "learning_rate": 3.2001795332136446e-06, + "loss": 0.3083, + "step": 73039 + }, + { + "epoch": 2.81, + "learning_rate": 3.1995383431649142e-06, + "loss": 0.2217, + "step": 73040 + }, + { + "epoch": 2.81, + "learning_rate": 3.198897153116184e-06, + "loss": 0.4601, + "step": 73041 + }, + { + "epoch": 2.81, + "learning_rate": 3.1982559630674534e-06, + "loss": 0.167, + "step": 73042 + }, + { + "epoch": 2.81, + "learning_rate": 3.1976147730187226e-06, + "loss": 0.165, + "step": 73043 + }, + { + "epoch": 2.81, + "learning_rate": 3.1969735829699926e-06, + "loss": 0.1741, + "step": 73044 + }, + { + "epoch": 2.81, + "learning_rate": 3.196332392921262e-06, + "loss": 0.2609, + "step": 73045 + }, + { + "epoch": 2.81, + "learning_rate": 3.1956912028725314e-06, + "loss": 0.0892, + "step": 73046 + }, + { + "epoch": 2.81, + "learning_rate": 3.195050012823801e-06, + "loss": 0.3441, + "step": 73047 + }, + { + "epoch": 2.81, + "learning_rate": 3.1944088227750706e-06, + "loss": 0.217, + "step": 73048 + }, + { + "epoch": 2.81, + "learning_rate": 3.19376763272634e-06, + "loss": 0.186, + "step": 73049 + }, + { + "epoch": 2.81, + "learning_rate": 3.19312644267761e-06, + "loss": 0.1534, + "step": 73050 + }, + { + "epoch": 2.81, + "learning_rate": 3.1924852526288794e-06, + "loss": 0.2405, + "step": 73051 + }, + { + "epoch": 2.81, + "learning_rate": 3.1918440625801486e-06, + "loss": 0.1745, + "step": 73052 + }, + { + "epoch": 2.81, + "learning_rate": 3.1912028725314186e-06, + "loss": 0.0983, + "step": 73053 + }, + { + "epoch": 2.81, + "learning_rate": 3.1905616824826878e-06, + "loss": 0.1351, + "step": 73054 + }, + { + "epoch": 2.81, + "learning_rate": 3.189920492433958e-06, + "loss": 0.1196, + "step": 73055 + }, + { + "epoch": 2.81, + "learning_rate": 3.189279302385227e-06, + "loss": 0.2147, + "step": 73056 + }, + { + "epoch": 2.81, + "learning_rate": 3.1886381123364966e-06, + "loss": 0.1824, + "step": 73057 + }, + { + "epoch": 2.81, + "learning_rate": 3.187996922287766e-06, + "loss": 0.501, + "step": 73058 + }, + { + "epoch": 2.81, + "learning_rate": 3.1873557322390358e-06, + "loss": 0.1738, + "step": 73059 + }, + { + "epoch": 2.81, + "learning_rate": 3.1867145421903054e-06, + "loss": 0.237, + "step": 73060 + }, + { + "epoch": 2.81, + "learning_rate": 3.186073352141575e-06, + "loss": 0.2845, + "step": 73061 + }, + { + "epoch": 2.81, + "learning_rate": 3.1854321620928446e-06, + "loss": 0.2607, + "step": 73062 + }, + { + "epoch": 2.81, + "learning_rate": 3.1847909720441137e-06, + "loss": 0.255, + "step": 73063 + }, + { + "epoch": 2.81, + "learning_rate": 3.1841497819953838e-06, + "loss": 0.1846, + "step": 73064 + }, + { + "epoch": 2.81, + "learning_rate": 3.183508591946653e-06, + "loss": 0.1834, + "step": 73065 + }, + { + "epoch": 2.81, + "learning_rate": 3.1828674018979225e-06, + "loss": 0.1685, + "step": 73066 + }, + { + "epoch": 2.81, + "learning_rate": 3.182226211849192e-06, + "loss": 0.2068, + "step": 73067 + }, + { + "epoch": 2.81, + "learning_rate": 3.1815850218004617e-06, + "loss": 0.1732, + "step": 73068 + }, + { + "epoch": 2.81, + "learning_rate": 3.1809438317517313e-06, + "loss": 0.2081, + "step": 73069 + }, + { + "epoch": 2.81, + "learning_rate": 3.180302641703001e-06, + "loss": 0.1828, + "step": 73070 + }, + { + "epoch": 2.81, + "learning_rate": 3.1796614516542705e-06, + "loss": 0.305, + "step": 73071 + }, + { + "epoch": 2.81, + "learning_rate": 3.1790202616055397e-06, + "loss": 0.2168, + "step": 73072 + }, + { + "epoch": 2.81, + "learning_rate": 3.1783790715568097e-06, + "loss": 0.2306, + "step": 73073 + }, + { + "epoch": 2.81, + "learning_rate": 3.177737881508079e-06, + "loss": 0.2246, + "step": 73074 + }, + { + "epoch": 2.81, + "learning_rate": 3.177096691459349e-06, + "loss": 0.3578, + "step": 73075 + }, + { + "epoch": 2.81, + "learning_rate": 3.176455501410618e-06, + "loss": 0.1873, + "step": 73076 + }, + { + "epoch": 2.81, + "learning_rate": 3.1758143113618877e-06, + "loss": 0.2698, + "step": 73077 + }, + { + "epoch": 2.81, + "learning_rate": 3.1751731213131573e-06, + "loss": 0.1753, + "step": 73078 + }, + { + "epoch": 2.81, + "learning_rate": 3.174531931264427e-06, + "loss": 0.1713, + "step": 73079 + }, + { + "epoch": 2.81, + "learning_rate": 3.1738907412156965e-06, + "loss": 0.2058, + "step": 73080 + }, + { + "epoch": 2.81, + "learning_rate": 3.173249551166966e-06, + "loss": 0.2805, + "step": 73081 + }, + { + "epoch": 2.81, + "learning_rate": 3.1726083611182357e-06, + "loss": 0.3506, + "step": 73082 + }, + { + "epoch": 2.81, + "learning_rate": 3.171967171069505e-06, + "loss": 0.203, + "step": 73083 + }, + { + "epoch": 2.81, + "learning_rate": 3.171325981020775e-06, + "loss": 0.1292, + "step": 73084 + }, + { + "epoch": 2.81, + "learning_rate": 3.170684790972044e-06, + "loss": 0.2481, + "step": 73085 + }, + { + "epoch": 2.81, + "learning_rate": 3.1700436009233137e-06, + "loss": 0.2097, + "step": 73086 + }, + { + "epoch": 2.81, + "learning_rate": 3.1694024108745833e-06, + "loss": 0.3088, + "step": 73087 + }, + { + "epoch": 2.81, + "learning_rate": 3.168761220825853e-06, + "loss": 0.2499, + "step": 73088 + }, + { + "epoch": 2.81, + "learning_rate": 3.1681200307771225e-06, + "loss": 0.1635, + "step": 73089 + }, + { + "epoch": 2.81, + "learning_rate": 3.167478840728392e-06, + "loss": 0.2559, + "step": 73090 + }, + { + "epoch": 2.81, + "learning_rate": 3.1668376506796617e-06, + "loss": 0.199, + "step": 73091 + }, + { + "epoch": 2.81, + "learning_rate": 3.1661964606309313e-06, + "loss": 0.1279, + "step": 73092 + }, + { + "epoch": 2.81, + "learning_rate": 3.165555270582201e-06, + "loss": 0.1936, + "step": 73093 + }, + { + "epoch": 2.81, + "learning_rate": 3.16491408053347e-06, + "loss": 0.2362, + "step": 73094 + }, + { + "epoch": 2.81, + "learning_rate": 3.1642728904847396e-06, + "loss": 0.208, + "step": 73095 + }, + { + "epoch": 2.81, + "learning_rate": 3.1636317004360092e-06, + "loss": 0.2856, + "step": 73096 + }, + { + "epoch": 2.81, + "learning_rate": 3.162990510387279e-06, + "loss": 0.186, + "step": 73097 + }, + { + "epoch": 2.81, + "learning_rate": 3.1623493203385484e-06, + "loss": 0.2388, + "step": 73098 + }, + { + "epoch": 2.81, + "learning_rate": 3.161708130289818e-06, + "loss": 0.1809, + "step": 73099 + }, + { + "epoch": 2.81, + "learning_rate": 3.1610669402410876e-06, + "loss": 0.1236, + "step": 73100 + }, + { + "epoch": 2.81, + "learning_rate": 3.1604257501923572e-06, + "loss": 0.2994, + "step": 73101 + }, + { + "epoch": 2.81, + "learning_rate": 3.159784560143627e-06, + "loss": 0.1405, + "step": 73102 + }, + { + "epoch": 2.81, + "learning_rate": 3.159143370094896e-06, + "loss": 0.1767, + "step": 73103 + }, + { + "epoch": 2.81, + "learning_rate": 3.158502180046166e-06, + "loss": 0.2019, + "step": 73104 + }, + { + "epoch": 2.81, + "learning_rate": 3.157860989997435e-06, + "loss": 0.1766, + "step": 73105 + }, + { + "epoch": 2.81, + "learning_rate": 3.157219799948705e-06, + "loss": 0.1251, + "step": 73106 + }, + { + "epoch": 2.81, + "learning_rate": 3.1565786098999744e-06, + "loss": 0.1183, + "step": 73107 + }, + { + "epoch": 2.81, + "learning_rate": 3.155937419851244e-06, + "loss": 0.1375, + "step": 73108 + }, + { + "epoch": 2.81, + "learning_rate": 3.1552962298025136e-06, + "loss": 0.33, + "step": 73109 + }, + { + "epoch": 2.81, + "learning_rate": 3.154655039753783e-06, + "loss": 0.1352, + "step": 73110 + }, + { + "epoch": 2.81, + "learning_rate": 3.154013849705053e-06, + "loss": 0.2934, + "step": 73111 + }, + { + "epoch": 2.81, + "learning_rate": 3.1533726596563224e-06, + "loss": 0.2261, + "step": 73112 + }, + { + "epoch": 2.81, + "learning_rate": 3.152731469607592e-06, + "loss": 0.2126, + "step": 73113 + }, + { + "epoch": 2.81, + "learning_rate": 3.152090279558861e-06, + "loss": 0.1558, + "step": 73114 + }, + { + "epoch": 2.81, + "learning_rate": 3.1514490895101308e-06, + "loss": 0.1428, + "step": 73115 + }, + { + "epoch": 2.81, + "learning_rate": 3.1508078994614004e-06, + "loss": 0.2506, + "step": 73116 + }, + { + "epoch": 2.81, + "learning_rate": 3.15016670941267e-06, + "loss": 0.2303, + "step": 73117 + }, + { + "epoch": 2.81, + "learning_rate": 3.1495255193639396e-06, + "loss": 0.2034, + "step": 73118 + }, + { + "epoch": 2.81, + "learning_rate": 3.148884329315209e-06, + "loss": 0.2679, + "step": 73119 + }, + { + "epoch": 2.81, + "learning_rate": 3.1482431392664788e-06, + "loss": 0.138, + "step": 73120 + }, + { + "epoch": 2.81, + "learning_rate": 3.1476019492177484e-06, + "loss": 0.2509, + "step": 73121 + }, + { + "epoch": 2.81, + "learning_rate": 3.146960759169018e-06, + "loss": 0.2653, + "step": 73122 + }, + { + "epoch": 2.81, + "learning_rate": 3.1463195691202876e-06, + "loss": 0.303, + "step": 73123 + }, + { + "epoch": 2.81, + "learning_rate": 3.145678379071557e-06, + "loss": 0.1622, + "step": 73124 + }, + { + "epoch": 2.81, + "learning_rate": 3.1450371890228263e-06, + "loss": 0.1359, + "step": 73125 + }, + { + "epoch": 2.81, + "learning_rate": 3.144395998974096e-06, + "loss": 0.1437, + "step": 73126 + }, + { + "epoch": 2.81, + "learning_rate": 3.1437548089253655e-06, + "loss": 0.2713, + "step": 73127 + }, + { + "epoch": 2.81, + "learning_rate": 3.143113618876635e-06, + "loss": 0.206, + "step": 73128 + }, + { + "epoch": 2.81, + "learning_rate": 3.1424724288279047e-06, + "loss": 0.2054, + "step": 73129 + }, + { + "epoch": 2.81, + "learning_rate": 3.1418312387791743e-06, + "loss": 0.2251, + "step": 73130 + }, + { + "epoch": 2.81, + "learning_rate": 3.141190048730444e-06, + "loss": 0.2008, + "step": 73131 + }, + { + "epoch": 2.81, + "learning_rate": 3.1405488586817135e-06, + "loss": 0.2053, + "step": 73132 + }, + { + "epoch": 2.81, + "learning_rate": 3.139907668632983e-06, + "loss": 0.276, + "step": 73133 + }, + { + "epoch": 2.81, + "learning_rate": 3.1392664785842523e-06, + "loss": 0.1643, + "step": 73134 + }, + { + "epoch": 2.81, + "learning_rate": 3.138625288535522e-06, + "loss": 0.2456, + "step": 73135 + }, + { + "epoch": 2.81, + "learning_rate": 3.1379840984867915e-06, + "loss": 0.1509, + "step": 73136 + }, + { + "epoch": 2.81, + "learning_rate": 3.137342908438061e-06, + "loss": 0.4518, + "step": 73137 + }, + { + "epoch": 2.81, + "learning_rate": 3.1367017183893307e-06, + "loss": 0.2475, + "step": 73138 + }, + { + "epoch": 2.81, + "learning_rate": 3.1360605283406003e-06, + "loss": 0.1541, + "step": 73139 + }, + { + "epoch": 2.81, + "learning_rate": 3.13541933829187e-06, + "loss": 0.3084, + "step": 73140 + }, + { + "epoch": 2.81, + "learning_rate": 3.1347781482431395e-06, + "loss": 0.1295, + "step": 73141 + }, + { + "epoch": 2.81, + "learning_rate": 3.134136958194409e-06, + "loss": 0.2921, + "step": 73142 + }, + { + "epoch": 2.81, + "learning_rate": 3.1334957681456787e-06, + "loss": 0.1682, + "step": 73143 + }, + { + "epoch": 2.81, + "learning_rate": 3.132854578096948e-06, + "loss": 0.2372, + "step": 73144 + }, + { + "epoch": 2.81, + "learning_rate": 3.1322133880482175e-06, + "loss": 0.103, + "step": 73145 + }, + { + "epoch": 2.81, + "learning_rate": 3.131572197999487e-06, + "loss": 0.3254, + "step": 73146 + }, + { + "epoch": 2.81, + "learning_rate": 3.1309310079507567e-06, + "loss": 0.2225, + "step": 73147 + }, + { + "epoch": 2.81, + "learning_rate": 3.1302898179020263e-06, + "loss": 0.2294, + "step": 73148 + }, + { + "epoch": 2.81, + "learning_rate": 3.129648627853296e-06, + "loss": 0.1881, + "step": 73149 + }, + { + "epoch": 2.81, + "learning_rate": 3.1290074378045655e-06, + "loss": 0.2239, + "step": 73150 + }, + { + "epoch": 2.81, + "learning_rate": 3.128366247755835e-06, + "loss": 0.1638, + "step": 73151 + }, + { + "epoch": 2.81, + "learning_rate": 3.1277250577071047e-06, + "loss": 0.2838, + "step": 73152 + }, + { + "epoch": 2.81, + "learning_rate": 3.1270838676583743e-06, + "loss": 0.1354, + "step": 73153 + }, + { + "epoch": 2.81, + "learning_rate": 3.1264426776096434e-06, + "loss": 0.1393, + "step": 73154 + }, + { + "epoch": 2.81, + "learning_rate": 3.125801487560913e-06, + "loss": 0.1396, + "step": 73155 + }, + { + "epoch": 2.81, + "learning_rate": 3.1251602975121826e-06, + "loss": 0.2114, + "step": 73156 + }, + { + "epoch": 2.81, + "learning_rate": 3.1245191074634522e-06, + "loss": 0.1502, + "step": 73157 + }, + { + "epoch": 2.81, + "learning_rate": 3.123877917414722e-06, + "loss": 0.1562, + "step": 73158 + }, + { + "epoch": 2.81, + "learning_rate": 3.1232367273659914e-06, + "loss": 0.3278, + "step": 73159 + }, + { + "epoch": 2.81, + "learning_rate": 3.122595537317261e-06, + "loss": 0.129, + "step": 73160 + }, + { + "epoch": 2.81, + "learning_rate": 3.1219543472685306e-06, + "loss": 0.1759, + "step": 73161 + }, + { + "epoch": 2.81, + "learning_rate": 3.1213131572198002e-06, + "loss": 0.2549, + "step": 73162 + }, + { + "epoch": 2.81, + "learning_rate": 3.12067196717107e-06, + "loss": 0.1284, + "step": 73163 + }, + { + "epoch": 2.81, + "learning_rate": 3.120030777122339e-06, + "loss": 0.2132, + "step": 73164 + }, + { + "epoch": 2.81, + "learning_rate": 3.1193895870736086e-06, + "loss": 0.1429, + "step": 73165 + }, + { + "epoch": 2.81, + "learning_rate": 3.118748397024878e-06, + "loss": 0.2697, + "step": 73166 + }, + { + "epoch": 2.81, + "learning_rate": 3.118107206976148e-06, + "loss": 0.1847, + "step": 73167 + }, + { + "epoch": 2.81, + "learning_rate": 3.1174660169274174e-06, + "loss": 0.1269, + "step": 73168 + }, + { + "epoch": 2.81, + "learning_rate": 3.116824826878687e-06, + "loss": 0.1705, + "step": 73169 + }, + { + "epoch": 2.81, + "learning_rate": 3.1161836368299566e-06, + "loss": 0.1867, + "step": 73170 + }, + { + "epoch": 2.81, + "learning_rate": 3.115542446781226e-06, + "loss": 0.2542, + "step": 73171 + }, + { + "epoch": 2.81, + "learning_rate": 3.114901256732496e-06, + "loss": 0.2039, + "step": 73172 + }, + { + "epoch": 2.81, + "learning_rate": 3.1142600666837654e-06, + "loss": 0.2249, + "step": 73173 + }, + { + "epoch": 2.81, + "learning_rate": 3.113618876635035e-06, + "loss": 0.1149, + "step": 73174 + }, + { + "epoch": 2.81, + "learning_rate": 3.112977686586304e-06, + "loss": 0.1113, + "step": 73175 + }, + { + "epoch": 2.81, + "learning_rate": 3.1123364965375738e-06, + "loss": 0.2331, + "step": 73176 + }, + { + "epoch": 2.81, + "learning_rate": 3.1116953064888434e-06, + "loss": 0.1828, + "step": 73177 + }, + { + "epoch": 2.81, + "learning_rate": 3.111054116440113e-06, + "loss": 0.1656, + "step": 73178 + }, + { + "epoch": 2.81, + "learning_rate": 3.1104129263913826e-06, + "loss": 0.1584, + "step": 73179 + }, + { + "epoch": 2.81, + "learning_rate": 3.109771736342652e-06, + "loss": 0.2077, + "step": 73180 + }, + { + "epoch": 2.81, + "learning_rate": 3.1091305462939218e-06, + "loss": 0.19, + "step": 73181 + }, + { + "epoch": 2.81, + "learning_rate": 3.1084893562451914e-06, + "loss": 0.1718, + "step": 73182 + }, + { + "epoch": 2.81, + "learning_rate": 3.107848166196461e-06, + "loss": 0.2157, + "step": 73183 + }, + { + "epoch": 2.81, + "learning_rate": 3.1072069761477306e-06, + "loss": 0.2553, + "step": 73184 + }, + { + "epoch": 2.81, + "learning_rate": 3.1065657860989997e-06, + "loss": 0.3498, + "step": 73185 + }, + { + "epoch": 2.81, + "learning_rate": 3.1059245960502693e-06, + "loss": 0.2925, + "step": 73186 + }, + { + "epoch": 2.81, + "learning_rate": 3.105283406001539e-06, + "loss": 0.1943, + "step": 73187 + }, + { + "epoch": 2.81, + "learning_rate": 3.1046422159528085e-06, + "loss": 0.1258, + "step": 73188 + }, + { + "epoch": 2.81, + "learning_rate": 3.104001025904078e-06, + "loss": 0.3798, + "step": 73189 + }, + { + "epoch": 2.81, + "learning_rate": 3.1033598358553477e-06, + "loss": 0.1614, + "step": 73190 + }, + { + "epoch": 2.81, + "learning_rate": 3.1027186458066173e-06, + "loss": 0.1914, + "step": 73191 + }, + { + "epoch": 2.81, + "learning_rate": 3.102077455757887e-06, + "loss": 0.2618, + "step": 73192 + }, + { + "epoch": 2.81, + "learning_rate": 3.1014362657091565e-06, + "loss": 0.2188, + "step": 73193 + }, + { + "epoch": 2.81, + "learning_rate": 3.100795075660426e-06, + "loss": 0.2367, + "step": 73194 + }, + { + "epoch": 2.81, + "learning_rate": 3.1001538856116953e-06, + "loss": 0.187, + "step": 73195 + }, + { + "epoch": 2.81, + "learning_rate": 3.099512695562965e-06, + "loss": 0.2318, + "step": 73196 + }, + { + "epoch": 2.81, + "learning_rate": 3.0988715055142345e-06, + "loss": 0.1519, + "step": 73197 + }, + { + "epoch": 2.81, + "learning_rate": 3.098230315465504e-06, + "loss": 0.1831, + "step": 73198 + }, + { + "epoch": 2.81, + "learning_rate": 3.0975891254167737e-06, + "loss": 0.2082, + "step": 73199 + }, + { + "epoch": 2.81, + "learning_rate": 3.0969479353680433e-06, + "loss": 0.2224, + "step": 73200 + }, + { + "epoch": 2.81, + "learning_rate": 3.096306745319313e-06, + "loss": 0.1779, + "step": 73201 + }, + { + "epoch": 2.81, + "learning_rate": 3.0956655552705825e-06, + "loss": 0.2509, + "step": 73202 + }, + { + "epoch": 2.81, + "learning_rate": 3.095024365221852e-06, + "loss": 0.2009, + "step": 73203 + }, + { + "epoch": 2.81, + "learning_rate": 3.0943831751731217e-06, + "loss": 0.2096, + "step": 73204 + }, + { + "epoch": 2.81, + "learning_rate": 3.093741985124391e-06, + "loss": 0.1178, + "step": 73205 + }, + { + "epoch": 2.81, + "learning_rate": 3.0931007950756605e-06, + "loss": 0.1446, + "step": 73206 + }, + { + "epoch": 2.81, + "learning_rate": 3.09245960502693e-06, + "loss": 0.2227, + "step": 73207 + }, + { + "epoch": 2.81, + "learning_rate": 3.0918184149781997e-06, + "loss": 0.2346, + "step": 73208 + }, + { + "epoch": 2.81, + "learning_rate": 3.0911772249294693e-06, + "loss": 0.0993, + "step": 73209 + }, + { + "epoch": 2.81, + "learning_rate": 3.090536034880739e-06, + "loss": 0.2208, + "step": 73210 + }, + { + "epoch": 2.81, + "learning_rate": 3.0898948448320085e-06, + "loss": 0.1902, + "step": 73211 + }, + { + "epoch": 2.81, + "learning_rate": 3.089253654783278e-06, + "loss": 0.3592, + "step": 73212 + }, + { + "epoch": 2.81, + "learning_rate": 3.0886124647345477e-06, + "loss": 0.2147, + "step": 73213 + }, + { + "epoch": 2.81, + "learning_rate": 3.0879712746858173e-06, + "loss": 0.1402, + "step": 73214 + }, + { + "epoch": 2.81, + "learning_rate": 3.087330084637087e-06, + "loss": 0.2465, + "step": 73215 + }, + { + "epoch": 2.81, + "learning_rate": 3.086688894588356e-06, + "loss": 0.241, + "step": 73216 + }, + { + "epoch": 2.81, + "learning_rate": 3.0860477045396256e-06, + "loss": 0.1622, + "step": 73217 + }, + { + "epoch": 2.81, + "learning_rate": 3.0854065144908952e-06, + "loss": 0.2891, + "step": 73218 + }, + { + "epoch": 2.82, + "learning_rate": 3.084765324442165e-06, + "loss": 0.1834, + "step": 73219 + }, + { + "epoch": 2.82, + "learning_rate": 3.0841241343934344e-06, + "loss": 0.2118, + "step": 73220 + }, + { + "epoch": 2.82, + "learning_rate": 3.083482944344704e-06, + "loss": 0.1504, + "step": 73221 + }, + { + "epoch": 2.82, + "learning_rate": 3.0828417542959736e-06, + "loss": 0.1295, + "step": 73222 + }, + { + "epoch": 2.82, + "learning_rate": 3.0822005642472432e-06, + "loss": 0.2256, + "step": 73223 + }, + { + "epoch": 2.82, + "learning_rate": 3.081559374198513e-06, + "loss": 0.3242, + "step": 73224 + }, + { + "epoch": 2.82, + "learning_rate": 3.0809181841497824e-06, + "loss": 0.22, + "step": 73225 + }, + { + "epoch": 2.82, + "learning_rate": 3.0802769941010516e-06, + "loss": 0.2946, + "step": 73226 + }, + { + "epoch": 2.82, + "learning_rate": 3.079635804052321e-06, + "loss": 0.2551, + "step": 73227 + }, + { + "epoch": 2.82, + "learning_rate": 3.078994614003591e-06, + "loss": 0.3527, + "step": 73228 + }, + { + "epoch": 2.82, + "learning_rate": 3.0783534239548604e-06, + "loss": 0.2694, + "step": 73229 + }, + { + "epoch": 2.82, + "learning_rate": 3.07771223390613e-06, + "loss": 0.1631, + "step": 73230 + }, + { + "epoch": 2.82, + "learning_rate": 3.0770710438573996e-06, + "loss": 0.2383, + "step": 73231 + }, + { + "epoch": 2.82, + "learning_rate": 3.076429853808669e-06, + "loss": 0.1201, + "step": 73232 + }, + { + "epoch": 2.82, + "learning_rate": 3.075788663759939e-06, + "loss": 0.2748, + "step": 73233 + }, + { + "epoch": 2.82, + "learning_rate": 3.0751474737112084e-06, + "loss": 0.2435, + "step": 73234 + }, + { + "epoch": 2.82, + "learning_rate": 3.074506283662478e-06, + "loss": 0.1431, + "step": 73235 + }, + { + "epoch": 2.82, + "learning_rate": 3.073865093613747e-06, + "loss": 0.1789, + "step": 73236 + }, + { + "epoch": 2.82, + "learning_rate": 3.0732239035650168e-06, + "loss": 0.2295, + "step": 73237 + }, + { + "epoch": 2.82, + "learning_rate": 3.0725827135162864e-06, + "loss": 0.1723, + "step": 73238 + }, + { + "epoch": 2.82, + "learning_rate": 3.071941523467556e-06, + "loss": 0.1467, + "step": 73239 + }, + { + "epoch": 2.82, + "learning_rate": 3.0713003334188256e-06, + "loss": 0.2607, + "step": 73240 + }, + { + "epoch": 2.82, + "learning_rate": 3.070659143370095e-06, + "loss": 0.1144, + "step": 73241 + }, + { + "epoch": 2.82, + "learning_rate": 3.0700179533213648e-06, + "loss": 0.1629, + "step": 73242 + }, + { + "epoch": 2.82, + "learning_rate": 3.0693767632726344e-06, + "loss": 0.2037, + "step": 73243 + }, + { + "epoch": 2.82, + "learning_rate": 3.068735573223904e-06, + "loss": 0.2425, + "step": 73244 + }, + { + "epoch": 2.82, + "learning_rate": 3.0680943831751736e-06, + "loss": 0.2708, + "step": 73245 + }, + { + "epoch": 2.82, + "learning_rate": 3.0674531931264427e-06, + "loss": 0.2877, + "step": 73246 + }, + { + "epoch": 2.82, + "learning_rate": 3.0668120030777123e-06, + "loss": 0.0901, + "step": 73247 + }, + { + "epoch": 2.82, + "learning_rate": 3.066170813028982e-06, + "loss": 0.2598, + "step": 73248 + }, + { + "epoch": 2.82, + "learning_rate": 3.0655296229802515e-06, + "loss": 0.2925, + "step": 73249 + }, + { + "epoch": 2.82, + "learning_rate": 3.064888432931521e-06, + "loss": 0.1418, + "step": 73250 + }, + { + "epoch": 2.82, + "learning_rate": 3.0642472428827907e-06, + "loss": 0.3132, + "step": 73251 + }, + { + "epoch": 2.82, + "learning_rate": 3.0636060528340603e-06, + "loss": 0.2322, + "step": 73252 + }, + { + "epoch": 2.82, + "learning_rate": 3.06296486278533e-06, + "loss": 0.2328, + "step": 73253 + }, + { + "epoch": 2.82, + "learning_rate": 3.0623236727365995e-06, + "loss": 0.1328, + "step": 73254 + }, + { + "epoch": 2.82, + "learning_rate": 3.061682482687869e-06, + "loss": 0.0659, + "step": 73255 + }, + { + "epoch": 2.82, + "learning_rate": 3.0610412926391383e-06, + "loss": 0.1767, + "step": 73256 + }, + { + "epoch": 2.82, + "learning_rate": 3.060400102590408e-06, + "loss": 0.1806, + "step": 73257 + }, + { + "epoch": 2.82, + "learning_rate": 3.0597589125416775e-06, + "loss": 0.2183, + "step": 73258 + }, + { + "epoch": 2.82, + "learning_rate": 3.059117722492947e-06, + "loss": 0.2195, + "step": 73259 + }, + { + "epoch": 2.82, + "learning_rate": 3.0584765324442167e-06, + "loss": 0.134, + "step": 73260 + }, + { + "epoch": 2.82, + "learning_rate": 3.0578353423954863e-06, + "loss": 0.1105, + "step": 73261 + }, + { + "epoch": 2.82, + "learning_rate": 3.057194152346756e-06, + "loss": 0.2122, + "step": 73262 + }, + { + "epoch": 2.82, + "learning_rate": 3.0565529622980255e-06, + "loss": 0.1439, + "step": 73263 + }, + { + "epoch": 2.82, + "learning_rate": 3.055911772249295e-06, + "loss": 0.1049, + "step": 73264 + }, + { + "epoch": 2.82, + "learning_rate": 3.0552705822005647e-06, + "loss": 0.0915, + "step": 73265 + }, + { + "epoch": 2.82, + "learning_rate": 3.0546293921518343e-06, + "loss": 0.1159, + "step": 73266 + }, + { + "epoch": 2.82, + "learning_rate": 3.0539882021031035e-06, + "loss": 0.3637, + "step": 73267 + }, + { + "epoch": 2.82, + "learning_rate": 3.053347012054373e-06, + "loss": 0.2612, + "step": 73268 + }, + { + "epoch": 2.82, + "learning_rate": 3.0527058220056427e-06, + "loss": 0.1801, + "step": 73269 + }, + { + "epoch": 2.82, + "learning_rate": 3.0520646319569123e-06, + "loss": 0.1696, + "step": 73270 + }, + { + "epoch": 2.82, + "learning_rate": 3.051423441908182e-06, + "loss": 0.1462, + "step": 73271 + }, + { + "epoch": 2.82, + "learning_rate": 3.0507822518594515e-06, + "loss": 0.1087, + "step": 73272 + }, + { + "epoch": 2.82, + "learning_rate": 3.050141061810721e-06, + "loss": 0.1893, + "step": 73273 + }, + { + "epoch": 2.82, + "learning_rate": 3.0494998717619907e-06, + "loss": 0.162, + "step": 73274 + }, + { + "epoch": 2.82, + "learning_rate": 3.0488586817132603e-06, + "loss": 0.1794, + "step": 73275 + }, + { + "epoch": 2.82, + "learning_rate": 3.04821749166453e-06, + "loss": 0.0899, + "step": 73276 + }, + { + "epoch": 2.82, + "learning_rate": 3.047576301615799e-06, + "loss": 0.3203, + "step": 73277 + }, + { + "epoch": 2.82, + "learning_rate": 3.0469351115670686e-06, + "loss": 0.2858, + "step": 73278 + }, + { + "epoch": 2.82, + "learning_rate": 3.0462939215183382e-06, + "loss": 0.1836, + "step": 73279 + }, + { + "epoch": 2.82, + "learning_rate": 3.045652731469608e-06, + "loss": 0.0731, + "step": 73280 + }, + { + "epoch": 2.82, + "learning_rate": 3.0450115414208774e-06, + "loss": 0.234, + "step": 73281 + }, + { + "epoch": 2.82, + "learning_rate": 3.044370351372147e-06, + "loss": 0.2137, + "step": 73282 + }, + { + "epoch": 2.82, + "learning_rate": 3.0437291613234166e-06, + "loss": 0.247, + "step": 73283 + }, + { + "epoch": 2.82, + "learning_rate": 3.0430879712746862e-06, + "loss": 0.1558, + "step": 73284 + }, + { + "epoch": 2.82, + "learning_rate": 3.042446781225956e-06, + "loss": 0.2223, + "step": 73285 + }, + { + "epoch": 2.82, + "learning_rate": 3.0418055911772254e-06, + "loss": 0.2332, + "step": 73286 + }, + { + "epoch": 2.82, + "learning_rate": 3.0411644011284946e-06, + "loss": 0.133, + "step": 73287 + }, + { + "epoch": 2.82, + "learning_rate": 3.0405232110797642e-06, + "loss": 0.3038, + "step": 73288 + }, + { + "epoch": 2.82, + "learning_rate": 3.039882021031034e-06, + "loss": 0.1626, + "step": 73289 + }, + { + "epoch": 2.82, + "learning_rate": 3.0392408309823034e-06, + "loss": 0.3048, + "step": 73290 + }, + { + "epoch": 2.82, + "learning_rate": 3.038599640933573e-06, + "loss": 0.1823, + "step": 73291 + }, + { + "epoch": 2.82, + "learning_rate": 3.0379584508848426e-06, + "loss": 0.2183, + "step": 73292 + }, + { + "epoch": 2.82, + "learning_rate": 3.037317260836112e-06, + "loss": 0.2263, + "step": 73293 + }, + { + "epoch": 2.82, + "learning_rate": 3.036676070787382e-06, + "loss": 0.1879, + "step": 73294 + }, + { + "epoch": 2.82, + "learning_rate": 3.0360348807386514e-06, + "loss": 0.3797, + "step": 73295 + }, + { + "epoch": 2.82, + "learning_rate": 3.035393690689921e-06, + "loss": 0.1478, + "step": 73296 + }, + { + "epoch": 2.82, + "learning_rate": 3.03475250064119e-06, + "loss": 0.1224, + "step": 73297 + }, + { + "epoch": 2.82, + "learning_rate": 3.0341113105924598e-06, + "loss": 0.199, + "step": 73298 + }, + { + "epoch": 2.82, + "learning_rate": 3.0334701205437294e-06, + "loss": 0.2166, + "step": 73299 + }, + { + "epoch": 2.82, + "learning_rate": 3.032828930494999e-06, + "loss": 0.218, + "step": 73300 + }, + { + "epoch": 2.82, + "learning_rate": 3.0321877404462686e-06, + "loss": 0.1416, + "step": 73301 + }, + { + "epoch": 2.82, + "learning_rate": 3.031546550397538e-06, + "loss": 0.1854, + "step": 73302 + }, + { + "epoch": 2.82, + "learning_rate": 3.0309053603488078e-06, + "loss": 0.2987, + "step": 73303 + }, + { + "epoch": 2.82, + "learning_rate": 3.0302641703000774e-06, + "loss": 0.25, + "step": 73304 + }, + { + "epoch": 2.82, + "learning_rate": 3.029622980251347e-06, + "loss": 0.2042, + "step": 73305 + }, + { + "epoch": 2.82, + "learning_rate": 3.0289817902026166e-06, + "loss": 0.3927, + "step": 73306 + }, + { + "epoch": 2.82, + "learning_rate": 3.0283406001538857e-06, + "loss": 0.3211, + "step": 73307 + }, + { + "epoch": 2.82, + "learning_rate": 3.0276994101051553e-06, + "loss": 0.2289, + "step": 73308 + }, + { + "epoch": 2.82, + "learning_rate": 3.027058220056425e-06, + "loss": 0.1747, + "step": 73309 + }, + { + "epoch": 2.82, + "learning_rate": 3.0264170300076945e-06, + "loss": 0.3387, + "step": 73310 + }, + { + "epoch": 2.82, + "learning_rate": 3.025775839958964e-06, + "loss": 0.1035, + "step": 73311 + }, + { + "epoch": 2.82, + "learning_rate": 3.0251346499102337e-06, + "loss": 0.1519, + "step": 73312 + }, + { + "epoch": 2.82, + "learning_rate": 3.0244934598615033e-06, + "loss": 0.1475, + "step": 73313 + }, + { + "epoch": 2.82, + "learning_rate": 3.023852269812773e-06, + "loss": 0.2117, + "step": 73314 + }, + { + "epoch": 2.82, + "learning_rate": 3.0232110797640425e-06, + "loss": 0.1835, + "step": 73315 + }, + { + "epoch": 2.82, + "learning_rate": 3.022569889715312e-06, + "loss": 0.2692, + "step": 73316 + }, + { + "epoch": 2.82, + "learning_rate": 3.0219286996665813e-06, + "loss": 0.1438, + "step": 73317 + }, + { + "epoch": 2.82, + "learning_rate": 3.021287509617851e-06, + "loss": 0.1074, + "step": 73318 + }, + { + "epoch": 2.82, + "learning_rate": 3.0206463195691205e-06, + "loss": 0.2302, + "step": 73319 + }, + { + "epoch": 2.82, + "learning_rate": 3.02000512952039e-06, + "loss": 0.2713, + "step": 73320 + }, + { + "epoch": 2.82, + "learning_rate": 3.0193639394716597e-06, + "loss": 0.1869, + "step": 73321 + }, + { + "epoch": 2.82, + "learning_rate": 3.0187227494229293e-06, + "loss": 0.0717, + "step": 73322 + }, + { + "epoch": 2.82, + "learning_rate": 3.018081559374199e-06, + "loss": 0.219, + "step": 73323 + }, + { + "epoch": 2.82, + "learning_rate": 3.0174403693254685e-06, + "loss": 0.2131, + "step": 73324 + }, + { + "epoch": 2.82, + "learning_rate": 3.016799179276738e-06, + "loss": 0.2035, + "step": 73325 + }, + { + "epoch": 2.82, + "learning_rate": 3.0161579892280073e-06, + "loss": 0.2553, + "step": 73326 + }, + { + "epoch": 2.82, + "learning_rate": 3.015516799179277e-06, + "loss": 0.2045, + "step": 73327 + }, + { + "epoch": 2.82, + "learning_rate": 3.0148756091305465e-06, + "loss": 0.2078, + "step": 73328 + }, + { + "epoch": 2.82, + "learning_rate": 3.014234419081816e-06, + "loss": 0.2341, + "step": 73329 + }, + { + "epoch": 2.82, + "learning_rate": 3.0135932290330857e-06, + "loss": 0.2292, + "step": 73330 + }, + { + "epoch": 2.82, + "learning_rate": 3.0129520389843553e-06, + "loss": 0.2126, + "step": 73331 + }, + { + "epoch": 2.82, + "learning_rate": 3.012310848935625e-06, + "loss": 0.1363, + "step": 73332 + }, + { + "epoch": 2.82, + "learning_rate": 3.0116696588868945e-06, + "loss": 0.4253, + "step": 73333 + }, + { + "epoch": 2.82, + "learning_rate": 3.011028468838164e-06, + "loss": 0.1517, + "step": 73334 + }, + { + "epoch": 2.82, + "learning_rate": 3.0103872787894337e-06, + "loss": 0.2776, + "step": 73335 + }, + { + "epoch": 2.82, + "learning_rate": 3.009746088740703e-06, + "loss": 0.1959, + "step": 73336 + }, + { + "epoch": 2.82, + "learning_rate": 3.0091048986919724e-06, + "loss": 0.1777, + "step": 73337 + }, + { + "epoch": 2.82, + "learning_rate": 3.008463708643242e-06, + "loss": 0.1199, + "step": 73338 + }, + { + "epoch": 2.82, + "learning_rate": 3.0078225185945116e-06, + "loss": 0.1323, + "step": 73339 + }, + { + "epoch": 2.82, + "learning_rate": 3.0071813285457812e-06, + "loss": 0.2014, + "step": 73340 + }, + { + "epoch": 2.82, + "learning_rate": 3.006540138497051e-06, + "loss": 0.1519, + "step": 73341 + }, + { + "epoch": 2.82, + "learning_rate": 3.0058989484483204e-06, + "loss": 0.2279, + "step": 73342 + }, + { + "epoch": 2.82, + "learning_rate": 3.00525775839959e-06, + "loss": 0.1528, + "step": 73343 + }, + { + "epoch": 2.82, + "learning_rate": 3.0046165683508596e-06, + "loss": 0.2101, + "step": 73344 + }, + { + "epoch": 2.82, + "learning_rate": 3.0039753783021292e-06, + "loss": 0.1887, + "step": 73345 + }, + { + "epoch": 2.82, + "learning_rate": 3.0033341882533984e-06, + "loss": 0.1469, + "step": 73346 + }, + { + "epoch": 2.82, + "learning_rate": 3.002692998204668e-06, + "loss": 0.2028, + "step": 73347 + }, + { + "epoch": 2.82, + "learning_rate": 3.0020518081559376e-06, + "loss": 0.1724, + "step": 73348 + }, + { + "epoch": 2.82, + "learning_rate": 3.0014106181072072e-06, + "loss": 0.2353, + "step": 73349 + }, + { + "epoch": 2.82, + "learning_rate": 3.000769428058477e-06, + "loss": 0.2609, + "step": 73350 + }, + { + "epoch": 2.82, + "learning_rate": 3.0001282380097464e-06, + "loss": 0.1712, + "step": 73351 + }, + { + "epoch": 2.82, + "learning_rate": 2.999487047961016e-06, + "loss": 0.2888, + "step": 73352 + }, + { + "epoch": 2.82, + "learning_rate": 2.9988458579122856e-06, + "loss": 0.1278, + "step": 73353 + }, + { + "epoch": 2.82, + "learning_rate": 2.998204667863555e-06, + "loss": 0.1448, + "step": 73354 + }, + { + "epoch": 2.82, + "learning_rate": 2.997563477814825e-06, + "loss": 0.2621, + "step": 73355 + }, + { + "epoch": 2.82, + "learning_rate": 2.996922287766094e-06, + "loss": 0.1284, + "step": 73356 + }, + { + "epoch": 2.82, + "learning_rate": 2.9962810977173636e-06, + "loss": 0.2406, + "step": 73357 + }, + { + "epoch": 2.82, + "learning_rate": 2.995639907668633e-06, + "loss": 0.114, + "step": 73358 + }, + { + "epoch": 2.82, + "learning_rate": 2.9949987176199028e-06, + "loss": 0.2257, + "step": 73359 + }, + { + "epoch": 2.82, + "learning_rate": 2.9943575275711724e-06, + "loss": 0.1802, + "step": 73360 + }, + { + "epoch": 2.82, + "learning_rate": 2.993716337522442e-06, + "loss": 0.2333, + "step": 73361 + }, + { + "epoch": 2.82, + "learning_rate": 2.9930751474737116e-06, + "loss": 0.2499, + "step": 73362 + }, + { + "epoch": 2.82, + "learning_rate": 2.992433957424981e-06, + "loss": 0.149, + "step": 73363 + }, + { + "epoch": 2.82, + "learning_rate": 2.9917927673762508e-06, + "loss": 0.2529, + "step": 73364 + }, + { + "epoch": 2.82, + "learning_rate": 2.9911515773275204e-06, + "loss": 0.1466, + "step": 73365 + }, + { + "epoch": 2.82, + "learning_rate": 2.9905103872787895e-06, + "loss": 0.2921, + "step": 73366 + }, + { + "epoch": 2.82, + "learning_rate": 2.989869197230059e-06, + "loss": 0.2197, + "step": 73367 + }, + { + "epoch": 2.82, + "learning_rate": 2.9892280071813287e-06, + "loss": 0.2753, + "step": 73368 + }, + { + "epoch": 2.82, + "learning_rate": 2.9885868171325983e-06, + "loss": 0.3693, + "step": 73369 + }, + { + "epoch": 2.82, + "learning_rate": 2.987945627083868e-06, + "loss": 0.1658, + "step": 73370 + }, + { + "epoch": 2.82, + "learning_rate": 2.9873044370351375e-06, + "loss": 0.137, + "step": 73371 + }, + { + "epoch": 2.82, + "learning_rate": 2.986663246986407e-06, + "loss": 0.0877, + "step": 73372 + }, + { + "epoch": 2.82, + "learning_rate": 2.9860220569376767e-06, + "loss": 0.2203, + "step": 73373 + }, + { + "epoch": 2.82, + "learning_rate": 2.9853808668889463e-06, + "loss": 0.1118, + "step": 73374 + }, + { + "epoch": 2.82, + "learning_rate": 2.9847396768402155e-06, + "loss": 0.1424, + "step": 73375 + }, + { + "epoch": 2.82, + "learning_rate": 2.984098486791485e-06, + "loss": 0.1751, + "step": 73376 + }, + { + "epoch": 2.82, + "learning_rate": 2.9834572967427547e-06, + "loss": 0.2476, + "step": 73377 + }, + { + "epoch": 2.82, + "learning_rate": 2.9828161066940243e-06, + "loss": 0.2653, + "step": 73378 + }, + { + "epoch": 2.82, + "learning_rate": 2.982174916645294e-06, + "loss": 0.1708, + "step": 73379 + }, + { + "epoch": 2.82, + "learning_rate": 2.9815337265965635e-06, + "loss": 0.2999, + "step": 73380 + }, + { + "epoch": 2.82, + "learning_rate": 2.980892536547833e-06, + "loss": 0.1509, + "step": 73381 + }, + { + "epoch": 2.82, + "learning_rate": 2.9802513464991027e-06, + "loss": 0.2485, + "step": 73382 + }, + { + "epoch": 2.82, + "learning_rate": 2.9796101564503723e-06, + "loss": 0.2411, + "step": 73383 + }, + { + "epoch": 2.82, + "learning_rate": 2.978968966401642e-06, + "loss": 0.1218, + "step": 73384 + }, + { + "epoch": 2.82, + "learning_rate": 2.978327776352911e-06, + "loss": 0.238, + "step": 73385 + }, + { + "epoch": 2.82, + "learning_rate": 2.9776865863041807e-06, + "loss": 0.1329, + "step": 73386 + }, + { + "epoch": 2.82, + "learning_rate": 2.9770453962554503e-06, + "loss": 0.1353, + "step": 73387 + }, + { + "epoch": 2.82, + "learning_rate": 2.97640420620672e-06, + "loss": 0.3238, + "step": 73388 + }, + { + "epoch": 2.82, + "learning_rate": 2.9757630161579895e-06, + "loss": 0.1387, + "step": 73389 + }, + { + "epoch": 2.82, + "learning_rate": 2.975121826109259e-06, + "loss": 0.149, + "step": 73390 + }, + { + "epoch": 2.82, + "learning_rate": 2.9744806360605287e-06, + "loss": 0.1374, + "step": 73391 + }, + { + "epoch": 2.82, + "learning_rate": 2.9738394460117983e-06, + "loss": 0.2793, + "step": 73392 + }, + { + "epoch": 2.82, + "learning_rate": 2.973198255963068e-06, + "loss": 0.2274, + "step": 73393 + }, + { + "epoch": 2.82, + "learning_rate": 2.9725570659143375e-06, + "loss": 0.2708, + "step": 73394 + }, + { + "epoch": 2.82, + "learning_rate": 2.9719158758656067e-06, + "loss": 0.1612, + "step": 73395 + }, + { + "epoch": 2.82, + "learning_rate": 2.9712746858168762e-06, + "loss": 0.3364, + "step": 73396 + }, + { + "epoch": 2.82, + "learning_rate": 2.970633495768146e-06, + "loss": 0.156, + "step": 73397 + }, + { + "epoch": 2.82, + "learning_rate": 2.9699923057194154e-06, + "loss": 0.2269, + "step": 73398 + }, + { + "epoch": 2.82, + "learning_rate": 2.969351115670685e-06, + "loss": 0.3298, + "step": 73399 + }, + { + "epoch": 2.82, + "learning_rate": 2.9687099256219546e-06, + "loss": 0.1301, + "step": 73400 + }, + { + "epoch": 2.82, + "learning_rate": 2.9680687355732242e-06, + "loss": 0.2447, + "step": 73401 + }, + { + "epoch": 2.82, + "learning_rate": 2.967427545524494e-06, + "loss": 0.2814, + "step": 73402 + }, + { + "epoch": 2.82, + "learning_rate": 2.9667863554757634e-06, + "loss": 0.2013, + "step": 73403 + }, + { + "epoch": 2.82, + "learning_rate": 2.966145165427033e-06, + "loss": 0.1782, + "step": 73404 + }, + { + "epoch": 2.82, + "learning_rate": 2.9655039753783022e-06, + "loss": 0.1752, + "step": 73405 + }, + { + "epoch": 2.82, + "learning_rate": 2.964862785329572e-06, + "loss": 0.2043, + "step": 73406 + }, + { + "epoch": 2.82, + "learning_rate": 2.9642215952808414e-06, + "loss": 0.137, + "step": 73407 + }, + { + "epoch": 2.82, + "learning_rate": 2.963580405232111e-06, + "loss": 0.1014, + "step": 73408 + }, + { + "epoch": 2.82, + "learning_rate": 2.9629392151833806e-06, + "loss": 0.1708, + "step": 73409 + }, + { + "epoch": 2.82, + "learning_rate": 2.9622980251346502e-06, + "loss": 0.0971, + "step": 73410 + }, + { + "epoch": 2.82, + "learning_rate": 2.96165683508592e-06, + "loss": 0.2979, + "step": 73411 + }, + { + "epoch": 2.82, + "learning_rate": 2.9610156450371894e-06, + "loss": 0.2372, + "step": 73412 + }, + { + "epoch": 2.82, + "learning_rate": 2.960374454988459e-06, + "loss": 0.3273, + "step": 73413 + }, + { + "epoch": 2.82, + "learning_rate": 2.9597332649397286e-06, + "loss": 0.0931, + "step": 73414 + }, + { + "epoch": 2.82, + "learning_rate": 2.9590920748909978e-06, + "loss": 0.3173, + "step": 73415 + }, + { + "epoch": 2.82, + "learning_rate": 2.9584508848422674e-06, + "loss": 0.1272, + "step": 73416 + }, + { + "epoch": 2.82, + "learning_rate": 2.957809694793537e-06, + "loss": 0.1897, + "step": 73417 + }, + { + "epoch": 2.82, + "learning_rate": 2.9571685047448066e-06, + "loss": 0.0683, + "step": 73418 + }, + { + "epoch": 2.82, + "learning_rate": 2.956527314696076e-06, + "loss": 0.2437, + "step": 73419 + }, + { + "epoch": 2.82, + "learning_rate": 2.9558861246473458e-06, + "loss": 0.0449, + "step": 73420 + }, + { + "epoch": 2.82, + "learning_rate": 2.9552449345986154e-06, + "loss": 0.1757, + "step": 73421 + }, + { + "epoch": 2.82, + "learning_rate": 2.954603744549885e-06, + "loss": 0.2115, + "step": 73422 + }, + { + "epoch": 2.82, + "learning_rate": 2.9539625545011546e-06, + "loss": 0.1796, + "step": 73423 + }, + { + "epoch": 2.82, + "learning_rate": 2.953321364452424e-06, + "loss": 0.2489, + "step": 73424 + }, + { + "epoch": 2.82, + "learning_rate": 2.9526801744036934e-06, + "loss": 0.3381, + "step": 73425 + }, + { + "epoch": 2.82, + "learning_rate": 2.952038984354963e-06, + "loss": 0.0678, + "step": 73426 + }, + { + "epoch": 2.82, + "learning_rate": 2.9513977943062325e-06, + "loss": 0.2584, + "step": 73427 + }, + { + "epoch": 2.82, + "learning_rate": 2.950756604257502e-06, + "loss": 0.1712, + "step": 73428 + }, + { + "epoch": 2.82, + "learning_rate": 2.9501154142087717e-06, + "loss": 0.2722, + "step": 73429 + }, + { + "epoch": 2.82, + "learning_rate": 2.9494742241600413e-06, + "loss": 0.3451, + "step": 73430 + }, + { + "epoch": 2.82, + "learning_rate": 2.948833034111311e-06, + "loss": 0.1361, + "step": 73431 + }, + { + "epoch": 2.82, + "learning_rate": 2.9481918440625805e-06, + "loss": 0.2381, + "step": 73432 + }, + { + "epoch": 2.82, + "learning_rate": 2.94755065401385e-06, + "loss": 0.0998, + "step": 73433 + }, + { + "epoch": 2.82, + "learning_rate": 2.9469094639651193e-06, + "loss": 0.1823, + "step": 73434 + }, + { + "epoch": 2.82, + "learning_rate": 2.946268273916389e-06, + "loss": 0.1379, + "step": 73435 + }, + { + "epoch": 2.82, + "learning_rate": 2.9456270838676585e-06, + "loss": 0.1706, + "step": 73436 + }, + { + "epoch": 2.82, + "learning_rate": 2.944985893818928e-06, + "loss": 0.2434, + "step": 73437 + }, + { + "epoch": 2.82, + "learning_rate": 2.9443447037701977e-06, + "loss": 0.3025, + "step": 73438 + }, + { + "epoch": 2.82, + "learning_rate": 2.9437035137214673e-06, + "loss": 0.1996, + "step": 73439 + }, + { + "epoch": 2.82, + "learning_rate": 2.943062323672737e-06, + "loss": 0.1847, + "step": 73440 + }, + { + "epoch": 2.82, + "learning_rate": 2.9424211336240065e-06, + "loss": 0.2692, + "step": 73441 + }, + { + "epoch": 2.82, + "learning_rate": 2.941779943575276e-06, + "loss": 0.1455, + "step": 73442 + }, + { + "epoch": 2.82, + "learning_rate": 2.9411387535265457e-06, + "loss": 0.1382, + "step": 73443 + }, + { + "epoch": 2.82, + "learning_rate": 2.940497563477815e-06, + "loss": 0.2717, + "step": 73444 + }, + { + "epoch": 2.82, + "learning_rate": 2.9398563734290845e-06, + "loss": 0.0986, + "step": 73445 + }, + { + "epoch": 2.82, + "learning_rate": 2.939215183380354e-06, + "loss": 0.1825, + "step": 73446 + }, + { + "epoch": 2.82, + "learning_rate": 2.9385739933316237e-06, + "loss": 0.2167, + "step": 73447 + }, + { + "epoch": 2.82, + "learning_rate": 2.9379328032828933e-06, + "loss": 0.1321, + "step": 73448 + }, + { + "epoch": 2.82, + "learning_rate": 2.937291613234163e-06, + "loss": 0.3107, + "step": 73449 + }, + { + "epoch": 2.82, + "learning_rate": 2.9366504231854325e-06, + "loss": 0.2935, + "step": 73450 + }, + { + "epoch": 2.82, + "learning_rate": 2.936009233136702e-06, + "loss": 0.2787, + "step": 73451 + }, + { + "epoch": 2.82, + "learning_rate": 2.9353680430879717e-06, + "loss": 0.2434, + "step": 73452 + }, + { + "epoch": 2.82, + "learning_rate": 2.9347268530392413e-06, + "loss": 0.1849, + "step": 73453 + }, + { + "epoch": 2.82, + "learning_rate": 2.9340856629905105e-06, + "loss": 0.2448, + "step": 73454 + }, + { + "epoch": 2.82, + "learning_rate": 2.93344447294178e-06, + "loss": 0.2075, + "step": 73455 + }, + { + "epoch": 2.82, + "learning_rate": 2.9328032828930497e-06, + "loss": 0.1464, + "step": 73456 + }, + { + "epoch": 2.82, + "learning_rate": 2.9321620928443192e-06, + "loss": 0.3156, + "step": 73457 + }, + { + "epoch": 2.82, + "learning_rate": 2.931520902795589e-06, + "loss": 0.1763, + "step": 73458 + }, + { + "epoch": 2.82, + "learning_rate": 2.9308797127468584e-06, + "loss": 0.1223, + "step": 73459 + }, + { + "epoch": 2.82, + "learning_rate": 2.930238522698128e-06, + "loss": 0.2434, + "step": 73460 + }, + { + "epoch": 2.82, + "learning_rate": 2.9295973326493976e-06, + "loss": 0.3579, + "step": 73461 + }, + { + "epoch": 2.82, + "learning_rate": 2.9289561426006672e-06, + "loss": 0.3151, + "step": 73462 + }, + { + "epoch": 2.82, + "learning_rate": 2.928314952551937e-06, + "loss": 0.195, + "step": 73463 + }, + { + "epoch": 2.82, + "learning_rate": 2.927673762503206e-06, + "loss": 0.1859, + "step": 73464 + }, + { + "epoch": 2.82, + "learning_rate": 2.9270325724544756e-06, + "loss": 0.1555, + "step": 73465 + }, + { + "epoch": 2.82, + "learning_rate": 2.9263913824057452e-06, + "loss": 0.2223, + "step": 73466 + }, + { + "epoch": 2.82, + "learning_rate": 2.925750192357015e-06, + "loss": 0.2326, + "step": 73467 + }, + { + "epoch": 2.82, + "learning_rate": 2.9251090023082844e-06, + "loss": 0.2176, + "step": 73468 + }, + { + "epoch": 2.82, + "learning_rate": 2.924467812259554e-06, + "loss": 0.1475, + "step": 73469 + }, + { + "epoch": 2.82, + "learning_rate": 2.9238266222108236e-06, + "loss": 0.1563, + "step": 73470 + }, + { + "epoch": 2.82, + "learning_rate": 2.9231854321620932e-06, + "loss": 0.1732, + "step": 73471 + }, + { + "epoch": 2.82, + "learning_rate": 2.922544242113363e-06, + "loss": 0.0923, + "step": 73472 + }, + { + "epoch": 2.82, + "learning_rate": 2.9219030520646324e-06, + "loss": 0.2771, + "step": 73473 + }, + { + "epoch": 2.82, + "learning_rate": 2.9212618620159016e-06, + "loss": 0.2309, + "step": 73474 + }, + { + "epoch": 2.82, + "learning_rate": 2.920620671967171e-06, + "loss": 0.2575, + "step": 73475 + }, + { + "epoch": 2.82, + "learning_rate": 2.9199794819184408e-06, + "loss": 0.2746, + "step": 73476 + }, + { + "epoch": 2.82, + "learning_rate": 2.9193382918697104e-06, + "loss": 0.1827, + "step": 73477 + }, + { + "epoch": 2.82, + "learning_rate": 2.91869710182098e-06, + "loss": 0.0737, + "step": 73478 + }, + { + "epoch": 2.83, + "learning_rate": 2.9180559117722496e-06, + "loss": 0.1307, + "step": 73479 + }, + { + "epoch": 2.83, + "learning_rate": 2.917414721723519e-06, + "loss": 0.1826, + "step": 73480 + }, + { + "epoch": 2.83, + "learning_rate": 2.9167735316747888e-06, + "loss": 0.362, + "step": 73481 + }, + { + "epoch": 2.83, + "learning_rate": 2.9161323416260584e-06, + "loss": 0.0464, + "step": 73482 + }, + { + "epoch": 2.83, + "learning_rate": 2.915491151577328e-06, + "loss": 0.2028, + "step": 73483 + }, + { + "epoch": 2.83, + "learning_rate": 2.914849961528597e-06, + "loss": 0.1831, + "step": 73484 + }, + { + "epoch": 2.83, + "learning_rate": 2.9142087714798668e-06, + "loss": 0.1711, + "step": 73485 + }, + { + "epoch": 2.83, + "learning_rate": 2.9135675814311364e-06, + "loss": 0.1802, + "step": 73486 + }, + { + "epoch": 2.83, + "learning_rate": 2.912926391382406e-06, + "loss": 0.3201, + "step": 73487 + }, + { + "epoch": 2.83, + "learning_rate": 2.9122852013336755e-06, + "loss": 0.2524, + "step": 73488 + }, + { + "epoch": 2.83, + "learning_rate": 2.911644011284945e-06, + "loss": 0.151, + "step": 73489 + }, + { + "epoch": 2.83, + "learning_rate": 2.9110028212362147e-06, + "loss": 0.1933, + "step": 73490 + }, + { + "epoch": 2.83, + "learning_rate": 2.9103616311874843e-06, + "loss": 0.1205, + "step": 73491 + }, + { + "epoch": 2.83, + "learning_rate": 2.909720441138754e-06, + "loss": 0.226, + "step": 73492 + }, + { + "epoch": 2.83, + "learning_rate": 2.909079251090023e-06, + "loss": 0.1441, + "step": 73493 + }, + { + "epoch": 2.83, + "learning_rate": 2.9084380610412927e-06, + "loss": 0.444, + "step": 73494 + }, + { + "epoch": 2.83, + "learning_rate": 2.9077968709925623e-06, + "loss": 0.1917, + "step": 73495 + }, + { + "epoch": 2.83, + "learning_rate": 2.907155680943832e-06, + "loss": 0.2732, + "step": 73496 + }, + { + "epoch": 2.83, + "learning_rate": 2.9065144908951015e-06, + "loss": 0.1732, + "step": 73497 + }, + { + "epoch": 2.83, + "learning_rate": 2.905873300846371e-06, + "loss": 0.189, + "step": 73498 + }, + { + "epoch": 2.83, + "learning_rate": 2.9052321107976407e-06, + "loss": 0.1903, + "step": 73499 + }, + { + "epoch": 2.83, + "learning_rate": 2.9045909207489103e-06, + "loss": 0.286, + "step": 73500 + }, + { + "epoch": 2.83, + "learning_rate": 2.90394973070018e-06, + "loss": 0.0952, + "step": 73501 + }, + { + "epoch": 2.83, + "learning_rate": 2.9033085406514495e-06, + "loss": 0.1694, + "step": 73502 + }, + { + "epoch": 2.83, + "learning_rate": 2.9026673506027187e-06, + "loss": 0.2117, + "step": 73503 + }, + { + "epoch": 2.83, + "learning_rate": 2.9020261605539883e-06, + "loss": 0.22, + "step": 73504 + }, + { + "epoch": 2.83, + "learning_rate": 2.901384970505258e-06, + "loss": 0.1901, + "step": 73505 + }, + { + "epoch": 2.83, + "learning_rate": 2.9007437804565275e-06, + "loss": 0.31, + "step": 73506 + }, + { + "epoch": 2.83, + "learning_rate": 2.900102590407797e-06, + "loss": 0.2481, + "step": 73507 + }, + { + "epoch": 2.83, + "learning_rate": 2.8994614003590667e-06, + "loss": 0.2631, + "step": 73508 + }, + { + "epoch": 2.83, + "learning_rate": 2.8988202103103363e-06, + "loss": 0.1339, + "step": 73509 + }, + { + "epoch": 2.83, + "learning_rate": 2.898179020261606e-06, + "loss": 0.1517, + "step": 73510 + }, + { + "epoch": 2.83, + "learning_rate": 2.8975378302128755e-06, + "loss": 0.139, + "step": 73511 + }, + { + "epoch": 2.83, + "learning_rate": 2.896896640164145e-06, + "loss": 0.3326, + "step": 73512 + }, + { + "epoch": 2.83, + "learning_rate": 2.8962554501154143e-06, + "loss": 0.1004, + "step": 73513 + }, + { + "epoch": 2.83, + "learning_rate": 2.895614260066684e-06, + "loss": 0.1882, + "step": 73514 + }, + { + "epoch": 2.83, + "learning_rate": 2.8949730700179535e-06, + "loss": 0.1845, + "step": 73515 + }, + { + "epoch": 2.83, + "learning_rate": 2.894331879969223e-06, + "loss": 0.2099, + "step": 73516 + }, + { + "epoch": 2.83, + "learning_rate": 2.8936906899204927e-06, + "loss": 0.2215, + "step": 73517 + }, + { + "epoch": 2.83, + "learning_rate": 2.8930494998717623e-06, + "loss": 0.1175, + "step": 73518 + }, + { + "epoch": 2.83, + "learning_rate": 2.892408309823032e-06, + "loss": 0.1582, + "step": 73519 + }, + { + "epoch": 2.83, + "learning_rate": 2.8917671197743014e-06, + "loss": 0.1726, + "step": 73520 + }, + { + "epoch": 2.83, + "learning_rate": 2.891125929725571e-06, + "loss": 0.1306, + "step": 73521 + }, + { + "epoch": 2.83, + "learning_rate": 2.8904847396768406e-06, + "loss": 0.2349, + "step": 73522 + }, + { + "epoch": 2.83, + "learning_rate": 2.88984354962811e-06, + "loss": 0.1813, + "step": 73523 + }, + { + "epoch": 2.83, + "learning_rate": 2.8892023595793794e-06, + "loss": 0.1573, + "step": 73524 + }, + { + "epoch": 2.83, + "learning_rate": 2.888561169530649e-06, + "loss": 0.2921, + "step": 73525 + }, + { + "epoch": 2.83, + "learning_rate": 2.8879199794819186e-06, + "loss": 0.2581, + "step": 73526 + }, + { + "epoch": 2.83, + "learning_rate": 2.8872787894331882e-06, + "loss": 0.2113, + "step": 73527 + }, + { + "epoch": 2.83, + "learning_rate": 2.886637599384458e-06, + "loss": 0.1239, + "step": 73528 + }, + { + "epoch": 2.83, + "learning_rate": 2.8859964093357274e-06, + "loss": 0.1997, + "step": 73529 + }, + { + "epoch": 2.83, + "learning_rate": 2.885355219286997e-06, + "loss": 0.1353, + "step": 73530 + }, + { + "epoch": 2.83, + "learning_rate": 2.8847140292382666e-06, + "loss": 0.3314, + "step": 73531 + }, + { + "epoch": 2.83, + "learning_rate": 2.8840728391895362e-06, + "loss": 0.2112, + "step": 73532 + }, + { + "epoch": 2.83, + "learning_rate": 2.8834316491408054e-06, + "loss": 0.288, + "step": 73533 + }, + { + "epoch": 2.83, + "learning_rate": 2.882790459092075e-06, + "loss": 0.2024, + "step": 73534 + }, + { + "epoch": 2.83, + "learning_rate": 2.8821492690433446e-06, + "loss": 0.2423, + "step": 73535 + }, + { + "epoch": 2.83, + "learning_rate": 2.881508078994614e-06, + "loss": 0.2081, + "step": 73536 + }, + { + "epoch": 2.83, + "learning_rate": 2.8808668889458838e-06, + "loss": 0.2422, + "step": 73537 + }, + { + "epoch": 2.83, + "learning_rate": 2.8802256988971534e-06, + "loss": 0.1001, + "step": 73538 + }, + { + "epoch": 2.83, + "learning_rate": 2.879584508848423e-06, + "loss": 0.1251, + "step": 73539 + }, + { + "epoch": 2.83, + "learning_rate": 2.8789433187996926e-06, + "loss": 0.0921, + "step": 73540 + }, + { + "epoch": 2.83, + "learning_rate": 2.878302128750962e-06, + "loss": 0.2806, + "step": 73541 + }, + { + "epoch": 2.83, + "learning_rate": 2.8776609387022318e-06, + "loss": 0.1784, + "step": 73542 + }, + { + "epoch": 2.83, + "learning_rate": 2.877019748653501e-06, + "loss": 0.1819, + "step": 73543 + }, + { + "epoch": 2.83, + "learning_rate": 2.8763785586047706e-06, + "loss": 0.1946, + "step": 73544 + }, + { + "epoch": 2.83, + "learning_rate": 2.87573736855604e-06, + "loss": 0.2162, + "step": 73545 + }, + { + "epoch": 2.83, + "learning_rate": 2.8750961785073098e-06, + "loss": 0.1651, + "step": 73546 + }, + { + "epoch": 2.83, + "learning_rate": 2.8744549884585794e-06, + "loss": 0.2469, + "step": 73547 + }, + { + "epoch": 2.83, + "learning_rate": 2.873813798409849e-06, + "loss": 0.1703, + "step": 73548 + }, + { + "epoch": 2.83, + "learning_rate": 2.8731726083611186e-06, + "loss": 0.1365, + "step": 73549 + }, + { + "epoch": 2.83, + "learning_rate": 2.872531418312388e-06, + "loss": 0.1577, + "step": 73550 + }, + { + "epoch": 2.83, + "learning_rate": 2.8718902282636577e-06, + "loss": 0.3325, + "step": 73551 + }, + { + "epoch": 2.83, + "learning_rate": 2.871249038214927e-06, + "loss": 0.2435, + "step": 73552 + }, + { + "epoch": 2.83, + "learning_rate": 2.8706078481661965e-06, + "loss": 0.2826, + "step": 73553 + }, + { + "epoch": 2.83, + "learning_rate": 2.869966658117466e-06, + "loss": 0.2553, + "step": 73554 + }, + { + "epoch": 2.83, + "learning_rate": 2.8693254680687357e-06, + "loss": 0.095, + "step": 73555 + }, + { + "epoch": 2.83, + "learning_rate": 2.8686842780200053e-06, + "loss": 0.2118, + "step": 73556 + }, + { + "epoch": 2.83, + "learning_rate": 2.868043087971275e-06, + "loss": 0.2155, + "step": 73557 + }, + { + "epoch": 2.83, + "learning_rate": 2.8674018979225445e-06, + "loss": 0.2255, + "step": 73558 + }, + { + "epoch": 2.83, + "learning_rate": 2.866760707873814e-06, + "loss": 0.202, + "step": 73559 + }, + { + "epoch": 2.83, + "learning_rate": 2.8661195178250837e-06, + "loss": 0.2441, + "step": 73560 + }, + { + "epoch": 2.83, + "learning_rate": 2.8654783277763533e-06, + "loss": 0.2366, + "step": 73561 + }, + { + "epoch": 2.83, + "learning_rate": 2.8648371377276225e-06, + "loss": 0.2316, + "step": 73562 + }, + { + "epoch": 2.83, + "learning_rate": 2.864195947678892e-06, + "loss": 0.1815, + "step": 73563 + }, + { + "epoch": 2.83, + "learning_rate": 2.8635547576301617e-06, + "loss": 0.1865, + "step": 73564 + }, + { + "epoch": 2.83, + "learning_rate": 2.8629135675814313e-06, + "loss": 0.1552, + "step": 73565 + }, + { + "epoch": 2.83, + "learning_rate": 2.862272377532701e-06, + "loss": 0.1037, + "step": 73566 + }, + { + "epoch": 2.83, + "learning_rate": 2.8616311874839705e-06, + "loss": 0.1477, + "step": 73567 + }, + { + "epoch": 2.83, + "learning_rate": 2.86098999743524e-06, + "loss": 0.2086, + "step": 73568 + }, + { + "epoch": 2.83, + "learning_rate": 2.8603488073865097e-06, + "loss": 0.2905, + "step": 73569 + }, + { + "epoch": 2.83, + "learning_rate": 2.8597076173377793e-06, + "loss": 0.3718, + "step": 73570 + }, + { + "epoch": 2.83, + "learning_rate": 2.859066427289049e-06, + "loss": 0.206, + "step": 73571 + }, + { + "epoch": 2.83, + "learning_rate": 2.858425237240318e-06, + "loss": 0.2679, + "step": 73572 + }, + { + "epoch": 2.83, + "learning_rate": 2.8577840471915877e-06, + "loss": 0.2196, + "step": 73573 + }, + { + "epoch": 2.83, + "learning_rate": 2.8571428571428573e-06, + "loss": 0.2223, + "step": 73574 + }, + { + "epoch": 2.83, + "learning_rate": 2.856501667094127e-06, + "loss": 0.3579, + "step": 73575 + }, + { + "epoch": 2.83, + "learning_rate": 2.8558604770453965e-06, + "loss": 0.1262, + "step": 73576 + }, + { + "epoch": 2.83, + "learning_rate": 2.855219286996666e-06, + "loss": 0.3011, + "step": 73577 + }, + { + "epoch": 2.83, + "learning_rate": 2.8545780969479357e-06, + "loss": 0.1371, + "step": 73578 + }, + { + "epoch": 2.83, + "learning_rate": 2.8539369068992053e-06, + "loss": 0.1696, + "step": 73579 + }, + { + "epoch": 2.83, + "learning_rate": 2.853295716850475e-06, + "loss": 0.1363, + "step": 73580 + }, + { + "epoch": 2.83, + "learning_rate": 2.8526545268017444e-06, + "loss": 0.1017, + "step": 73581 + }, + { + "epoch": 2.83, + "learning_rate": 2.8520133367530136e-06, + "loss": 0.1536, + "step": 73582 + }, + { + "epoch": 2.83, + "learning_rate": 2.8513721467042832e-06, + "loss": 0.2776, + "step": 73583 + }, + { + "epoch": 2.83, + "learning_rate": 2.850730956655553e-06, + "loss": 0.1445, + "step": 73584 + }, + { + "epoch": 2.83, + "learning_rate": 2.8500897666068224e-06, + "loss": 0.1555, + "step": 73585 + }, + { + "epoch": 2.83, + "learning_rate": 2.849448576558092e-06, + "loss": 0.2063, + "step": 73586 + }, + { + "epoch": 2.83, + "learning_rate": 2.8488073865093616e-06, + "loss": 0.1933, + "step": 73587 + }, + { + "epoch": 2.83, + "learning_rate": 2.8481661964606312e-06, + "loss": 0.2304, + "step": 73588 + }, + { + "epoch": 2.83, + "learning_rate": 2.847525006411901e-06, + "loss": 0.2191, + "step": 73589 + }, + { + "epoch": 2.83, + "learning_rate": 2.8468838163631704e-06, + "loss": 0.3663, + "step": 73590 + }, + { + "epoch": 2.83, + "learning_rate": 2.84624262631444e-06, + "loss": 0.1681, + "step": 73591 + }, + { + "epoch": 2.83, + "learning_rate": 2.845601436265709e-06, + "loss": 0.1698, + "step": 73592 + }, + { + "epoch": 2.83, + "learning_rate": 2.844960246216979e-06, + "loss": 0.2447, + "step": 73593 + }, + { + "epoch": 2.83, + "learning_rate": 2.8443190561682484e-06, + "loss": 0.2767, + "step": 73594 + }, + { + "epoch": 2.83, + "learning_rate": 2.843677866119518e-06, + "loss": 0.17, + "step": 73595 + }, + { + "epoch": 2.83, + "learning_rate": 2.8430366760707876e-06, + "loss": 0.1805, + "step": 73596 + }, + { + "epoch": 2.83, + "learning_rate": 2.842395486022057e-06, + "loss": 0.1779, + "step": 73597 + }, + { + "epoch": 2.83, + "learning_rate": 2.8417542959733268e-06, + "loss": 0.0852, + "step": 73598 + }, + { + "epoch": 2.83, + "learning_rate": 2.8411131059245964e-06, + "loss": 0.1759, + "step": 73599 + }, + { + "epoch": 2.83, + "learning_rate": 2.840471915875866e-06, + "loss": 0.2217, + "step": 73600 + }, + { + "epoch": 2.83, + "learning_rate": 2.8398307258271356e-06, + "loss": 0.2531, + "step": 73601 + }, + { + "epoch": 2.83, + "learning_rate": 2.8391895357784048e-06, + "loss": 0.3208, + "step": 73602 + }, + { + "epoch": 2.83, + "learning_rate": 2.8385483457296744e-06, + "loss": 0.0902, + "step": 73603 + }, + { + "epoch": 2.83, + "learning_rate": 2.837907155680944e-06, + "loss": 0.1931, + "step": 73604 + }, + { + "epoch": 2.83, + "learning_rate": 2.8372659656322136e-06, + "loss": 0.1363, + "step": 73605 + }, + { + "epoch": 2.83, + "learning_rate": 2.836624775583483e-06, + "loss": 0.0924, + "step": 73606 + }, + { + "epoch": 2.83, + "learning_rate": 2.8359835855347528e-06, + "loss": 0.2564, + "step": 73607 + }, + { + "epoch": 2.83, + "learning_rate": 2.8353423954860224e-06, + "loss": 0.0843, + "step": 73608 + }, + { + "epoch": 2.83, + "learning_rate": 2.834701205437292e-06, + "loss": 0.1853, + "step": 73609 + }, + { + "epoch": 2.83, + "learning_rate": 2.8340600153885616e-06, + "loss": 0.232, + "step": 73610 + }, + { + "epoch": 2.83, + "learning_rate": 2.8334188253398307e-06, + "loss": 0.1424, + "step": 73611 + }, + { + "epoch": 2.83, + "learning_rate": 2.8327776352911003e-06, + "loss": 0.1318, + "step": 73612 + }, + { + "epoch": 2.83, + "learning_rate": 2.83213644524237e-06, + "loss": 0.0825, + "step": 73613 + }, + { + "epoch": 2.83, + "learning_rate": 2.8314952551936395e-06, + "loss": 0.1507, + "step": 73614 + }, + { + "epoch": 2.83, + "learning_rate": 2.830854065144909e-06, + "loss": 0.1761, + "step": 73615 + }, + { + "epoch": 2.83, + "learning_rate": 2.8302128750961787e-06, + "loss": 0.2755, + "step": 73616 + }, + { + "epoch": 2.83, + "learning_rate": 2.8295716850474483e-06, + "loss": 0.1552, + "step": 73617 + }, + { + "epoch": 2.83, + "learning_rate": 2.828930494998718e-06, + "loss": 0.3255, + "step": 73618 + }, + { + "epoch": 2.83, + "learning_rate": 2.8282893049499875e-06, + "loss": 0.2876, + "step": 73619 + }, + { + "epoch": 2.83, + "learning_rate": 2.827648114901257e-06, + "loss": 0.3653, + "step": 73620 + }, + { + "epoch": 2.83, + "learning_rate": 2.8270069248525263e-06, + "loss": 0.2617, + "step": 73621 + }, + { + "epoch": 2.83, + "learning_rate": 2.826365734803796e-06, + "loss": 0.1435, + "step": 73622 + }, + { + "epoch": 2.83, + "learning_rate": 2.8257245447550655e-06, + "loss": 0.1548, + "step": 73623 + }, + { + "epoch": 2.83, + "learning_rate": 2.825083354706335e-06, + "loss": 0.2398, + "step": 73624 + }, + { + "epoch": 2.83, + "learning_rate": 2.8244421646576047e-06, + "loss": 0.1774, + "step": 73625 + }, + { + "epoch": 2.83, + "learning_rate": 2.8238009746088743e-06, + "loss": 0.1717, + "step": 73626 + }, + { + "epoch": 2.83, + "learning_rate": 2.823159784560144e-06, + "loss": 0.2535, + "step": 73627 + }, + { + "epoch": 2.83, + "learning_rate": 2.8225185945114135e-06, + "loss": 0.1598, + "step": 73628 + }, + { + "epoch": 2.83, + "learning_rate": 2.821877404462683e-06, + "loss": 0.2264, + "step": 73629 + }, + { + "epoch": 2.83, + "learning_rate": 2.8212362144139527e-06, + "loss": 0.0931, + "step": 73630 + }, + { + "epoch": 2.83, + "learning_rate": 2.820595024365222e-06, + "loss": 0.3875, + "step": 73631 + }, + { + "epoch": 2.83, + "learning_rate": 2.8199538343164915e-06, + "loss": 0.1537, + "step": 73632 + }, + { + "epoch": 2.83, + "learning_rate": 2.819312644267761e-06, + "loss": 0.2472, + "step": 73633 + }, + { + "epoch": 2.83, + "learning_rate": 2.8186714542190307e-06, + "loss": 0.0936, + "step": 73634 + }, + { + "epoch": 2.83, + "learning_rate": 2.8180302641703003e-06, + "loss": 0.0818, + "step": 73635 + }, + { + "epoch": 2.83, + "learning_rate": 2.81738907412157e-06, + "loss": 0.3516, + "step": 73636 + }, + { + "epoch": 2.83, + "learning_rate": 2.8167478840728395e-06, + "loss": 0.1301, + "step": 73637 + }, + { + "epoch": 2.83, + "learning_rate": 2.816106694024109e-06, + "loss": 0.1608, + "step": 73638 + }, + { + "epoch": 2.83, + "learning_rate": 2.8154655039753787e-06, + "loss": 0.2241, + "step": 73639 + }, + { + "epoch": 2.83, + "learning_rate": 2.8148243139266483e-06, + "loss": 0.0886, + "step": 73640 + }, + { + "epoch": 2.83, + "learning_rate": 2.8141831238779174e-06, + "loss": 0.1282, + "step": 73641 + }, + { + "epoch": 2.83, + "learning_rate": 2.813541933829187e-06, + "loss": 0.1561, + "step": 73642 + }, + { + "epoch": 2.83, + "learning_rate": 2.8129007437804566e-06, + "loss": 0.2633, + "step": 73643 + }, + { + "epoch": 2.83, + "learning_rate": 2.8122595537317262e-06, + "loss": 0.1347, + "step": 73644 + }, + { + "epoch": 2.83, + "learning_rate": 2.811618363682996e-06, + "loss": 0.1808, + "step": 73645 + }, + { + "epoch": 2.83, + "learning_rate": 2.8109771736342654e-06, + "loss": 0.1929, + "step": 73646 + }, + { + "epoch": 2.83, + "learning_rate": 2.810335983585535e-06, + "loss": 0.2563, + "step": 73647 + }, + { + "epoch": 2.83, + "learning_rate": 2.8096947935368046e-06, + "loss": 0.2637, + "step": 73648 + }, + { + "epoch": 2.83, + "learning_rate": 2.8090536034880742e-06, + "loss": 0.2201, + "step": 73649 + }, + { + "epoch": 2.83, + "learning_rate": 2.808412413439344e-06, + "loss": 0.1806, + "step": 73650 + }, + { + "epoch": 2.83, + "learning_rate": 2.807771223390613e-06, + "loss": 0.1186, + "step": 73651 + }, + { + "epoch": 2.83, + "learning_rate": 2.8071300333418826e-06, + "loss": 0.125, + "step": 73652 + }, + { + "epoch": 2.83, + "learning_rate": 2.806488843293152e-06, + "loss": 0.1138, + "step": 73653 + }, + { + "epoch": 2.83, + "learning_rate": 2.805847653244422e-06, + "loss": 0.2154, + "step": 73654 + }, + { + "epoch": 2.83, + "learning_rate": 2.8052064631956914e-06, + "loss": 0.2773, + "step": 73655 + }, + { + "epoch": 2.83, + "learning_rate": 2.804565273146961e-06, + "loss": 0.1657, + "step": 73656 + }, + { + "epoch": 2.83, + "learning_rate": 2.8039240830982306e-06, + "loss": 0.2229, + "step": 73657 + }, + { + "epoch": 2.83, + "learning_rate": 2.8032828930495e-06, + "loss": 0.1917, + "step": 73658 + }, + { + "epoch": 2.83, + "learning_rate": 2.8026417030007698e-06, + "loss": 0.2535, + "step": 73659 + }, + { + "epoch": 2.83, + "learning_rate": 2.8020005129520394e-06, + "loss": 0.1917, + "step": 73660 + }, + { + "epoch": 2.83, + "learning_rate": 2.8013593229033086e-06, + "loss": 0.3851, + "step": 73661 + }, + { + "epoch": 2.83, + "learning_rate": 2.800718132854578e-06, + "loss": 0.1192, + "step": 73662 + }, + { + "epoch": 2.83, + "learning_rate": 2.8000769428058478e-06, + "loss": 0.1965, + "step": 73663 + }, + { + "epoch": 2.83, + "learning_rate": 2.7994357527571174e-06, + "loss": 0.1096, + "step": 73664 + }, + { + "epoch": 2.83, + "learning_rate": 2.798794562708387e-06, + "loss": 0.2806, + "step": 73665 + }, + { + "epoch": 2.83, + "learning_rate": 2.7981533726596566e-06, + "loss": 0.2214, + "step": 73666 + }, + { + "epoch": 2.83, + "learning_rate": 2.797512182610926e-06, + "loss": 0.1821, + "step": 73667 + }, + { + "epoch": 2.83, + "learning_rate": 2.7968709925621958e-06, + "loss": 0.1696, + "step": 73668 + }, + { + "epoch": 2.83, + "learning_rate": 2.7962298025134654e-06, + "loss": 0.3222, + "step": 73669 + }, + { + "epoch": 2.83, + "learning_rate": 2.7955886124647345e-06, + "loss": 0.229, + "step": 73670 + }, + { + "epoch": 2.83, + "learning_rate": 2.794947422416004e-06, + "loss": 0.1442, + "step": 73671 + }, + { + "epoch": 2.83, + "learning_rate": 2.7943062323672737e-06, + "loss": 0.222, + "step": 73672 + }, + { + "epoch": 2.83, + "learning_rate": 2.7936650423185433e-06, + "loss": 0.1091, + "step": 73673 + }, + { + "epoch": 2.83, + "learning_rate": 2.793023852269813e-06, + "loss": 0.2273, + "step": 73674 + }, + { + "epoch": 2.83, + "learning_rate": 2.7923826622210825e-06, + "loss": 0.0759, + "step": 73675 + }, + { + "epoch": 2.83, + "learning_rate": 2.791741472172352e-06, + "loss": 0.2011, + "step": 73676 + }, + { + "epoch": 2.83, + "learning_rate": 2.7911002821236217e-06, + "loss": 0.2866, + "step": 73677 + }, + { + "epoch": 2.83, + "learning_rate": 2.7904590920748913e-06, + "loss": 0.1291, + "step": 73678 + }, + { + "epoch": 2.83, + "learning_rate": 2.789817902026161e-06, + "loss": 0.1664, + "step": 73679 + }, + { + "epoch": 2.83, + "learning_rate": 2.78917671197743e-06, + "loss": 0.2104, + "step": 73680 + }, + { + "epoch": 2.83, + "learning_rate": 2.7885355219286997e-06, + "loss": 0.0988, + "step": 73681 + }, + { + "epoch": 2.83, + "learning_rate": 2.7878943318799693e-06, + "loss": 0.2887, + "step": 73682 + }, + { + "epoch": 2.83, + "learning_rate": 2.787253141831239e-06, + "loss": 0.222, + "step": 73683 + }, + { + "epoch": 2.83, + "learning_rate": 2.7866119517825085e-06, + "loss": 0.1412, + "step": 73684 + }, + { + "epoch": 2.83, + "learning_rate": 2.785970761733778e-06, + "loss": 0.2285, + "step": 73685 + }, + { + "epoch": 2.83, + "learning_rate": 2.7853295716850477e-06, + "loss": 0.2471, + "step": 73686 + }, + { + "epoch": 2.83, + "learning_rate": 2.7846883816363173e-06, + "loss": 0.0924, + "step": 73687 + }, + { + "epoch": 2.83, + "learning_rate": 2.784047191587587e-06, + "loss": 0.2546, + "step": 73688 + }, + { + "epoch": 2.83, + "learning_rate": 2.7834060015388565e-06, + "loss": 0.1921, + "step": 73689 + }, + { + "epoch": 2.83, + "learning_rate": 2.7827648114901257e-06, + "loss": 0.2156, + "step": 73690 + }, + { + "epoch": 2.83, + "learning_rate": 2.7821236214413953e-06, + "loss": 0.195, + "step": 73691 + }, + { + "epoch": 2.83, + "learning_rate": 2.781482431392665e-06, + "loss": 0.0877, + "step": 73692 + }, + { + "epoch": 2.83, + "learning_rate": 2.7808412413439345e-06, + "loss": 0.111, + "step": 73693 + }, + { + "epoch": 2.83, + "learning_rate": 2.780200051295204e-06, + "loss": 0.258, + "step": 73694 + }, + { + "epoch": 2.83, + "learning_rate": 2.7795588612464737e-06, + "loss": 0.2834, + "step": 73695 + }, + { + "epoch": 2.83, + "learning_rate": 2.7789176711977433e-06, + "loss": 0.1047, + "step": 73696 + }, + { + "epoch": 2.83, + "learning_rate": 2.778276481149013e-06, + "loss": 0.193, + "step": 73697 + }, + { + "epoch": 2.83, + "learning_rate": 2.7776352911002825e-06, + "loss": 0.1719, + "step": 73698 + }, + { + "epoch": 2.83, + "learning_rate": 2.776994101051552e-06, + "loss": 0.3227, + "step": 73699 + }, + { + "epoch": 2.83, + "learning_rate": 2.7763529110028212e-06, + "loss": 0.2159, + "step": 73700 + }, + { + "epoch": 2.83, + "learning_rate": 2.775711720954091e-06, + "loss": 0.2344, + "step": 73701 + }, + { + "epoch": 2.83, + "learning_rate": 2.7750705309053604e-06, + "loss": 0.1628, + "step": 73702 + }, + { + "epoch": 2.83, + "learning_rate": 2.77442934085663e-06, + "loss": 0.2082, + "step": 73703 + }, + { + "epoch": 2.83, + "learning_rate": 2.7737881508078996e-06, + "loss": 0.1746, + "step": 73704 + }, + { + "epoch": 2.83, + "learning_rate": 2.7731469607591692e-06, + "loss": 0.1511, + "step": 73705 + }, + { + "epoch": 2.83, + "learning_rate": 2.772505770710439e-06, + "loss": 0.2418, + "step": 73706 + }, + { + "epoch": 2.83, + "learning_rate": 2.7718645806617084e-06, + "loss": 0.2083, + "step": 73707 + }, + { + "epoch": 2.83, + "learning_rate": 2.771223390612978e-06, + "loss": 0.1614, + "step": 73708 + }, + { + "epoch": 2.83, + "learning_rate": 2.7705822005642476e-06, + "loss": 0.0994, + "step": 73709 + }, + { + "epoch": 2.83, + "learning_rate": 2.769941010515517e-06, + "loss": 0.2786, + "step": 73710 + }, + { + "epoch": 2.83, + "learning_rate": 2.7692998204667864e-06, + "loss": 0.3065, + "step": 73711 + }, + { + "epoch": 2.83, + "learning_rate": 2.768658630418056e-06, + "loss": 0.1586, + "step": 73712 + }, + { + "epoch": 2.83, + "learning_rate": 2.7680174403693256e-06, + "loss": 0.1462, + "step": 73713 + }, + { + "epoch": 2.83, + "learning_rate": 2.767376250320595e-06, + "loss": 0.1723, + "step": 73714 + }, + { + "epoch": 2.83, + "learning_rate": 2.766735060271865e-06, + "loss": 0.2436, + "step": 73715 + }, + { + "epoch": 2.83, + "learning_rate": 2.7660938702231344e-06, + "loss": 0.1425, + "step": 73716 + }, + { + "epoch": 2.83, + "learning_rate": 2.765452680174404e-06, + "loss": 0.1637, + "step": 73717 + }, + { + "epoch": 2.83, + "learning_rate": 2.7648114901256736e-06, + "loss": 0.0934, + "step": 73718 + }, + { + "epoch": 2.83, + "learning_rate": 2.764170300076943e-06, + "loss": 0.1931, + "step": 73719 + }, + { + "epoch": 2.83, + "learning_rate": 2.7635291100282124e-06, + "loss": 0.2302, + "step": 73720 + }, + { + "epoch": 2.83, + "learning_rate": 2.762887919979482e-06, + "loss": 0.1616, + "step": 73721 + }, + { + "epoch": 2.83, + "learning_rate": 2.7622467299307516e-06, + "loss": 0.3084, + "step": 73722 + }, + { + "epoch": 2.83, + "learning_rate": 2.761605539882021e-06, + "loss": 0.1619, + "step": 73723 + }, + { + "epoch": 2.83, + "learning_rate": 2.7609643498332908e-06, + "loss": 0.1691, + "step": 73724 + }, + { + "epoch": 2.83, + "learning_rate": 2.7603231597845604e-06, + "loss": 0.152, + "step": 73725 + }, + { + "epoch": 2.83, + "learning_rate": 2.75968196973583e-06, + "loss": 0.2798, + "step": 73726 + }, + { + "epoch": 2.83, + "learning_rate": 2.7590407796870996e-06, + "loss": 0.1217, + "step": 73727 + }, + { + "epoch": 2.83, + "learning_rate": 2.758399589638369e-06, + "loss": 0.09, + "step": 73728 + }, + { + "epoch": 2.83, + "learning_rate": 2.7577583995896383e-06, + "loss": 0.1244, + "step": 73729 + }, + { + "epoch": 2.83, + "learning_rate": 2.757117209540908e-06, + "loss": 0.1756, + "step": 73730 + }, + { + "epoch": 2.83, + "learning_rate": 2.7564760194921775e-06, + "loss": 0.1027, + "step": 73731 + }, + { + "epoch": 2.83, + "learning_rate": 2.755834829443447e-06, + "loss": 0.2484, + "step": 73732 + }, + { + "epoch": 2.83, + "learning_rate": 2.7551936393947167e-06, + "loss": 0.437, + "step": 73733 + }, + { + "epoch": 2.83, + "learning_rate": 2.7545524493459863e-06, + "loss": 0.1852, + "step": 73734 + }, + { + "epoch": 2.83, + "learning_rate": 2.753911259297256e-06, + "loss": 0.2567, + "step": 73735 + }, + { + "epoch": 2.83, + "learning_rate": 2.7532700692485255e-06, + "loss": 0.1634, + "step": 73736 + }, + { + "epoch": 2.83, + "learning_rate": 2.752628879199795e-06, + "loss": 0.3073, + "step": 73737 + }, + { + "epoch": 2.83, + "learning_rate": 2.7519876891510647e-06, + "loss": 0.2843, + "step": 73738 + }, + { + "epoch": 2.84, + "learning_rate": 2.751346499102334e-06, + "loss": 0.2631, + "step": 73739 + }, + { + "epoch": 2.84, + "learning_rate": 2.7507053090536035e-06, + "loss": 0.2588, + "step": 73740 + }, + { + "epoch": 2.84, + "learning_rate": 2.750064119004873e-06, + "loss": 0.1216, + "step": 73741 + }, + { + "epoch": 2.84, + "learning_rate": 2.7494229289561427e-06, + "loss": 0.0991, + "step": 73742 + }, + { + "epoch": 2.84, + "learning_rate": 2.7487817389074123e-06, + "loss": 0.154, + "step": 73743 + }, + { + "epoch": 2.84, + "learning_rate": 2.748140548858682e-06, + "loss": 0.4217, + "step": 73744 + }, + { + "epoch": 2.84, + "learning_rate": 2.7474993588099515e-06, + "loss": 0.315, + "step": 73745 + }, + { + "epoch": 2.84, + "learning_rate": 2.746858168761221e-06, + "loss": 0.2641, + "step": 73746 + }, + { + "epoch": 2.84, + "learning_rate": 2.7462169787124907e-06, + "loss": 0.173, + "step": 73747 + }, + { + "epoch": 2.84, + "learning_rate": 2.7455757886637603e-06, + "loss": 0.3805, + "step": 73748 + }, + { + "epoch": 2.84, + "learning_rate": 2.7449345986150295e-06, + "loss": 0.0942, + "step": 73749 + }, + { + "epoch": 2.84, + "learning_rate": 2.744293408566299e-06, + "loss": 0.1396, + "step": 73750 + }, + { + "epoch": 2.84, + "learning_rate": 2.7436522185175687e-06, + "loss": 0.1537, + "step": 73751 + }, + { + "epoch": 2.84, + "learning_rate": 2.7430110284688383e-06, + "loss": 0.1533, + "step": 73752 + }, + { + "epoch": 2.84, + "learning_rate": 2.742369838420108e-06, + "loss": 0.2465, + "step": 73753 + }, + { + "epoch": 2.84, + "learning_rate": 2.7417286483713775e-06, + "loss": 0.2331, + "step": 73754 + }, + { + "epoch": 2.84, + "learning_rate": 2.741087458322647e-06, + "loss": 0.2043, + "step": 73755 + }, + { + "epoch": 2.84, + "learning_rate": 2.7404462682739167e-06, + "loss": 0.1583, + "step": 73756 + }, + { + "epoch": 2.84, + "learning_rate": 2.7398050782251863e-06, + "loss": 0.1362, + "step": 73757 + }, + { + "epoch": 2.84, + "learning_rate": 2.739163888176456e-06, + "loss": 0.2354, + "step": 73758 + }, + { + "epoch": 2.84, + "learning_rate": 2.738522698127725e-06, + "loss": 0.197, + "step": 73759 + }, + { + "epoch": 2.84, + "learning_rate": 2.7378815080789946e-06, + "loss": 0.1291, + "step": 73760 + }, + { + "epoch": 2.84, + "learning_rate": 2.7372403180302642e-06, + "loss": 0.2303, + "step": 73761 + }, + { + "epoch": 2.84, + "learning_rate": 2.736599127981534e-06, + "loss": 0.2635, + "step": 73762 + }, + { + "epoch": 2.84, + "learning_rate": 2.7359579379328034e-06, + "loss": 0.1457, + "step": 73763 + }, + { + "epoch": 2.84, + "learning_rate": 2.735316747884073e-06, + "loss": 0.2417, + "step": 73764 + }, + { + "epoch": 2.84, + "learning_rate": 2.7346755578353426e-06, + "loss": 0.2244, + "step": 73765 + }, + { + "epoch": 2.84, + "learning_rate": 2.7340343677866122e-06, + "loss": 0.1779, + "step": 73766 + }, + { + "epoch": 2.84, + "learning_rate": 2.733393177737882e-06, + "loss": 0.2636, + "step": 73767 + }, + { + "epoch": 2.84, + "learning_rate": 2.7327519876891514e-06, + "loss": 0.2235, + "step": 73768 + }, + { + "epoch": 2.84, + "learning_rate": 2.7321107976404206e-06, + "loss": 0.214, + "step": 73769 + }, + { + "epoch": 2.84, + "learning_rate": 2.73146960759169e-06, + "loss": 0.2123, + "step": 73770 + }, + { + "epoch": 2.84, + "learning_rate": 2.73082841754296e-06, + "loss": 0.1812, + "step": 73771 + }, + { + "epoch": 2.84, + "learning_rate": 2.7301872274942294e-06, + "loss": 0.2063, + "step": 73772 + }, + { + "epoch": 2.84, + "learning_rate": 2.729546037445499e-06, + "loss": 0.1506, + "step": 73773 + }, + { + "epoch": 2.84, + "learning_rate": 2.7289048473967686e-06, + "loss": 0.148, + "step": 73774 + }, + { + "epoch": 2.84, + "learning_rate": 2.728263657348038e-06, + "loss": 0.2006, + "step": 73775 + }, + { + "epoch": 2.84, + "learning_rate": 2.727622467299308e-06, + "loss": 0.2139, + "step": 73776 + }, + { + "epoch": 2.84, + "learning_rate": 2.7269812772505774e-06, + "loss": 0.1997, + "step": 73777 + }, + { + "epoch": 2.84, + "learning_rate": 2.726340087201847e-06, + "loss": 0.1648, + "step": 73778 + }, + { + "epoch": 2.84, + "learning_rate": 2.725698897153116e-06, + "loss": 0.2411, + "step": 73779 + }, + { + "epoch": 2.84, + "learning_rate": 2.7250577071043858e-06, + "loss": 0.2057, + "step": 73780 + }, + { + "epoch": 2.84, + "learning_rate": 2.7244165170556554e-06, + "loss": 0.1147, + "step": 73781 + }, + { + "epoch": 2.84, + "learning_rate": 2.723775327006925e-06, + "loss": 0.1956, + "step": 73782 + }, + { + "epoch": 2.84, + "learning_rate": 2.7231341369581946e-06, + "loss": 0.1356, + "step": 73783 + }, + { + "epoch": 2.84, + "learning_rate": 2.722492946909464e-06, + "loss": 0.2753, + "step": 73784 + }, + { + "epoch": 2.84, + "learning_rate": 2.7218517568607338e-06, + "loss": 0.1662, + "step": 73785 + }, + { + "epoch": 2.84, + "learning_rate": 2.7212105668120034e-06, + "loss": 0.2362, + "step": 73786 + }, + { + "epoch": 2.84, + "learning_rate": 2.720569376763273e-06, + "loss": 0.1978, + "step": 73787 + }, + { + "epoch": 2.84, + "learning_rate": 2.719928186714542e-06, + "loss": 0.1302, + "step": 73788 + }, + { + "epoch": 2.84, + "learning_rate": 2.7192869966658117e-06, + "loss": 0.3625, + "step": 73789 + }, + { + "epoch": 2.84, + "learning_rate": 2.7186458066170813e-06, + "loss": 0.1817, + "step": 73790 + }, + { + "epoch": 2.84, + "learning_rate": 2.718004616568351e-06, + "loss": 0.1445, + "step": 73791 + }, + { + "epoch": 2.84, + "learning_rate": 2.7173634265196205e-06, + "loss": 0.1735, + "step": 73792 + }, + { + "epoch": 2.84, + "learning_rate": 2.71672223647089e-06, + "loss": 0.3348, + "step": 73793 + }, + { + "epoch": 2.84, + "learning_rate": 2.7160810464221597e-06, + "loss": 0.1348, + "step": 73794 + }, + { + "epoch": 2.84, + "learning_rate": 2.7154398563734293e-06, + "loss": 0.3321, + "step": 73795 + }, + { + "epoch": 2.84, + "learning_rate": 2.714798666324699e-06, + "loss": 0.1111, + "step": 73796 + }, + { + "epoch": 2.84, + "learning_rate": 2.7141574762759685e-06, + "loss": 0.2075, + "step": 73797 + }, + { + "epoch": 2.84, + "learning_rate": 2.7135162862272377e-06, + "loss": 0.2043, + "step": 73798 + }, + { + "epoch": 2.84, + "learning_rate": 2.7128750961785073e-06, + "loss": 0.218, + "step": 73799 + }, + { + "epoch": 2.84, + "learning_rate": 2.712233906129777e-06, + "loss": 0.2151, + "step": 73800 + }, + { + "epoch": 2.84, + "learning_rate": 2.7115927160810465e-06, + "loss": 0.2753, + "step": 73801 + }, + { + "epoch": 2.84, + "learning_rate": 2.710951526032316e-06, + "loss": 0.2114, + "step": 73802 + }, + { + "epoch": 2.84, + "learning_rate": 2.7103103359835857e-06, + "loss": 0.2003, + "step": 73803 + }, + { + "epoch": 2.84, + "learning_rate": 2.7096691459348553e-06, + "loss": 0.299, + "step": 73804 + }, + { + "epoch": 2.84, + "learning_rate": 2.709027955886125e-06, + "loss": 0.2504, + "step": 73805 + }, + { + "epoch": 2.84, + "learning_rate": 2.7083867658373945e-06, + "loss": 0.3211, + "step": 73806 + }, + { + "epoch": 2.84, + "learning_rate": 2.707745575788664e-06, + "loss": 0.1479, + "step": 73807 + }, + { + "epoch": 2.84, + "learning_rate": 2.7071043857399333e-06, + "loss": 0.2786, + "step": 73808 + }, + { + "epoch": 2.84, + "learning_rate": 2.706463195691203e-06, + "loss": 0.1782, + "step": 73809 + }, + { + "epoch": 2.84, + "learning_rate": 2.7058220056424725e-06, + "loss": 0.2586, + "step": 73810 + }, + { + "epoch": 2.84, + "learning_rate": 2.705180815593742e-06, + "loss": 0.1546, + "step": 73811 + }, + { + "epoch": 2.84, + "learning_rate": 2.7045396255450117e-06, + "loss": 0.1337, + "step": 73812 + }, + { + "epoch": 2.84, + "learning_rate": 2.7038984354962813e-06, + "loss": 0.13, + "step": 73813 + }, + { + "epoch": 2.84, + "learning_rate": 2.703257245447551e-06, + "loss": 0.1699, + "step": 73814 + }, + { + "epoch": 2.84, + "learning_rate": 2.7026160553988205e-06, + "loss": 0.0649, + "step": 73815 + }, + { + "epoch": 2.84, + "learning_rate": 2.70197486535009e-06, + "loss": 0.2682, + "step": 73816 + }, + { + "epoch": 2.84, + "learning_rate": 2.7013336753013597e-06, + "loss": 0.2871, + "step": 73817 + }, + { + "epoch": 2.84, + "learning_rate": 2.700692485252629e-06, + "loss": 0.2872, + "step": 73818 + }, + { + "epoch": 2.84, + "learning_rate": 2.7000512952038984e-06, + "loss": 0.3002, + "step": 73819 + }, + { + "epoch": 2.84, + "learning_rate": 2.699410105155168e-06, + "loss": 0.2533, + "step": 73820 + }, + { + "epoch": 2.84, + "learning_rate": 2.6987689151064376e-06, + "loss": 0.2676, + "step": 73821 + }, + { + "epoch": 2.84, + "learning_rate": 2.6981277250577072e-06, + "loss": 0.2249, + "step": 73822 + }, + { + "epoch": 2.84, + "learning_rate": 2.697486535008977e-06, + "loss": 0.157, + "step": 73823 + }, + { + "epoch": 2.84, + "learning_rate": 2.6968453449602464e-06, + "loss": 0.1715, + "step": 73824 + }, + { + "epoch": 2.84, + "learning_rate": 2.696204154911516e-06, + "loss": 0.2113, + "step": 73825 + }, + { + "epoch": 2.84, + "learning_rate": 2.6955629648627856e-06, + "loss": 0.2069, + "step": 73826 + }, + { + "epoch": 2.84, + "learning_rate": 2.6949217748140552e-06, + "loss": 0.2021, + "step": 73827 + }, + { + "epoch": 2.84, + "learning_rate": 2.6942805847653244e-06, + "loss": 0.4101, + "step": 73828 + }, + { + "epoch": 2.84, + "learning_rate": 2.693639394716594e-06, + "loss": 0.2422, + "step": 73829 + }, + { + "epoch": 2.84, + "learning_rate": 2.6929982046678636e-06, + "loss": 0.1304, + "step": 73830 + }, + { + "epoch": 2.84, + "learning_rate": 2.692357014619133e-06, + "loss": 0.1244, + "step": 73831 + }, + { + "epoch": 2.84, + "learning_rate": 2.691715824570403e-06, + "loss": 0.3007, + "step": 73832 + }, + { + "epoch": 2.84, + "learning_rate": 2.6910746345216724e-06, + "loss": 0.1511, + "step": 73833 + }, + { + "epoch": 2.84, + "learning_rate": 2.690433444472942e-06, + "loss": 0.2107, + "step": 73834 + }, + { + "epoch": 2.84, + "learning_rate": 2.6897922544242116e-06, + "loss": 0.2466, + "step": 73835 + }, + { + "epoch": 2.84, + "learning_rate": 2.689151064375481e-06, + "loss": 0.2162, + "step": 73836 + }, + { + "epoch": 2.84, + "learning_rate": 2.688509874326751e-06, + "loss": 0.4273, + "step": 73837 + }, + { + "epoch": 2.84, + "learning_rate": 2.68786868427802e-06, + "loss": 0.1733, + "step": 73838 + }, + { + "epoch": 2.84, + "learning_rate": 2.6872274942292896e-06, + "loss": 0.2607, + "step": 73839 + }, + { + "epoch": 2.84, + "learning_rate": 2.686586304180559e-06, + "loss": 0.1797, + "step": 73840 + }, + { + "epoch": 2.84, + "learning_rate": 2.6859451141318288e-06, + "loss": 0.3309, + "step": 73841 + }, + { + "epoch": 2.84, + "learning_rate": 2.6853039240830984e-06, + "loss": 0.1047, + "step": 73842 + }, + { + "epoch": 2.84, + "learning_rate": 2.684662734034368e-06, + "loss": 0.2311, + "step": 73843 + }, + { + "epoch": 2.84, + "learning_rate": 2.6840215439856376e-06, + "loss": 0.1107, + "step": 73844 + }, + { + "epoch": 2.84, + "learning_rate": 2.683380353936907e-06, + "loss": 0.2176, + "step": 73845 + }, + { + "epoch": 2.84, + "learning_rate": 2.6827391638881768e-06, + "loss": 0.1977, + "step": 73846 + }, + { + "epoch": 2.84, + "learning_rate": 2.682097973839446e-06, + "loss": 0.326, + "step": 73847 + }, + { + "epoch": 2.84, + "learning_rate": 2.6814567837907155e-06, + "loss": 0.3429, + "step": 73848 + }, + { + "epoch": 2.84, + "learning_rate": 2.680815593741985e-06, + "loss": 0.2933, + "step": 73849 + }, + { + "epoch": 2.84, + "learning_rate": 2.6801744036932547e-06, + "loss": 0.2061, + "step": 73850 + }, + { + "epoch": 2.84, + "learning_rate": 2.6795332136445243e-06, + "loss": 0.2247, + "step": 73851 + }, + { + "epoch": 2.84, + "learning_rate": 2.678892023595794e-06, + "loss": 0.2211, + "step": 73852 + }, + { + "epoch": 2.84, + "learning_rate": 2.6782508335470635e-06, + "loss": 0.2402, + "step": 73853 + }, + { + "epoch": 2.84, + "learning_rate": 2.677609643498333e-06, + "loss": 0.1553, + "step": 73854 + }, + { + "epoch": 2.84, + "learning_rate": 2.6769684534496027e-06, + "loss": 0.2464, + "step": 73855 + }, + { + "epoch": 2.84, + "learning_rate": 2.6763272634008723e-06, + "loss": 0.1407, + "step": 73856 + }, + { + "epoch": 2.84, + "learning_rate": 2.6756860733521415e-06, + "loss": 0.2758, + "step": 73857 + }, + { + "epoch": 2.84, + "learning_rate": 2.675044883303411e-06, + "loss": 0.271, + "step": 73858 + }, + { + "epoch": 2.84, + "learning_rate": 2.6744036932546807e-06, + "loss": 0.269, + "step": 73859 + }, + { + "epoch": 2.84, + "learning_rate": 2.6737625032059503e-06, + "loss": 0.2732, + "step": 73860 + }, + { + "epoch": 2.84, + "learning_rate": 2.67312131315722e-06, + "loss": 0.2238, + "step": 73861 + }, + { + "epoch": 2.84, + "learning_rate": 2.6724801231084895e-06, + "loss": 0.1951, + "step": 73862 + }, + { + "epoch": 2.84, + "learning_rate": 2.671838933059759e-06, + "loss": 0.3141, + "step": 73863 + }, + { + "epoch": 2.84, + "learning_rate": 2.6711977430110287e-06, + "loss": 0.2895, + "step": 73864 + }, + { + "epoch": 2.84, + "learning_rate": 2.6705565529622983e-06, + "loss": 0.1198, + "step": 73865 + }, + { + "epoch": 2.84, + "learning_rate": 2.669915362913568e-06, + "loss": 0.1432, + "step": 73866 + }, + { + "epoch": 2.84, + "learning_rate": 2.669274172864837e-06, + "loss": 0.1297, + "step": 73867 + }, + { + "epoch": 2.84, + "learning_rate": 2.6686329828161067e-06, + "loss": 0.167, + "step": 73868 + }, + { + "epoch": 2.84, + "learning_rate": 2.6679917927673763e-06, + "loss": 0.1327, + "step": 73869 + }, + { + "epoch": 2.84, + "learning_rate": 2.667350602718646e-06, + "loss": 0.1968, + "step": 73870 + }, + { + "epoch": 2.84, + "learning_rate": 2.6667094126699155e-06, + "loss": 0.3087, + "step": 73871 + }, + { + "epoch": 2.84, + "learning_rate": 2.666068222621185e-06, + "loss": 0.1585, + "step": 73872 + }, + { + "epoch": 2.84, + "learning_rate": 2.6654270325724547e-06, + "loss": 0.1635, + "step": 73873 + }, + { + "epoch": 2.84, + "learning_rate": 2.6647858425237243e-06, + "loss": 0.1259, + "step": 73874 + }, + { + "epoch": 2.84, + "learning_rate": 2.664144652474994e-06, + "loss": 0.2264, + "step": 73875 + }, + { + "epoch": 2.84, + "learning_rate": 2.6635034624262635e-06, + "loss": 0.2854, + "step": 73876 + }, + { + "epoch": 2.84, + "learning_rate": 2.6628622723775326e-06, + "loss": 0.1772, + "step": 73877 + }, + { + "epoch": 2.84, + "learning_rate": 2.6622210823288022e-06, + "loss": 0.1676, + "step": 73878 + }, + { + "epoch": 2.84, + "learning_rate": 2.661579892280072e-06, + "loss": 0.3553, + "step": 73879 + }, + { + "epoch": 2.84, + "learning_rate": 2.6609387022313414e-06, + "loss": 0.3148, + "step": 73880 + }, + { + "epoch": 2.84, + "learning_rate": 2.660297512182611e-06, + "loss": 0.3096, + "step": 73881 + }, + { + "epoch": 2.84, + "learning_rate": 2.6596563221338806e-06, + "loss": 0.1069, + "step": 73882 + }, + { + "epoch": 2.84, + "learning_rate": 2.6590151320851502e-06, + "loss": 0.2474, + "step": 73883 + }, + { + "epoch": 2.84, + "learning_rate": 2.65837394203642e-06, + "loss": 0.1517, + "step": 73884 + }, + { + "epoch": 2.84, + "learning_rate": 2.6577327519876894e-06, + "loss": 0.229, + "step": 73885 + }, + { + "epoch": 2.84, + "learning_rate": 2.657091561938959e-06, + "loss": 0.1532, + "step": 73886 + }, + { + "epoch": 2.84, + "learning_rate": 2.656450371890228e-06, + "loss": 0.3151, + "step": 73887 + }, + { + "epoch": 2.84, + "learning_rate": 2.655809181841498e-06, + "loss": 0.1173, + "step": 73888 + }, + { + "epoch": 2.84, + "learning_rate": 2.6551679917927674e-06, + "loss": 0.1509, + "step": 73889 + }, + { + "epoch": 2.84, + "learning_rate": 2.654526801744037e-06, + "loss": 0.2774, + "step": 73890 + }, + { + "epoch": 2.84, + "learning_rate": 2.6538856116953066e-06, + "loss": 0.1388, + "step": 73891 + }, + { + "epoch": 2.84, + "learning_rate": 2.653244421646576e-06, + "loss": 0.1397, + "step": 73892 + }, + { + "epoch": 2.84, + "learning_rate": 2.652603231597846e-06, + "loss": 0.1873, + "step": 73893 + }, + { + "epoch": 2.84, + "learning_rate": 2.6519620415491154e-06, + "loss": 0.1852, + "step": 73894 + }, + { + "epoch": 2.84, + "learning_rate": 2.651320851500385e-06, + "loss": 0.2701, + "step": 73895 + }, + { + "epoch": 2.84, + "learning_rate": 2.6506796614516546e-06, + "loss": 0.2395, + "step": 73896 + }, + { + "epoch": 2.84, + "learning_rate": 2.6500384714029238e-06, + "loss": 0.1446, + "step": 73897 + }, + { + "epoch": 2.84, + "learning_rate": 2.6493972813541934e-06, + "loss": 0.3308, + "step": 73898 + }, + { + "epoch": 2.84, + "learning_rate": 2.648756091305463e-06, + "loss": 0.2108, + "step": 73899 + }, + { + "epoch": 2.84, + "learning_rate": 2.6481149012567326e-06, + "loss": 0.1898, + "step": 73900 + }, + { + "epoch": 2.84, + "learning_rate": 2.647473711208002e-06, + "loss": 0.2861, + "step": 73901 + }, + { + "epoch": 2.84, + "learning_rate": 2.6468325211592718e-06, + "loss": 0.273, + "step": 73902 + }, + { + "epoch": 2.84, + "learning_rate": 2.6461913311105414e-06, + "loss": 0.1664, + "step": 73903 + }, + { + "epoch": 2.84, + "learning_rate": 2.645550141061811e-06, + "loss": 0.2345, + "step": 73904 + }, + { + "epoch": 2.84, + "learning_rate": 2.6449089510130806e-06, + "loss": 0.2442, + "step": 73905 + }, + { + "epoch": 2.84, + "learning_rate": 2.6442677609643497e-06, + "loss": 0.1537, + "step": 73906 + }, + { + "epoch": 2.84, + "learning_rate": 2.6436265709156193e-06, + "loss": 0.1904, + "step": 73907 + }, + { + "epoch": 2.84, + "learning_rate": 2.642985380866889e-06, + "loss": 0.1788, + "step": 73908 + }, + { + "epoch": 2.84, + "learning_rate": 2.6423441908181585e-06, + "loss": 0.188, + "step": 73909 + }, + { + "epoch": 2.84, + "learning_rate": 2.641703000769428e-06, + "loss": 0.1536, + "step": 73910 + }, + { + "epoch": 2.84, + "learning_rate": 2.6410618107206977e-06, + "loss": 0.0968, + "step": 73911 + }, + { + "epoch": 2.84, + "learning_rate": 2.6404206206719673e-06, + "loss": 0.2406, + "step": 73912 + }, + { + "epoch": 2.84, + "learning_rate": 2.639779430623237e-06, + "loss": 0.3299, + "step": 73913 + }, + { + "epoch": 2.84, + "learning_rate": 2.6391382405745065e-06, + "loss": 0.1769, + "step": 73914 + }, + { + "epoch": 2.84, + "learning_rate": 2.638497050525776e-06, + "loss": 0.2101, + "step": 73915 + }, + { + "epoch": 2.84, + "learning_rate": 2.6378558604770453e-06, + "loss": 0.2385, + "step": 73916 + }, + { + "epoch": 2.84, + "learning_rate": 2.637214670428315e-06, + "loss": 0.2364, + "step": 73917 + }, + { + "epoch": 2.84, + "learning_rate": 2.6365734803795845e-06, + "loss": 0.3193, + "step": 73918 + }, + { + "epoch": 2.84, + "learning_rate": 2.635932290330854e-06, + "loss": 0.1832, + "step": 73919 + }, + { + "epoch": 2.84, + "learning_rate": 2.6352911002821237e-06, + "loss": 0.1331, + "step": 73920 + }, + { + "epoch": 2.84, + "learning_rate": 2.6346499102333933e-06, + "loss": 0.2631, + "step": 73921 + }, + { + "epoch": 2.84, + "learning_rate": 2.634008720184663e-06, + "loss": 0.3087, + "step": 73922 + }, + { + "epoch": 2.84, + "learning_rate": 2.6333675301359325e-06, + "loss": 0.2152, + "step": 73923 + }, + { + "epoch": 2.84, + "learning_rate": 2.632726340087202e-06, + "loss": 0.173, + "step": 73924 + }, + { + "epoch": 2.84, + "learning_rate": 2.6320851500384717e-06, + "loss": 0.1481, + "step": 73925 + }, + { + "epoch": 2.84, + "learning_rate": 2.631443959989741e-06, + "loss": 0.1738, + "step": 73926 + }, + { + "epoch": 2.84, + "learning_rate": 2.6308027699410105e-06, + "loss": 0.1506, + "step": 73927 + }, + { + "epoch": 2.84, + "learning_rate": 2.63016157989228e-06, + "loss": 0.1821, + "step": 73928 + }, + { + "epoch": 2.84, + "learning_rate": 2.6295203898435497e-06, + "loss": 0.127, + "step": 73929 + }, + { + "epoch": 2.84, + "learning_rate": 2.6288791997948193e-06, + "loss": 0.2121, + "step": 73930 + }, + { + "epoch": 2.84, + "learning_rate": 2.628238009746089e-06, + "loss": 0.2052, + "step": 73931 + }, + { + "epoch": 2.84, + "learning_rate": 2.6275968196973585e-06, + "loss": 0.2536, + "step": 73932 + }, + { + "epoch": 2.84, + "learning_rate": 2.626955629648628e-06, + "loss": 0.2251, + "step": 73933 + }, + { + "epoch": 2.84, + "learning_rate": 2.6263144395998977e-06, + "loss": 0.1955, + "step": 73934 + }, + { + "epoch": 2.84, + "learning_rate": 2.6256732495511673e-06, + "loss": 0.1155, + "step": 73935 + }, + { + "epoch": 2.84, + "learning_rate": 2.6250320595024364e-06, + "loss": 0.1165, + "step": 73936 + }, + { + "epoch": 2.84, + "learning_rate": 2.624390869453706e-06, + "loss": 0.1597, + "step": 73937 + }, + { + "epoch": 2.84, + "learning_rate": 2.6237496794049756e-06, + "loss": 0.2252, + "step": 73938 + }, + { + "epoch": 2.84, + "learning_rate": 2.6231084893562452e-06, + "loss": 0.1573, + "step": 73939 + }, + { + "epoch": 2.84, + "learning_rate": 2.622467299307515e-06, + "loss": 0.2613, + "step": 73940 + }, + { + "epoch": 2.84, + "learning_rate": 2.6218261092587844e-06, + "loss": 0.1413, + "step": 73941 + }, + { + "epoch": 2.84, + "learning_rate": 2.621184919210054e-06, + "loss": 0.2703, + "step": 73942 + }, + { + "epoch": 2.84, + "learning_rate": 2.6205437291613236e-06, + "loss": 0.2206, + "step": 73943 + }, + { + "epoch": 2.84, + "learning_rate": 2.6199025391125932e-06, + "loss": 0.3499, + "step": 73944 + }, + { + "epoch": 2.84, + "learning_rate": 2.619261349063863e-06, + "loss": 0.2452, + "step": 73945 + }, + { + "epoch": 2.84, + "learning_rate": 2.618620159015132e-06, + "loss": 0.2077, + "step": 73946 + }, + { + "epoch": 2.84, + "learning_rate": 2.6179789689664016e-06, + "loss": 0.3488, + "step": 73947 + }, + { + "epoch": 2.84, + "learning_rate": 2.617337778917671e-06, + "loss": 0.2643, + "step": 73948 + }, + { + "epoch": 2.84, + "learning_rate": 2.616696588868941e-06, + "loss": 0.3471, + "step": 73949 + }, + { + "epoch": 2.84, + "learning_rate": 2.6160553988202104e-06, + "loss": 0.3607, + "step": 73950 + }, + { + "epoch": 2.84, + "learning_rate": 2.61541420877148e-06, + "loss": 0.2311, + "step": 73951 + }, + { + "epoch": 2.84, + "learning_rate": 2.6147730187227496e-06, + "loss": 0.1839, + "step": 73952 + }, + { + "epoch": 2.84, + "learning_rate": 2.614131828674019e-06, + "loss": 0.2411, + "step": 73953 + }, + { + "epoch": 2.84, + "learning_rate": 2.613490638625289e-06, + "loss": 0.1119, + "step": 73954 + }, + { + "epoch": 2.84, + "learning_rate": 2.6128494485765584e-06, + "loss": 0.1949, + "step": 73955 + }, + { + "epoch": 2.84, + "learning_rate": 2.6122082585278276e-06, + "loss": 0.1889, + "step": 73956 + }, + { + "epoch": 2.84, + "learning_rate": 2.611567068479097e-06, + "loss": 0.186, + "step": 73957 + }, + { + "epoch": 2.84, + "learning_rate": 2.6109258784303668e-06, + "loss": 0.2394, + "step": 73958 + }, + { + "epoch": 2.84, + "learning_rate": 2.6102846883816364e-06, + "loss": 0.3983, + "step": 73959 + }, + { + "epoch": 2.84, + "learning_rate": 2.609643498332906e-06, + "loss": 0.09, + "step": 73960 + }, + { + "epoch": 2.84, + "learning_rate": 2.6090023082841756e-06, + "loss": 0.1705, + "step": 73961 + }, + { + "epoch": 2.84, + "learning_rate": 2.608361118235445e-06, + "loss": 0.1463, + "step": 73962 + }, + { + "epoch": 2.84, + "learning_rate": 2.6077199281867148e-06, + "loss": 0.1838, + "step": 73963 + }, + { + "epoch": 2.84, + "learning_rate": 2.6070787381379844e-06, + "loss": 0.1302, + "step": 73964 + }, + { + "epoch": 2.84, + "learning_rate": 2.6064375480892535e-06, + "loss": 0.2497, + "step": 73965 + }, + { + "epoch": 2.84, + "learning_rate": 2.605796358040523e-06, + "loss": 0.2386, + "step": 73966 + }, + { + "epoch": 2.84, + "learning_rate": 2.6051551679917927e-06, + "loss": 0.2021, + "step": 73967 + }, + { + "epoch": 2.84, + "learning_rate": 2.6045139779430623e-06, + "loss": 0.1566, + "step": 73968 + }, + { + "epoch": 2.84, + "learning_rate": 2.603872787894332e-06, + "loss": 0.2695, + "step": 73969 + }, + { + "epoch": 2.84, + "learning_rate": 2.6032315978456015e-06, + "loss": 0.299, + "step": 73970 + }, + { + "epoch": 2.84, + "learning_rate": 2.602590407796871e-06, + "loss": 0.1226, + "step": 73971 + }, + { + "epoch": 2.84, + "learning_rate": 2.6019492177481407e-06, + "loss": 0.4363, + "step": 73972 + }, + { + "epoch": 2.84, + "learning_rate": 2.6013080276994103e-06, + "loss": 0.1668, + "step": 73973 + }, + { + "epoch": 2.84, + "learning_rate": 2.60066683765068e-06, + "loss": 0.2236, + "step": 73974 + }, + { + "epoch": 2.84, + "learning_rate": 2.600025647601949e-06, + "loss": 0.2251, + "step": 73975 + }, + { + "epoch": 2.84, + "learning_rate": 2.5993844575532187e-06, + "loss": 0.233, + "step": 73976 + }, + { + "epoch": 2.84, + "learning_rate": 2.5987432675044883e-06, + "loss": 0.4467, + "step": 73977 + }, + { + "epoch": 2.84, + "learning_rate": 2.598102077455758e-06, + "loss": 0.1872, + "step": 73978 + }, + { + "epoch": 2.84, + "learning_rate": 2.5974608874070275e-06, + "loss": 0.194, + "step": 73979 + }, + { + "epoch": 2.84, + "learning_rate": 2.596819697358297e-06, + "loss": 0.0966, + "step": 73980 + }, + { + "epoch": 2.84, + "learning_rate": 2.5961785073095667e-06, + "loss": 0.1496, + "step": 73981 + }, + { + "epoch": 2.84, + "learning_rate": 2.5955373172608363e-06, + "loss": 0.2252, + "step": 73982 + }, + { + "epoch": 2.84, + "learning_rate": 2.594896127212106e-06, + "loss": 0.0936, + "step": 73983 + }, + { + "epoch": 2.84, + "learning_rate": 2.5942549371633755e-06, + "loss": 0.1898, + "step": 73984 + }, + { + "epoch": 2.84, + "learning_rate": 2.5936137471146447e-06, + "loss": 0.1725, + "step": 73985 + }, + { + "epoch": 2.84, + "learning_rate": 2.5929725570659143e-06, + "loss": 0.1755, + "step": 73986 + }, + { + "epoch": 2.84, + "learning_rate": 2.592331367017184e-06, + "loss": 0.2802, + "step": 73987 + }, + { + "epoch": 2.84, + "learning_rate": 2.5916901769684535e-06, + "loss": 0.2761, + "step": 73988 + }, + { + "epoch": 2.84, + "learning_rate": 2.591048986919723e-06, + "loss": 0.2729, + "step": 73989 + }, + { + "epoch": 2.84, + "learning_rate": 2.5904077968709927e-06, + "loss": 0.2699, + "step": 73990 + }, + { + "epoch": 2.84, + "learning_rate": 2.5897666068222623e-06, + "loss": 0.2468, + "step": 73991 + }, + { + "epoch": 2.84, + "learning_rate": 2.589125416773532e-06, + "loss": 0.1008, + "step": 73992 + }, + { + "epoch": 2.84, + "learning_rate": 2.5884842267248015e-06, + "loss": 0.2706, + "step": 73993 + }, + { + "epoch": 2.84, + "learning_rate": 2.587843036676071e-06, + "loss": 0.2233, + "step": 73994 + }, + { + "epoch": 2.84, + "learning_rate": 2.5872018466273402e-06, + "loss": 0.1065, + "step": 73995 + }, + { + "epoch": 2.84, + "learning_rate": 2.58656065657861e-06, + "loss": 0.2045, + "step": 73996 + }, + { + "epoch": 2.84, + "learning_rate": 2.5859194665298794e-06, + "loss": 0.1424, + "step": 73997 + }, + { + "epoch": 2.84, + "learning_rate": 2.585278276481149e-06, + "loss": 0.1977, + "step": 73998 + }, + { + "epoch": 2.85, + "learning_rate": 2.5846370864324186e-06, + "loss": 0.2267, + "step": 73999 + }, + { + "epoch": 2.85, + "learning_rate": 2.5839958963836882e-06, + "loss": 0.1401, + "step": 74000 + }, + { + "epoch": 2.85, + "learning_rate": 2.583354706334958e-06, + "loss": 0.207, + "step": 74001 + }, + { + "epoch": 2.85, + "learning_rate": 2.5827135162862274e-06, + "loss": 0.1584, + "step": 74002 + }, + { + "epoch": 2.85, + "learning_rate": 2.582072326237497e-06, + "loss": 0.2797, + "step": 74003 + }, + { + "epoch": 2.85, + "learning_rate": 2.5814311361887666e-06, + "loss": 0.1501, + "step": 74004 + }, + { + "epoch": 2.85, + "learning_rate": 2.580789946140036e-06, + "loss": 0.127, + "step": 74005 + }, + { + "epoch": 2.85, + "learning_rate": 2.5801487560913054e-06, + "loss": 0.3156, + "step": 74006 + }, + { + "epoch": 2.85, + "learning_rate": 2.579507566042575e-06, + "loss": 0.223, + "step": 74007 + }, + { + "epoch": 2.85, + "learning_rate": 2.5788663759938446e-06, + "loss": 0.1063, + "step": 74008 + }, + { + "epoch": 2.85, + "learning_rate": 2.578225185945114e-06, + "loss": 0.2552, + "step": 74009 + }, + { + "epoch": 2.85, + "learning_rate": 2.577583995896384e-06, + "loss": 0.1854, + "step": 74010 + }, + { + "epoch": 2.85, + "learning_rate": 2.5769428058476534e-06, + "loss": 0.2202, + "step": 74011 + }, + { + "epoch": 2.85, + "learning_rate": 2.576301615798923e-06, + "loss": 0.2751, + "step": 74012 + }, + { + "epoch": 2.85, + "learning_rate": 2.5756604257501926e-06, + "loss": 0.1833, + "step": 74013 + }, + { + "epoch": 2.85, + "learning_rate": 2.575019235701462e-06, + "loss": 0.167, + "step": 74014 + }, + { + "epoch": 2.85, + "learning_rate": 2.5743780456527314e-06, + "loss": 0.1719, + "step": 74015 + }, + { + "epoch": 2.85, + "learning_rate": 2.573736855604001e-06, + "loss": 0.2459, + "step": 74016 + }, + { + "epoch": 2.85, + "learning_rate": 2.5730956655552706e-06, + "loss": 0.2193, + "step": 74017 + }, + { + "epoch": 2.85, + "learning_rate": 2.57245447550654e-06, + "loss": 0.1113, + "step": 74018 + }, + { + "epoch": 2.85, + "learning_rate": 2.5718132854578098e-06, + "loss": 0.2287, + "step": 74019 + }, + { + "epoch": 2.85, + "learning_rate": 2.5711720954090794e-06, + "loss": 0.2075, + "step": 74020 + }, + { + "epoch": 2.85, + "learning_rate": 2.570530905360349e-06, + "loss": 0.1533, + "step": 74021 + }, + { + "epoch": 2.85, + "learning_rate": 2.5698897153116186e-06, + "loss": 0.2463, + "step": 74022 + }, + { + "epoch": 2.85, + "learning_rate": 2.569248525262888e-06, + "loss": 0.1667, + "step": 74023 + }, + { + "epoch": 2.85, + "learning_rate": 2.5686073352141573e-06, + "loss": 0.2892, + "step": 74024 + }, + { + "epoch": 2.85, + "learning_rate": 2.567966145165427e-06, + "loss": 0.3083, + "step": 74025 + }, + { + "epoch": 2.85, + "learning_rate": 2.5673249551166965e-06, + "loss": 0.2761, + "step": 74026 + }, + { + "epoch": 2.85, + "learning_rate": 2.566683765067966e-06, + "loss": 0.1088, + "step": 74027 + }, + { + "epoch": 2.85, + "learning_rate": 2.5660425750192357e-06, + "loss": 0.2132, + "step": 74028 + }, + { + "epoch": 2.85, + "learning_rate": 2.5654013849705053e-06, + "loss": 0.2714, + "step": 74029 + }, + { + "epoch": 2.85, + "learning_rate": 2.564760194921775e-06, + "loss": 0.2241, + "step": 74030 + }, + { + "epoch": 2.85, + "learning_rate": 2.5641190048730445e-06, + "loss": 0.1692, + "step": 74031 + }, + { + "epoch": 2.85, + "learning_rate": 2.563477814824314e-06, + "loss": 0.2903, + "step": 74032 + }, + { + "epoch": 2.85, + "learning_rate": 2.5628366247755837e-06, + "loss": 0.2617, + "step": 74033 + }, + { + "epoch": 2.85, + "learning_rate": 2.562195434726853e-06, + "loss": 0.276, + "step": 74034 + }, + { + "epoch": 2.85, + "learning_rate": 2.5615542446781225e-06, + "loss": 0.2714, + "step": 74035 + }, + { + "epoch": 2.85, + "learning_rate": 2.560913054629392e-06, + "loss": 0.147, + "step": 74036 + }, + { + "epoch": 2.85, + "learning_rate": 2.5602718645806617e-06, + "loss": 0.1725, + "step": 74037 + }, + { + "epoch": 2.85, + "learning_rate": 2.5596306745319313e-06, + "loss": 0.1745, + "step": 74038 + }, + { + "epoch": 2.85, + "learning_rate": 2.558989484483201e-06, + "loss": 0.1845, + "step": 74039 + }, + { + "epoch": 2.85, + "learning_rate": 2.5583482944344705e-06, + "loss": 0.2477, + "step": 74040 + }, + { + "epoch": 2.85, + "learning_rate": 2.55770710438574e-06, + "loss": 0.1635, + "step": 74041 + }, + { + "epoch": 2.85, + "learning_rate": 2.5570659143370097e-06, + "loss": 0.17, + "step": 74042 + }, + { + "epoch": 2.85, + "learning_rate": 2.5564247242882793e-06, + "loss": 0.2234, + "step": 74043 + }, + { + "epoch": 2.85, + "learning_rate": 2.5557835342395485e-06, + "loss": 0.1752, + "step": 74044 + }, + { + "epoch": 2.85, + "learning_rate": 2.555142344190818e-06, + "loss": 0.2031, + "step": 74045 + }, + { + "epoch": 2.85, + "learning_rate": 2.5545011541420877e-06, + "loss": 0.1957, + "step": 74046 + }, + { + "epoch": 2.85, + "learning_rate": 2.5538599640933573e-06, + "loss": 0.2503, + "step": 74047 + }, + { + "epoch": 2.85, + "learning_rate": 2.553218774044627e-06, + "loss": 0.1483, + "step": 74048 + }, + { + "epoch": 2.85, + "learning_rate": 2.5525775839958965e-06, + "loss": 0.2079, + "step": 74049 + }, + { + "epoch": 2.85, + "learning_rate": 2.551936393947166e-06, + "loss": 0.2664, + "step": 74050 + }, + { + "epoch": 2.85, + "learning_rate": 2.5512952038984357e-06, + "loss": 0.1549, + "step": 74051 + }, + { + "epoch": 2.85, + "learning_rate": 2.5506540138497053e-06, + "loss": 0.3308, + "step": 74052 + }, + { + "epoch": 2.85, + "learning_rate": 2.550012823800975e-06, + "loss": 0.1532, + "step": 74053 + }, + { + "epoch": 2.85, + "learning_rate": 2.549371633752244e-06, + "loss": 0.3187, + "step": 74054 + }, + { + "epoch": 2.85, + "learning_rate": 2.5487304437035136e-06, + "loss": 0.2421, + "step": 74055 + }, + { + "epoch": 2.85, + "learning_rate": 2.5480892536547832e-06, + "loss": 0.1732, + "step": 74056 + }, + { + "epoch": 2.85, + "learning_rate": 2.547448063606053e-06, + "loss": 0.2956, + "step": 74057 + }, + { + "epoch": 2.85, + "learning_rate": 2.5468068735573224e-06, + "loss": 0.2122, + "step": 74058 + }, + { + "epoch": 2.85, + "learning_rate": 2.546165683508592e-06, + "loss": 0.2494, + "step": 74059 + }, + { + "epoch": 2.85, + "learning_rate": 2.5455244934598616e-06, + "loss": 0.2018, + "step": 74060 + }, + { + "epoch": 2.85, + "learning_rate": 2.5448833034111312e-06, + "loss": 0.294, + "step": 74061 + }, + { + "epoch": 2.85, + "learning_rate": 2.544242113362401e-06, + "loss": 0.1706, + "step": 74062 + }, + { + "epoch": 2.85, + "learning_rate": 2.5436009233136704e-06, + "loss": 0.1587, + "step": 74063 + }, + { + "epoch": 2.85, + "learning_rate": 2.5429597332649396e-06, + "loss": 0.2541, + "step": 74064 + }, + { + "epoch": 2.85, + "learning_rate": 2.542318543216209e-06, + "loss": 0.1562, + "step": 74065 + }, + { + "epoch": 2.85, + "learning_rate": 2.541677353167479e-06, + "loss": 0.2337, + "step": 74066 + }, + { + "epoch": 2.85, + "learning_rate": 2.5410361631187484e-06, + "loss": 0.2014, + "step": 74067 + }, + { + "epoch": 2.85, + "learning_rate": 2.540394973070018e-06, + "loss": 0.2408, + "step": 74068 + }, + { + "epoch": 2.85, + "learning_rate": 2.5397537830212876e-06, + "loss": 0.2038, + "step": 74069 + }, + { + "epoch": 2.85, + "learning_rate": 2.539112592972557e-06, + "loss": 0.1901, + "step": 74070 + }, + { + "epoch": 2.85, + "learning_rate": 2.538471402923827e-06, + "loss": 0.1805, + "step": 74071 + }, + { + "epoch": 2.85, + "learning_rate": 2.5378302128750964e-06, + "loss": 0.1543, + "step": 74072 + }, + { + "epoch": 2.85, + "learning_rate": 2.5371890228263656e-06, + "loss": 0.3301, + "step": 74073 + }, + { + "epoch": 2.85, + "learning_rate": 2.536547832777635e-06, + "loss": 0.2362, + "step": 74074 + }, + { + "epoch": 2.85, + "learning_rate": 2.5359066427289048e-06, + "loss": 0.2829, + "step": 74075 + }, + { + "epoch": 2.85, + "learning_rate": 2.5352654526801744e-06, + "loss": 0.2857, + "step": 74076 + }, + { + "epoch": 2.85, + "learning_rate": 2.534624262631444e-06, + "loss": 0.2218, + "step": 74077 + }, + { + "epoch": 2.85, + "learning_rate": 2.5339830725827136e-06, + "loss": 0.282, + "step": 74078 + }, + { + "epoch": 2.85, + "learning_rate": 2.533341882533983e-06, + "loss": 0.1912, + "step": 74079 + }, + { + "epoch": 2.85, + "learning_rate": 2.5327006924852528e-06, + "loss": 0.1555, + "step": 74080 + }, + { + "epoch": 2.85, + "learning_rate": 2.5320595024365224e-06, + "loss": 0.2303, + "step": 74081 + }, + { + "epoch": 2.85, + "learning_rate": 2.531418312387792e-06, + "loss": 0.1792, + "step": 74082 + }, + { + "epoch": 2.85, + "learning_rate": 2.530777122339061e-06, + "loss": 0.2156, + "step": 74083 + }, + { + "epoch": 2.85, + "learning_rate": 2.5301359322903307e-06, + "loss": 0.2632, + "step": 74084 + }, + { + "epoch": 2.85, + "learning_rate": 2.5294947422416003e-06, + "loss": 0.2338, + "step": 74085 + }, + { + "epoch": 2.85, + "learning_rate": 2.52885355219287e-06, + "loss": 0.1629, + "step": 74086 + }, + { + "epoch": 2.85, + "learning_rate": 2.5282123621441395e-06, + "loss": 0.1743, + "step": 74087 + }, + { + "epoch": 2.85, + "learning_rate": 2.527571172095409e-06, + "loss": 0.2175, + "step": 74088 + }, + { + "epoch": 2.85, + "learning_rate": 2.5269299820466787e-06, + "loss": 0.2263, + "step": 74089 + }, + { + "epoch": 2.85, + "learning_rate": 2.5262887919979483e-06, + "loss": 0.1279, + "step": 74090 + }, + { + "epoch": 2.85, + "learning_rate": 2.525647601949218e-06, + "loss": 0.2096, + "step": 74091 + }, + { + "epoch": 2.85, + "learning_rate": 2.5250064119004875e-06, + "loss": 0.2045, + "step": 74092 + }, + { + "epoch": 2.85, + "learning_rate": 2.5243652218517567e-06, + "loss": 0.1057, + "step": 74093 + }, + { + "epoch": 2.85, + "learning_rate": 2.5237240318030263e-06, + "loss": 0.4782, + "step": 74094 + }, + { + "epoch": 2.85, + "learning_rate": 2.523082841754296e-06, + "loss": 0.1615, + "step": 74095 + }, + { + "epoch": 2.85, + "learning_rate": 2.5224416517055655e-06, + "loss": 0.1427, + "step": 74096 + }, + { + "epoch": 2.85, + "learning_rate": 2.521800461656835e-06, + "loss": 0.3497, + "step": 74097 + }, + { + "epoch": 2.85, + "learning_rate": 2.5211592716081047e-06, + "loss": 0.1594, + "step": 74098 + }, + { + "epoch": 2.85, + "learning_rate": 2.5205180815593743e-06, + "loss": 0.1513, + "step": 74099 + }, + { + "epoch": 2.85, + "learning_rate": 2.519876891510644e-06, + "loss": 0.3638, + "step": 74100 + }, + { + "epoch": 2.85, + "learning_rate": 2.5192357014619135e-06, + "loss": 0.1996, + "step": 74101 + }, + { + "epoch": 2.85, + "learning_rate": 2.518594511413183e-06, + "loss": 0.2244, + "step": 74102 + }, + { + "epoch": 2.85, + "learning_rate": 2.5179533213644523e-06, + "loss": 0.1629, + "step": 74103 + }, + { + "epoch": 2.85, + "learning_rate": 2.517312131315722e-06, + "loss": 0.2154, + "step": 74104 + }, + { + "epoch": 2.85, + "learning_rate": 2.5166709412669915e-06, + "loss": 0.1967, + "step": 74105 + }, + { + "epoch": 2.85, + "learning_rate": 2.516029751218261e-06, + "loss": 0.2166, + "step": 74106 + }, + { + "epoch": 2.85, + "learning_rate": 2.5153885611695307e-06, + "loss": 0.2187, + "step": 74107 + }, + { + "epoch": 2.85, + "learning_rate": 2.5147473711208003e-06, + "loss": 0.2406, + "step": 74108 + }, + { + "epoch": 2.85, + "learning_rate": 2.51410618107207e-06, + "loss": 0.1693, + "step": 74109 + }, + { + "epoch": 2.85, + "learning_rate": 2.5134649910233395e-06, + "loss": 0.1816, + "step": 74110 + }, + { + "epoch": 2.85, + "learning_rate": 2.512823800974609e-06, + "loss": 0.1365, + "step": 74111 + }, + { + "epoch": 2.85, + "learning_rate": 2.5121826109258787e-06, + "loss": 0.1705, + "step": 74112 + }, + { + "epoch": 2.85, + "learning_rate": 2.511541420877148e-06, + "loss": 0.1753, + "step": 74113 + }, + { + "epoch": 2.85, + "learning_rate": 2.5109002308284174e-06, + "loss": 0.1211, + "step": 74114 + }, + { + "epoch": 2.85, + "learning_rate": 2.510259040779687e-06, + "loss": 0.2121, + "step": 74115 + }, + { + "epoch": 2.85, + "learning_rate": 2.5096178507309566e-06, + "loss": 0.1463, + "step": 74116 + }, + { + "epoch": 2.85, + "learning_rate": 2.5089766606822262e-06, + "loss": 0.1982, + "step": 74117 + }, + { + "epoch": 2.85, + "learning_rate": 2.508335470633496e-06, + "loss": 0.1746, + "step": 74118 + }, + { + "epoch": 2.85, + "learning_rate": 2.5076942805847654e-06, + "loss": 0.2475, + "step": 74119 + }, + { + "epoch": 2.85, + "learning_rate": 2.507053090536035e-06, + "loss": 0.2773, + "step": 74120 + }, + { + "epoch": 2.85, + "learning_rate": 2.5064119004873046e-06, + "loss": 0.2205, + "step": 74121 + }, + { + "epoch": 2.85, + "learning_rate": 2.5057707104385742e-06, + "loss": 0.1847, + "step": 74122 + }, + { + "epoch": 2.85, + "learning_rate": 2.5051295203898434e-06, + "loss": 0.1541, + "step": 74123 + }, + { + "epoch": 2.85, + "learning_rate": 2.504488330341113e-06, + "loss": 0.2164, + "step": 74124 + }, + { + "epoch": 2.85, + "learning_rate": 2.5038471402923826e-06, + "loss": 0.1982, + "step": 74125 + }, + { + "epoch": 2.85, + "learning_rate": 2.503205950243652e-06, + "loss": 0.2424, + "step": 74126 + }, + { + "epoch": 2.85, + "learning_rate": 2.502564760194922e-06, + "loss": 0.2821, + "step": 74127 + }, + { + "epoch": 2.85, + "learning_rate": 2.5019235701461914e-06, + "loss": 0.124, + "step": 74128 + }, + { + "epoch": 2.85, + "learning_rate": 2.501282380097461e-06, + "loss": 0.2012, + "step": 74129 + }, + { + "epoch": 2.85, + "learning_rate": 2.5006411900487306e-06, + "loss": 0.075, + "step": 74130 + }, + { + "epoch": 2.85, + "learning_rate": 2.5e-06, + "loss": 0.2378, + "step": 74131 + }, + { + "epoch": 2.85, + "learning_rate": 2.4993588099512694e-06, + "loss": 0.1217, + "step": 74132 + }, + { + "epoch": 2.85, + "learning_rate": 2.498717619902539e-06, + "loss": 0.2695, + "step": 74133 + }, + { + "epoch": 2.85, + "learning_rate": 2.4980764298538086e-06, + "loss": 0.2521, + "step": 74134 + }, + { + "epoch": 2.85, + "learning_rate": 2.497435239805078e-06, + "loss": 0.2118, + "step": 74135 + }, + { + "epoch": 2.85, + "learning_rate": 2.4967940497563478e-06, + "loss": 0.218, + "step": 74136 + }, + { + "epoch": 2.85, + "learning_rate": 2.4961528597076174e-06, + "loss": 0.1601, + "step": 74137 + }, + { + "epoch": 2.85, + "learning_rate": 2.495511669658887e-06, + "loss": 0.1861, + "step": 74138 + }, + { + "epoch": 2.85, + "learning_rate": 2.4948704796101566e-06, + "loss": 0.3039, + "step": 74139 + }, + { + "epoch": 2.85, + "learning_rate": 2.494229289561426e-06, + "loss": 0.2595, + "step": 74140 + }, + { + "epoch": 2.85, + "learning_rate": 2.4935880995126958e-06, + "loss": 0.2953, + "step": 74141 + }, + { + "epoch": 2.85, + "learning_rate": 2.492946909463965e-06, + "loss": 0.1858, + "step": 74142 + }, + { + "epoch": 2.85, + "learning_rate": 2.4923057194152345e-06, + "loss": 0.1416, + "step": 74143 + }, + { + "epoch": 2.85, + "learning_rate": 2.491664529366504e-06, + "loss": 0.1243, + "step": 74144 + }, + { + "epoch": 2.85, + "learning_rate": 2.4910233393177737e-06, + "loss": 0.2649, + "step": 74145 + }, + { + "epoch": 2.85, + "learning_rate": 2.4903821492690433e-06, + "loss": 0.3572, + "step": 74146 + }, + { + "epoch": 2.85, + "learning_rate": 2.489740959220313e-06, + "loss": 0.1949, + "step": 74147 + }, + { + "epoch": 2.85, + "learning_rate": 2.4890997691715825e-06, + "loss": 0.1601, + "step": 74148 + }, + { + "epoch": 2.85, + "learning_rate": 2.488458579122852e-06, + "loss": 0.2107, + "step": 74149 + }, + { + "epoch": 2.85, + "learning_rate": 2.4878173890741217e-06, + "loss": 0.1432, + "step": 74150 + }, + { + "epoch": 2.85, + "learning_rate": 2.4871761990253913e-06, + "loss": 0.3482, + "step": 74151 + }, + { + "epoch": 2.85, + "learning_rate": 2.4865350089766605e-06, + "loss": 0.1608, + "step": 74152 + }, + { + "epoch": 2.85, + "learning_rate": 2.48589381892793e-06, + "loss": 0.2339, + "step": 74153 + }, + { + "epoch": 2.85, + "learning_rate": 2.4852526288791997e-06, + "loss": 0.2392, + "step": 74154 + }, + { + "epoch": 2.85, + "learning_rate": 2.4846114388304693e-06, + "loss": 0.1087, + "step": 74155 + }, + { + "epoch": 2.85, + "learning_rate": 2.483970248781739e-06, + "loss": 0.229, + "step": 74156 + }, + { + "epoch": 2.85, + "learning_rate": 2.4833290587330085e-06, + "loss": 0.1897, + "step": 74157 + }, + { + "epoch": 2.85, + "learning_rate": 2.482687868684278e-06, + "loss": 0.1535, + "step": 74158 + }, + { + "epoch": 2.85, + "learning_rate": 2.4820466786355477e-06, + "loss": 0.2793, + "step": 74159 + }, + { + "epoch": 2.85, + "learning_rate": 2.4814054885868173e-06, + "loss": 0.2347, + "step": 74160 + }, + { + "epoch": 2.85, + "learning_rate": 2.480764298538087e-06, + "loss": 0.2748, + "step": 74161 + }, + { + "epoch": 2.85, + "learning_rate": 2.480123108489356e-06, + "loss": 0.2133, + "step": 74162 + }, + { + "epoch": 2.85, + "learning_rate": 2.4794819184406257e-06, + "loss": 0.2271, + "step": 74163 + }, + { + "epoch": 2.85, + "learning_rate": 2.4788407283918953e-06, + "loss": 0.2164, + "step": 74164 + }, + { + "epoch": 2.85, + "learning_rate": 2.478199538343165e-06, + "loss": 0.262, + "step": 74165 + }, + { + "epoch": 2.85, + "learning_rate": 2.4775583482944345e-06, + "loss": 0.1269, + "step": 74166 + }, + { + "epoch": 2.85, + "learning_rate": 2.476917158245704e-06, + "loss": 0.2168, + "step": 74167 + }, + { + "epoch": 2.85, + "learning_rate": 2.4762759681969737e-06, + "loss": 0.3067, + "step": 74168 + }, + { + "epoch": 2.85, + "learning_rate": 2.4756347781482433e-06, + "loss": 0.2156, + "step": 74169 + }, + { + "epoch": 2.85, + "learning_rate": 2.474993588099513e-06, + "loss": 0.1485, + "step": 74170 + }, + { + "epoch": 2.85, + "learning_rate": 2.4743523980507825e-06, + "loss": 0.3762, + "step": 74171 + }, + { + "epoch": 2.85, + "learning_rate": 2.4737112080020516e-06, + "loss": 0.1009, + "step": 74172 + }, + { + "epoch": 2.85, + "learning_rate": 2.4730700179533212e-06, + "loss": 0.3261, + "step": 74173 + }, + { + "epoch": 2.85, + "learning_rate": 2.472428827904591e-06, + "loss": 0.2861, + "step": 74174 + }, + { + "epoch": 2.85, + "learning_rate": 2.4717876378558604e-06, + "loss": 0.1621, + "step": 74175 + }, + { + "epoch": 2.85, + "learning_rate": 2.47114644780713e-06, + "loss": 0.1991, + "step": 74176 + }, + { + "epoch": 2.85, + "learning_rate": 2.4705052577583996e-06, + "loss": 0.2002, + "step": 74177 + }, + { + "epoch": 2.85, + "learning_rate": 2.4698640677096692e-06, + "loss": 0.2641, + "step": 74178 + }, + { + "epoch": 2.85, + "learning_rate": 2.469222877660939e-06, + "loss": 0.2348, + "step": 74179 + }, + { + "epoch": 2.85, + "learning_rate": 2.4685816876122084e-06, + "loss": 0.1935, + "step": 74180 + }, + { + "epoch": 2.85, + "learning_rate": 2.467940497563478e-06, + "loss": 0.155, + "step": 74181 + }, + { + "epoch": 2.85, + "learning_rate": 2.4672993075147472e-06, + "loss": 0.3327, + "step": 74182 + }, + { + "epoch": 2.85, + "learning_rate": 2.466658117466017e-06, + "loss": 0.1991, + "step": 74183 + }, + { + "epoch": 2.85, + "learning_rate": 2.4660169274172864e-06, + "loss": 0.2435, + "step": 74184 + }, + { + "epoch": 2.85, + "learning_rate": 2.465375737368556e-06, + "loss": 0.2834, + "step": 74185 + }, + { + "epoch": 2.85, + "learning_rate": 2.4647345473198256e-06, + "loss": 0.1958, + "step": 74186 + }, + { + "epoch": 2.85, + "learning_rate": 2.464093357271095e-06, + "loss": 0.1152, + "step": 74187 + }, + { + "epoch": 2.85, + "learning_rate": 2.463452167222365e-06, + "loss": 0.281, + "step": 74188 + }, + { + "epoch": 2.85, + "learning_rate": 2.4628109771736344e-06, + "loss": 0.3343, + "step": 74189 + }, + { + "epoch": 2.85, + "learning_rate": 2.462169787124904e-06, + "loss": 0.1801, + "step": 74190 + }, + { + "epoch": 2.85, + "learning_rate": 2.461528597076173e-06, + "loss": 0.1479, + "step": 74191 + }, + { + "epoch": 2.85, + "learning_rate": 2.4608874070274428e-06, + "loss": 0.2701, + "step": 74192 + }, + { + "epoch": 2.85, + "learning_rate": 2.4602462169787124e-06, + "loss": 0.2111, + "step": 74193 + }, + { + "epoch": 2.85, + "learning_rate": 2.459605026929982e-06, + "loss": 0.2285, + "step": 74194 + }, + { + "epoch": 2.85, + "learning_rate": 2.4589638368812516e-06, + "loss": 0.2591, + "step": 74195 + }, + { + "epoch": 2.85, + "learning_rate": 2.458322646832521e-06, + "loss": 0.2372, + "step": 74196 + }, + { + "epoch": 2.85, + "learning_rate": 2.4576814567837908e-06, + "loss": 0.1782, + "step": 74197 + }, + { + "epoch": 2.85, + "learning_rate": 2.4570402667350604e-06, + "loss": 0.2721, + "step": 74198 + }, + { + "epoch": 2.85, + "learning_rate": 2.45639907668633e-06, + "loss": 0.2327, + "step": 74199 + }, + { + "epoch": 2.85, + "learning_rate": 2.4557578866375996e-06, + "loss": 0.2202, + "step": 74200 + }, + { + "epoch": 2.85, + "learning_rate": 2.4551166965888687e-06, + "loss": 0.171, + "step": 74201 + }, + { + "epoch": 2.85, + "learning_rate": 2.4544755065401383e-06, + "loss": 0.3484, + "step": 74202 + }, + { + "epoch": 2.85, + "learning_rate": 2.453834316491408e-06, + "loss": 0.1835, + "step": 74203 + }, + { + "epoch": 2.85, + "learning_rate": 2.4531931264426775e-06, + "loss": 0.1944, + "step": 74204 + }, + { + "epoch": 2.85, + "learning_rate": 2.452551936393947e-06, + "loss": 0.2222, + "step": 74205 + }, + { + "epoch": 2.85, + "learning_rate": 2.4519107463452167e-06, + "loss": 0.2992, + "step": 74206 + }, + { + "epoch": 2.85, + "learning_rate": 2.4512695562964863e-06, + "loss": 0.1378, + "step": 74207 + }, + { + "epoch": 2.85, + "learning_rate": 2.450628366247756e-06, + "loss": 0.3514, + "step": 74208 + }, + { + "epoch": 2.85, + "learning_rate": 2.4499871761990255e-06, + "loss": 0.2094, + "step": 74209 + }, + { + "epoch": 2.85, + "learning_rate": 2.449345986150295e-06, + "loss": 0.1821, + "step": 74210 + }, + { + "epoch": 2.85, + "learning_rate": 2.4487047961015643e-06, + "loss": 0.1944, + "step": 74211 + }, + { + "epoch": 2.85, + "learning_rate": 2.448063606052834e-06, + "loss": 0.1968, + "step": 74212 + }, + { + "epoch": 2.85, + "learning_rate": 2.4474224160041035e-06, + "loss": 0.3103, + "step": 74213 + }, + { + "epoch": 2.85, + "learning_rate": 2.446781225955373e-06, + "loss": 0.1216, + "step": 74214 + }, + { + "epoch": 2.85, + "learning_rate": 2.4461400359066427e-06, + "loss": 0.1528, + "step": 74215 + }, + { + "epoch": 2.85, + "learning_rate": 2.4454988458579123e-06, + "loss": 0.1956, + "step": 74216 + }, + { + "epoch": 2.85, + "learning_rate": 2.444857655809182e-06, + "loss": 0.186, + "step": 74217 + }, + { + "epoch": 2.85, + "learning_rate": 2.4442164657604515e-06, + "loss": 0.2412, + "step": 74218 + }, + { + "epoch": 2.85, + "learning_rate": 2.443575275711721e-06, + "loss": 0.1692, + "step": 74219 + }, + { + "epoch": 2.85, + "learning_rate": 2.4429340856629907e-06, + "loss": 0.3413, + "step": 74220 + }, + { + "epoch": 2.85, + "learning_rate": 2.44229289561426e-06, + "loss": 0.1592, + "step": 74221 + }, + { + "epoch": 2.85, + "learning_rate": 2.4416517055655295e-06, + "loss": 0.2295, + "step": 74222 + }, + { + "epoch": 2.85, + "learning_rate": 2.441010515516799e-06, + "loss": 0.2494, + "step": 74223 + }, + { + "epoch": 2.85, + "learning_rate": 2.4403693254680687e-06, + "loss": 0.1504, + "step": 74224 + }, + { + "epoch": 2.85, + "learning_rate": 2.4397281354193383e-06, + "loss": 0.1641, + "step": 74225 + }, + { + "epoch": 2.85, + "learning_rate": 2.439086945370608e-06, + "loss": 0.2516, + "step": 74226 + }, + { + "epoch": 2.85, + "learning_rate": 2.4384457553218775e-06, + "loss": 0.1873, + "step": 74227 + }, + { + "epoch": 2.85, + "learning_rate": 2.437804565273147e-06, + "loss": 0.1228, + "step": 74228 + }, + { + "epoch": 2.85, + "learning_rate": 2.4371633752244167e-06, + "loss": 0.2865, + "step": 74229 + }, + { + "epoch": 2.85, + "learning_rate": 2.4365221851756863e-06, + "loss": 0.1691, + "step": 74230 + }, + { + "epoch": 2.85, + "learning_rate": 2.4358809951269554e-06, + "loss": 0.1912, + "step": 74231 + }, + { + "epoch": 2.85, + "learning_rate": 2.435239805078225e-06, + "loss": 0.2319, + "step": 74232 + }, + { + "epoch": 2.85, + "learning_rate": 2.4345986150294946e-06, + "loss": 0.1183, + "step": 74233 + }, + { + "epoch": 2.85, + "learning_rate": 2.4339574249807642e-06, + "loss": 0.3483, + "step": 74234 + }, + { + "epoch": 2.85, + "learning_rate": 2.433316234932034e-06, + "loss": 0.2243, + "step": 74235 + }, + { + "epoch": 2.85, + "learning_rate": 2.4326750448833034e-06, + "loss": 0.2147, + "step": 74236 + }, + { + "epoch": 2.85, + "learning_rate": 2.432033854834573e-06, + "loss": 0.1368, + "step": 74237 + }, + { + "epoch": 2.85, + "learning_rate": 2.4313926647858426e-06, + "loss": 0.2184, + "step": 74238 + }, + { + "epoch": 2.85, + "learning_rate": 2.4307514747371122e-06, + "loss": 0.2579, + "step": 74239 + }, + { + "epoch": 2.85, + "learning_rate": 2.430110284688382e-06, + "loss": 0.0805, + "step": 74240 + }, + { + "epoch": 2.85, + "learning_rate": 2.429469094639651e-06, + "loss": 0.3396, + "step": 74241 + }, + { + "epoch": 2.85, + "learning_rate": 2.4288279045909206e-06, + "loss": 0.252, + "step": 74242 + }, + { + "epoch": 2.85, + "learning_rate": 2.4281867145421902e-06, + "loss": 0.1734, + "step": 74243 + }, + { + "epoch": 2.85, + "learning_rate": 2.42754552449346e-06, + "loss": 0.3035, + "step": 74244 + }, + { + "epoch": 2.85, + "learning_rate": 2.4269043344447294e-06, + "loss": 0.2473, + "step": 74245 + }, + { + "epoch": 2.85, + "learning_rate": 2.426263144395999e-06, + "loss": 0.363, + "step": 74246 + }, + { + "epoch": 2.85, + "learning_rate": 2.4256219543472686e-06, + "loss": 0.2155, + "step": 74247 + }, + { + "epoch": 2.85, + "learning_rate": 2.424980764298538e-06, + "loss": 0.1459, + "step": 74248 + }, + { + "epoch": 2.85, + "learning_rate": 2.424339574249808e-06, + "loss": 0.1853, + "step": 74249 + }, + { + "epoch": 2.85, + "learning_rate": 2.423698384201077e-06, + "loss": 0.2409, + "step": 74250 + }, + { + "epoch": 2.85, + "learning_rate": 2.4230571941523466e-06, + "loss": 0.2875, + "step": 74251 + }, + { + "epoch": 2.85, + "learning_rate": 2.422416004103616e-06, + "loss": 0.2215, + "step": 74252 + }, + { + "epoch": 2.85, + "learning_rate": 2.4217748140548858e-06, + "loss": 0.2507, + "step": 74253 + }, + { + "epoch": 2.85, + "learning_rate": 2.4211336240061554e-06, + "loss": 0.0975, + "step": 74254 + }, + { + "epoch": 2.85, + "learning_rate": 2.420492433957425e-06, + "loss": 0.2237, + "step": 74255 + }, + { + "epoch": 2.85, + "learning_rate": 2.4198512439086946e-06, + "loss": 0.1846, + "step": 74256 + }, + { + "epoch": 2.85, + "learning_rate": 2.419210053859964e-06, + "loss": 0.1541, + "step": 74257 + }, + { + "epoch": 2.85, + "learning_rate": 2.4185688638112338e-06, + "loss": 0.2364, + "step": 74258 + }, + { + "epoch": 2.86, + "learning_rate": 2.4179276737625034e-06, + "loss": 0.1732, + "step": 74259 + }, + { + "epoch": 2.86, + "learning_rate": 2.4172864837137726e-06, + "loss": 0.1053, + "step": 74260 + }, + { + "epoch": 2.86, + "learning_rate": 2.416645293665042e-06, + "loss": 0.216, + "step": 74261 + }, + { + "epoch": 2.86, + "learning_rate": 2.4160041036163117e-06, + "loss": 0.1711, + "step": 74262 + }, + { + "epoch": 2.86, + "learning_rate": 2.4153629135675813e-06, + "loss": 0.2426, + "step": 74263 + }, + { + "epoch": 2.86, + "learning_rate": 2.414721723518851e-06, + "loss": 0.222, + "step": 74264 + }, + { + "epoch": 2.86, + "learning_rate": 2.4140805334701205e-06, + "loss": 0.2301, + "step": 74265 + }, + { + "epoch": 2.86, + "learning_rate": 2.41343934342139e-06, + "loss": 0.1581, + "step": 74266 + }, + { + "epoch": 2.86, + "learning_rate": 2.4127981533726597e-06, + "loss": 0.1994, + "step": 74267 + }, + { + "epoch": 2.86, + "learning_rate": 2.4121569633239293e-06, + "loss": 0.1209, + "step": 74268 + }, + { + "epoch": 2.86, + "learning_rate": 2.411515773275199e-06, + "loss": 0.1268, + "step": 74269 + }, + { + "epoch": 2.86, + "learning_rate": 2.410874583226468e-06, + "loss": 0.3227, + "step": 74270 + }, + { + "epoch": 2.86, + "learning_rate": 2.4102333931777377e-06, + "loss": 0.2081, + "step": 74271 + }, + { + "epoch": 2.86, + "learning_rate": 2.4095922031290073e-06, + "loss": 0.2672, + "step": 74272 + }, + { + "epoch": 2.86, + "learning_rate": 2.408951013080277e-06, + "loss": 0.1915, + "step": 74273 + }, + { + "epoch": 2.86, + "learning_rate": 2.4083098230315465e-06, + "loss": 0.2301, + "step": 74274 + }, + { + "epoch": 2.86, + "learning_rate": 2.407668632982816e-06, + "loss": 0.1414, + "step": 74275 + }, + { + "epoch": 2.86, + "learning_rate": 2.4070274429340857e-06, + "loss": 0.16, + "step": 74276 + }, + { + "epoch": 2.86, + "learning_rate": 2.4063862528853553e-06, + "loss": 0.2031, + "step": 74277 + }, + { + "epoch": 2.86, + "learning_rate": 2.405745062836625e-06, + "loss": 0.2301, + "step": 74278 + }, + { + "epoch": 2.86, + "learning_rate": 2.4051038727878945e-06, + "loss": 0.1288, + "step": 74279 + }, + { + "epoch": 2.86, + "learning_rate": 2.4044626827391637e-06, + "loss": 0.1422, + "step": 74280 + }, + { + "epoch": 2.86, + "learning_rate": 2.4038214926904333e-06, + "loss": 0.2831, + "step": 74281 + }, + { + "epoch": 2.86, + "learning_rate": 2.403180302641703e-06, + "loss": 0.235, + "step": 74282 + }, + { + "epoch": 2.86, + "learning_rate": 2.4025391125929725e-06, + "loss": 0.1928, + "step": 74283 + }, + { + "epoch": 2.86, + "learning_rate": 2.401897922544242e-06, + "loss": 0.2432, + "step": 74284 + }, + { + "epoch": 2.86, + "learning_rate": 2.4012567324955117e-06, + "loss": 0.1991, + "step": 74285 + }, + { + "epoch": 2.86, + "learning_rate": 2.4006155424467813e-06, + "loss": 0.1669, + "step": 74286 + }, + { + "epoch": 2.86, + "learning_rate": 2.399974352398051e-06, + "loss": 0.2268, + "step": 74287 + }, + { + "epoch": 2.86, + "learning_rate": 2.3993331623493205e-06, + "loss": 0.1352, + "step": 74288 + }, + { + "epoch": 2.86, + "learning_rate": 2.39869197230059e-06, + "loss": 0.1488, + "step": 74289 + }, + { + "epoch": 2.86, + "learning_rate": 2.3980507822518593e-06, + "loss": 0.1898, + "step": 74290 + }, + { + "epoch": 2.86, + "learning_rate": 2.397409592203129e-06, + "loss": 0.2522, + "step": 74291 + }, + { + "epoch": 2.86, + "learning_rate": 2.3967684021543985e-06, + "loss": 0.3802, + "step": 74292 + }, + { + "epoch": 2.86, + "learning_rate": 2.396127212105668e-06, + "loss": 0.1983, + "step": 74293 + }, + { + "epoch": 2.86, + "learning_rate": 2.3954860220569376e-06, + "loss": 0.1751, + "step": 74294 + }, + { + "epoch": 2.86, + "learning_rate": 2.3948448320082072e-06, + "loss": 0.2283, + "step": 74295 + }, + { + "epoch": 2.86, + "learning_rate": 2.394203641959477e-06, + "loss": 0.2959, + "step": 74296 + }, + { + "epoch": 2.86, + "learning_rate": 2.3935624519107464e-06, + "loss": 0.236, + "step": 74297 + }, + { + "epoch": 2.86, + "learning_rate": 2.392921261862016e-06, + "loss": 0.1627, + "step": 74298 + }, + { + "epoch": 2.86, + "learning_rate": 2.3922800718132856e-06, + "loss": 0.1351, + "step": 74299 + }, + { + "epoch": 2.86, + "learning_rate": 2.391638881764555e-06, + "loss": 0.1441, + "step": 74300 + }, + { + "epoch": 2.86, + "learning_rate": 2.3909976917158244e-06, + "loss": 0.1506, + "step": 74301 + }, + { + "epoch": 2.86, + "learning_rate": 2.390356501667094e-06, + "loss": 0.0933, + "step": 74302 + }, + { + "epoch": 2.86, + "learning_rate": 2.3897153116183636e-06, + "loss": 0.2787, + "step": 74303 + }, + { + "epoch": 2.86, + "learning_rate": 2.3890741215696332e-06, + "loss": 0.4089, + "step": 74304 + }, + { + "epoch": 2.86, + "learning_rate": 2.388432931520903e-06, + "loss": 0.191, + "step": 74305 + }, + { + "epoch": 2.86, + "learning_rate": 2.3877917414721724e-06, + "loss": 0.3073, + "step": 74306 + }, + { + "epoch": 2.86, + "learning_rate": 2.387150551423442e-06, + "loss": 0.2044, + "step": 74307 + }, + { + "epoch": 2.86, + "learning_rate": 2.3865093613747116e-06, + "loss": 0.1105, + "step": 74308 + }, + { + "epoch": 2.86, + "learning_rate": 2.3858681713259808e-06, + "loss": 0.2608, + "step": 74309 + }, + { + "epoch": 2.86, + "learning_rate": 2.3852269812772504e-06, + "loss": 0.2544, + "step": 74310 + }, + { + "epoch": 2.86, + "learning_rate": 2.38458579122852e-06, + "loss": 0.2245, + "step": 74311 + }, + { + "epoch": 2.86, + "learning_rate": 2.3839446011797896e-06, + "loss": 0.1444, + "step": 74312 + }, + { + "epoch": 2.86, + "learning_rate": 2.383303411131059e-06, + "loss": 0.277, + "step": 74313 + }, + { + "epoch": 2.86, + "learning_rate": 2.3826622210823288e-06, + "loss": 0.1965, + "step": 74314 + }, + { + "epoch": 2.86, + "learning_rate": 2.3820210310335984e-06, + "loss": 0.1161, + "step": 74315 + }, + { + "epoch": 2.86, + "learning_rate": 2.381379840984868e-06, + "loss": 0.1801, + "step": 74316 + }, + { + "epoch": 2.86, + "learning_rate": 2.3807386509361376e-06, + "loss": 0.1283, + "step": 74317 + }, + { + "epoch": 2.86, + "learning_rate": 2.380097460887407e-06, + "loss": 0.2535, + "step": 74318 + }, + { + "epoch": 2.86, + "learning_rate": 2.3794562708386764e-06, + "loss": 0.1354, + "step": 74319 + }, + { + "epoch": 2.86, + "learning_rate": 2.378815080789946e-06, + "loss": 0.1798, + "step": 74320 + }, + { + "epoch": 2.86, + "learning_rate": 2.3781738907412156e-06, + "loss": 0.2081, + "step": 74321 + }, + { + "epoch": 2.86, + "learning_rate": 2.377532700692485e-06, + "loss": 0.1615, + "step": 74322 + }, + { + "epoch": 2.86, + "learning_rate": 2.3768915106437548e-06, + "loss": 0.1325, + "step": 74323 + }, + { + "epoch": 2.86, + "learning_rate": 2.3762503205950243e-06, + "loss": 0.2342, + "step": 74324 + }, + { + "epoch": 2.86, + "learning_rate": 2.375609130546294e-06, + "loss": 0.1925, + "step": 74325 + }, + { + "epoch": 2.86, + "learning_rate": 2.3749679404975635e-06, + "loss": 0.2836, + "step": 74326 + }, + { + "epoch": 2.86, + "learning_rate": 2.374326750448833e-06, + "loss": 0.1513, + "step": 74327 + }, + { + "epoch": 2.86, + "learning_rate": 2.3736855604001027e-06, + "loss": 0.2826, + "step": 74328 + }, + { + "epoch": 2.86, + "learning_rate": 2.373044370351372e-06, + "loss": 0.2249, + "step": 74329 + }, + { + "epoch": 2.86, + "learning_rate": 2.3724031803026415e-06, + "loss": 0.1703, + "step": 74330 + }, + { + "epoch": 2.86, + "learning_rate": 2.371761990253911e-06, + "loss": 0.1163, + "step": 74331 + }, + { + "epoch": 2.86, + "learning_rate": 2.3711208002051807e-06, + "loss": 0.1745, + "step": 74332 + }, + { + "epoch": 2.86, + "learning_rate": 2.3704796101564503e-06, + "loss": 0.1626, + "step": 74333 + }, + { + "epoch": 2.86, + "learning_rate": 2.36983842010772e-06, + "loss": 0.3096, + "step": 74334 + }, + { + "epoch": 2.86, + "learning_rate": 2.3691972300589895e-06, + "loss": 0.1644, + "step": 74335 + }, + { + "epoch": 2.86, + "learning_rate": 2.368556040010259e-06, + "loss": 0.3881, + "step": 74336 + }, + { + "epoch": 2.86, + "learning_rate": 2.3679148499615287e-06, + "loss": 0.1645, + "step": 74337 + }, + { + "epoch": 2.86, + "learning_rate": 2.3672736599127983e-06, + "loss": 0.2656, + "step": 74338 + }, + { + "epoch": 2.86, + "learning_rate": 2.3666324698640675e-06, + "loss": 0.2387, + "step": 74339 + }, + { + "epoch": 2.86, + "learning_rate": 2.365991279815337e-06, + "loss": 0.3016, + "step": 74340 + }, + { + "epoch": 2.86, + "learning_rate": 2.3653500897666067e-06, + "loss": 0.2061, + "step": 74341 + }, + { + "epoch": 2.86, + "learning_rate": 2.3647088997178763e-06, + "loss": 0.1579, + "step": 74342 + }, + { + "epoch": 2.86, + "learning_rate": 2.364067709669146e-06, + "loss": 0.2884, + "step": 74343 + }, + { + "epoch": 2.86, + "learning_rate": 2.3634265196204155e-06, + "loss": 0.1869, + "step": 74344 + }, + { + "epoch": 2.86, + "learning_rate": 2.362785329571685e-06, + "loss": 0.2386, + "step": 74345 + }, + { + "epoch": 2.86, + "learning_rate": 2.3621441395229547e-06, + "loss": 0.1985, + "step": 74346 + }, + { + "epoch": 2.86, + "learning_rate": 2.3615029494742243e-06, + "loss": 0.2713, + "step": 74347 + }, + { + "epoch": 2.86, + "learning_rate": 2.360861759425494e-06, + "loss": 0.1535, + "step": 74348 + }, + { + "epoch": 2.86, + "learning_rate": 2.360220569376763e-06, + "loss": 0.2653, + "step": 74349 + }, + { + "epoch": 2.86, + "learning_rate": 2.3595793793280327e-06, + "loss": 0.2706, + "step": 74350 + }, + { + "epoch": 2.86, + "learning_rate": 2.3589381892793023e-06, + "loss": 0.2709, + "step": 74351 + }, + { + "epoch": 2.86, + "learning_rate": 2.358296999230572e-06, + "loss": 0.129, + "step": 74352 + }, + { + "epoch": 2.86, + "learning_rate": 2.3576558091818415e-06, + "loss": 0.2628, + "step": 74353 + }, + { + "epoch": 2.86, + "learning_rate": 2.357014619133111e-06, + "loss": 0.1396, + "step": 74354 + }, + { + "epoch": 2.86, + "learning_rate": 2.3563734290843806e-06, + "loss": 0.2509, + "step": 74355 + }, + { + "epoch": 2.86, + "learning_rate": 2.3557322390356502e-06, + "loss": 0.144, + "step": 74356 + }, + { + "epoch": 2.86, + "learning_rate": 2.35509104898692e-06, + "loss": 0.1442, + "step": 74357 + }, + { + "epoch": 2.86, + "learning_rate": 2.3544498589381894e-06, + "loss": 0.1139, + "step": 74358 + }, + { + "epoch": 2.86, + "learning_rate": 2.3538086688894586e-06, + "loss": 0.1883, + "step": 74359 + }, + { + "epoch": 2.86, + "learning_rate": 2.3531674788407282e-06, + "loss": 0.1865, + "step": 74360 + }, + { + "epoch": 2.86, + "learning_rate": 2.352526288791998e-06, + "loss": 0.1155, + "step": 74361 + }, + { + "epoch": 2.86, + "learning_rate": 2.3518850987432674e-06, + "loss": 0.3039, + "step": 74362 + }, + { + "epoch": 2.86, + "learning_rate": 2.351243908694537e-06, + "loss": 0.3371, + "step": 74363 + }, + { + "epoch": 2.86, + "learning_rate": 2.3506027186458066e-06, + "loss": 0.192, + "step": 74364 + }, + { + "epoch": 2.86, + "learning_rate": 2.3499615285970762e-06, + "loss": 0.1402, + "step": 74365 + }, + { + "epoch": 2.86, + "learning_rate": 2.349320338548346e-06, + "loss": 0.1631, + "step": 74366 + }, + { + "epoch": 2.86, + "learning_rate": 2.3486791484996154e-06, + "loss": 0.1543, + "step": 74367 + }, + { + "epoch": 2.86, + "learning_rate": 2.3480379584508846e-06, + "loss": 0.2595, + "step": 74368 + }, + { + "epoch": 2.86, + "learning_rate": 2.347396768402154e-06, + "loss": 0.2699, + "step": 74369 + }, + { + "epoch": 2.86, + "learning_rate": 2.3467555783534238e-06, + "loss": 0.1897, + "step": 74370 + }, + { + "epoch": 2.86, + "learning_rate": 2.3461143883046934e-06, + "loss": 0.2556, + "step": 74371 + }, + { + "epoch": 2.86, + "learning_rate": 2.345473198255963e-06, + "loss": 0.1787, + "step": 74372 + }, + { + "epoch": 2.86, + "learning_rate": 2.3448320082072326e-06, + "loss": 0.3389, + "step": 74373 + }, + { + "epoch": 2.86, + "learning_rate": 2.344190818158502e-06, + "loss": 0.1356, + "step": 74374 + }, + { + "epoch": 2.86, + "learning_rate": 2.3435496281097718e-06, + "loss": 0.345, + "step": 74375 + }, + { + "epoch": 2.86, + "learning_rate": 2.3429084380610414e-06, + "loss": 0.1784, + "step": 74376 + }, + { + "epoch": 2.86, + "learning_rate": 2.342267248012311e-06, + "loss": 0.2044, + "step": 74377 + }, + { + "epoch": 2.86, + "learning_rate": 2.34162605796358e-06, + "loss": 0.1703, + "step": 74378 + }, + { + "epoch": 2.86, + "learning_rate": 2.3409848679148498e-06, + "loss": 0.2485, + "step": 74379 + }, + { + "epoch": 2.86, + "learning_rate": 2.3403436778661194e-06, + "loss": 0.1374, + "step": 74380 + }, + { + "epoch": 2.86, + "learning_rate": 2.339702487817389e-06, + "loss": 0.2303, + "step": 74381 + }, + { + "epoch": 2.86, + "learning_rate": 2.3390612977686586e-06, + "loss": 0.3335, + "step": 74382 + }, + { + "epoch": 2.86, + "learning_rate": 2.338420107719928e-06, + "loss": 0.1609, + "step": 74383 + }, + { + "epoch": 2.86, + "learning_rate": 2.3377789176711978e-06, + "loss": 0.238, + "step": 74384 + }, + { + "epoch": 2.86, + "learning_rate": 2.3371377276224673e-06, + "loss": 0.1202, + "step": 74385 + }, + { + "epoch": 2.86, + "learning_rate": 2.336496537573737e-06, + "loss": 0.2457, + "step": 74386 + }, + { + "epoch": 2.86, + "learning_rate": 2.3358553475250065e-06, + "loss": 0.2648, + "step": 74387 + }, + { + "epoch": 2.86, + "learning_rate": 2.3352141574762757e-06, + "loss": 0.1478, + "step": 74388 + }, + { + "epoch": 2.86, + "learning_rate": 2.3345729674275453e-06, + "loss": 0.1879, + "step": 74389 + }, + { + "epoch": 2.86, + "learning_rate": 2.333931777378815e-06, + "loss": 0.1719, + "step": 74390 + }, + { + "epoch": 2.86, + "learning_rate": 2.3332905873300845e-06, + "loss": 0.2504, + "step": 74391 + }, + { + "epoch": 2.86, + "learning_rate": 2.332649397281354e-06, + "loss": 0.2996, + "step": 74392 + }, + { + "epoch": 2.86, + "learning_rate": 2.3320082072326237e-06, + "loss": 0.1465, + "step": 74393 + }, + { + "epoch": 2.86, + "learning_rate": 2.3313670171838933e-06, + "loss": 0.1427, + "step": 74394 + }, + { + "epoch": 2.86, + "learning_rate": 2.330725827135163e-06, + "loss": 0.2005, + "step": 74395 + }, + { + "epoch": 2.86, + "learning_rate": 2.3300846370864325e-06, + "loss": 0.1611, + "step": 74396 + }, + { + "epoch": 2.86, + "learning_rate": 2.329443447037702e-06, + "loss": 0.2953, + "step": 74397 + }, + { + "epoch": 2.86, + "learning_rate": 2.3288022569889713e-06, + "loss": 0.2611, + "step": 74398 + }, + { + "epoch": 2.86, + "learning_rate": 2.328161066940241e-06, + "loss": 0.1948, + "step": 74399 + }, + { + "epoch": 2.86, + "learning_rate": 2.3275198768915105e-06, + "loss": 0.1587, + "step": 74400 + }, + { + "epoch": 2.86, + "learning_rate": 2.32687868684278e-06, + "loss": 0.2331, + "step": 74401 + }, + { + "epoch": 2.86, + "learning_rate": 2.3262374967940497e-06, + "loss": 0.173, + "step": 74402 + }, + { + "epoch": 2.86, + "learning_rate": 2.3255963067453193e-06, + "loss": 0.3321, + "step": 74403 + }, + { + "epoch": 2.86, + "learning_rate": 2.324955116696589e-06, + "loss": 0.1948, + "step": 74404 + }, + { + "epoch": 2.86, + "learning_rate": 2.3243139266478585e-06, + "loss": 0.1051, + "step": 74405 + }, + { + "epoch": 2.86, + "learning_rate": 2.323672736599128e-06, + "loss": 0.2378, + "step": 74406 + }, + { + "epoch": 2.86, + "learning_rate": 2.3230315465503977e-06, + "loss": 0.1346, + "step": 74407 + }, + { + "epoch": 2.86, + "learning_rate": 2.322390356501667e-06, + "loss": 0.1512, + "step": 74408 + }, + { + "epoch": 2.86, + "learning_rate": 2.3217491664529365e-06, + "loss": 0.4041, + "step": 74409 + }, + { + "epoch": 2.86, + "learning_rate": 2.321107976404206e-06, + "loss": 0.0962, + "step": 74410 + }, + { + "epoch": 2.86, + "learning_rate": 2.3204667863554757e-06, + "loss": 0.1699, + "step": 74411 + }, + { + "epoch": 2.86, + "learning_rate": 2.3198255963067453e-06, + "loss": 0.2506, + "step": 74412 + }, + { + "epoch": 2.86, + "learning_rate": 2.319184406258015e-06, + "loss": 0.1745, + "step": 74413 + }, + { + "epoch": 2.86, + "learning_rate": 2.3185432162092845e-06, + "loss": 0.2984, + "step": 74414 + }, + { + "epoch": 2.86, + "learning_rate": 2.317902026160554e-06, + "loss": 0.1497, + "step": 74415 + }, + { + "epoch": 2.86, + "learning_rate": 2.3172608361118236e-06, + "loss": 0.2681, + "step": 74416 + }, + { + "epoch": 2.86, + "learning_rate": 2.3166196460630932e-06, + "loss": 0.1713, + "step": 74417 + }, + { + "epoch": 2.86, + "learning_rate": 2.3159784560143624e-06, + "loss": 0.1422, + "step": 74418 + }, + { + "epoch": 2.86, + "learning_rate": 2.315337265965632e-06, + "loss": 0.2425, + "step": 74419 + }, + { + "epoch": 2.86, + "learning_rate": 2.3146960759169016e-06, + "loss": 0.1398, + "step": 74420 + }, + { + "epoch": 2.86, + "learning_rate": 2.3140548858681712e-06, + "loss": 0.3121, + "step": 74421 + }, + { + "epoch": 2.86, + "learning_rate": 2.313413695819441e-06, + "loss": 0.2515, + "step": 74422 + }, + { + "epoch": 2.86, + "learning_rate": 2.3127725057707104e-06, + "loss": 0.2357, + "step": 74423 + }, + { + "epoch": 2.86, + "learning_rate": 2.31213131572198e-06, + "loss": 0.34, + "step": 74424 + }, + { + "epoch": 2.86, + "learning_rate": 2.3114901256732496e-06, + "loss": 0.181, + "step": 74425 + }, + { + "epoch": 2.86, + "learning_rate": 2.3108489356245192e-06, + "loss": 0.2564, + "step": 74426 + }, + { + "epoch": 2.86, + "learning_rate": 2.310207745575789e-06, + "loss": 0.1275, + "step": 74427 + }, + { + "epoch": 2.86, + "learning_rate": 2.309566555527058e-06, + "loss": 0.1265, + "step": 74428 + }, + { + "epoch": 2.86, + "learning_rate": 2.3089253654783276e-06, + "loss": 0.2347, + "step": 74429 + }, + { + "epoch": 2.86, + "learning_rate": 2.308284175429597e-06, + "loss": 0.1937, + "step": 74430 + }, + { + "epoch": 2.86, + "learning_rate": 2.3076429853808668e-06, + "loss": 0.2213, + "step": 74431 + }, + { + "epoch": 2.86, + "learning_rate": 2.3070017953321364e-06, + "loss": 0.2011, + "step": 74432 + }, + { + "epoch": 2.86, + "learning_rate": 2.306360605283406e-06, + "loss": 0.1379, + "step": 74433 + }, + { + "epoch": 2.86, + "learning_rate": 2.3057194152346756e-06, + "loss": 0.1408, + "step": 74434 + }, + { + "epoch": 2.86, + "learning_rate": 2.305078225185945e-06, + "loss": 0.2041, + "step": 74435 + }, + { + "epoch": 2.86, + "learning_rate": 2.3044370351372148e-06, + "loss": 0.1991, + "step": 74436 + }, + { + "epoch": 2.86, + "learning_rate": 2.3037958450884844e-06, + "loss": 0.254, + "step": 74437 + }, + { + "epoch": 2.86, + "learning_rate": 2.3031546550397536e-06, + "loss": 0.2981, + "step": 74438 + }, + { + "epoch": 2.86, + "learning_rate": 2.302513464991023e-06, + "loss": 0.1389, + "step": 74439 + }, + { + "epoch": 2.86, + "learning_rate": 2.3018722749422928e-06, + "loss": 0.1954, + "step": 74440 + }, + { + "epoch": 2.86, + "learning_rate": 2.3012310848935624e-06, + "loss": 0.2459, + "step": 74441 + }, + { + "epoch": 2.86, + "learning_rate": 2.300589894844832e-06, + "loss": 0.2388, + "step": 74442 + }, + { + "epoch": 2.86, + "learning_rate": 2.2999487047961016e-06, + "loss": 0.2808, + "step": 74443 + }, + { + "epoch": 2.86, + "learning_rate": 2.299307514747371e-06, + "loss": 0.2999, + "step": 74444 + }, + { + "epoch": 2.86, + "learning_rate": 2.2986663246986408e-06, + "loss": 0.3504, + "step": 74445 + }, + { + "epoch": 2.86, + "learning_rate": 2.2980251346499103e-06, + "loss": 0.1936, + "step": 74446 + }, + { + "epoch": 2.86, + "learning_rate": 2.29738394460118e-06, + "loss": 0.2372, + "step": 74447 + }, + { + "epoch": 2.86, + "learning_rate": 2.296742754552449e-06, + "loss": 0.1696, + "step": 74448 + }, + { + "epoch": 2.86, + "learning_rate": 2.2961015645037187e-06, + "loss": 0.2005, + "step": 74449 + }, + { + "epoch": 2.86, + "learning_rate": 2.2954603744549883e-06, + "loss": 0.2103, + "step": 74450 + }, + { + "epoch": 2.86, + "learning_rate": 2.294819184406258e-06, + "loss": 0.2354, + "step": 74451 + }, + { + "epoch": 2.86, + "learning_rate": 2.2941779943575275e-06, + "loss": 0.1605, + "step": 74452 + }, + { + "epoch": 2.86, + "learning_rate": 2.293536804308797e-06, + "loss": 0.2489, + "step": 74453 + }, + { + "epoch": 2.86, + "learning_rate": 2.2928956142600667e-06, + "loss": 0.1553, + "step": 74454 + }, + { + "epoch": 2.86, + "learning_rate": 2.2922544242113363e-06, + "loss": 0.1431, + "step": 74455 + }, + { + "epoch": 2.86, + "learning_rate": 2.291613234162606e-06, + "loss": 0.2678, + "step": 74456 + }, + { + "epoch": 2.86, + "learning_rate": 2.2909720441138755e-06, + "loss": 0.1742, + "step": 74457 + }, + { + "epoch": 2.86, + "learning_rate": 2.2903308540651447e-06, + "loss": 0.2209, + "step": 74458 + }, + { + "epoch": 2.86, + "learning_rate": 2.2896896640164143e-06, + "loss": 0.1583, + "step": 74459 + }, + { + "epoch": 2.86, + "learning_rate": 2.289048473967684e-06, + "loss": 0.1903, + "step": 74460 + }, + { + "epoch": 2.86, + "learning_rate": 2.2884072839189535e-06, + "loss": 0.1309, + "step": 74461 + }, + { + "epoch": 2.86, + "learning_rate": 2.287766093870223e-06, + "loss": 0.2157, + "step": 74462 + }, + { + "epoch": 2.86, + "learning_rate": 2.2871249038214927e-06, + "loss": 0.2659, + "step": 74463 + }, + { + "epoch": 2.86, + "learning_rate": 2.2864837137727623e-06, + "loss": 0.2496, + "step": 74464 + }, + { + "epoch": 2.86, + "learning_rate": 2.285842523724032e-06, + "loss": 0.1522, + "step": 74465 + }, + { + "epoch": 2.86, + "learning_rate": 2.2852013336753015e-06, + "loss": 0.299, + "step": 74466 + }, + { + "epoch": 2.86, + "learning_rate": 2.284560143626571e-06, + "loss": 0.2343, + "step": 74467 + }, + { + "epoch": 2.86, + "learning_rate": 2.2839189535778407e-06, + "loss": 0.205, + "step": 74468 + }, + { + "epoch": 2.86, + "learning_rate": 2.28327776352911e-06, + "loss": 0.1352, + "step": 74469 + }, + { + "epoch": 2.86, + "learning_rate": 2.2826365734803795e-06, + "loss": 0.1263, + "step": 74470 + }, + { + "epoch": 2.86, + "learning_rate": 2.281995383431649e-06, + "loss": 0.1853, + "step": 74471 + }, + { + "epoch": 2.86, + "learning_rate": 2.2813541933829187e-06, + "loss": 0.1588, + "step": 74472 + }, + { + "epoch": 2.86, + "learning_rate": 2.2807130033341883e-06, + "loss": 0.2974, + "step": 74473 + }, + { + "epoch": 2.86, + "learning_rate": 2.280071813285458e-06, + "loss": 0.1746, + "step": 74474 + }, + { + "epoch": 2.86, + "learning_rate": 2.2794306232367275e-06, + "loss": 0.2543, + "step": 74475 + }, + { + "epoch": 2.86, + "learning_rate": 2.278789433187997e-06, + "loss": 0.2642, + "step": 74476 + }, + { + "epoch": 2.86, + "learning_rate": 2.2781482431392666e-06, + "loss": 0.0847, + "step": 74477 + }, + { + "epoch": 2.86, + "learning_rate": 2.2775070530905362e-06, + "loss": 0.2123, + "step": 74478 + }, + { + "epoch": 2.86, + "learning_rate": 2.2768658630418054e-06, + "loss": 0.1308, + "step": 74479 + }, + { + "epoch": 2.86, + "learning_rate": 2.276224672993075e-06, + "loss": 0.1777, + "step": 74480 + }, + { + "epoch": 2.86, + "learning_rate": 2.2755834829443446e-06, + "loss": 0.1097, + "step": 74481 + }, + { + "epoch": 2.86, + "learning_rate": 2.2749422928956142e-06, + "loss": 0.1328, + "step": 74482 + }, + { + "epoch": 2.86, + "learning_rate": 2.274301102846884e-06, + "loss": 0.1778, + "step": 74483 + }, + { + "epoch": 2.86, + "learning_rate": 2.2736599127981534e-06, + "loss": 0.2343, + "step": 74484 + }, + { + "epoch": 2.86, + "learning_rate": 2.273018722749423e-06, + "loss": 0.2095, + "step": 74485 + }, + { + "epoch": 2.86, + "learning_rate": 2.2723775327006926e-06, + "loss": 0.1846, + "step": 74486 + }, + { + "epoch": 2.86, + "learning_rate": 2.2717363426519622e-06, + "loss": 0.1554, + "step": 74487 + }, + { + "epoch": 2.86, + "learning_rate": 2.271095152603232e-06, + "loss": 0.2317, + "step": 74488 + }, + { + "epoch": 2.86, + "learning_rate": 2.270453962554501e-06, + "loss": 0.1894, + "step": 74489 + }, + { + "epoch": 2.86, + "learning_rate": 2.2698127725057706e-06, + "loss": 0.2234, + "step": 74490 + }, + { + "epoch": 2.86, + "learning_rate": 2.26917158245704e-06, + "loss": 0.1858, + "step": 74491 + }, + { + "epoch": 2.86, + "learning_rate": 2.2685303924083098e-06, + "loss": 0.1012, + "step": 74492 + }, + { + "epoch": 2.86, + "learning_rate": 2.2678892023595794e-06, + "loss": 0.2002, + "step": 74493 + }, + { + "epoch": 2.86, + "learning_rate": 2.267248012310849e-06, + "loss": 0.2493, + "step": 74494 + }, + { + "epoch": 2.86, + "learning_rate": 2.2666068222621186e-06, + "loss": 0.1726, + "step": 74495 + }, + { + "epoch": 2.86, + "learning_rate": 2.265965632213388e-06, + "loss": 0.1841, + "step": 74496 + }, + { + "epoch": 2.86, + "learning_rate": 2.2653244421646578e-06, + "loss": 0.1562, + "step": 74497 + }, + { + "epoch": 2.86, + "learning_rate": 2.2646832521159274e-06, + "loss": 0.1473, + "step": 74498 + }, + { + "epoch": 2.86, + "learning_rate": 2.2640420620671966e-06, + "loss": 0.3537, + "step": 74499 + }, + { + "epoch": 2.86, + "learning_rate": 2.263400872018466e-06, + "loss": 0.1187, + "step": 74500 + }, + { + "epoch": 2.86, + "learning_rate": 2.2627596819697358e-06, + "loss": 0.29, + "step": 74501 + }, + { + "epoch": 2.86, + "learning_rate": 2.2621184919210054e-06, + "loss": 0.1143, + "step": 74502 + }, + { + "epoch": 2.86, + "learning_rate": 2.261477301872275e-06, + "loss": 0.2173, + "step": 74503 + }, + { + "epoch": 2.86, + "learning_rate": 2.2608361118235446e-06, + "loss": 0.1668, + "step": 74504 + }, + { + "epoch": 2.86, + "learning_rate": 2.260194921774814e-06, + "loss": 0.1053, + "step": 74505 + }, + { + "epoch": 2.86, + "learning_rate": 2.2595537317260838e-06, + "loss": 0.2251, + "step": 74506 + }, + { + "epoch": 2.86, + "learning_rate": 2.2589125416773534e-06, + "loss": 0.2523, + "step": 74507 + }, + { + "epoch": 2.86, + "learning_rate": 2.258271351628623e-06, + "loss": 0.4479, + "step": 74508 + }, + { + "epoch": 2.86, + "learning_rate": 2.257630161579892e-06, + "loss": 0.1981, + "step": 74509 + }, + { + "epoch": 2.86, + "learning_rate": 2.2569889715311617e-06, + "loss": 0.2453, + "step": 74510 + }, + { + "epoch": 2.86, + "learning_rate": 2.2563477814824313e-06, + "loss": 0.2005, + "step": 74511 + }, + { + "epoch": 2.86, + "learning_rate": 2.255706591433701e-06, + "loss": 0.2529, + "step": 74512 + }, + { + "epoch": 2.86, + "learning_rate": 2.2550654013849705e-06, + "loss": 0.2592, + "step": 74513 + }, + { + "epoch": 2.86, + "learning_rate": 2.25442421133624e-06, + "loss": 0.203, + "step": 74514 + }, + { + "epoch": 2.86, + "learning_rate": 2.2537830212875097e-06, + "loss": 0.1898, + "step": 74515 + }, + { + "epoch": 2.86, + "learning_rate": 2.2531418312387793e-06, + "loss": 0.1245, + "step": 74516 + }, + { + "epoch": 2.86, + "learning_rate": 2.252500641190049e-06, + "loss": 0.1673, + "step": 74517 + }, + { + "epoch": 2.86, + "learning_rate": 2.2518594511413185e-06, + "loss": 0.316, + "step": 74518 + }, + { + "epoch": 2.87, + "learning_rate": 2.251218261092588e-06, + "loss": 0.2568, + "step": 74519 + }, + { + "epoch": 2.87, + "learning_rate": 2.2505770710438573e-06, + "loss": 0.1053, + "step": 74520 + }, + { + "epoch": 2.87, + "learning_rate": 2.249935880995127e-06, + "loss": 0.1293, + "step": 74521 + }, + { + "epoch": 2.87, + "learning_rate": 2.2492946909463965e-06, + "loss": 0.1551, + "step": 74522 + }, + { + "epoch": 2.87, + "learning_rate": 2.248653500897666e-06, + "loss": 0.1673, + "step": 74523 + }, + { + "epoch": 2.87, + "learning_rate": 2.2480123108489357e-06, + "loss": 0.1905, + "step": 74524 + }, + { + "epoch": 2.87, + "learning_rate": 2.2473711208002053e-06, + "loss": 0.2652, + "step": 74525 + }, + { + "epoch": 2.87, + "learning_rate": 2.246729930751475e-06, + "loss": 0.1495, + "step": 74526 + }, + { + "epoch": 2.87, + "learning_rate": 2.2460887407027445e-06, + "loss": 0.2755, + "step": 74527 + }, + { + "epoch": 2.87, + "learning_rate": 2.245447550654014e-06, + "loss": 0.1317, + "step": 74528 + }, + { + "epoch": 2.87, + "learning_rate": 2.2448063606052837e-06, + "loss": 0.1842, + "step": 74529 + }, + { + "epoch": 2.87, + "learning_rate": 2.244165170556553e-06, + "loss": 0.3356, + "step": 74530 + }, + { + "epoch": 2.87, + "learning_rate": 2.2435239805078225e-06, + "loss": 0.3714, + "step": 74531 + }, + { + "epoch": 2.87, + "learning_rate": 2.242882790459092e-06, + "loss": 0.2925, + "step": 74532 + }, + { + "epoch": 2.87, + "learning_rate": 2.2422416004103617e-06, + "loss": 0.287, + "step": 74533 + }, + { + "epoch": 2.87, + "learning_rate": 2.2416004103616313e-06, + "loss": 0.1832, + "step": 74534 + }, + { + "epoch": 2.87, + "learning_rate": 2.240959220312901e-06, + "loss": 0.0785, + "step": 74535 + }, + { + "epoch": 2.87, + "learning_rate": 2.2403180302641705e-06, + "loss": 0.1383, + "step": 74536 + }, + { + "epoch": 2.87, + "learning_rate": 2.23967684021544e-06, + "loss": 0.2066, + "step": 74537 + }, + { + "epoch": 2.87, + "learning_rate": 2.2390356501667097e-06, + "loss": 0.3039, + "step": 74538 + }, + { + "epoch": 2.87, + "learning_rate": 2.2383944601179792e-06, + "loss": 0.2519, + "step": 74539 + }, + { + "epoch": 2.87, + "learning_rate": 2.2377532700692484e-06, + "loss": 0.1543, + "step": 74540 + }, + { + "epoch": 2.87, + "learning_rate": 2.237112080020518e-06, + "loss": 0.1533, + "step": 74541 + }, + { + "epoch": 2.87, + "learning_rate": 2.2364708899717876e-06, + "loss": 0.1749, + "step": 74542 + }, + { + "epoch": 2.87, + "learning_rate": 2.2358296999230572e-06, + "loss": 0.1438, + "step": 74543 + }, + { + "epoch": 2.87, + "learning_rate": 2.235188509874327e-06, + "loss": 0.263, + "step": 74544 + }, + { + "epoch": 2.87, + "learning_rate": 2.2345473198255964e-06, + "loss": 0.2965, + "step": 74545 + }, + { + "epoch": 2.87, + "learning_rate": 2.233906129776866e-06, + "loss": 0.1475, + "step": 74546 + }, + { + "epoch": 2.87, + "learning_rate": 2.2332649397281356e-06, + "loss": 0.2134, + "step": 74547 + }, + { + "epoch": 2.87, + "learning_rate": 2.2326237496794052e-06, + "loss": 0.2192, + "step": 74548 + }, + { + "epoch": 2.87, + "learning_rate": 2.231982559630675e-06, + "loss": 0.1464, + "step": 74549 + }, + { + "epoch": 2.87, + "learning_rate": 2.231341369581944e-06, + "loss": 0.1688, + "step": 74550 + }, + { + "epoch": 2.87, + "learning_rate": 2.2307001795332136e-06, + "loss": 0.1589, + "step": 74551 + }, + { + "epoch": 2.87, + "learning_rate": 2.230058989484483e-06, + "loss": 0.1857, + "step": 74552 + }, + { + "epoch": 2.87, + "learning_rate": 2.229417799435753e-06, + "loss": 0.2813, + "step": 74553 + }, + { + "epoch": 2.87, + "learning_rate": 2.2287766093870224e-06, + "loss": 0.155, + "step": 74554 + }, + { + "epoch": 2.87, + "learning_rate": 2.228135419338292e-06, + "loss": 0.2293, + "step": 74555 + }, + { + "epoch": 2.87, + "learning_rate": 2.2274942292895616e-06, + "loss": 0.2801, + "step": 74556 + }, + { + "epoch": 2.87, + "learning_rate": 2.226853039240831e-06, + "loss": 0.2475, + "step": 74557 + }, + { + "epoch": 2.87, + "learning_rate": 2.2262118491921008e-06, + "loss": 0.2411, + "step": 74558 + }, + { + "epoch": 2.87, + "learning_rate": 2.2255706591433704e-06, + "loss": 0.1219, + "step": 74559 + }, + { + "epoch": 2.87, + "learning_rate": 2.22492946909464e-06, + "loss": 0.1953, + "step": 74560 + }, + { + "epoch": 2.87, + "learning_rate": 2.224288279045909e-06, + "loss": 0.3057, + "step": 74561 + }, + { + "epoch": 2.87, + "learning_rate": 2.2236470889971788e-06, + "loss": 0.162, + "step": 74562 + }, + { + "epoch": 2.87, + "learning_rate": 2.2230058989484484e-06, + "loss": 0.1477, + "step": 74563 + }, + { + "epoch": 2.87, + "learning_rate": 2.222364708899718e-06, + "loss": 0.2112, + "step": 74564 + }, + { + "epoch": 2.87, + "learning_rate": 2.2217235188509876e-06, + "loss": 0.2164, + "step": 74565 + }, + { + "epoch": 2.87, + "learning_rate": 2.221082328802257e-06, + "loss": 0.0881, + "step": 74566 + }, + { + "epoch": 2.87, + "learning_rate": 2.2204411387535268e-06, + "loss": 0.195, + "step": 74567 + }, + { + "epoch": 2.87, + "learning_rate": 2.2197999487047964e-06, + "loss": 0.1328, + "step": 74568 + }, + { + "epoch": 2.87, + "learning_rate": 2.219158758656066e-06, + "loss": 0.1691, + "step": 74569 + }, + { + "epoch": 2.87, + "learning_rate": 2.2185175686073355e-06, + "loss": 0.1544, + "step": 74570 + }, + { + "epoch": 2.87, + "learning_rate": 2.2178763785586047e-06, + "loss": 0.253, + "step": 74571 + }, + { + "epoch": 2.87, + "learning_rate": 2.2172351885098743e-06, + "loss": 0.2366, + "step": 74572 + }, + { + "epoch": 2.87, + "learning_rate": 2.216593998461144e-06, + "loss": 0.1438, + "step": 74573 + }, + { + "epoch": 2.87, + "learning_rate": 2.2159528084124135e-06, + "loss": 0.2223, + "step": 74574 + }, + { + "epoch": 2.87, + "learning_rate": 2.215311618363683e-06, + "loss": 0.3211, + "step": 74575 + }, + { + "epoch": 2.87, + "learning_rate": 2.2146704283149527e-06, + "loss": 0.2857, + "step": 74576 + }, + { + "epoch": 2.87, + "learning_rate": 2.2140292382662223e-06, + "loss": 0.2897, + "step": 74577 + }, + { + "epoch": 2.87, + "learning_rate": 2.213388048217492e-06, + "loss": 0.1066, + "step": 74578 + }, + { + "epoch": 2.87, + "learning_rate": 2.2127468581687615e-06, + "loss": 0.2138, + "step": 74579 + }, + { + "epoch": 2.87, + "learning_rate": 2.212105668120031e-06, + "loss": 0.2096, + "step": 74580 + }, + { + "epoch": 2.87, + "learning_rate": 2.2114644780713003e-06, + "loss": 0.2674, + "step": 74581 + }, + { + "epoch": 2.87, + "learning_rate": 2.21082328802257e-06, + "loss": 0.1839, + "step": 74582 + }, + { + "epoch": 2.87, + "learning_rate": 2.2101820979738395e-06, + "loss": 0.2065, + "step": 74583 + }, + { + "epoch": 2.87, + "learning_rate": 2.209540907925109e-06, + "loss": 0.1413, + "step": 74584 + }, + { + "epoch": 2.87, + "learning_rate": 2.2088997178763787e-06, + "loss": 0.1477, + "step": 74585 + }, + { + "epoch": 2.87, + "learning_rate": 2.2082585278276483e-06, + "loss": 0.1105, + "step": 74586 + }, + { + "epoch": 2.87, + "learning_rate": 2.207617337778918e-06, + "loss": 0.1793, + "step": 74587 + }, + { + "epoch": 2.87, + "learning_rate": 2.2069761477301875e-06, + "loss": 0.1267, + "step": 74588 + }, + { + "epoch": 2.87, + "learning_rate": 2.206334957681457e-06, + "loss": 0.2233, + "step": 74589 + }, + { + "epoch": 2.87, + "learning_rate": 2.2056937676327267e-06, + "loss": 0.2869, + "step": 74590 + }, + { + "epoch": 2.87, + "learning_rate": 2.205052577583996e-06, + "loss": 0.1914, + "step": 74591 + }, + { + "epoch": 2.87, + "learning_rate": 2.2044113875352655e-06, + "loss": 0.1395, + "step": 74592 + }, + { + "epoch": 2.87, + "learning_rate": 2.203770197486535e-06, + "loss": 0.2538, + "step": 74593 + }, + { + "epoch": 2.87, + "learning_rate": 2.2031290074378047e-06, + "loss": 0.1885, + "step": 74594 + }, + { + "epoch": 2.87, + "learning_rate": 2.2024878173890743e-06, + "loss": 0.1416, + "step": 74595 + }, + { + "epoch": 2.87, + "learning_rate": 2.201846627340344e-06, + "loss": 0.2329, + "step": 74596 + }, + { + "epoch": 2.87, + "learning_rate": 2.2012054372916135e-06, + "loss": 0.1318, + "step": 74597 + }, + { + "epoch": 2.87, + "learning_rate": 2.200564247242883e-06, + "loss": 0.1656, + "step": 74598 + }, + { + "epoch": 2.87, + "learning_rate": 2.1999230571941527e-06, + "loss": 0.3409, + "step": 74599 + }, + { + "epoch": 2.87, + "learning_rate": 2.1992818671454222e-06, + "loss": 0.2252, + "step": 74600 + }, + { + "epoch": 2.87, + "learning_rate": 2.1986406770966914e-06, + "loss": 0.178, + "step": 74601 + }, + { + "epoch": 2.87, + "learning_rate": 2.197999487047961e-06, + "loss": 0.1536, + "step": 74602 + }, + { + "epoch": 2.87, + "learning_rate": 2.1973582969992306e-06, + "loss": 0.2456, + "step": 74603 + }, + { + "epoch": 2.87, + "learning_rate": 2.1967171069505002e-06, + "loss": 0.1375, + "step": 74604 + }, + { + "epoch": 2.87, + "learning_rate": 2.19607591690177e-06, + "loss": 0.1821, + "step": 74605 + }, + { + "epoch": 2.87, + "learning_rate": 2.1954347268530394e-06, + "loss": 0.2237, + "step": 74606 + }, + { + "epoch": 2.87, + "learning_rate": 2.194793536804309e-06, + "loss": 0.0985, + "step": 74607 + }, + { + "epoch": 2.87, + "learning_rate": 2.1941523467555786e-06, + "loss": 0.1804, + "step": 74608 + }, + { + "epoch": 2.87, + "learning_rate": 2.1935111567068482e-06, + "loss": 0.2774, + "step": 74609 + }, + { + "epoch": 2.87, + "learning_rate": 2.192869966658118e-06, + "loss": 0.1811, + "step": 74610 + }, + { + "epoch": 2.87, + "learning_rate": 2.1922287766093874e-06, + "loss": 0.2894, + "step": 74611 + }, + { + "epoch": 2.87, + "learning_rate": 2.1915875865606566e-06, + "loss": 0.154, + "step": 74612 + }, + { + "epoch": 2.87, + "learning_rate": 2.190946396511926e-06, + "loss": 0.2557, + "step": 74613 + }, + { + "epoch": 2.87, + "learning_rate": 2.190305206463196e-06, + "loss": 0.2169, + "step": 74614 + }, + { + "epoch": 2.87, + "learning_rate": 2.1896640164144654e-06, + "loss": 0.1225, + "step": 74615 + }, + { + "epoch": 2.87, + "learning_rate": 2.189022826365735e-06, + "loss": 0.2132, + "step": 74616 + }, + { + "epoch": 2.87, + "learning_rate": 2.1883816363170046e-06, + "loss": 0.324, + "step": 74617 + }, + { + "epoch": 2.87, + "learning_rate": 2.187740446268274e-06, + "loss": 0.1181, + "step": 74618 + }, + { + "epoch": 2.87, + "learning_rate": 2.1870992562195438e-06, + "loss": 0.1751, + "step": 74619 + }, + { + "epoch": 2.87, + "learning_rate": 2.1864580661708134e-06, + "loss": 0.1574, + "step": 74620 + }, + { + "epoch": 2.87, + "learning_rate": 2.185816876122083e-06, + "loss": 0.1578, + "step": 74621 + }, + { + "epoch": 2.87, + "learning_rate": 2.185175686073352e-06, + "loss": 0.1168, + "step": 74622 + }, + { + "epoch": 2.87, + "learning_rate": 2.1845344960246218e-06, + "loss": 0.1268, + "step": 74623 + }, + { + "epoch": 2.87, + "learning_rate": 2.1838933059758914e-06, + "loss": 0.2335, + "step": 74624 + }, + { + "epoch": 2.87, + "learning_rate": 2.183252115927161e-06, + "loss": 0.1433, + "step": 74625 + }, + { + "epoch": 2.87, + "learning_rate": 2.1826109258784306e-06, + "loss": 0.1012, + "step": 74626 + }, + { + "epoch": 2.87, + "learning_rate": 2.1819697358297e-06, + "loss": 0.1677, + "step": 74627 + }, + { + "epoch": 2.87, + "learning_rate": 2.1813285457809698e-06, + "loss": 0.2546, + "step": 74628 + }, + { + "epoch": 2.87, + "learning_rate": 2.1806873557322394e-06, + "loss": 0.142, + "step": 74629 + }, + { + "epoch": 2.87, + "learning_rate": 2.180046165683509e-06, + "loss": 0.1641, + "step": 74630 + }, + { + "epoch": 2.87, + "learning_rate": 2.1794049756347785e-06, + "loss": 0.2401, + "step": 74631 + }, + { + "epoch": 2.87, + "learning_rate": 2.1787637855860477e-06, + "loss": 0.1519, + "step": 74632 + }, + { + "epoch": 2.87, + "learning_rate": 2.1781225955373173e-06, + "loss": 0.1829, + "step": 74633 + }, + { + "epoch": 2.87, + "learning_rate": 2.177481405488587e-06, + "loss": 0.2561, + "step": 74634 + }, + { + "epoch": 2.87, + "learning_rate": 2.1768402154398565e-06, + "loss": 0.1354, + "step": 74635 + }, + { + "epoch": 2.87, + "learning_rate": 2.176199025391126e-06, + "loss": 0.226, + "step": 74636 + }, + { + "epoch": 2.87, + "learning_rate": 2.1755578353423957e-06, + "loss": 0.1443, + "step": 74637 + }, + { + "epoch": 2.87, + "learning_rate": 2.1749166452936653e-06, + "loss": 0.1591, + "step": 74638 + }, + { + "epoch": 2.87, + "learning_rate": 2.174275455244935e-06, + "loss": 0.1585, + "step": 74639 + }, + { + "epoch": 2.87, + "learning_rate": 2.1736342651962045e-06, + "loss": 0.1252, + "step": 74640 + }, + { + "epoch": 2.87, + "learning_rate": 2.172993075147474e-06, + "loss": 0.1371, + "step": 74641 + }, + { + "epoch": 2.87, + "learning_rate": 2.1723518850987433e-06, + "loss": 0.287, + "step": 74642 + }, + { + "epoch": 2.87, + "learning_rate": 2.171710695050013e-06, + "loss": 0.1648, + "step": 74643 + }, + { + "epoch": 2.87, + "learning_rate": 2.1710695050012825e-06, + "loss": 0.2347, + "step": 74644 + }, + { + "epoch": 2.87, + "learning_rate": 2.170428314952552e-06, + "loss": 0.3294, + "step": 74645 + }, + { + "epoch": 2.87, + "learning_rate": 2.1697871249038217e-06, + "loss": 0.2197, + "step": 74646 + }, + { + "epoch": 2.87, + "learning_rate": 2.1691459348550913e-06, + "loss": 0.2151, + "step": 74647 + }, + { + "epoch": 2.87, + "learning_rate": 2.168504744806361e-06, + "loss": 0.188, + "step": 74648 + }, + { + "epoch": 2.87, + "learning_rate": 2.1678635547576305e-06, + "loss": 0.1902, + "step": 74649 + }, + { + "epoch": 2.87, + "learning_rate": 2.1672223647089e-06, + "loss": 0.2142, + "step": 74650 + }, + { + "epoch": 2.87, + "learning_rate": 2.1665811746601697e-06, + "loss": 0.2833, + "step": 74651 + }, + { + "epoch": 2.87, + "learning_rate": 2.1659399846114393e-06, + "loss": 0.2956, + "step": 74652 + }, + { + "epoch": 2.87, + "learning_rate": 2.1652987945627085e-06, + "loss": 0.2499, + "step": 74653 + }, + { + "epoch": 2.87, + "learning_rate": 2.164657604513978e-06, + "loss": 0.1397, + "step": 74654 + }, + { + "epoch": 2.87, + "learning_rate": 2.1640164144652477e-06, + "loss": 0.2281, + "step": 74655 + }, + { + "epoch": 2.87, + "learning_rate": 2.1633752244165173e-06, + "loss": 0.2134, + "step": 74656 + }, + { + "epoch": 2.87, + "learning_rate": 2.162734034367787e-06, + "loss": 0.1649, + "step": 74657 + }, + { + "epoch": 2.87, + "learning_rate": 2.1620928443190565e-06, + "loss": 0.1884, + "step": 74658 + }, + { + "epoch": 2.87, + "learning_rate": 2.161451654270326e-06, + "loss": 0.1549, + "step": 74659 + }, + { + "epoch": 2.87, + "learning_rate": 2.1608104642215957e-06, + "loss": 0.1661, + "step": 74660 + }, + { + "epoch": 2.87, + "learning_rate": 2.1601692741728652e-06, + "loss": 0.4023, + "step": 74661 + }, + { + "epoch": 2.87, + "learning_rate": 2.159528084124135e-06, + "loss": 0.174, + "step": 74662 + }, + { + "epoch": 2.87, + "learning_rate": 2.158886894075404e-06, + "loss": 0.2671, + "step": 74663 + }, + { + "epoch": 2.87, + "learning_rate": 2.1582457040266736e-06, + "loss": 0.2099, + "step": 74664 + }, + { + "epoch": 2.87, + "learning_rate": 2.1576045139779432e-06, + "loss": 0.2181, + "step": 74665 + }, + { + "epoch": 2.87, + "learning_rate": 2.156963323929213e-06, + "loss": 0.1964, + "step": 74666 + }, + { + "epoch": 2.87, + "learning_rate": 2.1563221338804824e-06, + "loss": 0.3278, + "step": 74667 + }, + { + "epoch": 2.87, + "learning_rate": 2.155680943831752e-06, + "loss": 0.1827, + "step": 74668 + }, + { + "epoch": 2.87, + "learning_rate": 2.1550397537830216e-06, + "loss": 0.1538, + "step": 74669 + }, + { + "epoch": 2.87, + "learning_rate": 2.1543985637342912e-06, + "loss": 0.1333, + "step": 74670 + }, + { + "epoch": 2.87, + "learning_rate": 2.153757373685561e-06, + "loss": 0.2711, + "step": 74671 + }, + { + "epoch": 2.87, + "learning_rate": 2.1531161836368304e-06, + "loss": 0.1562, + "step": 74672 + }, + { + "epoch": 2.87, + "learning_rate": 2.1524749935880996e-06, + "loss": 0.1765, + "step": 74673 + }, + { + "epoch": 2.87, + "learning_rate": 2.151833803539369e-06, + "loss": 0.1615, + "step": 74674 + }, + { + "epoch": 2.87, + "learning_rate": 2.151192613490639e-06, + "loss": 0.212, + "step": 74675 + }, + { + "epoch": 2.87, + "learning_rate": 2.1505514234419084e-06, + "loss": 0.2169, + "step": 74676 + }, + { + "epoch": 2.87, + "learning_rate": 2.149910233393178e-06, + "loss": 0.1856, + "step": 74677 + }, + { + "epoch": 2.87, + "learning_rate": 2.1492690433444476e-06, + "loss": 0.2147, + "step": 74678 + }, + { + "epoch": 2.87, + "learning_rate": 2.148627853295717e-06, + "loss": 0.246, + "step": 74679 + }, + { + "epoch": 2.87, + "learning_rate": 2.1479866632469868e-06, + "loss": 0.2441, + "step": 74680 + }, + { + "epoch": 2.87, + "learning_rate": 2.1473454731982564e-06, + "loss": 0.277, + "step": 74681 + }, + { + "epoch": 2.87, + "learning_rate": 2.146704283149526e-06, + "loss": 0.1206, + "step": 74682 + }, + { + "epoch": 2.87, + "learning_rate": 2.146063093100795e-06, + "loss": 0.2407, + "step": 74683 + }, + { + "epoch": 2.87, + "learning_rate": 2.1454219030520648e-06, + "loss": 0.1754, + "step": 74684 + }, + { + "epoch": 2.87, + "learning_rate": 2.1447807130033344e-06, + "loss": 0.2097, + "step": 74685 + }, + { + "epoch": 2.87, + "learning_rate": 2.144139522954604e-06, + "loss": 0.1595, + "step": 74686 + }, + { + "epoch": 2.87, + "learning_rate": 2.1434983329058736e-06, + "loss": 0.2227, + "step": 74687 + }, + { + "epoch": 2.87, + "learning_rate": 2.142857142857143e-06, + "loss": 0.2583, + "step": 74688 + }, + { + "epoch": 2.87, + "learning_rate": 2.1422159528084128e-06, + "loss": 0.1991, + "step": 74689 + }, + { + "epoch": 2.87, + "learning_rate": 2.1415747627596824e-06, + "loss": 0.2805, + "step": 74690 + }, + { + "epoch": 2.87, + "learning_rate": 2.140933572710952e-06, + "loss": 0.324, + "step": 74691 + }, + { + "epoch": 2.87, + "learning_rate": 2.1402923826622215e-06, + "loss": 0.2069, + "step": 74692 + }, + { + "epoch": 2.87, + "learning_rate": 2.1396511926134907e-06, + "loss": 0.1442, + "step": 74693 + }, + { + "epoch": 2.87, + "learning_rate": 2.1390100025647603e-06, + "loss": 0.1085, + "step": 74694 + }, + { + "epoch": 2.87, + "learning_rate": 2.13836881251603e-06, + "loss": 0.2325, + "step": 74695 + }, + { + "epoch": 2.87, + "learning_rate": 2.1377276224672995e-06, + "loss": 0.2576, + "step": 74696 + }, + { + "epoch": 2.87, + "learning_rate": 2.137086432418569e-06, + "loss": 0.2117, + "step": 74697 + }, + { + "epoch": 2.87, + "learning_rate": 2.1364452423698387e-06, + "loss": 0.2543, + "step": 74698 + }, + { + "epoch": 2.87, + "learning_rate": 2.1358040523211083e-06, + "loss": 0.1488, + "step": 74699 + }, + { + "epoch": 2.87, + "learning_rate": 2.135162862272378e-06, + "loss": 0.1523, + "step": 74700 + }, + { + "epoch": 2.87, + "learning_rate": 2.1345216722236475e-06, + "loss": 0.1554, + "step": 74701 + }, + { + "epoch": 2.87, + "learning_rate": 2.133880482174917e-06, + "loss": 0.2371, + "step": 74702 + }, + { + "epoch": 2.87, + "learning_rate": 2.1332392921261863e-06, + "loss": 0.2487, + "step": 74703 + }, + { + "epoch": 2.87, + "learning_rate": 2.132598102077456e-06, + "loss": 0.2312, + "step": 74704 + }, + { + "epoch": 2.87, + "learning_rate": 2.1319569120287255e-06, + "loss": 0.113, + "step": 74705 + }, + { + "epoch": 2.87, + "learning_rate": 2.131315721979995e-06, + "loss": 0.2432, + "step": 74706 + }, + { + "epoch": 2.87, + "learning_rate": 2.1306745319312647e-06, + "loss": 0.2185, + "step": 74707 + }, + { + "epoch": 2.87, + "learning_rate": 2.1300333418825343e-06, + "loss": 0.3151, + "step": 74708 + }, + { + "epoch": 2.87, + "learning_rate": 2.129392151833804e-06, + "loss": 0.23, + "step": 74709 + }, + { + "epoch": 2.87, + "learning_rate": 2.1287509617850735e-06, + "loss": 0.3393, + "step": 74710 + }, + { + "epoch": 2.87, + "learning_rate": 2.128109771736343e-06, + "loss": 0.1634, + "step": 74711 + }, + { + "epoch": 2.87, + "learning_rate": 2.1274685816876127e-06, + "loss": 0.2162, + "step": 74712 + }, + { + "epoch": 2.87, + "learning_rate": 2.126827391638882e-06, + "loss": 0.206, + "step": 74713 + }, + { + "epoch": 2.87, + "learning_rate": 2.1261862015901515e-06, + "loss": 0.2184, + "step": 74714 + }, + { + "epoch": 2.87, + "learning_rate": 2.125545011541421e-06, + "loss": 0.2378, + "step": 74715 + }, + { + "epoch": 2.87, + "learning_rate": 2.1249038214926907e-06, + "loss": 0.1434, + "step": 74716 + }, + { + "epoch": 2.87, + "learning_rate": 2.1242626314439603e-06, + "loss": 0.2324, + "step": 74717 + }, + { + "epoch": 2.87, + "learning_rate": 2.12362144139523e-06, + "loss": 0.2076, + "step": 74718 + }, + { + "epoch": 2.87, + "learning_rate": 2.1229802513464995e-06, + "loss": 0.254, + "step": 74719 + }, + { + "epoch": 2.87, + "learning_rate": 2.122339061297769e-06, + "loss": 0.3132, + "step": 74720 + }, + { + "epoch": 2.87, + "learning_rate": 2.1216978712490387e-06, + "loss": 0.1459, + "step": 74721 + }, + { + "epoch": 2.87, + "learning_rate": 2.121056681200308e-06, + "loss": 0.1844, + "step": 74722 + }, + { + "epoch": 2.87, + "learning_rate": 2.1204154911515774e-06, + "loss": 0.2174, + "step": 74723 + }, + { + "epoch": 2.87, + "learning_rate": 2.119774301102847e-06, + "loss": 0.2577, + "step": 74724 + }, + { + "epoch": 2.87, + "learning_rate": 2.1191331110541166e-06, + "loss": 0.1605, + "step": 74725 + }, + { + "epoch": 2.87, + "learning_rate": 2.1184919210053862e-06, + "loss": 0.3008, + "step": 74726 + }, + { + "epoch": 2.87, + "learning_rate": 2.117850730956656e-06, + "loss": 0.1449, + "step": 74727 + }, + { + "epoch": 2.87, + "learning_rate": 2.1172095409079254e-06, + "loss": 0.1747, + "step": 74728 + }, + { + "epoch": 2.87, + "learning_rate": 2.116568350859195e-06, + "loss": 0.2051, + "step": 74729 + }, + { + "epoch": 2.87, + "learning_rate": 2.1159271608104646e-06, + "loss": 0.2569, + "step": 74730 + }, + { + "epoch": 2.87, + "learning_rate": 2.1152859707617342e-06, + "loss": 0.1337, + "step": 74731 + }, + { + "epoch": 2.87, + "learning_rate": 2.1146447807130034e-06, + "loss": 0.1954, + "step": 74732 + }, + { + "epoch": 2.87, + "learning_rate": 2.114003590664273e-06, + "loss": 0.2351, + "step": 74733 + }, + { + "epoch": 2.87, + "learning_rate": 2.1133624006155426e-06, + "loss": 0.3625, + "step": 74734 + }, + { + "epoch": 2.87, + "learning_rate": 2.112721210566812e-06, + "loss": 0.3808, + "step": 74735 + }, + { + "epoch": 2.87, + "learning_rate": 2.112080020518082e-06, + "loss": 0.2574, + "step": 74736 + }, + { + "epoch": 2.87, + "learning_rate": 2.1114388304693514e-06, + "loss": 0.2608, + "step": 74737 + }, + { + "epoch": 2.87, + "learning_rate": 2.110797640420621e-06, + "loss": 0.1594, + "step": 74738 + }, + { + "epoch": 2.87, + "learning_rate": 2.1101564503718906e-06, + "loss": 0.2532, + "step": 74739 + }, + { + "epoch": 2.87, + "learning_rate": 2.10951526032316e-06, + "loss": 0.1491, + "step": 74740 + }, + { + "epoch": 2.87, + "learning_rate": 2.1088740702744298e-06, + "loss": 0.126, + "step": 74741 + }, + { + "epoch": 2.87, + "learning_rate": 2.108232880225699e-06, + "loss": 0.2781, + "step": 74742 + }, + { + "epoch": 2.87, + "learning_rate": 2.1075916901769686e-06, + "loss": 0.1246, + "step": 74743 + }, + { + "epoch": 2.87, + "learning_rate": 2.106950500128238e-06, + "loss": 0.2367, + "step": 74744 + }, + { + "epoch": 2.87, + "learning_rate": 2.1063093100795078e-06, + "loss": 0.2281, + "step": 74745 + }, + { + "epoch": 2.87, + "learning_rate": 2.1056681200307774e-06, + "loss": 0.1214, + "step": 74746 + }, + { + "epoch": 2.87, + "learning_rate": 2.105026929982047e-06, + "loss": 0.1037, + "step": 74747 + }, + { + "epoch": 2.87, + "learning_rate": 2.1043857399333166e-06, + "loss": 0.2209, + "step": 74748 + }, + { + "epoch": 2.87, + "learning_rate": 2.103744549884586e-06, + "loss": 0.187, + "step": 74749 + }, + { + "epoch": 2.87, + "learning_rate": 2.1031033598358558e-06, + "loss": 0.2362, + "step": 74750 + }, + { + "epoch": 2.87, + "learning_rate": 2.1024621697871254e-06, + "loss": 0.2189, + "step": 74751 + }, + { + "epoch": 2.87, + "learning_rate": 2.1018209797383945e-06, + "loss": 0.1511, + "step": 74752 + }, + { + "epoch": 2.87, + "learning_rate": 2.101179789689664e-06, + "loss": 0.1532, + "step": 74753 + }, + { + "epoch": 2.87, + "learning_rate": 2.1005385996409337e-06, + "loss": 0.2438, + "step": 74754 + }, + { + "epoch": 2.87, + "learning_rate": 2.0998974095922033e-06, + "loss": 0.1589, + "step": 74755 + }, + { + "epoch": 2.87, + "learning_rate": 2.099256219543473e-06, + "loss": 0.2407, + "step": 74756 + }, + { + "epoch": 2.87, + "learning_rate": 2.0986150294947425e-06, + "loss": 0.248, + "step": 74757 + }, + { + "epoch": 2.87, + "learning_rate": 2.097973839446012e-06, + "loss": 0.3068, + "step": 74758 + }, + { + "epoch": 2.87, + "learning_rate": 2.0973326493972817e-06, + "loss": 0.1646, + "step": 74759 + }, + { + "epoch": 2.87, + "learning_rate": 2.0966914593485513e-06, + "loss": 0.1684, + "step": 74760 + }, + { + "epoch": 2.87, + "learning_rate": 2.096050269299821e-06, + "loss": 0.1391, + "step": 74761 + }, + { + "epoch": 2.87, + "learning_rate": 2.09540907925109e-06, + "loss": 0.176, + "step": 74762 + }, + { + "epoch": 2.87, + "learning_rate": 2.0947678892023597e-06, + "loss": 0.2343, + "step": 74763 + }, + { + "epoch": 2.87, + "learning_rate": 2.0941266991536293e-06, + "loss": 0.3027, + "step": 74764 + }, + { + "epoch": 2.87, + "learning_rate": 2.093485509104899e-06, + "loss": 0.2066, + "step": 74765 + }, + { + "epoch": 2.87, + "learning_rate": 2.0928443190561685e-06, + "loss": 0.1403, + "step": 74766 + }, + { + "epoch": 2.87, + "learning_rate": 2.092203129007438e-06, + "loss": 0.2246, + "step": 74767 + }, + { + "epoch": 2.87, + "learning_rate": 2.0915619389587077e-06, + "loss": 0.2053, + "step": 74768 + }, + { + "epoch": 2.87, + "learning_rate": 2.0909207489099773e-06, + "loss": 0.1581, + "step": 74769 + }, + { + "epoch": 2.87, + "learning_rate": 2.090279558861247e-06, + "loss": 0.1361, + "step": 74770 + }, + { + "epoch": 2.87, + "learning_rate": 2.0896383688125165e-06, + "loss": 0.1493, + "step": 74771 + }, + { + "epoch": 2.87, + "learning_rate": 2.0889971787637857e-06, + "loss": 0.2115, + "step": 74772 + }, + { + "epoch": 2.87, + "learning_rate": 2.0883559887150553e-06, + "loss": 0.3088, + "step": 74773 + }, + { + "epoch": 2.87, + "learning_rate": 2.087714798666325e-06, + "loss": 0.2579, + "step": 74774 + }, + { + "epoch": 2.87, + "learning_rate": 2.0870736086175945e-06, + "loss": 0.187, + "step": 74775 + }, + { + "epoch": 2.87, + "learning_rate": 2.086432418568864e-06, + "loss": 0.1263, + "step": 74776 + }, + { + "epoch": 2.87, + "learning_rate": 2.0857912285201337e-06, + "loss": 0.1869, + "step": 74777 + }, + { + "epoch": 2.87, + "learning_rate": 2.0851500384714033e-06, + "loss": 0.1722, + "step": 74778 + }, + { + "epoch": 2.88, + "learning_rate": 2.084508848422673e-06, + "loss": 0.2001, + "step": 74779 + }, + { + "epoch": 2.88, + "learning_rate": 2.0838676583739425e-06, + "loss": 0.0992, + "step": 74780 + }, + { + "epoch": 2.88, + "learning_rate": 2.0832264683252116e-06, + "loss": 0.1309, + "step": 74781 + }, + { + "epoch": 2.88, + "learning_rate": 2.0825852782764812e-06, + "loss": 0.1557, + "step": 74782 + }, + { + "epoch": 2.88, + "learning_rate": 2.081944088227751e-06, + "loss": 0.1448, + "step": 74783 + }, + { + "epoch": 2.88, + "learning_rate": 2.0813028981790204e-06, + "loss": 0.14, + "step": 74784 + }, + { + "epoch": 2.88, + "learning_rate": 2.08066170813029e-06, + "loss": 0.2363, + "step": 74785 + }, + { + "epoch": 2.88, + "learning_rate": 2.0800205180815596e-06, + "loss": 0.2149, + "step": 74786 + }, + { + "epoch": 2.88, + "learning_rate": 2.0793793280328292e-06, + "loss": 0.2352, + "step": 74787 + }, + { + "epoch": 2.88, + "learning_rate": 2.078738137984099e-06, + "loss": 0.1166, + "step": 74788 + }, + { + "epoch": 2.88, + "learning_rate": 2.0780969479353684e-06, + "loss": 0.1514, + "step": 74789 + }, + { + "epoch": 2.88, + "learning_rate": 2.077455757886638e-06, + "loss": 0.2127, + "step": 74790 + }, + { + "epoch": 2.88, + "learning_rate": 2.076814567837907e-06, + "loss": 0.0929, + "step": 74791 + }, + { + "epoch": 2.88, + "learning_rate": 2.076173377789177e-06, + "loss": 0.3109, + "step": 74792 + }, + { + "epoch": 2.88, + "learning_rate": 2.0755321877404464e-06, + "loss": 0.1458, + "step": 74793 + }, + { + "epoch": 2.88, + "learning_rate": 2.074890997691716e-06, + "loss": 0.0807, + "step": 74794 + }, + { + "epoch": 2.88, + "learning_rate": 2.0742498076429856e-06, + "loss": 0.5056, + "step": 74795 + }, + { + "epoch": 2.88, + "learning_rate": 2.073608617594255e-06, + "loss": 0.2424, + "step": 74796 + }, + { + "epoch": 2.88, + "learning_rate": 2.072967427545525e-06, + "loss": 0.2161, + "step": 74797 + }, + { + "epoch": 2.88, + "learning_rate": 2.0723262374967944e-06, + "loss": 0.1936, + "step": 74798 + }, + { + "epoch": 2.88, + "learning_rate": 2.071685047448064e-06, + "loss": 0.1154, + "step": 74799 + }, + { + "epoch": 2.88, + "learning_rate": 2.0710438573993336e-06, + "loss": 0.1666, + "step": 74800 + }, + { + "epoch": 2.88, + "learning_rate": 2.0704026673506028e-06, + "loss": 0.1477, + "step": 74801 + }, + { + "epoch": 2.88, + "learning_rate": 2.0697614773018724e-06, + "loss": 0.2425, + "step": 74802 + }, + { + "epoch": 2.88, + "learning_rate": 2.069120287253142e-06, + "loss": 0.2303, + "step": 74803 + }, + { + "epoch": 2.88, + "learning_rate": 2.0684790972044116e-06, + "loss": 0.1625, + "step": 74804 + }, + { + "epoch": 2.88, + "learning_rate": 2.067837907155681e-06, + "loss": 0.2398, + "step": 74805 + }, + { + "epoch": 2.88, + "learning_rate": 2.0671967171069508e-06, + "loss": 0.1819, + "step": 74806 + }, + { + "epoch": 2.88, + "learning_rate": 2.0665555270582204e-06, + "loss": 0.1402, + "step": 74807 + }, + { + "epoch": 2.88, + "learning_rate": 2.06591433700949e-06, + "loss": 0.1314, + "step": 74808 + }, + { + "epoch": 2.88, + "learning_rate": 2.0652731469607596e-06, + "loss": 0.0851, + "step": 74809 + }, + { + "epoch": 2.88, + "learning_rate": 2.064631956912029e-06, + "loss": 0.1814, + "step": 74810 + }, + { + "epoch": 2.88, + "learning_rate": 2.0639907668632983e-06, + "loss": 0.16, + "step": 74811 + }, + { + "epoch": 2.88, + "learning_rate": 2.063349576814568e-06, + "loss": 0.2399, + "step": 74812 + }, + { + "epoch": 2.88, + "learning_rate": 2.0627083867658375e-06, + "loss": 0.2066, + "step": 74813 + }, + { + "epoch": 2.88, + "learning_rate": 2.062067196717107e-06, + "loss": 0.1358, + "step": 74814 + }, + { + "epoch": 2.88, + "learning_rate": 2.0614260066683767e-06, + "loss": 0.1601, + "step": 74815 + }, + { + "epoch": 2.88, + "learning_rate": 2.0607848166196463e-06, + "loss": 0.1364, + "step": 74816 + }, + { + "epoch": 2.88, + "learning_rate": 2.060143626570916e-06, + "loss": 0.1971, + "step": 74817 + }, + { + "epoch": 2.88, + "learning_rate": 2.0595024365221855e-06, + "loss": 0.2585, + "step": 74818 + }, + { + "epoch": 2.88, + "learning_rate": 2.058861246473455e-06, + "loss": 0.2071, + "step": 74819 + }, + { + "epoch": 2.88, + "learning_rate": 2.0582200564247247e-06, + "loss": 0.2027, + "step": 74820 + }, + { + "epoch": 2.88, + "learning_rate": 2.057578866375994e-06, + "loss": 0.3465, + "step": 74821 + }, + { + "epoch": 2.88, + "learning_rate": 2.0569376763272635e-06, + "loss": 0.1476, + "step": 74822 + }, + { + "epoch": 2.88, + "learning_rate": 2.056296486278533e-06, + "loss": 0.2488, + "step": 74823 + }, + { + "epoch": 2.88, + "learning_rate": 2.0556552962298027e-06, + "loss": 0.2179, + "step": 74824 + }, + { + "epoch": 2.88, + "learning_rate": 2.0550141061810723e-06, + "loss": 0.2342, + "step": 74825 + }, + { + "epoch": 2.88, + "learning_rate": 2.054372916132342e-06, + "loss": 0.136, + "step": 74826 + }, + { + "epoch": 2.88, + "learning_rate": 2.0537317260836115e-06, + "loss": 0.2487, + "step": 74827 + }, + { + "epoch": 2.88, + "learning_rate": 2.053090536034881e-06, + "loss": 0.209, + "step": 74828 + }, + { + "epoch": 2.88, + "learning_rate": 2.0524493459861507e-06, + "loss": 0.2506, + "step": 74829 + }, + { + "epoch": 2.88, + "learning_rate": 2.05180815593742e-06, + "loss": 0.2854, + "step": 74830 + }, + { + "epoch": 2.88, + "learning_rate": 2.0511669658886895e-06, + "loss": 0.1631, + "step": 74831 + }, + { + "epoch": 2.88, + "learning_rate": 2.050525775839959e-06, + "loss": 0.1386, + "step": 74832 + }, + { + "epoch": 2.88, + "learning_rate": 2.0498845857912287e-06, + "loss": 0.1893, + "step": 74833 + }, + { + "epoch": 2.88, + "learning_rate": 2.0492433957424983e-06, + "loss": 0.142, + "step": 74834 + }, + { + "epoch": 2.88, + "learning_rate": 2.048602205693768e-06, + "loss": 0.163, + "step": 74835 + }, + { + "epoch": 2.88, + "learning_rate": 2.0479610156450375e-06, + "loss": 0.2393, + "step": 74836 + }, + { + "epoch": 2.88, + "learning_rate": 2.047319825596307e-06, + "loss": 0.1875, + "step": 74837 + }, + { + "epoch": 2.88, + "learning_rate": 2.0466786355475767e-06, + "loss": 0.2128, + "step": 74838 + }, + { + "epoch": 2.88, + "learning_rate": 2.0460374454988463e-06, + "loss": 0.2936, + "step": 74839 + }, + { + "epoch": 2.88, + "learning_rate": 2.0453962554501154e-06, + "loss": 0.2084, + "step": 74840 + }, + { + "epoch": 2.88, + "learning_rate": 2.044755065401385e-06, + "loss": 0.2983, + "step": 74841 + }, + { + "epoch": 2.88, + "learning_rate": 2.0441138753526546e-06, + "loss": 0.2659, + "step": 74842 + }, + { + "epoch": 2.88, + "learning_rate": 2.0434726853039242e-06, + "loss": 0.1568, + "step": 74843 + }, + { + "epoch": 2.88, + "learning_rate": 2.042831495255194e-06, + "loss": 0.2493, + "step": 74844 + }, + { + "epoch": 2.88, + "learning_rate": 2.0421903052064634e-06, + "loss": 0.2637, + "step": 74845 + }, + { + "epoch": 2.88, + "learning_rate": 2.041549115157733e-06, + "loss": 0.2545, + "step": 74846 + }, + { + "epoch": 2.88, + "learning_rate": 2.0409079251090026e-06, + "loss": 0.1228, + "step": 74847 + }, + { + "epoch": 2.88, + "learning_rate": 2.0402667350602722e-06, + "loss": 0.2829, + "step": 74848 + }, + { + "epoch": 2.88, + "learning_rate": 2.039625545011542e-06, + "loss": 0.1307, + "step": 74849 + }, + { + "epoch": 2.88, + "learning_rate": 2.038984354962811e-06, + "loss": 0.2824, + "step": 74850 + }, + { + "epoch": 2.88, + "learning_rate": 2.0383431649140806e-06, + "loss": 0.2303, + "step": 74851 + }, + { + "epoch": 2.88, + "learning_rate": 2.03770197486535e-06, + "loss": 0.2262, + "step": 74852 + }, + { + "epoch": 2.88, + "learning_rate": 2.03706078481662e-06, + "loss": 0.2328, + "step": 74853 + }, + { + "epoch": 2.88, + "learning_rate": 2.0364195947678894e-06, + "loss": 0.2474, + "step": 74854 + }, + { + "epoch": 2.88, + "learning_rate": 2.035778404719159e-06, + "loss": 0.2186, + "step": 74855 + }, + { + "epoch": 2.88, + "learning_rate": 2.0351372146704286e-06, + "loss": 0.261, + "step": 74856 + }, + { + "epoch": 2.88, + "learning_rate": 2.034496024621698e-06, + "loss": 0.2899, + "step": 74857 + }, + { + "epoch": 2.88, + "learning_rate": 2.033854834572968e-06, + "loss": 0.1972, + "step": 74858 + }, + { + "epoch": 2.88, + "learning_rate": 2.0332136445242374e-06, + "loss": 0.2197, + "step": 74859 + }, + { + "epoch": 2.88, + "learning_rate": 2.0325724544755066e-06, + "loss": 0.1778, + "step": 74860 + }, + { + "epoch": 2.88, + "learning_rate": 2.031931264426776e-06, + "loss": 0.1526, + "step": 74861 + }, + { + "epoch": 2.88, + "learning_rate": 2.0312900743780458e-06, + "loss": 0.299, + "step": 74862 + }, + { + "epoch": 2.88, + "learning_rate": 2.0306488843293154e-06, + "loss": 0.2308, + "step": 74863 + }, + { + "epoch": 2.88, + "learning_rate": 2.030007694280585e-06, + "loss": 0.1802, + "step": 74864 + }, + { + "epoch": 2.88, + "learning_rate": 2.0293665042318546e-06, + "loss": 0.2706, + "step": 74865 + }, + { + "epoch": 2.88, + "learning_rate": 2.028725314183124e-06, + "loss": 0.2171, + "step": 74866 + }, + { + "epoch": 2.88, + "learning_rate": 2.0280841241343938e-06, + "loss": 0.0723, + "step": 74867 + }, + { + "epoch": 2.88, + "learning_rate": 2.0274429340856634e-06, + "loss": 0.0841, + "step": 74868 + }, + { + "epoch": 2.88, + "learning_rate": 2.026801744036933e-06, + "loss": 0.1456, + "step": 74869 + }, + { + "epoch": 2.88, + "learning_rate": 2.026160553988202e-06, + "loss": 0.1735, + "step": 74870 + }, + { + "epoch": 2.88, + "learning_rate": 2.0255193639394717e-06, + "loss": 0.1816, + "step": 74871 + }, + { + "epoch": 2.88, + "learning_rate": 2.0248781738907413e-06, + "loss": 0.2204, + "step": 74872 + }, + { + "epoch": 2.88, + "learning_rate": 2.024236983842011e-06, + "loss": 0.1512, + "step": 74873 + }, + { + "epoch": 2.88, + "learning_rate": 2.0235957937932805e-06, + "loss": 0.3152, + "step": 74874 + }, + { + "epoch": 2.88, + "learning_rate": 2.02295460374455e-06, + "loss": 0.1193, + "step": 74875 + }, + { + "epoch": 2.88, + "learning_rate": 2.0223134136958197e-06, + "loss": 0.1918, + "step": 74876 + }, + { + "epoch": 2.88, + "learning_rate": 2.0216722236470893e-06, + "loss": 0.1905, + "step": 74877 + }, + { + "epoch": 2.88, + "learning_rate": 2.021031033598359e-06, + "loss": 0.1702, + "step": 74878 + }, + { + "epoch": 2.88, + "learning_rate": 2.0203898435496285e-06, + "loss": 0.2314, + "step": 74879 + }, + { + "epoch": 2.88, + "learning_rate": 2.0197486535008977e-06, + "loss": 0.1682, + "step": 74880 + }, + { + "epoch": 2.88, + "learning_rate": 2.0191074634521673e-06, + "loss": 0.1887, + "step": 74881 + }, + { + "epoch": 2.88, + "learning_rate": 2.018466273403437e-06, + "loss": 0.1209, + "step": 74882 + }, + { + "epoch": 2.88, + "learning_rate": 2.0178250833547065e-06, + "loss": 0.2202, + "step": 74883 + }, + { + "epoch": 2.88, + "learning_rate": 2.017183893305976e-06, + "loss": 0.1587, + "step": 74884 + }, + { + "epoch": 2.88, + "learning_rate": 2.0165427032572457e-06, + "loss": 0.1423, + "step": 74885 + }, + { + "epoch": 2.88, + "learning_rate": 2.0159015132085153e-06, + "loss": 0.1204, + "step": 74886 + }, + { + "epoch": 2.88, + "learning_rate": 2.015260323159785e-06, + "loss": 0.1686, + "step": 74887 + }, + { + "epoch": 2.88, + "learning_rate": 2.0146191331110545e-06, + "loss": 0.2698, + "step": 74888 + }, + { + "epoch": 2.88, + "learning_rate": 2.0139779430623237e-06, + "loss": 0.3041, + "step": 74889 + }, + { + "epoch": 2.88, + "learning_rate": 2.0133367530135933e-06, + "loss": 0.2992, + "step": 74890 + }, + { + "epoch": 2.88, + "learning_rate": 2.012695562964863e-06, + "loss": 0.1014, + "step": 74891 + }, + { + "epoch": 2.88, + "learning_rate": 2.0120543729161325e-06, + "loss": 0.117, + "step": 74892 + }, + { + "epoch": 2.88, + "learning_rate": 2.011413182867402e-06, + "loss": 0.181, + "step": 74893 + }, + { + "epoch": 2.88, + "learning_rate": 2.0107719928186717e-06, + "loss": 0.1452, + "step": 74894 + }, + { + "epoch": 2.88, + "learning_rate": 2.0101308027699413e-06, + "loss": 0.1997, + "step": 74895 + }, + { + "epoch": 2.88, + "learning_rate": 2.009489612721211e-06, + "loss": 0.1568, + "step": 74896 + }, + { + "epoch": 2.88, + "learning_rate": 2.0088484226724805e-06, + "loss": 0.2437, + "step": 74897 + }, + { + "epoch": 2.88, + "learning_rate": 2.00820723262375e-06, + "loss": 0.1547, + "step": 74898 + }, + { + "epoch": 2.88, + "learning_rate": 2.0075660425750192e-06, + "loss": 0.4016, + "step": 74899 + }, + { + "epoch": 2.88, + "learning_rate": 2.006924852526289e-06, + "loss": 0.1549, + "step": 74900 + }, + { + "epoch": 2.88, + "learning_rate": 2.0062836624775584e-06, + "loss": 0.3034, + "step": 74901 + }, + { + "epoch": 2.88, + "learning_rate": 2.005642472428828e-06, + "loss": 0.2725, + "step": 74902 + }, + { + "epoch": 2.88, + "learning_rate": 2.0050012823800976e-06, + "loss": 0.2918, + "step": 74903 + }, + { + "epoch": 2.88, + "learning_rate": 2.0043600923313672e-06, + "loss": 0.2425, + "step": 74904 + }, + { + "epoch": 2.88, + "learning_rate": 2.003718902282637e-06, + "loss": 0.1962, + "step": 74905 + }, + { + "epoch": 2.88, + "learning_rate": 2.0030777122339064e-06, + "loss": 0.1901, + "step": 74906 + }, + { + "epoch": 2.88, + "learning_rate": 2.002436522185176e-06, + "loss": 0.2199, + "step": 74907 + }, + { + "epoch": 2.88, + "learning_rate": 2.0017953321364456e-06, + "loss": 0.1311, + "step": 74908 + }, + { + "epoch": 2.88, + "learning_rate": 2.001154142087715e-06, + "loss": 0.1743, + "step": 74909 + }, + { + "epoch": 2.88, + "learning_rate": 2.0005129520389844e-06, + "loss": 0.2471, + "step": 74910 + }, + { + "epoch": 2.88, + "learning_rate": 1.999871761990254e-06, + "loss": 0.1447, + "step": 74911 + }, + { + "epoch": 2.88, + "learning_rate": 1.9992305719415236e-06, + "loss": 0.1209, + "step": 74912 + }, + { + "epoch": 2.88, + "learning_rate": 1.998589381892793e-06, + "loss": 0.2381, + "step": 74913 + }, + { + "epoch": 2.88, + "learning_rate": 1.997948191844063e-06, + "loss": 0.1434, + "step": 74914 + }, + { + "epoch": 2.88, + "learning_rate": 1.9973070017953324e-06, + "loss": 0.1749, + "step": 74915 + }, + { + "epoch": 2.88, + "learning_rate": 1.996665811746602e-06, + "loss": 0.1878, + "step": 74916 + }, + { + "epoch": 2.88, + "learning_rate": 1.9960246216978716e-06, + "loss": 0.2105, + "step": 74917 + }, + { + "epoch": 2.88, + "learning_rate": 1.995383431649141e-06, + "loss": 0.1715, + "step": 74918 + }, + { + "epoch": 2.88, + "learning_rate": 1.9947422416004104e-06, + "loss": 0.1688, + "step": 74919 + }, + { + "epoch": 2.88, + "learning_rate": 1.99410105155168e-06, + "loss": 0.1436, + "step": 74920 + }, + { + "epoch": 2.88, + "learning_rate": 1.9934598615029496e-06, + "loss": 0.1855, + "step": 74921 + }, + { + "epoch": 2.88, + "learning_rate": 1.992818671454219e-06, + "loss": 0.181, + "step": 74922 + }, + { + "epoch": 2.88, + "learning_rate": 1.9921774814054888e-06, + "loss": 0.2777, + "step": 74923 + }, + { + "epoch": 2.88, + "learning_rate": 1.9915362913567584e-06, + "loss": 0.1111, + "step": 74924 + }, + { + "epoch": 2.88, + "learning_rate": 1.990895101308028e-06, + "loss": 0.2697, + "step": 74925 + }, + { + "epoch": 2.88, + "learning_rate": 1.9902539112592976e-06, + "loss": 0.2129, + "step": 74926 + }, + { + "epoch": 2.88, + "learning_rate": 1.989612721210567e-06, + "loss": 0.2504, + "step": 74927 + }, + { + "epoch": 2.88, + "learning_rate": 1.9889715311618368e-06, + "loss": 0.1206, + "step": 74928 + }, + { + "epoch": 2.88, + "learning_rate": 1.988330341113106e-06, + "loss": 0.1206, + "step": 74929 + }, + { + "epoch": 2.88, + "learning_rate": 1.9876891510643755e-06, + "loss": 0.1033, + "step": 74930 + }, + { + "epoch": 2.88, + "learning_rate": 1.987047961015645e-06, + "loss": 0.2161, + "step": 74931 + }, + { + "epoch": 2.88, + "learning_rate": 1.9864067709669147e-06, + "loss": 0.2121, + "step": 74932 + }, + { + "epoch": 2.88, + "learning_rate": 1.9857655809181843e-06, + "loss": 0.3272, + "step": 74933 + }, + { + "epoch": 2.88, + "learning_rate": 1.985124390869454e-06, + "loss": 0.2439, + "step": 74934 + }, + { + "epoch": 2.88, + "learning_rate": 1.9844832008207235e-06, + "loss": 0.0969, + "step": 74935 + }, + { + "epoch": 2.88, + "learning_rate": 1.983842010771993e-06, + "loss": 0.3689, + "step": 74936 + }, + { + "epoch": 2.88, + "learning_rate": 1.9832008207232627e-06, + "loss": 0.2733, + "step": 74937 + }, + { + "epoch": 2.88, + "learning_rate": 1.9825596306745323e-06, + "loss": 0.3427, + "step": 74938 + }, + { + "epoch": 2.88, + "learning_rate": 1.9819184406258015e-06, + "loss": 0.1857, + "step": 74939 + }, + { + "epoch": 2.88, + "learning_rate": 1.981277250577071e-06, + "loss": 0.1101, + "step": 74940 + }, + { + "epoch": 2.88, + "learning_rate": 1.9806360605283407e-06, + "loss": 0.1127, + "step": 74941 + }, + { + "epoch": 2.88, + "learning_rate": 1.9799948704796103e-06, + "loss": 0.2928, + "step": 74942 + }, + { + "epoch": 2.88, + "learning_rate": 1.97935368043088e-06, + "loss": 0.3947, + "step": 74943 + }, + { + "epoch": 2.88, + "learning_rate": 1.9787124903821495e-06, + "loss": 0.3435, + "step": 74944 + }, + { + "epoch": 2.88, + "learning_rate": 1.978071300333419e-06, + "loss": 0.2963, + "step": 74945 + }, + { + "epoch": 2.88, + "learning_rate": 1.9774301102846887e-06, + "loss": 0.2979, + "step": 74946 + }, + { + "epoch": 2.88, + "learning_rate": 1.9767889202359583e-06, + "loss": 0.2756, + "step": 74947 + }, + { + "epoch": 2.88, + "learning_rate": 1.9761477301872275e-06, + "loss": 0.218, + "step": 74948 + }, + { + "epoch": 2.88, + "learning_rate": 1.975506540138497e-06, + "loss": 0.2338, + "step": 74949 + }, + { + "epoch": 2.88, + "learning_rate": 1.9748653500897667e-06, + "loss": 0.2286, + "step": 74950 + }, + { + "epoch": 2.88, + "learning_rate": 1.9742241600410363e-06, + "loss": 0.1765, + "step": 74951 + }, + { + "epoch": 2.88, + "learning_rate": 1.973582969992306e-06, + "loss": 0.3229, + "step": 74952 + }, + { + "epoch": 2.88, + "learning_rate": 1.9729417799435755e-06, + "loss": 0.1373, + "step": 74953 + }, + { + "epoch": 2.88, + "learning_rate": 1.972300589894845e-06, + "loss": 0.1446, + "step": 74954 + }, + { + "epoch": 2.88, + "learning_rate": 1.9716593998461147e-06, + "loss": 0.2431, + "step": 74955 + }, + { + "epoch": 2.88, + "learning_rate": 1.9710182097973843e-06, + "loss": 0.2563, + "step": 74956 + }, + { + "epoch": 2.88, + "learning_rate": 1.970377019748654e-06, + "loss": 0.1436, + "step": 74957 + }, + { + "epoch": 2.88, + "learning_rate": 1.969735829699923e-06, + "loss": 0.2378, + "step": 74958 + }, + { + "epoch": 2.88, + "learning_rate": 1.9690946396511926e-06, + "loss": 0.1578, + "step": 74959 + }, + { + "epoch": 2.88, + "learning_rate": 1.9684534496024622e-06, + "loss": 0.1729, + "step": 74960 + }, + { + "epoch": 2.88, + "learning_rate": 1.967812259553732e-06, + "loss": 0.1068, + "step": 74961 + }, + { + "epoch": 2.88, + "learning_rate": 1.9671710695050014e-06, + "loss": 0.1756, + "step": 74962 + }, + { + "epoch": 2.88, + "learning_rate": 1.966529879456271e-06, + "loss": 0.2398, + "step": 74963 + }, + { + "epoch": 2.88, + "learning_rate": 1.9658886894075406e-06, + "loss": 0.2329, + "step": 74964 + }, + { + "epoch": 2.88, + "learning_rate": 1.9652474993588102e-06, + "loss": 0.2572, + "step": 74965 + }, + { + "epoch": 2.88, + "learning_rate": 1.96460630931008e-06, + "loss": 0.1263, + "step": 74966 + }, + { + "epoch": 2.88, + "learning_rate": 1.9639651192613494e-06, + "loss": 0.1972, + "step": 74967 + }, + { + "epoch": 2.88, + "learning_rate": 1.9633239292126186e-06, + "loss": 0.1555, + "step": 74968 + }, + { + "epoch": 2.88, + "learning_rate": 1.962682739163888e-06, + "loss": 0.0995, + "step": 74969 + }, + { + "epoch": 2.88, + "learning_rate": 1.962041549115158e-06, + "loss": 0.1773, + "step": 74970 + }, + { + "epoch": 2.88, + "learning_rate": 1.9614003590664274e-06, + "loss": 0.2591, + "step": 74971 + }, + { + "epoch": 2.88, + "learning_rate": 1.960759169017697e-06, + "loss": 0.1943, + "step": 74972 + }, + { + "epoch": 2.88, + "learning_rate": 1.9601179789689666e-06, + "loss": 0.1101, + "step": 74973 + }, + { + "epoch": 2.88, + "learning_rate": 1.959476788920236e-06, + "loss": 0.186, + "step": 74974 + }, + { + "epoch": 2.88, + "learning_rate": 1.958835598871506e-06, + "loss": 0.2903, + "step": 74975 + }, + { + "epoch": 2.88, + "learning_rate": 1.9581944088227754e-06, + "loss": 0.2121, + "step": 74976 + }, + { + "epoch": 2.88, + "learning_rate": 1.957553218774045e-06, + "loss": 0.1927, + "step": 74977 + }, + { + "epoch": 2.88, + "learning_rate": 1.956912028725314e-06, + "loss": 0.2031, + "step": 74978 + }, + { + "epoch": 2.88, + "learning_rate": 1.9562708386765838e-06, + "loss": 0.225, + "step": 74979 + }, + { + "epoch": 2.88, + "learning_rate": 1.9556296486278534e-06, + "loss": 0.2317, + "step": 74980 + }, + { + "epoch": 2.88, + "learning_rate": 1.954988458579123e-06, + "loss": 0.2479, + "step": 74981 + }, + { + "epoch": 2.88, + "learning_rate": 1.9543472685303926e-06, + "loss": 0.2052, + "step": 74982 + }, + { + "epoch": 2.88, + "learning_rate": 1.953706078481662e-06, + "loss": 0.122, + "step": 74983 + }, + { + "epoch": 2.88, + "learning_rate": 1.9530648884329318e-06, + "loss": 0.1993, + "step": 74984 + }, + { + "epoch": 2.88, + "learning_rate": 1.9524236983842014e-06, + "loss": 0.1302, + "step": 74985 + }, + { + "epoch": 2.88, + "learning_rate": 1.951782508335471e-06, + "loss": 0.2633, + "step": 74986 + }, + { + "epoch": 2.88, + "learning_rate": 1.9511413182867406e-06, + "loss": 0.1725, + "step": 74987 + }, + { + "epoch": 2.88, + "learning_rate": 1.9505001282380097e-06, + "loss": 0.0893, + "step": 74988 + }, + { + "epoch": 2.88, + "learning_rate": 1.9498589381892793e-06, + "loss": 0.2171, + "step": 74989 + }, + { + "epoch": 2.88, + "learning_rate": 1.949217748140549e-06, + "loss": 0.1766, + "step": 74990 + }, + { + "epoch": 2.88, + "learning_rate": 1.9485765580918185e-06, + "loss": 0.1315, + "step": 74991 + }, + { + "epoch": 2.88, + "learning_rate": 1.947935368043088e-06, + "loss": 0.1912, + "step": 74992 + }, + { + "epoch": 2.88, + "learning_rate": 1.9472941779943577e-06, + "loss": 0.188, + "step": 74993 + }, + { + "epoch": 2.88, + "learning_rate": 1.9466529879456273e-06, + "loss": 0.1541, + "step": 74994 + }, + { + "epoch": 2.88, + "learning_rate": 1.946011797896897e-06, + "loss": 0.152, + "step": 74995 + }, + { + "epoch": 2.88, + "learning_rate": 1.9453706078481665e-06, + "loss": 0.1991, + "step": 74996 + }, + { + "epoch": 2.88, + "learning_rate": 1.944729417799436e-06, + "loss": 0.1867, + "step": 74997 + }, + { + "epoch": 2.88, + "learning_rate": 1.9440882277507053e-06, + "loss": 0.0954, + "step": 74998 + }, + { + "epoch": 2.88, + "learning_rate": 1.943447037701975e-06, + "loss": 0.2114, + "step": 74999 + }, + { + "epoch": 2.88, + "learning_rate": 1.9428058476532445e-06, + "loss": 0.221, + "step": 75000 + }, + { + "epoch": 2.88, + "learning_rate": 1.942164657604514e-06, + "loss": 0.3358, + "step": 75001 + }, + { + "epoch": 2.88, + "learning_rate": 1.9415234675557837e-06, + "loss": 0.1944, + "step": 75002 + }, + { + "epoch": 2.88, + "learning_rate": 1.9408822775070533e-06, + "loss": 0.2207, + "step": 75003 + }, + { + "epoch": 2.88, + "learning_rate": 1.940241087458323e-06, + "loss": 0.2118, + "step": 75004 + }, + { + "epoch": 2.88, + "learning_rate": 1.9395998974095925e-06, + "loss": 0.2141, + "step": 75005 + }, + { + "epoch": 2.88, + "learning_rate": 1.938958707360862e-06, + "loss": 0.3135, + "step": 75006 + }, + { + "epoch": 2.88, + "learning_rate": 1.9383175173121313e-06, + "loss": 0.1487, + "step": 75007 + }, + { + "epoch": 2.88, + "learning_rate": 1.937676327263401e-06, + "loss": 0.1912, + "step": 75008 + }, + { + "epoch": 2.88, + "learning_rate": 1.9370351372146705e-06, + "loss": 0.1614, + "step": 75009 + }, + { + "epoch": 2.88, + "learning_rate": 1.93639394716594e-06, + "loss": 0.2293, + "step": 75010 + }, + { + "epoch": 2.88, + "learning_rate": 1.9357527571172097e-06, + "loss": 0.1799, + "step": 75011 + }, + { + "epoch": 2.88, + "learning_rate": 1.9351115670684793e-06, + "loss": 0.1855, + "step": 75012 + }, + { + "epoch": 2.88, + "learning_rate": 1.934470377019749e-06, + "loss": 0.1532, + "step": 75013 + }, + { + "epoch": 2.88, + "learning_rate": 1.9338291869710185e-06, + "loss": 0.1897, + "step": 75014 + }, + { + "epoch": 2.88, + "learning_rate": 1.933187996922288e-06, + "loss": 0.1714, + "step": 75015 + }, + { + "epoch": 2.88, + "learning_rate": 1.9325468068735577e-06, + "loss": 0.3529, + "step": 75016 + }, + { + "epoch": 2.88, + "learning_rate": 1.931905616824827e-06, + "loss": 0.2714, + "step": 75017 + }, + { + "epoch": 2.88, + "learning_rate": 1.9312644267760964e-06, + "loss": 0.1268, + "step": 75018 + }, + { + "epoch": 2.88, + "learning_rate": 1.930623236727366e-06, + "loss": 0.1878, + "step": 75019 + }, + { + "epoch": 2.88, + "learning_rate": 1.9299820466786356e-06, + "loss": 0.2462, + "step": 75020 + }, + { + "epoch": 2.88, + "learning_rate": 1.9293408566299052e-06, + "loss": 0.2682, + "step": 75021 + }, + { + "epoch": 2.88, + "learning_rate": 1.928699666581175e-06, + "loss": 0.1935, + "step": 75022 + }, + { + "epoch": 2.88, + "learning_rate": 1.9280584765324444e-06, + "loss": 0.2228, + "step": 75023 + }, + { + "epoch": 2.88, + "learning_rate": 1.927417286483714e-06, + "loss": 0.2029, + "step": 75024 + }, + { + "epoch": 2.88, + "learning_rate": 1.9267760964349836e-06, + "loss": 0.1621, + "step": 75025 + }, + { + "epoch": 2.88, + "learning_rate": 1.9261349063862532e-06, + "loss": 0.2543, + "step": 75026 + }, + { + "epoch": 2.88, + "learning_rate": 1.9254937163375224e-06, + "loss": 0.2889, + "step": 75027 + }, + { + "epoch": 2.88, + "learning_rate": 1.924852526288792e-06, + "loss": 0.28, + "step": 75028 + }, + { + "epoch": 2.88, + "learning_rate": 1.9242113362400616e-06, + "loss": 0.2196, + "step": 75029 + }, + { + "epoch": 2.88, + "learning_rate": 1.923570146191331e-06, + "loss": 0.1815, + "step": 75030 + }, + { + "epoch": 2.88, + "learning_rate": 1.922928956142601e-06, + "loss": 0.3517, + "step": 75031 + }, + { + "epoch": 2.88, + "learning_rate": 1.9222877660938704e-06, + "loss": 0.2052, + "step": 75032 + }, + { + "epoch": 2.88, + "learning_rate": 1.92164657604514e-06, + "loss": 0.1605, + "step": 75033 + }, + { + "epoch": 2.88, + "learning_rate": 1.9210053859964096e-06, + "loss": 0.3336, + "step": 75034 + }, + { + "epoch": 2.88, + "learning_rate": 1.920364195947679e-06, + "loss": 0.3524, + "step": 75035 + }, + { + "epoch": 2.88, + "learning_rate": 1.919723005898949e-06, + "loss": 0.1239, + "step": 75036 + }, + { + "epoch": 2.88, + "learning_rate": 1.919081815850218e-06, + "loss": 0.1357, + "step": 75037 + }, + { + "epoch": 2.88, + "learning_rate": 1.9184406258014876e-06, + "loss": 0.1686, + "step": 75038 + }, + { + "epoch": 2.89, + "learning_rate": 1.917799435752757e-06, + "loss": 0.2444, + "step": 75039 + }, + { + "epoch": 2.89, + "learning_rate": 1.9171582457040268e-06, + "loss": 0.2636, + "step": 75040 + }, + { + "epoch": 2.89, + "learning_rate": 1.9165170556552964e-06, + "loss": 0.4579, + "step": 75041 + }, + { + "epoch": 2.89, + "learning_rate": 1.915875865606566e-06, + "loss": 0.1663, + "step": 75042 + }, + { + "epoch": 2.89, + "learning_rate": 1.9152346755578356e-06, + "loss": 0.195, + "step": 75043 + }, + { + "epoch": 2.89, + "learning_rate": 1.914593485509105e-06, + "loss": 0.1456, + "step": 75044 + }, + { + "epoch": 2.89, + "learning_rate": 1.9139522954603748e-06, + "loss": 0.0939, + "step": 75045 + }, + { + "epoch": 2.89, + "learning_rate": 1.9133111054116444e-06, + "loss": 0.2239, + "step": 75046 + }, + { + "epoch": 2.89, + "learning_rate": 1.9126699153629135e-06, + "loss": 0.0771, + "step": 75047 + }, + { + "epoch": 2.89, + "learning_rate": 1.912028725314183e-06, + "loss": 0.1555, + "step": 75048 + }, + { + "epoch": 2.89, + "learning_rate": 1.9113875352654527e-06, + "loss": 0.1676, + "step": 75049 + }, + { + "epoch": 2.89, + "learning_rate": 1.9107463452167223e-06, + "loss": 0.2428, + "step": 75050 + }, + { + "epoch": 2.89, + "learning_rate": 1.910105155167992e-06, + "loss": 0.2375, + "step": 75051 + }, + { + "epoch": 2.89, + "learning_rate": 1.9094639651192615e-06, + "loss": 0.1902, + "step": 75052 + }, + { + "epoch": 2.89, + "learning_rate": 1.908822775070531e-06, + "loss": 0.2804, + "step": 75053 + }, + { + "epoch": 2.89, + "learning_rate": 1.9081815850218007e-06, + "loss": 0.1951, + "step": 75054 + }, + { + "epoch": 2.89, + "learning_rate": 1.9075403949730703e-06, + "loss": 0.1425, + "step": 75055 + }, + { + "epoch": 2.89, + "learning_rate": 1.90689920492434e-06, + "loss": 0.283, + "step": 75056 + }, + { + "epoch": 2.89, + "learning_rate": 1.9062580148756091e-06, + "loss": 0.1721, + "step": 75057 + }, + { + "epoch": 2.89, + "learning_rate": 1.9056168248268787e-06, + "loss": 0.1692, + "step": 75058 + }, + { + "epoch": 2.89, + "learning_rate": 1.9049756347781483e-06, + "loss": 0.1291, + "step": 75059 + }, + { + "epoch": 2.89, + "learning_rate": 1.904334444729418e-06, + "loss": 0.1729, + "step": 75060 + }, + { + "epoch": 2.89, + "learning_rate": 1.9036932546806875e-06, + "loss": 0.1601, + "step": 75061 + }, + { + "epoch": 2.89, + "learning_rate": 1.903052064631957e-06, + "loss": 0.1851, + "step": 75062 + }, + { + "epoch": 2.89, + "learning_rate": 1.9024108745832267e-06, + "loss": 0.2356, + "step": 75063 + }, + { + "epoch": 2.89, + "learning_rate": 1.9017696845344963e-06, + "loss": 0.3263, + "step": 75064 + }, + { + "epoch": 2.89, + "learning_rate": 1.901128494485766e-06, + "loss": 0.1798, + "step": 75065 + }, + { + "epoch": 2.89, + "learning_rate": 1.900487304437035e-06, + "loss": 0.2381, + "step": 75066 + }, + { + "epoch": 2.89, + "learning_rate": 1.8998461143883047e-06, + "loss": 0.1851, + "step": 75067 + }, + { + "epoch": 2.89, + "learning_rate": 1.8992049243395743e-06, + "loss": 0.1668, + "step": 75068 + }, + { + "epoch": 2.89, + "learning_rate": 1.8985637342908439e-06, + "loss": 0.1772, + "step": 75069 + }, + { + "epoch": 2.89, + "learning_rate": 1.8979225442421135e-06, + "loss": 0.2707, + "step": 75070 + }, + { + "epoch": 2.89, + "learning_rate": 1.897281354193383e-06, + "loss": 0.186, + "step": 75071 + }, + { + "epoch": 2.89, + "learning_rate": 1.8966401641446527e-06, + "loss": 0.1576, + "step": 75072 + }, + { + "epoch": 2.89, + "learning_rate": 1.8959989740959223e-06, + "loss": 0.1863, + "step": 75073 + }, + { + "epoch": 2.89, + "learning_rate": 1.8953577840471919e-06, + "loss": 0.2679, + "step": 75074 + }, + { + "epoch": 2.89, + "learning_rate": 1.8947165939984615e-06, + "loss": 0.1682, + "step": 75075 + }, + { + "epoch": 2.89, + "learning_rate": 1.8940754039497306e-06, + "loss": 0.1807, + "step": 75076 + }, + { + "epoch": 2.89, + "learning_rate": 1.8934342139010002e-06, + "loss": 0.1721, + "step": 75077 + }, + { + "epoch": 2.89, + "learning_rate": 1.8927930238522698e-06, + "loss": 0.2014, + "step": 75078 + }, + { + "epoch": 2.89, + "learning_rate": 1.8921518338035394e-06, + "loss": 0.1614, + "step": 75079 + }, + { + "epoch": 2.89, + "learning_rate": 1.891510643754809e-06, + "loss": 0.1346, + "step": 75080 + }, + { + "epoch": 2.89, + "learning_rate": 1.8908694537060786e-06, + "loss": 0.2478, + "step": 75081 + }, + { + "epoch": 2.89, + "learning_rate": 1.8902282636573482e-06, + "loss": 0.2363, + "step": 75082 + }, + { + "epoch": 2.89, + "learning_rate": 1.8895870736086178e-06, + "loss": 0.1544, + "step": 75083 + }, + { + "epoch": 2.89, + "learning_rate": 1.8889458835598874e-06, + "loss": 0.214, + "step": 75084 + }, + { + "epoch": 2.89, + "learning_rate": 1.888304693511157e-06, + "loss": 0.3427, + "step": 75085 + }, + { + "epoch": 2.89, + "learning_rate": 1.8876635034624262e-06, + "loss": 0.3714, + "step": 75086 + }, + { + "epoch": 2.89, + "learning_rate": 1.8870223134136958e-06, + "loss": 0.1421, + "step": 75087 + }, + { + "epoch": 2.89, + "learning_rate": 1.8863811233649654e-06, + "loss": 0.1587, + "step": 75088 + }, + { + "epoch": 2.89, + "learning_rate": 1.885739933316235e-06, + "loss": 0.2503, + "step": 75089 + }, + { + "epoch": 2.89, + "learning_rate": 1.8850987432675046e-06, + "loss": 0.3768, + "step": 75090 + }, + { + "epoch": 2.89, + "learning_rate": 1.8844575532187742e-06, + "loss": 0.3388, + "step": 75091 + }, + { + "epoch": 2.89, + "learning_rate": 1.8838163631700438e-06, + "loss": 0.0932, + "step": 75092 + }, + { + "epoch": 2.89, + "learning_rate": 1.8831751731213134e-06, + "loss": 0.2509, + "step": 75093 + }, + { + "epoch": 2.89, + "learning_rate": 1.882533983072583e-06, + "loss": 0.0932, + "step": 75094 + }, + { + "epoch": 2.89, + "learning_rate": 1.8818927930238526e-06, + "loss": 0.1259, + "step": 75095 + }, + { + "epoch": 2.89, + "learning_rate": 1.8812516029751218e-06, + "loss": 0.1999, + "step": 75096 + }, + { + "epoch": 2.89, + "learning_rate": 1.8806104129263914e-06, + "loss": 0.1373, + "step": 75097 + }, + { + "epoch": 2.89, + "learning_rate": 1.879969222877661e-06, + "loss": 0.123, + "step": 75098 + }, + { + "epoch": 2.89, + "learning_rate": 1.8793280328289306e-06, + "loss": 0.1401, + "step": 75099 + }, + { + "epoch": 2.89, + "learning_rate": 1.8786868427802002e-06, + "loss": 0.1367, + "step": 75100 + }, + { + "epoch": 2.89, + "learning_rate": 1.8780456527314698e-06, + "loss": 0.1966, + "step": 75101 + }, + { + "epoch": 2.89, + "learning_rate": 1.8774044626827394e-06, + "loss": 0.2212, + "step": 75102 + }, + { + "epoch": 2.89, + "learning_rate": 1.876763272634009e-06, + "loss": 0.2194, + "step": 75103 + }, + { + "epoch": 2.89, + "learning_rate": 1.8761220825852786e-06, + "loss": 0.225, + "step": 75104 + }, + { + "epoch": 2.89, + "learning_rate": 1.8754808925365482e-06, + "loss": 0.1729, + "step": 75105 + }, + { + "epoch": 2.89, + "learning_rate": 1.8748397024878173e-06, + "loss": 0.1977, + "step": 75106 + }, + { + "epoch": 2.89, + "learning_rate": 1.874198512439087e-06, + "loss": 0.1565, + "step": 75107 + }, + { + "epoch": 2.89, + "learning_rate": 1.8735573223903565e-06, + "loss": 0.2181, + "step": 75108 + }, + { + "epoch": 2.89, + "learning_rate": 1.8729161323416261e-06, + "loss": 0.2058, + "step": 75109 + }, + { + "epoch": 2.89, + "learning_rate": 1.8722749422928957e-06, + "loss": 0.2022, + "step": 75110 + }, + { + "epoch": 2.89, + "learning_rate": 1.8716337522441653e-06, + "loss": 0.1721, + "step": 75111 + }, + { + "epoch": 2.89, + "learning_rate": 1.870992562195435e-06, + "loss": 0.2509, + "step": 75112 + }, + { + "epoch": 2.89, + "learning_rate": 1.8703513721467045e-06, + "loss": 0.234, + "step": 75113 + }, + { + "epoch": 2.89, + "learning_rate": 1.8697101820979741e-06, + "loss": 0.2812, + "step": 75114 + }, + { + "epoch": 2.89, + "learning_rate": 1.8690689920492437e-06, + "loss": 0.2132, + "step": 75115 + }, + { + "epoch": 2.89, + "learning_rate": 1.868427802000513e-06, + "loss": 0.1661, + "step": 75116 + }, + { + "epoch": 2.89, + "learning_rate": 1.8677866119517825e-06, + "loss": 0.1111, + "step": 75117 + }, + { + "epoch": 2.89, + "learning_rate": 1.8671454219030521e-06, + "loss": 0.3079, + "step": 75118 + }, + { + "epoch": 2.89, + "learning_rate": 1.8665042318543217e-06, + "loss": 0.2119, + "step": 75119 + }, + { + "epoch": 2.89, + "learning_rate": 1.8658630418055913e-06, + "loss": 0.0733, + "step": 75120 + }, + { + "epoch": 2.89, + "learning_rate": 1.865221851756861e-06, + "loss": 0.2949, + "step": 75121 + }, + { + "epoch": 2.89, + "learning_rate": 1.8645806617081305e-06, + "loss": 0.2591, + "step": 75122 + }, + { + "epoch": 2.89, + "learning_rate": 1.8639394716594e-06, + "loss": 0.2212, + "step": 75123 + }, + { + "epoch": 2.89, + "learning_rate": 1.8632982816106697e-06, + "loss": 0.2005, + "step": 75124 + }, + { + "epoch": 2.89, + "learning_rate": 1.8626570915619389e-06, + "loss": 0.1721, + "step": 75125 + }, + { + "epoch": 2.89, + "learning_rate": 1.8620159015132085e-06, + "loss": 0.2012, + "step": 75126 + }, + { + "epoch": 2.89, + "learning_rate": 1.861374711464478e-06, + "loss": 0.2437, + "step": 75127 + }, + { + "epoch": 2.89, + "learning_rate": 1.8607335214157477e-06, + "loss": 0.277, + "step": 75128 + }, + { + "epoch": 2.89, + "learning_rate": 1.8600923313670173e-06, + "loss": 0.1479, + "step": 75129 + }, + { + "epoch": 2.89, + "learning_rate": 1.8594511413182869e-06, + "loss": 0.1729, + "step": 75130 + }, + { + "epoch": 2.89, + "learning_rate": 1.8588099512695565e-06, + "loss": 0.1405, + "step": 75131 + }, + { + "epoch": 2.89, + "learning_rate": 1.858168761220826e-06, + "loss": 0.2019, + "step": 75132 + }, + { + "epoch": 2.89, + "learning_rate": 1.8575275711720957e-06, + "loss": 0.1227, + "step": 75133 + }, + { + "epoch": 2.89, + "learning_rate": 1.8568863811233653e-06, + "loss": 0.1589, + "step": 75134 + }, + { + "epoch": 2.89, + "learning_rate": 1.8562451910746344e-06, + "loss": 0.1819, + "step": 75135 + }, + { + "epoch": 2.89, + "learning_rate": 1.855604001025904e-06, + "loss": 0.1669, + "step": 75136 + }, + { + "epoch": 2.89, + "learning_rate": 1.8549628109771736e-06, + "loss": 0.1297, + "step": 75137 + }, + { + "epoch": 2.89, + "learning_rate": 1.8543216209284432e-06, + "loss": 0.3255, + "step": 75138 + }, + { + "epoch": 2.89, + "learning_rate": 1.8536804308797128e-06, + "loss": 0.2079, + "step": 75139 + }, + { + "epoch": 2.89, + "learning_rate": 1.8530392408309824e-06, + "loss": 0.1658, + "step": 75140 + }, + { + "epoch": 2.89, + "learning_rate": 1.852398050782252e-06, + "loss": 0.3426, + "step": 75141 + }, + { + "epoch": 2.89, + "learning_rate": 1.8517568607335216e-06, + "loss": 0.1602, + "step": 75142 + }, + { + "epoch": 2.89, + "learning_rate": 1.8511156706847912e-06, + "loss": 0.1867, + "step": 75143 + }, + { + "epoch": 2.89, + "learning_rate": 1.8504744806360608e-06, + "loss": 0.2637, + "step": 75144 + }, + { + "epoch": 2.89, + "learning_rate": 1.84983329058733e-06, + "loss": 0.154, + "step": 75145 + }, + { + "epoch": 2.89, + "learning_rate": 1.8491921005385996e-06, + "loss": 0.2117, + "step": 75146 + }, + { + "epoch": 2.89, + "learning_rate": 1.8485509104898692e-06, + "loss": 0.2444, + "step": 75147 + }, + { + "epoch": 2.89, + "learning_rate": 1.8479097204411388e-06, + "loss": 0.2683, + "step": 75148 + }, + { + "epoch": 2.89, + "learning_rate": 1.8472685303924084e-06, + "loss": 0.1322, + "step": 75149 + }, + { + "epoch": 2.89, + "learning_rate": 1.846627340343678e-06, + "loss": 0.2012, + "step": 75150 + }, + { + "epoch": 2.89, + "learning_rate": 1.8459861502949476e-06, + "loss": 0.2592, + "step": 75151 + }, + { + "epoch": 2.89, + "learning_rate": 1.8453449602462172e-06, + "loss": 0.2056, + "step": 75152 + }, + { + "epoch": 2.89, + "learning_rate": 1.8447037701974868e-06, + "loss": 0.2452, + "step": 75153 + }, + { + "epoch": 2.89, + "learning_rate": 1.8440625801487564e-06, + "loss": 0.1046, + "step": 75154 + }, + { + "epoch": 2.89, + "learning_rate": 1.8434213901000256e-06, + "loss": 0.2208, + "step": 75155 + }, + { + "epoch": 2.89, + "learning_rate": 1.8427802000512952e-06, + "loss": 0.2242, + "step": 75156 + }, + { + "epoch": 2.89, + "learning_rate": 1.8421390100025648e-06, + "loss": 0.1319, + "step": 75157 + }, + { + "epoch": 2.89, + "learning_rate": 1.8414978199538344e-06, + "loss": 0.3432, + "step": 75158 + }, + { + "epoch": 2.89, + "learning_rate": 1.840856629905104e-06, + "loss": 0.2047, + "step": 75159 + }, + { + "epoch": 2.89, + "learning_rate": 1.8402154398563736e-06, + "loss": 0.2581, + "step": 75160 + }, + { + "epoch": 2.89, + "learning_rate": 1.8395742498076432e-06, + "loss": 0.286, + "step": 75161 + }, + { + "epoch": 2.89, + "learning_rate": 1.8389330597589128e-06, + "loss": 0.1693, + "step": 75162 + }, + { + "epoch": 2.89, + "learning_rate": 1.8382918697101824e-06, + "loss": 0.1216, + "step": 75163 + }, + { + "epoch": 2.89, + "learning_rate": 1.837650679661452e-06, + "loss": 0.3084, + "step": 75164 + }, + { + "epoch": 2.89, + "learning_rate": 1.8370094896127211e-06, + "loss": 0.2497, + "step": 75165 + }, + { + "epoch": 2.89, + "learning_rate": 1.8363682995639907e-06, + "loss": 0.0941, + "step": 75166 + }, + { + "epoch": 2.89, + "learning_rate": 1.8357271095152603e-06, + "loss": 0.1626, + "step": 75167 + }, + { + "epoch": 2.89, + "learning_rate": 1.83508591946653e-06, + "loss": 0.1675, + "step": 75168 + }, + { + "epoch": 2.89, + "learning_rate": 1.8344447294177995e-06, + "loss": 0.2284, + "step": 75169 + }, + { + "epoch": 2.89, + "learning_rate": 1.8338035393690691e-06, + "loss": 0.194, + "step": 75170 + }, + { + "epoch": 2.89, + "learning_rate": 1.8331623493203387e-06, + "loss": 0.2004, + "step": 75171 + }, + { + "epoch": 2.89, + "learning_rate": 1.8325211592716083e-06, + "loss": 0.3872, + "step": 75172 + }, + { + "epoch": 2.89, + "learning_rate": 1.831879969222878e-06, + "loss": 0.1792, + "step": 75173 + }, + { + "epoch": 2.89, + "learning_rate": 1.8312387791741475e-06, + "loss": 0.1342, + "step": 75174 + }, + { + "epoch": 2.89, + "learning_rate": 1.8305975891254167e-06, + "loss": 0.1972, + "step": 75175 + }, + { + "epoch": 2.89, + "learning_rate": 1.8299563990766863e-06, + "loss": 0.1348, + "step": 75176 + }, + { + "epoch": 2.89, + "learning_rate": 1.829315209027956e-06, + "loss": 0.1652, + "step": 75177 + }, + { + "epoch": 2.89, + "learning_rate": 1.8286740189792255e-06, + "loss": 0.1451, + "step": 75178 + }, + { + "epoch": 2.89, + "learning_rate": 1.8280328289304951e-06, + "loss": 0.1183, + "step": 75179 + }, + { + "epoch": 2.89, + "learning_rate": 1.8273916388817647e-06, + "loss": 0.2827, + "step": 75180 + }, + { + "epoch": 2.89, + "learning_rate": 1.8267504488330343e-06, + "loss": 0.1439, + "step": 75181 + }, + { + "epoch": 2.89, + "learning_rate": 1.826109258784304e-06, + "loss": 0.23, + "step": 75182 + }, + { + "epoch": 2.89, + "learning_rate": 1.8254680687355735e-06, + "loss": 0.1749, + "step": 75183 + }, + { + "epoch": 2.89, + "learning_rate": 1.8248268786868427e-06, + "loss": 0.1404, + "step": 75184 + }, + { + "epoch": 2.89, + "learning_rate": 1.8241856886381123e-06, + "loss": 0.226, + "step": 75185 + }, + { + "epoch": 2.89, + "learning_rate": 1.8235444985893819e-06, + "loss": 0.3736, + "step": 75186 + }, + { + "epoch": 2.89, + "learning_rate": 1.8229033085406515e-06, + "loss": 0.2013, + "step": 75187 + }, + { + "epoch": 2.89, + "learning_rate": 1.822262118491921e-06, + "loss": 0.1636, + "step": 75188 + }, + { + "epoch": 2.89, + "learning_rate": 1.8216209284431907e-06, + "loss": 0.135, + "step": 75189 + }, + { + "epoch": 2.89, + "learning_rate": 1.8209797383944603e-06, + "loss": 0.2496, + "step": 75190 + }, + { + "epoch": 2.89, + "learning_rate": 1.8203385483457299e-06, + "loss": 0.1204, + "step": 75191 + }, + { + "epoch": 2.89, + "learning_rate": 1.8196973582969995e-06, + "loss": 0.1989, + "step": 75192 + }, + { + "epoch": 2.89, + "learning_rate": 1.819056168248269e-06, + "loss": 0.1758, + "step": 75193 + }, + { + "epoch": 2.89, + "learning_rate": 1.8184149781995382e-06, + "loss": 0.1513, + "step": 75194 + }, + { + "epoch": 2.89, + "learning_rate": 1.8177737881508078e-06, + "loss": 0.1575, + "step": 75195 + }, + { + "epoch": 2.89, + "learning_rate": 1.8171325981020774e-06, + "loss": 0.3546, + "step": 75196 + }, + { + "epoch": 2.89, + "learning_rate": 1.816491408053347e-06, + "loss": 0.177, + "step": 75197 + }, + { + "epoch": 2.89, + "learning_rate": 1.8158502180046166e-06, + "loss": 0.2818, + "step": 75198 + }, + { + "epoch": 2.89, + "learning_rate": 1.8152090279558862e-06, + "loss": 0.1511, + "step": 75199 + }, + { + "epoch": 2.89, + "learning_rate": 1.8145678379071558e-06, + "loss": 0.1916, + "step": 75200 + }, + { + "epoch": 2.89, + "learning_rate": 1.8139266478584254e-06, + "loss": 0.2943, + "step": 75201 + }, + { + "epoch": 2.89, + "learning_rate": 1.813285457809695e-06, + "loss": 0.2, + "step": 75202 + }, + { + "epoch": 2.89, + "learning_rate": 1.8126442677609646e-06, + "loss": 0.3065, + "step": 75203 + }, + { + "epoch": 2.89, + "learning_rate": 1.8120030777122338e-06, + "loss": 0.1167, + "step": 75204 + }, + { + "epoch": 2.89, + "learning_rate": 1.8113618876635034e-06, + "loss": 0.2203, + "step": 75205 + }, + { + "epoch": 2.89, + "learning_rate": 1.810720697614773e-06, + "loss": 0.2319, + "step": 75206 + }, + { + "epoch": 2.89, + "learning_rate": 1.8100795075660426e-06, + "loss": 0.2075, + "step": 75207 + }, + { + "epoch": 2.89, + "learning_rate": 1.8094383175173122e-06, + "loss": 0.2828, + "step": 75208 + }, + { + "epoch": 2.89, + "learning_rate": 1.8087971274685818e-06, + "loss": 0.1407, + "step": 75209 + }, + { + "epoch": 2.89, + "learning_rate": 1.8081559374198514e-06, + "loss": 0.1524, + "step": 75210 + }, + { + "epoch": 2.89, + "learning_rate": 1.807514747371121e-06, + "loss": 0.1357, + "step": 75211 + }, + { + "epoch": 2.89, + "learning_rate": 1.8068735573223906e-06, + "loss": 0.2032, + "step": 75212 + }, + { + "epoch": 2.89, + "learning_rate": 1.8062323672736602e-06, + "loss": 0.2455, + "step": 75213 + }, + { + "epoch": 2.89, + "learning_rate": 1.8055911772249294e-06, + "loss": 0.1845, + "step": 75214 + }, + { + "epoch": 2.89, + "learning_rate": 1.804949987176199e-06, + "loss": 0.3662, + "step": 75215 + }, + { + "epoch": 2.89, + "learning_rate": 1.8043087971274686e-06, + "loss": 0.1867, + "step": 75216 + }, + { + "epoch": 2.89, + "learning_rate": 1.8036676070787382e-06, + "loss": 0.2119, + "step": 75217 + }, + { + "epoch": 2.89, + "learning_rate": 1.8030264170300078e-06, + "loss": 0.1639, + "step": 75218 + }, + { + "epoch": 2.89, + "learning_rate": 1.8023852269812774e-06, + "loss": 0.1498, + "step": 75219 + }, + { + "epoch": 2.89, + "learning_rate": 1.801744036932547e-06, + "loss": 0.1599, + "step": 75220 + }, + { + "epoch": 2.89, + "learning_rate": 1.8011028468838166e-06, + "loss": 0.2522, + "step": 75221 + }, + { + "epoch": 2.89, + "learning_rate": 1.8004616568350862e-06, + "loss": 0.1384, + "step": 75222 + }, + { + "epoch": 2.89, + "learning_rate": 1.7998204667863558e-06, + "loss": 0.2244, + "step": 75223 + }, + { + "epoch": 2.89, + "learning_rate": 1.799179276737625e-06, + "loss": 0.3612, + "step": 75224 + }, + { + "epoch": 2.89, + "learning_rate": 1.7985380866888945e-06, + "loss": 0.1854, + "step": 75225 + }, + { + "epoch": 2.89, + "learning_rate": 1.7978968966401641e-06, + "loss": 0.1572, + "step": 75226 + }, + { + "epoch": 2.89, + "learning_rate": 1.7972557065914337e-06, + "loss": 0.1583, + "step": 75227 + }, + { + "epoch": 2.89, + "learning_rate": 1.7966145165427033e-06, + "loss": 0.1063, + "step": 75228 + }, + { + "epoch": 2.89, + "learning_rate": 1.795973326493973e-06, + "loss": 0.1635, + "step": 75229 + }, + { + "epoch": 2.89, + "learning_rate": 1.7953321364452425e-06, + "loss": 0.1711, + "step": 75230 + }, + { + "epoch": 2.89, + "learning_rate": 1.7946909463965121e-06, + "loss": 0.2057, + "step": 75231 + }, + { + "epoch": 2.89, + "learning_rate": 1.7940497563477817e-06, + "loss": 0.1226, + "step": 75232 + }, + { + "epoch": 2.89, + "learning_rate": 1.7934085662990513e-06, + "loss": 0.1229, + "step": 75233 + }, + { + "epoch": 2.89, + "learning_rate": 1.7927673762503205e-06, + "loss": 0.2578, + "step": 75234 + }, + { + "epoch": 2.89, + "learning_rate": 1.7921261862015901e-06, + "loss": 0.2288, + "step": 75235 + }, + { + "epoch": 2.89, + "learning_rate": 1.7914849961528597e-06, + "loss": 0.3731, + "step": 75236 + }, + { + "epoch": 2.89, + "learning_rate": 1.7908438061041293e-06, + "loss": 0.2003, + "step": 75237 + }, + { + "epoch": 2.89, + "learning_rate": 1.790202616055399e-06, + "loss": 0.2652, + "step": 75238 + }, + { + "epoch": 2.89, + "learning_rate": 1.7895614260066685e-06, + "loss": 0.1383, + "step": 75239 + }, + { + "epoch": 2.89, + "learning_rate": 1.7889202359579381e-06, + "loss": 0.1385, + "step": 75240 + }, + { + "epoch": 2.89, + "learning_rate": 1.7882790459092077e-06, + "loss": 0.1737, + "step": 75241 + }, + { + "epoch": 2.89, + "learning_rate": 1.7876378558604773e-06, + "loss": 0.2866, + "step": 75242 + }, + { + "epoch": 2.89, + "learning_rate": 1.7869966658117465e-06, + "loss": 0.1493, + "step": 75243 + }, + { + "epoch": 2.89, + "learning_rate": 1.786355475763016e-06, + "loss": 0.2427, + "step": 75244 + }, + { + "epoch": 2.89, + "learning_rate": 1.7857142857142857e-06, + "loss": 0.1727, + "step": 75245 + }, + { + "epoch": 2.89, + "learning_rate": 1.7850730956655553e-06, + "loss": 0.1332, + "step": 75246 + }, + { + "epoch": 2.89, + "learning_rate": 1.7844319056168249e-06, + "loss": 0.1234, + "step": 75247 + }, + { + "epoch": 2.89, + "learning_rate": 1.7837907155680945e-06, + "loss": 0.1511, + "step": 75248 + }, + { + "epoch": 2.89, + "learning_rate": 1.783149525519364e-06, + "loss": 0.1642, + "step": 75249 + }, + { + "epoch": 2.89, + "learning_rate": 1.7825083354706337e-06, + "loss": 0.2315, + "step": 75250 + }, + { + "epoch": 2.89, + "learning_rate": 1.7818671454219033e-06, + "loss": 0.131, + "step": 75251 + }, + { + "epoch": 2.89, + "learning_rate": 1.7812259553731729e-06, + "loss": 0.2257, + "step": 75252 + }, + { + "epoch": 2.89, + "learning_rate": 1.780584765324442e-06, + "loss": 0.1971, + "step": 75253 + }, + { + "epoch": 2.89, + "learning_rate": 1.7799435752757116e-06, + "loss": 0.2285, + "step": 75254 + }, + { + "epoch": 2.89, + "learning_rate": 1.7793023852269812e-06, + "loss": 0.1878, + "step": 75255 + }, + { + "epoch": 2.89, + "learning_rate": 1.7786611951782508e-06, + "loss": 0.2467, + "step": 75256 + }, + { + "epoch": 2.89, + "learning_rate": 1.7780200051295204e-06, + "loss": 0.1844, + "step": 75257 + }, + { + "epoch": 2.89, + "learning_rate": 1.77737881508079e-06, + "loss": 0.1974, + "step": 75258 + }, + { + "epoch": 2.89, + "learning_rate": 1.7767376250320596e-06, + "loss": 0.1673, + "step": 75259 + }, + { + "epoch": 2.89, + "learning_rate": 1.7760964349833292e-06, + "loss": 0.2859, + "step": 75260 + }, + { + "epoch": 2.89, + "learning_rate": 1.7754552449345988e-06, + "loss": 0.127, + "step": 75261 + }, + { + "epoch": 2.89, + "learning_rate": 1.7748140548858684e-06, + "loss": 0.2123, + "step": 75262 + }, + { + "epoch": 2.89, + "learning_rate": 1.7741728648371376e-06, + "loss": 0.1676, + "step": 75263 + }, + { + "epoch": 2.89, + "learning_rate": 1.7735316747884072e-06, + "loss": 0.2038, + "step": 75264 + }, + { + "epoch": 2.89, + "learning_rate": 1.7728904847396768e-06, + "loss": 0.1635, + "step": 75265 + }, + { + "epoch": 2.89, + "learning_rate": 1.7722492946909464e-06, + "loss": 0.1185, + "step": 75266 + }, + { + "epoch": 2.89, + "learning_rate": 1.771608104642216e-06, + "loss": 0.1177, + "step": 75267 + }, + { + "epoch": 2.89, + "learning_rate": 1.7709669145934856e-06, + "loss": 0.2891, + "step": 75268 + }, + { + "epoch": 2.89, + "learning_rate": 1.7703257245447552e-06, + "loss": 0.2315, + "step": 75269 + }, + { + "epoch": 2.89, + "learning_rate": 1.7696845344960248e-06, + "loss": 0.0794, + "step": 75270 + }, + { + "epoch": 2.89, + "learning_rate": 1.7690433444472944e-06, + "loss": 0.2333, + "step": 75271 + }, + { + "epoch": 2.89, + "learning_rate": 1.768402154398564e-06, + "loss": 0.1343, + "step": 75272 + }, + { + "epoch": 2.89, + "learning_rate": 1.7677609643498332e-06, + "loss": 0.2746, + "step": 75273 + }, + { + "epoch": 2.89, + "learning_rate": 1.7671197743011028e-06, + "loss": 0.242, + "step": 75274 + }, + { + "epoch": 2.89, + "learning_rate": 1.7664785842523724e-06, + "loss": 0.2841, + "step": 75275 + }, + { + "epoch": 2.89, + "learning_rate": 1.765837394203642e-06, + "loss": 0.1243, + "step": 75276 + }, + { + "epoch": 2.89, + "learning_rate": 1.7651962041549116e-06, + "loss": 0.149, + "step": 75277 + }, + { + "epoch": 2.89, + "learning_rate": 1.7645550141061812e-06, + "loss": 0.2085, + "step": 75278 + }, + { + "epoch": 2.89, + "learning_rate": 1.7639138240574508e-06, + "loss": 0.1281, + "step": 75279 + }, + { + "epoch": 2.89, + "learning_rate": 1.7632726340087204e-06, + "loss": 0.1363, + "step": 75280 + }, + { + "epoch": 2.89, + "learning_rate": 1.76263144395999e-06, + "loss": 0.1862, + "step": 75281 + }, + { + "epoch": 2.89, + "learning_rate": 1.7619902539112596e-06, + "loss": 0.2747, + "step": 75282 + }, + { + "epoch": 2.89, + "learning_rate": 1.7613490638625287e-06, + "loss": 0.1674, + "step": 75283 + }, + { + "epoch": 2.89, + "learning_rate": 1.7607078738137983e-06, + "loss": 0.2599, + "step": 75284 + }, + { + "epoch": 2.89, + "learning_rate": 1.760066683765068e-06, + "loss": 0.0862, + "step": 75285 + }, + { + "epoch": 2.89, + "learning_rate": 1.7594254937163375e-06, + "loss": 0.2208, + "step": 75286 + }, + { + "epoch": 2.89, + "learning_rate": 1.7587843036676071e-06, + "loss": 0.1421, + "step": 75287 + }, + { + "epoch": 2.89, + "learning_rate": 1.7581431136188767e-06, + "loss": 0.1566, + "step": 75288 + }, + { + "epoch": 2.89, + "learning_rate": 1.7575019235701463e-06, + "loss": 0.1457, + "step": 75289 + }, + { + "epoch": 2.89, + "learning_rate": 1.756860733521416e-06, + "loss": 0.1396, + "step": 75290 + }, + { + "epoch": 2.89, + "learning_rate": 1.7562195434726855e-06, + "loss": 0.1655, + "step": 75291 + }, + { + "epoch": 2.89, + "learning_rate": 1.7555783534239551e-06, + "loss": 0.2612, + "step": 75292 + }, + { + "epoch": 2.89, + "learning_rate": 1.7549371633752243e-06, + "loss": 0.1355, + "step": 75293 + }, + { + "epoch": 2.89, + "learning_rate": 1.754295973326494e-06, + "loss": 0.3255, + "step": 75294 + }, + { + "epoch": 2.89, + "learning_rate": 1.7536547832777635e-06, + "loss": 0.2334, + "step": 75295 + }, + { + "epoch": 2.89, + "learning_rate": 1.7530135932290331e-06, + "loss": 0.1878, + "step": 75296 + }, + { + "epoch": 2.89, + "learning_rate": 1.7523724031803027e-06, + "loss": 0.2207, + "step": 75297 + }, + { + "epoch": 2.89, + "learning_rate": 1.7517312131315723e-06, + "loss": 0.1855, + "step": 75298 + }, + { + "epoch": 2.9, + "learning_rate": 1.751090023082842e-06, + "loss": 0.0838, + "step": 75299 + }, + { + "epoch": 2.9, + "learning_rate": 1.7504488330341115e-06, + "loss": 0.1745, + "step": 75300 + }, + { + "epoch": 2.9, + "learning_rate": 1.7498076429853811e-06, + "loss": 0.2383, + "step": 75301 + }, + { + "epoch": 2.9, + "learning_rate": 1.7491664529366503e-06, + "loss": 0.0993, + "step": 75302 + }, + { + "epoch": 2.9, + "learning_rate": 1.7485252628879199e-06, + "loss": 0.1618, + "step": 75303 + }, + { + "epoch": 2.9, + "learning_rate": 1.7478840728391895e-06, + "loss": 0.1877, + "step": 75304 + }, + { + "epoch": 2.9, + "learning_rate": 1.747242882790459e-06, + "loss": 0.1704, + "step": 75305 + }, + { + "epoch": 2.9, + "learning_rate": 1.7466016927417287e-06, + "loss": 0.2498, + "step": 75306 + }, + { + "epoch": 2.9, + "learning_rate": 1.7459605026929983e-06, + "loss": 0.2016, + "step": 75307 + }, + { + "epoch": 2.9, + "learning_rate": 1.7453193126442679e-06, + "loss": 0.2116, + "step": 75308 + }, + { + "epoch": 2.9, + "learning_rate": 1.7446781225955375e-06, + "loss": 0.1804, + "step": 75309 + }, + { + "epoch": 2.9, + "learning_rate": 1.744036932546807e-06, + "loss": 0.3095, + "step": 75310 + }, + { + "epoch": 2.9, + "learning_rate": 1.7433957424980767e-06, + "loss": 0.1322, + "step": 75311 + }, + { + "epoch": 2.9, + "learning_rate": 1.7427545524493458e-06, + "loss": 0.316, + "step": 75312 + }, + { + "epoch": 2.9, + "learning_rate": 1.7421133624006154e-06, + "loss": 0.1768, + "step": 75313 + }, + { + "epoch": 2.9, + "learning_rate": 1.741472172351885e-06, + "loss": 0.1459, + "step": 75314 + }, + { + "epoch": 2.9, + "learning_rate": 1.7408309823031546e-06, + "loss": 0.311, + "step": 75315 + }, + { + "epoch": 2.9, + "learning_rate": 1.7401897922544242e-06, + "loss": 0.209, + "step": 75316 + }, + { + "epoch": 2.9, + "learning_rate": 1.7395486022056938e-06, + "loss": 0.2203, + "step": 75317 + }, + { + "epoch": 2.9, + "learning_rate": 1.7389074121569634e-06, + "loss": 0.2577, + "step": 75318 + }, + { + "epoch": 2.9, + "learning_rate": 1.738266222108233e-06, + "loss": 0.2819, + "step": 75319 + }, + { + "epoch": 2.9, + "learning_rate": 1.7376250320595026e-06, + "loss": 0.2347, + "step": 75320 + }, + { + "epoch": 2.9, + "learning_rate": 1.7369838420107722e-06, + "loss": 0.2415, + "step": 75321 + }, + { + "epoch": 2.9, + "learning_rate": 1.7363426519620414e-06, + "loss": 0.0897, + "step": 75322 + }, + { + "epoch": 2.9, + "learning_rate": 1.735701461913311e-06, + "loss": 0.1226, + "step": 75323 + }, + { + "epoch": 2.9, + "learning_rate": 1.7350602718645806e-06, + "loss": 0.1801, + "step": 75324 + }, + { + "epoch": 2.9, + "learning_rate": 1.7344190818158502e-06, + "loss": 0.1685, + "step": 75325 + }, + { + "epoch": 2.9, + "learning_rate": 1.7337778917671198e-06, + "loss": 0.1938, + "step": 75326 + }, + { + "epoch": 2.9, + "learning_rate": 1.7331367017183894e-06, + "loss": 0.2105, + "step": 75327 + }, + { + "epoch": 2.9, + "learning_rate": 1.732495511669659e-06, + "loss": 0.3146, + "step": 75328 + }, + { + "epoch": 2.9, + "learning_rate": 1.7318543216209286e-06, + "loss": 0.0794, + "step": 75329 + }, + { + "epoch": 2.9, + "learning_rate": 1.7312131315721982e-06, + "loss": 0.1447, + "step": 75330 + }, + { + "epoch": 2.9, + "learning_rate": 1.7305719415234678e-06, + "loss": 0.2144, + "step": 75331 + }, + { + "epoch": 2.9, + "learning_rate": 1.729930751474737e-06, + "loss": 0.1361, + "step": 75332 + }, + { + "epoch": 2.9, + "learning_rate": 1.7292895614260066e-06, + "loss": 0.0713, + "step": 75333 + }, + { + "epoch": 2.9, + "learning_rate": 1.7286483713772762e-06, + "loss": 0.3859, + "step": 75334 + }, + { + "epoch": 2.9, + "learning_rate": 1.7280071813285458e-06, + "loss": 0.0998, + "step": 75335 + }, + { + "epoch": 2.9, + "learning_rate": 1.7273659912798154e-06, + "loss": 0.1252, + "step": 75336 + }, + { + "epoch": 2.9, + "learning_rate": 1.726724801231085e-06, + "loss": 0.1607, + "step": 75337 + }, + { + "epoch": 2.9, + "learning_rate": 1.7260836111823546e-06, + "loss": 0.3184, + "step": 75338 + }, + { + "epoch": 2.9, + "learning_rate": 1.7254424211336242e-06, + "loss": 0.167, + "step": 75339 + }, + { + "epoch": 2.9, + "learning_rate": 1.7248012310848938e-06, + "loss": 0.299, + "step": 75340 + }, + { + "epoch": 2.9, + "learning_rate": 1.7241600410361634e-06, + "loss": 0.2393, + "step": 75341 + }, + { + "epoch": 2.9, + "learning_rate": 1.7235188509874326e-06, + "loss": 0.1955, + "step": 75342 + }, + { + "epoch": 2.9, + "learning_rate": 1.7228776609387021e-06, + "loss": 0.2692, + "step": 75343 + }, + { + "epoch": 2.9, + "learning_rate": 1.7222364708899717e-06, + "loss": 0.2776, + "step": 75344 + }, + { + "epoch": 2.9, + "learning_rate": 1.7215952808412413e-06, + "loss": 0.1517, + "step": 75345 + }, + { + "epoch": 2.9, + "learning_rate": 1.720954090792511e-06, + "loss": 0.1233, + "step": 75346 + }, + { + "epoch": 2.9, + "learning_rate": 1.7203129007437805e-06, + "loss": 0.1441, + "step": 75347 + }, + { + "epoch": 2.9, + "learning_rate": 1.7196717106950501e-06, + "loss": 0.2487, + "step": 75348 + }, + { + "epoch": 2.9, + "learning_rate": 1.7190305206463197e-06, + "loss": 0.229, + "step": 75349 + }, + { + "epoch": 2.9, + "learning_rate": 1.7183893305975893e-06, + "loss": 0.2447, + "step": 75350 + }, + { + "epoch": 2.9, + "learning_rate": 1.717748140548859e-06, + "loss": 0.1251, + "step": 75351 + }, + { + "epoch": 2.9, + "learning_rate": 1.7171069505001281e-06, + "loss": 0.1721, + "step": 75352 + }, + { + "epoch": 2.9, + "learning_rate": 1.7164657604513977e-06, + "loss": 0.1963, + "step": 75353 + }, + { + "epoch": 2.9, + "learning_rate": 1.7158245704026673e-06, + "loss": 0.1452, + "step": 75354 + }, + { + "epoch": 2.9, + "learning_rate": 1.715183380353937e-06, + "loss": 0.3203, + "step": 75355 + }, + { + "epoch": 2.9, + "learning_rate": 1.7145421903052065e-06, + "loss": 0.2754, + "step": 75356 + }, + { + "epoch": 2.9, + "learning_rate": 1.7139010002564761e-06, + "loss": 0.1731, + "step": 75357 + }, + { + "epoch": 2.9, + "learning_rate": 1.7132598102077457e-06, + "loss": 0.2816, + "step": 75358 + }, + { + "epoch": 2.9, + "learning_rate": 1.7126186201590153e-06, + "loss": 0.2246, + "step": 75359 + }, + { + "epoch": 2.9, + "learning_rate": 1.711977430110285e-06, + "loss": 0.2329, + "step": 75360 + }, + { + "epoch": 2.9, + "learning_rate": 1.711336240061554e-06, + "loss": 0.1353, + "step": 75361 + }, + { + "epoch": 2.9, + "learning_rate": 1.7106950500128237e-06, + "loss": 0.2431, + "step": 75362 + }, + { + "epoch": 2.9, + "learning_rate": 1.7100538599640933e-06, + "loss": 0.1586, + "step": 75363 + }, + { + "epoch": 2.9, + "learning_rate": 1.7094126699153629e-06, + "loss": 0.2396, + "step": 75364 + }, + { + "epoch": 2.9, + "learning_rate": 1.7087714798666325e-06, + "loss": 0.2208, + "step": 75365 + }, + { + "epoch": 2.9, + "learning_rate": 1.708130289817902e-06, + "loss": 0.2358, + "step": 75366 + }, + { + "epoch": 2.9, + "learning_rate": 1.7074890997691717e-06, + "loss": 0.2555, + "step": 75367 + }, + { + "epoch": 2.9, + "learning_rate": 1.7068479097204413e-06, + "loss": 0.2682, + "step": 75368 + }, + { + "epoch": 2.9, + "learning_rate": 1.7062067196717109e-06, + "loss": 0.2123, + "step": 75369 + }, + { + "epoch": 2.9, + "learning_rate": 1.7055655296229805e-06, + "loss": 0.2707, + "step": 75370 + }, + { + "epoch": 2.9, + "learning_rate": 1.7049243395742497e-06, + "loss": 0.324, + "step": 75371 + }, + { + "epoch": 2.9, + "learning_rate": 1.7042831495255193e-06, + "loss": 0.0991, + "step": 75372 + }, + { + "epoch": 2.9, + "learning_rate": 1.7036419594767889e-06, + "loss": 0.3271, + "step": 75373 + }, + { + "epoch": 2.9, + "learning_rate": 1.7030007694280584e-06, + "loss": 0.1789, + "step": 75374 + }, + { + "epoch": 2.9, + "learning_rate": 1.702359579379328e-06, + "loss": 0.2951, + "step": 75375 + }, + { + "epoch": 2.9, + "learning_rate": 1.7017183893305976e-06, + "loss": 0.2292, + "step": 75376 + }, + { + "epoch": 2.9, + "learning_rate": 1.7010771992818672e-06, + "loss": 0.2683, + "step": 75377 + }, + { + "epoch": 2.9, + "learning_rate": 1.7004360092331368e-06, + "loss": 0.3808, + "step": 75378 + }, + { + "epoch": 2.9, + "learning_rate": 1.6997948191844064e-06, + "loss": 0.1623, + "step": 75379 + }, + { + "epoch": 2.9, + "learning_rate": 1.699153629135676e-06, + "loss": 0.2584, + "step": 75380 + }, + { + "epoch": 2.9, + "learning_rate": 1.6985124390869452e-06, + "loss": 0.1881, + "step": 75381 + }, + { + "epoch": 2.9, + "learning_rate": 1.6978712490382148e-06, + "loss": 0.0829, + "step": 75382 + }, + { + "epoch": 2.9, + "learning_rate": 1.6972300589894844e-06, + "loss": 0.1387, + "step": 75383 + }, + { + "epoch": 2.9, + "learning_rate": 1.696588868940754e-06, + "loss": 0.2137, + "step": 75384 + }, + { + "epoch": 2.9, + "learning_rate": 1.6959476788920236e-06, + "loss": 0.1794, + "step": 75385 + }, + { + "epoch": 2.9, + "learning_rate": 1.6953064888432932e-06, + "loss": 0.2066, + "step": 75386 + }, + { + "epoch": 2.9, + "learning_rate": 1.6946652987945628e-06, + "loss": 0.2138, + "step": 75387 + }, + { + "epoch": 2.9, + "learning_rate": 1.6940241087458324e-06, + "loss": 0.3615, + "step": 75388 + }, + { + "epoch": 2.9, + "learning_rate": 1.693382918697102e-06, + "loss": 0.0765, + "step": 75389 + }, + { + "epoch": 2.9, + "learning_rate": 1.6927417286483716e-06, + "loss": 0.1539, + "step": 75390 + }, + { + "epoch": 2.9, + "learning_rate": 1.6921005385996408e-06, + "loss": 0.1198, + "step": 75391 + }, + { + "epoch": 2.9, + "learning_rate": 1.6914593485509104e-06, + "loss": 0.1684, + "step": 75392 + }, + { + "epoch": 2.9, + "learning_rate": 1.69081815850218e-06, + "loss": 0.2331, + "step": 75393 + }, + { + "epoch": 2.9, + "learning_rate": 1.6901769684534496e-06, + "loss": 0.182, + "step": 75394 + }, + { + "epoch": 2.9, + "learning_rate": 1.6895357784047192e-06, + "loss": 0.1728, + "step": 75395 + }, + { + "epoch": 2.9, + "learning_rate": 1.6888945883559888e-06, + "loss": 0.1298, + "step": 75396 + }, + { + "epoch": 2.9, + "learning_rate": 1.6882533983072584e-06, + "loss": 0.1753, + "step": 75397 + }, + { + "epoch": 2.9, + "learning_rate": 1.687612208258528e-06, + "loss": 0.2965, + "step": 75398 + }, + { + "epoch": 2.9, + "learning_rate": 1.6869710182097976e-06, + "loss": 0.2439, + "step": 75399 + }, + { + "epoch": 2.9, + "learning_rate": 1.6863298281610672e-06, + "loss": 0.1158, + "step": 75400 + }, + { + "epoch": 2.9, + "learning_rate": 1.6856886381123364e-06, + "loss": 0.341, + "step": 75401 + }, + { + "epoch": 2.9, + "learning_rate": 1.685047448063606e-06, + "loss": 0.2299, + "step": 75402 + }, + { + "epoch": 2.9, + "learning_rate": 1.6844062580148756e-06, + "loss": 0.1636, + "step": 75403 + }, + { + "epoch": 2.9, + "learning_rate": 1.6837650679661452e-06, + "loss": 0.1509, + "step": 75404 + }, + { + "epoch": 2.9, + "learning_rate": 1.6831238779174147e-06, + "loss": 0.3253, + "step": 75405 + }, + { + "epoch": 2.9, + "learning_rate": 1.6824826878686843e-06, + "loss": 0.1875, + "step": 75406 + }, + { + "epoch": 2.9, + "learning_rate": 1.681841497819954e-06, + "loss": 0.1385, + "step": 75407 + }, + { + "epoch": 2.9, + "learning_rate": 1.6812003077712235e-06, + "loss": 0.1197, + "step": 75408 + }, + { + "epoch": 2.9, + "learning_rate": 1.6805591177224931e-06, + "loss": 0.1764, + "step": 75409 + }, + { + "epoch": 2.9, + "learning_rate": 1.6799179276737627e-06, + "loss": 0.2234, + "step": 75410 + }, + { + "epoch": 2.9, + "learning_rate": 1.679276737625032e-06, + "loss": 0.178, + "step": 75411 + }, + { + "epoch": 2.9, + "learning_rate": 1.6786355475763015e-06, + "loss": 0.233, + "step": 75412 + }, + { + "epoch": 2.9, + "learning_rate": 1.6779943575275711e-06, + "loss": 0.162, + "step": 75413 + }, + { + "epoch": 2.9, + "learning_rate": 1.6773531674788407e-06, + "loss": 0.1657, + "step": 75414 + }, + { + "epoch": 2.9, + "learning_rate": 1.6767119774301103e-06, + "loss": 0.1882, + "step": 75415 + }, + { + "epoch": 2.9, + "learning_rate": 1.67607078738138e-06, + "loss": 0.2551, + "step": 75416 + }, + { + "epoch": 2.9, + "learning_rate": 1.6754295973326495e-06, + "loss": 0.152, + "step": 75417 + }, + { + "epoch": 2.9, + "learning_rate": 1.6747884072839191e-06, + "loss": 0.1715, + "step": 75418 + }, + { + "epoch": 2.9, + "learning_rate": 1.6741472172351887e-06, + "loss": 0.1153, + "step": 75419 + }, + { + "epoch": 2.9, + "learning_rate": 1.6735060271864579e-06, + "loss": 0.2144, + "step": 75420 + }, + { + "epoch": 2.9, + "learning_rate": 1.6728648371377275e-06, + "loss": 0.2998, + "step": 75421 + }, + { + "epoch": 2.9, + "learning_rate": 1.672223647088997e-06, + "loss": 0.1469, + "step": 75422 + }, + { + "epoch": 2.9, + "learning_rate": 1.6715824570402667e-06, + "loss": 0.4055, + "step": 75423 + }, + { + "epoch": 2.9, + "learning_rate": 1.6709412669915363e-06, + "loss": 0.1866, + "step": 75424 + }, + { + "epoch": 2.9, + "learning_rate": 1.6703000769428059e-06, + "loss": 0.1159, + "step": 75425 + }, + { + "epoch": 2.9, + "learning_rate": 1.6696588868940755e-06, + "loss": 0.2574, + "step": 75426 + }, + { + "epoch": 2.9, + "learning_rate": 1.669017696845345e-06, + "loss": 0.154, + "step": 75427 + }, + { + "epoch": 2.9, + "learning_rate": 1.6683765067966147e-06, + "loss": 0.2057, + "step": 75428 + }, + { + "epoch": 2.9, + "learning_rate": 1.6677353167478843e-06, + "loss": 0.1056, + "step": 75429 + }, + { + "epoch": 2.9, + "learning_rate": 1.6670941266991535e-06, + "loss": 0.1859, + "step": 75430 + }, + { + "epoch": 2.9, + "learning_rate": 1.666452936650423e-06, + "loss": 0.2139, + "step": 75431 + }, + { + "epoch": 2.9, + "learning_rate": 1.6658117466016927e-06, + "loss": 0.1399, + "step": 75432 + }, + { + "epoch": 2.9, + "learning_rate": 1.6651705565529623e-06, + "loss": 0.1584, + "step": 75433 + }, + { + "epoch": 2.9, + "learning_rate": 1.6645293665042319e-06, + "loss": 0.2185, + "step": 75434 + }, + { + "epoch": 2.9, + "learning_rate": 1.6638881764555014e-06, + "loss": 0.2377, + "step": 75435 + }, + { + "epoch": 2.9, + "learning_rate": 1.663246986406771e-06, + "loss": 0.2404, + "step": 75436 + }, + { + "epoch": 2.9, + "learning_rate": 1.6626057963580406e-06, + "loss": 0.1813, + "step": 75437 + }, + { + "epoch": 2.9, + "learning_rate": 1.6619646063093102e-06, + "loss": 0.1975, + "step": 75438 + }, + { + "epoch": 2.9, + "learning_rate": 1.6613234162605798e-06, + "loss": 0.1266, + "step": 75439 + }, + { + "epoch": 2.9, + "learning_rate": 1.660682226211849e-06, + "loss": 0.1871, + "step": 75440 + }, + { + "epoch": 2.9, + "learning_rate": 1.6600410361631186e-06, + "loss": 0.124, + "step": 75441 + }, + { + "epoch": 2.9, + "learning_rate": 1.6593998461143882e-06, + "loss": 0.2942, + "step": 75442 + }, + { + "epoch": 2.9, + "learning_rate": 1.6587586560656578e-06, + "loss": 0.2314, + "step": 75443 + }, + { + "epoch": 2.9, + "learning_rate": 1.6581174660169274e-06, + "loss": 0.1404, + "step": 75444 + }, + { + "epoch": 2.9, + "learning_rate": 1.657476275968197e-06, + "loss": 0.2157, + "step": 75445 + }, + { + "epoch": 2.9, + "learning_rate": 1.6568350859194666e-06, + "loss": 0.0909, + "step": 75446 + }, + { + "epoch": 2.9, + "learning_rate": 1.6561938958707362e-06, + "loss": 0.3007, + "step": 75447 + }, + { + "epoch": 2.9, + "learning_rate": 1.6555527058220058e-06, + "loss": 0.1868, + "step": 75448 + }, + { + "epoch": 2.9, + "learning_rate": 1.6549115157732754e-06, + "loss": 0.2134, + "step": 75449 + }, + { + "epoch": 2.9, + "learning_rate": 1.6542703257245446e-06, + "loss": 0.1199, + "step": 75450 + }, + { + "epoch": 2.9, + "learning_rate": 1.6536291356758142e-06, + "loss": 0.1768, + "step": 75451 + }, + { + "epoch": 2.9, + "learning_rate": 1.6529879456270838e-06, + "loss": 0.2261, + "step": 75452 + }, + { + "epoch": 2.9, + "learning_rate": 1.6523467555783534e-06, + "loss": 0.3471, + "step": 75453 + }, + { + "epoch": 2.9, + "learning_rate": 1.651705565529623e-06, + "loss": 0.3176, + "step": 75454 + }, + { + "epoch": 2.9, + "learning_rate": 1.6510643754808926e-06, + "loss": 0.1971, + "step": 75455 + }, + { + "epoch": 2.9, + "learning_rate": 1.6504231854321622e-06, + "loss": 0.278, + "step": 75456 + }, + { + "epoch": 2.9, + "learning_rate": 1.6497819953834318e-06, + "loss": 0.1693, + "step": 75457 + }, + { + "epoch": 2.9, + "learning_rate": 1.6491408053347014e-06, + "loss": 0.2317, + "step": 75458 + }, + { + "epoch": 2.9, + "learning_rate": 1.648499615285971e-06, + "loss": 0.1792, + "step": 75459 + }, + { + "epoch": 2.9, + "learning_rate": 1.6478584252372402e-06, + "loss": 0.1916, + "step": 75460 + }, + { + "epoch": 2.9, + "learning_rate": 1.6472172351885098e-06, + "loss": 0.2994, + "step": 75461 + }, + { + "epoch": 2.9, + "learning_rate": 1.6465760451397794e-06, + "loss": 0.2611, + "step": 75462 + }, + { + "epoch": 2.9, + "learning_rate": 1.645934855091049e-06, + "loss": 0.1939, + "step": 75463 + }, + { + "epoch": 2.9, + "learning_rate": 1.6452936650423186e-06, + "loss": 0.1555, + "step": 75464 + }, + { + "epoch": 2.9, + "learning_rate": 1.6446524749935882e-06, + "loss": 0.1485, + "step": 75465 + }, + { + "epoch": 2.9, + "learning_rate": 1.6440112849448577e-06, + "loss": 0.1681, + "step": 75466 + }, + { + "epoch": 2.9, + "learning_rate": 1.6433700948961273e-06, + "loss": 0.2541, + "step": 75467 + }, + { + "epoch": 2.9, + "learning_rate": 1.642728904847397e-06, + "loss": 0.225, + "step": 75468 + }, + { + "epoch": 2.9, + "learning_rate": 1.6420877147986665e-06, + "loss": 0.2042, + "step": 75469 + }, + { + "epoch": 2.9, + "learning_rate": 1.6414465247499357e-06, + "loss": 0.2181, + "step": 75470 + }, + { + "epoch": 2.9, + "learning_rate": 1.6408053347012053e-06, + "loss": 0.2318, + "step": 75471 + }, + { + "epoch": 2.9, + "learning_rate": 1.640164144652475e-06, + "loss": 0.2953, + "step": 75472 + }, + { + "epoch": 2.9, + "learning_rate": 1.6395229546037445e-06, + "loss": 0.1211, + "step": 75473 + }, + { + "epoch": 2.9, + "learning_rate": 1.6388817645550141e-06, + "loss": 0.2431, + "step": 75474 + }, + { + "epoch": 2.9, + "learning_rate": 1.6382405745062837e-06, + "loss": 0.259, + "step": 75475 + }, + { + "epoch": 2.9, + "learning_rate": 1.6375993844575533e-06, + "loss": 0.2394, + "step": 75476 + }, + { + "epoch": 2.9, + "learning_rate": 1.636958194408823e-06, + "loss": 0.2042, + "step": 75477 + }, + { + "epoch": 2.9, + "learning_rate": 1.6363170043600925e-06, + "loss": 0.1639, + "step": 75478 + }, + { + "epoch": 2.9, + "learning_rate": 1.635675814311362e-06, + "loss": 0.2556, + "step": 75479 + }, + { + "epoch": 2.9, + "learning_rate": 1.6350346242626313e-06, + "loss": 0.2599, + "step": 75480 + }, + { + "epoch": 2.9, + "learning_rate": 1.6343934342139009e-06, + "loss": 0.1803, + "step": 75481 + }, + { + "epoch": 2.9, + "learning_rate": 1.6337522441651705e-06, + "loss": 0.234, + "step": 75482 + }, + { + "epoch": 2.9, + "learning_rate": 1.63311105411644e-06, + "loss": 0.229, + "step": 75483 + }, + { + "epoch": 2.9, + "learning_rate": 1.6324698640677097e-06, + "loss": 0.23, + "step": 75484 + }, + { + "epoch": 2.9, + "learning_rate": 1.6318286740189793e-06, + "loss": 0.183, + "step": 75485 + }, + { + "epoch": 2.9, + "learning_rate": 1.6311874839702489e-06, + "loss": 0.1807, + "step": 75486 + }, + { + "epoch": 2.9, + "learning_rate": 1.6305462939215185e-06, + "loss": 0.142, + "step": 75487 + }, + { + "epoch": 2.9, + "learning_rate": 1.629905103872788e-06, + "loss": 0.2142, + "step": 75488 + }, + { + "epoch": 2.9, + "learning_rate": 1.6292639138240575e-06, + "loss": 0.1726, + "step": 75489 + }, + { + "epoch": 2.9, + "learning_rate": 1.6286227237753269e-06, + "loss": 0.1135, + "step": 75490 + }, + { + "epoch": 2.9, + "learning_rate": 1.6279815337265965e-06, + "loss": 0.1978, + "step": 75491 + }, + { + "epoch": 2.9, + "learning_rate": 1.627340343677866e-06, + "loss": 0.1651, + "step": 75492 + }, + { + "epoch": 2.9, + "learning_rate": 1.6266991536291357e-06, + "loss": 0.1823, + "step": 75493 + }, + { + "epoch": 2.9, + "learning_rate": 1.6260579635804053e-06, + "loss": 0.1971, + "step": 75494 + }, + { + "epoch": 2.9, + "learning_rate": 1.6254167735316749e-06, + "loss": 0.1854, + "step": 75495 + }, + { + "epoch": 2.9, + "learning_rate": 1.6247755834829445e-06, + "loss": 0.1541, + "step": 75496 + }, + { + "epoch": 2.9, + "learning_rate": 1.624134393434214e-06, + "loss": 0.1718, + "step": 75497 + }, + { + "epoch": 2.9, + "learning_rate": 1.6234932033854836e-06, + "loss": 0.1726, + "step": 75498 + }, + { + "epoch": 2.9, + "learning_rate": 1.622852013336753e-06, + "loss": 0.2136, + "step": 75499 + }, + { + "epoch": 2.9, + "learning_rate": 1.6222108232880224e-06, + "loss": 0.1441, + "step": 75500 + }, + { + "epoch": 2.9, + "learning_rate": 1.621569633239292e-06, + "loss": 0.2192, + "step": 75501 + }, + { + "epoch": 2.9, + "learning_rate": 1.6209284431905616e-06, + "loss": 0.2084, + "step": 75502 + }, + { + "epoch": 2.9, + "learning_rate": 1.6202872531418312e-06, + "loss": 0.1672, + "step": 75503 + }, + { + "epoch": 2.9, + "learning_rate": 1.6196460630931008e-06, + "loss": 0.2237, + "step": 75504 + }, + { + "epoch": 2.9, + "learning_rate": 1.6190048730443704e-06, + "loss": 0.2021, + "step": 75505 + }, + { + "epoch": 2.9, + "learning_rate": 1.61836368299564e-06, + "loss": 0.0996, + "step": 75506 + }, + { + "epoch": 2.9, + "learning_rate": 1.6177224929469096e-06, + "loss": 0.2304, + "step": 75507 + }, + { + "epoch": 2.9, + "learning_rate": 1.6170813028981792e-06, + "loss": 0.1692, + "step": 75508 + }, + { + "epoch": 2.9, + "learning_rate": 1.6164401128494486e-06, + "loss": 0.1723, + "step": 75509 + }, + { + "epoch": 2.9, + "learning_rate": 1.6157989228007182e-06, + "loss": 0.2546, + "step": 75510 + }, + { + "epoch": 2.9, + "learning_rate": 1.6151577327519876e-06, + "loss": 0.2429, + "step": 75511 + }, + { + "epoch": 2.9, + "learning_rate": 1.6145165427032572e-06, + "loss": 0.3046, + "step": 75512 + }, + { + "epoch": 2.9, + "learning_rate": 1.6138753526545268e-06, + "loss": 0.2497, + "step": 75513 + }, + { + "epoch": 2.9, + "learning_rate": 1.6132341626057964e-06, + "loss": 0.2559, + "step": 75514 + }, + { + "epoch": 2.9, + "learning_rate": 1.612592972557066e-06, + "loss": 0.1651, + "step": 75515 + }, + { + "epoch": 2.9, + "learning_rate": 1.6119517825083356e-06, + "loss": 0.1886, + "step": 75516 + }, + { + "epoch": 2.9, + "learning_rate": 1.6113105924596052e-06, + "loss": 0.3372, + "step": 75517 + }, + { + "epoch": 2.9, + "learning_rate": 1.6106694024108748e-06, + "loss": 0.2069, + "step": 75518 + }, + { + "epoch": 2.9, + "learning_rate": 1.6100282123621442e-06, + "loss": 0.1393, + "step": 75519 + }, + { + "epoch": 2.9, + "learning_rate": 1.6093870223134138e-06, + "loss": 0.2926, + "step": 75520 + }, + { + "epoch": 2.9, + "learning_rate": 1.6087458322646832e-06, + "loss": 0.1542, + "step": 75521 + }, + { + "epoch": 2.9, + "learning_rate": 1.6081046422159528e-06, + "loss": 0.0948, + "step": 75522 + }, + { + "epoch": 2.9, + "learning_rate": 1.6074634521672224e-06, + "loss": 0.378, + "step": 75523 + }, + { + "epoch": 2.9, + "learning_rate": 1.606822262118492e-06, + "loss": 0.2564, + "step": 75524 + }, + { + "epoch": 2.9, + "learning_rate": 1.6061810720697616e-06, + "loss": 0.3312, + "step": 75525 + }, + { + "epoch": 2.9, + "learning_rate": 1.6055398820210312e-06, + "loss": 0.2107, + "step": 75526 + }, + { + "epoch": 2.9, + "learning_rate": 1.6048986919723007e-06, + "loss": 0.2526, + "step": 75527 + }, + { + "epoch": 2.9, + "learning_rate": 1.6042575019235701e-06, + "loss": 0.2258, + "step": 75528 + }, + { + "epoch": 2.9, + "learning_rate": 1.6036163118748397e-06, + "loss": 0.1814, + "step": 75529 + }, + { + "epoch": 2.9, + "learning_rate": 1.6029751218261093e-06, + "loss": 0.1826, + "step": 75530 + }, + { + "epoch": 2.9, + "learning_rate": 1.6023339317773787e-06, + "loss": 0.2572, + "step": 75531 + }, + { + "epoch": 2.9, + "learning_rate": 1.6016927417286483e-06, + "loss": 0.0774, + "step": 75532 + }, + { + "epoch": 2.9, + "learning_rate": 1.601051551679918e-06, + "loss": 0.3555, + "step": 75533 + }, + { + "epoch": 2.9, + "learning_rate": 1.6004103616311875e-06, + "loss": 0.2336, + "step": 75534 + }, + { + "epoch": 2.9, + "learning_rate": 1.5997691715824571e-06, + "loss": 0.2271, + "step": 75535 + }, + { + "epoch": 2.9, + "learning_rate": 1.5991279815337267e-06, + "loss": 0.1354, + "step": 75536 + }, + { + "epoch": 2.9, + "learning_rate": 1.5984867914849963e-06, + "loss": 0.2389, + "step": 75537 + }, + { + "epoch": 2.9, + "learning_rate": 1.5978456014362657e-06, + "loss": 0.3076, + "step": 75538 + }, + { + "epoch": 2.9, + "learning_rate": 1.5972044113875353e-06, + "loss": 0.1474, + "step": 75539 + }, + { + "epoch": 2.9, + "learning_rate": 1.596563221338805e-06, + "loss": 0.3212, + "step": 75540 + }, + { + "epoch": 2.9, + "learning_rate": 1.5959220312900743e-06, + "loss": 0.298, + "step": 75541 + }, + { + "epoch": 2.9, + "learning_rate": 1.5952808412413439e-06, + "loss": 0.2661, + "step": 75542 + }, + { + "epoch": 2.9, + "learning_rate": 1.5946396511926135e-06, + "loss": 0.2486, + "step": 75543 + }, + { + "epoch": 2.9, + "learning_rate": 1.593998461143883e-06, + "loss": 0.1297, + "step": 75544 + }, + { + "epoch": 2.9, + "learning_rate": 1.5933572710951527e-06, + "loss": 0.1048, + "step": 75545 + }, + { + "epoch": 2.9, + "learning_rate": 1.5927160810464223e-06, + "loss": 0.3669, + "step": 75546 + }, + { + "epoch": 2.9, + "learning_rate": 1.5920748909976919e-06, + "loss": 0.1119, + "step": 75547 + }, + { + "epoch": 2.9, + "learning_rate": 1.5914337009489613e-06, + "loss": 0.2486, + "step": 75548 + }, + { + "epoch": 2.9, + "learning_rate": 1.5907925109002309e-06, + "loss": 0.1822, + "step": 75549 + }, + { + "epoch": 2.9, + "learning_rate": 1.5901513208515005e-06, + "loss": 0.2388, + "step": 75550 + }, + { + "epoch": 2.9, + "learning_rate": 1.5895101308027699e-06, + "loss": 0.2958, + "step": 75551 + }, + { + "epoch": 2.9, + "learning_rate": 1.5888689407540395e-06, + "loss": 0.1855, + "step": 75552 + }, + { + "epoch": 2.9, + "learning_rate": 1.588227750705309e-06, + "loss": 0.2418, + "step": 75553 + }, + { + "epoch": 2.9, + "learning_rate": 1.5875865606565787e-06, + "loss": 0.3183, + "step": 75554 + }, + { + "epoch": 2.9, + "learning_rate": 1.5869453706078483e-06, + "loss": 0.2181, + "step": 75555 + }, + { + "epoch": 2.9, + "learning_rate": 1.5863041805591179e-06, + "loss": 0.1835, + "step": 75556 + }, + { + "epoch": 2.9, + "learning_rate": 1.5856629905103875e-06, + "loss": 0.1165, + "step": 75557 + }, + { + "epoch": 2.9, + "learning_rate": 1.5850218004616568e-06, + "loss": 0.1389, + "step": 75558 + }, + { + "epoch": 2.9, + "learning_rate": 1.5843806104129264e-06, + "loss": 0.3385, + "step": 75559 + }, + { + "epoch": 2.91, + "learning_rate": 1.583739420364196e-06, + "loss": 0.131, + "step": 75560 + }, + { + "epoch": 2.91, + "learning_rate": 1.5830982303154656e-06, + "loss": 0.2194, + "step": 75561 + }, + { + "epoch": 2.91, + "learning_rate": 1.582457040266735e-06, + "loss": 0.1757, + "step": 75562 + }, + { + "epoch": 2.91, + "learning_rate": 1.5818158502180046e-06, + "loss": 0.2462, + "step": 75563 + }, + { + "epoch": 2.91, + "learning_rate": 1.5811746601692742e-06, + "loss": 0.1366, + "step": 75564 + }, + { + "epoch": 2.91, + "learning_rate": 1.5805334701205438e-06, + "loss": 0.2427, + "step": 75565 + }, + { + "epoch": 2.91, + "learning_rate": 1.5798922800718134e-06, + "loss": 0.1873, + "step": 75566 + }, + { + "epoch": 2.91, + "learning_rate": 1.579251090023083e-06, + "loss": 0.154, + "step": 75567 + }, + { + "epoch": 2.91, + "learning_rate": 1.5786098999743524e-06, + "loss": 0.2034, + "step": 75568 + }, + { + "epoch": 2.91, + "learning_rate": 1.577968709925622e-06, + "loss": 0.2093, + "step": 75569 + }, + { + "epoch": 2.91, + "learning_rate": 1.5773275198768916e-06, + "loss": 0.109, + "step": 75570 + }, + { + "epoch": 2.91, + "learning_rate": 1.5766863298281612e-06, + "loss": 0.2269, + "step": 75571 + }, + { + "epoch": 2.91, + "learning_rate": 1.5760451397794306e-06, + "loss": 0.111, + "step": 75572 + }, + { + "epoch": 2.91, + "learning_rate": 1.5754039497307002e-06, + "loss": 0.1378, + "step": 75573 + }, + { + "epoch": 2.91, + "learning_rate": 1.5747627596819698e-06, + "loss": 0.1942, + "step": 75574 + }, + { + "epoch": 2.91, + "learning_rate": 1.5741215696332394e-06, + "loss": 0.1691, + "step": 75575 + }, + { + "epoch": 2.91, + "learning_rate": 1.573480379584509e-06, + "loss": 0.1851, + "step": 75576 + }, + { + "epoch": 2.91, + "learning_rate": 1.5728391895357786e-06, + "loss": 0.2768, + "step": 75577 + }, + { + "epoch": 2.91, + "learning_rate": 1.572197999487048e-06, + "loss": 0.2466, + "step": 75578 + }, + { + "epoch": 2.91, + "learning_rate": 1.5715568094383176e-06, + "loss": 0.2859, + "step": 75579 + }, + { + "epoch": 2.91, + "learning_rate": 1.5709156193895872e-06, + "loss": 0.1526, + "step": 75580 + }, + { + "epoch": 2.91, + "learning_rate": 1.5702744293408568e-06, + "loss": 0.1167, + "step": 75581 + }, + { + "epoch": 2.91, + "learning_rate": 1.5696332392921262e-06, + "loss": 0.2747, + "step": 75582 + }, + { + "epoch": 2.91, + "learning_rate": 1.5689920492433958e-06, + "loss": 0.2646, + "step": 75583 + }, + { + "epoch": 2.91, + "learning_rate": 1.5683508591946654e-06, + "loss": 0.1304, + "step": 75584 + }, + { + "epoch": 2.91, + "learning_rate": 1.567709669145935e-06, + "loss": 0.0796, + "step": 75585 + }, + { + "epoch": 2.91, + "learning_rate": 1.5670684790972046e-06, + "loss": 0.2224, + "step": 75586 + }, + { + "epoch": 2.91, + "learning_rate": 1.566427289048474e-06, + "loss": 0.2588, + "step": 75587 + }, + { + "epoch": 2.91, + "learning_rate": 1.5657860989997435e-06, + "loss": 0.2456, + "step": 75588 + }, + { + "epoch": 2.91, + "learning_rate": 1.5651449089510131e-06, + "loss": 0.2141, + "step": 75589 + }, + { + "epoch": 2.91, + "learning_rate": 1.5645037189022827e-06, + "loss": 0.1963, + "step": 75590 + }, + { + "epoch": 2.91, + "learning_rate": 1.5638625288535523e-06, + "loss": 0.1127, + "step": 75591 + }, + { + "epoch": 2.91, + "learning_rate": 1.5632213388048217e-06, + "loss": 0.2156, + "step": 75592 + }, + { + "epoch": 2.91, + "learning_rate": 1.5625801487560913e-06, + "loss": 0.1734, + "step": 75593 + }, + { + "epoch": 2.91, + "learning_rate": 1.561938958707361e-06, + "loss": 0.241, + "step": 75594 + }, + { + "epoch": 2.91, + "learning_rate": 1.5612977686586305e-06, + "loss": 0.2256, + "step": 75595 + }, + { + "epoch": 2.91, + "learning_rate": 1.5606565786099001e-06, + "loss": 0.1811, + "step": 75596 + }, + { + "epoch": 2.91, + "learning_rate": 1.5600153885611695e-06, + "loss": 0.2378, + "step": 75597 + }, + { + "epoch": 2.91, + "learning_rate": 1.559374198512439e-06, + "loss": 0.1685, + "step": 75598 + }, + { + "epoch": 2.91, + "learning_rate": 1.5587330084637087e-06, + "loss": 0.1582, + "step": 75599 + }, + { + "epoch": 2.91, + "learning_rate": 1.5580918184149783e-06, + "loss": 0.1485, + "step": 75600 + }, + { + "epoch": 2.91, + "learning_rate": 1.557450628366248e-06, + "loss": 0.1033, + "step": 75601 + }, + { + "epoch": 2.91, + "learning_rate": 1.5568094383175175e-06, + "loss": 0.1171, + "step": 75602 + }, + { + "epoch": 2.91, + "learning_rate": 1.5561682482687869e-06, + "loss": 0.1994, + "step": 75603 + }, + { + "epoch": 2.91, + "learning_rate": 1.5555270582200565e-06, + "loss": 0.2674, + "step": 75604 + }, + { + "epoch": 2.91, + "learning_rate": 1.554885868171326e-06, + "loss": 0.3148, + "step": 75605 + }, + { + "epoch": 2.91, + "learning_rate": 1.5542446781225957e-06, + "loss": 0.4031, + "step": 75606 + }, + { + "epoch": 2.91, + "learning_rate": 1.5536034880738653e-06, + "loss": 0.124, + "step": 75607 + }, + { + "epoch": 2.91, + "learning_rate": 1.5529622980251347e-06, + "loss": 0.3529, + "step": 75608 + }, + { + "epoch": 2.91, + "learning_rate": 1.5523211079764043e-06, + "loss": 0.2177, + "step": 75609 + }, + { + "epoch": 2.91, + "learning_rate": 1.5516799179276739e-06, + "loss": 0.177, + "step": 75610 + }, + { + "epoch": 2.91, + "learning_rate": 1.5510387278789435e-06, + "loss": 0.2383, + "step": 75611 + }, + { + "epoch": 2.91, + "learning_rate": 1.550397537830213e-06, + "loss": 0.1883, + "step": 75612 + }, + { + "epoch": 2.91, + "learning_rate": 1.5497563477814825e-06, + "loss": 0.2409, + "step": 75613 + }, + { + "epoch": 2.91, + "learning_rate": 1.549115157732752e-06, + "loss": 0.1924, + "step": 75614 + }, + { + "epoch": 2.91, + "learning_rate": 1.5484739676840217e-06, + "loss": 0.1397, + "step": 75615 + }, + { + "epoch": 2.91, + "learning_rate": 1.5478327776352913e-06, + "loss": 0.2318, + "step": 75616 + }, + { + "epoch": 2.91, + "learning_rate": 1.5471915875865609e-06, + "loss": 0.2526, + "step": 75617 + }, + { + "epoch": 2.91, + "learning_rate": 1.5465503975378302e-06, + "loss": 0.2829, + "step": 75618 + }, + { + "epoch": 2.91, + "learning_rate": 1.5459092074890998e-06, + "loss": 0.1262, + "step": 75619 + }, + { + "epoch": 2.91, + "learning_rate": 1.5452680174403694e-06, + "loss": 0.2515, + "step": 75620 + }, + { + "epoch": 2.91, + "learning_rate": 1.544626827391639e-06, + "loss": 0.213, + "step": 75621 + }, + { + "epoch": 2.91, + "learning_rate": 1.5439856373429086e-06, + "loss": 0.1685, + "step": 75622 + }, + { + "epoch": 2.91, + "learning_rate": 1.543344447294178e-06, + "loss": 0.1907, + "step": 75623 + }, + { + "epoch": 2.91, + "learning_rate": 1.5427032572454476e-06, + "loss": 0.2518, + "step": 75624 + }, + { + "epoch": 2.91, + "learning_rate": 1.5420620671967172e-06, + "loss": 0.1888, + "step": 75625 + }, + { + "epoch": 2.91, + "learning_rate": 1.5414208771479868e-06, + "loss": 0.3088, + "step": 75626 + }, + { + "epoch": 2.91, + "learning_rate": 1.5407796870992564e-06, + "loss": 0.2387, + "step": 75627 + }, + { + "epoch": 2.91, + "learning_rate": 1.5401384970505258e-06, + "loss": 0.1591, + "step": 75628 + }, + { + "epoch": 2.91, + "learning_rate": 1.5394973070017954e-06, + "loss": 0.2512, + "step": 75629 + }, + { + "epoch": 2.91, + "learning_rate": 1.538856116953065e-06, + "loss": 0.2281, + "step": 75630 + }, + { + "epoch": 2.91, + "learning_rate": 1.5382149269043346e-06, + "loss": 0.177, + "step": 75631 + }, + { + "epoch": 2.91, + "learning_rate": 1.5375737368556042e-06, + "loss": 0.196, + "step": 75632 + }, + { + "epoch": 2.91, + "learning_rate": 1.5369325468068736e-06, + "loss": 0.1861, + "step": 75633 + }, + { + "epoch": 2.91, + "learning_rate": 1.5362913567581432e-06, + "loss": 0.1611, + "step": 75634 + }, + { + "epoch": 2.91, + "learning_rate": 1.5356501667094128e-06, + "loss": 0.1417, + "step": 75635 + }, + { + "epoch": 2.91, + "learning_rate": 1.5350089766606824e-06, + "loss": 0.1258, + "step": 75636 + }, + { + "epoch": 2.91, + "learning_rate": 1.534367786611952e-06, + "loss": 0.1877, + "step": 75637 + }, + { + "epoch": 2.91, + "learning_rate": 1.5337265965632214e-06, + "loss": 0.2173, + "step": 75638 + }, + { + "epoch": 2.91, + "learning_rate": 1.533085406514491e-06, + "loss": 0.2305, + "step": 75639 + }, + { + "epoch": 2.91, + "learning_rate": 1.5324442164657606e-06, + "loss": 0.0802, + "step": 75640 + }, + { + "epoch": 2.91, + "learning_rate": 1.5318030264170302e-06, + "loss": 0.299, + "step": 75641 + }, + { + "epoch": 2.91, + "learning_rate": 1.5311618363682998e-06, + "loss": 0.1362, + "step": 75642 + }, + { + "epoch": 2.91, + "learning_rate": 1.5305206463195692e-06, + "loss": 0.1476, + "step": 75643 + }, + { + "epoch": 2.91, + "learning_rate": 1.5298794562708388e-06, + "loss": 0.1104, + "step": 75644 + }, + { + "epoch": 2.91, + "learning_rate": 1.5292382662221084e-06, + "loss": 0.1432, + "step": 75645 + }, + { + "epoch": 2.91, + "learning_rate": 1.528597076173378e-06, + "loss": 0.1505, + "step": 75646 + }, + { + "epoch": 2.91, + "learning_rate": 1.5279558861246476e-06, + "loss": 0.1765, + "step": 75647 + }, + { + "epoch": 2.91, + "learning_rate": 1.5273146960759172e-06, + "loss": 0.1764, + "step": 75648 + }, + { + "epoch": 2.91, + "learning_rate": 1.5266735060271865e-06, + "loss": 0.1854, + "step": 75649 + }, + { + "epoch": 2.91, + "learning_rate": 1.5260323159784561e-06, + "loss": 0.1807, + "step": 75650 + }, + { + "epoch": 2.91, + "learning_rate": 1.5253911259297257e-06, + "loss": 0.1829, + "step": 75651 + }, + { + "epoch": 2.91, + "learning_rate": 1.5247499358809953e-06, + "loss": 0.1084, + "step": 75652 + }, + { + "epoch": 2.91, + "learning_rate": 1.524108745832265e-06, + "loss": 0.1316, + "step": 75653 + }, + { + "epoch": 2.91, + "learning_rate": 1.5234675557835343e-06, + "loss": 0.1664, + "step": 75654 + }, + { + "epoch": 2.91, + "learning_rate": 1.522826365734804e-06, + "loss": 0.1745, + "step": 75655 + }, + { + "epoch": 2.91, + "learning_rate": 1.5221851756860735e-06, + "loss": 0.1112, + "step": 75656 + }, + { + "epoch": 2.91, + "learning_rate": 1.5215439856373431e-06, + "loss": 0.1454, + "step": 75657 + }, + { + "epoch": 2.91, + "learning_rate": 1.5209027955886127e-06, + "loss": 0.3178, + "step": 75658 + }, + { + "epoch": 2.91, + "learning_rate": 1.5202616055398821e-06, + "loss": 0.1306, + "step": 75659 + }, + { + "epoch": 2.91, + "learning_rate": 1.5196204154911517e-06, + "loss": 0.1912, + "step": 75660 + }, + { + "epoch": 2.91, + "learning_rate": 1.5189792254424213e-06, + "loss": 0.235, + "step": 75661 + }, + { + "epoch": 2.91, + "learning_rate": 1.518338035393691e-06, + "loss": 0.1382, + "step": 75662 + }, + { + "epoch": 2.91, + "learning_rate": 1.5176968453449605e-06, + "loss": 0.301, + "step": 75663 + }, + { + "epoch": 2.91, + "learning_rate": 1.5170556552962299e-06, + "loss": 0.1363, + "step": 75664 + }, + { + "epoch": 2.91, + "learning_rate": 1.5164144652474995e-06, + "loss": 0.0709, + "step": 75665 + }, + { + "epoch": 2.91, + "learning_rate": 1.515773275198769e-06, + "loss": 0.1906, + "step": 75666 + }, + { + "epoch": 2.91, + "learning_rate": 1.5151320851500387e-06, + "loss": 0.2727, + "step": 75667 + }, + { + "epoch": 2.91, + "learning_rate": 1.5144908951013083e-06, + "loss": 0.1371, + "step": 75668 + }, + { + "epoch": 2.91, + "learning_rate": 1.5138497050525777e-06, + "loss": 0.2528, + "step": 75669 + }, + { + "epoch": 2.91, + "learning_rate": 1.5132085150038473e-06, + "loss": 0.2074, + "step": 75670 + }, + { + "epoch": 2.91, + "learning_rate": 1.5125673249551169e-06, + "loss": 0.2526, + "step": 75671 + }, + { + "epoch": 2.91, + "learning_rate": 1.5119261349063865e-06, + "loss": 0.2392, + "step": 75672 + }, + { + "epoch": 2.91, + "learning_rate": 1.511284944857656e-06, + "loss": 0.0951, + "step": 75673 + }, + { + "epoch": 2.91, + "learning_rate": 1.5106437548089255e-06, + "loss": 0.2814, + "step": 75674 + }, + { + "epoch": 2.91, + "learning_rate": 1.510002564760195e-06, + "loss": 0.2098, + "step": 75675 + }, + { + "epoch": 2.91, + "learning_rate": 1.5093613747114647e-06, + "loss": 0.1931, + "step": 75676 + }, + { + "epoch": 2.91, + "learning_rate": 1.5087201846627343e-06, + "loss": 0.2181, + "step": 75677 + }, + { + "epoch": 2.91, + "learning_rate": 1.5080789946140036e-06, + "loss": 0.1545, + "step": 75678 + }, + { + "epoch": 2.91, + "learning_rate": 1.5074378045652732e-06, + "loss": 0.339, + "step": 75679 + }, + { + "epoch": 2.91, + "learning_rate": 1.5067966145165428e-06, + "loss": 0.2042, + "step": 75680 + }, + { + "epoch": 2.91, + "learning_rate": 1.5061554244678124e-06, + "loss": 0.3196, + "step": 75681 + }, + { + "epoch": 2.91, + "learning_rate": 1.505514234419082e-06, + "loss": 0.2585, + "step": 75682 + }, + { + "epoch": 2.91, + "learning_rate": 1.5048730443703514e-06, + "loss": 0.1619, + "step": 75683 + }, + { + "epoch": 2.91, + "learning_rate": 1.504231854321621e-06, + "loss": 0.176, + "step": 75684 + }, + { + "epoch": 2.91, + "learning_rate": 1.5035906642728906e-06, + "loss": 0.114, + "step": 75685 + }, + { + "epoch": 2.91, + "learning_rate": 1.5029494742241602e-06, + "loss": 0.2872, + "step": 75686 + }, + { + "epoch": 2.91, + "learning_rate": 1.5023082841754298e-06, + "loss": 0.3153, + "step": 75687 + }, + { + "epoch": 2.91, + "learning_rate": 1.5016670941266992e-06, + "loss": 0.0663, + "step": 75688 + }, + { + "epoch": 2.91, + "learning_rate": 1.5010259040779688e-06, + "loss": 0.2053, + "step": 75689 + }, + { + "epoch": 2.91, + "learning_rate": 1.5003847140292384e-06, + "loss": 0.2146, + "step": 75690 + }, + { + "epoch": 2.91, + "learning_rate": 1.499743523980508e-06, + "loss": 0.2386, + "step": 75691 + }, + { + "epoch": 2.91, + "learning_rate": 1.4991023339317776e-06, + "loss": 0.1602, + "step": 75692 + }, + { + "epoch": 2.91, + "learning_rate": 1.498461143883047e-06, + "loss": 0.249, + "step": 75693 + }, + { + "epoch": 2.91, + "learning_rate": 1.4978199538343166e-06, + "loss": 0.2343, + "step": 75694 + }, + { + "epoch": 2.91, + "learning_rate": 1.4971787637855862e-06, + "loss": 0.1725, + "step": 75695 + }, + { + "epoch": 2.91, + "learning_rate": 1.4965375737368558e-06, + "loss": 0.1313, + "step": 75696 + }, + { + "epoch": 2.91, + "learning_rate": 1.4958963836881254e-06, + "loss": 0.1915, + "step": 75697 + }, + { + "epoch": 2.91, + "learning_rate": 1.4952551936393948e-06, + "loss": 0.2646, + "step": 75698 + }, + { + "epoch": 2.91, + "learning_rate": 1.4946140035906644e-06, + "loss": 0.2167, + "step": 75699 + }, + { + "epoch": 2.91, + "learning_rate": 1.493972813541934e-06, + "loss": 0.1868, + "step": 75700 + }, + { + "epoch": 2.91, + "learning_rate": 1.4933316234932036e-06, + "loss": 0.2817, + "step": 75701 + }, + { + "epoch": 2.91, + "learning_rate": 1.4926904334444732e-06, + "loss": 0.4301, + "step": 75702 + }, + { + "epoch": 2.91, + "learning_rate": 1.4920492433957426e-06, + "loss": 0.1735, + "step": 75703 + }, + { + "epoch": 2.91, + "learning_rate": 1.4914080533470122e-06, + "loss": 0.1384, + "step": 75704 + }, + { + "epoch": 2.91, + "learning_rate": 1.4907668632982818e-06, + "loss": 0.211, + "step": 75705 + }, + { + "epoch": 2.91, + "learning_rate": 1.4901256732495514e-06, + "loss": 0.3148, + "step": 75706 + }, + { + "epoch": 2.91, + "learning_rate": 1.489484483200821e-06, + "loss": 0.2381, + "step": 75707 + }, + { + "epoch": 2.91, + "learning_rate": 1.4888432931520903e-06, + "loss": 0.3305, + "step": 75708 + }, + { + "epoch": 2.91, + "learning_rate": 1.48820210310336e-06, + "loss": 0.147, + "step": 75709 + }, + { + "epoch": 2.91, + "learning_rate": 1.4875609130546295e-06, + "loss": 0.1705, + "step": 75710 + }, + { + "epoch": 2.91, + "learning_rate": 1.4869197230058991e-06, + "loss": 0.2715, + "step": 75711 + }, + { + "epoch": 2.91, + "learning_rate": 1.4862785329571687e-06, + "loss": 0.2485, + "step": 75712 + }, + { + "epoch": 2.91, + "learning_rate": 1.4856373429084381e-06, + "loss": 0.1596, + "step": 75713 + }, + { + "epoch": 2.91, + "learning_rate": 1.4849961528597077e-06, + "loss": 0.1972, + "step": 75714 + }, + { + "epoch": 2.91, + "learning_rate": 1.4843549628109773e-06, + "loss": 0.1217, + "step": 75715 + }, + { + "epoch": 2.91, + "learning_rate": 1.483713772762247e-06, + "loss": 0.2414, + "step": 75716 + }, + { + "epoch": 2.91, + "learning_rate": 1.4830725827135165e-06, + "loss": 0.08, + "step": 75717 + }, + { + "epoch": 2.91, + "learning_rate": 1.482431392664786e-06, + "loss": 0.1832, + "step": 75718 + }, + { + "epoch": 2.91, + "learning_rate": 1.4817902026160555e-06, + "loss": 0.2337, + "step": 75719 + }, + { + "epoch": 2.91, + "learning_rate": 1.4811490125673251e-06, + "loss": 0.1195, + "step": 75720 + }, + { + "epoch": 2.91, + "learning_rate": 1.4805078225185947e-06, + "loss": 0.2294, + "step": 75721 + }, + { + "epoch": 2.91, + "learning_rate": 1.4798666324698643e-06, + "loss": 0.2579, + "step": 75722 + }, + { + "epoch": 2.91, + "learning_rate": 1.4792254424211337e-06, + "loss": 0.1342, + "step": 75723 + }, + { + "epoch": 2.91, + "learning_rate": 1.4785842523724033e-06, + "loss": 0.284, + "step": 75724 + }, + { + "epoch": 2.91, + "learning_rate": 1.4779430623236729e-06, + "loss": 0.2417, + "step": 75725 + }, + { + "epoch": 2.91, + "learning_rate": 1.4773018722749425e-06, + "loss": 0.1759, + "step": 75726 + }, + { + "epoch": 2.91, + "learning_rate": 1.476660682226212e-06, + "loss": 0.2629, + "step": 75727 + }, + { + "epoch": 2.91, + "learning_rate": 1.4760194921774815e-06, + "loss": 0.3232, + "step": 75728 + }, + { + "epoch": 2.91, + "learning_rate": 1.475378302128751e-06, + "loss": 0.1161, + "step": 75729 + }, + { + "epoch": 2.91, + "learning_rate": 1.4747371120800207e-06, + "loss": 0.2147, + "step": 75730 + }, + { + "epoch": 2.91, + "learning_rate": 1.4740959220312903e-06, + "loss": 0.28, + "step": 75731 + }, + { + "epoch": 2.91, + "learning_rate": 1.4734547319825597e-06, + "loss": 0.1669, + "step": 75732 + }, + { + "epoch": 2.91, + "learning_rate": 1.4728135419338293e-06, + "loss": 0.2347, + "step": 75733 + }, + { + "epoch": 2.91, + "learning_rate": 1.4721723518850989e-06, + "loss": 0.2247, + "step": 75734 + }, + { + "epoch": 2.91, + "learning_rate": 1.4715311618363685e-06, + "loss": 0.1257, + "step": 75735 + }, + { + "epoch": 2.91, + "learning_rate": 1.470889971787638e-06, + "loss": 0.2371, + "step": 75736 + }, + { + "epoch": 2.91, + "learning_rate": 1.4702487817389074e-06, + "loss": 0.2528, + "step": 75737 + }, + { + "epoch": 2.91, + "learning_rate": 1.469607591690177e-06, + "loss": 0.2036, + "step": 75738 + }, + { + "epoch": 2.91, + "learning_rate": 1.4689664016414466e-06, + "loss": 0.265, + "step": 75739 + }, + { + "epoch": 2.91, + "learning_rate": 1.4683252115927162e-06, + "loss": 0.1951, + "step": 75740 + }, + { + "epoch": 2.91, + "learning_rate": 1.4676840215439858e-06, + "loss": 0.1153, + "step": 75741 + }, + { + "epoch": 2.91, + "learning_rate": 1.4670428314952552e-06, + "loss": 0.1827, + "step": 75742 + }, + { + "epoch": 2.91, + "learning_rate": 1.4664016414465248e-06, + "loss": 0.2093, + "step": 75743 + }, + { + "epoch": 2.91, + "learning_rate": 1.4657604513977944e-06, + "loss": 0.154, + "step": 75744 + }, + { + "epoch": 2.91, + "learning_rate": 1.465119261349064e-06, + "loss": 0.1778, + "step": 75745 + }, + { + "epoch": 2.91, + "learning_rate": 1.4644780713003336e-06, + "loss": 0.23, + "step": 75746 + }, + { + "epoch": 2.91, + "learning_rate": 1.463836881251603e-06, + "loss": 0.2147, + "step": 75747 + }, + { + "epoch": 2.91, + "learning_rate": 1.4631956912028726e-06, + "loss": 0.2359, + "step": 75748 + }, + { + "epoch": 2.91, + "learning_rate": 1.4625545011541422e-06, + "loss": 0.4189, + "step": 75749 + }, + { + "epoch": 2.91, + "learning_rate": 1.4619133111054118e-06, + "loss": 0.2007, + "step": 75750 + }, + { + "epoch": 2.91, + "learning_rate": 1.4612721210566814e-06, + "loss": 0.2279, + "step": 75751 + }, + { + "epoch": 2.91, + "learning_rate": 1.4606309310079508e-06, + "loss": 0.2802, + "step": 75752 + }, + { + "epoch": 2.91, + "learning_rate": 1.4599897409592204e-06, + "loss": 0.2183, + "step": 75753 + }, + { + "epoch": 2.91, + "learning_rate": 1.45934855091049e-06, + "loss": 0.349, + "step": 75754 + }, + { + "epoch": 2.91, + "learning_rate": 1.4587073608617596e-06, + "loss": 0.1482, + "step": 75755 + }, + { + "epoch": 2.91, + "learning_rate": 1.4580661708130292e-06, + "loss": 0.1177, + "step": 75756 + }, + { + "epoch": 2.91, + "learning_rate": 1.4574249807642986e-06, + "loss": 0.1702, + "step": 75757 + }, + { + "epoch": 2.91, + "learning_rate": 1.4567837907155682e-06, + "loss": 0.1823, + "step": 75758 + }, + { + "epoch": 2.91, + "learning_rate": 1.4561426006668378e-06, + "loss": 0.1795, + "step": 75759 + }, + { + "epoch": 2.91, + "learning_rate": 1.4555014106181074e-06, + "loss": 0.2297, + "step": 75760 + }, + { + "epoch": 2.91, + "learning_rate": 1.454860220569377e-06, + "loss": 0.2395, + "step": 75761 + }, + { + "epoch": 2.91, + "learning_rate": 1.4542190305206464e-06, + "loss": 0.2029, + "step": 75762 + }, + { + "epoch": 2.91, + "learning_rate": 1.453577840471916e-06, + "loss": 0.2322, + "step": 75763 + }, + { + "epoch": 2.91, + "learning_rate": 1.4529366504231856e-06, + "loss": 0.215, + "step": 75764 + }, + { + "epoch": 2.91, + "learning_rate": 1.4522954603744552e-06, + "loss": 0.12, + "step": 75765 + }, + { + "epoch": 2.91, + "learning_rate": 1.4516542703257248e-06, + "loss": 0.3365, + "step": 75766 + }, + { + "epoch": 2.91, + "learning_rate": 1.4510130802769941e-06, + "loss": 0.1267, + "step": 75767 + }, + { + "epoch": 2.91, + "learning_rate": 1.4503718902282637e-06, + "loss": 0.2535, + "step": 75768 + }, + { + "epoch": 2.91, + "learning_rate": 1.4497307001795333e-06, + "loss": 0.15, + "step": 75769 + }, + { + "epoch": 2.91, + "learning_rate": 1.449089510130803e-06, + "loss": 0.2245, + "step": 75770 + }, + { + "epoch": 2.91, + "learning_rate": 1.4484483200820725e-06, + "loss": 0.2416, + "step": 75771 + }, + { + "epoch": 2.91, + "learning_rate": 1.447807130033342e-06, + "loss": 0.2532, + "step": 75772 + }, + { + "epoch": 2.91, + "learning_rate": 1.4471659399846115e-06, + "loss": 0.1798, + "step": 75773 + }, + { + "epoch": 2.91, + "learning_rate": 1.4465247499358811e-06, + "loss": 0.1927, + "step": 75774 + }, + { + "epoch": 2.91, + "learning_rate": 1.4458835598871507e-06, + "loss": 0.2896, + "step": 75775 + }, + { + "epoch": 2.91, + "learning_rate": 1.4452423698384203e-06, + "loss": 0.1663, + "step": 75776 + }, + { + "epoch": 2.91, + "learning_rate": 1.4446011797896897e-06, + "loss": 0.1787, + "step": 75777 + }, + { + "epoch": 2.91, + "learning_rate": 1.4439599897409593e-06, + "loss": 0.2149, + "step": 75778 + }, + { + "epoch": 2.91, + "learning_rate": 1.443318799692229e-06, + "loss": 0.1864, + "step": 75779 + }, + { + "epoch": 2.91, + "learning_rate": 1.4426776096434985e-06, + "loss": 0.2978, + "step": 75780 + }, + { + "epoch": 2.91, + "learning_rate": 1.4420364195947681e-06, + "loss": 0.2079, + "step": 75781 + }, + { + "epoch": 2.91, + "learning_rate": 1.4413952295460375e-06, + "loss": 0.2301, + "step": 75782 + }, + { + "epoch": 2.91, + "learning_rate": 1.440754039497307e-06, + "loss": 0.113, + "step": 75783 + }, + { + "epoch": 2.91, + "learning_rate": 1.4401128494485767e-06, + "loss": 0.1477, + "step": 75784 + }, + { + "epoch": 2.91, + "learning_rate": 1.4394716593998463e-06, + "loss": 0.2328, + "step": 75785 + }, + { + "epoch": 2.91, + "learning_rate": 1.4388304693511159e-06, + "loss": 0.2531, + "step": 75786 + }, + { + "epoch": 2.91, + "learning_rate": 1.4381892793023853e-06, + "loss": 0.2525, + "step": 75787 + }, + { + "epoch": 2.91, + "learning_rate": 1.4375480892536549e-06, + "loss": 0.2943, + "step": 75788 + }, + { + "epoch": 2.91, + "learning_rate": 1.4369068992049245e-06, + "loss": 0.192, + "step": 75789 + }, + { + "epoch": 2.91, + "learning_rate": 1.436265709156194e-06, + "loss": 0.302, + "step": 75790 + }, + { + "epoch": 2.91, + "learning_rate": 1.4356245191074635e-06, + "loss": 0.1548, + "step": 75791 + }, + { + "epoch": 2.91, + "learning_rate": 1.434983329058733e-06, + "loss": 0.1483, + "step": 75792 + }, + { + "epoch": 2.91, + "learning_rate": 1.4343421390100027e-06, + "loss": 0.1382, + "step": 75793 + }, + { + "epoch": 2.91, + "learning_rate": 1.4337009489612723e-06, + "loss": 0.2282, + "step": 75794 + }, + { + "epoch": 2.91, + "learning_rate": 1.4330597589125419e-06, + "loss": 0.2178, + "step": 75795 + }, + { + "epoch": 2.91, + "learning_rate": 1.4324185688638112e-06, + "loss": 0.2217, + "step": 75796 + }, + { + "epoch": 2.91, + "learning_rate": 1.4317773788150808e-06, + "loss": 0.2415, + "step": 75797 + }, + { + "epoch": 2.91, + "learning_rate": 1.4311361887663504e-06, + "loss": 0.2539, + "step": 75798 + }, + { + "epoch": 2.91, + "learning_rate": 1.43049499871762e-06, + "loss": 0.3151, + "step": 75799 + }, + { + "epoch": 2.91, + "learning_rate": 1.4298538086688896e-06, + "loss": 0.2453, + "step": 75800 + }, + { + "epoch": 2.91, + "learning_rate": 1.429212618620159e-06, + "loss": 0.1875, + "step": 75801 + }, + { + "epoch": 2.91, + "learning_rate": 1.4285714285714286e-06, + "loss": 0.2516, + "step": 75802 + }, + { + "epoch": 2.91, + "learning_rate": 1.4279302385226982e-06, + "loss": 0.4215, + "step": 75803 + }, + { + "epoch": 2.91, + "learning_rate": 1.4272890484739678e-06, + "loss": 0.2896, + "step": 75804 + }, + { + "epoch": 2.91, + "learning_rate": 1.4266478584252374e-06, + "loss": 0.3407, + "step": 75805 + }, + { + "epoch": 2.91, + "learning_rate": 1.4260066683765068e-06, + "loss": 0.3557, + "step": 75806 + }, + { + "epoch": 2.91, + "learning_rate": 1.4253654783277764e-06, + "loss": 0.1744, + "step": 75807 + }, + { + "epoch": 2.91, + "learning_rate": 1.424724288279046e-06, + "loss": 0.2192, + "step": 75808 + }, + { + "epoch": 2.91, + "learning_rate": 1.4240830982303156e-06, + "loss": 0.2329, + "step": 75809 + }, + { + "epoch": 2.91, + "learning_rate": 1.4234419081815852e-06, + "loss": 0.2616, + "step": 75810 + }, + { + "epoch": 2.91, + "learning_rate": 1.4228007181328546e-06, + "loss": 0.1889, + "step": 75811 + }, + { + "epoch": 2.91, + "learning_rate": 1.4221595280841242e-06, + "loss": 0.1624, + "step": 75812 + }, + { + "epoch": 2.91, + "learning_rate": 1.4215183380353938e-06, + "loss": 0.2141, + "step": 75813 + }, + { + "epoch": 2.91, + "learning_rate": 1.4208771479866634e-06, + "loss": 0.1799, + "step": 75814 + }, + { + "epoch": 2.91, + "learning_rate": 1.420235957937933e-06, + "loss": 0.1654, + "step": 75815 + }, + { + "epoch": 2.91, + "learning_rate": 1.4195947678892024e-06, + "loss": 0.0814, + "step": 75816 + }, + { + "epoch": 2.91, + "learning_rate": 1.418953577840472e-06, + "loss": 0.1634, + "step": 75817 + }, + { + "epoch": 2.91, + "learning_rate": 1.4183123877917416e-06, + "loss": 0.2562, + "step": 75818 + }, + { + "epoch": 2.91, + "learning_rate": 1.4176711977430112e-06, + "loss": 0.1889, + "step": 75819 + }, + { + "epoch": 2.92, + "learning_rate": 1.4170300076942808e-06, + "loss": 0.3132, + "step": 75820 + }, + { + "epoch": 2.92, + "learning_rate": 1.4163888176455502e-06, + "loss": 0.2736, + "step": 75821 + }, + { + "epoch": 2.92, + "learning_rate": 1.4157476275968198e-06, + "loss": 0.1808, + "step": 75822 + }, + { + "epoch": 2.92, + "learning_rate": 1.4151064375480894e-06, + "loss": 0.1973, + "step": 75823 + }, + { + "epoch": 2.92, + "learning_rate": 1.414465247499359e-06, + "loss": 0.3138, + "step": 75824 + }, + { + "epoch": 2.92, + "learning_rate": 1.4138240574506286e-06, + "loss": 0.2849, + "step": 75825 + }, + { + "epoch": 2.92, + "learning_rate": 1.413182867401898e-06, + "loss": 0.1528, + "step": 75826 + }, + { + "epoch": 2.92, + "learning_rate": 1.4125416773531675e-06, + "loss": 0.188, + "step": 75827 + }, + { + "epoch": 2.92, + "learning_rate": 1.4119004873044371e-06, + "loss": 0.2124, + "step": 75828 + }, + { + "epoch": 2.92, + "learning_rate": 1.4112592972557067e-06, + "loss": 0.0736, + "step": 75829 + }, + { + "epoch": 2.92, + "learning_rate": 1.4106181072069763e-06, + "loss": 0.1768, + "step": 75830 + }, + { + "epoch": 2.92, + "learning_rate": 1.4099769171582457e-06, + "loss": 0.1626, + "step": 75831 + }, + { + "epoch": 2.92, + "learning_rate": 1.4093357271095153e-06, + "loss": 0.1455, + "step": 75832 + }, + { + "epoch": 2.92, + "learning_rate": 1.408694537060785e-06, + "loss": 0.2984, + "step": 75833 + }, + { + "epoch": 2.92, + "learning_rate": 1.4080533470120545e-06, + "loss": 0.1292, + "step": 75834 + }, + { + "epoch": 2.92, + "learning_rate": 1.4074121569633241e-06, + "loss": 0.1856, + "step": 75835 + }, + { + "epoch": 2.92, + "learning_rate": 1.4067709669145935e-06, + "loss": 0.2203, + "step": 75836 + }, + { + "epoch": 2.92, + "learning_rate": 1.4061297768658631e-06, + "loss": 0.1459, + "step": 75837 + }, + { + "epoch": 2.92, + "learning_rate": 1.4054885868171327e-06, + "loss": 0.2028, + "step": 75838 + }, + { + "epoch": 2.92, + "learning_rate": 1.4048473967684023e-06, + "loss": 0.1495, + "step": 75839 + }, + { + "epoch": 2.92, + "learning_rate": 1.404206206719672e-06, + "loss": 0.2544, + "step": 75840 + }, + { + "epoch": 2.92, + "learning_rate": 1.4035650166709413e-06, + "loss": 0.1962, + "step": 75841 + }, + { + "epoch": 2.92, + "learning_rate": 1.402923826622211e-06, + "loss": 0.2394, + "step": 75842 + }, + { + "epoch": 2.92, + "learning_rate": 1.4022826365734805e-06, + "loss": 0.165, + "step": 75843 + }, + { + "epoch": 2.92, + "learning_rate": 1.40164144652475e-06, + "loss": 0.202, + "step": 75844 + }, + { + "epoch": 2.92, + "learning_rate": 1.4010002564760197e-06, + "loss": 0.235, + "step": 75845 + }, + { + "epoch": 2.92, + "learning_rate": 1.400359066427289e-06, + "loss": 0.3244, + "step": 75846 + }, + { + "epoch": 2.92, + "learning_rate": 1.3997178763785587e-06, + "loss": 0.2167, + "step": 75847 + }, + { + "epoch": 2.92, + "learning_rate": 1.3990766863298283e-06, + "loss": 0.2668, + "step": 75848 + }, + { + "epoch": 2.92, + "learning_rate": 1.3984354962810979e-06, + "loss": 0.1568, + "step": 75849 + }, + { + "epoch": 2.92, + "learning_rate": 1.3977943062323673e-06, + "loss": 0.1773, + "step": 75850 + }, + { + "epoch": 2.92, + "learning_rate": 1.3971531161836369e-06, + "loss": 0.1775, + "step": 75851 + }, + { + "epoch": 2.92, + "learning_rate": 1.3965119261349065e-06, + "loss": 0.1887, + "step": 75852 + }, + { + "epoch": 2.92, + "learning_rate": 1.395870736086176e-06, + "loss": 0.1549, + "step": 75853 + }, + { + "epoch": 2.92, + "learning_rate": 1.3952295460374457e-06, + "loss": 0.2492, + "step": 75854 + }, + { + "epoch": 2.92, + "learning_rate": 1.394588355988715e-06, + "loss": 0.2452, + "step": 75855 + }, + { + "epoch": 2.92, + "learning_rate": 1.3939471659399846e-06, + "loss": 0.1309, + "step": 75856 + }, + { + "epoch": 2.92, + "learning_rate": 1.3933059758912542e-06, + "loss": 0.13, + "step": 75857 + }, + { + "epoch": 2.92, + "learning_rate": 1.3926647858425238e-06, + "loss": 0.1874, + "step": 75858 + }, + { + "epoch": 2.92, + "learning_rate": 1.3920235957937934e-06, + "loss": 0.0618, + "step": 75859 + }, + { + "epoch": 2.92, + "learning_rate": 1.3913824057450628e-06, + "loss": 0.1405, + "step": 75860 + }, + { + "epoch": 2.92, + "learning_rate": 1.3907412156963324e-06, + "loss": 0.2848, + "step": 75861 + }, + { + "epoch": 2.92, + "learning_rate": 1.390100025647602e-06, + "loss": 0.2482, + "step": 75862 + }, + { + "epoch": 2.92, + "learning_rate": 1.3894588355988716e-06, + "loss": 0.1708, + "step": 75863 + }, + { + "epoch": 2.92, + "learning_rate": 1.3888176455501412e-06, + "loss": 0.3129, + "step": 75864 + }, + { + "epoch": 2.92, + "learning_rate": 1.3881764555014106e-06, + "loss": 0.195, + "step": 75865 + }, + { + "epoch": 2.92, + "learning_rate": 1.3875352654526802e-06, + "loss": 0.1523, + "step": 75866 + }, + { + "epoch": 2.92, + "learning_rate": 1.3868940754039498e-06, + "loss": 0.2015, + "step": 75867 + }, + { + "epoch": 2.92, + "learning_rate": 1.3862528853552194e-06, + "loss": 0.1218, + "step": 75868 + }, + { + "epoch": 2.92, + "learning_rate": 1.385611695306489e-06, + "loss": 0.2344, + "step": 75869 + }, + { + "epoch": 2.92, + "learning_rate": 1.3849705052577584e-06, + "loss": 0.2041, + "step": 75870 + }, + { + "epoch": 2.92, + "learning_rate": 1.384329315209028e-06, + "loss": 0.341, + "step": 75871 + }, + { + "epoch": 2.92, + "learning_rate": 1.3836881251602976e-06, + "loss": 0.1505, + "step": 75872 + }, + { + "epoch": 2.92, + "learning_rate": 1.3830469351115672e-06, + "loss": 0.162, + "step": 75873 + }, + { + "epoch": 2.92, + "learning_rate": 1.3824057450628368e-06, + "loss": 0.1843, + "step": 75874 + }, + { + "epoch": 2.92, + "learning_rate": 1.3817645550141062e-06, + "loss": 0.1939, + "step": 75875 + }, + { + "epoch": 2.92, + "learning_rate": 1.3811233649653758e-06, + "loss": 0.1746, + "step": 75876 + }, + { + "epoch": 2.92, + "learning_rate": 1.3804821749166454e-06, + "loss": 0.2373, + "step": 75877 + }, + { + "epoch": 2.92, + "learning_rate": 1.379840984867915e-06, + "loss": 0.2011, + "step": 75878 + }, + { + "epoch": 2.92, + "learning_rate": 1.3791997948191846e-06, + "loss": 0.1287, + "step": 75879 + }, + { + "epoch": 2.92, + "learning_rate": 1.378558604770454e-06, + "loss": 0.1777, + "step": 75880 + }, + { + "epoch": 2.92, + "learning_rate": 1.3779174147217236e-06, + "loss": 0.1998, + "step": 75881 + }, + { + "epoch": 2.92, + "learning_rate": 1.3772762246729932e-06, + "loss": 0.1001, + "step": 75882 + }, + { + "epoch": 2.92, + "learning_rate": 1.3766350346242628e-06, + "loss": 0.2795, + "step": 75883 + }, + { + "epoch": 2.92, + "learning_rate": 1.3759938445755324e-06, + "loss": 0.2108, + "step": 75884 + }, + { + "epoch": 2.92, + "learning_rate": 1.3753526545268017e-06, + "loss": 0.1302, + "step": 75885 + }, + { + "epoch": 2.92, + "learning_rate": 1.3747114644780713e-06, + "loss": 0.1729, + "step": 75886 + }, + { + "epoch": 2.92, + "learning_rate": 1.374070274429341e-06, + "loss": 0.2634, + "step": 75887 + }, + { + "epoch": 2.92, + "learning_rate": 1.3734290843806105e-06, + "loss": 0.157, + "step": 75888 + }, + { + "epoch": 2.92, + "learning_rate": 1.3727878943318801e-06, + "loss": 0.1662, + "step": 75889 + }, + { + "epoch": 2.92, + "learning_rate": 1.3721467042831495e-06, + "loss": 0.3025, + "step": 75890 + }, + { + "epoch": 2.92, + "learning_rate": 1.3715055142344191e-06, + "loss": 0.1872, + "step": 75891 + }, + { + "epoch": 2.92, + "learning_rate": 1.3708643241856887e-06, + "loss": 0.2443, + "step": 75892 + }, + { + "epoch": 2.92, + "learning_rate": 1.3702231341369583e-06, + "loss": 0.1113, + "step": 75893 + }, + { + "epoch": 2.92, + "learning_rate": 1.369581944088228e-06, + "loss": 0.189, + "step": 75894 + }, + { + "epoch": 2.92, + "learning_rate": 1.3689407540394973e-06, + "loss": 0.1283, + "step": 75895 + }, + { + "epoch": 2.92, + "learning_rate": 1.368299563990767e-06, + "loss": 0.14, + "step": 75896 + }, + { + "epoch": 2.92, + "learning_rate": 1.3676583739420365e-06, + "loss": 0.1837, + "step": 75897 + }, + { + "epoch": 2.92, + "learning_rate": 1.3670171838933061e-06, + "loss": 0.2582, + "step": 75898 + }, + { + "epoch": 2.92, + "learning_rate": 1.3663759938445757e-06, + "loss": 0.2291, + "step": 75899 + }, + { + "epoch": 2.92, + "learning_rate": 1.365734803795845e-06, + "loss": 0.2519, + "step": 75900 + }, + { + "epoch": 2.92, + "learning_rate": 1.3650936137471147e-06, + "loss": 0.1458, + "step": 75901 + }, + { + "epoch": 2.92, + "learning_rate": 1.3644524236983843e-06, + "loss": 0.2388, + "step": 75902 + }, + { + "epoch": 2.92, + "learning_rate": 1.363811233649654e-06, + "loss": 0.2188, + "step": 75903 + }, + { + "epoch": 2.92, + "learning_rate": 1.3631700436009235e-06, + "loss": 0.1221, + "step": 75904 + }, + { + "epoch": 2.92, + "learning_rate": 1.3625288535521929e-06, + "loss": 0.2111, + "step": 75905 + }, + { + "epoch": 2.92, + "learning_rate": 1.3618876635034625e-06, + "loss": 0.1625, + "step": 75906 + }, + { + "epoch": 2.92, + "learning_rate": 1.361246473454732e-06, + "loss": 0.1705, + "step": 75907 + }, + { + "epoch": 2.92, + "learning_rate": 1.3606052834060017e-06, + "loss": 0.14, + "step": 75908 + }, + { + "epoch": 2.92, + "learning_rate": 1.359964093357271e-06, + "loss": 0.1763, + "step": 75909 + }, + { + "epoch": 2.92, + "learning_rate": 1.3593229033085407e-06, + "loss": 0.2132, + "step": 75910 + }, + { + "epoch": 2.92, + "learning_rate": 1.3586817132598103e-06, + "loss": 0.3457, + "step": 75911 + }, + { + "epoch": 2.92, + "learning_rate": 1.3580405232110799e-06, + "loss": 0.2262, + "step": 75912 + }, + { + "epoch": 2.92, + "learning_rate": 1.3573993331623495e-06, + "loss": 0.1836, + "step": 75913 + }, + { + "epoch": 2.92, + "learning_rate": 1.3567581431136188e-06, + "loss": 0.1877, + "step": 75914 + }, + { + "epoch": 2.92, + "learning_rate": 1.3561169530648884e-06, + "loss": 0.1359, + "step": 75915 + }, + { + "epoch": 2.92, + "learning_rate": 1.355475763016158e-06, + "loss": 0.1002, + "step": 75916 + }, + { + "epoch": 2.92, + "learning_rate": 1.3548345729674276e-06, + "loss": 0.2372, + "step": 75917 + }, + { + "epoch": 2.92, + "learning_rate": 1.3541933829186972e-06, + "loss": 0.2599, + "step": 75918 + }, + { + "epoch": 2.92, + "learning_rate": 1.3535521928699666e-06, + "loss": 0.1985, + "step": 75919 + }, + { + "epoch": 2.92, + "learning_rate": 1.3529110028212362e-06, + "loss": 0.3126, + "step": 75920 + }, + { + "epoch": 2.92, + "learning_rate": 1.3522698127725058e-06, + "loss": 0.2221, + "step": 75921 + }, + { + "epoch": 2.92, + "learning_rate": 1.3516286227237754e-06, + "loss": 0.3318, + "step": 75922 + }, + { + "epoch": 2.92, + "learning_rate": 1.350987432675045e-06, + "loss": 0.3299, + "step": 75923 + }, + { + "epoch": 2.92, + "learning_rate": 1.3503462426263144e-06, + "loss": 0.2485, + "step": 75924 + }, + { + "epoch": 2.92, + "learning_rate": 1.349705052577584e-06, + "loss": 0.4546, + "step": 75925 + }, + { + "epoch": 2.92, + "learning_rate": 1.3490638625288536e-06, + "loss": 0.2163, + "step": 75926 + }, + { + "epoch": 2.92, + "learning_rate": 1.3484226724801232e-06, + "loss": 0.409, + "step": 75927 + }, + { + "epoch": 2.92, + "learning_rate": 1.3477814824313928e-06, + "loss": 0.2785, + "step": 75928 + }, + { + "epoch": 2.92, + "learning_rate": 1.3471402923826622e-06, + "loss": 0.2471, + "step": 75929 + }, + { + "epoch": 2.92, + "learning_rate": 1.3464991023339318e-06, + "loss": 0.1148, + "step": 75930 + }, + { + "epoch": 2.92, + "learning_rate": 1.3458579122852014e-06, + "loss": 0.2435, + "step": 75931 + }, + { + "epoch": 2.92, + "learning_rate": 1.345216722236471e-06, + "loss": 0.2098, + "step": 75932 + }, + { + "epoch": 2.92, + "learning_rate": 1.3445755321877406e-06, + "loss": 0.2332, + "step": 75933 + }, + { + "epoch": 2.92, + "learning_rate": 1.34393434213901e-06, + "loss": 0.2393, + "step": 75934 + }, + { + "epoch": 2.92, + "learning_rate": 1.3432931520902796e-06, + "loss": 0.2918, + "step": 75935 + }, + { + "epoch": 2.92, + "learning_rate": 1.3426519620415492e-06, + "loss": 0.3109, + "step": 75936 + }, + { + "epoch": 2.92, + "learning_rate": 1.3420107719928188e-06, + "loss": 0.2328, + "step": 75937 + }, + { + "epoch": 2.92, + "learning_rate": 1.3413695819440884e-06, + "loss": 0.1986, + "step": 75938 + }, + { + "epoch": 2.92, + "learning_rate": 1.3407283918953578e-06, + "loss": 0.3326, + "step": 75939 + }, + { + "epoch": 2.92, + "learning_rate": 1.3400872018466274e-06, + "loss": 0.117, + "step": 75940 + }, + { + "epoch": 2.92, + "learning_rate": 1.339446011797897e-06, + "loss": 0.3164, + "step": 75941 + }, + { + "epoch": 2.92, + "learning_rate": 1.3388048217491666e-06, + "loss": 0.1905, + "step": 75942 + }, + { + "epoch": 2.92, + "learning_rate": 1.3381636317004362e-06, + "loss": 0.2564, + "step": 75943 + }, + { + "epoch": 2.92, + "learning_rate": 1.3375224416517056e-06, + "loss": 0.1805, + "step": 75944 + }, + { + "epoch": 2.92, + "learning_rate": 1.3368812516029751e-06, + "loss": 0.2336, + "step": 75945 + }, + { + "epoch": 2.92, + "learning_rate": 1.3362400615542447e-06, + "loss": 0.0997, + "step": 75946 + }, + { + "epoch": 2.92, + "learning_rate": 1.3355988715055143e-06, + "loss": 0.1766, + "step": 75947 + }, + { + "epoch": 2.92, + "learning_rate": 1.334957681456784e-06, + "loss": 0.1143, + "step": 75948 + }, + { + "epoch": 2.92, + "learning_rate": 1.3343164914080533e-06, + "loss": 0.1778, + "step": 75949 + }, + { + "epoch": 2.92, + "learning_rate": 1.333675301359323e-06, + "loss": 0.2708, + "step": 75950 + }, + { + "epoch": 2.92, + "learning_rate": 1.3330341113105925e-06, + "loss": 0.4212, + "step": 75951 + }, + { + "epoch": 2.92, + "learning_rate": 1.3323929212618621e-06, + "loss": 0.152, + "step": 75952 + }, + { + "epoch": 2.92, + "learning_rate": 1.3317517312131317e-06, + "loss": 0.15, + "step": 75953 + }, + { + "epoch": 2.92, + "learning_rate": 1.3311105411644011e-06, + "loss": 0.1247, + "step": 75954 + }, + { + "epoch": 2.92, + "learning_rate": 1.3304693511156707e-06, + "loss": 0.1422, + "step": 75955 + }, + { + "epoch": 2.92, + "learning_rate": 1.3298281610669403e-06, + "loss": 0.1079, + "step": 75956 + }, + { + "epoch": 2.92, + "learning_rate": 1.32918697101821e-06, + "loss": 0.2152, + "step": 75957 + }, + { + "epoch": 2.92, + "learning_rate": 1.3285457809694795e-06, + "loss": 0.2815, + "step": 75958 + }, + { + "epoch": 2.92, + "learning_rate": 1.327904590920749e-06, + "loss": 0.1096, + "step": 75959 + }, + { + "epoch": 2.92, + "learning_rate": 1.3272634008720185e-06, + "loss": 0.1408, + "step": 75960 + }, + { + "epoch": 2.92, + "learning_rate": 1.326622210823288e-06, + "loss": 0.2863, + "step": 75961 + }, + { + "epoch": 2.92, + "learning_rate": 1.3259810207745577e-06, + "loss": 0.1927, + "step": 75962 + }, + { + "epoch": 2.92, + "learning_rate": 1.3253398307258273e-06, + "loss": 0.224, + "step": 75963 + }, + { + "epoch": 2.92, + "learning_rate": 1.3246986406770967e-06, + "loss": 0.2373, + "step": 75964 + }, + { + "epoch": 2.92, + "learning_rate": 1.3240574506283663e-06, + "loss": 0.1961, + "step": 75965 + }, + { + "epoch": 2.92, + "learning_rate": 1.3234162605796359e-06, + "loss": 0.2408, + "step": 75966 + }, + { + "epoch": 2.92, + "learning_rate": 1.3227750705309055e-06, + "loss": 0.2639, + "step": 75967 + }, + { + "epoch": 2.92, + "learning_rate": 1.3221338804821749e-06, + "loss": 0.2602, + "step": 75968 + }, + { + "epoch": 2.92, + "learning_rate": 1.3214926904334445e-06, + "loss": 0.247, + "step": 75969 + }, + { + "epoch": 2.92, + "learning_rate": 1.320851500384714e-06, + "loss": 0.3239, + "step": 75970 + }, + { + "epoch": 2.92, + "learning_rate": 1.3202103103359837e-06, + "loss": 0.1282, + "step": 75971 + }, + { + "epoch": 2.92, + "learning_rate": 1.3195691202872533e-06, + "loss": 0.1293, + "step": 75972 + }, + { + "epoch": 2.92, + "learning_rate": 1.3189279302385227e-06, + "loss": 0.1418, + "step": 75973 + }, + { + "epoch": 2.92, + "learning_rate": 1.3182867401897923e-06, + "loss": 0.144, + "step": 75974 + }, + { + "epoch": 2.92, + "learning_rate": 1.3176455501410619e-06, + "loss": 0.2209, + "step": 75975 + }, + { + "epoch": 2.92, + "learning_rate": 1.3170043600923314e-06, + "loss": 0.0802, + "step": 75976 + }, + { + "epoch": 2.92, + "learning_rate": 1.316363170043601e-06, + "loss": 0.1495, + "step": 75977 + }, + { + "epoch": 2.92, + "learning_rate": 1.3157219799948704e-06, + "loss": 0.1819, + "step": 75978 + }, + { + "epoch": 2.92, + "learning_rate": 1.31508078994614e-06, + "loss": 0.2186, + "step": 75979 + }, + { + "epoch": 2.92, + "learning_rate": 1.3144395998974096e-06, + "loss": 0.2054, + "step": 75980 + }, + { + "epoch": 2.92, + "learning_rate": 1.3137984098486792e-06, + "loss": 0.1398, + "step": 75981 + }, + { + "epoch": 2.92, + "learning_rate": 1.3131572197999488e-06, + "loss": 0.2372, + "step": 75982 + }, + { + "epoch": 2.92, + "learning_rate": 1.3125160297512182e-06, + "loss": 0.1718, + "step": 75983 + }, + { + "epoch": 2.92, + "learning_rate": 1.3118748397024878e-06, + "loss": 0.2821, + "step": 75984 + }, + { + "epoch": 2.92, + "learning_rate": 1.3112336496537574e-06, + "loss": 0.3025, + "step": 75985 + }, + { + "epoch": 2.92, + "learning_rate": 1.310592459605027e-06, + "loss": 0.1857, + "step": 75986 + }, + { + "epoch": 2.92, + "learning_rate": 1.3099512695562966e-06, + "loss": 0.2268, + "step": 75987 + }, + { + "epoch": 2.92, + "learning_rate": 1.309310079507566e-06, + "loss": 0.1088, + "step": 75988 + }, + { + "epoch": 2.92, + "learning_rate": 1.3086688894588356e-06, + "loss": 0.2713, + "step": 75989 + }, + { + "epoch": 2.92, + "learning_rate": 1.3080276994101052e-06, + "loss": 0.2688, + "step": 75990 + }, + { + "epoch": 2.92, + "learning_rate": 1.3073865093613748e-06, + "loss": 0.0951, + "step": 75991 + }, + { + "epoch": 2.92, + "learning_rate": 1.3067453193126444e-06, + "loss": 0.1088, + "step": 75992 + }, + { + "epoch": 2.92, + "learning_rate": 1.3061041292639138e-06, + "loss": 0.2457, + "step": 75993 + }, + { + "epoch": 2.92, + "learning_rate": 1.3054629392151834e-06, + "loss": 0.2262, + "step": 75994 + }, + { + "epoch": 2.92, + "learning_rate": 1.304821749166453e-06, + "loss": 0.1962, + "step": 75995 + }, + { + "epoch": 2.92, + "learning_rate": 1.3041805591177226e-06, + "loss": 0.3362, + "step": 75996 + }, + { + "epoch": 2.92, + "learning_rate": 1.3035393690689922e-06, + "loss": 0.0983, + "step": 75997 + }, + { + "epoch": 2.92, + "learning_rate": 1.3028981790202616e-06, + "loss": 0.1924, + "step": 75998 + }, + { + "epoch": 2.92, + "learning_rate": 1.3022569889715312e-06, + "loss": 0.1685, + "step": 75999 + }, + { + "epoch": 2.92, + "learning_rate": 1.3016157989228008e-06, + "loss": 0.1984, + "step": 76000 + }, + { + "epoch": 2.92, + "learning_rate": 1.3009746088740704e-06, + "loss": 0.1913, + "step": 76001 + }, + { + "epoch": 2.92, + "learning_rate": 1.30033341882534e-06, + "loss": 0.2903, + "step": 76002 + }, + { + "epoch": 2.92, + "learning_rate": 1.2996922287766094e-06, + "loss": 0.2306, + "step": 76003 + }, + { + "epoch": 2.92, + "learning_rate": 1.299051038727879e-06, + "loss": 0.2762, + "step": 76004 + }, + { + "epoch": 2.92, + "learning_rate": 1.2984098486791486e-06, + "loss": 0.1021, + "step": 76005 + }, + { + "epoch": 2.92, + "learning_rate": 1.2977686586304182e-06, + "loss": 0.311, + "step": 76006 + }, + { + "epoch": 2.92, + "learning_rate": 1.2971274685816877e-06, + "loss": 0.1257, + "step": 76007 + }, + { + "epoch": 2.92, + "learning_rate": 1.2964862785329571e-06, + "loss": 0.2092, + "step": 76008 + }, + { + "epoch": 2.92, + "learning_rate": 1.2958450884842267e-06, + "loss": 0.2053, + "step": 76009 + }, + { + "epoch": 2.92, + "learning_rate": 1.2952038984354963e-06, + "loss": 0.2415, + "step": 76010 + }, + { + "epoch": 2.92, + "learning_rate": 1.294562708386766e-06, + "loss": 0.1283, + "step": 76011 + }, + { + "epoch": 2.92, + "learning_rate": 1.2939215183380355e-06, + "loss": 0.152, + "step": 76012 + }, + { + "epoch": 2.92, + "learning_rate": 1.293280328289305e-06, + "loss": 0.1075, + "step": 76013 + }, + { + "epoch": 2.92, + "learning_rate": 1.2926391382405745e-06, + "loss": 0.2839, + "step": 76014 + }, + { + "epoch": 2.92, + "learning_rate": 1.2919979481918441e-06, + "loss": 0.2469, + "step": 76015 + }, + { + "epoch": 2.92, + "learning_rate": 1.2913567581431137e-06, + "loss": 0.2825, + "step": 76016 + }, + { + "epoch": 2.92, + "learning_rate": 1.2907155680943833e-06, + "loss": 0.1515, + "step": 76017 + }, + { + "epoch": 2.92, + "learning_rate": 1.2900743780456527e-06, + "loss": 0.1325, + "step": 76018 + }, + { + "epoch": 2.92, + "learning_rate": 1.2894331879969223e-06, + "loss": 0.1877, + "step": 76019 + }, + { + "epoch": 2.92, + "learning_rate": 1.288791997948192e-06, + "loss": 0.2306, + "step": 76020 + }, + { + "epoch": 2.92, + "learning_rate": 1.2881508078994615e-06, + "loss": 0.2141, + "step": 76021 + }, + { + "epoch": 2.92, + "learning_rate": 1.287509617850731e-06, + "loss": 0.1082, + "step": 76022 + }, + { + "epoch": 2.92, + "learning_rate": 1.2868684278020005e-06, + "loss": 0.1346, + "step": 76023 + }, + { + "epoch": 2.92, + "learning_rate": 1.28622723775327e-06, + "loss": 0.219, + "step": 76024 + }, + { + "epoch": 2.92, + "learning_rate": 1.2855860477045397e-06, + "loss": 0.1827, + "step": 76025 + }, + { + "epoch": 2.92, + "learning_rate": 1.2849448576558093e-06, + "loss": 0.2449, + "step": 76026 + }, + { + "epoch": 2.92, + "learning_rate": 1.2843036676070787e-06, + "loss": 0.171, + "step": 76027 + }, + { + "epoch": 2.92, + "learning_rate": 1.2836624775583483e-06, + "loss": 0.1565, + "step": 76028 + }, + { + "epoch": 2.92, + "learning_rate": 1.2830212875096179e-06, + "loss": 0.1681, + "step": 76029 + }, + { + "epoch": 2.92, + "learning_rate": 1.2823800974608875e-06, + "loss": 0.1153, + "step": 76030 + }, + { + "epoch": 2.92, + "learning_rate": 1.281738907412157e-06, + "loss": 0.1531, + "step": 76031 + }, + { + "epoch": 2.92, + "learning_rate": 1.2810977173634265e-06, + "loss": 0.1028, + "step": 76032 + }, + { + "epoch": 2.92, + "learning_rate": 1.280456527314696e-06, + "loss": 0.1631, + "step": 76033 + }, + { + "epoch": 2.92, + "learning_rate": 1.2798153372659657e-06, + "loss": 0.2172, + "step": 76034 + }, + { + "epoch": 2.92, + "learning_rate": 1.2791741472172353e-06, + "loss": 0.1912, + "step": 76035 + }, + { + "epoch": 2.92, + "learning_rate": 1.2785329571685049e-06, + "loss": 0.1571, + "step": 76036 + }, + { + "epoch": 2.92, + "learning_rate": 1.2778917671197742e-06, + "loss": 0.144, + "step": 76037 + }, + { + "epoch": 2.92, + "learning_rate": 1.2772505770710438e-06, + "loss": 0.2569, + "step": 76038 + }, + { + "epoch": 2.92, + "learning_rate": 1.2766093870223134e-06, + "loss": 0.1793, + "step": 76039 + }, + { + "epoch": 2.92, + "learning_rate": 1.275968196973583e-06, + "loss": 0.1263, + "step": 76040 + }, + { + "epoch": 2.92, + "learning_rate": 1.2753270069248526e-06, + "loss": 0.1689, + "step": 76041 + }, + { + "epoch": 2.92, + "learning_rate": 1.274685816876122e-06, + "loss": 0.2119, + "step": 76042 + }, + { + "epoch": 2.92, + "learning_rate": 1.2740446268273916e-06, + "loss": 0.1934, + "step": 76043 + }, + { + "epoch": 2.92, + "learning_rate": 1.2734034367786612e-06, + "loss": 0.1481, + "step": 76044 + }, + { + "epoch": 2.92, + "learning_rate": 1.2727622467299308e-06, + "loss": 0.2375, + "step": 76045 + }, + { + "epoch": 2.92, + "learning_rate": 1.2721210566812004e-06, + "loss": 0.2525, + "step": 76046 + }, + { + "epoch": 2.92, + "learning_rate": 1.2714798666324698e-06, + "loss": 0.2232, + "step": 76047 + }, + { + "epoch": 2.92, + "learning_rate": 1.2708386765837394e-06, + "loss": 0.2959, + "step": 76048 + }, + { + "epoch": 2.92, + "learning_rate": 1.270197486535009e-06, + "loss": 0.1802, + "step": 76049 + }, + { + "epoch": 2.92, + "learning_rate": 1.2695562964862786e-06, + "loss": 0.1487, + "step": 76050 + }, + { + "epoch": 2.92, + "learning_rate": 1.2689151064375482e-06, + "loss": 0.3116, + "step": 76051 + }, + { + "epoch": 2.92, + "learning_rate": 1.2682739163888176e-06, + "loss": 0.1596, + "step": 76052 + }, + { + "epoch": 2.92, + "learning_rate": 1.2676327263400872e-06, + "loss": 0.0817, + "step": 76053 + }, + { + "epoch": 2.92, + "learning_rate": 1.2669915362913568e-06, + "loss": 0.1689, + "step": 76054 + }, + { + "epoch": 2.92, + "learning_rate": 1.2663503462426264e-06, + "loss": 0.1593, + "step": 76055 + }, + { + "epoch": 2.92, + "learning_rate": 1.265709156193896e-06, + "loss": 0.1542, + "step": 76056 + }, + { + "epoch": 2.92, + "learning_rate": 1.2650679661451654e-06, + "loss": 0.1583, + "step": 76057 + }, + { + "epoch": 2.92, + "learning_rate": 1.264426776096435e-06, + "loss": 0.3391, + "step": 76058 + }, + { + "epoch": 2.92, + "learning_rate": 1.2637855860477046e-06, + "loss": 0.2545, + "step": 76059 + }, + { + "epoch": 2.92, + "learning_rate": 1.2631443959989742e-06, + "loss": 0.232, + "step": 76060 + }, + { + "epoch": 2.92, + "learning_rate": 1.2625032059502438e-06, + "loss": 0.1869, + "step": 76061 + }, + { + "epoch": 2.92, + "learning_rate": 1.2618620159015132e-06, + "loss": 0.1721, + "step": 76062 + }, + { + "epoch": 2.92, + "learning_rate": 1.2612208258527828e-06, + "loss": 0.2822, + "step": 76063 + }, + { + "epoch": 2.92, + "learning_rate": 1.2605796358040524e-06, + "loss": 0.2509, + "step": 76064 + }, + { + "epoch": 2.92, + "learning_rate": 1.259938445755322e-06, + "loss": 0.2653, + "step": 76065 + }, + { + "epoch": 2.92, + "learning_rate": 1.2592972557065916e-06, + "loss": 0.1435, + "step": 76066 + }, + { + "epoch": 2.92, + "learning_rate": 1.258656065657861e-06, + "loss": 0.1684, + "step": 76067 + }, + { + "epoch": 2.92, + "learning_rate": 1.2580148756091305e-06, + "loss": 0.1899, + "step": 76068 + }, + { + "epoch": 2.92, + "learning_rate": 1.2573736855604001e-06, + "loss": 0.1311, + "step": 76069 + }, + { + "epoch": 2.92, + "learning_rate": 1.2567324955116697e-06, + "loss": 0.2859, + "step": 76070 + }, + { + "epoch": 2.92, + "learning_rate": 1.2560913054629393e-06, + "loss": 0.176, + "step": 76071 + }, + { + "epoch": 2.92, + "learning_rate": 1.2554501154142087e-06, + "loss": 0.1987, + "step": 76072 + }, + { + "epoch": 2.92, + "learning_rate": 1.2548089253654783e-06, + "loss": 0.2117, + "step": 76073 + }, + { + "epoch": 2.92, + "learning_rate": 1.254167735316748e-06, + "loss": 0.3338, + "step": 76074 + }, + { + "epoch": 2.92, + "learning_rate": 1.2535265452680175e-06, + "loss": 0.1921, + "step": 76075 + }, + { + "epoch": 2.92, + "learning_rate": 1.2528853552192871e-06, + "loss": 0.2009, + "step": 76076 + }, + { + "epoch": 2.92, + "learning_rate": 1.2522441651705565e-06, + "loss": 0.1302, + "step": 76077 + }, + { + "epoch": 2.92, + "learning_rate": 1.251602975121826e-06, + "loss": 0.1848, + "step": 76078 + }, + { + "epoch": 2.92, + "learning_rate": 1.2509617850730957e-06, + "loss": 0.2021, + "step": 76079 + }, + { + "epoch": 2.93, + "learning_rate": 1.2503205950243653e-06, + "loss": 0.215, + "step": 76080 + }, + { + "epoch": 2.93, + "learning_rate": 1.2496794049756347e-06, + "loss": 0.2565, + "step": 76081 + }, + { + "epoch": 2.93, + "learning_rate": 1.2490382149269043e-06, + "loss": 0.203, + "step": 76082 + }, + { + "epoch": 2.93, + "learning_rate": 1.2483970248781739e-06, + "loss": 0.2483, + "step": 76083 + }, + { + "epoch": 2.93, + "learning_rate": 1.2477558348294435e-06, + "loss": 0.3287, + "step": 76084 + }, + { + "epoch": 2.93, + "learning_rate": 1.247114644780713e-06, + "loss": 0.0969, + "step": 76085 + }, + { + "epoch": 2.93, + "learning_rate": 1.2464734547319825e-06, + "loss": 0.2567, + "step": 76086 + }, + { + "epoch": 2.93, + "learning_rate": 1.245832264683252e-06, + "loss": 0.1789, + "step": 76087 + }, + { + "epoch": 2.93, + "learning_rate": 1.2451910746345217e-06, + "loss": 0.2138, + "step": 76088 + }, + { + "epoch": 2.93, + "learning_rate": 1.2445498845857913e-06, + "loss": 0.1857, + "step": 76089 + }, + { + "epoch": 2.93, + "learning_rate": 1.2439086945370609e-06, + "loss": 0.205, + "step": 76090 + }, + { + "epoch": 2.93, + "learning_rate": 1.2432675044883303e-06, + "loss": 0.1613, + "step": 76091 + }, + { + "epoch": 2.93, + "learning_rate": 1.2426263144395999e-06, + "loss": 0.2228, + "step": 76092 + }, + { + "epoch": 2.93, + "learning_rate": 1.2419851243908695e-06, + "loss": 0.1063, + "step": 76093 + }, + { + "epoch": 2.93, + "learning_rate": 1.241343934342139e-06, + "loss": 0.1, + "step": 76094 + }, + { + "epoch": 2.93, + "learning_rate": 1.2407027442934087e-06, + "loss": 0.1336, + "step": 76095 + }, + { + "epoch": 2.93, + "learning_rate": 1.240061554244678e-06, + "loss": 0.2249, + "step": 76096 + }, + { + "epoch": 2.93, + "learning_rate": 1.2394203641959476e-06, + "loss": 0.1674, + "step": 76097 + }, + { + "epoch": 2.93, + "learning_rate": 1.2387791741472172e-06, + "loss": 0.3061, + "step": 76098 + }, + { + "epoch": 2.93, + "learning_rate": 1.2381379840984868e-06, + "loss": 0.2284, + "step": 76099 + }, + { + "epoch": 2.93, + "learning_rate": 1.2374967940497564e-06, + "loss": 0.07, + "step": 76100 + }, + { + "epoch": 2.93, + "learning_rate": 1.2368556040010258e-06, + "loss": 0.1644, + "step": 76101 + }, + { + "epoch": 2.93, + "learning_rate": 1.2362144139522954e-06, + "loss": 0.2307, + "step": 76102 + }, + { + "epoch": 2.93, + "learning_rate": 1.235573223903565e-06, + "loss": 0.1311, + "step": 76103 + }, + { + "epoch": 2.93, + "learning_rate": 1.2349320338548346e-06, + "loss": 0.2683, + "step": 76104 + }, + { + "epoch": 2.93, + "learning_rate": 1.2342908438061042e-06, + "loss": 0.2642, + "step": 76105 + }, + { + "epoch": 2.93, + "learning_rate": 1.2336496537573736e-06, + "loss": 0.2045, + "step": 76106 + }, + { + "epoch": 2.93, + "learning_rate": 1.2330084637086432e-06, + "loss": 0.1956, + "step": 76107 + }, + { + "epoch": 2.93, + "learning_rate": 1.2323672736599128e-06, + "loss": 0.1345, + "step": 76108 + }, + { + "epoch": 2.93, + "learning_rate": 1.2317260836111824e-06, + "loss": 0.2912, + "step": 76109 + }, + { + "epoch": 2.93, + "learning_rate": 1.231084893562452e-06, + "loss": 0.2831, + "step": 76110 + }, + { + "epoch": 2.93, + "learning_rate": 1.2304437035137214e-06, + "loss": 0.2634, + "step": 76111 + }, + { + "epoch": 2.93, + "learning_rate": 1.229802513464991e-06, + "loss": 0.1013, + "step": 76112 + }, + { + "epoch": 2.93, + "learning_rate": 1.2291613234162606e-06, + "loss": 0.1446, + "step": 76113 + }, + { + "epoch": 2.93, + "learning_rate": 1.2285201333675302e-06, + "loss": 0.1574, + "step": 76114 + }, + { + "epoch": 2.93, + "learning_rate": 1.2278789433187998e-06, + "loss": 0.2435, + "step": 76115 + }, + { + "epoch": 2.93, + "learning_rate": 1.2272377532700692e-06, + "loss": 0.3515, + "step": 76116 + }, + { + "epoch": 2.93, + "learning_rate": 1.2265965632213388e-06, + "loss": 0.2777, + "step": 76117 + }, + { + "epoch": 2.93, + "learning_rate": 1.2259553731726084e-06, + "loss": 0.1467, + "step": 76118 + }, + { + "epoch": 2.93, + "learning_rate": 1.225314183123878e-06, + "loss": 0.172, + "step": 76119 + }, + { + "epoch": 2.93, + "learning_rate": 1.2246729930751476e-06, + "loss": 0.1834, + "step": 76120 + }, + { + "epoch": 2.93, + "learning_rate": 1.224031803026417e-06, + "loss": 0.2127, + "step": 76121 + }, + { + "epoch": 2.93, + "learning_rate": 1.2233906129776866e-06, + "loss": 0.201, + "step": 76122 + }, + { + "epoch": 2.93, + "learning_rate": 1.2227494229289562e-06, + "loss": 0.1375, + "step": 76123 + }, + { + "epoch": 2.93, + "learning_rate": 1.2221082328802258e-06, + "loss": 0.2108, + "step": 76124 + }, + { + "epoch": 2.93, + "learning_rate": 1.2214670428314954e-06, + "loss": 0.1835, + "step": 76125 + }, + { + "epoch": 2.93, + "learning_rate": 1.2208258527827647e-06, + "loss": 0.1868, + "step": 76126 + }, + { + "epoch": 2.93, + "learning_rate": 1.2201846627340343e-06, + "loss": 0.3465, + "step": 76127 + }, + { + "epoch": 2.93, + "learning_rate": 1.219543472685304e-06, + "loss": 0.1861, + "step": 76128 + }, + { + "epoch": 2.93, + "learning_rate": 1.2189022826365735e-06, + "loss": 0.2201, + "step": 76129 + }, + { + "epoch": 2.93, + "learning_rate": 1.2182610925878431e-06, + "loss": 0.3218, + "step": 76130 + }, + { + "epoch": 2.93, + "learning_rate": 1.2176199025391125e-06, + "loss": 0.1957, + "step": 76131 + }, + { + "epoch": 2.93, + "learning_rate": 1.2169787124903821e-06, + "loss": 0.1488, + "step": 76132 + }, + { + "epoch": 2.93, + "learning_rate": 1.2163375224416517e-06, + "loss": 0.1792, + "step": 76133 + }, + { + "epoch": 2.93, + "learning_rate": 1.2156963323929213e-06, + "loss": 0.2559, + "step": 76134 + }, + { + "epoch": 2.93, + "learning_rate": 1.215055142344191e-06, + "loss": 0.2381, + "step": 76135 + }, + { + "epoch": 2.93, + "learning_rate": 1.2144139522954603e-06, + "loss": 0.1474, + "step": 76136 + }, + { + "epoch": 2.93, + "learning_rate": 1.21377276224673e-06, + "loss": 0.1394, + "step": 76137 + }, + { + "epoch": 2.93, + "learning_rate": 1.2131315721979995e-06, + "loss": 0.2275, + "step": 76138 + }, + { + "epoch": 2.93, + "learning_rate": 1.212490382149269e-06, + "loss": 0.1997, + "step": 76139 + }, + { + "epoch": 2.93, + "learning_rate": 1.2118491921005385e-06, + "loss": 0.215, + "step": 76140 + }, + { + "epoch": 2.93, + "learning_rate": 1.211208002051808e-06, + "loss": 0.2149, + "step": 76141 + }, + { + "epoch": 2.93, + "learning_rate": 1.2105668120030777e-06, + "loss": 0.2815, + "step": 76142 + }, + { + "epoch": 2.93, + "learning_rate": 1.2099256219543473e-06, + "loss": 0.2057, + "step": 76143 + }, + { + "epoch": 2.93, + "learning_rate": 1.2092844319056169e-06, + "loss": 0.1521, + "step": 76144 + }, + { + "epoch": 2.93, + "learning_rate": 1.2086432418568863e-06, + "loss": 0.2229, + "step": 76145 + }, + { + "epoch": 2.93, + "learning_rate": 1.2080020518081559e-06, + "loss": 0.1896, + "step": 76146 + }, + { + "epoch": 2.93, + "learning_rate": 1.2073608617594255e-06, + "loss": 0.1345, + "step": 76147 + }, + { + "epoch": 2.93, + "learning_rate": 1.206719671710695e-06, + "loss": 0.1532, + "step": 76148 + }, + { + "epoch": 2.93, + "learning_rate": 1.2060784816619647e-06, + "loss": 0.2218, + "step": 76149 + }, + { + "epoch": 2.93, + "learning_rate": 1.205437291613234e-06, + "loss": 0.2587, + "step": 76150 + }, + { + "epoch": 2.93, + "learning_rate": 1.2047961015645037e-06, + "loss": 0.1555, + "step": 76151 + }, + { + "epoch": 2.93, + "learning_rate": 1.2041549115157733e-06, + "loss": 0.1763, + "step": 76152 + }, + { + "epoch": 2.93, + "learning_rate": 1.2035137214670429e-06, + "loss": 0.236, + "step": 76153 + }, + { + "epoch": 2.93, + "learning_rate": 1.2028725314183125e-06, + "loss": 0.1813, + "step": 76154 + }, + { + "epoch": 2.93, + "learning_rate": 1.2022313413695818e-06, + "loss": 0.2308, + "step": 76155 + }, + { + "epoch": 2.93, + "learning_rate": 1.2015901513208514e-06, + "loss": 0.2685, + "step": 76156 + }, + { + "epoch": 2.93, + "learning_rate": 1.200948961272121e-06, + "loss": 0.2537, + "step": 76157 + }, + { + "epoch": 2.93, + "learning_rate": 1.2003077712233906e-06, + "loss": 0.1718, + "step": 76158 + }, + { + "epoch": 2.93, + "learning_rate": 1.1996665811746602e-06, + "loss": 0.2719, + "step": 76159 + }, + { + "epoch": 2.93, + "learning_rate": 1.1990253911259296e-06, + "loss": 0.1322, + "step": 76160 + }, + { + "epoch": 2.93, + "learning_rate": 1.1983842010771992e-06, + "loss": 0.2073, + "step": 76161 + }, + { + "epoch": 2.93, + "learning_rate": 1.1977430110284688e-06, + "loss": 0.1994, + "step": 76162 + }, + { + "epoch": 2.93, + "learning_rate": 1.1971018209797384e-06, + "loss": 0.2212, + "step": 76163 + }, + { + "epoch": 2.93, + "learning_rate": 1.196460630931008e-06, + "loss": 0.3277, + "step": 76164 + }, + { + "epoch": 2.93, + "learning_rate": 1.1958194408822774e-06, + "loss": 0.1868, + "step": 76165 + }, + { + "epoch": 2.93, + "learning_rate": 1.195178250833547e-06, + "loss": 0.1475, + "step": 76166 + }, + { + "epoch": 2.93, + "learning_rate": 1.1945370607848166e-06, + "loss": 0.1094, + "step": 76167 + }, + { + "epoch": 2.93, + "learning_rate": 1.1938958707360862e-06, + "loss": 0.1867, + "step": 76168 + }, + { + "epoch": 2.93, + "learning_rate": 1.1932546806873558e-06, + "loss": 0.1888, + "step": 76169 + }, + { + "epoch": 2.93, + "learning_rate": 1.1926134906386252e-06, + "loss": 0.1324, + "step": 76170 + }, + { + "epoch": 2.93, + "learning_rate": 1.1919723005898948e-06, + "loss": 0.15, + "step": 76171 + }, + { + "epoch": 2.93, + "learning_rate": 1.1913311105411644e-06, + "loss": 0.1867, + "step": 76172 + }, + { + "epoch": 2.93, + "learning_rate": 1.190689920492434e-06, + "loss": 0.2816, + "step": 76173 + }, + { + "epoch": 2.93, + "learning_rate": 1.1900487304437036e-06, + "loss": 0.1751, + "step": 76174 + }, + { + "epoch": 2.93, + "learning_rate": 1.189407540394973e-06, + "loss": 0.2212, + "step": 76175 + }, + { + "epoch": 2.93, + "learning_rate": 1.1887663503462426e-06, + "loss": 0.3574, + "step": 76176 + }, + { + "epoch": 2.93, + "learning_rate": 1.1881251602975122e-06, + "loss": 0.1589, + "step": 76177 + }, + { + "epoch": 2.93, + "learning_rate": 1.1874839702487818e-06, + "loss": 0.1528, + "step": 76178 + }, + { + "epoch": 2.93, + "learning_rate": 1.1868427802000514e-06, + "loss": 0.1325, + "step": 76179 + }, + { + "epoch": 2.93, + "learning_rate": 1.1862015901513208e-06, + "loss": 0.1466, + "step": 76180 + }, + { + "epoch": 2.93, + "learning_rate": 1.1855604001025904e-06, + "loss": 0.2373, + "step": 76181 + }, + { + "epoch": 2.93, + "learning_rate": 1.18491921005386e-06, + "loss": 0.2067, + "step": 76182 + }, + { + "epoch": 2.93, + "learning_rate": 1.1842780200051296e-06, + "loss": 0.2722, + "step": 76183 + }, + { + "epoch": 2.93, + "learning_rate": 1.1836368299563992e-06, + "loss": 0.1445, + "step": 76184 + }, + { + "epoch": 2.93, + "learning_rate": 1.1829956399076685e-06, + "loss": 0.2776, + "step": 76185 + }, + { + "epoch": 2.93, + "learning_rate": 1.1823544498589381e-06, + "loss": 0.145, + "step": 76186 + }, + { + "epoch": 2.93, + "learning_rate": 1.1817132598102077e-06, + "loss": 0.1963, + "step": 76187 + }, + { + "epoch": 2.93, + "learning_rate": 1.1810720697614773e-06, + "loss": 0.2581, + "step": 76188 + }, + { + "epoch": 2.93, + "learning_rate": 1.180430879712747e-06, + "loss": 0.1001, + "step": 76189 + }, + { + "epoch": 2.93, + "learning_rate": 1.1797896896640163e-06, + "loss": 0.0692, + "step": 76190 + }, + { + "epoch": 2.93, + "learning_rate": 1.179148499615286e-06, + "loss": 0.0992, + "step": 76191 + }, + { + "epoch": 2.93, + "learning_rate": 1.1785073095665555e-06, + "loss": 0.2657, + "step": 76192 + }, + { + "epoch": 2.93, + "learning_rate": 1.1778661195178251e-06, + "loss": 0.1947, + "step": 76193 + }, + { + "epoch": 2.93, + "learning_rate": 1.1772249294690947e-06, + "loss": 0.1656, + "step": 76194 + }, + { + "epoch": 2.93, + "learning_rate": 1.1765837394203641e-06, + "loss": 0.1675, + "step": 76195 + }, + { + "epoch": 2.93, + "learning_rate": 1.1759425493716337e-06, + "loss": 0.0994, + "step": 76196 + }, + { + "epoch": 2.93, + "learning_rate": 1.1753013593229033e-06, + "loss": 0.2907, + "step": 76197 + }, + { + "epoch": 2.93, + "learning_rate": 1.174660169274173e-06, + "loss": 0.1327, + "step": 76198 + }, + { + "epoch": 2.93, + "learning_rate": 1.1740189792254423e-06, + "loss": 0.1508, + "step": 76199 + }, + { + "epoch": 2.93, + "learning_rate": 1.1733777891767119e-06, + "loss": 0.169, + "step": 76200 + }, + { + "epoch": 2.93, + "learning_rate": 1.1727365991279815e-06, + "loss": 0.2442, + "step": 76201 + }, + { + "epoch": 2.93, + "learning_rate": 1.172095409079251e-06, + "loss": 0.1928, + "step": 76202 + }, + { + "epoch": 2.93, + "learning_rate": 1.1714542190305207e-06, + "loss": 0.1672, + "step": 76203 + }, + { + "epoch": 2.93, + "learning_rate": 1.17081302898179e-06, + "loss": 0.1725, + "step": 76204 + }, + { + "epoch": 2.93, + "learning_rate": 1.1701718389330597e-06, + "loss": 0.2032, + "step": 76205 + }, + { + "epoch": 2.93, + "learning_rate": 1.1695306488843293e-06, + "loss": 0.2401, + "step": 76206 + }, + { + "epoch": 2.93, + "learning_rate": 1.1688894588355989e-06, + "loss": 0.1898, + "step": 76207 + }, + { + "epoch": 2.93, + "learning_rate": 1.1682482687868685e-06, + "loss": 0.162, + "step": 76208 + }, + { + "epoch": 2.93, + "learning_rate": 1.1676070787381379e-06, + "loss": 0.2142, + "step": 76209 + }, + { + "epoch": 2.93, + "learning_rate": 1.1669658886894075e-06, + "loss": 0.1778, + "step": 76210 + }, + { + "epoch": 2.93, + "learning_rate": 1.166324698640677e-06, + "loss": 0.2466, + "step": 76211 + }, + { + "epoch": 2.93, + "learning_rate": 1.1656835085919467e-06, + "loss": 0.3107, + "step": 76212 + }, + { + "epoch": 2.93, + "learning_rate": 1.1650423185432163e-06, + "loss": 0.0963, + "step": 76213 + }, + { + "epoch": 2.93, + "learning_rate": 1.1644011284944856e-06, + "loss": 0.1943, + "step": 76214 + }, + { + "epoch": 2.93, + "learning_rate": 1.1637599384457552e-06, + "loss": 0.1946, + "step": 76215 + }, + { + "epoch": 2.93, + "learning_rate": 1.1631187483970248e-06, + "loss": 0.1223, + "step": 76216 + }, + { + "epoch": 2.93, + "learning_rate": 1.1624775583482944e-06, + "loss": 0.1733, + "step": 76217 + }, + { + "epoch": 2.93, + "learning_rate": 1.161836368299564e-06, + "loss": 0.2487, + "step": 76218 + }, + { + "epoch": 2.93, + "learning_rate": 1.1611951782508334e-06, + "loss": 0.1958, + "step": 76219 + }, + { + "epoch": 2.93, + "learning_rate": 1.160553988202103e-06, + "loss": 0.2424, + "step": 76220 + }, + { + "epoch": 2.93, + "learning_rate": 1.1599127981533726e-06, + "loss": 0.1651, + "step": 76221 + }, + { + "epoch": 2.93, + "learning_rate": 1.1592716081046422e-06, + "loss": 0.2401, + "step": 76222 + }, + { + "epoch": 2.93, + "learning_rate": 1.1586304180559118e-06, + "loss": 0.2955, + "step": 76223 + }, + { + "epoch": 2.93, + "learning_rate": 1.1579892280071812e-06, + "loss": 0.2769, + "step": 76224 + }, + { + "epoch": 2.93, + "learning_rate": 1.1573480379584508e-06, + "loss": 0.1366, + "step": 76225 + }, + { + "epoch": 2.93, + "learning_rate": 1.1567068479097204e-06, + "loss": 0.1572, + "step": 76226 + }, + { + "epoch": 2.93, + "learning_rate": 1.15606565786099e-06, + "loss": 0.1693, + "step": 76227 + }, + { + "epoch": 2.93, + "learning_rate": 1.1554244678122596e-06, + "loss": 0.2351, + "step": 76228 + }, + { + "epoch": 2.93, + "learning_rate": 1.154783277763529e-06, + "loss": 0.1928, + "step": 76229 + }, + { + "epoch": 2.93, + "learning_rate": 1.1541420877147986e-06, + "loss": 0.1, + "step": 76230 + }, + { + "epoch": 2.93, + "learning_rate": 1.1535008976660682e-06, + "loss": 0.3135, + "step": 76231 + }, + { + "epoch": 2.93, + "learning_rate": 1.1528597076173378e-06, + "loss": 0.1462, + "step": 76232 + }, + { + "epoch": 2.93, + "learning_rate": 1.1522185175686074e-06, + "loss": 0.2373, + "step": 76233 + }, + { + "epoch": 2.93, + "learning_rate": 1.1515773275198768e-06, + "loss": 0.2962, + "step": 76234 + }, + { + "epoch": 2.93, + "learning_rate": 1.1509361374711464e-06, + "loss": 0.2373, + "step": 76235 + }, + { + "epoch": 2.93, + "learning_rate": 1.150294947422416e-06, + "loss": 0.1799, + "step": 76236 + }, + { + "epoch": 2.93, + "learning_rate": 1.1496537573736856e-06, + "loss": 0.1846, + "step": 76237 + }, + { + "epoch": 2.93, + "learning_rate": 1.1490125673249552e-06, + "loss": 0.1254, + "step": 76238 + }, + { + "epoch": 2.93, + "learning_rate": 1.1483713772762246e-06, + "loss": 0.1616, + "step": 76239 + }, + { + "epoch": 2.93, + "learning_rate": 1.1477301872274942e-06, + "loss": 0.1866, + "step": 76240 + }, + { + "epoch": 2.93, + "learning_rate": 1.1470889971787638e-06, + "loss": 0.176, + "step": 76241 + }, + { + "epoch": 2.93, + "learning_rate": 1.1464478071300334e-06, + "loss": 0.1571, + "step": 76242 + }, + { + "epoch": 2.93, + "learning_rate": 1.145806617081303e-06, + "loss": 0.2545, + "step": 76243 + }, + { + "epoch": 2.93, + "learning_rate": 1.1451654270325723e-06, + "loss": 0.2271, + "step": 76244 + }, + { + "epoch": 2.93, + "learning_rate": 1.144524236983842e-06, + "loss": 0.168, + "step": 76245 + }, + { + "epoch": 2.93, + "learning_rate": 1.1438830469351115e-06, + "loss": 0.1662, + "step": 76246 + }, + { + "epoch": 2.93, + "learning_rate": 1.1432418568863811e-06, + "loss": 0.1381, + "step": 76247 + }, + { + "epoch": 2.93, + "learning_rate": 1.1426006668376507e-06, + "loss": 0.3055, + "step": 76248 + }, + { + "epoch": 2.93, + "learning_rate": 1.1419594767889203e-06, + "loss": 0.208, + "step": 76249 + }, + { + "epoch": 2.93, + "learning_rate": 1.1413182867401897e-06, + "loss": 0.1318, + "step": 76250 + }, + { + "epoch": 2.93, + "learning_rate": 1.1406770966914593e-06, + "loss": 0.3287, + "step": 76251 + }, + { + "epoch": 2.93, + "learning_rate": 1.140035906642729e-06, + "loss": 0.1578, + "step": 76252 + }, + { + "epoch": 2.93, + "learning_rate": 1.1393947165939985e-06, + "loss": 0.1762, + "step": 76253 + }, + { + "epoch": 2.93, + "learning_rate": 1.1387535265452681e-06, + "loss": 0.3394, + "step": 76254 + }, + { + "epoch": 2.93, + "learning_rate": 1.1381123364965375e-06, + "loss": 0.0823, + "step": 76255 + }, + { + "epoch": 2.93, + "learning_rate": 1.1374711464478071e-06, + "loss": 0.1436, + "step": 76256 + }, + { + "epoch": 2.93, + "learning_rate": 1.1368299563990767e-06, + "loss": 0.0961, + "step": 76257 + }, + { + "epoch": 2.93, + "learning_rate": 1.1361887663503463e-06, + "loss": 0.1787, + "step": 76258 + }, + { + "epoch": 2.93, + "learning_rate": 1.135547576301616e-06, + "loss": 0.2206, + "step": 76259 + }, + { + "epoch": 2.93, + "learning_rate": 1.1349063862528853e-06, + "loss": 0.1577, + "step": 76260 + }, + { + "epoch": 2.93, + "learning_rate": 1.1342651962041549e-06, + "loss": 0.2529, + "step": 76261 + }, + { + "epoch": 2.93, + "learning_rate": 1.1336240061554245e-06, + "loss": 0.195, + "step": 76262 + }, + { + "epoch": 2.93, + "learning_rate": 1.132982816106694e-06, + "loss": 0.2855, + "step": 76263 + }, + { + "epoch": 2.93, + "learning_rate": 1.1323416260579637e-06, + "loss": 0.3073, + "step": 76264 + }, + { + "epoch": 2.93, + "learning_rate": 1.131700436009233e-06, + "loss": 0.1935, + "step": 76265 + }, + { + "epoch": 2.93, + "learning_rate": 1.1310592459605027e-06, + "loss": 0.2347, + "step": 76266 + }, + { + "epoch": 2.93, + "learning_rate": 1.1304180559117723e-06, + "loss": 0.25, + "step": 76267 + }, + { + "epoch": 2.93, + "learning_rate": 1.1297768658630419e-06, + "loss": 0.1792, + "step": 76268 + }, + { + "epoch": 2.93, + "learning_rate": 1.1291356758143115e-06, + "loss": 0.1478, + "step": 76269 + }, + { + "epoch": 2.93, + "learning_rate": 1.1284944857655809e-06, + "loss": 0.1978, + "step": 76270 + }, + { + "epoch": 2.93, + "learning_rate": 1.1278532957168505e-06, + "loss": 0.1952, + "step": 76271 + }, + { + "epoch": 2.93, + "learning_rate": 1.12721210566812e-06, + "loss": 0.2961, + "step": 76272 + }, + { + "epoch": 2.93, + "learning_rate": 1.1265709156193897e-06, + "loss": 0.2821, + "step": 76273 + }, + { + "epoch": 2.93, + "learning_rate": 1.1259297255706593e-06, + "loss": 0.2251, + "step": 76274 + }, + { + "epoch": 2.93, + "learning_rate": 1.1252885355219286e-06, + "loss": 0.2578, + "step": 76275 + }, + { + "epoch": 2.93, + "learning_rate": 1.1246473454731982e-06, + "loss": 0.193, + "step": 76276 + }, + { + "epoch": 2.93, + "learning_rate": 1.1240061554244678e-06, + "loss": 0.1941, + "step": 76277 + }, + { + "epoch": 2.93, + "learning_rate": 1.1233649653757374e-06, + "loss": 0.1488, + "step": 76278 + }, + { + "epoch": 2.93, + "learning_rate": 1.122723775327007e-06, + "loss": 0.1708, + "step": 76279 + }, + { + "epoch": 2.93, + "learning_rate": 1.1220825852782764e-06, + "loss": 0.2421, + "step": 76280 + }, + { + "epoch": 2.93, + "learning_rate": 1.121441395229546e-06, + "loss": 0.1307, + "step": 76281 + }, + { + "epoch": 2.93, + "learning_rate": 1.1208002051808156e-06, + "loss": 0.0902, + "step": 76282 + }, + { + "epoch": 2.93, + "learning_rate": 1.1201590151320852e-06, + "loss": 0.1408, + "step": 76283 + }, + { + "epoch": 2.93, + "learning_rate": 1.1195178250833548e-06, + "loss": 0.2452, + "step": 76284 + }, + { + "epoch": 2.93, + "learning_rate": 1.1188766350346242e-06, + "loss": 0.1787, + "step": 76285 + }, + { + "epoch": 2.93, + "learning_rate": 1.1182354449858938e-06, + "loss": 0.2158, + "step": 76286 + }, + { + "epoch": 2.93, + "learning_rate": 1.1175942549371634e-06, + "loss": 0.1741, + "step": 76287 + }, + { + "epoch": 2.93, + "learning_rate": 1.116953064888433e-06, + "loss": 0.1697, + "step": 76288 + }, + { + "epoch": 2.93, + "learning_rate": 1.1163118748397026e-06, + "loss": 0.1818, + "step": 76289 + }, + { + "epoch": 2.93, + "learning_rate": 1.115670684790972e-06, + "loss": 0.2756, + "step": 76290 + }, + { + "epoch": 2.93, + "learning_rate": 1.1150294947422416e-06, + "loss": 0.1817, + "step": 76291 + }, + { + "epoch": 2.93, + "learning_rate": 1.1143883046935112e-06, + "loss": 0.1164, + "step": 76292 + }, + { + "epoch": 2.93, + "learning_rate": 1.1137471146447808e-06, + "loss": 0.2666, + "step": 76293 + }, + { + "epoch": 2.93, + "learning_rate": 1.1131059245960504e-06, + "loss": 0.1508, + "step": 76294 + }, + { + "epoch": 2.93, + "learning_rate": 1.11246473454732e-06, + "loss": 0.1822, + "step": 76295 + }, + { + "epoch": 2.93, + "learning_rate": 1.1118235444985894e-06, + "loss": 0.2056, + "step": 76296 + }, + { + "epoch": 2.93, + "learning_rate": 1.111182354449859e-06, + "loss": 0.1235, + "step": 76297 + }, + { + "epoch": 2.93, + "learning_rate": 1.1105411644011286e-06, + "loss": 0.2277, + "step": 76298 + }, + { + "epoch": 2.93, + "learning_rate": 1.1098999743523982e-06, + "loss": 0.1018, + "step": 76299 + }, + { + "epoch": 2.93, + "learning_rate": 1.1092587843036678e-06, + "loss": 0.3074, + "step": 76300 + }, + { + "epoch": 2.93, + "learning_rate": 1.1086175942549372e-06, + "loss": 0.3082, + "step": 76301 + }, + { + "epoch": 2.93, + "learning_rate": 1.1079764042062068e-06, + "loss": 0.1874, + "step": 76302 + }, + { + "epoch": 2.93, + "learning_rate": 1.1073352141574764e-06, + "loss": 0.341, + "step": 76303 + }, + { + "epoch": 2.93, + "learning_rate": 1.106694024108746e-06, + "loss": 0.1991, + "step": 76304 + }, + { + "epoch": 2.93, + "learning_rate": 1.1060528340600156e-06, + "loss": 0.2473, + "step": 76305 + }, + { + "epoch": 2.93, + "learning_rate": 1.105411644011285e-06, + "loss": 0.2284, + "step": 76306 + }, + { + "epoch": 2.93, + "learning_rate": 1.1047704539625545e-06, + "loss": 0.3023, + "step": 76307 + }, + { + "epoch": 2.93, + "learning_rate": 1.1041292639138241e-06, + "loss": 0.2184, + "step": 76308 + }, + { + "epoch": 2.93, + "learning_rate": 1.1034880738650937e-06, + "loss": 0.1829, + "step": 76309 + }, + { + "epoch": 2.93, + "learning_rate": 1.1028468838163633e-06, + "loss": 0.3576, + "step": 76310 + }, + { + "epoch": 2.93, + "learning_rate": 1.1022056937676327e-06, + "loss": 0.2584, + "step": 76311 + }, + { + "epoch": 2.93, + "learning_rate": 1.1015645037189023e-06, + "loss": 0.1458, + "step": 76312 + }, + { + "epoch": 2.93, + "learning_rate": 1.100923313670172e-06, + "loss": 0.2228, + "step": 76313 + }, + { + "epoch": 2.93, + "learning_rate": 1.1002821236214415e-06, + "loss": 0.2073, + "step": 76314 + }, + { + "epoch": 2.93, + "learning_rate": 1.0996409335727111e-06, + "loss": 0.1137, + "step": 76315 + }, + { + "epoch": 2.93, + "learning_rate": 1.0989997435239805e-06, + "loss": 0.1559, + "step": 76316 + }, + { + "epoch": 2.93, + "learning_rate": 1.0983585534752501e-06, + "loss": 0.2476, + "step": 76317 + }, + { + "epoch": 2.93, + "learning_rate": 1.0977173634265197e-06, + "loss": 0.1739, + "step": 76318 + }, + { + "epoch": 2.93, + "learning_rate": 1.0970761733777893e-06, + "loss": 0.3533, + "step": 76319 + }, + { + "epoch": 2.93, + "learning_rate": 1.096434983329059e-06, + "loss": 0.1467, + "step": 76320 + }, + { + "epoch": 2.93, + "learning_rate": 1.0957937932803283e-06, + "loss": 0.1378, + "step": 76321 + }, + { + "epoch": 2.93, + "learning_rate": 1.095152603231598e-06, + "loss": 0.227, + "step": 76322 + }, + { + "epoch": 2.93, + "learning_rate": 1.0945114131828675e-06, + "loss": 0.2367, + "step": 76323 + }, + { + "epoch": 2.93, + "learning_rate": 1.093870223134137e-06, + "loss": 0.1583, + "step": 76324 + }, + { + "epoch": 2.93, + "learning_rate": 1.0932290330854067e-06, + "loss": 0.1555, + "step": 76325 + }, + { + "epoch": 2.93, + "learning_rate": 1.092587843036676e-06, + "loss": 0.2302, + "step": 76326 + }, + { + "epoch": 2.93, + "learning_rate": 1.0919466529879457e-06, + "loss": 0.2959, + "step": 76327 + }, + { + "epoch": 2.93, + "learning_rate": 1.0913054629392153e-06, + "loss": 0.1865, + "step": 76328 + }, + { + "epoch": 2.93, + "learning_rate": 1.0906642728904849e-06, + "loss": 0.2344, + "step": 76329 + }, + { + "epoch": 2.93, + "learning_rate": 1.0900230828417545e-06, + "loss": 0.18, + "step": 76330 + }, + { + "epoch": 2.93, + "learning_rate": 1.0893818927930239e-06, + "loss": 0.2004, + "step": 76331 + }, + { + "epoch": 2.93, + "learning_rate": 1.0887407027442935e-06, + "loss": 0.2271, + "step": 76332 + }, + { + "epoch": 2.93, + "learning_rate": 1.088099512695563e-06, + "loss": 0.0698, + "step": 76333 + }, + { + "epoch": 2.93, + "learning_rate": 1.0874583226468327e-06, + "loss": 0.0927, + "step": 76334 + }, + { + "epoch": 2.93, + "learning_rate": 1.0868171325981023e-06, + "loss": 0.1664, + "step": 76335 + }, + { + "epoch": 2.93, + "learning_rate": 1.0861759425493716e-06, + "loss": 0.1081, + "step": 76336 + }, + { + "epoch": 2.93, + "learning_rate": 1.0855347525006412e-06, + "loss": 0.2697, + "step": 76337 + }, + { + "epoch": 2.93, + "learning_rate": 1.0848935624519108e-06, + "loss": 0.2858, + "step": 76338 + }, + { + "epoch": 2.93, + "learning_rate": 1.0842523724031804e-06, + "loss": 0.2484, + "step": 76339 + }, + { + "epoch": 2.94, + "learning_rate": 1.08361118235445e-06, + "loss": 0.3125, + "step": 76340 + }, + { + "epoch": 2.94, + "learning_rate": 1.0829699923057196e-06, + "loss": 0.3062, + "step": 76341 + }, + { + "epoch": 2.94, + "learning_rate": 1.082328802256989e-06, + "loss": 0.1939, + "step": 76342 + }, + { + "epoch": 2.94, + "learning_rate": 1.0816876122082586e-06, + "loss": 0.2027, + "step": 76343 + }, + { + "epoch": 2.94, + "learning_rate": 1.0810464221595282e-06, + "loss": 0.1564, + "step": 76344 + }, + { + "epoch": 2.94, + "learning_rate": 1.0804052321107978e-06, + "loss": 0.1978, + "step": 76345 + }, + { + "epoch": 2.94, + "learning_rate": 1.0797640420620674e-06, + "loss": 0.0888, + "step": 76346 + }, + { + "epoch": 2.94, + "learning_rate": 1.0791228520133368e-06, + "loss": 0.2362, + "step": 76347 + }, + { + "epoch": 2.94, + "learning_rate": 1.0784816619646064e-06, + "loss": 0.1522, + "step": 76348 + }, + { + "epoch": 2.94, + "learning_rate": 1.077840471915876e-06, + "loss": 0.2148, + "step": 76349 + }, + { + "epoch": 2.94, + "learning_rate": 1.0771992818671456e-06, + "loss": 0.2373, + "step": 76350 + }, + { + "epoch": 2.94, + "learning_rate": 1.0765580918184152e-06, + "loss": 0.2162, + "step": 76351 + }, + { + "epoch": 2.94, + "learning_rate": 1.0759169017696846e-06, + "loss": 0.2976, + "step": 76352 + }, + { + "epoch": 2.94, + "learning_rate": 1.0752757117209542e-06, + "loss": 0.2761, + "step": 76353 + }, + { + "epoch": 2.94, + "learning_rate": 1.0746345216722238e-06, + "loss": 0.1637, + "step": 76354 + }, + { + "epoch": 2.94, + "learning_rate": 1.0739933316234934e-06, + "loss": 0.2185, + "step": 76355 + }, + { + "epoch": 2.94, + "learning_rate": 1.073352141574763e-06, + "loss": 0.307, + "step": 76356 + }, + { + "epoch": 2.94, + "learning_rate": 1.0727109515260324e-06, + "loss": 0.1317, + "step": 76357 + }, + { + "epoch": 2.94, + "learning_rate": 1.072069761477302e-06, + "loss": 0.2776, + "step": 76358 + }, + { + "epoch": 2.94, + "learning_rate": 1.0714285714285716e-06, + "loss": 0.2092, + "step": 76359 + }, + { + "epoch": 2.94, + "learning_rate": 1.0707873813798412e-06, + "loss": 0.2825, + "step": 76360 + }, + { + "epoch": 2.94, + "learning_rate": 1.0701461913311108e-06, + "loss": 0.0807, + "step": 76361 + }, + { + "epoch": 2.94, + "learning_rate": 1.0695050012823802e-06, + "loss": 0.1478, + "step": 76362 + }, + { + "epoch": 2.94, + "learning_rate": 1.0688638112336498e-06, + "loss": 0.2937, + "step": 76363 + }, + { + "epoch": 2.94, + "learning_rate": 1.0682226211849194e-06, + "loss": 0.3388, + "step": 76364 + }, + { + "epoch": 2.94, + "learning_rate": 1.067581431136189e-06, + "loss": 0.1875, + "step": 76365 + }, + { + "epoch": 2.94, + "learning_rate": 1.0669402410874586e-06, + "loss": 0.1586, + "step": 76366 + }, + { + "epoch": 2.94, + "learning_rate": 1.066299051038728e-06, + "loss": 0.1167, + "step": 76367 + }, + { + "epoch": 2.94, + "learning_rate": 1.0656578609899975e-06, + "loss": 0.193, + "step": 76368 + }, + { + "epoch": 2.94, + "learning_rate": 1.0650166709412671e-06, + "loss": 0.1366, + "step": 76369 + }, + { + "epoch": 2.94, + "learning_rate": 1.0643754808925367e-06, + "loss": 0.1513, + "step": 76370 + }, + { + "epoch": 2.94, + "learning_rate": 1.0637342908438063e-06, + "loss": 0.2594, + "step": 76371 + }, + { + "epoch": 2.94, + "learning_rate": 1.0630931007950757e-06, + "loss": 0.1577, + "step": 76372 + }, + { + "epoch": 2.94, + "learning_rate": 1.0624519107463453e-06, + "loss": 0.1832, + "step": 76373 + }, + { + "epoch": 2.94, + "learning_rate": 1.061810720697615e-06, + "loss": 0.0912, + "step": 76374 + }, + { + "epoch": 2.94, + "learning_rate": 1.0611695306488845e-06, + "loss": 0.2138, + "step": 76375 + }, + { + "epoch": 2.94, + "learning_rate": 1.060528340600154e-06, + "loss": 0.1785, + "step": 76376 + }, + { + "epoch": 2.94, + "learning_rate": 1.0598871505514235e-06, + "loss": 0.1484, + "step": 76377 + }, + { + "epoch": 2.94, + "learning_rate": 1.0592459605026931e-06, + "loss": 0.1741, + "step": 76378 + }, + { + "epoch": 2.94, + "learning_rate": 1.0586047704539627e-06, + "loss": 0.1415, + "step": 76379 + }, + { + "epoch": 2.94, + "learning_rate": 1.0579635804052323e-06, + "loss": 0.3764, + "step": 76380 + }, + { + "epoch": 2.94, + "learning_rate": 1.0573223903565017e-06, + "loss": 0.113, + "step": 76381 + }, + { + "epoch": 2.94, + "learning_rate": 1.0566812003077713e-06, + "loss": 0.2587, + "step": 76382 + }, + { + "epoch": 2.94, + "learning_rate": 1.056040010259041e-06, + "loss": 0.1767, + "step": 76383 + }, + { + "epoch": 2.94, + "learning_rate": 1.0553988202103105e-06, + "loss": 0.2013, + "step": 76384 + }, + { + "epoch": 2.94, + "learning_rate": 1.05475763016158e-06, + "loss": 0.1602, + "step": 76385 + }, + { + "epoch": 2.94, + "learning_rate": 1.0541164401128495e-06, + "loss": 0.1472, + "step": 76386 + }, + { + "epoch": 2.94, + "learning_rate": 1.053475250064119e-06, + "loss": 0.2187, + "step": 76387 + }, + { + "epoch": 2.94, + "learning_rate": 1.0528340600153887e-06, + "loss": 0.3084, + "step": 76388 + }, + { + "epoch": 2.94, + "learning_rate": 1.0521928699666583e-06, + "loss": 0.1339, + "step": 76389 + }, + { + "epoch": 2.94, + "learning_rate": 1.0515516799179279e-06, + "loss": 0.2384, + "step": 76390 + }, + { + "epoch": 2.94, + "learning_rate": 1.0509104898691973e-06, + "loss": 0.1371, + "step": 76391 + }, + { + "epoch": 2.94, + "learning_rate": 1.0502692998204669e-06, + "loss": 0.2895, + "step": 76392 + }, + { + "epoch": 2.94, + "learning_rate": 1.0496281097717365e-06, + "loss": 0.25, + "step": 76393 + }, + { + "epoch": 2.94, + "learning_rate": 1.048986919723006e-06, + "loss": 0.2176, + "step": 76394 + }, + { + "epoch": 2.94, + "learning_rate": 1.0483457296742757e-06, + "loss": 0.2445, + "step": 76395 + }, + { + "epoch": 2.94, + "learning_rate": 1.047704539625545e-06, + "loss": 0.1175, + "step": 76396 + }, + { + "epoch": 2.94, + "learning_rate": 1.0470633495768146e-06, + "loss": 0.1895, + "step": 76397 + }, + { + "epoch": 2.94, + "learning_rate": 1.0464221595280842e-06, + "loss": 0.1663, + "step": 76398 + }, + { + "epoch": 2.94, + "learning_rate": 1.0457809694793538e-06, + "loss": 0.2872, + "step": 76399 + }, + { + "epoch": 2.94, + "learning_rate": 1.0451397794306234e-06, + "loss": 0.2217, + "step": 76400 + }, + { + "epoch": 2.94, + "learning_rate": 1.0444985893818928e-06, + "loss": 0.2508, + "step": 76401 + }, + { + "epoch": 2.94, + "learning_rate": 1.0438573993331624e-06, + "loss": 0.1402, + "step": 76402 + }, + { + "epoch": 2.94, + "learning_rate": 1.043216209284432e-06, + "loss": 0.1597, + "step": 76403 + }, + { + "epoch": 2.94, + "learning_rate": 1.0425750192357016e-06, + "loss": 0.1372, + "step": 76404 + }, + { + "epoch": 2.94, + "learning_rate": 1.0419338291869712e-06, + "loss": 0.2929, + "step": 76405 + }, + { + "epoch": 2.94, + "learning_rate": 1.0412926391382406e-06, + "loss": 0.2329, + "step": 76406 + }, + { + "epoch": 2.94, + "learning_rate": 1.0406514490895102e-06, + "loss": 0.1779, + "step": 76407 + }, + { + "epoch": 2.94, + "learning_rate": 1.0400102590407798e-06, + "loss": 0.1131, + "step": 76408 + }, + { + "epoch": 2.94, + "learning_rate": 1.0393690689920494e-06, + "loss": 0.2197, + "step": 76409 + }, + { + "epoch": 2.94, + "learning_rate": 1.038727878943319e-06, + "loss": 0.1382, + "step": 76410 + }, + { + "epoch": 2.94, + "learning_rate": 1.0380866888945884e-06, + "loss": 0.2806, + "step": 76411 + }, + { + "epoch": 2.94, + "learning_rate": 1.037445498845858e-06, + "loss": 0.1096, + "step": 76412 + }, + { + "epoch": 2.94, + "learning_rate": 1.0368043087971276e-06, + "loss": 0.1519, + "step": 76413 + }, + { + "epoch": 2.94, + "learning_rate": 1.0361631187483972e-06, + "loss": 0.1159, + "step": 76414 + }, + { + "epoch": 2.94, + "learning_rate": 1.0355219286996668e-06, + "loss": 0.111, + "step": 76415 + }, + { + "epoch": 2.94, + "learning_rate": 1.0348807386509362e-06, + "loss": 0.0563, + "step": 76416 + }, + { + "epoch": 2.94, + "learning_rate": 1.0342395486022058e-06, + "loss": 0.2174, + "step": 76417 + }, + { + "epoch": 2.94, + "learning_rate": 1.0335983585534754e-06, + "loss": 0.1331, + "step": 76418 + }, + { + "epoch": 2.94, + "learning_rate": 1.032957168504745e-06, + "loss": 0.1827, + "step": 76419 + }, + { + "epoch": 2.94, + "learning_rate": 1.0323159784560146e-06, + "loss": 0.1477, + "step": 76420 + }, + { + "epoch": 2.94, + "learning_rate": 1.031674788407284e-06, + "loss": 0.3698, + "step": 76421 + }, + { + "epoch": 2.94, + "learning_rate": 1.0310335983585536e-06, + "loss": 0.2103, + "step": 76422 + }, + { + "epoch": 2.94, + "learning_rate": 1.0303924083098232e-06, + "loss": 0.2267, + "step": 76423 + }, + { + "epoch": 2.94, + "learning_rate": 1.0297512182610928e-06, + "loss": 0.1769, + "step": 76424 + }, + { + "epoch": 2.94, + "learning_rate": 1.0291100282123624e-06, + "loss": 0.1604, + "step": 76425 + }, + { + "epoch": 2.94, + "learning_rate": 1.0284688381636317e-06, + "loss": 0.2391, + "step": 76426 + }, + { + "epoch": 2.94, + "learning_rate": 1.0278276481149013e-06, + "loss": 0.2587, + "step": 76427 + }, + { + "epoch": 2.94, + "learning_rate": 1.027186458066171e-06, + "loss": 0.1223, + "step": 76428 + }, + { + "epoch": 2.94, + "learning_rate": 1.0265452680174405e-06, + "loss": 0.1422, + "step": 76429 + }, + { + "epoch": 2.94, + "learning_rate": 1.02590407796871e-06, + "loss": 0.2077, + "step": 76430 + }, + { + "epoch": 2.94, + "learning_rate": 1.0252628879199795e-06, + "loss": 0.2067, + "step": 76431 + }, + { + "epoch": 2.94, + "learning_rate": 1.0246216978712491e-06, + "loss": 0.2268, + "step": 76432 + }, + { + "epoch": 2.94, + "learning_rate": 1.0239805078225187e-06, + "loss": 0.2164, + "step": 76433 + }, + { + "epoch": 2.94, + "learning_rate": 1.0233393177737883e-06, + "loss": 0.2011, + "step": 76434 + }, + { + "epoch": 2.94, + "learning_rate": 1.0226981277250577e-06, + "loss": 0.1035, + "step": 76435 + }, + { + "epoch": 2.94, + "learning_rate": 1.0220569376763273e-06, + "loss": 0.2347, + "step": 76436 + }, + { + "epoch": 2.94, + "learning_rate": 1.021415747627597e-06, + "loss": 0.2822, + "step": 76437 + }, + { + "epoch": 2.94, + "learning_rate": 1.0207745575788665e-06, + "loss": 0.2047, + "step": 76438 + }, + { + "epoch": 2.94, + "learning_rate": 1.0201333675301361e-06, + "loss": 0.1977, + "step": 76439 + }, + { + "epoch": 2.94, + "learning_rate": 1.0194921774814055e-06, + "loss": 0.1544, + "step": 76440 + }, + { + "epoch": 2.94, + "learning_rate": 1.018850987432675e-06, + "loss": 0.2123, + "step": 76441 + }, + { + "epoch": 2.94, + "learning_rate": 1.0182097973839447e-06, + "loss": 0.229, + "step": 76442 + }, + { + "epoch": 2.94, + "learning_rate": 1.0175686073352143e-06, + "loss": 0.1411, + "step": 76443 + }, + { + "epoch": 2.94, + "learning_rate": 1.016927417286484e-06, + "loss": 0.2366, + "step": 76444 + }, + { + "epoch": 2.94, + "learning_rate": 1.0162862272377533e-06, + "loss": 0.0633, + "step": 76445 + }, + { + "epoch": 2.94, + "learning_rate": 1.0156450371890229e-06, + "loss": 0.1809, + "step": 76446 + }, + { + "epoch": 2.94, + "learning_rate": 1.0150038471402925e-06, + "loss": 0.2302, + "step": 76447 + }, + { + "epoch": 2.94, + "learning_rate": 1.014362657091562e-06, + "loss": 0.1536, + "step": 76448 + }, + { + "epoch": 2.94, + "learning_rate": 1.0137214670428317e-06, + "loss": 0.2666, + "step": 76449 + }, + { + "epoch": 2.94, + "learning_rate": 1.013080276994101e-06, + "loss": 0.3176, + "step": 76450 + }, + { + "epoch": 2.94, + "learning_rate": 1.0124390869453707e-06, + "loss": 0.2302, + "step": 76451 + }, + { + "epoch": 2.94, + "learning_rate": 1.0117978968966403e-06, + "loss": 0.335, + "step": 76452 + }, + { + "epoch": 2.94, + "learning_rate": 1.0111567068479099e-06, + "loss": 0.2599, + "step": 76453 + }, + { + "epoch": 2.94, + "learning_rate": 1.0105155167991795e-06, + "loss": 0.1909, + "step": 76454 + }, + { + "epoch": 2.94, + "learning_rate": 1.0098743267504488e-06, + "loss": 0.3659, + "step": 76455 + }, + { + "epoch": 2.94, + "learning_rate": 1.0092331367017184e-06, + "loss": 0.2449, + "step": 76456 + }, + { + "epoch": 2.94, + "learning_rate": 1.008591946652988e-06, + "loss": 0.1414, + "step": 76457 + }, + { + "epoch": 2.94, + "learning_rate": 1.0079507566042576e-06, + "loss": 0.0886, + "step": 76458 + }, + { + "epoch": 2.94, + "learning_rate": 1.0073095665555272e-06, + "loss": 0.2021, + "step": 76459 + }, + { + "epoch": 2.94, + "learning_rate": 1.0066683765067966e-06, + "loss": 0.3888, + "step": 76460 + }, + { + "epoch": 2.94, + "learning_rate": 1.0060271864580662e-06, + "loss": 0.2015, + "step": 76461 + }, + { + "epoch": 2.94, + "learning_rate": 1.0053859964093358e-06, + "loss": 0.2315, + "step": 76462 + }, + { + "epoch": 2.94, + "learning_rate": 1.0047448063606054e-06, + "loss": 0.2116, + "step": 76463 + }, + { + "epoch": 2.94, + "learning_rate": 1.004103616311875e-06, + "loss": 0.312, + "step": 76464 + }, + { + "epoch": 2.94, + "learning_rate": 1.0034624262631444e-06, + "loss": 0.2244, + "step": 76465 + }, + { + "epoch": 2.94, + "learning_rate": 1.002821236214414e-06, + "loss": 0.2063, + "step": 76466 + }, + { + "epoch": 2.94, + "learning_rate": 1.0021800461656836e-06, + "loss": 0.2445, + "step": 76467 + }, + { + "epoch": 2.94, + "learning_rate": 1.0015388561169532e-06, + "loss": 0.1768, + "step": 76468 + }, + { + "epoch": 2.94, + "learning_rate": 1.0008976660682228e-06, + "loss": 0.1695, + "step": 76469 + }, + { + "epoch": 2.94, + "learning_rate": 1.0002564760194922e-06, + "loss": 0.3152, + "step": 76470 + }, + { + "epoch": 2.94, + "learning_rate": 9.996152859707618e-07, + "loss": 0.1468, + "step": 76471 + }, + { + "epoch": 2.94, + "learning_rate": 9.989740959220314e-07, + "loss": 0.1406, + "step": 76472 + }, + { + "epoch": 2.94, + "learning_rate": 9.98332905873301e-07, + "loss": 0.2095, + "step": 76473 + }, + { + "epoch": 2.94, + "learning_rate": 9.976917158245706e-07, + "loss": 0.1826, + "step": 76474 + }, + { + "epoch": 2.94, + "learning_rate": 9.9705052577584e-07, + "loss": 0.1892, + "step": 76475 + }, + { + "epoch": 2.94, + "learning_rate": 9.964093357271096e-07, + "loss": 0.2418, + "step": 76476 + }, + { + "epoch": 2.94, + "learning_rate": 9.957681456783792e-07, + "loss": 0.3893, + "step": 76477 + }, + { + "epoch": 2.94, + "learning_rate": 9.951269556296488e-07, + "loss": 0.1568, + "step": 76478 + }, + { + "epoch": 2.94, + "learning_rate": 9.944857655809184e-07, + "loss": 0.1528, + "step": 76479 + }, + { + "epoch": 2.94, + "learning_rate": 9.938445755321878e-07, + "loss": 0.2514, + "step": 76480 + }, + { + "epoch": 2.94, + "learning_rate": 9.932033854834574e-07, + "loss": 0.1576, + "step": 76481 + }, + { + "epoch": 2.94, + "learning_rate": 9.92562195434727e-07, + "loss": 0.3677, + "step": 76482 + }, + { + "epoch": 2.94, + "learning_rate": 9.919210053859966e-07, + "loss": 0.0822, + "step": 76483 + }, + { + "epoch": 2.94, + "learning_rate": 9.912798153372662e-07, + "loss": 0.2689, + "step": 76484 + }, + { + "epoch": 2.94, + "learning_rate": 9.906386252885356e-07, + "loss": 0.1839, + "step": 76485 + }, + { + "epoch": 2.94, + "learning_rate": 9.899974352398051e-07, + "loss": 0.1265, + "step": 76486 + }, + { + "epoch": 2.94, + "learning_rate": 9.893562451910747e-07, + "loss": 0.3667, + "step": 76487 + }, + { + "epoch": 2.94, + "learning_rate": 9.887150551423443e-07, + "loss": 0.1901, + "step": 76488 + }, + { + "epoch": 2.94, + "learning_rate": 9.880738650936137e-07, + "loss": 0.169, + "step": 76489 + }, + { + "epoch": 2.94, + "learning_rate": 9.874326750448833e-07, + "loss": 0.2292, + "step": 76490 + }, + { + "epoch": 2.94, + "learning_rate": 9.86791484996153e-07, + "loss": 0.251, + "step": 76491 + }, + { + "epoch": 2.94, + "learning_rate": 9.861502949474225e-07, + "loss": 0.3621, + "step": 76492 + }, + { + "epoch": 2.94, + "learning_rate": 9.855091048986921e-07, + "loss": 0.2726, + "step": 76493 + }, + { + "epoch": 2.94, + "learning_rate": 9.848679148499615e-07, + "loss": 0.2446, + "step": 76494 + }, + { + "epoch": 2.94, + "learning_rate": 9.842267248012311e-07, + "loss": 0.2206, + "step": 76495 + }, + { + "epoch": 2.94, + "learning_rate": 9.835855347525007e-07, + "loss": 0.2403, + "step": 76496 + }, + { + "epoch": 2.94, + "learning_rate": 9.829443447037703e-07, + "loss": 0.2848, + "step": 76497 + }, + { + "epoch": 2.94, + "learning_rate": 9.8230315465504e-07, + "loss": 0.2572, + "step": 76498 + }, + { + "epoch": 2.94, + "learning_rate": 9.816619646063093e-07, + "loss": 0.2437, + "step": 76499 + }, + { + "epoch": 2.94, + "learning_rate": 9.81020774557579e-07, + "loss": 0.2027, + "step": 76500 + }, + { + "epoch": 2.94, + "learning_rate": 9.803795845088485e-07, + "loss": 0.1941, + "step": 76501 + }, + { + "epoch": 2.94, + "learning_rate": 9.79738394460118e-07, + "loss": 0.157, + "step": 76502 + }, + { + "epoch": 2.94, + "learning_rate": 9.790972044113877e-07, + "loss": 0.3437, + "step": 76503 + }, + { + "epoch": 2.94, + "learning_rate": 9.78456014362657e-07, + "loss": 0.2177, + "step": 76504 + }, + { + "epoch": 2.94, + "learning_rate": 9.778148243139267e-07, + "loss": 0.2206, + "step": 76505 + }, + { + "epoch": 2.94, + "learning_rate": 9.771736342651963e-07, + "loss": 0.1495, + "step": 76506 + }, + { + "epoch": 2.94, + "learning_rate": 9.765324442164659e-07, + "loss": 0.302, + "step": 76507 + }, + { + "epoch": 2.94, + "learning_rate": 9.758912541677355e-07, + "loss": 0.1529, + "step": 76508 + }, + { + "epoch": 2.94, + "learning_rate": 9.752500641190049e-07, + "loss": 0.0941, + "step": 76509 + }, + { + "epoch": 2.94, + "learning_rate": 9.746088740702745e-07, + "loss": 0.2368, + "step": 76510 + }, + { + "epoch": 2.94, + "learning_rate": 9.73967684021544e-07, + "loss": 0.1942, + "step": 76511 + }, + { + "epoch": 2.94, + "learning_rate": 9.733264939728137e-07, + "loss": 0.2944, + "step": 76512 + }, + { + "epoch": 2.94, + "learning_rate": 9.726853039240833e-07, + "loss": 0.1653, + "step": 76513 + }, + { + "epoch": 2.94, + "learning_rate": 9.720441138753527e-07, + "loss": 0.1353, + "step": 76514 + }, + { + "epoch": 2.94, + "learning_rate": 9.714029238266223e-07, + "loss": 0.157, + "step": 76515 + }, + { + "epoch": 2.94, + "learning_rate": 9.707617337778918e-07, + "loss": 0.1541, + "step": 76516 + }, + { + "epoch": 2.94, + "learning_rate": 9.701205437291614e-07, + "loss": 0.1563, + "step": 76517 + }, + { + "epoch": 2.94, + "learning_rate": 9.69479353680431e-07, + "loss": 0.327, + "step": 76518 + }, + { + "epoch": 2.94, + "learning_rate": 9.688381636317004e-07, + "loss": 0.2302, + "step": 76519 + }, + { + "epoch": 2.94, + "learning_rate": 9.6819697358297e-07, + "loss": 0.2064, + "step": 76520 + }, + { + "epoch": 2.94, + "learning_rate": 9.675557835342396e-07, + "loss": 0.2769, + "step": 76521 + }, + { + "epoch": 2.94, + "learning_rate": 9.669145934855092e-07, + "loss": 0.122, + "step": 76522 + }, + { + "epoch": 2.94, + "learning_rate": 9.662734034367788e-07, + "loss": 0.1577, + "step": 76523 + }, + { + "epoch": 2.94, + "learning_rate": 9.656322133880482e-07, + "loss": 0.0729, + "step": 76524 + }, + { + "epoch": 2.94, + "learning_rate": 9.649910233393178e-07, + "loss": 0.2983, + "step": 76525 + }, + { + "epoch": 2.94, + "learning_rate": 9.643498332905874e-07, + "loss": 0.1896, + "step": 76526 + }, + { + "epoch": 2.94, + "learning_rate": 9.63708643241857e-07, + "loss": 0.2351, + "step": 76527 + }, + { + "epoch": 2.94, + "learning_rate": 9.630674531931266e-07, + "loss": 0.2547, + "step": 76528 + }, + { + "epoch": 2.94, + "learning_rate": 9.62426263144396e-07, + "loss": 0.3128, + "step": 76529 + }, + { + "epoch": 2.94, + "learning_rate": 9.617850730956656e-07, + "loss": 0.102, + "step": 76530 + }, + { + "epoch": 2.94, + "learning_rate": 9.611438830469352e-07, + "loss": 0.1419, + "step": 76531 + }, + { + "epoch": 2.94, + "learning_rate": 9.605026929982048e-07, + "loss": 0.325, + "step": 76532 + }, + { + "epoch": 2.94, + "learning_rate": 9.598615029494744e-07, + "loss": 0.2531, + "step": 76533 + }, + { + "epoch": 2.94, + "learning_rate": 9.592203129007438e-07, + "loss": 0.2294, + "step": 76534 + }, + { + "epoch": 2.94, + "learning_rate": 9.585791228520134e-07, + "loss": 0.2651, + "step": 76535 + }, + { + "epoch": 2.94, + "learning_rate": 9.57937932803283e-07, + "loss": 0.2603, + "step": 76536 + }, + { + "epoch": 2.94, + "learning_rate": 9.572967427545526e-07, + "loss": 0.24, + "step": 76537 + }, + { + "epoch": 2.94, + "learning_rate": 9.566555527058222e-07, + "loss": 0.1576, + "step": 76538 + }, + { + "epoch": 2.94, + "learning_rate": 9.560143626570916e-07, + "loss": 0.3637, + "step": 76539 + }, + { + "epoch": 2.94, + "learning_rate": 9.553731726083612e-07, + "loss": 0.3206, + "step": 76540 + }, + { + "epoch": 2.94, + "learning_rate": 9.547319825596308e-07, + "loss": 0.1676, + "step": 76541 + }, + { + "epoch": 2.94, + "learning_rate": 9.540907925109004e-07, + "loss": 0.0716, + "step": 76542 + }, + { + "epoch": 2.94, + "learning_rate": 9.5344960246217e-07, + "loss": 0.1251, + "step": 76543 + }, + { + "epoch": 2.94, + "learning_rate": 9.528084124134394e-07, + "loss": 0.3093, + "step": 76544 + }, + { + "epoch": 2.94, + "learning_rate": 9.52167222364709e-07, + "loss": 0.1673, + "step": 76545 + }, + { + "epoch": 2.94, + "learning_rate": 9.515260323159786e-07, + "loss": 0.1139, + "step": 76546 + }, + { + "epoch": 2.94, + "learning_rate": 9.508848422672481e-07, + "loss": 0.1556, + "step": 76547 + }, + { + "epoch": 2.94, + "learning_rate": 9.502436522185175e-07, + "loss": 0.2, + "step": 76548 + }, + { + "epoch": 2.94, + "learning_rate": 9.496024621697871e-07, + "loss": 0.2816, + "step": 76549 + }, + { + "epoch": 2.94, + "learning_rate": 9.489612721210567e-07, + "loss": 0.173, + "step": 76550 + }, + { + "epoch": 2.94, + "learning_rate": 9.483200820723263e-07, + "loss": 0.2299, + "step": 76551 + }, + { + "epoch": 2.94, + "learning_rate": 9.476788920235959e-07, + "loss": 0.1813, + "step": 76552 + }, + { + "epoch": 2.94, + "learning_rate": 9.470377019748653e-07, + "loss": 0.2091, + "step": 76553 + }, + { + "epoch": 2.94, + "learning_rate": 9.463965119261349e-07, + "loss": 0.1506, + "step": 76554 + }, + { + "epoch": 2.94, + "learning_rate": 9.457553218774045e-07, + "loss": 0.202, + "step": 76555 + }, + { + "epoch": 2.94, + "learning_rate": 9.451141318286741e-07, + "loss": 0.0947, + "step": 76556 + }, + { + "epoch": 2.94, + "learning_rate": 9.444729417799437e-07, + "loss": 0.3061, + "step": 76557 + }, + { + "epoch": 2.94, + "learning_rate": 9.438317517312131e-07, + "loss": 0.1592, + "step": 76558 + }, + { + "epoch": 2.94, + "learning_rate": 9.431905616824827e-07, + "loss": 0.1315, + "step": 76559 + }, + { + "epoch": 2.94, + "learning_rate": 9.425493716337523e-07, + "loss": 0.1983, + "step": 76560 + }, + { + "epoch": 2.94, + "learning_rate": 9.419081815850219e-07, + "loss": 0.1367, + "step": 76561 + }, + { + "epoch": 2.94, + "learning_rate": 9.412669915362915e-07, + "loss": 0.2643, + "step": 76562 + }, + { + "epoch": 2.94, + "learning_rate": 9.406258014875609e-07, + "loss": 0.2225, + "step": 76563 + }, + { + "epoch": 2.94, + "learning_rate": 9.399846114388305e-07, + "loss": 0.1864, + "step": 76564 + }, + { + "epoch": 2.94, + "learning_rate": 9.393434213901001e-07, + "loss": 0.26, + "step": 76565 + }, + { + "epoch": 2.94, + "learning_rate": 9.387022313413697e-07, + "loss": 0.2436, + "step": 76566 + }, + { + "epoch": 2.94, + "learning_rate": 9.380610412926393e-07, + "loss": 0.2857, + "step": 76567 + }, + { + "epoch": 2.94, + "learning_rate": 9.374198512439087e-07, + "loss": 0.1456, + "step": 76568 + }, + { + "epoch": 2.94, + "learning_rate": 9.367786611951783e-07, + "loss": 0.3222, + "step": 76569 + }, + { + "epoch": 2.94, + "learning_rate": 9.361374711464479e-07, + "loss": 0.2016, + "step": 76570 + }, + { + "epoch": 2.94, + "learning_rate": 9.354962810977175e-07, + "loss": 0.3204, + "step": 76571 + }, + { + "epoch": 2.94, + "learning_rate": 9.348550910489871e-07, + "loss": 0.1824, + "step": 76572 + }, + { + "epoch": 2.94, + "learning_rate": 9.342139010002565e-07, + "loss": 0.2669, + "step": 76573 + }, + { + "epoch": 2.94, + "learning_rate": 9.335727109515261e-07, + "loss": 0.2109, + "step": 76574 + }, + { + "epoch": 2.94, + "learning_rate": 9.329315209027957e-07, + "loss": 0.1875, + "step": 76575 + }, + { + "epoch": 2.94, + "learning_rate": 9.322903308540653e-07, + "loss": 0.1316, + "step": 76576 + }, + { + "epoch": 2.94, + "learning_rate": 9.316491408053349e-07, + "loss": 0.2146, + "step": 76577 + }, + { + "epoch": 2.94, + "learning_rate": 9.310079507566042e-07, + "loss": 0.1659, + "step": 76578 + }, + { + "epoch": 2.94, + "learning_rate": 9.303667607078738e-07, + "loss": 0.2332, + "step": 76579 + }, + { + "epoch": 2.94, + "learning_rate": 9.297255706591434e-07, + "loss": 0.2228, + "step": 76580 + }, + { + "epoch": 2.94, + "learning_rate": 9.29084380610413e-07, + "loss": 0.2509, + "step": 76581 + }, + { + "epoch": 2.94, + "learning_rate": 9.284431905616826e-07, + "loss": 0.3305, + "step": 76582 + }, + { + "epoch": 2.94, + "learning_rate": 9.27802000512952e-07, + "loss": 0.1948, + "step": 76583 + }, + { + "epoch": 2.94, + "learning_rate": 9.271608104642216e-07, + "loss": 0.2747, + "step": 76584 + }, + { + "epoch": 2.94, + "learning_rate": 9.265196204154912e-07, + "loss": 0.1517, + "step": 76585 + }, + { + "epoch": 2.94, + "learning_rate": 9.258784303667608e-07, + "loss": 0.1155, + "step": 76586 + }, + { + "epoch": 2.94, + "learning_rate": 9.252372403180304e-07, + "loss": 0.1239, + "step": 76587 + }, + { + "epoch": 2.94, + "learning_rate": 9.245960502692998e-07, + "loss": 0.367, + "step": 76588 + }, + { + "epoch": 2.94, + "learning_rate": 9.239548602205694e-07, + "loss": 0.25, + "step": 76589 + }, + { + "epoch": 2.94, + "learning_rate": 9.23313670171839e-07, + "loss": 0.3162, + "step": 76590 + }, + { + "epoch": 2.94, + "learning_rate": 9.226724801231086e-07, + "loss": 0.122, + "step": 76591 + }, + { + "epoch": 2.94, + "learning_rate": 9.220312900743782e-07, + "loss": 0.2554, + "step": 76592 + }, + { + "epoch": 2.94, + "learning_rate": 9.213901000256476e-07, + "loss": 0.2016, + "step": 76593 + }, + { + "epoch": 2.94, + "learning_rate": 9.207489099769172e-07, + "loss": 0.1266, + "step": 76594 + }, + { + "epoch": 2.94, + "learning_rate": 9.201077199281868e-07, + "loss": 0.1195, + "step": 76595 + }, + { + "epoch": 2.94, + "learning_rate": 9.194665298794564e-07, + "loss": 0.1957, + "step": 76596 + }, + { + "epoch": 2.94, + "learning_rate": 9.18825339830726e-07, + "loss": 0.0701, + "step": 76597 + }, + { + "epoch": 2.94, + "learning_rate": 9.181841497819954e-07, + "loss": 0.2022, + "step": 76598 + }, + { + "epoch": 2.94, + "learning_rate": 9.17542959733265e-07, + "loss": 0.1553, + "step": 76599 + }, + { + "epoch": 2.95, + "learning_rate": 9.169017696845346e-07, + "loss": 0.2015, + "step": 76600 + }, + { + "epoch": 2.95, + "learning_rate": 9.162605796358042e-07, + "loss": 0.2661, + "step": 76601 + }, + { + "epoch": 2.95, + "learning_rate": 9.156193895870738e-07, + "loss": 0.1263, + "step": 76602 + }, + { + "epoch": 2.95, + "learning_rate": 9.149781995383432e-07, + "loss": 0.1957, + "step": 76603 + }, + { + "epoch": 2.95, + "learning_rate": 9.143370094896128e-07, + "loss": 0.0988, + "step": 76604 + }, + { + "epoch": 2.95, + "learning_rate": 9.136958194408824e-07, + "loss": 0.1066, + "step": 76605 + }, + { + "epoch": 2.95, + "learning_rate": 9.13054629392152e-07, + "loss": 0.1474, + "step": 76606 + }, + { + "epoch": 2.95, + "learning_rate": 9.124134393434213e-07, + "loss": 0.2043, + "step": 76607 + }, + { + "epoch": 2.95, + "learning_rate": 9.117722492946909e-07, + "loss": 0.1523, + "step": 76608 + }, + { + "epoch": 2.95, + "learning_rate": 9.111310592459605e-07, + "loss": 0.2869, + "step": 76609 + }, + { + "epoch": 2.95, + "learning_rate": 9.104898691972301e-07, + "loss": 0.1393, + "step": 76610 + }, + { + "epoch": 2.95, + "learning_rate": 9.098486791484997e-07, + "loss": 0.2298, + "step": 76611 + }, + { + "epoch": 2.95, + "learning_rate": 9.092074890997691e-07, + "loss": 0.2652, + "step": 76612 + }, + { + "epoch": 2.95, + "learning_rate": 9.085662990510387e-07, + "loss": 0.1597, + "step": 76613 + }, + { + "epoch": 2.95, + "learning_rate": 9.079251090023083e-07, + "loss": 0.3045, + "step": 76614 + }, + { + "epoch": 2.95, + "learning_rate": 9.072839189535779e-07, + "loss": 0.2019, + "step": 76615 + }, + { + "epoch": 2.95, + "learning_rate": 9.066427289048475e-07, + "loss": 0.2709, + "step": 76616 + }, + { + "epoch": 2.95, + "learning_rate": 9.060015388561169e-07, + "loss": 0.2023, + "step": 76617 + }, + { + "epoch": 2.95, + "learning_rate": 9.053603488073865e-07, + "loss": 0.3021, + "step": 76618 + }, + { + "epoch": 2.95, + "learning_rate": 9.047191587586561e-07, + "loss": 0.2644, + "step": 76619 + }, + { + "epoch": 2.95, + "learning_rate": 9.040779687099257e-07, + "loss": 0.2665, + "step": 76620 + }, + { + "epoch": 2.95, + "learning_rate": 9.034367786611953e-07, + "loss": 0.1805, + "step": 76621 + }, + { + "epoch": 2.95, + "learning_rate": 9.027955886124647e-07, + "loss": 0.2151, + "step": 76622 + }, + { + "epoch": 2.95, + "learning_rate": 9.021543985637343e-07, + "loss": 0.3202, + "step": 76623 + }, + { + "epoch": 2.95, + "learning_rate": 9.015132085150039e-07, + "loss": 0.1117, + "step": 76624 + }, + { + "epoch": 2.95, + "learning_rate": 9.008720184662735e-07, + "loss": 0.3766, + "step": 76625 + }, + { + "epoch": 2.95, + "learning_rate": 9.002308284175431e-07, + "loss": 0.3174, + "step": 76626 + }, + { + "epoch": 2.95, + "learning_rate": 8.995896383688125e-07, + "loss": 0.1529, + "step": 76627 + }, + { + "epoch": 2.95, + "learning_rate": 8.989484483200821e-07, + "loss": 0.2693, + "step": 76628 + }, + { + "epoch": 2.95, + "learning_rate": 8.983072582713517e-07, + "loss": 0.1212, + "step": 76629 + }, + { + "epoch": 2.95, + "learning_rate": 8.976660682226213e-07, + "loss": 0.1477, + "step": 76630 + }, + { + "epoch": 2.95, + "learning_rate": 8.970248781738909e-07, + "loss": 0.1883, + "step": 76631 + }, + { + "epoch": 2.95, + "learning_rate": 8.963836881251603e-07, + "loss": 0.1707, + "step": 76632 + }, + { + "epoch": 2.95, + "learning_rate": 8.957424980764299e-07, + "loss": 0.1991, + "step": 76633 + }, + { + "epoch": 2.95, + "learning_rate": 8.951013080276995e-07, + "loss": 0.17, + "step": 76634 + }, + { + "epoch": 2.95, + "learning_rate": 8.944601179789691e-07, + "loss": 0.1468, + "step": 76635 + }, + { + "epoch": 2.95, + "learning_rate": 8.938189279302387e-07, + "loss": 0.1952, + "step": 76636 + }, + { + "epoch": 2.95, + "learning_rate": 8.93177737881508e-07, + "loss": 0.2368, + "step": 76637 + }, + { + "epoch": 2.95, + "learning_rate": 8.925365478327776e-07, + "loss": 0.1268, + "step": 76638 + }, + { + "epoch": 2.95, + "learning_rate": 8.918953577840472e-07, + "loss": 0.1351, + "step": 76639 + }, + { + "epoch": 2.95, + "learning_rate": 8.912541677353168e-07, + "loss": 0.2593, + "step": 76640 + }, + { + "epoch": 2.95, + "learning_rate": 8.906129776865864e-07, + "loss": 0.2312, + "step": 76641 + }, + { + "epoch": 2.95, + "learning_rate": 8.899717876378558e-07, + "loss": 0.1502, + "step": 76642 + }, + { + "epoch": 2.95, + "learning_rate": 8.893305975891254e-07, + "loss": 0.2976, + "step": 76643 + }, + { + "epoch": 2.95, + "learning_rate": 8.88689407540395e-07, + "loss": 0.1638, + "step": 76644 + }, + { + "epoch": 2.95, + "learning_rate": 8.880482174916646e-07, + "loss": 0.216, + "step": 76645 + }, + { + "epoch": 2.95, + "learning_rate": 8.874070274429342e-07, + "loss": 0.1709, + "step": 76646 + }, + { + "epoch": 2.95, + "learning_rate": 8.867658373942036e-07, + "loss": 0.1363, + "step": 76647 + }, + { + "epoch": 2.95, + "learning_rate": 8.861246473454732e-07, + "loss": 0.1494, + "step": 76648 + }, + { + "epoch": 2.95, + "learning_rate": 8.854834572967428e-07, + "loss": 0.1335, + "step": 76649 + }, + { + "epoch": 2.95, + "learning_rate": 8.848422672480124e-07, + "loss": 0.1492, + "step": 76650 + }, + { + "epoch": 2.95, + "learning_rate": 8.84201077199282e-07, + "loss": 0.2299, + "step": 76651 + }, + { + "epoch": 2.95, + "learning_rate": 8.835598871505514e-07, + "loss": 0.2954, + "step": 76652 + }, + { + "epoch": 2.95, + "learning_rate": 8.82918697101821e-07, + "loss": 0.1987, + "step": 76653 + }, + { + "epoch": 2.95, + "learning_rate": 8.822775070530906e-07, + "loss": 0.2476, + "step": 76654 + }, + { + "epoch": 2.95, + "learning_rate": 8.816363170043602e-07, + "loss": 0.1465, + "step": 76655 + }, + { + "epoch": 2.95, + "learning_rate": 8.809951269556298e-07, + "loss": 0.1686, + "step": 76656 + }, + { + "epoch": 2.95, + "learning_rate": 8.803539369068992e-07, + "loss": 0.198, + "step": 76657 + }, + { + "epoch": 2.95, + "learning_rate": 8.797127468581688e-07, + "loss": 0.1321, + "step": 76658 + }, + { + "epoch": 2.95, + "learning_rate": 8.790715568094384e-07, + "loss": 0.2276, + "step": 76659 + }, + { + "epoch": 2.95, + "learning_rate": 8.78430366760708e-07, + "loss": 0.1501, + "step": 76660 + }, + { + "epoch": 2.95, + "learning_rate": 8.777891767119776e-07, + "loss": 0.1253, + "step": 76661 + }, + { + "epoch": 2.95, + "learning_rate": 8.77147986663247e-07, + "loss": 0.2455, + "step": 76662 + }, + { + "epoch": 2.95, + "learning_rate": 8.765067966145166e-07, + "loss": 0.1689, + "step": 76663 + }, + { + "epoch": 2.95, + "learning_rate": 8.758656065657862e-07, + "loss": 0.0987, + "step": 76664 + }, + { + "epoch": 2.95, + "learning_rate": 8.752244165170558e-07, + "loss": 0.1042, + "step": 76665 + }, + { + "epoch": 2.95, + "learning_rate": 8.745832264683251e-07, + "loss": 0.2069, + "step": 76666 + }, + { + "epoch": 2.95, + "learning_rate": 8.739420364195947e-07, + "loss": 0.1734, + "step": 76667 + }, + { + "epoch": 2.95, + "learning_rate": 8.733008463708643e-07, + "loss": 0.1647, + "step": 76668 + }, + { + "epoch": 2.95, + "learning_rate": 8.726596563221339e-07, + "loss": 0.2648, + "step": 76669 + }, + { + "epoch": 2.95, + "learning_rate": 8.720184662734035e-07, + "loss": 0.2038, + "step": 76670 + }, + { + "epoch": 2.95, + "learning_rate": 8.713772762246729e-07, + "loss": 0.2065, + "step": 76671 + }, + { + "epoch": 2.95, + "learning_rate": 8.707360861759425e-07, + "loss": 0.118, + "step": 76672 + }, + { + "epoch": 2.95, + "learning_rate": 8.700948961272121e-07, + "loss": 0.1764, + "step": 76673 + }, + { + "epoch": 2.95, + "learning_rate": 8.694537060784817e-07, + "loss": 0.2901, + "step": 76674 + }, + { + "epoch": 2.95, + "learning_rate": 8.688125160297513e-07, + "loss": 0.1078, + "step": 76675 + }, + { + "epoch": 2.95, + "learning_rate": 8.681713259810207e-07, + "loss": 0.1792, + "step": 76676 + }, + { + "epoch": 2.95, + "learning_rate": 8.675301359322903e-07, + "loss": 0.138, + "step": 76677 + }, + { + "epoch": 2.95, + "learning_rate": 8.668889458835599e-07, + "loss": 0.1497, + "step": 76678 + }, + { + "epoch": 2.95, + "learning_rate": 8.662477558348295e-07, + "loss": 0.1208, + "step": 76679 + }, + { + "epoch": 2.95, + "learning_rate": 8.656065657860991e-07, + "loss": 0.2118, + "step": 76680 + }, + { + "epoch": 2.95, + "learning_rate": 8.649653757373685e-07, + "loss": 0.2078, + "step": 76681 + }, + { + "epoch": 2.95, + "learning_rate": 8.643241856886381e-07, + "loss": 0.1165, + "step": 76682 + }, + { + "epoch": 2.95, + "learning_rate": 8.636829956399077e-07, + "loss": 0.1903, + "step": 76683 + }, + { + "epoch": 2.95, + "learning_rate": 8.630418055911773e-07, + "loss": 0.1699, + "step": 76684 + }, + { + "epoch": 2.95, + "learning_rate": 8.624006155424469e-07, + "loss": 0.2658, + "step": 76685 + }, + { + "epoch": 2.95, + "learning_rate": 8.617594254937163e-07, + "loss": 0.2732, + "step": 76686 + }, + { + "epoch": 2.95, + "learning_rate": 8.611182354449859e-07, + "loss": 0.1869, + "step": 76687 + }, + { + "epoch": 2.95, + "learning_rate": 8.604770453962555e-07, + "loss": 0.1825, + "step": 76688 + }, + { + "epoch": 2.95, + "learning_rate": 8.598358553475251e-07, + "loss": 0.2504, + "step": 76689 + }, + { + "epoch": 2.95, + "learning_rate": 8.591946652987947e-07, + "loss": 0.2851, + "step": 76690 + }, + { + "epoch": 2.95, + "learning_rate": 8.585534752500641e-07, + "loss": 0.1762, + "step": 76691 + }, + { + "epoch": 2.95, + "learning_rate": 8.579122852013337e-07, + "loss": 0.2241, + "step": 76692 + }, + { + "epoch": 2.95, + "learning_rate": 8.572710951526033e-07, + "loss": 0.2096, + "step": 76693 + }, + { + "epoch": 2.95, + "learning_rate": 8.566299051038729e-07, + "loss": 0.2366, + "step": 76694 + }, + { + "epoch": 2.95, + "learning_rate": 8.559887150551425e-07, + "loss": 0.2148, + "step": 76695 + }, + { + "epoch": 2.95, + "learning_rate": 8.553475250064118e-07, + "loss": 0.2165, + "step": 76696 + }, + { + "epoch": 2.95, + "learning_rate": 8.547063349576814e-07, + "loss": 0.1526, + "step": 76697 + }, + { + "epoch": 2.95, + "learning_rate": 8.54065144908951e-07, + "loss": 0.2648, + "step": 76698 + }, + { + "epoch": 2.95, + "learning_rate": 8.534239548602206e-07, + "loss": 0.166, + "step": 76699 + }, + { + "epoch": 2.95, + "learning_rate": 8.527827648114902e-07, + "loss": 0.1927, + "step": 76700 + }, + { + "epoch": 2.95, + "learning_rate": 8.521415747627596e-07, + "loss": 0.2765, + "step": 76701 + }, + { + "epoch": 2.95, + "learning_rate": 8.515003847140292e-07, + "loss": 0.0784, + "step": 76702 + }, + { + "epoch": 2.95, + "learning_rate": 8.508591946652988e-07, + "loss": 0.2103, + "step": 76703 + }, + { + "epoch": 2.95, + "learning_rate": 8.502180046165684e-07, + "loss": 0.219, + "step": 76704 + }, + { + "epoch": 2.95, + "learning_rate": 8.49576814567838e-07, + "loss": 0.4489, + "step": 76705 + }, + { + "epoch": 2.95, + "learning_rate": 8.489356245191074e-07, + "loss": 0.1556, + "step": 76706 + }, + { + "epoch": 2.95, + "learning_rate": 8.48294434470377e-07, + "loss": 0.1184, + "step": 76707 + }, + { + "epoch": 2.95, + "learning_rate": 8.476532444216466e-07, + "loss": 0.1924, + "step": 76708 + }, + { + "epoch": 2.95, + "learning_rate": 8.470120543729162e-07, + "loss": 0.1062, + "step": 76709 + }, + { + "epoch": 2.95, + "learning_rate": 8.463708643241858e-07, + "loss": 0.1676, + "step": 76710 + }, + { + "epoch": 2.95, + "learning_rate": 8.457296742754552e-07, + "loss": 0.2357, + "step": 76711 + }, + { + "epoch": 2.95, + "learning_rate": 8.450884842267248e-07, + "loss": 0.2076, + "step": 76712 + }, + { + "epoch": 2.95, + "learning_rate": 8.444472941779944e-07, + "loss": 0.1068, + "step": 76713 + }, + { + "epoch": 2.95, + "learning_rate": 8.43806104129264e-07, + "loss": 0.1393, + "step": 76714 + }, + { + "epoch": 2.95, + "learning_rate": 8.431649140805336e-07, + "loss": 0.14, + "step": 76715 + }, + { + "epoch": 2.95, + "learning_rate": 8.42523724031803e-07, + "loss": 0.0931, + "step": 76716 + }, + { + "epoch": 2.95, + "learning_rate": 8.418825339830726e-07, + "loss": 0.1976, + "step": 76717 + }, + { + "epoch": 2.95, + "learning_rate": 8.412413439343422e-07, + "loss": 0.1616, + "step": 76718 + }, + { + "epoch": 2.95, + "learning_rate": 8.406001538856118e-07, + "loss": 0.1947, + "step": 76719 + }, + { + "epoch": 2.95, + "learning_rate": 8.399589638368814e-07, + "loss": 0.2301, + "step": 76720 + }, + { + "epoch": 2.95, + "learning_rate": 8.393177737881508e-07, + "loss": 0.3686, + "step": 76721 + }, + { + "epoch": 2.95, + "learning_rate": 8.386765837394204e-07, + "loss": 0.1603, + "step": 76722 + }, + { + "epoch": 2.95, + "learning_rate": 8.3803539369069e-07, + "loss": 0.1361, + "step": 76723 + }, + { + "epoch": 2.95, + "learning_rate": 8.373942036419596e-07, + "loss": 0.239, + "step": 76724 + }, + { + "epoch": 2.95, + "learning_rate": 8.367530135932289e-07, + "loss": 0.201, + "step": 76725 + }, + { + "epoch": 2.95, + "learning_rate": 8.361118235444985e-07, + "loss": 0.1298, + "step": 76726 + }, + { + "epoch": 2.95, + "learning_rate": 8.354706334957681e-07, + "loss": 0.3505, + "step": 76727 + }, + { + "epoch": 2.95, + "learning_rate": 8.348294434470377e-07, + "loss": 0.2049, + "step": 76728 + }, + { + "epoch": 2.95, + "learning_rate": 8.341882533983073e-07, + "loss": 0.1852, + "step": 76729 + }, + { + "epoch": 2.95, + "learning_rate": 8.335470633495767e-07, + "loss": 0.2814, + "step": 76730 + }, + { + "epoch": 2.95, + "learning_rate": 8.329058733008463e-07, + "loss": 0.3395, + "step": 76731 + }, + { + "epoch": 2.95, + "learning_rate": 8.322646832521159e-07, + "loss": 0.252, + "step": 76732 + }, + { + "epoch": 2.95, + "learning_rate": 8.316234932033855e-07, + "loss": 0.2833, + "step": 76733 + }, + { + "epoch": 2.95, + "learning_rate": 8.309823031546551e-07, + "loss": 0.3493, + "step": 76734 + }, + { + "epoch": 2.95, + "learning_rate": 8.303411131059245e-07, + "loss": 0.2283, + "step": 76735 + }, + { + "epoch": 2.95, + "learning_rate": 8.296999230571941e-07, + "loss": 0.2288, + "step": 76736 + }, + { + "epoch": 2.95, + "learning_rate": 8.290587330084637e-07, + "loss": 0.1792, + "step": 76737 + }, + { + "epoch": 2.95, + "learning_rate": 8.284175429597333e-07, + "loss": 0.3627, + "step": 76738 + }, + { + "epoch": 2.95, + "learning_rate": 8.277763529110029e-07, + "loss": 0.1661, + "step": 76739 + }, + { + "epoch": 2.95, + "learning_rate": 8.271351628622723e-07, + "loss": 0.2535, + "step": 76740 + }, + { + "epoch": 2.95, + "learning_rate": 8.264939728135419e-07, + "loss": 0.16, + "step": 76741 + }, + { + "epoch": 2.95, + "learning_rate": 8.258527827648115e-07, + "loss": 0.2622, + "step": 76742 + }, + { + "epoch": 2.95, + "learning_rate": 8.252115927160811e-07, + "loss": 0.2382, + "step": 76743 + }, + { + "epoch": 2.95, + "learning_rate": 8.245704026673507e-07, + "loss": 0.5349, + "step": 76744 + }, + { + "epoch": 2.95, + "learning_rate": 8.239292126186201e-07, + "loss": 0.2979, + "step": 76745 + }, + { + "epoch": 2.95, + "learning_rate": 8.232880225698897e-07, + "loss": 0.2758, + "step": 76746 + }, + { + "epoch": 2.95, + "learning_rate": 8.226468325211593e-07, + "loss": 0.1421, + "step": 76747 + }, + { + "epoch": 2.95, + "learning_rate": 8.220056424724289e-07, + "loss": 0.138, + "step": 76748 + }, + { + "epoch": 2.95, + "learning_rate": 8.213644524236985e-07, + "loss": 0.1281, + "step": 76749 + }, + { + "epoch": 2.95, + "learning_rate": 8.207232623749679e-07, + "loss": 0.2554, + "step": 76750 + }, + { + "epoch": 2.95, + "learning_rate": 8.200820723262375e-07, + "loss": 0.2079, + "step": 76751 + }, + { + "epoch": 2.95, + "learning_rate": 8.194408822775071e-07, + "loss": 0.2254, + "step": 76752 + }, + { + "epoch": 2.95, + "learning_rate": 8.187996922287767e-07, + "loss": 0.1624, + "step": 76753 + }, + { + "epoch": 2.95, + "learning_rate": 8.181585021800463e-07, + "loss": 0.1107, + "step": 76754 + }, + { + "epoch": 2.95, + "learning_rate": 8.175173121313156e-07, + "loss": 0.1704, + "step": 76755 + }, + { + "epoch": 2.95, + "learning_rate": 8.168761220825852e-07, + "loss": 0.2737, + "step": 76756 + }, + { + "epoch": 2.95, + "learning_rate": 8.162349320338548e-07, + "loss": 0.1243, + "step": 76757 + }, + { + "epoch": 2.95, + "learning_rate": 8.155937419851244e-07, + "loss": 0.1027, + "step": 76758 + }, + { + "epoch": 2.95, + "learning_rate": 8.14952551936394e-07, + "loss": 0.1624, + "step": 76759 + }, + { + "epoch": 2.95, + "learning_rate": 8.143113618876634e-07, + "loss": 0.1774, + "step": 76760 + }, + { + "epoch": 2.95, + "learning_rate": 8.13670171838933e-07, + "loss": 0.2418, + "step": 76761 + }, + { + "epoch": 2.95, + "learning_rate": 8.130289817902026e-07, + "loss": 0.365, + "step": 76762 + }, + { + "epoch": 2.95, + "learning_rate": 8.123877917414722e-07, + "loss": 0.2022, + "step": 76763 + }, + { + "epoch": 2.95, + "learning_rate": 8.117466016927418e-07, + "loss": 0.0919, + "step": 76764 + }, + { + "epoch": 2.95, + "learning_rate": 8.111054116440112e-07, + "loss": 0.1867, + "step": 76765 + }, + { + "epoch": 2.95, + "learning_rate": 8.104642215952808e-07, + "loss": 0.1963, + "step": 76766 + }, + { + "epoch": 2.95, + "learning_rate": 8.098230315465504e-07, + "loss": 0.22, + "step": 76767 + }, + { + "epoch": 2.95, + "learning_rate": 8.0918184149782e-07, + "loss": 0.2298, + "step": 76768 + }, + { + "epoch": 2.95, + "learning_rate": 8.085406514490896e-07, + "loss": 0.2435, + "step": 76769 + }, + { + "epoch": 2.95, + "learning_rate": 8.078994614003591e-07, + "loss": 0.3105, + "step": 76770 + }, + { + "epoch": 2.95, + "learning_rate": 8.072582713516286e-07, + "loss": 0.2546, + "step": 76771 + }, + { + "epoch": 2.95, + "learning_rate": 8.066170813028982e-07, + "loss": 0.3233, + "step": 76772 + }, + { + "epoch": 2.95, + "learning_rate": 8.059758912541678e-07, + "loss": 0.097, + "step": 76773 + }, + { + "epoch": 2.95, + "learning_rate": 8.053347012054374e-07, + "loss": 0.204, + "step": 76774 + }, + { + "epoch": 2.95, + "learning_rate": 8.046935111567069e-07, + "loss": 0.138, + "step": 76775 + }, + { + "epoch": 2.95, + "learning_rate": 8.040523211079764e-07, + "loss": 0.4162, + "step": 76776 + }, + { + "epoch": 2.95, + "learning_rate": 8.03411131059246e-07, + "loss": 0.1711, + "step": 76777 + }, + { + "epoch": 2.95, + "learning_rate": 8.027699410105156e-07, + "loss": 0.2292, + "step": 76778 + }, + { + "epoch": 2.95, + "learning_rate": 8.021287509617851e-07, + "loss": 0.1945, + "step": 76779 + }, + { + "epoch": 2.95, + "learning_rate": 8.014875609130547e-07, + "loss": 0.2607, + "step": 76780 + }, + { + "epoch": 2.95, + "learning_rate": 8.008463708643242e-07, + "loss": 0.2205, + "step": 76781 + }, + { + "epoch": 2.95, + "learning_rate": 8.002051808155938e-07, + "loss": 0.1938, + "step": 76782 + }, + { + "epoch": 2.95, + "learning_rate": 7.995639907668634e-07, + "loss": 0.2786, + "step": 76783 + }, + { + "epoch": 2.95, + "learning_rate": 7.989228007181329e-07, + "loss": 0.2042, + "step": 76784 + }, + { + "epoch": 2.95, + "learning_rate": 7.982816106694025e-07, + "loss": 0.1111, + "step": 76785 + }, + { + "epoch": 2.95, + "learning_rate": 7.976404206206719e-07, + "loss": 0.2279, + "step": 76786 + }, + { + "epoch": 2.95, + "learning_rate": 7.969992305719415e-07, + "loss": 0.0894, + "step": 76787 + }, + { + "epoch": 2.95, + "learning_rate": 7.963580405232111e-07, + "loss": 0.2085, + "step": 76788 + }, + { + "epoch": 2.95, + "learning_rate": 7.957168504744806e-07, + "loss": 0.1435, + "step": 76789 + }, + { + "epoch": 2.95, + "learning_rate": 7.950756604257502e-07, + "loss": 0.3235, + "step": 76790 + }, + { + "epoch": 2.95, + "learning_rate": 7.944344703770197e-07, + "loss": 0.2854, + "step": 76791 + }, + { + "epoch": 2.95, + "learning_rate": 7.937932803282893e-07, + "loss": 0.2732, + "step": 76792 + }, + { + "epoch": 2.95, + "learning_rate": 7.931520902795589e-07, + "loss": 0.1888, + "step": 76793 + }, + { + "epoch": 2.95, + "learning_rate": 7.925109002308284e-07, + "loss": 0.4203, + "step": 76794 + }, + { + "epoch": 2.95, + "learning_rate": 7.91869710182098e-07, + "loss": 0.1555, + "step": 76795 + }, + { + "epoch": 2.95, + "learning_rate": 7.912285201333675e-07, + "loss": 0.3435, + "step": 76796 + }, + { + "epoch": 2.95, + "learning_rate": 7.905873300846371e-07, + "loss": 0.2855, + "step": 76797 + }, + { + "epoch": 2.95, + "learning_rate": 7.899461400359067e-07, + "loss": 0.2258, + "step": 76798 + }, + { + "epoch": 2.95, + "learning_rate": 7.893049499871762e-07, + "loss": 0.2236, + "step": 76799 + }, + { + "epoch": 2.95, + "learning_rate": 7.886637599384458e-07, + "loss": 0.2379, + "step": 76800 + }, + { + "epoch": 2.95, + "learning_rate": 7.880225698897153e-07, + "loss": 0.2772, + "step": 76801 + }, + { + "epoch": 2.95, + "learning_rate": 7.873813798409849e-07, + "loss": 0.229, + "step": 76802 + }, + { + "epoch": 2.95, + "learning_rate": 7.867401897922545e-07, + "loss": 0.2728, + "step": 76803 + }, + { + "epoch": 2.95, + "learning_rate": 7.86098999743524e-07, + "loss": 0.1977, + "step": 76804 + }, + { + "epoch": 2.95, + "learning_rate": 7.854578096947936e-07, + "loss": 0.178, + "step": 76805 + }, + { + "epoch": 2.95, + "learning_rate": 7.848166196460631e-07, + "loss": 0.2349, + "step": 76806 + }, + { + "epoch": 2.95, + "learning_rate": 7.841754295973327e-07, + "loss": 0.1336, + "step": 76807 + }, + { + "epoch": 2.95, + "learning_rate": 7.835342395486023e-07, + "loss": 0.2179, + "step": 76808 + }, + { + "epoch": 2.95, + "learning_rate": 7.828930494998718e-07, + "loss": 0.2455, + "step": 76809 + }, + { + "epoch": 2.95, + "learning_rate": 7.822518594511414e-07, + "loss": 0.1307, + "step": 76810 + }, + { + "epoch": 2.95, + "learning_rate": 7.816106694024109e-07, + "loss": 0.1276, + "step": 76811 + }, + { + "epoch": 2.95, + "learning_rate": 7.809694793536805e-07, + "loss": 0.3036, + "step": 76812 + }, + { + "epoch": 2.95, + "learning_rate": 7.803282893049501e-07, + "loss": 0.0814, + "step": 76813 + }, + { + "epoch": 2.95, + "learning_rate": 7.796870992562196e-07, + "loss": 0.2154, + "step": 76814 + }, + { + "epoch": 2.95, + "learning_rate": 7.790459092074892e-07, + "loss": 0.3123, + "step": 76815 + }, + { + "epoch": 2.95, + "learning_rate": 7.784047191587588e-07, + "loss": 0.22, + "step": 76816 + }, + { + "epoch": 2.95, + "learning_rate": 7.777635291100282e-07, + "loss": 0.2804, + "step": 76817 + }, + { + "epoch": 2.95, + "learning_rate": 7.771223390612978e-07, + "loss": 0.1176, + "step": 76818 + }, + { + "epoch": 2.95, + "learning_rate": 7.764811490125673e-07, + "loss": 0.206, + "step": 76819 + }, + { + "epoch": 2.95, + "learning_rate": 7.758399589638369e-07, + "loss": 0.2016, + "step": 76820 + }, + { + "epoch": 2.95, + "learning_rate": 7.751987689151065e-07, + "loss": 0.1354, + "step": 76821 + }, + { + "epoch": 2.95, + "learning_rate": 7.74557578866376e-07, + "loss": 0.2685, + "step": 76822 + }, + { + "epoch": 2.95, + "learning_rate": 7.739163888176456e-07, + "loss": 0.1574, + "step": 76823 + }, + { + "epoch": 2.95, + "learning_rate": 7.732751987689151e-07, + "loss": 0.1565, + "step": 76824 + }, + { + "epoch": 2.95, + "learning_rate": 7.726340087201847e-07, + "loss": 0.1989, + "step": 76825 + }, + { + "epoch": 2.95, + "learning_rate": 7.719928186714543e-07, + "loss": 0.1059, + "step": 76826 + }, + { + "epoch": 2.95, + "learning_rate": 7.713516286227238e-07, + "loss": 0.196, + "step": 76827 + }, + { + "epoch": 2.95, + "learning_rate": 7.707104385739934e-07, + "loss": 0.1629, + "step": 76828 + }, + { + "epoch": 2.95, + "learning_rate": 7.700692485252629e-07, + "loss": 0.2476, + "step": 76829 + }, + { + "epoch": 2.95, + "learning_rate": 7.694280584765325e-07, + "loss": 0.2251, + "step": 76830 + }, + { + "epoch": 2.95, + "learning_rate": 7.687868684278021e-07, + "loss": 0.2264, + "step": 76831 + }, + { + "epoch": 2.95, + "learning_rate": 7.681456783790716e-07, + "loss": 0.2402, + "step": 76832 + }, + { + "epoch": 2.95, + "learning_rate": 7.675044883303412e-07, + "loss": 0.1829, + "step": 76833 + }, + { + "epoch": 2.95, + "learning_rate": 7.668632982816107e-07, + "loss": 0.3083, + "step": 76834 + }, + { + "epoch": 2.95, + "learning_rate": 7.662221082328803e-07, + "loss": 0.2718, + "step": 76835 + }, + { + "epoch": 2.95, + "learning_rate": 7.655809181841499e-07, + "loss": 0.2555, + "step": 76836 + }, + { + "epoch": 2.95, + "learning_rate": 7.649397281354194e-07, + "loss": 0.1609, + "step": 76837 + }, + { + "epoch": 2.95, + "learning_rate": 7.64298538086689e-07, + "loss": 0.1263, + "step": 76838 + }, + { + "epoch": 2.95, + "learning_rate": 7.636573480379586e-07, + "loss": 0.2053, + "step": 76839 + }, + { + "epoch": 2.95, + "learning_rate": 7.630161579892281e-07, + "loss": 0.2122, + "step": 76840 + }, + { + "epoch": 2.95, + "learning_rate": 7.623749679404977e-07, + "loss": 0.2136, + "step": 76841 + }, + { + "epoch": 2.95, + "learning_rate": 7.617337778917672e-07, + "loss": 0.1526, + "step": 76842 + }, + { + "epoch": 2.95, + "learning_rate": 7.610925878430368e-07, + "loss": 0.2194, + "step": 76843 + }, + { + "epoch": 2.95, + "learning_rate": 7.604513977943064e-07, + "loss": 0.231, + "step": 76844 + }, + { + "epoch": 2.95, + "learning_rate": 7.598102077455759e-07, + "loss": 0.1262, + "step": 76845 + }, + { + "epoch": 2.95, + "learning_rate": 7.591690176968455e-07, + "loss": 0.1663, + "step": 76846 + }, + { + "epoch": 2.95, + "learning_rate": 7.585278276481149e-07, + "loss": 0.1193, + "step": 76847 + }, + { + "epoch": 2.95, + "learning_rate": 7.578866375993845e-07, + "loss": 0.2992, + "step": 76848 + }, + { + "epoch": 2.95, + "learning_rate": 7.572454475506541e-07, + "loss": 0.2811, + "step": 76849 + }, + { + "epoch": 2.95, + "learning_rate": 7.566042575019236e-07, + "loss": 0.1712, + "step": 76850 + }, + { + "epoch": 2.95, + "learning_rate": 7.559630674531932e-07, + "loss": 0.1089, + "step": 76851 + }, + { + "epoch": 2.95, + "learning_rate": 7.553218774044627e-07, + "loss": 0.241, + "step": 76852 + }, + { + "epoch": 2.95, + "learning_rate": 7.546806873557323e-07, + "loss": 0.2018, + "step": 76853 + }, + { + "epoch": 2.95, + "learning_rate": 7.540394973070018e-07, + "loss": 0.2752, + "step": 76854 + }, + { + "epoch": 2.95, + "learning_rate": 7.533983072582714e-07, + "loss": 0.2139, + "step": 76855 + }, + { + "epoch": 2.95, + "learning_rate": 7.52757117209541e-07, + "loss": 0.2739, + "step": 76856 + }, + { + "epoch": 2.95, + "learning_rate": 7.521159271608105e-07, + "loss": 0.1918, + "step": 76857 + }, + { + "epoch": 2.95, + "learning_rate": 7.514747371120801e-07, + "loss": 0.3849, + "step": 76858 + }, + { + "epoch": 2.95, + "learning_rate": 7.508335470633496e-07, + "loss": 0.2471, + "step": 76859 + }, + { + "epoch": 2.96, + "learning_rate": 7.501923570146192e-07, + "loss": 0.2729, + "step": 76860 + }, + { + "epoch": 2.96, + "learning_rate": 7.495511669658888e-07, + "loss": 0.1817, + "step": 76861 + }, + { + "epoch": 2.96, + "learning_rate": 7.489099769171583e-07, + "loss": 0.2045, + "step": 76862 + }, + { + "epoch": 2.96, + "learning_rate": 7.482687868684279e-07, + "loss": 0.1845, + "step": 76863 + }, + { + "epoch": 2.96, + "learning_rate": 7.476275968196974e-07, + "loss": 0.2794, + "step": 76864 + }, + { + "epoch": 2.96, + "learning_rate": 7.46986406770967e-07, + "loss": 0.1829, + "step": 76865 + }, + { + "epoch": 2.96, + "learning_rate": 7.463452167222366e-07, + "loss": 0.1396, + "step": 76866 + }, + { + "epoch": 2.96, + "learning_rate": 7.457040266735061e-07, + "loss": 0.2407, + "step": 76867 + }, + { + "epoch": 2.96, + "learning_rate": 7.450628366247757e-07, + "loss": 0.1928, + "step": 76868 + }, + { + "epoch": 2.96, + "learning_rate": 7.444216465760452e-07, + "loss": 0.1044, + "step": 76869 + }, + { + "epoch": 2.96, + "learning_rate": 7.437804565273148e-07, + "loss": 0.1147, + "step": 76870 + }, + { + "epoch": 2.96, + "learning_rate": 7.431392664785844e-07, + "loss": 0.3518, + "step": 76871 + }, + { + "epoch": 2.96, + "learning_rate": 7.424980764298539e-07, + "loss": 0.2005, + "step": 76872 + }, + { + "epoch": 2.96, + "learning_rate": 7.418568863811235e-07, + "loss": 0.1388, + "step": 76873 + }, + { + "epoch": 2.96, + "learning_rate": 7.41215696332393e-07, + "loss": 0.3924, + "step": 76874 + }, + { + "epoch": 2.96, + "learning_rate": 7.405745062836626e-07, + "loss": 0.2021, + "step": 76875 + }, + { + "epoch": 2.96, + "learning_rate": 7.399333162349322e-07, + "loss": 0.269, + "step": 76876 + }, + { + "epoch": 2.96, + "learning_rate": 7.392921261862016e-07, + "loss": 0.17, + "step": 76877 + }, + { + "epoch": 2.96, + "learning_rate": 7.386509361374712e-07, + "loss": 0.1761, + "step": 76878 + }, + { + "epoch": 2.96, + "learning_rate": 7.380097460887407e-07, + "loss": 0.178, + "step": 76879 + }, + { + "epoch": 2.96, + "learning_rate": 7.373685560400103e-07, + "loss": 0.2669, + "step": 76880 + }, + { + "epoch": 2.96, + "learning_rate": 7.367273659912798e-07, + "loss": 0.416, + "step": 76881 + }, + { + "epoch": 2.96, + "learning_rate": 7.360861759425494e-07, + "loss": 0.1489, + "step": 76882 + }, + { + "epoch": 2.96, + "learning_rate": 7.35444985893819e-07, + "loss": 0.2092, + "step": 76883 + }, + { + "epoch": 2.96, + "learning_rate": 7.348037958450885e-07, + "loss": 0.1072, + "step": 76884 + }, + { + "epoch": 2.96, + "learning_rate": 7.341626057963581e-07, + "loss": 0.2715, + "step": 76885 + }, + { + "epoch": 2.96, + "learning_rate": 7.335214157476276e-07, + "loss": 0.2121, + "step": 76886 + }, + { + "epoch": 2.96, + "learning_rate": 7.328802256988972e-07, + "loss": 0.2983, + "step": 76887 + }, + { + "epoch": 2.96, + "learning_rate": 7.322390356501668e-07, + "loss": 0.2943, + "step": 76888 + }, + { + "epoch": 2.96, + "learning_rate": 7.315978456014363e-07, + "loss": 0.2067, + "step": 76889 + }, + { + "epoch": 2.96, + "learning_rate": 7.309566555527059e-07, + "loss": 0.305, + "step": 76890 + }, + { + "epoch": 2.96, + "learning_rate": 7.303154655039754e-07, + "loss": 0.2426, + "step": 76891 + }, + { + "epoch": 2.96, + "learning_rate": 7.29674275455245e-07, + "loss": 0.0843, + "step": 76892 + }, + { + "epoch": 2.96, + "learning_rate": 7.290330854065146e-07, + "loss": 0.1719, + "step": 76893 + }, + { + "epoch": 2.96, + "learning_rate": 7.283918953577841e-07, + "loss": 0.1265, + "step": 76894 + }, + { + "epoch": 2.96, + "learning_rate": 7.277507053090537e-07, + "loss": 0.0978, + "step": 76895 + }, + { + "epoch": 2.96, + "learning_rate": 7.271095152603232e-07, + "loss": 0.2002, + "step": 76896 + }, + { + "epoch": 2.96, + "learning_rate": 7.264683252115928e-07, + "loss": 0.1538, + "step": 76897 + }, + { + "epoch": 2.96, + "learning_rate": 7.258271351628624e-07, + "loss": 0.2013, + "step": 76898 + }, + { + "epoch": 2.96, + "learning_rate": 7.251859451141319e-07, + "loss": 0.265, + "step": 76899 + }, + { + "epoch": 2.96, + "learning_rate": 7.245447550654015e-07, + "loss": 0.1769, + "step": 76900 + }, + { + "epoch": 2.96, + "learning_rate": 7.23903565016671e-07, + "loss": 0.1731, + "step": 76901 + }, + { + "epoch": 2.96, + "learning_rate": 7.232623749679406e-07, + "loss": 0.2576, + "step": 76902 + }, + { + "epoch": 2.96, + "learning_rate": 7.226211849192102e-07, + "loss": 0.2518, + "step": 76903 + }, + { + "epoch": 2.96, + "learning_rate": 7.219799948704797e-07, + "loss": 0.1705, + "step": 76904 + }, + { + "epoch": 2.96, + "learning_rate": 7.213388048217493e-07, + "loss": 0.1947, + "step": 76905 + }, + { + "epoch": 2.96, + "learning_rate": 7.206976147730187e-07, + "loss": 0.1277, + "step": 76906 + }, + { + "epoch": 2.96, + "learning_rate": 7.200564247242883e-07, + "loss": 0.2261, + "step": 76907 + }, + { + "epoch": 2.96, + "learning_rate": 7.194152346755579e-07, + "loss": 0.1714, + "step": 76908 + }, + { + "epoch": 2.96, + "learning_rate": 7.187740446268274e-07, + "loss": 0.2593, + "step": 76909 + }, + { + "epoch": 2.96, + "learning_rate": 7.18132854578097e-07, + "loss": 0.1472, + "step": 76910 + }, + { + "epoch": 2.96, + "learning_rate": 7.174916645293665e-07, + "loss": 0.2038, + "step": 76911 + }, + { + "epoch": 2.96, + "learning_rate": 7.168504744806361e-07, + "loss": 0.2957, + "step": 76912 + }, + { + "epoch": 2.96, + "learning_rate": 7.162092844319056e-07, + "loss": 0.1646, + "step": 76913 + }, + { + "epoch": 2.96, + "learning_rate": 7.155680943831752e-07, + "loss": 0.2508, + "step": 76914 + }, + { + "epoch": 2.96, + "learning_rate": 7.149269043344448e-07, + "loss": 0.2134, + "step": 76915 + }, + { + "epoch": 2.96, + "learning_rate": 7.142857142857143e-07, + "loss": 0.0932, + "step": 76916 + }, + { + "epoch": 2.96, + "learning_rate": 7.136445242369839e-07, + "loss": 0.1734, + "step": 76917 + }, + { + "epoch": 2.96, + "learning_rate": 7.130033341882534e-07, + "loss": 0.2985, + "step": 76918 + }, + { + "epoch": 2.96, + "learning_rate": 7.12362144139523e-07, + "loss": 0.1821, + "step": 76919 + }, + { + "epoch": 2.96, + "learning_rate": 7.117209540907926e-07, + "loss": 0.2377, + "step": 76920 + }, + { + "epoch": 2.96, + "learning_rate": 7.110797640420621e-07, + "loss": 0.214, + "step": 76921 + }, + { + "epoch": 2.96, + "learning_rate": 7.104385739933317e-07, + "loss": 0.2272, + "step": 76922 + }, + { + "epoch": 2.96, + "learning_rate": 7.097973839446012e-07, + "loss": 0.3024, + "step": 76923 + }, + { + "epoch": 2.96, + "learning_rate": 7.091561938958708e-07, + "loss": 0.2936, + "step": 76924 + }, + { + "epoch": 2.96, + "learning_rate": 7.085150038471404e-07, + "loss": 0.1477, + "step": 76925 + }, + { + "epoch": 2.96, + "learning_rate": 7.078738137984099e-07, + "loss": 0.2258, + "step": 76926 + }, + { + "epoch": 2.96, + "learning_rate": 7.072326237496795e-07, + "loss": 0.1894, + "step": 76927 + }, + { + "epoch": 2.96, + "learning_rate": 7.06591433700949e-07, + "loss": 0.1776, + "step": 76928 + }, + { + "epoch": 2.96, + "learning_rate": 7.059502436522186e-07, + "loss": 0.3594, + "step": 76929 + }, + { + "epoch": 2.96, + "learning_rate": 7.053090536034882e-07, + "loss": 0.1879, + "step": 76930 + }, + { + "epoch": 2.96, + "learning_rate": 7.046678635547577e-07, + "loss": 0.1782, + "step": 76931 + }, + { + "epoch": 2.96, + "learning_rate": 7.040266735060273e-07, + "loss": 0.2588, + "step": 76932 + }, + { + "epoch": 2.96, + "learning_rate": 7.033854834572968e-07, + "loss": 0.3363, + "step": 76933 + }, + { + "epoch": 2.96, + "learning_rate": 7.027442934085664e-07, + "loss": 0.1807, + "step": 76934 + }, + { + "epoch": 2.96, + "learning_rate": 7.02103103359836e-07, + "loss": 0.0592, + "step": 76935 + }, + { + "epoch": 2.96, + "learning_rate": 7.014619133111054e-07, + "loss": 0.1797, + "step": 76936 + }, + { + "epoch": 2.96, + "learning_rate": 7.00820723262375e-07, + "loss": 0.2581, + "step": 76937 + }, + { + "epoch": 2.96, + "learning_rate": 7.001795332136445e-07, + "loss": 0.179, + "step": 76938 + }, + { + "epoch": 2.96, + "learning_rate": 6.995383431649141e-07, + "loss": 0.2713, + "step": 76939 + }, + { + "epoch": 2.96, + "learning_rate": 6.988971531161836e-07, + "loss": 0.19, + "step": 76940 + }, + { + "epoch": 2.96, + "learning_rate": 6.982559630674532e-07, + "loss": 0.2301, + "step": 76941 + }, + { + "epoch": 2.96, + "learning_rate": 6.976147730187228e-07, + "loss": 0.2173, + "step": 76942 + }, + { + "epoch": 2.96, + "learning_rate": 6.969735829699923e-07, + "loss": 0.1734, + "step": 76943 + }, + { + "epoch": 2.96, + "learning_rate": 6.963323929212619e-07, + "loss": 0.1105, + "step": 76944 + }, + { + "epoch": 2.96, + "learning_rate": 6.956912028725314e-07, + "loss": 0.1588, + "step": 76945 + }, + { + "epoch": 2.96, + "learning_rate": 6.95050012823801e-07, + "loss": 0.2813, + "step": 76946 + }, + { + "epoch": 2.96, + "learning_rate": 6.944088227750706e-07, + "loss": 0.1639, + "step": 76947 + }, + { + "epoch": 2.96, + "learning_rate": 6.937676327263401e-07, + "loss": 0.2257, + "step": 76948 + }, + { + "epoch": 2.96, + "learning_rate": 6.931264426776097e-07, + "loss": 0.1701, + "step": 76949 + }, + { + "epoch": 2.96, + "learning_rate": 6.924852526288792e-07, + "loss": 0.3339, + "step": 76950 + }, + { + "epoch": 2.96, + "learning_rate": 6.918440625801488e-07, + "loss": 0.1453, + "step": 76951 + }, + { + "epoch": 2.96, + "learning_rate": 6.912028725314184e-07, + "loss": 0.2345, + "step": 76952 + }, + { + "epoch": 2.96, + "learning_rate": 6.905616824826879e-07, + "loss": 0.1983, + "step": 76953 + }, + { + "epoch": 2.96, + "learning_rate": 6.899204924339575e-07, + "loss": 0.3265, + "step": 76954 + }, + { + "epoch": 2.96, + "learning_rate": 6.89279302385227e-07, + "loss": 0.1413, + "step": 76955 + }, + { + "epoch": 2.96, + "learning_rate": 6.886381123364966e-07, + "loss": 0.1207, + "step": 76956 + }, + { + "epoch": 2.96, + "learning_rate": 6.879969222877662e-07, + "loss": 0.1059, + "step": 76957 + }, + { + "epoch": 2.96, + "learning_rate": 6.873557322390357e-07, + "loss": 0.2139, + "step": 76958 + }, + { + "epoch": 2.96, + "learning_rate": 6.867145421903053e-07, + "loss": 0.1149, + "step": 76959 + }, + { + "epoch": 2.96, + "learning_rate": 6.860733521415748e-07, + "loss": 0.3646, + "step": 76960 + }, + { + "epoch": 2.96, + "learning_rate": 6.854321620928444e-07, + "loss": 0.2765, + "step": 76961 + }, + { + "epoch": 2.96, + "learning_rate": 6.84790972044114e-07, + "loss": 0.1673, + "step": 76962 + }, + { + "epoch": 2.96, + "learning_rate": 6.841497819953835e-07, + "loss": 0.2476, + "step": 76963 + }, + { + "epoch": 2.96, + "learning_rate": 6.835085919466531e-07, + "loss": 0.1609, + "step": 76964 + }, + { + "epoch": 2.96, + "learning_rate": 6.828674018979225e-07, + "loss": 0.1793, + "step": 76965 + }, + { + "epoch": 2.96, + "learning_rate": 6.822262118491921e-07, + "loss": 0.1488, + "step": 76966 + }, + { + "epoch": 2.96, + "learning_rate": 6.815850218004617e-07, + "loss": 0.28, + "step": 76967 + }, + { + "epoch": 2.96, + "learning_rate": 6.809438317517312e-07, + "loss": 0.1153, + "step": 76968 + }, + { + "epoch": 2.96, + "learning_rate": 6.803026417030008e-07, + "loss": 0.2403, + "step": 76969 + }, + { + "epoch": 2.96, + "learning_rate": 6.796614516542703e-07, + "loss": 0.2435, + "step": 76970 + }, + { + "epoch": 2.96, + "learning_rate": 6.790202616055399e-07, + "loss": 0.2115, + "step": 76971 + }, + { + "epoch": 2.96, + "learning_rate": 6.783790715568094e-07, + "loss": 0.2642, + "step": 76972 + }, + { + "epoch": 2.96, + "learning_rate": 6.77737881508079e-07, + "loss": 0.2946, + "step": 76973 + }, + { + "epoch": 2.96, + "learning_rate": 6.770966914593486e-07, + "loss": 0.1532, + "step": 76974 + }, + { + "epoch": 2.96, + "learning_rate": 6.764555014106181e-07, + "loss": 0.0628, + "step": 76975 + }, + { + "epoch": 2.96, + "learning_rate": 6.758143113618877e-07, + "loss": 0.2723, + "step": 76976 + }, + { + "epoch": 2.96, + "learning_rate": 6.751731213131572e-07, + "loss": 0.1668, + "step": 76977 + }, + { + "epoch": 2.96, + "learning_rate": 6.745319312644268e-07, + "loss": 0.1474, + "step": 76978 + }, + { + "epoch": 2.96, + "learning_rate": 6.738907412156964e-07, + "loss": 0.1842, + "step": 76979 + }, + { + "epoch": 2.96, + "learning_rate": 6.732495511669659e-07, + "loss": 0.2029, + "step": 76980 + }, + { + "epoch": 2.96, + "learning_rate": 6.726083611182355e-07, + "loss": 0.2348, + "step": 76981 + }, + { + "epoch": 2.96, + "learning_rate": 6.71967171069505e-07, + "loss": 0.2326, + "step": 76982 + }, + { + "epoch": 2.96, + "learning_rate": 6.713259810207746e-07, + "loss": 0.2123, + "step": 76983 + }, + { + "epoch": 2.96, + "learning_rate": 6.706847909720442e-07, + "loss": 0.2435, + "step": 76984 + }, + { + "epoch": 2.96, + "learning_rate": 6.700436009233137e-07, + "loss": 0.1376, + "step": 76985 + }, + { + "epoch": 2.96, + "learning_rate": 6.694024108745833e-07, + "loss": 0.2485, + "step": 76986 + }, + { + "epoch": 2.96, + "learning_rate": 6.687612208258528e-07, + "loss": 0.1559, + "step": 76987 + }, + { + "epoch": 2.96, + "learning_rate": 6.681200307771224e-07, + "loss": 0.2602, + "step": 76988 + }, + { + "epoch": 2.96, + "learning_rate": 6.67478840728392e-07, + "loss": 0.1696, + "step": 76989 + }, + { + "epoch": 2.96, + "learning_rate": 6.668376506796615e-07, + "loss": 0.225, + "step": 76990 + }, + { + "epoch": 2.96, + "learning_rate": 6.661964606309311e-07, + "loss": 0.1927, + "step": 76991 + }, + { + "epoch": 2.96, + "learning_rate": 6.655552705822006e-07, + "loss": 0.2687, + "step": 76992 + }, + { + "epoch": 2.96, + "learning_rate": 6.649140805334702e-07, + "loss": 0.2788, + "step": 76993 + }, + { + "epoch": 2.96, + "learning_rate": 6.642728904847398e-07, + "loss": 0.3097, + "step": 76994 + }, + { + "epoch": 2.96, + "learning_rate": 6.636317004360093e-07, + "loss": 0.1108, + "step": 76995 + }, + { + "epoch": 2.96, + "learning_rate": 6.629905103872788e-07, + "loss": 0.1643, + "step": 76996 + }, + { + "epoch": 2.96, + "learning_rate": 6.623493203385483e-07, + "loss": 0.1211, + "step": 76997 + }, + { + "epoch": 2.96, + "learning_rate": 6.617081302898179e-07, + "loss": 0.0996, + "step": 76998 + }, + { + "epoch": 2.96, + "learning_rate": 6.610669402410874e-07, + "loss": 0.2002, + "step": 76999 + }, + { + "epoch": 2.96, + "learning_rate": 6.60425750192357e-07, + "loss": 0.1462, + "step": 77000 + }, + { + "epoch": 2.96, + "learning_rate": 6.597845601436266e-07, + "loss": 0.2671, + "step": 77001 + }, + { + "epoch": 2.96, + "learning_rate": 6.591433700948961e-07, + "loss": 0.2779, + "step": 77002 + }, + { + "epoch": 2.96, + "learning_rate": 6.585021800461657e-07, + "loss": 0.1469, + "step": 77003 + }, + { + "epoch": 2.96, + "learning_rate": 6.578609899974352e-07, + "loss": 0.1508, + "step": 77004 + }, + { + "epoch": 2.96, + "learning_rate": 6.572197999487048e-07, + "loss": 0.1519, + "step": 77005 + }, + { + "epoch": 2.96, + "learning_rate": 6.565786098999744e-07, + "loss": 0.1006, + "step": 77006 + }, + { + "epoch": 2.96, + "learning_rate": 6.559374198512439e-07, + "loss": 0.1937, + "step": 77007 + }, + { + "epoch": 2.96, + "learning_rate": 6.552962298025135e-07, + "loss": 0.2218, + "step": 77008 + }, + { + "epoch": 2.96, + "learning_rate": 6.54655039753783e-07, + "loss": 0.3632, + "step": 77009 + }, + { + "epoch": 2.96, + "learning_rate": 6.540138497050526e-07, + "loss": 0.1983, + "step": 77010 + }, + { + "epoch": 2.96, + "learning_rate": 6.533726596563222e-07, + "loss": 0.2755, + "step": 77011 + }, + { + "epoch": 2.96, + "learning_rate": 6.527314696075917e-07, + "loss": 0.2705, + "step": 77012 + }, + { + "epoch": 2.96, + "learning_rate": 6.520902795588613e-07, + "loss": 0.1539, + "step": 77013 + }, + { + "epoch": 2.96, + "learning_rate": 6.514490895101308e-07, + "loss": 0.2116, + "step": 77014 + }, + { + "epoch": 2.96, + "learning_rate": 6.508078994614004e-07, + "loss": 0.2326, + "step": 77015 + }, + { + "epoch": 2.96, + "learning_rate": 6.5016670941267e-07, + "loss": 0.3079, + "step": 77016 + }, + { + "epoch": 2.96, + "learning_rate": 6.495255193639395e-07, + "loss": 0.2054, + "step": 77017 + }, + { + "epoch": 2.96, + "learning_rate": 6.488843293152091e-07, + "loss": 0.1906, + "step": 77018 + }, + { + "epoch": 2.96, + "learning_rate": 6.482431392664786e-07, + "loss": 0.2534, + "step": 77019 + }, + { + "epoch": 2.96, + "learning_rate": 6.476019492177482e-07, + "loss": 0.2092, + "step": 77020 + }, + { + "epoch": 2.96, + "learning_rate": 6.469607591690178e-07, + "loss": 0.1082, + "step": 77021 + }, + { + "epoch": 2.96, + "learning_rate": 6.463195691202873e-07, + "loss": 0.2314, + "step": 77022 + }, + { + "epoch": 2.96, + "learning_rate": 6.456783790715569e-07, + "loss": 0.184, + "step": 77023 + }, + { + "epoch": 2.96, + "learning_rate": 6.450371890228264e-07, + "loss": 0.1289, + "step": 77024 + }, + { + "epoch": 2.96, + "learning_rate": 6.44395998974096e-07, + "loss": 0.1744, + "step": 77025 + }, + { + "epoch": 2.96, + "learning_rate": 6.437548089253655e-07, + "loss": 0.1619, + "step": 77026 + }, + { + "epoch": 2.96, + "learning_rate": 6.43113618876635e-07, + "loss": 0.1259, + "step": 77027 + }, + { + "epoch": 2.96, + "learning_rate": 6.424724288279046e-07, + "loss": 0.2476, + "step": 77028 + }, + { + "epoch": 2.96, + "learning_rate": 6.418312387791741e-07, + "loss": 0.1332, + "step": 77029 + }, + { + "epoch": 2.96, + "learning_rate": 6.411900487304437e-07, + "loss": 0.2755, + "step": 77030 + }, + { + "epoch": 2.96, + "learning_rate": 6.405488586817132e-07, + "loss": 0.1569, + "step": 77031 + }, + { + "epoch": 2.96, + "learning_rate": 6.399076686329828e-07, + "loss": 0.2967, + "step": 77032 + }, + { + "epoch": 2.96, + "learning_rate": 6.392664785842524e-07, + "loss": 0.1423, + "step": 77033 + }, + { + "epoch": 2.96, + "learning_rate": 6.386252885355219e-07, + "loss": 0.1376, + "step": 77034 + }, + { + "epoch": 2.96, + "learning_rate": 6.379840984867915e-07, + "loss": 0.1646, + "step": 77035 + }, + { + "epoch": 2.96, + "learning_rate": 6.37342908438061e-07, + "loss": 0.3219, + "step": 77036 + }, + { + "epoch": 2.96, + "learning_rate": 6.367017183893306e-07, + "loss": 0.2242, + "step": 77037 + }, + { + "epoch": 2.96, + "learning_rate": 6.360605283406002e-07, + "loss": 0.1896, + "step": 77038 + }, + { + "epoch": 2.96, + "learning_rate": 6.354193382918697e-07, + "loss": 0.3175, + "step": 77039 + }, + { + "epoch": 2.96, + "learning_rate": 6.347781482431393e-07, + "loss": 0.3438, + "step": 77040 + }, + { + "epoch": 2.96, + "learning_rate": 6.341369581944088e-07, + "loss": 0.2096, + "step": 77041 + }, + { + "epoch": 2.96, + "learning_rate": 6.334957681456784e-07, + "loss": 0.2395, + "step": 77042 + }, + { + "epoch": 2.96, + "learning_rate": 6.32854578096948e-07, + "loss": 0.1574, + "step": 77043 + }, + { + "epoch": 2.96, + "learning_rate": 6.322133880482175e-07, + "loss": 0.2727, + "step": 77044 + }, + { + "epoch": 2.96, + "learning_rate": 6.315721979994871e-07, + "loss": 0.1291, + "step": 77045 + }, + { + "epoch": 2.96, + "learning_rate": 6.309310079507566e-07, + "loss": 0.1227, + "step": 77046 + }, + { + "epoch": 2.96, + "learning_rate": 6.302898179020262e-07, + "loss": 0.2537, + "step": 77047 + }, + { + "epoch": 2.96, + "learning_rate": 6.296486278532958e-07, + "loss": 0.1314, + "step": 77048 + }, + { + "epoch": 2.96, + "learning_rate": 6.290074378045653e-07, + "loss": 0.1159, + "step": 77049 + }, + { + "epoch": 2.96, + "learning_rate": 6.283662477558349e-07, + "loss": 0.2374, + "step": 77050 + }, + { + "epoch": 2.96, + "learning_rate": 6.277250577071044e-07, + "loss": 0.278, + "step": 77051 + }, + { + "epoch": 2.96, + "learning_rate": 6.27083867658374e-07, + "loss": 0.139, + "step": 77052 + }, + { + "epoch": 2.96, + "learning_rate": 6.264426776096436e-07, + "loss": 0.2146, + "step": 77053 + }, + { + "epoch": 2.96, + "learning_rate": 6.25801487560913e-07, + "loss": 0.1709, + "step": 77054 + }, + { + "epoch": 2.96, + "learning_rate": 6.251602975121827e-07, + "loss": 0.3333, + "step": 77055 + }, + { + "epoch": 2.96, + "learning_rate": 6.245191074634521e-07, + "loss": 0.2173, + "step": 77056 + }, + { + "epoch": 2.96, + "learning_rate": 6.238779174147217e-07, + "loss": 0.293, + "step": 77057 + }, + { + "epoch": 2.96, + "learning_rate": 6.232367273659912e-07, + "loss": 0.2732, + "step": 77058 + }, + { + "epoch": 2.96, + "learning_rate": 6.225955373172608e-07, + "loss": 0.1952, + "step": 77059 + }, + { + "epoch": 2.96, + "learning_rate": 6.219543472685304e-07, + "loss": 0.1318, + "step": 77060 + }, + { + "epoch": 2.96, + "learning_rate": 6.213131572197999e-07, + "loss": 0.2269, + "step": 77061 + }, + { + "epoch": 2.96, + "learning_rate": 6.206719671710695e-07, + "loss": 0.2702, + "step": 77062 + }, + { + "epoch": 2.96, + "learning_rate": 6.20030777122339e-07, + "loss": 0.1746, + "step": 77063 + }, + { + "epoch": 2.96, + "learning_rate": 6.193895870736086e-07, + "loss": 0.4029, + "step": 77064 + }, + { + "epoch": 2.96, + "learning_rate": 6.187483970248782e-07, + "loss": 0.2173, + "step": 77065 + }, + { + "epoch": 2.96, + "learning_rate": 6.181072069761477e-07, + "loss": 0.2882, + "step": 77066 + }, + { + "epoch": 2.96, + "learning_rate": 6.174660169274173e-07, + "loss": 0.2392, + "step": 77067 + }, + { + "epoch": 2.96, + "learning_rate": 6.168248268786868e-07, + "loss": 0.2708, + "step": 77068 + }, + { + "epoch": 2.96, + "learning_rate": 6.161836368299564e-07, + "loss": 0.1536, + "step": 77069 + }, + { + "epoch": 2.96, + "learning_rate": 6.15542446781226e-07, + "loss": 0.2658, + "step": 77070 + }, + { + "epoch": 2.96, + "learning_rate": 6.149012567324955e-07, + "loss": 0.2448, + "step": 77071 + }, + { + "epoch": 2.96, + "learning_rate": 6.142600666837651e-07, + "loss": 0.1264, + "step": 77072 + }, + { + "epoch": 2.96, + "learning_rate": 6.136188766350346e-07, + "loss": 0.247, + "step": 77073 + }, + { + "epoch": 2.96, + "learning_rate": 6.129776865863042e-07, + "loss": 0.2551, + "step": 77074 + }, + { + "epoch": 2.96, + "learning_rate": 6.123364965375738e-07, + "loss": 0.2569, + "step": 77075 + }, + { + "epoch": 2.96, + "learning_rate": 6.116953064888433e-07, + "loss": 0.1374, + "step": 77076 + }, + { + "epoch": 2.96, + "learning_rate": 6.110541164401129e-07, + "loss": 0.3742, + "step": 77077 + }, + { + "epoch": 2.96, + "learning_rate": 6.104129263913824e-07, + "loss": 0.2074, + "step": 77078 + }, + { + "epoch": 2.96, + "learning_rate": 6.09771736342652e-07, + "loss": 0.2149, + "step": 77079 + }, + { + "epoch": 2.96, + "learning_rate": 6.091305462939216e-07, + "loss": 0.1756, + "step": 77080 + }, + { + "epoch": 2.96, + "learning_rate": 6.084893562451911e-07, + "loss": 0.1468, + "step": 77081 + }, + { + "epoch": 2.96, + "learning_rate": 6.078481661964607e-07, + "loss": 0.2377, + "step": 77082 + }, + { + "epoch": 2.96, + "learning_rate": 6.072069761477302e-07, + "loss": 0.2129, + "step": 77083 + }, + { + "epoch": 2.96, + "learning_rate": 6.065657860989998e-07, + "loss": 0.1828, + "step": 77084 + }, + { + "epoch": 2.96, + "learning_rate": 6.059245960502692e-07, + "loss": 0.2108, + "step": 77085 + }, + { + "epoch": 2.96, + "learning_rate": 6.052834060015388e-07, + "loss": 0.3217, + "step": 77086 + }, + { + "epoch": 2.96, + "learning_rate": 6.046422159528084e-07, + "loss": 0.2408, + "step": 77087 + }, + { + "epoch": 2.96, + "learning_rate": 6.040010259040779e-07, + "loss": 0.1298, + "step": 77088 + }, + { + "epoch": 2.96, + "learning_rate": 6.033598358553475e-07, + "loss": 0.1356, + "step": 77089 + }, + { + "epoch": 2.96, + "learning_rate": 6.02718645806617e-07, + "loss": 0.2074, + "step": 77090 + }, + { + "epoch": 2.96, + "learning_rate": 6.020774557578866e-07, + "loss": 0.2108, + "step": 77091 + }, + { + "epoch": 2.96, + "learning_rate": 6.014362657091562e-07, + "loss": 0.2768, + "step": 77092 + }, + { + "epoch": 2.96, + "learning_rate": 6.007950756604257e-07, + "loss": 0.2644, + "step": 77093 + }, + { + "epoch": 2.96, + "learning_rate": 6.001538856116953e-07, + "loss": 0.2056, + "step": 77094 + }, + { + "epoch": 2.96, + "learning_rate": 5.995126955629648e-07, + "loss": 0.1988, + "step": 77095 + }, + { + "epoch": 2.96, + "learning_rate": 5.988715055142344e-07, + "loss": 0.1347, + "step": 77096 + }, + { + "epoch": 2.96, + "learning_rate": 5.98230315465504e-07, + "loss": 0.1537, + "step": 77097 + }, + { + "epoch": 2.96, + "learning_rate": 5.975891254167735e-07, + "loss": 0.1982, + "step": 77098 + }, + { + "epoch": 2.96, + "learning_rate": 5.969479353680431e-07, + "loss": 0.3172, + "step": 77099 + }, + { + "epoch": 2.96, + "learning_rate": 5.963067453193126e-07, + "loss": 0.1995, + "step": 77100 + }, + { + "epoch": 2.96, + "learning_rate": 5.956655552705822e-07, + "loss": 0.2033, + "step": 77101 + }, + { + "epoch": 2.96, + "learning_rate": 5.950243652218518e-07, + "loss": 0.1676, + "step": 77102 + }, + { + "epoch": 2.96, + "learning_rate": 5.943831751731213e-07, + "loss": 0.2551, + "step": 77103 + }, + { + "epoch": 2.96, + "learning_rate": 5.937419851243909e-07, + "loss": 0.1941, + "step": 77104 + }, + { + "epoch": 2.96, + "learning_rate": 5.931007950756604e-07, + "loss": 0.2963, + "step": 77105 + }, + { + "epoch": 2.96, + "learning_rate": 5.9245960502693e-07, + "loss": 0.2265, + "step": 77106 + }, + { + "epoch": 2.96, + "learning_rate": 5.918184149781996e-07, + "loss": 0.1517, + "step": 77107 + }, + { + "epoch": 2.96, + "learning_rate": 5.911772249294691e-07, + "loss": 0.2329, + "step": 77108 + }, + { + "epoch": 2.96, + "learning_rate": 5.905360348807387e-07, + "loss": 0.3177, + "step": 77109 + }, + { + "epoch": 2.96, + "learning_rate": 5.898948448320082e-07, + "loss": 0.0888, + "step": 77110 + }, + { + "epoch": 2.96, + "learning_rate": 5.892536547832778e-07, + "loss": 0.1509, + "step": 77111 + }, + { + "epoch": 2.96, + "learning_rate": 5.886124647345474e-07, + "loss": 0.182, + "step": 77112 + }, + { + "epoch": 2.96, + "learning_rate": 5.879712746858169e-07, + "loss": 0.1753, + "step": 77113 + }, + { + "epoch": 2.96, + "learning_rate": 5.873300846370865e-07, + "loss": 0.266, + "step": 77114 + }, + { + "epoch": 2.96, + "learning_rate": 5.866888945883559e-07, + "loss": 0.2182, + "step": 77115 + }, + { + "epoch": 2.96, + "learning_rate": 5.860477045396255e-07, + "loss": 0.251, + "step": 77116 + }, + { + "epoch": 2.96, + "learning_rate": 5.85406514490895e-07, + "loss": 0.1955, + "step": 77117 + }, + { + "epoch": 2.96, + "learning_rate": 5.847653244421646e-07, + "loss": 0.187, + "step": 77118 + }, + { + "epoch": 2.96, + "learning_rate": 5.841241343934342e-07, + "loss": 0.192, + "step": 77119 + }, + { + "epoch": 2.97, + "learning_rate": 5.834829443447037e-07, + "loss": 0.4048, + "step": 77120 + }, + { + "epoch": 2.97, + "learning_rate": 5.828417542959733e-07, + "loss": 0.2241, + "step": 77121 + }, + { + "epoch": 2.97, + "learning_rate": 5.822005642472428e-07, + "loss": 0.2887, + "step": 77122 + }, + { + "epoch": 2.97, + "learning_rate": 5.815593741985124e-07, + "loss": 0.2067, + "step": 77123 + }, + { + "epoch": 2.97, + "learning_rate": 5.80918184149782e-07, + "loss": 0.148, + "step": 77124 + }, + { + "epoch": 2.97, + "learning_rate": 5.802769941010515e-07, + "loss": 0.3106, + "step": 77125 + }, + { + "epoch": 2.97, + "learning_rate": 5.796358040523211e-07, + "loss": 0.1631, + "step": 77126 + }, + { + "epoch": 2.97, + "learning_rate": 5.789946140035906e-07, + "loss": 0.2131, + "step": 77127 + }, + { + "epoch": 2.97, + "learning_rate": 5.783534239548602e-07, + "loss": 0.1645, + "step": 77128 + }, + { + "epoch": 2.97, + "learning_rate": 5.777122339061298e-07, + "loss": 0.3115, + "step": 77129 + }, + { + "epoch": 2.97, + "learning_rate": 5.770710438573993e-07, + "loss": 0.1542, + "step": 77130 + }, + { + "epoch": 2.97, + "learning_rate": 5.764298538086689e-07, + "loss": 0.1237, + "step": 77131 + }, + { + "epoch": 2.97, + "learning_rate": 5.757886637599384e-07, + "loss": 0.1669, + "step": 77132 + }, + { + "epoch": 2.97, + "learning_rate": 5.75147473711208e-07, + "loss": 0.1073, + "step": 77133 + }, + { + "epoch": 2.97, + "learning_rate": 5.745062836624776e-07, + "loss": 0.3668, + "step": 77134 + }, + { + "epoch": 2.97, + "learning_rate": 5.738650936137471e-07, + "loss": 0.1937, + "step": 77135 + }, + { + "epoch": 2.97, + "learning_rate": 5.732239035650167e-07, + "loss": 0.2043, + "step": 77136 + }, + { + "epoch": 2.97, + "learning_rate": 5.725827135162862e-07, + "loss": 0.1557, + "step": 77137 + }, + { + "epoch": 2.97, + "learning_rate": 5.719415234675558e-07, + "loss": 0.2252, + "step": 77138 + }, + { + "epoch": 2.97, + "learning_rate": 5.713003334188254e-07, + "loss": 0.1703, + "step": 77139 + }, + { + "epoch": 2.97, + "learning_rate": 5.706591433700949e-07, + "loss": 0.1739, + "step": 77140 + }, + { + "epoch": 2.97, + "learning_rate": 5.700179533213645e-07, + "loss": 0.3522, + "step": 77141 + }, + { + "epoch": 2.97, + "learning_rate": 5.693767632726341e-07, + "loss": 0.2148, + "step": 77142 + }, + { + "epoch": 2.97, + "learning_rate": 5.687355732239036e-07, + "loss": 0.1358, + "step": 77143 + }, + { + "epoch": 2.97, + "learning_rate": 5.680943831751732e-07, + "loss": 0.1567, + "step": 77144 + }, + { + "epoch": 2.97, + "learning_rate": 5.674531931264426e-07, + "loss": 0.2256, + "step": 77145 + }, + { + "epoch": 2.97, + "learning_rate": 5.668120030777122e-07, + "loss": 0.2492, + "step": 77146 + }, + { + "epoch": 2.97, + "learning_rate": 5.661708130289818e-07, + "loss": 0.1968, + "step": 77147 + }, + { + "epoch": 2.97, + "learning_rate": 5.655296229802513e-07, + "loss": 0.2059, + "step": 77148 + }, + { + "epoch": 2.97, + "learning_rate": 5.648884329315209e-07, + "loss": 0.231, + "step": 77149 + }, + { + "epoch": 2.97, + "learning_rate": 5.642472428827904e-07, + "loss": 0.3194, + "step": 77150 + }, + { + "epoch": 2.97, + "learning_rate": 5.6360605283406e-07, + "loss": 0.1464, + "step": 77151 + }, + { + "epoch": 2.97, + "learning_rate": 5.629648627853296e-07, + "loss": 0.1882, + "step": 77152 + }, + { + "epoch": 2.97, + "learning_rate": 5.623236727365991e-07, + "loss": 0.1527, + "step": 77153 + }, + { + "epoch": 2.97, + "learning_rate": 5.616824826878687e-07, + "loss": 0.1433, + "step": 77154 + }, + { + "epoch": 2.97, + "learning_rate": 5.610412926391382e-07, + "loss": 0.1815, + "step": 77155 + }, + { + "epoch": 2.97, + "learning_rate": 5.604001025904078e-07, + "loss": 0.1894, + "step": 77156 + }, + { + "epoch": 2.97, + "learning_rate": 5.597589125416774e-07, + "loss": 0.2163, + "step": 77157 + }, + { + "epoch": 2.97, + "learning_rate": 5.591177224929469e-07, + "loss": 0.1836, + "step": 77158 + }, + { + "epoch": 2.97, + "learning_rate": 5.584765324442165e-07, + "loss": 0.1317, + "step": 77159 + }, + { + "epoch": 2.97, + "learning_rate": 5.57835342395486e-07, + "loss": 0.1621, + "step": 77160 + }, + { + "epoch": 2.97, + "learning_rate": 5.571941523467556e-07, + "loss": 0.3294, + "step": 77161 + }, + { + "epoch": 2.97, + "learning_rate": 5.565529622980252e-07, + "loss": 0.1359, + "step": 77162 + }, + { + "epoch": 2.97, + "learning_rate": 5.559117722492947e-07, + "loss": 0.2042, + "step": 77163 + }, + { + "epoch": 2.97, + "learning_rate": 5.552705822005643e-07, + "loss": 0.4147, + "step": 77164 + }, + { + "epoch": 2.97, + "learning_rate": 5.546293921518339e-07, + "loss": 0.2734, + "step": 77165 + }, + { + "epoch": 2.97, + "learning_rate": 5.539882021031034e-07, + "loss": 0.3586, + "step": 77166 + }, + { + "epoch": 2.97, + "learning_rate": 5.53347012054373e-07, + "loss": 0.1568, + "step": 77167 + }, + { + "epoch": 2.97, + "learning_rate": 5.527058220056425e-07, + "loss": 0.0976, + "step": 77168 + }, + { + "epoch": 2.97, + "learning_rate": 5.520646319569121e-07, + "loss": 0.1608, + "step": 77169 + }, + { + "epoch": 2.97, + "learning_rate": 5.514234419081817e-07, + "loss": 0.1817, + "step": 77170 + }, + { + "epoch": 2.97, + "learning_rate": 5.507822518594512e-07, + "loss": 0.2523, + "step": 77171 + }, + { + "epoch": 2.97, + "learning_rate": 5.501410618107208e-07, + "loss": 0.1593, + "step": 77172 + }, + { + "epoch": 2.97, + "learning_rate": 5.494998717619903e-07, + "loss": 0.2226, + "step": 77173 + }, + { + "epoch": 2.97, + "learning_rate": 5.488586817132599e-07, + "loss": 0.1608, + "step": 77174 + }, + { + "epoch": 2.97, + "learning_rate": 5.482174916645295e-07, + "loss": 0.0779, + "step": 77175 + }, + { + "epoch": 2.97, + "learning_rate": 5.47576301615799e-07, + "loss": 0.1154, + "step": 77176 + }, + { + "epoch": 2.97, + "learning_rate": 5.469351115670685e-07, + "loss": 0.2042, + "step": 77177 + }, + { + "epoch": 2.97, + "learning_rate": 5.46293921518338e-07, + "loss": 0.1189, + "step": 77178 + }, + { + "epoch": 2.97, + "learning_rate": 5.456527314696076e-07, + "loss": 0.2501, + "step": 77179 + }, + { + "epoch": 2.97, + "learning_rate": 5.450115414208772e-07, + "loss": 0.1729, + "step": 77180 + }, + { + "epoch": 2.97, + "learning_rate": 5.443703513721467e-07, + "loss": 0.1553, + "step": 77181 + }, + { + "epoch": 2.97, + "learning_rate": 5.437291613234163e-07, + "loss": 0.133, + "step": 77182 + }, + { + "epoch": 2.97, + "learning_rate": 5.430879712746858e-07, + "loss": 0.1025, + "step": 77183 + }, + { + "epoch": 2.97, + "learning_rate": 5.424467812259554e-07, + "loss": 0.3989, + "step": 77184 + }, + { + "epoch": 2.97, + "learning_rate": 5.41805591177225e-07, + "loss": 0.2693, + "step": 77185 + }, + { + "epoch": 2.97, + "learning_rate": 5.411644011284945e-07, + "loss": 0.1745, + "step": 77186 + }, + { + "epoch": 2.97, + "learning_rate": 5.405232110797641e-07, + "loss": 0.169, + "step": 77187 + }, + { + "epoch": 2.97, + "learning_rate": 5.398820210310337e-07, + "loss": 0.209, + "step": 77188 + }, + { + "epoch": 2.97, + "learning_rate": 5.392408309823032e-07, + "loss": 0.1551, + "step": 77189 + }, + { + "epoch": 2.97, + "learning_rate": 5.385996409335728e-07, + "loss": 0.1784, + "step": 77190 + }, + { + "epoch": 2.97, + "learning_rate": 5.379584508848423e-07, + "loss": 0.2446, + "step": 77191 + }, + { + "epoch": 2.97, + "learning_rate": 5.373172608361119e-07, + "loss": 0.1385, + "step": 77192 + }, + { + "epoch": 2.97, + "learning_rate": 5.366760707873815e-07, + "loss": 0.2134, + "step": 77193 + }, + { + "epoch": 2.97, + "learning_rate": 5.36034880738651e-07, + "loss": 0.1808, + "step": 77194 + }, + { + "epoch": 2.97, + "learning_rate": 5.353936906899206e-07, + "loss": 0.1747, + "step": 77195 + }, + { + "epoch": 2.97, + "learning_rate": 5.347525006411901e-07, + "loss": 0.1854, + "step": 77196 + }, + { + "epoch": 2.97, + "learning_rate": 5.341113105924597e-07, + "loss": 0.1772, + "step": 77197 + }, + { + "epoch": 2.97, + "learning_rate": 5.334701205437293e-07, + "loss": 0.1523, + "step": 77198 + }, + { + "epoch": 2.97, + "learning_rate": 5.328289304949988e-07, + "loss": 0.2404, + "step": 77199 + }, + { + "epoch": 2.97, + "learning_rate": 5.321877404462684e-07, + "loss": 0.2276, + "step": 77200 + }, + { + "epoch": 2.97, + "learning_rate": 5.315465503975379e-07, + "loss": 0.2421, + "step": 77201 + }, + { + "epoch": 2.97, + "learning_rate": 5.309053603488075e-07, + "loss": 0.3118, + "step": 77202 + }, + { + "epoch": 2.97, + "learning_rate": 5.30264170300077e-07, + "loss": 0.156, + "step": 77203 + }, + { + "epoch": 2.97, + "learning_rate": 5.296229802513466e-07, + "loss": 0.2126, + "step": 77204 + }, + { + "epoch": 2.97, + "learning_rate": 5.289817902026162e-07, + "loss": 0.1977, + "step": 77205 + }, + { + "epoch": 2.97, + "learning_rate": 5.283406001538856e-07, + "loss": 0.2702, + "step": 77206 + }, + { + "epoch": 2.97, + "learning_rate": 5.276994101051552e-07, + "loss": 0.2823, + "step": 77207 + }, + { + "epoch": 2.97, + "learning_rate": 5.270582200564247e-07, + "loss": 0.1936, + "step": 77208 + }, + { + "epoch": 2.97, + "learning_rate": 5.264170300076943e-07, + "loss": 0.2048, + "step": 77209 + }, + { + "epoch": 2.97, + "learning_rate": 5.257758399589639e-07, + "loss": 0.3619, + "step": 77210 + }, + { + "epoch": 2.97, + "learning_rate": 5.251346499102334e-07, + "loss": 0.2178, + "step": 77211 + }, + { + "epoch": 2.97, + "learning_rate": 5.24493459861503e-07, + "loss": 0.155, + "step": 77212 + }, + { + "epoch": 2.97, + "learning_rate": 5.238522698127725e-07, + "loss": 0.0896, + "step": 77213 + }, + { + "epoch": 2.97, + "learning_rate": 5.232110797640421e-07, + "loss": 0.1487, + "step": 77214 + }, + { + "epoch": 2.97, + "learning_rate": 5.225698897153117e-07, + "loss": 0.1905, + "step": 77215 + }, + { + "epoch": 2.97, + "learning_rate": 5.219286996665812e-07, + "loss": 0.2717, + "step": 77216 + }, + { + "epoch": 2.97, + "learning_rate": 5.212875096178508e-07, + "loss": 0.2166, + "step": 77217 + }, + { + "epoch": 2.97, + "learning_rate": 5.206463195691203e-07, + "loss": 0.2211, + "step": 77218 + }, + { + "epoch": 2.97, + "learning_rate": 5.200051295203899e-07, + "loss": 0.2122, + "step": 77219 + }, + { + "epoch": 2.97, + "learning_rate": 5.193639394716595e-07, + "loss": 0.2017, + "step": 77220 + }, + { + "epoch": 2.97, + "learning_rate": 5.18722749422929e-07, + "loss": 0.1981, + "step": 77221 + }, + { + "epoch": 2.97, + "learning_rate": 5.180815593741986e-07, + "loss": 0.1426, + "step": 77222 + }, + { + "epoch": 2.97, + "learning_rate": 5.174403693254681e-07, + "loss": 0.1917, + "step": 77223 + }, + { + "epoch": 2.97, + "learning_rate": 5.167991792767377e-07, + "loss": 0.1632, + "step": 77224 + }, + { + "epoch": 2.97, + "learning_rate": 5.161579892280073e-07, + "loss": 0.1063, + "step": 77225 + }, + { + "epoch": 2.97, + "learning_rate": 5.155167991792768e-07, + "loss": 0.1536, + "step": 77226 + }, + { + "epoch": 2.97, + "learning_rate": 5.148756091305464e-07, + "loss": 0.2203, + "step": 77227 + }, + { + "epoch": 2.97, + "learning_rate": 5.142344190818159e-07, + "loss": 0.1879, + "step": 77228 + }, + { + "epoch": 2.97, + "learning_rate": 5.135932290330855e-07, + "loss": 0.1064, + "step": 77229 + }, + { + "epoch": 2.97, + "learning_rate": 5.12952038984355e-07, + "loss": 0.1932, + "step": 77230 + }, + { + "epoch": 2.97, + "learning_rate": 5.123108489356246e-07, + "loss": 0.2136, + "step": 77231 + }, + { + "epoch": 2.97, + "learning_rate": 5.116696588868942e-07, + "loss": 0.1677, + "step": 77232 + }, + { + "epoch": 2.97, + "learning_rate": 5.110284688381637e-07, + "loss": 0.1305, + "step": 77233 + }, + { + "epoch": 2.97, + "learning_rate": 5.103872787894333e-07, + "loss": 0.3818, + "step": 77234 + }, + { + "epoch": 2.97, + "learning_rate": 5.097460887407027e-07, + "loss": 0.1307, + "step": 77235 + }, + { + "epoch": 2.97, + "learning_rate": 5.091048986919723e-07, + "loss": 0.1782, + "step": 77236 + }, + { + "epoch": 2.97, + "learning_rate": 5.08463708643242e-07, + "loss": 0.1264, + "step": 77237 + }, + { + "epoch": 2.97, + "learning_rate": 5.078225185945114e-07, + "loss": 0.2168, + "step": 77238 + }, + { + "epoch": 2.97, + "learning_rate": 5.07181328545781e-07, + "loss": 0.1184, + "step": 77239 + }, + { + "epoch": 2.97, + "learning_rate": 5.065401384970505e-07, + "loss": 0.266, + "step": 77240 + }, + { + "epoch": 2.97, + "learning_rate": 5.058989484483201e-07, + "loss": 0.1044, + "step": 77241 + }, + { + "epoch": 2.97, + "learning_rate": 5.052577583995897e-07, + "loss": 0.1471, + "step": 77242 + }, + { + "epoch": 2.97, + "learning_rate": 5.046165683508592e-07, + "loss": 0.2509, + "step": 77243 + }, + { + "epoch": 2.97, + "learning_rate": 5.039753783021288e-07, + "loss": 0.1972, + "step": 77244 + }, + { + "epoch": 2.97, + "learning_rate": 5.033341882533983e-07, + "loss": 0.1738, + "step": 77245 + }, + { + "epoch": 2.97, + "learning_rate": 5.026929982046679e-07, + "loss": 0.1971, + "step": 77246 + }, + { + "epoch": 2.97, + "learning_rate": 5.020518081559375e-07, + "loss": 0.3188, + "step": 77247 + }, + { + "epoch": 2.97, + "learning_rate": 5.01410618107207e-07, + "loss": 0.2036, + "step": 77248 + }, + { + "epoch": 2.97, + "learning_rate": 5.007694280584766e-07, + "loss": 0.2067, + "step": 77249 + }, + { + "epoch": 2.97, + "learning_rate": 5.001282380097461e-07, + "loss": 0.1745, + "step": 77250 + }, + { + "epoch": 2.97, + "learning_rate": 4.994870479610157e-07, + "loss": 0.2053, + "step": 77251 + }, + { + "epoch": 2.97, + "learning_rate": 4.988458579122853e-07, + "loss": 0.2906, + "step": 77252 + }, + { + "epoch": 2.97, + "learning_rate": 4.982046678635548e-07, + "loss": 0.1328, + "step": 77253 + }, + { + "epoch": 2.97, + "learning_rate": 4.975634778148244e-07, + "loss": 0.2826, + "step": 77254 + }, + { + "epoch": 2.97, + "learning_rate": 4.969222877660939e-07, + "loss": 0.1525, + "step": 77255 + }, + { + "epoch": 2.97, + "learning_rate": 4.962810977173635e-07, + "loss": 0.2425, + "step": 77256 + }, + { + "epoch": 2.97, + "learning_rate": 4.956399076686331e-07, + "loss": 0.1817, + "step": 77257 + }, + { + "epoch": 2.97, + "learning_rate": 4.949987176199026e-07, + "loss": 0.1541, + "step": 77258 + }, + { + "epoch": 2.97, + "learning_rate": 4.943575275711722e-07, + "loss": 0.1762, + "step": 77259 + }, + { + "epoch": 2.97, + "learning_rate": 4.937163375224417e-07, + "loss": 0.2019, + "step": 77260 + }, + { + "epoch": 2.97, + "learning_rate": 4.930751474737113e-07, + "loss": 0.1221, + "step": 77261 + }, + { + "epoch": 2.97, + "learning_rate": 4.924339574249808e-07, + "loss": 0.3102, + "step": 77262 + }, + { + "epoch": 2.97, + "learning_rate": 4.917927673762504e-07, + "loss": 0.1554, + "step": 77263 + }, + { + "epoch": 2.97, + "learning_rate": 4.9115157732752e-07, + "loss": 0.1205, + "step": 77264 + }, + { + "epoch": 2.97, + "learning_rate": 4.905103872787895e-07, + "loss": 0.1634, + "step": 77265 + }, + { + "epoch": 2.97, + "learning_rate": 4.89869197230059e-07, + "loss": 0.2181, + "step": 77266 + }, + { + "epoch": 2.97, + "learning_rate": 4.892280071813285e-07, + "loss": 0.1611, + "step": 77267 + }, + { + "epoch": 2.97, + "learning_rate": 4.885868171325981e-07, + "loss": 0.0574, + "step": 77268 + }, + { + "epoch": 2.97, + "learning_rate": 4.879456270838677e-07, + "loss": 0.1165, + "step": 77269 + }, + { + "epoch": 2.97, + "learning_rate": 4.873044370351372e-07, + "loss": 0.251, + "step": 77270 + }, + { + "epoch": 2.97, + "learning_rate": 4.866632469864068e-07, + "loss": 0.1447, + "step": 77271 + }, + { + "epoch": 2.97, + "learning_rate": 4.860220569376763e-07, + "loss": 0.1873, + "step": 77272 + }, + { + "epoch": 2.97, + "learning_rate": 4.853808668889459e-07, + "loss": 0.2681, + "step": 77273 + }, + { + "epoch": 2.97, + "learning_rate": 4.847396768402155e-07, + "loss": 0.1699, + "step": 77274 + }, + { + "epoch": 2.97, + "learning_rate": 4.84098486791485e-07, + "loss": 0.1833, + "step": 77275 + }, + { + "epoch": 2.97, + "learning_rate": 4.834572967427546e-07, + "loss": 0.1355, + "step": 77276 + }, + { + "epoch": 2.97, + "learning_rate": 4.828161066940241e-07, + "loss": 0.2965, + "step": 77277 + }, + { + "epoch": 2.97, + "learning_rate": 4.821749166452937e-07, + "loss": 0.1969, + "step": 77278 + }, + { + "epoch": 2.97, + "learning_rate": 4.815337265965633e-07, + "loss": 0.221, + "step": 77279 + }, + { + "epoch": 2.97, + "learning_rate": 4.808925365478328e-07, + "loss": 0.1181, + "step": 77280 + }, + { + "epoch": 2.97, + "learning_rate": 4.802513464991024e-07, + "loss": 0.2327, + "step": 77281 + }, + { + "epoch": 2.97, + "learning_rate": 4.796101564503719e-07, + "loss": 0.3084, + "step": 77282 + }, + { + "epoch": 2.97, + "learning_rate": 4.789689664016415e-07, + "loss": 0.1399, + "step": 77283 + }, + { + "epoch": 2.97, + "learning_rate": 4.783277763529111e-07, + "loss": 0.1356, + "step": 77284 + }, + { + "epoch": 2.97, + "learning_rate": 4.776865863041806e-07, + "loss": 0.0971, + "step": 77285 + }, + { + "epoch": 2.97, + "learning_rate": 4.770453962554502e-07, + "loss": 0.1679, + "step": 77286 + }, + { + "epoch": 2.97, + "learning_rate": 4.764042062067197e-07, + "loss": 0.121, + "step": 77287 + }, + { + "epoch": 2.97, + "learning_rate": 4.757630161579893e-07, + "loss": 0.2306, + "step": 77288 + }, + { + "epoch": 2.97, + "learning_rate": 4.7512182610925877e-07, + "loss": 0.1366, + "step": 77289 + }, + { + "epoch": 2.97, + "learning_rate": 4.7448063606052837e-07, + "loss": 0.1723, + "step": 77290 + }, + { + "epoch": 2.97, + "learning_rate": 4.7383944601179797e-07, + "loss": 0.2121, + "step": 77291 + }, + { + "epoch": 2.97, + "learning_rate": 4.7319825596306746e-07, + "loss": 0.1901, + "step": 77292 + }, + { + "epoch": 2.97, + "learning_rate": 4.7255706591433706e-07, + "loss": 0.1808, + "step": 77293 + }, + { + "epoch": 2.97, + "learning_rate": 4.7191587586560655e-07, + "loss": 0.2555, + "step": 77294 + }, + { + "epoch": 2.97, + "learning_rate": 4.7127468581687615e-07, + "loss": 0.1796, + "step": 77295 + }, + { + "epoch": 2.97, + "learning_rate": 4.7063349576814575e-07, + "loss": 0.1443, + "step": 77296 + }, + { + "epoch": 2.97, + "learning_rate": 4.6999230571941524e-07, + "loss": 0.2112, + "step": 77297 + }, + { + "epoch": 2.97, + "learning_rate": 4.6935111567068484e-07, + "loss": 0.1305, + "step": 77298 + }, + { + "epoch": 2.97, + "learning_rate": 4.6870992562195434e-07, + "loss": 0.3051, + "step": 77299 + }, + { + "epoch": 2.97, + "learning_rate": 4.6806873557322393e-07, + "loss": 0.0983, + "step": 77300 + }, + { + "epoch": 2.97, + "learning_rate": 4.6742754552449353e-07, + "loss": 0.1939, + "step": 77301 + }, + { + "epoch": 2.97, + "learning_rate": 4.6678635547576303e-07, + "loss": 0.2272, + "step": 77302 + }, + { + "epoch": 2.97, + "learning_rate": 4.661451654270326e-07, + "loss": 0.2044, + "step": 77303 + }, + { + "epoch": 2.97, + "learning_rate": 4.655039753783021e-07, + "loss": 0.1896, + "step": 77304 + }, + { + "epoch": 2.97, + "learning_rate": 4.648627853295717e-07, + "loss": 0.1008, + "step": 77305 + }, + { + "epoch": 2.97, + "learning_rate": 4.642215952808413e-07, + "loss": 0.1378, + "step": 77306 + }, + { + "epoch": 2.97, + "learning_rate": 4.635804052321108e-07, + "loss": 0.1029, + "step": 77307 + }, + { + "epoch": 2.97, + "learning_rate": 4.629392151833804e-07, + "loss": 0.1091, + "step": 77308 + }, + { + "epoch": 2.97, + "learning_rate": 4.622980251346499e-07, + "loss": 0.198, + "step": 77309 + }, + { + "epoch": 2.97, + "learning_rate": 4.616568350859195e-07, + "loss": 0.1666, + "step": 77310 + }, + { + "epoch": 2.97, + "learning_rate": 4.610156450371891e-07, + "loss": 0.2228, + "step": 77311 + }, + { + "epoch": 2.97, + "learning_rate": 4.603744549884586e-07, + "loss": 0.2572, + "step": 77312 + }, + { + "epoch": 2.97, + "learning_rate": 4.597332649397282e-07, + "loss": 0.1123, + "step": 77313 + }, + { + "epoch": 2.97, + "learning_rate": 4.590920748909977e-07, + "loss": 0.1231, + "step": 77314 + }, + { + "epoch": 2.97, + "learning_rate": 4.584508848422673e-07, + "loss": 0.2367, + "step": 77315 + }, + { + "epoch": 2.97, + "learning_rate": 4.578096947935369e-07, + "loss": 0.1835, + "step": 77316 + }, + { + "epoch": 2.97, + "learning_rate": 4.571685047448064e-07, + "loss": 0.1522, + "step": 77317 + }, + { + "epoch": 2.97, + "learning_rate": 4.56527314696076e-07, + "loss": 0.1195, + "step": 77318 + }, + { + "epoch": 2.97, + "learning_rate": 4.5588612464734547e-07, + "loss": 0.3822, + "step": 77319 + }, + { + "epoch": 2.97, + "learning_rate": 4.5524493459861507e-07, + "loss": 0.1601, + "step": 77320 + }, + { + "epoch": 2.97, + "learning_rate": 4.5460374454988456e-07, + "loss": 0.2242, + "step": 77321 + }, + { + "epoch": 2.97, + "learning_rate": 4.5396255450115416e-07, + "loss": 0.2135, + "step": 77322 + }, + { + "epoch": 2.97, + "learning_rate": 4.5332136445242376e-07, + "loss": 0.2843, + "step": 77323 + }, + { + "epoch": 2.97, + "learning_rate": 4.5268017440369325e-07, + "loss": 0.2118, + "step": 77324 + }, + { + "epoch": 2.97, + "learning_rate": 4.5203898435496285e-07, + "loss": 0.114, + "step": 77325 + }, + { + "epoch": 2.97, + "learning_rate": 4.5139779430623234e-07, + "loss": 0.1981, + "step": 77326 + }, + { + "epoch": 2.97, + "learning_rate": 4.5075660425750194e-07, + "loss": 0.1308, + "step": 77327 + }, + { + "epoch": 2.97, + "learning_rate": 4.5011541420877154e-07, + "loss": 0.1745, + "step": 77328 + }, + { + "epoch": 2.97, + "learning_rate": 4.4947422416004104e-07, + "loss": 0.1281, + "step": 77329 + }, + { + "epoch": 2.97, + "learning_rate": 4.4883303411131064e-07, + "loss": 0.1427, + "step": 77330 + }, + { + "epoch": 2.97, + "learning_rate": 4.4819184406258013e-07, + "loss": 0.1745, + "step": 77331 + }, + { + "epoch": 2.97, + "learning_rate": 4.4755065401384973e-07, + "loss": 0.2061, + "step": 77332 + }, + { + "epoch": 2.97, + "learning_rate": 4.469094639651193e-07, + "loss": 0.3515, + "step": 77333 + }, + { + "epoch": 2.97, + "learning_rate": 4.462682739163888e-07, + "loss": 0.3089, + "step": 77334 + }, + { + "epoch": 2.97, + "learning_rate": 4.456270838676584e-07, + "loss": 0.2305, + "step": 77335 + }, + { + "epoch": 2.97, + "learning_rate": 4.449858938189279e-07, + "loss": 0.2202, + "step": 77336 + }, + { + "epoch": 2.97, + "learning_rate": 4.443447037701975e-07, + "loss": 0.1377, + "step": 77337 + }, + { + "epoch": 2.97, + "learning_rate": 4.437035137214671e-07, + "loss": 0.1374, + "step": 77338 + }, + { + "epoch": 2.97, + "learning_rate": 4.430623236727366e-07, + "loss": 0.2658, + "step": 77339 + }, + { + "epoch": 2.97, + "learning_rate": 4.424211336240062e-07, + "loss": 0.2671, + "step": 77340 + }, + { + "epoch": 2.97, + "learning_rate": 4.417799435752757e-07, + "loss": 0.1459, + "step": 77341 + }, + { + "epoch": 2.97, + "learning_rate": 4.411387535265453e-07, + "loss": 0.2619, + "step": 77342 + }, + { + "epoch": 2.97, + "learning_rate": 4.404975634778149e-07, + "loss": 0.2001, + "step": 77343 + }, + { + "epoch": 2.97, + "learning_rate": 4.398563734290844e-07, + "loss": 0.2275, + "step": 77344 + }, + { + "epoch": 2.97, + "learning_rate": 4.39215183380354e-07, + "loss": 0.2433, + "step": 77345 + }, + { + "epoch": 2.97, + "learning_rate": 4.385739933316235e-07, + "loss": 0.181, + "step": 77346 + }, + { + "epoch": 2.97, + "learning_rate": 4.379328032828931e-07, + "loss": 0.236, + "step": 77347 + }, + { + "epoch": 2.97, + "learning_rate": 4.3729161323416257e-07, + "loss": 0.178, + "step": 77348 + }, + { + "epoch": 2.97, + "learning_rate": 4.3665042318543217e-07, + "loss": 0.234, + "step": 77349 + }, + { + "epoch": 2.97, + "learning_rate": 4.3600923313670177e-07, + "loss": 0.2964, + "step": 77350 + }, + { + "epoch": 2.97, + "learning_rate": 4.3536804308797126e-07, + "loss": 0.1505, + "step": 77351 + }, + { + "epoch": 2.97, + "learning_rate": 4.3472685303924086e-07, + "loss": 0.2208, + "step": 77352 + }, + { + "epoch": 2.97, + "learning_rate": 4.3408566299051035e-07, + "loss": 0.1789, + "step": 77353 + }, + { + "epoch": 2.97, + "learning_rate": 4.3344447294177995e-07, + "loss": 0.2786, + "step": 77354 + }, + { + "epoch": 2.97, + "learning_rate": 4.3280328289304955e-07, + "loss": 0.3471, + "step": 77355 + }, + { + "epoch": 2.97, + "learning_rate": 4.3216209284431905e-07, + "loss": 0.1264, + "step": 77356 + }, + { + "epoch": 2.97, + "learning_rate": 4.3152090279558864e-07, + "loss": 0.2596, + "step": 77357 + }, + { + "epoch": 2.97, + "learning_rate": 4.3087971274685814e-07, + "loss": 0.2328, + "step": 77358 + }, + { + "epoch": 2.97, + "learning_rate": 4.3023852269812774e-07, + "loss": 0.1908, + "step": 77359 + }, + { + "epoch": 2.97, + "learning_rate": 4.2959733264939734e-07, + "loss": 0.2839, + "step": 77360 + }, + { + "epoch": 2.97, + "learning_rate": 4.2895614260066683e-07, + "loss": 0.1094, + "step": 77361 + }, + { + "epoch": 2.97, + "learning_rate": 4.2831495255193643e-07, + "loss": 0.2663, + "step": 77362 + }, + { + "epoch": 2.97, + "learning_rate": 4.276737625032059e-07, + "loss": 0.1491, + "step": 77363 + }, + { + "epoch": 2.97, + "learning_rate": 4.270325724544755e-07, + "loss": 0.1407, + "step": 77364 + }, + { + "epoch": 2.97, + "learning_rate": 4.263913824057451e-07, + "loss": 0.1394, + "step": 77365 + }, + { + "epoch": 2.97, + "learning_rate": 4.257501923570146e-07, + "loss": 0.324, + "step": 77366 + }, + { + "epoch": 2.97, + "learning_rate": 4.251090023082842e-07, + "loss": 0.2831, + "step": 77367 + }, + { + "epoch": 2.97, + "learning_rate": 4.244678122595537e-07, + "loss": 0.1775, + "step": 77368 + }, + { + "epoch": 2.97, + "learning_rate": 4.238266222108233e-07, + "loss": 0.183, + "step": 77369 + }, + { + "epoch": 2.97, + "learning_rate": 4.231854321620929e-07, + "loss": 0.2157, + "step": 77370 + }, + { + "epoch": 2.97, + "learning_rate": 4.225442421133624e-07, + "loss": 0.4417, + "step": 77371 + }, + { + "epoch": 2.97, + "learning_rate": 4.21903052064632e-07, + "loss": 0.163, + "step": 77372 + }, + { + "epoch": 2.97, + "learning_rate": 4.212618620159015e-07, + "loss": 0.3266, + "step": 77373 + }, + { + "epoch": 2.97, + "learning_rate": 4.206206719671711e-07, + "loss": 0.157, + "step": 77374 + }, + { + "epoch": 2.97, + "learning_rate": 4.199794819184407e-07, + "loss": 0.1718, + "step": 77375 + }, + { + "epoch": 2.97, + "learning_rate": 4.193382918697102e-07, + "loss": 0.1515, + "step": 77376 + }, + { + "epoch": 2.97, + "learning_rate": 4.186971018209798e-07, + "loss": 0.1608, + "step": 77377 + }, + { + "epoch": 2.97, + "learning_rate": 4.1805591177224927e-07, + "loss": 0.1125, + "step": 77378 + }, + { + "epoch": 2.97, + "learning_rate": 4.1741472172351887e-07, + "loss": 0.3579, + "step": 77379 + }, + { + "epoch": 2.98, + "learning_rate": 4.1677353167478836e-07, + "loss": 0.184, + "step": 77380 + }, + { + "epoch": 2.98, + "learning_rate": 4.1613234162605796e-07, + "loss": 0.1715, + "step": 77381 + }, + { + "epoch": 2.98, + "learning_rate": 4.1549115157732756e-07, + "loss": 0.3102, + "step": 77382 + }, + { + "epoch": 2.98, + "learning_rate": 4.1484996152859705e-07, + "loss": 0.122, + "step": 77383 + }, + { + "epoch": 2.98, + "learning_rate": 4.1420877147986665e-07, + "loss": 0.1787, + "step": 77384 + }, + { + "epoch": 2.98, + "learning_rate": 4.1356758143113615e-07, + "loss": 0.1484, + "step": 77385 + }, + { + "epoch": 2.98, + "learning_rate": 4.1292639138240575e-07, + "loss": 0.2323, + "step": 77386 + }, + { + "epoch": 2.98, + "learning_rate": 4.1228520133367535e-07, + "loss": 0.1483, + "step": 77387 + }, + { + "epoch": 2.98, + "learning_rate": 4.1164401128494484e-07, + "loss": 0.2003, + "step": 77388 + }, + { + "epoch": 2.98, + "learning_rate": 4.1100282123621444e-07, + "loss": 0.1464, + "step": 77389 + }, + { + "epoch": 2.98, + "learning_rate": 4.1036163118748393e-07, + "loss": 0.2375, + "step": 77390 + }, + { + "epoch": 2.98, + "learning_rate": 4.0972044113875353e-07, + "loss": 0.1682, + "step": 77391 + }, + { + "epoch": 2.98, + "learning_rate": 4.0907925109002313e-07, + "loss": 0.1748, + "step": 77392 + }, + { + "epoch": 2.98, + "learning_rate": 4.084380610412926e-07, + "loss": 0.1875, + "step": 77393 + }, + { + "epoch": 2.98, + "learning_rate": 4.077968709925622e-07, + "loss": 0.1518, + "step": 77394 + }, + { + "epoch": 2.98, + "learning_rate": 4.071556809438317e-07, + "loss": 0.1755, + "step": 77395 + }, + { + "epoch": 2.98, + "learning_rate": 4.065144908951013e-07, + "loss": 0.2262, + "step": 77396 + }, + { + "epoch": 2.98, + "learning_rate": 4.058733008463709e-07, + "loss": 0.2305, + "step": 77397 + }, + { + "epoch": 2.98, + "learning_rate": 4.052321107976404e-07, + "loss": 0.1923, + "step": 77398 + }, + { + "epoch": 2.98, + "learning_rate": 4.0459092074891e-07, + "loss": 0.2675, + "step": 77399 + }, + { + "epoch": 2.98, + "learning_rate": 4.0394973070017955e-07, + "loss": 0.1776, + "step": 77400 + }, + { + "epoch": 2.98, + "learning_rate": 4.033085406514491e-07, + "loss": 0.288, + "step": 77401 + }, + { + "epoch": 2.98, + "learning_rate": 4.026673506027187e-07, + "loss": 0.1333, + "step": 77402 + }, + { + "epoch": 2.98, + "learning_rate": 4.020261605539882e-07, + "loss": 0.1209, + "step": 77403 + }, + { + "epoch": 2.98, + "learning_rate": 4.013849705052578e-07, + "loss": 0.1659, + "step": 77404 + }, + { + "epoch": 2.98, + "learning_rate": 4.0074378045652733e-07, + "loss": 0.129, + "step": 77405 + }, + { + "epoch": 2.98, + "learning_rate": 4.001025904077969e-07, + "loss": 0.2415, + "step": 77406 + }, + { + "epoch": 2.98, + "learning_rate": 3.994614003590664e-07, + "loss": 0.1954, + "step": 77407 + }, + { + "epoch": 2.98, + "learning_rate": 3.9882021031033597e-07, + "loss": 0.108, + "step": 77408 + }, + { + "epoch": 2.98, + "learning_rate": 3.9817902026160557e-07, + "loss": 0.1309, + "step": 77409 + }, + { + "epoch": 2.98, + "learning_rate": 3.975378302128751e-07, + "loss": 0.2374, + "step": 77410 + }, + { + "epoch": 2.98, + "learning_rate": 3.9689664016414466e-07, + "loss": 0.2045, + "step": 77411 + }, + { + "epoch": 2.98, + "learning_rate": 3.962554501154142e-07, + "loss": 0.2467, + "step": 77412 + }, + { + "epoch": 2.98, + "learning_rate": 3.9561426006668376e-07, + "loss": 0.1685, + "step": 77413 + }, + { + "epoch": 2.98, + "learning_rate": 3.9497307001795335e-07, + "loss": 0.1083, + "step": 77414 + }, + { + "epoch": 2.98, + "learning_rate": 3.943318799692229e-07, + "loss": 0.2614, + "step": 77415 + }, + { + "epoch": 2.98, + "learning_rate": 3.9369068992049245e-07, + "loss": 0.2576, + "step": 77416 + }, + { + "epoch": 2.98, + "learning_rate": 3.93049499871762e-07, + "loss": 0.2179, + "step": 77417 + }, + { + "epoch": 2.98, + "learning_rate": 3.9240830982303154e-07, + "loss": 0.1826, + "step": 77418 + }, + { + "epoch": 2.98, + "learning_rate": 3.9176711977430114e-07, + "loss": 0.1017, + "step": 77419 + }, + { + "epoch": 2.98, + "learning_rate": 3.911259297255707e-07, + "loss": 0.299, + "step": 77420 + }, + { + "epoch": 2.98, + "learning_rate": 3.9048473967684023e-07, + "loss": 0.2248, + "step": 77421 + }, + { + "epoch": 2.98, + "learning_rate": 3.898435496281098e-07, + "loss": 0.1002, + "step": 77422 + }, + { + "epoch": 2.98, + "learning_rate": 3.892023595793794e-07, + "loss": 0.159, + "step": 77423 + }, + { + "epoch": 2.98, + "learning_rate": 3.885611695306489e-07, + "loss": 0.1791, + "step": 77424 + }, + { + "epoch": 2.98, + "learning_rate": 3.8791997948191847e-07, + "loss": 0.1358, + "step": 77425 + }, + { + "epoch": 2.98, + "learning_rate": 3.87278789433188e-07, + "loss": 0.2325, + "step": 77426 + }, + { + "epoch": 2.98, + "learning_rate": 3.8663759938445756e-07, + "loss": 0.2515, + "step": 77427 + }, + { + "epoch": 2.98, + "learning_rate": 3.8599640933572716e-07, + "loss": 0.1718, + "step": 77428 + }, + { + "epoch": 2.98, + "learning_rate": 3.853552192869967e-07, + "loss": 0.2605, + "step": 77429 + }, + { + "epoch": 2.98, + "learning_rate": 3.8471402923826625e-07, + "loss": 0.2392, + "step": 77430 + }, + { + "epoch": 2.98, + "learning_rate": 3.840728391895358e-07, + "loss": 0.192, + "step": 77431 + }, + { + "epoch": 2.98, + "learning_rate": 3.8343164914080534e-07, + "loss": 0.1921, + "step": 77432 + }, + { + "epoch": 2.98, + "learning_rate": 3.8279045909207494e-07, + "loss": 0.138, + "step": 77433 + }, + { + "epoch": 2.98, + "learning_rate": 3.821492690433445e-07, + "loss": 0.2104, + "step": 77434 + }, + { + "epoch": 2.98, + "learning_rate": 3.8150807899461403e-07, + "loss": 0.2677, + "step": 77435 + }, + { + "epoch": 2.98, + "learning_rate": 3.808668889458836e-07, + "loss": 0.1723, + "step": 77436 + }, + { + "epoch": 2.98, + "learning_rate": 3.802256988971532e-07, + "loss": 0.1883, + "step": 77437 + }, + { + "epoch": 2.98, + "learning_rate": 3.795845088484227e-07, + "loss": 0.1516, + "step": 77438 + }, + { + "epoch": 2.98, + "learning_rate": 3.7894331879969227e-07, + "loss": 0.1231, + "step": 77439 + }, + { + "epoch": 2.98, + "learning_rate": 3.783021287509618e-07, + "loss": 0.1918, + "step": 77440 + }, + { + "epoch": 2.98, + "learning_rate": 3.7766093870223136e-07, + "loss": 0.3578, + "step": 77441 + }, + { + "epoch": 2.98, + "learning_rate": 3.770197486535009e-07, + "loss": 0.2477, + "step": 77442 + }, + { + "epoch": 2.98, + "learning_rate": 3.763785586047705e-07, + "loss": 0.1208, + "step": 77443 + }, + { + "epoch": 2.98, + "learning_rate": 3.7573736855604006e-07, + "loss": 0.1899, + "step": 77444 + }, + { + "epoch": 2.98, + "learning_rate": 3.750961785073096e-07, + "loss": 0.2099, + "step": 77445 + }, + { + "epoch": 2.98, + "learning_rate": 3.7445498845857915e-07, + "loss": 0.2724, + "step": 77446 + }, + { + "epoch": 2.98, + "learning_rate": 3.738137984098487e-07, + "loss": 0.1159, + "step": 77447 + }, + { + "epoch": 2.98, + "learning_rate": 3.731726083611183e-07, + "loss": 0.1083, + "step": 77448 + }, + { + "epoch": 2.98, + "learning_rate": 3.7253141831238784e-07, + "loss": 0.2518, + "step": 77449 + }, + { + "epoch": 2.98, + "learning_rate": 3.718902282636574e-07, + "loss": 0.1833, + "step": 77450 + }, + { + "epoch": 2.98, + "learning_rate": 3.7124903821492693e-07, + "loss": 0.0993, + "step": 77451 + }, + { + "epoch": 2.98, + "learning_rate": 3.706078481661965e-07, + "loss": 0.2594, + "step": 77452 + }, + { + "epoch": 2.98, + "learning_rate": 3.699666581174661e-07, + "loss": 0.1758, + "step": 77453 + }, + { + "epoch": 2.98, + "learning_rate": 3.693254680687356e-07, + "loss": 0.2382, + "step": 77454 + }, + { + "epoch": 2.98, + "learning_rate": 3.6868427802000517e-07, + "loss": 0.1793, + "step": 77455 + }, + { + "epoch": 2.98, + "learning_rate": 3.680430879712747e-07, + "loss": 0.1457, + "step": 77456 + }, + { + "epoch": 2.98, + "learning_rate": 3.6740189792254426e-07, + "loss": 0.1463, + "step": 77457 + }, + { + "epoch": 2.98, + "learning_rate": 3.667607078738138e-07, + "loss": 0.1104, + "step": 77458 + }, + { + "epoch": 2.98, + "learning_rate": 3.661195178250834e-07, + "loss": 0.2758, + "step": 77459 + }, + { + "epoch": 2.98, + "learning_rate": 3.6547832777635295e-07, + "loss": 0.2745, + "step": 77460 + }, + { + "epoch": 2.98, + "learning_rate": 3.648371377276225e-07, + "loss": 0.1608, + "step": 77461 + }, + { + "epoch": 2.98, + "learning_rate": 3.6419594767889204e-07, + "loss": 0.4231, + "step": 77462 + }, + { + "epoch": 2.98, + "learning_rate": 3.635547576301616e-07, + "loss": 0.1874, + "step": 77463 + }, + { + "epoch": 2.98, + "learning_rate": 3.629135675814312e-07, + "loss": 0.0948, + "step": 77464 + }, + { + "epoch": 2.98, + "learning_rate": 3.6227237753270074e-07, + "loss": 0.2482, + "step": 77465 + }, + { + "epoch": 2.98, + "learning_rate": 3.616311874839703e-07, + "loss": 0.1113, + "step": 77466 + }, + { + "epoch": 2.98, + "learning_rate": 3.6098999743523983e-07, + "loss": 0.1917, + "step": 77467 + }, + { + "epoch": 2.98, + "learning_rate": 3.6034880738650937e-07, + "loss": 0.1669, + "step": 77468 + }, + { + "epoch": 2.98, + "learning_rate": 3.5970761733777897e-07, + "loss": 0.1608, + "step": 77469 + }, + { + "epoch": 2.98, + "learning_rate": 3.590664272890485e-07, + "loss": 0.2921, + "step": 77470 + }, + { + "epoch": 2.98, + "learning_rate": 3.5842523724031806e-07, + "loss": 0.141, + "step": 77471 + }, + { + "epoch": 2.98, + "learning_rate": 3.577840471915876e-07, + "loss": 0.196, + "step": 77472 + }, + { + "epoch": 2.98, + "learning_rate": 3.5714285714285716e-07, + "loss": 0.1273, + "step": 77473 + }, + { + "epoch": 2.98, + "learning_rate": 3.565016670941267e-07, + "loss": 0.2155, + "step": 77474 + }, + { + "epoch": 2.98, + "learning_rate": 3.558604770453963e-07, + "loss": 0.1552, + "step": 77475 + }, + { + "epoch": 2.98, + "learning_rate": 3.5521928699666585e-07, + "loss": 0.1093, + "step": 77476 + }, + { + "epoch": 2.98, + "learning_rate": 3.545780969479354e-07, + "loss": 0.1919, + "step": 77477 + }, + { + "epoch": 2.98, + "learning_rate": 3.5393690689920494e-07, + "loss": 0.1339, + "step": 77478 + }, + { + "epoch": 2.98, + "learning_rate": 3.532957168504745e-07, + "loss": 0.2998, + "step": 77479 + }, + { + "epoch": 2.98, + "learning_rate": 3.526545268017441e-07, + "loss": 0.1129, + "step": 77480 + }, + { + "epoch": 2.98, + "learning_rate": 3.5201333675301363e-07, + "loss": 0.1395, + "step": 77481 + }, + { + "epoch": 2.98, + "learning_rate": 3.513721467042832e-07, + "loss": 0.176, + "step": 77482 + }, + { + "epoch": 2.98, + "learning_rate": 3.507309566555527e-07, + "loss": 0.2307, + "step": 77483 + }, + { + "epoch": 2.98, + "learning_rate": 3.5008976660682227e-07, + "loss": 0.2772, + "step": 77484 + }, + { + "epoch": 2.98, + "learning_rate": 3.494485765580918e-07, + "loss": 0.1238, + "step": 77485 + }, + { + "epoch": 2.98, + "learning_rate": 3.488073865093614e-07, + "loss": 0.2562, + "step": 77486 + }, + { + "epoch": 2.98, + "learning_rate": 3.4816619646063096e-07, + "loss": 0.3148, + "step": 77487 + }, + { + "epoch": 2.98, + "learning_rate": 3.475250064119005e-07, + "loss": 0.2218, + "step": 77488 + }, + { + "epoch": 2.98, + "learning_rate": 3.4688381636317005e-07, + "loss": 0.2132, + "step": 77489 + }, + { + "epoch": 2.98, + "learning_rate": 3.462426263144396e-07, + "loss": 0.2191, + "step": 77490 + }, + { + "epoch": 2.98, + "learning_rate": 3.456014362657092e-07, + "loss": 0.2608, + "step": 77491 + }, + { + "epoch": 2.98, + "learning_rate": 3.4496024621697874e-07, + "loss": 0.1957, + "step": 77492 + }, + { + "epoch": 2.98, + "learning_rate": 3.443190561682483e-07, + "loss": 0.2528, + "step": 77493 + }, + { + "epoch": 2.98, + "learning_rate": 3.4367786611951784e-07, + "loss": 0.2306, + "step": 77494 + }, + { + "epoch": 2.98, + "learning_rate": 3.430366760707874e-07, + "loss": 0.18, + "step": 77495 + }, + { + "epoch": 2.98, + "learning_rate": 3.42395486022057e-07, + "loss": 0.207, + "step": 77496 + }, + { + "epoch": 2.98, + "learning_rate": 3.4175429597332653e-07, + "loss": 0.2978, + "step": 77497 + }, + { + "epoch": 2.98, + "learning_rate": 3.411131059245961e-07, + "loss": 0.3993, + "step": 77498 + }, + { + "epoch": 2.98, + "learning_rate": 3.404719158758656e-07, + "loss": 0.1178, + "step": 77499 + }, + { + "epoch": 2.98, + "learning_rate": 3.3983072582713517e-07, + "loss": 0.328, + "step": 77500 + }, + { + "epoch": 2.98, + "learning_rate": 3.391895357784047e-07, + "loss": 0.1822, + "step": 77501 + }, + { + "epoch": 2.98, + "learning_rate": 3.385483457296743e-07, + "loss": 0.1219, + "step": 77502 + }, + { + "epoch": 2.98, + "learning_rate": 3.3790715568094386e-07, + "loss": 0.2779, + "step": 77503 + }, + { + "epoch": 2.98, + "learning_rate": 3.372659656322134e-07, + "loss": 0.2016, + "step": 77504 + }, + { + "epoch": 2.98, + "learning_rate": 3.3662477558348295e-07, + "loss": 0.1304, + "step": 77505 + }, + { + "epoch": 2.98, + "learning_rate": 3.359835855347525e-07, + "loss": 0.2216, + "step": 77506 + }, + { + "epoch": 2.98, + "learning_rate": 3.353423954860221e-07, + "loss": 0.1994, + "step": 77507 + }, + { + "epoch": 2.98, + "learning_rate": 3.3470120543729164e-07, + "loss": 0.1813, + "step": 77508 + }, + { + "epoch": 2.98, + "learning_rate": 3.340600153885612e-07, + "loss": 0.153, + "step": 77509 + }, + { + "epoch": 2.98, + "learning_rate": 3.3341882533983073e-07, + "loss": 0.1984, + "step": 77510 + }, + { + "epoch": 2.98, + "learning_rate": 3.327776352911003e-07, + "loss": 0.1357, + "step": 77511 + }, + { + "epoch": 2.98, + "learning_rate": 3.321364452423699e-07, + "loss": 0.1694, + "step": 77512 + }, + { + "epoch": 2.98, + "learning_rate": 3.314952551936394e-07, + "loss": 0.2045, + "step": 77513 + }, + { + "epoch": 2.98, + "learning_rate": 3.3085406514490897e-07, + "loss": 0.2067, + "step": 77514 + }, + { + "epoch": 2.98, + "learning_rate": 3.302128750961785e-07, + "loss": 0.1706, + "step": 77515 + }, + { + "epoch": 2.98, + "learning_rate": 3.2957168504744806e-07, + "loss": 0.3564, + "step": 77516 + }, + { + "epoch": 2.98, + "learning_rate": 3.289304949987176e-07, + "loss": 0.1301, + "step": 77517 + }, + { + "epoch": 2.98, + "learning_rate": 3.282893049499872e-07, + "loss": 0.3223, + "step": 77518 + }, + { + "epoch": 2.98, + "learning_rate": 3.2764811490125675e-07, + "loss": 0.1419, + "step": 77519 + }, + { + "epoch": 2.98, + "learning_rate": 3.270069248525263e-07, + "loss": 0.2502, + "step": 77520 + }, + { + "epoch": 2.98, + "learning_rate": 3.2636573480379585e-07, + "loss": 0.1519, + "step": 77521 + }, + { + "epoch": 2.98, + "learning_rate": 3.257245447550654e-07, + "loss": 0.2541, + "step": 77522 + }, + { + "epoch": 2.98, + "learning_rate": 3.25083354706335e-07, + "loss": 0.3179, + "step": 77523 + }, + { + "epoch": 2.98, + "learning_rate": 3.2444216465760454e-07, + "loss": 0.3222, + "step": 77524 + }, + { + "epoch": 2.98, + "learning_rate": 3.238009746088741e-07, + "loss": 0.1664, + "step": 77525 + }, + { + "epoch": 2.98, + "learning_rate": 3.2315978456014363e-07, + "loss": 0.1312, + "step": 77526 + }, + { + "epoch": 2.98, + "learning_rate": 3.225185945114132e-07, + "loss": 0.2066, + "step": 77527 + }, + { + "epoch": 2.98, + "learning_rate": 3.218774044626828e-07, + "loss": 0.1771, + "step": 77528 + }, + { + "epoch": 2.98, + "learning_rate": 3.212362144139523e-07, + "loss": 0.2172, + "step": 77529 + }, + { + "epoch": 2.98, + "learning_rate": 3.2059502436522187e-07, + "loss": 0.2057, + "step": 77530 + }, + { + "epoch": 2.98, + "learning_rate": 3.199538343164914e-07, + "loss": 0.3284, + "step": 77531 + }, + { + "epoch": 2.98, + "learning_rate": 3.1931264426776096e-07, + "loss": 0.2092, + "step": 77532 + }, + { + "epoch": 2.98, + "learning_rate": 3.186714542190305e-07, + "loss": 0.231, + "step": 77533 + }, + { + "epoch": 2.98, + "learning_rate": 3.180302641703001e-07, + "loss": 0.2605, + "step": 77534 + }, + { + "epoch": 2.98, + "learning_rate": 3.1738907412156965e-07, + "loss": 0.2635, + "step": 77535 + }, + { + "epoch": 2.98, + "learning_rate": 3.167478840728392e-07, + "loss": 0.1387, + "step": 77536 + }, + { + "epoch": 2.98, + "learning_rate": 3.1610669402410874e-07, + "loss": 0.1667, + "step": 77537 + }, + { + "epoch": 2.98, + "learning_rate": 3.154655039753783e-07, + "loss": 0.258, + "step": 77538 + }, + { + "epoch": 2.98, + "learning_rate": 3.148243139266479e-07, + "loss": 0.2406, + "step": 77539 + }, + { + "epoch": 2.98, + "learning_rate": 3.1418312387791743e-07, + "loss": 0.3042, + "step": 77540 + }, + { + "epoch": 2.98, + "learning_rate": 3.13541933829187e-07, + "loss": 0.2995, + "step": 77541 + }, + { + "epoch": 2.98, + "learning_rate": 3.129007437804565e-07, + "loss": 0.2469, + "step": 77542 + }, + { + "epoch": 2.98, + "learning_rate": 3.1225955373172607e-07, + "loss": 0.3033, + "step": 77543 + }, + { + "epoch": 2.98, + "learning_rate": 3.116183636829956e-07, + "loss": 0.1807, + "step": 77544 + }, + { + "epoch": 2.98, + "learning_rate": 3.109771736342652e-07, + "loss": 0.2045, + "step": 77545 + }, + { + "epoch": 2.98, + "learning_rate": 3.1033598358553476e-07, + "loss": 0.1506, + "step": 77546 + }, + { + "epoch": 2.98, + "learning_rate": 3.096947935368043e-07, + "loss": 0.3112, + "step": 77547 + }, + { + "epoch": 2.98, + "learning_rate": 3.0905360348807386e-07, + "loss": 0.274, + "step": 77548 + }, + { + "epoch": 2.98, + "learning_rate": 3.084124134393434e-07, + "loss": 0.2764, + "step": 77549 + }, + { + "epoch": 2.98, + "learning_rate": 3.07771223390613e-07, + "loss": 0.3029, + "step": 77550 + }, + { + "epoch": 2.98, + "learning_rate": 3.0713003334188255e-07, + "loss": 0.2303, + "step": 77551 + }, + { + "epoch": 2.98, + "learning_rate": 3.064888432931521e-07, + "loss": 0.2408, + "step": 77552 + }, + { + "epoch": 2.98, + "learning_rate": 3.0584765324442164e-07, + "loss": 0.3733, + "step": 77553 + }, + { + "epoch": 2.98, + "learning_rate": 3.052064631956912e-07, + "loss": 0.1565, + "step": 77554 + }, + { + "epoch": 2.98, + "learning_rate": 3.045652731469608e-07, + "loss": 0.2216, + "step": 77555 + }, + { + "epoch": 2.98, + "learning_rate": 3.0392408309823033e-07, + "loss": 0.2564, + "step": 77556 + }, + { + "epoch": 2.98, + "learning_rate": 3.032828930494999e-07, + "loss": 0.1641, + "step": 77557 + }, + { + "epoch": 2.98, + "learning_rate": 3.026417030007694e-07, + "loss": 0.2388, + "step": 77558 + }, + { + "epoch": 2.98, + "learning_rate": 3.0200051295203897e-07, + "loss": 0.098, + "step": 77559 + }, + { + "epoch": 2.98, + "learning_rate": 3.013593229033085e-07, + "loss": 0.109, + "step": 77560 + }, + { + "epoch": 2.98, + "learning_rate": 3.007181328545781e-07, + "loss": 0.2822, + "step": 77561 + }, + { + "epoch": 2.98, + "learning_rate": 3.0007694280584766e-07, + "loss": 0.2574, + "step": 77562 + }, + { + "epoch": 2.98, + "learning_rate": 2.994357527571172e-07, + "loss": 0.1764, + "step": 77563 + }, + { + "epoch": 2.98, + "learning_rate": 2.9879456270838675e-07, + "loss": 0.1353, + "step": 77564 + }, + { + "epoch": 2.98, + "learning_rate": 2.981533726596563e-07, + "loss": 0.2789, + "step": 77565 + }, + { + "epoch": 2.98, + "learning_rate": 2.975121826109259e-07, + "loss": 0.3673, + "step": 77566 + }, + { + "epoch": 2.98, + "learning_rate": 2.9687099256219544e-07, + "loss": 0.1621, + "step": 77567 + }, + { + "epoch": 2.98, + "learning_rate": 2.96229802513465e-07, + "loss": 0.195, + "step": 77568 + }, + { + "epoch": 2.98, + "learning_rate": 2.9558861246473454e-07, + "loss": 0.1718, + "step": 77569 + }, + { + "epoch": 2.98, + "learning_rate": 2.949474224160041e-07, + "loss": 0.1896, + "step": 77570 + }, + { + "epoch": 2.98, + "learning_rate": 2.943062323672737e-07, + "loss": 0.189, + "step": 77571 + }, + { + "epoch": 2.98, + "learning_rate": 2.9366504231854323e-07, + "loss": 0.1891, + "step": 77572 + }, + { + "epoch": 2.98, + "learning_rate": 2.9302385226981277e-07, + "loss": 0.1582, + "step": 77573 + }, + { + "epoch": 2.98, + "learning_rate": 2.923826622210823e-07, + "loss": 0.2505, + "step": 77574 + }, + { + "epoch": 2.98, + "learning_rate": 2.9174147217235187e-07, + "loss": 0.2093, + "step": 77575 + }, + { + "epoch": 2.98, + "learning_rate": 2.911002821236214e-07, + "loss": 0.1593, + "step": 77576 + }, + { + "epoch": 2.98, + "learning_rate": 2.90459092074891e-07, + "loss": 0.1334, + "step": 77577 + }, + { + "epoch": 2.98, + "learning_rate": 2.8981790202616056e-07, + "loss": 0.1323, + "step": 77578 + }, + { + "epoch": 2.98, + "learning_rate": 2.891767119774301e-07, + "loss": 0.1202, + "step": 77579 + }, + { + "epoch": 2.98, + "learning_rate": 2.8853552192869965e-07, + "loss": 0.2449, + "step": 77580 + }, + { + "epoch": 2.98, + "learning_rate": 2.878943318799692e-07, + "loss": 0.2486, + "step": 77581 + }, + { + "epoch": 2.98, + "learning_rate": 2.872531418312388e-07, + "loss": 0.2168, + "step": 77582 + }, + { + "epoch": 2.98, + "learning_rate": 2.8661195178250834e-07, + "loss": 0.1596, + "step": 77583 + }, + { + "epoch": 2.98, + "learning_rate": 2.859707617337779e-07, + "loss": 0.1446, + "step": 77584 + }, + { + "epoch": 2.98, + "learning_rate": 2.8532957168504743e-07, + "loss": 0.2039, + "step": 77585 + }, + { + "epoch": 2.98, + "learning_rate": 2.8468838163631703e-07, + "loss": 0.1822, + "step": 77586 + }, + { + "epoch": 2.98, + "learning_rate": 2.840471915875866e-07, + "loss": 0.1972, + "step": 77587 + }, + { + "epoch": 2.98, + "learning_rate": 2.834060015388561e-07, + "loss": 0.1745, + "step": 77588 + }, + { + "epoch": 2.98, + "learning_rate": 2.8276481149012567e-07, + "loss": 0.1228, + "step": 77589 + }, + { + "epoch": 2.98, + "learning_rate": 2.821236214413952e-07, + "loss": 0.1322, + "step": 77590 + }, + { + "epoch": 2.98, + "learning_rate": 2.814824313926648e-07, + "loss": 0.2319, + "step": 77591 + }, + { + "epoch": 2.98, + "learning_rate": 2.8084124134393436e-07, + "loss": 0.1963, + "step": 77592 + }, + { + "epoch": 2.98, + "learning_rate": 2.802000512952039e-07, + "loss": 0.2153, + "step": 77593 + }, + { + "epoch": 2.98, + "learning_rate": 2.7955886124647345e-07, + "loss": 0.2249, + "step": 77594 + }, + { + "epoch": 2.98, + "learning_rate": 2.78917671197743e-07, + "loss": 0.198, + "step": 77595 + }, + { + "epoch": 2.98, + "learning_rate": 2.782764811490126e-07, + "loss": 0.3109, + "step": 77596 + }, + { + "epoch": 2.98, + "learning_rate": 2.7763529110028214e-07, + "loss": 0.1854, + "step": 77597 + }, + { + "epoch": 2.98, + "learning_rate": 2.769941010515517e-07, + "loss": 0.1672, + "step": 77598 + }, + { + "epoch": 2.98, + "learning_rate": 2.7635291100282124e-07, + "loss": 0.206, + "step": 77599 + }, + { + "epoch": 2.98, + "learning_rate": 2.7571172095409084e-07, + "loss": 0.2223, + "step": 77600 + }, + { + "epoch": 2.98, + "learning_rate": 2.750705309053604e-07, + "loss": 0.3026, + "step": 77601 + }, + { + "epoch": 2.98, + "learning_rate": 2.7442934085662993e-07, + "loss": 0.1767, + "step": 77602 + }, + { + "epoch": 2.98, + "learning_rate": 2.737881508078995e-07, + "loss": 0.144, + "step": 77603 + }, + { + "epoch": 2.98, + "learning_rate": 2.73146960759169e-07, + "loss": 0.2636, + "step": 77604 + }, + { + "epoch": 2.98, + "learning_rate": 2.725057707104386e-07, + "loss": 0.2057, + "step": 77605 + }, + { + "epoch": 2.98, + "learning_rate": 2.7186458066170816e-07, + "loss": 0.1436, + "step": 77606 + }, + { + "epoch": 2.98, + "learning_rate": 2.712233906129777e-07, + "loss": 0.2616, + "step": 77607 + }, + { + "epoch": 2.98, + "learning_rate": 2.7058220056424726e-07, + "loss": 0.2571, + "step": 77608 + }, + { + "epoch": 2.98, + "learning_rate": 2.6994101051551686e-07, + "loss": 0.1839, + "step": 77609 + }, + { + "epoch": 2.98, + "learning_rate": 2.692998204667864e-07, + "loss": 0.3516, + "step": 77610 + }, + { + "epoch": 2.98, + "learning_rate": 2.6865863041805595e-07, + "loss": 0.1591, + "step": 77611 + }, + { + "epoch": 2.98, + "learning_rate": 2.680174403693255e-07, + "loss": 0.1478, + "step": 77612 + }, + { + "epoch": 2.98, + "learning_rate": 2.6737625032059504e-07, + "loss": 0.2149, + "step": 77613 + }, + { + "epoch": 2.98, + "learning_rate": 2.6673506027186464e-07, + "loss": 0.2323, + "step": 77614 + }, + { + "epoch": 2.98, + "learning_rate": 2.660938702231342e-07, + "loss": 0.3722, + "step": 77615 + }, + { + "epoch": 2.98, + "learning_rate": 2.6545268017440373e-07, + "loss": 0.3326, + "step": 77616 + }, + { + "epoch": 2.98, + "learning_rate": 2.648114901256733e-07, + "loss": 0.3102, + "step": 77617 + }, + { + "epoch": 2.98, + "learning_rate": 2.641703000769428e-07, + "loss": 0.153, + "step": 77618 + }, + { + "epoch": 2.98, + "learning_rate": 2.6352911002821237e-07, + "loss": 0.2256, + "step": 77619 + }, + { + "epoch": 2.98, + "learning_rate": 2.6288791997948197e-07, + "loss": 0.3504, + "step": 77620 + }, + { + "epoch": 2.98, + "learning_rate": 2.622467299307515e-07, + "loss": 0.1384, + "step": 77621 + }, + { + "epoch": 2.98, + "learning_rate": 2.6160553988202106e-07, + "loss": 0.1926, + "step": 77622 + }, + { + "epoch": 2.98, + "learning_rate": 2.609643498332906e-07, + "loss": 0.1161, + "step": 77623 + }, + { + "epoch": 2.98, + "learning_rate": 2.6032315978456015e-07, + "loss": 0.1641, + "step": 77624 + }, + { + "epoch": 2.98, + "learning_rate": 2.5968196973582975e-07, + "loss": 0.2028, + "step": 77625 + }, + { + "epoch": 2.98, + "learning_rate": 2.590407796870993e-07, + "loss": 0.1888, + "step": 77626 + }, + { + "epoch": 2.98, + "learning_rate": 2.5839958963836884e-07, + "loss": 0.2124, + "step": 77627 + }, + { + "epoch": 2.98, + "learning_rate": 2.577583995896384e-07, + "loss": 0.2244, + "step": 77628 + }, + { + "epoch": 2.98, + "learning_rate": 2.5711720954090794e-07, + "loss": 0.229, + "step": 77629 + }, + { + "epoch": 2.98, + "learning_rate": 2.564760194921775e-07, + "loss": 0.2163, + "step": 77630 + }, + { + "epoch": 2.98, + "learning_rate": 2.558348294434471e-07, + "loss": 0.0917, + "step": 77631 + }, + { + "epoch": 2.98, + "learning_rate": 2.5519363939471663e-07, + "loss": 0.2369, + "step": 77632 + }, + { + "epoch": 2.98, + "learning_rate": 2.545524493459862e-07, + "loss": 0.1993, + "step": 77633 + }, + { + "epoch": 2.98, + "learning_rate": 2.539112592972557e-07, + "loss": 0.1867, + "step": 77634 + }, + { + "epoch": 2.98, + "learning_rate": 2.5327006924852527e-07, + "loss": 0.2961, + "step": 77635 + }, + { + "epoch": 2.98, + "learning_rate": 2.5262887919979487e-07, + "loss": 0.2917, + "step": 77636 + }, + { + "epoch": 2.98, + "learning_rate": 2.519876891510644e-07, + "loss": 0.1649, + "step": 77637 + }, + { + "epoch": 2.98, + "learning_rate": 2.5134649910233396e-07, + "loss": 0.312, + "step": 77638 + }, + { + "epoch": 2.98, + "learning_rate": 2.507053090536035e-07, + "loss": 0.269, + "step": 77639 + }, + { + "epoch": 2.99, + "learning_rate": 2.5006411900487305e-07, + "loss": 0.095, + "step": 77640 + }, + { + "epoch": 2.99, + "learning_rate": 2.4942292895614265e-07, + "loss": 0.2414, + "step": 77641 + }, + { + "epoch": 2.99, + "learning_rate": 2.487817389074122e-07, + "loss": 0.2362, + "step": 77642 + }, + { + "epoch": 2.99, + "learning_rate": 2.4814054885868174e-07, + "loss": 0.2469, + "step": 77643 + }, + { + "epoch": 2.99, + "learning_rate": 2.474993588099513e-07, + "loss": 0.2444, + "step": 77644 + }, + { + "epoch": 2.99, + "learning_rate": 2.4685816876122083e-07, + "loss": 0.215, + "step": 77645 + }, + { + "epoch": 2.99, + "learning_rate": 2.462169787124904e-07, + "loss": 0.1583, + "step": 77646 + }, + { + "epoch": 2.99, + "learning_rate": 2.4557578866376e-07, + "loss": 0.2683, + "step": 77647 + }, + { + "epoch": 2.99, + "learning_rate": 2.449345986150295e-07, + "loss": 0.2418, + "step": 77648 + }, + { + "epoch": 2.99, + "learning_rate": 2.4429340856629907e-07, + "loss": 0.2438, + "step": 77649 + }, + { + "epoch": 2.99, + "learning_rate": 2.436522185175686e-07, + "loss": 0.2563, + "step": 77650 + }, + { + "epoch": 2.99, + "learning_rate": 2.4301102846883816e-07, + "loss": 0.1803, + "step": 77651 + }, + { + "epoch": 2.99, + "learning_rate": 2.4236983842010776e-07, + "loss": 0.2461, + "step": 77652 + }, + { + "epoch": 2.99, + "learning_rate": 2.417286483713773e-07, + "loss": 0.1614, + "step": 77653 + }, + { + "epoch": 2.99, + "learning_rate": 2.4108745832264685e-07, + "loss": 0.1776, + "step": 77654 + }, + { + "epoch": 2.99, + "learning_rate": 2.404462682739164e-07, + "loss": 0.1683, + "step": 77655 + }, + { + "epoch": 2.99, + "learning_rate": 2.3980507822518595e-07, + "loss": 0.5295, + "step": 77656 + }, + { + "epoch": 2.99, + "learning_rate": 2.3916388817645555e-07, + "loss": 0.268, + "step": 77657 + }, + { + "epoch": 2.99, + "learning_rate": 2.385226981277251e-07, + "loss": 0.1605, + "step": 77658 + }, + { + "epoch": 2.99, + "learning_rate": 2.3788150807899464e-07, + "loss": 0.1749, + "step": 77659 + }, + { + "epoch": 2.99, + "learning_rate": 2.3724031803026418e-07, + "loss": 0.2761, + "step": 77660 + }, + { + "epoch": 2.99, + "learning_rate": 2.3659912798153373e-07, + "loss": 0.2351, + "step": 77661 + }, + { + "epoch": 2.99, + "learning_rate": 2.3595793793280328e-07, + "loss": 0.1659, + "step": 77662 + }, + { + "epoch": 2.99, + "learning_rate": 2.3531674788407287e-07, + "loss": 0.1951, + "step": 77663 + }, + { + "epoch": 2.99, + "learning_rate": 2.3467555783534242e-07, + "loss": 0.2854, + "step": 77664 + }, + { + "epoch": 2.99, + "learning_rate": 2.3403436778661197e-07, + "loss": 0.3004, + "step": 77665 + }, + { + "epoch": 2.99, + "learning_rate": 2.3339317773788151e-07, + "loss": 0.1773, + "step": 77666 + }, + { + "epoch": 2.99, + "learning_rate": 2.3275198768915106e-07, + "loss": 0.2096, + "step": 77667 + }, + { + "epoch": 2.99, + "learning_rate": 2.3211079764042066e-07, + "loss": 0.2212, + "step": 77668 + }, + { + "epoch": 2.99, + "learning_rate": 2.314696075916902e-07, + "loss": 0.207, + "step": 77669 + }, + { + "epoch": 2.99, + "learning_rate": 2.3082841754295975e-07, + "loss": 0.1636, + "step": 77670 + }, + { + "epoch": 2.99, + "learning_rate": 2.301872274942293e-07, + "loss": 0.3153, + "step": 77671 + }, + { + "epoch": 2.99, + "learning_rate": 2.2954603744549884e-07, + "loss": 0.3832, + "step": 77672 + }, + { + "epoch": 2.99, + "learning_rate": 2.2890484739676844e-07, + "loss": 0.1525, + "step": 77673 + }, + { + "epoch": 2.99, + "learning_rate": 2.28263657348038e-07, + "loss": 0.2654, + "step": 77674 + }, + { + "epoch": 2.99, + "learning_rate": 2.2762246729930753e-07, + "loss": 0.213, + "step": 77675 + }, + { + "epoch": 2.99, + "learning_rate": 2.2698127725057708e-07, + "loss": 0.2283, + "step": 77676 + }, + { + "epoch": 2.99, + "learning_rate": 2.2634008720184663e-07, + "loss": 0.1208, + "step": 77677 + }, + { + "epoch": 2.99, + "learning_rate": 2.2569889715311617e-07, + "loss": 0.3526, + "step": 77678 + }, + { + "epoch": 2.99, + "learning_rate": 2.2505770710438577e-07, + "loss": 0.2845, + "step": 77679 + }, + { + "epoch": 2.99, + "learning_rate": 2.2441651705565532e-07, + "loss": 0.2042, + "step": 77680 + }, + { + "epoch": 2.99, + "learning_rate": 2.2377532700692486e-07, + "loss": 0.1164, + "step": 77681 + }, + { + "epoch": 2.99, + "learning_rate": 2.231341369581944e-07, + "loss": 0.185, + "step": 77682 + }, + { + "epoch": 2.99, + "learning_rate": 2.2249294690946396e-07, + "loss": 0.1511, + "step": 77683 + }, + { + "epoch": 2.99, + "learning_rate": 2.2185175686073355e-07, + "loss": 0.1968, + "step": 77684 + }, + { + "epoch": 2.99, + "learning_rate": 2.212105668120031e-07, + "loss": 0.1725, + "step": 77685 + }, + { + "epoch": 2.99, + "learning_rate": 2.2056937676327265e-07, + "loss": 0.1059, + "step": 77686 + }, + { + "epoch": 2.99, + "learning_rate": 2.199281867145422e-07, + "loss": 0.1085, + "step": 77687 + }, + { + "epoch": 2.99, + "learning_rate": 2.1928699666581174e-07, + "loss": 0.2793, + "step": 77688 + }, + { + "epoch": 2.99, + "learning_rate": 2.1864580661708129e-07, + "loss": 0.1435, + "step": 77689 + }, + { + "epoch": 2.99, + "learning_rate": 2.1800461656835088e-07, + "loss": 0.3145, + "step": 77690 + }, + { + "epoch": 2.99, + "learning_rate": 2.1736342651962043e-07, + "loss": 0.2012, + "step": 77691 + }, + { + "epoch": 2.99, + "learning_rate": 2.1672223647088998e-07, + "loss": 0.1595, + "step": 77692 + }, + { + "epoch": 2.99, + "learning_rate": 2.1608104642215952e-07, + "loss": 0.0839, + "step": 77693 + }, + { + "epoch": 2.99, + "learning_rate": 2.1543985637342907e-07, + "loss": 0.2668, + "step": 77694 + }, + { + "epoch": 2.99, + "learning_rate": 2.1479866632469867e-07, + "loss": 0.2387, + "step": 77695 + }, + { + "epoch": 2.99, + "learning_rate": 2.1415747627596821e-07, + "loss": 0.1818, + "step": 77696 + }, + { + "epoch": 2.99, + "learning_rate": 2.1351628622723776e-07, + "loss": 0.1625, + "step": 77697 + }, + { + "epoch": 2.99, + "learning_rate": 2.128750961785073e-07, + "loss": 0.1573, + "step": 77698 + }, + { + "epoch": 2.99, + "learning_rate": 2.1223390612977685e-07, + "loss": 0.1425, + "step": 77699 + }, + { + "epoch": 2.99, + "learning_rate": 2.1159271608104645e-07, + "loss": 0.1422, + "step": 77700 + }, + { + "epoch": 2.99, + "learning_rate": 2.10951526032316e-07, + "loss": 0.2003, + "step": 77701 + }, + { + "epoch": 2.99, + "learning_rate": 2.1031033598358554e-07, + "loss": 0.1748, + "step": 77702 + }, + { + "epoch": 2.99, + "learning_rate": 2.096691459348551e-07, + "loss": 0.2464, + "step": 77703 + }, + { + "epoch": 2.99, + "learning_rate": 2.0902795588612464e-07, + "loss": 0.2201, + "step": 77704 + }, + { + "epoch": 2.99, + "learning_rate": 2.0838676583739418e-07, + "loss": 0.1178, + "step": 77705 + }, + { + "epoch": 2.99, + "learning_rate": 2.0774557578866378e-07, + "loss": 0.159, + "step": 77706 + }, + { + "epoch": 2.99, + "learning_rate": 2.0710438573993333e-07, + "loss": 0.161, + "step": 77707 + }, + { + "epoch": 2.99, + "learning_rate": 2.0646319569120287e-07, + "loss": 0.1438, + "step": 77708 + }, + { + "epoch": 2.99, + "learning_rate": 2.0582200564247242e-07, + "loss": 0.1965, + "step": 77709 + }, + { + "epoch": 2.99, + "learning_rate": 2.0518081559374197e-07, + "loss": 0.1602, + "step": 77710 + }, + { + "epoch": 2.99, + "learning_rate": 2.0453962554501156e-07, + "loss": 0.1811, + "step": 77711 + }, + { + "epoch": 2.99, + "learning_rate": 2.038984354962811e-07, + "loss": 0.2381, + "step": 77712 + }, + { + "epoch": 2.99, + "learning_rate": 2.0325724544755066e-07, + "loss": 0.2158, + "step": 77713 + }, + { + "epoch": 2.99, + "learning_rate": 2.026160553988202e-07, + "loss": 0.1959, + "step": 77714 + }, + { + "epoch": 2.99, + "learning_rate": 2.0197486535008978e-07, + "loss": 0.1806, + "step": 77715 + }, + { + "epoch": 2.99, + "learning_rate": 2.0133367530135935e-07, + "loss": 0.2504, + "step": 77716 + }, + { + "epoch": 2.99, + "learning_rate": 2.006924852526289e-07, + "loss": 0.1478, + "step": 77717 + }, + { + "epoch": 2.99, + "learning_rate": 2.0005129520389844e-07, + "loss": 0.2285, + "step": 77718 + }, + { + "epoch": 2.99, + "learning_rate": 1.9941010515516799e-07, + "loss": 0.3192, + "step": 77719 + }, + { + "epoch": 2.99, + "learning_rate": 1.9876891510643756e-07, + "loss": 0.2772, + "step": 77720 + }, + { + "epoch": 2.99, + "learning_rate": 1.981277250577071e-07, + "loss": 0.1938, + "step": 77721 + }, + { + "epoch": 2.99, + "learning_rate": 1.9748653500897668e-07, + "loss": 0.1593, + "step": 77722 + }, + { + "epoch": 2.99, + "learning_rate": 1.9684534496024622e-07, + "loss": 0.3027, + "step": 77723 + }, + { + "epoch": 2.99, + "learning_rate": 1.9620415491151577e-07, + "loss": 0.1482, + "step": 77724 + }, + { + "epoch": 2.99, + "learning_rate": 1.9556296486278534e-07, + "loss": 0.1242, + "step": 77725 + }, + { + "epoch": 2.99, + "learning_rate": 1.949217748140549e-07, + "loss": 0.1731, + "step": 77726 + }, + { + "epoch": 2.99, + "learning_rate": 1.9428058476532446e-07, + "loss": 0.1271, + "step": 77727 + }, + { + "epoch": 2.99, + "learning_rate": 1.93639394716594e-07, + "loss": 0.1233, + "step": 77728 + }, + { + "epoch": 2.99, + "learning_rate": 1.9299820466786358e-07, + "loss": 0.1085, + "step": 77729 + }, + { + "epoch": 2.99, + "learning_rate": 1.9235701461913313e-07, + "loss": 0.1757, + "step": 77730 + }, + { + "epoch": 2.99, + "learning_rate": 1.9171582457040267e-07, + "loss": 0.1469, + "step": 77731 + }, + { + "epoch": 2.99, + "learning_rate": 1.9107463452167224e-07, + "loss": 0.2312, + "step": 77732 + }, + { + "epoch": 2.99, + "learning_rate": 1.904334444729418e-07, + "loss": 0.2741, + "step": 77733 + }, + { + "epoch": 2.99, + "learning_rate": 1.8979225442421136e-07, + "loss": 0.1301, + "step": 77734 + }, + { + "epoch": 2.99, + "learning_rate": 1.891510643754809e-07, + "loss": 0.1175, + "step": 77735 + }, + { + "epoch": 2.99, + "learning_rate": 1.8850987432675046e-07, + "loss": 0.1788, + "step": 77736 + }, + { + "epoch": 2.99, + "learning_rate": 1.8786868427802003e-07, + "loss": 0.2236, + "step": 77737 + }, + { + "epoch": 2.99, + "learning_rate": 1.8722749422928957e-07, + "loss": 0.2268, + "step": 77738 + }, + { + "epoch": 2.99, + "learning_rate": 1.8658630418055915e-07, + "loss": 0.211, + "step": 77739 + }, + { + "epoch": 2.99, + "learning_rate": 1.859451141318287e-07, + "loss": 0.1835, + "step": 77740 + }, + { + "epoch": 2.99, + "learning_rate": 1.8530392408309824e-07, + "loss": 0.2875, + "step": 77741 + }, + { + "epoch": 2.99, + "learning_rate": 1.846627340343678e-07, + "loss": 0.1727, + "step": 77742 + }, + { + "epoch": 2.99, + "learning_rate": 1.8402154398563736e-07, + "loss": 0.1475, + "step": 77743 + }, + { + "epoch": 2.99, + "learning_rate": 1.833803539369069e-07, + "loss": 0.1718, + "step": 77744 + }, + { + "epoch": 2.99, + "learning_rate": 1.8273916388817648e-07, + "loss": 0.2455, + "step": 77745 + }, + { + "epoch": 2.99, + "learning_rate": 1.8209797383944602e-07, + "loss": 0.1267, + "step": 77746 + }, + { + "epoch": 2.99, + "learning_rate": 1.814567837907156e-07, + "loss": 0.0921, + "step": 77747 + }, + { + "epoch": 2.99, + "learning_rate": 1.8081559374198514e-07, + "loss": 0.1342, + "step": 77748 + }, + { + "epoch": 2.99, + "learning_rate": 1.8017440369325469e-07, + "loss": 0.2277, + "step": 77749 + }, + { + "epoch": 2.99, + "learning_rate": 1.7953321364452426e-07, + "loss": 0.2144, + "step": 77750 + }, + { + "epoch": 2.99, + "learning_rate": 1.788920235957938e-07, + "loss": 0.2017, + "step": 77751 + }, + { + "epoch": 2.99, + "learning_rate": 1.7825083354706335e-07, + "loss": 0.0975, + "step": 77752 + }, + { + "epoch": 2.99, + "learning_rate": 1.7760964349833292e-07, + "loss": 0.2825, + "step": 77753 + }, + { + "epoch": 2.99, + "learning_rate": 1.7696845344960247e-07, + "loss": 0.2225, + "step": 77754 + }, + { + "epoch": 2.99, + "learning_rate": 1.7632726340087204e-07, + "loss": 0.1962, + "step": 77755 + }, + { + "epoch": 2.99, + "learning_rate": 1.756860733521416e-07, + "loss": 0.1405, + "step": 77756 + }, + { + "epoch": 2.99, + "learning_rate": 1.7504488330341113e-07, + "loss": 0.2458, + "step": 77757 + }, + { + "epoch": 2.99, + "learning_rate": 1.744036932546807e-07, + "loss": 0.2044, + "step": 77758 + }, + { + "epoch": 2.99, + "learning_rate": 1.7376250320595025e-07, + "loss": 0.2175, + "step": 77759 + }, + { + "epoch": 2.99, + "learning_rate": 1.731213131572198e-07, + "loss": 0.1753, + "step": 77760 + }, + { + "epoch": 2.99, + "learning_rate": 1.7248012310848937e-07, + "loss": 0.2187, + "step": 77761 + }, + { + "epoch": 2.99, + "learning_rate": 1.7183893305975892e-07, + "loss": 0.2187, + "step": 77762 + }, + { + "epoch": 2.99, + "learning_rate": 1.711977430110285e-07, + "loss": 0.2044, + "step": 77763 + }, + { + "epoch": 2.99, + "learning_rate": 1.7055655296229804e-07, + "loss": 0.3689, + "step": 77764 + }, + { + "epoch": 2.99, + "learning_rate": 1.6991536291356758e-07, + "loss": 0.2072, + "step": 77765 + }, + { + "epoch": 2.99, + "learning_rate": 1.6927417286483716e-07, + "loss": 0.3263, + "step": 77766 + }, + { + "epoch": 2.99, + "learning_rate": 1.686329828161067e-07, + "loss": 0.2161, + "step": 77767 + }, + { + "epoch": 2.99, + "learning_rate": 1.6799179276737625e-07, + "loss": 0.2498, + "step": 77768 + }, + { + "epoch": 2.99, + "learning_rate": 1.6735060271864582e-07, + "loss": 0.2502, + "step": 77769 + }, + { + "epoch": 2.99, + "learning_rate": 1.6670941266991537e-07, + "loss": 0.1996, + "step": 77770 + }, + { + "epoch": 2.99, + "learning_rate": 1.6606822262118494e-07, + "loss": 0.2627, + "step": 77771 + }, + { + "epoch": 2.99, + "learning_rate": 1.6542703257245449e-07, + "loss": 0.1685, + "step": 77772 + }, + { + "epoch": 2.99, + "learning_rate": 1.6478584252372403e-07, + "loss": 0.2334, + "step": 77773 + }, + { + "epoch": 2.99, + "learning_rate": 1.641446524749936e-07, + "loss": 0.1219, + "step": 77774 + }, + { + "epoch": 2.99, + "learning_rate": 1.6350346242626315e-07, + "loss": 0.174, + "step": 77775 + }, + { + "epoch": 2.99, + "learning_rate": 1.628622723775327e-07, + "loss": 0.1844, + "step": 77776 + }, + { + "epoch": 2.99, + "learning_rate": 1.6222108232880227e-07, + "loss": 0.2071, + "step": 77777 + }, + { + "epoch": 2.99, + "learning_rate": 1.6157989228007181e-07, + "loss": 0.2336, + "step": 77778 + }, + { + "epoch": 2.99, + "learning_rate": 1.609387022313414e-07, + "loss": 0.1725, + "step": 77779 + }, + { + "epoch": 2.99, + "learning_rate": 1.6029751218261093e-07, + "loss": 0.1173, + "step": 77780 + }, + { + "epoch": 2.99, + "learning_rate": 1.5965632213388048e-07, + "loss": 0.0751, + "step": 77781 + }, + { + "epoch": 2.99, + "learning_rate": 1.5901513208515005e-07, + "loss": 0.1962, + "step": 77782 + }, + { + "epoch": 2.99, + "learning_rate": 1.583739420364196e-07, + "loss": 0.314, + "step": 77783 + }, + { + "epoch": 2.99, + "learning_rate": 1.5773275198768914e-07, + "loss": 0.2142, + "step": 77784 + }, + { + "epoch": 2.99, + "learning_rate": 1.5709156193895872e-07, + "loss": 0.3148, + "step": 77785 + }, + { + "epoch": 2.99, + "learning_rate": 1.5645037189022826e-07, + "loss": 0.2902, + "step": 77786 + }, + { + "epoch": 2.99, + "learning_rate": 1.558091818414978e-07, + "loss": 0.198, + "step": 77787 + }, + { + "epoch": 2.99, + "learning_rate": 1.5516799179276738e-07, + "loss": 0.1939, + "step": 77788 + }, + { + "epoch": 2.99, + "learning_rate": 1.5452680174403693e-07, + "loss": 0.1432, + "step": 77789 + }, + { + "epoch": 2.99, + "learning_rate": 1.538856116953065e-07, + "loss": 0.3791, + "step": 77790 + }, + { + "epoch": 2.99, + "learning_rate": 1.5324442164657605e-07, + "loss": 0.1736, + "step": 77791 + }, + { + "epoch": 2.99, + "learning_rate": 1.526032315978456e-07, + "loss": 0.34, + "step": 77792 + }, + { + "epoch": 2.99, + "learning_rate": 1.5196204154911517e-07, + "loss": 0.2092, + "step": 77793 + }, + { + "epoch": 2.99, + "learning_rate": 1.513208515003847e-07, + "loss": 0.2177, + "step": 77794 + }, + { + "epoch": 2.99, + "learning_rate": 1.5067966145165426e-07, + "loss": 0.1457, + "step": 77795 + }, + { + "epoch": 2.99, + "learning_rate": 1.5003847140292383e-07, + "loss": 0.0801, + "step": 77796 + }, + { + "epoch": 2.99, + "learning_rate": 1.4939728135419338e-07, + "loss": 0.186, + "step": 77797 + }, + { + "epoch": 2.99, + "learning_rate": 1.4875609130546295e-07, + "loss": 0.1194, + "step": 77798 + }, + { + "epoch": 2.99, + "learning_rate": 1.481149012567325e-07, + "loss": 0.1978, + "step": 77799 + }, + { + "epoch": 2.99, + "learning_rate": 1.4747371120800204e-07, + "loss": 0.2962, + "step": 77800 + }, + { + "epoch": 2.99, + "learning_rate": 1.4683252115927161e-07, + "loss": 0.1676, + "step": 77801 + }, + { + "epoch": 2.99, + "learning_rate": 1.4619133111054116e-07, + "loss": 0.168, + "step": 77802 + }, + { + "epoch": 2.99, + "learning_rate": 1.455501410618107e-07, + "loss": 0.2427, + "step": 77803 + }, + { + "epoch": 2.99, + "learning_rate": 1.4490895101308028e-07, + "loss": 0.2601, + "step": 77804 + }, + { + "epoch": 2.99, + "learning_rate": 1.4426776096434982e-07, + "loss": 0.167, + "step": 77805 + }, + { + "epoch": 2.99, + "learning_rate": 1.436265709156194e-07, + "loss": 0.1332, + "step": 77806 + }, + { + "epoch": 2.99, + "learning_rate": 1.4298538086688894e-07, + "loss": 0.1363, + "step": 77807 + }, + { + "epoch": 2.99, + "learning_rate": 1.4234419081815852e-07, + "loss": 0.2955, + "step": 77808 + }, + { + "epoch": 2.99, + "learning_rate": 1.4170300076942806e-07, + "loss": 0.1926, + "step": 77809 + }, + { + "epoch": 2.99, + "learning_rate": 1.410618107206976e-07, + "loss": 0.15, + "step": 77810 + }, + { + "epoch": 2.99, + "learning_rate": 1.4042062067196718e-07, + "loss": 0.1585, + "step": 77811 + }, + { + "epoch": 2.99, + "learning_rate": 1.3977943062323673e-07, + "loss": 0.1612, + "step": 77812 + }, + { + "epoch": 2.99, + "learning_rate": 1.391382405745063e-07, + "loss": 0.2253, + "step": 77813 + }, + { + "epoch": 2.99, + "learning_rate": 1.3849705052577585e-07, + "loss": 0.2354, + "step": 77814 + }, + { + "epoch": 2.99, + "learning_rate": 1.3785586047704542e-07, + "loss": 0.2358, + "step": 77815 + }, + { + "epoch": 2.99, + "learning_rate": 1.3721467042831496e-07, + "loss": 0.1785, + "step": 77816 + }, + { + "epoch": 2.99, + "learning_rate": 1.365734803795845e-07, + "loss": 0.1286, + "step": 77817 + }, + { + "epoch": 2.99, + "learning_rate": 1.3593229033085408e-07, + "loss": 0.1334, + "step": 77818 + }, + { + "epoch": 2.99, + "learning_rate": 1.3529110028212363e-07, + "loss": 0.1714, + "step": 77819 + }, + { + "epoch": 2.99, + "learning_rate": 1.346499102333932e-07, + "loss": 0.1861, + "step": 77820 + }, + { + "epoch": 2.99, + "learning_rate": 1.3400872018466275e-07, + "loss": 0.1532, + "step": 77821 + }, + { + "epoch": 2.99, + "learning_rate": 1.3336753013593232e-07, + "loss": 0.1624, + "step": 77822 + }, + { + "epoch": 2.99, + "learning_rate": 1.3272634008720187e-07, + "loss": 0.1568, + "step": 77823 + }, + { + "epoch": 2.99, + "learning_rate": 1.320851500384714e-07, + "loss": 0.1914, + "step": 77824 + }, + { + "epoch": 2.99, + "learning_rate": 1.3144395998974098e-07, + "loss": 0.1659, + "step": 77825 + }, + { + "epoch": 2.99, + "learning_rate": 1.3080276994101053e-07, + "loss": 0.394, + "step": 77826 + }, + { + "epoch": 2.99, + "learning_rate": 1.3016157989228008e-07, + "loss": 0.2497, + "step": 77827 + }, + { + "epoch": 2.99, + "learning_rate": 1.2952038984354965e-07, + "loss": 0.0988, + "step": 77828 + }, + { + "epoch": 2.99, + "learning_rate": 1.288791997948192e-07, + "loss": 0.1564, + "step": 77829 + }, + { + "epoch": 2.99, + "learning_rate": 1.2823800974608874e-07, + "loss": 0.2136, + "step": 77830 + }, + { + "epoch": 2.99, + "learning_rate": 1.2759681969735831e-07, + "loss": 0.1616, + "step": 77831 + }, + { + "epoch": 2.99, + "learning_rate": 1.2695562964862786e-07, + "loss": 0.2167, + "step": 77832 + }, + { + "epoch": 2.99, + "learning_rate": 1.2631443959989743e-07, + "loss": 0.188, + "step": 77833 + }, + { + "epoch": 2.99, + "learning_rate": 1.2567324955116698e-07, + "loss": 0.1808, + "step": 77834 + }, + { + "epoch": 2.99, + "learning_rate": 1.2503205950243652e-07, + "loss": 0.1621, + "step": 77835 + }, + { + "epoch": 2.99, + "learning_rate": 1.243908694537061e-07, + "loss": 0.1853, + "step": 77836 + }, + { + "epoch": 2.99, + "learning_rate": 1.2374967940497564e-07, + "loss": 0.1424, + "step": 77837 + }, + { + "epoch": 2.99, + "learning_rate": 1.231084893562452e-07, + "loss": 0.1229, + "step": 77838 + }, + { + "epoch": 2.99, + "learning_rate": 1.2246729930751476e-07, + "loss": 0.2519, + "step": 77839 + }, + { + "epoch": 2.99, + "learning_rate": 1.218261092587843e-07, + "loss": 0.1792, + "step": 77840 + }, + { + "epoch": 2.99, + "learning_rate": 1.2118491921005388e-07, + "loss": 0.2667, + "step": 77841 + }, + { + "epoch": 2.99, + "learning_rate": 1.2054372916132343e-07, + "loss": 0.2436, + "step": 77842 + }, + { + "epoch": 2.99, + "learning_rate": 1.1990253911259297e-07, + "loss": 0.2147, + "step": 77843 + }, + { + "epoch": 2.99, + "learning_rate": 1.1926134906386255e-07, + "loss": 0.2049, + "step": 77844 + }, + { + "epoch": 2.99, + "learning_rate": 1.1862015901513209e-07, + "loss": 0.1942, + "step": 77845 + }, + { + "epoch": 2.99, + "learning_rate": 1.1797896896640164e-07, + "loss": 0.0919, + "step": 77846 + }, + { + "epoch": 2.99, + "learning_rate": 1.1733777891767121e-07, + "loss": 0.2233, + "step": 77847 + }, + { + "epoch": 2.99, + "learning_rate": 1.1669658886894076e-07, + "loss": 0.1035, + "step": 77848 + }, + { + "epoch": 2.99, + "learning_rate": 1.1605539882021033e-07, + "loss": 0.1435, + "step": 77849 + }, + { + "epoch": 2.99, + "learning_rate": 1.1541420877147988e-07, + "loss": 0.1844, + "step": 77850 + }, + { + "epoch": 2.99, + "learning_rate": 1.1477301872274942e-07, + "loss": 0.2885, + "step": 77851 + }, + { + "epoch": 2.99, + "learning_rate": 1.14131828674019e-07, + "loss": 0.1705, + "step": 77852 + }, + { + "epoch": 2.99, + "learning_rate": 1.1349063862528854e-07, + "loss": 0.1521, + "step": 77853 + }, + { + "epoch": 2.99, + "learning_rate": 1.1284944857655809e-07, + "loss": 0.2131, + "step": 77854 + }, + { + "epoch": 2.99, + "learning_rate": 1.1220825852782766e-07, + "loss": 0.2567, + "step": 77855 + }, + { + "epoch": 2.99, + "learning_rate": 1.115670684790972e-07, + "loss": 0.1545, + "step": 77856 + }, + { + "epoch": 2.99, + "learning_rate": 1.1092587843036678e-07, + "loss": 0.1457, + "step": 77857 + }, + { + "epoch": 2.99, + "learning_rate": 1.1028468838163632e-07, + "loss": 0.2889, + "step": 77858 + }, + { + "epoch": 2.99, + "learning_rate": 1.0964349833290587e-07, + "loss": 0.2309, + "step": 77859 + }, + { + "epoch": 2.99, + "learning_rate": 1.0900230828417544e-07, + "loss": 0.1567, + "step": 77860 + }, + { + "epoch": 2.99, + "learning_rate": 1.0836111823544499e-07, + "loss": 0.1057, + "step": 77861 + }, + { + "epoch": 2.99, + "learning_rate": 1.0771992818671453e-07, + "loss": 0.3052, + "step": 77862 + }, + { + "epoch": 2.99, + "learning_rate": 1.0707873813798411e-07, + "loss": 0.3775, + "step": 77863 + }, + { + "epoch": 2.99, + "learning_rate": 1.0643754808925365e-07, + "loss": 0.1902, + "step": 77864 + }, + { + "epoch": 2.99, + "learning_rate": 1.0579635804052323e-07, + "loss": 0.2075, + "step": 77865 + }, + { + "epoch": 2.99, + "learning_rate": 1.0515516799179277e-07, + "loss": 0.2247, + "step": 77866 + }, + { + "epoch": 2.99, + "learning_rate": 1.0451397794306232e-07, + "loss": 0.1003, + "step": 77867 + }, + { + "epoch": 2.99, + "learning_rate": 1.0387278789433189e-07, + "loss": 0.1786, + "step": 77868 + }, + { + "epoch": 2.99, + "learning_rate": 1.0323159784560144e-07, + "loss": 0.1395, + "step": 77869 + }, + { + "epoch": 2.99, + "learning_rate": 1.0259040779687098e-07, + "loss": 0.1954, + "step": 77870 + }, + { + "epoch": 2.99, + "learning_rate": 1.0194921774814056e-07, + "loss": 0.1157, + "step": 77871 + }, + { + "epoch": 2.99, + "learning_rate": 1.013080276994101e-07, + "loss": 0.125, + "step": 77872 + }, + { + "epoch": 2.99, + "learning_rate": 1.0066683765067967e-07, + "loss": 0.2693, + "step": 77873 + }, + { + "epoch": 2.99, + "learning_rate": 1.0002564760194922e-07, + "loss": 0.0814, + "step": 77874 + }, + { + "epoch": 2.99, + "learning_rate": 9.938445755321878e-08, + "loss": 0.123, + "step": 77875 + }, + { + "epoch": 2.99, + "learning_rate": 9.874326750448834e-08, + "loss": 0.1745, + "step": 77876 + }, + { + "epoch": 2.99, + "learning_rate": 9.810207745575788e-08, + "loss": 0.1843, + "step": 77877 + }, + { + "epoch": 2.99, + "learning_rate": 9.746088740702744e-08, + "loss": 0.1639, + "step": 77878 + }, + { + "epoch": 2.99, + "learning_rate": 9.6819697358297e-08, + "loss": 0.3093, + "step": 77879 + }, + { + "epoch": 2.99, + "learning_rate": 9.617850730956656e-08, + "loss": 0.1674, + "step": 77880 + }, + { + "epoch": 2.99, + "learning_rate": 9.553731726083612e-08, + "loss": 0.2242, + "step": 77881 + }, + { + "epoch": 2.99, + "learning_rate": 9.489612721210568e-08, + "loss": 0.2201, + "step": 77882 + }, + { + "epoch": 2.99, + "learning_rate": 9.425493716337523e-08, + "loss": 0.286, + "step": 77883 + }, + { + "epoch": 2.99, + "learning_rate": 9.361374711464479e-08, + "loss": 0.1976, + "step": 77884 + }, + { + "epoch": 2.99, + "learning_rate": 9.297255706591435e-08, + "loss": 0.2079, + "step": 77885 + }, + { + "epoch": 2.99, + "learning_rate": 9.23313670171839e-08, + "loss": 0.3461, + "step": 77886 + }, + { + "epoch": 2.99, + "learning_rate": 9.169017696845345e-08, + "loss": 0.182, + "step": 77887 + }, + { + "epoch": 2.99, + "learning_rate": 9.104898691972301e-08, + "loss": 0.1996, + "step": 77888 + }, + { + "epoch": 2.99, + "learning_rate": 9.040779687099257e-08, + "loss": 0.1844, + "step": 77889 + }, + { + "epoch": 2.99, + "learning_rate": 8.976660682226213e-08, + "loss": 0.1819, + "step": 77890 + }, + { + "epoch": 2.99, + "learning_rate": 8.912541677353168e-08, + "loss": 0.3226, + "step": 77891 + }, + { + "epoch": 2.99, + "learning_rate": 8.848422672480124e-08, + "loss": 0.1887, + "step": 77892 + }, + { + "epoch": 2.99, + "learning_rate": 8.78430366760708e-08, + "loss": 0.1317, + "step": 77893 + }, + { + "epoch": 2.99, + "learning_rate": 8.720184662734035e-08, + "loss": 0.1574, + "step": 77894 + }, + { + "epoch": 2.99, + "learning_rate": 8.65606565786099e-08, + "loss": 0.2836, + "step": 77895 + }, + { + "epoch": 2.99, + "learning_rate": 8.591946652987946e-08, + "loss": 0.2804, + "step": 77896 + }, + { + "epoch": 2.99, + "learning_rate": 8.527827648114902e-08, + "loss": 0.2073, + "step": 77897 + }, + { + "epoch": 2.99, + "learning_rate": 8.463708643241858e-08, + "loss": 0.142, + "step": 77898 + }, + { + "epoch": 2.99, + "learning_rate": 8.399589638368812e-08, + "loss": 0.2728, + "step": 77899 + }, + { + "epoch": 3.0, + "learning_rate": 8.335470633495768e-08, + "loss": 0.2311, + "step": 77900 + }, + { + "epoch": 3.0, + "learning_rate": 8.271351628622724e-08, + "loss": 0.2094, + "step": 77901 + }, + { + "epoch": 3.0, + "learning_rate": 8.20723262374968e-08, + "loss": 0.194, + "step": 77902 + }, + { + "epoch": 3.0, + "learning_rate": 8.143113618876635e-08, + "loss": 0.2239, + "step": 77903 + }, + { + "epoch": 3.0, + "learning_rate": 8.078994614003591e-08, + "loss": 0.3665, + "step": 77904 + }, + { + "epoch": 3.0, + "learning_rate": 8.014875609130547e-08, + "loss": 0.2558, + "step": 77905 + }, + { + "epoch": 3.0, + "learning_rate": 7.950756604257503e-08, + "loss": 0.1136, + "step": 77906 + }, + { + "epoch": 3.0, + "learning_rate": 7.886637599384457e-08, + "loss": 0.2888, + "step": 77907 + }, + { + "epoch": 3.0, + "learning_rate": 7.822518594511413e-08, + "loss": 0.2528, + "step": 77908 + }, + { + "epoch": 3.0, + "learning_rate": 7.758399589638369e-08, + "loss": 0.2192, + "step": 77909 + }, + { + "epoch": 3.0, + "learning_rate": 7.694280584765325e-08, + "loss": 0.3782, + "step": 77910 + }, + { + "epoch": 3.0, + "learning_rate": 7.63016157989228e-08, + "loss": 0.3529, + "step": 77911 + }, + { + "epoch": 3.0, + "learning_rate": 7.566042575019236e-08, + "loss": 0.1474, + "step": 77912 + }, + { + "epoch": 3.0, + "learning_rate": 7.501923570146191e-08, + "loss": 0.1606, + "step": 77913 + }, + { + "epoch": 3.0, + "learning_rate": 7.437804565273147e-08, + "loss": 0.206, + "step": 77914 + }, + { + "epoch": 3.0, + "learning_rate": 7.373685560400102e-08, + "loss": 0.2046, + "step": 77915 + }, + { + "epoch": 3.0, + "learning_rate": 7.309566555527058e-08, + "loss": 0.19, + "step": 77916 + }, + { + "epoch": 3.0, + "learning_rate": 7.245447550654014e-08, + "loss": 0.2004, + "step": 77917 + }, + { + "epoch": 3.0, + "learning_rate": 7.18132854578097e-08, + "loss": 0.1255, + "step": 77918 + }, + { + "epoch": 3.0, + "learning_rate": 7.117209540907926e-08, + "loss": 0.0699, + "step": 77919 + }, + { + "epoch": 3.0, + "learning_rate": 7.05309053603488e-08, + "loss": 0.1897, + "step": 77920 + }, + { + "epoch": 3.0, + "learning_rate": 6.988971531161836e-08, + "loss": 0.1186, + "step": 77921 + }, + { + "epoch": 3.0, + "learning_rate": 6.924852526288792e-08, + "loss": 0.1188, + "step": 77922 + }, + { + "epoch": 3.0, + "learning_rate": 6.860733521415748e-08, + "loss": 0.2143, + "step": 77923 + }, + { + "epoch": 3.0, + "learning_rate": 6.796614516542704e-08, + "loss": 0.1911, + "step": 77924 + }, + { + "epoch": 3.0, + "learning_rate": 6.73249551166966e-08, + "loss": 0.2999, + "step": 77925 + }, + { + "epoch": 3.0, + "learning_rate": 6.668376506796616e-08, + "loss": 0.1016, + "step": 77926 + }, + { + "epoch": 3.0, + "learning_rate": 6.60425750192357e-08, + "loss": 0.1991, + "step": 77927 + }, + { + "epoch": 3.0, + "learning_rate": 6.540138497050527e-08, + "loss": 0.2073, + "step": 77928 + }, + { + "epoch": 3.0, + "learning_rate": 6.476019492177482e-08, + "loss": 0.2907, + "step": 77929 + }, + { + "epoch": 3.0, + "learning_rate": 6.411900487304437e-08, + "loss": 0.201, + "step": 77930 + }, + { + "epoch": 3.0, + "learning_rate": 6.347781482431393e-08, + "loss": 0.2657, + "step": 77931 + }, + { + "epoch": 3.0, + "learning_rate": 6.283662477558349e-08, + "loss": 0.3127, + "step": 77932 + }, + { + "epoch": 3.0, + "learning_rate": 6.219543472685305e-08, + "loss": 0.2481, + "step": 77933 + }, + { + "epoch": 3.0, + "learning_rate": 6.15542446781226e-08, + "loss": 0.2411, + "step": 77934 + }, + { + "epoch": 3.0, + "learning_rate": 6.091305462939215e-08, + "loss": 0.0843, + "step": 77935 + }, + { + "epoch": 3.0, + "learning_rate": 6.027186458066171e-08, + "loss": 0.144, + "step": 77936 + }, + { + "epoch": 3.0, + "learning_rate": 5.963067453193127e-08, + "loss": 0.2381, + "step": 77937 + }, + { + "epoch": 3.0, + "learning_rate": 5.898948448320082e-08, + "loss": 0.2372, + "step": 77938 + }, + { + "epoch": 3.0, + "learning_rate": 5.834829443447038e-08, + "loss": 0.1154, + "step": 77939 + }, + { + "epoch": 3.0, + "learning_rate": 5.770710438573994e-08, + "loss": 0.1915, + "step": 77940 + }, + { + "epoch": 3.0, + "learning_rate": 5.70659143370095e-08, + "loss": 0.1486, + "step": 77941 + }, + { + "epoch": 3.0, + "learning_rate": 5.642472428827904e-08, + "loss": 0.3379, + "step": 77942 + }, + { + "epoch": 3.0, + "learning_rate": 5.57835342395486e-08, + "loss": 0.2168, + "step": 77943 + }, + { + "epoch": 3.0, + "learning_rate": 5.514234419081816e-08, + "loss": 0.3016, + "step": 77944 + }, + { + "epoch": 3.0, + "learning_rate": 5.450115414208772e-08, + "loss": 0.1025, + "step": 77945 + }, + { + "epoch": 3.0, + "learning_rate": 5.385996409335727e-08, + "loss": 0.2911, + "step": 77946 + }, + { + "epoch": 3.0, + "learning_rate": 5.3218774044626827e-08, + "loss": 0.2446, + "step": 77947 + }, + { + "epoch": 3.0, + "learning_rate": 5.2577583995896386e-08, + "loss": 0.1272, + "step": 77948 + }, + { + "epoch": 3.0, + "learning_rate": 5.1936393947165945e-08, + "loss": 0.3595, + "step": 77949 + }, + { + "epoch": 3.0, + "learning_rate": 5.129520389843549e-08, + "loss": 0.199, + "step": 77950 + }, + { + "epoch": 3.0, + "learning_rate": 5.065401384970505e-08, + "loss": 0.1562, + "step": 77951 + }, + { + "epoch": 3.0, + "learning_rate": 5.001282380097461e-08, + "loss": 0.1426, + "step": 77952 + }, + { + "epoch": 3.0, + "learning_rate": 4.937163375224417e-08, + "loss": 0.1624, + "step": 77953 + }, + { + "epoch": 3.0, + "learning_rate": 4.873044370351372e-08, + "loss": 0.1951, + "step": 77954 + }, + { + "epoch": 3.0, + "learning_rate": 4.808925365478328e-08, + "loss": 0.2553, + "step": 77955 + }, + { + "epoch": 3.0, + "learning_rate": 4.744806360605284e-08, + "loss": 0.2231, + "step": 77956 + }, + { + "epoch": 3.0, + "learning_rate": 4.6806873557322393e-08, + "loss": 0.2689, + "step": 77957 + }, + { + "epoch": 3.0, + "learning_rate": 4.616568350859195e-08, + "loss": 0.2431, + "step": 77958 + }, + { + "epoch": 3.0, + "learning_rate": 4.5524493459861505e-08, + "loss": 0.214, + "step": 77959 + }, + { + "epoch": 3.0, + "learning_rate": 4.4883303411131065e-08, + "loss": 0.1553, + "step": 77960 + }, + { + "epoch": 3.0, + "learning_rate": 4.424211336240062e-08, + "loss": 0.2642, + "step": 77961 + }, + { + "epoch": 3.0, + "learning_rate": 4.360092331367018e-08, + "loss": 0.1617, + "step": 77962 + }, + { + "epoch": 3.0, + "learning_rate": 4.295973326493973e-08, + "loss": 0.2292, + "step": 77963 + }, + { + "epoch": 3.0, + "learning_rate": 4.231854321620929e-08, + "loss": 0.1463, + "step": 77964 + }, + { + "epoch": 3.0, + "learning_rate": 4.167735316747884e-08, + "loss": 0.3095, + "step": 77965 + }, + { + "epoch": 3.0, + "learning_rate": 4.10361631187484e-08, + "loss": 0.175, + "step": 77966 + }, + { + "epoch": 3.0, + "learning_rate": 4.0394973070017954e-08, + "loss": 0.38, + "step": 77967 + }, + { + "epoch": 3.0, + "learning_rate": 3.975378302128751e-08, + "loss": 0.1965, + "step": 77968 + }, + { + "epoch": 3.0, + "learning_rate": 3.9112592972557066e-08, + "loss": 0.263, + "step": 77969 + }, + { + "epoch": 3.0, + "learning_rate": 3.8471402923826625e-08, + "loss": 0.1967, + "step": 77970 + }, + { + "epoch": 3.0, + "learning_rate": 3.783021287509618e-08, + "loss": 0.2227, + "step": 77971 + }, + { + "epoch": 3.0, + "learning_rate": 3.718902282636574e-08, + "loss": 0.1748, + "step": 77972 + }, + { + "epoch": 3.0, + "learning_rate": 3.654783277763529e-08, + "loss": 0.1439, + "step": 77973 + }, + { + "epoch": 3.0, + "learning_rate": 3.590664272890485e-08, + "loss": 0.1089, + "step": 77974 + }, + { + "epoch": 3.0, + "learning_rate": 3.52654526801744e-08, + "loss": 0.1577, + "step": 77975 + }, + { + "epoch": 3.0, + "learning_rate": 3.462426263144396e-08, + "loss": 0.2041, + "step": 77976 + }, + { + "epoch": 3.0, + "learning_rate": 3.398307258271352e-08, + "loss": 0.2415, + "step": 77977 + }, + { + "epoch": 3.0, + "learning_rate": 3.334188253398308e-08, + "loss": 0.3004, + "step": 77978 + }, + { + "epoch": 3.0, + "learning_rate": 3.270069248525263e-08, + "loss": 0.2554, + "step": 77979 + }, + { + "epoch": 3.0, + "learning_rate": 3.2059502436522185e-08, + "loss": 0.2429, + "step": 77980 + }, + { + "epoch": 3.0, + "learning_rate": 3.1418312387791745e-08, + "loss": 0.2502, + "step": 77981 + }, + { + "epoch": 3.0, + "learning_rate": 3.07771223390613e-08, + "loss": 0.2359, + "step": 77982 + }, + { + "epoch": 3.0, + "learning_rate": 3.013593229033086e-08, + "loss": 0.1985, + "step": 77983 + }, + { + "epoch": 3.0, + "learning_rate": 2.949474224160041e-08, + "loss": 0.2206, + "step": 77984 + }, + { + "epoch": 3.0, + "learning_rate": 2.885355219286997e-08, + "loss": 0.2384, + "step": 77985 + }, + { + "epoch": 3.0, + "learning_rate": 2.821236214413952e-08, + "loss": 0.2732, + "step": 77986 + }, + { + "epoch": 3.0, + "learning_rate": 2.757117209540908e-08, + "loss": 0.1965, + "step": 77987 + }, + { + "epoch": 3.0, + "learning_rate": 2.6929982046678634e-08, + "loss": 0.342, + "step": 77988 + }, + { + "epoch": 3.0, + "learning_rate": 2.6288791997948193e-08, + "loss": 0.2494, + "step": 77989 + }, + { + "epoch": 3.0, + "learning_rate": 2.5647601949217746e-08, + "loss": 0.1529, + "step": 77990 + }, + { + "epoch": 3.0, + "learning_rate": 2.5006411900487305e-08, + "loss": 0.1472, + "step": 77991 + }, + { + "epoch": 3.0, + "learning_rate": 2.436522185175686e-08, + "loss": 0.1554, + "step": 77992 + }, + { + "epoch": 3.0, + "learning_rate": 2.372403180302642e-08, + "loss": 0.2465, + "step": 77993 + }, + { + "epoch": 3.0, + "learning_rate": 2.3082841754295976e-08, + "loss": 0.1261, + "step": 77994 + }, + { + "epoch": 3.0, + "learning_rate": 2.2441651705565532e-08, + "loss": 0.1722, + "step": 77995 + }, + { + "epoch": 3.0, + "learning_rate": 2.180046165683509e-08, + "loss": 0.1938, + "step": 77996 + }, + { + "epoch": 3.0, + "learning_rate": 2.1159271608104644e-08, + "loss": 0.183, + "step": 77997 + }, + { + "epoch": 3.0, + "learning_rate": 2.05180815593742e-08, + "loss": 0.1946, + "step": 77998 + }, + { + "epoch": 3.0, + "learning_rate": 1.9876891510643757e-08, + "loss": 0.2615, + "step": 77999 + }, + { + "epoch": 3.0, + "learning_rate": 1.9235701461913313e-08, + "loss": 0.2755, + "step": 78000 + }, + { + "epoch": 3.0, + "learning_rate": 1.859451141318287e-08, + "loss": 0.159, + "step": 78001 + }, + { + "epoch": 3.0, + "learning_rate": 1.7953321364452425e-08, + "loss": 0.134, + "step": 78002 + }, + { + "epoch": 3.0, + "learning_rate": 1.731213131572198e-08, + "loss": 0.1103, + "step": 78003 + }, + { + "epoch": 3.0, + "learning_rate": 1.667094126699154e-08, + "loss": 0.1351, + "step": 78004 + }, + { + "epoch": 3.0, + "learning_rate": 1.6029751218261093e-08, + "loss": 0.2017, + "step": 78005 + }, + { + "epoch": 3.0, + "learning_rate": 1.538856116953065e-08, + "loss": 0.2304, + "step": 78006 + }, + { + "epoch": 3.0, + "learning_rate": 1.4747371120800205e-08, + "loss": 0.1768, + "step": 78007 + }, + { + "epoch": 3.0, + "learning_rate": 1.410618107206976e-08, + "loss": 0.087, + "step": 78008 + }, + { + "epoch": 3.0, + "learning_rate": 1.3464991023339317e-08, + "loss": 0.156, + "step": 78009 + }, + { + "epoch": 3.0, + "learning_rate": 1.2823800974608873e-08, + "loss": 0.1277, + "step": 78010 + }, + { + "epoch": 3.0, + "learning_rate": 1.218261092587843e-08, + "loss": 0.1105, + "step": 78011 + }, + { + "epoch": 3.0, + "learning_rate": 1.1541420877147988e-08, + "loss": 0.1341, + "step": 78012 + }, + { + "epoch": 3.0, + "learning_rate": 1.0900230828417544e-08, + "loss": 0.0788, + "step": 78013 + }, + { + "epoch": 3.0, + "learning_rate": 1.02590407796871e-08, + "loss": 0.3105, + "step": 78014 + }, + { + "epoch": 3.0, + "learning_rate": 9.617850730956656e-09, + "loss": 0.3079, + "step": 78015 + }, + { + "epoch": 3.0, + "learning_rate": 8.976660682226212e-09, + "loss": 0.2484, + "step": 78016 + }, + { + "epoch": 3.0, + "learning_rate": 8.33547063349577e-09, + "loss": 0.1547, + "step": 78017 + }, + { + "epoch": 3.0, + "learning_rate": 7.694280584765324e-09, + "loss": 0.2098, + "step": 78018 + }, + { + "epoch": 3.0, + "learning_rate": 7.05309053603488e-09, + "loss": 0.2852, + "step": 78019 + }, + { + "epoch": 3.0, + "learning_rate": 6.4119004873044364e-09, + "loss": 0.1969, + "step": 78020 + }, + { + "epoch": 3.0, + "learning_rate": 5.770710438573994e-09, + "loss": 0.3379, + "step": 78021 + }, + { + "epoch": 3.0, + "learning_rate": 5.12952038984355e-09, + "loss": 0.2258, + "step": 78022 + }, + { + "epoch": 3.0, + "learning_rate": 4.488330341113106e-09, + "loss": 0.2255, + "step": 78023 + }, + { + "epoch": 3.0, + "learning_rate": 3.847140292382662e-09, + "loss": 0.2377, + "step": 78024 + }, + { + "epoch": 3.0, + "learning_rate": 3.2059502436522182e-09, + "loss": 0.1858, + "step": 78025 + }, + { + "epoch": 3.0, + "learning_rate": 2.564760194921775e-09, + "loss": 0.2155, + "step": 78026 + }, + { + "epoch": 3.0, + "learning_rate": 1.923570146191331e-09, + "loss": 0.1528, + "step": 78027 + }, + { + "epoch": 3.0, + "learning_rate": 1.2823800974608875e-09, + "loss": 0.1984, + "step": 78028 + }, + { + "epoch": 3.0, + "learning_rate": 6.411900487304438e-10, + "loss": 0.2773, + "step": 78029 + }, + { + "epoch": 3.0, + "learning_rate": 0.0, + "loss": 0.2441, + "step": 78030 + }, + { + "epoch": 3.0, + "eval_accuracy": 0.8949656261188632, + "eval_loss": 0.2753095030784607, + "eval_runtime": 252.0722, + "eval_samples_per_second": 1650.952, + "eval_steps_per_second": 25.798, + "step": 78030 + } + ], + "max_steps": 78030, + "num_train_epochs": 3, + "total_flos": 4.2690155243149176e+16, + "trial_name": null, + "trial_params": null +}